Water content and hydrogen behaviour during metasomatism in the uppermost mantle beneath Ray Pic volcano (Massif Central, France)

Abstract

To understand the deep cycle of water, upper mantle water content and distribution between nominally anhydrous minerals (NAMs) and hydrous minerals (e.g., amphibole) must be constrained. We need then to understand H behaviour during mantle melting and metasomatism. Major, minor and trace element compositions including water contents were obtained on ten xenoliths of spinel-bearing peridotites from the Ray Pic locality, in the Southern part of the Massif Central (France). The sample suite investigated here is composed of rather fertile lherzolites (89.4 ≤ Fo ≤ 90.8%; 11.3 ≤ cr# in spinel ≤ 21.1%; 0.942 ≤ [Yb]cpx ≤ 1.90 ppm; cpx: clinopyroxene), which can be best explained by batch melting, with degree of melting between 3 and 10%. These xenoliths contain up to 8% modal amphibole. Three groups are defined as a function of the amphibole modal abundance (above or below 1%) and equilibrium temperature (above or below 1000 °C). Results show no clear positive correlation between modal metasomatism (amphibole) and incompatible element enrichment in cpx. Trace element compositions in cpx show strong enrichments of the most incompatible elements (e.g., (La/Sm)PM as high as 15.7; PM: normalised to primitive mantle values), but strong negative anomalies of the high field strength elements (e.g., (Th/Nb)PM as high as ~ 680). Such trace element fractionations are usually ascribed to the so-called carbonatitic metasomatism involving the percolation of small volume melts which are enriched in volatiles. The hydrogen concentrations in cpx range from 203 to 330 ppm wt. H2O, in orthopyroxene from 66 to 160 ppm wt. H2O and in olivine from 2 to 6 ppm wt. H2O. These values are within the common concentration range of other spinel-bearing peridotites. Amphiboles contain 1.9 ± 0.5 wt.% of H2O.

Full text

Water content and hydrogen behaviour during metasomatism in the
uppermost mantle beneath Ray Pic volcano (Massif Central, France)
Carole M.M. Denis ⁎, Olivier Alard, Sylvie Demouchy
Geosciences Montpellier, University of Montpellier, place E. Bataillon, Montpellier, France
abstractarticle info
Article history:
Received 25 May 2015
Accepted 19 August 2015
Available online 2 September 2015
Keywords:
Hydrogen
Peridotite
Mantle
Melting
Metasomatism
Amphibole
To understand the deep cycle of water, upper mantle water content and distribution between nominally anhy-
drous minerals (NAMs) and hydrous minerals (e.g., amphibole) must be constrained. We need then to under-
stand H behaviour during mantle melting and metasomatism. Major, minor and trace element compositions
including water contents were obtained on ten xenoliths of spinel-bearing peridotites from the Ray Pic locality,
in the Southern part of the Massif Central (France). The sample suite investigated here is composed of rather fer-
tile lherzolites (89.4 ≤Fo ≤90.8%; 11.3 ≤cr# in spinel ≤21.1%; 0.942 ≤[Yb]
cpx
≤1.90 ppm; cpx: clinopyroxene),
which can be best explained by batch melting, with degree of melting between 3 and 10%. These xenoliths con-
tain up to 8% modalamphibole. Three groups are definedas a function of the amphibole modal abundance (above
or below 1%) and equilibrium temperature (above or below 1000 °C). Results show no clear positive correlation
between modal metasomatism (amphibole) and incompatible element enrichmentin cpx. Trace element com-
positions in cpx show strong enrichments of the most incompatible elements (e.g., (La/Sm)
PM
as high as 15.7;
PM: normalised to primitive mantle values), but strong negative anomalies of the high field strength elements
(e.g., (Th/Nb)
PM
as high as ~680). Such trace element fractionations are usually ascribed to the so-called
carbonatitic metasomatism involving the percolation of small volume melts which are enriched in volatiles.
The hydrogen concentrations in cpx range from 203 to 330 ppm wt. H
2
O, in orthopyroxene from 66 to
160 ppm wt. H
2
O and in olivine from 2 to 6 ppm wt. H
2
O. These values are within the common concentration
range of other spinel-bearing peridotites. Amphiboles contain 1.9 ± 0.5 wt.% ofH
2
O.
The effect of metasomatism on water abundances in NAMs is not straightforward. Hydrous metasomatism
(i.e., leading to the crystallisation of OH-bearing amphibole) has no effect on the water content of the co-
existing NAMs. This suggests thus that the occurrence of hydrous minerals, such as amphibole, does not system-
atically imply that the coexisting NAMs are water-rich (saturated). Further, the percolation of volatile-rich small
volume melts, which is fingerprinted by the strong enrichment of the incompatible elements, has also no clear
effect on the water content of the NAMs. These data are thus difficult to reconcile with the admitted highly
incompatible behaviour of H in upper mantle minerals.
© 2015 Elsevier B.V. All rights reserved.
1. Introduction
The percolation and reaction of melts and/or fluids are ubiquitous
processes within the mantle, which play a key role in theredistribution
of elements throughout the Earth's upper mantle (e.g., Bedini et al.,
1997; Bodinier et al., 1990; Menzies and Hawkesworth, 1987; Rudnick
et al., 1993). Classically, two types of metasomatism are recognised:
(1) modal metasomatism, for which new volatile-rich minerals precip-
itate, such as amphibole (am), phlogopite, apatite; and (2) cryptic meta-
somatism, for which no new phases crystallise, but primary phases, for
instance clinopyroxene (cpx), which displays trace element enrich-
ments not compatible with a simple melt depletion (e.g., Grégoire
et al., 2000; O'Reilly et al., 1991). The term of metasomatism encom-
passes a diversity of processes involving a variety of not well-defined
melt–fluid/rock interactions. These interactions occurred with different
fluid/rockregimes in various geological environments, with percolating
and metasomatising melts of variable compositions (O'Reilly and
Griffin, 2013). Usually the percolating melt/fluid is often considered
as a volatile-rich agent. Volatile such as H
2
OandCO
2
lower the
melt/fluid density, viscosity and the crystallisation temperature
(e.g., Dasgupta and Hirschmann, 2006; Wyllie and Ryabchikov,
2000) and therefore ease the percolation of such melt/fluids within
the cold mantle lithosphere. Deciphering the imprints of the various
metasomatic processes on the mantle lithosphere is a key to better
understand their effect on the physical properties of the mantle
and thus on the geodynamic of the lithosphere, but also to constrain
the volatile transfer at larger scale, from the deep Earth to the exo-
spheres (i.e., the deep water cycle).
Lithos 236–237 (2015) 256–274
⁎Corresponding author. Tel.: +33 4 67 14 39 39; fax: +33 4 67 14 36 42.
E-mail address: carole.denis@gm.univ-montp2.fr (C.M.M. Denis).
http://dx.doi.org/10.1016/j.lithos.2015.08.013
0024-4937/© 2015 Elsevier B.V. All rights reserved.
Contents lists available at ScienceDirect
Lithos
journal homepage: www.elsevier.com/locate/lithos

One of the key tools to decipher the type of metasomatism is the
abundances and fractionation of incompatible trace elements such as
Rare Earth Elements (REE). In addition, hydrogen (H) and carbon,
two other trace elements, play a crucial role since they are both a
potential indicator of metasomatic processes, but also a major con-
stituent of metasomatic melts/fluids. Indeed, molecular water and
water-derived species (H
+
,OH
−
,H
2
)canbefoundasamajorcon-
stituent (i.e., H
2
O≥1 wt.%) in metasomatic hydrous minerals such
as in amphibole and/or phlogopite. The precipitation of these hy-
drous minerals is canonically ascribed to the so-called hydrous
metasomatism (e.g., Agrinier et al., 1993; Bodinier et al., 2004;
McInnes et al., 2001). Nevertheless, H can also be incorporated in
the lattice of nominally anhydrous minerals (NAMs) using atomic
point defects (zero-dimension defect in the crystal lattice, e.g., Bell
and Rossman, 1992; Ingrin and Skogby, 2000; Miller et al., 1987),
such as olivine (ol), cpx and orthopyroxene (opx). Hydrogen is con-
sidered to behave like an incompatible trace element, with partition
coefficient similar to La and Ce (Dixon et al., 2002)orevenlower
(e.g., Novella et al., 2014; O'Leary et al., 2010). Hydrogen could
thus provide valuable and complementary information on metaso-
matic processes (e.g., nature of the melt/fluid, melt–fluid/rock
ratio). Hydrogen incorporation into NAMs was mainly investigated
through experimental petrology and mineralogy (e.g., Bai and
Kohlstedt, 1993; Férot and Bolfan-Casanova, 2012; Kohlstedt et al.,
1996; Mierdel et al., 2007) while data obtained on mantle xenolith
suites are only recent (e.g., Baptiste et al., 2015; Bell and Rossman,
1992; Bell et al., 2004; Bizimis and Peslier, 2015; Bonadiman et al.,
2009; Demouchy et al., 2006; Demouchy et al., 2015; Denis et al.,
2013; Doucet et al., 2014; Grant et al., 2007a, 2007b; Hao et al.,
2014a, 2014b; Li et al., 2008; Li et al., 2014; Peslier and Luhr, 2006;
Peslier et al., 2002, 2015; Xia et al., 2010, 2013; Yang et al., 2008;
Yu et al., 2011). However, H behaviour remains unclear during man-
tle processes and especially during metasomatic processes. Further,
except for Yang et al. (2008) and Bonadiman et al. (2009),thedistri-
bution of H among NAMs in a so-called hydrous mantle assemblages
such as amphibole-bearing peridotites, was not investigated.
We investigate here a suite of peridotite xenoliths from the Ray Pic
Volcano located in the French Massif Central (FMC) showing various
modal (amphibole) and cryptic metasomatic imprints. RayPic xenoliths
are a well-known xenolith locality from the FMC, which were previous-
ly studied by Downes and Dupuy (1987),Lenoir et al. (2000) and
Zangana et al. (1997, 1999). The main conclusions of these studies are
that the xenoliths from Ray Pic represent a heterogeneous mantle in
term of modal mineralogy, mineral chemistry, REE patterns and radio-
genic isotopes. Theseheterogeneities resulted from the superimposition
of metasomatic processes related to the impingement of the FMC
plume/diapir on a variably depleted protolith (Zangana et al., 1997,
1999).
To establish the relationships between mantle processes (melting
and metasomatism) and water content of the mantle minerals, we
first identify mantle processes which have affected the Ray Pic perido-
tites, using major and trace element compositions, and confront them
to water contents in NAMs measured by Fourier transform infrared
(FTIR) spectroscopy.
2. Geological setting and samples
Alkali-basaltic volcanism of theFMC began in the early Cenozoic and
ended ~0.01 Ma ago with a peak of volcanic activity at 4–6Ma(Maury
and Varet, 1980). This volcanic activity is related to an ascending as-
thenospheric plume/diapir beneath the FMC (e.g. Granet et al., 1995).
Seismic tomography data pinpointed the mantle anomaly beneath the
Eastern Massif Central between the 45°30′and 44°30′parallels to the
east of the Sillon Houiller fault (Granet et al., 1995). Geochemical studies
based on mantle xenoliths have also evidenced two contrasting mantle
lithosphere domains on either side of the 45°30′Nparallel(Lenoir et al.,
2000;Fig. 1). This contrast is based on lithology, trace element compo-
sitions, Pb, Lu–Hf, Sr and Nd isotope analyses (Downes et al., 2003;
Wittig et al., 2007) and metasomatism styles (Lenoir et al., 2000;
Touron et al., 2008). Peridotites from the southern domain are overall
characterised by a common occurrence of secondary coarse granular
microstructure (not protogranular), and less fractionated trace element
patterns compared to northern domain peridotites. These characteris-
tics were ascribed to the impingement of the FMC plume/diapir on a rel-
atively fertile lithosphere (Lenoir et al., 2000). It is worth noting that
hydrous metasomatic minerals such as phlogopite or amphibole are
more commonly found in peridotite xenoliths from the southern FMC
(i.e., in the Devès and the Velay, Fig. 1) than in the northern part. They
were attributed to the reaction between the peridotite and alkali
melts directly related to the asthenosphere plume/diapir (Lenoir et al.,
2000).
The ten samples of this study are from the Ray Pic volcano, a Quater-
nary alkali basalt volcano, in the Velay volcanic district of the southern
FMC (Fig. 1). Ray Pic volcano brought up numerous mantle xenoliths
reflecting a large range of mineralogical and geochemical composition
(Zangana et al., 1997, 1999). Variations in mineral microstructures,
whole rock (WR) major, trace and REE compositions, as well as radio-
genic isotope compositionsof Ray Pic xenolithswere previously report-
ed by Zangana et al. (1997, 1999). These variations were interpreted as
the result of two major processes, which have affected the lithospheric
mantle beneath Ray Pic: a depletion likely due to partial melting and
subsequent dominant cryptic metasomatism (modal metasomatism
was uncommon, Zangana et al., 1999).
3. Sample preparation and analytical method
Concentrations in major elements for the WR were obtained by
wide-angle X-ray fluorescence (WDXRF). Mineral compositions were
determined using an electron probe micro-analyser (EPMA) for major
elements, using a laser ablation induced coupled plasma mass spec-
trometer (LA-ICP-MS) for trace elements, and using a FTIR spectrometer
for hydrogen.
3.1. Sample preparation
All samples were cut far from weathered surfaces or from the
enclosing basalt. Part of each sample was crushed with an anvil and a
hammer, then reduced to powder using an agate ring mill for WR anal-
yses. In situ major and trace element compositions of minerals were de-
termined on 150 μm thick polished sections carbon-coated for EPMA
analyses. FTIR analyses were performed on separated minerals, doubly
polished by hand (~ 100–1000 μm of thickness) with a polishing jig
and diamond-lapping films (grid sizes ranging from 0.2 to 30 μm).
3.2. Major and trace elements
Major element contents in WR were measured by WDXRF by a se-
quential spectrometer Bruker S4 Pioneer at the analytical services of
the IACT (CSIC, Granada, Spain).
Major element contents in each mineral phase were determined at
Microsonde Sud facility, at University of Montpellier (France) using a
Cameca SX100 electron microprobe operating at 20 kV accelerating
voltage. The thin sections were analysed with a counting time on peak
and background of 20 s for the analysis of Fe, Mn and Ni, and 30 s for
the other elements. The spot size was ∼1μm.
Equilibrium temperature of each sample is calculated using the two
thermometers of Brey and Köhler (1990),one based on the Ca concen-
tration in opx (T°
Ca-in-opx
), and the other based on Fe–Mg exchange be-
tween opx and cpx (T
BKN
).
Using WR and mineral compositions, the mineralogical mode of
each sample is calculated usingmass balance and a least squares meth-
od (Herrmann and Berry, 2002).
257C.M.M. Denis et al. / Lithos 236–237 (2015) 256–274

