Internally consistent geothermometers for garnet peridotites and ...
Transcript of Internally consistent geothermometers for garnet peridotites and ...
ORIGINAL PAPER
Internally consistent geothermometers for garnet peridotitesand pyroxenites
Paolo Nimis • Herman Grutter
Received: 15 June 2009 / Accepted: 8 October 2009 / Published online: 27 October 2009
� Springer-Verlag 2009
Abstract Mutual relationships among temperatures esti-
mated with the most widely used geothermometers for
garnet peridotites and pyroxenites demonstrate that the
methods are not internally consistent and may diverge by
over 200�C even in well-equilibrated mantle xenoliths. The
Taylor (N Jb Min Abh 172:381–408, 1998) two-pyroxene
(TA98) and the Nimis and Taylor (Contrib Mineral Petrol
139:541–554, 2000) single-clinopyroxene thermometers
are shown to provide the most reliable estimates, as they
reproduce the temperatures of experiments in a variety of
simple and natural peridotitic systems. Discrepancies
between these two thermometers are negligible in appli-
cations to a wide variety of natural samples (B30�C). The
Brey and Kohler (J Petrol 31:1353–1378, 1990) Ca-in-Opx
thermometer shows good agreement with TA98 in the
range 1,000–1,400�C and a positive bias at lower T (up to
?90�C, on average, at TTA98 = 700�C). The popular Brey
and Kohler (J Petrol 31:1353–1378, 1990) two-pyroxene
thermometer performs well on clinopyroxene with Na
contents of *0.05 atoms per 6-oxygen formula, but shows
a systematic positive bias with increasing NaCpx (?150�C
at NaCpx = 0.25). Among Fe–Mg exchange thermometers,
the Harley (Contrib Mineral Petrol 86:359–373, 1984)
orthopyroxene–garnet and the recent Wu and Zhao
(J Metamorphic Geol 25:497–505, 2007) olivine–garnet
formulations show the highest precision, but systematically
diverge (up to ca. 150�C, on average) from TA98 estimates
at T far from 1,100�C and at T \ 1,200�C, respectively;
these systematic errors are also evident by comparison with
experimental data for natural peridotite systems. The older
O’Neill and Wood (Contrib Mineral Petrol 70:59–70,
1979) version of the olivine–garnet Fe–Mg thermometer
and all popular versions of the clinopyroxene–garnet
Fe–Mg thermometer show unacceptably low precision,
with discrepancies exceeding 200�C when compared to
TA98 results for well-equilibrated xenoliths. Empirical
correction to the Brey and Kohler (J Petrol 31:1353–1378,
1990) Ca-in-Opx thermometer and recalibration of the
orthopyroxene–garnet thermometer, using well-equili-
brated mantle xenoliths and TA98 temperatures as
calibrants, are provided in this study to ensure consistency
with TA98 estimates in the range 700–1,400�C. Observed
discrepancies between the new orthopyroxene–garnet
thermometer and TA98 for some localities can be inter-
preted in the light of orthopyroxene–garnet Fe3? parti-
tioning systematics and suggest localized and lateral
variations in mantle redox conditions, in broad agreement
with existing oxybarometric data. Kinetic decoupling of
Ca–Mg and Fe–Mg exchange equilibria caused by transient
heating appears to be common, but not ubiquitous, near the
base of the lithosphere.
Keywords Thermobarometry � Mantle xenoliths �Garnet peridotites � Garnet pyroxenites
Communicated by C. Ballhaus.
Electronic supplementary material The online version of thisarticle (doi:10.1007/s00410-009-0455-9) contains supplementarymaterial, which is available to authorized users.
P. Nimis (&)
Dipartimento di Geoscienze, Universita di Padova,
via Giotto 1, 35137 Padua, Italy
e-mail: [email protected]
P. Nimis
CNR-IGG, Padua, Italy
H. Grutter
BHP Billiton World Exploration Inc,
#800 Four Bentall, 1055 Dunsmuir Street,
Vancouver, BC V7X 1L2, Canada
123
Contrib Mineral Petrol (2010) 159:411–427
DOI 10.1007/s00410-009-0455-9
Introduction
Thermobarometry of garnet-bearing ultramafic rocks has
long supplied invaluable insight into the nature and evo-
lution of mantle rocks and high-grade crustal terrains.
Since the pioneering work of Davis and Boyd (1966) and
Boyd (1973), a large number of suitable geothermometers
have been proposed and several authors have attempted to
assess their reliability using constraints provided by
experiments and natural samples (e.g., Finnerty and Boyd
1984, 1987; Carswell and Gibb 1987; Brey and Kohler
1990; Taylor 1998; Smith 1999; Xu et al. 1999; Nimis and
Trommsdorff 2001a; Wu and Zhao 2007). The most widely
used geothermometers are those based on Ca–Mg equi-
libria between pyroxenes and on Fe–Mg equilibria between
garnet and olivine, orthopyroxene or clinopyroxene. These
methods allow estimation of equilibrium temperatures
from routine electron microprobe analyses, provided an
independent estimate of pressure is available.
The two-pyroxene thermometer is generally believed to
yield the most reliable estimates owing to the small pres-
sure dependence and the relatively small effect of minor
components, and specifically ferric iron, on pyroxene
mutual solubility in ultramafic systems. The Brey and
Kohler (1990) version of this thermometer (hereafter
referred to as BKN) has long represented a standard in
mantle studies. Taylor (1998) showed that the BKN for-
mulation tended to overestimate the temperature of fertile
peridotite compositions and provided a new version
(hereafter referred to as TA98), which incorporated
improved corrections for minor components, specifically
Ti, Fe and Na. This notwithstanding, the TA98 thermo-
meter has received little favor from mantle researchers,
who probably prefer to maintain consistency with previ-
ously published thermobarometric data.
The Ca-in-Opx thermometer of Brey and Kohler (1990;
hereafter BKNCa-in-Opx) provides a potential alternative to
the two-pyroxene formulation. Brey and Kohler (1990)
observed systematic deviations between BKNCa-in-Opx and
two-pyroxene temperatures for mantle xenoliths and sug-
gested that this possibly resulted from a neglected influence
of Na on the Ca content in orthopyroxene. Despite this
caveat, the BKNCa-in-Opx thermometer has been widely
used, employed as calibrant for other thermometers (Witt-
Eickschen and Seck 1991) and has even been suggested as
the best choice for peridotites at low T (Smith 1999).
Thermometers based on Fe–Mg exchange between
garnet and olivine (O’Neill and Wood 1979; O’Neill 1980;
Wu and Zhao 2007), garnet and orthopyroxene (Harley
1984; Lee and Ganguly 1988; Carswell and Harley 1990)
or garnet and clinopyroxene (Krogh 1988; Ai 1992;
Berman et al. 1995; Krogh Ravna 2000) are highly sensi-
tive to variations in Fe oxidation state. Considering all Fe
as Fe2?, generally improves agreement with two-pyroxene
temperature estimates (Canil and O’Neill 1996). Even so,
systematic discrepancies are common for natural xenoliths
and errors as large as ca. 250�C can be expected (Brey and
Kohler 1990; Canil and O’Neill 1996; Taylor 1998; Nimis
and Trommsdorff 2001a). Although the above consider-
ations pose serious doubts on their reliability, Fe–Mg
exchange thermometers remain popular for studies on
mantle xenoliths and inclusions in diamonds. They have
even been used as empirical calibrants for other ther-
mometers (Ryan et al. 1996; Creighton 2009) and still
represent the only viable major element-based option for
thermometry of samples lacking one of the two pyroxenes.
Despite over 40 years of mantle thermobarometry, an
internally consistent set of precise thermometers for garnet-
bearing ultramafic rocks is evidently still not available.
This notwithstanding, the agreement (or disagreement)
between independent temperature estimates has often been
tendered as proof of good (or poor) mineral equilibration
and reliable (or unreliable) thermobarometry (e.g., Franz
et al. 1996a, b; Woodland and Koch 2003; Stachel et al.
2004; Lazarov et al. 2009; Creighton et al. 2009). In the
present study, the best combination of major element
thermometric methods is assessed by cross-validation of T
estimates obtained for xenoliths of garnet peridotite, garnet
pyroxenite and polyphase megacrysts derived from diverse
mantle environments and for relevant experimental data.
An internally consistent empirical recalibration of the
orthopyroxene–garnet thermometer is also proposed.
Outdated thermometers that have already proved unreliable
based on previous evaluations on peridotitic systems
(cf. Brey and Kohler 1990; Taylor 1998) will not be
considered here. These include the more or less popular
versions of the pyroxene thermometer by Wells (1977),
Kretz (1982), Bertrand and Mercier (1985), and Finnerty
and Boyd (1987), the orthopyroxene–garnet Fe–Mg ther-
mometer of Lee and Ganguly (1988), and the clinopyro-
xene–garnet Fe–Mg thermometers of Ellis and Green
(1979) and Powell (1985).
The xenolith database
The database utilized in this work draws on mantle xenolith
data compilations previously used for mantle thermometry
(Grutter et al. 1999), for classification of mantle garnet
compositions (Grutter et al. 2004) and for investigation of
mantle pyroxene thermobarometers (Grutter and Moore
2003; Grutter 2009), further integrated with the most
recently published data. For assessment and calibration of
the thermometers in this study, we focused on xenolith
records comprising analyses for each of orthopyroxene,
clinopyroxene and garnet (plus olivine and spinel where
412 Contrib Mineral Petrol (2010) 159:411–427
123
available) that are derived from alkaline igneous rocks
erupted in a range of geodynamic environments. Records
for alpine-type or ultrahigh-pressure peridotite massifs
were specifically excluded from consideration, owing to
the high probable incidence of disequilibrium data result-
ing from partial reequilibration under low T metamorphic
conditions (Nimis and Morten 2000).
