Introduction

The likelihood of metamorphic overprinting increases with age, with many of our planet’s most ancient rocks now pervasively modified by later geological processes1. Specifically, many ancient mafic rocks are now amphibolites, reflecting the metamorphic redistribution of water. Dating the formation age for the protolith of amphibolites is notoriously difficult, due to the conversion of the primary mineralogy to phases that are more stable under subsequent temperature, pressure, and chemical conditions, and the general lack of primary zircon. The result of such overprinting processes is the variable loss of isotopically defined precursor age information. Nonetheless, magmatic dykes, converted to amphibolites, may retain an informative refractory xenocrystic cargo from deeper crustal levels or indeed the upper mantle, entrained on the primary magma’s emplacement pathway2,3.

Western Australia hosts an important archive of early Earth crustal evolution, with the oldest rocks on the continent preserved in the Manfred Complex, Narryer Terrane, with a magmatic crystallization age of c. 3730 Ma4. The denuded remains of even more ancient crust is retained as up to 4374 Ma detritus in the famous Narryer Terrane Jack Hills greenstone sequence5. Yet, outside the Narryer Terrane, elsewhere in the Yilgarn Craton, Nd and Hf isotope systematics have pointed towards a cryptic crustal substrate of Eoarchean to Paleoarchean age6. Specifically, very old > 4000 Ma whole rock Nd and zircon Hf model ages are known from samples of the northeastern part of the Archean Youanmi Terrane7. Additionally, detrital zircon grains from quartzite units of the Archean Southern Cross Domain in the central Yilgarn Craton have zircon ages ranging from c. 4350 Ma to c. 3130 Ma8 and rare >3100 Ma inherited zircon exists in some Youanmi Terrane granites and felsic volcanics9. Together these rather sparse data point towards ancient components, throughout the Yilgarn Craton, with temporal affinity to Narryer Terrane crust, the potential ancestral nucleus of the craton.

The South West Terrane defines the southwestern corner of the Yilgarn Craton and comprises granitic rocks, metasedimentary rocks, migmatite, and mafic to felsic gneisses10,11. The granitic rocks yield magmatic crystallization ages of 2704–2607 Ma, whereas metamorphic overprinting highlights the effect of the Proterozoic c. 1095–990 Ma Pinjarra Orogeny and 780–515 Ma Leeuwin Orogeny along the western margin12,13. Mafic rocks in the South West Terrane are relatively poorly dated, in part due to a lack of suitable geochronometers, and intense metamorphic overprinting14. Nevertheless, on a cratonic scale numerous dyke swarms are known to widely intrude the ancient cratonic basement, including the voluminous 2408–2401 Ma Widgiemooltha large igneous province15,16 and c. 1210 Ma Marnda Moorn large igneous province related suites17,18,19. Other dyke swarms that have been identified in the craton include; c. 2615 Ma Yandinilling Dolerite20, c. 1888 Ma Boonadgin Dolerite21,22, c. 1390 Ma Biberkine Dolerite23, c. 1075 Ma Warakurna large igneous province24, and c. 735 Ma Nindibillup Dolerite25. Despite some of these suites having known expressions only within the central and northern part of the craton, the NE-trending Yandinilling Dolerite, the E-trending Widgiemooltha Dolerite, the WNW-trending Boonadgin Dolerite, the NNW-trending Biberkine Dolerite and the NW-trending (c. 1210 Ma) Boyagin Dolerite crop out in the South West Terrane.

The South West Terrane forms the southwestern corner of the Archean Yilgarn Craton (Fig. 1). It is bounded to the south by the Paleo to Mesoproterozoic Albany–Fraser Orogen, to the west by the Mesoproterozoic Pinjarra Orogen and overlying Permian to recent Perth Basin and to the northeast by the Archean Youanmi Terrane. The South West Terrane has been recently redefined as exposing relatively young granitic rocks with 2704–2607 Ma magmatic crystallization ages compared to the neighbouring 3018–2600 Ma Youanmi Terrane10. Relatively undeformed biotite-monzogranite form the majority of the outcrops within the terrane and are comparatively well-dated. Granitic gneisses and migmatites are also locally voluminous, but their ages are poorly known, leaving the possibility that older basement components are exposed in parts of the South West Terrane14. Whole-rock Sm–Nd isotopes from granites in the South West Terrane indicate two-stage depleted mantle model ages of 3430–2880 Ma26, implying some Paleoarchean basement/inheritance. A suite of relatively young (2619–2607 Ma) granites are exposed towards the southwestern margin of the South West Terrane14,27.

Fig. 1: Geological map of the southwest Yilgarn Craton, Western Australia.
figure 1

Lithological divisions of the South West Terrane, overlying the 1:500,000-scale state map. Colours denote lithological units, where pinks are granitoids, greens are mafic intrusive rocks, and browns are sedimentary units. Inset depicts the terrane and domain structure of the craton. Red rectangle over the South West Terrane denotes detailed geology map provided as Supplementary Fig. A2.

Supracrustal rocks of the South West Terrane, with the exception of the Saddleback greenstone belt, are dominated by siliciclastic lithologies of largely unknown true depositional age due to a lack of datable volcanic horizons14. Psammitic and pelitic rocks collected from the Toodyay and Bridgetown areas show maximum depositional ages of c. 3200, 2670, and 2600 Ma14. These rocks commonly contain Meso- to Paleoarchean detrital zircons, with only a single Eoarchean detrital zircon dated at 3676 ± 12 Ma14. Unradiogenic Lu–Hf data measured from detrital zircon grains from modern beaches of southwestern Australia also suggest the presence of isotopically coherent Hadean-Eoarchean crustal vestiges associated with the southwestern border of the Yilgarn Craton28.