In situ trace element analyses were performed on cpx and amphi-
bole by LA-ICP-MS at Géosciences Montpellier (Montpellier, France),
using a ThermoFinnigan ELEMENT XR (eXtended Range) high resolu-
tion (HR) ICP-MS system. The ICP-MS is coupled with a Geolas
(Microlas) automated platform housing a ArF 193 nm Compex 102
laser from LambdaPhysik. Measurements were conducted in a Heatmo-
sphere, which enhances the sensitivity and reduces inter-elementfrac-
tionations. The helium gas stream and particles from the sample are
mixed with Ar before entering the plasma. Signals are measured in
time resolved acquisition mode, devoting 2 min for the blank, and
3 min for mineral analysis. The laser is fired using an energy density of
15 J/cm
2
at a frequency of 10 Hz and using a spot size of 77 μm. Oxide
levels, measured using the UO/U ratio, are below 0.7%. The standard
BIR-1G was included during the analytical runs and our measurements
are in good agreement with working values for this international stan-
dard (Jochum et al., 2005). The concentrations are calibrated against
the standard NIST 612 glass using the values given by Pearce et al.
(1997). Data are subsequently reduced using the GLITTER software
(Griffin et al., 2008) by carefully inspecting the time-resolved analysis
to check for homogeneity in the analysed volume.
3.3. Fourier transform infrared spectroscopy
Hydroxyl bonds in the samples are detected in transmission mode
by FTIR spectroscopy at the Laboratoire Charles Coulomb (LCC) at
University of Montpellier (France). Unpolarised infrared spectra were
acquired using a Bruker IFS66v, equipped with an MCT detector (Mer-
cury Cadmium Telluride) cooled with liquid nitrogen, and coupled to a
microscope BrukerHYPERION. This allows accurate optical visualisation
of the sample prior and during acquisitions. A Globar lightsource and a
Ge-KBr beam splitter are used to generate unpolarised mid-infrared ra-
diation (between 4000–400 cm
−1
). Measurements on olivine, opx, and
cpx were performed with a square aperture yielding a beam spot with
sizes ranging from 50 to 150 μm. For each measurement 200 scans are
accumulated with a resolution of 4 cm
−1
. The series of resulting inter-
ferograms is first ratioed against the reference spectrum (background),
then treated manually with a baseline correction using the OPUS
software, and finally normalised to 1 cm of thickness. From each spec-
trum, the quantification of the hydrogen concentration is deduced
from the hydroxyl stretching absorption in the sample, based on the
Beer–Lambertlaw.Atfirst, we choose to use the calibration of
Paterson (1982):
COH ¼χi
150ξ∫kυðÞ
3780−υdυð1Þ
This empirical and frequency-dependent equation gives the water
concentration (C
OH
) as a function of factor χ
i
, which is a function of
the density of the mineral i(χ
ol
= 2695 ppm wt H
2
O; χ
opx
2812 ppm
wt H
2
O; χ
cpx
= 2752 ppm wt H
2
O). ξis the orientation factor and equals
1/3 for non-polarised IR analyses (on non-oriented samples, Paterson,
1982); k(υ), the absorption coefficient is a function of the wavenumber
υ. The H concentration of the samples is calculated by integrating each
spectrum between 3610 and 3150 cm
−1
for olivine, 3670–2800 cm
−1
for opx, and 3770–3000 cm
−1
for cpx. This calibration allows a detec-
tion limit of about 0.5 ppm wt. H
2
O for an olivine sample of 1 mm of
thickness. Here H concentration is expressed as weight ppm of
dihydrogen oxide = H
2
O, also called water content by convention.
The estimated error from the empirical calibration in the resulting H
concentration is 30% (Paterson, 1982; Rauch, 2000). Average H concen-
trationsare obtained from averagingat least 10 spectra from each grain
and the maximal linear absorption of unnormalised spectra did not ex-
ceed 0.3 in agreement with the recommendations of Withers (2013) for
unpolarised FTIR measurements.
Mineral-dependent FTIR calibration should always be favour to re-
duce uncertainties on the resulting water content. However, the
mineral-dependent calibration for olivine from Bell et al. (2003) is not
used here since it overestimates hydrogen concentrations (Withers
et al., 2012). Ideally this calibration requires the use of crystallographi-
cally oriented olivine crystals and polarised IR radiation in order to
sum the integrated absorbances from the three polarised spectra corre-
sponding to the three crystallographic axis and for each analysed grain
(i.e., Absorbance
total
= absorbance
[100]
+ absorbance
[010]
+ absor-
bance
[001]
,e.g.,Bell et al., 2003; Withers et al., 2012). We chose to use
Fig. 1. Simplified geological mapof the French Massif Central(FMC) modified afterLenoir et al. (2000), showing the location of thestudied xenolithsfrom Ray Pic volcano(triangle), major
towns (dot ), and the main v olcanic distric ts (dark grey are as). The dashed lin e (45°30′parallel)represents the limitbetween the two lithospheric mantle domains, Northernand Southern
Massif Central defined by Lenoir et al. (2000).
258 C.M.M. Denis et al. / Lithos 236–237 (2015) 256–274

the calibration of Paterson (1982) since it allowed us to determine hy-
drogen concentrations for a larger number of analyses, to compare the
concentrations from the different mineral phases (i.e., one common
methods for all phases) and to avoid to cut each grains in oriented
cube, which will destroy the borders of the mineral grains (as in
Schmädicke et al., 2013). Conversion to the most recent mineral-
dependent calibration from Withers et al. (2012) for olivine can be cal-
culated using a factor of 1.8 (Withers et al., 2012, see in their Fig. S1) and
the concentration reported in results Table. (i.e., concentration from
Paterson calibration —unpolarised × 1.8 ≈average integrated
unpolarised and normalised absorbance × 3 × 0.119).
Regarding the use of the calibration of Bell et al. (1995) for H in py-
roxene, the H concentration can be calculated usingthe integrated area
reported in this study. The resulting conversion factor between the cal-
ibration of Paterson and the one from Bell et al. (1995) is equal to 1.03
for opx and equals to 1.74 for cpx. The discrepancy is likely to be due
to the use in their calibration of a single kimberlitic centimetric
megacryst of clinopyroxene (augite: Ca- and Al-poor and Mg and Fe-
rich) while the orthopyroxene is from a mantle peridotite. Thus only
the orthopyroxene standard has the adequate composition compared
to our samples and the application of this calibration for clinopyroxenes
to mantle diopside is questionable. Recently, Mosenfelder and Rossman
(2013b) have also document that a frequency-dependent calibration
yields better results than the calibration of Bell et al. (1995) for
clinopyroxene.
Another frequency-dependent IR calibration could be used as pro-
posed by Libowitzky and Rossman (1997) as well as by Mosenfelder
and Rossman (2013a, b). However, one must notice that for the
wavenumbers under consideration in NAMs (3600–3000 cm
−1
), the
calibration of Paterson (1982) and the one from Libowitzky and
Rossman (1997) are equivalent (see their Fig. 3 in Libowitzky and
Rossman, 1997) and yield results within the error bars.
As indicated in Férot and Bolfan-Casanova (2012, see their supple-
mentary material Figure S1), a factor of 3 can be used with confidence
to convert the hydrogen concentrations in olivine and pyroxenes re-
ported by unpolarised IR light to values equivalent to those obtained
with the polarised IR method of Bell et al. (1995, 2003;seealso
Kovács et al., 2008).
Finally, conversion factors to concentration in atomic H per 10
6
atoms of Si (at. H/10
6
Si = at. ppm H/Si) historically used
(e.g., Kohlstedt et al., 1996) are also given in the results' table.
The water contents of amphiboles from sample 13RP03 were mea-
sured by Karl–Fischer titration (KFT) using air as transporting gas and
muscovite as standard, see Behrens et al, (1996) for details on the
method.
4. Results
4.1. Petrography and mineralogy
Peridotite xenoliths from Ray Pic are all equilibrated in the spinel
(spl) facies and range from lherzolite to harzburgite as previously
reported by Downes and Dupuy (1987) and Zangana et al. (1997,
1999). The peridotite xenoliths investigated here are ten lherzolites
with variable modal composition of olivine (ol: 51–67%), opx (19–
33%), cpx (6–13%) and spl (1–3%) and one harzburgite (13RP14; ol:
77%; opx: 18%; cpx: 3%). Six samples contain significant amount of
amphibole (am: 0.2–8%) as shown in Fig. 2. The selected peridotites
have continuous textural variation evolving from porphyroclastic to
equigranular and coarse granular microstructure as reported by
Zangana et al. (1997, 1999). True protogranular textures according
to the definition of Mercier and Nicolas (1975) are not found, and
coarse-grained xenoliths are clearly secondary microstructure due to
recrystallisation processes at subsolidus conditions. Indeed, most of
our samples show transitional microstructure between porphyroclastic
and equigranular or coarse-granular texture. Sub-grain boundaries as
well as undulose extinctions are common in olivine. However, some of
the crystals are free of intra-crystalline deformation features,
progressing locally to triple junctions between minerals. The grain
sizes rangefrom 0.2 to 3 mm. Spinel defines the foliation in several sam-
ples, and occurs as holly-leafshaped grains with curvilinear boundaries.
Amphibole-bearing samples have relatively large pale brown amphi-
boles (Ø ≈1 mm) disseminated into the aggregate (i.e., in sample
13RP02, Fig. 2b). In some samples (i.e., sample 13RP03, Fig. 2)amphi-
boles are closely associated with spl and cpx forming brown patches.
The texture and mineralogical compositions of the samples are
summarised in Table 1.
4.2. Major element compositions of minerals and geothermometry
Major element compositions of WR and minerals are given in
Table 2.Whole-rockAl
2
O
3
concentrations vary between 1.54–
3.84 wt.%, which indicates a fertile composition in agreement with the
petrographic observations. The composition of our samples is compara-
ble to those previously reported in the southern domain of the FMC
(Alard et al., 1996; Downes and Dupuy, 1987; Downes et al., 2003;
Lenoir et al., 2000; Lorand and Alard, 2001; Touron et al., 2008; Xu
et al., 1998). They are fully within the compositional range reported
for Ray Pic xenoliths (Downes and Dupuy, 1987; Zangana et al., 1997,
1999;Fig. 3). The most fertile lherzolite is 13RP02, which also has the
Fig. 2. Thinsection micrographs in plane-polarizedlight for Ray Picspinel-peridotite xeno-
liths: (a)sample 13RP02 showing olivine(ol), orthopyroxene (opx), clinopyroxene (cpx),
spinel (spl) and disseminated amphibole (am), as well as the equigranular to coarse mi-
crostructure (b) sample 13RP03 showing a typical coarse granular microstructure. Note
that amphibole is associated with spl and cpx. Colour version available online.
Table 1
Sample list and summary of Ray Pic peridotite xenolith characteristics.
Ray Pic Texture
a
Modal mineralogy (%)
b
Equilibration
temp. (°C)
c
sample olivine opx cpx sp am T°Ca-opx T°BKN
13RP05 Eq 67 21 9 2 –841 710
13RP17 Crs 67 19 11 3 –903 859
13RP01 Eq 65 23 9 2 b1 840 822
13RP02 Crs to Prc 51 33 13 3 b1 925 928
13RP12 Eq to Crs 63 23 12 1 1 952 964
13RP04 Crs to Prc 57 28 11 2 2 880 903
13RP14 Crs 77 18 3 1 2 912 851
13RP03 Eq to Crs 65 19 6 1 8 933 877
13RP11 Crs to Prc 67 22 10 2 –1010 1044
14RP06 Crs 62 26 9 2 –1070 1084
a
Eq: equigranular; Crs: coarse; Prc: porphyroclastic.
b
opx: orthopyroxene; cpx: clinopyroxene; sp: spinel; am: amphibole.
c
T°(Ca-in-opx) and T° BKN from Brey and Köhler (1990),error±30°C.
259C.M.M. Denis et al. / Lithos 236–237 (2015) 256–274

