Journal of Petrology Advance Access originally published online on November 19, 2004
Journal of Petrology 2005 46(2):291-318; doi:10.1093/petrology/egh075
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Petrology vol. 46 issue 2 © Oxford University Press 2004; all rights reserved
Assigning Dates to Thin Gneissic Veins in High-Grade Metamorphic Terranes: A Cautionary Tale from Akilia, Southwest Greenland
1 SWEDISH MUSEUM OF NATURAL HISTORY, BOX 50007, SE-104 05 STOCKHOLM, SWEDEN
2 ADVANCED CENTRE FOR QUEENSLAND UNIVERSITY ISOTOPE RESEARCH EXCELLENCE (ACQUIRE), UNIVERSITY OF QUEENSLAND, ST LUCIA, QLD 4072, AUSTRALIA
RECEIVED MAY 7, 2004; ACCEPTED SEPTEMBER 6, 2004
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONREVIEW OF AKILIA GEOLOGY...SAMPLINGANALYTICAL METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
A granodiorite from Akilia, southwest Greenland, previously suggested to date putative life-bearing rocks to
3·84 Ga, is re-investigated using whole-rock major and trace-element geochemistry, and detailed cathodoluminescence image-guided secondary ion mass spectrometer analyses of zircon U–Th–Pb and rare earth elements. Complex zircon internal structure reveals three episodes of zircon growth and/or recrystallization dated to c. 3·84 Ga, 3·62 Ga and 2·71 Ga. Rare earth element abundances imply a significant role for garnet in zircon generation at 3·62 Ga and 2·71 Ga. The 3·62 Ga event is interpreted as partial melting of a c. 3·84 Ga grey gneiss precursor at granulite facies with residual garnet. Migration of this 3·62 Ga magma (or melt–crystal mush) away from the melt source places a maximum age limit on any intrusive relationship. These early Archaean relationships have been complicated further by isotopic reworking in the 2·71 Ga event, which could have included a further episode of partial melting. This study highlights a general problem associated with dating thin gneissic veins in polyphase metamorphic terranes, where field relationships may be ambiguous and zircon inheritance can be expected. KEY WORDS: Archaean; geochronology; Greenland; secondary ion mass spectrometry; zircon
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONREVIEW OF AKILIA GEOLOGY...SAMPLINGANALYTICAL METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
A major source of controversy in the debate about the origin and evolution of life on Earth is the assignment of both relative and absolute ages to key lithologies. Indeed, the significance of actual evidence for biogenicity or for key steps of biological evolution is impossible to evaluate in cases where no consensus can be reached about relative and absolute age. Nowhere is this problem more pronounced than on the island of Akilia, southwest Greenland, where evidence for Earth's earliest life has been claimed on the basis of isotopically light carbon, occurring as graphite inclusions hosted by apatite, in a rock originally interpreted as a banded iron formation (Mojzsis et al., 1996
). Since this claim was first published, several lines of evidence have emerged that question both the biogenicity and antiquity of the Akilia example, to the extent that most specialists now prefer to focus the discussion about earliest life on the much better preserved and unambiguous volcano-sedimentary sequences in the (near-) contemporaneous Isua Greenstone Belt and the somewhat younger Warrawoona Group in the East Pilbara Craton. Despite this sobering outcome of the Akilia debate, many lessons can still be learnt, e.g. regarding establishment of relative and absolute age relationships of key lithologies.
All lithologies on Akilia have experienced two major episodes of metamorphism at 3·60–3·65 and 2·71 Ga, reaching granulite and upper amphibolite facies respectively, as well as later thermal and minor structural overprints. In cases where age relationships are determined with granitoid rocks, such high-T metamorphism and high-strain deformation conspire to inextricably link relative and absolute ages. First, quartz and feldspar, the main constituents of granitoids, deform plastically under such conditions (e.g. Pryer, 1993) and, in the presence of an aqueous fluid, have potential to melt at temperatures exceeding 670°C (e.g. Clemens & Vielzeuf, 1987). Once an anatectic melt is present, the crust is further weakened (Hollister & Crawford, 1986), which facilitates and enhances deformation. As a result, partially molten rocks can easily lose their original relative field context and assume discordant relationships that preferentially reflect the latest preserved episode of deformation rather than original relationships. Second, although the U–Pb zircon geochronometer is very resilient, Pb-loss, recrystallization or even new growth occur typically under amphibolite- and granulite-facies metamorphism (e.g. Fraser et al., 1997; Vavra et al., 1999; Hoskin & Black, 2000; Hoskin & Schaltegger, 2003; Whitehouse & Platt, 2003). Memory loss and recrystallization are certainly aided by the presence of an aqueous fluid (Villa, 1998; Tomaschek et al., 2003) but crystallization of overgrowths on pre-existing grains or even entirely new grains is more likely in a partial melt (Roberts & Finger, 1997).
The combination of disruption of original relative age relationships with variable extent of zircon (re-)crystallization under high-grade tectonometamorphism inextricably links the tasks of establishing the relative and absolute ages of highly deformed and migmatized lithologies. In the Akilia example, the validity of cross-cutting field relationships is ambiguous at best. Here, we present new high-spatial resolution (c. 10–30 µm) U–Th–Pb isotopic and rare earth element (REE) abundance data for carefully characterized growth phases in polyphase zircon crystals from a granitoid whose intrusion age has been claimed, by some, to provide a minimum age for life on Earth. The complexity of field, structural, textural, chemical and isotopic relationships of these outcrops reflects the fact that the southwestern tip of the island is essentially a highly metamorphosed shear zone. The key to finding more convincing evidence for the establishment of life is to concentrate on low-strain outcrops, where unequivocal relative field relationships can be established and where absolute dates and isotopic integrity are preserved to the extent that the null hypothesis can be applied.
