Crystal-chemistry of micas belonging to the yangzhumingite-fluorophlogopite and phlogopite-fluorophlogopite series from the Apuan Alps (northern Tuscany, Italy)

The present work reports the crystal-chemical characterization of micas from the Monte Arsiccio and Buca della Vena mines (Apuan Alps, Italy) through electron microprobe analysis, single-crystal X-ray diffraction, and Raman spectroscopy. The sample from the Monte Arsiccio mine can be classified as an intermediate member of the yangzhumingite-fluorophlogopite series, with average crystal-chemical formula (K0.85Na0.01Ba0.09)Σ=0.95(Mg2.11Fe2+0.23Fe3+0.11Cr3+0.01Al0.20Ti0.04☐0.30)Σ=3.00(Si3.20Al0.80)Σ=4.00O10.00F1.90Cl0.02(OH)0.08. Unit-cell parameters are a ~ 5.30, b ~ 9.18, c ~ 10.14 Å, β ~ 100.12°, V ~ 486.22 Å3, corresponding to the 1M polytype. Structure refinements, performed in C2/m space group, converged to R1 = 3.54 and 4.46% and provided Mg plus Fe occupancy in the range 86–94% for the octahedral M1 and M2 sites. Raman spectroscopy shows very weak bands in the OH stretching region at ~ 3690 and 3580 cm−1. The sample from the Buca della Vena mine has been identified as an (OH)-rich fluorophlogopite, with average crystal-chemical formula (K0.84Na0.02Ca0.01)Σ=0.87(Mg2.12Fe2+0.55Fe3+0.10Al0.18☐0.05)Σ=3.00(Si2.99Al1.01)Σ=4.00O10.00F1.02Cl0.09(OH)0.89. Its unit-cell parameters are a ~ 5.33, b ~ 9.22, c ~ 10.23 Å, β ~ 100.09°, V ~ 494.39 Å3. Structure refinements gave good R1 values (3.27 and 4.37%) and revealed octahedral occupancy of 82–84% Mg and 16–18% Fe. Strong Raman signals at ~ 3702 cm−1 and 3595 cm−1 were observed in the OH stretching region. The findings allow to better understand not only the mineralogy of the Apuan Alps but, more generally, the crystal chemical details of intermediate dioctahedral-trioctahedral mica belonging to the yanzhumingite-fluorophlogopite series.


Introduction
Yangzhumingite, ideally KMg 2.5 Si 4 O 10 F 2 , is a tetrasilicic potassium fluor-mica found in few localities worldwide. The first natural occurrence of this very rare mineral was reported in a metamorphosed carbonate rock from Bayan Obo, Inner Mongolia, China (Miyawaki et al. 2011). The Bayan Obo yangzhumingite has crystal-chemical formula (K 0.70 Li 0.30 Na 0.01 ) Σ=1.01 (Mg 2.48 Fe 0.06 ) Σ=2.54 (Si 3.96 Al 0.03 ) Σ=3.99 O 10 F 1.92 (OH) 0.08 and unit-cell parameters, derived from X-ray powder diffraction data, a = 5.249(4), b = 9.095(5), c = 10.142(5) Å, β = 99.96(6)°, V = 476.9(5) Å 3 (Miyawaki et al. 2011). Yangzhumingite was also identified in a lamproitic dyke at the Kvaløya Island, North Norway (Kullerud et al. approach that includes electron microprobe (EMP) analysis, single-crystal X-ray diffraction (SCXRD) and Raman spectroscopy. It is noteworthy that notwithstanding more than 150 years of mineralogical studies on the mineralogy of the Apuan Alps, no accurate crystal-chemical studies on mica group minerals have been previously reported. For this reason, this contribution has a two-fold aim: to contribute to the knowledge of the mica mineralogy from this sector of the Northern Apennines; to improve our understanding of the intermediate dioctahedral-trioctahedral mica belonging to the yanzhumingite-fluorophlogopite series.

