Structural Breakdown of Natural Epidote and Clinozoisite in High-T and Low-P Conditions and Characterization of Its Products

A heat treatment was performed on selected epidote and clinozoisite crystals to establish the nature of any changes in the optical and crystal-chemical properties and to identify a breakdown product using a wide spectrum of analytical methods. Natural samples were heated from 900 to 1200 °C under atmospheric pressure in ambient oxidation conditions for 12 h. Epidote and clinozoisite were stable at 900 °C; those heated at 1000 °C, 1100 °C, and 1200 °C exhibited signs of breakdown, with the development of cracks and fissures. The average chemical composition of epidote is Ca2.000Al2.211Fe0.742Si2.994O12(OH), while that of clinozoisite is Ca2.017A12.626Fe0.319Si3.002O12(OH). The breakdown products identified by electron microanalysis, powder X-ray diffraction, Raman spectroscopy, and high-resolution transmission electron microscopy were anorthite, pyroxene compositionally close to esseneite, and wollastonite. The decomposition of the epidote-clinozoisite solid solution is controlled by the following reaction: 4 epidote/clinozoisite → 2 pyroxene + 2 wollastonite + 4 anorthite + 2 H2O. Pyroxene likely contains a significant proportion of tetrahedral Fe3+ as documented by the Mössbauer spectroscopy. Moreover, the presence of hematite in the Mössbauer spectrum of the clinozoisite sample heated at 1200 °C can result from the following reaction: 4 epidote → pyroxene + 3 wollastonite + 4 anorthite + hematite + 2 H2O.

The crystal structure and crystal chemistry of epidote-clinozoisite series at room or low temperature have been investigated extensively using several analytical methods samples were heated in a muffle furnace in an air atmosphere for 12 h. Samples were subsequently cooled to ambient temperature for 12 h. A list of samples is provided in Table 1. Chemical analysis was performed with the Cameca SX100 electron microprobe (Cameca company, part of Ametek group, Gennevilliers Cedex, France) operated in wavelengthdispersion mode at the Masaryk University, Brno, Czech Republic, under the following conditions: accelerating voltage 15 kV, beam current 20 nA, and beam diameter 5 µm. The samples were analyzed with the following standards: wollastonite (SiKα. CaKα), TiO 2 (TiKα), Al 2 O 3 (AlKα), pure Cr (CrKα), fayalite (FeKα), rhodonite (MnKα), MgO (Mg Kα), pure Ni (NiKα), albite (NaKα), orthoclase (KKα), Rb 2 ZnSi 5 O 12 glass (RbLα), pollucite (CsLα), barite (BaLα), SrTiO 3 (SrLα), BaF 2 (FKα), and NaCl (ClKα). Detection limits of the measured elements varied between 0.01 and 0.05 wt.% with Ba, F, and Cl always below the detection limit. Peak counting times were 10 to 40 s during measurements depending on the expected concentration of the element in the mineral phase. Major elements were measured using shorter times, whereas longer times were applied for elements with a low concentration. The chemical formula of epidote-supergroup minerals was calculated on the basis of ∑ (A + M + T) = 8 cations with the procedure following epidote-supergroup nomenclature [1].
Powder X-ray diffraction analyses were determined using a BRUKER D8 Advance diffractometer (Bruker AXS GmbH, Karlsruhe, Germany) at the Laboratory of X-ray diffraction SOLIPHA (Comenius University in Bratislava, Faculty of Natural Sciences, Bratislava, Slovakia) under the following conditions: Bragg-Brentano geometry (Theta-2Theta), Cu anticathode (Kα 1 = 1.5406 Å), accelerating voltage 40 kV, and beam current 40 mA. Ni Kβ filters were used for stripping Kβ radiation, and data were obtained using the BRUKER LynxEye detector. The step size was 0.01 • 2θ, the counting time was 3 s per step, and measurements ranged from 2 • to 65 • 2θ. Analyzed scans were fitted, and lattice parameters were refined with BRUKER DIFFRACplus TOPAS software using the structural models for appropriate mineral phases-anorthite [32], pyroxene esseneite [33], and wollastonite [34].
