See Appendix A (MC7).

The Kekem dykes are porphyritic with phenocrysts of olivine and rare clinopyroxene in a subophitic groundmass (Fig. MC2) of olivine, plagioclase, clinopyroxene, and opaque oxides. Olivine phenocrysts are euhedral to subeuhedral, 1 × 1.5 mm in size. Clinopyroxene with grain size between 0.2 × 0.3 mm and 0.2 × 0.1 mm mainly occurs as subhedral to anhedral grains in the groundmass. Plagioclase occurs as euhedral to subhedral lath-shaped crystals, with size varying from 0.3 to 0.1 mm. Some plagioclase crystals bear oxide inclusions. Opaque minerals are subeuhedral to anhedral crystals frequently associated with clinopyroxene and olivine or as needle-shaped crystals within plagioclase laths. Cr-bearing spinels are enclosed in olivine phenocrysts. Alteration effects (not widespread) are shown by the presence of clay minerals, chlorite, calcite, iddingsitization of olivine.

Olivine is relatively homogeneous in composition apart from thin more Fe-rich rims (Supplementary Table MC3). Olivine core composition of individual samples vary from 81 to 87 Fo mol% (where Fo  =  100  ×  Mg/[Mg + Fe]). The olivine compositions plot well below the equilibrium Kd band, implying forsterite contents too low to be in equilibrium with the whole-rock Mg# (66-68) values. Few olivine phenocrysts (sample D10) seem to be in equilibrium with their host rock (Fig. 3). NiO ranges from below detection limits to 0.5 wt.%. CaO ranges from 0.1 to 0.4 wt.% (Supplementary Table MC3). The olivines in the sample D10 are equilibrated at ∼1170 °C (according to Roeder and Emslie, 1970).

The clinopyroxene of the Kekem dykes falls in the diopside/salite field or close to the diopside–augite limit, with rare Fe–augite (Ca_49-41Fe_12-38Mg_44-17; Fig. 4a; Mg# = 32-81; where Mg# = 100*Mg/[Mg + Fe]). The most Mg-rich clinopyroxenes have been found in samples Tb, M3 and D10. Ti and Al increase with decreasing Mg in the sample Tb, whereas Ti and Al decrease with decreasing Mg in the other samples (Supplementary Table MC4; Fig. 4b,c). The chemical composition of clinopyroxene (particularly the TiO₂ content, which reaches values as high as 4.4 wt.% in groundmass salite crystals of sample Mb) is strongly indicative of different magma groups and crystallization histories for the Kekem dykes. The composition of the clinopyroxenes of the Kekem dykes is similar to that of the clinopyroxenes of the basaltic dykes from Bangangte, Dschang and Manjo, and the Bangangte plateau basalts (southern continental part of the Cameroon Volcanic Line; Tchouankoué et al., 2012). The clinopyroxenes of the Kekem dykes indicate a relatively alkaline affinity of the parental magmas.

Plagioclase shows a wide range in composition from labradorite (An₇₁) to oligoclase (An₁₇) (Supplementary Table MC5; Fig. 5a). The iron content (as FeO_t) ranges from below detection limits to 1.3 wt.%.

Cr-bearing spinel occurs as small inclusions in olivine phenocrysts. The compositional range observed is wide with Al₂O₃ varying from 26.2 to 42.2 wt.%, Cr₂O₃ varying from 19.7 to 33.7 wt.% (Cr# = 24–43; where Cr# = 100  ×  Cr/[Cr + Al]), and MgO varying from 5.5 to 15.9 wt.% (Mg# = 27–66; Supplementary Table MC6). Samples Mb, M3 and D10 have spinels that are distinctly richer in Cr than those of sample Tb (Fig. 5b). The lower Cr contents of spinels in the sample Tb clearly indicate different and unrelated parental magmas with respect to the basaltic dykes Mb, M3, and D10.

