Occurrence

Fluor-rossmanite was discovered at the Krutaya pegmatite (‘жила Крутая’ in cyrilic; 50°39'52''N, 109°55'35''E), Malkhan pegmatite field, Krasnochikoyskiy District, Zabaykalskiy Krai, Western Siberia, Russia (Fig. 1). The Malkhan pegmatite field covers an area of ~60 km² on the southern slopes of the Malkhan Ridge, at the interfluve of the Mogzon, Skakunia and Bolshaya Rechka rivers, which are tributaries of the Chikoi River. The Malkhan field was discovered by the ‘Baikalkvartzsamotsvety’ geological expedition in 1983. Its uniqueness lies in its extraordinary abundance of pegmatite dykes containing semi-precious stones, primarily jewellery and collectible tourmaline. The Malkhan field contains >300 pegmatites, 40 of which are tourmaline-rich (Peretyazhko et al., Reference Peretyazhko, Zagorskiy and Bobrov1989).

Figure 1. Krutaya pegmatite, Malkhan pegmatite field. Field of view ~16 m × 8 m. Photo by E. E. Novoselova, summer 2021.

Geologically, the Malkhan pegmatite field occurs in the southwest region of the Caledonian Malkhan–Yablonovaya structural-formational zone. This uplifted area is bound by the Khilok deep fault to the north-northwest and the Chikoi deep fault to the south-southeast. These faults control related Mesozoic basins. More detailed data on the geology and mineralogy of the Malkhan pegmatite field is given by Altukhov et al. (Reference Altukhov, Smirnov and Leontiev1973), Badanina (Reference Badanina1999), Peretyazhko et al. (Reference Peretyazhko, Zagorskiy and Bobrov1989), Vereshchagin et al. (Reference Vereshchagin, Kasatkin and Škoda2022), Zagorskiy (Reference Zagorskiy2010), Zagorskiy and Peretyazhko (Reference Zagorskiy and Peretyazhko1992a, Reference Zagorskiy and Peretyazhko1992b, Reference Zagorskiy and Peretyazhko2006, Reference Zagorskiy and Peretyazhko2008) and Zagorskiy et al. (Reference Zagorskiy, Peretyazhko and Shmakin1999).

The Krutaya pegmatite, where fluor-rossmanite was found, is one of the oldest-known in the Malkhan field. It was discovered by the geological expedition ‘Sosnovgeologiya’ in 1981, i.e. two years before the Malkhan field itself. However, the Krutaya pegmatite was studied much less than the Sosedka, Mokhovaya, Tabornaya, Oktyabrskaya and other world-famous pegmatite dykes that provided hundreds of first-class gem tourmaline specimens. Initial prospecting work in the mid-1980s revealed poor tourmaline mineralisation and crystal quality in the Krutaya pegmatite. It is located 250 m northeast of the mouth of the Zapadnyi stream, it occurs in fine-grained gneiss-diorites and metagabbro and dips to the north-northeast at an angle of 60–70°, hence the name – Krutaya (‘steep’) – see Fig. 1. The pegmatite dyke is 120 m long and up to 5 m thick with a poorly defined zonal structure. Most of the body is composed of graphic pegmatite, consisting of potassic feldspar, quartz and albite. The central part of the pegmatite contains orange spessartine, muscovite, beryl and tourmaline in addition to individual miaroles (up to 0.3 × 0.2 m in size) lined with quartz crystals (Ivanov and Chuev, Reference Ivanov and Chuev2021). In 2020, the geological company LLC ‘Technologiya’ resumed work on the pegmatite. Only very few crystals of gem-quality tourmaline were uncovered, one of which contained fluor-rossmanite. This crystal was obtained from the miners by one of the authors (Mikhail Yu. Anosov) during his trip to Zabaykalskiy Krai in October 2022.

Fluor-rossmanite is the fifth new mineral discovered at the Malkhan pegmatite field. These minerals include bismutocolumbite (Peretyazhko et al., Reference Peretyazhko, Zagorskiy, Sapozhnikov, Bobrov and Rakcheev1992), borocookeite (Zagorsky et al., Reference Zagorsky, Peretyazhko, Sapozhnikov, Zhukhlistov and Zvyagin2003), oxybismutomicrolite (Kasatkin et al., Reference Kasatkin, Britvin, Peretyazhko, Chukanov, Škoda and Agakhanov2020) and nioboixiolite-(Mn²⁺) (Chukanov et al., Reference Chukanov, Pekov, Zubkova, Yapaskurt, Shelukhina Yu, Britvin and Pushcharovsky2023).

