Christopher H.
Gammons
a and
Angela K.
Frandsen†
b
aMontana Tech of the University of Montana, Department of Geological Engineering, Butte, Montana 59701. E-mail: cgammons@mtech.edu
bMontana Tech of the University of Montana, Department of Environmental Engineering, Butte, Montana 59701
The aqueous geochemistry of Zn, Cu, Cd, Fe, Mn and As is discussed within the context of an anaerobic treatment wetland in Butte, Montana. The water being treated had a circum-neutral pH with high concentrations of trace metals and sulfate. Reducing conditions in the wetland substrate promoted bacterial sulfate reduction (BSR) and precipitation of dissolved metal as sulfide minerals. ZnS was the most common sulfide phase found, and consisted of framboidal clusters of individual spheres with diameters in the submicron range. Some of the ZnS particles passed through the subsurface flow, anaerobic cells in suspended form. The concentration of “dissolved” trace metals (passing through a 0.45 µm filter) was monitored as a function of H2S concentration, and compared to predicted solubilities based on experimental studies of aqueous metal complexation with dissolved sulfide. Whereas the theoretical predictions produce “U-shaped” solubility curves as a function of H2S, the field data show a flat dependence of metal concentration on H2S. Observed metal concentrations for Zn, Cu and Cd were greater than the predicted values, particularly at low H2S concentration, whereas Mn and As were undersaturated with their respective metal sulfides. Results from this study show that water treatment facilities employing BSR have the potential to mobilize arsenic out of mineral substrates at levels that may exceed regulatory criteria. Dissolved iron was close to equilibrium saturation with amorphous FeS at the higher range of sulfide concentrations observed (>0.1 mmol H2S), but was more likely constrained by goethite at lower H2S levels. Inconsistencies between our field results and theoretical predictions may be due to several problems, including: (i) a lack of understanding of the form, valence, and thermodynamic stability of poorly crystalline metal sulfide precipitates; (ii) the possible influence of metal sulfide colloids imparting an erroneously high “dissolved” metal concentration; (iii) inaccurate or incomplete thermodynamic data for aqueous metal complexes at the conditions of the treatment facility; and (iv) difficulties in accurately measuring low concentrations of dissolved sulfide in the field.
![]() | (1) |
![]() | (2) |
where Me = any divalent metal cation. Although reaction (1) produces 2 moles of bicarbonate alkalinity, the acidity produced by reaction (2) tends to offset this, such that protons are conserved in the overall reaction:
![]() | (3) |
However, in a treatment wetland or SRB bioreactor, it is often the case that more H2S is generated via reaction (1) than is consumed via precipitation of sulfide minerals. Such systems see a net increase in alkalinity, and often produce considerable quantities of excess H2S.9 The possible adverse affects of excess H2S have not received much attention in the engineering literature to date. Besides being a highly toxic substance, H2S is known to form stable aqueous compounds with many heavy metals. A good example is cadmium, for which the following general solubility reaction may be written:
![]() | (4) |
Fig. 1 shows the calculated solubility of amorphous CdS as a function of H2S concentration, based on equilibrium constants from Wang and Tessier10 for reaction (4), where n = 0 to 4. The “U-shaped” solubility dependence for CdS is very similar to the amphoteric relationship between metal oxide solubility and pH, which is much more familiar to water treatment engineers. Just as there is an optimal pH range for maximum metal removal during lime treatment of mining-impacted waters, there is an optimal H2S concentration for metal removal via sulfide treatment. Increases in H2S above this optimal level could theoretically result in enhanced metal solubility, compromising the performance of the treatment facility.
![]() | ||
Fig. 1
The solubility of amorphous CdS as a function of H2S concentration at pH = 7, 25![]() |
The behavior of iron and trace metals in low temperature H2S-rich waters has been extensively studied in natural environments, such as anoxic seawater,11 marine sediments,12,13 salt marshes and estuaries,14,15 and fresh water lakes.16–19 Although constructed wetlands are analogous in many respects to natural H2S-rich systems, they differ from the point of view that the waters being treated often have much higher metal concentrations. This is certainly true for the Wetlands Demonstration Project 1 (WDP1) site in Butte, Montana, which is the focus of this study. Although built as a treatment facility, data from WDP1 provide insight into the chemical, physical, and biological processes which interplay in low temperature, sulfidic waters. The closest natural analog to the WDP1 facility would be a wetland in a temperate climate that receives loadings of metals and sulfate, either due to anthropogenic impacts (e.g., acidic mine drainage), or as a natural consequence of weathering of sulfide-rich ore bodies. Such wetland systems are rare, but have been documented in the literature.20–23 The results of this study are also directly applicable to other treatment operations employing H2S, such as constructed wetlands, bioreactors, or reactive barriers.
![]() | ||
Fig. 2 Photograph of Butte's Metro Storm Drain, taken in November 1997. Note the high turbidity (despite low flow and no precipitation for several weeks), due to abundance of suspended iron oxy-hydroxide particles. |
WDP1 operated continuously for 3 years, from May 1996 through December 1998. The facility consisted of seven treatment cells with different dimensions and substrate compositions (Fig. 3, Table 1). MSD water was first pumped into a surge pond (Cell 0), which allowed settling of suspended particles, and provided hydraulic head and water storage for the rest of the facility. Water exiting the surge pond passed by gravity through one of four parallel, subsurface-flow, anaerobic cells (cells 1, 2, 3, 4). The main purpose of the anaerobic cells was to provide a suitable environment for sulfate-reducing bacteria (SRB), with subsequent precipitation of heavy metals as insoluble sulfide minerals. The anaerobic cells were lined on the bottom with HDPE, were filled with ca. 0.5 inch river gravel and limestone fragments, and (with the exception of cell 3) were planted at the surface with cattails (Typha latifolia) as a renewable source of organic carbon. A composted mixture of sewage sludge and wood chips was added to the substrates of cells 2 and 3 to further encourage SRB growth. Water exiting the anaerobic cells flowed into one of two aerobic polishing cells (cells 6 and 7), which consisted of a series of shallow pools and riffles. The purpose of these cells was to re-aerate the water, oxidize any excess H2S, and hopefully remove residual iron and manganese as oxy-hydroxide phases. The entire WDP1 facility treated between 15 and 30 gpm of MSD water during the 3 years of operation.
