Microstructural analysis was performed combining transmission light and scanning electron microscopy to determine average grain sizes of epidote and quartz, as well as their volume fractions in selected microstructural domains. Grain size analysis was carried out in ImageJ (IJ 1.46r; Ferreira and Wayne, 2012) using sketches of the microstructures drawn on transmitted-light, BSE, and FSE images. In domains where the microstructure is not suitable for an automated grain size analysis by ImageJ (Figs. 4b, 6, and 7c), epidote and quartz areas were directly measured on a transmitted-light microscope, approximated as ellipses. Epidote grains that are isolated among quartz grains are defined by the presence of epidote–quartz boundaries. Grain boundaries among or between epidote grains are defined as any irregular discontinuity between grains that is confirmed by different extinction angles at the optic microscope (i.e., misorientations of ≥ 2.5∘). Quartz grains are identified by the presence of visible grain boundaries in BSE images and confirmed by observations of differences in extinction angles at the optic microscope. The equivalent diameter of each grain (D) is calculated from the grain areas (A) as D=2×A/π.

X-ray compositional maps in wavelength dispersive mode of Si, Fe, Al, Ca, Mn, and Sr were obtained by an electron probe micro-analyzer (EPMA) on a JEOL-8200 microprobe. Accelerating voltage was 15 keV, specimen current was 100 nA, the step size was 4 µm, and dwell times were 190 ms (map of layers 1–3; Fig. 10a–c) and 180 ms (map of layer 2; Fig. 10d–f). For calibration of the X-ray maps, spot analyses were acquired with a specimen current of 10 nA and calibrated using the following standards: wollastonite (SiO2), olivine (MgO), anorthite (CaO, Al2O3), magnetite (FeO), pyrolusite (MnO), tugtupite (Cl), rutile (TiO2), and celestite (SrO). The processing of the X-ray compositional maps, including map calibration, was done by using XMapTools (Lanari et al., 2014, 2019). Minerals were identified based on the concentrations of specific elements (i.e., Ca, Fe, and Al for epidote; Si for quartz; K for micas) and classified. Maps of the structural formula of epidote were calculated on a 12.5 oxygen basis.

For trace element measurements, a RESOlutionSE 193 nm excimer laser system (Applied Spectra, USA) equipped with an S-155 large-volume, constant-geometry chamber (Laurin Technic, Australia) coupled with an Agilent 7900 ICP-QMS was employed. During ablation, a He atmosphere was used, and Ar was admixed to the carrier gas before reaching the plasma of the ICP-MS. NIST SRM612 was used for optimization of the analytical conditions, ensuring that the ThO production rate was <0.2 % and the Th/U sensitivity ratio >97 %. On-sample fluence was 5 J cm-2, with a repetition rate of 5 Hz. The size of the analysis spots ranged between 20–30 µm, and BSE images were used to plan the analyses to avoid locating any spots across zones with heterogeneous composition (e.g., chemical zoning) or on inclusions. External standardization was done with the USGS GSD-1G standard, and the SRM612 standard was measured as an unknown for quality control in the absence of a well-characterized epidote standard. Bracketing standardization enabled a true-time linear drift correction. Data reduction was carried out with the software SILLS (Guillong et al., 2008), using the sum of total oxides minus H2O (98.3 % for epidote and 100 % for SRM612) as an internal standard (see Halter et al., 2002). The formulation of Pettke et al. (2012) was employed to calculate limits of detection for each element in every analysis.

