Nomenclature of wöhlerite-group minerals

Abstract A nomenclature and classification scheme for wöhlerite-group minerals has been established. The general formula of minerals belonging to this group is given by X8(Si2O7)2W4, where X = Na+, Ca2+, Mn2+, Ti4+, Zr4+ and Nb5+; and W = F– and O2–. In addition, they may incorporate significant amounts of Mg2+, Fe2+, Y3+ and REE3+, where REE are the lanthanides. The main structural feature of these minerals is the four-columns-wide octahedral walls, which are interconnected through corner sharing and via the disilicate groups. The wöhlerite-group minerals crystallise in different unit-cell settings and symmetries, depending on the cationic ordering in the octahedral walls and the relative position of the disilicate groups. Different combinations of X and W constituents should be regarded as separate mineral species. In the case of coupled heterovalent substitutions at different crystallographic sites, it is advised to use the site-total charge approach to determine the correct end-member composition. Due to their structural and chemical features, wöhlerite-group minerals can easily form crystals with several micro domains, showing different crystal structures and chemical compositions. In addition, the crystallisation of polytypes is relatively common, although they should not be regarded as distinct mineral species. To date, ten minerals belonging to the wöhlerite group are considered as valid species: baghdadite, burpalite, cuspidine, hiortdahlite, janhaugite, låvenite, moxuanxueite, niocalite, normandite and wöhlerite. Låvenite and normandite are isostructural and are respectively the Zr and Ti end-members of a solid-solution series. Marianoite is discredited, as it is corresponds to wöhlerite. The ideal formula of hiortdahlite is revised as Na2Ca4(Ca0.5Zr0.5)Zr(Si2O7)2OF3, with one cationic site characterised by a valency-imposed double site-occupancy. These changes have been approved by the IMA–CNMNC (Proposal 20–D).


Introduction
The first mention of wöhlerite in the literature was made by Scheerer (1843) who studied the mineralogical paragenesis of the syenite pegmatites occurring on Løvøya island, Brevig area, Langesundsfjord, Norway. The chemical analysis performed by Scheerer (1843) indicated that wöhlerite is a silicate containing mainly Na, Ca, Zr and Nb, as well as minor amounts of Mg, Mn and Fe. At the time of wöhlerite's discovery niobium was not officially approved as a distinct chemical element, and consequently Scheerer (1843) had erroneously reported niobium in wöhlerite as tantalum. Among the wöhlerite-group minerals, cuspidine was the first to have its crystal structure solved (Smirnova et al., 1955) and the wöhlerite group is sometimes mentioned as the cuspidine group in the literature (e.g. Merlino and Perchiazzi, 1988;Chakhmouradian et al., 2008). However, as wöhlerite is the first described species of the group the name should be wöhlerite group in accordance with Mills et al. (2009). Merlino and Perchiazzi (1988) demonstrated that the nature of the crystal structure of the wöhlerite-group minerals (WGM) permits the crystallisation in different unit-cell settings and the formation of polytypes. In addition, they identified ten different structure-types that are possible within the fixed cell dimension a ≈ b ≈ 10.5 Å and c ≈ 7.3 Å. The WGM can form multi-domain crystals, as for instance in 'guarinite' from Monte Somma, Italy (Bellezza et al., 2012).
The new definition of the wöhlerite group has been approved by the Commission on New Minerals, Nomenclature and Classification (CNMNC) of the International Mineralogical Association (IMA) (Proposal 20-D; Miyawaki et al., 2020). The wöhlerite group includes mineral species that have the general formula X 8 (Si 2 O 7 ) 2 W 4 (Table 1), where X represents the cationic sites typically occupied by Na + , Ca 2+ , Mn 2+ , Fe 2+ , Ti 4+ , Zr 4+ and Nb 5+ ; and where W represents anionic sites with F -, (OH)and O 2-, which are not bonded to the silicate tetrahedra. The X sites have the same general topology and consequently a specific chemical element will have a different X site preference in different WGM species. The general formula proposed for WGM is similar to that of the rinkite group (seidozerite supergroup) minerals (Sokolova and Cámara, 2017), however rinkite-and wöhlerite-group minerals have different structures. The crystal structure of WGM is characterised by four-columns-wide octahedral walls, which interconnect through corner sharing and via the disilicate groups to create a framework (Fig. 1). The cationic ordering in the walls and the relative position of the disilicate groups lead to different symmetries (monoclinic and triclinic). The crystal structure of the borate minerals warwickite and yuanfuliite show the same type of framework, with isolated triangular BO 3 groups replacing the disilicate groups (Bigi et al., 1991;Appel et al., 1999).

