Arsenic-Bearing Serpentine-Group Minerals: Mineral Synthesis with Insights for the Arsenic Cycle

Abstract

When present at elevated levels in drinking water, arsenic is toxic, and magnesian clays are gaining recognition as a source of elevated arsenic in groundwater. In the crust and upper mantle of Earth, arsenic incorporation into clay minerals is influenced by geochemical conditions associated with hydrothermal fluids and metamorphic processes (e.g. serpentinization), meaning that As is a useful tracer of fluid-flow in the deep Earth. To improve understanding of arsenic speciation in groundwater, sediments, soils, and hydrothermal-metamorphic systems, the present study examined arsenic incorporation into magnesian clays by synthesis of serpentine minerals (200^oC, 10 d) with varied concentrations of Si, Al, As⁵⁺, and As³⁺. The synthesis experiments produced two distinct crystal types, tubular and platy serpentines, each with 10–15% randomly interstratified talc layers. X-ray absorption spectroscopy indicated that As⁵⁺ and As³⁺ occurred in the tetrahedral sheet. Single-crystal analysis revealed that tubular crystals contained up to 1 wt.% arsenic [Mg_2.8(Si_1.8As_0.2)O₅(OH)₄] (mean 0.2 wt.% As). The mean composition of platy, high-Al crystals is (Mg_1.8Al_0.7)(Si_2.0)O₅(OH)₄, and that of platy, medium-Al crystals with As³⁺ is (Mg_2.07Al_0.52) (Si_1.97As³⁺_0.03)O₅(OH)₄. Charge, geometry, and radius of tetrahedral AsO₄^3– oxyanions are similar to tetrahedral SiO₄^4–, and this facilitates fixation of As⁵⁺ into the tetrahedral sheet of clay minerals. The geometry and size of the larger As³⁺ in tetrahedral sites (as a pyramidal AsO₃^3– oxyanion) may limit incorporation relative to As⁵⁺. Arsenic-bearing Mg clays crystallize in alkaline environments where AsO₄^3– or AsO₃^3– are the dominant As species and where high pH accompanies crystallization of serpentine, talc, chlorite, or Mg-smectite. The presence of tetrahedral As in these clays raises the possibility of tetrahedral As in other Mg clays (e.g. sepiolite or kerolite) as well.

Introduction

The behavior of arsenic in rocks, sediments, and soils is controlled strongly by its speciation in minerals and amorphous solids, and this in turn has important implications for understanding arsenic fate and transport in systems ranging from low temperature (e.g. aquifers and soils) to medium and high temperature (e.g. hydrothermal, igneous, and metamorphic systems). In groundwater, arsenic is a toxic element that adversely affects the health of at least tens of millions of people globally (Bhattacharya et al. 2007; Navas-Acien et al. 2010; Shi et al. 2015), so improved understanding of arsenic speciation is crucial to predicting uptake and release in environmental systems. The presence of As in magnesian clays is a source of elevated As in aquifers (Guillot & Charlet 2007; Ryan et al. 2011; Masuda et al. 2012), and has also proven useful for tracing fluid flow associated with subduction and metamorphism (Hattori et al. 2005; Deschamps et al. 2010). To date, however, no study has used mineral synthesis to examine systematically factors controlling the crystallochemical occurrence of As in the mineralogical structure of magnesian clays, thus providing the impetus for this study.

The oxidation states of arsenic range from As^3– to As⁵⁺, where the low oxidation states (e.g. As^3–, As⁰) tend to occur in sulfides and arsenides and the higher oxidations states (e.g. As³⁺, As⁵⁺) in oxides, hydroxides, and silicates. The occurrence of As in sulfides (e.g. pyrite and arsenopyrite), arsenides (e.g. skutterudite, löllingite), and in iron (oxyhydr)oxides such as goethite and magnetite (Smedley & Kinniburgh 2002; Hattori et al. 2005) is well known. The occurrence of As in silicates is less well known, but X-ray absorption spectroscopy (XAS) indicates that As occurs in tetrahedral sites (Hattori et al. 2005; Charnock et al. 2007; Niu, 2011; Masuda et al. 2012; Masuda, 2018), where As⁵⁺ or As³⁺ may substitute for Si⁴⁺. Paired substitution of As⁵⁺ and As³⁺ may satisfy charge balance, and so too may paired substitution of As⁵⁺ and Al³⁺. All known occurrences of As-bearing phyllosilicates are associated with alkaline fluids and serpentinization (Hattori et al. 2005; Guillot & Charlet 2007; Deschamps et al. 2010; Ryan et al. 2011) or hydrothermal mineral precipitation (Pascua et al. 2005); from these origins, As-bearing phyllosilicates may be eroded and deposited in sediments, where they contribute to high arsenic in groundwater (Guillot & Charlet 2007; Ryan et al. 2011; Masuda et al. 2012).

Documented arsenic occurrences in Mg phyllosilicates include antigorite, chlorite, smectite, and talc. Hattori et al. (2005) used X-ray absorption near-edge spectroscopy (XANES) and extended X-ray absorption fine-structure spectroscopy (XAFS) to document tetrahedral As⁵⁺ in antigorites that contain 6 to 275 mg kg^–1 As in Himalayan serpentinites. Serpentinites from other orogenic belts have been recognized with >10 mg kg^–1 As, and many with >100 mg kg^–1 As (Deschamps et al. 2010; Ryan et al. 2011). Guillot & Charlet (2007) argued that antigorites are the primary source of As in the aquifer of the Bengal Fan. Ryan et al. (2011) and Niu (2011) determined (XAS, electron microprobe analysis [EMPA], and sequential extraction) that tetrahedral As³⁺ occurs in the silicate structure of antigorite in serpentinites from the northern Appalachians of the USA – these antigorites contain 20 to 450 mg kg^–1 As, and weathering of antigorite appears to be a source of elevated As in groundwater. Masuda et al. (2012) used micro-X-ray-fluorescence (μ-XRF) to document As⁵⁺ and As³⁺ in chlorite that is attributed to the high As concentration in groundwater of the Holocene aquifer system in Bangladesh. Pascua et al. (2005) used XAS in conjunction with chemical extraction to document the presence of both As⁵⁺ and As³⁺ in trioctahedral Mg- smectite with 1500 to 4000 mg kg^–1 As (~0.1 As per 4 tetrahedral sites) formed in a hydrothermal setting. Sequential chemical extraction and quantitative mineral analysis indicates that 50 to 140 mg kg^–1 of As occurs in talc formed in the alteration halo surrounding serpentinites (Ryan et al. 2011). Tetrahedral As is not restricted to Mg-phyllosilicates as it has also been detected in tetrahedral sites in andradite garnet (Charnock et al. 2007).

Considering radii and charges of ions in silicate minerals, the occurrence of arsenic in tetrahedral sites in phyllosilicates is predictable; in terms of crystal radii (Shannon, 1976), As⁵⁺ (0.48 Å) is a better fit in the tetrahedral sheet than is Al³⁺ (0.53 Å) (compared to Si⁴⁺ at 0.40 Å). The magnitude of the charge difference between As⁵⁺ and Si⁴⁺ is equal to that of Al³⁺ for Si⁴⁺. In fact, paired substitution of As⁵⁺ and Al³⁺ in tetrahedral sites for two Si⁴⁺ (As⁵⁺ + Al³⁺ ➔ 2 Si⁴⁺) was suggested to explain the occurrence of tetrahedral As⁵⁺ in Himalayan antigorites (Hattori et al. 2005). The As³⁺ ion forms a pyramidal geometry in AsO₃ ^3– that may limit compatibility in the tetrahedral sheet, but its crystal radius (0.54 Å) in tetrahedral coordination is comparable to that of Al³⁺ and is smaller than Fe³⁺ (0.63 Å), so the occurrence of tetrahedral As³⁺ in serpentine (antigorite; Niu, 2011) and potentially other phyllosilicates also should be explored.

One way to examine the potential for a mineral to incorporate trace elements is by mineral synthesis under controlled conditions. Such an approach can limit the number of variables and unknowns common to naturally occurring minerals, which is important when assessing trace-element speciation (Petit & Decarreau 1990; Iriarte et al. 2005; Bentabol et al. 2006, 2007, 2009). A prior synthesis study showed the potential of serpentine to contain up to 3000 mg kg^–1 As (0.02 mol As per 2 tetrahedral sites) (Lafay et al. 2016) (tetrahedral vs octahedral occupancy was not assessed). Considering synthesis of As-bearing Mg phyllosilicates, the ability to control for the absence of Fe simplifies the system, eliminating the possibility of As speciation into Fe-(oxyhydr)oxides. Synthesis also allows concentrations of Al³⁺, As⁵⁺, and As³⁺ to be adjusted, enabling analysis of controls on As incorporation. The method selected for synthesis of As-bearing serpentine group minerals uses salts of Mg, Al, Si, and As under alkaline conditions at 200^oC and 10 d.

