Mercury Reduction by Nanoparticulate Vivianite.

Mercury (Hg) is a toxic trace element of global environmental concern which has been increasingly dispersed into the environment since the industrial revolution. In aquatic and terrestrial systems, Hg can be reduced to elemental Hg (Hg0) and escape to the atmosphere or converted to methylmercury (MeHg), a potent neurotoxin that accumulates in food webs. FeII-bearing minerals such as magnetite, green rusts, siderite, and mackinawite are recognized HgII reducers. Another potentially Hg-reducing mineral, which commonly occurs in Fe- and organic/P-rich sediments and soils, is the ferrous iron phosphate mineral vivianite (FeII3(PO4)2·8H2O), but its reaction with HgII has not been studied to date. Here, nanoparticulate vivianite (particle size ∼ 50 nm; FeII content > 98%) was chemically synthesized and characterized by a combination of chemical, spectroscopic, and microscopic analyses. Its ability to reduce HgII was investigated at circumneutral pH under anoxic conditions over a range of FeII/HgII ratios (0.1-1000). For FeII/HgII ratios ≥1, which are representative of natural environments, HgII was very quickly and efficiently reduced to Hg0. The ability of vivianite to reduce HgII was found to be similar to those of carbonate green rust and siderite, two of the most effective Hg-reducing minerals. Our results suggest that vivianite may be involved in abiotic HgII reduction in Fe and organic/P-rich soils and sediments, potentially contributing to Hg evasion while also limiting MeHg formation in these ecosystems.

Introduction

Mercury (Hg) is a trace element that is released into the environment from natural processes (e.g., volcanic emissions) and anthropogenic sources (e.g., coal burning).[1] Since the beginning of the industrial revolution, concentrations of Hg have increased in all environmental compartments, including soils, sediments, water, atmosphere, and biota.[2] Hg can be present in different species and oxidation states in the environment.[3] Hg0 (metallic, elemental Hg) is the most abundant species in the atmosphere, allowing its long-range transport, while HgII (mercuric Hg) is the predominant oxidation state in many aquatic and terrestrial systems. Because of the lipophilicity and protein-binding properties of monomethylmercury (MeHg), Hg builds up to high levels along aquatic and terrestrial food webs,[4,5] thereby causing severe health effects in wildlife and humans.[6] Identifying new mechanisms involved in the transformations of Hg is key for a better understanding of its fate, to mitigate its toxic effects, and develop remediation techniques.

Reduction of HgII to Hg0 is an important pathway controlling its evasion from aquatic and terrestrial ecosystems,[1,7] and thus, its redistribution through the atmosphere while Hg would otherwise be buried and escape surficial cycling. In photic compartments, the reaction is mostly photochemically driven,[8] while in the absence of light, it can proceed through either biotic and abiotic pathways.[9] Some bacteria are specifically equipped with genes conferring resistance against Hg but the Hg concentration threshold to activate these genes suggests that their activities are restricted to highly contaminated sites.[10] Several dissimilatory metal-reducing bacteria are also able to reduce low concentrations of Hg in relation with the activity of respiratory electron transport chains[11] but under realistic environmental conditions, the reaction is probably prevented by Hg sorption onto cell surfaces by thiol groups.[12] Hg can also be reduced abiotically by organic matter (OM)[13] but at environmental OM and Hg concentrations, the reaction is also likely outcompeted by the strong complexation of thiol groups.[14] On the other hand, FeII-bearing minerals, such as phlogopite,[15] magnetite,[16] green rusts,[17,18] mackinawite,[19] and siderite[20] are well known to be efficient HgII reducers. The reduction reaction proceeds in two steps with Hg complexation at the mineral surface required prior to electron transfer from FeII. The reaction yields and kinetics are controlled by several factors, including the mineral specific surface area (SSA),[16,20] FeII concentration,[21] and surface charge determined by its point of zero charge and solution pH[18] but also by the presence of organic and inorganic ligands that will affect the dissolved Hg speciation and act as competitors for the mineral surface sites.[16,18,21,22]

