Formation of primary kimberlite melts – Constraints from experiments at 6–12 GPa and variable CO2/H2O

Liquidus phase relations were experimentally studied in model kimberlite melts at 6–12 GPa. Four starting materials were used with different CO2 and H2O contents but almost identical proportions of other components (~ 35 wt.% SiO2, 3 wt.% Al2O3, 10 wt.% FeO, 33 wt.% MgO, 15 wt.% CaO, 0.3 wt.% Na2O, and 1 wt.% K2O on a 100% volatile-free basis). The liquidus phase of the CO2 richest mixture (33 wt.% CO2 and no H2O) is coesite at P > 6–8 GPa and low-Ca pyroxene at lower pressures. In an H2O-free mixture with lower CO2 content (18 wt.%) low-Ca pyroxene is the main liquidus phase and coesite was never observed in the experimental products. The addition of water (mixtures with 10 wt.% H2O + 9 wt.% CO2 and 12 wt.% H2O + 5 wt.% CO2) depresses the liquidus temperature and expands the crystallization field of olivine. At high pressures (> 8 GPa), garnet crystallization dominates near-liquidus phase relations. Based on the experimental data, possible conditions of the simultaneous saturation of kimberlite melt with olivine + low-Ca pyroxene ± garnet were estimated. High-Ca pyroxene was never found in the near-liquidus assemblages despite the high-CaO compositions of the starting mixtures. It was shown that the supposed primary kimberlite melt can be in equilibrium with the garnet harzburgite assemblage at 6–8 GPa and variable volatile contents, from ~ 15 wt.% CO2 under anhydrous conditions to 20 wt.% H2O and low CO2. These conditions correspond to possible parameters of kimberlite magma generation by partial melting of carbonated garnet harzburgite in the lithospheric mantle. The primary CO2 content could be higher if the kimberlite magma was poorer in SiO2 (more ‘carbonatitic’). The maximum CO2 content is limited by the appearance of magnesite on the liquidus at 20–22 wt.%, which is much lower than CO2 solubility in kimberlite melt at 7–8 GPa (> 30 wt.%). The content of H2O in primary kimberlite magmas depends on the bulk H2O content in the source, because no hydrous phases are stable near the kimberlite liquidus.