The age and origin of the South Mountain Batholith (Nova Scotia, Canada) as constrained by zircon U–Pb geochronology, geochemistry, and O–Hf isotopes

Understanding processes responsible for the generation and construction of granitoid batholiths are fundamental to deciphering the evolution of Earth's continental crust. Based on field, geochronological, geophysical, and geochemical studies (e.g., Brown 1994; Petford et al. 2000; Vigneresse 2004), it has been inferred that composite granitoid batholiths are end products of crustal-scale magmatic plumbing systems with their assembly due to incremental emplacement of relatively small magma batches (e.g., Petford et al. 1993; Paterson and Vernon 1995; Cruden 1998; Brown and McClelland 2000; Schaltegger et al. 2009, 2019; Memeti et al. 2010; Smith et al. 2019). To further refine and understand the temporal processes related to the generation and emplacement of batholiths, U–Pb zircon geochronology integrates intra-grain textural observations from cathodoluminescence (CL) and backscattered electron (BSE) imagery to constrain the timescales involved in zircon growth, which can vary widely within an intrusion from ca. ten thousand to several million years (e.g., Brown and Fletcher 1999; Coleman et al. 2004; Schaltegger et al. 2009; Memeti et al. 2010; Schoene et al. 2012; Broderick et al. 2015; Samperton et al. 2015; Miles and Woodcock 2018). Importantly, detailed CL studies of zircon (e.g., Corfu et al. 2003; Harley et al. 2007; Erdmann et al. 2013) reveal complex textural domains now attributed to the dynamic and temporal evolution in former magma plumbing systems (e.g., Barboni et al. 2015; Miles and Woodcock 2018). These domains may include (cf. Miller et al. 2007): (1) xenocrysts from contamination at the source, along its path, or current wall rock; (2) antecrysts related to early magma gestation in its source prior to its migration and batholith assembly; (3) autocrysts that record final melt crystallization; and (4) partial replacement or overgrowth due to post-magma emplacement processes such as metamorphism (e.g., Hoskin and Black 2000), and (or) hydrothermal activity (e.g., Schaltegger 2007).

To address the age and textural complexities of zircons often requires both high spatial-precision and age-accurate data. Whereas chemical abrasion isotope dilution thermal ionization mass spectrometry (CA–ID–TIMS) is the most precise dating method (e.g., Mattinson 2005; Crowley et al. 2007), such work understandably avoids zircon with complex domains or inherited components since this compromises generating high-quality ages (i.e., not reflective of crystallization; cf. Mattinson 2011). In contrast, the sensitive high-resolution ion microprobe (SHRIMP) provides spatially-resolved analysis of zircon domains in an essentially nondestructive manner, but with less precision than CA–ID–TIMS (i.e., <1–2% versus <0.1%). The SHRIMP therefore has potential to reveal prolonged or episodic magmatic histories preserved in complexly-zoned zircon grains. Complementary to the latter, is the recent addition of both trace-element chemistry and isotopic (i.e., δ¹⁸O, Lu–Hf) fingerprinting to the dated domains which provides important insight to the petrogenesis of intrusive complexes (e.g., Gardiner et al. 2017; Sun et al. 2019; Schaltegger et al. 2019; Finch et al. 2021).

The purpose of this study is to apply a modern integrated analytical protocol to address the time and duration of melt generation, time span of assembly, and petrogenetic parameters for the South Mountain Batholith (SMB) of southern Nova Scotia, Canada. Given it is the largest exposed plutonic body (7300 km²) in the Appalachian orogen, such information has profound implications in regard to the geotectonic setting during emplacement. Previous U–Pb (zircon, monazite), whole-rock Rb–Sr and K–Ar, and ⁴⁰Ar/³⁹Ar dating constrain SMB crystallization from ca. 385 to ca. 370 Ma (Table 1), thus implying a ∼15 million year emplacement window; a duration which is highly anomalous and a multiple of the ∼5 million years typical for assembly of intrusive complexes (e.g., Petford et al. 2000; Coleman et al. 2004). To reassess the timeframe for SMB assembly, we present new in situ SHRIMP U–Pb dates for different zircon domains (i.e., xenocrystic, antecrystic, and autocrystic) in well-characterized samples from all its intrusive units in addition to the first high-resolution CA–ID–TIMS data for zircon autocrysts from the SMB; the latter provides the best constraints on crystallization ages (e.g., Parrish and Noble 2003; Mattinson 2005). In addition, in situ analysis using laser-ablation multicollector inductively coupled plasma mass spectrometry (LA–MC–ICP–MS) and secondary isotope mass spectrometry (SIMS) was done to couple age data with the geochemical signature (i.e., trace- and rare-earth (REE) elements, Lu–Hf and δ¹⁸O values) for each domain. Collectively, these data allow us to address several important aspects of the SMB: (1) its temporal reconstruction; (2) the petrogenesis for the many plutonic centres; and (3) its geotectonic setting during batholith assembly.

The SMB is part of the Meguma terrane of southern Nova Scotia (Fig. 1) which forms the most outboard of several Gondwana-derived lithotectonic terranes in eastern Canada (e.g., Ganderia, Avalonia, Meguma) that accreted to the eastern Laurentian margin during Paleozoic assembly of the Appalachian orogen (i.e., Fig. 1A; Bird and Dewey 1970; Williams and Hatcher 1983; Keppie 1985; van Staal 2007; van Staal et al. 2009; Waldron et al. 2015). The Meguma terrane records this accretion as a series of deformation events during the late Early Devonian (i.e., Acadian; 410–395 Ma; e.g., Muecke et al. 1988; Kontak et al. 1998) to the Early Carboniferous (375–350 Ma; e.g., Keppie and Dallmeyer 1995; Archibald et al. 2018) time interval. The latter tectonothermal events define the Neoacadian orogeny (van Staal 2007; van Staal et al. 2009, 2021) which involved dextral accretion of the Meguma terrane along the composite Laurentian margin (e.g., Keppie et al. 1991; Murphy et al. 2011; Fig. 2). Prior to accretion of the Meguma terrane, Avalonia accreted to Laurentia during the earlier Acadian orogeny (∼421–417 Ma) and is thought to have subsequently been structurally juxtaposed beneath the outboard Meguma terrane by ca. 380 Ma (Eberz et al. 1991; Shellnutt et al. 2019; Fig. 2). Importantly, the outline of the SMB and its internal features (e.g., faults/shear zones, jointing) combined with gravity, seismic, and aeromagnetic studies support a syn-tectonic model for its emplacement (e.g., Keen et al. 1991; Horne et al. 1992; Benn et al. 1997; MacDonald 2001).

The Meguma terrane is predominantly underlain by Ediacaran – Early Ordovician metasedimentary rocks of the Goldenville and Halifax groups and abundant peraluminous granitoid intrusions that include the SMB and smaller satellite plutons (Clarke et al. 1997; White 2010). The Goldenville and Halifax groups are unconformably overlain by metasedimentary and metavolcanic rocks of the Silurian–Devonian Rockville Notch Group; including ca. 440 Ma low-grade metavolcanic rocks of the White Rock Formation (White 2010). Lithostratigraphic and detrital zircon data from the metasedimentary rocks suggest a predominant provenance from the West African craton along the northern margin of Gondwana (Clarke and Halliday 1985; Waldron et al. 2009; White and Barr 2010; White et al. 2018).

Previous dating (Table 1) broadly constrains the emplacement of the SMB and spatially affiliated intrusions to ∼385–370 Ma; younger and smaller satellite intermediate- to felsic plutons in the southwestern Meguma terrane were emplaced at ca. 360 Ma (e.g., Wedgeport and Clayton Hill granites). The nature of plutons of similar age to the SMB in the southwestern Meguma terrane (e.g., Shelburne, Port Mouton, Barrington Passage, and Wedgeport plutons; Fig. 1A) is important to understanding SMB petrogenesis as some are interpreted to have originated from hybridization (i.e., magma-mixing) of mafic-intermediate and felsic magmas (e.g., Clarke et al. 1988; Tate and Clarke 1997; Tate et al. 1997). The much larger SMB is, however, ∼5 to 10 Ma older than both the relatively smaller tonalitic to dioritic intrusions (Table 1; e.g., Bog Island Lake, BG in Fig. 1A; Giles and Chatterjee 1987) and the rare lamprophyres (e.g., the Pope's Harbour dykes, PH in Fig. 1A; Owen et al. 1988; Kempster et al. 1989) along the SMB periphery.