Table 2
Whole-rock and mineral major element average compositions (wt. %) determined by XRF and EPMA respectively.
Whole-rock
Group G1 G2 G3
b1% am N1% am
Sample 13RP05 13RP17 13RP01 13RP02 13RP12 13RP04 13RP14 13RP03 13RP11 14RP06
T° Ca-opx 841 903 840 925 952 880 912 933 1010 1070
SiO
2
44.91 44.16 45.3 46.2 45.37 45.8 44.27 44.79 44.85 n.a.
TiO
2
0.11 0.18 0.15 0.21 0.19 0.18 0.11 0.18 0.13 n.a.
Al
2
O
3
2.31 2.76 2.48 3.84 2.63 2.93 1.54 2.7 2.55 n.a.
Fe
2
O
3
8.64 8.86 8.82 8.43 8.68 8.38 8.48 8.64 8.57 n.a.
MnO 0.13 0.14 0.14 0.14 0.13 0.13 0.13 0.14 0.13 n.a.
MgO 41.56 40.59 40.84 37.82 39.86 39.65 44.28 40.79 41.3 n.a.
CaO 2.28 2.64 2.13 3.09 2.9 2.73 1.14 2.51 2.31 n.a.
Na
2
O 0.04 0.09 0.13 0.24 0.21 0.2 0.03 0.19 0.12 n.a.
K
2
Ob0.02 b0.02 b0.02 b0.02 b0.02 b0.02 b0.02 b0.02 b0.02 n.a.
P
2
O
5
b0.02 0.02 b0.02 b0.02 0.02 b0.02 b0.02 0.02 0.03 n.a.
LOI 0 0.53 0 0 00000n.a.
Total 99.98 99.97 99.99 99.97 99.99 100 99.98 99.96 99.99 –
mg# (WR) 90.4 89.9 90.0 89.7 89.9 90.2 91.1 90.2 90.4 –
Olivine
Sample 13RP05 13RP17 13RP01 13RP02 13RP12 13RP04 13RP14 13RP03 13RP11 14RP06
n 451726817694
SiO
2
41.49 41.30 41.45 40.98 41.18 41.07 41.34 41.36 41.34 40.81
TiO
2
0.01 0.00 0.01 0.00 0.01 0.01 0.01 0.01 0.01 0.01
Al
2
O
3
0.00 0.01 0.01 0.02 0.02 0.01 0.01 0.02 0.02 0.04
Cr
2
O
3
0.01 0.01 0.03 0.00 0.01 0.01 0.06 0.01 0.01 0.03
FeO 9.45 9.66 9.75 9.97 10.05 9.75 9.22 9.46 9.37 9.24
MnO 0.13 0.13 0.13 0.16 0.13 0.13 0.13 0.13 0.13 0.13
MgO 48.15 48.11 47.94 47.93 47.93 48.54 48.52 48.60 48.40 48.43
CaO 0.02 0.04 0.05 0.05 0.08 0.04 0.04 0.06 0.07 0.10
Na
2
O 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.02 0.01
K
2
O 0.00 0.00 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00
NiO 0.42 0.39 0.40 0.38 0.40 0.42 0.41 0.37 0.39 0.38
Total 99.70 99.66 99.78 99.50 99.84 99.99 99.74 100.03 99.76 99.18
Fo% 90.0 89.8 89.6 89.4 89.4 89.8 90.3 90.0 90.8 90.2
Orthopyroxene = enstatite
Sample 13RP05 13RP17 13RP01 13RP02 13RP12 13RP04 13RP14 13RP03 13RP11 14RP06
n 255251017665
SiO
2
56.51 56.03 56.40 55.62 55.71 55.72 56.47 56.10 55.47 54.56
TiO
2
0.03 0.06 0.06 0.10 0.10 0.07 0.05 0.05 0.06 0.08
Al
2
O
3
2.81 3.63 3.33 3.94 3.93 3.58 3.07 3.51 4.33 4.43
Cr
2
O
3
0.24 0.26 0.30 0.31 0.27 0.29 0.36 0.34 0.39 0.56
FeO 6.28 6.47 6.48 6.51 6.41 6.35 5.89 6.31 6.16 6.03
MnO 0.14 0.14 0.13 0.12 0.15 0.15 0.13 0.15 0.14 0.12
MgO 33.19 32.69 32.75 32.47 32.58 33.18 33.27 33.02 32.32 32.00
CaO 0.37 0.51 0.37 0.56 0.63 0.48 0.49 0.60 0.82 1.05
Na
2
O 0.03 0.04 0.05 0.07 0.09 0.07 0.04 0.06 0.10 0.12
K
2
O 0.00 0.01 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.00
NiO 0.08 0.10 0.09 0.12 0.12 0.09 0.10 0.10 0.10 0.12
Total 99.68 99.95 99.96 99.82 99.98 99.98 99.88 100.25 99.89 99.08
mg# (opx) 90.4 90 90 89.9 90.1 90.3 91 90.3 90.3 90.3
[En] 86.6 85.5 85.5 85.2 85.5 86.6 86.9 86.2 85.2 85.5
[Fs] 11.3 11.5 12.1 11.1 10.7 9.5 10.7 10.4 10.5 9.7
[Wo] 2.1 3.0 2.4 3.6 3.8 3.9 2.5 3.4 4.2 4.7
Clinopyroxene = diopside
Sample 13RP05 13RP17 13RP01 13RP02 13RP12 13RP04 13RP14 13RP03 13RP11 14RP06
n 641326515474
SiO
2
53.31 52.64 52.81 52.03 52.06 52.20 53.07 53.26 52.16 51.35
TiO
2
0.12 0.28 0.49 0.45 0.54 0.42 0.18 0.20 0.28 0.34
Al
2
O
3
3.49 4.74 6.08 6.09 6.20 5.63 4.18 3.90 6.13 6.22
Cr2O3 0.65 0.58 0.94 0.75 0.67 0.70 0.80 0.47 0.86 1.04
FeO 2.42 3.04 2.42 3.02 3.00 2.80 2.50 2.94 3.19 3.20
MnO 0.08 0.08 0.07 0.09 0.08 0.08 0.07 0.09 0.08 0.09
MgO 15.68 15.42 14.34 14.83 14.97 15.27 15.70 16.12 15.28 15.73
CaO 23.19 22.20 20.98 21.06 20.73 21.07 22.18 22.20 20.14 19.31
Na
2
O 0.82 0.93 1.81 1.39 1.45 1.48 1.02 0.95 1.44 1.44
K
2
O 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.01
NiO 0.05 0.03 0.03 0.03 0.06 0.04 0.04 0.03 0.04 0.06
Total 99.80 99.95 99.96 99.73 99.75 99.68 99.75 100.14 99.60 98.78
mg# (cpx) 91.8 89.8 91.1 89.5 89.7 90.4 91.6 90.5 89.3 89.5
260 C.M.M. Denis et al. / Lithos 236–237 (2015) 256–274

lowest MgO content and the highest Al
2
O
3
and CaO content, whereas
the most refractory harzburgite is 13RP14.
The relatively high fertility of the Ray Pic peridotites is
reflected by their mineral compositions as well: forsterite
contents (Fo%) vary between 89.4 and 90.8 with an average of
89.9±0.4andcr#
spl
[100xCr/(Cr + Al)] varies coherently be-
tween 11.3 and 21.1 (Table 2). mg#
spl
varies between 72.6 and
77.0 and is correlated to cr#
spl
. Orthopyroxene are Mg-rich
enstatite (mg# = 89.7–90.8%; En
85.2–86.9
;Fs
9.5–12.1
;Wo
2.1–4.7
) and cpx
are Cr-diopsides (19.3 bCaO ≤23.2 wt.%; 6.8 bcr#
cpx
b11.1%) with
mg#
cpx
varying between 89.3 and 91.8%. No core–rim zoning was
observed.
Amphiboles show a relatively homogeneous composition with-
in and between each sample. There is no correlation between the
microstructural occurrence of amphiboles and their composition.
All amphiboles are Ti (calcic)-pargasite (CaO = 10.9 ± 0.3 wt.%;
TiO
2
= 1.8 ± 0.5 wt.%) with mg#
am
between 86.8 and 88.6%.
These amphiboles show constant Na
2
O content ca. 3.7 ± 0.2 wt.%.
K
2
O content varies between 0.01 and 0.61 wt.%. The sum of the
common oxides (Table 2) is c.a. 97.4 ± 0.2 wt.%. Halogen contents
are low, below the detection limit for F (i.e., b0.1 wt.%) and Cl con-
tents are bracketed between 0.007 and 0.080 wt.%. Karl–Fischer
titration analysis indicates H
2
Ocontentc.a.1.9±0.5wt.%for
amphibole in sample 13RP03. It is remarkably similar to water
content in amphibole for similar mantle xenoliths worldwide (e.g.
Bonadiman et al., 2014; Frezzotti et al., 2010; Litasov et al., 2000;
Moine et al., 2000).
Equilibration temperatures (T) of the peridotites, calculated using
the Ca-in-opx method from Brey and Köhler (1990), range from 840
to 1070 °C, while equilibration temperatures using the Fe–Mg distribu-
tion in two-pyroxene geothermometer (Brey and Köhler, 1990)yields
values between 710 and 1084 °C. Both calculations have a typical
standard deviation of ±30 °C. The occurrence and abundance of
amphiboles have no obvious effect on equilibrium temperatures. Fur-
thermore pyroxene geothermometers based on Fe–Mg, Na exchange
between cpx and opx, or for Ca in opx, are relatively consistent and no
Table 2 (continued)
Whole-rock
Group G1 G2 G3
b1% am N1% am
Sample 13RP05 13RP17 13RP01 13RP02 13RP12 13RP04 13RP14 13RP03 13RP11 14RP06
T° Ca-opx 841 903 840 925 952 880 912 933 1010 1070
Spinel
Sample 13RP05 13RP17 13RP01 13RP02 13RP12 13RP04 13RP14 13RP03 13RP11 14RP06
n 2582348444
SiO
2
0.16 0.13 0.15 0.16 0.14 0.16 0.15 0.15 0.19 0.13
TiO
2
0.04 0.09 0.05 0.12 0.12 0.06 0.03 0.08 0.10 0.21
Al
2
O
3
53.66 55.05 55.26 55.92 57.28 57.00 54.73 50.67 51.75 48.07
Cr2O3 13.78 11.53 13.09 11.64 10.41 10.87 13.26 16.42 16.05 19.12
FeO 12.77 13.74 12.23 12.14 11.62 11.94 12.54 14.30 12.56 12.47
MnO 0.15 0.13 0.13 0.12 0.11 0.11 0.13 0.14 0.13 0.14
MgO 18.49 19.06 18.85 19.44 20.04 19.84 18.92 18.66 19.38 18.72
CaO 0.02 0.01 0.02 0.00 0.10 0.00 0.02 0.01 0.02 0.03
Na
2
O 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
K
2
O 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
NiO 0.35 0.40 0.37 0.41 0.39 0.41 0.38 0.36 0.34 0.35
Total 99.44 100.13 100.16 99.97 100.21 100.40 100.16 100.81 100.54 99.24
mg# (spl) 73.2 74.3 73.1 75.3 77.0 76.4 73.0 73.5 75.6 72.6
cr# (spl) 14.7 12.3 13.7 12.2 10.8 11.3 14.0 17.8 17.2 21.1
Amphibole = Pargasite
Sample 13RP01 13RP02 13RP12 13RP04 13RP14 13RP03
% Am 0.2 0.5 1.1 1.6 1.8 8.1
n 624856
SiO
2
42.68 43.41 43.26 43.18 43.48 43.24
TiO
2
2.42 1.89 2.21 1.71 1.21 1.25
Al
2
O
3
14.54 15.21 15.25 15.31 14.78 15.04
Cr
2
O
3
1.06 0.98 0.94 1.07 1.64 1.28
FeO 3.85 4.40 4.34 4.22 3.97 4.52
MnO 0.05 0.07 0.06 0.06 0.06 0.07
MgO 17.02 16.93 17.02 17.11 17.27 16.99
CaO 11.28 10.68 10.55 10.78 11.03 11.24
Na
2
O 3.81 3.85 3.81 3.86 3.73 3.28
K
2
O 0.10 0.03 0.03 0.01 0.03 0.61
NiO 0.10 0.13 0.13 0.13 0.13 0.10
Total 96.91 97.60 97.61 97.44 97.32 97.62
mg# 88.6 87.1 87.3 87.7 88.4 86.8
F b.d. n.a. b.d. b.d. n.a. n.a.
Cl b.d. n.a. b.d. b.d. n.a. n.a.
H
2
O (KFT) n.a. n.a. n.a. n.a. n.a. 1.9 ± 0.5
n: for n measurements;T° (Ca-in-opx) fromBrey and Köhler (1990),with error ± 30 °C; mg#(Mg-number) andFo% (forsteritecontent) = 100.Mg/(Mg + Fe); cr#:Cr-number = 100.Cr/
(Cr + Al); Fo%, mg#and cr# are in italics, [En],[Fs], [Wo] are enstatite, Ferrosilite, Wollastonite end-membersrespectively; n.a.:not analysed; b.d.;below detection limit;ol: olivine; opx:
orthopyroxene; cpx: clinopyroxene; spl: spinel; am: amphibole.
261C.M.M. Denis et al. / Lithos 236–237 (2015) 256–274

