Résumé
Carbon emission at volcanic centers requires a constant balance between output (mostly by volcanism, either at plate boundaries or intraplate) and input (mostly at trench settings) of carbon from and to the Earth's mantle. The form of carbon that resides in the mantle is controlled by depth (pressure) and oxygen fugacity, the latter in turn depending on the depth and the concentration of iron in the mantle. In the shallow, lithospheric mantle, carbon is likely to be present in the oxidized form of CO2 (except under cratons where carbon is reduced to graphite or diamond). Below approximately 90 km, in the asthenosphere, the oxidized form of carbon is carbonate, either mineral or melt, depending on the thermal regime. At depths greater than approximately 150 km, the asthenospheric mantle is too reducing for carbon to stay in its oxidized form and only diamond is present, unless there is sufficient hydrogen to form reduced C–H fluids. Hence, the region located in the depth range of 90 to 150 km deep is where carbonatitic melts can most likely be produced and impregnate the surrounding mantle through metasomatism. The upper bound of this region is called the carbonate ledge. This limit prevents carbonate (either solid or molten) from ascending because of degassing and CO2 liberation. The lower bound is a redox front where redox melting (that is, melting caused by oxidation) may take place in an ascending portion of carbon-containing mantle. Carbonatite eruptions and presence of carbonate mineral inclusions in deep-seated diamonds provide evidence that these boundaries can be trespassed in some cases.An analysis of the experimental data that has bearing on silicate melting in the presence of carbon further shows that the carbonate ledge is a melting curve with a negative or flat Clapeyron (dP/dT) slope. In the carbonated ultrabasic (peridotite) systems, the carbonate ledge is located between ~ 2–3 GPa. The ledge divides the pressure–temperature space into a region of low-pressure silicate melt production, and a high-pressure region where carbonatites can be produced. Carbonatitic melts in equilibrium with mantle peridotite have compositions close to dolomitic (approximately equal amounts of Ca and Mg) with a general trend of becoming markedly more magnesian with increasing pressure. Calcic carbonatites may be stable at pressures < 2 GPa if clinopyroxene is absent. At mantle transition zone pressure range, there seems to be a melting temperature decrease (negative fusion slope), which may be caused by the stabilization of majoritic garnet.The carbonated basic (broadly eclogitic) system is more complex than the peridotitic one, because of the strong control of bulk silicate composition on melting temperatures, and hence, on melt composition. In carbonated eclogite systems, we propose that the effect of bulk composition upon all observed features can perhaps be related to silica super-saturation, or lack thereof. In some cases, high calcium (> 80 mol% CaCO3) melts can be produced, making the melting of carbonated eclogites an appealing scenario for the genesis of calcio-carbonatites in the Earth's mantle. Comparison with modeled pressure–temperature paths of subducted oceanic lithosphere shows that fusion of carbonated eclogite at depths shallower than 200 km should be expected for hot (Cascadian-type) subduction thermal regimes. On the other hand, in the case of cooler thermal regimes (Honshu-type, for instance), subducted carbonates may be stable to greater depths in Earth at trench settings, depending on the bulk composition of the system.Furthermore, high-pressure experiments show evidence of a continuum among carbonatitic, kimberlitic, melilititic, and basaltic liquids, for increasing melting degree of carbonated peridotite. This continuum has not been documented in the case of fusion of carbonated eclogite. It may be present, however, when certain sediments are fused, although the silicate melts are granitic to rhyodacitic instead of being kimberlitic in composition. Additional high-pressure work on phase relations in the simple binary system CaCO3–MgCO3 and specific focus on oxide solubility in the vapor phase have the potential to further clarify phase relations on complex silicate–carbonate systems at mantle conditions.