Evidence for the Multi-Stage Petrogenetic History of the Oka Carbonatite Complex ( Québec , Canada ) as Recorded by Perovskite and Apatite

The Oka complex is amongst the youngest carbonatite occurrences in North America and is associated with the Monteregian Igneous Province (MIP; Québec, Canada). The complex consists of both carbonatite and undersaturated silicate rocks (e.g., ijolite, alnöite), and their relative emplacement history is uncertain. The aim of this study is to decipher the petrogenetic history of Oka via the compositional, isotopic and geochronological investigation of accessory minerals, perovskite and apatite, using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS). The new compositional data for individual perovskite and apatite grains from both carbonatite and associated alkaline silicate rocks are highly variable and indicative of open system behavior. In situ Sr and Nd isotopic compositions for these two minerals are also variable and support the involvement of several mantle sources. U-Pb ages for both perovskite and apatite define a bimodal distribution, and range between 113 and 135 Ma, which overlaps the range of ages reported previously for Oka and the entire MIP. The overall distribution of ages indicates that alnöite was intruded first, followed by okaite and carbonatite, whereas ijolite defines a bimodal emplacement history. The combined chemical, isotopic, and geochronological data is best explained by invoking the periodic generation of small volume, partial melts generated from heterogeneous mantle.


Introduction
Previous studies indicate that carbonatites worldwide range in age from the Archean to present day, with the frequency of occurrences increasing with decreasing age (i.e., with those <200 Ma in age being the most abundant) [1].The oldest known carbonatite on Earth is the Tupertalik complex (3.0 Ga; western Greenland) [2], and the youngest is the active natrocarbonatite volcano, Oldoinyo Lengai, Tanzania [3][4][5][6][7].Among the 527 carbonatite occurrences identified and compiled by Woolley and Kjarsgaard [1], only 264 have been dated, with most ages determined by the K/Ar method and merely 6% investigated by U-Pb geochronology.
In North America, carbonatite and alkaline magmatism spans ~2.7 Ga [8,9], and Oka (Figure 1) is one of the youngest carbonatite complexes on the basis of available geochronological data for various minerals and/or rock types.Apatite fission track ages reported for Oka vary between 118 ± 4 and 133 ± 11 Ma [10], whereas Shafiquall et al. [11] document K-Ar ages that range between 107 and 119 Ma for the intrusive alnöites associated with the complex.Wen et al. [12] reported a Rb-Sr biotite-whole rock isochron age of 109 ± 2 Ma obtained by isotope dilution-thermal ionization mass spectrometry (ID-TIMS), whereas Cox and Wilton [13] obtained a U-Pb age of 131 ± 7 Ma for perovskite from carbonatite by laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS).Cox and Wilton [13] postulated that the variable ages obtained by either Rb-Sr or K-Ar methods for the Oka carbonatite complex most probably result from their lower closure temperatures (relative to the U-Pb isotope system).
In a recent study, Chen and Simonetti [14] conducted a detailed U-Pb geochronological investigation of apatite from carbonatite, okaite and melanite ijolite at Oka.The geochronological results define a bimodal age distribution with peaks at ~117 and ~125 Ma.Chen et al. [15] also reported in-situ U-Pb ages for niocalite [Ca 14 Nb 2 (Si 2 O 7 ) 2 O 6 F 2 ] from carbonatite at Oka, a first documented radiometric study for this mineral.The results from the latter study corroborate the variable apatite U-Pb ages [14] since they also define a similar bimodal distribution with peak 206 Pb/ 238 U weighted mean ages at 133.2 ± 6.1 and 110.1 ± 5.0 Ma.Hence, these detailed investigations reporting in-situ U-Pb ages for apatite and niocalite from Oka offer a more comprehensive geochronological view of this complex, and clearly suggest a protracted magmatic history that spanned at least ~10 million years [14].Of note, the protracted igneous activity outlined for Oka overlaps that for the entire range of ages reported for the remaining Monteregian Igneous Province (MIP)-related intrusions (118-135 Ma; Figure 1) [13][14][15][16].
Plutonic igneous bodies, such as the Oka carbonatite complex, may form as a result of sequential and episodic melting and crystallization events.These events may be traced by monitoring the chemical compositions of constituent minerals, whereas the timing of such events can be determined using precise geochronological methods (e.g., U-Pb dating).Rukhlov and Bell [9] emphasized the importance of incorporating data from several isotope systems (and mineral/rock phases) before concluding the definite emplacement ages for carbonatite complexes since these typically result from complicated petrogenetic histories.Moreover, accessory phases such as perovskite, apatite and niocalite are characterized by very high abundances (1000s of ppm) of incompatible elements such as Nd and Sr, which are important isotope tracers for delineating magmatic processes and potential mantle sources.Thus, the combined presence of several U-(and Pb-) bearing accessory minerals, such as apatite, niocalite and perovskite within carbonatite, together with the capability of obtaining precise and accurate, spatially resolved chemical, isotopic, and geochronological data by laser ablation inductively (multi-collector) coupled plasma mass spectrometry (LA-(MC)-ICP-MS) analysis on individual (single) mineral grains renders the investigation of these accessory minerals a powerful tool in deciphering the formational history of carbonatite complexes.
Perovskite, CaTiO 3 , is a widespread accessory mineral in SiO 2 -undersaturated and alkaline magmatic systems, and present within kimberlites, melilitites, foidites, olivinites, clinopyroxenites, ultramafic lamprophyres/carbonatites, and lamproites [18].It typically forms throughout the melt crystallization sequence and serves as a major host for incompatible trace elements, such as the Rare Earth Elements (REEs) [18,19].Furthermore, perovskite may preserve the original, magmatic radiogenic isotope signatures that are frequently obscured in whole rock compositions because it is relatively resistant to post-solidification alteration [19].Recently, perovskite has been the focus of petrogenetic studies involving alkaline magmatic systems because of its capacity to yield combined accurate geochronological, chemical and/or radiogenic isotope information [13,[19][20][21][22]. Cox and Wilton [13] were the first to report an in-situ U-Pb geochronological investigation of perovskite from Oka by LA-ICP-MS; however, their study was not combined with any geochemical data, and hence does not provide any insights into possible melt differentiation processes.For example, the exact petrogenetic relationship between carbonatite and associated alkaline silicate rocks within an individual carbonatite complex remains somewhat elusive; models proposed include liquid immiscibility [23][24][25][26], protracted fractional crystallization of a carbonate-rich, Si-undersaturated parental melt [27][28][29], and small volume partial melts derived from metasomatized mantle [30][31][32].
This study focuses on a detailed chemical, isotopic and geochronological investigation of perovskite and apatite associated with alnöite and jacupirangite from the Oka carbonatite complex.We report new, in-situ major and trace element chemical compositions, Sr and Nd isotopic data, and U-Pb ages for perovskite and apatite.Hence, this study provides additional insights into the formational history of the Oka carbonatite complex, and compliments the earlier in-situ U-Pb investigations of apatite [14] and niocalite [15].