systematic shift or clear inconsistency is observed for amphibole bear-
ing samples.
Using amphibole modal abundances and equilibrium temperatures,
the samples can be divided into three groups: Group 1 (G1): amphibole
free/poor samples (am b1%) with low equilibrium temperature
(T b1000 °C), Group 2 (G2): amphibole-rich peridotites (am N1%)
with low equilibrium temperature, Group 3 (G3): amphibole-free sam-
ples equilibrated at higher temperature (T N1000 °C).
4.3. Trace element compositions of clinopyroxenes and coexisting
amphiboles
The heavy REE (HREE) contents in cpx display significant variations
(Table 3). For instance Yb varies from 0.94 up to 1.90 ppm and thus sug-
gest variable degrees of melting. HREE abundances in cpx is broadly
negatively correlated with fertility indexes such as Fo% and cr#
spl
(Fig. 4a, d), or positively correlated with cpx modal abundance
(Fig. 4b, c) or CaO content in cpx (not shown). These correlations sug-
gest that HREE abundances in cpx are primarily governed by melting
and melt extraction processes. No such correlation is observed for
light-REE (LREE). Mid-REE show an intermediate behaviour, for in-
stance Sm remains correlated to fertility indexes (e.g., cpx %, Fig. 4c)
for the low temperature groups (G1 and G2); but Sm is enriched in G3
samples relative to the G1 and G2 groups. Chondrite normalised REE
patterns in Fig. 5a, b, d are quite variable in shape for Ray Pic mantle
xenoliths as previously observed by Zangana et al. (1997). Chondrite
normalised REE patterns of G1 and G2 show a continuum from MREE-
depleted to flat MREE–HREE segments (0.85 b(Sm/Yb)
N
≤1.18; N
indicates CI chondrite normalised; values after McDonough and Sun,
1995). In contrast,G3 patterns show overalla MREE enrichment relative
to the HREE (2.8 b(Sm/Yb)
N
≤3.6). LREE abundances are extremely var-
iable, in particular for G1 and G2 peridotites, where (La)
N
varies from
2.7 to 103, leading to a significant variation of the La/Sm ratio (0.3 b
(La/Sm)
N
≤15.7). There is no straightforward relationship between
the occurrence or abundance of amphibole and REE fractionation or
concentration. Indeed, the most LREE-enriched pattern is seen for an
6
5
4
3
2
1
0
Al2O3 in WR (wt.%)
543210
CaO in WR (wt.%)
G1, this study
G2, this study
G3, this study
Ray Pic
Massif Central
Fig. 3. Whole rock(WR) Al
2
O
3
content as a function of WR CaO content in wt.% for RayPic
spinel-peridotitexenoliths. Hollowcircle: G1 Ray Pic peridotites;grey triangle: G2 Ray Pic
peridotites; blackdiamonds: G3 Ray Pic peridotites (seetext and Table 2 for detai ls); black
circlesare previous data fromRay Pic (Zangana et al., 1997);small black dotsare data from
other Massif Central xenoliths from previous studies (Alard et al., 1996; Downes and
Dupuy, 1987; Lorand and Alard., 2001).
Table 3
LA-ICPMS trace and minor element concentrations in clinopyroxene and amphibole (ppm).
Clinopyroxene Amphibole BIR 1G
Group G1 G2 G3 G2 Standard
b1% am N1% am N1%
Sample 13RP05 13RP17 13RP01 13RP02 13RP12 13RP04 13RP14 13RP03 13RP11 14RP06 13RP12 13RP04 13RP03
T° Ca-opx 841 903 840 925 952 880 912 933 1010 1070 952 880 933
n 3 3 333 33 33 3 33311
Sc 77.8 77.9 67.1 67.3 63.3 87.3 77.5 60.3 59.5 63.5 42.1 41.3 57 45.4
V 220 n.a. 295 257 n.a. 270 n.a. 201 n.a. n.a. –––341
Co 17.2 n.a. 17.7 21.8 n.a. 17.2 n.a. 19 n.a. n.a. –––52.5
Rb n.d. b0.01 n.d. n.d. b0.0111 0.03 b0.0097 n.d. b0.0118 b0.112 9.38 0.118 7.37 0.178
Sr 53.7 201 45 72.2 150 53.2 49 304 228 264 416 129 801 104
Y 15.9 16.7 10.6 16 16 21.5 9.96 7.88 17.4 20.2 17.8 27.6 15.4 13.8
Zr 28.1 23.9 20.2 28.6 32.4 40.4 11.9 8.84 30.4 22.7 38.5 34.4 13.5 12.3
Nb 0.035 0.061 0.221 0.104 0.476 0.041 0.107 0.078 0.843 1.1 14.7 1.9 8.28 0.512
Cs n.d. b0.0052 n.d. n.d. b0.0055 b0.009 b0.0047 n.d. b0.0056 b0.0054 0.109 0.0161 0.0335 b0.01
Ba b0.005 0.2 0.15 0.03 0.3 b0.03 0.1 0.06 0.156 0.136 258 2.09 867 5.77
Hf 0.887 0.78 0.658 0.902 0.963 1.02 0.382 0.302 0.649 0.7 0.888 0.795 0.483 0.509
Ta 0.0024 0.0039 0.0152 0.0071 0.0347 0.01 0.006 0.0054 0.082 0.0792 0.296 0.0645 0.461 0.0363
Pb 0.169 0.668 0.0738 0.259 0.461 0.111 0.667 0.516 0.17 0.0809 1.18 0.412 2.03 3.63
Th 0.023 4.87 0.026 0.256 3.03 0.02 0.523 3.7 0.385 0.379 2.73 0.0181 6.64 0.0277
U 0.007 1.62 0.02 0.103 0.704 0.005 0.37 0.769 0.101 0.0811 0.665 0.006 1.35 0.0141
La 0.845 21.4 0.643 2.76 14.4 1.01 3.97 24.4 13.3 13.3 17.1 1.2 39.9 0.569
Ce 2.74 19.6 1.74 5.21 20.1 3.25 3.25 51.1 43.2 36 22.8 3.99 76.3 1.79
Pr 0.548 0.881 0.365 0.735 1.7 0.591 0.338 3.92 6.15 5.04 1.8 0.726 5.37 0.335
Nd 3.55 3.46 2.67 4.08 6.32 3.67 1.91 9.51 26.6 22.4 6.44 4.52 12.7 2.20
Sm 1.4 1.46 1.23 1.55 1.75 1.57 0.84 1 5.18 4.6 1.78 1.85 1.47 1.01
Eu 0.552 0.589 0.522 0.632 0.695 0.571 0.335 0.346 1.67 1.41 0.728 0.738 0.546 0.453
Gd 1.8 2.13 1.37 1.91 2.19 2.3 1.15 1.64 4.19 4.3 2.55 2.76 1.96 1.60
Tb n.a. 0.402 n.a. n.a. 0.414 0.443 0.235 n.a. 0.572 0.612 0.438 0.524 0.335 0.311
Dy 2.88 2.99 2.09 2.89 2.9 3.22 1.75 1.37 3.42 3.9 3.16 3.93 2.52 2.34
Ho n.a. 0.651 n.a. n.a. 0.602 0.709 0.385 n.a. 0.649 0.74 0.667 0.865 0.573 0.517
Er 1.79 1.88 1.2 1.82 1.72 1.97 1.12 0.921 1.78 2.08 1.89 2.5 1.7 1.52
Tm n.a. 0.273 n.a. n.a. 0.249 0.282 0.168 n.a. 0.234 0.292 0.257 0.338 0.261 0.225
Yb 1.74 1.9 1.2 1.73 1.69 1.89 1.1 0.942 1.59 1.85 1.67 2.39 1.7 1.53
Lu 0.253 0.262 0.154 0.239 0.231 0.275 0.147 0.139 0.218 0.255 0.233 0.312 0.243 0.224
Average on n measurements; T° (Ca-in-opx) from Brey and Köhler (1990); with error ± 30 °C; n.a.: note analysed; n.d.: not detected; am: amphibole.
262 C.M.M. Denis et al. / Lithos 236–237 (2015) 256–274

amphibole-rich sample (13RP03), while several amphibole-poor sam-
ples (e.g. 13RP17) also have LREE enriched patterns with an extremely
steep slope from Pr to La (Fig. 5a, b). Conversely, G2 cpx show LREE–
MREE depleted pattern (e.g. 13RP14). Amphiboles from G2 have REE
patterns parallel to the cpx with a D
am/cpx
(REE) = 1.05 ± 0.07 which
is in agreement with experimental and natural literature data (Coltorti
et al., 2000, 2004; Grégoire et al., 2000; Tiepolo et al., 2007; Vannucci
et al., 1995). We conclude that amphibole and cpx are in equilibrium
(Fig. 5b, c).
Extended trace element patterns of cpx from G1 and G2 shown
in Fig. 5e, f are marked by pronounced negative anomalies of Nb
and Ta relative to the LREE ((Nb/Ce)
PM
≤0.3; PM denotes primitive
mantle normalised values, after McDonough and Sun, 1995), and
strong positive anomalies of U, Th ± Sr (i.e., (U/Nb)
PM
up to 916).
While these anomalies also occur in G3 cpx, they are less promi-
nent ((U/Nb)
PM
≤4.1) than in the two other groups. Overall, G2
cpx show lower Rb–Ba content compared to amphibole-poor sam-
ples (G1), in agreement with their preferential partitioning into
amphibole. Indeed, the coexisting amphiboles have Rb and Ba
content at least 2 orders of magnitude higher that those of the
coexisting cpx as shown in Fig. 5g. High Field Strength Elements
(HFSE: Nb, Ta, Zr, Hf and Ti) are known to have more affinity for
amphibole than for pyroxenes (e.g., Coltorti et al., 1999). Indeed,
the HFSE content is significantly higher in the amphibole than in
the coexisting cpx in samples 13RP12, 13RP04 and 13RP03 (see in
Fig. 5f, g). However, only Ti shows a positive anomaly relative to
the HREE and MREE, while Nb–Ta still display a negative anomaly
relative to the neighbouring REE and LILE elements (U, Th). The
fractionation of U and Th relativetoLREEisextremelysimilarto
the one of cpx (kd
am/cpx
(U,Th) = 1.1 ± 0.4). However, Sr displays a
more marked anomaly relative to the neighbouring REE (1 ≤(Sr/Ce)
PM
in am ≤3) than observed in the cpx (kd
am/cpx
(Sr) = 2.7 ± 0.03 for
13RP12 and 13RP03).
4.4. Infrared spectra, OH absorption bands and calculated water contents
Water contents were obtained for 103 olivines, 90 opx and 83 cpx
grains. The olivine spectra show three groups of absorption bands:
14
12
10
8
6
4
2
0
YbNcpx
92919089
Fo %
14
12
10
8
6
4
2
0
YbNcpx
14
12
10
8
6
4
2
0
YbNcpx
50403020100
cr # spl
40
30
20
10
0
SmNcpx
20151050
modal cpx %
0.01
0.1
1
10
100
(La/Yb)Ncpx
1
10
(Sm/Yb)Ncpx
12001000800600
TBKN (°C)
(a)
(b)
(c)
(d)
(e)
(f)
Fig. 4. (a) Yb
N
in clinopyroxene (cpx)(N: normalisedto CI chondrite valuesafter McDonough and Sun(1995)) vs. forsterite content[Fo% = 100 × Mg/(Mg + Fe)]; (b) Yb
N
in cpx vs. modal
cpx %; (c) Sm
N
in cpx vs. modalcpx %; (d) Yb
N
in cpx vs.Cr-number [cr# = 100× Cr/(Cr + Al)] in spinel; (e) (La/Yb)
N
vs. in T
BKN
(°C) (equilibriumtemperature from Brey& Köhler(1990),
typicalerror is ± 30 °C); (f) (Sm/Yb)
N
in cpx vs. T
BKN
(°C). Dashlines in (e) and (f) represent the CI-chondritevalues; grey area:zone of amphibole crystallisation in Ray Picsamples; black
circles are from Ray Pic,black diamonds are for high temperature samples from Zangana et al. (1997). Other symbols are as in Fig. 3.
263C.M.M. Denis et al. / Lithos 236–237 (2015) 256–274

