Individual, free-grown miarolitic quartz crystals or pieces thereof from two pegmatite localities (Rosina pegmatite, Elba, Italy, and Misox pegmatite, Swiss Alps) were cut parallel to the c axis to prepare doubly polished thick sections of 200 to 500 µm thickness.

The Rosina pegmatite is a complexly and asymmetrically zoned, lithium–caesium–tantalum (LCT) pegmatite with several gem-bearing miarolitic cavities (Pezzotta, 2000). The main pegmatite dyke shows variable thickness, between 0.5 and 2 m. The Rosina pegmatite is emplaced into the San Piero facies of the Mt. Capanne monzogranite. The Mt. Capanne pluton formed through several pulses of S-type magma being emplaced between 7.3 to 7.0 Ma based on zircon U–Pb geochronology (Barboni et al., 2015), with San Piero facies being the youngest intrusion. Based on fluid inclusion studies of quartz, tourmaline, andalusite, and plagioclase in the leucogranitic dykes and pegmatite–aplite dykes in the Mt. Capanne pluton, Bakker and Schilli (2016) determined the temperature and pressure of pegmatite emplacement to be between 670 and 720 °C and 2.7 to 2.3 kbar. A sketch of Rosina pegmatite, with marked locations of studied quartz crystals, is shown in Fig. 2. We sampled three distinct miarolitic cavities in the complexly zoned Rosina pegmatite. The biggest miarolitic cavity, found in the core zone of the pegmatite, also bears the most evolved mineral assemblage, consisting of quartz, plagioclase, Li–Mn tourmalines, petalite, pollucite, and lepidolite (Orlandi and Pezzotta, 1996; Pezzotta, 2000). The second sampled miarolitic cavity is found at the border between the wall and the intermediate zone of the pegmatite body, where the assemblage is dominated by albite and quartz, with less white mica and black to pale-blue tourmaline. The third miarolitic cavity is found in a satellite vein of the Rosina main pegmatite body and has spessartine present in the assemblage.

The pegmatites of Misox valley, located in the Ticino region of Switzerland, are emplaced in the Simano Nappe of the Lepontine dome in the Central European Alps. The Lepontine dome exhibits a metamorphic gradient from migmatites in the south to lower amphibolite–greenschist facies in the north. The peak temperature reported for the migmatites along the southern margin of the dome is at ca. 700±50 °C (Burri et al., 2005). This peak temperature metamorphism lasted from around 32 to 22 Ma, shown by allanite (Gregory et al., 2012) and zircon (Rubatto et al., 2009) ages. These ages and distribution of zircon overgrowths suggested multiple melting episodes driven by fluid-assisted and muscovite dehydration partial melting at around 700±50 °C and 6 to 8 kbar (Burri et al., 2005; Berger et al., 2008; Rubatto et al., 2009). With no clear link to any nearby granitic pluton, the S-type magmas responsible for the formation of pegmatites in Misox are interpreted to have originated from partial melting at greater depths. Additionally, the pegmatites of Misox are renowned for producing quality mineral specimens of beryl, tourmaline, and garnet (Stroppini, 2019). The geological setting of the Misox pegmatites provides a unique opportunity to study the processes governing the formation and evolution of pegmatitic systems within the context of the tectono-metamorphic history of the central European Alps. Pegmatitic veins cutting orthogneisses in the Simano nappe of the Lepontine dome were dated to 23 to 21.5 Ma based on zircon U–Pb geochronology (Tagliaferri et al., 2023). A sketch of the studied Misox pegmatite vein is shown in Fig. 3. In the studied pegmatite vein, only one miarolitic cavity has been discovered so far, and all the studied quartz crystals are from the same miarolitic cavity assemblage (Stroppini, 2019).

The studied quartz crystals are free grown, often exceeding 10 cm in length. The images of representative quartz crystals are shown in Fig. 4, and the list of samples studied is presented in Table 1.

Optical scans of quartz sections were done using an Olympus BX50 microscope equipped with a microscopic image analyser (MIA) scan stage. Stacked images, with 15 % overlap, were composed using between 50 and 800 single images depending on the size of individual quartz crystals.

