The areas studied are located in Mexico, in the northwest (Buenavista del Cobre mine, Cananea; and El Tecolote mine tailings impoundment, Pitiquito; both in the Sonora State); in the north central part (a Zn refinery in San Luis Potosi city); and in the south central part (El Fraile mine tailing impoundment, in Taxco, Guerrero). The sampled points are shown in Figure 1. In the case of the Buenavista del Cobre mine, soils were collected along the dry river bed. The climates of the areas are categorized as semiarid and arid, except for Taxco, which is temperate (INEGI, 2019).

Figure 1 Figure 1 Map showing the study areas and location of samples: a) El Tecolote mine tailings in Pitiquito, Sonora, b) Buenavista del Cobre mine in Cananea, Sonora, c) El Fraile mine tailings in Taxco, Guerrero, and d) Solid hydrometallurgical wastes in San Luis, San Luis Potosi.

The Buenavista del Cobre mine currently produces Cu from porphyry copper mineralization, which is hosted mainly in volcanic rocks. The lithology of the porphyry inclusion presents quartz-monzonite, monzodiorite, and granodiorite, accompanied by potassic and quartz–sericite alterations and zinc–lead–copper (Zn-Pb-Cu) skarns with strata bound high-grade sulfide and iron-oxide deposits (Meinert, 1982; Ochoa-Landin et al., 2011). As mentioned before, the Cu extraction currently consists of spraying and leaching concentrated sulfuric acid through the Cu-rich rock piles.

In the Taxco mining area the mineralization appears mainly in hydrothermal veins, replacement ores, and stock works hosted in limestone, shale and schist. The flotation treatment plant El Fraile generated tailings that were deposited on calcareous shale. These wastes were deposited in two zones, one of smaller particles (covered with calcareous shale) and another of coarser particles, which is exposed since 1973 (Romero et al., 2007).

El Tecolote tailing impoundment comes from an inactive Cu-Zn- Ag mine. It was a skarn deposit with mineralization of Zn and Cu. In this area the material extracted underground was processed by flotation. The waste generated by this process was deposited directly on calcareous rocks (Cruz-Hernandez et al., 2018).

In San Luis Potosi, the study area is limited to the deposition of the solid wastes produced by a Zn electrolytic refinery; these wastes are composed mainly of jarosite, which was synthesized by the metallurgical operations since it takes up Fe and other present elements, and leaves Zn in a dissolved form devoid of other major elements for further purification.

All these other sites were selected for sampling because secondary Fe minerals were detected in previous works (Romero et al., 2007; Cruz-Hernandez et al., 2018); therefore, they were considered as good references for comparison with the Buenavista del Cobre mine.

Nine samples were analyzed in total, including four from the Buenavista del Cobre mine and five from other acidic mine-metallurgical environments. The details are shown in Table 1. Although this number may seem small, beamline work at the synchrotron facility (ALS, Lawrence Berkeley National Laboratory – cf. below) for each sample is extremely laborious. We were granted a total of four different sessions consisting of fifteen 8-hour cycles to work at the beamline, which is adequate for obtaining good analytical results. The Buenavista del Cobre mine samples were selected with the following criteria: 1) soil visibly contaminated by the acid spill and confirmed from previous wet-chemical analyses (Ca1 sample); 2) equivalent contaminated soil but treated with lime to neutralize pH (Ca2 sample) and determine this effect on the secondary Fe mineralogy on a very short time-scale; and 3) soil not evidently contaminated, from the same basin, found at 10–20 km distance downriver from the spill (samples Ca3 and Ca4), in order to detect any influence from the spill in the Fe mineralogy. Also, to compare the Fe mineralogy in different acidic mine and metallurgical environments, three samples from other mine tailing impoundments and two samples from hydrometallurgical wastes were processed and analyzed with the same techniques as the samples from the Buenavista del Cobre mine.

Table 1 Table 1 Location and description of the analyzed samples. a

Sampling dates: Ca1, Ca2 August 2014; Ca3, Ca4 February 2015.

