Quartz is formed under nearly all conditions realised within the Earth's crust, ranging from the Earth's surface down to the base of the continents. Depending on crystallisation-relevant physical and chemical parameters (pressure, stress, temperature, temperature gradients, time, presence and composition of melts and fluids) and geological circumstances (cracks and voids in the surrounding rock), the resulting habit and size of quartz crystals is very diverse, ranging from microcrystalline fibrous chalcedony to decimetre- and metre-sized euhedral crystals. Although both crystal size extremes are not representative of the main portion of quartz as a rock-forming mineral (12 vol % of the Earth's crust; Ronov and Yaroshevski, 1969), they are very prominent in our mind and important to human history because they were used since prehistoric times as precious stones (such as agate, amethyst, and rock crystal) or for tools and weapons (such as flint and rock crystal). Owing to the high appeal to the human eye and the technical applicability in modern times (Jung, 1992a, b), it is understandable that the first systematic studies on “water” in quartz were performed on large hydrothermal (natural or synthetic) crystals. The major portion of quartz, however, forms as millimetre-sized, often anhedral crystals in igneous systems (50 % of the Earth's continental crust consists of felsic igneous rocks (Wedepohl, 1995) that in turn consist of >20 % quartz). Furthermore, as a hard and chemically resistant mineral, quartz survives weathering and transport and comprises a significant fraction in siliciclastic sediments, sedimentary rocks, and their metamorphic counterparts.

This review summarises early findings on “water” in quartz and its influence on physical properties, followed by a review of theoretical and experimental considerations on OH point defects, and finally reports recent data on OH point defects in common quartz as a rock-forming mineral from igneous rocks, metamorphic rocks, and sedimentary archives. As will be outlined in this study, molecular water follows different incorporation laws than OH point defects, and, consequently, the wealth of studies on fluid inclusions in quartz, as well as on OH species (molecular water and OH point defects) in microcrystalline and/or amorphous silica such as opal and chalcedony (Langer and Flörke, 1974; Frondel, 1982; Adams et al., 1991; Chauviré et al., 2017) and moganite (Flörke et al., 1984; Heaney and Post, 1992; Hardgrove and Rogers, 2013), are not considered in this review.

The existence of a significantly polarised absorption band at 3400 cm-1 with a more pronounced absorption for E||o than E||e has been known since the end of the 19th century (Merritt, 1895) but was not linked to a specific impurity until several decades later. In the first half of the 20th century quartz became more and more interesting for the upcoming electronic industries (e.g. chronometry and radio engineering), and research on physical properties and chemical impurities of (preferentially large) natural and synthetic quartz crystals was promoted. The analysis of “water” in quartz reached its first prime in the early 1960s (Bambauer, 1961; Brunner et al., 1961; Kats, 1962; Bambauer et al., 1962, 1963), focussed on detailed investigations on the spatial distribution of chemical impurities, hydrogen mobility, charge balance, and assignment of infrared (IR) absorption bands to specific OH defects in natural hydrothermal quartzes. Although a distinction between different species such as molecular water and OH point defects (protons that are charge balanced by metal impurities or vacancies) was made, both species were often subsumed as water. The research on water in quartz was further fuelled by the discovery of hydrolytic weakening (Carter et al., 1964; Griggs and Blacic, 1965; Griggs et al., 1966) and its importance for structural geology. In the following decades research on OH in quartz experienced several flares accompanying technical improvements of microanalysis, such as secondary ion mass spectrometry (Rovetta et al., 1989), transmission electron microscopy (Gerretsen et al., 1989; Cordier and Doukhan, 1991), and calculation capacities (Purton et al., 1992; McConnell et al., 1995; Rosa et al., 2005). Predominant study objects during these periods still were large hydrothermal (natural and synthetic) single crystals. It was not until the early 21st century that common quartz – representing the main portion of quartz as a rock forming mineral – came into focus, and systematic studies on metamorphic (Müller and Koch-Müller, 2009), sedimentary (Stalder and Neuser, 2013; Stalder, 2014; Stalder et al., 2017, 2019; Jaeger et al., 2019), volcanic (Biró et al., 2016, 2017), and plutonic (Müller et al., 2009; Stalder et al., 2017; Potrafke et al., 2020) quartz grains were performed.

Impurities in quartz can be incorporated by a plethora of cation exchanges, either by the simple exchange of Si4+ by other tetravalent cations such as Ti4+ or Ge4+ or by coupled substitutions involving monovalent cations such as alkalis or protons. Within the scope of this work, defects involving protons are called “hydrous defects”, and those without protons are called “dry defects”.

