Redox control on the tungsten isotope composition of seawater

Significance

The fate and transport of tungsten (W) in aquatic environments is still poorly constrained. To identify the processes that control the abundance of dissolved W, we applied a sophisticated analytical approach that enables the accurate determination of the seawater W isotopic composition. Our results indicate that the removal of W from seawater mainly occurs via adsorption onto oxide minerals. The marine inventory of W is therefore intimately linked to the areal extension of oxic marine conditions. Concurrently, the limited scavenging of W in anoxic marine settings seems a unique characteristic of W, highlighting that W isotopes can help to reconstruct the earliest rise of oceanic oxygen in Earth’s history.

Abstract

Free oxygen represents an essential basis for the evolution of complex life forms on a habitable Earth. The isotope composition of redox-sensitive trace elements such as tungsten (W) can possibly trace the earliest rise of oceanic oxygen in Earth’s history. However, the impact of redox changes on the W isotope composition of seawater is still unknown. Here, we report highly variable W isotope compositions in the water column of a redox-stratified basin (δ^186/184W between +0.347 and +0.810 ‰) that contrast with the homogenous W isotope composition of the open ocean (refined δ^186/184W of +0.543 ± 0.046 ‰). Consistent with experimental studies, the preferential scavenging of isotopically light W by Mn-oxides increases the δ^186/184W of surrounding seawater, whereas the redissolution of Mn-oxides causes decreasing seawater δ^186/184W. Overall, the distinctly heavy stable W isotopic signature of open ocean seawater mirrors predominantly fully oxic conditions in modern oceans. We expect, however, that the redox evolution from anoxic to hypoxic and finally oxic marine conditions in early Earth’s history would have continuously increased the seawater δ^186/184W. Stable W isotope compositions of chemical sediments that potentially preserve changing seawater W isotope signatures might therefore reflect global changes in marine redox conditions.

Tungsten (W) belongs to the transition group VI elements and is dissolved as the tetrahedral oxyanion tungstate (WO₄²⁻) in the modern oceans. Despite its relatively low concentration of 0.041 to 0.067 nM, a moderately long residence time between 14,000 to 61,000 y suggests a homogenous distribution of W in the oceans (1–3). Riverine and hydrothermal sources, assumed to be the main inputs for marine W, have dissolved W concentrations that may be orders of magnitudes higher (rivers: 0.02 to 1029 nM; hydrothermal fluids: 0.22 to 123 nM), arguing for an efficient scavenging of W in estuarine and marine environments (3–6). While the stable W isotope composition of rivers and hydrothermal fluids is still unknown, a pioneering study of Fujiwara et al. (7) indicates that the seawater stable W isotope composition of the Northern Pacific is homogeneous with δ^186/184W of +0.55 ± 0.12 ‰ (Eq. 1). Thus, the seawater dissolved W inventory is isotopically significantly heavier than modern igneous rocks [δ^186/184W of +0.096 ± 0.076 ‰ (8, 9)], the ultimate source for marine W. This suggests stable W isotope fractionation during processes such as weathering, hydrothermal alteration of the oceanic crust, or the scavenging of dissolved marine W.

In euxinic environments, the sequestration of dissolved W is limited because of its increased solubility as thiotungstate species (WO_xS_4−x²⁻) that form when WO₄²⁻ is exposed to elevated H₂S_(aq) levels (e.g., ref. 10). However, particle shuttling by Mn-oxides and Fe-hydroxides acting at pelagic redoxclines efficiently scavenges dissolved WO₄²⁻ and may cause strong authigenic enrichments of W in modern chemical sediments (e.g., ref. 3). Adsorption of WO₄²⁻ onto these (hydr)oxides triggers a change in coordination from tetrahedral to octahedral (11). Because the bonding in octahedral coordination is longer and weaker, isotopically light W is preferentially adsorbed (12). Kashiwabara et al. (2017) experimentally determined similar stable W isotopic differences between dissolved and adsorbed W species for Mn-oxides and Fe-hydroxides, respectively, with Δ^186/184W_Mn-oxides = δ^186/184W_dissolved − δ^186/184W_adsorbed = 0.59 ± 0.14 ‰ and Δ^186/184W_{Fe-hydroxides} = δ^186/184W_dissolved − δ^186/184W_adsorbed = 0.51 ± 0.06 ‰ (12). Thus, adsorption processes in oxic marine environments may provide the key mechanism for the build-up of an isotopically heavy ocean.

