Quantitative elemental mapping of granulite‐facies monazite: Textural insights and implications for petrochronology

Texturally complex monazite grains contained in two granulite‐facies pelitic migmatites from southern Baffin Island, Arctic Canada, were mapped by laser ablation‐inductively coupled plasma‐mass spectrometry (using spot sizes ≤5 µm) to quantitatively determine the spatial variation in trace element chemistry (with up to 1,883 analyses per grain). The maps highlight the chemical complexity of monazite grains that have experienced multiple episodes of growth, resorption and chemical modification by dissolution–precipitation during high‐grade metamorphism. Following detailed chemical characterization of monazite compositional zones, a related U–Pb data set is re‐interpreted, allowing petrologically significant ages to be extracted from a continuum of concordant data. Synthesis of these data with pseudosection modelling of prograde and peak conditions allows for the temporal evolution of monazite trace element chemistry to be placed in the context of the evolving P–T conditions and major phase assemblage. This approach enables a critical evaluation of three commonly used petrochronological indicators: linking Y to garnet abundance, the Eu anomaly to feldspar content and Th/U to anatectic processes. Europium anomalies and Th/U behave in a relatively systematic fashion, suggesting that they are reliable petrochronological witnesses. However, Y systematics are variable, both within domains interpreted to have grown in a single event, between grains interpreted to be part of the same age population, and between samples that experienced similar metamorphic conditions and mineral assemblages. These observations caution against generalized petrological interpretations on the basis of Y content, as it suggests Y concentrations in monazite are controlled by domainal equilibria. The results reveal a c. 45 Myr interval between prograde metamorphism and retrograde melt crystallization in the study area, emphasizing the long‐lived nature of heat flow in high‐grade metamorphic terranes. Such long timescales of metamorphism would be assisted by the growth, retention and dominance of high‐Th suprasolidus monazite, as observed in this study, contributing to the radiogenic heating budget of mid‐ to lower‐crustal environments. Careful characterization of monazite grains suggests that continuum‐style U–Pb data sets can be decoded to provide insights into the duration of metamorphic processes.


| INTRODUCTION
Monazite geochronology is integral in establishing metamorphic timescales, and by association gaining insights into the rates of geodynamic processes (Catlos, 2013;Engi, 2017;Taylor, Kirkland, & Clark, 2016). Increasingly, isotopic analyses are combined with trace element geochemistry to provide additional constraints on the interpretation of the age(s) of a sample within a petrological context (Kylander-Clark, Hacker, & Cottle, 2013). Common inferences include the decrease in heavy rare earth element (HREE) content in monazite being concomitant with garnet growth, and the increase in negative Eu anomaly in monazite relating to feldspar growth (e.g. Dumond, Goncalves, Williams, & Jercinovic, 2015;Rubatto, Chakraborty, & Dasgupta, 2013). Additionally, Th/U in monazite is thought to increase with metamorphic grade, with divergent behaviour during anatexis leading to high ratios in supra-solidus monazite (Yakymchuk & Brown, 2014. Nevertheless, linking these processes to monazite growth requires a number of conditions to be met, including the precise identification of all growth zones in monazite, the analytical volume being smaller than the growth zone, and for monazite growth to be controlled by equilibration volumes larger than the analysed area within the sample. These requirements are particularly challenging for supra-solidus samples, where complex monazite textures and trace element signatures are frequently observed (Kohn, Wieland, Parkinson, & Upreti, 2005;Taylor et al., 2016). An alternative view is that correlations between monazite chemistry and metamorphic grade may be coincidental, and that while correlations may exist for some rocks, to generalize these relationships ignores the potential for complex reaction history in rocks and a range of potential trace element sources, including accessory minerals as well as major phases (Catlos, 2013).
Herein, we use quantitative 4 and 5 µm spot size laser ablation-inductively coupled plasma-mass spectrometer (LA-ICP-MS) mapping of monazite grains to decode the size, shape and trace element signature of growth zones in texturally complex monazite grains within two granulite-facies pelitic migmatites from Meta Incognita Peninsula, southern Baffin Island (Figures 1 and 2). The maps are integrated with a related in situ sensitive high-resolution ion microprobe (SHRIMP) U-Pb data set, thereby constraining the temporal evolution of the monazite chemistry. Combining these data with pressure-temperature (P-T) conditions and modal assemblage changes determined by phase equilibria modelling enables the monazite growth zones to be placed in the context of the evolving phase assemblage, and the metamorphic history of the Meta Incognita Peninsula to be characterized. Th/U and Eu anomalies appear to provide consistent petrochronological indicators, but complex and spatially dependent HREE systematics suggest domainal equilibria and sample specific controls that cannot be generalized.

| REGIONAL GEOLOGY
Southern Baffin Island forms part of the northeastern segment of the Trans-Hudson Orogen (THO), which is a Himalayanscale composite accretionary orogenic belt extending across North America that formed during gradual ocean basin closure between the lower Superior and upper Churchill plates from 1.92 to 1.80 Ga (Figure 1; Corrigan, Pehrsson, Wodicka, & de Kemp, 2009;Hoffman, 1988; St-Onge, Van that continuum-style U-Pb data sets can be decoded to provide insights into the duration of metamorphic processes. The three tectonic levels were progressively juxtaposed across two north-dipping crustal sutures during protracted deformation associated with the THO: the level 2-3 Soper River suture records the accretion of the Narsajuaq terrane to the Meta Incognita microcontinent at 1,845-1,842 Ma (Dunphy & Ludden, 1998;Scott, 1997); and the level 1-2 Bergeron suture formed during terminal collision of the upper plate assemblage with the lower plate Superior craton between 1,820 and 1,795 Ma (Scott & Wodicka, 1998;).

Incognita Peninsula
Meta Incognita Peninsula is underlain by quartzite, marble, psammite and semipelitic units that are lithologically similar to the metasedimentary strata of the contiguous Lake Harbour Group in its type locality of Kimmirut (Scott, St-Onge, Wodicka, & Hanmer, 1997;St-Onge et al., 1996;Figure 2). Two lithologically and geographically distinct sequences can be recognized across the length of the peninsula. On the western peninsula, the Lake Harbour Group comprises quartzite, garnetiferous psammite, minor semipelite F I G U R E 2 Geological map of southern Baffin Island (from St-Onge et al., 2020), showing the three structural levels and two study samples (blue boxes)  Scott et al., 1997). In contrast, extensive garnetiferous psammite interlayered with pelite and/or semipelite, with less than 5% marble and calc silicate rocks ('Markham Bay sequence'; Scott et al., 1997), is exposed on the eastern portion of the peninsula. Both sequences are cut by generally concordant sheets of mafic to ultramafic rocks (Liikane et al., 2015). The Lake Harbour Group is interpreted as a metamorphosed and deformed clastic-carbonate shelf succession (Jackson & Taylor, 1972), which shows a decrease in platformal quartzite and marble, commensurate with an increase in deeper water clastic sedimentary strata, from west to east on the Meta Incognita Peninsula (St-Onge, Rayner, Steenkamp, & Skipton, 2015).
St-Onge, Wodicka, and Ijewliw (2007) documented two regional metamorphic episodes (M 1,2 ) across the Meta Incognita Peninsula, with reported P-T conditions from phase equilibria modelling, and ages from U-Pb monazite and zircon dating by isotope dilution-thermal ionization mass spectrometry (ID-TIMS). M 1 is marked by a regional foliation (S 1 ), which progressively transposes from a schistosity that can be distinguished from bedding (S 0 ) in low-strain, low-grade areas, to a gneissosity that is parallel to lithological contacts in high-grade areas. Peak granulite-facies conditions of 8.3 kbar and 810°C are suggested as part of a counter-clockwise P-T path at 1,849-1,835 Ma, and ascribed to Narsajuaq terrane collision and Cumberland batholith emplacement. M 1 resulted in partial melting of metasedimentary units in the region, forming patch to stromatic migmatites ( Figure 3). M 2 is more localized in extent, restricted to the vicinity of D 2 thrusts, and defined as amphibolite-facies conditions of 6 kbar and 710°C as part of a clockwise P-T path at 1,820-1,813 Ma, causing retrograde metamorphism of granulite-facies assemblages and related to collision of the Superior craton with the Churchill plate.

