The Syenite–Carbonatite Complex of Ihouhaouene (Western Hoggar, Algeria): Interplay Between Alkaline Magma Differentiation and Hybridization of Cumulus Crystal Mushes

The 2 Ga-old Ihouhaouene alkaline complex (Western Hoggar, Algeria) is among the oldest known carbonatite occurrences on Earth. The carbonatites are calciocarbonatites hosted by syenites, the predominant rock type in the complex. Both rock types are characterized by medium-grained to pegmatitic textures and contain clinopyroxene, apatite, and wollastonite, associated with K-feldspar in syenites and a groundmass of calcite in carbonatites. The rock suite shows a continuous range of compositions from 57–65 wt.% SiO2 and 0.1–0.4 wt.% CO2 in red syenites to 52–58 wt.% SiO2 and 0.1–6.5 wt.% CO2 in white syenites, 20–35 wt.% SiO2 and 11–24 wt.% CO2 in Si-rich carbonatites (>10% silicate minerals), and <20 wt.% SiO2 and 24–36 wt.% CO2 in Si-poor carbonatites (<5% silicate minerals). Calculation of mineral equilibrium melts reveals that apatite and clinopyroxene are in disequilibrium with each other and were most likely crystallized from different parental magmas before being assembled in the studied rocks. They are subtle in the red syenites, whereas the white syenites and the Si-rich carbonatites bear evidence for parental magmas of highly contrasted compositions. Apatite was equilibrated with LREE-enriched (Ce/Lu = 1,690–6,182) carbonate melts, also characterized by elevated Nb/Ta ratio (>50), whereas clinopyroxene was precipitated from silicate liquids characterized by lower LREE/HREE (Ce/Lu = 49–234) and variable Nb/Ta ratios (Nb/Ta = 2–30). The Si-poor carbonatites resemble the Si-rich carbonatites and the white syenites with elevated REE contents in apatite equilibrium melts compared to clinopyroxene. However, apatite equilibrium melt in Si-poor carbonatite shows a majority of subchondritic values (Nb/Ta<10) and clinopyroxene has chondritic-to-superchondritic values (Nb/Ta = 15–50). Although paradoxical at first sight, this Nb-Ta signature may simply reflect the segregation of the carbonatite from highly evolved silicate melts characterized by extremely low Nb/Ta values. Altogether, our results suggest an evolutionary scheme whereby slow cooling of a silico-carbonated mantle melt resulted in the segregation of both cumulus minerals and immiscible silicate and carbonate melt fractions, resulting in the overall differentiation of the complex. This process was however counterbalanced by intermingling of partially crystallized melt fractions, which resulted in the formation of hybrid alkaline cumulates composed of disequilibrium cumulus phases and variable proportions of carbonate or K-feldspar.

However, whatever process is envisioned for the origin of carbonatites, the common association of carbonatites with alkaline silicate rocks must be taken into account as it is key stone to the understanding of their petrogenesis. Most often, carbonatites are associated with silica-undersaturated, highly alkaline silicate rocks such as nephelinites, ijolites, phonolites, and nepheline syenites (e.g., Bell et al., 1998;Woolley, 2003;Woolley and Kjarsgaard, 2008). The association of carbonatites with silica-saturated rocks is less common; it was estimated to represent 8% of carbonatites worldwide by Woolley and Kjarsgaard (2008).
The Ihouhaouene alkaline complex is one of those. Similar to Mountain Pass, California (Olson et al., 1954), Oka, Canada (Eby, 1975), and Eden Lake, Canada (Chakhmouradian et al., 2008), the complex shows the association of calciocarbonatites with silicasaturated syenites. The two rock types are closely associated in the field, being intermingled at various scales and showing similar pegmatitic/brecciated textures. Carbonatites and syenites also have similar Nd, Sr, and C isotopic signatures consistent with a mantle origin and ruling out significant crustal contamination (Bernard-Griffiths et al., 1988;Ouzegane et al., 1988;Fourcade et al., 1996). The rocks are also virtually free from subsolidus reactions and secondary alteration, making the Ihouhaouene complex a valuable case study to address petrogenetic relationships between carbonatites and silica-saturated syenites.
We report major and trace element data for whole rocks and minerals in all rock facies. However, our interpretations are principally based on the systematics of apatite and clinopyroxene equilibrium melts since the whole rocks contain a significant cumulate component and are not considered to represent melt compositions. Our results reveal that carbonatites and syenites from the Ihouhaouene complex are even more intimately related than suggested by geologic, isotopic, and petrographic evidence. The two rock types share a complex history of igneous differentiation partly counterbalanced by mixing of cumulus minerals crystallized from melt fractions of contrasting compositions, including previously segregated carbonate and silicate liquids.
The carbonatites and syenites occur as linear outcrops along shear zones in three sectors Lapin and Ploshko, 1988;Fourcade et al., 1996; Figure 1): 1) Oued Ihouhaouene characterized by several small hills-oriented EW and NW-SE; 2) the South Tirahart wadi complex with outcrops of approximately 4 km length and 1 km width, and 3) Tirahart wadi north, a hill spreading over 2.5 km length and 250 m width. Carbonatites outcrop as veins or massive formations in syenites. The contact between carbonatite and syenite is sharp (with no reactional selvage), and only local syenites have mylonitized edges (<20 cm).
According to their texture, carbonatites present as massive structures up to 50 m thickness and max. 500 m length with clinopyroxene and apatite medium grains in calcite matrix and can be brecciated including some syenite fragments with sizes ranging from 1 to 70 cm partially surrounded by clinopyroxene reaction aureoles (Figures 2A,B) or as pockets or small veins (<2 m width and <10 m length) coarse grained to pegmatitic with euhedral apatite (3-50 mm) ( Figures 2C,D) and calcite. Clinopyroxene and alkali feldspar euhedral megacrysts are present in these pegmatitic carbonatites ( Figure 2D).
The syenites form large outcrops in all three sectors. They are red ( Figure 2E) or white ( Figure 2F) according to the presence of red alkali feldspar, or white alkali feldspar and wollastonite, respectively. Both syenites present structures with magmatic segregation of clinopyroxene and feldspar minerals with the same orientation as the surrounding granulites. Red syenites are more abundant in Oued Ihouhaouene and Tirahart wadi North complexes. In Tirahart wadi North complex, white syenite fragments composed of wollastonite, garnet, and large crystals of pyroxene cemented by feldspars form feldspathic breccia. The contact between red and white syenites is brecciated. Only in Oued Ihouhaouene, sharp contacts are observed, with large white syenite bodies crosscutting red syenites (previously mapped as granites by Fourcade et al., 1996). The contact between syenites and granulitic basement is not obvious because of sand cover in this area, but the outcropping granulites (far from ∼20 m) in Oued Ihouhaouene and North wadi Tirahart present no evidence of fenitization. The granulitic crust preserved refractory chemical compositions (Ouzegane et al., 2003), not consistent with metasomatism and fenitization induced by a second local fluid (e.g., in situ alkali-metasomatism of country rock around intrusions of carbonatites and/or alkaline rocks, Kresten, 1988;Drüppel et al., 2004;Elliott et al., 2018). Syenite outcrops have homogenous mineralogy from the contact with carbonatite to the sand cover area separating the complex from granulite host rock. The sand cover area between syenites and granulite outcrops is not larger than 20 m in between outcrops and no fenitization aureole has been observed.

