Investigation on Sources of Contamination, for an Accurate Analytical Blank Estimation, during Measurements of the Fe Content in Seawater by Isotope Dilution Inductively Coupled Plasma Mass Spectrometry

Abstract

It is well known that direct measurement of the iron content in seawater by means of inductively coupled plasma mass spectrometry (ICPMS) is hardly possible and for open ocean water samples, typically, it requires a separation from the matrix associated to significant concentration factors. Furthermore, in the low ng kg-1 range a correct assessment of the analytical procedural blank is of crucial importance. Not only must this blank be low, but also realistic (i.e. obtained from the complete analytical sequence applied to real seawater) and reproducible. This is particularly difficult considering the ubiquity of iron and the complexity of the seawater matrix and, despite years of experiments and publications, this remains a fundamental analytical challenge and a great source of complexity for the realisation of reliable profiles of dissolved iron data. This paper re-visits the isotope dilution, ID, ICPMS procedure described by Wu and Boyle [ ], based on a multiple steps protocol, including a co-precipitation with magnesium hydroxide after ammonia loading and consecutive dissolution with hydrochloric acid, and compares two ways of evaluating realistic procedural blanks. The analytical protocol is optimised to achieve reproducible separation/pre-concentration of ~100% Fe under stable pH conditions. Typically, at ~30 ng Fe kg-1 level a concentration factor of up to 15 can be achieved leading to samples containing residual salinity ~0.06%. The first approach proposed for procedural blank determination applies mostly to Fe mass fractions >500 ng kg-1 and results from the comparison of two IDMS-based Fe content values of identical sample where the first is produced by Mg(OH)2 co-precipitation and the second – by dilution. For the second approach, adapted to lower Fe content samples (down to 3 ng Fe kg-1), the absolute procedural blank is the intercept of a linear regression between sample masses and corresponding absolute Fe contents for a given set of sample replicates. The estimated procedural blanks for both approaches were 16 +/- 12 ng kg-1 and 6 +/- 2 ng kg-1 respectively. We found that manipulations (i.e. sample handling) are by far the largest source of contamination as they contribute for ~ 75% of the total. The 2% nitric acid used to dissolve the precipitate and the instrumental background come after with respectively ~3-10% and ~5-9% contributions. The measurement process was validated through the use of reference materials, a systematic assessment of factors influencing the result, by calculating the combined measurement uncertainty and from the results obtained on a test material of a recent inter-laboratory comparison.