Measuring DAC metal-silicate partitioning experiments by electron microprobe: Thickness, fluorescence, and oxide spheres

The process by which the Earth's core segregated, and its resulting composition, can be inferred from the composition of the bulk silicate Earth if the partitioning of various elements into metal at relevant conditions is known. As such, partitioning experiments between liquid metal and liquid silicate over a wide range of pressures and temperatures are frequently performed to constrain the partitioning behaviour of many elements. The use of diamond anvil cell experiments to access more extreme conditions than those achievable by larger volume presses is becoming increasingly common. With a volume several orders of magnitude smaller than conventional samples, these experiments present unique analytical challenges. Typically, sample preparation is performed by FIB as a 2 mm thick slice, containing a small iron ball surrounded by a layer of silicate melt. This implies that analyses made by EPMA will be made near boundaries where fluoresced X-rays from the neighbouring phase may be significant. By measuring and simulating synthetic samples, we investigate thickness and fluorescence limitations. We find that for typical sample geometries, a thickness of 2 μm contains the entire analytical volume for standard 15kV analyses of metals. Fluoresced X-rays from light elements into the metal are below detection limits if there is no direct electron interaction with the silicate. Continuum fluorescence from higher atomic number elements from the metal into silicate poses significant difficulties [1]. This can cause metal-silicate partition coefficients of siderophile elements to be underestimated. Finally, we examine the origin and analytical consequences of oxide-rich exsolutions that are frequently found in the metal phase of such experiments. These are spherical with diameters of 100 nm and can be sparsely to densely packed. They appear to be carbon-rich and result in low analytical totals by violating the assumption of homogeneity in matrix corrections (e.g. φρz), which results in incorrect relative abundances. Using low kV analysis, we explore their origin i.e. whether they originate from quench exsolution or dynamic processes. Identifying their composition is key to understanding their origin and the interpretation of DAC experimental results.[1] Wade J & Wood B. J. (2012) PEPI 192-193, 54-58.