Hafnium isotope record of the Ancient Gneiss Complex, Swaziland, southern Africa: evidence for Archaean crust–mantle formation and crust reworking between 3.66 and 2.73 Ga

Abstract: Combined U–Pb and Lu–Hf isotope analyses of zircons from 16 tonalite–trondjemite–granodiorite (TTG) gneiss and granite samples from Swaziland reveal that the oldest rocks of the Ancient Gneiss Complex in southern Africa formed by reworking of Early Archaean or perhaps Late Hadean crust at 3.66 Ga, and that new crust was extracted from a depleted mantle source during Palaeoarchaean events between 3.54 and 3.32 Ga. This interpretation is supported by εHft of −1.6 ± 2.0 obtained from 3.66 Ga TTG gneisses, corresponding to hafnium model ages between 3.77 ± 0.18 Ga, for a presumed Hadean–Early Archaean chondritic mantle, and 4.08 ± 0.18 Ga, for a presumed Hadean depleted mantle reservoir, with the first model age being the most likely in the light of recent data from worldwide sources. Furthermore, it is reflected by superchondritic εHft up to +2.2 ± 2.0 for TTGs formed at 3.54, 3.45 and 3.32 Ga. The new datasets additionally show that the Palaeoarchaean crust formed between 3.54 and 3.32 Ga was intensely reworked afterwards, without significant addition of depleted mantle derived material, during orogenic and intracratonic melting processes at 3.23, 3.1 and 2.7 Ga. This is well reflected by an array of decreasing εHft from +2.2 to −7.2 between 3.3 and 2.7 Ga, which can be forced by 176Lu/177Hf of 0.0113, which is similar to that of present-day average continental crust, and might result from lower crust zircon fractionation during Archaean crust reworking. Supplementary material: Results of in situ U–Pb and Lu–Hf isotope zircon analyses and concordia diagrams are available at www.geolsoc.org.uk/SUP18465.

There are continuing debates and speculations about the origin of the oldest crust on Earth, as well as about the crust-mantle evolution during Hadean to Archaean times (e.g. Armstrong 1991;Amelin et al. 1999Amelin et al. , 2000Kamber et al. 2003Kamber et al. , 2005Condie et al. 2005;Harrison et al. 2005Harrison et al. , 2008Tolstikhin et al. 2006;Valley et al. 2006;Kramers 2007;Blichert-Toft & Albarède 2008;Nutman et al. 2008;Zeh et al. 2008Zeh et al. , 2009Kemp et al. 2009Kemp et al. , 2010Hawkesworth et al. 2010). Some of these speculations result from the scarcity of information about Hadean to Early Archaean rocks, owing to their limited exposure worldwide, but also from crust alteration processes, which can cause resetting or disturbance of certain isotope systems (e.g. Sm-Nd) during post-Hadean-Eoarchaean events, and decoupling of certain isotope systems during multiple geological events (e.g the Sm-Nd whole-rock from the U-Pb zircon system; Vervoort et al. 1996;Moorbath et al. 1997).
Despite these uncertainties, much progress has been made over the past 10 years, in particular by the application of combined U-Pb and Lu-Hf isotope analyses to zircons from (meta)sedimentary rocks and orthogneisses from many Archaean gneiss terranes worldwide (e.g. Patchett et al. 1981;Vervoort et al. 1996;Griffin et al. 2004;Davis et al. 2005;Halpin et al. 2005;Hartlaub et al. 2006;Zeh et al. 2008Zeh et al. , 2009Kemp et al. 2009Kemp et al. , 2010. In this context it is surprising to note that little is known so far about the crust-mantle processes that led to the formation and reworking of the oldest basement units of the Kaapvaal Craton, exposed in the Ancient Gneiss Complex of Swaziland ( Fig. 1). This basement complex consists of tonalitetrondjemite-granodiorite (TTG) gneisses as old as 3.66 Ga (Compston & Kröner 1988;Schoene et al. 2008Schoene et al. , 2009, and was additionally modified during magmatic events at c. 3.55, 3.45, 3.23, 3.1 and finally at 2.7 Ga (Kröner et al. 1989;Kröner & Tegtmeyer 1994; see summary by Schoene et al. 2009); that is, over a period of nearly 1 Ga. Present conclusions about the crust-mantle evolution of these rocks are mainly based on Sm-Nd isotope data (Carlson et al. 1983;Hegner et al. 1984;Kröner et al. 1996;Kröner & Tegtmeyer 1994;Schoene et al. 2009), whereas Lu-Hf data are reported only from two c. 3.24 Ga tonalite-trondjhemite gneisses east of Manzini ). The Sm-Nd dataset, which has been summarized by Schoene et al. (2009), reflects a wide scatter of åNd t data for rocks with ages around 3.55-3.50 Ga (À4 to +6), but also for the oldest gneisses formed at 3.66 Ga (0 and +7). In fact, these data point to an interpretation wherein the oldest basement of Swaziland was subject to intense reworking of older crust accompanied by the intrusion of magmas derived from a depleted or even strongly depleted mantle source (+7.0 at 3.66 Ga). We note, however, that most åNd t values obtained from these rocks were calculated on the basis of U-Pb zircon ages, which were only rarely directly obtained from the rocks used for Sm-Nd isotope analysis; in most cases ages were obtained from rocks that are assumed to belong to the same rock unit (see Schoene et al. 2009, table 1). Thus, it is not clear whether the åNd t of all rocks was calculated for the correct intrusion age. Furthermore, it is worth noting that the gneisses of the Ancient Gneiss Complex underwent a complex history between 3.66 and 3.0 Ga, as is reflected by complex age patterns obtained from single zircons using in situ U-Pb sensitive high-resolution ion microprobe (SHRIMP) dating (Compston & Kröner 1988;Kröner et al. 1989). At present, it is unclear whether the different zircon 'SHRIMP ages' reflect new zircon formation or just alteration, and if the different events were accompanied by the addition of new crustal material.