|
REVIEW OF AKILIA GEOLOGY AND GEOCHRONOLOGY |
|---|
|
TOPABSTRACTINTRODUCTIONREVIEW OF AKILIA GEOLOGY...SAMPLINGANALYTICAL METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
The lithologies on southwest Akilia that are key to inferences for early terrestrial life occur as enclaves of mafic–ultramafic rocks and putative metasediments [collectively defined as the Akilia association by McGregor & Mason (1977)
] within the regional early Archean Amîtsoq gneisses. The rock types in these enclaves are not amenable to precise dating with methods available at present; hence, the debate about their absolute age has revolved around the nature of their relative age relationship to their host gneisses and, on Akilia itself, focuses on a single, small outcrop where highly strained granitoid gneisses occur in contact with the putative life-bearing enclave. Despite their obviously complex structural context, these granitoid gneisses have, none the less, repeatedly been interpreted to cross-cut and therefore provide a minimum age for the mafic–ultramafic enclave (Nutman et al., 1997, 2000, 2002; Mojzsis & Harrison, 2000, 2002). In their comprehensive treatment of field relationships on southwest Akilia, Myers & Crowley (2000) concluded that the high-strain nature of the outcrop made it impossible to determine with any degree of confidence the relative age relationships between gneissic protoliths. Whitehouse & Fedo (2003) have presented further structural data from this specific outcrop and conclude that, with regard to relative dating constraints, the critical contact is not only tectonized at the point at which discordance has previously been claimed, but also documents a truncation of structures in the gneiss sheet by the mafic–ultramafic enclave, yielding the complete opposite of previously claimed relative age relationships. Based on these studies, there is clearly little point in obtaining absolute age data from the gneisses in order to constrain the age of the mafic–ultramafic enclave. Indeed, more recent studies of the putative metasediments (Fedo & Whitehouse, 2002; Bolhar et al., 2004) that have been claimed to be dated by the supposedly discordant gneiss sheet leave, in our view at least, little doubt that the original evidence for traces of life itself must be in error. This view is further reinforced by a reported absence of apatite-hosted graphite particles in a reinvestigation (Lepland et al., 2005) of the original sample. None the less, for scientists who remain persuaded by the original geological and geochemical evidence (Mojzsis et al., 1996), it is important to test whether a coherent absolute age interpretation of the quartz diorite sheet is attainable. As a prelude to our own study, we review below all previous geochronological data obtained from this sheet.
The only data of relevance are for rocks from the contact between the Akilia association enclave and Amîtsoq gneisses (Fig. 1) on the northern side of the small peninsula that forms the southwestern tip of Akilia. This so-called ‘northern contact’ is where Nutman et al. (1997) collected their sample, G93-05, which they described as a ‘quartz-diorite sheet’. The rock apparently yielded plenty of zircon (Nutman et al., 1997, p. 2478) ‘which range from approximately 100 to 300 µm in length with most of them prismatic and the remainder being stubby-prismatic to ovoid-equant in habit’. These workers further claimed (p. 2478) that ‘structurally older inherited cores were not discerned in the photomicrographs of the analysed zircons’. As will become apparent when discussing our new data, it is important to recall Nutman et al.'s (1997) statement (p. 2478) that ‘in selecting grains of the main population for analysis, the clearest (generally lowest U) grains were predominantly selected’. The clear, low-U grains analysed by Nutman et al. (1997) define a statistically relevant population with an average age of 3865 ± 11 Ma (2
), whereas the deeper-coloured, mostly ovoid grains yielded ages between 3630 and 3530 Ma. On the basis of the their data and observations regarding zircon morphology, size and colour, Nutman et al. (1997) concluded that the igneous protolith of the quartz-diorite gneiss sheet intruded before 3850 Ma.
|
Whitehouse et al. (1999)
also published ion-probe U–Pb zircon data for a so-called quartz-diorite gneiss from the same general outcrop (their sample SM/GR/97/7). They presented cathodoluminescence (CL) images of the prismatic, elongate zircons selected for analysis, which showed oscillatory banding and a paucity of core structures. Ion-probe data for SM/GR/97/7 zircons did not directly correspond to Nutman et al.'s (1997) results, either in age distribution (younger) or U-content (higher). Whitehouse et al. (1999) concluded based on their own data and CL imaging that the bulk of the zircon consisted of igneous growth at 3650 Ma, with minor inherited cores in excess of 3·8 Ga. Not surprisingly, Nutman et al. (2001) commented on the new results and inferred that Whitehouse et al. (1999) had either mislabelled or otherwise confused a sample, as the differences between data for G93-05 and SM/GR/97/7 appeared too substantial to permit a common origin of both samples. In their response, Whitehouse et al. (2001) confirmed that their sampling site of SM/GR/97/7 was no more than 25 cm from where Nutman et al. (1997) claimed to have taken G93-05. Whitehouse et al. (2001) acknowledged the substantial difference in data for zircon extracted from samples in such close proximity but reiterated their original explanation that this could have been because of a bias in selecting grains for mounting and/or that there could be true sample heterogeneity on a dm scale—something that is indeed evident from the now numerous detailed maps and photographs of the outcrop (Myers & Crowley, 2000; Nutman et al., 2000, 2002; Whitehouse & Fedo, 2003).
Subsequently, Nutman et al. (2000) published retrospective CL images for zircons of G93-05, which confirmed the paucity of the 3·6–3·7 Ga CL-zoned thick rims that Whitehouse et al. (1999) described from their sample SM/GR/97/7. Nutman et al. (2000) also provided new U–Pb ion-probe dates for additional zircon grains from G93-05. Their new data confirmed that among the grains selected by Nutman et al. (1997), biased to large, low-U, clear, prismatic material, the pre-3·8 Ga oscillatory grains clearly dominate. The only exception is their grain 16 [shown in fig. 10 of Nutman et al. (2000)], which is a strongly zoned, but equant grain, that is clearly 3·65 Ga old.