ANALYTICAL METHODS
Fifty-four intrusive silicate and carbonatite rocks representing the diversity of rock types in the Ihouhaouene area were selected for petrological and geochemical study. Samples were chosen according to the localities and the different types of carbonatites (brecciated, pegmatitic, or calcite veins) and the two types of red and white syenites. Brecciated carbonatites were well separated and cleaned off syenite fragments before milling. A significant amount (∼3 kg) of pegmatitic carbonatites was milled to have a representative bulk-rock composition.

Whole-Rock Major, Trace, and Volatile Element Analysis
All samples were cut, crushed, and milled in agate into fine powders (≥5 μm) without using water. Major elements were analyzed by ICP-OES (iCap ThermoFisher) at the SARM service of the CRPG (Nancy, France). The procedure followed is according to the protocol of Carignan et al. (2001). The wholerock powder was fused with LiBO 2 and then dissolved in a HNO 3 , H 2 O, and glycerol mixture. The relative standard deviation for the major elements is less than 5%. Carbon and sulfur contents of the samples were determined using an element analyzer at the SARM service, whereas F and Cl were determined by wet precipitation ferrithiocyanate spectrophotometry using a Varian Cary 50 spectrophotometer.
Trace elements were analyzed using a quadrupole 7,700x ICP-Mass Spectrometry at Geosciences Montpellier (AETE platform OSU-OREME). An acid attack on 0.1 g of sample powder weighed in teflon-lined vessel is carried out with 1 ml of HClO 4 and 2.5 ml of HF on a hot plate at 100°C for 24 h. Hydrofluoric acid breaks the silicate link of rocks and puts the released ions in solution, while perchloric acid acts as a powerful oxidant. Then, the samples were evaporated to dryness at 150°C. A second attack is applied with 0.5 ml of HClO 4 and 1 ml of HF for 24 h followed by evaporation, and the samples are then taken up in three steps, separated by evaporation, in 0.5 ml, 0.25 ml, and then 0.25 ml of HClO 4 . The sample is evaporated to a solid residue. The first dilution is accomplished with this residue by adding nitric acid and MilliQ water until having a solution of 20.6 g. Before mass spectrometer analysis, the sample solution is diluted with 9.45 ml of H 2 O, 0.25 ml of HNO 3 , and 0.1 of InBi according to a factor of 10,000, which has been chosen for the associated carbonatites and syenites. For each series of analysis (20 samples), four blanks (only acids were introduced into Savilex teflon-lined vessel) and four standards, BHVO-1, BEN, UBN, and COQ-1, were analyzed to monitor internal drift. The mass spectrometer used to quantify trace element concentrations in whole rock is Agilent 7,700x MC-ICPMS. The sensitivity is of the order of 200 × 106 cps/ppm on 115 In (indium). Internal standards A2 and A4 are calibrated for Zr-Nb and Hf-Ta concentrations, and calibration curves on Rb, Sr, Y, Zr, and Ce are obtained to verify the internal calibration.

