An evaluation of methodologies for calibrating Itrax X-ray fluorescence counts with ICP-MS concentration data for discrete sediment samples

摘要

Core-scanning X-ray Fluorescence (XRF-CS) is a well-established technique for rapid ( < 30 s/interval) analysis of sediment core geochemistry at sub-mm resolution with substantially less analytical cost compared to methods that rely on physical sub-sampling. Due to issues inherent in analyzing wet sediment of heterogeneous particle size and composition with irregular surface topography using XRF, XRF-CS results are generally considered semi-quantitative. The result of early efforts to calibrate XRF-CS data with conventional geochemical results (e.g. WD-or ED-XRF, ICP-AES, ICP-MS) showed weak correlations for less abundant or poorly detectable elements, however, more recent methods have been proposed to improve accuracy. These methods include: 1) converting XRF-CS results to dry mass concentration; 2) normalizing XRF-CS data to conservative elements (Si, Ca), total counts/second, or X-ray scatter (CIR); and 3) calibration of data using multivariate analysis of elemental logratios (MLC). These approaches are not yet widely employed, and require additional testing on a variety of sediment compositions. Recently developed equipment enables analysis of discrete sediment samples, providing > 30 replicate analyses for up to 180 samples in a single XRF-CS run. These replicate measurements allow for rigorous testing of precision and accuracy of XRF-CS data. To determine the ideal method of data transformation to improve XRF-CS calibration to quantitative geochemical concentration, 100 lake sediment-surface samples collected from Harvey Lake, New Brunswick, Canada, were analyzed using Itrax-XRF-CS, and then with ICP-MS analysis after multi-acid digestion. Normalization using the CIR and correction for water content showed strong correlation coefficients (Kendall's tau) for elements with atomic number > 18 and high concentrations in the sediment. Results for lighter elements and those with lower concentrations did not perform well using these calibration methods. The MLC provi