Understanding Preservation of Primary Signatures in Apatite by Comparing Matrix and Zircon‐Hosted Crystals From the Eoarchean Acasta Gneiss Complex (Canada)

A novel way to investigate the petrogenesis of ancient polymetamorphosed terranes is to study zircon‐hosted mineral inclusions, which are sensitive to melt evolution such as apatite. Recent contributions on such inclusions in unmetamorphosed granitoids can provide valuable petrogenetic information and, in turn, represent a way to circumvent effects of metamorphism. Yet the impact of metamorphism on apatite inclusion has never been studied in detail. To address the issue of chemical and isotopic preservation of primary signals in apatite crystals both in the matrix and armored within zircons, we have studied apatite crystals from four 3.6–4.0 Ga TTG granitoids from the Acasta Gneiss Complex (Canada). Our results demonstrate that U‐Th‐Pb isotope systematics in matrix apatite crystals were reset at 1.8–1.7 Ga (Wopmay orogen) whereas primary REE signatures were preserved in many crystals. In contrast, zircon‐hosted apatite inclusions all preserved primary REE signatures despite variable ages between 1.7 and 4.0 Ga. We interpret reset ages to be a consequence of metamorphism that managed to affect U‐Th‐Pb systematics because of advanced radiation damage accumulation in host‐zircon lattices. Only the most pristine zircon crystal has an apatite inclusion with a concordant age consistent with the magmatic age of the zircon (4.0 Ga). In addition, our results show that apatite crystals from TTG have distinct REE composition from post‐Archean granitoids apatites, that is preserved even in some apatites with reset ages. This capacity to retain primary information and discriminate granitoid types makes apatite a very valuable tool for reconstructing the nature and evolution of ancient crustal rocks through the use of detrital minerals.


Introduction
Rare-Earth bearing minerals, such as apatite (Ca 5 (PO 4 ) 3 (OH, Cl, F)), have proven to be extremely valuable in retrieving the nature of their host magmas and the crystallization history of granitoids (Chu et al., 2009;Jennings et al., 2011;Bruand et al., 2016). Hence, they can be very useful complements to zircon in crustal evolution studies (e.g., Belousova et al., 2010;Dhuime et al., 2012;Iizuka & Hirata, 2005). Apatite is ubiquitous in different rock types and contains measurable amounts of a large variety of trace elements (e.g., LREE-MREE, Sr, Pb, Mn, halogens; e.g., Prowatke & Klemme, 2006;Gregory et al., 2009;Miles et al., 2014). The concentration of trace elements in apatite is sensitive to melt evolution (in situ crystal fractionation, mixing, SiO 2 and ASI contents; e.g., Sha & Chappell, 1999;Belousova et al., 2001;Prowatke & Klemme, 2006;Zirner et al., 2015). Furthermore, apatite is also datable via U-Pb geochronology although prone to resetting during metamorphism, which makes it problematic to recover meaningful information about the magmatic history of old crustal segments. One way to overcome this limitation is to study apatite inclusions armored within zircon and, thus, combine the potential of apatite in retrieving magmatic information with the resilience of zircon to metamorphic processes. Previous workers have shown that zircon-hosted apatite inclusions can indeed retain important information about its host rock (Bruand et al., 2016;Emo et al., 2018;Jennings et al., 2011). For instance, Bruand et al. (2016) have studied an unmetamorphosed suite of granitoids and showed that zircon-hosted inclusions shared the same trace element signatures as cogenetic apatite crystals within the matrix. They have also shown that Sr/Sm ratio in apatite can discriminate felsic magmas from mafic ones and, as previously suggested by Belousova et al. (2001), that Sr content in apatite inclusions can be used as a proxy for the degree of differentiation of their parental magma (evaluation of the SiO 2 and Sr contents).
Very recently, Emo et al. (2018) demonstrated that zircon-hosted and matrix apatite crystals in two 3.6-3.7 Ga orthogneisses from the Acasta Gneiss Complex (AGC, Canada) exhibit a general discrepancy in their Sr isotope signatures. The authors interpreted the Sr isotopic signature in apatite inclusions as the first direct estimate of initial 87 Sr/ 86 Sr for these ancient rocks whereas Sr isotope signatures in matrix apatite were reset during later metamorphic event. This interpretation is in line with recent findings of Fisher et al. (2019) who demonstrated that Sm-Nd isotope systematics in matrix apatite from Acasta gneisses were reset during the Wopmay orogen at about 1.85 Ga. However, it is presently unclear to which extent zircon-hosted apatite inclusions can resist metamorphic resetting, especially in ancient zircon crystals that experienced significant radiation damage such as those in the AGC (e.g., Guitreau et al., 2018). Consequently, the aim of this paper is to study the preservation potential of trace element and U-Th-Pb isotope signatures in zircon-hosted inclusions and, hence, better constrain the extent to which such approach can allow retrieval of information about the formation of Earth's first crusts. For this purpose, we carefully and systematically measured trace element contents and U-Th-Pb isotope systematics in zircon-hosted apatite as well as in matrix apatite crystals in well-characterized 3.6-3.96 Ga samples from the AGC (Guitreau et al., 2012(Guitreau et al., , 2018Mojzsis et al., 2014).

