Biogeochemical cycling of chromium and chromium isotopes in the sub-tropical North Atlantic Ocean

1 Introduction

Chromium (Cr) is a transition metal, present in typical concentrations of 0.9 to 6.5 nM in seawater (e.g., Campbell and Yeats, 1981; Cranston, 1983; Jeandel and Minster, 1987; Achterberg and van den Berg, 1997; Sirinawin et al., 2000; Connelly et al., 2006; Scheiderich et al., 2015) and has a relatively long seawater residence time of ~3000 to 9500 years (Reinhard et al., 2013; McClain and Maher, 2016; Pöppelmeier et al., 2021). In some parts of the oceans, concentrations of dissolved Cr are modestly depleted in the surface layer, and Cr has been classified as intermediate between a ‘conservative’ and ‘recycled’ element (Jeandel and Minster, 1987; Sirinawin et al., 2000). In oxic seawater, Cr(VI) is predicted to be the thermodynamically stable form of Cr (Elderfield, 1970), although relatively high concentrations of dissolved Cr(III) have been reported in some studies (e.g., Achterberg and van den Berg, 1997; Connelly et al., 2006). High concentrations of Cr(III) in oxic waters are most likely due to the presence of Cr(III)-organic complexes in the colloidal fraction (Li et al., 2022) that may not be captured by some analytical techniques (Rue et al., 1997; Wang et al., 2019; Huang et al., 2021). Under oxygen deficient conditions, Cr(III) can be the dominant Cr species, accounting for up to ~64% of the total dissolved Cr (Murray et al., 1983; Rue et al., 1997; Huang et al., 2021). However, oxygen is not the sole control on Cr speciation (Goring-Harford et al., 2018; Janssen et al., 2020); field (Connelly et al., 2006; Janssen et al., 2020) and experimental (Døssing et al., 2011; Kitchen et al., 2012) studies have suggested that high levels of biological productivity, as well as the presence of Fe(II) and organic matter in surface waters, can also facilitate Cr reduction even in O₂-replete waters (Pettine et al., 1998). At the sea surface, Cr(VI) reduction most likely proceeds via photoreduction (Kieber and Helz, 1992; Kaczynski and Kieber, 1993). Cr(VI) is highly soluble as the chromate (CrO₄^2–) ion in oxic seawater, whilst Cr(III) is particle-reactive and readily adsorbed onto solid surfaces (Rai et al., 1987; Ellis et al., 2002).

Rivers are thought to be the main source of dissolved Cr to the ocean (Bonnand et al., 2013; Reinhard et al., 2013; McClain and Maher, 2016; Pöppelmeier et al., 2021). Chromium is mainly present in the form of Cr(III) in rocks; weathering reactions driven by Mn-oxides oxidise Cr(III) as it is released from silicate rocks (Eary and Rai, 1987; Frei et al., 2014; D’Arcy et al., 2016), so Cr can be in the form of both Cr(III) and Cr(VI) in river waters that are delivered to the oceans (e.g., Cranston and Murray, 1980; Goring-Harford et al., 2020). Atmospheric deposition is a major source of many trace metals to the ocean and can result in elevated concentrations of Cr in surface seawater at some locations (Achterberg and van den Berg, 1997), but not others (Goring-Harford et al., 2018; Janssen et al., 2023). Benthic Cr fluxes can be locally important (e.g., Shaw et al., 1990; Rigaud et al., 2013; Janssen et al., 2021), and may also be important for the overall oceanic Cr inventory (Pöppelmeier et al., 2021). High temperature hydrothermal fluids from the Kermadec Arc have Cr concentrations on the order of 10 to 30 nmol kg^-1, distinctly higher than surrounding seawater (Janssen et al., 2023). However, analyses of hydrothermal plume particles indicate that they have higher Cr/Fe ratios than predicted by mixing of vent fluids and seawater, suggesting that hydrothermal activity may be a net sink of Cr as it is scavenged from seawater onto Fe-(oxyhydr)oxides (Trocine and Trefry, 1988; German et al., 1991; Rudnicki and Elderfield, 1993; Feely et al., 1996; Bauer et al., 2019). Cr is removed from the oceanic inventory through reduction of Cr(VI), scavenging onto settling particles and burial as Cr(III) in reducing and/or anoxic sediments (Reinhard et al., 2014; Gueguen et al., 2016). Therefore, marine sediments are considered as the major sink for seawater Cr. The estimated fluxes of Cr inputs to, and outputs from, the ocean reported in the literature are summarized in Table 1.

Table 1 Range of concentrations and fluxes of Cr inputs and outputs to/from the ocean.

Stable chromium isotope ratios are expressed in delta notation relative to the NBS979 standard, as:

The δ⁵³Cr values of silicate rocks show a narrow range, −0.12 ± 0.10‰ (Schoenberg et al., 2008). Groundwaters with low levels of oxygen are relatively enriched in heavy Cr isotopes (δ⁵³Cr = 0.7 to 5.8‰: Ellis et al., 2002; Izbicki et al., 2008), which is thought to reflect isotope fractionation during oxidation of Cr(III) to Cr(VI) followed by partial back reduction of the Cr(VI). Theoretical studies and laboratory experiments have shown that reduction of Cr(VI) leads to large mass-dependent fractionation, with enrichment of light Cr isotopes in the Cr(III) that forms (Ellis et al., 2002; Døssing et al., 2011; Basu and Johnson, 2012; Kitchen et al., 2012). The direction of fractionation remains the same regardless of the reductant (Fe(II), Fe(II)-bearing minerals, organic matter), but the kinetic fractionation factors are variable (δ⁵³Cr(III) – δ⁵³Cr(VI) = −1.5 to −4.2‰). To date, there have only been three reports of the Cr isotopic composition of single Cr species in seawater in the peer-reviewed literature (Wang et al., 2019; Davidson et al., 2020; Huang et al., 2021). These studies all show that Cr(III) is isotopically lighter than co-existing Cr(VI), by between −0.07 to −3.1‰ (Davidson et al., 2020; Huang et al., 2021), consistent with the theoretical and laboratory studies.

There is a rapidly growing dataset on the distribution of dissolved Cr isotopes in the open ocean with δ⁵³Cr values ranging from 0.60 to 1.71‰ (Scheiderich et al., 2015; Goring-Harford et al., 2018; Moos and Boyle, 2019; Rickli et al., 2019; Janssen et al., 2020; Moos et al., 2020; Nasemann et al., 2020; Huang et al., 2021; Janssen et al., 2021). The average seawater δ⁵³Cr value yielded from the emerging dataset is 1.04 ± 0.19‰ (1SD, n=347; see Section 4.5 for more details), slightly higher than that of the global average river input (0.49 ± 0.23‰, 1SD, n=49: Frei et al., 2014; Paulukat et al., 2015; D’Arcy et al., 2016; Wu et al., 2017; Andronikov et al., 2019). Although some studies show evidence for fractionation of Cr isotopes during estuarine mixing (Roseburrough and Wang, 2021) and others do not (Goring-Harford et al., 2020), the difference between the mean river water and seawater values nevertheless suggests that the Cr isotopic signature of seawater is likely influenced by input of Cr from other sources (such as seabed sediments) and/or is modified by fractionation of Cr as it is removed from seawater (e.g., by reduction of Cr(VI) on the surface of particles that scavenge the Cr(III) that forms). The measured δ⁵³Cr values in the authigenic fractions of reducing and anoxic marine sediments are reported to be between 0.45 and 0.61‰ in Peru margin sediments (Gueguen et al., 2016; Bruggmann et al., 2019) and between 0.38 and 0.53‰ in Cariaco Basin sediments (Reinhard et al., 2014; Gueguen et al., 2016). These values are within the range of the δ⁵³Cr values of the riverine input, but are lower than most seawater values, supporting the proposition that Cr isotope fractionation occurs during reduction and scavenging of Cr (Janssen et al., 2022).

