2.1. Comoros archipelago

The Comoros Archipelago is located in the Mozambique Channel, between the northern tip of Madagascar and the Mozambican coast. This area results from the break-up of the Gondwana supercontinent ∼160 Ma ago and the southward drift of Madagascar relative to Africa (e.g., Chatterjee et al., 2013; Piqué, 1999). The archipelago (10–13°S; 43–46° E) is composed of four islands (Figure 1), from east to west: Mayotte (or Maore), Anjouan (Nzwani), Mohéli (Mwani) and Grande Comore (Ngazidja). Below sea level, the Comoros is a volcanic chain that also comprises seamounts, banks and atolls extending toward the east and up to the Cenozoic volcanoes of northern Madagascar (Rusquet, Famin, Michon, et al., 2025). This area has a seismic and volcanic activity among the most intense of the western Indian Ocean (Bachèlery, Morin, et al., 2016), the Karthala being the second most active volcano after Piton de la Fournaise (La Réunion Island).

Consensus is emerging that the Comoros Volcanic Chain is built over an oceanic basement (Phethean et al., 2016; Rolandone et al., 2022; Masquelet, Leroy, et al., 2022; Masquelet, Watremez, et al., 2024), approximately 150 Ma old, along a nascent border between the Lwandle and Somalia plates (Famin et al., 2020) also interpreted as a transfer area between the East African Rift System and Madagascar’s grabens (Feuillet et al., 2021). The volcanic ridges connecting the Comoros Islands are considered to be lithospheric fractures at the dextral transform frontier between the Lwandle and Somalia plates, controlling the volcanism (Famin et al., 2020).

The magmas of the Comoros Volcanic Chain share strong similarities in major and trace element compositions and isotopic signatures with those of the Cenozoic volcanism of northern Madagascar, suggesting similar melting sources and/or ascent processes (Cucciniello, Melluso, et al., 2011; Cucciniello, Tucker, et al., 2016; Cucciniello, Grifa, et al., 2022; Rusquet, Famin, Michon, et al., 2025). Comorian magmas have been produced by partial melting at different degrees of a mantle source composed of a mix between a HIMU (high μ = ²³⁸U/²⁰⁴Pb) and a DMM (Depleted MORB Mantle) part, as inferred from their Sr–Nd–Pb isotopic ratios and trace elements compositions (Späth et al., 1996; Pelleter et al., 2014; Bachèlery and Hémond, 2016). A possible explanation to the HIMU contribution in Comorian magmas is delamination of a continental lithosphere in the mantle induced by Gondwana breakup (Späth et al., 1996; Pelleter et al., 2014; Bachèlery and Hémond, 2016).

Mayotte is considered to be the oldest island of the Comoros, its construction on the seafloor would have started 27 Ma ago (Masquelet, Leroy, et al., 2022). Episodic subaerial volcanism in this island is documented over the past 6 Ma, with 3 different phases: 6–5 Ma and 4–3 Ma, occurred in the southern and northwestern parts, and the last one since about 2.4 Ma in the northeastern part. The subaerial volcanism of Anjouan has been recently dated between 900 ka to 11 ka (Quidelleur, Michon, et al., 2022) and the ages of Mohéli range from 3.25 Ma to 8 ka (Rusquet, Famin, Quidelleur, et al., 2023).

2.2. Grande comore

Grande Comore is an elongated island along an overall N–S axis (NW–SE in the south), with a length of 65 km and a width of around 20 km. It is composed of two main volcanic complexes: Karthala, a shield basaltic volcano forming the southern two thirds of the island, and La Grille, a shield volcano forming the northern part of the island. Bachèlery and Coudray (1993) have also identified the relict reliefs of a third eroded edifice beneath the Karthala volcano, south of Grande Comore, which they called “M’Badjini”. Whether this eroded M’Badjini relief is a third independent edifice truly older than Karthala and whether it has a distinct geochemical signature is unknown. Grande Comore is prolongated to the north by a series of submarine volcanic ridges called N’Droundé, to the west by the Vailheu Bank, an elongated reef outcropping occasionally at low tide, and finally to the southeast by the submarine volcanic ridge of Domoni, prolonging the southwest rift zone of Karthala and/or M’Badjini at sea and connecting Grande Comore and Mohéli (Tzevahirtzian et al., 2021; Thinon, Lemoine, Leroy, Paquet, et al., 2022).