General appearance, physical and optical properties

Fluor-rossmanite occurs as an intermediate light pink zone up to 3 mm thick in a chemically heterogeneous, concentrically zoned, polychrome tourmaline crystal 3.2 × 2 cm in size (Fig. 2). The above crystal is unique in terms of the quantity of tourmaline supergroup species it contains – we recorded as many as eight of them – and is the subject of a special publication (Kasatkin et al., Reference Kasatkin, Vereshchagin, Gorelova and Belakovskiy2024b). Here we report only briefly that its dark brown core consists of unusually Mn-rich (up to 9.60 wt.% MnO) fluor-tsilaisite with the average empirical formula (Na_0.50Ca_0.09□_0.41)_Σ1.00(Al_1.28Mn²⁺_1.24Li_0.35Ti_0.06Fe²⁺_0.03Sc_0.01Cr_0.01)_Σ2.98Al_6.00(Si_5.79Al_0.21)_Σ6.00B_2.99O₂₇(OH)₃[F_0.44(OH)_0.09O_0.47]_Σ1.00. Single local compositions corresponding to princivalleite and a potentially new tourmaline species, a Mn²⁺-F-analogue of foitite with the end-member formula □(Mn₂Al)Al₆(Si₆O₁₈)(BO₃)₃(OH)₃F, are also recorded in this zone. The greenish-yellow intermediate zone consists of Mn-rich fluor-elbaite with the average chemical composition (Na_0.49Ca_0.15□_0.36)_Σ1.00(Al_1.53Mn²⁺_0.84Li_0.55Ti_0.05Fe²⁺_0.01Sc_0.01Cr_0.01)_Σ3.00Al_6.00(Si_5.76Al_0.24)_Σ6.00B_2.99O₂₇[(OH)_2.86O_0.14]_Σ3.00(F_0.52O_0.48)_Σ1.00. Very few analyses of this zone correspond to the compositional field of darrellhenryite. This zone is surrounded by a Mn-poor, light pink zone composed mainly of fluor-rossmanite as described here. This zone also includes local chemical compositions corresponding to rossmanite and its oxy-analogue, another potentially new tourmaline species with the end-member formula □(Li_0.5Al_2.5)Al₆(Si₆O₁₈)(BO₃)₃(OH)₃O. Finally, the yellowish-green peripheral zone consists of late-generation, Mn-bearing fluor-elbaite with the average empirical formula (Na_0.56Ca_0.11□_0.33)_Σ1.00(Al_1.75Li_0.90Mn²⁺_0.24Fe²⁺_0.06Cr_0.02Ti_0.02)_Σ2.99Al_6.00(Si_5.82Al_0.18)_Σ6.00B_3.00O₂₇(OH)₃[F_0.49(OH)_0.02O_0.49]_Σ1.00. From the centre of the dark brown zone to the edge of the crystal, the Mn content decreases and the Li+Al content increases up to the middle of the pink zone where MnO reaches 0.21 wt.%. Subsequently, this compositional trend is reversed from the centre of the pink zone to the edge of the crystal. Unfortunately, of the potentially new tourmaline species, only fluor-rossmanite formed monomineralic areas large enough such that it could be described as a valid mineral.

Figure 2. Cross section of polychrome tourmaline crystal with fluor-rossmanite zone. Photo by Maria D. Milshina. Specimen no. 6049/1.

Fluor-rossmanite is pale pink, transparent, with white streak and vitreous lustre. It does not fluoresce under ultraviolet light. Cleavage and parting are not observed. Fluor-rossmanite is brittle with a conchoidal fracture. Its hardness on the Mohs scale based on scratch tests is 7. Its density measured by flotation in Clerici solution is 3.07(2) g cm^–3. A density value calculated using the empirical formula and the unit-cell parameters from single-crystal X-ray diffraction (XRD) data is 3.071 g cm^–3. In transmitted plane-polarised light fluor-rossmanite is non-pleochroic, uniaxial (–), ω = 1.647(2) and ɛ = 1.628(2) (589 nm).

Infrared spectroscopy

The Fourier-Transform InfraRed (FTIR) spectrum of fluor-rossmanite (Fig. 3) was collected using a Thermo Fisher Nicolet iN10 InfraRed microscope equipped with a KBr beam splitter and a LN-cooled MCT detector. The spectrum was collected at an operating resolution of 4 cm^–1 over the range 4000–700 cm^–1 by averaging 64 scans with a scan time of 5.6 seconds. Base-line corrections and peak identification were done using the OMNIC and Fityk V0.9.8 spectra software (Wojdyr, Reference Wojdyr2010). The FTIR spectrum of fluor-rossmanite was collected on an unoriented fragment of the X-ray crystal mounted on a BaF₂ background window.

