Water contents of nominally anhydrous orthopyroxenes from oceanic peridotites determined by SIMS and FTIR

This study presents new secondary ion mass spectrometry (SIMS) reference materials (RMs) for measuring water contents in nominally anhydrous orthopyroxenes from upper mantle peridotites. The enstatitic reference orthopyroxenes from spinel peridotite xenoliths have Mg#s between 0.83 and 0.86, Al2O3 ranges between 4.02 and 5.56 wt%, and Cr2O3 ranges between 0.21 and 0.69 wt%. Based on Fourier-transform infrared spectroscopy (FTIR) characterizations, the water contents of the eleven reference orthopyroxenes vary from dry to 249 ± 6 µg/g H2O. Using these reference grains, a set of orthopyroxene samples obtained from variably altered abyssal spinel peridotites from the Atlantic and Arctic Ridges as well as from the Izu-Bonin-Mariana forearc region was analyzed by SIMS and FTIR regarding their incorporation of water. The major element composition of the sample orthopyroxenes is typical of spinel peridotites from the upper mantle, characterized by Mg#s between 0.90 and 0.92, Al2O3 between 1.66 and 5.34 wt%, and Cr2O3 between 0.62 and 0.96 wt%. Water contents as measured by SIMS range from 68 ± 7 to 261 ± 11 µg/g H2O and correlate well with Al2O3 contents (r = 0.80) and Cr#s (r. = -0.89). We also describe in detail an optimized strategy, employing both SIMS and FTIR, for quantifying structural water in highly altered samples such as abyssal peridotite. This approach first analyzes individual oriented grains by polarized FTIR, which provides an overview of alteration. Subsequently, the same grain along with others of the same sample is measured using SIMS, thereby gaining information about homogeneity at the hand sample scale, which is key for understanding the geological history of these rocks.


