Full text of DOI:10.1007/s004100100289

Contrib Mineral Petrol ꢀ2001) 142: 244±259
DOI 10.1007/s004100100289
M. Gregoire á I. Jackson á S.Y. O'Reilly á J.Y. Cottin
Â
The lithospheric mantle beneath the Kerguelen Islands ꢀIndian Ocean):
petrological and petrophysical characteristics of mantle ma®c
rock types and correlation with seismic pro®les
Received: 14 August 2000 / Accepted: 22 May 2001 / Published online: 21 July 2001
Ó Springer-Verlag 2001
Abstract Deep-seated meta-igneous xenoliths brought measured VP agree well for rock types with micro-
to the surface by alkali basaltic magmas from the structures not complicated by kelyphitic breakdown of
Kerguelen Islands reveal that basaltic magmas have garnet and/or pervasive grain-boundary cracking.
intruded the upper mantle throughout their geological Pyroxene granulites have measured and calculated VP
evolution. These xenoliths record volcanic activity as- within the range 7.37±7.52 km/s; calculated velocities for
sociated with their early South East Indian Ridge loca- the garnet granulites and pyroxenites range from 7.69 to
tion and subsequent translation to an intraplate setting 7.99 km/s, whereas measured and calculated V for a
P
over the Kerguelen Plume. The meta-igneous xenoliths mantle harzburgite are 8.45 and 8.29 km/s respectively.
sample two distinctive geochemical episodes: one is The seismic structure observed beneath the Kerguelen
tholeiitic transitional and one is alkali basaltic. Geo- Islands can be explained by ꢀ1) a mixture of underplated
thermobarometry calculations provide a spatial context pyroxene granulites and ultrama®c rocks responsible for
for the rock type sequence sampled and for interpreting the 2±3 km low velocity transitional zone below the
petrophysical data. The garnet granulites equilibrated oceanic layer 3, ꢀ2) varying proportions of granulites
over a pressure range of 1.15 to 1.35 GPa and the garnet and pyroxenites in di€erent regions within the upper
pyroxenite at 1.8 GPa. Ultrasonic measurements of mantle producing the lateral heterogeneities, and ꢀ3)
compressional wave speed V have been carried out at intercalation of the granulites and pyroxenites
P
pressures up to 1 GPa, and densities measured for rep- throughout the entire upper mantle column, along with
resentative samples of meta-igneous xenoliths and for a elevated temperatures, accounting for the relatively low
harzburgite that represents the peridotitic mantle. VP mantle velocities ꢀ7.70±7.95 km/s).
and density have also been calculated using modal
proportions of minerals and appropriate elastic
properties for the constituent minerals. Calculated and
Introduction
M. Gregoire ꢀ&) á S.Y. O'Reilly
Â
Ultrama®c and ma®c xenoliths provide unique and
direct information about the nature, composition,
structure and evolution of the lower crust, upper mantle
and crust±mantle transition zone ꢀe.g. Frey and Prinz
1978; Grin et al. 1984; O'Reilly and Grin 1987;
GEMOC ARC National Key Centre,
Department of Earth and Planetary Sciences,
Macquarie University, NSW 2109, Australia
E-mail: mgregoire@geology.uct.ac.za
I. Jackson
Research School of Earth Sciences,
Australian National University,
Canberra, ACT 0200, Australia
Â
O'Reilly et al. 1990; Gregoire et al. 1998). Seismic pro-
les give indirect information on the properties of the
lower crustal and upper mantle rocks in situ ꢀe.g. Fin-
layson and Leven 1987; Recq et al. 1990). Laboratory
measurement of the petrophysical parameters, com-
®
J.Y. Cottin
Department of Geology-UMR 6524,
University J. Monnet, 23 rue P. Michelon,
pressional wave velocity ꢀV ) and density ꢀq) of the xe-
4
2023 St-Etienne, France
P
noliths allows comparison between these di€erent
sources of information on the deep part of the litho-
sphere and provides constraints for the interpretation of
the seismic data in a geologically meaningful way ꢀe.g.
O'Reilly et al. 1990; Rudnick and Fountain 1995;
Rudnick and Jackson 1995).
Present address: M. Gregoire
Â
Department of Geological Sciences,
University of Cape Town, Rondebosch,
7701 South Africa
Tel.: +27-21-6502909, Fax: +27-21-6503783
Editorial responsibility: J. Hoefs

2
45
Seismic refraction studies ꢀRecq et al. 1990, 1994;
Charvis et al. 1995) show evidence for thickened oce-
anic crust in the Kerguelen archipelago. They indicate
that the oceanic layer 2 ꢀVp: 4.6±4.8 km/s) is 8 to 9 km
thick with oceanic layer 3 ꢀV : 6.6±7.0 km/s) about 5 to
P
10 km thick and extending from a depth of about 10 km
to a base at 15±20 km in di€erent parts of the seismic
pro®les. Beneath layer 3, there is a seismically distinct
zone 2±3 km thick that shows a range in V increasing
P
from about 7.2 to 7.5 km/s with depth; these values are
signi®cantly lower than the usual V at this depth in
P
the oceanic lithosphere. This zone has been interpreted
as a transitional zone representing a layered crust±
mantle boundary ꢀe.g. Recq et al. 1990; Gre
998).
The seismic refraction data ꢀRecq et al. 1990; Charvis
Â
goire et al.
1
et al. 1995) also show that the P-wave velocities within
the upper mantle are relatively low ꢀ7.70±7.95 km/s) and
that there are lateral variations in V within the lower
P
crust and the upper mantle. These have been interpreted
previously as lithological heterogeneities due to the
presence of serpentinised mantle peridotite regions or to
magmatic underplating processes.
Gregoire et al. ꢀ1998) identi®ed the role of deep-
Â
seated magmatic additions to the lithosphere beneath
Kerguelen as an important part of crustal growth in
this region and a possible mechanism of continental
nucleation. This study builds on that petrological work
and focuses on interpreting the new petrophysical ꢀVP,
q) data using petrological and geochemical ꢀmajor and
trace element) information for both bulk rocks and
constituent minerals for a representative subset of the
meta-igneous xenoliths [termed Type II in accordance
with the terminology of Frey and Prinz ꢀ1978)] from
the lithospheric mantle beneath the Kerguelen Islands.
The xenoliths are frozen basaltic melts and cumulates
intruded into the upper mantle and subsequently
equilibrated to the temperature and pressure condi-
tions of granulite facies. They represent crystallisation
products from the magmas produced in multiple
melting episodes beneath the Kerguelen plateau ꢀWeis
et al. 1993; Yang et al. 1998) and record the intrusion
of successive sills and lenses of basaltic melts spanning
the long and complex volcanic history of this region
ꢀ
Gregoire et al. 1998). These xenoliths provide geo-
Â
chemical evidence to distinguish the di€erent contri-
butions of magmatism associated with both mid-ocean
ridge and plume activity. Their formation is therefore
an integral part of the geological evolution of the ar-
chipelago and especially re¯ects the vigorous activity
of the Kerguelen mantle plume. Because we can put
these rock types into a spatial context using geother-
mobarometry, we can assess the nature of construction
and the architecture of the plateau at depths from
about 10 to 50 km ꢀsampled by the xenoliths), and
integrate the geological data from these xenoliths with
their petrophysical characteristics to provide realistic
constraints to interpret the deep geophysical ꢀmainly
seismic) data for the Kerguelen Islands. The Type I

