GEMOC ARC National Key Centre

S. Graham1,3, S.R. Shee2, D.D. Lambert1

1. Victorian Institute of Earth and Planetary Sciences, Monash University
2. Stockdale Prospecting Limited
3. GEMOC, Macquarie

INTRODUCTION
We report the first Re and Os isotopic data for magnetite and ilmenite derived from kimberlites and related rocks.  In this study our main objective was to assess the applicability of the Re-Os isotopic system as a geochronological tracer.  This was achieved by obtaining Re and Os concentrations from clean groundmass magnetite and macrocryst ilmenite separated from orangeites, kimberlites, melnoites and carbonatites using a low-blank Carius tube digestion method.  A secondary goal was to assess the partitioning behaviour of Re and Os in these oxide minerals in an attempt to constrain the important hosts of these minerals in these rocks.
The occurrence of kimberlites, melnoites and carbonatites within a single province allows us to provide constraints on a perhaps the most controversial topic in alkaline ultramafic research- are kimberlites and carbonatites genetically related.  On one hand the Nd, Sr and Pb isotopic systematics suggest a close relationship to rocks derived from enriched plume melts.  Thus, the lithophile isotopic systems suggest that the rocks are all derived from similar deep sources, and have been throughout time.  However, the mode of emplacement and petrographic and major element characteristics, suggest that kimberlites and carbonatites are probably not derived from similar depths nor by inference, from similar sources.
The alkaline ultramafic rocks and carbonatite from this study occur in the easternmost region of the Archaean eastern Yilgarn Craton.  The rocks occur in three provinces, from north to south which have variable emplacement ages ranging from 850 Ma, ~2050 Ma and 1300 Ma, respectively.  In the central province (Leonora) from east to west the rocks grade from melnoite to carbonatite and from melnoite to kimberlite. There have been numerous attempts to obtain emplacement ages for these rocks, although these studies have been hampered by the altered nature of the samples.  Altered phlogopites from kimberlites and melnoites yield Rb-Sr model ages of 2050 to 2150 Ma, although with impossibly low initial 87Sr/86Sr values.  Zircon U-Pb data from one melnoite yield an emplacement age of 2032 ± 66 Ma, and an Rb-Sr isochron from the carbonatite yields an age of 2021 ± 13 Ma.
The secondary aim of this study is to assess the role of magnetite and ilmenite as hosts for Re and Os in these rocks.  During mantle melting, Re is partitioned into the melt, whereas Os is retained in metal alloys, oxides (e.g., chromite) and sulphides included in other restite minerals, such as olivine.  Os therefore behaves as a refractory siderophile-chalcophile element.  However, the geochemical behaviour of Re is more difficult to determine, although the Re enrichment in magmatic sulphide deposits suggests that Re behaves as a chalcophile element.  Further evidence for the chalcophile behaviour is that Re concentrations are always higher in sulphides than in co-crystallising oxides.
Chromites from mafic-ultramafic layered intrusions have very low Re/Os ratios (<<0.005) showing that Os is compatible in this mineral (E. Curl unpublished data).  In contrast, Al-spinels from Lewisian ultramafic bodies have higher Re/Os ratios (>0.15), and magnetites from the same body yield higher Re/Os ratios although a comparable Os concentration (L. Frick unpublished data).  These data suggest that Re becomes more compatible in magnetite during fractional crystallisation within silicate magma systems.  Where ilmenite and magnetite have been reported from the same bodies the data suggest different partitioning behaviour.  At Voisey's Bay magnetites and ilmenites both yield Re/Os ratios of 20, suggesting comparable partitioning. A layered sill on the Onenga plateau have very low Re/Os ratios for magnetites (0.08-0.176) and very high Re/Os ratios (2.7-30) for ilmenites, suggesting the preferential partitioning of Re into ilmenite.

