Sulfur abundance and heterogeneity in the MORB mantle estimated by copper partitioning and sulfur solubility modelling

Abstract

How much sulfur (S) in the upper mantle is important for understanding the global S cycle. The S contents of mid-ocean ridge basalts (MORBs) are usually used to estimate the mantle S abundance. However, this approach of estimation can be problematic due to possible sulfide-saturation during mantle melting. The partitioning behaviour of the chalcophile element copper (Cu) during mantle melting has the potential to track the fate of sulfide and the sulfur (S) abundance in the mantle. Sulfide is a minor phase and will be progressively consumed with mantle melting progress. At a given melting degree, Cu partitioning is controlled by both the silicate residue and sulfide. Therefore, accurate Cu partition coefficients (DCus) for silicate minerals in addition to sulfide are necessary to predict the Cu behaviour and sulfide fate. However, high-precision DCus for silicate minerals are still insufficient, leading to an incomplete understanding of the behaviour of Cu and S during mantle melting. Using the Cu-bearing capsule technique, we obtained accurate DCus between mantle minerals [olivine (ol), orthopyroxene (opx), clinopyroxene (cpx) and spinel (spl)] and basaltic melts under the P-T conditions for the MORBs generation. We found that Dol/meltCu=0.050±0.007, Dopx/meltCu=0.032±0.006, Dcpx/meltCu=0.049±0.010 and Dspl/meltCu=0.043 was nearly constant for each mineral, regardless of the P, T and compositions of minerals and melt under our experimental conditions. Based on mineral modes of the depleted mantle, a constant bulk Dmin/smCu of 0.045 between peridotite residue and basaltic melt is applicable for sulfide-exhausted mantle melting. With this Dmin/smCu, three important results are obtained in this study: (1) the calculated Cu abundances in the sources of six komatiite and picrite suites are 21–27 ppm; (2) the predicted Cu contents of the melts derived from sulfide-exhausted mantle are always higher than the real Cu contents of primary MORBs with melting degrees <∼15%, indicating sulfide-saturation is common during MORB generation; and (3) combination of Dmin/smCu (0.045) with the published models of Dsulfide/smCu and S-solubility, the S budget (residual sulfide S + dissolved melt S) calculations show that S abundance in the MORB mantle ranges from 120 to 300 ppm with an average of 206±25 ppm. Our results not only reveal sulfur in the upper mantle is highly heterogeneous but also demonstrate that the mantle S abundance previously obtained from magmatic S contents was underestimated.