Uniform “water” content in quartz phenocrysts from silicic pyroclastic fallout deposits – implications on pre-eruptive conditions

Abstract. Structural hydroxyl content of volcanic quartz phenocrysts was
investigated with unpolarized Fourier-transform infrared spectroscopy. The
phenocrysts originated from five pyroclastic fallout deposits from the
Bükk Foreland Volcanic Area (BFVA), Hungary, and two from the AD 1314
Kaharoa eruption (KH eruption), Okataina Volcanic Complex (Taupo Volcanic
Zone), New Zealand. All investigated quartz populations contain structural
hydroxyl content in a narrow range with an average of 9.3 (±1.7) wt ppm. The earlier correlated horizons in the BFVA had the same average
structural hydroxyl content (within uncertainty). Thus, it can be concluded
that the structural hydroxyl content does not depend on the geographical
distance of outcrops of the same units or the temperature or type of the
covering deposit. The rare outlier values and similar structural hydroxyl
contents show that the fallout horizons cooled fast enough to retain their
original structural hydroxyl content. The similarity of the structural
hydroxyl contents may be the result of similar P, T, and x (most importantly
H2O and the availability of other monovalent cations) conditions in the
magmatic plumbing system just before eruption. Therefore, we envisage common
physical–chemical conditions, which set the structural hydroxyl content in
the quartz phenocrysts and, consequently, the water content of the host
magma (∼ 5.5 wt %–7 wt % H2O) in a relatively narrow
range close to water saturation.