The purpose of this study was to carry out synthesis reactions to examine the potential for, and controls on, the incorporation of As⁵⁺ and As³⁺ into tetrahedral sites in serpentine group minerals. Data on mineral structures and composition were used to understand the incorporation of As into serpentine and the partitioning of As⁵⁺, As³⁺, and Al³⁺ relative to tetrahedral Si in structural sites – tetrahedral and octahedral – in the layered trioctahedral silicate. A testable hypothesis was that synthesized serpentines will incorporate As at levels at least comparable to – or greater than – natural serpentines. In addition, when charge, radius, and geometry of oxyanion (e.g. tetrahedral As⁵⁺O₄ vs octahedral As³⁺O₃) are considered, a preference was expected for As⁵⁺ over As³⁺ in tetrahedral sites in serpentine. Paired substitutions that maintain charge balance will foster greater incorporation of As than cases where vacancies are the only mechanism that could compensate for charge imbalance.

Materials and Methods

Synthesis of Serpentine-group Minerals

Stock solutions of salts containing Mg, Si, Al, As⁵⁺, and As³⁺ were prepared following the approach developed by Opiso et al. (2010) with some changes made to suit the objectives of this study and the synthesis equipment available. No Fe was used in the experiments. Sources of elements were as follows: Mg was from Mg(NO₃)₂ ^.6H₂O; Si was from Na₂SiO₃ ^.5H₂O; Al was from Al(NO₃)₃ ^.9H₂O; As⁵⁺ was from Na₂HAsO₄ ^.7H₂O; and As³⁺ was from NaAsO₂. With respect to the elements of interest, the molarities of stock solutions were 32.0 mM Mg and 22.5 mM for Si, Al, As⁵⁺, and As³⁺. The stock solutions were mixed to produce 3:2 molar ratios of Mg:Si for pure Mg-Si serpentines, and Si was adjusted downward in concert with quantities of Al, As⁵⁺, and As³⁺ added to the synthesis solution (Table 1). The total volume of mixed reactants (~12 mL) was mixed with 13.3 mL of a NaHCO₃-NaOH buffer (0.05 M NaHCO₃ and 0.3 M NaOH at pH = 12) to produce a 25 mL solution with initial pH of 10.0 ± 0.1. This solution was reacted at 200^oC for 10 d in 50 mL Teflon-lined Parr 4744 steel reactors at a pressure of 15 atm (Parr Instrument Co., Moline, Illinois, USA).

Table 1 Quantities of stock solutions of reagents (mL) mixed and added to Parr 4744 reactors for synthesis reactions. In each case, 13.325 mL of a pH 12 NaOH-NaHCO₃ buffer was added to produce a ~25 mL final volume at pH = 10. The right column shows the expected formula if resulting products are proportional to initial solution. See text for stock solution compositions

In order to avoid contamination, reactors were rinsed with milliQ water prior to synthesis then soaked overnight in the buffer solution. Reactors were then rinsed again with milliQ and dried, then filled with the synthesis reagents. In all, seven different compositions of serpentine were targeted (Table 1), from pure Mg-Si serpentine to six different stoichiometric ratios of Mg, Si, and variably Al, As⁵⁺, or As³⁺. Suppliers of reagents were as follows: Mg(NO₃)₂ ^.6H₂O, Al(NO₃)₃ ^.9H₂O, and NaOH (Panreac Quimica S.A., Barcelona, Spain); Na₂SiO₃ ^.5H₂O_, Na₂HAsO₄ ^.7H₂O, and NaAsO₂ (Sigma Aldrich Co., St. Louis, Missouri, USA); and NaHCO₃ (Merck, Darmstadt, Germany). Reactors were cooled quickly by submersing in cold water at the end of each experiment to ensure that the reactor was cool enough to open without spraying steam and synthesized product; as such, solids were separated as soon as possible from solutions by centrifugation.

The 10 d reaction produced white powders ranging in mass from 5 to 19 mg. In order to increase the amount of powder available for analysis, three replicate trials for each stoichiometric objective were carried out and powders were homogenized. To remove adsorbed Al, As, Mg, or Si, the synthesized powders were suspended in a 20 mL solution of 0.5 M CaCl₂ (from CaCl₂ ^.2H₂O, Panreac Quimica S.A., Barcelona, Spain), which was ultrasonicated to disaggregate the powder. This suspension was allowed to equilibrate overnight, and solids were obtained by centrifuging at 5000 rpm (~1600 × g) for 15 min. This process was repeated again with 0.05 and 0.005 M solutions of CaCl₂, pouring off supernatant in each case. The resulting Ca-saturated clays were then subsequently washed three times with deionized water to remove any Ca not electrostatically attracted to mineral particles. Ca-washed powders from the three replicate trials were dried at 60^oC, mixed, and ground gently to homogenize for instrumental analysis.

X-ray Diffraction (XRD)

X-ray diffraction analysis was performed on synthesized powders using a PANalytical X’Pert Pro system operating at 45 kV and 40 mA with an X’Celerator detector, CuKα radiation, a Ni filter, and 0.25^o divergence slit. Sample mounts were prepared by pipetting concentrated slurries of powders (in ethyl alcohol) onto glass slides and allowing the slurry to evaporate – these mounts were scanned continuously from 5 to 70^o2θ (0.4^o2θ/min) in ambient air-dried and ethylene glycol-solvated conditions. NEWMOD 2© (Release 1.3; https://newmod-for-clays.com/) was employed to model layer types using default parameters. The serpentine layer had a 7.3 Å d spacing and contained no Fe. The talc-like layer type was modeled with a 9.3 Å d spacing and no Fe, no interlayer cation and no interlayer water.

Transmission Electron Microscopy (TEM)

Transmission electron microscopy analyses were performed at the Centro de Instrumentación Científica (CIC) at the Universidad de Granada (Spain). All seven samples were first analyzed using a Philips CM-20 electron microscope (Almelo, The Netherlands) fitted with an ultrathin window and solid state Si(Li) detector for energy dispersive X-ray analysis (EDX; Mahwah, New Jersey, USA). The synthesized powders were suspended in pure ethyl alcohol and mounted on Cu grids, and analyses were performed at 200 kV with a 200 nm spot size. The compositions of single crystals were determined by analytical electron microscopy (TEM-AEM), where atomic proportions of Mg, Al, Si, and O were calculated from peak intensities (Kα lines) and converted into atomic concentrations using natural mineral standards (e.g. biotite, muscovite, olivine) with calibration carried out by the method of Cliff & Lorimer (1975). Arsenic was quantified using the Kα peak at 10.532 keV using theoretical basis for intensity, and the signal from the As 1.282 keV Lα peak was subtracted from the 1.253 Kα Mg peak to provide precise Mg intensity. The detection limit for As in TEM-AEM analyses of these synthesized serpentines was 0.1 wt.% or 1000 mg kg^–1 (this is the concentration above which the As peak is distinct and ≥3 standard deviations above the mean of the background).

Three selected samples (Serp 1, Serp 4, and Serp 5) were analyzed using an FEI Titan G2 TEM (FEI, Hillsboro, Oregon, USA), with an XFEG emission gun, a spherical aberration corrector, and a high-angle annular dark-field (HAADF) detector. The synthesized powders were mounted in epoxy and cut to emphasize as many possible crystallographic orientations as possible. Analyses were performed at 300 kV with a resolution of ~0.2 nm in the high-resolution (HRTEM) mode and ~2 nm in scanning TEM (STEM) mode (where the STEM method use an HAADF detector. Compositional analysis was performed with a Super X micro X-ray analyzer equipped with four detectors by X-ray energy dispersion (EDX). Elements were quantified by the same approach described above for the CM-20.

Inductively Coupled Plasma–Mass Spectrometry (ICPMS)

The chemical composition of synthesized powders was determined by ICPMS as follows: 10 mg of synthesized powder was dissolved in 10 mL of aqua regia (US EPA method 200.2, consisting of 2.9 mL of 50% HNO₃ and 7.1 mL of 20% HCl, trace metal grade [Sigma Aldrich Co., St. Louis, Missouri, USA], at 95^oC for 0.5 h), then diluted by pipetting 1.0 mL of this solution into 9.0 mL of 5% trace metal grade HNO₃. Then, 4.9 mL of this solution was mixed with 0.1 mL of an internal standard solution containing Sc and Ga. This 5.0 mL solution was analyzed using a Thermo iCAP Qc (Waltham, Massachusetts, USA) in KED mode at Middlebury College, calibrated using standards across the range of concentrations encountered, further calibrated by running NIST 1643f (National Institute of Standards and Testing, Gaithersburg, Maryland, USA) as an unknown, and drift corrected by monitoring intensities of Sc and Ga. Si could not be measured precisely due to problems with Si saturation, so the data reported pertain only to Mg, Al, and As, and are used as a means of assessing concentrations of those elements compared to TEM-AEM.

