Gypsum Amendment Induced Rapid Pyritization in Fe-Rich Mine Tailings from Doce River Estuary after the Fund ã o Dam Collapse

: Mine tailings containing trace metals arrived at the Doce River estuary, after the world’s largest mine tailings disaster (the Mariana disaster) dumped approximately 50 million m 3 of Fe-rich tailings into the Doce River Basin. The metals in the tailings are of concern because they present a bioavailability risk in the estuary as well as chronic exposure hazards. Trace metal immobilization into sulﬁdic minerals, such as, pyrite, plays a key role in estuarine soils; however, this process is limited in the Doce River estuarine soil due to low sulfate inputs. Thus, to assess the use of gypsum amendment to induce pyritization in deposited tailings, a mesocosm experiment was performed for 35 days, with vinasse added as carbon source and doses of gypsum (as a sulfate source). Chemical and morphological evidence of Fe sulﬁde mineral precipitation was observed. For instance, the addition of 439 mg of S led to the formation of gray and black spots, an Fe 2+ increase and decrease in sulﬁdes in the solution, an increase in pyritic Fe, and a greater Pb immobilization by pyrite at the end of the experiment. The results show that induced pyritization may be a strategy for remediating metal contamination at the Doce River estuary.


Introduction
The mineral pyrite (FeS 2 ) is widely known for its potential for metal immobilization in coastal wetland soils [1][2][3][4]. The formation of pyrite in soils is controlled by edaphic factors such as iron (Fe) and sulfide concentration in soil solutions, soil organic matter, microbial activity, redox potential (Eh), pH, soil moisture [5][6][7][8]. The formation of pyrite in coastal wetland soils results from anaerobic metabolic pathways for the degradation of organic matter using electron acceptors other than O 2 under reducing conditions (e.g., Eh < 100 mV) [5]. Accordingly, due to the high abundance of Fe and sulfate (SO 4 2− ) from the ocean, the microbial reduction of Fe 3+ and SO 4 2− is the most important anaerobic process for organic matter degradation in coastal wetlands soils [9,10]. In this sense, inducing the pyritization in soil has been identified as a strategy for increasing metal immobilization [11,12].
Since 2015, when the Fundão tailings dam in Brazil collapsed (the largest dam failure of the world, known as "Mariana disaster"), metal contamination has been identified as one of the most concerning phenomena along the Brazilian Coast [13][14][15][16][17]. After the Mariana disaster, a huge amount of Fe-rich mining tailings entered the Doce River estuary increasing the trace metal content in the soil [14]. Furthermore, the biogeochemical conditions of the estuarine soils have since shown an increase in trace metal bioavailability due to the dissimilatory Fe reduction process [14,16,18]. Additionally, due to geomorphological conditions, the Doce River estuary presents some unique characteristics such as low levels of seawater intrusion and salinity, which reduce the occurrence of sulfate reduction and increase the contamination risks [14,16].
According to the US Environmental Protection Agency (EPA) toxicity classifications, soils at River Doce estuary can be classified as highly polluted with Pb (Table 1) since the concentration of this toxic element in the estuarine mine tailing deposits is 48.8% higher than the threshold effect level (TEL) (35 mg kg −1 ) [19]. Lead (Pb) poisoning is a global concern and sulfidation of Pb is a relevant sink [4] because Pb is a metallic ion from "class B" or "soft sphere" and has a high affinity for forming strong soluble complexes with bisulfides and polysulfides [4]. Thus, improving pyritization through sulfur amendment is a useful strategy for inducing pedogenetic processes (i.e., sulfidation), and forming a Technosol, which can provide valuable ecosystem services, such as metal contamination remediation. In this sense, this study aimed to evaluate the use of different amendments to promote pyritization and reduce trace metal bioavailability from the Mariana disaster, supporting future estuarine reclamation.
Chemical immobilization is one of the most efficient in situ remediation techniques, especially when it uses cost-effective soil amendments [20,21]. We used gypsum (CaSO 4 ), a waste product of the phosphorus fertilizer industry, as a source of sulfate and vinasse, a waste from the sugarcane industry, as a source of carbon. These two waste materials were chosen because of their abundance and chemical characteristics, but also because of their low cost reduces the cost of remediation efforts [22,23]. For instance, gypsum is easily solubilized, which facilitates its action as a sulfate source in the field [24], whereas vinasse is rich in low molecular weight organic compounds (e.g., glycerol, lactic acid, ethanol, acetic acid, fructose, glucose, and sucrose) that are easily decomposed [25].
Therefore, we hypothesized that gypsum can be used as a source of sulfate for pyritization in Fe-rich mine tailings deposited in the Doce River estuary and that there is an optimum dose of gypsum for increasing pyritization. To test these hypotheses, we carried out mesocosm experiments for 35 days using mine tailings, with additions of vinasse and varying gypsum doses.

