Introduction

Determination of the redox state of iron and its spatial variations in sediments and rocks is of critical importance in both geosciences and environmental sciences, because of the need to understand redox state during their deposition or formation as well as subsequent changes in redox state due to weathering and other processes (e.g., de Andrade et al. 2006; Muñoz et al. 2006; Bernard et al. 2010; Benzerara et al. 2011a; Bolfan-Casanova et al. 2012; Stagno et al. 2013). In addition, quantification of Fe3+/ΣFe ratios can yield a better insight into the chemistry of complex geological materials (e.g., Muñoz et al. 2006), or a better estimation of P–T conditions by geothermobarometers, when variations of the Fe3+ content within the crystals are taken into account (e.g., Schmid et al. 2003; de Andrade et al. 2006; Bourdelle et al. 2013a). Therefore, assessment of the Fe3+/ΣFe ratio in minerals is an important and long-standing issue. Different techniques have been used extensively in the past for this purpose, including electron microprobe analysis (EMPA, e.g., Fialin et al. 2004), Mössbauer spectroscopy (e.g., Beaufort et al. 1992) X-ray photoelectron spectroscopy (XPS, e.g., Raeburn et al. 1997a, b) or X-ray absorption near edge structure (XANES) spectroscopy at the K edge (e.g., Waychunas et al. 1983; Bajt et al. 1994; Wilke et al. 2001, 2009; Berry et al. 2003, 2010). However, none of these methods provides spatial resolution at the few nanometres scale, which is particularly useful for studying chemical zonation patterns observed in low-temperature systems. Several studies (e.g., van Aken and Liebscher 2002) have shown that electron energy-loss spectroscopy (EELS) carried out in a transmission electron microscope (TEM) is a powerful method for determining the redox state of iron at a submicrometre resolution. However, it sometimes induces severe beam damage effects, such as electron beam-induced oxidation of iron (Lauterbach et al. 2000; Garvie et al. 2004), the effect of which can be corrected by measuring the signal as a function of time. Alternatively, XANES spectroscopy at the Fe L2,3 edges carried out with a scanning transmission X-ray microscope (STXM) has been increasingly used in the Earth and environmental sciences to infer qualitatively Fe3+/ΣFe ratios in geological and environmental samples at a spatial resolution of ~50 nm (e.g., Wasinger et al. 2003; Carlut et al. 2010; Lam et al. 2010; de Groot et al. 2010; Miot et al. 2011; Boulard et al. 2012). This technique has several advantages such as offering a high energy resolution (better than 0.1 eV at existing synchrotron facilities) and the possibility of maintaining samples under anoxic conditions before and during the measurement (e.g., Miot et al. 2009). However, no calibration of the STXM-based Fe L2,3-edge XANES approach has yet been carried out, whereas calibration of the EELS approach was quantified by van Aken and Liebscher (2002). Fe L2,3-edges result from 2p → 3d electronic transitions, as shown by Wasinger et al. (2003). These authors described in detail the physical basis of Fe L edges and showed that information about iron valency can be retrieved from XANES spectra by a multiplet calculation approach (e.g., van der Laan and Kirkman 1992; Cressey et al. 1993). This approach is difficult to apply when dealing with mineral phases for which we do not know the structure. Alternatively, fitting of XANES spectra with a linear combination of normalised reference spectra has been performed by Miot et al. (2009), but requires appropriate Fe2+ and Fe3+ end-member reference compounds with Fe in the same local coordination environment as in the sample of interest. Van Aken and Liebscher (2002) have shown the possibility of a third approach that they calibrated for EELS and which uses an empirical correlation between Fe3+/ΣFe ratios and a parameter (i.e., modified integral white-line intensity ratio) which is directly retrieved from EELS spectra at the Fe L2,3 edges and is independent of the coordination environment of Fe to a first-order approximation.

Here, we propose an empirical approach similar to that of van Aken and Liebscher (2002) to calibrate the correlation between Fe3+/ΣFe ratio and some parameters extracted from the STXM-derived XANES Fe L2,3-edge spectra of reference silicate glasses and phyllosilicates. Two empirical calibrations are proposed, both of which offer a compromise between speed and accuracy of the analytical measurement. An application of this approach to ultra-thin sections of natural chlorites and micas is presented to illustrate the methodology and to further assess the range of applicability of these calibrations for STXM.

