Narsarsukite in peralkaline granites from the Papanduva Pluton, Graciosa Province, south Brazil: Insights from textural and compositional features

Abstract

We report textural and compositional data for the titanosilicate narsarsukite [Na₂(Ti,Fe³⁺)Si₄(O,F)₁₁] in peralkaline granites from the Papanduva Pluton in Graciosa Province, south-southeastern Brazil. Two distinct narsarsukite generations, one late magmatic and the other post-magmatic, were identified on the basis of textural and compositional features. The magmatic generation consists of larger, euhedral to subhedral variably zoned crystals and late poikilitic intergrowths between narsarsukite and albite laths, representing the crystallisation of the latest melt pockets. The post-magmatic generation forms smaller, typically fibrous crystals and irregular aggregates that occur interstitially or replace the primary mafic minerals, particularly arfvedsonite. Compositions of narsarsukite from the Papanduva Pluton cover most of the compositional range described in known occurrences. The magmatic generation is enriched in Zr and depleted in Al. The Fe³⁺ and Al contents show a positive correlation for the magmatic crystals, but a negative correlation for the post-magmatic narsarsukite. The Al/Fe³⁺ ratios are higher in the post-magmatic crystals and can be used to discriminate between the two generations. The compositional variations are controlled mainly by the heterovalent substitution (Ti, Zr)⁴⁺ + O^2– = (Al, Fe)³⁺ + F^1–, and are compatible with (OH)^1– ions in the O site. Narsarsukite has significantly higher concentrations of mid REE+Y and HFSE and very low to absent contents of LREE and LILE relative to the host-rocks. The average REE pattern of the magmatic crystals is highly fractionated, consistent with the expected strong preference for the smaller HREE. The post-magmatic crystals that replace arfvedsonite have flatter patterns, with higher concentrations of LREE and MREE. Their compositions suggest a fluid phase relatively rich in HFSE, REE, Y, Pb, Th and U. Narsarsukite is a common phase in strongly peralkaline granites such as the Papanduva Pluton and the well-known Strange Lake Complex, and should be considered a diagnostic mineral of highly peralkaline SiO₂-oversaturated rocks containing rare Ti and Zr mineral assemblages.

Introduction

Titano- and zirconosilicates are typical phases constituting the mafic mineral assemblages of highly peralkaline SiO₂-undersaturated and -oversaturated rocks and contribute to the chemical budget of the high-field-strength elements (HFSE), the rare earth elements (REE) and Y during the evolution of these systems in both the magmatic and post-magmatic crystallisation stages. Furthermore, mineral assemblages and compositions may give relevant clues to access the main intensive parameters of these crystallisation environments, as well as the alkalinity and volatile fugacities in the magma (e.g. Nichols and Carmichael, 1969; Marks et al., 2011; Andersen et al., 2010; Siachoque et al., 2022).

Among the known titanosilicates, narsarsukite is a relatively rare inosilicate with the IMA accepted formula Na₂(Ti,Fe³⁺)Si₄(O,F)₁₁, crystallising in the I4/m space group (e.g. Anthony et al., 2003). It was first described in a pegmatite from Narsarsuk, Greenland, by Flink (1901) and its crystal structure was determined by Pyatenko and Pudovkina (1960) and Peacor and Buerger (1962). Recently, Schingaro et al. (2017) revisited the narsarsukite structure and suggested the extended formula Na₄{Ti_1–x⁴⁺,M_x ³⁺)[O_1–x(OH,F)_x]}₂(Si₈O₂₀) with 0 < x ≤ 1. The structure is composed of 4-periodic double chains of [Si₄O₁₀]₂ tetrahedra connected into tubes and to chains of corner-sharing Ti(O₅,OH,F)-octahedra. The tubes are parallel to the c-axis, and the 7-coordinated Na⁺ cations are located in the cavities between the tubes and the octahedral chains. According to the structure hierarchy of the chain-, ribbon- and tube-silicates developed by Day and Hawthorne (2020), based on the connectedness of one-dimensional polymerisation of the (TO₄)^{n –} tetrahedra, the narsarsukite structure belongs to the ³T₈ group. This means that the connectivity of the [SiO₄]^4– tetrahedron has a value of 3 and the number of these tetrahedra is 8 in the repeat structural unit.

Narsarsukite occurs in several volcanic and plutonic, mainly SiO₂-oversaturated, peralkaline rocks, pegmatites and hornfels. It is associated commonly, in addition to quartz and alkali-feldspars, with aegirine, arfvedsonite, pectolite, various titanosilicates, such as aenigmatite, astrophyllite, neptunite and bafertisite, and zirconosilicates, such as elpidite (e.g. Birkett et al., 1996; Siachoque et al., 2022).

The main crystallochemical features of narsarsukite are well-established in the mineralogical literature (e.g. Wagner et al., 1991; Read, 1991; Birkett et al., 1996; Schingaro et al., 2017). Wagner et al. (1991) discussed infrared and Mössbauer spectroscopic results, whereas Schingaro et al. (2017) provided micro-Fourier transform infrared and X-ray photoelectron spectroscopy data. Eu³⁺-doped narsarsukite was synthesised by the sol-gel method (e.g. Balmer et al., 1997) and was shown to have luminescence properties (Rainho et al., 2003).