Geological Setting and Description of Samples
The series of alkaline intrusions associated with the MIP define a linear E-W trend that roughly follows the Ottawa-Bonnechere Paleo-rift (Figure 1A) [33].The Oka carbonatite complex, which is the most westerly intrusion, is located entirely within the Grenville Province and does not contain any quartz-bearing rocks (Figure 1B).Moving in a southeastward direction, the remaining complexes have intruded two different tectonic/structural terrains.Mounts Royal, St. Bruno, St. Hilaire, Rougemont, and Johnson are hosted by St. Lawrence Lowlands Cambrian-Ordovician dolostones, carbonates and shales; Mounts Brome and Shefford intrude the metasediments and metamorphic rocks of the Appalachians (Figure 1A) [33,34].
Gold et al. [10] provided a detailed description of the Oka carbonatite complex (Figure 1B).In summary, the complex consists of both carbonatite (Figure 2A) and Si-undersaturated rocks (i.e., okaite, ijolite, alnöite, and jacupirangite; Figure 2B-E).The mineralogical descriptions of the samples investigated in this study are listed in Table 1, and all samples were retrieved from the Stop 2.3 locality with the exception of Oka88 (Stop 2.2; Figure 1B).Apatite is a common accessory mineral phase occurring in all rock types, and forms euhedral crystals, which vary between ~1 and ~100 μm in diameter (e.g., Figure 2A,E).Perovskite is present in most of the silicate rocks (i.e., okaite and alnöite) and usually occurs as euhedral crystals characterized by pseudocubo-octahedral habit up to several cm in diameter (i.e., Figure 2F,G).In two okaite samples (Oka229 and Oka137), perovskite exhibits zoning as evidenced in optical microscopy and back-scattered electron (BSE) imagery (e.g., Figure 2G,H).The zoning typically consists of a low average atomic number core and a high average atomic number rim of irregular thickness (Figure 2H).

Chemical Analysis
Major and minor element concentrations for apatite and perovskite were determined using a Cameca SX50 electron microprobe (EMP; Cameca, Gennevilliers, France) at the University of Chicago.All thin sections were carbon-coated prior to analysis.The EMP analyses were conducted using a 15 kV accelerating potential and 30-35 nA incident current.The natural and synthetic mineral and glass standards employed for calibration purposes were: natural olivine (for Fe, Mn, Mg, and Si), natural albite (for Na), durango apatite (for P, F, and Ca), synthetic glass of anorthite composition (for Al), strontianite (for Sr), zircon (for Zr), synthetic NB metal (for Nb), synthetic TiO 2 (for Ti), synthetic REE 3 metal (for La, Ce, and Pr), synthetic REE 2 metal (for Nd and Sm), and synthetic TA metal (for Ta).Calculated apatite and perovskite formulae were normalized by stoichiometry.
In-situ trace element analyses of individual apatite and perovskite grains were obtained using a UP213 nm laser ablation system coupled to a Thermo-Finnigan Element2 sector field high-resolution ICP-MS (Thermo Fisher Scientific, Dreieich, Germany) housed within the Midwest Isotope and Trace Element Research Analytical Center (MITERAC) at the University of Notre Dame, and following the protocol by Chen and Simonetti [14] and Chen et al. [15].The NIST SRM 610 international glass standard [35] was used for external calibration and 43 Ca ion signal intensities were employed as the internal standard with the CaO content (wt %) obtained by EMP analysis.The sample grains and standard were ablated using a 25 μm spot size, 4-5 Hz repetition rate, and corresponding energy density of ~10-12 J/cm 2 .Data reduction, including concentration determinations, method detection limits and individual run uncertainties were obtained with the GLITTER laser ablation software [36].

