Contrib Mineral Petrol (1998) 132: 34±47 Ó Springer-Verlag 1998

Thomas Stachel á Je€ W. Harris á Gerhard P. Brey

Rare and unusual mineral inclusions in diamonds

from Mwadui, Tanzania

Received: 23 August 1997 / Accepted: 7 January 1998

Abstract Syngenetic diamond inclusions from the

Mwadui kimberlite reveal that an unusually fertile sec-
Introduction
tion of lithospheric mantle beneath the Central African
Syngenetic mineral inclusions in diamonds are pristine
Craton was sampled. This is shown by a very high ratio
witnesses of the chemical and mineralogical environment
of lherzolitic to harzburgitic garnet inclusions (1:2) and
during diamond formation and thus represent a direct
low Mg/Fe-ratios in olivine and orthopyroxene. Geo-
window into the Earth mantle. The study of diamonds
thermometry applied to the peridotitic inclusions indi-
from the Mwadui Mine has the scope to provide data on
cates disequilibrium between non-touching inclusion
the state of the lithospheric upper mantle beneath Tan-
pairs to be common. Disequilibrium between garnet-
zania before the onset of rifting in Eastern Africa
olivine and garnet-orthopyroxene pairs suggests succes-
(Dawson 1992).
sive iron enrichment during diamond formation, e.g.
The Mwadui Mine (formerly known as Williamson
leading to the presence of harzburgitic garnet and
Mine) in Western Tanzania is the largest kimberlite pipe in
lherzolitic olivine in the same diamond. Apart from the
the world (146 ha) and lies within an area where 67 kim-
dominant peridotitic inclusion suite (88%), rare eclogitic
berlites (25 of which are diamondiferous) have been dis-
inclusions occur (2%) and a number of uncertain para-
covered (Gobba 1991). Located on the Tanzania Archon
genesis. Two diamonds, one with eclogitic garnets with
of the Central African Craton (Janse 1994), the pipes have
moderate pyroxene solid solution and the other with a
an eruption age of 40±53 Ma (Davis 1977; Raber 1978;
single ferro-periclase inclusion, suggest the contribution
Shee 1980 unpublished). The Mwadui Mine preserves
of a small sub-lithospheric component. The ®nding of
crater facies, including lake beds, a feature common to the
the association Fe-FeO-Fe3O4 in one single diamond
other pipes in the area (Gobba 1991). Petrographically,
indicates diamond formation over a large range of fO2
the absence of macrocrystal olivine and the presence of
conditions, possibly along redox fronts. Steep compo-
groundmass nepheline and analcime in many of the
sitional gradients may also be re¯ected by the joint oc-
kimberlites suggest possible transitions into para-kim-
currence of harzburgitic garnet and a SiO2-phase in the
berlites and olivine melilitite type rocks (Gobba 1991).
same diamond. Alternatively the formation of the SiO2-
phase may be due to extreme carbonation of the peri-
dotitic source. Further unusual ®ndings include the ex-
solution of a silicate phase from magnetite inclusions, Analytical methods and data base
(i.e. primary solution of c-olivine) and an ilmenite in-
clusion with an eskolaite (Cr2O3) component of Sample preparation and analytical techniques are the same as de-
14.5 mol%, the latter together with harzburgitic para- scribed in Stachel and Harris (1997a). In the same paper we list the
references for our diamond inclusion data base which has been ex-
genesis silicate inclusions. tended here by incorporation of the analytical data of Meyer and
Svisero (1975), Sobolev et al. (1976), Tsai (1978), Mvuemba Ntanda
et al. (1982), Meyer and McCallum (1986), Otter (1990), Chinn
(1995), McDade and Harris (1996), and Kopylova et al. (1997).
T. Stachel (&) á G. P. Brey
Institut fuÈr Mineralogie, UniversitaÈt Frankfurt,
Senckenberganlage 28, D-60054 Frankfurt, Germany
T. Stachel á J. W. Harris Chemical composition of the mineral inclusions
Department of Geology and Applied Geology,
University of Glasgow, Glasgow G12 8QQ, Scotland, UK From the initial population of several thousand dia-
Editorial responsibility: J. Hoefs monds, a total of 100 stones were crushed to release

representative mineral inclusions. Eight diamonds with
graphite coated rosette shaped fracture systems did not
yield any inclusions and one diamond had an inclusion
which was altered. The inclusion contents for the re-
maining 91 diamonds are listed in Table 1.
Garnet
Only 2 of 40 garnets, released from 36 diamonds, belong
to the eclogitic suite and they were found in the same
diamond. These garnets (Table 2, analyses MW52±77
and ±78) are distinct within the world-wide database
because of very high TiO2 contents (about 2 wt%) and
rather high Na2O (0.75 and 0.36 wt%), respectively. The
higher sodium content of MW52±77 is accompanied by
higher calcium and lower iron and aluminium. The
atomic proportions of Si (calculated on a basis of 24 [O]
per formula unit) are also signi®cantly di€erent in these
two garnets (6.179 and 5.920 respectively). The moder-
ate majoritic component only in MW52±77 indicates
crystallisation pressures between 80 and 100 kbar (Iri-
fune 1987). However, a di€erent picture arises, if the
high Ti contents of both garnets are taken into account.
Charge balance for the incorporation of Ti on the oc-
tahedral site may be reached either by substitution of an
equivalent amount of Na on the dodecahedral site or of
Al on the tetrahedral site. For majoritic garnets a sig-
ni®cant positive correlation between Ti and Na is neither
apparent from natural nor from experimental data (Iri-
fune 1987). On the contrary, the correlation between Si
and Al + Cr improves considerably if the sum of
Si + Ti is used (being equivalent to AlVI + Cr + Ti
versus Si + AlIV). For MW52±77 and ±78 then a much
higher excess in silica becomes apparent, with Si + AlIV
summing to 6.397 and 6.155 respectively (Fig. 1). Al-
though this calculation implies a majoritic component
for both garnets, the di€erence in excess silica remains
and might actually indicate continued diamond growth
at di€erent depth. Alternatively the co-variation of ex-
cess silica and sodium in the samples may indicate that,
depending on the availability of Na, Si occupies the
octahedral site for charge balance, e.g. as in Na-
Table 1 Inclusion abundance
table for Mwadui Mono-mineralic assemblages
Olivine (ol)
Cr-bearing pyrope (py)
Eclogitic pyrope-almandine
Enstatite (opx)
Chrome diopside (cpx)
Eclogitic omphacitic cpx
Chromite (chr)
Magnetite (Fe3O4)
Ferro-periclase
Sulphide (sulph)
Dolomite
Total diamonds
35
Ca2(AlSi)Si3O12 (Ringwood and Major 1971). Thus di-
amond growth may have occurred at a common depth.
On the other hand Fig. 2 shows that majoritic garnets
which were formed over a pressure range follow a linear
trend with the slope depending on the sodium content of
the speci®c source. Irifune's (1987) synthetic majorites
(Fig. 2) represent a trend for a source of moderate so-
dium content (i.e. for pyrolite minus 62% olivine as
starting material). For the two Mwadui garnets the
common high sodium trend line supports primary
crystallisation from a similar source at variable pres-
sures.
Of the peridotitic garnets, one third (an unusually
high proportion, cf. Meyer 1987) plot into the lherzolite
®eld de®ned by Sobolev et al. (1973, see Fig. 3). Seven of
these lherzolitic garnets are almost identical at 6 wt%
CaO with Cr2O3 at 5.0±5.5 wt% for four of them and
9.5 wt% for three. Lherzolitic garnet MW2±5b (Table 2)
in contact with a clinopyroxene inclusion, falls just
outside the lherzolite ®eld. Since this garnet contains no
detectable Na (which also shifts the lherzolite-ha-
rzburgite division to lower Ca contents, Sobolev et al.
1973), an unusually high crystallisation pressure and /or
temperature is indicated (Brey 1989; Brenker and Brey
1997). The harzburgitic garnets in Fig. 2 range from the
lherzolitic ®eld boundary, but do not extend to the very
high Cr2O3 versus low CaO region found for example, at
the Finsch Mine diamonds (Harris and Gurney 1979).
In general major and minor element contents for all
the garnets fall in the compositional range observed for
peridotitic garnet inclusions world-wide, except for some
rather high TiO2 contents (around 0.4 wt%) for several
lherzolitic and harzburgitic garnets and one harzburgitic
garnet with an uniquely high concentration of 1.07 wt%
(Table 2: MW38±66), which occurs together with an il-
menite inclusion.
Sample MW36±61a/b (Table 2) is a paired inclusion
(110 ´ 180 lm) of garnet and chromite. Analytical
pro®les (EPMA) over the garnet-chromite contact show
strong compositional gradients within the garnet with
Cr2O3 (12.5±9.4 wt%) and CaO (2.17±1.81 wt%) de-
creasing and Al2O3 (15.1±17.5 wt%) increasing towards
the garnet-chromite interface. Si, Ti, Mg, Mn, Fe, Ni are
n Bi-mineralic n Tri-mineralic n
assemblages assemblages
24 Ol-py 9 Ol-py-opx 1
17 Ol-opx 1 Ol-py-chr 1
1 Ol-cpx 2 Ol-py-ilmenite 1
8 Ol-chr 2 Ol-chr-sulph 1
0 Ol-sulph 1 Fe3O4-FeO-Fe 1
1 Py-opx 4
6 Py-cpx 1
1 Py-SiO2 1
1 Opx-cpx 1
4
1
64 22 5
 36