Scheiderich et al. (2015) showed that the δ⁵³Cr composition of waters in the Arctic and Pacific Ocean was highly heterogeneous (0.61 to 1.55‰) and hypothesized that this can be attributed to Cr reduction and scavenging in surface waters and oxygen minimum zones (OMZs), with subsequent release of Cr from sinking particles in deeper waters. These processes were used to explain the inverse correlation between δ⁵³Cr value and logarithmic Cr concentration in the global ocean, from which they derived an isotope fractionation factor (ϵ, the difference between the δ⁵³Cr value of Cr(III) and the δ⁵³Cr value of Cr(VI)) of ~ –0.8‰ (Scheiderich et al., 2015). More recent studies of the Cr isotopic composition of seawater from the Pacific (including oxygen-deficient zones) and the Southern Ocean (Moos and Boyle, 2019; Rickli et al., 2019; Janssen et al., 2020; Moos et al., 2020; Nasemann et al., 2020; Huang et al., 2021; Janssen et al., 2021) also follow this ‘global correlation’. Whilst evidence for the roles of redox transformations (Nasemann et al., 2020; Huang et al., 2021; Wang, 2021), water mass mixing (Rickli et al., 2019) and biological uptake and regeneration (Janssen et al., 2020; Janssen et al., 2021) on seawater δ⁵³Cr is emerging, the underlying mechanisms that produce the apparent correlation between δ⁵³Cr and Cr concentration are still under debate.

Here, we report total dissolved Cr (Cr(VI) + Cr(III)) concentrations and total dissolved Cr isotope compositions (δ⁵³Cr) for full-depth water column profiles for three stations from a GEOTRACES transect along the 22°N sub-tropical Atlantic Ocean (Figure 1A). This fills an important knowledge gap, because few full-depth water column profiles of Cr isotopes, together with ancillary data (biogeochemical parameters, concentrations of other trace metals, etc.) that aid the interpretation of Cr behavior, are available for the Atlantic ocean. The potential sources of Cr to the North Atlantic Ocean as well as the effects of internal cycling are explicitly investigated, and by considering these new data together with other data from the literature, we are able to provide new insights as to the removal and regeneration processes that shape the global distributions of Cr and Cr isotopes in seawater.

Figure 1 (A) Locations of sampling stations (1, 4 and 6) in the sub-tropical North Atlantic Ocean. Map courtesy of http://www.geomapapp.org (B) Salinity profiles for the three sampling stations. Water masses are delimited by horizontal dashed lines. WNACW, West North Atlantic Central Water; ESACW, East South Atlantic Central Water; AAIW, Antarctic Intermediate Water; MW, Mediterranean Water; UCDW, Upper Circumpolar Deep Water; LSW, Labrador Sea Water; NEADW, North East Atlantic Deep Water; NWABW, North West Atlantic Bottom Water; AABW, Antarctic Bottom Water. Water mass analysis after Artigue et al. (2020).

2 Materials and methods

2.1 Sample collection and oceanographic setting

Samples for this study were collected from Stations 1, 4, and 6 in the sub-tropical North Atlantic (Figure 1A) during RRS James Cook cruise JC150 (UK GEOTRACES GA_pr08 process cruise) between 26^th of June and 12^th of August 2017. The water depths for the three stations were 5408 m, 3505 m and 5810 m respectively. The cruise was specifically designed to investigate macro- and micro-nutrient co-limitation on nitrogen fixation in the oligotrophic gyre; thus, nutrient and trace metal distributions (e.g., Kunde et al., 2019; Artigue et al., 2021) were carefully quantified.

Seawater was collected using pre-cleaned 10 L Ocean Test Equipment (OTE) water sampling bottles that were mounted on a titanium rosette system and deployed from a Kevlar wire. On recovery, the OTE bottles were transferred into a trace metal clean container for sub-sampling. Seawater was filtered through a Sartobran 300 (Sartorius) filter capsule (0.2 µm) under gentle pressure and was collected into 2 L acid-cleaned low density polyethylene (LDPE) bottles. Filtered seawater samples were acidified with UpA-grade hydrochloric acid (HCl, Romil) to 0.024 M, and were stored for at least a year before the Cr isotope analysis, allowing complete conversion of Cr species to Cr (III) (Semeniuk et al., 2016).

Analyses of the water masses along the transect based on combined hydrographic and nutrient data are reported in Artigue et al. (2020) and summarized in Figure 1B. Briefly, surface waters were occupied by a shallow form of Eastern South Atlantic Central Water (ESACW), with a small contribution from the Amazon plume in the west. Below the surface, the waters mainly consisted of West North Atlantic Central Water (WNACW, 100-500 m) and 13°C-ESACW (500-900 m), although at approximately 700 m water depth, Antarctic Intermediate Water (AAIW) and Mediterranean Water (MW) were present, respectively, to the west and to the east of the Mid-Atlantic Ridge (MAR). Upper Circumpolar Deep Water (UCDW) was generally observed between 900-1250 m, below which Labrador Sea Water (LSW) was centered at ~1500 m. North East Atlantic Deep Water (NEADW), which includes a contribution from Iceland-Scotland Overflow Water (ISOW), was found below LSW and was centered at ~2500 m. At Station 1, to the west of the MAR, North West Atlantic Bottom Water (NWABW) was present between 3500 and 5000 m water depth, and deeper waters (>5000 m) consisted of Antarctic Bottom Water (AABW). The deep waters at Station 6 were dominated by NEADW. Note that LSW, NEADW and NWABW all contribute to North Atlantic Deep Water (NADW). Hereafter, we define waters from below the surface mixed layer to 700 m depth as subsurface waters (including the thermocline which was located between ~300 and ~700 m; Artigue et al., 2020), waters from between 700 and 2000 m depth as intermediate waters, and waters below ~2000 m water depth as deep waters (Sarmiento et al., 2007).

2.2 Cr isotope analysis

All acids used for chemical processing were thermally distilled. Milli-Q (MQ) water was used for diluting and for cleaning. LDPE bottles and Perfluoroalkoxy (PFA) vials were thoroughly cleaned for trace metal purposes. Samples were handled under laminar flow hoods, set within Class 100 clean laboratories at the National Oceanography Centre Southampton.

Dissolved Cr concentrations were initially determined using a Mg(OH)₂ co-precipitation method (Rickli et al., 2019; Davidson et al., 2020). Approximately 50 mL of filtered seawater was transferred into an acid-cleaned centrifuge tube, weighed, and amended with 10 ng of a ⁵³Cr single spike. Ammonia solution (SpA-grade, Romil) was then added to the sample until Mg(OH)₂ formed. For 50 mL of acidified seawater (pH ~1.7), ~500 µL concentrated ammonia was required; the size of Mg(OH)₂ pellet was kept as small as possible to minimize the potential for matrix effects. After centrifugation and removal of the supernatant, the Cr precipitate was re-dissolved in 5 mL of 0.45 M HNO₃. The Cr concentration of the seawater sample was derived by isotope dilution, based on the ⁵²Cr/⁵³Cr ratio of the sample/spike mixture measured by inductively coupled plasma mass spectrometry (ICP-MS; Thermo Scientific Element). These Cr concentration data were used to estimate the volume of water required for pre-concentration and to optimize the amount of double spike added. The accuracy of the Mg(OH)₂ co-precipitation method was assessed through the analysis of (1) the NASS-6 certified reference material (National Research Council Canada), which gave Cr = 1.95 ± 0.23 nmol kg^-1 (n=1; 2SE internal error), compared to certified value of 2.17 ± 0.15 nmol kg^-1 (2SD); and (2) an OSIL Atlantic seawater salinity standard (http://www.osil.co.uk), which gave Cr = 2.96 ± 0.12 nmol kg^-1 (n=1; 2SE internal error), consistent with a previously published value for the same OSIL batch (3.06 ± 0.04 nmol kg^-1 (2SD, n=6); Goring-Harford et al., 2018).

The Cr isotope compositions of seawater samples were determined using a method adapted from Bonnand et al. (2013). Samples of 1-2 L volume were amended with a ⁵⁰Cr-⁵⁴Cr double spike to achieve optimal target isotope ratios (Goring-Harford et al., 2018) and were left to equilibrate for ~24 hrs. Information on our double spike technique is given in Bonnand et al. (2011). The sample pH was then adjusted to pH 8-9 to facilitate precipitation of Cr. A freshly prepared suspended precipitate of Fe(II) hydroxide, made by addition of ammonia to a fresh ammonium Fe(II) sulfate solution, was added to the samples (10 mL L^-1 seawater), allowing oxidation of the Fe(II) hydroxide and reduction of any remaining Cr(VI). The Fe(III) hydroxide scavenges the Cr(III), resulting in quantitative precipitation of dissolved inorganic Cr (Connelly et al., 2006).

