Fate of cobalt and nickel in mackinawite during diagenetic pyrite formation

Mineral synthesis

Mackinawite (FeS_m) was synthesized from solutions of 0.6 M ferrous ammonium sulfate [Fe(NH₄)₂(SO₄)₂] and 0.6 M sodium sulfide (Na₂S) (Rickard 1997). Both solutions were made with ultrapure water (conductivity 0.052 µS) that had been boiled and cooled under an N₂ stream to make it anoxic. Sodium sulfide crystals were washed for 30 s with anoxic water to remove any oxidized S-species from the surface, then water was removed with a Pasteur pipette. Washed crystals were dried under an N₂ stream and stored in a N₂ atmosphere within a glass bottle with a butyl rubber stopper. Washed sodium sulfide crystals were weighed in an anoxic glovebox (100% N₂). The ferrous ammonium sulfate and sodium sulfide solutions were made in glass serum bottles and sealed with butyl rubber stoppers with an N₂ headspace. To synthesize FeS_m, equal volumes of each solution were added via N₂-purged syringes to a previously N₂-purged and stoppered serum bottle under a slight vacuum (0.5 mbar). FeS_m precipitated immediately. For the synthesis of FeS_m that contained either Ni or Co, 4.7 mg NiCl₂·6H₂O or 190 mg CoCl₂·6H₂O powders were weighed into serum bottles and stoppered, then flushed with N₂. This was accomplished by introducing N₂ through a long needle, which bubbled the solution, and letting the overpressure escape through a second, shorter needle. An appropriate volume of the ferrous ammonium sulfate solution was added to achieve concentrations of 500 µM Ni or 2000 µM Co when mixed with the sodium sulfide solution. These solutions were then mixed with sodium sulfide as described above.

Pyrite synthesis was attempted by two different methods. The first, oxidation of FeS_m with H₂S (Rickard 1997) (hereafter “H₂S oxidation”), was attempted to circumvent introduction of other solid S species (e.g., elemental sulfur, S₀) that would be detected with later microanalysis. The reaction 13323-4 of Rickard (1997) was chosen (400 mg ferrous ammonium sulfate, 80 ^°C, pH 6, –250 mV, 4 mM H₂S), as partial pyrite synthesis was reported from that reaction (80% pyrite and 20% mackinawite). Partial synthesis was desirable so that the partitioning of Co and Ni between the primary and secondary mineral could be evaluated. The anoxic solution of Ti(III) citrate and the phthalate buffer were made as described (Rickard 1997). The Na₂S solution was made as described for mackinawite synthesis but then adjusted to pH 6 with anoxic 1 N HCl. Dry FeS_m was weighed into a glass headspace vial with a small stir bar and closed with a stopper inside the glovebox. The buffer solution, Ti(III) citrate solution, and the Na₂S solution were introduced via syringes that had been flushed with N₂. The mixture was incubated in a sand bath at 80 ^°C on a temperature regulated hot plate, stirring, for either 4 or 20 days. At the end of the reaction, the headspace was flushed with N₂.

The second pyrite synthesis method reacted FeS_m with orthorhombic sulfur (S₈, with sulfur in the S₀ oxidation state) (Schoonen and Barnes 1991b), hereafter “sulfur oxidation.” FeS_m was synthesized as described above. Dried FeS_m (see below) was resuspended with pH 6 anoxic phthalate buffer in a stoppered serum vial. Orthorhombic sulfur was weighed and placed into a stoppered headspace vial, then the headspace was exchanged with N₂. Inside the glovebox, the orthorhombic sulfur and a small stir bar were added to the FeS_m-buffer solution and crimped with butyl rubber stoppers. The vials were placed in a sand bath at 65 °C on a temperature regulated hot plate, stirring, for 2 weeks. At the end of the reaction, the mineral suspension was centrifuged at 4000 rpm, the liquid removed with a 2-inch needle, and carbon disulfide (CS₂) was introduced from a stock under N₂ atmosphere to solubilize the unreacted orthorhombic sulfur. The suspension was centrifuged once again and CS₂ removed prior to washing of the suspension with water (see below).

After synthesis, minerals in solution were spun at 4000 rpm for 10 min while still sealed in glass serum bottles. The liquid was removed with a 2-inch needle while N₂ was being injected into the headspace using a second needle to maintain atmospheric pressure. Anoxic ultrapure water was added to the mineral suspension, and centrifugation repeated until the suspension had been washed twice. Then, vials were moved into the glovebox where the stoppers were removed. The serum bottles were covered with a paper tissue held in place by rubber bands. They were then placed in the vacuum chamber of the glovebox under a vacuum of –0.8 mbar until dried. Dried samples were stored in stoppered, N₂-filled glass bottles at room temperature until further use.