Table 2 Microprobe analyses of selected silicate, oxide and carbonate inclusions from Mwadui diamonds. Total iron expressed as FeO. Fe3+/Fe2+ for chromites and magnetities
calculated from stoichiometry after Droop (1987). (p peridotitic, p(L) peridotitic, lherzolitic, e ecologitic)

Mineral Garnet Cpx Cpx Garnet Garnet Opx Opx Opx Garnet Cpx Carbonate Chromite Olivine

Sample MW1-1 MW2-4 MW2-5a MW2-5b MW7-12 MW7-13 MW12-20 MW12-21 MW12-22 MW18-36 MW26-46 MW33-54 MW33-55
Paragenesis ? p (L) p p e ? p
Assemblage Gt, qtz Cpx.gt-cpx Gt, opx Gt, opx Cpx Carb Ol, chr, sul

P2O5 0.03 £0.02 £0.02 £0.02 £0.02 £0.02 £0.02 £0.02 £0.02 0.04 £0.02 £0.02 £0.02
SiO2 42.10 54.19 54.13 42.02 41.79 57.32 57.42 57.62 41.42 54.15 0.64 0.11 41.13
TiO2 £0.01 0.02 0.03 0.07 0.07 £0.01 0.05 0.06 0.34 0.83 £0.01 0.13 £0.01
Al2O3 18.25 1.25 0.46 19.77 19.45 1.10 0.66 0.68 16.73 8.59 £0.01 5.45 0.02
Cr2O3 8.33 0.75 0.54 5.41 6.06 0.44 0.49 0.50 9.92 0.05 £0.01 67.09 0.04
Fe2O3 calc 0.52
FeO 6.17 2.42 1.90 5.94 5.71 4.84 4.25 4.22 5.73 6.82 1.26 13.17 5.86
MnO 0.28 0.12 0.07 0.25 0.25 0.15 0.11 0.10 0.26 0.09 0.74 0.25 0.07
NiO £0.02 0.06 0.12 £0.02 £0.02 0.14 0.12 0.15 £0.02 0.03 £0.02 0.12 0.29
MgO 23.87 17.71 18.96 22.13 22.24 34.95 35.44 35.37 22.39 10.35 21.37 12.77 51.64
CaO 1.43 22.07 21.69 4.31 4.58 0.49 0.44 0.44 3.23 14.69 28.79 £0.01 £0.01
Na2O £0.02 0.55 0.34 £0.02 0.03 0.03 0.11 0.06 £0.02 4.10 0.09 £0.02 £0.02
K2 O £0.01 0.09 0.04 £0.01 £0.01 £0.01 £0.01 £0.01 £0.01 0.05 £0.01 £0.01 £0.01
Total 100.46 99.22 98.28 99.91 100.17 99.45 99.10 99.20 100.02 99.80 52.90 99.61 99.04

Mineral Chromite core Chromite rim Garnet rim Garnet core Ilmenite Olivine Garnet Magnetite Garnet Garnet Fe-periclase Olivine
Sample MW36-61a MW36-61a MW36-61b MW36-61b MW38-64 MW38-65 MW38-66 MW44-72 MW52-77 MW52-78 MW56-82 MW58-84

Paragenesis p p ? e ? p
Assemblage Ol, gt-chr Gt, ol, ilmenite Mag Gt Fe-per Ol, sulph

P2 O 5 £0.02 £0.02 £0.02 £0.02 £0.02 £0.02 £0.02 £0.02 0.03 0.08 £0.02 £0.02
SiO2 0.16 0.17 41.62 41.32 0.03 40.97 40.76 0.52 41.14 39.31 0.01 40.85
TiO2 0.04 0.05 0.02 0.02 49.27 £0.01 1.07 £0.01 1.93 2.07 £0.01 £0.01
Al2O3 6.21 6.24 17.68 15.26 0.99 0.02 14.89 0.05 17.82 19.90 0.02 0.02
Cr2O3 66.97 65.98 9.27 12.32 16.23 0.03 10.55 0.65 0.14 0.16 0.16 0.02
Fe2O3 calc 0.00 0.66 66.11
FeO 13.73 13.10 5.85 5.99 19.06 7.61 6.63 30.86 15.75 17.64 22.19 7.08
MnO 0.23 0.27 0.33 0.35 0.16 0.09 0.32 0.32 0.29 0.38 0.12 0.08
NiO 0.08 0.07 £0.02 £0.02 0.36 0.43 £0.02 0.04 £0.02 £0.02 1.27 0.37
MgO 12.54 12.87 23.59 23.00 14.24 50.47 20.67 £0.02 12.39 12.02 75.59 51.01
CaO £0.01 0.03 1.82 2.17 £0.01 0.03 5.31 £0.01 9.57 8.15 £0.01 0.02
Na2O £0.02 £0.02 £0.02 £0.02 £0.02 £0.02 0.03 £0.02 0.75 0.36 0.07 £0.02
K2 O £0.01 £0.01 £0.01 £0.01 £0.01 £0.01 £0.01 £0.01 £0.01 £0.01 £0.01 £0.01

Total 99.96 99.44 100.18 100.43 100.34 99.64 100.23 98.55 99.80 100.07 99.42 99.46
 37

constant over the entire pro®le within each mineral

phase. Compositional gradients were not observed in the
MW88-125
chromite.
Olivine

100.04
£0.02

£0.01

£0.02
£0.01
0.03

0.04
41.00

0.03

7.70
0.12
0.40
50.72
Co-variations between garnet composition and
equilibration temperature, calculated from olivine-gar-
net pairs (see below), do only exist for manganese (slope
MW84-117

±2.8 wt.ppm MnO/°C). Strong temperature dependence

Olivine

of the distribution of manganese between garnet and

£0.02

£0.01

£0.02
£0.01
0.03

0.03
40.84

99.10
0.03

7.49
0.09
0.38
50.21
pyroxene was already realised by Shimizu and AlleÁgre
(1978) and a good negative correlation between tem-
perature and Mn content in garnet also exists in the
MW84-116

experimental results of Brey et al. (1990, their Fig. 10e).

Garnet

£0.01

£0.02

£0.02
£0.01
18.19

4.92
0.04
41.38

99.96
7.48

6.38
0.33

21.23
Gt,ol
p

Olivine
MW82-114

The 43 olivines (from 43 diamonds) range in forsterite

100.19
£0.02

£0.01
0.52
57.55
0.02
1.04
0.41

4.96
0.11
0.14
35.37

0.06
Opx

content from 91.0 to 94.0 mol% (Fig. 4). Olivine in

equilibrium with cpx or lherzolitic garnet is restricted to
forsterite contents between 91.7 and 92.1. The mode of
MW80-109 MW80-110a MW80-110b MW82-113

olivine composition (at about 92 mol%, Fig. 4) coin-

Gt, opx
Garnet

£0.02

£0.02

£0.02
£0.01

cides with the narrow compositional range for the

19.24

5.78
41.63
0.07

99.90
6.12

5.83
0.24

21.00
p(L)

lherzolitic olivines. Although most Mwadui olivine in-

clusions are included singly, this mode would suggest
that lherzolitic olivines predominate. Potentially the
calcium content may be used to assess if olivine was in
£0.02

£0.01

21.67
54.58

1.40

99.86
0.82

2.45
0.13
0.07
17.90

0.77
0.08
Cpx

equilibrium with clinopyroxene (Fig. 5). KoÈhler and

Brey (1990) experimentally determined the calcium
content of olivine coexisting with clinopyroxene in nat-
ural lherzolitic compositions for pressures up to 60 kbar.
£0.02

