Optimizing sample and spike concentrations for isotopic analysis by double-spike ICPMS

Abstract

Double spike techniques are widely used for measuring the isotopic composition of natural samples. In order to achieve the most accurate results by double spike analysis, it is important to choose an appropriate double-spike composition, analyte concentration, and spike to natural ratio (Cspk/Cnat) where Cnat is the concentration of a sample or standard with a natural abundance of the isotopes and Cspk is the concentration of an added spike with an unnatural isotope composition. Here, the effect of varying these parameters is explored using a Monte Carlo technique which simulates error from counting statistics, Johnson noise, and isobaric interferences. Typically, optimal spike composition and Cspk/Cnat are calculated under the constraint that total concentration of spike plus sample (Cspk + Cnat) must remain constant, so that as the amount of double spike is increased, the amount of sample is decreased. In practice, there is no reason for Cspk + Cnat to be held constant, because an analyst with a fixed quantity of sample may add any amount of spike to this sample as long as detector limits are not exceeded. Therefore, here, double spikes are here optimized while allowing Cspk and Cnat to vary independently. For thirty three different elements, this new approach of allowing Cspk and Cnat to vary independently led to a decrease in theoretical error of up to ∼30% in the absence of isobaric interferences. In the presence of isobaric interferences, this approach can deliver even larger improvements in accuracy and precision. Theoretical error is then compared to observed error both for δ56Fe standards and for δ56Fe, δ66Zn, and δ114Cd measured in seawater. Theoretical error and measured error for real seawater samples are highly correlated, with 78%, 85%, and 96% of observed error in δ56Fe, δ66Zn, and δ114Cd, respectively, accounted for using an error model which includes only Johnson noise and counting statistics. This confirms that such models, which minimize theoretical error, can be used to optimize spike composition, Cspk, and Cnat in order to increase accuracy and precision for analysis of natural samples.