All glassware used in the above protocols was acid washed for at least 24 h in 1 N HCl. This was followed by a 24 h soak in ultrapure water, followed by three rinses in fresh ultrapure water.

Embedding and polishing

Dried samples of synthetic mineral grains were sprinkled into plastic 1-inch diameter rounds either on the bench or inside a glovebox (90% N₂/10% H₂). Samples were then immediately placed under vacuum within a Polyvac vacuum impregnator (Presi GmbH) at 80 kPa for 15 min. Then, Araldite epoxy 2020 A/B was slowly added (100:35 vol) until the sample was covered. Embedding proceeded at 80 kPa for 90 min. Samples were left to set for 24 h before removing from the impregnator.

Embedded samples were first ground with abrasive papers: 1200 (15 µm), then 2500 (8 µm) using a PRESI Mecatech 334. Polishing utilized a PRESI diamond stick LD 33 (3 µm) and Presi RAM polishing cloth, followed by a final polish with 0.05 µm deagglomerated g alumina on a Buehler Chemomet I cloth (BUEHLER Micropolisch II 0.05 µm). Sample mounts were stored and transported in a nitrogen atmosphere. Just prior to synchrotron-based microanalysis, samples were polished with a 1 µm cloth to remove oxidized coatings.

Bulk analysis

Particle size of synthesized samples was analyzed with a Mastersizer 2000 (Malvern Instruments GmbH). The surface area of solids was analyzed using a Micromeritics ASAP 2000 BET analyzer. Dried samples were loaded onto a holder and kept anoxic (100% N₂) until directly before analysis by m-X‑ray diffraction (XRD) on a Bruker D8 Discover instrument (Bruker, Germany) equipped with a CoKα X‑ray tube (k = 0.17902 nm, 30 kV, 30 mA) and GADDS area detector (Berthold et al. 2009). Mineral phases were identified using the International Center for Diffraction Data (ICDD) database.

Bulk trace element measurements were carried out by ICP-OES (Perkin Elmer Optima 5300 or Horiba Ultima 2) or an ICP-AES at University of Tübingen. Mineral samples (5–30 mg) were digested with aqua regia: 0.5 mL concentrated trace metal grade (TMG) HNO₃ (65%) and 1.5 mL concentrated HCl at 60 °C. Residues were resuspended in 8 N TMG HNO₃, and diluted 10-fold in 2% TMG HNO₃. Elemental sulfur that did not dissolve was filtered through a 0.2 µm Whatman filter. Aqueous samples were preserved with and further diluted in 2% TMG HNO₃.

Microscale analysis and spectroscopy

Elemental mapping by micro-X‑ray fluorescence (µXRF) and S and Fe K-edge X‑ray absorption near-edge structure (XANES) spectroscopy was performed at beamlines (BL) 14-3 and 2-3, respectively at Stanford Synchrotron Radiation Lightsource (SSRL). Sulfur distribution was mapped in approximately 1 × 1 mm areas at ~2500 eV on BL14-3, which has an energy range of 2100 to 5000 eV. The incident energy was selected using a Si(111) φ = 90 double crystal monochromator and the fluorescence signal was collected with a Si-drift diode Vortex detector (Hitachi) coupled to an Xspress3 pulse processing system (Quantum Detectors). The energy was calibrated to the thiosulfate pre-edge peak set to 2472.02 eV or a sulfate peak from CaSO₄ set to 2483 eV. The spot size was 5 × 5 µm, unless otherwise noted, and dwell times were between 50 to 200 ms. For a subset of samples, S maps were collected at multiple energies (ME) around the S K-edge that corresponded to characteristic absorption features for monosulfides, pyrite, elemental sulfur, sulfite/metal monosulfides, sulfate, and total sulfur, in order of increasing energy. In 2013 these energies were 2470, 2471.5, 2472.5, 2476.5, 2483, and 2490 eV. In 2014 they were 2469.7, 2471.2, 2472.2, 2476.2, 2482.7, and 2499 eV. In 2015 and 2016 these energies were 2470.7, 2471.1, 2472.6, 2478.5, 2482.5, and 2499 eV. These maps were then deadtime corrected, and subjected to a principle component analysis (PCA) in the Microanalysis Toolkit (http://smak.sams-xrays.com) (Webb 2011). Compositionally diverse spots based on the absorption at different energies above the S K-edge were revealed by PCA, and these spots were chosen for S K-edge spectroscopy. Fluorescence spectra were collected from 2460 to 2500 eV. Spectra were deadtime corrected and normalized using the SIXPACK software package (Webb 2005). Within SIXPACK, the collected spectra from each sample (3 to 23 spectra per sample) were subjected to PCA to reveal the number of unique spectra. These end-members (EM), representing the compositional heterogeneity of all µXANES from an individual sample, were then fitted with spectra of reference sulfur compounds available from the ESRF (https://www.esrf.eu/home/UsersAndScience/Experiments/XNP/ID21/php.html) and additional samples from mineral collections and laboratory synthesis that were analyzed on BL 14-3 [these include mackinawite (FeS) and polysulfide]. Mackinawite was bulk dried powder from the FeS_m synthesis described above, without added metals. The distribution of EM was then back-mapped using the characteristic fluorescence intensities at the specific energies in the MicroAnalysis Toolkit, similar to published methods (Mayhew et al. 2011; Farfan et al. 2018).