The 57 Fe Mössbauer spectra were obtained at room temperature (298 K) using a standard constant acceleration spectrometer with a 57 Co/Rh radioactive source. Calibration of the apparatus was performed by a 12.5 µm foil of metallic α-Fe and a commercial (Amersham) calibration pellet of sodium nitroprusside (surface density~5 mg Fe/cm 2 ) for high-(±10.5 mm/s) and low-velocity (±4.5 mm/s) regions, respectively. However, all isomer shifts herein are quoted with respect to the room-temperature Mössbauer spectrum of α-Fe. Errors in determination of particular spectral components including relative area, isomer shift (IS), quadrupole splitting (QS) (and/or quadrupole shift), line width (Γ), and hyperfine magnetic field (B hf ) were estimated to be ±1%, ±0.02 mm/s, ±0.03 mm/s, ±0.03 mm/s, and ±0.3 T, respectively (Institute of Nuclear and Physical Engineering, Faculty of Electrical Engineering and Information Technology, The Slovak University of Technology, Bratislava).
Raman spectra of low-P-treated epidote and clinozoisite were acquired on a LabRAM HR Evolution (Horiba, Jobin Yvon, Palaiseau, France) Raman spectrometer system with a Peltier-cooled CCD detector and Olympus BX-41 microscope (Masaryk University, Department of Geological Sciences, Brno, Czech Republic). The Raman spectra were excited by a 532 nm (diode) laser and collected in the range between 100 cm −1 and 4000 cm −1 with a resolution of 2 cm −1 . Multiple spot analyses on different areas of epidote produced similar spectra and confirmed the spectral reproducibility. No surface damage was observed after the laser illumination of the measurement. The acquired Raman spectra were processed using the Peakfit (Systat software, Inc., San Jose, CA, USA) software package. Band fitting was done using the Lorentz function with variable width; the fitting was gradually refined until it produced reproducible results with a square regression coefficient greater than 0.995.
High-resolution transmission electron microscopy (HRTEM) characterization was performed on a Jeol JEM ARM 200cF (JEOL Ltd., Akishima, Tokyo, Japan) analytical transmission electron microscope (TEM) operated at 200 kV that was equipped with a JEOL JED 2300 SDD detector for X-ray energy-dispersive spectroscopy (EDS) microanalysis (Center STU for Nanodiagnostic, The Slovak University of Technology, Bratislava, Slovakia). EDS microanalysis was used to study breakdown products and to distinguish between individual phases. Samples for TEM examination were prepared using a standard technique, which involved mechanical grinding of parent material followed by making a suspension of produced powders and ethanol, before undergoing sonication for 10 min. A drop of suspension was then deposited on a carbon-coated TEM Cu-grid. After drying in the air, the samples were examined by TEM.

Results
Both ES and CP samples displayed only weak chemical zoning in backscattered electron (BSE) images with small variations in Al and Fe 3+ content (Figures 1 and 2) and the average chemical composition listed in Table 2. The main difference between samples was in their Fe/Al ratio. Epidote from Sobotín was M3 Fe 3+ -dominant (0.68-0.80 apfu), while clinozoisite samples had Fe 3+ in the range 0.29-0.33 apfu (Table 2, Figure 3). The content of other cations including Ca, Mg, Ti, V, and Cr was very low or below the detection limit in both types of samples.
Raman spectra of low-P-treated epidote and clinozoisite were acquired on a LabRAM HR Evolution (Horiba, Jobin Yvon, Palaiseau, France) Raman spectrometer system with a Peltier-cooled CCD detector and Olympus BX-41 microscope (Masaryk University, Department of Geological Sciences, Brno, Czech Republic). The Raman spectra were excited by a 532 nm (diode) laser and collected in the range between 100 cm −1 and 4000 cm −1 with a resolution of 2 cm −1 . Multiple spot analyses on different areas of epidote produced similar spectra and confirmed the spectral reproducibility. No surface damage was observed after the laser illumination of the measurement. The acquired Raman spectra were processed using the Peakfit (Systat software, Inc., San Jose, CA, USA) software package. Band fitting was done using the Lorentz function with variable width; the fitting was gradually refined until it produced reproducible results with a square regression coefficient greater than 0.995.