Most of the problems mentioned above can be circumvented using recent developments in analytical techniques, in particular by applying combined cathodoluminescence (CL) imaging and in situ U-Pb and Lu-Hf isotope analyses to single zircon grains or well-defined zircon domains. This approach has been proven to be extremely useful in unravelling the magmatic-metamorphic history of complex polymetamorphic Archaean gneiss terranes (e.g. Kemp et al. 2009Kemp et al. , 2010Millonig et al. 2010;Zeh et al. 2010a). Besides the advantage of spatial resolution, combined U-Pb and Lu-Hf isotope patterns additionally allow distinction between zircon zones formed or altered during distinct metamorphic events ), even if primary features such as initial U-Pb ages or zoning patterns were erased during later alteration events (Zeh et al. 2010a,b). In this study we will apply this method to zircons from 16 TTG gneiss and granite samples from Swaziland to set unambiguous constraints for the crust-mantle evolution of the oldest rocks of the Kaapvaal Craton, and their subsequent Archaean evolution until 2.7 Ga. The sample locations and coordinates are shown in Figure 1 and Table 1.

Sample preparation
Zircon grains were selected from 2-5 kg rock samples by means of standard crushing techniques (jaw crusher, disc mill), a Wilfley table and heavy liquids. Subsequently, zircon grains of the respective samples were selected by hand-picking under a binocular microscope, mounted in epoxy resin (up to seven samples per 1 inch block), and ground down to expose their centres. After polishing, but prior to isotope analyses, all grains were imaged using a Jeol JSM-6490 scanning electron microscope (SEM) with Gatan MiniCL at Goethe University Frankfurt (GUF) to obtain information about their internal structure (Fig. 2). Based on the CL images, between 12 and 41 U-Pb and Lu-Hf analyses were obtained from selected zircon domains of each sample (see Table 2). In a first step, U-Pb analyses were carried out, and during a later session Lu-Hf isotope analyses were obtained from the same domains, by setting the Lu-Hf spot either directly 'on top' of the U-Pb spot or immediately beside it but within the same growth zone as obtained by CL (see Fig. 2).

LA-SF-ICP-MS U-Pb dating
Uranium, thorium and lead isotopes were analysed by laser ablation sector field inductively coupled plasma mass spectrometry (LA-SF-ICP-MS) using a Thermo-Finnigan Element 2 SF-ICP-MS system coupled to a New Wave Research UP-213 laser system with a teardrop low-volume cell at GUF following Gerdes & Zeh (2006) and Frei & Gerdes (2009. Laser spot-sizes were 30 ìm with a typical penetration depth of c. 15-20 ìm. A common-Pb correction based on the interference-and background-corrected 204 Pb signal and model Pb composition (Stacey & Kramers 1975) was carried out where necessary. The necessity of the correction was usually based on the 206 Pb/ 204 Pb value (,10 000). However, in case the interference-corrected 204 Pb could not be precisely detected (e.g. ,20 counts per second (c.p.s.)), this was applied only when the corrected 207 Pb/ 206 Pb was outside the internal errors of the measured ratios and yielded more concordant results (about 1-2% better concordance). The interference of 204 Hg (mean ¼ 255 AE 17 c.p.s.) on the mass 204 was estimated using a 204 Hg/ 202 Hg value of 0.2299 and the measured 202 Hg. All data were normalized relative to the GJ-1 reference. The total offset of the measured drift-corrected 206 Pb/ 238 U ratio from the 'true' ID-TIMS value of the analysed GJ-1 grain varied between 2 and 25% during the complete analytical session, but was much less during most sequences (protocols), which comprised 68 measurements: 13 standards and 55 unknowns. The drift during each sequence was corrected using the standard zircon GJ-1, which was measured together with the unknowns (11 unknowns followed by two standards). Reported uncertainties (2ó) of 206 Pb/ 238 U were propagated by quadratic addition of the external reproducibility (2SD) obtained from the standard zircon GJ-1 (n ¼ 13; 2SD c. 1.2%) during each analytical sequence, and the within-run precision of each analysis (2 standard error (SE)). For 207 Pb/ 206 Pb we used a 207 Pb-signal dependent uncertainty propagation ). For the Neoproterozoic GJ-1 zircon the reproducibility was 0.7 and 1.3% (2SD) for 207 Pb/ 206 Pb and 206 Pb/ 238 U, respectively. Reference zircon Plešovice was analysed to check quality and accuracy of the obtained data. Fifteen analyses yield a concordia age of 337.7 AE 1.4 Ma (MSWD CþE ¼ 0.63), which is in perfect agreement with published data (Slama et al. 2008). Concordia diagrams (2ó error ellipses), concordia ages and upper intercept ages (95% confidence level) were calculated using Isoplot/Ex 2.49 (Ludwig 2001).