Measurements of 207Pb / 206Pb ratios (as well as 87Sr / 86Sr ratios; Sect. 4.5) were made in epidote micro-separates following the procedure for sample digestion in acids and for column chemistry detailed by Peverelli et al. (2021; modified from Nägler and Kamber, 1996). Two epidote micro-separates – each mixing Epidote-A (epidote in layer 1) and Epidote-B (epidote grains in the microfold in layers 2–3; see Sect. 5.1) in unknown and different proportions – were prepared (Ep_A + B_ 1 and Ep_A + B_2). After hand-picking, the material was finely ground and washed with Milli-Q™ water. Two aliquots of each (Ep_A + B_1a and Ep_A + B_1b as well as Ep_A + B_2a and Ep_A + B_2b) were weighed in as replicates for each micro-separate, ensuring ca. 300 ng of Pb in each aliquot. During column chemistry with a Sr-spec™ resin (Horwitz et al., 1992), the Sr and Pb fractions were collected in sequence (Haeusler et al., 2016). One large Epidote-A grain was also handpicked and ground, and an amount of the powder corresponding to 250 ng of Sr was digested in acids before the extraction of the Sr fraction by column chemistry. The Pb fraction of this Epidote-A grain was not collected since Pb isotopic data are available from in situ U–Pb isotope measurements by LA-ICP-MS (Peverelli et al., 2021). Procedural blank samples were used to assess contamination during work in the laboratories. Lead isotope ratios were measured on a Thermo Fisher Neptune Plus MC-ICP-MS in desolvated plasma mode equipped with a CETAC Aridus 2 desolvating system. Instrumental mass fractionation was corrected within-run by means of a Tl spike. External reproducibility of the measurements was quantified by measuring the NIST NBS981 standard. The measured Pb isotope ratios were identical to those obtained by Rehkämper and Mezger (2000; their Table 4).

Strontium isotope ratios were measured on a ThermoFisher Triton™ thermal ionization mass spectrometer (TIMS) after loading 250 ng Sr diluted in 6.4 M of HCl on Re filaments using 1.5 µL of a Ta-oxide activator. The SRM987 standard (200 ppm; Weis et al., 2006) was measured for quality control. The detected masses were 84, 85, 86 (center cup), 87, and 88. The interference of 87Rb and within-run mass fractionation were corrected for by using the IUPAC 87Rb / 85Rb and 88Sr / 86Sr values of, respectively, 0.385617 and 8.735209. The SRM987 standard returned a weighted average 87Sr / 86Sr ratio of 0.710279 ± 0.000020 (2 standard deviations, 2 SD; number of replicates n=12), which is higher than the preferred reference value of 0.710248 (see Weis et al., 2006). Standards AGV-2 (n=1) and GSP-2 (n=2) of USGS were measured as unknowns for quality control and they returned 87Sr / 86Sr ratios of 0.704041 ± 0.000018 (2 standard errors, 2 SE) and 0.765202 ± 0.000008 (2 SE), respectively. These values are also higher than the preferred reference values of 0.703981 ± 0.00009 (2 SD) and 0.765144 ± 0.000075 (2 SD) for AGV-2 and GSP-2, respectively (see Weis et al., 2006). A correction based on the reference materials returning higher 87Sr / 86Sr values than their reference values would produce the same shift in all measured 87Sr / 86Sr ratios. Hence, we did not correct our data because only the variability among the samples is relevant in this study, while the interpretation of the absolute Sr isotope ratios is beyond the scope of this work. Rubidium concentrations were not measured, as a correction for 87Rb-derived 87Sr is not necessary in epidote. This is a valid approach because the incompatibility of Rb in the epidote crystal structure results in negligible Rb concentrations (see Frei et al., 2004; Feineman et al., 2007).

The different characteristics of epidote and its microstructures in layers 1–3 (Figs. 3–4) allow the distinction of epidote into Epidote-A and Epidote-B as illustrated below. The microstructural characteristics of epidote – described below – suggest different formation mechanisms between epidote in layer 1 and in layers 2–3.

The compositional maps in Fig. 10 (locations shown in Fig. 3) include layers 1–3, as marked in the figure. The compositions of FeO and Mn range between ca. 12.5 wt %–14 wt % and ca. 2500–5000 µg g-1, respectively, across all analyzed epidote grains. On the one hand, larger Epidote-A grains in layer 1 are zoned, with FeO and Mn concentrations increasing from core to rim. Epidote in layers 2–3, on the other hand, is characterized by uniform concentrations of FeO and Mn. It should be noted that, because the step size of the compositional maps is 4 µm, the chemical variability in epidote in layers 2–3 is better assessed among different grains across the overall microstructure rather than within each crystal (i.e., the majority of grains do not contain enough 4×4 µm pixels).