The precipitate was separated from the solution via vacuum filtration through pre-cleaned PTFE membrane filters (1 µm, Millipore Omnipore), and was subsequently leached from the filters using 6 M HCl before being dried down and taken up in 6 mL of 7 M HCl. The Cr was first separated from the Fe by anion exchange chromatography (~2 mL of Bio Rad AG1-X8 resin loaded in a Bio Rad Poly-Prep column). The resin was extensively cleaned with concentrated HNO₃, 0.5 M HCl and concentrated HCl, and was pre-conditioned with 7M HCl. The sample was loaded in 6 mL of 7M HCl onto the resin. The eluent was collected, dried down and then reconstituted in 6 mL of 0.5 M HCl. The column was cleaned with 0.5 M HCl to remove Fe and stored in 0.5 M HCl.

Any residual salts were removed by cation exchange chromatography (2.9 mL of BioRad AG 50W-X12 resin loaded in a 30 mL PFA Savillex column; Trinquier et al., 2008). The resin was cleaned with 10 mL of 8 M HNO₃, 30 mL of 6 M HCL and 30 mL of MQ water and was pre-conditioned with 12 mL of 0.5 M HCl. The sample was loaded in 6 mL of 0.5 M HCl and the Cr was immediately eluted and collected in a 15 mL Savillex vial. A further 4 mL of 0.5 M HCl was added to the column and collected. The resin was cleaned with 6M HCl to remove the remaining cations and stored in 0.5 M HCl. The Cr fraction was evaporated to dryness and was treated with 50 μL of concentrated H₂O₂ and HNO₃, respectively, to oxidize any remaining organic material, before being dried down once again and re-dissolved in 0.45 M HNO₃.

The isotopic composition of Cr was determined by multicollector inductively coupled plasma mass spectrometry (MC-ICP-MS; Thermo Fisher Neptune Plus) at the University of Southampton, using a method similar to that described in Goring-Harford et al. (2018). Purified samples at a concentration of ~50 ppb Cr were introduced using an Aridus 3 desolvator and signals from ⁵⁰Cr, ⁵²Cr, ⁵³Cr, ⁵⁴Cr and ⁴⁹Ti, ⁵¹V, ⁵⁶Fe were quantified. Medium resolution setting was used and a mass resolution of >5000 was achieved. The NBS979 standard was analyzed after every 3 sample measurements in the analytical sequence. Each sample/standard analysis consisted of 100 individual measurements. Polyatomic interferences were avoided by making measurements on peak shoulders. The typical ion beam size was 0.15-0.24 volts per ppb for ⁵²Cr on a 10¹¹ Ω amplifier. The mean signal intensity of a blank solution that was analyzed before and after each sample/standard was subtracted. Despite the large quantities of Fe added to the samples during co-precipitation, efficient removal of Fe by anion exchange chromatography ensured that the ⁵⁶Fe/⁵⁴Cr of the samples was always <0.50 and typically around 0.03, which would have no resolvable effect on δ⁵³Cr values (Bonnand et al., 2011). The raw Cr data were corrected for mass bias using an iterative deconvolution procedure (Albarède and Beard, 2004) and for the total procedural blank contribution. The final Cr isotope value is expressed in delta notation relative to the NBS979 isotope standard.

The Cr blank for the Fe(II) co-precipitation method was relatively constant, 18 ± 0.9 ng, the majority of which (~93%) came from the Fe(II) salt (Goring-Harford et al., 2018). The δ⁵³Cr value of the ammonium Fe(II) sulphate solution used in this study was –0.16 ± 0.04 ‰ (n=1; 2SE internal error), within the error of that measured previously (–0.34 ± 0.32‰, 2SD, n=6; Goring-Harford et al., 2018). The effect of the blank on the uncertainties of Cr and δ⁵³Cr is estimated to be ± 0.005 nmol kg^-1 and ± 0.02‰ (2SD), respectively (Goring-Harford et al., 2018).

The Cr concentration of each sample was determined simultaneously with the isotope ratios using isotope dilution equations, based on the known sample volume and the quantity of added spike. The results were generally within 10% of the concentrations derived from the Mg(OH)₂ co-precipitation method; the reported Cr concentration data are from the MC-ICP-MS measurements because of the better precision.

Analyses of the NBS979 chromium isotope standard gave δ⁵³Cr = 0.00 ± 0.05‰ (2SD, n=43). The precision of the methods was further assessed through multiple analyses of an OSIL Atlantic seawater salinity standard that yielded Cr = 3.07 ± 0.058 nmol kg^-1 (2SD, n=4) and δ⁵³Cr = 0.99 ± 0.04‰ (2SD, n=4) (Supplementary Info. 1; Table S1). These data are consistent with previously reported values (Cr = 3.10 ± 0.04 nmol kg^-1, δ⁵³Cr = 0.97 ± 0.10‰, 2SD, n=6; Goring-Harford et al., 2018) for the same batch of OSIL. By compiling the repeat analyses of the OSIL samples from this study and Goring-Harford et al. (2018), we apply 2% and ± 0.06‰ as an estimate of external reproducibility for Cr and δ⁵³Cr, respectively, for all samples in this study.

As OSIL is a salinity standard and different batches do not have identical Cr concentrations and δ⁵³Cr values (Rickli et al., 2019), we undertook a cross-calibration exercise with the University of Saskatchewan in efforts to better assess the accuracy of our Cr and δ⁵³Cr data. Both groups measured three seawater samples collected by the University of Saskatchewan in the Beaufort Sea. δ⁵³Cr data for all samples were identical within the analytical uncertainty (Supplementary Info. 1; Table S2); Cr concentrations measured in our laboratory were 12-14% lower than those measured in Saskatchewan (Table S2). We have not been able to pinpoint the source of this difference, although we note that it is within the range of the uncertainty of Cr concentrations reported in a recent inter-laboratory comparison of the trace metal composition of NASS seawater (Yang et al., 2018).

2.3 Analysis of dissolved Fe concentrations and other ancillary parameters

Filtered (at 0.2 μm) samples of seawater for analysis of dissolved Fe (dFe) were acidified on board and analyzed using flow injection analysis with chemiluminescence detection (FIA-CL) inside a Class 1000 clean laboratory either onboard or at the University of Southampton, as discussed in Kunde et al. (2019). The accuracy of the method was assessed by repeat quantification of dFe in reference samples (SAFe; Johnson et al., 2007). Filtered (at 0.2 μm) seawater for analysis of soluble Fe (sFe, i.e., truly dissolved Fe) was additionally filtered in-line through 0.02 μm syringe filters (Anotop, Whatman) before it was acidified (Kunde et al., 2019). Concentrations of sFe were then determined using the same method as dFe. Concentrations of colloidal Fe (cFe, operationally defined as Fe in the size fraction 0.02 - 0.2 μm; e.g., Fitzsimmons et al., 2015) were derived from the difference between dFe and sFe. These dissolved and colloidal Fe data have been previously published in Kunde et al. (2019) and are given in Supplementary Info. 1 (Tables S1; S3).

A Seabird 911 plus conductivity, temperature and depth (CTD) profiler system together with additional sensors was attached to the titanium frame during the seawater sampling. Sensors were cross-calibrated with analyses of discrete seawater samples on board. Salinity was calibrated using an Autosal 8400B salinometer (Guildline). Chlorophyll-a (Chl-a) was measured by a fluorescence sensor that was calibrated using a fluorimeter (Turner Designs Trilogy). Dissolved oxygen (O₂) was measured by a Seabird SBE43 sensor, calibrated against a photometric automated Winkler titration system. Turbidity was monitored using the WETLabs BBRTD light scattering sensor.

Macronutrient concentrations were analyzed on board using a 5-channel (nitrate, nitrite, phosphate, silicic acid, and ammonium) segmented flow auto-analyzer (Woodward and Rees, 2001). The macronutrient data from this cruise are discussed in Artigue et al. (2020).