Titaniferous magnetite and ilmenite are the main oxides found in the Kekem dykes. The Ti-magnetite has between 61 to 84 mol.% ulvöspinel and widely variable Al₂O₃ (from 1.5 to 3.3 wt.%) (Supplementary Table MC6). Ilmenites have a limited range of solid solution toward hematite and have low Al₂O₃ (< 0.5 wt.%) and MgO (< 2.7 wt%) contents (Supplementary Table MC6). Equilibrium temperatures and oxygen fugacities of coexisting magnetite and ilmenite, calculated using the ILMAT program of Lepage (2003), range from 1167 to 511 °C and from–9.4 to–27.8 log fO2 units, plotting close to the synthetic quartz–fayalite–magnetite (QFM) buffer. The ilmenite-magnetite pairs that re-equilibrated in the subsolidus plot towards the wüstite–magnetite (W–M) buffer.

The Kekem dykes are tholeiitic/transitional basalts according to the total alkali–silica (TAS; LeBas et al., 1986) diagram (Fig. 6) and CIPW norms. Samples Tb, Tc, Ma, M2, M3 and D20 are olivine-hypersthene normative (hy = 0.72%–13.96%) and samples D10 and M1 are nepheline (ne = 0.65%) and quartz (2.51%) normative, respectively. Dykes characterized by high MgO contents (MgO > 12 wt.%) can be classified as picrites (they are not ferropicrites having total iron as Fe₂O₃ lower than 13 wt.%). The Kekem dykes show a loss on ignition (LOI) value between 1.6 and 3.4 wt.%, suggesting that they underwent low to moderate alteration in agreement with their petrography. We therefore use in this paper only alteration-resistant elements (Ti, Zr, Nb, Y, Th and REEs and their ratios) for geochemical modeling and interpretation. In addition, major oxide compositions are based on analyses normalized to 100 wt.% on a volatile-free basis. The Kekem dykes are characterized by low to moderate TiO₂ (1.4–2.2 wt.%), Nb (6–37 ppm) and Zr (84–381 ppm) concentrations. MgO ranges from 7.4 to 12.4 wt.% (Mg# = 57–71, where Mg# = molar Mg*100/[Mg + Fe²⁺]). Compatible trace elements such as Cr and Ni show considerable range (Cr from 190 to 411 ppm; Ni from 15 to 234 ppm). The most primitive basaltic dyke (M2) with Mg# = 71, Cr = 411 ppm and Ni = 234 ppm. These values are consistent with mantle-derived primitive melts. Primitive mantle-normalized multi-element patterns of Kekem basaltic dykes are variably enriched in Rb, Ba and light rare earth elements (REEs) relative to Zr, Hf, Ti and the heavy REEs (Fig. 7a). The REE patterns (Fig. 7b) show moderate light REE enrichment (La_n/Yb_n = 5–8). The absence of negative Europium anomalies is consistent with the complete lack of significant feldspar fractionation.

The Kekem dykes show incompatible element concentrations and ratios that can be related either to the source heterogeneity and/or to shallow level evolutionary processes. The range of MgO (7.40–12.41 wt.%), Ni (15–234 ppm) and Cr (190–410 ppm) contents indicate that olivine fractionation (+ Cr-rich spinel) played an important role in the evolution of the Kekem dykes. However, the olivine fractionation will not significantly change the incompatible element ratios of magmas. The Kekem basaltic dykes show moderate Zr/Nb ratios (9–14) and high Zr/Y ratios (4–11), Nb/Yb ratios (5–8) and Ti/V ratios (48–71), typical of magmas rich in incompatible elements, such as ocean island basalts (OIB) (e.g., Sun and McDonough, 1989).

Trace element ratios such as Nb/U, Th/Nb and La/Nb are commonly used to assess the role of crustal contamination in basaltic rocks because they are not strongly modified from their source material by partial melting or fractional crystallization processes. The continental crust has low Nb/U (4.4–25) and high La/Nb (1.6–2.6) and Th/Nb (0.24–0.88) ratios (e.g., Rudnick and Gao, 2003). The Kekem dykes have Nb/U (29–32), Th/Nb (0.12–0.14) and La/Nb (1.5–1.8) ratios suggesting small crustal input. Furthermore, they show small negative Nb anomalies on the primitive mantle-normalized multi-element patterns (Fig. 7a), reflecting small crustal contamination. However, the Nb anomalies could be interpreted as source characteristics (Bensalah et al., 2013). In the Nb/Zr vs. Nb/La diagram (Fig. 8; Dadd et al., 2015), the data of Kekem dykes plot close to subcontinental lithospheric source. Therefore, the large range of geochemical compositions of the studied dykes is probably due to the heterogeneity of their mantle source region. Bensalah et al. (2013) distinguishes lithospheric from asthenospheric magma sources based on La/Ta and La/Nb ratios. The lithospheric sources are characterized by La/Ta > 22 and La/Nb > 1.5 whereas the asthenospheric sources have La/Ta < 22 and La/Nb < 1.5. The studied rocks display La/Ta (47–262) and La/Nb (1.5–1.8) suggesting a lithospheric origin.

