Eur. J. Mineral.

2011, 23, 5–16

Published online November 2010

Role of iron and reducing conditions on the stability of dolomite þ coesite

between 4.25 and 6 GPa – a potential mechanism for diamond formation during
subduction
AUDREY M. MARTIN1,2,3,4,* and TAHAR HAMMOUDA1,2,3

1
Clermont Université, Université Blaise Pascal, Laboratoire Magmas et Volcans, BP 10448, 63000 Clermont-
Ferrand, France
2
CNRS, UMR 6524, LMV, 63038 Clermont-Ferrand, France
3
IRD, R 163, LMV, 63038 Clermont-Ferrand, France
4
NASA Johnson Space Center, Mailcode KT, 2101 NASA Parkway, Houston, TX 77058, USA
*Corresponding author, e-mail: audrey.m.martin@nasa.gov

Abstract: We have investigated the effect of iron and oxygen fugacity on dolomite þ coesite stability during subduction. For redox
conditions buffered by the assemblage itself, the presence of iron (Fe/(Fe þ Mg) ca. 0.4) lowers the decarbonation reaction by about
200  C at 4.25 GPa and by 300  C at 5.5 GPa, compared to the iron-free reaction. Clinopyroxene and CO2 form by a decarbonation
reaction similar to the iron-free system. At low temperature, however, graphite replaces CO2 through redox interactions with iron in
the carbonate. Melting occurs approximately 100  C above decarbonation and a carbonatitic melt is produced. In a second series of
experiments, we imposed lower oxygen fugacity by adding molybdenum, in order to study the potential redox mechanisms in contact
with the peridotitic mantle. In these samples, we observe systematic carbon reduction producing the assemblage clinopyroxene þ
graphite. Our results show that the stability of dolomite þ coesite on a subduction path is limited by redox interactions, in addition to
pressure and temperature. As our experiments were run in the stability field of diamond, we also demonstrate that diamond may form
from dolomite þ coesite during subduction. Chemical diffusion of iron and oxygen in the slab and at the slab/mantle interface appears
to be a key parameter to determine at what pressure–temperature conditions this may happen.
Key-words: dolomite, coesite, subduction, high pressure, redox interactions, decarbonation, experimental petrology, diamond.

1. Introduction composition, but dolomite CaMg(CO3)2, ferroan dolomite

Analyses of mantle xenoliths show that carbon is present in
the Earth’s mantle in the form of oxidized or reduced
minerals (Ionov et al., 1993; Nixon, 1995; Viljoen, 1995;
Wang et al., 1996), fluids (Schrauder & Navon, 1993;
Klein-Ben David et al., 2007), or dissolved in silicate and
carbonate melts (Schiano et al., 1994; Yaxley et al., 1998;
Moine et al., 2004). Zhang & Zindler (1993) and Marty &
Tolstikhin (1998) infer from mass balance calculations that
the current concentration of carbon in the mantle (between
80 and 400 ppm) can only be explained by the subduction
of oceanic material, and in particular, oceanic carbonates.
Plank & Langmuir (1998) determined that 7 wt% of the
Global Subducting Sediment is of calcite CaCO3 composi-
tion. Magnesium and iron carbonate are also observed,
although more rare. Alteration of the oceanic crust by
hydrothermal activity also leads to the crystallization of
carbonates which represent a large potential carbon reser-
voir for high pressure recycling (Staudigel et al., 1989; Alt
& Teagle, 1999). These carbonates are mainly of calcite
(or ankerite) Ca(Mg,Fe)(CO3)2 and siderite FeCO3 are also
found (Laverne, 1993; Nakamura & Kato, 2004). In order
to constrain the destabilization conditions of carbonates
during subduction, low-pressure phase diagrams in the
system CaCO3 – MgCO3 – FeCO3 were determined by
experiments (Goldsmith et al., 1962) and thermodynamic
modelling (Anovitz & Essene, 1987; Davidson, 1994).
However, we still lack high-pressure data on ternary solid
solutions. Interactions with the surrounding material were
also considered. Luth (1993, 1995, 1999) claims that dec-
arbonation takes place at higher pressure in sediment or
altered basalt than in a peridotite. This would imply that a
depth interval exists, where the decarbonation of the super-
ficial layers of the slab can lead to the crystallization of
carbonate in the mantle wedge by CO2 transfer. Another
important parameter to consider is the oxidation state of the
system, because of the contrast between subducted mate-
rial and peridotitic mantle. Redox conditions of silicate
(peridotitic) lithologies at high pressure are imposed in
most cases by iron oxidation state (Ballhaus, 1993; Canil

0935-1221/10/0022-2067 $ 5.40
DOI: 10.1127/0935-1221/2010/0022-2067 # 2010 E. Schweizerbart’sche Verlagsbuchhandlung, D-70176 Stuttgart
6 A.M. Martin, T. Hammouda
& O’Neill, 1996), whereas oxygen fugacity (fO2) of carbo-
nate lithologies is likely to be controlled by carbon oxida-
tion state.
One of the decarbonation reactions investigated experi-
mentally by Luth (1995) for sediment or altered basalt was:
dolomite þ 2 coesite ¼ diopside þ 2 CO2
CaMgðCO3 Þ2 þ 2 SiO2 ¼ CaMgðSiO3 Þ2 þ 2 CO2
Luth found that the assemblage dolomite þ coesite is
stable to at least 6 GPa on a hot subduction path. In our
study, we first investigate the role of iron on this reaction
between 4.25 and 6 GPa over the temperature range 800 to
1425  C. Iron was chosen because (1) the presence of iron-
bearing dolomite (or ankerite) is reported in various places
(Laverne, 1993; Nakamura & Kato, 2004, see above) and
(2) iron is thought to control the fO2 in the mantle and
might diffuse into the subducted material. Iron (and mag-
nesium) might also diffuse into carbonates by exchange
reactions with subducted silicates occurring at high pres-
sure. For example, studies on carbonated eclogites by
Hammouda (2003) and Yaxley & Green (2004) show that
pure calcite is transformed to Fe-bearing dolomite by reac-
tions with garnet and clinopyroxene. Our experiments aim
at constraining iron/carbon interaction mechanisms.
Therefore, fO2 is buffered intrinsically by the samples
themselves, leading to relatively oxidizing conditions
(see Discussion). In a second series of experiments, we
study the role of reducing conditions on dolomite þ coesite
stability by adding molybdenum. Oxygen fugacity of the
Mo/MoO2 (MMO) buffer, studied by O’Neill (1986) at 1
bar, lies around IWþ0.5/þ1 and FMQ3.5/4.5 at 5 GPa
(Fig. 1). This value is in the range of oxidation conditions
determined by previous authors for peridotitic mantle
(Ballhaus, 1993; McCammon et al., 2001). In particular,
Frost & McCammon (2008) found that fO2 decreases when
pressure increases – FMQ3 to FMQ4.5 between 4 and 6
Fig. 1. Temperature – log (fO2) diagram comparing the position of
Mo–MoO2 buffer to Fe–FeO (IW), Fayalite–Magnetite–Quartz
(FMQ) and equilibria involving carbon species at 5 GPa.
Calculations are described in the text.
GPa – in the cratonic mantle. Redox conditions in the
mantle wedge might be a little higher – around or above
FMQ – due to metasomatism (Parkinson & Arculus, 1999;
Wang et al., 2007). However, percolation of reduced fluids
coming from the slab might also lower the fO2 in the
overlying mantle (Frost & McCammon, 2008).
(1) Considering both pressure effect and chemical uncertain-
ties at subduction zones, these last experiments aim at
constraining potential redox mechanisms in subducted
dolomite þ coesite in contact with the mantle wedge and,
in particular, diamond formation from the assemblage
dolomite þ coesite.
2. Experiments
2.1. Starting material and fO2 buffering
The starting material is a mixture of natural high-purity
quartz from Saint-Paul-La-Roche (France) and a dolomite
containing iron (Fe0.36Mg0.66Ca0.98(CO3)2) from Ben
Gasseur (Tunisia), which were crushed and dried at 140

