Three Si isotope materials have been used for an inter-laboratory comparison exercise to ensure reproducibility between international laboratories investigating natural Si isotope variations using a variety of chemical preparation methods and mass spectrometric techniques. These proposed standard reference materials are (i) IRMM-018 (a SiO₂ standard), (ii) Big-Batch (a fractionated SiO₂ material prepared at the University of California Santa Barbara), and (iii) Diatomite (a natural diatomite sample originally deposited as marine biogenic opal). All analyses are compared with the international Si standard NBS28 (RM8546) and are in reasonable agreement (<±0.22‰ 1σ_SDδ³⁰Si) given the different measurement techniques involved. These methods include both acid and alkaline dissolution/fusion, Si separation using cation exchange, selective co-precipitation, and gas-source versus plasma-ionization (high and low resolution) mass-spectrometric techniques. The average δ³⁰Si for Diatomite, IRMM-018, and Big-Batch are +1.26‰, −1.65‰ and −10.48‰, respectively, with corresponding δ²⁹Si values of +0.64‰, −0.85‰ and −5.35‰ for the same standards, respectively. For the most fractionated standard (Big-Batch), results demonstrate a kinetic mass-dependent fractionation effect for atomic Si (i.e., δ²⁹Si ∼ 0.51 × δ³⁰Si). There is almost no statistical difference between the mean values obtained by each participating laboratory, with the notable exception of the IRMM-018 standard. This effect could be caused by heterogeneity or contamination of this standard. The results for the other two standards indicate that data sets produced using any of the methods employed in this study will have similar precision and differences are limited to 0.2‰ in mean δ³⁰Si values for a given sample between laboratories, or differences of 0.13‰ in mean δ²⁹Si values.

An internal standard (IS) can be used to account for moderate, matrix-related shifts in mass bias using multi-collector inductively coupled plasma mass spectrometry through the empirical, linear relationship between measured isotope abundance ratios for different elements in ln-ln space. Unfortunately, erroneous mass bias corrected isotope abundance ratios may be returned by the model, requiring artificial adjustment of the true isotope abundance ratio of the IS. Although inadequate correction for peak tailing has been convincingly used to explain this problem, our analysis of the literature describing the development of the mass bias correction model using an IS reveals the presence of a source of systematic error. The origin of this error is purely mathematical and is eliminated in the revised model presented, in which mass bias corrected isotope abundance ratios are independent of the isotopic composition of the IS. An expression for computing the total combined uncertainty in the corrected ratio, incorporating contributions from the linear model, the isotopic reference material, and measurements of analyte element and IS in the sample, is also derived.