Skip Navigation

Journal of Petrology 2007 48(8):1605-1639; doi:10.1093/petrology/egm032
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (1)
Right arrowRequest Permissions
Google Scholar
Right arrow Articles by Zeh, A.
Right arrow Articles by Barton, J. M.
Right arrow Search for Related Content
GeoRef
Right arrow GeoRef Citation
Social Bookmarking
 Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

© The Author 2007. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org

Archaean to Proterozoic Crustal Evolution in the Central Zone of the Limpopo Belt (South Africa–Botswana): Constraints from Combined U–Pb and Lu–Hf Isotope Analyses of Zircon

Armin Zeh1,*, Axel Gerdes2, Reiner Klemd1 and Jackson M. Barton, Jr3

1Mineralogisches Institut Der Universität Würzburg, Am Hubland, D-97074 Würzburg, Germany
2Institut Für Geowissenschaften, Senckenberganlage 28, D-60054 Frankfurt Am Main, Germany
3Department of Geology, University of Fort Hare, Private Bag X1314, Alice, 5700, South Africa

RECEIVED JANUARY 9, 2007; ACCEPTED MAY 24, 2007


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL SETTING, CURRENT...
 ANALYTICAL TECHNIQUES
 RESULTS AND INTERPRETATION OF...
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
A combined set of U–Pb and Lu–Hf in situ laser ablation ICP-(MC)-MS zircon analyses were obtained from orthogneisses and granitoids in the Central Zone of the Limpopo Belt, which comprises the Beit Bridge and Mahalapye complexes. The results indicate that by combining the two isotope systems primary magmatic zircon domains can be distinguished from those formed during later metamorphic events, even if the distinct zircon domains underwent multiple Pb loss and the texture–age relationships, as obtained by cathodoluminescence images and U–Pb analyses, are ambiguous. Furthermore, the applied technique allows distinction of zircon grains formed in juvenile magmas from those generated by melting of older continental crust or affected by substantial crustal contamination. The combined U–Pb and Lu–Hf data reveal that the Sand River gneiss suite of the Beit Bridge Complex was emplaced at 3283 ± 8 Ma and formed from melting of an older Archaean crust, which was derived from a depleted mantle source at around 3·65 Ga. The hafnium model age (TDMHf) is significantly older than those obtained from zircons from numerous Neoarchaean granitoids of the Beit Bridge Complex, comprising the Singelele gneiss (2647 ± 12 Ma), the Bulai granite (2612 ± 7 Ma), the Regina gneiss (2649 ± 9 Ma) and two samples of the Zanzibar gneiss (2613 ± 6 Ma). These granitoids show initial {varepsilon}Hf(t) values between + 0·5 and –7·1, which correspond to initial TDMHf between 3·46 and 3·01 Ga. These variable TDMHfinitial and {varepsilon}Hf(t)initial values are interpreted to be the result of different mixtures of reworked 3·65 Ga Palaeoarchaean crust with juvenile magmas extracted from the depleted mantle during the Neoarchaean at ~2·65 Ga. This conclusion is supported by results obtained from the Mahalapye Complex, which was affected by migmatization and granite intrusions during the Palaeoproterozoic at 2·02–2·06 Ga. The Mokgware granite (2019 ± 9 Ma) contains zircon xenocrysts with Pb–Pb ages of 2·52–2·65 Ga and 2·93 Ga and hafnium model ages of 3·0–3·4 Ga, indicating that this granite is derived from remelting of Archaean crust. In contrast, uniform TDMHfinitial ages of 2·61–2·67 Ga obtained from a diorite gneiss (2061 ± 6 Ma) of the Mahalapye Complex indicate that its protolith may have been formed from remelting of a Neoarchaean juvenile crust. Variable {varepsilon}Hf(t)initial values from –3·7 to +6·3 of zircon cores (2711 ± 11 Ma) in an adjacent leucosome also support a model of mixing of juvenile mantle derived matter with older crust in the Neoarchaean.

KEY WORDS: Archaean; Palaeoproterozoic; Limpopo Belt; zircon, U–Pb dating; Lu–Hf isotopes; LA-ICP-MS


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL SETTING, CURRENT...
 ANALYTICAL TECHNIQUES
 RESULTS AND INTERPRETATION OF...
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Isotope methods applied to Archaean magmatic rocks provide important information about the formation and evolution of the continental crust during the early Earth's history. Such methods can provide answers to some of the most controversial questions; for instance, about the timing of Archaean to Proterozoic crust-forming events and the recycling of this crust during subsequent orogenic processes (e.g. Armstrong, 1968Go, 1991Go; O’Nions et al., 1979Go; O’Nions & Hamilton, 1981Go; Patchett et al., 1981Go; Allègre, 1987Go; Albarède & Brouxel, 1987Go; Allègre & Lewin, 1989Go; Stevenson & Patchett, 1990Go; Amelin et al., 1999Go; Rino et al., 2004Go; Condie et al., 2005Go; Harrison et al., 2005Go; Iizuka et al., 2005Go). In detail, such investigations provide valuable information on the timing of magma generation and its source, such as a juvenile depleted mantle, a reworked older crust or a combination of both (e.g. Vervoort & Blichert-Toft, 1999Go; Davis et al., 2005Go; Gerdes & Zeh, 2006Go; Hawkesworth & Kemp, 2006Go).

For such studies, the mineral zircon has been shown to be an ideal candidate, for the following reasons. (1) Zircon crystallization ages can be obtained very precisely using the U–Pb isotope system even for very old rocks (e.g. Krogh, 1973Go; Compston et al., 1984Go; Gerdes & Zeh, 2006Go; Iizuka et al., 2006Go). (2) Zircon can incorporate high concentrations of Hf (>1 wt %) but only minor Lu (~1 ppm) in its lattice and, thus, the initial Hf isotope compositions of the magma at the time of zircon crystallization can be obtained with high precision, in particular by means of inductively coupled plasma mass spectrometry (ICP-MS) techniques (e.g. Patchett et al., 1981Go; Vervoort & Blichert-Toft, 1999Go; Griffin et al., 2004Go; Gerdes & Zeh, 2006Go). (3) In situ analytical methods such as ion probes (e.g. sensitive high-resolution ion microprobe; SHRIMP) and laser ablation (LA)-ICP-MS allow the analyses of different parts of individual zircon grains. Consequently, it is possible to recognize inherited zircon domains as well as younger overgrowth and/or alteration zones (e.g. Gerdes & Zeh, 2006Go, and therein). Furthermore, the combined consideration of the U–Pb and Lu–Hf system for zircon grains in orthogneisses may help to constrain whether distinct zircon grains or domains were affected by single or multiple Pb loss, and/or if distinct zircon zones were formed at different times (e.g. Amelin et al., 2000Go, and therein).

