Crustal evolution, intra-cratonic architecture and the metallogeny of an Archaean craton

Abstract The generation of the Earth's continental crust modified the composition of the mantle and provided a stable, buoyant reservoir capable of capturing mantle material and ultimately preserving ore deposits. Within the continental crust, lithospheric architecture and associated cratonic margins are a first-order control on camp-scale mineralization. Here we show that the evolving crustal architecture of the Archaean Yilgarn Craton, Western Australia, played a key role in controlling the localization of camp-scale gold, iron and nickel mineralized systems. The age and source characteristics of Archaean lithosphere are heterogeneous in both space and time and are recorded by the varying Nd isotopic signature of crustal rocks. Spatial and temporal variations in isotopic character document the evolution of an intra-cratonic architecture through time, and in doing so map transient lithospheric discontinuities where gold, nickel and iron mineral systems were concentrated. Komatiite-hosted nickel deposits cluster into camps localized within young, juvenile crust at the isotopic margin with older lithosphere; orogenic gold systems are typically localized along major structures within juvenile crust; and banded iron formation (BIF)-hosted iron deposits are localized at the edge of, and within, older lithospheric blocks. Furthermore, this work shows that crustal evolution plays an important role in the development and localization of favourable sources of nickel, gold and iron by controlling the occurrence of thick BIFs, ultramafic lavas and fertile (juvenile) crust, respectively. Fundamentally, this study demonstrates that the lithospheric architecture of a craton can be effectively imaged by isotopic techniques and used to identify regions prospective for camp-scale mineralization.

The continental crust is a physical and geochemical record of Earth history and is unique within our solar system (Taylor & McLennan 1985;Blichert-Toft & Albarède 1997;Bizzarro et al. 2003;Hawkesworth et al. 2010). Its formation began c. 200 myr after the formation of the planet at around 4.4 Ga (Wilde et al. 2001;Harrison et al. 2005;Blichert-Toft & Albarède 2008), but large continental blocks did not stabilize until after c. 3.2 -3.0 Ga (Smithies et al. 2003;Shirey & Richardson 2011;Naeraa et al. 2012;Dhuime et al. 2012). The formation and stabilization of continental masses changed the composition of the mantle, creating two new geochemical reservoirs: the depleted upper mantle and the complementary enriched crust (Jacobsen & Wasserburg 1979;O'Nions et al. 1980;DePaolo 1981;Allègre et al. 1983;Hawkesworth & Kemp 2006;Hawkesworth et al. 2010). The creation of these reservoirs changed the geochemical budget of the Earth, whereby elements that preferentially partition into silicate melts (incompatible elements) were concentrated in the new crust (Rudnick 1995;Hawkesworth & Kemp 2006).
This new crustal reservoir was dominated by lower-density minerals that made it more buoyant and gravitationally stable than oceanic crust, which being relatively dense is recycled back into the asthenosphere at subduction zones (Carlson & Raskin 1984;Cloos 1993;Hawkesworth et al. 2010;Cawood et al. 2013). As a result, the preservation potential of the new continental crust relative to oceanic material was much greater (Hawkesworth & Kemp 2006;Hawkesworth et al. 2009;Hawkesworth et al. 2010). In addition, the new continental crust had a rheology conducive to capturing and storing mantle-derived magmas that allowed both the generation and the preservation of metallic ore deposits.
The continental crust is not only essential for the capture and storage of ore deposits, but also has a first-order control on their localization and formation at a variety of scales (Cassidy et al. 2002;Begg et al. 2009Begg et al. , 2010. The spatial evolution of the crust and the distribution of Precambrian cratons, Proterozoic fold belts and Phanerozoic crust create a diverse lithospheric architecture (Cawood et al. 2013) that effectively controls the preferential pathways of magmas and fluids as well as sites of structural complexity (Begg et al. 2009). Subsequently, Precambrian cratons and their boundaries are a major control on the location of mineral deposits worldwide (Groves & Batt 1984;Begg et al. 2009Begg et al. , 2010Blewett et al. 2010b). However, there is clearly potential for major mineralization events/ processes within cratonic blocks, as demonstrated by large nickel (Barrie et al. 1993;Houlé et al. 2008;Fiorentini et al. 2012;Houlé et al. 2012), platinum-group element (PGE; Zientek et al. 2002;Maier 2005), gold (Groves & Batt 1984;Robert et al. 2005;Ispolatov et al. 2008;Bateman et al. 2008), base metal (Ashley et al. 1988;Hannington et al. 1999a, b;Cantwell et al. 2009) and iron deposits (Khan & Naqvi 1996;Angerer & Hagemann 2010;Duuring et al. 2012;Angerer et al. 2012a) found within many Archaean cratons (e.g. Yilgarn, Superior, Kaapvaal). Consequently, understanding the internal temporal and spatial evolution of Archaean cratons is important in understanding the localization of Archaean mineral deposits (Ketchum et al. 2008;Blewett et al. 2010b).
To achieve this, Sm-Nd isotope data for Archaean crustal granites and felsic volcanic rocks of the Yilgarn Craton, Western Australia, were used to map spatial variations in the age and source of the crust. Previous Sm -Nd and Lu -Hf isotopic work by Cassidy et al. (2002), Griffin et al. (2004),  and Blewett et al. (2010b) showed that the Yilgarn Craton has an internal crustal architecture consisting of a number of discrete lithospheric blocks of varying age and origin. Several studies have linked this isotopic architecture to the localization of mineral systems (Barley et al. 2003;Begg et al. 2010); however a coherent, multicommodity study has been lacking. The primary aims of this work were to consolidate and extend the Sm-Nd isotope coverage of the Yilgarn Craton, develop a comprehensive understanding of the crustal evolution of the craton in space and time and relate this to the spatial and temporal occurrence of komatiite-hosted nickel, orogenic gold and banded iron formation (BIF)hosted iron systems.
The Yilgarn Craton is one of the largest preserved pieces of Archaean crust on Earth (Cassidy et al. 2006;) and records a history of continental crust from c. 4400 Ma (Jack Hills, Illara and Maynard Hills metasediments; Wilde et al. 2001;Wyche et al. 2004) to 2600 Ma (late, post-tectonic 'cratonizing' granites; Cassidy et al. 2002;, and a supracrustal record dating from 3010 to 2650 Ma (Wilde & Pidgeon 1986;Wang et al. 1996;Nelson 1997;Wang et al. 1998;Witt 1999;Chen et al. 2003;Robert et al. 2005;Kositcin et al. 2008;Ivanic et al. 2010). The Yilgarn Craton is also one of the most intensely mineralized crustal terranes on Earth, with multiple gold, iron and nickel camps (Groves 1993;Groves et al. 1995;Barley et al. 1998Barley et al. , 2003Barnes 2006a, b;Angerer & Hagemann 2010;Angerer et al. 2012a, b;Fiorentini et al. 2012;Duuring & Hagemann 2013a, b). As a result, it represents an excellent case study area to investigate the effects of crustal evolution and intra-cratonic architecture on the localization of multiple mineral systems.

Geological setting and previous work
The Archaean Yilgarn Craton is located in the SW of Western Australia and consists of approximately 70% granite -gneiss and 30% greenstone belts (meta-igneous and meta-sedimentary sequences; Myers 1995;Cassidy et al. 2006). The craton has been divided into a number of terranes and domains by Cassidy et al. (2006) based on detailed stratigraphic, structural, geochemical and geochronological data (see Fig. 1). The Eastern Goldfields Superterrane comprises the eastern half of the Yilgarn Craton and consists of the Yamarna, Burtville, Kurnalpi and Kalgoorlie Terranes (Cassidy et al. 2006;Pawley et al. 2012). The South West, Narryer and Younami Terranes occur to the west of the Ida Fault, and are collectively referred to here as the 'West Yilgarn' superblock. The Youanmi Terrane is further subdivided into the Murchison and Southern Cross Domains (Fig. 1a).
A summary of the preserved granite -greenstone geology of the Yilgarn Craton is presented in Cassidy et al. (2006), Wyche et al. (2012b), Table 1 and Figure 2.
The Yilgarn Craton has been the subject of numerous Sm -Nd isotopic studies, for example McCulloch & Compston (1981), Fletcher & Rosman (1982), McCulloch et al. (1983, McCulloch (1987), Fletcher et al. (1994), Champion & Sheraton (1997), Bateman et al. (2001), Cassidy et al. (2002), Barley et al. (2003), Griffin et al. (2004),  and Wyche et al. (2012a, b), that has led to a large body of spatially constrained data. Cassidy & Champion (2004) and  display this extensive isotopic dataset via contour mapping, demonstrating the significant spatial variation in the crustal history of the Yilgarn Craton. This map demonstrates that the Eastern Goldfields consist of much younger crust (c. 3000 Ma Nd model age) than the West Yilgarn (c. 3500-3300 Ma Nd model age). These two crustal domains are bounded by the Ida Fault, which is interpreted as a crustal-scale structure (Swager 1997;Drummond et al. 2000;Goleby et al. 2006;Dentith et al. 2012) representing the suture between two different lithospheric blocks. Mole (2012), Ivanic et al. (2012) and Wyche et al. (2012b) took the isotopic study of the Yilgarn Craton further by using Lu -Hf isotopes from magmatic and inherited zircons. Wyche et al. (2012b) identified five crustal growth and recycling events at c. 4200, 3500, 3100, 2800 and 2700 Ma, and demonstrated shared, craton-wide magmatism and juvenile input after c. 3000 Ma. Mole (2012) confirmed the significant isotopic difference between the West Yilgarn and Eastern Goldfields Superterrane, and mapped the evolution of the craton from 3050 to 2600 Ma. Cassidy & Champion (2004) and  used Sm-Nd data to produce a spatial analysis of crustal source regions for the entire Yilgarn Craton, although a number of areas had low sample density, particularly the southern Youanmi Terrane, which underwent targeted sampling for this work. Other isotopic studies (e.g. Griffin et al. 2004;Ivanic et al. 2012) focussed on smaller regions within the Yilgarn Craton. Such spatial isotopic work has also been performed in the Superior Craton (Boily et al. 2009) and Albany-Fraser Orogen (Kirkland et al. 2011). Griffin et al. (2004) used Lu-Hf analyses on detrital zircons from modern drainage systems to infer a series of crustal domains in the NW Yilgarn Craton. However, although the use of streamsampled detrital zircons allows the analysis of a large area, the resolution of the spatial component is highly uncertain. Ivanic et al. (2012) used Lu -Hf isotopes on magmatic zircons from granites of the Murchison Domain to better understand the crustal history through time; however, samples from this study were focused within a narrow region.

