Water is a key parameter in mantle rheology, magma genesis, magma evolution, and resulting eruption styles, because it controls the density and the viscosity, as well as the melting and crystallization behavior of a melt. The water content in nominally anhydrous minerals (NAMs) such as clinopyroxene recently has been used as a proxy for magmatic water contents. NAMs, however, may dehydrate during magma degassing and eruption. We performed rehydration experiments on potentially degassed clinopyroxene phenocrysts from various volcanic settings. The experiments were conducted in hydrogen gas at 1 atm or hydrothermal pressures ranging from 0.5 to 3 kbar to test the incorporation of water into natural clinopyroxene under water fugacities similar to those in a volcanic system. Our results show a dependence of the water content in the clinopyroxene crystals with pressure as the phenocrysts begin to dehydrate upon lower water fugacities in the experiments. Water loss or gain in a crystal occurs according to the relatively fast redox-reaction OH− + Fe2+ ↔ O2− + Fe3+ + ½ H2, which was confirmed by Mössbauer spectroscopy. The kinetics of this redox-process are independent of pressure and thus water fugacity. Water contents in rehydrated clinopyroxene crystals can be related to magmatic water contents at various levels in a volcanic system. Our results thus show that the water content in erupted clinopyroxene phenocrysts cannot be taken for granted to be representative of magmatic water contents prior to magma degassing. The conducted experiments indicate the simultaneous dehydration of clinopyroxene along with magma ascent and degassing. Rehydration experiments under hydrothermal pressures, however, may be able to reconstruct clinopyroxene water contents at crystallization prior to dehydration.
The majority of clinopyroxene crystals in this study come from the Canary Islands. More precisely, rock samples included are El Hierro island ankaramite lava vessels from Tanganasoga volcano and the Lomo Negro ankaramite lava flow (Carracedo et al. 2001; Longpré et al. 2009a; Villasante-Marcos and Pavón-Carrasco 2014). In addition, samples from basanite lava flows and one kaersutite cumulate xenolith from the 1971 Teneguia (Brändle et al. 1974) and 1949 Hoyo Negro (Klügel et al. 2000) eruptions on La Palma were included. To get broad comparisons regarding geological environments and chemical composition, additional clinopyroxene from the 2006 andesite lava from Merapi volcano (Gertisser 2001; Ratdomopurbo et al. 2013) on Java and basanite lavas from Shuangcai volcano in Zejiang province, China (Liu et al., unpublished manuscript) were studied. Two further samples consist of a mantle xenolith from Kilbourne Hole, New Mexico (Kil and Wendlandt 2004; Harvey et al. 2012) and a specimen from the well-studied martian Nakhla meteorite (Bunch and Reid 1975; Treiman 1993). A full description of the samples used in this study is given in Table 1.
Rock samples (n = 13) were crushed to obtain loose clinopyroxene crystals of a size suitable for analysis (≥300 μm). These were hand-picked under a binocular microscope and individual clinopyroxene crystals were then mounted in thermoplastic resin for further processing. With the help of crystal morphology and optical microscopy (extinction angles), the selected crystals (n = 37) were oriented along their crystallographic c-axis and their (100) and (010) crystal faces, on which the directions of the main refractive indices (α, β, and γ) occur. A detailed procedure of the crystal alignment is described in Stalder and Ludwig (2007). Various particle size-grades of Al2O3-grinding paper were used to thin and polish the oriented crystals to a thickness of a few hundred micrometers.
Electron probe microanalysis
Analyses of major elements (Al, Ti, Fe, Mg, Na, K, Si, Ca, Mn, Cr) in clinopyroxene crystals were carried out at the Department of Earth Sciences, Uppsala University, using a Field Emission-EPMA JXA-8530F JEOL hyperprobe. Between 5 and 8 spots were analyzed on each crystal using a beam current of 10 nA with an acceleration voltage of 15 kV with 10 s on peak and 5 s on lower and upper background. Standards used were fayalite (Fe2SiO4) for Fe, periclase (MgO) for Mg, pyrophanite (MnTiO3) for Mn and Ti, corundum (Al2O3) for Al, wollastonite (CaSiO3) for Ca and Si, eskolaite (Cr2O3) for Cr, nickel oxide (NiO) for Ni as well as albite (NaAlSi3O8), orthoclase (KAlSi3O8), and apatite Ca5(PO4)3(OH,F,Cl) for Na, K, and P, respectively. Several crystals showed weak zonation in backscattered electron (BSE) images and for each crystal an average composition was calculated from all analyzed spots. A detailed description of the EPMA procedure and the analytical uncertainties is presented in Barker et al. (2015). From the obtained weight percentages, the number of atoms per formula unit in each crystal was calculated on the basis of a four cation-normalization. To distinguish between and quantify Fe2+ and Fe3+ in clinopyroxenes, Mössbauer spectroscopy was applied (see below). Average compositional data for clinopyroxene from samples LP1971-1, LP1949, and EH-Ank were taken from Weis et al. (2015).
