Manganese, Arsenic, and Carbonate Interactions in Model Oxic Groundwater Systems

Abstract

Manganese and arsenic both threaten groundwater quality globally, but their chemical behavior leads to both co-contamination and separation of these contaminants from individual well to regional scales. Here we tested manganese and arsenic retention under conditions commonly found within aquifer redox fluctuating and transition zones where both arsenic and iron phases are present in oxidized forms, but manganese persists as reduced and soluble Mn(II). Analysis of column aqueous breakthrough data and characterization of solid-phase products using X-ray photoelectron (XPS) and absorption spectroscopies (XAS) show that the addition of bicarbonate increased manganese retention but decreased arsenic retention, while the presence of manganese and arsenic together increased both arsenic and manganese retention. In the presence of O₂ arsenic remained oxidized as arsenate under all conditions measured; however, reduced Mn(II) was oxidized to an average Mn oxidation state of ~3 in the absence of arsenate. The presence of arsenate partially inhibited Mn(II) oxidation likely by blocking ferrihydrite surfaces needed to catalyze Mn(II) oxidation by O₂ and by stabilizing Mn(II) via ternary complex formation. These results highlight the interactions between reduced and oxidized contaminants that can contribute to the co-occurrence or physical separation of manganese and arsenic in groundwater systems under changing or stratified redox conditions.

Graphical Abstract

INTRODUCTION

More than 2.5 billion people worldwide consume drinking water solely from groundwater sources¹ and groundwater provides 43% (and increasing) of global agriculture irrigation water.² While regional decreases in groundwater quantity due to over abstraction threaten the sustainability of groundwater resources,³ groundwater contamination also limits the suitability of groundwater uses, especially as a primary drinking water source.⁴ Further, since well drilling typically targets the shallowest aquifer that meets production (i.e. water quantity) requirements, wells are drilled and screened over a wide range of depth intervals depending on local geologic and hydrologic conditions. Thus, understanding of the relationships between well depth and the distribution of groundwater contaminants is necessary to ensure human safety and efficient use of resources.

Arsenic (As) occurs as a groundwater contaminant worldwide and presents a significant risk to the health of millions of people; as such, both its human health effects⁵ and geochemistry^6,7 have been extensively studied. Although manganese (Mn) is currently not widely accepted as a priority contaminant,^8-10 exposure to elevated Mn levels through drinking water may lead to numerous adverse health effects.^11-18 Manganese groundwater contamination is also prevalent and may occur both as a co-contaminant and at distinct depths from As^19,20 although little data exists on the health implications of co-exposure to contaminants such as As and Mn. As evidence for Mn toxicity grows, Mn groundwater contamination renders additional wells unusable as a drinking water source compared to As alone. Depth stratification of arsenic and manganese contamination has been reported at the global scale, with arsenic groundwater contamination occurring at deeper depths than manganese.¹⁹ One potential explanation is the contrasting effect of bicarbonate species on As and Mn mobility. The adsorption, redox, and precipitation properties of aqueous As and Mn also vary under oxic conditions typically experienced in redox-transition environments and may provide an additional explanation of separate well contamination by As and Mn.¹⁹ Although the processes governing the occurrence of geogenic groundwater As contamination have been extensively studied⁶ the chemical mechanisms that lead to separation of As and Mn in aquifers are not fully understood.

Carbonate concentration and speciation (CO₂/H₂CO₃, HCO₃⁻, and CO₃²⁻) in porewater varies significantly and is controlled by pH, organic carbon oxidation rate, and the presence of divalent cations leading to mineral dissolution/precipitation. Bicarbonate accumulates in porewater as a product of organic carbon oxidation, especially in reducing environments, and influences both As and Mn fate in aquifers. Dissolved Mn concentrations can be controlled by MnCO₃ precipitation under both oxic and anoxic conditions²¹ while bicarbonate competes for mineral adsorption sites with As(V) oxyanions.^22-24 Through these mechanisms, increasing bicarbonate concentration decreases aqueous Mn(II) through mineral precipitation and increases As(V) through surface competition.

