QUANTIFYING DETECTION LIMITS AND UNCERTAINTY IN X-RAY DIFFRACTION MINERALOGICAL ASSESSMENTS OF BIOGENIC CARBONATES

摘要

Mineralogical assessments of carbonates are widely used in environmental and diagenetic studies as well as to identify well-preserved archives for the reconstruction of past climate. Efforts to improve the quantification of phases in biogenic carbonates are ongoing and critical to accurately interpreting paleoclimate data. X-ray diffraction (XRD) is a standard quantitative mineralogical tool for many applications, but previous studies vary in their precision and determination of errors. Here we present a quantitative bimodal XRD calibration of marine, biogenic carbonates that we developed to assess the relative amounts of aragonite, low-Mg calcite, and high-Mg calcite in fossil coral samples. We compare the accuracy and precision of an ordinary-least-squares and a weighted-least-squares regression technique for low-concentration standard mixtures, and extend this comparison to two other published calibration studies. We also compare a polynomial fit and a spline fit for an overall calibration regression. The calibrations in this study are then compared with a model calibration using synthetic XRD data to test for sources of bias and uncertainty in the experimental calibrations. This test leads us to conclude that variation in the grain size and crystallinity of a natural aragonite standard is one source of uncertainty in these calibrations. Reproducibility of standard peak-area ratios has the largest control on calibration precision and limits of decision, detection, and quantification. Calibration accuracy is determined by the slope of the model regressions, which may be influenced by a number of factors, including mineralogical-standard crystallinity. Based on these results we include a set of best practices for XRD calibration procedures that arc dependent on the carbonate phase and precision requirements of the end-user's sample assessment. This method is made available and reproducible with an open-source R programming language code in GitHub and represents a r