Introduction
Understanding the role of water, its amount and distribution in the Earth's upper mantle and its influence on convective processes and geodynamics is a key task of igneous geochemistry. Water decreases the mantle's viscosity and mechanical strength via hydrolytic weakening of olivine (Demouchy et al. 2012;Tielke et al. 2017) and lowers peridotite melting temperatures (e.g. Hirth and Kohlstedt 1996), thereby affecting melting depth and thus the composition of the melt (Hauri et al. 2006).
Incorporated by various substitutions, such as protons attached to oxygen anions forming "hydroxyl defects" in the crystal structure (Bell and Rossman 1992;Libowitzky and Beran 2006), small amounts of water can be found in nominally anhydrous minerals (NAMs) such as pyroxenes and olivines derived from the Earth's upper mantle. Olivine is believed to dehydrate rapidly (e.g. Mackwell and Kohlstedt 1990) when decompressed during tectonic uplift or magmatic transport to the surface as recorded by core-rim zonation in natural grains (Demouchy et al. 2006;Peslier and Luhr 2006;Schmädicke et al. 2013). Although H loss has recently been observed in pyroxenes as well (Tian et al. 2016;Denis et al. 2018), orthopyroxene is generally considered to retain its upper mantle water concentration during exhumation Hao et al. 2014;Azevedo-Vannson et al. 2021) and can be used as a reliable tracer for water in the upper mantle (Denis et al. 2013;Warren and Hauri 2014). A number of studies on experimental and natural orthopyroxene have shown that the mineral can hold up to several hundred µg/g H 2 O, making it a significant reservoir for water in the upper mantle (Bell and Rossman 1992;Hirschmann et al. 2005;Peslier 2010).
Several laboratory methods are capable of quantifying hydrogen in NAMs: vacuum manometry is an easily available technique but is also destructive and requires large enough alteration-free sample volumes (Aubaud et al. 2007), which can be problematic in the case of NAMs in upper mantle rocks. Nuclear reaction analysis (NRA) and elastic recoil detection analysis (ERDA) are also reliable techniques for determining hydrogen contents in mantle minerals Withers et al. 2012) but the necessary facilities are often difficult to access (Aubaud et al. 2007). More widespread spectroscopic approaches such as nuclear magnetic resonance (NMR; Rauch and Keppler 2000;Kohn 2006) are difficult to use when iron is present in the matrix (Aubaud et al. 2007). A more detailed review of the different methods is given by Rossman (2006) and recent analytical advances are also discussed in Sheng et al. (2016).
For these reasons, FTIR has long been the method of choice for investigating the hydrogen distribution in NAMs (Rossman 2006;Mosenfelder et al. 2011). In contrast to other techniques, the method itself (after sample preparation) is non-destructive and yields information about the vibrational energy and orientation of OH-dipoles in the crystal structure. It can thus distinguish structurally incorporated hydrogen from water contained in impurities such as secondary hydrous phases, alteration products or fluid inclusions (Rossman 1996;Libowitzky and Beran 2004). Numerous studies have taken advantage of this specificity to document trace amounts of structurally bound water in olivine, ortho-, and clinopyroxene from abyssal peridotites which are often overprinted by alteration (Gose et al. 2009Peslier 2010;Schmädicke et al. 2011Schmädicke et al. , 2018Hesse et al. 2015;Gose and Schmädicke 2021). Generally, there are different approaches when using infrared spectroscopy: (i) analysis of oriented crystals using polarized radiation or (ii) analysis of unoriented crystals using unpolarized radiation (statistical method). Regarding the first method, the orthopyroxene grains in abyssal peridotites are either very small or broken into fragments, which makes it difficult to prepare oriented crystals to be analyzed by polarized radiation. For analyzing randomly oriented crystals using unpolarized radiation one would need to examine a large number of grains in order to get a statistically robust representation of the sample. In contrast to xenoliths, abyssal peridotites are often highly altered, i.e. serpentinized and exhibit a typical porphyroclastic texture, meaning there are often only two or three individual orthopyroxene grains available when preparing sections of hand samples. Even if one were able to find enough individual grains, with this method one would get the average water content for the sample with little or no information about the heterogeneity of the material. In either case, sample preparation for FTIR can be difficult and timeconsuming and strongly depends on sample quality and size. Moreover, there are significant, unresolved calibration issues regarding which molar absorption coefficients are appropriate (e.g. Mosenfelder and Rossman 2013) and there is no agreed procedure as to how the data should be processed (manual versus mathematical correction of absorption baseline), which can introduce major biases between data sets obtained by different working groups.
Because of these challenges, a number of recent studies have chosen SIMS as a means to quantify H 2 O contents in NAMs (e.g. Warren and Hauri 2014;Kumamoto et al. 2017). This method offers the advantage of high spatial resolution, often better than 10 µm, and insensitivity to crystal orientation (Koga et al. 2003). The main drawback of SIMS is that it detects the bulk hydrogen content of a material, meaning it cannot distinguish between different water species. This, however, is important in the case of abyssal peridotites which commonly contain secondary hydrous phases. Moreover, SIMS cannot directly determine elemental concentrations -it can only compare the secondary ion signal from well-characterized reference to that from an "unknown" sample. Crucially, such SIMS RMs must be closely matrixmatched to the actual samples being tested. Thus, for SIMS one needs to first establish a calibration curve, preferably based on multiple, well-characterized RMs. Only once this has been achieved can meaningful analyses of actual samples be undertaken.
The aims of this study are to (i) identify reference orthopyroxene crystals that are well suited for calibrating SIMS, (ii) quantify H 2 O concentrations in natural orthopyroxene samples using this calibration, (iii) compare our SIMS results to the results obtained independently by FTIR, and (iv) evaluate the methods in order to establish a robust methodology for determining intra-crystalline water contents in orthopyroxene recovered from abyssal peridotites (harzburgites), even when such material may have experienced extensive serpentinization.

Orthopyroxene reference materials
Eight orthopyroxenes from spinel peridotites occurring as xenoliths in alkali basalts from Western Mongolia (hand samples Mo21, Mo22, and Mo8531) and three orthopyroxenes from spinel peridotites from the Eifel volcanic field in Western Germany (hand samples E-K1 and 1b-5) were selected for characterization as reference minerals. These peridotites consist of the four-phase assemblage olivine, orthopyroxene, clinopyroxene, and spinel. For further details see Schmädicke et al. (2013) on samples from Western Mongolia and E-K1 and Stosch and Seck (1980) for sample 1b-5. The eleven, up to 3 mm sized orthopyroxene grains that were selected for use as RMs were optically clear and free of fractures.