In this study, the conditions were selected based on observations of the presence of dissolved Mn under “less reducing” or oxic conditions, specifically aqueous Mn occurrence independent of dissolved Fe(II)²⁵ or separation from Fe and As concentrations in deep oxidized Pleistocene sediments.^19,26 Specific conditions for this study were based on field data demonstrating elevated Mn and As concentrations from the Yangtze River Basin.^27,28 Time series field data indicated that As concentrations increased when reducing conditions prevailed and decreased during oxic periods, while dissolved Mn persisted through both oxic and anoxic periods (Figure S1). These results indicated close coupling of redox conditions and dissolved As concentration, but little change in Mn concentration during redox cycling in the same aquifer. Fe oxides are abundant in the relatively oxidized sediment near the screened interval of the monitoring well.^29,30 Therefore the objective of this study was to determine the interactions between As(V), Mn(II), and bicarbonate species on As and Mn retention on poorly crystalline iron (Fe) hydroxide (ferrihydrite, Fe(OH)₃), a common and reactive mineral in soils and sediments.

Our motivation was to determine the interactions between Mn, As, and bicarbonate during oxic periods of redox oscillating environments using a model column study with ferrihydrite-coated sand. We performed experiments using soluble and reduced Mn(II)_(aq) with oxidized As(V)_(aq) and Fe(III)_(s) species to determine the relevant interactions between these constituents during the following common groundwater scenarios: (1) oxic periods following anoxic conditions in redox oscillating soils and sediments; and (2) separation of dissolved Mn and As in redox stratified environments,¹⁹ where aqueous Mn persists in relatively oxic (shallower) conditions and aqueous As dominates in relatively reducing (deeper) conditions. We hypothesized that under our experimental conditions bicarbonate would increase Mn retention through precipitation of rhodochrosite (MnCO₃) and decrease As(V) retention through competitive sorption.²² Our results show that bicarbonate increased Mn retention and decreased As retention via Mn(II) oxidation, not solely through MnCO₃ precipitation. The presence of As(V) and Mn(II) together decreased the extent of Mn(II) oxidation but increased both Mn and As retention. The results highlight the complex interactions between Mn, As, and bicarbonate under oxidizing flow conditions and aid in interpreting and predicting the distribution of groundwater contaminants.

EXPERIMENTAL SECTION

Column Setup.

Columns containing ferrihydrite-coated sand were used to simulate aquifer processes under hydrodynamic conditions. Ferrihydrite paste was synthesized following the protocol of Schwertmann and Cornell³¹ and subsequently mixed with IOTA-6 quartz sand to produce ferrihydrite-coated sand (Fhy-sand). Fhy-sand was air dried, rinsed repeatedly with deionized (DI) water and again air-dried. A sample of Fhy-sand was finely ground with mortar and pestle, and analyzed using X-ray fluorescence to determine the initial chemical composition of the sand. The Fe content of Fhy-sand was 2.34% by mass and initially contained 41 ppm Mn and <0.5 ppm As (Table 1). Fhy-sand was wetted with DI water and approximately 22 g (dry mass) was packed into each of six borosilicate glass columns (Kimble Kontes FlexColumn Economy Columns, 1.5 cm inner diameter × 10 cm length), resulting in a porosity of 0.53–0.55. Columns were eluted with DI water for ~10 pore volumes to remove air and check for leakage prior to the experiment.

Table 1.

Solid-Phase Concentrations and Elemental Ratios of As, Mn, and Fe Following Column Deconstruction

treatment	Asμmol g−1	Mnμmol g−1	Feμmol g−1	As/Femol mol−1	Mn/Femol mol−1	As/Mnmol mol−1
ferrihydrite sand	Nd	0.7	419	nd	0.002	nd
Mn(II)	Nd	11.2	456	nd	0.025	nd
Mn(II) + CO32−	Nd	88.7	358	nd	0.248	nd
As(V)	14.7	0.5	321	0.046	0.002	nd
As(V) + CO32−	11.6	0.5	309	0.038	0.002	nd
Mn(II) + As(V)	17.1	24.3	343	0.050	0.071	0.702
Mn(II) + As(V) + CO32−	19.3	72.6	427	0.045	0.170	0.266

Column Flow Experiments.