3 Results

Profiles of Chl-a, O₂ concentration and turbidity at Stations 1, 4 and 6 are shown in Figure 2; vertical distributions of the macronutrients phosphate (PO₄) and silicic acid (Si) at these stations are illustrated in Supplementary Info. 2 (Figure S1). Overall, there was no significant inter-site variation in these biogeochemical properties. At each station, Chl-a peaked at approximately 140 m water depth, with concentrations up to 0.4 μg L^-1. This so-called deep chlorophyll maximum (DCM) is a common feature in oligotrophic sub-tropical regions where the nutrient supply to the euphotic zone is minimal and the phytoplankton biomass and rate of primary production are low throughout the year (e.g., Pérez et al., 2006; Mignot et al., 2014). The concentrations of PO₄ and Si were the lowest within the upper ~150 m waters, and then increased with depth; PO₄ reached a concentration maximum at approximately 800 to 1000 m water depth whereas Si concentrations were highest close to the seabed. Oxygen concentrations showed minimum values of 140 μmol/kg at depths between ~700 and ~900 m. Deeper waters were well oxygenated with O₂ concentrations of >230 μmol kg^-1 below ~2000 m water depth. Levels of turbidity were highest immediately below the surface (and above the DCM), which can be attributed to input of lithogenic particles of North African dust (Ye and Völker, 2017; Kunde et al., 2019), and gradually decreased with depth. Relatively high levels of turbidity were also found close to seafloor, which is likely due to resuspension of seabed sediments (Ohnemus and Lam, 2015). At Station 4 there was a local increase in turbidity between 3300 and 3500 m water depth that delineates the particle-rich hydrothermal plume (Kunde et al., 2019) above the Snake Pit hydrothermal vent field (Beaulieu, 2015) on the Mid-Atlantic Ridge.

Figure 2 Biogeochemical properties (dissolved oxygen, Chl-a concentration and turbidity) at Stations 1, 4, and 6: (A) full water column depth profiles; (B) upper 1100 m of the water column. The deep Chl-a maximum is marked by the red horizontal dashed line, and the oxygen minimum is highlighted by the purple band at each station.

Full water column depth profiles for dissolved Fe, Cr and δ⁵³Cr at the three stations are shown in Figure 3A (data given in Supplementary Info. 1). Concentrations of dissolved Cr ranged between 1.84 and 2.63 nmol kg^-1; the vertical distribution of Cr resembled that of dissolved Fe, notably in the euphotic (up to 200 m water depth) and mesopelagic (~200 to 1100 m water depth) zones. Measured δ⁵³Cr values varied from 1.06 to 1.42‰ and the δ⁵³Cr profiles appear to mirror the shape of the Cr concentration profiles. Lowest Cr concentrations (1.84 to 2.04 nmol kg^-1) and highest δ⁵³Cr values (1.36 to 1.42‰) coincided with lowest dFe concentrations (down to ~0.4 nM) between approximately 100 to 400 m water depth. Overall, subsurface waters (between the base of the surface mixed layer and 700 m water depth) were very slightly depleted in Cr and very slightly enriched in heavy Cr isotopes (average Cr = 2.14 ± 0.15 nmol kg^-1, δ⁵³Cr = 1.36 ± 0.04‰, 1SD, n=16), relative to intermediate (Cr = 2.31 ± 0.18 nmol kg^-1, δ⁵³Cr = 1.26 ± 0.06‰, 1SD, n=6) and deep (average Cr = 2.44 ± 0.18 nmol kg^-1, δ⁵³Cr = 1.19 ± 0.11‰, 1SD, n=9) waters.

Figure 3 (A) Full water column depth profiles of Cr concentration, δ⁵³Cr values, and dissolved Fe concentrations at Stations 1, 4, and 6. (B) Profiles of Cr concentrations, δ⁵³Cr values, and dissolved and colloidal Fe concentrations for the upper 1100 m of the water column at Stations 1, 4 and 6. Location of the hydrothermal plume at Station 4 is highlighted by the grey horizontal band. Error bars on Cr and Fe concentrations are smaller than the size of symbols. Error bars on δ⁵³Cr represent the external reproducibility of the measurements (± 0.06‰, 2SD) based on the repeat analyses of OSIL Atlantic seawater salinity standard. Fe data are from Kunde et al. (2019). All Cr data are given in Supplementary Info. 1 (Table S1).

Seawater Cr concentrations and δ⁵³Cr values measured in deep waters (>2000 m water depth) in this study are generally consistent with those reported for deep waters at a site further to the south (~12 °N) in the eastern sub-tropical north Atlantic (Cr = 2.5 to 2.9 nmol kg^-1, δ⁵³Cr = 1.08 to 1.20‰: Goring-Harford et al., 2018; Supplementary Info. 2; Figure S2). By contrast, Cr concentrations and δ⁵³Cr values in waters between the base of the surface mixed layer and 2000 m water depth were respectively, generally lower and higher at our sites at ~22 °N than they were at ~12 °N (Figure S2).

Although no other Cr isotope data have been published to date for seawater samples collected within ± 10° latitude of our sites, Cr concentration data have been reported by a number of authors (Jeandel and Minster, 1987; Mugo and Orians, 1993; Sirinawin et al., 2000; Connelly et al., 2006) and are highly variable, ranging from ~2 nmol kg^-1 to almost 5 nmol kg^-1. Our data occupy the lower end of this range (Figure S2). There are numerous potential explanations for the differences between the data sets. For example: (1) The sampling stations span almost the full width of the Atlantic Ocean, crossing many oceanographic boundaries. (2) These earlier data were obtained on samples that were not collected following GEOTRACES trace metal protocols that have been established more recently. While seawater samples collected using standard Niskin bottles from a conventional rosettes do not appear to be affected by Cr contamination (e.g., Scheiderich et al., 2015; Janssen et al., 2020; Moos et al., 2020), some types of filter can be associated with high and variable Cr blanks (e.g., Yiğiterhan et al., 2011). (3) Differences in sample treatment and analysis. Chromium concentrations were determined using various different techniques that may capture different forms of Cr. The Fe(II) co-precipitation used in this study (and also by Jeandel and Minster, 1987; Connelly et al., 2006) is unlikely to capture organically-bound dissolved Cr (Goring-Harford et al., 2020), as this fraction is thought to be resistant to reduction and adsorption (Nakayama et al., 1981). By contrast, the Mg(OH)₂ co-precipitation technique is expected to capture organically-bound Cr (Moos and Boyle, 2019; Davidson et al., 2020; Huang et al., 2021) and techniques that determine Cr through complexation with various ligands may capture the more labile organic-bound Cr (Mugo and Orians, 1993; Sirinawin et al., 2000). Note however that concentrations of organically-bound Cr are usually very low (e.g., Kaczynski and Kieber, 1994). In addition, most of the data reported in Jeandel and Minster (1987) were from analysis of unfiltered water samples; a handful of samples were filtered and Cr concentrations of filtered samples were between 0.15 and 1.1 nmol kg^-1 (mean 0.62 nmol kg^-1) lower than unfiltered samples. We also note that there are wide variations between more recent and earlier Cr concentration data from other locations. For example, Mugo and Orians (1993) reported total Cr data from 3000-4000 m water depth at Station P26 in the North Pacific of between 3.53 and 4.18 nmol kg^-1 (data converted from reported values in nM to nmol kg^-1 assuming salinity = 34.67; Janssen et al., 2021), whereas Janssen et al. (2021) measured values of between 4.52 and 4.54 nmol kg^-1 for the same station over the same depth range. While there is a clear need for more concerted inter-laboratory assessments of Cr and Cr isotope data (see also Moos et al., 2020 and Huang et al., 2021), we nevertheless believe our data are analytically robust as we have carried out inter-laboratory calibration exercises as discussed in Section 2.2 together with other tests (reported in Goring-Harford et al., 2018) to validate the accuracy and precision of our data.