RESULTS AND INTERPRETATIONS

(A) MAJOR ELEMENT ABUNDANCES
Magnetite from the samples is present as phenocryst and groundmass phases.  In the silicate rocks groundmass magnetite occurs as discrete grains, whereas in the carbonatite, magnetite is present as discrete grains and as intergrowths with sulphide.
As a group, the magnetites are characterised by variable FeOT, low MgO and Al2O3. 76% of the spinels are characterised by < 1.0 wt % Cr2O3, and 50% have MnO contents > 0.5 wt% (MnO ranges from 0 - 5.5 wt %).  These geochemical features suggest that the majority of magnetite crystallised very late during emplacement.
All of the spinels from the province adhere well to the magnetite-ulvospinel join on the Ti-Fe2+-Fe3+ ternary. Phenocrystic spinels from a calcite-rich carbonatite are stoichiometric Fe3O4 magnetite, whereas those from a more silica-rich carbonatite are magnetite-titanomagnetite solid-solutions (Usp25Mt75).  Magnetites from the kimberlites and melnoites are constantly displaced toward more reducing conditions than calcite-carbonatite magnetite.
Spinels with Cr2O3 contents between 21 - 24 wt % (FeOT = 40 - 60 wt %) are also present within the series.  If the spinels with the highest Cr concentrations are paragenetically early crystallising phases, then these spinels may represent the bulk compositions of the different magmas.  These data also suggest that there were fO2 contrasts between the different magmas (i.e., kimberlite magnetites crystallised from a more reduced magma than the melnoite magnetites.
Magnetites from the orangeite also adhere well to the magnetite-ulvospinel join on the Ti-Fe2+-Fe3+ ternary, although they are depleted in TiO2 compared to kimberlite and melnoite magnetites.  These geochemical features are consistent with the occurrence of poikilitic tetraferriphlogopite in the orangeite groundmass, and reflect the TiO2 poor nature of the orangeite magmas during the late stages of crystallisation.
Major element data for ilmenite are only available from two locations, an extremely evolved orangeite (Bulljah) and a region that straddles the southern margin of the Craton (Norseman). At Norseman ilmenite was obtained from two different melnoite dykes, one on the craton margin, the other 100 km on craton. In general, the ilmenites are low-Cr ilmenites and those from the craton margin are enriched in Fe3+ and depleted in Mg compared to those that occur on craton.  The geochemistry of the ilmenites is interpreted as being controlled by the bulk composition of the magmas from which they crystallised.  As rutile also occurs within the macrocryst suite it allows the calculation of oxygen fugacity at ilmenite crystallisation.  These data indicate that the craton margin melnoite magma was more oxidising than that of the on-craton magma.  Thus, ilmenite geochemistry and log fO2 are correlated with the proximity to the craton margin.
Ilmenite from the Bulljah orangeite has consistently lower MgO and higher Cr2O3 concentrations than the Norseman ilmenites.  These ilmenites may have been derived from disaggregated metasomatised xenoliths, although it is more likely that these ilmenites are a very early crystallising phase from the incompatible element enriched Bulljah magma.

(B) RE AND OS ABUNDANCES
Re and Os concentrations for groundmass magnetite are summarised on Figure 1.  The carbonatite magnetites have very low Os abundances (0.035 to 0.079 ppb) for their Re abundances (0.386 to 0.839), and yield consistent Re/Os ratios of 11.  Melnoite magnetite typically yields lower Re/Os ratios (~3.0), whereas the kimberlite and orangeite magnetite yield the lowest Re/Os ratios of 0.22 to 0.34.  The variability of the ratios is a result of increasing Os rather than decreasing Re (Fig. 1).  In general, the correlation between Os and Cr is attributed to increased digestion of xenocrystic oxides, sulphides and alloys by the kimberlite prior to magnetite crystallisation.  The low Cr and high Os content of the orangeite magnetites is probably because these magnetites crystallised from a magma depleted in Cr.  Fractionation of chromite is a possibility, although this would not explain the high Os abundance as Os is also compatible in chromite.  None-the-less, magnetite in this sample is interpreted as late stage, suggesting Cr depletion in the final fluid crystallisation fluid.
Re abundances are correlated with both Zn and Mn abundances (Fig. 1).  While Zn is a relatively chalcophile element Mn is not, showing that Re probably is not behaving as a chalcophile element in these magmatic systems. Although no petrograhic evidence was observed for sulphide-magnetite co-crystallisation.  The correlations may suggest that the DRe and DZn-Mn are similar.  The offset between the two correlations on the Mn vs Re diagram is similarly difficult to interpret, although it may be related to fO2 and the carbonate content of the melts.
Re and Os concentrations for macrocryst ilmenite are also summarised on Figure 1.  All of the ilmenites yield extremely low Os concentrations (0.01 to 0.04) and variable Re concentrations (0.07 to 0.49) and high Re/Os ratios (4 to 14).  High fO2 ilmenite at the craton margin has consistently higher Re and Os concentrations than the on-craton and orangeite ilmenites. Re was not observed to vary with any major element, whereas Os is correlated with Al, Cr and Ca (Fig. 1).  As the only geochemical parameter other than bulk composition that varied between the on- and off-craton ilmenites was fO2, these data suggest that as magmas become more oxidised, more Os and Re enter the ilmenite structure.

