Site-specific infrared OH absorption coefficients for water ... - MFGI

American Mineralogist, Volume 95, pages 292–299, 2010

Site-specific infrared O-H absorption coefficients for water substitution into olivine István Kovács,1,2,* HugH st.c. o’neIll,1 Jörg Hermann,1 and erIK H. HaurI3 1 Research School of Earth Sciences, The Australian National University, Building 61, Mills Road, Canberra ACT 0200, Australia Department of Data Management, Eötvös Loránd Geophysical Institute of Hungary, Columbus út 17-23, 1145, Budapest, Hungary 3 Department of Terrestrial Magnetism, Carnegie Institution of Washington, 5241 Broad Branch Road NW, Washington, D.C. 20015-1305, U.S.A.

2

abstract There are four ways by which OH– commonly substitutes into olivine, namely those associated with (1) Si vacancies, (2) Mg vacancies, (3) Ti, or (4) trivalent cations. The four mechanisms, which we label [Si], [Mg], [Ti], and [triv], respectively, may each be fingerprinted by their characteristic O-H stretching modes in the infrared spectrum. We show by comparing the integrated intensities of these characteristic infrared peaks against total water content analyzed by secondary ion mass spectrometry, obtained for a suite of synthetic olivines plus one natural olivine, that the different substitution mechanisms require different absorption coefficients (k). For [Ti], we find k = 0.18 ± 0.07, identical to the value previously obtained from natural olivines in which the water was mainly associated with [Ti] defects. Values of k for [Si] and [triv] are 0.57 ± 0.04 and 0.18 ± 0.05, respectively; that for [Mg] is too small to be accurately determined (0.03 ± 0.03). The values of k for [Ti] and [Si] defects differ by a factor of three even though their average wavenumbers are virtually the same. The [Ti] and [triv] defects, on the other hand, have very similar absorption coefficients at significantly different wavenumbers. This highlights the inadequacy of using wavenumber-dependent calibrations for olivine and presumably for NAMs in general. Different substitution mechanisms have their own crystallographic environments that determine their absorption coefficients. The large variation in absorption coefficients within a single mineral emphasizes the importance of distinguishing the substitution mechanism if meaningful quantitative results are to be obtained from infrared spectroscopy. Keywords: Nominally anhydrous minerals, olivine, infrared spectroscopy, secondary ion mass spectrometry, extinction coefficients

IntroductIon The variety and complexity of the infrared spectra of olivine associated with O-H stretching modes attest to the great number of ways by which hydroxyl (OH–, colloquially called “water”) can reside in this nominally anhydrous mineral (e.g., Beran and Putnis 1983; Miller et al. 1987; Libowitzky and Beran 1995; Matsyuk and Langer 2004). However, four substitution mechanisms, each identified from their characteristic bands in the infrared absorption spectrum, stand out as being commonly observed in natural mantle-derived olivine and in synthetic forsteritic olivine synthesized at mantle temperatures and pressures. These substitution mechanisms are associated with Si vacancies, Mg vacancies, octahedral Ti, and trivalent cations, which hereafter we label as [Si], [Mg], [Ti], and [triv]. The Kröger-Vink notation for these defects is: [Si] = (4H)XSi, [Mg] = (2H)XMg, [Ti] = {(Ti4+)••Mg ′ (2H)Si″ }, and [triv] = {(Me3+)•Mg(H)Mg }, respectively. Potentially, deconvoluting the infrared spectrum into the integrated absorption bands for each mechanism could yield the amounts of water substituting by that mechanism, but to obtain this information requires that the absorption coefficient for each mechanism be known. By contrast, methods of analysis such as secondary ion mass spectrometry (SIMS) give total water content, irrespective * E-mail: [email protected] 0003-004X/10/0203–292$05.00/DOI: 10.2138/am.2010.3313

