Metal–silicate partitioning experiments in the diamond anvil cell: A comment on potential analytical errors

We have performed simulations of the electron microbeam analysis of metal–silicate partitioning experiments in which a small (∼50 μm) ball of metal is embedded in a silicate matrix. This geometry approximates that found in many high pressure diamond-cell experiments aimed at determining the partition coefficients (Di) of elements between metal and silicate. The simulations involved using the (publicly available) Monte Carlo simulation package PENEPMA which is based on the general-purpose electron–photon transport code PENELOPE. The code allows complex geometrical structures to be defined by quadric surfaces, which enclose homogeneous materials specified by the user. The package simulates both electron and photon transport, keeps track of the X-ray production mechanism and includes the continuum contribution.Our principal results came from simulating the effects of secondary fluorescence on the apparent metal–silicate partition coefficient of Ni for the case of an Ni-rich metal (26.8% Ni, 73.2% Fe) embedded in a silicate matrix. If, as is frequently the case, analysis of the silicate is performed 5–10 μm from the silicate–metal interface then secondary fluorescence of Ni in the metal will increase the apparent NiO concentration in the silicate by ∼1000 ppm. In the case considered, this is 25% of the Ni content of the silicate and the fluorescence effect decreases the apparent DNi from 110 to 84. For this type of experiment, therefore, the fluorescence effect reduces the apparent partition coefficient.We performed similar simulations of the analysis of the centre of a 10 μm diameter metal ball embedded in the silicate matrix to determine apparent Si and O concentrations in the metal. In this case, because the electrons do not penetrate so far into the metal, the fluorescence effect is very small and can be neglected.Finally, we show that focused ion beam (FIB) sections, while providing the potential physically to separate metal from silicate present additional, but not insurmountable, problems of standardisation. The best results would be obtained with a section >3 μm in thickness.We recommend that, in general, experimentalists consider simulating analysis of a typical experimental product before publication of results which may be affected by secondary fluorescence.

► We model the electron microprobe analysis (EPMA) of experimental products.
► Secondary fluorescence (SF) of adjacent phases may lead to large analytical errors.
► These errors may be assessed using publicly available Monte Carlo simulation codes.
► Here a metal/silicate (M/S) Ni partitioning experiment and a FIB section are modelled.
► EPMA can underestimate the high-pressure Ni M/S partition coefficient by ∼30%.