Geological background and studied samples
The Buca della Vena (latitude 43° 59′ N, longitude 10° 18′ E) and Monte Arsiccio (latitude 43° 58′ N, longitude 10° 17′ E) mines exploited small ore bodies belonging to a series of pyrite ± baryte ± iron oxide ore deposits located in the southern Apuan Alps (D'Orazio et al. 2017, and references therein). In both localities, the ore bodies are hosted in the upper part of a metavolcanic-metasedimentary sequence of Paleozoic age, close to the contact with the overlying Triassic metadolostone (Grezzoni Formation). These rocks, as well as the ore bodies, have been affected by greenschist facies metamorphism during the Alpine orogeny. In addition, the basement rocks are affected by the Variscan orogeny. The geological setting of the Buca della Vena and Monte Arsiccio mines has been discussed by previous authors (e.g., Benvenuti et al. 1986;Costagliola et al. 1990). Since the end of the 1970s, Buca della Vena mine has been known for the occurrence of several rare mineral species (e.g., Orlandi and Dini 2004); among them there are thirteen new minerals, mainly represented by sulfosalts (e.g., Orlandi et al. 1999Orlandi et al. , 2007. The mineralogical discovery of the Monte Arsiccio mine is more recent and dates to the end of the 2000s. Since then, more than one hundred different minerals have been identified and, among them, ten new mineral species, mainly represented by sulfosalts (e.g., Orlandi et al. 2012;Biagioni et al. 2014) and sulfates .
Specimen of mica from both localities have been studied. The specimen from the Monte Arsiccio mine (hereafter indicated as MA) was collected at the Sant'Olga level and it is represented by brownish flakes hosted within metadolostone (Fig. 1a). The specimen from the Buca della Vena mine (hereafter indicated as BdV) is represented by dark green flakes of mica, associated with magnetite, and embedded in metadolostone (Fig. 1b). This specimen was sampled close to the contact between the ore bodies and the metadolostone lenses hosted within the ores; close to this contact, dolomite veins hosting greenish crystals of beryl have been collected.

Analytical methods
Several crystals from the specimen of the Monte Arsiccio and Buca della Vena mines were selected under an optical microscope and then checked for the crystal quality using SCXRD. Four of them, having a relatively good diffraction quality, were suitable for the intensity data collection and structure refinement; the same crystals were then used for chemical analyses and Raman spectroscopy.
EMP analyses were performed on crystals fixed with graphite glue to a graphite mount. A Superprobe JEOL JXA 8200 microprobe (WDS mode), installed at the Eugen F. Stumpfl laboratory (Leoben University, Austria) was used. The instrument operated at 10 kV accelerating voltage, 10 nA beam current, 3 μm beam diameter. Counting times were 15 s for peak and 5 s for background. The employed standards were: wollastonite (Si and Ca), sanidine (Na, K and Al), olivine (Fe and Mg), ilmenite (Ti), chromite (Cr), vanadium (V), rhodonite (Mn), sphalerite (Zn), skutterudite (Ni), baryte (Ba), fluorite (F) and tugtupite (Cl). The following diffracting crystals were selected: TAP for Na, Al, Mg; LIFH for Fe, Ti, Cr, V, Mn, Zn, Ni; PETJ for Si, Ca, K, Ba, Cl; and LDE1 for F. The Kα X-ray line was used for all the analyzed elements except for the Lα of Ba. Detection limits (μg/g) are: Si (300), Al (100), Mg (150), Fe (250), Ti (150), Cr (300), V (800), Mn (150), Zn (200), Ni (200), K (100), Na (150), Ba (200), Ca (100), F (300) and Cl (900). The obtained oxides (wt%) are the average of 3 to 8 spot analyses (see Table 1). Atomic proportions per formula unit (a.p.f.u.) based on 12(O, OH, F, Cl) are also listed in the table. The Li 2 O content was not measured for the studied samples. However, since lithium was previously found in yangzhumingite, the Li 2 O content was estimated for the studied MA samples basing on the Eq. 1 in Tischendorf (1999), see details in the Discussion section. The equation was also tested on samples whose lithium content was measured by ICP-MS or SIMS   Schingaro et al. 2014;Ti-phlogopite in Scordari et al. 2008) and the calculated results were found very close to the experimental ones. SCXRD data were collected by means of a Bruker AXS X8 APEXII automated diffractometer equipped with a CCD detector and graphite-monochromatized Mo Kα radiation (λ = 0.71073 Å) operating at 50 kV, 30 mA and 40 mm crystal-to-detector distance. The whole Ewald sphere (± h, ± k, ± l) was recorded up to θ ~ 35° with a scan width 1.0°/frame and an exposure time 10 s/frame. The APEX program suite allowed to optimize the collection strategy (Bruker 2010) whereas the SAINT package (Bruker 2007) was used for the integration of the intensities of reflections and the correction of the Lorentz and polarization effects. The SADABS software (Bruker 2009) was employed for the empirical absorption correction; XPREP (Sheldrick 2008) was used for the subsequent analysis of the intensity data and the assignment of the space group. Structure refinements were carried out by means of the program CRYSTALS (Betteridge et al. 2003) using the reflections with I > 3σ(I). Overall scale factor, atomic positions, cation occupancies and anisotropic atomic displacement parameters were refined, starting from the coordinates of fluorophlogopite given by Lacalamita et al. (2017). The occupancy of the tetrahedral site was constrained to 1. In the octahedral site, the Mg vs Fe occupancies were varied with full occupancy constraint for the BdV samples whereas for the MA crystals a restraint (0.8 ± 0.1), in keeping with the chemical analyses, on the occupancy of the Mg + Fe atoms was initially applied and then removed in the final cycles of the refinement. At the interlayer site, the K vs Ba occupancies were varied with a full occupancy constraint for the MA crystals; a restraint (1.00 ± 0.08) on the occupancy of the K atoms was, instead, used for the BdV crystals to allow the occupancy of this site to assume values higher or lower than 1. The analysis of the difference-Fourier map of the BdV samples showed the presence of a residual electron density peak ~ 1.1 e − /Å 3 which may be interpreted as the position of the hydrogen atom (being at ~ 0.67 Å from the O4 oxygen). The hydrogen atoms were refined with soft restraints on the bond lengths to regularize their geometry (O-H = 0.90 Å) as well as on U iso (H) (that could vary in the range 1.2-1.5 times U eq of the parent atom); the hydrogen atomic coordinates were also refined with riding constraints (Cooper et al. 2010).
The system was calibrated using the 520.5 cm −1 line of a silicon wafer before each experimental session. The spectrum was collected using a 100 × objective, an exposure time of l20 s and averaging three repeated acquisitions. Raman spectrum of the Monte Arsiccio sample was collected using a Horiba Jobin-Yvon XploRA Plus apparatus (installed at the Dipartimento di Scienze della Terra, University of Pisa), equipped with a motorized x-y stage and an Olympus BX41 microscope with a 10 × objective. Raman spectra were excited using a 532 nm line of a solidstate laser, attenuated to 50% (i.e., 12.5 mW). No sample damage was observed. The minimum lateral and depth resolution was set to a few μm. The system was calibrated using the 520.5 cm −1 Raman band of silicon before each experimental session. Spectra were collected through multiple acquisitions (3) with a counting time of 120 s. Backscattered radiation was analysed with a 1200 gr/mm grating monochromator. The spectra of the Monte Arsiccio and Buca della Vena samples are shown in Fig. 3a, b, respectively. The chemical composition of the BdV samples differs from that of the Monte Arsiccio crystals. In particular, the F amount of the BdV micas is one half that of the MA crystals (on average 1.1 vs 1.9 a.p.f.u., respectively, Table 1) whereas the Fe content of the BdV sample is almost twice than that of the Monte Arsiccio mica (on average 0.7 vs 0.3 a.p.f.u., respectively, Table 1).