SEM charge contrast imaging (CCI) of entire ∼4.5 cm crystals was acquired at low magnification to cover the complete thin section (image stacking). CCIs were collected according to the procedure described by Lehmann et al. (2009). Uncoated doubly polished thin sections were imaged using a Zeiss Evo 50 SEM, located at the Institute of Geological Sciences, University of Bern. The operating conditions were set to an accelerating voltage of 14 kV, a gas pressure of 10 Pa, a sample current of 2.3 nA, and a spot size of 567 nm. Laboratory air was used as the chamber gas, and the scan rate was set to 420 µs per pixel. CCIs were collected with a gaseous scintillation detection system (GSD), which detects the photons produced by excitation reactions of nitrogen gas and secondary electrons (Morgan and Phillips, 2006). Mosaics of CCIs were composed together with 15 % overlap using built-in Zeiss SEM software.

Charge contrast imaging of quartz was chosen as an alternative to cathodoluminescence (CL) imaging because of its potential to reveal compositional variations at a comparable resolution to CL on the same system (Lehmann et al., 2009). While the secondary electrons, which are generated by the electron beam, react very similarly to photons produced in CL imaging and thus reveal the same compositional variation, they have a significantly smaller interaction volume (Griffin, 2000; Watt et al., 2000).

Trace element contents of quartz were measured at the Institute of Geological Sciences, University of Bern, using a Resonetics RESOlutionSE 193 nm excimer laser system combined with an Agilent 7900 quadrupole ICP-MS system. The laser system is equipped with a S-155 dual volume constant geometry ablation cell (Laurin Technic, Australia). The ICP-MS was optimized for low oxide production (ThO/Th<0.2 %) and robust plasma conditions monitored by a Th/U sensitivity ratio >0.97). The atmosphere in which ablation was performed consisted of pure He (0.4 L min⁻¹), into which Ar (0.86 L min⁻¹) and N₂ (0.003 L min⁻¹) were admixed at the exit of the ablation cell. The beam size was 32–100 µm depending on the size of pure quartz domains or zoning as revealed on charge contrast images. Pre-ablation of the measurement spot was performed to clean the surface and was done with a slightly larger beam size compared to the actual measurement. The laser repetition rate was 5 Hz, and the fluence on the sample was ∼18 J cm⁻². The total acquisition time was ∼100 s, with gas background being measured for 50 s, 10 s of washout after pre-ablation, and 30–40 s of sample signal. GSD-1G was used as the external standard to calibrate analyte sensitivities. Data reduction was performed using the SILLS software (Guillong et al., 2008) with 95 % confidence limit of detection (LOD) filtering calculated according to the formulation reported in Pettke et al. (2012). Internal standardization was done using total element oxides of 100 wt %.

The placement of the LA-ICP-MS measurements on all samples was guided based on the zonation observed in SEM charge contrast and FTIR maps, overlain onto the optical microscope scan of the crystals. The placement of measurement spots aimed for full coverage of features observed on maps and avoidance of healed cracks (seen on charge contrast images with dark quartz fillings) and other kinds of damaged parts of the crystal detected in FTIR images (see below). All quartz trace element contents, as well as associated FTIR spectra quantifications and results of TiQ thermometry calculations, are available in Table S1 in the Supplement. Binary plots were generated using MinPlotX (Walters and Gies, 2025).

Transmission FTIR mapping of quartz crystals was carried out at the Institute of Geological Sciences of the University of Bern using a Bruker Tensor II spectrometer with a globar infrared source and a KBr beam splitter, coupled to a Bruker Hyperion 3000 microscope and a liquid-nitrogen-cooled mercury cadmium telluride (MCT) detector. The sample was placed in a closed Plexiglas chamber, which was purged with nitrogen gas during measurements to limit the effect of changing atmospheric CO₂ and H₂O. Because of the large size of the quartz crystals, most maps were acquired using the single MCT detector with an aperture of between 40 and 75 µm, 4 cm⁻¹ wavenumber resolution, and four scans between 600 and 6000 cm⁻¹. All FTIR maps were acquired with linearly polarized light, with the electric vector E parallel to no (i.e. perpendicular to the crystallographic c axis). The atmospheric correction and baseline correction were performed in OPUS^® version 8.5. The baseline correction was performed using the concave rubber-band method with 64 points and four iterations. The spectra were then exported and further processed with the software SpecXY (Gies et al., 2024).

