1. Introduction

Cubic Fe-Ti oxides, titanomagnetite (Fe_3-xTi_xO₄, 0 < x < 1) and its oxidized (non-stoichiometric) counterpart, titanomaghemite (Fe_(3-x)RTi_xR[ ]_3(1-R)O₄, where [ ] is the cation vacancy, R=8/[8+z(1+x)], 0 < z < 1), are most common magnetic minerals in oceanic crust. Magnetic properties of titanomagnetite or titanomaghemite strongly depend on its composition (x, z). Because titanomaghemites are the main carriers of remanent magnetization in submarine basalts, various approaches were tried in efforts to estimate their compositions. Yet, most estimates available up to date come from the measurements of composition-dependent parameters (such as Curie temperature and lattice constant). These parameters are not well known in the entire compositional field of titanomaghemites (0 < x < 1, 0 < z < 1) and depend on a number of variables rarely taken into account (e.g. the presence of impurities in natural samples).
Obtaining the compositions of titanomaghemite through analytical techniques (e.g., wet chemical analysis, X-ray spectrometry) is difficult because it requires measurements of oxygen content or Fe²⁺/Fe³⁺ ratio. Most analyses therefore simply ignore the oxidation state parameter (z) and give only the relative abundances of cations; the oxidation state must be determined independently to “complete” the analysis. This is an approach of conventional electron probe microanalysis (EPMA) applied to measuring the “cation compositions” of titanomaghemites [e.g. Akimoto et al., 1984; Furuta, 1993].
In a recent study of Zhou et al. [1999], a new technique for measuring z values of submicrometer-sized titanomaghemite grains was introduced. This technique utilizes the convergent-beam electron diffraction (CBED) observations of high-order Laue zone (HOLZ) diffraction patterns using a transmission electron microscope (TEM). A potential advantage of the TEM-CBED method is the ability to measure the oxidation degrees and x values simultaneously on a single titanomaghemite grain and thus to study the compositional variation between the grains in a sample. While this technique seems very promising, it requires specific TEM equipment and probably needs more thorough calibration.
 

2. Previous Work Summary

Most modern wavelength-dispersive (WD) and energy-dispersive (ED) detectors have the ability to detect the characteristic X-ray lines of light elements. However, the quantitative analysis of light elements is difficult, mostly because of the strong absorption of low-energy X-rays (within the sample and/or on the ED detector window/dead layer) and interferences of the light element Ka lines with the L- and M-families of higher atomic number elements. To the best of my knowledge, the only one study reports an EPMA analysis of 10-15 micron titanomaghemite grains in which the concentrations of oxygen were actually measured [Zhou et al., 1999]. In these experiments, a Cameca CAMEBAX electron microprobe was calibrated for O and Fe using magnetite (Fe₃O₄) and hematite (Fe₂O₃) standards and various silicate and oxide standards for other elements. WD spectra were measured at a 15 kV accelerating voltage, 20 nA beam current and normal incidence on a carefully polished surface of specimen. The PAP scheme [Pouchou and Pichoir, 1984] was used to calculate the matrix correction coefficients (Z, A). The analytical total was remarkably close to 100% (99.40-99.98%), suggesting that the effects of absorption and atomic number were appropriately corrected.

A WDS system has two advantages over an EDS system: it gives much higher spectral resolution and higher peak to background ratios. These are important when: (i) characteristic X-ray lines of analyzed elements have similar energies so that they can not be resolved in an ED spectrum and/or (ii) minor elements need to be analyzed. Fortunately, neither condition is met for titanomaghemite. Fe, Ti and O are all major elements. Oxygen is most abundant, so that O Ka has relatively high peak to background ratio despite it is strongly absorbed (Fig. 1). For idealized pure titanomagnetite or titanomaghemite (i.e. when the cations other than Fe and Ti are not present), there only pathological line overlap is that between O Ka (0.525 keV) and Ti La (0.452 keV). Fe La (0.705 keV) only slightly overlaps with O Ka; these two lines are well resolved in ED spectra (Fig. 1). Other lines which can result in a significant overlap with O Ka are V La (0.511 keV), Cr La (0.573 keV) and Mn La (0.637 keV). These elements can be present in titanomaghemite only as minor (a few wt.%) or trace (<1 wt.%) constituents, so their effect on measured O Ka is expected to be negligible. More common impurities (Ca, Mg, Al and Si) do not interfere with the characteristic radiation of oxygen.

Fig. 1.  Representative example of ED spectrum obtained on a ~20 micron grain of titanomaghemite from submarine basalt. EDAX Phoenix EDS system;15 kV.