4 Discussion

4.1 Potential sources of Cr to the North Atlantic

Dust supply from North African deserts directly affects the subtropical North Atlantic Ocean (Jickells et al., 2005). Total (wet + dry) deposition of lithogenic particles to surface waters ranged from ~1 × 10^-5 kg m^-2 d^-1 at Station 6 to ~4 × 10^-6 kg m^-2 d^-1 at Station 1 (Artigue et al., 2021), which may dominate over biogenic particles (Ohnemus and Lam, 2015). Input of Fe and aluminum (Al) from dust deposition across the GEOTRACES GA_pr08 transect is evidenced by elevated concentrations of dissolved Fe and Al in the surface mixed layer (SML), which extends to 20 m, 50 m and 60 m water depth respectively, for Stations 1, 4 and 6 (Kunde et al., 2019; Artigue et al., 2021). Only two samples were collected for Cr analysis from within the SML (at Stations 4 and 6); these had Cr concentrations of 2.28 nmol kg^-1 (Station 4) and 2.25 nmol kg^-1 (Station 6) (Figure 3B), similar to the Cr concentration for samples collected from the upper layer of intermediate waters at these stations (2.29 nmol kg^-1 at 1001 m of Station 4 and 2.12-2.35 nmol kg^-1 at 701-901 m of Station 6). The δ⁵³Cr values of dissolved Cr from the SML were 1.32‰ (Station 4) and 1.34‰ (Station 6). The δ⁵³Cr value of atmospheric dust is expected to be similar to that of crustal rocks (−0.12 ± 0.10‰; Schoenberg et al., 2008), although dissolution promoted by strong ligands may enrich heavier Cr isotopes in the dissolved phase (δ⁵³Cr up to 1.23‰; Saad et al., 2017). For total (wet + dry) deposition fluxes of ~8 × 10^-6 kg m^-2 d^-1 (Station 4) and ~1 × 10^-5 kg m^-2 d^-1 (Station 6; Artigue et al., 2021), with a Cr loading equivalent to average upper continental crust (~100 ppm; e.g., Rudnick and Gao, 2003), and assuming 10% of this Cr is soluble (Chester and Murphy, 1990), then the annual input of dust-derived Cr to the surface mixed layer at these stations can be expected to increase the Cr concentration by ~0.01 nM. This is within the analytical uncertainty of the measured Cr concentrations in the SML. By contrast, the expected increases in concentrations of dissolved Fe and Al due to dust deposition calculated in the same way are of the order of >1 nM, which is consistent with observations (Kunde et al., 2019; Artigue et al., 2021). It thus appears that input of dissolved Cr from atmospheric dust deposition to the surface ocean was not significant, at least at the time of sampling. In this connection, it is important to note that there can be considerable variability in dust transport and deposition in this region (Artigue et al., 2021) and potentially, variations in the Cr content of aerosol particles (Rädlein and Heumann, 1995), that may lead to increased input of dust derived Cr to surface waters at other times of the year.

The neutrally buoyant hydrothermal plume above the Snake Pit vent field was identified by a turbidity anomaly at water depths of between 3300-3500 m at Station 4 (Figure 2). Dissolved Fe concentrations reached 10.7 to 27.7 nM within the plume (Table S3), much higher than background seawater (~0.67 nM; Kunde et al., 2019), suggesting a substantial contribution of dFe from the Snake Pit vent fluids that have an end-member Fe concentration of ~3.5 mM (Findlay et al., 2015). The particulate Fe content in the plume was also high, up to ~20 nM (Kunde et al., 2019). Seawater samples from within the hydrothermal plume exhibit Cr concentrations of 2.48 to 2.57 nmol kg^-1 and δ⁵³Cr values of 1.06 to 1.25‰ (Figure 3A). These Cr concentrations are slightly higher (by 0.14 - 0.23 nmol kg^-1) than a single sample collected from a similar water depth within the same water mass (3500 m, NEADW) at Station 1; the δ⁵³Cr value of the sample from Station 1 (1.17‰) is within the range of the plume samples. By contrast, seawater samples collected from a site of diffuse flow in the North Fiji basin were observed to be strongly enriched in total dissolved Cr (up to 20 nmol kg^-1) up to 400 m above the seabed relative to background values (~10 nmol kg^-1; Sander and Koschinsky, 2000). However, as discussed in Janssen et al. (2023), these results may not offer unequivocal support for hydrothermally sourced Cr. We cannot make a direct comparison between our Snakepit data and those from the North Fiji basin hydrothermal site due to differences in analytical methods, but we nevertheless assert that the hydrothermal input to the Snakepit plume can be expected to be small, for the following reasons. Firstly, reduced species of Cr in high temperature hydrothermal fluids have limited solubility (Huang et al., 2019) and recent analyses of high temperature (up to 311°C) hydrothermal fluids suggest that Cr concentrations are <30 nmol kg^-1 (Janssen et al., 2023). Secondly, the highest Fe content (dissolved + particulate) sampled in the Snakepit plume was 47.7 nM (Kunde et al., 2019), versus a background Fe content of 0.67 nM (Kunde et al., 2019). Assuming that ~25% of the Fe in the Snakepit vent fluids precipitates immediately as Fe-sulfide as it is expelled at the seabed (James and Elderfield, 1996), then the total (dissolved + particulate) Fe data indicate that the seawater:vent fluid mixing ratio is ~56,000 at plume height. Even assuming that none of the vent fluid Cr precipitated as the fluids were expelled at the seabed (cf., German et al., 1991), the expected increase in Cr in the plume due to vent fluid input would be <5 × 10^-4 nmol kg^-1, which is smaller than the analytical uncertainty of our measurements.

At all three stations, the highest Cr concentrations (and the lowest δ⁵³Cr values) were recorded in samples taken from closest to the seabed. Slightly elevated Cr concentrations in deep waters have been reported in other studies and are generally attributed to release from sediments (e.g., Murray et al., 1983; Jeandel and Minster, 1987). In support of this, Cr concentrations in pore waters of oxic sediments can be up to an order of magnitude higher than Cr concentrations in overlying bottom waters (Shaw et al., 1990; Janssen et al., 2021). Our data suggest that Cr released by sediments is isotopically light relative to overlying bottom waters, which is consistent with release of Cr from lithogenic material (that has δ⁵³Cr = −0.12 ± 0.10 ‰; Schoenberg et al., 2008) and/or from authigenic sediment phases that have δ⁵³Cr values that range from −1.2 ‰ (Bauer et al., 2019) to 0.61 ‰ (Gueguen et al., 2016). Similarly, mass balance calculations based on bottom vs. deep water Cr concentrations and δ⁵³Cr values in the Tasman Sea have shown that oxic pore waters seem to be a source of relatively isotopically light Cr to bottom waters (δ⁵³Cr = 0.34 ± 0.25 ‰; Janssen et al., 2021).

A rough estimate of the Cr flux from sediments can be obtained from the gradient of the Cr concentration in deep waters and estimated rates of turbulent mixing above the seabed (Equation 2):

where Δ[Cr] is change in Cr concentration between the deepest water sample and the next deepest water sample, Δz is depth interval separating the deepest from the next deepest water sample and D is diapycnal diffusivity. Discounting Station 1 (as no samples were collected within <30 m above the seabed), then the estimated Cr flux from sediments at Station 6 is ~5 × 10^-3 nmol cm^-2 yr^-1 (for a D value of 0.1 cm² s^-1; Polzin et al., 1997), while the estimated Cr flux from sediments at Station 4 is considerably higher, ~2 nmol cm^-2 yr^-1. Note for Station 4, we use a higher value for D (5 cm² s^-1), because mixing rates have been shown to be much higher above the Mid-Atlantic Ridge than they are above the seabed in the abyssal ocean (Polzin et al., 1997). If these fluxes are applied to the areal extent of all oxic oceanic sediments (ca. 10⁸ km²), then the estimated global benthic flux of Cr would range from 5 × 10⁶ mol yr^-1 (Station 6) to 3 × 10⁹ mol yr^-1 (Station 4), equivalent to <0.3% to ~10 times the estimated river flux (Table 1). The value for F_Cr we calculate for Station 4 is comparable to that calculated based on pore water measurements in calcareous ooze in the Tasman Sea (up to ~ 3.2 nmol cm^-2 yr^-1; Janssen et al., 2021), supporting the suggestion that sediment inputs of Cr can be important at the local scale (Janssen et al., 2021). In the case of Station 4, situated above the Snakepit hydrothermal vent field, loss of Cr from the sediments is likely fueled by oxidation of Cr(III) associated with metalliferous Fe-(oxyhydr)oxides derived from the hydrothermal plume, followed by diagenetic remobilization of the Cr(VI) that forms (Bauer et al., 2019).

