Introduction

Arsenic (As) is a priority environmental pollutant. According to the World Health Organization, the concentration of this element in drinking water should not exceed 10 μg L−1 (Smedley and Kinniburgh 2002). An alternative technology to remove this metalloid from water matrices consists of using minerals. A study conducted in discharged water in a contaminated industrial site showed that 1–10 μm euhedral crystals of pyrite (FeS2) adsorbed and co-precipitated As, and the mineral retained up to 500–4000 mg/kg As (Saunders et al. 2018). In addition, accompanying bench experiments under strict anoxic conditions confirmed that the retention of As(III) (dissolved As2O3) on the mineral surfaces (crushed to several size fractions) increased with decreasing proton activity (5 < pH).

Ubiquitously found in the environment are clays, which are inorganic materials ≤ 2 μm, composed of clay minerals, i.e. specific minerals that mainly occur in the clay size fraction of the soil (e.g. kaolinite, illite, montmorillonite, smectite; Moore and Reynolds 1989). Despite its importance, the adsorption of As in its most toxic form, i.e., As(III), is far from being understood and remains an unresolved topic (Table 1 after Fendorf et al., 2010b). Thus, the present study addresses this complex but unresolved topic. A preceding study reported the adsorption envelopes, and the corresponding constant capacitance model calculations for As(III) adsorption on amorphous Al and Fe (surface areas 660 and 350 m2 g−1), kaolinite (KGa-2, 21.6 m2 g−1), montmorillonite (SAz-1, 68.9 m2 g−1) and illite (IMt-2, 22.6 m2 g−1) showed maximum values of 17 μmol g−1 (7 ≤ pH ≤ 9), ca. 40 μmol g−1 (2 ≤ pH ≤ 11), 0.27 μmol g−1 (8 ≤ pH ≤ 10), 0.2 μmol g−1 (7.5 ≤ pH ≤ 9.5) and 0.25 μmol g−1 (7.5 ≤ pH ≤ 9.5; Goldberg 2002). That study also fitted data to the constant capacitance model, and the adsorption of As on am-Fe oxide and illite quantitatively fit. However, the fitting of the model to the adsorption on am-Al oxide and kaolinite was underpredicted and overpredicted compared to the experimental values at 7 ≤ pH ≤ 9, respectively. Finally, in that study, the adsorption of As was attributed to inner-sphere surface complexation, while fitting of the model was not attempted for data corresponding to montmorillonite.

Far-from-equilibrium experiments to assess the sorption of As(III) on clay mineral surfaces [kaolinite (KGa-1), montmorillonite (SWy-1) and nontronite (iron-rich clay mineral, NAu-1 and NAu-2; Ghorbanzadeh et al. 2015)] showed that the sorption kinetics were described by pseudo-second-order equations. Based on the high values of the r2 coefficients, the adsorption behaviour was well described by Freundlich isotherms for all clays except kaolinite. The authors of that study concluded that nontronite NAu-1 was the most effective clay for removing As(III) from solutions (rate constant ca. 0.056 g mg−1 min−1; Ghorbanzadeh et al. 2015).

High-resolution scanning electron microscopy and energy-dispersive spectrometry (HRSEM EDS). Clay minerals are composed of assemblages of tetrahedral and octahedral sheets, whereby the tetrahedral Si–O and O–O bond distances are approximately 0.162 nm and 0.264 nm, respectively, and the Al octahedral O–O and OH–OH distances are 0.267 nm and 0.294 nm, respectively (Sparks 2003, p. 51). Here, the authors studied the atomic environment of As(III) on clay mineral surfaces with 1 nm precision using HRSEM–EDS. This experimental approach guaranteed results with high accuracy, especially if the mineral As concentrations were low to very low. The outcome of this study well complemented reports on the atomic environment of As on mineral surfaces determined by wet chemistry, computational modelling and X-ray absorption near edge spectroscopy [XANES; high energy resolution (< 0.5 eV) and high spatial resolution (100 μm); (Bostick and Fendorf 2003; Sverjensky and Fukushi 2006; Fendorf et al. 2010a)]. Unveiling the atomic environments of mineral As and bond energies that control the release of As into the solution with high precision will help predict the fate of this metalloid in aquifers, particularly in those containing low to very low amounts of organic carbon. Then, the cycling of the elements can be regulated by the chemistry of the minerals (Fendorf et al. 2010b).

