Introduction
Tourmalines are complex borosilicates whose general chemical formula may be written as: XY3Z6T6O18(BO3)3V3W, where X = Na+, K+, Ca2+ and ▫ (= vacancy); Y = Al3+, Fe3+, Cr3+, V3+, Mg2+, Fe2+, Mn2+ and Li+; Z = Al3+, Fe3+, Cr3+, V3+, Mg2+ and Fe2+; T = Si4+ Al3+ and B3+; B = B3+; V = (OH)– and O2–; W = (OH)–, F– and O2– (Henry et al., Reference Henry, Novák, Hawthorne, Ertl, Dutrow, Uher and Pezzotta2011). In this representation unitalicised letters X, Y, Z, T and B represent groups of cations hosted at the [9]X, [6]Y, [6]Z, [4]T and [3]B crystallographic sites (letters italicised), whereas V and W represent groups of anions accommodated at the [3]-coordinated O(3) and O(1) crystallographic sites, respectively.
Tourmaline has been studied extensively in terms of crystal structure and crystal chemistry (e.g. Foit, Reference Foit1989; Grice and Ercit, Reference Grice and Ercit1993; Hawthorne and Henry, Reference Hawthorne and Henry1999; Ertl et al., Reference Ertl, Hughes, Pertlik, Foit, Wright, Brandstatter and Marler2002; Novák et al., Reference Novák, Povondra and Selway2004; Bosi and Lucchesi, Reference Bosi and Lucchesi2007; Bosi, Reference Bosi2018; Henry and Dutrow, Reference Henry and Dutrow2011; Henry et al., Reference Henry, Novák, Hawthorne, Ertl, Dutrow, Uher and Pezzotta2011; Cempírek et al., Reference Cempírek, Houzar, Novák, Groat, Selway and Šrein2013; Bačík and Fridrichová, Reference Bačík and Fridrichová2020). Results show that the tourmaline structure is remarkably flexible in a chemical sense, accommodating ions of a wide range of size and charge, which in turn leads to Mg–Fe–Al–Cr–V disorder over the Y and Z sites.
Povondraite, ideally NaFe3+3(Fe3+4Mg2)(Si6O18)(BO3)3(OH)3O, was described by Grice et al. (Reference Grice, Ercit and Hawthorne1993). Until now, however, a chemical and structural study dealing with a statistically significant dataset of Fe3+-dominant oxy-tourmalines is missing. To explore the crystal-chemical aspects and their implications on the tourmaline supergroup, single-crystal structure refinements and electron microprobe data have been collected on five crystals from the type locality for povondraite: San Francisco Mine, Villa Tunari, Alto Chapare, Cochabamba, Bolivia (Walenta and Dunn, Reference Walenta and Dunn1979; Grice et al., Reference Grice, Ercit and Hawthorne1993; Žáček et al., Reference Žáček, Frýda, Petrov and Hyršl2000). All five single crystals were extracted from povondraite sample 110379 from the American Museum of Natural History, New York, USA. This is the first time that Mössbauer spectroscopic and single-crystal infrared spectroscopic data have been presented for this mineral.
Analytical methods and results
General comment
Initially, several crystal fragments of sample 110379 were analysed by electron microprobe; these proved to be chemically inhomogeneous, as reported in Hovis et al. (Reference Hovis, Tribaudino, Altomare and Bosiin press). More recent analyses, reported in the present study, were obtained on different relatively small fragments, which proved to be quite homogeneous as shown by the relatively low standard deviation values of the analysed elements (Table 4).
Single-crystal structure refinement (SREF)
Single-crystal X-ray diffraction (XRD) was undertaken on five crystal fragments of povondraite by mounting on a Bruker KAPPA APEX-II single-crystal diffractometer (Sapienza University of Rome, Earth Sciences Department) equipped with a CCD area detector (6.2 × 6.2 cm active detection area, 512 × 512 pixels) and a graphite-crystal monochromator using MoKα radiation from a fine-focus sealed X-ray tube. The sample-to-detector distance was 4 cm. A total of 1296 exposures (step = 0.4° and time/step = 20 s) covering a full reciprocal sphere were collected using ω and φ scan modes. Final unit-cell parameters were refined using the Bruker AXS SAINT program on reflections with I > 10 σI in the range 5° < 2θ < 75°. The intensity data were processed and corrected for Lorentz polarisation and background effects using the APEX2 software program of Bruker AXS. The data were corrected for absorption using a multi-scan method (SADABS). The absorption correction led to an improvement in R int. No violation of R3m symmetry was detected.