NW-striking shear zones, possibly originating during the late Archean and reactivating during the Proterozoic, dissect the southwest corner of the South West Terrane10. These structures are truncated by north-striking faults and shear zones of the proto-Darling Fault. These proto-Darling Fault structures affect c. 1210 Ma dolerite dykes and are likely related to the 1095–990 Ma Pinjarra Orogeny and the 780–515 Ma Leeuwin Orogeny29. A time space diagram of the region is provided as Supplementary Fig. A1.

The Archean pressure and temperature conditions of the South West Terrane are mostly unknown due to the general lack of suitable metamorphic assemblages. Nonetheless, greenschist- and amphibolite- to granulite-facies conditions have been inferred for some exposed gneisses and migmatite12. Along the Darling Fault, amphibolite facies pressure-temperature conditions overprinted Archean mineral assemblages with metamorphic zircon and monazite growth at 1840, 1190, 1145, 1040, and 700 Ma14. Biotite Rb–Sr data from the Yilgarn Craton define a strong E-W trend in cooling ages, from 2480 Ma in the central Yilgarn Craton to younger than 500 Ma at the western margin of the craton close to the Darling Fault30,31. These ages have been interpreted to reflect reworking or slow cooling in the central Yilgarn Craton after cratonisation, with the western margin also cooling following Gondwana assembly during the Ediacaran–Cambrian32. Near the dyke sample reported here (SW-1), the peak pressure–temperature estimate from pseudosection modelling is 5.5 kbar and 633 °C, best constrained at 707 ± 5 Ma based on metamorphic zircon and 667 ± 25 Ma by metamorphic monazite14.

Here we present the results of U–Pb geochronology on an amphibolite dyke from the South West Terrane (Supplementary Fig. A2) whose U-bearing accessory mineral cargo appears to track much of the Proterozoic and Mesozoic geological history of this margin. Importantly, the zircon inclusion cargo within titanite implies an inherited component of 3440 Ma age transported within the dyke. We complement these findings with detrital zircon geochronology from a sample of the Swan-Avon river system. This river system represents the largest fluvial system in the region, draining over 120,000 km2 of the South West and Youanmi terranes, with a palaeo-Swan River inferred to have been flowing in broadly the same position for at least 60 Ma33. These detrital zircon results also support the presence of a Paleoarchean crustal component well to the south of the Narryer Terrane, or the widespread mobilisation of ancient detritus of this age.

Results

Sample SW-1, investigated in this work, is an amphibolite dyke (Fig. 2) within the southwestern edge of the South West Terrane, approximately 12 km east of the Darling Fault (Supplementary Fig. A3). The major element chemistry of SW-1 reflects a basaltic andesitic protolith, with a relatively evolved composition (SiO2 = 54.5 wt%, MgO = 4.7 wt%, Mg# = 45), including high TiO2 (1.8 wt%), with clear tholeiitic affinity. Relatively steep whole rock LREE/MREE ([La/Sm]N = 2.65) but flat MREE/HREE ([Tb/Yb]N = 1.33) patterns further indicate a relatively evolved composition (Fig. 3).

Fig. 2: Photomicrographs of SW-1.
figure 2

a transmitted light, b reflected light, c automated mineralogy map. ap apatite, qtz quartz, amph amphibole, ttn titanite, ep epidote, plag plagioclase.

Fig. 3: Whole rock geochemistry multi-element plot.
figure 3

Geochemistry for sample SW-1 compared to the Biberkine, Boonadgin, and Marnda Moorn dykes. Lcc Lower continental crust, OIB Ocean island basalt, NMORB Mid ocean ridge basalt, EMORB Enriched MORB. Refer to reference23 for comparison source data.

Dyke apatite

Apatite U–Pb analyses mainly define a single mixing line between radiogenic and common Pb isotopic components (Fig. 4). Thirty-six analyses have low U content and negligible radiogenic Pb ingrowth, excluding two statistical outliers, 34 individual apatite analyses define a weighted mean 207Pb/206Pb ratio of 0.937 ± 0.012 (MSWD = 1.7). This isotopic ratio is interpreted as an estimate of the common Pb signature of the rock, consistent with titanite and apatite data fanning away from this locus in 207Pb/206Pb versus 238U/206Pb space (Fig. 4). Excluding one old outlier, a regression through those apatite analyses with U concentrations >0.5 ppm, anchored on this common Pb estimate, yields an age of 211 ± 19 Ma (MSWD = 1.6; n = 14), interpreted as the dominant time of apatite closure to radiogenic Pb diffusion.

Fig. 4: U–Pb isotopic results from SW-1.
figure 4

a Weighted mean 207Pb/206Pb for common Pb component within apatite (blue filled bars denotes outliers). b Stacked histograms and probability density plot for different minerals using apparent 207Pb-corrected ages (colours as per part c), with best estimate of dyke crystallization age shown by vertical dotted line. c Inverse concordia plot for titanite, zircon, and apatite. Grey triangle denotes field of mixtures between ancient zircon inclusion component and the common–radiogenic titanite mixing array. Error bars are all at the two sigma level.