The most recently published dataset (Mojzsis & Harrison, 2002) for zircon from another sample (GR9716) from this locality, claimed to be equivalent to G93-05, has added considerably to the complexity. Unfortunately, Mojzsis & Harrison (2002) did not provide CL images of their zircon. Instead, they obtained two datasets using two different experimental approaches. One dataset was obtained by depth profiling into the centre of a single, clear (presumably low-U) euhedral, prismatic zircon (grain GR9716_69) similar to the majority of zircons analysed by Nutman et al. (1997, 2000) from their sample G93-05. The profiling technique yielded clear evidence for this single grain of a 3828 ± 8 Ma core, mantled by a c. 3·6–3·65 Ga inner rim, which itself is mantled by a c. 2·7 Ga outer rim. This observation is entirely consistent with the general picture commonly obtained from tonalite–trondhjemite–granodiorite (TTG) suite gneisses in the area (e.g. Whitehouse et al., 1999; Nutman et al., 2002) when dates are obtained from suites of handpicked zircon grains selected for CL-image-guided U–Pb ion-probe dating. The second dataset presented by Mojzsis & Harrison (2002), however, is potentially more insightful than the depth profiling because it was obtained in situ in petrographic thin sections. Of the 71 grains analysed, only a single grain, embedded in biotite, yielded an age in excess of 3·8 Ga. This grain (g1-1) yielded a near-concordant 207Pb/206Pb age of 4079 ± 18 Ma and was interpreted by Mojzsis & Harrison (2002) as an inherited xenocryst at a relatively late stage of protolith emplacement. This is by far the oldest zircon yet reported from the North Atlantic Craton and has no correspondence to any other previously studied sample. However, the important observation from the in situ study is that almost all the grains analysed yielded ages between 3·5 and 3·7 Ga. This prompted Mojzsis & Harrison (2002) to propose that in a 
3·6 Ga migmatization event, a significant fraction ‘(50%?)’ of the protolith zircon dissolved and re-precipitated via partial melting, dissolution–reprecipitation or Ostwald ripening. However, provided that the dated zircons in this section are representative of the rock's entire population (i.e. only one grain in 71 analysed is older than 3·8 Ga), the estimate of the fraction of reworked zircons ought to be moved upward from ‘(50%?)’ to 98%. Although the in situ method has obvious advantages for preserving the petrographic enviroment of the analysed zircon, it should be realized that unlike separated zircon grain mounts, where polishing is optimized to reveal as much internal structure as possible, the zircon surface exposed in a thin section is essentially random. Without additional control of CL (or BSE) imaging, application of this technique to potentially complex zircons will inevitably yield ambiguous results owing to mixing of different growth phases. Indeed, a plot of Th/U against apparent age [fig. 7 of Mojzsis & Harrison (2002)] shows poor correspondence between in situ analyses and results of an earlier (Mojzsis & Harrison, 2000) grain mount study on the very same sample. The apparent paucity of >3·8 Ga zircon growth zones in the grains exposed in the thin section studied by Mojzsis & Harrison (2002) could thus partly stem from overlapping the primary ion beam onto ubiquitous younger mantles around old cores.
When the entire zircon evidence from this locality is considered, two obvious implications for further studies (such as ours) emerge. First, there appears to be very significant sample heterogeneity on the dm-scale as far as the distribution of zircon with different age, structure and chemistry is concerned. Only this can explain why Nutman et al. (1997) obtained a maximum age of 3865 ± 11 Ma [later revised to 3841 ± 6 Ma by Nutman et al. (2000)], and Mojzsis & Harrison (2002) determined (from a single grain) a slightly younger age of 3828 ± 8 Ma, whereas Whitehouse et al. (1999) found only a single >3·8 Ga age from the core of a c. 3·65 Ga grain. As it is physically impossible to sample exactly the same material as previous workers (unless a concerted effort is made to obtain a single large sample that is later split into aliquots and distributed to different laboratories), we have reprocessed one of our samples (SM/GR/97/8, taken immediately adjacent to SM/GR/97/7) which we now believe corresponds to the original G93-05 [see discussion also by Whitehouse & Fedo (2003)] in an effort to obtain a higher yield of zircon for a new imaging and dating session. The second implication is that irrespective of the heterogeneity in zircon distribution, it is obvious that the selection of rare, clear (low-U), prismatic grains for ion-probe analysis very strongly biases to the oldest grains. The picture gained from such grains is not at all representative of the entire zircon population. Although this bias may be logical and intentional when directed towards estimating the potential age of a pre-metamorphic protolith, it can be misleading with regard to the geological history experienced by a sample. In the case of the much-discussed Akilia locality, the question is clearly that of how old the relative field relationships are (which some interpret as igneous cross-cutting), and it is, therefore, critically important to test whether these rocks underwent significant, pervasive partial melting that would have obliterated older structural relationships pertaining to their original, magmatic emplacement. For this reason, we have tried to select the complete variety of zircon present in these rocks, and to analyse the various growth and recrystallization domains observed in the CL images. Furthermore, building on our previous work (Whitehouse & Kamber, 2002, 2003), we have obtained REE compositional data for the dated zircon spots.
|
SAMPLING |
|---|
|
TOPABSTRACTINTRODUCTIONREVIEW OF AKILIA GEOLOGY...SAMPLINGANALYTICAL METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
The sample studied here, SM/GR/97/8, was collected from the northern contact locality (Fig. 1) by S. Moorbath in the company of the late V. R. McGregor and S. J. Mojzsis. It is closely related spatially to sample SM/GR/97/7 of Whitehouse et al. (1999)
, coming from <25 cm west across regional strike of that sample. Whitehouse & Fedo (2003) noted that the latter sample preserves gneissic banding in thin section, whereas SM/GR/97/8 is essentially homogeneous (Fig. 2), with equigranular quartz and feldspar mosaic texture implying largely static, post-deformational recrystallization. This led them to suggest that SM/GR/97/8 is the most likely equivalent of the so-called ‘homogeneous tonalite’ sheet worked on by Nutman et al. (1997, 2000) and Mojzsis & Harrison (2000, 2002), with SM/GR/97/7 representing the banded gneiss immediately adjacent to the east, which, they noted, in places, appears to be cross-cut by the homogeneous sheet [see fig. 4d of Whitehouse & Fedo (2003)].