Structure refinement was done using the SHELXL-2013 program (Sheldrick, Reference Sheldrick2015). Starting coordinates were taken from Grice et al. (Reference Grice, Ercit and Hawthorne1993). Variable parameters included scale factor, extinction coefficient, atom coordinates, site-scattering values (for X, Y and Z sites) and atomic-displacement factors. Each structure was refined as a two-component inversion twin. Fully ionised-oxygen scattering factor and neutral-cation scattering factors were used. In detail, the X site was modelled using the Na vs. K scattering factors. The occupancy of the Y site was obtained by considering the presence of Fe vs. Mg, and the Z site with Fe vs. Al. Although there is more Mg than Al content, the latter was preferred to ZMg, having produced slightly better statistical indices and standard uncertainties. The T, B and anion sites were modelled, respectively, with Si, B and O scattering factors and with a fixed occupancy of 1, as refinement with unconstrained occupancies showed no significant deviations from this value. The position of the H atom bonded to the oxygen at the O(3) site in the structure was taken from the difference-Fourier map and incorporated into the refinement model; the O(3)–H(3) bond length was restrained (by DFIX command) to be 0.97 Å with isotropic displacement parameter constrained to be equal to 1.2 times that obtained for the O(3) site. Table 1 lists crystal data, data-collection information, and refinement details. Table 2 gives the fractional atom coordinates and equivalent isotropic-displacement parameters of a typical povondraite crystal (Pov1). Table 3 reports selected bond lengths for all studied crystals. The crystallographic information files showing all structural data have been deposited with the Principal Editor of Mineralogical Magazine and are available as Supplementary material (see below).
* Notes: R int = merging residual value; R 1 = discrepancy index, calculated from F-data; wR 2 = weighted discrepancy index, calculated from F 2-data; GoF = goodness of fit; diff. peak = maximum and minimum residual electron density. Space-group type = R3m and Z = 3. Radiation: MoKα = 0.71073 Å. Range for data collection (2θ) = 5°–75°. Data collection temperature = 293 K. Axis = phi-omega, frame width = 0.4°, time per frame = 20 s. Absorption correction method: multi-scan (SADABS, Bruker). Refinement method: full-matrix last-squares on F 2. Structural refinement program: SHELXL-2013 (Sheldrick, Reference Sheldrick2015).
* Isotropic-displacement parameters (U iso) for H(3) constrained to have a U iso 1.2 times the U eq value of the O(3) oxygen atom, respectively.
* Note: T-site occupancy = Si1.00 and B-site occupancy = B1.00
Electron microprobe analysis (EMPA)
The crystals used for X-ray diffraction refinement were analysed by wavelength dispersive spectrometry (WDS mode) using a Cameca SX50 instrument (CNR-Istituto di Geologia Ambientale e Geoingegneria, Rome, Italy) operating at an accelerating potential of 15 kV and a sample current of 15 nA, with a 10 μm beam diameter. The following standards, X-ray Kα lines and analyser crystals were used: jadeite (Na; TAP), periclase (Mg; TAP), orthoclase (K; PET), rutile (Ti; PET), wollastonite (Si, Ca; PET), metallic Zn and Mn (Zn, Mn; LIF), vanadinite (V; PET), fluorophlogopite (F; TAP), metallic Cr (Cr; PET), corundum (Al; TAP) and magnetite (Fe; LIF). The ‘PAP’ routine was applied (Pouchou and Pichoir Reference Pouchou and Pichoir1991). The results (Table 4) represent mean values of several spot analyses. Vanadium, Cr, Mn and Zn were below detection limits (< 0.03 wt.%).
* Notes: wt.% = weighted percent; apfu = atoms per formula unit (normalised to 31 anions); epfu = electrons per formula unit; b.d.l. = below detection limit; errors for oxides and fluorine are standard deviations (in brackets).
a Calculated by stoichiometry, (Y+Z+T) = 15.00 apfu.
b Fe oxidation state determined by Mössbauer spectroscopy.