Vivianite (FeII3(PO4)2·8H2O), the most common stable iron phosphate mineral, has been reported to occur worldwide in a large variety of natural aquatic and terrestrial environments, such as freshwater or coastal sediments and sinking particles, bogs, waterlogged and organic-rich soils, but also in anthropogenic systems such as sewage sludges (see Rothe et al.,[23] for a comprehensive review on the topic). Its formation occurs in oligotrophic as well as in eutrophic ecosystems where the production of sulfides is limited relative to the Fe present.[24] It appears to be favored under organic-rich conditions where readily degradable organic debris enable the development of reducing microenvironments, promoting Fe reduction, and thus, the release of FeII and orthophosphates previously sequestered by Fe oxyhydroxides. Although the quantitative importance of vivianite remains difficult to establish, it is one of the most important sinks of phosphorus in reducing natural environments, which could represent up to 40–50% of total P in sulfide-poor environments.[25−27] It has also been found to be the dominant Fe phase in the deep anoxic waters and sediments of the ferruginous and low-sulfate Lake Pavin.[28] The reactivity of vivianite toward trace elements, and especially, its ability to immobilize and/or reduce them, has already been established for several important metal(loids) including, for example, As,[29,30] U,[31] Zn,[32] Cu,[33] Ni, and Co[34] or Cr.[35] However, the reactivity of vivianite toward HgII has not yet been studied, and we hypothesize that it may contribute to the reduction of HgII in organic/P-rich soils and sediments, where Hg also accumulates thus promoting its evasion and redistribution while also limiting its methylation.

In this work, vivianite was abiotically synthesized by precipitation and characterized by a combination of complementary techniques: X-ray diffraction (XRD), Mössbauer spectroscopy, X-ray absorption spectroscopy (XAS, Fe K-edge), and transmission electron microscopy (TEM). Nanoparticles of pure vivianite were obtained and the main objective of this work was to demonstrate their ability to reduce HgII to Hg0 and determine the reaction yields and kinetics at various FeII/HgII ratios (0.1, 1, 100, and 1000) at circumneutral pH under anoxic conditions. The mass balance between HgII reduction, Hg0 production, and FeII oxidation was determined for the FeII/HgII ratio 1, as well as the change in mineralogy of the HgII-reacted vivianite. The kinetic parameters of the reduction reaction were determined for all ratios and compared to the other FeII-bearing minerals.

Materials and Methods

All solutions were prepared from analytical grade chemicals with boiled, N2-outgassed doubly deionized (DDI) water (Milli-Q, Millipore, 18.2 MΩ cm). Brown glass vials used for Hg reduction experiments were washed successively with 10% nitric acid (HNO3, v/v) and 10% hydrochloric acid (HCl, v/v), and finally rinsed with anoxic DDI water. Teflon caps were washed in the same way but with 1% acids (v/v).

Vivianite Synthesis

Vivianite was synthesized in an anoxic chamber (MBraun, UNIlab Plus, N2 atmosphere, <1 ppm (v/v) O2), at 45 °C by precipitation from 1 mmol L–1 solutions of (NH4)2HPO4 and FeIISO4·7H2O mixed in appropriate proportions[36] (i.e. a slight excess of P to scavenge all Fe2+ species) to have final concentrations of FeII ranging from 50 nmol L–1 to 50 μmol L–1. At this stage, pH was slightly acidic and readjusted to 7.0 ± 0.2 with the dropwise addition of a 4 M NaOH solution. A white precipitate (Figure S1) appeared immediately and was continuously stirred at 400 rpm for 1 h to entirely consume free FeII species. Because vivianite is sensitive to light exposure (Figure S1c), its synthesis was conducted in amber glass flasks.

HgII Reduction Kinetic Experiments

A first set of experiments was carried out outside the anoxic chamber with a purge and trap setup (Figure S2) to precisely determine the Hg mass balance with respect to Hg0 production and the reaction stoichiometry with respect to Fe oxidation. Briefly, the purge and trap setup was assembled in the anoxic chamber, sealed, removed, and then the brown glass reactor containing vivianite (final concentration of 8.4 mg L–1) was spiked with a HgIISO4 solution (final concentration of 50 μmol L–1), that is, an FeII/HgII ratio of 1 (Table S1). Elemental mercury (Hg0) produced during these HgII reduction experiments was stripped off from the vivianite solutions by a constant stream of ultrapure N2 gas (4.5 quality N2, 200 cm3 min–1 transiting through PTFE tubing) and trapped in a 0.6% w/v KMnO4 solution acidified by 10% v/v HCl. Both the initial vivianite solution to which Hg2+ was added and the KMnO4 solutions were analyzed for Hg by gas chromatography–inductively coupled plasma–mass spectrometry (GC–ICP–MS) (see below).