MINERAL CHEMISTRY
Two stromatic migmatites from the Meta Incognita Peninsula were selected for integrated petrographic, thermobarometric and geochronological analysis (blue boxes, Figure 2; Figure S1). The two pelitic samples were acquired from the Palaeoproterozoic Lake Harbour Group, with sample 14SAB-O31A2 (hereafter referred to as O31A2) sourced from the Kimmirut sequence, and sample 14SAB-A27A2 (hereafter referred to as A27A2) from the Markham Bay sequence. Petrographic images of the samples are given in Figure 4, whole-rock major-element oxides are provided in Table S1, and electron probe microanalysis (EPMA) mineral compositional analyses are listed in Tables S2 and S3 (Methods).

| O31A2
Sample O31A2 exhibits a layered fabric, comprising coarsegrained leucocratic quartzo-feldspathic domains alternating with a fine-medium-grained mesocratic matrix ( Figure S1a). The coarse-grained domains mostly comprise K-feldspar-plagioclase-quartz (Figure 4a), with some K-feldspar replaced by myrmekite, and are interpreted as crystallized former melt (leucosome), whereas the fine-medium-grained domains additionally contain garnet-sillimanite-ilmenite-cordieritebiotite (Figure 4b), and are interpreted as melanosome. The  preferred orientation of biotite laths and sillimanite needles defines a planar fabric (S 1 ; Figure 4b), which is sub-parallel to the mesoscopic layering. Garnet is present as anhedral, elongate to skeletal poikiloblasts in the melanosome. Chemical profiles across garnet grains indicate that almandine (∼0.61), grossular (∼0.03), pyrope (∼0.34) and spessartine (∼0.02) components are constant across grains, except at their margins where pyrope content decreases, and almandine and spessartine increase (Table S2). Based on the morphology and zoning style, the garnet grains are interpreted to have been homogeneous at peak conditions, and to have undergone retrograde net transfer reaction and dissolution (Kohn & Spear, 2000). Garnet grains contain sillimanite needles and biotite laths that are continuous with the external fabric ( Figure 4c). The external fabric also wraps around the garnet grains, suggesting that the porphyroblasts are syn-tectonic with respect to the formation of S 1 .Cordierite (X Mg ∼0.80) forms grains in contact with all other phases throughout the melanosome, as well as forming films around garnet grains ( Figure 4c). Cordierite grains exhibit similar syn-tectonic fabric relations as garnet, but the enclosed aligned sillimanite grains are more prismatic, suggesting that cordierite grew after garnet once the matrix had coarsened. Ilmenite is present as inclusions in garnet and throughout the melanosome, and is occasionally cored by rutile in both petrographic locations (Figure 4d). Ilmenite is interpreted to have replaced rutile as the major Ti-bearing phase during the prograde history of the sample. Accessory apatite, monazite and zircon are present throughout the sample.
Plagioclase grains are variably seriticized (particularly in the leucosome; Figure 4a), cordierite grains are commonly pinitized, and rare biotite flakes in the leucosome are ragged and altered, suggesting that the sample has experienced some post-peak, fluid-assisted retrogression. Undulose extinction in quartz, deformation twins in plagioclase and kink banding in sillimanite prisms indicate that deformation outlasted peak thermal conditions.

| A27A2
Sample A27A2 was collected from an outcrop that exhibits a layered fabric, comprising discontinuous coarse-grained leucocratic quartzo-feldspathic segregations containing cordierite porphyroblasts, which are embedded within a fine-medium-grained mesocratic matrix (Figure 3). The coarse-grained domains dominantly comprise K-feldspar and quartz, and are interpreted as crystallized former melt (leucosome), with rare cm-scale cordierite porphyroblasts that contain sillimanite inclusions interpreted as a peritectic product of biotite-dehydration melting (Figure 4e). The fine-medium-grained domains additionally contain garnetcordierite-sillimanite-magnetite-ilmenite-hercynite-biotite (Figure 4f), and are interpreted as melanosome. The preferred orientation of prismatic sillimanite and elongate garnet grains sub-parallel to the layering define a planar fabric (S 1 ; Figure 4f).
In the melanosome, cordierite is present as porphyroblasts that contain aligned sillimanite needles that are concordant with the external fabric (Figure 4g), as well as thin films mantling magnetite grains (Figure 4f). Cordierite porphyroblasts have a uniform composition (X Mg ∼0.81) throughout all petrographic locations. Garnet is present in two petrographic associations: as inclusion-poor, anhedral porphyroblasts that sometimes contain aligned sillimanite needles concordant with the external fabric, and as grains adjacent to or mantling oxide-rich regions exhibiting symplectic growth with magnetite, hercynite and ilmenite (Figure 4f,h). The former association is suggested to represent prograde growth, and the latter is interpreted to have formed during retrograde metamorphism. Garnet chemistry in both settings is characterized by flat core profiles with increased X Fe and spessartine at the rim, consistent with retrograde dissolution of homogenized grains (Table S3). Magnetite commonly contains blebs of dark green hercynite (X Mg ∼0.31; Figure 4g), which is interpreted to have exsolved during cooling from a high-T solid solution (p. 561, Deer, Howie, & Zussman, 1992). Ilmenite grains occasionally contain rutile. Biotite (X Mg ∼0.58) is present as minor, euhedral S 1aligned flakes in the melanosome, and more commonly as ragged grains located at the margins of garnet, ilmenite and cordierite. The former association is suggested to represent prograde biotite, with the latter interpreted as retrograde in formation. However, it is equivocal if biotite formed part of the peak assemblage as prograde biotite may have been armoured by melanosome phases during peak metamorphism. Accessory apatite, monazite and zircon are present throughout the sample, and similar to sample O31A2, most phases exhibit evidence of high-T deformation.

| Petrographic summary
Both samples are interpreted as pelites that underwent partial (O31A2) to potentially complete (A27A2) biotitedehydration melting, followed by a post-peak evolution characterized by: melt crystallization; minor back reaction of cordierite and garnet to biotite and sillimanite; localized formation of retrograde textures including exsolution, symplectite and corona structures; and minor post-peak, fluid-assisted alteration. Interpreted peak assemblages are biotite-cordierite-garnet-ilmenite-K-feldspar-plagioclase-quartz-sillimanite-melt for sample O31A2, and ±biotite-cordierite-garnet-ilmenite-K-feldpsar-magnetite/ hercynite-quartz-sillimanite-melt for sample A27A2. Rutile inclusions in ilmenite and garnet (e.g. Figure 4d) in both samples suggest rutile was present on the prograde path. Similar porphyroblast-fabric relations in both samples indicate a common tectono-metamorphic history.