The samples were collected by the environmental geochemistry group of the Laboratorio Nacional de Geoquimica y Mineralogia, Instituto de Geologia (LANGEM), Universidad Nacional Autonoma de Mexico (UNAM) according to the mexican norm MNX-AA-132- SCFI-2006 (SE, 2006), from the top 5 cm of material; with the exception of Ta2, which was a combined sample of a mine tailings vertical profile collected from 40 to 180 cm in Taxco, Guerrero. The samples were air-dried and passed through a 2-mm round-hole sieve. The pH and E.C. (electrical conductivity) were determined in a 1:5 solid:water ratio following the ISO norm 10390:2005 (ISO, 2005) for the contaminated soils from the Sonora river basin and in a 1:1 ratio in accordance with Thomas (1996) for the samples from the mine tailings. All aqueous solutions were prepared with Milli-Q water (>18.2 MΩ・cm, Millipore). The color of the air-dried soil samples was determined in the laboratory using Munsell soil color charts (Munsell Color, 1975).

Eleven elements were quantified by a modification of the EPA method 3051a (Environmental Protection Agency; U.S. EPA, 2007). The samples were sieved to 74 μm, 0.5 g of each sample was weighed into a Teflon vessel followed by addition of 9 mL concentrated nitric acid and 3 mL concentrated hydrochloric acid (analytical grade reagents from Sigma-Aldrich). The vessels were placed in the microwave oven (Milestone brand model ETHOS UP) with parameters of 5 min ramp to reach a temperature of 180 °C held for 10 min. After cooling to room temperature, the resulting solution was filtered through Whatman No. 5 filter paper (Whatman International Ltd., Hillsboro, OR) and diluted to 50 mL with Milli-Q water. The resulting solution was additionally filtered with a pore size membrane of 0.05 μm (MFMillipore membrane). All analyses were performed in duplicate. The solutions were analyzed by ICP-OES (inductively coupled plasma optical emission spectrometry, Perkin Elmer, Optima 8300) at the Instituto de Geologia, UNAM. The LoD (limit of detection, μg/L) were: Al (1.2), As (13.6), Ca (9.1), Cd (1.3), Cr (1.1), Cu (1.0), Fe (1.0), Mn (0.9), Ni (1.3), Pb (3.5), Zn (1.5).

The SSEs allow quantification of the labile elements associated with specific solid phases, but also of different major solid constituents (Hayes et al., 2014). The present investigation focused on the Fe phases. Hence, a SSE of three steps was chosen that provides information on the proportion of (1) water-soluble Fe, predominantly as Fe(II) salts, (2) Fe in poorly crystalline phases, and (3) in crystalline phases; these last two items are mainly Fe(III) minerals and in smaller proportion mixedvalence Fe minerals. The following methodology was based on the work published by Drahota et al. (2014). Step 1 (soluble Fe phases): 0.4 g of sample was weighed into 50 mL centrifuge tubes, 40 mL of Milli-Q water was added and placed for 10 h in orbital shaking at 200 rpm (in a Lumistell™ AOP-70 stirrer). The suspensions were centrifuged at 4050 g (relative centrifugal force) for 10 min at 20 °C (Centurion Scientific brand PrO-Research centrifuge) and the supernatants were filtered through 0.05 μm membrane filters; these last two steps were repeated at the end of each extraction step. Step 2 (poorly crystalline Fe phases): 40 mL of 0.2 M NH₄-oxalate solution brought to pH 3.0 by 0.2 M oxalic acid (analytical grade reagents by Sigma-Aldrich) was added to the tube with the centrifuged sample at the end of step 1, stirred for 2 h at room temperature at 200 rpm with the tubes covered to keep them in the dark. Step 3 (crystalline Fe phases): the same solution as in step 2 was added to the tube with centrifuged sample at the end of step 2, stirred for 4 h at 80 °C. The filtered samples from step 2 and 3 were diluted with water (ratio 1:3). 10 mL aliquots of each extraction step were taken and 0.5 mL concentrated nitric acid was added to each tube. The solutions were kept in refrigeration until 24 h prior to analysis. The acidified solutions were analyzed by ICP-OES. The LoD (μg/L) for the oxalate matrix analyses were: Al (24), As (132), Ca (196), Cd (17), Cr (17), Cu (25), Fe (55), Mn (23), Ni (20), Pb (57), Zn (30).