2
46
xenolith, a harzburgite ꢀsample GM92-453, described
in detail in Gregoire et al. 2000) was included in this
study in order to assess the petrophysical characteris-
tics of the peridotitic upper mantle wall-rock beneath
Ultrama®c and ma®c xenoliths from the Kerguelen Islands
occur in dykes, lava-¯ows and breccia pipes of the youngest and
more alkaline basaltic rocks ꢀGregoire et al. 1994, 1997, 1998).
Â
The mantle and meta-igneous xenoliths studied here were col-
lected from four di€erent localities of the Kerguelen archipelago
Â
the Kerguelen archipelago. Previous studies ꢀe.g. ꢀFig. 1 and Table 1). They have sub-rounded shapes and range
from 10 to 30 cm in diameter. They correspond to Type I and
Type II Kerguelen xenoliths de®ned by Gregoire et al. ꢀ1997,
998, 2000).
Gregoire et al. 1997, 2000) have detailed the petrology
Â
and composition of the Kerguelen peridotitic upper
mantle.
Â
1
Geological setting
Sampling and analytical methods
The Kerguelen Islands are located in the oceanic domain of the Six xenoliths large enough for laboratory determination of acoustic
Antarctic plate and represent the thickest section of the Kerguelen velocity were chosen to represent the di€erent types of meta-igne-
oceanic plateau, which is the second largest oceanic plateau ous xenoliths identi®ed from Kerguelen.
3
ꢀ
1
25´106 km ) after the Ontong Java plateau ꢀCon and Eldhom
Samples were taken from the central parts of the xenoliths and
993). The Kerguelen Islands have evolved from a location near ground in an agate mill. Major and minor elements ꢀCr, Ni) in bulk
the SEIR ꢀSouth East Indian Ridge) to a present-day intraplate rocks were analysed by X-ray ¯uorescence spectrometry ꢀXRF) at
setting and the magmatic activity has extended over 45 m.y. ꢀe.g. Macquarie University ꢀsee O'Reilly and Grin 1988 for methods).
Giret 1993). Their geodynamic evolution records a progressive The concentration of 30 minor and trace elements in bulk rocks
change in composition of basaltic magmas from tholeiitic to ꢀREE, Ba, Cs, Rb, Th, U, Nb, Ta, Pb, Sr, Zr, Ti, Y, Sc, V, Co, Cu
alkaline ꢀGautier et al. 1990; Weis et al. 1993). Therefore, the and Zn) were analysed using a Perkin-Elmer Sciex ELAN 6000
Kerguelen Islands present a unique geological setting combining ICP-MS instrument at Macquarie University. Mineral major and
characteristics of both the Iceland and Hawaiian regions ꢀGiret minor element compositions were determined by a Cameca
et al. 1997). This is an unusual oceanic tectonic environment ꢀand Camebax SX 50 microprobe at Macquarie University using a
possibly analogous to that of the smaller Agulhas plateau; Uen- wavelength-dispersive spectrometric ꢀWDS) technique. The
zelmann-Neben et al. 1999; Gohl and Uenzelmann-Neben 2001) microprobe was used with 15 kVaccelerating voltage, sample
but it may be relevant to an important mechanism for the current of 20 nA, a beam diameter of 2±3 lm, and natural and
nucleation and growth of continental nuclei ꢀGre
Â
goire et al. 1995, synthetic minerals as standards. Count times were 20±40 s and no
1
998).
values are reported below detection limits ꢀ0.01±0.04 wt%).
Fig. 1 Location of ultrama®c
and ma®c xenolith-bearing
alkali basalts of the Kerguelen
Islands ꢀmodi®ed after Gregoire
Â
et al. 1997, 2000). a Ice caps;
b moraines; c alkaline silica-
oversaturated volcano-plutonic
complexes; d alkaline silica-
undersaturated volcano-
plutonic complexes; e tholeiitic±
transitional plutonic complexes;
f ¯ood basalts of transitional to
alkaline type. Ultrama®c and
ma®c xenolith-outcrops shown
as solid squares; numbered open
squares refer to sample locality
ꢀsee Table 1 for the naming of
each outcrop). Inset shows
location of Kerguelen Islands,
South West Indian Ridge
ꢀSWIR) and the South East
Indian Ridge ꢀSEIR)

2
47
Fig. 2 Photomicrographs
of the Kerguelen xenoliths.
A Granoblastic-mosaic tex-
ture ꢀsamp1e 0B93-57, ´25).
B Garnet and sapphirine
coronites around spinel
ꢀ
sample GM92-347, ´50).
C Garnet granulites of Type
IIc ꢀsample GM92-390)
showing signi®cant replace-
ment of garnet by kelyphite
ꢀ
´25). D Poikilitic spinel-
bearing harzburgite ꢀsample
GM92-453). Clinopyroxene
encloses spinel grains ꢀ´50)
plagioclase are garnet+clinopyroxene2 in samples
GM92-390 and 394, garnet+clinopyroxene2+sapphi-
rine in sample GM92-347 and clinopyroxene2+sapphi-
Petrography
The samples studied here in detail are all Type II
ꢀ
rine in sample GM92-412 ꢀGregoire et al. 1994, 1998).
Â
Frey and Prinz 1978) and comprise two-pyroxene
granulites ꢀsamples GM92-412 and OB93-57), garnet
granulites ꢀsamples GM92-347, GM92-390 and GM92-
The garnet and sapphirine coronites and symplectites
are especially abundant in sample GM92-347 ꢀFig. 2).
The rare secondary amphiboles occurring in the Type
IIb and IIc xenoliths are derived by metasomatic reac-
tion of pre-existing clinopyroxenes.
3
94) and a garnet pyroxenite ꢀGM92-165), as sum-
marised in Table 1. The xenoliths can be grouped into
two main geochemical types as detailed in Gregoire
et al. ꢀ1998).
Â
The grain size in the Type II xenoliths commonly
ranges from 0.1 to 2±3 mm but goes up to 5 mm in
clinopyroxenite GM92-165. Two garnet-bearing granu-
lites ꢀGM92-390 and 394; Fig. 2) show signi®cant
replacement of garnet by kelyphite ꢀ20 to 80% from one
garnet to another), similar to that described by Grin
et al. ꢀ1987) and Rudnick and Jackson ꢀ1995) for the
garnet granulite xenoliths from the Central and Chud-
leigh volcanic provinces of north Queensland ꢀAustralia).
The clinopyroxenite GM92-165 displays a large amount
of grain boundary alteration and hematite staining.
The Type I harzburgite chosen for this study ꢀsample
GM92-453) is a poikilitic harzburgite consisting of
olivine ꢀ84 wt%), opx ꢀ13 wt%), cpx ꢀ2 wt%) and spinel
The ®rst group has basaltic tholeiitic±transitional
characteristics and includes Type IIa ꢀe.g. samples GM92-
4
12, OB93-57, GM92-347), which are cumulates and
Type IIc ꢀe.g. sample GM92-390), which are frozen melts
Gregoire et al. 1998). Sample GM92-394 shows petro-
ꢀ
Â
graphic characteristics that are intermediate between
those of Type IIa and Type IIc and is designated IIa/IIc.
The second group has alkali basaltic anities and is
represented by the garnet clinopyroxenite ꢀsample GM92-
Â
165) and designated Type IIb ꢀGregoire et al. 1998).
The main microstructures of Type II xenoliths are
allotriomorphic±granular heterogranular, but some are
hypidiomorphic±granular. All the samples show local
recrystallisation to granoblastic-mosaic textures ꢀFig. 2).
Most of the granulite samples ꢀGM92-412, GM92-347,
GM92-390 and GM92-394) have evidence of subsolidus
re-equilibration characterised by the widespread devel-
opment of coronitic and symplectitic mineral parage-
neses. The coronitic and symplectitic minerals at the
grain boundaries between pyroxenes, spinel and
ꢀ
1 wt%). The poikilitic microstructure is similar to that
described in some harzburgitic xenoliths from the
French Massif Central ꢀCoisy and Nicolas 1978) and is
characterised by the occurrence of large olivine grains
ꢀ
up to 5 cm) enclosing orthopyroxene inclusions and by
the habit of clinopyroxene ꢀFig. 2), which encloses
olivine, orthopyroxene and spinel crystals ꢀGregoire et
al. 1997, 2000).
Â