292

of the substitution mechanism, and thus provide no information on how the water is incorporated. For olivine and probably other nominally anhydrous minerals (NAMs), such information is needed to assess at what stage of the mineral’s history the water was acquired—for example, whether the water was present at the time of the original crystallization of the host rock, or was added subsequently by hydrothermal activity. Here we obtain infrared absorption coefficients specific to each substitution mechanism by determining total water by SIMS and relating this to the infrared spectra of olivine specimens containing the four major types of OH-related defects in different proportions. The four major substitution mechanisms for water in olivine have the following characteristics: (1) [Si]. Four H atoms charge-balance a Si vacancy in a tetrahedral site by being bound to O atoms at the apices of the tetrahedron, producing the stoichiometry Mg2H4O4 (Lemaire et al. 2004; Walker et al. 2007). This mechanism produces O-H absorption peaks between 3630 and 3400 cm–1, the most prominent of which are at 3613, 3580, 3567, and 3480 cm–1. The identification of these absorption peaks as related to Si vacancies follows their appearance in olivine from a silica-deficient environment and is supported by ab initio calculations (Walker et al. 2007). On the other hand, Kudoh et al. (2006) and Smyth et al. (2006) attributed the same set of IR peaks to Mg vacancies,

293

KOVáCS ET AL.: SITE-SPECIFIC INFRARED O-H ABSORPTION COEFFICIENTS

based on X-ray diffraction studies. However, ab-initio calculations have shown that the OH-dipoles are oriented in a different way in a Si vacancy than in a Mg vacancy (Walker et al. 2007). The comparison of these calculated orientations with polarized IR spectroscopy supports the assignment of these peaks at high wavenumbers to Si vacancies (Walker et al. 2007). (2) [Mg]. Mg vacancies, where two H+ substitute for divalent cations (commonly Mg) in the octahedral sites with stoichiometry MgH2SiO4 (Berry et al. 2005; Lemaire et al. 2004; Walker et al. 2007). This mechanism results in IR absorption peaks between 3300–3100 cm–1 including two prominent peaks at 3220 and 3160 cm–1. (3) [Ti]. A titanoclinohumite-like point-defect, MgTi[6]H2O4. In this mechanism, Ti4+ is in an octahedral site charge-balanced by a substitution of Si4+ by two protons on a neighboring tetrahedral site. This mechanism produces two prominent absorption peaks at 3572 and 3525 cm–1 (Berry et al. 2005, 2007a; Walker et al. 2007). (4) [triv]. The absorption peaks between 3400 and 3300 cm–1 in olivine are related to the incorporation of trivalent cations (Berry et al. 2007b). The mechanism is a coupled substitution of M3+ and H+ for 2M2+ producing the stoichiometry M3+HSiO4. The absorption bands between 3650 and 3450 cm–1, which include both [Si] and [Ti] peaks, have been referred to widely in the literature as “group I” peaks. The [triv] and [Mg] peaks have been referred to collectively as “group II,” although the two kinds are rather distinct (Bai and Kohlstedt 1993; Lemaire et al. 2004; Matveev et al. 2005). In the almost complete absence of empirical data, early work concerning absorption coefficients for quantifying water in NAMs by Fourier transform infrared (FTIR) spectroscopy adopted some very broad assumptions. The simplest approach is the wavenumber specific calibration (Paterson 1982; Libowitzky and Rossman 1997), where it is assumed that the integrated molar extinction coefficient (ε) of water in different substances (i.e., glass, minerals) shows an inverse correlation with wavenumber (ν). The integrated molar extinction coefficient (ε) is related to the absorption coefficient (k) through k=

cM A ρ⋅ε

(1)

where the thickness is unity (cm), MA is the molar weight of the absorbing species (18.02 g/mol for water), ρ is density of the absorbing mineral (3240 g/L for olivine), and the factor c is to convert from units of moles per liter to µg of H2O per g of olivine (i.e., ppm water by weight). Part of the initial attraction of this approach was that the calibration of extinction coefficient against wavenumber could be obtained from substances with high water contents, which could be determined by the then available methods (such as Karl-Fischer titration on glasses), or by stoichiometry; but theoretical investigation has shown that this assumption is an oversimplification that is likely to produce erroneous results if anything more than order-of-magnitude estimates are desired (Balan et al. 2008). A more advanced approach is to calibrate the absorbance in each mineral individually. For NAMS, mineralspecific calibration factors have been determined for olivine, orthopyroxene, clinopyroxene, and garnet (Bell et al. 1995,