Figure 3. The FTIR spectrum of fluor-rossmanite in the range 4000–700 cm^–1. Bands are observed that are associated with ^O(1)/O(3)[OH]-stretching modes, [BO₃]^3– and [SiO₄]^4– stretching and bending modes.

A series of peaks is observed in the range ~3660–3320 cm⁻¹ due to (OH) stretching modes associated with ^O(1)(OH) and ^O(3)(OH) where the O(1) site is occupied by the W anions (OH)⁻, F⁻ and O²⁻ and the O(3) site is occupied by the V anion, (OH)^–. Each component band comprising the composite absorption in this region must be associated with the nearest-neighbour arrangements [1] YYY–O(1) or [2] YZZ–O(3) (Bosi, Reference Bosi2013; Hawthorne, Reference Hawthorne2016; Bronzova et al., Reference Bronzova, Babushkina, Frank-Kamenetskaya, Vereshchagin, Rozhdestvenskaya and Zolotarev2019) where Y = Al³⁺ and Li⁺ (and minor amounts of other cations), and Z = Al³⁺ in fluor-rossmanite. In the (OH)-stretching region of fluor-rossmanite, several relatively intense bands are observed at ≤ 3600 cm⁻¹, such bands in tourmaline are generally attributed to local arrangements involving O(3) = (OH)⁻ (Gonzalez-Carreño et al., Reference Gonzalez-Carreño, Fernández and Sanz1988; Bosi et al., Reference Bosi, Skogby, Lazor and Reznitskii2015; Hawthorne, Reference Hawthorne2016) the most abundant of which in fluor-rossmanite is probably LiAlAl–(OH). The bands observed at ≥ 3600 cm⁻¹ in the spectrum of fluor-rossmanite are associated with several distinct YYY–O(1) arrangements where Y = Al³⁺ and Li⁺. These bands are relatively weak, which is in accord with the (OH) contents of the O(1) site compared to the O(3) site. Absorption observed from 1420–1250 cm⁻¹ and 1150–800 cm⁻¹ is due to different symmetric and asymmetric stretching modes associated with [BO₃]³⁻ and [SiO₄]⁴⁻ groups, respectively. Absorption observed in the range 800–700 cm⁻¹ is probably associated with N–O–N and O–N–O bending modes where N = Si, B, Al and Li (Mashkovtsev and Lebedev, Reference Mashkovtsev and Lebedev1991; Robert et al., Reference Robert, Fuchs and Gourdant1996). Absorbance due to atmospheric CO₂ transitions are observed in the range ~2370–2330 cm⁻¹ and regions with relatively more noise observed at ~3900–3740 cm⁻¹ and ~1740–1460 cm⁻¹ are due to atmospheric contamination by H₂O vapour.

Raman spectroscopy

The Raman spectra of fluor-rossmanite (Fig. 4) were obtained from an oriented crystal (E ‖ c and E ⊥ c) by using a Horiba Labram HR Evolution spectrometer. This dispersive, edge-filter-based system is equipped with an Olympus BX 41 optical microscope, a diffraction grating with 600 grooves per millimetre, and a Peltier-cooled, Si-based charge-coupled (CCD) detector. The Raman spectra were collected using a 532 nm laser. The nominal laser beam energy of 50 mW was attenuated to 10% using a neutral density filter to prevent thermal damage of the analysed area. The Raman spectra were collected in the range of 80–4000 cm^–1 using a 50× objective in confocal mode with a beam diameter of ~2.6 μm and an axial resolution of ~5 μm. Time acquisition was 240 s per spectral window; 5 accumulations and 7 spectral windows were applied to cover the 80–4000 cm^–1 range. Wavenumber calibration was done using the Rayleigh line and low-pressure Ne-lamp emissions. The wavenumber accuracy is ~0.5 cm^–1, and the spectral resolution is ~2 cm^–1. Band fitting was done after background correction using Voight functions and assuming combined Lorentzian–Gaussian band shapes (RamanCrystalHunter Software, Nestola et al., Reference Nestola, Zhang, Day, Lorenzon, Pamato, Rocchetti, Bendazzoli, Novella, Mazzoli, Sassi, Pearson, Smith, Scott, Barbaro, Brenker, Santello, Molinari, Qu, Wang, Škoda, Alvaro, Gilio, Murri and Kasatkin2024). Due to the presence of luminescence peaks in the Raman spectra of fluor-rossmanite, the sample was also analysed using a 473 nm laser to distinguish the Raman bands from the luminescence peaks.