2
48
Table 2 Bulk rock major ꢀwt%) and trace element ꢀppm) abundances and calculated modal compositions of Type II xenoliths from the
Kerguelen Islands
Sample
Type IIa
GM92-412
Type IIa
OB93-57
Type IIa
GM92-347
Type IIa/IIc
GM92-394
Type IIb
GM92-165
Type IIc
GM92-390
SiO
TiO
Al
FeO
MnO
MgO
CaO
2
44.9
0.06
25.66
3.96
0.05
10.31
13.16
1.45
0.08
0.02
49.3
0.35
16.95
5.81
0.10
10.33
14.48
1.85
0.05
0.02
44.8
0.04
23.35
5.25
0.07
13.30
11.85
1.06
0.07
0.02
49.1
0.83
16.80
6.71
0.12
11.10
11.57
2.69
0.16
0.04
46.2
1.95
10.46
9.23
0.16
12.42
16.91
1.65
0.12
0.04
45.3
2.26
14.25
12.91
0.24
2
2
O
3
8.41
12.58
2.42
0.14
0.05
1.13
0.25
0.20
100.1
2
Na O
K
2
O
P
2
O
5
H
H
2
O+
O-
0.37
0.15
0.07
0.62
0.12
0.13
0.57
0.16
0.03
1.31
0.31
0.05
0.69
0.06
0.11
2
CO
Sum
2
100.3
100.1
100.5
100.7
100.0
Mg number
Sc
V22.6
Cr ꢀXRF)
Co
82.27
4.15
185
730
43
76.03
40.1
9.8
480
43
81.88
4.30
250
505
52
74.66
36.3
375
840
42
70.57
52.3
400
270
47
53.74
49.8
300
52
Ni ꢀXRF)
Cu
Zn
Rb
Ba
Sr
Pb
Th
U
Nb
Ta
Ti
Zr
La
Ce
Pr
Nd
Sm
Eu
Gd
Dy
Ho
Er
Yb
Lu
Y
300
170
430
250
21.7
35
5.32
47.4
224
135
26.6
56
90
56.6
160
6.79
21.5
149
0.307
0.053
0.024
2.04
0.146
12100
42
15.0
15.5
1.47
26.6
274
0.100
0.016
0.003
0.114
0.007
521
1.65
0.47
0.830
0.122
0.57
0.15
0.145
0.18
0.174
0.037
0.11
0.09
0.01
1.18
4.00
30.8
0.49
10.7
164
5.95
21.3
2.77
17.4
156
0.220
0.057
0.032
0.514
0.038
308
2.15
0.86
1.528
0.187
0.71
0.14
0.158
0.14
0.133
0.028
0.08
0.08
0.01
0.83
3.70
18.0
120
0.770
1.64
0.640
5.09
0.370
12500
81
0.131
0.020
0.003
0.103
0.014
2200
7.01
0.52
1.57
0.294
1.71
0.72
0.424
0.96
1.178
0.255
0.70
0.62
0.09
6.93
0.231
0.176
0.101
1.50
0.094
4900
17.4
2.05
4.68
0.651
3.46
1.41
0.719
2.0
2.325
0.496
1.37
1.17
0.17
13.5
7.4
1.32
5.6
1.34
8.5
3.32
1.239
4.2
4.948
1.049
2.97
2.54
0.37
28.3
18.9
3.13
16.3
5.0
1.750
5.5
4.920
0.960
2.45
1.86
0.25
26.7
The two sapphirine-bearing Type IIa granulite xe-
noliths ꢀGM92-412 and GM92-347) have high mg
numbers and Al O contents and extend the Ma®c 1
Major and trace element compositions
2
3
Whole rock composition
®eld de®ned by Kempton and Harmon ꢀ1992) for
granulite xenoliths worldwide to lower SiO /Al O val-
2
2
3
Major elements and transition trace elements ꢀTTE)
ues ꢀFig. 3). Three other Kerguelen xenoliths ꢀType IIa
granulite OB93-57, Type IIa/IIc granulite GM92-394
and Type IIb clinopyroxenite GM92-165) overlap the
All of the studied Type II xenoliths have ma®c composi-
tions, with SiO ranging from 44.7 to 49.3 wt% and MgO
2
®eld of Ma®c 1 granulites, but the Type IIc sample
contents varying from 8.4 to 13.3 wt% ꢀTable 2). The
Types IIb ꢀalkaline) and IIc ꢀtholeiitic±transitional melt)
xenoliths are higher in TiO than the Type IIa ꢀtholeiitic±
ꢀ
ꢀ
GM92-390) plots in the `primitive' basaltic magma ®eld
Fig. 3). The Type IIa granulites and the intermediate
2
IIa/IIc sample ꢀGM92-394) have higher Cr ꢀ480±
40 ppm) and Ni ꢀ170±430 ppm) contents than the Type
IIb clinopyroxenite ꢀCr 272 ppm and Ni 135 ppm) and
transitional cumulate) xenoliths. Sample GM92-394 ꢀIIa/
IIc) has an intermediate TiO content ꢀTable 2).
8
2

2
49
Fig. 3 Mg number vs.
SiO /Al diagram for
2
2 3
O
Kerguelen granulite and
pyroxenite xenoliths com-
pared with ®elds for granu-
lite xenoliths worldwide.
De®nition of ®elds from
Kempton and Harmon
ꢀ
1992). Ma®c 1 ma®c gran-
ulites ꢀi.e. SiO <54 wt%,
mg number<0.6; intermedi-
ate SiO >54 wt% and
66 wt%; silicic
SiO >66 wt%. Note that,
2
2
<
2
for this diagram only, mg
number is de®ned as Mg/
ꢀMg+0.85 FeOtot), to be
consistent with Kempton
and Harmon ꢀ1992). Unbro-
ken arrows show generalised
direction of compositional
change with accumulation of
indicated mineral phase
the Type IIc granulite ꢀCr 302 ppm and Ni 91 ppm) but
lower Cu, Zn, Sc and Vcontents ꢀTable 2). The Type IIa
granulite GM92-347 is especially low in Sc ꢀ4.30 ppm)
and Vꢀ9.80 ppm) and high in Ni ꢀ430 ppm) in com-
parison to the other samples ꢀTable 2).
Other trace elements ꢀREE,Y,HFSE,LILE)
The Type IIa ꢀtholeiitic±transitional cumulate) granulite
xenoliths have low REE contents ꢀless than three times
primitive mantle values) and have consistent positive Eu
anomalies, especially pronounced in the two sapphirine-
bearing samples which are also characterised by a
LREE/HREE ratio >1 ꢀFig. 4). The Type IIb ꢀalkaline)
clinopyroxenite and the Type IIc ꢀtholeiitic±transitional
melt) granulite are higher in total REE than Type IIa
Fig. 4 Primitive mantle-normalised REE patterns for whole rock
of Kerguelen granulite and pyroxenite xenoliths. Normalising
values after McDonough and Sun ꢀ1995)
ꢀ
tholeiitic±transitional cumulate). Type IIb shows slight
LREE enrichment while Type IIc is characterised by a Mineral compositions
at pattern from Pr to Lu with depletion in La and Ce.
¯
The Type IIa/IIc sample ꢀGM92-394) displays a similar The clinopyroxenes of Type II xenoliths ꢀTable 3) are
REE pattern to those of the Type IIa sample OB93-57, aluminous diopsides ꢀAl O : 6.65±7.80 wt%) poor in
but its REE content is higher, its positive Eu anomaly is chromium ꢀCr O <0.20 wt%). Mg numbers [100 Mg/
2
3
2
3
smaller and it plots between the Type IIa and the Type ꢀMg+Fe)] decrease from the Type IIa and IIa/IIc
IIc xenoliths ꢀFig. 4).
granulites ꢀ78.5±89.8) to Type IIb clinopyroxenite ꢀ73.8)
All the studied Type II xenoliths have negative Zr and to the Type IIc granulite ꢀ69.2).
anomalies ꢀFig. 5). The high Sr content of the Type
IIa and IIa/IIc granulites results in the large positive tite ꢀAl O : 3.50±7.90 wt%). It shows a similar range in
The orthopyroxene ꢀTable 3) is an aluminous ensta-
2
3
Sr anomalies in their incompatible trace element pat- its mg number to those of the clinopyroxenes from Type
terns ꢀFig. 5). These patterns have very similar shapes IIa and IIa/IIc granulites ꢀ74.5±88.5) through to the
for the elements from Ce to Lu, but the two garnet- Type IIb clinopyroxenite ꢀ73) and to the Type IIc
bearing rocks are higher in Th, U, Nb, Ta and La granulite ꢀ70).
than the two garnet-free samples. The Type IIb
Plagioclase ꢀTables 3 and 4) is very high in CaO in the
clinopyroxenite trace element pattern is characterised three Type IIa granulites ꢀAn 68±80; samples OB93-57,
by positive Th and U anomalies and negative Pb and GM92-412 and GM92-347). The three other samples
Sr anomalies. The Type IIc granulite has negative Th, ꢀGM92-165, GM92-390 and GM92-394) have more
U, and Pb anomalies.
sodic plagioclase ꢀAn 34±42).

2
50
TiO2 ꢀ0.4±0.45 wt%) than those of Type IIa
<0.10 wt%) ꢀTable 3).
As already pointed out by Gre
ꢀ
Â
goire et al. ꢀ1998), il-
menite ꢀTable 3) in the Type IIb clinopyroxenite is
higher in TiO ꢀ51.75 wt%) and MgO ꢀ4.5 wt%) than
2
ilmenite in the Type IIc granulite ꢀTiO : 41.15 wt% and
2
MgO: 2.9 wt%).
Amphiboles ꢀTable 3) of Type IIb and Type IIc xe-
noliths have similar mg numbers ꢀ63 and 63.35 respec-
tively) but the Type IIb amphibole is slightly higher in
TiO ꢀ3.75 wt%), Al O ꢀ14.5 wt%) and K O ꢀ0.95 wt%)
2
2
3
2
than those of Type IIc ꢀTiO : 3 wt%, Al O : 13.6 wt%
2
2
3
and K O: 0.20 wt%).
2
Sapphirine ꢀTable 3) in the garnet-bearing Type IIa
granulite ꢀGM92-347) is more magnesian ꢀmg number:
89), higher in Al O ꢀ62.70 wt%) and lower in SiO
2 3 2
ꢀ
granulite ꢀGM92-412, mg number: 83; Al O : 59 wt%
13.20 wt%) than its equivalent in garnet-free Type IIa
2
3
and SiO : 14.30 wt%).
2
Equilibration temperatures and pressures
Equilibration temperatures of the xenoliths studied
were estimated using two-pyroxene geothermometers
ꢀ
È
Wells 1977; Brey and Kohler 1990) and clinopyrox-
ene±garnet equilibria ꢀEllis and Green 1979) in the
Kerguelen xenoliths. Equilibration pressures were esti-
mated using orthopyroxene±garnet equilibria ꢀHarley
1984). Core compositions of the primary large grains
and FeOꢀtotal) have been used for the calculations. Our
goal was to establish temperature and pressure ranges,
using a consistent methodology, and these procedures
are consistent with the protocols recommended in Xu et
al. ꢀ1998). The chosen geothermometers and geoba-
rometers are appropriate for the bulk compositions and
relevant pressure and temperature ranges and have
been shown by Xu et al. ꢀ1998) to give consistent re-
sults for the di€erent bulk compositions of peridotites
with adjacent pyroxenites or granulites in composite
xenoliths. The results are given in Table 5. The two
garnet-free granulitic samples, GM92-412 and OB93-
5
7, give temperatures ranging from 800 to 900 °C and
according to their mineral assemblages have been
equilibrated in the `pyroxene granulite stability ®eld',
i.e. within the pressure range from 0.5±0.8 to 1.1 GPa
Fig. 5 Primitive mantle-normalised trace element patterns for
whole rock from Kerguelen granulite and pyroxenite xenoliths.
Normalising values after McDonough and Sun ꢀ1995)
Â
ꢀGasparik 1984; Wood and Holloway 1984 ; Gregoire
et al. 1994, 1998). However, we can infer that these
The chemical composition of Type II Kerguelen xenoliths equilibrated at the higher pressure end of this
garnets is controlled by Mg/Fe substitution, as shown range by comparison with the calculated equilibration
by decreasing pyrope content with increasing alman- pressures of garnet-bearing xenoliths with similar
dine content from Type IIa sapphirine-bearing granu- temperatures ꢀsee below).
lite to the Type IIa/IIc granulite, to Type IIb
clinopyroxenite and to Type IIc granulite ꢀTables 3 and net-bearing granulites and the clinopyroxenite give 1.15±
). The proportion of the grossular component does 1.20 GPa at 800±900 °C ꢀGM92-390 and GM92-165),
not vary.
Geothermobarometry calculations for the four gar-
4
1.35 GPa at 800±900 °C ꢀGM92-347) and 1.80 GPa at
Most Spinels are pleonastes ꢀTable 3) with a low 925±1000 °C ꢀGM92-394). These are within the experi-
Cr O content ꢀmg number: 55±77.5 and cr number: mentally determined `garnet granulite stability ®eld'
0.3±2.3). Type IIb and Type IIa/IIc spinels are higher in ꢀIrving 1974; Gasparik 1984).
2
3

