Résumé
Orogenic lherzolites exhibit a specific (although reproducible) platinum-group element signature characterized by slight positive deviations of Pd/Ir, Rh/Ir and Ru/Ir ratios from the canonical chondritic model. Such a signature was alternatively considered to be a compositional feature of the primitive upper mantle or resulting from sulfide melt addition by refertilisation reactions in the continental lithosphere. To shed additional light on this conundrum, the distribution of PGE in FON B 93, an unserpentinized, fertile (3.24% Al2O3) orogenic lherzolite (French Pyrenees) used as in-house standard by our group, has been studied to different scales, from the bulk rock to trace minerals. In addition to new ICPMS analyses after separation into Nis and Te coprecipitation, its whole-rock PGE concentrations were redetermined by three different laboratories using very precise ID-ICP-MS methods after digestion of the sample in Carius tube at T=250 degrees C and 320 degrees C or in high-pressure asher (HPA-S). The four methods produce reproducible Ru, Rh and Pd contents (7.1 +/- 0.18 ppb; 1.43 +/- 0.05 ppb; 7.1 +/- 0.30 ppb) whereas the non-ID NiS-fire assay method underestimates the Os and Ir concentrations by c.a. 10 and 15% compared to ID-ICP-MS analyses (4.40 +/- 0.07 and 4.00 +/- 0.17 ppb, respectively). Platinum was the most difficult to analyse. If performed on powder aliquots smaller than 3 g, the ID-ICP-MS analyses generate strong nugget effects while the non-ID NiS-fire assay method yields statistically lower (but highly reproducible) Pt concentrations (6.92 +/- 0.26 ppb). These features reflect the strong partitioning of Pt into trace phases that SEM and laser ablation-ICP-MS analyses on thin sections identify as both high-temperature Pt-Ir-Os alloys and Pt-(Pd) tellurides of likely subsolidus origin; ten to fifteen grains ranging in maximum dimension from a few micrometres to a few hundreds of nanometers, were identified by standard polished thin section. LA-ICP-MS data on base metal sulfides (25 grains analysed), coupled with the whole-rock S concentration (277 +/- 10 ppm) and modal composition of the BMS (90% pentlandite+accessory pyrite and secondary pyrrhotite + 10% chalcopyrite) allowed the contribution of the BMS phase to the PGE budget of FON B 93 to be estimated. Except Pt that exhibits a 95% deficit in the BMS phase, the PGE concentrations measured by ID-ICP-MS can be balanced by BMS while Al-spinel is a negligible contributor, accounting for less than 0.5% of the Ru budget. The occurrence of Pt in trace phases may bias the whole-rock Pt concentrations because 1) mechanical collection of Pt-rich trace phases remains problematic with the Nis button, and 2) Pt-Ir-Os alloys may be prone to digestion problems in conventional Carius tube procedures. Since they could be stable at mantle depth, such refractory alloys that contain Os and Ir may also have enhanced the heavy PGEs/light PGEs fractionation. Our observations likely pertain to orogenic lherzolites as a whole because BMS assemblages in these mantle rocks record evolution at low sulfur fugacity, which prevents Pt-Ir-Os alloy from entering the Mss in the mantle; moreover, at subsolidus temperature, pentlandite, the major BMS, cannot accommodate 0 valence state Pt.