Dyke titanite

Titanite U–Pb isotopic ratios predominantly define a cone of data on the Tera-Wasserburg plot, scattering between an apparently single common Pb composition with variable proportions of radiogenic Pb. Assuming the apatite-defined common Pb value, 207Pb-corrected ages range from c. 1941 to 347 Ma, expressing a parametric distribution, with an age mode at c. 950 Ma and common Pb proportions (F207%) between 93% and 13%. Trace element compositions from split stream analyses reveal depletions in LREE relative to heavy rare earth element (HREE) and a slightly negative Eu anomaly relative to chondrite ([Eu/Eu*]CN = 0.86, where Eu* = √Sm × Gd; Fig. 5a), consistent with titanite growth in a mafic–intermediate rock34. Older apparent ages are broadly associated with higher total REE compared to younger dates, implying coupled Pb–REE mobility led to younger dates (Fig. 5a). Rarely, some titanite grains are enriched in LREE, likely a function of inclusions, and such analyses are unlikely to yield reliable ages. Using only the oldest portion of grains, consistent with LREE-depleted chemistry, defines a mixing line between radiogenic and common components, with a concordia intercept age of 1364 ± 56 Ma and a 207Pb/206Pbi intercept of 0.947 ± 0.034 (n = 22; MSWD = 0.9).

Fig. 5: Trace element compositions of SW-1 titanite compared to U–Pb geochronology.
figure 5

a REE spider diagram (normalised to chondrite), colour-coded for apparent 207-corrected age. Violin plot of element distribution overlaid on REE profiles. b λ1 parameter (slope) plotted against apparent age, colour-coded for apparent Zr-in-titanite temperature, with arrows representing interpreted processes.

The slope of the chondrite-normalized titanite REE patterns, quantified using the first order shape coefficient35 λ1 increases towards younger apparent ages. Similar REE patterns have been interpreted to reflect coupled dissolution of a trace element rich titanite with simultaneous reprecipitation of a new, trace element poor titanite (Fig. 5b). A small number of analyses have anomalously elevated LREE contents, which positively correlate with Th, but not U, consistent with incorporation of allanite inclusions. The Zr-in-titanite thermometer (calculated with an average crustal pressure of 5.5 kbar and titania and quartz activities of 1 and 0.7, respectively;36) implies a median (re)crystallization temperature of ~680 °C, with outlier higher calculated temperatures of ~950 °C associated with older apparent ages (Fig. 6c). Such higher calculated temperatures are due to the presence of zircon inclusions in the titanite and regarded as geologically meaningless.

Fig. 6: Grain geochemical signatures from SW-1.
figure 6

a Hf isotope composition of small metamict discrete zircon grains analysed by split stream laser ablations. Black squares denote ɛHf = apparent 207Pb-corrected age. Uncertainties are shown at the two sigma level. The red Pb loss line depicts the trajectory for radiogenic-Pb loss from a a 1390 Ma neoblastic dyke component and b an Archean xenocrystic component inherited into the dyke. The Hf isotopic compositions indicate the zircon cannot be Archean xenocrysts that have lost Pb, yet retained primary Hf. MORB-DM depicts the average composition of modern MORB extrapolated to ɛHf = 0 at 4500 Ma. The CHUR parameters are provided in the Supplementary Information. Whole rock Nd converted to ɛHf assuming a terrestrial array relationship plots at the blue star symbol. Hf isotopic data were not collected for zircon inclusions in titanite as these inclusions are below the spatial resolution of the method. b Laser ablation signal (ppm) concentrations of Ti, Zr, Th, and U during ablation highlighting the ablation of a zircon inclusion in titanite. Time-resolved integrations are set on consistent periods to subsample the desired mineral component. c Plot of Zr versus F207% in titanite from SW-1. Data points are colour-coded for apparent 207Pb-corrected age. Zircon inclusions show higher Zr content, less discordance, and older apparent ages.

Dyke matrix-hosted zircon

Discrete zircon grains under CL emission reveal low response, high-U cores, overgrown by high CL response rims (Supplementary Fig. A4). Cores have mottled textures and are inclusion rich, consistent with their metamict state. Nineteen analyses are concordant to discordant and spread within the broad 1000–600 Ma section of the concordia curve. Two <10% discordant core analyses with the lowest U ( ~ 60 ppm) and Th ( ~ 80 ppm), and thus least likely to have lost radiogenic Pb, yielded 206Pb/238U ages of c. 1044 and 1012 Ma. The youngest concordant analysis was located on a low CL response, low U rim, and indicates a 206Pb/238U age of c. 581 Ma. The range of these apparent ages are interpreted to reflect physical mixtures between the intimately inter-grown cores (c. 1000 Ma) and rims (c. 600 Ma), and also the effects of Pb mobility. Lower U overgrowths appear to have higher proportions of common Pb (a function of lower absolute radiogenic Pb), whereas the higher U cores seem more influenced by radiogenic Pb loss.

Hf isotopes are less susceptible to diffusion in zircon than Pb, nonetheless, only a few analyses likely sampled entirely core or rim components. On a Hf evolution plot, the data scatter from CHUR-like compositions at c. 1000 Ma to slightly more evolved values through time (Fig. 6a), consistent with radiogenic Pb loss.