Introduction
Water content significantly affects the physical properties (density, viscosity) of magma and has a large impact on the magnitude and type of the volcanic eruptions especially by moderating explosivity (Sparks, 1978;Dixon and Stolper, 1995;Dingwell, 1996;Popa et al., 2019Popa et al., , 2020Popa et al., , 2021Allabar et al., 2020). Therefore, it is important to find proxies which allow estimation of the original water content of the magma, and for this reason, investigation of the water content of nominally anhydrous minerals (NAMs hereafter) has come to the forefront of this research. For estimating the original magmatic water content, the commonly used NAMs are clinopyroxene (Weis et al., 2015;Lloyd et al., 2016;Turner et al., 2017), olivine (Portnyagin et al., 2008;Kovács et al., 2010;Barth et al., 2019) and plagioclase (Johnson andRossman, 2003, 2013;Johnson, 2006;Hamada et al., 2011Hamada et al., , 2013. Quartz is one of the most abundant minerals in the Earth's continental crust and occurs very often in the most differentiated, silicic magmatic rocks (Ronov and Yaroshevski, 1969). Hence, for calculation of magmatic water content quartz can be a good candidate due to its abundance and resistivity against environmental alteration factors (i.e., weathering). Quartz is nominally anhydrous but incorporates certain amounts of OH by coupled substitutions (Kats, 1962;Müller and Koch-Müller, 2009;Stalder and Konzett, 2012;Frigo et al., 2016;Biró et al., 2016Biró et al., , 2017Jollands et al., 2020a). An effective method to determine the structural hydroxyl content of NAMs is micro-Fourier transform infrared (FTIR hereafter) due to its sensitivity to the hydroxyl vibrations in minerals. It is a cost-effective method with a relatively simple preparation procedure compared to that of melt inclusion studies. FTIR spectra of quartz have been studied for decades (e.g., Kats, 1962), and the incorporation of trace species (most importantly H + ) is well-established: the most common heterovalent substitution in quartz is Si 4+ ← Al 3+ + H + , which results in AlOH defects causing three characteristic absorption bands at 3310, 3378, and 3430 cm −1 , respectively in the FTIR spectra (Kats, 1962;Stalder and Konzett, 2012;Frigo et al., 2016;Biró et al., 2017;Jollands et al., 2020a). In the case of Li-bearing systems, a sharp band occurs at 3483 cm −1 , which is caused by LiOH defects (Frigo et al., 2016), and commonly appears in the absorption spectra of volcanic quartz phenocrysts (Biró et al., 2017;Jollands et al., 2020a). A broad and weak absorption band can occur between ∼ 3480 and 3100 cm −1 , centered at 3400 cm −1 , which is usually interpreted as molecular H 2 O present in (nano)inclusions (Aines and Rossman, 1984;Stenina, 2004;Biró et al., 2016). Another subordinated incorporation mechanism of H+ is evidenced by a band centered at 3585 cm −1 corresponding to weakly H-bonded silanol groups in pure quartz (Stalder and Konzett, 2012). In addition, a characteristic peak (centered at ∼ 3200 cm −1 ) may represent more strongly bonded H to the mineral surface (Biró et al., 2016;Stalder, 2021), or it can be assigned to Si-O bonds (Kats, 1962). Nevertheless, assignation of this peak remained controversial (Stalder, 2021). The estimation of original magmatic water content using quartz as a NAM and its melt inclusions can be challenging due to syn-and post-eruptive processes, which can modify (reduce in most cases) the structural hydroxyl content of the NAM (Biró et al., 2017). H diffusion in natural quartz is isotropic and relatively fast at volcanic temperatures (700-800 • C) on the order of 10 −10 to 10 −14 m 2 s −1 (Kats 1962;Kronenberg et al., 1986;Jollands et al., 2020b). This diffusivity is fast enough to produce diffusion profiles, which are applicable for the determination of timescales of magma ascent rates and eruption history (Myers et al., 2018(Myers et al., , 2019Tollan et al., 2019;Jollands et al., 2020b). However, this rapid diffusion is also sufficient to lose a significant portion of the original H content of the quartz phenocrysts exposed to high temperatures (300-700 • C) for time intervals ranging from days to months under decreasing water activity (e.g., degassing, slow cooling at the surface; Biró et al., 2017;Stalder et al., 2017). Another problem is the Li vs. H competition for the charge-balancing position connected to Al, because Li is also highly mobile and can be characterized with fast diffusion . However, recent studies imply that Li compensates for the Al caveats to the detriment of H, when the water activity in melt drops significantly, but Li activity remains the same Jollands et al., 2020b). This means that during ascension of the magma, a little amount of H diffuses out from the rims of quartz crystal, and Li enters into the structure to compensate for the charge balance which was broken by the dehydration . Other elements (e.g., major elements, like Na) possibly do not play a significant role in charge balancing of Al in competition with H because they diffuse much more slowly (Cherniak and Dimanov, 2010;Zhang et al., 2010).
In this study, we report AlOH structural hydroxyl concentrations of quartz phenocrysts from five Miocene pyroclastic fallout deposits of the Bükk Foreland Volcanic Area (BFVA hereafter), northern Hungary, Carpathian-Pannonian region (Fig. 1), and two pyroclastic fallout deposits of the AD 1314 Kaharoa eruption, Tarawera, New Zealand (KH eruption hereafter). The studied pyroclastic fallout deposits were selected in order to minimize the effect of post-and syn-eruptive hydrogen loss (Wallace et al., 2003;Jollands et al., 2020b); thus presumably their AlOH structural hydroxyl content can be used to infer the original water content of the host magma. Furthermore, in this way, we investigate whether the structural hydroxyl content of quartz phenocrysts depends on any volcanological conditions (such as lateral distance from the vent, stratigraphic position, welding of the covering deposit, etc.). The most important factors controlling the structural hydroxyl content of quartz in felsic volcanic systems producing explosive eruptions are discussed.
2 Geodynamic setting and volcanism 2.1 Silicic magmatic systems -BFVA eruptions and KH eruption In silicic systems the uppermost level of the magmatic plumbing system is characterized by distinct magma batches containing crystal mush and melt lenses (Gualda et al., 2018). The eruptible magma could be stored at around 1.5 to 2 kbar pressure (i.e., 6-7 km below the surface), characterized by a bunch of sill-and dyke-shaped bodies and other small, long-lived magma pockets (Matthews et al., 2012;Kennedy et al., 2018;Gualda et al., 2019;Huber et al., 2019). However, recent studies suggest a shallower (0.5-1.5 kbar pressure zone) emplacement of magma bodies, where the last step of the crystallization and segregation of the eruptible melt occur just before eruption (Graeter et al., 2015;Tol-lan et al., 2019). The root of the magmatic plumbing system may go down as far as the mantle-crust boundary, which is the so-called MASH zone (melting, assimilation, storage and homogenization; Hildreth and Moorbath, 1988).
Magmatic storage system of the BFVA was presumably initiated in the Early Miocene with long-lasting (starting from ca. 18.5 Ma according to Lukács et al., 2018) subvolcanic activity, which created a complex, multi-level magmatic reservoir in the upper crust (for hundreds of kiloyears) with several shallow silicic melt pockets, which fed the eruptions via magma mingling Lukács et al., 2005Lukács et al., , 2015Lukács et al., , 2018. The bimodality of the chemical composition of some units (andesitic and rhyolitic;  suggests that the partial melting of the asthenosphere and lithospheric mantle produced mafic melts, which were stalled in the vicinity of the mantle-crust boundary (Lukács et al., , 2018 and evolved at mid-crustal levels Cooper et al., 2016;Kósik et al., 2019Kósik et al., , 2020. The extensional thinning of the lithosphere had already taken place by the Late Miocene, which favored the ascent of mafic magmas from lower crustal reservoirs into the upper crust Lukács et al., 2018;Petrik et al., 2019). The best analogue for the magmatic system and for the volcanism of the BFVA could be the Taupo Volcanic Zone in terms of volcanic morphology (nested calderas; Wilson, 1993Wilson, , 2001, topography (lowland setting, possible influence of sea or lake water; Wilson, 2001;Biró et al., 2020), geodynamic setting (extensional back-arc, normal fault systems), existence of rift zones (Wilson and Walker, 1985, 1993, 2001Graham et al., 1995;Nairn et al., 2004;Wilson and Rowland, 2016;Stagpoole et al., 2020), and volcanic eruption styles (silicic Plinian ignimbrite-forming eruptions, phreatoplinian eruptions, small-scale silicic monogenetic eruptions; Lukács et al., 2005Lukács et al., , 2018Barker et al., 2020;Biró et al., 2020;Kósik et al., 2020;Hencz et al., 2021a;b).
In recent models of the magmatic system of the KH eruption, a single, stratified magma chamber was suggested based on geochemical features of different magmatic products (Nairn et al., 2004). Mafic magma recharge was obvious throughout the lifetime of the magmatic system, which may have triggered the eruptions but could not destroy the layered (in terms of silica content) magma chamber (Leonard et al., 2002;Nairn et al., 2004Nairn et al., , 2005. The depth of the uppermost level of the Taupo Volcanic Zone (TVZ) system (where the KH magma originated from) was calculated to be between 6 and 7 km (1.6-2 kbar pressure) based on mineral thermobarometry and glass composition (Ewart et al., 1975;Nicholls et al., 1992;Blundy and Cashman, 2001).

Tectonic setting of the BFVA
During the Early to Mid-Miocene large-scale silicic volcanism was taking place in the Carpathian-Pannonian region (Schréter, 1939;Szakács et al., 1998;Lukács et al., 2018). This felsic calc-alkaline magmatism was preceded by various subduction stages, which re-fertilized the lithospheric mantle and asthenosphere, allowing partial melting via volatile fluxing (Szabó et al., 1992;Kovács and Szabó, 2008). During the eastward movement and rotation of the ALCAPA (Alpine-Carpathian-Pannonian) microplate ( Fig. 1), lithospheric thinning was taking place in the inner part of the ALCAPA microplate of which the exact driving mechanism is still debated Horváth, 1993;Márton and Márton, 1996;Márton et al., 2007;Kovács et al., 2012;Horváth et al., 2015;Balázs et al., 2016). During the rotation normal faults were formed in the crustal part of AL-CAPA . Due to the re-fertilization of the mantle (i), crustal thinning in the ALCAPA block (ii), and normal fault propagation in the crust (iii), silica-rich melts formed via fractional crystallization and crustal assimilation, erupting through large-scale ignimbrite-forming explosive volcanic activity on the eastern part of ALCAPA between ca. 18 and 14 Ma based on zircon U-Pb ages (Lukács et al., 2015(Lukács et al., , 2018. The deposits generated by the prevailingly explosive volcanism (pyroclastic fallout deposits, PDC deposits, reworked tephras) are now buried under Late Miocene-Quaternary sediments and crop out continuously in coherent condition in a narrow (40 × 10 km) belt between the Mesozoic sedimentary block of the Bükk Mountains and the Great Hungarian Plain (Fig. 1). The most voluminous deposits are ignimbrites showing different facies (welded, nonwelded, distal, etc.), but pyroclastic fallout horizons occasionally also occur (Capaccioni et al., 1995;Szakács et al., 1998;Biró et al., 2020;Hencz et al., 2021a, b).
For the present study, we sampled pyroclastic fallout layers from two different parts of the stratigraphic column (Table 1 and Fig. 2), which represents different episodes (i.e., age intervals) of the volcanism: from the middle part of the Lower Pyroclastic Complex (LPC, base of the Mangó ignimbrite unit of Lukács et al., 2018) and from the lowest part of the Upper Pyroclastic Complex (UPC, right above the Jató member distinguished by Biró et al., 2020).