All spectra were normalized to 1 cm thickness, and the integrated absorbances for different maps were generated by extracting the chosen part of the signal, performing a linear baseline correction between the end points of the linear baseline range, and integrating the selected integration range. The average spectra of the LA-ICP-MS spot positions were extracted from the maps using the region of interest (ROI) circle tool of SpecMaps and exported to SpecDB, where the integrated absorbance and H₂O content in µgg-1 were calculated for each extracted mean spectrum. The applied linear baseline ranges, integration ranges, band positions, absorbance coefficients, correction factors, and references for the calculation of OH coupled to Li, B, and Al, as well as total OH absorbance from FTIR measurements, are provided in Table 2.

Pleochroism of the different OH defects was accounted for by applying directional weighting factors that reflect the orientation of the OH dipole relative to the crystallographic c axis. For LiOH and AlOH defects, absorbance is confined to E⊥c, with no contribution for polarization E∥c, corresponding to weighting factors of 1 and 0, respectively. In contrast, BOH defects exhibit absorbance in both directions, with weighting factors of 2/3 for E⊥c and 1/3 for E∥c, reflecting the distribution of OH dipole orientations (after Baron et al., 2015; Potrafke et al., 2019).

Defect-specific factors (Fdefect) relate the integrated absorbance to OH contents expressed as H₂O equivalent concentration by accounting for the molar mass of H₂O (MH2O), quartz density (ρQtz), and the integrated molar absorption coefficient of the respective OH defect (ε) (Eq. 1a). The orientation coefficient (Korientation) corrects the maximum polarized absorbance measured (absorbance ∥a) by incorporating the relative contributions of the ordinary and extraordinary polarization components (Eq. 1b).

Total H₂O mass fractions [µgg-1] associated with each point defect (cH2Odefect) were therefore obtained by multiplying the maximum polarized absorbance (Fabsorbance) by the orientation coefficient Korientation, which accounts for both the reconstruction of the total integrated absorbance from polarized measurements and the integrated molar absorption coefficient. The differences in anisotropic behaviour of the BOH defect in comparison to AlOH and LiOH defects are expressed in the different Korientation obtained by the added contribution of Ane.

The final calculation for the molar AlOH, LiOH, and BOH concentrations was done by multiplying the values of total integrated absorbance by a thickness-normalized maximum absorbance, calculated according to the following formula (Eq. 1c): 1aFDefect=MH2O⋅106ρQtz⋅ε,1bKOrientation=Ano+Ano+AneAno,1ccH2ODefect=FDefect⋅KOrientation⋅Anot, where MH2O is the molar mass of H₂O (18.01528 g mol⁻¹, ρQtz is the density of quartz (2.650 g cm⁻³), ε is the integrated molar absorption coefficient for the selected band, 10⁶ is the conversion factor to convert the units into mug g⁻¹, A is the integrated polarized absorbance, and t is the thickness in cm.

The absorption coefficients from Jollands et al. (2020) were used for the quantification of LiOH, BOH, and AlOH defect concentrations. The absorption coefficient for the AlOH region was modified to be between the values of 174 000 and 218 000 cm⁻² L (mol H₂O)⁻¹, whereas the values of 161 000 and 80 000 cm⁻² L (mol H₂O)⁻¹ were used for the LiOH and BOH defect, respectively.

Finally, profiles were extracted from the maps referenced to the rim of the crystal using the stripe profile method of SpecMaps.

In Table S1 in the Supplement, all integrated absorbances used for quantification are reported, as well as values for H₂O, Li, B, and Al mass fractions associated with each defect.

Overview SEM charge contrast images reveal details of the compositional zonation of quartz crystals. Quartz from the Rosina pegmatite shows internal zonation, often with typical core–rim domains (Fig. 5a). Oscillatory zoning is present only in smaller quartz crystals from Rosina, for example, in sample MB-IT-R2-QC (Fig. 5b). The cores have a brighter contrast than the rims, indicating higher trace element contents in the cores of the crystals. The bases of individual quartz crystals commonly display a network of dark, irregular domains resembling quartz-filled cracks. The abundance of these domains prominently diminishes towards the tips of the crystals.

