Background

10 Sulphide-mineral oxidation is commonly cited as the cause of poor water quality at mine sites worldwide because it results in acid-mine drainage, which is the release of acidity, sulphate, and dissolved metals (e.g., Alpers and Blowes 1994; Jambor and Blowes 1994; Jambor et al. 2003). Therefore, understanding sulphide-mineral oxidation is an important step toward predicting the geochemical evolution of drainage water from tailings and implementing mitigation measures that will ensure environmentally acceptable mine closure. One of our research interests is related to the oxidation of arsenopyrite (FeAsS) in the tailings from the Mount Pleasant Tungsten Mine, which is located approximately 60 km south of Fredericton, NB. The geochemistry, hydrogeology, and mineralogy of the Mount Pleasant Mine tailings has been previously presented by Petrunic and Al (2005), and from this study it is evident that elevated As concentrations occur in the near-surface pore water (maximum of 7.1 mg/L). The abiotic and biotic oxidation of arsenopyrite under a variety of conditions (e.g., air and aqueous solutions) has been studied previously (e.g., Buckley and Walker 1988; Richardson and Vaughan 1989; Nesbitt et al. 1995; Fernandez et al. 1996a, b; Nesbitt and Muir 1998; Schaufuss et al. 2000; Jones et al. 2003; Mikhlin and Tomashevich 2005). Many of these studies have used spectroscopic techniques to detect changes in oxidation state for As, Fe, and S as arsenopyrite oxidizes and secondary minerals form. From previous studies, it is evident that arsenopyrite oxidation leads to the formation of various As- and/or Fe-bearing secondary minerals. The aim of this study is to build on previous research by employing analytical (S)TEM to study secondary minerals that have formed as a result of arsenopyrite oxidation and interactions with the local pore-water chemistry. The results from this work are summarized here but are explained in detail by Petrunic et al. (2006).

TEM sample preparation

11 The loose tailings were collected using aluminum casing (7.5 cm diameter), which was later cut in half lengthwise. The tailings were then dried and impregnated with low-viscosity epoxy. Double-polished thin sections were prepared from the epoxy-impregnated blocks. Oxidized arsenopyrite grains were identified in the thin sections using SEM in the backscattered electron mode, and the FIB was used to prepare two sections for (S)TEM analysis from specified locations at the margins of two arsenopyrite grains (Fig. 1A). The FIB sample preparation technique was chosen for this study because it is possible to prepare site-specific sections from minerals of contrasting hardness while maintaining spatial relationships. The FIB method described by Patterson et al. (2002) was used to prepare samples in this study. It was possible to obtain a sample from each grain at a specified location, with the section extending from the epoxy-filled pore space, through the soft secondary-mineral coating, and into the relatively hard arsenopyrite grain (Fig. 1B). In this way, the spatial relationship between the pore space, the coating, and the arsenopyrite was maintained, allowing for a meaningful interpretation of the results. Although FIB was initially developed for materials science, it is a useful tool for the preparation of high quality specimens for (S)TEM analysis from a wide variety of geological samples (e.g., Heaney et al. 2001; Dobrzhinetskaya et al. 2003; Lee et al. 2003; Seydoux-Guillaume et al. 2003; Wirth 2004; Benzerara et al. 2005).

Display large image of Figure 1

Summary of results

12 The secondary coating extends several microns from the arsenopyrite grain boundary (Fig. 1B). A higher magnification HAADF image acquired from the rectangular area outlined in Fig. 1B, demonstrates the heterogeneity of the coating (Fig. 1C). In this image it is possible to note the presence of several distinct phases which were characterized by EDS analyses. These phases include an Fe– As– O coating, a Cu– S-rich phase, a Pb– Mo– O-rich phase, and a Bi– Pb– As– O-rich phase. Examples of each of these phases are presented at still higher magnification in Figs. 2A-D, and representative STEM– EDS analyses are shown in Fig. 2E. Unique mineral identification requires chemical as well as structural information, therefore, electron diffraction patterns were collected from each of the phases in either SAED and/or microdiffraction mode and these patterns were compared against simulated electron diffraction patterns of known minerals. These analyses indicate that the Fe– As– O coating is amorphous, the Cu– S-rich phase is crystalline and the patterns are consistent with chalcocite (Cu₂S) and/ or djurleite (Cu₃₁S₁₆), the Pb– Mo– O-rich phase is crystalline and the patterns are consistent with wulfenite (PbMoO4), and the Bi– Pb– As– O-rich phase is crystalline but the diffraction patterns could not be indexed conclusively.