Granites of the Yilgarn Craton
Geochronology shows that granites (sensu lato) were emplaced during several discrete episodes dating back to c. 3700 Ma (Cassidy et al. 1998(Cassidy et al. , 2002Mole et al. 2012), with the majority of preserved granites emplaced at c. 2680-2670, 2660-2650 and 2640-2620 Ma (Cassidy et al. 2002;Mole et al. 2012;Pawley et al. 2012). The 2660-2650 group is the most widespread and was broadly synchronous with late Archaean volcanism, deformation and metamorphism (Cassidy et al. 2002;Barley et al. 2003;Kositcin et al. 2008;Mole et al. 2012). The 2640 -2620 Ma group is also common and appears to be the last magmatic event in the Yilgarn Craton, correlating with the craton-wide gold mineralization Kent & McDougall 1995). This suggests that both events are intimately associated with final cratonization of the Yilgarn . Metamorphism throughout the craton is typically of prehnite-pumpellyite to upper amphibolite (Ahmat 1984;Cassidy et al. 2006;Goscombe et al. 2009), although granulite facies metamorphism, concurrent with emplacement of the c. 2640-2620 Ma charnockite granites, is found in the South West Terrane (Nemchin et al. 1994).
Compositionally, late Archaean granites across the craton belong to two primary groups, the High-Ca and Low-Ca granites, which account for c. 60 and 20% of the total surface area of Yilgarn granites, respectively (Champion & Sheraton 1997;Czarnota et al. 2010). In addition, there are three minor groups: the Mafic, Syenitic and High-HFSE (high field strength element) groups. The Mafic granites are of particular interest here as they have been proposed as a source of gold in the Yilgarn, and also host gold mineralization at several deposits (e.g. Granny Smith, Great Eastern, Lady Bountiful and Porphyry; Cassidy et al. 1998). They comprise up to 10% by surface area of the granites in the craton, are typically located within or marginal to greenstone belts, and are characterized by their diverse form, mineralogy and geochemistry (Champion & Sheraton 1997).
The majority of Yilgarn granites are interpreted to have a low to mid-crustal derivation (Champion & Sheraton 1997;Cassidy et al. 2002;Barley et al. 2003;. Granites of the Low-Ca group typically have a rare-earth element (REE) pattern suggestive of a plagioclasebearing source, while the High-Ca granites typically demonstrate a garnet-bearing source. This suggests that the High-Ca granites are from a deeper source (.35 km), while the Low-Ca granites originate from shallower crustal melting (Cassidy et al. 2002;Mole 2012).

Mineral systems in the Yilgarn Craton
The Yilgarn Craton is one of the most endowed geological terranes on Earth (Fig. 3) and hosts a number of world-class orogenic gold (e.g. Golden Mile, Sons of Gwalia, Sunrise Dam) and komatiite-hosted nickel deposits (Mt Keith, Kambalda camp), as well as numerous large BIF-hosted iron deposits (Windarling, Koolyanobbing, Weld Range). These three commodities dominate the mineral systems of the craton; however, volcanic-hosted massive sulphide Cu -Zn (i.e. Golden Grove, Teutonic Bore/Jaguar), vanadium (i.e. Windimurra) and Sn -Ta deposits (i.e. Greenbushes) also occur, although not presently at camp scale.
Gold mineralization occurs in all terranes of the Yilgarn Craton (Fig. 3), with most deposits concentrated in number and resource size in the Eastern Goldfields Superterrane. Orogenic gold deposits are the most common type of gold system in the craton (Fig. 3), although there are rare exceptions such as the Boddington Cu -Au deposit (Archaean porphyry-type with an orogenic gold overprint; McCuaig et al. 2001;Stein et al. 2001). Gold deposits are hosted by a variety of rocks types, with variable structural setting, alteration and ore mineralogy (Witt & Vanderhor 1998;Duuring et al. 2007). However, common parameters suggest they represent a coherent group of epigenetic deposits that formed during a widespread (500 000 km 2 ) hydrothermal event at c. 2650-2630 Ma (Groves 1993;Kent & McDougall 1995;Witt & Vanderhor 1998;Robert et al. 2005;Duuring et al. 2007) during the closing stages of the late Archaean tectono-thermal evolution of the Yilgarn Craton. Gold deposits cluster in the Eastern Goldfields Superterrane as well as the north-central Murchison Domain and the central Southern Cross Domain in the Youanmi Terrane (Fig. 3). In the Eastern Goldfields Superterrane, deposits are concentrated in the Norseman-Wiluna Belt in the Kalgoorlie Terrane and the Laverton belt in the Kurnalpi Terrane, particularly along major regional structures and their subsidiary faults (e.g. Robert et al. 2005).
Iron deposits in the Yilgarn Craton are hosted by banded iron formation, which represents a widely distributed, but volumetrically minor, lithology in exposed greenstone belts (Gole 1981). Economic iron ore bodies are commonly the product of several superimposed early hypogene and late supergene hydrothermal alteration stages (Duuring & Hagemann 2013a). This study is not concerned with the supergene processes, as these are usually late, and unrelated to crustal evolution. To form hypogene BIF-hosted iron ore, BIF (c. 30 wt% Fe) is enriched in iron (.50% Fe) via the dissolution of primary quartz-, iron silicate-or carbonate-rich bands by fluids and the addition of iron oxides to BIF. As in gold deposits, structure plays an important role during the enrichment process, with enriched/upgraded BIF, now as iron ore, often concentrated in late shear zones (e.g. Koolyanobbing; Angerer & Hagemann 2010) and reactivated fault zones located along the margins of the BIF. The BIF sequences within the West Yilgarn are estimated at 3.0-2.8 Ga (Gole 1981;Angerer & Hagemann 2010); however, the fluids and localizing structures that upgrade BIF to iron ore are younger, possibly c. 2.7 Ga. Crustal architecture has the potential to control both these features. BIF-hosted iron deposits are concentrated in the central Southern Cross Domain, Murchison Domain (external to most gold occurrences) and the Narryer Terrane (within the Jack Hills metasedimentary gneiss belt).
Nickel deposits in the Yilgarn Craton are hosted by Archaean komatiites, which are characterized by extremely high eruption or emplacement temperatures (.1600 8C), high MgO (.18% MgO), turbulent magma flow and high-flux emplacement (Nisbet et al. 1993;Hill et al. 1995;Hill 2001;Barnes 2006a, b;Herzberg et al. 2007;Arndt et al. 2008). Typically, komatiite-hosted nickel deposits form through the addition of external sulphur (sediment or exhalative) to the ultramafic magma. This drives the system to sulphur saturation and the formation of an immiscible sulphide liquid, which subsequently concentrates chalcophile elements into ore-grade accumulations (Groves et al. 1986;Arndt et al. 2008).  Cassidy et al. (2006) and Pawley et al. (2012). Important greenstone belts referred to in this study are labelled as follows: JH, Jack Hills; WR, Weld Range; MD, Marda-Diemals; SC, Southern Cross; KG, Koolyanobbing; FO, Forrestania; RAV, Ravensthorpe, LJ, Lake Johnston; AW, Agnew-Wiluna; KAL, Kalgoorlie/Kambalda; NM, Norseman and DK, Duketon. (b) Map showing the distribution of Sm-Nd samples throughout the Yilgarn Craton. Red points represent data collated from previous studies (see Table 2) and black points show primary data collected and analysed by this study.  (2007) and Wyche et al. (2012a, b).
Examples of major mineral deposits or prospects within the specific terrane/domain.
The 2.7 Ga komatiites form an almost continuous c. 700 km belt of high MgO, adcumulate-rich ultramafic lava flows and sills within the Norseman-Wiluna greenstone belt in the Eastern Goldfields Superterrane (Fig. 3). This belt represents the greatest outpouring of komatiite magma and some of the hottest melts preserved on Earth Barnes 2006b;. In accordance with these features, the belt hosts two world-class nickel sulphide camps (Hoatson et al. 2006): (1) the Agnew-Wiluna belt in the northern Kalgoorlie Terrane, hosting numerous deposits such as Mt Keith, Cliffs, Perseverance, Honeymoon Well and Cosmos (Barnes 2006a;Fiorentini et al. 2010Fiorentini et al. , 2012; and (2) the Kambalda camp in the south Kalgoorlie Terrane (Gresham & Loftus-Hills 1981;Beresford et al. 2002) containing many relatively small, high-grade (c. 30 -10% Ni) deposits, such as Long, Victor and Lunnon (Gresham & Loftus-Hills 1981).

Methodology
In this study, the Sm-Nd isotopic technique was used to evaluate crustal evolution through time, with age constraints based on U -Pb secondary ion mass spectrometry (SIMS; using sensitive highresolution mass spectrometry or SHRIMP) zircon geochronology where available (some collated samples from other studies use approximate ages based on regional stratigraphy and/or cross-cutting relationships; see Table 2). Previously published Sm-Nd isotope data (259 samples: Table 2) were augmented with new data from 60 samples for this study. The isotopic data (Table 2, Fig. 1) were investigated in space and time to constrain regions of crust with a common history. The result of the new dataset is a spatially diverse, high-resolution, isotopic understanding of the Archaean crust of the Yilgarn Craton.

Sm -Nd isotope analysis
The new Sm-Nd data acquired for this study were obtained on samples that were pulverized to X-ray fluorescence (XRF) grade and subsequently analysed at two laboratories (Table 2): (1) the University of Melbourne, (Victoria, Australia); and (2) Geosciences Rennes Laboratory (Rennes, France).
At the University of Melbourne, Sm-Nd isotopic data were obtained on sample powders spiked with a 149 Sm-150 Nd tracer and dissolved in Krogh-type high-pressure vessels. Subsequently, Sm and Nd were extracted using EICHROM TM RE-and LN-resin (Maas et al. 2005). Total analytical blanks were well below 100 pg and negligible compared with the amounts of sample analyte. Isotopic analyses were carried out on an NU Instruments multicollector inductively-coupled plasma mass spectrometer with sample introduction via a CETAC Aridus desolvation system. The Nd isotope ratios were measured with signals of 12 -20 V total Nd and corrected for mass bias and spike impurities using an online iterative procedure involving internal normalization to 146 Nd/ 145 Nd ¼ 2.0719425 (equivalent to 146 Nd/ 144 Nd ¼ 0.7219; Vance & Thirlwall 2002) with the exponential law. Typical in-run precision for 143 Nd/ 144 Nd is +0.000010 (2s). Data are reported relative to the La Jolla standard (Nd ¼ 0.511850). External precision, or reproducibility, is c. +0.000020 (2s). US Geological Survey basalt standard BCR-2 yielded average 147 Sm/ 144 Nd and 143 Nd/ 144 Nd of 0.1383 + 0.0002 and 0.512640 + 20, respectively (2s, n ¼ 10, analyses from 2009-2010), consistent with thermal ionisation mass spectrometry (TIMS) reference values (Raczek et al. 2003).
At the Geosciences Rennes Laboratory, samples were spiked with a 149 Sm -150 Nd mixed solution and dissolved in HF-HNO 3 . Rare-earth elements were separated using BioRad AG 50W × 8H + 200-400 mesh cationic resin. Subsequently, Sm and Nd were separated and collected by passing the solution through a further set of ion exchange columns loaded with LN spec EICHROM TM resin. The Sm and Nd were loaded onto double Re filaments with HNO 3 reagent and analysed in a Finnigan MAT262 multicollector mass spectrometer in static mode. In each analytical session, the unknowns were analysed together with the Ames nNd21 Nd standard, which yielded an average of 0.511965 during the course of this study. All analyses of the unknowns are adjusted to a nominal 143 Nd/ 144 Nd value of 0.511850 for the La Jolla standard. Mass fractionation was monitored and corrected using the value 146 Nd/ 144 Nd ¼ 0.7219. Procedural blanks analysed during the period of these analyses were c. 190 pg and are considered to be negligible compared with the total quantity of Nd in the samples.