Ferrihydrite-coated sand columns were connected to feedstock solutions using Tygon S3 tubing and flow was maintained at a rate of 35 mL d⁻¹ using an ISMATEC IPC High Precision Multichannel Dispenser. All feedstock solutions were prepared using deionized water (18.2 MΩ-cm) and contained 0.1 M NaCl and 10 mM PIPES buffer. Manganese and As(V) were added using MnCl₂ and Na₂HAsO₄, respectively, and bicarbonate was added using NaHCO₃. Solution pH was adjusted to 7.00 (±0.01) following addition of all reagents. Feedstocks were freshly prepared every 28 days or less, and precipitates were not observed in the feedstock bottles. Homogeneous oxidation of Mn(II) by molecular oxygen is kinetically-limited with a half-life of at least 400 days at pH < 8;⁴² therefore, the maximum amount of Mn(II) oxidized within a feedstock bottle was 4.7%. A total of six column treatments were performed. Columns were run with continuous flow in three phases:

Phase 10–147 days, three treatments in duplicate: Duplicate columns were fed stocks of 10 mg L⁻¹ Mn(II) or 1.6 mg L⁻¹ As(V), or both 10 mg L⁻¹ Mn(II) and 1.6 mg L⁻¹ As(V). These values were chosen based on the highest concentrations detected in the National Hydrochemical Survey performed on Bangladesh groundwaters³² and are consistent with high values of As and Mn concentrations reported throughout Asia.^{6,7,19,20,33,34}

Phase II 147–261 days, bicarbonate addition to one of each duplicate column: Sodium bicarbonate (4.2 mM NaHCO₃) was added to the influent of one of each duplicate column, and solution pH was adjusted to 7 following NaHCO₃ addition. Bicarbonate was added to these columns through Phase II and Phase III.

Phase III 261–350 days, As(V) concentration increased: The As(V) influent concentration was increased from 1.6 to 5 mg L⁻¹ for all columns containing As to ensure breakthrough to assess both enhancement or inhibition of As mobilization from columns with the addition of bicarbonate.

Aqueous Analyses.

Column outflow was collected in 50 mL centrifuge tubes and aliquoted every 1–3 days. Aqueous solutions were acidified to a final concentration of ~2% v/v using concentrated trace metal grade HNO₃ and analyzed for As, Fe, and Mn, using inductively coupled plasma optical emission spectrometry (ICP-OES, Perkin-Elmer Optima 7300DV). Stock solutions were also analyzed using ICP-OES to account for variation in input concentrations. The detection limit was 0.005 mg L⁻¹ for As and 0.01 mg L⁻¹ for Mn and Fe.

Column Deconstruction and Solid-Phase Analyses.

Following cessation of flow, columns were dismantled and solids from each treatment were harvested post-reaction and air-dried. Total elemental distribution of post-reaction solids was determined by finely grinding ~3 g of sample and analyzing it using energy dispersive X-ray fluorescence (ED-XRF) as described previously.²⁹ Samples were also analyzed using X-ray absorption spectroscopy (XAS) at the As K-edge to identify the major binding modes of As as a function of solution conditions. Samples were prepared by finely grinding with an agate mortar and pestle, then loading into a multisample holder sealed with 25 micron (1 mil) Kapton tape. Samples were analyzed at beamline 4-1 at the Stanford Synchrotron Radiation Lightsource (SSRL) in a liquid N₂ cryostat (~77 K). Energy selection was achieved using a Si(220) double monochromator and beam spot was focused to 1 × 10 mm. Energy was calibrated to the first inflection point of gold foil (11,919 eV). Energy was scanned from 11,638 to 11,848 eV in 10 eV steps, 11,848 to 11,883 in 0.3 eV steps, and to k = 15 Å in 0.05 Å⁻¹ steps. A minimum of three scans were collected for each sample and compared to check for beam damage (photo oxidation/reduction), which was not observed in any of the samples. Incident flux (I₀) was monitored using an ion chamber and fluorescent X-rays were detected using a PIPS detector and all fluorescence signal was normalized to I₀. Data normalization and background removal were performed using Athena software.³⁵ Linear combination fitting (LCF) analysis of the As XANES region was also performed in Athena. Ten standards were initially used and nonrelevant standards were eliminated stepwise. Data were ultimately fit using only two standards: arsenite and arsenate adsorbed on ferrihydrite. Shell-by-shell modeling of the EXAFS region was performed using FEFF implemented in Artemis. The ferrihydrite structure proposed by Michel et al.³⁶ was used and modified to include arsenate or Mn tetrahedron in place of a single Fe tetrahedron in the structure depending on the experimental conditions. Specific details of the fitting procedure along with details for Mn XANES and EXAFS data processing are provided in the Supporting Information.

X-ray photoelectron spectroscopy (XPS) characterization was carried out by using a Kratos AXIS ULTRA^DLD XPS system equipped with an Al Kα monochromated X-ray source and a 165 mm mean radius electron energy hemispherical analyzer with a slot and electrostatic lens mode. Vacuum pressure was kept below 3 × 10⁻⁹ torr during the acquisition, and a neutralizer was applied to compensate sample charging during the measurement. Additional data acquisition parameters including atomic sensitivity factors are listed in Supporting Information Table S1, and a more detailed methodology for XPS peak deconvolution is provided in the Supporting Information.

