To determine the chemical compositions of fluorcarletonite and associated minerals, a JEOL JXA-8200 electron microprobe was employed. It is equipped with a high-resolution scanning electron microscope, an energy-dispersive spectrometer with the Si(Li) detector (resolution – 133 eV) and five wavelength-dispersive spectrometers (WDSs).

The analyzed sample was a polished section of a charoitite rock. The sample images in back-scattered electrons obtained by scanning over the area showed that it consists of small intergrowths of several minerals that were WDS-diagnosed.

The conditions for the excitation and recording of analytical signals were as follows: accelerating voltage =20 kV; probe current =2 nA; beam diameter =10–20 µm, depending on the analyzed object; counting time on peak =10 s; and counting time on background =5 s. The following standards were used: albite (Na, Al), diopside (Mg, Ca, Si), ilmenite (Ti), pyrope (Fe), orthoclase (K), apatite (Cl, P) and F-apatite (F). A conversion from X-ray counts to oxide weight percentages was obtained with the ZAF data reduction system (ZAF correction is combination of corrections for atomic number effect (Z), for absorption (A), and for fluorescence phenomena (F)).

Table 1 offers the average compositions of some areas of the studied sample compared with carletonite (Chao, 1971). The H2O and CO2 contents were determined and/or confirmed by the TG–DSC, FTIR and SCXRD investigations (see below).

The powder X-ray diffraction data of the mineral were collected using a Bruker D8 ADVANCE diffractometer equipped with a Göbel mirror and scintillation detector with radial Soller slits on the diffraction beam. Data were recorded in step scan mode in the range of diffraction angles 2θ from 5 to 80∘, using CuKα radiation. Experimental conditions were as follows: 40 kV, 40 mA, time per step – 1 s and step size – 0.02∘2θ. The unit-cell parameters of fluorcarletonite refined using TOPAS 4 (Bruker, 2008) are a=13.219(1) Å, c=16.707(2) Å, V=2919.4(6) Å3. The XRD pattern is similar to that obtained for carletonite by Chao (1971) (Table S1, Supplement S1, freely available online as a Supplement linked to this article).

A fragment of blue fluorcarletonite single crystal was mounted on a Bruker AXS D8 VENTURE dual-source diffractometer with a Photon 100 detector under monochromatized MoKα radiation. Low-temperature (100 K) data collection was done using a Bruker Cobra nitrogen cryostat. Two sets of 12 frames were used for initial cell determination, whereas the entire Ewald sphere (±h, ±k, ±l) up to θmax⁡∼40∘ was recorded by a combination of several φ and ω rotation sets, with 0.5∘ scan width and 6 s exposure time per frame. Operating conditions were crystal-to-detector distance of 40 mm, 50 kV and 1 mA. The data collection strategy was optimized by the APEX2 program suite (Bruker, 2003) and the reflection intensities were extracted and corrected for Lorentz polarization by the SAINT package (Bruker, 2007). A semiempirical absorption correction was applied by means of the SADABS software (Bruker, 2009). The XPREP software assisted in the determination of the space group and in the calculation of the intensity statistics. Finally, least-squares refinement was performed using the program CRYSTALS (Betteridge et al., 2003). Information concerning the data collection and refinement is listed in Table 2. The refined parameters included scale factors, atom coordinates, atomic displacement parameters and occupancies (for alkaline and alkaline earth sites and O(w) positions). The initial atom coordinates for the structure refinement were taken from Chao (1972). The refinement converged to the R1 value of 1.90 % (wR=2.28 %). Final atomic coordinates and displacement parameters of fluorcarletonite sample are given in Table 3; selected interatomic bond distances and angles are listed in Tables S2 and S3, Supplement S1.

A polycrystalline sample of the mineral was investigated by differential scanning calorimetry (DSC) on a synchronous analysis device STA 449 F1 Jupiter in an argon atmosphere. The sample was heated from 45 to 1500 ∘C at the rate of 5 ∘C min-1. The qualitative and quantitative composition of gaseous thermolysis product was monitored using a 403 C Aëolos quadrupole mass spectrometer. The initial mass was 33.54 mg. The energy of electron impact is 70 eV.

The Fourier transform infrared spectrum of fluorcarletonite powder in a KBr pellet was recorded by a Vertex 70 FTIR Bruker infrared spectrometer and a Varian 3 100ATR/FTIR (RAM II) spectrophotometer. Infrared spectrum was acquired from 4000 to 400 cm-1, as displayed in Fig. 3.

