Contrib Mineral Petrol (2011) 162:1249–1257 DOI 10.1007/s00410-011-0649-9 ORIGINAL PAPER Microanalysis of the iron oxidation state in (Mg,Fe)O and application to the study of microscale processes Micaela Longo • Catherine A. McCammon Steven D. Jacobsen • Received: 25 February 2010 / Accepted: 6 May 2011 / Published online: 20 May 2011 Ó Springer-Verlag 2011 Abstract We report application of the flank method using the electron microprobe to a suite of twelve (Mg,Fe)O samples with composition 2–47 wt% Fe and Fe3?/RFe = 1 to 11%, where Fe3?/RFe was determined independently using Mössbauer spectroscopy on the same grains used for the flank method measurements. A calibration curve of the form Fe2? = A ? B 9 (RFe)2 ? C 9 (Lb/La) was fit to the data and gave excellent agreement between Fe3?/RFe calculated from the flank method and Fe3?/RFe determined using Mössbauer spectroscopy. We found the method to be sufficiently sensitive to determine meaningful variations in Fe3?/RFe for geophysically relevant compositions of (Mg,Fe)O (\25 wt% Fe), and calibration parameters remained constant within experimental uncertainty over the course of the entire study (20 months). Flank method measurements on an inhomogeneous sample of synthetic (Mg,Fe)O showed evidence of diffusion processes resulting from rupture of the capsule during the high-pressure experiment and the possibility to measure Lb/La variations with a spatial resolution of a few microns. We detected the presence of exsolved magnesioferrite in a suite of Communicated by M. W. Schmidt. M. Longo (&) C. A. McCammon Bayerisches Geoinstitut, University of Bayreuth, 95440 Bayreuth, Germany e-mail: [email protected] S. D. Jacobsen Department of Earth and Planetary Sciences, Northwestern University, Evanston, IL 60208, USA Present Address: M. Longo Department of Geosciences, University of Padua, 35122 Padua, Italy (Mg,Fe)O single crystals using transmission electron microscopy and Mössbauer spectroscopy. Flank method measurements on the same suite of single crystals showed enhanced Fe3?/RFe values, consistent with the presence of magnesioferrite even though the grains were too small to be resolved by conventional electron microprobe measurements. Keywords Iron oxidation state Flank method Electron microprobe Ferropericlase Introduction Ferropericlase, (Mg,Fe)O, is probably the most abundant non-silicate oxide within the Earth. Current models suggest that it makes up roughly 20% of the lower mantle, which comprises more than half of the Earth by volume. (Mg,Fe)O can incorporate a significant amount of Fe3?, where its concentration and incorporation mechanism (e.g. point defects) strongly influence the rheological and transport properties. The strong contrast between these properties of (Mg,Fe)O and those of (Mg,Fe)(Si,Al)O3 perovskite, believed to be the Earth’s most abundant phase, suggests that the lower viscosity and higher atomic diffusivity and electrical conductivity of (Mg,Fe)O may dominate those properties of the lower mantle (e.g. Dobson et al. 1997; Yamazaki and Karato 2001; van Orman et al. 2003). (Mg,Fe)O is stable over an extremely wide pressure and temperature range, and it is the only lower mantle phase that is also thermodynamically stable at ambient conditions. Additionally, it is the only lower mantle phase for which natural samples have been recovered from the Earth’s interior, namely as inclusions in diamonds of deep mantle origin (e.g. McCammon 2001). The behaviour of 123 1250 Fe3? in (Mg,Fe)O and its relation to mantle properties can therefore be studied either by experimental synthesis or through examination of natural samples. Studies of Fe3? in (Mg,Fe)O have been restricted up until now with regard to sample size due to the limited number of methods to analyse Fe3? on the microscale. For example, Mössbauer spectroscopy measurements have been performed on ferropericlase inclusions from lower mantle diamonds using a high specific activity source (‘point source’) (McCammon et al. 1997, 2004), but the smallest inclusion that could be studied was 100 lm due to physical limitations of the 57Co point source (McCammon 1994). Since the majority of inclusions in lower mantle diamonds are less than 50 lm in diameter, the population cannot be considered yet to have been adequately sampled. The study of microscale processes in synthetic (Mg,Fe)O has also been restricted with respect to Fe3? determination. For example, diffusion of transition elements in MgO was studied nearly 50 years ago (e.g. Wuensch and Vasilos 1962), but so far, it has not been possible to independently measure diffusion coefficients of Fe2? and Fe3?. One promising method to determine Fe3?