2
51

2
52
Table 4 Calculated end member compositions for the constituent minerals of Type I harzburgite and Type II granulite and pyroxenite
xenoliths
Tholeiitic±transitional basalt anity
Type II pyroxene granulites Type II garnet two-pyroxene granulites
Type IIb
Type I
GM92-453
Type I
harzburgite
Peridotite
mantle wall-rock
GM92-165
Type II garnet
pyroxenite
Alkali basalt
anity
Type IIa
OB93-57a
Type IIa
GM92-412
Type IIa
GM92-347
Type IIa/IIc
GM92-394
Type IIc
GM92-390
Clinopyroxene
40.50%
13.50
7.70
23%
13.90
3.20
3.60
69.16
10.14
25%
12.30
2.10
5.50
73.73
6.37
40%
11.70
3.50
10.00
63.51
11.29
46.50%
14.00
11.90
2.00
57.03
15.07
81.50%
15.60
9.10
2.60
59.43
13.27
2%
8.50
10.70
4.50
67.16
9.14
CaTs
Ac
Jd
Di
Fs
±
67.83
10.97
Orthopyroxene
18%
18%
2%
7.50%
9.59
70.30
20.12
1%
5.50%
12.23
62.13
25.64
13%
CaTs
En
Fs
14.02
58.88
27.10
18.21
56.10
25.69
10.86
77.72
11.42
10.40
61.25
28.35
10.58
76.92
12.50
Plagioclase
Xan
Xab
40%
0.68
0.32
0.01
49%
0.80
0.20
±
26%
0.72
0.26
0.02
27.50%
0.42
0.57
18%
0.41
0.58
±
3%
0.34
0.62
0.03
Xor
0.01
Spinel
1.50%
0.38
0.62
7%
0.28
0.72
2.50%
0.25
0.75
1.50%
0.35
0.65
±
4%
0.37
0.63
1%
Hercynite
Pleonaste
Picochromite
Chromite
31.86
21.98
27.14
18.72
Garnet
Xalm
Xpyr
Xgro
36%
0.18
0.66
0.15
23%
0.28
0.57
0.14
30%
0.40
0.42
0.16
1%
0.36
0.45
0.18
Ilmenite
Ilmenite
Geikielite
Hematite
±
±
±
±
4.50%
0.66
0.08
1.50%
0.78
0.18
0.26
0.04
Olivine
Fo
Fa
84%
89.1
10.9
Other phases
Sa: 3%
±
Sa: 8.5%
Ru: 0.5%
Am: 3.5%
The ultrasonic velocity measurements were per-
formed in a Harwood pressure vessel at the petrophysics
laboratory of the Research School of Earth Sciences,
Australian National University. The vessel operates to
Petrophysical characteristics
Measured density ꢀq) and P-wave velocity ꢀVp)
1
GPa con®ning pressure at room temperature. The
Three orthogonal cores, 15±35 mm in length and 15 mm
in diameter, were prepared from each of the selected
xenoliths except sample GM92-165, which has only one
core due to its small size. Most samples do not show
shape preferred mineral orientation ꢀSPO), but, where
observed ꢀsample GM92-412), the cores were cut parallel
and orthogonal to it. The ends of the cores were ground
cores ꢀon which densities have been measured) were
jacketed in copper sleeves sealed with O-rings against
hardened steel end pieces in order to exclude the petro-
leum spirit pressure medium. After each successful run,
the steel end pieces were removed and the sample ex-
amined to ensure that ¯uid had not penetrated into the
pore space of the specimen. Detailed descriptions of the
ultrasonic technique have been given by Jackson and
Arculus ꢀ1984), O'Reilly et al. ꢀ1990) and Rudnick and
Jackson ꢀ1995). Rock velocities were calculated from the
zero-pressure core length and measured high-pressure
travel time of the pulse. Estimated uncertainties of the
measurements are 1% for compressional wave velocity
¯
at and parallel within ‹0.01 mm. Densities were mea-
sured by immersion in ethanol. The signi®cance of the
measured densities depends upon the degree of inter-
connectivity of the network of cracks and pores. Isolated
ꢀ
i.e. non-interconnected) porosity within a specimen will
be re¯ected in reduced bulk density, whereas a `grain'
density closely comparable with that calculated from the
modal mineralogy ꢀbelow) would be determined for a
specimen with fully interconnected pore space.
3
and 0.01 g/cm for density.
The experimental results of the measurements of the
compressional wave velocities for each core of each

2
53
Table 5 Calculated pressures and temperatures of equilibration for Type II granulite and pyroxenite xenoliths from the Kerguelen
Islands. In Brey and Kohler column ꢀ0.8 Gpa) is assumed P
P ꢀGpa)
Harley ꢀ1984)
T ꢀ °C)
Wells ꢀ1977)
T ꢀ °C)
Brey and Kohler ꢀ1990)
T ꢀ °C)
Ellis and Green ꢀ1979)
Tholeiitic±transitional anity
Two-pyroxene granulites
OB93-57a Type IIa
GM92-412 Type IIa
Garnet two-pyroxene granulites
±
±
862
900
800 ꢀ0.8)
890 ꢀ0.8)
±
±
GM92-347
GM92-394
GM92-390
Type IIa
Type IIa/IIc
Type IIc
1.35
1.80
1.15
865
925
815
800
930
890
850
1000
860
Alkali basalt anity
Garnet pyroxenite
GM92-165
Mantle harzburgite
GM92-453
Type II
Type I
1.20
±
850
±
890
900
±
1125
Fig. 6 Velocity vs. pressure
for the three orthogonal
cores ꢀX, Y and Z) of
Kerguelen granulite and
pyroxenite xenoliths. For
sample GM92-165 only one
core was made because of
the small size and disaggre-
gated state of this rock
sample are shown as functions of pressure to 0.9 or granulites, regardless of whether or not they contain
3
1
GPa in Fig. 6. The harzburgite ꢀGM92-453) and the garnet ꢀ3.04±3.25 g/cm ).
garnet clinopyroxenite ꢀGM92-165) have higher densi- Three granulites ꢀGM92-412, GM92-347, GM92-394)
ties ꢀ3.30 and 3.37 g/cm respectively) than the ma®c show a marked increase in VP with increasing P to
3