2003, 2004). However, if the absorption coefficient depends on the crystallographic environment, then the existence of different kinds of substitution mechanisms means that even these mineralspecific calibration factors should only be used for those samples that are identical to the calibration specimens, not only in the kinds of mechanisms, but also their relative proportions. For olivine, the only available calibration factor is that of Bell et al. (2003), based on just three olivine grains that have predominantly [Ti] peaks in their IR spectra. Nevertheless, this calibration has been applied to all olivine regardless of their water substitution mechanisms, which, as we shall show, is inappropriate.

experImental metHods The aim is to determine absorption coefficients (k) for the different substitution mechanisms in olivine using multiple linear regression to fit the relationship g] CH2O(SIMS) = k[Si] · A[Stoi]t/cm + k[Ti] · A[toTi]t/cm + k[triv] · A[ttorit/v]cm + k[Mg] · A[toMt/cm .

(2)

We used SIMS to measure absolute water concentration, and the method of Kovács et al. (2008) for determining the total integrated absorbances for the different substitution mechanisms by FTIR spectroscopy with unpolarized light. We require a selection of olivine specimens containing water substituting by all four major substitution mechanisms in different proportions, which can be achieved by high-pressure, high-temperature syntheses. The natural “Pakistani” olivine from Kovács et al. (2008), which contains predominantly [Si] peaks, has also been included in this study. One difficulty is to resolve the contributions of the [Si] and [Ti] peaks because these substitutions coincide in the spectrum and only the position of the main peaks can be distinguished; this obstacle can be circumvented experimentally by synthesizing olivine under conditions selected to produce [Si] peaks alone or [Ti] peaks alone.

experImental procedures To produce different substitution mechanisms in olivine, the following experiments were undertaken (Table 1): (1) periclase-buffered (p) experiments in the MgO+SiO2+H2O system for [Si] peaks; (2) enstatite-buffered (e) experiments in the MgO+SiO2+H2O for [Mg] bands; (3) enstatite-buffered experiments in the MgO+SiO 2+TiO2+H2O system (et) for [Ti] peaks; and (4) enstatite-buffered experiments in the MgO+SiO2+Sc2O3+H2O systems (es) for [triv] peaks. Two such enstatite-buffered syntheses also contained ZrO2 (ez) or ReO2 (er) (10 and 1 wt%, respectively) in an attempt to investigate whether Zr4+ or Re4+ produced OH defects similar to those with Table 1. Experimental conditions for olivine standards Label*

P (kbar)

T (°C)

Duration (h)

Dry mix

Excess Maximum water grain size (µm) MgO SiO2 H2O (wt%) (wt%) (wt%) p_15_1100 15 1100 48 68.9 31.1 ~10% 50–100 p_25_1000 25 1000 48 68.9 31.1 ~10% 100–150 p_25_1100 25 1100 48 68.9 31.1 ~10% 50–120 p_25_1200 25 1200 48 68.9 31.1 ~10% 200–300 p_25_1300 25 1300 48 68.9 31.1 ~10% 80–120 p_25_1400 25 1400 48 68.9 31.1 ~10% 80–120 e_15_1100 15 1100 48 47.3 52.7 ~10% 70–150 e_35_1100 35 1100 48 47.3 52.7 ~10% 50–100 er (Re4+)† 15 1400 24 52.1 47.9 ~7% 100–400 ez (Zr4+)† 15 1400 24 52.1 47.9 ~7% 100–400 et (Ti4+)† 15 1400 24 52.1 47.9 ~7% 100–400 es (Sc3+)† 15 1400 24 52.1 47.9 ~7% 100–400 * p = periclase-buffered experiments, e = enstatite-buffered experiments. † 1 wt% TiO2, 1 wt% Sc2O3, 1 wt% ReO2, and 10 wt% ZrO2 was added to the dry mix of et, es, er, and ez, respectively.