In Situ Mineral Analysis for Major, Trace, and Volatile Elements
Major and volatile element concentrations in minerals were analyzed using electron microprobe (CAMECA SX-100 at Geosciences Montpellier, France). Peaksight software was used to perform calibration, overlap correction, and quantification. Raw data processing for matrix corrections was performed using ZAF and PAP models (Castaing, 1963). Acceleration voltage and beam current were set to 20 kV and 10 nA, and a focused beam (1 µm) was used except for carbonates that were analyzed with a defocused beam (10 µm). The background counting time was fixed at 20 s for major elements and 40 s for F, Cl, and S. The analyses were calibrated on reference material standards chosen for each element: Al 2 O 3 for Al, TiO 2 for Ti, wollastonite for Ca and Si, albite for Na, orthoclase for K, forsterite for Mg, hematite for Fe, apatite for P, native metal for Ni, Mn, and Cu, celestine for Sr, ThO 2 for Th, glasses for REE, baryte for S and Ba, fluorite for F, and chloroapatite for Cl. We observed F migration in apatite for the analytical conditions used for major and minor element analyses (Goldoff et al., 2012). The F content has been estimated using reference apatite crystal oriented with its c-axis perpendicular to the incident electron beam with 3.4 wt.% F. Trace element analyses of minerals were determined by LA-HR-ICP-MS at Geosciences Montpellier (AETE platform, OSU-OREME), using GeoLas Q + Excimer CompEx102. A laser beam of diameter 26 µm was used for apatite and britholite, 51 µm for silicate minerals, and 77 µm for calcite and wollastonite with a repetition rate of 7 Hz, a laser power of 5 J/cm 2 , and a mixture of helium and argon as carrier gas. Each analysis consisted of a preablation pulse of 10 s, 2 min measurement of the background, and 60 s integration while the laser fires. The concentrations were calibrated with NIST 612 and the SiO 2 -CaO concentrations determined previously with electron microprobe for each mineral. BIR-G1 standard is used as external standard. The detection limits depended on the day-to-day tuning of the ICPMS, the "gas blank" data obtained from the mass spectrometer before the laser is turned (background), the laser spot size, and the element being analyzed. Two-minute measurement of the background was performed for optimization of the peak integration. The instrument sensitivity due to analytical conditions was determined from the average across all days of repeat measurements of the synthetic NIST 612 glass (Pearce et al., 1997). Data were processed with Glitter software (Griffin et al., 2008), and elements were considered below detection and filtered out of the dataset if the concentration had a standard deviation (1 sigma) that was larger than the value of the detection limit. Values for standards and detection limit are reported in Supplementary Table S2.
Si-rich carbonatite REE patterns are similar to those of Sipoor carbonatites. They are enriched in 2,633 to 7,551 times chondrite with a slight negative anomaly in Eu (Eu/Eu* 0.69 to 0.82; Figure 6C) and La N /Yb N ratio from 88 to 145. The concentrations of LILE and high-field-strength elements (HFSEs) in Si-rich carbonatites are higher than in Si-poor carbonatites (0.31-1.46 ppm Cs, 29.79-183 ppm Rb, 559-3,104 ppm Ba) with high contents in U and Th (Table 1; Figure 6D) and Nb-Ta and Zr-Hf fractionation (Zr/Hf 22.28-59.20, Nb/Ta 3.39-11.14).

MINERAL CHEMISTRY Clinopyroxene
FIGURE 5 | Bulk rock major oxide compositions of the Ihouhaouene carbonatites and syenites. Symbols as in Figure 4.  Figure 8A). They display positive anomalies in Ba, U, Pb, and Sr and Zr-Hf and Nb-Ta fractionation (Zr/Hf 9.26-20.05, Nb/Ta 10.33-17.37; Figure 8B; Supplementary   Table S3). Diopside and hedenbergite have low TiO 2 contents (TiO 2 0.02-0.33 wt.%, Figure 7B), and Al 2 O 3 contents range from 0.49 to 1.42 wt.% and 0.22 to 1.04 wt.%, respectively. REE patterns have a sinusoidal to asymmetric U-shape with slightly high concentrations in LREE compared to HREE (La N /Yb N 1.17-4.89; Figure 8C; Supplementary Table S4). Large variations in LILE concentrations are observed in Si-rich carbonatite pyroxenes with strong positive anomalies for Pb and Sr in addition to Zr-Hf and Nb-Ta fractionation (Zr/Hf 20.71-33.93, Nb/Ta 1.57-50.73; Figure 8D).  Table S3). Chondrite-normalized patterns of diopside in red syenites display a convex LREE section (La N / Sm N 0.13-1.30) and a negative Eu anomaly (Eu/Eu* 0.11 to 0.28; Figure 8E; Supplementary Table S4). Relative to the primitive mantle, pyroxenes have high fractionation in Nb-Ta (Nb/ Ta 1.63-28.55) with negative anomalies in Ba, Pb, Sr, and Zr and enrichment in Th ( Figure 8F; Supplementary Table S4). It should be noted that clinopyroxenes that crystallized around red or white syenite fragments are diopside or hedenbergite, respectively, with the same composition compared to interstitial clinopyroxene in red or white syenite.  Table S3). The REE patterns of hedenbergite in white syenites have low REE content and the same sinusoidal to asymmetric U-shaped pattern compared to Si-rich carbonatites, with high LREE and HREE concentrations compared to MREE (La N /Yb N 0.58-3.12, La N /Sm N 0.73-2.27) ( Figure 8G). They have variable trace element contents down to very low values (i.e., Rb, Nb, Ta, Pb, and Sr) compared to pyroxenes in red syenites ( Figure 8H; Supplementary Table S4).