2
54
400 MPa ꢀFig. 6) ± presumably re¯ecting the suppres- method of Jackson et al. ꢀ1990) using the modal pro-
sion of a substantial amount of crack porosity ± with portions of the constituent minerals olivine, orthopy-
milder pressure sensitivity at higher pressures. This be- roxene, clinopyroxene, plagioclase, spinel, garnet and
haviour is typical of measurements on moderately ilmenite, along with the appropriate single-crystal
coarse-grained rock specimens, especially xenoliths ꢀe.g. elasticity and density data from the literature ꢀe.g. Ta-
O'Reilly et al. 1990; Rudnick and Jackson 1995). The ble 7). The modes ꢀTable 4) were calculated by a least-
densities ꢀq) and the compressional ꢀVp) wave velocities squares mixing program using the major element com-
measured at 0.9 GPa are tabulated in Table 6. Specimens positions of minerals and bulk rock composition. We
GM92-453 and OB93-57 yield relatively high velocities did not include amphibole and sapphirine because
at 0.9±1.0 GPa but without the much stronger pressure amphibole has a very low modal abundance and there
sensitivity at lower pressures ± suggesting an unusual are no elasticity and density data for sapphirine. The
degree of mechanical competence in these specimens modes used for calculations were adjusted for the re-
such that much of the crack porosity is closed at pres- moval of sapphirine.
sures lower than 100±200 MPa. Specimens GM92-165
and GM92-390 also display little pressure sensitivity, but modelled as proportional molar averages of the end
at much lower velocities. member properties ꢀTable 4). This time-honoured
The elastic moduli for mineral solid solutions are
Comparison of the velocities measured on the approach has been documented previously ꢀe.g. Leitner
orthogonal cores prepared from each xenolith reveals et al. 1980) and has recently been shown to reproduce
appreciable elastic anisotropy ꢀi.e. >1%) persisting to accurately the measured elastic moduli for omphacitic
ꢀ
1 GPa only for specimens GM92-394 ꢀ1.4%), OB93-57 pyroxene ꢀBhagat et al. 1992). Rarely, signi®cant
2.5%) and GM92-412 ꢀ3.5%). For this last specimen, a deviations from the molar average are observed and may
ꢀ
weak shape preferred orientation ꢀSPO) is evident in be attributable to cation ordering e€ects ꢀBass and
hand specimen and thin section ꢀsee the petrographic Weidner 1984).
description).
The plagioclase and alkali feldspar end member
The two garnet-free pyroxene granulites ꢀOB93-57 and properties of Jackson et al. ꢀ1990) are here replaced with
GM92-412), both of tholeiitic±transitional geochemical those tabulated by Bass ꢀ1995) for albite, anorthite and
anity, have mean V at 0.9 GPa of 7.37 and 7.52 km/s microcline. We also add the properties for ilmenite given
P
respectively. The two garnet granulites GM92-390 and by Liebermann ꢀ1976) and Bass ꢀ1995). In addition there
GM92-394 ꢀalso tholeiitic±transitional) have mean Vp of has been a dramatic expansion during the past two de-
7.08 and 7.06 km/s respectively. The more magnesian cades of the single-crystal elasticity data set for pyrox-
tholeiitic±transitional garnet-bearing granulite contain- enes resulting from the use of opto-acoustic techniques
ing sapphirine ꢀGM92-347) has the highest mean Vp of all on microcrystals ꢀe.g. Kandelin and Weidner 1988; Chai
the granulites ꢀ7.90 km/s at 0.9 GPa). The mantle harz- et al. 1997). It is accordingly now feasible to model the
burgite GM92-453 has a mean Vp of 8.45 km/s, the elasticity of these phases much more realistically than
highest value for samples from Kerguelen studied here, as has previously been attempted ꢀe.g. Jackson et al. 1990).
expected for its high mg number ꢀ89.93) and the high In particular, clinopyroxene elasticity data are now
modal proportion of olivine ꢀ84%, Table 4).
available not only for the quadrilateral end members
enstatite, ferrosilite, diopside and hedenbergite) but also
for acmite, jadeite and Tschermak components.
ꢀ
Calculated density and P-wave velocities
The approach adopted here is to recalculate the
Densities, elastic moduli and P-wave velocities for the analysed orthopyroxene compositions within the ter-
xenoliths of this study have been calculated by the nary system En±Fs±CaTs and the clinopyroxene com-
Table 6 Measured and calculated densities and seismic velocities ꢀ0.9 GPa) of the Type I and Type II Kerguelen xenoliths
Density
ꢀ
measured
Density
ꢀg/cm )
calculated
Consistency ꢀ%)
Vp ꢀkm/s)
measured
ꢀ0.9 GPa)
Vp ꢀkms)
calculated
Consistency
ꢀ%)
3
g/cm )
3
Harzburgite ꢀperidotite mantle wall-rock)
3.30
GM92-453
3.34
1
8.45
8.29
2
Pyroxene granulites ꢀType IIa): tholeiitic±transitional basaltic anity
GM92-412
OB93-57
3.06
3.08
3.09
3.13
1
1.5
7.37
7.52
7.43
7.39
1.5
2
Pyroxene garnet granulites ꢀType IIa, IIa/IIc and IIc): tholeiitic±transitional basaltic anity
GM92-347
GM92-394
GM92-390
3.25
3.04
3.17
3.32
3.27
3.48
2
7
9
7.9
7.06
7.08
7.99
7.69
7.78
1.5
8
9
Garnet pyroxenite ꢀType IIb): alkali basaltic anity
GM92-165 3.37 3.41
1
7.03
7.88
11

2
55
Table 7 Densities and elastic moduli for pyroxene end members. Unless otherwise indicated, densities are from Smyth and McCormick
1995) and elasticity data from Bass ꢀ1995)
ꢀ
3
)
Descriptor
Composition
Density ꢀg/cm
K ꢀbulk modulus) ꢀGPa)
G ꢀshear modulus) ꢀGPa)
Orthopyroxene
Enstatite
Ferrosilite
Mg
Fe Si
En80Fs
ꢀMg0.65Ca0.35)AlAlSiO
2
Si
2
O
6
3.204
4.002
3.304c
3.376
108
101
116
76
52
78
100
a
2
2
O
6
b
`Kilbourne Hole'
9
MgTs
7
CaTs
4
d
d
ꢀMg,Ca)Ts
6
183
Clinopyroxene
Diopside
Hedenbergite
Jadeite
Acmite
CaTs
CaMgSi
CaFeSi
NaAlSi
NaFeSi
CaAlAlSiO
2
O
O
6
O
6
6
3.279
3.656
3.341
3.576
3.438
113
120
143
112e
143
67
61
85
59
2
2
2
O
6
e
6
85
a
d
Molar averages based on G=72 GPa provide a better ®t to the
moduli measured on solid solutions of bronzite composition
Chai et al. ꢀ1997)
Required to reconcile data for the Kilbourne Hole aluminous
orthopyroxene ꢀChai et al. 1997) with the molar average of end
member moduli
b
c
e
Estimated from data tabulated for clinoenstatite, diopside and
CaTs by Smyth and McCormick ꢀ1995)
As for jadeite ꢀsee text)
positions as a combination of the ®ve end members Di, garnet granulites GM92-347). The other two garnet
Hd, Ac, Jd and CaTs. This approximation to the granulites ꢀGM92-394 and GM92-390) give substan-
measured compositions of the generally aluminous tially lower measured V and densities than are calcu-
P
pyroxenes consistently accounts for more than 95% of lated from the modal mineralogy ꢀTable 3). Garnet
the analysis. The bulk and shear moduli of each of clinopyroxenite ꢀGM92-165) has similar calculated and
these end member phases appropriate for elastically measured densities but lower measured than calculated
isotropic polycrystals as calculated or inferred from P-wave velocity ꢀFig. 7).
single-crystal elasticity measurements are presented in
Table 7.
There are several possible explanations for discrep-
ancies commonly observed between the P-wave veloci-
The substantially higher bulk and shear moduli for ties measured on xenoliths and those calculated from the
jadeite than for Al-free pyroxenes are attributed mainly modal mineralogy as discussed by Rudnick and Jackson
to the sti€ening of the Al-accommodating M1 octahe- ꢀ1995) for ma®c granulite xenoliths from the Chudleigh
dral site ꢀKandelin and Weidner 1988). Signi®cantly, this volcanic province ꢀNorth Queensland, Australia). Scat-
feature of the jadeite structure is partly shared by CaTs tering of elastic waves at grain boundaries is not appli-
in which Al is distributed between the M1 octahedral cable to the Kerguelen xenoliths for the same reasons
and the tetrahedral sites. Na in jadeite is replaced by Ca given by Rudnick and Jackson ꢀ1995). Neither are
in the M2 sites of CaTs, and half of the Si in jadeite is compositional e€ects due to the pyroxenes relevant be-
replaced by Al in the tetrahedral sites of CaTs. The ꢀCa/ cause we have accounted in this study for the Al and Na
M2
T
Na)
and ꢀAl/Si) substitutions may be expected to
have opposing in¯uences upon the elastic moduli that
will tend to cancel. Accordingly, the elastic moduli of
CaTs can reasonably be approximated by those for
M2
jadeite ꢀTable 7). For MgTs, the substitution ꢀMg/Ca)
is expected to result in moduli substantially higher even
than for CaTs and jadeite, in accord with the inferences
drawn from the properties of the Kilbourne Hole orth-
opyroxene ꢀTable 7).
The correlation between the measured densities and
mean ꢀ0.9±1.0 GPa) compressional wave velocities and
the primary mineralogy of the xenoliths has been as-
sessed by comparison with the corresponding properties
calculated from the modal mineralogy ꢀTable 4) and
appropriate density and single-crystal elasticity data ꢀas
described above in the Methods section).
The calculated densities and P-wave velocities ꢀFig. 7;
Table 6) are in good agreement ꢀwithin 1±2 and 1.5±2%
respectively) with their measured equivalents for four
samples ꢀharzburgite GM92-453, the pyroxene-granu-
Fig. 7 Measured velocity ꢀ0.9 GPa) plotted against calculated
velocity ꢀmean values) of Kerguelen granulite and pyroxenite
lites GM92 412 and OB93-57, and one of the three xenoliths