Notwithstanding the wealth of data available on narsarsukite, there is a paucity of systematic studies that combine textural and compositional features, particularly those comparing the characteristics of magmatic and post-magmatic crystallisation generations. These are relevant aspects to consider, as more than one textural generation of the mafic minerals (e.g. magmatic vs. hydrothermal) is common in peralkaline rocks (e.g. Siachoque et al., 2022). Additionally, although some electron microprobe data for Y and Ce have been made by Read (1991) and Birkett et al. (1996), REE data are still scarce and complete analyses exist only for peralkaline granites from the Strange Lake Complex (Vasyukova and Williams-Jones, 2019).

Narsarsukite is a common and widespread accessory mineral in the strongly peralkaline alkali-feldspar granites of the Papanduva Pluton, Morro Redondo Complex, in the Neoproterozoic A-type Graciosa Province, south-southeastern Brazil (Gualda and Vlach, 2007). This work presents and discusses textural and compositional data (major, minor and trace elements) for magmatic and post-magmatic narsarsukite generations. We also examine the behaviour of REE in this mineral and contribute to the mineralogical knowledge of the province and similar occurrences worldwide.

The Papanduva Pluton

The Papanduva Pluton, with ~60 km² and irregular surface contours (Fig. 1), is a peralkaline intrusion of the Morro Redondo Complex in the Neoproterozoic (ca. 580 Ma) Graciosa Province, south-southeastern Brazil (Gualda and Vlach, 2007; Vlach et al., 2011; Vilalva and Vlach, 2014; Vilalva et al., 2019). The pluton is composed of slightly greyish hypersolvus holo- to leucocratic alkali-feldspar granites showing significant variations in structure, texture and mineralogy. These variations have been grouped and mapped as massive, cataclastic, and foliated fine-, medium- and coarse-grained petrographic facies, accompanied by some microgranites. See Vilalva (2007), Vilalva and Vlach (2014) and Vilalva et al. (2016) for detailed petrographic descriptions and textural and compositional characterisation of the rock-forming felsic and mafic minerals. The granites are highly ferroan [FeO^T/(FeO^T+MgO) ≥ 0.96], and have SiO₂ contents of 74–78 wt.% and Na₂O+K₂O contents of 8.9–9.3 wt.%. The peralkaline index [(Na₂O+K₂O)/Al₂O₃ molar] is between 1.04 and 1.28, and the HFSE contents are relatively high, up to 2430 ppm (Vilalva and Vlach, 2014).

Figure 1. Simplified geological map of the Papanduva Pluton, Morro Redondo Complex, South Brazil. (1) Silicic and basic-intermediate volcanics associated with the intrusive rocks; granitic petrographic facies: (2) foliated; (3) cataclastic; (4) massive and (5) microgranites. (6) Metaluminous to slightly peraluminous syeno- and monzogranites from the Quiriri Pluton. (7) Metasediments from the Paranaguá Terrain; (8) gneisses and granulites from the Luis Alves microplate. Asterisks indicate locations of the samples analysed. Modified from Vilalva and Vlach (2014).

The most evolved and highly peralkaline foliated facies (Fig. 2) have sub-magmatic, close-to-solidus, deformational structures with variable intensity. They have a fine- to medium-grained porphyroclastic to protomylonitic textures, with orientated and deformed megacrysts of mesoperthite, arfvedsonite and quartz in a fine-grained, sometimes saccharoidal, matrix with alkali-feldspars (microcline and albite), quartz and mafic minerals. Millimetric miarolitic cavities or vugs (partially filled with accessory phases) occur in some samples. In most hand samples, yellowish minerals (mainly narsarsukite and zirconosilicates) can be seen. The accessory and rare phases include earlier-, late- and/or post-magmatic generations of aenigmatite, astrophyllite, narsarsukite, elpidite and other unidentified zirconosilicates of Na and K, neptunite, britholite-(Ce), nacareniobsite-(Ce), turkestanite, bastnäsite-(Ce), allanite-(Ce), titanite, anatase, together with fluorite and several unidentified phases (see Vilalva and Vlach, 2010, Vilalva et al., 2013; Siachoque et al., 2022). Siachoque et al. (2022) present an interpretation for the relative sequence of crystallisation of these minerals based on textural evidence. Narsarsukite is the most abundant rare accessory mineral in these rocks, reaching up to 3.2 vol.%.

Figure 2. (a) and (b) General structural and textural aspects of the peralkaline alkali-feldspar granites from the foliated facies of the Papanduva Pluton; (c) texture of the most deformed rock type under the microscope with crossed polarisers, showing mesoperthitic alkali-feldspar, quartz and arfvedsonite (commonly zoned) porphyroclasts in a recrystallised matrix with quartz, alkali-feldspars (microcline and albite) and mafic minerals. Yellow bar in (c) measures 1 mm. Mineral abbreviations as recommended by the IMA (Warr, 2021).