Materials and methods

Reference samples

The samples used in this study were reference synthetic silicate glasses, natural phyllosilicates and fayalite, prepared as powders or ultra-thin sections cut by focused ion beam (FIB) milling. The bulk chemical compositions of the five synthetic glasses were previously determined by Magnien et al. (2004). All samples are composed of similar proportions of Si, Mg, Ca, Na and Fe. The SiO2 and FeO contents are ~52 and ~13 wt%, respectively. Bulk Fe3+/ΣFe ratios were determined by wet chemistry, Mössbauer spectroscopy and EMPA and range from 0.09 to 0.94 (Table 1; Magnien et al. 2004). For STXM-XANES analyses, we ground these samples in deaerated and deionised water, inside an anoxic glove box (p(O2) <50 ppm) to avoid oxidation during sample preparation.

Table 1 Reference samples used for XANES—Fe3+/ΣFe ratio quantification
Full size table

The phyllosilicate samples have 2:1 and 2:1:1 structures, and their bulk compositions were investigated previously by Joswig et al. (1986); Keeling et al. (2000); Shingaro et al. (2005); Rigault (2010) and in the present study by EMP analyses. Total Fe contents vary significantly between samples and bulk Fe3+/ΣFe ratios ranging between 0.03 and 1.0 were measured by Mössbauer spectroscopy, EXAFS (extended X-ray absorption fine structure) and/or EELS (Table 1). In addition, a fayalite sample was used as a pure Fe2+ reference. For STXM-XANES analyses, some samples (smectite Nau-2, chlorites GAB 42, VNI 92, VNI 114, fayalite) were prepared by grinding in deaerated and deionised water in an anoxic glove box (p(O2) <50 ppm). Other samples (clintonite, chlorite “prochlorite”, chlorite Ch1, Ti-mica) were prepared by FIB milling.

Samples transparent to soft X-rays are needed to measure XANES spectra in the transmission mode of STXM, therefore requiring the preparation of thin samples. FIB foils were cut with a FEI Model 200 TEM FIB system at University Aix-Marseille using the protocol detailed by Heaney et al. (2001). A 30 kV Ga+ beam operating at ~20 nA excavated the sample to a depth of 5 μm. The sample foil was then further thinned to ~80–100 nm at lower beam voltage (5 kV) and current (~ 100 pA), in order to remove the layer damaged by high-energy ions (Bourdelle et al. 2012).

Glarus field samples

The Glarus Alps (Switzerland) belongs to the Helvetic zone of the northern margin of the Central Alps and was affected by low-grade metamorphism. Details about the location and composition of the samples analysed by STXM in the present study are provided in Lahfid et al. (2010). The selected rock samples are metapelites, more or less clayey or sandy marls, with various proportions of quartz, calcite and clay minerals. Three samples (noted Glarus GL07 13, 16 and 20, as in Lahfid et al. 2010), containing chlorites and K-deficient micas, were milled by FIB. The compositions of the chlorites and micas were obtained on the FIB foils by analytical electron microscopy analyses described elsewhere (Bourdelle et al. 2012).

XANES spectroscopy

Part of the STXM analyses were performed at the advanced light source (ALS) (Lawrence Berkeley National Laboratory) on branch line 11.0.2.2 following the procedures described in Miot et al. (2009). The ALS storage ring was operated at 1.9 GeV and 500 mA current in a top-up mode. More details on the branch line 11.0.2.2 and beam characteristics are given by Bluhm et al. (2006 ). Stacks of images were obtained by scanning the sample in the x–y directions of selected sample areas over the 690–730 eV energy range (Fe L2,3-edge) using an energy increment of 0.789 eV between 690 and 705 eV, 0.10 eV in the 705–713 eV energy range, 0.19 eV in the 713–719 eV energy range, 0.155 eV in the 719–726 eV energy range and 0.475 eV in the 726–730 eV energy range. The dwell time per pixel and energy point was 1.3 ms.