The AD 1314 Kaharoa eruption, New Zealand (KH eruption)
The KH eruption occurred at AD 1314 (±12) in the North Island of New Zealand Newnham et al., 1998;Hogg et al., 2003;Klemetti and Cooper, 2007;Robertson, 2007) in an extensional geotectonic setting inside a continental plate (Graham et al., 1995). The main rhyolitic volcanic zone in New Zealand is situated in the Taupo Volcanic Zone (TVZ hereafter), where the volcanic activity began around 2 Ma with andesitic activity, which marks the transition of the volcanic activity from the older Coromandel Volcanic Zone (CVZ) to the TVZ (Wilson et al., 1995;Gravley et al., 2016) (Fig. 3). The volcanism of the TVZ can be divided into two large time periods: 2-0.7 Ma, when andesite-dominated volcanism occurred, and  Szakács et al., 1998;Márton and Pécskay, 1998;Márton et al., 2007). The sampling points are indicated with stars. 0.7-present, which is dominated by rhyolitic volcanism in the form of large-scale ignimbrite-forming explosive volcanism and caldera-forming eruptions (Deering et al., 2011). The ascendent transport of the melts was facilitated by fault zones (Wright, 1992;Graham et al., 1995). In recent times, several calderas occupied the central TVZ (e.g., Whakamaru, Mangakino, Kapenga; Wilson et al., 1995), and the eruption center of the KH eruption occurred at the Tarawera volcano intra-caldera rhyolite-dome complex within the Haroharo caldera of the Okataina Volcanic Centre, located at the northern-central part of the TVZ (Graham et al., 1995;Lowe et al., 1998;Nairn et al., 2001Nairn et al., , 2004Sahetapy-Engel et al., 2014) (Fig. 3). Explosive volcanic activity occurred from seven vents spread along an 8 km long lineament (Graham et al., 1995;Nairn et al., 2001). The eruptions were fed by three (and one mixed) rhyolite magma batches with slightly varying geochemistry (Nairn et al., 2004). Largescale pyroclastic fallout activity dominated the volcanism, producing well-identifiable fallout horizons across the region (Sahetapy-Engel et al., 2014). For this study, two pyroclastic fallout horizons of the KH eruption were sampled representing the early, high-silica content rhyolitic phase of the volcanism, originating from well-known outcrops (Table 1).  Biró et al. (2020) This sequence (Unit III/subunit A in Biró et al., 2020) built up by 3-20 cm thick coarse, well-sorted tuff layers containing pumice clasts, lithic clasts, and millimeter-sized quartz, feldspar, and biotite phenocrysts. These layers were interpreted as pyroclastic fallout horizons.
The separated quartz phenocrysts are up to 1 mm in diameter and are euhedral, bipyramidal crystals.   A pyroclastic fallout horizon was sampled at the base of the Mangó ignimbrite unit (following the nomenclature of Lukács et al., 2018). This layer was sampled in three different geographical sites (E, O, S). This is a well-sorted, topography-mantling layer and contains pumice fragments, lithic clasts, ash matrix in lower volume, and phenocrysts, such as quartz, biotite, and feldspar. The deposit is the thickest in the eastern BFVA at Sály (Latorvár, in the southern slope of the hill; S), where it is ∼ 70 cm, and it is the thinnest in a mine near the town of Eger, where it is only 20 cm thick. Underneath there is a brown paleosol, and over the fallout horizon there is deposit from dilute PDC showing faint crossbedding (20 cm in thickness) and topped with a min. 30 m thick non-welded ignimbrite in Eger and Ostoros and with a welded ignimbrite in Sály. The pyroclastic fallout material was deposited in an early phase of the eruption which created the overlying thick Mangó ignimbrite. The eruption center could have been in the larger vicinity of Sály.
The separated quartz phenocrysts are relatively large, over 1 mm in diameter, euhedral, and bipyramidal. Melt inclusions are common.  2014), which consists of fallout-lapilli with the capping ash. The dispersal was mainly towards the southeast. This is the first lapilli-ashfall of the KH eruption with regional dispersal. Below this bed, an explosion breccia crops out, very visible in the case of site KH2. The capping pyroclastic sediment is a lapilli fall deposit (layer B in Sahetapy-Engel et al., 2014) The separated quartz phenocrysts are small, about 500 µm in diameter, euhedral, and bipyramidal.
Melt inclusions are very common and sometimes contain bubbles. Nairn et al. (2001Nairn et al. ( , 2004 3 Methods and sampling

Sampling and sample preparation
The middle part of individual pyroclastic fallout layers was sampled. When the rock was slightly cemented, the sampling was made by using a hammer. Then, the sample was crushed by hand and washed. After drying (at 60 • C for 180 min) the crushed bulk sample was sieved to separate the 0.5-2 mm fraction. About 30 quartz phenocrysts were separated from each layer in the mentioned grain-size fraction under a binocular microscope. Euhedral and subhedral bipyramidal quartz phenocrysts were preferred for selection.
The phenocrysts were polished on two sides following the protocol described in Biró et al. (2016). The phenocrysts were then placed on a glass plate and were embedded into gel glue (Loctite ® Super Bond Power Flex Gel glue) in an unoriented manner. The mounts were ground using Al 2 O 3 powder until all the separated phenocrysts became uniformly exposed (∼ 600-700 µm in thickness), and then they were  Graham et al. (1995), Briggs et al. (2005), and Kósik (2018). polished with Al 2 O 3 suspension. The mounts were dissolved with acetone from the supporting glass slide and then the phenocrysts were turned to their other side and were again attached to the glass slide with liquid adhesive (Loctite ® Super Bond Universal liquid glue). The mounts were ground to ∼ 250 µm thickness to be suitable for micro-FTIR measurements (Biró et al., 2016), then they were polished again with Al 2 O 3 . They were separated from the glass plate, washed in running water, and dried out at 90 • C for 2 h.