Our starting database contained 1,839 xenoliths and was
dominated by kimberlite-borne suites (see supplementary
material). Electron microprobe analyses of minerals were
screened for quality according to the following restrictions:
for pyroxenes and garnets, oxide totals in the range 98.5–
101.5 wt%; for pyroxenes, cation sums on a 6-oxygen basis
C3.990 apfu. A restriction for cation sums in garnets was
considered unnecessary, owing to the relatively small
influence of garnet composition on thermobarometric
estimates. To remain broadly applicable to peridotite and
pyroxenite bulk compositions, we limited garnet compo-
sitions to Cr2O3 [ 1 wt% or Cr2O3 \ 1 wt% and Mg/
(Mg ? Fe)mol [ 0.6 (cf. Grutter et al. 2004). In cases of
reported chemical zoning, the compositions of mineral
cores were generally selected. This choice minimized dis-
turbance by any heating event associated with host mag-
matism (cf. Boyd et al. 2004) and almost invariably
produced the smallest scatter in P–T estimates. Rim com-
positions were only selected for a few exsolved websterites
from Malaita (Solomon Islands), for which good evidence
of rim reequilibration was provided (Ishikawa et al. 2004).
Choice of the reference barometer
Except where noted, input pressure for temperature cal-
culation was obtained by combining the TA98 thermo-
meter with the Al-in-Opx barometer of Nickel and Green
(1985; hereafter NG85). The choice of the TA98 version of
the two-pyroxene thermometer as our reference thermo-
meter over the more popular BKN of Brey and Kohler (1990)
will be justified in the following section. The NG85 Al-in-
Opx barometer was favored over the more popular BKN of
Brey and Kohler (1990) because (1) NG85 reproduces with
better precision pressures of experiments on variably
depleted to fertile peridotite compositions to 60 kbar
(Fig. 1), (2) BKN increasingly overestimates experimental
pressures at P [ 50 kbar, whereas NG85 only slightly
underestimates at P [ 60 kbar (Fig. 1), (3) NG85 shows
better consistency of carbon species in lherzolite xenoliths
with the graphite–diamond curve (Fig. 2), and (4) NG85
tends to minimize scatter around geotherms in P–T plots
for individual localities (Grutter 2009).
The modification to the NG85 barometer proposed by
Carswell (1991) apparently shows better consistency with
experiments at P [ 60 kbar than the native NG85 formu-
lation (Fig. 1). Carswell’s (1991) modification only affects
P calculations for highly sodic orthopyroxenes with
(Na–Cr–Ti)mol [ 0, for which it is theoretically more cor-
rect. Indeed, in some cases, the modified version was found
to reduce further scatter around geotherms. Nonetheless,
Fig. 1 Discrepancies between calculated and experimental pressures
for experiments on natural peridotite compositions (Ai 1992; Brey
et al. 1990, 2008; Robinson and Wood 1998; Taylor 1998) using
different versions of the Al-in-Opx barometer (PBKN: Brey and
Kohler 1990; PBBG08: Brey et al. 2008; PNG85: Nickel and Green
1985; PNG85mod TA98: PNG85 with XAlM1 in Opx calculated as
recommended by Taylor 1998; PNG85mod CA91: PNG85 with XAlM1 in
Opx calculated as recommended by Carswell 1991). Open symbolsTaylor’s (1998) experiments on highly fertile Hawaiian pyrolite (TiO2
in Opx = 0.23–0.97 wt%). NG85 and NG85mod TA98 show the highest
accuracy and precision for ‘‘normal’’ peridotite compositions to
60 kbar, and progressive underestimation at higher P. NG85mod CA91
shows comparable accuracy and precision to 50 kbar and slightly
overestimates at higher P. BBG08 shows the highest overall accuracy
to 80 kbar, but lower precision at low P. BKN significantly
overestimates above 50 kbar
Contrib Mineral Petrol (2010) 159:411–427 413
123
Carswell’s (1991) modification was also found to yield
unreasonably high pressures for some of these highly sodic
orthopyroxenes and was therefore abandoned. A detailed
reevaluation of Al-in-Opx barometry is beyond the aim of
this study and we cannot exclude that Carswell’s (1991)
version is more accurate, at least for certain compositions. It
is worth noting, however, that the P-dependence of the
TA98 thermometer is only 12–20�C per 5 kbar, so that our
reference temperatures will only marginally be affected by
our preference for the unmodified NG85 barometer.
The modification to the NG85 barometer proposed by
Taylor (1998), which was expressly designed to improve
agreement with experiments on extremely fertile perido-
tites at P B 35 kbar (Fig. 1), was not adopted here because
of its poor consistency with the constraint provided by
graphite or diamond in natural xenoliths (Fig. 2) and
increased scatter in the P–T plots. The latest version of
Brey et al. (2008) was also discarded owing to its lower
overall precision at moderate P (Fig. 1) and severe
inconsistency with the graphite–diamond curve (Fig. 2).
Fig. 2 P–T estimates for
graphite-bearing (solid circles)
and diamond-bearing (opendiamonds) lherzolites using
different versions of the Al-in-
Opx barometer (PBKN: Brey and
Kohler 1990; PBBG08: Brey
et al. 2008; PNG85: Nickel and
Green 1985; PNG85mod Taylor98:
PNG85 with XAlM1 calculated as
recommended by Taylor 1998;
PNG85mod Carswell91: PNG85 with
XAlM1 calculated as recommended
by Carswell 1991; PSI08:
Simakov 2008). Tie-lines
connect estimates obtained
using different published
analyses of the same sample.
The NG85 version and its
modification by Carswell (1991)
are the most consistent with
diamond and graphite stability.
The diamond–graphite
boundary was calculated from
thermodynamic properties of
carbon after Chatterjee (1991)
and is in excellent agreement
with the experiments of
Kennedy and Kennedy (1976) at
T = 1,100–1,600�C
414 Contrib Mineral Petrol (2010) 159:411–427
123
Among barometers based on equilibria other than the
garnet–orthopyroxene Al-transfer reaction, the recent
clinopyroxene–garnet barometer of Simakov (2008) was
discarded because of the unreasonably low pressure esti-
mates obtained for several graphite- or diamond-bearing
samples (Fig. 2). The single-mineral garnet barometer of
Ashchepkov (2006) was not considered owing to previ-
ously demonstrated low precision against Al-in-Opx
methods (discrepancies up to several tens of kbar; see
Fig. 4 in Ashchepkov 2006). The Cr-in-Cpx barometer of
Nimis and Taylor (2000), which might represent a valid
alternative method for well-equilibrated coarse-textured
peridotites, tends to underestimate at P greater than ca.
50 kbar (Nimis 2002) and may be highly sensitive to dis-
equilibrium in chemically and thermally perturbed sheared
xenoliths, owing to the low diffusivity and low activity
of Cr in Cpx at high P. For this reason, it will not be
considered in this study.
A quest for equilibrium
Due to the variable detail of available petrographic
descriptions, no limitations were generally adopted for the
degree of textural equilibration of the xenoliths. An
exception occurred for high T samples from the southern
Wyoming craton (Hearn and McGee 1984; Hearn 2004),
which showed significant within-grain and grain-to-grain
chemical heterogeneities, yielded highly scattered P–T
values and showed evidence of late melt addition (Hearn
2004). These samples were excluded from the data set. A
few samples with TTA98Cpx–Opx \ 700�C were also discarded in
view of the poor reliability of pyroxene thermometry at
very low T. Peridotites with PNG85 \ 15 kbar were also
discarded as such pressures fall below the minimum P
required for garnet to be stable in lherzolite (O’Neill 1981;
Klemme and O’Neill 2000). A lower minimum limit of
10 kbar was adopted for pyroxenites, given the extended
stability of garnet in olivine-free ultramafic rocks
(Gasparik 1984). All other samples were checked for
chemical equilibrium by comparing thermometric results
obtained using independent thermometers. Since internally
consistent thermometers were not yet available, a step-by-
step selection was made by considering systematic rela-
tionships between temperatures obtained by different
methods, as described below.
A first check of mineral equilibration was made by
comparing temperature estimates obtained with TA98 and
BKNCa-in-Opx. Although both are based on the clinopy-
roxene–orthopyroxene equilibria, the former thermometer
is essentially dependent on the composition of the clino-
pyroxene, whereas the latter depends only on the compo-
sition of the orthopyroxene. We find that most (94%) of the
selected samples show good agreement of BKNCa-in-Opx
with TA98 over the range 1,400–1,000�C, although a
systematic positive deviation is evident at lower T (ca.
90�C on average at TTA98Cpx–Opx \ 700�C; Fig. 3a). Despite
the systematic deviation, the good correlation between
these two independent thermometers indicates that in most
mantle xenoliths, pyroxene equilibrium was probably
achieved to T as low as 700�C (Fig. 3a). Such observation
Fig. 3 a Correlation betweenTBKNCa-in-Opx and TTA98
Cpx–Opx calculated at Pgiven by the combination of NG85 barometer and TA98 thermometer.