(1) directly on depleted mantle or (2) as a wellconstrained evolution line defined by a series of data points, is it possible to be certain of a singular source for that sample, and that the calculated model age is representative of crust formation. However, most Nd datasets for evolved crustal rocks do not plot near the depleted mantle evolution line (DePaolo et al. 1991;Taylor & McLennan 1985Fletcher et al. 1994). The possible significance of such data is explored in schematic form in Figure 4.
In Figure 4a, a granite sample (green square) has a geological age T 2 with associated 1Nd well below the depleted mantle curve. This can be explained if the granitic magma is derived from older, felsic, light REE-enriched (low-Sm/Nd) crust. Extrapolation from the sample data-point to the depleted mantle curve using an average crustal Sm/Nd ratio will then yield an age T 1 , the Nd model age or 'crustal extraction' age (T DM 2 ). The situation in Figure 4a is complicated if the magma source of the sample is a mixed source, or if a mafic magma with juvenile 1Nd assimilates older, low-1Nd crust to form a granitic body (Fig. 4b). As a result, the 1Nd and model age of the sample are not representative of the new crust (depleted mantle) or the old crust, but represent a mixture between the two sources. However, complete mixing very rarely occurs, and as a result, when multiple analyses are performed from an area, arrays of data ( Fig. 4b) will often link the original sources involved with the final, mixed source (Arndt & Goldstein 1987;DePaolo et al. 1991). This is important as it allows the estimation of the age and character of the mixing components.

Results
To complement the existing Sm-Nd dataset, this study collected 36 field samples, together with 10 samples from the Geological Survey of Western Australia and 14 from the PhD study of Qiu (1997), for a targeted Nd isotope programme across the SW-central Yilgarn Craton. Previously published Sm -Nd data from 259 samples (Fletcher & Rosman 1982;McCulloch et al. 1983;McCulloch 1987;Watkins et al. 1991;Nutman et al. 1993;Fletcher et al. 1994;Champion & Sheraton 1997;Cassidy et al. 2002) were collated, resulting in a craton-wide database of 319 Sm -Nd analyses of granites and felsic volcanic rocks (see Table 2). Using this information, it is possible to investigate the spatial variation in lithospheric architecture of the Yilgarn Craton. First, the Nd isotopic data are examined through time (Figs 5-9) to establish the source character and age of the Yilgarn crust. These data are then plotted as maps  to understand the spatial distribution of those features. Data are presented as two-stage model ages (T DM 2 ; Liew & Hofmann 1988) and 1Nd values. The T DM 2 is preferred over depleted mantle model ages as this method uses the Sm/Nd ratio of the continental crust (0.11) rather than that of the sample. This produces more realistic model ages regarding crustal rocks. Youanmi Terrane: Southern Cross Domain. Samarium -Nd data are available for 72 samples from the Southern Cross Domain, with U -Pb ages ranging from 2983 to 2615 Ma (Fig. 8e). The T DM 2 ages from this domain range from c. 3500 to 2750 Ma, peaking at 3100 Ma, with 1Nd values between 26.9 and 3.9, peaking at 22.0 ( Fig. 6a, b). This distribution indicates that, while old, reworked crust dominates this domain, a minor juvenile component is also present (Fig. 8e).
A small number of samples have U -Pb ages .2750 Ma; the 2983 Ma sample has a T DM 2 of 3150 Ma (1Nd of 1.8), while the 2813 Ma sample has a T DM 2 of 2810 Ma (1Nd of 3.9), suggesting that relatively juvenile crust was present at 2983 Ma, with the addition of mantle-derived material at c. 2810 Ma. Most samples younger than 2750 Ma fall into two groups with T DM 2 ages of c. 3300 and 3100 Ma, with a lesser number overlapping with T DM 2 ages c. 3200 Ma. Two samples with T DM 2 c. 3500 Ma (1Nd of 26.5) indicate the presence of a minor, older component (Palaeoarchaean crust) (Fig. 8e).
The Southern Cross Domain displays four distinct but overlapping felsic magmatic events younger than 2750 Ma (Fig. 8e). The first event, at 2750-2700 Ma, shows a heterogeneous crustal source with T DM 2 ages of 3300-3000 Ma. The second, at 2690-2680 Ma, consists of crustal sources with T DM 2 ages of 3300 (1Nd 24.5 to 23.5) and 3100 Ma (1Nd 20.9 to 20.5). The c. 2665-2650 Ma group consists of two very similar subgroups: (1) a dominant subgroup at T DM 2 3100 Ma and 1Nd 22.0 to 21.0; and (2) a relatively minor subgroup at T DM 2 3300 Ma and 1Nd c. 24.5. The final c. 2640 -2630 Ma event dominantly displays a heterogeneous crustal source with T DM 2 ages of 3300-2900 Ma (1Nd 24.2 to 0.5), and a more continuous Nd 'array' than the previous two events.
Youanmi Terrane: Murchison Domain. Thirty-six samples with U-Pb ages ranging from 2950 to 2602 Ma are available from the Murchison Domain. The T DM 2 values from this domain range from c. 3660 to 2950 Ma (peak at c. 3200 Ma), with 1Nd values between 27.2 and 2.9, peaking at 23.2 (subordinate peaks at 21.3 and 0.5; Fig. 6c). The wide range of the T DM 2 and 1Nd probability density curves (Fig. 6c, d) demonstrates the heterogeneous nature of crustal sources in this domain.
Samples with U -Pb ages .2750 Ma demonstrate a wide range of source compositions, from a reworked source at c. 2950 Ma (T DM 2 values of c. 3700-3300 Ma, 1Nd 25.9 to 20.9) to a juvenile source at c. 2920 Ma (T DM 2 values of c. 3010-2980 Ma, 1Nd 1.7 and 2.9). After 2750 Ma, three broad, overlapping T DM 2 groups occur at c. 3400-3300, 3200 and 3000 Ma (Fig. 8f ).
The Murchison Domain displays three welldefined felsic magmatic events younger than 2760 Ma (Fig. 8f). The first, at c. 2760-2740 Ma, consists of the three main T DM 2 groups at c. 3300, 3200-3100 and 3000 Ma, with corresponding 1Nd values of c. 24.4 to 23.7, 21.6 to 21.0 and     (Fig. 6e, f ), and 1Nd values between 25.5 and 2.6, with the major peak at 21.6. Additional subordinate peaks occur at 23.4 and 2.5 (Fig. 6e). Samples with U -Pb ages of c. 3000 Ma have two apparent sources with T DM 2 of c. 3250 and 3100 Ma and 1Nd of 0.2 and 2.5, respectively (Fig. 8g) ranging from 3400 to 3000 Ma (1Nd 25.5 and 20.3; Fig. 8g), clustering at a T DM 2 ages of 3200 Ma (1Nd 23.5).
Narryer Terrane. In the NW of the Yilgarn Craton, 16 samples aged between c. 3005 and 2620 Ma are available from the Narryer Terrane (Figs 6g, h &  8h). The T DM 2 values for this terrane range from c. 3700 to 3150 Ma, peaking at c. 3400-3300 Ma (Fig. 6h). The 1Nd values for this area range from 29.0 to 1.0, and peak at 26.8 and 21.5, with relatively minor peaks at 29.2, 24.2 and 0 (Fig. 6g). age. These data suggest that this terrane is very long lived, with isotopic evidence of crust formation as old as c. 3700 Ma. The 1Nd values decrease consistently through time (Fig. 8h)  Analysis performed at the University of Melbourne.
14 Analysis performed at the Geosciences Rennes Laboratory.

15
A standard error of 0.5 1 units is used for the 1Nd data as this corresponds to the average analytical error.

Spatial evaluation of Sm -Nd isotope data
The temporal analysis of regional Sm -Nd isotopic data presented above documents the crustal history of the individual terranes/domains of the Yilgarn Craton. However, the isotopic data are sorted using an independently derived terrane system (Cassidy et al. 2006), developed in conjunction with additional litho-stratigraphic and geochronological data. The use of these datasets, while beneficial in the understanding of the complete tectono-stratigraphic evolution of an area, also introduces an external level of interpretation that may not be truly representative of the isotopic  Fig. 4. Diagrams demonstrating the multiple interpretations of unradiogenic Sm-Nd data. (a) This 1Nd isotope plot demonstrates a potential interpretation for an isolated data-point crystallized at T 2 , with no direct connection with the depleted mantle. Here, the sample is inferred to be the reworked product of precursor, juvenile crust that was extracted from the mantle at T 1 . Hence this rock has a T 1 model age. (b) An alternative interpretation for this data-point is mixing of two contrasting Nd sources. Here, new, highly radiogenic mantle-derived crust is emplaced into older, precursor crustal material (model age T 1 ) at T 2 . During emplacement, the juvenile magma assimilates and mixes with the precursor crust. This produces a rock with Nd systematics intermediate between the two components. As a result, the model age here is a mixed age, and does not represent a mantle extraction event. This produces a crustal source with a multistage isotopic history. In this situation, data 'arrays' often form between the two sources. These represent magmas with varying degrees of contamination. CHUR, chondritic uniform reservoir. D. R. MOLE ET AL. 50 history of the crust alone. As a result, the Sm-Nd isotope characteristics (T DM 2 and 1Nd) were plotted as contour maps (Figs 10 & 11) in order to constrain the spatial relationships in crustal evolution without the use of external datasets. Only granites and felsic volcanic rocks with U -Pb crystallization ages of 2800-2600 Ma were mapped in order to reduce potential mixing signatures between distinct events. As a result, this map represents a snapshot of the lithospheric architecture of the Yilgarn Craton at 2800-2600 Ma.
As well as the T DM 2 and 1Nd contour maps (Fig. 10a, b), two key isotopic cross sections are plotted (Fig. 11b, c) to show the change in isotopic character of the crust across key sections of the craton. The T DM 2 and 1Nd values presented in this section are broad regional approximations based on the maps of Figure 10a, b, used in conjunction with Figure 8. Individual sample sites are displayed on the maps, and it is important to account for the location and density of sampling, as they provide a fundamental first-order control on the precision of the contour mapping.