In summary, the geochemical signatures and particularly the immobile incompatible element ratios of the Kekem dykes closely correspond to magmatic compositions. Fractional crystallization and crustal contamination have not notably overprinted the geochemical signature of Kekem dykes. The Kekem dykes with high MgO, Ni and Cr contents record geochemical characteristic of primary melts, and can be regarded as geochemical tracers of mantle source regions. The Visual Basic Excel spreadsheet of Lee et al. (2009) was used to constrain the major element compositions, temperature and depth of primary melts. This program is only valid for basaltic rocks that have fractionated along olivine-control line and reverse-fractionates the olivine crystallization (or accumulation for picrites). In addition, the average pressures and temperatures calculated are reliable only for primary mantle-derived magmas that came from a peridotitic (ol + opx + cpx) source (see Lee et al., 2009 for further details). The Kekem dykes have only fractionated olivine and so are suitable for modeling considering magmas with > 9.0 wt.% MgO and assume Fe³⁺/Fe = 0.1 and forsterite of the mantle source of 0.9. The temperatures and pressures of the primary magmas are respectively of 1392–1435 °C and 1.8–2.1 GPa (∼60–75 km), falling above the dry lherzolite solidus and the spinel–garnet transition (e.g., 80–100 km; Robinson and Wood, 1998). The Kekem data yield mantle potential temperatures around ∼1415 °C, which is slightly hotter than the average ambient mantle ∼1350 °C (e.g., Lee et al., 2009 and references therein).

To evaluate the melting processes involved in the formation of the Kekem dykes, we used the non-modal fractional melting mode (Shaw, 1970). The estimated composition of the primitive mantle (PM; Lyubetskaya and Korenaga, 2007) has been chosen as the potential mantle source, considering, as extreme cases, a spinel and a garnet facies. The source modal composition range is 53–60% olivine, 27–20% orthopyroxene, 17–10% clinopyroxene, and 3% spinel or 10% garnet. The amounts of minerals participating in the melt used in the model are as follows: spinel-bearing lherzolite source–6% olivine, 28% orthopyroxene, 67% clinopyroxene and 11% spinel; garnet-bearing lherzolite source 3% olivine,–16% orthopyroxene, 88% clinopyroxene, and 9% garnet. Mineral/melt distribution coefficients are from Niu et al. (1996). The results indicate that the Kekem dykes could be produced by 2–8% melting of a primitive source in the spinel stability field (Fig. 9). Our modeling implies that the Kekem magmas were generated at shallow depth (∼80–100 km; Robinson and Wood, 1998) in a within plate environment.

Several dyke swarms crop out in the central domain of the Cameroon Pan-African fold belt (also referred to as the South Continental part of the Cameroon Volcanic Line; Tchouankoué et al., 2014). Fig. 7a shows that the Kekem dykes have incompatible element patterns similar to those of the Dschang, Maham and Kendem basaltic dykes (Tchouankoué et al., 2014), suggesting a common mantle source. The basalt dykes of the southern continental part of the CVL were emplaced in the same geological environment, characterized by Paleoproterozoic to Archean crust remobilized during Pan-African orogenesis (Ganwa et al., 2016; Toteu et al., 2001).