In this study we investigate zircon from orthogneisses and granitoids in the Central Zone of the Limpopo Belt, which represents a mobile belt squeezed between the Zimbabwe Craton to the north and the Kaapvaal Craton to the south (Fig. 1). As shown by previous studies, the (meta)magmatic rocks of the Limpopo Belt were formed during at least three distinct magmatic periods, during the Palaeoarchaean at 3·25 Ga, the Neoarchaean at 2·5–2·7 Ga and the Palaeoproterozoic at c. 2·0 Ga (e.g. Barton et al., 1994Go; Barton & Sergeev, 1997Go; Jaeckel et al., 1997Go; Kröner et al., 1998Go, 1999Go; Chavagnac et al., 2001Go). In addition, a few detrital zircon grains and xenocrysts point to magmatic events as old as 3·8 Ga (Armstrong et al., 1988Go; Kröner et al., 1998Go). The available geochronological data show that the Limpopo Central Zone represents a unique orogenic belt where Archaean to Proterozoic magmatism, which took place over a period of more than 1300 Myr, can be studied in a very restricted area. Furthermore, the position of the Limpopo Belt between the Kaapvaal and Zimbabwe Cratons makes it important for obtaining information about the crust–mantle processes that took place during the amalgamation of these two cratons.


Figure 1
View larger version (86K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 1. Geological map of the Limpopo Belt with sample locations. 1, Mahalapye granite; 2, Mokgware granite; 3, Bulai granite; 4, Matok granite; 5, Razi granite.

 
In contrast to the abundant geochronological data mentioned above, little is known about the source(s) of the magmatic rocks that were emplaced during the three distinct events. A few Sm–Nd isotope whole-rock analyses indicate that the orthogneisses with ages of 3·25–3·3 Ga and 2·6 Ga were derived from a source with an average crustal residence age (TDMNd) between 2·8 and 3·4 Ga (Harris et al., 1987Go; Barton, 1996Go; Kröner et al., 1998Go; 1999Go), whereas some 2·0 Ga migmatitic–magmatic rocks yielded younger crustal residence ages between 2·4 and 2·9 Ga (Chavagnac et al., 2001Go). To clarify the crustal evolution of the Limpopo Belt's Central Zone in more detail, we present new in situ U–Pb LA-ICP-MS age data from magmatic zircons and combine this information with multicollector (MC-)ICP-MS Lu–Hf isotope analyses from the same zircon domains.


    GEOLOGICAL SETTING, CURRENT INTERPRETATIONS AND SAMPLES
 TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL SETTING, CURRENT...
 ANALYTICAL TECHNIQUES
 RESULTS AND INTERPRETATION OF...
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
The Limpopo Belt is a high-grade metamorphic province bounded by the Zimbabwe and Kaapvaal Cratons, comprising Archaean to Palaeoproterozoic rocks (Fig. 1) (e.g. Roering et al., 1992Go; Berger et al., 1995Go; Kamber et al., 1995aGo, 1995bGo; Rollinson & Blenkinsop, 1995Go; Berger & Rollinson, 1997Go; Holzer et al., 1998Go, 1999Go). Lithologically and structurally, it can be subdivided into three distinct domains, which are separated by major shear zones (McCourt & Vearncombe, 1987Go, 1992Go; Holzer et al., 1999Go; Schaller et al., 1999Go). The Central Zone is structurally bounded by the >10 km wide Palala–Zoetfontain and Triangle shear zones against the Southern and Northern Marginal zones, respectively. These steeply dipping shear zones were formed and/or activated under high-grade metamorphic conditions at about 2·0 Ga (Holzer et al., 1998Go, 1999Go; Chavagnac et al., 2001Go; Kreissig et al., 2001Go). The Southern Marginal Zone consists of high-grade, intensely deformed and metamorphosed, tonalitic, trondjemitic and granodioritic rocks (TTGs) with minor greenstones, interpreted as reworked counterparts of the Kaapvaal Craton (e.g. Coward et al., 1976Go; van Reenen et al., 1987Go, 1992Go), whereas the Northern Marginal Zone is dominated by enderbite and charnoenderbitic orthogneisses with some greenstones and granitoid rocks (e.g. Mason, 1973Go; Berger et al., 1995Go; Kamber & Biino, 1995Go; Berger & Rollinson, 1997Go; Jelsma & Dirks, 2002Go). The rocks of both marginal zones were thrust onto their adjacent cratons along >10 km wide, steeply dipping shear zones, the Hout River and Umlali shear zones, respectively (Fig. 1). According to Kreissig et al. (2001Go), magmatic intrusions and metamorphism of the Southern Marginal Zone occurred between 2·69 and 2·62 Ga, coeval with or slightly younger than the magmatic–metamorphic evolution in the Northern Marginal zone between 2·72 and 2·58–2·62 Ga (Berger et al., 1995Go; Kamber & Biino, 1995Go; Mkweli et al., 1995Go).

What was traditionally called the Central Zone proper was subdivided by Aldiss (1991Go) into three complexes, which were interpreted by Barton et al. (2006Go) as terranes: (1) the Mahalapye Complex, which is dominated by c. 2·02–2·0 Ga granites with minor high-grade sedimentary rocks (Hisada & Miyano, 1996Go; McCourt & Armstrong, 1998Go; Holzer et al., 1999Go; Chavagnac et al., 2001Go); (2) the Phikwe Complex, which is dominated by Archaean, hornblende-bearing, tonalitic and trondjemitic gneisses and igneous rocks (Brandl, 1992Go); (3) the Beit Bridge Complex. The Beit Bridge Complex hosts the c. 3·2–3·3 Ga Sand River TTG suite and Messina layered intrusion (Barton, 1983Go; Barton & Sergeev, 1997Go; Kröner et al., 1999Go) and numerous Neoarchaean granitic to granodioritic orthogneisses with ages between 2·73 and 2·60 Ga, such as the Alldays, Singelele, Bulai, Zanzibar and Zoetfontain gneisses (e.g. Jaeckel et al., 1997Go; Kröner et al., 1999Go). Based on detrital zircon ages and lithological–structural differences, the supracrustal paragneisses of the Beit Bridge Complex can be divided into three unconformity-bounded successions, with ages >3·1 Ga, 3·1–2·6 Ga and < 2·6 Ga (e.g. Erikson et al., 1988Go; Brandl, 1992Go; Kröner et al., 1998Go, 1999Go; Barton et al., 2003Go; Buick et al., 2003Go). Quartzites and metapelites from within the oldest unconformity-bounded succession contain zircons with U–Pb ages of 3·2–3·8 Ga, suggesting that a yet unrecognized, very old protolith must exist within the Beit Bridge Complex (Armstrong et al., 1988Go; Barton & Sergeev, 1997Go; Kröner et al., 1998Go).

Further subdivision of the Limpopo Belt comes from Pb isotopic data, indicating that the Zimbabwe Craton, Northern Marginal Zone, and Beit Bridge and Phikwe terranes were all derived from a source with a long-lived, high µ value (>11–12), whereas the rocks in the Southern Marginal Zone and the Kaapvaal Craton were derived from a source with lower µ values close to Bulk Earth (Barton et al., 1983Go; Taylor et al., 1991Go; Barton, 1996Go; Berger & Rollinson, 1997Go; Kreissig et al., 2000Go). Hence, the rocks of the Zimbabwe Craton, Northern Marginal Zone, and Beit Bridge and Phikwe terranes may stem from the same mantle source, but they cannot be genetically related in a straightforward way to those of the Kaapvaal Craton and Southern Marginal Zone (Barton et al., 2006Go). Based on geochemical and Sm–Nd–Pb isotope data and petrogenetic modelling, Berger & Rollinson (1997Go) suggested that the 2·6–2·7 Ga enderbitic to charnoenderbitic rocks of the Northern Marginal Zone represent a mixture of a pre-existing Archaean TTG–crust, which was partially remelted and mixed with a more juvenile magma derived from the depleted mantle at 2·6–2·7 Ga.