![]() | ||
Fig. 3 Schematic diagram of the Wetlands demonstration project 1 (not to scale). See Table 1 for further specifications. |
Cell | Description | Specifications | HRTa and design flow |
---|---|---|---|
a HRT = hydraulic residence time. b gpm = gallons per minute (1 gallon = 4.54609 dm3). | |||
0 | Settling pond | Area = 7000 ft2 (1 foot = 0.3048 m), volume = 40![]() |
10 days and 20 gpmb |
1 | Anaerobic wetlands | Horizontal subsurface flow volume = 30![]() |
9.4 days and 5 gpm |
2 | Anaerobic wetlands | Horizontal subsurface flow volume = 20![]() |
6.2 days and 5 gpm |
3 | Anaerobic wetlands | Upwards subsurface flow volume = 12![]() |
3.7 days and 5 gpm |
4 | Anaerobic wetlands | Horizontal subsurface flow volume = 15![]() |
4.8 days and 5 gpm |
6 | Aerobic wetlands | Shallow ponds and riffles area ca. 2000 ft2 | Ca. 2 days and 10 gpm |
7 | Aerobic wetlands | As for cell 6, area ca. 6500 ft2 | Ca. 6 days and 10 gpm |
All metal samples were acidified to 2% HCl or HNO3 using Fisher Trace Metal Grade acid, and were refrigerated prior to analysis. The samples were analyzed for the elements As, Al, Ca, Cd, Cr, Co, Cu, Fe, S, P, Pb, Ni, K, Mg, Mn, Na, Si, and Zn, using a Perkin Elmer Optima 3000 DV ICP-AES spectrometer. Analytical protocol for this instrument followed SW-846 method 6010B, inorganic analysis by ICP-AES and EPA 600 method 200.7, inorganic analysis by inductively coupled plasma analysis. Filtration was performed in the field using disposable 0.45 µm CAMEO 25GA acetate-plus membrane syringe filters with 1 µm glass prefilters. The inorganic anions fluoride, chloride, nitrite, bromide, nitrate, orthophosphate, and sulfate were determined using a Dionex System DX-500 ion chromatograph, following EPA Method 300.0. Dissolved sulfide was determined using the Methylene Blue method (HACH method 8131), and a HACH DR-2010 portable spectrometer. Samples collected for sulfide analysis were obtained directly with a syringe from the low volume flow cell during pumping. The samples were stabilized in the field by filtration to 0.22 µm and immediate transfer to disposable cuvettes containing the methylene blue reagent. Duplicate sampling indicated minimal loss of H2S occurred due to volatilization or oxidation during sampling and prior to analysis.
During the last year of operation, each internal piezometer in cells 1, 2 and 4 was equipped with a small cup near the base of the sampling tube. These cups collected solid particles that passed through the piezometer screens during pumping. The cups were sampled in November 1998, and the solids were stored in tightly sealed glass vials filled completely with original site water, and placed in a refrigerator. The solids were examined periodically in 1999 and early 2000 by SEM and XRD, using equipment in the Department of Metallurgical Engineering, Montana Tech, and the Instrumental and Chemical Analysis Laboratory (ICAL), Montana State University.
Before discussing each metal–sulfide system in detail, a brief summary of seasonal changes in selected water quality parameters at WDP1 is presented. Aspects of the overall performance of the WDP1 facility have been summarized elsewhere.9,26–30 Additional details may be found in Montana Tech MSc theses.31–37
Parameter | Summer (8/17/98) | Winter (2/19/98) | ||||
---|---|---|---|---|---|---|
Influent | Effluent | Removal efficiency | Influent | Effluent | Removal efficiency | |
a Heavy metals = µg L−1, filtered (0.45 µm); alkalinity = mg CaCO3 L−1. b DIC = dissolved inorganic carbon. | ||||||
T/°C | 19.0 | 19.1 | 2.0 | 2.7 | ||
pH | 6.80 | 6.84 | 6.42 | 6.73 | ||
SC/µS cm−1 | 872 | 1070 | 821 | 897 | ||
Alkalinity | 66.3 | 178 | 72 | 149 | ||
DICb/mg C L−1 | 16.1 | 54.4 | 18.3 | 36.4 | ||
Eh/mV | 476 | −153 | 493 | 76 | ||
As | 8.6 | 11.9 | (−38%) | 7.3 | 5.2 | 28.8% |
Cd | 40.5 | 0.51 | 98.7% | 32.7 | 0.52 | 98.4% |
Cu | 123 | 3.6 | 97.0% | 216 | 1.0 | 99.5% |
Fe | 28.8 | 18.0 | 37.5% | 171 | 528 | (−209%) |
Mn | 6680 | 3640 | 45.5% | 7070 | 8080 | (−14%) |
Zn (filtered) | 9330 | 10.5 | 99.9% | 9950 | 101 | 99.0% |
Zn (unfiltered) | 9850 | 42.9 | 99.6% | 10,000 | 640 | 93.6% |
Sulfate/mg SO4 L−1 | 472 | 289 | 38.8% | 336 | 313 | 6.8% |
H2S/mg H2S L−1 | 0.0 | 34.1 | 0.0 | 2.0 |
Based on the observed decreases in dissolved sulfate concentration, Gammons et al.9 estimated a maximum mid-summer BSR rate of ca. 0.4 millimoles of H2S produced per liter of water per day. Of the total H2S produced, a significant percentage was immobilized as sulfide minerals, an unknown fraction was volatilized to the air, and the remainder exited the cells in dissolved form, or as colloidal particles of elemental sulfur. Machemer et al.1 conducted a more detailed S mass balance for a similar treatment wetland in Colorado, and concluded that loss of H2S due to volatilization was minor (<1% of the total sulfide generated). Nonetheless, the WDP1 facility did smell foul most of the year, indicating potentially significant H2S discharges to the atmosphere.