The differences between the basaltic dykes of CCSZ and Cenozoic alkali basalts of the Cameroon Volcanic Line (Mt Manengouba; Kagou Dongmo et al., 2001; Bafang region; Tchuimegnie Ngongang et al., 2015) and the post-collisional Kekem Neoproterozoic Gabbro (Kwékam et al., 2013) are highlighted in Fig. 7. The Kekem dykes are different from those of the Bafang and Manengouba basalts, which have higher LREE, MREE contents and no Nb anomalies in the primitive mantle-normalized diagram (Fig. 7a). The alkali basalt of the Cameroon Volcanic Line appears to have been derived from a distinctly more enriched mantle source, with no crustal contamination. The Kekem gabbros have higher incompatible element contents than Kekem dykes and negative Nb anomalies in the primitive mantle-normalized multi-element patterns (Fig. 7). Kwékam et al. (2013) argued that the Kekem gabbros were derived from the partial melting of a garnet-spinel lherzolite mantle source. The negative Nb anomalies indicate that this mantle source was modified by the contribution of a subduction-related material (Kwékam et al., 2013).

As discussed above, the Kekem dykes and the other dyke swarms of CCSZ originated from a source area different from those of post Pan-African collision gabbros and CVL basalts in the studied area. This fact precludes any relationship between CCSZ dyke swarms and alkali basalts of the Cameroon Volcanic Line, and reinforces the hypotheses of Tchouankoué et al., 2012, 2014) that the dyke swarms represent magmatic events different from those of the Cenozoic Cameroon Volcanic Line.

Trace elements ratios are currently used to constrain the tectonic setting of dyke swarms (Bensalah et al., 2013). In Fig. 10 most of the Kekem dykes as well as those of the CCSZ plotted in within plate basalts field. These dykes swarm are associated with a network of post-pan-African fractures whose arrangement is compatible with Riedel's fracture model (Moreau et al., 1987; Mvondo Ondoa, 2009).

The N70–90°E oriented dykes swarm yielded an age of 421.3 ± 3.5 Ma (⁴⁰Ar/³⁹Ar) in Dschang (Tchouankoué et al., 2014), 404.2 ± 3.2 (⁴⁰Ar/³⁹Ar) and 417 ± 8.1 (⁴⁰K/³⁹Ar) in Maham (Tchouankoué, 2005; Tchouankoué et al., 2014). The ⁴⁰Ar/³⁹Ar age interval overlaps with the less imprecise age of K/Ar obtained in Maham, which suggests that they might belong to a same magmatic event. The field study, the geochemical features as well as the mineralogy data are similar to those of the Dschang and Maham areas, whose emplacement age is bracketed between 420 and 404 Ma (⁴⁰Ar/³⁹Ar age). Therefore, such a time interval can be constrained to the Kekem dyke swarm. Such a magmatic event is not widespread in Cameroon. However, a 428 ± 5.2 Ma magmatic event is reported in Senegal at the eastern border of the West African craton (Fullgraf et al., 2013). These magmatic events display a similar magma source region. Furthermore, both regions occupied the central Gondwana in the pre-drift reconstruction. These similarities suggest that the syn-rift magmatism as well as the opening of the Rheic Ocean ascribed to the Senegal Paleozoic can be envisaged for Cameroon dyke, and therefore the latter belongs to the same event.

The N25°–35°E oriented fractures in Bangoua and Kendem yielded ages of 214 ± 6.6 Ma (⁴⁰K/³⁹Ar) and 192.1 ± 7.5 Ma (⁴⁰Ar/³⁹Ar), respectively (Tchouankoué, 2005; Tchouankoué et al., 2014). Both ages might belong to a same magmatic event and the slight difference should be due to the imprecision of K/Ar method (Maluski et al., 1995). Given that ⁴⁰Ar/³⁹Ar age is more reliable that ⁴⁰K/³⁹Ar age, the age of 192.1 Ma recorded in Kendem fall within the interval of Triassic–Jurassic magmatism constrained at 200–190 Ma (⁴⁰Ar/³⁹Ar; Vérati et al., 2007) that affected America, Africa and the European continent. However, its geodynamic significance is diversely interpreted. It is considered as the early stage of the opening of the South Atlantic Ocean in South America (Thomaz-Filho et al., 2000) or of the lithospheric extension that results in central Africa rift (Youbi et al., 2003). This magmatic event displays similarities with the Kendem dyke swarm that yielded an age of 192 Ma (⁴⁰Ar/³⁹Ar, Tchouankoué et al., 2014), which falls in the interval between 200 and 190 Ma. These features suggest that the Kendem dyke in Cameroon is likely the manifestation of this magmatism in Cameroon. However, the lithospheric extension can be envisaged for Kendem dykes during the Triassic–Jurassic rift in the Central Africa rift, as the Kendem fracture trend (N25°E) is consistent with an ENE–WSW extensional direction suggested for the Benue Trough dyke swarm dated back to 148 Ma (Maluski et al., 1995), interpreted as the forerunners of the opening of the South Atlantic ocean (Maluski et al., 1995), which suggests that the latter have previously experienced an extension. This argument precludes any correlation with an early opening of the South Atlantic Ocean, as reported by Tchouankoué et al. (2014).