£0.01

£0.01

There is strong indication that natural diamonds crys-

57.20

1.08

99.69
0.31

5.62
0.13
0.13
34.47
0.71
0.04
Opx

tallise along a geothermal gradient corresponding to 40±

Cpx, cpx-opx

42 mW/m2 surface heat ¯ow (e.g. Boyd and Gurney

1986; Stachel and Harris 1997a). The experimental data
of KoÈhler and Brey (1990) then indicate a minimum
£0.02

£0.01
1.45
0.82

2.40
0.09
0.07

0.70
0.08
21.99
54.08

17.64

99.31
p(L)
Cpx

CaO content in olivine of about 250 ppm at the inter-

section (45 kbar/1130 °C) between the 40 mW/m2 ge-
otherm and the diamond stability curve (Kennedy and
MW77-105

Kennedy 1976). Extrapolation of the experimental re-

Olivine

sults to pressures >60 kbar shows that along a 42 mW/

£0.02

£0.01

£0.02
£0.01
40.28

0.02

98.30
0.04

8.00
0.10
0.40
49.43
0.02

m2 geotherm a maximum CaO content of about

900 ppm is reached at 80 kbar/1470 °C. Due to the
MW77-104

curvature of the geotherm the CaO content in olivine

Ol, cpx

probably does not rise further towards even greater

£0.02

£0.01
54.80

1.23

99.56
2.36

2.08
0.09
0.04
16.77
20.57
1.27
0.35
p(L)
Cpx

depth. Thus CaO contents of 1000 ppm and more would

indicate geothermal gradients higher than cratonic geo-
therms or locally elevated temperatures. Mwadui oli-
MW64-92

vines of known lherzolitic paragenesis range between

Garnet

200 and 700 ppm CaO (Fig. 5). A large number of oli-
£0.02

£0.01

£0.02

£0.02
£0.01
7.53

5.81
0.28

2.45
41.46

18.51

23.20

99.24

vines included singly but with similar forsterite contents

(<92.5 mol%) also show CaO contents between 200±
900 ppm. This suggests a lherzolitic paragenesis for all
MW64-91
Olivine

these olivines consistent with the high proportion of

100.07
Assemblage Gt, ol
Table 2 (continued)

£0.02

£0.01

£0.02
£0.01
41.22

0.02
0.03

7.30
0.09
0.41
50.97
0.02

lherzolitic paragenesis garnets. The large variation in

Paragenesis p

CaO content of these lherzolitic olivines implies crys-

tallisation over a considerable depth range. All olivines
calc

with forsterite contents greater 93 mol% have CaO

Mineral

Sample

Cr2O3
Fe2O3
Al2O3

Na2O

contents too low (<150 ppm) for crystallisation in

Total
MnO

MgO
P2O5

TiO2

CaO
SiO2

K2O
FeO

NiO

equilibrium with clinopyroxene. Olivines of both para-

38

Table 2 (continued)

Mineral Opx Garnet Garnet Magnetite Wustite rim

Sample MW88-126a MW88-126b MW88-127 MW96-136 MW96-137a

Paragenesis p ?
Assemblage Gt, ol, gt-opx Mag, wu-Fe

P2O5 £0.02 £0.02 £0.02 £0.02 £0.02

SiO2 58.52 41.65 41.35 0.80 0.83
TiO2 £0.01 0.07 0.07 £0.01 £0.01
A2O3 0.51 18.49 18.26 0.06 0.12
Cr2O3 0.35 7.46 7.45 0.57 1.74
Fe2O3 calc 65.70
FeO 3.46 5.75 5.71 31.22 94.98
MnO 0.08 0.22 0.22 0.38 0.78
NiO 0.22 £0.02 £0.02 0.06 £0.02
MgO 36.90 21.71 21.50 £0.02 £0.02
CaO 0.36 4.96 4.89 £0.01 £0.01
Na2O 0.03 £0.02 £0.02 £0.02 £0.02
K2O £0.01 £0.01 £0.01 £0.01 £0.01
Total 100.42 100.32 99.44 98.78 98.46

5 0.4
Mwadui
World-wide
Irifune (1987)
50 kbar
4 0.3
80
100
Al + Cr + Ti

120
137
Na

3 0.2
VI

150
175 kbar

2 165 0.1
Mwadui
World-wide
Irifune (1987)

1 0
6 IV 7 6 IV
7
Si + Al Si + Al

Fig. 1 Octahedral versus tetrahedral cations for eclogitic garnet
inclusions from Mwadui and from world-wide occurrences. Cations
per formula unit calculated on a basis of 24 oxygens. Tetrahedral
Al3+ was assigned to balance an equivalent amount of octahedral
Ti4+. Crosses correspond to pressures from experiments of Irifune
(1987)
geneses are possibly present at forsterite contents be-
tween 92.5 and 93 mol%, with CaO contents between
150±250 ppm.
Otherwise major and minor elements fall completely
within the ranges of our world-wide database, with only
some of the NiO contents being comparatively high (up
to 0.43 wt%) for the generally low forsterite contents.
Orthopyroxene
In accord with the olivines the chemical composition of
orthopyroxenes (opx) (18 inclusions released from 15
Fig. 2 Sodium versus ``silica excess'' [on a basis of (O) ˆ 24] for
eclogitic garnet inclusions from Mwadui and from world-wide
occurrences, and experimental compositions from Irifune (1987)
diamonds) falls on the iron-rich side of our world-wide
inclusion data base, with the Mg-numbers [100 Mg/
(Mg+Fetotal)] showing a mode between 92.5 and
93 mol% (Fig. 4). A few orthopyroxenes show CaO
contents too low (down to 0.03 wt%) to have been in
equilibrium with clinopyroxene (cpx), but all others
form a narrow cluster between 0.44 and 0.52 wt% CaO,
irrespective of their paragenesis. One opx (MW80±110a,
Table 2), intergrown with cpx, has an unusually low
Mg-number of 91.6 mol% and a rather high CaO con-
tent of 0.71 wt%.
Four orthopyroxenes released from two diamonds,
one with a lherzolitic garnet the other one with opx only,
show very high TiO2 concentrations of 0.06 and
0.09 wt% respectively. Also, orthopyroxene MW88±
 39

10 83 84 85 86 87 88 89 90 91 92 93 94 95 96
Lherzolitic 40
Harzburgitic %
Olivine
World-wide 30
8 Mwadui
World-wide
20

6
10
CaO

0
4 30
% Opx

2 20

10
0
0 10 20
Cr2O3
0
Fig. 3 Diagram of CaO versus Cr2O3 (wt%) in garnet with the
lherzolite ®eld (shaded area) as de®ned by Sobolev et al. (1973) Cpx
40
%

126a, which touches a harzburgitic garnet, has an un- 20

usually high NiO content of 0.22 wt% (Table 2) and
falls outside the world-wide database, while the rest of
the Mwadui opxs clusters between 0.12±0.15 wt% NiO. 0

Garnet
20
Clinopyroxene %

The chemical composition of the single omphacitic in-

clusion recovered falls within the range of eclogitic
clinopyroxene inclusions world-wide except for a TiO2
content of 0.83 wt% which lies at the high end of the
database range.
The seven peridotitic clinopyroxenes (from four dia-
monds) have a Ca2Si2O6 component between 43.8±
45.7 mol%, with the majority falling above the
45 mol% which separate augites from diopsides (Mori-
moto 1988). These Mwadui clinopyroxenes, therefore,
form part of the calcium-rich suites in our world-wide
database, indicating formation at slightly lower tem-
peratures than usual. The cpx with the highest enstatite
component also shows an extreme Ni content of
0.12 wt%. Three clinopyroxenes (released from two di-
amonds) have uncommonly high K2O contents of 0.31±
0.35 wt%, indicating not only a K-rich environment but
also crystallisation at unusually high pressures (Erlank
and Kushiro 1970; Harlow 1997), (see later).
10
0
83 84 85 86 87 88 89 90 91 92 93 94 95 96
100 Mg
Mg + Fe
Fig. 4 Histogram (frequency in %) of the molar Mg-numbers of
olivine, orthopyroxene, clinopyroxene, and garnet inclusions from
Mwadui and from world-wide sources (world-wide inclusion data
base: 217 olivines, 135 opx, 40 peridotitic cpx, 196 peridotitic garnets).
The unusually high proportion of lherzolitic inclusions for Mwadui is
well displayed in generally low Mg-numbers, relative to the world-
wide database. For lherzolitic inclusions from Mwadui the mode in
Mg-number increases slightly from olivine (92) to cpx (93±93.5),
whereas the world-wide database shows identical modes at 92.5. Since
distribution coecients become more equal with increasing temper-
ature, this implies on average a slightly lower temperature of
formation for Mwadui lherzolitic diamonds than commonly observed