The Karthala volcano, rising at 2361 m, is the highest summit of the archipelago. The Karthala volcano displays two rift zones extending from the summit caldera, one toward the north, the other toward the southeast. The steep flanks of Karthala and the presence of hummocky surfaces on the seafloor (Audru et al., 2006; Tzevahirtzian et al., 2021), notably in the east and to a lesser extent in the south of the island, are evidence of ancient landslides (e.g. Bachèlery and Coudray, 1993; Tzevahirtzian et al., 2021), as commonly observed for basaltic shield volcanoes (e.g., Moore et al., 1994). The volcanism of Karthala is only temporally constrained by a series of ¹⁴C ages near the summit between 625 and 5045 years BP (Bachèlery and Coudray, 1993), and by two K–Ar age on whole rock at 10 ± 10 ka (Hajash and Armstrong, 1972) and 130 ± 20 ka (Emerick and Duncan, 1982). The Karthala volcano is the conundrum of the controversy about the geodynamic origin of the Comorian volcanism, because it is the only edifice to show a mantle plume signature according to Sr–Nd–Pb–Os isotopes, distinct from the other magmatic products of the Comoros Volcanic Chain (Class, Goldstein, Stute, et al., 2005; Class, Goldstein and Shirey, et al., 2009; Bachèlery and Hémond, 2016).

North of Karthala, the La Grille edifice forms an elongated ridge toward the north. La Grille is covered by many large scoria cones, with alignments linked to its rift zone, and tuff-rings, remnants of violent hydromagmatic eruptions. Recent eruptions on La Grille are documented by two ¹⁴C ages of 1300 ± 65 years BP and 740 ± 130 years BP (Bachèlery and Coudray, 1993), but there is no geochronological constrain on the construction history of La Grille. The La Grille volcano has a HIMU-DMM Sr–Nd–Pb–Os isotopic signature also observed on the other islands, but at variance with the Karthala signature (Class, Goldstein, Stute, et al., 2005; Class, Goldstein and Shirey, et al., 2009; Bachèlery and Hémond, 2016; Bordenca et al., 2023).

3.1. Sampling

Ten volcanic rock samples were collected in June 2022 (Figure 2; Table 1). Samples 22CM01, 22CM02 and 22CM03 are from lava flows, southeast of the island, in an area mapped as belonging to the M’Badjini massif (Bachèlery and Coudray, 1993). Samples 22CM07 and 22CM08 are from dykes, also from the M’Badjini massif, located to the west of the above samples. Sample 22CM04 (Figure 2) is from an ankaramite lava flow (a melanocrate and porphyric clinopyroxene-dominated basalt) located in the same area. Samples 22CM05 and 22CM06, which belong to the massif of Karthala, are ankaramite lava flows located on the eroded flanks in the eastern part of Grande Comore. In the northern part of the island, 22CM09A and 22CM11 were sampled from lava flows belonging to La Grille volcano. As K–Ar dating requires a relatively large amount of fresh groundmass with low vesicularity, some of the above samples were not considered suitable for geochronological analysis. It is the case of samples 22CM04, which was too porphyric and vesiculated, and 22CM11, which showed evidence of secondary mineralization in vesicles (see Supplementary Material). The eight remaining samples underwent K–Ar dating at the GEOPS laboratory (Orsay, France).

3.2. K–Ar dating

The Cassignol–Gillot technique (Gillot, Hildenbrand, et al., 2006), described in detail along with uncertainties calculation in Germa, Quidelleur, Lahitte, et al. (2011), was applied to perform K and Ar measurements. Thin sections of each sample selected were first examined to confirm their freshness (i.e., absence of secondary phases) and to determine the groundmass fraction size (typically 125–250 μm) that should be analyzed. Then, samples were crushed and sieved in order to isolate the groundmass for dating. Samples were then washed with diluted nitric acid (10% concentrate) in an ultrasonic bath to eliminate any possible alteration. A heavy liquid separation using diiodomethane was performed to isolate the groundmass from the remaining dense phenocrysts and the light phases. When needed magnetic sorting was realized to further separate crystals from the groundmass.