/RFe on the microscale is electron energy loss spectroscopy (EELS) on a transmission electron microscope, which offers nanometre-scale spatial resolution (e.g. van Aken et al. 1998). Although sample preparation involves a degree of sample destruction, the advent of focussed ion beam thinning may offer improved possibilities to study rare inclusions from the lower mantle as well as optimise selection of the sample region to be studied, such as perpendicular to diffusion profiles. Another promising method for Fe3? studies is X-ray Absorption Near Edge Structure (XANES) spectroscopy, which has successfully been applied at a spatial resolution of roughly 5 lm (e.g. Berry et al. 2008). However, neither of these methods has yet been applied to determine Fe3?/RFe in (Mg,Fe)O, and more crucially, neither method can be considered to be a routine application accessible to a large proportion of the geoscience community to the extent of, for example, the electron microprobe. In the past decades, many studies have been devoted to Fe3?/RFe determination from the FeLb and FeLa emission spectra using the electron microprobe. One of the most successful is the so-called flank method, which has shown an unambiguous correlation of measured intensity ratios of Fe Lb/La X-ray emission with the oxidation state in garnet, independent of composition (Höfer and Brey 2007 and references therein). Höfer and Brey (2007) demonstrated that the sensitivity of Lb/La to Fe3?/RFe is a function of both the total iron content and the coordination number of the iron atoms, where the large coordination number of iron in garnet gives the highest sensitivity. Earlier work applying the flank method to FexO demonstrated that Fe3?/ 123 Contrib Mineral Petrol (2011) 162:1249–1257 RFe could be measured, even though the Lb/La variation was less than for garnet (Höfer et al. 2000). For (Mg,Fe)O, particularly samples with iron compositions relevant for the lower mantle (\25 wt% Fe), it is not obvious whether the sensitivity of the flank method is sufficient to determine meaningful variations in Fe3?/RFe, particularly since values in lower mantle compositions may range only between 0 and 10%. The goal of this study, therefore, is to investigate the potential of the flank method to measure Fe3?/ RFe in geophysically relevant compositions of (Mg,Fe)O and to demonstrate its application to the study of microscale processes. Experimental methods Sample synthesis We synthesised a suite of (Mg,Fe)O samples covering a wide range of composition (2–47 wt% Fe) and Fe3?/RFe (1–11%). To synthesise the samples, Mg and Fe metals were mixed in stoichiometric proportions to give a wide compositional range of Fe in (Mg,Fe)O. The mixture was enriched in 57Fe to approximately 10% of the Fe total to optimise Mössbauer measurements. Metals were first dissolved in HNO3, and then, the mixture was heated to 50°C to allow HNO3 to slowly evaporate. Subsequently, NH4OH was added to obtain oxide precipitates. Excess HNO3 and NH4OH were then removed by drying the obtained gels using a Bunsen burner (1,200–1,500°C). All synthetic powders were then equilibrated in a gas-mixing furnace under CO/CO2 at 1,300°C and controlled oxygen fugacity in order to obtain a wide range of Fe3?/RFe. Oxygen fugacities were varied over the range from 10-7 to 10-11 and were monitored during the experiments using an oxygen fugacity sensor. X-ray powder diffraction was used to verify the structure of the polycrystalline powders. In order to obtain the high-quality surfaces needed for electron microprobe measurements, the grain size of the polycrystalline samples was coarsened through annealing at high pressure and temperature using a multianvil apparatus. The (Mg,Fe)O polycrystalline powders were loaded into Re capsules of 1.6 mm diameter and 2 mm long. Samples were compressed up to 15 GPa and heated to 1,800–2,000°C for about 1 h. The resulting samples were analysed using Mössbauer spectroscopy to determine Fe3?/ RFe and then were mounted on epoxy or glass for the electron microprobe measurements. For Mössbauer measurements, the multianvil run products were cut into sections 100 lm thick and centred behind a hole of 150 lm diameter in 25 lm thick Ta foil. Mössbauer spectra were recorded at room temperature (293 K) in transmission mode on a constant acceleration Contrib Mineral Petrol (2011) 162:1249–1257 spectrometer with a nominal 370 MBq 57Co high specific activity source in a 12-lm Rh matrix. The velocity scale was calibrated relative to 25 lm thick a-Fe foil using the positions certified for (former) National Bureau of Standards standard reference material no. 1541; line widths of 0.36 mm/s for the outer lines of a-Fe were obtained at room temperature. On average, spectra took one day each to collect. The spectra were analysed using the commercially available fitting program NORMOS written by R.A. Brand (distributed by Wissenschaftliche Elektronik GmbH, Germany). We fitted the Mössbauer spectra of (Mg,Fe)O to two Fe2? doublets and one Fe3? doublet, all with Voigt lineshape according to previous models (e.g. McCammon et al. 1998). Fe3?