The Comoros archipelago is located in the Indian Ocean, more precisely in the Mozambique Channel, halfway between Madagascar and Africa [10] . From north to south and west to east, it comprises four (04) islands ( Figure 1 ): Grande Comore; Mohéli; Anjouan and Mayotte ( Figure 1 ).

The Comoros archipelago was formed in the Cenozoic, following the detachment of Africa in the Lower Jurassic [13] . These islands were formed as a result of age-dependent progressive volcanism from the ESE to the WNW of the archipelago [3] [7] [14] . These authors deduce that the oldest volcanism began in the Cretaceous-Miocene, after volcanism had migrated westwards to form the four islands. Thus, Maoré (Mayotte) existed in the Mio-Pliocene, Anjouan and Mohéli in the Pliocene [13] and Grande Comore or Ngazidja in the Quaternary [13] .

The Mozambique Channel, which lies between the relatively steep and narrow continental slopes of the African continent and the island of Madagascar, began its formation during the break-up of Gondwana and continues to the actual [10] and [15] :

The island of Grande Comore has been subdivided into three (03) massifs based on geomorphological considerations [18] : Grille in the north, Karthala in the center and Badjini in the extreme south ( Figure 2 ).

However, it should be noted that Grande Comore is made up of two (02) large shield volcanoes: the Grill massif to the north and the Karthala massif (the center and south of the island) [19] . The latter, the site of our study area, is a large active volcano with a poly-lobed summit caldera whose morphology is marked by the presence of two craters: the Chungmu Chahalé crater in the center and the Chagnoumeni crater in the north, a few hundred metres from the Itsandra gate. Major volcanic eruptions have produced lava flows that have emerged either from the summit crater or, more often, from lateral fissures marked by a large number of small adventitious strombolian cones [13] .

The formations of the Karthala massif have been subdivided by [18] into three (03) volcanic units according to the presence or absence of surface flow structures (see Figure 3 below).

Karthala (K1, K2 and K3), devoid of vegetation and with the marked surface characteristics of well-preserved flows.

Lithological and petrographic studies were carried out mainly on the three volcanic units of the Karthala massif (ancient, recent and actual), the sampling points of which are located on the volcanotectonic map of Grande Comore ( Figure 3 ).

The lithological study shows that the Karthala massif is characterized by an eruptive dynamism that gives rise to flows of different shapes divided into three units:

Porphyritic vesicular basalts have been subdivided into two types according to the abundance and nature of the phenocrysts: vesicular basalts with clinopyroxene and orthopyroxene phenocrysts and vesicular basalts with olivine phenocrysts.

Microscopically, both types of basalt show the same vacuolated porphyritic microlitic texture, composed of plagioclases, clinopyroxenes, orthopyroxenes or olivines as the case may be, opaque minerals and vesicles bathed in little or no mesostasis ( Figure 4(c) and Figure 4(d) ).

Plagioclases are very abundant (55% to 65%) and occur as microlites.

Clinopyroxenes (30% to 35%), sometimes macerated, occur as automorphic to subautomorphic phenocrysts (0.4 to 1.2 mm) and small crystals.

Orthopyroxenes are rarely found (1% to 2%) as automorphic to subautomorphic phenocrysts (0.3 to 1 mm), only associated with clinopyroxene phenocrysts.

Olivines (10% to 15%) occur as automorphic to subautomorphic phenocrysts (0.6 to 1.3 mm). Some individuals contain opaque mineral crystals.

The vesicles (3% to 6%) are generally rounded to subrounded and show no filling. Their walls are lined with plagioclase microlites.

Non-vesicular porphyritic basalts ( Figure 4(e) ) outcrop to the northwest of the Karthala massif, around 6 km from the Bangaani sector and 13.8 km from the Karthala summit. These basalts have a very limited thickness (20 m). Microscopically, the rock shows a porphyritic microlitic texture with phenocrysts of clinopyroxenes and olivines containing crystals of opaque minerals (2% - 5%) and a mesostasis containing micolites of plagioclase (60% - 65%) and clinopyroxenes ( Figure 4(f) ).

Oceanites occur mainly in the vicinity of the large Chahalé crater, in the form of scattered light-gray decimetric to millimetric blocks with abundant olivine phenocrysts ( Figure 4(g) ). They are embedded in a platform composed of heterogeneous elements (volcanic ash, black sand, slag and lapillis) from the Karthala volcanic eruptions. Microscopically, the rock shows a porphyritic microlitic texture composed of olivine phenocrysts (30% - 35%) and clinopyroxenes (5% - 7%), and a mesostasis composed of plagioclase microlites (55% - 60%), clinopyroxenes and opaque minerals ( Figure 4(h) ).

Ankaramites outcrop in the Ourovéni area in the extreme south of Grande Comore, 19.7 km from the Karthala volcano, in the form of massive centimetric to metric vesicular blocks. Petrographically, ankaramites have the same texture and mineralogical composition as oceanites, but are distinguished from the latter by the abundance of clinopyroxene phenocrysts ( Figure 4(i) and Figure 4(j) ). Some clinopyroxene phenocrysts include olivine crystals ( Figure 4(k) ).

