An assessment of the accuracy of stable Fe isotope ratio measurements on samples with organic and inorganic matrices by high-resolution multicollector ICP-MS

Abstract

Multicollector ICP-MS-based stable isotope procedures provide the capability to determine small variations in metal isotope composition of materials, but they are prone to substantial bias introduced by inadequate sample preparation. Such a “cryptic” bias is not necessarily identifiable from the measured isotope ratios. The analytical protocol for Fe isotope analyses of organic and inorganic materials described here identifies and avoids such pitfalls. In medium-mass resolution mode of the ThermoFinnigan Neptune MC-ICP-MS, a 1-ppm Fe solution with an uptake rate of 50–70 μL min −1 yielded 3 × 10 −11 A on 56Fe for the ThermoFinnigan stable introduction system and 1.2–1.8 × 10 −10 A for the ESI Apex-Q uptake system. Sensitivity was increased again 3–5-fold when using Finnigan X-cones instead of the standard H-cones. The combination of the ESI Apex-Q apparatus and X-cones allowed the determination of the isotope composition on as little as 50 ng of Fe. Fe isotope compositions were corrected for mass bias with both the standard-sample bracketing (SSB) method, and by using the 65Cu/ 63Cu ratio of added synthetic copper (Cu-doping) as internal monitor of mass discrimination. Both methods provide identical results on high-purity Fe solutions of either synthetic or natural samples. We prefer the SSB method because of its shorter analysis time and more straightforward correction of instrumental mass bias compared to Cu-doping. Strong error correlations of the data are observed in three isotope diagrams. Thus, we suggest that the quality assessment in such diagrams should be performed with error ellipses rather than error bars. Reproducibility of δ 56Fe, δ 57Fe and δ 58Fe values of natural samples alone is not a sufficient criterion for accuracy. A set of tests is lined out that identify cryptic matrix effects and ensure a reproducible level of quality control. Using these criteria and the SSB correction method, we determined the external reproducibilities for δ 56Fe, δ 57Fe and δ 58Fe at the 95% confidence interval from 318 measurements of 95 natural samples to be 0.049, 0.071 and 0.28‰, respectively.