256
contents of the pyroxenes. Our method of calculation of Kempton and Harmon ꢀ1992) showed that the rela-
the modal composition of the samples is based on rep- tionship between mg number and SiO /Al O can be
2
2
3
resentative bulk rock and mineral analyses and is un- used to identify granulites that had formed by accumu-
likely to be a source of signi®cant error. It is also lation of speci®c mineral phases from those representing
unlikely that there is an error in the modal composition solidi®ed melts ꢀFig. 3). The Type IIa, IIa/IIc and IIb
for only three samples that moreover produces a sys- xenoliths lie near or in the Ma®c 1 ®eld in this plot,
tematic shift to higher values for the velocities calculated indicating that they are cumulates of original plagioclase
for these samples.
and pyroxene ꢀor olivine). The Type IIc xenolith plots in
Two likely possibilities therefore remain to explain the `primitive' basaltic magma ®eld, indicating that the
the observed discrepancy between measured and calcu- Type IIc xenoliths represent frozen basaltic melts. The
lated Vp: grain boundary alteration in some xenoliths petrological and geochemical characteristics of sample
and/or failure to close all cracks and pore spaces at GM92-394 are intermediate between those of the Type
0.9 GPa. IIa and Type IIc xenoliths and indicate that a compo-
Petrographic studies of the Kerguelen xenoliths have sitional continuum exists between the deep tholeiitic±
shown that one feature common to samples GM92-390 transitional cumulates and frozen melts.
and GM92-394 is the partial but signi®cant replacement
of garnet with kelyphite ꢀfrom 20 to 80%). Those
kelyphites are a secondary alteration product of lower Signi®cance of petrophysical data and comparison
3
density ꢀꢀ3.3 versus ꢀ3.8 g/cm ) and compressional with the seismic data and models
wave speed ꢀ8.0 versus 8.8±8.9 km/s), resulting from
breakdown of the garnet due to decompression The V data derived in this study, along with the geo-
P
ꢀ
Rudnick and Jackson 1995) or heating ꢀe.g. Grin et chemical and geothermobarometry data from the same
al. 1984, 1987). The measured densities and wave speeds xenoliths, allow a re-interpretation of existing seismic
of the two xenoliths GM92-390 and GM92-394 are data ꢀRecq et al. 1990, 1994; Charvis et al. 1995) and the
lower than the calculated ones and could be explained construction of likely rock-type sections for the crust±
by signi®cant conversion of garnet to kelyphite; this mantle boundary and the lithospheric mantle column
suggests the presence of porosity that is not intercon- ꢀi.e. below the oceanic layer 3) beneath the Kerguelen
nected, presumably associated with the transformation archipelago.
of garnet to kelyphite. The coarse-grained garnet clin-
Geothermobarometry calculations of this study show
opyroxenite GM92-165 is pervasively fractured and that the studied Type II xenoliths are derived from the
shows clear evidence of grain-boundary alteration. In- depth range of about 35 to 55 km beneath the Kerguelen
terconnectivity of the fracture network resulting in the Islands ꢀcorresponding to the calculated pressure range
measurement of a grain density would explain the ap- of about 1.1 to 1.8 GPa). The associated equilibrium
proximate consistency between measured and calculat- temperatures inferred from the primary mineralogy of
ed densities for this specimen. The low and pressure- the granulites and pyroxenite vary between about 800
insensitive wave speeds suggest that pressure is ine€ec- and 1000 °C. It is important to note that granulites may
tive in closing the grain-boundary porosity in this occur intercalated with peridotites in the upper mantle
specimen.
ꢀe.g. Pearson and O'Reilly 1991; Pearson et al. 1991 and
In view of the discrepancies between measured and references therein) and do not exclusively signify crustal
calculated wave speeds for three of the seven xenoliths, facies. The Type I ꢀharzburgite) xenolith ꢀmantle wall-
the wave speeds calculated for the seven studied rock) has an equilibration temperature of 1125 °C ꢀusing
Kerguelen samples, suitably corrected to in situ P±T the two-pyroxene thermometer of Brey and Kohler 1990
conditions, provide the more robust basis for the in Gre
following discussion.
Â
goire et al. 2000) and thus is derived from a level
deeper than the Type II xenoliths. The transitional zone,
which is about 3 km thick, has relatively low velocity,
and is interpreted as representing the crust±mantle
boundary from the seismic data ꢀRecq et al. 1990). It lies
below oceanic layer 3 ꢀshown between 18 to 21 km in
Fig. 8). Therefore, no xenoliths within the extensive suite
Discussion
Origin of the xenoliths:
studied so far ꢀe.g. Gre
Based on their petrographic, major element and REE samples of the lithospheric column shallower than about
characteristics, Gregoire et al. ꢀ1998) have proposed that 30±35 km. The mantle sample provided by xenoliths is
Â
goire et al. 1994, 1998) provide
Â
the Type II Kerguelen xenoliths are either deep basaltic always plagued by the capricious entrainment charac-
cumulates of the tholeiitic±transitional ꢀType IIa) or teristics of the host magmas ꢀe.g. O'Reilly 1989). How-
alkaline ꢀType IIb) magmatic series of the archipelago or ever, Gre
else deep frozen tholeiitic±transitional basaltic melts the di€erent xenolith types evidencing multiple events of
Type IIc). The major and trace element features of the magmatic intrusion in the mantle beneath Kerguelen
studied Type II xenoliths in this paper emphasise these that correlate well with the tectonic and magmatic
origins and highlight some new characteristics. history of the region. We are therefore con®dent that the
Â
goire et al. ꢀ1998) have been able to recognise
ꢀ

2
57
for pyroxene granulites, and the garnet granulites from
Kerguelen are distinctly higher in wave speed than those
from other localities. The high wave speeds are attrib-
utable to the distinctive major element compositions of
these Kerguelen xenoliths. The cumulates that dominate
this granulite suite are highly magnesian [100 Mg/
ꢀ
and 26 wt% Al O for GM92-347 and GM92-412).
Mg+Fe)=75±82] and also very aluminous ꢀup to 23
2
3
These compositional characteristics are re¯ected in the
aluminous compositions of the pyroxenes and the rela-
tively large amounts of modal garnet in the garnet
granulites. These factors combine to explain the
unusually high wave speeds calculated ꢀand measured)
for the Kerguelen granulite suite. The calculated and
ꢀ
1 GPa) measured wave speeds for the harzburgite
GM92-453 and the calculated wave speed for the
pyroxenite GM92-165 are unexceptional and fall within
the ®eld identi®ed by O'Reilly et al. ꢀ1990) and Rudnick
and Fountain ꢀ1995) for ultrama®c rocks.
For meaningful comparison with seismic wave speeds
for the deep crust and upper mantle beneath Kerguelen,
the calculated ꢀor measured) V need to be adjusted for
P
in situ conditions of pressure and temperature. The
appropriate temperature and pressure derivatives,
determined in high-frequency ultrasonic studies,
À4
are @Vp=@T ˆ À 5  10 km=s=K and @Vp=@P ˆ
0:1km=s=Gpa ꢀe.g. Jackson 1991). The additional e€ect
of viscoelastic relaxation ꢀMinster and Anderson 1981,
equation 43) is negligible ꢀ<0.05 km/s) for relatively
short-period P-waves under conditions appropriate to
the upper mantle beneath Kerguelen.
The granulites and pyroxenites are interpreted as
magmatic intrusions underplated and/or intruded
around the crust±mantle boundary and intercalated
Fig. 8 a Vp±depth models for the Kerguelen Islands ꢀafter Recq
et al. 1990, 1994). b Model for crust±mantle boundary ꢀCMB)
beneath the Kerguelen Islands. The 2±3 km transitional zone below
oceanic layer 3 of the Kerguelen plateau is best explained by a throughout the lithospheric mantle ꢀrepresented by the
varying mixture of pyroxene granulites and ultrama®cs with the
proportion of the latter material increasing downwards. Lateral
sections ꢀe.g. O'Reilly and Grin 1987; Deemer and
heterogeneities seen seismically in the Kerguelen upper mantle and
harzburgite) analogous to many continental lithospheric
Hurich 1994) and other oceanic regions where there has
the low average wave speeds can be explained by variable
proportions of the garnet granulites and pyroxenites and harz- been signi®cant volcanic activity ꢀe.g. Agulhas: Uenzel-
burgite rock types characterised in this study, along with an
anomalously high average temperature. Such lithological hetero-
geneity results from intrusions of basaltic magmas into the mantle
wall-rock, consistent with other evidence of sustained voluminous
magmatic activity in this region
mann-Neben et al. 1999; Gohl and Uenzelmann-Neben
2
1
001; Iceland: Operto 1995; Marquesas: Caress et al.
992).
The V values calculated for these xenoliths, suitably
P
adjusted to in situ pressures and temperatures, are
xenolith suite available represents the deep-seated rock- 7.1 km/s for the pyroxene granulites, 7.4±7.7 km/s for
type variability within this lithospheric column and can the garnet granulites and pyroxenite, and 8.0 km/s for
be used to infer the rock-type mix at shallower levels the harzburgite. It follows that the transitional layer
between about 55 km and oceanic layer 3.
ꢀV ꢀ7.2±7.5 km/s at depths of 18±21 km) beneath layer
P
The compressional wave speeds calculated ꢀand in 3 of the oceanic crust is most plausibly interpreted as
favourable circumstances measured) for the xenoliths of an interlayering of pyroxene granulites and ultrama®c
the Kerguelen suite indicate VP is ꢀ7.4 km/s for the rocks with increasing proportions of mantle wall-rock
pyroxene granulites and 7.7±8.0 km/s for the garnet with increasing depth through the transition
granulites and pyroxenite respectively. These values zone ꢀFig. 8). As garnet is not stable in granulite
compare with the ranges of measured ꢀ600 MPa) V for compositions at this relatively shallow depth, the garnet
P
garnet-free ꢀ6.79±7.45 km/s) and garnet-bearing granulites would not contribute to this part of the
ꢀ
given by Rudnick and Fountain ꢀ1995). The garnet-free about 30 km ꢀGregoire et al. 1994, 1998). A similar
7.03±7.50 km/s) ma®c granulites from various localities lithospheric column: they become stable at a depth of
Â
Kerguelen granulites thus have calculated ꢀand mea- mixture of rock types has been well-constrained for
sured) wave speeds as high as any previously reported some continental regions with more continuous xeno-

