Importance of Ca-rich fluids for the formation of secondary Sc-minerals. An example from a metaluminous pegmatite Kožichovice II, Czech Republic

Scandium is considered as a critical metal, owing to its high demand in technologically strategic materials in next-generation industries (e.g. COM, 2020; Sovacool et al., 2020). It is a relatively rare transitional metal with an average concentration of 25.3 and 7 ppm in the lower and upper continental crust, respectively (Wedepohl, 1995). According to the International Union of Pure and Applied Chemistry (IUPAC) (Connelly et al., 2005), Sc belongs with yttrium and lanthanides (Ln) in the rare earth elements (REE), however its ionic radius (^VIII0.87 Å; ^VI0.745 Å) is substantially smaller than the ionic radius of Ln and Y (^VIII1.16–0.98 Å). Consequently scandium substitutes for smaller ions of similar ionic radii (Fe, Mg, Zr, Sn, Ti and Al; see Shannon, 1976) and is easily incorporated into certain rock-forming (pyroxene, garnet, amphibole, biotite; Neumann, 1961; Eby, 1973; Glassley and Piper, 1978; Chassé et al., 2018; Gion, 2020) and accessory (zircon, columbite, wolframite, beryl, epidote, tourmaline) minerals (Das et al., 1971; Novák et al., 2008, 2011; Přikryl et al., 2014; Čopjaková et al., 2015; Výravský et al., 2017; Williams-Jones and Vasyukova, 2018). Hence, individual scandium minerals and economic-grade deposits are very rare. To date, 23 minerals approved by the International Mineralogical Association (IMA) with Sc as an essential component have been described (Mindat.org, 2024). Four of them are known only from meteorites (Ma et al., 2014, 2015, and references therein), and eight of them contain Ca as an essential element in their crystal structure. On the Earth, Sc-minerals are known from several geological environments including: carbonatites (Amli, 1977; Liferovich et al., 1998), high-temperature skarns (Galuskina et al., 2005), hydrothermal phosphate veins (Bernhard et al., 1998), diagenetic ironstone (Moëlo et al., 2002), and various supergene to very low-temperature hydrothermal systems (Frondel et al., 1968; Dill et al., 2006; Spandler et al., 2004). However, granitic pegmatites are by far the most important environment for both the diversity of species and the size of the individual crystals. Elevated concentrations of Sc and the presence of Sc-minerals are typical for pegmatites of the NYF geochemical family (sensu Černý and Ercit, 2005); however, pretulite (ScPO₄) as a decomposition product of Sc-rich zircon and ixiolite from an LCT-type pegmatite (sensu Černý and Ercit, 2005) has been investigated by Výravský et al. (2017). Thortveitite, monoclinic (C2/m) Sc₂Si₂O₇, is the only primary magmatic, terrestrial Sc-mineral. Regardless of its rarity, magmatic thortveitite is found in the blocky zone of some NYF granitic pegmatites (e.g. Evje-Iveland pegmatites in southern Norway; Schetelig, 1922; Bjørlykke, 1935; Williams-Jones and Vasyukova, 2018) as prismatic crystals up to several tens of cm long. Nevertheless, more commonly, thortveitite and other Sc minerals, typically Ca-bearing (kristiansenite, cascandite, scandiobabingtonite), are among the most recent minerals in miarolic cavities (e.g. Pezzotta et al., 2005), or are found as probable proximal replacement products of the Sc-bearing precursors (e.g. Raade et al., 2004; Novák and Filip, 2010; Výravský et al., 2017).

In this contribution, we report a unique assemblage of secondary Sc-minerals from the intraplutonic NYF pegmatite Kožichovice II, of the Třebíč Pluton (Moldanubicum) consisting of kristiansenite, Ca₂ScSn(Si₂O₇)(Si₂O₆OH), thortveitite, Sc₂Si₂O₇, and nioboheftetjernite, ScNbO₄, which originated by hydrothermal replacement of Sc-enriched columbite-(Mn). This investigation focuses on: (1) the composition of the Sc-bearing and associated minerals; (2) the conditions of their formation and mass balance; (3) importance of Ca-rich fluids for the formation of Sc-minerals in granitic pegmatites; and (4) the behaviour of the Sc from the magmatic crystallisation through metasomatic replacements to the hydrothermal stage.

The Třebíč Pluton (343–335 Ma, Janoušek et al., 2019) forms a large flat body composed of durbachites (K-Mg-rich melagranites to melasyenites), which intruded high-grade metamorphic rocks (sillimanite–biotite paragneiss with intercalations of amphibolites, marbles and quartzites) in the eastern part of the Moldanubian Zone, Bohemian Massif, Czech Republic (Fig. 1) (Leichmann et al., 2017; Janoušek et al., 2020). The origin of durbachite magma is interpreted to be a result of the melting of mantle domains previously metasomatised by fluids released from a subducted slab of mature continental crust and physically contaminated by that crustal material. The melt reflects simultaneous enrichment in typical mantle (Mg, Cr, Ni) and lithophile elements (K, Rb, Cs, Th, U). Noticeably, the durbachites of the Třebíč Pluton yielded quite high concentrations of Sc of 8–30 ppm (Janoušek and Holub, 2007; Janoušek et al., 2020).

The intraplutonic pegmatites that are related genetically to the Třebíč Pluton have variable degrees of textural and geochemical evolution. They form minor primitive segregations as well as more common dykes, up to ∼2 m in thickness and several tens of metres in length, with transitional to sharp contacts with the host durbachite (Škoda and Novák, 2007, Zachař et al., 2020). Zoned dykes consist of a granitic graphic and blocky unit, albite unit typically located between blocky K-feldspar and a quartz core, and locally small miarolitic cavities (Zachař and Novák, 2013; Zachař et al., 2020). The pegmatites belong to the NYF petrogenetic family although they are F-poor (Škoda and Novák, 2007; Čopjaková et al., 2013). They are also an example of Group 2 pegmatites generated from residual melts of granite magmatism (RMG) in the proposed nomenclature of granitic pegmatites (Wise et al., 2022). On the basis of their mineralogy, two types of pegmatites were distinguished: (1) Simply zoned primitive allanite-type pegmatites (Kfs+Qz+Pl+Bt±Amp) containing allanite, ilmenite, zircon, titanite ± tourmaline; and (2) more evolved euxenite-type pegmatites (Kfs+Qz+Pl+Ab+Bt±Tur), with euxenite-(Y) and aeschynite-group minerals ± Be-minerals (primary: beryl, phenakite, helvine–danalite; secondary: bavenite–bohseite, milarite–agakhanovite-(Y), hingganite-(Y), bazzite), Sn-minerals (primary: cassiterite, herzenbergite; secondary: stokesite) and common greenish K-feldspar (amazonite variety). Mineral symbols are abbreviated according to Warr (2021).