Yangzhumingite-fluorophlogopite series
The two studied MA crystals exhibit almost the same unitcell parameters which are, on average, a ~ 5.30, b ~ 9.18, c ~ 10.14 Å, β ~ 100.12°, V ~ 486.22 Å 3 (see Table 2). The value of the c cell parameter in mica strictly depends on the anionic content (Cruciani and Zanazzi 1994;Ventruti et al. 2008;Ottolini et al. 2010;Zema et al. 2010;Scordari et al. 2013). Figure  The values of the a and b parameters of the studied MA samples, instead, are related to the composition of the octahedral sites (Zema et al. 2010). X-ray site scattering refinement of the MA_1 crystal provided similar Fe and Mg occupancies for the M1 and M2 sites (Table S1). Analogous results were also obtained for the MA_3 crystal. In addition, for both samples the < M1-O > distance is longer than the < M2-O > one (~ 2.074 vs 2.061 Å, see Table 3), likely due to a preference of vacancies for the M1 site. Therefore, the crystals are geometrically "meso-octahedral" (Weiss et al. 1985(Weiss et al. , 1992. Regarding individual distances, the M1-O4 and M2-O4 ones are the shortest due to the F for OH substitution at the O4 site. Similar structural features were previously found for the Kvaløya yangzhumingite by Schingaro et al. (2011). The synthetic tetrasilicic fluormica in Toraya et al. (1976Toraya et al. ( , 1978, having only Mg in the octahedral sites, showed short values (~ 2.063 Å) for both the < M1-O > and < M2-O > distances.
Finally, the mean tetrahedral bond length (~ 1.647 Å) of the studied MA crystals is close to that of fluorophlogopite (see Lacalamita et al. 2017