The HRSEM–EDS analysis was complemented with hydride generation–cryotrapping–atomic absorption spectrometry (HG-CT-AAS) analysis. Taken together, these techniques enabled us to achieve low detection limit analyses of both supernatant solutions and surfaces. The work was conducted at the referred institutions on June 2019.

Materials and methods

Expandable clay minerals

Hectorite {SHCa-1; (Na)0.80[Si7.90Al0.1][Mg5.30Li0.70]O20(OH)4; Hector, California, USA; 50% hectorite, 43% calcite, 3% quartz, 3% dolomite (Chipera and Bish 2001), 0.32% Fe2O3 with no detectable Fe2+ (Mermut and Cano 2001; Środoń and McCarty 2008)} and nontronite {NAu-1; (Na1.05[Si6.98Al0.95Fe0.07][Al0.36Fe3.61Mg0.04]O20(OH)4; Uley Mine, Australia; nontronite and trace amounts of feldspar and quartz (Keeling et al. 2000)} were purchased from the repository of the clay minerals society and used as received.

Reaction between As(III) and expandable clay minerals (hectorite and nontronite)

The source of As(III) was sodium (meta)arsenite (Merck CAS Number 7784–46-5; ≤ 90%). A solution of 10–4 M sodium arsenite in water (pH 0 = 7.2) was prepared. Then, 10 mL of the resulting solution was added to 1 g of clay mineral and allowed to react at room temperature (25 ± 1 °C) for 1 week. Solids were recovered, placed in Teflon weighing plates, covered with Parafilm, allowed to dry under ambient conditions for a week, and finally prepared for microscopic analysis (Sect. 2.4).

HG-CT-AAS

Arsenic species in supernatant solutions were analysed by HG-CT-AAS, which enables the quantification of seven arsenic species, including arsenite, arsenate, methylarsenite, methylarsenate, dimethylarsenite, dimethylarsenate and trimethylarsine oxide (Supplementary material after Del Razo et al. 2001; Devesa et al. 2004; Hernández-Zavala et al. 2008).

HRSEM–EDS

The clay mineral samples were placed in an Al sample holder, fixed with carbon tape, and analysed by HRSEM. We used an environmental scanning electron microscope Philips XL30 ESEM (amplification range 10–200,000 × ; resolution 3.5 nm) and a solid-state energy-dispersive spectrometer equipped with an X-ray Si (Li) ultrathin window to detect elements from Be and onwards.

Table 1 Retention maxima from adsorption isotherms at fixed pH and adsorption envelopes for As(III) on various solidsa
Full size table

Results and discussion

As in supernatant solutions

As evidenced by HG-CT-AAS, As(III) in the solution showed no or below-detection-limit transformation to As(V), while its concentration decreased upon exposure to clay minerals (Fig. 1).

Fig. 1
figure 1

Retention of As(III) in the clay minerals

Full size image

Characteristic equations

As determined by the EDS analysis, the clay minerals were primarily composed of Si, O, Fe, Al, As, S (%atom; Table 2) and other elements (footnotes in Table 2).

Table 2 EDS analyses of the clay minerals reacted with As(III)
Full size table

The elemental associations were determined by calculating the ratios As/S, Fe/S, As/O, Fe/O, Si/O, Al/O, As/Fe, and Al/Fe; then, the ratios were plotted for comparison (Table 3; Figs. 2, 3, 4 and 5). Data within precision were considered, and outliers were discarded.