The euxenite-type pegmatite Kožichovice II crops-out in the fields ∼ 1.5 km NE of the Kožichovice village, in the southeastern vicinity of Třebíč town. It was found in the 1980s, and currently the material is available as boulders and small fragments in the field and in the collection of the Moravian Museum, Brno. The dyke is ∼0.5 m in thickness and several tens of metres in length and consists of granitic, graphic, blocky zones and a quartz core. The albite unit is developed between the blocky K-feldspar and the quartz core (Novák and Filip, 2010). The investigated sample is an ∼10 cm large fragment of massive quartz and pale green blocky amazonite. The quartz hosts a euhedral crystal of fresh phenakite, ∼ 20 × 5 mm in size, and a dark brown, dull to metallic heterogeneous aggregate, ∼10 mm in size, which contains the Sc mineralisation pertinent to this investigation.

Two polished epoxy discs (RS88 and RS169 samples) containing fragments of the Sc-rich aggregate were prepared for electron-probe microanalyses, TIMA and Raman investigation.

The compositions of minerals were determined by means of a CAMECA SX100 electron-probe microanalyser (EPMA) at the Department of Geological Sciences, Masaryk University, Brno. The wavelength-dispersive mode, an accelerating voltage of 15 kV, a beam current of 20 nA, and an electron beam diameter of 1–5 µm were used. The following characteristic X-ray lines and standards were used for oxide minerals quantiitative analysis: Kα lines – F (topaz); Na (albite); Si (sanidine); Sc (ScVO₄); Mg (MgAl₂O₄); Ti (anatase); Mn (Mn₂SiO₄); Fe (almandine); Zn, Al (gahnite); Kβ lines – Ca (wollastonite); Lα lines – Y (YAG); Nb (columbite Ivigtut); W (metallic W); Sn (SnO₂); Lβ lines – Zr (zircon); Sb (metallic Sb); Mα lines – Ta (CrTa₂O₆); Hf (metallic Hf); Th (CaTh(PO₄)₂); Mβ lines – Pb (vanadinite); U (metallic U); Bi (metallic Bi). Silicates were analysed using the same standards as above plus Lα lines – La (LaPO₄), Ce (CePO₄); Sm (SmPO₄); Lβ lines – Pr (PrPO₄), Nd (NdPO₄). Raw X-ray intensities were corrected for matrix effects with a ϕρ(z) algorithm of the X-PHI routine (Merlet, 1994). On the basis of the counting statistics, the measurement error expressed as 2σ is approximately <1 rel.% for concentrations of ∼20 wt.% and ∼8 rel.% for concentrations of ∼1 wt.%. Due to an interference of the ZrLα line with the 2^nd-order ScKα line, the Zr was measured on the ZrLβ line using pentaerythritol monochromator.

Most of the investigated phases were identified and classified with confidence on the basis of the EPMA data. However, as the mineral with composition CaNb₂O₆ can crystallise with a columbite (fersmite) or aeschynite (vigezzite) structure, and the natural ScNbO₄ phase was described just recently as nioboheftetjernite with a wolframite structure, these two phases were investigated using Raman spectroscopy in order to determine their structural properties. The spectra were obtained from uncoated polished sections using an Horiba Labram HR Evolution spectrometer equipped with an Olympus BX 41 optical microscope, a diffraction grating with 600 gr/mm, and a Peltier-cooled, Si-based charge-coupled device (CCD) detector, at the Department of Geological Sciences, Masaryk University, Brno. The Raman signal was excited by a 532 nm laser and collected in the range of 50–4000 cm^–1 using a 100× objective. Experimental conditions and band fitting procedure were similar to those described in Škoda et al. (2018).

The Raman spectroscopy unequivocally established that the CaNb₂O₆ phase corresponds to fersmite and ScNbO₄ to nioboheftetjernite. The observed vibration bands together with published data are shown in Supplementary materials – Table ST1 and Figs S1 and S2.

The panoramic phase- and element-distribution maps (Figs S3 and S4 in supplementary material) together with numerical values for volume ratios of individual minerals were obtained using a TIMA (Tescan Integrated Mineral Analyzer) mounted on a Tescan MIRA FEG electron microscope at TESCAN GROUP, Brno, using the following analytical conditions: working distance of 15 mm, accelerating voltage of 25 kV and probe current of 10 nA. The data were acquired using liberation analysis, high-resolution mapping mode (see Hrstka et al., 2018 for more details). The pixel spacing for both back-scattered electron (BSE) and energy dispersive spectroscopy (EDS) data was 1 µm/pixel. The computation of the mineral volumetric proportions was based on spatial ratios determined by TIMA; the EPMA data for individual phases were used as representative compositions and mineral densities were calculated for these compositions using published unit-cell volumes of the closest end-members (Table 1). Although unit cell volume varies with composition, in our case the changes in density with respect to the end-members are predominantly caused by changes in Nb/Ta ratio. Coincidently, Nb and Ta have virtually the same ionic radius (Shannon, 1976), therefore, their substitution should not change the unit cell volume. We believe that potential errors in the density calculations are small enough to have no significant influence on the interpretations discussed below.

A detailed study of the Sc-rich assemblage in BSE images and TIMA phase- and element-distribution maps revealed complex mineral assemblages and textural relations (Table 2).

Zircon and columbite-(Mn) are evidently the early minerals in the assemblage and formed during a late magmatic stage of the pegmatite-forming process. Zircon forms euhedral crystals and crystal aggregates, up to 500 µm in diameter, typically enclosed in columbite-(Mn) or later fersmite, locally however it also crystallised in cavities in columbite-(Mn) and was subsequently overgrown by younger thortveitite and kristiansenite (Fig. 2a). Columbite-(Mn) forms aggregates of crystals reaching ∼2.5 mm in size. The crystals exhibit complex patchy to irregular zoning in BSE, with lighter areas being Ta (and Sn, Ti, Sc) enriched. The edge of the columbite-(Mn) grains commonly have embayment features and grains are evidently intensively corroded by later fersmite (Figs S3, S4).

