Clinopyroxene and garnet are believed to be the principal minerals containing trace elements in a major part of the upper mantle, and variations in their trace element partition coefficients, therefore, have a wide bearing on trace element distribution and the mobilisation of trace elements in melts or other fluids. In an attempt to determine variations in Cpx/Grt partitioning as a function of temperature, we have determined their REE abundances in several kimberlite-derived mantle peridotite xenoliths from Jagersfontein (JAG) and Matsoku (MAT) and a coexisting Cpx-Grt pair from a Monastery megacryst (MON). The trace element abundances were measured with the NERC-supported CAMECA ims-4f instrument at the University of Edinburgh, using standard procedures (8-10 nA primary beam of O- ions, collecting high-energy secondary ions, and calibrating against NBS 610 glass and natural Cpx and Grt standards).
An equilibrium temperature was estimated for each sample using the thermobarometric calibrations recommended for peridotites by Brey and Köhler (1990). Fig. 1 shows the measured partition coefficients, together with other data from the literature (Shimizu, 1975) plotted against 1/T. The data from the literature are for several high-temperature deformed peridotites (DEF) and a low-moderate temperature granular peridotite (PHN 2032). In addition, a set of Cpx/Grt partition coefficients (XP) have been calculated from experimental Cpx/melt (Grutzek et al., 1974; Hart and Dunn, 1993) and Grt/melt (Shimizu and Kushiro, 1975) data, relevant to temperatures of about 1300C. The regression lines on Fig. 1 were calculated using all the data except those from granular xenolith 2302 and the composite experimental data. They indicate, as generally expected, a decrease in the range of partition coefficients with increasing temperature. The change in partition coefficient is greater than 1.0 ln unit for LREE and Lu.
Sample PHN 2302 and the experimental data show a poor fit to the regression lines, particularly for La and Ce. A possible explanation of this for PHN 2302 might be a small amount of contamination in the bulk garnet separate used for analysis; given the extremely low concentrations of La and Ce in peridotitic garnet only a very small amount of contaminant with relatively high LREE content is needed to produce a marked effect. A similar explanation may hold for the composite experimental data, or alternatively the observed deviation from natural data may be because the partition coefficients were calculated from two independent experimental studies rather than from crystals in direct equilibrium. In either case, the present evidence suggests that caution should be used in regarding the experimental clinopyroxene/melt and garnet-melt partition coefficients as complementary for the LREE. Given the reasonable agreement of several sets of experimental Cpx/melt data, it may be appropriate to use these data in conjunction with the natural Cpx/Grt data to obtain Grt/melt partition coefficients for REE and other trace elements.
Brey, G.P. & Köhler, T., J. Petrology 31, 1353-1378 (1990).
Grutzek, M.W., Kridelbaugh, S.J. & Weill, D.F., Geophys. Res. Lett. 1, 273-275 (1974).
Hart, S.R. & Dunn, T., Contr. Miner. Petrol. 113, 1-8 (1993).
Shimizu, N., Earth planet. Sci. Lett. 25, 26-32 (1975).
Shimizu, N. & Kushiro, I., Geophys. Res. Lett. 2, 413-416 (1975).
Fig. 1: Partition coefficients for REE plotted against 1/T for samples identified in the text.