Samples and analytical procedures

The analytical work was carried out at the GeoAnalitica facility, Institute of Geosciences, University of São Paulo. After conventional petrographic analysis, seven samples from the foliated facies were chosen for major, minor and some trace-element analysis with the electron microprobe analyser (EMPA). Three of these samples were further analysed for trace elements using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS). The locations of the samples are shown in Fig. 1.

Electron microprobe analyses

Electron microprobe analyses were performed using the JEOL-JXA8600S and JEOL JXA-FE8530 electron microprobes. The latter instrument is equipped with a field emission electron gun. Both these instruments have five wavelength dispersive (WD) and one energy dispersive (ED) spectrometers. Quantitative WDS spot analysis was performed with the JEOL-JXA8600S equipment, and compositional mapping was obtained using a combination of WDS and EDS signals with the JEOL JXA-FE8530 equipment. The WDS analytical conditions were 15 kV, 20 nA and 5–10 μm for the column accelerating voltage, beam current and diameter, respectively. The analytical routine is summarised in Supplementary Table S1. Fluorine and Na were measured simultaneously in the first spectrometric round to minimise their loss. The matrix effect correction and conversion of the raw data to mass oxides were performed using PROZA software (Bastin and Heijligers, 1990). The analytical errors, checked against standards readings, are lower than 2% relative for the major, between 5 and 10% relative for the minor, and higher for trace elements. Back-scattered electron (BSE) images and compositional maps were obtained with a focused beam and the same analytical conditions. The dwell time for maps was set to 50 ms.

Literature data suggest that Fe in narsarsukite occurs mainly as Fe³⁺ (Wagner et al., 1991; Birkett et al., 1996; Schingaro et al., 2017). Its structural formula has been alternatively computed considering seven total cations or 11 (O+F). The first scheme does not account for potential vacancies in the Na site (Wagner et al., 1991) and might result in some excess in the Si and Ti sites in addition to in the O site, and the second scheme does not allow for the possible entry of (OH)^1– anions in the structure. Considering these drawbacks, we computed the structural formulae according to both approaches. We used the first scheme for the analytical data to better examine the substitution of O^2– by F^–1 in the O sites.

Laser ablation inductively coupled plasma mass spectrometry analysis

Laser ablation inductively coupled plasma mass spectrometry analysis was performed using Quadrupole Elan 6100DRC equipment from Perkin Elmer coupled with a UP-213 laser ablation system from NewWave Research, provided with a 213 nm Nd-doped YAG laser, following the procedures described in Andrade (2016). The analyses were carried out in raster mode with spot sizes varying from 30 to 50 μm. Back-scattered electron images were used to guide the selection of areas to be analysed.

The laser beam fluence and repetition rate were set at 8.5 J/cm² and 4 Hz, respectively. The following isotopes were measured: ⁷Li, ⁹Be, ²⁵Mg, ³¹P, ⁴²Ca, ⁴⁵Sc, ⁴⁹Ti, ⁵¹V, ⁵²Cr, ⁵⁵Mn, ⁵⁹Co, ⁶⁵Cu, ⁶⁶Zn, ⁶⁹Ga, ⁸⁵Rb, ⁸⁸Sr, ⁸⁹Y, ⁹⁰Zr, ⁹³Nb, ⁹⁵Mo, ¹¹⁸Sn, ¹³³Cs, ¹³⁷Ba, ¹³⁹La, ¹⁴⁰Ce, ¹⁴¹Pr, ¹⁴³Nd, ¹⁴⁷Sm, ¹⁵¹Eu, ¹⁵⁵Gd, ¹⁵⁹Tb, ¹⁶³Dy, ¹⁶⁵Ho, ¹⁶⁶Er, ¹⁶⁹Tm, ¹⁷³Yb, ¹⁷⁵Lu, ¹⁷⁹Hf, ¹⁸¹Ta, ²⁰⁸Pb, ²³²Th and ²³⁸U. The total acquisition time was set to 120 s, distributed equally between blank and ion signal measures. The signal integration and dwell times were 8.33 and 1.66 ms, respectively. The daily rate of oxide generation was controlled holding ThO⁺ formation below 1%. Glitter software (Griffin et al., 2008) was used for data acquisition, treatment, and conversions to concentrations. The synthetic NIST glass SRM-610 and the average TiO₂ content previously measured by WDS were used as external and internal standards, respectively.

Regardless of our efforts, reliable quantitative trace-element data for the post-magmatic narsarsukite generation was considerably more difficult to acquire. This is because of the small dimensions of the crystals, their common lamellar or fibrous habit, and their close association with arfvedsonite and aegirine. We were only able to obtain two reliable analyses, which were for crystals replacing arfvedsonite. Most of the other analyses were discarded as the measured Ca and Mg contents were relatively high, well above 400 and 300 ppm, respectively, and thus were considered to be mixed, non-representative values.

Results

Textural features

Two generations of narsarsukite were identified on the basis of their textural and compositional features. The most typical characteristics of each generation, as observed under the petrographic microscope, are shown in Fig. 3.