Dissolved sulfide concentrations of cell 2 effluent waters were typically >1 mg L−1, sometimes as high as 30 mg L−1. In the winter months, BSR rates and effluent H2S concentrations were approximately ten times lower than in summer, and most of the observed increases in alkalinity and DIC in winter have been attributed to dissolution of limestone in the substrate.9 In cells 6 and 7, excess H2S leaving the anaerobic cells in dissolved form was incompletely oxidized to elemental sulfur by a combination of white and purple bacteria (Fig. 4). This phenomenon caused water in the down-gradient aerobic polishing ponds to take on the appearance of dilute skim milk (Fig. 5).
![]() | ||
Fig. 4 Photograph of purple and white S-oxidizing bacteria near the outlet distribution pipe for cells 2 and 3. The purple bacteria are incompletely oxidizing H2S to elemental sulfur. Similar growths of purple bacteria have been reported for H2S-rich effluent waters at other constructed wetlands (James Gusek, personal communication, 1999). Although bacterial identification was not attempted in this study, it is likely78 that the purple microbes include the Chromatiaceae and possibly Rhodospirillaceae, whereas an anonymous reviewer of this article suggested the filamentous white bacteria could be Beggiatoacea. |
![]() | ||
Fig. 5 Photograph of the aerobic polishing pond cell 7, looking upgradient from the outlet. Photo taken in November 1997. Note the turbid, milky white color imparted by suspended particles of elemental sulfur. |
Removal of Cd and Cu was excellent year-round in all of the anaerobic cells. The most likely attenuation mechanism was precipitation of insoluble sulfide minerals. Sulfide minerals of cadmium and copper have exceedingly low solubility products;38 consequently, these metals would have been immediately scavenged by H2S as it was formed by BSR. Geochemical modeling indicates that cadmium could also have been attenuated as impurities in calcite (i.e., CdCO3–CaCO3 solid solution). There is a strong thermodynamic driving force for cadmium to be scavenged to very low levels by carbonate minerals.39,40 The reductive nature of the substrates could also have induced precipitation of Cu2+ as Cu2O (cuprite) or metallic Cu. Metallic copper nodules have been described in natural wetlands in alpine watersheds in Montana that received pre-historic acid rock drainage.41,42 However, in the presence of measurable quantities of H2S, it is certain that any copper or cadmium previously deposited in a carbonate, oxide or elemental state would have quickly tarnished to a sulfide form.
Zinc removal efficiency varied widely from cell to cell, and from season to season, with especially poor performance during the first winter (1996–7). Zinc is thought to have been removed from solution primarily as ZnS, with low removal efficiencies coinciding with decreased hydraulic residence time and low BSR rates.9 Zinc attenuation was more erratic than copper or cadmium, partly because influent Zn concentrations were much higher (ca. 10 mg L−1), and partly because the solubility product of ZnS is greater than that of CdS, CuS or Cu2S.38 Manganese removal was nonexistent in winter, although partial attenuation was noted in the anaerobic cells during the summer months. The most likely mechanism for manganese removal in summer was co-precipitation with ZnS (see below), although some Mn may also have been scavenged as impurities in calcite.
Dissolved iron and arsenic levels were generally low in the oxidized influent waters, although particulate Fe and As were sometimes elevated.30 Slight increases in dissolved Fe and As were noted as the treatment water passed through the anaerobic cells, presumably due to reductive dissolution of ferric oxy-hydroxide compounds and concomitant release of adsorbed arsenic. This was particularly true during the first few months of operation, when concentrations of dissolved arsenic leaving the anaerobic cells were consistently highly elevated (>50 µg L−1).
Adsorption of metals onto organic and inorganic phases in the substrate could have played an important role in the first few months of operation of WDP1. However, extrapolation of the results of adsorption isotherm experiments conducted on substrate materials35,36 indicate that available adsorption sites probably became saturated in the first 6 months of operation, due to the continuous metal loadings. A similar conclusion was reached for a constructed wetland receiving metal-rich acid mine drainage in Colorado.2 Likewise, chemical analysis of cattail tissues collected after one year of operation suggests that metabolic uptake of metals by Typha latifolia was minimal compared to the total metal loadings.
Sulfidic sediments collected in the sampling cups of the internal piezometers from cell 2 were examined by SEM-EDX at the ICAL laboratory, Montana State University. The samples were prepared by pipeting a tiny amount of water-saturated sediment onto a silicon disk, and immediately evacuating with application of a carbon coating. The abundance of detrital mineral fragments and organic material made the process of finding sulfide particles for SEM-EDX analysis somewhat difficult. Best results were obtained by scanning the samples at low power in electron backscatter mode (see also Morse and Cornwell43). Clusters of metal sulfide particles appeared as bright spots, due to their higher average atomic mass. Fig. 6 is a backscatter image revealing the distribution of metal sulfide particles on the surface of a sample collected near the outlet to cell 2. This particular field of view had a higher density of sulfide particles than average: nonetheless, sulfide particles were common, and were found in all of the samples examined.
![]() | ||
Fig. 6 SEM backscatter image of particles collected near the outlet to cell 2, approximately 4 ft below the surface. The bright spots have a higher mean atomic mass, and are mostly ZnS particles. The darker regions are a mix of organic and mineral matter inherited from the cell substrates. |
Upon closer examination, the bright spots in Fig. 6 were found to be mainly zinc sulfide particles. Close-up SEM photographs of the ZnS particles are shown in Fig. 7a and 7b. Most of the clusters were on the order of a few tens of micrometers in diameter, and were composed of loosely aggregated spheres approximately 1 µm in diameter, with some larger and some smaller. Farrand44 produced similar synthetic CuS, FeS and ZnS precipitates, which consisted of spherical particles with diameters from 0.05 to 0.2 µm. The WDP1 precipitates are larger than this. Because the samples were stored for nearly 1 year before SEM examination, it is likely that some ripening occurred, and that the precipitates were initially finer grained. It is significant to note that particles in the 0.05 to 0.4 µm range could have passed through the 0.45 µm membrane filters used in this study to collect water samples.