/RFe was calculated from the relative areas of the Fe3? absorption corrected for differences in recoil-free fraction of Fe2? and Fe3?. Waychunas et al. (1994) found that Mössbauer determinations of Fe3?/RFe in (Mg,Fe)O were systematically higher than for thermogravimetric results, which is consistent with a higher recoil-free fraction for Fe3? in the tetrahedral site compared to Fe2? in the octahedral site. We corrected the relative areas as described by McCammon (2004) using Mössbauer Debye temperatures of 390 and 550 K for Fe2? and Fe3?, respectively. Electron microprobe analysis We determined the FeLa and FeLb positions on the Jeol XA-8200 microprobe at Bayerisches Geoinstitut following the procedure established by Höfer and Brey (2007). For each electron microprobe, it is necessary to collect the FeLa and FeLb emission spectra of standards in order to establish the exact La and Lb positions needed for the measurements. As references for Fe2? and Fe3?, we collected spectra on almandine and andradite standards, respectively, at 15 kV and 100 nA (Table 1). Spectra were collected using a TAP spectrometer with detector slits set to the smallest aperture possible (which corresponds to 300 lm for a TAP spectrometer type) in order to reduce as much as possible the background contribution during flank method measurements. The FeLa and FeLb positions were identified from the difference spectrum as the maximum and minimum points, respectively (Höfer and Brey 2007). For comparison, FeO and Fe2O3 spectra were also collected to determine the reproducibility of FeLb and FeLa energies using Fe2?- and Fe3?-bearing phases other than almandine and andradite. We found that La and Lb positions determined using the two different Fe2?- and Fe3?-bearing phase couples were consistent within the estimated uncertainty (Table 1). We corrected the daily shift of the FeLa and FeLb peak positions using the software PEAK FIT (Höfer and Brey 2007) at 25 kV and 80 nA. 1251 Table 1 Flank method positions La and Lb for the Jeol JXA-8200 microprobe at Bayerisches Geoinstitut Energy (eV) mma Almandine-andradite FeO–Fe2O3 FeLa FeLa 0.710 190.74 FeLb 0.700 188.09 0.711 190.82 FeLb 0.700 187.98 a Distance between the position of the Fe La and Fe Lb lines on the spectrometer crystal and the beam spot in the Rowland circle geometry The flank method parameter, Lb/La, is determined from the intensity ratio of the count rates recorded at each of the positions described above. In order to perform flank method measurements at the same time as major element analysis, we added two fictitious elements to the major element list with energies set to the FeLa and FeLb positions determined above in order to measure the intensities at those positions. Flank method measurements were always performed on the TAP spectrometer. We performed major element analysis and qualitative analysis simultaneously at an acceleration voltage of 15 kV and a current of 80 nA according to the procedure described by Höfer and Brey (2007). We measured the X-ray emission counts for the two positions La and Lb for 180 s each, for a total time of 360 s for each measured spot. Whenever possible, we performed measurements along line profiles with a minimum of 20–30 spots for each sample during each microprobe session to obtain reliable statistics. The Lb/La ratio was then used to determine the calibration for Fe3?/RFe based on the Mössbauer measurements. To test the flank method on the Jeol XA-8200 electron microprobe at Bayerisches Geoinstitut, we performed simultaneous major element analysis plus flank method measurements on a suite of natural garnets that had also been measured using Mössbauer spectroscopy (Longo 2009). We confirmed the observations of Höfer and Brey (2007) that Lb/La increased linearly as function of Fe2? (wt%) and that flank method Fe3?