The crystal chemical formula was calculated on the basis of (Si + Al) =8 atoms per formula unit (apfu) and assuming H2O content consistent with the refined occupancies of the O11(w) and O12(w) sites (Table 3). The average formula of the fluorcarletonite is K1.04Na3.89Ca3.87Ti0.01Si7.99Al0.01O18(CO3)3.86(F0.72OH0.28)⋅1.11H2O. It is noteworthy that Al is almost absent in the fluorcarletonite compared to the carletonite from the Mont Saint-Hilaire massif (Table 1). In our case, the Ti concentrations in the sample are low. The Canadian mineral was described as nonstoichiometric with the deficiencies in K, Na, Ca, CO3 and F (Chao, 1971, 1972).

Electron microprobe analysis of fluorcarletonite revealed that individual crystals are very inhomogeneous with respect to the fluorine content (from 0.9 wt % to 1.7 wt %) and the average F content in the sample is 1.3 wt % (vs. ∼0.7 wt % in the carletonite from Mont Saint-Hilaire, Chao, 1971). Accordingly, the calculated atomic proportion of F, substituting by the OH group in the crystal structure, for the studied Murun fluorcarletonite ranges from 0.53 to 0.98 apfu compared with 0.41 apfu in the carletonite from Mont Saint-Hilaire (Chao, 1971).

Some of the analyzed areas contain a reduced amount of fluorine. Basically, these are the peripheral zones of crystalline fluorocarletonite aggregates, characterized by the less saturated blue color. Analysis points with F content of 0.17 wt %–0.78 wt %, with an average of 0.54 wt %, were noted. This corresponds to 0.31 fluorine atoms per formula unit, indicating a small content of carletonite, K1.04Na3.84Ca3.97Mg0.01Si7.99Al0.01O18(CO3)3.93 (OH0.69F0.31)⋅nH2O, in the charoitic rock sample.

The mass spectra exhibited the presence of H2O (mass number, m.n., 18), CO2 (m.n. 44) and F (m.n. 19). The endothermic effects at 45–127 and 127–315 ∘C are accompanied by the weight losses of 0.15 % and 1.02 % most likely related to the release of surface and structural H2O, respectively (Fig S1, Supplement S1). The endothermic effect at 630–1134 ∘C accompanied by the weight loss of 14.9 % results from the release of CO2 (Fig. 2). It should be noted that the loss of CO2 in this temperature range shows several endothermic effects, which suggest that the crystal structure contains carbonate ions in different coordination environments. The exothermic effect at 1136–1500 ∘C is accompanied by the weight loss of 1.22 % and is most probably due to the loss of fluorine (Fig. S2, Supplement S1). It is worthy to note that the release of fluorine continues while cooling the sample to about 1450 ∘C (Fig. S3, Supplement S1).

The infrared spectra features found in this study are in agreement with the previous ones reported for carletonite and related minerals of similar chemical composition (Table S4, Supplement S1).

The spectral range from 400 to 1200 cm-1 on Fig. 3 displays overlapping of bands, which can be divided into two sets. The first set of lines observed at 435, 455, 499, 524 and 592 cm-1 may be assigned to the bending Si–O vibration. Two bands at 1049 and 1196 cm-1 can be assigned to the stretching vibrations of the Si–O bonds. The bands at 662, 690, 728, 782 and 875 cm-1 are attributed to the bending vibrations of CO3 groups, whereas the bands observed at 1393, 1450, 1477 and 1525 cm-1 are assigned to the CO3 stretching vibrations.

The bands of H-containing groups occur in the region extending from 4000 to 1500 cm-1. The H2O stretching band is observed at 3436 cm-1, and the band at 1623 cm-1 is assigned to the bending vibration of H2O. Finally, the peak at 3555 cm-1 is attributed to the stretching vibration of hydroxyl groups.

The structural features of fluorcarletonite are reported in Tables 3, 4 and 7; Tables S1 and S2 (Supplement S1); and Figs. 4–5.

Figure 4 shows the crystal structure of fluorcarletonite in projection down the b (Fig. 4a) and c axes (Fig. 4b), whereas Fig. 5 displays the details of the crystal structure with the emphasis on the tetrahedral linkage. The crystal structure consists of silicate layers, sheets of Na- and Ca-centered polyhedra, and K+ cations occupying the cavities within the eight-membered silicate rings. Additional H2O molecules are situated within pores of the silicate double layers. Finally, isolated CO3 triangles are linked to the Na- and Ca-centered polyhedra sheets.