U-Pb Age Dating by Laser Ablation (Multi-Collector) Inductively Coupled Plasma Mass Spectrometry
The instrumental configuration described above for the trace element determinations was also employed for the in-situ U-Pb isotope analyses.The analytical protocol adopted here is similar to that described in Simonetti and Neal [37] and Chen and Simonetti [14].Data acquisition typically consisted of the first ~30 s for measurement of the background ion signals, followed by 30 s of ablation, and a minimum 15 s of washout time.Single mineral grains were ablated using a 40-55 μm spot size and corresponding fluence of ~3 J/cm 2 and repetition rate of 5 Hz.The following ion signals were acquired: 202 Hg, 204 (Pb + Hg), 206 Pb, 207 Pb, 208 Pb, 232 Th, 235 U, 238 U and 232 Th 16 O. 202Hg was measured to monitor the 204 Hg interference on 204 Pb (using a 204 Hg/ 202 Hg value of 0.229883) [36].In addition, U-Pb dates for some samples were determined using a NWR193 nm laser ablation system (ESI New Wave Research, Huntingdon, UK) coupled to a Nu Plasma II MC-ICP-MS instrument (Nu Instruments Ltd., Wrexham, UK) within the MITERAC facility at the University of Notre Dame.All masses of interest ( 202 Hg, 204 (Pb + Hg), 206 Pb, 207 Pb, and 208 Pb) can be simultaneously acquired using a combination of ion counters (Hg and Pb ion signals), and Faraday cups ( 232 Th and 238 U).Samples and standards were ablated employing a 55-75 μm spot size with corresponding fluence of ~10 J/cm 2 and 7 Hz repetition rate.Data acquisition consisted of the first ~40 s for background measurement, followed by ~60 s of ablation, and a minimum ~2 min of washout time.
Instrumental drift and laser induced elemental fractionation (LIEF) was monitored using a "standard-sample bracketing" technique.The Madagascar apatite [38] and Ice River perovskite [19] were adopted as the external standards for the U-Pb dating of apatite and perovskite, respectively.Each set of 10-12 unknown analyses was bracketed with five analyses of the pertinent standard both prior and after the unknown analyses.Instrumental drift and Pb-U laser induced fractionation were corrected based on the 206 Pb/ 238 U and 207 Pb/ 235 U ratios for the standards, i.e., for the Madagascar apatite, the ratios are 0.0781 and 0.6123 [38], respectively; whereas for the Ice River perovskite, the adopted values are 0.0575 and 0.4270, respectively [19].
Apatite and perovskite are U-bearing accessory minerals that may contain a significant amount of common Pb.The 207 Pb-correction method was adopted here, which employs the Tera-Wasserburg Concordia plot and consequently is an approach that does not require knowledge of the accurate abundance of 204 Pb.This method was successfully employed in previous studies for a variety of common Pb-bearing accessory minerals, such as titanite [20,[39][40][41], perovskite [13,19,20], apatite [14,41], and niocalite [15].The 207 Pb-correction method does require knowledge, however, of the Pb isotope composition of the common Pb component.In this study, the latter is defined by the Pb isotope composition of the associated and ubiquitously present U-free calcite [14].The 207 Pb/ 206 Pb ratio of 0.792 ± 0.06 [14,42] obtained for the latter is then applied to correct the measured 206 Pb/ 238 U ratios using well established common lead-radiogenic lead mixing equations [13,39].
Fragments of Emerald Lake and Durango apatites were used as secondary apatite standards and both are well characterized with ages of 90.5 ± 3.1 and 30.6 ± 2.3 Ma, respectively [43].Repeated analyses of these two standards obtained during the course of this study yielded weighted mean 206 Pb/ 238 U ages of 92.6 ± 1.6 Ma (n = 17) and 31.9 ± 1.3 Ma (n = 10), respectively, and both are identical (given their associated uncertainties) to the ages reported by Chew et al. (Figure 3A,B) [41].Of importance, U-Pb ages were obtained by both laser ablation multi-collector inductively coupled plasma mass spectrometry (LA-MC-ICP-MS) and LA-ICP-MS methods for the same apatite grains in one carbonatite sample (Oka147), and these yield identical dates (Figure 3C,D).These corroborative results in turn serve to further validate the analytical methods employed here.
Uncertainties associated with individual analyses, which include propagation of errors from individual measurements (based on counting statistics) and the relative standard deviation associated with repeated analyses of the Madagascar apatite and Ice River perovskite standards, were determined using the quadratic equation [38,44,45].Isoplot v3.0 program was employed for constructing Tera-Wasserburg diagrams and determination of Concordia lower intercept ages, and 206 Pb/ 238 U weighted mean age calculations [46].Figure 3. U-Pb isotopic ages for apatite secondary standards and sample.The U-Pb age of the secondary standard-Durango apatite is shown (A), and it is identical to the reported value (B) by Chew et al. [41].Apatite from carbonatite Oka147 was dated by the laser ablation multi-collector inductively coupled plasma mass spectrometry (LA-MC-ICP-MS) (C) and LA-ICP-MS (D).

