Fig. 5: Estimates of ƒO₂ of convecting mantle sources from which the oceanic crustal protoliths to eclogites formed.

Shown are V abundances (ppm) in reconstructed bulk eclogite xenoliths and DI as a function of MgO content (wt.%), filtered to exclude samples with gabbroic cumulate protoliths (conservative cut-off Al₂O₃/FeO = 1.7) and, for xenoliths, metasomatism (NMORB-normalised Ce/Yb > 1 and Cr₂O₃ > 0.1 wt.%). Shown for comparison is the modelled evolution of melts for a mantle potential temperature (T_P) of 1450 °C (olivine crystallisation temperature of 1340 °C) after crystallisation of spinel, olivine and plagioclase from a parental melt formed at a melt fraction of 0.2, appropriate for warmer Archaean spreading ridges, and various ƒO₂. A difference of ±50 °C in T_P and corresponding crystallisation temperature is shown as blue envelope for ∆logƒO₂(FMQ) = −2, whereby V becomes more incompatible with increasing temperature³⁵, causing the isopleths to shift upward relative to samples. Modelled effect of 30% melt extraction at FMQ-4 and FMQ (as in Fig. 4a) is shown for comparison (purple arrows), illustrating that melt extraction at FMQ-4 biases V abundances toward higher apparent ƒO₂ whereas at FMQ melt extraction shifts samples with MgO > ∼9 wt.% along olivine control lines. Modelling parameters and rationale in Methods; Supplementary Table 3. Error bars for reconstructed bulk rocks show average propagated 1σ uncertainties on V abundances of 29 ppm for xenoliths and 69 ppm for DI (Methods). Data sources in Supplementary Data 1.