Dyke zircon inclusions in titanite

Several titanite grains contain small inclusions of zircon, which are typically less than 30 microns in width (Fig. 7). Analysis of small ( < 30 µm-wide) zircon inclusions in titanite can be facilitated by cropping the integration window of the time-resolved U–Pb signal, with reference to the concurrent element compositions. Furthermore, because of the much higher U content in the zircon relative to the titanite, zircon inclusions will dominate any mixed age (Fig. 6b). One highly important example of a zircon inclusion being intersected within titanite is provided by grain 13 (Supplementary Data). Figure 8 depicts a time-resolved ablation signal for this grain, which in its last 15 s intersected material with higher U and Th content, in comparison to titanite ablated in the first 15 s. Considering a sequence of time resolved integrations through this analysis, as a mechanism to better understand the changing isotopic ratios through this analysis, we plot 238U/206Pb and 207Pb/206Pb ratios from a moving three second integration window through the entire ablation, on an inverse concordia plot (Fig. 8). This visualisation reveals late ablations in the high U and Th component, interpreted as zircon (steps 7–9), clustering on the concordia curve around c. 3400 Ma. Earlier ablations (steps 1–5) in titanite scatter along a mixing line defined by Proterozoic common Pb and a radiogenic component at c. 1100 Ma, consistent with other titanite analyses in this sample. An ablation directly on the zircon side of the inclusion boundary is discordant and within the common-radiogenic mixing array, likely reflecting a combination of signals. By way of example, a deliberate mixture between components was calculated and sits along a second mixing line that ranges between the concordant zircon component and the locus of the pure titanite analyses. Taking a longer integration from 30 to 45 s, within the apparently unmixed portion of the time-resolved signal, yields a concordia age of c. 3440 Ma (integration ID: 13; Supplementary Data), hosted in the core of a titanite grain with a 207Pb-corrected age of c. 1100 Ma (Fig. 4). The Paleoarchean age of this zircon inclusion is interpreted as the magmatic crystallization age of an inherited component. A further two analyses of inclusions are concordant, indicate high Zr concentration, and yield concordia ages of c. 1602 Ma and c. 1387 Ma, interpreted as the ages of other inherited (or in the latter case likely neoblastic) zircon components. Other inclusions yield a range of 207Pb-corrected ages between 2093 and 956 Ma, peaking at 1260 Ma. Based on trace element chemistry and apparent U–Pb systematics, these ages are variably contaminated with titanite. Those mixtures dominated by zircon are less discordant, whereas those dominated by titanite are more discordant, due to a higher proportion of common Pb, consistent with less U (Fig. 6c). Importantly, the distribution of data on the inverse concordia plot supports the presence of an inherited Paleoarchean zircon component, variably mixed into the titanite common-radiogenic mixing array (Fig. 4c).

Fig. 7: Images of titanite grains with mineral inclusions.
figure 7

a, b Transmitted light photomicrograph mosaic of titanite grains within SW-1, revealing large number of zircon inclusions. c Backscatter electron image of zircon inclusion within titanite. Ttn titanite, amph amphibole, zr zircon.

Fig. 8: Time-integrated laser ablation signal for SW-1.
figure 8

a Time-integrated laser ablation signal with multiple integrations set through the run to demonstrate the effect of sampling titanite versus an Archean zircon inclusion. Apparent concentrations in ppm based on MKED titanite reference material. b Inverse concordia plot for the same integrations with the step number corresponding to the section of the signal as highlighted with boxes in “a”. Uncertainty ellipses are shown at the 2 sigma level.

River sediment detrital zircon

This sample yielded abundant rounded zircon grains that range from colourless to pale brown to black. They are 50–300 µm long with aspect ratios up to 5:1. In cathodoluminescence (CL) images, many grains display concentric growth zoning that is truncated at grain edges (Supplementary Fig. A5). Three hundred detrital zircon analyses from sample SW-2 (Swan-Avon river system), yield 213 concordant analyses, specified as having a log ratio distance of less than ≤ 5%, to the calculated concordia age point37. Of the concordant component three analyses are excluded from further consideration as they show erratic down hole ablation response. The remaining 210 analyses define detrital zircon age modes at (in significance order), 2645 Ma, 2699 Ma, 3243 Ma, 3206 Ma, and 3172 Ma, by 55, 22, 19, 18, and 15 analyses, respectively. Furthermore, there are several subordinate age modes including those at 2799 Ma, 3334 Ma, 3493 Ma, defined by nine, four, and four analyses, respectively. These modes are interpreted as the ages of zircon-crystallizing rocks in the detrital source region(s), or as the ages of derivative detrital components within precursor (meta)sediments that have been reworked into this sediment. There are four analyses younger than 2500 Ma with the youngest yielding an age of 517 ± 4 Ma, while there are 71 analyses greater than 3000 Ma in the sample, with 61 between 3353 and 3141 Ma and six greater than 3400 Ma.

Discussion

Apatite, titanite, and zircon have different propensities to resetting of their radiogenic Pb cargo, depending on the thermal and fluid conditions in their local environment, the physical condition of the mineral’s crystal lattice, and also the nature of any fast diffusion pathways connecting to the grain38,39,40. As we discuss below, apatite provides information on late-stage exhumation, titanite on dyke crystallization and Neoproterozoic orogenic events, discrete zircon crystals on Neoproterozoic events and shielded zircon inclusions in titanite on dyke crystallization and a potential window into the ancient components of the southwestern Yilgarn Craton. These findings on the deep geology of the South West Terrane from zircon inclusions within titanite crystals of a dyke, are supported by analysis of detrital zircon grains from a large-scale modern river, in part draining the South West Terrane.