X-ray Absorption Near Edge Spectroscopy (XANES) and X-ray Absorption Fine Structure (XAFS)

In order to assess As oxidation state and coordination, XANES and XAFS were used to study synthesized powder Serp 5, the one that contained both As⁵⁺ and As³⁺ in the initial solution. The sample was placed on Kapton tape for analysis. XANES and XAFS spectra were collected at the As K-edge (11867 eV) at the Inner Shell Spectroscopy beamline (8-ID) at the National Synchrotron Light Source II at Brookhaven National Laboratory. The electron storage ring was operating at 3 GeV with a beam current of 400 mA. Spectra were acquired in fluorescence mode at room temperature using a cryogenically cooled Si (111) double crystal monochromator and a passivated, implanted, planar silicon (PIPS) detector. A germanium Z1 filter was used to decrease flux, minimizing sample degradation. Spectra were calibrated to the Au L-edge (11919 eV) using reference foils measured with each sample (Müller et al. 2010). Multiple scans were collected and averaged to improve the signal to noise ratio, and no oxidation or reduction of samples occurred during analysis. Data reduction and analysis of XANES spectra were performed using the ATHENA program (Ravel & Newville 2005). To determine relative percentages of As⁵⁺ and As³⁺ within the sample, linear combination fitting (LCF) was performed using the ATHENA program (Arai et al. 2001; Ravel & Newville 2005). The standards were 100 mMol solutions of NaAsO₂ and Na₂HAsO₄ ^.7H₂O (Sigma Aldrich Co., St. Louis, Missouri, USA). Analysis of XAFS spectra was performed using the ATHENA (Newville 2001) and Artemis (Ravel & Newville 2005) programs. Uncertainties of fit reported were calculated by ATHENA and Artemis as R-Factor = Σ(data – fit)² / Σ(data²) and chi-squared.

Fourier-transform Infrared Analysis (FTIR)

The FTIR spectra were recorded at the Instituto Andaluz de Ciencias de la Tierra (Granada, Spain) on a Perkin-Elmer Spectrum One FTIR spectrometer (Waltham, Massachusetts, USA) in absorbance mode (4000 to 400 cm^–1 range) with a resolution of 4 cm^–1. Samples were prepared as KBr pressed pellets by diluting 1 mg of sample in 150 mg of dried KBr (Merck, Darmstadt, Germany). The pellets were heated overnight at 120^oC before analysis.

Results

XRD

Each of the seven synthesis experiments produced crystalline phases with 00l peaks that are consistent with the presence of serpentine group minerals as the dominant solid phase in the powders (Fig. 1). None of the powders exhibited change in XRD patterns with ethylene glycol solvation. The synthesis involving only Mg and Si as cations (Serp 1) resulted in 00l peaks at 7.5 to 7.6 Å (001), 3.65 Å (002), and 1.84 Å (004). The 060 peak occurs at 1.53 Å and the 020 peak is at 4.52–4.54 Å; the 060 indicates a trioctahedral b-axis spacing of 9.12 Å whereas the 020 peak indicates a 9.04 to 9.08 Å b dimension consistent with trioctahedral but also smaller than typical serpentine group minerals. The strong peak at 2.57 Å is probably the 202 peak and the weaker peak between 1.73 and 1.70 Å is likely to be the 206 peak (Brindley & Brown 1980). The low-angle shoulder on the 001 peak and the non-integer relationship of 001 to higher-order 00l peaks (mainly the higher-than-normal 7.5 to 7.6 Å 001 peak) indicated randomly interstratified 2:1 layers (Moore & Reynolds 1997) with structure and composition akin to talc or kerolite. Modeling XRD patterns as an R0 interstratified mineral with 85% serpentine layers and 15% talc layers reproduced the low-angle asymmetry and ~7.5 Å position of the 001 peak and the high-angle asymmetry of the 002 peak (Fig. 1B). One place where the model was not a perfect match was the 0.07 Å difference in position of the 002 peak. The spacing of the broad serpentine 001 peak (~7.5 Å) was between nominal (end-member) 001 spacings of serpentine (7.3 Å) and talc (9.3 Å). The high-angle asymmetry of the 002 peak indicated that it was a composite peak between nominal serpentine 002 (3.65 Å) and talc 003 (3.1 Å) peaks. This type of peak shifting and asymmetry was predicted by the interstratification model developed by Méring (1949) (see also Moore & Reynolds 1997).

Fig. 1 (a) XRD data of random mounts of seven powders produced by synthesis. Data shown are for the air-dried state, digits above peaks indicate spacing in Å, and ions in addition to Mg and Si are listed with sample IDs on the right; (b) XRD data from specimen Serp 2 and a model of R0 interstratified serpentine(0.85)/talc. The model does not show a 02,11 peak because NEWMOD models 00l peaks (Newmod 2). The 02,11 in the experimental data is a consequence of the tubular crystal form.

Five of the seven powders – Serp 1, Serp 2, Serp 3, Serp 5, and Serp 7 – yielded XRD patterns that were virtually identical to each other, including the pure Mg-Si end-member (i.e. Serp 1). Two of the specimens, however, displayed some notable variations relative to the other five powders – these were Serp 4 (containing Al in addition to Mg and Si) and Serp 6 (containing both Al³⁺ and As³⁺ in addition to Mg and Si). Serp 4 and Serp 6 were characterized by sharper 00l peaks than the others, as well as an 001 peak that occurred at a smaller spacing than the others: 7.25 Å (Serp 4) and 7.22 Å (Serp 6) vs ~7.5 Å for the other 5 powders. Serp 4 had a comparatively sharp 020 peak at 4.61 Å and the 020 peak from Serp 6 occurred at 4.59 Å, peaks consistent with a larger trioctahedral b-axis dimension of ~9.2 Å (compared to the other five synthesized powders). The position of the 001 peak was sensitive to Al content (Serna et al. 1979) – in the case of pure Mg-Si (Al-free) serpentines, the 001 occurs at 7.32 Å, but with increasing Al, the 001 spacing decreased, reaching a low of 7.15 Å for serpentines containing 1.0 Al per O₅(OH)₄ unit formula. Based on empirical data from Serna et al. (1979), peak positions of 7.25 Å for Serp 4 and 7.22 Å for Serp 6 correspond to ~0.4 to 0.5 Al per O₅(OH)₄ unit formula, respectively. The presence of broad and relatively weak peaks at 9.5 Å in addition to serpentine peaks described above indicated the presence of a polyphase mixture dominated by serpentine but also containing discrete packets of talc- or kerolite-like 2:1 layers, likely as zones within serpentine crystals, or as discrete talc-like crystals.

TEM

Morphologically, the majority of crystals produced by the synthesis reactions were tubular (Serp 1, 2, 3, 5, and 7); tube length was ~2000 Å (0.2 μm), and tube diameter was ~200 Å (Fig. 2), with one exception, Serp 3, where tube length was generally <500 Å and crystals were difficult to image due to instability under the electron beam. The two specimens that were distinct in XRD analysis, Serp 4 (containing Al³⁺) and Serp 6 (containing Al³⁺ and As³⁺), were also distinct in the TEM, consisting of platy crystals 1000–2000 Å in diameter and 100 to 200 Å thick, often hexagonal with either rounded or sharp edges.

Fig. 2 TEM images of tubular serpentines (Serp 1, Serp 2, Serp 5, Serp 7, top four images) and platy serpentines with hexagonal morphology (Serp 4, Serp 6). Serp 3 (not shown) is also tubular, but is distinct in its very fine size, making it difficult to image crystals compared to the six specimens shown here.

Analysis by EDX of dozens of crystals indicated that some were stoichiometrically serpentine-group minerals, yet others contained excess Si which suggested the presence of 2:1 layers interstratified with 1:1 serpentine layers (Table 2). Measurements of crystal chemistry (single crystals) were summarized individually from Serp 1 through Serp 7, and mineral formulas are presented on the basis of O₅(OH)₄ units; so for an ideal 1:1 clay, the sum of charges of cations is +14. Anything <14 implies vacancies or interstratified 2:1 layers. In terms of speciation, Al is commonly partitioned preferentially into the octahedral sheet, a decision based on analyses of natural serpentines which has shown that octahedral Al is often greater than tetrahedral Al (e.g. Fuchs et al. 1998; Deer et al. 2009). Partitioning Al into the octahedral sheet also produces compositions that are most consistent with the large ratio of serpentine to talc layers observed in all of the analytical methods. Partitioning Al into the tetrahedral sheet would elevate the number of tetrahedral sites, resulting in a greater abundance of 2:1 layers than was indicated by XRD and HRTEM results in the present study. This dominantly octahedral apportionment of Al resulted in the unit-cell compositions shown, but it did not rule out the potential for tetrahedral Al in cases where it was not shown. All arsenic, whether As⁵⁺ or As³⁺, was partitioned into the tetrahedral sheet (Hattori et al. 2005).