The K content was measured by Atomic Absorption Spectroscopy, using an Agilent AA240 spectrometer. Solutions with K concentrations from 1 to 2 mg/l were used to calibrate measurements, then compared to references standards BCR-2 (K = 1.481%; Raczek et al., 2001) and MDO-G (K = 3.510%; Gillot, Cornette, et al., 1992). ⁴⁰K content was calculated using decay constants and K isotopic ratio of Steiger and Jäger (1977).

The sample gases were released from another aliquot of each sample using a high frequency induction furnace above 1400°C. They were purified on a titanium foam heated at 800°C, which eliminates all the non-noble gases, leaving argon nearly as the only remaining gas. The sample underwent further purification stages using Al–Zr getters. Argon measurement was performed using a 180° sector multi-collector mass spectrometer (e.g., Gillot, Hildenbrand, et al., 2006). Age determination by the Cassignol–Gillot technique relies on the comparison of argon 36 and 40 from the samples and atmospheric aliquots, with a detection limit of 0.1% for radiogenic argon (⁴⁰Ar*; Quidelleur, Gillot, et al., 2001). A calibration of the ⁴⁰Ar signal was made with an air pipette, based on repeated measurements of HD-B1 standard (Fuhrmann et al., 1987) using the age of 24.18 Ma (Schwarz and Trieloff, 2007).

Each K and Ar measurement was at least duplicated. Finally, the age was obtained by calculating a weighted mean of replicated measurements of K and Ar. Age uncertainties presented in this study are given at the 1σ level.

3.3. Major and trace element geochemistry

The ten samples underwent geochemical analyses, including whole-rock major and trace elements and REE measurements. The analyses were performed at the University of Bretagne Occidentale, in the Géosciences Océan laboratory (Brest, France), by ICP-AES (Inductively Coupled Plasma-Atomic Emission Spectrometry) for major elements and by ICP-MS (Inductively Coupled Plasma–Mass Spectrometry) for trace elements. Relative uncertainties are <2% for major elements and <5% for trace elements at 2σ levels. To compare our samples with the existing literature, geochemical compositions of rocks from all the archipelago were extracted from the GEOROC database: for Grande Comore (Reisberg et al., 1993; Späth et al., 1996; Class, Goldstein, Altherr, et al., 1998; Class, Goldstein, Stute, et al., 2005; Class, Goldstein and Shirey, et al., 2009; Deniel, 1998; Rusquet, Famin, Michon, et al., 2025; Class and Goldstein, 1997), Mayotte (Debeuf, 2004; Pelleter et al., 2014; Rusquet, Famin, Michon, et al., 2025), Mohéli (Rusquet, Famin, Quidelleur, et al., 2023; Rusquet, Famin, Michon, et al., 2025), and Anjouan (Quidelleur, Michon, et al., 2022; Rusquet, Famin, Michon, et al., 2025).

3.4. Geomorphology

A geomorphological investigation was conducted on the edifices of La Grille, Karthala and M’Badjini where the eight dated rocks were sampled (22CM01, 22CM02, 22CM03, 22CM05, 22CM06, 22CM07, 22CM08, 22CM09A). The analysis is based on the bathymetry data from Tzevahirtzian et al. (2021) and a digital elevation model derived from the ALOS World 3D–30 m program of the Japan Aerospace Exploration Agency (JAXA; Takaku et al., 2020).