Chromite
Elemental abundances for chromites (ten inclusions re-
leased from eight diamonds) fall completely within the
ranges de®ned previously. They tend to plot towards the
Cr2O3- and FeO-rich and MgO- and Fe2O3-poor side of
the database (Fe2+/Fe3+ calculated from stoichiometry
after Droop 1987).
Sulphide
One lone inclusion of Fe-monosulphide and four inclu-
sions of Fe-Ni-monosulphide (21±22 wt% NiO, two
coexisting with silicate inclusions of peridotitic para-
genesis) were recovered. Deines and Harris (1995) have
shown that assignment of Ni-rich monosulphides to the
peridotitic paragenesis is ambiguous without further
evidence from coexisting silicate inclusions, since strong
40

0.10 the diamond and surrounded only by small graphite

Lherzolitic coated fractures thus excluding an epigenetic origin. In-
Harzburgitic version of coesite to quartz during decompression is in-
0.08 dicated and is conceivably related to the formation of the
small disc shaped fractures surrounding the inclusion.
The coexistence of a SiO2-phase (coesite is normally
0.06 associated with the eclogitic inclusion suite) and harz-
burgitic garnet (peridotitic suite) within one diamond
CaO

could indicate repeated stages of diamond growth,

0.04 connected with a strong change in chemical environ-
ment. Mixed (eclogitic and peridotitic) inclusion para-
geneses are very rare but have been previously observed
0.02 (Hall and Smith 1985; Moore and Gurney 1989; Otter
and Gurney 1989; McDade and Harris 1996). Kopylova
et al. (1997) discovered a SiO2-phase, harzburgitic garnet
0
90 91 92 93 94 95 and olivine within a single diamond but did not ®nd
Mg# evidence for successive and distinct growth stages by
cathodoluminescence. They argued a case of ultrahigh
Fig. 5 CaO versus Mg-number plotted for olivine inclusions from pressure origin of this inclusion paragenesis. We suggest
Mwadui
alternatively that the coexistence of harzburgitic garnet
and free SiO2 is due to successive carbonation reactions,
which ultimately produce calcic magnesite and SiO2
indications for a separate sulphide inclusion paragenesis from a peridotitic bulk composition (Wyllie and Huang
exist. The variable Cr2O3 contents (Table 3) of the sul- 1976). The required extreme CO2 in¯ux can only be
phide inclusions cover the entire previously known range envisaged as a localised phenomenon, such as in highly
from below 0.01 up to 0.70 wt%, otherwise they are in carbonated metasomatic veins, or related to a block of
good agreement with the data previously published on subducted sediments, as suggested by Schrauder and
sulphide inclusions from Tanzania (Deines and Harris Navon (1993). In such a scenario the occurrence of oli-
1995). vine as an additional inclusion phase (Kopylova et al.
1997) relates source carbonation and diamond forma-
tion, since olivine would need to be encapsulated before
Rare or unusual inclusions it is completely eliminated.

SiO2-phase plus harzburgitic garnet

A SiO2-phase inclusion was released from a diamond
together with a typical harzburgitic garnet (1.43 wt%
CaO, MW1±1, Table 2). Since both, the colourless SiO2
inclusion with distinct crystallographical faces and the
purple garnet, were realised as distinct entities before
breakage of the diamond, contamination may be ex-
cluded. Single crystal X-ray di€raction showed the SiO2-
phase to be quartz. The inclusion was located centrally in
Dolomite
The single dolomite [100Ca/(Ca+Mg) equals 49.2 mol%]
inclusion contained minor amounts of iron (1.26 wt%
FeO), manganese (0.74 wt% MnO) and silica (0.64 wt%
SiO2). Carbonate inclusions in diamond have previously
been described (calcite by Meyer and McCallum, 1986,
and magnesite by Phillips and Harris, 1995, and Wang
et al., 1996) but we report here the ®rst ®nding of do-

Table 3. Microprobe analyses of sulphides and a native iron inclusion from Mwadui diamonds

Mineral FeNiS FeNiS FeS FeNiS FeNiS Fe

Sample MW33±56 MW49±76 MW57±83 MW58±85 MW95±135 MW96±137b
Assemblage Ol,chr,sulph Sulph Sulph Ol,sulph Sulph Mag,wu-Fe

Si 0.02 0.02 0.02 £0.01 0.02 0.05

Al £0.01 £0.01 £0.01 £0.01 £0.01 0.02
Cr 0.70 0.17 £0.01 0.45 0.26 0.09
Fe 40.33 41.04 58.50 41.15 40.18 98.04
Mn £0.02 £0.02 £0.02 £0.02 0.03 0.03
Ni 20.88 21.83 1.26 21.96 21.34 0.15
Mg 0.03 £0.01 0.02 £0.01 0.02 £0.01
Co 0.26 0.41 0.23 0.41 0.37 £0.02
Cu 0.74 1.13 0.86 0.74 1.98 0.24
S 35.47 36.13 38.65 36.31 35.86 £0.02
Total 98.42 100.72 99.54 101.02 100.06 98.61
 41

lomite as an inclusion phase. Although the inclusion did (1981) observed an inclusion of native iron rimmed by
not appear to be connected to any surface fracture, the magnetite and also attributed the magnetite rim to epi-
diamond from which the dolomite inclusion was re- genetic oxidation. However, the two Mwadui diamonds
leased was a fractured dodecahedron and thus an epi- from which magnetite was released did not have any
genetic origin cannot not entirely be excluded. On the external fractures. Both magnetites are almost identical
other hand, the occurrence of dolomite ®ts into a sce- in composition with SiO2, Cr2O3, MnO, and CuO being
nario of high, localised CO2 in¯ux where the extent of the only impurities (Table 2). Also the very low con-
carbonation reactions has gone as far as to eliminate centrations of TiO2, MgO (both below detection limit),
clinopyroxene by the reaction 2 Mg2SiO4 + CaMg- and Al2O3 di€er markedly from the possibly epigenetic
Si2O6 + 2 CO2 ˆ CaMg(CO3)2 + 4 MgSiO3 (Eggler magnetites of Meyer and McCallum (1986).
1978). Chemical traverses across the magnetite inclusions
show large and opposing variations of silica and chro-
mium (Fig. 6). This zigzag pattern and the distribution
Geikielite-ilmenite-eskolaite solid solution pattern of silica as obtained from X-ray mapping
(Fig. 7) indicate exsolution of an originally homoge-
An ilmenite group mineral inclusion (MW38±64, Ta-
ble 2) was observed occurring with a harzburgitic garnet 1.50
and a Ca-poor olivine. Expressed in terms of end- wt% Cr2O3
member compositions the ilmenite group mineral is SiO2
47.5 mol% geikielite (MgTiO3), 35.7 mol% ilmenite 1.00
(FeTiO3), 14.5 mol% eskolaite (Cr2O3) and 1.3 mol%
Al2O3, with virtually no haematite component. Al-
though eskolaite is very soluble in geikielite-rich end- 0.50
members of the ilmenite group, Haggerty (1991) has
shown that the level rarely exceeds 10 mol%. Hence the
chemical composition of this inclusion is most unusual. 0
8 16 24 32 40 48 56 µm 64
Extensive substitution of chromium into ilmenite re- 0
quires high pressures and is temperature dependent Fig. 6 Analytical pro®le (spacing 8 lm) across magnetite inclusion
(Green and Sobolev 1975; Haggerty 1991). For upper MW44±72, beginning and ending 5 lm from the inclusion rim
mantle derived ilmenites high chromium contents are
characteristic for ilmenite-spinel (‹rutile) assemblages
formed by subsolidus reduction (Haggerty 1991).
Ilmenite is very rare as an inclusion in diamond with
only nine analyses being reported (compilation of
Chinn, 1995). Of these, only four have chromium above
the detection limit (0.44±3.33 wt% Cr2O3) with the il-
menite richest in chromium being classi®ed as poten-
tially epigenetic by Meyer and McCallum (1986). Thus
the ilmenite group inclusion described here has an un-
precedented chromium content not only among dia-
mond inclusions but also among ilmenite xenocrysts in
kimberlites (Haggerty 1991).