The structural basement to the Meguma terrane is not exposed, but is inferred from metaigneous and metasedimentary granulite-facies xenoliths in the Late Devonian Pope's Harbour dyke on the eastern shore of the terrane (Fig. 1A; Greenough et al. 1999). Zircons from such xenoliths (Shellnutt et al. 2019) have a detrital age population distribution ranging from Proterozoic to Late Silurian and younger (ca. 412–388 Ma; 387–364 Ma; and 363–343 Ma). Shellnutt et al. (2019) interpreted these data to reflect the following, respectively: (1) an Avalonian provenance for the metasedimentary granulite xenoliths; (2) a lower crustal expression of Rockville Notch Group-related magmatic events at ca. 440 Ma; and (3) tectono-magmatic events associated with the Neoacadian orogeny. These data thus collectively constrain the composition and age of tectono-magmatic events that affected the lower crust underlying the Meguma terrane.

The younger ca. 360 Ma granitic magmatism (reviewed by Kontak et al. 2013) was synchronous with deposition of basin-fill molasse deposits of the late Famennian-Early Mississippian Horton Group (Martel et al. 1993). These clastic rocks are the oldest stratigraphic unit to unconformably overstep the Avalon-Meguma terrane boundary and they also onlap the SMB. The Horton Group clast inventory indicates provenance from Meguma metasedimentary rocks, the Rockville Notch Group, and the SMB (Murphy 2000). These data therefore indicate that at least part of the SMB was exhumed by ca. 355 Ma (e.g., Archibald et al. 2018).

The SMB is a composite intrusive complex dominated by granodiorite to monzogranite with lesser more evolved leucogranite units (MacDonald et al. 1992; MacDonald 2001). Overall, the SMB has a peraluminous bulk composition (i.e., has normative corundum) and is characterized by relatively high P₂O₅ (e.g., McKenzie and Clarke 1975; Clarke et al. 2005). Earlier interpretations suggest it originated via partial melting of unexposed basement rocks (Clarke et al. 1988), but subsequent models have favoured assimilation-fractional crystallization (Clarke et al. 2004) with the former involving Meguma metasedimentary rocks (e.g., MacDonald et al. 1992; Clarke et al. 1997; Clarke and Carruzzo 2007; Shellnutt and Dostal 2012), a conclusion consistent with field evidence (e.g., MacDonald 2001; MacDonald and Clarke 2017).

The SMB is characterized by two distinct magmatic stages (MacDonald 2001): Stage I granodiorite that form a border phase; and Stage II evolved monzogranite plutons (n = 5) that include, from west to east, the Davis Lake (DLP), West Dalhousie (WDP), East Dalhousie (EDP), New Ross (NRP), and Halifax (HP) plutons (Fig. 1B). A summary of the features of these phases, after MacDonald (2001), is presented in Supplementary Material S1.

Each of the main phases of the SMB were sampled for U–Pb dating (i.e., zircon and monazite) to complement previous age constraints for both it and the peripheral plutons (see Table 1). Locations are shown in Fig. 1B with precise coordinates in Table 2, whereas descriptions of all samples collected are presented in Supplementary Material S1. Zircon mineral separation undertaken used ∼10 kg of rock collected from surface outcrops. These samples were prepared for either U–Pb dating (SHRIMP, CA–ID–TIMS), δ¹⁸O analysis (SIMS) or combined trace-element, U–Pb dating and Lu–Hf isotopic analyses (LA–ICP–MS). Full details on sample preparation protocols specific to each analytical procedure and the relevant laboratories are given in Supplementary Material S2.

U–Pb crystallization ages of zircon and monazite were calculated from a total of 18 samples employing three different U–Pb dating methods: (1) in situ U–Pb isotopic analysis of zircon using the SHRIMP at the J.C. Roddick Ion Microprobe Laboratory at the Geological Survey of Canada, Ottawa; (2) CA-ID-TIMS analysis of single zircon and monazite grains by abrasion and dissolution in Krogh-type TEFLON bombs prior to analysis by a multi-collector MAT 262 mass spectrometer at the radiogenic isotope facility housed at Memorial University of Newfoundland (Canada); and (3) simultaneous in situ U–Pb isotopic and trace-element analysis of zircon using a Photon Machines Analyte G2 193-nm excimer laser ablation system coupled to a Thermo Neptune Plus with Jet Interface MC–ICP–MS (U–Pb isotopes) and Thermo iCan-TQ Q–ICP–MS (trace-elements) at the Mineral Exploration Research Centre Isotope Geochemistry Lab (MERC-IGL), Laurentian University, Sudbury (Canada). The latter instrument was also used for Lu–Hf isotopic analysis in a subsequent session targeting the pre-existing U–Pb/trace-element ablation pits. Reduction of the U–Pb data from the MC–ICP–MS used the VizualAge package (Petrus and Kamber 2012) whereas processing of trace-element data from the Q–ICP–MS used the trace-element internal standard DRS (Si = 15.28 wt.% in zircon) in Iolite v. 3.32 (Paton et al. 2011). Oxygen isotopic analysis of the same zircons used a Cameca IMS 1280 multicollector ion microprobe at the Canadian Centre for Isotopic Microanalysis (CCIM), University of Alberta, Edmonton (Canada). Full details on experimental conditions for each analytical method are provided in Supplementary Material S3.

Zircons analysed in this study are from representative samples that span the entire lithological range of the SMB. Less-evolved phases of the SMB likely are more abundant in accessory phases such as zircon than those in evolved phases due to the spatial association of these minerals with biotite (Clarke et al. 2021). In general, the zircons are from 25 to 500 µm length with external morphologies that include irregular stubby and partially resorbed, regular elongate prismatic, and the most dominant equant prismatic. Furthermore, collectively these zircons have up to four types of internal domains as distinguished using CL and BSE imagery and verified with geochemical and U–Pb isotopic analyses. Zircon domain types (see below) are depicted in Fig. 3 and include: xenocrysts (inherited), antecrysts (inherited), autocrysts, and post-emplacement growth or recrystallized.

Where zircon domains are narrow, mineral dissolution (i.e., CA–ID–TIMS) or in situ analysis (i.e., LA–ICP–MS or SHRIMP) could inadvertently sample multiple domains resulting in a mixed age with ambiguous geological significance (cf. Mundil et al. 2001; Miller et al. 2007). To avoid this possibility, zircon for CA–ID–TIMS analysis targeted the clearest (i.e., least damaged) and most homogeneous crystals, whereas homogeneous portions of grains from both interior (cores or mantling cores) and exterior (rims) domains were used for in situ analysis.

Xenocrystic cores are typically rounded or irregular in shape, have convoluted patterns or lack zoning (e.g., Figs. 3E and 3F), and are mantled by well-zoned magmatic zircon (i.e., either antecrystic or autocrystic). Antecrystic domains are characterized by a combination of oscillatory- to sector-zoning with irregular to partially resorbed margins that are typically rimmed by autocrystic overgrowths. Antecrystic domains include those crystallized in earlier intrusive phases that were either assimilated or transported from depth by later magmas of the same magmatic system (Schoene et al. 2012).

Autocrystic domains are recognized herein by continuous oscillatory zoning and bright CL response relative to other zircon domains; they typically mantle the earlier domains (i.e., xenocrystic and antecrystic; e.g., Figs. 3D–3F). Late overgrowth and (or) replacement domains, where present, truncate earlier zones (Fig. 3F). These domains typically have chaotic- to weakly convoluted zoning as is characteristic for metamorphic or metasomatic origins (Corfu et al. 2003). The widespread occurrence of these late overgrowth domains suggests zircon dissolution, resorption, and (or) recrystallization across the SMB, as is typical of zircon in intermediate peraluminous granites (cf. Watson and Harrison 1983).

The CL emission in zircon is negatively correlated with its metamictization and elemental content of U, Hf, and Y (e.g., Poller et al. 2001; Rubatto and Gebauer 2000), whereas emission activators in zircon include Dy and Tb (Hanchar and Rudnick 1995). The latter authors also showed that variable Hf is primarily responsible for variability in BSE intensity. Thus, assuming saturation of ZrSiO₄ occurs as the granitic magma cools and crystallizes zircon, the fine-growth banding and commensurate strong CL and BSE contrasts (e.g., Figs. 3D–3F) are likely due to variations in U and other trace-element concentrations at the crystal-melt interface (e.g., Rubatto and Gebauer 2000).

