Uranium-lead (U-Pb) accessory mineral petrochronology has been increasingly used to constrain the timing of tectonometamorphic events. However, because mafic rocks commonly lack minerals with a high U/Pb ratio, they may be underrepresented in the chronologic record. This study on polymetamorphic mafic granulites from the Archean Rae craton (northern Canada) provides a striking example of a metamorphic cycle that has been entirely overlooked. We utilized Lu-Hf garnet geochronology and equilibrium phase diagram modeling to characterize two high-pressure granulite-facies mineral assemblages that affected the 2.6 Ga protolith. Zircon and garnet recrystallization occurred at 1.87 Ga within a gneissic foliation, while a coarse-grained garnet precursor nucleated 230 m.y. earlier during a stage of high heat flow within thickened lower crust, the latter of which is nearly absent in the zircon and monazite age record except for rare igneous occurrences. Combined garnet geochronology and petrological modeling reinforce a ca. 1.9 Ga age for high-grade overprinting in the southern Rae craton and clearly show within the same sample that U-Pb accessory minerals did not grow during a newly identified 2.11 Ga granulite-facies event.

Uranium-bearing accessory minerals are common in crustal rocks and widely used to constrain the timing and duration of igneous, metamorphic, and deformation events. Petrochronology using minerals with a high U/Pb ratio (high-U/Pb) enables important linkages between age records and petrological processes (e.g., Kohn et al., 2017); nevertheless, these records may be incomplete, especially in mafic rocks where high-U/Pb minerals may be lacking. Zircon and monazite, for instance, may not be stable together, such that single stages are recorded by one or the other depending on rock composition (Kooijman et al., 2017) and some stages may not be recorded at all (Kelsey et al., 2008; Yakymchuk and Brown 2014; Kohn et al., 2015). Additionally, in rocks with a polymetamorphic history, linkages between accessory mineral petrochronology and a sample’s petrological record may be ambiguous (e.g., Dragovic et al., 2016). Herein, we demonstrate that direct dating of the petrogenetic indicator mineral garnet can provide crucial time constraints for poorly recorded metamorphic processes in high-pressure (HP) polymetamorphic mafic rocks.

The southeastern Rae craton (northern Canada; Fig. 1A) exemplifies a polymetamorphic amphibolite- to granulite-facies region (Berman et al., 2007) where U-bearing accessory minerals have been used to constrain the timing of numerous tectonometamorphic events that occurred between initial crust formation at 2.7–2.6 Ga to stabilization after 1.82 Ga (Fig. 1B; e.g., Berman et al., 2013; Bethune et al., 2013; Dumond et al., 2015). HP (>13 kbar) relict assemblages in mafic rocks that remained intact after 1.9 Ga pervasive deformation and recrystallization are argued to have crystallized during collisional orogenesis at 2.55 and/or 1.9 Ga (e.g., Flowers et al., 2008; Martel et al., 2008). However, these assemblages have only been dated indirectly by U-Pb zircon (e.g., Baldwin et al., 2004; Flowers et al., 2008) or monazite in metasedimentary rocks adjacent to the undated HP mafic units (Dumond et al., 2017). Importantly, the highest metamorphic pressures are commonly preserved by mineral assemblages within mafic units (e.g., O’Brien et al., 2001; Carswell et al., 2003; Baldwin et al., 2004), so direct dating of their crystallization ages is necessary for ascertaining accurate reconstructions of their tectonic histories. To better evaluate the timing and nature of HP metamorphism during the complex history of the Rae craton, we have dated two texturally distinct garnet generations in a HP mafic granulite. Using Lu-Hf garnet geochronology with complementary petrological modeling, garnet dates presented herein correspond to a HP stage that has remained largely undetected in the U-Pb mineral record of the southeastern Rae craton.