4.2 Removal of Cr in subsurface waters

Lowest Cr concentrations were found in subsurface waters at between ~100 and ~500 m water depth (Figure 3B). The most Cr-depleted samples in subsurface waters had Cr concentrations of as low as 1.84 nmol kg^-1, between ~9 and 20% lower than the samples in the surface mixed layer (2.25-2.28 nmol kg^-1) and ~8 to 18% lower than the uppermost intermediate waters (2.25 ± 0.10 nmol kg^-1, 1SD, n=4). δ⁵³Cr values of the most Cr-depleted samples were also very slightly higher, by up to ~0.14‰ than the uppermost intermediate waters, which is greater than our analytical uncertainty of ± 0.06‰ (2SD). As this interval of very slightly higher δ⁵³Cr is a feature for all 3 stations, we tentatively assert that it may be real. In support of this, Jeandel and Minster (1987) also noted a weak but non-negligible subsurface Cr minimum at water depths between ~100 and 500 m in the North Atlantic Ocean (24 °N to 39 °N).

Removal of Cr does not appear to be associated with biological uptake in these oligotrophic waters. Lowest Cr concentrations occurred below the DCM (and below where macronutrient concentrations were the lowest; Figure S1), and there is no significant correlation (at p<0.05) between δ⁵³Cr and Chl-a concentration in the subsurface waters (Figure 4A). Low productivity waters in the North Pacific also showed no evidence for Cr removal (Janssen et al., 2020). Additionally, there is no significant (at p<0.05) correlation between δ⁵³Cr values and O₂ concentrations, or Cr and O₂ concentrations in the subsurface waters (Figure 4B), suggesting that there is no redox control on Cr isotope behavior in this slightly O₂ deficient depth interval (O₂ concentrations >130 μmol kg^-1). Recent studies on oxygen deficient waters (with O₂ concentrations down to ~13.2 μmol kg^-1) in the eastern sub-tropical Atlantic and the North Pacific have drawn the same conclusion (Goring-Harford et al., 2018; Moos and Boyle, 2019; Janssen et al., 2020) and reduction of Cr(VI) has only been demonstrated in waters with extremely low levels of oxygen (e.g., < 2 μmol kg^-1; Murray et al., 1983; Rue et al., 1997; Moos et al., 2020; Nasemann et al., 2020; Huang et al., 2021).

Figure 4 Cr and δ⁵³Cr as a function of (A) Chl-a concentration; (B) dissolved oxygen concentration; (C) dFe concentration; and (D) apparent oxygen utilization. SML (■), subsurface (●), intermediate (○), and deep (◊) waters are identified by the different symbols. Blue, orange, and green colors represent Stations 1, 4 and 6, respectively, as for Figure 3. Solid lines show significant correlations (p<0.05) between Cr and dFe, and δ⁵³Cr and dFe data in subsurface waters.

Figure 3 shows that the lowest Cr waters coincide with lowest dissolved and colloidal Fe concentrations, suggesting that cycling of Cr and Fe may be coupled in the subsurface waters. In this study area, depletion of dFe has been shown to be associated with removal of cFe from the water column and is not exclusively a result of biological uptake (Kunde et al., 2019). This is because: (1) experimental studies have indicated that sFe is biologically preferred over cFe (Chen and Wang, 2001; Hurst and Bruland, 2007); (2) lowest cFe concentrations occur below the DCM (~140 m water depth) which is, although not exclusively, an indicator of biomass maximum (Figure 3B); and (3) an enrichment of particulate Fe (pFe), which would capture the cellular Fe pool, was not observed (Kunde et al., 2019). The simultaneous minima of pFe and cFe (Kunde et al., 2019) suggests that cFe is not simply transferred from the dissolved to the particulate phase, but both phases are removed together as they sink through the DCM. The subsurface deficit of dFe may therefore be explained by active aggregation of colloidal Fe (e.g., Honeyman and Santschi, 1989), in the form of Fe-(oxyhydr)oxides, into filterable particles and/or scavenging of cFe onto settling dust-derived particles (Kunde et al., 2019). A similar Fe removal mechanism has also been proposed to explain changes in Fe size partitioning in other parts of the North Atlantic Ocean (Fitzsimmons et al., 2015; Ohnemus and Lam, 2015).

As Cr concentrations and δ⁵³Cr values only show a significant correlation (p<0.05) with dFe concentrations in the subsurface waters (and show no significant correlations with any of the other ancillary parameters that were measured; Figure 4), we suggest that removal of Cr proceeds via scavenging by newly-formed colloidal aggregates that subsequently fall out of the water column, as discussed above. This process has not been recognized previously, but is likely enhanced in this area of exceptionally high dust deposition (Section 4.1; Jickells et al., 2005), as the rate of colloidal aggregation is predicted to increase with particle loading due to so-called ‘colloidal pumping’ (e.g., Honeyman and Santschi, 1989). These aggregates ‘shuttle’ rapidly through the SML and into the sub-surface waters, where they are predicted to scavenge trace metals (Ohnemus and Lam, 2015), including presumably Cr. To date, only a handful of measurements of Cr concentrations in the colloidal and particulate fractions have been reported (Klun et al., 2019; GEOTRACES Intermediate Data Product Group, 2021), but these are clearly critical for proper assessment of the role of colloid-particulate interactions in Cr cycling.

Adsorption of Cr(VI) onto Fe-oxides (Pettine, 2000) or SiO₂-Al₂O₃ minerals (Frank et al., 2019) is limited in seawater because high concentrations of sulfate and/or chloride compete with chromate for adsorption sites (Pettine, 2000; Frank et al., 2019). However, Cr(III) species can be effectively scavenged by Fe-(oxyhydr)oxides (Frei et al., 2009), clay minerals and sand (Richard and Bourg, 1991) as well as biogenic particles (Semeniuk et al., 2016). Therefore, it seems likely that removal of Cr occurs mainly via scavenging of Cr(III). Cr(III) may be sourced from biologically (Connelly et al., 2006; Janssen et al., 2020; Huang et al., 2021) and/or photochemically (Kieber and Helz, 1992; Kaczynski and Kieber, 1993; Achterberg and van den Berg, 1997; Li et al., 2009) mediated reduction of seawater Cr(VI), or possibly from regeneration of Cr (III) from particles exported from the deep chlorophyll maximum (Janssen et al., 2020). If real, then the very small increase in δ⁵³Cr in the Cr-depleted waters would lend support to this idea. Cr(III) has been widely shown to be isotopically light relative to Cr(VI), hence removal of Cr(III) can be expected to leave the residual dissolved Cr pool isotopically heavy (Ellis et al., 2002; Døssing et al., 2011; Kitchen et al., 2012; Janssen et al., 2020; Huang et al., 2021). Species-specific Cr and Cr isotope measurements of water samples, as well as the particulate phase, will be essential for validating these ideas.

4.3 Regeneration of Cr in deeper waters

While there is some evidence for scavenging of Cr in subsurface waters (Section 4.2), concentrations of total dissolved Cr were highest in intermediate and deep waters, suggesting that Cr may be re-released. Although concentrations of Cr and Fe (and δ⁵³Cr and Fe) were correlated in the subsurface waters (Figure 4C), the cycling of Cr and Fe is apparently decoupled in intermediate and deep waters. In intermediate waters (700-2000 m water depth), there is a positive correlation between dFe concentrations and apparent oxygen utilization (AOU) respectively at Stations 4 and 6 (p<0.05; Supplementary Info. 2; Figure S3) that reflects remineralization of sinking organic material within the water column (Kunde et al., 2019), as also observed in other areas (Hatta et al., 2015). While such a correlation is not observed at Station 1, this is likely due to the small number of data points, the influence of atmospheric dust (Ye and Völker, 2017; Kunde et al., 2019), and/or other additional external Fe sources (Artigue et al., 2021). By contrast, there is no positive relationship between Cr concentrations and AOU, or δ⁵³Cr values and AOU, within the same depth interval at any of our sampling stations (Figure 4D). Consistent with our observation that there is no obvious biological control on Cr concentrations and δ⁵³Cr in subsurface waters, and recent observations of Cr behavior in intermediate and deep water masses in the Southern, Pacific and Atlantic oceans (Janssen et al., 2021), regeneration of Cr in intermediate waters via organic matter respiration does not appear to be an important control on Cr distributions in this part of the North Atlantic Ocean; thus, the release of Cr appears to be mechanistically different from Fe.