/RFe values were in excellent agreement with those from Mössbauer spectroscopy (Longo 2009). However, we did find that calibration parameters were different (their Eq. 3), which is likely due to slight differences in analyser crystal geometry between the two electron microprobes. All synthetic (Mg,Fe)O samples previously analysed by Mössbauer spectroscopy were mounted on epoxy or glass for the electron microprobe measurements. For major element analysis on synthetic (Mg,Fe)O, only Fe and Mg were calibrated. Fe metal and MgO were used as standards for Fe and Mg, respectively. Counting time for both was 20 s (peak) and 10 s (background), and the point beam was focused on the smallest diameter possible (1 lm). The 123 1252 Contrib Mineral Petrol (2011) 162:1249–1257 correction used for self-absorption and matrix effect was the phi-rho-z type, and the oxygen concentration was calculated indirectly from the FeO and MgO oxides. Before starting the calibration measurements, the sample homogeneity was tested for all synthesised samples, and outer regions of each sample that were potentially affected by diffusion processes during synthesis using the multianvil apparatus were excluded from the analysis. One of the synthetic (Mg,Fe)O samples was chosen as a standard that was measured before and after the measurements of unknown samples in order to monitor possible spectrometer drifts which could cause a shift in Lb/La ratios and therefore affect the accuracy in the determination of Fe3?/RFe. Results and discussion Calibration of the flank method on synthetic (Mg,Fe)O We measured the variation of Lb/La as a function of RFe for the twelve synthetic (Mg,Fe)O ferropericlase samples that showed the highest homogeneity and followed a similar approach used for garnets by Höfer and Brey (2007) using least-squares regression to fit the three variables RFe, Lb/La and Fe2?. Through many fitting trials, we found that the best and simplest fit for Fe2? in (Mg,Fe)O with RFe between 2 and 47 wt% is given by: Fe2þ ¼ A þ B ðRFeÞ2 þC ðLb=LaÞ A comparison of Fe3?/RFe determined using the flank method with Fe3?/RFe calculated using Mössbauer spectroscopy shows good agreement within experimental error (Fig. 2). Error bars for the flank method were calculated by ð1Þ In all other formulations with more parameters, the system was overdetermined, producing highly correlated parameters with consequently large uncertainties. Our calibration gave the following parameters: A = -18.16(51), B = 0.0078(2) and C = 34.61(66) where both Fe2? and RFe are in wt%. RFe was calculated from electron microprobe data assuming no cation vacancies, since inclusion of their effect would require a priori knowledge of Fe3? concentrations. However, we estimated the influence of non-stoichiometry on the calculations and found it to be smaller than experimental error for compositions with RFe \50 wt% and Fe3?/RFe \0.2. The experimental data are plotted in Fig. 1 with the calibration curves calculated for constant values of Fe3?/ RFe, which shows the variation of Lb/La with both RFe and Fe3?/RFe. Similar to the results for garnet reported by Höfer and Brey (2007), the sensitivity of the flank method increases with increasing Fe concentration. From Eq. 1, the detection limit for Fe3?/RFe can be estimated based on the iron concentration and the precision of Lb/La determination. For example, a precision of 0.01 enables a minimum difference in Fe3?/RFe of roughly 0.05 to be detected for ferropericlase with a composition of Mg0.95Fe0.05O. 123 P Fig. 1 Lb/La variation as a function of Fe (wt%) for twelve synthetic (Mg,Fe)O samples plotted with isopleths of Fe3?/RFe (values indicated on each line) determined using (1) Fig. 2 Comparison of Fe3?/RFe determined using the flank method according to (1) with Fe3?/RFe determined using Mössbauer spectroscopy. Error bars are indicated for all points (solid circles) except for those with RFe \3 wt% (open circles), where the error bars were omitted due to their large size (± 0.015 for x values and ± 0.20 for y values). The dotted line indicates 1:1 correspondence Contrib Mineral Petrol (2011) 162:1249–1257 1253 Table 2 Flank method results for synthetic (Mg,Fe)O samples used for calibration Run RFe (wt%) Lb/La Fe3?/RFe Mössbauer Fe3?/RFe Flank method 0.006 (205) S4068 2.44 (15) 0.594 (20) 0.027 (15) S4099 2.46 (23) 0.594 (16) 0.015 (15) 0.001 (195) S4117 12.57 (20) 0.846 (11) 0.024 (15) 0.016 (37) S4139 13.23 (20) 0.816 (11) 0.107 (25) 0.136 (35) S4044 22.79 (6) 0.999 (15) 0.109 (20) 0.102 (22) S4123 22.98 (20) 1.005 (15) 0.099 (25) 0.097 (23) S3939 29.58 (75) 1.104 (18) 0.098 (30) 0.092 (32) S3941 33.03 (47) 1.170 (15) 0.070 (30) 0.066 (22) S3996 33.15 (32) 1.143 (11) 0.084 (30) 0.096 (18) S4028 33.44 (7) 1.189 (17) 0.070 (25) 0.064 (16) S4153 S4155 39.72 (26) 46.81 (26) 1.232 (27) 1.285 (32) 0.075 (20) 0.069 (20) 0.074 (19) 0.073 (17) Fe3?/RFe determined by Mössbauer spectroscopy and by the flank method using (1) propagating the uncertainties in the equation parameters, as well as in RFe and Lb/La obtained from the mean value of 20–30 independent measurements at 15 kV and 80 nA. Results obtained for flank method measurements plus Fe bulk compositions, as well as a comparison between Fe3?