LA-MC-ICP-MS Lu-Hf isotope analyses
Hafnium isotope measurements were made with a Thermo-Finnigan Neptune multicollector (MC)-ICP-MS system at GUF coupled to a New Wave Research UP-213 laser system with a  teardrop-shaped, low-volume laser cell following Gerdes & Zeh (2006. Multiple analyses of Lu-and Yb-doped JMC 475 solutions show that results with a similar precision and accuracy can be achieved also if Yb/Hf and Lu/Hf are 5-10 times higher than in most magmatic zircons. All data were adjusted relative to the JMC 475 176 Hf/ 177 Hf ratio of 0.282160 and quoted uncertainties are quadratic additions of the within-run precision and the reproducibility of the 40 ppb JMC 475 solution (2SD ¼ 0.0033%, n ¼ 5 during the analytical session). The correctness of the applied adjustment is reflected by the data for the standard zircon GJ-1 (0.282011 AE 0.000032 2SD, n ¼ 21), which were obtained during the same session as the sample zircon data. For calculation of åHf t the chondritic uniform reservoir (CHUR) was used as recommended by Bouvier et al. (2008;176 Lu/ 177 Hf ¼ 0.0336 and 176 Hf/ 177 Hf ¼ 0.282785), and a decay constant of 1.867 3 10 À11 (average of Scherer et al. 2001;Söderlund et al. 2004). Initial 176 Hf/ 177 Hf t and åHf t for all analysed zircon domains were calculated using the apparent Pb-Pb ages obtained for the respective domains, and for all cogenetic zircon domains by using the intrusion ages of the respective granitoids. Mean values for 176 Hf/ 177 Hf t and åHf t (and related errors (AE2ó)) for the respective granitoids are summarized in Table 2. Depleted mantle hafnium model ages (T DM ) were calculated using values for the depleted mantle as suggested by Blichert-Toft & Puchtel (2010), with 176 Hf/ 177 Hf ¼ 0.283294 and 176 Lu/ 177 Hf ¼ 0.03933, corresponding to a straight DM evolution line with åHf today ¼ +18 and åHf 4:558Ga ¼ 0.0. T DM ages for all data were calculated by using the measured 176 Lu/ 177 Hf of each spot for the time since zircon crystallization, and a mean 176 Lu/ 177 Hf of 0.0113 for the Palaeoproterozoic-Archaean crust (mean of average continental crust as suggested by Taylor & McLennan (1985) and Wedepohl (1995)). In addition, T DM ages were calculated using 176 Lu/ 177 Hf ¼ 0.02 as recommended for Hadean-Archaean mafic (protocrust) crust , in combination with depleted mantle parameters of Blichert-Toft & Puchtel (2010). These model ages are on average between 0.07 and 0.38 Ga older than those calculated with 176 Lu/ 177 Hf ¼ 0.0113. Because the existence of a depleted mantle reservoir during the Hadean-Early Archaean is highly speculative (see Hawkesworth et al. 2010;Kemp et al. 2010; and discussion below), we additionally calculated T CHUR model ages for the oldest zircons from Swaziland, using the CHUR parameters of Bouvier et al. (2008) in combination with 176 Lu/ 177 Hf ¼ 0.02 for mafic crust. The T CHUR ages are on average 0.3 Ga younger than the corresponding T DM ages. It must be noted, however, that there are many other options to obtain model ages. For example, a lower 176 Lu/ 177 Hf of 0.007, which is typical for present-day upper crust (Taylor & McLennan 1985;Wedepohl 1995), would decrease the T DM of our zircon analyses by about 0.01-0.2 Ga, relative to 176 Lu/ 177 Hf ¼ 0.0113, whereas a higher 176 Lu/ 177 Hf of 0.015, as proposed by Rudnick & Gao (2003) for the bulk continental crust, would increase the  (1993) T DM by 0.12 Ga (on average). Additional uncertainties with similar influence on the calculated T DM result from the 176 Lu/ 177 Hf heterogeneity of depleted mantle melts (e.g. Chauvel & Blichert-Toft 2001). The application of hafnium model ages to Archaean-Hadean rocks is discussed below in the section 'Crustal evolution'.