|
|
ANALYTICAL METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONREVIEW OF AKILIA GEOLOGY...SAMPLINGANALYTICAL METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
High spatial resolution U–Th–Pb and REE data were generated using a Cameca IMS1270 large-format ion microprobe at the Nordsim facility, Swedish Museum of Natural History. Detailed analytical methods have been described previously for U–Th–Pb (Whitehouse et al., 1997
, 1999) and REE (Whitehouse & Kamber, 2003; Whitehouse & Platt, 2003). In all cases, a defocused O2-primary beam was used to project the image of either a 100 µm or a 50 µm aperture onto the sample, generating elliptical, flat-bottomed craters of nominal c. 30 µm and c. 15 µm (long axis), respectively. For some analyses, the spatial resolution was further enhanced by use of the field aperture in the secondary ion optics, restricting the actual analysed region admitted to the mass spectrometer to a c. 11 µm square on the illuminated object, regardless of the size of primary probe. Complete U–Th–Pb analyses at a mass resolution (M/
M) of c. 5000 were performed using a peak switching routine, with a single ion-counting electron multiplier (EM) as the detection device. An energy window of 60 eV was used throughout, with energy adjustments made using the 90Zr216O peak. Precise mass calibration was maintained by using an automatic routine in the Cameca CIPS software to scan over large peaks and extrapolate the mass to B-field curve for peaks between these reference points (e.g. Pb-isotopes were calibrated by centring the 94Zr216O peak at nominal mass 204 and the 177Hf16O2 peak at nominal mass 209). Pb/U ratios, elemental concentrations and Th/U ratios were calibrated relative to the Geostandards zircon 91500 reference, which has an age of 1065 Ma (Wiedenbeck et al., 1995). In some cases, particularly the highest spatial resolution analyses of very small zircon domains, only Pb-isotopic ratios were measured using either a short mono-collection peak-switching routine, or a multi-collection routine using four ion-counting EMs positioned at the appropriate mass spacing and operating at a mass resolution of c. 8000 (determined by the fixed exit slit width of the multicollector system). For the multicollector measurements, the relative efficiencies of the EMs were first matched as closely as possible using pulse height analysis to set appropriate high voltage and thresholds, and then calibrated by direct measurement on the NIST SRM610 standard, assuming the composition determined by TIMS (Belshaw et al., 1994). Decay constants follow the recommendations of Steiger & Jäger (1977).
Simultaneous high mass resolution ion counting was also used for most of the analyses of Light REE (LREE) (La–Eu) coupled with low mass resolution, energy filtered mono-collection measurements of the middle to heavy REE (MREE to HREE) following the method detailed by Whitehouse & Kamber (2003) and Whitehouse (2004). For a few analyses, selected high-abundance HREE (i.e. Sm, Gd, Dy, Er, Yb) were also analysed at high mass resolution using simultaneous ion counting, with a single peak switch to measure 92Zr28Si16O2 at nominal mass 152 as a matrix reference peak. REE abundances were normalized relative to NIST SRM610 [working values of Pearce et al. (1996)] with the exception of the multi-collector HREE measurements where the Geostandards 91500 zircon was used, assuming the concentrations determined by Whitehouse & Platt [2003; these were themselves calibrated against NIST SRM610 in this earlier study, ensuring compatibility of the datasets generated by the different methods; furthermore, these concentrations are in good agreement with the ‘working values’ of Wiedenbeck et al. (2004)]. Tables 2 and 3 present the U–Th–Pb and REE analytical data, respectively, and also document the spatial resolution and method used for each analysis.
Cathodoluminescence images of zircon were obtained using a Hitachi S4300 scanning electron microscope at the Swedish Museum of Natural History.
Whole-rock major and trace-element analyses of SM/GR/97/8 were obtained from an aliquot that was milled in tungsten carbide at Oxford University. No data for W and Ta are reported because of contamination from the mill. Major element compositions were determined at the Department of Earth Sciences (UQ) by ICP–OES from a solution obtained by dissolving a fused bead. Results are reported in Table 1. Trace-element analysis was performed at the ACQUIRE laboratory (UQ) by ICP–MS from a solution obtained by digesting the sample in a steel-jacketed PTFE (Teflon), vessel. Data are reported in Table 1. Analytical details were the same as those reported by Kamber et al. (2003).
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONREVIEW OF AKILIA GEOLOGY...SAMPLINGANALYTICAL METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Whole-rock major and trace-element geochemistry
At least a part of the confusion regarding equivalence of samples from this outcrop stems from the disparate petrographic names assigned by various workers. The original sample, G93-05, was called a ‘quartz-diorite’ by Nutman et al. (1997)
, a term also adopted by Whitehouse et al. (1999) for their sample, SM/GR/97/7, but was later referred to as a homogeneous ‘tonalite’ by Nutman et al. (2000). Mojzsis & Harrison (2002) referred to their supposedly equivalent sample GR9716 also as tonalite. Major and selected trace-element (Zr, Pb, U, Th) compositions of these samples are listed in Table 1 alongside that of our sample, SM/GR/97/8.
We note the close correspondence of all analyses from these samples, in accord with our belief that they do indeed originate from the same, homogeneous part of the gneiss sheet. The normative mineralogy calculated from bulk-rock composition also indicates that all three samples are in fact granodiorites according to the widely accepted IUGS classification of plutonic rocks (Streckeisen, 1976; le Maitre, 2002), richer in quartz (c. 27%) than quartz-diorite (5–20%) and lower in plagioclase as a proportion of total feldspar (c. 85%) than tonalite (90–100%). On this basis, we use granodiorite throughout this paper. We note, however, that although quartz-diorite is definitely an inappropriate name, use of other plutonic rock-classification schemes adds ambiguity to the distinction between granodiorite and tonalite. In an ab–an–or diagram (Barker, 1979), the percentage of normative orthoclase (c. 14%) means that all three analyses plot within the tonalite field. The use of normative mineralogy in classification schemes such as QAP or ab–an–or that are intended principally for modal mineralogy can introduce additional ambiguity when dealing with hydrated rocks (SM/GR/97/8 contains abundant modal biotite) because of the tendency of norms to assign K entirely to orthoclase and to consider only orthoclase as an alkali feldspar. In the only commonly used classification scheme that utilizes elemental abundances instead of normative or modal mineralogy, the R1–R2 cation proportion scheme of de la Roche et al. (1980), these analyses fall on the boundary between the granodiorite and tonalite field.