(C) MAGNETITE AND ILMENTE AS RE AND OS HOSTS
This discussion is limited to the Norseman melnoites where Re-Os data were obtained for both ilmenite and magnetite.  Mass balance calculations based on modal proportions (750 points) of magnetite and ilmenite (~15% and ~1.5%, respectively) show that together these minerals account for >95% of the Os budget in the melnoites (~0.07 ppb).  Applying the same mass balance to Re suggests that almost 45% of the Re in the whole rocks (0.35 ppb) cannot be accounted for by magnetite and ilmenite.  Further modelling encompassing the high Re/Os ilmenite shows that the missing Re can be attributed to a sulphide with a Re/Os ratio of ~20.  Furthermore, these models are consistent with contamination by ~0.02% of sulphide similar in composition to those included in eclogitic diamonds.  The best candidate for this phase is a sulphide mineral such as pyrrhotite.
Since the Os concentration of the whole rocks can be primarily explained by two minerals, ilmenite and magnetite, a partition coefficient for Os (DOs) for each of the phases can be estimated.  It has to be assumed that the magma never formed cumulates from phases, such as sulphide or chromite, which may have been able to incorporate large amounts of Os or Re.  Furthermore, the rocks must be representative of melt compositions. Petrographic observations suggest that the Norseman melnoites crystallised from melts.  Using the equation Delement = [mineral]/[melt], ilmenite has a calculated DOs <1 (0.2 - 0.5) and magnetite has a calculated DOs >1 (~5.0).  DRe values are calculated as 0.2 - 0.5 for ilmenite and 2.5 for magnetite. The DOs for magnetites are consistent with the absence of known Os hosting minerals in the phenocryst and other groundmass phases.  The DRe for magnetite is lower than its calculated DOs value, consistent with the low Re/Os ratio of the magnetites in this igneous system.

(D) MAGNETITE AND ILMENTE AS GEOCHRONOMETERS
Magnetite samples from the Leonora Province (Fig. 2) yield extremely variable 187Re/188 Os ratios ranging from <1.3 to 70.5.  These samples are well correlated on an isochron diagram, and although the samples yield a high MSWD (perhaps due to lithospheric mantle assimilation by the kimberlite) the samples provide a precise emplacement age of 2021 ± 18 Ma and a radiogenic initial isotopic composition (gOs = +4.2; Fig. 2).  The significance of these data is three-fold.  Firstly, the isochron is indistinguishable from the SHRIMP-zircon and Rb-Sr isochron emplacement ages previously determined for rocks within the province.  Secondly this isochron represents the first isochron obtained for minerals other than phlogopite for alkaline ultramafic rocks and suggests that the Re-Os isotopic system provides a useful alternative geochronometer for altered alkaline ultramafic rocks.  Thirdly, because the kimberlites, melnoites and carbonatite lie on the same isochron it implies a genetic relationship.  This last point is incredibly significant as it provides the first hard evidence that these diverse rock types can be genetically related, by source.
A second Re-Os isochron, for the Norseman Province, is also shown on Figure 2.  On this diagram the magnetite and ilmenites are correlated and yield an imprecise age of 1000 ± 200 Ma.  While the age is greater than the accepted emplacement age (850 Ma) it is within error of this age.  This is a positive result and suggests that the analytical improvements made in the pre-concentration of Re and Os may lead to these minerals, particularly groundmass high MgO ilmenite, also providing precise emplacement isochrons for alkaline ultramafic rocks.  The radiogenic isotopic composition (gOs = +54) suggests derivation from an enriched mantle source and that the lithospheric mantle beneath the eastern Yilgarn Craton has not remained geochemically unchanged for the 1000 Ma between alkaline ultramafic emplacement.