Molybdenum (Mo), the geochemical twin of W, also exists as a dissolved oxyanion (MoO₄²⁻) species and is homogeneously distributed in the modern ocean with a higher concentration of around 105 nM (13). As for W, the stable Mo isotope composition of seawater [δ^98/95Mo = 2.34 ± 0.10 ‰ (14, 15)] is distinctly heavy compared to the bulk crust [δ^98/95Mo = +0.47 ± 0.12 ‰ (16)]. In analogy to WO₄²⁻, adsorption of MoO₄²⁻ onto Mn-oxides also causes a change in coordination from tetrahedral to octahedral. However, there are some key differences in the redox sensitivity of W and Mo. During adsorption of MoO₄²⁻ onto Fe-hydroxides, the tetrahedral coordination is partially retained (11). The experimentally determined isotopic difference between dissolved MoO₄²⁻ and adsorbed Mo species is therefore significantly smaller for Fe-hydroxides [Δ^98/95Mo_{Fe-hydroxides} = δ^98/95Mo_dissolved − δ^98/95Mo_adsorbed = 1.11 ± 0.15 ‰ (17)] than for Mn-oxides [Δ^98/95Mo_Mn-oxides = δ^98/95Mo_dissolved − δ^98/95Mo_adsorbed = 2.67 ± 0.18 ‰ (18)]. In euxinic environments, MoO₄²⁻ is successively thiolated with increasing H₂S_(aq), forming predominantly thiomolybdate (MoS₄²⁻) at H₂S_(aq) > 11 µM (19). In contrast to WS₄²⁻, MoS₄²⁻ is particle reactive and readily removed from solution by forming Fe–Mo–S clusters (20). In restricted and euxinic basins such as the modern Black Sea, the near quantitative sequestration of MoS₄²⁻ results in the preservation of the seawater Mo isotopic signature in Black Sea sediments (21). However, incomplete scavenging of MoO_xS_4−x²⁻ species in less restricted and/or weakly euxinic environments with H₂S_(aq) < 11 µM leads to the preferential burial of isotopically light Mo (21, 22). Therefore, adsorption processes in oxic marine environments but also the incomplete sequestration of Mo in euxinic settings may increase the δ^98/95Mo of seawater. On the other hand, temporal variations in the δ^186/184W of seawater and authigenic sediments might be more intimately linked to adsorption and deposition of oxide minerals. The difference in sensitivity to changing redox environments suggests that combined W and Mo proxy data may be a powerful way to trace subtle changes in oxygenation levels of the early surface Earth, particularly if further combined with other redox-sensitive trace metal proxies. Increasingly, multiple-proxy approaches are being used to remove the ambiguities that may be associated with individual paleoredox proxies (e.g., refs. 23–26). Thus, stable W isotope analyses may represent a complementary tool for paleo-redox reconstructions that is particularly sensitive to oxide mineral formation in oxic marine environments. However, future paleo-redox applications require a better understanding of the processes that fractionate W isotopes and their definite implication on the seawater and sedimentary δ^186/184W as offered by modern systems characterized by changing redox conditions.

In this study, we present W concentrations and stable W isotope compositions of oceanic water column profiles from the Southern Atlantic Ocean, the South China Sea in the Western Pacific Ocean, and from the Landsort Deep, a redox-stratified basin within the more restricted Baltic Sea (SI Appendix, Fig. S1). The dataset is complemented by stable W isotope compositions of suspended particles that form at the pelagic redoxcline of the Landsort Deep and euxinic porewaters from this site. Our results confirm the initial assumption of a deep ocean homogeneous in isotopically heavy W. Furthermore, we highlight the major redox-related processes that cause variations in the abundance and stable isotope composition of marine W. Overall, these findings provide the initial framework essential for the future application of stable W isotopes as a paleo-redox proxy.

