Redox evolution of carbon- and sulphur-bearing clinopyroxenites as proxies of deeply recycled crustal material.

Fossil subduction zones are key to studying the deep geochemical cycles of C, O, and S. We analysed graphite-sulphide-bearing garnet clinopyroxenites from the External Ligurian (Northern Apennines, Italy), which serve as indicators of deep recycling of subducted crust. These rocks are among the only three known pyroxenite occurrences worldwide (along with Beni Bousera, Morocco, and Ronda, Spain) that formed through the crystallisation of eclogite-derived melts (P ≥ 3 GPa and 1100 °C), following a prolonged recycling history in off-craton mantle. In particular, a MOR-type heterogeneous gabbroic sequence was recycled into the mantle as eclogite 1.5–1.0 Ga ago, then partially melted and crystallised in the convective mantle, followed by subsolidus re-equilibration and exhumation. We analysed the redox state of garnets and clinopyroxenes associated with graphite and sulphides in garnet clinopyroxenites, that crystallised from a liquid produced by partial melting of recycled eclogite. Electron Energy Loss Spectroscopy within a transmission electron microscope and Synchrotron micro-Mössbauer analyses revealed heterogeneities in Fe3+/Fe2+ distribution and its partitioning among mineral phases (Fe3+/ΣFe error is 0.03 for Electron Energy Loss Spectroscopy and 0.01–0.03 for Synchrotron micro-Mössbauer). The analysed clinopyroxenites display three generations of clinopyroxenes: unexsolved crystals included in garnet cores with Fe3+/ΣFe = 0.16–0.38 (representative of eclogite-derived melt crystallisation in the asthenosphere), clinoenstatite exsolution-bearing grains with Fe3+/ΣFe = 0.03–0.10 (related to a first stage of exhumation in the lithospheric mantle), and Al-poorer rims without Fe3+ (related to the final stage of exhumation). In contrast, garnets have Fe3+/ΣFe-poor cores (<0.03) and slightly higher ratios in the rims (0.04–0.07). When considered together with the markedly higher Fe3+ contents in the earliest clinopyroxene generation, this pattern is consistent with a pressure–temperature–dependent partitioning of ferric iron from garnet to clinopyroxene during cooling from 1100 to 950 °C along the exhumation path. fO2 calculations suggest a variation from more oxidised samples (ΔFMQ = −1.25 to 0) to more reduced ones (ΔFMQ = −4.2 to −1.6) at 3 GPa. At 1.5 GPa, ΔFMQ ranges from −1.2 to −0.6 down to < −5, indicating that graphite may have formed through reduction of a previously oxidised carbon phase. The oxidation state variations are linked to sub-solidus decompression, and not to S-C-related redox reactions, describing a closed system with no fluid/melt-rock interaction. Our results show that sulphur plays no role in controlling the redox state of these graphite-bearing mantle rocks, even over prolonged geological histories, and that variations in the redox state of carbon and iron in garnet and clinopyroxene can depend on pressure and temperature changes only, rather than from redox reactions, indicating that the intensive variable fO2 can be decoupled from redox processes in a closed system.