The full data for in situ U–Pb zircon dating on samples collected from each major phase of the SMB are provided in Supplementary Tables S1 (SHRIMP zircon analyses) and S2 (LA–ICP–MS zircon analyses). The in situ zircon U–Pb results presented are categorized into domain types, which were assigned prior to analysis based on the criteria described above. Results are presented in sequential order: xenocryst, antecryst, autocryst, and overgrowth/recrystallized. Spots analysed using LA–MC–ICP–MS yielded dates (not corrected for common-Pb; see Supplementary Material S3) that are consistent with, though less concordant than, the ²⁰⁴Pb-corrected ²⁰⁶Pb/²³⁸U dates from equivalent SHRIMP spots (see below). Thus, our interpretation of LA–MC–ICP–MS data relies on the trace-element and Lu–Hf results for these zircons. Additionally, where data with error ellipses overlap Concordia, dates are based on the weighted mean average ²⁰⁶Pb/²³⁸U ages. These data are presented below with the corresponding 2σ error.

Xenocrystic zircon domains collectively yield age populations with three distinct sub-groups based on peaks in probability density space (Fig. 4): (1) ∼400 to 470 Ma; (2) ∼480 to 670 Ma; and (3) >700 Ma. Low-density clusters also occur at ∼700 to 850 Ma, ca. 1.5 Ga, and ∼1.9 to 2.2 Ga. Whereas the ²⁰⁶Pb/²³⁸U chronometer is applied to data that yields dates <850 Ma, the ²⁰⁷Pb/²⁰⁶Pb chronometer is applied for dates >850 Ma for a more robust measure of concordance in these age populations. Overall the xenocryst ages are similar to the detrital zircon populations for the upper Goldenville and Halifax groups and overlying Silurian–Devonian rocks (White Rock and Torbrook formations; summary in Waldron et al. 2009). The ∼540 to 770 Ma zircon population is also shared by clastic sedimentary units of the adjacent Avalonia terrane (cf. Barr et al. 2012; Henderson et al. 2016). A small anomalous xenocryst population also occurs at ∼1.5 Ga (6 analyses across 4 samples; Fig. 4).

Antecrystic domains record dates ∼3 to 15 million years older than their respective overgrown autocrystic domains (Table 3; Figs. 5A and 5B and 6A–6J). Furthermore, these domains collectively yield ²⁰⁴Pb-corrected weighted mean ²⁰⁶Pb/²³⁸U (SHRIMP) dates from 391.0 ± 3.1 to 385.9 ± 3.8 Ma in the less evolved phases (16BIC-080 and 16BIC-075, respectively; Figs. 6A and 6C), and from 388.3 ± 3.9 to 377.8 ± 3.6 Ma in the more evolved phases (16BIC-079 and 16BIC-082, respectively; e.g., Figs. 6I and 6J).

Autocrystic domains yield ²⁰⁴Pb-corrected ²⁰⁶Pb/²³⁸U dates ranging from ∼381 to 369 Ma. These dates (Table 3) may be sub-divided spatially and temporally into two groups: 1) less evolved granodiorite and the WDP monzogranite which range from 381.1 ± 3.8 to 375.9 ± 4.1 Ma (16BIC-075 and 16BIC-074, respectively); and 2) the more evolved EDP, NRP, HP, and DLP that range from 376.8 ± 3.8 to 370.2 ± 3.7 Ma (16BIC-084 and 16BIC-079, respectively) with an overlapping younger date at 368.5 ± 3.7 Ma (16BIC-082; Table 3, Fig. 5A).

An exception is noted in regard to the above subdivision for a Stage II WDP monzogranite sample (16BIC-083) which has lithological characteristics and a SHRIMP date similar to the less evolved Stage I granodiorite (378.1 ± 3.8 Ma, Table 3; Figs. 5A and,6D). This date is complemented by a CA–ID–TIMS sample from the mapped WDP (below, Fig. 5A); thus, for the purpose of this study, the host pluton is categorized as part of the early-stage suite in the SMB (i.e., Stage I; purple fields in Fig. 5).

A number of zircon grains analysed in situ for U–Pb have unzoned overgrowths or replacement domains. Analyses on the overgrowth/replacement domains yield concordant dates which are 15–10 Ma younger than the autocrysts dated in their respective samples (i.e., analyses with blue error bars/ellipses in Figs. 5 and 6).

Monazite was not measured by in situ methods, but the most pristine grains that yielded similar CA–ID–TIMS dates to those of representative zircon are included in age interpretations for this study. The full data set of CA-ID-TIMS zircon and monazite analyses is presented in Table S3.

Monzogranite sample 19BIC-087 was collected from the Panuke Lake intrusion within the NRP (MacDonald et al. 1992). This sample yielded a date of 377.5 ± 1.1 Ma (Fig. 7A) — i.e., ∼5 Ma older than other NRP samples dated in this study (Fig. 5A; Table 4). Therefore, we suggest sample 19BIC-087 represents a monzogranitic phase from the early-stage crystallization of the SMB rather than the NRP. Given this proposed reclassification, the ²⁰⁶Pb/²³⁸U dates for Stage I SMB samples (including the WDP) are between 378.8 ± 1.1 and 375.2 ± 1.1 Ma (Figs. 7A–7D), whereas Stage II plutons are between 375.4 ± 0.8 and 371.8 ± 0.8 Ma (Figs. 7E–7H).

Notably, zircon and monazite from the DLP leucogranite sample (16BIC-071) yielded an upper discordia intercept date of 374 + 17/–12 Ma (Fig. 7H). To resolve the large associated error, we interpret the U–Pb analysis by LA–MC–ICP–MS on the sample split, which yielded a weighted mean ²⁰⁶Pb/²³⁸U date of 373.3 ± 2.2 Ma (n = 4, MSWD = 0.39; Table 3), as the best estimate of its crystallization age (Fig. 5B).

Trace-element and REE concentrations for representative zircons from each of the main SMB plutons and domain types are displayed in Figs. 8–10 and the complete data set is available in Supplementary Table S2. The data set was filtered to exclude zircon domains with anomalously high common-Pb (²⁰⁴Pb) by omitting analyses with ²⁰⁴Pb greater than 4 ppm, final discordance % greater than 60%, and final fraction common-Pb greater than 0.025 (Supplementary Table S2; probability plots in Supplementary Fig. S2f). The crystallization temperature for each zircon spot analysis (Fig. 11) was determined using the Ti-in-zircon geothermometer (Ferry and Watson 2007) at fixed activities of SiO₂ and TiO₂ assumed to be 1 and 0.5, respectively. The SiO₂ activity is based on Si saturation, while the latter is likely a minimum, given the presence of magmatic rutile has been inferred, albeit poorly constrained by a decline of Nb/Ta in whole rock data (Carruzzo et al. 2006). Importantly, evaluation of this parameter by Schiller and Finger (2019) using MELTS modelling suggests this is the most reasonable value for peraluminous granites, hence suitable for the SMB.

Magmatic oxygen fugacity (ƒO₂) was estimated using an internally-consistent model for zircon-melt partitioning of Ce and a redox equilibrium calibrated for silicate melts (Smythe and Brenan 2015, 2016). This model incorporates Ti-in-zircon temperatures, Ce-in-zircon anomalies (i.e., Ce/Ce*), the REE and high field strength element (HFSE) concentrations of both the measured zircon and representative whole rock, and a presumed concentration of dissolved H₂O (estimated at ∼2 wt.%). The latter is based on LOI values (∼0.5–1 wt.%) from the SMB geochemical database of MacDonald (2001) that, when pressure-corrected to emplacement depth (i.e., ∼10–12 km, Kontak and Kyser 2011; Jamieson et al. 2012), equates to a minimum of ∼2 wt.% H₂O (cf. Holtz et al. 2001). The ƒO₂ values of zircon xenocrysts were not determined due to unknown parameters (e.g., any assumption of equilibration with a peraluminous melt) for the inferred detrital source of the zircons, as well as the metamict character for many of the xenocrystic cores which influences zircon chemistry. Thus, we do not estimate ƒO₂ for these data.

The trace-element and REE data for xenocrysts (n = 89) are summarized for the three dominant age groups in plots of Hf content and Th/U versus age and chondrite-normalized REE (REE_N) patterns (Fig. 8). There is no discernable trend of Hf with decreasing age, but there is a notably very low Th/U signature (i.e., <0.4) for the youngest group (∼470–400 Ma; Fig. 8A) which reflects their high U contents. In contrast, the 700–470 Ma group has relatively low- to moderate U and a wide range of Th/U ratios (0.02–2.40; Fig. 8B). The >700 Ma group have low U and variable Th/U ratios (0.07–1.01), similar to the 700–470 Ma group (Fig. 8C). All xenocrysts exhibit characteristic REE_N trends for zircon (i.e., HREE > LREE), have positive Ce anomalies, and negative Eu anomalies (Fig. 8) which are typically less pronounced than for the antecrystic and autocrystic domains.