• • Serp 1, synthesized with only Mg and Si as cations (no Al or As), had molar Mg:Si ratios ranging from 1.1 to 1.4, where 1.5 is ideal for a serpentine with Mg₃Si₂O₅(OH)₄ stoichiometry, and 0.75 is ideal for a talc 2:1 layer. The average Serp 1 was a Mg-deficient tubular crystal with mean composition Mg_2.6 Si_2.0 O₅(OH)₄. The lack of Mg relative to Si led to a total cation charge of +13.2, consistent with ~15% interstratified 2:1 layers, as indicated by XRD data and HRTEM (below). Using terminology for interstratified clay minerals (Moore & Reynolds 1997), these could be described as “serpentine(0.85)/talc.”

• • Serp 2 contained Mg-deficient tubular crystals with mean composition ~Mg_2.3Al_0.16(Si_1.9Al_0.08 As⁵⁺_0.04)O₅(OH)₄. The sum of octahedral and tetrahedral charge was ~ +13.2 (where +14 is ideal serpentine), and the deficiency appeared to be in the octahedral sheet, where the sum of charges was +5.14 (+6.0 is the ideal Mg₃ end-member). In Serp 2, the mean tetrahedral charge was +8.04. The small octahedral charge implied the presence of interstratified talc-like layers (~15%), much like Serp 1.This is the only Al-bearing specimen occurring as chrysotile-like tubes rather than platy crystals.

• • Serp 3 contained tubular crystals with a mean composition of Mg_2.6 (Si_1.98 As⁵⁺_0.02)O₅(OH)₄, very similar to Serp 1, with a deficiency of Mg consistent with some interstratified talc-like layers. Similar to Serp 1 and Serp 2, the Mg:Si ratio of 1.3 was consistent with ~15% 2:1 talc-like layers in interstratified serpentine/talc. The tubular crystals in this specimen were much finer grained than in the other six; the majority had tube lengths <500 Å, making it difficult to obtain images or compositional analyses due to instability. As was detected in only two crystals (of 11).

• • Serp 4 consisted of Al-bearing platy crystals the mean formula of which, if all Al was partitioned into octahedral sites, was (Mg_1.76Al_0.72)(Si_2.0)O₅(OH)₄ and the cation charge was +13.7 (and if some Al were partitioned into tetrahedral sites, cation charge would drop further below the ideal value of +14.0). This charge deficiency (i.e. < +14.0) is consistent with a deficiency of octahedral sheets relative to tetrahedral sheets, implying the presence of talc-like 2:1 layers. Some single crystals produced compositions such as (Mg_2.40Al_0.43)(Si_1.91Al_0.09)O₅(OH)₄ where the ideal total charge is +14.0 for a serpentine-group mineral. In other crystals, compositions did not cast well as serpentines, e.g. (Mg_1.87Al_0.38)(Si_2.36)O₅(OH)₄, stoichiometry which for this individual crystal indicated the occurrence of talc-like layers in a ~60% serpentine:40% talc proportion. Overall, Serp 4 contained ~10% talc layers.

• • Serp 5 contained tubular crystals for which the mean composition was Mg_2.79 (Si_1.81 As_0.20) O₅(OH)₄. Assuming equal amounts of As³⁺ and As⁵⁺ in tetrahedral sites, cation charges summed to +13.6, a low value compared to the ideal +14.0 for a 1:1 clay. If all As were As⁵⁺, the cation charge sum would be +13.8. Values <14 indicate a 1:1 to 2:1 layer ratio that is smaller than that of an ideal serpentine, consistent with ~10% interstratified talc-like layers. Compositions of selected single crystals ranged from electrically neutral serpentine group stoichiometry, e.g. Mg_2.9 (Si_1.64 As_0.36) O₅(OH)₄, to crystals that appeared to be deficient in octahedral sheets for a serpentine group mineral, e.g. (Mg_2.44)(Si_2.28As_0.11)O₅(OH)₄. This was clearly not a stoichiometrically correct serpentine, and a reasonable explanation is the presence of interstratified talc-like layers in the dominantly serpentine mineral, consistent with XRD data. The most As-rich crystal analyzed by TEM-AEM had a composition of (Mg_2.38)(Si_1.37As⁵⁺_0.51 As³⁺_0.12)O₅(OH)₄, where As oxidation states were apportioned based on XANES data.

• • Serp 6 contained sub-hexagonal platy crystals with mean composition (Mg_2.07Al_0.52) (Si_1.97As³⁺_0.03) O₅(OH)₄. The sum of cationic charges was +13.7, consistent with ~10% talc-like 2:1 layers. The crystal with the largest As concentration had a formula of (Mg_2.22Al_0.56) (Si_1.89As³⁺_0.11) O₅(OH)₄ and the sum of cation charges was +14.0.

• • Serp 7 contained tubular crystals with an average formula (Mg_2.66) (Si_1.97As³⁺_0.03) O₅(OH)₄, yielding a cation charge sum of +13.3 that implied ~15% talc-like layers in single crystals. Interestingly, the two crystals with the largest As contents produced stoichiometries closest to 1:1 trioctahedral serpentine stoichiometry: (Mg_2.98) (Si_1.93As³⁺_0.07) O₅(OH)₄ and (Mg_2.96) (Si_1.86As³⁺_0.14) O₅(OH)₄. Cation charge sums for these two were +13.9 and +13.8, respectively.

  Table 2 Concentrations of the cations used in synthesis, presented in g kg^–1 for idealized synthesis stoichiometry ("Stoich"), as well as results of chemical analysis by TEM-EDS and ICPMS. These values are for hydrous serpentine where the O₅(OH)₄ component accounts for 520 ± 10 g/kg (variability depends on mass of cations). Si could not be analyzed by ICPMS. Sample IDs with "TU" indicate tubular crystals and "PL" indicates platy crystals

  

The HRTEM images of vertically oriented tubular crystals revealed diameters (perpendicular to the vertical b axis) of 150 to 200 Å with central holes of ~120 to 160 Å diameter. The rolled layers had thicknesses of 36 to 45 Å (Fig. 3a), but instability of the crystals in the presence of the electron beam prevented precise analysis of the layer structure. Crystals where the rolled layers were a total of 36 Å thick likely consisted of five stacked 1:1 7.2 Å layers; in other cases (e.g. thickness of 45 Å), these zones possibly contained five 1:1 serpentine layers and one 2:1 talc layer. While determining precisely the presence of interstratified 2:1 layers was not possible with HRTEM, this suggestion is consistent with XRD and TEM-AEM compositional measurements. In the case of platy hexagonal crystals, the majority of c-axis spacings were 7 Å but clear examples of clusters of interstratified ~ 9.5 Å layers were also visible (Fig. 3b), including lateral transitions between 1:1 (7 Å) layers and 2:1 (9.5 Å) layers. This evidence for serpentine with packets of interstratified talc is consistent with XRD and TEM-AEM data.

Fig. 3 HRTEM images showing (a) cross-sections of vertical tubular crystals in Serp 5 showing the thickness of rolled layers as well as the diameter of the central hole; and (b) interstratification of 7 Å and 9.5 Å layers in a platy crystal from Serp 4 that was oriented vertically to emphasize c-axis spacings.

ICPMS of Dissolved Serpentine Powders

Data from ICPMS (Table 3) provided information on Al, As, and Mg content of dissolved serpentines. Like TEM-AEM, analysis by ICPMS revealed the presence of Al and As at measureable levels in tubular and platy serpentines; unlike TEM, ICPMS analysis did not return reproducible results for Si, a problem attributed to low solubility of Si. Comparison of Al measured by TEM-EDS and ICPMS showed that TEM Al measurements were larger than those reported by ICPMS by up to an order of magnitude. Buffering of aqua regia solutions during dissolution of serpentine crystals could lead to formation of nano-scale Al hydroxides that were filtered prior to ICPMS analysis, thus producing an anomalously low bias to ICPMS Al data. Considering elemental As concentration in the synthesized crystals, TEM average values were sometimes larger than ICPMS (e.g. Serp 2 and 5), they were equal in one case (Serp 6), and lower in two specimens (Serp 3 and 7). In Serp 3, ICPMS detected As at 14.1 g/kg while TEM returned an average (N = 11) of only 1.2 g/kg, resulting from As detected in only two Serp 3 crystals in TEM. These differences were, at least in part, due to the small sample size of TEM analyses of single crystals (N = 7 to 12 per specimen) compared to analysis of a dissolved mineral powder in the case of ICPMS. Serp 3 consisted of many tubular nanoparticle serpentines (<20 nm) that were difficult to image and were unstable under the electron beam. Low Mg in Serp 4 measured by ICPMS may indicate incomplete dissolution of this particular powder when preparing the solution for analysis, which would also contribute to low Al measurement by ICPMS.