All rehydration experiments were performed at a temperature of 700 °C, at which hydrogen diffusion coupled to redox-reaction 1 has been shown to be active (Skogby and Rossman 1989; Skogby 1994; Stalder and Skogby 2003; Bromiley et al. 2004; Koch-Müller et al. 2007; Sundvall et al. 2009; Sundvall and Skogby 2011). Clinopyroxene crystals in this study show XFe/(Fe+Mg) > 0.08 (Table 2) and are thus above the threshold value suggested by Hercule and Ingrin (1999) (XFe/(Fe+Mg > 0.07). Therefore, the hydration kinetics for Equation 1 are expected to be similar to the kinetics for hydrogen self-diffusion (Ingrin and Blanchard 2006). For rehydration it is also relevant how much Fe3+ is present in the crystals. Fe3+ generated by the redox-process discussed above would probably follow similar kinetics. Additional Fe3+ may have been incorporated during crystallization by other charge-balancing mechanisms such as an aegirine component (NaFe3+). This Fe3+, however, seems to be more stable and to take part in the redox-processes only in association with the occurrence of trivalent cations (e.g., Al3+, Fe3+) in the tetrahedral site, which are associated with hydrogen defects (Skogby and Rossman 1989; Skogby 1994; Purwin et al. 2009). However, a component involving (NaHFe2+) otherwise seems less likely. At 700 °C the reaction kinetics for hydrogen self-diffusion are as high as −logD = 11.3 m2/s and hence at least 4.5 orders of magnitude higher than for vacancy diffusion (−logD ≤ 16 m2/s) (Hercule and Ingrin 1999; Ingrin and Skogby 2000; Ingrin and Blanchard 2006; Cherniak and Dimanov 2010). Therefore, for the given temperature and time intervals, a relatively fast hydrogen diffusion is expected, whereas defects other than those generated by the redox-process are not expected to be significantly annealed. This is different to previous high-pressure and high-temperature experiments where the hydrogen solubility in clinopyroxene has been tested (e.g., Bromiley et al. 2004), where all structural defects are annealed. It is important to note that the experimental conditions did not intend to mimic conditions in a natural magmatic system, but simply to reverse redox-reaction 1.
The rehydration experiments were not buffered to control the oxygen fugacity. However, due to the hedenbergite component in the clinopyroxene crystals (see Table 2), we assume that the oxygen fugacity is controlled by the hedenbergite-magnetite-quartz buffer, which imposes redox conditions similar to QFM (cf. Gustafson 1974; Xirouchakis and Lindsley 1998). Water fugacities for pressure experiments behave in a linear relation to and do not deviate strongly from the nominal pressure (cf. Pitzer and Sterner 1994).
Rehydration at 1 atm
To rehydrate the defects in clinopyroxene that lost hydrogen during oxidation (see Eq. 1), the crystals were heated under a stream of H2 gas for different time intervals (see Table 3 for details) at a temperature of 700 °C and an ambient pressure of 1 atm. The crystals were kept in a gold sample holder and placed into the middle of a horizontal glass-tube furnace after the target temperature had been reached. The temperature was measured with a Pt100–Pt90Rh10 thermocouple placed directly above the samples, which has an estimated uncertainty of ±2 °C. Prior to the start of the hydrogen flow, the glass tube was flushed with CO2 so that reaction between the minerals and ambient oxygen as well as an explosion-risk of the hydrogen gas was prevented. Clinopyroxenes were then heated. After thermal annealing, the glass tube was again flushed with CO2 and the crystals were removed from the furnace. Heating and cooling times of the samples for this method are on the order of ~1 min.
Rehydration under pressure
For pressure experiments (Table 3), selected crystals were welded into Au-capsules with an outer (inner) diameter of 5.0 (4.6) mm together with 12 μL of water. To prevent any dissolution of the clinopyroxene during the experiments, about 2 mg of diopside powder (CaMgSi2O6) were added to the capsule. Au-capsules were sealed using a Lampert PUK U3 welding device (equipped with tungsten electrode, flushed with argon gas). Possible leaks causing water loss were identified by weighing the capsules before and after heating in an oven at 120 °C for 15 min. Pressure treatment was performed in Rene 41 steel-vessels in cold-seal pressure vessels at Innsbruck University using water as pressure medium. All experiments were performed at a temperature of 700 °C and pressures between 0.5 and 3 kbar (Table 3). The temperature was measured by Ni-CrNi thermocouples and pressures were measured with a Heise gauge and kept constant within 0.05 kbar during the whole run duration.