As removal of Cr in subsurface waters at our study sites, where lithogenic suspended particles are abundant, appears to proceed via scavenging of Cr(III) by colloid aggregates that consist of authigenic Fe-(oxyhydr)oxides and/or dust particles (Section 4.2), increased concentrations of dissolved Cr in intermediate waters could be due to re-oxidation of Cr(III) to the less particle reactive Cr(VI), potentially driven by reduction (and dissolution) of manganese(III, IV) oxides (MnO_x) that are present throughout the oxygenated water column (Jones et al., 2020). However, as concentrations of particulate Mn are very low (sub-nanomolar) in oceanic environments (Jones et al., 2020; Xiang et al., 2021), the predicted oxidation rate of Cr(III) is extremely slow, ~2 × 10^-5 nmol kg^-1 yr^-1 (van der Weijden and Reith, 1982). The residence time of fine lithogenic mineral particles (~1 to 5 μm diameter) in the upper 2000 m of the North Atlantic Ocean water column is months to years (Ohnemus and Lam, 2015; Ohnemus et al., 2019), over which time oxidation of Cr(III) can be expected to increase the Cr concentration of seawater by approximately 10^-5 to 10^-4 nmol kg^-1 in the intermediate waters. Clearly, this is too small to account for the observed 0.12 to 0.47 nmol kg^-1 increase in dissolved Cr (Section 3) in intermediate waters relative to the lowest Cr subsurface waters (i.e., those between 100 and 400 m water depth).

It thus seems more likely that regeneration of Cr occurs either via dissolution of authigenic Fe-(oxyhydr)oxides and/or dust particles and/or desorption of Cr(III) from these particles. The latter would mean that the scavenging of Cr from subsurface waters is reversible. In deeper waters, concentrations of particulate matter are lower (Figure 2), so the equilibrium between Cr(III) in the dissolved and particulate phases will shift towards the dissolved phase (e.g., Bacon and Anderson, 1982). Reversible exchange between dissolved and particulate phases has been shown to successfully reconcile the oceanic distributions of dissolved copper and zinc whose concentrations also generally increase with depth (e.g., Little et al., 2013; John and Conway, 2014; Weber et al., 2018), although unlike Cr, copper and zinc are directly influenced by biological uptake and are strongly complexed with organic ligands in seawater (e.g., Little et al., 2013).

Regeneration of Cr is unlikely to fractionate Cr isotopes to any great extent. Cr isotope fractionation caused by oxidation of Cr(III) by MnO_x minerals is, as yet poorly constrained (Bain and Bullen, 2005; Ellis et al., 2008), but recent laboratory studies suggested that differences in δ⁵³Cr values between Cr(VI) and Cr(III) (of up to 0.39‰) cannot be sustained over time (Ansari and Johnson, 2022). Adsorption and desorption processes also do not appear to fractionate Cr isotopes due to rapid isotopic exchange on particle surfaces (Ellis et al., 2004; Wang et al., 2015); thus, the isotopic compositions of dissolved and scavenged Cr(III) are generally assumed to be identical (Huang et al., 2021). Consistent with our observations, the δ⁵³Cr value of regenerated Cr, either in the form of Cr(VI) or Cr(III), can be expected to be isotopically light compared to dissolved Cr(VI) that remains in the subsurface waters.

4.4 Accumulation of Cr in deep waters and effects of water mass mixing

Excluding samples collected from within 30 m of the seabed, deep waters (>2000 m water depth) at our stations in the sub-tropical North Atlantic have lower Cr concentrations (2.40 ± 0.19 nmol kg^-1; n=7) and higher δ⁵³Cr values (1.22 ± 0.10‰; n=7) than Pacific waters (average Cr = 4.75 ± 0.42 nmol kg^-1 and δ⁵³Cr = 0.77 ± 0.10‰, 1SD, n=25; Moos and Boyle, 2019; Nasemann et al., 2020; Huang et al., 2021; Janssen et al., 2021) from >2000 m depth (including PDW, UCDW and LCDW water masses). Previous studies have suggested that Cr may accumulate in deep waters along the thermohaline flow path from the Atlantic to the Pacific Ocean (Jeandel and Minster, 1987; Janssen et al., 2021), and this behaviour has also been predicted by a model of the ocean Cr cycle (Pöppelmeier et al., 2021). Comparing our data with others from the literature (Goring-Harford et al., 2018; Moos and Boyle, 2019; Nasemann et al., 2020; Huang et al., 2021; Janssen et al., 2021), it is further apparent that Cr concentrations increase as the conventional radiocarbon (¹⁴C) age of the deep waters (Supplementary Info. 1; Table S4; Supplementary Info. 2; S1) increases, whereas the δ⁵³Cr value of total dissolved Cr decreases (Figure 5). Thus, isotopically light Cr accumulates in deep waters presumably via release from sinking particles as the deep water masses age as they are transported laterally from the Atlantic Ocean to the Southern Ocean, the South Pacific and finally into the North Pacific Ocean (DeVries and Holzer, 2019).

Figure 5 Cr and δ⁵³Cr vs radiocarbon (¹⁴C) age of deep water masses (≥ 2000 m water depth). Cr data are from: this study (sub-tropical North Atlantic); Goring-Harford et al. (2018) (eastern subtropical Atlantic); Rickli et al. (2019) (Southern Ocean); Moos and Boyle (2019) (North Pacific); Janssen et al. (2021) (Southern Ocean, North Pacific, eastern tropical South Pacific and South Pacific); Nasemann et al. (2020) (eastern tropical South Pacific); Huang et al. (2021) (eastern tropical North Pacific). ¹⁴C age calculated based on Δ¹⁴C data extracted from de Lavergne et al. (2017), as described in Supplementary Info. 2 (S1). Note that the ¹⁴C age of newly-formed North Atlantic Deep Water is ~400 years (Matsumoto, 2007). Data are provided in Supplementary Info. 1 (Table S4).

The deep waters therefore act as a reservoir for Cr ‘raining’ down in particles descending from subsurface waters. While the close similarity between deep water Cr and macronutrient concentrations (Supplementary Info. 2; Figure S4) implies that deep water Cr concentrations are to a degree regulated by remineralization of organic material, especially in high productivity regions (Janssen et al., 2021), we show here that deep water Cr concentrations also reflect ‘reversible scavenging’ of Cr(III) taken up in subsurface waters in regions where lithogenic suspended particles are abundant (Section 4.2). Note that it is also likely that deep water Cr concentrations beneath OMZs are affected by Cr released by re-oxidation of Cr(III) (Nasemann et al., 2020; Huang et al., 2021; Janssen et al., 2021). Whilst sediments can represent an important source of Cr to the deep ocean (Janssen et al., 2021), their influence on a global scale remains uncertain (Pöppelmeier et al., 2021).

Water mass mixing can also influence the spatial distribution of Cr in deep waters at local (Rickli et al., 2019; Janssen et al., 2023) and ocean basin (Janssen et al., 2021) scales. Although there is some variation in the composition of deep water masses in our study area (Figure 1), to a first approximation they can be considered to consist of North Atlantic Deep Water (NADW) and Antarctic Bottom Water (AABW). The relative contributions of these end members can be estimated from the dissolved oxygen and phosphate concentrations of the seawater samples as follows (Broecker, 1991; Sarmiento et al., 2007):

where f_NADW is the fraction of NADW, 1.59 and 0.74 are the PO₄* values (in μmol kg^-1) of, respectively, AABW and NADW (Artigue et al., 2020), and PO₄* is given by:

where [PO₄] and [O₂] are, respectively, the PO₄ and O₂ concentrations of a seawater sample and 170 is the stoichiometric ratio of oxygen to phosphate (Anderson and Sarmiento, 1994).

Excluding one sample that has f_NADW > 1 (due to input of Mediterranean seawater; Sarmiento et al., 2007) and samples collected from within 30 m of the seabed, there is a correlation (significant at p<0.1) between dissolved Cr and f_NADW (r^{2 =} 0.42) in sub-tropical Atlantic deep waters (>2000 m) (Figure S4), suggesting that water mass mixing can, to some degree, act to homogenize distinct Cr pools. However, there is no significant correlation between δ⁵³Cr and 1/Cr for the same set of samples. From Figure S4 it appears likely that the NADW end-member has lower Cr and a higher δ⁵³Cr value relative to the AABW end-member, but as no samples of ‘pure’ NADW (i.e., from ~45°N in the North Atlantic at ~4000 m water depth; Sarmiento et al., 2007) or ‘pure’ AABW (i.e., from ~35°S in the South Atlantic at ~4000 m water depth; Sarmiento et al., 2007) have been analyzed for Cr and δ⁵³Cr to date, this cannot be confirmed.