Results and interpretations
For data interpretation we use the combined set of CL images, and U-Pb and Lu-Hf isotope analyses (Figs 2-4; Table 2). The CL images of zircons from most granitoids revealed typical oscillatory magmatic zoning and monophase growth patterns (not shown), except for zircons of the TTG gneisses from the Piggs Peak inlier (samples AG7 and AG6a-c) and from the Mankyane area (sample DA16). These commonly show core-rim relationships (Fig. 2 Table 2). In contrast to the 'monophase' zircons, the zoned zircons as observed in samples AG7, AG6a-c and DA16 yielded significantly different concordant U-Pb ages and 176 Hf/ 177 Hf for their cores and rims (Figs 2 and 3), far outside the external reproducibility of the method (equal to the case 4 scenario of Zeh et al. 2009). In situ U-Pb analyses of zircon cores from the TTG gneiss sample AG7, taken near the southern boundary of the Phophonyane shear zone (Fig. 1), gave a concordant U-Pb age of 3.644 AE 0.007 Ga, and a few rims gave a younger upper intercept U-Pb age of 3.191 AE 0.009 Ga. Similar ages were also obtained from samples AG6a-c, which were taken from an outcrop at the road near 'the falls', where dark banded tonalitic gneisses are transected by dykes and bodies of granodiorite composition.
Zircons from the dark banded tonalite gneiss (sample AG6c) yielded three ages at 3.662 AE 0.017 Ga (n ¼ 10), 3.219 AE 0.013 Ga (n ¼ 2), and 3.127 AE 0.009 Ga (n ¼ 7). The oldest age was obtained from zircon cores, and the two younger ages from zircon rims, as well as from needle-like zircons (Fig.  2). In contrast, zircons from a granodiorite dyke (sample AG6a) and body (sample AG6b), which cross-cut the dark gneiss, yielded concordant ages of 3.221 AE 0.012 Ga and 3.230 AE 0.009 Ga, respectively, and only a few zircon cores gave older ages of 3.52-3.56 Ga (Table 2).

Age and hafnium isotope data
The presented CL images and U-Pb-Lu-Hf isotope data indicate that the igneous protoliths of the oldest gneisses of the Ancient Gneiss Complex, which occur in the Piggs Peak inlier, were emplaced at 3.66-3.64 Ga, and were affected by two subsequent events at 3.22 and 3.13 Ga. The two older ages are in agreement (within error) with those previously obtained by highprecision U-Pb single-grain dating on zircons from other banded tonalitic gneisses samples (3.663 Ga) and related granite dykes (3.223 Ga) (Schoene & Bowring 2007;Schoene et al. 2008), and by U-Pb SHRIMP dating of zircon cores (3.644 Ga : Compston & Kröner 1988;Kröner et al. 1989). In contrast to our dataset (and to Schoene & Bowring 2007), Compston & Kröner (1988) and Kröner et al. (1989) reported a large number of additional U-Pb zircon SHRIMP ages of 3.58, 3.50, 3.43, 3.20, 3.0 and 2.99-2.87 Ga, which they interpreted to reflect different postintrusive thermal-metamorphic-tectonic events. However, the meaning of the different ages in a geological context is still unclear. An unambiguous interpretation of these data is hampered by the lack of CL or SEM images for these zircons, by very complex Pb loss patterns, and by the negative concordance of many analyses. It is likely that the 3.58-3.  (Fig. 1, Table 2). In contrast, zircons from the contemporaneous Kaap Valley and Nelshoogte plutons, which are exposed north of the Saddleback-Inyoka fault system, yielded significantly more radiogenic 176 Hf/ 177 Hf of 0.28076 AE 0.00003 ). In fact, the Lu-Hf isotope data of this study support previous conclusions that the Saddleback-Inyoka fault system (or the central part of the Barberton Greenstone Belt) represents an important terrane boundary, which separates a southern, less radiogenic terrane (Barberton South) from a northern, more radiogenic terrane (Barberton North). This separation is reflected not only by the Lu-Hf isotope data , this study; Fig. 4f), but also by Sm-Nd isotope data (Schoene et al. 2009). Schoene et al. (2009) interpreted this isotope difference (together with many other field observations and time constraints) to result from magmatism above a doubly vergent subduction zone at c. 3.2 Ga, whereby the slab-induced magmas formed underneath the southern terrane (comprising the Ancient Gneiss Complex, and Stolzburg and Steyndorp terranes of Swaziland) assimilated higher amounts of older crust than the magmas formed underneath the northern terrane. Assimilation of pre-3.23 Ga gneisses during magmatism at 3.23 Ga in the southern terrane is well reflected by zircon xenocrysts with ages between 3.65 and 3.43 Ga (and with lower radiogenic 176 Hf/ 177 Hf) in rocks from the Piggs Peak inlier, and in the area west of Mankyane (samples AG6a, AG6b and DA16; this study) and around Manzini (samples AGC1 and AGC2; Zeh et al. 2009).
