Three-D Mineralogical Mapping of the Kovdor Phoscorite-Carbonatite Complex, NW Russia: III. Pyrochlore Supergroup Minerals

: The pyrochlore supergroup minerals (PSM) are typical secondary phases that replace (with zirconolite–laachite) earlier Sc-Nb-rich baddeleyite under the inﬂuence of F-bearing hydrothermal solutions, and form individual well-shaped crystals in surrounding carbonatites. Like primary Sc-Nb-rich baddeleyite, the PSM are concentrated in the axial carbonate-rich zone of the phoscorite-carbonatite complex, so their content, grain size and chemical diversity increase from the pipe margins to axis. There are 12 members of the PSM in the phoscorite-carbonatite complex. Fluorine- and oxygen-dominant phases are spread in host silicate rocks and marginal carbonate-poor phoscorite, while hydroxide-dominant PSM occur mainly in the axial carbonate-rich zone of the ore-pipe. Ti-rich PSM (up to oxycalciobetaﬁte) occur in host silicate rocks and calcite carbonatite veins, and Ta-rich phases (up to microlites) are spread in intermediate and axial magnetite-rich phoscorite. In marginal (apatite)-forsterite phoscorite, there are only Ca-dominant PSM, and the rest of the rocks include Ca-, Na- and vacancy-dominant phases. The crystal structures of oxycalciopyrochlore and hydroxynatropyrochlore were reﬁned in the Fd 3 m space group with R 1 values of 0.032 and 0.054 respectively. The total difference in scattering parameters of B sites are in agreement with substitution scheme B Ti 4+ + Y OH - = B Nb 5+ + Y O 2- . The perspective process ﬂow diagram for rare-metal “anomalous ore” processing includes sulfur-acidic cleaning of baddeleyite concentrate from PSM and zirconolite–laachite impurities followed by deep metal recovery from baddeleyite concentrate and Nb-Ta-Zr-U-Th-rich sulfatic product from its cleaning. Single-crystal X-ray diffraction studies of oxycalciopyrochlore and hydroxynatropyrochlore were performed at 293 K using a Bruker Kappa APEX DUO diffractometer equipped with the I µ S microfocus source (beam size of 0.11 mm, Mo K α radiation, λ = 0.71073 Å and operated at 45 and 0.6 and a CCD area detector. The intensity data was reduced and corrected for Lorentz, polarization and background effects. The APEX2 software Billerica, MA, USA, 2014) a multi-scan absorption-correction. Crystal structures were reﬁned


Introduction
The Kovdor phoscorite-carbonatite complex is the largest source of magnetite, hydroxylapatite and baddeleyite in the NW of Russia [1][2][3][4]. Besides, the deposit contains significant amounts of Sc, Ln, Ta, Nb and U concentrated mainly in baddeleyite and the products of its alteration, first of all, in zirconolite and minerals of the pyroclore supergroup [5][6][7][8][9][10]. Therefore, it is important to highlight

Materials and Methods
For this study, we used 548 samples of phoscorite, carbonatites and host rocks taken from 108 exploration holes drilled at the interval from −80 to −650 m within the Kovdor phoscorite-carbonatite complex [16]. Bulk-rock samples were analyzed in the Tananaev Institute of Chemistry of KSC RAS (Apatity) by means of inductively coupled plasma-mass spectrometry (ICP-MS) performed with an ELAN 9000 DRC-e mass spectrometer (Perkin Elmer, Waltham, MA, USA). For the analyses, the samples were dissolved in a mixture of concentrated hydrofluoric and nitric acids with distillation of silicon and further addition of hydrogen peroxide to a cooled solution to suppress hydrolysis of polyvalent metals [19].
Composition of PSM grains smaller than 20 µm in diameter was determined using a LEO-1450 scanning electron microscope (Carl Zeiss Microscopy, Oberkochen, Germany) with a Quantax 200 energy-dispersive X-ray spectrometer (Bruker, Ettlingen, Germany). The same equipment with 500-pA beam current and 20 kV acceleration voltage was used to obtain back-scattered electron Grains larger 20 µm in diameter were then studied using a Cameca MS-46 electron probe microanalyzer (EPMA) (Cameca, Gennevilliers, France) operating in wavelength-dispersive mode at 20 kV and 20-30 nA. The analyzes were performed with the beam size of 5-10 µm and the counting time of 10-20/10 s on peaks/background for every chemical element and every of 5-10 measurement points. Table 1 presents the used standards, detection limits and precisions based on repeated analyses of standards. Fluorine was determined with a Quantax 200 energy-dispersion instrument and standard-less ZAF method, based on a detection limit of 0.5 wt %. Coefficients in crystallochemical formulas were calculated using the MINAL program by D. V. Dolivo-Dobrovolsky [20]. Single-crystal X-ray diffraction studies of oxycalciopyrochlore and hydroxynatropyrochlore were performed at 293 K using a Bruker Kappa APEX DUO diffractometer equipped with the IµS microfocus source (beam size of 0.11 mm, MoKα radiation, λ = 0.71073 Å and operated at 45 kV and 0.6 mA) and a CCD area detector. The intensity data was reduced and corrected for Lorentz, polarization and background effects. The APEX2 software (Bruker-AXS, Billerica, MA, USA, 2014) applied a multi-scan absorption-correction. Crystal structures were refined with the SHELX program [21] and drawn using the VESTA 3 program [22].