294


Ti4+, but the samples displayed mainly the [Mg] peaks expected from the buffering by orthopyroxene (Table 1). The mixes were made up from analytical grade oxides, dried at 1100 °C, and ground under acetone, which was then driven off by heating to 300–500 °C in a furnace. The first two groups of experiments (excluding er and ez) were originally undertaken to study the effect of pressure, temperature, and chemistry on water solubility in simple experimental systems, but were subsequently found to be suitable for calibration purposes. These experiments (p and e) were conducted at different pressures (15, 25, and 35 kbar) and temperatures (1000–1400 °C), with a duration of 48 h (Table 1). For these experiments, ~50 mg of the mix and 5–6 mg of deionized water (~10 wt%) were loaded in a 3.5 mm sized Pt capsule. All other enstatite-buffered experiments (et, es, ez, and er) used forsterite plus 10 wt% excess SiO2, which was additionally doped with 1 wt% TiO2 or Sc2O3 for et and es experiments, respectively (Table 1). Small amounts of each mix were loaded in 3.5 mm platinum capsules, with 7 wt% of water added with a high precision syringe. These experiments were conducted at 15 kbar and 1400 °C for 24 h. All capsules were sealed by arc welding while wrapped in wet tissue to prevent water loss. The capsules were also checked against water loss by weighing before and after welding. The capsule was placed in a MgO tube, inside a Teflon foil-NaCl-graphite-pyrex assembly. The experiments were conducted in an end-loaded, ½ inch piston-cylinder apparatus. These experiments lasted for 24 h, which allowed the assembly to compact and relieve internal friction. Pressure was calculated from the direct conversion of load to pressure assuming negligible friction, and is thought accurate to ±0.1 GPa. Temperature was controlled using type B thermocouples (Pt94Rh6/Pt70Rh30) and is accurate to ±10 °C. Sample preparation and analytical techniques FTIR. The recovered experimental charges were cut in half. One half was mounted in epoxy and ground down until a representative section was exposed. This section was polished on high precision diamond and alumina plates progressively to the 1 µm grade. The phase relations and grain size were studied by optical microscopy and SEM imaging. The section was then removed from the epoxy mount, the polished side attached to a

glass slide with crystal-bond glue, and the charge polished on the opposite side to make a double polished thin section for FTIR analysis. The thickness of the doubly polished section was measured with a Mitotuyo analog micrometer, which is nominally accurate to within 2 µm (Table 2). In some cases (er, ez, et, and es), the production of double polished thin sections was not feasible due to the loose consistency of the experimental products, therefore, the thicknesses of individual grains was estimated by the method of Matveev and Stachel (2007), which uses the integrated area of the second-order silica overtones between 1625 and 2150 cm–1 in olivine, which varies linearly with thickness. A linear fit forced through the origin gives Atot = 0.6366*t, where Atot is the integrated area of the second-order silica overtones between 1625 and 2150 cm–1 and t is the sample thickness in micrometers. The method estimates the sample thickness with a claimed accuracy of ±15%. A Bruker IFS-28 infrared spectrometer mounted with an A590 Bruker infrared microscope, supplied with a nitrogen-cooled MCT detector, was used for IR analyses (see Berry et al. 2005 for further details). A KBr beam splitter was utilized. Spectra were recorded in the range of 600 to 5000 cm–1. The spectra have a resolution of 2 cm–1. Analyses were made with a circular aperture of 70 µm diameter, while the microscope stage was continuously flushed with nitrogen. Spectra were processed using the OPUS software (Bruker Inc.). Background subtraction was made with the “interactive concave rubber band correction” (ICRC) tool within the OPUS software. Similar numbers of iterations were used (depending on the quality of the spectra) to set the absorbance beyond the region of structurally bound water bands to 0. Applying an unnecessarily large number of iterations may lead to underestimation of the real absorbance, which becomes particularly a problem for water-rich olivine with broad absorption peaks. The “Integration” and “Peak pick” tool of the OPUS software were used to measure the areas and heights of peaks. The total absorbance was estimated from the average unpolarized spectra. The uncertainty in an individual analysis is ~3%; the uncertainty in the estimation of total absorbance changes with the number of grains used for the unpolarized average (Sambridge et al. 2008), but is never higher than 20%. For some of the pbuffered experiments and the Sc-doped e-buffered experiment