Orthopyroxene samples from abyssal peridotites
The twelve samples to be analyzed by SIMS include orthopyroxenes recovered from abyssal spinel peridotites (harzburgites) from various locations along the Mid-Atlantic Ridge (MAR), the Gakkel Ridge, Arctic Ocean, and the Izu-Bonin-Mariana forearc region (IBM; Table 1). The samples were recovered by either drilling (DSDP and ODP samples) or dredging and diving expeditions. Where large enough hand samples were available, thin sections were made on which the degree of serpentinization was assessed qualitatively (Fig. 1).
The rocks are moderately to highly altered. The primary mineral assemblage consists of olivine (ol), orthopyroxene (opx), brown spinel (spl), and minor clinopyroxene (cpx). In some samples olivine has been entirely replaced by serpentine, whereas in other samples fresh olivine kernels are preserved in the centers of concentric layers of serpentine ("mesh texture"). Orthopyroxene invariably forms up to mm-sized crystals, heavily broken into fragments, which are partially preserved (porphyroclastic texture). Most of the preserved orthopyroxene fragments contain amphibole lamellae, others are entirely altered or replaced by bastite. Primary clinopyroxene is rare to absent in these samples. Spinel grains formed interstitially, leading to highly irregular shapes. For further information on the petrology of the samples, see Regelous et al. (2016).

Sample preparation
For the FTIR analyses, the peridotite reference orthopyroxenes were cut out of the hand samples using a saw. Xenoliths were slightly crushed and easily disaggregated thanks to their less compact structure. The RM grains were then prepared as cuboid bodies and polished on all six planar sides. The thickness of the cuboids was measured with a precision of ≤3 µm using an electronic micrometer and varied between 516 and 1746 µm.
Since the abyssal peridotite samples are highly altered and the orthopyroxene grains are commonly fragmented, few crystals suitable for FTIR analysis could optically be identified in the hand rock samples and thin sections. We therefore proceeded to crush (instead of sawing) selected rock portions. Because of their greater mechanical strength, the peridotite samples required crushing with severe force. Suitable orthopyroxene grain fragments, optically clear and without visible fractures, inclusions, and alteration products were hand-picked under a binocular microscope, individually embedded in epoxy, grinded down and polished on two sides. For the final polishing step a 3 µm diamond paste was used. This way, numerous orthopyroxene fragments from all samples could be prepared for FTIR analysis. The measurements were performed in three sessions in three randomly oriented but orthogonal planes. Sample thicknesses varied between 80 and 812 µm.
Upon having determined all water contents via FTIR, the RM Mo22-opx3 was dehydrated in a furnace at 1000 °C in air for 24 hours such that the grain could be used as a blank to determine the level of hydrogen background during the subsequent SIMS analyses. Dehydration was verified by FTIR according to the analytical procedure described below.
After completion of the FTIR analyses, pieces of the eleven reference grains were embedded in Epo-Tek® 301  (Epoxy Technology epoxy resin) and prepared as a 1-inch diameter round mount, which was polished and carboncoated for the EPMA (electron probe micro-analyzer) analyses. This mount was also used for the calibration of the hydrogen quantification by SIMS. Prior to the SIMS analyses, the carbon coating was taken off, the mount was washed in distilled water, followed by ultrasonic cleaning in highpurity ethanol followed by argon sputter coating to produce a 35 nm thick, high-purity conductive gold film across the sample surface. It was then loaded into the sample holder and stored in the SIMS' high vacuum airlock for 72 hours.
Since sample outgassing became a problem, the reference grains were removed from the epoxy, cleaned and recast in indium metal using a hand press. Unfortunately, during this process, two of the reference grains (Mo22-opx6 and E-K1-opx3) were lost. The new indium mount was washed, cleaned and gold-coated as described above. For analyzing the hydrogen contents of the abyssal peridotite orthopyroxene samples, these grains together with three RMs (Mo21-opx1, Mo22-opx5, and 1b-5-opx1) were embedded in a second indium mount following the same procedure.
Major elements of the orthopyroxene samples were analyzed on this second indium mount after the SIMS session. Therefore, the gold coating was taken off by polishing the mount by hand with 1 µm diamond paste. Afterwards, the mount was cleaned and sputtered with carbon.

