Environ. Sci. Technol. 2008, 42, 37–42
Speciation of Cu in a Contaminated Agricultural Soil Measured by XAFS, µ-XAFS, and µ-XRF DANIEL G. STRAWN* AND LESLIE L. BAKER P.O. Box 442339, University of Idaho, Moscow, Idaho 83844-2339
Received June 29, 2007. Revised manuscript received October 11, 2007. Accepted October 22, 2007.
Contamination of agricultural soils with Cu as a result of fungicide application and spills threatens environmental quality and reduces soil quality for crop growth. In this paper advanced spectroscopic and microscopic methods were used to elucidate the Cu speciation in a calcareous soil contaminated since the 1940s. Microscopically focused synchrotron-based XRF (µ-SXRF) was used to map the elemental distribution in the soils. Results indicated that most of the Cu was not associated with metal oxides, silicates, phosphates, or carbonates. Bulk and microscopically focused X-absorption near edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) spectra indicated that the Cu in the soil was predominantly Cu adsorbed on soil organic matter (SOM). Interpretation of the fitting results suggests that the Cu is complexed to SOM via bidentate inner-sphere coordination with carboxyl or amine ligands.Resultspresentedinthispaperprovidedetailedinformation on the molecular coordination of Cu in a contaminated soil. Such information is critical for understanding the long-term fate and best management practices for Cu in the environment.
Introduction Previous studies have shown that soil organic matter (SOM) is a significant component for metal ion complexation in soils (1–3). Carbonate minerals, iron, manganese, and aluminum oxides, and clay minerals are also important sinks for metals in soils. In soils with mixtures of minerals and SOM, adsorption on SOM is an important controlling phase. The specifics of how SOM affects metal solubility and lability in a given system appear to depend strongly on the metal and soil physicochemical properties. McBride and Bouldin (4) examined the behavior of Cu in a contaminated agricultural soil from Hubbardsville, New York. This soil was contaminated more than seven decades ago with several thousand mg kg-1 Cu, likely as a result of a spill of CuSO4 fungicide. McBride and Bouldin (4) extracted the soil and determined that the soluble Cu fraction was associated with SOM. They observed indications of malachite coatings on dolomite fragments in the soil. However, leaching experiments suggested that the soil solution was undersaturated in malachite or tenorite. They concluded that although most of the soluble Cu in this sample was complexed with SOM, the nonwater-soluble Cu fraction was likely in inorganic form, such as hydroxide or hydroxycarbonate species sorbed on mineral surfaces. * Corresponding author e-mail:
[email protected]; phone: 208-885-2713; fax: 208-885-7760. 10.1021/es071605z CCC: $40.75
Published on Web 11/23/2007
2008 American Chemical Society
Sauvé et al. (5) examined the leachability of Cu in a number of soils including the Hubbardsville soil studied by McBride and Bouldin (4). In all soils except the Hubbardsville soil, total Cu content was correlated with total SOM. They proposed that the correlation was due to suppression of SOM oxidation as a result of microbe toxicity to Cu, and that the absence of a correlation in the Hubbardsville soil was due to agricultural perturbations. However, the contrasting Cu behavior in the Hubbardsville soil has not been definitively explained. Information on the speciation of Cu in the soil will provide insight into the observed Cu leachability and correlations. Studies on the molecular characterization of Cu in soils are difficult because multiple species may exist in a single soil. One approach used to understand Cu reactivity in natural environments is to study interactions with single minerals (6–12). In recent years, investigations into speciation in multicomponent systems have been done to simulate the complexity in natural systems (13–16). Recent advancements in microscopic and molecular-scale tools and data analysis have made possible investigation of trace element speciation in natural samples (17). These powerful