In the present study zircon grains from nine granitoid samples were investigated, six from the Beit Bridge Complex (Sand River gneiss, Singelele gneiss, Bulai granite, Regina gneiss, grey and leucocratic Zanzibar gneiss) and three from the Mahalapye Complex (Mogkware granite, a garnet–biotite gneiss and a leucosome from the Lose quarry). The sample localities and co-ordinates are shown in Fig. 1 and Table 1, respectively.


View this table:
[in this window]
[in a new window]

 
Table 1: Co-ordinates of sample localities

 

    ANALYTICAL TECHNIQUES
 TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL SETTING, CURRENT...
 ANALYTICAL TECHNIQUES
 RESULTS AND INTERPRETATION OF...
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
U–Pb and Lu–Hf LA-ICP(MC)-MS analyses
Zircon concentrates were prepared at Würzburg, Giessen and Frankfurt Universities using standard crushing and heavy mineral separation techniques. Selected grains were separated under alcohol and set in epoxy resin to form discs. Subsequently, the zircons were polished to expose their centres. Prior to analysis all mounts were photographed and zircon grains were imaged by SEM cathodoluminescence (CL) to identify homogeneous growth domains, using the JEOL JSM-6400 electron microprobe at the Institute of Geosciences, Frankfurt University. Selected zircon domains were analysed, following the procedure as outlined in detail by Gerdes & Zeh (2006Go). First, U–Pb analyses were carried out by LA-ICP-MS at Frankfurt University using a Thermo-Finnigan Element II sector field ICP-MS system coupled to a New Wave UP213 ultraviolet laser system. A teardrop-shaped, low-volume laser cell was used to allow the precise detection of heterogeneous material (e.g. inclusion or different growth zones) during time-resolved data acquisition (see Janousek et al., 2006Go). Each analysis consisted of c. 20 s background acquisition followed by 35 s data acquisition, using a laser spot-size of 30 µm. A common-Pb correction based on the interference- and background-corrected 204Pb signal and a model Pb composition (Stacey & Kramers, 1975Go) was carried out if necessary. The necessity of the correction is judged on whether the corrected 207Pb/206Pb lies outside the internal errors of the measured ratios. Discordant analyses were generally interpreted with care. Reported uncertainties were propagated by quadratic addition of the external reproducibility (standard deviation) obtained from the standard zircon GJ-1 (n = 12; ~0·5% and 0·5–0·9% for 207Pb/206Pb and 206Pb/238U, respectively) during individual analytical sessions (33 analyses of unknowns and 12 standards) and the within-run precision of each analysis (standard error). Concordia diagrams (2{sigma} error ellipses), concordia ages and upper intercept ages (95% confidence level) were calculated using Isoplot/Ex 2.49 (Ludwig, 2001Go). Subsequently, the same zircon domains were studied using the same laser system with a 40 µm spot-size, and the Lu, Hf and Yb isotopes were measured with a Thermo-Finnigan Neptune multi-collector ICP-MS system. The isotopes 172Yb, 173Yb and 175Lu were simultaneously monitored during each analysis step to allow the correction of isobaric interferences between Lu and Yb isotopes on mass 176. 176Yb and 176Lu were calculated using 176Yb/173Yb = 0·796218 (Chu et al., 2002Go) and 176Lu/175Lu = 0·02658 (see Gerdes & Zeh, 2006Go), and by taking the instrumental mass fractionation of each individual analysis into account. For instrumental mass bias correction Yb isotope ratios were normalized to 172Yb/173Yb = 1·35274 (Chu et al., 2002Go) and Hf isotope ratios to 179Hf/177Hf = 0·7325 using an exponential law. The protocol was tested by replicate analysis of a 50 ppb JMC 475 solution that was doped with variable amounts of pure Yb and Lu (see Gerdes & Zeh, 2006Go). The accuracy of the interference correction could be verified even for a 176Yb/177Hf ratio of 0·241 and a 176Lu/177Hf ratio of 0·021, which clearly exceed the highest observed Yb and Lu amounts in the studied zircon grains. Analyses of a pure 50 ppb JMC 475 solution yielded 176Hf/177Hf = 0·282146 ± 8 (2{sigma}, n = 22). For better comparison all analyses are reported relative to the JMC 475 176Hf/177Hf value of 0·282160. Ten and 16 LA-MC-ICP-MS analyses of the GJ-1 zircon (c. 9600 ppm Hf) during the two analytical session gave a 176Hf/177Hf ratio of 0·282007 ± 15 (2{sigma}) and 0·282006 ± 15 (2{sigma}), respectively (Table 3). This is identical, within error, to results obtained by solution MC-ICP-MS analyses of the Lu- and Yb-free Hf fraction (0·281999 ± 8, 2{sigma}, n = 10). Multiple LA-MC-ICP-MS analyses of the 91500 zircon standard over a period of 9 months yielded 176Hf/177Hf = 0·282297 ± 22 (2{sigma}, n = 112). Our results are shown in Tables 2 and 3, and in Figs 2–8GoGoGoGoGoGo. All uncertainties in this paper are quoted at the ±2{sigma} level.


Figure 2
View larger version (115K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 2. CL images of zircon grain from the Beit Bridge Complex: (a) sample SR, Sand River gneiss; (b, c) sample Bu, Bulai granite; (d–f) sample Sin, Singelele gneiss. White circles indicate position of laser spots and the estimated Pb–Pb ages with 2{sigma} error (in Ma).

 

Figure 3
View larger version (47K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 3. Concordia diagrams for samples from the Beit Bridge Complex: (a) sample SR; (b) sample Bu; (c) sample Sin; (d) sample Reg; (e) sample ZAG; (f) sample ZAL. (For further explanation see text.)

 

Figure 4
View larger version (35K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 4. Results of combined Lu–Hf and U–Pb zircon spot analyses of samples from the Beit Bridge Complex presented as 176Hf/177Hf(t) vs 207Pb/206Pb age, {varepsilon}Hf(t) vs 207Pb/206Pb age and TDM vs 207Pb/206Pb age. 176Hf/177Hf(t), initial 176Hf/177Hf calculated at the time t; t, apparent 207Pb/206Pb age obtained from the respective spot analysis; {varepsilon}Hf(t) and TDM, (hafnium model age) calculated for time t, using the constants and parameters as described in the text. •, Concordant analyses (98–102% concordance level); {circ}, discordant analyses.

 

Figure 5
View larger version (52K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 5. Cathodoluminescence images of zircon grains from the Beit Bridge Complex: (a) sample Reg, Regina gneiss; (b, c) sample ZAG, Zanzibar granodiorite gneiss; (d) sample ZAL, Zanzibar granite gneiss. White circles indicate position of the laser spots and the estimated Pb–Pb ages with 2{sigma} error (in Ma).

 

Figure 6
View larger version (113K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 6. Cathodoluminescence images of zircon grains from the Mahalapye Complex: (a–e) sample Mo, Mokgware granite; (f) sample Ma1h, garnet–biotite gneiss; (g, h) sample Ma1b, garnet-bearing leucosome. White circles indicate position of the laser spots and the estimated Pb–Pb ages with 2{sigma} error (in Ma). co, core with oscillatory zoning; cp, core with a patchy zoning, rd, dark U-rich rim; rp, rim with a patchy zoning. (For further explanation see text.)

