Whole-rock major and some trace elements were measured on the Horiba-Jobin-Yvon^® Ultima 2 ICP-AES at the IUEM (European Institute for Marine Studies, Pôle de Spectrométrie Océan, Brest, France). A description of the analytical procedure is given in Cotten et al. (1995). Elements were determined from an H₃BO₃ solution, boron being used as internal standard for ICP-AES analysis. For major elements, relative standard deviation is 1% for SiO₂ and 2% for the other major elements, except for low values (< 0.50 wt.%), for which the absolute standard deviation is ± 0.01 wt.%.

Trace element concentrations were measured with a Thermo Element 2 HR-ICP-MS in Brest (France), after a repeated HF-HClO₄ digestion, and HNO₃ dilutions. Details of the analytical procedure are given in Li and Lee (2006). The repeated analysis of the international standard BCR2 demonstrated an external reproducibility better than 5–10% depending on the element and concentration. Major and trace element data are presented in Table 1.

Pb isotope compositions were obtained from single HF-HNO3 dissolutions of 500 mg of sample and analysed using the Thermo Neptune MC–ICP–MS of Ifremer-Brest. 2σ internal errors never exceeded the last significant digit (1E-5) in each analysis.

Sr–REE dried fractions, previously collected at the beginning of Pb column separation, were dissolved and dried two times with HNO₃ at 95̊C until complete dryness. Dry samples were dissolved again in HCl prior to elution. Chemical separation for Sr and REE was performed on cationic DOWEX^® AG50X8 200–400 mesh columns. Sr was kept and proceeded another time in the same column to efficiently separate Sr from Rb and Ca. Nd was further eluted on LnSpec Eichrom resin. Isotopic measurements were performed on a Thermo Scientific Triton at the IUEM in Brest (France). Sr was run on a single W filament with Ta activator, while Nd was run on a Re double filament. International standards were run regularly to check our measurements:

• • Sr: NBS987, average value ⁸⁷Sr/⁸⁶Sr = 0.710262 ± 12 (2σ, n = 10);
• • Nd: JNdi-1, average value ¹⁴³Nd/¹⁴⁴Nd = 0.512106 ± 14 (2σ, n = 6);
• • La Jolla, average value ¹⁴³Nd/¹⁴⁴Nd = 0.511848 ± 12 (2σ, n = 3).

Blanks were very low and considered negligible. According to previously published geochronological data and to observations on the field, the Sr and Nd isotopic data on basalts and dacites have been recalculated to 45 Ma initial ratios while the basanites data have not been changed as these rocks are very young. Measured and initial isotopic ratios are presented in Table 1.

In the Na₂O + K₂O vs. SiO₂ TAS classification diagram (Fig. S1, Le Bas et al., 1986), the Bamoun samples define three groups of rocks. Three samples (ZB08, ZB09 and ZB10) are basanites with olivine and nepheline normative composition. A second group is made of basalts with few samples plotting slightly in the basaltic andesite field. All these samples will be referred to as basalts. They are mostly sub-alkaline and contain quartz and hyperstene in their normative composition (except ZB01 and ZB101). The third group comprises two alkali-poor dacites (ZB68 and ZB69).

The basalts have a fairly restricted range in major element composition, with SiO₂ ranging from 47.9 to 52.8 wt% and MgO ranging from 2.8 to 5.1 wt%. The basanites have very different major element compositions, with lower SiO₂, Al₂O₃, and TiO₂, and higher CaO, MgO, P₂O₅, and Na₂O contents than the basalts. Similar chemical compositions were already noticed by Moundi et al. (1996) and Okomo Atouba et al. (2016). As expected, the two dacites have high SiO₂ and K₂O contents, but low CaO, MgO, and TiO₂ concentrations.

Chondrite-normalized extended incompatible element patterns are shown in Fig. 2. All the Bamoun samples are enriched in LREE and in incompatible elements compared to HREE, but the three groups of samples (basanites, basalts, dacites) display also clearly distinct trace element contents. The basalts have homogeneous REE and trace element compositions, with a slight enrichment in LREE over HREE, with La/Yb_N ratios ranging from 7.1 to 13. A few samples show a small positive Eu anomaly. The three basanites are very homogeneous in REE, with steeper patterns than the basalts. They have similar HREE but higher LREE concentrations, with La/Yb_N ratio = 32. Dacites have flatter REE patterns with a higher HREE content than the mafic rocks, but an LREE content intermediate between the basalts and the basanites. Their La/Yb_N ratio ranges from 7.6 to 9.4, and they have a slight negative Eu anomaly.

When considering all the trace elements, the basanites confirm that they are highly enriched in incompatible elements, with very high Th, U, Nb, Ta, and Sr contents as well as higher Zr content than the basalts for similar Hf content. The basalts have low Th and U concentrations and very variable Rb and Ba contents. The dacites have a contrasted trace element pattern, with the highest Ba, Zr and Hf concentrations, but lower Th, U, Nb, and Ta contents than the basanites. They have a very pronounced Sr negative anomaly, probably reflecting some amount of plagioclase fractionation.