Southern Rae craton basement comprises Neoarchean (ca. 2.7–2.5 Ga) intermediate gneiss and intrusions with minor mafic gneiss metamorphosed between 2.60 and 2.50 Ga (e.g., Davis et al., 2015; Dumond et al., 2015; Regis et al., 2017). The Arrowsmith orogeny (Fig. 1B) mainly affected the northwestern margin of the southern Rae craton from 2.54 to 2.28 Ga (e.g., Berman et al., 2013). From 2.28 to 2.00 Ga, regional extension is manifested by sedimentation, anorthosite magmatism, and formation of mafic dikes (Rainbird et al., 2010; Card et al., 2014; Regan et al., 2017). These processes did not favor zircon growth (Fig. 1B), and, thus, their records are dwarfed by those of other events. Widespread 1.94–1.89 Ga regional metamorphism during the Taltson and Snowbird orogenies (Fig. 1B; Berman et al., 2007; Bethune et al., 2013) preceded melt generation and exhumation facilitated by crustal-scale shear zones that occurred between 1.90 and 1.76 Ga (Flowers et al., 2006; Mahan et al., 2006; Regis et al., 2017; Thiessen et al., 2018).

To better understand the aforementioned HP relicts, we analyzed a mafic granulite (sample 15ET249) that has comparable metamorphic assemblages to other HP (>13 kbar) rocks preserved in the southern Rae craton (e.g., Baldwin et al., 2004; Mahan et al., 2008). Our sample was collected from a belt of highly-tectonized granulite units within the Wholdaia Lake shear zone (Thiessen et al., 2018), a >300-km-long crustal-scale shear zone that exhumed mid- to lower-crustal rocks adjacent to the Snowbird tectonic zone (STZ) between 1.90 and 1.86 Ga. Sample 15ET249 from this study has 2.6 and 1.9 Ga U-Pb zircon age components and contains a rare relict metamorphic domain (M1) that was not recrystallized at 1.9 Ga during pervasive ductile shearing along the Wholdaia Lake shear zone.

The M1 domain contains coarse-grained, granoblastic garnet + clinopyroxene + quartz + ilmenite ± plagioclase with >5-mm-diameter garnet porphyroblasts (Fig. 2A). These minerals typically display polygonal boundaries and do not preserve a solid-state tectonic fabric. Garnet (Grt1) occurs as a roughly equant porphyroblast with an irregular outer margin surrounded by 1-mm-diameter Grt1 fragments suggesting a larger initial crystal size by at least 10 vol% (Fig. 2A). Major element composition profiles in Grt1 are generally flat; however, Grt1 margins and two internal fractures have relatively low Mn (Fig. 2B), suggesting a locally focused secondary equilibration event, which is the opposite of what is expected for resorbed garnet margins (e.g., Kohn and Spear, 2000). Lutetium concentrations vary between 0.5 and 2.5 ppm across Grt1; the higher concentrations correlate with zones of low Mn (Fig. 2B). Plagioclase is closely associated with Grt1 grain boundaries and was likely produced during decompression and Grt1 consumption. Amphibole occurs along fractures and is considered a late, retrograde phase.

The M2 domain envelops the M1 domain and consists of a fine-grained <1 mm granoblastic-polygonal assemblage of garnet + plagioclase + clinopyroxene + quartz + ilmenite + titanite (Fig. 2A), which defines a prominent gneissic foliation. Quartz occurs as lobate inclusions in garnet (Grt2) or within interstitial spaces. Major element compositions are generally homogeneous within all major phases and are comparable to M1 phase compositions (Table DR1 in the GSA Data Repository1).

Pressure-temperature (P-T) isochemical phase diagrams were constructed (Fig. 3; Figs. DR2–DR3 in the Data Repository) for sample 15ET249 to constrain equilibrium conditions recorded by M1 and M2. We segmented the thin section into the two texturally distinct domains and combined modal phase proportions with their representative compositions to calculate independent effective bulk compositions for each domain (Table DR1). Based on the calculated stable equilibrium assemblage field of the observed major silicates Grt1 and clinopyroxene (>1 vol%) and complemented by isomodes for garnet (originally ∼35 vol% or higher assuming resorption) and plagioclase (initially plagioclase free), minimum P-T conditions for the M1 assemblage are 12 ± 1 kbar and >700 °C (Fig. 3). Accessory rutile was not preserved in the M1 assemblage and may have been lost during pervasive overprinting in the ilmenite stability field. However, it should be noted that the stability of minor phases is not accurately predicted with this technique (e.g., Forshaw et al., 2019). Three 1–3 mm fragments of the Grt1 porphyroblast shown in Figure 2, together with a M1 matrix clinopyroxene, were subjected to Lu-Hf analysis, yielding a statistically valid isochron regression (Fig. 3; Table DR2) with an apparent age of 2111 ± 3 Ma (mean square weighted deviation [MSWD] = 0.41). The M2 assemblage equilibrated at P-T conditions of 8–11 kbar and >750 °C (Fig. DR3). Lu-Hf analyses of two separates of clean, single 1 mm Grt2 crystals, with no anchoring mineral or whole-rock point, lay along a 1.87 Ga reference isochron and yield a two-point date of 1870 ± 40 Ma (Fig. 3), which is within error of metamorphic zircon age components in the same sample (1890 ± 33 Ma; Thiessen et al., 2018).