The most striking observation is that, as indicated by previous studies (Champion & Sheraton 1997;Cassidy et al. 2002;Champion & Cassidy 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 -12 -10 -8 -6 -4 -2 0 2 4 6 εNd T DM 2 (Ga) 2007), the Eastern Goldfields Superterrane forms a distinct crustal block, with typical T DM 2 values of c. 3000 -2700 Ma (Fig. 10b) and 1Nd of c. 20.2 to 3.6 (Fig. 10a). This superterrane is younger and more juvenile than the West Yilgarn, which dominantly consists of crust with T DM 2 of 3300-3000 Ma and 1Nd of 24.0 to 21.0. The margin between the isotopically distinct West Yilgarn and Eastern Goldfields Superterrane represents a lower crustal boundary between contrasting source regions at the time of granite magmatism. This boundary correlates with the surface expression of the Ida (south : Swager 1997;Drummond et al. 2000;Goleby et al. 2004Goleby et al. , 2006 and Waroonga faults (north; Cassidy et al. 2006). This association empirically suggests that these major crustal faults (Drummond et al. 2000;Goleby et al. 2004Goleby et al. , 2006 represent the contact between two proto-cratonic blocks (Cassidy & Champion 2004;, although at present this relationship is unconfirmed. Figure 11b illustrates the significant difference between the West Yilgarn and Eastern Goldfields Superterrane and demonstrates (1)  As highlighted previously (Fig. 8c), the Kurnalpi Terrane has an older, reworked component at c. 2680-2670 Ma, which has a T DM 2 of c. 3200-2800 Ma and 1Nd values from 23.2 to 1.8. Figure 10 shows that this older component forms a relatively old 'block' of crust at the margin between the Kurnalpi and Burtville Terranes, and may be the result of reworking associated with terrane amalgamation, or alternatively a small fragment/sliver of older crust sitting at the terrane boundary. To the west, the structural boundaries of the Kurnalpi Terrane (Cassidy et al. 2006) correlate with a north-south belt of young, relatively juvenile crust with T DM 2 values of c. 2900 -2700 Ma and 1Nd of c. 1.0 to 3.0.
In contrast, the Kalgoorlie Terrane demonstrates a mixed source as identified in the temporal analysis section (Fig. 8d). The eastern side of the terrane dominantly comprises young, juvenile material similar in age to that of the Kurnalpi Terrane. The T DM 2 and 1Nd values increase and decrease, respectively, from east to west with the increasing interaction and addition of the older Southern Cross Domain component (Fig. 10a, b).
Furthermore, there is subtle variation along the north -south isotopic margin between the Kalgoorlie Terrane and Southern Cross Domain. In the south and north Kalgoorlie Terrane, the juvenile material extends close to the surface positions of the Ida and Waroonga faults, respectively, specifically in the areas underlying the Agnew -Wiluna and Kalgoorlie -Norseman greenstone belts. However, in the central Kalgoorlie Terrane, there is a 200-150 km embayment in this isotopic margin where older material extends further east into the Kalgoorlie Terrane, shifting the north-southtrending isotopic boundary eastward (see Figs 10 -13). This feature correlates spatially with the poorly mineralized area between the gold-nickelrich Agnew -Wiluna and Kalgoorlie-Norseman greenstone belts (Figs 3, 12 & 13;. Unlike elsewhere in the Kalgoorlie Terrane,   Fig. 8. Sm-Nd isotope data (as 1Nd) v. age for all samples from the Yilgarn Craton (see Table 2). this area does not contain high-MgO komatiites or associated nickel sulphide deposits (Fig. 13). As a result, this feature appears to have a significant physical control on geological processes and features of the terrane.
In contrast to the relatively homogenous isotopic character of the Eastern Goldfields Superterrane, the West Yilgarn is heterogeneous with respect to spatial variations in crustal source, creating a more complex, intra-cratonic lithospheric architecture. As a result, a number of discrete crustal blocks can be identified that do not generally match the terrane subdivision of Cassidy et al. (2006). This is probably because the current terrane structure was dominantly controlled by late, major tectonothermal and structural events in the Yilgarn at c. 2680-2620 Ma (Cassidy et al. 2002Mole et al. 2012), whereas the Nd isotopic map contains time-resolved information that accounts for previous tectonic and magmatic events. The crustal blocks identified here (Figs 10 & 11) may represent preserved cratonic elements that collectively form the West Yilgarn protocraton.
The Southern Cross Domain is characterized by a northern area with T DM 2 values of c. 3400-3200 Ma and 1Nd of 26.0 to 21.0, and a less evolved southern region with T DM 2 values of c. 3200-3000 Ma and 1Nd of 22.0 to 0. The northern, more reworked area is designated as the Marda block, and the southern area as the Lake Johnston block (Fig. 11a).
The north-south isotopic cross-section of the crust across the craton (Fig. 11c) displays similar features to that observed across the West Yilgarn-Eastern Goldfields Superterrane boundary (Fig.  11b). The crustal 1Nd values decrease significantly to the north of the Lake Johnston block margin, which occurs at the southern edge of the Marda greenstone belt. The Marda block contains a small component of the Lake Johnston block and vice versa, suggesting that their source regions interacted during tectono-thermal development of the region.
Available T DM 2 and 1Nd data from the South West Terrane largely suggest it has a similar crustal history to the Lake Johnston block, although the northern boundary (around the Koolanooka fault; Fig. 10a, b) is more similar to the Marda block. This suggests that the reworked Marda block may extend west across the West Yilgarn into the South West Terrane (Figs 10 & 11), cross-cutting previously proposed terrane/domain boundaries including the Southern Cross-Murchison Domain and Youanmi-South West Terrane boundaries. This is in agreement with the assertion of Cassidy et al. (2006) that many of the domain/terrane boundaries are the structural expression of late tectono-thermal events in the Yilgarn Craton.
The crust of the Murchison Domain appears to have two spatially distinct crustal sources, which correlate with the T DM 2 values of the c. 3400, 3200 and 3000 Ma groups identified previously (Fig. 8f ). The c. 3400 group is very minor and spatially difficult to discriminate from the 3200 Ma (1Nd 23.9 to 22.2) group, which is generally restricted to the margins of the domain. The  (2007) and Ivanic et al. (2010Ivanic et al. ( , 2012, which is flanked by regional structures (i.e. Carbar, Big Bell, Cuddingwarra and Mt Magnet faults) of the same orientation (Spaggiari 2006;Van Kranendonk & Ivanic 2008). This NE to SW band of juvenile crust has been interpreted as a failed continental rift (Ivanic et al. 2010). It is unlikely to represent a back-arc rift formed during the docking of the Narryer Terrane, as this amalgamation probably occurred at 2750-2650 Ma, along the Balbalinga Fault (Occhipinti et al. 2001), which is later than the inferred age of the rift (c. 3000-2800 Ma). The majority of the greenstone sequences in this domain occur within this 'rift' zone (i.e. Meekatharra -Wydgee belt ;Spaggiari 2006), and the large layered intrusions of the Murchison Domain (Ivanic et al. 2010) occur on its flanks (Figs 10 & 11). The rift may have been formed at c. 2810-2800 Ma during the emplacement of the mafic -ultramafic intrusions (e.g. Windimurra; Cassidy & Champion 2004;Ivanic et al. 2010), which would have required significant crustal thinning to create space for the large magma volumes as well as the ascent of mantle-derived magmas (Ivanic et al. 2010). Alternatively, this event may have Histograms of U-Pb geochronology are also shown for each terrane/domain. Some ages used for these histograms were acquired using stratigraphic, cross-cutting and/or structural relationships; these are highlighted in Table 2. Hence, although most ages presented here are U-Pb SHRIMP on zircon and the age groups plotted agree with previous studies , Cassidy et al. 2002, Champion & Sheraton 1997, these groups should be used tentatively. Crustal reservoirs referred to in the text are shown, where blue indicates the initial crust in the terrane (where known) and red highlights later additions of juvenile or ancient crustal material and the subsequent mixed reservoirs. 2940 Ma juvenile material (Fig. 8f ). This scenario appears more likely, as the majority of the crust within the rift has a T DM 2 of c. 3000 Ma (Figs 8f  & 10b).
In the NW corner of the Yilgarn Craton, the Narryer Terrane is by far the oldest crustal block based on Nd isotope data, with T DM 2 ages .3300 Ma and up to c. 3700 Ma (1Nd 29.0 to 24.0) consistent with the occurrence of the oldest granites and mafic magmatic rocks in the Yilgarn Craton (Kinny et al. 1988;Nutman et al. 1993Nutman et al. , 1991.

Evaluation of Sm-Nd data in space and time
The data collected in this study and corresponding analysis presented in the results section can be used to evaluate the crustal history of the individual terranes that make up the Yilgarn Craton. The potential geodynamic settings driving this crustal evolution are suggested where possible. The Nd isotopic system differentiates between crust and mantle sources for magmas, and subsequently cannot uniquely discriminate between all tectonic settings, only those which require juvenile input (continental rift, island arc, back-arc, plume magmatism) or crustal melting (collisional orogenies, continental arcs).

Eastern Goldfields Superterrane: Yamarna
Terrane. The Sm-Nd data from the Yamarna Terrane cluster relatively tightly above CHUR at 0.7 to 1.4 1Nd and T DM 2 of c. 2950-2850 Ma and shows little variation toward more juvenile or reworked ages. Based on this data distribution, it appears that the Yamarna Terrane was extracted from the mantle at c. 2950-2850 Ma. The possibility of crustal source mixing was disregarded based on the juvenile (1Nd) and clustered nature of the data, together with the lack of any pre-existing .2900 Ma crust. This suggests that the model age represents a real crustal growth event. This theory could be further evaluated by collecting Sm-Nd and/or Lu -Hf isotopes from supracrustal rocks of this terrane, which span a considerable geological history (.2832 -2630 Ma; Pawley et al. 2012). Understanding of this terrane is at an early stage, and as the available dataset is very small, the interpretation presented here should be considered speculative.
Eastern Goldfields Superterrane: Burtville Terrane. The Nd isotope distribution of the Burtville Terrane could be interpreted using a number of crustal evolution models: (1) Evolution of the crust from a single source.
This appears unlikely owing to the spread in Sm-Nd data and the observation of two sources prior to 2700 Ma with T DM 2 2900 Ma (at 2755 Ma) and 3100 Ma (at 2940 Ma). In addition, if the crust of this terrane was derived from a mantle extraction event at the dominant T DM 2 age of c. 2900-2850 Ma, the 2940 Ma sample should record a depleted mantle signature and represent the original crust, which is not the case (Fig. 8b).
(2) Derivation of the crust of the Burtville Terrane from multiple crustal sources. Initially, at c. 2940 Ma, the crust had a T DM 2 of c. 3100 Ma and 1Nd 1.1 (Figure 8b). At some point between 2940 and 2755 Ma, this crust was rejuvenated by the addition of juvenile material, speculated here to occur at 2810-2750 Ma, corresponding with highly juvenile Lu -Hf data from the Mapa Igneous Complex and Swincer Dolerite (Wyche et al. 2012b;Pawley et al. 2012), as well as the juvenile sample at c. 2755 Ma in Figure 8b. The Hf data from the c. 2812 Ma Swincer Dolerite (Pawley et al. 2012) support the crustal mixing model suggested here, showing 1Nd values ranging from 22.9 to +6.3 (Wyche et al. 2012b). This mixed reservoir was later reworked during the ,2755 Ma granite events. This model explains the range of T DM 2 ages for these later granites. As a result, at ,2700 Ma, the majority of granites were formed by reworking of a mixed 2950-2850 Ma reservoir. If the crustal history presented above is accurate, it suggests that a tectono-thermal event at c. 2900-2800 Ma led to the input of juvenile material into older c. 3100 Ma crust. The Duketon komatiites (Duketon greenstone belt; Fig. 1a) are dated at c. 2805 Ma , and these ultramafic lavas may represent the surface manifestation of a mantle plume. Crustal attenuation and thinning associated with this plume may have led to intracontinental rifting at this time, and the addition of juvenile mantle material into the crust.
