Nanoscale Secondary Ion Mass Spectrometry determination of the water content of staurolite

Rationale Staurolite is an important mineral that can reveal much about metamorphic processes. For instance, it dominates the Fe–Mg exchange reactions in amphibolite‐facies rocks between about 550 and 700°C, and can be also found at suprasolidus conditions. Staurolite contains a variable amount of OH in its structure, whose determination is a key petrological parameter. However, staurolite is often compositionally zoned, fine‐grained, and may contain abundant inclusions. This makes conventional water analysis (e.g., Fourier transform infrared (FTIR) spectroscopy or by chemical titration) unsuitable. With its high sensitivity at high spatial resolution, Nanoscale Secondary Ion Mass Spectrometry (NanoSIMS) is potentially a valuable tool for determining water contents in staurolite. However a calibration with relevant standards covering a large range of water content is required to obtain accurate and reliable analyses, because matrix effects typically prevent direct quantification of water content by SIMS techniques. Methods In this study, a calibration for NanoSIMS analyses of water content by using minerals with crystallographic structures comparable to that of staurolite (i.e., amphibole and kyanite, an inosilicate and a nesosilicate, respectively) has been developed. Results Water measurements in an inclusion‐free crystal from Pizzo Forno, Ticino, Switzerland, by FTIR spectroscopy (1.56 ± 0.14 wt% H2O) and by Elastic Recoil Detection Analysis (ERDA) (1.58 ± 0.15 wt% H2O) are consistent with NanoSIMS results (1.56 ± 0.04 wt% H2O). Conclusions This implies that our approach can accurately account for NanoSIMS matrix effects in the case of staurolite. With this calibration, it is now possible to investigate variations in water content at the microscale in metamorphic minerals exhibiting high spatial variability and/or very small size (few micrometers).