group I from 3598 to 3420 cm
−1
; group II from 3420 to 3280 cm
−1
and
group III from3280 to 3195 cm
−1
as shown in Fig. 6. The group II and/or
III are not present in every spectrum, which could be due to the anisot-
ropy of the crystal lattice, but are here mainly due to the low concentra-
tion of the stretching hydroxyl groups in the olivine lattice (e.g., Beran
and Putnis,1983; Denis et al., 2013; Miller et al., 1987). The absorption
spectra ofopx and cpx each show three major absorption bands at 3590,
3519, 3419 cm
−1
and 3633, 3525 and 3452 cm
−1
,respectively,as
displayed in Fig. 6. These absorption bands are typical of olivine, opx
and cpx from peridotite xenoliths in alkali basalt (Bell and Rossman,
1992; Demouchy et al., 2006, 2015; Denis et al., 2013; Grant et al.,
2007a, 2007b; Ingrin and Skogby, 2000; Peslier and Luhr, 2006; Peslier
et al., 2002; Skogby, 2006; Xia et al., 2010, 2013; Yang et al., 2008; Yu
et al., 2011).
The range and average of water contents in minerals expressed in
ppm wt. H
2
O is reported in Table 4. In G1, the water content averages
of olivineand cpx from 13RP17 (i.e., 6 and 329 ppm wt. H
2
O for olivine
and cpx respectively) are higher than for theother sample within group
G1, which are around 3 and 220 ppm wt. H
2
O for olivine and cpx, re-
spectively, as presented in Fig. 7. The average water contents in NAMs
for G2 are within the same range of concentration than G1 for cpx
(220 ppm wt. H
2
O), slightly lower for olivine (3 ppm wt. H
2
O) and
higher for opx (115 ppm wt. H
2
O). The water contents averages for
each minerals of the G3 are higher than both G1 and G2, i.e., 6, 140
and 280 for olivine, opx and cpx, respectively (Fig. 7). These water con-
tents are typical of olivine, opx and cpx from spl-bearing peridotite xe-
noliths in alkali basalts whatever the calibration used (see Table 4;
e.g., Baptiste et al., 2015; Bonadiman et al., 2009; Demouchy et al.,
2015; Denis et al., 2013; Ingrin and Skogby, 2000; Peslier et al., 2002,
2015; Peslier and Luhr, 2006; Peslier, 2010; Yu et al., 2011; Xia et al.,
2010, 2013).
FTIR measurements along transepts in three large olivine grains
(length N2 mm) and one pyroxene (length N1.5 mm) did not reveal
heterogeneous water content across mineral grains, which would be in-
dicative of H loss (e.g., Demouchy et al., 2006; Denis et al., 2013; Peslier
and Luhr, 2006; Thoraval and Demouchy, 2014).
At last, there is no positive correlation between H
2
O content of the
NAMs and the occurrence and/or abundance of amphibole as display
in Fig. 8 andasreportedbypreviousstudies(e.g.,Bonadiman et al.,
2009; Demouchy et al., 2015).
1
10
100 (c) G2 am
1
10
100
LaCePrNd SmEuGdTbDyHoErTmYbLu
(a) G1 cpx
13RP01
0.001
0.01
0.1
1
10
100 (h) G3 cpx
0.001
0.01
0.1
1
10
(f) G2 cpx
0.001
0.01
0.1
1
10
100 (g) G2 am
0.001
0.01
0.1
1
10
100 (e) G1 cpx
1
10
100 (d) G3 cpx
1
10
100
(b) G2 cpx
13RP12
13RP04
13RP03
13RP14
La Pr Sm Gd Dy Er Yb
Ce Nd Eu Tb Ho Tm Lu
Ba U Ta Ce Sr Sm Hf Ti Tb Ho Er Yb
Rb Th Nb La Pr Nd Zr Eu Gd Dy Y Tm Lu
0.0001
100
Concentration / Primitive Mantle
Concentration / Chondrite
Fig. 5. (a), (b), (c) and (d) Chondrite-normalised REE compositions of clinopyroxene (cpx) and amphibole (am) for Ray Pic spinel-peridotite xenoliths. (a) cpx for G1; (b) cpx for G2;
(c) am for G2; (d) cpx for G3. (e), (f), (g) and (h) primitive mantle normalised trace element compositions of cpx and am; (e) cpx for G1; (f) cpx for G2 ; (g) am for G2; (h) cpx for
G3. CI-chondrite and primitive mantle from McDonough and Sun (1995). Thin grey dashed lines denote Ray Pic cpx from Zangana et al. (1997). Grey arrows in the right panel for Rb
and Ba indicate a concentration below the detection limit (see Table 3).
264 C.M.M. Denis et al. / Lithos 236–237 (2015) 256–274

4.5. Recalculated whole-rock H
2
O contents
Using mineral modal abundances and mineral water contentobtain-
ed by FTIR (see above) and KFT (see Section 4.2), one can compute
whole-rock H
2
O content. We here define an ‘amphibole-free whole-
rock’, which neglects the effect of amphibole on the whole-rock H
2
O
content (i.e., only NAMs are considered) and a ‘whole-rock’which in-
clude NAMs and amphibole. H
2
O content in amphibole-free whole-
rock ranges between 34 and 71 ppm, while whole-rock may reach
values as high as 1663 ± 201 ppm wt. H
2
O for the amphibole-rich
sample 13RP03 (Fig. 8d). These values are in good agreement with
experimental data obtained on am-bearing peridotite at lithosphere
conditions (1.5–2.5 GPa; 1000–1150 °C; Green et al., 2010). Thus,
whole-rock H
2
O contents are dominated by amphibole as shown in
Fig. 8d. The H
2
O content of amphibole-freewhole-rocks are over-all cor-
related with fertility indexes reported in Fig. 9; only the G3 samples
show a small shift toward higher H
2
O content.
5. Discussion
5.1. Record of chemical depletion and metasomatism
The chemical composition of peridotite xenoliths results from two
main processes: (1) melting, with variable degrees of melt extraction
during partial melting, and (2) metasomatism, due to the percolation
of a melt or a fluid through the peridotite matrix (e.g., Frey and Green,
1974; O'Reilly and Griffin, 2013).
We observe continuous depletion trend of the lithophile element
content in minerals (e.g., Al and Na in cpx and opx; Al in spl); which is
correlated with the decreasing modal abundances of cpx ± opx. At
the whole rock scale, these covariations translate a decreasing trend of
Al
2
O
3
and CaO (Fig. 3). Such covariations are identical to the one report-
ed for intracontinental spinel-peridotite from worldwide occurrence
and are interpreted as resulting from melt depletion (e.g., McDonough
and Sun, 1995). Correlation between fertility indexes related either to
WR major element contents, mineral composition and mineral modal
abundances, is here self-consistent (Figs. 8d; 9). This suggests that peri-
dotites from Ray Pic representa residue from variable degrees of partial
melting and melt extraction from a fertile lherzolite source. The rela-
tionships between HREE content and melt depletion indexes such as
cr#spl, Fo%, CaO content of cpx, suggest that HREE abundance was pri-
marily controlled by melt depletion and extraction processes (Fig. 4)
and even for the G3 or amphibole-rich samples (see below). The varia-
tion of HREE (i.e., Yb) content in cpx indicates two different ranges of
melt-depletion (Fig. 5). Using the approach from Norman (1998) and
a primitive mantle composition from McDonough and Sun (1995),we
calculate that the most Yb-rich samples (e.g., 13RP17) have undergone
3 to 4% of partial melting using a batch melting model or 3 to 6% of par-
tial melting using a fractional melting model as shown in Fig. 10. The
significantly lower Yb values for samples 13RP01, 13RP14 and 13RP03
suggest a higher degree of melting estimated between 8 and 10% and
between 12 and 16%, using both batch and fractional melting models,
respectively (Fig. 10). We thus consider that modal abundances and
major element characteristic of olivine and pyroxenes results from
small to moderate percentage of partial melting and melt extraction.
The results from Zangana et al. (1997; 1999), based on a larger
number of xenoliths from Ray Pic,encompass our range of melt deple-
tion (1–10% for batch and 1 to 18% for fractional melting). The high
143
Nd/
144
Nd (0.51347–0.51305) and low
87
Sr/
86
Sr (0.7019–0.7029), re-
ported by Zangana et al. (1997) for LREE-depleted cpx, attest a long
term melt-depletion of the mantle lithosphere beneath the Ray-Pic
volcano, possibly since the late Proterozoic.
The occurrence of amphibole and/or the extremely variable abun-
dances and fractionation of the LREE ± MREE, LILE and HFSE relative
to the HREE show that the slightly to moderately depleted protolith
(i.e., the parent rock) was overprinted by metasomatic events. Modal
metasomatism is here shown by the occurrence of amphibole, with
modal abundance up to 8%. Microstructural context, major element
chemistry and trace element content of the amphibole show that
12
10
8
6
4
2
0
Absorption coef. (cm-1)
4000 3600 3200 2800
Wavenumber (cm-1)
Ol integration field
3571
3355
3225
13RP01
13RP02
13RP11
13RP12
13RP14
13RP17
13RP03
13RP04
13RP05
(a)
Ol
unpol.
ppm wt. H2O
14RP06 5.5
3.0
3.0
2.2
2.1
3.5
5.9
3.2
3.7
5.2
4000 3600 3200 2800
Wavenumber (cm-1)
Cpx integration field
3633
3525
3452
(c)
Cpx
unpol.
ppm wt. H2O
233
226
211
203
210
213
330
240
247
329
60
40
20
0
Absorption coef. (cm-1)
4000 3600 3200 2800
Wavenumber (cm-1)
Opx integration field
3590
3519
3419
13RP01
13RP11
13RP12
13RP14
13RP17
13RP02
13RP03
13RP04
13RP05
(b)
Opx
unpol.
ppm wt. H2O
14RP06 114
66
111
102
97
96
160
141
132
107
Fig. 6. Representative unpolarized FTIR spectra for Ray Pic spinel-peridotite xenoliths: (a) olivine (ol), (b) orthopyroxene (opx), and (c) clinopyroxene (cpx). Spectrawith minimal and
maximalabsorbance aregiven for each sample.All spectra are normalised to 1 cm of thickness. Longblack arrows indicate the integration rangeused for the IR calibration; theshort dashed
arrows indicate the major OH absorption bands of the mineral.
265C.M.M. Denis et al. / Lithos 236–237 (2015) 256–274

