View article

Methods: In our model, we use a crystallization sequence [4], where we assume a 600 km deep LMO, the lower and upper limits of DH for each mineral-melt pair (Table 1), and vary the initial bulk H2O (100 and 1000 ppm) assuming no residual melt in the crystal mush after compaction (Figure 1). We track the H2O content of the crystallized phases, and compare the H2O content of the plagioclase cumulates with the measured H2O equivalent contents of plagioclase in FANs and Mg-suites [5, 6]. We also track the H2O content (OH-+ molecular H2O) of the residual liquid at each stage of crystallization, and when the dissolved H2O content exceeds the solubility limit at a given pressure and temperature [7], we assume that the excess H2O is degasses and outgassed efficiently from the system. We also evaluate the effect of bulk BSM or LMO H2O content and fraction of interstitial liquid on the average crustal thickness using a combination of the codes SPICES [8] and alphaMELTS [9]. The combination of the two codes yields the best fit to experimentally produced data [10]. We have not considered the presence of molecular H2 in the system, even though up to 20% of H in the system may be present as H2 under LMO conditions [11]. H2 will be considered in our future simulations.