Dyke crystallization age: The U–Pb systematics of titanite imply it dominantly reflects a two-component mixture between radiogenic and common Pb, with this common component best estimated through low U apatite with a 207Pb/206Pb ratio of ~0.937, consistent with the Pb growth model for Proterozoic crust41. An additional complication of a third radiogenic Pb contribution exists from apparently older zircon inclusions shielded within the titanite. Nonetheless, when screened for inclusions using trace element chemistry the radiogenic end of the titanite mixing array supports an interpretation of a partial recrystallization event during various events in the Neoproterozoic. The oldest component within the titanite, when not deleteriously affected by zircon inclusions, defines a mixing line between common Pb and a c. 1360 Ma radiogenic component, whereas other analyses suggest susceptibility to later events (Fig. 4).

A relationship of decreasing REE slope with younger age for inclusion-free titanite is also consistent with a recrystallization process for these younger dates. The implication is that the c. 1360 Ma radiogenic component provides, conservatively, a minimum crystallization age for the dyke. Furthermore, several zircon inclusions yield ages consistent with Proterozoic inherited components, with the youngest concordant zircon inclusion yielding an age of c. 1390 Ma, near coeval with the oldest titanite generation, supporting dyke magmatic crystallization at around this time. Other mildly discordant analyses that are zircon inclusion–titanite mixtures imply similar Mesoproterozoic lower concordia intercept model ages (i.e., removing the titanite component through extrapolating towards concordia).

Hf analyses in discrete zircon grains suffer, in terms of precision, from the small crystal size, zonation, metamict nature of most grains, and necessity to use a small spot split stream approach. Nonetheless, the Hf isotopic composition of matrix-hosted, discrete zircon grains also lends support to a crystallization age of c. 1390 Ma for the dyke. Whilst discrete zircon grains are all younger than this age due to variable amounts of radiogenic Pb loss, Hf diffusion in zircon is much more sluggish than Pb, even in metamict grains, due to the similarity of Zr4+ and Hf4+ charge and ionic radii42. With this in mind, if individual zircon Hf data are back-calculated to 1390 Ma along a Pb-loss trend (i.e., 176Lu/177Hf = 0), then the most radiogenic data overlap with depleted mantle (Fig. 6a). Given that the dyke is a basaltic andesite and likely derived from melting of variably enriched subcontinental lithospheric mantle, Hf isotopes and a crystallization age of c. 1390 Ma is consistent with juvenile melt extraction at this time. Moreover, whole rock 143Nd/144Nd1390 Ma with two stage model ages of c. 2.2 Ga are consistent with inheritance of older crustal components into the primary magma.

Comparison of our new whole rock geochemistry to published geochemical data provides another line of evidence to link the dyke sampled by SW-1 to the c. 1390 Ma Biberkine dykes23. Specifically, the enriched LREE pattern is characteristic of these dykes23, and generally more LREE-enriched than the c. 1890 Ma Boonadgin and c. 1210 Ma Marnda Moorn dyke swarms (Fig. 3). SW-1 also shows a negative Eu anomaly and Al3O3/TiO2 below the mantle range ( ~ 20), consistent with low-pressure fractionation of plagioclase. Strongly elevated concentrations of Sr ( ~ 700 ppm) and LREE, coupled with a modest negative Nb anomaly, indicates either assimilation of evolved crust or extraction from an enriched source. Published evolved Nd (and arguably radiogenic Sr) isotopic ratios and a lack of correlation with Mg# from the Biberkine suite have been considered consistent with a subcontinental lithospheric component, enriched by previous subduction or some other form of assimilation of crust, within the primary magma23.

Detrital zircon components in river sediment: The major detrital zircon age components within the river sediment can be readily ascribed to known local crystalline basement in the Yilgarn Craton (Fig. 9). For example, the age mode at 2645 Ma corresponds to the widely distributed late low-Ca granite bloom. The mode at 2699 Ma corresponds to the age of the Gibb Rock granite in the South West Terrane and is also recognised from volcanics within this terrane6. The older age modes in SW-2 at 3243 Ma, 3206 Ma, and 3172 Ma, are all similar to detrital age modes in South West Terrane metasedimentary rocks6,14. Of particular interest, in the context of the age components within the dyke sample (SW-1), is the >3400 Ma component in the river sediment. This component is defined by minor detrital age modes at 3334 Ma and 3493 Ma which correspond to detrital zircon ages in metasedimentary rocks within the wider South West Terrane and Southern Cross Domain of the Youanmi Terrane and also inheritance within the Narryer Terrane14 (Fig. 9). Such old ages are, in contrast, comparatively lacking from the Eastern Goldfields Superterrane. The catchment area of the Swan-Avon river system whilst extensive is nonetheless a substantial distance south of the Narryer Terrane ( > 550 km), and implies sourcing of the ancient detritus via a direct or multi-cycle pathway from rocks within the South West Terrane itself. This southern and relatively localised ancient source is consistent with the observation of northwards fluvial transport prior to Australia breakup with Antarctica and a general paucity of these ages in Perth Basin sedimentary rocks43.

Fig. 9: Zircon age probability density plots.
figure 9

Age of detrital zircons in modern river sediment from the Swan-Avon river system (SW-2) compared to igneous (magmatic and xenocrystic ages) and (meta)sedimentary (detrital ages) samples of the Yilgarn Craton. Yellow bar denotes best estimate of zircon inheritance age in dyke sample SW-1. All ages collated in the GSWA geoview system (http://www.dmp.wa.gov.au/geochron) and the Geoscience Australia geochronology delivery system (http://www.ga.gov.au/geochron-sapub-web/geochronology/shrimp/search.htm).

