All solutions used in the experiments were prepared under an inert atmosphere using a glovebox (N₂ atmosphere) and using MQ-water (Ω < 18.2) degassed by purging boiling water with N₂ for 20 min and as it cooled to room temperature. Sulfide minerals (FeS_(s) and CdS_(s)) were synthesized and characterized as described elsewhere (Jonsson et al., 2016). Briefly, disordered FeS_(s) was synthesized by adding 100 ml of 0.6 M Na₂S to 100 ml of 0.6 M Mohr’s salt ((NH₄)₂Fe(II)(SO₄)₂∙6H₂O); and CdS_(s) by adding 25 ml of 0.6 M Na₂S to 25 ml of 0.6 M Cd(NO₃)₂·4H₂O.

The minerals were characterized using X-ray Diffraction Crystallography (XRD) and Brunauer–Emmett–Teller (BET) measurements (Jonsson et al., 2016). XRD studies were conducted by Rigaku UltimaIV diffractometer with Cu Kα radiation (λ = 1.5418 Å) operating at a beam voltage of 40 kV and beam current of 45 mA. The patterns were acquired at a scan rate of 2°min⁻¹, from 0 to 80 degrees in the 2θ range. BET surface-area measurements were performed using nitrogen sorption experiments conducted on a Quantochrome Nova 2000e instrument. All the samples were degassed for 5 h before analysis. Specific surface area was calculated using the adsorption isotherm within 0.05 < P/P0 < 0.3 range, where P/P0 is the relative pressure.

The Hg_(aq) working standard was prepared from a 1,000 ppm Hg_(aq) stock solution (Merck, Allemagne, 1,000 mgL⁻¹ Hg in 1.00 M HNO₃) and then adjusted using 2–8 M KOH_(aq) to obtain the desired pH. Mercury working solutions were prepared daily for each experiment. The ferrous iron solution was prepared by dissolving Mohr’s salt in MQ-water. The DOM isolates used were extracted from surface waters collected at the shelf break of the North Atlantic Ocean and on the western side of Long Island Sound (United States) (Mazrui et al., 2018). The extraction procedure involved passing 0.2 µm filtered seawater through a modified benzene styrene polymer cartridge (Bond Elut) at a rate of <4 ml/min (Dittmar et al., 2008). The cartridge was then rinsed with dilute HCl, dried and the adsorbed DOM eluted with methanol and acetone. DOM dissolved in organic solvent was dried at 40°C using a Nitrogen evaporator (N-EVAP 111). Stock solutions of DOM were prepared by dissolving approximately 0.1 g of the DOM in 100 ml of degassed purified water. The solutions were then filtered through a 0.02 µm PTFE syringe filter, adjusted to pH 7–8, using dilute HCl/KOH and stored in the dark in airtight containers at 4°C until use.

The reduction of Hg^II in the presence of FeS_(s), CdS_(s), Fe^II _(aq) or DOM was tested by adding Hg^II _(aq) to slurries of FeS_(s) or CdS_(s) or solutions of Fe^II _(aq) or DOM in acid cleaned glass vials (total volume of 10 ml). The samples were then incubated in the glove box under anoxic and dark conditions (foil-wrapped sealed serum bottles) to prevent photochemical reactions. Each experimental set was done in triplicate (n = 3) at room temperature. At the end of each experiment, vials were removed from the glove box and produced Hg⁰ _(g) collected onto Goldtrap™(Supelco) traps. For the collection, two tubes were inserted through the septum of the vial. One tube was used to purge the headspace of the vial with Argon (Ar) at a rate of 200 ml/min for 20 min, while the other collected the purged gasses onto a gold trap. Collected Hg⁰ _(g) was then analyzed using a Cold Vapor Atomic Fluorescence Spectrophotometer (CVAFS) (Tekran, model 2,500) after thermal desorption of the Hg⁰ from the gold-traps. A calibration curve was prepared by analyzing 10–200 µL of air saturated with Hg⁰ _(g) from a vial containing Hg⁰ _(g) at a known temperature.

