Anatectic reworking and differentiation of continental crust along the active margin of Gondwana: a zircon Hf–O perspective from West Antarctica

Abstract The Fosdick migmatite–granite complex of West Antarctica preserves evidence of two crustal differentiation events along a segment of the former active margin of Gondwana, one in the Devonian–Carboniferous and another in the Cretaceous. The Hf–O isotope composition of zircons from Devonian–Carboniferous granites is explained by mixing of material from two crustal sources represented by the high-grade metamorphosed equivalents of a Lower Palaeozoic turbidite sequence and a Devonian calc-alkaline plutonic suite, consistent with an interpretation that the Devonian–Carboniferous granites record crustal reworking without input from a more juvenile source. The Hf–O isotope composition of zircons from Cretaceous granites reflects those same two sources, together with a contribution from a more juvenile source that is most evident in the detachment-hosted, youngest granites. The relatively non-radiogenic ϵHf isotope characteristics of zircons from the Fosdick complex granites are similar those from the Permo-Triassic granites from the Antarctic Peninsula. However, the Fosdick complex granites contrast with coeval granites in other localities along and across the former active margin of Gondwana, including the Tasmanides of Australia and the Western Province of New Zealand, where the wider range of more radiogenic ϵHf values of zircon suggests that crustal growth through the addition of juvenile material plays a larger role in granite genesis. These new results highlight prominent arc-parallel and arc-normal variations in the mechanisms and timing of crustal reworking v. crustal growth along the former active margin of Gondwana. Supplementary material: Figs S1 and S2 are available at www.geolsoc.org.uk/SUP18625

Abstract: The Fosdick migmatite-granite complex of West Antarctica preserves evidence of two crustal differentiation events along a segment of the former active margin of Gondwana, one in the Devonian-Carboniferous and another in the Cretaceous. The Hf-O isotope composition of zircons from Devonian -Carboniferous granites is explained by mixing of material from two crustal sources represented by the high-grade metamorphosed equivalents of a Lower Palaeozoic turbidite sequence and a Devonian calc-alkaline plutonic suite, consistent with an interpretation that the Devonian-Carboniferous granites record crustal reworking without input from a more juvenile source. The Hf-O isotope composition of zircons from Cretaceous granites reflects those same two sources, together with a contribution from a more juvenile source that is most evident in the detachment-hosted, youngest granites. The relatively non-radiogenic 1Hf isotope characteristics of zircons from the Fosdick complex granites are similar those from the Permo-Triassic granites from the Antarctic Peninsula. However, the Fosdick complex granites contrast with coeval granites in other localities along and across the former active margin of Gondwana, including the Tasmanides of Australia and the Western Province of New Zealand, where the wider range of more radiogenic 1Hf values of zircon suggests that crustal growth through the addition of juvenile material plays a larger role in granite genesis. These new results highlight prominent arc-parallel and arc-normal variations in the mechanisms and timing of crustal reworking v. crustal growth along the former active margin of Gondwana.
Supplementary material: Figs S1 and S2 are available at www.geolsoc.org.uk/SUP18625 Convergent margins are the primary location for the growth and differentiation of the continental crust (e.g. Taylor 1967;Rudnick 1995;Davidson & Arculus 2006;Bahlburg et al. 2009;Cawood et al. 2009;Kemp et al. 2009;Mišković & Schaltegger 2009). The relative proportions of crustal growth v. crustal reworking that operate in magmatic arcs can be deciphered using the geochemistry and isotope characteristics of granites (sensu lato) that were emplaced at crustal levels significantly higher than where they were generated. Granites are commonly mixtures of more than one source component. The relative contribution of the source components can be evaluated by modelling the proportions of each required to achieve the isotope composition of the granites, where the source components are usually a well-constrained supracrustal end-member and a less well-constrained (commonly inferred) infracrustal or more juvenile end-member (Kemp et al. 2006a(Kemp et al. , 2007Appleby et al. 2009;Tulloch et al. 2009). In the case of predominantly crustal reworking, this approach may be extended to exposed crustal sections that record evidence of in situ partial melting of the putative sources as well as granites that represent melts trapped during migration to higher crustal levels, to obtain a quantitative evaluation of the proportion of crustal reworking v. crustal growth in a convergent margin system.
In this study, we report the results of Hf -O isotope analyses of zircons from two generations of granite in the Fosdick migmatite-granite complex (hereafter the Fosdick complex) of West Antarctica, which is located along the former active margin of Gondwana (e.g. Siddoway & Fanning 2009). The granites are hosted in the high-grade equivalents of a supracrustal sequence and an infracrustal calc-alkaline intrusive suite (Korhonen et al. 2010a, b) in the Ford Ranges of western Marie Byrd Land (Adams 1987;Weaver et al. 1991). The supracrustal sequence has been correlated with similar units in western New Zealand and the Lachlan Belt of eastern Australia (Gibson & Ireland 1996;Ireland et al. 1998;Pankhurst et al. 1998;Scott et al. 2009;Nebel-Jacobsen et al. 2011). The Hf -O isotope signatures of zircons from Devonian -Carboniferous granites in the Fosdick complex suggest that crustal reworking was the dominant process for the production of granites in this segment of the former active margin of Gondwana at that time. However, an isotopically more juvenile source is required to explain the wider range of 1 Hf values in zircons from the Cretaceous granites, and this source was particularly important as a component in the youngest granites emplaced in the detachment zone that was responsible for unroofing the Fosdick complex at c. 100 Ma.
Temporal variations in isotope characteristics may be attributable to different plate margin characteristics in the Devonian -Carboniferous compared with the Cretaceous, and also to different stages in strain evolution within the Cretaceous event. Although the Hf-O and Nd isotope characteristics of Fosdick complex granites are similar to the Permo-Triassic granites from the Antarctic Peninsula, they contrast with Hf-O and Nd isotope results from other localities along and across the former active margin of Gondwana, such as the Tasmanides of Australia (e.g. Kemp et al. 2009) and the Western Province of New Zealand (Bolhar et al. 2008), suggesting significant margin-parallel and margin-perpendicular variations in both the amount and nature of crustal reworking v. crustal growth with time.