Trace Element Analysis by LA-ICP-MS
Trace element analysis of zircon-hosted apatite inclusions was done by laser ablation-inductively coupled plasma mass spectrometry (LA-ICP-MS) using the Resonetics Resolution M-50E 193 nm excimer laser coupled to the Thermo Element XR at LMV. The spot sizes used ranged between 4 and 12 μm (depending on the phases analyzed) with a repetition rate of 1 Hz and a fluence of 2.88 J/cm 2 . We purposely set a low frequency to avoid drilling through apatite too quickly. Each run lasted 2 min with 20 s of background acquisition. Each sequence was bracketed by two analyses of GSE-1G standards (9 μm spot size) at the beginning and at the end of the sequence and one GSC-1G in the middle of the sequence (Jochum et al., 2005). Hf content was monitored during analysis of apatite inclusions in order to identify potential concurrent sampling of the host zircon. All analyses with Hf content above the background values were discarded.
Matrix apatites were analyzed by LA-ICP-MS using the same laser as for inclusions but coupled to an Agilent 7500 ICP-MS with a spot size of 40 μm, a frequency of 4 Hz, and a fluence of 4.5 J/cm 2 . The same amount of time as for inclusion apatites was used for data acquisition, and each run started and finished with two measurements of NIST 610 standard (Pearce et al., 1997). For all runs (matrix and inclusions), Durango apatite (Marks et al., 2012) was regularly analyzed throughout the analytical session to monitor data quality at the same spot size as the analyzed unknowns to account for downhole fractionation (Table S2). Internal normalization for LA-ICP-MS data was done using 43 Ca, and data processing was conducted using the GLITTER software (Griffin, 2008).

U-Th-Pb Geochronology by LA-ICP-MS
Apatite crystals were dated by LA-ICP-MS at LMV using a Resonetics Resolution M-50E laser-ablation system coupled to a Thermo Element XR ICP-MS. During U-Th-Pb measurements, the laser frequency was set to 2 Hz, the fluence to 2.5 J.cm −2 , and the spot size between 40 and 9 μm so as to adapt to the size of the crystals targeted for analysis. Analyses consisted in~20s background counting followed by 60s of signal acquisition in peak-jumping mode (5 ms dwell time and 20 μs settling time). Bahia and Xuxa apatite were analyzed for quality check (Rosa et al., 2017;Schuch, 2018). The apatite standard MAD (Thomson et al., 2012) was used for external normalization. Unknowns were analyzed as batches of six and were bracketed by analyses of MAD apatite. Data were processed using in-house spreadsheets based on the t 0 intercept method (Chew et al., 2011(Chew et al., , 2014. Because we do not know the true 207 Pb/ 206 Pb of MAD apatite, which is a combination of radiogenic Pb and common Pb, and that it is not reported in Thomson et al. (2012), one cannot determine the mass bias for 207 Pb/ 206 Pb during LA-ICP-MS analyses. Consequently, we derived the Pb isotope mass bias by using the 208 Pb/ 206 Pb expected from the "true" 232 Th/ 238 U and the age of MAD apatite. Similarly, we estimated the amount of common Pb within analyzed apatite crystals using the difference between the mass-bias corrected 208 Pb/ 206 Pb and that expected from the mass-bias corrected 232 Th/ 238 U in a way akin to that of Chew et al. (2011). Common Pb corrections were further done using the Pb evolution model of Stacey and Kramers (1975) and a convergence method similar to that of Thomson et al. (2012) except that we monitored 206 Pb/ 238 U and 208 Pb/ 232 Th for convergence. U-Th-Pb age calculations were done using a present-day 238 U/ 235 U of 137.818 (Hiess et al., 2012) and decay constants for 238 U, 235 U and 232 Th of 1.55125.10 −10 , 9.8485.10 −10 , and 4.9475.10 −11 , respectively (Jaffey et al., 1971;Le Roux & Glendenin, 1963). The Isoplot software (Ludwig, 2008) was used to build Tera-Wasserburg plots.
As illustrated in Figure S1, common Pb correction results in migration of data points toward the Concordia curve but not complete converge to a concordant age, especially for Bahia apatite data. This can be explained by our correction method, which is well suited for minerals with low Th/U but becomes less precise and less accurate as Th/U increases as a result of radiogenic 208 Pb/ 206 Pb getting closer to that of common Pb (Ireland & Williams, 2003). Consequently, the influence of common Pb becomes less detectable in high Th/U mineral. As an illustration, a 10% contamination of a 2,056 Ma Bahia apatite (Th/U = 15), a 573 Ma Xuxa apatite (Th/U = 30), and a 1,700 Ma Acasta apatite (Th/U = 0.5) decreases determined 208 Pb/ 206 Pb ratios by 5% and 7% in Bahia and Xuxa apatites, respectively, while increasing the 208 Pb/ 206 Pb by almost 150% in the Acasta apatite. This observation can explain why, later in the manuscript, the common Pb corrections work much better for Acasta apatites, which commonly have low Th/U, as most data points become concordant once corrected for common Pb.
Apatite crystals from Acasta gneisses were analyzed in three different contexts: (1) as separate grains within an epoxy mounts (40, 20, and 9 μm spots on the same apatite crystals), (2) as grains within polished sections (matrix apatite), and (3) as inclusions within zircon crystals mounted in epoxy. All results are available in Table S4.