To study the kinetics of HgII reduction by vivianite, experiments were carried out in the anoxic chamber at four FeII/HgII ratios, that is, 0.1, 1, 100, and 1000 (Table S1); the HgII solution was prepared from a HgSO4 salt (Merck, 99%) in 0.5 M HCl, then its pH was adjusted to 7.0 ± 0.2 with 1 M NaOH just before its addition to 200 mL vivianite suspensions (contained in 250 mL bottles leaving a headspace volume larger than 50 mL) to reach a final concentration of 0.5 μmol L–1 Hg. The suspensions were continuously magnetically stirred at 400 rpm to allow for the produced Hg0 to escape into the headspace volume. A small volume (0.5 mL) of the unfiltered HgII-reacted suspension (i.e., with dissolved and adsorbed Hg) was regularly sampled over 24 h and immediately dissolved in 6 M HCl to stop the reaction by dissolving the mineral phase and stabilize Hg. Control experiments were performed under the exact same conditions except that vivianite was replaced by aqueous Fe2+ (50 μmol L–1, pH 6.5) or deoxygenated water to assess the reactivity of HgII species toward free Fe2+aq versus structural FeII in vivianite, and the potential losses of HgII.

Chemical Analyses: pH Monitoring and Fe and Hg Speciation

The initial pH value of vivianite suspensions before HgII spike was measured using a microelectrode immersed in 1 mL of the sample maintained in an anoxic chamber. A data-reading time of 30 s was used to obtain stable values. Total FeII and total Fe (FeTOT = FeII + FeIII) in the suspensions (i.e., solids + solution) were determined using the ferrozine method.[37] At regular time intervals, the suspensions were sampled using syringes and part of the samples were dissolved in 2 M HCl for 15 min, and then diluted with DDI water. FeII species were measured immediately after adding the ferrozine by UV absorption at 562 nm (Cary 60 UV–vis, Agilent). FeTOT species were evaluated at the same wavelength after reduction with 1.4 M NH2OH and HCl for 10 min in darkness and stabilization of pH by the addition of 10 M CH3CO2NH4.

The concentrations of HgII were determined by GC–ICP–MS (Agilent 8800 ICP-QQQ-MS and 7890B GC fitted with an Agilent J&W HP-5 column, 30 m long, 0.320 mm I.D., 0.25 μm film thickness) according to Monperrus et al.[38] but using external calibration. Briefly, aliquots of the samples dissolved in 6 M HCl were added to 0.5 M sodium acetate buffer and the pH was adjusted to pH 4 with a NH3 solution if necessary. Proper volumes of isooctane and 5% (w/v) sodium tetrapropylborate were then added, and the vials were placed on an orbital shaker for 10 min. After the derivatization reaction was complete, the organic phase was transferred to GC vials and injected into the GC–ICP–MS using an autosampler.

Fe-Bearing Mineral Characterization

X-ray Diffraction

XRD data for unreacted and HgII-reacted vivianites were collected with a Bruker D8 ADVANCE diffractometer equipped with a high-resolution energy-dispersive one-dimensional (1-D) detector (LYNXEYE). The X-ray source was a Cu anode (kα1= 1.5406 Å; kα2= 1.5444 Å). The diffractograms were recorded in the Bragg–Brentano geometry in the 10°–70° 2θ range with a 0.02° step size and an acquisition time of 4 s per step in the dark. In order to perform XRD analyses under inert atmosphere conditions (N2 atmosphere), unreacted and HgII-reacted vivianites were centrifuged (10,000g at 22 °C for 10 min), concentrated, and washed three times with O2-free DDI water before being deposited onto polished silicon wafer (Sil’tronix Silicon Technologies, France) to form a homogeneous film after a 24 h drying period in an anoxic chamber protected against light oxidation by Al foil. A plexiglas dome equipped with an anti-scatter knife edge (Bruker, A100B138-B141) was used to seal the dried film and allow anoxic conditions to be maintained during the XRD analysis. The diffractograms were analyzed by Rietveld fitting in TOPAS (V.5.0; Bruker, Germany) based on CIF files for vivianite (American Mineralogist Crystal Structure Database, #0015722) and metavivianite (International Crystal Structure Database, #188922), respectively.