Eastern Goldfields Superterrane: Kurnalpi Terrane. The Kurnalpi Terrane has a less extensive magmatic   10. Sm-Nd isotopic contour maps showing the spatial variation of 1Nd and T DM 2 values from granitic (a) 1Nd data mapped using the geometric interval (GI) method, where all groups are sized using a geometric series that highlights the differences between areas and removes the effects of extreme values; (b) T DM 2 data also mapped using the geometric interval (GI) method. The location and corresponding average MgO (plus standard deviation) content is shown for all komatiites (both c. 2.9 and 2.7 Ga), and the locations of the Murchison large layered intrusions are also plotted. Isotopic contour maps were produced using the inverse distance weighted interpolation method in ArcGIS w . This method, which used 15 nearest neighbours (minimum of 10) at a 'power' of 2, produced the most robust spatial representation of the isotopic dataset. Other methods, such as kriging, were not viable owing to the relatively small dataset. Data were grouped using the geometric interval method (a protocol built into ArcGIS w ). This method designates class breaks based on intervals that have a geometrical series. This ensures that each class range has approximately the same number of values and that the change between intervals is fairly consistent. This algorithm, which was specifically designed to accommodate continuous data such as the isotope data, produces a result that minimizes variance within classes, and can even work reasonably well on data that are not normally distributed. As a result, this method enhances the contrast among different crustal regions as it accounts for extreme values and designates groups and group size accordingly. history than most other terranes, especially those of the West Yilgarn, with the oldest sample presented here dated at 2714 Ma. The evolution of this terrane occurred through a series of felsic magmatic events constrained by U-Pb data (Fig. 8c): (1) The 2715-2700 Ma felsic magmatism has an 1Nd of 2.0 and was derived from reworking of a source with T DM 2 c. 2850 Ma. Data from this event forms a tight cluster suggesting little to no mixing, possibly associated with the establishment of a rift between the Burtville and Youanmi Terranes. At the initiation of the extension, juvenile melts would be contaminated by c. 3100-2900 Ma Burtville crust; a feature not shown by the data in Figure 8c The .3100 Ma material is interpreted as 'initial crust' in Figure 8c, suggesting it may represent an older end member of Burtville Terrane crust. However, it may also have been sourced from the Southern Cross Domain (Fig. 11b). This suggests collisional tectonics and crustal thickening associated with the possible closure of a rift or rifts at these margins. This is supported by a transition in the composition of supracrustal rocks from mafic volcanics/intrusives, calc-alkaline complexes and feldspathic sedimentary rocks at 2720-2700 Ma to bimodal rhyolite -basalt volcanism and felsic alkaline complexes at 2690-2680 Ma (Cassidy et al. 2006;Barley et al. 2008). This older, c. 3200 Ma component may represent thinned Burtville Terrane crust which has been overwelmed with juvenile input at c. 2720 Ma.  (Fig. 8c). This material may represent preexisting Burtville Terrane crust which was thinned during extension at c. 2720 Ma. Felsic magmatism, while known to occur at 2720-2700 Ma in the Agnew-Wiluna belt (Barley et al. 2003;Kositcin et al. 2008), is relatively rare in comparison to the Kurnalpi Terrane, and is not represented in this dataset. This indicates the predominance of plumederived mafic -ultramafic magmatism at this time (Claoué-Long et al. 1988;Nelson 1997;Barley et al. 2003;Cassidy et al. 2006). Subsequently, the felsic magmatic activity for this terrane occurs in four well-defined, overlapping events between 2690 and 2620 Ma: (1) The first event occurs at c. 2685-2680 Ma and forms an array of T DM 2 and 1Nd at c. 3300-2900 Ma and 22.9 to 1.7, respectively. This suggests that juvenile input, probably during the c. 2700 plume event, has occurred since 2760 Ma, reducing the T DM 2 from c. 3100-3000 Ma. Despite this, significant reworking dominates this event, with the addition of an unradiogenic crustal component with T DM 2 c. 3300 Ma. Based on Figures 10  and 11, it is likely that the juvenile addition and reworking signatures are spatially controlled, with the more juvenile source occurring in the east of the terrane and reworked magmas occurring in the west. The reworked areas are likely to be part of, or heavily contaminated by, the Southern Cross Domain (Champion & Sheraton 1997;Cassidy et al. 2002;Czarnota et al. 2010). In terms of geodynamics, the Sm-Nd array could represent crustally contaminated melts from the retreating or collapsing (Kumagai et al. 2007;Kumagai et al. 2008) c. 2700 Ma plume (Campbell & Hill 1988), or tectonic movement of the crust away from plume-sourced juvenile material. The process may have been coupled with the closure of the rift basin formed during komatiite emplacement and collision of the Eastern Goldfields and West Yilgarn proto-cratons.
(2) The second recorded magmatic event in the Kalgoorlie Terrane occurred at 2670-2660 Ma and appears to be more juvenile in nature, dominantly consisting of material with T DM 2 3000-2800 Ma and 1Nd of 0.3 to 2.8. There is also a minor T DM 2 c. 2700 Ma component which correlates with the c. 2700 Ma plume event, inferring crustal growth directly related to a mantle plume (Fig. 8d). The absence of the 3300-3200 Ma component indicates that the influence of the Southern Cross Domain waned during this period, possibly due to rifting of the Kalgoorlie Terrane away from the West Yilgarn. Alternatively, melting may have shifted to a deeper, more juvenile reservoir formed during previous plume magmatism. This scenario would correspond with deeper thermal effects of a retreating/collapsing plume.
(3) The third event occurred at 2655-2650 Ma and displays a similar Nd array to that of the 2685-2680 Ma event. While High-Ca granites are still abundant, the Low-Ca group begins to dominate at c. 2655 Ma (Cassidy et al. 2002) and felsic volcanism terminates , suggesting a major tectono-thermal change at this time (Cassidy et al. 2002Czarnota et al. 2007;Mole et al. 2012). In terms of geodynamics, this may be due to shallower melting of a more reworked source, as suggested by the chemistry of the Low-Ca granites (Champion & Sheraton 1997;Cassidy et al. 2002). Alternatively, reduced melt production and decreased mantle input associated with the 'retreat' of the mantle plume, together with crustal thickening associated with rift closure, may have led to an increase in crustal reworking. (4) The fourth and final event at 2635-2620 Ma displays further reworking, with data clustered on the c. 3000 and 3200 Ma T DM 2 lines. This event infers that the cratonizing granites of the Kalgoorlie Terrane intruded relatively late, c. 20 -10 myr after those in the other terranes of the Yilgarn Craton with the exception of the South West Terrane, with which they are synchronous (Sircombe et al. 2007).
Although the Kalgoorlie Terrane appears to be more reworked than the Kurnalpi Terrane (Fig.  10), Figures 9d and 11b show that, if the contaminated material associated with the Southern Cross Domain is removed, the terranes have a similarly juvenile signature. This suggests that both terranes had similar juvenile histories prior to the contamination of the west Kalgoorlie Terrane with material from the Southern Cross Domain at c. 2685 Ma.
Youanmi Terrane: Southern Cross Domain. This domain is characterized by two spatially distinct crustal domains, with the T DM 2 c. 3300 Ma Marda block in the central-north of the domain, and the T DM 2 c. 3100 Ma Lake Johnston block in the south (Fig. 11a). This, together with the occurrence of younger T DM 2 c. 2800 Ma (1Nd 2.8 to 3.9) and older c. 3500 Ma (1Nd 27.0) material, suggests that multiple crustal sources were involved during the evolution of this domain.
Despite these observations, the crustal reservoirs of the Southern Cross Domain remain fairly consistent through time, suggesting that the older 3500 Ma source was very minor and/or almost completely rejuvenated by the input of younger material. As the c. 3100 Ma source of the Lake Johnston block can be traced back to 2983 Ma (Fig. 8e), it appears that this T DM 2 is representative of a crustal growth (mantle extraction) event, which may have assimilated the 3500 Ma component. In the Marda block, T DM 2 ages of c. 3300 Ma suggest that either mantle input at 3100 Ma was less extensive in this area, or that this crust was extracted earlier.
These scenarios assume a pre-existing c. 3500 Ma crust in both blocks, as suggested by single data points (Figs 8e & 10) and .3400 Ma zircon xenocrysts found in granites from these blocks (Cassidy et al. 2002;Mole 2012;Mole et al. 2012). This is especially likely in the Marda block, where Lu-Hf data on zircon indicated model ages .4000 Ma (Wyche et al. 2012b), and detrital zircons from basal quartzite units are 4350-3130 Ma (Wyche et al. 2004).
As with the terranes of the Eastern Goldfields Superterrane, the Southern Cross Domain displays a series of overlapping magmatic events: (1) The first, at 2750-2700 Ma, is diffuse and forms two broad, spatially controlled groups. The Lake Johnston block displays a less reworked signature (c. 1Nd 22 to 0.8), while the Marda block consists mainly of crust with T DM 2 ages of c. 3300 Ma (1Nd c. 24.0).
(2) The second event at 2690-2680 Ma forms two groups clustered at T DM 2 ages of c. 3050 and 3300 Ma (Fig. 8e). Magmatism in this event again appears to be spatially controlled, with each block reworking its respective pre-existing crust. This period correlates with the 2685-2680 Ma reworking period in the Kalgoorlie Terrane, suggesting some form of collisional event between the two blocks ( Fig. 9f).
(3) The third event at c. 2665 -2650 Ma dominantly displays reworking of the c. 3100 Ma source, suggesting that felsic magmatism at this time was mainly restricted to the Lake Johnston block. This event correlates with the juvenile pulse observed for the Kalgoorlie and Kurnalpi Terranes, suggesting a shared felsic magmatic event that tapped different source regions. (4) The fourth magmatic event at c. 2640-2630 Ma forms an almost continuous array of Nd-isotope compositions from 1Nd c. 25.0 to 21.0 and T DM 2 ages 3300-3100 Ma, indicating that crustal granites were tapping the source regions of both the Marda and Lake Johnston blocks (Fig. 9a-c), producing magmas with mixed isotopic compositions and suggesting that these blocks had amalgamated.