Fersmite extensively replaces columbite-(Mn) as irregular aggregates or veinlets and locally exhibits a weak zonation in BSE-images (Fig. 2b, c). The euhedral, prismatic crystals of nioboheftetjernite up to ∼100 µm in length, typically grew on surfaces of columbite-(Mn) corroded by fersmite (Fig. 2b, c). Nioboheftetjernite crystals are surrounded by fersmite or grew into vugs within fersmite (Fig. 2b–e), or subsequently were overgrown by younger minerals, typically titanite (Figs S3, S4).

Irregular aggregates of fluorcalciomicrolite, up to ∼1 mm in size, appear to be younger than associated fersmite (Figs 2, S3, S4,). They are typically strongly hydrothermally altered with complex zoning in BSE images, locally with late Pb-enriched replacements and overgrowths (Fig. 2d). Very rare anhedral thortveitite fills small cavities, up to ∼50 µm, commonly associated spatially with euhedral zircon (Fig. 2a). Kristiansenite and titanite also occur as cavity- and fracture-filling minerals, up to several hundred µm in size. Kristiansenite is typically homogeneous in BSE-images (Fig. 2e), whereas titanite commonly displays fine oscillatory zonation, reflecting the variable Ta content (Fig. 2f). Textural relationships indicate the following crystallisation sequence of primary to secondary minerals: zircon + columbite-(Mn) → fersmite + nioboheftetjernite → kristiansenite + thortveitite + titanite + fluorcalciomicrolite.

The zircon is enriched in Hf (0.06–0.10 apfu; 6.4–10.9 wt.% of HfO₂, Table 3). Fresh zircon contains an average 0.01 apfu of Sc (0.43 wt.% Sc₂O₃), and 0.02 apfu of Y+HREE (Dy, Er, Yb). The infrequent altered domains have elevated actinide contents, lower analytical totals, uptake of non-formula elements (Ca, Al, Fe, F) as well as elevated contents of Sc and P, up to ∼2 wt.% of Sc₂O₃ and ∼7 wt.% of P₂O₅.

Columbite-(Mn) contains 1.42–1.62 apfu of Nb, 0.26–0.39 apfu of Ta, 0.52–0.57 apfu of Mn. and 0.32–0.38 apfu of Fe, resulting in a Ta/(Ta + Nb) ratio of 0.14–0.22 and a Mn/(Mn + Fe) ratio of 0.58–0.64 (Fig. 3). Titanium (avg. 0.14 apfu, 3.06 wt.% TiO₂) and Sc (avg. 0.08 apfu, 1.6 wt.% of Sc₂O₃) are the most important minor elements. Tin, Y and W were detected, though with concentrations ≤ 0.01 apfu (Table 4).

Nioboheftetjernite corresponds to a simplified empirical formula (N = 61 analyses) (Sc_0.58Fe_0.15Sn_0.10Mn_0.08Zr_0.02)_Σ0.93(Nb_0.67Ta_0.33W_0.01Ti_0.04)_Σ1.05O₄. The ratio Ta/(Ta + Nb) = 0.26–0.39 and Mn/(Fe+Mn) = 0.31–0.46 (Fig. 3), whereas Sc contents vary between 0.50 and 0.64 apfu (14.08–17.98 wt.% of Sc₂O₃), Table 4. Kolitsch et al. (2010) assigned Ti into the octahedral site in chemically related heftetjernite, ScTaO₄. Adopting the same approach, the Sc- and Nb-site stoichiometry of the phase would be close to 1:1 ratio. The characteristic Raman bands are similar to the spectra published by Lykova et al. (2021) for recently described nioboheftetjernite from the type locality as well as the synthetic ScNbO₄ phase of wolframite structure (Ouahrani et al., 2022), see Table ST1 and Fig. S1.

Fersmite is more enriched in Nb than columbite-(Mn) with 1.58–1.83 apfu Nb, 0.17–0.37 apfu Ta and a Ta/(Ta + Nb) ratio of 0.09–0.19 (Fig. 3). Apart from major Ca (0.83–0.94 apfu) fersmite contains 0.06 apfu of Ti, Σ Mn + Fe = 0.02 apfu (Mn ˃ Fe), ∼0.01 apfu of Y, W and Sn, and only 0.007 apfu of Sc (0.16 wt.% of Sc₂O₃, Table 4). The main Raman bands are consistent with those published by Husson et al. (1977) and Moreira et al. (2010) for Pbcn CaNb₂O₆, a synthetic analogue of fersmite (Table ST1, Fig. S2)

Fluorcalciomicrolite (to rare fluorcalciopyrochlore) has variable Ta (0.77–1.13 apfu), Nb (0.57–0.94 apfu) and Ti (0.20–0.46 apfu) content, resulting in Ta/(Ta + Nb) = 0.45–0.65 (Fig. 3). In addition, it contains 1.14 Ca, 0.06 Na, 0.13 Sb, 0.14 Sn, 0.04 Mn, 0.36 F, and ≤ 0.01 Y, U, Th, Si, Fe and W (all avg. in apfu). The calculated vacancy in the Ca position ranges between 0.20–0.90 apfu, avg. 0.42 apfu. Scandium is systematically low ∼ 0.03 wt.% of Sc₂O₃, close to the detection limit. The mineral is classified as fluorcalciomicrolite as F slightly prevails over OH and O in that particular structural site. However, the exact identification requires additional study, as the amount of OH was not determined directly but derived from the electroneutrality. Fluorcalciomicrolite is locally penetrated by late Pb-enriched domains (≤0.23 apfu of Pb). Compositions are shown in Table 4.

The simplified empirical formula of kristiansenite (N = 30) corresponds to: Ca_2.04[Sn_0.94Sc_0.73Fe_0.15(Nb,Ta)_0.08Zr_0.03Mn_0.02Ti_0.02]_Σ1.97(Si_3.95Al_0.05)_Σ4.00O₁₃OH. The amount of Sc varies between 0.56–0.98 apfu (6.62–11.85 wt.% of Sc₂O₃, Table 3).

Thortveitite composition (Table 3) ranges from an almost pure end-member with 1.95 apfu Sc (50.56 wt.% of Sc₂O₃) to a composition with 1.57 apfu of Sc (38.94 wt.% of Sc₂O₃). Remarkably, the measured Sn content of ≤0.12 apfu (6.72 wt.% of SnO₂) is the highest reported so far.