3. Proposed Work

Argument given in previous section suggests that the intensities of characteristic X-ray lines of all major elements (including oxygen) as well as minor elements (impurities) in titanomaghemite can be measured with acceptable accuracy using an EDS system. The main advantages of a WDS system do not seem to be critical for quantitative X-ray analysis of titanomaghemite composition. The success of Zhou’s et al. [1999] EPMA analyses suggests that appropriate matrix correction coefficients (Z and A) can be calculated by modeling the depth distribution of generated X-rays, Phi(Rho*z), using established formulations [Pouchou and Pichoir, 1984; Bastin and Heijligers, 1991]. Fluorescence corrections (F) for characteristic radiation and continuum are expected to be minor and can be calculated using standard approaches [e.g. Goldstein et al., 1992, and references therein].

Therefore I propose to measure the ED spectra of titanomaghemites and other Fe and Ti oxides using a LEO 982 scanning electron microscope equipped with an EDAX Phoenix EDS system. The main question I want to address is: Can a reliable quantitative analysis of oxygen content in Fe-Ti oxides be obtained from the EDS X-ray data?

A simple test can be performed [Zhou et al., 1999]. Different oxides (e.g., Fe₃O₄, Fe₂O₃, TiO₂, FeTiO₃, etc.) are used as calibration standards for oxygen. Then a titanomaghemite grain of unknown composition is measured at the same operation conditions as standards and its oxygen content is calculated. If the oxygen contents of unknown calculated using different calibrations are consistent and the analytical totals are close to 100%, the matrix corrections (ZAF) were appropriate and the calculated composition is reliable. (Note that the use of standards with compositions close to that of analyzed titanomaghemite is needed to minimize the errors arising from the uncertainties of mass absorption coefficients, which are especially important for oxygen.)

Standardless quantitative EDS analysis (incorporated in the data processing software of EDAX Phoenix EDS system) can be used to estimate the relative cation abundances in titanomaghemites and calculate their x values. It is believed that standardless analysis gives acceptable results when the Ka lines of elements with atomic numbers from 20 to 32 (energies >3 keV) are analyzed. Fe and Ti Ka lines satisfy these conditions, while most of Ka lines of the impurities (e.g. Al, Mg, Si) do not (Fig. 1). Therefore, it is interesting to test how accurate the results of standardless analysis are. This can be done by direct comparison of cation composition data quantified using a set of calibration standards (oxides and/or silicates) with the data obtained through standardless analysis. In this case, the need in standards with compositions close to that of analyzed titanomaghemite is less critical because the mass absorption coefficients for the elements of interest (Fe, Ti, Al, Si, Mg, Ca,…) are relatively well known and oxygen is not analyzed.

4. Samples

Polished thin sections (similar to those used for reflected-light petrography) of submarine basalt samples containing 1-50 micron-sized grains of titanomaghemite will be prepared at the Department of Earth and Environmental Sciences. SEM-specific preparation is the coating with gold to ensure the electrical conductivity of samples.
 

5. Microscopic Techniques

Secondary electron (SE) imaging will be used to observe the surfaces of thin sections to check the quality of polishing and to avoid the effect of rough surfaces on measured oxygen contents.
Backscattered electron (BSE) imaging will be used for identification of titanomaghemite grains in thin sections (due to compositional contrast they look significantly brighter than the silicate matrix minerals, the shapes are also characteristic) and for selection of homogeneous grains with no fractures or inclusions for X-ray analyses.
ED X-ray spectrometry will be used for initial qualitative analysis of titanomaghemite. This will help to decide which standards should be used for quantitative analysis. Availability of high-quality standards with compositions close to that of analyzed titanomaghemite (preferably Fe and Ti oxides) is critical for quantification of oxygen contents.
ED X-ray spectra will be obtained from the standards and titanomaghemite grains of unknown composition under the same operation conditions. Acceleration voltage of 15 kV is sufficient for efficient excitation of all Ka lines of interest (Fig. 1) and provides an interaction region about 1-2 micron in size.
Quantification of X-ray data will be performed using both the standardless analysis (for cation contents) and X-ray data measured on standards. The calculation of matrix correction coefficients (Z, A, F) for specimen-standard measurements will be done by using Phi(Rho*z) method [Pouchou and Pichoir, 1984; Bastin and Heijligers, 1991]. The results of standardless standard-calibrated analyses for the same grains will be compared.
 