Before and after thermal and/or hydrothermal treatment, polarized FTIR spectra in the range 2000–5000 cm−1 were acquired on the oriented clinopyroxene crystals along the directions of the main refractive indices (α, β, and γ) to obtain the total absorbance: Aα + Aβ + Aγ = Atotal. Aα and Aγ were measured on the (010) crystal face and Aβ on (100). The polished crystals were measured in the sample compartment of a Bruker Vertex 70 spectrometer equipped with a NIR source (halogen lamp), a CaF2 beamsplitter, a wiregrid polarizer (KRS-5), and an InSb detector. Another set of crystals was measured at Innsbruck University using a Bruker Hyperion 3000 microscope equipped with a Globar source, a KBr beamsplitter, and a MCT detector. Crystal thickness varied between 150 and 800 μm, with most crystals having a thickness between 200 and 400 μm for both the (100) and (010) orientations. Cracks and inclusions in the crystals were avoided by applying small apertures (100 to 400 μm) for masking during analysis. In some cases impurities (e.g., magnetite inclusions) were present in the beam path, but these appeared not to have had any significant effect on the OH range of the spectra. For each individual spectrum, 128 scans were performed and averaged. No significant differences between the analyses in the sample compartment and microscope measurements were observed. The obtained spectra were baseline corrected by a polynomial function and the individual OH bands were fitted with the software PeakFit and used for further calculations. The corresponding water contents were then calculated using both the wavenumber-dependent calibration function established by Libowitzky and Rossman (1997) and the mineral-specific (augite) calibration of Bell et al. (1995).
The oxidation states of iron in clinopyroxenes before and, where possible, after the rehydration experiments were obtained by Mössbauer spectroscopy. To obtain a sample average for different iron valence states, powdered crystal separates were analyzed with a 57Co standard source (active diameter 5.0 mm), while selected single crystals were analyzed using a point source (active diameter 0.5 mm) to investigate oxidation and reduction during rehydration experiments. Several powdered crystals (10 mg in total) from individual rock samples were mixed and ground with acrylic resin and pressed to a thin disk under mild heat (150 °C) for analysis with the standard source. Untreated, oriented single crystals for the point source were cut in half and one piece was used for rehydration experiments first. After the annealing and FTIR analysis the individual single-crystal pieces were powdered, mixed, and ground with thermoplastic resin and formed into a ~1 mm3 cylinder that was mounted on a strip of tape for analysis with the point source. The Mössbauer measurements were performed at incident angles of 90° and 54.7° to the γ-rays for the point and the standard source, respectively. All obtained spectra were calibrated against an α-Fe foil, folded and reduced from 1024 to 512 channels. The spectra fitting was done with the Mössbauer spectral analysis software MossA (cf. Prescher et al. 2012). During the fitting process, one doublet each was assigned to Fe2+ and Fe3+ in the octahedral positions. From the area of the doublets, the percentage of each oxidation state relative to the total iron content of the sample was obtained, assuming similar recoil-free fractions for Fe2+ and Fe3+. The estimated analytical error for the obtained Fem+/Fetot ratios is ±1%. No crystal-specific compositional analysis was carried out on the Mössbauer sample set due to the destructive nature of sample preparation. Mössbauer results for La Palma and El Hierro lava samples were taken from Weis et al. (2015). Data for the Nakhla meteorite were taken from Dyar (2003) due to the limited amount of sample material available.
Electron probe microanalysis
The clinopyroxene chemical data obtained by EPMA are shown in Table 2. The chemical composition of Tanganasoga clinopyroxenes was measured on a sample set separate to the one taken for rehydration experiments and is represented as average values. The full data set is available in Appendix1 A1. The analyzed clinopyroxenes from the Canary Islands (n = 22) are all titanium-rich diopsides (2.5 to 3.3 wt% TiO2) (Table 2, Fig. 1) with Mg# between 74 and 79 (mean = 78). With only very few exceptions, the individual crystals are homogeneous in composition and show limited zonation. Differences in chemical composition between crystals of individual rock samples occur on occasion. Clinopyroxenes from Merapi, Shuangcai, and Kilbourne Hole (n = 8) are also diopsides with Mg# from 76 to 82. The crystal from the Nakhla meteorite is an augite (Mg# = 62). The structural formulas for clinopyroxene phenocrysts from this study are reported in Table 2.
FTIR before rehydration
All analyzed clinopyroxenes showed absorption maxima at 3630, 3530, and 3460 cm−1 in the IR spectra (Fig. 2), which corresponds to the typical vibrational bands expected for OH in diopside (e.g., Skogby 2006). The OH band at around 3630 cm−1 is prominent for E‖α and E‖β, while the two bands around 3530 and 3460 cm−1 dominate for E‖γ. This infrared-pleochroic behavior is typical for clinopyroxene OH bands (Fig. 2) (e.g., Beran 1976) and thus excludes the influence of possible OH-bearing impurities (e.g., hydrous minerals, melt/fluid inclusions).
Infrared analyses before rehydration experiments revealed a significant spread in the water content of the crystals from different rock samples (Table 3, Fig. 3). In some cases, water contents of clinopyroxene within individual rock samples varied significantly despite homogenous crystal chemistry. Water contents (given in parts per million weight H2O) for clinopyroxene from Tanganasoga samples ranged from 141 to 790 ppm. Clinopyroxenes from the Lomo Negro and La Palma lava flows showed values between 36 and 364 ppm. Crystals from Merapi volcano were completely dry with water contents below detection limit, while those from Shuangcai volcano varied between 84 and 292 ppm. The crystals from mantle xenoliths from Kilbourne Hole showed values from 357 to 379 ppm. The martian clinopyroxene from the Nakhla meteorite was also completely dry with a water content below detection limit. Water contents corresponding to the spectra differed for the calibrations of Bell et al. (1995) and Libowitzky and Rossman (1997). Values determined with the mineral-specific calibration by Bell et al. (1995) were about 25% higher, but the authors note that their mineral-specific (augite) calibration is primarily valid for samples with similar OH-spectra. Our diopside spectra mostly do not show a strong band at ~3460 cm−1 for E‖α and E‖β and thus differ from the spectra presented in Bell et al. (1995). We therefore used the values derived through the calibration by Libowitzky and Rossman (1997), which previously has been used successfully for synthetic as well as natural clinopyroxene samples (e.g., Stalder 2004; Stalder and Ludwig 2007; Sundvall and Stalder 2011) and that was reconfirmed by Mosenfelder and Rossman (2013).