Results and Discussion

To further constrain the δ^186/184W of open ocean seawater, samples were analyzed from the Southern Atlantic and the South China Sea. The δ^186/184W and W concentrations of Southern Atlantic seawater are relatively constant throughout the water column (+0.510 to +0.579 ‰ and 0.049 to 0.054 nM; Fig. 1A). We note that very small differences between Antarctic Intermediate Water (+0.510 to +0.543 ‰; 0.053 to 0.054 nM) and subjacent North Atlantic Deep Water (+0.553 to +0.579 ‰; 0.049 nM) can be barely resolved. In contrast, Western Pacific waters from the South China Sea show a clearly resolvable variability in δ^186/184W and W concentrations. Surface waters exhibit highest W concentrations (up to 0.067 nM) and correspondingly have the lowest δ^186/184W (as low as +0.445 ‰; Fig. 1B). With increasing water depth, W concentrations (as low as 0.050 nM) and δ^186/184W (up to +0.556 ‰) approach an Atlantic-like open ocean value. Importantly, deep water masses in the South China Sea reflect open Pacific waters that intrude through the Luzon Strait as indicated by Pacific-like ε_Nd values (27). However, the proximity of this site to the continent (SI Appendix, Fig. S1) makes it prone to riverine discharges (e.g., Pearl River) to surface waters, showing a slightly lower salinity (Dataset S1) and lower ε_Nd values (27). Moreover, elevated concentrations of dissolved Mn in surface waters (Dataset S1) might indicate redissolution of particles enriched in Mn-oxide supplied by rivers or aerosols and the concurrent release of adsorbed W. Thus, increasing W concentrations and decreasing δ^186/184W in shallower waters might reflect binary mixing of W supplied from rivers/aerosols and the open Pacific, respectively. Excluding the locally affected surface seawater samples from the South China Sea, the open ocean seawater samples have an average δ^186/184W of +0.543 ± 0.046 ‰ (2SD; n = 10) and an average W concentration of 0.050 ± 0.007 nM (2SD; n = 10; Datasets S1 and S2). The new seawater δ^186/184W data from both the South Atlantic Ocean and the South China Sea are perfectly consistent but more precise than a previous estimate on Northern Pacific seawater [+0.55 ± 0.12 ‰ (7)] and clearly indicate a relatively homogeneous seawater stable W isotope composition beneath the permanent thermocline.

Fig. 1.

Seawater W concentrations (open symbols) and W isotope compositions (closed symbols) for water column profiles of the Southern Atlantic Ocean (A) and the South China Sea in the Western Pacific Ocean (B). The error bars of δ^186/184W values indicate the long-term repeatability of our in-house standard or the internal error of the measurement (2SE), whichever is larger. For W concentration measurements, we assume a relative SD (2RSD) of 5% based on repeated analyses of various reference materials (41). In the Southern Atlantic, Antarctic Intermediate Water exhibits higher W concentrations and slightly lower δ^186/184W than North Atlantic Deep Water below. In the South China Sea, W concentrations decrease with depth, whereas δ^186/184W increases, indicating a riverine and/or aerosol contribution of W to surface waters. Overall, our data suggest an open ocean δ^186/184W of +0.543 ± 0.046 ‰ (2SD; n = 10; blue shading).

In contrast to the open ocean, water samples from the Landsort Deep (site LD1), the deepest basin in the more restricted Baltic Sea, show large variability in dissolved W concentrations (0.051 to 0.232 nM) and δ^186/184W (+0.347 to +0.810 ‰; Fig. 2 and Dataset S3). In this basin, a pelagic redoxcline at around 80 to 88 m water depth (3, 28) separates oxic surface waters with high δ^186/184W (+0.615 to +0.638 ‰) and low dissolved W (0.072 to 0.073 nM) from a deeper weakly euxinic water body (≥150 m) with lower δ^186/184W (+0.409 to +0.417) and higher dissolved W (0.159 to 0.174 nM). The largest variability in δ^186/184W of dissolved W is observed in the intermediate water column (80 to 105 m) including the redoxcline. In the upper part of the redoxcline, upward-diffusing dissolved Mn^2+/3+ ions are oxidized to Mn(IV)-oxide particles (29), which efficiently scavenge trace metals such as W (Fig. 2; e.g., refs. 3, 30). Due to the preferential adsorption of isotopically light W onto Mn-oxides (12), the formation of Mn-oxides causes an increase in the δ^186/184W of dissolved W from +0.638 ‰ in oxic surface waters up to +0.810 ‰ in the upper part of the redoxcline (Fig. 2). The chemically and/or microbially mediated dissolution of sinking Mn-oxides below and the concurrent release of adsorbed trace metals increases concentrations of dissolved W (up to 0.232 nM) and shifts the δ^186/184W of dissolved W back to considerably lower values (down to +0.347 ‰ in 105 m depth). Consistently, Mn-oxide–rich particles that were filtered from the seawater across the redoxcline show even lower δ^186/184W (+0.096 to +0.335 ‰; Fig. 2 and Dataset S4). The observed isotopic difference between the dissolved and the particulate phase is roughly in line with the experimentally predicted isotopic difference of Δ^186/184W = δ^186/184W_dissolved − δ^186/184W_adsorbed = 0.59 ± 0.14 ‰ (12). However, unlike dissolved W, the particulate δ^186/184W shows no distinct trend with depth, which is most likely due to slowly sinking Mn-oxide particles (31) that obscure a direct relation between depth of particle formation and sampling depth. The isotopic difference between dissolved and particulate W is therefore not expected to be constant with depth. Interestingly, the particulate W component (average δ^186/184W of +0.211 ‰) is isotopically lighter than seawater but still significantly heavier than modern igneous rock suites [δ^186/184W: +0.096 ± 0.076 ‰ (8, 9)]. Thus, extremely high δ^186/184W values of Baltic seawater still cause relatively high δ^186/184W values in the authigenic W component, although Mn-oxides preferentially adsorb isotopically light W (Fig. 3).