This confusion over precise petrographic name is far from trivial when set against the backdrop of arguments about zircon saturation and the ability of melts to retain inherited zircon that has long been a feature of the Akilia geochronology debate. As noted recently by Hanchar & Watson (2003), the actual Zr-concentration and cation ratio (Na + K + 2Ca)/(AlSi) value (M) have only a small influence on zircon saturation temperature (± a few tens of °C), the most significance factor being the assumed temperature of crystallization. The higher magmatic temperature expected for a tonalite implies a greater ability to dissolve xenocrystic zircon compared with a lower-temperature granodiorite, although, as we shall demonstrate below, several additional factors need to be taken into account in discussing zircon solubility, not the least of which is the presence of water, which will significantly lower the solidus of any granitic composition.
With regard to incompatible trace elements, sample SM/GR/97/8 has general characteristics similar to those of arc (sensu lato) magmas (Fig. 3a). The key features are over-enrichment (relative to similarly compatible elements) in Cs, Pb, Be, Li and relative depletion in Nb (note that the reported value is an overestimate as a result of likely contamination from the tungsten–carbide mill). Although these features do not necessarily imply an origin of this chemistry from supra-subduction zone melting, we emphasize that enrichment in Cs, Pb, Be and Li is a very clear indication that source melting was triggered by aqueous fluids. Nutman et al. (1999) and Kamber et al. (2002) have reported similar incompatible trace-element patterns for a variety of early Archaean TTGs from southwest Greenland. The former researchers commented on the fact that trace-element systematics such as those of sample SM/GR/97/8 cannot be generated without preferential element enrichment from aqueous fluids. The latter workers provided additional evidence for the role of aqueous fluids in generating the earliest preserved continental crust. In the context of the Akilia debate, it is very significant that trace-element distribution strongly argues in favour of hydrous rather than anhydrous melting because previous consideration of zircon solubility (Nutman et al., 1997; Mojzsis & Harrison, 2002) has assumed that the igneous precursor of the studied rock crystallized from a ‘dry’ melt which, together with inappropriate petrographic identification discussed above, implies a much higher magmatic temperature and, hence, ability to dissolve inherited zircon. This point will be discussed in more detail after presentation of the zircon data.
|
A final aspect of the whole-rock geochemistry that is of relevance to the present study is the observation that the studied granodiorite does not show the strong LREE over HREE enrichment that is so characteristic of typical Archaean TTGs. In fact, the chondrite-normalized La/Yb of our sample is only 9·0, whereas the average of reported >3·8 Ga tonalite and quartz-diorite data from south of the Isua Greenstone Belt is 18—almost twice as fractionated as sample SM/GR/98/8. For this reason, this sample only just plots into the Archaean TTG field (Fig 3b) defined by Martin (1986)
. The more potassic 3·65 Ga Amîtsoq gneisses, reported by Kamber et al. (2002), show a bimodal distribution of chondrite-normalized La/Yb: one group with very high values of c. 50, the other with much lower values of 6–17. This implies that at least in terms of REE systematics, the studied sample from Akilia is rather atypical for an Archaean TTG but akin to more potassic 3·65 Ga granitoids in the area. The origin of extreme HREE depletion in the more typical TTG may be a function of fractional crystallization processes rather than direct source processes as suggested by Kleinhanns et al. (2003), possibly due in part to differing H2O-contents (Muentener et al., 2001).
Zircon populations in SM/GR/97/8
CL imaging reveals complex, polyphase internal structures in most zircon grains from sample SM/GR/97/8 (Fig. 4). On the basis of this imaging, we identify three principal types of zircon growth, commonly more than one of which is represented within a single grain. Group 1a comprises oscillatory-zoned cores occurring in euhedral to slightly rounded prismatic grains with length/width ratio of c. 3–4. The oscillatory zoned part of the grain is generally CL-bright (subgroup 1a) and its zoning is commonly sharply truncated, in some cases (e.g. grain 2–34) showing evidence for fracturing. Several cores (subgroup 1b) exhibit considerably darker CL and less distinct zoning (e.g. grains 1–4, 1–9 and 1–16). Group 2 ‘soccer ball’ zircon grains are generally equant and multifaceted, showing very low cathodoluminescence, although vaguely zoned internal structures are occasionally visible (e.g. grain 2–13, Fig. 4). Group 3 zircon grains comprise CL-dark, sometimes zoned, rims or embayments that occur on both group 1a zircon grains (in which case, they are classified as subgroup 3a) and group 1b and 2 zircon grains (classified as subgroup 3b). This type of zircon varies from substantial tip overgrowths of prismatic grains (e.g. grain 2–4, Fig. 4) to thin rims developed on typical group 2 grains (e.g. grain 2–14, Fig. 4). A fourth group of zircon is defined as outer rims that are found on grains with two earlier phases of growth.
|
Zircon U–Th–Pb geochronology
SIMS U–Th–Pb analytical data for zircons from SM/GR/97/8 are presented in Table 2 and in a plot of 207Pb/206Pb dates in increasing order (Fig. 5). These are discussed below in the context of the grouping of zircon growth phases based on CL characteristics.
