Published online 29 June 2007
Published in Soil Sci Soc Am J 71:1288-1291 (2007)
DOI: 10.2136/sssaj2007.0007
© 2007 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
SOIL CHEMISTRY
Experimental Validation of Quantitative XANES Analysis for Phosphorus Speciation
Babasola Ajiboyea,
Olalekan O. Akinremia,* and
Astrid Jürgensenb
a Dep. of Soil Science, Univ. of Manitoba, 13 Freedman Crescent, Winnipeg, MB Canada R3T 2N2
b Canadian Synchrotron Radiation Facility, Synchrotron Radiation Center, Univ. of Wisconsin-Madison, Stoughton, WI 53589-3097
* Corresponding author (Akinremi{at}ms.umanitoba.ca).
 |
ABSTRACT
|
|---|
The quantitative approach used in x-ray absorption spectroscopy (XAS) experiments is oftentimes based on statistical goodness-of-fit criteria, which do not explain the accuracy of the components obtained from the fittings. This study was performed to validate the linear combination (LC) approach used in quantitative XAS analysis by estimating the accuracy of this procedure. Near-edge K
1 fluorescence XAS spectra were acquired for known binary mixtures of Ca, Al, and Fe phosphates in varying proportions and for the individual compounds. All combinations of the spectra of model compounds were fitted to the spectra of the known mixtures to obtain their relative abundance. The binary combinations produced the best fit with 
2 values ranging from 0.02 to 0.25. The relative error associated with the fitting ranged from as low as 0.8 to 17% for thoroughly mixed samples. The relative error was small when the proportion of Ca phosphate in the mixture was high but the error was large at low abundance of this component in the mixture. Because the interpretation of the XANES result largely depends on the relative proportion of species in the sample obtained by LC, we therefore recommend acquiring a spectrum for a mixture of certified reference compounds that mimics the composition of the sample being investigated at the beamline to estimate the accuracy of the proportions obtained from quantitative x-ray absorption near-edge structure (XANES) analysis.
Abbreviations: HAP, hydroxyapatite • LC, linear combination • PSIDER, phosphosiderite • VAR, variscite • XANES, x-ray absorption near-edge structure • XAS, x-ray absorption spectroscopy
|
INTRODUCTION
|
|---|
X-ray absorption spectroscopy (XAS) is fast becoming a versatile tool for the speciation of P in complex environmental samples. The near-edge region of XAS spectra, also known as x-ray absorption near-edge structure (XANES), is especially sensitive to the local structural and electronic environment of the P atom and can distinguish between precipitated and adsorbed phases of P (Hesterberg et al., 1999; Beauchemin et al., 2003; Sato et al., 2005). Studies have shown that each P species has its own fingerprints in the XANES region (Hesterberg et al., 1999; Khare et al., 2004), and the spectra of two or more P species can be used quantitatively to reveal the abundance of these species in a sample. Quantitative speciation using XANES data is highly dependent on the quality of the data and how well the chosen standards match the real species in the samples of unknown composition (Pickering et al., 1995; Beauchemin et al., 2003). In addition, the spectral features of identified components must be clearly distinguishable from one another to perform a reliable quantitative XANES analysis.
The quantitative approach used in the speciation of P and other elements in XAS usually involves principal component analysis (PCA) combined with target transformation (TT), and least-squares linear combination (LC) fitting (Beauchemin et al., 2002, 2003; Ressler et al., 2005). The PCA uses a multivariate statistical procedure to identify the number of independent orthogonal components that constitute the sample spectra, while TT identifies the actual chemical species. The LC can be used as a stand-alone procedure for quantitative XANES analysis or in conjunction with PCA and TT to reduce the number of reference compounds needed in the fitting, which could be a big time saver. The LC is usually done by fitting the reference spectra directly to unknown spectra while adjusting the fraction of each component assumed in the fit model and energy offset (O'Day et al., 2004). Beauchemin et al. (2003) reported that PCA lacked sensitivity to P 1s XANES data in that the maximum number of chemical species that can be accepted based on TT depends on the number of orthogonal components considered sufficient to reconstruct the XANES data in a reduced space. In fact, they achieved the best characterization of their soils by using ternary combinations of P standards in the LC fitting, contrary to the binary combination suggested by the result from PCA analysis. In using S XANES to characterize chemical species of humic acid in soil, Beauchemin et al. (2002) also reported that the number of potential targets (identified by TT) was more than the number of dominant components indicated by PCA. In a recent study on P speciation in fertilized soil with granular and liquid monoammonium phosphate, only two of the three species accepted by TT (SPOIL value <3) were used in the LC fitting because PCA indicated that only two orthogonal components are sufficient to reconstruct the P XANES data (Lombi et al., 2006). Therefore, using either a binary or ternary combination of P compounds in the LC fitting may affect the LC result and the conclusion made from such a procedure. For example, in the study by Lombi et al. (2006), the relative contribution of the third component that was discarded could be significant and may have a different implication from an environmental perspective.