Iron, sulfur, and trace metals were mapped in approximately 0.5 × 0.5 mm areas at 8500 eV on BL2-3, which has an energy range of 4500 to 24 000 eV. The incident energy was selected using a Si(111) φ = 0 double crystal monochromator, and the fluorescence signal was collected with a Si-drift diode Vortex detector (Hitachi) coupled to an Xspress3 pulse processing system (Quantum Detectors). The energy was calibrated using the first inflection of an Fe₀ foil as a standard (7112 eV). The spot size was either 2 × 2 or 5 × 5 µm, and dwell times were between 50 to 200 ms. Spots with different element intensities were chosen for Fe K-edge XANES (3 to 10 spectra per sample). Data were processed as described above.

Reference spectra used for fitting Fe K-edge XANES have been previously published (O’Day et al. 2004). Additional reference spectra were for Fe XANES collected as bulk samples on BL 4-1 at SSRL. Powdered samples were analyzed in sandwiches of “Kapton” tape, or in “Teflon” (polytetrafluoroethylene) holder windows sealed with Kapton tape. Where necessary, powders were diluted with boron nitride to reduce self-absorption effects. The XANES spectra were collected in transmission mode as well as in fluorescence mode using a Lytle detector. The E₀ of an Fe₀ metal foil placed behind the first transmission detector was adjusted to 7112 eV to energy calibrate the resulting spectra. Replicate spectra were then averaged and background subtracted in SIXPACK (Webb 2005).

Semi-quantitative analysis of elemental abundance was accomplished by calibrating to NIST-traceable standards of the elements of interest, which were deposited on mylar films measured at the same detector distance, incident energy, dwell time, spot size, and detector gain as the samples. A calibration file was assembled containing the information from the standard maps in the MicroAnalysis Toolkit, and applied to the sample. Because the samples were in thick mounts, the sample thickness was approximated as the absorption length for the mineral host, either FeS_m or FeS₂. Because individual grains may be thinner, especially on the margins than this absorption thickness, it is important to emphasize that this approach is only semi-quantitative. The amount of element per area was converted to concentrations as milligram metal per gram sample assuming one absorption length of either FeS_m or pyrite as the sample depth. A mask was applied to the image to eliminate pixels that did not represent Fe- and S-containing particles, and the statistics were calculated on the selected pixels by the MicroAnalysis Toolkit.

Characteristics of synthesized minerals

The bulk mineralogy of synthetic samples was determined by XRD. Reference minerals were mackinawite, greigite, pyrite, pyrrhotite, and elemental sulfur. Synthesized FeS_m matched diffraction patterns for mackinawite, with minor greigite and pyrrhotite (Fig. 1). No peaks attributable to the iron oxides hematite, goethite, magnetite, or ferrihydrite were present (data not shown). The XRD pattern for FeS_m with no metals was very similar to that with added Co or Ni. Transformation of FeS_m with or without Co or Ni to pyrite via H₂S oxidation (Rickard 1997) did not produce pyrite. However, peaks for mackinawite were sharpest in the FeS_m sample with no metals that reacted the longest—20 days—whereas FeS_m with either Co or Ni (reacted 4 days) had lower intensity peaks, the weakest in the sample that contained Ni. Pyrite was synthesized from FeS_m both with and without Co or Ni using the sulfur oxidation protocol. Residual elemental sulfur was also detected in Co- and Ni-containing samples. The pyrite peaks had similar intensity irrespective of metal additions.