Even before quartz was systematically investigated for impurities and chemical zoning, it was known that β-eucryptite (LiAlSiO4) is isostructural to high quartz (Winkler, 1948; Cohen, 1960; London, 1984) and that a metastable solid solution series exists in the system SiO2 – LiAlO2 (Roy and Osborn, 1949), indicating the possibility of a “dry” AlLi defect through the exchange Si4+=Al3++Li+. Consequently, the first charge balance equations involving incorporation of defect protons into the quartz structure considered Al3+ and alkali cations as the most relevant impurities (Bambauer, 1961; Kats, 1962, Chakraborty and Lehmann, 1976a). Other tri- and pentavalent cations were subsequently added to the equation, suggesting a charge balance equation, 1[Al,B,Fe3+]=[H,Na,K,Li,P], for natural samples (Müller and Koch-Müller, 2009) that later was confirmed for experimentally grown quartz crystals from natural starting material (Baron et al., 2015; Potrafke et al., 2019, 2020), while a modified equation for small Li contents and crystallisation at moderate pressure (<10 kbar) was proposed [Al,B,Li]=[H,Na,K,P], taking into account a neutral LiOH species (Frigo et al., 2016).

Large hydrothermal quartz crystals are very heterogeneous with respect to OH (Kats, 1962), and the highest concentrations are often found in the centre of the crystal (Chakraborty and Lehmann, 1976a). A similar systematic zoning from core to rim was observed for metal impurities such as Al3+ (Müller et al., 2003; Miyoshi et al., 2005), suggesting that metal impurities are linked to OH defects and that the impurity incorporation is controlled by the fluid chemistry and growth rate. Strong chemical zoning was recently also discovered within individual phenocrysts from the Mesa Falls Tuff pyroclastic succession (Tollan et al., 2019) and from Bishop Tuff (Jollands et al., 2020b), where a decrease in OH from core to rim mirrored a strong zoning for Al and the formation of a “dry” LiAl defect at the outermost rim by the exchange of H by Li.

As a chemical component, water in quartz is hosted in at least two totally different ways: (1) molecular water such as fluid inclusions and (2) OH point defects, in which protons are charge balanced by vacancies or metal impurities. Further hosts of water are micro inclusions of melts and hydrous minerals such as mica (Stalder and Neuser, 2013; Kronenberg et al., 2017), as well as amorphous gel-like material (Brunner et al., 1961). The different OH species were identified and characterised by a number of different strategies, such as freezing behaviour, D/H exchange, and absorption features in the IR region. Amorphous gel-like material, for example, causes a rather broad, isotropic absorption band at 3400 or 3500 cm-1 and leads to the formation of molecular water upon heat treatment (2SiOH=Si-O-Si+H2O) that gives the crystal a milky appearance (Bambauer et al., 1969). It further does not show evidence for H/D exchange (Brunner et al., 1961) and probably is formed during low temperature processes. Such absorption bands are typical for precious varieties of quartz such as amethyst, citrin, and rock crystal (Chakraborty and Lehmann, 1976b).

It has already been repeatedly mentioned since the first studies that molecular water (mostly as fluid inclusions) is the dominant OH species in quartz (Bambauer, 1961; Aines et al., 1984; Gerretsen et al., 1989; Cordier and Doukhan, 1991; Müller and Koch-Müller, 2009; Kronenberg et al., 2017) and that these fluid inclusions in turn are strongly enriched in alkalis (Bambauer, 1961; Müller et al., 2003). Some of the molecular water is hosted in nano-inclusions (Gerretsen et al., 1989; Cordier and Doukhan, 1991) that are often too small to build ice upon freezing (Aines and Rossman, 1984; Müller and Koch-Müller, 2009). Furthermore, no correlation could be found between the amount of molecular water incorporation and OH point defects (Aines et al., 1984; Biró et al., 2016), which implies that the amount of molecular water in quartz is not a diagnostic for the formation conditions and thermal history of an individual grain. While point defect formation is controlled by thermochemical parameters (Paterson, 1986), the amount of fluid inclusions is often heterogeneously distributed even within one individual and otherwise chemically homogeneous grain. Fluid inclusions are often concentrated along healed cracks and were taken up under deformation, which in early studies led to estimates of water solubility that are too high (Gerretsen et al., 1989). Therefore, the present article will be focused on OH point defects, and molecular water will not be discussed in detail.

