DETERMINING THE COMPOSITION OF HIGH-PRESSURE MANTLE MELTS USING DIAMOND AGGREGATES

We present a new experimental technique for circumventing the quenchin g problems that have plagued high-pressure peridotite melting studies. A thin layer of approximately 50 mum diamonds is placed above a layer of peridotite powder. Partial melt extracted from the peridotite laye r collects in the pore spaces between the diamonds and equilibrates di ffusively with the residual peridotite mineralogy. Isolated from the c rystalline residue, the melt quenches to a glass that records the comp osition of the liquid coexisting with the residual crystalline phases under the conditions of the experiment. We have used this technique to investigate partial melting of a fertile mantle composition at 10 kba r and a temperature range of 1270-1390-degrees. Oxide concentrations i n the liquids from the longest duration runs (up to 151 hours) vary sy stematically with increasing temperature: TiO2, Al2O3, and Na2O decrea se monotonically, while Cr2O3, FeO, and MgO increase steadily. CaO sh ows more complicated behavior, first increasing and then decreasing, w ith the crest in the temperature-CaO trend approximately coincident wi th the disappearance of clinopyroxene from the residue between 1330 an d 1350-degrees-C. Overall variation in silica content with temperature is small, and there appears to be a minimum at about 12% melting. The compositions of liquids produced in time series, temperature reversal , and two-stage experiments (conducted to test the technique) all indi cate that our experimentally determined liquid compositions represent close approaches to equilibrium. Calculated melt fractions (F) also va ry systematically with temperature. The slope of the T(degrees-C)-F cu rve is not constant over the spinel lherzolite melting interval, but d ecreases as temperature increases from 1270 to 1330-degrees-C. Extrapo lating the curve back to zero melt suggests that the anhydrous solidus temperature for our peridotite starting composition is approximately 1240-degrees-C. At temperatures below the cpx-out curve, melt generati on occurs via the reaction, 0.38 opx + 0.71 cpx + 0. 13 sp --> 0.22 ol iv + 1.0 liq, and the proportions of minerals that enter the melt appe ar to be independent of temperature. At temperatures above cpx-out, th e less well constrained melting reaction is: 1.06 opx + 0.04 sp = 0.1 oliv + 1 liq. The fact that all of the 10 kbar melts have FeO content s that are substantially lower than those reported in any primitive MO RB glasses further strengthens the conclusions that these glasses are not 10 kbar primary melts, that they involve a component of higher pre ssure partial melting, and that they have evolved by significant olivi ne fractionation from more primitive liquids. Our experimental data al so provide an independent check of the results of recent peridotite pa rtial melting calculations. Efforts to parameterize the experimental d atabase on peridotite melting, and to calculate melt compositions as a function of P, T, and F are partially successful in reproducing the c ompositional trends determined in this study.