Samples were taken during RV SONNE cruise SO268 in the period between February and May 2019 as part of the JPI Oceans MiningImpact project (Haeckel and Linke, 2021). The cruise covered the German and Belgian license areas for manganese nodule prospection in the Clarion Clipperton Zone (CCZ), central equatorial Pacific (Figure 2; Table 1).

Surface sediments (ca. 20-30 cm) were collected with a multicorer (MUC) or ROV push core (ROV-PUC) and sediments down to ca. 5 m depth with a gravity corer (GC). Surface sediments up to ca. 75 cm were lost at the top of the GCs. The loss was variable among cores and could not be reconstructed. Therefore, all data shown here for the GCs are sampling depth within the core. Sediment cores were immediately transferred into the cold room of RV SONNE (4-7°C). The MUC bottom waters overlying the collected sediments in the liners were sampled with 0.1 M suprapure HCl and deionized water (DI; 18.2 MΩ•cm) pre-cleaned syringes and directly filtered through 0.2 µm polyethersulfone (PES; Chromafil, Macherey-Nagel) syringe filters that were pre-cleaned with DI during SO268 Leg 1 and with 0.1 M suprapure HCl and DI during SO268 Leg 2 (Table 1). Syringes were rinsed with samples first and the first drops of sample going through the filter were discarded. To sample the pore water, sediment was transferred into acid (technical grade 2% HNO₃ and 0.2% HF) and DI cleaned 50 mL centrifuge tubes with acid (10% HNO₃) and DI cleaned plastic spatulas. MUCs and the ROV PUC were sampled in 2 – 4 cm layers. For higher resolution dissolved and soluble pool comparisons, such as 136MUC and 171MUC, samples from the same depth intervals of two MUC liners were pooled to acquire sufficient pore fluids. Similarly looking cores were chosen. GCs were sampled upon eye inspection in different layers, approx. 3 samples per 1 m core section. All samples were centrifuged at 4°C, 2465 x g (3500 rpm), for 40 minutes to separate the pore water from the solid phase. The supernatant pore water was subsequently filtered through 0.2 µm PES syringe filters that were pre-cleaned with DI during SO268 Leg 1 and with 0.1 M suprapure HCl and DI during SO268 Leg 2 (Table 1). The syringes were always pre-cleaned with 0.1 M suprapure HCl and DI. Syringes were rinsed with sample first, which was discarded, and the first drops of sample filtered through the syringe filter were also discarded. For 0.02 µm filtration, sample was taken from the 0.2 µm filtrate with a clean syringe after pre-rinsing with 0.2 µm sample, and filtered through Anotop syringe filters (alumina based Anopore membrane; Whatman) directly after the 0.2 µm filtration was completed. This is well within the 24 h time range recommended by Jensen et al. (2020) to get reliable soluble and dissolved metal concentrations. Anotop filters were cleaned with DI during SO268 Leg 1 and 0.015 M suprapure HCl and DI during SO268 Leg 2. Filtration of oxic samples was performed in a clean bench. Based on the data of onboard ex-situ oxygen measurements (Haeckel and Linke, 2021), samples from suboxic GC sediments were filtered in a glove tent filled with nitrogen. Sediment subsampling of the 1 m GC liner sections was performed in ambient air as is standard procedure for GCs and has been successfully carried out in previous studies (Volz et al., 2018; Paul et al., 2019). Diffusion of oxygen is only a few mm in the 2-3 hours of sampling (Paul et al., 2019) and the few mm of air-exposed surface sediment were removed prior to sampling the sediment into the centrifuge tubes.

Trace metal samples were acidified with 1.5 µL ultrapure HCl per 1 mL sample to ~pH 1.8. Trace metal samples were stored in LDPE bottles or 5 mL PP vials cleaned with detergent, technical grade 2% HNO₃ and 0.2% HF, and 1 M suprapure HCl and DI, at ca. 4-9°C until analysis.