M1 and M2—Together, Yet 230 Million Years Apart

Major and minor element concentrations provide no evidence for the presence of multiple age domains in Grt2. The crystals are large enough to be unaffected by 176Hf* (radiogenic Hf) diffusion, which would have required temperatures of at least 900 °C (Smit et al., 2013). Although evidence of diffusive uptake of Lu in these grains is lacking, such an effect remains possible (Bloch et al., 2015), and thus, the Lu-Hf date for M2 may be subject to slight “younging”, which in extreme cases, may be a few tens of millions of years (Smit et al., 2013). Regardless, the ca. 1.87 Ga date for Grt2 overlaps metamorphic zircon ages obtained from this sample (1890 ± 33 Ma; Thiessen et al., 2018) and is consistent with the time of regional metamorphism (1.92–1.89 Ga; Martel et al., 2008) and shearing along the Wholdaia Lake shear zone (1.90–1.86 Ga; Thiessen et al., 2018). The date is slightly older than the age for Wholdaia Lake shear zone footwall anatexis (1845–1824 Ma; Regis et al., 2017), which immediately preceded exhumation documented by 40Ar-39Ar hornblende ages of 1840–1769 Ma (Regis and Kellett, 2018).

The Grt1 Lu-Hf age provides a striking new result, distinct from previously recognized zircon age components for this sample (2.6 and 1.9 Ga). For several reasons, the 2.11 Ga Grt1 date likely reflects primary growth rather than resetting or mixing with older 2.6 Ga garnet. Significant diffusive loss of 176Hf* is unlikely, as closure temperatures (>1000 °C) for the grain size of Grt1 are well beyond the peak estimated temperature, even when considering an unrealistically slow cooling rate of 2 °C/m.y. (Smit et al., 2013). Also, diffusive reuptake of Lu as observed by Kelly et al. (2011) is not evident. However, we do observe small increases in Lu toward the rim and within healed fractures (Fig. 2B), but they do not display typical Rayleigh fractionation patterns and are, therefore, likely secondary features. Furthermore, the low-Mn Grt1 rims and fractures are not typical features for resorbed garnet, in which Mn usually is enriched due to diffusive reuptake. The reason for the coupled low Mn and higher Lu is unclear, yet this feature appears to be part of Grt1. The higher Lu concentrations in the Grt1 rim and fractures correspond to those obtained from the solution analyses (Table DR2), which could indicate that the 2111 ± 3 Ma age is skewed toward the rims. However, the sampled 1–3 mm garnet fragments may preferentially preserve core domains over rims given the irregular and fragile outer margins of Grt1 (Fig. 2). Even if the isochron age is skewed toward the volumetrically significant rim domains, the analyzed fragments include both core and rim isotopic values that together yield a precise age with no resolvable scatter indicating that both features formed at 2111 ± 3 Ma. Also, age skewing due to dissolution of inclusions rich in 176Hf is unlikely due to the high Lu/Hf ratios obtained (Table DR2). It is possible that the 2.11 Ga Lu-Hf age is the result of mixing between Lu-enriched domains (possible 1.87 Ga M2) and a 2.6 Ga core. Such extreme age mixing or partial resetting has been described in the literature (Herwartz et al., 2011; Dragovic et al., 2016), and has yielded a high degree of scatter for the calculated isochrons. If affected by these processes, similar degrees of scatter should be expected for Grt1 considering that each data point consisted of large fragments (>1 mm), each presumably with variable proportions of Lu-rich material. However, this scatter is not observed; the Grt1 isochron is remarkably precisely constrained. This may not theoretically exclude a mixing hypothesis but sets such an unrealistic number of exact requirements to the volume and concentration of the mixing components that this option is considered highly implausible. Age skewing due to diffusive gain of Lu (Bloch et al., 2015) is also possible especially because we observe a Lu enrichment along fractures and rims of M1 garnet. However, if diffusive gain occurred well after initial M1 garnet growth, then notable scatter should be seen in the M1 isochron reflecting distinct Lu/Hf domains, which is not observed. Lastly, we note that the current decay constant uncertainties for 176Lu (Söderlund et al., 2004) do not facilitate overlap of the M1 Lu-Hf and U-Pb ages determined for this sample. Thus, we conclude that 2111 ± 3 Ma very closely approximates the time of growth of M1, including Lu/Mn heterogeneities observed in Grt1, at pressures of 12 kbar or higher.