Table 3 Comparison of TEM-AEM to ICPMS analysis of mineral composition. In all cases, formulae are presented relative to an idealized O₅(OH)₄ unit cell, and in many cases, the sum of cationic charges is <14, a consequence of interstratified talc layers

If ICPMS results reflect the mean composition of dissolved serpentine, then they present a more representative analysis of average mineral composition. The problem is that pre-ICPMS dissolution may incorporate phases other than serpentine, or be affected by incomplete dissolution. The advantage of TEM-AEM is the capacity to determine the composition of individual crystals, but there is no guarantee that they encompass the full range of compositions for a powder sample, nor necessarily a representative average. What is clear in this study is that repeatable TEM-AEM measurements of single crystals provide reliable evidence of Al and As in tubular and platy serpentines, and the occurrence and magnitude of Al and As in these samples is supported by ICPMS analysis of dissolved powders.

XANES/XAFS

The normalized As K-edge XANES spectra obtained from Serp 5 and the liquid standards of As⁵⁺ and As³⁺ (Fig. 4a) indicated the predominance of As⁵⁺ (over As³⁺) (Table 4) in a synthesis experiment that contained equal amounts of those two species in the original synthesis solution. E ₀, measured as the white line, was found to be 11875.1 eV (As⁵⁺), 11871.7 eV (As³⁺), and 11874.8 eV (Serp 5). By fitting Serp 5 to the liquid standards using linear combination fitting in ATHENA (Table 4), the sample was 78% As⁵⁺ and 22% As³⁺ (Arai et al. 2001; Takahashi et al. 2003; Ravel & Newville 2005; Müller et al. 2010). The small R-Factor of 0.016 indicates the goodness of fit (Ravel & Newville 2005; Liu et al. 2008). Considering XANES data, the mean molar formula unit of Serp 5 was Mg_2.8 (Si_1.8 As⁵⁺_0.16As³⁺_0.04) O₅(OH)₄.

Fig. 4 XAS data: (a) Stacked, normalized, flattened As K-edge XANES spectra for Serp 5 and liquid standards of As⁵⁺ and As³⁺. The dashed line indicates a fit determined via LCF; (b) As K-edge XAFS spectra of Serp 5 as the Fourier transform of the k ³ * χ(k). Solid lines indicate raw data and dashed lines the fit to theoretical standards.

Table 4 Results of linear combination fitting showing combinations of standards yielding the best fits to arsenic K-edge XANES

Note: ^aUncertainties of fit reported were calculated by Athena.

^bR-Factor = Σ(data – fit)² / Σ(data²). R-factor is given as the error measure, but expressed as chi-squared, in this case, 1.666. Liquid standards of As⁵⁺ and As³⁺ were used as the standards for As (III) and As (V) components, respectively.

Arsenic K-edge XAFS analysis was utilized to investigate the position of As within the clay structure. The XAFS spectra in R space for Serp 5 (solid line) and fitted to theoretical standards (dashed line) are shown in Fig. 4b. Background removal, normalization, and Fourier transform were performed using the ATHENA program (Newville 2001). For the Fourier transform, a Hanning window was used with a k range of 2.7 to 12.2 Å^–1 and a k ³ weighting. The XAFS fitting was done using the Artemis program (Ravel & Newville 2005) from an R range of 1 to 3.5 Å. The theoretical scattering paths used in the fitting were generated using the FEFF 9.0 calculation code (Rehr et al. 2010) based on an antigorite structure with As substituted for a Si atom within the tetrahedral sheet (Dódony et al. 2002). During the fitting, only single scattering paths were considered. The amplitude reduction factor, S₀ ², was fixed at 0.96, the value obtained from fitting the XAFS spectrum for Na₂HAsO_4-7H₂O.

The first shell was fitted successfully as As-O with a coordination number of 3.3 and an interatomic distance of 1.71 Å (Fig. 4b, Table 5). The second shell was fitted successfully as As-Si with a coordination number of 3 and an interatomic distance of 3.20 Å and As-Mg with a coordination number of 3 and an interatomic distance of 3.45 Å. These fits agreed with XANES results indicating that both As⁵⁺ and As³⁺ were present in the tetrahedral sheet of Serp 5, As⁵⁺ with a tetrahedral coordination to four O atoms, and As³⁺ with a trigonal pyramid coordination to three O atoms, both with coordination to three Si within the tetrahedral sheet and three Mg within the octahedral sheet. Further analysis via Si K-edge XAFS would help to further probe the presence of tetrahedral Al and its relationship to As in the tetrahedral sheet. The calculated R-factor of 0.016 showed the relative goodness of fit, with good fits defined as R <0.05 (Liu et al. 2008). Online Resource 1 presents a K-edge EXAFS k3 * χ(k) spectrum of Serp 5.

Table 5 EXAFS fit parameters for Serp 5 and As⁵⁺ and As³⁺ liquid standards

Note: ^aR-Factor = Σ(data – fit)² / Σ(data²)

^bamplitude reduction factor

^cenergy shift

^dinteratomic distance

^ecoordination number

^fDebye-Waller factor

*fixed parameter.

FTIR

Tubular serpentines

The infrared spectra of the synthesized tubular serpentines contained a strong 980 cm^–1 peak that corresponds to Si–O stretching (Si and basal oxygen of the tetrahedral sheet, approximately parallel to the a-b plane) and a 1087 cm^–1 peak caused by Si–O stretching (Si and apical oxygen, perpendicular to a-b plane) (Fig. 5). Serp 2 differed slightly from the other tubular synthesized powders in that the basal Si-O stretching peak occurred at 1003 cm^–1 (rather than 980 cm^–1). The 3697 cm^–1 peak in the tubular serpentine spectra was produced by stretching of the OH bond in the presence of 3 Mg per hexagonal ring in the octahedral sheet (Mg₃O-H), and the weak shoulder at 3640–3648 may be related to octahedral vacancies.

Fig. 5 FTIR spectra of the synthesized serpentine group minerals. Peak positions are indicated in units of cm^–1.

The infrared spectra of the synthesized tubular serpentines were similar to those of natural chrysotiles (Farmer 1974; Craw et al. 1987; Suquet 1987; Rozalén et al. 2014), especially in terms of the strong 980 cm^–1 basal Si-O stretching vibration and the Mg₃O-H vibration at 3697 cm^–1 (Table 6). Natural chrysotiles produce a basal Si–O peak that ranges from 980 cm^–1 (Farmer 1974; Craw et al. 1987) to 960 cm^–1 (Yariv & Heller-Kallai 1975; Suquet 1987; Rozalén et al. 2014). Natural chrysotiles produce a Mg₃O-H vibration at 3695 ± 5 cm^–1. The ~1020 cm^–1 shoulder on the 960–980 cm^–1 peak is probably related to Si–O stretching in tetrahedral sheets of small amounts of talc-like or kerolite-like 2:1 layers (e.g. Farmer 1974; Čavajda et al. 2015). Spectra of tubular crystals with the greatest amount of As (Serp 5, Serp 7) contained a small peak at 880 cm. Peaks that occur in the ~590–640 cm^–1 range in synthesized samples in the present study and in natural chrysotiles are probably related to vibrations of the Mg–OH bond in the octahedral sheet (Suquet 1987).

Table 6 FTIR peaks (cm^–1) for tubular and platy serpentines synthesized for this study compared to data from natural chrysotile (Craw et al. 1987; Farmer, 1974; Rozalén et al. 2014, Suquet, 1987), antigorite (Mellini et al. 2002), and lizardite (Viti & Mellini, 1997), as well as serpentine synthesized by Serna et al. (1979)

sh = shoulder; wk = weak; br = broad; tr = "trace" i.e. < 0.03 p. f. u. (O₅(OH)₄), and Al values in the bottom row are p. f. u. O₅(OH)₄.

Bold indicates strong stand-alone peak, rather than shoulder, composite peak or weak peak.