Youanmi Terrane: Murchison Domain. The Murchison Domain shows broadly similar crustal source  Figure 12. This diagram demonstrates the changing lithospheric architecture between terranes of different isotopic composition and history. The thicker, older crust to the west is inferred to have focused the plume into the shallower area to the east as demonstrated here. This either created a set of rift-related terranes in the Kalgoorlie-Kurnalpi Terranes or reactivated a previous margin, which allowed high flux passage of hot, pre-existing komatiitic magmas to the surface. A similar process is inferred to have formed the c. 2.9 Ca komatiites and associated deposits in the Southern Cross Domain, but this cannot be proven using Sm-Nd isotopes as this map only covers the 2.8-2.6 Ga period. The approximate thickness of developed Archaean lithosphere (c. 250-150 km) was taken from Boyd et al. (1985) and Begg et al. (2009). The approximate scale of the plume head (c. 1600 km), tail (200-100 km) and thickness (150-100 km) were taken from Campbell et al. (1989) and . These values are proxies based on modern analogues and experiments. characteristics to those of the Southern Cross Domain (Fig. 9g). However, the Murchison Domain appears to display a more well-defined temporal grouping, with four major felsic magmatic events: (1) The oldest event occurs at c. 2950-2920 Ma and indicates that the crust of the Murchison Domain was already reworked, with a maximum T DM 2 of 3700 Ma and 1Nd 25.9 at 2950 Ma. However, the occurrence of relatively juvenile material at c. 2920 Ma (T DM 2 of c. 3000 Ma and 1Nd of c. 2.0) suggests juvenile input into older, heterogeneous crust. This interpretation is supported by the Lu -Hf data of Ivanic et al. (2012), which showed evidence for pre-existing 3800-3700 Ma crust and mantle input at c. 3040 Ma as well as mixing between a juvenile (mantle) source and a reworked, c. 3800-3700 Ma source at 2980 Ma. This juvenile input at c. 3000 Ma possibly represents the opening of the NE -SW 'rift' zone (see Figs 10 & 11) within pre-existing 3700-3300 Ma crust, and formation of the 2960-2930 Ma Golden Grove Group greenstone sequence (Wang et al. 1998, Spaggiari 2006Van Kranendonk & Ivanic 2008). This juvenile zone possibly represents a 'failed continental rift' as postulated by Ivanic et al. (2010), and may have been reactivated during the emplacement of the c. 2810-2800 Ma mafic -ultramafic intrusions (Ivanic et al. 2010). This is supported by the moderately reworked nature of the coeval Norie Group mafic volcanics (1Nd 20.5 to 0.9), suggesting contamination with older unradiogenic crustal material (Wyman & Kerrich 2012). whereby these younger granites were derived from more unradiogenic middle crust, as opposed to the more mafic lower crust inferred for older granites (Cassidy et al. 2002;Ivanic et al. 2012). Alternatively, felsic magmatism at this time was isolated in the older crust at the margins of the juvenile zone.
This interpretation of the Murchison Domain crustal history remains tentative owing to known geological events that are not represented in the compiled dataset. First, the formation of large layered intrusions/sills (e.g. Windimurra, Barambie) at c. 2810-2800 Ma and komatiitic basalts at c. 2735-2710 (Glen Group and Yalgowra suite; Ivanic et al. 2010;Wyman & Kerrich 2012) represents the input of a large amount of mafic, mantlederived material into the crust (Ivanic et al. 2012), which is poorly represented in the Sm -Nd data. Second, the 2670-2650 Ma granite event, found throughout the Yilgarn Craton and clearly identified in the Lu -Hf work of Ivanic et al. (2012) from this domain, is not represented in the Sm -Nd dataset. If this group is added to the events identified above, the felsic magmatic episodes from the Southern Cross and Murchison Domains correlate well (Figs 6,7,9 & 11), with events in the Murchison occurring at c. 2760-2740, 2700-2680, 2670-2650and 2640-2620Ma, and at c. 2750-2700, 2690-2680, 2665-2650, 2640-2630 Ma in the Southern Cross Domain.
South West Terrane. In general, the South West Terrane shows the reworking of a heterogeneous c. 3400-3000 Ma crustal source (dominantly 3200 Ma), from c. 3000 to 2630 Ma (Fig. 9g). The only major magmatic event in the compiled dataset occurs at c. 2645 -2630 Ma and represents reworking of this source. This dominant T DM 2 age of c. 3200 Ma is likely to represent a mixing age owing to the heterogeneous nature of the crustal source for this terrane. The end members of this mixing trend, more representative of the added and pre-existing crust, occur at T DM 2 ages c. 3000 and 3400 Ma, broadly correlating with events from the Youanmi Terrane.
Narryer Terrane. Nutman et al. (1993) summarized their Sm-Nd data from 3730-2620 Ma granites and granite -gneisses of the Narryer Terrane as demonstrating two broad crustal sources. The early Archaean (3700-3350 Ma) gneisses are typically tonalitic in composition and have T DM of variable reworking of a 3400-3200 Ma crustal reservoir, with the younger 2750-2620 Ma granites the result of further reworking of this source.
Data presented here are ,3100 Ma and record a c. 3500-3200 Ma T DM 2 age (Fig. 8h). This appears to be analogous to the 3400-3200 Ma source that produced the middle Archaean gneiss and late granites of Nutman et al. (1993). Consequently, the data presented here only capture the reworking history of a 3500-3200 Ma source, apart from a single 2648 Ma granite with T DM 2 of c. 3700 Ma.
The change in source composition and age of crustal rocks of the Narryer Terrane at c. 3300-3000 Ma suggests that a juvenile event created new crust at this time (see Nutman et al. 1993), which underwent variable mixing with the older, 3700-3500 Ma early Archaean material. This led to a mixed 3400 -3200 Ma source, demonstrated by the wide range of 1Nd (23.7 to +3.2) for the middle Archaean granite-gneisses, which represent variable mixing between the new and old crust. Reworking of this mixed source at 2750-2620 Ma then formed the late Archaean granites. This conclusion is supported by the occurrence of c. 3700 Ma xenocrystic zircons in the younger granites of the Narryer Terrane (Nutman et al. 1993), as well as the 3700 Ma T DM 2 age for a 2648 Ma granite (Fig. 8h). the earlier sections. Interestingly, the Southern Cross Domain demonstrates a very minor juvenile component (Figs 8e & 11c), which correlates with the signature of the Eastern Goldfields Superterrane.
This information infers a minimum 1100 myr crustal history for the West Yilgarn (Fig. 7). Initially, pre-existing 3700-3500 Ma crust was rifted in the Cue area of the Murchison Domain at c. 3000 Ma, thinning the crust and extracting new material from the mantle. This event broadly correlates with the extraction of the Lake Johnston block and Burtville Terrane crust (Wyche et al. 2012b), potentially owing to the c. 3000 Ma plume event that produced the Southern Cross Domain komatiites (Perring et al. 1995;Wang et al. 1996;Perring et al. 1996). The Marda block, inferred to represent the core of the West Yilgarn, appears to represent older crust that had less juvenile input at this time. At c. 2810-2800 Ma, the Murchison rift was reactivated during emplacement of the mafic-ultramafic intrusions (Ivanic et al. 2010) and synchronous greenstone sequences (Norie Group;Van Kranendonk & Ivanic 2008). This event correlates with the c. 2805 Ma Duketon komatiites and the inferred juvenile input to the Burtville Terrane crust. Felsic magmatic activity from 2800 to 2600 Ma appears to have dominantly reworked the older, spatially segregated crustal sources. However, the occurrence of boninites in the Polelle Group suggests that subductionderived magmas, probably related to the docking of the Narryer Terrane, were emplaced at c. 2800-2700 Ma (Wyman & Kerrich 2012;Occhipinti et al. 2001). This event is not clearly recorded in the Sm -Nd data (Fig. 8f).
The result of the comparison between the West Yilgarn and the Eastern Goldfields Superterrane indicates two major features: (1) The Eastern Goldfields Superterrane is c. 400 -600 myr younger than the West Yilgarn, and up to 800 myr if the oldest material of the Narryer Terrane is considered (Fig. 7).
(2) Both crustal blocks have been interacting (shared tectono-magmatic events and isotopic signatures) since c. 3100 -3000 Ma, as suggested by contamination of basalts with older unradiogenic crust (Said et al. 2010;McCulloch & Compston 1981;Barley 1986), shared greenstone events (Burtville -Youanmi; Pawley et al. 2012;Van Kranendonk & Ivanic 2008) and plume magmatism (Ivanic et al. 2010;Kositcin et al. 2008), as well as the isotopic work of Wyche et al. (2012b), which demonstrated that after c. 2960 Ma the rocks of the Eastern Goldfields Superterrane and Youanmi Terrane were affected by contemporaneous magmatic events. These significant differences in crustal evolution between the Eastern Goldfields Superterrane and the West Yilgarn suggest that the Ida Fault represents a long-lived lithospheric boundary between two cratonic blocks (Swager 1997;Drummond et al. 2000).
Influence of lithospheric architecture on komatiite-hosted nickel, orogenic gold and BIF-hosted iron mineral systems Komatiite-hosted nickel Spatial correlations with crustal architecture. When the spatial occurrences of the komatiites of the c. 2.9 Ga Southern Youanmi Terrane and c. 2.7 Ga Norseman-Wiluna belt are compared with the Sm-Nd isotope map (Fig. 13), there are some striking correlations. The 2.7 Ga komatiites of the Norseman-Wiluna belt (Kalgoorlie Terrane) occur on the juvenile side of a major isotopic boundary (structurally represented by the Ida Fault), and run adjacent to this boundary for c. 700 km (Fig. 13). This, together with the lack of synchronous komatiites in the Southern Cross Domain, suggests that the nature of the crust and associated margins provides a major first-order control on the emplacement of voluminous komatiite magmatism. Figure 13 identifies a distinctive 'gap' in the komatiite belt around the central Kalgoorlie Terrane (Leonora district), possibly owing to a lack of greenstone material in this area (as shown in regional mapping and aeromagnetic datasets). However, this gap correlates with a well-defined, c. 200-150 km-long, eastward deviation in the Ndisotope trend (Figure 10), from a generally linear, north-south path parallel to the Ida Fault (Figs 10 -13). This potentially suggests that the high MgO komatiites may also be deflected to the east in this area (Fig. 13).
The 2.9 Ga komatiites technically fall outside of the temporal scope of the Nd map (Fig. 13), which was constructed using only 2.8 -2.6 Ga crustal rocks. However, assuming that the Neoarchaean crustal architecture presented in Figure 10 is inherited from previous 'parent' architecture, it is possible to make some tentative correlations regarding large-scale controls.
The Southern Cross Domain hosts c. 2.9 Ga komatiites whose geochemistry, prospectivity and volcanology change markedly throughout the domain (Perring et al. 1995(Perring et al. , 1996Chen et al. 2003;Heggie et al. 2012a, b). The greenstone belts in the more reworked (T DM 2 c. 3300 Ma) Marda block are typically dominated by basalt (i.e. the lower sequence of the Marda greenstone belt; Chen et al. 2003), with komatiite occurring as rare, thin, low MgO, sheet flows. As a result, no economic nickel sulphide mineralization is known to occur in komatiites of this belt (Barnes 2006a). In contrast, the less reworked (T DM 2 c. 3100 Ma) Lake Johnston block, hosting the Southern Cross, Forrestania, Lake Johnston and Ravensthorpe greenstone belts (Cassidy et al. 2006), contains abundant komatiites, although only the latter three host nickel sulphide mineralization (Barnes 2006a). These mineralized belts contain thick, channelized, high MgO (MgO 20-40%) flows with abundant ortho-adcumulate bodies (Perring et al. 1995(Perring et al. , 1996Heggie et al. 2012a, b), which suggest high-flux eruption of komatiite magmas and high volume flow through of lava in large, submarine tubes (Hill et al. 1990Arndt et al. 2008). As a result of these favourable features, numerous nickel sulphide deposits occur within these belts including Flying Fox, Spotted Quoll (Forrestania), Maggie Hays (Lake Johnston) and RAV-8 (Ravensthorpe).