The pore waters in surface sediments (< 25 cm) were oxic and aerobic respiration largely controlled organic matter degradation in the sediments shallower than 1 m (Volz et al., 2018). Suboxic conditions (absence of oxygen, co-occurrence of nitrate and elevated (µM range) pore-water Mn, absence of sulfide) occurred at depths > 1 m and were variable between the study sites (Volz et al., 2018). The area was characterized by low but variable inputs of organic matter and the total organic carbon (TOC) content in surface sediments was 0.2-0.6 wt.% (Volz et al., 2018). Reductive dissolution of Mn oxides only occurred at depths of ca. 1 to 10 m (Volz et al., 2018) which is also indicated by increasing Mn concentrations to the µmol/kg range for two examples from the German area (GCs Figure 3). Manganese and Fe (oxyhydr)oxide colloids must therefore be stable in the oxic CCZ surface pore waters.

Manganese, Co, and Ni showed similar depth profiles (Figure 3) because Co and Ni are often adsorbed or incorporated onto or into Mn oxide minerals and then these elements are cycled together between the solid and dissolved as well as between the colloidal and soluble phase (Klinkhammer, 1980; Sawlan and Murray, 1983; Shaw et al., 1990; Koschinsky and Halbach, 1995). All three elements displayed a peak in the upper 6 cm in the dissolved and soluble fractions, with highest concentrations at ca. 2 cm, just below the SWI (Figure 3). This is due to the release of these metals during organic matter degradation, which has been shown for Cu and V in oxic pore waters (Klinkhammer, 1980; Heggie et al., 1986; Shaw et al., 1990; Paul et al., 2018), but rarely for Mn, Co, and Ni, and the exact mechanisms were not discussed in these previous publications. In the case of Mn, Co, and Ni, this peak was often not discoverable, potentially due to a lack of analytical sensitivity or because studies were mostly conducted in areas with a more condensed redox zonation (e.g., shelves) where the Mn, Co, and Ni peaks at the oxic-suboxic boundary dominated the profiles (Klinkhammer, 1980; Sawlan and Murray, 1983; Shaw et al., 1990). Heggie et al. (1986) observed Mn and Ni peaks in oxic pore waters at the SWI. They attributed the Mn peak to the reduction of solid-phase Mn oxides through dissolved organic matter (Heggie et al., 1986). In the solid phase of the cores presented in this study, Mn was also enriched in the upper ca. 5 cm (Supplementary Figure S1), sometimes considerably in the upper ca. 3 cm, suggesting the presence of Mn micronodules. The solid-phase Mn oxide enrichment can therefore be a source for the reductive Mn (and associated metal) release to the pore waters. Alternatively, in anoxic microenvironments, different forms of Mn oxides that are more easily reduced, or different types of organic matter have been suggested as possible reasons for the metal release at the SWI (Kalhorn and Emerson, 1984; Burdige, 1993). For some organic matter break-down processes, MnO₂ might be the preferred electron acceptor even though O₂ is present (Kalhorn and Emerson, 1984). Additionally, nitrite maxima of up to 0.6 µM in the uppermost cm were observed at the same sites (Volz et al., in prep; Haeckel and Linke, 2021), suggesting that nitrification or denitrification are occurring, releasing nitrite into the pore water at the same depth as the metals are released. The co-occurrence of nitrite and dMn peaks have previously been reported for the oxygen minimum zones of the Pacific and Arabian Sea, linking denitrification and Mn reduction based on the same noticeable profiles we observed in the CCZ pore waters (Klinkhammer and Bender, 1980; Saager et al., 1989). In the pore waters, nitrite is, however, usually produced by oxidation of ammonia and not by denitrification (Volz et al., 2018). The coinciding nitrite and Mn peaks could therefore also suggest that Mn oxides are utilized for ammonia oxidation (“Mn-anammox”) in these Mn oxide-rich sediments (Mogollón et al., 2016) (Supplementary Figure S1), subsequently releasing nitrite and sMn into the pore waters. The oxidation of ammonia by Mn oxides in the presence of oxygen has been previously described for surface sediments at the Nova Scotia margin (Luther et al., 1997).