modern techniques allow for precise determination of Cu distribution and speciation in soil. Bulk X-ray absorption fine structure (XAFS) spectroscopy can provide information about the bonding of Cu to soil components; however interpretation may be difficult because the bulk XAFS spectrum represents an average of multiple phases that may be present. Microscopically focused XAFS holds promise for overcoming this limitation because it allows for analysis of Cu in micron-size spots, which may represent single Cu phases (end-members). Flogeac et al. (18) used EPR and XAFS to investigate Cu adsorption mechanisms on calcareous soils spiked with Cu in the laboratory. Based on the XANES and EPR spectra, and the first-shell coordination environment derived from fitting the EXAFS spectrum, they concluded that the Cu was complexed to SOM coated minerals. Frenkel and Korshin (19) measured XAFS spectra on a Cu-spiked soil and proposed that the Cu was coordinated primarily to SOM. Liu and Wang (20) used XAFS spectroscopy to analyze a Cu-contaminated soil collected from a circuit board recycling plant. They used linear combination fitting of the XANES spectrum and firstshell fitting of the EXAFS spectrum and proposed that the Cu in the soil existed as Cu-SOM, CuCO3, Cu2O, and CuO, with Cu-SOM predominating. While XANES and EXAFS spectra in these three studies on Cu speciation in soils were shown to be useful for interpreting Cu speciation, the precise molecular structure of Cu in the soils has not been determined because of limited analysis of the second shell structure in the soil EXAFS and the averaging nature of bulk XAFS spectra. In the current study Cu speciation in the long-term contaminated soil from Hubbardsville, New York is reinvestigated using µ-SXRF element mapping and both bulk and microscopic XAFS. Due to the intense X-rays generated by synchrotron sources, these techniques allow for precise determination of Cu distribution and speciation in the soil. Specific research objectives for the contaminated soil are (1) determine the Cu and correlated elemental distribution, (2) model the molecular structure based on XANES and EXAFS spectra, and (3) compare our results to previous results obtained using best-available techniques two decades ago (4). The new insights into Cu speciation afforded by the synchrotron methods will allow for better risk analysis and improved remediation strategies, and help explain results from recent studies that have investigated Cu lability and bioavailability in the soil (5). VOL. 42, NO. 1, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
37
Methods and Materials Soil Sample. A sample of Cu-contaminated soil from a farm near Hubbardsville (near the town of Hamilton), New York, was kindly provided by Dr. Murray McBride of Cornell University. Detailed sampling and site characteristics are described in McBride and Bouldin (4). In brief, the soil is characterized as a Howard series (Mesic Glossic Hapludalf) (21), and contains free carbonates with a pH of 7.1. The soil was collected from a cultivated field in which residents observed poor crop growth due to Cu toxicity for more than 50 years before samples were collected (4). The Cu contamination source is presumed to be CuSO4, which was used as a fungicide on potato crops in the area in the first half of the 20th century (22). We further characterized the soil chemistry and mineral content. Mineralogy was determined by XRD (Siemens D5000 with Cu anode and Bruker Diffracplus Evaluation program version 10 for mineral identification). Total SOM was measured by rapid dichromate oxidation (23). Total concentration of 35 elements was analyzed by HNO3-HClO4-HF-HCl digestion and ICP-AES (Acme Laboratories, Vancouver, BC). Sample Preparation. The less than 2-mm size fraction of the Hubbardsville soil was analyzed by bulk XAFS. For the microprobe analysis the less than 50 µm-size fraction was separated by sieving and mounted as both a thin section and powder dispersed on Scotch tape. Thin section and standards preparation are described in the Supporting Information (SI). X-ray Spectroscopy. Synchrotron-based microprobe experiments were done on Beamline 2-3 at