 

Figure 7
View larger version (23K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 7. Concordia diagrams for U–Pb analyses of samples from the Mahalapye Complex; (a) sample Mo; (b) sample Ma1h; (c) sample Ma1b. (For further explanation see text.)

 

Figure 8
View larger version (26K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 8. Results of combined Lu–Hf and U–Pb zircon spot analyses of samples from the Mahalapye Complex presented as 176Hf/177Hf(t) vs 207Pb/206Pb age, {varepsilon}Hf(t) vs 207Pb/206Pb age and TDM vs 207Pb/206Pb age. •, Concordant analyses (98–102% concordance level);{circ}, discordant analyses. Black open ellipses in (g–i) mark analyses of zircon cores with oscillatory zoning (co); cp, core with a patchy zoning. Grey ellipses in (g–i) mark uranium-rich zircon overgrowths (rims). Large filled circles in (h) and (i) represent the range of initial {varepsilon}Hf(t) and TDM values, recalculated for cp zircons using the crystallization age of 2·71 Ga. (For further explanation see caption of Fig. 4 and text.)

 

View this table:
[in this window]
[in a new window]

 
Table 2: Results of in situ U–Pb zircon dating

 

View this table:
[in this window]
[in a new window]

 
Table 3: Results of in situ Lu–Hf zircon analyses

 
Constants and calculation parameters
Depleted mantle Hf model ages (TDMHf) were calculated using a decay constant of 1 · 867 x 10–11 (Scherer et al., 2001Go; Söderlund et al., 2004Go), 176Lu/177Hf = 0·0384 and 176Hf/177Hf = 0·28325 for the depleted mantle (Chauvel & Blichert-Toft, 2001Go), a mean 176Lu/177Hf value of 0·0113 for the Palaeoproterozoic–Archaean crust [mean of average continental crust as suggested by Taylor & McLennan (1985Go) and Wedepohl (1995Go)], and the measured 176Lu/177Hf value of each spot for the time since zircon crystallization. The applied decay constant is, according to various workers (e.g. Davis et al., 2005Go; Albarède et al., 2006Go; Scherer et al., 2007Go), more appropriate to derive information about Archaean crustal growth processes than that estimated by the analyses of meteorites [(1·958 ± 0·044) x 10–11; Albarède et al., 2006Go, and references herein). The use of the ‘meteoritic’ constant would increase the calculated {varepsilon}Hf(t) for our samples on average by about 2·5 epsilon units, and would give slightly younger (0·10–0·003 Ga) hafnium model ages. A more significant impact on the calculated hafnium model ages results from the assumed 176Lu/177Hf value for the Palaeoproterozoic–Archaean crust. A lower 176Lu/177Hf value of 0·007, which is typical for present-day upper crust (Taylor & McLennan, 1985Go; Wedepohl, 1995Go), would decrease the TDMHf of our zircon analyses by about 0·01–0·25 Ga (mean for all analyses = 0·10 Ga). A higher 176Lu/177Hf value of 0·015, as proposed by Rudnick & Gao (2003Go) for the bulk continental crust, would increase our calculated TDMHf by about 0·12 Ga (average). Additional uncertainties with similar influence on the calculated TDMHf result from the 176Lu/177Hf heterogeneity of depleted mantle melts (e.g. Chauvel & Blichert-Toft, 2001Go).

For calculation of {varepsilon}Hf(t) we used the chondritic uniform reservoir (CHUR) recommended by Blichert-Toft & Albarède (1997Go; 176Lu/177Hf and 176Hf/177Hf values of 0·0332 and 0·282772, respectively), and the apparent Pb–Pb ages obtained for the respective zircon domains. By means of this procedure many zircon analyses from individual granitoid samples plot on linear arrays in the {varepsilon}Hf(t) vs apparent Pb–Pb age diagram, in the initial 176Hf/177Hf vs apparent Pb–Pb age diagram, and in the TDMHf vs apparent Pb–Pb age diagram (Figs 4 and 8). Given the case that the initial 176Hf/177Hf values of all of these zircon analyses are identical, within error, such an array suggests that all zircon domains were formed at the same time, but that some of them underwent multiple Pb loss afterwards, while maintaining their initial 176Hf/177Hf (see Amelin et al., 2000Go). Thus, for all zircon analyses that plot on such an array initial {varepsilon}Hf(t) values [={varepsilon}Hf(t)initial] were recalculated using the respective magma crystallization ages (e.g. Fig. 9a and b). The same criterion was used to calculate initial TDMHf values (=TDMHfinitial), which are younger than the TDMHf obtained by using the respective apparent Pb–Pb age (Table 3). The depleted mantle curve, as shown in Fig. 9, is extrapolated from average modern-day values of mid-ocean ridge basalts (MORB) (176Lu/177Hf = 0·0384, 176Hf/177Hf = 0·28325; Chauvel & Blichert-Toft, 2001Go) assuming a linear behaviour.


Figure 9
View larger version (28K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 9. (a, b) {varepsilon}Hf(t)initial vs age diagram, showing a synopsis of all Lu–Hf results obtained from magmatic zircon domains of samples from the Central Zone of the Limpopo Belt. (a) Individual {varepsilon}Hf(t)initial values recalculated using the intrusion age of the respective granitoids (see Figs 3 and 7). The chondritic uniform reservoir (CHUR) and the depleted mantle array were calculated with parameters defined in the text. The grey bands represent {varepsilon}Hf(t) bulk-rock evolution trends, calculated using 176Lu/177Hf of 0·0113. (b) Summary of {varepsilon}Hf(t) values obtained from granitoids from the Beit Bridge and Mahalapye complexes. The small black circles (TDM) mark the time of juvenile crust formation in the Limpopo Belt. The arrows starting at TDM = 3·65 and 2·65 Ga illustrate the evolution trend of a bulk rock with a 176Lu/177Hf of 0·0113. The dotted vertical line defines the mixing line between evolved Palaeoarchaean crust and juvenile Neoarchaean crust at 2·65 Ga. The black arrows starting at 3·283 Ga (Sand River gneiss data) define the {varepsilon}Hf(t) evolution trends of zircon (*) crystallized at 3·283 Ga in the Sand River gneiss, and of melts, which may have formed by partial anatexis (+ zircon fractionation) of the Sand River gneiss. (c) {varepsilon}Nd(t) vs age diagram showing data for rocks from the Kaapvaal Craton, the Zimbabwe Craton and the Limpopo Belt. It should be noted that {varepsilon}Nd(2·6–2·7 Ga) values of rocks from the Zimbabwe Craton and the Northern Marginal Zone are commonly higher than those for the Kaapvaal Craton and the Southern Marginal Zone. Data sources: (1) Maier et al. (2000Go); (2) Schmitz et al. (2004Go); (3) Wilson & Carlson (1989Go) and Kreissig et al. (2000Go); (4) Barton et al. (1995, 1996); (5) Berger et al. (1995Go); (6) Barton (1996Go); (7) Zhai et al. (2006Go); (8) Hamilton et al. (1977Go), Moorbath et al. (1986Go), Taylor et al. (1991Go) and Jelsma et al. (1996Go). (For further explanation see text.)

