- © 2012 E. Schweizerbart'sche Verlagsbuchhandlung Science Publishers
Single crystals of pure Mg2Si2O6 orthoenstatite synthesized at 25–80 kbar were analysed by secondary ion mass spectrometry (SIMS) and Fourier transform infrared spectroscopy (FTIR) with respect to water. The IR spectra of the crystals exhibited two groups of OH-absorption bands in varying proportions, ascribed to a varying ratio of two different OH-defects. Different strategies for quantification of the water content based on the comparison of the IR spectra and the SIMS data are evaluated. Results suggest that one single mineral-specific absorption coefficient for enstatite is not useful, and that site-specific absorption coefficients (i.e., coefficients specific to each absorption band) are not necessary. Instead, a function for the absorption coefficient dependence on wavelength could be established. This function is similar to previously formulated wavelength-dependent absorption coefficients for nominally anhydrous minerals, hydrous minerals and water-bearing glasses.
The Earth’s mantle mainly consists of nominally anhydrous minerals, which are able to incorporate traces of water as OH-point defects or lamellae of hydrous minerals. H2O is an important component in geochemical cycles and petrologic processes such as partial melting, and the presence and concentration of protons affect significantly the physical properties of minerals (Karato, 1990; Mei & Kohlstedt, 2000). Therefore, the exact amount of water stored in the mantle is of particular interest to many branches of geosciences, such as mineral physics, geophysics, geochemistry, petrology and volcanology.
The different types of OH-point defects in nominally anhydrous minerals give rise to characteristic infrared (IR) absorption bands, and IR spectroscopy is a powerful method to characterize both the local environment and the concentration of OH-defects, provided that a calibration with a direct method for water analysis exists. Several strategies for the quantification of the water content based on IR spectra have been put forward, including (1) mineral-specific absorption coefficients (e.g., Bell et al., 1995, 2003), (2) wavelength-specific absorption coefficients (Paterson, 1982; Libowitzky & Rossman, 1997), and (3) site-specific absorption coefficients (Kovacs et al., 2010).
1.1. Wavelength-specific calibrations
In general OH-vibrations in hydrous minerals show a broad energy range due to different strength of the OH-dipole and H-bonding in O-H · · · O arrangements (where H · · · O symbolizes the H bond), and an increasing H · · · O bond length is correlated with an increasing wavenumber (energy) of the OH-absorption band (Libowitzky, 1999). Previous attempts to quantify OH from IR spectroscopy measurements on silicate glasses (Paterson, 1982) and hydrous minerals (Skogby & Rossman, 1991; Libowitzky & Rossman, 1997) found a linear negative correlation between the absorption coefficient ɛ and the wavenumber of the respective OH-band. This general trend has been confirmed by ab initio and density functional theory calculations (Kubicki et al., 1993; Balan et al., 2008), in which both the slope and the intersection with the abscissa vary significantly (Fig. 1).
1.2. Mineral-specific calibrations
In order to improve the quality of quantitative water measurements in nominally anhydrous minerals derived by IR spectroscopy, mineral-specific absorption coefficients were determined by calibration against other analytical methods, e.g. vacuum extraction manometry (Bell et al., 1995), NMR spectroscopy (Johnson & Rossman, 2003), secondary ion mass spectrometry (Aubaud et al., 2007), and proton-proton scattering (Thomas et al., 2009). Mineral-specific absorption coefficients often deviate strongly from the values obtained from the wavelength-dependent calibration, and for olivine the application of mineral-specific rather than wavelength-specific calibration systematically leads to higher water contents (Bell et al., 2003; Aubaud et al., 2007). However, the overall pattern is not conclusive: on the one hand, a general mineral-specific absorption coefficient for all feldspars (Johnson & Rossman, 2003) could be established, even if the average wavenumber of IR bands in sanidine is much higher than in plagioclase, on the other hand it was pointed out for olivine that the application of a mineral-specific calibration can only be applied, when IR characteristics of specimen and calibration standard are similar (Bell et al., 2003). The latter conclusion is confirmed by a study on water in synthetic clinopyroxene (Stalder & Ludwig, 2007), where the application of the mineral-specific absorption coefficient determined by Bell et al. (1995) led to a systematic overestimation of water content, because of fundamental differences with respect to the shape of the IR spectra of analyzed samples and reference material.
1.3. Structure-specific calibrations
In a polymorph series OH absorption coefficients vary systematically as a function of the structure, and the mineral-specific absorption coefficient is negatively correlated to the molar volume, e.g. SiO2: ɛStish > 5 · ɛQz (Thomas et al., 2009) and Mg2SiO4 (Koch-Müller & Rhede, 2010). Within one solid-solution series the absorption coefficients follow the same trend as expected from the wavelength-specific calibrations.
