Determination of multiple sulfur isotopes in glasses: A reappraisal of the MORB δ34S

We report on improvements to the sulfur extraction method out of silicate glasses by the use of HF + CrCl2 solution for the precise determination of S multi-isotope compositions. This protocol has been designed to be sulfide specific and if needed, can be easily extended to oxidized sulfur. The complete method was validated on a synthetic powder and three laboratory mid-ocean ridge glassy internal standards (CH98 DR12, ED DR46 1-6 and ED DR26 type 1). S extraction yields of the three basalt glass standards were 100 +/- 4% (1 sigma, n = 12), 102 +/- 6% (1 sigma, n = 4) and 97 +/- 6% (1 sigma, n = 4) respectively. Their delta S-34 show little variation, at -120 +/- 0.08 parts per thousand, -1.09 +/- 0.14 parts per thousand and -125 +/- 0.12 parts per thousand (all 1 sigma) with respect to V-CDT. Delta S-33 and Delta S-36 are both negative with respect to our SF6 tank, between -0.018 and -0.021 parts per thousand. (+/- 0.012 parts per thousand maximum 1 sigma) for Delta S-33 and between -0.216 and -0282 (+/- 0.106 maximum 1 sigma) for Delta S-36. The method was then applied to sixteen additional glasses from worldwide mid-ocean ridges, including three samples for which delta S-34 was reported previously. Our results show that previous data are affected by a systematic delta S-34 shift toward positive values, of similar or equal to 0.4 to 1.59 parts per thousand. We infer that this shift originates from an incomplete S recovery when the Kiba extraction protocol is used. delta S-34 values range between -1.80 and 0.02 parts per thousand, with a mean value of -0.91 +/- 0.50 parts per thousand (1 sigma n = 19). Such negative values contrast with the positive range previously reported leading us to suggest a revised mantle delta S-34, mainly dominated by negative values. Delta S-33 and Delta S-36 are strikingly homogeneous with mean respective values of -0.019 +/- 0.005 parts per thousand and -0.193 +/- 0.093 +/- (1 sigma n = 19) versus our SF6 tank. These estimates are indistinguishable from our CDT measurements and define the best present-day upper mantle estimate. (C) 2012 Elsevier B.V. All rights reserved.