Manganese mobilization has also been observed as reprecipitation of Mn oxide on burrow walls from mobilization of oxygen-depleted microenvironments rich in organic matter (Jung and Lee, 1999) and the enrichment of Mn in the shallow subsurface proves this (recent) Mn mobilization, likely in local microenvironments.

In summary, the release of predominantly soluble Mn, Co, and Ni just below the SWI is related to specific conditions and types of Mn oxides or organic matter that allow for reductive dissolution of Mn oxides in this shallow layer of high reactivity.

Once the labile organic matter has been consumed or Mn oxides are not the preferred electron acceptor anymore, the “usual” electron acceptor sequence consisting of O₂, NO₃^-, and MnO₂ continues downcore (Froelich et al., 1979). The data from Heggie et al. (1986), redrawn in Burdige (1993) at a higher resolution, clearly show the difference between Mn reduction at the SWI in the oxic sediments and in the “main” Mn reduction zone. As the “normal” Mn reduction zone is reached below 30 cm or at several m in the CCZ (Volz et al., 2018), the small Mn reduction zone at the SWI may play an especially important role in trace metal cycling between the pore water and water column, as diffusion from high concentration suboxic pore waters is not occurring – in contrast to sediments underlying more productive surface waters with a narrow redox zonation. However, the peak at 2 cm suggests a benthic flux to the overlying bottom water, likely fueling polymetallic nodule formation as the oxic diagenetic flux. This flux and its contribution to polymetallic nodule formation has been previously suggested for Mn (Kalhorn and Emerson, 1984; Heggie et al., 1986) but we could here also show it for Co and Ni, two of the major metals of interest in polymetallic nodules (Hein et al., 2013) – to our knowledge for the first time.

Manganese, Co, and Ni were predominantly present in the soluble phase (sMn 57-97%, sCo 88-105%, sNi 84-107%) in the upper 10 cm of the oxic pore waters (Figure 3). Soluble Mn, Co, and Ni can either be present as the free ions or bound in the small colloidal fraction < 0.02 µm e.g., to organic ligands. The formation of e.g., Mn(III)-ligands which have been previously found in the < 0.02 µm fraction could result from the interaction of Mn with organic matter (Oldham et al., 2017). These Mn(III)-ligands can be produced during ligand-promoted reductive dissolution of Mn oxides (Oldham et al., 2015; Oldham et al., 2020), potentially another pathway for Mn oxide dissolution at the SWI.

Below 10-15 cm, the concentrations of sMn decreased to 54%, sCo and sNi to ca. 70%, indicating that with increasing depth, colloids are becoming more important. Moreover, there were single layers (e.g., 171MUC 20 cm, 136MUC 8 cm) with up to 99% colloidal fraction, which were special microenvironments, potentially in the form of Mn oxides or adsorbed to Fe nanoparticles (also see Figure 6) originating from buried polymetallic nodules in the surrounding sediment. Overall, the dissolved concentrations of all three metals were, however, decreasing with depth. The smallest decrease was visible for sNi and dNi concentrations were higher than dMn and dCo concentrations, indicating that dNi cannot only be controlled by Mn (oxides). A large fraction of dNi is likely controlled by organic ligands, which can be up to 50% in seawater of the tropical South East Pacific Ocean (Boiteau et al., 2016) and might also be important in controlling pore-water Ni. Besides from Mn oxides, Ni is also released during organic matter degradation and silica dissolution at the sediment water interface (Heggie et al., 1986), which would explain the higher concentrations here.