Transmission Electron Microscopy

TEM was conducted using a dedicated scanning transmission electron microscope (STEM, HD2700Cs, Hitachi, Japan) equipped with a high-angle annular dark field (HAADF) detector to visualize contrasts in atomic weights, and a secondary electron detector to image the morphology of the particles. Elemental analyses, including elemental distributions, were conducted on a (S)TEM (Talos F200X, Thermo Fisher) coupled to an energy-dispersive X-ray (EDX) system (Super-X EDX, 4 detector configurations, FEI), also equipped with a HAADF detector. Both microscopes were operated at an acceleration voltage of 200 kV. Carbon-coated Cu grids were functionalized with 0.1% poly-l-lysine solution for 10–15 min and washed with DDI water before centrifuging unreacted and HgII-reacted vivianite suspensions at 14,000g for 1 h on TEM grids. Afterward, TEM grids were rinsed three times with DDI water to prevent the formation of precipitates from dissolved salts. The characterization of the particle morphology and size were determined from 300 individual particles using ImageJ software. The SSA (m2 g–1) was calculated for each individual vivianite particle based on their respective diameter, assuming that particles are spherical, according to eq (39)where d (nm) is the diameter and ρ (g cm–3) is the density (2.69 g cm–3 for vivianite).

57Fe Mössbauer Spectroscopy

57Fe Mössbauer spectroscopy was used to quantitatively characterize unreacted and HgII-reacted vivianites. Dried powder of unreacted vivianite (∼50 mg) was collected after centrifugation (10,000g at 22 °C for 10 min), rinsed three times with anoxic DDI water, and filtered in an anoxic chamber. Then, the dried unreacted vivianite was loaded onto a Plexiglas holder (1 cm2). The wet filtered HgII-reacted sample was prepared by passing liquid material through a 0.44 μm filter, which was then sealed between two layers of Kapton tape. The sample sandwich was kept in a freezer (−20 °C) under anoxic conditions until measurement. All samples were sealed in Schott bottles and only exposed to air immediately prior to loading inside a closed-cycle exchange gas (He) cryostat (Janis cryogenics). Measurements were collected at 77 K with a constant acceleration drive system (WissEL) in transmission mode with a 57Co/Rh source and calibrated against 7 μm thick α-57Fe foil measured at room temperature. All spectra were analyzed using Recoil (University of Ottawa) by applying a Voight Based Fitting routine. The half width at half maximum was fixed to a value of 0.130 mm/s for all samples.

Bulk Fe XAS Analyses

Unreacted vivianite was analyzed by bulk Fe K-edge (7112 eV) extended X-ray absorption fine structure (EXAFS) spectroscopy at the SAMBA beamline (SOLEIL Synchrotron, St. Aubin, France). For this analysis, the dried sample material was pressed into 1.3 cm pellets and sealed with Kapton tape. To prevent changes in the Fe oxidation state, samples for XAS analyses were prepared in an anoxic glovebox, where they were doubly sealed in Al foil for transport to the synchrotron to be measured under anoxic conditions. To this end, immediately prior to sample mounting, Al-sealed bags were opened in ambient air and the Kapton-sealed pellet was mounted onto the sample holder which was immediately plunged into liquid N2. The N2 frozen-mounted sample was then inserted into a cryostat [He(l), ∼25 K], which was then flushed and purged with He(g) 3 times. The spectra then were recorded in continuous scan mode as transmission data at ∼25 K using a Si(220) monochromator, which was calibrated to the first-derivative maximum of the K-edge absorption spectrum of metallic Fe foil (7112 eV). Foil was continuously monitored to account for small energy shifts (<1 eV) during the sample measurements. Higher harmonics in the beam were eliminated by use of Si mirrors. 10–15 scans were collected and averaged. Details on data reduction and analyses are given in the Supporting Information (Section S3).

Results and Discussion

Characterization of Vivianite Nanoparticles

The white precipitate obtained in the synthesis was first analyzed by XRD and the resulting diffractogram was consistent with the vivianite structure as shown by Rietveld fitting[40] (Figure a). Only a minor percentage (1.6%) of metavivianite, the partially oxidized form of vivianite FeIIFeIII2(PO4)2(OH)2·6H2O,[41] was detected by one small reflection at 12.6 °2θ, highlighting the purity of the synthesized vivianite. The spherical morphology of vivianite nanoparticles was evidenced by TEM (Figure b) and their good dispersion on the grids allowed for an assessment of the particle size distribution. The size distribution of vivianite nanoparticles was monomodal with a davg value of 53 ± 10 nm (n = 300, normal distribution, p-value = 0.2), leading to an average SSA distribution (eq ) of 42.0 ± 0.4 m2 g–1. This is four times higher than previously reported by Luna-Zaragoza et al.[42] because of the nanoparticulate nature of the synthesized mineral.