 

    RESULTS AND INTERPRETATION OF U–PB AND LU–HF ZIRCON ISOTOPE ANALYSES
 TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL SETTING, CURRENT...
 ANALYTICAL TECHNIQUES
 RESULTS AND INTERPRETATION OF...
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Beit Bridge Complex
Sand River gneiss (sample SR)
This tonalitic gneiss sample, which mainly consists of plagioclase, quartz, biotite, minor hornblende, and orthopyroxene, was taken from the bed of the Sand River near the Causeway locality south of Messina (Fig. 1, Table 1). It represents a grey banded gneiss that is truncated and cut by numerous later melt batches. CL images of most zircon grains show cores with oscillatory zoning patterns, which are in some grains overgrown by structureless rims (Fig. 2a). In addition, a few completely structureless zircon crystals are observed (Fig. 2a). During this study 27 U–Pb spot analyses were carried out on cores and rims of 21 zircon grains (Table 2). In addition, Lu–Hf isotope spot analyses of nine zircon cores were obtained (Table 3). As shown in Fig. 3a, four groups of concordant to near-concordant U–Pb ages can be distinguished. Zircon cores yield concordant ages of about 3·28 Ga and one of 3·14 Ga, whereas zircon overgrowths and structureless zircon grains yield concordant ages of about 2·02 Ga and one of c. 2·61 Ga (Figs 2a and 3a). All four concordant age groups are also reflected in the Pb–Pb intercept ages of discordant zircon analyses (Table 2, Fig. 3a).

We suggest that the oldest concordant zircon grains, which yield a mean age of 3283 ± 8 Ma, identical to an upper intercept age of 3280 ± 7 Ma (Fig. 3a), reflect the time of the Sand River gneiss protolith intrusion, whereas the younger concordant age of 3143 ± 13 Ma dates the time of zircon growth during a Palaeoarchaean anatectic event. Our estimated intrusion age is identical to or slightly younger than that obtained by SHRIMP U–Pb analyses from the Sand River granitoid suite at the Causeway locality (3314 ± 5 Ma, 3240 ± 5 Ma), the Macloutse farm (3290 ± 17 Ma), Bellvue farm (3297 ± 7 Ma), Verbaard farm (3296 ± 4 Ma) and Vrouenroom farm (3292 ± 4 Ma), but older than zircon SHRIMP ages obtained from the Esmefour farm (3205 + 69/–62 Ma) and the Sand River bed (3240 ± 5 Ma; 3197 ± 30 Ma) (Kröner et al., 1998Go, 1999Go). The suggested Palaeoarchaean anatectic event at 3143 ± 13 Ma conforms with field relationships and geochronological data; the highly deformed and migmatitized Sand River gneiss is locally cut by basalt dykes, which yielded Rb–Sr and Pb–Pb ages of 3060 Ma and 2922 Ma, respectively (Barton et al., 1977Go; Barton et al., 1990Go).

The upper intercept age of 2614 ± 11 Ma is interpreted to reflect zircon formation in the Sand River gneiss during a Neoarchaean metamorphic–magmatic event, whereas the upper intercept age of 2014 ± 10 Ma (five spots, MSWD = 1· 0) dates zircon growth during the Palaeoproterozic granulite-facies metamorphic overprint. The latter age is slightly younger than U–Pb isotope dilution thermal ionization mass spectrometry (ID-TIMS) and Pb–Pb zircon evaporation ages of 2031 ± 6 and 2026 ± 1 Ma, respectively, which were obtained on metamorphic zircon grains from pelitic gneisses of the Causeway locality (Jaeckel et al., 1997Go), but identical within error to the upper intercept ages of strongly discordant U–Pb ID-TIMS zircon analyses of 2006 ± 8 Ma and 2003 ± 11 Ma, which were obtained from zircon grains from melt patches in the Sand River gneisses and from a pegmatitic granite (Jaeckel et al., 1997Go). The 2614 Ma age is identical within error to the intrusion age of several orthogneisses found throughout the Central Zone (Jaeckel et al., 1997Go; Kröner et al., 1999Go; this study). However, it is significantly younger than the two concordant U–Pb SHRIMP analyses of about 2734 ± 4 Ma in a c. 3·29 Ga quartz monzonitic augen gneiss, which were interpreted to represent a local melting event during the Neoarchaean (Kröner et al., 1999Go).

The Lu–Hf analyses of eight zircon cores yielded relative homogeneous initial 176Hf/177Hf values between 0·28063 and 0·28066, which correspond to {varepsilon}Hf(t)initial between –0·1 and –1· 6, and provide evidence that the zircon cores crystallized in a magma, which was derived from a source with an average crustal residence age of TDMHfinitial 3·64 ± 0·04 Ga. A slightly higher initial 176Hf/177Hf value (0·28071) was obtained from the 3·14 Ga zircon domain (spot 32c/zrc19) interpreted to be formed during Palaeoarchaean anatexis about 140 Myr after the intrusion of the Sand River gneiss protolith (Figs 2a and 3a). The increased initial 176Hf/177Hf is consistent with the incorporation of additional radiogenic hafnium during anatectic zircon crystallization. This additional 176Hf was formed by 176Lu decay in the Sand River gneiss matrix between 3·28 and 3·14 Ga.

The presented hafnium model ages for the 3·28 Ga zircon cores are in line with the U–Pb age data, but older than the TDM values of 3·1–3·4 Ga, which were derived by Sm–Nd whole-rock isotope analysis of the Sand River gneisses (Kröner et al., 1999Go). The younger neodymium model ages may result from Palaeoarchaean open-system behaviour, which led to a resetting of the Sm–Nd system (see Moorbath et al., 1997Go). In combination with the U–Pb ages, our TDMHfinitial values indicate that the crustal source was extracted from a depleted mantle reservoir about 300 Myr prior to the formation of the Sand River granitoid suite.

Bulai granite (sample Bu)
Sample Bu is a porphyric, weakly deformed granite containing K-feldspar, plagioclase, biotite and quartz, and was taken from the type locality of the Bulai pluton (Table 1). From this sample 27 U–Pb spot analyses on 24 zircon grains and 19 Lu–Hf analyses were carried out (Tables 2 and 3; Fig. 4d–f). CL images of most zircon crystals reveal an oscillatory magmatic zoning and two bright zones (cores and rims), which are separated by a dark band of weak luminescence (Fig. 2b). Some zircon grains show additional overgrowths with a diffuse luminescence (Fig. 2c). U–Pb analyses obtained from all of these distinct zircon domains yielded within error identical ages (Fig. 2b and c), indicating that they were formed during the same magmatic event. Fourteen concordant U–Pb zircon analyses (99–101% concordance level) gave an age of 2612 ± 7 Ma, which is within error identical to an upper intercept age of 2607 ± 7 Ma obtained from five discordant analyses plus the 14 concordant zircon analyses (Fig. 3b). The lower intercept age of 790 ± 67 Ma is interpreted as geologically meaningless and may result from multiple Pb loss. Furthermore, there are seven U–Pb analyses that group together with the 14 concordant analyses on a discordia with a lower intercept at 1943 ± 150 Ma (Fig. 3b). This suggests that some zircon domains were affected by Pb loss during the Palaeoproterozoic, which is in agreement with previous age data of Barton & Sergeev (1997Go), Jaeckel et al. (1997Go) and Holzer et al. (1998Go).