Sample AG09009
Sample AG09009 is mainly made of quartz, alkali feldspar, plagioclase, biotite, muscovite, chlorite, and accessory phases such as zircon, apatite, monazite, allanite, thorite, ilmenite, and rutile ( Figure 2a). Locally, chlorite partly replaces micas and can occur as small veins. Felspars are slightly sericitized. In this sample, apatite is the main accessory phase and most commonly occurs as euhedral crystals (<150 microns in length) either directly in contact with biotite or as inclusions in quartz and feldspar. Apatite can be zoned from light gray (core) to dark gray (rim) in cathodoluminescence ( Figure 2d) and can contain zircon inclusions. Apatite occasionally appears as secondary mineral associated with allanite and monazite. In these rare cases, a monazite grain is surrounded by a corona of apatite, itself surrounded by a corona of fibrous allanite ( Figure 2b). Numerous small inclusions of monazite are persistent inside the apatite corona.
Biotite, alkali feldspar, plagioclase, chlorite, apatite, chromite, and chalcopyrite have been identified as inclusions in zircon ( Figure 2c) and are usually <20 microns. Most zircon crystals are metamict (damaged by radioactivity), which leads to the formation of fractures that can be filled by several minerals (e.g., biotite, chlorite, plagioclase) thereby forming polymineralic inclusions. Most apatite inclusions are monomineralic and euhedral except for one case where apatite is surrounded by chlorite (Figure 2c).

Sample AG09014
Sample AG09014 is made of quartz, biotite, plagioclase, and accessory phases. These latter are zircon, apatite, pyrite, monazite, and rutile crystals. In this sample, apatite is the main accessory phase (Figures 2a-2j), it is euhedral, small (<100 microns) and can contain zircon inclusions. Apatite can be found either as inclusion in quartz and plagioclase or in contact with biotite. Apatite is usually unzoned (Figures 2a-2j). In this sample, inclusions in zircon are monomineralic and small (<10 microns). They are mostly apatite (Table 1), quartz, chlorite, and monazite crystals.

Sample AG09015
In sample AG09015, the main rock forming minerals are quartz, biotite, plagioclase, and oxides (rutile, ilmenite; Table 1). Accessory phases are zircon, apatite, sulfides (pyrite, chalcopyrite), titanite, and allanite. As for samples AG09014, AG09015 is only slightly altered. Apatite is the main accessory phase; it is euhedral and <200 microns. It is either in direct contact with biotite or as inclusion within quartz and feldspar. Most crystals are zoned with a light gray core and a dark gray rim in CL ( Figure 2h). They can also be found as clusters where zonation is oscillatory with a dark gray core ( Figure 2g).
In this sample, inclusions found in zircon are apatite, biotite, plagioclase, and chlorite. These inclusions are small (<5 microns) and monomineralic.