Historical synopsis
Cuspidine Cuspidine, ideally Ca 8 (Si 2 O 7 ) 2 F 4 (Z = 2), was described by Scacchi (1876) from Monte Somma, Somma-Vesuvius complex, Italy. Cuspidine occurs in different geological environments such as skarns (Tilley, 1947;Taner et al., 2013), tuff ejecta (Federico and Peccerillo, 2002), pegmatitoid facies of venanzite (Bellezza et al., 2004a), calc-silicate xenoliths (Owens and Kremser, 2010), natrocarbonatite (Mitchell and Belton, 2004) and alkaline rocks (Andreeva et al., 2007). S.G.space group. *see Merlino and Perchiazzi (1988  Cuspidine is monoclinic, P2 1 /a, with a = 10.906, b = 10.521, c = 7.518 Å and β = 109.90°. The first structural investigation was provided by Smirnova et al. (1955) who considered the structure as an array of chains of edge-sharing Ca(O,F) 6 octahedra running parallel to the c axis, by analogy to the structure of ilvaite, epidote and tilleyite. Subsequent refinement of the cuspidine structure by Saburi et al. (1977) concluded that the coordination environments of the Ca sites are not solely octahedral but vary between six-, seven-and eight-fold. There are four Ca sites in total, with an average bond distance of 2.367, 2.404, 2.428 and 2.449 Å. In each column of the wall, the sites alternate between being small or large. Bellezza et al. (2004a) reported the incorporation of up to 0.22 Zr atoms per formula unit (apfu) and 0.32 Na apfu in cuspidine, following the substitution mechanism 2Ca 2+ + F -↔ Na + + Zr 4+ + O 2-. According to their structural model, Zr is incorporated on the small octahedral site lying in the outer columns of the wall (X1), whereas Na is incorporated on the large site (X3) lying in internal columns and connected by edge-sharing to the Zr-bearing octahedra (Fig. 2). Note that Taner et al. (2013) reported cuspidine from the Güneyce-Ikizdere Region in Turkey, with an unusually low F content (1.36 apfu), which may correspond to a hydroxide equivalent. Finally, Krzątała et al. (2018) reported from the Hatrurim complex, Israel, a 'uranian cuspidine' containing up to 0.64 U apfu, and only 0.98 F apfu. The oxidation state of uranium remains uncertain, as well as the exact insertion mechanism. Although, considering the similar ionic radii of Ca 2+ and U 4+ (U 5+ ) in octahedral coordination, the occurrence of uranium-bearing cuspidine is plausible.
The crystal structure of låvenite was refined on samples from the Lovozero alkaline massif, Russia (Simonov and Belov, 1960), Langesundsfjord, Norway (Mellini, 1981) and Los Archipelagos, Guinea (Biagioni et al., 2012). Låvenite is monoclinic, P2 1 /a, with a = 10.83, b = 9.98, c = 7.17 Å and β = 108.1°. These structural data, as well as the chemical data published elsewhere (see the references listed above) indicate clearly the occurrence of cationic substitution on the four X sites (Fig. 2). The larger X2 and X4 sites are dominated by Ca and Na, respectively. Note that X2 usually contains a high amount of Na leading to the mix occupancy close to Ca 0.60 Na 0.40 . The smallest X1 site is dominated by Zr, and the main substitution observed is Zr 4+ ↔ Ti 4+ . The last site, X3, has an intermediate size (≈ 2.23 Å) and is occupied by a mix of Ca, Fe, Mn and Zr. In låvenite, Mn 2+ is dominant on X3 though the high Ca contents reported in some samples may indicate that Ca could also be dominant, thus leading to a new end-member composition (Na 2 Ca 4 Zr 2 (Si 2 O 7 ) 2 O 2 F 2 ). The W2 site, bonded to the X3 and X4 sites, is fully occupied by F -, while the W1 site, bonded to the X1, X3 and X4 sites, is populated by O 2which is partially substituted by F -.