Paleoarchean ages in the South West Terrane: A zircon inclusion within titanite preserves an apparently much older age than the dyke, implying it is a c. 3440 Ma inherited component (Fig. 4). Zircon saturation is much lower in felsic than in mafic magmas, resulting in late and typically minor zircon crystallization in the latter44. This relationship is consistent with mafic melts ability to dissolve zircon inherited from other rocks45. While there are numerous factors expected to influence zircon survivorship versus its dissolution in silicate liquids, generally a mafic magma would dissolve (inherited) zircon more efficiently than (neoblastic) titanite45,46. Hence, encapsulating zircon within a titanite armour, which is more geochemically stable in the dyke magma, may explain the retention of older zircon as inclusions in titanite, which have been lost from the matrix.

It is fortuitous that such ancient zircon has been sampled by the dyke and preserved, via shielding due to its location within titanite (Fig. 10). This dyke is >150 km further south of the Swan-Avon river system, which also records detrital zircon evidence of an ancient magmatic source component. Measurements of other zircon inclusions plot within a triangle on the inverse concordia plot, defined by the titanite common–radiogenic mixing array and a c. 3440 Ma component. This distribution of data is interpreted as evidence of other Paleoarchean grains within the included cargo (Fig. 4). Measurement of these small zircon inclusions must also have incorporated some variable component of the host titanite, consistent with trace element signatures. Similarly, two other titanite analyses are situated within this triangle that, when coupled with trace element data, indicate zircon inclusions of probable Paleoarchean age (Fig. 4). Hence, despite there only being a single concordant Paleoarchean zircon inclusion analysis, there are at least five other analyses implying this age of inheritance. Moreover, such material cannot be contaminants introduced during processing as the host titanite for the zircon inclusions yields Proterozoic ages consistent with the dominant age of titanite in this rock.

Fig. 10: Schematic block diagram of crust in the South West Terrane of the Yilgarn Craton and adjacent Pinjarra Orogen.
figure 10

The emplacement of Biberkine Dykes is depicted as transporting Narryer-like ancient xenocrystic zircon grains from the deep crust. The + and 0 denote out and into section movement. Inset shows a representative thin section sketch of the amphibolite dyke hosting inherited zircon inclusions within titanite crystals. Ep epidote, qtz quartz, Ti titanite, plag plagioclase, ap apatite.

No ancient ( > 3000 Ma) U–Pb dates from primary igneous rocks have been reported from the South West Terrane, nonetheless evidence for old crust in this region is recognised from detrital and xenocrystic zircon grains, and whole rock Nd and zircon Hf model ages. Very rare detrital zircon grains in relatively proximal Archean metasedimentary rocks to the dyke have yielded isolated U–Pb ages of up to c. 3400 Ma in a sample c. 60 km south west and c. 3300 Ma in a sample 20 km north14. Other rocks in the South West Terrane more distal to SW-1 contain more abundant relicts of an ancient basement, including zircon in metasedimentary rocks from the Toodyay, Moora, and Julimar areas (e.g. GSWA samples 177901, 177904, 248205, 215316 and 19948214), and also from the south-western-most Youanmi Terrane (e.g. GSWA samples 208381 and 19858014). However, based on the current data it is impossible to distinguish between inheritance of sedimentary or magmatic material into the dyke. Furthermore, detrital zircon U–Pb data in beach sediments in SW Australia have yielded additional putative evidence for Paleoarchean crust, with two c. 3500 Ma grains found28. In any case, the c. 3440 Ma inherited component reported herein supports an interpretation of ancient >3000 Ma components within the South West Terrane (Fig. 10). This ancient material may be a direct basement signature or first or multicycle transported detritus.

The occurrence of ancient zircon components in the South West Terrane could support the concept of a Paleoarchean substrate which was more widespread, beyond the Narryer Terrane in the NW of the Yilgarn Craton, which hosts some of the oldest rocks on the planet47. Indeed, a compilation of detrital zircon ages from the South West Terrane and Southern Cross Domain (Youanmi Terrane), has several Paleoarchean age modes6. Specifically, a c. 3440 Ma detrital zircon age mode is present in Archean metasedimentary rocks of the South West Terrane. Furthermore, a c. 3440 Ma age mode is present in South West Terrane and Youanmi Terrane (both Southern Cross and Murchison domains) inherited zircon within magmatic rocks. The presence of ancient zircon in the South West Terrane is further supported by Paleoarchean detrital zircon components within the Swan-Avon river system that in part drains the South West Terrane. The most similar crystallization ages within the Narryer Terrane to the dyke’s c. 3440 Ma zircon inclusion component, identified herein, is the Eurada Gneiss, dated at 3439 ± 3 Ma48. However, c. 3730 to 3300 Ma monzogranitic gneiss together with minor tonalitic and trondhjemitic components are widespread across the Narryer Terrane Meeberrie Gneiss;49, with old detrital components similarly abundant50. Mafic components at 3400 Ma also exist, with some deformed and metamorphosed mafic lenses in the Narryer Terrane interpreted as synplutonic dykes, intruded near contemporaneously with granites (the Dugel Gneiss)51,52. These Eo- to Paleoarchean components indicate Nd-depleted mantle model ages of c. 3700–3600 Ma53.

