The oldest crust in the Ukrainian Shield – Eoarchaean U–Pb ages and Hf–Nd constraints from enderbites and metasediments

Abstract The oldest crust in the Ukrainian Shield occurs in the Podolian and Azov domains, which both include Eoarchaean components. U–Pb age data for Dniestr–Bug enderbites, Podolian Domain, indicate that these are c. 3.75 Ga old, and Lu–Hf isotope data indicate extraction from chondritic to mildly isotopically depleted sources with ɛHf up to c. +2. Nd model ages support their Eoarchaean age, while model ages for Dniestr–Bug metasedimentary gneisses indicate that these also include younger crustal material. Most of the Hf-age data for metasedimentary zircon from the Soroki greenstone belt, Azov Domain, reflects Eoarchaean primary crustal sources with chondritic to mildly depleted Hf isotope signatures at 3.75 Ga. A minor portion is derived from Mesoarchaean crust with a depleted ɛHf signature of c. +4 at 3.1 Ga. U–Pb zircon ages from Fedorivka greenstone belt metasediments are consistent with the Soroki age data, but also include a 2.7–2.9 Ga component. Nd whole rock model ages provide support for a younger crustal component in the latter. Both domains have been subject to Neoarchaean, c. 2.8 Ga, and Palaeoproterozoic, c. 2.0 Ga, metamorphism. The spatial distribution indicates that the Podolian and Azov domains evolved independently of each other before the amalgamation of the Ukrainian Shield.

The nature of continental crust formation processes, the rates of crust formation and destruction and the extent of continental crust in the early Archaean are matters of much debate. While there is a considerable consensus that modern-type Wilson cycle processes have been in operation since c. 3 Ga, it is not clear when such processes became dominant and when a global depleted mantle reservoir was established. Modern-type subduction processes have been suggested to have been in operation already in the earliest Archaean (e.g. Harrison et al. 2005Harrison et al. , 2008Nutman et al. 2009;Hiess et al. 2009), but a number of lines of evidence suggest that early Archaean crust primarily formed by other processes (e.g. Shirey & Richardson 2011;Dhuime et al. 2012) and that at least regionally there was a mid-Archaean transitional period between an ancient crustal evolutionary regime and the onset of modern-type plate tectonics (Naeraa et al. 2012).
Key evidence about both ancient and modern crustal formation processes is provided by isotopic age determinations combined with Nd, Hf, Pb and O isotopic data from whole rock samples, and the Hf and O isotope compositions of dated zircon crystals (e.g. Hawkesworth & Kemp 2006). Old Archaean crust is commonly reworked, and the characterization of its primary age and composition is often difficult. In such rocks, where deformation and metamorphism may have modified the U -Pb systems in zircon and the original isotope record in whole rock samples, zircon crystals commonly retain a robust memory of their original Hf and O isotopic compositions, even when the U -Pb system in the crystals has been disturbed.
The East European Craton, with its three segments Sarmatia, Volgo -Uralia and Fennoscandia (Bogdanova 1993), comprises a significant fraction of the Archaean crust exposed at the Earth, including Palaeo-and Eoarchaean components (Fig. 1). Much of this crust has been strongly reworked during the Palaeoproterozoic. While the Archaean in Fennoscandia has been documented in some detail (e.g. Slabunov et al. 2006;Lauri et al. 2011), the Archaean of Sarmatia and Volgo-Uralia is less well known. The Ukrainian Shield, a central part of Sarmatia, includes Eoarchaean components as well as younger Archaean and Proterozoic rocks (Claesson et al. 2006;Bibikova et al. 2010). In this contribution we present U -Pb and Hf zircon results and Sm -Nd whole rock data from enderbites and metasedimentary rocks, which provide new information about the origin and complex Archaean and Palaeoproterozoic evolution of this old segment of crust. We demonstrate that two separate domains in the Ukrainian Shield include significant components of rocks and metasedimentary detrital minerals which are up to 3.75 Ga in age. There is no evidence in these new results of more than mildly depleted mantle reservoirs 3.75 Ga ago, but 3.1 Ga rocks were derived from clearly depleted mantle sources.

The Ukrainian Shield
The Ukrainian Shield is commonly described as several blocks, or domains, separated by suture  zones (Fig. 1). The Archaean high-grade Azov Domain in the east and the Podolian Domain in the SW were strongly reworked in the Palaeoproterozoic. In contrast, the Meso-Neoarchaean granitegreenstone Middle Dniepr Domain, in the central part of the Shield, was virtually untouched by Palaeoproterozoic orogenic processes. Palaeoproterozoic rocks compose most of the Kirovograd (Ingul) Domain in the central part of the Shield, and also the Ros-Tikich and Volyn domains in its northwestern part. The Orekhiv-Pavlograd, Krivyy Rih and Golovanivsk suture zones (Fig. 1) which separate main domains have complex tectonic fabrics including strong shearing of rocks from the adjacent domains on both sides of the suture zones. Tectonically the structure of the Ukrainian basement can be described as a collage of Archaean and Palaeoproterozoic terranes (e.g. Kalyaev 1976;Glevassky & Glevasska 2002) which have been amalgamated around Palaeoarchaean cores at different times, both in the Archaean and in the Palaeoproterozoic.
The oldest components of the Ukrainian Shield appear in the Azov and Podolian domains (Figs 1 & 2). The oldest identified rocks in the east belong to the Novopavlovka complex, which appears within the Orekhiv-Pavlograd Suture Zone, separating the Azov and Middle Dniepr domains. These have been discovered by deep drilling and consist mainly of ultramafic rocks and tonalites, metamorphosed to amphibolite and granulite facies. The tonalites have been dated at 3.65-3.6 Ga by Sm-Nd isochron dating (Bibikova & Baadsgaard 1986), and both conventional (Scherbak et al. 1984) and ion-microprobe (Bibikova & Williams 1990) U -Th -Pb dating of zircon. A second phase of tonalite emplacement has been dated at 3.4 Ga (Bibikova & Williams 1990). Bibikova et al. (2010) reported ages of 3.5-3.6 Ga, and zircon cores older than 3.7 Ga for metasedimentary zircon from the Soroki greenstone belt, Azov domain. From the Podolian Domain, U -Pb dating of zircon from granulite-facies granitoids (enderbites) from the Dniestr-Bug Series has yielded ages up to 3.65 Ga, and indications of material up to 3.75 Ga old (Claesson et al. 2006).