Some data (chlorite Ch1, chlorite “prochlorite” and Ti-mica) were acquired on the Pollux beamline at the Swiss Light Source (SLS, Villigen, Switzerland). The SLS synchrotron storage ring was operated at 2.4 GeV and 300 mA current in a top-up mode during data collection, and the characteristics of the beamline are detailed by Raabe et al. (2008). Stacks were obtained over the 690–730 eV energy range (Fe L2,3-edge) using an energy increment of 0.667 eV between 690 and 700 eV, 0.15 eV in the 700–715 eV energy range, 0.40 eV in the 715–727 eV energy ranges and 0.89 eV in the 727–730 eV energy range. The dwell time per pixel and energy point was 3.5 ms.

At both the ALS and the SLS, focus was achieved systematically for each sample, and precision in the determination of the focus position was better than the focus depth. Image stacks were aligned, and XANES spectra were derived from areas of interest using the aXis2000 software (Hitchcock 2012). Potential beam damage caused by the incident photon beam was assessed by monitoring spectral changes at the Fe L2,3-edge with increasing dwell times up to a hundred milliseconds (10, 50 and 100 ms).

Spectra processing

Energy calibration was performed using the gaseous neon 1s → 3p electronic transition at 867.3 eV. As explained in Fig. 1, the processing of spectra consisted of two steps. First, the contribution of lower energy absorption edges (i.e., background) was removed so that in the end, the pre-edge region is set to 0 optical density (noted OD) with a slope of zero. For that purpose, a “linear background” correction was applied to the spectrum. Second, the two edge steps resulting from transitions to unoccupied states in the continuum were subtracted using the following double arctan function (Chen et al. 1995; van Aken and Liebscher 2002; Brotton et al. 2007):

$$ f\left( {\Updelta E} \right) = \frac{{h_{1} }}{\pi }\left( {\arctan \left[ {\frac{\pi }{{w_{1} }}\left( {\Updelta E - E_{1} } \right)} \right] + \frac{\pi }{2}} \right) + \frac{{h_{2} }}{\pi }\left( {\arctan \left[ {\frac{\pi }{{w_{2} }}\left( {\Updelta E - E_{2} } \right)} \right] + \frac{\pi }{2}} \right) $$
(1)

where h 1 and h 2 are the step heights of the two arctan functions, w 1 and w 2 are fixed peak widths and E 1 and E 2 are the positions of the inflection points resulting in an energy near the edge onset. Here, w 1 and w 2 are fixed to 1 eV (Fig. 1). Brotton et al. (2007) proposed setting the function slope w at 5 eV, to account for the slow onset of the continuum. They argued that a value smaller than 5 eV could induce spurious structures in the background-corrected spectrum. We observed that values of w = 1 eV or w = 5 eV provided similar results.

Fig. 1
figure 1

Subtraction of background from XANES spectra at Fe L-edge, using linear (a) and double arctan (b) functions (w 1  = w 2  = 1 eV), for chlorite GAB 42

Full size image

Results and discussion

Evolution of Fe L2,3-edge XANES spectra with changes in Fe3+/ΣFe

X-ray absorption near edge structure spectra at the Fe L2,3-edges of the reference phyllosilicates, fayalite and five Fe-bearing silicate glasses, corrected for continuum absorption, are shown in Fig. 2, and the positions of major peaks are summarised in Table 1. These spectra are qualitatively similar to those described in several previous studies and were obtained using different analytical techniques (e.g., Crocombette et al. 1995; Heijboer et al. 2003; van Aken and Liebscher 2002). Four major Fe L2,3-edge XANES peaks are present in all samples. The two major peaks on the L3 edge are noted as “L3-a” and “L3-b”, and similarly, the major peaks on the L2 edge are noted as “L2-a” and “L2-b”. For all samples examined, the measured separations of the Fe L3 and L2 maxima, due to spin–orbit splitting (van Aken and Liebscher 2002), are 12.9 ± 0.4 eV and 14.2 ± 1.4 eV for peaks a and b, respectively, in agreement with previous EELS and XANES studies (e.g., de Smit et al. 2008; de Groot et al. 2010). However, although most of the spectra show a single asymmetrical L3-a peak, some of them (i.e., VNI 92, VNI 114 and GAB 42, fayalite) display an “L3-a” split into two peaks. In addition, these specific spectra show additional peaks on the L3-a side at ~706.3 and ~706.8 eV. According to Wasinger et al. (2003), the presence of these minor peaks may be due to a specific atomic environment and/or orbital co-valency of iron in these mineral phases. Similarly, several minor peaks can be observed at around 719.8 eV on the L2-edge for several samples (VNI 92, VNI 114, GAB 42, PyrNa 17R, fayalite).