Proton incorporation as OH defects leads to the formation of characteristic OH dipoles that can be detected and distinguished depending on the charge compensation by their absorption frequency in the IR range of electromagnetic radiation. Since oxygen is the only charge-balancing anion in the crystal, the relevant geochemical component in mineralogical and petrological literature is often expressed as a neutral water component. Therefore, throughout this article OH contents are expressed as weight parts per million (wt ppm) water (identical to µg/g H2O), in which 0.01 wt % water corresponds to 100 ppm water, 11.1 (=100/9) ppm H, or 667 (=100/0.15) H/106Si.

Although several different instrumental methods to measure OH in nominally anhydrous minerals has been established, the by far most used technique to analyse OH in quartz is IR spectroscopy that detects absorption of IR radiation due to vibrations in the respective material. Different structural environments differ in their bond strength, length, and orientation and thus have different vibration properties. Based on the absorption band characteristics such as vibration frequency, anisotropy, and band sharpness (Fig. 1a), IR spectroscopy is able to distinguish between different OH species, such as (1) molecular water, (2) nominally OH-bearing minerals, and (3) several specific OH point defects. Since the absorbance is linearly correlated to the concentration (Eq. 2), the sensitivity depends on the sample thickness, and for 200 µm thick sections of crystals a detection limits down to the parts per million (ppm) level can be reached. For thicker samples even lower concentrations can be detected. Analysis can now be performed on surfaces below 0.001 mm2 (compared to 10 mm2 in early studies such as Brunner et al., 1961), enabling the possibility to reveal micrometre-sized colourless inclusions (Stalder and Neuser, 2013) and complex internal zonings (Potrafke et al., 2020). In order to turn IR absorptions into concentrations, OH contents have to be determined by an independent method on reference material, and in the past 6 decades different methods were used, such as (1) H-alkali exchange via electrolysis (Brunner et al., 1961), (2) secondary ion mass spectrometry (Rovetta et al., 1989; Thomas et al., 2009), (3) proton–proton-scattering (Thomas et al., 2009), and (4) the theoretical calculation of extinction coefficients (Balan et al., 2008; Jollands et al., 2020a). The formulations deviate from each other with respect to the measurement protocol (unpolarised versus polarised) and in terms of the resulting units (concentrations expressed as wt ppm water or H/106Si), and thus different integral extinction coefficients ε (Eq. 2) were obtained, which finally all lead to the similar results: the mineral-specific calibration of Thomas et al. (2009) determined a value for the extinction coefficient as ε=89000±15000 L mol-1 cm-2, which is in excellent agreement with ε=246.6(3753-ν) of Libowitzky and Rossman (1997) that would receive the same value for ε for a wavenumber ν=3390 cm-1. As the main absorption band of OH in quartz is exactly in this region for most quartz specimens, both calibrations obtain nearly identical results. Only in a few cases in which high-wavenumber bands are over-represented (Stalder and Konzett, 2012) does the wavenumber-dependent calibration lead to systematically higher OH concentrations. Considering that OH dipoles are nearly totally polarised ||no (perpendicular to the c axis) and using the Lambert–Beer equation, 2ac=At⋅ε, with the concentration c, the extinction coefficient ε, and the thickness normalised absorbance A/t, Thomas et al. (2009) would receive for one polarised measurement ||no the relation 2bc(ppmH2O)=At⋅2⋅MH2O⋅1000DQz⋅89000=At⋅0.153, with the molar mass of water M=18 g mol-1, the density of quartz D=2.65 g cm-3, and the thickness t in centimetres. The factor 2 represents the twofold contributions of no (out of the three directions in space), and the factor 1000 is obtained from the conversion of