The 3.13 Ga event detected in the zircons from sample AG6c from the Piggs Peak inlier correlates well with emplacement ages obtained from the potassium-rich granite batholiths, which form voluminous, sheet-like bodies in Swaziland and around the Barberton greenstone belt, comprising the Piggs Peak batholith (3.10 Ga), Mpuluzi batholith (3.08 Ga) and Nelspruit batholith (3.11 Ga), and the Boesmanskop syenite (3.10 Ga) (age data from Zeh et al. 2009). In further agreement, the 3.13 Ga magmatic zircons from sample AG6c have, within error, identical 176 Hf/ 177 Hf (0.28077 AE 0.00004) to the zircons from the batholiths (0.28075 AE 0.00003 to 0.28078 AE 0.00004; data from Zeh et al. 2009). This relationship hints that the 3.66 Ga banded tonalitic gneisses of the Piggs Peak inlier (sample AG6c) were infiltrated by younger magma, which fed the nearby Piggs Peak batholiths at 3.10 Ga (Fig. 1), and caused the crystallization of new zircon of this age as discrete crystals and rims around older cores. It is worth noting, however, that clear evidence for infiltration has not been observed in sample AG6c, although the occurrence of tiny, banding-parallel younger layers cannot be excluded.
Zircons from the Sinceni granodiorite yielded a U-Pb zircon age of 3.067 AE 0.012 Ga, which is, within error, identical to the age obtained by Maphalala & Kröner (1993 Maphalala & Kröner (1993), but significantly older than the age of 2.691 AE 0.002 Ga reported by Layer et al. (1989) for the Mbabane granite. Furthermore, it should be noted that the Mpageni granite, which intruded the northern part of the Barberton Mountain Land (Fig. 1)

Crustal evolution
The new datasets reveal that the protoliths of the oldest TTG gneisses of Swaziland were emplaced at 3.66 Ga by reworking of an even older crust, as is reflected by the subchondritic åHf 3:66Ga of the oldest zircons. However, putting an exact age on this older crust is difficult and involves many uncertainties, which amongst others result from the choice of the parameters used to calculate appropriate hafnium model ages (see section 'LA-MC-ICP-MS Lu-Hf isotope analyses'), and especially from the problem that the Hf isotope evolution of the Earth's mantle during the Hadean-Early Archaean is not well constrained (see the discussions by Valley et al. 2006;Zeh et al. 2008Zeh et al. , 2009Hoffmann et al. 2010;Kemp et al. 2010). In fact, it is still a matter of debate whether or not a voluminous depleted mantle reservoir, which formed in response to important continental crust extraction, already existed during the Hadean to Early Archaean. Although the existence of a depleted reservoir (with respect to hafnium isotopes) was suggested on the basis of zircon solution Hf isotope data from Greenland and the Jack Hills (Vervoort et al. 1996;Vervoort & Blichert-Toft 1999;Blichert-Toft & Albarède 2008), and by the early zircon laser ablation data from the Jack Hills (Harrison et al. 2005), its existence could not be verified by more recent in situ laser ablation Lu-Hf-U-Pb zircon isotope studies (see Fig. 5a), which have revealed only chondritic to subchondritic åHf t (Harrison et al. 2008;Kemp et al. 2009Kemp et al. , 2010. Kemp et al. (2009Kemp et al. ( , 2010 argued that the systematically higher radiogenic hafnium values obtained by the zircon solution Hf data from Greenland and the Jack Hills result from younger zircon overgrowths that could not be removed prior to dissolution, and that the large laser spots used by Harrison et al. (2005) sampled both cores and rims of the complexly zoned Jack Hills zircons. Taking this into account, the most reliable (trustworthy) zircon dataset available at present, as compiled by Kemp et al. (2010), provides no evidence for the existence of a depleted (global-scale) mantle reservoir during the Hadean-Eoarchaean. In contrast, it points to the existence of a volumetrically insignificant 'KREEP'-like protocrust (KREEP ¼ high potassium, REE and phosphorus) that was continuously reworked by remelting over nearly 400 Ma (without addition of new depleted mantle material), and finally disappeared with the onset of new, juvenile crust formation during the Archaean at ,4.0 Ga (also see Kamber et al. 2003Kamber et al. , 2005Zeh et al. 2008). Based on the most reliable zircon datasets, Kemp et al. (2009Kemp et al. ( , 2010 also argued that most of the juvenile Early Archaean crust was extracted directly from a chondritic mantle reservoir rather than from a depleted mantle source. This final interpretation is, apart from the most reliable zircon datasets , also supported by solution Hf isotope data obtained from 3.72-3.75 Ga Isua pillow basalts and metasediments (Polat et al. 2003;Hoffmann et al. 2010), which all scatter around CHUR ( Fig. 5a  and b). Nevertheless, highly positive åHf values up to +13.5 obtained from pristine 3.75 Ga Isua boninites (Hoffmann et al. 2010) provide evidence that the Early Archaean mantle was heterogeneous, and contained spatially restricted, highly depleted (long-term isolated) reservoirs (Fig. 5b). Similar mantle reservoirs have also been suggested to be the source for some Barberton komatiites (åHf 3:45Ga ¼ +2.3 to +7.5), which extruded contemporaneously with much less depleted tholeiitic basalts (åHf 3:45Ga ¼ +2.3 to À0.5) in the Barberton Greenstone Belt, to the north of Swaziland (Blichert-Toft & Arndt 1999).