Figure 1

XRD of (a) FeS with and without Co and/or Ni, (b) FeS with and without Co and/or Ni after pyrite synthesis for 4 or 20 days using the H₂S oxidation protocol, and after pyrite synthesis with and without Co and/or Ni formed using the sulfur oxidation protocol. Reference reflections are shown for mackinawite (red), pyrite (black), elemental sulfur (yellow), greigite (green), and pyrrhotite (blue). Elemental sulfur was a residual phase following the sulfur oxidation protocol.

During synthesis with 0.1 L each of 0.6 M ferrous ammonium sulfate and Na₂S solutions, 5.27 g of FeS_m is calculated to have formed. FeS_m had a BET surface area of 25.96 ± 0.29 m²/g, within the range of values reported in the literature for similar methodologies. Other BET measurements on synthetic freeze-dried FeS_m resulted in 36.5 m²/g (Rickard 1997), 16–21 m²/g (Benning et al. 2000), and 47 m²/g (Wolthers et al. 2003). Particle size ranged from 2 to 138 µm, with greater than 90% of particles being less than 10 µm. Pyrite synthesized by sulfur oxidation had a BET surface area of 13.84 ± 0.21 m²/g. The surface area of Co-containing pyrite synthesized by sulfur oxidation was 18.06 ± 0.12 m²/g.

The results of S K-edge XRF and S XANES mapping of FeS_m synthesized with Co or Ni are shown in Figure 2, with the composition and fitting results of identified EM spectra reported in Table 1. FeS_m synthesized without metals was not mapped at ME through the S K-edge, so a similar map is not available. However, grains synthesized without metals were composed of FeS and sulfate, based on analysis of the spot S XANES spectra

Figure 2

Phase distribution maps generated from multiple XRF maps within the S K-edge and point XANES in FeS_m synthesized with Co or Ni. For each sample, the end-member XANES points labeled on the map (numbers) correspond to the spectra at right, and their distribution is shown in the same color as the corresponding spectra. Fits are dotted black lines. The fit compositions, as reported in Table 1, are also given. Note that FeS_m synthesized without metals was not mapped at multiple XRF energies. (a) Results of FeS_m synthesis with Co added. (b) Results FeS_m synthesis with Ni added.

(Fig. 3; Table 1). The standard quenstedtite [a hydrated ferric sulfate, Fe₂(SO₄)₃·11H₂O] was used as a proxy for the presence of an inorganic sulfate, as this fit all of the samples in the study requiring it fairly well. Substituting another sulfate mineral (e.g., anhydrite) did not change the fit significantly (e.g., more than 10% improvement on Σ² values; data not shown). It is important to note that inclusion of quenstedtite in the fit is not equivalent to detection of this mineral, as XANES is not a crystallographic technique. The use of quenstedtite in fits rather estimates the

Figure 3

(top) S XANES spectra from oxically embedded FeS_m compared to the same sample embedded anoxically (bottom). Data are black circles, and fits are thin black lines.

amount of an inorganic sulfate-bearing phase (Zeng et al. 2013). The FeS_m synthesized in the presence of Co or Ni similarly were fit with FeS and sulfate (Fig. 2). Although ME mapping was not performed at the Fe K-edge, Fe XANES from points on the FeS_m with added Co sample map were fit by FeS and pyrrhotite (Table 2). In this case, the FeS spectra used was collected from a bulk powder of the FeS_m synthesized without any metals. FeS_m

with added Ni was fit with mostly FeS and some pyrrhotite.

The synthesized FeS_m was sensitive to oxidation during embedding, polishing, and oxic handling during analysis. Figure 3 shows the EM S XANES spectrum of the FeS_m embedded oxically, which was fit by the reference spectra for FeS (0.68) and sulfate (0.37) (Table 1). The EM spectra from FeS_m embedded anoxically also fit with FeS and sulfate, but polysulfide was also necessary for a good fit (Fig. 3; Table 1). No ME map was collected, so it was not possible to assess the distribution of the phases. However, these spectra were determined to be representative of the 10 individual XANES collected on each sample by a PCA in SIXPACK. The FeS_m with no metal sample was not analyzed at the Fe K-edge.