litres to cubic centimetres (cm3). The obtained value is equivalent to 2ccH106Si=At⋅1.02, while Aines et al. (1984) receive only very slightly higher values: 2dcH106Si=At⋅1.05. Taking into account the different measurement protocols and different units, the extinction coefficients of ε=14000 cm mol-1 used by Brunner et al. (1961), Bambauer (1961), and Chakraborty and Lehmann (1976a) would be recalculated to the sixfold value ε=84000 L mol-1 cm-2, which is in excellent agreement with the other calibrations. The mineral-specific calibration of Rovetta et al. (1989) leads to 30 % higher OH contents. There is an ongoing debate whether a mineral- or wavenumber-specific calibration is more reliable. On the one hand, a pronounced dependence of the extinction coefficient on the wavenumber for OH dipoles in quartz was theoretically predicted by density functional theory (DFT) calculations (Balan et al., 2008; Jollands et al., 2020a), and a nearly linear correlation has been observed for OH dipoles in hydrous minerals (Libowitzky and Rossman, 1997). On the other hand, a less pronounced dependence was observed for different OH absorption bands in quartz in the calibration of Thomas et al. (2009), and finally, one mineral-specific extinction coefficient of 89 000 L mol-1 cm-2 was proposed. While the theoretically calculated extinction coefficients (Balan et al., 2008; Jollands et al., 2020a) are significantly higher than suggested by the analytical calibrations and would lead to lower OH concentrations by up to a factor a 2, the general wavenumber-specific (and not quartz-specific) calibration of Libowitzky and Rossman (1997) leads nearly to identical OH contents for most samples compared to the mineral-specific calibrations of Brunner et al. (1961) and Thomas et al. (2009). Since all calibrations finally obtain very similar OH contents, no a posteriori corrections concerning the originally published values were made in the data compilation (Table 1) of published OH contents in quartz. A more complicating factor for the comparison of published data from different references is the circumstance that some studies did not make a distinction between molecular water and OH from point defects, leading to higher OH for some samples, especially those in which molecular water was an important species. Since molecular water gives rise to broad isotropic absorption bands and hydrous defects are nearly perfectly polarised ||no (Brunner et al., 1961), a distinction between both fundamentally different groups of species was made early by only taking into account the sharp polarised absorption bands (Bambauer et al., 1962) or by the separate quantification of point defects and molecular water (Müller and Koch-Müller, 2009; Biró et al., 2016). Another strategy to overcome the contribution of non-defect OH (actually originally proposed by Brunner et al., 1961) is a revised protocol for the treatment of IR absorption spectra (Stalder and Konzett, 2012), in which all measurements are performed on oriented sections ||c for both crystallographic directions (E||no and E||ne) on the same spot. Subtraction of both spectra (E||no-E||ne) eliminates the isotropic contributions of molecular water, hydrous melts, and randomly oriented mica inclusions (Fig. 1a). For the BOH defect that exhibits one third of its absorbance ||ne, the proposed protocol was later refined (Stalder and Neuser, 2013; Baron et al., 2015). This strategy is in general in accordance with many previous studies, in which all measurements were performed ||no (Bambauer, 1961) and/or only the contribution ||no was used for OH quantification (Brunner et al., 1961; Aines et al., 1984). Besides the OH absorption bands, IR spectroscopy also reveals information on the orientation and thickness of a sample by using lattice overtones (Biró et al., 2016; Stalder et al., 2017; Jollands et al., 2020b) and offers an independent alternative to mechanical measurements of the sample thickness that is needed to calculate concentrations from the measured absorbance (Eq. 2).