High-resolution transmission electron microscopy (HRTEM) characterization was performed on a Jeol JEM ARM 200cF (JEOL Ltd., Akishima, Tokyo, Japan) analytical transmission electron microscope (TEM) operated at 200 kV that was equipped with a JEOL JED 2300 SDD detector for X-ray energy-dispersive spectroscopy (EDS) microanalysis (Center STU for Nanodiagnostic, The Slovak University of Technology, Bratislava, Slovakia). EDS microanalysis was used to study breakdown products and to distinguish between individual phases. Samples for TEM examination were prepared using a standard technique, which involved mechanical grinding of parent material followed by making a suspension of produced powders and ethanol, before undergoing sonication for 10 min. A drop of suspension was then deposited on a carbon-coated TEM Cu-grid. After drying in the air, the samples were examined by TEM.

Results
Both ES and CP samples displayed only weak chemical zoning in backscattered electron (BSE) images with small variations in Al and Fe 3+ content (Figures 1 and 2) and the average chemical composition listed in Table 2. The main difference between samples was in their Fe/Al ratio. Epidote from Sobotín was M3 Fe 3+ -dominant (0.68-0.80 apfu), while clinozoisite samples had Fe 3+ in the range 0.29-0.33 apfu (Table 2, Figure 3). The content of other cations including Ca, Mg, Ti, V, and Cr was very low or below the detection limit in both types of samples.   The formation of cracks was connected with volume changes after the transformati and dehydration with a visible increase in size with rising temperature. This phenomen was probably related to structural changes in the transition of ES and CP to the breakdow products. The breakdown process is well documented in the XRD pattern (Figures 4 a 5) and Raman spectra ( Figures 6 and 7). The unit-cell dimensions of breakdown produ are provided in Tables 5 and 6. The Raman spectra of samples heated at 900 °C clea show that at this temperature did not take place ( Figure 6, Table 7); however, spec measured from the matrix developed after heating at 1000-1200 °C could be interpret as a composite of anorthite, pyroxene and wollastonite ( Figure 7, Table 8), confirming t results from powder XRD.       After heating at 900 • C, neither sample showed signs of phase transitions, whereas, in samples heated at more than 1000 • C, a new mineral association with plagioclase (anorthite), pyroxene (with a composition close to esseneite), and wollastonite evolved. The breakdown products formed a homogeneous mass retaining the original composition of epidote and clinozoisite in samples heated up to 1100 • C, but samples heated at 1200 • C contained domains enriched or depleted in Fe 3+ (Figures 1 and 2; Tables 3 and 4). This suggests that breakdown products were intergrown on the nanoscale. Moreover, in both sets of samples heated from 1000 • C to 1200 • C, a system of crystallographically oriented cracks and fissures developed (Figures 1c,d and 2c,d).
The formation of cracks was connected with volume changes after the transformation and dehydration with a visible increase in size with rising temperature. This phenomenon was probably related to structural changes in the transition of ES and CP to the breakdown products. The breakdown process is well documented in the XRD pattern (Figures 4 and 5) and Raman spectra (Figures 6 and 7). The unit-cell dimensions of breakdown products are provided in Tables 5 and 6. The Raman spectra of samples heated at 900 • C clearly show that at this temperature did not take place ( Figure 6, Table 7); however, spectra measured from the matrix developed after heating at 1000-1200 • C could be interpreted as a composite of anorthite, pyroxene and wollastonite ( Figure 7, Table 8), confirming the results from powder XRD.   Figure 4. Powder X-ray diffraction patterns of epidote experimental samples (ES) compared to patterns of wollastonite [39], anorthite [40], esseneite [33], and epidote [41] (modeled from structural data using pseudo-Voigt function).  [39], anorthite [40], esseneite [33], and epidote [41] (modeled from structural data using pseudo-Voigt function). Figure 4. Powder X-ray diffraction patterns of epidote experimental samples (ES) compared to patterns of wollastonite [39], anorthite [40], esseneite [33], and epidote [41] (modeled from structural data using pseudo-Voigt function).