Raman spectra of 6 typical PSM were produced using a Jobin-Yvon LabRam HR 800 spectrometer (Horiba, Kyoto, Japan) with a 514 nm laser under the same conditions for each sample. The band component analysis was performed using the OriginPro 8.1 SR2 program [23], with Lorenzian peak function.
Statistical analyses were implemented with the STATIATICA 8.0 (StatSoft) program [24]. Geostatisical studies and 3D modeling were conducted with the MICROMINE 16 program [25]. Interpolation was performed by ordinary kriging. Automatic 3D geological mapping was developed by conversion of rock chemical composition to mineral composition using logical computation [26].

Occurrence and Morphology
The pyrochlore supergroup minerals are common accessories in all rocks of the Kovdor alkaline-ultrabasic massif. The occurrence and content of PSM decrease from host foidolite and diopsidite-phlogopitite to earlier (apatite)-forsterite phoscorite, and then increase to intermediate phoscorite and, finally, to the latest carbonatites (Table 2). In the phoscorite-carbonatite complex, the pyrochlore supergroup minerals result mainly from the alteration of Nb-rich baddeleyite formed in the pipe axial zone due to the substitution 2Zr 4+ ↔ Sc 3+ Nb 5+ . On this reason, the pyrochlore areal coincides with that of Sc-Nb-rich baddeleyite ( Figure 1). Carbonate-rich phoscorite and carbonatites enriched in PSM form an intensive radioactive anomaly (about 200 m in diameter and >900 m in depth), known as the "Anomalous Zone" [1]. Content of pyrochlore in rocks of the "Anomalous Zone" gradually increases with depth at the account of baddeleyite [27].

Occurrence and Morphology
The pyrochlore supergroup minerals are common accessories in all rocks of the Kovdor alkalineultrabasic massif. The occurrence and content of PSM decrease from host foidolite and diopsiditephlogopitite to earlier (apatite)-forsterite phoscorite, and then increase to intermediate phoscorite and, finally, to the latest carbonatites ( Table 2). In the phoscorite-carbonatite complex, the pyrochlore supergroup minerals result mainly from the alteration of Nb-rich baddeleyite formed in the pipe axial zone due to the substitution 2Zr 4+ ↔ Sc 3+ Nb 5+ . On this reason, the pyrochlore areal coincides with that of Sc-Nb-rich baddeleyite ( Figure 1). Carbonate-rich phoscorite and carbonatites enriched in PSM form an intensive radioactive anomaly (about 200 m in diameter and >900 m in depth), known as the "Anomalous Zone" [1]. Content of pyrochlore in rocks of the "Anomalous Zone" gradually increases with depth at the account of baddeleyite [27].   The pyrochlore group minerals, namely oxycalciopyrochlore (OCP), oxynatropyrochlore (ONP), hydroxycalciopyrochlore (HCP), hydroxynatropyrochlore (HNP), hydroxykenopyrochlore (HKP), fluorcalciopyrochlore (FCP), fluornatropyrochlore (FNP) and fluorkenopyrochlore (FKP), are common in foidolite, diopsidite, phlogopitite, phoscorite and carbonatite (OCP, HCP and HNP are predominant). Minerals of microlite and betafite groups play a subordinate role, mainly, as separate parts of PSM inhomogeneous crystals. Oxycalciobetafite (OCB) and its cation-deficient analogue occur in phoscorite and carbonatite of the ore-pipe intermediate and axial zones; while fluorcalciomicrolite (FCM), hydroxycalciomicrolite (HCM) and hydroxykenomicrolite (HKM) are found only in calcite-rich phoscorite and calcite carbonatite.
Within the phoscorite-carbonatite complex, the equivalent circle diameter of the analyzed PSM grains widely varies from 1 to 1000 µm (see Table 2) increasing from host foidolite and diopsidite to marginal (apatite)-forsterite phoscorite, and then to intermediate low-carbonate magnetite-rich phoscorite and axial calcite-rich phoscorite and carbonatite (see Figure 1c). In fact, the PSM grain size is proportional to their content in the rock: in PSM-poor rocks, there are small (up to 100 µm in diameter) separate grains of these minerals; while PSM-rich rocks contain irregularly-shaped and lens-like segregations, bands and veinlets of much larger (up to 1 mm in diameter) crystals of the pyrochlore supergroup minerals.