Mössbauer spectroscopy
Several crystal fragments from povondraite sample 110379 were used for 57Fe Mössbauer spectroscopy performed at the Swedish Museum of Natural History, Stockholm, Sweden, using a conventional spectrometer system operated in constant-acceleration mode. Data were collected over 1024 channels; these were folded and calibrated against the spectrum of α-Fe foil. The spectrum (Fig. 1) was fit using the software MossA (Prescher et al., Reference Prescher, McCammon and Dubrowinsky2012) with two absorption doublets consistent with Fe3+ (Table 5). No indications of absorption due to Fe2+ was observed. In line with the site population results from SREF (see below, Table 6), the two doublets can be related to the occurrence of Fe3+ at both the Y and Z sites. However, the low resolution of the two doublets does not allow a definite site assignment.
* δ = centre shift, ΔE Q = quadrupole splitting, FWHM = full width at half-maximum.
* Notes: All crystals have the B and O(2,4,5,6,7,8) sites fully occupied by B3+ and O2–, respectively; obs = observed, calc = calculated from the site population.
The Mössbauer spectroscopy results are consistent with a synchrotron XANES study that reported 87–100% Fe as Fe3+ (Levy et al., Reference Levy, Henry, Roy and Dutrow2018).
Single-crystal infrared spectroscopy
Polarised Fourier-transform infrared (FTIR) absorption spectra were measured on a 33 μm thick doubly polished single-crystal section oriented parallel to the c-axis. A Bruker Vertex spectrometer attached to a Hyperion 2000 microscope and equipped with a halogen lamp source, CaF2 beamsplitter, ZnSe wiregrid polariser and InSb detector was used to collect spectra in the range 2000–13000 cm–1 at a resolution of 4 cm–1. Spectra recorded in polarised mode parallel to the crystallographic c-axis (E||c) show a weak band at 3440 cm–1, a very intense band around 3550 cm–1, a significant band at 3593 cm–1 and a weak band at 3699 cm–1 (Fig. 2). As observed typically for polarised tourmaline spectra in the (OH) range, the main band is off-scale for the E||c direction due to excessive absorption. Spectra obtained perpendicular to the c-axis show considerably weaker bands.
Bands above 3600–3650 cm–1 are normally considered to be due to (OH) at the W position [≡ O(1) site] (e.g. Gonzalez-Carreño et al., Reference Gonzales-Carreño, Fernández and Sanz1988; Bosi et al., Reference Bosi, Skogby, Lazor and Reznitskii2015). For the samples studied, the comparatively weak intensity of the band at 3699 cm–1 indicates low amounts of W(OH). On the basis of studies by Bosi et al. (Reference Bosi, Skogby and Balić-Žunić2016), Watenphul et al. (Reference Watenphul, Burgdorf, Schlüter, Horn, Malcherek and Mihailova2016) and Gatta et al. (Reference Gatta, Bosi, McIntyre and Skogby2014), the main FTIR bands at ~3440 cm–1, ~3550 and ~3593 cm–1 are probably caused by the occurrence of the atomic arrangements 3[Y(Fe3+)Z(Fe3+,Al)Z(Al)]–O(3)(OH)3, {2[Y(Fe3+)Z(Fe3+)Z(Mg)]–[Y(Fe3+)Z(Fe3+)Z(Fe3+)]}–O(3)(OH)3 and 3[Y(Fe3+)Z(Fe3+)Z(Mg)]–O(3)(OH)3, respectively, whereas the band at ~3699 cm–1 may be caused by the arrangements Y(Fe3+MgMg)–O(1)(OH)–X(Na,K).
Determination of number of atoms per formula unit (apfu)
In agreement with the structure-refinement results, the boron content was assumed to be stoichiometric (B3+ = 3.00 apfu). In fact, both the site-scattering results and the bond lengths of B and T are consistent with the B site fully occupied by B3+ and with the T site free of B3+ (e.g. Bosi and Lucchesi, Reference Bosi and Lucchesi2007). Iron oxidation state was determined by Mössbauer spectroscopy, which shows the exclusive presence of Fe3+. In accordance with Pesquera et al. (Reference Pesquera, Gil-Crespo, Torres-Ruiz, Torres-Ruiz and Roda-Robles2016), Li concentrations were considered insignificant as MgO > 2 wt.% in the povondraite crystals studied. The (OH) content and the formula were then calculated by charge balance with the assumption (T + Y + Z) = 15 apfu and 31 anions. The excellent agreement between the number of electrons per formula unit (epfu) derived from EMPA and SREF (within 1 epfu for all studied crystals) supports the stoichiometric assumptions.