A common ancient substrate in the western margin of the Yilgarn craton has been proposed based on Hf depleted mantle model ages from detrital zircon grains28. The identification of a Paleoarchean inherited zircon component provides additional evidence for a connection between the South West Terrane and crustal ages with similarity to the Narryer Terrane. The eastern margin of this possible crustal nucleus appears delimited by a train of iron ore and gold deposits e.g. Katanning, Karara54, which also corresponds to a mercury anomaly in the regolith55. This anomaly runs parallel to NW-striking regional shear zones of the Corrigin Tectonic Zone associated with a deep-seated structure10, which may reflect an ancient crustal boundary. Ultimately, we posit that a swathe of Paleoarchean, and perhaps even Eoarchean substrate (based on Nd and Hf isotopic arrays extending to 3800 Ma56, is situated underneath the western part of the Yilgarn Craton, which served as the nucleus for crustal growth of the craton.

Protracted history of the Darling Fault: Matrix-hosted zircon grains, titanite, and apatite all reveal a history of post-crystallization (i.e., post-1390 Ma) isotopic perturbations in sample SW-1. Discrete zircon grains characteristically have much lower common Pb content than apatite and titanite but, when not shielded by titanite, these grains have been reset at c. 1040 Ma, if not more recently (Fig. 4). Such a process is consistent with fluid mobility following metamictization57,58. By way of example, consider a c. 3440 Ma primary magmatic zircon and an overprinting process commencing at 1360 Ma (i.e., a minimum of 2080 Ma of time between crystallization and overprinting), for a grain containing 342 ppm U and 450 ppm Th (i.e., concentration averages of unshielded zircon cores). By the Proterozoic, such a grain would be metamict, with a density of only 4.44 g/cm3 relative to crystalline zircon of 4.7 g/cm359, and thus highly susceptible to fluid-mediated radiogenic Pb mobility and recrystallization. Titanite shows similar late Mesoproterozoic to Neoproterozoic lower concordia intercept model ages, consistent with the matrix-hosted zircon data. Apatite yields much younger ages, with its radiogenic Pb content implying most of it closed to Pb diffusion at c. 210 Ma. The closure temperature for Pb in apatite is likely somewhere in the range of 375 to 600 °C40,60,61, implying the rock passed through these conditions during the Triassic.

The spread in ages from the Paleoarchean to Triassic is a function of the sample’s proximity to the Darling Fault, a > 1000 km transcrustal terrane boundary between the Precambrian Yilgarn Craton and Phanerozoic Perth Basin (Supplementary Fig. A2). Whilst the Darling Fault may have originally developed as an Archean structure62, it was certainly active during the Proterozoic and Phanerozoic, and into the Neogene63. Known basement rocks west of the Darling Fault, the Pinjarra Orogen64,65, comprise metagranitoids as old as c. 2180 Ma66,67 and c. 1150–1110 Ma Mesoproterozoic metasedimentary rocks66,68,69. Importantly, magmatic ages from crystalline basement intersected in the Woodleigh 1 drillhole, which penetrated the Carnarvon Basin north of the Northampton Complex67, yielded ages of c. 1350 Ma. This magmatic age is consistent with a c. 1350 Ma detrital zircon age component within paragneisses of the Northampton Complex66,68. What this Mesoproterozoic (Ectasian) event reflects is challenging to decipher based on limited material of this age in the exposed crust of Western Australia. Nonetheless, given the presence of appropriately aged detrital zircon grains and the evolved lithology of the Woodleigh 1 drill core basement, this event would appear to have involved emplacement of felsic magmatism into the Pinjarra Orogen at broadly the same time as emplacement of the Biberkine dyke suite into the now adjoining Yilgarn Craton. Farther afield, feasible correlations can be made to alkaline rock emplacement and rifting along the Krishna Province of the Eastern Ghats Belt, on the now separated margin in India70.

Proterozoic overprinting in sample SW-1 is consistent with thermal and fluid perturbation, driven by metamorphism in the Pinjarra Orogen. The igneous and sedimentary protoliths of the Pinjarra Orogen were metamorphosed during the Pinjarra Orogeny at c. 1100–1020 Ma, possibly due to collision with an island arc66,68,69,71 and again during the Kuunga Orogeny at c. 750–520 Ma72. This overprinting is well-represented in the post-1300 Ma titanite ages, but apparently, even more so in the discrete zircon grain fraction. These metamict zircon grains appear especially susceptible to resetting between 1000 and 580 Ma, broadly consistent with c. 630–620 Ma biotite resetting ages in the SW Yilgarn Craton73. Newly grown zircon rims of this age have also been reported (GSWA sample 198551).

The final event recorded in SW-1 is through a Triassic apatite U–Pb age. This apatite date corresponds with previously reported apatite fission track data from the southwest Yilgarn Craton74. Apatite fission track data75 and vitrinite reflectance data have been interpreted to reflect the denudation of the sedimentary cover across the Yilgarn Craton76, as the western margin of the craton rifted. The existence of such sedimentary cover is independently supported by (i) preservation of Palaeozoic to Mesozoic sedimentary rocks in isolated downfaulted sub-basins just south of the study area76 (Supplementary Fig. A2), and (ii) detrital zircon studies suggesting sedimentary reworking of a cover sequence into younger basins and modern coastal settings77,78. Sedimentological evidence supports the development of an emergent rift shoulder along the Darling Fault in the Late Permian to early Triassic79.