Sample AG09016
Sample AG09016 is made of quartz, plagioclase, biotite, and muscovite. Accessory phases are zircon, apatite, rutile, and titanite. This sample shows limited alteration. Apatite is ubiquitous, euhedral, and <250 microns in size and occurs in contact with biotite or in inclusions in quartz and plagioclase. Apatite crystals have two types of zonation: (i) a light gray core with a dark rim or (ii) an oscillatory zoning with a dark core ( Figure 2f). Some apatite grains are occasionally fractured with biotite filling these cracks. In this sample, inclusions in zircon can be either monomineralic or polymineralic with grain sizes <20 microns. The identified inclusions are apatite (Figure 2e), chlorite, biotite, plagioclase, quartz, and thorite. For this sample, apatite inclusions are always monomineralic.
Contrary to samples AG09009 and AG09016, only matrix apatite has been analyzed in AG09014 and AG09015 samples because inclusions were too small (<10 μm in size).
Sample AG09009. Chondrite-normalized REE patterns of apatites from the matrix define two groups: Group 1 shows depleted LREE compared to HREE and a negative Eu anomaly (Eu/Eu* = Eu N /√[(Sm N ).
(Gd) N ] < 0.5; Figure 3a, Table 2) whereas Group 2 patterns are subparallel to Group 1 but with lower REE content and no Eu anomaly. Apatite inclusions have the same REE patterns as Group 1 with some grains having a higher REE concentration (Figures 3a and 4a). Overall, apatites from AG09009 have the   Geochemistry, Geophysics, Geosystems Geochemistry, Geophysics, Geosystems highest concentrations in La (~166 ppm) and Yb (~353 ppm, Table S4) among all samples analyzed in this study (Figure 4a). These highest values are systematically associated with analyses done within crystal cores whereas lower concentrations are linked to analyses done within rims (Figures 4b-4c). On the other hand, Group 2 apatite compositions have only been found in the outer-part of grains that had core compositions belonging to Group 1 apatites ( Figure 2d) and have the lowest REE concentrations (Figure 4a).
Sample AG09014. Chondrite-normalized REE patterns for apatite from sample AG09014 show similar LREE depletion compare to HREE. In this sample, two groups of apatite can be distinguished. The first group (Group 1) is similar to Group 1 identified in other samples from this study (Figure 3c). The second group is depleted in all REE compared to Group 1 and has a distinctive impoverishment in LREE.
Sample AG09015. Contrary to previous samples, apatite from AG09015 displays multiple and strongly distinguishable compositions. Five groups can be defined based on chondrite-normalized patterns ( Figure 3d). Group 1 REE pattern is comparable to Group 1 from the other samples with a depletion in LREE compared to HREE ( Figure 4a) and a negative Eu anomaly (0.4 < Eu/Eu* < 0.6). All other groups have a strong depletion in LREE, which distinguishes them from Group 1. In Figure 4a, it can be seen that the other groups (2-5) have low La and Yb contents and are undistinguishable from each other (gray dots in Figure 4d). These groups have highly depleted LREE chondrite-normalized patterns (La N < 100) compared to HREE (Yb N < 1,000). The best chemical criterion that allows these groups to be discriminated is their respective Eu anomaly (Figure 3d). Group 2 has no Eu anomaly (0.9 < Eu/Eu* < 1.3), Group 3 a slightly negative Eu anomaly (Eu/Eu* ≈ 0.5), and Group 4 a positive Eu anomaly (Eu/Eu* > 1.2). Group 5 apatite has slightly higher LREE chondrite-normalized values compared to Group 4, and a small negative Eu anomaly (Eu/Eu* ≈ 0.75). Groups 2, 3, and 5 are always neoformed (no zonation, low REE content) or recrystallized grains, while Group 4 can be found as new generation grains or as an overgrowth surrounding Group 2 apatites.
Sample AG09016. Much as for sample AG09009, chondrite-normalized REE patterns of matrix apatites from AG09016 define two groups. Group 1 is depleted in LREE compared to HREE (mean (La/Yb) N = 0.14) with a consistent negative Eu anomaly (~0.3; Figure 3b, Table S4). Group 2 defines a subparallel pattern to Group 1, with lower REE content, a stronger depletion in LREE compared to HREE (c.a. (La/Yb) N = 0.03) and small Eu anomalies (0.6 < Eu/Eu* < 0.7). Chondrite-normalized REE patterns of zircon-hosted apatite inclusions have strong similarities with Group 1 apatite, although the LREE content in the former is more variable (Figure 3b).

Apatite U-Th-Pb Geochronology 5.2.1. Matrix Apatite
Most of matrix apatite crystals in AG09009 define a major trend that intersects the Concordia curve in a Tera-Wasserburg plot ( Figure 5a) at 1,797 ± 27 Ma (MSWD = 3.6; n = 27). Once corrected for common Pb (red ellipses in Figure 5a), most data points plot onto the Concordia curve with a weighted average 206 Pb * / 238 U age of 1,794 ± 24 Ma (MSWD = 1.4). One data point lingers outside the major trend and has a Concordia age of 3,197 ± 82 Ma (MSWD = 0.09) and 3,184 ± 120 Ma (MSWD = 0.11; Figure 5a) before and after common Pb correction, respectively.
Data for AG09014 apatite crystals form a robust trend with a lower intercept at 1,700 ± 15 Ma (MSWD = 1.6; n = 12; Figure 5b). Once corrected for common Pb, most data points plot on the Concordia curve and exhibit a weighted average 206 Pb * / 238 U age of 1,710 ± 20 Ma (MSWD = 0.92).
With the exception of one data point, all analyses from AG09015, show a trend that gives a lower-intercept age of 1,713 ± 21 Ma (MSWD = 4.1; n = 35; Figure 5c). Once corrected for common Pb, all data points plot on or very near the Concordia curve and give a weighted average 206 Pb * / 238 U age of 1,723 ± 20 Ma (MSWD = 1.9). One data point plots on the Concordia and gives an age of 2,721 ± 63 Ma.
All data points for apatites from AG09016 plot along a linear trend toward a lower intercept of 1,746 ± 46 Ma (MSWD = 3.7; n = 37; Figure 5d). Once corrected for common Pb, most data points are concordant and have a weighted average 206 Pb * / 238 U age of 1,721 ± 23 Ma (MSWD = 2.2).

Zircon-Hosted Apatite
Over the large quantity of zircons mounted in epoxy and examined within thin sections (~200), only a few apatite inclusions larger than 10 μm were found. One apatite was analyzed in sample AG09009 whereas