Points more than 60�C off the correlation line (open circles) were
discarded. b Correlation between TKR00Cpx–Grt and THA84
Opx–Grt for samples
selected according to a, calculated at same P as above. Arrows point
to outliers off the plot field. Points more than 80�C off the correlation
line (open circles) were discarded. Source of data is given in the
online supplementary material
Contrib Mineral Petrol (2010) 159:411–427 415
123
contradicts the common belief that pyroxene-based ther-
mometry is relatively insensitive to T below 900�C, and
that pyroxene disequilibrium should be expected in most
cases under these conditions (e.g., Smith 1999). The cor-
relation between TTA98Cpx–Opx and TBKN
Ca-in-Opx was modeled with
a third-order polynomial fit, and samples plotting [60�C
off the fit curve were discarded (open circles in Fig. 3a). In
doing this, uncertainties on both thermometers were
allowed, i.e., the minimum distance from the fit curve was
considered. As the rejected samples do not show any
compositional peculiarity, they most likely reflect depar-
ture from clinopyroxene–orthopyroxene equilibrium or
poor chemical analyses.
An independent check for possible Fe–Mg disequili-
brium was made by comparing results of the clinopyroxene–
garnet (Krogh Ravna 2000) and orthopyroxene–garnet
(Harley 1984; hereafter HA84) Fe–Mg exchange ther-
mometers. Although both Fe–Mg thermometers fail in
many cases to reproduce within reasonable uncertainties
the temperatures estimated from pyroxene thermometry,
their results show a significant mutual correlation (Fig. 3b).
The correlation was again modeled with a third-order
polynomial fit, and samples plotting[80�C off the fit curve
were discarded. This screening eliminated an additional 6%
of previously accepted samples that likely suffer from Fe–
Mg disequilibrium between pyroxenes and garnet.
A final cross-validation of pyroxene-based and Fe–Mg
exchange thermometry was made by comparing results
for TA98 and HA84. When samples from individual
localities are considered, the difference between THA84Opx-Grt
and TTA98Cpx–Opx (hereafter DTHA84) tends to be linearly
correlated with temperature (i.e., TTA98Cpx–Opx) (Fig. 41). The
DTHA84 correlation lines may show variable slopes or
slight T displacement from one xenolith locality to the
next, but a general tendency for decreasing DTHA84 at
higher TTA98Cpx–Opx appears to be systematic. Scatter around
the correlation lines is mostly restricted to the highest T
regions (Fig. 4 and online supplementary Fig. 1) and can
be interpreted in the light of existing data on Fe3? par-
titioning between garnet and orthopyroxene (Canil and
O’Neill 1996).
THA84Opx–Grt values depend on ln KGrt�Opx
Fe2þ�Mg; which can be
written as
ln KGrt�Opx
Fe2þ�Mg¼ ln
ðFe2þÞGrt
ðFe2þÞOpxþ lnðMgÞOpx
ðMgÞGrt:
Treating total Fe as Fe2?, as done in common practice,
will affect calculated ln K as follows:
lnKGrt�OpxFetot�Mg � ln KGrt�Opx
Fe2þ�Mg¼ ln
Fetotð ÞGrt
Fetotð ÞOpx� ln
ðFe2þÞGrt
ðFe2þÞOpx
¼ ln1� ðFe3þ=FetotÞOpx
1� ðFe3þ=FetotÞGrt:
Figure 5 shows that in mantle garnet peridotites, Fe3?/
Fetot ratios are higher in orthopyroxene at low P–T, but
favor garnet at high P–T. This observation is in keeping
with the stabilization of Fe3Fe23?Si3O12 (skiagite) com-
ponent in garnet at higher P (Gudmundsson and Wood
1995) and the enhanced partitioning of Fe2? into ortho-
pyroxene at higher T (e.g. Harley 1984). Given T and P are
strongly correlated in the available data, it is difficult to
differentiate thermal and baric effects on Fe3?/Fetot rela-
tions. In any case, Fig. 5 suggests that redox variations will
have an opposite influence on calculated KGrt�Opx
Fe2þ�Mgat low
and high P–T, respectively. Accordingly, outliers displaced
to higher DTHA84 values relative to DTHA84 correlation
lines for individual localities (Fig. 4) may reflect (1)
localized oxidation at low P–T, with Fe3?/Fetot higher in
the orthopyroxene, or (2) reduction at high P–T, with Fe3?/
Fetot higher in the garnet, or (3) disequilibrium. Localized
oxidation is typically induced by metasomatism (e.g., Zhao
et al. 1999; McCammon et al. 2001; Creighton et al. 2009).
Disequilibrium may result from transient heating and
kinetic decoupling of ‘‘faster’’ Fe–Mg and ‘‘slower’’
Ca–Mg equilibria, and is the most likely explanation for
elevated DTHA84 values occurring at high T and near the
base of the lithosphere (cf. Smith and Boyd 1992; Franz
et al. 1996a; Griffin et al. 1996a, b). Outliers displaced to
lower DTHA84 values at high P–T may instead reflect oxi-
dation related to metasomatism or, again, disequilibrium.
To clean our database from any potentially ill-equili-
brated sample, we discarded eye-selected outliers in plots of
DTHA84 versus TTA98Cpx–Opx for individual localities (Fig. 4 and
supplementary Fig. 1). Most likely, this operation also
eliminated several samples equilibrated under redox condi-
tions unlike those typical for unperturbed lithospheric mantle
(cf. Frost and McCammon 2008). A total of 84 xenolith
records were additionally abandoned for localities where
DTHA84 versus TTA98Cpx–Opx trends could not be adequately
recognized owing to excessive scatter or to limited number
of data (see supplementary Fig. 1). Finally, three suspicious
samples with P–T estimates well off the local xenolith geo-
therm were rejected. The remaining 764 samples in our
go-forward data set virtually cover the whole range of P–T
conditions for garnet-facies lithospheric mantle (Fig. 6) and
can be considered as the most comprehensive available ref-
erence data set of well-equilibrated, garnet-bearing ultra-
mafic rocks. Despite culling some 1,000 records in total from
our starting database of over 1,800 two-pyroxene ? garnet-
bearing xenoliths, the mineral compositions for peridotites
1 A complete set of locality plots showing relations between various
thermometers and TA98 is given in supplementary online Fig. 1.
416 Contrib Mineral Petrol (2010) 159:411–427
123
and pyroxenites in our go-forward data set cover a substan-
tially broader overall range in Cr, Al, Ti, Mg/Fe, Ca/Mg
and particularly Na content than has been compiled to date
for any evaluation of mantle thermometers by either
empirical or experimental methods. The go-forward data set
is available from either author on request.
Pyroxene thermometry
Two-pyroxene thermometers
The BKN two-pyroxene thermometer generally overesti-
mates temperatures relative to the TA98 thermometer
(Fig. 7a). The discrepancy is strictly correlated with the Na
content of the clinopyroxene, with a change in sign at
NaCpx & 0.05 apfu (Fig. 7b). For the most sodic samples
(NaCpx & 0.38 apfu), the difference is ca. 180�C. Although
such extremely sodic samples are relatively uncommon in
mantle-derived ultramafic rocks, this strong compositional
effect is worrying and should be considered as most signi-
ficant for the choice of the optimum pyroxene thermometer.
The choice is not straightforward. On the one hand, the
experimental data used for calibration of the TA98 thermo-
meter cover a broader range of Na-, Fe- and Ti-rich
compositions, which may have allowed improved appraisal
of the effects of these minor components on the pyroxene
solvus. On the other hand, 36 of 62 calibration experiments
were unreversed and were run for relatively short times,
suggesting a potentially poor approach to equilibrium.
There are two good reasons, however, to prefer the TA98
two-pyroxene thermometer over the BKN formulation.
Fig. 4 Relationships between
THA84Opx–Grt–TTA98
Cpx–Opx and
temperature for representative
and well-known xenolith
localities. Circles peridotites,
triangles pyroxenites. Opensymbols indicate points falling
off the main linear trends, which
were excluded from the
database. The complete set of
locality plots is given in the
supplementary Fig. 1. See text
for further explanation
Contrib Mineral Petrol (2010) 159:411–427 417
123
First, the TA98 thermometer shows better agreement with
BKNCa-in-Opx, whereas the BKN two-pyroxene thermometer
shows significant systematic deviations at T between
900 and 1,400�C, as well as larger scatter in TBKNCa-in-Opx versus
TCpx–Opx plots (compare Figs. 3a, 8). Second, the TA98
thermometer reproduces temperatures of experiments in
highly sodic NCMAS systems (NaCpx = 0.12–0.44 apfu;
NaOpx = 0.027–0.104 apfu; Bulatov et al. 2002), while
the BKN thermometer significantly overestimates them
(Fig. 9). Accordingly, we suggest that discrepancies
between TBKNCpx–Opx and TTA98
Cpx–Opx are essentially due to inap-
propriate treatment of NaCpx in the BKN formulation. Not
surprisingly, the best agreement between TA98 and BKN
two-pyroxene thermometers is observed for NaCpx contents
of about 0.05 apfu, which is close to the average NaCpx in
Brey and Kohler’s (1990) experiments (0.059 apfu;
range = 0.015–0.089 apfu). Following Bertrand and Mer-
cier (1985), Brey and Kohler (1990) expressed the enstatite
activity in pyroxenes as 1 - Ca/(1 - Na), which evidently
over-emphasizes the influence of Na. The more conventional
expression adopted by Taylor (1998), which is based on a
three-site solid solution model, i.e. (1 - Ca - Na) �(1 - [6]Al – Cr - Ti) � (1 - [4]Al/2)2, appears to be more
robust over the range of pyroxene compositions observed in
mantle-derived ultramafic rocks.
