Dy/Dy*; a tool for resolving petrogenetic processes from REE data

REE patterns are an invaluable tool for the interpretation of magmatic sources and processes. The REE as a group vary systematically in cation size (decreasing from La to Lu). As a consequence partitioning behavior, which is controlled by cation size relative to a given lattice site size, also varies systematically for a given mineral. This means that REE patterns for melts, once normalized and plotted on a log scale, are smooth lines reflecting partitioning during melt production and differentiation from a source with REE characteristics which are themselves the result of partitioning. But our qualitative recognition of processes from REE patterns may be hindered by the "crowding" which occurs when multiple patterns from a suite of rocks are plotted, and by the reduction in spread caused by use of the log scale. An alternative way of considering REE data is through simple binary diagrams - the La/Yb ratio, for instance, is widely used as a measure of LREE enrichment, and binary plots enable more data to be represented since each sample is represented by a single point rather than a line. But the La/Yb ratio cannot convey the shape of the REE pattern, so again information may be lost. We propose that the shape of a REE pattern - concave up or concave down - can be represented by the parameter Dy/Dy*. This parameter works rather like the Europium anomaly Eu/Eu* in comparing the measured Dy value with the interpolated value. Dy is a MREE, so if a REE pattern is a straight line then Dy will lie on the interpolation between La and Yb at either end of the pattern. If Dy/Dy*>1 the pattern is concave down, <1 it is concave up. Plotting Dy/Dy* against Dy/Yb proves to be a powerful way of representing REE data. On this plot, MORB and arc data form a negative array from LREE-depleted (low Dy/Yb) concave-down patterns to LREE enriched (high Dy/Yb) concave-up. Individual arc volcanoes commonly show very tight positive trends of decreasing Dy/Yb and Dy/Dy* with silica, most likely reflecting amphibole (or perhaps cpx) control. The global data set suggests that arc mantle sources vary along a fertility control trend, but magmas derived from these sources differentiate under the control of a mineral(s) with high relative MREE partitioning. Continental crust falls on these same arc differentiation trends. OIB all appear to fall to much higher Dy/Yb, indicating significant garnet control, and precluding OIB source mantle as a significant contributor to arcs.