Sr and Nd Analysis by LA-(MC)-ICP-MS
In-situ Sr and Nd isotope ratios for perovskite and apatite were determined with a NWR193 laser ablation system coupled to a Nu Plasma II MC-ICP-MS instrument (Nu Instruments Ltd., Wrexham, UK).In-situ Sr isotope measurements involve correction of critical spectral interferences that include Kr, Rb, and doubly charged REEs [47,48].These detailed corrections are adopted in this study and are identical to those reported in Chen et al. [15].A modern-day coral (Indian Ocean) served as an external, in-house standard, which is well characterized for its 87 Sr/ 86 Sr isotopic composition by ID-TIMS [49].The coral standard and perovskite grains were analyzed using a 75~100 μm spot size, 7 Hz repetition rate, and an energy density ~11 J/cm 2 .The average 87 Sr/ 86 Sr ratio obtained for the coral standard is 0.70915 ± 0.00003 based on five measurements, and is indistinguishable (given uncertainties) compared to the corresponding TIMS value of 0.70910 ± 0.00002 [49].
In order to obtain accurate measurement of Nd isotope ratios, it is important to identify and correct isobaric interferences and monitor for instrumental mass discrimination.The isobaric interferences for in-situ Nd isotope determinations are principally related to Sm, Ce and Ba, and the correction of the 144 Sm ion signal on 144 Nd is critical.The 146 Nd/ 144 Nd ratio is traditionally selected to correct for instrumental mass discrimination with the 146 Nd/ 144 Nd ref = 0.7219 [50].As described in Yang et al. [51], the mass bias for both Sm and Nd were set to be identical (βSm = βNd).The mass bias correction for Sm is based on the 149 Sm/ 144 Sm, and consequently the 144 Nd ion signal can be calculated according to the following equation: The Durango apatite was adopted as the external standard with the accepted 143 Nd/ 144 Nd = 0.512483 [52].The "standard-sample" bracketing method was used and both standard and samples were analyzed using a 75-100 μm spot size, 7 Hz repetition rate, and corresponding energy density of ~9 J/cm 2 .

Major and Trace Element Data
The major and trace element data for perovskite investigated in this study are listed in Tables 2 and 3.In total, >30 chemical analyses of perovskite from okaite (Oka229, Oka137, and Oka208), alnöite (Oka73), and jacupirangite (Oka70) were obtained.Based on their major element compositions, the data for perovskite plot into two groups, in particular relative to their Nb 2 O 5 wt % abundances (Figure 4A).One group (Nb-E: Nb-enriched) contains high Nb 2 O 5 contents (between 7.25 and 10.80 wt %), whereas the other (Nb-D: Nb-depleted) is characterized by lower Nb 2 O 5 abundances (between 1.56 and 4.92 wt %).Table 2 shows that perovskite from both alnöite and jacupirangite belongs to the Nb-D group, whereas compositions for those from okaite are variable.Of importance, both Nb-E and Nb-D types of perovskite are present within individual samples (i.e., Oka137 and Oka229) and even within singular zoned grains (e.g., Figure 5; in total two well-zoned grains have been identified).For example, the zoned perovskite grain shown in Figure 5 has a rim that consists of a Nb-E composition, whereas the central area is characterized by the Nb-D component.In general, Nb-E perovskite is characterized by high contents of Na, Al, Fe, Sr, Ta and REEs (Figure 4B,C; Tables 2 and 3).The composition for the Nb-E perovskite is not that of the ideal end-member, but can be described by involving components of lueshite (NaNbO 3 ), latrappite (CaNb 0.5 Fe 0.5 O 3 ), and (LREE)FeO 3 (light REEs), which form by elemental substitutions into the structure at both Ca and Ti sites [53,54].Molar percentages of different perovskite endmembers are also listed in Table 2.The most significant substituents in the Ca site are REEs and Na (Figure 4B), which comprise up to 8.70 mol % of the (LREE)FeO 3 component in some Nb-E perovskite (Table 2).Another example is the coupled substitution between Ti and Nb + Fe 3+ or Nb + Al, which accounts for up to 10.96 mol % of CaNb 0.5 Fe 0.5 O 3 (latrappite) or CaNb 0.5 Al 0.5 O 3 (Figure 4C; Table 2).
The total REE budget for perovskite is dominated by the light REEs (LREEs; i.e., La, Ce, Pr, Nd) abundances with (La/Yb) CN ratios that vary between 364 and 1652 (Table 3), and as illustrated by the pronounced, negatively-sloped chondrite-normalized REE patterns (Figure 6).Pb and Th abundances for perovskite both define negative correlations with Ca contents and suggest their substitution within the same Ca site (Figure 7A,B); in contrast, U abundances do not show any covariance with Ca contents (not shown).Of note, some elements are reported for both EMP and LA-ICP-MS analysis (e.g., LREEs).For example, the relative difference in the measured abundances of Pr obtained by these two methods is ~10%.Newly obtained chemical compositions for apatite from alnöite, ijolite and jacupirangite are listed in Tables 4 and 5. Figure 8 plots the major element and LREE compositions for apatite from all rock types investigated here, along with those from carbonatite and okaite (Chen and Simonetti [14]).The compositions of fluorapatite from alnöite and jacupirangite are chemically distinct relative to the remaining rock types at Oka (Figure 8), i.e., they contain a higher Ca content for a given P abundance (Figure 8A).As explained by Chen and Simonetti [14], REE abundances for apatite exhibit a positive correlation with Si contents due to their co-substitution within Ca and P structural sites.Once again, the same substitution scheme is evidenced here for all the apatites with the exception of those from alnöite and jacupirangite (Figure 8B).Chondrite normalized REE patterns for apatite investigated here are also negatively-sloped (Figure 9), but are more variable compared to those for perovskite (Figure 6).Of note, the chondrite normalized REE patterns for apatite from alnöite and jacupirangite exhibit less negative slopes among all rock types, with lower LREE abundances and comparable heavy REE (HREE) contents (Figure 9).(La/Yb) N ratios vary from 45 to 161 for apatite from alnöite and jacupirangite (Table 5), whereas ratios range between 106 and 695 for the remaining apatite [14]; these ratios for apatite are generally lower compared to those for perovskite (Table 3).