T CHUR model ages of 3.77 AE 0.18 Ga obtained from the oldest magmatic zircons from Swaziland during this study (sample AG6, AG7) are in agreement with the conclusions of Kemp et al. (2010); thereafter abundant mafic crust (with 176 Lu/ 177 Hf ¼ 0.02) was directly extracted from a CHUR mantle during the Early Archaean. In fact, the 3.66 Ga Swaziland TTGs plot on the same crust evolution trend as meta-igneous, granitic and metasedimentary rocks from Greenland and the Jack Hills (see Fig. 5b).
Nevertheless, despite this coincidence, it cannot be ruled out completely that a (global-scale) depleted mantle source already existed during the Late Hadean-Eoarchaean, although there is no clear evidence for it so far. Taking this into account, the T DM model age of 4.08 AE 0.18 Ma obtained from the oldest zircons (Fig. 5b) might be considered as a maximum age for the protolith of the Swaziland TTGs. In addition, there is the possibility that the 3.66 Ga TTGs resulted from reworking of an Eoarchaean or even Hadean TTG crust, which was admixed with juvenile, mantle-derived magmas at 3.66 Ga. This final option, however, seems to be less likely in the light of the latest Hf isotope data of Kemp et al. (2009Kemp et al. ( , 2010, which refute the existence of a Hadean TTG crust, as was originally suggested by Harrison et al. (2008) and Blichert-Toft & Albarède (2008). None the less, the existence of a (global) depleted mantle reservoir at 3.66 Ga is very likely, and is in fact required by the zircon Hf isotope data from Greenland, the Jack Hills and the 3.54-3.32 Ga granitoids of Swaziland (see Fig. 5b), as well as by the Hf isotope solution data for Archaean mafic rock, comprising those from the 3.45 Ga Barberton greenstone belt (tholeiitic basalts + some komatiites; Blichert-Toft & Arndt 1999), the 3.75 Ga Isua basalts (Polat et al. 2003), the 2.82 Ga Kostomuksha komatiites (Blichert-Toft & Puchtel 2010), and the 2.7 Ga Superior province basalts (Polat & Münker 2004). Most of these data (excluding the abnormally depleted rocks from Greenland and Barberton) can be forced to fit a straight depleted mantle array between 4.0 Ga and today using a present-day 177 Lu/ 176 Hf of 0.04017, and an average present-day 177 Hf/ 176 Hf of 0.283294 Ma (Vervoort & Blichert-Toft 1999) (see Fig. 5b). This line is only slightly steeper than that proposed by Pietranik et al. (2009) based on the compilation of zircon hafnium isotope data.
Formation of abundant (juvenile) crust in the Swaziland terrane during the Palaeoarchaean is well reflected by super-  Zeh et al. 2007Zeh et al. , 2009Zeh et al. , 2010aMillonig et al. 2010Millonig et al. (Z'2007Millonig et al. , 2009Millonig et al. , 2010Millonig et al. and M'2010), combined with zircon data of Kemp et al. (2009Kemp et al. ( , 2010Kemp et al. ( ) (K'2009Kemp et al. ( , 2010. In addition, the diagram shows the composition of Archaean mafic and ultramafic rocks (B&A'1999, Blichert-Toft & Arndt 1999P'2003P' , Polat et al. 2003P'2004, Polat & Münker 2004B&P'2010, Blichert-Toft & Puchtel 2010, and of Isua boninites (H'2010(H' , Hoffmann et al. 2010. SCT, Swaziland crustal trend with 176 Lu/ 177 Hf ¼ 0.0113; MC, mafic crust with 176 Lu/ 177 Hf ¼ 0.02; T CHUR and T DM define the respective model ages (average) obtained from the oldest Swaziland zircons. DM, depleted mantle evolution using the parameters of Blichert-Toft & Puchtel (2010); DM1, evolution of the depleted mantle between 4.0 Ga and today. chondritic åHf t of +0.5 to +2.2 obtained from nearly all TTGs that were emplaced between 3.54 and 3.32 Ga (Fig. 3). At present, the reason for this c. 200 Ma period of juvenile granitoid magmatism is unclear. One explanation could be that several juvenile island arcs were successively accreted between 3.54 and 3.32 Ga, although there is little structural control that supports this scenario. Alternatively, the same pattern could also be achieved by periodic slab melting during more or less steady subduction of hydrated oceanic crust beneath a relatively stationary proto-craton. However, because of the lack of additional geochemical data, this scenario remains speculative too. The formation of new continental crust (by extraction from a depleted mantle source) obviously continued until 3.23 Ga, as can be concluded from åHf t of À0.3 to À0.5 (AE1.5) obtained from the granitoids of the Piggs Peak inlier, as well as from those of the Manzini and Mankyane areas. The lower åHf t values of these granitoids, compared with the 3.54-3.32 Ga TTGs, could be explained by the assimilation of a significant amount of older crust at 3.23 Ga. This conclusion is well supported by zircon xenocrysts found in many of the 3.23 Ga granite gneisses ( Table  2, and data of Zeh et al. 2009). Schoene et al. (2009) suggested that the 3.23 Ga magmatism in Swaziland was the result of a southward-directed subduction of an oceanic crust, which vanished by c. 3.2 Ga as a result of the amalgamation of a southern terrane (comprising the Swaziland + Stolzburg + Steyndorp terranes ¼ Barberton South of Zeh et al. 2009) with a northern terrane (comprising the northern part of the Barberton greenstone belt ¼ Barberton North of Zeh et al. 2009).