RESULTS

Aqueous Concentrations and Breakthrough Curves.

Manganese Columns.

Manganese breakthrough was observed within the first 22 pore volumes, and effluent Mn concentrations remained at steady state for the remainder of the first phase of the experiment (Figure 1a). At ~530 pore volumes, influent was altered to increase Mn concentration to 25 mg L⁻¹ in one column and bicarbonate was added with continued addition of 10 mg L⁻¹ Mn in the other column. An increase in effluent Mn concentration was observed in the column with increased Mn influent, but addition of bicarbonate lead to a decrease in effluent Mn concentration in the second column. The Mn concentration in the first column was reduced back to ~10 mg L⁻¹ at 640 pore volumes and the effluent concentration remained at steady state, nominally in balance with the influent concentration, for the duration of the experiment. In the manganese only column, input Mn(II) concentration was temporarily increased from 10 ppm to 25 ppm to test whether higher input Mn(II) concentration would lead to additional Mn retention to ferrihydrite-coated sand. Results show that a small amount of additional Mn(II) was retained, but that retention following the increase in Mn(II) concentration was minimal as evidenced by the rapid increase in effluent Mn concentration. This observation points to a finite capacity for Mn retention that is nearly independent of input Mn(II) concentration in the absence of additional aqueous species such as As(V) or bicarbonate. Finally, a temporary increase in Mn(II) concentration from 10 ppm to 25 ppm had less net effect on Mn retention than the co-presence of As(V), bicarbonate, or both As(V) and bicarbonate indicating that Mn concentration alone is not the dominant factor determining overall Mn retention to ferrihydrite-coated sand (Table 1). In the second column with bicarbonate, effluent Mn concentration was more variable but overall lower than in the absence of bicarbonate. Some variability in the Mn(II) + bicarbonate column was associated with changing stock solution which may have been due to a minor shift in pH over the length of time one bottle of feed solution is used (28 days), after which new feedstock is provided.

Figure 1.

(a, c) Dissolved manganese and (b,d) arsenic concentration in effluent of columns with influent containing (a) Mn(II), (b) As(V), or (c, d) Mn(II) + As(V). In each treatment columns were initially replicates until >500 pore volumes represented by the vertical gray line, when bicarbonate was added to columns indicated by open squares. In columns containing arsenic (b, c, d) the influent As concentration was increased from 1.6 to 5 mg L⁻¹ at 940 pore volumes indicated by the solid black line.

Arsenic Columns.

Addition of bicarbonate resulted in faster breakthrough of As (Figure 1b). Both columns with influent As(V) had effluent As concentrations below detection until >700 pore volumes, when As concentration in effluent reached 0.05–0.1 mg L⁻¹ in the presence of bicarbonate and increased to >0.5 mg L⁻¹ by 850 pore volumes (Figure 1b). Effluent As concentration remained below detection in the absence of bicarbonate until >900 pore volumes. At 940 pore volumes, the influent As(V) concentration was increased to 5 mg L⁻¹ in both columns to ensure breakthrough to determine whether bicarbonate addition may in fact have an inhibitory effect on As(V) elution. Following the increase in As(V) influent concentration, breakthrough was observed from 940 to 1000 pore volumes in both columns before reaching a plateau nominally at the input concentration of 5 mg L⁻¹.

Manganese and Arsenic Columns.

To determine if As and Mn co-contamination influences As and Mn mobility in the presence and absence of bicarbonate, a third set of columns were run with Mn(II) and As(V) in the influent. The presence of both Mn(II) and As(V) in influent resulted in an initial rapid increase in effluent Mn concentration to ~8.5 mg L⁻¹, then a further increase to ~11 mg L⁻¹ at 390 pore volumes (Figure 1c). Similar to columns with only Mn(II) influent (Figure 1a), bicarbonate addition at ~530 pore volumes lead to both a decrease in Mn effluent concentration and an increase in Mn effluent concentration variability (Figure 1c). Following the influent As(V) concentration increase to 5 mg L⁻¹ at ~950 pore volumes, Mn effluent concentration increased in the absence of bicarbonate and then gradually increased as As(V) breakthrough was observed (Figure 1c,d). In the presence of As(V) and bicarbonate Mn effluent concentration remained variable, but a decreasing trend was observed following the increase in As(V) input concentration (Figure 1c).