Site populations
The povondraite site populations at the X, B, T, O(3) (≡ V) and O(1) (≡ W) sites of crystals Pov1,2,3,4,5 follow the standard site preference suggested for tourmaline (Henry et al., Reference Henry, Novák, Hawthorne, Ertl, Dutrow, Uher and Pezzotta2011) and are coherent with the information from FTIR absorption spectra. In particular, the presence of ~0.10 Al apfu at the T site is consistent with observed <T–O> distances ranging from 1.621–1.625 Å, which are larger than the expected value for <TSi–O> = 1.619(1) Å (Bosi and Lucchesi, Reference Bosi and Lucchesi2007). The Fe3+, Al and Mg site populations at the octahedrally coordinated Y and Z sites were optimised according to the procedure of Bosi et al. (Reference Bosi, Reznitskii, Hålenius and Skogby2017b) and Wright et al. (Reference Wright, Foley and Hughes2000), as well as by fixing the minor elements Ti4+ at the Y site. The resulting site populations are reported in Table 6, which also includes a comparison between the values of observed mean atomic number (as defined by Hawthorne et al., Reference Hawthorne, Ungaretti and Oberti1995) and those calculated from the site populations. The agreement between the refined and calculated values is very good and validates the distribution of cations over the X, Y, Z and T sites in the crystals studied. This site population is also supported by the comparison of weighted bond-valence sums (BVS) and weighted atomic valence (or mean formal charge) calculated from the site populations (Table 7). It is worth noting that the presence of W(OH) at the O(1) site, revealed by FTIR spectra, has been quantified by using the empirical equation W(OH) = {2 – [1.01⋅BVS(O1)] – 0.21 – F} of Bosi (Reference Bosi2013). As a result, O and (OH) are partially disordered over the O(1) and O(3) sites.
* Note: The O(2,4,5,6,7,8) sites are fully occupied by O2–. Bond valence parameters from Gagné and Hawthorne (Reference Gagné and Hawthorne2015).
Discussion
All the crystals studied can by identified as povondraite (Table 6). More specifically, they are consistent with oxy-tourmalines belonging to the alkali group (Henry et al., Reference Henry, Novák, Hawthorne, Ertl, Dutrow, Uher and Pezzotta2011), Na-dominant at the X position and oxy-dominant at the W position with O2– > (F + OH) in the tourmaline general formula. The Y position is dominated by Fe3+ and the Z position requires a double site-occupancy (Fe3+4Mg2) for formula electroneutrality. Collectively, these constituents lead to the povondraite end-member NaFe3+3(Fe3+4Mg2)(Si6O18)(BO3)3(OH)3O.
The five analysed crystals show a substitution series dominated by Fe3+ and Al, which leads to bosiite, ideally NaFe3+3(Al4Mg2)(Si6O18)(BO3)3(OH)3O, by the substitution ZFe3+ ↔ ZAl, and to oxy-dravite, ideally Na(Al2Mg)(Al5Mg)(Si6O18)(BO3)3(OH)3O, by the substitution YFe3+3 + ZMg ↔ 2YAl2 + YMg + ZAl. As a result, the comprehensive substitution reaction along the povondraite–bosiite–oxy-dravite series is: YFe3+3 + ZMg + ZFe3+4 ↔ YAl2 + YMg + ZAl5. The latter can be summarised as Fe3+Al–1, as shown in Fig. 3 where the substitution of Fe3+ for Al defines a line having a slope of ~45°. Any deviation from this line may be ascribed to other substitutional mechanisms such as coupled substitutions related to O–(OH) at the O(1) (= W) and O(3) (= V) sites. A similar Fe3+Al–1 substitution is reported in Žáček et al. (Reference Žáček, Frýda, Petrov and Hyršl2000).
