Götzenite, ideally NaCa₆Ti(Si₂O₇)₂OF₃ (Sokolova and Cámara, 2013), belongs to the rinkite group of minerals (Christiansen et al. (2003a) have named this group the rosenbuschite group; in Sokolova (2006) and Sokolova and Cámara (2013), it is Group I) in the seidozerite supergroup of TS-block (TS denotes titanium silicate) minerals (Sokolova and Cámara, 2017). The TS block consists of a central octahedral sheet O and two adjacent heteropolyhedral sheets H. These HOH layers are characteristic sheets of minerals in the polysomatic series of heterophyllosilicates, as described by Ferraris (2008). Here, we follow the classification based on the crystallographic description of the titanosilicates with TS blocks, as described by Sokolova (2006). The general formula of the TS block is A2PB2PM2HM4O(Si₂O₇)₂X_4+n, with the cations M2H of the H sheet; the cations M4O of the O sheet; A^P and B^P cations at peripheral (P) sites; and X4+n= 4X^O + nX^P, with the apical anions XP= XMP + XAP at the periphery of the TS block (see Eq. 10 in Sokolova, 2006, and Sokolova and Cámara, 2013, 2017). Thus, the ideal structural formula of götzenite can be described as A2P= Ca₂; M2H= Ca₂; M4O= NaCa₂Ti; (Si₂O₇)₂; and XO= XMO + XAO, with (XMO)2= (OF) and (XAO)2= F₂. Rinkite-group minerals including götzenite are characterized as having a fixed content of 1 apfu in the Ti-bearing site (Ti + Nb in götzenite) normalized to one unit of (Si₂O₇)₂ (Sokolova and Cámara, 2013, 2017; Day et al., 2022). Götzenite was first described by Sahama and Hytönen (1957) from a nephelinite of the extinct Mt. Shaheru volcano (Democratic Republic of Congo). It was observed that almost every crystal showed lamellar twinning with the twin axis b, which in the unit cell setting used here, corresponds to axis c. Fleischer (1958a, b) pointed out in his discussion that the minerals götzenite and Ca rinkite appear to be identical, and Sahama (1960) has proven that the calcium rinkite described by Slepnev (1957) is identical to götzenite and “may be termed a rare earth- and niobium-bearing strontian götzenite”. The structural relationship between götzenite and rinkite is discussed by Christiansen and Rønsbo (2000). Consequently, the name Ca rinkite was discredited by the Commission on New Minerals and Mineral Names of the International Mineralogical Association (Min. Mag. 33, 1962, 260–263). Under this aspect, the occurrence of götzenite termed Ca rinkite was mentioned by Chirvinskiy (1935) and Starynkevich-Borneman (1935), quoted after Val'ter et al. (1963). Neumann (1962) and Sahama et al. (1966) discussed the close relationship between rosenbuschite and götzenite by comparing the X-ray powder diffraction data and chemical compositions. The crystal structure of götzenite was solved by Cannillo et al. (1972) on a twinned crystal from Mt. Shaheru (Democratic Republic of Congo). It is closely related to the other rinkite-group minerals rinkite–Ce) Na₂Ca₄(REE)Ti(Si₂O₇)₂OF₃ (Cámara et al., 2011; Sokolova and Cámara, 2017), rinkite-(Y) Na₂Ca₄YTi(Si₂O₇)₂OF₃ (Pautov et al., 2019), nacareniobsite-(Ce) Na₃Ca₃(REE)Nb(Si₂O₇)₂OF₃ (Sokolova and Hawthorne, 2008), mosandrite-(Ce) (Ca₃REE)[(H₂O)₂Ca0.5□0.5]Ti(Si₂O₇)₂(OH)₂(H₂O)₂ (Sokolova and Hawthorne, 2013; Sokolova and Cámara, 2017), seidozerite Na₄MnZr₂Ti(Si₂O₇)₂O₂F₂ (Christiansen et al., 2003a), grenmarite Na₄MnZr₃(Si₂O₇)₂O₂F₂ (Bellezza et al., 2004a; Sokolova and Cámara, 2017), hainite-(Y) Na₂Ca₄YTi(Si₂O₇)₂OF₃ (Lyalina et al., 2015; Sokolova and Cámara, 2017), fogoite-(Y) Na₃Ca₂Y₂Ti(Si₂O₇)₂OF₃ (Cámara et al., 2017), batievaite-(Y) Ca₂Y₂Ti(Si₂O₇)₂(OH)₂(H₂O)4 (Lyalina et al., 2016), kochite Na₃Ca₂MnZrTi(Si₂O₇)₂OF₃ (Christiansen et al., 2003b; Sokolova and Cámara, 2017), rosenbuschite Na₆Ca₆Zr₃Ti(Si₂O₇)₄O₂F₆ (Christiansen et al., 2003a; Sokolova and Cámara, 2017), and bortolanite Na₂Ca_4.5Zr_0.5Ti(Si₂O₇)₂OF₃ (Day et al., 2022). In particular, the members fogoite-(Y), batievaite-(Y), hainite-(Y), and bortolanite are closely related to götzenite, all being triclinic in space group P1‾ with similar lattice parameters. The following criteria are used to distinguish between götzenite and the related minerals. In fogoite-(Y), Ca is replaced by Y in the H sheet of götzenite and by Na in the O sheet. In batievaite-(Y), Ca is replaced by Y in the H sheet of götzenite, and the Na site in the O sheet is essentially vacant. Hainite-(Y) differs from götzenite by containing more Na and Y, which occupy separate sites in the crystal structure (Lyalina et al., 2015). In bortolanite, 0.5 Ca in the H sheet of götzenite is replaced by Zr, and Ca in the O sheet is replaced by Na (Day et al., 2022).