EPMA analysis
All major-element analyses were performed using a JEOL JXA-8200 electron microprobe equipped with five wavelength dispersive spectrometers. The analyses were run at a 15 kV acceleration voltage and a 15 nA probe current. A ~5 µm beam diameter was used for Na, whereas a <1 µm diameter was used for all other elements. Peak and background counting times were 20 s each. The following natural and synthetic materials were used for calibration: Al 2 O 3 (Al), wollastonite (Ca, Si), Cr 2 O 3 (Cr), Fe 2 O 3 (Fe), MgO (Mg), MnSiO 3 (Mn), albite (Na), NiO (Ni), rutile (Ti), and sphalerite (Zn). Kα lines were analyzed for all elements. The data were corrected using an iterative ZAF algorithm. Dispersed between samples a diopside crystal with known composition was analyzed both as a quality control material and for checking instrumental drift. The relative analytical uncertainties are approximately 1 % for elemental concentrations above 2 wt%, whereas relative uncertainties of ~5 % are estimated for element abundances below 2 wt%.
Eight analyses per grain were performed on the reference orthopyroxenes to check for compositional zoning. Element maps including the major elements Ti, Mg, Mn, Fe, Ca, Al, Cr, and Ni were recorded for assessing chemical homogeneity. For the calculation of mineral formulae, all iron was treated as ferrous iron (Fe 2+ ).

FTIR spectroscopy
A Bruker Vertex 70 FTIR spectrometer, equipped with a KBr beam splitter and coupled to a Hyperion 3000 microscope with an MCT detector was used for all FTIR analyses in this study. The measurements were performed in three randomly oriented but orthogonal planes of the prepared crystals with perpendicular polarizer positions parallel to minimum and maximum absorption according to the method of Libowitzky & Rossman (1996). The RMs and samples were analyzed with a spectral resolution of 2 cm -1 across the 7500-550 cm -1 wavenumber range. At least 64 scans were acquired for each spectrum. Rectangular apertures between 80x100 µm 2 and 150x150 µm 2 were used to define the analyzed area in the case of the reference orthopyroxenes. For the unknown samples, apertures between 30x30 and 80x80 µm 2 depending on crystal quality were used. Special care was taken to solely analyze optically clear and inclusion-and alteration-free spots. For internal comparison, three different spots were analyzed per grain, each containing a set of six spectra. A background correction for atmospheric H 2 O and CO 2 was applied following each measurement.
Water concentrations (in µg/g H 2 O) were derived from the integral absorbance normalized to 1 cm thickness applying the mineral-specific molar absorption coefficient from Bell et al. (1995). For baseline correction, a third degree polynomial function was utilized to subtract the background between 3700 and 2800 cm -1 . The analytical uncertainties for spectra with small interferences from amphibole bands in the same wavenumber range as the observed orthopyroxene bands are estimated to amount to 10-15%. Upon request, extensive data can be obtained from the corresponding author.

SIMS calibration
Determinations of H 2 O contents by SIMS were conducted using a Cameca 1280-HR instrument. The first series of analyses of the reference orthopyroxene crystals used a mass filtered 133 Cs + primary ion beam with a total impact energy of 20 keV. 16 O 1 Hions were measured as a proxy for the hydrogen content as this molecule provided more stable within-run data than did 1 H -. Despite the SIMS' highvacuum conditions of 5 × 10 -7 Pa in the sample chamber, we were unable to produce consistent 16 O 1 H -/ 18 Oratios. Even the most hydrogen enriched RM (1b-5-opx1, 249 ± 6 µg/g H 2 O according to FTIR) was affected by hydrogen adsorption onto the sample surface at the point of analysis during data acquisition. Much of this background could be attributed to electron-induced desorption of OH that resulted from the need to use low energy electron flooding when operating in negative secondary ion polarity.
Therefore, an alternative analytical set-up was tested using a duoplasmatron to generate 16 Oas primary ions and recording the 28 Si 1 H + / 29 Si + ion ratio. This alternative approach provided no advantages and was further complicated by an unstable primary ion current that dropped from roughly 2.0 nA to 0.8 nA during the analytical session. This resulted in an increase in the observed 28 Si 1 H + / 29 Si + measured ratio as one would expect when hydrogen was mostly coming from vacuum contamination adsorbed during the analysis. We therefore decided it was necessary to remove all crystals from the epoxy and recast them in indium.
For the next series of analyses, the 16 Oprimary current was increased to 30 nA, the field of view was reduced to only 10 µm by closing the field aperture, and the primary beam raster was turned off. Furthermore, the positioning of external permanent magnets was adjusted such that H could be run as an ion rather than a molecule. A long duration liquid nitrogen cold trap was used during the run to improve the vacuum (1.2 × 10 -7 Pa). Limit of detection under these conditions was around 5 to 10 µg/g H 2 O. This new approach, however, required a spot size of over 30 µm.

