Fig. 4

H₂O and CO₂ systematics during crustal differentiation of OPB and MORB. a H₂O and b CO₂ versus [MgO] for OPB and MORB. Because S can strongly partition into a co-existing H₂O−CO₂ vapor phase (e.g. ref. ⁵), it is important to constrain whether OPB were vapor-saturated or not during crustal differentiation. Like MORB, the H₂O contents of OPB increases with decreasing [MgO] and analyses closely overlap with a 0.1 GPa MELTS model (ref. ⁵⁷) of fractional crystallization of the most primitive (highest [MgO]) Kroenke sample (see accompanying supplementary note for further details regarding MELTS modeling). These systematics indicate that OPB were H₂O-under-saturated during differentiation. Silicate melts typically degas CO₂ before H₂O (e.g. ref. ³⁷) and this is consistent with the lower CO₂ of many of the OPB magmas compared to the trend predicted using MELTS modeling. Hence, with the exception of the S-degassed Tamu OPB, degassing of OPB was typically restricted to loss of CO₂. A few of the Kwaimbaita samples plot on the MELTS line, indicating that CO₂ saturation was reached at ~8.5−7.5 wt.% MgO and melts started to degas. MORB H₂O data are taken from Le Roux et al.³⁶ and Cottrell and Kelley³⁵. MORB CO₂ data are taken from Le Roux et al.³⁶. Assuming that the Kroenke samples are under-saturated and have not degassed H₂O or CO₂ and were generated by ~30% partial melting, we estimate that the OPB source mantle contained ≥700 ppm H₂O and ≥40 ppm CO₂