13 Based on imaging, EDS, and electron diffraction data collected from the two samples, it was concluded that arsenopyrite oxidation leads to the precipitation of the amorphous Fe– As– O coating. This is consistent with previous studies, which have identified a range of As- and (or) Fe-rich coatings as arsenopyrite oxidation products (Buckley and Walker 1988; Richardson and Vaughan 1989; Nesbitt et al. 1995; Nesbitt and Muir 1998; Schaufuss et al. 2000). The main difference between this and previous studies arises from the ability to collect images, chemical analyses, and diffraction data from the same sample at the nanoscale while preserving spatial relationships. This allows for a better understanding of the influence of the pore-water composition on secondary-mineral formation. With the exception of the Fe– As– O coating, the formation of the secondary minerals that are observed requires the input of additional elements (Cu, Pb, Mo, Bi) to the local geochemical system. The addition of these elements is accomplished by advection or diffusion of aqueous solutes in the pore space adjacent to the arsenopyrite grain. For instance, in addition to finding discrete Cu sulphides disseminated in the bulk coating, Cu sulphides were observed along the arsenopyrite grain boundary. This suggests that a reaction between aqueous Cu(II) and arsenopyrite occurs in the tailings, resulting in the precipitation of a Cu sulphide along the boundary according to the following simplified reaction (scorodite represents the Fe– As– O coating):

Display large image of Figure 2

14 Continued oxidation of arsenopyrite and retreat of the grain boundary leads to the isolation of Cu sulphides in the Fe– As– O coating in the form of layers aligned parallel with the arsenopyrite grain boundary (Fig. 2B) and aggregates (not shown). The precipitation of wulfenite and the Bi– Pb– As– O-rich phase occurs in the open voids within the Fe– As– O coating in response to changes in saturation state for these phases in the local pore-water.

15 More details regarding the processes mentioned above can be found in Petrunic et al. (2006), but the overall conclusion is that there is a complex assemblage of secondary minerals in this system and their compositional and mineralogical characteristics can only be adequately assessed at the nanometre scale.

Background

16 Chlorinated solvent contamination is common in ground-water systems, resulting from subsurface accumulations of immiscible solvents which are part of a larger group of Dense Non-Aqueous Phase Liquids known as DNAPLs. Trichloroethylene (TCE; C₂Cl₃H) and tetrachloroethylene (also known as perchloroethylene or PCE; C₂Cl₄) are two of the most widely used solvents. A common approach to aquifer restoration is the in situ destruction of chlorinated ethane sources using chemical oxidation, referred to as In Situ Chemical Oxidation (ISCO). The permanganate ion (MnO₄ ^-) is a strong oxidant that is frequently employed in the form of either the potassium or sodium salt (KMnO₄ and NaMnO₄). The injection of MnO₄ ^- into aquifers, and subsequent reactions with TCE and PCE, results in profound changes in the redox state and may lead to a large decrease in pH as well:

17 The reduction of Mn(VII), the overall increase in the aquifer redox state, and the addition of acidity to the aquifer trigger a variety of mineral-water interactions in the aquifer. The two most important reactions are the precipitation of insoluble Mn oxides with Mn valence ranging between II and IV (herein referred to generally as MnO_x), and pH buffering from dissolution of carbonate minerals (Nelson et al. 2001; Huang et al. 2002; Crimi and Siegrist 2004; Li and Schwartz 2004). Aquifer systems commonly contain small amounts of heavy metals that are naturally bound in a variety of stable mineral forms. The radical changes in redox and pH conditions that accompany ISCO could lead to the mobilization of these metals. The acid-generating and pH-buffering reactions represent important controls on trace-metal solubility and adsorption behaviour (Pretorius and Linder 2001; Kanungo et al. 2004; Tonkin et al. 2004).