Our results for the colloidal pool in the top 10 cm corroborate findings from a study of contaminated beach sands in which ca. 80-85% Mn and ca. 70% of Co were soluble (Huerta-Diaz et al., 2007) but it needs to be recognized that the setting was considerably different. The dominance of the soluble pool in the top 10 cm also reflects observations for seawater in which only ca. 4% of Mn is in the colloidal phase (1 kDa-0.2 µm) (Wells et al., 2000), in line with our bottom water results (sMn 96-108%). Our results for colloidal Mn at depths > 10 cm were in line with results from Homoky et al. (2011) who found on average 61% of Mn in the colloidal phase in oxic deep-water sediments around the Crozet Islands (Indian Ocean sector of the Southern Ocean). The occurrence and extent of colloidal Mn is therefore likely related to the prevalent lithology. The increase of the colloidal fraction with depth might be related to a decrease of organic ligand concentrations that can bind Mn, Co, and Ni and the substitution by larger inorganic colloids or for Ni also a shift from low to high molecular weight (HMW) organic ligands. Additionally, as suggested above, there might be a thin layer of Mn oxide reduction at the SWI, where the Mn oxide colloids are not thermodynamically stable, but below this ca. 2-6 cm layer, Mn oxide colloids are more stable again. In deeper suboxic pore waters, Mn oxide colloids are not stable (sMn=dMn and sCo=dCo; Figure 3 GCs) due to reductive dissolution of MnO₂, as has also been shown e.g., in Homoky et al. (2011).

The suboxic Mn and Co concentrations are especially high at the “No Nodule” site, up to 23 µmol/kg and 73 nmol/kg, respectively. Oxygen penetration is shallowest leading to elevated pore-water Mn (and Co) concentrations below 20 cm (Figure 3). Similar trends have been observed at another “No Nodule” site in the German area (Mewes et al., 2014), highlighting that the geochemistry of the “No Nodule” site is different from the nodule covered areas in the Belgian and German areas. The reasons could, however, not be unraveled yet and there are no clear trends from TOC or sedimentation rate data (Mewes et al., 2014) on what the reasons for the different redox zonation could be.

The dFe profiles (Figure 6) showed some exceptionally high peaks for oxic pore waters (up to 3.5 µmol/L). These maxima were only present in the dissolved fraction but not in the soluble fraction, therefore they must be due to Fe colloids present in specific layers (see 158ROV PUC8 and 171MUC, Figure 6). The sFe profiles were smoother (Figure 6), showing the same subsurface peak as Mn, Co, Ni, Cu, and V. Hence, Fe must also be released during organic matter degradation at the SWI through co-release with other metals or by reductive dissolution of Fe (oxyhydr)oxides in this microenvironment – similar to Mn (section 4.1). Reductive dissolution is likely because sFe was also present but non-reductive dissolution has been suggested to release dFe to oxic pore waters (Homoky et al., 2021). On average, only 40% of dFe was sFe (excluding the peaks where we suspect nanoparticles where the colloidal fraction is even larger), showing that the dominant fraction of dFe in these oxic pore waters was colloidal. This is consistent with findings from Homoky et al. (2021); Homoky et al. (2011), who found a large range of 17-99% of colloidal Fe in oxic pore waters. For the water column, it has been shown that dFe can be mainly colloidally bound, e.g., up to 50% in the North Pacific predominantly below 200 m depth, but this is seasonally and spatially variable throughout the oceans (Nishioka et al., 2001).