The antecrysts (n = 34) yield a wide range for U (161 to 2204 ppm) with lower contents typical for Stage I samples (i.e., 200–867 ppm; Supplementary Table S2). Furthermore, these domains have low-to moderate Th contents (20–183 ppm) throughout the SMB except for the DLP which has higher values (at 81–403 ppm; Supplementary Table S2). Most of the antecrysts have similar Hf contents and Th/U ratios as the autocrysts (Fig. 9) and display no discernable increase or decrease with age and calculated Ti-in-zircon temperatures (Fig. 10).

Antecrysts display typical REE_N patterns with HREE > LREE (i.e., >3 or 4 orders of magnitude; Fig. 9) and are similar throughout all lithologies, but a slight increase in LREE is noted for these domains from some Stage II plutons (i.e., EDP and DLP; Figs. 9D and 9E) along with enrichment of HREE relative to middle REE (MREE; i.e., [Yb/Gd]_N) which generally decreases with higher Th/U (Fig. 10A). The Ce-in-zircon anomalies (Ce/Ce*) show no discernable trend to their respective (Yb/Gd)_N values (Fig. 10B; Supplementary Table S2). The Eu anomalies (Eu/Eu*) for antecrysts broadly overlap (Fig. 10C) and are generally indistinguishable from their respective autocrystic domains.

The derived crystallization temperatures are lower for antecrysts compared to the younger autocrystic domains (742 to 981 °C in the DLP and 671 to 829 °C in all other plutons; Fig. 10D). Additionally, these domains have ƒO₂ values that are low, relative to FMQ (Fig. 11A), with ΔFMQ ranging from −1.3 to −7.2.

The autocrysts from Stage I granodiorite and the WDP samples (n = 26) have values and ranges in U (70–863 ppm) and Th (17–217 ppm) similar to their respective antecrysts (Supplementary Table S2), with Th/U ratios that are notably lower in the younger plutons (Fig. 9). The ΣREE and ΣLREE contents, in addition to REE_N profiles, are similar to those in their respective antecrystic domains (Fig. 9). The low Eu/Eu* values overlap those of the autocrysts in Stage II samples (Fig. 10A), whereas the Ce/Ce* values are variably positive (2.1 to 53.6; Fig. 10B). The Ti-in-zircon values for early units have a wide range of crystallization temperatures (669 to 874 °C) and overlap data for the autocrysts across the SMB (Fig. 10D). Autocrysts in Stage I samples have ƒO₂ values similar to their respective antecrysts (ΔFMQ from −1.6 to −6.8; Fig. 11B), and are consistent with ƒO₂ values for Stage II samples (see below).

Autocrysts from samples in more evolved plutons (i.e., the EDP, NRP, HP, and DLP; n = 50) have a wide range in U contents (119 to 2617 ppm) that is generally similar to those in their respective antecrystic domains (Supplementary Table S2). The majority of these domains have moderate- to low Th values (18–268 ppm), whereas autocryst domains in the DLP have a wider range and higher values (66–442 ppm; Supplementary Table S2). The Th contents for DLP autocrysts are elevated compared to the rest of the SMB, but are similar to values in their respective antecrysts.

The ΣREE in these autocrysts have ranges like those in their respective antecrystic domains, as well as autocrysts for Stage I samples (Fig. 9). Typical REE_N profiles are noted for these domains with weak to moderately positive Ce/Ce* values (1.5–50; Fig. 10B) and weakly positive Eu/Eu* values (<0.01–0.11; Fig. 10C) that are generally higher than their respective antecrystic domains but similar to the autocrysts in the less evolved parts of the SMB.

The Ti-in-zircon data for autocrysts from the Stage II plutons indicate a wide range of crystallization temperatures: 663–894 ° C in the EDP, NRP, and HP with higher values in the DLP (742–981 ° C) (Fig. 11B). These temperatures are higher on average, but generally overlap with the antecrysts and autocrysts from Stage I samples (Fig. 11A). The calculated ƒO₂ values for these domains are similar to their respective antecrystic domains, with ΔFMQ from −0.1 to −6.4 (Fig. 11B). In the DLP, however, these domains have a narrower range (ΔFMQ from −0.8 to −4.5) and are thus less reduced than the other data.

The youngest zircon domains from the SMB (n = 45) have widely variable U (83 to 1693 ppm), Hf (8430 to 13 050 ppm), and Th (17 to 304 ppm) contents (Supplementary Table S2). These ranges, as well as their ΣREE and ΣLREE, generally overlap with that for autocrysts and antecrysts for each respective zircon (Fig. 9). The Ce/Ce* values are also highly variable, but generally more positive than their respective autocrystic and antecrystic domains (i.e., 1.4 to 49.4; two outliers at 78.0 and 79.8; Supplementary Table S2).

The Ti-in-zircon thermometer and Ce-in-zircon oxygen barometer were not applied to these data as calculations assume the measured zircons have equilibrated with a magmatic reservoir or metamorphic in origin (Ferry and Watson 2007; Smythe and Brenan 2016). These domains typically reflect Pb loss or have modified compositions, thus making them unsuitable for the zircon-based thermometer/barometers utilized above (e.g., Siégel et al. 2018).

Spots for in situ δ¹⁸O analysis were located within the same zircon domains used for U–Pb dating and trace-element analysis. These spots were selected based on CL imaging (e.g., Fig. 3). The δ¹⁸O data are shown Fig. 12 and detailed in SupplementaryTable S4.

Xenocrystic zircon cores are the most variable in their δ¹⁸O values (Fig. 12A) and weakly correlate with age: 1) the ∼400 to 470 Ma domains vary from 8.37‰ ± 0.17‰ to 9.79‰ ± 0.18‰ (n = 12); 2) the ∼480 to 670 Ma domains vary from 4.30‰ ± 0.27‰ to 11.38‰ ± 0.21‰ (n = 36); and 3) the >700 Ma domains vary from 3.45‰ ± 0.18‰ to 9.92‰ ± 0.18‰ (n = 11; Supplementary Table S4).

For the nonxenocrystic domains the δ¹⁸O values are: (1) early stage granodiorite antecrysts (ca. 390–385 Ma) range from 7.55‰ ± 0.19‰ to 9.16‰ ± 0.16‰ (n = 15, Fig. 12B); (2) the ca. 385–378 Ma antecrysts are from 7.33‰ ± 0.20‰ to 8.96‰ ± 0.20‰ (n = 17, Fig. 12C); (3) autocrysts in Stage I samples are from 7.81‰ ± 0.20‰ to 9.19 ± 0.21‰ (n = 26, Fig. 12D) and overlap with the higher values in their respective antecryst populations; (4) autocrysts in Stage II samples, excluding those from the DLP, vary from 8.26‰ ± 0.23‰ to 9.13‰ ± 0.23‰ (n = 41; Fig. 12E); (5) autocrysts in DLP samples may define two sub-populations, one between 7.31‰ ± 0.28‰ and 7.95‰ ± 0.26‰ (n = 7), and the other between 8.10‰ ± 0.25‰ and 8.88‰ ± 0.15‰ (n = 12; Fig. 12F); and (6) late overgrowth/recrystallized domains have a wide range from 7.68‰ ± 0.23‰ to 9.16‰ ± 0.24‰ (n = 25; Fig. 12G).

Zircon Hf isotopic data were obtained by ablating directly on top of earlier sites where U–Pb and trace-element analyses were determined. A total of 377 analyses on 321 domains were obtained for all intrusive phases as follows: (1) 92 for xenocrysts; (2) 71 for antecrysts; (3) 105 for autocrysts; and (4) 49 on overgrowth/recrystallized areas. Sixty analyses were excluded based on domain overlap or interference by microinclusions (Supplementary Table S5).

Analysis of xenocrystic zircon populations yields ¹⁷⁶Lu/¹⁷⁷Hf ratios that vary with respect to U–Pb age and a range of ε_Hf(t) values that become more variable with age (Fig. 13A). The antecrystic domains also have a wide range of ¹⁷⁶Lu/¹⁷⁷Hf values with two populations of ε_Hf(t) values based on the host plutons: +1.8 ± 0.5 to +4.6 ± 0.8 for the DLP, and −3.7 ± 0.5 to +2.0 ± 0.6 for the rest of the SMB (Supplementary Table S5). Autocrysts have overlapping ¹⁷⁶Lu/¹⁷⁷Hf and ε_Hf(t) values (n = 28; Fig. 13B) that are consistent with their respective antecrystic domains for each plutonic phase. The overgrowth and recrystallized domains have a wide range for both ¹⁷⁶Lu/¹⁷⁷Hf and ε_Hf(t) values (n = 43; Supplementary Table S5).