Site Characterization
Mine tailings were collected in August 2019 in the Doce River estuary in Regência, Espirito Santo state, Brazil ( Figure 1A). Since 2015, the Doce River estuary has been affected by mine tailings from the "Fundão" dam collapse. A thin layer of tailings (almost 5 cm) within the estuary was recognized 15 days after the tailings' arrival in Regência [14].
In August 2019, mine tailings deposited on islands in the estuary reached over 30 cm ( Figure 1B). In 2015 pseudo-total Fe, previously reported by [14], was 8.8 times higher than the pseudo-total Fe in the tailings collected in 2019. The water of the Doce River estuary has low salinity and a low sulfate content [17], which increases the risks of metal pollution through the microbial reduction of Fe and a lack of sulfate reduction to precipitate pyrite. This corroborates the low degree of pyritization in the estuary (<1%) reported by [14]. To evaluate the use of gypsum and vinasse as soil amendments we conducted a mesocosm experiment whereby different doses of gypsum and vinasse were added to mine tailings. In August 2019, mine tailings deposited on islands in the estuary reached over 30 cm ( Figure  1B). In 2015 pseudo-total Fe, previously reported by [14], was 8.8 times higher than the pseudo-total Fe in the tailings collected in 2019. The water of the Doce River estuary has low salinity and a low sulfate content [17], which increases the risks of metal pollution through the microbial reduction of Fe and a lack of sulfate reduction to precipitate pyrite. This corroborates the low degree of pyritization in the estuary (<1%) reported by [14]. To evaluate the use of gypsum and vinasse as soil amendments we conducted a mesocosm experiment whereby different doses of gypsum and vinasse were added to mine tailings.

Characterization of Mine Tailings and Amendments
Samples of mine tailings deposited at the Doce River estuary were stored at approximately 4 °C in a laboratory to prevent chemical and biological alterations [18]. During the sampling procedure, redox potential (Eh) and pH values were determined using a platinum electrode and a previously calibrated glass electrode, respectively. The collected mine tailings were characterized according to their particle size distribution, mineralogical composition, organic carbon content, nitrogen content, and total metal content ( Table 1).
The particle size distribution of mine tailings was determined using the pipette method [26]. The mine tailing mineralogical assemblage was determined using X-ray diffraction (XRD) with Cu-Kα radiation, at 0.02° 2θ s −1 in the range of 3-60° 2θ. The XRD analysis was performed using non-oriented powder mine tailings (Supplementary Figure  S1). The total organic carbon (TOC) and nitrogen (TN) contents were obtained using an elementary analyzer (LECO 144 SE-DR, TrusPec®, Saint Joseph, MI, USA) [27]. The total metal contents were obtained after tri-acid digestion in a microwave using the Environmental Protection Agency (EPA) method 3052 [28] and determined by Inductively coupled plasma optical emission spectrometry (ICP-OES) (Thermo Fisher Scientific, Waltham, MA, USA), according to the 6010C protocol [29]. Prior to the mesocosm experiment, the gypsum and vinasse were chemically characterized [30], in terms of pH and TOC, as well as TN, P, K, Mg, Ca, and S content ( Table 2).