The morphology and microstructure of pyrochlore particles show different aspects, such as irregularly shaped and rounded grains (Figure 2a (Figure 2g), rims around baddeleyite and lueshite grains followed by partial pseudomorphs (Figure 2h), epitaxial intergrowth with baddeleyite (Figure 2i), as well as the finest inclusions in exsolved titanomagnetite. In all rocks of the phoscorite-carbonatite complex, there are zoned PSM crystals with primary regular zonation corresponding to crystal shape, secondary irregular zoning caused by mineral alteration, and both of these zoning types, which is typical for PSM from carbonatites [28]. The primary zoning is caused usually by increases of Ca, Ta, Ti, Zr and F contents from the crystal core to rim at the expense of Na, Th, U, REE, Nb and (OH) amounts. Secondary zoning appears, first of all, due to leaching of Na, Ca and F from marginal parts of metamict grains of U-Th-rich PSM. Sometimes, separate zones of one crystal are formed by different minerals of the pyrochlore supergroup (see Figure 2f).
Irregularly shaped and drop-like yellowish-brown pyrochlore grains dominate in host silicate rocks and carbonate-poor phoscorite (see Figure 2a); while in carbonate-rich phoscorite and related carbonatite, they also form dark brown truncated octahedral crystals, sometimes with baddeleyite relics (see Figure 2e,i). In calcite carbonatite veins, there are reddish-brown to creamy and yellow octahedral, cubic, truncated octahedral and cubooctahedral crystals of pyrochlores (see Figure 2c), and their close intergrowth with baddeleyite and zirconolite [1,4,29,30]. In vein dolomite carbonatite, yellow to brown truncated octahedral to cubic pyrochlore crystals (see Figure 2d) occur in voids in typical association with zircon and endemic phosphates [1,27,31,32]. Betafite group minerals occur as separate irregularly shaped, drop-like and ellipsoidal grains (up to 100 µm) or form marginal zones of pyrochlore crystals (see Figure 2d) and rims around baddeleyite grains. Microlite group members form marginal zones of pyrochlore crystals and inclusions (up to 20 µm in diameter) in zirconolite, titanomagnetite and exsolution ilmenite.
Like PSM, minerals of the zirconolite-laachite series are typical products of baddeleyite alteration; however, they can also replace pyrochlores ( Figure 3a). Besides, grains of U-Th-rich kenopyrochlores are often replaced/rimmed by pyrite ( Figure 3b) and valleriite ( Figure 3c) and sometimes by strontianite (Figure 3d), probably, due to radiolytic splitting of water into hydrogen peroxide and molecular hydrogen. In particular, with radiation dose growth, a water solution of H 2 SO 4 (below 200 • C) becomes significantly rich in H 2 due to H 2 O 2 [33,34], and this, probably, causes precipitation of sulfides around the radiation source.  Table 3 shows the results of precision EPMA analyses conducted on 12 different members of the pyrochlore supergroup found in the Kovdor massif, and Table 4 presents statistical data on the PSM composition in different rocks of this massif. Most of the analyzed grains (92%) correspond to Ca-, Na-and vacancy-dominant members of the pyrochlore group and the rest 8% are represented by minerals of microlite and betafite groups (about 4% of each). Pyrochlores occur in all rocks of the Kovdor massif (Figure 4), and microlites and betafites co-exist with pyrochlore in the central part of the phoscorite-carbonatite complex.  [11] and relations between Nb, Ta and Ti contents.  Table 3 shows the results of precision EPMA analyses conducted on 12 different members of the pyrochlore supergroup found in the Kovdor massif, and Table 4 presents statistical data on the PSM composition in different rocks of this massif. Most of the analyzed grains (92%) correspond to Ca-, Na-and vacancy-dominant members of the pyrochlore group and the rest 8% are represented by minerals of microlite and betafite groups (about 4% of each). Pyrochlores occur in all rocks of the Kovdor massif (Figure 4), and microlites and betafites co-exist with pyrochlore in the central part of the phoscorite-carbonatite complex.  Table 3 shows the results of precision EPMA analyses conducted on 12 different members of the pyrochlore supergroup found in the Kovdor massif, and Table 4 presents statistical data on the PSM composition in different rocks of this massif. Most of the analyzed grains (92%) correspond to Ca-, Na-and vacancy-dominant members of the pyrochlore group and the rest 8% are represented by minerals of microlite and betafite groups (about 4% of each). Pyrochlores occur in all rocks of the Kovdor massif (Figure 4), and microlites and betafites co-exist with pyrochlore in the central part of the phoscorite-carbonatite complex.   [11] and relations between Nb, Ta and Ti contents. Table 3. Precision microprobe analyses of pyrochlore supergroup minerals (wt %) (see abbreviations above) and the corresponding crystallochemical formulas calculated on the basis of B = 2, X = 6 and Y = 1 atoms per formula unit (apfu).     6.00 6.00 6.00 6.00 6.00 6.00 6.00 6.00 6.00 6.00 6.00 6.00       Over 92% of the analyzed PSM are represented by O-and (OH)-dominant phases, and the rest are fluorine-dominant. Fluorine shows positive correlations with Ca and Nb and negative correlations with Ti and U ( Figure 5). Correspondingly, fluorine-dominant phases occur mainly among calciopyroclores; while natropyrochlores, betafites and microlites are usually represented by hydroxyl-dominant and sometimes oxygen-dominant phases. As for pyrochlore, Ca-dominant phases prevail in ≈45% of the analyzed samples; about 20% of the samples include Na-dominant pyrochlore, and the rest are formed by kenopyrochlore (Figure 6a) as a result of heterovalent substitutions: 2Ca 2+ → Na + REE 3+ , 2Ca 2+ → U 4+ , Ca 2+ O 2− → Na + (OH) − , etc. Simultaneously, Nb is replaced with Ta and Ti up to the formation of microlite and betafite.

PSM
In about 35% of the analyzed PSM, the sum of cations in the A-position does not exceed 1 apfu (see Figure 6a). Kenopyrochlores occur in 59% of the analyzed samples of vein calcite carbonatite, 45% of diopsidite and phlogopitite samples, 25% of phoscorite samples, and 23% of vein dolomite carbonatite samples. The deficit of cations in the A-position is caused by both the presence of high-charge cations of U 4+ , Th 4+ , REE 3+ instead of Na + and Ca 2+ (4N a+ ↔ 3 U 4+ , 3Ca 2+ ↔ 2REE 3+ , etc.), and cation loss during pyrochlore metamictization and hydration. The last processes are typical for pyrochlore with >15 wt % of (U,Th)O 2 , and accompanied by destruction of the mineral crystal structure and leaching of Na, then Ca, and finally REE, U and Th (Figure 6b). In kenopyrochlore, uranium is a predominant high-charge cation; while Th-dominant phases occur much rarely (Figure 6c). Rare-earth elements are represented mainly by light lanthanides La through Nd, with total average content of La and Ce of about 87%. The highest REE content in pyrochlore is typical for host diopsidite and phlogopitite, as well as vein calcite carbonatite.  Figure 5). Correspondingly, fluorine-dominant phases occur mainly among calciopyroclores; while natropyrochlores, betafites and microlites are usually represented by hydroxyl-dominant and sometimes oxygen-dominant phases. As for pyrochlore, Ca-dominant phases prevail in ≈45% of the analyzed samples; about 20% of the samples include Na-dominant pyrochlore, and the rest are formed by kenopyrochlore (Figure 6a) as a result of heterovalent substitutions: 2Ca 2+ → Na + REE 3+ , 2Ca 2+ → ☐U 4+ , Ca 2+ O 2− → Na + (OH) − , etc. Simultaneously, Nb is replaced with Ta and Ti up to the formation of microlite and betafite.
In about 35% of the analyzed PSM, the sum of cations in the A-position does not exceed 1 apfu (see Figure 6a). Kenopyrochlores occur in 59% of the analyzed samples of vein calcite carbonatite, 45% of diopsidite and phlogopitite samples, 25% of phoscorite samples, and 23% of vein dolomite carbonatite samples. The deficit of cations in the A-position is caused by both the presence of highcharge cations of U 4+ , Th 4+ , REE 3+ instead of Na + and Ca 2+ (4N a+ ↔ 3☐U 4+ , 3Ca 2+ ↔ ☐2REE 3+ , etc.), and cation loss during pyrochlore metamictization and hydration. The last processes are typical for pyrochlore with >15 wt % of (U,Th)O2, and accompanied by destruction of the mineral crystal structure and leaching of Na, then Ca, and finally REE, U and Th (Figure 6b). In kenopyrochlore, uranium is a predominant high-charge cation; while Th-dominant phases occur much rarely ( Figure  6c). Rare-earth elements are represented mainly by light lanthanides La through Nd, with total average content of La and Ce of about 87%. The highest REE content in pyrochlore is typical for host diopsidite and phlogopitite, as well as vein calcite carbonatite.    Figure 5). Correspondingly, fluorine-dominant phases occur mainly among calciopyroclores; while natropyrochlores, betafites and microlites are usually represented by hydroxyl-dominant and sometimes oxygen-dominant phases. As for pyrochlore, Ca-dominant phases prevail in ≈45% of the analyzed samples; about 20% of the samples include Na-dominant pyrochlore, and the rest are formed by kenopyrochlore (Figure 6a) as a result of heterovalent substitutions: 2Ca 2+ → Na + REE 3+ , 2Ca 2+ → ☐U 4+ , Ca 2+ O 2− → Na + (OH) − , etc. Simultaneously, Nb is replaced with Ta and Ti up to the formation of microlite and betafite.