![]() | ||
Fig. 7 SEM photomicrographs of ZnS particles collected from cell 2. Note 1 micron bar for scale. |
The approximate composition of the ZnS particles was determined by semi-quantitative EDX analysis, and is summarized in Table 3. Atomic ratios of (Zn + Mn + Fe)/S, Fe/Zn, and Mn/Zn were calculated for each particle. The average ratios were 0.93 ± 0.21, 0.11 ± 0.057, and 0.030 ± 0.0062, respectively (errors represent one standard deviation). The average composition of a zinc sulfide particle from cell 2 was therefore (Zn0.87Fe0.10Mn0.03)S. These data indicate that significant and consistent amounts of iron and manganese are present as impurities in the zinc sulfide particles. This is the most likely reason for the aforementioned manganese attenuation in the wetland cells observed in the summer months.
Sample | Atomic%a | |||
---|---|---|---|---|
S | Zn | Mn | Fe | |
a Compositions normalized to 100%. b FeSx particle, (Zn + Fe)/S≈1 ∶ 1.6. c FeSx particle, pyrite framboid in Fig. 8, (Fe + Zn + Mn)/S≈1 ∶ 2.5. | ||||
2B2L #1 | 57.0 | 37.1 | 1.3 | 4.6 |
2B2L #2 | 55.3 | 38.2 | 1.3 | 5.2 |
2D1M #4 | 55.1 | 39.1 | 1.2 | 4.6 |
2D1XL #1 | 55.9 | 40.3 | 1.1 | 2.7 |
2D1XL #2 | 52.7 | 43.4 | 1.1 | 2.8 |
2D1XL #3 | 40.9 | 51.9 | 1.6 | 5.6 |
2D1XL #5 | 52.1 | 44.1 | 1.5 | 2.3 |
2D1XL #6 | 55.4 | 34.8 | 1.2 | 8.6 |
2D1XL #7 | 48.2 | 44.7 | 0.8 | 6.3 |
2D1XL #8 | 51.5 | 43.7 | 1.6 | 3.2 |
2D1M Fe #1b | 61.8 | 5.1 | — | 33.1 |
2D1M Fe #2c | 71.2 | 0.3 | 1.3 | 27.2 |
Phase relations in the Zn–Fe–S system45 indicate that natural sphalerite grains in equilibrium with stoichiometric FeS should contain approximately 20 mole% FeS. The lower % of Fe in precipitates in WDP1 could be a function of higher redox potentials (FeS-unstable), or could reflect disequilibrium. For example, the high Zn/Fe ratio of influent waters may have locally precluded saturation with a separate Fe-sulfide phase. This is consistent with our observation that Fe-sulfide particles were rare compared to the much more abundant ZnS particles.
Fig. 8 is an SEM photograph of an Fe-sulfide framboid from cell 2. The composition of this particle (Table 3) suggests that it is pyrite. Very similar pyrite framboids have been described from synthesis experiments, as well as from natural environments (e.g., Black Sea sediments).46,47 Following the reasoning of Wilkin and Barnes,47 it is likely that the initial Fe-sulfide precipitate in WDP1 was amorphous or crystalline FeS (mackinawite), and that the monosulfide subsequently transformed to pyrite over a period of days to months. This hypothesis is supported by metal solubility constraints (see below). Semi-quantitative EDX analysis indicates that the Fe-sulfide particles contained significant amounts of Mn and Zn (Table 3), although the paucity of data and scatter in the results preclude determination of an average composition. Experimental studies have shown that Mn and Zn can partition into Fe-monosulfide in amounts consistent with our observations.48,49
Source | Solid | T/°C | Reactions | Log K |
---|---|---|---|---|
a Adjusted from S2− using Gubeli’s53 dissociation constant: HS− = H+ + S2−, log K = −13.48. b Adjusted using the reaction from Shea and Helz:58 4S(s) + HS− = S52− + H+, log K = −9.44. c Adjusted using the reaction from Shea and Helz:58 3S(s) + HS− = S42− + H+, log K = −9.56. d Adjusted from S2− using Ste. Marie’s62 dissociation constant: HS− = H+ + S2−, log K = −13.48. | ||||
65 | Amorphous FeS | 20 | FeS + H+ = Fe2+ + HS− | −3.00 |
FeS + H+ + HS− = Fe(HS)2° | +3.45 | |||
51 | Crystalline sphalerite | 25 | ZnS + H2S° = Zn2+ + 2HS− | −18.47 |
ZnS + H2S° = Zn(HS)2° | −5.65 | |||
ZnS + 2HS− = ZnS(HS)22− | −5.33 | |||
ZnS + H2S° + 2HS− = Zn(HS)42− | −3.83 | |||
ZnS + HS− = ZnS(HS)− | −4.64 | |||
52 | Crystalline sphalerite | 25 | ZnS + H2So = Zn(HS)2° | −5.3 |
ZnS + H2So + HS− = Zn(HS)3− | −3.3 | |||
ZnS + H2So + 2HS− = Zn(HS)42− | −3.4 | |||
ZnS + H2O + HS− = Zn(OH)(HS)2− | −4.4 | |||
ZnS + H2O + 2HS− = Zn(OH)(HS)32− | −4.9 | |||
53 | Amorphous ZnS | 25 | ZnS + H+ = Zn2+ + HS− | −10.89a |
ZnS + H2O = Zn(OH)(HS)° | −5.87 | |||
54 | Crystalline covellite, CuS | 25 | CuS + H+ = Cu2+ + HS− | −20.95 |
CuS + 2HS− = CuS(HS)22− | −4.97 | |||
CuS + 3HS− = CuS(HS)33− | −4.04 | |||
CuS + 4S(s) + HS− = CuS(S5)2− + H+ | −12.19b | |||
CuS + 6.5S(s) + 1.5HS− = CuS4S53− + 1.5H+ | −15.71bc | |||
CuS + 7.5S(s) 1.5HS− = Cu(S5)23− + 1.5H+ | −16.06bc | |||
55 | Crystalline chalcocite, Cu2S | 22 | Cu2S + H+ = 2Cu+ + HS− | −34.62 |