MODELLING
To determine the P-T conditions of metamorphism, pseudosections were constructed using thermocalc v3.40 F I G U R E 5 Pseudosection modelling of peak conditions. Bold red font and lines highlight the peak assemblage field. Other coloured lines show zero-mode isolines for garnet (green), H 2 O (light blue), K-feldspar (pink), liquid (black) and plagioclase (orange). (a) P-T pseudosection for the whole rock composition of sample O31A2. (b) T − M H 2 O pseudosection (P = 7 kbar) for sample O31A2, where X = 0 has H 2 O = 0.1 mol.%, and X = 1 has the water content of the whole rock shown in (a) = 2.81 mol.%. The interpreted peak assemblage is in equilibrium with the solidus at reduced water content. (c) As (b), but a P − M H 2 O pseudosection (T = 850°C). (d) P-T pseudosection for the whole rock composition of sample A27A2. All model bulk compositions given in  (Powell, Holland, & Worley, 1998) with data set ds-62 (Holland & Powell, 2011). Although any thermobarometric analysis of migmatite samples requires numerous assumptions to be made regarding peak assemblages, equilibration volumes and water activities, a pseudosection approach is preferred in this study versus conventional thermobarometry because the former relies more on preserved phase assemblage than phase composition, which is more readily defensible in terms of preserved equilibrium (Powell & Holland, 2008). For this reason, we focus below on assemblage rather than the composition of phases, particularly as the interpreted peak phases exhibit textures suggesting that their chemistry has been modified on the retrograde path (e.g. Figure 4g). Modelling was performed in the 11-component MnO-Na 2 O-CaO-K 2 O-FeO-MgO-Al 2 O 3 -SiO 2 -H 2 O-TiO 2 -O 2 system, using the activity-composition relations of White, Powell, and Johnson (2014). The initial bulk-rock compositions used for modelling (Table S4) were calculated by conversion of the whole-rock ICP-MS analyses (Table S1) to model-ready bulk compositions (Palin, Weller, Waters, & Dyck, 2016). Measured P 2 O 5 was used to adjust CaO to account for apatite contributions, by assuming that half of the measured P 2 O 5 was derived from apatite (with the rest from monazite; Weller et al., 2013). The measured loss-on-ignition was used to infer a H 2 O content of the samples, following correction for C, S and FeO determined by titration (Lechler & Desilets, 1987). As the samples used for whole-rock analysis contained both leucosome and melanosome ( Figure S1), the model bulk compositions are considered broadly representative of the samples at peak conditions. However, as the samples exhibit some retrogression, such as sericitization of plagioclase grains (e.g. Figure 4a), the measured H 2 O content is likely an overestimation of the peak H 2 O content. Therefore, P-and T − M H 2 O pseudosections are constructed below to assess the effect of reduced water content on the peak assemblage. Furthermore, the preservation of granulite-facies assemblages in the samples suggests that some melt has been lost from the samples (White & Powell, 2002). Therefore, melt-reintegrated pseudosections are calculated below when considering prograde history. Figure 5a shows a P-T pseudosection for sample O31A2. The inferred peak assemblage (red text) is stable over a small triangular P-T area at ∼7 kbar and 850°C. At lower temperatures, magnetite joins the assemblage, at higher pressures cordierite is unstable and at lower pressures biotite is lost from the assemblage. Although covering a small area of P-T space, phases exhibit relatively rapid modal changes across the peak assemblage field due to the breakdown of biotite. The coincidence of peak conditions within this region could be a consequence of the highly endothermic nature of biotite-dehydration melting, which causes thermal buffering as the reaction proceeds (Schorn, Diener, Powell, & Stüwe, 2018). However, the modelled solidus (bold black line) is ∼200°C lower than proposed peak conditions, suggesting that the modelled water content is too high.

| O31A2
To assess the effect of lower water content in the petrological model, a T − M H 2 O pseudosection was constructed at the peak pressure (7 kbar), with T varying from 670 to 900°C, and M H 2 O ranging from 0.1 mol.% at X = 0 to the measured whole-rock H 2 O content at X = 1 ( Figure 5b).
The solidus position (bold black line) increases in temperature as the water content is reduced, as excess H 2 O (blue line), then muscovite, are lost from the sub-solidus assemblage. At relatively low water content (M H 2 O ∼0.10 to 0.15), the interpreted peak assemblage (red text) is in equilibrium with the solidus. Notably, the temperature of the peak assemblage field does not vary with water content, staying constant at ∼850°C. An equivalent P − M H 2 O pseudosection (at T = 850°C; Figure 5c) shows similar behaviour, with the peak assemblage in equilibrium with the solidus at reduced water content, and the pressure of the peak field insensitive to varying water content, staying constant at ∼7 kbar. Collectively, this analysis suggests that although there is significant uncertainty on the peak H 2 O content (and associated melt mode), the spectrum of possible hydration states all yield the same peak P-T estimate of 7 kbar and 850°C. Figure 5d shows a P-T pseudosection for sample A27A2. If biotite is included as part of the peak assemblage, identical peak P-T conditions of ∼7 kbar and 850°C are suggested. Pressure is constrained by Crd-Bt coexistence as above, but in this case the lower temperature limit is bound by the onset of plagioclase stability (orange line). The calculated magnetite phase in the peak assemblage field contains y(mt)=molar Al/ (Al+Fe 3 +2Ti)=0.10, consistent with the interpretation that the currently observed magnetite (measured y(mt) = 0.005; F I G U R E 6 Pseudosection modelling of prograde metamorphism. Melt reintegrated P -T pseudosection for sample (a) O31A2 and (b) A27A2. Thick coloured lines show zero-mode isolines for garnet (green), H 2 O (light blue), K-feldspar (pink), liquid (black), plagioclase (orange), rutile (purple) and staurolite (yellow). (c-f) Associated garnet (c-d) and total feldspar (e-f) vol.% maps for sample O31A2 (c, e) and A27A2 (d, f). White cross highlights peak P-T estimate from Figure 5 (±1σ). All model bulk compositions given in Table S4 | 861 WELLER Et aL.   Table S3) with blebs of hercynite (measured y(mt) = 0.95) may have formed a solid-solution at peak conditions. If biotite was not stable at peak conditions, higher temperatures at <7 kbar are required. Seriticized cordierite grains ( Figure 5e) and biotite flakes rimming garnet and ilmenite ( Figure 5g) attest to some post-peak fluid infiltration. Therefore, a similar logic is suggested for sample A27A2 as for 031A2: the effective bulk composition during peak metamorphism may have had a lower H 2 O content than calculated in Figure 5d, which would raise the solidus temperature, but not be influential in changing the peak P-T estimate. For both samples the peak assemblage P-T area is smaller than the uncertainty of the technique, which is suggested to be at least ±1 kbar and 50°C (Palin et al., 2016;Powell & Holland, 2008). The agreement between samples suggests similar peak P-T conditions were experienced across the Meta Incognita Peninsula.