The new U–Pb dating of zircon and monazite from the SMB presented herein constrains its absolute age, duration of crystallization and assembly, and provide a general age profile for its inherited zircons. In addition, the trace-element chemistry and complementary Lu–Hf and oxygen isotopic analyses for zircon domain types (i.e., xenocrysts, antecrysts, autocrysts, overgrowths) provide insights into source reservoirs, longevity of magmatism, and temperatures and redox conditions of both magmatism and subsequent modification. These data, when combined with field relationships and petrographic observations, constrain models for the origin, evolution and setting of the SMB.

Trace-element and REE data from temporally-constrained magmatic zircon domains document intra-grain and inter-pluton variability relevant to the evolution of the crustal-scale SMB magmatic plumbing system. Such data can also distinguish zircon formed during progressive magma fractionation in crustal melts from zircon whose composition was influenced by magma mixing or host rock contamination (e.g., Claiborne et al. 2010; Wotzlaw et al. 2013; Marsh and Stockli 2015; Buret et al. 2016). In this regard, we note the zircon geochemistry shows evidence for both magma fractionation, as exemplified by increasing (Yb/Gd)_N with decreasing Th/U (e.g., Claiborne et al. 2010; Buret et al. 2016), and contamination/mixing by different lithologies based on ranges in Th/U, Ce/Ce*, and Eu/Eu* (e.g., Belousova et al. 2002; Fig. 10).

Of the two-age populations identified for xenocrystic domains, (1) 470–400 Ma and (2) >470 Ma, the chemical signature of the former (i.e., low Th/U, negative Eu/Eu*, HREE_N-enriched, wide range of LREE_N values; Fig. 8A) suggests a metamorphic origin (e.g., Rubatto 2002). Of the younger xenocrystic age population, those at ca. 400 Ma overlap in time with metamorphic zircons from granulite xenoliths in the Pope's Harbour dyke (Fig. 1A) and are considered to represent Avalonian lower crust that underthrust Meguma rocks by ca. 380 Ma (Eberz et al. 1991; Shellnutt et al. 2019). The older (>470 Ma) xenocrystic domain population has moderate Th/U values, HREE enrichment relative to chondrite, moderately negative Eu/Eu* values, and strongly positive Ce/Ce* values (Fig. 8B and 8C), and thus have a signature typical of zircons crystallizing from felsic magmas (e.g., Hinton and Upton 1991; Hoskin and Ireland 2000). We interpret range in ages in the older xenocrystic domains to collectively represent a detrital age population.

The lack of well-defined trends in Hf content and Th/U values in magmatic zircon (Fig. 9) suggests these measures may not be effective in monitoring fractionation (cf. Kirkland et al. 2015) in composite magmatic complexes with extensive contamination. In addition, the similar REE and trace element geochemical signatures suggest: (1) melts crystallizing both domain types (autocrystic and antecrystic) originated from a similarly equilibrated chemical reservoir; and (or) (2) progressive crystallization of the magmas did not generate systematic compositional changes in the zircon.

The samples that contain zircon with ca. 360 Ma domains (i.e., 16BIC-071, -072, -076, -080, -083 and -084; Fig. 5) are located either within or along strike of known regional shear zones (Fig. 1B). These structures are locally characterized by a distinct and variably pervasive hematite alteration (e.g., shear zones cutting the EDP; Horne et al. 1992). Zircon recrystallization or overprinting occurs either through hydrothermal growth (i.e., direct crystallization from a zircon-saturated aqueous-based fluid; e.g., Kirkland et al. 2009) or solid-state recrystallization during granulite-grade metamorphism (e.g., Hoskin and Black 2000). The ca. 360 Ma overgrowth/recrystallized zircon domains do not exhibit typical hydrothermal zircon chemical signatures (i.e., enrichment in LREEs or nonradiogenic Pb; e.g., Schaltegger 2007), but reflect partial inheritance of the precursor autocrystic domains with somewhat flatter LREE_N profiles and more pronounced Ce anomalies (Fig. 9). Although not typical, similar Ce anomalies and LREE profiles have been attributed to hydrothermal transitional settings for Sn-hosting granites (e.g., Mole Granite, Australia; Pettke et al. 2005). Thus, we suggest a similar interpretation for the ca. 360 Ma zircon domains to have had their U–Pb systematics reset by zircon-saturated aqueous-based fluids focused along regional shear zones.

Ti-in-zircon thermometry (Ferry and Watson 2007) provides an estimate of zircon crystallization temperature in the host magma; although widely applied (e.g., Kellett et al. 2009; Claiborne et al. 2010; Dilles et al. 2015; Smythe and Brenan 2016), we note it is not without problems (see discussions in Siégel et al. 2018; Schiller and Finger 2019). Our results suggest zircon in the SMB crystallized in low- to moderate-temperature magmas with values of ∼675 to 825° C for antecrysts (Fig. 11A) and ∼700 to 900 °C for autocrysts (Fig. 11B). Importantly, these and the values proposed below are considered minimum estimates by 50–100 ° C if factors such as P (Ferriss et al. 2008) and elemental substitution (e.g., Hf; Lee and Hattori 2017) are considered (Fig. 11). We further note that the presence of abundant xenocrysts and partially resorbed antecrysts are considered typical of low-temperature peraluminous melts (e.g., Miller et al. 2003). In marked contrast to most of the SMB, data for the DLP suggest hotter and more varied temperatures (i.e., ∼750 to 1000 °C) which are anomalous given the highly evolved, F-rich nature of the DLP where these samples originate (see Dostal and Chatterjee 1995) versus that expected (cf. Černý et al. 2005). We interpret the higher temperatures for the DLP to reflect different melt generating processes than for the rest of the SMB, such as water-undersaturated melting of a lower crustal source (e.g., Brown 2007; Collins et al. 2021).

The derived oxygen fugacities (ƒO₂) values for all zircon domains indicate values of −0.1 to −7 ΔFMQ for the SMB (Figs. 11A and 11B), hence from inception to batholith assembly the ƒO₂ varied from slightly- to strongly reduced relative to FMQ. These values overlap with ΔFMQ values of 1.05 to −5.23 reported by Chavez (2019) for selected SMB plutons (i.e., HP and NRP) using the same protocol but assuming ∼5 wt.% H₂O in the melt. Notably, error propagation of ƒO₂ values in the Smythe and Brenan (2016) method results in ∼1 log unit uncertainty and is particularly sensitive to changes of wt.% H₂O in the melt, e.g., anhydrous melt has 2.5 log units lower ƒO₂ compared to a melt with ∼5 wt.% H₂O.

In contrast to our results, representative I-type granites of the northern Appalachians (e.g., Magaguadavic Granite suite of southern New Brunswick) record typical fO₂ signatures (+2 to +7 ΔFMQ; blue field in Fig. 11B) for I-type granites, although some fractionated I-type granites from the same suite (e.g., Mount Douglas, New Brunswick; peach field in Fig. 11B) do record fO₂ values similar to SMB plutons. The latter is also consistent with similarly low fO₂ signatures for other strongly contaminated I-type intrusions elsewhere (grey field in Fig. 11B; Ague and Brimhall 1988).

Many earlier interpretations of the SMB evolution were based on field, petrographic and geochemical evidence (Clarke and Chatterjee 1988; MacDonald and Horne 1988; MacDonald 2001; Clarke et al. 2004, 2009; Erdmann et al. 2007; Clarke and Carruzzo 2007; MacDonald and Clarke 2017), in addition to isotopic data (O, S; Clarke and Halliday 1980; Longstaffe et al. 1980; Eberz et al. 1991; Poulson et al. 1991), to suggest the SMB was substantially contaminated by metasedimentary rocks of the Meguma terrane. As these potential contaminants are variably carbonaceous (e.g., White 2010), they may have partly contributed to the reduced nature of the SMB. Similar redox conditions to the SMB, however, are recorded globally for other peraluminous granitoid suites (e.g., Sierra Nevada batholith (Lackey et al. 2006), the Chaîne des Puy series in the Massif Central (France et al. 2016), and batholiths in the Lachlan Fold Belt, southeast Australia (Clemens 2003)), suggesting such granites may be inherently reduced.