Geochronological Data
New, in-situ U-Pb ages for apatite are reported here from three alnöites, one ijolite, two okaites, and one jacupirangite (Table 6).As with the in-situ U-Pb dating results documented previously for apatite from Oka by Chen and Simonetti [14], the newly obtained ages for several alkaline silicate rock samples also indicate a bimodal distribution (e.g., Oka132 and Oka229; Figure 10B,C).For example, apatite from alnöite sample Oka75 yields bimodal 206 Pb/ 238 U weighted mean ages of 111.7 ± 3.4 and 131.6 ± 2.4 Ma.In general, samples with only one age peak (i.e., Oka209, Figure 10A) yield a relatively young age.
In total, ~40 U-Pb analyses for perovskite from four okaites and one alnöite obtained here are listed in Table 7.Of interest, U-Pb ages for perovskite from okaite sample Oka229 also yields a bimodal distribution (Figure 10D).Moreover, individual ages correlate with their corresponding chemical compositions, i.e., older perovskites that define a 2 °6Pb/ 238 U weighted mean age of 139.4 ± 2.5 Ma are characterized by high Nb 2 O 5 contents (Nb-E group), whereas younger perovskites with an age of 115.7 ± 3.9 Ma belong to the Nb-D group (Figure 5).Of note, the young ages for perovskite obtained in this study (Table 4 and Figure 10E) are younger than the previously reported (single) U-Pb age of 131 ± 7 Ma for the same mineral from carbonatite at Oka [13].Thus, as with the recently published apatite and niocalite ages for Oka [14,15], the U-Pb perovskite ages obtained here also suggest a rather protracted crystallization history for Oka.Of importance, the Th/U ratios for perovskite investigated in this study are all >1 with the highest value up to 31.Chew et al. [41] pointed out that using the 208 Pb-correction method in conjunction with determining 208 Pb-232 Th ages only yields reliable geochronological results when 232 Th/ 238 U ratios are <0.5.Consequently, we do not report the 232 Th-208 Pb ages for perovskite investigated here.All reported uncertainties are at 2σ level as determined by Isoplot [46].The Mean Square Weighted Deviation (MSWD) is used as a statistical validity of the regression line according to the criteria defined by Wendt and Carl [56].A recent geochronological study by Chen et al. [15] focused on the Nb-disilicate mineral, niocalite, for which Oka is the type locality.Niocalite from one of the carbonatite samples investigated by Chen et al. [15] also indicates a bimodal age distribution with weighted mean 206 Pb/ 238 U ages of 110.1 ± 5.0 and 133.2 ± 6.1 Ma, and overlaps that of co-existing apatite for the same sample [15].Niocalite from two other carbonatite samples yield younger ages of 110.6 ± 1.2 and 115.0 ± 1.9 Ma [15].
In summary, a total of 293 in-situ U-Pb apatite ages yield a bimodal distribution pattern using the Kernel Density Estimation (KDE) diagram (Figure 11A; KDE is a standard statistical technique used for estimating the density distribution in geochronlogical studies) [57], with two peaks at ~126 and ~115 Ma.The variable perovskite ages indicate an additional older age peak at 135.4 ± 3.2 Ma (Figure 11D), which is similar (given the associated uncertainties) to the age of 131 ± 7 Ma for perovskite obtained by Cox and Wilton [13].In contrast, the niocalite U-Pb dating results tend to converge toward the younger age, with a peak at 112.6 ± 1.2 Ma (Figure 11C) [15].The majority of the combined in situ U-Pb dating results for apatite, perovskite, and niocalite from Oka clearly support a protracted history of magmatic activity in the order of ~10-15 million years (Figure 11A).