Titanite is significantly enriched in Ta and Nb, with ≤0.2 apfu of Ta + Nb and Ta/(Ta + Nb) = 0.5–0.85, (Fig. 3, Table 3) and has variable enrichment of Sn 0.01–0.11 apfu and Sc ≤0.03 apfu (0.04–1.93 wt.% of Sc₂O₃).

The high-resolution phase- and element-distribution maps of the whole sample are very useful because they reveal textural features and mineral relationships otherwise unrecognisable by conventional scanning electron microscopy (EDS mode) investigation. Variability in the composition of columbite and fersmite causes their mean Z (atomic number) to overlap, making it impossible to distinguish them by BSE, hence a combination of BSE with EDS mapping is essential. In addition, changing the view field from mm to cm scale revealed that the metasomatic alteration propagated inwards and that columbite is mostly replaced along cracks (Figs S3, S4).

Fersmite is the volumetrically dominant mineral (∼51 vol.% of the aggregate), followed by relics of columbite (∼29 vol.%), and fluorcalciomicrolite (∼10 vol.%), respectively. The most abundant Sc-dominant mineral is kristiansenite comprising 3.3 vol.%, followed by nioboheftetjernite (1.4 vol.%) and thortveitite, which accounts for only 0.08 vol.%. The amount of titanite corresponds to 4.6 vol.% and the remaining 0.7 vol.% is represented by zircon. Unfortunately, the volume of voids cannot be properly measured, because they are commonly above the BSE background threshold. The significant loss of Fe and Mn and substantial gain of Ca, Si and Sn in secondary mineral assemblages are evident. The calculated (see Analysis section on TIMA) bulk composition of the secondary mineral assemblage (Table 5) shows that it contains only 0.57 wt.% Sc compared to 1.05 wt.% in primary columbite, and its bulk content of 1.77 wt.% Sn is approximately ten times higher than in the primary precursor.

The very high compatibility of Sc resulting from it having an ionic radius similar to Fe²⁺, Fe³⁺, Mg, Zr, Sn, Ti, and Al (Shannon, 1976) enables easy incorporation into numerous major to accessory minerals. Two principal substitution mechanisms, homovalent and heterovalent, have been recognised (Raade et al., 2002). Homovalent substitutions include Sc = Al, Fe³⁺ (garnet, beryl, amphibole, pyroxene: Frondel et al., 1968; Quartieri et al., 2006; Neumann, 1961; Eby, 1973; Chassé et al., 2018; Novák and Filip, 2010; Přikryl et al., 2014) and rather exceptionally Sc = Y, HREE (keiviite: Langhoff, 1996; xenotime: Bernhard et al., 1998). The heterovalent substitutions typically involve R³⁺ + R⁴⁺ = R²⁺ + R⁵⁺ (Novák and Černý, 1998; Wise et al., 1998), 2R³⁺ = R²⁺ + R⁴⁺ (e.g. helvine: Raade et al., 2002; ilmenite: Kalashnikov et al., 2016), R³⁺ + R⁵⁺ = R⁴⁺ + R⁴⁺ (e.g. zircon: Moëlo et al., 2002; Výravský et al., 2017; baddeleyite: Kalashnikov et al., 2016) and R³⁺ + R⁺ = R²⁺ + R²⁺ (e.g. jervisite: Mellini et al., 1982; Vignola et al., 2019) where R²⁺= Fe, Mn, Mg, Ca; R³⁺ = Sc, Y, Ln; R⁴⁺ = Ti, Sn, Zr, Si; and R⁵⁺= Nb, Ta, P.

Columbite-(Mn) from the Kožichovice II pegmatite has a very strong negative correlation between Sc and Nb, and a strong positive correlation between Sc and Ta (R = –0.84 and +0.82, respectively; Fig. 4). A positive correlation of Sc with Ti (R = +0.92) indicates the euxenite-type substitution R³⁺ + R⁴⁺ = R²⁺ + R⁵⁺. However, the amount of Ti significantly exceeds the content of Sc(+Y)³⁺ and due to very low amounts of calculated Fe³⁺, the remaining Ti is possibly incorporated via the rutile-type substitution 3Ti⁴⁺ = Fe²⁺ + 2Nb⁵⁺.

Very uniform ΣNb + Ta (0.99–1.02 apfu, avg. 1.00 apfu) in the nioboheftetjernite and low correlation coefficients with other elements (strongest –0.37 for Ti) indicate, that there is only very limited substitution in the Nb site. In contrast all of the remaining elements (except Zr) correlate negatively with Sc. Noticeably, Sc (and Zr) are enriched in the Ta-rich compositions (Fig. 4), whereas all the other elements (Fe, Mn, Sn, Ti) trend to Nb-rich compositions, indicating solid solution between Sc₂Ta₂O₈ and M²⁺M⁴⁺Nb₂O₈ and the substitution 2M³⁺ = M²⁺ + M⁴⁺. However, regardless of the low calculated Fe³⁺ (avg. 0.04 apfu), the amount of M²⁺ surpasses that of M⁴⁺, leaving a slight deficit in charge in the Sc site, which is roughly equivalent to the surplus of charge caused by presence of W in the Nb site. This is in contrast with the observations of Kolitsch et al. (2010) in heftetjernite, where the amount of M²⁺ and M⁴⁺ matched almost perfectly, and indicates a possible substitution mechanism involving R²⁺ and W⁶⁺.

The composition of kristiansenite is comparable to the type kristiansenite from Heftetjern pegmatite (Raade et al., 2002, 2004) except for elevated Nb and Ta (avg. 0.04 apfu each). The Nb + Ta correlates positively (R = 0.8) with (Fe,Mn)²⁺, indicating possible substitution Sc³⁺ + Sn⁴⁺ = Fe²⁺ + (Nb, Ta)⁵⁺. As the slope of regression line is ≈0.5, other substitution mechanisms might take place simultaneously. The most important substitution in the samples of thortveitite is 2Sc³⁺ = (Fe, Ca, Mn, Mg)²⁺ + (Sn, Ti, Zr)⁴⁺. The incorporation of Ta and Nb in titanite is facilitated mainly by a 2Ti⁴⁺ = (Al, Sc, Fe)³⁺ + (Nb, Ta)⁵⁺ and subordinately also by Na⁺ + (Ta, Nb)⁵⁺ = Ca²⁺ + Ti⁴⁺ substitution, respectively (see e.g. Lussier et al., 2009).