Table 2. Summary of FTIR properties of olivine standards FTIR Total integrated absorbance/cm* Sample label Thickness n.o.a. Major absorption bands (cm–1) a.u.a. corr [Si] corr [Ti] [triv] corr (see Table 1) (µm) “Pakistani” ol – polarized 3611, 3594, 3581, 3567, 3480 134 p (15_1100) 135 14 3612, 3590, 3579, 3567, 3555, 3544, 3534, 3478, hydrous 0.02 2421 p (25_1000) 133 11 3613, 3579, 3567, 3553, 3533, 3500, 3478, hydrous 0.56 1.50 4700 7050 p (25_1100) 110 6 3613, 3579, 3566, 3552, 3533, 3478 0.40 1.35 3798 5128 p (25_1200) 115 11 3612, 3579, 3566, 3552, 3533, 3509, 3478, hydrous 0.36 1.25 3578 4472 p (25_1300)‡ 145 6 3612, 3579, 3567, 3554, 3543, 3573, 3477 0.33 1.15 3104 3569 p (25_1400) 110 5 3612, 3579, 3567, 3554, 3543, 3573, 3478 0.34 1.15 1864 2143 e (15_1100) 85 5 3613, 3589, 3568, 3218, 3158,† 3525,† 3503,† 3220 0.02 183 e (35_1100) 105 11 3613, 3590, 3579, 3567, 3554, 3546, 3478, 3218,† hydrous 0.09 585 er (Re4+) 276§ 10 3613, 3579, 3567, 3537, 3476, 3388,† 3160 0.15 174 3 ez (Zr4+) 132§ 7 3613, 3579, 3567, 3551, 3480, 3389,† 3380,† 3355† 0.11 163 8 et (Ti4+) 177§ 8 3612, 3572, 3524, 3483, 3447, 3403, 3350, 3312, 3160 0.16 63 217 50 es (Sc3+) 213§ 7 3613, 3593, 3580, 3535, 3521, 3483, 3450, 3407, 3356, 3321, 3160 0.63 1.80 200 736 1325 Note: n.o.a. = number of unpolarized analysis, a.u.a. = average unpolarized linear absorbance, corr = correction factor (see text for more detail). * The total integrated absorbance is estimated to be accurate around 20%. † Peaks highlighted with daggers are close to the limit of detection. ‡ This sample was not analyzed by SIMS. § Thickness is the average of the analyzed, individual grains and estimated by using the method of Matveev and Stachel (2007) (see text for more detail).

[Mg]

117 5 832 587 228 212

295


a H2O wt% (NRA, FTIR)

0.03 0.025 0.02 0.015

ROM177 (FTIR) ROM250-13 (FTIR) ROM250-2 (FTIR) GRR1012 (NRA) KLV-23 (NRA)

0.01 0.005 0 0

0.02

0.04

0.06 16

0.08

0.1

0.12

0.14

OH/30Si (SIMS)

b 0.03 H2O wt% (manometry, FTIR)

(es), the average unpolarized absorbance was greater than that deemed to provide an acceptable approximation, as detailed by Kovács et al. (2008), and these data were corrected empirically using the ratio of measured to true values as plotted in Appendix Figure 1 of Sambridge et al. (2008). To apply such a correction, the maximum linear polarized absorbance is needed. We calculated the maximum linear absorbance using the observation of Kovács et al. (2008) that the minimum linear polarized absorbance in olivine is almost zero and the maximum is twice the measured average linear unpolarized absorbance (Table 2). SIMS. Samples were analyzed at the Carnegie Institution in Washington by an IMS6f ion microprobe with analytical parameters set out in Hauri et al. (2002) and Koga et al. (2003). The sample preparation for SIMS requires special attention to ensure the lowest possible H background. Therefore, the grains and the remaining parts of the experimental charges were imbedded in liquid indium and after the In solidified the mount was polished. The In has low H-content compared to other widely used organic mounting materials that degas H under high vacuum (i.e., epoxy). The sample was then placed in an oven (usually for a day) at high temperature and low pressure to drive off water contamination prior to application of a 30 nm thick gold coat. The mount was inserted into the sample chamber through air locks a minimum of 12 h prior to analysis. The sample was sputtered by primary Cs+ ions, which were accelerated to 10 kV through an electromagnetic field. The negatively charged target ions were counted by a CAMECA counting system (dead time 44 ns). The counting time was 10 s on 16OH and 2 s on the normalizing isotope 30Si. Each measurement was preceded by a 5 min pre-sputter period to clean the sample from potential surface contamination. The mass resolution enabled us to resolve isobaric overlapping between 16OH and 17O. The sputter crater was normally ~15–20 µm across and ions were extracted through an aperture that further reduced the spatial resolution to the central 10 µm of the crater; the depth of the sputter crater was usually 1–3 µm. The detection limit was usually around 5–10 ppm H2O and largely depended on the amount of background H2O in the system, which depended, in turn, mainly on sample preparation and vacuum quality. The background level of H2O was monitored by analysis of a synthetic forsterite that has less than 1 ppm water (Koga et al. 2003). For the quantification of water, the 16OH/30Si ratio was used. The analyzed 16OH/30Si ratios were always corrected for background, which became especially important at very low concentration levels. Ablation pits were imaged with SEM after the SIMS session to check if inclusions or phases other than the target phase were included in the ablation volume. The 16OH/30Si ratio provided us with a robust analysis that was not influenced by primary beam fluctuations or by ionization efficiency changes from matrix effects (Koga et al. 2003). The SIMS calibration line for olivine (Fig. 1) was determined by using natural olivine KLV23, GRR-1012, ROM250-2, ROM250-13, and ROM177 from Bell et al. (2003, 2004). KLV23 and GRR-1012 were analyzed by NRA in Bell et al. (2003) and have predominantly [Ti] peaks, whereas the other specimens were analyzed solely by FTIR using the Bell et al. (2003) calibration. Figure 1a shows that the calibration factor is about 20% higher when all the standards are included (solid line) with respect to the case when only KLV23 and GRR-1012