Potential uncertainties for calculated water contents can arise from baseline correction and measurements of the crystal thickness. However, due to the quality of the spectra and the relatively large thickness of the crystals, a maximum error of ±10% is assumed for the precision of the calculated clinopyroxene water contents. In addition, the uncertainty regarding the accuracy of the values due to the calibration for absorption coefficients is another ±10% (cf. Libowitzky and Rossman 1997) resulting in an overall uncertainty of ±20% for the calculated clinopyroxene water contents. No hydrogen diffusion profiles were observed in the clinopyroxene crystals.
FTIR after rehydration at 1 atm
The thermal treatment caused no fractures or other visible effects on the crystals. The absorption bands increased significantly in height in all three directions (α, β, and γ) after thermal annealing in some samples while they decreased in others. The strongest increase as well as decrease in absorbance was notably always observed for the band at 3630 cm−1. No change in band positions was observed, however. Treatment in H2-atmosphere at 1 atm led to different hydration trends in clinopyroxene crystals from Tanganasoga. Water-poor crystals from sample T4 increased their hydrogen content to values between 226 and 307 ppm, whereas crystals with higher water contents from samples T1, T2, T3 (569 to 790 ppm) showed a decrease to hydrogen contents between 394 and 565 ppm. Clinopyroxene from Lomo Negro and the kaersutite cumulate xenolith from La Palma reached water contents between 241 and 434 ppm. The dry Merapi crystals increased to values between 151 and 177 ppm. Clinopyroxene from Shuangcai also showed a decrease as well as an increase in water contents with values between 142 and 183 ppm. One annealed crystal from Kilbourne Hole showed no significant change in water content. The Nakhla clinopyroxene increased to a water content of 16 ppm. Results for rehydration at 1 atm are presented in Table 3.
FTIR after rehydration under pressure
During the pressure experiments some crystals fractured along the cleavage planes and surfaces had to be re-polished. However, no entire crystal had to be discarded. Results for rehydration under pressure are presented in Table 4. Hydrothermal pressure annealing resulted exclusively in an increase in absorbance with the same pattern as for thermal annealing at 1 atm. All crystals seem to be close to saturation after 100–150 h of annealing, and no important changes within the analytical error of the water content occur or are to be expected. After pressure treatment Tanganasoga crystals reached water contents between 747 and 1129 ppm. Clinopyroxene from La Palma and Lomo Negro lava flows increased to values between 668 and 1180 ppm. Pressure-annealed clinopyroxene crystals from Merapi range between 712 and 794 ppm. Crystals from Shuangcai and Kilbourne Hole revealed water contents between 425 and 704 ppm. The clinopyroxene from Nakhla showed a water content of 134 ppm after pressure annealing.
Hydrogen treatment of pressure-annealed samples
Several pressure-annealed clinopyroxene crystals from Tanganasoga [T3(3) and T4(5)], Lomo Negro (LN-3), and Kilbourne Hole (KBH-Cpx2) were again annealed at 1 atm and then showed a decrease in water content between 20 and 56%.
The results for the Mössbauer spectroscopy of clinopyroxenes are shown in Tables 5 and 6. A representative spectrum with doublets for Fe2+ and Fe3+ is shown in Figure 4a. On average, the analyzed clinopyroxenes in this study show Fe3+/Fetot ratios between 18 and 39%. Both reduction and oxidation of iron are observed upon hydrogen treatment (Table 6). Crystals that lost hydrogen during annealing showed an increase in Fe3+/Fetot while those that incorporated hydrogen showed a reduction of the same. Neither metallic iron, which would indicate excessive reduction, nor oxidation products such as magnetite or hematite were observed after the experiments.
