Fig. 4. A Discordant spinifex veins intruding the lower olivine cumulate komatiite zone of a flow. Hammer for scale is 40 cm in length. B Photograph of a thin auto-brecciated bx komatiitic basalt flow. Hammer for scale is 40 cm in length. C Fine-grained
subhedral and coarser grained, more skeletal relict olivine grains Fo
91.7
; Table 3 set in an altered glassy groundmass of an orthocumulate komatiite. Field of view is 6 mm. D Coarse-grained, randomly-oriented serpentine pseudomorphs of skeletal olivine
in the top of a differentiated komatiite flow. Coin for scale is 1 cm in diameter.
5. Geochemical modelling
Possible indications of crustal contamination could be deviations from uncontaminated model
fractionation trends, enrichments in incompatible immobile elements that are significantly enriched
in the continental crust and, perhaps, the presence of igneous hydroxy-amphibole Stone et al.,
1997. In particular, deviations from uncontami- nated model fractionation trends and LREE en-
richments in the LKH and UKH rocks are investigated below.
The immobile elements in MgO variation dia- grams Fig. 5 define three distinct trends. First,
as the content of MgO decreases from 50 to 20 to B
10, FeO
t
increases, then decreases, and finally remain somewhat constant. Second, Ni contents
in the same MgO interval first decrease and then remain relatively constant. Third, the contents of
Al
2
O
3
and TiO
2
increase with decreasing MgO. The olivine incompatible elements define trends
relative to increasing magnesium contents that extrapolate approximately to the compositions of
olivines analysed from dunite and olivine cumu- late samples Table 3. These incompatible ele-
ment trends extrapolate to : 51 MgO at 0 Al
2
O
3
and TiO
2
, and to : 42 SiO
2
, : 5 FeO
t
, :
0.08 MnO, : 4000 ppm Ni, and : 1200 ppm Cr at 51 MgO. All the LKH and UKH geo-
chemical trends together with thin section petro- graphic studies, suggest evolution of LKH and
UKH parental melts and derivative rock types by
Fig. 5. Whole rock geochemistry. Fields: 1, dunite, 2, wehrlite, 3, clinopyroxenite, 4, amphibole clinopyroxenite, 5, cumulate amphibole gabbro, 6, noncumulate amphibole gabbro gabbro liquid, A, olivine cumulate, B, olivine spinifex komatiite liquid, C,
clinopyroxene cumulate and D, clinopyroxene spinifex komatiitic basalt liquid.
liquid-crystal fractionation of olivine to clinopy- roxene to plagioclase.
The geochemical similarities of the LKH and UKH rocks Fig. 5 and their close spatial and
apparent stratigraphic associations are considered to be strong evidence for a genetic relationship
Muir, 1979; Pyke, 1982. Possible genetic rela- tionships between the different LKH and UKH
rock types, and between the LKH and the UKH, have been modelled utilising the fractional crys-
tallisation and assimilation – fractional crystallisa- tion program MELTS Ghiorso and Sack, 1995.
MELTS was chosen because most thermodynamic properties except thermal conductivity are con-
sidered for individual phases and liquid under a wide variety of equilibrium conditions. However,
trace element compositions are not modelled and are therefore considered separately. Prior to de-
scribing the results of the MELTS modelling, an overview of the model assumptions is presented
see also Appendices A and B.
5
.