0.025 0.02 0.015 India OPX (FTIR) ROM273 (FTIR) A288 (FTIR) KBH-1 (manometry)

0.01 0.005

0

0.02

0.04

0.06 0.08 OH/30Si (SIMS)

0.1

0.12

0.14

16

FIgure 1. A SIMS calibration line for olivine (a) and orthopyroxene (b) based on different natural mineral standards (see text for more detail): the solid line represents the calibration for all the olivine standards, whereas the dashed line is based on the two olivines (KLV-23, GRR1012), where the water content has been determined with a fully independent Figure 1, Kovács et al. (2009) method (NRA, Bell et al. 2003). The calibration lines are least squares fits forced through the origin. The data for the standards are reported in Bell et al. (1995, 2003, 2004) and Koga et al. (2003).

are used (dashed line), for which a truly independent water determination exists. Additionally, the calibration line for hydrous Mg-rich orthopyroxenes (Fig. 1b), where the water content has been obtained by manometry (KBH-1, Bell et al. 1995), is very similar to the calibration line obtained for olivine. This is in agreement with sputtering theory, which argues that the SIMS H2O calibrations for Mg-rich orthopyroxene and olivine should be nearly identical (Koga et al. 2003). A further source for minor errors comes from an imperfect compositional match between samples and standards. The olivine standards are Fo90, whereas the experimental samples are Fo100. However, detailed SIMS investigations on MgO-rich hydrous glasses with and without FeOtot (total Fe as FeO) show that the presence of up to ~10 wt% FeOtot does not introduce a significant error (Kovács 2008). It is important to note that the main aim of this paper is to show the relative difference between the absorption coefficients (k) for the different substitution mechanisms in olivine, and this relative difference is independent of the SIMS calibration used. The calibration factor was 3205 ppm H2O per unit of 16OH/30Si

296


for the analytical session in which all the e-buffered experiments (e, er, ez, et, and es) and one p-buffered experiment at 15 kbar were analyzed, and 3477 ppm H2O per unit of 16OH/30Si (i.e., 8.4% higher) for the analytical session in which the other pbuffered experiments and the natural “Pakistani” olivine were analyzed. The uncertainty in the calibration for water is typically around 10% over a wide range of chemical compositions (Hauri et al. 2002).

other experiment showing a distinct [triv] peak is the Ti-doped sample, where likely small amounts of trivalent Ti are present (Berry et al. 2007b). The “Pakistani” olivine has predominantly [Si] peaks at 3611, 3594, 3581, and 3480 cm–1 and a likely serpentine-related peak at 3700 cm–1 (referred to as [hydrous] in Fig. 2). The presence of a small contribution from [Ti] peaks cannot be excluded, due to their overlap with the [Si] peaks, but the “Pakistani” olivine has only 2 ppm Ti (Kovács et al. 2008).

results Texture

[hydrous]

[Si]

[triv]

[Mg]

