Environ. Sci. Technol. 2002, 36, 5021-5028
Combining Selective Sequential Extractions, X-ray Absorption Spectroscopy, and Principal Component Analysis for Quantitative Zinc Speciation in Soil ANDREAS C. SCHEINOST,* RUBEN KRETZSCHMAR, AND SABINA PFISTER Institute of Terrestrial Ecology, ETH Zurich, CH-8952 Schlieren, Switzerland DARRYL R. ROBERTS Department of Physics, University of Ottawa, Ottawa K1N 6N5, Canada
Selective sequential extractions (SSE) and, more recently, X-ray absorption fine-structure (XAFS) spectroscopy have been used to characterize the speciation of metal contaminants in soils and sediments. However, both methods have specific limitations when multiple metal species coexist in soils and sediments. In this study, we tested a combined approach, in which XAFS spectra were collected after each of 6 SSE steps, and then analyzed by multishell fitting, principal component analysis (PCA) and linear combination fits (LCF), to determine the Zn speciation in a smelter-contaminated, strongly acidic soil. In the topsoil, Zn was predominately found in the smelter-emitted minerals franklinite (60%) and sphalerite (30%) and as aqueous or outer-sphere Zn2+ (10%). In the subsoil, aqueous or outersphere Zn2+ prevailed (55%), but 45% of Zn was incorporated by hydroxy-Al interlayers of phyllosilicates. Formation of such Zn-bearing hydroxy-interlayers, which has been observed here for the first time, may be an important mechanism to reduce the solubility of Zn in those soils, which are too acidic to retain Zn by formation of inner-sphere sorption complexes, layered double hydroxides or phyllosilicates. The stepwise removal of Zn fractions by SSE significantly improved the identification of species by XAFS and PCA and their subsequent quantification by LCF. While SSE alone provided excellent estimates of the amount of mobile Zn species, it failed to identify and quantify Zn associated with mineral phases because of nonspecific dissolution and the precipitation of Zn oxalate. The systematic combination of chemical extraction, spectroscopy, and advanced statistical analysis allowed us to identify and quantify both mobile and recalcitrant species with high reliability and precision.
Introduction Contamination of soils with trace metals is a worldwide problem, impeding the safety of food and drinking water, the production of food and fiber, and the stability of * Corresponding author phone: 41-1-633 61 47; fax: 41-1-633 11 18; e-mail:
[email protected]. Mailing address: ETH-ITO, Grabenstrasse 3, CH-8952 Schlieren, Switzerland. 10.1021/es025669f CCC: $22.00 Published on Web 11/02/2002
2002 American Chemical Society
ecosystems. The short-term toxicity of trace metals in soils depends on the amount and the chemical nature of the most mobile species, while the long-term toxicity is determined by the resupply of the mobile pool from more recalcitrant phases. Thus, quantitative speciation of trace metals as well as its variation with time is a prerequisite for long-term risk assessments. Trace metals may be retained by soil components through a number of processes such as electrostatic adsorption, formation of inner-sphere sorption complexes or multinuclear surface complexes, and precipitation of new mineral phases (1-3). Some or all of these retention processes occur simultaneously in soils, where different organic and organomineral substances, phyllosilicates and metal oxohydroxides provide reactive surfaces. Due to the heterogeneous distribution of voids, minerals, organic matter, microorganisms and roots, diverse chemical microenvironments may coexist on micrometer to millimeter scales, which may further increase the number of species in a given bulk sample (4-8). Despite their proximity, gradients between such microenvironments, and consequently the diversity of chemical species, are maintained for long time scales due to low solubilities, slow reaction kinetics, slow diffusion kinetics or passivated surfaces. Selective sequential extraction (SSE) procedures have been developed in the past to investigate multiple species, including physically and chemically sorbed metal ions, metals occluded in carbonates, Mn (hydr)oxides, crystalline and amorphous Fe (hydr)oxides and metal sulfides (9-12). In principle, SSE would allow for the identification and quantification of as many species as there are specific extraction steps. However, SSE is prone to a variety of shortcomings including the incomplete dissolution of a target phase (13, 14), the dissolution of nontarget species (15, 16), the incomplete removal of dissolved species due to readsorption (16-18) or reprecipitation (19), and the modification of oxidation state (17). Therefore, SSE fractions are only operationally defined through the extraction procedure and may or may not represent chemical species. XAFS spectroscopy has been used to identify chemical species in situ with minor or no pretreatments. The shortrange structural information extracted by multishell fitting is usually sufficient to determine the most predominant species in a sample. Under favorable conditions a second species may be identified (20-22). In the presence of minerals with strong second-shell backscattering, however, the soluble species with weak or missing second-shell backscattering are easily overlooked (23). Alternatively, XAFS spectra may be evaluated by fitting linear combinations of standard spectra to the unknown sample spectra (LCF). LCF has been successfully employed to identify and quantify up to three major species, including minerals and sorption complexes (16, 24, 25). Even when a large database of reference spectra is available and a good fit is achieved, however, there may always remain some doubt whether all species have been found and whether the fit gives a unique solution (26). Therefore, another statistical approach, principal component analysis (PCA), is increasingly used in combination with LCF (27-29). With this method, the number and the spectral characteristics of the principal components () species) can be determined even without an explicit database. There are two basic requirements for the success of PCA: first, the number of species present in a set of samples must be smaller than the number of sample spectra, and second, the species composition of these samples must vary. These requirements may be ideally fulfilled by varying the species composition VOL. 36, NO. 23, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5021
TABLE 1. Element Concentrations and pH of Palmerton Samples pH C S Mn Fe Zn Pb
g/kg g/kg g/kg g/kg mg/kg mg/kg
topsoil
subsoil
sediment
3.2 320 6.4 1.5 33 6200 7000
3.9 50 0.8 0.5 25 900 62
4.5 25 1.0 0.1 22 2500 406
of a sample by physical or chemical separation methods (30). Therefore, the purpose of this research is to systematically combine a 6-step SSE with X-ray absorption spectroscopy and then to evaluate the resulting XAFS data sets with PCA and LCF. We expected the following benefits from this combination: (I) The stepwise removal of metal fractions reduces the number of species, which eases the subsequent spectroscopic identification of the remaining species. (II) Having a set of varying spectra for each sample should help to identify number and kind of species by PCA. (III) SSE may be specifically suited to detect the most mobile species, which are most toxic. (IV) Finally, the SSE method itself can be validated by monitoring the success of SSE steps with XAFS spectroscopy (14, 19). We applied this approach to identify and quantify Zn species in a smelter-contaminated soil profile. Since we were using the same samples that have been investigated earlier using a range of conventional and microspectroscopic techniques and leaching experiments (31), we were able to determine the advantages and disadvantages of our approach.
Materials and Methods Soil Samples. The sampling site is located on Blue Mountain, about 1000 m SE of the former sphalerite smelter (Smelter II or East Plant) at Palmerton, Pennsylvania. The topsoil consists of a 10 cm thick layer of dark, dry organic debris of only partially decomposed plant residue (230 g/kg organic C); its decomposition is most likely prevented by the high content of heavy metals. The subsoil, 10 to 30 cm in depth, is a consolidated, skeletal silty loam (26% sand, 59% silt, 15% clay). Only shallow soil profiles such as this remained in erosion-protected depressions of the bedrock, after the elsewhere dense forest vegetation had been destroyed due to the intense deposition of sulfuric acid and heavy metals in 80 years of smelter operation. In addition, a sediment sample was collected from an artificial pond nearby, which was dry at the time of sampling (8/1/1999). Chemical properties of these samples are compiled in Table 1. The samples were air-dried, sieved to collect the