Further to this, the highest concentration of high MgO komatiites, which occurs in the Forrestania greenstone belt, is found at the terrane margin between the Youanmi Terrane/Southern Cross Domain and the South West Terrane (Fig. 13). Lu -Hf isotope work by Mole et al. (2010) demonstrates that, at 3050-2820 Ma, the eastern South West Terrane is more reworked than the south Southern Cross Domain, which has positive 1Hf values. Subsequently, during the emplacement of the c. 2.9 Ga komatiites, the South West Terrane-Lake Johnston block boundary displayed a similar isotopic architecture to the Youanmi Terrane-Kalgoorlie Terrane boundary at 2.7 Ga.
Large-scale controls. The last few decades have seen a significant increase in our understanding of komatiites, particularly their generation, volcanology, geochemistry and associated Ni-Cu -PGE mineralization (Arndt et al. 1979;Gresham & Loftus-Hills 1981;Lesher et al. 1984;Huppert & Sparks 1985b;Arndt & Jenner 1986;Groves et al. 1986;Gole et al. 1987;Campbell et al. 1989;Herzberg 1992;Hill et al. 1995;Blichert-Toft & Arndt 1999;Barnes et al. 2004;Fiorentini et al. 2010Fiorentini et al. , 2011. However, most studies were focused at the deposit scale, whereas the understanding of the controls of komatiite emplacement at the terrane to craton scale has remained largely incomplete. Building on previous work by Thompson & Gibson (1991) and Sleep (1997Sleep ( , 2005, Begg et al. (2010) used seismic tomography to demonstrate that the lithosphere can physically control the flow and lateral 'ponding' of plume material and subsequently the location of Ni -Cu -PGE deposits/ camps. The Sm-Nd isotopic data can be used to infer an approximate three-dimensional intra-cratonic lithospheric architecture (Figs 11 & 13), where highly reworked crust, typically with 1Nd , 23.0, was assigned cratonic lithospheric thicknesses of 200-250 km (Boyd et al. 1985;Begg et al. 2009), while relatively juvenile crust, with 1Nd of 1.0-3.0, was assigned Phanerozoic lithospheric thicknesses of 80 -60 km (Cawood et al. 2013). Crustal domains with 1Nd c. 0 were assigned intermediate crustal thicknesses. This proxy has been shown to work well in the Yellowstone area of the western USA (Ormerod et al. 1988;Nash et al. 2006;Manea et al. 2009;Pierce & Morgan 2009), and broadly agrees with recent geophysical studies from the Yilgarn Craton (Blewett et al. 2010b;Dentith et al. 2012, Fishwick & Rawlinson 2012. This three-dimensional crustal architecture, together with the 'plume-deflection' model of Begg et al. (2010), can be used to explain the localization of komatiite magmatism in the Yilgarn Craton.
At c. 2.7 Ga, a plume impinged on the thick, old lithosphere of the West Yilgarn  and was focused eastward into the thinner, more juvenile Kalgoorlie Terrane (Fig. 13). Concurrently, attenuation caused by the rising plume re-activated a pre-existing lithospheric weakness along the isotopic boundary, leading to north-south orientated lithosphere-scale rifting. These trans-lithospheric faults or permeability zones were essential for the unrestricted rise of komatiite magma to the upper crust/surface. Owing to the high density of komatiitic magma relative to the continental crust (Herzberg et al. 1983;Huppert & Sparks 1985a), ultramafic melts can only ascend to the surface if they are able to form a continuous column of magma from the melting zone to the top of the crust (Naldrett 2010;). Rifting at the isotopic margin allows the formation of these pathways.
As a result of these features, komatiites emplaced at the isotopic margin are highly primitive and discharged at a high rate, resulting in channelization  and the formation of multiple, world-class nickel sulphide deposits (e.g. Mt Keith, Perseverance, Kambalda camp; Rosengren et al. 2005;Fiorentini et al. 2012). In contrast, the West Yilgarn consists of old, reworked and hence thicker crust, and as a result no komatiite magmatism of any form occurs here. The closest volcanic event is the eruption of the c. 2730 Ma andesitesrhyolites (Chen et al. 2003) of the Marda Complex within the Marda block. Their Nd isotopic signatures (1Nd of c. 23; Table 2) suggest derivation from a crustal source, with geochemistry similar to a modern Andean-type margin (Chen et al. 2003), although evidence of other subduction-related features is lacking, that is, high-grade metamorphic rocks, tectonic mélanges, ophiolites, and so on (Bédard et al. 2012).
The lack of major, thick, high MgO, cumulaterich komatiite sequences further to the east of the isotopic margin, in the Kurnalpi, Burtville and Yamarna Terranes (Fig. 13), suggests that the isotopic margin provides a fundamental physical control on komatiite localization (Figs 10-13). Komatiites are known to occur in these eastern terranes ), but they are neither as thick or hot, nor as extensive as those in the Kalgoorlie Terrane, suggesting a less focused melt source.
Within the Kalgoorlie Terrane itself, there are significant differences between the Agnew-Wiluna camp (felsic volcanic-hosted) and Kambalda camp (basalt-hosted) komatiites. The flows/ sills of the northern camp typically demonstrate higher MgO values, more extensive cumulate (olivine adcumulate) zones and larger nickel deposits (Barnes 2006a;Barnes 2006b;. We suggest that this variation is controlled by the character of the crustal block and isotopic margin adjacent (west) of the related nickel camp. The crust to the west of the Kambalda camp (the Lake Johnston block) is slightly less reworked (thinner) than the block west of the Agnew-Wiluna belt (the Marda block; see Figs 10 -13). This, together with the more pronounced isotopic boundary adjacent to the Agnew-Wiluna belt, suggests that plume magmas were focused to a greater extent at the northern margin.
The c. 2.9 Ga komatiites in the Southern Cross Domain also display a strong correlation between Nd isotopic architecture and komatiite location, character and nickel sulphide endowment. Rising plume magmas at this time would have been impeded by the presence of a thick overlying lithospheric 'lid' in the South West Terrane and Marda block. Consequently, magmatic fractionation, differentiation and crustal assimilation led to basaltdominated greenstone sequences in the Marda block (Chen et al. 2003), and a complete lack of preserved volcanic rocks in the South West Terrane. In these environments, ore-forming processes that require turbulent emplacement of primitive magmas are not favoured (Lesher et al. 1981(Lesher et al. , 1984Bekker et al. 2009).
In contrast, the Forrestania and Ravensthorpe belts are located along an inferred craton margin. This creates a lithospheric setting similar to that in the Eastern Goldfields at 2.7 Ga, where the South West Terrane forms the old, thick crustal block and the Lake Johnston block the thinner crust (see Fig. 13). As a result, similar plume-deflection processes (Begg et al. 2010) inferred for the 2.7 Ga komatiites led to the emplacement of komatiite magmas with the properties needed to form nickel sulphide deposits. In addition, the magnetotelluric work of Dentith et al. (2012) potentially shows the 'fossil' remnants of the lithospheric discontinuities that allowed komatiitic magmas to reach the surface unimpeded.
The Southern Cross komatiites appear to form a subgroup between the unprospective, low-MgO Marda flows and the prospective, high-MgO, Forrestania, Ravensthorpe and Lake Johnston magmas ). These komatiites have high MgO contents, but appear to be unchannelized and have significantly less cumulate than the flows to the south. The transitional nature of these flows correlates with their location at the margin of the Marda and Lake Johnston blocks (Fig. 13).
In summary, the variable spatial location, character and nickel sulphide endowment of komatiites in two separate areas of the Yilgarn Craton, at two distinct times, appear to be governed by the same basic lithospheric architecture. The occurrence of a thin, young, juvenile crustal block adjacent to an older, thicker, more reworked block, together with the impingement of a mantle plume, appear to be vital to the formation of thick, channelized, high flux MgO-rich komatiite sequences and the onset of nickel sulphide ore-forming systems.

Orogenic gold
Spatial correlations with crustal architecture. While gold deposits can be found throughout the Yilgarn Craton (see Figs 3 & 12), they are particularly concentrated in the Kalgoorlie and Kurnalpi Terranes (e.g. Laverton belt) in the Eastern Goldfields Superterrane (Robert et al. 2005) and the central Murchison and Southern Cross Domains of the Youanmi Terrane.
In the Murchison Domain, gold deposits (e.g. Mt Magnet gold camp) are concentrated internal to, and on the margins of, the 'rift-like' juvenile architecture (Spaggiari 2006;McCuaig et al. 2010). The density of deposits appears to increase in the northern area of the 'rift' where large crustal structures appear to converge (Fig. 12).
In the Eastern Goldfields Superterrane, gold deposits are concentrated in the Kalgoorlie and Kurnalpi Terranes (Figs 3 & 12), with the Kalgoorlie -Norseman belt particularly prospective (Blewett et al. 2010a, b). Most deposits occur along major, terrane bounding structures or secondary structures (Groves et al. 1995;Witt & Vanderhor 1998;Robert et al. 2005). In general, the gold systems in the Kalgoorlie and Kurnalpi Terranes follow the isotopic boundary between the Eastern Goldfields Superterrane and West Yilgarn. Despite being the most juvenile crustal domain in the Yilgarn Craton (Fig. 10), the Kurnalpi Terrane does not appear to be more prospective than the Kalgoorlie Terrane. This suggests that, while juvenile crust is important for the formation of gold systems, large-scale structural networks may be more important controls (Vearncombe 1998;Cox 1999;Blewett et al. 2010b).
This relationship is exemplified by the Southern Cross gold belt (e.g. Marvel Loch gold camp), which occurs along a major structure between the South West Terrane and Southern Cross Domain: two isotopically coherent regions which are both relatively reworked compared with the Murchison Domain and Kalgoorlie/Kurnalpi Terranes (Fig.  12). Although this margin is poorly defined in the Sm-Nd maps (Figs 10 -12), time-resolved Lu -Hf mapping (Mole et al. 2010), together with granite age (U -Pb) mapping , show that this margin was much more pronounced at c. 2820-2720 Ma. As a result, the Southern Cross area demonstrates the significance of inherited, early architecture c. 200 myr older than the gold mineralizing event.
Large-scale controls. Gold deposits of the Yilgarn Craton generally formed in a craton-wide event at c. 2650-2630 Ma (Kent & McDougall 1995;Kent & Hagemann 1996), typically along lines of steep 1Nd gradient (Fig. 12) and in areas where multiple regional structures intersect and/or change direction (Cox 1999;Chen et al. 2001b;Blewett et al. 2010a, b). The large-scale hydrothermal systems that drove gold mineralization used lithospheric-crustal scale structures localized by the lithospheric architecture (Blewett et al. 2010b). This relationship is particularly clear in the gold-rich areas of the Murchison Domain, Southern Cross Domain and Kalgoorlie Terrane (Figs 3 & 12).
Gold mineralization in these areas is controlled by pre-existing lithospheric architecture, in tandem with major structures. This relationship is due to a number of factors: (1) The addition of juvenile material into the crust before the initiation of a gold-forming event is critical to developing gold fertility Bateman & Bierlein 2007, Hronsky et al. 2012). This can happen at anytime before gold mineralization, and suggests that the fertility of the Southern Cross and Murchison gold sources developed 350 -150 myr before the gold mineralizing event at c. 2650-2630 Ma, as this is when juvenile material was added to the crust, creating a fertile source. However, the delay between juvenile input and gold mineralization in the Southern Cross district may explain the relatively small size of these deposits ). In the Eastern Goldfields Superterrane, relatively juvenile crust (1Nd . 0) dominated before and during gold mineralization. This suggests that the input of juvenile crust before and during gold mineralization created an especially gold-rich source, possibly explaining why deposits in this area are typically larger than those in the West Yilgarn.