| INTRODUCTION
In metamorphic rocks water can be stored as the hydroxyl group (water that is structurally bound) in hydrous minerals like amphibole and talc in addition to nominally anhydrous minerals (NAMs) like pyroxene, garnet and rutile. 1 One of these hydrated minerals is staurolite. Staurolite is a monoclinic nesosilicate with chemical formula (Fe,Mg,Zn,Co) 3-4 (Al,Fe) [17][18] (Si,Al) 8 O 48 H 3-4 . 2 It is an index metamorphic mineral common in metapelites equilibrated in the lower amphibolite facies 3 of Barrovian-type metamorphism, where it is often associated with garnet and Al 2 SiO 5 polymorphs. 4 More rarely, it occurs in metapelites in the eclogite-facies. 5 Mg-rich staurolite has been observed in high-pressure metabasites, 6 whereas Fe-rich staurolite has been synthesized experimentally at suprasolidus conditions in metapelitic bulk compositions. 7,8 The crystal chemical formula of staurolite is not fully known to date, in particular as concerns its hydroxyl content. 2,9 Such variable OH content determines values between 1 and 2 wt% H 2 O in most reported staurolite analyses. Two types of reactions seem to control the water content of staurolite 10 : homogeneous reactions with cation-hydroxyl substitutions and heterogeneous reactions with redox and dehydration equilibria. The latter appear to be favored by an increase in temperature. 9 Staurolite has great significance during metamorphic processes. In common Ms-Qz-bearing metapelites it breaks down to garnet, biotite, and Al 2 SiO 5 , whereas the dehydration of staurolite in Qz-absent protoliths may produce hercynitic spinel. 11,12 Little data exist concerning the water content of staurolite and its implications for metamorphic processes. 10,11 It is therefore necessary to collect more information on staurolite water contents in order to better understand its relevance in fluid control during metamorphic processes.
Nanoscale Secondary Ion Mass Spectrometry (NanoSIMS) is a powerful instrument for the determination of water contents at the micrometer scale in geological samples. 9,[13][14][15][16][17][18][19] However, as with other SIMS instruments, several biases (e.g., instrumental fractionation), including the so-called matrix effect, 20,21 need to be considered in order to obtain accurate quantitative results. During analysis, a primary ion beam of O À or Cs + sputters the surface of the sample. Secondary ions are emitted from the sample surface as a secondary ion beam, which is subsequently analyzed by a double focusing mass spectrometer. In order to calibrate measurements and obtain accurate results, the analysis of reference materials is required. 20,21 For a proper calibration, it is necessary to measure standards exhibiting the same or a similar matrix. This implies choosing standards sharing similar chemical composition 22 and crystallographic structure with the samples, in order to obtain accurate data.
The set of standards must cover a wide range of water concentrations and must be minerals of gem quality, i.e., homogeneous and large enough to carry out multiple and independent analyses. However, there are few standards of staurolite that can satisfy these criteria, as they can be zoned or rich of inclusions 23,24 ( Figure 1A). Hence, minerals with similar composition or crystallographic structure to those of staurolite can be used to define the calibration curve and thus to correct data for the matrix effect. This effect depends on the secondary ionization probability of a species (e.g., H) at the sample surface. In other words, it characterizes the emission yield of a given ion within different materials. We have chosen two minerals to test this approach: amphibole and kyanite. Both minerals, in particular kyanite, can be found in metamorphic rocks, and can be associated with staurolite at amphibolite-facies conditions. 25 27 Amphibole has a structure based on a double chain of tetrahedra and octahedra, 27,28 and has an O/OH ratio similar to that of staurolite. Kyanite, a triclinic nesosilicate, has a crystallographic structure very similar to that of the monoclinic staurolite. 25,29 In fact, the staurolite structure can be envisaged as an alternation between a kyanite module (Al 2 SiO 5 ) and one of Fe 2 Al 0.7 O 2 (OH) 2 composition along [010]. 30 In the second module, Fe 2+ is in tetrahedral coordination. Hydrogen is linked to oxygens from octahedra to form OH groups. The strong structural analogy explains the frequent epitaxial intergrowths or replacements between the two minerals. 25,31 F I G U R E 1 BSE images of A, staurolite from the Armorican massif (sample from the mineral collection of the Museum National d'Histoire Naturelle de Paris) and B, the staurolite from Pizzo Forno, Ticino, Switzerland used in this study. The Armorican crystal (A) contains many inclusions, unlike the mineral used in this study (B). FTIR and ERDA analyses of the Armorican crystal are hence more difficult In this study, we present an original approach for correcting NanoSIMS measurements of water content in staurolite using kyanite and amphibole as standards. The accuracy of the corrected NanoSIMS measurements was evaluated by independent measurements of the same staurolite crystal by Fourier-transform infrared (FTIR) spectroscopy and Elastic Recoil Detection Analysis (ERDA). The limitations of the NanoSIMS method and the influence of the crystal structures on the matrix effect are discussed.

| Samples
This study involves a crystal of staurolite from Pizzo Forno, Ticino,

| NanoSIMS measurements
One polished section (001) of the staurolite crystal without resin was stuck on a double-sided copper tape and analyzed with the Cameca NanoSIMS 50 installed at the Muséum National d'Histoire Naturelle in Paris. Three other polished sections of the staurolite crystal were analyzed to characterize the impact of the crystal orientation on the measurements. Each sample was polished to a quarter micrometer with alcohol and cleaned with ethanol in an ultrasonic cleaner. All samples and standards were gold coated (20 nm thick) before NanoSIMS analysis. Their surfaces were rastered by a 16 keV Cs + primary beam, set to 23 pA (probe size around 200 nm). Secondary ions were recorded in multicollection mode: 12 C À , 16 OH À , and 28 Si À , using electron multipliers with a 44 ns dead time. The mass resolving power was set to 8000, sufficient to resolve any interferences on the recorded masses. A flooding electron gun with a current of 8009 V was used for charge compensation. Measurement of 12 C À attested that analyses were not made at the edge of the sample or on a crack or a hole at the sample surface. Presputtering was carried out over a surface area of 5 Â 5 μm 2 for 300 s with a 200 pA primary Cs + beam to remove surface contamination, gold coating, and to reach a steadystate sputtering regime. 36 Analyses were made on a 3 Â 3 μm 2 surface area during 100 cycles of 1.24 s each for a total measurement time per point of 431 s. Counts were collected only from the inner 1 Â 1 μm 2 using the beam blanking mode to reduce contamination from the edge of the area of interest. 17 During the session, the vacuum never exceeded 3 Â 10 À10 Torr in the analysis chamber.