The minimum spot size used for trace element measurements by LA-ICP-MS is 20 µm, which is ca. 4 times larger than the smallest epidote grains. Therefore, chemical variability is assessed throughout each microstructural layer by relying on measurements in large-enough epidote grains to avoid contamination from grain boundaries. Epidote-A is also addressed here with no specific reference to intra-grain zoning but only across the overall Epidote-A microstructure (i.e., layer 1). Data of Epidote-B are collected in anhedral epidote grains in layer 2. The different extents of chemical variability between Epidote-A and Epidote-B noted in the compositional maps (Fig. 10) are reflected by trace element data (Fig. 11, Table 2).

The concentrations of Sr and Pb in Epidote-B overlap with the trend defined by the same elements measured in Epidote-A, but they cover a more limited range of values (Fig. 11a). The CI chondrite-normalized rare earth element (REE) patterns (Fig. 11b) of Epidote-A and Epidote-B have similar trends, characterized by positive slopes (LaN / YbN of 0.1–0.4 in Epidote-A and 0.03–0.5 in Epidote-B) and variably positive Eu anomalies (1.5–2.4 in Epidote-A and 1.2–2.3 in Epidote-B). The REE trends of Epidote-B fall within the range of Epidote-A and confirm the lesser extent of chemical variability in Epidote-B relative to Epidote-A. Selected elements, plotted as values normalized to the concentrations in the vein's host Central Aar Granite in Fig. 11c, reinforce chemical affinity between Epidote-A and Epidote-B, as well as the lesser chemical variability in the latter relative to the former. Cathodoluminescence (CL) images of the recrystallized quartz grains (Fig. 12) qualitatively indicate trace element variability also in quartz grains (see Ramseyer et al., 1988; Götze et al., 2001; Nègre et al., 2022).

Strontium and Pb isotopic data (Table 3) of the epidote micro-separates and those of the Epidote-A grain are plotted in Fig. 13. To compare the Pb and Sr isotopic composition of Epidote-A with those of the micro-separates mixing Epidote-A and Epidote-B, data obtained from three different techniques are combined in one 207Pb / 206Pb-versus-87Sr / 86Sr plot. All 87Sr / 86Sr ratios and the 207Pb / 206Pb ratios of the micro-separates mixing Epidote-A and Epidote-B were measured by bulk techniques – namely TIMS and solution ICP-MS – which homogenize ca. 9×109 µm3 of epidote material. Epidote-A 207Pb / 206Pb ratios are taken from LA-ICP-MS measurements by Peverelli et al. (2021), which include 23 in situ analyses sampling ca. 20–24×103 µm3 each. The small volume of epidote material sampled in LA-ICP-MS analyses causes much larger variation among the single 207Pb / 206Pb analyses. In order to plot one datum point for each micro-separate aliquot against one representing the Epidote-A end-member, an average 207Pb / 206Pb ratio has to be calculated from the LA-ICP-MS data points. To obtain one datum point for Epidote-A, thus, we use analysis no. 10 of Peverelli et al. (2021; see their Table 5) to represent the 207Pb / 206Pb ratio of Epidote-A in combination with the 87Sr / 86Sr ratio measured by TIMS (Table 3). This is the datum point closest to the average (0.786) and to the median (0.787) values of the 207Pb / 206Pb ratios measured by LA-ICP-MS, hence the best representative value of 207Pb / 206Pb in Epidote-A. It should be noted that there is a fundamental difference between the Pb and Sr isotope systems in epidote: Sr isotopes reflect the Sr isotope composition of the fluid, given the negligible Rb contents of epidote. In contrast, the measured 207Pb / 206Pb ratios contain both initial (i.e., inherited from the fluid during crystallization) and radiogenic (i.e., ingrown and U-derived after crystallization) Pb. Therefore, Fig. 13 is a plot of total (initial + radiogenic) 207Pb / 206Pb vs. initial 87Sr / 86Sr ratios.