Concentrations of dissolved Cr are also positively correlated to the ¹⁴C water mass age (Supplementary Info. 1) in Atlantic deep waters (significant at p<0.1), supporting the accumulation of Cr, but the correlation (r^{2 =} 0.32) is weaker than it is for Cr- f_NADW. The Atlantic Ocean is rapidly ventilated by NADW and high-latitude source waters and has developed a relatively weaker accumulation signal of dissolved metals, such as zinc, compared to other ocean basins (Weber et al., 2018; Middag et al., 2019). This also appears to be the case for Cr.

4.5 Global correlation between seawater Cr and δ⁵³Cr

Data from this study confirm an inverse correlation between δ⁵³Cr and logarithmic Cr concentration (r^{2 =} 0.53, n=33; Figure 6A) and the slope of the linear regression is −0.79 ± 0.13‰. This is consistent with the previously proposed global δ⁵³Cr-ln[Cr] relationship for the open ocean, which was considered to reflect a Rayleigh-type fractionation of Cr isotopes as Cr is removed from seawater (Scheiderich et al., 2015; Figure 6):

Figure 6 (A) Relationship between δ⁵³Cr and ln[Cr] for new data from this study. (B) Relationship between δ⁵³Cr and ln[Cr] for the global open ocean seawater. Literature data are from: Scheiderich et al. (2015) (Arctic); Goring-Harford et al. (2018) (eastern subtropical Atlantic); Rickli et al. (2019) (Southern Ocean); Moos and Boyle (2019) (North Pacific); Janssen et al. (2020) (North Pacific); Moos et al. (2020) (eastern tropical North Pacific); Huang et al. (2021) (eastern tropical North Pacific); Nasemann et al. (2020) (eastern tropical South Pacific); Janssen et al. (2021) (subtropical Atlantic, Southern Ocean, North Pacific, eastern tropical South Pacific, South Pacific, South Atlantic). Data sets from the South Atlantic (Bonnand et al., 2013) and the eastern Pacific OMZ (Bruggmann et al., 2019; Wang et al., 2019) are not shown as more recent studies from similar locations (Huang and Boyle, 2018; Moos et al., 2020; Nasemann et al., 2020) indicate that these data may not be representative of the true seawater signal or are likely not correct.

where δ⁵³Cr₀ represents the initial Cr isotope composition, f represents the fraction of Cr remaining in seawater, and ε is the isotope fractionation factor between Cr(III) and Cr(VI). The Cr isotope fractionation factor derived from all open ocean seawater samples reported in the literature to date, including the present study, is ε = −0.71 ± 0.08‰ (r² = 0.83, n=347: Scheiderich et al., 2015; Goring-Harford et al., 2018; Moos and Boyle, 2019; Rickli et al., 2019; Janssen et al., 2020; Moos et al., 2020; Nasemann et al., 2020; Huang et al., 2021; Janssen et al., 2021; this study) (Table S5). This value is smaller than determined in the laboratory for Cr reduction by Fe(II) and/or organic matter (ϵ = −1.5 to −4.2‰: Døssing et al., 2011; Basu and Johnson, 2012; Kitchen et al., 2012), or biotic Cr reduction (ϵ = −1.6 to −4.3‰: Zhang et al., 2018; Zhang et al., 2019). It is now clear that although Cr is reduced in the euphotic zone due to biological and/or photochemical processes (e.g., Li et al., 2009; Janssen et al., 2020), as well as in the OMZs by organic matter, microbial activity, and possibly Fe(II) (Moos et al., 2020; Nasemann et al., 2020; Huang et al., 2021), the intrinsic Cr isotope fractionation is diminished as a portion of isotopically light Cr(III) remains in the dissolved phase (Moos et al., 2020; Nasemann et al., 2020; Huang et al., 2021; Wang, 2021). It is also possible that (1) the fractionation factor may be influenced by the rate of Cr reduction and removal (Jamieson-Hanes et al., 2014); and (2) scavenging/adsorption of Cr onto particles may cause a small (but as yet unconstrained) isotope fractionation (Ellis et al., 2004).

As discussed in Section 4.4, progressive regeneration of Cr that is removed from the upper water column in deep water causes accumulation of Cr with a lower δ⁵³Cr value along the global ocean deep water flow path (Section 4.4). In theory, the Rayleigh-type δ⁵³Cr-Cr curve requires that the δ⁵³Cr difference between regenerated Cr and total dissolved Cr in the water column is approximately equal to ϵ = −0.71‰. However, simple mass balance modelling (Figure S5) shows that incremental addition of isotopically light Cr (with δ⁵³Cr ≈ 0.08 to 0.36‰) to deep waters can partly reproduce the shape and slope of the global δ⁵³Cr-ln[Cr] relationship. Given the relatively long seawater residence time of Cr (~3000 to 9500 years: Reinhard et al., 2013; McClain and Maher, 2016; Pöppelmeier et al., 2021) compared to the time scale of ocean ventilation (~1000 years: e.g., DeVries and Holzer, 2019), this implies that, while local Cr and Cr isotope variations caused by removal of Cr fit with the global δ⁵³Cr-Cr relationship (e.g., Huang et al., 2021), regeneration of Cr from sinking particles and accumulation of isotopically light Cr as water masses age are also important processes that shape the large scale heterogeneity of Cr and Cr isotopes in the world’s major ocean basins and contribute to the systematic δ⁵³Cr-Cr relationship discussed by other studies.

5 Conclusions

This study reports full water column depth profiles of dissolved Cr and δ⁵³Cr at three stations in the sub-tropical North Atlantic. Considered together with ancillary data including dissolved, and colloidal Fe concentrations, turbidity, Chl-a, and dissolved oxygen as well as macronutrient concentrations, we have assessed the processes that regulate the behaviour of Cr and Cr isotopes in the modern sub-tropical North Atlantic Ocean. There is no clear evidence for significant inputs of Cr from atmospheric dust and hydrothermal vents in this area, but benthic inputs of Cr may be locally important, notably in the vicinity of hydrothermal vents. The distribution of dissolved Cr and δ⁵³Cr rather appears to principally reflect internal cycling of Cr. Subsurface waters (above 700 m depth) are very slightly depleted in Cr, and very slightly enriched in heavy Cr isotopes, relative to deeper waters. Removal of Cr in subsurface waters is not directly controlled by levels of oxygen or biological uptake in this region where biological productivity is low and waters show only modest oxygen depletion, but does coincide with lowest concentrations of dFe and cFe. We suggest that Cr, most likely in the form of relatively isotopically light Cr(III), is taken up onto colloid aggregates of Fe-(oxyhydr)oxides and/or dust particles. Despite the concurrent removal of Cr and Fe in the upper 700 m water column, the cycling of Cr and Fe is apparently decoupled in deeper waters. While regeneration of Fe occurs via respiration of biogenic particles (Kunde et al., 2019), we show that subtly increased levels of Cr (with relatively low δ⁵³Cr values) in intermediate waters most likely reflect reversible scavenging in this part of the North Atlantic Ocean. In deep waters, water mass mixing plays a role in controlling the Cr and δ⁵³Cr distributions, but regeneration of Cr from particles is more obvious at least at the global scale.

Data availability statement

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

Author contributions

WW performed the chromium analyses and data interpretation and prepared the original draft of the manuscript. RJ designed the study, mentored WW and edited subsequent drafts. HG-H supported method validation, sample analysis, and data interpretation. KK collected samples in the field, performed the iron analyses, and edited the draft. EMW performed nutrient analysis in the field. ML acquired funding for the research cruise and led sample collection. DC supported the analytical work and data interpretation. All authors commented on drafts of the manuscript and have approved the submitted version.

Funding

This cruise was funded by the NERC-funded ZIPLOc project (NE/N001125/1). WW’s PhD studentship was funded by the Chinese Scholarship Council and the Graduate School of the National Oceanography Centre Southampton; KK’s PhD studentship was funded by the Graduate School of the National Oceanography Centre Southampton.

Acknowledgments

We thank the scientific team, captain, and crew of the RRS James Cook cruise JC150 for their invaluable help in the collection of the samples and for generously sharing ancillary data that have greatly assisted the interpretation of Cr data. We thank David Janssen for invaluable comments on a previous version of this manuscript.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmars.2023.1165304/full#supplementary-material

References