Figure 4. The Raman spectrum of fluor-rossmanite in the range of 80–1250 cm^–1 and 3000–3800 cm^–1 excited by a 532 nm laser. The upper spectrum corresponds to the orientation of the tourmaline c axis perpendicular to the laser polarisation, the lower spectrum corresponds to orientation of the c axis parallel to the laser polarisation. The measured spectrum is shown with a dashed line. All observed peaks were fitted with Voigt functions and the resultant fit-line (sum) is shown with a solid black line. The luminescence peaks are marked with an asterisk.

Several Raman bands are observed in the (OH)-stretching region of fluor-rossmanite (Fig. 4). Where E ‖ c, bands are observed at 3469, 3596, 3652 and 3525 cm⁻¹. Where E ⊥ c, three relatively weaker bands are observed at 3473, 3594 and 3652 cm⁻¹, and the 3525 cm⁻¹ band is not observed. The two most intense bands at 3469 cm⁻¹ (3473 cm⁻¹ for E ⊥ c) and 3596 cm⁻¹ (3594 cm⁻¹ for E ⊥ c) probably correspond to (OH)-stretching modes of the YZZ–O(3) arrangements; AlAlAl–(OH) and LiAlAl–(OH), respectively (Fantini et al., Reference Fantini, Tavares, Krambrock, Moreira and Right2014). The weaker band at 3652 cm⁻¹ probably corresponds to the (OH)-stretching mode of the YYY–O(1) arrangement; LiAlAl–(OH). The majority of bands in the lower frequency region from 80–1250 cm⁻¹ correspond to vibrational modes associated with [BO₃]³⁻ and [SiO₄]⁴⁻ groups and [Si₆O₁₈]¹²⁻ rings. Bands observed from ~800–1150 cm⁻¹ correspond to different symmetric and antisymmetric Si–O and B–O stretching modes. Bands observed at < 800 cm⁻¹ correspond to many different vibrational modes that correspond to N–O–N and O–N–O bending modes where N = Si, B, Al and Li. This region also contains bands due to vibrational modes associated with [Si₆O₁₈]¹²⁻ rings (e.g. ring breathing, puckering and compression) and [BO₃]³⁻ groups (Mihailova et al., Reference Mihailova, Gasharova and Konstantinov1996; McKeown, Reference McKeown2008; Fantini et al., Reference Fantini, Tavares, Krambrock, Moreira and Right2014).

Chemical data

Chemical analyses (10 spots) were done with a JEOL JXA-8230 electron microprobe (WDS mode, 15 kV, 20 nA and 10 μm beam diameter). Special care was taken in measuring fluorine (F). A thorough FKα peak search was performed prior to the analysis. The use of a TAP crystal made it possible to avoid the possible overlap of FKα line with MnLα and MnLβ lines and higher order MnKα lines. The counting time for F was 60 s at the peak position and 30 s at high- and low-energy backgrounds. Contents of other elements with atomic numbers higher than that of carbon are below detection limits. The raw intensities were processed for matrix effects using the PAP correction algorithm (Pouchou and Pichoir, Reference Pouchou, Pichoir and Armstrong1985). The theoretical amount of B₂O₃, H₂O and Li₂O were included in the computation.

Lithium and boron contents were determined by inductively coupled plasma – atomic emission spectroscopy (ICP–AES). For this purpose, 0.02 g of the sample was put into an EasyPrep iWave vessel where 2 ml of HNO₃, 3 ml of HCl and 5 ml of HF were added. The vessel was capped and placed in the MARS 6 iWave microwave for digestion at 200°C for 4 hours. Upon digestion, the solution was left to cool at room temperature before being diluted to a final volume of 50 ml with de-ionised water. The solution was then filtered and analysed with a Shimadzu ICPE-9820 atomic emission spectrometer with inductive coupled plasma. Water content was calculated from H content determined by CHNS-analysis carried out using a Thermo Flash 2000 organic elemental analyser.