It is reasonable to start the description of the crystal structure of fluorcarletonite from silicate layers. According to Liebau (2012), the loop-branched sechser double-silicate layer in carletonite, as well as in fluorcarletonite (Fig. 5a), results from the condensation of two branched single layers (Fig. 5b). The crystal structure of fluorcarletonite and carletonite is closely related to those of delhayelite, K4Na2Ca2[AlSi7O19]F2Cl; hydrodelhayelite, KCa2AlSi7O17(OH)2⋅6H2O (Cannillo et al., 1969; Pekov et al., 2009); rhodesite, KNa2Ca2Si8O19(OH)⋅6H2O (Hesse and Liebau, 1992); macdonaldite, BaCa4Si16O36(OH)2⋅10H2O (Cannillo et al., 1968); and monteregianite-(Y), Na4K2Y2Si16O38⋅10H2O (Ghose et al., 1987) (Table S5, Supplement S1). However, all these silicates have branched dreier double layers [(Si, Al)8O19] of the same topology (Liebau, 2012), whereas fluorcarletonite and carletonite contain [Si8O18] layers. Two independent silicate tetrahedra of the single layer form eight-membered rings connected by four-membered rings with four adjacent eight-membered rings. The two sublayers are linked via sharing a common oxygen atom O6 as indicated in Fig. 5a. The size of the eight-membered rings is 7.178(1)×4.664(1) Å, while that of four-membered rings is 4.027(1)×3.203(1) Å. These values are quite close to those found in other silicates from the Murun charoitites (see, for example, Lacalamita et al., 2017) and silicates with heterogeneous frameworks (see, for example, Mesto et al., 2014). There is a further eight-membered ring with a size of 4.466(1)×4.466(1) Å in the straighter windows and 6.144(1)×6.144(1) Å in the wider ones configured with the equivalent eight-membered ring of the adjacent SiO4 layer. These cages host a H2O group (at the O11 site) that is shared by two Na atoms coordinated with the apex of the smaller window.

For the studied fluorcarletonite crystals, SiO4 tetrahedra are quite regular as confirmed by the average Si–O distances (1.608(1) and 1.617(1) Å; Table S2, Supplement S1). The BLD (bond length distortion: Renner and Lehmann, 1986) values for Si1O4 and Si2O4 tetrahedra are ∼0.67 % and ∼1.24 %, respectively, and the TAV (tetrahedral angle variance, Robinson et al., 1971) values are ∼5.9 and 14.5, respectively (Table 4), which shows that the Si2 site has a slightly greater bond length distortion and angle variance. The calculation of the distortion parameters was also performed with the data for carletonite taken from Chao (1972), exhibiting similar relative variations.

Silicate layers are linked by vertices to the Na1 and Na2 octahedra and Ca polyhedra. The Na3 site is surrounded by eight oxygen atoms belonging to carbonate groups. Two symmetrically independent Na-centered octahedra are geometrically different: (1) the Na2 site has longer average Na–O distances than Na1 (∼2.438(2) Å and 2.386(2) Å, respectively; Table S3, Supplement S1) and, as a consequence, the greater octahedral volume (18.58 Å3 vs. 16.64 Å3, respectively; Table 4); (2) the angle variances for Na1 are greater than those for Na2 (OAVs, octahedral angle variances, Robinson et al., 1971, are 209.772 and 86.522 for Na1 and Na2, respectively; Table 4); (3) the BLD for Na2 is greater than that for Na1 (4.513 and 3.685, respectively; Table 4). Concerning the 8-coordinated Na3 and Ca polyhedra, the Na–O and Ca–O distances are 2.498(1) Å and 2.455(1) Å, respectively (Table S3, Supplement S1).

The K+ ions are located in a fully occupied 10-coordinated site within the eight-membered ring of the silicate anion with the average K–O bond length of 2.978(1) Å (Table S3, Supplement S1).

Two independent CO3 groups in the crystal structure are linked to the Na2, Na3 and Ca polyhedra. The C2O3 carbonate group is also bonded to K+ cations. The observed C–O, O–O distances and O–C–O angles of CO3 triangle (Table S3, Supplement S1) are similar to those reported by Chao (1972) for carletonite.

The O atoms of the H2O groups occupy the O11(w) and O12(w) sites and coordinate the Na1 and Na2 sites, respectively (Fig. 4a and b). According to the structure refinement, the O11(w) site is almost completely occupied (0.840(8); Table 3), while the O12(w) site has the occupancy of ∼50 % (Table 3; the distances between two neighboring O12(w) positions is 2.361(3) Å). We could not locate the H sites from the analysis the difference-Fourier maps, presumably due to the disorder and partial occupancy of the H2O sites.

Finally, in fluorcarletonite samples the interatomic distances are very close to those observed for carletonite from the Mont Saint-Hilaire massif as reported by Chao (1972) (see Tables S1 and S2, Supplement S1). The same applies to the distortion parameters for the fluorcarletonite and carletonite samples (Table 4).

Table 5 yields the results of a bond-valence analysis for the sample studied. For calculation of the bond-valence sums (BVSs), the parameters suggested by Brese and O'Keeffe (1991) and Gagnè and Hawthorne (2015) have been used. The BVSs are satisfactory or in some cases somewhat higher than expected, taking into account the absence of contributions from the H sites.