Several dozens of OH absorption bands in quartz have been reported in the literature (for a summary see Aines and Rossman, 1984) and linked to specific defect species based on chemical correlations, thermal stability, H/D exchange experiments, and spectral characteristics such as sharpness and polarisation. It is understandable that reported band positions deviate from each other by some wavenumbers due to different spectrometer calibrations and different temperatures during analysis (e.g. room and liquid nitrogen temperature). A particular example of a band shift is reported for the most prominent AlOH band at 3378 cm-1, which is shifted to 3395 cm-1 at 550 ∘C and drops to 3386 cm-1 at the α/β transition of quartz at 573 ∘C (Suzuki and Nakashima, 1999). During cooling from room temperature to liquid nitrogen temperature, the AlOH band is shifted by 15–20 cm-1 towards lower wavenumbers, while the LiOH band at 3470 cm-1 (at room temperature) is only shifted by 10 cm-1 (Brunner et al., 1961; Suzuki and Nakashima, 1999). Therefore, wavenumbers given below may deviate slightly from the values reported in the original publications.

Apart from the aforementioned broad and isotropic absorption feature for molecular water, the most important OH absorption bands are strongly polarised and have been assigned to specific metal impurities. The by far most important metal impurity is Al3+, followed by Li+ and B3+.

Several attempts have been made to model OH defect species based on thermodynamics (Doukhan and Trépied, 1985; Paterson, 1986; Rovetta, 1989) and ab initio calculations (Purton et al., 1992; McConnell et al., 1995; Rosa et al., 2005; Jollands et al., 2020a). Until recently calculations were strongly focussed on the two concurring intrinsic defects: hydrogarnet defect [4H]Si and the so called Griggs defect according to the reaction Si-O-Si + H2O = 2 SiOH (Griggs et al., 1966). There is consent that [4H]Si concentrations are positively correlated with pressure up to 10 kbar (Doukhan and Trépied, 1985; Paterson, 1986) at oxygen fugacities above NNO-2. As to temperature, both negative (Doukhan and Trépied, 1985) and positive (Paterson, 1986) correlations were predicted. Assuming a [4H]Si formation by the reaction of the dissociated species (Paterson, 1986), 32H2 (gas)+O2 (gas)+SiO2 (crystal)=H4O2 (crystal)+SiO2 (surface), where H4O2 (crystal) represents the [4H]Si defect. It was predicted that [4H]Si concentrations drop to nearly zero below oxygen fugacities of NNO-6 (Paterson, 1986). Similarly, thermodynamic models for the Griggs-type SiOH defect calculated for pressure conditions at 15 kbar suggest increasing concentrations with decreasing oxygen fugacity and a positive correlation to temperature (Rovetta, 1989) and maximum OH concentrations corresponding to up to 300 wt ppm water. Ab initio calculations (using density functional theory, DFT) suggest that (i) [4H]Si is energetically the most favourable intrinsic defect species (Purton et al., 1992; McConnell et al., 1995; Rosa et al., 2005), which (ii) converts to molecular water during the development of dislocations (McConnell et al., 1995), and (iii) that the Griggs defect is unstable in unstrained quartz (Purton et al., 1992).

The incorporation of OH species influences the physical properties in different ways: (1) the high mobility of protons enhance the electrical conductivity by acting as charge carrier, and (2) protons modify the chemical bonds and thereby the mechanical properties of the silica network itself. The electrical conductivity depends on the concentration and mobility of mobile species and can be described by the Nernst–Einstein equation: 4σ=∑iDicizi2kBT, where σ is the bulk electrical conductivity, kB is the Boltzmann constant, T is the absolute temperature, and Di, ci, and zi are the diffusion coefficients, the concentration, and the charge, respectively, of the mobile species i. In quartz, the principle charge carriers are protons and thus crucial for long-term DC (direct current) conductivity (Kronenberg and Kirby, 1987). In order to calculate the electrical conductivity according to Eq. (4), the diffusivities of protons from different OH species (H not linked to Al, and H linked to Al) have to be known.

A number of different strategies to determine H diffusivities have been followed, such as hydrogen uptake (Shaffer et al., 1974; Kronenberg et al., 1986), hydrogen extraction (Bachheimer, 1998), H/D exchange (Brunner et al., 1961; Kats, 1962; Rovetta et al., 1986), and H/Li exchange (Jollands et al., 2020b). All methods represent different processes with different rate-limited steps. In addition, diffusion experiments were performed with different starting materials (although mostly with natural hydrothermal rock crystals with AlOH > LiOH). Consequently, they reflect different diffusion mechanisms, and the results spread by several log units at a given temperature (Fig. 2): the H/D inter-diffusion coefficient is a good approximation for the self-diffusion of H; in contrast, H uptake or extraction can be rate-limited by other diffusivities depending on the charge balance it is based on and may in fact be slower by orders of magnitude. The hitherto reported H/D exchange experiments above 700 ∘C all fall into a narrow range (Fig. 2) and exhibit a rather high activation energy around 200 kJ/mol. Experiments at lower temperatures reveal a marked drop in the activation energy to approximately 70 kJ/mol (Kats, 1962). This remarkable change has been interpreted as the transition between two regimes, in which the diffusing species is dominated by H not linked to Al at low temperatures in contrast to the results at high temperatures, in which H is predominantly linked to Al (Kronenberg and Kirby, 1987). H uptake experiments by Kronenberg et al. (1986) show a similar diffusion law as H/D exchange until the charge neutrality [Al] = [H] is reached, which suggests that the diffusion probably relies on a similar mechanism to H/D exchange. IR spectra revealed a defect ratio AlOH > LiOH at the beginning of the experiment, and AlOH >> LiOH (with nearly no LiOH) at the end of the experiments. This finding insinuates that to some extent H/Li exchange (and H exchange with other alkalis) occurred, but the majority of the protons incorporated during the experiment has to combine with the existing AlSi′ defects (a dry defect, in which Al is on a tetrahedral position and exhibits a local charge deficit) to an AlOH defect. It is furthermore noteworthy that initial samples used for H extraction (Bachheimer, 1998) reveal IR absorption spectra that are similar to the hydrated sample of Kronenberg et al. (1986), but H extraction still does not seem to follow the reverse process of H uptake. Some of the differences between these two studies may partly be explained by missing exchange options by alkalis in the extraction experiments. Interestingly, the activation energies are similar, but diffusivities for H extraction are shifted towards lower values by 3 log units (compare H uptake and dehydration in Fig. 2). For crystals without other defects than those that follow the charge neutrality [Al] = [H], a lower diffusivity for H extraction than from self-diffusion is expected since another – negatively charged – species is involved. A possible candidate for this negatively charged species is oxygen, which is several orders of magnitude slower. Other explanations for the discrepancies between H uptake and dehydration are the redox behaviour of trace amounts of Fe (not reported in these studies) or pre-existing other defects.