Similar to As(V) columns, effluent As concentration was below detection until >800 pore volumes in the presence of Mn(II) (Figure 1d). With added bicarbonate As(V) breakthrough occurred in the As+Mn columns prior to increasing the influent concentration to 5 mg L⁻¹, whereas without bicarbonate, breakthrough occurred ~100 pore volumes after the increase in influent As(V) concentration. Bicarbonate addition to the As + Mn columns also increased variability in As(V) effluent concentration (Figure 1d).

Solid-Phase Analyses

Iron Oxide Stability.

The Fe(OH)₃ surface coating remained stable throughout each column experiment and contained only Fe(III) at the end of the experiment (Supporting Information Figure S3) and dissolved Fe was below detection (0.05 mg L⁻¹) in aqueous measurements in all columns (data not shown). These results imply that the overall redox potential remained above the threshold for Fe(III) reduction throughout the experiment.

Manganese and Arsenic Retention.

Ferrihydrite-coated sand retained both Mn and As (Table 1). Because all columns achieved breakthrough of As(V) and Mn(II) (Figure 1), differences in solid-phase concentration between treatments are interpreted as differences in steady-state retention capacity averaged over the entire column. To account for variation in ferrihydrite coating of quartz sand between columns (1.72–2.54% Fe; Fe/Si 0.049–0.074), ratios of solid-phase As and Mn content normalized by Fe content are also presented (Table 1).

In the column with only Mn(II) addition, Mn solid-phase concentration increased from 0.7 to 11.2 μmol g⁻¹ (41–618 ppm); in the column with only As(V) addition, the solid-phase As concentration increased from below detection to 14.7 μmol g⁻¹ (<0.5–1100 ppm) (Table 1). The addition of bicarbonate decreased As retention by 25% compared to As-only treatment while Mn retention increased by nearly 8-fold compared to Mn-only treatment (Table 1). When As(V) and Mn(II) were both added to influent, solid-phase retention of each increased compared to treatments with As(V) or Mn(II) alone (Table 1). Both As(V) and CO₃²⁻ increase Mn(II) retention individually compared to Mn(II) alone, but Mn(II) retention when all three species were present was lower than Mn(II) + CO₃²⁻ and higher than Mn(II) + As(V) (Table 1).

Mn and As X-ray Absorption Spectroscopy.

Arsenic K-edge XANES spectra showed that arsenate comprised >90% of the As species in all four samples that included As, indicating As(V) remained the dominant and likely only As phase throughout each experiment (within measurement error of XAS). Arsenic EXAFS data and nonlinear least-squares shell fits were optimized using a first shell of 3 O atoms with a bond distance of 1.67 Å (Table 2) and 1 O atom with a bond distance of 1.74 Å (Figure 2 and Table 2). In the column with only As(V), a single As–Fe scattering path with distance of 3.28 A reproduced the data well. A similar As–Fe path was included in fits for each treatment (Table 2). In the As(V) + Mn(II) treatment, addition of an As–Mn path with a distance of 3.35 Å significantly improved the fit compared to As–O and As–Fe paths alone. Data of As(V) + bicarbonate treatments were best fit by including a C scattering path with an As–C bond distance of 3.26 Å. Similar Mn and C paths were used to fit the As(V) + Mn(II) + bicarbonate treatment, but an As–Mn path representing MnCO₃ was also added and significantly improved the fit (Figure 2 and Table 2).

Table 2.

Arsenic K-Edge EXAFS Shell Fitting Parameters Including Path Degeneracy (N), Radial Interatomic Distance (R), and Debye–Waller Factor (σ²)^a