/ RFe ratios determined by flank method and Mössbauer spectroscopy data, are summarised in Table 2. We found that subsequent measurements of the ferropericlase standards gave the same Lb/La values as previous measurements within experimental error and that the calibration (Eq. 1) remained constant with time over the entire period of the study (20 months). Sample homogeneity and the potential to investigate diffusion processes Nearly all of the synthetic (Mg,Fe)O samples appeared to be homogeneous as shown by the random variation of Lb/ La versus Fe (wt%) within the experimental error of the measurements. However, one of the multianvil run products (S4149) showed a large variation in flank method measurements taken over different parts of the sample. We therefore omitted that particular sample from the calibration fitting procedure, but performed further analysis to explore the potential of the flank method to quantify inhomogeneity. The sample prepared from multianvil run S4149 showed a large incursion of metal into the ferropericlase as well as uneven electron reflectivity of the ferropericlase itself (Fig. 3). Energy dispersive X-ray (EDX) analysis showed the metal to contain Re, which likely came from the sample capsule, and also La and Cr, which likely came from the LaCrO3 heater, strongly suggesting that the capsule broke during the experiment. EDX analysis showed that ferropericlase also contains 2–4 wt% Cr and that the Fig. 3 Electron backscattered image of synthetic ferropericlase sample S4149 suggesting a fracture of the Re capsule (right), leading to incursion of Re as well as La and Cr (from the LaCrO3 furnace) into the sample concentrations of Cr and Fe appear to be inversely correlated. We performed flank method measurement profiles on two of the ferropericlase grains in the inner part of the sample, as well as one on ferropericlase near the outer edge. Results show variations of Lb/La versus RFe that are displaced significantly below the calibration line for zero Fe3?/RFe, indicating that a reasonable proportion of Fe3? is present (Fig. 4a). The actual Fe3?/RFe values determined from the flank method calibration show a consistent trend with total Fe concentration, as well as a measureable difference between the inner and outer parts of the sample (Fig. 4b). Possible explanations for the difference in Fe3?/ RFe values across the sample include a gradient in oxygen fugacity, interactions between Fe2?/Fe3? and Cr2?/Cr3? 123 1254 Contrib Mineral Petrol (2011) 162:1249–1257 (a) (e.g. Eeckhout et al. 2007), or a matrix effect due to Cr that changes the Fe3?/RFe calibration. While more detailed interpretation of the processes that occurred during run S4149 is not justified due to the uncontrolled nature of the experiment, the results do illustrate the ability of the flank method to resolve variations in Lb/La across a sample and demonstrate how the method could be applied to study dynamic processes such as diffusion. Synthetic (Mg,Fe)O containing exsolved magnesioferrite Fig. 4 Flank method measurements for sample S4149: a Lb/La variation as a function of RFe wt% for measurement profiles from a region in the inner part of the sample (open circles), from a second region in the inner part of the sample (solid circles), and from the outer part of the sample (grey circles). The lines show the isopleths of the calibration equation for the indicated values of Fe3?/RFe; b Fe3?/ RFe variation calculated using the flank method for the same regions of sample S4149 We measured a set of synthetic (Mg,Fe)O single crystals consisting of nine samples with an iron compositional range between 8 and 67 wt% that had been synthesised by interdiffusion of Fe and Mg between single-crystal MgO and (Mg,Fe)O prereacted powders (Reichmann et al. 2000; Jacobsen et al. 2002). Previous Mössbauer measurements had revealed the presence of minor amounts of magnesioferrite (Reichmann et al. 2000; Jacobsen et al. 2002), so to further investigate the microstructure of the crystals, we performed TEM analysis. Results confirmed the presence of sub-micron grains of magnesioferrite and showed that its distribution in the crystal is not homogeneous (Fig. 5). One possible explanation for the presence of magnesioferrite in the (Mg,Fe)O single crystals are the oxidising conditions during the annealing experiments (near the NiNiO buffer). Jacobsen et al. (2002) reported two series of synthesis experiments: the first involved annealing at 1,450°C and fO2 = 10-7, while the second involved annealing at more reducing conditions, namely 1,300°C and fO2 = 10-10 bar. Mössbauer spectroscopy detected magnesioferrite in the samples from the first series of Fig. 5 TEM dark field microphotographs showing a comparison between (Mg,Fe)O samples with similar iron concentrations for a single crystal from Jacobsen et al. (2002) (Fe15) (left) and a sample from the present study (S3855 with 18.7 wt% Fe and 5% Fe3?