For the H/Li exchange process in quartz a lower activation energy of 100 kJ/mol was recently determined (Jollands et al., 2020b), which is rather similar to the low temperature H/D exchange data of Kats (1962) and in accordance with the interpretation that the diffusing H species is not linked to Al (Kronenberg and Kirby, 1987). H/Li exchange is influenced by the Li mobility. Estimates for Li diffusion in quartz yield similar to slightly higher values than for H (Charlier et al., 2012). Considering the inter-diffusivity, 5DH-Li=XH+XLiDHDLiXHDH+XLiDLi, with D= diffusivity and X= mole fraction, the mismatch between the observed H mobility in H/Li exchange ||c (Jollands et al., 2020b) and H/D exchange experiments < 600 ∘C ||c (Kats, 1962) by up to a factor of 3 (half a log unit) may be explained (Fig. 2). Taking into account that the samples from both studies differed strongly with respect to absolute H and Li content, diffusion can be considered concentration independent and DH and DLi very similar.

No significant dependence of H mobility on pressure (Kronenberg et al., 1986) and oxygen fugacity (Jollands et al., 2020b) was observed. Similarly, no pronounced anisotropy (difference between diffusivity ||c and ⊥c) was found for H uptake and H/D exchange (Kronenberg et al., 1986) and H extraction (Bachheimer, 1998); the effect for H/Li exchange was not quantified with regard to crystal orientation in these studies, but a slower diffusivity for diffusion ⊥c than ||c was reported (Jollands et al., 2020b).

The presence of defect protons has an indirect impact on other physical properties. In this context it has been shown that the incorporation of protons enhance the diffusivity of O within the crystal lattice and thus indirectly influence the mechanical properties of quartz (Elphick and Graham, 1988). After the first observation of hydrolytic weakening of quartz (Griggs and Blacic, 1965; Griggs et al., 1966), the nature of the relevant OH species assumed to facilitate the deformation was extensively debated. The discussion primarily concentrated on intrinsic defects such as SiOH, while AlOH as a chemically very stable defect was not considered as a relevant species for hydrolytic weakening (Cordier and Doukhan, 1991). The initial model assumed that weakening was caused by the formation of Griggs-type defects by the following reaction (Griggs et al., 1966; Aines et al., 1984): 6Si-O-Si+H2O=Si-OHHO-Si. Later on it was suggested that actually hydrogarnet defects [4H]Si are the relevant species that help activating dislocations and thus facilitate hydrolytic weakening (Doukhan and Trépied, 1985; Cordier et al., 1994). Based on the observation of the precipitation of molecular water bubbles during the formation of dislocations (Cordier et al., 1988; Cordier and Doukhan, 1991), a conversion of hydrogarnet defects to molecular water during dislocation (according to Eq. 4) was suggested (McConnell et al., 1995). A transformation between different OH species was also observed during the thermal annealing of previously deformed quartz, in which recrystallised samples contained less molecular water due to the conversion into point defects (Niimi et al., 1999).

It has been shown early by hydrothermal annealing at 15 kbar and 900 ∘C (Mackwell and Paterson, 1985; Rovetta, 1989) that quartz is able to incorporate high amounts of OH (up to 150 wt ppm water) and, furthermore, that OH incorporation is enhanced at low oxygen fugacities (Rovetta, 1989). The interpretation of incorporation as point defects was soon challenged, and it has been argued that OH is mainly hosted in fluid inclusion (Gerretsen et al., 1989). Experimental research activities in the following years were focussed on the role of OH during deformation of quartz and studying OH in other nominally anhydrous minerals, and relevant data from experimental petrological studies on OH in quartz, which were applicable to granitic systems, were not published until the beginning of the past decade. By using a new analytical protocol (Stalder and Konzett, 2012), the isotropic contribution of molecular water (and species in melts and other amorphous material) could be erased from the IR absorption spectrum, and the contribution from OH point defects could be determined even in sub-millimetres, inclusion-rich crystals from high-pressure experiments.