	As-O−			As-OH			Fe
	N*b	R(Å)	σ (Å2)	N*	R(Å)	σ(Å2)	N*	R(Å)	σ(Å2)
As only	3	1.67 ± 0.00737	0.00335 ± 0.0000825	1	1.74 ± 0.000737	0.00335 ± 0.0000825	2	3.28 ± 0.0259	0.00836 ± 0.00257
As+CO32−	3	1.67 ± 0.0075	0.00252 ± 0.000829	1	1.74 ± 0.0075	0.00252 ± 0.000829	2	3.28 ± 0.0767	0.01269 ± 0.00902
As+Mn	3	1.67 ± 0.00639	0.00261 ± 0.000711	1	1.74 ± 0.00639	0.00261 ± 0.000711	2	3.28 ± 0.271	0.0107 ± 0.00282
As+Mn + CO32−	3	1.67 ± 0.00789	0.00199 ± 0.000873	1	1.74 ± 0.00789	0.00199 ± 0.000873	2	3.28 ± 0.767	0.02796 ± 0.0109
	C (carbonate)−			Mn(II) (sorbed)			Mn (rhodochrosite)
	N*	R(Å)	σ (Å2)	N*	R(Å)	σ(Å2)	N*	R(Å)	σ(Å2)
As only
As+CO32−	2	3.26 ± 0.0683	0.0008 ± 0.00615
As+Mn				1	3.35 ± 0.0271	0.0107 ± 0.00282
As+Mn + CO32−	2	3.25 ± 0.044	0.00048 ± 0.00407	1	3.38 ± 0.0767	0.02796 ± 0.0109	6	3.8 ± 0.0767	0.02796 ± 0.0109

^aErrors for R and σ² are ± standard deviation.

^bN*Coordination number fixed.

Figure 2.

X-ray absorption spectroscopy data collected at the arsenic K-edge. (a) X-ray absorption near edge structure (XANES) spectra show As(V) is the dominant species in all treatments. The dashed line at 11,875 eV represents the white line position of sodium arsenate for reference. The proportion of As(V) was >90% in all XANES linear combination fits. (b) k³-weighted extended X-ray absorption fine structure (EXAFS) spectra and (c) radial distribution function (RDF) show the local binding environment of As atoms, displayed visually in (d). Shell fit parameters of the EXAFS data are given in Table 2.

Manganese K-edge XANES data show that Mn(II) was oxidized in all treatments but to varying extents (Figure 3 and Table 3). In the absence of As, Mn average oxidation state (AOS) was ~3 with and without bicarbonate addition. The treatment with influent containing As(V) + Mn(II) had the lowest Mn AOS of 2.62, indicating similar amounts of Mn(II) and Mn(III). In the treatment with As(V) + Mn(II) + bicarbonate Mn AOS was 2.80 which is between the AOS of As(V) + Mn(II) and no As treatments. In addition, Mn EXAFS analyses confirmed results from XANES data where the first shell of Mn is well described by oxygens corresponding to Mn–O octahedron of groutite, a Mn(III) oxide, in the Mn(II) + bicarbonate and As(V) + Mn(II) + bicarbonate treatments (Figure S8). However, Mn within the Mn(II) + As(V) column exhibited longer Mn–O distances that indicated dominance of Mn(II); this finding is consistent with XANES results showing that Mn(II) oxidation was less extensive in the presence of As(V) (Table S3). In the Mn(II) + bicarbonate column, two C were detected at a distance of 3.02 Å, indicating the presence of rhodochrosite cross-validating results from XRD (Figure S9).

Figure 3.

Mn K-edge XANES spectra of treatments with Mn(II) added. Solid lines represent measured data and dashed lines the best linear combination fit using the three standards shown as solid gray lines.

Table 3.

Manganese XANES Linear Combination Fitting (LCF) Results^a

	proportion of LCF fit
	MnIISO4(aq)	β- MnIIIOOH	(Fe,MnIII) SiO3	AOS	χ 2
Mn	0.06	0.76	0.19	2.94	3.2x10−4
Mn + HCO3	b.d.b	0.95	b.d.	2.96	7.3x10−5
Mn + As	0.39	0.32	0.29	2.62	2.8x10−4
Mn + As +HCO3	0.20	0.62	0.19	2.80	1.2x10−4

^aAverage oxidation state (AOS) was calculated using the relative proportion of each standard, but the reference standards used should only be interpreted in terms of Mn oxidation state in this analysis.³⁸

^bbelow detection (<0.05).

X-ray Photoelectron Spectroscopy.

