Partitioning of Trace Elements During Primary Melting of MORB Mantle

Vincent J. M. Salters National High Magnetic Field Lab. and Dept. of Geology, FSU, Tallahassee, Florida, USA

salters@magnet.fsu.edu

John Longhi Lamont Doherty Earth Observatory, Palisades, New York, USA

We have determined the partition coefficients between clinopyroxene, orthopyroxene, garnet, and melt at a range of pressures (2.8 to 1.0 GPa) and temperatures (1570oC-1375oC) and compositions in an effort to mimic as closely as possible the conditions at which MORB are formed.

The experiments were run in two steps. First the multi-saturated liquidus phase relations of upper mantle minerals (olivine, orthopyroxene, clinopyroxene, and garnet or spinel) were determined for relevant portions of the synthetic system SiO2, TiO2, Al2O3, FeO, MgO, MnO, CaO, Na2O, K20, Cr2O3, and P2O5. The goal was to produce assemblages of mineral phases and melt that are similar in composition to those expected from low degrees of anhydrous melting of natural peridotite, but that have low crystal/melt ratios in order to promote equilibrium and to facilitate analysis. Although none of the experiments to date has yielded a full lherzolitic assemblage, the composition of the lherzolitic melts can be closely bracketed. these lherzolitic melts can be assigned a percent mantle melt by assuming a reference mantle composition (e.g. primitive mantle minus 1% mantle melt) and then using the K concentration in the synthetic melt to calculate percent melt.

After determination of the phase relations of synthetic mixtures, the most promising mixtures were doped with U and Th at 3000ppm level, Nb and Hf at 500 ppm level, and Y, Zr, Ce, Nd, Sm, Er, Yb and Lu at 100ppm level. This material was then run to yield run products on which mineral melt distribution coefficients were determined. The trace element compositions of minerals and melt were analyzed by with the CAMECA 3f ion microprobe at WHOI. The distribution coefficient for Nb is extremely low in clinopyroxene, garnet and orthopyroxene, thus Nb was always measured to ensure that the analysis spot was free of melt inclusions.

The U and Th distribution coefficients (Ds) for clinopyroxene range from 0.014 to 0.003 for U and 0.002 to 0.012 for Th such that clinopyroxenes with low CaO content have the lowest distribution coefficients. The CaO content in the clinopyroxenes ranges from 6 wt.% in the high pressure clinopyroxene (2.8 GPa) to 16 wt.% at 1.2 GPa. The effect of pressure on pyroxene composition has not been appreciated until recently and, consequently, Ds for clinopyroxene have not determined at these low CaO contents. The change in Ds as a consequence of the change in composition is most dramatic for the REE. The high pressure low-CaO clinopyroxene has Ds for the REE which are within error to those of orthopyroxene (DLu=0.149 for cpx and DLu=0.136 for opx) and are up to three times lower than high CaO clinopyroxene (DLu=0.435-0.502). Thus although the modal abundance of clinopyroxene has to increase with pressure (as to accommodate the calcium) the bulk distribution coefficient for the REE decreases with increasing pressure.

DU and DTh for garnet range from 0.016-0.021 and 0.005-0.007 respectively which shows less U/Th fractionation than other studies. The Ds for the REE in garnet are similar to previous studies. Garnet is able to significantly fractionate the Zr/Sm and Hf/Sm ratio as Dzr and DHf are 1.5 to 2 times higher than Dsm. Consequences of these changes in distribution coefficients for MORB genesis will be discussed at the meeting.