Fig. 2.

Water column profiles of dissolved and particulate Mn (A) and W (B) concentrations in the Landsort Deep (Baltic Sea). Oxic surface waters (blue shading) are separated from euxinic deeper waters (gray shading). (C) Water column profile of stable W isotope compositions. The error bars in all panels are smaller than symbol sizes. The preferential adsorption of isotopically light W onto Mn oxides causes increasing seawater δ^186/184W in the upper part of the redoxcline but decreasing seawater δ^186/184W below, where Mn-oxides are dissolved again. The diagenetic release of W from Mn-rich sediments might cause relatively low porewater δ^186/184W (see also SI Appendix, Fig. S2 for more details).

Fig. 3.

Stable W isotope compositions of various reservoirs on Earth and their interrelations. The δ^186/184W of igneous rocks (8, 9, 43), the ultimate source of marine W, is distinctly lower than the δ^186/184W of seawater. It is still unclear if stable W isotope fractionation occurs during weathering, adsorption of W during riverine transport or hydrothermal alteration of the oceanic crust, thereby changing the δ^186/184W of the marine input. However, our results indicate that the adsorption of marine W onto Mn-oxides is the most likely process that causes very high δ^186/184W of open ocean seawater (mass balance constraints in the SI Appendix, Fig. S3). We highlight that the δ^186/184W of open ocean seawater is lower than the δ^186/184W of (shallow) Baltic seawater possibly due to the efficient sequestration of isotopically light W via adsorption onto Mn-oxides in deep basins such as the Landsort Deep and only restricted water exchange with the open ocean via Skagerrak. Accordingly, δ^186/184W values of Mn-oxides deposited in open marine settings are expected to be lower than the δ^186/184W of Mn-oxide particles from the Landsort Deep. The partial dissolution of Mn-oxides during diagenesis in the Landsort Deep sediments may result in δ^186/184W of sediment pore waters and basinal deep waters that are distinctly lower than the δ^186/184W of (shallow) Baltic seawater.

