New chemical and Sr-Nd isotopic data on the Late Cretaceous mafic dike swarm intruding the Archean-Proterozoic crystalline basement in the Tamatave-Sainte Marie Island sector (northeast coast passive margin), and on lavas and dikes of the northeastern part of the Mahajanga sedimentary basin (passive margin after the opening of the Jurassic-Cretaceous Somali basin), allow better knowledge of the chemical variations observed in the northern part of the Madagascan igneous province. Two distinct basalt groups have been identified. Group 1 basalts have low light to heavy rare earth element (REE) ratios [(La/Yb) n = 2.2-2.9], low Zr/Y and Nb/Y (4-6 and 0.2-0.4, respectively), low ( 87 Sr/ 86 Sr) 88 (0.7034-0.7042), and high to moderate e Nd (88) (+5.1 to +1.5). Subgroup la comprises basalts with the same light to heavy REE ratios [(La/Yb)n = 2.7-3], Zr/Y and Nb/Y (4.5-5.8 and 0.2-0.3, respectively), and slightly high ( 87 Sr/ 86 Sr) 88 (0.7042-0.7048) at the same e Nd (88) (+5.4 to +4.4) of the group 1 basalts. Group 2 basalts have high light to heavy REE ratios [(LaNb) n = 5.3-7.8], high Zr/Y and Nb/Y (7-11 and 0.5-0.8, respectively), relatively high ( 87 Sr/8 6 Sr) 88 (0.7045-0.7057), and low e Nd (88) (+3.8 to +1). The basalts of the groups 1, la, and 2 cannot be linked by closed-system magma-differentiation processes, and require distinct mantle sources. The major and trace element variations of the Tamatave dikes of the group 1-la are compatible with moderate degrees of crystal fractionation (∼60%) from the least (MgO = 7.3 wt%) to the most evolved compositions (MgO = 4.2 wt%), involving the separation of plagioclase, augite, pigeonite, and minor oxides, perhaps accompanied by crustal contamination or differences in the 87 Sr/ 86 Sr ratios. The mantle sources of the group 1-la basalts seem to be located well within the spinel stability field, whereas a larger contribution of melts derived from garnet-bearing residual mantle is observed in the geochemistry and in the melting models of the group 2 basalts. The chemical and isotopic composition of both rock groups indicate their ultimate provenance from variably enriched lithospheric mantle sources; there is no clear evidence of a hotspot component like that found in the present-day lavas of the Marion-Prince Edward archipelago. The sources of this volcanism seem to be significantly similar to those of the Mahableshwar and Ambenali basalts of the later erupted Deccan Traps, located on formerly contiguous parts of the Gondwana lithosphere.
The northern part of the Late Cretaceous flood-basalt province of Madagascar is formed by lava flows, dykes and intrusive complexes. These range from tholeiitic to weakly alkaline and have composition from basalts to basaltic andesites (with rare, more evolved, rocks). The magmatic evolution is dominated by low-pressure crystal fractionation of olivine, Cr-spinel, plagioclase and clinopyroxene. The erupted rocks form two subprovinces, in terms of incompatible element concentrations or ratios and geographical position. The mafic and intermediate samples from the western subprovince have low abundances of high field strength elements (e.g. Nb 2–9 ppm; Zr 50–230 ppm), negative Nb anomalies on mantle-normalized diagrams, and a wide range of initial (88 Ma) 143 Nd/ 144 Nd (0.51293–0.51197; ε Nd +7.9 to −10.9). These elemental and isotopic variations are compatible with melting of predominantly N-MORB-like mantle and variable low-pressure crustal contamination. The mafic samples of the eastern subprovince are mildly enriched in incompatible elements (e.g. Nb 6–25 ppm; Zr 109–330 ppm), with the highest values found in tholeiites from the east Mahajanga basin, Tampoketsa Kamoreen and Tamatave. They also have a comparatively narrow range of initial 143 Nd/ 144 Nd (0.51276–0.51258; ε Nd +5.1 to +1.0), consistent with derivation from incompatible element enriched mantle sources and minor subsequent crustal contamination. The marked differences in the chemical and isotopic characteristics of the samples imply variable degrees of partial melting of hotter than normal, chemically heterogeneous, mantle sources at pressures where spinel to garnet are residual phases. No clear evidence is seen for a chemical component similar to that represented at present by the Marion hotspot lavas.