1
. Liquid compositions The initial komatiitic liquid composition was
estimated utilising variations in whole-rock Mg numbers of the LKH. The highest Mg number
corresponds to dunite \ 90 olivine, sample c
22493; Larson, 1996 and constrains the maxi- mum Mg number of the most primitive cumulate
olivine and the initial liquid. The remaining liquid components were estimated from the geochemical
trends on MgO variation diagrams at 32 MgO, following Lesher 1989 see Nisbet et al., 1993 for
a revised view of such high MgO contents. Thus, the initial liquid composition is estimated to have
contained 32 MgO, 0.3 TiO
2
, 5.5 Al
2
O
3
, 11 FeO
t
, 1750 ppm Ni, 75 ppm Co, and 2700 ppm Cr. This magma would have crystallised
Fo
94.9
olivine assuming K
D
= 0.30 and Fe
3 +
SFe=0.1; Roeder and Emslie, 1970, which is more magnesian than the analysed olivines
Fo
91.7 – 93.1
; Table 3, but similar to that inferred from the olivine adcumulate whole-rock composi-
tions Fo
95
; Table 4. The Na content was derived from the : 28 MgO liquid composition of
Arndt 1986b sample c M666 and adjusted to corresponding values at 32 MgO. The P content
was estimated to be 0.01 P
2
O
5
. The geochemical composition of the estimated initial liquid Table
5 is similar to compositions of other highly mag- nesian komatiitic liquids, except that the FeO
t
content appears to be higher. Clinopyroxenite compositions in the MgO vari-
ation diagrams appear to have formed from derivative liquids with 10 – 15 MgO and the
amphibole gabbros from liquids with 4 – 6 MgO. The presence of igneous amphibole rather than
clinopyroxene indicates that the presence of hy- drous residual liquids. The H
2
O content of these amphibole-supersaturated liquids is assumed to be
3, based on experimental studies of basalt sys- tems Semet and Ernst, 1981; Gilbert et al., 1982.
The H
2
O content of the initial liquid is therefore estimated to be 1 – 2, adjusting these liquid com-
positions for 60 – 70 fractionation. Hence, FC models were run anhydrous and with 1 and 2 wt
H
2
O added to the liquid. However, the results of the anhydrous and hydrous modelling were not
significantly different. The A-FC models were run with the anhydrous estimated initial komatiitic
liquid composition and with 1 water added to the initial komatiite liquid composition.
The UKH olivine and clinopyroxene spinifex textured samples and aphyric samples from mar-
ginal flow breccia zones were preferentially sam- pled and analysed to determine erupted liquid
compositions. Aphyric rocks are interpreted to represent liquid compositions and spinifex tex-
tured rocks represent near-liquid compositions cf. Donaldson, 1982. These samples are charac-
terised by lower MgO contents up to 30 MgO, less cumulus olivine and clinopyroxene and a high
proportion of glassy interstitial areas. The FeMg ratio of the liquid is further constrained by the
composition of relict olivines in three komatiite samples Fo
91.7-93.1
; Table 3 which suggests crys- tallisation from liquid containing up to 30 wt
MgO. The geochemical variation trends suggest that the komatiitic basalts formed from derivative
liquids with B 10 – 15 MgO.
5
.
2
. Contamination and other physical assumptions
The presence of iron-formation xenoliths in a
wehrlite dyke stratigraphically down section of the LKH represents tangible evidence for at least
localised assimilation.
However, calc-alkalic
basalt, andesite and dacite are more abundant than iron-formation and most of the latter is
stratigraphically upsection of the LKH Pyke, 1982. Therefore, iron-formation is not considered
to be a significant contaminant. Average composi- tions of calc-alkalic basalt and dacite mafic and
felsic end-members utilised in the modelling are listed in Table 7. Geochemical modelling indi-
cated no significant change in the liquid composi- tion until assimilation of \ 10 of the mass of
the initial liquid.
If the UKH formed from LKH liquids, the latter may have extruded prior to or following
assimilation of calc-alkalic basalt. In addition, iron-formation assimilation might possibly have
been an important factor in the evolution of the UKH Green and Naldrett, 1981. The average
composition of sulphide iron-formation is listed in Table 6.
The system pressure was constrained from the apparent map-scale thickness of the calc-alkalic
volcanic pile 4 km average; Pyke, 1982 intruded by the LKH and an assumed ocean water depth
of 1 km. The latter depth would be that of UKH extrusion ocean floor.
A lithostatic pressure gradient of 0.3 kbkm and a hydrostatic pressure gradient of 0.1 kbkm
yield a pressure estimate of 1.3 kb.
5
.