4.1. K–Ar ages

The eight new K–Ar ages obtained in this study range from 0.007 ± 0.002 to 2.209 ± 0.032 Ma (Table 1 and Figure 2). The groundmass contains between 0.859% and 1.371% of potassium and between 0.3% and 38.1% of ⁴⁰Ar^∗. The ages can be divided into three stages. (1) Two ages are close to 10 ka (22CM05 and 22CM06). They are both ankaramites, from the base of the steep east flank of the Karthala massif. (2) Four ages range between 230 and 283 ka (22CM01, 22CM02, 22CM03, and 22CM09A). Three of these samples (22CM01, 22CM02, and 22CM03) were collected on the M’Badjini massif according to the map of Bachèlery and Coudray (1993), while the fourth sample (22CM09A) was taken on the lowermost, presumably oldest outcropping units of La Grille volcano. (3) Two samples (22CM07 and 22CM08) display much older ages (2.167 ± 0.031 and 2.209 ± 0.032 Ma), undistinguishable from each other at the 1σ level. Note that the mean age of sample 22CM08 (Table 1) was obtained from dating two different density fractions (3.16–3.08 and 2.93–3.08), with different K content (0.925 and 1.166%, respectively), which further reinforces the confidence in our K–Ar groundmass ages.

4.2. Major and trace elements

Samples from this study are basalts, basanites and trachybasalts (Figure 3), characterized by phenocrysts (olivine, clinopyroxene, and plagioclase) of various sizes and proportions, in crystalline to micro-crystalline groundmass (see Supplementary Material). Major and trace elements contents have been measured for the ten samples of this study and are given in Table 2 and shown in Figures 3 and 4 along with previous data. Note that all samples have losses on ignition (LOI) lower than 0.5%. Measurements are displayed in a total alkali versus silica (TAS) diagram (Figure 3), highlighting their low SiO₂ content (<48%). All the samples are relatively close to each other in this diagram. Rocks from Karthala and M’Badjini belong to the alkaline series, while samples from La Grille belong to the hyperalkaline series, with primitive magmas (Bachèlery and Hémond, 2016). Samples from La Grille fall in the field of basanites (22CM09A and 22CM11), those from M’Badjini fall in the basalt field (22CM07, 22CM08, and 22CM04), and samples from Karthala are also basalts (22CM06, 22CM01, 22CM02, and 22CM03), except for 22CM05, which falls in the trachy-basalt field. These results are consistent with previous data for Grande Comore (Hajash and Armstrong, 1972; Emerick and Duncan, 1982; Bachèlery and Coudray, 1993). There is a clear separation between the hyperalkaline tephrites of La Grille and the (trachy-) basalts from Karthala and M’Badjini, with Karthala samples slightly above M’Badjini’s in the TAS diagram, although sample 22CM06 is very close to the M’Badjini group.

Figure 4a shows spider diagrams of trace elements normalized to primitive mantle (Lyubetskaya and Korenaga, 2007). Patterns of the new data in the spider diagrams are similar to those obtained previously for Grande Comore (Reisberg et al., 1993; Späth et al., 1996; Class, Goldstein, Altherr, et al., 1998; Class, Goldstein, Stute, et al., 2005; Class, Goldstein and Shirey, et al., 2009; Deniel, 1998; Rusquet, Famin, Michon, et al., 2025; Class and Goldstein, 1997), with lower values for U, K, Pb, and P. In detail, however, samples from La Grille (22CM09A and 22CM11) are more enriched than those of Karthala (22CM01, 22CM02, 22CM03, 22CM05, and 22CM06), consistent with previous data, presenting a small enrichment in most trace elements. On the contrary, samples 22CM04, 22CM07, and 22CM08 from M’Badjini massif show values slightly depleted in most trace elements, even more than previously published data from Grande Comore. In a Rare Earth Elements (REE) diagram normalized to chondrite (Sun and Mcdonough, 1989), all samples show a much higher light-REE contents than heavy-REE contents (Figure 4b), which may indicate a low degree of partial melting (Rollinson, 1993). The distribution of samples is the same as above, with similar behavior for Karthala and La Grille and lower values for M’Badjini. Note that in major and trace elements diagrams, sample 22CM06 presents concentrations in between those of the M’Badjini and Karthala samples.