Analyses of abyssal peridotite orthopyroxene samples
For analyzing the hydrogen contents of the abyssal peridotite orthopyroxene samples, the SIMS determinations employed a 16 Oprimary ion beam operating with a 30 nA current with a total impact energy of 23 keV in Gaussian mode, providing a circa 25 µm diameter beam at the point of impact on the polished sample surface. Each analysis was preceded by a 90 s pre-sputtering using a 50 × 50 µm 2 raster. Prior to taking data, the raster was turned off. In order to assure that the crater was precisely centered on the field aperture, a centering scan was conducted prior to each analysis using two sets of orthogonal electrical deflection plates. To avoid hydrogen adsorbed from the residual glasses in the analysis chamber (total pressure range from 2.7 to 2.3 × 10 -7 Pa) a small field aperture of 550 × 550 µm 2 was used, equivalent of a field of view of only 6 × 6 µm 2 . Thus, the field of view of the mass spectrometer was limited to the very center of the primary beam's Gaussian distribution. The mass spectrometer was operated in dynamic mono-collection mode at a mass resolution of M/ΔM ≈ 2500 at 10 % peak height. A 150 eV wide energy bandpass was used. A single analysis consisted of 16 cycles of peak switching between 1 H (4 seconds per cycle) and 30 Si (1 s), thus a single analysis required a little over three minutes including the pre-sputtering process. All data were recorded using an ETP133H pulse counter to which deadtime correction of 46.2 ns was applied based on a delay circuit in the pre-amplifier electronics. The 1 H/ 30 Si repeatability on n = 13 determinations spread over the two-day analytical session on our most hydrogen enriched reference material (1b-5-opx1, 249 ± 6 µg/g H 2 O) was ± 3.2 % (1s). The best-fit regression line for the n = 41 analyses conducted in total on the three RMs yielded R 2 = 0.90, so we infer that a linear calibration between [H (µg/g) / SiO 2(wt%) ] true vs. [ 1 H + / 30 Si + ] measured is reasonable. However, this regression line had a negative intercept value, equivalent to -8 µg/g H 2 O. Since the vacuum conditions during this run were comparable to those during the calibration of the mount containing the blank, we conclude that the H 2 O background for these analyses was comparably low, probably in the single digit µg/g range, and therefore decided to use a linear fit forcing the calibration line through zero for the calculation of the H 2 O contents of the samples. We have thus not applied any further background correction, but one must bear in mind that for samples with lower hydrogen contents there may be a few µg/g background which has not been addressed and that such SIMS data might have a small systematic offset.
Depending on crystal orientation, distinctive bands can be found in all samples around 3600, 3570, 3520, and 3420/3410 cm -1 , and less prominent bands around 3300, 3210, and 3060 cm -1 . These bands are ascribed to intrinsic OH in aluminous orthopyroxene (Beran and Zemann 1986;Skogby et al. 1990;Peslier et al. 2002;Stalder 2004). Pleochroism between the different orientations A, B, and C is variably strong since the individual grains are not oriented according to their crystallographic axes. Fig. 2d depicts   Fig. 2 Orthopyroxene FTIR spectra from three different spots (1, 2, 3) on the RMs that were used for the SIMS analyses of the abyssal peridotite orthopyroxene samples: a) Mo21-opx1, b) Mo22-opx5, c) 1b-5-opx1. Three mutually orthogonal grain orientations (A, B, C) collected in horizontal (h) and vertical (v) polarization mode were recorded for each spot. Absorption bands attributed to structural OH are labelled with the corresponding wavenumbers. Wavenumbers in brackets are ascribed to the epoxy introduced during the sample preparation. d) Data from dehydration experiment: one spectrum for each orientation of Mo22-opx3 is shown in black/grey before heat treatment and in red after heat treatment. All spectra are normalized to 1 cm thickness and have been shifted vertically for clarity    infrared spectra recorded both prior to and after dehydration of the grain Mo22-opx3. No hydrogen-bearing species are present in the dehydrated spectra. However, the different slopes and shapes of the spectra, depending on the crystal orientation, emphasize the problem of applying a proper background correction to each spectrum for calculating water contents. Total H 2 O concentrations of the reference orthopyroxenes measured by FTIR vary between 0 (dehydrated grain) and 249 ± 6 µg/g H 2 O (Table 3). Within the range of uncertainty, multiple grains from one sample have the same or similar water contents: 148 ± 5 and 142 ± 4 µg/g H 2 O for Mo21; 84 ± 4, 86 ± 3, 76 ± 5, 89 ± 5, 96 ± 3, and 77 ± 3 µg/g H 2 O for Mo22; and 233 ± 8 and 227 ± 5 for E-K1. The good internal precision as indicated by low standard deviations is also notable.