18 One of the objectives of this study was to use analytical (S)TEM to investigate metal sequestering by MnO_x that form during ISCO in a controlled laboratory column experiment. The input solution to the sand-packed columns contained trace metals (Cu, Pb, Zn, Ni, Mo) in addition to MnO₄ and TCE. A flow gradient was induced along the columns and the reaction was monitored under calcite (CaCO₃) buffered and unbuffered conditions. The results that pertain to the carbonate-buffered system are summarized here, but readers are referred to Al et al. (2006a) for a more detailed description of the research.

TEM sample preparation

19 The MnO_x that form as a result of ISCO are expected to coat the silica and calcite sand grains present in the columns. This process presents a problem because it can be very challenging to prepare samples for TEM analysis that contain both hard and soft components such as MnO_x-coated silica grains. Although the thin section/FIB method described earlier could have been suitable for this study, the columns could not be dismantled to allow sampling for thin-section preparation because the columns were required in a second experiment. Therefore, a technique that would minimize disturbance to the system was necessary. Although hard minerals can be successfully sectioned by ultramicrotomy using a diamond knife, damage to the knife can easily occur, making the technique costly. As an alternative to these methods, we mixed polypropylene threads with the silica and calcite sand during construction of the columns so that they could be removed later through side ports installed along the length of the columns. As the KMnO₄— TCE reaction proceeded and MnO_x formed, the oxides coated the threads. As an artificial substrate, the threads were easily sectioned with a diamond-knife mounted on an ultramicrotome. Prior to ultramicrotomy, the MnO_x coated threads were embedded in a resin block immediately after extracting them from the column to avoid mineralogical changes due to drying and desiccation. NANOPLAST® FB101 was used to embed the threads because it can accommodate small amounts of water during curing.

Summary of results

20 The MnO_x coatings were characterized by analytical (S)TEM. In the plane of section, the coatings occur as fibrous masses that extend outward, generally normal to the surface of the thread (Fig. 3A). The fibres are surrounded by finer-grained, interstitial MnO_x that has no discernable crystal habit. Although the fibrous habit within the coatings suggests that there is some ordered structure, diffraction patterns obtained by SAED suggest that the material is amorphous.

21 Analyses of the coatings by EDS display detectable Si, Mo, Fe, Ni, and Cu (Fig. 3B). Note that the large Cu peaks in spectra 1-3 of Fig. 3B are artifacts of the Cu-mesh grids used to support the samples in the TEM. Similarly, the large Ni peaks in spectra 4 and 5 are artifacts from Ni grids. Despite these artifacts, Ni is detected in the samples that were supported on Cu grids (spectra 1-3) and Cu is detected in samples supported on Ni grids (spectra 4 and 5).

22 The general conclusion from the (S)TEM analyses is that secondary MnO_x formed as a result of ISCO is capable of sequestering a variety of metals either through adsorption and/or co-precipitation mechanisms. The MnO_x is extremely fine grained and intergrown with oxides and hydroxides of Fe, Si, and Al, and it is therefore not possible with techniques other than (S)TEM to obtain the discrete analyses necessary to support this conclusion.

Background

23 Groundwater conditions are typically reducing in fractured crystalline rocks at depths of 500 to 1000 m, potentially providing a stable geochemical environment for waste isolation as proposed in the deep geologic repository concept of nuclear waste management. However, over tens of thousands of years, the glaciation events that are expected would cause changes in the hydrogeological and geochemical environment possibly leading to the ingress of dissolved O2 to greater depths than at present. Paleohydrogeological evidence for the depth of penetration of oxygenated fluids within fractured crystalline shield environments in the past may provide insight into future changes due to variations in climate and groundwater flow conditions.