C for 10 h. Compositions of the starting mixtures are
given in Tables 1 and 2. In the first series of experiments,
oxygen fugacity is, in theory, imposed intrinsically by the
oxidation state of carbon whose concentration in the mix-
ture (expressed in moles) is five times greater than that of
iron. In practice, oxidation conditions were established
using reactions occurring in our samples (see
Discussion). AuPd capsules welded at the two extremities
were used in order to limit oxygen and hydrogen exchange
with the surrounding materials. As our experiments were
anhydrous, a double capsule technique using an external
buffer and H was excluded. For reducing experiments,
samples were loaded into a rolled and folded Mo foil that
was introduced into a platinum capsule welded at the two
extremities. Computation of MMO equilibrium shows that
it lies approximately 3.5–4 log unit below the fayalite-
magnetite-quartz buffer (FMQ), 0.5–1 log unit above the
iron-wüstite (IW) buffer, and about 2–3 log units below
diamond/CO2 (CCO from Frost 1998) or diamond/carbo-
nate buffers (DCDD, dolomite þ 2 coesite ¼ diopside þ 2
diamond þ 2 O2, relevant to eclogite or sediment after
Luth (1993), and EMOD, enstatite þ magnesite ¼ forster-
ite þ diamond þ O2, relevant to peridotite after Eggler &
Baker (1982)) at 5 GPa (Fig. 1). Thermodynamic calcula-
tions were made using the relations:
dG
f O2 ¼  (2)
RT
dG ¼ dH  TdS þ ðP  1Þ dV solid (3)
where G represents the Gibbs free energy, R the gas con-
stant (¼ 8.3144 J K1 mol1), T the temperature, H the
enthalpy, S the entropy, P the pressure and V the volume.
Data used for calculations were taken from Holland &
Powell (1998), Robie & Hemingway (1995) or the NIST
 Role of iron and low fO2 on dolomite þ coesite stability 7

Table 1. Experimental conditions of the first series of experiments (intrinsic fO2).

Serie 1 Run# P (GPa) T ( C) Duration Wt% Dolomite Wt% Quartz Assemblage

ME625 4.25 800 6h 74.8 25.2 dol þ cs þ graph
ME710 4.25 950 94 h 75.0 25.0 dol þ cs
ME831 4.25 1100 18 h 74.5 25.5 dol þ cs þ cpx
ME380 4.25 1250 6h 61.7 38.3 dol þ cs þ cpx þ melt
ME614 5.5 800 6 h 30 min 74.7 25.3 dol þ arag þ mc þ cs þ graph
ME729 5.5 1000 2 h 15 min 75.0 25.0 dol þ cs
ME384 5.5 1300 6h 61.7 38.3 dol þ cs þ melt
ME382 5.5 1350 6h 64.1 35.9 dol þ cs þ melt
ME398 5.5 1425 6h 71.5 28.5 cs þ cpx þ melt
ME619 6 800 6 h 30 min 73.1 26.9 arag þ Fe-mc þ cs þ graph
ME592 6 900 6h 72.9 27.1 dol þ arag þ Fe-mc þ cs þ graph
ME645 6 1000 6h 75.2 24.8 dol þ cs þ graph
ME708 6 1100 34 h 75.0 25.0 dol þ cs
ME832 6 1200 18 h 74.5 25.5 dol þ cs þ cpx

Abbreviations: cpx ¼ clinopyroxene, cs ¼ coesite, dol ¼ dolomite, graph ¼ graphite, mc ¼ magnesite, arag ¼ aragonite. The presence of
CO2 is not indicated here.

Table 2. Experimental conditions of the second series of experiments (reducing conditions).

Serie 2 Run# P (GPa) T ( C) Duration Wt% Dolomite Wt% Quartz Assemblage

ME504 4.25 950 6h 74.6 25.4 dol þ cs þ cpx þ graph þ Fe-Mg oxide
ME470 4.25 1050 6h 72.9 27.1 dol þ cs þ cpx þ melt þ graph
ME464 4.25 1150 6h 63.4 36.6 arag þ cs þ cpx þ melt þ graph
ME483 5 1150 6h 74.8 25.2 dol þ cs þ cpx þ melt
ME474 5.5 1150 6h 75.1 24.9 dol þ cs þ cpx þ melt þ graph
ME517 5.5 1250 6h 75.1 24.9 dol þ cs þ cpx þ melt
ME469 5.5 1300 6h 74.8 25.2 cpx þ melt þ graph
ME826 6 900 48 h 74.5 25.5 dol þ cs þ cpxþ graph
ME505 6 1050 6h 74.4 25.6 dol þ cs þ cpx þ graph
ME484 6 1100 6h 73.9 26.1 dol þ cs þ cpx
ME471 6 1150 9 min 74.3 25.7 dol þ cs þ cpx þ graph
ME485 6 1200 6h 74.3 25.7 dol þ cs þ cpx þ melt
ME468 6 1250 6 h 30 min 73.5 26.5 cpx þ cs þ melt þ graph
ME512 6 1300 6h 74.5 25.5 cpx þ cs þ melt