Figure 3. Main textural features of magmatic and post-magmatic narsarsukite generations from the Papanduva Pluton, Graciosa Province. (a) Larger magmatic crystals showing colour zoning and abundant minute inclusions in the yellow–orange intermediate zone, in plane-polarised light; (b) same crystal with cross polarisers, depicting the high birefringence mainly in crystal cores; (c) deformed crystal, bent around a quartz porphyroclast and with some translational slip; (d) zoned crystal partially altered by secondary mineral aggregates; (e) partially altered narsarsukite intergrown with albite laths; (f) post-magmatic lamellar and fibrous crystals associated with a somewhat sizeable magmatic crystal; (g) completely altered crystal associated with aegirine, note narsarsukite (probably post-magmatic) development along aegirine fractures and cleavage; (h) irregular post-magmatic narsarsukite replacing arfvedsonite; (i) crystal aggregate of narsarsukite replacing arfvedsonite. Images (d) (g) and (h) are plane-polarised light, images (c), (e), (f) and (i) are under crossed polarisers. White and black scale bars measure 0.5 mm.

The first generation occurs as interstitial euhedral to subhedral (1–2 mm long) crystals formed during the late-magmatic crystallisation stage and show some textural features similar to those described for narsarsukite in the Strange Lake granites (Birkett et al., 1996). These crystals have platy to tabular habits (a >> c) with high relief and second-order birefringence colours, with a weak colourless to pinkish pleochroism. Most larger crystals are zoned, with pinkish colours, high birefringence in their cores, and almost colourless and lower birefringence rims (Fig. 3a–d). Some crystals have a relatively homogenous and clean core surrounded by an intermediate poikilitic zone populated with a significant number of minute inclusions and a thin clean rim (Fig. 3a,b; cf. also Fig. 4). Deformed (bent and with some translational slip features) crystals are common (Fig. 3c). These crystals are very susceptible to alteration and several samples have been replaced by tiny brownish aggregates composed of late alteration minerals, including titanite, bastnäsite, phyllosilicate minerals and other unidentified phases. A remarkable texture observed in interstitial and relatively smaller crystals and in the intermediate and sometimes the rim zones of larger crystals, is a poikilitic intergrowth between narsarsukite and euhedral to subhedral albite laths (Fig. 3e).

Figure 4. BSE and compositional maps for the narsarsukite crystal shown in Fig. 3a,b. (a) BSE image (compositional mode); (b), (c), (d), (e) and (f) compositional maps for the indicated elements. EDS signals for Ti and WDS signals for the other elements are represented. The two elongated pits, which are most evident in the BSE images, are laser ablation raster lines. The white bar in the BSE image measures 250 μm; the colour bar represents signal intensities. See text.

The second generation consists of tiny crystals (≤ 1 mm long) with subhedral, lamellar and fibrous habits and seems to be less susceptible to late alteration. They occur mainly as minute inclusions in recrystallised quartz rims, as isolated crystals or crystal aggregates developed along the contacts between the other minerals (Fig. 3f), along fractures and cleavage planes of arfvedsonite and aegirine (Fig. 3g,h) or as irregular aggregates within arfvedsonite (Fig. 3i). The observed textural relationships, particularly those shown in Fig. 3h,i indicate that this generation replaces previous arfvedsonite and possibly aegirine.

A back-scattered electron image and qualitative compositional maps for Ti, Zr, Al, Fe and Nb for the magmatic crystals depicted in Fig. 3a,b are presented in Fig. 4. The images show the main zoning pattern observed in most large narsarsukite crystals, with relatively Zr- and Nb-rich and Fe- and Al-poor crystal cores. The poikilitic intermediate and the external rim crystalline zones are almost Zr-free. Furthermore, some areas have irregular Fe and Nb (sectorial?) zoning. The high-Zr and low-Ti phase intergrowth with narsarsukite in the intermediate poikilitic zone is an unidentified zirconosilicate with some Na and K, as determined from EDS qualitative analysis. This intergrowth might alternatively represent the co-precipitation of narsarsukite and a zirconosilicate still in the magmatic stage, indicating the saturation of alkali-bearing zirconosilicates in the melt or fluid-induced post-magmatic exsolution of the Zr-rich phase leaving a Zr-poor narsarsukite host. The most external and Ti-, Al-rich zones might represent late- to post-magmatic overgrowths.

Compositions and variability

Representative compositions for magmatic and post-magmatic narsarsukite are given in Tables 1 (WDS) and 2 (LA-ICP-MS). The complete dataset is given in Supplementary Tables S2 and S3, including structural formulae computed on the basis of seven cations and 11 (O+F) for the WDS data, and analytical errors and average standard readings (SRM-610 and SRM-612, from NIST, and BCR-2G, from USGS) used for analytical control for the LA-ICP-MS data.