Geological setting
Marie Byrd Land, West Antarctica, is situated along the former active margin of Gondwana (Fig. 1a, b; Ireland et al. 1998;Adams 2004;Tulloch et al. 2009). Pankhurst et al. (1998) divide Marie Byrd Land into an inboard Ross Province and outboard Amundsen Province that became amalgamated in the Lower Cretaceous (DiVenere et al. 1995). The Amundsen province records two episodes of calc-alkaline magmatism, from 450 to 420 Ma and at c. 275 Ma, respectively, which yield Nd model ages of 1000-1300 Ma (Pankhurst et al. 1998).
The geology of the Ross Province is dominated by the Lower Palaeozoic Swanson Formation, an unmetamorphosed to low-metamorphic-grade turbidite sequence that accumulated outboard of the Ross-Delamerian Orogen (Bradshaw et al. 1983;Adams 1986Adams , 2004Ireland et al. 1998), and Devonian-Carboniferous intrusive calc-alkaline magmatic rocks (c. 375-355 Ma) designated as the Ford Granodiorite suite ( Fig. 1c; Weaver et al. 1991;Pankhurst et al. 1998;Mukasa & Dalziel 2000). The high-grade metamorphosed equivalents of these rock units crop out in the Fosdick complex (Fig 1c, d;Siddoway et al. 2004a). Additionally, there is evidence of a Proterozoic basement underlying Ross Province (Flowerdew et al. 2007;Adams & Griffin 2012).
The Swanson Formation has been correlated with the Greenland Group in New Zealand, the Lachlan Supergroup in Eastern Australia and the Robertson Bay Group in northern Victoria Land (Adams 1986(Adams , 2004Ireland et al. 1998). The Ford Granodiorite suite is the manifestation, in Marie Byrd Land, of Devonian -Carboniferous calcalkaline magmatism along the former active margin of Gondwana. Coeval suites along the margin include the Karamea Batholith in New Zealand (Muir et al. 1996;Tulloch et al. 2009), the Admiralty Intrusives in northern Victoria Land (Muir et al. 1996) and granites in the Melbourne terrane of the Lachlan Belt in Australia (Chappell et al. 1988;Tulloch et al. 2009). In Marie Byrd Land, both the Swanson Formation and the Ford Granodiorite suite are intruded by the Lower Cretaceous Byrd Coast Granite suite (Fig. 1c;Adams 1987;Siddoway 2008), which is the main plutonic phase emplaced during the intracontinental extension that preceded break-up of Marie Byrd Land (Siddoway et al. 2004b(Siddoway et al. , 2005McFadden et al. 2010a) and Zealandia (Mortimer 2004).
Cretaceous oblique divergence and extreme lithospheric thinning preceding break-up of Gondwana in the Pacific sector has exposed a high-grade migmatite -granite complex in the Fosdick Mountains (McFadden et al. 2010a, b). Nd -Sr isotope geochemistry (Fig. 2) and U -Pb ages of paragneisses and orthogneisses exposed in the Fosdick complex suggest that these are the products of partial melting due to high-grade metamorphism of Swanson Formation and Ford Granodiorite suite protoliths, respectively (Siddoway et al. 2004a;Korhonen et al. 2010bKorhonen et al. , 2012McFadden et al. 2010b). A comparison of zircon age spectra from the Swanson Formation, paragneisses in the Fosdick complex and the cores of inherited zircons in granites support this interpretation (Fig. 3).
Two episodes of high-grade metamorphism and partial melting in the Fosdick complex have been identified: a Devonian -Carboniferous event  Tulloch et al. 2006). (b) West Antarctica and the location of the Ford Ranges. The inferred location of the boundary between the Ross and Amundsen provinces is from Mukasa & Dalziel (2000). (c) Geological map of the Ford Ranges (Siddoway 2008) with the locations for sample sites outside the Fosdick migmatite-granite complex. (d) Geological map of the Fosdick migmatite-granite complex (McFadden et al. 2010b). Samples used in this study are indicated in (c) and (d). Figures modified from Korhonen et al. (2010b). at pressures of 0.75 -1.15 GPa and temperatures of c. 870 8C, and a Cretaceous event at pressures of 0.60-0.75 GPa and temperatures of 830 -870 8C (Korhonen et al. 2010a(Korhonen et al. , 2012. Elevated temperatures and crustal melting in the Palaeozoic are attributed to thickening of the continental margin during calc-alkaline arc magmatism represented by the Ford Granodiorite suite (Pankhurst et al. 1998;Siddoway & Fanning 2009), and in the Mesozoic are attributed to moderate thickening followed by lithosphere thinning during intracontinental extension across the West Antarctic rift system (Siddoway 2008). Coeval with Cretaceous rifting, mafic dykes of c. 113 Ma to c. 98 Ma age were intruded throughout the Fosdick complex (Saito et al. 2013).
Phase equilibria modelling of the protolith compositions indicates that the metamorphic pressures and temperatures were sufficient during both episodes to partially melt both the Ford Granodiorite suite and the Swanson Formation (Korhonen et al. 2010a). Devonian -Carboniferous granites within the Fosdick complex are found as metre-to decimetre-scale layers within residual paragneiss exposed at Mt Avers and in the western Fosdick range, and metre-to decimetre-scale laterally continuous sheets in the 'layered plutonic complex ' (McFadden et al. 2010a) in the central to western Fosdick Mountains (Fig. 1d). Cretaceous granites   (Pankhurst et al. 1998); (b) zircon ages .400 Ma from paragneiss in the Fosdick migmatite-granite complex (Siddoway et al. 2004a); (c) inherited cores of zircon ages .400 Ma from granites within the Fosdick migmatite -granite complex (Korhonen et al. 2010b;McFadden et al. 2010a, b; this study). occupy three distinct structural settings: (1) subvertical sheets concordant with foliation in host gneisses on the north side of the complex, emplaced during a Cretaceous wrench deformation phase along the Balchen Glacier fault (Fig. 1d); (2) subhorizontal sheets concordant with layering in the host gneisses in the eastern and southern Fosdick Mountains that form a sheeted leucogranite complex (Fig. 1d) associated with the South Fosdick detachment zone; and (3) discordant granites emplaced late in the structural development of the South Fosdick detachment zone (Fig. 1d;McFadden et al. 2010a, b).

Sample descriptions
Fifteen samples comprising thirteen granites, one mafic gneiss and one calcareous argillite (Table 1), for which U -Pb zircon ages were determined in this study or for which previously dated zircons were available, were selected for Lu -Hf and O isotope analysis at the Research School of Earth Sciences, Australian National University. In addition, zircon separates from two samples of the Ford Granodiorite suite from outside the Fosdick Mountains (MB.214.W and MB.219.W), collected by Pankhurst et al. (1998) and made available for this study, were analysed for Lu -Hf and O isotope compositions to provide an important baseline for the study of crustal differentiation in the Fosdick complex. The 13 granites from the Fosdick complex selected for this study have been interpreted as the products of crustal melting either in the Devonian -Carboniferous or the Cretaceous (Siddoway & Fanning 2009;Korhonen et al. 2010b). Descriptions of the samples selected for Hf and O isotope analysis follow, with sample localities shown in Figure 1.
Exposures of the Swanson Formation at Mount Woodward (for location see Fig. 1c) consist of greenschist-grade red-brown argillite containing centimetre-to decimetre-scale layers and nodules of pale green calc-silicate rock. The calcareous argillite was sampled (8D27-10) to assess the detrital zircon population of the lower-grade metasedimentary rocks.
The Devonian -Carboniferous calc-alkaline Ford Granodiorite suite comprises granodiorite to monzogranite that is metaluminous to weakly peraluminous in composition (Weaver et al. 1991). Sample MB.214.W is a hornblende-biotite -titanite granodiorite from Gutenko Nunataks (Fig. 1c) to the south of the Fosdick Mountains. Sample MB.219.W from Hermann Nunataks in the eastern Phillips Mountains (Fig. 1c) is a massive equigranular granodiorite that hosts mafic enclaves. For these two samples, Hf and O isotope analyses were obtained for the same zircon grains used for SHRIMP U -Pb age determinations reported in Pankhurst et al. (1998).
Four Devonian-Carboniferous granites from the Fosdick complex ( Fig. 1d) were sampled as follows. Sample M5-G175 is a representative dark, medium-grained, equigranular biotite granodiorite, affected by only limited anatexis and subsequent solid-state deformation, that forms plutons at Mt Getz and Mt Richardson within host migmatitic orthogneisses and paragneisses (Siddoway & Fanning 2009). Sample C5-Is51A, from Mt Iphigene, is a migmatitic monzogranite orthogneiss with large Carlsbad-twinned feldspars, which is a common metaplutonic rock type in the Fosdick complex. The gneissic foliation forms decimetre-scale folds. A second sample from Mt Iphigene, C5-I26, is a porphyritic cordierite leucogranite from an irregular narrow dyke that cuts across foliation and folds. A leucogranite, C6-Aw86.1, was collected from Mt Avers, a large massif in the central Fosdick complex that is made up of residual paragneiss and intermediate metaplutonic rocks. The sample is from a 2-m-thick concordant sill of white Kfeldspar granite within residual paragneiss. The granite exhibits a weak foliation defined by sparse biotite.
For the Cretaceous, samples were selected to be representative of granites emplaced within three structural settings that developed successively during wrench to transtensive deformation (McFadden et al. 2010b), which led to detachment-controlled exhumation of the Fosdick complex (McFadden et al. 2010a). The distribution of samples allows an evaluation of whether there were short-term tectonic controls on the source evolution of the Cretaceous granites.
Two samples, K6-Bb47 and C6-BB112 from Bird Bluff, which were emplaced during oblique motion on the South Fosdick detachment zone (Fig. 1d;McFadden et al. 2010a), are representative of granites forming the sheeted leucogranite complex. K6-Bb47 is a massive concordant to discordant, coarse-grained K-feldspar granite that contains cordierite (Korhonen et al. 2010b). C6-BB112 was sampled from a 20-m-thick dyke of unfoliated K-feldspar porphyritic granite containing biotite and smokey quartz. Finally, four samples are representative of granites emplaced late in the structural development of the South Fosdick detachment zone. The granites cut host rock foliation but lack well-developed internal foliation. C6-T101 is a sample of grey quartz-rich leucogranite that contains nodular to dendritic cordierite, intruded into biotite -quartzfeldspar migmatitic gneisses at Thompson Ridge in the northwestern Fosdick Mountains. C6-Aw87.3 is K-feldspar -biotite granite collected from the margin of a sigmoidal-shaped granite body that intruded orthogneisses within the layered plutonic complex of Mt Avers. Sample C5-R60B is a light grey, medium-grained hypidiomorphicgranular K-feldspar leucogranite, collected within the South Fosdick detachment zone at Mt Richardson. The contacts of the leucogranite are both concordant and discordant to solid-state foliation in host gneisses, and the leucogranite lacks a foliation. The youngest sample, M5-G174 from Mt Getz, is one in an array of equigranular, mediumgrained, two-mica granite dykes that crosscut the gneissic foliation (Fig. 1d). Discordant dykes of this generation exist throughout the central Fosdick complex from Ochs Glacier to Mts Getz and Bitgood (Fig. 1d).
A single mafic gneiss sample, C5-Mj74, is an apatite-rich garnet -biotite gneiss collected from Marujupu Peak within the layered plutonic complex of the central Fosdick Mountains (Fig. 1d). Garnets are spatially associated with coronae of leucocratic material that suggests a peritectic origin.