| Prograde history
The preservation of granulite-facies assemblages requires that some or all melt has been lost from a sample (White & Powell, 2002). Therefore, when considering the prograde history of a granulite-facies migmatite, it is generally necessary to apply a 'melt reintegration' correction (White, Powell, & Halpin, 2004). There are many different methodologies for such a correction (Bartoli, 2017), with most approaches aiming to achieve a protolith solidus that is minimally water-saturated at a relevant pressure, achieved via the introduction of 'lost' melt in one or more steps.
Here, we apply a simple one-step melt reintegration calculation, as our interest is only on the topology of a plausible protolith (i.e. not on quantifying melt loss history in detail). The presence of rutile inclusions in both samples (e.g. Figure 4d) compared with ilmenite in the matrix suggests that higher pressures were experienced during prograde metamorphism (e.g. Weller et al., 2013). As such, melt integration is applied to the whole-rock bulk composition such that the new solidus is water-saturated at 10 kbar. The melt composition used is that of the 10 kbar whole-rock solidus. An example T-X pseudosection showing the methodology is given in Figure S2. Following this approach, 16 and 1 mol.% melt are required to refertilize the A27A2 and O31A2 bulk compositions respectively (Table S4). Figure 6a,b shows P-T pseudosections for the melt-reintegrated bulk compositions, with the pseudosections extended to lower temperatures than Figure 5. As per the assumptions above, the solidus (black line) is now water-saturated (blue line) at 10 kbar for both samples. The O31A2 topology is relatively unchanged owing to the similar bulk composition, although at lower temperatures rutile (purple line) is stable above 9 kbar at 550°C, consistent with the interpretation that the sample experienced a clockwise approach to peak conditions. The topology of A27A2 is similar at high temperatures as the whole-rock pseudosection, but the increased fertility of the sample results in staurolite (yellow line) becoming part of the stable prograde assemblage.
Figure 6c-f shows the corresponding garnet and feldspar vol.% changes, and are of relevance in future discussions of monazite trace element signatures. The calculations show similar garnet systematics and abundances throughout the considered prograde P-T range (Figure 6c,d). Both samples also feature similar feldspar systematics (Figure 6e,f), although sample O31A2 contains approximately half of the total feldspar content, and features a much smaller K-feldspar increase associated with muscovite breakdown (the contributions made by plagioclase and K-feldspar to the total feldspar content shown in Figures S3 and S4), likely owing to the less aluminous bulk composition.

| MONAZITE GEOCHRONOLOGY
To constrain the timing of metamorphism, in situ U-Pb isotopic analyses of monazite were carried out using a SHRIMP II at the Geological Survey of Canada (GSC). Automated full thin section scans were performed on the GSC's Zeiss Evo Scanning Electron Microscope to locate monazite suitable for geochronological analysis. Targets from all petrographic contexts were prepared for analysis by coring multiple thinsections using a diamond-tipped drill bit and mounting in epoxy (Rayner & Stern, 2002). Back-scattered electron (BSE) images and EPMA Y and Th maps were generated for each monazite grain to identify internal compositional domains, and to guide analytical spot placement (Figures S5 and S6). A 10 µm diameter spot was used for SHRIMP analyses. Data reduction protocol and U-Pb calibration details are reported as footnotes in Tables S5 and S6. Isoplot (version 3.0; Ludwig, 2003) was used to generate mean ages with related statistics. 207 Pb*/ 206 Pb* dates are reported as these values provide the most precise age estimate for Palaeoproterozoic rocks (Mattinson, 1987;Schoene, 2014).

| O31A2
Monazite is present throughout sample O31A2. Leucosomehosted monazite is characterized by homogeneous to oscillatory-zoned appearances, whereas melanosome-and garnet-hosted monazite typically has patchy cores surrounded by homogenous to oscillatory-zoned rims ( Figure S5). Twentyone analyses were acquired from 19 grains, yielding concordant dates from 1,870 to 1,803 Ma ( Figure 7a; Table S5). The
older dates are typically associated with patchy core domains with relatively high Y content (>3,000 ppm) and low Th/U (<∼20), whereas the younger dates are associated with oscillatory-zoned or homogeneous monazite domains with low Y content (<3,000 ppm) and high Th/U (>∼20).
If the data set is separated by Y content, monazite analyses with Y >3,000 ppm yield a weighted mean 204 Pb-corrected 207 Pb/ 206 Pb age of 1,862.4 ± 6.1 Ma (MSWD = 1.8, n = 7), and monazite analyses with Y < 3,000 ppm yield an age of 1,823.0 ± 7.2 Ma (MSWD = 1.5, n = 12; Figure 7b). The oldest analysis in the Y < 3,000 ppm population (2-66.1) and the youngest analysis in the Y > 3,000 ppm population (1-21.1) both straddle high-Y grain cores ( Figure S5), and thus likely represent mixed domains and are excluded from the calculations of the means (dashed outlines, Figure 7b). The c. 1,862 Ma monazite age is initially interpreted as a prograde constraint for sample O31A2, as high Y content is thought to be associated with crystallization of monazite concomitant with low garnet abundance on the prograde path (Rubatto et al., 2013). The c. 1,823 Ma monazite age is interpreted to record crystallization of the melt following peak metamorphism, because the population is associated with all of the monazite located in the leucosome (Table S5), and the associated oscillatory zoning in some analysed grains is indicative of crystallization from a melt (Aleinikoff, Burton, Lyttle, Nelson, & Southworth, 2000). Both interpretations are re-evaluated following LA-ICP-MS mapping of a sub-set of the grains, as described below.

| A27A2
Monazite in sample A27A2 is present in the leucosome, the melanosome matrix and as inclusions in garnet. The F I G U R E 7 SHRIMP U-Pb data. All underlying data given in Tables S5 and S6. (a) Sample O31A2 concordia diagram, colour coded by petrographic position. (b) Sample O31A2 data displayed by age order, colour coded by Y content (as measured by the SHRIMP). Dashed outlines indicate analyses excluded from the mean; see text for details. (c) Sample A27A2 concordia diagram, colour coded by petrographic position. (d) Sample A27A2 data displayed by age order, colour coded by Y content (as measured by the SHRIMP). Note that these data are re-interpreted in monazite grains exhibit complex textures, with leucosome monazite typically having patchy, oscillatory and/or homogenous domains, whereas garnet-and melanosome-hosted monazite has a concentric or patchy appearance, with all monazite variably outlined by homogenous rims ( Figure S6). Thirty-four analyses were acquired from 21 monazite grains, yielding a spread of concordant dates from 1,861 to 1,737 Ma ( Figure 7c; Table S6). Older dates are generally associated with higher Y content and patchy monazite domains, whereas the younger dates typically have lower Y content and are located in rim domains ( Figure S6). Nevertheless, the data set cannot be partitioned into statistically valid age populations associated with any consistent chemical or petrographic criteria (Figure 7d). It remains unclear whether this spread reflects geological processes that are not temporally discrete, or the complicated monazite textures reflect episodic growth with chemical signatures that cannot be resolved using BSE, Th and Y systematics alone. To test these two scenarios, LA-ICP-MS mapping of a sub-set of grains was conducted.