The Podolian Domain
The Podolian Domain is composed of high-grade Archaean and Palaeoproterozoic igneous and supracrustal rocks. The oldest Palaeoarchaean rocks have been found in outcrops along the banks of the Pivdenny (South) Bug River and in nearby open pit quarries (Fig. 2a). The crust in this region, which is strongly deformed and generally metamorphosed in granulite facies, has been divided into the older Dniestr-Bug Series, and the younger Bug Series which is inferred to be of late Archaean age (Yesypchuk et al. 2003;Stepanyuk et al. 2004). A major component of the Dniestr-Bug Series is enderbitic gneisses, commonly referred to as enderbites. These are granulite-facies granitoids, mainly of tonalitic composition, typically composed of 3 -5% orthopyroxene, up to 10% clinopyroxene, 1 -5% biotite, 35 -60% plagioclase and 25-35% quartz. Minor minerals include apatite and zircon. The enderbites are typically intercalated with mafic rocks (two-pyroxene and amphibole-pyroxene schists). On weathered surfaces the strongly deformed gneissic nature of these rocks, with a persistent intense banding and ubiquitous tight folding on a centimetre-to metre-scale, is clearly visible. This strong deformation is less clearly visible on fresh surfaces, where the medium-to coarse-grained enderbitic gneisses typically are greenish-grey in colour and have a more massive homogeneous appearance. Enderbites and mafic schists may be genetically linked, but the relation between the two rock types is not interpreted to be migmatitic. The high-grade metamorphism of this tonalitemafite-ultramafite sequence makes the interpretation of the primary nature of the rocks as plutonic or volcanic rocks difficult, and the enderbite sequence has also been described as the Gayvoron intrusive complex (Kryvdik et al. 2011). The igneous origin of the enderbites is not contested.
The Dniester -Bug Series is traditionally subdivided into five strata, which in addition to the characteristic association of enderbites and mafic rocks also includes schists and calciphyres interpreted to have a supracrustal origin, and garnetbiotite bearing leucogneisses. The Bug Series is divided into two suites. The lower Kosharo-Oleksandrivka Suite includes quartzites, high-Al gneisses and schists which commonly are graphitebearing, while the upper Khashuvato -Zavallya Suite includes carbonate rocks (marbles and calciphyres) which are associated with graphite -biotite, garnet-biotite, biotite and pyroxene gneisses. Febearing quartzites are also present in places. Bug Series metasediments have been interpreted to have been deposited in depressions developed in the older Archaean basement.
For the present study samples of enderbite from the Dniester-Bug Series were collected in two open pit quarries, the Odesa quarry (sample 06-BG38) and the Kozachy Yahr quarry (sample C10-U4), located at opposite banks of the Pivdenny (South) Bug river (Fig. 2a).

The Azov Domain
The Azov Domain (Fig. 2b) is dominated by Archaean to Palaeoproterozoic, heavily metamorphosed supracrustal rocks and granitoids. Unlike the Middle Dniepr Domain to the west, which is dominated by Meso-and Neoarchaean Tonalite-Trondhjemite-Granodiorite (TTG) gneisses and greenstone belts, greenstone belts are not widespread in the Azov block but some structures identified as such have been described. Among these is the Soroki greenstone belt, situated in the SW part of the Azov domain (Bobrov et al. 2000), which is one of the targets of the present study. It is about 35 km long and up to 1.2 km wide. Volcano-sedimentary rocks, attributed to the Osipenkovo Series, form a symmetrical syncline with a metamorphic zonation from greenschist to epidote-amphibolite facies. The Osipenkovo Series is subdivided into two suites: the lower Olgino and the upper Krutobalka Suite. The total thickness of the Osipenkovo Series is about 500 m. Volcanic rocks prevail in the Olgino suite while meta-terrigeneous rocks dominate in the Krutobalka suite. The latter is composed of sandstones, conglomerates and high-Al schists with metamorphic biotite, muscovite, staurolite, sillimanite, tourmaline and garnet. Rare amphibolite bodies are also present. An Archaean age of the Osipenkovo Series was identified when granitoids of the Shevchenko complex were dated at 2.8 Ga (Artemenko 1997). These granitoids intrude the Olgino suite of the Osipenkovo Series. Zircon from garnet -biotite paragneisses underlying the mafic rocks of the Olgino suite has yielded an age of 3350 + 120 Ma (Artemenko 1997). Three samples of zircon separated from metasedimentary rocks of the Krutobalka suite are included in the present study. Two of them, 5/88, and CU-1, were collected in the Sobach'ya gully, whereas sample 92/218 was collected next to the Soroki village. The samples from the Sobach'ya gully were taken from a homogeneous outcrop (5 × 10 m) of mica schist. This is a grey, generally schistose rock with lepido-to granoblastic texture consisting of 15-20% biotite, 40-45% plagioclase (albite-oligoclase), 40% quartz and minor chlorite, apatite and zircon. Sample 92-218 is a paragneiss, consisting of 55-60% plagioclase (albite), 15% biotite, 5% muscovite, 20% quartz and some carbonate.
The Fedorivka structure is located in the northern part of the Azov domain (Fig. 2b). It occurs as a 20 km-long and 4 km-wide graben-syncline elongated in mainly north-south direction with some branching. It is hosted by metamorphic rocks of the Zakhidnopryazovska Series, and consists of a 2.25 km-thick sequence of biotite-muscovitegarnet-sillimanite bearing schists and gneisses, some with amphibole and pyroxene. It also includes metamorphic graphite-bearing and carbonate horizons (Bibikova et al. 2012). The supracrustal rocks of the Fedorivka syncline are cut by numerous pegmatite veins, in contrast to the surrounding basement rocks which have not been affected by migmatization. The age of the schist-carbonate sequence in the Fedorivka syncline is poorly constrained; a minimum age of deposition for Fedorivka metasediments is provided by the 2085 Ma age for the Anadol granites (Vasilchenko et al. 1992), while a tighter constraint is given by the recently reported age 2735 + 30 Ma for the likewise intrusive Yanvarsk granite (Isakov et al. 2010).
Sample 89/551 in the present study was taken from a borehole in the central part of the Fedorivka graben-syncline, located on the left side of the Mokrye Yaly River 1.9 km south of the Fedorivka settlement. It is a paragneiss composed of c. 83% plagioclase, 5% biotite, 5% quartz, 2% microcline, 2% muscovite and 2% carbonate. Minor minerals include apatite and zircon.

Analytical methods
Zircons were extracted from all samples using standard methods. Separated zircon crystals were mounted into a 25 mm epoxy puck along with the c. 1065 Ma Geostandard 91500 reference zircon (Wiedenbeck et al. 1995) and polished approximately half way through. Before isotope analysis, all zircon crystals were cathodoluminescence (CL)imaged to clarify internal structures and to identify portions suitable for analysis.
Following CL imaging the mount was coated with c. 30 nm of gold. Secondary ion mass spectrometer (SIMS) U -Th -Pb analyses were carried out using a large geometry Cameca IMS 1270 instrument at the Nordsim facility, Swedish Museum of Natural History, Stockholm, Sweden. The U-Th-Pb instrumental set up broadly follows that of Whitehouse & Kamber (2005). The following procedures were fully automated: (a) pre-sputtering with a 25 mm raster for 120 s; (b) centring of the secondary ion beam in the 4000 mm field aperture; (c) mass calibration optimization; and (d) optimization of the secondary beam energy distribution. Age interpretations were performed using the routines of Isoplot/Ex (Ludwig 2003). The reported discordance, a measure of the Pb loss, refers to the position of data points in relation to the concordia curve in conventional 207 Pb/ 235 U-206 Pb/ 238 U space. Decay constants used follow the recommendations of Steiger & Jäger (1977).
Hf isotope analyses were performed at Bristol University, UK. The data were acquired with a Thermo-Scientific Neptune multicollector inductively coupled plasma mass spectrometer (ICP-MS) coupled to a New Wave 193 nm ArF laser ablation sampling system operating at 4 Hz and using a 50 mm spot size over a 60 s ablation period. The Yb isotope compositions of Segal et al. (2003) were adopted for interference corrections following the procedures developed by Kemp et al. (2009). Details of the analytical procedure are discussed by Kemp et al. (2009) (Chauvel & Blichert-Toft 2001), and chondritic (Bouvier et al. 2008) parameters were used for model age calculations.
Sm-Nd isotope measurements were conducted in the Laboratory of Isotope Geochemistry and Geochronology (Institute of Geochemistry and Analytical Chemistry, Russian Academy of Sciences), Moscow, and at the Laboratory for Isotope Geology, Swedish Museum of Natural History, Stockholm. In Moscow, a mixed 150 Nd + 149 Sm tracer was added to 20 -30 mg of rock powder, which was then dissolved in a mixture of hydrofluoric and nitric acids (5:1) using a thermostat at a temperature of 200 8C. This solution was evaporated and the residue was transformed into chloride form. Samarium and neodymium were extracted using two-stage ion exchange chromatography. In the first stage, all rare earth elements were extracted using a DOWEX 50W_X8 cation exchanger, and in the second, samarium and neodymium were separated using the reagent HDEHP. The blank contamination was 0.03 and 0.1 ng for Sm and Nd, respectively. Nd and Sm isotope compositions were measured using a Triton multicollector mass spectrometer using the isotopic dilution method. The La Jolla Nd standard yielded a 143 Nd/ 144 Nd ratio of 0.511857+7 (2s, n ¼ 21 Chemical preparation procedures and Nd analysis in Stockholm followed standard routines in use at the Laboratory for Isotope Geology (De Ignacio et al. 2006). A Thermo-Finnigan Triton TIMS instrument was used for the isotope analyses and data were normalized to 146 Nd/ 144 Nd ¼ 0.7219. The accuracy of the measurements was monitored by running a series of BCR-1 and La Jolla standards. Depleted mantle model ages were calculated using the values by DePaolo (1981) for the depleted mantle.