Figure 1

Characterization of unreacted and HgII-reacted vivianite: (a) XRD diffractograms of unreacted vivianite (upper graph) and after 24 h incubation with HgII at a FeII/HgII ratio of 1 (lower graph). (b) Particle size distribution of vivianite nanoparticles, determined from three independent 100-particle samples (inset, STEM-secondary electron image). (c) 77 K transmission Mössbauer spectra of vivianite at t = 0 (upper graph) and t = 24 h (lower graph) were fitted using the presence of three doublets D1 (light green) and D2 (dark green) for structural ferrous iron, and D3 (red) for structural ferric iron.

Mössbauer spectroscopy was used to quantify the proportion of FeII and FeIII species in solid samples. The Mössbauer spectrum of the unreacted vivianite obtained at 77 K was fitted with three doublets, D1, D2, and D3 (Figure c) with the corresponding hyperfine parameters, as presented in Table . Doublets D1 and D2 are characterized by large center shift (CS) and quadrupole splitting (ΔEQ) values, which are similar to previously published values for vivianite at 77 K,[43,44] and correspond to the two FeII octahedral sites. One site corresponds to Fe2+O2(H2O)4 octahedra where the trans O2– corners are the apices of PO4 tetrahedra, whereas the second site constitutes Fe2+O4(H2O)2 octahedra that share edges in pairs. A third doublet D3 is characterized by small CS and ΔEQ values of 0.59 and 0.76 mm s–1, respectively. These values are typical of FeIII and are potentially consistent with those of metavivianite:[41,43] CS ∼ 0.43 mm s–1 and ΔEQ ∼ 0.77–0.87 mm s–1. Thus, the contribution of FeII and FeIII was estimated at 93.8 and 6.2%, respectively. This indicates that some degree of oxidation occurred but only to a limited extent, possibly during the transfer of the sample into the cryostat.

Table 1

Mössbauer Hyperfine Parameters at 77 K of Unreacted and HgII-Reacted Vivianitea

	type of sites	CS (mm s–1)	ΔE (mm s–1)	σ (ΔEQ)	R.A. (%)	χ2
unreacted vivianite	D1 (FeII)	1.34 ± 0.001	3.22 ± 0.004	0.08	56.4 ± 0.8	0.85
	D2 (FeII)	1.30 ± 0.002	2.62 ± 0.009	0.15	37.4 ± 0.8
	D3 (FeIII)	0.59 ± 0.027	0.76 ± 0.043	0.30	6.2 ± 0.4
HgII-reacted vivianite	D1 (FeII)	1.34 ± 0.002	3.23 ± 0.005	0.00	10.3 ± 0.1	0.59
	D2 (FeII)	1.31 ± 0.004	2.64 ± 0.012	0.14	16.8 ± 0.4
	D3 (FeIII)	0.50 ± 0.090	0.83 ± 0.120	0.48	72.9 ± 0.1

CS is the center shift with respect to α-Fe (mm s–1), ΔEQ the quadrupole splitting (mm s–1), with σ(ΔEQ) indicating sigma broadening of the quadrupole splitting, R.A. the relative abundance of each site (%), and χ2 the goodness of fit. D refers to each doublet used during fitting.

Shell fit analysis of the Fe K-edge EXAFS spectra for the unreacted vivianite was conducted as a complimentary technique to XRD and Mössbauer to determine local Fe coordination and bond lengths as well as the presence of amorphous phases. The results are shown in Figure S3, and the corresponding fitting parameters are detailed in Table S2. The first coordination shell was fitted with 4.8 O at a distance of 2.13 Å. Additional paths for Fe–Fe (at 3.02 Å) and Fe–P (at 3.35 Å) were used to fit the second and third shells, respectively. Additional features of the EXAFS spectrum were fitted with Fe–Fe paths (Fe–Fe2–4) at ca. 4.7, 5.25, and 6.25 Å. These values are in good agreement with theoretical path distances for neighbor shells in Fe(II) phosphates.[45−47] The inclusion of an Fe-O path and a Fe–O–O multiple scattering path at 3.98 and 4.46 Å, respectively, likewise improved the fit.