Minerals 2022, 12, x FOR PEER REVIEW 10 of 19 Figure 6. The Raman spectra of epidote (a) and clinozoisite (b), natural samples and heated at 900 °C in the 1200-100 cm −1 region; our spectra were compared with the published spectrum of epidote [42].
Both sets of samples heated at 1000 °C or more displayed significant changes in the Mössbauer spectra compared to original samples and samples heated at 900 °C (Tables 9 and 10, Figure 8). According to the mineral phases in breakdown products, it is most likely that Fe was incorporated into the pyroxene structure at both octahedral and tetrahedral sites according to isomer shift (IS) and quadrupole splitting (QS) of measured doublets. Doublets with QS above 0.97 mm/s were attributed to Fe 3+ at the T site, whereas those with QS below 0.70 mm/s were interpreted as Fe 3+ at the M1 site. Moreover, in the CP sample heated at 1200 °C, a magnetic component occurred, which could be interpreted as hematite.  Figure 6. The Raman spectra of epidote (a) and clinozoisite (b), natural samples and heated at 900 • C in the 1200-100 cm −1 region; our spectra were compared with the published spectrum of epidote [42].     390 Both sets of samples heated at 1000 • C or more displayed significant changes in the Mössbauer spectra compared to original samples and samples heated at 900 • C (Tables 9 and 10, Figure 8). According to the mineral phases in breakdown products, it is most likely that Fe was incorporated into the pyroxene structure at both octahedral and tetrahedral sites according to isomer shift (IS) and quadrupole splitting (QS) of measured doublets. Doublets with QS above 0.97 mm/s were attributed to Fe 3+ at the T site, whereas those with QS below 0.70 mm/s were interpreted as Fe 3+ at the M1 site. Moreover, in the CP sample heated at 1200 • C, a magnetic component occurred, which could be interpreted as hematite.  Fitting curves for all data points are gray. Decomposed peaks are also colored with different shades of gray. Shaded sub-spectra correspond in color to Fe 3+ sites from Tables 7 and 8.  Table 9. Parameters of 57 Mössbauer spectra for epidote samples. Color of dublets is for their identification in the Figure 8.  M1  Table 10. Parameters of 57 Mössbauer spectra for clinozoisite samples. Color of dublets is for their identification in the Figure 8.

No. Doublet Area IS (mm/s) QS (mm/s) Γ (mm/s) Mineral Site
CP-0 Epidote-clinozoisite structural breakdown and breakdown products were documented on the nanoscale by HRTEM. Breakdown products formed particles several tens of nm (Figure 9a). It was also used to confirm the presence of both pyroxene and wollastonite, because these XRD peaks and Raman bands were partly overlapping (not in the case of well-evidenced anorthite); therefore, this observation could be taken as their definitive proof. The particle with O, Si, and Ca identified in the EDS spectrum was wollastonite (Figure 9b), whereas the nanocrystal with O, Si, Ca, Fe, Al, and Ti peaks was pyroxene (Figure 9c).

Discussion
In the present experiment, the epidote-clinozoisite solid solution stability was studied under ambient conditions, atmospheric pressure, and without addition of any fluid phase. There was no evidence of phase transition or breakdown of epidote or clinozoisite at 900 °C; however, at T > 1000 °C, a breakdown of both epidote and clinozoisite was observed. The Raman spectra of samples heated at temperatures greater than 1000 °C contained relatively broad bands and were probably the composite of both anorthite and esseneite-like pyroxene contributions. The broadness of the bands suggested a very low structural order of the breakdown product, which was supported by other analytical data.
All products were convincingly identified by the selected analytical methods. Anorthite and wollastonite presumably had low chemical variability. The original epidote and clinozoisite had low Na content, below the EMPA detection limit. This limited possible Figure 9. The HRTEM images of epidote breakdown products: (a) nanocrystals of breakdown products; (b) detailed view of wollastonite with its EDS spectrum; (c) detailed view of pyroxene with its EDS spectrum. Note that Cu in the EDS spectrum is from the Cu TEM grid.