Stanford Synchrotron Radiation Laboratory (SSRL). The incident beam energy was 14000 eV. Details of beamline instrumentation are described in the Supporting Information. µ-XANES data from 10 points of interest were collected; however, signal-to-noise was poor for all except two spots, which are presented in this paper. The element distribution data are presented as tricolor maps that allow for the spatial distribution of three elements to be shown. Pixel brightness is displayed in RGB. Pixel brightness is directly correlated to fluorescence intensity. Absolute concentrations of elements cannot be accurately determined from the µ-SXRF data on these samples. Acquisition of the Cu K-edge (8979 eV) bulk XAFS data for the soil and Cu adsorbed on goethite, calcite, and HA standards was done on beamline 10-2 at SSRL. Copper XAFS for the Cu(OH)2, malachite, and aqueous Cu(NO3)2 were collected on beamlines 4-1 and 4-2 at SSRL (7). A Cu-foil was used in all experiments to calibrate the beam energy to 8979 eV. Beamline 10-2 monochromator consists of two parallel Si(220) crystals with an entrance slit of 1 mm. The beam was detuned by 40% for these measurements to minimize harmonics. Step size through the XANES region was 0.35 eV. The spectra were collected at room temperature (∼298 K). Fluorescence was detected using a Lytle or 13-element Ge detector. For XAFS data analysis several scans were merged and calibrated. Theoretical backscattering phase and amplitude functions for backscatterers were calculated using the FEFF8.0 code, with an input file based on a bidentate Cu chelate with carboxyl functional groups modeled after the Cu five ring chelate structure used by Karlsson et al. (24). Fitting of the EXAFS spectra was done using the WinXAS program (version 3.1) (25). Additional details of EXAFS data processing and fitting are described in the Supporting Information.
Results Bulk Soil Mineralogy and Elemental Content. Analysis of the XRD data indicated that the Hubbardsville soil contains calcite, dolomite, quartz, feldspar (albite), chlorite, jacobsite, and Illite. The total SOM content is 6.75%. The metal contents measured in a split of our sample were 4167 mg kg-1 Cu, 278 38
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 1, 2008
mg kg-1 Zn, 57 mg kg-1 Cr, 1.4 mg kg-1 Cd, and 260 mg kg-1 Pb. Dolomite and calcite were verified by optical examination of grain mounts and thin sections in the less than 50 µm-size fraction that was mapped by µ-SXRF. Copper Distribution by µ-SXRF. The Hubbardsville µ-SXRF soil maps show that Cu is distributed heterogeneously in the soil (Figure 1). Despite the high Cu concentration in the soil, only two spots had high enough Cu counts to collect XAFS data. The scarcity of Cu hot spots was also verified using elemental mapping with SEM-EDX. Three separate µ-SXRF element maps were collected, but only one representative map is shown in this paper. The brightest points in the maps are only a few pixels across, indicating that a significant fraction of Cu in the soil is composed of small Cu-rich particles. Larger regions of moderate Cu concentrations are also present. The highest Cu counts are not correlated with any other mapped elements (Figure 1 and SI Figure S-1). Sulfur, Si, and P have low-energy XRF, thus detection of these elements in the µ-SXRF spectrum on beamline 2-3 is limited to pixels with high concentrations. µ-SXRF maps of S, Si, and P (SI Figures S-3 and S-4) indicated that there are pixels elevated in these low-z elements, which likely correspond to phosphate, silicate, and sulfate minerals. Clay minerals can be estimated by the distribution of K and Si, using the ratio of K to other elements such as Ca, S, and P to rule out K present in other phases (26). The lack of high peak counts of S, Si, K, and P peaks in the µ-SXRF maps for the pixels with the highest Cu fluorescence (SI Figure S-4) suggests that the Cu phases are not Cu silicate, phosphate, sulfate, or clay minerals. Similarly, the lack of significant Ca, Fe or Mn fluorescence at the pixels with the highest Cu counts indicates that the Cu is not associated with metal oxides or carbonate phases of these metals. Remaining