MONAZITE GRAINS
Quantitative elemental mapping of seven monazite grains from sample A27A2 and two monazite grains from sample O31A2 was performed using LA-ICP-MS, with elemental map data acquired on 24 isotopes ( Figure S7, Table S7). The objective of the mapping is to decode the complex monazite textures by identifying domains in each monazite grain that share the same extended trace element signature, both to provide further constraints on the SHRIMP data set, and to explore trace element systematics in monazite that have experienced granulite-facies metamorphism. Data were acquired by performing a number of contiguous, parallel line scan ablations across selected grains. All mapping was performed using a spot diameter of 4 µm, except for samples 27-1-120 and 27-2-132, which used a spot diameter of 5 µm.
Only pixels having a recorded Ce signal of >15% of the typical Ce signal for pure monazite were included in the maps. This criterion effectively removed minerals other than monazite from the maps, yet did not over-aggressively remove pixels along grain margins where narrow, chemically distinct monazite overgrowths occurred in some grains (see results below and Figure S7). Where relevant, possible contamination or dilution of trace element signals by adjacent grains or inclusions was screened through consideration of Si, Al and Fe pixel maps, and mass balance considerations, as discussed below. Twenty-four elemental maps were generated for each of the nine analysed grains, with each set of maps consisting of over 24,000 data points. Figure S7 shows an example set of maps acquired for one grain (A27A2-2-132). Given the size of the data set, analysis of these data are presented in Figures 8-11 and 13 for a representative selection of grains, whereas the remainder are presented in Figures S9-S12. Each figure focusses on one monazite grain, and features the BSE images and qualitative EPMA Th and Y maps associated with the SHRIMP dating, as well as LA-ICP-MS data presented in map form and as chondrite-normalized (CN) REE diagrams. As discussed below, the Y map (d) in each figure is overlain by an interpretation of the growth zones of the grain, following analysis of the LA-ICP-MS data, with grains shown in the supplementary figures exhibiting similar systematics to grains described in this paper.

| A27A2 leucosome grain 2-132
Monazite grain 2-132 is located in the leucosome (Figure 8a). Viewed with BSE, the grain exhibits a patchy interior and a homogeneous exterior that contains abundant sillimanite inclusions (Figure 8b). Patchy core regions are associated with lower Th and higher Y contents, with a thin low-Y second rim also apparent (Figure 8c,d).
During mapping, LA-ICP-MS data were acquired for 1,285 locations (pixels) on the grain. All data feature a decrease in CN concentration from light to HREE, with a prominent negative Eu anomaly, with both the extent of HREE fractionation and the size of the Eu anomaly varying across the grain (Figure 8e). Colour coding the data by Th/U shows that the data also exhibit systematic variation in Th/U, for example analyses with the lowest Th/U have the smallest Eu anomaly and shallowest REE slopes. However, as so many data are displayed on one plot, it is difficult to ascertain whether the data are indicative of continuous variation, or define chemically distinct groups. Therefore, bivariate Y-Eu/Eu* plots (colour coded by Th/U in Figure 8f, and data density in Figure 8g) are displayed to evaluate this proposition, where Eu/Eu* = Eu CN /(Sm CN x Gd CN ) 1/2 . Yttrium content is considered to be a proxy for HREE content (Rubatto et al., 2013), such that plots of Y versus Eu/Eu* capture the dominant intra-grain trace element variation noted above. While the slope of Yb/Gd could be calculated directly to avoid use of a proxy, Y content is used as an axis such that the plot can be compared with the Y EPMA map (Figure 8d). Furthermore, Figure S8 shows a comparison of Y and Yb/Gd and demonstrates that the same conclusions would be drawn from either analysis. Figure 8f indicates that the data define two trends between three end-members (numbered dashed circles). Data density (Figure 8g) indicates that most data are located at end-member 2, and that data intermediate to the end-members are mostly a mixing signal from LA pits that straddled domain boundaries. The interpretation that the trace element data array is indicative of multiple distinct chemistries (rather than continuous variation) is supported by the qualitative Y map (Figure 8d) being characterized by three main intensities.
Analysis of the quantitative Y 2 O 3 , Eu/Eu* and Th/U LA-ICP-MS maps (Figure 8h-j) allow the three end-members identified in Figure 8f,g to be located with respect to the grain (annotations, Figure 8d). End-member 1, which is characterized by the highest Y and Eu/Eu*, and lowest Th/U, corresponds to the patchy core domains apparent in BSE, Th and Y images. Dominant end-member 2, which has the lowest Eu/Eu* and highest Th/U, corresponds to the bulk of the rest of grain. End-member 3, which features the lowest Y content, corresponds to the thin rim observed on the Y map, as well as forming isolated patches within the grain. The low-Y rim is multiple pixels wide in places, and also characterized by La and U enrichment ( Figure S7), indicating that the signal is not the result of contamination by silicate phases marginal to the grain. The maps confirm that domains 1-2 and 2-3 are common neighbours, whereas domains 1-3 are rarely adjacent to each other; hence only two main mixing trends are observed on Figure 8f. The location of the three SHRIMP analyses (dashed white circles, Figure 8d) correspond to domains 1 and 2, with the oldest analysis present in domain 1, consistent with the implied sequence of growth.

| A27A2 leucosome grain 1-100
Monazite grain 1-100 is located in the leucosome (Figure 9a). Viewed with BSE, the grain is relatively homogeneous, with a bright and irregular inner rim (Figure 9b). Thorium and, in particular, Y maps indicate that most of the grain features oscillatory zoning (Figure 9c,d). LA-ICP-MS data (Figure 9e) reveal that the grain is characterized by HREE fractionation and a negative Eu anomaly. In Y-Eu/Eu* space (Figure 9f,g), two main groups are observed: a major population of low Eu/Eu*, variable Y and high Th/U analyses (labelled '2' due to chemical similarities with domain 2 in grain 2-132; Figure 8f); and a population of relatively high Eu/Eu* analyses with low Y and Th/U ('3').
By consideration of Figure 9h-j, group 2 corresponds to the majority of the grain, which is characterized by oscillatory zoning. Unlike grain 2-132, where low-density spreads in Y-Eu/Eu* data between narrow areas of high density were interpreted as mixing between discrete end-members (Figure 8g), the spread of high-density data for domain 2 on Figure 9g indicates that domain 2 chemistry is heterogeneous. Nevertheless, as these pixels correspond to the oscillatory zoned region observed in Figure 9d, the range in trace element data is still considered to represent a discrete monazite growth domain. Low Y and high Eu/Eu* group 3 occurs along the rim of the grain.
The dashed line on Figure 9f,g represents an interpreted mixing trend. The mixing line occurs at the low-Y end of the domain 2 range, which is also characterized by the highest Th/U, suggesting that these traits are characteristic of the bright inner rim apparent in BSE that marks the boundary between domains 2 and 3. The location of the SHRIMP analyses (dashed white circles, Figure 9d) correspond to domains 2 and 3, with the youngest analysis present in domain 3, consistent with the implied sequence of growth.
Monazite grain 1-100 closely resembles grain 1-116 ( Figure S9), with both grains located in the leucosome of 3 High KDE Low sample A27A2 and featuring a zoned core exhibiting variable Y at low Eu/Eu* of ∼0.05, with a high Eu/Eu* and low Th/U rim. Associated SHRIMP dates for monazite grain 1-100 are consistent with the implied sequence of growth events ( Figure S9d).