The bulk mineralogy of samples was determined via PXRD (powder X-ray diffraction) using an EMPYREAN diffractometer equipped with a fine focus Cu tube, Ni filter, and PIXCell3D detector operating at 40 mA and 45 kV at the LANGEM, UNAM. Samples were dispersed with an agate pestle and mortar to <75 μm and mounted in back side Al holders. The analyses were carried out on randomly oriented samples by the step scan method using the measurement range (2θ) of 5° to 70° with an integration time of 40 s and step size of 0.003° at room temperature. Phase identification (and semi-quantification by RIR method) was made using HighScore version 4.5 and Data Viewer version 1.8 software from PANalytical and current PDF-2 and ICSD databases.

The XAS analyses were carried out at the Advanced Light Source (ALS) in Lawrence Berkeley National Laboratory, Berkeley, California, with a beamline 10.3.2 X-ray fluorescence microprobe (Marcus et al., 2004). The EXAFS data were collected in transmission and fluorescence modes, the monochromator used was Si(111) crystal from 2.1 keV to 17 keV. Fe K-edge spectra were collected using a 7-element Ge solid state detector. The X-ray beam size on the sample was 15 μm. The μ-XAS analysis for real samples shows variability between spots; since it is a microfocused technique it allows the analysis of one particle of at least 15 μm. Natural samples are not homogeneous and the analyzed spots were limited by the beamtime availability, limiting in turn the complete identification of Fe phases. Therefore, care must be taken in the interpretation of μ-XAS data and its limitations on the representative nature of the whole sample. The spatial distribution of the elements in the sample was mapped by μ-XRF (X-ray fluorescence). We focused the XAS analysis on spots rich in Fe but also simultaneously containing other PTEs, such as Pb and As, to identify Fe phases that were acting as PTE sequestering minerals. Five to ten scans were collected for each spot. All spectra were calibrated in energy, deadtime corrected (only in fluorescence mode); pre-edge background subtracted, and post-edge normalized and averaged using a suite of custom LabVIEW-based programs available at the beamline. The relative proportions of Fe phases in the samples were determined by LCF (linear combination fitting) of the EXAFS spectra using the Athena software (Ravel and Newville, 2005) performed over the range 2-9 k (where k is wavenumber in A-1) and for XANES (X-ray absorption near edge structure) from 2 to 6 k. Fe-bearing standards include the ones in the XAS database already at the ALS beamline 10.3.2, and those read by the research team in several cycles, mainly (Pb-)As-jarosite (Aguilar-Carrillo et al., 2018). The complete list can be consulted in Table 2, the effectiveness of LCF from collected XAS data depends on the completeness of the database used. The weighting factors were forced to be between 0 and 1; the weighting factor is associated with each standard and quantifies the importance of the standard in reproducing the entire spectral series. This factor prevents a standard to have a negative contribution or to be larger than 1. The sum weights were not constrained in order to make sure that the references were suitable. LCF errors are reported as χ2, a statistical indicator of “goodness of fit” calculated from the sum of the squared error divided by the degrees of freedom in the fit. The result are presented as the k3-weighted and normalized Fe K-edge EXAFS spectra, together with the LCF results. In some of the spots chosen the LCF were not satisfactory, most likely because the samples contain other minerals that are not found in the EXAFS database used (Table 2). Consequently, the analysis was restricted to XANES, because this library contains more references. However, only XANES LCF results were reported when there was an improvement. Slight, but significant differences may be found between minerals with different degrees of EPT incorporation or adsorption, such as jarosite with different amounts of As incorporated in their structures or Pb sorbed on them (Aguilar-Carrillo et al., 2018). Another advantage of XAS over XRD is the possibility to identify amorphous or poorly crystalline phases (e.g., schwertmannite and ferrihydrite).