Table 1. Representative compositions (WDS, wt.%) and atomic proportions for narsarsukite from the Papanduva Pluton, Graciosa Province, south-southeastern Brazil.*

* All Fe as Fe₂O₃; n.d. = not detected; 1= magmatic, 2 = post-magmatic; c = core, i = intermediate and r = crystalline rim zones; pk = poikilitic, ra = replacing arfvedsonite. See also Supplementary Table S2.

Table 2. Representative trace-element concentrations (LA-ICP-MS, ppm) for narsarsukite from the Papanduva Pluton, Graciosa Province, south-southeastern Brazil.

MR-02A, MR-06 and MR-21 = magmatic; MR-2A(*) = post-magmatic, replacing arfvedsonite. n.d. = not detected; n.a. = not analysed. See also Supplementary Table S3.

The most significant compositional variations are observed for Zr, Al, Fe³⁺ and Nb, all of which occupy the octahedral [Ti]-site. The contents (wt.%) of ZrO₂, Al₂O₃, Fe₂O₃ and Nb₂O₅ vary in the ranges 5.95–0.28, 0.96–0.39, 7.24–3.75, 1.21–0.08, respectively, in the magmatic narsarsukite, and in the ranges 0.27–0.00, 3.10–0.50, 7.42–4.17, 0.76–0.00, respectively, in the post-magmatic narsarsukite. The magmatic generation is richer in Zr and, on average, poorer in Al compared to the post-magmatic generation. Among the divalent cations, MnO has concentrations lower than 0.12 and is higher in magmatic crystals. BaO and CaO are lower than 0.16, and their contents lie in similar ranges in both generations, whereas K₂O is ≤ 0.15 (all quantities in wt.%).

The variations of Al and Fe³⁺ with Zr are shown in Fig. 5a. A slight decrease of both Al and Fe³⁺ with increasing Zr is observed in the compositional zoning of the larger crystals of the magmatic generation. These compositional relationships strongly support the textural interpretations that indicate two contrasted crystallisation generations. In addition to its higher contents, Al also has a significantly wider variation range in the post-magmatic crystals. The Al/Fe³⁺ ratio is a suitable variable to discriminate between narsarsukite generations in the Papanduva Pluton, being almost constant around 0.2 in the magmatic generation and between 0.2 and 1.2 in the post-magmatic generation. Notably, Al and Fe³⁺ contents show a positive correlation in the magmatic crystals, but a negative correlation in the post-magmatic crystals (Fig. 5b). The Fe³⁺ contents in narsarsukite reached maximum values at the transition between the magmatic and post-magmatic crystallisation stages. In the first case, Al and Fe³⁺ increase at an almost constant ratio of 1:4, a feature observed in all larger and zoned analysed crystals. In the second case, Al increases and Fe³⁺ decreases in a proportion close to 1:1.

Figure 5. Main compositional variations in narsarsukite from the Papanduva Pluton. (a) Al vs. Zr and Fe³⁺ vs. Zr (in the inset) and (b) Fe³⁺ vs. Al diagram. Note that Fe³⁺ and Al contents show a positive correlation in the magmatic generation, but a negative correlation in the post-magmatic generation. All quantities are expressed in atoms per formula unit (apfu). Colours represent magmatic (blue) and post-magmatic (red) generations.

In addition to the homovalent substitutions of Ti by Zr and Fe³⁺ by Al, the entry of trivalent cations in the narsarsukite structure is described by the heterovalent substitutions 2R⁴⁺ = R³⁺+ R⁵⁺ (1) and R⁴⁺+ O^2– = R³⁺ + F^1– (2), where R³⁺, R⁴⁺ and R⁵⁺ mainly represent (Al, Fe³⁺), (Ti, Zr) and Nb, respectively (Wagner et al., 1991; Birkett et al., 1996; Schingaro et al., 2017). These substitutions are plotted in Fig. 6. Wagner et al. (1991) also suggested that the substitution R⁴⁺ + 2O^2– = R²⁺ + 2F^–1 accounts for the entry of divalent cations (Mn, Mg, Ca and Ba), which occur in much lower concentrations in our samples (see Tables 1 and S2). Reaction (1) gives an excellent negative correlation for our dataset, with a determination coefficient R ² = 0.93 (Fig. 6a). However, it plays a minor role in the entry of the R³⁺ cations, as the Nb contents are much lower (≤ 0.036 atoms per formula unit, apfu) than Al and Fe³⁺, which sum up to 0.456 apfu. Thus, reaction (2), for which we only obtained a moderate negative correlation with a slope of –1.19 and R ² = 0.61 (Fig. 6b), is by far the most important in explaining the observed compositional variations. There are no significant and systematic variations in F or Nb between the magmatic and post-magmatic generations and the higher values of R³⁺+F^1– plotted in Fig. 6b reflect mainly the higher quantities of Al+Fe³⁺ in post-magmatic narsarsukite. The data points have a significant dispersion in this diagram and the slope departs from the ideal value of –1. These features are in part due to the non-consideration of some R³⁺ cations, such as the REE and Y, however it is most probably due to the entry of variable contents of (OH)^–1 anions in the O site, according to the extended narsarsukite formula suggested by Schingaro et al. (2017).