References:

Akimoto, T., H. Kinoshita and T. Furuta, Electron probe microanalysis study of the process of low-temperature oxidation of titanomagnetite, Earth Planet. Sci. Lett., 71, 263—268, 1984.

Bastin, G. F., and H. J. M. Heijligers, Quantitative electron probe microanalysis of ultra-light elements (boron-oxygen), in Electron probe quantitation, K. J. F. Heinrich and D. E. Newbury eds., Plenum, New York, 1991, 145—161 pp.

Furuta, T., Magnetic properties and ferromagnetic mineralogy of oceanic basalts, Geophys. J. Int., 113, 95—114, 1993.

Pouchou, L. J., and F. Pichoir, New model quantitative x-ray microanalysis, 1. Application to the analysis of homogeneous samples, Rech. Aerosp., 3, 13—38, 1984.

Zhou, W., D. R. Peacor and R. Van der Voo, Determination of lattice parameter, oxidation state, and composition of individual titanomagnetite/titanomaghemite grains by transmission electron microscopy, J. Geophys. Res., 104, 17689—17702, 1999.
 
 

1. Overview

Quantitative X-ray microanalysis of Fe-Ti oxide minerals (titanomaghemites) was performed using a Leo 982 Scanning Electron Microscope equipped with an Edax Phoenix Energy-dispersive Spectrometry (EDS) system. Quantification of the EDS data was done with a set of natural and synthetic oxide standards. The concentrations of oxygen and cations were calculated from the intensities of measured Ka lines using the Edax PhiZAF matrix correction algorithm; the performance of this algorithm was tested with standard samples. Standardless quantification of the EDS data was compared with the results of standard-calibrated analyses.
Direct measurements of oxygen contents in natural titanomagnetite/titanomaghemite were made previously only by electron microprobe analysis using wavelength-dispersive spectrometry (WDS) [Zhou et al., 1999]. This study suggests that compositional data of similar quality can be obtained from EDS measurements.
 

2. Samples

Two types of samples were used in this study: polished whole rock sections and loose mineral grains. Polished thin sections were made from samples of the Late Cretaceous (~76 Ma) submarine basalt from Detroit Seamount (northwestern Pacific Ocean) using common sample preparation techniques for the reflected light microscopy (Fig. 1). Loose grains of Fe-Ti oxides with known compositions (these were used as standards for O, Ti and Fe for the quantification of EDS data, see Section 5) were embedded into a ~3 mm-thick cylindrical block of epoxy resin, which was ground from one side to expose the grains and polished (Fig. 2). Samples were coated by a thin layer of gold (~30-40 nm), mounted on aluminum stubs and grounded to the stubs with carbon paint to ensure electrical conductivity.

Fig. 1. Polished, gold coated thin section of a submarine basalt.

Fig. 2. Block of oxide standards.

The surfaces of the samples were observed using the secondary electron (SE) imaging to check the quality of the polish. Backscattered electron (BSE) imaging was used to identify Fe-Ti oxide (titanomaghemite) grains in thin sections (Fig. 3).

Fig. 3. BSE image obtained from the sample shown in Fig. 1. White grains are the crystals of titanomaghemite. Gray and dark gray grains are the silicate matrix minerals.

3. Simulations of Electron-Specimen Interactions

Monte Carlo electron trajectory simulations (using the Electron Flight Simulation Software, v. 3.1) were performed to estimate the size of the beam interaction volume in Fe-Ti oxides of various compositions. Input parameters (accelerating voltage, detector type and geometry, etc.) were chosen to match the SEM/EDS conditions at which the ED spectra were measured (Section 4). From these simulations, the depth distributions of generated and emitted O Ka X-rays were estimated.

Figure 3 shows the results of this experiment for the two end-members of titanomagnetite solid solution series: magnetite (Fe₃O₄) and ulvospinel (Fe₂TiO₄). The size of the beam interaction region is almost the same for all compositions of titanomagnetite and titanomaghemite (~2 micrometers). The emitted intensity of O Ka X-rays, however, decreases sharply with the increasing Ti contents due to the strong absorption of O Ka in Ti.
 
 

Fig. 3. Simulations of electron beam interaction volume in magnetite and ulvospinel. Blue lines are electron trajectories. Green dots, generated O Ka X-rays; red dots, emitted O Ka X-rays. Depth distributions of generated and emitted O Ka X-rays are shown as green and red histograms, respectively.