Hg Mass Balance and Reaction Stoichiometry with FeII

The ability of vivianite to reduce HgII (initial concentration of 50 μmol L–1) was first evidenced at FeII/HgII ratio of 1, where the Hg0 released was trapped and re-oxidized in acidic KMnO4 solution. In contact with vivianite nanoparticles, HgII concentration decreased concomitantly with the production of a roughly equivalent volatile Hg0 concentration (Figure ), according to an exponential decay (eq ) and an exponential rise (eq. ), respectivelywhere [HgII]0 and [Hg0]0 are the initial concentrations of cationic divalent mercury and elemental mercury, respectively; [HgII] and [Hg0] are the concentrations of cationic divalent mercury and elemental mercury at the plateau, respectively; and kobs is the apparent rate coefficient (min–1). After a 6 h incubation period, a color change was observed in the KMnO4 solution from dark purple to colorless (Figure S2), and the reaction between HgII and vivianite was almost complete with a HgII reduction of 46 ± 3% and an equivalent release of Hg0. The stoichiometry of the redox reaction between FeII and HgII was determined to be 1.98 ± 0.03 (Figure ), in agreement with the expected stoichiometry involving the oxidation of 2 mol of FeII to FeIII for the reduction of 1 mol HgII to Hg0. In this case, the Hg mass balance reached 92 ± 1% and the missing fraction is likely liquid Hg0 that remained bound to vivianite, which is then lost during the acidification of the suspension. The elemental distribution maps presented in Figure clearly show the presence of spherical or platy Hg enrichments, with a homogeneous size distribution ranging from about 20 to 40 nm (average 28 ± 5 nm, n = 37) associated with vivianite. No other element was detected in association with Hg, suggesting that these enrichments consisted of nanodroplets of Hg0. The formation of liquid Hg0 is favored under our experimental conditions with high Hg concentrations because its solubility is 0.3 μmol L–1, but it is however questionable if Hg0 formed under environmental conditions would also remained associated with the mineral phases. Bouffard and Amyot[48] found a significant proportion of the total Hg as Hg0 in the solid fraction of lake sediments (up to 28%) and hypothesized that Hg0 formed in porewater and then adsorbs to the solid phase through OM. Our finding represents an alternative explanation to why Hg0 can be found associated with solid phases in soils and sediments.

Figure 2

Time courses of inorganic Hg species (HgII: close circles, Hg0: open circles) during the reduction of HgII by vivianite at a FeII/HgII ratio of 1 with an initial HgII concentration of 50 μmol L–1. The inset shows the FeII consumption as a function of the HgII reduction between 0 and 6 h. Error bars represent standard deviation for two independent assays.

Figure 3

Elemental distribution of HgII-reacted vivianite at a FeII/HgII ratio of 1 after 5 h of reaction. The violet color of vivianite nanoparticles (a) results from the combination of Fe (red) and P (blue) colors. Hg enrichments (yellow) are clearly seen associated with vivianite after HgII reduction (b). Two spectra (c) were extracted from the elemental distribution maps for a Hg enrichment (b, area #1) and vivianite nanoparticle (b, area #2) and the size distribution of the Hg enrichments was also estimated (d).

Characterization of HgII-Reacted Vivianite

The changes in the Fe redox state and mineralogy of the HgII-reacted vivianite were investigated for a FeII/HgII ratio of 1 ([HgII]0 = 50 μmol L–1) and the Fe oxidation was visually confirmed when the mineral color changed from white to dark blue, typical of oxidized vivianite.[49] The chemical analysis of the reacted vivianite by the ferrozine method indicated that 76 ± 6% of FeII was oxidized to FeIII during HgII reduction (data not shown). Three doublets D1, D2, and D3 were also required to fit the Mössbauer spectrum of the reacted vivianite (Figure c and Table ). The CS and ΔEQ values were close to those obtained for the unreacted vivianite but the contribution of FeII and FeIII was reversed with 27.1 ± 0.3 and 72.9 ± 0.1%, respectively. Along with the reduction of HgII by vivianite, the Mössbauer data indicated a decrease of the FeII/FeTOT ratio from 0.95 to 0.23, in agreement with the results of the chemical analyses.