Wöhlerite, ideally Na₂Ca₄ZrNb(Si₂O₇)₂O₃F (Dal Bo et al., 2022), was first described by Scheerer (1843). The mineral was found on Løvøya Island close to the city of Brevig in Norway. Chemical analysis revealed a silicate containing mainly Na, Ca, Zr, and Nb, with minor amounts of Mg, Mn, and Fe, where Nb was erroneously given as Ta, as discussed by Dal Bo et al. (2022). Following Biagioni et al. (2012) and Dal Bo et al. (2022), wöhlerite belongs to a group of minerals usually named the wöhlerite group but sometimes also designated as the cuspidine group (Merlino and Perchiazzi, 1988; see Dal Bo et al., 2022, for details), with the general composition X₈(Si₂O₇)₂W₄. Other members (for references, see Dal Bo et al., 2022) besides wöhlerite are cuspidine Ca₈(Si₂O₇)₂F₄, låvenite Na₂Ca₂Mn₂Zr₂(Si₂O₇)₂O₂F₂, normandite Na₂Ca₂Mn₂Ti₂(Si₂O₇)₂O₂F₂, niocalite Ca₇Nb(Si₂O₇)₂O₃F, janhaugite Na₃Mn₃Ti₂(Si₂O₇)₂(OH)₂OF, burpalite Na₄Ca₂Zr₂(Si₂O₇)₂F₄, baghdadite Ca₆Zr₂(Si₂O₇)₂O₄, hiortdahlite Na₂Ca₄(Ca_0.5Zr_0.5)Zr(Si₂O₇)₂OF₃, moxuanxueite NaCa₆Zr(Si₂O₇)₂OF₃, madeiraite Na₂Ca₂Fe₂Zr₂(Si₂O₇)₂O₂F₂ (Mills et al., 2021), and pilanesbergite Na₂Ca₂Fe₂Ti₂(Si₂O₇)₂O₂F₂ (Dal Bo et al., 2024). Goßner and Kraus (1933) determined the lattice parameters and assigned space group P2₁ or P21/m to wöhlerite, with a preference for the latter. Shibaeva and Belov (1962) solved the crystal structure representing a sorosilicate with cation polyhedra linked by Si₂O₇ groups. A detailed refinement of the crystal structure was done by Mellini and Merlino (1979). The mineral marianoite (Chakhmouradian et al., 2008) was discredited because it was found to be equivalent to wöhlerite (Miyawaki et al., 2020; Dal Bo et al., 2022; see also the discussion by Merlino and Mellini, 2009, and Chakhmouradian and Mitchell, 2009). Wöhlerite was also identified in a multi-phase mineral species named “guarinite”, consisting of three distinct domains with hiortdahlite II in domain I, wöhlerite in domain II, and hiortdahlite I in domain IV (Bellezza et al., 2012).