amphibole is in equilibrium with the other minerals (e.g., cpx). Despite
variable modal abundance and distinct microstructural occurrence, all
amphiboles show similar major element composition, suggesting that
they are all related to the same metasomatic event involving the same
mantle source.
However, amphibole-bearing samples show variable REE fraction-
ation going from depleted [(La/Sm) b1] to enriched [(La/Sm) N1]
pattern. The shape of REE pattern is clearly independent of the occur-
rence and/or abundance of amphibole (Fig. 5). Therefore, amphibole
could have been a component of the protolith, perhaps formed during
an earlier metasomatic event (n), postdating a partial melting event
(n −1). Then, within this scenario, the occurrence of amphibole is not
related to the current trace element enrichments and fractionations
due to, at least one more recent metasomatic event (n + 1), and
which has overprinted the trace element signature related to the am-
phibole crystallisation events. However such scenario is hardly reconcil-
able with the fact that some amphibole shows LREE depleted pattern
(13RP04). Alternatively, the necessary increase of H
2
O fugacity in the
percolating melt/fluid, which is required for crystallisation of amphi-
bole, could have occurred at a higher distance from the source than
the chromatographic front for LREE for instance (e.g., Le Roux et al.,
2007), and produced an offset between the LREE enrichment zone
and the amphibole crystallisation zone. This scenario is supported by
the fact that LREE depleted amphibole occurred mostly in low-
temperature samples (e.g., Alard et al., 2011; Touron et al., 2008),
13RP04 has also a low equilibrium temperature (T° Ca-in-opx =
880 ± 30 °C).
On the one hand, the strong LREE and LILE (U, Th, Sr) enrichment
without concomitant enrichment of the HFSE and especially Nb–Ta
(G1 and G2), is a trace element signature, which has been previously
identified for peridotites from many localities and is generally ascribed
to a carbonatitic or carbonated metasomatism (Alard et al., 2011;
Coltorti et al., 1999; Dautria et al., 1992; Ionov et al., 1993; Rudnick
et al., 1993; Yaxley et al., 1998). Sun and Kerrich (1995) has ascribed
it to be H
2
O–CO
2
fluids. While the true nature of the metasomatic
melt/fluid remainsunclear, all studies agree on (1) the volatile-richna-
ture of such metasomatic melt and (2) the need for this metasomatism
to occur at low melt/rock ratio (≪1%; e.g., Bedini et al., 1997). The
volatile-rich nature of the fluid is required to significantly lower the vis-
cosity, dihedral angles and crystallisation temperature of the fluid/melts
and thus allow a more efficient percolation through the upper and
colder part of the subcontinental lithospheric mantle (SCLM; Minarik
Table 4
Water content in olivineand pyroxenes (ppm wt H
2
O, based on thecalibration of Paterson, 1982 (P82) and also calculated forother calibrations, B2003= Bell et al., 2003;W12=Withers
et al., 2012,B95=Bell et al., 1995.
Group G1 G2 G3
b1% am N1% am
Sample 13RP05 13RP17 13RP01 13RP02 13RP12 13RP04 13RP14 13RP03 13RP11 14RP06
Olivine: n 6 15 9 8 12 11 12 8 18 4
Calc. Area 18.6 34.0 16.2 15.5 18.8 12.3 21.1 12.8 31.0 26.2
±1SD 6.1 8.3 4.7 5.9 3.5 4.3 7.0 3.3 8.0 7.5
min P82 1.9 3.1 2 2.0 1.9 1.3 1.4 1.5 3.5 3.9
max P82 5.2 7.1 3.9 5.3 4.2 4.2 5.9 3.6 8.7 7.8
C
OH
avg P82 3.5 5.2 3 3.0 3.2 2.1 3.7 2.2 5.9 5.5
±1SD 1.1 1.2 0.7 1.2 0.6 0.8 1.3 0.6 1.4 1.5
C
OH
avg B2003 10.5 19.2 9.1 8.7 10.6 6.9 11.9 7.2 17.5 14.8
C
OH
avg W12 6.6 12.1 5.8 5.5 6.7 4.4 7.5 4.6 11.1 9.4
Opx: n 6 6 10 11 9 10 8 12 12 6
Calc. area 496 531 335 566 733 494 694 505 847 555
±1SD 40 77 37 91 90 58 102 63 145 58
min P82 88 92 58 89 118 87 101 92 136 105
max P82 106 124 75 136 177 118 154 122 196 124
C
OH
avg P82 96 107 66 111 141 97 132 102 160 114
±1SD 6 10.6 4.8 12.7 13.1 9 15.0 8.1 20.1 7.7
C
OH
avg B95 100 107 68 114 148 100 140 102 171 112
Cpx: n 10 5 11 8 8 9 9 13 5 5
Calc. area 843 1320 901 857 923 792 1138 801 1472 977
±1SD 183 158 292 34 163 178 80 160 147 247
min P82 138 268 145 181 189 126 214 128 287 183
max P82 275 398 375 221 302 266 289 248 374 306
C
OH
avg P82 213 329 226 211 240 210 247 203 330 233
±1SD 44 46.2 70.1 7.9 26.4 43 23.5 31.4 30.1 45.7
C
OH
avg B95 357 559 381 363 391 335 482 339 623 413
C
OH
WR (w/. am) 42 62 38 65 64 52 34 34 71 55
±1σ7 101111119 6 6 129
C
OH
WR (w. am) 42 62 78 165 285 378 389 1663 72 55
±1σ7 10 11 23 37 48 48 201 12 9
Kd
cpx/opx
(H
2
O) 2.2 3.1 3.4 1.9 1.7 2.2 1.9 2.0 2.1 2.0
±1σ0.5 0.5 1.1 0.2 0.3 0.5 0.3 0.4 0.3 0.4
Kd
cpx/ol
(H
2
O) 61 63 75 71 75 100 67 92 56 42
±1σ24 18 31 30 17 44 29 32 15 15
Kd
opx/ol
(H
2
O) 27 20 22 37 44 46 36 47 27 21
±1σ8.8 5.1 5.4 15.4 9.1 18.0 13.4 14.1 7.3 5.8
Factor
a
ol 16.32 16.33 16.35 16.36 16.36 16.33 16.30 16.32 16.26 16.31
Factor
a
opx 11.58 11.60 11.61 11.59 11.59 11.55 11.58 11.57 11.58 11.56
Factor
a
cpx 12.12 12.13 12.07 12.08 12.09 12.07 13.75 12.11 12.10 12.09
Average (avg) on n grains; 1σ: absolute uncertainties calculated by algebraic method;
a
denotes factor to convert water contents in ppm wt. H
2
O to H/10
6
Si; opx: orthopyroxene; cpx:
clinopyroxene; am: amphibole. In average, the intercalibration conversion factors (using unpolarizedIR) are 3.14, 1.99, 1.03 and 1.73 for (B2003/P82)
ol
, (W12/P82)
ol
, (B95/P82)
opx
and
(B95/P82)
cpx
, respectively.
266 C.M.M. Denis et al. / Lithos 236–237 (2015) 256–274

and Watson, 1995; Watson and Brenan, 1987). Samples from the G1
and G2 groups have equilibrium temperature below 1000 °C. Strong
fractionation between REE as observed in G1 and G2 samples, could
be easily produced by melt percolation at low melt/rock ratio (Navon
and Stolper, 1987). On the other hand, MREE enriched pattern is com-
monly observed for cpx metasomatised at high melt–rock ratio such
as the one found near or within pyroxenite veinlets/dykes (Kourim
et al., 2014; Witt-Eickschen and Kramm, 1998). Melt/fluid-percolation
models (e.g., Bodinier et al., 1990; Navon and Stolper, 1987; Vasseur
et al., 1991; Vernières et al., 1997) indicate that MREE enrichment
over HREE can only be obtained for relatively large melt/rock ratios
(i.e., ≥1%, Bodinier et al., 1990; Navon and Stolper, 1987). Melt percola-
tion/reaction at higher melt/rock ratio will produce more steady and
continuous enrichment from the HREE to the LREE (i.e., linear pattern)
and reduce fractionation between incompatible elements as observed
in G3. Although melt percolation and temperature propagation are
two distinct processes, this type of interaction should occur closer to
the melt source and thus peridotite samples should have a higher equi-
librium temperature than measured in G1 and G2 samples (see also
Alard et al., 1996; Xu et al., 1998). It is the case for the equilibrium tem-
perature of G3 samples (N1000 °C). The relationship between REE frac-
tionation and equilibrium temperature in the Ray Pic xenoliths is quite
remarkable (Fig. 4e, f) and was first reported by Zangana et al. (1997).
Accordingly, Fig. 4f shows that only samples equilibrated at high tem-
perature (i.e., T N1000 °C, G3 samples) are enriched in MREE relative
to the HREE (i.e., (Sm/Yb)
N
N1). The large fractionation ofthe LREE rel-
ative to the HREE (and MREE not shown) requiring interactions at low
melt/rock ratio that occur for xenoliths equilibrated at relatively low
temperature (down to 850 °C). Samples equilibrated at low tempera-
ture (b850 °C) are not affected by the selective LREE enrichment
(Fig. 4e), they are thus located ahead of the La-chromatographic front.
12
10
8
6
4
2
0
COH
0
5
10
15
20
COH (W12) (ppm wt. H2O)
G1 G2 G3
(a)
600
500
400
300
200
100
0
COH (P82)
13RP05
13RP17
13RP01
13RP02
13RP12
13RP04
13RP14
13RP03
13RP11
14RP06
0
200
400
600
800
1000
COH (B95)
(c)
250
200
150
100
50
0
COH (P82)
0
50
100
150
200
250
COH (B95)
(b)
OLIVINE
OPX
CPX
(ppm wt. H2O)
(P82)
Fig. 7. Watercontents in nominally anhydrous minerals for eachRay Pic spinel-peridotite
xenoliths (in ppm wt. H
2
O, based on calibration of Paterson, 1982 (P82); Withers et al.,
2012 (W12); and Bell et al., 1995 (B95)). Symbols are average values from N analyses
for each sample. Verticallines across each symbol give the standard deviation of the aver-
age valuesof water content per sample(i.e., 1SD, see Table 4), and horizontallines give the
uncertainties on the average values per group (using 30% of uncertainty from calibration
of Paterson, 1982). Opx: orthopyroxene; cpx: clinopyroxene; symbols are as in Fig. 3.
400
350
300
250
200
150
100
COH in cpx
200
300
400
500
600
CHO )59B(
CPX (c)
250
200
150
100
50
0
COH (P82)
0
50
100
150
200
250
CHO )59B(
OPX (b)
8
6
4
2
0
COH (P82) (ppm wt.H2O)
0
2
4
6
8
10
12
14
CHO H.twmpp()21W( 2)O
OLIVINE (a)
80
60
40
20
0
COH (P82) dry-WR calc. (ppm)
2000
1500
1000
500
0
COH (P82) wet-WR calc. (ppm)
12840
modal am %
(d)
Fig. 8. Water contents for Ray Pic spinel-peridotite xenoliths (in ppm wt. H
2
O, based on
calibration of Paterson, 1982 (P82); Withers et al., 2012 (W12); and Bell et al., 1995
(B95); see Table 4) as a function of the mineral mode of amphibole in % for (a) olivine
(ol), (b) orthopyroxene (opx), and (c) clinopyroxene (cpx), error bar on the average
values of water content per sample is 1SD (Table 4); (d) water contents for amphibole-
free whole rock (WR, C
OH
WR (w/. am)), and whole rock (C
OH
WR (w. am)). WR water
contents(C
OH
WR (w. am)) in grey refers to the grey axis on the right hand side; absolute
uncertainty on the WR water content was calculated by algebraic method for error prop-
agation. Symbols are as in Fig. 3.
267C.M.M. Denis et al. / Lithos 236–237 (2015) 256–274

This relationship suggests that both types of metasomatic finger-
prints are related to the same melt/fluid. The percolating melt evolved
along the conductive geotherm by successive volume reduction, due
to the reaction with the peridotite matrix and/or new crystallised
phases. The resulting melt/fluids is thus enriched in themost incompat-
ible elements, including the volatiles (e.g., Bedini et al., 1997). The scale
at which these iterative metasomatic melt reactions operate is still
under debate. Indeed, two hypotheses are possible: (1) a regional
scale heterogeneity involving mantle diapir or plume (e.g., Bedini
et al., 1997) or (2) a metre to centimetre scale heterogeneity involving
melt channels (Bodinier et al., 1991; Grégoire et al., 2000; Kourim
et al., 2014) and leaving behind pyroxenite dykes, veinlets and/or
metasomatised wall-rocks. To favour one or the other scenarios de-
pends on the interpretation of equilibrium temperatures: (i) either
they record temperatures representing in situ conditions in the litho-
spheric mantle, then they are equivalent to the emplacement depth of
the xenoliths; or (ii) they are related to the metasomatic process itself
and its heating effect. On the basis of both geophysical data, which
have imaged an asthenosphere diapir/plume beneath the FMC (Granet
et al., 1995); and the isotopic composition of the LREE enriched xeno-
liths; Zangana et al. (1997) favoured the first hypothesis. Other authors
working on the mantle lithosphere of the FMC (e.g., Alard et al., 1996;
Lenoir et al., 2000; Xu et al., 1998) have also favoured this interpreta-
tion. However, the common occurrence of pyroxenites among the man-
tle xenoliths from Ray Pic (wherlites, websterites), the similarity of
pyroxenite equilibrium temperatures (≈1030–1050 °C, Zangana et al.,
1997) with temperatures obtained for the G3 peridotite xenoliths
(≈1040 ± 30 °C) lead us to consider that the second scenario is here
a relevant alternative.
5.2. The preservation of the initial water content of the mantle source
Given the high diffusivity of H in olivine (e.g., Demouchy and
Mackwell, 2006; Mackwell and Kohlstedt, 1990), the relevance of the
low H concentration in olivine as an indicator of the water content in
mantle source was questioned (Demouchy and Mackwell, 2006; Hao
et al., 2014a, 2014b; Ingrin and Skogby, 2000). It was argued that the
low water content of olivine (b5ppmwt.H
2
O) in mantle xenoliths
might be due to a rapid ionic diffusion (dehydration) during ascent to-
ward the surface (Demouchy et al., 2006; Denis et al., 2013; Peslier
and Luhr, 2006). However, such dehydration profiles are not ubiquitous
in all mantle olivine and other parameters than temperature, control-
ling H ionic diffusion, remain to be identified and quantified. In this
study, concentration profiles along transects across olivine and pyrox-
ene grains evidence long concentration plateaus (e.g., to the contrary
to short plateau), which permit to rule out significant dehydration
(see Thoraval and Demouchy, 2014).
Since olivine appears to be a weak indicator of the water source, the
emphasis was put on pyroxenes (e.g., Denis et al., 2013; Hao et al.,
2014a, 2014b; Xia et al., 2010). Indeed, the relatively constant water
content ratio between cpx and opx in spl-bearing peridotites hosted in
alkali basalts suggests that they preserve the initial H
2
O contents of
the mantle source (e.g., Denis et al., 2013; Hao et al., 2014a, 2014b;
Peslier et al., 2012; Xia et al., 2010). For the Ray Pic xenoliths, the ratio
of water concentration between opx and cpx range between 1.7 and
3.4, this variation is irrespective of amphibole occurrence, cpx or opx
2015105
modal cpx %
100
80
60
40
20
0
COH (P82) WR (w/. am)
2O)
(a)
321
Ybcpx (ppm)
(b)
(ppm wt. H
Fig. 9. Water contents for amphibole-free whole rock (C
OH
WR (w/. am)) in ppm wt. H
2
O
as a function of(a) modalpercentage of clinopyroxene (cpx in %); (b) Yb in cpx given in
ppm. Hollow circle: G1 Ray Pic peridotites;grey triangle: G2 Ray Pic peridotites; blackdi-
amonds:G3 Ray Pic peridotites (seetext and Table 2 for details). Concentrationsare given
for the calibration of Paterson (1982).
800
700
600
500
400
300
200
100
COH in cpx (P82)
2015105
YbN in cpx
200
400
600
800
1000
1200
COH in cpx (B95)
(c) CPX
350
300
250
200
150
100
50
COH in opx (P82)
100
150
200
250
300
350
COH in opx (B95)
(b) OPX
10
8
6
4
2
COH (P82) (ppm wt. H2O)
5
10
15
20
COH (W12) (ppm wt. H2O)
(a) OLIVINE
FM1
FM2
FM1
FM2
FM1
FM2
Fig. 10. Water contents for Ray Pic spinel-peridotite xenoliths (in ppm wt. H
2
O, based on
calibration of Paterson, 1982 (P82);Withers e t al., 2012 (W12); andBell et al., 1995 (B95))
as a functionof Yb
N
in cpx (N: normalised to CI chondrite valuesfrom McDonoughand Sun
(1995)) for (a) ol; (b) opx and (c) cpx. Symbols are as in Fig. 3. Solid lines are for batch
meltingusing the methodof No rman (1998). Fertilemantle (FM) contains18% of cpx hav-
ing either700 ppm wt. H
2
O (FM1), or 400ppm wt. H
2
O (FM2). Thisrange (FM2–FM1) en-
compasses the variationsof H
2
O content in spinel reported in the literature (e.g., Bell and
Rossman, 1992; Grant et al., 2007a,2007b; Hao et al., 2014a, 2014b; Li etal., 2014; Peslier
and Luhr, 2006; Peslier et al.,2002). Black lines are for a D
cpx/melt
(H
2
O) = 0.25, grey lines
are for a D
cpx/melt
(H
2
O) = 0.1 (see maintext for details). D
ol/melt
(H
2
O) and D
opx/melt
(H
2
O)
were derived using Kd
cpx/ol
(H
2
O) = 70 and Kd
cpx/opx
(H
2
O) = 2.2 (see text and Fig. 11).
White, grey and black stars indicate 5%, 10% and 15% of melting degrees, respectively.
268 C.M.M. Denis et al. / Lithos 236–237 (2015) 256–274