Peculiarities of the sample preparation
In the case of quartz phenocrysts with homogenous water content (when melt inclusions are relatively rare in the crystal), unpolarized absorbances of hydroxyl-related bands should ideally vary between predicted minimum and maximum values as follows from the unpolarized absorbance indicatrix of quartz Sambridge et al., 2008;Stalder and Konzett, 2012;Biró et al., 2017). A theoretical consideration for unpolarized measurements was given earlier Sambridge et al., 2008), a recent work also referred to the strategy of polarized measurements on oriented grains (Stalder and Konzett, 2012), while the connection was given between the polarized and unpolarized absorbance indicatrix of Al-related hydroxyl defects in quartz (Biró et al., 2016). Applying unpolarized IR radiation for the measurements, we get the maximum absorbance when the incident infrared radiation travels parallel to the crystallographic c axis. In contrast, unpolarized absorbance is minimal when the incoming light travels per- pendicular to the c axis; in this case the absorbance should be ideally half of what is measured in the former case, because of the orientation of the hydroxyl dipole perpendicular to the c axis (as summarized by Biró et al., 2017). Our first round of measurements implied that the applied sample preparation will result in an overrepresentation of lower absorbances (i.e., the incident light is most often almost perpendicular to the c axis). This is due to the fact that the euhedralsubhedral bipyramidal crystals were laid on the glass plate on their pyramidal facets or prism surfaces (Fig. 4). Thus, the measured absorbances may be slightly lower than if the grains were perfectly unoriented. Perfectly unoriented grains mean that the direction of the incident light is equally distributed over a sphere with respect to the absorbance indicatrix (see Biró et al., 2017, for more discussion). Since the preparation of the samples was consistent and similar in each case, the results are comparable to each other, and the real concentrations are expected to be underestimated by the same extent. To follow up this "unperfect unorientation" issue, manually crushed crystals were prepared (see Appendix A). The results assured us that (1) the crushed crystals can retrieve accurate absolute concentrations.
(2) Thus we could quantitatively constrain the degree of underestimation with the currently applied methodology (in contrast with the one applied in Biró et al., 2017).
(3) The determined concentrations with the applied methodology are robust for exploring relative variations quantitatively.

FTIR methodology
The IR spectra of the quartz wafers were recorded in a temperature-controlled laboratory at the HAS RCAES with a Bruker ® Hyperion 2000 IR microscope attached to a Bruker ® Vertex 70 spectrometer. We used unpolarized FTIR methodology in light of indicatrix theory for unpolarized light Sambridge et al., 2008). This makes it possible to determine the structural hydroxyl content of anisotropic crystals when the following conditions are met: (1) at least a few crystals are measured from the same population (> 5), (2) the crystals are unoriented with respect to the incident IR light, (3) structural hydroxyl content is homogenous within and among crystals, and (4) maximum linear absorbance of the spectra in the hydroxyl region (roughly between 3000 and 3500 cm −1 ) does not exceed 0.15 Sambridge et al., 2008).
During the FTIR measurements the aperture was 50 × 50 µm. The spectra were recorded with 128 scans, with 4 cm −1 spectral resolution between 400-4000 cm −1 . During the measurements, atmospheric compensation of the OPUS ® software was used to eliminate the contribution of water vapor and CO 2 from the atmosphere. The spectra were obtained carefully from parts of the crystal wafers where melt or fluid inclusions were not visible. Measurements along transects and in crystal cores and rims were also acquired in order to investigate the intra-crystal variations in the structural hydroxyl content.
The obtained spectra then were processed using the OPUS ® software. The integration was calculated using the "B-type" integration method of the OPUS ® software ( Fig. 5; see Fig. 2 of Biró et al., 2016, andFig. S3 of Reynes et al., 2018). The B-type integration includes the area above a line connecting the intersections of the lower and upper limits of integration (Fig. 5). Thus, baseline correction was not necessary, because it could cause a significant underestimation of the real structural hydroxyl content. Moreover, using the Btype integration, the possible contribution of molecular water can be minimalized, because it takes into account only the bands that correspond to the structural hydroxyl signal and not the abovementioned wide and flat band of molecular H 2 O. Structural hydroxyl is more stable and better preserved in the phenocryst than the molecular H 2 O, which is more sensitive to external influences (heat, pressure, chemical activity, etc.). The integration boundaries are given in Table 2. The structural hydroxyl content of the quartz phenocrysts was calculated using the modified Beer-Lambert law (Eq. 1): where c is the calculated water concentration (in H 2 O wt ppm); A tot is the total polarized absorbance; M i is the molar weight of the water (in g/mol); ε i is the integrated absorption coefficient (where 94 000 ± 20 000 L/mol cm −2 was used according to Thomas et al. (2009), because this number had been established from measurements from natural quartz phenocrysts, of which the spectrum was identical to the spectra in the case of volcanic quartz phenocrysts); ρ is the density of the phenocryst (2650 g/L); and t is the thickness of the sample. The total polarized absorbance was calculated as 3 times the (average) integrated unpolarized absorbance of bands caused by AlOH structural hydroxyl substitution (centered at 3430, 3378, and 3315 cm −1 ), which are dominant in the hydroxyl region of the IR spectra. The minor bands (such as at 3483 or at 3200 cm −1 ) are negligible and were not included into the total structural hydroxyl content. The thickness was calculated using the method described by Biró et al. (2016) according to Eq. (2): where x is the sample thickness in micrometers and y is the integrated area of Si-O bands between 2110-1440 cm −1 . The structural hydroxyl contents are expressed in molecular water equivalent and are related to the AlOH structural hydroxyl content (Table 2) and characterized by typically ±15 % cumulative uncertainty. The unit we used ("wt ppm") means wt ppm H 2 O. Note that we also calculated "indicative" water concentrations for individual measurements from a given sample population, which, in the ideal case, varies only due to the variation in the direction of the incident light with respect to the absorbance indicatrix. Also note that the average of these measurements from a population of unoriented crystals can be used to estimate the true absolute structural hydroxyl content.