Based on the BET determined surface area (Jonsson et al., 2016), concentrations of FeS_(s) and CdS_(s) in the experiments were adjusted to have a concentration, reported as surface area to volume of solution ratio, of 1, 5, and 30 m²L⁻¹. This is equivalent to 0.02, 0.09, and 0.54 g L⁻¹ for FeS_(s) and 0.01, 0.07, and 0.41 g L⁻¹ for CdS_(s), respectively. Samples containing FeS_(s) or CdS_(s) and 50 pM Hg^II were equilibrated for 24 h under dark conditions at pH 5–8 for the initial experiments. The production of purgeable Hg⁰ was measured after 24 h in the mixtures and control solutions consisting of degassed MQ water and 50 pM Hg^II. In a similar manner, reduction of Hg^II by DOM or Fe^II was tested by incubating 10 ml of experimental solutions containing 5.0 mg C/L of DOM (24 h and pH 7–8) or 1 mM Fe^II (0–4.5 h and pH 5 and 7.5) and 50 pM of Hg^II. Experiments were also performed over time at a pH of 7–8 in the presence of FeS_(s) at different Hg^II:FeS_(s) ratios to examine the impact on the reaction rate.

After the incubation period, experimental solutions containing FeS_(s) slurries were filtered through a 0.02 µm PTFE syringe filter and prepared for Fe^II analysis inside the glove box. Samples for Fe^II analysis were removed from the glove box and immediately analyzed for the concentration of aqueous Fe^II using the ferrozine method (Vollier et al., 2000). Briefly, ferrozine (monosodium salt hydrate of 3-(2-pyridyl)-5, 6-diphenyl-1, 2, 4-triazine-p,p′-disulfonic acid) was reacted with dissolved iron to form a stable magenta complex which absorbs in the visible region at 562 nm. A PharmaSpec UV-1700 UV−Vis spectrometer (Shimadzu) was then used to detect the complex before and after a reduction step with hydroxylamine.

The potential reactions that could occur were examined using calculations of the respective equilibrium constants and the free energy (ΔG) of the reaction under the experimental conditions. The concentration of dissolved Fe (Fe(II)_T) and sulfide (S(-II)_T), and the individual species (principally Fe²⁺ and FeS⁰, with the potential for FeOH⁺, FeCl⁺, and FeSO₄ ⁰ being present at higher pH and anion concentrations) was calculated using the solubility model for FeS_(s) of Rickard (2006) which considers that the Fe(II) concentration is determined by a solubility reaction and an equilibrium reaction: Fe S ( s ) +2 H + =F e 2+ + H 2 S           log   K=3 .5 Fe S ( s ) =Fe S 0                     Log   K = − 5 .7 where FeS⁰ represents a series of (FeS)_x cluster compounds that form in the presence of FeS_(s). The Hg speciation and interaction with FeS_(s) was modeled using constants from Skyllberg and Drott (2010), Stumm and Morgan (1996). Equilibrium constants for the redox reactions were from Stumm and Morgan (1996). The results of the thermodynamic calculations are detailed in Supplementary Tables S1, S2; Tables 1–3. Supplementary Table S1 details the solubility of FeS_(s) across the pH range used in the experiments, Supplementary Table S2 contains a listing of the examined reactions while Table 1 details the concentrations used in the calculations at pH 7. The calculated free energies of the various reactions are contained in Tables 2, 3. The concentrations of Hg⁰ were those measured in the experiments and it was assumed that the total concentration of oxidized forms (sulfate and Fe(III)) were low, (respectively, 0.1 µM and 1 nM) given that these were primarily produced by reduction of Hg^II, or were present as trace constituents in the experimental solution. Their dissolved speciation was taken into account in the calculations.

The rates of reaction were calculated using the Hg⁰ data and with an assumption of a pseudo first order reaction as the concentration of Hg⁰ (<50 pM) is at least five orders of magnitude higher than the concentrations of Fe^II, S^−II in solution or in the solid phase. Additionally, under the experimental conditions, <1% of the solid is dissolved at equilibrium. Therefore, the assumption of a pseudo first order is valid.