Apatite
Apatites in carbonatites and syenites are fluorapatites (1.8-4.5 wt.% F, Supplementary Table S5). They have different silica, phosphate, calcium, and REE contents depending of the host rock (Figure 9; Supplementary Table  S5), but despite their relatively large size (up to 5 mm), apatite crystals do not have significant zonation from core to rim or dissolution/recrystallization texture.
White syenites have silica-rich apatites with 1.  Figure 9). REE patterns of apatite in white syenites display the same regular shape compared to Si-rich   Figure 10H).

Accessory Minerals
Accessory minerals (<1 vol.%) are present as interstitial minerals, corona, or small inclusions in some samples (Supplementary Table S1; Section 3) with texture of crystallization at subsolidus conditions. For clarity, we present only a brief overview of the texture and composition of accessory minerals as they represent very small volume of the rocks and do not significantly change the composition of magmatic minerals and bulk rocks. It is important to note that inclusion-free and inclusion-bearing apatites have the same composition and associated clinopyroxenes are unzoned with homogenous composition in each rock facies.

Monazite
Monazite is likely also exsolution Lorenz et al., 2019); it is present as inclusions in apatite and around apatite crystals in Si-poor carbonatites and red syenites. It has more than 60% of REE with high La and Ce concentrations

The Ihouhaouene Complex: Case Study for the Association of Carbonatites With Silica-Saturated Syenites
When associated with alkaline silicate rocks, carbonatites worldwide are mostly associated with silica-undersaturated rocks (e.g., Brava Island, Cape Verde- Weidendorfer et al., 2016;Kovdor, Russia-Verhulst et al., 2000); only thirty-two occurrences are distinguished by their association with silicasaturated rocks (i.e., 8% of carbonatites worldwide- Woolley and Kjarsgaard, 2008). Among those, Mountain Pass, California (Olson et al., 1954), Oka, Canada (Eby, 1975;Chen and Simonetti, 2013), and Eden Lake, Canada (Chakhmouradian et al., 2008) show the association of calciocarbonatites with silica-saturated syenites and are furthermore comparable to Ihouhaouene in terms of mineral assemblages (diopside-hedenbergite, apatite) and major and trace element compositions (Figure 4). In most cases, a cogenetic origin has been assumed for carbonatite and silicate rocks, whether saturated or undersaturated. However, the original magmatic features were often obscured by late hydrothermal/metamorphic imprints, precluding full understanding of petrogenetic relationships between rock facies (e.g., Jones and Wyllie, 1985;Ramasamy, 1986;Smith et al., 2000;Chakhmouradian et al., 2008;Saha et al., 2011). In the Ihouhaouene complex, the FIGURE 10 | Normalized REE and trace element contents of apatite from Ihouhaouene carbonatites and associated silicate rocks. Chondrite values and primitive mantle values are obtained from Sun and McDonough (1989 coexistence of silica-saturated syenites with carbonatites characterized by a wide range of silica content (3-35 wt.% SiO 2 ) and the lack of significant postmagmatic imprint makes it possible to address the relationships between these rock types and shed light on the origin of the syenite-carbonatite association. We discuss in the following the salient features of the Ihouhaouene complex that should be considered in a petrogenetic model for the rock suite. This includes the geological framework of the intrusions, petrographic observations, and trace element systematics of rock-forming minerals and equilibrium melts.

Geological and Isotopic Evidence for a Single Igneous Event and a Common Magma Source for Carbonatites and Syenites
The Ihouhaouene carbonatites and syenites are closely associated in the field, where they form elongated bodies emplaced along major shear zones Ouzegane et al., 2003). In several complexes worldwide, the same pathway has been used by more than one magmatic event (e.g., Eifel, Vogelsberg, Kaiserstuhl-Jung et al., 2005;Walter et al., 2018;Giebel et al., 2019). In Ihouhaouene, however, carbonatites and syenites are intermingled at various scales and display similar pegmatitic/brecciated textures ( Figure 2) and mineral compositions, suggesting that they were emplaced during a single event. Cogenetism of carbonatites and syenites is also supported by the parallelism between pervasive igneous layering in the complex and ductile flow foliation in the host granulite, indicating that the whole alkaline complex behaved coherently during late crystallization stages Ouzegane et al., 2003). Finally, carbonatites and syenites have similar Nd, Sr, and C isotopic signatures consistent with a mantle origin (εNd (T) −6.3 to −9.9‰, 87 Sr/ 86 Sr (T) 0.7093 to 0.7104, δ 13 C −3.5-9.7‰-Bernard- Griffiths et al., 1988;Ouzegane et al., 1988;Fourcade et al., 1996). These values also indicate that the Ihouhaouene parental magmas were not significantly affected by granulite contamination, even though carbonate melt intrusion into a tectonically active setting can potentially result in the assimilation of fragmented wall-rock materials (Chakhmouradian et al., 2008;Giebel et al., 2019).