(OH)-rich fluorophlogopite
The BdV crystals show the following unit-cell parameters: a ~ 5.33, b ~ 9.22, c ~ 10.23 Å, β = 100.09°, V = 494.4 Å 3 ( Table 2). The c axis is intermediate between that of the MA crystals and fluorophlogopite (see above), and the endmember phlogopite (c = 10.310(6) Å, Redhammer and Roth 2002). The samples fall out of the trend in Fig. 2, showing a c parameter larger than those of other literature (OH)-rich fluorphlogopites (Joswig 1972;Russell and Guggenheim 1999;Schingaro et al. 2014) which accounts for differences in the tetrahedral and octahedral site composition.
The a and b axes of the BdV crystals are elongated with respect to those of the MA samples, in agreement with the higher iron content in the octahedral sites (e.g. M1 = 84% Mg and 16% Fe, M2 = 82% Mg and 18% Fe in BdV_1 crystal). Coherently, the < M-O > distances are larger than those of MA (~ 2.074 Å vs 2.065 Å, respectively). The BdV crystals differ from the MA ones also in the tetrahedral size, showing a larger < T-O > distance (1.654 Å vs 1.647 Å, respectively) which suggests a lower Si/Al content in the former samples.  Seifert and Schreyer (1971); cross symbol: hypersilic Cl-mica in Nazzareni et al. (2008); square with vertical line: montdorite in Robert and Maury (1979). Open symbols for literature fluorophlogopite [square: BdV crystals in this study; circle: Schingaro et al. (2014); triangle pointing upward: Takeda and Donnay (1966); triangle pointing downward: Joswig (1972); diamond: Hazen and Burnham (1973);pentagon: McCauley et al. (1973); square with horizontal line: Takeda and Morosin (1975); circle with horizontal line: Hazen et al. (1981); triangle pointing upward with horizontal line: Russell and Guggenheim (1999); triangle pointing downward with horizontal line: Gianfagna et al. (2007); diamond with horizontal line: Balassone et al. (2013); pentagon with horizontal line: Lacalamita et al. (2017); star: Schingaro et al. (2011); asterisk: Scordari et al. (2013)]. The line is a guide to the eye The difference between the < K-O > inner and < K-O > outer distances leads to Δ K-O = 0.29 in the BdV and Δ K-O ~ 0.21 (Table S2) in the MA crystals, highlighting slight differences in the composition of the interlayer site. Figure 3a illustrates the Raman spectrum (3800-100 cm −1 ) of the MA_2 sample. In the low frequency range (1100-100 cm −1 ), broad and weak bands at ~ 1077 and 890 cm −1 due to Si-O antisymmetric internal modes as well as a sharp band at 679 cm −1 due to symmetric Si-O-Si modes are observed. Below 600 cm −1 , a series of bands (at 534, 319, 270 and 187 cm −1 ) occur; they can be assigned to Si-O deformation modes, internal MO 6 vibrations and M-(OH) librations. Similar spectral features were observed for the BdV sample (Fig. 3b).

Raman spectra
In the O-H stretching region (4000-3000 cm −1 ), the Raman spectrum of the MA mica is noisy and only very weak signals may be identified at ~ 3690 and 3580 cm −1 . On the contrary, in the same region, a well-defined Raman signal was found for the BdV mica (Fig. 3b). In particular, a broad band at 3595 cm −1 (with shoulders at 3611 and 3565 cm −1 ), likely due to OH groups bonded to 2 Mg and adjacent to an octahedral vacancy, is associated to a high intensity peak centered at 3702 cm −1 which may be assigned to MgMgMg-OH and/or MgMgFe 2+ -OH local environments.
The Raman band assignment agrees with that reported in literature for phyllosilicates (e.g. Wang et al. 2015;Moro et al. 2017). In summary, the Raman analysis highlights differences in the OH content between the two studied mica specimen as well as the occurrence of octahedral vacancy substitutions. These evidences are in keeping with the crystal-chemical formulae in Table 4 and with the derived substitution mechanisms (see more details below). Table 4 lists the crystal-chemical formulae of the studied crystals which were calculated combining the EMP data with the results of the SCXRD and Raman spectroscopy following the procedure explained elsewhere (see, for instance, Schingaro et al. 2013;Lacalamita et al. 2017).
It can be noted that in the MA samples, the octahedral vacancies are justified by the mechanisms (2) and (3) (see crystal chemical formulae in Table 4) whereas in the Kvaløya yangzhumingite the dominant mechanism is (2). This difference is reflected by the values of selected distortion parameters (see Table S2). In particular, the presence of high-charge cations in the octahedral sites of MA crystals entails greater distortion of the parameters D.M., ELD and OAV and lower t oct and volume M2 in comparison to the Kv_Yan_sc sample in Schingaro et al. (2014). The low values of the shift M2 parameter is commonly related to the absence of oxy-type substitutions. Other distortion parameters (volume M1 , ψ angles and BLD parameters in Table S2) seem less affected by the cation substitutions at the octahedral sites.   Table S2) whereas it causes an increase of the in-plane rotation angle (on average α ~ 4.7 vs 1.4°, Table S2) which represents the rotation of the individual tetrahedra to match the difference in sizes between the tetrahedral and octahedral sheets (Mercier et al. 2005). The higher the α rotation angle, the greater the difference between the K-O outer and K-O inner distances (Δ K-O parameter in Table S2). In the case of the MA samples, this parameter is also affected by the relatively high amount of barium (0.10 Ba a.p.f.u.) in the interlayer site. In addition, the IV Al → IV Si substitution seems to regularize the tetrahedra as testified by the lower values of the BLD T (0.2 vs 0.6, Table S2) and TAV (about 4.4 vs 8.9°, Table S2) parameters in the MA crystals with respect to the yangzhumingite. The MA samples also show a less corrugated tetrahedral sheet (Δz ~ 0.003 Å in the MA vs 0.008 Å in the Kv_Yan_sc, see Table S2).
Finally, the estimated Li 2 O content (0.02 wt%) in the studied samples corresponds to only 0.01 Li a.p.f.u., which has a negligible effect on the crystal chemistry of the MA crystals. Low Li 2 O concentration (0.006 wt%) was also detected in the Kvaløya yangzhumingite in comparison to the 0.30 Li a.p.f.u. of the Bayan Obo yangzhumingite in Miyawaki et al. (2011).

