Metasomatism-induced Melting in Mantle Xenoliths from Mongolia

Abstract

Mantle xenoliths from two locations in Mongolia contain patches of glass-phenocryst aggregates (‘melt pockets’) up to 1 cm in diameter, including one ‘composite’ xenolith, which shows a complete transition from unaltered spinel Iherzolite to a zone containing melt pockets surrounded by a cpx and spinel-free peridotite matrix. We have analyzed major elements by wet chemistry, X-ray fluorescence (XRF), and electron microprobe, trace elements by ion microprobe and inductively coupled plasma mass spectrometry (ICP-MS) techniques, and Sr and Nd isotopes by mass spectrometry, to elucidate the origin of these melt pockets. Petrographic and chemical evidence shows that the melt pockets were formed neither by infiltration of the host basalt nor by dehydration melting of hydrous phases, such as amphibole. Instead, melting was induced by the interaction of a metasomatic fluid with clinopyroxene and spinel. The reaction produced melts of variable composition, with SiO2 ranging from 52 to 68% and MgO from 4.5 to 0.5%. The melts contain euhedral grains of olivine, clinopyroxene, and spinel, and a large number of (now empty) vugs. The melt shows no sign of having invaded the Iherzolite matrix surrounding the pockets. There is some evidence for fractional crystallization, but some of the major element chemical trends, such as the negative correlation between Na2O and SiO2, cannot be accounted for by such a mechanism. The glasses and clinopyroxene phenocrysts are very rich in light rare earth elements (LREE) and Sr, and completely dominate the bulk contents of these and some other incompatible elements in the rocks with melt pockets. The invading fluid introduced high concentrations of LREE, Th, U, Pb, and Sr, but was relatively depleted in Ba, Rb, Nb, Ta, Zr, Hf, and Ti, and had unusually high Zr/Hf and Nb/Ta ratios. Ion microprobe analyses of fresh glass directly adjacent to clinopyroxene microphenocrysts yield a series of cpx-melt partition coefficients for REE and several other trace elements. DYb (cpx-melt) varies between 0–3 and 1.6 and is positively correlated with the A12O3+SiO2 and Na2O contents of the glass, and negatively correlated with MgO, FeO, and CaO contents. These correlations are consistent with qualitative predictions from considerations of silicate melt structure. The clinopyroxenes in the unaltered zones of the composite xenolith show evidence of an earlier phase of metasomatism which enriched Ce, La, and Sr, but did not affect the other REE. Clinopyroxenes from these zones have high εNd values of + 14 and +19, indicating a history of low Nd/Sm ratios. At the same time, 87Sr/86Sr ratios are high (>0.704), indicating infiltration of relatively radiogenic Sr during the early stage of metasomatism. Ion microprobe traverses show no zoning of La/Nd ratios. Therefore, there was enough time to equilibrate the metasomatic effects in the grain interiors, and we estimate the time required for this equilibration to be of the order of 105 years. In sharp contrast, the second, or main, metasomatic event that caused the formation of the melt pockets must have been extremely short-lived and probably lasted only hours or days before the xenolith was captured by the magma and erupted at the surface. This short duration is required by the preservation of fresh glass and by the lack of equilibration of the melt pockets with their surrounding matrix. The isotopic compositions of Sr and Nd are identical between melt pockets and host basalts in both localities. Therefore, we conclude that the metasomatic fluids were probably derived from the same source rocks as the host basalts. We speculate that the xenoliths originally resided in an upper-mantle region which was intruded by a partially molten diapir. Volatiles were expelled from the unmelted margin of the diapir and invaded the adjacent upper-mantle peridotites. The fluid infiltration triggered formation of the melt pockets, whereupon the material was picked up by rapidly ascending magma and erupted at the surface. The fluids appear to have been poor in water, as no hydrous minerals are present among phenocryst or quench phases in melt pockets. The major component of the fluid may have been CO2 or liquid carbonate.