Figure 6. Heterovalent substitution mechanisms accounting for the main compositional variations for Papanduva narsarsukite. (a) 2R⁴⁺ vs. R³⁺+R⁵⁺. (b) R⁴⁺+O^2– vs. R³⁺+F^1– plot. R³⁺, R⁴⁺ and R⁵⁺ are (Al, Fe³⁺), (Ti, Zr) and Nb, respectively. Symbols as in Fig. 5. See text.

The magmatic and post-magmatic narsarsukite compositions from the Papanduva Pluton are compared in the ternary Fe³⁺–Al–Zr diagram (Fig. 7) with those of other important worldwide occurrences, which includes peralkaline SiO₂-oversaturated rocks, an alkaline pegmatite from Narsarsuk, the type locality, and hornfels associated with SiO₂-undersaturated rocks (nepheline syenites). They show an extensive variation range that overlaps with most of the available compositions. For example, the magmatic narsarsukite from Papanduva has almost as much Zr as narsarsukite in comendite from the Sirwa Massif, near Ouarzazate (Southern Morrocco), the occurrence with the highest Zr contents reported to date (Wagner et al., 1991). In contrast, the post-magmatic Papanduva narsarsukite can have almost as low Fe³⁺ contents as the specimen in the Illutalik trachyte dyke (Upton et al., 1976), from Illutalik (formerly spelled Igdlutalik) Island, Kujalleq (Greenland). The narsarsukite from lamprophyre dykes of the Murun Massif (Aldan Shield, Russia) has the highest and lowest known Fe³⁺ and Al contents, respectively (Schingaro et al., 2017), whereas the specimen in comendites/pantellerites and related vugs from Mayor Island (New Zealand) present relatively high Zr contents for given Al values (Read, 1991). The Papanduva magmatic narsarsukite has a similar compositional trend to the peralkaline granites from the Strange Lake in northern Québec–Labrador, Canada (Birkett et al., 1996), though with relatively lower Nb and Al and higher Fe³⁺ contents on average. The cores of the largest crystals also have higher Zr contents.

Figure 7. Fe³⁺–Al–Zr ternary diagram showing the compositional trends for the magmatic and post-magmatic narsarsukite from the Papanduva Pluton, compared with worldwide available compositions in peralkaline SiO₂-oversaturated rocks, pegmatites and hornfels associated with nepheline syenites. Data sources: Illutalik dyke (Kujalleq, Greenland): Upton et al. (1976); Gouré region (Niger): Jérémine and Christophe-Michel-Lévy (1961); Mayor Island (Bay of Plenty region, New Zealand) and Mont Saint-Hilaire hornfels (southern Quebec, Canada): Read (1991); Murun Massif (Aldan Shield, Russia): Schingaro et al. (2017); Narsarsuk pegmatite (type area, Narsarsuk, Kujalleq, Greenland) and Sirwa Massif (north of Ouarzazate, Southern Morrocco): Wagner et al. (1991); Shira Complex (northern Nigeria): Bennet et al. (1984); Papanduva Pluton: this work; Strange Lake granites (northern Québec-Labrador, Canada): Birkett et al. (1996); Sweetgrass Hills (north-central Montana, USA): Graham (1935, in Wagner et al., 1991). See text.

Trace elements

The most abundant trace elements in narsarsukite are the HFSE, Y and the mid to heavy rare earth elements (MREE–HREE), whereas the LILE (large-ion lithophile element) contents are low. In Fig. 8a, we plot selected trace elements normalised to the host rocks for samples MR-2A (magmatic and post-magmatic generations) and MR-21 (magmatic). The post-magmatic narsarsukite from sample MR-2A has higher Ba and light rare earth element (LREE) contents and lower HREE and Hf contents than the magmatic narsarsukite. Tantalum was not plotted in the diagram due to data unavailability for the host rocks, however its abundances range from 38 to 500 ppm and 64 to 268 ppm in magmatic and post-magmatic crystals, and its behaviour closely follows Nb (cf. Tables 2 and S3).

Figure 8. (a) Multi-element diagram showing the behaviour of selected trace elements in magmatic (1^st gen.) and post-magmatic (2^nd gen.) narsarsukite, normalised to the respective host rocks. Average data are presented for the magmatic generation; for the post-magmatic case, this is the only reliable result for this sample. Host-rock data from Vilalva and Vlach (2014). (b) REE patterns for magmatic (1^st) and post-magmatic (2^nd) narsarsukite from the Papanduva Pluton. Strange Lake data from Vasyukova and Williams-Jones (2019). Normalisation factors from Boynton (1984). See text.