In using LC for quantifying the relative proportion of reference standards in the XANES spectra of the sample, very few studies have examined the accuracy of the estimated proportion of standard compounds in known mixtures of these compounds (O'Day et al., 2004). In fact, no study has explored the accuracy of the quantitative methods of XAS analysis of P using known mixtures. The goodness-of-fit criteria reported for the LC fitting are usually statistical measures of error, which do not explain whether the components assumed in the fits are correct or not. This systematic error has not been well characterized and is difficult to estimate by statistical measure alone (Sayers, 2000; O'Day et al., 2004). Therefore, an empirical measure of accuracy of the XANES fit of known components may be necessary as a first step in using LC for quantitative XANES analysis to obtain a reliable fit. This will serve as a more robust estimation of the error in using LC fitting than the goodness-of-fit criteria obtained for the LC fitting.
The objective of the study was to use known mixtures of P compounds as a first attempt to validate the LC fitting procedure used in quantitative XANES analysis as a means of estimating its accuracy. A more intensive study involving additional mixtures of P compounds found in acidic and alkaline soils may be undertaken in the future.
|
MATERIALS AND METHODS
|
|---|
Three classes of P compounds commonly found in the environment were selected for this study. The compounds were: CaP as hydroxyapatite [HAP, Ca10(PO4)6(OH)2]; AlP as variscite (VAR, AlPO4·2H2O); and FeP as phosphosiderite (PSIDER, monoclinic FePO4·2H2O). Reagent-grade samples of these compounds were ground with mortar and pestle and mixed in a gram-weight proportion equivalent to [HAP]x + [VAR]1–x, and [HAP]x + [PSIDER]1–x with x = 0.25, 0.5, and 0.75. The relative contribution of P, based on atomic mass fraction, by the individual reference compounds constituting the mixtures varied. The total P in the mixtures was similar, however, with 
19% (w/w) in the HAP + VAR mixtures and 17 to 18% (w/w) in the HAP + PSIDER mixtures (Table 1). The mixtures and pure standard compounds were diluted with BN to bring the P content to 5% (w/w) and then spread as a thin film on a double-sided C tape to minimize self-absorption, which may distort fluorescence measurements. The XANES spectra of the mixtures and individual P compounds were acquired using the DCM beamline of the Canadian Synchrotron Radiation Facility at the Synchrotron Radiation Center, University of Wisconsin-Madison, operating at 800 MeV ring energy and maximum beam current of 230 mA. Standard operating parameters for this beamline have been explained in detail elsewhere (Lombi et al., 2006). Duplicate spectra of K1 fluorescence emission, involving the ejection of a 1s electron into the continuum and simultaneous filling of the core hole with a 2p3/2 electron, were recorded from 2140 to 2200 eV for each mixture.
View this table:
[in this window]
[in a new window]
|
Table 1. Binary mixtures of hydroxyapatite (HAP) and variscite (VAR) or phosphosiderite (PSIDER) showing the total P in each fraction of the mixtures.
|
|
The energy range of the spectra was recalibrated to correct for any edge shift in the spectra of the standards used in calibrating the beamline. This was done by using the edge energy offset between the energy of the white line of Na4P2O7 analyzed with the sample and that recorded during the calibration of the beamline. This offset value was used as a preprocessing parameter for all other spectra. The FY data were averaged and background corrected by a linear regression fit through the pre-edge region and a cubic spline through the post-edge region. The spectra were then normalized to a unit edge jump. All the data reduction and XANES analyses were performed using Athena Version 0.8.049 (Ravel and Newville, 2005). Linear combination fitting was performed without a priori assumptions about the number of components in the mixtures by using all combinations of the three model spectra (HAP, VAR, and PSIDER). The fitting was performed across the relative energy range of –11 to 49 eV. During the LC fitting, the threshold energy, E0, was allowed to vary, but the maximum energy shift observed for any of the components in the mixture was much less than the step size (0.25 eV) used during data acquisition. The goodness-of-fit was judged by the residual factor (R factor) and 2 values; the fit with the least R factor and 2 and the least standard deviation of the estimated proportion was chosen as the most likely fit.