Pahoehoe-type flows or corded lavas appear as pleated rope clusters ( Figure 4(l) ). They are in the form of cow dung, with small folds wound around each other on their surface to form entrainment folds.

Type aa flows have a rough, irregular surface with decimetric to metric sharp blocks colonized by lichens and moss ( Figure 4(m) ). It should be noted, however, that we were unable to make thin slides of the pahoehoe and aa-type flows due to their very brittle nature.

Basalt stalactites formed by dripping lava are very common in the study area ( Figure 4(n) ).

The geochemical study will be carried out according to the three units already identified on the Karthala massif, i.e. ancient, recent and actual Karthala. A total of 13 rock samples were selected for geochemical analysis, with concentrations of major and trace elements shown in Table 1 . Of these samples, 2 belong to ancient Karthala, including 1 porphyritic vesicular basalt and 1 porphyritic non-vesicular basalt, 5 to recent Karthala, including 1 porphyritic vesicular basalt, 1 porphyritic non-vesicular basalt, 1 ankaramite and 2 oceanites, and 6 to actual Karthala, including 4 porphyritic vesicular basalts and 2 porphyritic non-vesicular basalts. It should be noted, however, that all these samples have very low losses on ignition (LOI) ranging from 0.16% to 0.91%, suggesting that chemical elements such as LILE and HFSE are very little affected by post-magmatic phenomena such as metamorphism, hydrothermalism, seawater, tectonics and oxidation. Therefore, the geochemical characters of all rock samples will be examined in this section. In addition, REEs are normalized to the NWA974 chondrites of [20] , while incompatible trace elements are normalized to the primitive mantle of [21] .

Plotting the geochemical data of the Karthala massif rocks on the Zr/TiO₂ versus Nb/Y diagrams of [22] and Zr/Ti versus Nb/Y diagram of [22] modified by [23] shows that all samples have an alkaline basalt composition ( Figure 5(a) and Figure 5(b) ). However, the ancient Karthala basalts have higher contents of major elements such as SiO₂ (48% - 48.7% vs. 46.4% - 47.9%), TiO₂ (2.6% vs. 1.6% - 2.5%), Na₂O (3.1% vs. 1.8% - 3.1%) and K₂O (1.2 vs. 0.6% - 1.1%) and incompatible trace elements such as Y (25.8% - 26.1% vs. 16.5% - 23.9%) and Yb (2% vs. 1%, 3% - 1.8%) but lower values in Mg# (50.2% - 52.1% vs. 53.6% - 72.8%) and compatible trace elements such as Ni (90 - 140 ppm vs. 140 - 550 ppm), Cr (70 - 130 ppm vs. 150 - 930 ppm) and Co (42 - 44 ppm vs. 47% - 68%) than basalts, ankaramite and oceanites from recent Karthala ( Figure 6 ). These geochemical characteristics suggest that the magma of ancient Karthala is more evolved than that of recent Karthala. In addition, the basalts of actual Karthala have more variable chemical compositions (SiO₂ = 46.1% - 48.5%; TiO₂ = 1.9% - 2.7%; Al₂O₃ = 10.7% - 14.9%; Na₂O = 2.4% - 3.4%; K₂O = 0.9% - 1.3%; Ni = 70 - 550 ppm; Cr = 30 - 740 ppm; Co = 39 - 68 ppm; La = 27.7 - 40 ppm; Y = 18.9 - 27.3 ppm and Yb = 1.5 - 2.3 ppm) that overlap those of ancient and recent Karthala ( Figure 6 ). Finally, the chemical compositions of Karthala massif rocks (ancient, recent and actual) show with Mg# (Mg# = ((MgO/40)/(MgO/40) + (Fe₂O₃/72)) × 100) used as an index of magmatic differentiation good negative correlations for SiO₂, TiO₂, Al₂O₃, Na₂O, K₂O, P₂O₅, La, Y and Yb and good positive correlations for Ni, Cr and Co ( Figure 6 ). These geochemical trends are consistent with the successive crystallization of olivines, pyroxenes, plagioclases and ferrotitanium oxides.

Early Karthala alkaline basalts with ΣREE ranging between 960 and 1011 ppm have highly enriched LREEs of 216 to 236 times chondrites and highly depleted HREEs of 12.6 to 12.9 times chondrites with highly fractionated spectra (LaN/YbN = 16.1 - 17.2) and very weak negative europium anomalies (Eu/Eu* =