Analytical data are given in Table 1. The empirical formula calculated on the basis of 31 anions (O+OH+F) is: (□_0.46Na_0.32Ca_0.20Pb_0.02)_Σ1.00(Al_1.84Li_1.05Mn²⁺_0.05Fe²⁺_0.02Ti_0.02Cr_0.01)_Σ2.99Al_6.00(Si_5.79Al_0.21)_Σ6.00B_2.99O₂₇(OH)₃[F_0.44(OH)_0.20O_0.36]_Σ1.00. The ideal formula is □(Al₂Li)Al₆(Si₆O₁₈)(BO₃)₃(OH)₃F which requires Li₂O 1.61, B₂O₃ 11.28, Al₂O₃ 44.06, SiO₂ 38.94, F 2.05, H₂O 2.92, O=F –0.86, total 100 wt.%. The Gladstone–Dale compatibility index (1 – K _p/K _c) calculated for fluor-rossmanite using its empirical formula and the unit-cell parameters determined from single-crystal XRD data is 0.019 using D _calc and 0.018 using D _meas, both values rated as superior (Mandarino, Reference Mandarino1981).

Table 1. Chemical composition of fluor-rossmanite.

* From ICP-AES

** All Mn and Fe considered as MnO and FeO, respectively, for classification purposes. The green colour of the associated fluor-elbaite is probably due to Fe²⁺ (Bosi et al., Reference Bosi, Andreozzi, Skogby, Lussier, Abdu and Hawthorne2013)

*** Calculated from 0.338 wt.% H determined by CHN-analysis.

S.D. – Standard deviation

X-ray crystallography and crystal structure

Powder X-ray diffraction data (Table 2) were obtained using a DRON-2.0 diffractometer with FeKα radiation, a Mn-filter and quartz as an internal standard. The instrument is installed at the Fersman Mineralogical Museum of Moscow, Russia. The parameters of the trigonal unit cell refined from the powder data using the UNITCELL software by Holland and Redfern (Reference Holland and Redfern1997) are as follows: a = 15.7846(13), c = 7.0895(12) Å, V = 1529.81(27) ų and Z = 3.

Table 2. Powder X-ray diffraction data (d in Å) for fluor-rossmanite.

Note: The eight strongest reflections are shown in bold. Only reflections with a relative intensity ≥ 5 are reported.

Single-crystal X-ray diffraction (SCXRD) analysis of fluor-rossmanite was done using a Supernova Rigaku-Oxford Diffraction diffractometer equipped with micro-source MoKα radiation (λ = 0.71073 Å; 50 kV and 0.8 mA) and a Pilatus 200K Dectris detector at the University of Padova, Italy. The data were collected in 1638 frames over 33 runs; the exposure time was 7 seconds per frame for a total time of 12 hours and 17 minutes covering the full reciprocal sphere up to 2θ_max = 63.2° with a redundancy of 23.7 and 100% data completeness. The data were processed by CrysAlisPro 1.171.41.123a software (Rigaku Oxford Diffraction, 2018). The new mineral is trigonal, space group R3m, a = 15.7951(3), c = 7.08646(17) Å, V = 1531.11(7) ų and Z = 3.

The crystal structure of fluor-rossmanite was solved by direct methods and refined to R ₁ = 0.0211 for 1178 unique reflections with I > 2σ(I). The crystal data, data collection information, and structure refinement details are given in Table 3, atomic coordinates and equivalent anisotropic thermal parameters are reported in Table 4, whereas selected bond distances are provided in Table 5. The bond valence calculation is reported in Table 6. The crystallographic information file has been deposited with the Principal Editor of Mineralogical Magazine and is available as Supplementary material (see below).

Table 3. Data from single-crystal X-ray diffraction analysis and structure refinement.

Weighting scheme: w = 1/[s ²(F _o²) + (0.0364×P)² + 0.6233 × P] where P = (F _o² + 2F _c²)/3

Table 4. Atom coordinates and equivalent anisotropic displacement parameters U _eq (in Ų) (the complete list of anisotropic displacement parameters is provided with the crystallographic information file).

Table 5. Selected interatomic distances (Å) for fluor-rossmanite.

Table 6. Bond valence sums (BVS) for fluor-rossmanite.

Note: F1 was calculated considering 44% and 56% O. Y and T sites were calculated using the cation population from the experimental chemical formula. BVS = Bond valence sum. In order to perform the bond valence calculation, we used the mean R₀ and B₀ values from the following references: Brown and Altermatt (Reference Brown and Altermatt1985); Urusov (Reference Urusov2006); Yu and Xue (Reference Yu and Xue2006); Cabana et al. (Reference Cabana, Ling, Oro-Sole, Gautier, Tobias, Adams, Canadell and Palacin2004); Brese and O'Keeffe (Reference Brese and O'Keeffe1991); Czerwinska et al. (Reference Czerwinska, Madura and Zachara2016); Gagné and Hawthorne (Reference Gagné and Hawthorne2015); Adams (Reference St.2001); Kanowitz and Palenik (Reference Kanowitz and Palenik1998); Liu and Thorp (Reference Liu and Thorp1993); Allmann (Reference Allmann1975); Krivovichev and Brown (Reference Krivovichev and Brown2001); Wood and Palenik (Reference Wood and Palenik1999).