In order to support the identification of the mineral species investigated by X-ray diffraction methods and to provide possible models for the cation distribution in the crystal structure, chemical analyses were done by electron probe microanalyses (EPMA) using two different instruments at the University of Freiburg (UF) and the University of Bremen (UB). The first analysis at UF yielded a chemical composition close to that of wöhlerite which did not match the identification of the mineral by X-ray diffraction (XRD), indicating a götzenite-like species. This prompted us to search for another crystal with a similar appearance. Finally, several crystals could be separated among which both götzenite and wöhlerite could be identified. Both phases were analyzed by EPMA at UB. Since the sum of the weight fractions was significantly lower than 100 %, additional analyses were done using laser ablation inductively coupled plasma mass spectrometry (ICP-MS) to determine the contents of elements not routinely analyzed by EPMA.

Götzenite and wöhlerite crystals were analyzed at the Institut für Geo- und Umweltwissenschaften (University of Freiburg, UF) and the Faculty of Geosciences (University of Bremen, UB), each with a Cameca SX 100 equipped with five wavelength dispersive spectrometers (WDSs). Analytical conditions included 15 kV acceleration voltage, 15 nA beam current, and a focused beam (1 µm) at Freiburg and a 15 kV/20 nA/5 µm at Bremen, respectively. Raw data were corrected with the software PeakSight™ using the built-in PAP matrix correction (Pouchou and Pichoir, 1991). Spectrometer crystals and calibration standards are listed in Table 1. Figure 2 shows the back-scattered electron (BSE) image of a sample region where wöhlerite and götzenite are in direct contact with each other.

Laser ablation (LA) ICP-MS analyses of Rb, Y, Ba, rare earth elements (REE), Hf, Ta, Pb, Th, and U were carried out on a polished slice using a NewWave UP193ss laser coupled to a Thermo Element 2 at the Faculty of Geosciences, University of Bremen. Ablation was done with a laser pulse rate of 5 Hz, an irradiance of ∼ 1 GW cm⁻², and circular spots of 35 or 50 µm diameter for samples and 75 µm for standards. Helium (∼ 0.8 L min⁻¹) was used as the carrier gas, and argon (∼ 0.8 L min⁻¹) was added as the make-up gas; the plasma power was 1200 W. All isotopes were analyzed at low resolution with five samples in a 20 % mass window and with a total dwell time of 25 ms per isotope. Oxide formation was low (ThO/Th ≤ 0.1 %) so that no mass interference corrections were applied. The analyses were quantified using the Cetac GeoProTM software with NIST610 glass standard reference material (SRM) for external calibration and with Ca as the internal standard element. The Ca concentrations of the samples were derived from the previous EPMA analyses. Accuracy was monitored by analyses of the USGS basalt glasses BHVO-2G and BCR-2G (Jochum et al., 2005), along with the samples.

The results of the electron probe microanalysis (EPMA, wt %) and the laser ablation spectrometry analyses (LA-ICP-MSs, µg g⁻¹) are listed in Table 2. The major element concentrations are principally in accordance with the ideal end-member compositions if some allowance is made for cation exchanges such as Ti–Zr, Ti–Nb, or Ca–Sr. The low totals of the EPMA analyses are largely explained by the high concentrations of trace elements not analyzed, such as the rare earth elements (∑REE = 2.58 wt %–4.18 wt % in götzenite and 0.23 wt %–0.44 wt % in wöhlerite); Ta, Th, and U in götzenite; and Hf, Ta, and U in wöhlerite. The high REE and Sr contents of götzenite suggest substantial replacement of Ca²⁺ by Sr²⁺ and REE³⁺, with the latter possibly being a coupled substitution of 2 Ca²⁺ <=> REE³⁺ + Na⁺, which might explain the relatively low Ca and high Na contents of götzenite when compared to the ideal composition and to the composition of wöhlerite (see Table 2). The chondrite-normalized REE spectra of götzenite show a weak s shape with systematic enrichment of the lighter REE relative to the heavier REE. Those of wöhlerite show a pronounced s shape with a maximum at La and Ce, a minimum at Ho, and a gradual increase from Er to Lu (Fig. 3). The weight fractions of the main elements and trace elements obtained from EPMA and LA-ICP-MS, respectively, are listed in Table S1 in the Supplement. The formula units calculated from the weight fractions listed in Tables 2 and S1 are presented in Table 3, scaled to four Si atoms (two Si₂O₇ groups) per unit cell.

Single-crystal X-ray diffraction (SXRD) measurements of götzenite and wöhlerite crystals were carried out on a Bruker D8 Venture diffractometer using MoKα radiation (λ=0.71076 Å) at the University of Bremen, Fachbereich Geowissenschaften. The instrument is equipped with a curved TRIUMPH monochromator, a 0.3 mm collimator, a four-circle goniometer in κ geometry, and a Photon 100 (götzenite data) or a Photon III (wöhlerite data) CMOS area detector. Data collection parameters and single-crystal data are listed in Table 4.