The total REE and Y contents of primary narsarsukite range between 505–2353 and 473–2169 ppm, respectively. The post-magmatic crystals have lower contents, ranging from 910–864 ppm (MR-02A) and 602–2165 ppm (MR-06), respectively. The REE patterns of the magmatic generation (Fig. 8b) show high fractionation of the HREE over the LREE and within the HREE and the LREE, with average Ce_N/Yb_N, Ce_N/Sm_N and Gd_N/Yb_N ratios of 0.003, 0.295 and 0.031, respectively, and a strong negative Eu anomaly, with an average Eu* [=Eu_N/(Sm_N*Gd_N))^0.5] = 0.09. The post-magmatic crystals have a relatively flat distribution pattern, with a similar Eu anomaly. The average data for the Strange Lake narsarsukite (Vasyukova and Williams-Jones, 2019) is also plotted in Fig. 8b for comparison. This shows considerably higher total REEs and Y (averaging 14,453 and 4872 ppm, respectively), with a significant high-fractionated average pattern and a less pronounced negative Eu anomaly.

Discussion and final remarks

Textures and compositions of narsarsukite

Narsarsukite is a common accessory in strongly peralkaline granites from the Papanduva Pluton. It formed both in the magmatic and the post-magmatic crystallisation environments. Late-magmatic, relatively larger and compositionally zoned crystals were followed by poikilitic intergrowths between narsarsukite and albite formed in the latest melt pockets. Similar aegirine + albite, arfvedsonite + albite, and aenigmatite + albite intergrowths occur in the Papanduva Pluton. The albite laths within these intergrowths are almost pure (Ab ≥ 98% molar) and have significantly high Fe₂O₃ contents (Vilalva, 2007). These intergrowths suggest co-precipitation of these phases and closely resemble the so-called khibinitic texture typical of the SiO₂-undersaturated eudialyte-bearing, agpaitic, nepheline syenites (khibinites, e.g. Gerasimovsky et al., 1974; Ulbrich and Ulbrich, 2000). Therefore, albite (as an Al₂O₃-saturated phase) stabilised at the end of the magmatic crystallisation. Conversely, the fibrous narsarsukite crystals or crystals aggregates that appear interstitially or replace arfvedsonite and aegirine are typically post-magmatic (hydrothermal).

The Papanduva narsarsukite covers the compositional range for Fe³⁺, Al and Zr described in the literature. A well-defined Fe³⁺ inflection marks the compositional limit between the magmatic and the post-magmatic generations, with the latter being Zr-poor and Al-rich. This inflection reflects that Fe³⁺ was less available in the fluids and preferentially partitioned into other Fe³⁺-bearing phases, such as aegirine (cf. below). The increasing contribution of Al-bearing late fluids also contributed to this inflection. The relative proportions for the main R⁴⁺, R³⁺ and R⁵⁺ cations and O^2– and F^1– suggest the presence of some (OH)^1– anions in the O site, in agreement with the findings of Schingaro et al. (2017).

The narsarsukite investigated has significant total REE and Y (up to 2350 and 2170 ppm) contents. However, these contents are considerably lower than those reported for the narsarsukite in the Strange Lake Complex (Vasyukova and Williams-Jones, 2019), which also has highly fractionated REE patterns and smaller negative Eu anomalies. This difference could be due to the higher contents of HFSE and REE in the granites from Strange Lake, or to the occurrence of a different set of accessories competing for the REE and Y. The REE and other R³⁺ cations enter the six-fold coordinated [Ti]-site, which is significantly distorted in narsarsukite (Schingaro et al., 2017). The effective ionic radii (Shannon, 1976) of ^VITi⁴⁺ is 61 pm, and this explains the increasing preference for the smaller HREE in this site. However, even ^VILu³⁺, the smallest REE, has an effective radius (~86 pm), significantly higher than ^VITi⁴⁺, and this limits the maximum REE and Y contents in narsarsukite. The REE patterns obtained for the hydrothermal crystals replacing amphibole are characterised by relatively flat distribution patterns with higher LREE abundances, characteristics that are similar to those observed in hydrothermal zircon (e.g. Hoskin, 2005; Vlach, 2022).

The replacement of arfvedsonite and aegirine by narsarsukite

The replacement of arfvedsonite and possibly aegirine by narsarsukite during the post-magmatic crystallisation stage suggests the involvement of relatively Ti-rich late fluids in the mineralogical evolution of the host granites. This is supported by the occurrence of hydrothermal Ti-minerals such as neptunite, titanite and anatase in miarolitic cavities and vugs within the granites (Vilalva, 2007).

In this sense and considering idealised end-members, narsarsukite might be formed through the following replacement reactions:

(1)$$2[ {{\rm N}{\rm a}_3{\rm F}{\rm e}_4^{2 + } {\rm F}{\rm e}^{3 + }{\rm Si}_8{\rm O}_{22}{( {{\rm OH},{\rm F}} ) }_2} ] _{{\rm Arf}} + \, 3 [ {{\rm Ti}{\rm O}_2} ] _{{\rm Fluid}} + 3{\rm O}_2 = 3[ {{\rm N}{\rm a}_2{\rm TiS}{\rm i}_4{\rm O}_{11}( {{\rm OH},{\rm F}} ) } ] _{{\rm Nar}} + 5[ {{\rm F}{\rm e}_2{\rm O}_3} ] _{{\rm Hem}} + 4[ {{\rm SiO}_2 } ] _{{\rm Qz}} + [ {( {{\rm OH},{\rm F}} ) } ] _{{\rm Fluid}}$$

and

(2)$$2[ {{\rm NaFeS}{\rm i}_2{\rm O}_6} ] _{{\rm Aeg}} + [ {{\rm Ti}{\rm O}_2} ] _{{\rm Fluid}} + [ {( {{\rm OH},{\rm F}} ) } ] _{{\rm Fluid}} = [ {{\rm N}{\rm a}_2{\rm TiS}{\rm i}_4{\rm O}_{11}( {{\rm OH},{\rm F}} ) } ] _{{\rm Nar}} + [ {{\rm F}{\rm e}_2{\rm O}_3} ] _{{\rm Hem}}$$