Similar to MnO₂ colloids, Fe (oxyhydr)oxides and nanoparticles can form and persist in oxic pore waters. In the CCZ sediments investigated here, Fe (oxyhydr)oxides are not even reduced at several meters depth, as only Mn reduction is occurring but not Fe reduction (Heller et al., 2018). Hence, Fe (oxyhydr)oxide colloids can persist even in the suboxic pore waters. In oxic sediments with low organic C supply and a gradually declining O₂ gradient [as is the case for the CCZ (Volz et al., 2018)], and therefore slow diagenetic reactions over years/decades/centuries, goethite is the favored Fe (oxyhydr)oxide (Homoky et al., 2011) with higher stability than the less crystalline Fe (oxyhydr)oxide ferrihydrite, which is often assumed to be the dominant Fe (oxyhydr)oxide in marine sediments (Thamdrup, 2000). As this goethite is nanoparticulate with an average size of 5 nm (van der Zee et al., 2003), it is in the colloidal range and can even occur in the operationally defined soluble phase < 20 nm. As colloids tend to aggregate (Honeyman and Santschi, 1988; Jensen et al., 2020), colloidal aggregates exceeding 20 nm, i.e. the 0.02 µm filter pore size, including colloidal goethite, can be found in our 0.02-0.2 µm fraction. In contrast to redox sensitive Fe (oxyhydr)oxides, Fe-bearing clay colloids can exist in suboxic pore waters irrespective of redox conditions. Fe-bearing smectite is a common clay colloid and especially stable in pore waters with high Si (> 500 µM) (Florindo et al., 2003; Homoky et al., 2011). The CCZ pore waters reach concentrations of 400-550 µmol/kg dSi at depth in the GCs from the German area (37GC and 157GC), and around 300-400 µmol/kg in the Belgian area (84GC) (data not shown). These concentrations are sufficiently high to make smectite colloids likely, especially at several meters depth in the GCs. Fe-bearing clay colloids might therefore be important inorganic sorbents for other trace metals once organic ligands are less abundant and MnO₂ is not stable anymore. Mineralogical data (not shown) for the sediments revealed that smectite was present in the CCZ surface sediments. The persisting Fe colloids can scavenge metals, some of which (e.g., Co) might have been previously adsorbed to MnO₂ colloids. These inorganic Fe colloids are also competing with organic ligands, which become less abundant with depth in the pore water (Paul et al., 2021).

Fe-binding organic ligand analyses in seawater detected 99% Fe(III) bound to organic ligands in the dissolved phase at analytical pH 6.9, which could be slightly less at pH 8 because of different behavior of especially inorganic Fe at the slightly lower analytical pH (Gledhill and van den Berg, 1994). Combined approaches using electrochemical analyses and sequential filtrations (0.4 and 0.02 µm) indicated that the organic ligands are found in the soluble and colloidal pools, but that an additional significant amount of inorganic Fe colloids must be present, which could potentially be Fe (oxyhydr)oxides (Cullen et al., 2006; Fitzsimmons et al., 2015). The differentiation between organic and inorganic ligands/sorbents in seawater has not been straight forward (Boye et al., 2010) and the most prominent and likely explanation is therefore that Fe (oxyhydr)oxides in combination with organic matter are the predominant colloid (Raiswell and Canfield, 2012). There is a continuum of organic-inorganic-nanoparticle-colloid formation and how the metals are bound in each step (Aiken et al., 2011). Processes involving organic and inorganic material include sorption of organic material onto inorganic surfaces and heterogenic precipitation of organic-inorganic clusters/nanoparticles/colloids (Aiken et al., 2011). Adsorption of organic matter onto colloids can change their surface charge and alter their reactivity (Tiller and O’Melia, 1993; Aiken et al., 2011). A complex combination of organic-inorganic ligand complexation of Fe in pore waters is likely and a lot of questions remain about the nature of Fe colloids. More research is needed to explain the variability of the Fe colloidal pool which was also visible in our sFe and dFe data (Figure 6).

Copper was predominantly soluble in the upper 10 cm of the pore waters and predominantly colloidal at greater depths (Figure 4). Huerta-Diaz et al. (2007) found that the subsurface peak of Cu was only present in the small colloidal fraction (3 kDa-0.45 µm) but not in the truly dissolved phase (< 3 kDa). Therefore, Cu released during organic matter degradation at the SWI appears to be stabilized in the small colloidal size fraction, but likely smaller than 0.02 µm as our results here suggest.

Dissolved and soluble Cu concentrations generally showed the same trends in the depth profiles (Figure 4), but the percentage of soluble Cu decreased with increasing depth at all sites. While ca. 80-100% of dCu was sCu in the top 5-10 cm at all sites, around 10 cm the sCu made up only ca. 60% of the dCu pool. With increasing depth, the fraction of colloidal Cu became dominant, which was already at ca. 15 cm depth in the Belgian area and below roughly 1 m in the German area. This was similar to the behavior of Mn, Co, and Ni, but the colloidal fraction was larger at depth for Cu.