Structure refinements were done with the program SHELX-97 (Sheldrick, 1997, 2008) as part of the WINGX suite (Farrugia, 1999) using scattering factors of the neutral elements. If possible, the occupancies of atoms were refined or they were fixed to the values determined from the chemical analyses if more than two elements with minor amounts were sharing one site. Crystal structure projections were drawn with the program STRUPLO (Fischer and Messner, 2025). Some entries with atom parameters were taken from the Inorganic Crystal Structure Database (ICSD; Belsky et al., 2002) when the atom parameters were not available in the original publications.

Bond valences (bond strengths) sij for pairs of cations i and anions j are calculated with the program VaList (Wills, 2018) according to the relation 1sij=eRo-RijB, with the bond length Rij and the empirical parameters Ro and B listed in Table 5, as used for the calculations in Sect. 4. It should be noted that bond valence values might be erroneous if they are calculated for mixed occupancies at cation and/or anion sites, as discussed by Bosi (2014). Systematic errors might be introduced by using the mean distances Rij in Eq. (1) resulting from the simultaneous refinement of the positional parameters of different atoms on the same site which are, in fact, displaced from the mean position with different bond lengths to the next neighbors in the local environment. However, the averaged bond valence values listed in Sect. 4.1 for götzenite and in Sect. 4.2 for wöhlerite can still be used as a rough guide to assess the choice of cations and anions at a certain site.

Crystals were prepared from the yellowish-brown part of the rock shown in Fig. 1. Only fragments could be separated with an anhedral shape not allowing routine identification. A fragment showing uniform extinctions under the polarizing microscope was selected and mounted on a goniometer head for single-crystal X-ray diffraction. Inspection of sections in reciprocal space (calculated with Bruker APEX2 instrument control software) yielded a preliminary C-centered monoclinic unit cell with space group C2/m and lattice parameters a=37.346(2) Å, b=7.3299(3) Å, c=5.7259(2) Å, and β=96.866°. Attempts to solve the crystal structure yielded rosenbuschite-like fragments with Si₂O₇ groups linked by MO₆ octahedra in a disordered arrangement where two components seemed to overlap. The R1 residual could not be refined below about 18 %. A possible explanation was the presence of twinning. Based on this hypothesis, a smaller unit cell with triclinic symmetry was found, which could be derived from the monoclinic unit cell according to the transformation 1/4 (a+b), c, -b. Refinement of lattice parameters after data collection in the triclinic setting yielded a=9.6191(3) Å, b=5.7342(2) Å, c=7.3386(2) Å, α=89.986(1)°, β=101.040(1)°, and γ=100.485(1)° in space group P1‾ with rotation twinning on [001] according to (-0.99997, 0.00023, -0.50199; -0.00008, -1.00002, 0.00034; -0.00001, -0.00002, 1.00000), which is close to the vector representation -a-1/2c, -b, c, with contributions of 40 % and 60 % from the two twin domains, respectively. Based on these lattice parameters and on the structural fragments found in the structure determination, the mineral could be identified as götzenite, though it was closely related to other species in the rinkite group of minerals as well (see the discussion in Sect. 4.1). We cross-checked the derived twinning law for rotation twinning on [001] with the software Geminography (Nespolo and Ferraris, 2006), confirming the result with some additional information: searching in the lambda region (20, 0, 6), the almost vertical plane was found to be 1‾04, and the twin obliquity was 0.06°, suggesting twinning by reticular pseudo-merohedry with twin index 2, referring to a pseudo-monoclinic I-centered twin lattice with parameters a=5.734 Å, b=7.339 Å, c=37.131 Å, α=90.04°, β=91.95°, and γ=89.99°. After transformation to the equivalent monoclinic C setting (matrix -1 0 1, 0 1 0, -1 0 0, as given in the International Tables for Crystallography, vol. A; Hahn, 2002), the lattice parameters are a=37.37778 Å, b=7.339 Å, c=5.734 Å, α=89.99°, β=96.869°, and γ=89.962°, which are very close to those of the above reported C-centered cell that was initially found from the experimental data. Similar calculations with Geminography for a potential reflection twinning on (100) show an obliquity of 1.95° and a twin index 2 (for definitions of twin obliquity and twin index, see Grimmer and Nespolo, 2006). Using initial atomic parameters from Cannillo et al. (1972), the refinement converged with a residual R1 = 8.2 %, and all anisotropic displacement parameters were positive within their standard deviations. Rotation twinning on [001] was already observed by Sahama and Hytönen (1957) (after transformation of lattice parameters according to a+b, c, -b). Figure 4 shows the resulting arrangement of lattice points for rotation twinning with the observed reciprocal lattice points in Fig. 4a and the lattice points simulated for rotation twinning in Fig. 4b using the software Blender, ver. 4.3.2 (Blender Foundation, 2025). The simulations were limited to ±3 for h, k, and l. In the l=2n layers in Fig. 4b, the lattice points of the blue domain cover the exactly overlapping lattice points of the red domain, except at the positions where no blue lattice points were generated.