The variation in clinopyroxene water contents between individual rock samples, especially from the same location, correlates with the Mössbauer results. For example, clinopyroxene crystals in sample T4 from Tanganasoga show low water and higher Fe3+ contents than crystals from Tangansoga sample T1 where crystals have much higher water and lower Fe3+ contents. Mössbauer spectroscopy also shows that redox-reaction 1 was active during annealing experiments and Mössbauer data for the studied Tanganasoga crystals demonstrate reduction as well as oxidation of Fe3+ and Fe2+ after thermal annealing (Table 6). The increase and decrease of Fe3+ parallels the change in water content in the crystals (Fig. 5). Crystals that incorporated hydrogen under thermal annealing at 1 atm in hydrogen gas or under pressure annealing showed a reduction of Fe3+ and a decrease in Fe3+/Fetot. The measured changes in Fe3+/Fetot for the two investigated crystals from sample T4 match, within error, the expected changes in Fe3+/Fetot based on the hydrogen incorporation during rehydration (Table 6). For example, crystal T4(7) incorporated 908 ppm of water upon thermal annealing, which corresponds to 0.022 atoms per formula unit (apfu). Considering the 1:1 ratio of redox-reaction 1 this intake would imply a reduction of 0.022 apfu Fe3+ corresponding to a change of 10.9% in Fe3+/Fetot (Fetot = 0.203 apfu). Mössbauer analysis revealed a change in Fe3+/Fetot of 9.2% (corresponding to 0.019 apfu). Those clinopyroxene crystals that showed a loss of hydrogen upon annealing at 1 atm revealed an increase in Fe3+. The expected rise in Fe3+/Fetot, matches again, within analytical error, the measured change in Fe3+/Fetot obtained by Mössbauer spectroscopy. The minor discrepancies in expected vs. measured changes could be a consequence of small compositional differences between the investigated crystals and the sample average determined by EMPA. However, the results confirm that the hydrogen diffusion is linked to redox-reaction 1. Interesting is that hydrogen diffuses out of the clinopyroxene at 1 atm and that the charge-balancing oxidation of Fe2+ occurs despite the strongly reducing hydrogen atmosphere in the glass tube furnace. This implies that the hydrogen diffusion out of clinopyroxene crystals is not only influenced by their environment (oxidizing vs. reducing) but is also pressure related.
De- and rehydration kinetics
Rehydration at different pressures
From the obtained data it is obvious that the rehydration under pressure results in higher water contents in the crystals than for annealing at 1 atm. A positive relation between pressure and clinopyroxene water content can be seen after the rehydration experiments (Fig. 7). The two crystals that have been subsequently annealed at lower pressures (2 and 3 kbar then at 0.5 kbar and finally at 1 atm) showed water loss. Plotting the water contents obtained for annealing experiments against the different pressures (1 atm to 3 kbar) on a logarithmic scale reveals a positive correlation. A perfect relation between pressure and water content cannot be seen for the total of the Tanganasoga crystals on first glance, which might be due to minor variations in crystal compositions (Tables 3 and 4). However, plotting the average water contents of annealed Tanganasoga clinopyroxene crystals for each pressure reveals a trend similar to that observed for the individual crystals. The same pressure relation is observed for clinopyroxene crystals from Shuangcai (sample SC-2). Samples from Lomo Negro and Merapi annealed at different pressures do not show this trend very clearly. For Lomo Negro crystals, which are very homogenous in composition, the trend still holds within the analytical error. The three Merapi crystals annealed at pressure, on the contrary, show differences in their compositions. Crystal MP-8 has significantly more tetrahedral Al3+ (Table 2), which is most likely the reason for the high water content at 0.5 kbar compared to the other crystals at 2 kbar.
Reconstruction of initial water contents in clinopyroxenes
The hypothesis behind the rehydration of clinopyroxenes proposes that their “initial” water content, i.e., their equilibrium water content prior to possible dehydration, can be restored. This previously has been tested in various studies with various different clinopyroxenes and annealing experiments conducted at 1 atm (Sundvall et al. 2009; Sundvall and Skogby 2011; Weis et al. 2015). It remains difficult to show to what extent the initial water content of the crystals can be reached by performing rehydration experiments. For example, de- and subsequent rehydration of synthetic iron-poor diopside showed that only ~75% of the crystal’s initial water content could be restored (Sundvall et al. 2009). Despite this aspect of insufficient rehydration there might be also a risk of excessive hydration of the clinopyroxene crystals. Synthetic, originally dry but Fe3+-rich clinopyroxene has been shown to incorporate substantial amounts of hydrogen upon annealing (Skogby 1994). Furthermore, due to fast ascent with limited to no magma degassing and violent eruption at Tanganasoga volcano (cf. Pinel and Jaupart 2000; Manconi et al. 2009; Stroncik et al. 2009; Pedrazzi et al. 2014), the clinopyroxene crystals from the pyroclastic material most likely were quenched. In this way, the crystals are expected to preserve their initial values (cf. Wade et al. 2008). Yet, our rehydration experiments exceed the water contents in untreated crystals, which might indicate an excessive hydration. Another issue is the possible effect that CO2 and other volatiles might have on the water activity (e.g., Kovács et al. 2012; Yang et al. 2014) and thus on the rehydration of hydrogen-associated defects during the hydrothermal experiments. However, there are also arguments against insufficient as well as excessive rehydration.