The age of 148 ± 38 Ma (obtained with K/Ar method) in Bangoua (Tchouankoué, 2005) is close to the dyke swarm ages (146–106 Ma, obtained with Ar/Ar method) of tholeiitic affinity, well documented in the northern area of the Benue Through in Nigeria (Maluski et al., 1995), and to the 145–127.5 Ma ages of the alkaline igneous rocks from the Parana–Etendeka large igneous province in the South American continent (Gibson et al., 2006). Both events were previously interpreted as forerunners of the extension of the South Atlantic Ocean (Maluski et al., 1995). However, Gibson et al. (2006) considered them as resulted from a plume–lithosphere interaction. In Cameroon, in the stage of knowledge, no data about large igneous province associated with a dyke swarm of such an age was reported. Rather, magmas generally consisting of high Mg and Ni contents suggest plume activity related magmatism (Gibson et al., 2006) and tholeiitic affinity is indicative of a lithospheric extension (Gibson et al., 2006; Kuepouo et al., 2006; Maluski et al., 1995; Youbi et al., 2003). Therefore, the role played by the lithospheric extension should not be excluded and the interplay between both phenomena should be taken in account. This implies that a plume–lithosphere interaction is the most plausible explanation, as suggested by Gibson et al. (2006). From a regional point of view, the northern Benue trough and the Bangoua area are located in central Africa, which suggests that both dyke swarms might belong to a same magmatic episode. Accordingly, the hypothesis of forerunners of the extension of the South Atlantic Ocean applied in Northern Benue can be envisaged in the Bangoua dyke swarm in Cameroon. This interpretation is consistent with structural studies that suggest an ENE–WSW extensional direction that reactivated pre-existing Pan-African fractures, leading to the formation of intracontinental NNE–SSW- and east–west-oriented graben systems transferred along N60°E sinistral faults (Maluski et al., 1995), as the N25–35°E oriented dyke swarm in Kekem matches well with one of the two directions.

The field relation (this study) and geochronological data (⁴⁰Ar/³⁹Ar and ⁴⁰K/³⁹Ar ages) on the mafic dyke swarm show that they post-date the pan-African deformation (Tchouankoué et al., 2014). N20 and N70 °E striking tholeiitic dyke swarms were reported in NE Nigeria, especially in the Benue Through (Maluski et al., 1995). The orientation of dyke swarms in Cameroon and NE Nigeria is consistent with Riedel's fractures kinematic model (Moreau et al., 1987; Mvondo Ondoa, 2009), suggesting that they resulted from a same tectonic event. In fact, both areas are located within the Central Africa Pan-African Shear Zone, and it has been demonstrated that the activity of such a large-scale shear zone is capable to originate similar structural associations (Moreau et al., 1987). Therefore, the late dextral shearing movement dated at 552 Ma (EMP dating on monazite; Tchaptchet Tchato et al., 2009) along the CCSZ, which served as pathways for the ascent into the surface of tholeiitic magma, is consistent with the geometric arrangement of this Pan-African fracture network and the hypothesis of a reactivation of Pan-African structures that originated from the Cretaceous sedimentary basins with which the dyke swarms are associated (Noutchogwe et al., 2010). These within-plate basalts derived from lithospheric extension (Kuepouo et al., 2006; Maluski et al., 1995; Youbi et al., 2003), which is consistent with the geophysical data of Tokam et al. (2010), in which the crustal thickness in Cameroon is significantly thinned under Cretaceous sedimentary basins.