Primary Melts at Mid-Ocean Ridges, Mantle Plumes
and Subduction-Related Plate Margins:
Constraints on Mantle Melting

A. V. Sobolev Vernadsky Institute of Geochemistry, Russian Acad. of Sciences, Kosygin str.,19,

Moscow 117975, Russia

The main objective of this study is to contribute to the ongoing debate (McKenzie and Bickle, 1988; Plank and Langmuir, 1992) on the nature and mechanisms of mantle melting and magma segregation, transport and storage, by a study of magmatic inclusions in near-liquidus phenocrysts (Sobolev and Shimizu, 1993) from various mantle-derived melts combined with a study of the mineralogy of magmatic cumulates (Ross and Elthon, 1993) and residual mantle after melt extraction (Johnson et al., 1990). Specific goals include reconstruction of major and trace element compositions of mantle-derived primary melts, reconstruction of the mantle source compositions and melting scenarios, and modeling of fractionation trends of these primary melts. Studied samples containing magnesian olivines came from the type localities of 3 major geodynamic environments: mid-oceanic ridges (MOR- Mid Atlantic Ridge , 9-16 N, East Pacific Rise, Siqueiros F.Z., Iceland), mantle plumes (MP- Mauna Loa, Hawaii) and suprasubduction zones (SSZ- major volcanic suits of the Troodos ophiolites, Cyprus). Studied samples of magmatic cumulates and residual mantle peridotites came from the Troodos intrusive complex and metamorphic base. The study includes electron and ion microprobe analyses
of experimentally and naturally quenched glassy melt
inclusions in early phenocrysts as well as clinopyroxenes
from cumulates and peridotites to characterize source composition and reveal different fractions of primary melts, and modeling of mantle melting processes using petrological
and geochemical results from published high-pressure

Widely accepted idea of general mixing of primitive melts in the mantle assumes very rare natural occurrence of unmodified primary melts (McKenzie and Bickle, 1988). However massive ion probe studies of melt inclusions in high Mg-olivines in all studied objects show that from a few to 10% of the measured populations could be treated as unmixed or moderately mixed nearly primary melts whereas the rest are well mixed samples close in composition to bulk rocks or glasses. Nearly primary melts manifest extremely large range in incompatible element concentrations from ultra-enriched to ultra-depleted even in a single sample suggesting very efficient element fractionation during melting. Similar fractionation pattern of highly incompatible elements is indicated by data on clinopyroxene compositions from cumulates and residual peridotites.

The modeling of compositions of instantaneous primary melts suggests the following conclusions: (1) mantle under MOR, MP and in SSZ acts as effectively open system for the melt fractions of less than 1-3 wt %; (2) magmagenerating mantle columns under MOR commonly exceed 50 km in thickness and melting starts in the presence of garnet (less than 5% degree of melting); (3) the garnet is required in the source in the entire melting range (1-12 % degree) for studied MP; (4) some primary melts could escape magma mixing or interaction with wall rocks and thus argue for the fast channeling from the segregation level. (5) The melt inclusions data for mantle melting in SSZ is confirmed by data on Troodos cumulates and peridotites (Batanova et al., 1994) as well as similar data for MOR is consistent with results of study of MOR cumulates (Ross and Elthon, 1993) and peridotites (Johnson et al., 1990).


Batanova, V.G., Sobolev, A.V. & Schmincke, H.-U., Min. Mag. 58A, 57-58 (1994).

Johnson, K.T.M., Dick, H.J.B. & Shimizu, N., J. Geophys. Res. 97, 9219-9241 (1990).

McKenzie, D. & Bickle, M.J., J. Petrol. 29, 625-679 (1988).

Plank, T. & Langmuir, C.H., J. Geophys. Res. 97, 19749-19770 (1992).

Ross, K. & Elthon, D., Nature 365, 826-829 (1993).

Sobolev, A.V. & Shimizu, N., Nature 363, 151-154 (1993).