Mössbauer spectroscopy

Determinations of iron oxidation state were integral to the work presented here, and all measurements were made with transmission Mössbauer spec-troscopy. This technique uses a radioactive source to produce gamma (γ) radiation, which is transmitted through the sample. Some of the γ rays are absorbed and re-emitted by atoms of the target material, while the remainder pass through the sample without interacting and are collected by a detector behind the sample (see figure 2.5). Therefore, the spectrum that is produced records dips in the count rate, at discrete energies that were absorbed (e.g., Dyar et al., 2006).

Atoms can only absorb and emit gamma rays at certain discrete, quantized energies, and so the source must emit radiation that is at the necessary energy

for the isotope of interest. In the case of iron, a source of⁵⁷Co is conventionally used. Radioactive sources emit monochromatic gamma rays; i.e., at a fixed energy. This value can, however, be modulated by moving the source (or the target) and thereby Doppler-shifting the emitted γ radiation. Thus, gamma rays with a range of energies can be transmitted through the sample, and the detector can record count rate as a function of the source velocity (e.g., McCammon, 2004; Dyar et al., 2006).

Figure 2.5: Cartoon of a Mössbauer spectrometer. The radioactive source is mounted on a drive, moving it relative to the sample and thereby Doppler-shifting the incident γ rays. Some of the atoms in the sample scatter the incident waves, so the detector behind the sample records only transmitted rays; the signal is in the absence of the rays that were scattered. Figure adapted from Dyar et al. (2006) and McCammon (2004).

Further, the energies that are absorbed can be modified (shifted or split) de-pending on the specific environment (coordination, number of electrons, sym-metry of the site) of the iron nucleus in the sample. This is a result of inter-actions between the nucleus and electrons, or the hyperfine interaction (e.g., Dyar et al., 2006). The result is a spectrum that can be used to determine valence state, spin state, and coordination of the target atoms.

Mössbauer spectra comprise sets of peaks, usually Lorentzian doublets and sextets, the shapes of which are dependent on the hyperfine parameters from every structural site occupied by iron in the sample. The primary hyperfine parameters used are the isomer centre shift and quadrupole splitting. The former is a shift of the peak energy absorbed (relative to α-Fe), and falls into a defined range as a result of coordination, valence state, and spin state

(Ban-croft et al., 1967; McCammon, 2004). The latter is a split of a single peak into a doublet, and reflects an asymmetry of valence electron charge distri-bution, as well as non-cubic symmetry in a crystalline lattice (Ingalls, 1964;

McCammon, 2004). Fitting a spectrum involves deconvolving the sum spec-trum into its component doublets and sextets, which are each then assigned to Fe³⁺ or Fe²⁺ depending on the hyperfine parameters determined from them (Figure 2.6). Fe3+ / ^PFe is then determined from the relative areas beneath the component spectra.

Figure 2.6: Assigning a component subspectrum to Fe2+ or Fe3+ is accom-plished with the values of the hyperfine parameters. Figure adapted from Dyar et al. (2006)

Spectra can be quite complicated, with several sets of overlapping peaks. It may not be possible to determine a unique solution (McCammon, 2004). For this study, this work of fitting the spectra was done with the MossA software (Prescher et al., 2012). This program allows for a visual inspection of the spectrum, enabling the user to define parameters of the subspectra that are physically plausible. Deviations in a subspectra could be accounted for by effects such as site distortion, preferred orientation, or a signal from iron in another crystallographic site (McCammon, 2004). The proper selection of a model can effect the final result, and care must be taken to account for not

only a good fit of the model to the data but ensuring that the result is phys-ically possible (Prescher et al., 2012). If possible, additional constraints from other analysis techniques (e.g. to determine the identity of the phases present) may be necessary.

Figure 2.7:Example of a Mössbauer spectrum collected for this study (chap-ter 5). Evident in the spectrum is a sextet from the iron in the FeS sulfide, as well as two doublets, assigned to Fe³⁺ (light green) and Fe²⁺ (dark green) based on their centre shift and quadropole splitting.

Many of the samples analysed for this study were quenched silicate glasses.

Glassy and other amorphous materials do not have a repeating structural array, so the sites that contain iron are therefore locally less regular and more distorted. This generally manifests as a broadening of the spectra obtained from amorphous samples. The hyperfine parameters obtained, however, arise from interactions of the iron nucleus with atoms up to two coordination shells out, and so structural information can still be obtained.

As for crystalline materials, the relative areas of the subspectra are used to de-termine the site abundances. Due to the broadening of the spectra, however, this procedure is more sensitive to an appropriate choice of fitting model. For example, a broad spectrum can be fit by adding successive Lorentzian lines

for multiple quadrupole doublets until a fit is achieved. Another approach assumes hyperfine parameters, and then solves for their probability (Vanden-berghe et al., 1994), thus finding the most likely set of Lorentzian doublets to fit the data. A third technique is built on the approximately Gaussian distribu-tions of hyperfine parameter probabilities, and so spectra are fit using sums of Gaussian (Rancourt and Ping, 1991). Whichever technique is used, the centre shift and quadrupole splitting of each doublet can then be used to determine relative abundance of Fe³⁺and Fe²⁺as described above for crystalline material (McCammon, 2004).

Broadening of the spectrum can also occur if the sample is not prepared ade-quately. The optimum thickness for a sample depends on the chemical compo-sition, as there are trade-offs between signal strength (for which more iron is better) and nonresonant scattering which creates noise. For the compositions used in this study (∼10 wt % FeO), samples were thinned to ∼ 500 µm, giving an apparent thickness of ∼ 10 mg Fe/cm² (Long et al., 1983). For this work, all spectra were taken at room temperature on a constant acceleration Möss-bauer spectrometer in transmission mode. The velocity scale was calibrated by collecting a spectrum from Fe foil at the same conditions of measurement.