/RFe) (right). The left image clearly shows cubic-shaped magnesioferrite impurities smaller than 1 lm in the (Mg,Fe)O matrix, whereas the right image shows no evidence of impurities. Laminar structures visible in both images are caused by stacking faults in the (Mg,Fe)O structure. The insets illustrate electron diffraction images taken from the respective samples. The left pattern shows a combination of ferropericlase and magnesioferrite, while the right pattern shows only ferropericlase (b) 123 Contrib Mineral Petrol (2011) 162:1249–1257 annealing experiments, but not in the samples from the second series. Although both series of oxygen fugacity conditions are within the (Mg,Fe)O stability field, the first series conditions are quite close to the phase boundary with magnesioferrite, and cooling could cause the phase equilibrium to shift into the magnesioferrite stability field, causing exsolution of a secondary magnesioferrite phase from the primary (Mg,Fe)O crystals. Such exsolution is already well known in the Fe–O system, where magnetite can exsolve from single-phase FexO on cooling (e.g. Hazen and Jeanloz 1984). We measured the Fe3? concentration in the residual ferropericlase as well as the abundance of magnesioferrite using Mössbauer spectroscopy (Fig. 6). The presence of magnesioferrite can be detected even at low abundance due to its occurrence as two magnetic sextets whose intense peaks do not overlap those of ferropericlase (De Grave et al. 1979), and its abundance could be estimated from its relative area corrected for the compositions based on the electron microprobe data and assuming that the Fe2?/Mg ratio in the residual ferropericlase and the exsolved magnesioferrite are approximately the same (Speidel 1967). We collected Mössbauer spectra at two velocity ranges, one small (–5 to ?5 mm/ s) in order to resolve the Fe3? component in ferropericlase (Fig. 6), and one large (-9 to ?9 mm/s) in order to observe all of the magnesioferrite peaks. The results show that the crystals from series 1 contain significantly more magnesioferrite than those from series 2, where the abundance of magnesioferrite in the latter samples was below the detection limit (Table 3). A sample that had been annealed at even more oxidising conditions (T = 1,300°C and fO2 = 10-7.5 bar) showed the highest magnesioferrite concentration (sample Fe78ox, Table 3). We calculated the bulk Fe3?/RFe ratio for each sample based on the individual Fe3? contents of each phase, their iron concentrations, and the phase abundance in the assemblage. Due to the small size of the exsolved magnesioferrite, the phase will contribute to both the major element analysis and the flank method measurements; thus, both RFe and the Lb/La ratios will be affected. Figure 7 illustrates the flank method results obtained for the nine (Mg,Fe)O samples, five from the first series (open circles), three from the second series (grey triangles) and one from a third experiment that was the most oxidizing of all the annealing series (1,300°C and fO2 = 10-7.5; solid circles). Results show that the samples from the most reducing conditions (grey triangles) are closest to the Fe3?/RFe = 0 calibration line as expected. Fe3?/RFe values for sample Fe24 fall between 0 and 0.1, which is roughly consistent with the results from Mössbauer spectroscopy, particularly considering the difference in spatial resolution (discussed in more 1255 (a) (b) (c) Fig. 6 Room temperature Mössbauer spectra of (Mg,Fe)O: a pure (Mg,Fe)O from the present study (S4074 with 18.8 wt%Fe and 8% Fe3?/RFe); b sample Fe15 containing 0.3 mol% magnesioferrite (seen as a small deviation in the residual in the regions marked by arrows); c sample Fe37 containing 4 mol% magnesioferrite (fitted with sextets corresponding to magnesioferrite and the most intense lines in the displayed velocity range are marked by arrows). The Fe3? in ferropericlase components are shaded grey, and the residuals are shown above each spectrum detail below). The other samples in series 2, Fe53 and Fe78, fall outside the calibration range, and it is evident that the curvature of the calibration curve affects the accuracy of the Fe3?/RFe determination outside of the range (Fig. 7). 123 1256 Contrib Mineral Petrol (2011) 162:1249–1257 Table 3 Quantification of Fe3? content and magnesioferrite fraction for synthetic (Mg,Fe)O single crystals containing (Mg,Fe)Fe2O4 using Mössbauer spectroscopy Sample Seriesa Fe3?/RFe (Mg,Fe)O Mole fraction (Mg,Fe)Fe2O4 Fe3?