The concordant 2612 ± 7 Ma U–Pb zircon age of the Bulai pluton is interpreted to reflect the time of emplacement. This age is identical within error to a U–Pb zircon SHRIMP upper intercept age of 2620 ± 8 Ma from a granitic phase (Kröner et al., 1998Go) and an ID-TIMS U–Pb zircon upper intercept age of 2605 ± 2 Ma from a deformed enderbitic phase of the Bulai pluton (Barton et al., 1994Go). However, it is significantly older than a U–Pb ID-TIMS zircon age of 2572 ± 4 Ma from a granitic phase (Barton et al., 1994Go) and a Pb–Pb zircon evaporation age of 2587 ± 1 Ma from a deformed enderbitic phase (Kröner et al., 1999Go). It should be noted that the younger zircon age reported by Barton et al. (1994Go) is an upper intercept U–Pb age, which results from the fitting of only two highly discordant zircon analyses. Thus, it cannot be excluded that these zircon grains were subjected to multiple Pb loss, an effect that is well documented by our in situ LA-ICP-MS analyses (Fig. 3b). Multiple Pb loss may also account for the young Pb–Pb evaporation age (2587 ± 1 Ma) obtained by Kröner et al. (1999Go).

Nineteen Lu–Hf analyses of concordant and discordant zircons were carried out. As shown in Fig. 4d, the zircons have initial 176Hf/177Hf ratios of about 0·28102 ± 0·00003 (2{sigma} SD = standard deviation), which correspond to {varepsilon}Hf(t)initial between –2·3 and –4·3 when the 2612 Ma Bulai granite crystallization age is applied to all Hf analyses. This procedure is justified, as the initial 176Hf/177Hf values of all analysed zircon domains show relatively minor variations. Nevertheless, there seems to be a small increase of the initial 176Hf/177Hf with decreasing Pb–Pb age (Fig. 4d), a feature that can be explained in different ways. Taking the errors into account, the initial 176Hf/177Hf of the apparently younger domains is identical to that obtained from most of the apparently older domains. Thus, the observed array can be interpreted (in the most simple way) to reflect multiple Pb loss in younger zircon domains, which preserved their initial 176Hf/177Hf incorporated during magma crystallization. Alternatively, the small increase of the initial 176Hf/177Hf could reflect some minor incorporation of more radiogenic Hf in the younger zircon domains, perhaps during partial recrystallization of metamict zircon domains caused by fluid–zircon–matrix interactions. Initial hafnium model ages of 3·22 ± 0·06 Ga indicate that the Bulai granite was derived from a crustal source, which was extracted on average from a depleted mantle reservoir 600–700 Myr prior to the Bulai granite intrusion.

Singelele granodiorite gneiss (sample Sin)
Sample Sin is a high-grade granite gneiss, which was collected from the type locality at the Singelelekop near Messina (see Kröner et al., 1999Go). It is characterized by the mineral assemblage plagioclase–K-feldspar–biotite–quartz–garnet. A total of 53 spots on 43 zircon grains from this sample were analysed for U–Pb and 45 spots for Lu–Hf (Tables 2 and 3; Figs 3c and 4g–i). The calculated 207Pb/206Pb ages show large variations from 1·54 to 2·66 Ga with 38 of 53 analyses yielding discordant results (below or above 95–105% concordance; Table 2, Fig. 3c). For many zircon grains core and rim domains were distinguished by their CL images (Fig. 2d–f). However, this textural relationship is not always reflected by the geochronological data. Based on the texture–(Pb–Pb)-age relationships three zircon groups can be distinguished: (1) zircon grains with significantly younger cores than rims (Fig. 2d); (2) zircon grains with nearly identical Pb–Pb ages of cores and rims (Fig. 2d); (3) zircon grains with older cores and younger rims (Fig. 3f). These different texture–age relationships are interpreted to reflect that the zircon cores and rims grew either at the same time or at different times [textures (2) and (3)], and/or that the distinct zircon domains were affected by alteration processes causing non-zero or multiple Pb loss, effects that have been described in detail by several researchers (e.g. Vavra et al., 1996Go; Geisler et al., 2001Go).

Whereas the texture–age relationships are difficult to interpret, the additional Hf isotope data reveal a relatively simple story. In fact, they allow us to distinguish between zircon zones that either are affected by alteration or are the result of new zircon growth. Most zircon domains (Fig. 4g) have, despite their different apparent Pb–Pb ages, which range from 1539 to 2658 Ma, very similar initial 176Hf/177Hf values of about 0·28099 ± 0·00005 (2{sigma} SD). Therefore, they display a simple trend in the {varepsilon}Hf(t) vs age diagram, which is characterized by a positive correlation between apparent zircon Pb–Pb ages and {varepsilon}Hf(t) (Fig. 4h). This trend is similar to that observed for the Bulai zircons (Fig. 4e), although more pronounced. It suggests that all of these zircon domains or grains (including most zircon cores and rims of sample Sin1; Tables 2 and 3) were formed during the same magmatic event, but were subjected to Pb loss with a different intensity. It should be noted that the calculated TDMHf of these zircons become apparently older with decreasing apparent Pb–Pb age (Fig. 4i). This effect results from the two-stage model used. Thus, for the geological interpretation only initial hafnium model ages (TDMHfinitial) can be used. These are between 3·2 and 3·5 Ga for the Singelele gneiss.

The Hf isotope analyses indicate that nearly all zircon grains or domains in sample Sin crystallized during a single magmatic event, and that the zircon grains are free of older inheritance. Thus, the oldest Pb–Pb and concordant ages obtained from these zircon domains are nearest to the crystallization age. Eleven spot analyses define a discordia with an upper intercept at 2646 ± 10 Ma, and nine of those define an identical concordia age of 2647 ± 12 Ma, which is interpreted as the intrusion age of the Singelele granite gneiss (Fig. 3c). In contrast, all younger concordant and discordant zircons, which have similar initial 176Hf/177Hf values (0·28099 ± 0·00005, 2{sigma} SD) are assumed to have undergone Pb loss. Among them are seven that fall on a discordia with a lower intercept age of 1989 ± 160 Ma (Fig. 3c), indicating a Palaeoproterozoic Pb loss similar to that obtained for sample Bu (Fig. 3b). However, the large scatter of the discordant analyses indicates that most zircon domains have undergone multiple Pb loss. In this context, it should be noted that the Pb–Pb evaporation ages and discordant U–Pb ID-TIMS ages of 2568–2582 Ga for the Singelele gneiss (Jaeckel et al., 1997Go) could date the time of zircon alteration (Pb loss) rather than zircon formation.