Following the 3.23-3.22 Ga collision event, the eastern part of the Kaapvaal craton was affected by voluminous magmatism at 3.1 Ga, which caused the formation of numerous potassium-rich, sheet-like granite batholiths with slightly subchondritic åHf t of À0.1 AE 1.4 to À1.7 AE 2.2 (this study; Zeh et al. 2009). Zeh et al. (2009) suggested that these batholiths formed by the reworking of older crust, with the addition of minor components from a depleted mantle source. It is likely that the magmas of the batholiths originate from lower crustal melting at 3.1 Ga, triggered by incubational heating owing to K-U-Th decay, in combination with rift-or transtension-related magmatic underplating (see Schoene et al. 2009). Crustal recycling is also in agreement with the data from the 3.07 Ga Sinceni granodiorite with åHf t ¼ À2.8 AE 2.0, and in particular with those from the 2.73 Ga potassium-rich granites (Mbabane, Sincunusa and Ngwempisi), which show highly subchondritic åHf t between À6.6 and À7.2.
In general, the combined U-Pb-Lu-Hf datasets from the Swaziland granites define two arrays, an older array with increasing åHf t from À1.6 to +2.2 between 3.66 and 3.32 Ga, and a younger array with a relatively steady decrease of åHf t from +2.2 to À7.2 between 3.32 and 2.73 Ga (Fig. 3). The younger array can be explained by several scenarios, with the most simple forcing all rocks with ages ,3.32 Ga on a single crust evolution trend with 176 Lu/ 177 Hf ¼ 0.0113 (see Figs 3 and 5b), a value that is typical for present-day continental crust (Taylor & McLennan 1985;Wedepohl 1995). However, the oldest rocks with ages between 3.32 and 3.07 Ga also fit on an evolution trend with 176 Lu/ 177 Hf ¼ 0.005, a value that is typical for Archaean TTGs (Kamber et al. 2002;Condie 2005; Fig. 3b). Alternatively to a simple crust evolution model (forcing all granites on one trend), the obtained datasets might also be explained by granitoid formation from different sources, with the 3.32-3.07 Ga granitoids formed by reworking of a TTG crust (derived from a depleted mantle source at c. 3.32 Ga), and the 2.7 Ga granites by melting of an older mafic crust, the same from which the 3.66 Ga TTGs were extracted (see Fig. 3b). This final scenario, however, seems to be less likely, as melting of hydrated mafic crust would inevitably lead to the formation of TTGs and not to potassium-rich granites. Furthermore, it appears highly unlikely that an old mafic crust survived in a pristine state in the deeper part of the Swaziland crust, which was affected by several phases of juvenile magma addition between 3.54 and 3.32 Ga. Formation of the 2.7 Ga granites by direct remelting of the 3.66 Ga TTGs is unlikely as well, as the final granitoids are volumetrically unimportant (restricted to the Piggs Peak inlier; Fig. 1), and would provide melts with a very low åHf 2:7Ga of about À20 (Fig.  3b). Taking this into account, it seems to be most likely that the 2.7 Ga granites predominantly formed by reworking of the 3.32 Ga TTGs, although a substantial contribution from the juvenile 3.43 Ga TTGs cannot completely be excluded, considering that the final granitoids are abundant in Swaziland and the adjacent Barberton Greenstone Belt (Fig. 1), and that the 3.23 Ga granites contain 3.43-3.53 Ga zircon xenocrysts (this study and Zeh et al. 2009). However, if the 2.7 Ga granites evolved from the ,3.32 Ga TTGs, there must have been a mechanism that caused an increase of the 176 Lu/ 177 Hf from 0.005 (as typical for TTG crust) to 0.0113 (or even higher) (see Fig. 3b). One possibility to explain this increase is that the primary TTG crust of the Ancient Gneiss Complex was subject to zircon fractionation during post-3.32 Ga melting processes, whereby zircon with less radiogenic hafnium was retained in the restite lower crust during the extraction of the granite at 3.07 and 2.7 Ga. However, minor addition of juvenile melts during magmatism between 3.32 and 2.7 Ga would lead to the same patterns. Furthermore, a combination of the two processes is also possible. To discriminate between the various scenarios additional geochemical data are required.

Summary and conclusions
Combined U-Pb and Lu-Hf isotope analyses of zircons reveal that the oldest TTGs of Swaziland, which were emplaced at 3.66 Ga, have subchondritic åHf t of À1.6 AE 2.0. These data indicate that these TTGs formed by reworking of an even older crust, which most probably was derived from an Early Archaean chondritic uniform mantle reservoir at about 3.8 Ga, immediately prior to TTG formation, although an older crustal source of Hadean age cannot completely be excluded, but would require complex mixing models.
