MELTING OF THE SHALLOW UPPER MANTLE: A NEW PERSPECTIVE

Abstract

Detailed examination of liquidus phase relationships in binary and ternary joins of the CFMAS +Cr system has permitted a rigorous determination of the dry melting path of an initially fertile spinel peridotite composition resembling Bulk Silicate Earth or MORB-pyrolite. It is demonstrated that it is impossible to model mantle melting accurately using only one set of ratios of phases entering the melt; this implies that the melting process is primarily controlled by solid solution rather than eutectic behaviour. The proportions of phases entering a melt depend on whether a phase reacts and/or disappears from a system, and on the choice of the initial and final peridotite compositions. Four discrete domains in the melting regime of upper-mantle peridotites are distinguished, each characterized by different phase melting coefficients, relating to the melting of: (1) lherzolites, (2) clinopyroxene-bearing harzburgites (i.e., free-clinopyroxene), (3) clinopyroxene-saturated harzburgites (i.e., clinopyroxene in solid solution in orthopyroxene), and (4) clinopyroxene-free harzburgites (i.e., no clinopyroxene). The proposed non-linear fashion in which mantle lithologies melt explains the inadequacy of all previous models to reproduce the observed compositions of upper-mantle peridotite melting residues. It is suggested that: (1) olivine and orthopyroxene will melt cotectically; (2) clinopyroxene and spinel will lose most of their aluminous component after ̃8% melting within the first 4 kb(̃ 12 km) of ascent from the dry solidus; and that (3) clinopyroxene will disappear completely from a MORB-pyrolite mantle after ̃42% melting. Although such a number is significantly higher than that dictated by the position of the clinopyroxene-out curves from peridotite isobaric equilibrium melting experiments (̃22%), it is emphasized that the latter are a gross oversimplification of the natural melting process and are not equivalent to melting during adiabatic upwelling. It is concluded that the commonly postulated disappearance of clinopyroxene from fertile peridotite compositions at ̃22% melting is greatly in error if melting in an adiabatically rising mantle is considered, thus providing an explanation for many unsuccessful attempts by various authors to model the behaviour of transition elements in sub-oceanic and supra-subduction-zone mantle and derivative magmas.