4.3. Geomorphology

Grande Comore is known to present contrasted morphologies between La Grille and Karthala volcanoes (Bachèlery, Morin, et al., 2016). La Grille is capped by numerous, large scoriascoria cones scattered on its flat summit and most of its subaerial and submarine flanks (Figure 5a). The maximum concentration of cones along a NNW–SSE axis delineates the summit rift zone of the volcano extending from the saddle between Karthala and La grille to the northern flank. The rift zone western limit corresponds to a NNE–SSW trending break in slope that separates the summit and the western flank, which is covered by few scoria cones (Figure 5a). In submarine domain, the flanks of La Grille present regular slopes on the edifice northwestern, northern and northeastern parts. Conversely, they show a concave geometry in the west (Figure 5c).

Compared to La Grille, the morphology of Karthala is marked by well-developed coalescent collapse calderas dissecting the flat summit, steep flanks with average slope of 15° (average of 8° for La Grille), and small scoria cones mostly concentrated along two rift zones north and south of the summit (Figure 5b). Although most of Karthala edifice is composed of regular homogeneous slopes, the subaerial east and southwest flanks of Karthala present obvious morphological discontinuities characterized by steep slopes of 30–40°. In the southwest, these steep slopes form two clear U-shaped scars intersecting each other, and a third arcuate structure parallel to the SE rift zone (Figure 5b). East of the summit, the flank shows two main slope breaks at high (1500–1800 m) and low (100–300 m) elevations (Figure 5d). The lower slope break is composed of a curved scarp opened toward the submarine east flank between our samples 22CM05 and 22CM06, and, south of 22CM06, a second scarp that progressively merges with the east slope of the SE rift zone (Figure 5b). Noteworthily, these scarps and the flank between the upper and lower slope breaks are incised by a network of narrow valleys (Figure 5a). East of the lower slope breaks, the edifice slope is formed by coalescent lava deltas.

The southern part of the SE rift zone, mapped as belonging to the M’Badjini massif (Bachèlery and Coudray, 1993), is characterized by contrasting flank geometries (Figure 5e). The apex and the eastern flank of the SE rift zone are covered by large scoria cones scattered on a regular slope. Conversely, the western flank presents a partially drowned dissected relief limited by steep slopes along which samples 22CM04, 22CM07, and 22CM08 were collected (Figure 5b and e).

5.1. Construction of the radiometric ages database

Besides K–Ar, other dating methods can be applied for volcanic rocks but they need specific mineral phases, such as zircon or apatite for (U–Th)/He, or charcoal for ¹⁴C, that are seldom found in basalts. The ⁴⁰Ar/³⁹Ar technique provides high-precision ages when applied to K-rich minerals, but it is challenging when used for low-K high-Ca samples. This is due to the requirement of sample irradiation prior to analyses in order to transform ³⁹K into ³⁹Ar. Unfortunately, it also produces ³⁶Ar from Ca, making it less precise for basic rocks, especially when they are young and have a high natural ³⁶Ar content. Therefore, we prefer to use the K–Ar technique applied to the groundmass for young Quaternary basaltic rocks (Quidelleur and Famin, 2024). The groundmass is preferred because, as this phase is the last to crystallize (i.e., during the eruption), it contains the largest amount of potassium, which is an incompatible element. In addition, the removal of phenocrysts prevents possible age bias due to excess argon that would be introduced by minerals crystallized prior to the eruption (Germa, Quidelleur, Shea, et al., 2023). This is the reason why K–Ar dating applied to whole rock must be considered with caution in crystal-bearing rocks, because this technique often yields “too old” ages compared to other techniques (e.g., Samper et al., 2007; Quidelleur, Michon, et al., 2022).