Quartz crystals from the Misox pegmatite show significantly more complex zonation patterns (Fig. 5c). While the common core–rim zonation pattern is visible, distinct domains of crystal growth, belonging to two different crystals, twinned according to the Dauphiné law, are observed. Dauphiné-twinned quartz domains incorporate different amounts of trace elements, resulting in different contrasts of the same growth zone for different twin domains (Fig. 5d; Lehmann et al., 2009; this work). The tips of the quartz crystals generally show brighter contrast in comparison to the cores of the crystals. Moreover, oscillatory zoning is evident within twin domains (Fig. 5c). While cracking is still present in the bases of the crystals (not illustrated here), cracks are much less abundant compared to quartz samples from the Rosina pegmatite.

Based on the results of the SEM charge contrast images, areas for FTIR mappings of quartz crystals were selected. Figure 6 displays a representative FTIR spectrum of quartz from Rosina and from Misox pegmatites, illustrating absorption bands at different wavenumbers assigned to specific OH-related defects. The Misox quartz is characterized by high absorption of Al- and Li-related OH bands, whereas the BOH defect is barely detected. In contrast, the B- and Li-related defects are more prominent in the Rosina quartz, while AlOH defects are less prominent. It is important to note that the dominance of the LiOH defect in the FTIR spectra of quartz is rather rare and only restricted to evolved pegmatitic environments.

The zonation observed in integrated maps of FTIR polarized band absorptions was directly correlated with SEM charge contrast images of the corresponding area in the quartz crystals (Figs. 7 and 8). The polarized FTIR maps of selected regions of Rosina quartz crystals reveal that brighter SEM charge contrast regions correlate with higher absorbance (Fig. 7). The FTIR map of the integrated B–OH absorbance reveals an oscillatory zoning pattern, different from the core–rim zonation pattern revealed by the SEM charge contrast images, as well as integrations based on other peaks in the FTIR spectrum (Fig. 7). The maps of the integrated LiOH, BOH, and AlOH absorbance of quartz crystals from Rosina show a maximum of 1.9, 3.3, and 8.9 µgg-1 of OH-coupled Li, B, and Al, respectively (Table S1). These values decrease by up to an order of magnitude in the rim of the crystals (Fig. 7). The sum of total Li-, B-, and Al-defect-related structural H₂O mass fractions calculated according to Jollands et al. (2020) reaches up to 12 µgg-1 in the core zone and drops to between 2 and 5 µgg-1 in the rim of the Rosina quartz crystals (Fig. 7). The total H₂O calculated by the method of Thomas et al. (2009) has significantly higher values for the same analyses, reaching up to 97 µgg-1 in the core zone and dropping to between 3 and 7 µgg-1 in the rim of the Rosina quartz crystals (Table S1).

The FTIR maps of selected regions of Misox quartz crystals reveal that brighter SEM charge contrast domains again correlate with higher absorption band intensities (Fig. 8). The Al defect is much more pronounced in Misox samples than in Rosina (Fig. 6). The zonation observed in Misox quartz crystals is significantly more complex than zonations in Rosina quartz. The zonation between the core and the rim of the crystals is still visible, showing higher amounts of OH defects and total H₂O mass fractions in the rims of the crystals. As seen in SEM charge contrast images, all quartz crystals from Misox display Dauphiné law twinning, showing different trace element incorporation in neighbouring domains of the twins. This is also seen in the FTIR maps (Fig. 8). Furthermore, within a single twin domain, crystals exhibit oscillatory zoning that spans both the core and the rim of the crystal (e.g. zonation observed within the domain marked in blue in Fig. 8b). The OH-coupled Li, B, and Al absorbance integration maps of quartz crystals from Misox yield maximum mass fractions in the tips of the crystals, amounting up to 5.8, 1.5, and 68 µgg-1, respectively. A sum of total Li-, B-, and Al-defect-related structural H₂O reaches up to 51 µgg-1 in Misox quartz crystals (Fig. 8). The total H₂O calculated by the method of Thomas et al. (2009) has, once again, significantly higher values for the same analyses, reaching up to 125 µgg-1 (Table S1).