The experiments, both de- and rehydration, conducted by Sundvall et al. (2009) on very iron-poor crystals were performed at much higher temperatures (1000 °C) on a timescale of several days. Due to much faster kinetics under these temperatures, vacancy and cation diffusion is much more likely and cannot be excluded. The hydrogen incorporation was not according to the 1:1 relation of redox-reaction 1 and despite the fact that more hydrogen was incorporated than Fe3+ was reduced, yet the original water content was not reached (Sundvall et al. 2009). Considering that the 1:1 relation thus was not followed during dehydration either, vacancy diffusion seems reasonable since the crystal otherwise would have been left without a charge balance upon complete dehydration. The vacancy diffusion also could explain hydrogen intake exceeding the Fe3+ reduction during rehydration. New defects such as vacancies incorporating hydrogen were generated; however, not so many that the initial water contents could be restored. Our crystals, on the contrary, are iron rich. Any dehydration could easily have been compensated by the oxidation of Fe2+. Redox-processes in our experiment also follow more closely the 1:1 relation of reaction 1 (Fig. 5). In addition, experimental conditions in this study make cation and vacancy diffusion very unlikely.
The risk of an excessive rehydration may mostly be linked to the amount of Fe3+ available in the clinopyroxenes. The problem could be that additional available Fe3+, which has not been linked to prior dehydration, is reduced and hydrogen incorporated. The hydration of initially dry, synthetic crystals by Skogby (1994) can be linked to the quantity of trivalent cations in the tetrahedral site. The hydrogen intake of the crystals equals almost 1:1 the amount of the latter (Skogby 1994). Mössbauer analysis further showed additional reduction of Fe3+; however, no additional hydrogen intake occurred. Also, as pointed out previously, not all Fe3+ in the clinopyroxene crystals can be reduced since some of it may be, for example, associated with an aegirine component that only favors a NaHFe2+ component in association with defects suitable for hydrogen incorporation (cf. Skogby and Rossman 1989; Purwin et al. 2009). Thus the defects become the limiting factor for the rehydration. Furthermore, the synthetic crystals in Skogby (1994) were grown under dry conditions that, however, are very unlikely in a magmatic system.
The higher water contents in annealed Tanganasoga clinopyroxene compared to untreated crystals could be a consequence of minor degassing of Tanganasoga magmas and subsequent re-equilibration of the crystals to the new fluid pressure (e.g., Hamada et al. 2011). Our experiments show that pressure differences cause significant changes in clinopyroxene hydrogen content within a few hours. In addition, moderate dehydration by oxidation when exposed to the atmosphere prior to quenching of the clinopyroxenes during eruption can be assumed. Hydrogen diffusion in clinopyroxene is fast enough to evolve diffusion profiles over several millimeters within minutes at magmatic temperatures (cf. Woods et al. 2000; Lloyd 2014) and thus an effect of dehydration even on rapidly erupted crystals cannot be ruled out. Dehydration upon eruption may be indicated by the variation in clinopyroxene water contents (e.g., sample T2, ΔH2OCpx > 120 ppm) within individual rock samples despite homogenous crystal chemistry. Furthermore, pressure-annealed crystals from Shuangcai volcano reach water contents similar to the highest values reported in clinopyroxenes from Gaoping volcano in the Zhejiang province, China (Liu et al. unpublished manuscript) that are interpreted to have preserved more often their initial water content, in contrast to the samples from Shuangcai volcano where crystals are interpreted to have dehydrated.
Another issue regarding excessive hydration of the clinopyroxene crystals is the effect of CO2 on the activity of water. It has been shown that other volatiles such as CO2 or halogens within a magmatic system reduce the incorporation of hydrogen in NAMs (e.g., Stalder et al. 2008; Kovács et al. 2012; Yang et al. 2014). As such, the water contents in olivines annealed in a H2O + CO2 fluid phase, for example, were about half of those treated in pure H2O (Yang et al. 2014). However, these studies concentrated on the solubility of hydrogen in NAMs, and the conducted experiments were done at pressures of several gigapascals and temperatures exceeding 1000 °C. Thus the annealed crystals experienced a resetting of their structural defects, and the influence of CO2 on the generation of hydrogen-associated defects was tested. Experimental conditions in this study, however, do not intend a reset of hydrogen associated defects but only a controlled reversal of redox-reaction 1 and thus a rehydration of already existing yet potentially dehydrated hydrogen-associated defects.
Furthermore, CO2 activities in magmas from the Western Canaries might be low. Degassing of mostly CO2-dominated fluids (~90%) commences at pressures above 1000 MPa in the upper mantle while most clinopyroxene crystallization is below this level (500 to 800 MPa) (e.g., Klügel et al. 2005; Longpré et al. 2008; Stroncik et al. 2009; Weis et al. 2015). Extensive degassing of H2O might only occur upon slow magma ascent in the shallower, i.e., crustal, magma storage levels (e.g., Klügel et al. 2005; Longpré et al. 2008). A similar scenario may be the case for the Merapi clinopyroxenes. The main zone of clinopyroxene crystallization occurs between 300 and 600 MPa where most likely degassing of CO2 and SO2 from water-rich magmas occurs (Costa et al. 2013; Chadwick et al. 2013). Exsolution and degassing of water, on the contrary, has been modeled to commence below 300 MPa (Costa et al. 2013). Thus the influence of CO2 on the crystals in this study may initially have been low. A precise effect of CO2 or other volatiles on the rehydration experiments, however, is unknown and remains to be tested.