|
|
Group 1 zircon cores exhibit a wide but bimodal range of 207Pb/206Pb dates from c. 2·7 to >3·8 Ga. The CL bright, oscillatory-zoned majority of this group (subgroup 1a) show a more restricted range of 207Pb/206Pb dates from c. 3·57 to 3·88 Ga. These analyses are concordant to slightly discordant with low-U concentrations (c. 70–220 ppm) and moderate Th/U ratios (0·14–0·70). The similar CL characteristics and Th/U systematics suggest that these grains could belong to a single igneous protolith with an age in excess of 3·8 Ga which has experienced variable amounts of ancient Pb-loss, probably during the high-grade events recorded in the Godthåbsfjord at c. 3·65 and c. 2·7 Ga, resulting in a spread of ages along concordia. In this context, the oldest coherent group of 207Pb/206Pb dates that define a prominent plateau on the cumulative age diagram (Fig. 5) yields a weighted average of 3842 ± 5 Ma (n = 7, MSWD = 0·62, 95% confidence). This interpretation is in accord with previous interpretations of similarly spread dates from other gneisses in this region (e.g. Nutman et al., 1996
, 2000; Whitehouse et al., 1999), but the possibility that younger magmatic zircon is represented in this population cannot be entirely ruled out. A single analysis yielding a 207Pb/206Pb age of 3876 ± 26 Ma (2) is slightly older than the main c. 3·84 Ga group and might represent a genuinely older grain. Four analyses of CL-dark cores of subgroup 1b have high-U concentrations (1800–8300 ppm), and low Th/U ratios (<0·04). Only one concordant date has been obtained at c. 2·72 Ga, with the other analyses plotting between 27 and 47% discordant. One of the discordant analyses has a 207Pb/207Pb age of c. 2·78 Ga, suggesting that it cannot simply represent an equivalent of the concordant 2·72 Ga analysis that has undergone later Pb-loss but probably contains an inherited component or is a much older grain (i.e. early-Archaean) that has experienced extreme, possibly multiple, Pb-loss.
Group 2 ‘soccer ball’ zircon grains mostly yield a small range of 207Pb/206Pb dates from c. 3·57 to c. 3·62 Ga. Th/U ratios are very low (<0·03) and U-concentrations fall in the range c. 500–800 ppm. The spread to younger dates is attributed to minor Pb-loss from originally 3·62 Ga zircons, probably during the c. 2·71 Ga regional high-grade event (although still concordant at the 2 error level, the two analyses with youngest 207Pb/206Pb dates are displaced to the right of the inverse concordia). One zircon core (1–15a) that is provisionally classified into group 2 on the basis of its CL-dark, rounded appearance yields a concordant age of c. 2·71 Ga, with low Th/U (0·025) and high U-concentration (1100 ppm).
Group 3a zircon phases comprise distinct, rather thin rim, overgrowths or embayments developed on oscillatory zoned CL-bright zircon cores of group 1a. Pb-isotope ratios only were obtained from this group using a small primary beam spot (c. 15 µm), together with further enhancement of spatial resolution of c. 11 µm using a field aperture in the secondary ion beam (see Fig. 4). Despite the overall similarity of appearance in CL as well as consistently low Th/U ratios (<0·06), this group yields a distinctly bimodal set of 207Pb/206Pb dates in the range c. 3·70–3·50 and 2·72–2·58 Ga. In line with our earlier interpretations, we attribute this limited spread in zircon dates to Pb-loss affecting zircon grown originally at c. >3·60 and c. 2·71 Ga (Fig. 5). Group 3b zircon (rims on group 1b and group 2 zircon), together with group 4 zircon (rims on polyphase grains) show 207Pb/206Pb dates around 2·7 Ga or lower and consistently low Th/U ratios (<0·1). Six of the seven analyses in group 3b for which full U–Pb data are available lie on a regression with an upper intercept date of 2709 ± 13 Ma (MSWD = 2·2, 95% confidence); a subset of four of these analyses define a concordia age of 2709 ± 7 Ma (MSWD = 0·73, 2; decay constant errors included). A somewhat older, discordant age of c. 2·95 Ga from analysis 1–12a probably represents a mixed analysis between a c. 2·7 Ga rim and older core, as a 207Pb/206Pb age of 2741 ± 26 Ma (2) was obtained from a subsequent higher spatial resolution analysis.
The U–Th–Pb data that we obtain from sample SM/GR/98/8 are in good agreement, both in maximum age and age range, with data reported by Nutman et al. (1997) and Mojzsis & Harrison (2002) from their respective samples of the homogeneous tonalite from this locality, again suggesting that SM/GR/97/8 is the equivalent of these previously investigated samples.
Zircon REE geochemistry
REE concentration data are presented in Table 3, in a series of normalized [to C1 chondrite values of Anders & Grevasse (1989)] abundance diagrams (Figs 6–9), and in plots showing the degree of LREE and HREE fractionation against age (Fig. 10a and b). Zircon samples from group 1a, i.e. oscillatory zoned c. 3·6 to >3·8 Ga cores of grains, show a limited range of compositions in Fig. 6. Most analyses show typical ‘igneous’ zircon REE profiles with closely parallel steep slopes from La to Lu, marked positive-Ce and small negative-Eu anomalies that are very similar to the profiles reported from other early-Archaean zircon grains in the Godthåbsfjord (e.g. Whitehouse & Kamber, 2002, 2003). Two of these five profiles (2a–x and 10a–x) represent analyses of the same growth phase (based on CL imaging) that was analysed for U–Th–Pb but in a slightly different location. Analyses from the original dated location show markedly increased LREE concentrations (omitted from Fig. 6 but data are retained in Table 3) that might be related to later LREE mineralization along cracks, perturbing the original profile. The exact nature of this mineralization is unclear but, given the clearly LREE-depleted nature of the original zircon, only a very small amount of a strongly LREE-enriched phase (e.g. monazite or allanite) would be required to influence the REE profiles, whereas other characteristic enrichments (e.g. higher levels of Th in monazite) remain undetected. Analysis 2–5a exhibits a highly unusual REE profile that is enriched in both LREE and MREE. Because it is unlikely that zircon can incorporate REE in this pattern [by see discussion in Whitehouse & Kamber (2002)], we conclude that the REE profile of this particular grain must be controlled by inclusions (and/or microcrack fillings) of both LREE- and MREE-enriched minerals, such as monazite, allanite or apatite. This analysis is not considered further for the purposes of protolith characterization. In general, the REE profiles from group 1a, together with Th/U ratios, are typical of zircon crystallized in equilibrium with a primary melt.
|
|
|
|
|
|
Group 1b zircon samples also represent cores, but these show markedly different characteristics from those of group 1a, particularly with regard to the LREE, which are consistently flat at chondrite-normalized values between c. 40 and 1000, with absent to small positive Ce-anomalies (Fig. 6). The HREE profiles of this group are either highly fractionated, similar to those of group 1a, or in one case (analysis 1–4b), flat at c. 100x chondritic.