Analysed materials and analytical results
The internal structures of zircon in all samples are highly variable, reflecting extensive metamorphic reworking of the host rocks. CL images of selected crystals are shown in Figure 3 (enderbite samples C10-U4 (KozachyYahr) and 06-BG38 (Odesa)) and Figure 4 (Fedorivka metasediment sample 89/ 551). For comparison, CL images from the Soroki metasediment samples CU-1, 5/88 and 92/218 reproduced from Bibikova et al. (2010) have also been included in Figure 4. Many crystals have cores of variable appearance with one or more distinct overgrowths, the latter demonstrating that the host rocks have been through more than one metamorphic or magmatic episode, including growth of new zircon. Overgrowths without distinct CL structure, interpreted to indicate growth under metamorphic conditions, are common (e.g. Fig. 3g, h, l; Fig. 4B(g, h, i)). Some crystals display internal variations interpreted as either magmatic zonation or later textural overprints, but no clear corerim structure. Internal zonation patterns indicating metamorphic recrystallization are shown in, for example, Figures 3g, k-m & 4A(h, i), B(h, k). Other crystals display structureless interiors of variable CL brightness. CL-light to -intermediate, ovalshaped crystals which are optically transparent without visible internal structure, but which show zonation in CL, represent a zircon type commonly associated with granulite facies metamorphic conditions, and they are not uncommon (Fig. 4B(l)).
All analysed spots in enderbite zircon were characterized with respect to zircon type and position in the crystal, as shown by CL images. Analysed zircon and zircon domains in crystals from enderbite sample 06-BG38 and C10-U4 were classified as spots in cores, in crystals without clear corerim structure, and in rims. Some spots which may straddle both core and rim domains were also identified (Table 1). In C10-U4, the cores have been further divided into CL-light cores which usually are zoned, and CL-dark cores without discernible zonation (Table 1). In 06-BG38, roundish crystals which are optically transparent and structureless have been distinguished as a separate type (Table 2). These tend to have low U concentrations, in most cases less than 50 ppm. Crystals which display optically and CL-visible internal structures but no clear core-overgrowth relation, another common zircon type in 06-BG38, do on the other hand show the same range of U concentrations as the analysed zircon cores ( Table 2). The characterization of analysed spots into different types is subjective.
Laser ablation inductive coupled plasma (LA-ICP) isotope analysis of Hf in zircon was performed in 74 spots in 73 zircon crystals from Soroki metasediment samples CU-1, 5/88 and 92/ 218 which have been analysed previously for U, Pb (Bibikova et al. 2010), and in 60 spots in 59 zircon crystals from enderbite sample 06-BG38. The Soroki metasediment Hf analyses were centred around the pits made by Cameca 1270 ion microprobe during U -Pb dating. 06-BG38 was analysed for Pb isotopes in conjunction with the LA-ICP Hf isotope analyses, and thereafter for U -Pb isotopes by SIMS in Stockholm in order to clarify the discordancy of analysed zircon. These SIMS analyses in 06-BG38 zircon were made in the same zircon domains as the Hf analyses, and complemented by another 15 U -Pb spot analyses in crystals which had not been analysed for Hf isotopes. Sixty-two spots in 57 zircon crystals from the Kozachy Yahr enderbite C10-U4 and 78 spots in 68 zircons from the Fedorivka metasediment sample 89/551 were analysed for U -Pb by SIMS in Stockholm.
Uranium-lead results for zircon from enderbite sample C10-U4, Kozachy Yahr, are presented in Table 1 and in a Tera-Wasserburg diagram in Figure 5. Hafnium and U -Pb results for zircon from enderbite sample 06-BG38, Odesa quarry, are presented in Table 2, Figure 6 (U -Pb Tera -Wasserburg diagram) and Figure 8 (Hfage diagram). Uranium-lead results for Fedorivka metasediment sample 89/551 are presented in Table 3 and in a Tera -Wasserburg diagram in Figure 7. Hafnium results for Soroki metasediment samples CU-1, 5/88 and 92/218 are presented in Table 4 and Figure 9 (Hf-age diagram). Owing to the complexity of the presented data, the interpretation not straightforward. Our best estimates of primary and metamorphic ages, and Hf results, are presented in conjunction with the discussion of the analytical results.
Sm-Nd analysis of whole-rock samples from the Dniestr-Bug region and the Azov Domain was performed in Stockholm and Moscow. Sampled rocks from Dniestr-Bug include the enderbites 06-BG38 (Odesa) and C10-U4 (Kozachy Yahr) analysed for Hf and U -Pb in this study, two samples from a quartzite enclosed within the strongly tectonized Odesa enderbite, one mafic granulite from the Odesa locality and a garnet-mica schist from the nearby Zavallya open pit graphite quarry. Samples from the Azov Domain include metasediment 89/551 from the Fedorivka structure, analysed for U -Pb in this study, Soroki samples 7/11 and 5/88 which have been analysed for Hf in this study and for U -Pb by Bibikova et al. (2010), and one additional Soroki metasediment sample. Results and Nd model ages are presented together with some data from the literature in Table 5, and discussed in the following section.