The formation of a hydrothermal vein entails the crystallization of a mineral assemblage from a mineralizing fluid that fills a fracture (Bons et al., 2012). Although the original morphology of the studied epidote–quartz vein is obliterated by deformation, the euhedral, often elongate, shapes of Epidote-A grains (Fig. 4d) suggest that the formation of Epidote-A in layer 1 is related to vein-filling mineralization, with crystallization occurring in equilibrium with a fluid as observed in less deformed epidote veins (e.g., Fig. 1a and c of Peverelli et al., 2021) and in other vein-filling minerals (e.g., Oliver and Bons, 2001; Bons et al., 2012; and references therein). The reasons why layer 1 is less deformed than layers 2–3 may be linked to the closer proximity of layer 1 to the host rock, to a role of mechanically strong epidote clusters forming a load-bearing network, or to a combination of both (Handy, 1990, 1994; Masuda, 1990, 1995; Tullis, 2002; Passchier, 2005). The microstructures in layers 2–3 differing from those in layer 1 (Sect. 5.1), along with the lesser extent of chemical variability across epidote grains in layers 2–3 compared to Epidote-A in layer 1 (Figs. 10–11), suggest that the mechanism of formation of Epidote-B is different than that of Epidote-A. The Zener relation between quartz and epidote in layers 2–3 indicates an interplay between epidote and quartz during the formation of this microstructural domain. At the same time, the alignment of Epidote-B grains along the axial planes (Figs. 4a and 6) and the presence of minute Epidote-B crystals along grain boundaries of dynamically recrystallized quartz grains (Fig. 9) suggest that quartz dynamic recrystallization and Epidote-B formation are linked processes and thus that Epidote-B formed during deformation. The overlap in major and trace element compositions of Epidote-B and Epidote-A (Figs. 10–11), though, demonstrates that these epidote generations are chemically related. One way to reconcile this geochemical affinity with different formation mechanisms is a scenario in which Epidote-B formed via dissolution of Epidote-A grains and (re)precipitation during deformation. In fact, a few euhedral epidote grains can still be recognized in layer 2 (Fig. 6a), which are interpreted as Epidote-A relicts inherited from the original vein morphology. The gradation from epidote-rich/quartz-poor domains to virtually epidote-free/quartz-dominated ones in layers 2–3 (Figs. 3, 4, 6a, 7, and 14g) suggests that veining processes produced a morphology similar to that shown in Fig. 14a and that the original vein microstructure was subsequently deformed to form the present-day epidote–quartz spatial distribution. In fact, shifts between epidote-rich and quartz-rich domains are frequent in epidote–quartz veins (e.g., Fig. 1a and c in Peverelli et al., 2021). Further evidence for fluid-mediated mass transfer is also given by CL images of quartz (Fig. 12): these indicate trace element variability within single quartz grains, which is most likely the result of recrystallization in the presence of a fluid that mediated mass-transfer processes (e.g., Ramseyer et al., 1988; Götze et al., 2001; Nègre et al., 2022). The presence of a fluid during deformation is also supported by fluid inclusions in dynamically recrystallized quartz (Fig. 6b–c).