3
. Summary One hundred and two MELTS models were
run. The most likely FC and the spectrum of A-FC model liquid evolution trends for the LKH
and UKH rocks are presented in Fig. 6 and discussed below.
5
.
4
. Fractional crystallisation All but one of the LKH FC model runs can
reproduce the most primitive olivine compositions observed \ Fo
94.6
. Although the liquid evolu-
Table 5 Model initial komatiitic liquid compositions
Shaw Dome Kambalda
Dumont Alexo
Yakabinde Duke 1986
Arndt 1986b Duke and Naldrett 1978
Lesher 1983 This study
45.6 45.6
44.9 45.4
47.5 SiO
2
wt 0.27
0.27 TiO
2
0.25 0.35
0.32 5.60
5.60 Al
2
O
3
7.00 5.40
6.50 0.38
Cr
2
O
3
0.43 0.40
0.39 0.30
11.1 FeO
t a
9.6 10.4
10.8 11.1
0.18 0.17
0.2 MnO
0.17 0.17
32.0 MgO
32.0 32.0
27.9 27.5
0.21 0.22
NiO 0.16
0.20 0.14
0.01 0.01
0.01 CoO
0.01 0.01
4.50 5.70
CaO 5.30
6.50 6.00
Na
2
O 0.35
0.26 0.35
0.29 0.15
0.09 0.03
K
2
O na
0.10 na
0.01 0.03
P
2
O
5
na na
na 100.00
100.00 Total
99.57 99.12
99.70 20
Al
2
O
3
TiO
2
21 22
20 20
CaOAl
2
O
3
0.83 1.02
0.95 0.93
0.92 Fo
6b
95.0 95.6
95.3 94.1
94.2
a
FeO
t
, total iron as FeO.
b 6
Fo, equilibrium olivine composition calculated assuming K
D
value for FeOMgO olivinemelt partitioning = 0.30 and 10 Fe
3+
.
Fig. 6. Geochemical modelling results. Solid lines represent the minimum and maximum A-FC model trends as defined by MELTS. Field labels are the same as per Fig. 5. cab, calc-alkaline basalt.
Table 6 Average whole-rock XRF analyses 1 standard deviation of possible contaminants altered in the Shaw Dome
Calc-alkalic Calc-alkalic
Calc-alkalic Sulfide
Basalt Andesite
Dacite Iron-formation
12 4
6 n
3 2.3
61.8 2.6
65.0 0.33
53.5 15.5
SiO
2
wt 59.9
0.141 0.615
0.102 0.352
0.729 0.095
TiO
2
0.079 0.073
Al
2
O
3
15.4 0.76
16.3 0.88
17.1 0.41
2.1 1.1
0.019 Cr
2
O
3
0.010 0.011
0.004 B
0.010 0.003
B 0.010
0.002 1.1
6.3 1.0
FeO
t a
3.3 7.2
0.27 35.8
16.6 0.01
MnO 0.1
0.10 0.03
0.05 0.01
0.19 0.04
0.95 3.3
0.27 2.0
4.5 0.34
MgO 2.6
1.1 0.012
NiO 0.005
0.008 0.002
B 0.010
0.001 0.004
0.002 0.001
CoO 0.000
0.003 0.001
B 0.010
0.000 B
0.010 0.001
1.2 4.3
2.4 CaO
4.9 6.4
1.4 4.8
2.6 Na
2
O 2.5
1.5 3.6
1.2 3.0
1.3 0.68
0.60 0.76
0.66 0.41
1.5 0.99
0.67 K
2
O 0.07
0.11 0.05
0.16 0.04
0.12 P
2
O
5
0.04 0.19
0.19 0.12
S 1.82
1.63 H
2
O
b
2.1 0.71
2.8 2.0
2.7 1.6
0.00 0.00
6 26
16 Rb ppm
15 11
19 6
2 Ba
115 107
247 220
147 169
31 54
Sr 208
96 217
111 215
158 49
34 14
27 12
16 30
9 Co
7 7
Ni 84
29 88
55 57
44 32
16 69
21 26
30 45
29 Cu
139 153
21 47
23 48
Zn 14
65 52
2 61
139 58
73 131
41 V
51 28
48 160
Cr 220
136 46
49 8
14 1
20 5
16 17
5 Ga
1 1
4 20
8 12
Y 6
21 11
7 19
135 38
118 146
27 Zr
32 15
4 B
1 6
2 B
1 Nb
5 315
264 408
43 210
87 S
18212 16339
a
FeO
t
, total iron as FeO.