Similar Mineral Assemblages as Evidence for Close Relationships Between Carbonatites and Syenites Throughout Parental Melt Differentiation
One of the most conspicuous features of the Ihouhaouene complex is the presence of texturally and compositionally identical mineral assemblages (chiefly pyroxenes and apatite) occurring in both carbonatites and syenites. This is particularly true for the silica-rich carbonatites and the white syenites (Figures 7, 9) that differ from each other only by the presence of carbonates in the carbonatite while the white syenite contains alkali feldspar instead. Other mineralogical phases (clinopyroxene, wollastonite, and apatite) are almost indistinguishable in both rock types. This observation strongly supports a close cogenetic relationship between the two rock types. However, the presence of silicate minerals in carbonatites is rather unusual and must be addressed as it is most likely a key feature for deciphering the relationship between carbonatites and syenites.
Silicate minerals in carbonatites were sometimes ascribed to subsolidus recrystallization or interpreted as xenocrysts entrained by the magma (Barker, 2001;Giebel et al., 2019). Secondary-likely subsolidus-reactions are observed in the Ihouhaouene complex. Their imprint is however restricted to the development of narrow garnet coronae after K-felspar or pyroxene, crystallization of accessory minerals such as allanite, or exsolution of britholite in apatite. By no means can these processes explain the common occurrence of subhedral clinopyroxenes in the Ihouhaouene carbonatites. Since they show no evidence of resorption or zonation nor any reaction rims in contact with calcite or apatite, the clinopyroxenes are also unlikely to represent xenocrysts. Rather, they represent primary, igneous minerals crystallized during the magmatic evolution and solidification of the Ihouhaouene complex.  Taylor and McLennan (1985). Altered wollastonite is shown in dashed lines.
Frontiers in Earth Science | www.frontiersin.org February 2021 | Volume 8 | Article 605116 A primary, igneous origin of minerals in the Ihouhaouene carbonatites is further attested by their trace element composition. The chondritic-to-near-chondritic Y/Ho ratio in clinopyroxene (23-30), wollastonite (27-34), apatite (27-32), and calcite (28-40) in both carbonatites and syenites strongly suggests a magmatic origin at high temperature and crystallization over a temperature range evident by the increasing Y/Ho ratios in calcite (e.g., Chakhmouradian et al., 2017;Anenburg et al., 2020). However, while the Si-rich carbonatites contain hedenbergitediopside (Mg# 0.42-0.77- Figure 7) associated with Si-rich apatite and wollastonite (Figures 9, 11), the Si-poor carbonatites contain diopsides (Mg# 0.63-0.72) associated with Si-poor apatite. This feature corroborates the sensitivity of silicate mineral stability and composition in carbonatites to silica activity that is usually low (Barker, 2001;Massuyeau et al., 2015), whereas the crystallization of wollastonite and hedenbergite is expected in a highly silica and calcic carbonatite with slightly low magnesium activity . The crystallization of hedenbergite over diopside in all rock facies is related to different Mg and Fe 2+ activities. It should be noted that all clinopyroxene contain the same high Fe 3+ content indicating similar oxidation state during crystallization.
Silica activity in carbonatites can increase as a result of assimilation of cogenetic igneous silicate rocks or contamination by country rocks at high temperature, recently described as "antiskarn metasomatism" (Chakhmouradian et al., 2008;Giebel et al., 2019;Anenburg et al., 2020). Due to the close intermingling of FIGURE 12 | Chondrite normalized REE contents of recalculated apatite and clinopyroxene equilibrium liquids from Ihouhaouene carbonatites and associated syenites compared to nephelinite, melilitite, and alkali basalt REE compositions. Chondrite and alkali basalt values are obtained from Sun and McDonough (1989). Hanang nephelinite and Labait melilitite are obtained from Baudouin et al. (2016) and Baudouin and Parat (2020 syenites and carbonatites in the complex and the existence of brecciated textures, the Ihouhaouene carbonatites often contain syenite fragments. These fragments may be surrounded by aureoles of clinopyroxene with the same composition as euhedral pyroxene crystals, but no further evidence for syenite assimilation was observed. In particular, the syenite fragments display no trace of zoning as would be expected for reaction with carbonatite. Although contamination by the syenite cannot be definitely excluded, the relatively high silica activity required to crystallize clinopyroxene in the Ihouhaouene carbonatites was most likely achieved by differentiation and unmixing from a carbonate-rich primary silicate melt, as previously suggested by Dawson (1998) and Baudouin et al. (2016).