(2) Deep-seated granite magmatism that drives large-scale hydrothermal systems is controlled by lithospheric architecture, that is, the geometry of the lithosphere controls where melts can rise and decompress. The late, 2650-2620 Ma Low-Ca granites, broadly synchronous with gold mineralization Cassidy et al. 1998;Cassidy et al. 2002), are often localized at terrane boundaries and major structures where they can drive hydrothermal systems. These granites transferred large amounts of high heatproducing elements (U, Th, Rb) from the lower to upper crust (Champion & Sheraton 1997;Cassidy et al. 2002;Czarnota et al. 2007). This led to rapid cooling and cratonization of the lower crust, and also a significant increase in heat-flow in the upper crust, potentially responsible for the large hydrothermal systems involved in the craton-wide gold event.
(3) The 'Mafic' group of granites (sanukitoids; Cassidy et al. 2002) are associated with juvenile crust and appear to be an important factor in gold mineralization (Cassidy et al. 1998(Cassidy et al. , 2002. These granites are relatively high in Ni, Cr and Mg# as well as large ion lithophile elements (Ba, Sr, light REE; Champion & Sheraton 1997) and such alkalic mafic magmas may have represented an important gold source (Wyman & Kerrich 1988;Müller 2002;Hronsky et al. 2012;Duuring et al. 2007). The Mafic granites utilized the same structures as gold and were emplaced at the start of the gold event at c. 2650 Ma (Cassidy et al. 2002). As a result, they are the preferred host for granite-hosted orogenic gold deposits (Cassidy et al. 1998;Duuring et al. 2007). These granites represent a link between the mantle, crust, gold source and deposition site and demonstrate the importance of juvenile crust to the gold fertility of a terrane.  (Groves 1993, Sibson 1994Witt & Vanderhor 1998;Vearncombe 1998;Cox 1999). These boundaries are also likely to be reactivated multiple times in their history, leading to increased permeability and the formation of the larger gold systems (Blewett et al. 2010a). In the Murchison Terrane, these structures appear to reach the mid-crust at c. 30 -20 km, as shown by regional seismic data (Ivanic et al. 2013). (5) Fluid flow -isotopic contrasts typically represent the boundary between rheologically and chemically different crustal blocks (Burov et al. 1998). These features, together with strain partitioning and inherited tectonic weakness, encourage the creation of largescale, crustal and lithospheric faults together with complex subsidiary structural regimes Blewett et al. 2010a, b). This complex and deep-tapping structural network allows the establishment of large hydrothermal cells (Groves et al. 1995;Vearncombe 1998;Cox 1999), possibly fuelled by deep-seated granites (Cassidy et al. 2002).
The fluids would then have potentially fluxed through mafic alkaline igneous rocks, such as the Mafic granite suite (Cassidy et al. 2002;Duuring et al. 2007), picking up gold and depositing it further down the system in a trap or buffer (Groves et al. 1995Hronsky & Groves 2008;McCuaig et al. 2010;Hronsky et al. 2012). (6) Heat flow -the presence of younger, juvenile crust indicates that the lithosphere in that area may be thinner than in more reworked areas. This is because juvenile crust requires a variable amount of mantle input, whereas reworked signatures indicate crustal melting (Kemp et al. 2007). This is important as thinned crust allows more rapid and effective heat transfer to the upper crust, encouraging the generation of large hydrothermal cells and maintaining them over time (Champion & Sheraton 1997;Cassidy et al. 2002Cassidy et al. , 2005Cassidy & Champion 2004;).

BIF-hosted iron
Spatial correlations with crustal architecture. Iron deposits of the Yilgarn Craton are spatially associated with older, more reworked crustal domains, with groups of deposits occurring internal to, and on the margins of, older crustal blocks of the West Yilgarn (Fig. 12). In contrast, BIF-hosted iron deposits appear rare to absent in the juvenile Eastern Goldfields Superterrane, where sulphidic black shales dominate sedimentary sequences and thick BIF units are lacking (Swager 1997;Barley et al. 2003;Bekker et al. 2009).
Within the West Yilgarn, three main iron camps occur in the Jack Hills, Murchison and central-north Southern Cross Domain areas (see Fig. 12). The Jack Hills camp (Crosslands, Mt Narryer, Taylor Range, Mt Hale) occurs within the Narryer Terrane (Occhipinti et al. 2001;Spaggiari et al. 2007), which is the oldest crustal terrane delineated by the Nd isotopes (see Figs 6g, h & 8h). BIF-hosted iron deposits in the Murchison Domain are mainly located on the margins of the NW-SE-trending juvenile crust (Figs 10 & 12). For example, the Weld Range deposits (e.g. Beebyn, Weld Range; Duuring et al. 2013)  Large-scale controls. BIF-hosted iron deposits require the deposition of thick primary BIF sequences. These provide the iron source, which is later upgraded by silica-poor fluids localized by structures. This section demonstrates how lithospheric architecture potentially controls the location and extent of both processes.
The distinct absence of documented high-grade, BIF-hosted iron deposits in greenstone belts in the Eastern Goldfields Superterrane may be due to the rarity of laterally continuous, thick BIF in this region (Swager 1997;Gole 1981). Differences in ocean chemistry and crustal evolution during the Archaean may influence the primary abundance, distribution and composition of BIF (Gole 1981;Lascelles 2007;Bekker et al. 2010;Evans et al. 2012). Thus, the different crustal evolution of the West Yilgarn compared with the Eastern Goldfields Superterrane best explains the first-order differences in iron ore abundance between these areas.
In the West Yilgarn, the localization of iron deposits in the more reworked areas of the crust (Fig. 12) correlates with regional variations in the abundance of BIF (Gole 1981). This suggests that the older, more evolved and stable crustal regions were analogous to continental platform settings (Bekker et al. 2010). The edge of these crustal blocks could have represented passive margins, and this is supported by the presence of lowtemperature hydrothermal VMS systems in the BIF and felsic volcanic footwall rocks of the Weld Range iron (Duuring & Hagemann 2013a, b). In these areas volcanism was episodic and of a relatively low-flux, resulting in a more stable environment for the accumulation of thick BIF sequences (Bekker et al. 2010). In contrast, relatively juvenile regions represent more active, unstable environments where more continuous, rapid emplacement of komatiites and basalts, as well as associated hydrothermal systems, led to thinner BIF sequences (Lascelles 2007). This spatial variation can be observed in the different stratigraphy of the Marda (thick BIFs, basaltic volcanism) and Forrestania (thin BIFs and abundant komatiites) greenstone belts (Chen et al. 2003;Angerer & Hagemann 2010;Angerer et al. 2012a).
In general, there is an antithetic relationship between orogenic gold and BIF-hosted iron camps in the Yilgarn Craton (Fig. 12). This is because the fluids needed to form iron ore deposits must be silica-poor in order to remove chert from the BIF (Duuring & Hagemann 2013b). However, gold is invariably found associated with quartz, suggesting that a silica-rich fluid is involved (Phillips & Groves 1983). This is exemplified at the Mt Morgans gold camp in the Kurnalpi Terrane, where extensive hydrothermal alteration by the SiO 2 -H 2 O-S-CO 2 -rich gold-bearing fluids resulted in the formation of quartz -carbonate-sulphide -gold veins, whereas addition of silica to the BIF resulted in the dilution of the iron content (Vielreicher et al. 1994). Consequently, as iron and gold show a preference for crustal domains with different histories, there is a correlation between fluid composition and crustal source, whereby older crust appears to favour silica-poor fluids, and juvenile crust silica-rich fluids. Further work is required to understand this relationship between fluid source and crustal evolution.
Large-scale structures, such as those described above for gold systems, are also fundamental in iron systems (Angerer & Hagemann 2010;Duuring et al. 2013). Iron camps (e.g. Weld Range, Windarling) form along major structures orientated parallel to isotopic margins (Angerer et al. 2012a;Duuring & Hagemann 2013a, b), suggesting an intimate regional relationship between major crustal boundaries and these localizing faults (Fig. 12). † Periods of reworking are inferred to represent compressive events, which led to orogenesis and the production of infra-crustal granites. Juvenile addition signifies crustal thinning, allowing magmatism to tap more mantlederived, younger sources. † Mantle input into the crust after c. 3000 Ma appears to be minor in the West Yilgarn. This may be due to thicker crust preventing mantle melts interacting with the mid-upper crust. Alternatively, owing to the older signature of the West Yilgarn crust, juvenile input does not necessarily result in .0 1Nd values. † Multiple periods of craton-wide granite magmatism, which tap a number of shared crustal sources at ,2700 Ma, infer that the Yilgarn Craton was assembled by this time. Older shared features, such as greenstone ages and plume magmatism in the Burtville and Youanmi Terranes, suggest an even older relationship as far back as c. 3000 Ma. † Crustal growth events, marked by the addition of mantle material, are inferred to have occurred at c. 3100-3000 Ma (formation of the Burtville Terrane, Murchison rift and Southern Cross komatiites), c. 2900 Ma (Yamarna Terrane formation?), c. 2800 Ma (Murchison mafic-ultramafic intrusions, Duketon komatiites, decrease in Eastern Goldfield Superterrane model ages) and c. 2720-2700 Ma (crustal stretching and thinning between Youanmi-Burtville and formation of Kalgoorlie and Kurnalpi Terranes, Norseman-Wiluna komatiites). † BIF-hosted iron systems are concentrated in the older, more reworked crust of the West Yilgarn, internal to, and on the margins of, old crustal blocks. † Orogenic gold systems typically occur internal to, and on the margins of, juvenile blocks. However, the occurrence of the Southern Cross gold belt in reworked crust demonstrates that the addition of juvenile material, which creates gold fertility in the crust, does not have to be synchronous with gold mineralization. The structural complexity common at the margins of crustal domains, together with juvenile crustal material where available, is a favourable location for gold mineralization. † Gold mineralization throughout the craton appears to be coeval with cratonization and the last tectono-thermal event at c. 2640-2620 Ma. † Komatiite-hosted nickel deposits are preferentially localized on the juvenile side of a margin with a more reworked crustal block. This architecture focused the plume source into the thinner juvenile crustal domain. Rifting occurred at or close to the isotopic margin, which represented a pre-existing crustal weakness. The formation of trans-lithospheric faults/pathways at this margin allowed komatiite magmas to be erupted unfractionated and with minimal crustal contamination. † Lithospheric architecture is fundamental in localizing the source reservoir for a given commodity. In gold and nickel, the emplacement of juvenile magmas is governed by the age and configuration of crustal domains. In BIF-hosted iron systems, the development of thick BIF sequences that form the iron source is controlled by the age and stability of a crustal block.
Lithospheric architecture is a first-order control of mineral systems, and an important tool in assessing and understanding regional prospectivity. Our understanding of the effects of large-scale crustal architecture on camp/deposit localization is at an early stage, and further work needs to be done to understand how evolving lithospheric architecture controls the movement of prospective zones in space and time.