Methods
U -Pb zircon age determinations were made using SHRIMP II at the Research School of Earth Sciences, Australian National University (RSES-ANU) and SHRIMP IIe at Geoscience Australia, following the procedures described in Williams (1998 and references therein). Data were reduced using the SQUID Excel Macro of Ludwig (2001). The zircon U -Pb ratios were normalized relative to a value of 0.0668 for the Temora reference zircon, equivalent to an age of 417 Ma (Black et al. 2003); analytical uncertainties for the respective analytical sessions are given in the footnotes to Table 2. Uncertainties reported in Table 2 for individual analyses (ratios and ages) are given at the 1s level. Tera-Wasserburg concordia plots (Tera & Wasserburg 1972), probability density plots with stacked histograms, and weighted mean 238 U-206 Pb ages were calculated using ISOPLOT/EX (Ludwig 2003). Where appropriate, the 'Mixture Modelling' algorithm of Sambridge & Compston (1994) via ISOPLOT/EX was used to unmix statistical age populations or groups. From such groupings, weighted mean 238 U-206 Pb ages were calculated and the uncertainties reported as 95% confidence limits, including incorporation, in quadrature, of the uncertainty in the reference zircon calibration.
Following the U -Pb analyses, the SHRIMP U -Pb pits (1-2 mm deep and c. 20 mm in diameter) were lightly polished away, and oxygen isotope analyses were made in exactly the same location using SHRIMP II fitted with a Cs ion source and electron gun for charge compensation as described by Ickert et al. (2008). Oxygen isotope ratios were determined in multiple collector mode using an axial continuous electron multiplier (CEM) triplet collector, and two floating heads with interchangeable CEM-Faraday Cups. The Temora 2, Temora 3 (unpublished RSES-ANU internal reference standard) and FC1 reference zircons were analysed to monitor and correct for isotope fractionation.  (Ickert et al. 2008). Reproducibility in the Duluth Gabbro FC1 reference zircon d 18 O value varied between analytical sessions, with most of the reference zircon analytical uncertainties in the range 0.32-0.41‰ (+2s). As a secondary reference, the Temora 2 or Temora 3 zircons were analysed in the same sessions, and gave d 18 O values of +8.2‰ and +7.59‰, respectively, in agreement with data reported by Ickert et al. (2008) for Temora 2 and unpublished data for Temora 3.
Lu -Hf isotope measurements were conducted by laser ablation multicollector inductively coupled plasma mass spectroscopy (LA-MC-ICPMS) using the RSES Neptune MC-ICPMS coupled with a HelEx 193 nm ArF Excimer laser, following procedures similar to those described in Munizaga et al. (2008). For all analyses of unknowns or secondary standards, the laser spot size was c. 47 mm in diameter. Laser ablation analyses targeted the same locations within single zircon grains used for both the U-Pb and oxygen isotope analyses described above. The mass spectrometer was first tuned to optimal sensitivity using a large grain of zircon from the Mud Tank carbonatite. Isotopic masses were measured simultaneously in staticcollection mode. A gas blank was acquired at regular intervals throughout the analytical session (every 12 analyses). Typically, the laser was fired with a 5-8 Hz repetition rate and 50 -60 mJ energy. Data were acquired for 100 s, but in many cases only a selected interval from the total acquisition was used in data reduction.
monitor data quality and reproducibility. Signal intensity for unknowns was typically c. 5-6 V for total Hf at the beginning of ablation, and decreased over the acquisition time to 2 V or less.

Results of SHRIMP U -Pb analyses
New SHRIMP zircon U -Pb data for four granite samples (C6-T101, C6-Aw87.3, M6-B248A and K6-Bb47), one mafic gneiss sample (C5-Mj74) and the calcareous argillite (8D27-10) were determined as part of this study. The results are plotted on Tera-Wasserberg concordia diagrams and as probability density plots (with stacked histograms) in Fig. 5; the U -Pb isotope data are provided in Table 2. Populations of zircon were defined on the basis of morphological and cathodoluminescence characteristics (Fig. 4), the details of which are discussed for each new sample below. SHRIMP U -Pb zircon data were previously published for eight of the granites included in this study (Table 1). Table 1 summarizes the U -Pb zircon ages, sample locations and rock types for all samples. Examination of the Hf and O isotope characteristics of the full suite of granite samples (Table 3; Figs S1, S2) will allow an assessment of the degree of crustal reworking v. crustal growth for the two episodes of granitic magmatism, as well as any trends that may be linked to stages in the tectonic evolution.

M6-B248A (two-mica granite)
The heterogeneous zircon population from this sample can be subdivided into two morphological Fig. 4. Cathodoluminescence images of representative zircon grains from granites in the Fosdick complex dated in this study. Ellipses mark the location of U -Pb, O and Hf spot analyses. Cathodoluminescence images were collected on different dates and so exhibit some variation in quality/sharpness. The contrast of images of individual zircons was adjusted to best display internal zoning. Most zircons display core-rim variability in Hf and O isotopes (Supplementary Figs S1 and S2). External reproducibility for Hf isotope measurements is +21 units. types: (a) relatively coarse prismatic moderately zoned grains of moderate brightness in cathodoluminescence, and (b) sub-equant small grains with nonsymmetrical oscillatory growth zones upon rounded inherited cores (Fig. 4). Truncation of sector and oscillatory zoning in the central areas against overgrowths indicates that these areas are inherited zircon components, consistent with 206 Pb -238 U ages of ≥360 Ma for such areas. The majority of the rim areas analysed are notable in that the Th-U ratios are ≤0.10, typical for metamorphic zircon, or zircon formed due to anatexis (i.e. enriched in U, so the apparent low Th -U is due to U enrichment rather than Th depletion as is common in high-grade metamorphic zircon; Hoskin & Schaltegger 2003). The weighted mean 206 Pb -238 U age calculated from nine low Th-U rim analyses is 114.8 + 1.4 Ma (MSWD ¼ 1.4; Fig. 5a, b) and is interpreted as the crystallization age of the granite.

C6-Aw87.3 (biotite -granite)
Zircons from this sample are large prismatic, moderately zoned grains with round terminations (Fig. 4). The grains are dark under cathodoluminescence, with indications that the central dark cathodoluminescence areas have truncated zonation. Six of the areas analysed yield Devonian to Carboniferous ages (Table 2). A weighted mean 206 Pb -238 U age calculated from 14 of 16 analyses of Cretaceous age, both rims and core areas, gives 101.6 + 0.8 Ma (MSWD ¼ 1.4; Fig. 5c, d) and is interpreted as the crystallization age of the granite. It is noteworthy that the majority of Cretaceous areas analysed have Th -U ratios ,0.10, arising more from an enrichment in U (c. 600-900 ppm) rather than a depletion in Th (c. 50-60 ppm).

K6-Bb47 (cordierite -granite)
The dominant zircon population consists of prismatic blocky grains that are moderately dark under cathodoluminescence imaging, with few, wide oscillatory zones, with or without inherited cores (Fig. 4). Inherited central areas, if present, are unzoned and three analysed yield ages ranging from c. 465 Ma to c. 575 Ma (Table 2; not shown on Fig. 5). Zircon areas that yield Cretaceous ages form two distinct 206 Pb-238 U age groupings, one at 104.7 + 0.7 Ma (n ¼ 11, MSWD ¼ 0.6), which is interpreted as the crystallization age of the granite, and 101.6 + 0.9 (n ¼ 7, MSWD ¼ 0.4; Fig. 5e, f ). All these Cretaceous zircon analyses have Th-U ratios of ≤0.12, corresponding to the highest U and lowest Th values of the four granite samples.

C6-T101 (cordierite-leucogranite)
Acicular prismatic zircon grains with oscillatory zoning are the dominant morphology in this sample (Fig. 4). A weighted mean 206 Pb-238 U age of 101.7 + 0.6 Ma (n ¼ 21, MSWD ¼ 0.87) for analyses of grain central areas is considered to date the time of magmatic zircon crystallization (Fig.  5g, h). The zircons have relatively high U and Th concentrations, and a majority of analyses yield elevated Th-U ratios that are in the normal range for igneous zircon (Table 2). Older zircon components yield 206 Pb-238 U ages that range to c. 910 Ma.