Platy serpentines

The platy specimens (Serp 4 and Serp 6) produced spectra that were distinct from the tubular forms and similar to synthesized Al-bearing Mg-serpentines that contain 0.2 to 1.0 Al per O₅(OH)₄ (Serna et al. 1979), and natural antigorite with ~0.2 Al per per O₅(OH)₄ (Mellini et al. 2002). Those synthesized in the current study produced a strong basal plane Si–O stretching peak at 1007 cm^–1 (Serp 4) or 1003 cm^–1 (Serp 6), similar to the 995 cm^–1 peak of synthesized Mg-serpentines of Serna et al. (1979), or the 987 cm^–1 Si–O basal vibration from natural antigorite (Mellini et al. 2002). The shift in this peak towards higher wavenumbers in the synthesized serpentines was likely caused by interstratified talc layers that pull the Si-O stretching peak towards 1020 cm^–1 (Farmer & Russell 1967).

The 3662 and 3660 cm^–1 Mg–OH hydroxyl stretching peak for Serp 4 and Serp 6 corresponds to the 3660 cm^–1 peak for serpentine with 0.4 Al per O₅(OH)₄ unit formula (Serna et al. 1979), and the position of this peak is sensitive to Al, migrating to lower wavenumbers with increasing Al. The peak position for serpentines with no Al is 3685 cm^–1, whereas the peak for serpentines with 1.0 Al per O₅(OH)₄ unit formula is at 3650 cm^–1. This peak occurs at 3678 in natural antigorite with 0.16 Al (Mellini et al. 2002), also consistent with this effect. Thus, the 3660 position is consistent with 0.4 to 0.5 Al per O₅(OH)₄ unit formula in the octahedral sheet (TEM-AEM, for reference, indicates 0.5 to 0.7 Al per O₅(OH)₄). The 608 and 644 cm^–1 peaks were probably caused by Mg-OH vibrations in the chrysotile octahedral sheet (Farmer 1974). The weak 800–810 cm^–1 peak may have been produced by tetrahedral Al given that it occurs in synthesized Al-bearing serpentines from the current study as well as those of Serna et al. (1979), and in Al-bearing lizardite (Viti & Mellini 1997). If the peaks between 800 and 900 cm^–1 were to represent vibrations in the tetrahedral sheet corresponding to greater bond length or instability compared to Si–O bonds, they may be a signal of Al or As. Theoretically, peaks corresponding to As⁵⁺ in the tetrahedral sheet would fall between Si–O (~1000 cm^–1) and Al–O (~800 cm^–1; Serna et al. 1979). The FTIR peak at 880 cm^–1 (Fig. 5) appeared most prominently in the two synthesized tubular serpentines with the greatest As content, Serp 5 and Serp 7. In serpentines with smaller amounts of As, this peak was barely detectable. Four compounds with tetrahedral coordination of As⁵⁺ that contain peaks in this range are: AsO₄ ^3–, with ν1 symmetric stretching at 810–837 cm^–1 and ν2 anti-symmetrical stretching at 810–878 cm^–1; Ca₃AsO₄, with ν1 symmetric stretching at 840 cm^–1 and ν2 anti-symmetrical stretching at 880 cm^–1 (Farmer 1974); FeAsO₄, with a ν1 symmetric stretching at 823 cm^–1 (Chukanov 2014); and PbAl₃(AsO₄)₂(OH)₅ ^.H₂O ,with a composite ν1 symmetric stretching/ν2 anti-symmetrical stretching at 830–840 cm^–1 (Frost et al. 2013). Modeling indicates that the As-O stretching vibration caused by tetrahedral AsO₄ ^3– in crystalline solids should occur in the range 820–960 cm^–1 (Myneni et al. 1998), where the variability depends on coordination to other species in the mineral structure.

Discussion

Crystallochemical Characterization of the Synthesized Serpentines

The main mineral structure formed by the synthesis experiments was trioctahedral Mg phyllosilicate dominated by 1:1 layers with the capacity to incorporate arsenic into tetrahedral sites. Data from XRD, TEM, and FTIR indicated that the synthesized powders – whether tubular or platy – consisted mainly of 1:1 layers (serpentine) with 10–15% talc-like 2:1 layers, which in tubular crystals was randomly interstratified and in platy crystals occurred as discrete packets. Evidence for the occurrence of 2:1 layers included TEM images and compositions, lattice spacings of ~9.5 Å in electron diffraction, XRD data of tubular crystals indicating 15% randomly interstratified talc-like 2:1 layers in a dominantly 1:1 mineral, XRD data of platy crystals showing a broad ~9.5 Å peak caused by discrete packets of talc-like layers, and XRF data that showed evidence of talc layers in spectra otherwise dominated by serpentine.

Morphology, composition and terminology

The tubular morphology and evidence for interstratified talc layers in five of the seven specimens indicated the presence of what could be termed R0 chrysotile(0.85)/talc. The platy crystals in Serp 4 and Serp 6 possessed morphologies, compositions, and mineral structures that contain attributes of antigorite and lizardite. XRD patterns of the platy serpentines indicated that talc-like layers were physically intermixed with serpentine layers, and HRTEM images showed the talc-like layers as clusters within serpentine crystals. With insufficient precision to distinguish antigorite from lizardite layer type, the platy varieties could be termed platy serpentine with talc layers occurring in packets and comprising 10–15% of the crystal. These synthesized crystals are the first known occurrence of interstratified serpentine/talc. For the purpose of brevity, consistency, and context, from this point onward, the terms tubular serpentine and platy serpentine will be used.

The two crystal forms produced in these experiments were the result of variations in crystal radii of the cations in tetrahedral and octahedral sites. The tubular serpentines ranged compositionally from Mg-Si end-member (Serp 1) to crystals containing As (Serp 3, 5, and 7) or a mix of Al and As (Serp 2); the FTIR spectra for these were similar to that of chrysotile. The two specimens with platy crystals contained either Al (Serp 4) or Al and As³⁺ (Serp 6). The FTIR spectra for the platy crystals were more similar to naturally occurring antigorite than lizardite, but the data do not fit either perfectly, suggesting that the platy serpentines may possess elements of both antigorite-like and lizardite-like zones (Table 6).

In end-member Mg-Si serpentine, where the tetrahedral sheet b-axis dimension is less than the octahedral b-axis dimension, tetrahedral sheets must flex outward to increase distance between apical oxygen atoms and enable them to bond with the octahedral sheet (Deer et al. 2009). Chrysotile contains carpet-like rolls of 1:1 layers, where the octahedral sheets are on the outside of the roll for a given 1:1 layer. Substitution of As⁵⁺ (0.48 Å) or As³⁺ (0.54 Å) for Si (0.40 Å) at the levels achieved in the current experiments was not sufficient to overcome the tetrahedral-octahedral mismatch, and tubular crystals formed. In the platy serpentines (Serp 4 and 6), Al substitution (perhaps with some inversions) occurred at levels sufficient to overcome the mismatch (Fig. 6).

Fig. 6 Ternary diagram showing relative molar proportions of Mg, Al, and As (Table 3). Al-rich platy Mg-serpentines (Serp 4, 6) are indicated by hexagons to distinguish them from tubular serpentines.

By comparing the concentration of an element in serpentine relative to its concentration in the initial solution (Table 2), the serpentines appeared to strongly partition Al from solution. Similar to the synthesized Al-lizardites of Bentabol et al. (2010), and based on indirect evidence from TEM and FTIR, Al appeared to be partitioned into the octahedral sheet preferentially over the tetrahedral sheet. Si was slightly more concentrated in serpentine relative to the original solution, and Mg was stoichiometrically slightly lower than was targeted in the initial solution (i.e. 3 Mg:2 Si), with the greatest difference noted in the Al-rich Serp 4. Arsenic was also stoichiometrically lower in serpentine than in the initial solution with one exception, the slight partitioning of As⁵⁺ into Serp 5.

Earlier research has documented compositional overlap among the three main naturally occurring serpentine-group minerals; in general, lizardite and antigorite (i.e. platy forms) have greater Al contents than chrysotile (Deer et al. 2009). This is consistent with the crystals synthesized in this study, where the two platy serpentines contained greater amounts of Al than the tubular serpentines (Fig. 6), and Al in the synthesized platy serpentines occurred at levels comparable to natural Al-rich lizardites and Al-rich antigorites (up to 0.4 Al per O₅(OH)₄). In the case of lizardite, Al substitution enables the mineral to satisfy the mismatch, because when substituted for octahedral Mg (0.86 Å), the relatively small Al (0.72 Å) would shrink the octahedral sheet; when substituted for tetrahedral Si (0.40 Å), the relatively large Al (0.53 Å) could expand the tetrahedral sheet. Also, incorporation of Al into the octahedral sheet may lead to vacancies that also would shrink the octahedral sheet. Thus, by means of octahedral and/or tetrahedral substitution, Al can minimize the tetrahedral-octahedral mismatch. Some platy Al-rich serpentines do not solve the mismatch merely by Al substitution, but rather contain a modulating layer structure characteristic of antigorite (Fig. 7).