Discussion and interpretations
Most of the zircons analysed display open U -Pb systems with more or less discordant U -Pb ages, which makes it difficult to determine their primary ages. Furthermore, in rocks with complex thermal histories and in particular in the Archaean, even U -Pb ages which are concordant within analytical error may no longer be that of the primary age. This is due to the curvature of the concordia curve in Archaean time, which causes discordia lines connecting the times of primary crystallization and metamorphism to run close to the concordia curve even when there is a significant time span between primary crystallization and metamorphism. The 100 m μ criteria used in this study for identification of reliable ages are described below. Hafnium and O are essential constituents of the zircon lattice, and zircon is renowned for being very robust against subsequent disturbance. The Hf and O isotopic compositions of zircon reflect the isotope compositions of the magma from which the zircons crystallized, and they are typically preserved even in crystals which have been open to U -Pb exchange. Initial isotopic compositions of Hf can convey information about both the crustal residence time of the material from which the zircons crystallized and the nature of the mantle source from which the crust originated. However, calculation of initial Hf isotopic compositions and Hf model ages requires that the crystallization age of the zircon that is its primary age is known and for zircon with disturbed U -Pb systems this is difficult to establish. Failure to determine zircon ages accurately has been identified as an important reason for uncertainties about the interpretation of the early Archaean evolution of Earth, and the composition of the early Archaean mantle based on the Hf isotope record in zircon (e.g. Hawkesworth et al. 2010;Guitreau et al. 2012).
For recrystallized zircon, and for new zircon crystallizing on older cores, another complication is that the crystals may take up Hf from the surrounding rock, and given that rock must have had a different, higher Lu/Hf and therefore a different Hf isotope ratio, the resulting newly crystallized zircon could have a hybrid signature resulting from the mix of Hf from the old zircon and the surrounding rock.
Despite these complications, the Hf isotopic composition of zircon in combination with U -Pb isotope data and knowledge about zircon texture, as visualized by, for example, CL imaging, still provides a powerful tool to convey information about the origin and evolution of rocks even in old polymetamorphic crust. In this contribution we use U -Pb isotope data to estimate the primary ages of different rock units and the ages of metamorphism, and combined Hf and U -Pb systematics to constrain the crustal residence ages for magmatic enderbite and for source rocks of metasedimentary zircon, and the isotope composition of early Archaean mantle sources.
For identification of reliable primary and metamorphic ages based on the presented zircon U -Pb data we use a combination of criteria. For zircon populations which on the basis of their internal structure in CL images can be interpreted to be of magmatic origin and can be expected to be co-genetic, and which do not display any traces of older inherited components and have concordant or close to concordant U-Pb ages, the oldest 207 Pb/ 206 Pb ages are taken to provide minimum estimates of the true geological crystallization ages. The same applies for 207 Pb/ 206 Pb ages of distinct overgrowths which fulfill the above criteria. If several analysed spots in such zircon form a cluster with similar ages, this indicates that the age of the cluster is close to the true crystallization age. In this study, one cluster of Eoarchaean age is identified in enderbite sample C10-U4. In enderbite sample 06-BG38, a precise Palaeoproterozoic age is defined by several concordant and close to concordant zircon overgrowths which provide the same 207 Pb/ 206 Pb age.
More generally, clusters of zircon analyses with concordant and close to concordant U-Pb ages can provide estimates of ages of zircon growth or recrystallization. Depending on their CL appearance, U, Th concentrations and locations of analysed spots in the zircon structure, such ages can be interpreted to be either magmatic or metamorphic. The identification of such clusters can be equivocal, in particular for clusters of recrystallized zircon defining metamorphic ages. This is due to the possible combined effects of inherited Pb and subsequent Pb loss which can cause calculated ages to be either older or younger than the age of zircon growth and recrystallization. In this study, one cluster of Mesoneoarchaean age is identified in enderbite sample C10-U4 and another, albeit less precisely defined, in metasediment sample 89/551. Fig. 3. Cathodoluminescence images of selected zircons from Dniestr-Bug enderbites S10-U4 and 06-BG38, with SIMS analysed spots shown as ellipses. Discordancy is ,5% unless stated otherwise. 207 Pb/ 206 Pb ages in Ga and discordancy of .5% discordant spots are shown on the images, precise ages with errors are given in Tables 1 & 2. (a -i) Kozachy Yahr quarry, sample S10-U4 (Table 1). (a-f) Spots in cores belonging to the cluster of oldest concordant and nearly concordant ages of 3.70-3.79 Ga which is interpreted to reflect the primary age of the enderbite, while core ages in (g -i) are too young owing to Pb loss. (f-i) Crystals analysed in more than one spot. Overgrowths in (f-h) have 207 Pb/ 206 Pb ages of 2.79-2.94 Ga and reflect a period of metamorphism at c. 2.8 Ga. The age of overgrowth in i is 2.2 Ga and interpreted to be too young owing to Pb loss. (a) Spot 46; (b) spot 19; (c) spot 50; (d) spot 54; (e) spot 12; (f) spots 47 a, b (core); (g) spots 20 a (core), b; (h) spots 43 a (core), b; (i) spots 18 a (core), b. ( j-n) Odesa quarry, sample 06-BG38, (Table 2). Cores in ( j -l) are interpreted to be the same age as in sample C10-U4, c. 3.75 Ga, and 207 Pb/ 206 Pb ages to be too young owing to Pb loss. Rims in (l -n) reflect zircon growth at 2.0 Ga, while the rim in (k) is interpreted to be too old owing to inherited Pb. (j) Spot 7; (k) spots 48 and 62 (core); (l) spots 31 (core) and 63; (m) spots 66 (core) and 67; (n) spots 39 (core) and 64.   Table 3. SIMS analysed spots are shown as ellipses. Length of scale bar in all images is 100 mm. Analyses are ≤1% discordant unless stated otherwise. 207 Pb/ 206 Pb ages in Ga and discordancy of .1% discordant spots are shown on the images; precise ages with errors are given in Table 3. Zircon and zircon cores with magmatic zonation patterns (a-f) are interpreted to give minimum ages for source rocks, which however may be much too young owing to Pb loss. Cores with zonation patterns indicating alteration or recrystallization (g-i) do generally provide younger ages which not are considered to be geologically relevant. The age of the inner overgrowth in (d) falls within the age interval 2.7-2.9 Ga identified as a period of major Neoachean metamorphic reworking, while the geological relevance of the age of the overgrowth in g is unclear. (a) OLDEST CRUST UKRAINIAN SHIELD The combination of U -Pb age data with Hf isotope data for the same zircon can contribute to the identification of geologically meaningful ages from complex zircon populations. In magmatic rocks this may help to identify crystals with a common primary magmatic origin, which have not taken up Hf from the ambient rock during metamorphism. In metasedimentary rocks, combined U-Pb-Hf data may be used to distinguish zircon populations derived from rocks with different crustal provenance ages, which under favourable circumstances can help to identify geologically meaningful rock ages. Much of the combined Hf and U -Pb zircon data presented in this study form more or less well-defined linear trends in Hf -time space. The upper age ends of such trends which extend into the region in the Hf -time diagram with positive 1 Hf values can, even if the trends themselves reflect subsequent Pb loss (e.g. Kemp et al. 2009), provide estimates of the ages of periods of crustal formation. This is further discussed in a later section.
Initial 1 Hf values, which are used for the interpretation of zircon isotope systematics and are displayed in Hf -time space have been calculated using the SIMS 207 Pb/ 206 Pb ages. In some of the crystals from 06-BG38, which were analysed both for Pb isotopes in Bristol and subsequently for U-Pb isotopes in Stockholm, Pb -Pb ages from the LA-ICP and SIMS analysis differ considerably even though the analyses were performed as far as possible in the same zircon domains (Table 2). This may be due to variable U -Pb discordancy; the discordancy of the LA-ICP analyses in this study is not known, but a contributing factor may also be the larger volumes of zircon sampled by the LA-ICP analyses than by SIMS. Older LA-ICP ages from rims can be because the laser also has sampled older portions of the crystals, while older LA-ICP ages for cores can be explained if the laser has sampled and used up old portions of the crystals which therefore were not available for subsequent ion microprobe analysis.
In the following paragraphs we discuss first the ages of enderbites and other Dniestr -Bug rocks and age patterns in Fedorivka and Soroki metasediments, and then the constraints on the crustal evolution of the Dniestr-Bug region and the Azov Domain based on combined U -Pb and Hf isotope data on zircon and Sm-Nd whole rock data. Subsequently we discuss the Hf and U -Pb systematics of zircon resulting in the Hf-time patterns shown in Figures 8-10. Finally we briefly compare the early Archaean evolution in the Podolian and Azov domains, and the Ukrainian Shield with other early Archaean crustal segments.