10.1029/2020GC008923
Geochemistry, Geophysics, Geosystems eight were analyzed in sample AG09016. Most zircon crystals in which we found apatite inclusions do not have pristine internal textures, except for zircon AG09016-88 which exhibits fine-oscillatory zoning. Ages for six of the nine host zircons are reported in Figure 6. The three remaining zircon crystals were not dated because of their disturbed internal zoning which would have resulted in meaningless discordant ages. Apatite ages mentioned in this section correspond to data corrected for common Pb unless specified otherwise. Overall, U-Pb isotope systematics in apatite inclusions exhibit contrasting behavior compared to matrix apatite. The former essentially distribute along the Concordia curve between 1,700 and 4,000 Ma whereas the latter forms a robust trend between 1,700 and 1,800 Ma and common Pb (Figure 7). In detail, zircon-hosted apatite inclusions display three distribution behaviors: (i) on or close to the Concordia curve, (ii) on the matrix apatite trend, and (iii) inverse discordance.
The oldest inclusion (AG09016-88) has a concordant age of 4,003 ± 89 Ma ( 207 Pb/ 206 Pb age corrected for common Pb; 2 SE; 0.7% discordance; Figure 7) which is consistent with data obtained on the pristine-looking host zircon (3,967 ± 49 Ma, Guitreau et al., 2018). A second analysis in the rim of the same inclusion shows a reversely discordant age, though consistent with the first one (4,111 ± 90 Ma). Two other data points from apatites AG09016-100 and 108 plot on the left of the Concordia curve at 207 Pb/ 206 Pb ages corresponding to~4,000 Ma. Apatites AG09016-24 and 37 show concordant ages between 3,200 and 4,000 Ma. Apatite AG09016-15 is the biggest inclusion and is not fully enclosed within the host zircon ( Figure 6). The zircon exhibits a dark patchy zoning indicative of advanced radiation damage accumulation.
In this apatite inclusion, three analyses from core to rim were done. Obtained U-Th-Pb ages decrease progressively from core (~3,200 My) to rim (~1,800 Ma old; Figure 7). Apatite AG09016-62 displays two ages consistent with the oldest domains of apatite AG09016-15 (i.e., 3,200 and 2,400 Ma). The single data point for AG09009 (08) corresponds to an age of 1,867 ± 143 Ma ( 207 Pb/ 206 Pb age; 2 SE) and comply well with the major trend defined by matrix apatites, much as inclusion AG09016-11 (Figure 7). It is important to note that inclusion AG09009-08 is surrounded by chlorite and chlorite-filled cracks, suggesting later fluid circulation within the zircon crystal ( Figure 6). Similarly, apatite AG09016-11 is traversed by a major crack that runs through the zircon.

Preservation of Magmatic Chemical Signature
In all studied TTG samples, apatite exhibits at least two chemical groups of which Group 1 has the same characteristics across the entire range of samples. Group 1 apatites are characterized by chondrite-normalized REE patterns depleted in LREE compare to HREE and a negative Eu anomaly (Figure 3). In a Nd versus Eu/Eu* diagram, Group 1 apatites (both matrix and inclusions) have systematically lower Eu/Eu* values and higher Nd content than that of other apatite groups ( Figure 8). Texturally, this group corresponds to core composition when a core to rim systematic is present (Figures 4b and 4c). Moreover, all zircon-hosted inclusions have Group 1 chemical characteristics. Finally, recent data on apatite from adakite (TTG-like signature; Bruand et al., 2020 and TTGs from two other cratons (Kaapvaal, South Africa and Karelia, Finland) show comparable chemistry to Group 1 apatite. All these results indicate that Group 1 apatite composition represents primary (magmatic) apatite REE signature.

Matrix Apatite
Our new U-Th-Pb ages on matrix apatite indicate that the vast majority of analyzed crystals exhibit ages that are consistent with the latest pervasive metamorphic event recorded by Acasta gneisses, that is, the 1,800-1,700 Ma Wopmay orogen ( Figure 5). This is in line with apatite ages obtained by Sano et al. (1999) and Fisher et al. (2019) and indicate that metamorphic conditions associated with the Wopmay orogen exceeded the closure temperature of apatite (>350-600°C e.g., Cherniak et al., 1991;Chamberlain & Bowring, 2001;Cochrane et al., 2014). We can, therefore, assume that any previously recorded metamorphic events had been overprinted by that related to the Wopmay orogen (Table S1). Rare concordant matrix apatite ages at 3,197 ± 82 Ma and 2,721 ± 63 Ma can point to possible local preservation of older ages, in line with titanite geochronology (Davidek et al., 1997;Fisher et al., 2019; Figure 5), but overall magmatic ages are not preserved in matrix apatites despite the local preservation of primary REE signature in the same apatites (Group 1). The dichotomy between the primary chemical signature of Group 1 matrix apatite and the secondary record of its U-Th-Pb data is consistent with diffusion work done on REE by Cherniak (2000) who showed that REE in apatite can diffuse relatively slowly even in the case of high 10.1029/2020GC008923 Geochemistry, Geophysics, Geosystems Figure 6. Cathodoluminescence (CL) and back-scattered-electron (BSE) images for zircons containing apatite inclusions analyzed in this study. Also reported are locations and labels of U-Th-Pb isotope measurements together with zircon ages, when available (Guitreau et al., 2012(Guitreau et al., , 2018.