Mo22-opx5-
As the range of orthopyroxene water contents from abyssal peridotites in former studies varies between 0 and 330 µg/g H 2 O (Peslier et al. 2007;Warren and Hauri 2014;Hesse et al. 2015;Schmädicke et al. 2018) and we expect our samples from equivalent rocks to have similar contents, the selected reference minerals are well suited for use as SIMS calibration materials.

SIMS calibration
The subsequent SIMS determinations were calibrated using the reference orthopyroxenes that had been characterized by FTIR for their H 2 O contents. These results are reported in Table 4. We concluded that a linear relationship exists between H 2 O (wt%, FTIR) vs. ( 1 H + / 30 Si + ) SIMS × SiO 2(wt%, EPMA) . Therefore, the averaged measured 1 H + / 30 Si + ratios are normalized by multiplication with SiO 2 (in wt%, EPMA data) and plotted against the H 2 O concentrations determined by FTIR (Fig. 3). This relationship has a slope of 2.074 × 10 -4 and a coefficient of determination of R 2 = 0.97. The limit of detection is <10 µg/g. In contrast to the FTIR values, where within the range of uncertainty, different orthopyroxene grains from the same sample record the same water content, the SIMS concentrations indicate that the water contents from different orthopyroxene grains of sample Mo22 are heterogenous. However, no differences between core and rim analyses could be detected in the two to five spots measured per grain, such that we assume intra-grain homogeneity.

Water contents measured by SIMS
For eight out of twelve samples H 2 O concentrations could successfully be quantified. The ODP304 orthopyroxenes from the Atlantis Massif and the ODP209 sample from the Mid-Atlantic Ridge are so highly altered that the indicated water contents lie far outside the calibrated range. Grain 1b-5-opx1 was used as the primary RM for the SIMS hydrogen determinations. Based on FTIR analyses this material has a hydrogen content of 249 ± 6 µg/g H 2 O. On 13 determinations on 1b-5-opx1 a mean water content of 261 with a relative standard deviation of ± 8 % was obtained, which we believe is a reasonable estimate for the overall data quality at the higher end of our concentration range. Two other RMs, Mo21-opx1 with a nominal hydrogen content of 148 ± 5 µg/g H 2 O and Mo22-opx5 with a nominal hydrogen content of 96 ± 3 µg/g H 2 O via FTIR, were measured as SIMS quality control materials and to check for instrumental drift. These materials yielded concentration values of 125 ± 14 and 114 ± 6 µg/g H 2 O, respectively (Table 6).
For reasons outlined above a calibration line forced through zero was deployed, plotting the averaged measured 1 H + / 30 Si + ratios normalized by multiplication with SiO 2 (in wt%, EPMA data) against the H 2 O concentrations determined by FTIR of the reference materials contained in the sample mount. This relation has a slope of 6.62 × 10 -4 with a coefficient of determination of R 2 = 0.99 (Fig. 4). This calibration was used for the calculation of the water contents of the abyssal peridotite orthopyroxene samples.