FTIR spectra and structural hydroxyl content of the quartz phenocrysts
Total evaluation, including the exact calculations, can be found in Supplement files S1 and S2. The summarized analytical results can be found in Table 3. The average unpolarized spectra for each layer are shown in Fig. 5, including the calculated structural hydroxyl contents and the cumulative uncertainty in brackets. The results depicted in the box-andwhisker plot are displayed in Fig. 6.

Eger (E), Ostoros (O), and Sály (S) sites (all samples from the same fallout horizon of the LPC)
The hydroxyl region of the spectra is dominated by three sharp bands, the largest of which is centered at 3379 cm −1 , and the two smaller ones are centered at 3314 and 3430 cm −1 . A minor weak band is present at 3478 cm −1 , but in some wafers it is absent. Another minor band can also be observed at 3197 cm −1 in the case of all samples. The broad band centered at ∼ 3400 cm −1 associated with molecular water was not registered. The indicative structural hydroxyl content of the phenocrysts is in the range of 6.8-15.1, 7.0-14.1, and 6.4-12.8 wt ppm in the case of E, O, and S samples, respectively.  The calculated structural hydroxyl contents are 9.2 (±1.7), 9.4 (±1.6), and 9.2 (±1.6) wt ppm (standard deviations in brackets).

Tibolddaróc (TD) and Bogács (BG) sites (all samples from the same fallout horizon of the UPC)
The triplet with a main band at 3379 cm −1 and minor bands at 3314 and 3430 cm −1 is clearly visible and dominates the hydroxyl region of the spectra. The LiOH band (3478 cm −1 ) is not visible (or very minor) in both the TD and BG samples. However, a sharper but minor band can be clearly seen at 3197 cm −1 . The calculated structural hydroxyl content in TD is 9.1 (±1.3) wt ppm, and the indicative concentrations are in the range of 7.1 to 12.4 wt ppm. In the case of BG, the calculated value is 9.1 (±1.3) wt ppm, and the measured values are in the 7.1-12.0 wt ppm range.

KH eruption sites
The hydroxyl region is dominated by the AlOH triplet centered at 3314, 3379, and 3430 cm −1 , similar to the BFVA samples. The LiOH band (3478 cm −1 ) is visible with low intensity. A very minor band can be seen at 3197 cm −1 .
The calculated structural hydroxyl content of the KH1 sample is 9.0 (± 1.4) wt ppm. The indicative structural hydroxyl contents are in the range of 6.0 to 11.4 wt ppm. In the case of the KH2 sample, the structural hydroxyl content is slightly higher, 10.2 (±2.6) wt ppm. The range of the struc-  tural hydroxyl contents is 1.4 to 15.7 wt ppm, but both values are considered outliers (this is also the reason for the relatively high standard deviation).

Transect and core-rim measurements
The results of a representative transect measurement (a) and of two representative core-rim measurements (b, c) are shown in Fig. 7. In general, no significant variability can be recognized along the transects. This also stands for the corerim measurements, where there is no significant and consistent difference in the structural hydroxyl content of the different parts of the measured crystals in the case of both KH and BFVA samples. Hence, possibly there is also no significant H + loss via diffusion, so all the sampled fallout deposits (BFVA and KH eruption alike) probably cooled down fast enough to avoid this post-emplacement diffusion loss.

Band assignment
In all studied samples the hydroxyl region of the IR spectra is dominated by the AlOH triplet (at 3430, 3378, and 3315 cm −1 , respectively) (Kats, 1962;Stalder and Konzett, 2012). These bands are related to coupled Al 3+ + H + substitution into the tetrahedral site replacing Si 4+ . In case of E, O, S, BG, and TD the band at 3197 cm −1 is very sharp and may be connected to the strongly bonding H + at the quartz crystal surface (Biró et al., 2016;Stalder, 2021) or can be assigned to Si-O bonds (Kats, 1962). The minor but sharp band at 3478 cm −1 could be linked to the LiOH structural hydroxyl defects (Potrafke et al., 2020). The results of the measurements along transects and results of core-rim measurements showed that no intensity growth of the Li band (as described in the case of quartz from pyroclastic and granitic igneous bodies by Tollan et al., 2019, andPotrafke et al., 2020) or significant variations in the AlOH bands are visible in the spectra ( Fig. 7; Tollan et al., 2019;Potrafke et al., 2020).