10.1029/2020GC008923
Geochemistry, Geophysics, Geosystems temperature metamorphic events. This study demonstrates that two different types of substitution are responsible for diffusion in apatite; simple substitution (REE 3+ ↔ REE 3+ ) and coupled substitutions (i.e. REE 3 + + Si 4+ ↔ Ca 2+ + P 5+ ). In diffusion experiments, Cherniak (2000) showed that Sr, Pb, and O are more readily reset in apatite but can keep their primary signature in the core of the crystal (10 μm zoning) for metamorphic episodes below 750°C and for a duration of 1.05·10 4 years. On the other hand, apatite grains (100 μm in size) keep their primary REE signature for metamorphic episodes of less than 10 5 years in simple substitutions and 10 7 years in coupled substitutions for the same temperature.

Apatite Inclusions
Zircon-hosted apatite inclusions within AG09016, all exhibit primary REE signatures and, contrary to matrix apatites, essentially distribute along the Concordia curve with ages ranging between~1,700 and 4,000 Ma ( Figure 7; Table S6). This demonstrates that apatite inclusions have not been completely overprinted by the Wopmay orogen but, on the contrary, kept the memory of older metamorphic events and even original igneous U-Pb ages in some cases. This is precisely the case for apatite AG09016-88 which displays a concordant age of 4,003 ± 89 Ma ( 207 Pb/ 206 Pb age, 0.7% discordant) that is very consistent with that obtained on the host zircon (3,967 ± 49 Ma; Guitreau et al., 2018). The large range of ages exhibited by apatite inclusions is consistent with the fact that the AGC recorded several tectono-thermal events from 3,750 to 1,700 Ma, although it is difficult to associate all ages to specific events (Table 2). Yet three data points correspond to an age of~3,200 Ma consistent with that of titanite and garnet within Acasta samples (Fisher et al., 2019;Maneiro, 2016). We interpret the general distribution to represent punctual thermally activated resetting during metamorphic events (e.g., Cochrane et al., 2014). Three other data points are consistent with the matrix apatite trend, which suggests that the U-Pb system in these crystals were reset in a similar way as matrix apatite, despite encapsulation within zircon. Therefore, apatite inclusions have been either reset within zircon without any influence from the external media or in connection with this latter via cracks and porous zircon domains.
Reversely discordant data can be seen as either analytical artifacts or evidence for Pb contamination, which would significantly lower the measured U/Pb ratio. An analytical artifact would mean odd fractionation between U and Pb, which is inconsistent with the relatively good agreement between U, Th, and Pb concentrations determined during trace element analyses and U-Th-Pb isotope measurements. In addition, such extreme displacement from the Concordia curve would mean an instrumental fractionation much too large compared to the rest of our U/Pb data. Therefore, we think that the U-Pb isotope signatures of these inclusions are real and come from Pb contamination of apatite inclusions. Common Pb cannot be responsible for this contamination because the 208 Pb/ 206 Pb of reversely discordant apatite inclusions is too low. Figure 9 shows that matrix apatite Pb isotope composition is controlled to a large extent by common Pb contamination because data points form a positive correlation toward the common Pb end-member. In contrast, apatite inclusions have a horizontal distribution and, hence, no visible influence from common Pb, except for a few apatites that exhibited U-Pb data distribution consistent with the matrix apatite trend in Figure 7. Therefore, we conclude that common Pb contamination cannot account for the reversely discordant apatite data. Instead, we propose that reversely discordant apatite inclusions were contaminated by radiogenic Pb coming from the host zircon during a thermal event such as the Wopmay orogen. Radiogenic Pb migration has already been described in Archean zircons, notably from the Napier Complex, Antarctica (  . Tera-Wasserburg diagram for apatite inclusions in AG09009 and AG09016 corrected for common Pb. Also reported are matrix apatite data with (black dashed circles) and without common Pb correction (gray dashed circles). Numbers next to colored ellipses correspond to analysis numbers as reported in Table S6 and in Figure 7.

10.1029/2020GC008923
Pb isotope composition of a zircon with a Th/U of 0.5 is presented in Figure 9, and it is clearly visible that contamination by such radiogenic Pb would result in horizontal spread of data in a 208 Pb/ 206 Pb versus 207 Pb/ 206 Pb plot.
The zircons that host apatite inclusions analyzed in this study are often not pristine-looking ( Figure 6). Instead, they show advanced signs of radiation damage, a common feature in Archean zircon, have variably discordant ages, and some of them even display metamorphic ages (see BSE images in Figure 6). Because zircons underwent disturbances in their U-Pb isotope systematics by thermal events (Guitreau et al., 2018), one can easily expect that apatite inclusions might also be affected. This would be consistent with our geochronological results. An important conclusion from our apatite dating is that compared to matrix apatite, no common Pb seems to influence the data distribution in AG09016 (Table S6 and Figure 9). This means that host zircons satisfactorily isolated apatite inclusions from external fluids possibly bringing common Pb, which accounts for both the resetting of U-Pb ages, the existence of reversely discordant U-Pb data and the fact that Sr isotope data for apatite inclusions from Emo et al. (2018) are less radiogenic than those in the matrix despite zircons showing advanced radiation damage in CL images. If external fluids were to penetrate within zircon, apatite Sr isotope composition should be influenced by the fluids, which are more radiogenic. In summary, our results highlight that U-Th-Pb ages can be compromised in apatite inclusions hosted in zircons with advanced radiation damage but this is not the case for REE concentrations. Pb plot for apatite data from this study. Also reported are the evolution of the terrestrial lead, as given in Stacey and Kramers (1975), and the Pb isotope evolution of a 4 Ga zircon that has a present-day Th/U of 0.5. Matrix and zircon-hosted apatites exhibit a clear dichotomy with the former defining a positive trend toward common Pb whereas the latter distribute horizontally, except for a few data points that have U-Pb ages consistent with those of matrix apatite (see Figure 7). Note that apatite 207 Pb/ 206 Pb ratios vary between values that correspond to the age of the Wopmay orogen and that of the oldest Acasta gneisses, except for three data points that we surmise incorporated zircon-derived radiogenic Pb during metamorphism related to Wopmay orogen (see text for details).