In most studies looking at the interactions of Hg^II and FeS_(s), the products obtained were the stable species β-HgS_(s), and Hg⁰ was not detected suggesting that the primary interaction was an exchange reaction with the release of Fe²⁺ from FeS_(s) with concomitant β-HgS formation. This is essentially a cation exchange reaction driven by the thermodynamic favorability of precipitating HgS_(s) (Jeong et al., 2008; Jeong et al., 2010; Skyllberg and Drott, 2010): F e S ( s ) + H g ( S H ) 2 = H g S ( s ) + 2 H S − + F e 2 +       log   K = − 4.4 The reaction is favorable under the experimental conditions except at the lower pH (ΔG = −4.72 kJ/mol at pH 7 and 7.8 kJ/mol at pH 5; Table 3). Therefore, reduction of dissolved Hg could only occur initially, before Hg^II is co-precipitated. Bone et al. (2014), however, suggested that Hg⁰ was generated from the reduction of Hg^II by FeS_(s). They formulated a reduction hypothesis starting from Hg^II adsorption to the mineral. Nevertheless, they could not conclusively verify the role of the reductant—S^−II or Fe^II in Hg^II reduction, and overall, the role of S^−II or Fe^II as electron donors in Hg^II reduction appears to vary according to the ratio of Hg^II:FeS_(s). Similar to the Bone et al. hypothesis, others (Hua and Deng, 2008; Hyun et al., 2012) suggested that U(VI) reduction by mackinawite or amorphous FeS_(s) occurred following U(VI) adsorption onto the mineral surface and simultaneous release of Fe^II. They proposed that once sorbed to the mackinawite surface, either the surface U(VI) is reduced by S^−II at the Fe^II depleted mackinawite surface or the dissolved U(VI) is reduced by dissolved HS⁻ released by congruent dissolution of mackinawite. Kirsch et al. (2008) come to a similar conclusion for their studies of antimony. However, in contrast to U(VI), Hg is known to have a high affinity for reduced S even in substantially oxic environments (Wolfenden et al., 2005). In our experiments, co-precipitation reduces the dissolved Hg in solution and so while the reactions in the dissolved phase with Fe^II or S^−II may be favorable (Reactions 3 and 6 in Table 2), they are unlikely to be the only reactions occurring over time. Indeed, the reactions were slower in the presence of the FeS_(s) suggesting that interaction of Hg with the solid is occurring, as predicted by the thermodynamic calculations at pH 7 and 8 (Tables 2, 3).

The reaction of Hg^II with the surface is pH dependent, as pH affects both the dissolved speciation of Hg^II and sulfide, and the surface charge on the mineral. The point of zero charge (PZC) for mackinawite is around 7.5 (Wolthers et al., 2005) and so under the experimental conditions the surface is either positively charged or near neutral, and thus would not hinder the interaction of the dissolved Hg complexes with the surface. At the lower pH values, the uncharged Hg-sulfide complex dominates in solution but becomes less important as the pH increases. Overall, the noted pH effect on the reduction reaction is likely not related to the impact of pH on the interaction of Hg with the mineral surface. However, the precipitation of HgS_(s) becomes less favorable at low pH. The reactions on the surface and in solution involving Hg^II reduction become more favorable at higher pH, primarily due to the decreasing concentrations of Fe(II) and HS⁻ with increasing pH (Table 3). Furthermore, the experimental pH effect is relatively small with the increase in Hg⁰ production increasing by less than a factor of 2 for a change in [H⁺] of 10³.