| A27A2 leucosome grain 1-120
Monazite grain 1-120 is located in the leucosome (Figure 10a), and exhibits a patchy texture in BSE that is closely mirrored by zoning in Th, and also apparent on the Y map (Figure 10b-d). LA-ICP-MS data (Figure 10e) reveal that the grain is characterized by HREE fractionation and a negative Eu anomaly, with the lowest Eu CN analyses having the highest Th/U. In Y-Eu/Eu* space (Figure 10f), a triangular cluster of data are observed. Consideration of the data density (Figure 10g), and the zoning observed in Figure 10h-j, suggests that the triangular data cluster comprises four groups (dashed circles, Figure 10f,g), with intermediate data representing mixed analyses. Three of the groups define a linear trend, featuring variable Y at a relatively constant Eu/Eu* of ∼0.05. Due to their overlapping chemistry with domain 2 from A27A2 leucosome grains 2-132, 1-100 and 1-116, the linear trend groups are labelled 2a-c, with the a, b, c designations representing the interpreted sequence of growth events based on their spatial distribution (e.g. 2a forms islands within 2b, with both rimmed by 2c; Figure 10d). The 2a-c sequence suggests that monazite growth first decreased and then increased in Y at constant Eu/Eu*. The fourth group-the minor population of relatively high Eu/Eu* analyses ('3')-belong to the rim of the grain. Associated SHRIMP dates are all consistent with the implied sequence of growth events (Table S6).

| A27A2 melanosome grain 1-39
Monazite grain 1-39 is located in an aggregate of hercynite-magnetite-sillimanite in the melanosome (Figure 11a). Viewed with BSE, the grain exhibits a patchy core, with a homogeneous but discontinuous rim (Figure 11b). Similar appearances are observed in Th and Y maps (Figure 11c,d). LA-ICP-MS data (Figure 11e) reveal that the grain is characterized by HREE fractionation and a negative Eu anomaly, with a greater variation in Eu CN content than previous examples.
In Y-Eu/Eu* space (Figure 11f,g) the data are relatively diffuse, reflective of the patchy zoning. With the context provided by the other samples, three chemical trends are suggested, but as the systematics are less well-defined than previous examples, large dashed ovals are used to demarcate interpreted growth zones on Figure 11d (via analysis of Figure 11h-j), with no attempt made to separate out mixed analyses. Nevertheless, similar spatial systematics as in previous examples are observed, with a relatively high Y, high Eu/Eu*, low Th/U domain ('1') corresponding to an irregular region internal to the grain, mantled by a low Eu/Eu* (∼0.05) domain ('2') exhibiting variable Y and high Th/U, and rimmed by a high Eu/Eu* region ('3'). Domain 3 is also present as an isolated patch internal to the grain. The two SHRIMP analyses correspond to domains 1 and 3, with the oldest analysis present in domain 1.
Two other grains (1-25, Figure S10, and 1-27, Figure S11) were also analysed from the same melanosome area of the thin section (Figure 11a). Both grains show similar systematics to 1-39, with patchy and anhedral internal domains exhibiting domain 1-type chemistry, overgrown by mantles and rims exhibiting domain 2-and 3-type chemistry, respectively. Associated SHRIMP dates are all consistent with the implied sequence of growth events.

| A27A2 synthesis
LA-ICP-MS data from all seven grains analysed in sample A27A2 are plotted in Figure 12a, partitioned into leucosome-(green) and melanosome-hosted (blue) monazite, with the leucosome data further separated into the two analysed leucosome areas (dark green leucosome 1 relates to grain 2-132). Overlying the data are the chemical domains interpreted for each grain (Figures 8-11 and Figures S9-S11), with purple ovals corresponding to domain 1, orange ovals to domain 2, and grey ovals to domain 3. Figure 12a indicates that data from monazite present in the leucosome exhibit tighter linear trends with reduced total variation in Eu/Eu*, compared with more diffuse patterns in the melanosome. Nevertheless, similar systematics are observed between all grains, with the coloured ovals defining three broad and distinct regions: a population of high Y analyses with Eu/Eu* of ∼0.10 to 0.40, which belonged to the core of some analysed grains (purple ovals, domain 1); a dominant population of variable Y, low  Eu/Eu* (<0.10) analyses that generally formed the mantle of the analysed grains (orange ovals, domain 2); and a minor population of low Y, high Eu/Eu* (up to ∼0.70) analyses, which formed the rim of the analysed grains (grey ovals, domain 3). The three chemical domains also feature distinct Th/U, with domain 2 analyses consistently characterized by the highest ratios (≥40), compared with lower Th/U in domains 1 and 3 (Figure 12b). Collectively, the implication is that sample A27A2 monazite experienced at least three phases of growth, with distinct trace element signatures.

| SHRIMP data re-interpretation
If the A27A2 SHRIMP U-Pb data set ( Figure 7d, Table S6) is reduced to focus on grains analysed by LA-ICP-MS (coloured bars, Figure 12c), and associated SHRIMP spots are assigned to their interpreted LA-ICP-MS chemical domain, each of the three domains yields a statistically valid single population: domain 1 core regions have an age of 1,849.8 ± 8.2 Ma (MSWD = 0.22, n = 3), domain 2 mantle regions have an age of 1,805 ± 13 Ma (MSWD = 1.6, n = 10) and domain 3 rim regions have an age of 1,790 ± 12 Ma (MSWD = 1.1, n = 3). With the context given by the LA-ICP-MS data, further SHRIMP U-Pb data could theoretically be incorporated into these populations using the associated SHRIMP measurements of Th/U and Y (Table S4), and grain textures ( Figure S6). While such an approach would be relatively objective for domain 2-type analyses, as they have a clear high Th/U association, it is difficult to apply this analysis to domains 1 and 3, as they exhibit the same Th/U and feature overlap in Y content (for the analysed grains). As such, given the complex textures and possibility of mixed analyses, and the potential that grains not chemically mapped by LA-ICP-MS may exhibit different systematics, additional SHRIMP data are not incorporated, nor considered further for sample A27A2. The domain 2 c. 1,805 Ma age is interpreted as the age of melt crystallization during cooling of the sample to the solidus on the retrograde path, based on: the association of this domain with oscillatory-zoned monazite (e.g. Figure 9d; Aleinikoff et al., 2000); the dominance of this signature in leucosome-hosted monazite (Figure 12a); the low Eu/Eu*, interpreted to result from feldspar in the leucosome sequestering Eu; and the high Th/U signature (Figure 12b), which is common for anatectic monazite (Yakymchuk & Brown, 2019).
The domain 3 c. 1,790 age is within error of domain 2, but domain 3 features notably higher Eu/Eu*. Following similar signatures in granulite-facies pelitic monazite reported by Prent et al. (2019), domain 3 growth is interpreted as the result of fluid-rock interaction following melt crystallization, causing alteration of feldspar and liberating Eu. Domain 3 is notable for forming isolated patches within grains (Figures 8d  and 11d), as well as forming thin rims. Assuming that domain 3 represents an interconnected volume, the presence of domain 3 chemistry internal to the grain suggests that the domain 2-3 boundary is irregular, and features lobate incursions into domain 2, which is a feature of modification by dissolution-precipitation (Harlov, Wirth, & Hetherington, 2007Putnis & John, 2010).
Interpretation of the domain 1 c. 1,850 Ma age is less constrained, but it most likely represents a prograde age constraint for sample A27A2, based on the association of this domain with the core regions of monazite grains that exhibit irregular outlines, the increased preservation of this domain in melanosome-hosted monazite, and the lower Th/U relative to domain 2. All of these features are consistent with modelling by Yakymchuk and Brown (2014), which suggests that sub-solidus monazite would mostly be consumed during partial melting, with kinetics not expected to hinder monazite dissolution, except possibly for large grains, or those shielded in the melanosome, with subsequent higher Th/U monazite growth predicted to form during cooling to the solidus. Given that monazite in pelites typically first form from the breakdown of allanite around staurolite-grade conditions (Berman, Sanborn-Barrie, Stern, & Carson, 2005;Kohn & Malloy, 2004;Smith & Barreiro, 1990;Weller et al., 2013), the earliest prograde conditions datable using monazite are likely amphibolite-facies. The Eu/Eu* content of domain 1 supports this interpretation, as modelling of the prograde evolution of the sample indicates that feldspar would have been present, consistent with the observed negative Eu anomaly, and that feldspar content would have increased during prograde metamorphism (Figure 6f), compatible with Eu/ Eu* decreasing towards domain 2.