b
L.O.I. values are assumed to represent the H
2
O content of the sample because they lack carbonate alteration and except the iron-formation are virtually devoid of sulphide minerals.
tion trends for all the FC runs are similar, no single model liquid evolution trend can reproduce
the trend of all the whole-rock compositions. The liquid evolution models that most closely approxi-
mate the composition of the most differentiated amphibole gabbro samples are produced by frac-
tionation of hydrous komatiitic liquid. The addi- tion of H
2
O suppresses plagioclase crystallisation to lower Mg contents, but does not closely repro-
duce the compositions of the noncumulate amphi- bole gabbro samples.
Titanium and Al are important elements to consider when evaluating the validity of the liquid
evolution models, because of their incompatibility in olivine. Considering this, the FC liquid evolu-
tion model that best approximates the whole-rock geochemical trend of the LKH compositions is an
anhydrous model at 2.0 kb : 6 km crustal
depth. As illustrated in Fig. 6, Ti increases and Al decreases in the residual liquid at low Mg
contents B 10 wt MgO, during plagioclase fractionation and accumulation. However, the
model trend on the Al
2
O
3
versus MgO variation diagram exhibits a sharp decrease in Al prior to
evolving to the most differentiated amphibole gabbro compositions Fig. 6. This suggests that
the model liquid is crystallising plagioclase too early or conversely that the samples crystallised
plagioclase later than predicted. A similar trend is evident on the FeOMgO variation diagram
Fig. 6. Thus, the results of the FC modelling do not provide a single liquid evolution trend
that closely represents the actual compositions of the noncumulate LKH rocks. Furthermore,
this result suggests that, either the amphibole gabbro is not petrologically related to the asso-
ciated dunite and wehrlite, or that the LKH formed by different crystallisation processes,
possibly A-FC.
5
.
5
. Assimilation-fractional crystallisation The range of A-FC trends that fractionate
olivine as magnesian, as the most magnesian olivine composition in the Shaw Dome samples,
and that can produce the most differentiated amphibole
gabbro compositions,
are shown
graphically in Fig. 6. These trends were gener- ated by assimilation of calc-alkalic basalt by an-
hydrous komatiitic liquid and FC at 2.0 kb pressure : 6 km crustal depth. No single calc-
alkalic basalt A-FC model can produce the olivine
compositions and
amphibole gabbro
compositions required. This suggests that if A- FC was responsible, variable amounts of assimi-
lation were involved in producing the LKH rocks. This is a reasonable interpretation, be-
cause differential heat loss and flux, as well as different flow velocities and therefore erosive po-
tential, are likely in a relatively large magmatic – volcanic system the size of that inferred for the
Shaw Dome. Assimilation of dacite produces similar liquid evolution trends. In detail, assimi-
lation of smaller amounts of dacite is required to produce trends with similar silica contents.
On Al
2
O
3
and TiO
2
versus MgO variation di- agrams, the composition of the calc-alkalic as-
similant cab plots close to the extrapolated position of olivine and clinopyroxene fractiona-
tion curves control lines. This closeness sug- gests
these rocks
could have
formed by
contamination of komatiites by upper crustal material.
A scenario yet to be considered is that amphi- bole gabbro is not petrologically related to
wehrlite and the associated pyroxenitic rocks. Instead, the amphibole gabbro units could rep-
resent melted and recrystallised wall rocks or they could be intrusions unrelated to the ko-
matiitic rocks. If they are recrystallised wall rocks,
the amphibole
gabbro compositions
should all plot in about the same location as the calc-alkalic basalt composition on MgO varia-
tion diagrams Fig. 6. This is not the case. The second possibility also does not appear to be
reasonable see Section 4.1.