Table 2

a) The formulas of the (Pb-)As-jarosites are (Aguilar-Carrillo et al., 2018): As0.07-jarosite= (H₃O)0.37Na0.07K0.56Fe2.55(SO₄)1.929(AsO₄)0.071(OH,H₂O)₆, As0.14-jarosite= (H₃O)0.35Na0.08K0.57Fe2.77(SO₄)1.86(AsO₄)0.14(OH,H₂O)₆, As0.32-Jarosite= (H₃O)0.40Na0.09K0.51Fe2.90(SO₄)1.68(AsO₄)0.32(OH,H₂O)₆, Pb-As0.14-Jarosite= (H₃O)0.64Na0.18Pb0.18Fe2.59(SO₄)1.86(AsO₄)0.14(OH,H₂O)₆, Pb-As0.23-Jarosite= (H₃O)0.56Na0.28Pb0.16Fe2.62(SO₄)1.77(AsO₄)0.23(OH,H₂O)₆, Pb-As0.40-Jarosite= (H₃O)0.50Na0.28Pb0.22Fe2.49(SO₄)1.60(AsO₄)0.40(OH,H₂O)₆. bXANES (X-ray Absorption Near Edge Structure) refers to the XAS energy area closest to the photoelectron leaving the atom, until about 50 eV above this value, called the X-ray absorption edge. The reference compounds used in the EXAFS LCF were also used in LCF of the Fe XANES spectra.

It is interesting to note that jarosite was only identified in the Ca1 sample where low pH and most probably high sulfate concentrations are consistent with its formation (Stoffregen, 1993). Although jarosite forms in more acidic environments (around pH 2 to pH 2.5; Li et al., 2014), it may be still present at pH 4.1 because of insufficient time for its transformation, considering the initial pH from the spill was very low (pH 2.2 and 2.3; UNAM, 2016). Jarosite has a large capacity to incorporate both cationic and anionic PTEs in its structure, serving as a natural attenuating mineral under very acidic conditions.

Schwertmannite and ferrihydrite were alternately found in all samples. Both are nanominerals and metastable phases, but ferrihydrite can persist for long time periods in a pH range of 2 to 7.5 (Hayes et al., 2014; Li et al., 2014). Schwertmannite precipitates at slightly higher pH than jarosite, from 2.5 to 4.5 (Walter et al., 2003; Li et al., 2014). Being nanoparticulated, both Fe(III) phases show considerable surface adsorption capacities, and schwertmannite additionally can incorporate oxyanions into its structure. Under semi-arid climate conditions, there is a delay of the metastable phase transformation. Hayes et al. (2014) observed ferrihydrite and schwertmannite in mine tailings that have been air-exposed for almost 50 years in IKMHSS (Iron King Mine and Humboldt Smelter Superfund; Dewey-Humboldt, Arizona, USA). On the other hand, Dold and Fontobe (2001) detected schwertmannite in mine tailings in Chile where evaporation exceeds precipitation, probably because in these situations sulfate concentrations are conservative (without leaching effects) and allow stability of this hydroxy-sulfate, provided the pH conditions are also adequate. Another important factor that inhibits their transformation is the presence of high concentrations of different ions or other precipitates incorporated in their structure, for example As(V), or Cr(VI) (Fukushi et al., 2003; Regenspurg and Peiffer, 2005). Schwertmann and Cornell (2000) indicate that crystallization inhibitors (e.g. organics and silicate species) stabilize ferrihydrite and retard its transformation to more stable minerals. Swedlund et al. (2014) confirmed that arsenate ions stabilize ferrihydrite against transformation to goethite in solutions with pH of 6, 8, and 10. Therefore, since the formation of schwertmannite and ferrihydrite occurs from the gradual neutralization of an acidic solution enriched in PTEs, the incorporation of the latter into the structure or on the surface of these minerals is highly probable and, in consequence, may favor their stabilization.

Some reducing conditions are suggested by the presence of Fe(II)- containing minerals such as vivianite and green rusts, which may be expected under submerged conditions by the running river water during the rainy season. Green rust is produced under reducing conditions and from slightly acidic to slightly alkaline environments, often as an intermediate phase in the formation of Fe oxides (Schwertmann and Fechter, 1994; Abdelmoula et al., 1998). Green rust was identified in the samples with either one of two oxyanions, SO₄ ^2- at pH 4.1 to 5.4, and CO₃ ^2- at pH 6.3, which makes sense if contaminated soils were remediated with lime.