Magnetite, wustite, and native iron

Two dodecahedral Mwadui diamonds contained iron /

iron-oxide inclusions associated with black central
fracture systems. In one diamond magnetite occurred as
a single inclusion (MW44±72, Table 2) whilst in the
second, a separate magnetite was associated with an
inclusion of native iron rimmed by wustite (MW96±136
and ±137a/b, Tables 2 and 3). All three inclusions from
the two diamonds are perfect spheres and vary in size Fig. 7 Image of the distribution of Si in magnetite inclusion MW96±
between 70 and 80 lm. 136. A matrix of 200 ´ 200 points (0.1 seconds per point) was
Magnetite inclusions are rare and of unknown para- collected as beam scan at 15 kV and 20 nA. The contrast was
enhanced after data collection by histogram equalisation, for noise
genesis (Meyer 1987). Meyer and McCallum (1986) reduction a median ®lter was applied. Bright colours imply high Si
suggested that magnetite inclusions precipitate from contents. The grain boundaries can be seen in the four corners of the
¯uids invading along cracks into diamond. Sobolev et al. image. Grain diameter is about 70 lm
42
neous magnetite high in SiO2. Silica contents up to
2.4 wt% were observed which may underestimate the
true silica content of the exsolved mineral, solely because
of the spatial resolution of the electron microprobe.
Under the experimental conditions chosen, the electron
beam excited a volume of about 2 lm in diameter. Ex-
solution lamellae well below 1 lm across, therefore,
would be analysed with a very large overlap of the sur-
rounding magnetite matrix. Silica becomes soluble in
magnetite as the olivine c-phase at high pressures with c-
fayalite requiring lower pressures than c-forsterite
(Katsura and Ito 1989; Woodland and O'Neill 1993).
New experimental data on the binary system Fe2SiO4-
Fe3O4 (Woodland 1997) indicate that at pressures above
30 kbar (at 1100 °C) an intermediate spinelloid phase
appears. Assuming that the spinelloid phase is not re-
stricted to Mg-free systems, this result excludes coexis-
tence of magnetite and olivine in the diamond stability
®eld. Woodland (1997) ®nds that at pressures within the
diamond stability ®eld magnetite in equilibrium with
spinelloid contains at least 20 mol% fayalite component
in solid solution. Thus the observed SiO2 contents of up
to 2 wt% actually are surprisingly low for magnetite
inclusions in diamond and indicate that magnetite was
not in equilibrium with a peridotitic environment.
Diamond MW96 also contained a second inclusion,
consisting of a spherical core of native iron (40 lm di-
ameter) rimmed by wustite (20 lm thick). As can be seen
from Fig. 8 the wustite rim is of constant thickness with
a straight, sharp contact to the iron core, which would
strongly suggest that secondary oxidation is not re-
sponsible for the boundary. The native iron core of the
inclusion is extremely pure, with only trace amounts of
Cu (0.24 wt%), Ni (0.15 wt%) and Cr (0.09 wt%). Co-
balt was not detected in contrast to the Yakutian iron
inclusions (Co 0.14±0.10 wt%, Sobolev et al. 1981). The
wustite rim is signi®cantly di€erent from the minor and
trace element composition of the core (Cr2O3 1.74 wt%,
SiO2 0.83 wt%, MnO 0.78 wt% and Al2O3 0.12 wt%),
indicating a distinct and primary formation of this rim.
Native iron has previously been described as inclu-
sion in diamonds from Yakutia (Sobolev et al. 1981) and
the Sloan kimberlites (Meyer and McCallum 1986).
Ballhaus (1995) evaluated the decrease in fO2 and silica
Fig. 8 Inclusion MW96±137a/b consisting of a core of native iron
rimmed by wustite. Diameter of the inclusion is 80 lm. The dark
cavities in the inclusion originate from polishing of the sample
activity with increasing pressure and predicted that Fe
metal saturation in the upper mantle may occur at
pressures around 90 kbar. The observed low Ni content
of 0.15 wt% of the Mwadui metal inclusion seems to
contradict equilibrium with normal mantle peridotite,
with Ni contents e.g. in olivine of about 0.3 wt%.
However, Sobolev et al. (1981) also observed low Ni
contents (0.06±0.15 wt%) in three native iron inclusions
from Yakutia coexisting with subcalcic pyrope +
chromite ‹ olivine ‹ sulphide.
Wustite inclusions of unknown paragenesis are de-
scribed by Bulanova (1995) as central inclusions (seeds)
in Yakutian diamonds, magnesiowustite (7.29 wt%
MgO) from Koefontein is ascribed to ultradeep
(>400 km) and highly reduced source regions by Moore
et al. (1986). Apart from any depth implications, the
presence of native iron, wustite and magnetite in the
same diamond indicates a variation in oxygen fugacity
from below WI (FeO-Fe) to above MW (Fe3O4-FeO).
This contradicts the suggestion of Bulanova (1995) that
diamond formation takes place around WI but is con-
sistent with the fact that along cratonic geothermal
gradients (40 mW/m2) the MW-bu€er represents fO2
conditions below or around DCO (diamond-carbon
oxide) for pressures up to about 60±70 kbar.
Ferropericlase
Ferropericlase inclusions in diamond are very rare, but
have been described from Orroroo (Scott Smith et al.
1984), Koefontein (Scott Smith et al. 1984; Moore et
al. 1986) Monastery (Moore and Gurney 1989), Sloan
(Otter and Gurney 1989), Sao Luiz (Wilding 1990),
Letseng-la-terrai (McDade and Harris 1996), and River
Ranch (Kopylova et al. 1997).
From Mwadui a single, cubo-octahedral, brown fer-
ropericlase inclusion (Table 2, MW56±82) was released
from a strongly etched, irregular shaped diamond. The
inclusion was partly graphite coated and surrounded by
small disc shaped fractures. With a molar Mg-number
(total iron as FeO) of 85.9 mol% this ferropericlase lies
just above the large compositional range described for
Mg-wustite and ferropericlase from Sao Luiz (38.5±
85.5 mol%) but within the narrow cluster formed by
ferropericlase from other known localities (85.5±
89.0 mol%). Compared to the world-wide database the
inclusion is poor in Cr2O3 (0.16 wt%) but high in NiO
(1.27 wt%).
The potential implication of ferropericlase as an in-
clusion in diamond, i.e. the possible derivation from the
lower mantle, was ®rst recognised by Scott Smith et al.
(1984). Data on the partitioning of Mg, Fe, Ni, Mn and
Cr between ferropericlase and magnesian silicate per-
ovskite (Kesson and Gerald 1992) indicate that dispro-
portionation of typical mantle olivine (Fo 90) would
indeed produce ferropericlase with a Mg-number
around 85 mol% together with the observed trace ele-
ment characteristics. However, ferropericlase by itself is
 43

not necessarily an indication for an ultrahigh pressure Harzburgitic or peridotitic unspecified

origin. It may also be the result of locally reducing Lherzolitic Presumably lherzolitic
conditions and low silica activity at shallower mantle
depth, i.e. in the stability ®eld of olivine. A possible
900 1000 1100 1200 1300 1400
scenario would be the reduction of originally magnesite 2
bearing dunite to below the EMOD bu€er (Eggler and Opx-gt
MW7
Baker 1982) which would result in an olivine-ferroper- [Harley]
iclase-diamond paragenesis. 1