4. Measurements of the ED Spectra and Qualitative X-ray Analysis

The size of the primary electron interaction volume was found to be approximately 2 micrometers for all compositions of  titanomagnetite or titanomaghemite for a focused beam at 15 kV accelerating voltage (Section 3). However, the range of characteristic fluorescence of O Ka (by Ti and Fe Ka,b) and Ti Ka (by Fe Ka,b) is expected to be at least an order of magnitude greater than the primary excitation volume. Therefore, only homogeneous grains larger than 20 micrometers were used for EDS measurements to ensure that the Ka intensities obtained from the ED spectra are accurate for all elements present in the grain. Smaller grains and those with multiple cracks subdividing the grain into the regions smaller than 20 micrometers were not used.

The ED spectra were measured at 15 kV accelerating voltage, which is optimal for excitation of Fe K shell. The energy resolution of the EDS system was estimated to be 128 eV FWHM for Mn Ka. Working distance was 20 mm, resulting in a 43.3^o takeoff angle. The acquisition time was preset to 120 live seconds for all measurements.

Identification of elements (qualitative analysis) in measured spectra and spectral data processing (background subtraction, peak deconvolution, calculation of the Ka intensities and their uncertainties) were done using the Edax SEM Quant v.3.2 software. Qualitative analysis showed that Mg and Al are present in minor amounts in titanomaghemite grains from submarine basalt (Fig. 4). Other impurities (such as Si, Cr and Mn) either were not detected or yielded very low peak intensities (peak to background ratios less than 1), suggesting that only trace amounts of these elements might be present in titanomaghemite from our samples.

Fig. 4. Representative example of the ED spectrum from titanomaghemite.

5. Quantification of EDS Data Using Standards

Quantification of EDS data was done with a set of natural and synthetic oxide standards. Natural magnetite (Fe₃O₄) was used as a standard for O and Fe (Fig. 5). Stoichiometry of magnetite was confirmed by thermomagnetic measurements, which showed Curie temperature of 580 ^oC and sharp Verwey transition at 153 ^oC, characteristic for pure, unoxidized Fe₃O₄. Natural rutile (TiO₂) sample was used as a standard for Ti (Fig. 6); synthetic MgO and Al₂O₃ were used as standards for Mg and Al.
 

a.

b.

a.

b.

Edax PhiZAF calculation scheme (an Edax implementation of the Phi(Rho*z) method included into the SEM Quant software package) was used to calculate pure element intensities from standards (reverse ZAF calculation) and matrix correction coefficients (Z, A and F) for samples with unknown compositions.

To test the performance of quantification procedure, the ED spectra were measured from synthetic ilmenite (FeTiO₃) (Fig. 7) and quantified with magnetite and rutile standards. The ilmenite powder (particle size: 5-100 microns) was obtained from Alfa Aesar; it is certified as Fe²⁺-Ti oxide of 99.8% purity, the rhombohedral structure of ilmenite was confirmed by X-ray powder diffraction.
 

a.

b.

The results of quantitative EDS analysis of large grains (50-100 microns) of synthetic ilmenite are shown in Table 1. Measured concentrations of O, Ti and Fe are reasonably consistent with the expected (theoretical) values; the discrepancy is generally less than 1 wt.% for O and <1.5 wt.% for Ti and Fe. The analytical total is remarkably close to 100%. This suggests that our quantification procedure (standardization with magnetite and rutile + PhiZAF matrix correction scheme) is appropriate for the measurements of oxygen and cation contents of Fe-Ti oxides with the accuracy of ~1-1.5 wt.%
 
 

6. Compositions  and Oxidation States of Titanomaghemite

Figure 8 shows the examples of EDS data collected from large (~30-50 microns) grains of titanomaghemite from a submarine basalt sample. Compositions of titanomaghemite were calculated using the quantification procedure described in previous section. Oxide standards were measured along with titanomaghemite. The calibration for analyzed elements (pure element intensities calculated from standards) was checked between the measurements of individual titanomaghemite grains to ensure that the data from standards and unknown are collected at the same beam conditions. The beam was usually quite stable and no recalibration was required.
 
 

a.	b.
c.	d.

The results of quantitative microanalysis (Table 2) suggest that the ulvospinel content of original (not oxidized) titanomagnetite (x₀ parameter, see Table 2 for the definition) varies within the sample between x₀ = 0.57 and x₀ = 0.67. This range is typical for oceanic basalts. The oxidation state (z, Table 2) of titanomaghemite grains ranges from <0.2 (slightly non-stoichiometric titanomagnetite) to ~0.8 (severely oxidized titanomagnetite).