The intensity data of fluor-rossmanite were processed and corrected for Lorentz, polarisation, and background effects using CrysAlisPro 1.171.41.123a software. No violation of R3m symmetry was detected. Structure refinement was done using the SHELXL-2019/3 program (Sheldrick, Reference Sheldrick2015). Starting coordinates were taken from Bosi et al. (Reference Bosi, Pezzotta, Altieri, Andreozzi, Ballirano, Tempesta, Cempírek, Škoda, Filip, Čopjaková, Novák, Kampf, Scribner, Groat and Evans2022). Variable parameters were scale factor, extinction coefficient, atom coordinates, site-scattering values (for the X and Y sites), and atomic-displacement factors. The absorption correction was carried out by the interframe scaling (CrysAlis Pro 1.171.41.123a). Neutral scattering factors were used for the cations and oxygen atoms. As for the atomic model refinement, the X site was refined using the Na scattering factor; the Y site was refined using the Al scattering factor. The occupancies of Z and B sites were not refined because of their full occupancies from the chemical composition (e.g. ^ZAl = 6 apfu and B = 2.99 apfu). For the T site, we preferred to fix its occupancy to the chemical composition to avoid refinement of Si versus Al, which would not provide reliable results by X-ray diffraction.

In terms of site occupancy, the observed mean atomic number (M.A.N.; the value is referred to electrons) obtained by the chemical composition at the X and Y sites are 9.16 and 9.87, respectively. The refined M.A.N. from the crystal structure data for the X site is 9.59, which is in good agreement with the EMPA data with a difference of only 4.5%, slightly better that the rossmanite refinement (Selway et al., Reference Selway, Novák, Hawthorne, Černý, Ottolini and Kyser1998), where the difference between the refined M.A.N. and the EMPA data is 6%. Such small discrepancies could be reasonable considering that the X sites in rossmanite and fluor-rossmanite are largely vacant. For the Y site, the refined M.A.N. from the crystal structure data for fluor-rossmanite is 9.93, which is also in good agreement with the EMPA data (9.87). These observations definitively indicate that the chemical composition provided for fluor-rossmanite is accurate. To better show this, we performed a bond valence calculation (Table 6) in order to apply the equation for the F, OH, O distribution at the W site (Bosi, Reference Bosi2013). This equation is given as follows: ^W(OH) = 2 – [1.01 × BVS(F1)] – 0.21 – F. The original equation reports O1 instead of F1 but for fluor-rossmanite we indicated this site as F1. Using our experimental F content from the fluor-rossmanite formula, which is 0.44 apfu, and the bond-valence sum (BVS) at F1, which is ~1.17 vu, we obtain OH = 0.18 apfu compared to the experimentally determined value of 0.20. Thus, using the above equation, we get the following W site occupancy: ^W[F_0.44(OH)_0.29O_0.27]. The occupancy obtained experimentally from the chemical analyses is as follows: ^W[F_0.44(OH)_0.20O_0.36]. Although differences between the calculated and experimental W site occupancy are observed, in both the cases, we confirm that F dominates the W position and that F + OH > O, which are the required conditions to use the prefix ‘fluor-’ in ‘fluor-rossmanite’.

With regards to the crystal structure of fluor-rossmanite, a direct comparison can be made with rossmanite (Selway et al., Reference Selway, Novák, Hawthorne, Černý, Ottolini and Kyser1998). The unit-cell volume of rossmanite is slightly smaller (V ≈ 1531 ų fluor-rossmanite, V ≈ 1526 ų rossmanite); the main reason for this difference is made apparent by comparison of the bond lengths in each mineral. With regards to the X site, <X–O> = 2.678 Å in fluor-rossmanite compared to 2.666 Å in rossmanite. With regards to the Y site, <Y–O> = 1.985 Å in fluor-rossmanite compared to 1.966 Å in rossmanite. As expected, <Z–O> distances are practically identical for both rossmanite and fluor-rossmanite (1.904 and 1.905 Å, respectively) and consistent with the Z site occupied solely by Al (e.g. Kutzschbach et al., Reference Kutzschbach, Wunder, Krstulovic, Ertl, Trumbull, Rocholl and Giester2017; Vereshchagin et al., Reference Vereshchagin, Frank-Kamenetskaya, Rozhdestvenskaya and Zolotarev2018; Bačík, Reference Bačík2018). We observe the same for the B site, which shows an average value of 1.371 Å for rossmanite and fluor-rossmanite. The situation is the same for the T site with an average value of 1.614 Å for both the minerals, which indicates a slight inclusion of aluminium in the TO₄ tetrahedron (e.g. Vereshchagin et al., Reference Vereshchagin, Gritsenko, Vigasina, Dedushenko, Gorelova, Pautov, Agahanov, Chernyshova and Zolotarev2024). To conclude, the larger unit-cell volume of fluor-rossmanite with respect to that of rossmanite is mainly related to the Y site as fluor-rossmanite has a <Y–O> almost 1% larger than in rossmanite. This is due to differences in the occupancy of the Y site, which shows an average cation radii of 0.618 Å in fluor-rossmanite compared to 0.597 Å in rossmanite (Shannon, Reference Shannon1976).