Ages of enderbites and other rocks from the Dniestr -Bug region
Enderbite is a dominating rock type in the area along South Bug river where most samples for this study were collected (Fig. 2a). Typically it is tectonically intercalated with mafic bands or enclaves on a decimetre-to metre-scale. Similar-looking occurrences of enderbite along the South Bug river have been interpreted to form part of either the Dniestr-Bug Series or the younger Bug Series, indicating that all enderbite in the region may not be of the same age. Previous U-Pb dating of multigrain zircon fractions from Dniestr -Bug high-grade rocks have given upper intercept ages for discordant zircon fractions up to 3.1-3.4 Ga (Bibikova 1984;Lesnaya et al. 1995), but did not provide precise results owing to the complex internal structure of analysed zircons. U -Pb SIMS data for enderbite zircon from the Kozachy Yahr and Odesa open rock quarries, located in the Dniestr-Bug Series on opposite banks of the South Bug river, were presented for the first time by Claesson et al. (2006), who reported complex age patterns reflecting the internal zircon structure and ages up to 3.75 Ga from Odesa (three concordant analyses in the same crystal), and up to 3.65 Ga from Kozachy Yahr. Claesson et al. (2006) also identified a period of metamorphic zircon growth at c. 2.8 Ga indicated by overgrowths on older cores and new, low-U, isometric light-coloured zircon crystals, and a second period of metamorphic zircon growth at 2.0 Ga.
The enderbite samples 06-BG38 and C10-U4 investigated in the present study were also collected in the Odesa and Kozachy Yahr open rock quarries, respectively, but not from the same outcrops as the samples analysed by Claesson et al. (2006). The metamorphic reworking of enderbite, and by inference of the U -Pb systems in zircon, varies on outcrop-scale, and the purpose of the new sampling was to obtain a better insight into the primary age and metamorphic evolution of the enderbite. As shown below, the new U -Pb results from samples 06-BG38 and C10-U4 confirm and add more detail to the results from the reconnaissance study by Claesson et al. (2006), the main difference being a revised best estimate of the primary age of the enderbite.
The age data presented here indicate that the zircon populations in both enderbite samples, 06-BG38 and C10-U4 from the Odesa and Kozachy Yahr quarries, have been strongly modified by metamorphic reworking of the host rocks. The data indicate Pb loss, and probable new zircon growth during or after those Pb loss epochs. In detail, zircon from the Kozachy Yahr sample C10-U4 have preserved a better record of the original age of the rock, while the metamorphic overprint is better registered in zircon from the Odesa sample 06-BG38. This is in general accordance with a field observation that the immediate surroundings of C10-U4 appear to be more homogeneous and less affected by tectonic reworking. In general, however, the field appearance indicates that degree of deformation in both Odesa and Kozachy Yahr quarries varies on outcrop scale.
In enderbite sample C10-U4 from Kozachy Yahr, the oldest ages were obtained for CL-light to -intermediate cores with more or less well developed zonation patterns (Fig. 3). Nine analyses in cores yield a cluster of ,5% discordant U -Pb ages with 207 Pb/ 206 ages of 3.70-3.79 Ga, which we interpret to reflect the primary magmatic age of the enderbite (Fig. 5). Our best estimate of this age is c. 3.75 Ga. Younger core ages are typically more discordant, and not considered to be geologically meaningful. Most light-coloured cores have 100-1000 ppm U while dark-coloured cores are richer in uranium, up to 3600 ppm. These typically are more discordant, consistent with a high degree of Pb loss.
Most analysed zircon rims and crystals with metamorphic CL-appearance in sample C10-U4 form a group around c. 2.7-2.9 Ga (Fig. 5). This includes eight analyses in rims, six in metamorphic crystals and four in zircon structures interpreted as cores, which all are ,3% discordant. The U concentrations in rims and metamorphic grains in this group are generally ,300 ppm, while the cores have .700 ppm U. We interpret this cluster of ages at c. 2.8 Ga to reflect a previously documented period of metamorphic reworking, including granulite facies metamorphism, in the Podolian Domain at this time. A well-defined cluster of zircon ages at c. 2.8 Ga in an enderbite sample was also reported by Claesson et al. (2006).
Younger ages in rims and metamorphic zircon from sample C10-U4 are not interpreted to be geologically meaningful. These ages could be due to Pb loss and incorporation of some old radiogenic lead when new zircon crystallized, probably during a period of Palaeoproterozoic metamorphism around 2.0 Ga, which is better displayed in enderbite sample 06-BG38. Metamorphism reaching amphibolite facies is known to have affected the region at this time, and this is corroborated by the recent identification of c. 2.0-2.1 Ga zircon, interpreted to be metamorphic, in dykes which postdate the c. 2.8 Ga regional metamorphism.
The U -Pb age pattern shown for zircon from enderbite sample 06-BG38, Odesa quarry (Fig. 6), is similar to that for C10-U4, Kozachy Yahr, but a larger fraction of analyses in 06-BG38 zircon are discordant and 06-BG38 also lacks the clusters of concordant ages around c. 3.75 and 2.8 Ga which are seen in C10-U4. Further, 06-BG38 displays a group of more or less concordant rim ages at 2.0 Ga.
A minimum estimate of the primary age of enderbite sample 06-BG38 is provided by the oldest cores in 06-BG38, which are c. 3.6 Ga and slightly discordant. However local field observations indicate that Odesa and Kozachy Yahr, which are located less than 1 km from each other, are parts of the same enderbite complex and by inference probably the same age at c.3.75 Ga. This is corroborated by data from an unzoned zircon core in another enderbite sample in the same complex presented by Claesson et al. (2006), which in three analysed spots yielded within error concordant ages of 3.73 -3.77 Ga. A common origin is also supported by the Hf data discussed further below. The younger core ages in 06-BG38, most of which are strongly discordant, are not considered to be geologically meaningful. In a Tera -Wasserburg diagram in Figure 6, a diffuse trend defined by most of the structureless type zircon, and some cores, possible cores and rims, may be discerned. The least discordant analyses in this trend have a range of 207 Pb/ 206 Pb ages of 2.6-2.8 Ga; two structureless zircons (spots 16 and 23) are less than 1% discordant and have 207 Pb/ 206 Pb ages of c. 2.65 Ga. This is likely to reflect zircon crystallization and recrystallization during the c. 2.8 Ga metamorphic event which is better recorded in sample C10-U4. We interpret the slightly younger ages of the nearly concordant analyses in 06-BG38 to be due to a second period of metamorphism which was more intensely recorded in 06-BG38 than in C10-U4.
The youngest, c. 2.0 Ga ages in 06-BG38 were with one exception obtained from distinct rim structures identified by CL. The SIMS analyses of CL-identified rims include a group of four rims with an age of c. 2040 Ma and one slightly older at 2077 Ma. Most of these rims were not analysed in the LA-ICP session. There was obviously a zircon-forming event at c. 2.0 Ga, which we correlate with strong Palaeoproterozoic metamorphic reworking which also has disturbed the U -Pb isotope systems and contributed to the strong discordancy in many of the older crystals. One rim in the 2.0 Ga group is almost 5% discordant, but this has not affected the 207 Pb/ 206 Pb age, indicating that the discordancy is due to recent Pb loss.
One SIMS analysis in a light structureless crystal in 06-BG38 (spot 17) is concordant and it has an even lower 207 Pb/ 206 Pb age of 1855 Ma. There is no obvious reason why this analysis gives a young age. However, it might be geologically meaningful. There is evidence of an important zircon-forming  (Fig. 2b, Table 3), with zircon types indicated. The oldest, concordant and close to concordant .3.5 Ga ages in cores provide minimum ages for the oldest rocks in the source region of the metasediment. The concordant 2.7-2.9 Ga ages in crystals without core-rim structure, and in rims, reflect a period of metamorphic reworking. Many of these have low U concentrations, ,100 ppm, and high Th/U ratios indicating growth under granulite facies conditions. (b) Metasediment samples from the Soroki greenstone belt (Fig. 2b), with zircon types indicated. Diagram is based on data previously reported in Bibikova et al. (2010). event at c. 1.8 Ga in the western Ukrainian Shield, both to the north in Novohrad-Volynskyy region and to the east in the vicinity of Krivyy Rih. Monazite separated from Bug Series quartzite near the location of sample 06-BG38 has yielded an age of 1857.5 + 1.1 Ma (Stepanyuk et al. 2004). Cases of other outliers are two analyses in 06-BG38 (spots 40 and 75), which are more than 10% reversely discordant. Both appear to be technically acceptable, and there is no obvious reason for this reverse discordancy. In summary, our interpretation of the new U -Pb zircon data presented here from the Kozachy Yahr and Odesa quarries, taken together with the results from Claesson et al. (2006), is that the primary age of the enderbite in both quarries is the same, and our best estimate of this age is 3.75 Ga. The enderbite has been subject to high-grade metamorphism, probably to granulite facies, at c. 2.8 Ga, which is reflected in the recrystallization of older zircon, crystallization of new, low-U metamorphic zircon and metamorphic zircon rims. A second period of metamorphic reworking at 2.0 Ga is reflected in a second generation of low-U zircon rims in enderbite sample 06-BG38.
Additional insight into the ages and age relations for enderbites and other rocks in the Dniestr -Bug Series is provided by the Nd model ages presented in Table 5. The oldest Nd model ages are obtained for the Kozachy Yahr (T DM ¼ 3.8 Ga) and Odesa (T DM ¼ 3.7 and 3.9 Ga) enderbites. T DM ages for Odesa quartzite and granulitic gneiss are 3.4-3.5 Ga, while the ages for the garnet-mica schist and gneiss samples from Zavallya are 3.1-3.2 Ga. For comparison, two enderbite samples from another locality, near Litin c. 180 km to the NW of Kozachy Yahr and Odesa, yield T DM model ages of 3.6 and 3.65 Ga (Stepanyuk et al. 1998). These samples are also distinguished by their higher 147 Sm/ 144 Nd ratios of 0.13 -0.16, compared with 0.10 -0.12 for Odesa and Kozachy Yahr enderbites.
Bearing in mind the possibility that the Nd whole-rock systems might have been disturbed during later metamorphism, and the uncertainty about the existence of a modern-type depleted mantle reservoir in the early Archaean, the results for the Kozachy Yahr and Odesa enderbites are in good agreement with the U -Pb ages and Hf data, and support the conclusion that these remobilized rocks are also the oldest rocks so far identified in  Black symbols, less than 2% discordant; grey symbols, 2 -8% discordant; open symbols, more than 8% discordant. Most analyses plot along an array, indicated in the diagram with a dashed line, with a slope corresponding to a 176 Lu/ 177 Hf ratio close to zero. This array reflects Pb loss but closed Hf isotope systems in zircon crystals during metamorphic reworking, while data points on the upper left side are interpreted to indicate uptake of Hf from the ambient rock. The initial 1 Hf ratio of zircon crystals can be approximated from the extension of the upper right end of the array to 3.75 Ga, which based on U -Pb data for sample C10-U4 is interpreted to be the primary age of the enderbite.  Character of analysed zircon domain as shown in CL image and by Hf isotopic composition. Core, Spot in core or in crystals without clearly CL-visible core -rim structure; Rim, spot in rim; y, spot in zircon belonging to the younger population with crustal residence age c. 3.1 Ga.
the Podolian Domain. The model ages for enderbites from Litin are slightly younger, and also younger than the rock age 3.75 Ga for Odesa and Kozachy Yahr. This provides some support for the interpretation that the Dniestr-Bug enderbites are not all of the same age. The Zavallya gneiss and garnetmica schist, which are interpreted to form part of the Bug Series, also have the lowest Nd crustal provenance ages.