Fig. 2
figure 2

Representative XANES spectra at the Fe L2,3-edges for the reference silicates. The spectra have been normalised to the integral Fe L3-edge intensity, and some of the spectra have been shifted vertically for clarity (normalised intensity with arbitrary units). The dotted lines represent the energies fixed to determine the Fe3+ concentration from the Fe L3-peaks’ intensity ratio. The solid line underlines the position of L2-b maximum intensity, which is identical for all spectra. X = Fe3+/ΣFe ratios of Table 1

Full size image

The relative intensities of the different major peaks vary depending on the Fe3+/ΣFe ratio (Fig. 2). With increasing Fe3+/ΣFe ratios, the relative intensity of the L3-a peak decreases compared to that of the L3-b peak; L3-a is more intense than L3-b in the XANES spectrum of the VNI 92 sample (Fe3+/ΣFe = 0.35), whereas the opposite is observed for PyrNa 5R (Fe3+/ΣFe = 0.61). Likewise, the relative intensity of L2-a progressively decreases, whereas that of L2-b increases as Fe3+/ΣFe increases. The energy position of L2-b changes very weak between the samples, whereas peaks L3-a and L2-a shift slightly towards higher energies when Fe3+/ΣFe increases.

Quantification of Fe3+/ΣFe from XANES Fe L2,3-edge intensity ratios

As documented in Fig. 2, the main variations in the XANES spectra of reference samples with varying Fe3+/ΣFe ratios involve the L3-b/L3-a intensity ratio. More precisely, the L3-b/L3-a intensity ratio is linearly correlated with Fe3+/ΣFe ratio with only a little scatter (R² = 0.96) for both the phyllosilicates and silicate glasses (Fig. 3). The correlation is described by Eq. (2):

$$ \frac{{{\text{Fe}}^{3 + } }}{{\sum {\text{Fe}} }} = \frac{{R_{{L_{3} }} - 0.1867}}{0.01991}\; {\text{with}}\;R_{{L_{3} }} = \frac{{I\left( {L_{3} - b} \right)}}{{I\left( {L_{3} - a} \right)}} $$
(2)

This approach requires only five XANES images to map Fe3+/ΣFe (see Fig. 4 and below): two images in the pre-edge (needed to apply the “linear background correction” at each pixel of the image), one at 708.7 eV to quantify the L3-a peak, one at 710.25 eV to quantify the L3-b peak and one at 718 eV, to remove the edge step of the arctan function. Finally, the ratio of the resulting 708.7 and 710.25 eV images can be used to determine the R L3 parameter at each pixel of the image.

Fig. 3
figure 3

L3-edge intensity ratio I(L 3 -b)/I(L 3 -a) from XANES spectra versus ferric iron concentration Fe3+/ΣFe quantified by independent methods for the selected silicates. CI confidence interval (95 %)

Full size image
Fig. 4
figure 4

Determination of the Fe3+/ΣFe ratio from 5 selected energy images: two images in the pre-edge (to apply the “linear background correction” at each pixel of the image), one at 708.7 eV to quantify the L3-a peak, one at 710.25 eV to quantify the L3-b peak and one at 718 eV, to remove the edge step of the arctan function. Finally, the ratio of the resulting 708.7 and 710.25 eV images can be used to determine the R L3 parameter at each pixel of the image and obtain iron redox mapping. All images are OD images (70 × 90 pixels), where the illite and chlorite are the dark- and light-grey phases, respectively. As an illustration, spectrum #1 was retrieved from 110 images (i.e., 110 energy points) on a chlorite area (dark rectangle on image C); spectrum #2 was obtained after the linear function subtraction from spectrum #1 and spectrum #3, after the actan function subtraction from spectrum #2. Case of FIB foil of Glarus GL07 20 sample

Full size image

This calibration is useful but has some limitations. The L3 peaks, which are much more intense than the L2 peaks, are more susceptible to absorption saturation (see Hanhan et al. 2009, where saturation effects are described for Ca 2p edge spectra). This phenomenon occurs when the sample is too thick and/or highly concentrated in Fe implying that few photons are transmitted. This may trigger a nonlinear response of the detection and an artifactual modification of the relative peak heights. The use of a spectral parameter correlated with Fe3+/ΣFe based on the less absorbing L2-edge may provide in this case an interesting way of circumventing absorption saturation issues encountered with the L3-edge.