Goethite was identified in samples from the most distal away from the spill, but was absent in samples Ca1 and Ca2 closest to the spill, most probably because the high sulfate concentrations and low pH resulting from the spill select for hydroxy-sulfate minerals such as jarosite and schwertmannite. As sulfate is diluted and/or removed from solution, e.g., by lime additions, the conditions for goethite formation may improve, which is the probable mechanism in samples farthest away from the spill. Goethite is one of the most ubiquitous Fe minerals in soils because it is thermodynamically most stable under humid conditions (Hayes et al., 2014). In addition to its well-known surface adsorption capacity, isomorphic substitutions in goethite are known (Gerth, 1990), which makes this Fe(III) phase a crucial player in the mobility control of ions such as those of PTEs under more circumneutral pH conditions.

Ramos-Perez (2017) indicated the following mineralogy associated to pristine soils in the study area affected by the acid spill of Buenavista del Cobre mine: quartz, minerals from the mica group (such as muscovite), intermediate members of the plagioclase group (e.g., labradorite), phyllosilicates (as montmorillonite), and some possible trace minerals, such as birnessite (Mn oxide) and hematite. However, hematite was not detected by XRD or EXAFS in our samples Ca1 to Ca4. Also, neither jarosite nor goethite were identified in pristine soil samples. Therefore, their presence is most likely related to the conditions produced by the influence of the acid spill. No ferrihydrite or schwertmannite are expected were identified by XRD in soil samples in the work of Ramos-Perez (2017), because of their poor crystallinity, but the conditions for schwertmannite precipitation also correspond to the low pH and high sulfate concentrations generated by the acid spill.

Samples Ta1, Ta2 (Taxco, Guerrero), Te (Tecolote, Sonora), SLP1, and SLP2 (San Luis Potosi) showed an acidic pH range from 3.1 to 4.8 (Table 3), and gypsum is present in all sample, which is a sign of acidity and sulfate generation from sulfur oxidation, and proton reactions with calcium carbonates. Gypsum detection by XRD is also confirmed in very high water-soluble Ca concentrations (Table 4). The E.C. values may be imposed mainly by the presence of gypsum (high soluble Ca concentrations, Table 3), with lower aqueous complexation of Ca than in the Buenavista del Cobre mine samples and in minor proportion by metal water-soluble salts (Romero et al., 2008; Murray et al., 2014; Nordstrom et al., 2015; Ramos-Perez, 2017). The highest E.C. values were those from the solid wastes from the SLP1 and SLP2 zinc processing sites, consistent with the detection of gypsum by XRD and the high quantities of soluble elements (Table 3 and Table 4), especially Zn, which is soluble in the same high order of magnitude as Ca.

The low pH and high sulfate concentration in all these samples lead to jarosite and schwertmannite formation, which is the case for the Buenavista del Cobre mine samples. Both minerals were identified by XAS (Table 5, Figure 3), but only crystalline jarosite was also detected by XRD (Table 3). These minerals are common in AMD and mining environments (Acero et al., 2006; French et al., 2012; Hayes et al., 2014). The yellow hue in the samples might be attributed to jarosite (Lynn and Pearson, 2000). Schwertmannite was not evident in the analyzed spots of sample Ta2. It is noteworthy that in all tailing samples the Fe concentrations quantified in step 3 for crystalline phases considerably surpassed the amount for non-crystalline phases (step 2, Table 4). This fact explains why several known crystalline phases were detected by XAS additionally to jarosite, including goethite, hematite, lepidocrocite, magnetite, and sulfated green rust (Table 5). The latter two with semi-reduced Fe, perhaps in intermediate transformation steps from reduced Fe sulfide to ultimate Fe(III)-containing oxide minerals. It is also notable that no ferrihydrite was detected by XAS, except for SLP2, probably because the low pH and high sulfate concentrations favored schwertmannite formation as the non-crystalline Fe phase (French et al., 2012).

Figure 3 Figure 3 k3-weighted Fe K-edge EXAFS spectra (solid lines) and LCF (dark dashed lines) of: a) Ta1, b) Ta2, c) Te, and d) SLP1 and SLP2.

Pyrite was detected in Te sample by XRD, but pyrite spots were avoided in the XAS analysis to better focus on the presence of secondary Fe phases. This justifies that almost 40% of the detected Fe was not an oxide (difference between T.A.D. and T.S.E. Table 4). This is consistent with the primary mineralogy of the El Tecolote tailings for Fe, which includes pyrite, pyrrhotite, and cubanite (Cruz-Hernandez et al., 2018). While FeSO₄ is the only Fe(II) salt determined by EXAFS LCF in sample Ta2, it should be noted that all samples have a small amount of Fe in the soluble phase (Table 4). Actually, Fe-sulfate salt precipitation by capillary transport in the tailings during the dry season is a well-documented phenomenon (Li et al., 2014; Murray et al., 2014).