Age patterns in Fedorivka and Soroki metasediments, Azov Domain
The 78 zircon U-Pb analyses from the Fedorivka sample 89/551 are mainly from zircon cores. Twelve of these cores give 207 Pb/ 206 Pb ages .3.5 Ga, while the remainder give 207 Pb/ 206 Pb ages ranging from 3.3 to 3.5 Ga or younger, down to 3.0 Ga. Many are strongly discordant (Fig. 7a), and do not provide meaningful estimates of rock ages in the source region of the metasediments. Some 20 crystals without any distinct core-rim structures, and zircon rims, give concordant or close to concordant U -Pb ages of 2.7-2.9 Ga. In the following we discuss the Fedorivka age data in conjunction with the U -Pb zircon data for four samples of Soroki greenstone belt metasediments reported by Bibikova et al. (2010). For comparison, a Tera -Wasserburg diagram showing the Soroki U -Pb zircon data is included here (Fig. 7b).
The Soroki study was focused on the oldest components in the Soroki metasediments. Bibikova et al. (2010) identified a group of ages in the range 3.5 -3.6 Ga, and several zircon cores older than 3.7 Ga, and concluded that Azov domain metasediments include an important, previously unidentified Palaeoarchaean component of that age and also Eoarchaean .3.6 Ga material, demonstrating the existence of crust of such age in the provenance region for these rocks. Bibikova et al. (2010) also identified a minor, 3.05-3.08 Ga-old population of zircon without cores characterized by a rounded prismatic shape and a characteristic CL zonation (Fig. 4B, images k and l) Lu/ 177 Hf ratio close to zero, demonstrating the direction of movement in the diagram of zircon crystals affected by Pb loss, which causes a lowering of apparent 207 Pb/ 206 Pb ages. The bimodal age distribution indicates that the analysed material is dominantly derived from rocks with two distinct crustal formation ages. The younger array extends into the region in the diagram with distinctly positive 1 Hf values, demonstrating that the source rocks of these zircons were derived from depleted mantle sources. The zircons in the older array appear to be derived from mildly depleted mantle sources. conclusion by Bibikova et al. (2010) that the Azov Domain includes a regionally important component of Paleoarchaean or older crust. In contrast to Soroki, the Fedorivka zircon population analysed here does not show any evidence of 3.05 -3.08 Ga zircon growth. Furthermore, the 2.7-2.9 Ga age population in sample 89/551 is not represented in any of the Soroki samples. The 2.7-2.9 Ga structureless zircons and rims commonly have low uranium concentrations of less than 100 ppm, in many cases as low as 20-30 ppm, and high Th/U ratios of about 1.0. This indicates growth under granulitic conditions. Based on the data presented here, it is not clear if 2.7 -2.9 Ga metamorphic zircon represents a single metamorphic event, or if the metamorphism was older than or postdated sedimentation. Some of the 2.7-2.9 Ga zircon may be clastic and in that case demonstrate the occurrence of Mesoarchaean material in the metasediment. Such an interpretation is supported by the Nd T DM whole rock model age of 3.0 Ga for the Fedorivka sample 89/551, which is younger than the model Nd ages for the three samples of Soroki metasediments of 3.3-3.4 Ga ( Table 5).