Corresponding magmatic water contents
It has been shown that the water content in clinopyroxene phenocrysts can be used to calculate the magmatic water content of their parent melt by applying appropriate crystal/melt partitioning data (Wade et al. 2008; O’Leary et al. 2010; Nazzareni et al. 2011; Okumura 2011; Xia et al. 2013; Weis et al. 2015; Liu et al., unpublished manuscript). This approach can be applied to test the validity of water contents in our rehydrated crystals. The presumption is that the magmatic water content determined through rehydrated clinopyroxenes correlates with values that are either observed or expected for the volcanic environment in which they crystallized.
The values for magmatic water contents for the La Palma melts are considerably higher than those determined by Weis et al. (2015) (0.71 ± 0.07 to 1.49 ± 0.15 wt% H2O) and are more similar to melts from El Hierro’s Tanganasoga volcano (2.84 ± 0.56 to 3.88 ± 0.78 wt% H2O) (Table 6). This overlap is expected considering that both islands are part of the Western Canaries and that Tanganasoga melts can be seen as the volatile-rich primitive magma source in the upper mantle (e.g., Manconi et al. 2009). Magmatic water contents for ankaramite magmas from Teno volcano on Tenerife, a volcanic system similar to Tanganasoga, have been estimated to have been up to 3 wt% (Longpré et al. 2009b; Fig. 8) supporting the high values obtained through pressure-annealed clinopyroxenes. In addition, magmatic water contents for the Western Canaries can be estimated using compositional data of amphibole and the thermobarometric formulations by Ridolfi and Renzulli (2012) (Eqs. 1a and 4). Using compositional data for amphibole pheno- and xenocrysts in basanite and phonolite lavas from La Palma and the kaersutite cumulate xenolith from this study (sample LP1971-B) (Klügel et al. 2000; Barker et al. 2015) reveals magmatic water contents between 3.37 ± 0.78 and 4.48 ± 0.78 wt% (Appendix1 A1) with an average of 3.76 ± 0.78 wt% H2O. These values overlap with the range obtained through pressure-annealed clinopyroxene from the Western Canaries. Most amphibole crystals in the 1971 lavas on La Palma are xenocrysts; however, amphibole in the Western Canaries crystallizes at the same depth as the clinopyroxene (20–45 km) (Klügel et al. 2005; Longpré et al. 2009a; Barker et al. 2015; Weis et al. 2015). Thus the xenocrysts provide a guideline for the H2O content in the volcanic system at this depth. Furthermore, the values obtained from amphibole phenocrysts in phonotephrites (3.47 ± 0.78 to 4.48 ± 0.78 wt% H2O) from the 1949 eruption on La Palma (Klügel et al. 2000) provide a direct comparison for magmatic water contents at depth. Amphibole from sample LB1971-B reveals magmatic water contents between 3.25 ± 0.78 and 4.17 ± 0.78 wt% while pressure-annealed clinpyroxene revealed 4.32 ± 0.86 wt% H2O. The available melt inclusion and volcanic glass data for the Canaries and other ocean island basalts correspond to entrapment pressures between 0.3 and 0.5 kbar (e.g., Dixon et al. 1997; Wallace 1998), which are far below the crystallization pressures of Western Canary clinopyroxene (cf. Klügel et al. 2005; Longpré et al. 2009a; Barker et al. 2015; Weis et al. 2015). Thus the melt inclusions and water contents in clinopyroxene annealed at 1 atm may represent the water contents in the upper parts of the volcanic system after potential H2O loss from the magma while pressure-annealed clinopyroxenes most likely represent the undegassed magmatic water content at crystallization in the upper mantle.
Similar to the results from the Canaries, are the results for Merapi volcano. Merapi clinopyroxenes rehydrated at 1 atm reveal magmatic water contents of 1.25 ± 0.24 and 1.46 ± 0.30 wt%. These values overlap with melt inclusion data reported in the literature (Gertisser 2001; Nadeau et al. 2013; Preece et al. 2014; Fig. 8). Most melt inclusions from Merapi are considered to have undergone H2O loss (Gertisser 2001; Preece et al. 2014), to be re-equilibrated (Preece et al. 2014) or to be of secondary origin (Nadeau et al. 2013). Entrapment and re-equilibration of the melt inclusions occurred at depth between 0.6 and 9.7 km (Nadeau et al. 2013; Preece et al. 2014). Pressure-annealed clinopyroxene phenocrysts, however, reveal magmatic water contents around 6.0 ± 1.2 wt% that overlap with magmatic H2O contents determined through amphibole in the lava samples (~6.0 ± 0.9 wt% H2O; Costa et al. 2013; Nadeau et al. 2013). Crystallization of host clinopyroxene and amphibole occurred much deeper in the volcanic system between 14 and 25 km (Gertisser 2001; Chadwick et al. 2013; Costa et al. 2013; Nadeau et al. 2013; Preece et al. 2014). Furthermore, the clinopyroxene-based magmatic water contents fall within the range of water contents for vapor-saturated basaltic andesite at depths between 14 and 25 km (6 to 10 wt% H2O) (Pineau et al. 1998). The magmatic water content of 2.66 ± 0.54 wt% determined by crystal MP-8, annealed at 0.5 kbar, lies within range of values determined by melt inclusions. Together with clinopyroxene annealed at 1 atm this sample represents the magmatic water contents in the shallower part of the volcanic system after various stages of magma degassing. Phenocrysts annealed at higher pressure, on the contrary, are representative for magmatic water contents at the depth of their crystallization and prior to degassing.