Scandium as a compatible element is incorporated in mafic silicates (amphibole, pyroxene, garnet, biotite) or zircon during the solidification of granitic magmas (Glassley and Piper, 1978; Williams-Jones and Vasyukova, 2018). The Třebíč Pluton consisting of amphibole- and biotite-rich melasyenite to quartz syenite (durbachite) contains 8–30 ppm Sc (Holub, 1997; Janoušek et al., 2020). The residual magma that was more felsic and volatile-rich solidified locally as pegmatite bodies with a NYF signature; although these are typically F-poor. The NYF pegmatites of the Třebíč Pluton commonly host biotite, tourmaline (schorl–dravite–dutrowite solid solution), beryl and scarce actinolite, all these minerals can potentially incorporate Sc. Rare actinolite is present only in the outermost granitic unit, whereas common biotite occurs in the less-evolved textural-paragenetic units (granitic, graphic). Tourmaline is a rare accessory-to-minor mineral in some NYF pegmatites of the Třebíč Pluton. It occurs in the graphic, blocky and aplitic unit and contains 30–500 ppm Sc (Novák et al., 2011; Čopjaková et al., 2013, 2015). This is, together with biotite, the most common mineral with elevated concentrations of Sc.

At the Kožichovice II pegmatite, crystallisation of volumetrically dominant feldspars and quartz, together with the absence of primary tourmaline, and relative scarcity of biotite, resulted in enrichment of Sc in a residual pegmatitic melt. Consequently, Sc entered the structure of beryl and columbite-(Mn) which crystallised at the late magmatic stage. Very rare tourmaline (schorl–dravite) is present only as a late fracture-filling morphological type of subsolidus origin (Novák et al., 2011). Thus its crystallisation did not influence the concentration of Sc during the magmatic stage. Scandium-enriched beryl (0.05–0.68 wt.% Sc₂O₃) is rare (Novák and Filip, 2010) and not related spatially to the Sc-enriched columbite-(Mn) (avg. 1.6 wt.% Sc₂O₃). Regardless of a complete solid solution between ZrSiO₄ and ScPO₄, even at rather low temperatures (Möelo et al., 2002; Výravský et al., 2017), the very low content of P in the pegmatites of the Třebíč Pluton (Škoda et al., 2006, Zachař 2021) severely limited this substitution in zircon (avg. 0.43 wt.% Sc₂O₃) associated with columbite-(Mn). Further, zircon is volumetrically subordinate in the mineral assemblage (0.7 vol.%).

The columbite-(Mn) crystal aggregates were corroded by Ca-enriched residual fluids and replaced in situ by fersmite mainly along fractures propagating into the central parts of the grains (Fig. S3, S4). Almost no (Fe + Mn) and only ∼10 wt.% of Sc (related to the primary columbite) were accommodated in fersmite (avg. 0.16 wt.% Sc₂O₃), which has also significantly lower content of Ti and a slightly lower Ta/(Ta+Nb) ratio (Fig. 3). Nioboheftetjernite commonly growing on the corroded surfaces of columbite-(Mn) is usually enclosed in fersmite and both newly-formed minerals appear to be in textural equilibrium (Fig. 2). Very low Sc in newly formed fersmite as a consequence of crystal structural constraints (^VISc 0.745 Å and ^VICa 1.000 Å; Shannon, 1976) resulted in a local supersaturation of Sc and allowed crystallisation of nioboheftetjernite. However, fersmite and nioboheftetjernite contain a higher amount of Sn (0.49 and 6.12 wt.% SnO₂, respectively) relative to primary columbite-(Mn) (0.25 wt.% SnO₂). This results in Sn-enrichment of the residual fluids together with Ca.

In the late hydrothermal stage, the compositions of the crystallising phases still indicate a very high activity of Ca, Sc and Sn, accompanied by the introduction of Si and enrichment in Ta, demonstrated by high Ta/(Ta + Nb) ratios for kristiansenite, microlite and titanite: avg. 0.50, 0.56 and 0.73, respectively (Fig. 3). Ilmenite, intensively replaced by Sn-enriched titanite from several NYF pegmatite bodies (Škoda et al., 2006), the occurrence of kristiansenite after Sc-enriched columbite-(Mn) at the pegmatite Číměř I (Zachař, 2021), and common bavenite–bohseite replacement of primary Be minerals (beryl, phenakite, helvine–danalite; Novák and Filip, 2010; Zachař et al., 2020; Novák et al., 2023), demonstrates that Ca- and Sn-enriched residual fluids are a common feature of the NYF pegmatites of the Třebíč Pluton.

The last recognisable processes are formation of Pb-enriched veinlets in the fluorcalciomicrolite aggregates and alteration of earlier fluorcalciomicrolite, as is known from other pegmatites (see e.g. Chládek et al., 2021). Additionally, some of the zircon grains are locally hydrothermally altered (see Results); however, the timing of this alteration is remains unclear.

Despite having detailed information (TIMA and EMPA data) regarding the composition and abundance of all minerals involved in the evolution of these pegmatites, definitive conclusions about the behaviour of particular elements cannot be drawn as the initial amount of columbite-(Mn) and its porosity are not known. Three possible scenarios are discussed below.

Scenario 1: The replacement process was isovolumetric, all minerals from the metasomatic and late hydrothermal stage directly replaced columbite. In this scenario, more than 50% of the Sc, originally present in columbite-(Mn), was leached out from this mineral. The Nb, Ta and Ti do not show significant influx or leaching. Our data show that a significant majority of Ca and Sn was derived from fluids and almost all Mn and Fe were leached out of the system (see Table 6). This scenario is rather unlikely because most of the minerals from the late hydrothermal stage (microlite, titanite, kristiansenite, thortveitite) crystallised in open spaces or filled fractures and do not replace earlier-formed minerals (Figs 2, S3, S4). Moreover, the presence of porosity is crucial for the formation of secondary mineral assemblages by fluid driven dissolution–reprecipitation processes (Putnis and Putnis, 2007).