The transport of dissolved epidote-forming material towards the microfold axial planes (Fig. 14d) calls for a mechanism allowing the fluid to move throughout the deforming microstructural domain. The presence of ca. 10 µm sized Epidote-B grains along quartz grain boundaries and at triple junctions (Fig. 9) implies that dissolved epidote-forming material is transported and precipitated into locations of the new small Epidote-B sites. This is achieved by the presence of an interconnected network of porosity hosting an intergranular fluid that allows mass transfer processes. Although transport of dissolved epidote material may be possible by diffusion through an interconnected porosity network, fluid circulation at the centimeter scale enabling mass transfer processes is possible by cavitation with the formation of nucleation sites (Fig. 14e–f; also described as “dynamic granular fluid pump”; see Fusseis et al., 2009; Menegon et al., 2015) among sliding dynamically recrystallized quartz grains (Figs. 9 and 12). A first deformation step in which quartz grain size is reduced by dynamic recrystallization via subgrain rotation (Fig. 14c) is suggested by a crystallographic preferred orientation in quartz-dominated domains (Fig. 8). Once the new quartz grains are formed, grain boundary sliding produces creep cavities (Fig. 14e–f; see Herwegh and Jenni, 2001; Fusseis et al., 2009; Gilgannon et al., 2017, 2021), thus creating nucleation loci for Epidote-B (Fig. 9). A similar process is described by Gottardi and Hughes (2022) in quartzites deformed at the brittle–ductile transition in the crust, in which fluid inclusions are redistributed in the quartz matrix by dynamic recrystallization of quartz by subgrain rotation and grain boundary migration. A similar microstructure and a similar interplay of processes are also discussed in mafic high-temperature mylonites (Kruse and Stünitz, 1999) and quartzites (Nègre et al., 2021; Pongrac et al., 2022). This first step is followed by one in which the grain size and volume abundance of the newly precipitated Epidote-B grains control the grain size of quartz (Fig. 7), hence exerting control on quartz deformation mechanisms by pinning the migrating quartz boundaries (see Jessel, 1987; Olgaard, 1990; Humphreys and Ardakani, 1996; Bate, 2001; Herwegh and Berger, 2004; Herwegh et al., 2011; Cyprych et al., 2016). This suggests a feedback process: (1) quartz grain boundary sliding creates creep cavities in which Epidote-B crystallizes, (2) Epidote-B keeps quartz grain size small, (3) more creep cavities are formed, and (4) more Epidote-B grains are formed. This process is referred to as viscous granular flow (e.g., Fitz Gerald and Stünitz, 1993; Stünitz and Fitz Gerald, 1993; Paterson, 1995; see also Kruse and Stünitz, 1999). The dissolution of epidote may have been facilitated by increased surface area achieved via brittle grain size reduction in Epidote-A grains (Figs. 5c–d and 14b).