The results of S K-edge XRF and S XANES mapping on samples treated with the sulfur oxidation method are presented in Figure 4, with the EM spectra displayed and used to determine phase distributions and the spectral fitting results reported in Table 1. The FeS_m with no metals added oxidized with the sulfur oxidation method resulted in predominately pyrite (Fig. 4a). Minor amounts of FeS were present in this and Co- and Ni-containing samples (Figs. 4b and 4c), visible as a shoulder at 2470 eV along the pyrite pre-edge absorption feature. Sulfate was also present, as with FeS_m samples. No sulfur was detected by XRF/XANES, although it was observed by XRD (Fig. 1). When Co was present, the grains were predominantly pyrite, with some polysulfide and residual FeS (Fig. 4b). The matrix was composed of sulfur and polysulfide, which was not seen in the metal-free sample. When Ni was added, most grains were made of pyrite with some polysulfide, or pyrite with FeS (Fig. 4c).

Figure 4

Phase distribution maps generated from multiple XRF maps within the S K-edge and point XANES in experiments where pyrite was produced from sulfur oxidation. For each sample, the end-member XANES points labeled on the map (numbers) correspond to the spectra at right, and their distribution is shown in the same color as the corresponding spectra. Fits are dotted black lines. The fit compositions, as reported in Table 1, are also given. (a) Results of pyrite synthesis from FeS_m with no metals added. (b) Results of pyrite synthesis from FeS_m with Co added. (c) Results of pyrite synthesis from FeS_m with Ni added.

The FeS_m containing either Co or Ni treated with the H₂S oxidation pathway did not produce pyrite, according to S XANES and XRF. The composition of EM S XANES spectra determined fitting was FeS with minor sulfate (Table 1). The Fe XANES spectra fit with an FeS standard (Table 2).

Trace element additions

The elemental concentrations of solutions used to synthesize FeS_m were analyzed, but S was not calibrated, and Fe was often above the range of the calibration curve and could not be quantified (Table 3). The ferrous ammonium sulfate solutions with nickel chloride or cobalt chloride were measured before adding the sodium sulfide solution and had a concentration of 65.28 mg/L (1112 µM) Ni, which was diluted by an equal volume of sodium sulfide solution to achieve ca. 500 µM Ni final

concentration. The solution with Co had 223.2 mg/L (3878 µM) Co, and was also diluted by an equal volume of the sodium sulfide solution to achieve ca. 2000 µM Co.

The Fe content of bulk solids determined with ICP-OES increased after applying the H₂S oxidation protocol (Table 3). The Fe content was only quantified in the Ni-containing sample after sulfur oxidation, due to the range of ICP-AES and standard concentrations. For the Co- and Ni-bearing solids, Co and Ni were quantified in bulk by either ICP-OES (FeS_m and H₂S oxidation samples) or ICP-AES (sulfur oxidation samples; Table3) or semi-quantitatively mapped by X‑ray fluorescence at 11 keV on individual pixels (Table 4). The measured values of Co and Ni in FeS_m, measured by ICP-OES were 1.05 ± 0.26 mg Ni per g FeS_m, and 3.71 ± 0.06 mg Co per g FeS_m, higher than the range of the median values obtained by synchrotron-based XRF mapping on beamline 2–3 at 11 keV: of 0.4 ± 0.2 mg Ni per g FeS_m and 3.1 ± 0.7 mg Co per g FeS_m. Quantification of Ni after sulfur oxidation diverged the most between methods. ICP-AES measured 3.4 ± 0.1 mg Ni per g mineral, while XRF maps contained median values of 0.5 ± 0.4 mg Ni per g mineral. Results for Co after sulfur oxidation were comparable by bulk and microscale methods. Maps of particles used for XRF quantification are shown in Supplementary^[1] Figure S1.

Mineral transformation

For the FeS_m synthesized in this study, XRD documented the mineral mackinawite, with minor greigite and pyrrhotite (Fig. 1). The best fitting standard for S XANES spectra was with FeS, a mackinawite standard, occasional polysulfide, and varying amounts of sulfate (Table 1). Even after embedding in the glovebox, the FeS_m still shows signs of oxidation, shown by the inclusion of sulfate and polysulfide in the fits, although it may be less oxidized than samples embedded in the air (Fig. 3). Less sulfate was needed to fit the anoxically embedded samples, however, indicating some success of the protection from oxygen. The presence of polysulfide may have resulted from some sample hydration during storage, as the sample was anoxically embedded a year after oxically embedded samples, although they were stored under dry N₂ in airtight containers. As sulfate minerals were not detected in XRD (detection limit 5 wt%), oxidation likely occurred during the storage, embedding or analysis of samples, and not during synthesis or transformation protocols. There was a major reflection at about 35° in the FeS_m that could be greigite or pyrite, but other major reflections for these minerals were absent. The addition of Co or Ni did not appear to alter the mineralogy of the FeS_m, or to change the intensity or width of reflections (Fig. 1).