C5-Mj74 (mafic gneiss)
Zircon grains from this sample are dominantly stubby, anhedral to subhedral grains that are complexly zoned, resorbed and irregular, with a subordinate population of blocky, prismatic, oscillatory-zoned grains (Fig. 4). This is a heterogeneous population of zircon grains as is evident for the resultant U-Pb ages, which range from c. 865 Ma to Cretaceous. Of 26 analyses of cores and rims, 22 yield Cretaceous ages that range from c. 146 Ma to c. 114 Ma (Fig. 5i, j), with 73% of these between c. 146 Ma and c. 130 Ma in age. However, no significant age grouping or geological age can be drawn from this data set.

8D27-10 (calcareous argillite)
This sample yielded a heterogeneous zircon population that includes small rounded cores (,180 mm) with uniform cathodoluminescence properties (range of luminescence in cores, from dark to bright) and narrow rim overgrowths; large prismatic grains (.225 mm) or grain fragments with oscillatory growth zones, rimmed by a narrow overgrowth that is medium to bright in cathodoluminescence; subangular uniform-coloured cores with a narrow rim overgrowth that is dark in cathodoluminescence. There is truncation of faint sector and oscillatory zoning within the rounded cores against overgrowths, consistent with the derivation of the cores as inherited detrital grains. SHRIMP U -Pb analysis did not yield a coherent age group. Of 25 randomly selected grains, 8 fall in the range c. 552 Ma to c. 675 Ma, with just one analysis that is ,500 Ma (492 Ma). The majority of analysis spots yield older ages that scatter between c. 650 Ma and c. 1150 Ma, with outliers at c. 1.8 Ga and 2.0 Ga (Fig. 5k, l; Table 2).

Results of Hf -O analyses
The Hf and O isotope compositions of zircons from the Ford Granodiorite suite and zircons from granites from the Fosdick complex are summarized in Figures 6 and 7 and Table 3. To characterize both intra-and intersample variations of Hf and O isotope ratios in zircon, the results are presented by granite type (Ford Granodiorite suite; and Devonian-Carboniferous granites and Cretaceous granites within the Fosdick complex). An examination of core-rim relationships in zircons and zircon populations from individual samples is provided in the Supplementary publications (Figs S1, S2).
Magmatic zircons from Ford Granodiorite samples MB.214.W and MB.219.W display a limited range of 1 Hf(t) values from -1 to +4, with d 18 O values ranging from 5.7‰ to 11.5‰ with most between 6.4‰ and 7.1‰ (Figs 6 & 7). The samples are characterized by relatively radiogenic 1 Hf(t) values, low d 18 O signatures, and Proterozoic Lu -Hf depleted mantle model ages (Table 3).
Devonian -Carboniferous granites from the Fosdick complex show a wide range in 1 Hf(t) and d 18 O values, from -7 to +1 and 5.8‰ to 12.6‰, respectively, and are more crust-like than in the Ford Granodiorite suite (Figs 6 & 7). Although some individual spot ages, as well as 1 Hf (t (Figs 6 & 7b).
Cretaceous granites show a comparatively large spread in 1 Hf(t) , with values ranging from -16 to      (Vervoort & Blichert-Toft 1999) and a bulk earth 176 Lu/ 177 Hf value of 0.015 (Goodge & Vervoort 2006). *Uncertainties on t DM are +0.1 Ga, at minimum, and may approach 1 Ga in cases of old inherited grains with younger ages (Vervoort et al. 2011), so the t DM are not here given in Ma.

Discussion
The new U -Pb zircon ages for granites reported here are representative of the two pulses of magmatism recorded in the Fosdick complex at c. 115 Ma and 100 Ma, associated with Cretaceous anatexis and magmatism in the Ross Province during the change from oblique plate convergence to divergence (e.g. Siddoway 2008). Hf and O isotopes in zircon have been used to characterize the putative sources of granites in the Tasmanides and the Western Province of New Zealand (Hawkesworth & Kemp 2006;Kemp et al. 2006aKemp et al. , 2007Bolhar et al. 2008;Tulloch et al. 2009), and the new data presented here allow a similar evaluation to be made for the Ross Province of West Antarctica.
In western Marie Byrd Land, zircons from Ford Granodiorite suite samples have a restricted range of radiogenic 1 Hf(t) values and low d 18 O values (Table 3, Fig. 6a, b), suggesting derivation from a relatively uniform igneous source. In contrast, most granites in the Fosdick complex contain zircons with less radiogenic 1 Hf(t) and higher d 18 O values that are more characteristic of continental realms. For the Devonian -Carboniferous granites, there is no requirement for input from a more juvenile source, for example as a result of underplating during Palaeozoic convergence. However, input from a more juvenile source is required to explain the wider range of 1 Hf(t) values for the Cretaceous granites, and in particular to explain the low d 18 O values for samples C5-R60B and M5-G174. This juvenile source may reflect mafic magmatism associated with attenuation of the subcontinental lithospheric mantle during Mesozoic intracontinental extension (Siddoway 2008) due to ridge-trench interaction (Weaver et al. 1994;Fig. 6. Histograms and probability density plots of d 18 O (zircon) and zircon 1 Hf(t) results from the Ford Granodiorite suite (a, b), from the Ford Ranges and from Devonian -Carboniferous granites (c, d), and from Cretaceous granites (e, f) from the Fosdick complex. Luyendyk 1995), or arising from other causes (Saito et al. 2013).

Infracrustal v. supracrustal source
Previous work based on the Nd-Sr whole-rock isotope composition of a relatively small number of samples proposed that the granites exposed in the Fosdick complex were derived by partial melting of a mixed source comprising Ford Granodiorite suite and Swanson Formation protoliths ( Fig. 2; Korhonen et al. 2010b). In the present study, this petrogenetic model is tested using Hf and O isotope ratios in zircons from granites, which should provide a higher-resolution record of mixing between the putative supracrustal and infracrustal sources and may provide information about any additional juvenile source requirement (Hawkesworth & Kemp 2006).
Both the Devonian-Carboniferous and the Cretaceous granites are inhomogeneous with respect to d 18 O in zircon, and both have relatively high d 18 O values for most samples, indicative of a predominantly crustal derivation (Fig. 6). In addition to primary igneous zircon, a majority of the inherited Palaeozoic cores from zircon within Cretaceous granites (five samples) are significantly more crust-like, with higher d 18 O values than those obtained from Ford Granodiorite samples (Supplementary Fig. S2). However, it is ambiguous whether the inherited cores were sampled from the source of the magma or incorporated from the wall rock during magma ascent and emplacement. Hf isotope ratios in zircons from the Devonian-Carboniferous and the Cretaceous granites are less radiogenic than zircons from the Ford Granodiorite suite, and zircons from Cretaceous granites in particular show a comparatively large spread in 1 Hf(t) values in comparison with the Devonian-Carboniferous granites (Figs 6 & 7). These features suggest that both the Devonian -Carboniferous and Cretaceous granites were derived from more than one isotopically distinct source, and may also implicate a more juvenile source component for the Cretaceous granites.
The Hf-O isotope ratios of zircons from two Ford Granodiorite suite samples and the Hf isotope values of zircons from sample 8D27-10, a calcareous argillite that has a population of detrital zircons similar to that from the Swanson Formation metaclastic rocks, are taken to be representative of these source rocks at depth.  (Valley et al. 1994;Valley 2003), are taken to be representative of the Swanson Formation in general. This is justified because whole-rock d 18 O values reported from the Greenland Group are similar to values expected for most sedimentary rocks (12‰; O'Neil & Chappell 1977). Although there are 1 Hf(t) zircon data from the Proterozoic basement rocks that crop out at the Haag Nunataks in West Antarctica (Flowerdew et al. 2007), in the absence of d 18 O data this representative of a potential Proterozoic source cannot be evaluated adequately in this study.
Zircons from Devonian -Carboniferous granites preserve 1 Hf(t) values intermediate between values expected for zircons that would have crystallized from anatectic melt derived solely from either the Swanson Formation or the Ford Granodiorite suite (Fig. 6d). In contrast, zircons from the Cretaceous granites preserve a wider range of 1 Hf(t) values than those from the Devonian -Carboniferous granites, although the pooled average 1 Hf(t) value of all zircons from Cretaceous granites is statistically indistinguishable from the 1 Hf(t) value of zircons from the Ford Granodiorite suite samples recalculated to 100 Ma (Fig. 6f). Similarly, the majority of d 18 O zircon values from Devonian-Carboniferous granites are intermediate between values expected for zircons that would have crystallized from anatectic melt derived solely from either the Ford Granodiorite suite or the Greenland Group, a proxy for the Swanson Formation (Fig. 6c). A majority of zircon d 18 O values from the Cretaceous granites also lie between the zircon d 18 O values of these two putative sources, but the range extends to lower values and overlaps the zircon d 18 O data from zircons of the Ford Granodiorite suite (Fig. 6e). The Hf and O isotope signatures of zircons from both the Devonian -Carboniferous and the Cretaceous granites therefore require the input of material derived from both supracrustal and infracrustal sources.
The large spread in isotope ratios of the magmatic zircon populations, particularly for the Cretaceous granites, which is greater than the range of values for inherited zircon grains, suggests opensystem behaviour and the involvement of Hf and O from an external source (Kemp et al. 2005(Kemp et al. , 2007Lackey et al. 2005;Yang et al. 2007;Zheng et al. 2007;Bolhar et al. 2008). In other words, the Hf and O isotope characteristics are not indicative of closed-system dissolution and precipitation within the Fosdick complex (cf. Flowerdew et al. 2006;Villaros et al. 2012). Also, the lack of homogeneous Hf and O isotope signatures for zircons from the granites is another indication of opensystem melting of deeper fertile crust, with extraction of the melt to higher crustal levels, which is an effective process of crustal differentiation (Brown 1994(Brown , 2007Solar et al. 1998).