The hydrothermal replacement of arfvedsonite by narsarsukite is probably favoured in relatively oxidising conditions. If Na metasilicate (Na₂Si₂O₅) or a similar compound and SiO₂ species are available in the fluid, aegirine will be formed preferentially instead of hematite, according to the reaction:

(3)$$[ {{\rm F}{\rm e}_2{\rm O}_3} ] _{{\rm Hem}} + [ {{\rm N}{\rm a}_2{\rm S}{\rm i}_2{\rm O}_5} ] _{{\rm Fluid}} + 2[ {{\rm Si}{\rm O}_2} ] _{{\rm Fluid}} = 2[ {{\rm NaFeS}{\rm i}_2{\rm O}_6} ] _{{\rm Aeg}}$$

Importantly, the occurrence of contrasted magmatic and hydrothermal aegirine generations and hydrothermal aegirine replacing arfvedsonite are standard features in these rocks (e.g. Vilalva et al., 2016; Vasyukova and Williams-Jones, 2019).

The compositions of the involved primary and post-magmatic crystallising phases can be used to infer the compositional characteristics of the post-magmatic fluid. The average compositions of arfvedsonite and primary aegirine in the narsarsukite-bearing granites are given in Supplementary Table S4 (see complete data set in Vilalva et al., 2016). In a gain–loss diagram (Fig. 9), the replacing narsarsukite has higher contents of Be, Al, Sc, Ba, HFSE, Y, REE, Th, U and Pb, in addition to Ti, Na, Al and Si, than arfvedsonite. This suggests that the fluid phase was relatively enriched in these elements. Conversely, K, Ca, Mn, Fe and Rb were released preferentially from the amphibole to the fluid phase.

Figure 9. Gain-loss diagram illustrating the ratios between major, minor and trace element concentrations, expressed as log(C_Nar/C_Afs), for the post-magmatic narsarsukite and the replaced arfvedsonite shown in Fig. 3i. The red lines correspond to the indicated actual concentration ratios (C_Nar/C_Afs).

Narsarsukite in peralkaline granites and a comment on the use of the term agpaitic

Narsarsukite is a significant and widely distributed primary accessory phase in some highly peralkaline granites such as those from the Papanduva Pluton and the Strange Lake Complex. It should be included as a characteristic and diagnostic Na–Ti-bearing mineral in the assemblages of these rocks, together with other rare Ti- and Zr-minerals, such as aenigmatite, astrophyllite and elpidite.

Given the inherent difficulties and compositional complexities, experimental constraints on the crystallisation of narsarsukite and other rare Na–K–Ti–Zr-bearing phases starting from peralkaline SiO₂-oversaturated compositions are rare or absent. Upton et al. (1976) simulated the crystallisation of a narsarsukite-bearing peralkaline trachyte under 100 MPa and saturated H₂O conditions at the magnetite–hematite buffer, however narsarsukite was not observed among the products. A qualitative analysis on the relative stability of narsarsukite and other Ti-minerals for the Strange Lake case was presented by Birkett et al. (1996), based on chemographic relations in the Na–Fe²⁺–Fe³⁺–Ti space.

The term ‘agpaitic’ was introduced by Ussing (1912) for some nepheline syenites from the Ilímaussaq Complex (south Greenland). It is specifically recommended for peralkaline nepheline syenites containing complex Zr- and Ti-silicate minerals, with corresponding particular geochemical signatures (see also Sørensen, 1960; Le Maitre, 1989). However, Vlach and Gualda (2007) used the this term to describe the mineralogy of the most evolved peralkaline granites in the Papanduva Pluton considering the widespread occurrence of titano- and zirconosilicate among the rare accessory phases. Later, Marks et al. (2011) proposed to extend the term ‘agpaitic’ to include peralkaline granitic rocks with (K, Ti)-, (K, Zr)-, (Na, Ti)- and (Na, Zr)-bearing rare mineral assemblages such as eudialyte + aenigmatite, astrophyllite + dalyite, aenigmatite + dalyite and astrophyllite + dalyite. We agree with Marks et al. (2011) and believe that these highly peralkaline granites, together with rhyolites, containing rare Zr- and Ti-silicate minerals might also be characterised as agpaitic to distinguish them from more common peralkaline SiO₂-oversaturated rocks. However this proposal needs to be discussed further by the petrological community and is beyond the scope of this work.