2
58
lith sampling of the relevant depth range ꢀe.g. O'Reilly the garnet granulites and pyroxenites and harzburgite
and Grin 1987). rock types characterised in this study, along with an
The seismic data reveal two other characteristics of anomalously high average temperature. Such lithologi-
the mantle beneath the Kerguelen plateau ꢀRecq et al. cal heterogeneity results from intrusions of basaltic
1
990): lateral variations of the velocity gradient within magmas into the mantle wall-rock, consistent with other
the upper mantle and relatively low P-wave velocities evidence of sustained voluminous magmatic activity in
7.7±7.95 km/s) for the upper mantle. These observa- this region ꢀGautier et al. 1990; Giret 1993; Weis et al.
tions might be explained by variable proportions of 1993; Gregoire et al. 1998).
ma®c ꢀgarnet granulites and pyroxenites) and ultrama®c The relatively low density of the Kerguelen mantle
harzburgite) lithologies within the upper mantle. This is to about 80 km is also consistent with the relatively low
ꢀ
Â
ꢀ
consistent with the indications from the geothermoba- density of the depleted mantle and the intercalated
rometry that the ma®c rock types are present to depths basaltic rock types equilibrated to mantle conditions.
of at least 55 km and therefore are mixed with the
mantle wall-rocks. The generally low and spatially Acknowledgements This work has been made possible by the
variable wave speeds are probably also indicative of a generous assistance and technical expertise of N.J. Pearson,
A. Sharma, C. Lawson ꢀGEMOC Geochemical Analysis Unit)
H. Kokkonen and R. Guziak ꢀPetrophysic laboratory, RSES,
ANU). This work has been supported by a Macquarie University
Research Fellowship and Grant Schemes ꢀM.G.), Australian Re-
broad thermal anomaly. For a given lithology, the sen-
sitivity of VP to temperature variation should be
ꢀ
±0.1 km/s per 200 °C.
The mantle harzburgite of this study has a calculated search Council Large and Small Grants ꢀS.Y. O'R., M.G.) and
CNRS and IFRTP grants ꢀJ-Y. C.). We appreciate the continuing
3
density of 3.34 g/cm ꢀthe measured value is even lower
3
at 3.30 g/cm ), which is signi®cantly less dense than
advice and input of A. Giret. We also thank the French Polar
Research and Technology Institute ꢀIFRTP, Brest, France) and the
University Jean Monnet ꢀSt. Etienne, France) for their support.
3
young subcontinental mantle ꢀ3.37 g/cm ) or primitive
3
mantle ꢀ3.39 g/cm ). It is similar to relatively buoyant D. Smith, H. Kern, K. Mengel, J. Hoefs and an anonymous
3
Proterozoic mantle ꢀ3.34 g/cm ), as documented in
reviewer are gratefully acknowledged for providing constructive
and thorough reviews. This is Publication 244 of the ARC National
Key Centre for the Geochemical Evolution and Metallogeny of
Continents ꢀwww.es.mq.edu.au/GEMOC/).
Grin et al. ꢀ1999), but has a higher density than de-
pleted very buoyant Archean mantle ꢀ3.31 g/cm ) be-
3
cause of its higher Fe content. In addition, the densities
of the pyroxenites and garnet granulites range to values
3
3.27 g/cm ) signi®cantly lower than the harzburgite
ꢀ
References
density. These results are consistent with those of
Charvis et al. ꢀ1995) and those inferred from the geoid
data by Marks and Sandwell ꢀ1991), indicating the oc-
currence of relatively low-density material to about
80 km deep in the mantle beneath Kerguelen in order to
achieve the local isostatic compensation.
Bass JD ꢀ1995) Elasticity of minerals, glasses and melts. In: Ahrens
TJ ꢀed) Mineral physics and crystallography: a handbook of
physical constants. Am Geophys Union, Washington, pp 45±63
Bass JD, Weidner DJ ꢀ1984) Elasticity of single-crystal orthofer-
rosilite. J Geophys Res 89:4359±4371
Bhagat SS, Bass JD, Smyth JR ꢀ1992) Single-crystal elastic prop-
erties of omphacite-C2/c by Brillouin spectroscopy. J Geophys
Res 97:6843±6848
Conclusions
È
Brey GP, Kohler T ꢀ1990) Geothermometry in four-phase lherzo-
lites. II. New thermobarometers, and practical assessment of
existing thermobarometers. J Petrol 31:1353±1378
Caress DW, Mutter JC, McNutt MK, Detrick RS ꢀ1992) An OBS
refraction experiment across the Marquesas islands: the deep
crustal structure of a hot-spot trace. EOS Trans Am Geophys
Union 73:489±490
Measured and calculated values of V for rock samples
P
representative of the crust±mantle boundary and upper
mantle beneath the Kerguelen plateau de®ne three
ranges: ꢀ7.4 km/s for pyroxene granulites, 7.7 to 8.0 km/s
for garnet granulites and 8.3 km/s for harzburgite ꢀper-
idotite mantle wall-rock). Measured and calculated
Chai M, Brown JM, Slutsky LJ ꢀ1997) The elastic constants of an
aluminous orthopyroxene to 12.5 GPa.
02:14779±14785
J Geophys Res
1
values of V are in good agreement for samples free of Charvis P, Recq M, Operto S, Brefort D ꢀ1995) Deep structure of
P
the northern Kerguelen plateau and hotspot-related activity.
Geophys J Int 122:899±924
Con MF, Eldhom O ꢀ1993) Scratching the surface: estimating
dimensions of large igneous provinces. Geology 21:515±518
kelyphitic transformation of garnet and grain-boundary
alteration. These wave speeds need to be adjusted
downward by 0.3±0.4 km/s for in situ conditions ꢀ1±
2
mobarometry.
The 2±3 km transitional zone from about 18 to 20 km
below oceanic layer 3 of the Kerguelen plateau is best
explained by a varying mixture of pyroxene granulites
GPa, 800±1100 °C) determined through geother- Coisy P, Nicolas A ꢀ1978) Structure et ge
Â
odynamique du manteau
rieur sous le Massif Central ꢀFrance) d'apres l'etude des
enclaves des basaltes. Bull Mineral 4:424±436
supe
Â
Á
Â
Deemer SJ, Hurich CA ꢀ1994) The re¯ectivity of magmatic un-
derplating using the layered ma®c intrusion analog. Tectono-
physics 232:239±255
and ultrama®cs with the proportion of the latter mate- Ellis DJ, Green DH ꢀ1979) An experimental study of the e€ect of
Ca upon garnet±clinopyroxene Fe±Mg exchange equilibria.
Contrib Mineral Petrol 71:13±22
Finlayson DM, Leven JH ꢀ1987) Lithospheric structures and pos-
sible processes in eastern Australia from deep seismic investi-
gations. Tectonophysics 133:199±215
rial increasing downwards. Lateral heterogeneities seen
seismically in the Kerguelen upper mantle and the low
average wave speeds ꢀRecq et al. 1990, 1994; Charvis
et al. 1995) can be explained by variable proportions of