potential Cu solid phases are Cu (hydr)oxide, Cu carbonate, or Cu sorbed onto SOM. A subpopulation of pixels with moderate Cu fluorescence intensity relative to the high Cu intensity pixels was linearly correlated with Mn (SI Figure S-2). All the pixels with relatively high Mn fluorescence also had some Cu fluorescence, but not all the pixels that had Cu fluorescence had Mn fluorescence. In fact, the highest Cu counts were from pixels that had relatively low Mn counts. Iron was also present at the pixels with high Mn fluorescence. The codistribution of Cu, Mn, and Fe suggests that some Cu is associated with Fe-Mn oxides, or that Mn and Fe are sorbing onto another phase, such as SOM, in conjunction with Cu. No other elements are present at high concentrations in the Mn and Cu-rich subpopulation of pixels. Bulk and Micro XAFS. The bulk and µ-XANES spectra for the Hubbardsville soil and standards are shown in Figure 2. Copper K-edge XANES spectral features are very sensitive to molecular coordination (27, 28). Thus, interpretation of the XANES spectra can be used to determine molecular structure of Cu in soil samples. The pre-edge peak at 8976 eV is due to the 1s to 3d electron transition, and is invariant in the Cu XANES spectra in this study. The shoulder between 8979 and 8984 eV in the derivative spectra results from the 1s to 4pz electron transition in conjunction with a ligand to metal charge transfer (27, 28). The third and fourth XANES features in the Cu K-edge spectra are the 1s to 4p main-edge electron transitions (labeled R and β in Figure 2). Splitting of the mainedge peak results from anisotropic symmetry of Cu(II) compounds (Jahn–Teller distortion). The R peak arises from the 1s to 4pz transition and the β peak results from the 1s to 4px/py transitions (27, 28). The energy separation and intensity of the R and β peaks is dependent on molecular composition and coordination, and can be used to qualitatively assess Cu speciation in the soils. The spectra for several Cu(II) standards that are possible end-members for Cu in the soil are shown in Figure 2; each
FIGURE 1. Element map of Hubbardsville soil thin section. Color assignments: red ) copper, green ) calcium, blue ) manganese. Pixel size is 5 µm. The total map area is 0.870 × 0.875 mm. Magenta shades indicate Cu-Mn-rich particles. The two labeled points show the regions where µ-XAFS spectra were collected. The inset graphs show the multichannel analyzer (MCA) spectra of these regions. In each inset graph, the blue curve shows the summed MCA spectra of the high-Cu region and the red curve shows the individual MCA spectrum of the highest-Cu pixel in the region. spectrum has unique features. When comparing the XANES spectrum from the Hubbardsville soil to the Cu sorbed on Elliot HA standard there is an exact match of the R and β peaks. Fitting of the Hubbardsville soil XANES using the Cu sorbed on the Elliot HA standard XANES (Figure 2) had a residual of 8.9%. Using Cu(OH)2 XANES spectrum as a standard the fit residual was 15%. Fitting the soil XANES spectrum using both the Elliot HA and Cu(OH)2 spectra resulted in 90% Elliot HA and 7% Cu(OH)2, with only a small improvement in fit residual (8.6%). Thus we conclude that the best fit for the XANES spectra from the bulk Hubbardsville soil is the Cu sorbed on Elliot HA standard. The XANES spectrum from spot 2 in the Hubbardsville soil is very similar to the bulk Hubbardsville soil, suggesting that this spot is a dominant end-member for the Cu in the soil. Knowing that this is a dominant end-member allows us to extrapolate the elemental associations from Spot 2 to the Cu speciation in the bulk soil. The XANES from spot 1 is distinct from the bulk and spot 2 Hubbardsville spectra. There is a significant peak at 8982 eV (8980 eV in first derivative) in the spot 1 spectrum that is distinct from any peaks observed in the XANES spectra from the Cu(II) compounds. This peak occurs where Cu(I) oxidation would be expected (8981.5 eV in CuCl salt (29)). Tracking the size of this peak over sequential scans showed growth, suggesting that radiation-induced reduction of Cu(II) was occurring. We also observed a very small amount (