Fig. 7 Schematic diagrams of (a) geometries and scaled sizes of SiO₄ ^4–, AsO₄ ^3– , AlO₄ ^5–, AsO₃ ^3–, and FeO₄ ^5– anions, arranged in order of increasing cation–O bond distance (Table 6), and (b) examples of substitutions of Al, Fe³⁺, As⁵⁺, and As³⁺ in tetrahedral sheets of an antigorite crystal structure (re-drawn from Capitani & Mellini 2004). The central parts of the crystals (in the broken-line box) represents a small portions of a chrysotile crystal, including an As³⁺ ion showing the problem of its pyramidal geometry (rather than tetrahedral).

Insights into arsenic speciation in serpentine

The tubular serpentines are associated with larger As content than the platy serpentines, especially tubular Serp 5 with 2 wt.% As or 20 g kg^–1 (21 g kg^–1 by TEM-AEM and 11 g kg^–1 by ICPMS) in the mineral structure (based on a hydrous O₅(OH)₄ composition). This amount is equivalent to 10% of tetrahedral sites occupied by As (up to 0.2 p.f.u. As⁵⁺). Serp 3 – with only As⁵⁺ in addition to Mg and Si – had the lowest As concentration, likely because the only way to account for As⁵⁺ incorporation in the absence of paired substitution by a trivalent cation is vacancies. Serp 2 and Serp 6 represent an interesting comparison; they are both syntheses involving Mg, Si, Al, and As, where the only difference from onset of synthesis was that As⁵⁺ was in the Serp 2 initial solution (tubular crystals formed), compared to As³⁺ in the Serp 6 initial solution (platy crystals formed). The difference in morphology and composition implies that the oxidation state of arsenic influenced crystal nucleation and growth. Platy crystals formed when the substituted cation was Al³⁺ or Al³⁺ plus As³⁺, but when Al³⁺ was paired with As⁵⁺, tubular crystals formed. In Serp 2, the Al content apparently was too small to solve the tetrahedral-octahedral mismatch. This suggests that As⁵⁺ limited the incorporation of Al. Interestingly, the As content of tubular Serp 2 crystals was comparable to the other As-bearing synthesized serpentines.

For a given cation site, the most important controls on isomorphous substitution are ionic charge, crystal radius, and molecular geometry. Of the two As species with the potential to occur in the serpentine structure, both have charges within ±1 of Si⁴⁺. The crystal radius of As⁵⁺ in tetrahedral coordination (0.48 Å) is closer to that of Si (0.40 Å) than is tetrahedral Al (0.53 Å), and tetrahedral As³⁺ (0.54 Å) is nearly identical to Al. Thus, the smaller As⁵⁺ appears to be a better fit than As³⁺ or Al³⁺ in the tetrahedral sheet in terms of radius considerations, and XAFS data indicating the As-O bond length of 1.71 Å in Serp 5 is a very good match with theoretical As⁵⁺-O distance of 1.72 Å (Table 7).

Table 7 Crystal radii of pertinent cations in tetrahedral coordination as well as tetrahedral bond lengths (± 0.03 Å) for each cation

Crystal radii are from Shannon (1976) except for (a) EXAFS data from Niu (2011), and (b) Meunier (2005).

The crystal radius of oxygen in tetrahedral coordination is 1.22 to 1.24 Å.

Considering valence and molecular geometry, As⁵⁺ readily forms tetrahedral anions with O^2–, e.g. AsO₄ ^3–, which can substitute for SiO₄ ^4– in the tetrahedral sheet with little special accommodation required. The difference in valence structure of AsO₄ ^3– compared to SiO₄ ^4– could be accommodated by the absence of a hydrogen atom where bonding to the octahedral sheet occurs. As³⁺, on the other hand, tends to form AsO₃ ^3– with a pyramidal shape that effectively lacks an apical oxygen atom (Fig. 7). This geometry would probably not bond as well with the adjacent octahedral sheet as AsO₄ ^3–, although an As coordination number of 3.3 in Serp 5 (Fig. 4B, Table 5) and the XANES fit (Fig. 4A) are consistent with both As⁵⁺ and As³⁺ in the tetrahedral sheet (when both were included in the synthesis solution, i.e. Serp 5). From these first-principle considerations, As⁵⁺ is crystallographically well suited to occur in the tetrahedral sheet. As³⁺ may also occur in the tetrahedral sheet, but the pyramidal structure of AsO₃ ^3– may limit its incorporation relative to As⁵⁺. XANES analysis of Serp 5 is consistent with this, indicating that the highest oxidation state of arsenic, As⁵⁺, has a greater tendency to substitute into the tetrahedral sheet than As³⁺.

The charge imbalance of +1 that accompanies substitution of Si⁴⁺ by As⁵⁺ can be satisfied by paired substitution as follows (Hattori et al. 2005):

$\begin{array}{c}{{\mathrm{As}}^{5+}}_{\mathrm{tet}}+{{\mathrm{Al}}^{3+}}_{\mathrm{tet}}\to 2{{\mathrm{Si}}^{4+}}_{\mathrm{tet}}\\ {{\mathrm{As}}^{5+}}_{\mathrm{tet}}+{{\mathrm{As}}^{3+}}_{\mathrm{tet}}\to 2{{\mathrm{Si}}^{4+}}_{\mathrm{tet}}\end{array}$

or by omission of an H⁺ in the octahedral sheet, or by vacancies in the octahedral sheet, e.g.

$2{{\mathrm{As}}^{5+}}_{\mathrm{tet}}➔2{{\mathrm{Si}}^{4+}}_{\mathrm{tet}}+{{\mathrm{Mg}}^{2+}}_{\mathrm{oct}}$

The XAS data from Niu (2011) and the present study document the occurrence of As³⁺ in the tetrahedral sheet of serpentine, and Pascua et al. (2005) identified As³⁺ in hydrothermal Mg-smectites with 1500 to 4000 mg kg^–1 As (~0.01–0.02 As per Si₄O₁₀). The charge on As³⁺ does not limit incorporation into tetrahedral sheets, nor should size; for example, As³⁺ is smaller than the Fe³⁺ atom that occurs in tetrahedral sites of antigorite as well as other clay minerals (Wicks & O’Hanley 1988; Petit & Decarreau 1990; Petit et al. 2017). So, based on considerations of charge and radius, As³⁺ is compatible with substitution into tetrahedral sites and charge balance can be maintained as follows:

${{\mathrm{As}}^{3+}}_{\mathrm{tet}}+{{\mathrm{As}}^{5+}}_{\mathrm{tet}}➔2{{\mathrm{Si}}^{4+}}_{\mathrm{tet}}$

or:

${{\mathrm{As}}^{3+}}_{\mathrm{tet}}+{{\mathrm{Al}}^{3+}}_{\mathrm{oct}}➔{{\mathrm{Si}}^{4+}}_{\mathrm{tet}}+{{\mathrm{Mg}}^{2+}}_{\mathrm{oct}}$

The latter example is similar to Tschermak substitution cited to explain the occurrence of Al in serpentine (Serna et al. 1979):

${{\mathrm{Al}}^{3+}}_{\mathrm{tet}}+{{\mathrm{Al}}^{3+}}_{\mathrm{oct}}➔{{\mathrm{Si}}^{4+}}_{\mathrm{tet}}+{{\mathrm{Mg}}^{2+}}_{\mathrm{oct}}$

One important difference is that, unlike Si⁴⁺, As⁵⁺, Al³⁺, and Fe³⁺, which form tetrahedral anions in nature, As³⁺ does not (Fig. 7). The mostly likely orientation of the pyramidal AsO₃ in the tetrahedral sheet is one without an apical oxygen atom, leaving only an electron pair to bond with the adjacent octahedral sheet. The greater length of the As³⁺–O bond compared to the Si–O bond – plus the lack of an apical O atom (Fig. 7) – indicates that As³⁺ will have a tendency to distort the crystal structure in order to fit, so it may occupy tetrahedral sites at edges where size and bonding is less of a constraint. In antigorite, relatively large cations occur where tetrahedral sheets undergo inversion to satisfy the octahedral-tetrahedral mismatch (Moore & Reynolds 1997). The thicknesses (c-axis dimension) of the largest tetrahedral sites in antigorite (at inversions; Fig. 7) are 0.10–0.15 Å greater than non-inversion sites and tetrahedral bond lengths are commensurately greater at inversion sites (Capitani & Mellini 2004). This implies that the stability of the antigorite structure would be less impacted if As³⁺ (or Al³⁺, As⁵⁺, or Fe³⁺) were to occupy the larger tetrahedral inversion sites. The tubular structure of chrysotile involves distortions in the tetrahedral sheet that may accommodate As³⁺, and the often-high levels of Al in lizardite imply the potential to incorporate As³⁺.