Discussion
In the present experiment, the epidote-clinozoisite solid solution stability was studied under ambient conditions, atmospheric pressure, and without addition of any fluid phase. There was no evidence of phase transition or breakdown of epidote or clinozoisite at 900 • C; however, at T > 1000 • C, a breakdown of both epidote and clinozoisite was observed. The Raman spectra of samples heated at temperatures greater than 1000 • C contained relatively broad bands and were probably the composite of both anorthite and esseneite-like pyroxene contributions. The broadness of the bands suggested a very low structural order of the breakdown product, which was supported by other analytical data.
All products were convincingly identified by the selected analytical methods. Anorthite and wollastonite presumably had low chemical variability. The original epidote and clinozoisite had low Na content, below the EMPA detection limit. This limited possible albite substitution.
Pyroxene was the most intriguing breakdown product. It had the largest compositional variability and could accommodate Si, Ca, and both Al and Fe. This composition indicates that pyroxene in our samples had QUAD composition [47]. The composition of hedenbergite proposed for epidote breakdown [28] was not very likely for our reaction because hedenbergite contains only divalent Fe, and, without Na, there is no substitution that could balance the increased charge of Fe 3+ indicated by the Mössbauer study of our samples. However, a tschermakite-type substitution with trivalent cation substituting for Si 4+ could charge-balance the exclusive Fe 3+ content in studied pyroxene. The tschermakitetype substitution is the most probable mechanism allowing the incorporation of trivalent cations such as Fe 3+ into Ca pyroxene, similarly to esseneite [33].
The doublets in the Mössbauer spectra of breakdown products could be divided into two groups. Both had an isomer shift below 0.40 mm/s, indicating that all Fe in pyroxene was ferric [37,[48][49][50]. However, there were differences between both groups according to their quadrupole splitting. Doublets with quadrupole splitting between 0.60 and 0.70 mm/s were similar to those of octahedral Fe 3+ at the M1 site of aegirinic pyroxene [37,48,50], has albeit with slightly higher QS than in pyroxenes from a hedenbergite-aegirine solid solution (up to 0.49 mm/s [37]). This could have result from the different distortion level of M1 octahedron in the studied pyroxene.
The second group of doublets had distinctly higher quadrupole splitting than octahedral Fe 3+ doublets. Their QS and IS values were similar to tetrahedral Fe 3+ in Fe-esseneitic or Ca-Fe-tschermakitic pyroxenes [49,51,52]. The presence of Fe 3+ at the tetrahedral site caused an expansion of the T tetrahedron. This expansion was anisotropic; the length of nonbridging T-O1 and T-O2 bonds was very similar in pyroxene with tetrahedral Al and Fe 3+ , while the bridging bonds were larger by 0.02 Å in pyroxene with tetrahedral Fe 3+ than those in CaTs [49].
The quadrupole splitting had a negative correlation to the area of tetrahedral Fe 3+ doublets but was positively correlated for octahedral Fe 3+ (Figure 10a,b). The decrease in tetrahedral Fe 3+ quadrupole splitting could be attributed to the reduced distortion of the tetrahedra with increasing T Fe 3+ or T Al occupancy due to M1 Fe 2+ + T Si ↔ M1 Fe 3+ + T Fe 3+ substitution [49,53]. The positive correlation of octahedral Fe 3+ in studied samples was unusual. The difference to the published data [49] could be explained by the different substitution. The electric field was more strongly perturbed by changing the valence of neighboring M1 or M2 ions, and the change in quadrupole splitting of M1 Fe 3+ was also influenced by a change in valence of the tetrahedral site ions [48]. In the synthetic Sideficient pyroxenes of CaFe 3+ AlSiO 6 − CaTiAl 2 O 6 composition, the octahedral distortion and change in quadrupole splitting of M1 Fe 3+ were influenced by the M1 Fe 3+ ↔ M1 Ti 4+ and T Si ↔ T Al substitutions [49]. However, in the studied samples, the Ti content was low. Therefore, the accommodation of Fe 3+ at the octahedral site was allowed by M1 R 2+ + T Si ↔ M1 R 3+ + T R 3+ , where R is either Fe or Al. Consequently, if a random distribution of Al and Fe 3+ is assumed, an increase in the octahedral distortion and quadrupole splitting would have been in a linear dependence of the M1 Fe 3+ proportion. The evolution of quadrupole splitting values against the heating temperature is shown in Figure 10c,d.