Binary mixing: the nature of the end-members
The Hf and O isotope signatures of granites from the Fosdick complex, which indicate open-system behaviour and more than one source, potentially provide constraints on the amount and nature of crustal reworking in the Ross Province of Marie Byrd Land. The Hf -O isotope composition of an anatectic granite can be modified from that of its source by (1) hybridization with granite that has a different isotope composition and/or (2) incorporation of suprasolidus or subsolidus material with a different isotope composition.
These alternative scenarios are investigated for both Devonian -Carboniferous and Cretaceous granites using binary mixture modelling between the two putative end-member sources, as identified based on the Nd and Sr isotope characteristics of similar granites from the Fosdick complex ( Fig. 2; Korhonen et al. 2010b), but with an additional juvenile source in the case of the Cretaceous granites. We build upon the approach of Kemp et al. (2006aKemp et al. ( , 2007 by integrating the Hf -O data from zircons with the results from phase equilibria modelling (Korhonen et al. 2010a). The exposed Ford Granodiorite end-member represents a relatively juvenile source that contains zircons with relatively radiogenic 1 Hf(t) values and relatively low d 18 O values, whereas the Swanson Formation endmember represents a supracrustal source with relatively non-radiogenic 1 Hf(t) and higher d 18 O. The additional juvenile source is inferred to have more radiogenic 1 Hf(t) and lower d 18 O values than the Ford Granodiorite end-member.
Phase equilibria modelling predicts that both the Swanson Formation and the Ford Granodiorite would have reached P-T conditions above their solidi during both the Devonian -Carboniferous and Cretaceous events, with high-grade metamorphism causing two separate episodes of partial melting of the Swanson Formation and Ford Granodiorite end-members (Korhonen et al. 2010a(Korhonen et al. , 2012. Modelling of seven Swanson Formation and four Ford Granodiorite compositions indicates that Devonian -Carboniferous metamorphism could have yielded 4-40 mol% melt from the Swanson Formation and 2-7 mol% melt from the Ford Granodiorite suite at the level exposed in the Fosdick complex (Korhonen et al. 2010a, Yakymchuk unpublished data; mol% in the modelling is approximately equal to vol% in nature). Modelling further suggests that the subsequent Cretaceous metamorphism could have produced 8-48 mol% melt from fertile Swanson Formation compositions and 3-17 mol% melt from fertile Ford Granodiorite suite compositions at the level exposed in the Fosdick complex, but less from protoliths meltdepleted from the Devonian -Carboniferous event (Korhonen et al. 2010a, Yakymchuk unpublished data). Thus, on the one hand, the amount of melt produced by anatexis of the Swanson Formation at the level exposed in the Fosdick complex generally exceeded the melt connectivity transition (c. 7 vol%, Rosenberg & Handy 2005) and was probably mostly lost to shallower crustal levels (Korhonen et al. 2010a(Korhonen et al. , b, 2012, whereas the amount of melt produced by anatexis of the Ford Granodiorite suite at the level exposed in the Fosdick complex may not have exceeded the melt connectivity transition and most likely was trapped in source. On the other hand, the volumetric proportion of granite exposed in the Fosdick complex demonstrates that melt derived from deeper in the crust was trapped locally during ascent to shallower crustal levels (cf. Brown 2010).
Alternative mechanisms investigated using binary mixture modelling are mixing between liquids derived from the partial melting of the Swanson Formation and Ford Granodiorite suite, and mixing between liquid derived from one source and solid material from the other source. The latter process is interpreted to represent the incorporation of suprasolidus source rock or solid or suprasolidus wall rock into migrating melt. In addition, for the Cretaceous granites, mixing is considered between liquids derived from the partial melting of the Swanson Formation or Ford Granodiorite and a more juvenile source, and mixing between liquid derived from the juvenile source and solid Swanson Formation or Ford Granodiorite.
In this study, the Hf concentrations of the Ford Granodiorite and Swanson Formation are averages taken from bulk rock values from the low-grade protoliths exposed outside the Fosdick complex (Korhonen et al. 2010b). The D Hf values between silicate phases and melt are calculated from the mineral mole proportions obtained by phase equilibria modelling of both the Ford Granodiorite suite and Swanson Formation end-member sources for a pressure of 0.7 GPa and a temperature of 8208C (Korhonen et al. 2010a). At these conditions, a representative Swanson Formation composition is predicted to yield 27 mol% melt and a representative Ford Granodiorite composition is expected to produce 3 mol% melt. These melt proportions are combined with Hf distribution coefficients taken from Rubatto & Hermann (2007) for garnet and zircon and from the GERM database (http:// www.earthref.org/GERM) for other phases to estimate the concentrations of Hf in the model melt. Zircon modes are estimated using Zr and Hf concentrations (e.g. Wark & Miller 1993) from bulk rock analyses of the Swanson Formation and Ford Granodiorite suite (Korhonen et al. 2010b). This zircon modal estimate represents an upper limit because some Zr may have been partitioned into major phases, such as garnet (Villaseca et al. 2003), during high-grade metamorphism. The concentration of Hf in melt derived from deeper in the Fosdick complex will be different because the melt volume produced is expected to have been larger. However, the mixing envelopes shown in Figures  8 and 9 are relatively insensitive to changes in the amount of melt considered in the calculations.
The d 18 O value of the two end-members used in the binary mixing calculations represent the range of measured isotope values of zircon from the Ford Granodiorite samples and a fictive range of isotope values for zircon from the Swanson Formation adjusted for oxygen isotope fractionation. Because no d 18 O whole rock values from the Swanson Formation are available, the range of whole rock d 18 O values of the correlative Greenland Group in New Zealand (13.7-16.2‰) is used (Tulloch et al. 2009). These values are within the range documented for siliciclastic sediments (10-20‰; Eiler 2001). Because of zircon/melt O isotope fractionation, the d 18 O value of zircon that has crystallized from a granitic melt will be 1-2‰ less than that of the melt (Valley et al. 1994;Valley 2003). Using a zircon/melt d 18 O fractionation factor (D 18 O Zrc-WR ) of -2‰, zircons that have crystallized from a melt derived from partial melting of the Greenland Group (assumed to be similar to the Swanson Formation) are expected to yield d 18 O values of 11.7-14.2‰ with an average value of 12.9‰. In addition, a potential juvenile source is added as a third end-member. The 1 Hf(t) range of a potential juvenile source is constrained by the 1 Nd values for mafic dykes from Marie Byrd Land (Saito et al. 2013) using the correlation for juvenile mantle-derived rocks of Vervoort & Blichert-Toft (1999). The potential juvenile source is inferred to have an average d 18 O value similar to that for zircons from mantle-derived magmas (5.7 + 0.3‰; Hawkesworth & Kemp 2006) and a Hf concentration of 4.6 ppm -the average concentration for Cretaceous mafic rocks in Marie Byrd Land (Storey et al. 1999).
The ranges of d 18 O and 1 Hf values for zircon from the end-members are plotted together with individual zircon analyses from the Devonian -Carboniferous granites (Fig. 8) and the Cretaceous granites (Fig. 9). A high and low 1 Hf value for each end-member, considered to be representative of the spread in values, is connected with the calculated mixing curves for the scenarios discussed above. This procedure takes into account the spread of d 18 O and 1 Hf values of each end-member, and, in principle, mixing may occur between batches of melt derived from sources with d 18 O and 1 Hf values that vary between the mixing curves shown in Figures 8 and 9.