1.4. Site-specific calibration
It is widely accepted from results of experimental and analytical work that water incorporation in nominally anhydrous minerals is strongly enhanced by cationic substitutions. Theoretical calculations have shown that substitutions can shift absorption bands by up to 100 cm−1 (Kubicki et al., 1993). Thus, each local environment actually has to be considered separately and, consequently, a site-specific calibration for water determination in olivine has recently been proposed (Kovacs et al., 2010). It has been shown that each substitution mechanism (revealed by distinct IR absorption features) needs an own specific absorption coefficient. In particular, the specific absorption coefficient for the IR band generated by hydrogarnet substitution (tetrahedral defects) differs by a factor of 20 from the coefficient for the IR band generated by protons charge balancing Mg-vacancies (octahedral defects). According to all existing wavelength-specific calibrations these two coefficients should only deviate by a factor of about 3.
From the plethora of previous publications on IR calibrations for OH in minerals it is clear that a general calibration of IR absorbance for all minerals cannot lead to accurate water contents (Balan et al., 2008; Koch-Müller & Rhede, 2010) and that mineral-specific absorption coefficients are only reliable if the IR spectra are similar in shape. Therefore, for each OH-defect in each mineral the absorption coefficient is in principal a priori unknown and has to be determined experimentally.
One hitherto unanswered question refers to the specific absorption coefficients of different types of OH-defects in orthopyroxene, the second most abundant mineral of the Earth’s upper mantle. Here we present IR and secondary ion mass spectrometry (SIMS) measurements on a series of pure, synthetic Mg2Si2O6 enstatite – an important end-member of orthopyroxene in the Earth’s mantle.
2. Mineral synthesis and analytical procedure
2.1. Mineral synthesis
Starting mixtures were prepared from Mg(OH)2 (Alfa Aesar) and SiO2 (Alfa Aesar, 99.995 %) and sealed in a platinum capsule with an outer (inner) diameter of 3.0 (2.6) mm. Synthesis runs were performed between 1150 and 1250 °C and between 40 and 80 kbar in a 1000 t press equipped with a Walker-type multianvil module using 25/15 assemblies made of MgO + 5 % Cr2O3 with a graphite-furnace. The temperature was measured with a Pt100-Pt90Rh10 thermocouple, and both pressure and temperature were computer-controlled during the entire duration of the runs. The synthesis procedure is described in more detail in Prechtel & Stalder (2010 in Prechtel & Stalder (2011).
2.2. FTIR spectroscopy
Single crystals of enstatite were handpicked from the experimental run product and aligned parallel (010) or (100) in a thermoplastic resin. All crystals were polished double-sided to a thickness of 85–320 μm. After preparation, the thermoplastic resin was removed by rinsing in acetone. The thickness of each crystal was measured with a mechanical micrometer with an accuracy of ± 2 μm. Mid-infrared absorption spectra were recorded at room temperature in transmission mode using a Bruker Vertex 70 FTIR spectrometer, coupled to a Hyperion 3000 microscope equipped with nitrogen-cooled MCT-D316-025 detector, a silicon carbide (SiC) globar, a KBr beam splitter, and a ZnSe wire grid polarizer. From each experimental charge one crystal section//(010) and one//(100) were measured parallel to main refractive indices (i.e., E//nα and E//nγ in the (100) section, and E//nβ and E//nγ in the (010) section). Each spectrum was acquired by 64 scans in the 550–8000 cm−1 range with a spectral resolution of 2 cm−1. Spectra recorded for E//nα were very similar to the ones recorded for E//nγ and constantly showed lower intensity by a factor of 2 (Prechtel & Stalder, 2010). Since only the (010) section was measured by SIMS, measurements from the (100) section were only used to derive the absorbance ratio A//nα to A//nγ. This ratio was then multiplied with the A//nγ measured on the (010) section to derive the value for A//nα. The total absorbance Atot was then derived by adding A//nα, A//nβ and A//nγ. This procedure to determine A//nα was performed in order to eliminate variation in the total amount of water between different crystals from one charge. The obtained spectrum was then corrected by a linear baseline between 2800 and 3730 cm−1, and divided into four spectral regions: 3730–3640 cm−1, 3640–3475 cm−1, 3475–3290 cm−1, 3290–2800 cm−1, from which the respective absorbance were calculated (Table 1).
2.3. Secondary ion mass spectrometry (SIMS)
Only (010) oriented crystals were measured by SIMS. Hydrogen analyses (Table 1) were performed at the University of Heidelberg on a Cameca ims 3f at low mass resolution (m//Δm = 400), at an offset of 75 eV and with the energy window set to 40 eV. The primary beam was formed by 16O-ions at a net energy of 14.5 keV and with beam currents ranging from 5 to 50 nA. In situ contamination with water was minimised with a liquid-nitrogen cooling finger in the sample chamber. For most of the samples the method (denoted as method “A” in Table 1) described in Ludwig & Stalder (2007), using multiple beam currents, was applied. Since the in situ H contamination was found to be very low compared to the actual H concentration in the samples, some analyses were performed with a focused primary beam at a fixed 40 nA beam current (denoted as method “B”). Sample 028c2 was analysed using both methods: method “A” resulted in 410 μg/g compared to 406 μ/g using method “B”, thus proving that the difference between method “A” and “B” was negligible. Quantification was done using relative ion yields (reference isotope 30Si). Three different tourmalines (Dyar et al., 2001) and a hydrated glass (equivalent to glass “CG 1” in Acosta-Vigil et al., 2006) were used as reference material. The range of relative ion yields for these tourmalines (0.08–0.12) and the hydrated glass (0.093) suggests an accuracy of approx. ±20 % for the SIMS analyses of H.