The two-pyroxene Na-partitioning thermometer of Brey
and Kohler (1990) shows very low precision relative to the
TA98 thermometer (see supplementary Fig. 1). It is
unclear whether this results from a simplified formulation,
restricted bulk Na range of calibration experiments or
inadequate Na determinations in routine electron micro-
probe analyses. At present, its application in mantle studies
cannot be recommended.
Single-Cpx thermometers
The en-in-Cpx version of the pyroxene thermometer (Nimis
and Taylor 2000; NT00) yields estimates very similar to the
TA98 thermometer (Fig. 10). The excellent agreement
between the two thermometers (±30�C) is not surprising,
Fig. 5 Variation ofðFe3þ=RFeÞOpx
ðFe3þ=RFeÞGrt
ratio versus T and P in mantle
xenoliths of various provenance, calculated using Mossbauer data for
garnet, clinopyroxene and, where available, orthopyroxene and Canil
and O’Neill’s (1996) orthopyroxene–clinopyroxene Fe3? partitioning
systematics. For clinopyroxene-bearing samples (solid symbols), only
samples satisfying TTA98Cpx–Opx versus TBKN
Ca-in-Opx relationships discussed
in the text were considered and T was estimated with the TA98
thermometer (±50�C and ±4 kbar uncertainties were assumed); for
clinopyroxene-free samples (open symbols), T was estimated with the
new orthopyroxene–garnet thermometer (see Eq. 4; ±70�C and
±5 kbar uncertainties were assumed). Conservative uncertainties onðFe3þ=RFeÞOpx
ðFe3þ=RFeÞGrt
were calculated assuming an absolute error of 0.02 on
garnet and orthopyroxene Fe3?/RFe data and normal error propaga-
tion. Source of data: circles Canil and O’Neill (1996); trianglesWoodland and Peltonen (1999); diamond McCammon et al. (2001;
garnet core analyses). McCammon et al’s (2001) data referred to
strongly zoned, metasomatized samples that fell off the general trends
and were excluded from regressions
Fig. 6 P–T conditions for the selected well-equilibrated mantle
xenoliths, calculated by the combination of NG85 Al-in-Opx
barometer and TA98 two-pyroxene thermometer. Reference geo-
therms after Pollack and Chapman (1977)
418 Contrib Mineral Petrol (2010) 159:411–427
123
since they use a similar expression for enstatite activity in
clinopyroxene and similar corrections for minor compo-
nents. The NT00 calibration was based on a much larger
experimental data set, which included also experiments on
CMS, CMASCr and variably Cr-enriched natural peridotite
systems, and may thus be more robust against composi-
tional factors. On the other hand, the single-mineral for-
mulation is intrinsically simplified, as the enstatite activity
in the coexisting orthopyroxene is considered as fixed over
the whole P–T–X space of interest. Since potential dis-
crepancies between TA98 and NT00 are within commonly
accepted uncertainties (e.g., ±15–30�C at the 1r level; Brey
and Kohler 1990), the two thermometers can be considered
as ideal alternatives. Significantly, their mutual inter-
changeability permits consistency to be maintained between
temperatures estimated from garnet lherzolite xenoliths and
single-clinopyroxene xenocrysts that are assumed to be in
equilibrium with orthopyroxene (e.g., Grutter 2009).
Simakov (2008) recently proposed a new single-clino-
pyroxene thermometer, which is claimed to reproduce
temperatures of experiments in peridotite systems better
than the NT00 formulation. In his Figs. 9 and 10, however,
calculated TNT00en-in-Cpx values for Taylor’s (1998) experiments
are erroneously reported and about 30–40�C too high.2
Moreover, several experiments from Walter (1998), which
are free of orthopyroxene and may therefore only yield
minimum T estimates, are unduly used to test the perfor-
mance of the thermometers. In fact, the agreement between
NT00 and experimental temperatures is excellent from
900�C to at least 1,500�C (r = 30–40�C for different
clinopyroxene ? orthopyroxene-bearing experimental
sets; Nimis and Taylor 2000) and somewhat better than for
the Simakov (2008) version within this temperature range.
Ashchepkov et al. (2008) also proposed an empirical
correction to the NT00 thermometer to improve consis-
tency with the BKN two-pyroxene thermometer. Based on
our evaluation of the BKN thermometer (see above), such a
correction cannot be recommended.
The Ca-in-Opx thermometer
The BKNCa-in-Opx thermometer does not reproduce run
temperatures of NCMAS experiments as well as the TA98
thermometer, but BKNCa-in-Opx is nevertheless more robust
to Na variations than the BKN two-pyroxene thermo-
meter (Fig. 9). Based on systematic discrepancies between
TBKNCa-in-Opx and TBKN
Cpx–Opx for peridotite xenoliths, Brey
and Kohler (1990) suggested that a correction could be
needed for Na in orthopyroxene. As shown above, the
Fig. 7 Discrepancies between TBKNCpx–Opx and TTA98
Cpx–Opx for well-equil-
ibrated mantle xenoliths, calculated at P given by the combination of
NG85 Al-in-Opx barometer and TA98 two-pyroxene thermometer,
plotted against (a) TTA98Cpx–Opx and (b) NaCpx. Arrows indicate the
additional minor effects of increasing FeCpx and TiCpx on these
discrepancies. The regression line does not include the pyroxenite
samples, for which the Fe–Ti effects are the most evident
Fig. 8 Correlation between TBKNCa-in-Opx and TBKN
Cpx–Opx calculated at Pgiven by the combination of NG85 barometer and BKN two-pyroxene
thermometer. Open circles indicate samples falling more than 60�C
off the fit curve. Arrow points to outliers off the plot field. Compared
with TA98, the BKN two-pyroxene thermometer shows a poorer
agreement with the BKNCa-in-Opx thermometer (cf. Fig. 3a)
2 Note that symbols have been erroneously reported in the captions to
Simakov’s (2008) Figs. 10 and 11: squares and triangles refer to
Taylor’s (1998) and Walter’s (1998) experiments, respectively; solid
and open symbols indicate temperatures calculated using the NT00
and Simakov (2008) thermometers, respectively.
Contrib Mineral Petrol (2010) 159:411–427 419
123
discrepancies rather appear to be related to inappropriate
treatment of Na in the clinopyroxene.
The reason for the systematic, progressive overestimate
at T \ 1,000�C (\30�C at TTA98Cpx–Opx = 900�C, but up to
90�C, on average, at TTA98Cpx–Opx = 700�C) relative to the
TA98 thermometer (Fig. 3a) is not understood. Only one
reversed experiment at the relevant T is available (850�C;
CMS system; Lindsley and Dixon 1976) and its tempera-
ture is reproduced reasonably well by both thermometers
TTA98Cpx–Opx = 845�C; TBK90
Ca-in-Opx = 815�C. Therefore, it
remains unclear which thermometer is more accurate for
very low T peridotites. Although the effect of minor
components may be stronger at low T for chemically more
complicated clinopyroxene than for orthopyroxene, the
sensitivity of orthopyroxene to T is much lower at low T
(Lindsley and Dixon 1976). In addition, BKNCa-in-Opx uses
a simplified single-mineral formulation, which may be
insufficiently robust against variations of diopside activity
in clinopyroxene at low T. Therefore, the suggestion made
by Smith (1999) that the Ca-in-Opx thermometer is
potentially more precise and accurate at very low T does
not appear sufficiently substantiated.
To improve internal consistency between two-pyroxene
and Ca-in-Opx thermometry estimates over the whole
range of mantle temperatures (Fig. 3a), the following
empirical correction can be adopted:
TCa�in�Opxcorr: ð�CÞ ¼ �628:7þ 2:0690 � TCa�in�Opx
BKN
� 4:530 � 10�4ðTCa�in�OpxBKN Þ2 ðR2 ¼ 0:950Þ: ð1Þ
After this correction, Tcorr.Ca-in-Opx and TTA98
Cpx–Opx estimates
for our go-forward xenolith data set agree to within
±106�C, and the positive bias at TTA98Cpx–Opx = 700�C is
reduced to ca. 30�C on average (Fig. 11a). The standard
error of estimate relative to TA98 of the corrected Ca-
in-Opx thermometer decreases from 45�C at T \ 900�C to
36�C at T = 900–1,200�C to 25�C at T [ 1,200�C,
suggesting decreasing precision with decreasing T.
Other single-mineral thermometers
Ashchepkov et al. (2008) proposed several single-mineral
thermometers for kimberlite-borne minerals, including
garnet, chromite and ilmenite. The precision of these
methods against pyroxene thermometers appears too low
(discrepancies up to ca. 200�C; see Fig. 1 in Ashchepkov
et al. 2008) to recommend their use in mantle studies. For
this reason, they will not be considered further here.