Goethite, which can be the responsible for the yellow-brown hue in some mine tailing samples (Lynn and Pearson, 2000; Schwertmann and Cornell, 2000), was not observed in SLP1 and SLP2 samples, probably because these are solid wastes from Zn processing. Jarosite precipitation is promoted in the Zn refining process, where Fe and other elements are removed as jarosite by precipitation in high sulfate concentration and acidic media, which avoids the formation of goethite (Dutrizac, 1979; Ramos-Azpeitia et al., 2015). Therefore, jarosite was detected by XRD and as predominant phase by XAS in these samples (close to 80%). Franklinite was identified by XRD in samples SLP1 and SLP2, but was not a standard available in the library to include in the LCF of XAS spectra. It constitutes a common secondary mineral in residues from the Zn process (Burke and Kieft, 1972; Zhang et al., 2011; Ramos- Azpeitia et al., 2015), and its presence may be explained by the dark value in the Munsell color determination of the solid wastes (Table 3; Burke and Kieft, 1972). Scorodite was also spotted in sample SLP1 with a pH of 4.8 (Table 3).

Surprisingly, hematite was found in the sample Ta1 with the lowest pH (3.1, Table 3), but it was previously identified in other mine environments at low pH (Bernstein and Waychunas, 1987; Romero and Gutierrez-Ruiz, 2010) and it was detected in minor quantity as part of the ore in the Taxco zone (Romero et al., 2007). The low abundance of this mineral in the samples might be due to the conditions, because soluble Fe salts, jarosite, and schwertmannite are more prevalent secondary minerals widely found in the unsaturated oxidation zones rich in sulfates and low-water fluxes (Hayes et al., 2014; Li et al., 2014).

In the analysis of Fe phases based on pH conditions of the samples from the acid spill, jarosite and schwertmannite were identified at pH 4.1; while at pH 5.4 ferrihydrite, sulfate green rust and goethite were detected. Schwertmannite, ferrihydrite, scorodite, and carbonate green rust were observed at the highest pH of 6.3. Finally, schwertmannite, goethite, maghemite, sulfate green rust, vivianite, and andradite were found in sample Ca3 (unknown pH).

The Fe mineralogy in the mine tailing samples is represented by the jarosite group, specifically in Pb-As-jarosite (Table 5) in the pH range 3.1 to 4.8. Jarosite compounds have the versatility of incorporating a great number of different chemical elements in their structural sites, such as the Fe octahedral position, the S tetrahedral position, and in the coordination cation site, commonly occupied by K (Li et al., 2014). Incorporation of PTEs into the jarosite structure was observed by identifying (plumbo)-arsenical jarosite in different samples (Table 5). Goethite is the second Fe phase found almost ubiquitously in the waste samples. Other minor phases were spotted in these samples, such as sulfate green rust, schwertmannite, ferrihydrite, etc.

The Fe contribution quantified by SSEs of poorly crystalline phases is larger in the acid spill-affected soil samples than in the mine tailing samples and it is consistent with the XAS results (Table 4 and Table 5). These finding is compatible with a fresher formation of these poorly crystalline Fe phases as the acid is neutralized and since of the sampling time was relatively short after the spill event (Table 1). Time is an important factor in the increase of crystallinity of Fe phases in mine tailing samples. The occurring processes are opposite from those in the acid spill, as the pH is gradually decreased by the generation of AMD, and the low pH values and high sulfate concentrations are decisive factors for the increasing stability of jarosite in comparison to schwertmannite and ferrihydrite (Hayes et al., 2014; Schwertmann and Cornell, 2000). Metastable phases can persist over time in the semi-arid climate conditions (Hayes et al., 2014) of the Sonora region. Another factor for the abundance of poorly crystalline phases in the acid spill samples is the existence of crystallization inhibitors in the soil (including organic compounds, phosphates, and silicate species, which are widespread in natural environments). These inhibitors also stabilize metastable Fe(III) minerals (Schwertmann and Cornell, 2000).