In about 35% of the analyzed PSM, the sum of cations in the A-position does not exceed 1 apfu (see Figure 6a). Kenopyrochlores occur in 59% of the analyzed samples of vein calcite carbonatite, 45% of diopsidite and phlogopitite samples, 25% of phoscorite samples, and 23% of vein dolomite carbonatite samples. The deficit of cations in the A-position is caused by both the presence of highcharge cations of U 4+ , Th 4+ , REE 3+ instead of Na + and Ca 2+ (4N a+ ↔ 3☐U 4+ , 3Ca 2+ ↔ ☐2REE 3+ , etc.), and cation loss during pyrochlore metamictization and hydration. The last processes are typical for pyrochlore with >15 wt % of (U,Th)O2, and accompanied by destruction of the mineral crystal structure and leaching of Na, then Ca, and finally REE, U and Th (Figure 6b). In kenopyrochlore, uranium is a predominant high-charge cation; while Th-dominant phases occur much rarely ( Figure  6c). Rare-earth elements are represented mainly by light lanthanides La through Nd, with total average content of La and Ce of about 87%. The highest REE content in pyrochlore is typical for host diopsidite and phlogopitite, as well as vein calcite carbonatite.   To determine the chemical evolution of the PSM in natural rock sequence, we implemented factor analysis of 314 PSM compositions using the method of principal components (with normalization and varimax rotation of factors) ( Table 5). The resultant factors enable us to specify five schemas of isomorphic substitutions (elements with high factor loadings are bolded): Correspondingly, in the natural sequence of the Kovdor rocks (Figure 7), content of Ba, Sr, U, Ta, Fe and Zr in PSM gradually increases due to Na, Ca, REE, Th and Nb (F2-F5), while higher contents of Ti, U and vacancies are observed in host silicate and calcio-carbonatite rocks (F1). To determine the chemical evolution of the PSM in natural rock sequence, we implemented factor analysis of 314 PSM compositions using the method of principal components (with normalization and varimax rotation of factors) ( Table 5). The resultant factors enable us to specify five schemas of isomorphic substitutions (elements with high factor loadings are bolded): (1) Na + Ca 2+ Nb 5+ ↔ U 4+ Ti 4+ ; (2) (Th 4+ , REE 3+ ) ↔ U 4+ ; (3) Na + Nb 5+ ↔ Sr 2+ Zr 4+ ; (4) Nb 5+ ↔ (Ta 5+ , Fe 3+ ); (5) Ca 2+ ↔ Ba 2+ . Correspondingly, in the natural sequence of the Kovdor rocks (Figure 7), content of Ba, Sr, U, Ta, Fe and Zr in PSM gradually increases due to Na, Ca, REE, Th and Nb (F2-F5), while higher contents of Ti, U and vacancies are observed in host silicate and calcio-carbonatite rocks (F1).   Table 5) in natural sequence of the Kovdor rocks.
All above substitutions cause complex zonation of the ore-pipe in terms of the PSM composition (   Table 5) in natural sequence of the Kovdor rocks.
All above substitutions cause complex zonation of the ore-pipe in terms of the PSM composition (  In the unaltered PSM crystals from diopsidite, phlogopitite, phoscorite and calcite carbonatite, there are irregular variations of chemical composition between separate zones, without any clear trends from cores to margins. However, in host foidolite, marginal zones of pyrochlore crystals are constantly enriched in Ca, Na, Th and Nb in comparison with REE-U-Ta-Ti-rich cores, and the latest dolomite carbonatite, pyrochlore grains have Na-Nb-dominant cores and Ca-U-Zr-rich margins. In the unaltered PSM crystals from diopsidite, phlogopitite, phoscorite and calcite carbonatite, there are irregular variations of chemical composition between separate zones, without any clear trends from cores to margins. However, in host foidolite, marginal zones of pyrochlore crystals are constantly enriched in Ca, Na, Th and Nb in comparison with REE-U-Ta-Ti-rich cores, and the latest dolomite carbonatite, pyrochlore grains have Na-Nb-dominant cores and Ca-U-Zr-rich margins. Besides, U-Th-rich (keno)pyrochlore grains often have secondary zonation due to leaching of A-cations and then B-cations from the grain marginal parts. Fluorine content increases from the core to rim in fresh pyrochlore crystals (Figure 9a). The PSM alteration under the influence of self-irradiation causes loss of fluorine ( Figure 9b); therefore, the content of fluorine in the PSM is directly proportional to the amounts of Ca and Nb, and inversely proportional to the content of U and Th (see Figure 5).