Abbreviations: cpx ¼ clinopyroxene, cs ¼ coesite, dol ¼ dolomite, graph ¼ graphite, arag ¼ aragonite.

website. The only molybdenum oxide observed in the
samples was MoO2 and Mo was still present at the end of
all experiments.
2.2. Experimental protocol
Experiments were conducted on a 1000 ton multi-anvil
apparatus (octahedral symmetry) at Laboratoire Magmas
et Volcans (Hammouda, 2003) using an 18/11 type assem-
bly (octahedron edge length in mm/truncation length of the
anvils in mm). Experimental conditions are reported in
Tables 1 and 2. In the first series of experiments (buffered
intrinsically), the starting mixture was loaded in a 2.0 mm
diameter Au80Pd20 (by weight) capsule. This capsule was
dried at 1000  C before loading. It was then welded at the
two extremities in order to prevent CO2 or melt loss. In the
second series of experiments, the samples were contained
in Mo internal capsules introduced in outer platinum
2.0 mm diameter capsules (see above). Before loading,
platinum capsules were dried using the same method as
for AuPd capsules and Mo foils were dried at 140  C for 1
h. Stepped LaCrO3 furnaces with Mo electrodes were used
for the two series of experiments. The furnace was sepa-
rated from the capsule by an MgO spacer. A W95Re5-
W74Re26 thermocouple was inserted axially such that the
junction was in contact with one end of the capsule. Cr2O3-
doped MgO octahedra were used in our experiments.
Zirconia was placed between the furnace and the octahe-
dron for thermal insulation. Pressure calibration was made
using the coesite/stishovite (Akaogi, 1995) and fayalite/
spinel (Katsura & Ito, 1989) transitions at 1000  C.
Pressure is believed to be accurate within 0.5 GPa
and temperature uncertainty is about 20  C. The
temperature gradient inside the capsules is not known;
however, considering the short length of every capsule
(2.5–3.0 mm), we believe that this gradient is very small.
Pressure was first increased in about 2–3 h (depending on
8 A.M. Martin, T. Hammouda
pressure). Then the samples were heated in less than 20 min
by increasing the electrical power in the LaCrO3 furnace.
Temperature was maintained at constant value by an auto-
matic regulation (Eurotherm controller). Running duration
for most experiments was 6 h, based on Luth (1995). Some
longer-duration experiments were performed to test kinetic
effects, in particular at low temperature (Tables 1 and 2).
Temperature was quenched by shutting off the electrical
power (drop to ,100  C temperatures in 4–8 s).
Decompression was programmed for durations of 3–7 h.
2.3. Sample preparation and analyses
Recovered samples were mounted in resin and polished
using alumina powder and ethanol. Bright veins or pockets
of graphite were identified by optical reflected-light micro-
scopy. Raman spectroscopy (Jobin-Yvon T64000, 514.5
nm) was used to study the carbon species (graphite or
diamond). The volume of sample excited by the laser is
about 1 mm3. The laser power on the sample was main-
tained at a low value (,15 mW) during analyses to mini-
mize sample alteration (which was controlled by optical
microscopy). Phase relations were determined by scanning
electron microscopy using energy-dispersive X-ray spec-
trometry (Jeol 5910-LV). Chemical analyses were
obtained with an electron microprobe (Cameca SX100
using (rz) data reduction procedure) using wavelength-
dispersive spectrometry (accelerating voltage of 15 kV;
beam current of 15 nA for silicate minerals and 8 nA for
carbonate minerals or melts with a 5 mm defocused beam).
No damage was observed at the carbonate surface after
analyses. The standards used for the microprobe calibra-
tions are wollastonite for Ca and Si, Al2O3 for Al, forsterite
for Mg, fayalite for Fe, albite for Na and molybdenum for
Mo. Counting time was 10 s on the peak for all elements
except for Mo (30 s) and Fe (40 s). Because of complex
quenching textures, melt composition was determined,
when possible, either with a large defocused beam or
with a scanning beam over a large area.
3. Results
3.1. Phase relations
Phase assemblages observed in the first and second series
of experiments are reported in Tables 1 and 2, respectively.
Typical sample textures are shown in Fig. 2 and 3 and
phase diagrams of the two series of experiments are
reported in Fig. 4.
In the first series of experiments, the dolomite þ coesite
decarbonation reaction is bracketed between 950 and 1100  C
at 4.25 GPa, and between 1100 and 1200  C at 6 GPa
(Table 1; Fig. 2a). In the dolomite þ coesite stability field,
graphite is also found at 4.25 GPa – 800  C and at 6
GPa – 1000  C. At 5.5 GPa – 800  C and 6 GPa below 900  C,
dolomite breaks down to aragonite plus magnesite
(Hammouda et al., 2010) with a eutectoid texture
(Fig. 2b) similar to that observed by Luth (2001).
Graphite is present in all these samples. At 4.25 GPa –
1100  C and 6 GPa – 1200  C, clinopyroxene forms at
grain boundaries between dolomite and coesite (Fig. 2c).
Melting occurs below 1250  C at 4.25 GPa, producing a
carbonate-silicate melt (Fig. 2d) in association with
dolomite, coesite and clinopyroxene. At 5.5 GPa, carbo-
nate-silicate melt appears at 1300  C, in association with
coesite and dolomite up to 1350  C, and with coesite
and clinopyroxene at 1425  C. The presence of a gas
phase was observed in some samples, either in the form
of bubbles going out from the capsule during polishing,
or of rounded holes in some sample sections. Because
we could not perform systematic observations of fluid
occurrence in our sample, we did not report them in the
Tables. The presence of clinopyroxene is sufficient evi-
dence for the decarbonation of the starting material.
In the second series of experiments, graphite is found in
almost all experimental samples (plucking during grinding
and polishing is believed to be the cause of the absence of
graphite in some samples). Clinopyroxene is also observed
below 1000  C at 4.25 GPa and below 1200  C at 6 GPa
(Table 2; Fig. 3a). Above these temperatures, carbonate –
silicate melt forms in association with coesite, dolomite,
clinopyroxene and graphite in various proportions
(Fig. 3b–d).
To determine if the graphite found in our samples formed
during the experiments or during the quench (at the end of
the experiments), we investigated its structure by using
Raman spectroscopy (Fig. 5). We focused on the first-
order region of the Raman spectra of carbon, which
shows several bands corresponding to different vibration
modes of the carbon bonds (Beyssac et al., 2003). We looked
at three bands in particular: the G band (at 1580 cm1)
which is due to the stretching vibration in aromatic layers
of graphite (E2g2 mode), the D1 band (at 1350 cm1)
called the defect band (A1g mode) and the D2 band (at
1620 cm1) which is always present when D1 is present.
The last two bands characterize poorly ordered graphite
or carbonaceous material (CM). In our first series of
experiments, Raman spectra of graphite, which was only
found in the low-temperature experiments, show that it
is disordered. The intensity of D1 (at 1351–1360 cm1)
is slightly larger than that of G (at 1592–1600 cm1).
D2 appears at approximately 1620 cm1, but cannot be
resolved from G and causes an offset of G to higher
frequencies. In Series 2 (reducing experiments), where
graphite was found in almost all samples, ordering of
graphite is a function of the presence or absence of melt.
At subsolidus conditions, D1 is about the same height as
G, implying disorder. Above the solidus, G (at
1354–1356 cm1) is about three times higher than D1
(at 1578–1581 cm1), i.e. the graphite is quite ordered
and it appears that the presence of melt favours graphite
ordering. From this strong link between ordering and
melting, we conclude that graphite formed at high tem-
perature (during the experiments) and is not a quench
phase – otherwise, the D1/G value would be much more
 Role of iron and low fO2 on dolomite þ coesite stability 9