Water contents measured by FTIR
Several orthopyroxene fragments per sample were prepared for the FTIR analyses. The refined method of crushing the highly altered abyssal peridotite samples, however, failed to provide the desired data quality: in addition to secondary amphibole bands around 3690 cm -1 in many of the spectra (Skogby et al. 1990), absorption bands at ~2920 and ~2845 cm -1 interfered with our measurements (Fig. 5). We attribute these bands to traces of the epoxy that was used for the preparation of the less than mm-sized specimens and is possibly trapped in microfractures due to the crushing process (Grant et al. 2007;Tollan and Hermann 2019). We noted that this contamination extends as far as in the 3400 cm -1 region in our spectra, such that the contribution of the epoxy to the spectrum could not be assessed, effectively prohibiting the recovery of any meaningful information about the orthopyroxene H 2 O content. Only for three grains from two samples (DSDP45-395-18R-1W-112 grains opx1 and opx4 and PS55-75-30 grain opx2) water contents could successfully be quantified. For two of these grains the water contents obtained by FTIR are significantly higher than the ones from the same samples obtained by SIMS: 100 ± 5 µg/g H 2 O (SIMS) vs. 102 (+2 %) and 127 µg/g H 2 O (+27 %; FTIR) for the two orthopyroxene fragments from DSDP45-395-18R-1W-112, and 206 ± 8 µg/g H 2 O (SIMS) vs. 244 µg/g H 2 O (+18 %; FTIR) for sample PS55-75-30 (Table 6). Whether there is a true difference between these SIMS and FTIR values or the discrepancy in these specific determinations is partially attributed to the epoxy cannot be evaluated at the present state of knowledge. However, the peridotites from both locations are highly altered, so generally one would expect the SIMS values to be higher if water-bearing alteration products were present in the analyzed sample volumes -as for example in the Atlantis Massif samples. As this is not the case, we conclude that for all other samples structurally incorporated water with little or no component derived from alteration processes was measured by SIMS.
The highest water content in this study is detected in the HLY-102-D32 sample from Gakkel Ridge, which contains 261 ± 11 µg/g H 2 O. This is somewhat surprising since Peslier et al. (2007) measured distinctively lower water concentrations of 25-60 wt. ppm by FTIR for enstatites of PS66-238 samples dredged at the Gakkel Ridge. The orthopyroxene grains from samples PS59-249-Br4 and PS59-236-80-500-900 in this study, also from Gakkel Ridge, contain significantly less water: 68 ± 7 µg/g and 98 ± 9 µg/g, respectively. Nonetheless, the second value is still higher than in the Peslier et al. (2007) study. Warren and Hauri (2014) also examined two orthopyroxenes from  in their SIMS study, which contain 99 ± 10 and 145 ± 18 wt. ppm H 2 O, respectively. These water contents are consistent with our PS59-236-80-500-900 sample. For the orthopyroxene grain from sample PS55-75 from the Arctic Ocean, another very high H 2 O concentration of 206 ± 8 µg/g was determined.
The ODP125-0779A-26R-2W orthopyroxene from the Izu-Bonin-Mariana forearc region in this study contains 181 ± 9 µg/g H 2 O. A recent study by Gose and Schmädicke (2021) also investigated ODP Leg 125 and measured comparable orthopyroxene water contents between 122 and 363 In contrast to experimental data on doped crystals where correlations of water contents with major elements are evident (Rauch and Keppler 2002;Stalder and Skogby 2002;Stalder 2004;Stalder et al. 2005), correlations in natural samples are less obvious due to their greater chemical complexity. However, several studies on natural upper mantle orthopyroxene have found well-defined correlations between the mineral's water content and its Al 2 O 3 , FeO, and/or MgO contents (Peslier and Bizimis 2015;Kilgore et al. 2018;Schmädicke et al. 2018;Schaffer et al. 2019;Ashley et al. 2020). The present set of samples also indicates that major element composition and SIMS determined water contents are related. Our abyssal peridotite orthopyroxene samples from diverse locations cover a broad range of Mg# and Cr# (Fig. 6a). In particular Cr# (Fig. 6c) and Al 2 O 3 (Fig. 6d) correlate well with the H 2 O contents measured by SIMS (r = -0.89 and 0.80, respectively). Correlation coefficients with TiO 2 (Fig. 6b), FeO (Fig. 6f) and MgO are 0.59, 0.56, and -0.55, respectively. Other major elements do not seem to be related to the amount of structural water.

