One-atmosphere experiments conducted on a synthetic glass similar to Galapagos Spreading Center (GSC) FeTi basalt (POO.82N2), (Byerly et al., 1976) define liquid lines of descent at f02 values between the quartz-fayalite-magnetite (QFM) buffer and 2 log units more oxidizing than the nickle-nickle oxide (NNO) buffer. The experiments provide a framework for understanding the development of FeTi basalts by fractionation at near-ocean floor conditions. GSC lavas from near 85¿W initially follow a compositional trend, distinguished by FeO (=FeO+0.9Fe2O3) enrichment and SiO2 depletion, which is nearly identical to the trend observed in experiments at QFM to which olivine seeds were added. This compositional trend can be produced by crystallization along an olivine→pigeonite reaction boundary in a shallow crystal-rich magma reservoir. In contrast, GSC lavas from 95¿W do not mimic the 1-atm liquid line of descent, but appear to have fractionated at somewhat higher pressure. Basaltic liquids from 95¿W underwent fractional crystallization at 1--2 kbar, did not experience FeO enrichment along an olvine→low-Ca pyroxene reaction boundary, and developed FeO enrichment concomitant with SiO2 enrichment. This compositional variation is consistent with a differentiation process in which crystals are continually removed from contact with liquid. Ryodacites from 95¿W cannot be related to the basalts and FeTi basalts recovered at 95¿W by parents at greater depth and higher PH2O, or formed by fractionation of different parents. Andesite formed by mixing between basaltic and ryodacitic liquids.

As a consequence, mixed andesites define a trend of decreasing P2O5 which has been previously interpreted to represent apatite saturation at approximately 0.22 wt% P2O5 significantly earlier than at 85¿W (where P2O5 decreases at approximately 0.7 wt% P2O5). Our experiments suggest that the fO2 when titanomagnetite first saturates at the GSC was approximately at the NNO buffer. Together with the Fe2O3/FeO data of Byers et al. (1983, 1984) and Christie et al. (1986), this requires an increase in fO2 during crystallization in excess of that produced during closed-system fractional crystallization. We suggest that this increase in fO2 results from interaction with oxidizing surroundings in an open-system process. ¿ American Geophysical Union 1989