0.5Cu2S + 1.5HS− + 0.5H+ = Cu(HS)2− | −0.13 | |||
Cu2S + 2HS− = Cu2S(HS)22− | −4.75 | |||
10 | Crystalline greenockite (gr) and two amorphous precipitates | 25 | CdS(gr) + H+ = Cd2+ + HS− | −14.82 |
(a1 and a2) | CdS(a1) + H+ = Cd2+ + HS− | −14.40 | ||
CdS(a2) + H+ = Cd2+ + HS− | −14.15 | |||
Cd2+ + HS− = Cd(HS)+ | +7.38 | |||
Cd2+ + 2HS− = Cd(HS)2° | +14.43 | |||
Cd2+ + 3HS− = Cd(HS)3− | +16.28 | |||
Cd2+ + 4HS− = Cd(HS)42− | +18.43 | |||
63 | Crystalline greenockite | 25 | CdS + H+ = Cd2+ + HS− | −14.36 |
Cd2+ + 3HS− = Cd(HS)3− | +16.44 | |||
Cd2+ + 4HS− = Cd(HS)42− | +17.89 | |||
Cd2+ + HS− + H2O = CdOHS− + 2H+ | −2.47 | |||
62 | Amorphous CdS | 25 | CdS + H+ = Cd2+ + HS− | −12.28d |
Cd2+ + HS− = Cd(HS)+ | +7.55 | |||
Cd2+ + 2HS− = Cd(HS)2° | +14.61 | |||
Cd2+ + 3HS− = Cd(HS)3− | +16.49 | |||
Cd2+ + 4HS− = Cd(HS)42− | +19.35 |
Source | Solid | T/°C | Reactions | Log K |
---|---|---|---|---|
67 | Crystalline orpiment | 22 | As2S3 + 6H2O = 2H3AsO3° + 3H2S | −25.8 |
68 | Crystalline orpiment | 22 | As2S3 + H2O = H2As2S3O° | −6.45 |
As2S3 + HS− = HAs2S4− | −1.47 | |||
As2S3 + OH− + HS− = As2S42− + H2O | +4.05 | |||
69 | Crystalline orpiment | 25 | 0.5As2S3 + 3H2O = H3AsO3° + 1.5HS− + 1.5H+ | −23.11 |
1.5As2S3 + 1.5HS− + 0.5H+ = H2As3S6− | +3.61 | |||
70 | Amorphous As2S3 | 25 | 0.5As2S3 + 3H2O = H3AsO3° + 1.5H2So | −11.9 |
1.5As2S3 + 1.5H2S° = H2As3S6− + H+ | −5.0 | |||
71 | Amorphous As2S3 | 25 | 0.5As2S3 + 3H2O = H3AsO3° + 1.5HS− + 1.5H+ | −22.38 |
1.5As2S3 + 1.5HS− = As3S5(SH)2− + 0.5H+ | +1.78 | |||
1.5As2S3 + 1.5HS− = 1.5H+ = HAs3S4(SH)2° | +14.74 |
For each sample, bisulfide concentration was computed from the measured pH and total dissolved sulfide values using published values for the dissociation of H2S. Activity coefficients for dissolved ionic compounds were calculated with MINTEQA2,50 using the Davies version of the extended Debye–Huckel equation. The ionic strength of water at WDP1 was approximately constant at ca. 0.016 (mol kg−1). For these conditions, the computed activity coefficients were 0.88, 0.61, and 0.32, for monovalent, divalent, and trivalent ions, respectively. Uncharged aqueous species were assumed to have an activity coefficient of 1.0.
In the diagrams that follow, results are shown for all analyses in which paired H2S and filtered metal concentrations were taken at WDP1. Attempts to separate the database according to individual cells showed no meaningful trends, other than the fact that cells 2 and 3 (the cells with compost added) tended to produce more H2S, whereas cells 1 and 4 tended to have lower H2S concentrations. For reference, the average dissolved metal concentration of the anaerobic cell influent (i.e. the surge pond effluent), is also superimposed on each figure.
![]() | (5) |
![]() | ||
Fig. 9 Comparison of concentrations of filtered zinc at WDP1 vs. predicted values based on solubility models from literature sources. |
Both of these solubility studies were based on crystalline sphalerite, and appear to underestimate zinc mobility in the WDP1 wetlands, especially at bisulfide concentrations <1 mmol. The two curves labeled MINTEQ in Fig. 9 show the predicted solubility of crystalline (cr) and amorphous (am) ZnS, as predicted using the MINTEQ database. The MINTEQ estimates have a similar shape to those of ref. 50 and 51, with amorphous ZnS being ca. 2.5 log units more soluble than crystalline ZnS. Gubeli and Ste. Marie53 measured the solubility of freshly precipitated (amorphous) ZnS as a function of dissolved sulfide (0.004 to 0.01 molar) and pH (1 to 14). ZnS solubility was independent of both variables at pH > 3, being constant at ca. 1.3 × 10−6 mol kg−1. To explain these observations, Gubeli and Ste. Marie proposed the following reaction:
![]() | (6) |
Of the previous studies investigated, the flat nature of the WDP1 data at sulfide concentrations >10−6 mol kg−1 is best matched by the results of Gubeli and Ste. Marie.53 Both data sets are independent of bisulfide concentration, and have comparable solubility values. However, Daskalakis and Helz51 have argued that the constantly elevated Zn concentrations measured by Gubeli and Ste. Marie were due to the presence of colloidal ZnS particles. It is possible that colloidal ZnS particles could have passed through the filters used by Gubeli and Ste. Marie (filter size not specified), and also the samples taken in this study (0.45 µm filter). This idea is discussed in greater detail below.