quadrupole splitting of M1 Fe 3+ were influenced by the M1 Fe 3+ ↔ M1 Ti 4+ and T Si ↔ T Al substitutions [49]. However, in the studied samples, the Ti content was low. Therefore, the accommodation of Fe 3+ at the octahedral site was allowed by M1 R 2+ + T Si ↔ M1 R 3+ + T R 3+ , where R is either Fe or Al. Consequently, if a random distribution of Al and Fe 3+ is assumed, an increase in the octahedral distortion and quadrupole splitting would have been in a linear dependence of the M1 Fe 3+ proportion. The evolution of quadrupole splitting values against the heating temperature is shown in Figure 10c,d. Pyroxenes in the studied samples showed some similarities, as well as some differences. The a size was similar to hedenbergite [37] in the ES samples, whereas, in the CP Pyroxenes in the studied samples showed some similarities, as well as some differences. The a size was similar to hedenbergite [37] in the ES samples, whereas, in the CP samples heated at >1100 • C, it was significantly smaller, even smaller than that of aegirine [37], shifting toward jadeite [38]. The b size of ES pyroxene was very similar to the published esseneite [35] and aegirine [37] data. The data for CP pyroxene were very variable, increasing with the temperature from a value similar to jadeite [38] for the sample heated at 1000 • C to a value approaching that of hedenbergite [37]. The c size in the CP samples was very similar to aegirine [37] or esseneite [35], whereas, in ES pyroxene, it was even larger than esseneite.
On the basis of the lattice parameters of studied samples, it can be concluded that a and b indicate variable Al 3+ and Fe 3+ occupancy at the M sites. Observed variations could result from the various degree of ordering of these cations. The c sizes in all samples clearly evidence a relatively high substitution of trivalent cations for Si 4+ at the T site. The CP pyroxene likely had a composition close to esseneite. Esseneite was described with the composition of (Ca 1.01 Na 0.01 ) (Fe 3+ 0.72 Mg 0.16 Al 0.04 Ti 0.03 Fe 2+ 0.02 ) Si 1.19 Al 0.81 O 6.00 [33] and end-member formula CaFe 3+ AlSiO 6 . However, the ES pyroxene with the c size larger than esseneite very likely had Fe 3+ at the T site, expanding the tetrahedra and, as the result of its larger ionic radius (0.49 Å, [54]) than Si (0.26 Å, [54]) and Al (0.39 Å, [54]), extending the tetrahedral chain. There was no detectable Fe 3+ in the tetrahedra of natural esseneite, but experimentally synthesized CaFe 3+ AlSiO 6 contained tetrahedral Fe 3+ , implying a cooling rate dependence of Fe 3+ and Al ordering at the M1 and T sites [33]. Consequently, the proportion of Fe 3+ and Al in pyroxene varied to replicate the composition of original epidote and clinozoisite. On the basis of the Fe 3+ content in the original samples and with it is possible to assume that epidote-group minerals can also be precursors for tschermakitic pyroxenes and pyroxenoids, if the temperature exceeds their stability.

Conclusions
In the present experiment, the epidote-clinozoisite solid solution stability was studied under ambient conditions, atmospheric pressure and without addition of any fluid. The structural breakdown was observed in samples heated at 1000-1200 • C. The breakdown mineral association comprised anorthite, pyroxene, and wollastonite with additional hematite occurring after heating at 1200 • C. The XRD and Mössbauer spectroscopy data suggest that pyroxene contained a significant proportion of Ca-tschermakitic molecules, resulting in a composition similar to esseneite but with the substitution of Al 3+ and Fe 3+ at both tetrahedral and octahedral sites depending on their proportion in the original epidote and clinozoisite sample. The high-temperature and low-pressure breakdown of epidote to the specific mineral association suggests that a similar process could also take place naturally.