Figure 4a demonstrates nicely that the observed lattice points of the two unit cells related by rotation twinning are almost perfectly aligned in an orientation parallel to b*, assuming a larger unit cell with monoclinic symmetry. The layers in the 3D reciprocal space patterns, shown to be nearly horizontal, have variable h and k indices and constant l indices. Accordingly, in Fig. 4a and b, with rotation twinning about [001] being assumed for both, in the comparatively sparsely populated layers with l=2n, the spots of both twin components coincide exactly and therefore cannot be distinguished; in the twice as densely populated layers with l=2n+1, the spots form well-separated alternating rows (red versus blue in 4b), extending along their b* directions. Overall, the pattern in Fig. 4b perfectly matches the pattern in Fig. 4a, both representing rotation twinning. Consequently, all refinements were done assuming rotation twinning on [001].

Another crystal was prepared from the same piece from which the götzenite crystal was obtained (see Fig. 1). SXRD immediately showed that it is not identical to götzenite, revealing reflection twinning on (100) according to -a-c, b, c, with contributions of 31 % and 69 % from the two twin domains, respectively, and with the lattice parameters a=10.842(1) Å, b=10.249(1) Å, c=7.2673(8) Å, and β=109.343(4)° in the monoclinic space group P2₁. This corresponds to the symmetry and unit cell parameters of wöhlerite. Assuming and testing rotation twinning on [001] according to -a-c, -b, c yielded a plausible model as well but resulted in slightly higher residuals. Analyzing the two potential twinning laws for wöhlerite with Geminography, we found pseudo-merohedry for reflection twinning on (100) with the normal direction [201], obliquity of 0.24°, and twin index 1. Similarly, for rotation twinning on [001], pseudo-merohedry was found as well, with 1‾02 for the normal plane, obliquity of 0.24°, and a twin index 1. From these results, [001] and (100) twinning were equivalent. Therefore, due to the slightly better residuals in the crystal structure refinement, reflection twinning is assumed here for the twin model of wöhlerite. Using initial parameters from Biagioni et al. (2012) (ICSD entry no. 187604), the crystal structure could be refined to R1 = 1.3 %.

Based on the results of EPMA and LA-ICP-MS (see Tables 2, 3, and S1), the resulting chemical formula can be written as Na_1.50(7)Ca_5.18(11)Sr_0.13(3)Fe0.03(1)2+Mn_0.014(4)Zr_0.06(2)La_0.08(1)Ce_0.11(2)Nd_0.017(2)TE_0.021(2)Ti_0.80(7)Nb_0.18(6)(Si₂O₇)₂O_0.9(1)F_3.5(2), with the trace elements TE = Y_0.002(1)Pr_0.008(2)Ta_0.0020(7)Th_0.007(2)U_0.0024(2) and minor traces of Rb, Ba, Eu, Gd, Tb, Dy, Ho, Er, Tm, Yb, Lu, Hf, and Pb at <0.001 atoms per formula unit (apfu). Thus, the following chemical composition is used for further discussion: Na_1.50Ca_5.18Sr_0.13Fe0.032+Mn_0.01Zr_0.06La_0.08Ce_0.11Nd_0.02Ti_0.81Nb_0.19(Si₂O₇)₂O_1.2F_2.8 omitting elements <0.01 apfu. Considering the fact that the sum of O (outside of Si₂O₇) and F, distributed over two sites in the crystal structure, should be limited to 4 apfu and that the chemical composition should be charge balanced, the number of F and O atoms was adjusted within a few standard deviations to meet these criteria. Similarly, the Ti and Nb content was adjusted to 0.81 Ti and 0.19 Nb, respectively, yielding 1 apfu in total. The chemical composition can be compared with the crystal chemical assignments derived from the site occupancies listed in Sect. 4.1.