![]() | (7) |
![]() | (8) |
Reaction (7) has a calculated free energy change of −3.75 kcal mol−1,56 indicating that covellite is the stable copper sulfide in the presence of elemental sulfur. In the absence of sulfur, the relative stability of the two phases depends on the Eh, pH and H2S concentration (reaction 8). Using thermodynamic data,56 it can be shown that covellite is the stable phase for the conditions of the WDP1 anaerobic cells (pH ca. 7, Eh ca. −200 mV, H2S > 10−6 mol kg−1). Also, at this Eh, aqueous Cu+ should predominate over Cu2+.
To calculate the concentrations of the Cu(II) polysulfide complexes proposed by Shea and Helz,54 it was assumed that solid elemental sulfur was present in the WDP1 cells, with an activity of unity. Although elemental sulfur was not directly observed in subsurface water samples, it was present in abundance in cells 6 and 7 near the outlet to the anaerobic cells. It is reasonable to assume that elemental sulfur particles were also forming in the subsurface, due to mixing of water from different micro-environments, diffusion of oxygen from the atmosphere, and/or respiration of oxygen by cattails and other plants in the root zone.57 In the event that the subsurface waters were undersaturated with elemental sulfur, the estimated solubilities based on the model of Shea and Helz54 are still useful in that they provide a maximum limit on the stability of Cu(II) polysulfide complexes.
Fig. 10 shows the predicted copper solubility trends compared with wetland effluent data. The detection limit for Cu ranged from 9 × 10−9 to 2 × 10−8 mol kg−1. The majority of measured Cu concentrations were more than one order of magnitude above the IDL. The models based on Mountain and Seward55 and Shea and Helz54 appear to match the data from WDP1 fairly well at the upper range of bisulfide concentrations presented. However, the two models significantly diverge from each other at lower sulfide concentrations, and both underestimate the WDP1 field data. To explain this discrepancy, a number of possibilities are considered. First, it is possible that copper concentrations at WDP1 were controlled by a metastable or poorly crystalline phase. Shea and Helz58 reported that the solubility of poorly crystalline copper monosulfide (approximate composition Cu1.18S) was roughly three orders of magnitude greater than that of crystalline covellite. Also, Thompson and Helz59 proposed that chalcocite can metastably control dissolved copper in sulfidic waters in cases where covellite is predicted to be the stable solid phase. Thompson and Helz showed that deep Black Sea water was ca. 2 to 2.5 orders of magnitude supersaturated with respect to covellite, but close to equilibrium saturation with chalcocite. The main problem with invoking amorphous or metastable Cu–S phases to explain the WDP1 data is that, although the solubilities of these phases are much higher than crystalline covellite, there still should exist a positive correlation between total dissolved copper and total dissolved sulfide. Such a correlation is absent in the WDP1 data shown in Fig. 10.
![]() | ||
Fig. 10 Comparison of concentrations of filtered copper at WDP1 vs. predicted values based on solubility models from literature sources. |
It is also quite possible that the elevated copper concentrations at WDP1 were due to sampling of colloidal Cu-sulfide particles with diameter⊕<⊕0.45 µm. Cu-sulfide sols formed by rapid precipitation with H2S have been described,60,61 and it has been suggested that the sulfide nanocrystals in these sols may remain in suspension for years, provided the total metal concentration and ionic strength are both low.60 This hypothesis would explain the relatively high concentration of copper in filtered samples, as well as the lack of influence of copper mobility on dissolved sulfide shown in Fig. 10.
![]() | ||
Fig. 11 Comparison of concentrations of filtered cadmium at WDP1 vs. predicted values based on solubility models from literature sources. |
The same general solubility model for CdS was derived from the experiments of Wang and Tessier10 and Ste. Marie et al.62 In both studies, all of the complexes of interest had the general formula Cd(HS)x2 − x, and similar dissociation constants were obtained. The curve generated from MINTEQ also reflects the general shape of these models. For reasons that are not obvious, amorphous CdS from Ste. Marie et al. is about two orders of magnitude more soluble than the amorphous phase in Wang and Tessier. The curve representing the model by Daskalakis and Helz63 has a different shape, reflecting the fact that the dominant aqueous species using their data for the conditions of WDP1 is a mixed hydroxy-sulfide complex, CdOHS−.
Fig. 12 shows the predicted iron solubility trends according to the data in the recent compilation of Davison et al.,65 as compared with the WDP1 wetland effluent data. The detection limit for Fe ranged from about 9 × 10−9 to 3 × 10−8 mol kg−1. The models of Davison et al. and MINTEQA2, based upon amorphous FeS solid phases, bracket the observed iron concentrations well once bisulfide concentrations exceed ca. 5 × 10−5 mol kg−1. At lower sulfide concentration, it is likely that iron solubility in the wetland cells was limited by ferric oxide/hydroxide phases. For the following reactions:
![]() | (9) |
![]() | (10) |
![]() | ||
Fig. 12 Comparison of concentrations of filtered iron at WDP1 vs. predicted values based on solubility models from literature sources. |
Log K at 25°C is computed to be 14.84 and 8.95, respectively.64,66 Assuming pH = 7 and Eh = −200 mV (a reasonable value for neutral waters containing small amounts of sulfide), ferric oxy-hydroxide should convert to amorphous FeS at H2S concentrations greater than 3 × 10−11 mol kg−1
(ferrihydrite-stable) or greater than 2 × 10−5 mol kg−1
(goethite stable). A comparison of these values with Fig. 12 suggests that goethite may be the solubility-limiting phase for iron in the anaerobic cells when H2S concentrations are less than roughly 10−5 mol kg−1. In contrast, ferrihydrite (poorly crystalline ferric hydroxide) is predicted to sulfidize to FeS at extremely low H2S concentrations.