Geothermobarometry
0
Information on equilibration temperatures for mineral 3
Opx-gt [BKN]
inclusions in Mwadui diamonds was obtained from the MW7
various geothermometers listed in Fig. 9. The results, 2
also shown in Fig. 9 assume a pressure of 50 kbar. From
garnet-opx pairs temperatures may be calculated either 1
by using the Mg-Fe exchange (Harley 1984) or by ap-
plying the garnet-opx barometer (based on the Al-ex- 0
change) of Brey and KoÈhler (1990) as a thermometer. 4
The results of both computations agree very well, bearing Ol-gt [O'Neill]
in mind that the application of the garnet-opx barometer 3
as a thermometer will result in rather large errors. Peri-
2
dotitic garnets may be assigned to lherzolitic or harz- MW11
burgitic parageneses according to their CaO and Cr2O3 1
contents (Fig. 3). The garnet released from diamond
0
MW7 falls compositionally on the line separating the
harzburgitic and lherzolitic ®elds in this diagram. Com- 4
Ca in Ol [KB]
bined with coexisting opx, this garnet gives a high 3
equilibration temperature typical for the Mwadui MW11
lherzolitic paragenesis (around 1250±1300 °C, Fig. 9), as 2
compared to relatively low temperatures observed for the
harzburgitic paragenesis (around 1000±1150°C). Thus, 1
this garnet may well belong to the lherzolitic paragenesis. 0
For olivine-garnet pairs temperatures estimated from
3
the Mg-Fe exchange (O'Neill and Wood 1979; O'Neill Cpx [BKN]
1980) show a large range, up to very high temperatures,
2
but with a mode at 1150 °C. Of the two observed
lherzolitic garnet-olivine pairs one yields the highest
temperature and one falls onto the mode of the harz- 1
burgitic paragenesis.
Assuming crystallisation in equilibrium with clino- 0
pyroxene the Ca content of olivine may be used to cal- 10
culate temperatures (KoÈhler and Brey 1990). From 8
Ca in Opx [BK]
coexisting phases (garnet or cpx), ®ve olivines de®nitely
belong to the lherzolitic paragenesis. It is apparent from 6
Fig. 5 that all olivines with Mg-numbers below 92.1 (the 4
upper limit of lherzolitic olivines from Mwadui) have Ca 2
contents at least as high as the olivines of proven Cpx-Opx intergrowth
lherzolitic paragenesis. These eleven olivines are shown 0
900 1000 1100 1200 1300 1400
with a shaded pattern in Fig. 9. Compared to the garnet-
olivine temperatures the Ca content in olivine gives an Fig. 9 Thermometry of silicate inclusions from Mwadui. (Harley
identical mode (Fig. 9) but a less extreme range of Harley 1984, BKN Brey and KoÈhler 1990; O'Neill O'Neill and Wood
temperatures. If any of the olivines did not form in 1979 and O'Neill 1980, KB KoÈhler and Brey 1990, BK Brey and
KoÈhler 1990)
equilibrium with cpx, then calculated temperatures are
minimum values. However, four olivines of demonstra-
ble lherzolitic paragenesis are in the low temperature tail Clinopyroxene inclusions in equilibrium with ortho-
of the distribution indicating that systematic underesti- pyroxene enable the application of the two-pyroxene
mation of temperatures is unlikely. thermometer, also of Brey and KoÈhler (1990). Where no
44
coexisting orthopyroxene was observed equilibrium was
assumed and an estimated chemical composition (the
average of lherzolitic opx) was used. The possible errors
introduced by assuming an orthopyroxene composition
are small, since calculated temperatures mainly depend
on the composition of the clinopyroxene. As Fig. 9
shows, clinopyroxene thermometry results in a narrow
temperature distribution around a mode of 1050 °C.
The Ca content of orthopyroxene crystallised in the
presence of clinopyroxene also may be used as a ther-
mometer (Brey and KoÈhler 1990). Four orthopyroxene-
clinopyroxene inclusion pairs were recovered from
Mwadui diamonds. Ten more orthopyroxenes were as-
signed to the lherzolitic paragenesis by excluding all in-
clusions with a Mg-number above 94.25, as well as all
those coexisting with harzburgitic garnets. The resulting
temperature distribution is very narrow and the mode is
identical (1050 °C) with the one observed for clinopy-
roxene. One rather high equilibration temperature was
estimated from a clinopyroxene-orthopyroxene inter-
growth and may be due to later (syn-eruptive?) re-
equilibration.
In addition to the temperatures derived from silicate
inclusions the trace element thermometer Zn in spinel
(Grin et al. 1994; Ryan et al. 1996) may be applied.
Eight diamonds with chromite inclusions yielded Zn-
temperatures between 1166±1238 °C, with an average of
1198 ‹ 25 °C (not shown in Fig. 9).
Comparing the results of the various thermometers
substantial inconsistency becomes apparent. As exam-
ples (Fig. 9), clinopyroxenes and orthopyroxenes indi-
cate that crystallisation of lherzolitic diamonds occurred
at rather low temperatures around 1050 °C. The Ca
contents in olivine, however, indicate crystallisation
temperatures for lherzolitic diamonds with a mode at
1150 °C and a tail up to about 1300 °C. Olivine-garnet
temperatures, derived mainly from pairs of harzburgitic
paragenesis, have the same mode, but range to even
higher temperatures.
The di€erence in the temperatures calculated for the
lherzolitic garnet-olivine pair released from diamond
MW11 (1402 °C from the Mg-Fe exchange and 1246 °C
from Ca in olivine) indicates disequilibrium between
olivine and garnet, possibly explaining the large range in
olivine-garnet temperatures. Disequilibrium between
olivine and garnet can also be demonstrated for dia-
mond MW88, which contains an olivine and a garnet
inclusion plus a touching garnet-opx pair. Combination
of the olivine inclusion with the single and the touching
garnet gives consistent temperatures of 1365 and
1369 °C respectively. The combination of the opx in-
clusion with both garnets, however, produces tempera-
tures of 983 and 982 °C. Both thermometers (garnet-
olivine and garnet-opx) are based on Fe-Mg exchange
and give consistent results for both garnets, implying
that re-equilibration of the touching garnet-opx pair
towards lower temperatures has not occurred. Thus, the
large temperature di€erence is due to the Fe-rich olivine
inclusion (Mg-number of 92.2) not being in equilibrium
with the Mg-rich opx (Mg-number of 95.0). Using the
Mg-numbers as indicators of paragenesis (Harris 1992),
the opx is harzburgitic, as are the garnet inclusions,
whereas the olivine is lherzolitic. Thus the inclusion of
the olivine represents a distinct stage of diamond
growth, which occurred after the inclusion of garnet and
opx.
Harzburgitic garnet-opx pairs yield temperatures that
are comparable with the mean temperatures of
lherzolitic diamonds derived from olivine, clinopyroxene
and orthopyroxene thermometry. Compared to the
temperatures for harzburgitic olivine-garnet pairs, ha-
rzburgitic garnet-opx pairs have slightly lower temper-
atures. The three or four (Fig. 9) lherzolitic garnet-opx
pairs appear to indicate a high temperature origin, but
this value is inconsistent with the Ca content of the or-
thopyroxenes involved. Again disequilibrium between
coexisting inclusions is demonstrated, which may have
two possible causes: (1) The garnet compositions clearly
indicate a lherzolitic paragenesis, but this may not apply
to the orthopyroxenes, which would then mean that the
opx is not Ca-saturated and in consequence the tem-
peratures derived from Ca in opx are only minimum
ones. (2) Similarly with the garnet-olivine disequilibria,
the orthopyroxenes re¯ect an environment with a much
lower Mg/Fe ratio than the apparently coexisting gar-
net, resulting in temperature overestimation using the
Mg-Fe exchange thermometer.
In view of common evidence for disequilibrium,
combinations of thermometers and barometers have not
been used to calculate PT conditions by iterative meth-
ods for the Mwadui inclusion suite.
Implications for diamond formation
and mantle metasomatism
Major and rare earth element characteristics of mineral
inclusions in diamonds (e.g. Stachel and Harris 1997b)
document an evolution of the surrounding in which
diamond grows in at least two stages: an early melt
extraction event followed later by metasomatic re-en-
richment. Evidence for disequilibrium among Mwadui
inclusions is ubiquitous which implies steep composi-
tional gradients during diamond formation, as they
could evolve in progressing metasomatic fronts. De-
creasing Mg/Fe ratios of mineral inclusions from the
centre towards the rim within single host diamonds (e.g.
Bulanova 1995) support our interpretation that mixed
harzburgitic-lherzolitic inclusion parageneses form
during progressive metasomatic enrichment of iron and
calcium thus relating source enrichment and diamond
formation.
Subsolidus equilibration temperatures for harzburgi-
tic inclusion parageneses (Boyd and Gurney 1986;
Grin et al. 1992; Stachel and Harris 1997b) and the
carbon isotopic composition of diamond (Deines and
Harris 1995) indicate that metasomatic re-enrichment
occurred by percolating methane-rich ¯uids from which