Binary mixture modelling: results
The approach presented above is a necessary simplification of a naturally complex anatectic system. However, it does provide important insights into the nature and relative proportions of source components and the mechanisms of mixing. Points along the model curves in Figures 8 and 9 represent the isotope signature of zircon that would crystallize from magma that records mixing between two end-members in various proportions. The sets of mixing curves between the Swanson Formation and the Ford Granodiorite suite calculated for the scenarios discussed above yield similar results and are separated from one another by less than three 1 Hf units. Less than three 1 Hf units also separate the calculated binary mixing curves between liquid derived from the juvenile source and material derived from either the Swanson Formation or the Ford Granodiorite suite. The close spatial association of these mixing curves on the Hf -O plot does not allow us to determine the physical process by which mixing occurred (Figs 8 & 9), but these curves may be used to evaluate the relative contribution of each source to the isotope signature of the zircons from granites in the Fosdick complex.
For the Devonian -Carboniferous granites, most individual zircon 1 Hf(t) and d 18 O values plot within the calculated mixing envelope. Exceptions include two analyses from M5-G175 and one from C5-Is51A. Samples C5-I26 and C6-Aw86.1 define tightly clustered groups, whereas samples C5-Is51A and M5-G175 contain comparatively large spreads in d 18 O and 1 Hf(t) zircon values (Fig. 8a). The zircon d 18 O values from samples C5-I26 and C6-Aw86.1 plot closer to the Ford Granodiorite source, whereas the 1 Hf(t) values of zircons from these samples are similar to the most radiogenic values from detrital zircons in the Swanson Formation (Fig. 8a, b). The majority of zircon data from the Devonian -Carboniferous samples are consistent with the earlier Sr-Nd Hf isotope values are plotted for zircons from sample 8D27-10, a calcareous argillite rock that contains detrital zircons with a U-Pb age distribution similar to that of Swanson Formation metaclastic rocks (Fig. 3a), as a means to bracket the range of 1 Hf(t) values for the Swanson Formation. Crosses on the mixing curves are 10 vol% increments and represent the isotope signature of zircon that would crystallize from magma that records mixing between the two putative end-members in various proportions. All zircon 1 Hf(t) values have been recalculated to 350 Ma. The cross in the upper right of each diagram shows approximate +2s uncertainties for Hf and O isotope values. Three sets of curves are calculated to evaluate (1) mixing between melts derived by anatexis of the Swanson Formation (Hf ¼ 3.9 ppm) and the Ford Granodiorite suite (Hf ¼ 5.0 ppm), (2) mixing of melt derived by anatexis of the Ford Granodiorite suite with solid or suprasolidus Swanson Formation (Hf ¼ 3.2 ppm), and (3) mixing of melt derived by anatexis of the Swanson Formation with solid or suprasolidus Ford Granodiorite suite (Hf ¼ 3.4 ppm). An 1 Hf range of a potential juvenile source is constrained by the 1 Nd signature of mafic dykes from Marie Byrd Land (Saito et al. 2013) using the correlation for juvenile mantle-derived rocks of Vervoort & Blichert-Toft (1999). The potential juvenile source is assigned an average d 18 O value of zircons from mantle-derived magmas (5.7 + 0.3‰; Hawkesworth & Kemp 2006 isotope results that suggest these granites were derived from a two-component mixture of material from the Swanson Formation and the Ford Granodiorite suite, with the latter being the dominant component ( Fig. 2; Korhonen et al. 2010a). These results are also consistent with phase equilibria modelling, which suggests that much of the melt derived from anatexis of the Swanson Formation was likely to have been lost to shallower crustal levels than exposed in the Fosdick complex, whereas the Devonian -Carboniferous granites within the Fosdick complex were likely to have been dominated by melts derived from the Ford Granodiorite suite trapped during ascent from deeper in the crust where this protolith is expected to occur in greater abundance than the supracrustal rocks (Korhonen et al. 2010a). Zircon analyses for samples C5-I26 and C6-Aw86.1 plot in a region of the mixing envelope consistent with 30 -50% involvement of Swanson Formation material (Fig. 8a, b). The large spread in values for samples C5-Is51A and M5-G175, including some outside the mixing envelope, may suggest that there was significant isotope heterogeneity Crosses on the mixing curves are 10 vol% increments and represent the isotope signature of zircon that would crystallize from magma that records mixing between two of the putative end-members in various proportions. All zircon 1 Hf(t) values have been recalculated to 100 Ma. Seven sets of curves are calculated to evaluate (1) mixing between melt derived from partial melting of the Swanson Formation (Hf ¼ 3.9 ppm) and the Ford Granodiorite suite (Hf ¼ 5.0), (2) of melt derived by anatexis of the Ford Granodiorite suite with solid or suprasolidus Swanson Formation (Hf ¼ 3.2 ppm), (3) mixing of melt derived by anatexis of the Swanson Formation with solid or suprasolidus Ford Granodiorite suite (Hf ¼ 3.4 ppm), (4) mixing of melt derived from a juvenile source (Hf ¼ 4.6 ppm) mixing with solid or suprasolidus Swanson Formation, (5) mixing of melt derived from a juvenile source with melt derived from anatexis of the Swanson Formation, (6) mixing of melt derived from a juvenile source mixing with solid or suprasolidus Ford Granodiorite suite, and (7) mixing of melt derived from a juvenile source with melt derived from anatexis of the Ford Granodiorite suite. Also plotted are the isotope values for zircons from sample 8D27-10, an argillaceous rock that contains detrital zircons with a U-Pb age distribution similar to that of Swanson Formation metaclastic rocks (Fig. 3a)  within the magma, and for this reason the zircon analyses from these samples were not used to evaluate the proportional contribution of each putative end-member component.
Binary mixing curves for various mixing scenarios for zircons from the Cretaceous granites are shown in Figure 9. A majority of zircon data from samples C6-Aw87.3, C6-T101, C6-BB112 and M6-B248A lie within or along the radiogenic 1 Hf boundary of the mixing envelope and, again, data plot closer to the Ford Granodiorite source, which suggests that it provided the dominant contribution for many of the granites (Fig. 9a, b). Zircons from these samples plot in a region of the mixing envelope consistent with incorporation of 20-60% Swanson Formation material in melt derived from a Ford Granodiorite source. Zircons from sample K6-Bb47 also plot within the mixing envelope but they contain less radiogenic 1 Hf(t) values than the majority of zircons from other granites. This suggests that a larger proportion of supracrustal material contributed to the isotope composition of this granite compared to other samples, which corroborates the interpretation by Korhonen et al. (2010b) that this granite was derived predominantly from partial melting of the Swanson Formation.
Samples C5-Is54, M6-L188B, C5-R60B and M5-G174 contain zircons with radiogenic 1 Hf(t) values that are similar or greater than zircons from the Ford Granodiorite suite, which points to the involvement of a more juvenile source that is not exposed at the surface. With the exception of two zircon analyses from sample C5-Is54, the remaining zircons in this sample and zircons from sample C5-R60B define a spread in 1 Hf(t) values from +4 to -3 but a limited range in d 18 O from 6.5‰ to 8.5‰, which suggests a much smaller contribution of supracrustal material compared with the other granites analysed from the Fosdick complex. This may indicate mixing between the Ford Granodiorite suite and melt derived from a deep-seated juvenile source (Fig. 9c, d) not exposed in the Ford Ranges. Sample M6-L188B contains zircons that yield Hf and O isotope values within the range expected for the Ford Granodiorite suite, but also contains several analyses with higher d 18 O values and some with more radiogenic 1 Hf(t) values. The spread in values recorded by this sample suggests the involvement of the Ford Granodiorite suite, the Swanson Formation, and an unexposed juvenile source. The addition of a relatively juvenile source to the mixing models for the Cretaceous granites constrains most of the data that display more radiogenic values than zircons from the Ford Granodiorite suite, including those of sample M5-G174, representing the array of felsic dykes that crosscut all migmatite structures (Siddoway et al. 2005) and the South Fosdick detachment zone. Binary mixing between the juvenile source and the Ford Granodiorite suite encompasses most of the data for sample C5-R60B, whereas a ternary mix of all three putative sources is required to constrain the majority of data from samples C5-Is54 and M6-L188B (Fig. 9c, d).
What are the petrogenetic implications?
The Hf isotope signature of zircons from most Devonian -Carboniferous and Cretaceous granites in the Fosdick complex requires a large component derived from the Ford Granodiorite source or for some of the Cretaceous granites from a more juvenile source, but many of the granites also have relatively high d 18 O values, indicating the involvement of supracrustal material. This could reflect magma mixing and/or the incorporation of supracrustal material into migrating melt. The results of the binary mixture modelling suggest that a magma derived by partial melting of the Ford Granodiorite source and/or a more juvenile source must incorporate 20 -60 vol% solid or suprasolidus Swanson Formation material to comply with the isotope constraints (Figs 8 & 9). Similar studies from the Lachlan Belt of Australia suggest that 40 -85 vol% crustal material was incorporated into granites derived from more juvenile sources to produce the observed Hf-O isotope signatures (Kemp et al. 2006a(Kemp et al. , 2007. However, whether felsic magma can incorporate this much material by processes such as melting and disaggregation is uncertain. The incorporation of subsolidus or suprasolidus material into magma is an energy-intensive process that may be self-limiting (e.g. Koyaguchi 1986;Bowen 1922;Glazner 2007). Thermodynamic modelling of xenolith assimilation suggests that mafic magma can only incorporate a few tens of percent of warm (400 8C) granite (Glazner 2007). The upper limit of this estimate is suppressed for more felsic melts due to their lower temperatures compared with basaltic melts. Granites now exposed in the Fosdick complex were emplaced at mid-crustal levels where ambient temperatures were up to c. 870 8C in the Carboniferous and the Cretaceous (Korhonen et al. 2010a), and both the Swanson Formation and the Ford Granodiorite suite are inferred to have been partially molten at the time of granite emplacement.
The Ford Granodiorite suite is predicted to have produced c. 2-7 mol% melt at the crustal level exposed (Korhonen et al. 2010a). However, at slightly deeper structural levels, the Ford Granodiorite suite could have produced significantly more melt through hydrate-breakdown melting, sufficient to be extracted and emplaced higher in the crust, including the present exposure level of the Fosdick complex (Korhonen et al. 2010a, b). The temperatures beneath the Fosdick complex are interpreted to have exceeded the stability of biotite (Korhonen et al. 2010a), and the Swanson Formation at depth was likely to be limited in volume and residual in nature and is not expected to have contributed much melt from deeper structural levels. Geophysical models based on airborne magnetic and gravity surveys across the Ford Ranges indicate that the rocks underlying the Fosdick complex are similar to plutonic rocks exposed elsewhere in the Ford Ranges (Ferraccioli et al. 2000;Luyendyk et al. 2003). Therefore, the putative plutonic crust beneath the Fosdick complex, which is probably dominated by the Ford Granodiorite suite, is interpreted to represent the primary source of melts making up the Devonian -Carboniferous and many of the Cretaceous granites now exposed in the Fosdick complex.
Isentropic ascent of melt sourced from the Ford Granodiorite suite would allow it to become superheated (Stolper & Asimow 2007) and capable of incorporating more material than would be possible in the source. At the level of emplacement for this melt, most of the melt generated from the Swanson Formation is inferred to have been lost to higher structural levels to allow preservation of the high-grade mineral assemblages in the paragneisses (Korhonen et al. 2010a). However, some residual melt would have been retained along grain boundaries (Holness & Sawyer 2008), with the amount being dependent on the percolation threshold for the microstructure of the melt-bearing rock (Cheadle et al. 2004). Thus, migrating moredeeply sourced melt has the potential to incorporate suprasolidus Swanson Formation rocks by disaggregation.
Melt produced through anatexis may be extracted as a single pulse (batch melting) or more likely it may be extracted as several smaller pulses (fractional melting). A consequence of this process is that melt pulses extracted from an initially isotopically homogeneous source may evolve different isotope compositions due to isotope fractionation of Lu from Hf. The Lu -Hf ratios are expected to change if melt is extracted from a source that contains garnet or zircon (Vervoort & Patchett 1996;Hawkesworth & Kemp 2006). This process is particularly important in high-grade polymetamorphic terranes, such as the Fosdick complex, where multiple episodes of melt production and extraction have been suggested (Korhonen et al. 2010a(Korhonen et al. , b, 2012. Modelling of melting in the lower crust suggests that residual rocks can develop an anomalously high 176 Hf/ 177 Hf signature within 300 -400 million years (Vervoort & Patchett 1996). This time frame is too long for significant differences in the Hf isotope signature to develop in residual material in the Fosdick complex. Although fractional melting may have played an important role in producing the granites in the Fosdick complex, the effects cannot be resolved using Hf isotopes, and any modification of the 176 Hf/ 177 Hf values recorded in zircon due to fractional melting is probably insignificant when compared with the Hf isotope heterogeneity of zircons in the putative protoliths.