24 Conceptually, the movement of oxygenated groundwater through a fracture zone leaves geochemical evidence in the form of a zone of alteration around the fracture. The strongest alteration occurs in minerals at the fracture surface where advectively transported oxygen oxidizes the reduced species, especially Fe²⁺, at the mineral surfaces. As a result, secondary Fe³⁺-bearing minerals are often precipitated on the edges of the fracture. A diffusive flux of oxygen away from the fracture into the matrix causes a wider zone of oxidation extending for some distance into the wall rock adjacent to the fracture. Oxygen diffuses along the grain boundaries between minerals, reacting first with the mineral surface and then slowly diffusing into the grain. Minerals in this region may show zonation from reacted rims to unreacted cores, producing a trend of decreasing Fe³⁺/ ∑Fe with distance into the mineral. Reactions with oxygen and water may also destabilize the crystal lattice of primary minerals, such as biotite (K(Mg,Fe)3(AlSi3)O10(OH)2), leading to the formation of secondary clay minerals in the alteration zone. The aim of this study was to demonstrate the use of analytical TEM in the identification of oxidative alteration products in biotite samples obtained from a rock core adjacent to a fracture in the Lac du Bonnet granite batholith. More details regarding this study can be found in Cavé and Al (2005).

Display large image of Figure 3

TEM sample preparation

25 The Precision Ion Polishing System (PIPS) was used to prepare samples for TEM analysis. This technique is a routine method for the preparation of geological samples. The general steps to obtaining such samples include the identification of regions of interest in a standard 30 µm double-polished petrographic thin section mounted on a glass slide with Crystalbond^TM thermo polymer adhesive, the attachment of a standard 3 mm diameter TEM grid (aperture or slot grid) over the region of interest, and thinning the sample to electron transparency in the PIPS.

Summary of results

26 Biotite grains located at various distances from the fracture were prepared for (S)TEM analysis. Figure 4A presents an ADF image of a biotite grain at an intermediate magnification that was found 5 mm from the fracture. It is evident from the variable contrast in the image that the material is not homogeneous. In an attempt to explain the inhomogeneity, additional characterization was done using EDS mapping, thereby providing compositional images (Figs. 4B-H) from the same area as the ADF image (Fig. 4A). Consistent with the ADF imaging, the EDS elemental mapping reveals that on the submicron scale, the composition of biotite adjacent to the oxidized fracture is not uniform, as was suggested by observations made at the micron scale using EMP (McMurry and Ejeckam 2002) and SEM– EDS (Cavé and Al 2005) analyses.

27 The original biotite (K(Mg,Fe,Ti)3(AlSi3)O10(OH)2), contains Ti and K which are lost along with some of the Fe as the mineral alters to vermiculite (Mg0.33(Mg,Al,Fe³⁺)3(AlSi3)O10(OH)2). The final stage of alteration is manifest by the presence of kaolinite (Al2Si2O5(OH)4) and iron oxyhydroxide. The Fe map (Fig. 4B) highlights areas of high Fe content between 50 and 100 nm thick, which are Si and Al poor, representing Fe oxyhydroxide layers identified by SAED patterns as goethite (FeO(OH)). The Al and Si maps (Figs. 4E, 4G) contain a small wedge-shaped region in the lower left with elevated Al, Si, and O only, representing kaolinite.

28 The low temperature weathering of igneous biotite generally follows the sequence (Dong et al. 1998; Jeong and Kim 2003):

29 The weathering products identified in Fig. 4 provide evidence for oxidative alteration of the biotite. The aluminosilicate alteration products of biotite represent robust indicators of past redox changes because, unlike Fe oxyhydroxides, they are not susceptible to reductive dissolution if the system later reverted to reducing conditions.

30 This study was a test of concept, and it was known that these samples had undergone oxidative alteration. Although the alteration is not evident at the scale of SEM and EMP investigations, the STEM imaging and EDS elemental mapping techniques demonstrate that these alteration products are detectable at the nanoscale.