0.83 - 0.84) ( Figure 7(a) ). Alkaline basalts from recent Karthala (ΣREE = 816 - 938 ppm) also show very marked REE enrichment and fractionation (187 to 220 times chondrites in LREE and LaN/YbN = 17.5 - 17.9) with similarly low europium anomalies (Eu/Eu* = 0.87 - 0.88) ( Figure 7(a) ). It should be noted, however, that recent Karthala alkaline basalts are less enriched in REE, with slightly steeper spectra and less pronounced negative europium anomalies than ancient Karthala alkaline basalts ( Figure 7(a) ). Recent Karthala ankaramite (ΣREE = 581 ppm) shows a highly fractionated REE spectrum marked by a strong LREE enrichment of 129 times the chondrites and a clear HREE depletion of 8 times the chondrites with a very steep spectrum (LaN/YbN = 15) and a very weak negative europium anomaly (Eu/Eu* = 0.9) ( Figure 7(a) ). Recent Karthala oceanites (ΣREE = 799 - 809 ppm) also show very marked enrichment (189 times chondrites in LREE) and fractionation of REE spectra (LaN/YbN = 17.2 - 17.6) and very slight negative europium anomalies (Eu/Eu* = 0.87 - 0.9) ( Figure 7(a) ). Compared with the alkaline basalts and ankaramite of recent Karthala, these oceanites have REEs

intermediate between the two ( Figure 7(a) ). Finally, alkaline basalts from actual Karthala (ΣREE = 749 - 1061 ppm) are slightly more enriched in REE (compared with chondrites 175 to 253 times in LREE and 10 to 14 times in HREE) with spectral fractionation (LaN/YbN = 15.5 - 19) and negative europium anomalies (Eu/Eu* = 0.83 - 0.89) similar to alkaline basalts from ancient and recent Karthala ( Figure 7(a) ).

On the expanded incompatible trace element diagrams normalized to the primitive mantle ( Figure 7(b) ), all samples from ancient, recent and actual Karthala show positive anomalies in Rb, Ba, Ta, Nb and Nd, and negative anomalies in Th, U, Sr and Zr. Pb, on the other hand, shows negative anomalies, with the exception of three samples with positive anomalies (samples: océanite A5 from recent Karthala; ankaramite A21 from recent Karthala and alkaline basalt A14 from actual Karthala).

In this section, we will use all the samples from the Karthala massif, which are supposed to be representative of magmatic liquids. In the Sm/Yb (ppm) vs. La/Yb (ppm) diagram by [24] ( Figure 8(a) ), Karthala massif lavas lie in the enriched mantle domain and correlate perfectly with asthenospheric mantle melting curves in garnet lherzolite facies. This result is confirmed by the TiO₂/Yb (ppm) versus Nb/Yb (ppm) diagram [25] ( Figure 8(b) ) proposed for estimating the source depth of lavas from oceanic domains not influenced by subduction. In the latter diagram, which confirms the alkaline nature of the rocks, we observe that all lavas lie within the garnet stability field, suggesting that the source of the magma would be very deep garnet lherzolite. The result is also consistent with the very strong

depletion of HREEs. However, the heterogeneity of the chemical composition of the magmas that generated the alkaline rocks of ancient, recent and actual Karthala can be explained by a single mantle source, but resulting either from variable rates of partial melting, or from the recycling of subducted oceanic crust residues into the deep mantle [26] . The variable degree of partial melting has been shown to be a function of the LREE/HREE ratio, which can be >1 or <1. Thus, a low partial melting rate gives a LREE/HREE ratio >1, while a high partial melting rate gives a LREE/HREE ratio <1. All the samples from the Karthala massif (ancient, recent and actual) have LREE/HREE (La/Lu) ratios >1. This implies a low partial melting rate for the magmas from which these rocks originate, and this result was confirmed by the Sm/Yb (ppm) vs La/Yb (ppm) diagram [24] ( Figure 8(a) ), which places the rocks from the Karthala massif between 1% and 0.1% partial melting rate. However, the almost identical LREE/HREE ratios between ancient, recent and actual Karthala rocks do not reflect variations in partial melting rates. The differences in chemical composition between ancient, recent and present-day Karthala alkaline rocks are therefore likely to have resulted from the recycling of subducted oceanic crustal residues into the deep mantle, and the interpretation lies in the presence of positive Nb and negative Pb anomalies in the alkaline rocks of the Karthala massif. The distance of the Comoros Islands from the mid-ocean ridge suggests an interaction between the mantle plume and a thick oceanic lithosphere, which could also explain the variability in the chemical composition of the Karthala basalts [27] .

To constrain geodynamic environments, we have used discrimination diagrams involving immobile elements such as HFSEs and REEs. Examples include the TiO₂/Yb (ppm) versus Nb/Yb (ppm) ( Figure 8(b) ) or Th/Yb versus Nb/Yb diagrams by [25] ( Figure 9(a) ) and the Nb/Y versus Zr/Y diagram by [28] ( Figure 9(b) ), where the Karthala Massif lavas are located in the Ocean Island Basalt (OIB) field. Furthermore, in [28] diagram ( Figure 9(b) ), all samples from the Karthala massif lie above the ΔNb line, indicating that the source is related to a mantle plume from the enriched lower mantle and the rocks were emplaced at a hot-spot. The oceanic island context (OIB) for the alkaline rocks of the Karthala massif is also in agreement with the highly enriched LREE spectra and the positive Nb and Ta and negative Pb anomalies (with the exception of three samples).