| O31A2 results
Two monazite grains were mapped in sample O31A2 (the only grains with multiple SHRIMP analyses); grain  Figure 13, and grain 1-105 is displayed in Figure S12. Both grains are hosted by garnet porphyroblasts, and exhibit almost identical textures, trace element systematics and age relations. Consequently, the discussion below applies to both, although focusses on grain 1-130 ( Figure 13).

± 9
High KDE Low 6.3.1 | O31A2 melanosome grain 1-130 Monazite grain 1-130 is located within a fractured, anhedral garnet porphyroblast, which contains aligned sillimanite needles and biotite blades concordant with the external S 1 fabric (Figure 13a). Viewed with BSE, the grain exhibits a mottled core, which contains a weak fabric consistent with S 1 , with a clear rim (Figure 13b). The core area displays patchy zoning with respect to Th and Y contents, with more homogeneous high-Th and low-Y rims (Figure 13c,d). As per sample A27A2, trace element data (Figure 13e) indicate that the monazite grain exhibits variation in the size of the negative Eu anomaly and extent of HREE fractionation. In Y-Eu/ Eu* space, all data can be explained by three end-member chemistries, with data intermediate to the end-members interpreted to be mixed signals from LA pits that straddled domain boundaries (Figure 13f,g). The extent of the three corresponding domains are shown on Figure 13d (via spatial analysis of Figure 13h-j), with the labels being a suggested growth sequence based on inclusion relationships, but with no implied link to sample A27A2. Domains 1 and 2 are both internal to the grain with irregular outlines, and contain low Th/U (≲10) and similar Y content, but markedly different Eu/ Eu* (∼0.64 for 1 versus 0.16 for 2). Domain 3 belongs to the rim, and is characterized by extremely low Y CN (<2,000) and Eu/Eu* (∼0.03), and relatively high Th/U (>20).
The SHRIMP analytical locations (dashed black ovals) are also shown on Figure 13d. The SHRIMP analysis from domain 3 is part of the c. 1,823 Ma population identified on Figure 7b. The low Y and high Th/U of all other analyses associated with this age (Table S5) indicates that domain 3 can be linked to the c. 1,823 Ma age, which was interpreted above to represent the timing of melt crystallization in the sample. However, the SHRIMP analysis of the 'core' area, which is part of the c. 1,862 Ma population (Figure 7d), straddles the chemically distinct domains 1 and 2, and thus represents a mixing age between these two growth events. Consideration of other SHRIMP analytical locations associated with this population ( Figure S3 and S11) indicates that most analyses also likely represent mixing between composite core domains, indicating that the c. 1,862 Ma age (Figure 7b) does not correspond to a single monazite growth event, and therefore is not as geologically meaningful as proposed above. Nevertheless, the low MSWD value related to the c. 1,862 Ma age suggests that the two growth events are not of resolvably different ages. Therefore, this age is still considered an approximate prograde age constraint for sample O31A2.

| DISCUSSION
The integration of petrographic observations (Figure 4), pseudosection modelling of prograde ( Figure 6) and peak ( Figure 5) metamorphic conditions, SHRIMP U-Pb data (Figure 7), and LA-ICP-MS mapping of a sub-set of grains (Figures 8-13) has allowed for the monazite trace element systematics of two granulite-facies pelitic migmatite samples from southern Baffin Island (Figure 2) to be placed in the context of evolving P-T conditions and major phase assemblages ( Figure 14). Analysis of the monazite chemistry focusses on Y, Eu/Eu* and Th/U, as these three parameters effectively capture all of the intra-grain variation (e.g. Figure S7), and are also the criteria most commonly used for petrochronological interpretations of monazite ages (e.g. Dumond et al., 2015;Rubatto et al., 2013;Yakymchuk & Brown, 2014). We first briefly outline the major compositional trends documented for the monazite grains contained in two pelitic migmatites, before comparing and discussing the implications of these results for petrochronological interpretations.
In sample A27A2, three compositional domains are defined within the sub-set of grains analysed by LA-ICP-MS ( Figure 12a). These domains are used to extract petrologically significant ages from a continuum of SHRIMP U-Pb data ( Figure 12c). The oldest population of monazite (domain 1) is characterized by relatively high Y content (Y CN > ∼5,000, equivalent to Y > 8,850 ppm), Th/U <∼50 and Eu/ Eu* of ∼0.10 to 0.40 (Figure 14a). Domain 1 is present as an anhedral core within all melanosome-hosted monazite (e.g. Figure 11), compared with forming isolated patches in just one of the four analysed leucosome grains (Figure 8), and is interpreted to represent prograde growth, which was in large part resorbed during partial melt processes, particularly for leucosome-hosted monazite (Yakymchuk & Brown, 2014).
Domain 2 regions represent the bulk of the analysed monazite (Figure 14c), and feature consistently high Th/U> ∼40, with low Eu/Eu* (<0.10) and a wide range of YCN (2,500-12,500). Based partly on the presence of oscillatory zoning in some of the grains (Figure 9), domain 2 is interpreted to have formed during melt crystallization on the retrograde path, following peak conditions of 850°C and 7 kbar ( Figure 5).
Domain 3, which forms a thin rim to all analysed grains, as well as isolated patches within some grains (e.g. Figure 11d), features relatively low Y CN (2,500-7,000), low Th/U <∼50 and a wide range of Eu/Eu* (∼0.1 to 0.7). This domain, which overlaps in age with domain 2, is interpreted to have resulted from fluid-rock interaction associated with alteration of feldspar, which liberated Eu into the matrix. The interface between domains 2 and 3 is highly irregular, consistent with modification by dissolution-precipitation (Harlov, Wirth, & Hetherington, 2007Putnis & John, 2010).
Three chemical domains are also observed within the two grains analysed in sample O31A2 (Figure 14b), with the numerical order representing the implied growth sequence, but with no implied link to equivalently numbered   (Figure 14d) is attributed to melt crystallization, and features extremely low Y CN (<1,500) and Eu/Eu* (<0.09), with relatively high Th/U (>20).

| Implications for monazite petrochronology
A comparison of the monazite growth systematics documented in the two analysed samples allows for a critical evaluation of what trace element characteristics may be generalized as petrochronological tools, as the samples are both pelitic migmatites that feature similar mineral parageneses and experienced similar granulite-facies peak conditions and a common tectono-metamorphic history.