Fig. 2. SEM images (BSE) of typical experimental textures observed in samples of the first series of experiments. (a) Dolomite (light grey) and
coesite (dark grey) in a non-decarbonated sample. (b) Fine-grained eutectoid texture formed by aragonite (lighter) and magnesite (darker). Black
phase is graphite. Dark grey phase is coesite. White phase is a small piece of the capsule which moved during polishing. (c) Clinopyroxene (white
phase) formed at grain boundaries between dolomite (light grey) and coesite (dark grey) in a decarbonated sample. (d) Carbonate-silicate melt
with dendritic clinopyroxene exsolved during quench. Dark grey phase is coesite and light grey phase is clinopyroxene.

scattered, as it would depend on quench speed and phase
relations. Moreover, a graphite precipitation from CO2
fluid during quenching (Wopenka & Pasteris, 1993;
Cesare, 1995) would not explain why only low-tempera-
ture, non-decarbonated samples contain graphite in
Series 1. In addition, Luque et al. (1998) also showed
that natural (low temperature) fluid-deposited graphite
(except that contained in inclusions) exhibits a high
degree of order, whereas the graphite formed in our
low-temperature experiments is disordered. Considering
all this, we can definitely rule out graphite precipitation
from a fluid during quench. Rather, graphite formed
during the experiments by a redox reaction between Fe
and C inside carbonates in Series 1 (see Discussion).
3.2. Phase compositions
The composition of the low-temperature carbonates is
given in Table S1, freely available online as
Supplementary Material on the GeoScienceWorld website
of the journal, linked to the relevant article at http://eur-
jmin.geoscienceworld.org/. Subsolidus decarbonation of
the samples produces clinopyroxene (cpx), the
composition of which varies depending on redox condi-
tions (Fig. 6 and supplementary material S2 and S3). In the
first series of experiments, Mg# (defined by Mg/(Mg þ
Fe)) of cpx is 0.60 while Mg# of carbonate is 0.67. In
Series 2 (reducing conditions), Mg# of cpx and carbonate
are equivalent but with a larger variability, between 0.60
and 0.72. A more remarkable feature is the difference
between XCa values of cpx (defined by Ca/(Ca þ Mg þ
Fe) in Series 1 and 2. While XCa in carbonate remains
constant (0.50), XCa in cpx is 0.35 in Series 1, but 0.43
to 0.49 in Series 2. Therefore, cpx from the first series of
experiments is less calcic than those formed at more redu-
cing conditions. Moreover, the Ca content of cpx from
reduced experiments is close to that of carbonate, as
expected from mass-balance calculations. As no Ca-rich
carbonate or calcite was found in the samples, the low Ca
content of cpx in Series 1 requires the presence of an
additional phase that takes up calcium relative to magne-
sium. The only possibility is a CO2 fluid. Although cal-
cium solubility in CO2 is not well documented from
experiments at high pressure, natural examples of Ca-
bearing high density fluids (HDFs) in diamond inclusions
have been reported (Klein-Ben David et al., 2007). This
kind of fluid is thought to be stable in the diamond
10 A.M. Martin, T. Hammouda

Fig. 3. SEM images (BSE) of typical experimental textures observed in samples of the second series of experiments (reducing conditions).
(a) Decarbonated sample with graphite veins (black) and clinopyroxene (white) which formed at grain boundaries between dolomite (light grey)
and coesite (dark grey). (b) Clinopyroxene rims (white) around coesite (dark grey) in a melted sample (melt is not shown on this picture).
Dolomite is light grey. (c) Carbonate-silicate melt (grey) with dendritic clinopyroxene exsolved during quench (white stripes). Light grey
euhedral crystals are clinopyroxene. Coesite is dark grey and dolomite is grey (at the top of the image). (d) SEI image of molybdenum oxide
(medium grey) embedded in the carbonate-silicate melt (dark grey with white stripes). Molybdenum capsule is at the bottom (light grey).