X-ray powder diffraction of the Kovdor PSM showed good crystallinity of U/Th-poor calcio-and natropyrochlores; while all analyzed kenopyrochlores became amorphous. Therefore, a single-crystal X-ray study was performed only for low-vacant pyrochlore. Besides, U-Th-rich (keno)pyrochlore grains often have secondary zonation due to leaching of Acations and then B-cations from the grain marginal parts. Fluorine content increases from the core to rim in fresh pyrochlore crystals (Figure 9a). The PSM alteration under the influence of self-irradiation causes loss of fluorine ( Figure 9b); therefore, the content of fluorine in the PSM is directly proportional to the amounts of Ca and Nb, and inversely proportional to the content of U and Th (see Figure 5). X-ray powder diffraction of the Kovdor PSM showed good crystallinity of U/Th-poor calcio-and natropyrochlores; while all analyzed kenopyrochlores became amorphous. Therefore, a single-crystal X-ray study was performed only for low-vacant pyrochlore.

Crystal Structure
Single-crystal X-ray diffraction data were obtained for well-crystalline oxycalciopyrochlore 917-318.5 (see Figure 2e) and hydroxynatropyrochlore K-017-4 (see Figure 2d). A quadrant of threedimensional data was collected with frame widths of 1° in ω, and with 220 s in the range 2θ 6.8°-55°. Scattering factors were calculated from initial model with all Ca and Na at A, all Nb and Ti at B, O at O1, and O at Y1 site. All cation site-occupancies are given in accordance with electron-microprobedetermined values (normalized on a basis of 2B cations per formula unit), the O-populated sites (O1 and Y1) was set at full occupancy. Table 6 shows data collection and refinement parameters for the single-crystal X-ray experiments. Atom coordinates, displacement parameters and site occupancies are given in Table 7, and bond lengths in Table 8. Anisotropic displacement parameters are attached in Supplementary Materials (CIF data is available).

Crystal Structure
Single-crystal X-ray diffraction data were obtained for well-crystalline oxycalciopyrochlore 917-318.5 (see Figure 2e) and hydroxynatropyrochlore K-017-4 (see Figure 2d). A quadrant of three-dimensional data was collected with frame widths of 1 • in ω, and with 220 s in the range 2θ 6.8 • -55 • . Scattering factors were calculated from initial model with all Ca and Na at A, all Nb and Ti at B, O at O1, and O at Y1 site. All cation site-occupancies are given in accordance with electron-microprobe-determined values (normalized on a basis of 2B cations per formula unit), the O-populated sites (O1 and Y1) was set at full occupancy. Table 6 shows data collection and refinement parameters for the single-crystal X-ray experiments. Atom coordinates, displacement parameters and site occupancies are given in Table 7, and bond lengths in Table 8. Anisotropic displacement parameters are attached in Supplementary Materials (CIF data is available).
Crystal structures of oxycalciopyrochlore 917/318.5 and hydroxynatropyrochlore K-017-4 ( Figure 10 1.00 respectively. The mean A1-O bond ranges from 2.530 to 2.533 Å, which is more suitable for ideal Ca-O distance 2.54 Å than for ideal Na-O distance 2.60 Å [35]. For hydroxynatropyrochlore K-017-4, displacement parameters for the Y site are slightly higher than those for oxygen, and are consistent with its occupancy by (OH)-groups (Table 7); while for oxycalciopyrochlore 917-318-5, displacement parameters for Y1 and O1 are almost equal.

Raman Spectroscopy
The PSM Raman spectra obtained under the same conditions showed significantly different observed intensities Iobs of absorption bands (Figure 11a). The observed intensity of the spectrum depends on mineral crystallinity, which, in turn, gradually decreases with growth of U and Th total content from 0.13 apfu in HNP 972/86.9 to 0.47 apfu in ONP 966/62.9 (Table 9). For the same reason, the stability of the UTh-rich PSM decreases under the influence of laser beams. Corrected absorption band intensities I were calculated as I = Iobs/(n(ω) + 1), where n(ω) is the Bose (Einstein factor). Figure  11b shows the results for a zoned HKP-HNP crystal from calcite carbonatite 972/86.9.