Table 3 Elemental ratios plotted in Figs. 2 and 3
Full size table
Fig. 2
figure 2

Relations between %As and %Fe in nontronite

Full size image
Fig. 3
figure 3

Relations between As/S and Fe/S in nontronite

Full size image
Fig. 4
figure 4

Relations between As/O and Fe/O in nontronite

Full size image
Fig. 5
figure 5

Relations between As/Fe and Al/Fe in nontronite

Full size image

Unlike for hectorite, the surface analysis of nontronite confirmed the As accumulation, which was related to compositional and structural units. %As increased with %Fe (Fig. 2):

$$ \left\{ {\% {\text{As}}} \right\} \, = \, 0.099 \, \left\{ {\% {\text{Fe}}} \right\}^{2} {-} \, 0.09\left\{ {\% {\text{Fe}}} \right\} + \, 0.65,r^{2} = \, 0.93 $$
(1)

Additionally, the data trends of As/O, As/S, Fe/O and Fe/S were related, as described by Eqs. 2 and 3 (Figs. 3, 4):

$$ \left\{ {{\text{As}}/S} \right\} \, = \, 0.17\left\{ {{\text{Fe}}/S} \right\}^{2} - 0.11\left\{ {{\text{Fe}}/S} \right\} \, + \, 0.43,r^{2} = \, 0.95 $$
(2)
$$ \left\{ {{\text{As}}/O} \right\} \, = \, 0.35\left\{ {{\text{Fe}}/O} \right\}^{2} - 0.07\left\{ {{\text{Fe}}/O} \right\} \, + \, 0.01,r^{2} = \, 0.93 $$
(3)

Notably, the fitting coefficients of Eqs. 1, 2 and 3 were similar; however, a direct comparison of the polynomial coefficients showed that Fe/O served as a good predictor for the As accumulation, surpassing %Fe and Fe/S. Additionally, a positive, shallower relation was found to hold true between the data trends of As/Fe and Al/Fe (Eq. 4 after Fig. 5). Finally, no relation held true between the Si content and the As accumulation (Tables 2 and 3).

$$ \begin{aligned}\left\{ {{\text{As}}/{\text{Fe}}} \right\} & = \, - 0.0017\left\{ {{\text{Al}}/{\text{Fe}}} \right\}^{2} + 0.0502\left\{ {{\text{Al}}/{\text{Fe}}} \right\} \\&\quad + \, 0.0498,r^{2} = \, 0.83 \end{aligned} $$
(4)

Surface Fe

Arsenic accumulated in clay minerals; nevertheless, the data show that the values of %As for hectorite (≤ 4% As) could surpass those found in nontronite (≤ 2.7% As; Table 2). Nontronite is an Fe-rich clay mineral [30% (or more) of Fe as Fe2O3; Cornell and Schwertmann 2006]; therefore, the total content of structural Fe in the clay mineral per se did not serve as a predictor for the As accumulation. The findings herein expanded on a report showing that Fe in clay minerals guaranteed the preferential retention of As(III) (Ghorbanzadeh et al. 2015).

Surface O

The As accumulation in both clay minerals (hectorite and nontronite) was not related to Si (as Si/O) regardless of dissolved O2 (Figs. S1 and S2). With that said, As/O vs Si/O graphs showed that the As accumulation (as As/O) occurred at different intervals of Si/O: 0.15 ≤ Si/O ≤ 0.4 (nontronite), 0.01 ≤ Si/O ≤ 0.5 (hectorite, ox), and 0.01 ≤ Si/O ≤ 0.7 (hectorite, an). Consequently, the As accumulation in hectorite occurred in the presence of a higher content of surface O, arguably due to the binding and formation of As-O structural units. In contrast, the As accumulation in nontronite was not limited by its association with surface O.

Factors controlling the accumulation of As in nontronite

Equation 4 shows a positive, shallow relation between As/Fe and Al/Fe in nontronite, which strongly suggested that As was substituted for Fe, similar to Al (Cornell and Schwertmann 2006). Equations 2 and 3 show the significant relations between As/O, Fe/O, As/S and Fe/S in nontronite, which implies that As binds surface O in a similar manner to Fe binding surface O. Finally, a direct comparison of the polynomial coefficients (Eqs. 2 and 3) shows that As was associated with surface O and S, with the former being preferred by a factor of 2:1.