Comparison with formerly published results
The water content of natural quartz crystals from several rock types (sandstone, hydrothermal veins, granite) as well as of laboratory-grown crystals was in between 0 and a few hundred weight parts per million (Miyoshi et al., 2005;Müller and Koch-Müller, 2009;Stalder and Neuser, 2013;Stalder, 2014;Baron et al., 2015;Frigo et al., 2016;Stalder et al., 2017;Potrafke et al., 2019Potrafke et al., , 2020Jollands et al., 2020b;Stalder, 2021). Hydroxyl content of volcanic quartz was investigated only in a few studies. The "water" content of quartz phenocrysts originating from a pyroclastic fallout layer of the Bishop Tuff is between 7 and 12 wt ppm (recalculated from Jollands et al., 2020b). Quartz crystals separated from rhyolites originating from Germany and Italy (Großer Inselsberg and Bolzano) showed 1.4 to 12 wt ppm water content (Stalder and Neuser, 2013). In the BFVA volcanic quartz from a fallout horizon and several ignimbrite sequences was investigated recently (Biró et al., 2016(Biró et al., , 2017. The results show a continuous decrease in the hydroxyl content from ∼ 13 to ∼ 3 wt ppm (or even in some cases close to zero) towards the middle part of the ignimbrite; thus they cannot be regarded as the original water content of the phenocrysts (Biró et al., 2017). Quartz phenocrysts from the fallout horizon right below the Mangó ignimbrite at the Eger site showed 12.1 wt ppm structural hydroxyl content (E-1 in Biró et al., 2017). This is in good agreement with the results obtained on quartz from the lowest part of the covering ignimbrite with 11.4 wt ppm (E-2 in Biró et al., 2017). The reason for the difference between this result (12.1 wt ppm) and our calculated structural hydroxyl content for quartz phenocrysts originating from this layer (9.2 wt ppm) is due to the different sample preparation methodology including using whole vs. fragmented crystals for the measurements (see Appendix A about single-crystal measurements).
There are several frequency distribution plots of the measured structural hydroxyl content of quartz phenocrysts (obtained using FTIR measurements with polarized light) originating mostly from granite bodies, metamorphic rocks, or detrital grains reported by Neuser (2013), Stalder (2014) and Stalder et al. (2017). The present OH defect contents are close to the global average of 10 wt ppm (Stalder and Neuser, 2013;Stalder, 2014;Stalder et al., 2017). The 0-10 wt ppm range dominates these distribution plots (Fig. 7 of Neuser, 2013, andFig. 8 of Stalder et al., 2017), and the range we measured agrees well with these values. Quartz-bearing pyroclastic fallout deposits can cover large areas (up to a few tens of thousands of square kilometers; e.g., Wilson, 1993); hence volcanic quartz population originating from explosive eruptions (just like the sediments which formed after reworking of the fallout deposit) can also contribute to the crustal average. Vice versa, when a quartz population is found in an accumulated, reworked deposit with ∼ 10 wt ppm characteristic structural hydroxyl content, volcanic origin can be assumed in the provenance analyses.

Uniform P , T , and x conditions in silicic magma reservoirs recorded by OH in quartz
The structural hydroxyl contents of the quartz phenocryst populations are in a narrow range of 9.0-10.2 wt ppm in the correlated horizons of both BFVA and KH eruption samples (Figs. 5, 6 and Table 3). This narrow range in the average structural hydroxyl content could reflect similarities in the P , T , and x (mainly H 2 O) activity conditions in the magma storage system when the quartz phenocrysts equilibrated with the magma just before the significant degassing and/or eruption events. Water contents of other NAMs (clinopyroxene, feldspar, etc.) can record conditions at different levels of the magmatic system, because they crystallize at different levels characterized by different P , T , and x conditions (e.g., plagioclase; Hamada et al., 2011). For hydrogen a few minutes to hours are enough to attain equilibrium between the melt and the solid phase (crystals) due to its fast diffusion rates (10 −10 to 10 −14 m 2 s −1 at 800 • C in case of quartz and 10 −12 m 2 s −1 in the case of plagioclase; Hamada et al., 2013;Johnson and Rossman, 2013;Lloyd et al., 2016;Myers et al., 2018;Jollands et al., 2020b). Quartz usually crystallizes in the upper(most) part of the magmatic plumbing system over a geologically short period of time (a few hundred to few thousand years; Gualda et al., 2012;Matthews et al., 2012;Graeter et al., 2015;Tollan et al., 2019;Jollands et al., 2020b). The narrow range of the measured structural hydroxyl contents Figure 7. Example of structural hydroxyl concentration along a transect (a) and observed in core-rim relation (b, c). Note that no significant difference in structural hydroxyl content was observed along a transect nor in core vs. rim relation.
suggests a relatively uniform P , T , and x environment and fast ascension (Myers et al., 2018). The amount of dehydration during or just before the eruption via Li-H interdiffusion (e.g., Jollands et al., 2020b) was very minor based on the fact that (1) there is no strong zonation of OH towards the rims of the crystals and (2) the intergranular variability of obtained OH concentrations is rather small. If Li-H interdiffusion was significant just before and during the eruption, various grain sizes should give various OH contents according to the grain size dependence of diffusion. During this latest stage of the crystallization there was enough time for re-equilibration of the quartz phenocrysts to eliminate any zoning or difference of the structural hydroxyl content in the crystal's structure (as seen in transect and core-rim measurements, Fig. 7).
Consequently, it can be inferred that the quartz phenocrysts (separated from pyroclastic fallout deposits) have almost the same structural hydroxyl content throughout the whole evolution of the felsic explosive volcanism in the BFVA. The two sampled layers can represent two large eruptive complexes (consist of several eruptions) of the BFVA (Fig. 2), besides several more eruption events (fed by one magma pocket; took up to several days or weeks) that occurred during the volcanism of the BFVA (Biró et al., 2020;Hencz et al., 2021a, b). This feature may be universal, since we measured identical structural hydroxyl contents in another much younger (essentially recent) volcanic field in New Zealand.