modal abundances, melting or metasomatism indexes. These ratios of
water concentration between opx and cpx can be interpreted as an
apparent partition coefficient (Kd
cpx/opx
(H
2
O)). For the 10 Ray Pic-
xenoliths Kd
cpx/opx
(H
2
O)
RP
is c.a., 2.2 ± 0.6 in average using the
calibration of Paterson (grey area in Fig. 11), which is within error of
the literature average Kd
cpx/opx
(H
2
O) = 1.8 ± 0.8 obtained on a
range of spl-bearing xenoliths in alkali basalts from worldwide occur-
rences (Demouchy et al., 2015; Denis et al., 2013; Soustelle et al.,
2013). Using the calibration of Bell et al. (1995),Kd
cpx/opx
(H
2
O)
RP
range between 2.9 and 5.7 and average c.a. 3.8 ± 0.9. The range of
Kd
cpx/opx
(H
2
O) for spl-bearing xenoliths, excluding the one from cra-
tons, varies from 0.5 up to 6.5 and yield an average value c.a. 2.6 ± 0.9
(N = 182, Bonadiman et al., 2009; Falus et al., 2008; Grant et al.,
2007a, 2007b; Li et al., 2008; Peslier and Luhr, 2006; Peslier et al.,
2002; 2015; Xia et al., 2010; Yu et al., 2011). Thus, although the average
Kd
cpx/opx
(H
2
O) for Ray-Pic xenoliths are slightly higher than the“global”
average, it remains within error. The dataset used for calculated
these Kd
cpx/opx
(H
2
O) is shown in Fig. 12 where data are presented
according to the calibration of Bell et al. (1995) as well as the cali-
bration of Paterson (1982) for our data. Indeed, the water contents
for Ray-Pic xenoliths are slightly higher than the “global”average
after conversion but remains within the main trend (Fig. 12).
These concentration ratios for natural pyroxenes are in relative
good agreement with partition coefficients obtained by experi-
mental petrology, notably with Kd
cpx/opx
(H
2
O) = 1.4 from
Aubaud et al. (2004) at equilibrium conditions (1 GPa, 1200 °C);
Kd
cpx/opx
(H
2
O) = 1.8 from Novella et al. (2014) at equilibrium condi-
tions (6 GPa, 1400 °C) and Kd
cpx/opx
(H
2
O) = 1.7 from Kovács et al.
(2012) at equilibrium conditions (2.5 GPa, 1000 °C; Fig. 11a). We also
note that the Kd
opx/ol
and Kd
cpx/ol
for the Ray Pic xenoliths are, like the
Kd
cpx/opx
, relatively constant (same relative standard deviation) and
equal to 33 ± 10 and 70 ± 17, respectively (Fig. 11b, c). These relatively
constant water content ratios further support that water content in
olivine has not been affected by significant dehydration process. Finally,
dehydration processes are unlikely to preserve correlations of olivine
water contents with fertility indexes (see next section). Therefore,
these observations suggest that the water content of NAMs in the Ray
10
8
6
4
2
0
COH (P82) in ol
4003002001000
COH (P82) in cpx (ppm wt. H2O)
0
5
10
15
COH (B95) in ol
0 200 400 600
COH (B95) in cpx
Kd
cpx/ol = 70 ±17
(c)
12.5
16.6
21.3
25
9.2
10
8
6
4
2
0
COH (P82) in ol
250200150100500
COH (P82) in opx (ppm wt. H2O)
0
5
10
15
COH (W12) in ol
Kd
opx/ol = 33 ± 10
(b)
1
2
3
4
5
6
7
6.5
9.1
9.7
15.5
11.7
500
400
300
200
100
0
COH (P82) in cpx (ppm wt. H2O)
0
200
400
600
800
COH (B95) in cpx
0 50 100 150 200 250
COH (B95) in opx
Kd
cpx/opx = 2.2 ± 0.6
G1
G2
G3
(a)
1.4
1.2
1.8
1.7
Fig. 11. Water contents for Ray Pic spinel-peridotite xenoliths (in ppm wt. H
2
O, based on
calibration of Paterson, 1982 (P82); Withers et al., 2012 (W12); and Bell et al., 1995
(B95)).(a) In ol as a functionof water content inopx; (b) in cpx as a function of watercon-
tent in opx; (c) in ol as a function of water content in cpx. Symbols are as in Fig. 3; grey
areas are the average water content ratio using 30% of uncertainty for mantle-derived
mineralsof G1, G2 and G3. Blacklines are experimentally determined partition coefficients
from (1) Ardia et al., 2012 (5–8 GPa; 1400–1450 °C); (2) Aubaud et al., 2004 (1–2 GPa;
1230–1380 °C); (3) Grant et al., 2007a, 2007b (1.5 GPa; 1295–1320 °C); (4) Hauri et al.,
2006 (cpx/opx: 1.2–1.6 GPa; 1185–1370 °C; cpx/ol and opx/ol: minimum and maximum
values at 1.2–1.6 GPa; 1185–1370 °C); (5) Tenner et al., 2009 (3–5GPa;1350–1440 °C);
(6) Novellaet al., 2014 (6 GPa; 1400 °C); (7)Kovács et al., 2012 (2.5 GPa; 1000°C). Values
from IR-based studies (Grant et al.,2007a, 2007b; Kovácset al., 2012) were converted to
Paterson's calibration for homogeneity.
1000
800
600
400
200
0
C
OH
in cpx (ppm wt. H
2
O)
5004003002001000
COH in opx (ppm wt. H2O)
B95-pol.
B95-unpol.
P82 conv. to B95
This study (P82)
This study (B95)
Fig. 12. Water contents in cpx as a function of water content in opx for spinel-peridotite
xenoliths in alkali basalt (in ppm wt. H
2
O, based on calibration of Paterson, 1982 (P82)
and Bell et al.,1995 (B95)). Open squaresare for polarized IR measurements using the cal-
ibration of Bell et al. (1995) (Grant et al., 2007a, 2007b; Li et al., 2008;Peslier et al., 2002,
2015; Peslier and Luhr, 2006; Peslier and Bizimis, 2015); open triangles are for unpolar-
ized measurements using the calibration Bell et al. (1995) (Bonadiman et al., 2009;
Falus et al., 2008; Xia et al., 2010;Yu et al., 2011); open circles are for values using the cal-
ibration Paterson (1982) with unpolarized measurements, then converted to the calibra-
tion of Bell et al. (1995) (Demouchy et al., 2015; Denis et al., 2013; Soustelle et al., 2013).
Filled blackand grey diamonds are for results from this study usingthe calibration of Pat-
erson (1982) and using calibration of Bell et al. (1995) respectively. For clarity, error bars
(1SD) are only shown for the results from this study.
269C.M.M. Denis et al. / Lithos 236–237 (2015) 256–274

Pic xenoliths have not been significantly modified during ascent toward
the surface.
5.3. Hydrogen behaviour during mantle processes
5.3.1. Partial melting
Water contents in ol, opx and cpx are roughly correlated with [Yb]
cpx
(Fig. 10), which is used here as a proxy of melt depletion (Fig. 4). Simi-
larly, calculated H
2
O content in the amphibole-free WR is correlated to
fertility indexes (Fig. 9). In particular for G1 and G2 samples. Therefore
these correlations suggest that water abundances in Ray-Pic NAMs
could be mainly related to melting processes. Despite some scatter,
melting curves reproduce broadly the observed distribution of the Ray
Pic samples within the [Yb]
cpx
–C
OH
cpx,opx, ol
spaces (Fig. 10). The best data
fits are obtained for D
cpx/melt
(H
2
O) ranging between 0.1 and 0.25 and
for an initial cpx water content between 400 and 700 ppm (see also
Fig. 10 for further detail). The range of D
cpx/melt
(H
2
O) used here is at
least one order of magnitude higher than results from experimental
petrology in modelled systems (D
cpx/melt
=0.013–0.048; Aubaud
et al., 2004; Aubaud et al., 2008; Hauri et al., 2006; Novella et al.,
2014; O'Leary et al., 2010; Tenner et al., 2009). Nevertheless, such
low experimental D
cpx/melt
(H
2
O) (i.e., 0.01–0.05) implies that the
mantle will be completely deprived of its water (i.e., C
OH
cpx
b50 ppm,
using a batch-melting model) from the first melting increments (i.e., 1–
5%). On the other hand, D
cpx/melt
(H
2
O) ranging between 0.1 and 0.25
are in good agreement with the water content in pyroxenes from
Hao et al. (2014a, 2014b),whosuggestD
peridotite/melt
(H
2
O) ≈0.1.
Relative to lithophile elements, D
cpx/melt
(H
2
O) ≈0.1–0.25 would
suggest that H behaves as a moderately incompatible element as Sm
(D
cpx/melt
(Sm) ≈0.25; e.g., Hart and Dunn, 1993; Witt-Eickschen and
Kramm, 1998) and not as a very incompatibleelements (D
cpx/melt
(Ce) ≈
0.075; D
cpx/melt
(La) ≈0.055; Hart and Dunn, 1993; Norman et al.,1996)as
it was previously proposed (Dixon et al., 2002).
Although, we do not intend to ascribe all variationsof water contents
in NAMs to a single melting processes, it appears that for the Ray-Pic
xenoliths under consideration, melting processes can possibly account
for a large part of water contents variability within NAMs. Nevertheless,
the residual scatter of the data around the melting curves (Fig. 10)also
suggests that other mantle processes besides partial melting have been
involved.
5.3.2. Modal hydrous metasomatism
While it was shown that amphibole could be hydroxyl-poor and
halogen-rich (e.g., Hawthorne et al., 1998; King et al., 1999; Oberti
et al., 2007), a large number of mantle pargasites were proven to be
water-rich (H
2
O≈1–2 wt.%; e.g., Bonadiman et al., 2014; Moine et al.,
2000). Therefore, the occurrence of amphibole in mantle rocks is recur-
rently taken as a water indicator. Numerous authors refer, sometimes
abusively, to amphibole-bearing mantle rock as “wet”mantle, while
amphibole-free samples are usually coined as “dry”mantle. In the xeno-
liths from Ray Pic, no correlation between the water content of the
NAMs and the occurrence or abundance of amphibole is found
(Fig. 8). Indeed the NAMs in amphibole-poor/free xenoliths from G1
and amphibole-rich xenoliths from G2 show the same range of water
contents. Major element and trace element contents of the amphibole
show that they are in equilibrium with the coexisting silicates. There-
fore, this could lead to the conclusion that the pargasite from Ray Pic
samples could be water-poor. However, here the pargasites from Ray
Pic are water-rich amphiboles. Karl–Fischer titration analyses of
pargasite from 13RP03 sample has yield content ca. 1.9 ± 0.5 wt.%
H
2
O. Given the compositional homogeneity of the amphiboles from
Ray Pic, all amphiboles from Ray Pic are considered water-rich.
This lack of effect and/or correlation between amphibole abundance
and the water content of NAMs was also recently observed by
Schmädicke et al. (2013) for olivine in Eifel xenoliths, by Demouchy
et al. (2015) for NAMs in Ontong Java Plateau xenoliths and by
Bonadiman et al. (2009) for cpx and opx in the Baker rock xenoliths
(Antartica). The study of Bonadiman et al. (2009; 2014) shows that
the amphiboles contain 0.84–1.42 wt.% of H
2
O, with water contents of
opx and cpx ranging from low water contents (39 and 83 ppm wt.
H
2
O for opx and cpx, respectively), to fairly typical contents (166 and
399 ppm wt. H
2
O for opx and cpx) according to spinel-bearing perido-
tites worldwide (e.g., Bolfan-Casanova, 2005; Ingrin and Skogby,
2000; Peslier, 2010). Although, it may be seen as counterintuitive, wet
amphibole containing up to 2 wt.% of H
2
O, has no effect on the water
content of coexisting NAMs, nor on water distribution between NAMs.
In other words, the so-called hydrous metasomatism has no effect on
the water content of NAMs and its only effect is to crystallise a water-
rich mineral: amphibole. Further NAMs in equilibrium with water rich
pargasite are not water richer (or poorer) than NAMs in amphibole
free peridotite, suggesting thus that H diffusion from amphibole to
NAMs was extremely reduced. It may be due to the difference in speci-
ation of H in these two solid phases. Indeed, in amphibole, H occurs as
an independent hydroxyl group (OH) in the O3 structural site of the lat-
tice, while in NAMs, only H (proton, no external oxygen companion) is
located in the lattice mineral asimpurities in cationic vacant sites (point
defect).
Using our data, from 13RP03, we derive a concentration ratio which,
assuming that thermodynamic equilibrium was achieved in the natural
rock, could be equivalent to partition coefficient at high temperature
between pargasite and NAMs: Kd
ol/am
(H
2
O) = 0.11 ± 0.04 × 10
−3
;
Kd
opx/am
(H
2
O) = 5.9 ± 0.6 × 10
−3
and Kd
cpx/am
(H
2
O) = 11.4 ±
2.5 × 10
−3
. These results (once converted to the same calibration)
agree with the data of Bonadiman et al. (2009, 2014) for which
we can derived an estimate Kd
opx/am
(H
2
O)
BR avg
≈9.8 × 10
−3
and
Kd
cpx/am
(H
2
O)
BR avg
≈17 × 10
−3
(BR avg: average for Baker Rockperi-
dotite xenoliths).
5.3.3. Cryptic metasomatism
The relationship between cryptic metasomatism and water content
of the NAMs is not straightforward. Indeed there is no obvious
correlation between the classic metasomatic indexes such as LREE
content (e.g., (La)
N
;Fig. 13a, b, c) or REE fractionation (e.g., (La/Sm)
N
;
(La/Yb)
N
not shown;or (Sm/Yb)
PM
Fig. 13d, e, f) and the water content
of NAMs. This suggests that H does not behave as a highly incompatible
element (e.g., like La) as discussed above for melt-depletion processes.
Indeed, despite strong chromatographic enrichment of the LREE relative
to the HREE and MREE, there is no correlatedwater enrichment. The G3
samples show overall a relatively high water content, in particular for
olivine (Fig. 13 d), which would suggest that water enrichment is only
possible for NAMs previously modified by reactions at high melt/rock
ratio.
The lack of correlation between water content of the NAMs and the
Th/Nb
PM
ratio (Fig. 12g, h, i) is of particular interest. Indeed, the frac-
tionation of the LILEand LREE relative to the HFSE is seen as symptom-
atic of the so-called carbonatitic metasomatism (Dautria et al., 1992;
Ionov et al., 1993; Rudnick et al., 1993; Yaxley et al., 1998). This type
of metasomatic melt/fluid is meant to be rich in volatile (H
2
O–CO
2
)as
proposed by Sun and Kerrich (1995). However, for the Ray-Pic xeno-
liths, there is no significant enhancement effect on hydrogen incorpora-
tion into NAMs, related to the percolation of such melt/fluid. Thus,
either the volatile inventory of this melt/fluid does not include water,
or the coexistence of CO
2
and H
2
O has drastically lowered the fugacity
of water in this system (Baptiste et al., 2015; Shirey et al., 2013). How-
ever in both cases, the low ƒH
2
O should preclude the crystallisation of
H
2
O–bearing pargasite and thus lead to consider that amphiboles
were part of a previous metasomatic event, which is not our preferred
interpretation.
It could be argued that the water contents in opx showssome degree
of correlation with metasomatic indexes (Fig. 13b, e). However, water
incorporation into opx seems linked to the Ca (Fig. 14) and Al contents
(not shown). This was also observed by previous works (e.g., Bell et al.,
270 C.M.M. Denis et al. / Lithos 236–237 (2015) 256–274