Characterization of Mine Tailings and Amendments
Samples of mine tailings deposited at the Doce River estuary were stored at approximately 4 • C in a laboratory to prevent chemical and biological alterations [18]. During the sampling procedure, redox potential (Eh) and pH values were determined using a platinum electrode and a previously calibrated glass electrode, respectively. The collected mine tailings were characterized according to their particle size distribution, mineralogical composition, organic carbon content, nitrogen content, and total metal content ( Table 1).
The particle size distribution of mine tailings was determined using the pipette method [26]. The mine tailing mineralogical assemblage was determined using X-ray diffraction (XRD) with Cu-Kα radiation, at 0.02 • 2θ s −1 in the range of 3-60 • 2θ. The XRD analysis was performed using non-oriented powder mine tailings (Supplementary Figure  S1). The total organic carbon (TOC) and nitrogen (TN) contents were obtained using an elementary analyzer (LECO 144 SE-DR, TrusPec ® , Saint Joseph, MI, USA) [27]. The total metal contents were obtained after tri-acid digestion in a microwave using the Environmental Protection Agency (EPA) method 3052 [28] and determined by Inductively coupled plasma optical emission spectrometry (ICP-OES) (Thermo Fisher Scientific, Waltham, MA, USA), according to the 6010C protocol [29]. Prior to the mesocosm experiment, the gypsum and vinasse were chemically characterized [30], in terms of pH and TOC, as well as TN, P, K, Mg, Ca, and S content ( Table 2).

Experimental Design
A mesocosm experiment composed of six treatments was analyzed weekly for 35 days to evaluate pyrite formation after the addition of gypsum and vinasse. he mine tailings (25 g) were incubated in 50 mL Falcon ® tubes, in triplicate, forming a factorial scheme, composed of 6 treatments and 5 evaluation times (destructive samples obtained weekly), totaling 90 experimental units ( Figure 2). The treatments were set up as follows: -Control: only deionized water (0 mg of S); -Vinasse: only 6.2 mL of vinasse addition (as a carbon source: 51.7 mg of C, but with 2.4 mg of S); -Vin + G1: 6.2 mL of vinasse + 0.5 g of gypsum (89.7 mg of S); -Vin + G2: 6.2 mL of vinasse + 2.5 g of gypsum (439 mg of S); -Vin + G3: 6.2 mL of vinasse + 5 g of gypsum (875 mg of S); -Vin + G4: 6.2 mL of vinasse + 10 g of gypsum (1748 mg of S).
The treatments were randomized. The experimental units were maintained at a controlled temperature (25 ± 2 • C) and covered with aluminum foil to block any light. The aluminum foil cover was removed, and the tubes opened only during the evaluations.

Experimental Design
A mesocosm experiment composed of six treatments was analyzed weekly for 35 days to evaluate pyrite formation after the addition of gypsum and vinasse. he mine tailings (25 g) were incubated in 50 mL Falcon ® tubes, in triplicate, forming a factorial scheme, composed of 6 treatments and 5 evaluation times (destructive samples obtained weekly), totaling 90 experimental units ( Figure 2). The treatments were set up as follows: The treatments were randomized. The experimental units were maintained at a controlled temperature (25 ± 2 °C) and covered with aluminum foil to block any light. The aluminum foil cover was removed, and the tubes opened only during the evaluations.
Destructive samples were collected weekly for Fe and Pb solid-phase fractionation analysis. Geochemical fractionation was carried out using a combination of methods proposed by [33][34][35], which enable the extraction of six operationally distinct fractions: Exchangeable