Scenario 2: The volume of the replaced columbite-(Mn) corresponds to the volume of the minerals from the metasomatic stage (fersmite and nioboheftetjernite). Only 74 vol.% of the whole secondary mineral assemblage directly replaced the columbite-(Mn). This scenario is consistent with our textural observations (see above, Fig. S3, S4). However, ∼40 proportional wt.% (hereinafter prop. wt.%) of Sc and most of Fe and Mn from the primary columbite-(Mn) was leached out (Table 6). A significant influx of Sn is necessary, and some Nb, Ta and Ti is required to be added to the system (Table 6). The leaching of Sc is supported by the occurrence of kristiansenite in the cracks commonly situated near the rim or in a vicinity of the aggregate (Fig. S3, S4). Similar occurrences of kristiansenite on cracks near or distal from a Sc-mineral precursor have been found at other pegmatites of the Třebíč Pluton, e.g. Číměř (Zachař, 2021). Apparently, distal kristiansenite originated from fluids at the conditions where Sc is more mobile relative to the nioboheftetjernite typically found in a proximity or directly on the surface of corroded columbite-(Mn). The mobility of Sc in fluids is also supported by the occurrence of kristiansenite as a fracture or miarolitic cavity filling mineral at other occurrences: Baveno, Italy (Pezzotta et al., 2005), Heftetjern, Norway (Raade et al., 2002, 2004), Věžná I, Czech Rep. (Toman and Novák, 2020), Szklarska Poręba, Poland (Evans et al., 2018) and Cadalso de los Vidrios, Spain (Prado-Herrero et al., 2009; Correcher and Garcia-Guinea, 2020).

Scenario 3: Sc was not leached out of the system. This scenario is unlikely because it requires a very high initial porosity of the original columbite-(Mn) aggregate or the presence of a significant amount of a now completely dissolved Sc-free mineral. The results of the mass balance calculations demonstrate that the quantity of dissolved columbite-(Mn) required for the formation of the current secondary mineral assemblage would occupy only 47 vol.% of the total amount of the secondary mineral assemblage (Table 6). Furthermore, there is evidence of high Sc mobility in the textural position of kristiansenite (see above), and high mobility of Sc has been observed at other localities (see the review below of Sc-bearing minerals in granitic pegmatites).

Scenario 2 illustrates most closely the real behaviour of elements during the alteration process based mainly on our textural evidence and mass-balance calculations. Consequently, the following conclusions can be drawn: (1) All Sc necessary for the formation of the secondary Sc-bearing minerals has been supplied by decomposition of Sc-enriched columbite-(Mn) coupled with leaching out of a significant portion of the Sc. The alteration of Sc-bearing zircon could yield only a negligible amount of Sc. (2) Approximately 3–7 prop. wt.% of Mn and Fe from the parental columbite-(Mn) remained in the secondary mineral assemblage with the remainder being leached out. (3) Only ∼5–11 prop. wt.% Sn contained in the secondary mineral assemblage could have been supplied by the decomposing columbite-(Mn). (4) Because the Si, Ca and Sb contents are below the detection limit of EPMA in columbite-(Mn), they must have been supplied by the incoming fluids together with most of the Sn. (5) Metasomatic alteration of the Sc-enriched precursors could be a source of Sc for the distal secondary hydrothermal Sc-rich minerals in granitic pegmatites including those developed in miarolitic pockets.

Scandium-bearing minerals from granitic pegmatites have been distinguished on several compositional and paragenetic criteria. We use the term Sc-mineral for the minerals with nominal Sc (e.g. thortveitite, bazzite) and the term Sc-enriched mineral for the minerals where Sc enters as a minor to trace cation at distinct positions in the crystal structure (Table 7). On the basis of their position in pegmatite evolution the following paragenetic types of Sc-bearing minerals have been defined: (1) primary magmatic in massive rock; (2) primary hydrothermal in miarolitic pockets; and (3) secondary hydrothermal defined as proximal (3a) in pseudomorphs after Sc-bearing precursor and distal (3b) developed on tectonic fractures close to, or far from from the primary precursor. A similar paragenetic classification was previously used for cookeite (Novák et al., 2015), secondary Be-minerals (Zachař et al., 2020; Novák et al., 2023) and Nb–Ta oxide minerals (Chládek et al., 2021) from other occurrences.

The source of Sc for the formation of hydrothermal Sc-bearing minerals might be fluids exsolved from residual magma or altered primary Sc-minerals or Sc-enriched minerals. Examples of Sc-bearing minerals and their relationships are given in Table 7. The only known terrestrial primary magmatic Sc-mineral is thortveitite. However, Sc is present in substantial amounts in several primary minerals: keiviite (isostructural with thortveitite), zircon, beryl, Nb-rich rutile, ixiolite, columbite-group minerals, wolframite, garnet, biotite, epidote and tourmaline (see Table 7 for details and references). Note that many of these are prone to alteration by hydrothermal fluids.

At the Kožichovice II pegmatite hydrothermal alteration of two distinct Sc-bearing precursors, beryl and columbite-(Mn), yielded two distinct assemblages. Alteration of beryl by alkaline Ca-rich fluids at T < 250–350°C produced an Al-enriched assemblage (secondary beryl, bavenite–bohseite, smectite) + bazzite (Novák and Filip, 2010). The elevated activity of Ca is evident from the common presence of bavenite–bohseite (Table 7). The assemblage is also Ca-rich, though compositionally the very distinct oxide precursor columbite-(Mn) facilitated the formation of a different, dominantly oxide, secondary mineral assemblage. The textural relationships between columbite-(Mn), nioboheftetjernite and fersmite (Fig. 2, Supplementary Figs S3, S4) unambiguously testify for the metasomatic origin of nioboheftetjernite related to the replacement of Sc-enriched columbite-(Mn) by fersmite.

The most significant pegmatite-hosted occurrences of Sc mineralisation are the Heftetjern pegmatite in the Tørdal area of southern Norway (Bergstøl and Juve, 1988; Raade et al., 2002, 2004; Cooper et al., 2006; Kolitsch et al., 2010 and references therein), and intraplutonic miarolitic pegmatites in the Baveno area, northern Italy (Gramaccioli et al., 2000; Pezzotta et al., 2005; Guastoni et al., 2012 and references therein). These occurrences have remarkable similarities with the Kožichovice II pegmatite especially with respect to the high activity of Ca in fluids and paragenetic position of Sc-minerals.