The Lachlan Fold Belt granitoid suite includes granites attributed to sedimentary protoliths (i.e., traditional S-type) in addition to ones originating via fusion of mafic and chemically primitive igneous material that underplated the crust (i.e., chemically-evolved I-type; Chappell and Stephens 1988). As the lithospheric upper mantle is inferred to be locally heterogeneous in fO₂ (from FMQ−3 to FMQ+1; Ballhaus 1993), intrusive rocks derived from these reservoirs, such as the aforementioned evolved I-type granites, provide the means to impart a reduced signature to melts and thus obviate the need for contamination by carbonaceous crustal lithologies. Lower-crustal protoliths that can likely produce such reduced peraluminous magmas include abyssal spinel peridotites, such as the Beni Bousera and Ronda lherzolite massifs of northern Morocco and southern Spain, respectively (maroon field in Fig. 11A). Notably, the Beni Bousera and Ronda lherzolite massifs represent variably metasomatized lithospheric mantle that underlies large granitoid plutons in southern Spain (Woodland et al. 1992, 2006; Frets et al. 2014) that are similar in character to the SMB.

In contrast to most of the SMB, autocrystic zircon in the DLP have higher fO₂ values (ΔFMQ from −0.8 to −4.5), thus implying its host magma was less reduced in addition to having hotter crystallization temperatures as noted above. The parental DLP magmas therefore likely interacted with a less reduced magma at depth and (or) was less contaminated by reduced metasedimentary wall rocks, a possibility that is further addressed below.

The δ¹⁸O values of zircon constrain sources and contamination of melts with the following assumptions (cf. Roberts et al. 2013): 1) for mantle-derived melts, a δ¹⁸O value of +5.3‰ ± 0.3‰ (Valley et al. 1998); 2) values > 6.5‰ are attributed to melting of altered igneous rocks and (or) assimilation of sedimentary rocks; and 3) values < 5‰ are attributed to melting of hydrothermally altered material (Valley et al. 2005). There is evidence, however, for some δ¹⁸O heterogeneity in the mantle reservoir such that melts may become enriched in ¹⁸O by up to +2‰ due to hybridization of the mantle wedge by subducted sediments containing ¹⁸O-enriched fluids (e.g., Eiler et al. 2005; Auer et al. 2009; Johnson et al. 2009; Roberts et al. 2013).

For the SMB, the overall range in δ¹⁸O_zircon data (+7.10‰ and +9.19‰; Fig. 12) corresponds to the only other available data for the SMB, that of Lackey et al. (2011) who reported values of 8.14‰ ± 0.23‰ (n = 21) for the HP and NRP. Collectively, these data are consistent with the following characteristics with respect to melt source and contamination: (1) the values are higher than expected for primitive mantle-derived melts; (2) they suggest crystallization of zircon from melts with a significant crustal-sourced ¹⁸O component; (3) the slight increase in δ¹⁸O_zircon average values from antecrystic to autocrystic domains is consistent with minor contamination by wall rock but not with crystal fractionation (Valley et al. 2005); and (4) values for the overgrowth/recrystallized domains of ca. 360 Ma suggests late-stage interaction of variably sourced fluids (e.g., Carruzzo et al. 2003; Kontak and Kyser 2011).

We infer from the available δ¹⁸O_zircon data that the SMB crystallized from an isotopically homogeneous magma (in regard to δ¹⁸O), likely derived from a source with a relatively uniform δ¹⁸O or with similar degrees of subsequent contamination everywhere. Importantly, after correcting for zircon-melt fractionation (Δ_WR-zircon ≈ +1.5‰; Valley et al. 1994), the δ¹⁸O_melt values of ca. +9.30‰ to +10.69‰ overlap with the +9‰ to +13‰ for whole rock data of the SMB (compilation in Kontak et al. 2002). Such high δ¹⁸O_zircon and whole rock values are typical of intrusions interpreted to reflect crustal-derived magmas or where extensive contamination of more primitive source melts has occurred (e.g., Taylor and Sheppard 1986).

To assess potential contamination of the SMB melt by the Meguma metasedimentary rocks (e.g., Clarke et al. 2004), the δ¹⁸O_zircon data are compared to whole rock values in the latter which range from +10.1‰ to +13.3‰ (Fig. 12; Longstaffe et al. 1980). Using a mean value of +11.5‰ for this contaminant, mass balance indicates that to obtain the final average δ¹⁸O_melt value of +10‰ using initial δ¹⁸O_melt value of 7.5, 8.5‰ and 9.5‰ (i.e., δ¹⁸O_zircon of 6‰, 7‰, and 8‰, respectively), and assuming similar oxygen contents for melt and contaminant, requires 23%, 13%, and 4% contamination (respectively). For comparison, Clarke et al. (2004) suggested ≤ 30% and Shellnutt and Dostal (2012) ≤20% contamination by Meguma metasedimentary rocks. Although the δ¹⁸O_zircon data are consistent with these estimates, several issues are noted regarding timing and source of contamination: (1) the general uniformity of δ¹⁸O_zircon across the SMB which indicates contamination has no spatial bias; (2) contamination must predate melt emplacement at the current exposure level given δ¹⁸O_zircon values for older antecrystic zircons overlap with autocrystic domains; and (3) contamination by Meguma sedimentary rocks is but one plausible source, with similar rocks in Avalonian basement may be an alternative (see below). The latter possibilities would be inconsistent with a singular model involving contamination of the SMB melts at or near the current level of exposure via stoping. In the DLP, the lower subpopulation of δ¹⁸O_zircon values for antecrystic and autocrystic domains (i.e., +7.3‰ to +7.7‰, Fig. 12) suggests the degree of contamination is less than for the rest of the SMB, thus supporting earlier observations in the DLP from whole rock δ¹⁸O values (Kontak 1990; Kontak et al. 2002).

The Lu–Hf isotopic signature and derived ε_Hf values for zircon domains provide insights into the influence of the composition and the age of isotopic reservoirs on magma composition. For example, variation in ε_Hf is often attributed to source heterogeneity and (or) crustal assimilation during magma evolution (e.g., Gagnevin et al. 2011; Wang et al. 2013; Kemp et al. 2017). Where an isotopically heterogeneous reservoir is tapped, melts can have a more juvenile ε_Hf signature compared to the rest of the complex (e.g., Schaltegger et al. 2019), such as documented herein for zircon from the DLP compared to the rest of the SMB. Below we explore the implications of the isotopic data for the different zircon domains.

The xenocrystic zircon domains have ε_Hf values representative of source and (or) conduit rocks partially assimilated by the SMB melts. Some of these values yield model ages that imply an earlier history of Mesoproterozoic to Archean magmatic events in addition to younger Neoproterozoic and Lower Paleozoic events (Fig. 13A). Two scenarios are explored in regard to these data. The first is in relation to ε_Hf versus U–Pb age distribution patterns that overlap the detrital zircon populations from Ordovician sedimentary rocks in the Anti-Atlas of Morocco (i.e., grey field in Fig. 13A; Gärtner et al. 2016). Thus the data not only support previous studies suggesting the Meguma terrane strata were derived from the West African craton (e.g., Clarke and Halliday 1985; White et al. 2006; Waldron et al. 2009; White and Barr 2010; Henderson et al. 2016), but that these same rocks were also assimilated by the SMB, as noted above. The second scenario relates to the ε_Hf values and associated ages from a majority of SMB xenocrysts closely match the zircon data in Pollock et al. (2015); yellow field in Fig. 13A) for Avalonia igneous and sedimentary rocks from northern Nova Scotia and contiguous eastern Newfoundland (see Fig. 1). Thus, although the xenocrystic data cannot unequivocally discriminate these two identified provenances which both include zircon populations of widely varying Precambrian ages, they do confirm assimilation as part of the SMB petrogenesis.

The ε_Hf values from both autocrystic and antecrystic zircon in the Stage I and II plutons have a combined range that is lower than ε_Hf values of zircon in the DLP (Fig. 13B). The more juvenile DLP values are inconsistent with melting of crust with Mesoproterozoic model ages (i.e., around −10.5 to +0.5, Fig. 13A; Griffin et al. 2004), but <10 epsilon units lower than expected for melts derived from a depleted mantle reservoir (i.e., around +13, Fig. 13B). These data suggest, therefore, the SMB magma petrogenesis involved a somewhat juvenile component, either as a source (i.e., melting mafic crust) or as a mixed/mingled melt phase (i.e., mantle-derived), that subsequently partially assimilated evolved crustal material. This scenario is consistent with the δ¹⁸O zircon values, and permissive with the SMB contaminant being chemically similar to its wall rocks.