Fe–Mg exchange thermometry
Figure 1 (see supplementary material) shows that none of
the existing Fe–Mg exchange thermometers satisfactorily
reproduce TA98 temperatures for mantle xenolith suites. In
particular: (1) the popular olivine–garnet thermometer of
O’Neill and Wood (1979; corrected by O’Neill 1980)
shows very low precision and errors exceeding 200�C,
notwithstanding our meticulous selection of well-equili-
brated samples; (2) the version of this thermometer
proposed by Wu and Zhao (2007), which incorporates a
more sophisticated solution model for garnet, improves
precision, but seemingly also introduces a systematic
positive bias at TTA98Cpx–Opx \ 1,200�C (ca. ?150�C on aver-
age at TTA98Cpx–Opx = 700�C); (3) available versions of the
clinopyroxene–garnet thermometer show such large
uncertainties that their use may appear unwise even for
qualitative purposes; (4) the orthopyroxene–garnet ther-
mometer of Harley (1984) shows the highest precision
Fig. 9 Calculated temperatures
for experiments in NCMAS
system (Bulatov et al. 2002)
using the TA98 and BKN two-
pyroxene thermometers and the
BKNCa-in-Opx thermometer
Fig. 10 Correlation betweenTNT00en-in-Cpx and TTA98
Cpx–Opx calculated at Pgiven by the combination of NG85 Al-in-Opx barometer and TA98
thermometer, for well-equilibrated mantle xenoliths. A polynomial
best-fit curve is shown as a dashed line
420 Contrib Mineral Petrol (2010) 159:411–427
123
relative to two-pyroxene thermometry, but distinct sys-
tematic deviations at low and high T.
The systematic bias observed with the Harley (1984)
and Wu and Zhao (2007) thermometers for natural samples
is also reproduced by experiments on natural peridotitic
compositions (see Fig. 2d in Brey and Kohler 1990, and
Fig. 1c in Wu and Zhao 2007). This indicates that the
discrepancies relative to TA98 are not due to systematic
errors within the TA98 formulation, but rather reflect
inconsistencies in the adopted thermodynamic parameters
when extrapolated to natural compositions and redox
conditions. The particularly low precision of the Cpx–Grt
thermometers is probably due to the relatively high affinity
of Fe3? for clinopyroxene (Canil and O’Neill 1996) and,
possibly, to the stronger influence of minor components
such as Al, Cr and Na on Fe and Mg activities in
clinopyroxene.
A new orthopyroxene–garnet thermometer
Opx–Grt thermometers for ultramafic rocks are based on
the equilibrium
1
3Mg3Al2Si3O12
Grt
þ FeSiO3Opx
$ 1
3Fe3Al2Si3O12
Grt
þMgSiO3Opx
;
ð2Þ
and have been calibrated against experimental data in FMAS
and CFMAS systems run in iron or graphite capsules (Harley
1984; Lee and Ganguly 1988; Carswell and Harley 1990).
The systematic deviations from TA98 make all available
versions of the Opx–Grt thermometer unsuitable for inter-
nally consistent thermobarometry, in spite of their relatively
high precision (Fig. 4; additional plots can be found in
supplementary Fig. 1; cf. also Figs. 2d and 2e in Brey and
Kohler 1990). The Lee and Ganguly (1988) version (not
shown) yields temperatures that are systematically higher by
ca. 140�C, on average, than those of Harley (1984), while
those of Carswell and Harley (1990), which are based on
regression of both Harley’s and Lee and Ganguly’s data, lie
somewhere in between. The relationships with TA98,
however, remain similar for all three versions.
Our attempts to recalibrate the orthopyroxene–garnet
thermometer including also experiments on natural peri-
dotite compositions (Brey et al. 1990, 2008; Ai 1992;
Taylor 1998) were unsuccessful as they revealed system-
atic discrepancies between the different experimental data
sets and did not reduce the disagreement with TA98 for
natural samples at low and high T. The systematic rela-
tionships between HA84 and TA98 (Fig. 4) prompted an
alternative ‘‘natural’’ approach to recalibrate the Opx–Grt
Fe–Mg exchange thermometer. We empirically fitted the
expression
T ¼�DH� � DV�Pþ DW XGrt
Ca þ XGrtMn
� �
R ln KOpx�GrtFe�Mg � DS�
; ð3Þ
where
KGrt�OpxFe�Mg ¼
FeGrtMgOpx
MgGrtFeOpx
;
XGrtCa ¼ Ca=ðCaþ FeþMgþMnÞ;
XGrtMn ¼ Mn=ðCaþ FeþMgþMnÞ;
and Fe ¼ Fetot;
using mineral compositions of the best equilibrated natural
ultramafic rocks in combination with P and T values esti-
mated by the TA98 two-pyroxene thermometer and the
NG85 Al-in-Opx barometer. Compared to conventional
calibration based on experimental data, calibrations based
on natural data suffer from much larger uncertainties in the
P and T of equilibration.3 On the other hand, experimental
Fig. 11 Calibration residuals of (a) the corrected Ca-in-Opx ther-
mometer (Eq. 1) and of (b) the new Opx–Grt thermometer (NG09;
Eq. 4). Black dashed lines are at two standard errors of estimate.
Polynomial best-fit curves (halftone dashed lines) evidence the slight
residual positive bias at very low T (ca. 30�C and 20�C at
TTA98Cpx–Opx = 700�C, respectively)
3 Uncertainties on T can be assumed to be ca. ±50�C, considering the
reported 1r calibration uncertainties of the TA98 thermometer
(±31�C), and propagation of analytical errors and of uncertainties
on P estimates. Uncertainties on P are believed to be ca. ±4 kbar,
considering the T-dependence of the NG85 barometer (ca. 3 kbar per
50�C) and an additional contribution of analytical and calibration
errors on P uncertainties. These rough estimates are consistent with
constraints imposed by diamond and graphite stability (Fig. 2). Some
systematic underestimation of pressure at P [ 60 kbar can also be
expected (cf. Fig. 1).
Contrib Mineral Petrol (2010) 159:411–427 421
123
data may suffer from incomplete equilibration even in
reversed runs, owing to their much shorter duration relative
to natural equilibration processes (cf. Pattison 1994;
Grutter 2009) and from unnatural redox conditions.
Regression through our selected xenolith data (N = 764)
produced the following thermometric expression:
From this equation, a -DH� of 10.1 ± 0.2 kJ/mol, a
-DS� of 6.1 ± 0.1 J/K mol, a -DV� of 0.145 ± 0.002 J/
bar mol, and a DW of 12 ± 1 kJ can be estimated for
reaction (2) (errors at the 1r level).
Our regressed -DH� and -DS� are lower than those
obtained by Harley (1984; -DH� = 15.7 ± 2.6 kJ/mol;
-DS� = 8.2 ± 0.2 J/K mol) and Lee and Ganguly (1988;
-DH� = 16.4 ± 1.4 kJ/mol; -DS� = 8.1 ± 0.1 J/K mol)
from their experiments in the FMAS system at T = 800–
1,400�C, but more similar to those calculated from Berman
and Aranovich’s (1996) internally consistent thermo-
dynamic database at our average calibration temperature
(-DH�1 bar, 1,273 K = 12.25 kJ/mol; -DS�1 bar, 1,273 K =
4.93 J/K mol). Most likely, our reduced -DH� and -DS�values in part serve to counteract the effect of different
overall redox conditions in the mantle and in the
experiments.
Our fitted -DV� is much larger than expected from
previous work (0.089–0.099 J/bar mol; Harley 1984; Lee
and Ganguly 1988; Carswell and Harley 1990; Berman and
Aranovich 1996). Significantly, its value remained virtu-
ally unchanged if samples from off-craton settings, some of
which were characterized by much steeper dT/dP gradients,
were excluded from the calibration, and even increased if a
larger -DH�, more similar to that expected from experi-
mental data, was imposed. The augmented -DV� cannot be
the result of systematic errors in input PNG85 values, as
similar results were obtained if PBKN pressures, which are
known to overestimate at high P (Fig. 1), were used. The
pressure term in Eq. 4 evidently incorporates additional P-
dependent effects, which were recorded by lithospheric
mantle rocks, but were not reproduced in the experiments.
A possible candidate is the systematic variation in oxygen
fugacity with depth in garnet-facies mantle sections (ca.
-0.1 DlogfO2 [FMQ] per kbar; Woodland and Koch 2003;
McCammon and Kopylova 2004; Frost and McCammon
2008; Lazarov et al. 2009), which is accompanied by a
decrease inðFe3þ=FetotÞOpx
ðFe3þ=FetotÞGrt
ratio (Fig. 5). If not properly
counterbalanced, this variation would produce a
progressive increase of KGrt�OpxFe�Mg and T underestimation
with depth if all Fe is treated as Fe2?.
Despite the relatively restricted range of (XCaGrt ? XMn
Grt)
values in the calibration database (0.08–0.21), the calcu-
lated net interaction parameter for garnet is within error of
that estimated by Lee and Ganguly (1988) from thermo-
dynamic data (DW = 12.6 kJ/mol). Incorporation of a
more sophisticated solid solution model for garnet
(Ganguly et al. 1996; Holdaway 2000) produced no
statistical improvement, even if a large WFeMg for ortho-
pyroxene was included.
Although the overall precision of the new orthopyrox-
ene–garnet thermometer is slightly worse than that of the
experimentally derived HA84 version, the calculated
standard error of estimate is reasonably low (±34�C),
given the comparable uncertainties of input TTA98Cpx–Opx val-
ues. More important, the overall systematic deviations
from TA98 observed with previous orthopyroxene–garnet
thermometry formulations were almost totally eliminated
(Fig. 11b). Small discrepancies between TA98 and our new
thermometer still exist for some localities (Fig. 12; addi-
tional plots can be found in supplementary Fig. 1). Rigo-
rous inspection of our data indicates that the discrepancies
are independent of geothermal gradients and the presence/
absence of coexisting spinel, and can largely be explained
in the light of Fe3? partitioning systematics between
orthopyroxene and garnet (cf. Fig. 5).