Trace Elements Reveal Some Complexity in the Petrogenetic Evolution of the Igneous Suite
Trace element distribution in the Ihouhaouene complex shows several features that cannot be straightforwardly explained by a simple scheme of differentiation and unmixing from a single parental melt. These features notably include (1) The extremely wide range of Nb/Ta ratios in minerals, from subchondritic-to-superchondritic values (0.7-177 in apatite and clinopyroxene; Nb/Ta chondrite 18- Sun and McDonough, 1989). Both partial melting of heterogeneously carbonated mantle and separation of immiscible carbonate melts from silicate melts have been advocated to explain substantial Nb-Ta fractionation (Koster van Groos and Wyllie, 1968;Wyllie and Huang, 1975;Freestone and Hamilton, 1980;Kjarsgaard and Peterson, 1991;Dalton and Wood, 1993;Kjarsgaard et al., 1995;Mitchell, 2009;Brooker and Kjarsgaard, 2011;Martin et al., 2013;Novella et al., 2014). These processes would notably account for the elevated, superchondritic Nb/Ta values frequently observed in carbonatites. In the Ihouhaouene complex, however, both carbonatites and syenites show mostly subchondritic Nb/Ta values, with paradoxically somewhat lower values in carbonatites (0.5-17.3- Table 1) than in syenites (4.2-21.5). However, considering the unusual mineralogy and major element chemistry of many carbonatites, as well as their coarse grain size in several instances, the Ihouhaouene wholerock compositions are probably largely governed by the accumulation of "cumulus" phases, such as apatite and clinopyroxene, and accessory titanite. Therefore, these compositions are unlikely to represent true parental melt compositions. As regards to Nb-Ta, the variability of parental melts is probably more adequately reflected by the wide range of Nb/Ta values registered in minerals. Most samples, including the syenites, contain high Nb/Ta minerals (Figures 8, 10) that were likely equilibrated with carbonate melts.
(2) The paradoxically low LREE content minerals of the Si-poor carbonatites compared with other rock types (Figures 8, 10).
Carbonate-rich melts produced by low melting degree of a carbonated mantle source or by unmixing of carbonate melt from primary silicate melt are expected to be LREE-enriched (Hamilton et al., 1989;Jones et al., 1995;Martin et al., 2013), whereas recent experimental studies point out the role of calcium and silica activities in REE exchange reactions and mineral compositions. In clinopyroxene, the incorporation of  Klemme et al. (1995) and Blundy and Dalton (2000). For syenites, we selected partition coefficients between silica-rich melt and apatite from Mahood and Stimac (1990) and Prowatke and Klemme (2006) and partition coefficients between silica-rich melt and clinopyroxene from Larsen (1979) and Mahood and Stimac (1990 (Anenburg et al., 2020), whereas HREE can enter in the Mg site, Mg 2+ + Si 4+ Lu 3+ + Al 3+ (Olin and Wolff, 2010;Dygert and Yan, 2014). High calcium and low aluminum activities would shift the LREE and HREE reactions to the left, meaning low LREE and HREE in the clinopyroxene. The unusual nonsinusoidal shape of REE pattern (Reguir et al., 2012) and slightly higher MREE and HREE contents in clinopyroxene in Si-poor carbonatite could be then explained by the convex upward pattern of partition coefficients for REE between clinopyroxene and calciocarbonatitic melt, indicating that clinopyroxene prefers MREE relative to LREE and HREE (Prowatke and Klemme, 2006;Reguir et al., 2012;Anenburg et al., 2020). HREE are much smaller than MREE and LREE cations and can fit into the tetrahedral silicon site, Ca 2+ + Si 4+ 2HREE 3+ , without requiring charge balancing with Al 3+ (Anenburg et al., 2020). In apatite, the REE content correlates positively with SiO 2 content ( Figure 9) indicating that the REE incorporation in the Ca site was accompanied by Si substitution for P, REE 3+ + Si 4+ Ca 2+ + P 5+ (Pan and Fleet, 2002;Hammouda et al., 2010). The exchange reaction can explain the low REE content in Si-poor apatite compared to Si-rich apatite. The low LREE concentration in apatite ( Figure 10) can be explained by lower partition coefficient for LREE between apatite and carbonatitic melt than MREE and HREE (convex upward pattern of partition coefficients for REE between apatite and calciocarbonatitic melt; Hammouda et al., 2010;Klemme and Dalpé, 2003) and equilibrium melt composition ( Figure 12; Section 7.5; Supplementary Table S9).
Moreover, the low, subchondritic Nb/Ta values observed in several minerals from this rock type (down to 1.9- Figures 8, 10) are opposite to the values that would be expected for differentiation from a primary melt source, either by magma unmixing (high Nb/Ta in the carbonatitic melt, Woolley and Kjarsgaard, 2008;Mitchell, 2009) or by crystal fractionation as the latter would likely involve high Nb/Ta melt after crystallization of Nb-Ta-bearing minerals (e.g., D Nb/Ta titanite/silicate liquid 0.07-0.28- Prowatke and Klemme, 2006). The trace element signature of the Ihouhaouene Si-poor carbonatites rather suggests a genesis of this rock type by segregation from low-Nb/Ta melt. In this scheme, the trace element signature of the Sipoor carbonatites is the reflection of the parental melt composition, characterized by low Nb/Ta and low LREE signature inherited from former differentiation stages involving several steps of fractional crystallization and unmixing.