Figure 2 shows that an increase in Fe3+/ΣFe is associated with a decrease in the intensity of L2-a. Figure 5 shows the correlation between Fe3+/ΣFe and R L2 , a ratio that reflects the importance of L2-b relative to the total L2. Similar to the modified intensity defined by van Aken and Liebscher (2002), the L2-b contribution is computed as an integration window of 2 eV width centred around the maximum L2-b intensity; the ratio R L2 is calculated from this modified L2-b intensity and the total integral intensity of L2-edge. The correlation is high (R² = 0.97), and is described by Eq. (3):

$$ \frac{{{\text{Fe}}^{3 + } }}{{\sum {\text{Fe}} }} = \frac{{R_{{L_{2} }} - 0.1476}}{0.00297}\;{\text{with}}\;R_{{L_{2} }} = \frac{{I\left( {L_{2} - b} \right)_{\bmod } }}{{A\left( {L_{2} } \right)_{total} }} $$
(3)

This approach requires the acquisition of a complete stack of images (i.e., as many images as energy points are required to obtain a complete spectrum with a given spectral resolution) between 715 and 730 eV, to cover the entire L2-edge, and to calculate the double arctan function (Eq. (1)). As a consequence, the acquisition time required for this method is longer than for the L3-b/L3-a intensity ratio method (e.g., 30–40 min versus 5–10 min for an area of 150 by 150 pixels). However, this second method seems to be more accurate, especially because (1) the calibration data are less scattered (Fig. 5 versus Fig. 3) and (2) the intensity integration improves the signal-to-noise ratio. Several other methods of calibration have been tested, sometimes giving equation with a high correlation (with a R² up to 0.95), but the two methods proposed here seem to be a good trade-off between Fe3+/ΣFe estimation accuracy, acquisition time and ease of use.

Fig. 5
figure 5

L2-edge integral intensity ratio (i.e., integral intensity of maximum L2-b ± 0.1 eV over total integral intensity (area) of L2-edge) from XANES spectra versus ferric iron concentration Fe3+/ΣFe quantified by independent methods for the reference silicates. CI confidence interval (95 %)

Full size image

Assessment of saturation and beam damage effects

When particles are sufficiently thin, the intensity of each spectral feature changes linearly with thickness. However, Hanhan et al. (2009) showed that in the case of samples that are too thick, one can observe distortions of the Ca 2p spectrum due to a saturation effects. These observations led the authors of that study to set a maximum peak intensity, which should not be exceeded to avoid saturation phenomena.

Similarly, we determined the maximum peak intensity below which the Fe L23 spectra are undistorted and vary linearly. For this purpose, a powder of the smectite Nau-2 sample with grains of various sizes was analysed by STXM. Figure 6 plots the difference between the intensities at 710.35 (L3-b) and 723.54 eV (L2-b) (corrected from the pre-edge slope) versus the intensity at 710.35 eV (L3-b, i.e., the peak of maximum intensity for Nau-2, hence the most susceptible to saturation) for each pixel of the stack of images (i.e., a total of 6,336 pixels). The difference between L3-b and L2-b intensities increases linearly when L3-b intensity is lower than ~1.5 OD. Once the L3-b intensity exceeds 1.5 OD, the L3-b–L2-b difference increases more slowly than L3-b, underlining (1) the distortion of the spectra for the considered pixels and (2) the faster increase in L2-b intensity compared to that of L3-b with increasing sample thickness. All the data presented in this study were therefore collected from areas presenting a L3 peak intensity lower than 1.5 OD.

Fig. 6
figure 6

Difference, pixel by pixel, of intensity detected between the 710.25 and the 723.54 eV images (in which a pre-edge image was not subtracted) versus the intensity of the 710.25 eV image of a Smectite Nau-2 STXM-map (Nau-2, 72 × 88 pixels = 6336 points), that is, the L3-b – L2-b intensity difference versus the L3-b intensity for each pixel. The dashed line was calculated from a quadratic equation. Insets: representative spectra and optical density image (710.25 eV) for Nau-2 sample

Full size image

The spectrum may be also influenced by the crystal orientation relative to the direction of polarisation of the X-ray beam, a process called linear dichroism. Therefore, several XANES spectra were measured on the same part of a FIB foil after sequential rotation of the linear polarisation (see Benzerara et al. 2011b for details on the procedure). The variation of resulting Fe3+/ΣFe estimates is negligible, showing that sample orientation does not affect the Fe3+ quantification.

