Correction of secondary X-ray fluorescence near grain boundaries in electron microprobe analysis: Application to thermobarometry of spinel lherzolites

A correction procedure is proposed to account for the effect of secondary X-ray fluorescence near grain boundaries in electron microprobe analysis. The procedure is based on the Monte Carlo simulation method, which is used to calculate the X-ray spectrum emitted by the mineral couple (i.e., the mineral of interest with the neighboring mineral). The contribution of secondary fluorescence from the neighboring mineral, which appears in the simulated spectrum in a natural way, is then subtracted from the measured k-ratio and thereafter conventional matrix corrections are applied. The Monte Carlo simulation algorithm used is largely based on the general-purpose simulation package PENELOPE. In order to assess the reliability of this code, simulated “apparent” element profiles are compared with electron microprobe measurements found in the literature, in which the effect of secondary fluorescence was characterized and the reliability of the different assumptions underlying the proposed procedure is discussed. Finally, the procedure is used to assess data from the olivine-clinopyroxene thermobarometer in a spinel lherzolite xenolith. The application of the secondary fluorescence correction leads to: (1) higher systematic pressure estimations than those obtained from uncorrected data, with lower uncertainties; and (2) a better agreement between the olivine-clinopyroxene temperature estimations and those estimated using the two-pyroxene thermometer. Estimated P-T conditions indicate a decompressional path with a slight decrease in temperature from core to crystal rim. However, if the effect of secondary fluorescence is not taken into account, an apparent heating event is observed. LLOVET AND GALAN: CORRECTION OF SECONDARY FLUORESCENCE 122 However, measurement and quantification using L-lines is difficult and may lead to larger uncertainties. Besides, for some elements such as Ca, the use of the L-line is not a realistic alternative. Several methods have been proposed for correcting SF near phase boundaries (see Dalton and Lane 1996 and references therein). For example, Bastin et al. (1983) developed a numerical method to correct for SF effects in metal couples that can be incorporated into a matrix correction procedure. Myklebust and Newbury (1994) used the Monte Carlo (MC) simulation of electron transport in metal couples to predict SF induced at phase boundaries. These numerical corrections were applied mainly to single-element materials and, moreover, SF induced by the continuum was approximated by very crude methods or even neglected. In this work, a new correction method is proposed to account for the effect of SF near phase boundaries. The method is based on the simulation of the full X-ray spectrum emitted by the mineral couple, using the MC simulation method. The MC algorithm used is the one developed by Acosta et al. (1998), which is based on a modified version of the general-purpose simulation package PENELOPE (Baró et al. 1995). This code allows the simulation of coupled electron-photon transport in complex material structures consisting of homogeneous regions of arbitrary composition, limited by quadric surfaces. Because both electron and photon transport is considered, the contribution of SF appears in the X-ray spectrum in a natural way and also includes the continuum contribution. In this work, minerals are assumed to be separated by a planar interface; however PENELOPE allows the simulation of other geometries, such as inclined or curved boundaries (Llovet et al. 2000). To assess the reliability of this MC algorithm, simulated element profiles are compared with the EMPA measurements of Dalton and Lane (1996). Those measurements consisted of apparent Ca profiles, which are the result of the effect of SF near grain boundaries, obtained from mineral couples of olivine of increasing Fe content with diopside, and from San Carlos olivine with minerals of increasing Ca content. As an application, we address the problem of analyzing Ca in olivine close to clinopyroxene in spinel lherzolites for purposes of thermobarometry. Indeed, olivine can contain a large amount of Fe and thus emitted FeKa X-rays can be absorbed in coexisting clinopyroxene by Ca atoms, with the result of the emission of secondary CaKa X-rays. This contribution is not negligible, as the Ca content in olivine is typically below 1 wt%. Using conventional EMPA correction procedures, the Ca content in olivine was always found to depend on the distance of the electron beam point of impact from the olivine border: the closer the point, the higher the Ca concentration (Adams and Bishop 1986; Jurewicz and Watson 1988; Kölher and Brey 1990; Dalton and Lane 1996). To avoid such contamination prudently, the latter authors recommended that olivine crystals should be physically separated, mounted on a Ca-free matrix, and analyzed separately. The aim of the present study is to provide an alternative method in an attempt to avoid the physical separation of crystals, which is always tedious and sometimes difficult, e.g., when the phase of interest encloses the contaminant mineral as an inclusion (cf., Feenstra and Engi 1998). MONTE CARLO SIMULATION OF SECONDARY X-RAY