End-member formula and relation to other species

Within the tourmaline supergroup (Henry et al., Reference Henry, Novák, Hawthorne, Ertl, Dutrow, Uher and Pezzotta2011), fluor-rossmanite belongs to the X-site vacant tourmaline group and at the lower hierarchical level – to the vacant-subgroup 2. During the CNMNC voting procedure on fluor-rossmanite, several voting members expressed their concern about its validity as a new species because the total charge at the X site in the empirical formula (+0.76) appears to be closer to 1 than 0. It should be noted that according to Bosi et al. (Reference Bosi, Biagioni and Oberti2019), this value can deviate from the correct value due to a lack of consideration of all the potential end-members involved in chemical substitution. In our case, the occurrence of Ca²⁺ and Pb²⁺, for example, via the substitution Ca(Pb) + Li → □ + Al, affects the total charge at the X site. Based on the empirical formula, several end-member charge arrangements can be written such as:

1. (1) [^X(0) ^Y(1⁺3⁺₂) ^Z(3⁺)₆ (^T4⁺₆2⁻₁₈) (^B3⁺2⁻₃)₃ ^V(1^–)₃ ^W(1^–)]_0.46 limited by vacancy content;

2. (2) [^X(1⁺) ^Y(1⁺3⁺₂) ^Z(3⁺)₆ (^T4⁺₆2^–₁₈) (^B3⁺2^–₃)₃ ^V(1^–)₃ ^W(2^–)]_0.32 limited by 1⁺-cations content;

3. (3) [^X(1⁺) ^Y(1⁺_1.53⁺_1.5) ^Z(3⁺)₆ (^T4⁺₆2^–₁₈) (^B3⁺2^–₃)₃ ^V(1^–)₃ ^W(1^–)]_0.32 limited by 1⁺-cations content;

4. (4) [^X(2⁺) ^Y(1⁺₂3⁺) ^Z(3⁺)₆ (^T4⁺₆2⁻₁₈) (^B3⁺2⁻₃)₃ ^V(1⁻)₃ ^W(1⁻)]_0.22 limited by 2⁺-cations content.

The prevailing charge arrangement is the first one, which aligns with the end-member composition of fluor-rossmanite as ^X□ ^Y(Al₂Li)^ZAl₆(^TSi₆O₁₈)(BO₃)₃^V(OH)₃^WF.

Fractions of individual end-members of the tourmaline studied can be deciphered from its empirical formula by repetitive extraction of the dominant end-member and subsequent application of classification procedures (Bosi et al., Reference Bosi, Biagioni and Oberti2019) to the remaining composition. Apart from dominant fluor-rossmanite (46 mol.%) component, the other participating end-members include darrellhenryite (32 mol.%), liddicoatite (13.5 mol.%) and several other end-member compositions with quantities of <5 mol.% each (Table 7).

Table 7. Proportions of individual end-members in fluor-rossmanite.

Note. The symbols R and X indicate that more elements of the same charge were involved in end-member calculations. Unequivocally defined elements are depicted by their chemical symbols.

Constituents: ^XR²⁺ = Ca, Pb; ^YR²⁺ = Mn²⁺, Fe²⁺; ^YR³⁺ = Al³⁺, Cr³⁺; ^WX^1– = F^–, OH^–1

