Oxidation of the Kaapvaal lithospheric mantle driven by metasomatism

Abstract

The oxidation state, reflected in the oxygen fugacity (fO₂), of the subcratonic lithospheric mantle is laterally and vertically heterogeneous. In the garnet stability field, the Kaapvaal lithospheric mantle becomes progressively more reducing with increasing depth from Δlog fO₂ FMQ-2 at 110 km to FMQ-4 at 210 km. Oxidation accompanying metasomatism has obscured this crystal-chemical controlled depth-fO₂ trend in the mantle beneath Kimberley, South Africa. Chondrite normalized REE patterns for garnets, preserve evidence of a range in metasomatic enrichment from mild metasomatism in harzburgites to extensive metasomatism by LREE-enriched fluids and melts with fairly unfractionated LREE/HREE ratios in phlogopite-bearing lherzolites. The metasomatized xenoliths record redox conditions extending up to Δlog fO₂ = FMQ, sufficiently oxidized that magnesite would be the stable host of carbon in the most metasomatized samples. The most oxidized lherzolites, those in or near the carbonate stability field, have the greatest modal abundance of phlogopite and clinopyroxene. Clinopyroxene is modally less abundant or absent in the most reduced peridotite samples. The infiltration of metasomatic fluids/melts into diamondiferous lithospheric mantle beneath the Kaapvaal craton converted reduced, anhydrous harzburgite into variably oxidized phlogopite-bearing lherzolite. Locally, portions of the lithospheric mantle were metasomatized and oxidized to an extent that conversion of diamond into carbonate should have occurred.

Appendix: Modifications to the flank method

The flank method protocol we have used to measure Fe³⁺ concentrations in mantle-derived garnets follows the pioneering work of Höfer et al. (1994) (as detailed by Höfer and Brey 2007). From this base, we have made modifications to the method to improve the precision and accuracy for the analysis of mantle-derived pyropic garnets with low ΣFe and Fe³⁺ concentrations. Our modifications include: (1) increasing the number of analytical points and decreasing the dwell time per spot; (2) operating the electron probe with a beam current of 150-160 nA and an accelerating voltage of 20 kV; and (3) using garnets’ standards compositionally similar to the unknowns to establish calibration curves (e.g., Fig. 16 of Höfer and Brey 2007).

We have increased the number of analytical points to a 10 × 10 grid with a spacing of 2–3 μm between each spot. The spacing is set large enough to avoid overlapping activation volumes and gives an effective spatial resolution of 20 × 20 μm or 30 × 30 μm. Our flank method analyses were conducted with spectrometers in a fixed position to avoid errors introduced by minor irreproducibility in the mechanical positioning of the spectrometer crystal. In our protocol, measurements are first made on the Lβ flank position for a grid on all garnets and then the spectrometer is driven to the Lα flank position and the grid sequence is repeated. Instrumental drift is monitored by repeat analysis of one of the standard garnets.

The data set of 100 points, counted for 60 s each on Lβ and Lα (total analysis time of ~3.5 h per sample), has a Gaussian distribution and, therefore, allows the use of simple statistics to describe the population. Thus, the Fe Lβ/Lα ratio of the garnet is taken as the mean of the analyses and the error in Fe³⁺/ΣFe is estimated by propagating the standard error of the mean. As indicated in the main text, repeat analyses of a single garnet fragment over a period of several months had a reproducibility of ±0.008 in Fe³⁺/ΣFe.

Operating with high beam current increases the intensity (count rate) of the weak L-line X-Ray emissions which is important for measuring samples with low total (<10 wt% FeO_T) and ferric iron (<1.0 wt.% Fe₂O₃) concentrations. The increased count rate improves the sensitivity of the flank method making it applicable to mantle-derived pyropic garnets; however, the increased X-Ray intensity of major element emissions (e.g., Mg, Si) increases “dead time” of gas-flow proportional counters making simultaneous quantitative analysis using first order Kα emissions unreliable. With proper calibration, it may, however, be possible to use second order lines for quantitative analysis.

Höfer and Brey (2007; Fig. 14) show that in the region of key interest for mantle garnets (<10 wt.% FeO_T, and Fe³⁺/ΣFe < 0.15), there is considerable scatter and disagreement between the flank method and Mössbauer spectroscopy for natural samples. One potential source of the observed discrepancy may be the sensitivity of the flank method to crystallographic site distortions due to cation substitution (Höfer 2002). By measuring fragments of natural homogeneous garnet crystals, similar in major element composition to the unknowns (especially in Ca concentration), using a Mössbauer ‘milliprobe’ with a young high specific activity point source (McCammon 1994; with experimental setup as in McCammon and Kopylova 2004), we have been able to improve the agreement between the two techniques (Fig. 2). The calibration curves necessary for calculating the Fe³⁺/ΣFe of unknown garnets (e.g., Fig. 16 of Höfer and Brey 2007) were constructed using the same grains measured with Mössbauer spectroscopy, eliminating any error that may be introduced by slight variations in ferric iron concentrations in the standard garnets. Because of minor variations in spectrometer reproducibility, it is important that the calibration curves be derived for each analytical session. Using these modifications, the flank method can now be applied to accurately and precisely measure Fe³⁺/Fe²⁺ of mantle-derived garnets.

Another potential application of the high spatial resolution of the flank method is the ability to measure variations in Fe³⁺ in zoned garnets. Garnets in the present study were found to be unzoned, so this ability was not needed. But to illustrate the potential, we were able to measure ferric iron zoning in a large (~3 cm diameter) garnet in xenolith PHN 1611 from Thaba Putsoa, Lesotho (Fig. 12) using the flank method. The ferric iron zoning in this sample indicates an increase in Δlog fO₂ with the metasomatic enrichment apparent in this sample.

Fig. 12

Iron zoning in a large (3.2 cm diameter) garnet from xenolith PHN1611. Total iron reported as FeO_T shows a slight increase from core to rim. Ferric iron measured with the flank method shows a dramatic increase from 2.60 to 2.94 wt% Fe₂O₃ within 60 μm from the garnet rim. The 1σ error bars are smaller than the symbols used