Most studies on the interactions between Hg^II and minerals show that the production of Hg⁰ increases with the pH of the solution (Wiatrowski et al., 2009; Bone et al., 2014; Ha et al., 2017). The results reported by Andersson (1979) on the interaction between Hg^II and Fe₂O₃.nH₂O found that the amount of reduced Hg^II increased with pH, from pH values of 6.2–8.5. The same result has been observed by Patterson et al. (1997) with the interaction between chromium and FeS_(s). Thus, our results agree with these studies with the best rate of reduction at pH 7–8. The influence of pH on the production of Hg⁰ in the range 7–8 could be explained by the formation of the dissolved species FeOH⁺ in these studies, which increases in relative concentration with pH, as shown by Amirbahman et al. (2013), although this reaction is unlikely to be occurring in our experiments (Table 3). Our results showed an increase in reduction with increased mineral surface area, suggesting that co-precipitated HgS_(s) is likely involved in the reaction. Other studies on heterogeneous reduction have demonstrated that the formation of surface complexes is responsible for the enhanced reaction rate (Liu et al., 2008; Pecher et al., 2002; Schwarzenbach and Stone, 2003) and this adsorption depends on the pH (Kim et al., 2004; Miretzky et al., 2005). We conclude that this explains why the increase in pH promotes the reduction of Hg^II.

At the low concentrations of Hg used in our experiments compared to the other studies mentioned above, the amount of Fe²⁺ released into solution from the co-precipitation of Hg onto the mineral surface is small compared to the Fe²⁺ in solution in equilibrium with the solid phase. The following reactions determine the dissolved Fe(II) and total S(-II) concentrations in equilibrium with the solid (Stumm and Morgan, 1996; Wolthers et al., 2005; Rickard, 2006; Supplementary Tables S1, S2): F e S ( s ) + H + = F e 2 + + H S −       l o g   K = − 2.23 F e S ( s ) = F e S ( a q ) 0               l o g   K = − 5.7 H 2 S = H + + H S −             l o g   K = − 7.1 F e 2 + +   O H − = F e O H +         l o g   K = 4.5 While FeOH⁺ is the dominant complex formed in solution in the absence of sulfide, the principal form is the free ion. In the presence of sulfide, FeS⁰ is also found where this represents a series of cluster compounds with 1:1 stoichiometry (Rickard, 2006), and is present at a fixed concentration in equilibrium with the solid. The total dissolved Fe^II concentration depends on both the pH and the sulfide concentration (Supplementary Table S1). At pH 5.5, FeOH⁺ is insignificant but increases to about 10% of the total Fe^II at pH 8, according to the thermodynamic calculations. The calculations, based on Rickard (2006), predict a dissolved Fe^II concentration of 182 μM at pH 5 and 3.7 μM at pH 8. Our measurements (Table 5) found slightly higher concentrations at higher pH and less of a pH effect (e.g., 137 μM at pH 7–8 and 166 μM at pH 5–6 for 5 m²/L FeS_(s)), and also that the dissolved Fe^II increased with the amount of FeS_(s) added, suggesting that the assumption of a pure solid (activity = 1) is likely not completely valid for our studies, likely due to the amorphous nature of the solid used. While we did not measure sulfide concentrations, the predicted dissolved concentration (total S^−II _(aq) ∼ Fe^II _(aq)) is not high enough to precipitate the majority of the dissolved Hg as HgS_(s) (Table 1), and much of the Hg^II is in solution initially as Hg(SH)₂, and its deprotonated forms (HgS₂H⁻ and HgS₂ ²⁻) (Skyllberg and Drott, 2010).

The source of the electrons for the reduction of Hg^II is either from a redox reaction on the surface involving the mineral constituents, or a reaction with dissolved reduced ions, either Fe^II or S^−II. If sulfide was being oxidized during the reduction of Hg, then one would predict this should have occurred in the presence of the CdS_(s), but no reduction was observed. The equilibrium dissolved S^−II and Cd^II concentrations in the presence of the solid (K = −14.36 for CdS_(s) + H⁺ = Cd²⁺ + HS⁻) are lower, however, than in the presence of FeS_(s), and thus the reduction in the presence of CdS_(s) would be less favorable even if S^−II was the reductant. Thermodynamic calculations suggest the concentration of sulfide is at low nM levels in the presence of CdS_(s) and that the co-precipitation reaction of Hg with the CdS_(s) is not thermodynamically favorable (Table 2). Thus, the lack of reaction in this case does not necessarily negate the role of sulfide oxidation in Hg reduction.