|
RESULTS AND DISCUSSION
|
|---|
The LC fit of known binary mixtures of P standard compounds are shown in Fig. 1. The various proportions of the standard compounds and the overall LC fit are plotted together with normalized XANES data acquired for the mixture. Visual inspection confirmed that the LC is a good representation of the XANES data (Fig. 1). The statistical goodness-of-fit value ranges from 0.15 to 0.25 for different combinations of HAP and VAR mixtures and from 0.02 to 0.09 for HAP and PSIDER mixtures (Table 2). The mixing efficiency of these P standards, though not verified, was assumed to be within the range for dry mixing of two-powder systems, which was reported to be between 0.62 and 0.82 (Chen and Yu, 2004). The relative error between the theoretical proportions of the standards in the mixtures and proportions obtained by LC fitting of the XANES spectra ranged from as low as 0.8 to 17% for the HAP + VAR mixtures and from 2.9 to 41.6% for the HAP + PSIDER mixtures (Table 2). The relative error was small when the proportion of HAP in the mixture was high, but the error was large at low abundance of this component in the mixture (Table 2). It should be noted that the XAS spectrum of HAP has rich features, especially in the near-edge region, and this result suggests that such features are easily optimized for in the LC fitting when they are more abundant than other component features in the mixture. Although the 25% HAP + 75% PSIDER mixture had the greatest relative error, the LC fit was almost perfect with the best statistical goodness-of-fit (Fig. 1d, Table 2). There are some explanations for this observation. First, it is possible that the mixing efficiency was not perfect, which in turn resulted in hot spots of PSIDER in the mixture analyzed by XAS. Another reason may be the abundance of PSIDER in the mixture coupled with similarity in the energy position of the post-edge resonance occurring at approximately 2162 eV in both PSIDER and HAP, and the lack of a shoulder feature in the XANES data of the mixture. This makes the pre-edge feature the only structure that was optimized during the LC fitting, hence the overestimation of PSIDER. In addition, the low proportion of HAP in the mixture and the absence of a feature that distinguishes it from PSIDER in the post-edge region may be limiting the quantitative XANES analysis. Studies have shown that quantitative XANES cannot detect a trace of one chemical species of an element in the presence of a large excess of another species of the same element (Pickering et al., 1995, 2000) or, for that matter, another element with similar XANES features.
View larger version (30K):
[in this window]
[in a new window]
|
Fig. 1. Linear combination (LC) fittings of binary mixtures of phosphate standards containing (a) 25% hydroxyapatite (HAP) + 75% variscite (VAR), (b) 50% HAP + 50% VAR, (c) 75% HAP + 25% VAR, (d) 25% HAP + 75% phosphosiderite (PSIDER), (e) 50% HAP + 50% PSIDER, and (f) 75% HAP + 25% PSIDER.
|
|
View this table:
[in this window]
[in a new window]
|
Table 2. Summary of linear combination (LC) fittings of XANES spectra of binary mixtures of hydroxyapatite (HAP) and variscite (VAR) or phosphosiderite (PSIDER) used in the experiment.
|
|
Because the interpretation of the XANES result depends on the relative proportion of species in the sample obtained by LC, we propose that the accuracy of the fitting should be reported alongside the fit result. This can be accomplished by acquiring the XANES spectrum for a mixture of certified reference compounds that mimics the composition of the sample under investigation at the beamline. Since this is a first attempt to validate the LC fitting, future investigations should include mixtures of P compounds commonly found in a wide range of soil pHs. Such investigations may include an analysis of mixtures of loosely adsorbed and aqueous phosphates, additional binary mixtures involving AlP and FeP, ternary and more complex mixtures like freeze-dried insoluble minerals and a wet "sludge" mixture of insoluble minerals, and a phosphate solution in a liquid cell. Other intensive studies to determine the accuracy and precision limit of the XANES technique are also needed. In addition, the combination of XANES with other solid-state analytical techniques like synchrotron radiation based x-ray diffraction and solid-state nuclear magnetic resonance should be applied whenever possible.
|
CONCLUSIONS
|
|---|
This study examined the accuracy of LC fitting of reference standards to samples of known composition. Overall, this result confirms that the relative proportion of different species in a mixture can be obtained from an LC fit of XANES data with some confidence. The relative error associated with the fitting ranged from as low as 0.8 to 17% for thoroughly mixed samples. An unusually large relative error was associated with a particular mixture involving Ca and Fe phosphates, which suggested that the capability of quantitative XANES in detecting a species of small proportion with no distinct spectral feature in comparison with other species in the post-edge region may be limited. We therefore recommend that a mixture of certified reference materials similar to the sample being investigated should be analyzed during the XAS experiment to estimate the accuracy of the proportions obtained from quantitative XANES analysis.