Further Complexity in the Carbonatite-Syenite Relationship Unveiled by Apatite and Clinopyroxene Equilibrium Melts
As noted above, the Ihouhaouene whole-rock compositions are unlikely to represent melt compositions and therefore cannot be straightforwardly used to constrain the igneous evolution of the rock suite. To overcome this difficulty and shed light on magma compositions and their evolution, we calculated trace element compositions of equilibrium melts for apatite and clinopyroxene, using experimentally determined mineral/melt partition coefficients. For apatite, we took the apatite/carbonatite partition coefficients of Hammouda et al. (2010), who used apatite from Ihouhaouene for their experiments, and the apatite/silicate melt and apatite/trachyandesite of Prowatke and Klemme (2006) and Mahood and Stimac (1990), respectively. For clinopyroxene, we used the clinopyroxene/ carbonatite partition coefficients of Green et al. (1992) and Klemme et al. (1995) and the clinopyroxene/trachyte partition coefficients of Blundy and Dalton (2000), Mahood and Stimac (1990), and Larsen (1979) (see Supplementary Table S9 for details).
Considering the strong similarities of apatite and clinopyroxene major element compositions in carbonatites and white syenites (see Section 7.3), we made no a priori assumption on parental melt compositions in terms of major elements. Accordingly, we calculated trace element equilibrium melts using both mineral/carbonatite and mineral/silicate melt partition coefficients for each rock type. Results are illustrated in Figure 12 for chondrite-normalized REE and in Figure 13 for Ce/Lu and Nb/Ta. Equilibrium melts show considerable variation ranges, encompassing roughly four orders of magnitude for LREE ( Figure 12) and three for Ce/Lu and Nb/Ta ( Figure 13). Some systematics are however observed according to rock facies.
The white syenites and the Si-rich carbonatites are both distinguished by much higher REE content and Ce/Lu ratios in apatite equilibrium melts than in their clinopyroxene counterparts (Figures 12, 13A,B) and high, superchondritic Nb/Ta ratios (>50) in apatite equilibrium melts compared with subchondritic (<10) to superchondritic ratios in the clinopyroxene equilibrium melts (Figures 13A,B).
These results are observed with both carbonatite and silicate melt partition coefficient sets. They indicate that apatite and clinopyroxene are in disequilibrium with each other and-unexpectedly and most importantly-were most likely crystallized from different melt compositions before being eventually assembled in the studied rocks. Combined together, the REE and Nb-Ta signatures suggest that apatite was equilibrated with REE-enriched carbonate melts, whereas coexisting clinopyroxene was more likely precipitated from silicate liquids. The nonprimary (evolved) composition of the latter is indicated by their low REE content compared with primary alkaline mantle melts (Sun and McDonough, 1989;Baudouin et al., 2016;Baudouin and Parat, 2020). Based on the few analyses available, this is particularly true for the white syenites, whose clinopyroxene equilibrium melts are strongly depleted in REE ( Figure 12).
The Si-poor carbonatites resemble the Si-rich carbonatites (and the white syenites) in showing elevated REE contents in apatite equilibrium melts compared with clinopyroxene equilibrium melts and primary mantle melts (Figure 12) since the apatite equilibrium melt in this rock type shows a wide range of Nb/Ta ratios (<10- Figure 13A) and clinopyroxene equilibrium melts show chondritic-to-superchondritic values. In fact, this paradoxical Nb-Ta signature may simply reflect the segregation of the carbonatite from highly, although variably, evolved silicate melts characterized by low Nb/Ta values. The red syenites are distinguished from the other rock types by showing roughly similar REE contents in apatite and clinopyroxene equilibrium melts, at least at first glance (Figure 12), suggesting crystallization from a common silicate melt. Moreover, these compositions are roughly comparable to those of alkaline primary melts (Sun and McDonough, 1989;Baudouin et al., 2016), indicating that the red syenites have crystallized from silicate melt that did not reach carbonate saturation and immiscibility, compared with the other facies. In detail, however, equilibrium melts calculated from apatite and clinopyroxene slightly differ from each other by higher Ce/Lu ratios in the majority of the clinopyroxene equilibrium melts-higher in fact than in most clinopyroxene equilibrium melts from the other facies, and more comparable to the values observed in apatite equilibrium melts ( Figure 13B)-somewhat higher Nb/Ta values in the apatite equilibrium melts, compared with their clinopyroxene counterparts ( Figure 13B).
These differences are not consistent with the carbonate vs. silicate melt duality of parental magmas as suggested for the other facies. The possibility that these subtle differences are simply a reflection of uncertainties in values of the experimental partition coefficients and/or in their application to our samples cannot be definitely excluded. However, similar differences are observed with both carbonatite and silicate melt partition coefficient sets. Hence, somewhat distinct parental melts were possibly involved in the formation of red syenite minerals as well. The comparatively less contrasted parental melt compositions may reflect the formation of this facies in an earlier differentiation stage of the complex.
Finally, trace element systematics of apatite and clinopyroxene equilibrium melts reveal that carbonatites and syenites from the Ihouhaouene complex are even more intimately related than suggested by geologic, isotopic, and petrographic evidence. All studied samples, whether carbonatites or syenites, contain mineral assemblages formed from distinct melts and as such must be viewed as 'hybrid' igneous rocks. Moreover, the compositional contrast between coexisting parental melts tends to increase with the overall degree of differentiation. While only mildly heterogeneous silicate melts were-possibly-involved in the formation of the poorly evolved red syenite, the white syenite and the carbonatites bear evidence for parental magmas of highly contrasted compositions, including coexisting carbonate and silicate melts.
The mechanism where distinct magma batches that were first differentiated from each other by fractional crystallization and liquid unmixing were thereafter mixed back after incomplete crystallization is puzzling and will clearly deserve further investigation. The key might be sought in the peculiar geodynamic setting in which the Ihouhaouene complex was emplaced, i.e., lower crust affected by a high-T thermal event (Ouzegane et al., 2003). In this context, alkaline magmas injected from the mantle into shear zones may have remained partially molten for a long period of time, allowing intermingling and mixing of magma batches and crystal mushes characterized by variable degrees of differentiation and solidification.