In contrast to the zircon analyses discussed above, six zircon rim analyses, which show significantly higher initial 176Hf/177Hf values of up to 0·28166, do not follow the general {varepsilon}Hf(t) trend and show significantly younger TDMHf ages extending to 2·2 Ga (Fig. 4g–i). A concordant U–Pb analysis of one of these rims (spot 27r/zrc3) indicates that it was formed at 2042 ± 19 Ma. This age is slightly younger than two Pb–Pb upper intercept ages of 2081 ± 20 Ma and 2091 ± 13 Ma, which were obtained for two other overgrowths (spots 15r/zrc19 and 20r/zrc23), but significantly older than three discordant Pb–Pb zircon ages of about 2196, 2444 and 2538 Ma (spots 54r/zrc38, 47r/zrc35 and 55r/zrc39). The three younger zircon ages suggest that these zircon overgrowths were formed during the Palaeoproterozoic, perhaps during the high-grade metamorphic event that was dated at 2·02–2·05 Ga (e.g. Barton & Sergeev, 1997Go; Jaeckel et al., 1997Go; Holzer et al., 1999Go). During this event the Singelele orthogneiss underwent partial anatexis, as evident from a few melt patches observed in the outcrops at the Singelelekop. The three older Pb–Pb ages could reflect mixing ages, which result from the analyses of older (2·65 Ga) and younger (2·04 Ga) zircon domains. These ages could not be resolved by the technique used. Consequently, their 176Hf/177Hf initial ratios should also represent mixtures.

The distinctly higher initial 176Hf/177Hf values of the six rims are thought to be due to partial melting of the Singelele orthogneiss at c. 2·04 Ga. During that melting event most of the c. 2·65 Ga zircon grains remained undissolved. Consequently, only minor amounts of Hf, incorporated by these zircon crystals (= zircon hafnium), were released into the melt from which the new zircon overgrowths were formed. Thus, the zircon overgrowths incorporated abundant radiogenic 176Hf formed by 176Lu decay in the granite gneiss matrix (= matrix hafnium) between 2·65 and 2·04 Ga. This assumption is best supported by spot 27r/zrc3, which shows nearly identical zircon crystallization and TDMHf ages of 2·04 and 2·2 Ga, respectively (Table 2 and 3; Fig. 4i). In contrast, the other five spot analyses yield less radiogenic Hf, thereby indicating that these zircon domains incorporated a mixture of highly radiogenic ‘matrix hafnium’, and weakly radiogenic ‘zircon hafnium’, released by partial dissolution of 2·65 Ga zircons during anatexis.

In summary, the combined U–Pb and Lu–Hf analyses of the zircon grains indicate that the protolith of the Singelele orthogneiss intruded at about 2646 ± 10 Ma and that the magma contained considerable amounts of an older crust, as reflected by average crustal residence ages between 3·2 and 3·5 Ga. The Singelele gneiss underwent partial anatexis during the Palaeoproterozoic at about 2·0 Ga, which caused Pb loss in many zircon domains and the formation of a few zircon overgrowths incorporating abundant highly radiogenic ‘matrix hafnium’.

Regina granite gneiss (Reg)
This gneiss sample was collected from the Regina farm c. 30 km north of Alldays, and contains mainly plagioclase, K-feldspar, quartz, amphibole, garnet, and magnetite. The Regina gneiss is part of the Krone metamorphic terrane forming the basement below the Venetia klippen complex, which underwent an amphibolite-facies overprint at about 2·06–2·0 Ga (Barton et al., 2003Go; Klemd et al., 2003Go; Zeh et al., 2005aGo, 2005bGo). So far, nothing is known about the timing of the orthogneiss protolith intrusions in this area. From sample Reg, 22 U–Pb spot analyses on 13 zircon grains and 11 Lu–Hf spots were analysed (Tables 2 and 3; Figs 3d and 4j–l).

The CL images indicate that most zircon grains have oscillatory zoning and some of them distinct cores and rims (Fig. 5a). Twelve U–Pb spot analyses yield similar 207Pb/206Pb ages of 2580–2660 Ma for the different domains, indicating that the grains formed during a Neoarchaean magmatic event. In detail, however, more than one age population seems to exist. Five grains yielded a concordia age of 2649 ± 9 Ma, which is identical to an upper intercept age of 2651 ± 6 Ma as defined by eight analyses (Fig. 3d). Three other zircon analyses, however, yielded a significantly younger concordia age of 2600 ± 11 Ma and one analysis a concordia age of 2530 ± 16 Ma. The remaining 10 spots gave 20–79% discordant results with 207Pb/206Pb ages that vary from 1684 to 2487 Ma. We interpret the older concordant age to represent the intrusion age of the Regina gneiss protolith, whereas the younger ages result from partial Pb loss either at about 2·0 Ga or during multiple events (inset in Fig. 3f).

All zircon grains have within error identical initial 176Hf/177Hf values of 0·28096 ± 0·00004 (2{sigma} SD), which correspond to {varepsilon}Hf(t)initial of –3·4 to –5·5 and TDMHfinitial of 3·3–3·4 Ga. It is worth noting that these values are similar to that of the Bulai granite and to most values obtained from the Singelele gneiss.

Zanzibar granodiorite gneiss (sample ZAG)
The investigated grey gneiss sample was taken from the Seoka river bed near the Botswana–South Africa boundary (see Barton & Key, 1983Go). From this sample 23 U–Pb and Lu–Hf spot analyses were performed on 17 zircon grains (Tables 2 and 3; Figs 3e and 4m–o). CL images indicate complexly zoned zircon grains and most of them reveal a core and rim structure (Fig. 5b and c). In addition, there are abundant zircon grains that show parallel banding (Fig. 5b). U–Pb spot analyses provided a wide scatter of apparent Pb–Pb ages for all domains ranging between 2227 and 2629 Ma (Table 2). Some zircon grains yield within error identical ages for their core and rim (e.g. spots 9 and 10, Fig. 5c) whereas other zircon grains show apparently younger cores and older rims (e.g. spots 24 and 25, Fig. 5b). The latter indicates that the analysed core domains underwent a much stronger Pb loss than the rim domains of the same zircon.

Eight U–Pb analyses (99–101% concordance) yielded a concordia age of 2613 ± 7 Ma (Fig. 3e), which is interpreted as the time of granodiorite intrusion. Five apparently concordant zircon analyses (96–98% concordance) gave ages of around 2540 and 2596 Ma (Table 2). These ages are interpreted to result from partial Pb loss during a Palaeoproterozoic (c. 2·0 Ga) metamorphic event. The remaining eight spot analyses are 77–94% discordant, reflecting either partial Pb loss to zero or multiple Pb loss. It should be noted that the intrusion age of 2613 ± 7 Ma is significantly younger than the Rb–Sr whole-rock isochron age of 3227 ± 40 Ma from the same gneiss (Barton & Key, 1983Go). However, it is identical to or slightly younger than two SHRIMP U–Pb zircon ages of 2614 ± 13 Ma and 2659 ± 10 Ma of a granodiorite and monzonitic gneiss, respectively, which were collected from the nearby Tapalaphala river bed (Kröner et al., 1999Go).

Lu–Hf analyses reveal that all 23 zircon spots have within error identical initial 176Hf/177Hf values of 0·28108 ± 0·00005 (2{sigma} SD; Fig. 4m). This supports the conclusion that all investigated zircon domains were formed during the same magmatic event and that all of them underwent partial Pb loss during multiple events. This conclusion is also reflected by the array of the {varepsilon}Hf(t) values, which shows a straight line between –1·2 at 2·62 Ga and –9·4 at 2·2 Ga (Fig. 4n). Initial hafnium model ages between 3·01 and 3·18 Ga reveal that the Zanzibar gneiss results from remelting of substantial amounts of an older Archaean crust, which was formed on average c. 400–500 Myr prior to the intrusion.