3. Results and discussion
The IR spectra show two groups of absorption bands (Fig. 2), group 1 at high wavenumber with maximum intensity for the polarization E//nβ, and group 2 at lower wavenumber with maximum intensity for E//nγ (Prechtel & Stalder, 2010, 2011). Group 1 and group 2 were further subdivided in two absorption bands, respectively (A1 to A4). For each peak the area was determined as described in the previous section, and the weighed centre of each of the regions was determined, which in all cases was within a few wavenumbers close to the maximum of the respective absorption band (A1 = 3690 cm−1, A2 = 3590 cm−1, A3 = 3360 cm−1, A4 = 3090 cm−1). In the next step from the integral absorbances obtained in this way (Table 1) water contents were calculated using different methods: (1) only the total absorbance Atot and the mineral-specific absorption coefficient of Bell et al. (1995) was taken into account; (2) for each of the four peak regions an absorption coefficient according to Libowitzky & Rossman (1997) was calculated; (3) a modified wavelength-dependent function for the absorption coefficient was applied. The obtained water contents determined by the three above mentioned methods were plotted against the water contents determined by SIMS (Fig. 3). Crystal 034c1 – the only specimen which showed significant deviations from alignment perpendicular to the acute bisectrix – is an outlier in all three correlations. The misorientation leads to considerably lower absorbances for the OH-bands polarised//nβ (i.e., the high-wavenumber bands) and therefore to disproportionately lower calculated water contents in the dependent-dependent calibrations. Therefore this data point was not further considered. Water concentrations calculated using the mineral-specific calibration show a large scatter, especially for the samples with high water content, which in turn show the largest variations in the intensity of group 1 peaks (Table 1). Water concentrations calculated using the wavelength-specific calibration of Libowitzky & Rossman (1997), i.e. ɛ = 246.6 · (3753-υ), show a much lower scatter and overall concentrations of about 60 % of the SIMS data (Fig. 3), in accord to Stalder et al. (2005). Modification of υ0 of the wavelength-specific calibration curve yields still better correlation coefficients, and for the present data set a value of 3728 cm−1 furnishes the best correlation (Fig. 4). It has, however, to be noted that the improvement in correlation coefficient by the modification of υ0 has only limited significance, and in principal a scaling of the calibration of Libowitzky & Rossman (1997) with a factor of 1.7 would also do for the present data set. In comparison to other wavelength-dependent curves (Fig. 1), the correlation proposed here, i.e. ɛ = 172.4 · (3728-υ), is rather shallow and has a slightly lower υ0 than all other wave-number-dependent calibrations. The two largest sources of error in our study are the background correction of the IR spectra and the accuracy of the SIMS data, but even both together would not be able to eliminate the discrepancies between slopes of previous and the current study. The background correction in this study was linear and therefore less rigorous than in other studies, where the area of the absorption bands was much more reduced during background subtraction. If the background would be subtracted more rigorously, the discrepancy between the calibration of Libowitzky & Rossman (1997) towards the SIMS data presented here would be still larger. Other sources of error are negligible compared to the aforementioned sources: (a) the thickness was accurate within 2 μm compared to sample thicknesses between 85 and 320 μm, which gives an average error of 1 %. (b) The position of the weighed centre of the absorption bands is accurate within a few wavenumbers, leading to a wavenumber-dependent error of the extinction coefficient (1 % for A3 and A4 to 6 % for A1). If it is taken into account that the area of A1 is always smaller than 15 % of the total absorbance, the total error arising from the uncertainty in the peak centre positions is between 1 and 2 %. (c) The error arising from the misorientation is even smaller than (a) and (b), since all samples were aligned within a few degrees resulting in an error less than 1 %. All together the errors in orientation, peak position and thickness sum up to less than 4 %, which is within the size of the symbols in Fig. 3.
The aim of this study is not a general revision of the wavenumber-specific IR calibration. Therefore, the proposed calibration curve has no general application to other minerals and other pyroxene end-members, it just highlights the superiority of wavenumber-specific over mineral-specific absorption coefficients for our special case (note that other mineral systems such as feldspars show different behaviour, as has been summarized in the introduction). As proposed in other studies on OH in pyroxenes (e.g., Stalder & Ludwig, 2007), the use of a mineral-specific absorption coefficient is only advisable if the shape of the IR spectra of sample and standard are similar (Bell et al., 2003). It has also to be stressed that any trial to improve the correlation by using site-specific absorption coefficients (i.e., one individual absorption coefficient for each absorption band) failed to produce a better correlation than the general wavenumber-specific calibration. Therefore, in contrast to olivine, where each OH-band (caused by a specific OH-defect) seems to require its own absorption coefficient (Kovacs et al., 2010), site-specific absorption coefficients for group 1 and group 2 OH absorption bands in pure enstatite are not appropriate.
- Received 4 March 2011.
- Modified version received 20 August 2011.
- Accepted 16 December 2011.