Analysis of particle surfaces using XPS provided information on the oxidation state of Fe, Mn, and As. In all samples the binding energy of the Fe 2p_3/2 peak was ~710 eV, indicating the presence of only Fe(III) in the system (Supporting Information Figure S3). Manganese XPS data show the presence of Mn(III) at the surface of the Mn + bicarbonate treatment based on Mn 2p_3/2 peak centered at 641.6 eV (Supporting Information Figure S4). Treatments containing Mn(II) and As(V) had a 2p_3/2 peak shifted 0.6 eV lower to 641.0 eV, which was consistent with Mn K-edge XANES data (Figure 3) showing more reduced Mn in treatments containing As as compared to Mn + bicarbonate. The Mn signal in the column with only Mn was below the resolution of XPS (<1000 mg kg⁻¹). Arsenic 3d XPS spectra contained a single peak at 45 eV indicating the presence of As(V) at particle surfaces (Supporting Information Figure S5), which was in agreement with bulk As XANES data (Figure 2). The binding energy between 44.8 and 45.0 eV is indicative of AsO₄³⁻ species³⁷ and is in agreement with equilibrium speciation calculations (PHREEQC) using hydrous ferric oxide strong and weak sites that show the Hfo_(s,w)OHAsO₄³⁻ species is ~3 orders or magnitude more abundant than any other surface As species under the range of experimental conditions.

DISCUSSION

Homogeneous oxidation of Mn(II) by O₂ is thermodynamically favorable under common environmental conditions but is kinetically limited at pH <8.^39-42 In the absence of favorable Mn(II) mineral precipitation or surface catalysts, aqueous Mn(II) can persist in oxic water for years.⁴⁰ Field data indicate that Mn(II) oxidation may be limited even during oxidizing events (Figure S1). Further, bicarbonate produced during anoxic periods enhances Mn retention through precipitation of MnCO₃ but competes with As(V) for surface adsorption sites and decreases As(V) retention to ferrihydrite, producing contrasting effects on Mn and As solubility.

Manganese Retention and Oxidation.

Manganese retention was lowest when only Mn(II) was added (Table 1) and solid-associated Mn had an average oxidation state of 2.94 based on Mn K-edge XANES (Figure 3 and Table 3). Manganese AOS of ~3 is consistent with previous results showing the oxidation of Mn(II) catalyzed by ferrihydrite in the presence of O₂.³⁹ Similar oxidation of Mn(II) is also catalyzed by lepidocrocite and goethite.⁴¹

The presence of As(V) and bicarbonate each led to increased Mn retention (Table 1 and Figure 4), but resulted in different Mn solid-phase products. When Mn(II) and bicarbonate were added together, Mn AOS was 2.96, which was similar to Mn(II) addition alone. Saturation index calculations indicate MnCO₃ precipitation was favorable, and comparison of XRD patterns for solids exposed to Mn(II) with bicarbonate shows formation of a small peak with d-spacing of ~2.84 Å coinciding with the primary peak for MnCO₃ (Figure S9).

Figure 4.

Dominant processes affecting As and Mn retention under each experimental condition. Brown color represents iron oxides (Fe(OH)₃), gray parts are quartz sand on which ferrihydrite is coated at the start of the experiment, and pink represents rhodochrosite (MnCO₃).

Addition of Mn(II) with As(V) doubled Mn retention compared to Mn(II) alone (Figure 4 and Table 1). However, Mn AOS was 2.62 (Figure 3 and Table 3), lower than Mn(II) alone (2.94) or Mn(II) with bicarbonate (2.96). An AOS less than 3 indicated the presence of solid-associated Mn(II) and suggests that As(V) inhibited Mn(II) (re)oxidation relative to Mn(II) addition alone or with bicarbonate. Manganese(II) arsenate minerals (e.g. krautite, MnHAsO₄·H₂O,⁴³ and Mn₃(AsO₄)₂·8H₂O) remained undersaturated even when calculating the ion activity product (IAP) using input (highest) concentrations of Mn(II) and As(V). Therefore, it is unlikely that direct Mn^II–As mineral precipitation leads to the observed increase in Mn retention in the presence of As(V), and no evidence for Mn–As minerals was found in spectroscopic or diffraction data. Instead, increased Mn retention combined with lower Mn AOS could be explained by a combination of (i) As(V) blocking ferrihydrite surface sites needed to catalyze Mn(II) oxidation, (ii) formation of bridging complexes similar to FeO–Mn²⁺–HAsO₄²⁻ where As(V) could block O₂ from oxidizing Mn(II) at the ferrihydrite surface and (iii) regeneration of Mn(II) due to oxidation of trace As(III) or other reduced species by Mn(III).⁴⁴ However, in the case of (iii) Mn(III) reduction would be in competition with O₂ or reactive Fe(III) phases which have been shown to oxidize As(III).⁴⁵

When Mn(II) was added in the presence of both As(V) and bicarbonate, Mn retention decreased by ~20% relative to the Mn(II) + bicarbonate treatment (Figure 4 and Table 1) and Mn AOS was 2.80, which is at the midpoint between treatments of Mn(II) added with As(V) and bicarbonate alone (Table 3). Manganese retention and AOS are consistent with Mn(III) product formation in the absence of As(V) as well as the stabilization of Mn(II) against (re)oxidation.