(OH)-rich fluorophlogopite
The crystal-chemical formulae of the BdV crystals (Table 4) show that the main substitution mechanisms are: Al, Fe 3+ -Tschermak [ VI M 2+ + IV Si 4+ ↔ VI (Al 3+ , Fe 3+ ) + IV Al 3+ ]; R 3+ -vacancy (3 VI R 2+ ↔ 2 VI R 3+ + VI ☐, with R 3+ = Al, Fe 3+ ); (OH) − ↔ F − . This mica differs from the MA for the higher IV Al content (about 1 vs 0.8 a.p.f.u., Table 4) which explains the lower values of the BLD T (~ 0.1 vs 0.2, Table S2) and TAV (~ 2.4 vs 4.4°, Table S2) distortion parameters and the higher values of the α rotation angle (~ 6.4 vs 4.6°, Table S2). As concern the octahedral site, the BdV crystals exhibit a higher VI R 2+ (i.e. Mg + Fe 2+ ) content (~ 2.7 vs 2.4 a.p.f.u., Table 4) and less vacancies (~ 0.1 vs 0.3 p.f.u.) with respect to the MA crystals. These features cause minor distortion of the M1 and M2 octahedra of the BdV crystals as testified by the values of the BLD, and OAV parameters in Table S2. Finally, the low extent of the (OH) − ↔ F − substitution in the Buca della Vena mica with respect to the MA crystals is responsible for the larger interlayer separation (t int ~ 3.40 vs 3.31 Å, Table S2).

Conclusions
The present paper represents the first contribution to the knowledge of the crystal-chemistry of micas from the ore deposits of the Apuan Alps. The detailed comparative crystal-chemistry of the present study allowed to identify a mica with a peculiar composition. Indeed, the specimen from the Monte Arsiccio mine can be classified as an intermediate member in the yangzhumingite-fluorophlogopite series since it is characterized by the same substitution mechanisms active in the structure of the yangzhumingite, although its tetrahedral composition approximate that of a fluorophlogopite. This finding provides evidences of a solid solution between these two minerals.
The specimen of mica from Buca della Vena shows, instead, the typical crystal-chemical features of a (OH)-rich fluorophlogopite affected by vacancies substitutions.
In both cases, the occurrence of high F contents is particularly intriguing and could explain the finding of some fluo-sulfates recently described from the Apuan Alps ores and related to the interaction between H 2 SO 4 released during pyrite weathering and (F-rich mica)-bearing country rocks (e.g., wilcoxite, khademite- Mauro et al. 2019Mauro et al. , 2020. In addition, the sample from the Monte Arsiccio mine contains relatively high Ba contents. This is not an unexpected feature owing to the occurrence of several Ba-minerals (e.g., "hyalophane", cymrite, mannardite- Biagioni et al. 2009), in addition to widespread baryte. Consequently, considering also the variable geochemistry of the ore deposits from the southern Apuan Alps, extensive studies on mica group minerals should be carried out, in order to get new insights into the crystal-chemistry of Ba in micas, as well as to describe the chemical variability of these minerals in the Apuan Alps hydrothermal ores.