The synthesis experiments usually returned 100–400 µm olivine crystals; the ones selected for analyses were transparent, colorless, and macroscopically free of inclusions and lamellae of hydrous phase impurities. The crystals can be easily distinguished from the matrix of quench precipitate, which displays a porcelain-like appearance.

p (15_1100) p (25_1000) p (25_1100)

SIMS and FTIR

p (25_1200) p (25_1400) e (15_1100)

*

*

e (35_1100)

absorbance

The integrated absorbances of the different substitution mechanisms were evaluated separately. Peaks occurring around 3700 cm–1 are due to hydrous impurities (probably serpentine; Post and Borer 2000). These impurities are variably present in some samples at the analytical scale of the FTIR spectroscopy, indicated by their considerable standard deviation, but may be avoided at the much smaller sampling volume of the SIMS analyses. The presence of hydrous phases can be excluded by monitoring the 16OH/30Si ratio during SIMS analysis. Water determined by SIMS is given in Table 3. The uncertainty in the calibration line is ca. 5–10%, and the net uncertainty, including the uncertainty in individual analysis, is taken as ±10%. The natural sample, the “Pakistani” olivine, has 72 ppm water (Table 3). The specimens in the p- and e-buffered experiments have 927–4393 and 91–278 ppm water, respectively. Those in the doped e-buffered experiments (er, ez, et, and es) have water concentrations between 87 and 360 ppm (Table 3), the highest value being for the olivine doped with Sc. The [Si] peaks (at 3613, 3593, 3579, 3567, 3551, 3535, 3478, 3450, and 3405 cm–1) are the only absorption peaks in the periclase-buffered experiments, and are also present in all other samples (Table 2; Fig. 2). The [Mg] peaks at 3220 and 3160 cm–1 are present in all enstatite-buffered experiments (Fig. 2). The [Ti] peaks (3572 and 3525 cm–1) appear only in the Ti-doped enstatitebuffered experiment (et), in which the intensity of the [Si] peak at 3613 cm–1 is negligible relative to those in the periclase-buffered experiments. The [triv] peaks occur between 3300 and 3400 cm–1 and are most intense for the Sc-doped experiment (es). The only

er ez [Ti]

et

es

3600

3400 3200 wavenumber (cm-1)

FIgure 2. Average unpolarized infrared spectra for olivine samples Figure 2, Kovács et al. (2009) (absorbance is given in arbitrary units and the spectra are stacked for clarity). The major absorption bands are indicated by dashed lines. Asterisks mark peaks that are not intrinsic water peaks but represent organic contamination [e (15_1100)] or overlapping orthopyroxene [e (35_1100)]. The label [hydrous] indicates absorption peaks at >3600 cm–1 for water stored in hydrous phases such as serpentine, and not in the olivine structure.

Table 3. SIMS analysis of olivine standards Sample label “Pakistani” p (Table 1) ol (15_1100) n.o.a. 15 8 16 OH/30Si 0.0208 0.5730 standard deviation 0.0017 0.0363 calibration factor (see text) 3477 3205 H2O (ppm wt%) 72 1836 standard deviation 6 116 Note: n.o.a. = number of analysis.

p p p p e (25_1000) (25_1100) (25_1200) (25_1400) (15_1100) 5 6 5 8 5 1.2633 1.2318 1.0381 0.2665 0.0282 0.0965 0.1003 0.0209 0.0186 0.0024 3477 3477 3477 3477 3205 4393 4283 3610 927 91 336 349 73 65 8

e (35_1100) 2 0.0868 0.0064 3205 278 20

er (Re4+) 8 0.0273 0.0012 3205 87 4

ez (Zr4+) 9 0.0302 0.0015 3205 97 5

et es (Ti4+) (Sc3+) 11 3 0.0285 0.1124 0.0014 0.0049 3205 3205 91 360 4 16


dIscussIon Calibration of the different substitution mechanisms in olivine The combined data from FTIR (Table 2) and SIMS (Table 3) shows that the investigated samples contain a variety of defect types, with a large range of water contents that contribute in different proportions to the total absorbance, enabling the different absorption coefficients k[Si], k[Mg], k[Ti], and k[triv] in Equation 2 to be determined by multiple linear regression. The initial regression showed that k[Mg] was not resolvable given experimental uncertainties and was certainly