The average diameter of the HgII-reacted nanoparticles remained similar to the initial vivianite, that is, 45 ± 5 nm (data not shown) and the XRD diffractogram of the HgII-reacted vivianite was essentially the same as the starting material but with the presence of a small background feature (Figure a), which was interpreted as being from an amorphous phase(s), most likely ferric phosphate. Rouzies and Millet[50] already demonstrated that upon air oxidation at room temperature the monoclinic crystal system of vivianite was maintained until about 50% of the iron was present as FeIII and then alters to metavivianite. In our sample that was not exposed to air, no formation of crystalline metavivianite could be detected beyond the small amount originally present. Altogether, our results suggest that a significant proportion of the Fe is present as oxidized vivianite, which structural integrity was maintained after oxidation by HgII. This oxidized vivianite could potentially be regenerated under reducing conditions, as previously demonstrated for green rust[51] and magnetite,[39,52] and serves several times as an electron donor but this warrants further investigations.

Kinetics of HgII Reduction by Vivianite

Significant HgII reduction occurred in vivianite suspensions with FeII/HgII ratios of 1000, 100, and 1 but not at 0.1 (Figures a and S4). Control tests performed without vivianite or with free Fe2+aq (50 μmol L–1) did not lead to a significant removal of HgII (less than 5% in both cases) over the 24 h of the experiment. This indicates that HgII losses observed in the presence of vivianite were neither because of sorption onto the vessel walls or Teflon caps, nor because of reduction by free aqueous FeII species.

Figure 4

Reduction of HgII by vivianite. (a) Percent HgII reduced in suspension 24 h after its spike at 0.50 μmol L–1. Four FeII/HgII ratios were studied: 1000 (red), 100 (blue), 1 (black), and 0.1 (green). Two control experiments were performed with free Fe2+ (red stripes) and deoxygenated water (light blue). (b) Linearized kinetic data with dotted lines indicating the pseudo first-order law fit for FeII/HgII ratios of 1000 (red triangles), 100 (blue diamonds), 1 (black circles), and 0.1 (green squares). Error bars represent standard deviation of experimental replicates (n = 2–5).

For FeII/HgII ratios greater than or equal to 1, HgII reduction can be described by a pseudo first-order reaction (Figure S4, all fits presenting r2 ≥ 0.91), which is consistent with previous studies performed on HgII removal in the presence of FeII-bearing minerals.[16−20] Before 1 h (Figure b), the data for the 100:1 ratio show a slightly higher degree of scattering around the linear model, while for the 1000:1 ratio, the data do not fit the linear model. This is likely because the reaction is too fast with this large excess of Fe and the time sampling did not adequately resolve that period. However, when considering the other ratios and previous studies, it is very likely that the reaction also follows a pseudo-first order kinetic, given the large excess of Fe. The kobs values of 1.80 ± 0.03 × 10–3, 6.17 ± 0.07 × 10–3, and 11.7 ± 0.3 × 10–3 min–1, obtained for FeII/HgII ratios of 1, 100, and 1000, respectively, were in the same order of magnitude as those reported for siderite,[20] carbonate green rust,[18] and magnetite[16] (Table ). As expected, the higher the FeII concentration, the faster the rate of HgII reduction. However, for both FeII/HgII ratios of 100 and 1000, the reaction was still not complete after 24 h, and 20% of the added Hg remained as HgII, suggesting that a fraction of Hg was non-specifically adsorbed to vivianite. To allow a better comparison of the Hg reduction kinetic with other FeII-bearing minerals, it is crucial to consider the FeII surface site concentration and the FeII/HgII ratio. In previous studies investigating HgII removal by microcrystalline Fe minerals,[18] the SSA (m2 g–1) was evaluated by N2-BET measurements. Here, Avivianite was calculated to be 3.68 m2 L–1 according to the following equationwhere SSA is the SSA (42.0 m2 g–1, estimated from the size of the vivianite nanoparticles as described in the Materials and Methods section), MW is vivianite’s molecular weight (501.61 g mol–1), and [FeII3(PO4)2·8H2O] is the actual concentration of Fe in vivianite (500 μmol L–1 of Fe corresponding to a vivianite concentration of 167 μmol L–1). Because the average value of vivianite surface site density is 6 FeII sites per nm2,[53] vivianite should have 2.2 × 1019 FeII sites L–1 (i.e., 6 × 3.68 × 1018 nm2 L–1) and a total concentration of FeII surface sites of 36.7 μmol L–1 (Table ). Thus, kobs values were normalized to the total concentration of FeII surface sites to define ks values, and the ability of vivianite to reduce HgII is similar to those of carbonate green rust[18] and siderite[20] at FeII/HgII ratios of 400 and 7855, respectively. Nonetheless, the interpretation of t1/2 values in regard to FeII/HgII ratios should place vivianite as an ideal candidate to reduce HgII to Hg0 in FeII and P-rich environments, such as in freshwater or coastal sediments where the vivianite/HgII ratio is potentially even higher than the maximum FeII/HgII ratio (1000:1) studied here.