| Th/U
In both samples, Th/U in monazite increases with metamorphic grade. However, as a generalization, only relative changes within a sample rather than rather than absolute ratios between samples are meaningful in interpreting the data, as the 'high' values of Th/U in sample O31A2 characteristic of melt crystallization (>20) overlap with the 'low' values in Th/U in sample A27A2 characteristic of either prograde growth or late fluid alteration (<50; Figure 14a,b). In both samples, the increase in Th/U from prograde to melt crystallization conditions is driven by Th content increasing ∼3 to 5 fold (Figure 14e,f), and U decreasing by approximately an order of magnitude ( Figure S7). Coupled with the dominance of monazite domains resulting from melt crystallization in each sample (Figure 14c,d), the net effect is to yield a highly Th-enriched signature in the study samples.

| Eu/Eu*
In both samples, a negative Eu anomaly (Eu/Eu* <1) is measured in all monazite grains, consistent with the presence of feldspar (Figure 6e,f). In addition, Eu/Eu* is shown to decrease with metamorphic grade, consistent with an increase in the feldspar content of the external assemblage (both the melanosome, and the crystallizing leucosome). Extremely low Eu/Eu* of ≤0.10 are characteristic of monazite associated with melt crystallization. The low and consistent Eu/Eu* suggests that the melt effectively buffered the Eu budget, likely due to concomitant crystallization of feldspar-rich leucosome. High Eu/Eu* monazite rims in sample A27A2 may be a sign of post-peak alteration of feldspar associated with fluid-rock interaction (Prent et al., 2019).

| Yttrium
Yttrium, which is an effective proxy for HREE fractionation in the analysed samples (e.g. Figure S8), exhibits complex and variable systematics, both within domains interpreted to have grown in a single event, and between samples that theoretically experienced similar garnet behaviour during metamorphism (Figure 6c,d), suggesting no clear link between the HREE content of monazite and garnet abundance. This point is best exemplified by domain 2 monazite in sample A27A2, which is interpreted to have crystallized in the leucosome. In this domain, Y content varies between grains, exhibiting a narrow range in Y (Figure 8g), a continuous spread in Y (Figure 9g, Figure S9), and a linear array of distinct Y (Figure 10g), resulting in a population that exhibits a range of Y content that spans the entire data set (Figure 14a). The variable and oscillating Y content of domain 2 in particular suggests domainal equilibria are controlling the Y content of monazite, rather than a sample-wide, major phase control such as garnet. Further work is required to develop our understanding of the likely myriad (and sample-specific) controls on HREE distribution within a sample, with accessory and precursor phases likely playing an important role (Catlos, 2013), and leucosome dynamics feasibly influential, but LA-ICP-MS mapping studies of a range of phases are providing important insights (e.g. George, Gaidies, & Boucher, 2018).

| Decoding monazite textures
This study highlights the usefulness of quantitative LA-ICP-MS mapping of monazite grains (synthesized with BSE images and EPMA elemental maps) to decode the shape and number of growth events in texturally complex monazite grains. In particular, the maps can highlight complex domain outlines, which can lead to 'internal' rim domains (Figures 8d and 11d), and demarcate where patchy zoning likely corresponds to different growth domains (Figures 11d  and 13d). All of these features may otherwise cause confusion when trying to reduce a data set that uses qualitative textural interpretations to identify different analytical domains. This point is evidenced in this study, where the LA-ICP-MS maps provide new insights into the SHRIMP data set, both assisting in extraction of ages from a continuum of concordant data (A27A2, Figure 12b), and revealing that previously defined age populations were hybrid (O31A2, Figure 13d).

WELLER Et aL.
For sample A27A2, the interpretation that the monazite grew episodically rather than continuously is supported by the observation of distinct monazite growth zones by various imaging techniques (e.g. Figure 8b-d). The implication is that 'continuum' data, as commonly documented in granulite-facies terranes (Taylor et al., 2016), may contain distinct monazite growth events. Conversely, for sample O31A2, the LA-ICP-MS maps reveal that a monazite grain, which was qualitatively interpreted to contain only a core and rim (Figure 13b), actually comprised three distinct growth zones. The implication is that monazite grains may contain more complex growth zoning than apparent, which reinforces the importance of good imaging prior to analysis, and confirms the benefit of integrated U-Pb and trace element data sets (e.g. Kylander- Clark et al., 2013).

| Timescales of granulite-facies metamorphism
Finally, our results characterize the P-T conditions and timescales of granulite-facies metamorphism on Meta Incognita Peninsula, providing a case-study of mid-crustal metamorphism during orogenesis. Samples A27A2 and O31A2 both record peak conditions of ∼7 kbar and 850°C ( Figure 5), similar to reported peak conditions of 8.1 kbar and 830°C for Meta Incognita Peninsula (St-Onge et al., 2007), and ∼6 kbar and 820°C for Lake Harbour Group pelites from Foxe Peninsula, south-west Baffin Island (Figure 2; Smye, St-Onge, & Waters, 2009). Collectively, these results indicate a relatively consistent thermal architecture across southern Baffin Island. Peak metamorphism in this study is suggested to have been reached during a clockwise P-T cycle, owing to observations of higher-pressure rutile being replaced by lower-pressure ilmenite in both study samples (Figures 4d and 6a). St-Onge et al. (2007) suggested an anticlockwise P-T path, but this interpretation was derived from preserved mineral zoning, which is known to be unreliable in the granulite facies due to retrograde diffusion (e.g. Frost & Chacko, 1989).
Prograde metamorphism is constrained at 1,849.8 ± 8.2 Ma for sample A27A2, with melt crystallization at 1,805 ± 13 Ma (Figure 12c), with both ages within error of equivalent constraints for sample O31A2 (1,862.4 ± 6.1 Ma and 1,823.0 ± 7.2 Ma; Figure 7b). These timescales overlap with, but extend, previous suggestions of M1 granulite-facies metamorphism at 1,849-1,835 Ma (St-Onge et al., 2007). The latter age range was mostly constrained using U-Pb ID-TIMS analysis of unabraded monazite grains, which exhibit similar textures to composite grains in this study, and thus the data could in part represent mixed ages that underestimate the duration of M 1 metamorphism. Interestingly, St-Onge et al. (2007) also reported clockwise, retrograde M 2 amphibolite-facies conditions at 1,820-1,813 Ma, associated with thrust structures and fabrics related to D 2 terminal collision of the Superior craton (Figure 2), which overlap with the c. 1,823-1,805 Ma age of M 1 melt crystallization in this study. These data are all consistent with regional metamorphism on southern Baffin Island being driven by a single long-lived thermal event (c. 45 Myr) from c. 1,850-1,860, related to Narsajuaq terrane collision and emplacement of the Cumberland batholith, with M 2 representing zones where complete retrograde recrystallization was permissible due to enhanced fluid-infiltration along D 2 thrusts that overlapped with the end of granulite-facies conditions.
The suggestion of protracted M 1 metamorphism in this study is consistent with calculated timescales of metamorphism from thermal modelling of thickened crust (e.g. Clark et al., 2015;McKenzie & Priestley, 2008). Extended timescales of granulite-facies metamorphism are assisted by the retention of heat-producing elements in mid-to lower-crustal environments. The lack of observed correlation between metamorphic grade and rate of radiogenic heating in global compilations suggests that anatexis does not necessarily lead to crustal stratification of radiogenic heating rates (Hasterok, Gard, & Webb, 2018). The growth, retention and dominance of high-Th supra-solidus monazite in granulite-facies terranes, as observed in this study, could in part explain this behaviour.