The highest measured LA-ICP-MS trace element contents are present in the cores of the quartz crystals from Rosina. Measured element mass fractions in Rosina quartz range from 11 to 450 µgg-1 Li, 1 to 36 µgg-1 B, 43 to 1780 µgg-1 Al, and 0.16 to 12 µgg-1 Ti, <0.004 to 0.4 µgg-1 Ga, and 1.4 to 11.7 µgg-1 Ge.

The Li and Al contents in quartz are always linearly correlated (Fig. 9a, b). In Rosina samples, this correlation follows the 1:1 molar ratio closely. Both Li and Al, as well as Ga mass fractions, are highest in the core of the quartz sample from the core zone of the Rosina pegmatite (Fig. 9, Table S1). All the samples show a linear correlation trend between Ga and Li. The sample from the core zone of the Rosina pegmatite, however, follows a linear correlation with a different slope, incorporating less Ga per mol of Li (Fig. 9c).

In Misox quartz samples, LA-ICP-MS element mass fractions range from 0.5 to 100 µgg-1 Li, 0.05 to 11 µgg-1 B, 4.5 to 760 µgg-1 Al, and 0.02 to 5.7 µgg-1 Ti. In Misox quartz, all trace elements, including Ti, display lower contents in crystal cores, increasing towards the complexly zoned tips of the crystals. While, within each quartz crystal from Misox, the molar Li to Al is still strongly correlated linearly, the ratio does not follow a 1:1 line like in Rosina samples but rather incorporates variably less Li per mole of Al (Fig. 9b). In Misox, Ga and Ge mass fractions in quartz reach up to 0.01 and 7 µgg-1, respectively.

There is a clear difference between the two localities in terms of trace element incorporation into quartz. Rosina quartz samples are more enriched in Ga, Ti, B, and Li compared to quartz from Misox (Fig. 10). Gallium, Ge, Al, and total H₂O all increase from core to rim in Misox quartz but decrease from core to rim in Rosina quartz (Fig. 8a, b, d).

Titanium contents decrease from the core to the rim of the quartz crystals in both localities. This trend is much more obvious in quartz from Rosina, where, in all samples, it gradually decreases from core to rim, irrespective of the lateral position of the transect through the crystal (Fig. 11). The total Ti mass fractions in quartz crystals from Rosina range from 12.4 to 0.16 µgg-1 (Table S1). Titanium distribution in Misox quartz correlates with extensive twinning domains and complex zoning displayed by the crystals (Fig. 12). The highest Ti mass fractions are generally found in the cores of the crystals, reaching up to 3.2 µgg-1 Ti in sample MB-CH10B. At the rims of the crystals, the Ti mass fractions drop down to 0.03 µgg-1. However, in some samples, such as in MB-CH3, it rises again up to 2.7 µgg-1 in the latest rims of the crystal (Table S1).

Temperatures during quartz crystallization were calculated using the Ti-in-Quartz (TiQ) thermometer from Huang and Audétat (2012). Pressure for Rosina pegmatite formation was estimated to 2.3 kbar, which is assumed to be the lower end of pressure of formation of pegmatitic veins in San Piero facies of Mt. Capanne pluton (Bakker and Schilli, 2016), and Ti activity in the system was assumed to be 0.5 based on the presence of ilmenite in the pegmatite mineral assemblage, identified from scanning electron microscope energy dispersive X-ray spectroscopy (SEM-EDS) spectra. The resulting temperatures are presented in Table S1, including temperatures calculated with the thermometer proposed by Zhang et al. (2020) for comparison. Temperatures decrease from the core to the rim of the crystals in all quartz samples from Rosina. The temperatures obtained by TiQ thermometry range from 590 to 450 °C for the quartz from the core zone of the Rosina pegmatite (Fig. 11). This quartz crystal records the highest temperature of all quartz crystals analysed. TiQ temperatures decrease from the core to the rim of the crystal, without abrupt changes as displayed by other trace element data; there is no sudden drop in temperature recorded at any point. All the other samples from Rosina yield TiQ temperatures between 560 and 320 °C, along with the same gradual decrease from core to rim.