The other potential reductant is Fe^II for the reactions in the presence of the FeS_(s), as it is with the reactions in the presence of dissolved Fe^II, and no surfaces (Figure 5). Thermodynamically the reaction is favorable in solution (ΔG = −71.3 kJ/mol at pH 7) under the experimental conditions (Tables 2, 3), even given the low concentration of Hg^II relative to Fe^II and HS⁻, and with the assumption that Fe^III is low (Table 1): H g ( S H ) 2 + 2 F e 2 + + 2 H 2 O = H g 0 + 2 F e ( O H ) 2 + + 4 H + + 2 H S −   L o g   K = 55.4 Thus, while Hg remains in solution, the reaction will proceed and this likely accounts for the initial formation of Hg⁰ in the initial time period, and could account for some of the trend seen with pH. However, as the concentration of dissolved Hg^II decreases as Hg is precipitated onto the FeS_(s), this reaction will no longer occur. The time series measurements were made at pH 7–8 where co-precipitation is a favorable reaction, thereby decreasing the dissolved concentration of Hg over time. The differences in the rate of reaction in homogeneous solution (Figure 5) and in the presence of FeS_(s) (Figure 2) indicates that the majority of the Hg^II is being co-precipitated or surface absorbed to the solid.

Another mechanism is therefore needed to account for the Hg⁰ formation at later times. The reaction of co-precipitated HgS_(s) with Fe(II) is not favorable (Table 3) and therefore the reaction that occurs with the precipitated Hg does not involve Fe(II): H g S ( s ) + 2 F e 2 + + 4 H 2 O = H g ( a q ) 0 + 2 F e ( O H ) 2 + + 3 H + + 2 H S −   log   K = − 54.5 Overall, we conclude that Fe^II or FeOH⁺ is not the reductant in our experiments after the Hg has co-precipitated onto the solid. Thus, there is a difference in the mechanisms for the reduction of Hg^II in the presence of FeS_(s) and with dissolved Fe^II (Jeong et al., 2010; Richard et al., 2016). Note, however, that at pH 5 the precipitation reaction is not thermodynamically favorable and therefore the reactions in solution dominate, with Fe^II being the primary reductant (Table 3).

Alternatively, the reductant could be S^−II, and the reason for the low reaction in the presence of CdS_(s) is probably because of the low sulfide concentration in equilibrium with the solid. The potential reaction is thermodynamically favorable, even under the low concentrations of the experimental conditions, for both the reactions in the water and that with the solid, except at the lower pH levels (Table 3): H g ( S H ) 2 + H 2 O = H g ( a q ) 0 + 1 4 S O 4 2 − + 1 3 4 H S − + 2 1 4 H +   log   K = − 25.3 and H g S ( s ) + H 2 O = H g ( a q ) 0 + 1 4 S O 4 2 − + 3 4 H S − + 1 4 H +   log   K = − 24.4 The equations represent either oxidation of dissolved sulfide or of the S^−II associated with the HgS_(s). Overall, again, these calculations suggest that the reaction later in the experimental time period does not involve dissolved reduced species and that the reduction involves reactions within the solid, with the electrons being provided from the oxidation of S^−II by an electron transfer reaction at the surface, followed by the release of Hg⁰ into solution. Overall, these results suggest that initially in the experiments, the dissolved Hg is being reduced by either Fe(II) or HS⁻, but that later in the experiment the reaction only involves reduced S. If Fe^II is being oxidized, the Fe^III produced would likely remain adsorbed on the solid, but it is also likely that the Fe(III) would be reduced back to Fe(II) by the sulfide in solution as this reaction is favorable under the experimental conditions (ΔG = −28.8 kJ/mol at pH 7; Tables 2, 3). Thus, once the Hg is co-precipitated onto the FeS_(s) surface, whether the reaction involves initially S(−II) or Fe(II) is somewhat academic as the final products will be the same because of the reduction of any Fe(III) produced.