2004; Ingrin and Skogby, 2000; Jurewicz and Watson, 1988; Peslier
et al., 2002). Ca partitioning between cpx and opx is highly temperature
dependent (e.g., Brey and Khöler, 1990). As discussed above (see
Section 5.1;Fig. 4b) REE fractionationsand enrichments are here linked
to the equilibrium temperature (see also Zangana et al., 1997)(Fig. 14)
and thus are related to the sample position relative to the metasomatic
melt source. Therefore the apparent correlation between water content
in opx and REE fractionation, reflect rather a temperature effect
(i.e., which could be related to metasomatism, see Section 5.1). It yields
to the incorporation of a higher amount of Ca, producing then a higher
abundance of impurities at atomic scale and associated hydrogen-
bearing point defects in opx.
As shown here, the relationships between the H
2
O content of NAMs
and metasomatism is not straightforward. Indeed, our observations
suggest that H partitioning between melt/fluid and NAMs is dependent
of the nature of the melt/fluid and in particular of its H
2
O and/or volatile
fugacity. In one hand we observe H
2
O enrichment in NAMs when these
latters interact with relatively large melt fraction (melt/rock ratio ≥1%)
and thus when the ƒH
2
O is relatively low, suggestingthat H behaves as a
mid-REE. On the other hand, the crystallisation of LREE-depleted am-
phiboles would suggest that H
2
O behave in the melts as an highly in-
compatible elements; its chromatographic front is further away from
the source than the LREE chromatographic front thus D(H
2
O) b
D(LREE), as suggested by Demouchy et al., 2015. Then, when peridotites
interact with such small volume melt or fluid, which are volatiles-rich
and have a high ƒH
2
O leading to the crystallisation of amphibole, then
NAMs are not enriched in H
2
O. Is it due to the crystallisation of amphi-
bole, which then sequesters all the water available? Or is it due to a
drastic change of partitioning between NAMs and volatile-rich fluids?
0.1 1 10 100 1000
(Th/Nb)PM in cpx
(i)
(h)
PM
(g)
4
3210
(Sm/Yb)PM in cpx
(f)
(e)
PM
(d)
8
6
4
2
0
COH in ol (ppm wt. H2O)
PM
(a)
250
200
150
100
50
COH in opx
(b)
500
400
300
200
100
COH in cpx
1 10 100
LaN in cpx
(c)
8
6
4
2
200
150
100
400
300
200
100
4 0.1
0
250
50
500
50
500
Fig. 13. Watercontents in minerals (ol, opx and cpx) as a function of trace element indexes of metasomatism. Water contents are given in ppm wt. H
2
O using the c alibration of Paterson
(1982).(a), (b) and (c) water content in mineralsas a function of La
N
in cpx. (d),(e) and (f) watercontents in mineralsas a function of(Sm/Yb)
PM
in cpx. (g),(h) and (i) are water contents
in minerals as a function of (Th/Nb)
PM
in cpx. PM and N: normalised to Primitive mantle andCI chondrite values, respectively,from McDonough and Sun (1995); symbolsare as in Fig. 3;
dashed lines denote primitive mantle values; error bar on the average values of water content per sample is 1SD.
Fig. 14. Water contents in opx (in ppm wt. H
2
O, based on calibration of Paterson, 1982
(P82); and Bell et al., 1995 (B95)) as a function of [Ca] a.p.f.u (atom per formula unit cal-
culated for 6 bonding oxygens) for Ray Pic spinel-peridotite xenoliths. Corresponding
temperature (in °C) was calculated using Brey and Köhler (1990) ‘T°
Ca-in-opx
’
geothermometer. Error bar on the average values of water content per sample is 1SD.
271C.M.M. Denis et al. / Lithos 236–237 (2015) 256–274

These complexities are probably due, at least in part, to the dual behav-
iour of H as major vs. trace elements and to its related different types of
incorporation into the mineral structure.
6. Conclusion
In this study, water content in NAMs can be broadly explained by de-
pletion during partial melting event(s) using relatively straightforward
melting models. The metasomatism, expressed by LREE and LILE
enrichment without concomitant HFSE enrichment, thus involving
small volume melt enriched in volatiles (also coined as carbonatitic meta-
somatism), has no obvious effect on the water content of NAMs. Hydro-
gen in NAMs is not correlated with canonical metasomatism indicator
such as La and Ce abundances of La/Sm La/Yb ratio, and hydrogen
seems to rather behave similarly to a MREE during melting and metaso-
matism. However some of our data suggest that water contents in
NAMs were then possibly modified afterward by melt–rock reactions at
relatively high melt/rock ratio (i.e., ≥1%). Our results for the Ray Pic xeno-
liths call for further studies, specially integrating data on trace elements
and water content of NAMs necessary to truly assess the behaviour of
water during melting and metasomatism.
Hydrous metasomatism (i.e., leading to the crystallisation of OH-
bearing amphibole) has no effect on water content of NAMs. Thus the
occurrence of amphibole does not warrant the water-saturation of co-
existing NAMs. Therefore, its presence should not be used as a justifica-
tion for the application of weak and wet flow law for olivine.
Acknowledgements
CNRS supported this study through INSU 2011, 2012 & 2013
programmes to S. Demouchy and O. Alard. The University of Montpellier
supported this study through a Ph.D fellowship “Bourse Président”to C.
Denis. We thank F. Parat for informal review and H. Behrens for KFT mea-
surements. We also thank C. Nevado and D. Delmas for providing impec-
cable thin sections. One anonymous reviewer and Prof. Xia are thanked
for detailed reviews, which have improved the quality of the manuscript.
Electron microprobe analyses were performed at ‘Microsconde SUD’and
carried out with the help of B. Boyer and C. Merlet. FTIR analyses were
carried out with the assistance of D. Maurin at the Laboratoire Charles
Coulomb, University of Montpellier, France.
References
Agrinier,P., Mével, C., Bosch, D.,Javoy, M., 1993. Metasomatic hydrous fluids in amphibole
peridotites from Zabargad Island (Red Sea). Earth and Planetary Science Letters 120,
187–205. http://dx.doi.org/10.1016/0012-821X(93)90239-6.
Alard, O., Dautria, J.-M., Bodinier, J.-L., 1996. Nature du manteau supérieur et processus
métasomatiques de part et d'autre du Sillon Houiller (Massif Central). Comptes
rendus de l'Académie des sciences. Série 2. Sciences de la terre et des planètes 323,
763–770.
Alard, O., Lorand, J.-P., Reisberg, L., Bodinier, J.-L., Dautria, J.-M., O'Reilly, S.Y., 2011.
Volatile-rich metasomatism in Montferrier xenoliths (Southern France): implications
for the abundances of chalcophile and highly siderophile elements in the subconti-
nental mantle. Journal of Petrology 52, 2009–2045. http://dx.doi.org/10.1093/
petrology/egr038.
Ardia, P., Hirschmann, M.M., Withers, A.C., Tenner, T.J., 2012. H
2
O storage capacity of oliv-
ine at 5–8GPaand consequences for dehydration partial melting ofthe upper mantle.
Earth and Planetary Science Letters 345–348, 104–116. http://dx.doi.org/10.1016/j.
epsl.2012.05.038.
Aubaud, C., Hauri, E.H., Hirschmann, M.M., 2004. Hydrogen partition coefficients between
nominally anhydrous minerals and basaltic melts. Geophysical Research Letters 31.
http://dx.doi.org/10.1029/2004GL021341.
Aubaud, C.,Hirschmann, M.M.,Withers, A.C., Hervig, R.L., 2008. Hydrogenpartitioning be-
tween melt, clinopyroxene, and garnet at 3 GPa in a hydrous MORB with 6 wt.% H2O.
Contributions to Mineralogy and Petrology 156, 607–625. http://dx.doi.org/10.1007/
s00410-008-0304-2.
Bai, Q., Kohlstedt, D.L., 1993. Effects of chemical environment on the solubility and incor-
poration mechanism for hydrogen in olivine. Physics and Chemistry of Minerals 19,
460–471. http://dx.doi.org/10.1007/BF00203186.
Baptiste,V., Tommasi, A., Vauchez, A., Demouchy, S., Rudnick, R.L.,2015. Deformation,hy-
dration, and anisotropy of the lithospheric mantle in an active rift: constraints from
mantle xenoliths from the North Tanzanian Divergence of the East African Rift.
Tectonophysics 639, 34–55. http://dx.doi.org/10.1016/j.tecto.2014.11.011.
Bedini, R.M., Bodinier, J.-L., Dautria, J.-M., Morten, L., 1997. Evolution of LILE-enriched
small melt fractions in the lithospheric mantle: a case study from the East African
Rift. Earth and Planetary Science Letters 153, 67–83. http://dx.doi.org/10.1016/
S0012-821X(97)00167-2.
Behrens, H., Romano, C., Nowak, M., Holtz, F., Dingwell, D.B., 1996. Near-infrared spectro-
scopic determination of water species in glasses of the system MAlSi 3O 8 (M = Li,
Na, K): an interlaboratory study. Chemical Geology 128, 41–63.
Bell, D.R., Rossman, G.R., 1992. Water in Earth's mantle: therole of nominally anhydrous
minerals. Science 255, 1