For the Misox pegmatite, pressure was estimated to be 5 kbar, slightly lower than the lowest estimated pressure conditions of partial melting of migmatite leucosomes in the Lepontine dome (Burri et al., 2005), and the same Ti activity of 0.5 was assumed based on the presence of ilmenite in the pegmatite mineral assemblage, which was identified based on the EDS spectra on the SEM scans. The resulting temperatures are presented in Table S1, including temperatures calculated with the thermometer proposed by Zhang et al. (2020) for comparison. These temperatures indicate a gradual decrease from the core to the rim of the crystals in all samples. The quartz from the miarolitic cavity of the Misox pegmatite gives TiQ temperatures of between 545 and 315 °C when going from the core to the rim of the crystal (Fig. 12) (see the Discussion section for an interpretation).

Quantifying H₂O and corresponding OH-coupled Li, B, and Al contents in quartz using FTIR involves several challenges. A strict protocol was followed for sample preparation, of the analytical method, and quantification calibration, which is now addressed in detail.

Firstly, we used only oriented crystals, cut parallel to the c axis, and measured the FTIR spectra in polarized light with E⊥c to obtain the best possible signal-to-noise ratio. The orientation of the quartz crystal relative to the infrared beam can significantly affect absorbance and, thus, resulting OH quantification. Using maximum polarized FTIR measurements to quantify OH content in quartz represents the best-controlled approach. In this configuration, the optical anisotropy of quartz is accounted for by aligning the sample to enhance the OH-related absorption features, minimizing orientation-related variability. In this section, the absorbance for E∥c can also be measured, testing the different anisotropies of the BOH peak with respect to the LiOH and AlOH peaks (Sect. S2 in the Supplement).

Points with large H₂O broadband contributions were filtered out of the dataset by masking all of the pixels with absorbance higher than the signal noise in the range between 3200 and 3100 cm⁻¹. This process by itself removed a significant portion of uncertainty elaborated upon by Stalder (2021). The linear baseline was used to avoid the addition of minor H₂O broadband contribution to the integration of the respective bands. This was tested using selected spots and comparing different methods. The quantification of H₂O in BOH defects using a maximum polarized measurement resulted in less than 10 % difference when compared to values obtained from spectra averaged over measurements parallel to no and ne following the approach by Stalder (2021). However, the AlOH and LiOH absorption in spots with high total OH absorbance differ more significantly when compared to the no–ne method (Sect. S2 in the Supplement). This is not only a result of imperfect absorbance along no but also likely a combination of higher signal noise and higher H₂O broadband contribution. Therefore, if the signal noise is higher, the no–ne method might also have a larger error. This is further supported by a greater difference between the quantification of H₂O using the sum of defects from Jollands et al. (2020) and Thomas et al. (2009) at higher defect concentrations. Nevertheless, in all cases, the uncertainty regarding the absorption coefficients (Table 2) is still far greater than the difference in results when comparing the two methods just addressed (Stalder, 2021; this work).

Because of extreme concentrations of fluid inclusions in some areas of the quartz crystals, peak deconvolution of the FTIR spectra was not done. The H₂O mass fractions were calculated based on absorption coefficients from Jollands et al. (2020) applied to integration ranges covering targeted absorption bands for all samples. This is especially problematic for the quantification of AlOH because the absorption peak positions of the Al triplet may shift. Jollands et al. (2020) provide two different absorption coefficients for the 3382 and 3306 cm⁻¹ peaks; 174 000 and 218 000 cm⁻² L (mol H₂O)⁻¹, respectively. In this study, we used an intermediate absorption coefficient of 180 000 cm⁻² L (mol H₂O)⁻¹ for the complete integration range of the AlOH triplet. However, even when using either end-member coefficient from Jollands et al. (2020) instead of the intermediate value employed here, the resulting variation in H₂O mass fractions related to AlOH remains minor – typically no more than ∼15 µgg-1 (∼10 % relative), even for the highest AlOH absorption. This difference is negligible in the context of total Al mass fractions in quartz, which can reach up to ∼1800 µgg-1.

Combining the LA-ICP-MS and FTIR measurements of quartz gives us a unique opportunity to study the quartz crystallization environment. By measuring the total trace element mass fractions via LA-ICP-MS and the H₂O mass fraction that is coupled to Li, B, and Al mass fractions via FTIR we are able to track the changes in the crystallizing environment and the changes in the mechanisms of their incorporation into the quartz crystal structure.