Normandite, ideally Na 2 Ca 2 Mn 2 Ti 2 (Si 2 O 7 ) 2 O 2 F 2 , is the titanium analogue of låvenite described initially from the Poudrette quarry, Mont Saint-Hilaire, Quebec, Canada (Chao and Gault, 1997). Note that a mineral with similar physical properties and composition had been reported prior to the normandite description from the Khibiny massif and Lovozero massif, Russia (Vlasov, 1966); Tenerife, Canary Islands (Ferguson, 1978) and Tamazeght, Morocco (Khadem Allah, 1993). Perchiazzi et al. (2000) refined the crystal structure of normandite from the Poudrette quarry and Amdrup Fjord, Greenland. Normandite is monoclinic, P2 1 /a, with a = 10.799, b = 9.801, c = 7.054 Å and β = 108.08°. The normandite structure confirmed the structural model and the cation distribution established in låvenite. Perchiazzi et al. (2000) also noted that their samples had an excess of Na and Ca with respect to the expected value of 4 apfu in total, and also had an excess of high-charge cations (Zr and Ti). At the same time, the sum of cations (Mn, Fe and Mg) located on the X3 site is significantly below 2 apfu, suggesting that the excess of Ca and Zr is hosted on the X3 site with an average bond distance of ≈2.20 Å (Fig. 2).
Baghdadite is monoclinic, P2 1 /a, with a = 10.432, b = 10.163, c = 7.356 Å and β = 90.96° (Biagioni et al., 2010). In addition to its chemical composition, the crystal structure of baghdadite is also unique for WGM as it shows the edge-sharing of two ZrO 6 octahedra in the internal columns of the wall (Biagioni et al., 2010) (Fig. 2). This structural feature is at odds with the Pauling's fourth rule, which states that high-valence cations tend to not share polyhedron elements (Pauling, 1929). In all other WGM the high-valence cations (Y 3+ , Ti 4+ , Zr 4+ and Nb 5+ ) do not share any ligands.

Burpalite
Burpalite, Na 4 Ca 2 Zr 2 (Si 2 O 7 ) 2 F 4 , was found for the first time within a fenitised sandstone in the contact zone of the Burpalinskii alkaline massif, North Transbaikal, Russia (Merlino et al., 1990). It is reported in only a few other localities around the world: Umbozero mine, Lovozero massif, Russia; Vesle Arøya, Langesundsfjorden, Norway; and Nanna pegmatite, Igaliku, Greenland (Friis et al., 2010). Chemical data on burpalite are scarce and the published data on the type material indicates a composition close to the end-member formula.
Burpalite is monoclinic, pseudo-orthorhombic, P2 1 /a, with a = 10.117, b = 10.445, c = 7.255 Å and β = 90.04° (Merlino et al., 1990). Site occupancies indicate that the X1 and X2 sites are fully occupied by Zr and Ca, respectively. The larger X3 and X4 sites are mainly populated by Na with minor substitutions of Ca. Bond valence analysis confirms the presence of only F on the W2 site, and the replacement of a small amount of F by O on the W1 site bonded to the Zr polyhedron (Fig. 2).
Structural investigations performed by Merlino et al. (1990) also indicate that some crystals of burpalite contain domains with a låvenite-type structure. Burpalite-and låvenite-type structures are related, as they are two distinct ordered members in a family of order-disorder structures. In addition, a so-called 'orthorhombic låvenite' was reported from the Burpalinskii massif (Portnov et al., 1966;Portnov and Sidorenko, 1975). This mineral has the same chemical composition as burpalite, but with a B-centred orthorhombic cell, a = 21.01, b = 10.05 and c = 7.23 Å, and is polysynthetically twinned on (100). That unit-cell however, can be transformed to monoclinic, a = 11.11, b = 10.05, c = 7.23 Å and β = 108.99°, twinned on (100), analogous to the unit-cell of låvenite. Consequently 'orthorhombic låvenite' is a polytype of burpalite (burpalite-1M 2 ), corresponding to a twinned maximum degree of order (MDO) polytype. Merlino et al. (1990) suggested that isotypic series may occur between burpalite and baghdadite through the substitution Na + + F -↔ Ca 2+ + O 2-. However, a burpalite-baghdadite series is unlikely because in burpalite the ZrO 6 octahedra are lying in the outer columns of the walls, while in baghdadite they are in the central columns (Fig. 2). Therefore, a transformation of burpalite into baghdadite requires a complete reordering of the cationic octahedral sites.