During the following 200 Ma, new juvenile crust was added to the Ancient Gneiss Complex, as is well reflected by superchondritic åHf t of +2.2 to +0.5 obtained from TTG gneisses with ages between 3.54 and 3.32 Ga. The mechanism that formed this new crust is not well understood. It could be explained either by successive accretion of primitive island arcs or by successive subduction and periodic slab melting beneath a relatively stationary proto-craton.
The new datasets presented here additionally reveal that during subsequent magmatic processes the Palaeoarchaean crust formed between 3.54 and 3.32 Ga was intensely reworked, as is reflected by a linear array of decreasing åHf t from +2.2 to À7.2 between 3.32 and 2.72 Ga. This reworking process started with the amalgamation of Swaziland's Ancient Gneiss Complex with terranes of the Barberton Greenstone Belt at 3.23 Ga, and continued during subsequent intracratonic melting processes at 3.1, 3.06 and 2.72 Ga. The derived array with an average 176 Lu/ 177 Hf of 0.0113 can be explained either by zircon fractionation during lower TTG(?) crust melting or by minor addition of mantle-derived melts during the granite formation events, or by a combination of both.  Edited by M. Sosson, N. Kaymakci, R. A. Stephenson, F. Bergerat and V. Starostenk This wide area of the Alpine-Himalayan belt evolved through a series of tectonic events related to the opening and closure of the Tethys Ocean. In doing so it produced the largest mountain belt of the world, which extends from the Atlantic to the Pacific oceans. The basins associated with this belt contain invaluable information related to mountain building processes and are the locus of rich hydrocarbon accumulations. However, knowledge about the geological evolution of the region is limited compared to what they offer. This has been mainly due to the difficulty and inaccessibility of cross-country studies. This Special Publication is dedicated to the part of the Alpine-Himalayan belt running from Bulgaria to Armenia, and from Ukraine to the Arabian Platform. It includes twenty multidisciplinary studies covering topics in structural geology/tectonics; geophysics; geochemistry; palaeontology; petrography; sedimentology; stratigraphy; and subsidence and lithospheric modelling. This volume reports results obtained during the MEBE (Middle East Basin Evolution) Programme and related projects in the circum Black Sea and peri-Arabian regions. Reservoir compartmentalization, the segregation of a petroleum accumulation into a number of individual fluid/pressure compartments, controls the volume of moveable oil or gas that might be connected to any given well drilled in a field, and consequently impacts on reserves 'booking' and operational profitability. This is a general feature of modern exploration and production portfolios, and has driven major developments in geoscience, engineering and related technology. Given that compartmentalization is a consequence of many factors, an integrated subsurface approach is required to better understand and predict compartmentalization behaviour, and to minimize the risk of it occurring unexpectedly. This volume reviews our current understanding and ability to model compartmentalization. It highlights the necessity for effective specialist discipline integration, and the value of learning from operational experience in: detection and monitoring of compartmentalization; stratigraphic and mixed-mode compartmentalization; and fault-dominated compartmentalization.

Edited by G. P. Goffey, J. Craig, T. Needham and R. Scott
Onshore fold-thrust belts are commonly perceived as 'difficult' places to explore for hydrocarbons and are therefore often avoided. However, these belts host large oil and gas fields and so these barriers to effective exploration mean that substantial unexploited resources may remain. Over time, evaluation techniques have improved. It is possible in certain circumstances to achieve good 3D seismic data. Structural restoration techniques have moved into the 3D domain and increasingly sophisticated palaeo-thermal indicators allow better modelling of burial and uplift evolution of source and reservoirs. Awareness of the influence of pre-thrust structure and stratigraphy and of hybrid thick and thin-skinned deformation styles is augmenting the simplistic geometric models employed in earlier exploration. But progress is a slow, expensive and iterative process. Industry and academia need to collaborate in order to develop and continually improve the necessary understanding of subsurface geometries, reservoir and charge evolution and timing; this publication offers papers on specific techniques, outcrop and field case studies.

• Special Publication 355
The SE Asian Gateway: History and Tectonics of the Australia-Asia collision Edited by R. Hall, M. Cottam and M. E. J. Wilson Collision between Australia and SE Asia began in the Early Miocene and reduced the former wide ocean between them to a complex passage which connects the Pacific and Indian Oceans. Today, the Indonesian Throughflow passes through this gateway and plays an important role in global thermohaline flow, and the region around it contains the maximum global diversity for many marine and terrestrial organisms. Reconstruction of this geologically complex region is essential for understanding its role in oceanic and atmospheric circulation, climate impacts, and the origin of its biodiversity.
The papers in this volume discuss the Palaeozoic to Cenozoic geological background to Australia and SE Asia collision, and provide the background for accounts of the modern Indonesian Throughflow, oceanographic changes since the Neogene, and aspects of the region's climate history.

From the Geological Society Publishing House
For full details see the Online Bookshop: www.geolsoc.org.uk/bookshop