Secondary Apatite Signatures
The secondary groups of apatites are chemically distinct from the primary group described above. All of them show a strong decrease in REE compared to Group 1 apatites (or magmatic apatites), with HREE depleted to a lesser extent than LREE. All apatites from Groups 2 and more (except AG09014), display no or more rarely a positive Eu anomaly (Figure 3d). These groups texturally appear as new or recrystallized grains or are rimming Group 1 cores. Their growth are related to secondary processes such as metamorphism and/or metasomatism which in the light of current literature on those phases are sometimes difficult to distinguish (Bau, 1991;Henrichs et al., 2018).
In the studied samples, we identified different late apatite generations: i In one sample (AG09009), a core (Group 1) to rim (Group 2) systematic has been identified ( Figure 4b).
The depletion in LREE and the absence of Eu anomaly characterizing the rim is interpreted as metamorphic ( Figure 3a). This type of overgrowth is also observed in metamorphosed post-Archean apatites and has similar low REE + Y, U + Th rims around igneous apatite cores (Henrichs et al., 2018). ii For the three other samples, second generation of apatite crystals (Figure 3, Groups 2 to 5 depending on samples have been observed. In samples AG09015 and AG09016, these groups are interpreted as metamorphic and/or metasomatic. On the other hand, Group 2 from sample AG09014 is the only second generation displaying a strong negative Eu anomaly and rather magmatic characteristics. This could be related to a late magmatic interaction with a different magma or the consequence of the crystallization of another LREE-mineral during the crystallization of the granitoid (Hoskin et al., 2000;Bruand et al., 2014; Table 2). Here it probably points to a lower grade of metamorphism (than for AG09015) because of its higher content in Cl and lower content in Y (Table S5). iii Apatites derived from primary monazite. These secondary apatites have only been identified in sample AG09009 (Figure 2b). In this case, primary monazite is destabilized in a metamorphic corona of apatite and fibrous allanite. This type of secondary textures has been described occasionally in natural samples (Gasser et al., 2012 in metapelite under amphibolite facies; Finger et al., 1998 metamorphosed I and S-type granites under amphibolite facies) and in experimental studies (Budzyń et al., 2017) and is characteristic of amphibolite facies metamorphic conditions.
All secondary apatite compositions have U-Th-Pb ages corresponding to the Wopmay orogen 1,800-1,700 Ma and have nucleated or were reset during this event ( Figure 5). Again, this is consistent with young ages obtained by Sano et al. (1999) and Fisher et al. (2019).

Apatite Chemistry and Identification of Host Granitoid
Archean TTGs are known for their specific geochemical signature, notably their high (La/Yb) N and low HREE concentrations (Figure 10c; e.g., Moyen & Martin, 2012). They can be easily discriminated from post-Archean granitoids in (La/Yb) N versus Yb N diagrams. In this study, we found that the same diagram discriminates very well TTG apatite from post-Archean granitoid apatite (Figure 10a) alike in Bruand et al. (2020). However, the chemical characteristics define an opposite trend when compared to the whole rock data. Indeed, apatites from TTGs have systematically low (La/Yb) N ratios, high HREE contents and notably high Y concentrations, whereas apatites from post-Archean granitoids commonly show the opposite. Interestingly, apatite chemistry from peraluminous post-Archean granitoids are marginally overlapping with the apatite from TTG (Figures 10a-10d). Their (La/Yb) N are nonetheless systematically higher than apatite TTG from Archean cratons (Figure 10d). Yet a very important conclusion of our study is that we observed distinct generations of apatite that we attribute to magmatic, late-stage, metamorphic, and/or metasomatic processes. The late events affecting the AGC have an impact on the LREE chemistry of apatite whereas the HREE signature is similar. The consequence of this, is that the (La/Yb) N ratio of secondary apatite in TTG will be even lower than the magmatic one ( Figure 10b). However, no matter which generation analyzed apatites belong to, they all show the peculiar chemistry of TTG apatite in a (La/Yb) N versus Yb N diagram, hence, allowing identification of their source rock ( Figure 10). This signature is not specific to Acasta samples as shown by Bruand et al. (2020). Apatites analyzed in TTG from Kaapvaal and Karelia cratons (South Africa and Finland, respectively) have the same chemical characteristics.