Assuming the complete growth zone of a single quartz crystal grew at the same time – thus, under constant P–T–X conditions – one can compare single LA-ICP-MS spot measurements and their total trace element mass fractions with the FTIR spectra of each pixel with the same distance from the rim of the crystal. This way, a transect of total trace element contents measured by LA-ICP-MS can be overlapped with the integrated mass fractions of OH-coupled Li, B, and Al and total H₂O (Fig. 13).

Aluminium and Li mass fractions measured via FTIR follow the trend of total mass fractions of these elements measured via LA-ICP-MS. Such a correlation between the total and H₂O-coupled mass fractions of trace elements shows that the Al and Li incorporations into quartz are closely related to each other throughout the whole crystallization sequence (Fig. 13c, d). Germanium seems to follow a very similar zonation pattern through the quartz crystal when compared to B (Fig. 13e, f). The correlation between the total and OH-coupled B is not as clear as that for Li and Al because of lower contents and associated higher noise. While the FTIR data record three local mass fraction maxima for the BOH defect (at around 1000, 1800, and 3800 µm from the rim of the crystal), the first two local maxima are not discernible in LA-ICP-MS spot measurements (Fig. 13e). This shows the capability of FTIR measurements to offer additional information on crystal chemical zonation in comparison to single spot analyses from LA-ICP-MS.

Figure 13b shows that the sudden decrease in Li, Al, and H₂O mass fractions in quartz is not related to a drastic decrease in crystallization temperature, thus suggesting that parameters other than temperature dominate OH-related trace element incorporation. Possibilities include the transition from α to β quartz, the magmatic–hydrothermal transition, or the crystallization of one or more new minerals (such as lepidolite) that affects the Li–Al–OH distribution between minerals. A detailed study of fluid inclusions is needed to clarify this issue.

The charge balance of trace elements in the studied quartz was evaluated by considering the bulk incorporation of trivalent cations (Al³⁺ and B³⁺) and their charge compensation by monovalent cations (H⁺ and Li⁺), representing the dominant trace elements. The Li and Al contents linearly correlate (Fig. 9a, b) because of their coupled substitution Al3+Li+↔Si4+, where Li⁺ in the structural channels compensates for the charge deficit created by the substitution of Al³⁺ for Si⁴⁺ (Fig. 1b). However, the slope of that correlation is not the same in all samples (Fig 9b), meaning that Li and Al contents are not always correlated in a 1:1 molar ratio. Since Li⁺ and H⁺ are the only significant monovalent species detected in our samples, we calculated the total available charge compensation as the sum of structurally bound H₂O (measured by FTIR) and total Li⁺ minus 2 times the molar concentration of the LiOH defect to account for the Li and H not related to the incorporation of trivalent cations: 2Cmonovalent=H++Li+-2⋅LiOH. This approach does not attempt to allocate specific monovalent cations to individual trivalent cations but rather evaluates whether, in bulk, the total positive charge from H⁺ and Li⁺ is sufficient to compensate charge for the trivalent trace element content.

All of the measurements on Rosina quartz (except two outliers) fall on the 1:1 line, showing the same amount of trivalent and monovalent cations and a neutral charge balance (Fig. 14). However, all the measurements on Misox quartz plot to the left of the 1:1 reference line, indicating an excess of trivalent species, not fully charge balanced by the monovalent species (Fig. 14). There are also not enough pentavalent cations for charge balancing, and the excess Al in Misox quartz might be an indication of the presence of nanoparticle inclusions of aluminosilicates in these samples.

Comparison of LA-ICP-MS element data with those obtained from FTIR reveals that only a small fraction of the total Li and Al is incorporated into Rosina quartz through coupled OH-related defects, never exceeding 4 % and 10 % of total Li and Al. For the Misox quartz, OH-coupled defects represent a more important incorporation mechanism for trace elements in the quartz crystal structure. A total of 10 %–30 % of total Al and <10 % of Li are incorporated into Misox quartz through coupling with an OH defect, with higher values at lower contents of elements. In both Rosina and Misox quartz, about 50 % of B is associated with OH defects, significantly higher than the main trivalent cation Al (Table S1). At low B concentrations, it is possible that the entirety of B is hosted in BOH defects.