The errors of calculated x₀ and z values were estimated from the uncertainties of measured intensities of the characteristic X-ray lines. As an approximation, I considered the standard deviation of concentration of an element (in %) to be equal that of its Ka line intensity (which was estimated from counting statistics). The 95% confidence errors are generally ~0.018-0.02 for x₀ and ~0.1-0.15 for z. These errors, however, are likely to be underestimated because the errors related to standardization procedure (Section 5) were not accounted for. A 1 wt.% systematic error in measured contents of O, Ti and Fe can add up to ~0.1 (~15%) to the total error of x₀ and up to ~0.3 (~40-60% for z > 0.5) to the total error of z, but this is perhaps too conservative. More calibration tests (similar to that with ilmenite sample, described in Section 5) are needed to estimate the errors of x₀ and z more precisely.
 
 

The oxidation degree (z) depends on permeability of titanomaghemite grains to oxidizing agents (i.e. hydrothermal fluids), which is determined mostly by surface to volume ratio and by the presence of cracks within the grain, creating additional surfaces to promote the oxidation. The oxidation parameter is therefore expected to be lower for large, continuous grains and higher for smaller grains or grains with multiple cracks. It is also expected to vary within large grains, being highest near the surface and lowest in the central part of the grain. Both the distribution of the grain sizes and the intrinsic inhomegeneity of large grains probably contribute to the large variation of titanomaghemite oxidation states in our basalt sample.
 
 

7. Standardless Quantitative Analysis

Standardless quantitative EDS analysis is believed to work reasonably well when the Ka lines of elements with atomic numbers from 20 to 32 (> 3 keV) are used (e.g. Fe and Ti Ka). To test this "rule of the thumb", I performed the standardless analysis of the ED spectra from Fe-Ti oxide standards and titanomaghemite grains and compared its results with the standard-calibrated quantification data (Table 3, Fig. 9). The PhiZAF matrix correction scheme and the Edax default table of Standardless Element Coefficients (SEC) were used for standardless quantification.
As it was anticipated, the standardless quantification procedure fails to accurately estimate the absolute oxygen content in Fe-Ti oxide samples (Table 3). Concentrations of O were always lower (by ~9 wt.%) than their true values, suggesting that the absorption is not adequately accounted for in standardless calculation. Because the totals are always normalized to 100% in a standardless analysis, the cation concentrations are artificially higher than their true values. However, the relative abundances of Fe and Ti cations (i.e. Fe/Ti atomic ratios) calculated using the standardless approach are very close to those estimated from quantification with standards (Fig. 9).  The standardless values of Fe/Ti ratio are slightly, but systematically higher than those calculated using standards. Although the difference is not statistically significant at 95% confidence level, this suggests a small systematic error arising from slightly inaccurate SEC values or standard compositional data, or both. It is not clear at this point which quantification procedure is preferable for the measurements of Fe/Ti ratios in Fe-Ti oxides. Both quantification methods seem to produce results of acceptable quality.
 
 

Fig. 9. Comparison of Fe/Ti cation ratios in titanomaghemites estimated from standardless and standard-calibratied quantitative X-ray microanalysis. Error bars are 95% confidence intervals.

8. Summary

Energy-dispersive spectroscopy can be successfully used for the analysis of chemical compositions of natural Fe-Ti oxide minerals. The use of a thin window/windowless ED detector capable of detecting the low-energy characteristic X-rays of light elements and calibration of the EDS system with a set of well characterized standards (preferably oxides with compositions close to those of analyzed minerals) make it possible to directly measure the absolute oxygen contents, eliminating thus the necessity of indirect estimation of oxidation state. The error of oxygen measurements achieved in this study is estimated to be not greater than 1 wt.%. The accuracy can probably be significantly improved by using higher quality standards. The performance of the Edax PhiZAF algorithm was found to be satisfactory for calculation of the matrix correction coefficients for all analyzed elements, including oxygen, when standards were used. Standardless quantification of EDS data showed that the relative abundances of cations in Fe-Ti oxides (Fe/Ti ratios) calculated without standards are reliable. The Fe/Ti ratios calculated with standards do not differ significantly from standardless estimates. Standardless quantitative analysis can thus be used to estimate the x values of titanomagnetite.
 
 
 

Acknowledgments

I wish to thank Brian McIntyre for his help and guidance, Gery Kloc for preparation of polished thin sections, and Alexei Smirnov, who sacrificed his favorite magnetite crystal in the name of Science.
 
 
 

Please contact me if you have any questions, criticism or suggestions:

Pavel Doubrovine
Hutchison Hall 227
Earth & Environmental Sciences
University of Rochester
Rochester, NY 14627
(585) 275-8810
pavel@earth.rochester.edu