Surface (As)(O)

The association of As(III) with Fe-(hydr)oxides occurs through multiple inner-sphere complexes (Ona-Nguema et al. 2005), outer-sphere complexes and hydrogen bonding interactions, all of which only conform to a weak colligative force. Therefore, in time, a facile detachment of surface As(III) was sought to proceed in hectorite and nontronite.

Surface (As)(S)

The results show that the preferential association of As with S were consistent with evidence that in reducing environments, the solubility of As(III) is limited by sulphide precipitates (Moore et al. 1988). In metal sulphides, adsorbed As(III) becomes incorporated into the mineral lattice (Bostick and Fendorf 2003). This mechanism was responsible for the nonreversible association of As(III) with nontronite surfaces.

Surface (As)(O) and (As)(S)

Reported studies show that minerals determine the mechanism of As adsorption. For example, on pyrite (FeS2), As strongly binds to form surfaces, inner-sphere complexes with Fe-S (5 ≤ pH) or ligand exchange with ≡FeOH surface sites (higher pH values). In contrast, on Fe (oxy)hydroxides, anionic As participates in ligand exchange with surface hydroxyl groups (Bonnisel-Gissinger et al. 1998; Bostick and Fendorf 2003; Saunders et al. 2018).

Surface As-ferrihydrite

Mössbauer spectroscopic evidence shows that if anoxic conditions prevailed, the structural (octahedral) Fe of nontronite NAu-1 caused very limited oxidation of adsorbed As(III) (Ilgen et al. 2017). Meanwhile, at circumneutral pH, the dominant As species are H2AsO4− and HAsO2− (pKa = 6.9) and H3AsO3− (pKa = 9.2; Fendorf et al. 2010a, b). Thus, in nontronite dispersions, As remained as As(III).

The Fe/As ratio indicates the phases containing As (e.g. Frau et al. 2005; Paktunc et al. 2008). Thus, the obtained As/Fe ratios (Table 3) were used to estimate the corresponding Fe/As ratios, which showed values of 3.3 ≤ Fe/As ≤ 10. The obtained Fe/As values were compared to those reported elsewhere (Frau et al. 2005). A study on mine waste materials and stream-bed sediments from the Baccu Locci mine area in Sardinia, Italy, using transmission electron microscopy (TEM) and X-ray photoelectron spectroscopy (XPS) showed Fe/As of 0.9–26 (Frau et al. 2005). There, the registered ratios Fe/As > 1 were associated with the formation of ferrisymplesite [Fe3(AsO4)2(OH)3·5H2O], crystalline scorodite [Fe(III)AsO4·2H2O] and arsenical ferrihydrite. Another unrelated study reported the characterization of synthetic scorodite and arsenical ferrihydrite using TEM, X-ray diffraction (XRD) and XANES techniques (Paktunc et al. 2008). Therein, in aqueous dispersions with 1 ≤ pH ≤ 4.5 and 1 < Fe/As, the predominant mineral formed was ferric arsenate, which eventually transformed to scorodite. However, in dispersions with pH = 4.5 and 1 ≤ Fe/As ≤ 4, the minerals consisted of admixtures of ferric arsenate, arsenical ferrihydrite and As-rich hydrous ferric oxide. If 5 ≤ Fe/As, the growth of ferrihydrite was favourable, and the two-line ferrihydrite gradually transformed to six-line ferrihydrite. However, ferrihydrite growth was inhibited by As adsorption (Paktunc et al. 2008). Thus, at circumneutral pH, As(III) adsorbed onto nontronite surfaces was concluded to react with surface ferrihydrite to form arsenical ferrihydrite.

Conclusion

This contribution expands on current knowledge on the interaction between As(III) and expandable clay minerals. The higher content of structural Fe in nontronite vs. hectorite favours the retention of As(III). Arguably, adsorbed As(III) on nontronite surfaces (1) transforms to arsenical ferrihydrite or (2) bonds with naturally occurring S; both processes facilitate its nonreversible surface bonding. In addition, adsorbed As(III) on nontronite and hectorite surfaces reversibly bonds to surface O.