Behaviour of the minor elements and their potential use

Figure 5.5:Sulfide/silicate partitioning behaviour of molybdenum and vana-dium as a function of oxygen fugacity, FeO content of the silicate, and tem-perature. The diamond symbol denotes the experiment performed at a lower temperature. Literature data from Li and Audétat (2012), Li (2014), and Lod-ders and Palme (1991) is plotted for comparison. Data from Li and Audétat (2012) was used with our data to fit a parameterised model for D_Mo^SL/SM as a function of fO₂, FeO content of the silicate melt, and temperature (see text);

the resulting curves are also plotted.

If an oxygen fugacity dependence can be calibrated for minor element par-titioning between sulphide and silicate melt it could be used in later experi-ments to determine the oxygen fugacity without the need to determine the melt ferric/ferrous ratio. The nickel concentrations in the silicate glasses of our re-covered samples were in general very close to the detection limit of the electron microprobe. As such our results have extremely large uncertainties, and so are not considered useful. Molybdenum concentrations in the silicate melt were also low but certainly above detection limits and, as will be shown, could be routinely used to determine the oxygen fugacities in such experiments.

The obtained molybdenum sulfide liquid/silicate melt partition coefficients,

D^SL/SM_Mo , appear to have a weak dependence on the silicate melt FeO

concentra-tion but no clear fO₂ dependence (Figure 5.5). Li and Audétat (2012) studied a basanitic melt composition at 1.5 and 2.5 GPa and ∼1200-1300 ^◦C. Their data, also shown in figure 5.5, have a stronger fO2 trend but were also col-lected over a narrower range in silicate melt FeO content. Li and Audétat (2015) proposed an empirical expression to describe D_Mo^SL/SM as a function of T, silicate melt FeO content (in weight %) and oxygen fugacity normalised to the FMQ oxygen buffer, i.e.,

D^SL/SM_Mo = 8.66−94000/T −0.27·(∆FMQ)−2.34·log(FeO_melt). (5.1)

This equation provides a poor description of our experimental results, to some extent because our data extend to a higher temperature than that at which the expression was calibrated. As the partitioning of Mo is clearly fO2 dependent, at least in the data of Li and Audétat (2012), a change of oxidation state of Mo between the sulfide and silicate melt is implied. The most likely change would be from Mo⁴⁺ in the sulphide to Mo⁶⁺ in the melt (Lodders and Palme, 1991; O’Neill and Eggins, 2002), which can be described through the exchange reaction:

MoO^silicate₃ + 2FeS^sulfide=MoS^sulfide₂ + 2FeO^silicate+ 1

2O2. (5.2) If it is assumed that the equilibrium is not pressure dependent, then the Gibbs

free energy for the equilibrium can be written as

∆rG= ∆H⁰−T∆S⁰+RTln

hX_Mo^{SLi h}X_FeO^SMi2[fO₂]^0.5

hX_Mo^{SMi h}X_FeS^{SL i2}

(5.3)

Where SL, SM and X denote sulfide liquid, silicate melt, and mole fraction, respectively, ∆H⁰ and ∆S⁰ are the standard state enthalpy and entropy of the equilibrium, and R is the universal gas constant. The activity-composition relations are, therefore, ignored and X_FeS is then assumed to be 1. The ex-pression can be rearranged: which has a very similar form to equation 5.1. This expression, however, predicts that the fO2 dependence should have the coefficient -0.5, and the X_FeO dependence -2, which is slightly different to the terms obtained by Li and Audétat (2015). Through a combined fit of the data from this study and that of Li and Audétat (2012) we derive the expression:

logD_Mo^SL/SM= −25300

T + 14.75−0.41 logfO₂−1.988 logX_FeO^SM. (5.5) The derived coefficients for X_FeO is exactly that predicted for equilibrium 5.3, and that for fO2 is very close. In fact, very little difference in the quality of the fit occurs if an fO₂ coefficient of 0.5 is used, in accordance with the most plausible redox change of Mo. Curves for this equation are plotted in Figure 5.5. As can be seen, the fO2 dependence is quite strong but is to some extent obscured in the current dataset by the FeO dependence. This seems to confirm the proposed change in oxidation state of Mo between the sulfide and the silicate melts, and also raises the possibility that the Mo partition coefficient could be employed to determine the oxygen fugacity in such experiments.

Li and Audétat (2015) also report Mo sulfide/silicate melt partition coefficients but the silicate melt compositions employed range through andesite to dacite, and the partition coefficients do not agree with the parameterisation in

equa-tion 5.5. This is most likely due to the raised level of polymerisaequa-tion in the melts. Similarly, the hydrous basanitic melts employed by Li (2014), which contain several weight % H₂O, are in poor agreement with equation 5.5 and with the results of Li and Audétat (2012). This is possibly due to the very raised H₂O contents. Finally, data from Lodders and Palme (1990) from ex-periments performed at 1 bar are also shown in Figure 5.5. In these data, the effects of varying Fe/S ratio on Mo partitioning was also examined, and only the data close to stoichiometric FeS melt are shown in Figure 5.5. Nev-ertheless, these data seem to show a much shallower fO₂ dependence, and they also extend to lower values of fO₂. This probably implies a change in the redox state of Mo in the silicate melt either at lower pressures or lower oxygen fugacities.

By rearranging equation 5.5 we obtain

logfO2 = 2

−25300

T + 14.75−logD^SL/SM_Mo −1.988 logX_FeO^SM

.

By assuming that the fO₂ dependence is that expected from equation 5.4, there is a slight improvement in the reproduction of the experimental oxygen fugacities for the current data set, where the oxygen fugacities are reproduced to ∼0.3 log units with only one outlier of 0.5 log units difference.

As shown in figure 5.5, there is no clear fO2 dependence in the partitioning of V, either in the current data set or in that of Li and Audétat (2012). This is somewhat surprising, because V is known to undergo a change in oxida-tion state in silicate melts as a funcoxida-tion of fO2, and mineral/melt partition coefficients have been used to infer oxygen fugacity (Mallmann and O’Neill, 2009).

However, it should also be noted that the partition coefficients obtained in the current study are about 1 order of magnitude higher than those obtained by Li and Audétat (2012). Although T may have some effect, there are data points from both studies at 1573 K and they differ by amounts that are far outside of the uncertainties. The difference is also probably not caused by pressure, as Li and Audétat (2012) examine one different pressure and no noticeable difference occurs. The data of Li and Audétat (2015), on the other hand are in

much better agreement with the results of this study. Further studies would be required to identify the cause of this difference, which may arise through some quite subtle, and therefore very interesting, chemical difference.