Gold Open Access: This paper is published under the terms of the CC-BY license.


It has recently been demonstrated that hydrothermal vents are an important source of dissolved Fe (dFe) to the Southern Ocean. The isotopic composition (δ56Fe) of dFe in vent fluids appears to be distinct from other sources of dFe to the deep ocean, but the evolution of δ56Fe during mixing between vent fluids and seawater is poorly constrained. Here we present the evolution of δ56Fe for dFe in hydrothermal fluids and dispersing plumes from two sites in the East Scotia Sea. We show that δ56Fe values in the buoyant plume are distinctly lower (as low as −1.19‰) than the hydrothermal fluids (−0.29‰), attributed to (1) precipitation of Fe sulfides in the early stages of mixing, and (2) partial oxidation of Fe(II) to Fe(III), >55% of which subsequently precipitates as Fe oxyhydroxides. By contrast, the δ56Fe signature of stabilized dFe in the neutrally buoyant plume is −0.3‰ to −0.5‰. This cannot be explained by continued dilution of the buoyant plume with background seawater; rather, we suggest that isotope fractionation of dFe occurs during plume dilution due to Fe ligand complexation and exchange with labile particulate Fe. The δ56Fe signature of stabilized hydrothermal dFe in the East Scotia Sea is distinct from background seawater and may be used to quantify the hydrothermal dFe input to the ocean interior.


The Southern Ocean is of significant importance to the global carbon cycle and is a major sink for atmospheric carbon dioxide (CO2) (Pollard et al., 2009). The micronutrient iron (Fe) is a key regulator of primary productivity and therefore CO2 uptake in the Southern Ocean (e.g., Martin et al., 1990). The impact of past and future climate variability has been shown to be mediated by modifications to the supply of Fe to the biota in this area (e.g., Watson et al., 2000). Understanding the pathways that govern Fe supply to and removal from the Southern Ocean is therefore critical to quantifying the impact of Fe on global productivity. However, the relative importance of different sources of Fe to the oceans is not well known, and flux estimates from atmospheric dust, hydrothermal vents, icebergs, and oceanic sediments vary by orders of magnitude (Boyd and Ellwood, 2010).

A number of recent studies have demonstrated that as much as 46% of hydrothermal Fe may remain in the dissolved (<0.2 μm) phase, in the form of either colloids or organic complexes (Bennett et al., 2008; Hawkes et al., 2013, 2014; Saito et al., 2013; Fitzsimmons et al., 2014; Kleint et al., 2016). In support of this, modeling studies have shown that observations of the distribution of dissolved Fe (dFe) in the Southern Ocean can only be replicated when the Fe flux from hydrothermal sources is included (Tagliabue et al., 2010, 2014), and analyses of dFe on transects across the North Atlantic, South Atlantic, and eastern equatorial Pacific have provided evidence for advection of hydrothermal dFe for thousands of kilometers away from the mid-ocean ridge (Saito et al., 2013; Conway and John, 2014; Fitzsimmons et al., 2014; Resing et al., 2015).

Attempts have been made to parameterize hydrothermal dFe using dissolved Fe/3He ratios, but this approach is imprecise because hydrothermal fluids have widely variable Fe/3He (Tagliabue et al., 2010; Saito et al., 2013). Moreover, the controls on the proportion of hydrothermal Fe that is stabilized in the dissolved phase are unknown (Kleint et al., 2016). In principle, one way to circumvent some of these problems is by analysis of stable Fe isotopes. However, δ56Fe values of dFe {δ56Fe, where δ56Fe = [(56Fe/54Fe)sample / (56Fe/54Fe)IRMM-14 – 1] × 1000; IRMM—Institute for Reference Materials and Measurements} proximal to vent sites vary from −1.35‰ to +0.56‰ (Conway and John, 2014; Fitzsimmons et al., 2016), whereas δ56Fe values reported for hydrothermal fluids range from −0.69‰ to +0.28‰ (Beard et al., 2003; Severmann et al., 2004; Rouxel et al., 2008, 2016; Bennett et al., 2009). The likely reason for this is that the δ56Fe composition of hydrothermal dFe is modified on mixing with seawater due to precipitation of Fe sulfides, oxidation of Fe(II) to Fe(III), and Fe(III)-oxyhydroxide formation (e.g., Welch et al., 2003; Rouxel et al., 2008, 2016; Bennett et al., 2009; Roy et al., 2012).