Background

31 Iron is the fourth most abundant element in Earth's crust. Iron-bearing minerals are found in a large variety of geologic environments, ranging from high pressure and temperature deep within the crust and mantle, to low temperature and pressure at Earth's surface. In Earth's crust, Fe occurs mainly in two valence states, ferrous (Fe²⁺) and ferric (Fe³⁺), and it cycles between these states in response to prevailing reduction-oxidation (redox) conditions. Consequently, measurements of Fe valence ratios (Fe³⁺/∑Fe) in minerals can provide a useful measure of past redox states in geologic systems. For example, measurements of Fe³⁺/∑Fe in minerals and silicate glasses can be used to investigate oxygen fugacity in the mantle (McCanta et al. 2004; Williams et al. 2004; Bezos and Humler 2005), and estimate temperature and pressure conditions for igneous and metamorphic rocks (Schumacher 1991; Andersen et al. 1991; Holdaway et al. 1997; Wu et al. 2004). In low temperature environments, measurements of the Fe³⁺/∑Fe ratio may provide a tool for estimating the depth of infiltration of oxygenated groundwater into bedrock, which has applications in fields such as climate change studies and the safe siting of geological repositories for nuclear waste disposal.

32 Traditional methods of quantifying Fe³⁺/∑Fe have involved wet chemical analysis, Mössbauer spectroscopy, or calculations from microprobe data using mineral stoichiometry and charge balance considerations. One of the challenges of these bulk methods has been the problem of spatial resolution, especially in fine-grained materials with spatial heterogeneity. Advances in X-ray microbeam and synchrotron methods have allowed Fe³⁺/∑Fe measurements to be made on individual mineral grains using techniques such as X-ray Photoelectron Spectroscopy (XPS) (Raeburn et al. 1997), X-ray absorption spectroscopy (Cressey et al. 1993; Dyar et al. 2002) and Synchrotron Mössbauer Spectroscopy (SMS) (Jackson et al. 2005). However, the best spatial resolution to date for measuring Fe³⁺/∑Fe ratios has been obtained from EELS in the TEM (Garvie and Buseck 1998; van Aken et al. 1998; van Aken and Liebscher 2002; Cavé et al. 2006).

Display large image of Figure 4

33 The main objective of this research was to develop a method for one- and two-dimensional mapping of Fe valence in oxide minerals using coupled STEM– EELS. Modified from the procedures developed by van Aken et al. (1998), the Fe³⁺/∑Fe ratios are measured by a calibrated relationship between Fe valence and Fe L3/L2 white line intensity ratios in EELS spectra. The method is demonstrated with spectrum images of the Fe³⁺/∑Fe changes occurring across mineral interfaces between ilmenite (Fe²⁺TiO₃) and magnetite (Fe²⁺Fe³⁺₂O₄) in a sample with microscale exsolution lamellae. A summary of the results is presented here but readers are referred to Cavé et al. (2006) for more details.

TEM sample preparation

34 Iron valence measurements were calibrated using Fe-bearing mineral standards of known Fe valence: fayalite (Fe²⁺₂SiO₄), magnetite, and hematite (Fe³⁺_2O3). Samples of ilmenite, goethite (Fe³⁺O(OH)) and a different sample of hematite were analyzed as unknowns to verify that the method produced reliable results for samples of known Fe valence. Grain mounts of the calibration standards were examined by SEM and analyzed by SEM– EDS to confirm their purity.

35 Mineral standards were prepared for STEM– EELS by grinding to a fine powder in an agate mortar and pestle. Only a small amount of fayalite was available, so the crystals were crushed to powder between glass slides. Crushed material was suspended in 100% ethanol and sonicated. The suspension was withdrawn during sonication using a micropipette and dropped on a holey carbon coated molybdenum TEM grid. The standards were lightly carbon coated to prevent specimen drift during TEM analysis. This method provides areas on the edges of crushed grains which are thin enough for EELS analysis.

36 An intergrown magnetite/ilmenite sample from the University of New Brunswick Mineral Collection (Sample UNB45-1078) was used for Fe valence mapping. This magnetite was a different sample from that used as a calibration standard. A double-polished, 30 µm petrographic thin section was prepared on a glass slide, areas of finely intergrown magnetite and ilmenite were identified under reflected light, and the PIPS method of preparing samples for TEM analysis was used. Following thinning, samples of magnetite/ilmenite were carbon coated.