The heterogeneity in As concentration in these synthesized serpentines (TEM-AEM data) and in naturally occurring antigorites implies that the incorporation of As into serpentine is kinetically controlled and dictated by solution composition. The synthesized serpentine crystals contained As concentrations ranging from < 1 mg kg^–1 to 20,000 mg kg^–1 (equivalent to as high as 0.2 mol As per 2 mol tetrahedral sites) and EMPA shows that As in natural antigorite can range from < 100 mg kg^–1 to 1300 mg kg^–1 over a scale of ~100 µm (Niu 2011). The heterogeneity indicates that As content in antigorite is related to local-scale differences in As availability during antigorite crystallization, consistent with influence of kinetic control on the variable levels of As incorporation into serpentine. It may also be a consequence of rapid crystallization. Compositional heterogeneity extends to Al and Fe in natural serpentines as well, as Yariv & Heller-Kallai (1975) documented, by EMPA analysis, extensive variability in isomorphous substitutions of Al that led them to state “some serpentines are so inhomogeneous that it is questionable whether the average chemical composition has any real significance.” Given these observations, two questions that this current study does not directly address are the effect of Fe on incorporation of As into Mg clays, and whether or not longer reaction time during synthesis would produce more homogeneous crystal compositions.

As-bearing Trioctahedral Clay Minerals in the Context of the Arsenic Cycle

Serpentinization of ultramafic rocks and crystallization of smectite in an intermediate-to-alkaline hydrothermal system are the two known environments that foster incorporation of arsenic into tetrahedral sites in Mg-phyllosilicates (Fig. 8). Evidence for this particular speciation includes XANES and XAFS analyses of antigorite, chlorite, and synthesized serpentine, indicating that both As⁵⁺ and As³⁺ occur in tetrahedral sites (Hattori et al. 2005; Ryan et al. 2011; Masuda et al. 2012). The EMPA and TEM analysis of natural antigorites, chlorites, and Mg-smectites indicate As concentrations that attain ~10³ ppm (0.1 wt.%, 0.01 to 0.02 mol per Si₂O₅ or Si₄O₁₀ unit cell), and compositions of synthesized serpentines indicate the potential to reach the order of 10⁴ ppm As (1 wt.%, 0.1–0.2 As per 2 tetrahedral sites). Sequential chemical extraction and EMPA of talc is consistent with the occurrence of As at concentrations of 65 to 80 mg kg^–1, presumably in tetrahedral sites (Ryan et al. 2011; Boskabadi et al. 2017).

Fig. 8 Schematic illustration of the arsenic cycle, emphasizing occurrence of trioctahedral Mg silicates (hexagonal symbols). Concentrations in most reservoirs are quite variable, so the values shown at the lower left, with one exception (seawater), attempt to represent this variability. The symbol “ms” indicates metasedimentary rock. Sources for As reservoir concentrations are Smedley & Kinniburgh (2002), Milham & Craw (2009), Ryan et al. (2011), and Breuer & Pichler (2013). Diagram adapted from Deschamps et al. (2010) and Masuda (2018).

In high-temperature systems, As is transported with other fluid-mobile elements (e.g. B, Sb, Cs) in waters derived from subducting sediments (Fig. 8), and the fate of As in these waters is controlled to some extent by Mg-rich clay minerals. If the waters react with upper-mantle rock, they may form serpentine or chlorite that fixes As, providing insight into processes associated with subduction, serpentinization, and subsequent metamorphism (Hattori et al. 2005; Deschamps et al. 2010). If the waters evolve into shallow alkaline hydrothermal systems, As may become fixed into tetrahedral sheets when Mg-smectite crystallizes (Pascua et al. 2005). While shallow hydrothermal systems are recognized as a source of elevated As in groundwater, more research is needed to ascertain the role that As-bearing clays might exert in controlling As fate and transport in these systems (López et al. 2012).

Given the association of Mg clays with alkaline environments of formation, including serpentinizing fluids and intermediate-to-alkaline hydrothermal systems (Inoue 1995; Lafay et al. 2016), and the occurrence of the AsO₄ ^3– oxyanion in alkaline waters, it is likely that the optimal conditions for fixing As into clay-mineral tetrahedral sites occurs when clays precipitate from alkaline fluids. Conditions like these occur in near-surface hydrothermal (~100^oC; Pascua et al. 2005) and deeper lithosphere serpentinizing systems (~250^oC; Deschamps et al. 2010), and perhaps in alkaline surface waters where Mg-smectite, kerolite, sepiolite, and associated Mg clays crystallize (Jones & Conko 2011).

In earth surface hydrologic systems, As in tetrahedral sites of antigorite and chlorite has been identified as a source of elevated As in groundwater, emphasizing the importance of understanding this particular speciation. Trioctahedral clay minerals may function either as a direct source of As to aquifers by way of in situ chemical weathering (Ryan et al. 2011; Masuda et al., 2012) or by weathering in a sedimentary source area where As becomes fixed into secondary minerals (e.g. Fe (oxyhydr)oxides) that are then transported into the aquifer (Hattori et al. 2005; Guillot & Charlet 2007). In both scenarios, understanding of the role of silicate clays as a potential As source is needed for accurate models (Mukherjee et al. 2014; Masuda 2018).

Incorporation of As depends on many environmental factors, including fluid composition, reduction-oxidation potential (as fO₂, Eh or pe; Fig. 9), pH, temperature, co-precipitation, and likely microbial influences (depending on setting). The concentration of As in hydrothermal waters ranges generally from <1 to 50 mg/L (Ballantyne & Moore 1988; Webster & Nordstrom 2003), making the As concentrations in the synthesis solutions an apt comparison (16.7 or 33.3 mg/L As; Table 2). While it is not possible to measure concentrations of fluids deep in suprasubduction zones, evidence from shallower systems suggests that As concentrations may range from <10 to 10² or 10³ mg/L (Breuer & Pichler 2013). Thus, the incorporation of As into up to 10% of tetrahedral sites of Mg phyllosilicates synthesized in the current study, i.e. when [As] in solution is on the order of 10¹ to 10² mg/L, provides a reference point for natural systems.

Fig. 9 fO₂ -pH diagram for a fluid system at 250^oC showing speciation of arsenic as a function of pH and redox conditions, where oxygen fugacity (fO₂) is a measure of oxidation-reduction potential of mineralizing fluids (higher values indicating more oxidizing). This system includes As, S, Fe, and O, and indicates the propensity for As to occur as tetrahedral anions at moderate to high oxygen fugacity across a wide range of pH. Modified from James-Smith et al. (2010).

Aqueous As⁵⁺ species are stable over a wide range of redox conditions, pH, and temperature (Shock et al. 1997), and the suprasubduction zone fluids that foster serpentinization of peridotites evolve towards alkaline solutions with pH ≥ 10 (e.g. Barnes & O’Neil 1969), conditions where tetrahedral AsO₄ ^3– would be available for incorporation in serpentine group minerals. In neutral to acidic systems, protonation of AsO₄ ^3– to HAsO₄ ^2–, H₂AsO₄ ^–, and H₃AsO₄ could limit incorporation into the tetrahedral sheet. In reducing conditions and with decreasing pH, reduction of As⁵⁺ to As³⁺ and eventually As^3– or As⁰ would preclude fixation into the silicate structure, and instead fix As into arsenides or sulfides (Fig. 9; Ishimaru & Arai 2008; Deschamps et al. 2010). Redox conditions that favor the presence of As³⁺ in mineralizing solutions may result in serpentines with no As⁵⁺ and relatively large amounts of As³⁺, and if Eh conditions are low-intermediate, a mix of As⁵⁺ and As³⁺ may occur in the same crystal; in this way, As oxidation state may record redox conditions of lithospheric waters (Hattori et al. 2005).

Conclusions

Tubular and platy arsenic-bearing serpentines were synthesized at 200^oC in alkaline solutions for reactions of 10 days' duration. The arsenic occurs as As⁵⁺, As³⁺, or both, in tetrahedral sites of serpentine at levels reaching 10⁴ mg kg^–1 (ppm); future analyses using Al NMR and Si EXAFS would help to corroborate these results. The tetrahedral AsO₄ ^3– oxyanion appears to be particularly well suited for substitution into the tetrahedral sheet, and alkaline conditions that favor crystallization of Mg clays coincide with As occurrence as AsO₄ ^3– (or in reduced waters, AsO₃ ^3–). This explains the occurrence of tetrahedral As in serpentine-group minerals, chlorite, Mg-smectites, and talc, and suggests that As may also occur in other Mg clays that form in alkaline environments, e.g. sepiolite and palygorskite.

Electronic supplementary material

The online version of this article (https://doi.org/10.1007/s42860-019-00040-1) contains supplementary material, which is available to authorized users.