In the German area the colloidal Cu fraction increased at depth, in the suboxic zone (GCs), to up to 90% (157GC, Figure 4). In the Belgian area (136MUC), sCu was already considerably decreased to 23% within the MUC in the oxic pore water. Therefore, the change in Cu complexation was likely not related to redox conditions but to a decreasing and more refractory organic matter content with depth. The shift from soluble to the colloidal pool might be due to the lack of low molecular weight (LMW) organic ligands that keep Cu in the soluble pool. As shown in Figure 1, organic ligands are predominantly present in the soluble fraction. In surface seawater, Cu is up to 99% complexed by organic ligands (Coale and Bruland, 1988). Recently, it was also shown that deep-sea surface pore water Cu is predominantly complexed by organic ligands (Paul et al., 2021), which had previously been reported for Chesapeake Bay pore waters (Skrabal et al., 1997). The analysis of the upper 20-30 cm pore waters showed, however, that the ligand excess decreases with depth similar to dCu profiles, with highest concentrations in the upper 10 cm (Paul et al., 2021). While still > 99% Cu was complexed by organic ligands, the overall ligand concentration decreased with depth. Our findings here corroborate that Cu complexation changes and that while organic ligands (small colloidal size fraction) might become less available, larger inorganic colloids or HMW organic ligands gain importance as Cu ligands. The shift from predominantly soluble to colloidal Cu does not correspond to the redox zonation, so that the physical speciation of Cu might not be related to differences in redox zonation, as it is partly the case for Mn and Fe, but to the availability of small organic ligands. In the deeper layers, Cu could be partially adsorbed to Fe colloids, which are ubiquitous as shown above (Figure 6), but dCu concentrations exceed dFe concentrations, requiring another Cu host-phase or HMW ligand.

Pore-water concentrations of As and V in surface sediments were elevated by a factor of 1.3 and 6, respectively, in the upper 2 cm compared to bottom seawater. Vanadium is therefore released during organic matter degradation at the SWI, like Mn, Ni, Co, Fe, and Cu. This is in line with results from other studies and which suggested that V is stabilized by organic material in this layer as V⁴⁺ (Wehrli and Stumm, 1989; Emerson and Huested, 1991). The reduced V species (often present as VO²⁺ or VO(OH)⁺) is more particle reactive and likely to bind to reactive surfaces (Wehrli and Stumm, 1989), wherefore V could also be released from MnO₂ during reductive dissolution at the SWI as explained in section 4.1. The same is possible for As, which can be cycled with Mn as well (Telfeyan et al., 2017) and usually shows a small peak in the upper 2-4 cm (Figure 7).

The dissolved and soluble concentrations for Mo, V, U, As, and Cd were indistinguishable (Figures 5, 7, 8). They are predominantly present as negatively charged complexes and stable in their oxidized, soluble form: Mo⁶⁺ (as MoO₄^2-), U⁶⁺ (as UO₂(CO₃)₃^4-), V⁵⁺ (as H₂VO₄^- or HVO₄^2-), As⁵⁺ (Andreae, 1979; Beck et al., 2008), therefore not found in the colloidal fraction. Vanadium could also be present in the reduced form stabilized by dissolved organic complexes, as described above (Wehrli and Stumm, 1989; Emerson and Huested, 1991). The consistent concentration of dV and sV with depth below ca. 4 cm – unlike for Cu, for which the sCu fraction decreases with depth – could, however, suggest inorganic speciation as more likely. Vanadium does not seem to be affected by changes in organic ligand availability. Vanadium was also found to be in the soluble phase by Huerta-Diaz et al. (2007), while Mo, U, and As were not studied there. In contrast to Huerta-Diaz et al. (2007), Cd is fully soluble in the pore waters studied here, while Cd was up to 60% colloidal in pore waters in the literature comparison conducted by the authors.