If the reaction involves sulfide oxidation, the fate would depend on the degree of oxidation of S^−II. It is likely that some intermediate product, such as elemental sulfur (S⁰), could result, rather than complete oxidation to sulfate. Indeed, an electron exchange reaction between Hg^II and S^−II could potentially occur with the formation of Hg⁰ and elemental S. Given the uncertainty in the equilibrium constants (Stumm and Morgan, 1996; Skyllberg and Drott, 2010), and assuming pure solids are formed, the reaction is near equilibrium at pM Hg⁰ concentrations (i.e., [Hg⁰] ∼ K; Table 2, Reaction #7): H g S ( s ) = H g ( a q ) 0 + S ( s ) 0   log   K = − 12.5   and   Δ G  =   6 . 84   kJ / mol   at   5   pM   Hg 0 As mentioned earlier, the lack of a reaction with CdS_(s) is likely because of the lack of precipitation of HgS_(s) on the surface at the low sulfide concentrations found in equilibrium with the solid phase, and the low sulfide concentration in solution. This is because of the stronger M-S bond in CdS_(S) compared to FeS_(S).

Furthermore, as noted above, the increase in the amount of mineral surface of mackinawite (Figure 2) slightly influences the quantities of Hg⁰ produced. With 30 m²L⁻¹ of FeS_(s), the reduction reached a maximum of 11 pM of Hg⁰ after 24 h of reaction, while this maximum was 6.9 pM for 1 m²L⁻¹ of FeS_(s). These results indicate that a surface catalytic role of precipitated HgS_(s) on mackinawite is involved in the production of Hg⁰. Wiatrowski et al. (2009) demonstrated that the kinetics of Hg^II reduction by magnetite systematically varies as a function of magnetite concentration. Amirbahman et al. (2013)’s study on the kinetics of Hg^II reduction by Fe^II suggested that the mineral phases are important factors affecting the rate of the mercury reductive pathways, and O’Loughlin et al. (2020) showed that there were differences in the reaction rates in the presence of Fe^II-containing clays. However, Jeong et al. (2010) have shown that the adsorption of Hg^II onto the surfaces of mackinawite only occurs below a certain molar ratio of Hg^II and FeS_(s), which implies that the ratio of Hg^II:FeS_(s) could also influence the production of Hg⁰. This ratio in our study was 5–7 orders of magnitude lower than that of Jeong et al. (2010).

In summary, our study indicated that mercury reduction by FeS_(s) is kinetically slow and the production of Hg⁰ is small compared to other potential reduction pathways in environmental ecosystems, such as Hg^II reduction in the presence of dissolved Fe^II or DOM, and also appears to occur via a different mechanism. The experiments were carried out with excess FeS_(s) concentrations so the reaction can therefore be described according to pseudo first order kinetics. The overall reaction rate constants obtained are k = 67 × 10⁻³ h⁻¹; 85 × 10⁻³ h⁻¹; and 92 × 10⁻³ h⁻¹, respectively, for 1, 5, and 30 m²/L of FeS_(s). These values are similar in terms of the link between reaction rate constant and mineral concentration noted in some studies (Wiatrowski et al., 2009; Amirbahman et al., 2013; Ha et al., 2017). However, our experimental data show that the average net rate of Hg^II reduction by FeS_(s), assuming first order kinetics over the experimental time period, is lower than Hg^II reduction by humic substances (1.6–2.1 × 10⁻² h⁻¹; Chakraborty et al., 2015), minerals such as clay (1.74 × 10⁻¹ h⁻¹) (Peretyazhko et al., 2006a), hematite from phlogopite (6.60 × 10⁻¹ h⁻¹) or magnetite (Wiatrowski et al., 2009). These results confirm that the production rate of Hg⁰ is a function of the nature of the mineral (i.e., oxide or sulfide), and likely the form of Hg adsorbed or precipitated on the surface of the mineral.