Synthesis experiments in simple systems allow for the assignment of OH bands to metal impurities by investigating pure systems and systems doped with the respective metal ions. This allowed us to verify the nature of intrinsic defects (in the following referred to as hydrogarnet) and impurity-specific defects in particular, and the substantial influence of B on OH incorporation was shown (Fig. 1).

Quartz from high-pressure experiments in granitic systems exhibit OH defect contents between 30 and 600 wt ppm water, depending on the pressure and composition of the system (Fig. 3). The influence of crystallisation temperature is not very well constrained because melt viscosities and the P-T circumstances in the phase diagrams for realistic scenarios set rather narrow limits for successful experiments which are expected to provide sufficiently large quartz crystals for further analysis. However, no clear temperature trend on OH incorporation was established in a haplogranitic system (Stalder and Konzett, 2012). Between 1 and 5 kbar, pressure seems to have a rather small systematic influence (Potrafke et al., 2019), except for a narrow region at 4 to 4.5 kbar close to the high/low quartz transition in which a sudden increase in OH and metal impurities was observed. Between 5 and 25 kbar, a negative pressure trend (Fig. 3) was observed in each studied system (Stalder and Konzett, 2012; Baron et al., 2015; Frigo et al., 2016), leading to the suggestion that the most pure quartz (inclusive of the charge-balancing metal ions) is formed at the quartz/coesite transition (Frigo et al., 2019). IR spectra recorded on quartz from high-pressure syntheses in granitic systems are usually dominated by the AlOH triplet (Fig. 1c). If the system is saturated with tourmaline (Baron et al., 2015) and/or Li (Frigo et al., 2016; Potrafke et al., 2019), specific BOH and LiOH are observed in addition. More detailed studies (Stalder and Konzett, 2012; Baron et al., 2015) observed a negative pressure trend for AlOH that may be explained by the unfavourable size of Al3+ in the tetrahedral site at high pressure (Potrafke et al., 2019). Furthermore, LiOH is strongly decreased with pressure especially around the high/low quartz transition (Frigo et al., 2016), while BOH does not show a clear pressure trend (Baron et al., 2015). These observed trends lead to the situation that at low pressure the linear absorbance of the LiOH band may nearly reach the absorbance of the AlOH main band (Potrafke et al., 2019), and at high pressure the BOH band may even exceed the linear absorbance of the main AlOH band in some cases (Baron et al., 2015). The hydrogarnet defect shows a positive pressure dependence, possibly culminating in a maximum at 20–25 kbar (Stalder and Konzett, 2012), but it never reaches the absorbance of the AlOH band in granitic systems.

In contrast to experimentally grown quartz from granitic systems (Baron et al., 2015; Frigo et al., 2016; Potrafke et al., 2019) in which high OH defect contents ranging between 30 wt ppm and 600 wt ppm water were observed (Fig. 3), natural quartz from granitic and hydrothermal systems show large variation and median values around 20 ppm water (Table 1). Average crustal quartz represented by siliciclastic sediments hosts only 10 wt ppm water as OH defects (Stalder, 2014), while metamorphic and volcanic samples cluster around 5 wt ppm (Table 1). The large discrepancy between natural and experimentally grown quartz from water-saturated systems may have several reasons, such as (1) reduced water activity during crystallisation in natural systems leading to reduced OH incorporation (Stalder and Konzett, 2012) and (2) metamorphic overprint leading to partial or total OH loss (Stalder et al., 2017). In accordance with the high-pressure experiments, the highest reported values in natural quartz closely reach values above 200 wt ppm water (Bambauer et al., 1962), and even in the sedimentary archives a handful of grains with OH defect contents >100 ppm water are preserved (Stalder, 2014; Jaeger et al., 2019, see Table 1). Interestingly, most of the OH-rich grains from sediments that retained a high OH defect content are dominated by AlOH defects, which is in contrast to the late-stage igneous samples with high OH contents (Fig. 1d). Possibly, these grains underwent very low-grade thermal treatment after formation (and before sedimentation) that led to a transformation of the LiOH to AlOH absorption bands, comparable to thermal annealing experiments on quartz (Brunner et al., 1961; Rovetta et al., 1986; Stalder et al., 2017).