Here we study the evolution of δ56Fe during mixing between hydrothermal fluids and seawater, presenting the first compilation of δ56Fe in hydrothermal fluids, buoyant and nonbuoyant plumes at two vent sites, E2 and E9N, located on the East Scotia Ridge (see the GSA Data Repository1 for sample site descriptions and sampling strategies). We use these data to determine the controls on the isotopic signature of hydrothermal dFe delivered to the Scotia Sea in the Southern Ocean and show that Fe isotopes are likely to be effective tracers of hydrothermal dFe throughout most parts of the world oceans.


Vent fluids with minimal seawater mixing (Mg < 1.64 mM) display Fe concentrations of 1070 μM at Site E2 and 580 μM at E9N, and have similar δ56Fe values (−0.28‰ at E2 and −0.30‰ at E9N). Water samples obtained from the buoyant part of the hydrothermal plume have the lowest δ56Fe values (as low as −1.19‰ at E2 and as low as −0.76‰ at E9N) and highest dFe concentrations (as high as 83.5 nM at E2 and as high as 23.0 nM at E9N; Fig. 1). As the hydrothermal plume is dispersed and further diluted in the neutrally buoyant part of the hydrothermal plume, dFe concentrations decrease, whereas δ56Fe values increase. Slight differences in δ56Fe values and dFe concentrations can be observed between the two vent sites: for the same degree of dilution, samples from E9N tend to have slightly higher δ56Fe and lower dFe compared to samples from E2. Total (dissolved + particulate) concentrations of Fe (tFe) are, on average, 40% ± 10% and 70% ± 30% lower at E2 and E9N, respectively, than calculated, assuming that Fe is conserved during mixing between the end-member vent fluid and seawater (Fig. 1). Partitioning of Fe between different size fractions, and δ56Fe values, may, however, be slightly modified in the interval between sampling and processing (see the Data Repository).


The missing tFe is likely to have been removed by precipitation of iron sulfides immediately on venting at the seafloor (e.g., Rudnicki and Elderfield, 1993), and higher Fe sulfide removal at E9N may be due to the slightly higher sulfide concentrations found in E9N vent fluids (James et al., 2014). Field and experimental studies have shown that light Fe isotopes are preferentially incorporated into Fe sulfides, leaving the remaining dFe isotopically heavy (Rouxel et al., 2008, 2016; Bennett et al., 2009; Roy et al., 2012). Assuming that the fractionation factor associated with Fe sulfide formation [ΔFe(II)-FeS] is +0.66‰ (Rouxel et al., 2008; Bennett et al., 2009; Roy et al., 2012), then according to the Rayleigh fractionation model, the δ56Fe value of the remaining dFe would be +0.07‰ ± 0.05‰ at E2 and +0.49‰ ± 0.05‰ at E9N. Assuming that a small part (<10%) of the dFe in the buoyant plume may be present as pyrite nanoparticles (Yücel et al., 2011) that would contribute light isotopes to the dissolved fraction, this would still produce δ56Fe values (−0.02‰ at E2, +0.39‰ at E9N) that are higher than we measure in the buoyant plumes.

The low Fe isotopic values in the buoyant part of the plume are, however, consistent with Fe sulfide formation combined with partial oxidation of Fe(II) to Fe(III). Oxidation of Fe(II) to Fe(III) results in enrichment of the heavy Fe isotopes in Fe(III) such that the δ56Fe value of the remaining Fe(II) is up to 3.56‰ lower than Fe(III) (Welch et al., 2003). Fe(III) is not stable in seawater and rapidly forms Fe(III) oxyhydroxides, which tend to aggregate and coagulate into larger particles (Field and Sherrell, 2000; Statham et al., 2005). Because these heavy Fe(III) oxyhydroxide particles are no longer part of the dFe fraction (<0.2 μm), the δ56Fe value of the remaining dFe pool decreases (e.g., Severmann et al., 2004; Bennett et al., 2009; Rouxel et al., 2016).