Fluor-rossmanite is related to rossmanite (Selway et al., Reference Selway, Novák, Hawthorne, Černý, Ottolini and Kyser1998) by the monovalent substitution ^WF⁻ ↔ ^WOH⁻ and to fluor-elbaite, Na(Li_1.5Al_1.5)Al₆(Si₆O₁₈)(BO₃)₃(OH)₃F (Bosi et al., Reference Bosi, Andreozzi, Skogby, Lussier, Abdu and Hawthorne2013) by the heterovalent substitution 2^X□ + ^YAl³⁺ ↔ 2^XNa⁺ +^YLi⁺. Alumino-oxy-rossmanite, □Al₃Al₆(Si₅AlO₁₈)(BO₃)₃(OH)₃O (Ertl et al., Reference Ertl, Hughes, Prowatke, Ludwig, Lengauer, Meyer, Giester, Kolitsch and Prayer2022) can be derived from fluor-rossmanite by the substitution ^YLi⁺ + 2^WF⁻ ↔ ^YAl³⁺ + 2^WO²⁻ leading to a potentially new ‘oxy-rossmanite’ species, □(Al_2.5Li_0.5)Al₆(Si₆O₁₈)(BO₃)₃(OH)₃O and then by ^YLi⁺ + 2^TSi⁴⁺ ↔ ^YAl³⁺ + 2^TAl³⁺.

Fluor-rossmanite is the seventh member of the tourmaline supergroup with a dominantly vacant X site along with rossmanite, □(LiAl₂)Al₆(Si₆O₁₈)(BO₃)₃(OH)₃(OH) (Selway et al., Reference Selway, Novák, Hawthorne, Černý, Ottolini and Kyser1998), foitite, □(Fe²⁺₂Al)Al₆(Si₆O₁₈)(BO₃)₃(OH)₃(OH) (MacDonald et al., Reference MacDonald, Hawthorne and Grice1993), magnesio-foitite, □(Mg₂Al)Al₆(Si₆O₁₈)(BO₃)₃(OH)₃(OH) (Hawthorne et al., Reference Hawthorne, Selway, Kato, Matsubara, Shimizu, Grice and Vajdak1999), oxy-foitite, □(Fe²⁺Al₂)Al₆(Si₆O₁₈)(BO₃)₃(OH)₃O (Bosi et al., Reference Bosi, Skogby and Hålenius2017), celleriite □(Mn²⁺₂Al)Al₆(Si₆O₁₈)(BO₃)₃(OH)₃(OH) (Bosi et al., Reference Bosi, Pezzotta, Altieri, Andreozzi, Ballirano, Tempesta, Cempírek, Škoda, Filip, Čopjaková, Novák, Kampf, Scribner, Groat and Evans2022) and alumino-oxy-rossmanite, □Al₃Al₆(Si₅AlO₁₈)(BO₃)₃(OH)₃O (Ertl et al., Reference Ertl, Hughes, Prowatke, Ludwig, Lengauer, Meyer, Giester, Kolitsch and Prayer2022).

However, fluor-rossmanite is the first species with simultaneous dominance of a vacancy at the X site and F at the W site. Henry (Reference Henry2005) and Henry and Dutrow (Reference Henry and Dutrow2011) observed the general aversion to ^X□–^WF in empirical data of tourmalines; they suggested that it is caused by a combination of both F concentrations in the coexisting fluid and crystallographic factors. The W site which exclusively hosts F, is coordinated by three Y site cations; bond valence requirements related to the short-range ordering of F suggest that it should be present at the W site only if the average charge at the three Y sites is below +7. On the other hand, the X-site vacancy typically compensates for substitution of Al³⁺ at the Y site which increases the total charge at the Y sites. Therefore, the presence of ^XCa in fluor-rossmanite may act to reduce the charge at the Y site. A Y site charge of +6.82, calculated from the empirical formula of fluor-rossmanite, is close to the structural limit although structural strain is partially relaxed by the presence of ^TAl. The most important geochemical factor related to the aversion of tourmaline to ^X□–^WF, is the gradual increase of alkalinity together with F contents in pegmatite melts. This results in the typical evolution of tourmaline composition from foitite to schorl, elbaite, fluor-elbaite and fluor-liddicoatite (e.g. Selway et al., Reference Selway, Novak, Černý and Hawthorne1999; Henry and Dutrow, Reference Henry and Dutrow2011). This trend is partially broken in the final stage of tourmaline crystallisation when both Na and F contents in tourmaline decrease (the ‘darrellhenryite loop’; Skřápková et al., Reference Skřápková, Cempírek, Vašinová Galiová and Cempírek2017) before their renewed increase in the final stage of magmatic crystallisation. This decrease in Na and F is explained either by crystallisation of other F-rich phases in the system, e.g. Li-mica, or by fluid exsolution after pocket rupture and system opening (e.g. Henry and Dutrow Reference Henry and Dutrow2011, Bosi et al., Reference Bosi, Pezzotta, Skogby, Altieri, Hålenius, Tempesta and Cempírek2022).