Arsenic Retention.

Arsenic sorption was typical for column flow experiments using ferrihydrite sand and sorption densities were similar to previous studies⁴⁶ (0.045 mol As (mol Fe)⁻¹). The presence of only As(V) species (Figure 2 and Figure S5) confirms the columns remained oxic throughout the course of the experiment.

Addition of bicarbonate to columns with As(V) resulted in faster breakthrough (Figures 1 and 3) and lower As solid-phase retention (Figure 4 and Table 1) likely due to two factors: (i) a slight increase in pH even in the presence of PIPES buffer and (ii) bicarbonate competition for surface adsorption sites.^22,23 Arsenate sorption capacity on ferrihydrite is sensitive to pH changes and an increase in pH would lead to lower retention capacity.⁴⁷ Other studies performed in batch conditions have found only slight effects of bicarbonate on As(V) sorption to ferrihydrite.⁴⁸ A combination of these factors likely explains the decrease in As(V) retention in the presence of bicarbonate.

Co-addition of Mn(II) with As(V) increased As(V) retention compared to As(V) alone both on a total As mass basis and when normalized by Fe content (Table 1). Arsenate retention to phyllosilicate minerals is enhanced through divalent cation-bridging compounds (e.g. Ca²⁺ and Mg²⁺),⁴⁹ therefore it is also reasonable to expect that Mn²⁺ could provide a similar effect at negatively charged sites on the ferrihydrite or exposed SiO₂ surface. Manganese L-edge XAS spectra of columns with Mn(II) and As(V) have a more pronounced peak at 639.7 eV compared to spectra of Mn(II) with carbonate or Mn(II) alone (Figure S7), indicating that the presence of As(V) contributes both to retention of total Mn (Table 1) as well as stabilization of Mn(II). Stabilization of Mn(II) is further supported by XPS data that show a shift to lower energy of the 2p_3/2 in columns with Mn(II) and As(V) compared to the column with Mn(II) + carbonate (Figure S4) and the lower AOS observed in Mn K-edge XANES (Figure 3). Taken together, these data indicate that As(V) stabilizes Mn(II) at particle surfaces relative to experiments without As(V).

Analysis of As EXAFS spectra showed adsorbed arsenate with an average As–O bond distance of 1.67 Å in all samples (Figure 2 and Table 2). Sherman and Randall⁵⁰ showed using density functional theory calculations that the bidentate binuclear corner-sharing complex (²C) is more stable than the bidentate binuclear edge-sharing complex (²E), and EXAFS results from all columns with As(V) yielded an As–Fe bond distance in agreement with a ²C complex (~3.30 Å). Ratios of As/Fe similar to this study (0.045 mol As/mol Fe) showed a combination of monodentate mononuclear corner-sharing (¹V) complexes and ²C complexes when ferrihydrite dominated a ferrihydrite-goethite biphasic mineral.⁵¹ Therefore, it is reasonable to expect both ¹V and ²C complexes to form under the experimental conditions, however, inclusion of a ¹V complex As–Fe distance in modeling of the EXAFS data did not improve the fits. The way in which arsenate is reacted or added to ferrihydrite may impact the formation of ¹V and ²C complexes. In this study relatively low concentrations of arsenate were added to columns over nearly a year which limited local As(V) concentrations in contrast to coprecipitation experiments where ¹V complexes have been observed at nearly identical As/Fe loadings.⁵¹

ENVIRONMENTAL IMPLICATIONS

Groundwaters in Bangladesh with high concentrations of dissolved Mn and low dissolved Fe and As have been previously explained with thermodynamic “redox buffering” by Mn-oxides.⁵² Field data from China indicate that redox buffering is not a controlling mechanism in a redox oscillating environment because Mn_aq concentration increases along with Fe(II) and As_aq during reducing periods (Figure S1), indicating Fe(III) reduction occurs even in the presence of oxidized Mn. The stability of Mn_aq in the presence of O₂ provides a mechanism for separation of dissolved Mn and As in groundwater¹⁹ as well as opportunities for interactions between Mn, As, and bicarbonate during redox transition periods. Bicarbonate and Mn(II) are both produced during reducing conditions but remain stable during oxic periods; however, they play contrasting roles in As(V) retention to ferrihydrite during oxic periods.