Detailed explanations of XAS principles, techniques, and applications in soil science can be found in a number of excellent reviews (Bertsch and Hunter, 2001; Brown and Sturchio, 2002; Fendorf et al., 1994; Kelly et al., 2008) while details on the physics of XAS appear in several books (Koningsberger and Prins, 1988; Stửhr, 1992; Teo, 1986).
In a standard XAS measurement, the absorption spectra are typically collected by scanning the monochromator energy through the absorption edge of the element of interest and then recording the incident (I0) and absorbed (I) photon intensities with ion chambers placed before and after the sample. While this configuration directly measures the energy depen- dence of the absorption coefficient (à), à(E) = log I0/I, these excited atoms decay within a few femtoseconds via XRF or Auger emission. As these emissions are proportional to X-ray absorption, either of these processes can be used to measure the absorption coefficient, and in an XRF configu- ration à(E) ∝ If/I0, where If is the monitored fluorescence intensity associ- ated with the absorption process. A fluorescence detector with low-energy resolution and large-solid angle (e.g. a Lytle detector) is ideal for analysis of a single element that is the highest concentration detectable element in the sample, whereas analysis of a trace element in a chemically complex material typically requires a multi-element solid-state detector array. With an energy-discriminating detector, an XRF spectrum is recorded at each energy point in the XAS scan, and the fluorescence intensity for the ele- ment of interest is extracted, normalized to the incident beam intensity,
and plotted as a function of the incident energy. When full XRF spectra are preserved as a part of the scan at each energy step, Gaussian peak fit- ting of the fluorescence spectra can be used to mitigate effects of spectral overlap.
2.4.1. Anatomy of XAS Spectrum
An XAS spectrum is conventionally divided into two regimes: X-ray absorption near-edge structure (XANES) and EXAFS (Fig. 1.7). The energy region extending from 50 eV below the absorption edge to circa 200 eV above the absorption edge is the XANES portion of the spec- trum. Fingerprint information can be rich in this region. XANES provides information on valence state of a selected element, the local symmetry of its unoccupied orbitals, and in some cases, the molecular species by com- parison of measured spectroscopic data to a spectral library of known com- pounds. It is important to collect XANES data for the references on the same beamline, preferably during the same run, as used for the samples.
XANES EXAFS
Energy (eV)
Normalized absorbance Re
Figure 1.7 Ni K-edge àXAS spectrum of Alyssum murale leaf (collected from the primary vein) showing the region of the spectrum representing the XANES and EXAFS. Inset:
k2-weighted χ(k) EXAFS spectrum (left) and corresponding Fourier transform (including real and imaginary parts). For color version of this figure, the reader is referred to the online version of this book.
Since the binding energies of the valence orbitals are higher for more oxi- dized atoms, the energy position of the absorption edge and the pre-edge features are easily correlated with the valence state of the absorbing atom in the sample. The main step-like feature of the absorption spectrum is due to the excitation of the photoelectron into the continuum. The absorption edge is usually identified by the inflection point of this main absorption feature, and its position is dependent on the chemical environment of the absorbing atom. The position of the absorption edge generally increases by 1–3 eV for each electron removed from the valence shell due to the increas- ing binding energy of the core levels. However, the position of the absorp- tion edge is also influenced by the bonding environment of the absorbing atom such that spectra of two reference compounds containing an element in the same formal valence state can have slightly different edge positions (typically <1 eV).
The EXAFS part of the spectrum is the normalized oscillatory part of the absorption coefficient above the absorption edge to circa 800 eV or higher, which contains the critical information required to deter- mine the local coordination environment of the element of interest.
Beyond the edge, oscillations are observed which arise from interfer- ence effects involving the photoelectron wave ejected from the absorb- ing atom and the fraction of the photoelectron wave backscattered by atoms surrounding the absorbing atom. The frequency of the oscil- lations is inversely related to the bond distance between the absorber and neighboring atoms, and the amplitude is related to the number and identity of the neighboring atoms in a particular shell (i.e. group of atoms at a unique radial distance from the absorber). Fourier trans- formation of the oscillatory fine structure (obtained after background subtraction) yields a radial structure function (RSF) in real space with peaks revealing the local environment of the target atom (Manceau et al., 2002). A plot of the oscillatory fine structure and corresponding Fourier transform for a Ni K-edge àXAS spectrum collected from an Alyssum leaf is shown as an inset in Fig. 1.7.
2.4.2. Data Analysis for Complex, Mixed-Component Systems
Traditional shell-by-shell XAS analysis, involving Fourier transforming, backtransforming and filtering, generally does not work well for multi- component systems with a mixture of metal species. The atomic shells from the different species overlap, so that one cannot separate them out when there is a complex mixture (Manceau et al., 2002). One way to untangle
a complex mixture is to use a microprobe to examine multiple spots, and use PCA to determine how many independent components are needed to reproduce the spectral dataset. A primary goal of a àXAS experiment is to identify all the unique chemical forms for the element of interest. The PCA technique then determines if the dataset can be described as weighted sums of a smaller number of components (i.e. the principle components). Target transformation (TT) is used to identify the principle components by taking a spectrum of a candidate reference compound and mathematically remov- ing from it anything that is not reconstructed by the set of principle com- ponents identified by PCA (Malinowski, 1978). One can compare various models by fitting the unknown spectrum with different combinations of reference spectra and tabulating the fit results to compare the goodness-of- fit. Once the best-fit components have been identified, their proportion can be determined in each sample by linear combination fitting. It is important to note that the accuracy of this fitting approach is dependent upon the data quality, the completeness and relevance of the standards dataset, and the range over which the data were fit. It should also be emphasized that consistent data treatment for both standards and sample spectra is required.
Furthermore, this approach can be limited by the lack of unique spectral features for the candidate reference compounds, and it is well known that XAS techniques are inherently less sensitive to metals bound to lighter ele- ments (Sarret et al., 2004). In such a case, it may not be possible to constrain the exact speciation without additional data from ancillary measurements.
Linear combination fitting subroutines are available in XAS data analysis programs such as Athena (Ravel and Newville, 2005) and Sixpack (Webb, 2007). Additional discussions on XAS data analysis for heterogeneous sys- tems can be found in Wasserman (1997, 1999), Manceau et al. (2002), and Kelly et al. (2008).
2.4.3. Self-Absorption
An important consideration for fluorescence XAS measurements is the potential for self-absorption to occur with thick samples or samples that are highly concentrated in the absorbing element. When samples are too thick or concentrated, the penetration depth of the incident beam will vary as a function of à(E). As à(E) increases above the absorption edge, the penetration depth decreases and, thus, attenuates the oscillatory structure of the XAS spectra. Some algorithms exist for calculating the magnitude of such effects but these rely on precise knowledge of the density of all atoms in the path of the X-rays which is rarely available (standards are an
exception), thus it is generally preferable to avoid self-absorption effects than to attempt mathematical corrections to the data. However, if one has a relevant reference spectrum that is free of self-absorption effects, then a straightforward correction to an experimental XANES spectrum can be made using the equation given by Sarret et al. (2007) [ycorrected = yexperimental/ (1 + a(1 − yexperimental))], where the self-absorption parameter (a) equals 0 in the absence of a self-absorption effect and increases with this effect; as the precise composition of the sample is not known, “a” is adjusted iteratively to match the amplitude of the experimental spectrum with the standard. A detailed discussion of the origin of self-absorption (or “overabsorption”) is presented in Manceau et al. (2002).