stability field, i.e. 4 GPa and higher. Additionally, Luth 4. Discussion

(2006) reported significant Mg and Ca solubility in CO2
fluid in equilibrium with diopside at pressures between 3
and 5 GPa, with preferential uptake of Ca, as in the
present experiments. Unfortunately, polishing difficul-
ties did not allow us to determine precisely carbonate,
cpx and CO2 fractions in the samples, thus we cannot
estimate how much Ca dissolved in the vapour in our
experiments.
Melting produces carbonate liquids that have a small
fraction of dissolved silicate (,15 wt% SiO2;
Supplementary material S4). At the P–T conditions of
our study, no immiscibility was observed between the
carbonate and silicate fractions of the melts. In the first
series of experiments, melts have a dolomitic composition
with an XCa value of 0.5. The Mg# value of the solid
phases formed in equilibrium with the melts is 0.60–0.80
for cpx and 0.60–0.72 for carbonates. In clinopyroxene,
XCa is between 0.35 and 0.50, whereas in carbonates XCa
is between 0.50 and 0.70. In the second series of experi-
ments, some molybdenum was found in the melt (up to 5
wt% MoO3). Thus, the use of the Mo/MoO2 buffer is not
appropriate for high-temperature – supersolidus – experi-
ments in the presently studied chemical system.
4.1. Effect of iron on the stability of dolomite þ coesite
4.1.1. Phase relations
In the first series of experiments (self-buffered) the
Fe-bearing dolomite starting material breaks down to mag-
nesite þ aragonite at about the same conditions as those
found by Luth (2001) and Buob et al. (2006) for the Mg
end-member (Fig. 4a). Considering the experimental
brackets, the dolomite breakdown boundary in this study is
also consistent with the results of Morlidge et al. (2006) on
Fe end-member (ankerite) stability. According to Luth
(1995), the decarbonation reaction dolomite þ 2 coesite
¼ diopside þ 2 CO2 occurs at 1500  C at 6 GPa and
1200  C at 4 GPa. In the iron-bearing system, the
high-temperature assemblage is favoured and the decarbona-
tion reaction is shifted to lower temperature by at least 200  C
at 4.25 GPa and 300  C at 5.5 GPa (Fig. 4a). Melting
occurs between 1100 and 1250  C at 4.25 GPa and between
1200 and 1400  C at 6 GPa. Compared to the iron-free system
studied by Luth (2006), we note a solidus depression of at
least 325  C at 4 GPa. In our experiments, the solidus P–T
 Role of iron and low fO2 on dolomite þ coesite stability
Fig. 4. Pressure–temperature diagrams of (a) the first series of
experiments (intrinsic fO2) and (b) the second series of experi-
ments (reducing conditions). The precision on the pressure value is
less than 0.5 GPa; the precision on temperature is 20  C.
Black points indicate the experimental conditions; diamond sym-
bols represent samples with graphite. Black curves are reactions
deduced from our samples. Dashed grey curve represents the Fe-
free decarbonation reaction dolomite þ 2 coesite ¼ diopside þ 2
CO2 and the solidus calibrated by Luth (2006). Dol ¼ dolomite,
cs ¼ coesite, cpx ¼ clinopyroxene, arag ¼ aragonite, mc ¼
magnesite, C ¼ graphite. Solidus in Series 1 is indicated for
information; however, our experiments do not allow one to deter-
mine if its slope is positive or if it is vertical.
slope appears to be either positive or vertical whereas Luth
(2006) found a negative slope between 3 and 8 GPa in the
iron-free system. This is probably an effect of the presence of
iron in our system. From all the above discussion, we can
deduce that the sole addition of iron to dolomite lowers the
dolomite þ coesite stability regarding both decarbonation
and melting.
4.1.2. Oxidation conditions
In these experiments, fO2 is assumed to be buffered by the
samples themselves. Carbonate reduction to graphite was
observed at low temperature. This can only be explained
by a redox reaction with the iron contained in the starting
dolomite, which could be written in a simplified way:
dolomite1 ¼ dolomite2 þ iron oxide þ graphite (4)
11
Fig. 5. Raman spectra of graphite in the experimental samples (S1 ¼
Series 1, S2 ¼ Series 2).
with dolomite2 containing slightly less iron than dolo-
mite1. A few iron oxide grains were found in the gra-
phite-bearing samples, but they were too small to obtain
good microprobe analyses for iron valence state deter-
mination. The loss of iron from the dolomite may be
either compensated by slight Ca and Mg content varia-
tions in the dolomite structure, or by the formation of
new phases (carbonates or oxides). We could not, how-
ever, observe any variations or new phases, probably
because of the very small amounts of iron oxide and
graphite produced.
In order to quantify the oxygen fugacity in the samples,
we modelled the stability of siderite with regard to
magnetite and graphite over a P–T range corresponding to
our experiments. Siderite was chosen as iron-carbonate end-
member because thermodynamic values for ankerite are
poorly known. The results are illustrated on a polybaric
temperature–fO2 section in Fig. 7. Using the redox reaction:
6 siderite ¼ 2 magnetite þ 6 C þ 5 O2 (5)
6 FeCO3 ¼ 2 Fe3 O4 þ 6 C þ 5 O2
we estimate a value of log(fO2) 7.5 at 6 GPa – 1000  C
in our samples. Below this temperature (carbonate and
graphite also present), log(fO2) decreases with temperature
along reaction (5). Above this temperature (graphite
absent), log(fO2) is above reaction (5). At 5 GPa – 900  C,
log(fO2) 9.4 (FMQ–0.3). Below this temperature,
12 A.M. Martin, T. Hammouda

Fig. 6. Ternary projection (Ca–Mg–Fe) of coexisting carbonate and clinopyroxene compositions at subsolidus conditions: dol ¼ dolomite, cpx
¼ clinopyroxene. Series 1 ¼ experiments with intrinsic fO2; Series 2 ¼ experiments at reducing conditions. Note clustering within each
mineral type, although pressure and temperature conditions vary.

log(fO2) decreases with temperature along reaction (5). dolomite þ 2 coesite þ 2 Mo