Temporal trends in the zircon Hf and O isotope characteristics
Devonian -Carboniferous granites contain zircons that display less radiogenic 1 Hf(t) and higher d 18 O values than zircons from the Ford Granodiorite suite (Fig. 10a, b), which is consistent with the previously proposed two-component mixing model (Korhonen et al. 2010b). The lowest d 18 O values come from sample M5-G175 -the oldest of the Devonian -Carboniferous granites collected from the Fosdick complex -in which zircons have highly variable Hf and O values suggesting that the melts were not well homogenized (Fig. 10a). Zircons from the three younger Devonian -Carboniferous granites contain significantly elevated d 18 O values that not only exceed those of the Ford Granodiorite suite zircons but also show a tendency towards higher d 18 O values with decreasing age (Fig. 10a). This is interpreted as an indication of an increasing contribution of Swanson Formation supracrustal material in the granite magmas during the evolution of the Devonian -Carboniferous melting event.
Excluding samples C5-R60B and M5-G174 (discussed in the following), the data from Cretaceous granites also show a slight tendency towards higher d 18 O values with decreasing age (Fig. 10c). Results from zircon in granites emplaced during wrench deformation (c. 116-115 Ma; C5-Is54, M6-L188B and M6-B248) are slightly more variable, with a wide range of d 18 O values (6.2-11.6‰, most ≥7.5‰), and the lowest d 18 O values for zircons from Cretaceous granites excluding samples C5-R60B and M5-G174 (Fig. 10c). Zircons from younger granites emplaced during transtension (c. 109-102 Ma; K6-Bb47, C6-BB112, C6-T101 and C6-Aw87.3) have more elevated d 18 O on average and less radiogenic 1 Hf(t) values (Fig. 10c,  d), reflecting a greater contribution from the supracrustal Swanson Formation. This generally supports the petrogenetic modelling of Korhonen et al. (2010aKorhonen et al. ( , 2012, which postulated that the younger, detachment-controlled sheeted leucogranite complex was derived primarily from a Swanson Formation source. The Cretaceous granites were emplaced during a transition in tectonic regimes (McFadden et al. 2010a, b), which may have affected the ability of melt to escape the source region now exposed in the Fosdick complex. On the one hand, granites emplaced during wrench deformation form steeply dipping to vertical sheets (McFadden et al. 2010a, b) that would have provided conduits for melt migration from and through the exposed level of the Fosdick complex to shallower crustal levels, with only a low probability of entrapment. The 1 Hf(t) and d 18 O values of zircons from these granites suggest that a high proportion of the melt from which they crystallized was derived from plutonic rocks, which may include an unexposed juvenile source responsible for the relatively radiogenic 1 Hf(t) values in these older Cretaceous granites (Fig. 10d), consistent with a deeper source with only a limited Swanson Formation component. On the other hand, granites emplaced during transtension form subhorizontal sheets (McFadden et al. 2010a, b). The less radiogenic 1 Hf(t) and high d 18 O values of zircons in these granites suggest they include a larger proportion of melt derived from the Swanson Formation, consistent with a higher degree of retention of locally derived melt due to the change from steep to shallow transport channels. Therefore, one explanation for the secular trend away from more juvenile 1 Hf(t) and lower d 18 O values in the older Cretaceous granites to less radiogenic 1 Hf(t) and high d 18 O values in the younger Cretaceous granites is the change from wrench to transtensional tectonics and the consequent effect on trapping v. escape of melt (McFadden et al. 2010b;cf. Scott & Cooper 2006).
The isotope compositions of granites C5-R60B and M5-G174 are distinctly different from the other Cretaceous granites. C5-R60B was emplaced directly into the South Fosdick detachment zone at 102.4 + 0.7 Ma and records the most radiogenic 1 Hf(t) values obtained from any granite studied and has low d 18 O values (Fig. 10c, d). M5-G174, one in an array of leucocratic dykes that crosscut the gneissic foliation in the central Fosdick complex, exhibits low d 18 O values, overlapping those from the Ford Granodiorite suite, and heterogeneous 1 Hf(t) values ranging from -16 to -1, with most between -8 and -1 (Fig. 10c, d). The appearance of a granite with lower d 18 O zircon values and, in one case, with juvenile 1 Hf(t) values within the South Fosdick detachment zone may record a more radiogenic, mantle-like source brought in by lithospheric thinning induced by the change from wrench to transtensional tectonics in the region (McFadden et al. 2010a, b). A change in chemistry is also documented for older v. younger mafic dykes intruded over approximately the same time interval as the Cretaceous granites (c. 113 to c. 98 Ma, based on LA-ICP-MS U-Pb zircon ages for the dykes; Saito et al. 2013). The mafic dykes have positive 1 Sr and negative to slightly positive 1 Nd values (calculated at 100 Ma), consistent with derivation from a more enriched mantle source that is interpreted to be a metasomatized sub-arc mantle that underwent decompression melting during intracontinental extension (Saito et al. 2013).
Comparison with the Antarctic Peninsula, the Tasmanides of Australia and the Western Province of New Zealand Convergent continental margins represent the primary locus for the production of juvenile crust and its differentiation into stable continental crust (Brown & Rushmer 2006). The Fosdick complex within Ross Province represents a segment of one of the most extensive and long-lived convergent plate margins of the Phanerozoic, the active margin of Gondwana. The landmasses of Australia, New Zealand (Zealandia) and West Antarctica ( Fig. 1a; Gibson & Ireland 1996) were contiguous during the Lachlan phase (c. 485 -340 Ma) of Tasmanide orogenesis.
In eastern Australia, the Tasmanide orogen is attributed to alternating extensional and contractional tectonics associated with a west-dipping subduction zone that migrated oceanward from the Cambrian to the Permian (Foster & Gray 2000;Collins 2002;Gray & Foster 2004;Glen 2005;Cawood 2005;Foster et al. 2005). The Tasmanides of eastern Australia occupied a relatively inboard position along the active margin from the Devonian to the Cretaceous. Two to three thousand kilometres east (present coordinates) of the Tasmanides along the active margin of Gondwana, the Western Province of New Zealand and the Ross Province of West Antarctica occupied more outboard and more inboard positions, respectively, and both record Devonian -Carboniferous calc-alkaline and Cretaceous alkaline plutonism (Tulloch et al. 2009. A protracted period of Silurian to Devonian silicic magmatism is recorded in the Tasmanides with minor episodes of Carboniferous magmatism. In contrast, granites from the Western Province of New Zealand and the Ross Province were emplaced over relatively short periods in the Devonian -Carboniferous and the Cretaceous (Adams 1987;Siddoway & Fanning 2009).
Because all three regions record silicic magmatism in the Phanerozoic and they are spatially distributed along the Gondwana margin (Fig. 1a), a comparison of the Nd and Hf isotope signatures of granites from the three regions may be used to evaluate variations in the extent of juvenile and evolved crustal sources along and across the strike of the former active margin of Gondwana. Figure 11 is an evolution diagram comparing whole rock 1 Nd and zircon 1 Hf data for igneous rocks from the three regions during the Palaeozoic and Mesozoic.
Granites from the Ross Province show nonradiogenic 1 Hf (Fig. 11) and elevated d 18 O isotope signatures (Fig. 10a, c) within a setting dominated by crustal reworking, as can be expected for the more inboard location of this province in comparison with the Western Province of New Zealand (Fig. 1a). The Western Province of New Zealand includes Devonian -Carboniferous and Cretaceous granites with 1 Nd signatures indicative of both a more juvenile source and a source similar to the Ross Province granites (Tulloch et al. 2009), although granites with positive 1 Nd and 1 Hf values dominate, in strong contrast to those from Ross Province (Fig. 11). These isotope signatures are consistent with significant Phanerozoic crustal growth of Zealandia, which was located further outboard in an active continental arc setting, with significant differences in the source and style of contemporaneous magmatism. The zircon Hf isotope and whole-rock Nd signatures in the Western Province reflect partial melts derived from a more juvenile source (e.g. a basaltic underplate) that incorporated smaller volumes of crustal materials than did partial melts in the Ross Province.
In contrast to the episodic history of granite emplacement in the Ross Province and the Western Province of New Zealand, granites from the Antarctic Peninsula and Thurston Island record protracted magmatism from the Carboniferous through the Cretaceous with relatively non-radiogenic whole rock 1 Nd and zircon 1 Hf signatures. An exception is a Late Silurian orthogneiss from Mt Eissenger on the Antarctica Peninsula, a sample that contains zircon cores with juvenile 1 Hf values (Fig. 11;Flowerdew et al. 2006). The Tasmanides contain relatively few Palaeozoic igneous rocks of comparable age to those in Ross Province and Western Province of New Zealand, and Devonian granites in the Tasmanides generally have more juvenile 1 Hf and 1 Nd values than the Devonian -Carboniferous granites in Ross Province.
Hafnium isotope-time trends for igneous rocks from the Tasmanides show several linear arrays that indicate a progression from relatively nonradiogenic to radiogenic sources that are corroborated by whole-rock Nd isotope data (Kemp et al. 2009). Relatively high d 18 O zircon values from several suites of granites show that they contain a supracrustal component, whereas Hf -O binary mixture modelling suggests assimilation of 40 -85 vol% crustally derived material into a melts derived from a more juvenile source (Kemp et al. 2006a, b). Thus, the c. 300 million year span of magmatism in the Tasmanides involved both crustal reworking and the addition of new crustal material Fig. 11. Compilation of whole rock Nd (a) and zircon Hf (b) isotope data for igneous rocks from the eastern Gondwana margin. Sources of Nd data include McCulloch et al. (1987), Pankhurst et al. (1993Pankhurst et al. ( , 1998, Muir et al. (1995Muir et al. ( , 1996Muir et al. ( , 1998 (Vervoort & Blichert-Toft 1999) and Nd (DePaolo 1981). from juvenile sources. The significant contributions from more juvenile sources are attributed to periodic extension phases punctuated by short intervals of contraction in Palaeozoic times (e.g. Collins 2002;Kemp et al. 2009). Only very subtle secular trends over short time intervals (c. 20 myr) are observed in the granites from Ross Province, with a possible progression towards a greater contribution of supracrustal material with decreasing age in both the Devonian -Carboniferous and Cretaceous granites (Fig. 10).
These results highlight the significant differences in crustal production and reworking that may occur both along (Antarctic Peninsula, Ross Province and the Tasmanides) and across (Ross Province and Western Province) a former active continental margin arc system. It is therefore important to evaluate the relative position and tectonic style of different segments of extensive ancient continental margin arc systems in order to test tectonic reconstructions and models of supercontinent evolution.