Summary of results

37 The sample of finely intergrown magnetite and ilmenite (Fig. 5) was ideal for developing Fe valence mapping and testing the spatial sensitivity of the technique because the ex-solution lamellae provide sharp boundaries between zones of markedly different Fe valence states. Ilmenite lamellae in the sample range in width from 1 to >100 µm. Data for the average composition of ilmenite suggest that there is substitution of Mn²⁺ and Mg²⁺ for Fe²⁺ and that about 10% of the Ti⁴⁺ sites are occupied by Fe³⁺. Numerous irregularly-shaped inclusions of spinel (MgAl₂O₄) were evident from optical microscopy and SEM, particularly within the ilmenite. Trace quantities of Al, Ti, and Cr may substitute within the magnetite structure, but these elements could also be present in the form of spinel and ilmenite inclusions, which are not resolved at the scale of the SEM– EDS analysis.

38 Point determinations of Fe³⁺/∑Fe by STEM– EELS were made on magnetite and ilmenite before acquiring spectrum images. Magnetite Fe³⁺/∑Fe values cover a small range, with a mean of 0.67 ± 0.03 (mean ± 1ó), which is in agreement with the theoretical value for magnetite. In contrast, STEM– EELS determinations of Fe³⁺/∑Fe ratios for ilmenite are more scattered, ranging between 0 and 0.20. The STEM– EELS valence measurements on ilmenite are consistent with calculations of Fe³⁺/∑Fe based on SEM– EDS analyses and mineral stoichiometry, which suggest that the average Fe³⁺/∑Fe ratios are 0.18 ± 0.02.

Display large image of Figure 5

39 An example of a Fe³⁺/∑Fe line scan, obtained using the STEM– EELS imaging method developed in this study, is shown in Fig. 5. Line scan data are presented as a scatter plot of Fe³⁺/∑Fe versus distance along the path shown in Fig. 5A. This method has also been used to acquire 2D analytical images of the Fe valence across the magnetite/ilmenite interface (Cavé et al. 2006). The data demonstrate that the Fe valence in magnetite is relatively constant, suggesting that in this magnetite sample there is no substitution of Fe²⁺ in the Fe³⁺ sites or vice versa. Variations in the Fe³⁺/∑Fe for magnetite may arise from substitutions in the magnesioferrite-jacobsite-magnet-ite mineral series (Mg²⁺Fe³⁺₂O₄ – (Mn,Fe,Mg)²⁺(Fe,Mn)³⁺₂O₄ – Fe²⁺Fe³⁺₂O₄), but this specimen does not contain sufficient Mg or Mn to have a significant effect on the Fe³⁺/∑Fe. There is a measurable variation in Fe³⁺/∑Fe within the ilmenite grains, with fluctuations between 0 and 0.20 over submicron distances. Substitution of Fe³⁺ is allowed within the ilmenite lattice. At high temperatures, there is a continuous solid-solution series between ilmenite and the Fe³⁺ end-member, hematite, which is achieved by coupled substitution of Fe³⁺ for Ti⁴⁺ and Fe³⁺ for Fe²⁺ (Andersen et al. 1991).

40 Intermediate Fe³⁺/∑Fe values were found for 5 points along the interface region between the two minerals in the line scan (Fig. 5). Since no attempt was made to ensure that the mineral interface intersects the sample plane at 90°, it is likely that the intermediate values reflect a mixed ratio from overlap of the minerals along an inclined interfacial boundary plane. When the 10 nm probe approaches very close to the grain boundary, the electron beam passing through the sample may interact with material on both sides of the interface to produce a weighted average signal for the two minerals.

41 The collection of STEM– EELS data as described in this study provides an extremely useful method of obtaining quantitative Fe valence measurements at the nanoscale, which can be applied to many geological issues. A similar approach has been taken to determine the Mn oxidation state in minerals and this is discussed in detail by Loomer et al. (2007).