Beam damage can also potentially alter assessment of the Fe3+/ΣFe ratio. Here, beam damage was evaluated by monitoring spectral changes at the Fe L2,3-edge with increasing dwell times from 10 up to 100 ms. Figure 7 shows that Fe3+/ΣFe ratios derived from XANES spectra are only slightly affected by increasing dwell time. In particular, no significant change was observed for typical dwell times used during routine analyses of the samples (i.e., ~1.3 and 3.5 ms per energy- and image-point for ALS and SLS synchrotrons, respectively).

Fig. 7
figure 7

Beam-induced radiation damage during STXM analyses of chlorite “prochlorite” (XFe3+ = 30 %) and clintonite (XFe3+ = 69 %). Evolution of the Fe3+/ΣFe ratios as a function of dwell time, estimated by Eqs. (2) and (3) from XANES spectra. Data were fit by a quadratic function. The beam radiations (increasing dwell) involve (1) a decrease in XFe3+ calculated from L3-edge (Eq. 2) and (2) an increase in XFe3+ calculated from L2-edge (Eq. 3). Spectra of reference samples and Glarus samples (see text) were recorded with a dwell time of 1.3-3.5 ms per point and energy: the beam radiation damage is thus negligible with our analytical conditions for data collection

Full size image

Application to a geological case: chlorites and micas from Glarus (Central Alps, Switzerland)

To go further, we have applied the methods proposed here on micrometre-sized chlorite and mica/illite-like grains sampled in the Glarus area of Switzerland and cut by FIB milling. The temperatures of chlorite formation were calculated from analytical electron microscopy (AEM) chemical analyses, based on the thermometer by Bourdelle et al. (2013b), which does not require Fe3+/ΣFe input, and the thermometer by Inoue et al. (2009), which needs a previous estimation of Fe3+ content. The Fe3+/ΣFe ratios were estimated for each FIB foil by XANES from Eqs. (2) and (3). The results are given in Fig. 8 and Table 2.

Fig. 8
figure 8

Scanning transmission X-ray microscopy (STXM) and XANES analysis and Fe3+/ΣFe estimations for FIB foils of Glarus samples (chlorite and illite). [Left] Optical density images of FIB foils at 708.7 eV. The illite and chlorite are the dark- and light-grey phases, respectively. [Right] XANES spectra of areas of interest and calculated Fe3+ concentrations associated (crystals rims)

Full size image
Table 2 AEM chemical representative analyses of Glarus chlorites and comparison of thermometers results taking into account Fe3+/ΣFe ratios (Eq. (3)); regarding the scatter of data on Fig. 3, we infer a precision of ±5 % on the Fe3+/ΣFe
Full size table

From images converted to optical density units taken at 708.7 eV, we can easily distinguish Fe-rich and Fe-poor minerals: chlorites appear as light grey and represent the Fe-rich phase, whereas micas are dark, that is, Fe-poor. XANES spectra, acquired along the mica-chlorite contacts show that the Fe3+/ΣFe ratio is higher in illite than in chlorite: the Fe3+/ΣFe ratios estimated by Eq. (3) range from 22.3 to 27.9 % in chlorite, whereas these ratios vary between 30 and 65.5 % in illite-like phase. Equation (2) provides consistent estimations, suggesting that both calibrations are reliable. This analysis shows that K-deficient micas can contain a high proportion of ferric iron (e.g., samples 13 and 20). Despite the relatively high Fe3+/ΣFe ratio in some illite-like crystals, the total Fe3+ content remains higher in chlorite.

