Fe³⁺ partitioning systematics between orthopyroxene and garnet in mantle peridotite xenoliths and implications for thermobarometry of oxidized and reduced mantle rocks

Abstract

We have investigated the partitioning of Fe³⁺ between orthopyroxene (Opx) and garnet (Grt) in well-equilibrated mantle xenoliths using Mössbauer spectroscopy. The samples cover a wide range of P–T conditions (2.1–6.6 GPa, 690–1,412 °C) and geothermal gradients, and are thus representative for Earth’s upper mantle in both on-craton and off-craton continental settings. Garnet has Fe³⁺/Fe_tot ratios of 0.03–0.13 and Fe₂O₃ contents of 0.24–1.00 wt%. Orthopyroxene has, on average, lower Fe³⁺/Fe_tot ratios (0.01–0.09) and Fe₂O₃ contents (0.05–0.63 wt%). In low-pressure, high-temperature samples, however, Opx is systematically richer in Fe₂O₃ than the coexisting Grt. The Fe³⁺ Opx/Grt partition coefficient $\left( {D_{{{\text{Fe}}^{ 3+ } }}^{\text{Opx/Grt}} } \right)$ shows no obvious relationship with temperature, but increases with decreasing pressure and with increasing Na^Opx. The observed Opx/Grt Fe³⁺ systematics imply that the Opx–Grt Fe–Mg exchange thermometer is not robust against redox changes if total Fe is treated as Fe²⁺. An approximate evaluation of errors on T estimates due to redox effects predicts negligible deviations for strongly reduced conditions (<65 °C), but potentially large deviations (> to ≫100 °C) for strongly oxidized conditions, especially at very high pressure and when both P and T are calculated by iteration.

Appendix: Estimation of maximum bias on Opx–Grt temperature estimates due to changing redox conditions

Figure 8 shows a compilation of existing fO₂ data for mantle xenoliths worldwide, recalculated using input P–T values obtained with the thermobarometer combinations recommended by Nimis and Grütter (2010). This choice significantly reduced the scatter of points (especially for Diavik) compared to earlier published versions of this plot (e.g., Stagno et al. 2013). Correction of Canil and O’Neill’s (1996) Mössbauer data for different recoil-free fractions in Grt (Table 4) produced a slight decrease in calculated fO₂ of about 0.6 Δlog units. The plot shows the well-known overall decrease in FMQ-normalized oxygen fugacity with increasing mantle depth and a range for fO₂ at each depth.

From this compilation, we selected five xenoliths coming from different depths and recording “average” redox conditions for their particular depths of provenance (Table 6; Fig. 8). We calculated the Opx–Grt temperatures for these xenoliths with the thermometer version of Nimis and Grütter (2010) (hereafter TNG10) at P given by the thermobarometers combination recommended by the same authors, using total Fe. The TNG10 thermometer was calibrated against a large set of mantle xenoliths from localities worldwide and should therefore be robust when applied to mantle rocks characterized by “average” redox conditions. All selected xenoliths showed very good agreement (ΔT < 60 °C) between thermometric estimates using the internally consistent thermometers recommended by Nimis and Grütter (2010). This indicates good equilibrium and also confirms that redox conditions in the xenoliths were indeed “average” and compatible with the TNG10 thermometer calibration (cf Nimis and Grütter 2010). Therefore, the calculated P–T conditions should be reliable.

We then allowed fO₂ for each of the selected xenoliths to vary to the maximum and minimum values expected for the mantle at the corresponding depths, as indicated by our compilation in Fig. 8. We estimated the Fe³⁺/Fe_tot ratios in the garnets at these maximum and minimum redox conditions by reversing the oxybarometer of Stagno et al. (2013), and those in the coexisting orthopyroxenes by using the Fe³⁺ partitioning systematics obtained in our work (cf Eq. 6). The mineral compositions were modified using the new Fe³⁺/Fe_tot ratios while kee** $K_{{D_{{{\text{Fe}}^{2 + } - {\text{Mg}}}} }}^{{{\text{Grt}} - {\text{Opx}}}}$ unvaried––the latter depends essentially on T, therefore kee** it fixed corresponds to kee** T fixed. An increase (or decrease) in the Fe³⁺/Fe_tot ratio thus determined a net increase (or decrease) in the total Fe content (actually Fe³⁺), which was compensated by varying the Al³⁺ + Cr³⁺ contents by the same magnitude at constant Al/Cr ratio. Since the solid solution model for garnet which is used in the oxybarometer of Stagno et al. (2013) is sensitive to the Al and Cr contents, the Fe³⁺/Fe_tot ratios had to be readjusted by iteration, although the effect of this correction was found to be minimal.

We then recalculated the TNG10 temperatures using the modified total Fe contents in both orthopyroxenes and garnets, either kee** P fixed or recalculating both P and T iteratively. The P–T estimates obtained for the selected xenoliths using the original mineral compositions and the compositions modified for their respective maximum and minimum redox conditions are reported in Table 6. We emphasize that the aim of this exercise was to assess “relative” variations on final P–T estimates and that possible small interlab discrepancies in Fe³⁺/Fe_tot ratios for Grt extracted from the literature and from this work do not significantly alter our results.