From the morphology observed on the field, it is clear that the basanites represent a very recent volcanic episode, while the basalts and the dacites were emplaced long time ago. In agreement with the K–Ar ages previously reported for the old volcanic episodes in the area (Moundi et al., 2007; Okomo Atouba et al., 2016), we calculated initial Sr and Nd isotopic compositions back to 45 Ma for the basalts and the dacites. The Bamoun samples cover a large range in Sr and Nd compositions (Fig. 3), with ⁸⁷Sr/⁸⁶Sr_i ranging from 0.7034 to 0.7063 and ¹⁴³Nd/¹⁴⁴Nd_i from 0.51225 to 0.51281 for the basalts. The basanites have very restricted isotopic compositions with ⁸⁷Sr/⁸⁶Sr ranging from 0.70352 to 0.70356 and ¹⁴³Nd/¹⁴⁴Nd from 0.512885 to 0.512892. The analyzed dacite has a high ⁸⁷Sr/⁸⁶Sr_i ratio (0.70555) and ¹⁴³Nd/¹⁴⁴Nd_i intermediate (0.512512) between the most and the less radiogenic basalts.

Similar observations can be made for the Pb isotopic ratios, with the highest and very restricted isotopic ratios for the basanites (for example ²⁰⁶Pb/²⁰⁴Pb from 19.812 to 19.828), while the basalts cover a larger range of isotopic compositions, with ²⁰⁶Pb/²⁰⁴Pb ranging from 16.694 to 19.631 (Fig. 6). As for Nd, the dacite has Pb isotopic composition intermediate between the less and the most radiogenic basalts, with ²⁰⁶Pb/²⁰⁴Pb = 18.031.

The three basanites are very rich in incompatible trace elements, and are very difficult to contaminate by crustal products. They have very low silica content (SiO₂ < 44 wt%), which makes unlikely their interaction with high silica crustal melts, in agreement with their very unradiogenic Sr isotopic composition. At the opposite, the basalts have higher silica content and a large range in isotopic compositions. Their SiO₂ content is strongly correlated with Sr and Nd isotopic ratios, with increasing Sr (Fig. S2) and decreasing Nd isotopic values for increasing SiO₂. These correlations are clear indicator that, during their differentiation, the mafic magmas have interacted with high silica rocks with more radiogenic Sr and less radiogenic Nd compositions, similar to what expected from rocks from the continental crust.

Nb is highly depleted in all continental crust rocks (Taylor and McLennan, 1985) and the La/Nb ratio is often used as an indicator of crustal contamination. La/Nb ratio is correlated with Sr, Nd and Pb isotopic ratios for the Bamoun basalts and this indicates interactions with a crustal component with low Pb isotopic ratios. However, some basalts (ZB01, ZB07 and ZB101) have isotopic composition different from the basanites, but with similar and low La/Nb ratio. For these samples, their different isotopic composition reflects differences in the composition of the mantle source. A detailed study of this contamination process shows that it affected selectively some elements like Ba and Nb, but Th and the REE were not modified by the contamination. The contaminated samples will be identified with special symbols in all the isotopic diagrams discussing the origin of the Bamoun volcanic rocks.

In the Ni vs. Th diagram (Fig. 4), all the analyzed mafic samples have Ni concentrations lower than 120 ppm, and are clearly not primary magmas in equilibrium with mantle minerals. Two trends compatible with olivine fractionation are visible, one with Th < 4 ppm with all the basalts and some already published data from the Bamoun area, and another one with higher Th content, with the basanites, a few published data from the literature, and overlapping with the data from the Mount Cameroon lavas. These two trends point towards two different chemical compositions for their primary magma, one with low Th concentration, probably lower than 2 ppm, and the other one with a much more enriched composition.

The Th/La ratio is strongly correlated with the different isotope ratios and, when plotted against ²⁰⁶Pb/²⁰⁴Pb, two different groups defined two parallel correlations (Fig. 5). One group comprises all the basalts but three and the dacite, and defines a positive correlation on the low Pb isotopic ratio of the diagram. The second group defines a nearly parallel positive correlation, but with higher Pb isotopic ratios, and comprises the basanites and the three basalts ZB01, ZB07, and ZB101, for which no or negligible crustal contamination has been observed. Samples from Mount Cameroon extend this trend towards higher values of ²⁰⁶Pb/²⁰⁴Pb. Clearly most of the basalts as well as the dacites have been shifted from the main trend towards low Pb isotopic values for similar Th/La ratio during the crustal contamination process. These correlations clearly highlight a mixing between at least two different mantle sources beneath the CVL. One source has high Pb and low Sr isotopic ratios. On the other side of this correlation, the second mantle source has low Pb isotopic composition, higher Sr isotopic ratio, and low Th content. The dacites plot on the same correlation as the contaminated basalts and have similar La/Yb ratios, indicating that they were formed by crystal fractionation and assimilation from the basaltic magmas.