5.1.2. Whole comoros archipelago

Rusquet, Famin, Michon, et al. (2025) have updated the age database for the whole archipelago by evaluating the quality of each radiometric data, including whole-rock K–Ar ages. Thus, the following part is a summary of the selection carried out by these authors, but with the more severe criterion of rejecting most whole-rock ages for reasons explained above. We realize that we might reject valid ages among erroneous data, but we prefer here to build a very robust ages dataset, albeit limited, for the whole Comoros Archipelago. Mayotte Island has been subjected to the greatest number of dating, with 25 groundmass ⁴⁰Ar/³⁹Ar ages (Debeuf, 2004; Pelleter et al., 2014), 33 K–Ar whole-rock ages (Hajash and Armstrong, 1972; Emerick and Duncan, 1982; Emerick and Duncan, 1983; Nougier et al., 1986) and 4 K–Ar groundmass ages (Rusquet, Famin, Michon, et al., 2025). From these, the 4 K–Ar groundmass ages can be considered as reliable, as they were obtained after a severe selection in the field and thorough mineral preparation in the laboratory. Among the 25 groundmass ⁴⁰Ar/³⁹Ar ages, only 3 satisfy the quality criterions of this technique, namely presenting a plateau age or a linear isochron with three continuous steps. Thus, only 7 ages remain for Mayotte (Figure 6). 8 K–Ar whole-rock ages are available for Anjouan (Hajash and Armstrong, 1972; Emerick and Duncan, 1982; Emerick and Duncan, 1983; Nougier et al., 1986), but are not further considered here for the above reasons. Recently, 13 K–Ar groundmass ages and one ¹⁴C age (Quidelleur, Michon, et al., 2022) completed the database. These 14 remaining ages make Anjouan the island with the highest number of reliable ages for the Comoros Archipelago. The island of Mohéli has 10 K–Ar groundmass ages (Rusquet, Famin, Quidelleur, et al., 2023), all considered reliable. Only 2 K–Ar whole-rock ages and 2 very young ¹⁴C ages were available for Grande Comore before this study (Hajash and Armstrong, 1972; Emerick and Duncan, 1982; Emerick and Duncan, 1983; Nougier et al., 1986), making it the least geochronologically constrained island of the Comoros. The island has now 8 reliable K–Ar groundmass ages available (Figure 6). Recently, submarine volcanic edifices have been identified during oceanic campaigns (Thinon, Lemoine, Leroy, Ali, et al., 2021; Thinon, Lemoine, Leroy, Paquet, et al., 2022; Berthod, Médard, et al., 2021), some of which have been dated, making them of great importance to understand the whole Comorian volcanism. It is the case for the Domoni Ridge, between Grande Comore and Mohéli, dated by Rusquet, Famin, Michon, et al. (2025), with two ⁴⁰Ar/³⁹Ar groundmass plateau ages around 1.23 and 1.43 Ma. One K–Ar groundmass age has been obtained for the Chistwani Ridge (Rusquet, Famin, Quidelleur, et al., 2023). Submarine samples have also been dated on Mwezi (two K–Ar groundmass ages), Jumelles (two K–Ar groundmass ages and one ⁴⁰Ar/³⁹Ar groundmass age) and Anjouan (one ⁴⁰Ar/³⁹Ar groundmass age) (Rusquet, Famin, Michon, et al., 2025). All ages from our selection are shown in Figure 6.

5.2. Volcanic history of Grande Comore

As presented above, the volcanism of Grande Comore presents three main distinct stages, each evidenced by the age relative probability plot of Deino and Potts (1992, Figure 7). The most recent phase is of Holocene age, with 3 ages around 10 ka for the Karthala massif (Hajash and Armstrong, 1972; samples 22CM05 and 22CM06 of this study) and the 12 very recent ages around <5 ka for La Grille and Karthala volcanoes (¹⁴C ages). These data shows that both volcanoes are still active, with the last eruption of Karthala in 2007 AD. Our results show that a widespread volcanic activity was present on the island between around 130 ka and 280 ka. Note that ages from our study show a coeval period of activity for Karthala (samples 22CM01, 22CM02, and 22CM03) and La Grille (sample 22CM09A) during this interval. No subaerial activity has been recorded in our study and earlier works between 280 ka and 2.2 Ma, which could represent a period of volcanic quiescence of nearly 2 Ma for Grande Comore. However, the ages between 1.2 and 1.4 Ma obtained for the N155° E Domoni Ridge (Rusquet, Famin, Michon, et al., 2025) could be related to the volcanism of Grande Comore, as suggested by the trace element diagrams (Figure 4a and b), and therefore would question the hypothesis of volcanic quiescence during this interval. Our study is the first to report a phase of volcanism around 2.2 Ma on Grande Comore (Table 1). The break in slope in the geomorphology of the southern area (Figure 5b and e) suggests that our two samples 22CM07 and 22CM08 were collected on the edges of an ancient structure now eroded and partly covered by younger volcanic units. These two samples are dykes belonging to a series of intrusions oriented N120° E and following the break in slope. These pieces of evidence further corroborate the existence of a M’Badjini edifice as identified on the geologic map (Bachèlery and Coudray, 1993), preexisting to the Karthala and covered by it.