Group 2 zircon grains exhibit an internally coherent set of REE profiles (Fig. 7) that are notably different from those of either group 1a or group 1b. The most prominent difference is the marked depletion of the HREE with flat or even concave-up profiles from Dy to Lu, irrespective of whether the analysed grain is early or late Archaean. These zircon grains show more variable fractionation of LREE relative to group 1a, with the early Archaean grains showing the steepest profiles. In general, these analyses show smaller-magnitude Ce-anomalies and larger negative Eu-anomalies. Interestingly, the least suppressed positive Ce-anomaly (analysis 14a2) is accompanied by the least enhanced negative Eu-anomaly—a feature that would be consistent with variable redox control of the behaviour of these two elements during the event generating the c. 3·6 Ga zircon. The characteristic HREE depletion of this group (and possibly in analysis 1–4b of group 1b) is the most important observation, as it suggests crystallization in an environment depleted in HREE, probably because of the role of garnet.
Group 3a zircon grains show a wide range of REE profiles (Fig. 8) that probably reflect different origins within this group. The similarities of individual analyses to those of either group 1a and 1b or group 2 are apparent in comparison of profiles in chondrite-normalized abundance diagrams and are particularly well defined in plots of (Pr/Sm)N and (Dy/Yb)N against age. The characteristic HREE depletion of group 2 zircon is evident in some of both the early and late Archaean analyses of this group, suggesting crystallization from a garnet-influenced medium, although fractionated HREE profiles are also seen in both age groups. LREE profiles show a range of fractionation in the early Archaean analyses from steep profiles similar to those of group 1a through to somewhat flatter (but still fractionated) profiles. The late Archaean analyses show only flatter LREE profiles.
Analyses of zircon in group 3b, all of which are late Archaean in age, exhibit a wide range of fractionation in both HREE and LREE (Fig. 9a). In two cases, the corresponding core (group 1b) and rim (group 3b) analyses may be compared. For zircon grain 1–4, the core (1–4b) and rim (1–4a) REE systematics are very similar, whereas for grain 1–16, the rim (1–16a) shows a similar degree of fractionation of HREE at lower concentrations relative to the core (1–16b), but a strongly reverse fractionated LREE profile.
Group 4 zircon analyses both show fractionated HREE, in one case (2–36c) at a low overall concentration level, whereas the LREE of analysis 2–23c show a flat profile (Fig. 9b). It is interesting to note that the HREE of analysis 2–23c contrasts with the flat profile seen in the early Archaean group 3a core overgrowth in this same zircon grain (analysis 2–23bx).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONREVIEW OF AKILIA GEOLOGY...SAMPLINGANALYTICAL METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Equivalence of studied samples
The first point to note about our new zircon geochronological data from SM/GR/97/8 is that they exhibit the same range of ages and, within specific age groups, Th/U ratios as data previously presented (Nutman et al., 1997
; Mojzsis & Harrison, 2002) from a homogeneous granitoid sheet within the much debated c. 0·7 m wide composite gneiss layer on Akilia. Taken together with our whole-rock major- and trace-element data, which also show strong similarity to these earlier studies, we are, therefore, convinced that this sample is an appropriate representative of the homogeneous granodiorite sheet from which claims for the >3·83 Ga age for the mafic–ultramafic enclave on southwest Akilia have been derived. A previous claim (Whitehouse et al., 1999) that another sample from the composite gneiss sheet (SM/GR/97/7) was the equivalent, despite its younger c. 3·65 Ga zircons, was based on an incomplete understanding of the complexity of the gneiss sheet (Nutman et al., 1997), that was subsequently revealed by Myers & Crowley (2000), Nutman et al. (2002) and Whitehouse & Fedo (2003); this claim is now considered invalid, although data from this sample remain valid.
In this paper, we also reiterate the detailed field observations of Whitehouse & Fedo (2003), which indicate (1) that the homogeneous granodiorite sheet cannot be demonstrated to cross-cut the mafic–ultramafic enclave and hence cannot provide an age constraint, and (2) that the homogeneous granodiorite itself appears to cross-cut a banded gneiss (SM/GR/97/7) that contains typically igneous oscillatory-zoned zircons with an age of c. 3·65 Ga (Whitehouse et al., 1999). Thus, the possibility needs to be considered that the preserved discordance could indicate intrusive remobilization of the granodiorite 
3·65 Ga. Against this background, >3·65 Ga oscillatory zoned zircons in the granodiorite could be remnants from a c. 3·84 Ga protolith that were not digested during the c. 3·62 and 2·71 Ga remelting and remobilization events.
Although it is beyond any doubt that 3·75–3·84 Ga granitoids existed in the Godthåbsfjord region, the critical issue on Akilia Island is not to provide evidence for relict original >3·8 Ga features (such as zircon cores), but to establish which aspects of the gneiss as it presents itself today can be attributed to known stages of the regional geology. The key question in this regard is to what extent the gneiss was remelted and deformed during the 3·6 and 2·71 Ga events. Only then can relative age relationships seen in the outcrop be dated. In the following sections, we discuss our new data together with previously published observations in the context of three possible emplacement scenarios.