diamond was precipitated during redox reactions oc-
curring in contact with more oxidised lithospheric
peridotite (Taylor and Green 1989; Stachel and Harris
1997b). A lower solidus temperature and on average
higher equilibration temperatures for lherzolitic inclu-
sions from southern Africa (Grin et al. 1992) and
Ghana (Stachel and Harris 1997b) indicate that melt
involvement during metasomatic enrichment and dia-
mond formation may also occur. However, transitions
from harzburgitic to lherzolitic inclusion parageneses
within single Mwadui diamonds contradict separate
growth environments at least for this speci®c source.
Concluding remarks
From their mineral inclusion content Mwadui diamonds
can be grouped into peridotitic (88%), eclogitic (2%),
mixed (low-Ca pyrope and SiO2-phase, 1%) and un-
known (sulphides, native iron and iron oxides, ferro-
periclase, dolomite, 9%) suites. Inclusion chemistry
shows Mwadui diamonds to re¯ect sampling of a
lithosphere with an unusually high proportion of
undepleted or re-fertilised mantle peridotite. This is
documented mainly by the high proportion of calcium-
saturated peridotitic garnets as well as by the low
Mg-numbers of olivines and orthopyroxenes. High Ni
contents in olivine could suggest that this fertility actu-
ally represents re-enrichment of a previously depleted
source. An explanation may lie in the positive correla-
tion of the proportion of orthopyroxene and the Ni
contents in olivine from peridotitic xenoliths. Kelemen
and Hart (1996) interpret this as a metasomatic reaction
which creates orthopyroxene at the expense of olivine
(which increases the Ni content of the remaining oli-
vine). Metasomatic enrichment processes may also be
indicated by high titanium contents in garnet and or-
thopyroxene and high potassium in clinopyroxene.
Application of geothermometers based on garnet-
olivine, garnet-opx and cpx-opx equilibria indicate that
the mode of equilibration temperatures for Mwadui di-
amonds lies between 1050±1150 °C. However, the ap-
plication of various thermometers to the same inclusion-
pairs reveals evidence for common disequilibrium be-
tween non-touching inclusions. High temperatures cal-
culated for garnet-opx and garnet-olivine pairs may be
attributed to diamond formation in the course of a
metasomatic enrichment event whereby successive in-
clusions become richer in Fe. Further indications of
large variations in the chemical environment during di-
amond growth may come from the coexistence of a
SiO2-phase and harzburgitic garnet in the same dia-
mond. However, localised extreme carbonation of
mantle peridotite could eventually also result in the
formation of a free SiO2-phase.
In addition to the standard peridotitic mineral suite
Mwadui diamonds contain a number of rare or unusual
inclusions. Most noticeable among these is an ilmenite
45
group inclusion with unprecedented eskolaite content
(14.5 mol%) occurring together with harzburgitic garnet
and olivine, and magnetite inclusions with sub-lm ex-
solution lamellae of a high silica phase (possibly fayalite)
indicative of primary solution of the olivine c-phase
within the magnetite.
The presence of two diamonds, one containing in-
clusions of moderately majoritic eclogitic garnet, the
other of ferropericlase, may indicate a small diamond
contribution from the sub-lithosphere. Although the
evidence of a single ferropericlase inclusion is ambigu-
ous, its presence might testify a lower mantle origin. An
inclusion of native iron rimmed by wustite (plus a co-
existing inclusion of magnetite) also might be indicative
for an ultra-deep origin and indicates rather large vari-
ations in oxygen fugacity from below WI (FeO-Fe) to
above MW (Fe3O4-FeO).
To account for the disequilibrium conditions during
diamond formation beneath Mwadui, we conclude that
diamonds grew in metasomatic fronts, where steep
compositional gradients existed and where carbon was
deposited as diamond from redox reactions between the
metasomatic ¯uid or melt and the peridotitic wall rocks.
Acknowledgements Colin Farrow (Glasgow) is thanked for gener-
ating the microprobe image shown in Fig. 7. Werner Joswig's
(Frankfurt) identi®cation of the SiO2-phase inclusion by single
crystal X-ray di€raction is acknowledged. Barry Dawson (Edin-
burgh) kindly provided information on volcanism in Tanzania and
the formation of the East Africa Rift. T.S. received funding of the
European Union under the Human Capital and Mobility scheme
(Marie Curie Fellowship). Financial support of the project by
DeBeers Consolidated Mines Ltd. and the Department of Geology
and Applied Geology, University of Glasgow is gratefully ac-
knowledged. F.R. Boyd and B. Harte are thanked for thoughtful
reviews
References
Ballhaus C (1995) Is the upper mantle metal-saturated? Earth
Planet Sci Lett 132: 75±86
Boyd FR, Gurney JJ (1986) Diamonds and the African lithosphere.
Science 232: 472±477
Brenker FE, Brey GP (1997) Reconstruction of the exhumation
path of the Alpe Arami garnet-peridotite body from depths
exceeding 160 km. J Metamorphic Geol 15: 581±592
Brey GP (1989) Geothermobarometry for lherzolites: experiments
from 10 to 60 kbar, new thermobarometers and application to
natural rocks. Habilitation thesis, Technische Hochschule,
Darmstadt
Brey GP, KoÈhler T (1990) Geothermobarometry in four-phase
lherzolites. II. New thermobarometers, and practical assessment
of existing thermobarometers. J Petrol 31: 1353±1378
Brey GP, KoÈhler T, Nickel KG (1990) Geothermobarometry in
four-phase lherzolites. I. Experimental results from 10 to
60 kbar. J Petrol 31: 1313±1352
Bulanova GP (1995) The formation of diamond. J Geochem
Explor 53: 1±23
Chinn IL (1995) A study of unusual diamonds from the George
Creek K1 kimberlite dyke, Colorado. PhD thesis, Univ Cape
Town, RSA
Davis CL (1977) The ages and uranium contents of zircons from
kimberlites and associated rocks. Carnegie Inst Washington
Yearb 76: 631±654
46
Dawson JB (1992) Neogene tectonics and volcanicity in the North
Tanzania sector of the Gregory Rift Valley ± contrasts with the
Kenya sector. Tectonophysics 204: 81±92
Deines P, Harris JW (1995) Sul®de inclusion chemistry and carbon
isotopes of African diamonds. Geochim Cosmochim Acta 59:
3173±3188
Droop GTR (1987) A general equation for estimating Fe3+ concen-
trations in ferromagnesian silicates and oxides from microprobe
analyses, using stoichiometric criteria. Geol Mag 51: 431±435
Eggler DH (1978) The e€ect of CO2 upon partial melting of
peridotite in the system Na2O-CaO-Al2O3-MgO-SiO2-CO2 to
35 kbar, with an analysis of melting in a peridotite-H2O-CO2
system. Am J Sci 278: 305±343
Eggler DH, Baker DR (1982) Reduced volatiles in the system C-O-
H: implications for mantle melting, ¯uid formation and dia-
mond genesis. In: Akimoto S, Manghnani MH (eds) High
pressure research in geophysics. Cent Acad Publ, Tokio, pp
237±250
Erlank AJ, Kushiro I (1970) Potassium contents of synthetic py-
roxenes at high temperatures and pressures. Carnegie Inst
Washington Yearb 68: 439±443
Green DH, Sobolev NV (1975) Coexisting garnets and ilmenites
synthesized at high pressures from pyrolite and olivine basanite
and their signi®cance for kimberlitic assemblages. Contrib
Mineral Petrol 50: 217±229
Grin WL, Gurney JJ, Ryan CG (1992) Variations in trapping
temperatures and trace elements in peridotite-suite inclusions
from African diamonds: evidence for two inclusion suites, and
implications for lithosphere stratigraphy. Contrib Mineral
Petrol 110: 1±15
Grin WL, Ryan CG, Gurney JJ, Sobolev NV, Win TT (1994)
Chromite macrocrysts in kimberlites and lamproites: geo-
chemistry and origin. In: Meyer HOA, Leonardos OH (eds)
Kimberlites, related rocks and mantle xenoliths. CPRM Spec
Publ 1/A Jan/94, Brasilia, pp 366±377
Gobba JM (1991) The geology and mineralogy of some kimberlites
in the Mwadui area. 5(abstract) In: 5th Int Kimberlite Conf
Araxa Extended Abstr. CPRM Spec Publ 2/91, pp 116±118
Haggerty SE (1991) Oxide mineralogy of the upper mantle. In:
Lindsley DH (ed) Oxide minerals: petrology and magnetic sig-
ni®cance. (Reviews in mineralogy, 25) Mineral Soc Am,
Washington, DC, pp 355±416
Hall AE, Smith CB (1985) Lamproite diamonds ± are they dif-
ferent? In: Glover JE, Harris PG (eds) Kimberlite occurrence
and origin. Publ Geol Dep Univ Ext, Univ W Aust 8, pp 167±
212
Harley SL (1984) An experimental study of the partitioning of iron
and magnesium between garnet and orthopyroxene. Contrib
Mineral Petrol 86: 359±373
Harlow GE (1997) K in clinopyroxene at high pressure and tem-
perature: an experimental study. Am Mineral 82: 259±269
Harris JW (1992) Diamond geology. In: Field JE (ed) The prop-
erties of natural and synthetic diamond. Academic Press,
London, pp 345±393
Harris JW, Gurney JJ (1979) Inclusions in diamond. In: Field JE
(ed) The properties of diamond. Academic Press, London, pp
555±594
Irifune T (1987) An experimental investigation of the pyroxene-
garnet transformation in a pyrolite composition and its bearing
on the constitution of the mantle. Earth Planet Sci Lett 45: 324±
336
Janse AJA (1994) Is Cli€ord's rule still valid? Armative examples
from around the world. In: Meyer HOA, Leonardos OH (eds)
Diamonds: characterization, genesis and exploration. CPRM
Spec Publ 1/B Jan/94, Brasilia, pp 215±235
Katsura T, Ito E (1989) The system Mg2SiO4-Fe2SiO4 at high
pressures and temperatures ± precise determination of stabilities
of olivine, modi®ed spinel, and spinel. J Geophys Res 94±11:
15663±15670
Kelemen PB, Hart SR (1996) Silica enrichment in the continental
lithosphere via melt/rock interaction (abstract). In: V.M.
Goldschmidt Conf Heidelberg 1996, J Conf Abstr 1±1, p 308
Kennedy CS, Kennedy GC (1976) The equilibrium boundary be-
tween graphite and diamond. J Geophys Res 81: 2467±2470
Kesson SE, Gerald JDF (1992) Partitioning of MgO, FeO, NiO,
MnO and Cr2O3 between magnesian silicate perovskite and
magnesiowustite ± implications for the origin of inclusions in
diamond and the composition of the lower mantle. Earth Planet
Sci Lett 111: 229±240
KoÈhler T, Brey GP (1990) Calcium exchange between olivine and
clinopyroxene calibrated as geothermobarometer for natural
peridotites from 2 to 60 kbar with applications. Geochim
Cosmochim Acta 54: 2375±2388
Kopylova MG, Gurney JJ, Daniels LRM (1997) Mineral inclusions
in diamonds from the River Ranch kimberlite, Zimbabwe.
Contrib Mineral Petrol 129: 366±384
McDade P, Harris JW (1996) Multiple diamond sources at the
Kalahari Craton Margin: Letseng-la-terai, Lesotho (abstract).
In: V.M. Goldschmidt Conf Heidelberg 1996, J Conf Abstr 1±1,
p 391
Meyer HOA (1987) Inclusions in diamond. In: Nixon PH (ed)
Mantle xenoliths. J Wiley and Sons, Chichester, pp 501±522
Meyer HOA, McCallum ME (1986) Mineral inclusions in dia-
monds from the Sloan kimberlites, Colorado. J Geol 94: 600±
612
Meyer HOA, Svisero DP (1975) Mineral inclusions in Brazilian
diamonds. Phys Chem Earth 9: 785±795
Moore RO, Gurney JJ (1989) Mineral inclusions in diamond from
Monastery kimberlite, South Africa. In: Ross J, et al (eds)
Kimberlites and related rocks. (Geol Soc Aust Spec Publ 14, vol
2) Blackwell, Carlton, pp 1029±1041
Moore RO, Otter ML, Rickard RS, Harris JW, Gurney JJ (1986)
The occurrence of moissanite and ferro-periclase as inclusion in
diamonds (abstract). In: 4th Int Kimberlite Conf Perth 1986,
Extended Abstr, pp 409±411
Morimoto N (1988) Nomenclature of pyroxenes. Mineral Mag 52:
535±550
Mvuemba Ntanda F, Moreau J, Meyer HOA (1982) ParticulariteÂs
des inclusions cristallines primaires des diamants du Kassai,
ZaõÈ re. Can Mineral 20: 217±230
O'Neill HStC (1980) An experimental study of the iron-magnesium
partitioning between garnet and olivine and its calibration as a
geothermometer: corrections. Contrib Mineral Petrol 72: 337
O'Neill HStC, Wood BJ (1979) An experimental study of the iron-
magnesium partitioning between garnet and olivine and its
calibration as a geothermometer. Contrib Mineral Petrol 70:
59±70
Otter ML (1990) Diamonds and their mineral inclusions from the
Sloan diatremes, Colorado-Wyoming State Line kimberlite
district, North America. PhD thesis, Univ Cape Town, RSA
Otter ML, Gurney JJ (1989) Mineral inclusions in diamond from
the Sloan diatremes, Colorado-Wyoming State Line kimberlite
district, North America. In: Ross J, et al (eds) Kimberlites and
related rocks. (Geol Soc Aust Spec Publ 14, vol 2) Blackwell,
Carlton, pp 1042±1053
Phillips D, Harris JW (1995) Geothermobarometry of diamond
inclusions from the De Beers Pool Mines, Kimberley, South
Africa (abstract). In: 6th Int Kimberlite Conf Novosibirsk 1995,
Extended Abstr, pp 441±443
Raber E (1978) Zircons from diamond-bearing kimberlites: oxide
reactions, ®ssion track dating and mineral inclusions study.
MSc thesis, Univ Massachusetts
Ringwood AE, Major A (1971) Synthesis of majorite and other
high pressure garnets and perovskites. Earth Planet Sci Lett 12:
411±418
Ryan CG, Grin WL, Pearson NJ (1996) Garnet geotherms ±
pressure-temperature data from Cr-pyrope garnet xenocrysts in
volcanic rocks. J Geophys Res 101 B3: 5611±5625
Schrauder M, Navon O (1993) Solid carbon-dioxide in a natural
diamond. Nature 365, 6441: 42±44
Scott Smith BH, Danchin RV, Harris JW, Stracke KJ (1984)
Kimberlites near Orroroo, South Australia. In: Kornprobst
J (ed) Kimberlites I: kimberlites and related rocks. Elsevier,
Amsterdam, pp 121±142