Radiogenic Isotope Data
The Sr and Nd isotope results for perovskite and apatite obtained here are listed in Table 8 and shown in Figure 12.Overall, Rb concentrations are below (or close to) the detection limit, and consequently the calculated Rb/Sr ratios are extremely low so that the age correction of the measured 87 Sr/ 86 Sr ratio is negligible.For the Sm-Nd data, a correction for radiogenic 143 Nd was applied, and ages used for the correction were based on the U/Pb dating results obtained here.The in-situ Sr and Nd isotope data for both perovskite and apatite overlap the entire range defined by previously reported whole rock data for carbonatite from Oka [12], but definitely indicate a larger variation ( 87 Sr/ 86 Sr: 0.70312-0.70367; 143Nd/ 144 Nd: 0.51270-0.51286),and is not consistent with closed-system behavior (Figure 12A).Of interest, the Nd and Sr isotope data from Oka overlap the upper end of the East African Carbonatite Line (EACL; Figure 12B) [58].The EACL is defined by the Nd-Sr isotope values for young (<40 Ma old) East African carbonatites, and was interpreted to represent mixing between two end-member mantle components: HIMU (mantle component with time integrated, high 238 U/ 204 Pb ratio)-and EMI (enriched mantle 1)-like.okaite between 120 and 127 Ma; and (3) Lastly, at ~114 Ma, emplacement of the vast majority of the carbonatite, along with okaite, ijolite, and a minor amount of alnöite occurred.
Oka is not the sole alkaline complex that is characterized by an extended formational history spanning millions of years.Several previous studies of carbonatite and kimberlite alkaline complexes also define a protracted history of magmatic activity up to 40 million years [36,[59][60][61][62][63][64][65][66][67].Of note, based on U-Pb ages for ~30 kimberlite complexes in North America, Heaman and Kjarsgaard [59] stated that discrete kimberlite emplacement events within individual fields can occur over time intervals of up to 20 Myrs.For example, the majority of the kimberlite complexes located within the region of Timiskaming, which is located ~1000 km northwest from the MIP, were emplaced between 155 and 134 Ma (i.e., over ~20 Myrs period).
The protracted emplacement history (and ensuing melt differentiation) that occurred at Oka may be explained by invoking either one of two models: (1) Melt generation occurred at ~135 Ma, followed by magma differentiation in a closed-system over a period of ~10-15 million years; or (2) There was periodic generation of small volume, partial melts from a metasomatized, CO 2 -bearing mantle source over a period ~10-15 million years, with each melt fraction undergoing an independent crystallization/differentiation path.Given the extremely large variations in trace element abundances recorded by apatite (Table 5) [14], and those depicted by perovskite investigated here (Figures 4-7; Tables 2 and 3), it is difficult if not impossible to attribute these variations to closed-system melt differentiation involving solely one parental melt, regardless of whether this melt was carbonatitic, or a carbonate-rich, alkaline, silica-undersaturated parental melt [14].Chen and Simonetti [14] advocated for open-system behavior, possibly involving magma mixing, which is an interpretation also put forward by Zurevinski and Mitchell [68] to explain the chemical variations documented by pyrochlore from Oka.In this study, Figures 8 and 9 (and Tables 4 and 5) clearly indicate that the chemical compositions for apatite from alnöite and jacupirangite are distinct relative to those from other rock types.Their major and trace element contents and REE chondrite normalized patterns suggest derivation from a different mantle source.Of interest, Nb-E perovskites are only present in okaite and some represent the rim of zoned perovskite grains (Figure 5).The latter texture has been described as reverse zoning (i.e., an increase of REE and Th contents from core to rim) [53], which is uncommon and possibly results from re-equilibration of perovskite with magma modified by assimilation or contamination processes, or later surrounded by a melt of different composition [53].Thus, based on the combined chemical and geochronological data obtained for all rock types at Oka, we believe that the second hypothesis involving periodic generation of small volume melts and subsequent magma mixing best explains the petrogenetic history of the complex.

Relationship between Oka, Monteregian Igneous Province (MIP)-Related Intrusions, and Mantle Plumes?
There exist two competing hypotheses for the formation of Oka and the associated MIP-related intrusions in southeastern Québec (Figure 1A).One model proposes that they formed as the result of intraplate melting in an extensional setting associated with opening of the Atlantic Ocean [69,70].The alternative view is that the MIP results from the passage of the North American plate over the Great Meteor hotspot [59,[71][72][73][74].The main criticism with the latter is the lack of a precise correlation between the radiometric ages of the MIP-related intrusions and lithospheric plate migration (i.e., geographic position).However, the majority of the geochronological data for the MIP-related intrusions were obtained either by apatite fission-track or K-Ar methods, and only a small number of analyses were conducted for each intrusion.Thus, a more thorough and robust geochronological evaluation is required for each of the MIP intrusions before the plume hypothesis is completely ruled out.Moreover, the results from this study and Chen and Simonetti [14] both report ages for Oka that overlap the entire MIP age range, which further complicate matters in relation to evaluating a temporal relationship for the MIP intrusions relative to a plume hypothesis.
Carbonatites can provide valuable information for deciphering the geochemical nature of the upper mantle as their isotopic ratios inherited from their source region are buffered against crustal contamination due to their extremely high concentrations of incompatible elements (e.g., Sr and Nd).For example, in their study of the carbonatites and associated Si-undersaturated rocks from the Chilwa Island carbonatite complex, Simonetti and Bell [75] clearly indicate that an unrealistic amount of crustal assimilation is needed in order to explain the variable Nd and Sr isotope ratios.Hence, they advocated for melt derivation from a chemically and isotopically heterogeneous (metasomatized) mantle source region.In this study, the Nd and Sr isotope data for both apatite and perovskite overlap those previously obtained for whole rock samples from Oka (Figure 12) [12], but the former are clearly much more variable.This simply reflects the fact that whole rock analyses represent a weighted average of the Sr and Nd isotope composition of the (Sr-and Nd-bearing) constituent minerals (e.g., apatite, calcite, perovskite, and niocalite), and mask subtle differences between phases; however, the latter provide important details for deciphering the petrogenetic history of a complex.Evidence for "open-system" behavior at Oka was already evident from the TIMS generated whole rock data as these define a range of Sr and Nd isotope values that are well outside the typical in-run analytical precision (Figure 12).In Figure 12B, the Nd and Sr isotope compositions for apatite and perovskite from Oka are compared to those for well-established mantle components (i.e., HIMU, EMI, EMII (enriched mantle 2), and DMM (depleted mid-ocean ridge basaltic mantle)) [76] and East African carbonatite complexes [58].The Nd and Sr isotope data from Oka plot proximal to the field for the HIMU mantle component and most lie along a HIMU-EMI mixing array.Both HIMU and EMI are prevalent mantle components that underlie most of East Africa and also characterize the isotope compositions of ocean island basalts (OIBs) worldwide.Several previous investigations have advocated for the involvement of HIMU, EMI, and FOZO (Focus Zone) mantle components in the generation of most young (<200 Ma) carbonatites on a global scale [77][78][79][80][81]. On the basis of a compilation of both radiogenic and stable isotopic data from carbonatites worldwide, Bell and Simonetti [82] made the argument that parental carbonatitic magmas are derived from a sub-lithospheric source that is associated with either asthenospheric "upwellings" or more deep-seated, plume-related activity.Amongst the important evidences that support the generation of carbonated melts from sub-lithospheric mantle are: the petrogenetic and temporal association of carbonatites with large igneous provinces (LIPs; e.g., Deccan, Parana), carbonatites with primitive noble gas isotopic signatures, and their radiogenic isotope ratios similar to OIBs.
Numerous previous studies have advocated for a direct link between carbonatite melt generation and mantle plumes [74,78,82,83].As pointed out by Rukhlov and Bell [9], the presence of carbonatites may mark the initiation of mantle-generated magmatism because of the very fluid nature of carbonatitic melts, and the fact that they are produced by low degrees of partial melting (i.e., precursors to basaltic activity, and perhaps are associated with changes in mantle dynamics).In relation to the southeastern region of Québec and location of the MIP intrusions, tomographic data clearly indicates the presence of a low-velocity anomaly in the upper mantle region beneath the Ottawa-Bonnechere Rift [84].This anomaly is further interpreted to extend over a broad region at a depth of ~200 km beneath the Great Lakes, where lithosphere was partially breached by the Great Meteor plume [85].Hence, in relation to the MIP-related magmatism, we propose that carbonatite-like melts and volatile-bearing fluids first metasomatized the upper mantle at ~135 Ma, which gave rise to the older alkaline silicate rocks at Oka (e.g., alnöite).Subsequently, based on the limited geochronological data for the remaining MIP-related intrusions (Table 9) [16,86], the slightly undersaturated to critically saturated complexes of Mounts Royal, Bruno, Rougemont, Yamaska, Shefford, and Brome were emplaced between ~135 Ma and ~128 Ma.The last magmatism to occur involved generation of the moderately-to-strongly undersaturated silicate melts at Royal, Johnson, Yamaska, Shefford, and Brome that occurred ~117 Ma.