Both titanite and apatite ordinate intercepts in Terra-Wasserburg space imply a similar initial Pb composition. Interestingly, this initial Pb composition matches that expected for the Yilgarn Craton at c. 1100 Ma80, supporting the interpretation that a component of these minerals recrystallized during the Pinjarra Orogeny, consistent with the probability peak in titanite U–Pb ages. Apatite must have subsequently been reset via loss of radiogenic-Pb by thermally activated volume diffusion, but not recrystallization, at c. 210 Ma, to generate the observed isochron. Such resetting cannot have caused the complete loss of structurally bound common Pb in the apatite or else the initial 207Pb/206Pb ratio would have decreased to <0.9, as would be predicted for Triassic crustal fluids.

Hence, the c. 210 Ma date defined by the apatite can be interpreted as a function of cooling driven by uplift and denudation of a Yilgarn Craton sedimentary cover, with the date correlating with an interval of peak denudation rate interpreted from apatite fission track studies at ~250–200 Ma75. Alternatively, the apatite U–Pb age may reflect re-heating, related to an Early Triassic phase of rift related magmatism at c. 210 Ma81.

Conclusions

Given age, geochemical, and isotopic relationships, the most parsimonious interpretation is that the protolith to the sampled amphibolite was a 1390–1360 Ma intrusive, related to the Biberkine dyke swarm. Either in its source or emplacement pathway, this rock inherited a cargo of originally magmatic zircon crystals, including a Paleoarchean component. This dyke was hydrated and recrystallized during various stages of Proterozoic tectonothermal activity, as recorded in the Pinjarra Orogen. Later apatite closure to Pb diffusion tracks crustal denudation or heating due to rift magmatism in the Triassic. The proximity to the translithospheric Darling Fault may have permitted the dyke to sample an ancient tectonic sliver, residing at depth in this region. When considered along with the >3000 Ma detrital zircon within (meta)sedimentary rocks of the South West Terrane, modern detrital zircon in the major dewatering system of the region, and the Nd and Hf isotopic evidence of an ancient crustal component, we posit that a collaged section of ancient crust and/or its denuded remains resides at depth along the western margin of the Yilgarn Craton. This material appears to have at least some age similarity to the Narryer Terrane and specifically the Eurada Gneiss. Mineral inheritance within kimberlites has often been used to understand the sub-continental lithosphere. River detrital samples also support such investigations by providing information on potentially hidden crust over a wide area. However, the sampling of the mineral inclusion cargo of crystals within dykes offers another widely available sampling medium, which also, perhaps, helps probe hidden deep geology that may otherwise remain enigmatic.

Methods

Based on limited outcrop, the amphibolite (SW-1) appears as a series of discontinuous NNE-trending dykes of up to 10 metres wide intruding into granitic gneiss (Supplementary Fig. A3). The rock is strongly foliated with amphibole defining a lineation and thin leucocratic veins cutting the foliation. The amphibolite comprises 55% Ca-amphibole (hornblende), 16% plagioclase, 16% quartz, 6% epidote, 5% titanite, 1% apatite, and accessory zircon and opaque oxide minerals (Fig. 2).

A polished thin section and apatite, zircon, and titanite mounts from sample SW-1 were prepared. Mineral separates were generated using pulsed electric discharge (SelFrag) followed by conventional magnetic and heavy liquid processing with both hand picking and bulk mounting. All material was scanned using a Tescan Integrated Mineral Analyzer (TIMA) and imaged using transmitted and reflected light, backscattered electron, secondary electron, and cathodoluminescence (CL) response. This material (both mounts and thin section) was analysed over several laser ablation sessions using an array of different mass spectrometry configurations including: a split stream Agilent 8900 (U–Pb) and Nu Plasma II (Lu–Hf) zircon session, a single stream Agilent 8900 (U–Pb) titanite and apatite session, and a split stream Agilent 8900 (U–Pb) and Agilent 7900 (trace element) titanite and zircon inclusion session. The split stream U–Pb and trace elements session was designed to avoid the need for the counting system to swap modes (e.g. pulse to analogue counting) across variable count rates on the isotopes of interest for both U–Pb and trace elements. All uncertainties in the text are reported at the two standard error level (2σ). Analyses of discrete mineral grains (titanite, zircon, apatite) along with inclusions (e.g. zircon) within mineral grains (e.g. titanite) was made, with the realisation that the latter may consist of mixtures that would require compositionally based subsampling of components.

Whole rock geochemistry was carried out at Australian Laboratory Services (ALS) Global Pty Ltd laboratories, Perth, using standard X-ray fluorescence spectrometry techniques for major elements and inductively coupled plasma mass spectroscopy for trace elements.

To complement the geochronology of SW-1, a sample of river sediment was recovered from the Swan-Avon river system, denoted SW-2. Several kilograms of sand were collected from a shore bar of the Avon River near Cobbler Pool. Heavy mineral grains were concentrated using heavy liquids and magnetic separation, with the resultant entire heavy mineral fraction mounted in epoxy resin, polished to half grain thickness, and imaged using the TIMA, to highlight zircon crystals of interest for further analysis. Identified zircon grains were analysed for U–Pb, using a rapid laser ablation technique82. Ages of individual detrital zircon grains are reported using the calculated spot concordia age83. Given the potential wide temporal range of detritus in SW-2, this approach avoids the need for an arbitrary swap in age defining isotope ratio and provides optimum use of both 238U/206Pb and 207Pb/206Pb ratios.

Detailed analytical methods are provided in the Supplementary Methods and a compilation of data tables is provided within the Supplementary Data.