The evolution of δ56Fe in the buoyant plumes at E2 and E9N after sulfide formation is complete can therefore be modeled as a function of the proportion of Fe(II) oxidized to Fe(III) and the proportion of Fe(III) that leaves the dissolved phase (i.e., coagulates to form particles larger than 0.2 μm) (Fig. 2; see the Data Repository for a detailed description). According to the model, the δ56Fe values measured in the buoyant plume are consistent with oxidation of 30%–75% of Fe(II) to Fe(III), with removal of >50% of the Fe(III) produced, at both E2 and E9N. The average proportion of tFe present as dFe in the buoyant plume (∼50%; Fig. 1) is also consistent with the modeled amount of Fe(III) precipitation.

As the plume moves upward through the water column it continues to mix with seawater until the density within the plume equals that of surrounding seawater, at which point it spreads out to form the neutrally buoyant plume. During this process, dFe decreases and δ56Fe increases. The most dilute part of the neutrally buoyant plume sampled at E2 has dFe ≈15 nM and δ56Fe ≈ −0.5‰; at E9N dFe ≈7 nM and δ56Fe ≈ −0.3‰. By contrast, background seawater (Weddell Sea Deep Water) has dFe ≈ 0.7 nM and δ56Fe ≈ −0.1‰ (Abadie et al., 2017). However, although the decrease in dFe concentrations in the neutrally buoyant plume is broadly consistent with simple mixing between the buoyant plume and surrounding seawater, mixing cannot explain the evolution of δ56Fe, because measured δ56Fe values are higher than predicted (Fig. 3).

Relatively high δ56Fe values cannot be attributed to limited fallout of Fe sulfide nanoparticles (Yücel et al., 2011), because these would have to have unrealistically low δ56Fe (<−7.5‰) to reproduce the δ56Fe values we measure for dFe in the neutrally buoyant plume. Our data therefore imply that a small proportion of hydrothermal Fe is stabilized within the <0.2 μm size fraction. This Fe could be in the form of colloidal Fe oxyhydroxides and/or Fe sulfide nanoparticles, or Fe complexed by ligands (FeL) (Yücel et al., 2011; Hawkes et al., 2013; Fitzsimmons et al., 2016). Studies of Fe speciation in the neutrally buoyant plumes from the East Scotia Ridge indicate that ∼50% of dFe is in the colloidal fraction and ∼30% of dFe is in the form of FeL (Hawkes et al., 2013). Relatively high δ56Fe values in the neutrally buoyant plume may therefore reflect exchange between these dFe species, and labile Fe in the particulate fraction (e.g., neoformed FeOOH particles, adsorbed Fe; e.g., Ellwood et al., 2015).


Our data demonstrate that the δ56Fe value of hydrothermal Fe stabilized in the dissolved fraction and delivered to the East Scotia Sea in the Southern Ocean is −0.5‰ (E2) and −0.3‰ (E9N), significantly higher than the value assigned (−1.35‰) in a recent study that aimed to quantify dissolved Fe sources to a North Atlantic transect (Conway and John, 2014). Our work shows that this value is critically dependent on the δ56Fe value of the hydrothermal fluid, but also on the proportion of the Fe that precipitates as sulfides immediately on venting; the higher this is (e.g., E9N), the higher the δ56Fe value of the stabilized dFe. Nevertheless, the δ56Fe value of stabilized hydrothermal Fe from both vent sites in the East Scotia Sea is distinct from the background seawater (Weddell Sea Deep Water), as well as from other water masses surrounding mid-ocean ridges, and from other deep-ocean Fe sources (Fig. 4). There is, however, potential for overlap with the δ56Fe signature of Fe derived from reducing sediments.

While our study confirms the importance of sulfide precipitation and Fe oxidation for setting the δ56Fe value of hydrothermal iron delivered to the ocean interior, it also reveals the possibility for FeL complexation and continued exchange of Fe between dFe and particulate Fe. Understanding the physicochemical speciation of dFe remains essential for quantifying the longevity of hydrothermal iron and for modeling climate.

This work was supported by UK Natural Environment Research Council (NERC) National Capability funds and the Graduate School of the National Oceanography Centre, Southampton. Research cruises JC042 and JC055 were funded by NERC consortium grant NE/D01249X/1. We thank the Captain and crew of RRS James Cook for support on both cruises. We also thank D. Green and B. Alker for shipboard and onshore laboratory assistance, K. Nakamura for the loan of his Eh sensor, and two anonymous reviewers for their considered and constructive reviews.

1GSA Data Repository item 2017101, supplemental information on study area, methods, isotope modeling, and data, is available online at or on request from