A combination of the different approaches to study OH point defects in quartz enables a link between understanding the incorporation behaviour depending on physical and chemical conditions, the evolution of igneous systems, and the recycling in sedimentary systems. Ideally, clusters of sedimentary quartz grains can be identified and then traced back to their possible sources and can give insights to their geological past. In detail the link may not be unequivocal, especially if the original sources are removed by erosion. However, a rough general estimate concerning the thermal history of the source can still be made. If the origin of quartz grains is unknown, one may divide them into three large classes: (1) pristine igneous and/or hydrothermal origin with moderate to high total OH and variable LiOH/AlOH, (2) mildly thermally annealed (late-stage) igneous grains with moderate to high total OH and low LiOH/AlOH, and (3) strongly dehydrated grains with very low OH contents and low LiOH/AlOH such as metamorphic grains or grains from the middle part of pyroclastic density current deposits characterised by slow cooling. Different classes are not strictly divided and exhibit a continuous transition between each other: while group (1) covers nearly the whole range in Fig. 4, group (2) is confined to the low LiOH/AlOH edge, and (3) is confined to the low LiOH/AlOH and to the low OH corner. The very close match of OH contents of 250 wt ppm water in quartz from high-pressure experiments ≤4 kbar under water-saturated conditions in a granitic system (Potrafke et al., 2019) and the highest contents observed in natural grains (Jaeger et al., 2019) suggests that the natural processes can be realistically simulated by high-pressure experiments. Linking the sedimentary grains to quartz specimens found in situ (such as directly sampled from igneous bodies or hydrothermal veins and fissures) is not straightforward, and candidates that might be the precursors for the most OH-rich sedimentary grains were not identified yet. Basically all hitherto identified grains from sediments with very high (>100 wt ppm water) OH contents (Fig. 1d) would belong to the above-defined class (2) but otherwise cannot be linked to natural specimens found in situ, which either belong to class (1) and show a much higher LiOH/AlOH (as the mimetic quartzes; Bambauer, 1961) or have lower total OH contents (Table 1). Possibly, the OH-rich grains were originally formed with high LiOH/AlOH and were later annealed at sufficiently low temperature (and/or over a sufficiently short time) so that LiOH was destroyed, but the high OH content was preserved. With respect to the low thermal stability of LiOH the available data led to a consensus that fully explains that LiOH is absent in old plutonic bodies and metamorphic rocks.

Generally, OH incorporation is probably controlled predominantly by the availability of Li and water (provided that also Al is present), and an increase in Li leads to a decrease in LiOH by the formation of dry LiAl defects (Frigo et al., 2016), and, further, an increase in water activity leads to an increase in total OH. The observed trend on a larger scale suggests a general decrease in OH with pressure and depth – that is mainly driven by an unfavourable Al incorporation on the tetrahedral site (Stalder and Konzett, 2012; Frigo et al., 2016) – implying that quartz from large batholiths should exhibit less OH defects in their roots. Consequently, deep (late) erosion furnishes less OH-rich quartz grains compared to shallow (early) erosion (Fig. 9). This effect is further reinforced by the longer (and stronger) subjection to metamorphic overprint of the more basal portions of the plutonic bodies.

In conclusion, OH defects in quartz may be used to quantify geological processes, provided that a sufficiently large data set is present.

A statistical distribution of OH defects in metamorphic rocks could be used as a geothermometer, in which a higher degree of equilibration between the individual grains is reached at higher temperature.

Despite the progress during the past decade, several knowledge gaps persist. Further research directions could be addressed to (1) the formation and stability of hydrous defects during deformation, (2) experiments in a more narrow grid at 1–5 kbar, (3) the influence of high/low quartz transition, (4) experiments <1 kbar to verify the rapid increase in the low pressure regime and for application to volcanic rocks, (5) Li-H exchange in quartz from volcanic samples for quantification of eruption timescales, and (6) the nature of the band at 3200 cm-1.

Other hitherto unsolved questions refer to the link between IR spectroscopy and other spectroscopic methods such as cathodoluminescence (CL) and optical stimulated luminescence spectroscopy (OSL). Despite several systematic studies a clear correlation between the spectral characteristics of IR and CL could so far not be established, and both methods seem to give complementary information (Müller et al., 2009; Stalder and Neuser, 2013; Potrafke et al., 2020). Very few studies combining IR and OSL were so far performed, and the existing data set does not reveal a clear correlation between IR and OSL characteristics. Hydrous point defects show a vague but barely significant positive correlation to the intensity of the OSL signal (Thamóné et al., 2020), and for molecular water (fluid inclusions) contrasting results – positive trends (Sharma et al., 2017) and no correlation (Thamóné et al., 2020) – were reported. The fact that for these methods the ideal grain size is different – typically 90–250 µm for OSL (Rhodes, 2011) and 250–1000 µm for IR – is unfavourable for combining both methods for the same sample material.