Aqueous and Solid-Phase Analysis
Solution aliquots were collected weekly from mesocosm experiments to determine the concentration of dissolved Fe 2+ [31], sulfides (HS − and H 2 S) [32] and Pb [29].
Destructive samples were collected weekly for Fe and Pb solid-phase fractionation analysis. Geochemical fractionation was carried out using a combination of methods proposed by [33][34][35], which enable the extraction of six operationally distinct fractions: Exchangeable and soluble metals (EX) were extracted after 30 min (min) of agitation in 30 mL of 1 mol·L −1 MgCl 2 solution (pH 7.0).
Metals associated with carbonates (CA) were extracted after 5 h of agitation in 30 mL of 1 mol·L −1 NaOAc solution (pH 5.0).
Fe and other metals associated with ferrihydrite (FR) were extracted after 6 h at 30 • C of agitation in 30 mL of hydroxylamine solution 0.04 mol·L −1 + 25% acetic acid (vol/vol) solution.
Fe and other metals associated with lepidocrocite (LP) were extracted after 6 h at 96 • C of agitation in 30 mL of hydroxylamine solution 0.04 mol·L −1 + 25% acetic acid (vol/vol) solution.
Fe and other metals associated with goethite/hematite (OX) were extracted after 30 min at 75 • C of agitation in 20 mL of 0.25 mol·L −1 sodium citrate + 0.11 mol·L −1 sodium bicarbonate with 3 g of sodium dithionite.
Fe and other metals associated with pyrite (PY) were extracted after eliminating the Fe and metals associated with silicates (using 10 mol·L −1 HF) and the organic material (using concentrated H 2 SO 4 ), the pyritic phase was extracted by shaking the residue for 2 h at room temperature with 10 mL of concentrated HNO 3 . A more detailed description of the combined method is presented in [3,36]. Using this method, it is possible to calculate the degree of pyritization (DOP, %) and degree of trace metals pyritization (DTMP, %), which represent the proportion of pseudo-total contents incorporated into pyrite, as follows: where Fe − PY and ME − PY is the concentration of Fe and other metals associated with the pyritic fraction (PY, in mg kg −1 ), ΣFe − EX → Fe − PY is the sum of all Fe fractions and Σ ME − EX → ME − PY is the sum all respective metal fractions.

Statistical Analysis
Variables were subjected to the non-parametric Kruskal-Wallis test to assess differences between treatments with the level of significance set to p < 0.05 (XLSTAT version 2014.5.03) [37]. Spearman correlation analysis was performed, and a model for trace metal pyritization was obtained using mixed-effect linear models through stepwise backward elimination. These analyses were performed in R [38].

Effects of Time and Amendment Dose
Pb in the aqueous phase was lower than the detection limit (0.01 mg·L −1 ) in all treatments. Significant differences in the behavior of oxygen-sensitive chemical species (reduced iron, Fe 2+ , and sulfide species, HS − and H 2 S) were observed in treatments over time (Table S1). The Fe 2+ concentrations increased from the third week, especially in treatments including gypsum (Figure 3), followed by a decrease in the Fe 2+ concentration from the fourth week. H 2 S and HS − concentrations exhibited similar behavior to Fe 2+ with a significant increase from the second week and followed by a decrease (Figure 3). The addition of more gypsum resulted in more Fe 2+ in the liquid phase, but the sulfide maximum concentration was observed with the Vin + G2 treatment (439 mg of S).
The increase in Fe 2+ and sulfide concentration indicates the Fe 3+ and SO 4 2− reduction processes, whereas the subsequent decrease in their concentration indicates precipitation [7]. Previous studies have reported models of pyritization [7,39] showing that dissolved Fe 2+ reacts with HS − and precipitates as FeSx species, such as mackinawite, greigite or other acid volatile sulfides (AVS), and pyrite [40,41]. The pathway of pyrite formation is very dependent on the Fe/S molar ratio and an excess of Fe 2+ is required to precipitate FeS 2 [7,38]. In our experiment, the sulfide concentration in the solution was in the range of 29-600 µM, representing common sulfide concentration in coastal wetland soils and sediments [36,[42][43][44], and allowing pyritization. These behaviors of oxygen-sensitive chemical species (reduced iron, Fe 2+ , and sulfide species, HS − and H2S) led to morphological changes over time ( Figure 4) with an increase in gray and black spots in the mine tailings, especially in the treatments comprising gypsum and vinasse. Such changes are morphological evidence of sulfide formation [12]. In the vinasse only treatment the tailings were less red than those in the control treatment. In other words, there were more black spots in the treatments with vinasse and gypsum than in the treatments with vinasse only. These behaviors of oxygen-sensitive chemical species (reduced iron, Fe 2+ , and sulfide species, HS − and H 2 S) led to morphological changes over time ( Figure 4) with an increase in gray and black spots in the mine tailings, especially in the treatments comprising gypsum and vinasse. Such changes are morphological evidence of sulfide formation [12]. In the vinasse only treatment the tailings were less red than those in the control treatment. In other words, there were more black spots in the treatments with vinasse and gypsum than in the treatments with vinasse only.