The earliest crust in the Dniestr -Bug region
In 1 Hf -time space (Fig. 8), most of the zircon crystals from enderbite sample 06-BG38 which were analysed for both Hf and U -Pb isotopes form a linear array. In general such arrays can give insight into the nature of the primary reservoir, particularly its Lu/Hf, but this is not the case here. The Pb data for discordant zircons had previously been interpreted to result from Pb loss and to give no real age information. The Lu -Hf system supports this conclusion. Essentially the Hf isotope composition of these zircons is constant. This results in a calculated Lu/Hf for the array of close to zero, an artefact of the 207 Pb/ 206 Pb data having no time significance. Felsic and mafic crustal reservoirs (real rocks) typically have 176 Lu/ 177 Hf ratios of c. 0.01 and c. 0.02, respectively. Given our best estimate of the primary age of the zircon in the Odesa and Kozachy Yahr enderbite rocks, the corresponding 1 Hf value is about c. +2.
The production of magma with the geochemical composition of the enderbite, with a SiO 2 content over 60%, requires differentiation from a more mafic juvenile crustal precursor. Since the isotopic composition of the mantle from which the primary crust was extracted is unknown, the age of this crustal precursor is poorly constrained. It can have preceded the enderbite-forming differentiation by several hundred million years, or have taken place shortly before the generation of the enderbite. A minimum age is c. 3.75 Ga, the age of the enderbite, while an upper limit may be estimated assuming a mafic precursor with a 176 Lu/ 177 Hf ratio of 0.02 derived from depleted mantle with the model composition of Griffin et al. (2002). The latter results in a crustal provenance age of 3.9 -4.0 Ga. Such crustal provenance ages are also consistent with the whole-rock depleted mantle Nd model ages for the Odesa and Kozachy Yahr enderbites of 3.7-3.9 Ga ( Table 5).

Formation of Archaean crust in the Azov Domain
The majority of the Hf-age data from the three samples of Soroki metasediments analysed for both Hf and U -Pb isotopes form two groups on the 1 Hf -time plot (Fig. 9), with most analyses scattering along two more or less distinct arrays towards younger ages and more negative 1 Hf values (Fig. 9). Since most of the 207 Pb/ 206 Pb ages for individual data points shown in Figure 9 do not reflect accurate primary crystallization ages but have been modified by metamorphic reworking, these cluster-arrays are attributed to Pb loss and not to the evolution of distinct segments of crustal material. As discussed above, the oldest (and most U-Pb concordant) ages on any such array are taken to be the most reliable estimates of the primary crystallization ages, in this case at c. 3.5-3.6 Ga and 3.08 Ga for the two arrays on Figure 9.
Thus, the cluster of c. 3.5 -3.6 Ga zircons with 1 Hf between 26 and 0 provides a minimum age for one period of magmatism. Zircon older than 3.6 Ga with 1 Hf of 22 to +2, which define the high 1 Hf end of the older cluster-array in Figure 9, may represent slightly older crustal magmatism. The positive 1 Hf values for some of these oldest zircons suggests their host magmas were juvenile or, depending on the composition of the mantle source, that the source material for these magmas had a short crustal residence time. The cluster of zircon with ages up to 3.08 Ga and 1 Hf +1 to +4 defines the high 1 Hf end of the younger cluster-array in Figure 9. Three of these are 3-6% discordant, and the discordancy of the remaining data points in this array is undetermined.
Taken together, this suggests that the magmatic rocks in which the bulk of the analysed zircons have crystallized were derived from continental crust dominated by two periods of crust generation. A minimum crustal formation age for the source rocks of the zircon which define the older clusterarray is 3.75 Ga. In the same way as for Dniestr -Bug enderbites, an upper limit of 3.9 -4.0 Ga may be estimated assuming a mafic crustal precursor with a 176 Lu/ 177 Hf ratio of 0.02 which is derived from depleted mantle with the model composition of Griffin et al. (2002).
Similarly, 3.1 Ga is a minimum crustal residence age for the source rocks of the zircon which define the younger cluster-array in Figure 9. With a primary crust of mafic composition derived from a modern-type depleted mantle, the crustal formation age may be up to c. 3.25 Ga. This cluster-array includes all zircons from the 3.05-3.08 Ga old population of zircon without cores and with a characteristic CL zonation which was identified by Bibikova et al. (2010). The CL appearance indicates these zircons may be metamorphic rather than magmatic, but if so the Hf data require a short premetamorphic crustal residence period. This zircon population occurs primarily in Soroki sample 92/18, but also in sample 5/88. It may be derived from a local source.
We note that the estimated 1 Hf values of c. +2 at 3.75 Ga for the oldest rocks identified here in both the Azov and Podolian domains, and up to +4 at 3.1 Ga in the Azov Domain (Fig. 10), show an excellent fit with the evolution model for new crust by Dhuime et al. (2011), which is shown in Figure 10.
In addition to the two periods of crustal generation described above, the new data include some indication of material with other crustal ages. Two core analyses from Soroki metasediment samples, CU1-39 and 5/88 -14, stand out as slightly older than those within the main cluster-array. These are both less than 5% discordant and give within error identical ages of 3.785 Ga and slightly negative initial 1 Hf values of 21 to 22. Optically, these zircons are not different in appearance from zircon in the main cluster; both display cores with wide rims. This combination of 207 Pb/ 206 Pb -1 Hf data on the old side of the main cluster-array cannot be readily explained by a mixture of other identified age components in these samples. The 3.56 Ga-old analysis CU1-34, in a distinct core, may also represent material with a different crustal residencetime. It has a positive 1 Hf value of 3.4 and plots closer to the depleted mantle curve than any other analysis in Figure 9. The sparse nature of these data make interpretation speculative at best.