Implications on the water content of the host melt
Determining the dissolved water content of the silicate melt is essential to understand the dynamics of the volcanic eruptions (besides other magmatic volatiles, like CO 2 , S, Cl, and F; Cashman, 2004) and to be able to predict the course of future eruptions based on the physical and chemical properties of the magma including its possible ascent rates (Lloyd et al., 2014;Myers et al., 2019;Popa et al., 2019;Tollan et al., 2019;Allabar et al., 2020). Water content of NAMs can be in equilibrium with the dissolved water content of the melt in each level of the magma storage system, so their water content can be a good proxy to estimate the magmatic water content (Wallace, 2005;Portnyagin et al., 2008;Lloyd et al., 2013;Weis et al., 2015;Tollan et al., 2019;Jollands et al., 2020b). While FTIR is a useful tool to measure the water content of different NAMs as was demonstrated above, calculating magmatic water contents requires the application of a partition coefficient (i.e., the quantitative relation between water content of a given nominally anhydrous phase and the melt; Qin et al., 1992;Johnson andRossman, 2003, 2013;Hauri et al., 2006;O'Leary et al., 2010;Lloyd et al., 2016).
Crystal and melt partition coefficients are available for several NAMs (e.g., clinopyroxene, plagioclase, olivine; Hauri et al., 2006;Grant et al., 2007;Hamada et al., 2013;Liu et al., 2015) but are not available for quartz. According to our best knowledge, only one single calculated partition coefficient is found in the published literature (i.e., D quartz/melt = 0.0001; Qin et al., 1992), where the partition coefficient was tentatively calculated based on applying a mathematical model which investigates the diffusive re-equilibration of H in melt inclusions with the host melt (Qin et al., 1992). In this partition coefficient the molecular water was also included (Qin et al., 1992;Myers et al., 2019). Using our new structural hydroxyl contents of quartz phenocrysts presented above (∼ 9.6 wt ppm, which is the average of the calculated structural hydroxyl contents of KH eruption samples), it is possible to calculate the partition coefficient between quartz and silicate melts in the case of KH eruptions, because there are available magmatic water contents presented recently based on melt inclusion analyses (6.1 wt %-6.5 wt % H 2 O; Nairn et al., 2004). Using these data, the crystal-melt partition coefficient for quartz is 0.000157-0.000148, thus 0.000153 on average. This partition coefficient is on the same order of magnitude as that of Qin et al. (1992), but according to our knowledge this is the first attempt to constrain D quartz/melt H 2 O based on coexistent natural mineral -silica glass pairs. The new calculated water contents (using this new partition coefficient) for the BFVA and KH eruption magmas is presented in Table 3.
All the measured samples originating from the fallout deposits of both BFVA (E-O-S and TD-BG) and KH eruptions (KH1 and KH2) represent the first fallout event of a largescale Plinian eruption. In the BFVA both sampled pyroclastic layers are mantling a paleosol (representing a dormant period in volcanism of a maximum of a few tens of kiloyears; Solleiro-Rebolledo et al., 2003). Similarly, sampling of the KH eruption fallout deposits was carried out from their basal part representing the first phase of the eruption (fed by highly differentiated T1 and T2 rhyolitic magma; Nairn et al., 2004). As was suggested earlier, the first phase of silicic eruptions used to be characterized with the highest SiO 2 (76 wt %-77 wt %) content and with the largest magmatic H 2 O content (5 wt %-6.5 wt %; e.g., Blake, 1984;Anderson et al., 2000;Wallace et al., 2003;Nairn et al., 2004;Graeter et al., 2015). This can be explained by density-driven differentiation in silicic magma chambers: highly evolved/fractionated melts accumulate in the upper parts of a magma chamber, having the highest volatile content (including, of course, water contents, about 6 wt %-7 wt %; Blake, 1984;Graeter et al., 2015). This water-rich melt part of the magma chamber is tapped first during eruption (Nairn et al., 2004).

Further implications
A study found that mafic arc magmas contain about 4 wt % dissolved water, based on investigations of melt inclusions trapped in volcanic phenocrysts . The identical water content of mafic arc magmas was explained by two things: (1) water contents are limited by the vapor saturation at the last storage level by degassing until ca. 4 wt % is reached, which seems to be optimal for the P and T conditions in arc settings, at the depth of magma stalling in the crust , and (2) mantle-source melting processes themselves limit the water content of the host melt, which maintains constant water content at mantle temperatures . The second scenario assumes a deeper magma chamber (10-12 km), where the melt stalls prior to the eruption, and the eruption is fed directly from that level . These models were developed only for mafic magmas based on the measured water content of olivine-hosted melt inclusions. In the case of silicic caldera-forming volcanism, however, the second scenario does not seem to be a realistic option (the uppermost magma chamber at 10 km depth), because in silicic systems the shallowest stalling level of magma is usually located at 2-7 km depth (Blundy and Cashman, 2001;Gualda et al., 2012Gualda et al., , 2018Graeter et al., 2015;Pamukçu et al., 2020). In the case of KH eruption 6.6 km depth was proposed (Nairn et al., 2004). As was mentioned before, the main crystallization episode of the quartz also takes place at such shal-low levels of the magmatic plumbing system (Graeter et al., 2015;Tollan et al., 2019;Jollands et al., 2020b;Pamukçu et al., 2020); thus we suggest a potential crustal water content "regulator" in the case of the BFVA (i.e., scenario 1 of , limiting the water content of the melt to ∼ 5 wt %-7 wt %. The melt was in equilibrium with the structural hydroxyl content of the crystallized quartz phenocrysts (to the measured ∼ 10 wt ppm). This means that the physicochemical conditions (P -T , oxygen fugacity, stress, etc.) may have been fairly uniform at the shallowest level of the magma plumbing system just before the eruption at several silicic magmatic provinces developed in an extension-dominated regime. A possible explanation may be the rheological features of the crust, which can be functioned as precursors for the depth and pressure conditions where magma will accumulate (Huber et al., 2019). As was suggested in recent studies (e.g., Huber et al., 2019), the optimal pressure conditions for the formation of a subvolcanic silicic magma chamber are in between 1.5 and 2.5 kbar, representing a ∼ 4 to 7 km depth interval. This is in good agreement with the suggested depth for the KH eruption magma chamber (6.6 km) and may also suggest a similar depth for the BFVA eruptions. Interestingly, these depths are consistent with the depth where the water saturation occurs based on experimental calculations of Zhang et al. (2007). Thus, in this depth range large-scale bubble formation can start due to water saturation and exsolution, which can contribute significantly to triggering an eruption (Zhang et al., 2007). Consequently, the pressureand-depth range can be estimated for the equilibrium between quartz and melt based on our fairly uniform structural hydroxyl (and thus magmatic water) contents. Possibly, this crustal region was the last stall of the magma prior to the eruption.
Here we suggest that the "crustal regulator" of magma water content may be a general phenomenon in the Earth's upper crust in case of silicic magma plumbing systems (especially in an extensional back-arc setting such as the BFVA and TVZ), able to limit the explosive potential of the silicic magmas (as assumed by . As Blake (1984) and Zhang et al. (2007) suggested, the oversaturation of water in the abovementioned pressure conditions can be a general worldwide phenomenon, which eliminates the possibility of further volatile enrichment of the magma, thus limiting the explosivity of the eruption (excluding other internal (e.g., CO 2 content) or external (e.g., the presence of surface water driving phreatomagmatic eruptions) factors). To substantiate this hypothesis, more investigation is necessary in silicic volcanic fields, focused on the water content of the melt, whereas structural hydroxyl content of quartz phenocrysts could be a good proxy for that.