Epidote dissolution–precipitation and the transport of dissolved epidote material to the loci of Epidote-B crystallization imply the presence of a fluid whose nature can be assessed by Pb–Sr isotope data. We have mentioned that the epidote micro-separates used for solution ICP-MS measurements mix Epidote-A and Epidote-B to unknown proportions, since the mechanical separation of pure Epidote-B is not feasible due to its small grain size. Therefore, the observation that the 207Pb / 206Pb and 87Sr / 86Sr ratios of the micro-separates are different (Fig. 13) suggests that the micro-separates include different epidote generations and hence that Epidote-A and Epidote-B crystallized from fluids with different Pb and Sr isotope compositions. This is supported by the Pb–Sr isotopic ratios of the micro-separates being different than that measured in an Epidote-A grain (Fig. 13, Table 2). There are two principal processes that can induce changes in the Pb and Sr isotope compositions: (1) radioactive decay of 235,238U into 207,206Pb and 87Rb into 87Sr or (2) open-system conditions allowing for advection of extraneous Pb and Sr with different isotopic compositions. Process (1) is unlikely in this case because ingrown radiogenic Pb and Sr would not be sufficient to cause the observed isotopic differences among micro-separates and Epidote-A grain (Fig. 13). Consequently, the Pb–Sr isotope heterogeneity of the two epidote generations infers fluid-mediated addition of extraneous Sr and Pb during crystallization of Epidote-B. Because the vein microstructure carries evidence of a single deformation event (Fig. 14g), it is likely that a second fluid type entered the system during vein deformation and formation of Epidote-B (Fig. 14d–f). Thus, the differences in Pb and Sr isotopic data of Epidote-A grain and micro-separates are interpreted to document mixing between different contributions: (1) one reflecting the isotopic composition of Epidote-A incorporated upon vein formation and (2) one reflecting the mixing of dissolved Epidote-A material with extraneous Pb and Sr brought to the site of Epidote-B crystallization. The existence of open-system conditions bears the potential for fluid-mediated transfer of chemical constituents (Fig. 14e–f). The fact that the major and trace element composition of Epidote-B maintains the same patterns as, and represents a narrower range of, that of Epidote-A (Figs. 10–11) suggests similarly (either rock- or fluid-)buffered conditions. Hence, during fluid-mediated crystallization of Epidote-A and Epidote-B, physico-chemical (e.g., P, T, fO2) conditions – thus fluid compositions – were closely comparable. The initial compositional variability in Epidote-A thereby became homogenized upon dissolution and crystallization into Epidote-B. Considering that the 299 ± 2 Ma old host granitoid (Schaltegger and Corfu, 1992; Ruiz et al., 2022) contains minerals with high concentrations of U, Th, and Rb (e.g., allanite, biotite), even small and variable contributions of radiogenic Pb and Sr leached from these minerals may produce a measurable shift in Pb–Sr isotope ratios without a resolvable effect on Sr and Pb fluid concentrations. Also, crystallization of Epidote-A and Epidote-B most likely occurred under comparable T conditions (ca. 400–450 ∘C as inferred from the presence of biotite in the vein; see Goncalves et al., 2012). Hence, fluid–epidote partition coefficients of all measured elements remained similar between the two epidote crystallization events. Such a scenario can reconcile the compositional uniformity with Pb and Sr isotopic differences between Epidote-A and Epidote-B, hence allowing at least a fraction of the fluid involved in the deformation of layers 2–3 to be of external origin. Peverelli et al. (2022) measured the H isotope composition of epidote in this sample, obtaining values that can only be explained as a mixture of end-member waters (e.g., meteoric, seawater, etc.) as a source for the epidote-forming fluids. In light of the present data, it is possible that their measured δD value reflects the mixing of internal and external fluids upon deformation of the epidote–quartz vein as described above.

The present major and trace element data show that Epidote-B is geochemically similar to, and less variable than, Epidote-A (Figs. 10–11), and this result might be linked to differences in crystallization mechanisms between the two epidote generations. In a rock-dominated system, the larger geochemical variability in Epidote-A may be due to (1) fluid distillation during epidote crystallization, (2) slight variations in physico-chemical conditions (e.g., P, T) throughout the veining processes, or (3) pulsating fluid fluxes. Trace element data (Fig. 11a–b) defining trends and not distinct epidote populations suggest that the geochemical variability in Epidote-A is most readily explained by an evolving epidote-forming fluid chemistry along with more and more epidote precipitating, thus readily removing compatible elements from the fluid (see also Anenburg et al., 2015). This does not exclude multiple pulses of the same fluid (i.e., same fluid source and chemistry) and a role for varying physico-chemical conditions, such as small temperature variations (e.g., Trincal et al., 2015). The lesser extent of chemical variation in Epidote-B indicates homogenization of the whole trace element budget dissolved into the Epidote-B-forming fluid. Viscous granular flow is a dynamic process in which continuous feedback exists among dissolution of minerals, grain boundary sliding, creep cavitation, mass transfer, and mineral precipitation (e.g., Fitz Gerald and Stünitz, 1993; Stünitz and Fitz Gerald, 1993; Paterson, 1995). This implies that repeated dissolution and reprecipitation of the same material is likely to occur, and the chemical budget gets compositionally homogenized with ongoing deformation. Consequently, in the case of the epidote–quartz microfold in layers 2–3, such a mechanism may account for the chemical homogeneity of Epidote-B. In this respect, combined grain boundary sliding, cavitation, and nucleation promote recycling and homogenization of fluids in deforming polymineralic aggregates.