The attempts to transform FeS_m, with or without Co and/or Ni,to pyrite via H₂S oxidation were not successful. From XRD, the sample still appeared to be mackinawite, with the most intense reflections in the Co-containing samples (Fig. 1). The Ni- and Co-containing samples were only reacted 4 days and yet had much more intense reflections for mackinawite than FeS_m without metals, reacted either 4 or 20 days. The failure of the H₂S oxidation method to produce pyrite from FeS_m synthesized with Co or Ni was also verified by S XANES and XRF mapping (Table 1), as well as by point Fe XANES analysis (Table 2). These results are consistent with several studies, which found very slow reactions between mackinawite and H₂S to form pyrite (Berner 1970; Schoonen and Barnes 1991b; Wilkin and Barnes 1996; Benning et al. 2000). It has been suggested that the freeze-dried FeS, used in the successful synthesis of pyrite via H₂S oxidation (Rickard 1997), underwent minor oxidation by oxygen during drying, providing oxidized surface sites for the formation of pyrite via reaction with H₂S (Benning et al. 2000). Secondary electron microscopy (SEM) studies of freeze-dried mackinawite in that second study noted an oxidized sulfate mineral phase, as well as a change in morphology from freshly precipitated mackinawite (Benning et al. 2000). In this same study, the freeze-dried material produced pyrite in reaction with H₂S (100% transformation after 9 days),

whereas freshly precipitated mackinawite reacted with H₂S did not. Our finding that N₂-dried mackinawite does not react with H₂S to form pyrite further support the claim (i.e., Benning et al. 2000) that H₂S oxidation of mackinawite is not a significant reaction in the formation of pyrite at low temperatures.

In contrast, pyrite was formed from the sulfur oxidation method, evident in clear reflections for pyrite from XRD in metal-free samples (Fig. 1). After pyritization, most XRD reflections for mackinawite had disappeared, indicating a near-complete reaction. From the S ME mapping and EM XANES, pyrite was also the main product (Fig. 4; Table 1), with minor FeS and sulfate, likely reflecting oxidation during sample mounting and/or analysis (e.g., Fig. 2). The spot Fe XANES analysis also documented pyrite, but the product still retained some FeS_m (Table 4). In addition, marcasite was a component of Fe XANES fits (Table 4), which likely reflects the same sulfur oxidation state and similar orbital energy levels of Fe in pyrite and marcasite (Jones 2006), rather than the presence of marcasite. The Fe XANES of pyrite and marcasite are very similar (O’Day et al. 2004). The inclusion of pyrrhotite in some fits could reflect some oxidation during embedding, storage, or analysis.

Pyrite was also produced in the Co-containing solids, as shown by XRD, S XANES, and Fe XANES. In the Co-containing samples, an additional intense reflection for sulfur was present at about 26°, and several minor reflections between 30–35° in the XRD data (Fig. 1). These likely correspond to residual sulfur retained after CS₂ washing. The single CS₂ wash must have been insufficient to remove sulfur. It is significant that evidence for residual sulfur is not seen in the metal-free product. In Co-containing samples analyzed by S XANES, FeS was still present in pyrite grains (visible as a shoulder at 2470 eV), as was polysulfide (Fig. 4). Sulfur and polysulfide comprised the matrix, consistent with the intense reflections for sulfur by XRD. Polysulfides form in the reaction of elemental sulfur with water, and themselves react with FeS_m to form pyrite (Schoonen and Barnes 1991b), and so are expected S phases during pyrite formation by the S⁰ oxidation method. By Fe XANES, both pyrite/marcasite and FeS were components of fits. Taken together, these data indicate that the presence of Co impeded reaction of FeS_m to pyrite in the time course of the experiments (2 weeks). Although FeS_m was still detectable in metal-free product, the absence of any detectable sulfur may signify that metal-free FeS_m reacted more completely than Co-amended FeS_m. Inhibition of FeS transformation to pyrite has been observed in the presence of arsenate [As(V)], although this is linked to arsenate oxidizing both Fe and S (Wolthers et al. 2007). An analogous pathway can be excluded for Co²⁺, as its reduction in the aqueous phase is unlikely. Arsenate and arsenite ([As(III)] can also prevent FeS transformation to pyrite by sorption and blockage of surface sites (Wolthers et al. 2007). Although, Co does not adsorb even as strongly to FeS as Ni (Morse and Arakaki 1993), yet Ni-containing FeS appears to have more completely reacted in our experiments (see below). The lower concentrations of Co in pyrite compared to FeS (Tables 3 and 4) indicate that some Co could have been mobilized from solid phases during the reaction and available for sorption. It is impossible to rule out inhibition of pyrite formation due to Co sorption with our data.