For instance, samples from several Kaapvaal localities
(including samples discarded after the HA84 vs. TA98 test
illustrated in Fig. 4) yield calculated TNG09Opx–Grt–TTA98
Cpx–Opx
values (DTNG09) close to zero at moderate T and scatter
toward higher DTNG09 at high T (Fig. 12): this pattern is
indicative of transient heating, with decoupling of ‘‘faster’’
Fe–Mg and ‘‘slower’’ Ca–Mg equilibria, near the base of
the lithosphere. Samples from the North and South Slave
craton (Canada) show a significant scatter toward lower
DTNG09 at T [ 1,100�C (Fig. 12): this pattern may reflect
the documented stronger and variable oxidation of the deep
Slave lithosphere (cf. McCammon and Kopylova 1994). A
similar oxidation probably also affected the deep litho-
spheric mantle at Nikos (Somerset Island), and in the
Jagersfontein–Koffiefontein area (Kaapvaal, South Africa)
(Fig. 12). Samples from Central Slave, Kirkland Lake
(Canada), Montana (USA), Karoo (South Africa), Malaita
(Solomon Islands) and Gibeon–Mier (Namibia, South
Africa) show a distinct decrease of DTNG09 with increasing
TNG09ðKÞ ¼1; 215ð�26Þ þ 17:4ð�0:2ÞPðkbarÞ þ 1; 495ð�120Þ XGrt
Ca þ XGrtMn
� �
ln KOpx�GrtFe�Mg þ 0:732ð�0:017Þ
: ð4Þ
422 Contrib Mineral Petrol (2010) 159:411–427
123
T and, for Gibeon–Mier and Malaita, a scatter toward
higher DTNG09 at high T (Fig. 12 and supplementary
Fig. 1). These patterns may reflect more oxidized overall
conditions than is typical for cratonic mantle and, for
Gibeon and Malaita, transient heating at deep levels or
during short-lived residence of the xenoliths in a shallower
magma chamber (cf. Franz et al. 1996a). The pattern for
Kimberley (Kaapvaal, South Africa) is obscured by the
relatively large scatter, but also suggests a slight overall
decrease of DTNG09 with T (Fig. 12). Variable oxidation in
the Kimberley mantle, generally with a three order of
magnitude variation in fO2 at a given depth, was docu-
mented by Creighton et al. (2009).
Equation 4 reproduces the temperatures of experiments
in natural peridotite systems with reasonable accuracy if
experimental uncertainties are considered (Fig. 13). The
slight increase in DTNG09 with P indicates that the built-in
correction for redox variations with depth may be exces-
sive for P–T–fO2 conditions achieved in experiments in
graphite or olivine capsules.
Equation 4 will yield biased estimates for lithosphere
sections characterized by anomalous relationships between
redox conditions and depth and should be used with cau-
tion. Potential discrepancies are expectedly larger for
samples equilibrated at either very low or very high P–T,
i.e., at conditions under which theðFe3þ=FetotÞOpx
ðFe3þ=FetotÞGrt
ratio
becomes significantly different from unity (Fig. 5). In such
cases, potential errors may exceed 150�C, as suggested by
DTNG09 values for samples excluded from calibration
(Fig. 12 and supplementary Fig. 1). The temperature at
which such discrepancies are minimized apparently
Fig. 12 Difference between
temperatures calculated with the
new Opx–Grt thermometer
(NG09; Eq. 4) and TA98, for
representative and well-known
xenolith localities (compare
with Fig. 4). Dashed lines are at
2 standard errors of estimate
(±70�C). Relationships of
TNG09Opx–Grt–TTA98
Cpx–Opx with
temperature are discussed in the
text. Symbols are as in Fig. 4.
The complete set of locality
plots is given in the
supplementary Fig. 5
Contrib Mineral Petrol (2010) 159:411–427 423
123
decreases from suites characterized by low dT/dP gradients
(e.g., Central Slave, ca. 1,200�C; Kirkland Lake, 1,100�C)
to those characterized by progressively higher dT/dP gra-
dients (e.g., Karoo, 1,000�C; Namibia, 950�C; Malaita,
850�C) (Fig. 12 and supplementary Fig. 1). This may be an
artifact caused by the augmented pressure term in Eq. 4,
but may also indicate a significant baric effect on Fe3?
partitioning between orthopyroxene and garnet.
We provide an independent test of our new thermometer
in combination with the NG85 barometer, using constraints
from graphite- or diamond-bearing garnet peridotite
xenoliths and touching orthopyroxene–garnet inclusions in
diamond (Fig. 14). Most PNG85–TNG09 points plot close to
a 40 mW/m2 geotherm, whereas PNG85–THA84 pairs
crosscut theoretical geotherms, as expected from THA84
versus TTA98 relationships (cf. Fig. 4). Also, most PNG85–
TNG09 points are within the respective stability fields of the
associated carbon polymorphs or within less than 1.5 kbar
of the graphite–diamond curve. Only two graphite-bearing
harzburgites (PHN2492 of Pearson et al. 1994, and U-233/
82 of Solovjeva et al. 1995) fall well into the diamond field,
irrespective of the thermometer used. These aberrant
Fig. 13 Difference between temperatures calculated with the new
Opx–Grt thermometer (NG09; Eq. 4) and experimental temperatures
for experiments in natural peridotite systems run in graphite (Ai 1992;
Taylor 1998; Robinson and Wood 1998) or olivine capsules (Brey
et al. 1990). The KLB50 series by Taylor (1998) used a carbonate
flux. Dashed lines are at 2 standard errors of estimate (SEE) of the
thermometer calibration. Double arrows indicate maximum variations
of TNG09Opx–Grt–TTA98
Cpx–Opx between experiments at the same P and T(within ±5�C) for each experimental data set, and give an idea of the
possible effect of experimental uncertainties on calculated T. All
experimental temperatures are reproduced to within 2 SEE if such
uncertainties are considered
Fig. 14 P–T estimates for graphite- or diamond-bearing peridotites
and pyroxenites and touching orthopyroxene–garnet inclusions in
diamonds, obtained using the HA84 (Harley 1984) and NG09 (Eq. 4
of this work) orthopyroxene–garnet thermometers in combination
with the NG85 barometer. Tie-lines connect estimates obtained using
different published analyses of the same sample. Most PNG85–TNG09
points plot close to the 40 mW/m2 geotherm of Pollack and Chapman
(1977), whereas PNG85–THA84 points crosscut theoretical geotherms.
Two aberrant diamond-bearing samples showed strongly diverging
orthopyroxene–garnet and two-pyroxene temperatures (DT = 80–
230�C), suggesting disequilibrium. Note that many harzburgitic
samples, including the two graphite-bearing samples falling within
the diamond field, could not be checked for equilibrium, owing to a
lack of internally consistent alternative thermometers. Source of data
for graphite- and diamond-bearing peridotites is given in the online
Appendix. Only samples satisfying the quality criteria for orthopy-
roxene and garnet analyses adopted for the xenolith database were
considered. The diamond–graphite boundary was calculated from
thermodynamic properties of carbon from Chatterjee (1991)
424 Contrib Mineral Petrol (2010) 159:411–427
123
samples are clinopyroxene-free and could not be checked
for equilibrium, owing to a lack of internally consistent,
alternative thermometers. Their P–T estimates thus remain
suspicious and do not necessarily indicate inconsistency of
the NG09 thermometer.
Conclusions
The most widely used thermometers for mantle-derived
garnet peridotites and garnet pyroxenites are not internally
consistent and may diverge by over 200�C for well-equil-
ibrated samples. The following scheme is suggested for
optimum thermobarometry4:
(a) Calculate P–T conditions using a combination of the
TA98 two-pyroxene thermometer and the NG85 Al-
in-Opx barometer. Using the same NG85 pressure,
also calculate temperatures with BKNCa-in-Opx (Brey
and Kohler 1990) and NG09 (Eq. 4 of this work).
BKNCa-in-Opx temperature should be corrected as in
Eq. 1 to ensure consistency with TA98 and NG09
estimates.
(b) Check for agreement between temperatures obtained
with the TA98 and corrected BKNCa-in-Opx ther-
mometers. Discrepancies larger than 90�C (at
TTA98Cpx–Opx \ 900�C), 70�C (at TTA98
Cpx–Opx = 900–
1,200�C), or 50�C (at TTA98Cpx–Opx [ 1200�C), which is
about twice the standard errors of estimate relative to
TA98 of the corrected Ca-in-Opx thermometer in the
different T intervals, may reflect (1) disequilibrium
between the pyroxenes or (2) poor chemical analyses.
(c) Check relationships between TTA98Cpx–Opx and TNG09
Cpx–Grt
estimates. Discrepancies larger than 70�C, that is
about twice the standard error of estimate relative to
TA98 of the new NG09 thermometer, may indicate
(1) decoupling of Fe–Mg and Ca–Mg equilibria due to
transient thermal perturbations, (2) significantly dif-
ferent redox conditions than in typical unperturbed
cratonic mantle sections or (3) poor chemical
analyses.