Euxinic deep water samples from site LD1 (dissolved H₂S_total between 30.2 and 34.2 µM) exhibit constant δ^186/184W of +0.414 ± 0.008 ‰ (2SD; n = 4) as well as constant concentrations of dissolved W (0.164 ± 0.011 nM; 2SD; n = 7) and dissolved Mn (2.48 ± 0.27 µM; 2SD; n = 7), arguing against significant scavenging or dissolution in water depths below ∼150 m, as expected from the virtual absence of particulate Mn (Fig. 2). In the corresponding sediments at site LD1, the porewater concentrations of dissolved Mn (83.4 to 901 µM), dissolved W (23.0 to 219 nM), and dissolved H₂S_total (935.7 to 3,042 µM) increase with sediment depth and are significantly higher than in the water column [SI Appendix, Fig. S2 and Dataset S5 (3)]. The increasing dissolved H₂S_total levels cause the successive thiolation of WO₄²⁻ (32). While WO₄²⁻ is the predominant species in the weakly euxinic water column of the Landsort Deep (>98%), it is significantly less abundant in the highly euxinic porewaters (e.g., only 15.6% in the deepest porewater sample; SI Appendix, Fig. S2 and Dataset S5). Because W–S bonds are expected to be slightly longer and weaker than W–O bonds (33), more thiolated W species might be isotopically lighter. Indeed, the δ^186/184W of the euxinic porewater (+0.301 ± 0.059 ‰; 2SD; n = 6) is resolvably lower than the δ^186/184W of the weakly euxinic water column (Fig. 2). However, relatively constant porewater δ^186/184W values at highly variable porewater H₂S_(aq) concentrations and variable abundances of different WO_xS_4−x²⁻ species argue against this hypothesis (SI Appendix, Fig. S2). Alternatively, the partial diagenetic release of isotopically light W that was previously adsorbed onto Mn-oxides could cause relatively lower porewater δ^186/184W values. This assumption is consistent with strong authigenic Mn and W enrichments in modern sediments from the Landsort Deep (3). The strong gradient between the W concentration of the uppermost porewaters and the deeper water column suggests the release of isotopically lighter porewater W back into the water column. Thus, the constant and intermediate δ^186/184W of basinal deep waters might reflect a zone of chemical equilibrium between isotopically lighter porewaters and isotopically heavier oxic seawater.

Summary and Implications.

Relative to a clearly defined homogenous open ocean δ^186/184W value of +0.543 ± 0.046 ‰, seawater from more restricted basins in the vicinity of continents can exhibit larger variation in dissolved W concentrations and δ^186/184W (Fig. 4). For example, Western Pacific seawater from the South China Sea shows lower δ^186/184W in surface waters that are influenced by river discharge. The large internal variation in Baltic seawater δ^186/184W values is most likely related to W scavenging by Mn-oxides in redox stratified basins (e.g., Landsort Deep) and the limited water exchange with the open ocean. As such, the preferential scavenging of isotopically light W by Mn-oxides causes lower δ^186/184W in suspended particles, in euxinic deep waters where these particles are redissolved, and in the sediment porewaters. In contrast, residual seawater from the upper oxic water column of the Landsort Deep exhibits higher δ^186/184W than the open ocean (Fig. 4). Thus, globally high seawater δ^186/184W is likely a result of the preferential scavenging of isotopically light W onto oxide minerals such as Mn-oxides [Figs. 3 and 4 and SI Appendix, Fig. S3 (7)]. In times of reduced Mn-oxide mineral formation (e.g., the predominantly anoxic Archean), seawater δ^186/184W as well as the δ^186/184W of authigenic sediments such as carbonates or iron formations may have been significantly lower (mass balance constraints in SI Appendix, Fig. S3). Such a relationship has already been demonstrated for other redox-sensitive transition metals, which have an affinity to Mn-oxides (34, 35). For example, the preferential adsorption of isotopically light Mo onto Mn-oxides is the main cause for the modern isotopically heavy seawater inventory of Mo (15, 18). However, nonquantitative scavenging of Mo in euxinic settings can additionally fractionate Mo isotopes and also influence the seawater Mo isotope composition in the same direction (21). Increasing δ^98/95Mo in sediments that might capture the seawater Mo isotope composition could therefore reflect enhanced burial of Mn-oxide minerals in oxic marine settings or the relative extension of euxinic conditions. In contrast to Mo, W is more soluble in euxinic settings (3, 10). Thus, the stable W isotope composition of seawater is expected to be independent of the global extension of euxinic conditions but more intimately linked to Mn-oxide formation in oxic marine settings. We also note that WO₄²⁻ has a relatively stronger affinity to Mn-oxides than MoO₄²⁻ (11), causing a distinctly lower Mo/W_molar ratio in ferromanganese crusts (∼7) than in seawater [∼1,800 (1)]. Consistently, the particulate fraction from the water column of the Landsort Deep exhibits a strong covariation between Mn and W contents (r² = 0.97) but lacks a similarly clear covariation between Mn and Mo contents (r² = 0.51; SI Appendix, Fig. S4 and Dataset S3). Thus, Mn-oxide formation in well-oxygenated marine environments has a particularly strong control on the inventory and isotope composition of marine W.

Fig. 4.