The absorption bands (see Table 9) were assigned by analogy with other pyrochlore-like compounds [36][37][38] taking into account theoretical considerations by McCauley [39] and Arenas et al. [40]. According to McCauley [39], pyrochlore yields six Raman-active modes and one acoustic. These modes involve four vibrations of F1u, F2g, Eg, and A1g symmetry. Theoretical calculations by [40] (Figure 12), which enables us to estimate the content of major impurities using the PSM Raman spectra.

Raman Spectroscopy
The PSM Raman spectra obtained under the same conditions showed significantly different observed intensities I obs of absorption bands (Figure 11a). The observed intensity of the spectrum depends on mineral crystallinity, which, in turn, gradually decreases with growth of U and Th total content from 0.13 apfu in HNP 972/86.9 to 0.47 apfu in ONP 966/62.9 (Table 9). For the same reason, the stability of the UTh-rich PSM decreases under the influence of laser beams. Corrected absorption band intensities I were calculated as I = I obs /(n(ω) + 1), where n(ω) is the Bose (Einstein factor). Figure 11b shows the results for a zoned HKP-HNP crystal from calcite carbonatite 972/86.9.

Discussion
Within the Kovdor phoscorite-carbonatite complex, the PSM are concentrated in the axial carbonate-rich zone, so their content, grain size and diversity increase from the pipe margins towards its axis (see Figure 1). Besides, in this direction, we observe gradual growth of Ba, Sr, U, Ta, Fe and Zr content in PSM due to Na, Ca, REE, Th and Nb amounts (see Figures 7 and 8). This trend is the reverse of known ranges of carbonate/silicate melt partition coefficients [41][42][43][44]: Al < Si < Ti < Fe < Mg < K < Na < Ca < F < P < CO 2 (for major elements), and Hf < Zr < Th < U < Ta < Y < Nb < Nd < Sr < Ba (for trace elements). In other words, we can assume that most of the high-field-strength elements (HFSE: Ti, Zr, Hf, Nb, Ta and Y) are predominantly distributed in silicate melt; while alkaline-earth and most of rare-earth elements will be localized in carbonate melt. If we compare the earliest silicate rocks (including forsterite-dominant phoscorite) and the latest carbonatite veins (Table 10), we can see that this assumption is correct. However, the highest concentrations of Zr, Hf, Ta and F as well as the local maximum of Nb content characterize magnetite-rich phoscorite, and can be caused by initial concentration of HFSE in Mg-Al-Ti-rich magnetite. Magnetite exsolution produces ScNb-rich baddeleyite ( Figure 13) as a co-product of spinel and ilmenite-geikielite [45], and later alteration of such baddeleyite with fluorine-bearing hydrothermal solutions enriches them in Zr, Nb and Sc. When the concentration of HFSE in the solution reaches a critical level, the PSM start crystallizing. One part of the PSM is formed in situ as rims around baddeleyite grains, and the other part is crystallized as individual crystals in carbonatites that act as a geochemical barrier for Nb(OH) 3 F 2 , Ta(OH) 3   Since Sc-Nb-rich baddeleyite and U-Ta-rich PSM of the ore-pipe axial zone contain most of the Kovdor's Nb, Ta, U, as well as a major part of Sc, this "Anomalous" zone can be regarded as a complex rare-metal deposit [6][7][8]. Prevailing close intergrowths of Sc-rich baddeleyite and U-rich kenopyrochlore significantly complicate the production of a high-quality baddeleyite concentrate. A perspective approach for the deposit development includes selective mining of "anomalous ore", sulfur-acidic cleaning of baddeleyite concentrate from pyrochlore and zirconolite impurities ( Figure  14a), followed by deep metal recovery from baddeleyite concentrate (Figure 14b) and Nb-Ta-Zr-U-Th-rich sulfatic product of its cleaning [49][50][51].  [49]) and deep processing (b, after [50,51]). Gray rectangles show final products.

