The oxidation of elemental carbon to carbonate beneath mid-ocean ridges

Carbonatites or carbonate-rich melts are likely to be the first liquids produced during mantle up welling (McKenzie, 1985). Though they are implicated in the origins of kimberlitic magmas, it is possible that mid ocean ridge melting also commences with the production of small degree carbonate rich melts

at conditions much deeper in the mantle than the main phase of silicate partial melting. Seismological and geoelectrical anomalies represent the evidence for deep melting in MOR settings and are confirmed by experimental studies, which showed a typical high electrical conductivity of CO2-rich liquid.

The depth of 200-300 km beneath the East Pacific Rise where this body characterized by high electrical conductivity was identified has been also the subject of experimental petrology. In order to explain deep melting, experimental studies suggested that the melting of a peridotite source with 2.5 % of CO2 is able to produce carbonatitic melts in small amounts occurring when the mantle adiabat intersects the solidus curve of the appropriate mantle assemblage. However, the crucial role of oxygen fugacity in the upper mantle was totally neglected. While it is possible that the coexistence of carbonate and graphite/diamond may buffer the oxygen fugacity of the mantle, as a result of the relatively low carbon content, ferric/ferrous equilibria probably exert a more dominant influence and may drive the oxygen fugacity of the mantle down with increasing depth. In this scenario the onset of carbonatitic melting in upwelling mantle will be controlled by the point where the ambient oxygen fugacity crosses the oxygen fugacity imposed by the appropriate carbon-carbonate equilibria, rather than the solidus of carbonate-bearing mantle.

Using MORB glasses as a source to determine the fo2 range in the asthenosphere and oxy-thermobarometry methods we estimated that the bulk Fe³⁺/ΣFe ratio of MORB source peridotite is likely in the range 0.02-0.03, which is consistent with estimates for the melt-residue partition coefficient of Fe³⁺ and with estimates based on mantle xenoliths (O’Neill et al 1993; Canil et al., 1994;

Canil and O’Neill 1996).

We assume that the general trend of decreasing fo2 with depth found in mantle xenoliths from cratonic areas might also reflect the fo2 in asthenospheric mantle in down welling when a whole convection mantle is considered. Calculations were performed along an adiabat with a potential temperature of 1320 °C using the bulk silicate Earth (BSE) composition of McDonough and Sun

(1995) andassuming a fixed garnet Fe³⁺/∑Fe ratio of 0.12, which corresponds to a bulk rock Fe³⁺/∑Fe ratio of approximately 0.03.

By determining the fo2 buffered by equilibria involving both elemental carbon (graphite or diamond) and carbonate minerals or melts at pressures between 2.5 and 11 GPa and temperatures at and above the carbonated peridotite solidus, we are now able to determine the depth at which a redox melting may occur. Based on our results we have examined how the fo2 of the carbon/carbonate equilibrium varies with melt composition as the solidus carbonate melt evolves towards a silicate liquid. The dilution of the carbonate component in SiO2-bearing melts was found to lower the relative fo2, expanding the melt stability field with respect to reduced carbon, which well applies to a context of decompression melting along an adiabat beneath mid ocean ridges. Comparing with previous studies for the plausible oxygen fugacity trend in the upper mantle based on redox equilibria involving the skiagite garnet end member, the stability of carbonate-bearing melts should enter the domain of asthenospheric mantle fo2 at depths of approximately 120 km, where redox melting (Taylor and Green, 1988) would occur. Only when up-welling mantle reaches this depth will graphite start to oxidize, through reduction of Fe2O3 in silicates, to form a carbonate melt containing 3-15 wt. % SiO2. Small fractions of CO2-rich liquids will not affect the main production of melt from the spinel peridotite facies. Isotopic signatures and trace elements patterns, however, will record features that result from the scavenging of elements incompatible in the presence of carbonate melt from much greater volumes of the mantle than the main phase of melting involves.

However, some caution must be taken in the acceptance of the current ideas on the variation in fo2

of the mantle with depth and the calibration of the existing oxy-thermobarometer for garnet peridotite (Gudmunnson and Wood, 1995). The garnet oxy-barometer predicts that at pressures of approximately 6 GPa garnets buffered at oxygen fugacities compatible with the presence of graphite and carbonate

melt should contain in excess of 40 % of Fe in the ferric state. However, in this study no garnet was synthesised with more than approximately 12 % ferric Fe. In fact, although greater amount of ferric iron (~40%) in garnet would be required in order to promote redox melting of a carbonated peridotite, those amounts have been shown to be totally in contrast with garnet compositions found in these experiments and even in natural rocks, which recorded coexistence with CO2-rich melt. If results from this study are confirmed, the oxygen fugacity of the upper mantle may be in fact more oxidized than previously thought.

In this scenario, during up welling carbon oxidation will reduce the Fe2O3 content of the mantle causing the fo2 to be buffered by the graphite/carbonate-bearing melt equilibrium for some depth interval until graphite is exhausted. The extent to which this is important depends on the mantle carbon content, for which estimates for the upper mantle range from approximately 20-250 ppm.