Decoupling of Lu-Hf and U-Pb Chronometers

Microtextural analyses and petrologic modeling of sample 15ET249 reveal a two-stage evolution separated by ∼230 m.y. with high-grade garnet growth at 2111 ± 3 Ma and 1870 ± 40 Ma. The sample preserves U-Pb zircon populations at 2.6 and 1.9 Ga, while no concordant 2.11 Ga zircon dates occur (Thiessen et al., 2018). Lu-Hf garnet and U-Pb mineral dates are commonly offset due to their growth along distinct intervals of a single P-T path or owing to their contrasting responses to reequilibration of petrological assemblages (e.g., Smit et al., 2014). Zircon growth within HP mafic granulites is predicted to occur mainly upon cooling and decompression rather than during prograde metamorphism (Kohn et al., 2015) due to the breakdown of Zr-bearing phases. Because the M1 assemblage is anhydrous, cooling alone at 2.11 or 1.9 Ga may not have been sufficient to liberate Zr and grow new zircon. Additionally, metamorphic zircon within sample 15ET249 at 1.9 Ga is interpreted as recrystallized rather than neocrystallized (Thiessen et al., 2018). This suggests that minimal Zr was liberated during the 1.9 Ga M2 event and that deformation may have been necessary for recrystallization and isotopic resetting along 2.6 Ga zircon rims and, thus, may explain the lack of zircon recrystallization during the static M1 event.

High-pressure metamorphism at 2111 ± 3 Ma occurred during an interval of time when no major tectonic events have been recognized within the southern Rae craton (Fig. 1B); however, rare igneous events dated at 2.1 Ga possibly related to regional (failed) rifting have been documented (Card et al., 2014; Regan et al., 2017). Regardless, the M1 assemblage is an example of a refractory mafic granulite in thickened lower crust that has persisted through extreme metamorphic cycles (e.g., Cutts and Smit, 2018) and preserves a unique age component that is absent in the zircon age record.

The 2.11 Ga garnet age highlights a need to reexamine the timing of similar HP events in polymetamorphic terranes where growth decoupling of U-Pb accessory minerals and garnet in mafic granulites may have occurred. The results obtained here show that substantial improvements and reevaluations of tectonic models dealing with the HP evolution of the southern Rae craton can be made through garnet chronology, and that further work to discern the nature and extent of 2.1 Ga metamorphism is required. This study also highlights how a major metamorphic process within the lower crust can remain undetected if the affected rocks do not co-stabilize U-Pb accessory minerals and have been subsequently extensively deformed.

Funding was provided by a GEM2 Natural Resources Canada bursary to Thiessen, a Natural Sciences and Engineering Research Council of Canada (NSERC) grant to Gibson (03808), the Northwest Territories Geological Survey, a NSERC Scholarship to Cutts (475186), and an NSERC grant (04080) and the Canadian Foundation of Innovation and British Columbia Knowledge Development Fund (229814) to Smit. Thanks to Kyle Larson for chemical mapping and laser ablation–inductively coupled plasma–mass spectrometry analyses, to Edith Czech for electron probe microanalyses, and to Matt Power for scanning electron microscopy maps. This manuscript benefited from an early review by Bill Davis and discussions with Brendan Dyck, Tyler Ambrose, Derek Thorkelson, and Jacob Forshaw. We thank Ethan Baxter, Mike Williams, Besim Dragovic, and an anonymous reviewer who greatly improved the clarity of the manuscript and helped refine our arguments.

1GSA Data Repository item 2019207, methods and supporting data for phase equilibrium modeling and major element determinations; Lu-Hf isotope analyses; and trace element determinations of garnet, is available online at http://www.geosociety.org/datarepository/2019/, or on request from editing@geosociety.org.
Gold Open Access: This paper is published under the terms of the CC-BY license.