Graph illustrating W concentrations and stable W isotope compositions of dissolved and particulate compounds. The open ocean (black squares: Southern Atlantic; black circles: South China Sea) shows relatively homogeneous W concentrations and δ^186/184W. Larger variation in the δ^186/184W of Baltic seawater from the Landsort Deep is related to preferential adsorption of isotopically light W onto Mn-oxides that form along the redoxcline. Porewaters that are highly enriched in W exhibit lower δ^186/184W than seawater, likely due to the diagenetic release of light W associated with Mn-rich sediments. The error bars for samples of this study are smaller than symbol sizes. The yellow circles with larger internal SEs (2SE) indicate Northern Pacific waters (7).

In early Earth history, the oceans were predominantly anoxic, thereby inhibiting Mn-oxide formation. Because Fe-hydroxides already form at lower O₂ than Mn-oxides and possibly even in fully anoxic environments via photoautotrophic bacteria (36), the adsorption of WO₄²⁻ onto Fe-hydroxides might have controlled the marine inventory of W in times of O₂ deficiency. In favor of this argument, W (but also Mo) contents in ca. 2.44 Ga old stromatolitic carbonates that were deposited before a major environmental oxygenation event (i.e., the Great Oxidation Event) positively covary with Fe contents (r² = 0.82; SI Appendix, Fig. S5) but not with Mn contents (37). The enrichment of Fe and siderite formation in stromatolitic carbonates was explained by oxidation of organic compounds via Fe-hydroxides during diagenesis as indicated by a negative correlation between Fe content and δ¹³C_Carb (37). Thus, either organic matter or Fe-hydroxides might have been the carrier phase that caused the authigenic enrichment of W and Mo in these stromatolitic carbonates. In analogy to Mn-oxides, Fe-hydroxides would have preferentially adsorbed isotopically light W and Mo, thereby increasing the δ^186/184W and δ^98/95Mo of seawater (12, 17). We note, however, that WO₄²⁻ forms stronger complexations with Fe-hydroxides than MoO₄²⁻, causing distinctly larger fractionation during adsorption (12) and a more distinct enrichment of heavy isotopes in seawater. Thus, the stable W isotope composition of authigenic sediments might be a more sensitive tracer for Fe-hydroxide deposition and the earliest and slightest increase of marine oxygen concentrations in early Earth history.

The relationship between seawater δ^186/184W and authigenic, chemically precipitated sediments needs to be determined in more detail by future studies in order to constrain whether ancient sedimentary archives really preserve changes in the global seawater δ^186/184W. However, authigenic enrichments of marine W in carbonates and iron formations (SI Appendix, Fig. S5) indicate that there is a high potential that the δ^186/184W of such chemical sediments help to better reconstruct the evolution of the marine redox state through Earth history. The relatively homogeneous isotopic composition of modern seawater is clearly advantageous for using stable W isotopes as a global paleo-redox proxy. Furthermore, the combination of multiple paleo-redox proxies may allow a particularly detailed understanding of ancient environmental conditions and their evolution through time (23–26). For example, enhanced deposition of Mn-oxides in well-oxygenated oceans may have increased the δ^98/95Mo and δ^186/184W of seawater, while enhanced deposition of Fe-hydroxides in still barely oxygenated oceans may have mainly increased the δ^186/184W of seawater. In contrast, the extension of (weakly) euxinic conditions may have raised the δ^98/95Mo of seawater while keeping the δ^186/184W of seawater unchanged.