In the Heftetjern pegmatite, three paragenetic types of the Sc mineralisation were recognised. (1) Some textural varieties of thortveitite and bazzite in massive rock are probably primary magmatic (Juve and Bergstøl, 1990; Raade et al.,2004; Kristiansen, 2009). (2) Late Sc-minerals from small miarolitic (?) cavities in albite (kristiansenite, bazzite, scandiobabingtonite, cascandite, heftetjernite, oftedalite – Raade et al., 2002; Kolitsch et al.,2010; Raade and Erambert, 1999; Cooper et al., 2006), show no direct evidence of an altered Sc-bearing precursor. (3) Secondary Sc-minerals (kristiansenite, bazzite, oftedalite) formed by a direct metasomatic replacement of a Sc-rich precursor (thortveitite, Raade et al.,2004). The source of Sc necessary for crystallisation of Sc-minerals from the small pockets (2) is not clear. Residual postmagmatic Sc-enriched fluids are possible; however, there is significant amount of intensively chloritised Sc-enriched biotite (≤ 2800 ppm Sc; Rosing-Schow et al., 2018), whereas chlorite is Sc-poor (EPMA; < 300 ppm, below detection limit of EPMA; unpublished data of the authors). Because we observed mobility of Sc on a cm scale from the altered precursors (Výravský et al., 2017, discussion above), we assume, that the Sc necessary for the formation of late hydrothermal stage Sc-minerals in small cavities in albite or along the microfractures in the central part of the pegmatite body was derived from altered biotite or perhaps another altered primary Sc-enriched mineral (see list of Sc-enriched minerals and description of textural types of Sc-minerals in Kristiansen, 2009). There is common almandine–spessartine in the Heftetjern pegmatite containing ≤2000 ppm of Sc (Steffenssen et al., 2020); however, its role as a source of Sc for late-stage mineralisation is unlikely due to the absence of alteration.

In the Baveno pegmatites, the only known primary magmatic Sc-bearing mineral is beryl with a concentration of Sc ≤ 0.34 wt.% Sc₂O₃ (Pezzotta et al., 2005). Scandium minerals (cascandite, kristiansenite, scandiobabingtonite, bazzite, jervisite, thortveitite) associated with fluorite and Ca-zeolites occur as hydrothermal minerals in miarolitic pockets. Pezzotta et al. (2005) ascribed the source of Sc to subsolidus alteration of magmatic siderophyllite–‘zinnwaldite’ and gadolinite-(Y).

Černý et al. (2000) described an exsolved phase (Fe³⁺,Sc)NbO₄, possible rossovskyite, as tiny lamellas in Nb-rich rutile from the NYF pegmatite Håverstad, Evje-Iveland pegmatite district, Norway. It was interpreted as a process of cooling-related exsolution and the early exsolution was followed by coarsening and extended exsolution to a product with Sc > Fe³⁺, possible nioboheftetjernite, developed along microfractures presumably assisted by a fluid phase.

In the Věžná I pegmatite, zoned hübnerite to W-rich ixiolite (0.64 wt.% of Sc₂O₃) from the most evolved albite–pollucite unit (Toman and Novák, 2018, 2020) was strongly hydrothermally altered to a mixture of variable pyrochlore-group minerals, a Mn-hydroxide close to vernadite, and adularia. A few millimetres from this replacement assemblage, fracture filling kristiansenite associated with stokesite were found. Abundant Ca-rich pyrochlore minerals, stokesite and fluorapatite suggest elevated activity of Ca in the acting fluids.

Scandium-enriched wolframite to W-rich ixiolite and Sc-enriched zircon from the Dolní Bory pegmatite No. 3 underwent a hydrothermal replacement at P ≈200 MPa and T < 300–350°C and released Sc precipitated as proximal pretulite ScPO₄ within the pseudomorphs after the precursors (Novák et al., 2008; Výravský et al., 2017). Rare distal pretulite also was identified (Table 7). These secondary assemblages of Sc-rich minerals differ from all others by high activity of P and Al and particularly by total absence of Ca-rich minerals except for very rare scheelite found in pseudomorphs after W-rich ixiolite and not associated with pretulite (Novák et al., 2008). Recently, Výravský (2022) described the breakdown of Sc-enriched zircon from nearby pegmatite Dolní Bory No. 4, where in contrast with the pegmatite No. 3, the hydrothermal recrystallisation led to the formation of small amounts of secondary thortveitite instead of pretulite and leaching of most of the released Sc.

The incompatibility of Sc in volumetrically dominant newly formed secondary minerals is the main factor allowing separation of Sc and formation of Sc-rich minerals. This is caused by: (a) rather low temperature, which makes the crystal structures more rigid and less favourable to Sc substitution; and mainly by (b) favourable composition of the fluids. The Ca-rich fluids are particularly favourable for the formation of Sc-dominant minerals, because Sc does not preferentially enter the Ca-crystallographic site in substantial amounts due to its difference in ionic radius (100 pm vs. 75 pm for octahedrally coordinated Ca and Sc, respectively; Shannon, 1976). The hydrothermal Sc-minerals from granitic pegmatites are commonly Ca-rich (kristiansenite, cascandite, scandiobabingtonite; see Table 7), where Sc enters the octahedral site. In contrast, the crystallisation of a large volume of Sc-compatible Ca-minerals from hydrothermal fluids, e.g. titanite-group minerals, would prevent the precipitation of Sc-minerals. Thus, the overall chemical composition of the Ca-rich fluids as well as the composition of the Sc-enriched precursor are crucial for the formation of Sc-minerals.

A summary of similarities and differences between three important known occurrences of Sc-minerals in granitic pegmatites is given in Table 8. The pegmatites belong to the NYF compositional family (Černý and Ercit, 2005) and the intra-granitic pegmatites Kožichovice II and Baveno have an obvious magmatic source. In the exo-plutonic Heftetjern pegmatite, Bergstøl and Juve, (1988) speculated about Sc-enrichment originating from the surrounding amphibolites. The Sc-minerals are either late hydrothermal minerals in cavities or could be directly linked to metasomatic replacement of some older Sc-enriched precursor. In addition, the high activity of Ca in parental fluids manifested by the presence of Ca-rich minerals (Table 8) in close paragenetic association with Sc-minerals is typical at all localities.