Mixing/assimilation of SMB source melts and contaminants should result in systematic variation of the zircon isotopic data in δ¹⁸O versus ε_Hf space, as seen in Fig. 14. This figure presents potential contaminants and zircon values from end-member melts analogous to those involved in generation of the SMB as well as evolutionary pathways of coupled δ¹⁸O-ε_Hf signatures of zircon from globally representative composite intrusive suites: (1) the Tertiary I-type Adamello complex (ADC) in Italy (Schaltegger et al. 2019); (2) two Caledonian S-type granites (CSG) in Scotland (Appleby et al. 2010); (3) a suite of Cretaceous-Tertiary mineralized granites from paired magmatic belts in Myanmar (MG; Gardiner et al. 2017); (4) a suite of Cretaceous barren- and ore-bearing adakitic intrusions in the Handan-Xingtai district, North China (HXD; Sun et al. 2019); and (5) a suite of meta- to peraluminous Late Cretaceous Sn mineralized granites from the Tibetan plateau, China (TPG; Sun et al. 2020).

The potential contaminant and source melt reservoirs of the SMB magmas include (Fig. 14): (1) metasedimentary rocks from the strata in the Meguma terrane (i.e., the Halifax and Goldenville Groups); (2) lower crust, i.e., arc crustal material of the Avalonian terrane that has been suggested to represent a basement to the Meguma terrane (e.g., Keppie et al. 1991; Keppie and Krogh 2000) or to have underthrust the Meguma terrane prior to SMB magmatism (e.g., Keen et al. 1991; Tate and Clarke 1995; Greenough et al. 1999; Shellnutt et al. 2019); and (3a) typical depleted mantle (DM in Fig. 14); or (3b) hybridized mantle material contaminated by limited amounts (10%–20%) of crustal material during continental subduction, similar to the interpreted source for ca. 425–410 Ma Caledonian post-orogenic mafic melts (cf. Couzinié et al. 2016; CMM in Fig. 14). The limitations on characterizing source reservoirs and potential contaminants include: (1) a lack of published Lu–Hf isotopic data for Meguma metasedimentary rocks (note wide range in ε_Hf shown in Fig. 14); (2) the unknown δ¹⁸O-ε_Hf signature of inferred juvenile source melts that contributed to early SMB magmatism; and (3) the range of ε_Hf values shown for Avalonian crustal sources (−10 to +12) as derived from igneous zircon (Pollock et al. 2015; Fig. 13A) whereas the wide range of δ¹⁸O values (+7.0 to +9.5‰; Harris et al. 1997) are from typical arc (I-type) magmas.

Two noteworthy aspects of the SMB zircon data relative to the representative intrusive suites are: (1) the general trend of more negative ε_Hf values commensurate with increasing δ¹⁸O for the reference suites is, in all cases, attributable to a crustal influence; and (2) that the SMB data are highly anomalous in terms of their δ¹⁸O–ε_Hf signatures, as all zircons represented here (along with many examples not shown) have depleted ε_Hf values for similar δ¹⁸O_zircon values to the SMB. Thus, despite the overwhelming evidence that the SMB records some degree of assimilation, the ε_Hf signature of the zircons does not detect this contamination. We suggest that the latter can be accommodated if the assimilated material had similar ε_Hf values, as is the case for Avalonian crust, based on data of Pollock et al. (2015; i.e., most ε_Hf values for analysed zircons fall between −10 and +12, purple field in Fig. 14).

The U–Pb age data presented here constrain the temporal intrusive history of the composite SMB and imply a protracted magmatic history leading to its final assembly. The CA–ID–TIMS dating of oscillatory-zoned zircon autocrysts (Figs. 5A and 8) together with similar ages via in situ SHRIMP and LA–MC–ICP–MS document ca. 8 million years of continuous magma emplacement. This robust data set confirms the less constrained inferences from earlier K–Ar and ⁴⁰Ar/³⁹Ar ages (see Table 1). This time-frame is typical for mid- to upper-crustal batholiths constructed by incremental melt injection over a 5–10 million years (e.g., Brown and McClelland 2000; Coleman et al. 2004; Broderick et al. 2015; Samperton et al. 2015; Schaltegger et al. 2019).

The age population distribution and isotopic characteristics of xenocrystic zircons reflect partial assimilation of various rock packages during magmatic evolution (Fig. 15A): (1) the ca. 440 Ma metavolcanic rocks of the Rockville Notch Group (cf. White et al. 2018; Fig. 15B) and (or) their intrusive equivalents (e.g., the 439 + 4/−3 Ma Brenton Pluton; Keppie et al. 2018); (2) the Ediacaran-Cambrian Goldenville Group (Waldron et al. 2009, and Pothier et al. 2015; Fig. 15C); and (3a) the lower crust as represented by granulite xenoliths in the Pope's Harbour dyke (Owen et al. 1988; Eberz et al. 1991; Shellnutt et al. 2019; Fig. 15D) or (3b) zircons from Avalonian Neoproterozoic arc crust (Owen et al. 1988; Eberz et al. 1991; Greenough et al. 1999; Keppie et al. 2018; Shellnutt et al. 2019).

The age and chemistry of antecrystic and autocrystic zircon domains typically record incremental stages of magma generation and evolution, including magma-mixing and host-rock assimilation (e.g., Hoskin and Schaltegger 2003). In the SMB, the partially preserved antecrysts are chemically indistinguishable from their respective autocrystic domains (i.e., wide range of Hf, Ce/Ce*, and Eu/Eu*, and bivariant spread of Th/U and Yb/Gd values), which indicates zircon formation, resorption, and reprecipitation under similar magmatic conditions. Furthermore, Ti-in-zircon geothermometry indicates crystallization temperatures for antecrysts overlap with those calculated for their respective autocryst domains (Fig. 11). We interpret the antecrysts to represent the earliest record of SMB magma that crystallized in its melt-rich roots ∼3 to 15 million years before batholith assembly.

In contrast with the consistent δ¹⁸O and ε_Hf values from the SMB magmatic zircon (∼8.5‰–9‰ and −3 to +1.5, respectively; Figs. 12B–12F and 13B), the DLP zircons have a subpopulation of lower δ¹⁸O values (∼7.5‰; Figs. 12B, 12F) and overall higher ε_Hf values (∼+1.5 to +4.0; Fig. 13B and 13C). This contrast implies the DLP formed from a more juvenile source or was less contaminated in its evolution relative to other SMB plutons. Analogous to the early DLP evolution may be that of the ca. 373 Ma Shelburne Pluton diorite (Fig. 1A), located ∼20 km southwest of the DLP. It has been suggested that hybridization and mingling of a mantle-derived magma play a significant role in the origin of this pluton (Currie et al. 1998). The water-undersaturated melting of a lower crustal source that we have interpreted from higher crystallization temperatures in the DLP zircon would require the presence of such mafic melts early in the DLP petrogenesis and thus likely sourcing from a replenished melt reservoir with respect to the rest of the SMB (Fig. 14).

Subaerial exposure of the SMB by ca. 360 Ma implies rapid uplift (from initial depths of ∼10–12 km, e.g., Kontak and Kyser 2011; Jamieson et al. 2012) and exhumation of the Meguma terrane relative to Avalonia between ∼375 and 360 Ma (Fig. 2; e.g., Keppie 1979; Keppie and Dallmeyer 1995; Murphy 2000). Between ca. 370 and 360 Ma, dextral transpression along shear zones related to the Meguma-Avalonia terrane boundary (i.e., the Minas Fault Zone, Fig. 1; Murphy et al. 2011) is recorded by dynamically recrystallized muscovite and biotite (⁴⁰Ar/³⁹Ar; e.g., Keppie and Dallmeyer 1987). As noted by Horne et al. (1992), the outline of the SMB and internal pluton contacts reflect a structural control by basement features. Thus, crustal-scale faults (Fig. 1B) were likely active during assembly of the SMB and acted as magma conduits focusing incremental melt addition to an inflating upper crustal complex (i.e., repetition of Figs. 16A and 16B) undergoing extension. These faults also likely accommodated the marked differential uplift in the region between Avalonia and Meguma.

The youngest U–Pb zircon ages are for domains of both recrystallized and unzoned neomorphic zircon overgrowths at ca. 360 Ma (± 10 Ma; Fig. 6) and the observed discordance in some analyses is interpreted cautiously as the result of surface-related Pb-loss and (or) significant underestimation of common Pb in these domains (cf. Heaman and Parrish 1991). However, these ages do correlate with known events: (1) exposure of the SMB, as constrained by the clasts of the SMB in the late Famennian – Early Mississippian Horton Group (Martel et al. 1993; Murphy 2000); (2) regional fluid flux, as recorded ages for new mineral growth (e.g., muscovite, apatite) and partial resetting of others (e.g., Ravenhurst et al. 1987; Culshaw and Reynolds 1997; Hicks et al. 1999; Keppie et al. 2002; Reynolds et al. 2004; Archibald et al. 2018); (3) ca. 360 Ma magmatism in the southwest Meguma terrane noted previously (e.g., Wedgeport and Clayton Hill plutons); and (4) regional transtension that provided the accommodation space for deposition of Horton Group strata (Murphy 2000; Waldron et al. 2015). Thus, the anomalously young zircon domains that occur in samples proximal to faults through the SMB may be local expressions of upwelling geothermal fluids along regional structures.