Zanzibar granite gneiss (sample ZAL)
The Zanzibar granite gneiss sample was also collected in the Seoka river bed. Twenty U–Pb and 15 Lu–Hf spot analyses were obtained on 16 zircon grains from this sample (Tables 2 and 3; Figs 3f and 4p–r). Many zircon grains have cores, which are separated from their rims by dark, U-rich bands (Fig. 5d). Most of the cores show oscillatory zoning, whereas some of the rims are structureless.

U–Pb spot analyses of all zircon domains yield discordant results (97–48% discordance) with apparent Pb–Pb ages between 1785 and 2619 Ma (Table 2). However, the initial 176Hf/177Hf values of all spot analyses are identical within error (0·28108 ± 0·00005; 2{sigma} SD). This indicates that all analysed zircon domains, comprising the different core and rim areas, must have been formed during the same magmatic event, which is also supported by the {varepsilon}Hf(t) and TDMHf vs Pb–Pb age trends (Fig. 4q and r). The TDMHfinitial values indicate that the granite gneiss protolith stems from a crustal source with an average hafnium model age of 3·02–3·18 Ga, which is identical to that of the Zanzibar granodiorite.

Nine U–Pb spot analyses fall on a discordia (MSWD = 1· 9) with intercepts at 2614 ± 9 Ma and 716 ± 46 Ma (Fig. 3f). The upper intercept age of 2614 ± 9 Ma is interpreted to date the time of the ZAL granite intrusion. In fact, this age is identical to that obtained from the nearby sample ZAG, which shows an identical Hf isotope composition (see above). The scatter of various data points in the concordia diagram of sample ZAL indicates than many zircon domains underwent post-crystallization multiple Pb loss, perhaps during Palaeoproterozoic and recent time (see inset in Fig. 3f).

Mahalapye Complex
Mokgware granite (sample Mo)
The investigated sample represents a medium-grained granite variety, which contains mainly K-feldspar, plagioclase, biotite and quartz. Seventeen U–Pb spot analyses on 14 zircon grains and an 14 Lu–Hf spot analyses were carried out (Tables 2 and 3). CL images indicate different zircon zoning patterns (Fig. 6a–e). Type-I zircons show U-poor cores with a bright luminescence and ‘dark’ U-rich overgrowths (Fig. 6c and d). Some of these cores display oscillatory zoning, whereas others are structureless–diffuse (Fig. 6d). The latter are similar to recrystallized metamict zircons as described by Vavra et al. (1996Go). Type-II zircons show a balanced luminescence throughout the entire crystal with a weak oscillatory zoning, although some of them have distinct cores and rims (Fig. 6a, b and e).

U–Pb spot analyses of most zircon domains from type-I zircon and from the oscillatory cores of type-II zircon define a discordia with intercepts at 10 ± 12 and 2019 ± 8 Ma (Fig. 7a). The upper intercept age is within error identical to a concordia age of 2026 ± 10 Ma, which was calculated from three zircon analyses (Fig. 7a). It is interpreted to reflect the time of granite emplacement. In contrast, two younger, strongly discordant Pb–Pb analyses, which have apparent Pb–Pb ages of about 1463 and 1721 Ma, are interpreted to result from multiple Pb loss after 2·03 Ga. These two analyses show the highest U contents (1432 and 2832 ppm, respectively). In addition, a few zircon cores give significantly older ages. One zircon core yields a concordant age of 2929 ± 14 Ma (spot 8c/zrc6), and three others have discordant Pb–Pb ages between 2520 and 2645 Ma (Fig. 7a). The age data indicate that the Mokgware granite intruded at 2019 ± 8 Ma and contains some Archaean crustal components. This is in agreement with the results obtained from the Lu–Hf spot analyses (Fig. 8a–c), which reveal Archaean hafnium model ages between 3·0 and 3·4 Ga for the different zircon domains (Fig. 8c).

The 10 spots that define the 2·02 Ga Mokgware discordia age show slightly variable initial 176Hf/177Hf of 0·28120 ± 0·00005 (2{sigma} SD). This suggests that magma homogenization during zircon crystallization was incomplete (TDMHfinitial = 3·1–3·2 Ga; {varepsilon}Hf(t)initial = –10·5 ± 1· 9). The inherited components with variable age (>2·02 Ga) and Hf isotope composition ({varepsilon}Hf(at 2019 Ma) = –11· 3 to –20·1) clearly point to a heterogeneous crustal source.

It should be noted that the assimilation–melting of Archaean material during the Palaeoproterozoic granite formation in the Mahalapye Complex is not restricted to the Mokgware granite, but has also been reported from the nearby Mahalapye granite (Fig. 1). This granite intruded at 2023 ± 11 Ma and contains abundant zircon xenocrysts with Pb–Pb ages between 2·45 and 3·15 Ga (McCourt & Armstrong, 1998Go). The age of the oldest xenocryst is well within the range of the hafnium model ages, which are estimated in this study for the Mogkware granite.

Lose quarry samples
The Lose quarry is situated c. 50 km south of Palapye (Fig. 1) and exposes ortho- and paragneisses that are transected by garnet-bearing leucosomes. The leucosomes form veins of centimetre to several metres width, which occupy c. 25% of the outcrop volume. A detailed description of this outcroup, lithologies and metamorphic conditions has been given by Chavagnac et al. (2001Go) and Hisada et al. (2005Go). For this study two samples were collected, a dark garnet–biotite gneiss and a leucosome.

Garnet–biotite gneiss (sample Ma1h). This sample shows melanocratic and leucocratic domains on the centimetre scale, both of which contain garnet, biotite, plagioclase, quartz and gem quality zircon grains. CL images show typical magmatic zoning patterns such as oscillatory zoning, sector zoning and a similarly bright luminescence for all zircon grains (Fig. 6f). Twenty U–Pb spot analyses on 20 zircon grains and 15 Lu–Hf spot analyses were carried out (Table 2 and 3). All U–Pb analyses fall on a discordia with intercepts at zero and 2058 ± 5 Ma, which is identical to the concordia age of 2061 ± 6 Ma as defined by 17 analyses (Fig. 7b). The Lu–Hf analyses of all zircon grains yielded identical initial 176Hf/177Hf of 0·28145 ± 0·00002 (Fig. 8d), which corresponds to {varepsilon}Hf(t)initial between 0·1 and –1·2 (Fig. 8e) and TDMHfinitial of 2·61–2·67 Ga (Fig. 8f).

Similar characteristic CL patterns, Hf isotope data and identical U–Pb ages of all zircon grains indicate that the protolith of the garnet–biotite gneiss was a magmatic rock, perhaps a diorite, which was emplaced at 2061 ± 6 Ma, and syn- or post-intrusively deformed and metamorphosed. Our LA-ICP-MS U–Pb age is identical to, but more precise than, the zircon SHRIMP U–Pb upper intercept age of 2053 ± 21 Ma, which was obtained by McCourt & Armstrong (1998Go) from a granodiorite dyke of the same outcrop. In contrast to McCourt & Armstrong (1998Go), we did not find zircon xenocrysts with ages between 2·6 and 3·19 Ga in our sample.

The proposed magmatic origin for the garnet–biotite gneiss contradicts the findings of Chavagnac et al. (2001Go), who designated texturally similar rocks as migmatitic paragneisses, which were interpreted to be due to in situ anatexis of metagreywackes. However, in such a scenario one would expect to find a more heterogeneous zircon population with different inherited cores (see Zeh et