Overall, under our experimental conditions, the homogeneous Hg reduction in presence of aqueous Fe^II without mineral surfaces was more favorable than the experiments in presence of FeS_(s) mineral (Figure 5), which is consistent with the thermodynamic calculations (Table 3). The initial reaction rate was 20–70% higher for the aqueous Fe^II experiments. However, the concentration of Fe^II in the homogeneous experiments was much higher than in the mineral studies and this could potentially account for the higher conversion rate to Hg⁰, although the rate should be similar given that the initial Hg^II concentration was the same, and the Fe^II concentrations in both cases is substantially higher and not rate limiting Rather, the mechanisms are likely different for the two situations. Although several authors have shown the role of surface-catalysis by iron minerals on the rate of mercury reduction, our data (Figure 5) shows fast reduction of mercury in presence of aqueous Fe^II. In contrast, Ha et al. (2017) indicated that mercury reduction by aqueous ferrous iron in the absence of a solid phase was kinetically slow. Pasakarnis et al. (2013), Amirbahman et al. (2013) suggested that Hg^I sorbed onto the mineral surface during the transformation of Hg^II to Hg⁰ and acts as a surface-catalyst in this reaction. Peretyazhko et al. (2006b) demonstrated that adsorption of Fe^II to the haematite surface created very reactive sites for the reduction of Hg^II, while in the absence of haematite particles, no production of Hg⁰ occurred. The difference between this study and previous studies mentioned above might be due to the low concentration of Fe^II in the FeS_(s) suspensions or more likely because the reaction proceeds via a different mechanism once the Hg is co-precipitated.

It is well known that dissolved organic matter (DOM) has a strong interaction with mercury and other trace metals affecting their speciation, mobility and toxicity (Buffle, 1988). Under abiotic dark conditions in aquatic systems, DOM participates in the conversion of Hg^II to Hg⁰ but also contributes to the strong complexation of Hg^II (Ravichandran et al., 1998; Deonarine and Hsu-Kim, 2009; Zheng and Hintelmann 2010; Zheng et al., 2012; Han et al., 2007). This complexation is attributed to reduced sulfur ligands (Waples et al., 2005; Merritt and Amirbahman, 2007). Indeed, DOM is a mixture of molecular organic compounds with a large number of hydrophilic functional groups: carboxylic (COOH), phenolic and/or alcoholic (OH), carbonyl (C=O) and amine groups (NH₂). Reduced sulfur groups also exist in different oxidation states (R-SH, R-S=S-R and R-SO₃H). Chakraborty et al. (2015) showed that the ratio of the–COOH/−OH groups and the sulfur content in the humic substances reveal a strong competition between complexation and reduction of Hg^II. They suggested that several parameters such as pH, total sulfur content, the −COOH/−OH ratio and salinity influenced the reduction of Hg^II in presence of DOM. In our studies, the less humic DOM1 reduced Hg at a higher rate than that with DOM2, and this is consistent with the data of Chakraborty et al. (2015) who showed that the rate of reduction was higher for humic material with less total S, or a higher ratio of carboxylic to thiol groups. As discussed above and shown in Supplementary Figure S1, DOM2 has more humic character while DOM1 is more protein-like in terms of its fluorescence.

We observed that the Hg^II reduction by DOM was diminished in presence of FeS_(s), whatever the characteristics of the experiment (Figure 4). Calculations of the speciation of dissolved Hg in the presence of FeS_(s) and DOM at the concentrations used in the experiment, using the RSH:DOM ratios determined by Seelen (2018) for comparable coastal waters and the Hg(SR)₂ binding constant from Skyllberg and Drott (2010), suggest that a small fraction of the Hg—5–10% depending on the DOM was organically complexed during the experiments. This is consistent with the results that showed the extent of reduction in the presence of FeS_(s) and DOM was lower than that of FeS_(s) alone. Overall, we conclude that the presence of DOM increases the barrier to Hg reduction by sulfide surfaces, and likely also has a similar effect for other reductive surfaces.