The activity of F is very variable among the individual localities as is demonstrated by the mineral assemblages. Fluorite and/or topaz is common in the Baveno and Heftejern pegmatites but absent in Kožichovice II and all pegmatites from the Třebíč Pluton. The composition of fluorcalciomicrolite is very close to its not yet described OH analogue (0.36 apfu F and 0.35 apfu OH) and resulted from the low F activity. Consequently, a high activity of F is evidently not essential for the formation of Sc-dominant minerals in pegmatites as was proposed by Gramaccioli et al. (2000). The origin of Sc mineralisation at the relevant localities could be explained by metasomatic alteration of Sc-enriched precursors by Ca-rich fluids, where Sc was not compatible with the majority of coeval secondary minerals and therefore formed its own Sc-minerals.

Late Ca-enrichment is typical in the Kožichovice II pegmatite (Novák and Filip, 2010) and in many granitic pegmatites worldwide (Novák et al., 1999, 2013, 2023; Teertstra et al., 1999; Tindle et al., 2002, 2005; Martin and De Vito, 2014); however, the origin of Ca in these fluids is still not fully understood. Four principal sources of Ca in hydrothermal fluids are viable. (1) Pegmatite melt was primarily enriched in Ca from its parental granite or (2) Ca was derived by Ca-contamination in the early stages of pegmatite crystallisation (pre-emplacement and/or post-emplacement stage; Novák, 2013) as has been documented at pegmatites cutting metacarbonates, calc-silicate rocks and/or skarns (Novák et al., 1999, 2013; Buřival and Novák, 2018). The late, but primary, fluorapatite, liddicoatite, fluorite, calcite and zeolites are evidence that some of the Ca might have survived to later stages of primary crystallisation. Martin and De Vito (2014) suggested an additional source of Ca by (3) a late Ca ‘miniflood’ originating from albitisation of Ca-rich plagioclase in host rocks facilitated by the fluids which escaped from a pegmatite (see also Pieczka et al., 2019). (4) Serpentinite-derived fluids are an additional important source of Ca (Evans et al., 2013) as documented at primitive pegmatites with accessory tourmaline enclosed in serpentinite (Čopjaková et al., 2021).

At the intra-granitic pegmatites of the Třebíč Pluton including the Kožichovice II pegmatite, residual fluids are the most probable source of Ca. Evidence for this is in the common secondary bavenite–bohseite ± milarite after beryl, phenakite and helvine–danalite (Novák and Filip, 2010; Zachař et al., 2020) and common secondary titanite after ilmenite (Škoda et al., 2006; Zachař and Novák, 2013). The secondary Ca-rich minerals are typically proximal in pseudomorphs after a primary precursor (Novák et al., 2023). Rare primary actinolite in the outermost primitive unit and locally abundant Ca-enriched primary tourmaline (Novák et al., 2011) show that high Ca was present in the original pegmatite melt. Further, the total absence of distal secondary Be-minerals on late fractures (Zachař et al., 2020) and other fracture-filling Ca-bearing minerals, which are potential channel-ways for externally derived Ca-rich fluids, support the evidence that residual fluids are the source of Ca.

The unique assemblage of Sc-minerals in the Kožichovice II pegmatite has enabled an investigation of in situ processes of formation of Sc-minerals in granitic pegmatites and revealed the conditions favourable for their formation in magmatic-hydrothermal systems. The process includes three principal phases.

(1) Pre-concentration of Sc in the NYF pegmatitic melt typically derived from A- or I-type granites with elevated concentrations of Sc. This is because A-type granites are the products of melting of upper mantle and overlying lower crust containing 25.4 ppm Sc, compared to upper crust containing 7 ppm Sc (Wedepohl, 1995). The A- and I-type granites are enriched in Sc compared to upper crust-derived S-type granites. Importantly, S-type granites are usually rich in P, which facilitates incorporation of Sc into early crystallised zircon and thus impeding concentration of Sc in a residual melt. Consequently, in granitic systems Sc-rich minerals occur mainly in NYF pegmatites. The Třebíč Pluton has I-type affinity and evident enrichment in Sc (8–30 ppm; Holub, 1997, Janoušek et al., 2020) is reflected in the presence of Sc-rich minerals in its pegmatites.

(2) Incompatibility of Sc in the volumetrically dominant rock-forming minerals (feldspars and quartz; Hreus et al., 2021; Steffenssen et al., 2020) during magmatic crystallisation leads to its pre-concentration in minor and accessory minerals (biotite-group micas, tourmalines, epidote-group, zircon, columbite-group, rutile, cassiterite, garnets). At this stage, the precursor minerals contain hundreds of ppm to units of wt.% Sc. Only very exceptionally, Sc-dominant minerals, such as thortveitite, could form in the late-magmatic stage (e.g. Evje-Iveland pegmatites, Norway) after solidification of large quantities of feldspars and quartz in the central parts of pegmatite, which resulted in saturation of Sc in melt (Bjørlykke, 1935; Williams-Jones and Vasyukova, 2018). In the Kožichovice II pegmatite, Sc preferentially entered primary columbite-(Mn) and beryl (Novák and Filip, 2010).

(3) Subsolidus hydrothermal alteration of the precursor Sc-rich minerals by Ca-rich fluids, coupled with the incompatibility of Sc in volumetrically dominant secondary minerals, locally facilitated saturation of minor secondary Sc-minerals. In the Kožichovice II pegmatite, the metasomatic replacement of Sc-bearing columbite-(Mn) by fersmite led to the formation of nioboheftetjernite and later kristiansenite and thortveitite.

On the basis of detailed study of minerals assemblages of Sc-enriched primary and secondary Sc-rich minerals, it is reasonable to assume, that Sc-minerals at the localities Heftetjern, Baveno and Kožichovice II originated during similar primary solidus and subsolidus processes. However, F activity was very low in Kožichovice II, indicating the F complexation is not needed for the formation of secondary Sc-minerals in granitic pegmatites, whereas the presence of Ca-rich fluids is essential.

The supplementary material for this article can be found at https://doi.org/10.1180/mgm.2024.85.

The authors are grateful to reviewers Pietro Vignola, and two anonymous reviewers and also to handling editor Fabrizio Nestola for constructive comments that improved the manuscript. This research was supported by OP RDE [grant number CZ.02.1.01/0.0/0.0/16_026/0008459 (Geobarr) from the ERDF] for MN and RŠ. This paper is dedicated to our friend Alessandro Guastoni who intensively studied Sc-bearing minerals from pegmatites during his scientific career. The authors also thank Production Editor Helen Kerbey and Principal Editor Roger Mitchell for editorial assistance and copy editing.

The authors declare none.