Above this temperature, log(fO2) is above reaction (5).
These values are about 0.2 log unit fO2 below the CCO
buffer at the same P–T conditions (Fig. 7).
At higher temperature, decarbonation of the sample pro-
duces cpx and CO2. Considering the intersection between
CCO and the decarbonation reaction dolomite þ 2 coesite ¼
cpx þ 2 CO2 (DCDV) that we calibrated experimentally
around 1100  C in the iron-bearing system, we deduce that
the fO2 increases to above 7.5 log units fO2 (FMQ–0.7) at
5 GPa when the decarbonation occurs (all phases of DCDV
present and graphite absent). No constraints can be given on
the fO2 value between the graphite stability boundary and
the decarbonation temperature, nor above the decarbonation
temperature. However, a temperature dependence of the
oxidation conditions in the samples (log(fO2)9.4 at
900  C to above 7.5 at 1100  C in the 5 GPa experiments)
is demonstrated.
4.2. Effect of reducing conditions on dolomite þ coesite
stability
All the samples synthesized at reducing conditions (Series
2) contain clinopyroxene. Therefore, reducing conditions
imposed by Mo dramatically contract the dolomite þ coe-
site stability field – by more than 300  C compared to
oxidizing conditions. Reaction mechanisms can be divided
into two parts: a decarbonation reaction similar to that
occurring in Series 1 (reaction (1)), and a redox reaction:
Mo þ CO2 ¼ MoO2 þ C
in agreement with our thermodynamic calculations
(Fig. 1). The global reaction can be written:
¼ cpx þ 2 C þ 2 MoO2 (7)
CaMgðCO3 Þ2 þ 2 SiO2 þ 2 Mo
¼ CaMgðSiO3 Þ2 þ 2 C þ 2 MoO2
CO2 absence is deduced from textural evidence and clin-
opyroxene composition (see Results), and can be explained
by a simultaneous occurrence of reactions (1) and (6). By
lowering the CO2 activity, Mo lowers the temperature of
reaction (1). Oxidation conditions in our samples are buf-
fered by reaction (7), which means that, at 5 GPa, the fO2
lies somewhere between FMQ–0.5 and FMQ–4.5 at 1000  C
and between FMQ–1 and FMQ–3.5 at 1300  C (Fig. 1).
Although the oxidation conditions could not be deter-
mined more precisely, we demonstrate that dolomite þ
coesite stability during subduction may be lower than
previously thought due to redox interactions, and that
diamond or graphite may form from the subduction of
this assemblage. In the natural case, however, its stabi-
lity should be dependent on the FeO/Fe2O3 equilibria in
the surrounding silicates.
4.3. Reactions kinetics and equilibrium achievement
4.3.1. Self-buffered experiments
In our first series of experiments, the starting mixture is
never completely decarbonated. According to Lüttge &
(6) Metz (1991, 1993), kinetics of the reaction dolomite þ 2
quartz ¼ diopside þ 2 CO2 is very sluggish as soon as
clinopyroxene grains (which grow at the boundary
between dolomite and quartz) separate the two starting
 Role of iron and low fO2 on dolomite þ coesite stability
Fig. 7. Temperature – log(fO2) diagram illustrating the location of
siderite (sid) breakdown to magnetite (mt) þ graphite (C) (black
curves). Siderite breakdown equilibrium was modelled using
Thermocalc software (Holland & Powell, 1998) version 3.26. At
each pressure investigated, the observed stability limit of graphite
was reported in order to delimit (black dashed curve) the upper
stability of graphite. Graphite – CO2 equilibrium (grey curves) was
positioned using the experimental data of Frost (1998). Note that the
siderite breakdown reaction should form CO2 instead of C þ O2 after
it intersects CCO (thus, the higher temperature part of the carbonate
breakdown curve is metastable).
minerals, in spite of the presence of a CO2 fluid. Our
samples exhibit identical clinopyroxene compositions
within each sample; thus, we consider that a temporary
chemical equilibrium is attained. The determination of the
precise reaction temperatures would require either very
long duration experiments (months, years?), or reversal
experiments (which, in our case, were limited by the melt-
ing temperature of the capsule metal). Therefore, our dec-
arbonation temperatures are maximum temperatures, and
the decrease of the decarbonation temperature caused by
iron represents a minimum. This implies that the stability of
dolomite þ coesite during subduction is much lower than
previously determined (Luth, 1995) if the presence of iron
is considered. Kinetics of the low-temperature graphite
formation reaction are unknown but might also be slow
enough that a temporary chemical equilibrium was
reached. Therefore, the fO2 calculated for these experi-
ments represent maximum values.
4.3.2. Reducing experiments
In Series 2 (reducing conditions), molybdenum oxides
formed between the Mo0 capsule and samples in all experi-
ments that used the reducing buffer. Oxide stoichiometry
deduced from microanalyses indicates a MoO2 (IV) com-
position. No other molybdenum oxide formed. No iron was
detected either in the Mo oxide, or in the remaining Mo
capsule. Graphite is observed in most samples, but carbo-
nate is also still present. Thermodynamic calculations
(Fig. 1) show that the reaction Mo þ O2 ¼ MoO2
(MMO) lies 2–4 log units fO2 below the reaction CO2 ¼
C þ O2 (CCO) at the P–T conditions of our experiments;
13
Fig. 8. Pressure – temperature diagram showing the destabilization
conditions of the dolomite þ coesite assemblage at intrinsic fO2
(Series 1) and reducing conditions (Series 2). (1) Graphite or dia-
mond is produced by the reduction of the subducted dolomite in
contact with the mantle. (2) ‘Hot’ subduction path crosses the dolo-
mite decarbonation reaction at intrinsic conditions, i.e. far from the
mantle. In the shaded range, graphite or diamond forms by a redox
reaction between Fe and C in the subducted dolomite.
Graphite–diamond transition is from Sung (2000).
therefore, the interaction between Mo and C equilibria was
probably not complete. The remaining dolomite and coe-
site were chemically ‘‘isolated’’ by a mechanism similar to
that observed in self-buffered experiments – i.e. growth of
cpx at grain boundaries – as shown by the large number of
dolomite and coesite inclusions in cpx (Fig. 3b). Therefore,
we expect that reaction (7) occurs at much lower tempera-
ture than the lower experimental temperatures we tested.
This strengthens our conclusion that the assemblage dolo-
mite þ coesite may be destabilized at very shallow depth
during subduction due to interactions with the mantle.
5. Implications for dolomite þ coesite high-
pressure stability and a potential way of
diamond formation during subduction
Our results show that the destabilization of dolomite þ
coesite is favoured by the presence of iron and by reducing
conditions. If this assemblage is present in the lower part of
the subducted crust, i.e. far enough from the mantle wedge
to prevent redox interactions, and if there is no interaction
with other minerals or fluids (which means, at an intrinsic
fO2), CO2 is produced from less than 6.5 GPa (195 km) to
7.5 GPa (225 km) and melt above 7.5 GPa (Fig. 8). At
low pressure and temperature, graphite forms by the break-
down of the carbonate ferrous end-member up to 4.5
GPa. In contact with the peridotitic mantle wedge, slab
delamination and chemical diffusion probably produce
more reducing conditions for the subducted material. In
this case, dolomite is reduced to graphite above 3.5 GPa
(90 km). Diamond is absent in our samples, although
most of the experiments were performed in the diamond
stability field. This feature is common in experiments
performed at moderate pressure (Gunn & Luth, 2006)
and is related to the high energy required for diamond
14 A.M. Martin, T. Hammouda

Fig. 9. Sketch view of the subducting slab illustrating the possibilities for dolomite þ coesite reaction. Case (1) represents material remaining
in the slab, whereas case (2) corresponds to material assimilated to the mantle (delamination). Diamond symbols represent graphite or
diamond formation (a) by reduction at the mantle interface and (b) by a Fe/C redox reaction inside the dolomite. Grey areas represent portions
of the slab that would not be affected by the reducing conditions of the mantle, i.e. far enough to prevent oxygen diffusion. Details of the
various reactions are given in the text.

