Characterization of a series of absolute isotope reference materials for magnesium: ab initio calibration of the mass spectrometers, and determination of isotopic compositions and relative atomic weights

Abstract

For the first time, an ab initio calibration for absolute Mg isotope ratios was carried out, without making any a priori assumptions. All quantities influencing the calibration such as the purity of the enriched isotopes or liquid and solid densities were carefully analysed and their associated uncertainties were considered. A second unique aspect was the preparation of three sets of calibration solutions, which were applied to calibrate three multicollector ICPMS instruments by quantifying the correction factors for instrumental mass discrimination. Those fully calibrated mass spectrometers were then used to determine the absolute Mg isotope ratios in three candidate European Reference Materials (ERM)-AE143, -AE144 and -AE145, with ERM-AE143 becoming the new primary isotopic reference material for absolute isotope ratio and delta measurements. The isotope amount ratios of ERM-AE143 are n(²⁵Mg)/n(²⁴Mg) = 0.126590(20) mol mol⁻¹ and n(²⁶Mg)/n(²⁴Mg) = 0.139362(43) mol mol⁻¹, with the resulting isotope amount fractions of x(²⁴Mg) = 0.789920(46) mol mol⁻¹, x(²⁵Mg) = 0.099996(14) mol mol⁻¹ and x(²⁶Mg) = 0.110085(28) mol mol⁻¹ and an atomic weight of A_r(Mg) = 24.305017(73); all uncertainties were stated for k = 2. This isotopic composition is identical within uncertainties to those stated on the NIST SRM 980 certificate. The candidate materials ERM-AE144 and -AE145 are isotopically lighter than ERM-AE143 by −1.6‰ and −1.3‰, respectively, concerning their n(²⁶Mg)/n(²⁴Mg) ratio. The relative combined standard uncertainties are ≤0.1‰ for the isotope ratio n(²⁵Mg)/n(²⁴Mg) and ≤0.15‰ for the isotope ratio n(²⁶Mg)/n(²⁴Mg). In addition to characterizing the new isotopic reference materials, it was demonstrated that commonly used fractionation laws are invalid for correcting Mg isotope ratios in multicollector ICPMS as they result in a bias which is not covered by its associated uncertainty. Depending on their type, fractionation laws create a bias up to several per mil, with the exponential law showing the smallest bias between 0.1‰ and 0.7‰.