SIMS versus FTIR
Once a calibration and analytical routine have been established, SIMS offers the advantages of being a rapid method for quantifying total hydrogen contents with the potential of analyzing targeted samples in the low nanogram mass range. In the case of abyssal peridotite, SIMS also has the major advantage of much simpler sample preparation requirements as compared to FTIR, because the latter requires the preparation of crystals in three orientations. On the other hand, a significant challenge of the SIMS technology is the need for well-characterized, matrix-matched reference materials, which ideally include a blank material. Another aspect is the need to assure exceptionally good vacuum quality in the analysis chamber, as the residual H 2 O and H 2 in the chamber can be a significant source of background due to molecules adsorbed at the analysis location during the course of data acquisition. For our samples with water contents of (several) hundred µg/g H 2 O a possible hydrogen background of a few µg/g H 2 O was not a major concern. Koga et al. (2003) report hydrogen counts of tens of ions per second, corresponding to a blank of 2-4 µg/g H 2 O, which is similar to what can be achieved with FTIR. An FTIR detection limit below ~1 µg/g H 2 O is very challenging because of issues related to the optical quality of the sample, possible biases introduced by the various calibration strategies, and the method of baseline correction (Mosenfelder and Rossman 2013;Kumamoto et al. 2017). The main disadvantage of FTIR is that depending on the method (polarized measurements in three orthogonal orientations vs. non-polarized/statistical The arrows indicate bands attributed to amphibole (3690 cm -1 ) and residual traces of epoxy (2920 and 2845 cm -1 ). The spectra are normalized to 1 cm thickness and shifted vertically for clarity. Depending on crystal quality, apertures smaller than 50x50 µm 2 were used which resulted in less smooth spectra Black symbols are assigned to the three RMs used for the SIMS calibration of our sample analyses. All other RMs are shown in grey. The error bars correspond to 1 standard deviation measurements in only one orientation) sample preparation can easily become very time-intensive or even impossible, particularly for small or poor-quality samples such as abyssal peridotite, which usually contains only few suitable grains on thin section scale.
As spatial resolution is higher, SIMS is a reliable technique for detecting compositional zoning within a crystal.
Depending on the quality of the crystal, the volume analyzed by FTIR can easily become very large (150x150x1660 µm 3 as in the case of the largest RM in this study, but with typical volumes of 50x50x250 µm 3 for abyssal peridotite orthopyroxene grains), and one only gets an averaged value for the water content integrated over this volume. The sample volume of SIMS is smaller, often less than 10x10x1 µm 3 , such that zonation due to diffusional water loss/gain or compositional variation resulting from differing crystal growth events can effectively be identified and quantified.
While SIMS records total water contents, FTIR provides the significant advantage of also yielding information about hydrogen speciation, i.e. the type of substitution mechanisms, as well as possible alteration products, hydrous fluid inclusions, and secondary phases. This additional information can be essential for understanding the geological history of a sample, including such issues as thermobarometric conditions at the time of hydrogen incorporation, the assessment of redox conditions, and bulk dewatering trends brought about by high grade regional metamorphism.

Conclusion
This study has demonstrated that when investigating pyroxenes from highly altered oceanic peridotite samples, SIMS offers significant advantages for quantifying intracrystalline H 2 O contents. The SIMS sampling volume is much smaller than that offered by FTIR, allowing cracks and zones of alteration to be avoided. When SIMS analysis targets are selected carefully using optical and/or SEM imaging, water contents un-affected by alteration or inclusions can be obtained by SIMS. We observed that the water contents calculated from the SIMS analyses, although about 20% lower than the corresponding FTIR values for two samples, are similar within the range of uncertainty of the two methods, which indicates that the SIMS results do correspond to crystal bound, structural OH with little or no component derived from younger alteration processes. If this were not the case, water contents from differing localities should expectedly be much higher and scatter over a wide range of values, which is not seen in the present SIMS data.
As FTIR has the important strength of being able to distinguish between the different hydrogen species, which is needed in order to get a better understanding of the evolution of the water contents of such rocks, we suggest a combination of both methods, especially for "difficult" samples such as abyssal peridotites. We conclude that the optimal strategy combines both methods by first analyzing an individual grain by FTIR to get an overview of alteration and water content and then measure the same grain at multiple, carefully targeted locations using SIMS from which the homogeneity of the H 2 O distribution can be assessed. This two-step approach provides the possibility of gaining a complete picture of the evolution of such complex materials such that implications about the geological history of the rocks can be made.
Equally important, an additional benefit of this project is that it has resulted in a set of natural reference orthopyroxenes which have been carefully cross-calibrated between FTIR (used to define the absolute hydrogen abundances) and SIMS (used to confirm micron-scale homogeneity of the hydrogen distribution). These eleven reference minerals yielding reproducible data and a SIMS limit of detection of <10 µg/g for H 2 O in orthopyroxene are described in detail in this manuscript.
The water contents of orthopyroxenes from a suite of highly serpentinized abyssal peridotites from the Atlantic and Arctic Ridges as well as from the Izu-Bonin-Mariana forearc region, being difficult to impossible to analyze by FTIR, were measured using these reference orthopyroxenes according to the outlined strategy. The eight samples determined via SIMS gave water contents ranging between 68 ± 7 and 261 ± 11 µg/g H 2 O, which is consistent with literature data for abyssal peridotites (see Gose and Schmädicke 2021; and references therein). All samples and reference orthopyroxenes have homogenous water concentrations at grain scale, such that no signs of water gain or loss could be detected for either tectonically exhumed peridotites or mantle xenoliths. provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.