Figure 8 also shows the variations of Fe3+/ΣFe ratios versus the temperature of formation, which was estimated by chlorite thermometry (Table 2). In this respect, Fe3+/ΣFe ratio increases slightly in chlorites with increasing temperature, whereas this ratio decreases in K-deficient micas. It should be noted that, contrarily to the Bourdelle et al. (2013b) model, some geothermometers based on thermodynamic models for chlorite (e.g., Inoue et al. 2009) require prior determination of the Fe3+/ΣFe ratio. When this value is not known, it is set to zero as the default in these types of models. Interestingly, the comparison of results provided by different thermometers in Table 2 shows that the Inoue and Bourdelle geothermometers yield very different temperature results (differences of up to 76 °C) when Fe3+/ΣFe ratio is not known. In contrast, taking into account the Fe3+/ΣFe, the two thermometers provide more similar temperatures estimates (a maximum difference of less than 28 °C, i.e., within the uncertainty of the thermometers), showing the cross-check validity of the Fe3+/ΣFe estimation. A variation of the Fe3+/ΣFe ratio from 0 to ~23 % in chlorites implies a decrease in the temperatures calculated by the Inoue model of 20, 40 and 46 °C depending on the sample.

Figure 9 displays an example of Fe3+/ΣFe mapping at the nanometre-scale derived from images at 706, 708.7, 710.25 and 718 eV using Eq. (2) (see Fig. 4). The analysis was carried out on the Glarus GL07 20 FIB foil. The scanned area measures 3.3 × 3.5 micrometres with a pixel size of 88 × 88 nm. The analysis of the illite-chlorite contacts by AEM showed that they are approximately perpendicular to the FIB foil surface, that is, there is only a little overlap between the two minerals at their contact. The spatial averaging effect of the X-ray beam over the pixel size (i.e., 88 nm) sets the limit of the minimum distance over which illite-chlorite contacts can be discriminated. Beyond this distance, the intracrystalline variation of the Fe3+/ΣFe ratio in the illite-like phase can be interpreted as an authentic zonation, from ~55 % in crystal rims (conforming to the spectra presented in Fig. 8) to ~85 % in several crystal core clusters. In the same way, the Fe3+/ΣFe ratio distribution draws a subtle zonation in the chlorite, with a Fe3+/ΣFe ratio ranging from 18 to ~23 % on the crystal rim, in accordance with the spectra shown in Fig. 8. Such variations of the Fe3+/ΣFe ratio within the crystals are equivalent to several degrees or tens of degrees in the temperature estimation, especially when this variation is associated with a variation in composition. One can expect that this zonation is a crucial issue in application of geothermometers (de Andrade et al. 2006; Bourdelle et al. 2013a), and the redox gap between illite and chlorite raises the issue of the crystallisation processes.

Fig. 9
figure 9

Quantitative Fe redox nanomapping on FIB foil of Glarus GL07 20 sample. a Optical density image at 708.7 eV of Glarus GL07 20 FIB section, where the illite and chlorite are the dark- and light-grey phases, respectively. b Optical density image at 708.7 eV of the area of interest. c iron redox mapping, calculated from the 708.7 to 710.25 eV images ratio coupled with Eq. (2). The illite-chlorite contacts were analysed by AEM to check that they are approximately perpendicular to the FIB foil surface, that is, there is only a small overlap between the two minerals at their contact. The spatial averaging effect of the X-ray beam over the pixel size (i.e., 88 nm) sets the limit of the minimum distance over which illite-chlorite contacts can be discriminated. Beyond this distance, the intracrystalline variation of Fe3+/ΣFe ratio in the illite-like phase can be interpreted as an authentic zonation, from 55 to 85 % in several crystal core clusters

Full size image

In summary, the STXM-based XANES study of FIB foils from the Glarus, Switzerland samples enables (1) estimation of the Fe3+/ΣFe ratio in each phase and (2) establishment of iron redox mapping with high spatial resolution.

Conclusion

In this study, we have demonstrated the reliability of two methods that allow quantitative determination of Fe3+/ΣFe ratios in silicate phases using STXM coupled with XANES spectroscopy at the Fe L2,3-edges. These approaches are similar to those proposed by van Aken and Liebscher (2002) for EELS measurements but are here calibrated for STXM. The two calibrations are based on reference samples with variable but known Fe3+/ΣFe ratios, which were prepared as powders or as FIB foils. We tested these calibrations on three FIB foils extracted from field samples of phyllosilicates (Glarus, Switzerland chlorite and illite samples from metapelites), demonstrating the potential of these methods for quantifying Fe3+/ΣFe ratios at the submicrometre scale. This approach will allow more quantitative mineralogical or geomicrobiological studies requiring estimation of the iron redox state at the nanoscale for terrestrial or extraterrestrial Fe-rich samples.