Materials and Methods

A detailed description of sampling sites and the sampling procedure can be found in SI Appendix. Tungsten was preconcentrated from its seawater matrix in the clean laboratory facilities of the Institute of Geochemistry and Petrology, ETH Zurich, using an ethylenediaminetriacetic acid chelating resin, sold commercially as Nobias PA1 [Hitachi High Technologies (38)] and using the protocols as described in ref. 39. Briefly, prior to preconcentration, 1 to 2 L of seawater samples acidified to pH 2 were equilibrated for 24 h with an appropriate amount of ¹⁸⁰W–¹⁸³W double spike. The addition of the double spike and sample-spike homogenization prior to preconcentration accounts for kinetic mass-dependent isotope fractionation during chemical purification of W and subsequent mass spectrometric measurement of W isotope abundances (40, 41). Samples were then adjusted to pH 5 ± 0.3 using an ammonium acetate buffer, made up to a final acetic acid concentration of 30 mM, before loading onto the column. Matrix cations, principally Na⁺, Mg²⁺, and Ca²⁺ were eluted with 100 mL of 30 mM ammonium acetate buffer, followed by the elution of W using 30 mL 1 M HNO₃. The subsequent chemical separation of W from preconcentrated water samples, filtered particles, and powdered reference materials was carried out at the clean laboratory facilities of the University of Cologne using 18.2 MΩ cm water, distilled acids, and precleaned Savillex PFA labware. An adequate amount of a ¹⁸⁰W–¹⁸³W double spike (41) was added to powdered reference materials and filtered particles prior to dissolution in a concentrated HF–HNO₃ mixture (3:1). Dried down sample material was subsequently taken up in concentrated HCl to dissolve potentially formed fluorides. The separation of W comprises a three-step column chemistry (BioRad AG 50 W-X8, 200 to 400 mesh; BioRad AG 1-X8, 100 to 200 mesh; Eichrom TEVA) that is described in detail in SI Appendix, Table S1 and in ref. 41. For preconcentrated water samples, the first column (BioRad AG 50 W-X8, 200 to 400 mesh) was skipped because the abundance of cations was already very low. The setup of the mass spectrometer for W concentration and W isotope measurements (ThermoFisher Scientific© NeptunePlus multicollector inductively coupled plasma mass spectrometer at the University of Cologne) is summarized in SI Appendix, Table S2 and in ref. 41. We present our data in the δ-notation and relative to National Institute of Standards and Technology Standard Reference Material (NIST SRM) 3163 in ‰:

$${\delta }^{186/184}\text{W}=\left(\frac{{\left(\frac{{}_{\hspace{0.17em}}{}^{186}W}{{}_{\hspace{0.17em}}{}^{184}W}\right)}_{\text{Sample}}}{{\left(\frac{{}_{\hspace{0.17em}}{}^{186}W}{{}_{\hspace{0.17em}}{}^{184}W}\right)}_{\text{NIST}\hspace{0.17em}\text{SRM}\hspace{0.17em}3163}}-1\right)\hspace{0.17em}\times \hspace{0.17em}\mathrm{1,000.}$$ [1]

During measurement sequences, the standard NIST SRM 3163 and an Alfa Aesar reference solution were repeatedly measured. The NIST SRM 3163 has a defined δ^186/184W of 0.000 ‰ and a long-term repeatability of ±0.012 ‰ (2SD; n = 192). The Alfa Aesar reference solution is isotopically slightly heavier with an average δ^186/184W of +0.055 ‰ and a long-term repeatability of ±0.015 ‰ (2SD; n = 192). SI Appendix, Fig. S6 shows the long-term δ^186/184W of NIST SRM 3163 and Alfa Aesar measurements (circles) and their sessional averages (squares) since May 2017. The average isotopic difference between Alfa Aesar and NIST SRM 3163 during 29 different measurement sessions was very constant with Δ^186/184W = 0.055 ± 0.009 ‰ (2SD), which is consistent with previous measurement sessions at the University of Tübingen (41). The United States Geological Survey reference material NOD-P-1 was repeatedly measured, also including individual processing through the complete chemical separation procedure. Results for NOD-P-1 (+0.159 ± 0.012 ‰; 2SD; n = 9) are consistent with previous reports for this reference material (41, 42) and indicate a 2SD intermediate precision of ±0.012 ‰ for samples that consist of Mn-oxide–rich suspended particles. We note, however, that the error bars presented in our figures are slightly larger because the long-term repeatability of our in-house standard is with ±0.015 ‰ (2SD) somewhat inferior. The error bars of seawater samples also indicate this long-term repeatability of our in-house standard or the internal error of the measurement (2SE), whichever is larger. Larger internal errors for some seawater samples stem from lower signal intensities during measurements due to lower sample weights. The relative SD (2RSD) of W concentration measurements is ∼5% based on repeated analyses of various reference materials (41). Two procedural blanks of 17 and 24 pg, respectively, indicate a blank contribution of less than 1% of total processed sample W during the processing of filtered particles. The blank contribution to water samples (including preconcentration and subsequent separation of W) was between 132 and 170 pg (n = 3) accounting for less than 1% of total processed W.

Data Availability

All study data are included in the article and/or supporting information.