Hf-age isotope systematics during crustal reworking
The 1 Hf -time array formed by the majority of the zircon analyses from enderbite 06-BG38 (Fig. 8) has a slope corresponding to evolution in an environment with a 176 Lu/ 177 Hf ratio close to zero and a well-defined, straight lower right boundary  and CU-1, illustrating the evolution of analysed zircons and their host rocks, isotopic compositions of mantle sources and model crustal residence ages. In addition to CHUR and the evolution of depleted mantle (Griffin et al. 2002), the evolution curve for new crust by Dhuime et al. (2011) is also shown. Yellow stars show estimated ages for crust formation events (3.75 Ga for both Podolian Domain enderbite and most metasedimentary zircon in Soroki greenstone belt metasediments, and 3.10 Ga for the younger portion of Soroki metasediment zircon), and corresponding initial 1 Hf values of c. +2 and +4. The stars show an excellent fit to the evolution line for new crust. The grey band shows the effect of Pb loss on 3.75 Ga zircon, while the green band shows the evolution of 3.75 Ga bulk crust with a 176 Lu/ 177 Hf ratio of 0.015, which is intermediate between typical felsic and mafic crust with 176 Lu/ 177 Hf ratios of c. 0.010 and 0.020 respectively. More detail information about individual samples is shown in Figures 8 and 9. towards older ages and more negative 1 Hf , while there is a scatter of data points on the young side of the array. Zircon has Lu/Hf close to zero, and the observed Hf-age array in Figure 8 can be explained if all the zircon along the array originally crystallized during a magmatic event and individual zircon crystals since then have closed Lu -Hf reservoirs, but they have suffered Pb loss during subsequent metamorphic reworking. This requires that recrystallization of primary magmatic zircon, and also growth of zircon rims which plot along the array, has taken place without significant exchange of Hf between zircon and the ambient rock. Similar Pb loss trends have been presented by, for example, Zeh et al. (2008) and Guitreau et al. (2012).
The scattering data points to the left of and above the array in Figure 8 are interpreted to reflect in-mixing of Hf from the ambient rock during metamorphic reworking. This is supported by the observation that most of these data points are from crystals without CL-visible internal structures, or with internal structure indicating recrystallization. Two core analyses which do not plot along the array may be metamorphic zircon overgrown by younger rims.
The slopes of the clusters-arrays for Soroki metasediments, Azov Domain (Fig. 9), are less well constrained, but they are similar to the slope for the Podolian Domain enderbite 06-BG38. This is best explained by Pb loss from zircons defining these arrays during later strong metamorphic overprint, similar to our interpretation of the array for enderbite 06-BG38. Data points scattering between the two arrays in Figure 9 may reflect inherited zircon components in the rocks where these zircons crystallized.

Comparison with other early Archaean crustal segments
The oldest crust in the Ukrainian Shield has long since been identified in the Bug region, Podolian Domain, and in the Azov Domain. Both have been metamorphically reworked both c. 2.7-2.9 and c. 2.0 Ga ago, and the two domains have also been correlated with each other. Geochronological and isotope-geochemical results presented here and in Bibikova et al. (2010) have identified fragments of Archaean crust as old as 3.75 Ga in both domains. However, in spite of the geochronological similarities, we argue that the Podolian and Azov domains probably have evolved independently of each other before the amalgamation of the Ukrainian Shield. In addition to the general observation that present-day geographical proximity between early Archaean crustal segments not is a good criterion for a common origin, our argument is that the intervening, granite-greenstone belt-dominated Middle Dniepr Domain does not include any crust older than 3.2 Ga. Furthermore, the Middle Dniepr Domain has, in contrast to the Podolian and Azov domains, only been very mildly affected by post-Archaean metamorphism. Bleeker (2003) correlated the early development of different cratons and suggested the existence of three main Supercratons in the early Archaean -Sclavia, Superia and Vaalbara -based on similarities in their stratigraphic and structural evolution. Recent progress in single zircon dating technique has resulted in rapidly growing age databases. Zircons as old as 3.8-3.9 Ga have been identified from most cratons, especially from Archaean sedimentary rocks Zheng et al. 2004;Davis et al. 2005;Lee et al. 2007;Liu et al. 2008;Pietranik et al. 2008;Willner et al. 2008;Wu et al. 2008;Lauri et al. 2011). Thus, it is almost impossible to use early Archaean U -Pb zircon ages to correlate the earliest stages in formation and evolution of ancient cratons.
Another geochronological reference mark is the time of emplacement of sanukitoids, high-Mg subalkaline granitoids, which are interpreted to manifest the time of cratonization (Stern & Hanson 1991;Martin et al. 2010). Sanukitoids have been recognized and dated in nearly all cratons. Their ages vary from 2.95 to 2.97 in the Pilbara and Kaapvaal cratons (Smithies & Champion 2000), around 2.8-2.7 Ga in Superior and Fennoscandia (Stern & Hanson 1991;Bibikova et al. 2005Bibikova et al. , 2008Halla 2005) to about 2.6 Ga in the Slave province (Bleeker 2003) and 2.5 Ga in the Darwhar craton, India (Moyen et al. 2003). Sanukitoids in the Azov domain have been dated at 2.90-2.93 Ga (Bibikova et al. 2008), that is, most similar to ages in the Pilbara and Kaapvaal cratons which form parts of the Vaalbara Supercraton of Bleeker (2003).

Summary and conclusions
U -Pb and Hf data for zircon and Sm-Nd whole rock data from the Podolian and Azov Domains, Ukrainian Shield, provide information about the origin, ages and metamorphic evolution of old Archaean crust in these domains. The zircon populations are complex, overgrowths of new zircon on older cores are common and the internal structures of many crystals show these have been recrystallized. The majority of U -Pb dated zircons are discordant. We demonstrate that both domains include rocks as old as 3.75 Ga, and identify the ages of Archaean and Palaeoproterozoic metamorphic reworking of this old crust.
In the Podolian Domain, U -Pb zircon data from the Kozachy Yahr enderbite show that it is 3.75 Ga or older. We interpret the nearby Odesa enderbite to be coeval with Kozachy Yahr at 3.75 Ga, and to have been derived from a source with a chondritic to mildly depleted 1 Hf isotopic signature of up to +2. The age of the primary crust from which it was extracted can be constrained to within 3.75-3.9 Ga, consistent with Sm-Nd whole rock data for the same rocks (Fig. 10).
The oldest U -Pb ages from metasediments in the Fedorivka greenstone belt, Azov Domain, are consistent with previously published zircon U -Pb data from the Soroki greenstone belt. This supports that the Eoarchaean, c. 3.7 Ga or older zircons from Fedorivka and Soroki metasediments in the Azov Domain reflect ages of rocks which were regionally distributed at the time when the material was eroded.
For the Soroki metasediments in the Azov Domain, combined Hf and U -Pb data indicate that most of the analysed material was derived from sources with a minimum crustal formation age of 3.75 Ga and a corresponding initial, mildly depleted 1 Hf isotopic signature of up to +2. A maximum age for the primary crust can be constrained to c. 3.9 Ga. A younger, Mesoarchaean crustal component in Soroki metasediments is represented by close to concordant 3.05-3.08 Ga old zircons with initial 1 Hf up to +4. The crustal residence age of this material can be constrained to 3.1-3.25 Ga.
The Hf isotopic signatures of the 3.75 Ga crust identified in the Azov and Podolian domains, and the 3.1 Ga crust in the Azov Domain (Fig. 10), show an excellent fit with the evolution model for new crust by Dhuime et al. (2011). The presented data also lend some support to the model by Pietranik et al. (2008) with episodic formation of crust in the Archaean, but the Mesoarchaean episode identified here is younger than the 3.4 Ga episode proposed by Pietranik et al. (2008).
In the enderbite data from the Dniestr-Bug region, Podolian Domain, a period with zirconrecrystallization/new growth, including granulite facies metamorphism, is identified at c. 2.8 Ga. The 2.04 Ga zircon overgrowths are interpreted to reflect a second period of metamorphic reworking. Neoarchaean, 2.7-2.9 Ga metamorphism has been identified in metasedimentary rocks from the Fedorivka greenstone belt, Azov Domain, and magmatism of this age may also have occurred in the metasediment source region.
The linear arrays in Hf -time space defined by most analysed zircons (Fig. 10) are interpreted in terms of Pb loss. Many of these zircon crystals appear to have been closed systems with respect to Hf since the primary crystallization. Recrystallization and growth of new zircon have in several cases also taken place with limited exchange of Hf between zircon and the host rock.