2
59
Frey FA, Prinz M ꢀ1978) Ultrama®c inclusions from San Carlos, Leitner BJ, Weidner DJ, Liebermann RC ꢀ1980) Elastic of single
Arizona. Petrologic and geochemical data bearing on their crystal pyrope and implications for garnet solid solution series.
petrogenesis. Earth Planet Sci Lett 38:139±176 Phys Earth Planet Inter 22:111±121
Gasparik T ꢀ1984) Two-pyroxene thermobarometry with new ex- Liebermann RC ꢀ1976) Elasticity of ilmenites. Phys Earth Planet
perimental data in the system CaO±MgO±Al
Mineral Petrol 87:87±97
2
O
3
±SiO
2
. Contrib
Inter 12:5±10
Marks KM, Sandwell DT ꢀ1991) Analysis of geoid height versus
topography for oceanic plateaux and swells using nonbiased
linear regression. J Geophys Res 96:8045±8055
Gautier I, Weis D, Mennessier JP, Vidal P, Giret A, Loubet M
1990) Petrology and geochemistry of the Kerguelen Archipela-
ꢀ
go basalts ꢀSouth Indian Ocean): evolution of the mantle sources McDonough WF, Sun S ꢀ1995) The composition of the Earth.
from ridge to intraplate position. Earth Planet Sci Lett 100:59±76
Giret A ꢀ1993) Les etapes magmatiques de l'edi®cation des õ
Kerguelen. Bull Soc Geol Fr/Bull APBG, Spec Issue, pp 273±282
Giret A, Gre
Chem Geol 120:223±253
les Minster JB, Anderson DL ꢀ1981) A model of dislocation-controlled
rheology for the mantle. Philos Trans R Soc Lond 299:319±356
Â
Â
Ã
Â
goire M, Cottin JY, Michon G ꢀ1997) The Kerguelen Operto S ꢀ1995) Structure et origine du Plateau de Kerguelen
islands: the third type of oceanic islands. In: Ricci CA ꢀed) The
Antarctic region: geological evolution and processes. Terra
Antarctica Publication, pp 735±741
ꢀOcean Indien Austral): implications geodynamiques. These
d'Universite, Univ Paris VI
O'Reilly SY ꢀ1989) Australian xenolith types, distribution and
transport. In: Johnson RW ꢀed) Intraplate volcanism in eastern
Australia and New Zealand. Cambridge University Press,
Cambridge, pp 249±253
Gohl K, Uenzelmann-Neben G ꢀ2001) The crustal role of the
Agulhas Plateau, southwest Indian Ocean: evidence from seis-
mic pro®ling. Geophys J Int 144:632±646
Gre
Shimizu N, Giret A ꢀ1994) Oceanic ma®c granulite from the
Kerguelen Archipelago. Nature 367:360±363
Gre
Â
goire M, Mattielli N, Nicollet C, Cottin JY, Leyrit H, Weis D, O'Reilly SY, Grin WL ꢀ1987) Eastern Australia ± 4000 kilome-
ters of mantle samples. In: Nixon PH ꢀed) Mantle xenoliths.
Wiley, Chichester, pp 267±280
goire M, Cottin JY, Mattielli N, Nicollet C, Weis D, Giret A O'Reilly SY, Grin WL ꢀ1988) Mantle metasomatism beneath
Â
ꢀ
1995) The Kerguelen archipelago: an hypothetic continental
ma®c protolith. Terra Antarctica 2ꢀ1):1±6
Gre
western Victoria, Australia, I: Metasomatic processes in Cr-
diopside lherzolites. Geochim Cosmochim Acta 52:433±447
Â
goire M, Lorand JP, Cottin JY, Giret A, Mattielli N, Weis D O'Reilly SY, Jackson I, Bezant C ꢀ1990) Equilibration tempera-
ꢀ
1997) Petrology of Kerguelen mantle xenoliths: evidence of a
tures and elastic wave velocities for upper mantle rocks from
eastern Australia: implications for the interpretation of seis-
mological models. Tectonophysics 185:67±82
refractory oceanic mantle percolated by basaltic melts beneath
the Kerguelen archipelago. Eur J Mineral 9:1085±1100
Gre
Â
goire M, Cottin JY, Giret A, Mattielli N, Weis D ꢀ1998) The Pearson NJ, O'Reilly SY ꢀ1991) Thermobarometry and P±T±t
paths: the granulite to eclogite transition. J Metamorph Geol
9:349±359
metaigneous xenoliths from Kerguelen archipelago: evidence of
a continent nucleation in an oceanic setting. Contrib Mineral
Petrol 133:259±283
Gregoire M, Moine BN, O'Reilly SY, Cottin JY, Giret A ꢀ2000)
Pearson NJ, O'Reilly SY, Grin WL ꢀ1991) The granulite to
eclogite transition beneath the eastern margin of the Australian
craton. Eur J Mineral 3:293±322
Â
Trace element residence and partitioning in mantle xenoliths
metasomatised by high alkaline silicate- and carbonate-rich Recq M, Brefort D, Malod J, Veinante JL ꢀ1990) The Kerguelen
melts ꢀKerguelen Islands, Indian Ocean). J Petrol 41:477±509 Isles ꢀsouthern Indian Ocean): new results on deep structure
Grin WL, Wass SY, Hollis JD ꢀ1984) Ultrama®c xenoliths from from refraction pro®les. Tectonophysics 182:227±248
Bullenmeri and Gnotuk maars, Victoria, Australia: petrology of Recq M, Le Roy I, Charvis P, Goslin J, Brefort D ꢀ1994) Structure
a subcontinental crust±mantle boundary. J Petrol 25:53±87
Grin WL, Sutherland FL, Hollis JD ꢀ1987) Geothermal pro®le
and crust±mantle transition beneath east-central Queensland: Rudnick RL, Fountain DM ꢀ1995) Nature and composition of the
profonde du mont Ross d'apre
ocean Indien austral). Can J Earth Sci 31:1806±1821
Â
Á
s la sismique ꢀ õÃ les Kerguelen,
volcanology, xenolith petrology and seismic data. J Volcanol continental crust: a lower crustal perspective. Rev Geophys
Geotherm Res 31:177±203 33:267±310
Grin WL, O'Reilly SY, Ryan CG ꢀ1999) The composition and Rudnick RL, Jackson I ꢀ1995) Measured and calculated elastic
origin of sub-continental lithospheric mantle. In: Fei Y, Bertka
CM, Mysen BO ꢀeds) Mantle petrology: ®eld observations and
high-pressure experimentation. A tribute to Francis R. ꢀJoe)
Boyd, Geochemical Society, Spec Pub No 6, pp 13±43
wave speeds in partially equilibrated ma®c granulite xenoliths:
implications for the properties of an underplated lower conti-
nental crust. J Geophys Res 100:10211±10218
Smyth JR, McCormick TC ꢀ1995) Crystallographic data for min-
erals. In: Ahrens TJ ꢀed) Mineral physics and crystallography: a
handbook of physical constants. Am Geophys Union, Wash-
ington, pp 45±63
Harley SL ꢀ1984) The solubility of alumina in orthopyroxene co-
existing with garnet in FeO±MgO±Al
2 3 2
O ±SiO and CaO±FeO±
MgO±Al ±SiO . J Petrol 25:665±696
3
2
O
2
Irving AJ ꢀ1974) Geochemical and high pressure experimental Uenzelmann-Neben G, Gohl K, Ehrhardt A, Seargent M ꢀ1999)
studies for garnet pyroxenite and pyroxene granulite xenoliths Agulhas Plateau, SW Indian Ocean: new evidence for excessive
from the Delegate basaltic pipe, Australia. J Petrol 15:1±40 volcanism. Geophys Res Lett 26:1941±1944
Jackson I ꢀ1991) The petrophysical basis for the interpretation of Weis D, Frey FA, Leyrit H, Gautier I ꢀ1993) Kerguelen Archi-
seismological models for the continental lithosphere. In:
Drummond BJ ꢀed) The Australian lithosphere. Geol Soc Aust
Spec Publ 17, pp 81±114
pelago revisited: geochemical and isotopic study of the S.E.
Province lavas. Earth Planet Sci Lett 118:101±119
Wells PRA ꢀ1977) Pyroxene thermometry in simple and complex
systems. Contrib Mineral Petrol 62:129±139
ments on lower crustal xenoliths from Calcutteroo, South Wood BJ, Holloway JR ꢀ1984) A thermodynamic model for
Jackson I, Arculus RJ ꢀ1984) Laboratory wave velocity measure-
Australia. Tectonophysics 101:185±197
Jackson I, Rudnick RL, O'Reilly SY, Bezant C ꢀ1990) Measured
subsolidus equilibria in the system CaO±MgO±Al
Geochim Cosmochim Acta 48:159±176
and calculated elastic wave velocities for xenoliths from the Xu X, O'Reilly SY, Grin WL ꢀ1998) The geotherm of the litho-
2 3 2
O ±SiO .
lower crust and upper mantle. Tectonophysics 173:207±210
Kandelin J, Weidner DJ ꢀ1988) The single-crystal elastic properties
of jadeite. Phys Earth Planet Inter 50:251±260
Kempton PD, Harmon RS ꢀ1992) Oxygen isotope evidence for
large-scale hybridization of the lower crust during magmatic
underplating. Geochim Cosmochim Acta 56:971±986
sphere beneath Qilin, SE China: a reappraisal and implications
for P±T estimation of Fe-rich pyroxenites. Lithos 47:115±119
Yang H-Y, Frey FA, Weis D, Giret A, Pyle D, Michon G ꢀ1998)
Petrogenesis of the ¯ood basalts forming the Northern Ker-
guelen Archipelago: implications for the Kerguelen Plume.
J Petrol 39:711±748