Background

42 The Murray Brook volcanogenic massive-sulphide deposit in northern New Brunswick consists mostly of pyrite and Cu — Zn— Pb sulphides, and it is overlain by a Cu-rich supergene-enrichment zone and a surficial gossan (Boyle and Smith 1994; Boyle 1995). The gossan is composed primarily of goethite, quartz (SiO2), beudantite (PbFe3(AsO4)(SO4)(OH)6), jarosite ((K,Na)Fe3(SO4)2(OH)6), plumbojarosite (PbFe6(SO4)4(OH)12), argentojarosite (AgFe3(SO4)2(OH)6), bindheimite (Pb2Sb2O6 6 (O,OH)) and scorodite with accessory cinnabar (HgS), native Ag, Au, and cassiterite (SnO2) (Boyle 1995). The gossan was mined by open pit methods between 1989 and 1992. The ore was crushed, treated with cement and hydrated lime, and sprayed with Ca— CN solution for Au/Ag extraction. The CN leaching process also extracted Hg from the ore, and a total of 1000 kg of Hg were recovered during the Au-production process. The tailings, which contained residual Hg and CN, were stored on a prepared pad of compacted glacial till (Lemon and Pheeney 1996).

43 Residual CN in the tailings causes mobilization of Hg to the groundwater beneath the tailings in the form of Hg-CN complexes (Boyle and Smith 1994; Shaw et al. 2006), which in turn discharges into the headwaters of Gossan Creek. Gossan Creek, a first-order stream in the Upsalquitch River catchment contains highly elevated concentrations of Hg (Maprani et al. 2005). Previous work by Leybourne et al. (2000) indicated that residual CN has also been responsible for the mobilization of Au from the tailings to the creek via the groundwater flow system.

44 Despite ongoing efforts to understand the geochemistry and ecotoxicology of Hg, there has been relatively little research effort to identify the discrete mineral hosts for Hg that accumulate in sediment in response to contamination events. However, this information is necessary for the evaluation of long-term environmental impacts related to bioavailability and leachability. Numerous authors present the results of analyses that indicate high concentrations of Hg in stream sediment proximal to Hg mines and other sources of Hg contamination (Gosar et al. 1997; Lechler et al. 1997; Frazier et al. 2000; Gray et al. 2002; Gray et al. 2004). Gray et al. (2004) report that some of the elevated Hg in the stream sediments at Azogado Creek in Spain can be explained by the presence of detrital cinnabar and calcine material that has been transported to the stream from waste piles. Similarly, Lechler et al. (1997) report the presence of Au– Hg amalgam and elemental Hg in sediments, and suggest that these were released directly during mining activities near Carson River, Nevada. Results of chemical extractions provide indirect evidence that Hg may partition from water to the organic and inorganic fractions of sediments (Lechler et al. 1997; Drexel et al. 2002). Kim et al. (2004a,b) make use of X-ray absorption spectroscopy to demonstrate the adsorption of Hg onto Fe- and Al-hydroxides, and Barnett et al. (1997), using electron microscopy techniques, identified metacinnabar in Hg-contaminated floodplain sediments near Oak Ridge, TN.

45 The overall objective of this work is to understand the transport and fate of Hg that enters the surface flow system at the headwaters of Gossan Creek and partitions to the sediments. Using SEM, Leybourne et al. (2000) identified Au– Hg particles, < 1 µm diameter, in the sediment from Gossan Creek and suggested that these formed in the stream system in response to the degradation of Hg– and Au– CN complexes and subsequent reduction of Hg(II) and Au(I) to their elemental forms. One of the specific objectives was to build on previous research by Leybourne et al. (2000) and investigate the mechanisms of Hg mass loss from the stream to the sediments using analytical (S)TEM. Consequently, mineralogical investigations focused on the characterization of the amalgam particles, and identification of additional Hg-bearing secondary mineral phases. The results from this aspect of the study are presented here but additional details about the aqueous and sediment geochemistry and hydrology can be found in Al et al. (2006b).