Geochemistry, Geophysics, Geosystems
Sr/Y versus Y diagram is another well-known discriminating diagram for TTG whole rock. However, this diagram only moderately discriminates the TTG from other granitoids when using apatite chemistry. Indeed, Sr and Y contents in apatite have been shown to be strongly influenced by SiO 2 , Sr, and the ASI of the melt (Bea & Montero, 1999;Sha & Chappell, 1999;Belousova et al., 2001;Jennings et al., 2011;Bruand et al., 2014). Consequently, in an Sr/Y versus Y diagram, apatite from peraluminous and TTG granitoid are undistinguishably defining the same field. REE partitioning in apatite are intimitaly linked to the host rock ASI, substitutions and co-crystallizing phases. To explain the dichotomy between the LREE depleted and enriched HREE characters of TTG apatite and the LREE-enriched and HREE-depleted characters of TTG (whole rock; Figures 10a-10c), we need to invoke the presence of other LREE-bearing phases. Recently, Bruand et al. (2020) have highlighted that the presence of monazite is not rare in TTG granitoids. In the studied samples, only few occurrences of monazite interpreted as primary have been identified. The occurrences of monazite in one of the samples suggest that the presence of monazite could be partially responsible for the specific chemistry of TTG apatite. In post-Archean granitoids cocrystallizing monazite would mean a reduced host magma with ASI > 1 and relatively low Ca content (Hsieh et al., 2008). Trace elements allow making the distinction between mafic I-type granitoids and I/S-type granitoids (Sha & Chappell, 1999), but the different types of granites can only be discriminated by using Sr and Th. The concentrations in both elements are rather different from the ones from post-Archean apatite, which shows that the composition of apatite in TTG are most likely due to different melt composition and probably fO 2 during crystallization of the apatite in the Archean granitoid. However, the competition between REE-bearing phases in TTG are not well known and the absence of partition coefficient for those REE-bearing phases in such granitoid preclude further interpretation of the data but highlight that more work is needed on this subject.

Implications
The results of the present study demonstrate that matrix apatites in polymetamorphosed Acasta gneisses can preserve primary REE signatures, whereas U-Th-Pb isotope systematics have been fully reset. On the contrary, zircon-hosted apatite inclusions can both preserve primary REE signatures and U-Th-Pb isotope  Figure 10. (a) (La/Yb) N versus Yb N discriminating diagram for apatites from Archean granitoids (TTG) and post-Archean granitoids. Literature data from (Belousova et al., 2001;Bruand et al., 2014Bruand et al., , 2020Chu et al., 2009;Dempster et al., 2003;Hoskin et al., 2000;Jennings et al., 2011). (b) Details on the distinction between magmatic and metamorphic apatites from this study. (c) (La/Yb) N versus Yb N discriminating diagram between TTG and post-Archean rocks modified from Hansen et al. (2002). (d) Log (La/Yb) N versus Yb N discriminating diagram between TTG and post-Archean granitoids.

10.1029/2020GC008923
Geochemistry, Geophysics, Geosystems systematics provided that the host zircon experienced limited radiation damage and remained isolated from the external medium. Interestingly, apatite can be used as a discriminating tool for different granitoid types (This study ;Jennings et al., 2011;Bruand et al., 2014Bruand et al., , 2016Bruand et al., , 2020. Therefore, our results strongly suggest that the geochemistry and geochronology of detrital apatite and/or detrital zircon-hosted apatite inclusions can be used as a new tool to unravel the origin, nature, and evolution of complex multistage Archean cratons as well as the type of granitoid Hadean detrital zircons crystallized from. However, one needs to be aware of the limitations of this approach that essentially pertain to metamorphic resetting of isotope systematics. This limits the use of detrital apatite but also suggests that when one focuses on zircon-hosted inclusions, the host zircon should be well characterized for, at least, internal textures and U-Pb ages so as to understand the pristineness of the information encapsulated in the apatite inclusion. This is because all inclusions within zircon are not systematically primary but could grow within old radiation-damaged detrital zircons (e.g., Rasmussen et al., 2011). Consequently, analyzed inclusions should be monomineralic to avoid perturbation from later events and should be contained within a zircon with limited radiation damage.

Conclusions
The aim of this study was to decipher whether or not, matrix apatites and apatite inclusions in zircons from Archean TTGs from the AGC (Canada) preserved primary compositions. Here, we were able to demonstrate that primary REE composition of apatite in the matrix can be preserved for large crystals (in this study mostly over 40 μm), whereas their U-Th-Pb isotope systematics were fully reset by HT metamorphism around 1,800-1,700 Ma (Wopmay Orogen). Our study also shows that apatites from TTGs have geochemical signatures (La/Yb) N vs. Yb N ) that are very distinct from those of apatite from other granitoid type, regardless of their primary or secondary origin. In addition, our results show that Archean apatites trapped within zircon that have withstood multiple HT metamorphic events were effectively protected and kept their primary isotopic and geochemical signature, provided that zircon lattice underwent moderate radiation damage. Altogether, these new results have important implications for studies of Archean and Hadean crustal evolution because this tool can be applied to detrital apatites, knowing the limitations of such an approach, but more importantly to detrital zircon-hosted apatites. In the latter case, it is important to target the most pristine-looking zircon crystals, to characterize them for, at least, U-Pb ages, and look for monomineralic apatite inclusions in priority.