This demonstrates that OH defects can play a major role in the incorporation of both monovalent and trivalent cations. Sharp zoning patterns in OH distribution maps (Figs. 7, 8) indicate that these OH-related defects are primary and faithfully record changes in fluid composition and trace element availability during hydrothermal quartz crystallization as limited diffusional equilibration preserved the observed zonation. Compositional zoning further reveals systematic core–rim variations: total trace element cations (µmolmol-1) are generally higher in crystal cores of Rosina quartz but lower in the cores of Misox quartz, relative to their respective rims. In Misox quartz, total Al cations dominate over Li in the rims, generating excess trivalent cations, whereas Rosina quartz cores exhibit two distinct populations of total cation concentrations.

Understanding the mechanisms of trace element incorporation in quartz is crucial for tracing the fluid evolution and the changes in P–T conditions during the formation of quartz-bearing ore deposits. There has been significant discussion regarding the optimization and use of TiQ as a geothermometer to determine the temperature of quartz crystallization (e.g. Huang and Audétat, 2012; Zhang et al. 2020; Audétat, 2021; Osborne et al., 2022). However, little to no attention has been given so far to intra-grain variability related to small-scale zoning and twinning of quartz crystals. Our measurement data document significant variability in TiQ temperatures for identical growth zones of different quartz twins. Data reported here result in over 20 °C difference as a function of where the LA-ICP-MS spot measurement was placed within the same growth zone (Fig. 16). It is also visible that the trace element contents are significantly higher in rhombohedral faces of the quartz rim region when compared to contemporaneously crystallized prism faces (Fig. 8b). Recently, Lueder et al. (2024) documented for rutile that Zr shows distinct zonation that correlates with other trace elements by using trace LA-ICP-MS element mapping. Resulting Zr-in-rutile temperatures varied by up to 10 % (±35 °C). This variation is of the same order as the temperature variations reported here (Fig. 15a) and is not a consequence of Ti mineral inclusions in quartz (Fig. 15b). Our results demonstrate that element mapping is essential for improving the understanding and interpretation of mineral thermometry data, while also underscoring the careful attention to detail required for the reliable interpretation of TiQ temperatures.

The geological context of the miarolitic cavity in the Misox pegmatite indicates that it likely formed under water-saturated, eutectic conditions. The haplogranite eutectic at ∼0.6 GPa occurs at ∼750 °C (e.g. Holtz et al., 2001), which might be further lowered by the presence of B to ∼600 °C. This implies that the crystallization temperatures were higher than estimated from TiQ measurements. Consequently, the most plausible explanation is that Ti activity was lower than the value of ∼0.5 inferred from the sporadic occurrence of ilmenite in the samples. A lower TiO₂ activity of 0.3 or 0.2 would result in 30 or 60 °C higher temperatures (Table S1). The key point in the Misox quartz is not the absolute temperatures but the complications that can arise from twinning, resulting in temperature variations of up to 30 °C for adjacent domains in a growth zone.

The variability in trace element contents among quartz crystals from different cavities of Rosina highlights the localized nature of crystallization environments within a single magmatic–hydrothermal ore body. This suggests that each miarolitic cavity evolved in isolation and at local equilibrium. This localized variability is also documented by the relationship between Li mass fractions and TiQ temperatures (Fig. 16). The measured differences in trace element (TE) contents in quartz from three different miarolitic cavities reflect distinct physico-chemical evolutions during crystallization for each cavity. The highest temperatures and Li mass fractions were measured in the core of the sample MB-IT-ROS1 (quartz from the core zone of the pegmatite). At lower temperatures the three samples display parallel trends of Li decrease with decreasing temperatures. This suggests that fluid Li mass fractions from which the quartz crystallized were different in each of these cavities but followed a general trend of decreasing Li incorporation with decreasing temperature. This likely reflects a decrease in Li solubility with decreasing temperature.

The Rosina and Misox pegmatites both contain tourmaline, but only the Rosina samples have Li-rich minerals such as Li mica or Li tourmaline. By plotting Li against total H₂O (Fig. 10d), two distinct compositional fields exist. For a given H₂O content, the Rosina samples have up to an-order-of-magnitude-higher Li contents. Our data indicate that Li contents in quartz can represent a prospecting tool for individuating Li-rich pegmatites based on the trace element and H₂O analyses of detrital quartz (see also Stalder et al., 2017).