|
ACKNOWLEDGMENTS
|
|---|
The support of the National Science Foundation under award no. DMR-0537588 to the Synchrotron Radiation Center (SRC), University of Wisconsin-Madison, where part of this research was conducted, is duly acknowledged. The Canadian Synchrotron Radiation Facility is funded by the Open Access Grant of the Natural Sciences and Engineering Council of Canada and the National Research Council Canada, and these supports are duly acknowledged. Supports by the University of Manitoba via a graduate fellowship and SRC via a travel supplement to B. Ajiboye are duly acknowledged.
|
NOTES
|
|---|
Present address: Canadian Light Source Inc., Univ. of Saskatchewan, 101 Perimeter Rd, Saskatoon, SK Canada S7N 0X4
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Permission for printing and for reprinting the material contained herein has been obtained by the publisher.
Received for publication January 4, 2007.
|
REFERENCES
|
|---|
- Beauchemin, S., D. Hesterberg, and M. Beauchemin. 2002. Principal component analysis approach for modeling sulfur K-XANES spectra of humic acids. Soil Sci. Soc. Am. J. 66:83–91.[Abstract/Free Full Text]
- Beauchemin, S., D. Hesterberg, J. Chou, M. Beauchemin, R.R. Simard, and D.E. Sayers. 2003. Speciation of phosphorus in phosphorus-enriched agricultural soils using x-ray absorption near-edge structure spectroscopy and chemical fractionation. J. Environ. Qual. 32:1809–1819.[Abstract/Free Full Text]
- Chen, C., and C. Yu. 2004. Two-dimensional image characterization of powder mixing and its effects on the solid-state reactions. Mater. Chem. Phys. 85:227–237.[CrossRef]
- Hesterberg, D., Z. Weiqing, K.J. Hutchison, S. Beauchemin, and D.E. Sayers. 1999. XAFS study of adsorbed and mineral forms of phosphate. J. Synchrotron Radiat. 6:636–638.[CrossRef][Web of Science][Medline]
- Khare, N., D. Hesterberg, S. Beauchemin, and S.L. Wang. 2004. XANES determination of adsorbed phosphate distribution between ferrihydrite and boehmite in mixtures. Soil Sci. Soc. Am. J. 68:460–469.[Abstract/Free Full Text]
- Lombi, E., K.G. Scheckel, R.D. Armstrong, S. Forrester, J.N. Cutler, and D. Paterson. 2006. Speciation and distribution of phosphorus in a fertilized soil: A synchrotron-based investigation. Soil Sci. Soc. Am. J. 70:2038–2048.[Abstract/Free Full Text]
- O'Day, P.A., N. Rivera, Jr., R. Root, and S.A. Carroll. 2004. X-ray absorption spectroscopic study of Fe reference compounds for the analysis of natural sediments. Am. Mineral. 89:572–585.[Abstract/Free Full Text]
- Pickering, I.J., G.E. Brown, and T.K. Tokunaga. 1995. Quantitative speciation of selenium in soils using x-ray absorption spectroscopy. Environ. Sci. Technol. 25:2456–2459.
- Pickering, I.J., R.C. Prince, D.E. Salt, and G.N. George. 2000. Quantitative, chemically specific imaging of selenium transformation in plants. Proc. Natl. Acad. Sci. 97:10717–10722.[Abstract/Free Full Text]
- Ravel, B., and M. Newville. 2005. ATHENA, ARTEMIS, HEPHAESTUS: Data analysis for x-ray absorption spectroscopy using IFEFFIT. J. Synchrotron Radiat. 12:537–541.[CrossRef][Web of Science][Medline]
- Ressler, T., J. Wong, J. Roos, and I.L. Smith. 2005. Quantitative speciation of Mn-bearing particulates emitted from autos burning (methycyclopentadienyl) manganese tricarbonyl-added gasoline using XANES spectroscopy. Environ. Sci. Technol. 34:950–958.
- Sato, S., D. Solomon, C. Hyland, Q.M. Ketterings, and J. Lehmann. 2005. Phosphorus speciation in manure and manure-amended soils using XANES spectroscopy. Environ. Sci. Technol. 39:7485–7491.[Medline]
- Sayers, D.E. (chair). 2000. Report of the International XAFS Society Standards and Criteria Committee. (Available at fisica.unicam.it/IXS/OLD/subcommittee_reports/sc/SC00report.pdf; verified 15 May 2007). Int. XAFS Soc., Camerino, Italy.
TOP
ABSTRACT
INTRODUCTION