/RFe bulk Fe6 1 0.016 (10) 0.001 (1) 0.04 (2) Fe15 1 0.009 (10) 0.003 (2) 0.04 (2) Fe24 2 0.006 (10) 0.000 (1) 0.01 (1) Fe27 1 0.052 (30) 0.004 (2) 0.08 (4) Fe37 1 0.060 (30) 0.038 (19) 0.20 (7) Fe53 2 0.068 (30) 0.000 (1) 0.07 (3) Fe75 Fe78 1 2 0.123 (50) 0.072 (30) 0.047 (24) 0.000 (1) 0.21 (7) 0.07 (3) Fe78ox 3 0.090 (30) a Conditions for annealing—series 1: T = 1,450°C and fO2 = 10 fO2 = 10-7.5 bar 0.28 (14) -7 bar; series 2: T = 1,300°C and fO2 = 10 Fig. 7 Lb/La variation as a function of RFe for nine previously studied (Mg,Fe)O single crystals (open and solid circles and grey triangles) (Reichmann et al. 2000; Jacobsen et al. 2002) and a single crystal of magnetite (grey stars). The samples indicated by open circles (Fe6, Fe15, Fe27, Fe37, Fe75) were annealed at 1,450°C and fO2 = 10-7 bar, the sample indicated by the solid circles (Fe78ox) was annealed at 1,300°C and fO2 = 10-7.5 bar, and the samples indicated by grey triangles (Fe24, Fe53, Fe78) were annealed at 1,300°C and fO2 = 10-10 bar. Thirty to sixty measurements made on each crystal are mostly clustered at discrete values of RFe, but show significant data scatter with Lb/La. The lines show the isopleths of the calibration equation for the indicated values of Fe3?/RFe; solid lines indicate values within the calibration range, while dashed lines indicate values outside the calibration limits The flank method results from the crystals of series 1 show varying agreement with the estimated bulk Fe3?/ RFe values determined using Mössbauer spectroscopy. For samples Fe6, Fe15 and Fe27, all of which showed the presence of magnesioferrite in Mössbauer spectra, Fe3?/RFe is systematically overestimated by the flank method (Fig. 7) compared to the Mössbauer estimates, which fall in the range 0.04–0.08 (Table 3). In contrast, for sample Fe37, Mössbauer data indicate that the bulk 123 0.44 (6) -10 bar; series 3: T = 1,300°C and Fe3?/RFe ratio should be *0.2, while the flank method shows values less than 0.1 (Fig. 7). The difference in behaviour may lie in the different spatial resolution of the two methods, combined with the effect on Lb/La ratios of different crystal structures (rocksalt versus spinel). Mössbauer spectroscopy as applied in this study is a transmission method that probes the entire sample and gives information about the average environment in each of the phases. In contrast, the electron microprobe (and hence the flank method) probes only the upper few microns of the sample with a spatial resolution also on the order of a few microns. Our TEM study showed that the distribution of magnesioferrite is not homogeneous (Fig. 5), which means that its distribution between the surface and the bulk may also vary. If magnesioferrite were concentrated near the surface, flank method measurements would overestimate the amount of Fe3?, as we observed in our study (samples Fe6, Fe15 and Fe27). For samples with greater amounts of iron, the size of the magnesioferrite grains would be larger, for example, Reichmann et al. (2000) reported that magnesioferrite in sample Fe75 could be observed with the optical microscope. In this case, the structure of magnetite would have a stronger influence on the Lb/La ratios. We collected flank method measurements on a single crystal of magnetite, which gave Lb/La values (grey stars, Fig. 7) that represent the limit of maximum exsolution. Indeed, the Lb/La values for the most oxidised samples (Fe75 and Fe78ox) tend towards this limit. However, our calibration model predicts that Lb/La values should be much lower for such oxidised compositions (dashed curve labelled 0.67 in Fig. 7), which is likely due to both applying the model outside its calibration limits and the effect of different structures in modifying the Lb/La ratio (Höfer and Brey 2007). Contrib Mineral Petrol (2011) 162:1249–1257 Conclusions and further work We successfully demonstrated the use of the flank method to determine Fe3?/RFe ratios in geophysically relevant compositions of (Mg,Fe)O with an uncertainty comparable to that of Mössbauer spectroscopy. The flank method offers a number of advantages over Mössbauer spectroscopy, however: measuring times of only a few hours (compared to 1–2 days for Mössbauer spectroscopy) and spatial resolution of 1–10 lm (compared to no less than 100 lm for Mössbauer spectroscopy). Thus, the flank method is a suitable technique to determine Fe3?/RFe in ferropericlase inclusions in diamond, which are commonly less than 50 lm in size. Results from measurements on inhomogeneous samples demonstrate the possibility to quantify the variation of Fe3?/RFe with a spatial resolution of a few microns in samples subject to dynamic processes such as diffusion or exsolution. For example, the diffusion coefficients of Fe2? and Fe3? could be measured independently, providing important insight into the kinetics of oxidation–reduction reactions. Measurements of (Mg,Fe)O single crystals containing exsolved (Mg,Fe)Fe2O4 show enhanced Fe3?