TEM sample preparation

46 Sediments were sieved to < 45 µm and analyzed for total Hg. A portion of the sample with the highest Hg concentration (1 636 µg/g) was saturated with a low-viscosity epoxy (EPO-TEK 301-2FL) and then pressed into a 1 cm diameter mould to form a pellet. A double-polished thin section was prepared from the pellet and used for SEM– EDS analysis. Mercury-bearing sediment particles were identified with the SEM in the backscattered electron mode. In all cases, these particles were less than 1 to 2 µm in diameter; too small to obtain satisfactory image resolution, or spatial resolution for discrete EDS analyses. For this reason, analytical (S)TEM was used to further investigate the Hg-bearing particles. Samples were prepared for (S)TEM analysis from the standard petrographic thin section using the FIB method described by Patterson et al. (2002).

Summary of results

47 Mercury was identified by SEM– EDS in fine grains (< 2 µm) in association with Au, Ag, Cu and S. Two discrete forms of Hg-bearing mineral grains were observed with (S)TEM– EDS, HgS (Fig. 6) and Hg– Au amalgam with or without Ag. Silver was commonly observed in association with the HgS grains (Fig. 6), suggestive of the presence of a separate Ag-bearing phase, or a Hg– Ag sulphide such as imiterite (Ag₂HgS₂), but this possibility has not been resolved. Selected-area electron diffraction patterns were collected from the HgS grains whenever possible by tilting the specimen stage to acquire patterns from multiple zones on the same grain. Analysis of the patterns indicates that the HgS is crystalline with a structure consistent with metacinnabar (HgS). This result is consistent with the findings of Barnett et al. (1997) who identified metacinnabar in Hg-contaminated soils using TEM– EDS and electron diffraction, and with the experimental results of Charnock et al. (2003) who used X-ray absorption spectroscopy and XRD to define a sequence of reactions that culminate in the precipitation of stable, crystalline metacinnabar from HS^- and Hg(II)-containing solutions. The morphology of the metacinnabar grains presented in Fig. 6A is very delicate, and it is highly unlikely that that these grains could have been derived from mechanical transport from the gossan. We believe that they are secondary precipitates formed in situ within the stream sediments.

48 Analyses of the Hg-bearing amalgam grains suggest that Au is ubiquitous, and the Hg and Ag content of the grains varies widely. Two of these grains are presented in Fig. 7, one with abundant Ag (Figs. 7A, 7B), and the second with virtually no Ag (Figs. 7C-E). The relative proportions of these metals in the amalgam grains are likely to be controlled by their redox speciation and the activity of reduced S species. For example, in SO₄ reducing conditions, Hg and Ag are likely to partition to sulphide minerals.

Display large image of Figure 6

49 Abundant sub-micron size Cu-sulphide grains were also noted in the (S)TEM– EDS investigations (Figs. 7A, 7B). Many of them are fine-grained with a delicate morphology (not shown) and, as is the case for the Cu sulphide shown in Fig. 7A, Cu sulphide commonly occurs intergrown with, or encapsulated by the amalgam. Similar to metacinnabar, these grains are thought to have formed in situ within the sediment. The EDS analyses of the grains consistently indicate a high Cu– S ratio which is suggestive of a cuprous sulphide such as chalcocite or djurleite.

50 The results are consistent with Leybourne et al. (2000) in that Hg is attenuated in the stream sediments in the form of Hg– Au amalgam, but it is clear from the (S)TEM analyses that the amalgam is chemically heterogeneous at the nanoscale. For instance, Ag may or may not occur in association with the Hg– Au amalgam. In addition, the results from the (S)TEM analyses indicate that Hg has also partitioned to the sediments in the form of nanoscale metacinnabar. Gold is detected in the form of an amalgam only and Ag occurs in the Hg-, Au-containing amalgam but also in association with Hg and S, perhaps as a discrete Ag or Ag-sulphide phase, or a Hg– Ag-sulphide phase. As a result of the (S)TEM analyses it was also possible to identify the presence of nanoscale secondary Cu minerals in the stream sediments.

Display large image of Figure 7