Unlike zinc and copper, ferrous iron does not undergo significant complexation with bisulfide until the concentration of the latter exceeds about 10−3 mol kg−1,65 which is near the upper limit of the measured values at WDP1.
Selected thermodynamic data for the arsenic-sulfide system are presented in Table 5. Considerable disagreement persists over the exact nature of the arsenic sulfide complexes. In particular, it is uncertain whether the complexes exist as monomers, dimers, or trimers. Mironova et al.67,68 proposed dimeric complexes. The other authors advocate monomers or trimeric arsenic sulfide complexes. Models by Webster69 and Eary70 include the same suite of trimeric complexes, but differ in the crystallinity of the solid phase. Young and Robins 71 obtained very high solubilities for amorphous As2S3 in H2S solutions (solubility minimum = 45 ppm), and proposed yet another set of polynuclear aqueous complexes. Helz et al.72 argued that, whereas polynuclear As-sulfide complexes probably dominate in As2S3-saturated solutions, mononuclear complexes are more likely at the low As concentrations of most natural waters.
Fig. 13 shows the predicted arsenic solubility trends compared with wetland effluent data. The average influent arsenic concentration is shown for reference. The detection limit for arsenic ranged from about 1 × 10−8 to 3 × 10−8 mol kg−1. In contrast to the metals discussed previously, the WDP1 data do not show arsenic removal as a function of increased sulfide concentration. Instead, a slight increase is observed at higher bisulfide concentrations, indicating mobilization rather than attenuation in the anaerobic cells. The increase in dissolved arsenic concentrations is attributed to dissolution of particulate or adsorbed arsenic from the influent water, or dissolution of trace amounts of arsenic present in the original gravel substrates themselves.
![]() | ||
Fig. 13 Comparison of concentrations of filtered arsenic at WDP1 vs. predicted values based on solubility models from literature sources. |
Arsenic solubility curves in Fig. 13 based upon crystalline and amorphous As2S3 show that the WDP1 waters are significantly undersaturated with respect to these solid phases. In contrast to the other metals of concern, precipitation of As2S3 is not a viable mechanism to remove dissolved arsenic to regulatory standards. For example, the Montana drinking water standard for arsenic is 18 ppb,73 and the EPA maximum contaminant level is 50 ppb (this may drop pending research and legislation in progress). Both of these values are below the minimum predicted solubility of As2S3 based on the various models shown in Fig. 13. The undersaturation of WDP1 waters with As2S3 is therefore explained by a combination of high solubility and low influent As concentrations. It is also possible that arsenic levels were kept low by adsorption or co-precipitation of arsenic onto Zn- and Fe-sulfide phases. However, we have no evidence for this mechanism.
![]() | (11) |
Log K
(25°C) is estimated at +3.8.66 If we assume ideal mixing, the activity of MnS in the ZnS particles of WDP1 would be roughly 0.03. Using this value, and the measured pH, bisulfide and manganese concentrations, the WDP1 waters are still undersaturated with MnS by >1 log unit. Thus, unless the MnS–ZnS solution is strongly non-ideal, it is more likely that manganese was removed by a non-equilibrium co-precipitation process, or by adsorption onto freshly precipitated ZnS particles.
Even at the more elevated DOC levels discussed above, it is exceedingly unlikely that organic ligands could out-compete with bisulfide for most of the metals of concern. For example, for the following reactions, (where Ac = acetate):
![]() | (12) |
![]() | (13) |
the respective log β2 values are +12.82 51 and +3.45.75 The ligand exchange reaction can be written:
![]() | (14) |
for which log K = +9.37. The condition {Zn(Ac)2} = {Zn(HS)2} would require log(Ac−/HS−) = 4.68. In other words, for the Zn–acetate complex to be important, acetate would need to be roughly 4 orders of magnitude more abundant than bisulfide. This was rarely the case for the WDP1 waters, even if we make the exceedingly unlikely assumption that 100% of the measured DOC was present as acetate. A similar conclusion can be made comparing the stability of sulfide vs. acetate complexes for cadmium and copper. Although organic complexation of iron, manganese and arsenic is potentially more important, no correlation was noted in this study between DOC and these metals.
Metal | Mean (log mol kg−1) | n | Standard deviation | Range of H2S |
---|---|---|---|---|
Zn | −6.2 | 81 | 0.6 | 10−6 to 10−3 mol kg−1 |
Cu | −7.4 | 81 | 0.5 | 10−6 to 10−3 mol kg−1 |
Cd | −8.2 | 81 | 0.5 | 10−6 to 10−3 mol kg−1 |
Fe | −5.6 | 81 | 0.5 | 10−6 to 10−3 mol kg−1 |
As | −6.7 | 81 | 0.6 | 10−6 to 10−3 mol kg−1 |
One possible explanation for the discrepancy between the field and laboratory data is that additional aqueous complexes are present in the waters of WDP1 for which thermodynamic data are unavailable. In this respect, it is important to realize that most of the previous solubility experiments were conducted at relatively high total sulfide concentrations. For example, the lowest total sulfide values used by Shea and Helz,54 Mountain and Seward,55 and Daskalakis and Helz51 were 0.001, 0.01 and 0.0004 mol kg−1, respectively. Thus, the maximum sulfide concentrations measured in the WDP1 treatment wetland (which are similar to the upper limit of H2S in natural sulfidic waters) are at the lower end of the range of total sulfide employed in the experimental studies. In general, the lack of experimental information at low ligand concentration is a common problem in solubility studies: in order to obtain quantities of reactants and products that can be accurately measured, it is often necessary to perform experiments at unrealistically high ligand concentrations. Although procedurally difficult, further experiments are needed to more closely mimic the conditions of natural sulfidic waters.