The fractional crystallisation model that best defines the geochemical trend of the UKH
whole-rock and olivine compositions involves variable amounts of calc-alkalic basalt assimila-
tion Fig. 6. The komatiitic basalt samples ap- parently crystallised from liquids significantly
more crustally contaminated than the liquids from which the komatiite samples crystallised
cf. SiO
2
MgO variation diagram of Fig. 6. The A-FC models that involve iron-formation
produce olivine of appropriate Mg contents. However, these liquids do not reproduce the ko-
matiitic basalt compositions. Therefore, if iron- formation assimilation occurred, it may have
done so only locally and iron-formation contam- inated rock types were not sampled.
The petrologic modelling and the whole-rock geochemical
trends suggest
that the
UKH formed mainly as a result of olivine fractiona-
tion, similar to dunite and wehrlite of the LKH. The komatiitic basalt geochemical compositions
are remarkably similar to those of the LKH am- phibole gabbros.
5
.
6
. REE 6ariations Assimilation-fractional crystallisation produces
marked changes in REE profiles, particularly in LREE relative to HREE. This results from the high
LREE contents and high LREEMREE and low MREEHREE and low HREE contents of most
crustal rocks. The LREE profiles of contaminated komatiites should therefore markedly differ from
the parental melt. Consequently, REE geochemical compositions of the LKH and UKH rocks are
utilised to further constrain the possibility of relationships by A-FC processes.
Examination of the primitive mantle-normalised REE profiles Fig. 7 indicates marked differences
between the LKH and UKH rock types. All but a single LKH sample sample c 22388; Larson,
1996 are slightly enriched in LREE relative to MREE
and HREE
[LaSm
n
= 1.461 – 2.613;
LaYb
n
= 1.388 – 2.969]. In contrast, the UKH
komatiite and olivine cumulate samples are all depleted
in LREE
[LaSm
n
= 0.353 – 0.627;
LaYb
n
= 0.264 – 0.400], whereas the komatiitic
basalt samples are relatively more enriched in LREE [LaSm
n
= 0.918 – 1.083; LaYb
n
= 0.912
– 1.201] Table 4. The fact that these LREE variations are at relatively high MgO cannot be
explained by the FC models. The marked REE compositional gap between the
LKH and UKH is explained by genetic differences. However, it is not possible to conclusively
distinguish the effects of source heterogeneity, melting processes, and A-FC on the basis of
geochemical
compositions. Nevertheless,
the LREE variations are consistent with A-FC
processes. The LKH amphibole gabbros and the UKH komatiitic basalts are considered to represent
A-FC products, whereas the LKH dunite and the UKH komatiites are considered to represent
depleted, primary melts e.g. Lesher and Arndt, 1995.
Fig. 7. Primitive mantle-normalised REE diagrams. Closed symbols, aphyric and spinifex textured rocks, representing liquid and near-liquid compositions. Open symbols, cumulate rocks. Normalisation values from Sun and McDonough 1995.
5
.
7
. Accumulation, differentiation and contamination
As discussed, the evolution of the Shaw Dome komatiitic rocks can be attributed to variable
amounts of olivine accumulation followed by liq- uid differentiation and or crustal contamination
Fig. 8. The massive dunite and wehrlite sills and dykes and massive komatiitic basalt flows can be
explained by fractional olivine accumulation. In contrast, the differentiated wehrlite – pyroxenite –
gabbro and olivine cumulate to spinifex textured to aphyric komatiite units can be explained by
differentiation and or A-FC. Differentiation can be distinguished from A-FC on the basis of geo-
chemical and mineralogical parameters as inferred from different rock types. For example, A-FC
processes are represented by LREE enrichment and, perhaps, by the presence of igneous amphi-
bole in the amphibole gabbro lithology.
6. Evolution of the Shaw Dome complex