nucleation. In nature, one can anticipate the role of seeds,
nuclei or high strain rates at slab/mantle interface (King &
Bebout, 2006) to promote diamond nucleation. Therefore,
our results illustrate a potential mechanism for diamond
formation by the reduction of dolomite þ coesite inside the
subducted plate, far from or in contact with the mantle
(Fig. 9). Chemical diffusion – in particular, of iron and
oxygen – at slab/mantle interface and in the slab appears to
be a key parameter.
Acknowledgements: D. Laporte and D. Andrault are sin-
cerely acknowledged for their comments. The authors are
also grateful to K. Righter for English corrections and
interesting comments. Dolomite starting material was
kindly provided by B. Devouard. Assistance by J.-L.
Devidal on electron microprobe, by J.-M. Hénot and F.
Faure on scanning electron microscope, by J.-M. Nédélec
on Raman microprobe and by N. Bolfan-Casanova, F.
Pointud and J.-L. Fruquière on multi-anvil apparatus is
gratefully acknowledged. We also thank R.W. Luth and
G.M. Yaxley for their constructive reviews, and B. Fritz,
C. Shaw and C. Chopin for their editorial work. Financial
support from CNRS-INSU (DyETI program) is also
acknowledged. The multi-anvil apparatus of Laboratoire
Magmas et Volcans is financially supported by the Centre
National de la Recherche Scientifique (Instrument
National de l’INSU).
References
Akaogi, M. (1995): Thermodynamic properties of a-quartz, coesite,
and stishovite and equilibrium phase relations at high pressure
and high temperatures. J. Geophys. Res., 100, 22337–22347.
Alt, J.C. & Teagle, D.A.H. (1999): The uptake of carbon during altera-
tion of ocean crust. Geochim. Cosmochim. Acta, 63, 1527–1535.
Anovitz, L.M. & Essene, E.J. (1987): Phase equilibria in the system
CaCO3 – MgCO3 – FeCO3. J. Petrol., 28, 389–414.
Ballhaus, C. (1993): Redox states of lithospheric and asthenospheric
upper mantle. Contrib. Mineral. Petrol., 114, 331–348.
Beyssac, O., Goffé, B., Petitet, J.-P., Froigneux, E., Moreau, M.,
Rouzaud, J.-N. (2003): On the characterization of disordered and
heterogeneous carbonaceous materials by Raman spectroscopy.
Spectrochim. Acta Pt. A, 59, 2267–2276.
Buob, A., Luth, R.W., Schmidt, M.W., Ulmer, P. (2006):
Experiments on CaCO3-MgCO3 solid solutions at high pressure
and temperature. Am. Mineral., 91, 435–440.
Canil, D. & O’Neill, H.St.C. (1996): Distribution of ferric iron in
some upper-mantle assemblages. J. Petrol., 37, 609–635.
Cesare, B. (1995): Graphite precipitation in C-O-H fluid inclu-
sions: closed system compositional and density changes, and
thermobarometric implications. Contrib. Mineral. Petrol.,
122, 25–33.
Davidson, P.M. (1994): Ternary iron, magnesium, calcium carbo-
nates: A thermodynamic model for dolomite as an ordered deri-
vative of calcite-structure solutions. Am. Mineral., 79, 332–339.
Eggler, D.H. & Baker, D.R. (1982): Reduced volatiles in the system
C-O-H: implications to mantle melting, fluid formation, and
diamond genesis. in ‘‘High-Pressure Research in Geophysics’’,
S. Akimoto & M.H. Manghnani, eds. Advances in Earth and
Planetary Sciences, 12, Tokyo, 237–250.
Frost, D.J. (1998): Erratum to Frost & Wood (1997) ‘‘Experimental
measurements of the fugacity of CO2 and graphite/diamond
stability from 35 to 77 kbar at 925 to 1650 C’’. Geochim.
Cosmochim. Acta, 62, 725.
Frost, D.J. & McCammon, C.A. (2008): The redox state of Earth’s
mantle. Annu. Rev. Earth Planet. Sci., 36, 389–420.
Frost, D.J. & Wood, B.J. (1997): Experimental measurements of the
fugacity of CO2 and graphite/diamond stability from 35 to 77 kbar
at 925 to 1650 C. Geochim. Cosmochim. Acta, 61, 1565–1574.
Goldsmith, J.R., Witters, J., Northrup, D.A. (1962): Studies in the
system CaCO3-MgCO3-FeCO3: A method for major-element
 Role of iron and low fO2 on dolomite þ coesite stability
spectrochemical analysis; 3. Composition of some ferroan dolo-
mites. J. Geol., 70, 659–687.
Gunn, S.G. & Luth, R.W. (2006): Carbonate reduction by Fe-S-O
melts at high pressure and high temperature. Am. Mineral., 91,
1110–1116.
Hammouda, T. (2003): High pressure melting of carbonated eclogite
and experimental constraints on carbon recycling and storage in
the mantle. Earth Planet. Sci. Lett., 214, 357–368.
Hammouda, T., Andrault D., Koga, K., Katsura, T., Martin, A.M.
(2010): Ordering in double carbonates and implications for
processes at subduction zones. Contrib. Mineral. Petrol., DOI:
10.1007/s00410-010-0541-z.
Holland, T.J.B. & Powell, R. (1998): An internally consistent ther-
modynamic data set for phases of petrological interest. J.
Metamorphic Geol., 16, 309343.
Ionov, D.A., Dupuy, C., O’Reilly, S.Y., Kopylova, M.G., Genshaft,
Y.S. (1993): Carbonated peridotite xenoliths from Spitsbergen:
implications for trace element signature of mantle carbonate
metasomatism. Earth Planet. Sci. Lett., 119, 283–297.
Katsura, T. & Ito, E. (1989): The system Mg2SiO4-Fe2SiO4 at high
pressures and temperatures: precise determination of stabilities
of olivine, modified spinel, and spinel. J. Geophys. Res., 94,
15663–15670.
Klein-Ben David, O., Izraeli, E.S., Hauri, E., Navon, O. (2007):
Fluid inclusions in diamonds from the Diavik mine, Canada
and the evolution of diamond-forming fluids. Geochim.
Cosmochim. Acta, 71, 723–744.
King, R.L. & Bebout, G.E. (2006): Metamorphic evolution along the