(d) For samples within the above limits, the combination
of the TA98 two-pyroxene thermometer and the
NG85 Al-in-Opx barometer will yield the most robust
P–T estimates. Accuracy of temperature estimates
below 900�C and of pressure estimates above 60 kbar
cannot yet be assured from available data. For
samples that do not satisfy the above criteria, P–T
estimates should be taken with caution. For clino-
pyroxene-free garnet harzburgites and garnet
orthopyroxenites, temperatures obtained with the
new orthopyroxene–garnet thermometer (Eq. 4) will
be the most reliable and consistent with TA98
estimates, albeit potentially biased because of tran-
sient thermal perturbations (especially for high T
xenoliths) or different mantle redox conditions.
(e) The use of the available versions of the olivine–garnet
and clinopyroxene–garnet thermometers cannot be
recommended given the results of our work.
Application of the above scheme may be extended to
orogenic garnet peridotites, with the caveat that these rocks
may have equilibrated under very different redox condi-
tions (cf. Canil and O’Neill 1996) and may therefore yield
strongly biased TNG09Cpx–Grt values. An additional problem in
applications to orogenic peridotites is that low T and/or
short-lived metamorphic processes may induce kinetic
decoupling of P- and T-dependent chemical equilibria (cf.
Nimis and Trommsdorff 2001b), which might not be
detected by the proposed screening method. For relatively
low-pressure rocks (\50 kbar), a comparison between Al-
in-Opx (Nickel and Green 1985) and Cr-in-Cpx (Nimis and
Taylor 2000) pressure estimates may further help to dis-
tinguish poorly equilibrated samples.
Systematic variations of DTNG09 (=TNG09Opx–Grt–TTA98
Cpx–Opx)
values may help reveal localized thermal or redox pertur-
bations, as well as lateral variations of oxidation state
relative to typical lithospheric mantle. Further analyses of
ferric iron in mantle-derived orthopyroxene–garnet pairs
would be highly desirable, however, to gain a better insight
into Fe3? distribution in these minerals and to obtain a
more robust calibration of the orthopyroxene–garnet ther-
mometer, which will be independent of redox conditions
and applicable to all types of orthopyroxene–garnet-bear-
ing rocks.
Acknowledgments PN acknowledges support by MIUR ex60% and
IGG-CNR (TA.P01.0004.002). Reviews by T. Stachel and an anon-
ymous reviewer helped us improve the paper. We are particularly
grateful to T. Stachel for pointing out a small bug in our early
calculations.
References
Ai Y (1992) Major and minor element systematics in the lherzolite
system: a petrological and experimental study. PhD thesis,
University of Tasmania
Ai Y (1994) A revision of the garnet–clinopyroxene Fe2?–Mg
exchange geothermometer. Contrib Mineral Petrol 115:467–473
Ashchepkov IV (2006) Empirical garnet thermobarometer for mantle
peridotites. Russian Geol Geophys 47:1071–1085
Ashchepkov IV, Pokhilenko NP, Vladykin NV, Rotman AY,
Afanasiev VP, Logvinova AM, Kostrovitsky SI, Pokhilenko
LN, Karpenko MA, Kuligin SS, Malygina EV, Stegnitsky YB,
Alymova NA, Khmelnikova OS (2008) Reconstruction of mantle
4 A spreadsheet is available at http://www.geoscienze.unipd.it/
*paolon/PTMantle.xls, which performs P–T calculations and rele-
vant checks according to the proposed scheme.
Contrib Mineral Petrol (2010) 159:411–427 425
123
sections beneath Yakutian kimberlite pipes using monomineral
thermobarometry. In: Coltorti M, Gregoire M (eds) Metasoma-
tism in Oceanic and Continental Lithospheric Mantle, vol 293.
Geol Soc London Spec Publ, pp 335–352
Berman RG, Aranovich LY (1996) Optimized standard state and
solution properties of minerals. Contrib Mineral Petrol 126:1–24
Berman RG, Aranovich LY, Pattison DR (1995) Reassessment of the
garnet–clinopyroxene Fe–Mg exchange thermometer: II. ther-
modynamic analysis. Contrib Mineral Petrol 119:30–42
Bertrand P, Mercier J-CC (1985) The mutual solubility of coexisting
ortho- and clinopyroxene: toward an absolute geothermometer
for the natural system? Earth Planet Sci Lett 76:109–122
Boyd FR (1973) A pyroxene geotherm. Geochim Cosmochim Acta
37:2533–2546
Boyd FR, Pearson DG, Hoalc KO, Hoald BG, Nixon PH, Kingstonf
MJ, Mertzman SA (2004) Garnet lherzolites from Louwrensia,
Namibia: bulk composition and P/T relations. Lithos 77:573–592
Brey GP, Kohler T (1990) Geothermobarometry in four-phase
lherzolites. II. New thermobarometers, and practical assessment
of existing thermobarometers. J Petrol 31:1353–1378
Brey GP, Kohler T, Nickel KG (1990) Geothermobarometry in four-
phase lherzolites. I. Experimental results from 10 to 60 kb.
J Petrol 31:1313–1352
Brey GP, Bulatov VK, Girnis AV (2008) Geobarometry for perido-
tites: Experiments in simple and natural systems from 6 to
10 GPa. J Petrol 49:3–24
Bulatov VK, Ryabchikov ID, Brey GP (2002) Two pyroxene–garnet
equilibria in the Na2O–CaO–MgO–Al2O3–SiO2 system at
2–5 GPa pressure. Geochem Int 40:929–942
Canil D, O’Neill HStC (1996) Distribution of ferric iron in some
upper-mantle assemblages. J Petrol 37:609–635
Carswell DA (1991) The garnet–orthopyroxene Al barometer:
problematic application to natural garnet lherzolite assemblages.
Mineral Mag 55:19–31
Carswell DA, Gibb FG (1987) Evaluation of mineral thermometers
and barometers applicable to garnet lherzolite assemblages.
Contrib Mineral Petrol 95:499–511
Carswell DA, Harley SL (1990) Mineral barometry and thermometry.
In: Carswell DA (ed) Eclogite facies rocks. Chapman & Hall,
New York, pp 83–110
Chatterjee ND (1991) Applied mineralogical thermodynamics.
Selected topics. Springer, Berlin, p 321
Creighton S (2009) A semi-empirical manganese-in-garnet single-
crystal thermometer. Lithos XX: XX-XX, doi:10.1016/j.lithos.
2009.05.011
Creighton S, Stachel T, Matveev S, Hofer H, McCammon C, Luth
RW (2009) Oxidation of the Kaapvaal lithospheric mantle driven
by metasomatism. Contrib Mineral Petrol 157:491–504
Davis BTC, Boyd FR (1966) The join Mg2Si2O6–CaMgSi2O6 at
30 kilobars pressure and its application to pyroxenes from
kimberlites. J Geophys Res 71:3567–3576
Ellis DS, Green DH (1979) An experimental study on the effect of Ca
upon garnet–clinopyroxene Fe–Mg exchange equilibria. Contrib
Mineral Petrol 71:13–22
Finnerty AA, Boyd FR (1984) Evaluation of thermobarometers for
garnet peridotites. Geochim Cosmochim Acta 48:15–27
Finnerty AA, Boyd FR (1987) Thermobarometry for garnet perido-
tites: basis for the determination of thermal and compositional
structure of the upper mantle. In: Nixon PH (ed) Mantle
xenoliths. John Wiley & Sons, London, pp 381–402
Franz L, Brey GP, Okrusch M (1996a) Reequilibration of ultramafic
xenoliths from Namibia by metasomatic processes at the mantle
boundary. J Geol 104:599–615
Franz L, Brey GP, Okrusch M (1996b) Steady-state geotherm,
thermal disturbances, and tectonic development of the lower
lithosphere underneath the Gibeon Kimberlite Province, Nami-
bia. Contrib Mineral Petrol 126:181–198
Frost DJ, McCammon CA (2008) The redox state of Earth’s mantle.
Annu Rev Earth Planet Sci 36:389–420
Ganguly J, Cheng WJ, Tirone M (1996) Thermodynamics of
aluminosilicate garnet solid solution: new experimental data,
an optimized model, and thermometric applications. Contrib
Mineral Petrol 126:137–151
Gasparik T (1984) Two-pyroxene thermobarometry with new exper-
imental data in the system CaO–MgO–Al2O3–SiO2. Contrib
Mineral Petrol 87:87–97
Griffin WL, Kaminsky FV, Ryan CG, O’Reilly SY, Win TT, Ilupin IP
(1996a) Thermal state and composition of the lithospheric
mantle beneath the Daldyn kimberlite field, Yakutia. Tectono-
physics 262:19–33
Griffin WL, Smith D, Ryan CG, O’Reilly SY, Win TT (1996b) Trace-
element zoning in mantle minerals; metasomatism and thermal
events in the upper mantle. Can Mineral 34:1179–1193
Grutter HS (2009) Pyroxene xenocryst geotherms: techniques and
application. Lithos. doi:10.1016/j.lithos.2009.03.023
Grutter H, Moore R (2003) Pyroxene geotherms revisited—an
empirical approach based on Canadian xenoliths. Extended
Abstract 8th International Kimberlite Conference no. 272 (CD-
ROM)
Grutter HS, Apter DB, Kong J (1999) Crust–mantle coupling:
evidence from mantle-derived xenocrystic garnets. In: Gurney
JJ, Gurney JL, Pascoe MD, Richardson SH (eds) Proceedings of
the 7th International Kimberlite Conference, vol 1. Red Roof
Design, Cape Town, pp 307–313
Grutter H, Gurney JJ, Menzies AH, Winter F (2004) An updated
classification scheme for mantle-derived garnet, for use by
diamond explorers. Lithos 77:841–857
Gudmundsson G, Wood BJ (1995) Experimental tests of garnet
peridotite oxygen barometry. Contrib Mineral Petrol 119:56–67
Harley SL (1984) An experimental study of the partitioning of Fe and
Mg between garnet and orthopyroxene. Contrib Mineral Petrol
86:359–373
Hearn BC Jr (2004) The homestead kimberlite, central Montana,
USA: mineralogy, xenocrysts, and upper-mantle xenoliths.