For Ni-containing samples, the predominant product of sulfur oxidation was pyrite, based on XRD, S XANES, and Fe XANES analyses. Some residual FeS_m was present based on S and Fe XANES (Tables 1 and 2). One Fe XANES EM also fit with maghemite, indicating Fe oxidation, which likely occurred during sample embedding, storage, or analysis, as it was not detected by XRD prior to these steps. Sulfur was also present based on XRD, but the diffraction peaks were not as sharp or as large as for the Co-containing product. Polysulfide was detectable in S XANES (Fig. 4). The reaction of FeS_m to pyrite also seems to be incomplete in the presence of Ni, but may have progressed further given the evidence for more abundant sulfur/polysulfide in the Co-containing grains. Another line of evidence is the more abundant distribution of sulfur-rich grains in Co-containing product, as revealed by the 8000 eV XRF map (Supplementary^[1] Fig. S2) as compared to the Ni-containing product, although the sample size is just a few grains.

These results of incomplete pyrite formation in Ni-containing samples are in contrast to the recent observation that the presence of Ni accelerates the rate of pyrite formation (Morin et al. 2017). In that work, pyrite was formed from de novo synthesis from slightly acidic (ca. pH = 5.5) solutions of ferric chloride and H₂S (Noël et al. 2014, 2015). In this mechanism, Fe(III) must be the oxidant, whereas our protocol started with solid FeS_m and utilized sulfur as the oxidant. The authors proposed that Ni impurities accelerated pyrite nucleation from solution (Morin et al. 2017), which is generally slow. However, formation of pyrite in marine sediments likely proceeds from the nucleation and subsequent sulfidization of FeS minerals, such as mackinawite and greigite (Schoonen and Barnes 1991a), rather than through reaction of aqueous Fe(III) and H₂S. This inference is supported by the observation of FeS phases being converted to pyrite with depth in numerous anoxic sediments (Cutter and Velinsky 1988; Huerta-Diaz and Morse 1992; Scholz and Neumann 2007). Our experiments, therefore, suggest that Ni (and Co) impurities might impede the rate of pyritization in pH circum-neutral sediments, although others note that Ni and Co may more completely pyritize than other metals (Morgan et al. 2012b).

The weight percent of Fe in FeS_m calculated from bulk digestion data in Table 3 was about 30–40%, less than the predicted 63.5% for a 1:1 Fe:S stoichiometry. This lower weight percent iron may have been caused by sulfate precipitation during drying, as discussed above. After the H₂S oxidation protocol, the weight percent of iron increased to 42–51%, closer to stoichiometric FeS, but still lower than expected. This may be due to increasing crystallinity of the FeS_m in the aqueous phase (Rickard 1995), which has support from the sharpening of diffraction peaks (Fig. 1) or washing out of excess sulfate with the fluids used for H₂S oxidation. Although the Fe content was not determined on most solids produced by the sulfur oxidation protocol, the Fe content for pyrite for the Ni-containing sample was ca. 33 wt%, whereas stoichiometric pyrite should have 46.55% Fe. This likely reflects residual sulfur in the solid sample after the sulfur oxidation method, as well as the persistence of FeS_m.

Trace element incorporation

The distribution coefficient for Co in FeS_m at 25 °C is 29 ± 3 (Morse and Arakaki 1993), which predicted 16 ± 2 mg Co per g FeS_m in our synthesized mineral. The measured values of Co were 2.6 to 3.7 mg Co per g mineral were well under this range. Based on the published distribution coefficient of 280 ± 181 for Ni in FeS_m at 25 °C (Morse and Arakaki 1993), we expected to see 3.9 ± 2.5 mg Ni per g FeS_m. The measured value of Ni in FeS_m (0.6 to 1.05 mg Ni per g mineral) was below this range (Table 3). In those experiments, mackinawite was synthesized by slow titration, whereas our FeS_m was synthesized rapidly by mixing a Co- or Ni-bearing iron solution with a sulfide solution. It is likely that this fast precipitation did not allow full uptake of the likely equilibrium values of Co and Ni observed in slow precipitation experiments (Morse and Arakaki 1993). This may be due to the slower exchange of H₂O ligand for S^2– on Co²⁺ and Ni²⁺ ions as compared to Fe²⁺ ions (Morse and Luther 1999).