Emplacement models consistent with zircon data and field observations
Model 1: 3·60–3·65 Ga emplacement of the granodiorite with inheritance of older zircon
As previously discussed for other TTG gneisses from the Godthåbsfjord, the distribution of zircon ages in SM/GR/97/8 and equivalent samples is formally consistent with a magmatic age of 3·60–3·65 Ga. This age is within error of that of the immediately adjacent sample SM/GR/97/7, which appears structurally older (Myers & Crowley, 2000; Whitehouse & Fedo, 2003). There is thus no conflict between relative and absolute age dates in this interpretation, in which >3·65 Ga zircon is viewed as inherited. The possibility that melts of tonalitic–granodioritic composition can entrain older zircon has long been a vexed issue in discussion of the protolith ages of rocks from the Gothåbsfjord region. Several workers, applying the zircon solubility arguments of Watson & Harrison (1983), considered that the assumed high temperature and low-Zr concentration of most TTG melts a priori preclude survival of older zircon (Nutman et al., 1997; Mojzsis & Harrison, 2002). We note that these studies universally ignore the factors considered by Watson (1996) to favour the survivability of inherited zircon, namely (1) presence of large zircon in the source rock or occlusion of zircon grains within major silicates; (2) a brief or cool magmatic event; and (3) failure of the melt to coalesce into a large body. In our view, two additional important issues that seriously question the validity of rejecting the hypothesis of inherited zircon concern the assumed solidus temperature for the studied rocks and the estimated fraction of zircon which actually survived as xenocrystic grains.
With regard to the homogeneous granodiorite sheet on Akilia, no evidence has ever been presented that it represents, as claimed (e.g. Mojzsis & Harrison, 2000), a high-temperature (c. 900°C) dry melt. Rather, the new trace-element data presented here for SM/GR/97/8, in accordance with published data for similar rocks (Nutman et al., 1999; Kamber et al., 2002), strongly argue against a dry (and hence hot) solidus. Experiments of wet melting of tonalite compositions have shown that very significant melt fractions (30–70%) are still present at temperatures below 750°C (e.g. Peters & Kamber, 1994; Sawyer, 1994). As pointed out by Kamber & Moorbath (1998), a large proportion of typical Amîtsoq gneisses will retain a ‘critical melt fraction’, sufficient still to be mobile, at c. 775°C and at this temperature, they would be Zr-oversaturated and hence capable of preserving inherited zircon. Under these conditions, the criteria noted by Watson (1996) of cool magmatic events and failure of melts to coalesce into large bodies would be met. Because temperature is the single most important parameter determining zircon saturation (Hanchar & Watson, 2003), advocates of a c. 900°C solidus for the studied rock have to produce conclusive supporting evidence, before discarding the possibility of inheritance.
Inheritance of xenocrystic zircon in exactly these rocks has been demonstrated with the >4·0 Ga grain analysed in situ by Mojzsis & Harrison (2002). These workers suggested, however, that survival of a significant proportion of inherited zircon was unlikely (assuming, of course, that their solidus temperature estimate of 900°C is correct). The question, then, is what should be regarded as a significant proportion of inherited grains? As discussed earlier, for understandable reasons, zircon separation, grain selection and ion-probe dating all bias towards large grains of high-quality (low-U) zircon. In our study, we have strived to date a representative selection of grains, although even with these efforts, the majority of zircon grains (many of which are too small for routine handling) have remained untouched. Although it is impossible to estimate reliably the true proportion of >3·8 Ga zircon in this rock, it is sufficient to note from the CL images (Fig. 4) that group 1a zircon grains are not necessarily a predominant component of the total zircon fraction.
We thus maintain that if the homogeneous granitoid sheet was emplaced as a crystal-rich magma, instead of as a completely liquid melt, the observation by Mojzsis & Harrison (2002) that the oldest zircon grains are found in the middle of feldspars and must therefore represent the protolith age is irrelevant, because they would merely represent the age of the source rock from which the homogeneous granodiorite sheet was ultimately derived and have no bearing on the emplacement age of the granodiorite sheet itself, which has been the singular important goal of these studies.
Although a magmatic age of 3·6–3·65 Ga appears plausible from the data discussed thus far, our new REE data for group 2 zircon grains (which best represent growth of zircon in this time period) suggest that this event more probably represented lower-crustal remelting of an older protolith (in the presence of garnet or another HREE-retaining phase). We investigate this in the following two models.
Model 2: remelting of 3·84 Ga gneiss at 3·62 Ga
Structural and geochronological constraints discussed by Whitehouse & Fedo (2003) preclude a >3·65 Ga magmatic age for the homogeneous granitoid sheet. It is therefore likely that the actual magmatic age is represented by one of the two remaining age groups identified: either c. 3·62 or c. 2·7 Ga. In this section, we investigate the constraints imposed by our REE data on these ages.
The distinctive HREE-depleted profiles seen in the group 2 zircon grains, as well as c. 3·62 Ga overgrowths on older cores (group 3a), are typical of zircon that has crystallized and/or recrystallized in equilbrium with a strongly HREE-enriched phase of which garnet is most likely (e.g. Rubatto, 2002; Whitehouse & Platt, 2003). As garnet is not present in this granodiorite, we propose that it was a residual phase during melting of an older protolith. Melting of a garnet-bearing mafic source, although capable of producing a quartzofeldspathic melt, is considered unlikely because a much steeper whole-rock REE profile similar to typical TTGs (Fig. 3a) would be expected (e.g. Martin, 1986). Furthermore, a non zircon-bearing mafic precursor would not yield older zircon, as seen in the cores of many of the SM/GR/97/8 grains. Instead, we propose melting of a tonalite or quartz–diorite protolith during granulite-facies metamorphism at c. 3·62 Ga. Several experimental studies have documented garnet as a reaction product formed by dehydration melting of biotite-bearing quartzofeldspathic rocks (e.g. Rutter & Wyllie, 1988; Skjerlie & Johnston, 1993, 1996; Patiño-Douce & Beard, 1996; Singh & Johannes, 1996a, 1996b). The study of Singh & Johannes (1996a, 1996b) specifically investigated melting of a tonalite and showed that garnet appeared in the melt residues at pressures above 10 kbar and temperatures exceeding c. 750°C, whereas the melts themselves were of granodioritic composition.
An intriguing aspect of the REE composition of 3·62 Ga zircon in this sample is that it appears to record different growth conditions, with the group 2 and some of the group 3a analyses recording the garnet influence on HREE described above,