Conclusions
Hf and O isotope data from zircon separates from samples from the Fosdick complex are used to investigate the petrogenesis of granites within the complex. Devonian-Carboniferous granites (370-355 Ma) were formed by crustal melting and mixing of either melts or melts and residual materials derived from the calc-alkaline Ford Granodiorite suite and the metasedimentary rocks of the Swanson Formation. In contrast, most Cretaceous granites (116-115 Ma and 105 -96 Ma) require an unexposed more juvenile source in addition to the same crustal sources as the Devonian-Carboniferous granites. These results demonstrate that Devonian-Carboniferous granites in the Fosdick complex essentially record crustal reworking, whereas Cretaceous granites require a component of recent crustal growth to provide the more juvenile source in addition to crustal reworking. The most juvenile 1 Hf(t) and lowest d 18 O zircon values come from the youngest granites emplaced during and after detachment faulting, a possible indication of input from a more radiogenic, mantle-like source. The substantial intrasample variations of Hf and O isotope values in zircon are consistent with the involvement of material derived from isotopically heterogeneous supracrustal and infracrustal sources, as well as a mantle source for the Cretaceous granites, during partial melting and differentiation of the continental crust.
The dominance of crustal reworking suggested by the isotope characteristics of zircons from granites in the Ross Province of West Antarctica contrasts with correlative granite suites across the former active margin of Gondwana, such as those of the Western Province of New Zealand, which occupied a more outboard position with respect to the trench, where more radiogenic 1 Hf isotope signatures in zircon from granites suggest derivation from more juvenile sources and less crustal reworking. The Devonian -Carboniferous granites in the Fosdick complex do not exhibit evidence for protracted magmatism nor for the temporal trends in source evolution that are present in granites from the Tasmanides, which were generated in a similar inboard position from the subduction front during the Devonian -Carboniferous. The results of this study indicate significant variations in the petrogenetic processes producing granite magma along and across the former active margin of Gondwana.