Variation of Solid-Phase Content over Time and as a Function of Amendment Dose
A larger decrease was observed in the pseudo-total Fe content when more sulfur was added ( Figure 5), with Fe loss of −6.6% (Vin + G1), −13.6% (Vin + G2), −19.4% (Vin + G3), and −27.4% (Vin + G4). The decrease in pseudo-total Fe was associated with losses from oxide forms, mostly Fe-CR (on average −21%; Figure 6) which corroborates the Fe reduction process.

Variation of Solid-Phase Content over Time and as a Function of Amendment Dose
A larger decrease was observed in the pseudo-total Fe content when more sulfur added ( Figure 5), with Fe loss of −6.6% (Vin + G1), −13.6% (Vin + G2), −19.4% (Vin + and −27.4% (Vin + G4). The decrease in pseudo-total Fe was associated with losses f oxide forms, mostly Fe-CR (on average −21%; Figure 6) which corroborates the Fe re tion process. On the other hand, treatments with vinasse and gypsum enhanced the Fe content associated with carbonates (+784.1 mg·kg −1 or 50 times more Fe-CA than the control) and pyrite (+471.2 mg·kg −1 or 22 times more Fe-PY than the control). The Fe-CA increase was time-dependent: on average, there was 180 times more Fe-CA after 35 days of incubation than after 7 days of incubation. The increase in Fe-CA content in the mesocosm experiment may be a combination of CO 3 2− formation due to microbial respiration [45] and greater supply of Ca 2+ from both gypsum and vinasse ( Table 2). The CO 2 released from organic matter mineralization leads to the formation of CO 3 2− which in the presence of free Ca 2+ precipitates as CaCO 3 [46,47]. The formation of carbonates (i.e., Fe-CA) may play a key role in metal immobilization since Fe-CA may be more stable mineral fraction than Fe oxyhydroxides in estuarine soil with limited pyritization processes [14,48].     The increase in Fe associated with pyrite was dependent on the amount of S ( Figure 6) and the highest amount of Fe-PY was obtained using 2 to 439 mg of S. These results indicate that even the sulfur content present in the vinasse (2 mg) may be sufficient to increase the pyritization in Fe-rich mining tailings deposited in the Doce River estuarine soil. In particular, in the present study on reactive Fe (i.e., a sum of Fe-EX to Fe-CR), the Fe/S ratio ranged from 3 to 919 (Table S2) revealing conditions favorable for pyritization [39,41], as several studies have reported that an Fe/S molar ratio above two favors the pyritization [40,41].
Fe oxides (Fe-FR, Fe-LP and Fe-OX) showed negative correlations with the amount of S added (Figure 7). On the other hand, high sulfur amendment contributes to a decrease in the concentration of Fe-oxides, resulting in a lower Fe pseudo-total content ( Figure 5). The addition of sulfur leads to the formation of sulfides (Figure 4) and other sulfur species (for instance, sulfite (S 2 O 3 2− ) or elemental sulfur (S 0 )) that are expected to reductively dissolve Fe(III) minerals, through the sulfidation process [11,40] according to the following reactions:

Pb Pyritization and Modeling
We evaluated Pb to assess the potential for other metals' pyritization. In the control treatment almost 45% of all Pb was already associated with pyrite, but the last 55% was mostly associated with iron oxides (Table S4), which can be reductively dissolved in estuarine conditions, mobilizing the trapped Pb.
Significantly higher Pb-DTMP was observed with the addition of gypsum (Figure 8), and there was a strong negative correlation (Figure 7) between Pb associated with lepidocrocite and the amount of S added to the mine tailings, indicating that the addition of gypsum affected the solid phase speciation of Pb, and Pb associated with lepidocrocite can be a source to Pb associated with pyrite. An addition of 439 mg of S was sufficient to raise the DTMP-Pb to 63%, lowering the labile or reactive Pb to less than TEL (35 mg•kg −1 ).