Major and trace elements geochemistry also show a clear difference between samples from La Grille and Karthala and samples from M’Badjini, the latter being less enriched in incompatible elements than the others (Figure 4). Olivine and clinopyroxene enrichment may partly explain this relative depletion in two samples (22CM04 and 22CM07) but not in the third one (22CM08). These results strengthen the idea that M’Badjini and Karthala are two distinct massifs, erupting at a different time with a different chemistry. Finally, the submarine continental shelf prolongating the south of the M’Badjini is the most developed one of the whole island (Figure 1; Tzevahirtzian et al., 2021), which may represent the continuation of this old massif completely eroded today. These arguments support the existence of an old, eroded massif in the southern Grande Comore, which is clearly distinct from the Karthala massif.

In contrast, our three samples (22CM01, 22CM02, and 22CM03) were taken from volcanic units mapped as part of the M’Badjini massif (Bachèlery and Coudray, 1993). The ages and chemistry of these samples, and the absence of any topographic difference with the modern Karthala (Figure 5e), show that they in fact belong to the Karthala massif. This result implies that the history of the modern Karthala volcano may date back to at least 280 ka, the modern subaerial edifice being built on the remnants of a much older eroded volcano, emerged >2.2 Ma ago. A consequence of our geochronological results is that some samples collected by previous workers on the southeastern slope of Grande Comore and purported representative of the M’Badjini massif (Class, Goldstein, Altherr, et al., 1998; Class, Goldstein, Stute, et al., 2005), in fact belong to the 130 ka and 230–280 ka units of Karthala. It is thus very likely that the true M’Badjini edifice has not yet been subjected to any detailed geochemical investigation, as inferred from the range of values shown in Figure 4.

The evolution of the La/Sm ratio as a function of the age is given in Figure 8a for Grande Comore. It presents a slight increase of the incompatible element concentration ratios with time, which might be interpreted as a decrease in the degree of partial melting of the lithosphere with time. This situation is reminiscent of Mohéli’s geochemical evolution, also characterized by gradual increases in incompatible element ratios such as Ba/Ti and La/Yb within the last ∼3.8 Ma, and interpreted as decreasing partial melting (Rusquet, Famin, Quidelleur, et al., 2023). Figure 8b shows that such overall increase is also observed for Mayotte and Anjouan islands. Whatever the cause of this temporal geochemical evolution, it is thus a common pattern of Grande Comore and other islands from the Comoros archipelago.

Finally, Grande Comore seems to have had a discontinuous volcanic activity. First an old volcanic phase at more than 2 Ma built the M’Badjini edifice, in agreement with the onset of Grande Comore volcanism inferred from seismic data at 2 ± 1 Ma (Masquelet, Sauter, et al., 2025, accepted). This volcano likely experienced a series of dismantling processes to make it a remnant subsequently drowned by lava flows of the 200–300 ka volcanic phase. Karthala and La Grille volcanoes likely built during this latter period as suggested by widespread ages. Whether La Grille experienced a westward flank destabilization after this period of activity to explain the morphological discontinuity and the concave geometry of the submarine flank is still unclear (Figure 5c). In contrast, the obvious scars on the Karthala’s east and southwest flanks indicate that several successive flank failures occurred after this building period. The presence of a network of narrow valleys in the east flank points to a lull of activity after the flank failure long enough to enable fluvial incision. Sporadic eruptions may have occurred during this period as suggested by the 130 ± 20 ka age along the SE rift zone. The activity definitely resumed in the last 15 ka in the two massifs of Karthala and La Grille. Radiometric ages reveal that eruptions fed lava flows channelized down to the seashore on the flanks of Karthala and built scoria cones of the summit of La Grille.