Chemical Zoning of Perovskite
Two perovskite grains (from a total of 25) investigated here display reverse zoning, with Nb and REE abundances that are enriched in the rim relative to their respective central regions of the crystals; the latter may have undergone Pb loss since these are characterized by younger ages resulting from higher U abundances relative to the rim (Figure 5).Minerals yielding younger ages within any given sample should always be carefully examined for Pb loss or U addition, especially since the Nb-D and Nb-E groups of perovskite have different trace element compositions (e.g., Pb, Th; Figure 7 and Tables 2 and 3). Figure 13A plots U-Pb ages against their respective U abundances for perovskite and it is possible that the young ages corresponding to the higher U contents in samples Oka229, Oka73, and Oka137 can be attributed to Pb loss.The reason being that perovskite with higher U abundances (~200 ppm in this case) will undergo a higher amount of alpha decay from the radioactive disintegration of U, which consequently damages the crystal structure; this then enhances the possibility of losing loosely bound radiogenic Pb.In contrast, perovskite from okaite samples Oka208 and Oka209 yield relatively young ages of 112.2 ± 1.9 Ma (Figure 10E) and 116.4 ± 2.9 Ma, respectively, with U contents comparable to those for the older perovskite (Figure 13A).A previous study of perovskite attributed the higher abundances of incompatible elements at grain boundaries to secondary processes; e.g., alteration in intergranular regions as a result of interaction with a melt, an aqueous fluid, or a gas phase [53].As stated above, the core areas of two large perovskite grains may have undergone a Pb loss event, and these are characterized by more radiogenic Sr (and comparable Nd) isotopic ratios compared to the remaining perovskite (Figure 12C).In addition, the core regions are marked by distinct chemical compositions (i.e., lower Nb/Zr; Figure 14A,B).Hence, the distinct, more radiogenic 87 Sr/ 86 Sr isotope compositions for the cores of these two perovskite grains may be attributed to either crystallization from a melt derived from a distinct mantle source, or perturbation by a contamination/alteration process (Figures 12C and 14).Thus, a possible formational history for the reversely zoned perovskite grains is as follows: (1) the cores formed from a first batch of magma; (2) this was followed by the influx of a new (distinct) batch of magma with lower 87 Sr/ 86 Sr ratio (relative to the cores), which resulted in the crystallization of the rims; and (3) the latter was associated with an "autometasomatic" event in which fluids scavenged the Nb (and certain other trace elements) from the core towards the rim.Obviously, this process was not widespread since only two of the perovskite grains investigated here exhibit this anomalous, reversely zoned texture.It is possible that this "autometasomatic" activity produced the vast majority of the late-stage pyrochlore and/or niocalite resulting in the Nb ore deposits at Oka. Chen et al. [15] discussed the issue of late-stage replacement of niocalite by pyrochlore (or vice versa) at Oka.In addition, Samson et al. [87] also advocated for the occurrence of late-stage hydrothermal activity at Oka as recorded by fluid inclusions within constituent minerals.
The highly variable chemical compositions, Nd and Sr isotope ratios, and ages documented for perovskite, pyrochlore, niocalite, and apatite from the different rock types associated with the Oka carbonatite complex indicate that these formed as a result of episodic, small volume partial melting and subsequent magma mixing [14,15,68].However, Figure 13B shows that there is a positive correlation between the total REE contents and U-Pb ages for the perovskite grains investigated here (excluding the two reversely zoned grains).Thus, a possible interpretation is that the Nb-E perovskite formed first within a melt produced at ~135 Ma, and their enriched geochemical nature is the result of low-degree partial melting of a metasomatized (carbonated) mantle source region.The Nb-D perovskite formed later at ~114 Ma from a less-enriched magma and possibly reflects derivation from a more depleted (less metasomatized) mantle source region.