Sodium and K show similar variations, 0.79–0.89 apfu and 0.12–0.24 apfu, respectively, which are strongly correlated with each other (coefficient of determination, r 2 = 0.997) and affect the <X–O> mean bond-length variation (2.729–2.744 Å). In particular, the relatively high K content is related to the increase in Fe3+ (Table 4). As noted by Bačík et al. (Reference Bačík, Uher, Sykora and Lipka2008), the incorporation of a relatively large cation such as K (+ Na) into the povondraite structure should be favoured by the larger unit-cell of Fe3+ – relative to Al-dominant tourmalines such as bosiite or oxy-dravite (see below). This mechanism is different from that involved in maruyamaite (K- and Al-dominant tourmaline) in which the substitution K → Na occurs only under high-pressure conditions (Berryman et al., Reference Berryman, Wunder and Rhede2014; Lussier, et al., Reference Lussier, Ball, Hawthorne, Henry, Shimizu, Ogasawara and Ota2016).
As for the octahedrally coordinated cations, Mg varies from 1.84 to 2.02 apfu and occupies both the Y and Z sites, whereas Al varies from 0.18 to 1.50 apfu and is ordered at the Z site. Ferric iron varies from 5.44 to 7.07 apfu, showing a rather disordered distribution over the Y and Z sites.
Despite the significant YFe3+ variations in the present povondraite crystals, the <Y–O> values are practically constant; the increase in contents of the smaller cation YFe3+ (2.26–2.69 apfu), accompanied by decrease in contents of the larger cation YMg (0.31–0.66 apfu), do not produce any decrease in <Y–O> (2.043–2.045 Å). Therefore, we may infer that the accommodation of Fe at the Y site should produce a <Y–O> expansion that compensates for the differences in size between Fe3+ and Mg2+ substituent ions. This expansion may be shown by the smaller values of bond-valence sum at Y (2.69–2.74 vu) with respect to the weighted atomic valence at Y (2.81–2.90 vu) (Table 7), indicating that the Y-cation is underbonded and bond lengths in the YO6 polyhedron are stretched (Bosi, Reference Bosi2014).
In general, the variation of the structural parameters is dominated by Fe3+ (or Al). No significant correlation occurs between <Y–O> and YFe3+ (Fig. 4a), whereas the <Z–O> variation (1.997–2.017 Å) is positively correlated with ZFe3+ (Fig. 4b) and negatively correlated with ZAl (not shown; r 2 = 0.98). Similarly, the a- and c-parameter are positively related to ZFe3+ (Fig. 5).
The a- and c-parameters show similar variations in the studied crystals, 16.1679(2)–16.2366(3) Å and 7.4122(1)–7.4688(2) Å, respectively, which are positively correlated (r 2 = 0.98). Povondraite has relatively large unit-cell parameters with respect to other tourmalines due to the larger size of Fe3+ compared to other trivalent cations V3+ > Cr3+ > Al3+ (Bosi, Reference Bosi2018). The plot of a against c (Fig. 6) shows the variation of these parameters in the tourmaline-supergroup minerals (a range ~15.60–16.25 Å and c range ~7.00–7.50 Å) and their increase with increasing Fe3+ or decreasing Al. In particular, the smallest a- and c-parameters are those of synthetic Al-B-tourmalines, whose compositions lead to the end-members NaAl3Al6(Si3B3O18)(BO3)3(OH)3(OH) (Schreyer et al., Reference Schreyer, Wodara, Marler, van Aken, Seifert and Robert2000; Marler et al., Reference Marler, Borowski, Wodara and Schreyer2002) and NaAl3Al6(Si4B2O18)(BO3)3(OH)3O (Kutzschbach et al., Reference Kutzschbach, Wunder, Rhede, Koch-Mueller, Ertl, Giester, Heinrich and Franz2016), whereas the largest ones are of povondraite crystal pov5 of the present study.
Acknowledgements
The authors are grateful to George E. Harlow (American Museum of Natural History, New York, USA) for kindly furnishing povondraite sample 110379 to G.H. (Hovis et al., Reference Hovis, Tribaudino, Altomare and Bosiin press) along with well wishes for the further study reported here. Chemical analyses were done with the kind assistance of Beatrice Celata to whom the authors express their gratitude. Comments by the Structural Editor (P. Leverett) and reviewers Peter Bačík and Darrell Henry are very much appreciated. F.B. acknowledges funding by Sapienza University of Rome (Prog. Università 2020) and by the Italian Ministry of Education (MIUR)–PRIN 2020, ref. 2020WYL4NY.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1180/mgm.2022.132
Competing interests
The authors declare none.