ARTICLE pubs.acs.org/est
Sulfur Speciation in Soil by S K-Edge XANES Spectroscopy: Comparison of Spectral Deconvolution and Linear Combination Fitting J€org Prietzel,* Anna Botzaki, and Nora Tyufekchieva Technische Universit€at M€unchen, Lehrstuhl f€ur Bodenkunde, D-85354 Freising, Germany, Emil-Ramann-Strasse 2
Mareike Brettholle and J€urgen Thieme† Georg-August-Universit€at G€ottingen, Institut f€ur R€ontgenphysik, D-37077 G€ottingen, Germany, Friedrich-Hund-Platz 1
Wantana Klysubun Synchrotron Light Research Institute, Nakhon Ratchasima 30000, Thailand, Muang District, 111 University Avenue
bS Supporting Information ABSTRACT:
Defined, quartz-diluted mixtures of sulfur (S) compounds with different oxidation state (OS) were analyzed by K-edge XANES spectroscopy using linear combination fitting (LCF) and spectrum deconvolution by fitting several Gaussian and arctangent functions (GCF). Additionally, for different soils the S speciation as calculated by both methods was compared with results of a wetchemical S speciation. For mixtures of FeS, L-cysteine, and Na2SO4, the S speciation was recovered with satisfactory accuracy and precision by both methods at the 2 and 0.2 mg S g1 level. For GCF, white-line peaks must be normalized with respect to their OSspecific absorption cross-section. LCF must be conducted with dilute reference compounds to avoid self-absorption effects. For mixtures of FeS, FeS2, S°, and L-cysteine, both procedures showed poor accuracy. For the soils, similar percentages of reduced inorganic S, organic S, and sulfate were calculated by LCF, GCF, and wet chemical S speciation. GCF allows a fair estimation of S species groups with different OS (inorganic reduced S, organic reduced S, organic intermediate S, oxidized S) in soils without standards. If dilute standards of all S compounds assumed to be present in a sample are available, LCF is more objective and allows a more detailed S speciation.
’ INTRODUCTION Sulfur (S) in environmental samples exists in a large variety of organic and inorganic forms with different electronic oxidation states (OS), ranging from 2 (inorganic sulfide) to þ6 (sulfate). This makes traditional wet-chemical S speciation complicated, and prone to artifacts and biases.1,2 In recent years, synchrotron-based X-ray absorption near-edge structure (XANES) r 2011 American Chemical Society
spectroscopy has emerged as powerful, noninvasive method for a direct speciation of various elements in environmental Received: June 30, 2010 Accepted: March 2, 2011 Revised: February 28, 2011 Published: March 15, 2011 2878
dx.doi.org/10.1021/es102180a | Environ. Sci. Technol. 2011, 45, 2878–2886
Environmental Science & Technology samples.3,4 The method is particularly sensitive to distinguish among different OS of an element; therefore, it has become especially attractive for the speciation of S in soil and other environmental samples, most prominently at the S K-edge.1,3,513 The uncertainty of quantitative determination of S species by S K-edge XANES spectroscopy in such samples has been reported (e.g., refs 6,7,1416) as (510%. However, this estimation was inferred from studies on coal or asphaltenes with S contents between 5 and 50 mg g1. For soils, which generally have smaller S concentrations (0.15 mg S g1), until now no specific error assessment has been conducted. Particularly, it has not been investigated which of the two widely used methods of spectrum deconvolution: (i) Linear combination fitting (LCF; 810,14,15,17,18), and (ii) Gaussian curve fitting (GCF; 1,2,6,12,13), modeled by several Gaussian (white-lines) and arc tangent functions (edge steps), is more accurate and precise in determining the S speciation in soils. The white-line energies of S species with OS 3) S species, respectively) instead of specific arc tangent functions for each identified S species were used in the GSF procedure, as done in earlier studies,1,12 recovery deteriorated (mean difference of calculated from real percentage: ( 3.7% for sulfate, ( 6.1% for FeS and (8.5% for L-cysteine); moreover, sulfate was systematically overestimated at the expense of L-cysteine (Supporting Information Table S4). For set 2 (Table 3), GCF resulted in larger inaccuracies than LCF, on average about (16 up to (26% for different S species. As with LCF, L-cysteine was often overestimated, FeS and S° underestimated. Quantification of Different S Species in Soils. For several well-aerated and water-logged soil horizons with different humus 2880
dx.doi.org/10.1021/es102180a |Environ. Sci. Technol. 2011, 45, 2878–2886
Environmental Science & Technology
ARTICLE
Figure 1. Results of Linear combination fitting (left) and spectral deconvolution by Gaussian curve/arctangent function fitting and subsequent correction for absorption cross section (right) applied to spectra of (a,b) defined mixtures containing 33.3 mass percent sulfur as FeS, L-cysteine, and Na sulfate, respectively (a) 2000 mg S g1, (b) 200 mg S g1, and (c) a soil sample (Stagnosol He horizon).
content (spectra shown in Supporting Information Figure S5), similar percentages of reduced inorganic (=sulfide) S (OS < 0), organic S with different OS (0 to þ5), and sulfate (OS þ6) were calculated by LCF (Figure 1c; left panel), GCF (Figure 1c; right panel), and traditional wet chemical S speciation. Compared to wetchemical S speciation, XANES provided a more detailed speciation of organic S (Figure 2). With GCF, generally larger percentages of sulfate, sulfone and sulfoxide, and smaller percentages of sulfonate and organic disulfide were recovered than with LCF.
’ DISCUSSION For most reference compounds, normalized K-edge XANES spectra of subsamples with different dilution levels looked fairly identical. However, some exceptions from this general pattern were noticed. The different spectra of the dilute variants of dimethyl sulfone (Supporting Information Figure S1h) can be assigned to artifacts during grinding that compound with quartz in N2 atmosphere as identified by a typical smell which was absent for the concentrated variants. The sulfate peak noticed for
the dilute variants of Na sulfite (Supporting Information Figure S1g) can be explained by photo-oxidation during X-ray irradiation; also minor irregularities in the spectra of other dilute reference compounds were probably caused by radiation damage. Compared to the dilute subsamples of the reference compounds, undiluted subsamples of the same compound often showed damped or distorted white-lines. Such a distortion of the spectrum is caused by the self-absorption effect, which is known to occur for samples of high (>5 mass%) concentration.1,3,6,7,19 The edge-normalized spectra of the dilute variants (0.2, 2, in most cases also 20 mg S g1) in most cases had white-lines with similar magnitudes and shapes; thus self-absorption can be considered negligible for these samples. Our results show that self-absorption in S K-edge XANES analyses using fluorescence detection is insignificant, if samples and standards are fineground and their S concentrations are 20 mg g1 or less. This is in line with ref 5 which reported that self-absorption is no problem for thin samples with S concentrations