5.3. Volcanic evolution of the Comoros Archipelago

Relative ages probability plots (Deino and Potts, 1992) obtained for Grande Comore are shown in Figure 9 along with the literature data of the Comoros Archipelago selected with the criteria discussed above. The close-up for the last 1 Myr allows us to clearly distinguish several phases of Comorian volcanism. The recent phase starting at ∼10 ka described for Grande Comore matches the recent volcanism of the Comoros Archipelago, with volcanism up to ∼30 ka on Mohéli and Anjouan, and the rather recent activity suggested for Petite Terre at Mayotte (e.g., Lacombe et al., 2024). The period without volcanism evidenced on Grande Comore between ∼50 and ∼130 ka is also observed for all the other Comoros islands. Then, between ∼200 ka and ∼300 ka, the volcanic activity on Grande Comore occurs concomitantly with volcanic activity on Mohéli and Anjouan, as well as submarine volcanism. The archipelago undergoes several other distinct periods of volcanism: a submarine episode at ∼400 ka, activity on Mohéli and Anjouan around ∼530 ka, and then a more diffuse stage nearly uninterrupted of volcanism from ∼700 ka to ∼1.5 Ma on Mayotte, Anjouan, Mohéli and submarine volcanism. The M’Badjini stage identified in Grande Comore at ∼2.2 Ma slightly precedes the ∼1.8 Ma volcanism on Mohéli. Finally, a volcanic episode occurs between ∼3.2–3.5 Ma on Mohéli and Mayotte, and another one is recorded in Mayotte at ∼5.9 Ma. Altogether, these ages show that the emersion of Grande Comore is older than that of Anjouan, and thus that Grande Comore is not the youngest island of the Comoros Archipelago as previously thought.

We note that Mayotte and Mohéli islands have carbonate platforms developed on edifice slopes aged of less than 2 Ma (Rusquet, Famin, Quidelleur, et al., 2023; Rusquet, Famin, Michon, et al., 2025), whereas Anjouan and Grande Comore are nearly devoid of shelves despite being of similar age. Mayotte and Mohéli also differ from Anjouan and Grande Comore by showing evidence of subsidence whereas the two later islands do not (Debeuf, 2004; Tzevahirtzian et al., 2021; Rusquet, Famin, Quidelleur, et al., 2023; Rusquet, Famin, Michon, et al., 2025). This comparison suggests that subsidence is the primary factor of coral growth and shelf development around volcanic islands, whereas activity rate and age of the volcano is only a secondary factor. Thus, we think that Anjouan and Grande Comore did not develop a large carbonate platform because they never subsided, or if they did, the eroded reliefs were too small to allow a carbonate platform to develop.

5.4. Tectonic evolution of the Comoros Archipelago

An element of answer for the origin of the Comorian volcanism could be found in the repartition of the volcanism in the archipelago over time. In Figure 6, samples with similar age ranges are represented together on a line. This enables us to distinguish several alignments of volcanic periods. Volcanism older than 3 Ma is present on Mohéli and Mayotte. The south of Grande Comore, Anjouan, and the Jumelles Ridges have their oldest volcanism ranging from 2 to 1 Ma. The northern part of Grande Comore and the Mwezi Volcanic Province have a volcanism younger than 200–300 ka. These observations could imply that the volcanism is propagating northward in the Comoros Archipelago, resulting in a presently wider active volcanic area than it was before 3 Ma.

Recently, new hypotheses have been raised to link volcanism with a transtensive (Feuillet et al., 2021) or strike-slip (Famin et al., 2020) tectonic context, resulting in fractures and faults between Somali and Lwandle plates. In both cases, there could be a reactivation of deep lithospheric fractures (Nougier et al., 1986; Feuillet et al., 2021; Boymond et al., 2025), facilitating the magma ascent from the base of the lithosphere to the surface, possibly in link with changes in motion of the Somalia/Lwandle/Rovuma plate system (Rusquet, Famin, Michon, et al., 2025). Our results suggest that the Comorian volcanism is widening with time as in volcanic zones caused by lithospheric thinning (Figure 6), which is an additional argument supporting these hypotheses.