Lithos 77:473–491
Hearn BC Jr, McGee ES (1984) Garnet peridotites from Williams
kimberlites, north-central Montana, USA. In: Kornprobst J (ed)
Kimberlites II: the Mantle and crust–mantle relationships. Proc
3rd Int Kimb Conf. Elsevier, Amsterdam, pp 57–70
Holdaway MJ (2000) Application of new experimental and garnet
Margules data to the garnet–biotite geothermometer. Am
Mineral 85:881–892
Ishikawa A, Maruyama S, Komiya T (2004) Layered lithospheric
mantle beneath the Ontong Java Plateau: implications from
xenoliths in alnoite, Malaita, Solomon Islands. J Petrol 45:2011–
2044
Kennedy CS, Kennedy GC (1976) The equilibrium boundary between
graphite and diamond. J Geophys Res 81:2467–2470
Klemme S, O’Neill HStC (2000) The near-solidus transition from
garnet lherzolite to spinel lherzolite. Contrib Mineral Petrol
138:237–248
Kretz R (1982) Transfer and exchange equilibria in a portion of the
pyroxene quadrilateral as deduced from natural and experimental
data. Geochim Cosmochim Acta 46:411–421
Krogh EJ (1988) The garnet–clinopyroxene Fe–Mg geothermometer:
a reinterpretation of existing experimental data. Contrib Mineral
Petrol 99:44–48
Krogh Ravna E (2000) The garnet–clinopyroxene Fe2?–Mg geother-
mometer: an updated calibration. J Metamorphic Geol 18:211–
219
426 Contrib Mineral Petrol (2010) 159:411–427
123
Lazarov M, Woodland AB, Brey GP (2009) Thermal state and redox
conditions of the Kaapvaal mantle: a study of xenoliths from the
Finsch mine, South Africa. Lithos. doi:10.1016/j.lithos.2009.
03.035
Lee HY, Ganguly J (1988) Equilibrium compositions of coexisting
garnet and orthopyroxene: experimental determinations in the
system FeO–MgO–Al2O3–SiO2, and applications. J Petrol
29:93–113
Lindsley D, Dixon SA (1976) Coexisting diopside and enstatite at
20 kbar and 900–1,200�C. Am J Sci 276:1285–1301
McCammon C, Kopylova MG (2004) A redox profile of the Slave
mantle and oxygen fugacity control in the cratonic mantle.
Contrib Mineral Petrol 148:55–68
McCammon CA, Griffin WL, Shee SH, O’Neill HSC (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
Nickel KG, Green DH (1985) Empirical geothermobarometry for
garnet peridotites and implications for the nature of the
lithosphere, kimberlites and diamonds. Earth Planet Sci Lett
73:158–170
Nimis P (2002) The pressures and temperatures of formation of
diamond based on thermobarometry of chromian diopside
inclusions. Can Mineral 40:871–884
Nimis P, Morten L (2000) P–T evolution of ‘‘crustal’’ garnet
peridotites and included pyroxenites from Nonsberg area (Upper
Austroalpine), NE Italy: from the wedge to the slab. J Geodyn
30:93–115
Nimis P, Taylor WR (2000) Single-clinopyroxene thermobarometry
for garnet peridotites. Part I. Calibration and testing of a Cr-in-
Cpx barometer and an enstatite-in-Cpx thermometer. Contrib
Mineral Petrol 139:541–554
Nimis P, Trommsdorff V (2001a) Revised thermobarometry of Alpe
Arami and other garnet peridotites from the Central Alps.
J Petrol 42:103–115
Nimis P, Trommsdorff V (2001b) Comment to ‘‘New constraints on
the P–T evolution of the Alpe Arami garnet peridotite body
(Central Alps, Switzerland)’’ by Paquin and Altherr (2001).
J Petrol 42:1773–1779
O’Neill HStC (1980) An experimental study of Fe–Mg-partitioning
between garnet and olivine and its calibration as a geothermom-
eter: corrections. Contrib Mineral Petrol 72:337
O’Neill HStC (1981) The transition between spinel lherzolites and
garnet lherzolites, and its use as a geobarometer. Contrib Mineral
Petrol 77:185–194
O’Neill HStC, Wood BJ (1979) An experimental study of Fe–Mg-
partitioning between garnet and olivine and its calibration as a
geothermometer. Contrib Mineral Petrol 70:59–70
Pattison DRM (1994) Are reversed Fe–Mg exchange and
solid solution experiments really reversed? Am Mineral
79:938–950
Pearson DG, Boyd FR, Haggerty SE, Pasteris JD, Field SW, Nixon
PH, Pokhilenko NP (1994) The characterisation and origin of
graphite in cratonic lithospheric mantle: a petrological carbon
isotope and Raman spectroscopic study. Contrib Mineral Petrol
115:449–466
Pollack HN, Chapman DS (1977) On the regional variations of heat
flow, geotherms and lithospheric thickness. Tectonophysics
38:279–296
Powell R (1985) Regression diagnostic and robust regression in
geothermometer/geobarometer calibration: the garnet–clinopy-
roxene geothermometer revisited. J Metamorphic Geol 3:231–243
Robinson JAC, Wood BJ (1998) The depth of the spinel to garnet
transition at the peridotite solidus. Earth Planet Sci Lett
164:277–284
Ryan CG, Griffin WL, Pearson NJ (1996) Garnet geotherms:
Pressure–temperature data from Cr-pyrope garnet xenocrysts in
volcanic rocks. J Geophys Res 101:5611–5625
Simakov SK (2008) Garnet–clinopyroxene and clinopyroxene geo-
thermobarometry of deep mantle and crust eclogites and
peridotites. Lithos 106:125–136
Smith D (1999) Temperatures and pressures of mineral equilibration
in peridotite xenoliths: review, discussion, and implications. In:
Fei Y, Bertka CM, Mysen BO (eds) Mantle petrology: field
observations and high pressure experimentation: a tribute to
Francis R. (Joe) Boyd, vol 6. The Geochemical Society Spec
Publication, pp 171–188
Smith D, Boyd FR (1992) Compositional zonation in garnets in
peridotite xenoliths. Contrib Mineral Petrol 112:134–147
Solovjeva LV, Egorov KN, Dneprovskaya LV, Kharkiv AD, Popol-
itov KE (1995) The role of fO2 regime in evolution of mantle
metasomatism and diamond formation. 6th Int Kimb Conf Ext
Abstr, pp 566–568
Stachel T, Viljoen KS, McDade P, Harris JW (2004) Diamondiferous
lithospheric roots along the western margin of the Kalahari
Craton—the peridotitic inclusion suite in diamonds from Orapa
and Jwaneng. Contrib Mineral Petrol 147:32–47
Taylor WR (1998) An experimental test of some geothermometer and
geobarometer formulations for upper mantle peridotites with
application to the thermobarometry of fertile lherzolite and
garnet websterite. N Jb Min Abh 172:381–408
Walter MJ (1998) Melting of garnet peridotite and the origin of
komatiite and depleted lithosphere. J Petrol 39:29–60
Wells PR (1977) Pyroxene thermometry in simple and complex
systems. Contrib Mineral Petrol 62:129–140
Witt-Eickschen G, Seck HA (1991) Solubility of Ca and Al in
orthopyroxene from spinel peridotite: an improved version of an
empirical geothermometer. Contrib Mineral Petrol 106:431–439
Woodland AB, Koch M (2003) Variation in oxygen fugacity with
depth in the upper mantle beneath the Kaapvaal craton, Southern
Africa. Earth Planet Sci Lett 214:295–310
Woodland AB, Peltonen P (1999) Ferric iron contents of garnet and
clinopyroxene and estimated oxygen fugacities of peridotite
xenoliths from the eastern Finland kimberlite province. In: Gurney
JJ, Gurney JL, Pascoe MD, Richardson SH (eds) Proceedings of
the VIIth international kimberlite conference, The P.H. Nixon
Volume. Red Roof Design, Cape Town, pp 904–911
Wu CM, Zhao GC (2007) A recalibration of the garnet–olivine
geothermometer and a new geobarometer for garnet peridotites
and garnet–olivine–plagioclase-bearing granulites. J Metamor-
phic Geol 25:497–505
Xu Y, Lin C, Shi L (1999) The geotherm of the lithosphere beneath
Qilin, SE China: a re-appraisal and implications for P–Testimation of Fe-rich pyroxenites. Lithos 47:181–193
Zhao DG, Essene EJ, Zhang YX (1999) An oxygen barometer for
rutile-ilmenite assemblages: oxidation state of metasomatic
agents in the mantle. Earth Planet Sci Lett 166:127–137
Contrib Mineral Petrol (2010) 159:411–427 427
123