Large differences in the La/Yb ratio between the basalts, the basanites and the Mount Cameroon samples can be ascribed either to different partial melting rates or to mantle source heterogeneities. When combined with Th/La, these two trace element ratios provide key information about the mantle processes at the origin of the CVL mafic magmas (Fig. S3). The whole data for the mafic lavas from the CVL cover a quite restricted range of Th/La ratios, with values ranging from 0.8 to 0.14 for most of the samples. This small range is quite surprising as the plotted data represent a large number of mafic rocks covering the whole continental part of the CVL with magmas emplaced from more than 40 Ma ago to the present-day Mount Cameroon eruptions. It implies that from the point of view of trace elements, source heterogeneities are rather limited in the mantle beneath the CVL. The basalts from the Bamoun plateau have among the lowest Th/La ratios of the CVL, while the basanites plot at the middle of the range. Variations of the La/Yb ratio are larger, with values ranging from less than 10 up to 47. Most of these variations can be ascribed to various partial melting degrees, even if source heterogeneities can also add to this variability. Interestingly, the mafic rocks of the Bamoun plateau have the most extreme La/Yb ratios, with basalts having whole CVL values among the lowest ones, while basanites have the highest La/Yb ratios. Clearly, the degree of partial melting has changed dramatically beneath this area of the CVL from 45 Ma to the present day. This is also well illustrated in a zoom of the mafic part of the classical TAS diagram (Fig. S4). A large majority of the mafic samples from the CVL plot on the alkali part of the diagram, above the dividing lines defined by Irvine and Baragar (1971) and Bellieni et al. (1982). Moundi et al. (1996) already analyzed some mafic rocks from the Bamoun plateau with transitional affinity, and this is well confirmed by our new data. The mafic rocks from the Bamoun plateau are once more divided into two extreme groups, with basanites having the more alkali composition among the rocks of the CVL, while basalts plot on the lower right of the diagram, entirely into the transitional and sub-alkaline domains. Considering that the basalts are clearly older than the basanites, the melting regime in the mantle changed dramatically beneath the Bamoun plateau to form transitional basalts older than 40 Ma old and more alkali-rich basanites, more recently.

The isotopic ratios of mafic rocks from the CVL cover a limited range of compositions in Sr, Nd, and Pb, and part of the variability is probably due to small amounts of crustal assimilation. Our new Bamoun data plot in the middle of this range (Fig. 6 and Fig. S5), except for some samples with very low Pb isotopic ratios, which have been affected by crustal contamination, as already shown in the same area by Moundi et al. (2009) and Okomo Atouba et al. (2016). Interestingly, the Mount Cameroon mafic rocks have isotopic compositions that do not overlap with those of other CVL samples, with higher Pb isotopic ratios. These Mount Cameroon compositions are also different from those of the mafic rocks from Saint-Helena Island, and this observation is against some ideas that the source of the CVL magmas is similar to the mantle plume at the origin of this island and represents the old track of the Saint-Helena mantle plume. In the Pb isotopic spaces, where mixing between different mantle sources defines straight lines, the non-contaminated Bamoun samples define a trend compatible with a mixing between two different sources, one with low Pb isotopic compositions, but also low Nd and high Sr isotopic ratios and pointing towards FOZO and EMI compositions, and a second one with higher Pb isotopic composition pointing towards a HIMU mantle reservoir. Two important observations can be added for the understanding of the origin and nature of the mantle sources beneath the CVL:

• • the isotopic trends defined by the Bamoun mafic rocks, but also by the majority of the CVL mafic rocks, do not point towards the MORB compositions and the DMM mantle component. The asthenospheric mantle is not (or very slightly) participating to the genesis of the magmas beneath the CVL;
• • towards high Pb isotopic values, the isotopic trends defined by the Bamoun mafic rocks do no point towards Saint-Helena basaltic compositions, nor to Mt Cameroon isotopic values.

This confirms that the magmatism of the CVL is not related to the Saint-Helena hotspot track, but also that Mount Cameroon has a peculiar mantle source composition, which is not an end-member of the mixing trend defined by the whole CVL magmas.

The isotopic trend shown by our new Bamoun data, and the already published data from this area, are consistent with a two end-member mixing between two different mantle sources with very contrasted compositions. These two sources are located in the subcontinental lithospheric mantle, and the participation of the asthenosphere is clearly very minor. Okomo Atouba et al. (2016) showed that the EMI-type mantle source produces some basalts with a clear positive Eu anomaly and interpreted this feature as an evidence for the presence of pyroxenitic lithologies in this part of the mantle. This kind of small positive anomaly is present in some of our new Bamoun basalts (e.g., ZB51, ZB60, ZB62) and confirms that pyroxenites probably participated in the genesis of some magmas beneath the CVL. These basalts also have the lowest La/Yb ratios and accordingly are generated by the highest partial melting degrees, which is consistent with the lower melting temperature for pyroxenite lithologies (Hirschmann and Stolper, 1996). On the opposite, the basanites with the highest Pb isotopic ratios point toward the HIMU-like mantle component, and have the highest La/Yb ratios and are also highly enriched in the most incompatible elements (Fig. 2). They are generated in a different mantle source, highly enriched in incompatible elements, and by lower partial melting degrees. However, while having high Pb isotopic composition, their mantle source is different from the Mount Cameroon mantle source, and they are formed by even lower partial melting degrees than the Mount Cameroon basalts.