slab/mantle interface within subduction zones. Geochim.
Cosmochim. Acta, 70, 18 (Supplement 1), A319.
Laverne, C. (1993): Occurrence of siderite and ankerite in young
basalts from the Galapagos Spreading Center (DSDP Holes
506G and 507B). Chem. Geol., 106, 27–46.
Luque, F.J., Pasteris, J.D., Wopenka, B., Rodas, M., Barrenechea,
J.F. (1998): Natural fluid-deposited graphite: mineralogical
characteristics and mechanisms of formation. Am. J. Sci., 298,
471–498.
Luth, R.W. (1993): Diamonds, eclogites, and oxidation state of the
Earth’s mantle. Science, 261, 66–68.
— (1995): Experimental determination of the reaction dolomite þ
2coesite ¼ diopside þ 2CO2 to 6 GPa. Contrib. Mineral. Petrol.,
122, 152–158.
— (1999): Carbon and carbonates in the mantle. in: ‘‘Mantle Petrology:
Field Observations and High Pressure Experimentation. A Tribute
to Francis (Joe) Boyd’’, Y. Fei, C.M. Bertka, B.O. Mysen, eds.
Geochem. Soc. Spec. Publ., 6, 297–316.
— (2001): Experimental determination of the reaction aragonite þ
magnesite ¼ dolomite at 5 to 9 GPa. Contrib. Mineral. Petrol.,
141, 222–232.
— (2006): Experimental study of the CaMgSi2O6–CO2 system at
3–8 GPa. Contrib. Mineral. Petrol., 151, 141–157.
Lüttge, A. & Metz, P. (1991): Mechanism and kinetics of the reac-
tion 1 Dolomite þ 2 Quartz ¼ 1 Diopside þ 2 CO2. Can.
Mineral., 29, 803–821.
—, — (1993): Mechanism and kinetics of the reaction 1 Dolomite þ
2 Quartz ¼ 1 Diopside þ 2 CO2: a comparison of rock-sample
and powder experiments. Contrib. Mineral. Petrol., 115,
155–164.
15
Marty, B. & Tolstikhin, I.N. (1998): CO2 fluxes from mid-ocean
ridges, arcs and plumes. Chem. Geol., 145, 233–248.
McCammon, C.A., Griffin, W.L., Shee, S.R., O’Neill, H.S.C. (2001):
Oxidation during metasomatism in ultramafic xenoliths from the
Wesselton kimberlite, South Africa: implications for the survival
of diamond. Contrib. Mineral. Petrol., 141, 287–296.
Moine, B.N., Grégoire, M., O’Reilly, S.Y., Delpech, G., Sheppard,
S.M.F., Lorand, J.P., Renac, C., Giret, A., Cottin, J.-Y. (2004):
Carbonatite melt in oceanic upper mantle beneath the Kerguelen
Archipelago. Lithos, 75, 239–252.
Morlidge, M., Pawley, A., Droop, G. (2006): Double carbonate
breakdown reactions at high pressures: an experimental study
in the system CaO–MgO–FeO–MnO–CO2. Contrib. Mineral.
Petrol., 152, 365–373.
Nakamura, K. & Kato, Y. (2004): Carbonatization of oceanic crust
by the seafloor hydrothermal activity and its significance as a
CO2 sink in the Early Archean. Geochim. Cosmochim. Acta, 68,
4595–4618.
Nixon, P.H. (1995): A review of mantle xenoliths and their role in
diamond exploration. J. Geodyn., 20, 305–329.
O’Neill, H.St.C (1986): Mo-MoO2 (MOM) oxygen buffer and the
free energy of formation of MoO2. Am. Mineral., 71,
1007–1010.
Parkinson, I.J. & Arculus, R.J. (1999): The redox state of subduction
zones: insights from arc-peridotites. Chem. Geol., 160, 409–423.
Plank, T. & Langmuir, C.H. (1998): The chemical composition of
subducting sediment and its consequences for the crust and
mantle. Chem. Geol., 145, 325–394.
Robie, R.A. & Hemingway, B.S. (1995): Thermodynamic properties
of Minerals and Related Substances at 298,15K and 1Bar (105
Pascals) Pressure and at Higher Temperatures. U.S. Geol. Surv.
Bull., 2131.
Schiano, P., Clocchiatti, R., Shimizu, N., Weis, D., Mattielli, N.
(1994): Cogenetic silica-rich melts trapped in mantle minerals in
Kerguelen ultramafic xenoliths: implications for metasomatism in
the oceanic upper mantle. Earth Planet. Sci. Lett., 123, 167–178.
Schrauder, M. & Navon, O. (1993): Solid carbon dioxide in a natural
diamond. Nature, 365, 42–44.
Staudigel, H.R., Hart, S.R., Schmincke, H.-U., Smith, B.M. (1989):
Cretaceous ocean crust at DSDP Sites 417 and 418: Carbon
uptake from weathering versus loss by magmatic outgassing.
Geochim. Cosmochim. Acta, 53, 3091–3094.
Sung, J. (2000): Graphite-diamond transition under high pressure: a
kinetics approach. J. Mater. Sci., 35, 6041–6054.
Viljoen, K.S. (1995): Graphite- and diamond-bearing eclogite xeno-
liths from the Bellsbank kimberlite, Northern Cape, South
Africa. Contrib. Mineral. Petrol., 121, 414–423.
Wang, A., Pasteris, J.D., Meyer, H.O.A., Dele-Duboi, M.L. (1996):
Magnesite-bearing inclusion assemblage in natural diamond.
Earth Planet. Sci. Lett., 141, 293–306.
Wang, J., Hattori, K.H., Kilian, R., Stern, C.R. (2007):
Metasomatism of sub-arc mantle peridotites below southern-
most South America: reduction of fO2 by slab-melt. Contrib.
Mineral. Petrol., 153, 607–624.
Wopenka, B. & Pasteris, J.D. (1993): Structural characterization
of kerogens to granulite-facies graphite: Applicability of
Raman microprobe spectroscopy. Am. Mineral., 78,
533–557.
16 A.M. Martin, T. Hammouda

Yaxley, G.M. & Green, D.H. (2004): Phase relations of carbonate- Zhang, Y. & Zindler, A. (1993): Distribution and evolution of
bearing eclogite assemblages from 2.5 to 5.5 GPa: implications carbon and nitrogen in Earth. Earth Planet. Sci. Lett., 117,
for petrogenesis of carbonatites. Contrib. Mineral. Petrol., 146, 331–345.
606–619.
Yaxley, G.M., Green, D.H., Kamenetsky, V. (1998): Carbonatite Received 10 March 2010
metasomatism in the Southeastern Australian lithosphere. J. Modified version received 28 June 2010
Petrol., 39, 1917–1930. Accepted 22 September 2010