Analysis of dissolved iron isotopes in seawater

Iron is an important nutrient in the ocean. Measuring the stable isotopes of dissolved Fe in seawater may help to answer important biogeochemical questions such as what are the sources and sinks for Fe to the oceans, and how is Fe biologically cycled. Because Fe concentrations in seawater are very low, typically less than 1 nM, there are significant challenges both to separate and purify Fe from seawater without introducing contamination, and to accurately analyze δ 56Fe on the small quantities of Fe extracted. New techniques are presented here for separation and purification of Fe from seawater by bulk extraction onto a resin with NTA functional groups, followed by anion exchange chromatography. This method recovers 89% of the Fe from 1 L samples of seawater without causing any fractionation of Fe isotopes, with a total blank of 1.1 ± 0.6 ng Fe. To optimize the analytical procedure for small amounts of Fe, the different sources of error in measurement of δ 56Fe have been analyzed. For 252 individual analyses of standards and samples, the internal error is well described by the combination of errors from electronic noise on the detectors (Johnson noise), counting statistics, and a third source of error hypothesized to be short-timescale flicker in instrumental mass bias. With the small amounts of iron found in natural seawater samples, error is dominated by Johnson noise and counting statistics. Our analyses also include 160 pairs of “intermediate” replicates in which the same post-purification sample was measured during different analytical sessions, and 141 pairs of “external” replicate analyses for samples prepared from the same original seawater carboy but which were extracted and purified separately. The portion of overall mass spectrometry error that derives from intermediate error has been evaluated by comparing the variance in δ 56Fe for a single sample measured during multiple analytical sessions with the internal variance in δ 56Fe for the multiple cycles of data that make up each single analysis. The portion of total external error that derives from internal error was determined from variance in δ 56Fe for external replicates, compared with internal error based on the variance in cycles for each single analysis. We find that the error for multiple analyses of a sample during different analytical sessions is 1.06 times the internal error, and the external error for analysis of Fe samples which have been separately purified and extracted from the