Inception of SMB magmatism at ca. 395 Ma was coeval with regional deformation (i.e., ∼415–390 Ma) associated with the dextral transpression between the Avalon and Meguma terranes and destruction of the Rheic Ocean. A re-entrant of the Rheic Ocean is, however, interpreted to have existed between the Meguma terrane and western Avalonia during this time (van Staal et al. 2021). Thus the transpressional boundary between these two terranes likely transitioned to a subduction zone along the Laurentian margin (Fig. 17A); the presence of ca. 400 Ma subduction-related magmatism in the NE United States supports this tectonic setting (e.g., Gibson et al. 2021; Fig. 1A). For this magmatism to commence at ca. 395 Ma and terminate by ∼370 Ma in the Meguma terrane, the Rheic oceanic re-entrant may have formed a divergent double subduction beneath both Laurentia and the Meguma terrane (Fig. 17B), analogous to the Cenozoic evolution of the Molucca Sea of SE Asia (e.g., Zhang et al. 2017). A similar model of re-entrant closure by double subduction has also been applied to the western Lachlan fold belt (SE Australia) to explain the spatial and temporal relationships of wide-scale Early- to Middle Devonian granitoid and later bimodal magmatism (Soesoo et al. 1997).

Sustained magmatic temperatures, as recorded by zircon, have previously been attributed to internal (radioactive) heat production of over-thickened crust (e.g., Lucazeau and Vasseur 1989; Cuney 1990; Jamieson et al. 1998). Although the Meguma crust was thickened and the lower crust likely dehydrated (as indicated by the presence of granulite basement xenoliths; Owen et al. 1988; Eberz et al. 1991) during Neoacadian convergence, heating from thermal relaxation (e.g., England and Thompson 1984; Gaudemer et al. 1988) of such a crust is typically expected to only produce ∼5 wt.% melt over 40 million years after thickening (e.g., Thompson and Connolly 1995). Modelling of water-fluxed lower crust melting processes (∼8 kbar), however, show that addition of ∼4.5 wt.% water into a dioritic crust at 900 °C can generate up to 40% melt, and at 750 °C will produce ∼30% tonalitic to trondhjemitic melts (Collins et al. 2020). Thus, the addition of water is important for generation of large volumes of melt in an anhydrous basic lower crust (Castro 2020). For the SMB magmas, melt derivation from a re-hydrated mantle indicates direct infiltration of water during subduction by ca. 395 Ma (Fig. 17B). Regional magmatism of this age elsewhere in the northern Appalachians is also interpreted to reflect subduction-related processes, followed by slab-breakoff and mantle upwelling (e.g., van Staal et al. 2009, 2021; Wilson et al. 2017; Kellett et al. 2021).

The terminal stages of SMB emplacement is represented by ∼375 to 370 Ma Stage II plutons and includes the DLP with the most isotopically juvenile zircons. The Lu–Hf and O isotopic signatures of these zircons suggest the DLP parental melts formed from either metasomatized mantle-derived magmas or melting of an unexposed mafic lower crust. A candidate for the latter crustal component is that of the Avalonian arc crust, which would have previously (∼380-375 Ma; Fig. 17C) underthrust the Meguma terrane during convergence between these terranes.

Analogous to the Molucca Sea (cf. Zhang et al. 2017), double subduction collision between the Meguma-Avalonia boundary may have been followed by sinking of a subducted slab into the mantle (Fig. 17C and 17D). Initial subduction would have subjected the underthrust Avalonian sub-continental lithospheric mantle (SCLM; Fig. 17D) to partial melting, generating SMB magma that gestated in a zone of melting, assimilation (i.e., Avalonia rocks), storage, and homogenization in the lower crust (i.e., MASH in Fig. 16A). Eventually, SMB magma ascended and became contaminated by Meguma terrane metasedimentary rocks at its emplacement level (Fig. 16B). Subsequent sinking of a subducted fossil slab would have subjected the underthrust Avalonian crust to an upwelling asthenosphere with the increased temperatures resulting in lower crustal melting (Fig. 17D). Underthrusting by the Avalon terrane, we suggest, can account for the more juvenile isotopic signature of DLP zircon compared to other parts of the SMB. Asthenospheric upwelling would have resulted in thinning of the lithosphere beneath the Avalonia/Meguma terrane boundaries and thus re-activation of deep crustal faults, allowing for the significant differential uplift between these terranes along the Laurentian margin (Fig. 17D). In addition, the thinned lithosphere is more likely to have been susceptible to oroclinal bending (e.g., Schliffke et al. 2021), as has been suggested for the Meguma terrane during the Alleghanian orogeny after ca. 320 Ma (Warsame et al. 2021).

A comprehensive U–Pb zircon dating study of the 7300 km² composite peraluminous SMB yields a broad range in ages for various CL-defined zircon domain types. The CA–ID–TIMS autocryst zircon dates indicate that batholith construction was long-lived, but crystallized in two stages: (1) early granodiorite to monzogranite between 378.7 ± 0.6 and 375.6 ± 0.5 Ma and (2) the less voluminous but highly fractionated plutons between 375.4 ± 0.8 and 371.8 ± 0.8 Ma. Although two intrusive events indicate ca. 8 million years of overlapping magmatic activity, an expanded record of this magmatism is indicated by in situ SHRIMP and LA–MC–ICP–MS analyses of antecrysts that record multiple crystallization events that are ca. 3–15 Ma older than dated autocrysts (i.e., ∼395 to 380 Ma); hence, magma generation in the lower crust significantly preceded magma emplacement into the middle crust.

In situ SHRIMP, SIMS, and LA–MC–ICP–MS analyses for all zircon domain types document: (1) abundant xenocrysts (∼400 Ma to 2.2 Ga), a majority of which share similar age distribution and isotopic signatures to both Meguma metasedimentary detrital zircons and Avalonian igneous and sedimentary rocks; and (2) similar (with the exception of the DLP) δ¹⁸O, REE, and ε_Hf signatures, as well as derived fO₂ values. These data support magma generation via a modified MASH model involving partial melting in a metasomatized mantle or SCLM regions followed by contamination at different crustal levels that involved Avalonia igneous suites (deep) and Meguma metasedimentary rocks (shallow). The DLP has a zircon ε_Hf signature (+1.74 to +4.38) that is higher than those in the rest of the SMB (−2.99 to +1.68), on average higher Ti-in-zircon crystallization temperatures and lower δ¹⁸O values. These data collectively suggest a different model for its generation that involved less contamination by Meguma metasedimentary rocks than suggested by Clarke et al. (2004; ≤30%) and Shellnutt and Dostal (2012; ≤20%).

The documented longevity of SMB magmatism reflects the geotectonic evolution of this region. Initial melt generation was synchronous with regional deformation that records destruction of the Rheic Ocean, preceded by subduction that initiated incipient melting of a metasomatized mantle source. Cessation of subduction and terrane collision was followed by slab failure, asthenospheric upwelling, and crustal attenuation. Collectively these processes provided the means to generate large volumes of hybrid melt (i.e., mantle and crust) and facilitate its ascent into a brittle upper crust undergoing dextral transpression and post-accretionary relaxation. The latter allowed upward migration of the magma and final emplacement where it assimilated the host Meguma metasedimentary rocks. Its final stage of evolution is recorded by widespread overgrowth and annealing of zircons at 360 Ma which coincides with its uplift, denudation and exposure and more distal emplacement of granites of similar age.

This project was variably funded from an NSERC-CRD (with Avalon Advanced Materials Inc.), NSERC Discovery Grants to DJK, IMS, and JBM, and Natural Resources Canada as part of the Targeted Geoscience Initiative (TGI-5) program of the Geological Survey of Canada (DAK). The authors acknowledge Isabelle Coutand for assistance with sample preparation and thank James Brenan, Nadia Mohammadi, David Lentz, Bill Mercer, Fergus Tweedale, and late Trevor McHattie for thoughtful discussions that aided geological interpretations, but the authors are wholly responsible for any factual errors or misinterpretations. We would like to thank Nicole Rayner for assistance with SHRIMP analysis. The authors would also like to thank the reviewers Barrie Clarke and Chris MacFarlane for detailed and constructive comments.

Supplementary data are available with the article at https://doi.org/10.1139/cjes-2021-0097.