CONCLUSION
At first sight, the extreme diversity of trace element compositions recorded by the Ihouhaouene equilibrium melts might suggest that the complex was formed from a range of primary mantle melts generated by variable degrees of melting of a carbonated mantle source. This scheme is however at odds with the occurrence of LREE-depleted minerals in the Si-poor carbonatites, whereas the melting model would predict the opposite, i.e., selective LREE enrichment in carbonate-rich melts produced by the lowest melting degrees. In addition, variable melting degrees would hardly account for the considerable range of Nb/Ta variations in equilibrium melt, a feature, which cannot be explained by fractional crystallization alone either.
Unmixing of carbonated melt from a silico-carbonated parental magma may account for substantial Nb-Ta fractionation (e.g., Martin et al., 2013). However, carbonatites generated by liquid-liquid separation are expected to show elevated Nb/Ta values. This is the case for the equilibrium melts calculated from apatite in the Si-rich carbonatites, but not for the apatite equilibrium melts in the Si-poor carbonatites. The latter are mostly characterized by low, subchondritic Nb/Ta ratios, a signature that may be explained by segregation of carbonated melts from an evolved parental melt residual after earlier stages of unmixing.
Previous studies of volcanic alkaline/carbonatite systems have pointed out the role of fractional crystallization to reach carbonate saturation and immiscibility (Holm et al., 2006;Weidendorfer et al., 2016). Moreover, multistage processes involving alternating steps of crystal fractionation and carbonate-melt unmixing were advocated to account for the geochemical complexity of certain suites (Dawson, 1998;Baudouin et al., 2016). In the Ihouhaouene complex, a multistage process would account for (1) the increasing importance of carbonate rock types and carbonate equilibrium melts in the presumed order of magma differentiation, i.e., from the red syenites to the Sipoor carbonatites through the white syenites and the Si-rich carbonatites; (2) the more evolved character and wider range of equilibrium melt compositions in the white syenites and the carbonatites, compared with the red syenites; (3) the paradoxically Nb/Ta and low LREE signature of the Sipoor carbonatites, implying unmixing of carbonatitic melt from a differentiated parental melt.
However, additional complexity in the Ihouhaouene complex arises from the fact that all rock facies contain minerals crystallized from distinct melt types. Mineral assemblages from the most differentiated facies, in particular, include minerals Frontiers in Earth Science | www.frontiersin.org February 2021 | Volume 8 | Article 605116 equilibrated with both carbonate and silicate liquids. This peculiar feature is tentatively ascribed to the high thermal gradient inferred for the lower crust during emplacement of the Ihouhaouene complex, with temperatures reaching 1,000°C in the host granulites (Ouzegane et al., 2003). Magmas injected from the mantle into narrow shear zones may have remained partially molten for a long period of time, allowing intermingling and mixing of crystal mushes characterized by variable degrees of differentiation and solidification. Altogether, our observations suggest an evolutionary scheme for the Ihouhaouene alkaline complex whereby slow cooling of a silicocarbonated mantle melt resulted in the segregation of both cumulus minerals (clinopyroxene, wollastonite, apatite, etc.) and immiscible silicate and carbonate melt fractions, resulting in the differentiation of the complex. This process was however counterbalanced by intermingling of partially crystallized melt fractions. This resulted in the formation of hybrid alkaline cumulates composed of disequilibrium cumulus phases and variable proportions of carbonates or K-felspar depending on the composition of the last melt fraction that was locally present upon solidification.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, and further inquiries can be directed to the corresponding author/s.

AUTHOR CONTRIBUTIONS
AD collected the samples and carried out the microprobe, ICPMS, and LA-ICPMS analyses and defined the model for data interpretation. AD, FP, J-LB, KO, and J-MD contributed to the interpretations and writing. All authors contributed to the article and approved the submitted version.