Conclusions
The structural hydroxyl content of volcanic quartz phenocrysts was measured with unpolarized FTIR methodology. The quartz phenocrysts were separated from five pyroclastic fallout horizons from the Bükk Foreland Volcanic Area (northern Hungary) and from the first phase of the AD 1314 Kaharoa eruption (TVZ, New Zealand). The results of the study can be summarized as follows.
2. Structural hydroxyl content of quartz does not depend on either the geographic or stratigraphic position of the hosting pyroclastic fallout horizon. Quartz phenocrysts from previously correlated pyroclastic fallout deposits show nearly the same structural hydroxyl content regardless of the thermal and/or pressure impact of the covering deposit.
3. The calculated structural hydroxyl contents fit well in the range of earlier published values from quartz phenocrysts derived from different geological environments (granitic bodies, metamorphic rocks, detrital grains), including volcanic quartz phenocrysts.
4. The calculated magmatic water contents (around 6 wt % H 2 O) also fit well with earlier published water contents, especially in the case of the early explosive phase of Plinian eruptions of highly evolved felsic magma (e.g., the KH eruption).
5. The structural hydroxyl contents are nearly the same in quartz phenocrysts of pyroclastic fallout deposits in the BFVA. Hence, it can be assumed that P , T , and x (mainly H 2 O) activity could be similar in the uppermost (i.e., shallowest) storage level of the magmatic plumbing system, where quartz phenocrysts crystallized and chemically equilibrated with the melt. This geodynamic and petrologic environment thus constrained the structural hydroxyl content of the quartz phenocrysts, i.e., the water content of the host magma, which could have acted as an upper crustal-level regulator from a volcanic explosivity point of view.
6. The similarity of the average magmatic water content in the studied pyroclastic deposits is pointed out throughout the large-scale silicic caldera volcanism of both BFVA and TVZ (KH eruption), which could be a common feature in volcanic areas located in similar geodynamic environments. More occurrence areas of largescale silicic explosive volcanism worldwide need to be involved in future studies to prove this hypothesis. This could probably be the pressure-depth level at which the water saturation is achieved and water exsolution begins and triggers large-volume Plinian-type explosive eruptions.
The preliminary results underline the importance of measuring water contents of NAMs in fallout pyroclastic layers originating from extension-related silicic explosive eruptions in various volcanic fields.

Appendix A: Single-crystal preparations
To more accurately analyze the average unpolarized absorbance of our phenocrystal population, we crushed some quartz phenocrysts (from Eger site) manually in a mortar to be able to reproduce the methodology of Biró et al. (2017), who made the measurements on crushed (or fragmented) grains instead of separated euhedral or subhedral grains (Table 1; E-1 in Biró et al., 2017). The sample preparation methodology was almost the same as described above; however, at the end of the preparation the glue slab was dissolved with acetone, and the single phenocryst wafers were individually picked up and stored. The prepared phenocrysts were measured with the same unpolarized micro-FTIR methodology, but the single crystals were emplaced on a CaF supporting slide during the measurements.
The measurements resulted in higher structural hydroxyl content: 11.5 (±2.4) wt ppm (Supplement File S3). The measured maximal indicative structural hydroxyl content is 15.5 wt ppm. This could be the theoretical maximal indicative structural hydroxyl content of the quartz phenocrysts of the BFVA if the incident light travels parallel to the c axis. In this case, it follows from the absorbance indicatrix theory of quartz Biró et al., 2017) that the average unpolarized absorbance should be 66 % of this value, which is 10.3 wt ppm. This agrees well with the empirically determined 11.5 wt ppm. Hence, the reason for the difference between the results of this investigation and the measurements of Biró et al. (2017) is due to the different sample preparation methodology including the shape of the separated phenocrysts (fragmented vs. euhedral/subhedral crystals; see Sect. 3.2) Data availability. Our datasets are available in the Supplement.
Author contributions. MH wrote and prepared the original draft, performed the sampling in the BFVA, sample preparation, measurements, and calculations. KN carried out the sampling in New Zealand. TB, IJK, AS, and KN carried out the conceptualization and reviewed and edited the manuscript. RS played a role in methodological specifications, investigations, and editing of the manuscript. ZsP made additional calculations for this methodology and reviewed the manuscript. ZoP and DK designed and concep-tualized the work. IJK and DK also provided the financial background.
Competing interests. Some authors are members of the editorial board of European Journal of Mineralogy. The peer-review process was guided by an independent editor, and the authors have also no other competing interests to declare.
Disclaimer. Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Special issue statement. This article is part of the special issue "Probing the Earth: reviews of OH groups in anhydrous and hydrous minerals". It is not associated with a conference.
Acknowledgements. Volcanological research of Tamás Biró was supported by the ÚNKP-20-4 New National Excellence Program of the Ministry of Innovation and Technology from the source of the National Research, Development and Innovation Found (ÚNKP-20-4-II-ELTE-32). Volcanological research of Mátyás Hencz was supported by the ÚNKP-20-3 New National Excellence Program of the Ministry for Innovation and Technology from the source of the National Research, Development and Innovation Fund (ÚNKP-20-3-II-ELTE-102). The research was supported by the European Union and the state of Hungary, co-financed by the European Regional Development Fund in the project GINOP-2.3.2.-15-2016-00009 "ICER". Assistance of Szabolcs Kósik, Tamás Sági, and Sándor Józsa during the sampling; Dóra Kesjár during the FTIR measurements; and Zoltán Cseri in figure drawing is highly acknowledged.
We thank Mike Jollands and Hélène Balcone-Boissard for the in-depth review which improved the manuscript. We thank Etienne Balan for handling the manuscript.
Financial support. This research has been supported by the Hungarian Scientific Research Fund (grant nos. K115472, K128122, and K131894).
Review statement. This paper was edited by Etienne Balan and reviewed by Michael Jollands and Hélène Balcone-Boissard.