Pb Pyritization and Modeling
We evaluated Pb to assess the potential for other metals' pyritization. In the control treatment almost 45% of all Pb was already associated with pyrite, but the last 55% was mostly associated with iron oxides (Table S4), which can be reductively dissolved in estuarine conditions, mobilizing the trapped Pb.
Significantly higher Pb-DTMP was observed with the addition of gypsum (Figure 8), and there was a strong negative correlation (Figure 7) between Pb associated with lepidocrocite and the amount of S added to the mine tailings, indicating that the addition of gypsum affected the solid phase speciation of Pb, and Pb associated with lepidocrocite can be a source to Pb associated with pyrite. An addition of 439 mg of S was sufficient to raise the DTMP-Pb to 63%, lowering the labile or reactive Pb to less than TEL (35 mg·kg −1 ). Similar results were obtained in a study using K 2 SO 4 as a sulfate source to induce the formation of Pb sulfides in wetland tailings [11]. Linear regression using the stepwise methodology resulted in two equations for modeling Pb pyritization, each one with DTMP-Pb and Pb-PY as dependent variables ( Table  3). The selected equations had the highest conditional R² and highest Akaike's information criterion (AIC), minimizing data loss ( Table 3). As expected, modeling showed that Pb pyritization is highly dependent on the sulfides content in the solution (Table 3) and this variable had the highest degree of fixed effects on Pb-PY and on the DTMP-Pb models. In addition, the degree of pyritization was highly relevant on Pb-pyritization. In other words, the precipitation of Pb with carbonates had negative effects on Pb pyritization, so in this batch experiment pyritization competed with the formation of carbonates (Table  3). This is corroborated by the high Fe-CA contents observed in the final weeks of experimentation with vinasse and gypsum addition ( Figure 5) and the strong and positive correlation between Fe-CA and experiment time (Figure 7). Table 3. Summary of the mixed-effect linear model for Pb pyritization obtained via stepwise backward elimination. Linear regression using the stepwise methodology resulted in two equations for modeling Pb pyritization, each one with DTMP-Pb and Pb-PY as dependent variables ( Table 3). The selected equations had the highest conditional R 2 and highest Akaike's information criterion (AIC), minimizing data loss ( Table 3). As expected, modeling showed that Pb pyritization is highly dependent on the sulfides content in the solution (Table 3) and this variable had the highest degree of fixed effects on Pb-PY and on the DTMP-Pb models. In addition, the degree of pyritization was highly relevant on Pb-pyritization. In other words, the precipitation of Pb with carbonates had negative effects on Pb pyritization, so in this batch experiment pyritization competed with the formation of carbonates (Table 3). This is corroborated by the high Fe-CA contents observed in the final weeks of experimenta-tion with vinasse and gypsum addition ( Figure 5) and the strong and positive correlation between Fe-CA and experiment time (Figure 7).

Conclusions
The results show that it is possible to increase the pyritization in Fe-rich mine tailings using sulfur amendments (e.g., gypsum and vinasse), since the concentration of Fe and Pb associated with pyrite increased in a short time (5 weeks) when compared with the control treatment. An optimum gypsum dose of 439 mg of S per 25 g of tailings was determined for increasing the pyritization of Fe and Pb. The results obtained in this study not only provide a base on which to design strategies for estuarine soil trace metal remediation and River Doce estuary reclamation, but also for other wet environments exposed to a huge amount of Fe. Future research should examine how to improve sulfate reduction (for instance, inoculation of the tailings with sulfate reducing bacteria) and consider the effects of real-world conditions (such as open atmosphere, tide effects, and long-term gypsum dissolution) to develop effective technologies and techniques.
Supplementary Materials: The following are available online at https://www.mdpi.com/2075-1 63X/11/2/201/s1, Figure S1: Bulk samples XRD patterns from recently mine tailing deposited at Doce river estuary, Table S1: Average, standard deviation and p-values from the Kruskal-Wallis test, Table S2: Reactive Fe (sum of Fe-EX to Fe-OX) and sulfide (H 2 S and HS − ) molar ratio, Table S3: Correlation coefficients (in bold) and p-value of Spearman correlation analysis, Table S4: Pb solidphase fractionation from the mesocosm experiment over time.