Mishra et al. (2011) observed that the Hg^II reduction by magnetite and green rust was severely diminished in the presence of bacterial biomass, suggesting inhibition by surface sulfhydryl groups. These experiments suggest that the conditions of the experiment likely determine whether Hg is primarily bound to the reduced S in DOM or the inorganic reduced sulfide in FeS_(s), or is removed by co-precipitation. Furthermore, in most of the studies on the interaction between Hg^II and FeS_(s), the products obtained were the stable solids metacinnabar, cinnabar, and Hg associated with iron sulfides (Jeong et al., 2008; Liu et al., 2008; Skyllberg and Drott, 2010) suggesting Fe^II present in FeS_(s) suspension acts as an electron donor in the production of Hg⁰. However, in the presence of DOM and FeS_(s), this mechanism could be changed as DOM likely keeps the Hg in solution and prevents its interaction with the solid phase, although our calculations show that the extent of complexation was small. However, depending on the pH, the DOM can also interact with the mineral surface and therefore hinder the co-precipitation of Hg^II and any surface reactions. Dissolved organic matter is known to play a dual role in HgS_(s) formation and stabilization (Slowey, 2010; Gerbig et al., 2011a), creating a competition between its complexation of Hg and Hg adsorption to the iron sulfide (Skyllberg and Drott, 2010) and, influencing, through its complexation of dissolved Hg, the dissolution of cinnabar (Ravichandran et al., 1998; Waples et al., 2005). We conclude that our data showing that Hg^II reduction in presence of both DOM and FeS_(s) was less than found in the presence of either DOM or FeS_(s) only, is because of the competition between FeS_(s) and DOM for complexation and the extent of HgS formation (Skyllberg and Drott, 2010). The Hg^II would be less available for reduction by DOM, Fe^II or FeS_(s) under these conditions. Zhu et al. (2013) have shown that the strength of Fe^II as a reducing agent is affected by DOM during the reduction of 2-nitrophenol (2-NP) in TiO₂ suspensions. Overall, the Hg^II reduction in presence of DOM or mineral phases involves complicated reaction pathways but the presence of DOM increases the barrier to reduction.

Our study demonstrates that Hg^II can be reduced to Hg⁰ in the presence of FeS_(s) but the extent of reduction is slow compared to that found with hydrous ferric oxide, with dissolved Fe(II) and in the presence of DOM. The data presented herein show clearly that in the presence of sulfide surfaces, Hg^II is less available for reduction. However, our results also showed that there was no Hg⁰ production in presence of CdS_(s) in contrast to FeS_(s), suggesting that the presence of a sulfide surface is not sufficient for this reaction to occur (Figure 4). The concentration of sulfide in solution also plays a role in controlling the extent of the reaction. Neither FeS_(s) nor CdS_(s) enhanced Hg^II reduction compared to DOM or Fe^II. Based on thermodynamic calculations (Tables 2, 3; Supplementary Table S2), we suggest that S^−II was the likely electron donor for reduction of precipitated Hg^II in the presence of FeS_(s), and its higher concentration in the FeS_(s) solutions compared to the CdS_(s) solutions accounts for the differences in the Hg⁰ formation. At low pH in the presence of FeS_(S), precipitation of Hg^II is unlikely to occur and in this instance, reactions in solution are likely controlling the rate of reduction. We therefore suggest that while the Hg does not need to be adsorbed to the surface for the reaction to proceed, this is the likely fate of Hg in the presence of FeS solids under environmental conditions.

Extrapolating these findings to environmental conditions, we suggest that chemical reduction of Hg^II is complex in anoxic environments, such as sediment, with many potential reaction pathways. This reaction is influenced by ferrous iron, minerals, sulfide, DOM and interactions between the different compounds and solid phases. However, we conclude that the presence of FeS_(s) in environmental sediments is not the major driver of the formation of Hg⁰ in such systems as the reactions are slow once the Hg interacts with the mineral surface. Other reduction pathways are much more favorable with dissolved reductants (reduced Fe and S species). Furthermore, this study shows the influence of DOM on the reaction between Hg and FeS_(s) and that its presence needs to be considered because DOM affects mercury transformation and mercury reactivity toward minerals, as shown by Skyllberg and Drott (2010). The type of DOM also influences the rate of reaction, as it does complexation.

Overall, processes that convert Hg^II to Hg⁰ under anoxic conditions are important mitigators of the production and bioaccumulation of CH₃Hg as reduction potentially removes ionic Hg from the system where it could otherwise be methylated. More research at lower Hg concentrations is needed to further understand the primary reactions that are occurring and the potential role of DOM and pH in controlling the rates of Hg reduction.