We present here a study of plagioclase/melt partitioning of trace elements at their natural concentration levels, using sample charges from the widely cited plagioclase/melt partitioning experiments of Drake and Weill (1975). In these experiments, sample charges were doped to ~1 wt% with Sr, Ba, rare earth elements (REE) and Y, but each change was only doped with one to four elements. Thus, these samples provide an opportunity to compare partition coefficients (Di) at natural concentration levels with those for doped concentration levels for the same composition of plagioclase, melt and temperature. Plagioclase-glass pairs of seventeen runs at four different plagioclase compositions and temperatures were analyzed by electron microprobe for major elements and some of the doped trace elements and by ion microprobe for undoped Li, Be, B, F, Mg, P, Cl, K, Sc, Ti, V, Cr, Mn, Fe, Co, Ni, Rb, Sr, Y, Zr, Nb, Cs, Ba, La, Ce, Pr, Nd, Sm, Eu, and Pb. Partitioning of the homovalent substituting ions Sr2+ and Ba2+ show no differences between doped and natural concentration levels. Ion microprobe measured Di heterovalent substituting ions REE3+ and Y3+ at natural concentration levels (0.3-3 ppm) in samples doped with wt% levels of Sr or Ba are up to three times higher than at doped concentration levels and cannot be explained by analytical artifacts. We discuss possible reasons for this. All trace element Di data show linear relationships of the forms in (Di)= a XAn + b and RT ln (Di) = a XAn + b in 0.4 < XAn < 0.8 range. Alkalies, alkaline earths, and lanthanides exhibit the same type of compositional dependence within each group of elements. Slopes a and a vary with the increase of the ionic radius within each valence group. The smaller ions of each of these groups exhibit no or positive slopes a and a; the larger ions show negative slopes. The magnitudes of the slopes increase linearly with ionic radii within the same valence group. This relationship allows extrapolation and prediction of the compositional dependence of elements of the same group whose concentrations could not be measured in this work. We present best fit approximation parameters for the RT in (Di) = aXAn + b relationship. These can be used in various petrologic applications to reconstruct the primary trace elemental composition of the parental melt from which plagioclase crystallized.
Bindeman, I. N., Davis, A. M., & Drake, M. J. (1998). Ion microprobe study of plagioclase-basalt partition experiments at natural concentration levels of trace elements. Geochimica et Cosmochimica Acta, 62(7), 1175–1193. https://doi.org/10.1016/S0016-7037(98)00047-7