Shimizu N, AlleÁgre CJ (1978) Geochemistry of transition elements
in garnet lherzolite nodules in kimberlites. Contrib Mineral
Petrol 67: 41±50
Sobolev NV, Lavrent'ev YuG, Pokhilenko NP, Usova LV (1973)
Chrome-rich garnets from the kimberlites of Yakutia and their
paragenesis. Contrib Mineral Petrol 40: 39±52
Sobolev NV, Ye®mova ES, Koptil VI, Lavrent'yev YuG, Sobolev
VC (1976) Inclusions of coesite, garnet and omphacite in dia-
monds of Yakutia ± ®rst ®nd of coesite paragenesis. Trans Dokl
USSR Acad Sci Earth Sci Sec 230: 1442±1444
Sobolev NV, E®mova ES, Pospelova LN (1981) Native iron in
Yakutian diamonds and its paragenesis (in Russian). Geol
Geophys Akad Nauk SSSR Sib 12: 25±29
Stachel T, Harris JW (1997a) Syngenetic inclusions in diamond
from the Birim Field (Ghana) ± a deep peridotitic pro®le with a
history of depletion and re-enrichment. Contrib Mineral Petrol
127: 336±352
Stachel T, Harris JW (1997b) Diamond precipitation and mantle
metasomatism ± evidence from the trace element chemistry of
silicate inclusions in diamonds from Akwatia, Ghana. Contrib
Mineral Petrol 129: 143±154
Taylor WR, Green DH (1989) The role of reduced C-O-H ¯uids in
mantle partial melting. In: Ross J (ed) Kimberlites and related
47
rocks: their composition, occurrence, origin and emplacement.
(Geol Soc Aust Spec Publ No 14, vol 1) Blackwell, Carlton, pp
592±602
Tsai HM (1978) Mineralogical and geochemical investigations of
mineral inclusions in diamond, kimberlite and related rocks.
PhD thesis, Purdue Univ, West Lafayette, USA
Wang A, Pasteris JD, Meyer HOA, Deleduboi ML (1996) Ma-
gnesite-bearing inclusion assemblage in natural diamond. Earth
Planet Sci Lett 141: 293±306
Wilding MC (1990) (unpublished). PhD thesis. Univ Edinburgh,
UK
Woodland AB (1997) The system Fe2SiO4-Fe3O4: phase relations
to 9 GPa and molar volumes of spinel solid solutions (abstract).
AGU Fall Meet Abstr
Woodland AB, O'Neill HStC (1993) Synthesis and stability of
Fe32+Fe23+Si3O12 garnet and phase relations with Fe3Al2-
Si3O12 - Fe32+Fe23+Si3O12 solutions. Am Mineral 78: 1002±
1015
Wyllie PJ, Huang W-L (1976) Carbonation and melting reactions
in the system CaO-MgO-SiO2-CO2 at mantle pressures with
geophysical and petrological applications. Contrib Mineral
Petrol 54: 79±107