Conclusions
This study reports combined geochemical, isotopic, and geochronological data for both perovskite and apatite from the Oka carbonatite complex, and clearly demonstrates that a more detailed petrogenetic history can be deciphered for complexly zoned igneous centers.Of importance, the U-Pb results from this study indicate the need for conducting a thorough geochronological investigation rather than defining the age of any one alkaline intrusive complex solely on the basis of a single or small number of radiometric age determinations.
The combined chemical, isotopic and geochronological data for apatite suggest that its crystallization occurred during the entire magmatic history of the complex.Moreover, the lack of any significant correlations between geochemical and geochronological results for apatite and niocalite indicate a complicated petrogenetic history involving magma mixing [14,15].On the basis of correlations between chemical compositions and U-Pb ages for the perovskite investigated in this study, these formed during two main episodes of melt generation.
The geochronological results from this study offer valuable insights into the emplacement relationships between carbonatite, okaite, ijolite and alnöite at Oka.It is proposed that carbonatite-like melts/fluids were the first to emanate from an enriched, volatile-bearing mantle plume, and these interacted with the overlying lithosphere; ensuing alkaline silicate melts were formed and generated the first alnöite, ijolite and okaite emplaced at ~135 Ma.Periodic, small volume partial melting subsequently continued, with a second major pulse of magmatism that occurred at ~114 Ma.Later generation melts mixed with earlier-formed rocks and minerals so as to yield samples with multiple-aged accessory minerals; these are considered as cognate crystals [14].The Sr and Nd isotopic compositions for perovskite and apatite indicate the involvement of at least two mantle endmembers, HIMU-and EMI-like, within their mantle source region, although dominated by the former component.Given the results reported here and from previous investigations on Oka [14,15], it is difficult to assign either component to a mantle region; i.e., lithosphere vs. asthenosphere (or plume).Alternatively, the mantle plume itself may be isotopically heterogeneous as proposed for the magmatic/tectonic regime for the East African alkaline province [58].Regardless of which model is preferred, infiltration/refertilization of the lithosphere by enriched, volatile-bearing melts/fluids from a plume component will "swamp" the geochemical and isotopic composition of the overlying lithosphere [82].

Figure 5 .
Figure 5. Petrographic image showing the zonation of perovskite.(A) U (ppm) concentrations obtained by LA-ICPMS with a 25 μm spot size; (B) Nb 2 O 5 (wt %) abundances analyzed by EMP with a 5 µm beam; and (C) 206 Pb/ 238 U weighted mean ages (determined by LA-MC-ICPMS using a spot size of 75 μm) across a zoned perovskite grain from sample Oka229 (okaite).

Figure 6 .
Figure 6.Chondrite normalized Rare Earth Element (REE) patterns for perovskite from alkaline silicate samples at Oka. Chondrite values are from McDonough and Sun [55].

Figure 9 .
Figure 9. Chondrite normalized REE patterns for apatite from different rock types at Oka.As with Figure 8, additional REE abundances for apatite from carbonatite, okaite and melanite ijolite are from Chen and Simonetti [14].The grey shaded area outlines the normalized patterns for apatite from alnöite and jacupirangite.Chondrite values are from McDonough and Sun [55].

Figure 10 .
Figure 10.U-Pb isotopic ages for apatite and perovskite from the Oka carbonatite complex.Examples are illustrated for samples with a single age for apatite (A) and bimodal age distributions (B,C).Diagrams (D) and (E) illustrate examples of perovskite age results, (D) gives a bimodal distribution age and (E) yields a single young age.All reported uncertainties are at 2σ level as determined by Isoplot[46].The Mean Square Weighted Deviation (MSWD) is used as a statistical validity of the regression line according to the criteria defined by Wendt and Carl[56].

Figure 12 .
Figure 12. (A) Diagram of 143 Nd/ 144 Nd vs. 87 Sr/ 86 Sr shows data obtained in this study and by Wen et al. [12].(B) Plot of 143 Nd/ 144 Nd vs. 87 Sr/ 86 Sr.Also shown are the East African Carbonatite Line (EACL) from Bell and Blenkinsop [58], and CHUR and Bulk Earth (BE) values for comparison.(C) Diagram of 143 Nd/ 144 Nd vs. 87 Sr/ 86 Sr values for the different groups of perovskite.

Figure 14 .
Figure 14.(A) Diagram of 87 Sr/ 86 Sr vs. Nb/Zr illustrating the different groups of perovskite; and (B) Plot of 143 Nd/ 144 Nd vs. Nb/Zr for the different perovskite groups.

Table 1 .
Summary of the mineralogy of the main rock types at Oka.
Note: Mineral occurrences are reported in volume percent.

Table 2 .
Major element compositions of perovskite from the Oka complex.

Table 2 .
Cont.Major element compositions are presented in wt %.Structural formulae are calculated based on 3 atoms of oxygen.The mol % of the endmembers are calculated following the sequence: (1) Ti 4+ is assigned for CaTiO 3 ; (2) latrappite is calculated based on the availability of Ca 2+ , Nb 5+ , Fe 3+ and Al 3+ ; (3) depends on the available Na + and Nb 5+ , mol % of NaNbO 3 is calculated; and (4) the rest of the Fe 3+ is combined with the LREE 3+ to (LREE)FeO 3 .

Table 3 .
Trace element abundances of perovskite.
Note: Trace element concentrations are listed in ppm.
Notes: Major element compositions are presented in wt %; Structural formulae of apatite are calculated based on 12 atoms of oxygen.
Notes: Trace element concentrations are listed in ppm; b.d.= below detection limit.