One problem with comparing the clinopyroxene data to magmatic water contents derived by amphibole is the timing of crystallization. NAMs and in this case clinopyroxene usually crystallize earlier than amphibole. Significant fractionation of NAMs thus might enrich the residual melt in H2O that then leads to the onset of amphibole crystallization. Thus the comparison of the two minerals might actually indicate an excessive rehydration of the crystals. However, especially for Merapi, modeling showed that amphibole and clinopyroxene crystallization occurred simultaneously (e.g., Costa et al. 2013). Most Merapi clinopyroxenes in this study are low in Al2O3 (Table 2). The low-Al clinopyroxene has been determined by thermobarometry and modeling to have crystallized in the main magma storage level between 10 and 20 km depth (300 to 600 MPa) together with amphibole and plagioclase at magmatic water contents between 4 and 6 wt% (Chadwick et al. 2013; Costa et al. 2013; Nadeau et al. 2013). Amphibole and high-Al clinopyroxene crystallization already commences at deeper levels (Costa et al. 2013). Results from plagioclase-melt hygrometry indicate magmatic water content of 5.0 ± 0.5 wt% in the main magma storage zone (Surono et al. 2012). Thus the comparison of magmatic water contents derived through amphibole and clinopyroxene seems applicable. However, it remains difficult to establish similar relations between clinopyroxene and amphibole for the samples from the Canary Islands. The phonotephrites of the 1949 eruption are seen as a separate magma batch compared to the basanites (Klügel et al. 2000). Also crystals of both minerals in the kaersutite cumulate sample LP1971-B were accumulated from the same magma, but it is not known if they actually crystallized simultaneously (Barker et al. 2015). Similar, feldspar hygrometry from the La Palma basanites indicates a water content of around 4 wt% in the magma, but crystallization most likely occurred after clinopyroxene (Barker et al. 2015). However, the three independent methods all point toward water contents between 3 and 4 wt%. Furthermore, the highest water contents among untreated clinopyroxene crystals from Tanganasoga would result in magmatic water contents close to 3 wt% and again overlap with values derived through rehydrated crystals. As outlined above, dehydration of the Tanganasoga crystals upon magma degassing and eruption cannot be excluded, which would indicate even higher magmatic water contents at depth.
It cannot be claimed with entire certainty that our rehydration experiments reconstructed the actual initial water contents of clinopyroxenes prior to possible dehydration. However, our results outlined above indicate that it is likely the case.
The rehydration experiments conducted in this study imply that clinopyroxene crystals begin to dehydrate upon magma ascent within a volcanic system. This dehydration is a consequence of the decrease in pressure with vicinity to the surface and the adaption of the crystals to the new conditions. Furthermore, the crystals may undergo dehydration through oxidation during the eruption and eventually may fully dehydrate. Both types of dehydration occur according to redox-reaction 1. Thus the water content of erupted clinopyroxene phenocrysts cannot be taken for granted as a proxy for magmatic or mantle water contents. The pressure- and oxidation-related dehydration may be hindered by fast magma ascent with minimal magma degassing and more violent eruptions that subsequently lead to quenching of the erupted crystals. In this way the initial water content at crystallization may be preserved to large extents. To reach a rehydration, corresponding to the water content at crystallization, of dehydrated clinopyroxene phenocrysts, experiments at 1 atm in hydrogen gas are not sufficient. It requires the input of pressure to restore water contents of clinopyroxene in the mantle. A precise threshold value for pressure at which the water content in clinopyroxene will be held stable has not been established. Further investigation of rehydration at different pressures in comparison with detailed studies on the crystals’ origin and eruption history could help to establish a more precise relation between pressure and clinopyroxene water content. The rehydration at different pressures of dehydrated clinopyroxene phenocrysts from lava samples seems to be able to serve as a proxy for magmatic water contents at different levels in a volcanic system. Thus different pressure experiments could be used to investigate the degassing history of magma in a volcanic system that, in turn, has implications on petrological processes.
We thank Valentin Troll, Frances Deegan, Shaochen Liu, Kirsten Zaczek, and Juan Carlos Carracedo for providing the samples from the Canary Islands and Indonesia, Alexander Nemchin for providing crystals from the Nakhla meteorite, and Shaochen Liu for providing the samples from Shuangcai volcano. Lisa Samrock, Soophie Omidian, and Per-Olof Persson are thanked for their help with sample preparation. We thank Jannick Ingrin for valuable comments on an earlier version of the manuscript and István Kovács and Geoffrey Bromiley for their constructive reviews that all helped to improve this study. Financial support for this project was provided by the Swedish Research Council. The data for this paper are available in the text, tables, and references therein and from the corresponding author on request.