H2O LOSS FROM HYDROUS MELTS DURING FLUID-ABSENT PISTON CYLINDER EXPERIMENTS

We have documented H2O loss from vapor-absent melting (dehydration-mel ting) experiments with durations of 5-31 d in a piston cylinder appara tus at 7, 10, and 15 kbar and 925-1000-degrees-C. In experiments on an amphibole gneiss (PB-92-2), the most obvious manifestation of H2O los s is a decrease in melt fraction and an increase in plagioclase abunda nce with increasing temperature. In the most extreme case (at 10 kbar) , the melt fraction decreases from 33 to 15% between 975 and 1000-degr ees-C, whereas modal plagioclase increases from 18 to 29 wt%. The tota l H2O content of this sample, estimated from microprobe 0 analyses, de creased from a starting value of 1.4 wt% to 0.6 wt% at 1000-degrees-C, although no such decrease was evident at 975-degrees-C. Similar, but smaller, effects were observed in high-temperature experiments on the amphibole gneiss at 7 and 15 kbar. The bulk H2O content of a sample co ntaining a biotite gneiss composition (PB-92-1) decreased from 1.7 to 1.0% between 975 and 1000-degrees-C at 10 kbar with no obvious effects on phase relations. Nominal H2O losses observed in long-duration expe riments (31 d, 10 kbar, 950-degrees-C) on both starting compositions w ere not resolvable within analytical uncertainty. Although dehydration is accompanied by an increase in f(O2) that is attributable to H loss , the low abundance of Fe3+ and the low f(O2) (QFM - 1) of even the mo st dehydrated samples require a mechanism other than H loss for most o f the observed dehydration. We suggest that molecular H2O is diffusing out of the samples. Unlike H, molecular H2O is a common species in hy drous silicate melts. The little information that exists on the diffus ivities of large volatile species (e.g., O2, N2) in metals suggests th at the diffusivities can approach that of H at high temperatures. If H 2O diffusivity is within even 2-3 orders of magnitude of H diffusivity at 1000-degrees-C, diffusive loss of H2O could account for our observ ations.