It is also possible that the nearly constant and elevated metal concentrations obtained from WDP1 are due to sampling of metal-sulfide colloid or cluster compounds. It is widely known that colloidal ZnS sols form during rapid precipitation of Zn-sulfide from aqueous solution. Daskalakis and Helz51 found that colloidal ZnS particles persisted for at least several days, and that the aged solutions did not equilibrate with crystalline sphalerite, even after several months. To obtain equilibrium solubility values, Daskalakis and Helz went to great lengths to avoid supersaturation (they could not reverse their experiments), lest colloids were allowed to form and persist. Along the same lines, Luther et al.76 discuss the formation of aqueous zinc sulfide clusters that form at relatively low concentrations of metal and bisulfide. Their results indicate that these clusters are precursors to solid precipitate formation, and are “thermodynamically strong”. Similar copper sulfide sols have been shown to be stable for years, depending on the solution pH, metal concentration, and ionic strength.60,61 Rozan et al.77 describe sulfide cluster compounds containing zinc, iron and copper with diameters⊕<⊕0.2 µm in oxic, polluted rivers in the eastern US. These compounds apparently persist for days, even in the presence of oxygen, and can contribute to “dissolved” metal concentrations of up to 6 × 10−7 mol kg−1, which, coincidently, is close to the filtered copper and zinc concentrations obtained from the WDP1 wetlands (Table 6). It is very possible that similar colloidal metal sulfide particles could form and persist over the time frame of a rapid treatment scheme such as WDP1. The very cold water temperatures in the winter months could have further retarded flocculation and growth of sulfide particles to filterable dimensions. Ultra-filtration to 0.001 µm would have been necessary to discriminate between colloidal vs. dissolved metal, but even this pore size would not filter out the cluster compounds proposed by Luther et al.76 and Rozan et al.77 such as M2S4, M3S3, and M4S6 (M = Cu, Fe, or Zn). No attempts to collect ultra-filtered samples were made at WDP1 during its life of operation.
A third factor to consider regarding the discrepancy between field and experimental results deals with the quality of the WDP1 data. With the exception of cadmium, all of the metals discussed in this paper were typically well above their instrument detection limits. There is, however, considerable error associated with the dissolved sulfide measurements at the lower range of values reported (<10−5 mol kg−1). For one thing, the instrument detection limit for the colorimetric procedure employed is on the order of 10−7 to 10−8 mol kg−1 S2−, and the suggested method detection limit quoted by HACH is even higher (ca. 3 × 10−7 mol kg−1). Because sulfide was not a priority analyte, no rigorous quality control procedures were performed at WDP1 to test the accuracy or precision of the analytical results at very low sulfide concentrations. A second problem is that, in some cases, a significant amount of the total sulfide measured during the colorimetric determination could have been initially complexed with dissolved metal. Mass balance calculations indicate that the only metal in this study that could significantly influence the total sulfide mass balance at low S2− concentrations is zinc. The geometric mean concentration of dissolved zinc for 81 samples with total sulfide >10−6 mol kg−1 was 6.3 × 10−7 mol kg−1 (Table 6). Even at very low sulfide concentrations, the data of Daskalakis and Helz51 suggest that the dominant aqueous zinc complex was Zn(HS)2−. If so, this implies that ca. 10−6 mol kg−1 of dissolved sulfide could easily have been tied up as an aqueous zinc complex. Thus, the true concentration of “free” sulfide may have been overestimated in Fig. 9 to 13, especially at ΣS2−⊕<⊕10−5 mol kg−1. To make matters worse, Thompson and Helz59 have shown that S2−/Zn mole ratios near unity favor the formation of ZnS cluster particles, which for the most part have not been thermodynamically well-characterized. This is analogous to the polymerization of metal cations undergoing hydrolysis, where the final Me/OH ratios are near unity.78
Finally, it is instructive to re-examine the possible errors introduced from two of the more dubious assumptions made in our solubility calculations. Firstly, the calculations assumed pure phases, whereas it is possible or indeed likely that the sulfide particles consisted of solid solutions or intimately mixed precipitates. Incorporation of a solid solution model would lower the activity of the end-member metal sulfide component, resulting in lower calculated solubilities. For zinc, copper and cadmium, this would only worsen the agreement between observed and theoretical dissolved concentrations. For arsenic and manganese, the gap would be lessened, although probably not enough to result in a close agreement (see section 4.6). The second assumption of a uniform temperature of 25°C also needs to be examined. The WDP1 water was typically much cooler than 25
°C, approaching 0°C in the winter months. Unfortunately, recalculation to the true field temperature was not possible in this study, because accurate data on the temperature dependence of the relative solubility-controlling reactions (including formation constants for individual metal sulfide complexes) are missing or incomplete. However, a rough idea of the direction and possible magnitude of the temperature correction can be obtained by examination of the enthalpies of the solubility products of each metal sulfide system. Expressed in the following format:
![]() | (15) |
the standard enthalpies of reaction (15) for crystalline ZnS, CuS, CdS, and MnS are +34.5, +100.5, +68.5 and −24.2 kJ mol−1, respectively.66 Thus, for zinc, copper and cadmium, sulfide mineral dissolution is endothermic, and will decrease with decrease in temperature, whereas the reverse is true for manganese. Using a Van't Hoff extrapolation of the above solubility product of sphalerite, a drop in temperature from 25 to 2°C translates into a drop in the theoretical zinc solubility of just under one order of magnitude. This will worsen the discrepancy between our calculated Zn solubilities and the observed concentrations in the field. This is also true for most of the other metals of interest, including manganese, which was shown to be strongly undersaturated with respect to pure MnS.
Footnote |
† Present address: CDM Federal Programs Corporation, Helena, Montana 59601. E-mail: E-mail: frandsenak@cdm.com |
This journal is © The Royal Society of Chemistry 2001 |