/RFe in (Mg,Fe)O up to compositions of 40 wt% Fe, consistent with the presence of ferrite grains even when these are too small to be resolved by conventional electron microprobe measurements. Acknowledgments We are grateful to Stephen Mackwell for supplying some of the samples examined in this study, to Hugh O’Neill for valuable advice concerning (Mg,Fe)O synthesis, to Heidi Höfer for her support on the flank method calibration, to Detlef Krauße for his technical support at the electron microprobe, to Nobuyoshi Miyajima for TEM investigations and to Florian Heidelbach for SEM analyses. The manuscript was significantly improved by the comments of Andrew Berry and an anonymous reviewer. ML received financial support from the European Commission under the Marie Curie Action for Early Stage Training of Researchers within the 6th Framework Programme (contract number MEST-CT-2005-019700). SDJ is supported in part by the US NSF (EAR-0748707) and by the David and Lucile Packard Foundation. References Berry AJ, Danyushevsky LV, HStC O’Neill, Newville M, Sutton SR (2008) The oxidation state of iron in komatiitic melt inclusions indicates hot Archean mantle. Nature 455:961–963 De Grave E, Govaert A, Chambaere D, Robbrecht G (1979) Mössbauer effect study of MgFe2O4. Physica 96B:103–110 Dobson DP, Richmond NC, Brodholt JP (1997) A high-temperature electrical conduction mechanism in the lower mantle phase (Mg, Fe)1-xO. Science 275:1779–1781 1257 Eeckhout SG, Bolfan-Casanova N, McCammon C, Klemme S, Amiguet E (2007) XANES study of the oxidation state of Cr in lower mantle phases: periclase and magnesium silicate perovskite. Am Mineral 92:966–972 Hazen RM, Jeanloz R (1984) Wüstite (Fe1-xO): a review of its defect structure and physical properties. Rev Geophys Space Phys 22:37–46 Höfer HE, Brey GP (2007) The iron oxidation state of garnet by electron microprobe: its determination with the flank method combined with major-elements analysis. Am Mineral 92:873–885 Höfer HE, Weinbruch S, McCammon CA, Brey GP (2000) Comparison of two electron probe microanalysis techniques to determine ferric iron in synthetic wüstite samples. Eur J Mineral 12:63–71 Jacobsen SD, Reichmann HJ, Spetzler HA, Mackwell SJ, Smyth JR, Angel RJ, McCammon CA (2002) Structure and elasticity of single-crystal (Mg,Fe)O and a new method of generating shear waves for gigahertz ultrasonic interferometry. J Geophys Res 107(B2):2037. doi:10.1029/2001JB000490 Longo M (2009) Iron oxidation state of (Mg,Fe)O: calibration of the flank method (EMPA) on synthetic samples and applications on natural samples from lower mantle diamonds. Dissertation. University of Bayreuth, Germany McCammon CA (1994) A Mössbauer milliprobe: practical considerations. Hyperfine Interact 92:1235–1239 McCammon C (2001) Deep diamond mysteries. Science 293:813–814 McCammon CA (2004) Mössbauer spectroscopy: Applications. In: Beran A, Libowitsky E (eds) Spectroscopic Methods in Mineralogy, vol 6. Eötvös University Press, Budapest, pp 369–398 McCammon CA, Hutchison M, Harris J (1997) Ferric iron content of mineral inclusions in diamonds from São Luiz: a view into the lower mantle. Science 278:434–436 McCammon CA, Peyronneau JP, Poirier J-P (1998) Low ferric iron content of (Mg, Fe)O at high pressures and high temperatures. Geophys Res Lett 25:1589–1592 McCammon CA, Stachel T, Harris JW (2004) Iron oxidation state in lower mantle mineral assemblages II. Inclusions in diamonds from Kankan, Guinea. Earth Planet Sci Lett 222:423–434 Reichmann HJ, Jacobsen SD, Mackwell SJ, McCammon CA (2000) Sound wave velocities and elastic constants for magnesiowüstite using gigahertz interferometry. Geophys Res Lett 27:799–802 Speidel DH (1967) Phase equilibria in the system MgO-FeO-Fe2O3: the 1300°C isothermal section and extrapolations to other temperatures. J Amer Ceram Soc 50:243–248 van Aken PA, Liebscher B, Styrsa VJ (1998) Quantitative determination of iron oxidation state in minerals using FeL23-edge electron energy-loss near-edge structure spectroscopy. Phys Chem Mineral 25:323–327 van Orman JA, Fei Y, Hauri EH, Wang J (2003) Diffusion in MgO at high pressures: constraints on deformation mechanisms and chemical transport at the core-mantle boundary. Geophys Res Lett 30: doi:10.1029/2002GL016343, 2003 Waychunas GA, Dollase WA, Ross C II (1994) Short-range order measurements in MgO-FeO and MgO-LiFeO, solid solutions by DLS simulation-assisted EXAFS analysis. Am Miner 79:274–288 Wuensch BJ, Vasilos T (1962) Diffusion of transition metal ions in single-crystal MgO. J Chem Phys 36:2917–2922 Yamazaki D, Karato S (2001) Some mineral physics constraints on the rheology and geothermal structure of Earth’s lower mantle. Am Miner 86:385–391 123
© Copyright 2026 Paperzz