After the H₂S oxidation protocol, the measured value was 4.05 ± 1.92 mg Co per g mineral, and 1.1 mg Co per g mineral after the sulfur oxidation protocol. Cobalt enrichments in FeS_m are generally present in all particles based on XRF mapping of the synthesized minerals at 8500 eV (Supplementary^[1] Fig. S2). In the samples mapped after sulfur oxidation, Co was distributed throughout the Fe-rich particles. From cross-plots of the Fe and S pixels, the Fe-rich particles always contain some S, and so are inferred to be FeS_m/pyrite. The sulfur-rich grains comprise a very low-Fe population of pixels and are likely sulfur/polysulfide-rich grains (analyses not shown). Cobalt is visible in the FeS_m/pyrite grains, but absent in the sulfur/polysulfide grains, indicating that Co is preferentially incorporated into sulfide minerals. From the bulk (ICP-OES or ICP-AES) and XRF metal quantification, there was general agreement between the techniques for the quantity of Co in both FeS_m and sulfur oxidation products (Fig. 5). However, Co was quantitatively lost from the samples during transformation. The lower concentration of Co in Fe-rich particles by XRF after pyrite transformation (Supplementary^[1] Figs. S1 and S2) indicates this is not a phenomenon of solid-phase dilution with residual sulfur for the ICP analyses. Therefore, it seems likely that Co is lost from the sulfide phase during the transformation of FeS_m to pyrite.

Figure 5

Amount of metal in the solids produced from FeS_m and pyrite synthesis in the presence of Co or Ni. The average of triplicate bulk measurements by ICP-AES or ICP-OES are in black, with error bars representing the standard deviation. The median of individual pixels in grains determined on XRF maps collected at 8500 eV. The error bars are the standard deviation of these data.

After treating FeS_m with the H₂S oxidation protocol, Ni increased to 1.43 ± 0.87 mg Ni per g mineral. The amount in pyrite produced from the sulfur oxidation protocol was 3.4 ± 0.1 mg Ni per g mineral. In Ni-containing samples, the FeS_m has Ni distributed throughout all grains. In the Ni-containing product oxidized by sulfur, both S- and Fe-rich grains are visible in the XRF map, similar to the Co-containing sample (Supplementary^[1] Fig. S2). In the product, some Fe-rich grains are preferentially enriched in Ni, while others are Ni-poor, distinct from the Co-rich product. The S-rich grains have little Ni. Asno multi-energy maps were made at these higher energies on BL2-3 (e.g., the Fe K-edge), it is not possible to say whether Ni preferentially incorporated into discrete iron sulfide phases, such as FeS_m or pyrite. Other studies have documented a preference of Ni for the pyrite vs. mackinawite mineral (Noël et al. 2015). This may be consistent with preferential Ni concentration in pyrite-bearing grains, as opposed to FeS_m grains, although we cannot validate this with the current data set. The ICP data always detected more Ni in the samples than is apparent with the synchrotron-based XRF maps (Fig. 5), likely indicating the need for further matrix correction for XRF standards. Both techniques show an increase of Ni in the solid after sulfur oxidation, although this increase is unlikely to be statistically significant in the XRF data. It is possible that in the case of dissolution of FeS, and incomplete pyrite formation, mobilized Ni was effectively sequestered in fewer grains, increasing the concentration.

A final observation is that the addition of Co or Ni shifted thepre-edge peak position of the FeS in the S XANES spectra to

more positive values (Fig. 6), which was more extreme for Ni. A similar peak shift was noted for pyrite, although only for Ni. Other authors noted longer bond distance (2.28 Å) for Ni-S than for Fe-S (2.26 Å) based on Ni EXAFS of mackinawite synthesized with Ni (Wilkin and Beak 2017). However, shorter bond distances generally require higher excitation energy. It may more likely reflect the ordering of stability constants of the transition metal series, e.g., the Irving-Williams order, Fe < Co < Ni < Cu > Zn (Rickard and Luther 2006). Although the amount of the non-Fe metal is small, these pre-edge peaks may reflect the increased energy needed to achieve electronic transitions to empty hybridized Me-S and S-S anti-bonding orbitals.

Figure 6

Overplots of end-member FeS_m and pyrite S XANES from samples synthesized with and without Co and Ni. The presence of Ni causes a shift for the pre-edge feature of FeS_m and pyrite to be shifted to higher energy. A subtle shift is observed with Co, but only for FeS_m.