Conclusions
According to mineralogical, geochemical, crystallochemical and spectroscopic data obtained for the PSM from the Kovdor phoscorite-carbonatite complex, we can make the following conclusions: (1) High-temperature magmatic magnetite is a primary concentrator of Zr, Ti, Nb and Sc in the Kovdor phoscorite-carbonatite complex. The magnetite exsolution under cooling produces spinel and ilmenite-geikielite inclusions containing, in turn, the smallest baddeleyite particles. Besides, separate baddeleyite crystals are crystallized within and around magnetite grains due to their self-cleaning from listed impurities. The content of Nb and Sc in baddeleyite gradually increase from marginal (apatite)-forsterite phoscorite to axial carbonate-magnetite-rich Since Sc-Nb-rich baddeleyite and U-Ta-rich PSM of the ore-pipe axial zone contain most of the Kovdor's Nb, Ta, U, as well as a major part of Sc, this "Anomalous" zone can be regarded as a complex rare-metal deposit [6][7][8]. Prevailing close intergrowths of Sc-rich baddeleyite and U-rich kenopyrochlore significantly complicate the production of a high-quality baddeleyite concentrate. A perspective approach for the deposit development includes selective mining of "anomalous ore", sulfur-acidic cleaning of baddeleyite concentrate from pyrochlore and zirconolite impurities (Figure 14a), followed by deep metal recovery from baddeleyite concentrate (Figure 14b) and Nb-Ta-Zr-U-Th-rich sulfatic product of its cleaning [49][50][51]. Since Sc-Nb-rich baddeleyite and U-Ta-rich PSM of the ore-pipe axial zone contain most of the Kovdor's Nb, Ta, U, as well as a major part of Sc, this "Anomalous" zone can be regarded as a complex rare-metal deposit [6][7][8]. Prevailing close intergrowths of Sc-rich baddeleyite and U-rich kenopyrochlore significantly complicate the production of a high-quality baddeleyite concentrate. A perspective approach for the deposit development includes selective mining of "anomalous ore", sulfur-acidic cleaning of baddeleyite concentrate from pyrochlore and zirconolite impurities ( Figure  14a), followed by deep metal recovery from baddeleyite concentrate (Figure 14b) and Nb-Ta-Zr-U-Th-rich sulfatic product of its cleaning [49][50][51]. Figure 14. Process flow diagrams of the PSM-bearing baddeleyite concentrate acidic cleaning (a, after [49]) and deep processing (b, after [50,51]). Gray rectangles show final products.

Conclusions
According to mineralogical, geochemical, crystallochemical and spectroscopic data obtained for the PSM from the Kovdor phoscorite-carbonatite complex, we can make the following conclusions: (1) High-temperature magmatic magnetite is a primary concentrator of Zr, Ti, Nb and Sc in the Kovdor phoscorite-carbonatite complex. The magnetite exsolution under cooling produces spinel and ilmenite-geikielite inclusions containing, in turn, the smallest baddeleyite particles. Besides, separate baddeleyite crystals are crystallized within and around magnetite grains due to their self-cleaning from listed impurities. The content of Nb and Sc in baddeleyite gradually increase from marginal (apatite)-forsterite phoscorite to axial carbonate-magnetite-rich Figure 14. Process flow diagrams of the PSM-bearing baddeleyite concentrate acidic cleaning (a, after [49]) and deep processing (b, after [50,51]). Gray rectangles show final products.

Conclusions
According to mineralogical, geochemical, crystallochemical and spectroscopic data obtained for the PSM from the Kovdor phoscorite-carbonatite complex, we can make the following conclusions: (1) High-temperature magmatic magnetite is a primary concentrator of Zr, Ti, Nb and Sc in the Kovdor phoscorite-carbonatite complex. The magnetite exsolution under cooling produces spinel and ilmenite-geikielite inclusions containing, in turn, the smallest baddeleyite particles. Besides, separate baddeleyite crystals are crystallized within and around magnetite grains due to their self-cleaning from listed impurities. The content of Nb and Sc in baddeleyite gradually increase from marginal (apatite)-forsterite phoscorite to axial carbonate-magnetite-rich phoscorite and carbonatite, which is made possible following formation of apo-baddeleyite Nb-Sc-minerals including PSM; (2) The PSM are secondary minerals that replace (together with zirconolite-laachite) grains of Sc-Nb-rich baddeleyite and use them as seed crystals. Content, grain size and chemical diversity of PSM increase from the pipe margins to axis against the background of a gradual decreasing of temperature of subsolidus processes. In particular, Ca-(Nb,Ti)-F-rich PSM are spread in marginal (apatite)-forsterite phoscorite, Na-(Nb,Ta)-OH-rich phases occur mainly in intermediate low-carbonate magnetite-rich phoscorite, and U-(Nb,Ti)-OH-rich PSM are localized in axial carbonate-rich phoscorite and carbonatites. Subsolidus PSM crystals usually have primary zoning, with increases of Ca, Ta, Ti, Zr and F contents from core to rim at the expense of Na, Th, U, REE, Nb and (OH) amounts; (3) In addition to comparatively high-temperature subsolidus PSM, there are hydrothermal (Na,Ca)-OH-rich pyrochlores that form well-shaped crystals in voids and fractures (in characteristic association with low-temperature Sc-phosphates). Primary zoning of hydrothermal PSM crystals is characterized by growth of U and F contents at the expense of Na, Ca and (OH) amounts. Besides, high U and Th contents cause radiation destruction of the PSM crystal structure, with the following loss of Na, Ca and F under the influence of hydrothermal solutions and the formation of corresponding secondary zoning.