4. Results
4.3 S, H, and C Contents of Melt Inclusions
It is worth pointing out that the length-scale of the Fe depleted zone of olivine adjacent to the melt inclusion in the FMQ+4.7 experiment is ~40 µm which is about 2x longer than that the diffusion distance (i.e., x = (D∙t)1/2) estimated for Fe-Mg interdiffusion at the experimental condition (x ≈ 16 µm, extrapolated to high ƒO2 from Dohmen & Chakraborty 2007). It is possible that an Fe3+-defect-rich olivine increases bulk diffusivities, especially of charged defect species, making it easier to transmit ƒO2
information across the crystal volume at high ƒO2; the fact that dislocations in olivine are rapidly decorated at oxidizing conditions (e.g., Wu and Kohlstedt 1988), and symplectites are found to follow traces of cracks in the experimental olivines (Figure 5A), indicates that fast paths for diffusion of oxygen or oxidant-bearing defects were likely present (e.g., Spandler and O’Neill 2010, Burgess and Cooper 2013).
from experiments run at or below FMQ+0.35 was designated as either sulfide-saturated if sulfides were observed or sulfide-undersaturated if they were not observed. Note that this designation was usually made after half of the inclusion was ground away during polishing; if there were sulfides in the half that was destroyed and none in the remaining half then there may be some false-negative cases where inclusions were wrongly identified as sulfide-undersaturated. Despite this uncertainty, the petrographic designation is clearly consistent with the S contents of the inclusions: sulfide-saturated inclusions have uniformly high S (all >660 ppm) and all but two sulfide-undersaturated inclusions have S <520 ppm (Figure 9). The two sulfide-undersaturated inclusions with S between 602-633 ppm may be wrongly identified, or only slightly undersaturated with respect to sulfide. It has become increasingly evident that the sulfur concentration at sulfide saturation (SCSS) is sensitive to not only melt composition (e.g., O’Neill &
Mavrogenes 2002), but also the composition of the coexisting sulfide (Smythe et al.
2017), the concentrations of siderophile elements in the melt (e.g., Ni and Cu, O’Neill 2020), and to H2O (Smythe et al. 2017, Liu et al. 2021). The effect of H2O is negligible for the low water contents of the natural (<0.30 wt%) and experimental (<0.08 wt%) inclusion; the inferred effect is to increase SCSS by about 100 ppm per 1 wt% dissolved H2O. Ni and Cu were not measured in the experimental glasses. Only three experiments at FMQ-1 had sulfides large enough for quantitative WDS analyses (see Chapter 1), and the Ni+Cu was variable (from ~0.1 wt% to ~0.3 wt%) (Supplementary Figure 1).
In some olivine grains there are both sulfide-saturated and sulfide- undersaturated melt inclusions present and in a few cases both types of inclusions were exposed at the polished surface and could be analyzed. For two olivines equilibrated at FMQ-1 (H9X2 and H2XC2), sulfide-saturated inclusions have S = 776 and 803 ppm,
Figure 9. FeO* wt% vs. S wt% in experimental melt inclusions. Black squares – FMQ-1; Green diamonds – FMQ+0.35; Blue circles – FMQ+2.8;
Red triangles – FMQ+4.7. Filled symbols had sulfides visibly identified either in backscattered electron images or optically with transmitted light.
Open symbols had no sulfides visible.
respectively, whereas the nearby sulfide-undersaturated glasses have S = 590 and 633 ppm. A special olivine grain SF11-3, contained a high density of trapped melt inclusions, which allowed for simultaneous exposure of many melt inclusions (Figure 10). In this single grain, which was homogenized for 24 hr at 1225 °C and FMQ+0.35, 16 melt inclusions were analyzed for their S contents. Four of the sixteen exposed inclusions had visible sulfides; coincidentally this is the same fraction of melt inclusions (25%) that were observed to have sulfides by inspection of the whole population of unheated natural olivines from Papakōlea. The sulfide-saturated inclusions have S between 710-780 ppm, two of the sulfide-undersaturated inclusions have intermediate S = 514 and 520 ppm, and the remaining ten inclusions all have S < 254 ppm (cloud of green diamonds at FeO* = 7.7 wt% in Figure 9). Although the number of inclusions in each group are small, the major and minor element compositions are the indistinguishable except for their P2O5 contents, which fail a t-test for equal means and variances (p>0.05), and which are lower in the sulfide-saturated liquids. Comparing the sulfide-saturated and -undersaturated groups, the FeO* contents are 7.91±0.03 wt%
(n=4 inclusions) and 7.7±0.11 wt% (n=12), respectively, and the P2O5 contents are 0.26±0.04 (n=4) and 0.34±0.04 (n=12), where the number after the plus-minus symbol is 1σ of the distribution of the average inclusion compositions based on three analyses of each inclusion. Because all the inclusions were hosted in the same olivine (Fo88.4), these differences likely reflect heterogeneities in the compositions of the initially trapped melt inclusions. Note that P2O5 has been shown to have a very large effect on stabilizing Fe3+ over Fe2+ in silicate melts (~10x larger per mole compared to the effect of Na2O, for example, Jayasuriya et al. 2004), which could possibly explain the lower P contents measured in sulfide-saturated inclusions versus sulfide-undersaturated inclusions of otherwise indistinguishable liquid compositions.
Figure 10. Backscattered electron images of experimental olivine SF11-3, which contained many melt inclusions that were simultaneously exposed on the polished sample surface. The melt inclusions are circular to ovoid shapes with numbers, and are brighter than the surrounding olivines. Sulfides are bright white and are visible in some of the melt inclusions. Scale bars are 140 µm.
Figure caption is on the next page.
Figure 11