| FTIR spectroscopy measurements
Fourier-transform infrared (FTIR) spectroscopy is a non-destructive method, which has a low detection limit 37 (<1 ppm H 2 O). It is straightforward to identify traces of epoxy, for example, and to determine the speciation of water. The main drawback of this technique is the demanding sample preparation needed to obtain a doubly polished thin section for analysis in transmission mode.
Polishing defects may affect the IR signal and the thickness of the doubly polished thin section determination uncertainties, and thus the error of the result. To perform analyses and evaluate the total integrated absorbance, the mineral was prepared along each crystallographic axis. For the staurolite there is absorbance only in E//a and E//c. 33 Hence, only the (010) plane was studied. All analyses were carried out at the spectroscopy platform of the Institut de which is equal to 83,000 ± 5,000 LÁmol H2O À1 cm À2 according to Koch-Müller and Langer; 33 and t is the section thickness (cm). The total relative uncertainty is 10%. All parameters are summarized in  for all measurements (Figure 3). We thus conclude that, in the case of staurolite, NanoSIMS is insensitive to orientation effects for water concentration measurements. Hence, it will provide accurate results on any crystallographic orientation of the sample.
Data collected on the three amphiboles and the kyanite standards determine a consistent calibration curve (Figure 2A). The H 2 O concentration calculated from NanoSIMS measurements of OH À /Si À ratios is 1.56 (± 0.04) wt% (  Figure 2; Figure S1, supporting information; and Table 2).
Polarized spectra of the staurolite section (010) determined by the FTIR method are similar to those previously reported in the literature. 8,33,[38][39][40][41] They show typical bands at 3345 cm À1 , 3460 cm À1 , 3580 cm À1 , and 3680 cm À1 ( Figure S2, supporting information), which correspond to three crystallographically different OH groups with diverse proton positions (H1; H2; H3). 41,42 Each spectrum depends on crystallographic direction. To determine water contents in staurolite by FTIR spectroscopy, the analysis has to be made on the (010) section (perpendicular to the b-axis). Along this plane, the two perpendicular crystallographic axes a and c were investigated. The water concentration was derived from the total absorbance and the normalization was made from the integrated absorbance. The H 2 O concentration recalculated from FTIR spectroscopy was found to be 1.56 (± 0.14) wt% H 2 O (Figure 4; Table 2).
Hence, the determination of water content by NanoSIMS appears consistent with FTIR spectroscopy and ERDA. Furthermore, NanoSIMS has a better precision and measurements are made on a smaller sample volume.

| DISCUSSION
We have shown here that NanoSIMS, ERDA and FTIR spectroscopy   (Figure 2A). For instance, Bellatreccia et al 43 used minerals as standards and not glasses, and applied this method (i.e., OH À /Si À vs H 2 O/SiO 2 ) to define a calibration curve. Like Thomen et al, 36 we can define a simple model of secondary ion currents where: Si where I p corresponds to the current density of the primary beam, Y is the total sputtering yield, [OH] and [Si] are the surface densities of corresponding atoms. α represents the ionization probability of OH À and Si À . Finally, T OH À and T Si À are transmission factors for OH À and Si À , respectively. With Equations (1) and (2), the OH À /Si À ratio can be expressed as: T depends on the NanoSIMS optics and the setting for the analytical session, while α is characteristic of the sample. The Then, we obtain: where β corresponds to the slope value of the calibration curve and depends only on the instrumental parameters and the emissivity of ions, which in turn are constant from one sample to another during the same session. If all standards are aligned on the calibration curve, then the matrix effect can be considered as corrected. Hence, with a simple model of secondary ion current, it is possible to show the link between the measured ion ratios and the true elemental ratios (i.e., OH À

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1002/rcm.9331.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are openly available in "NanoSIMS determination of the water content of staurolite" at https://doi.org/10.48579/PRO/SNGYCN.