Partitioning of elements between silicate melt and immiscible fluoride, chloride, carbonate, phosphate and sulfate melts with implications to the origin of natrocarbonatite

Liquid–liquid partitioning of 42 elements between synthetic silicate melts and immiscible fluoride, chloride, carbonate, phosphate and sulfate liquids was studied at temperatures of 650–1100 _C, pressures 72–100 MPa, with 0–11 wt.% H2O. One series of experiments was performed in a rotating internally-heated autoclave where separation of the immiscible liquids was assisted by centrifugal forces. An analogous series of experiments was done in static rapid-quench cold-seal pressure vessels. The experimentally determined liquid–liquid distribution coefficients (D’s) vary over several orders of magnitude, as a result of variable Coulombic interactions between cations and anions. For alkaline, alkaline earth and rare earth elements there is a strong and systematic dependence of the liquid/liquid D values on the ionic potential Z/r for all the examined systems. In contrast, highly charged cations (e.g., HFSE) show no systematic relationships between the D’s and Z/r. New experimental constraints on the carbonate/silicate liquid–liquid D values presented here confirm that rare metals such as Nb, Zr, REE, Th and U concentrate in silicate liquids, and therefore carbonatites that carry economical rare metal mineralization are not likely to have formed by liquid immiscibility. The comparison between experimentally-determined carbonate– silicate liquid–liquid D values and bulk-rock natrocarbonatite vs. nephelinite compositions at the Oldoinyo Lengai in Tanzania reveals significant discrepancies for Cs, Rb, Ba, Be, Zn, heavy REE, Ti, Mo and W, thus rendering a simple, one-stage immiscibility model for Oldoinyo Lengai questionable