Table 2

Comparison of Kinetics Parameters for HgII Reduction by Vivianite, Siderite, Green Rusts, and Magnetite at Circumneutral pH under Anoxic Conditionsa

	FeII/HgII ratio	A (m2L–1)	FeII surface site (μmol L–1)	kobs (×10–3 min–1)	kS (L mmol–1 min–1)	t1/2 (min)
vivianiteb	1000	3.68	36.7	11.7	0.32	59
	100	3.68 × 10–1	3.67	6.17	1.68	112
	1	3.68 × 10–3	3.67 × 10–2	1.80	49	385
	0.1	3.68 × 10–4	3.67 × 10–3	0.14	38	4951
sideritec	31,382	24.72	246	14.7	0.06	47
	15,691	12.36	123	20.4	0.17	34
	7855	6.18	62	13.9	0.22	50
green rustsd	400e	10	83	28	0.34	25
	400f	4.6	38.2	130	3.40	5
magnetiteg	8300	2.0	75.2	96	1.3	7
	4150	1.0	37.6	54	1.4	13
	2075	0.5	18.8	24	1.3	29

Kinetic rate coefficients (kobs, min–1) normalized to the FeII surface site concentration (kS, L mmol–1 min–1) by taking into account the mineral surface area concentration (A, m2 L–1). HgII half-life (t1/2, min) of the pseudo first-order reaction is calculated from ln(2)/kobs.

This study.

Reference (20).

Reference (18).

Carbonate green rust.

Sulphate green rust.

Reference (16).

Environmental Implications

In the present study, we demonstrated that HgII is efficiently reduced by vivianite, with reaction yield and kinetics similar to those of carbonate green rust and siderite, considered up to now as the two most effective FeII minerals for Hg reduction. In P-rich environments, this pathway deserves special attention considering that the reactivity of green rusts is severely decreased in the presence of even low amounts of phosphates[18,54,55] and that vivianite could potentially be regenerated under reducing conditions as demonstrated for other Fe minerals.[39,51,52] The reducing capacity of vivianite toward HgII, however, requires further investigations, especially regarding the impact of dissolved organic ligands that may decrease the availability of Hg, and the presence of competing ions, such as chloride.[21] Nonetheless, this represents a newly discovered pathway for Hg reduction which could be of significant environmental relevance for the Hg biogeochemical cycle because vivianite has been detected worldwide in various settings. It might be especially important in organic-poor environments where the speciation of dissolved Hg is shifted toward inorganic Hg complexes, such as coastal sediments and groundwaters. Because of the presence of strong organic ligands, the reaction might be slower and/or quantitatively less important in organic-rich environments, such as eutrophic aquatic ecosystems, wastewater sludges, and paddy or organic soils where vivianite and Hg accumulates[23] but this would need to be systematically studied. While the methylation and re-emission of Hg from thawing permafrost soils is now attracting much attention,[56,57] vivianite was recently shown to be the main FeII-bearing mineral in organic horizons of an arctic tundra soil.[58]

The formation of vivianite under reducing microenvironments is also relevant for Hg methylation that also occur in these microniches, including in otherwise oxic and suboxic compartments. In lakes and reservoirs, it has been demonstrated that Hg methylation is promoted by eutrophication,[59−61] and thus the phosphorus delivery to the ecosystem. Sulfur-poor, Fe and organic/P-rich soils and sediments are potential hotspots for Hg methylation, where HgII is efficiently methylated by dissimilatory Fe(III) reducing bacteria,[62−64] such as Shewanella oneidensis MR-1, Geobacter sulfurreducens PCA, and Geobacter metallireducens GS-15.[65−67] However, it was also observed that the production of MeHg was suppressed in the presence of high concentrations of Fe. This was speculated to be caused by the scavenging of Hg by Fe minerals and/or its microbial reduction,[62,63,67] but the reduction of HgII by authigenic FeII-bearing minerals, including vivianite, should also be considered as a competitive pathway to methylation in such environments, thereby limiting its subsequent bioaccumulation.