Niocalite crystals typically show polysynthetic or contact twinning and therefore the determination of the correct space group was not straightforward. Nickel et al. (1958) reports the unit-cell parameters a = 10.83, b = 10.42, c = 7.38 Å and β = 109.40°, and the probable space group Pa or P2/a. They showed that the twinned crystals have a pseudo-orthorhombic symmetry that is achieved through the twin plane ( 102). Li et al. (1966) conducted a structural investigation and proposed a model in P2 1 , though without taking the twinning into account. The presence of Si 2 O 7 groups ambiguously connected to the short edge of the NbO 6 octahedra in the Li et al. model prompted Mellini (1982) to reinvestigate the crystal structure of niocalite using the space group Pa. Mellini (1982) interpreted the diffraction pattern from twinned niocalite as the result of a twinning on the (100) compositional plane. The same author confirmed the presence of microstructural domains in niocalite through transmission electron microscopy investigations. Although, the structure refinement seems to indicate a disordered distribution of Ca and Nb within two different crystallographic sites (X4 and X6), it was concluded by Mellini (1982) that Nb and Ca are ordered on their respective sites (Fig. 2). The apparent disorder is the result of the averaging of the intensity data from the two twin-related domains, in which the X4 and X6 site positions are mutually inverted.
Note that the chemical analysis shows the presence of approximately two F apfu in niocalite, while the end-member formula has only one F apfu. The bond valence analysis shows that only the W3 site, which is shared solely by Ca polyhedra, is entirely populated by F. The remaining F content is distributed randomly on the other W sites. The incorporation of F on the W sites results from the charge balance mechanism linked to the Ca 2+ ↔ Na + substitution, and to the partial replacement of Nb 5+ by Ti 4+ and Zr 4+ .
Wöhlerite is monoclinic, P2 1 , with a = 10.823, b = 10.244, c = 7.290 Å and β = 109.00° (Mellini and Merlino, 1979). Shibayeva and Belov (1962) and Golyshev et al. (1973) performed the first structure refinements on wöhlerite and showed the presence of four-columns-large octahedral walls interconnected by corner sharing and Si 2 O 7 diorthosilicate groups. Mellini and Merlino (1979) provided a structure refinement of wöhlerite from Brevig, Norway, confirming the space group and showing that the structure is based on four independent Si sites, four Ca sites, two Na sites, one Zr site and one Nb site (Fig. 2). The bondvalence analysis indicates that only one anionic site is dominated by a monovalent anion. Biagioni et al. (2012) have refined the crystal structure of wöhlerite from Los Archipelagos, Guinea, which contains more Mn and F, and less Nb than the Norwegian material.
Chemical data provided by Mariano and Roeder (1989) on wöhlerite from different localities indicate that the chemical composition of wöhlerite is relatively consistent, and they note that the largest variations are observed for the Nb, Ti and F contents. According to the structural model they establish the coupled substitution Nb 5+ + O 2-↔ Ti 4+ + F -. However, the Ti increase is only half of the decrease of the Nb content, thus indicating a replacement of Nb by another chemical element. This is confirmed by subsequent crystal structure refinements in which the Nb site is populated by a significant amount of Mn or Mg (Bellezza et al., 2012;Biagioni et al., 2012). Andersen et al. (2010) and Sunde et al. (2018) showed only minor chemical variations in wöhlerite from different localities in the Larvik plutonic complex in Norway.

Janhaugite
Described from a sodium-rich alkali feldspar granite (ekerite) at Gjerdingen, Oslo region, Norway (Raade and Mladeck, 1983), janhaugite, ideally Na 3 Mn 3 Ti 2 (Si 2 O 7 ) 2 (OH) 2 OF, is an extremely rare mineral displaying some unique chemical features among the WGM. The Mn content in janhaugite (up to 2.4 apfu) is the highest recorded for any WGM. Electron microprobe analytical (EMPA) data show the presence of roughly one F apfu. Infrared spectroscopy confirmed the presence of OH groups, and the splitting of the O-H stretching frequencies (3550, 3510 and 3460 cm -1 ) may indicate that OH groups are distributed on three different crystallographic sites (Raade and Mladeck, 1983).
Janhaugite is monoclinic, P2 1 /n, with a = 10.668, b = 9.787, c = 13.931 Å and β = 107.82° (Annehed et al., 1985). The refinement of the crystal structure, coupled with the bond-valence analysis, indicate that only two W sites (W2 and W4) are populated mainly by monovalent anions (Fig. 2). While the sum of (OH + F) should be equal to three to keep the electroneutrality of the mineral, one can assume that the remaining monovalent anions are distributed randomly on the W1 and W3 sites.
Considering the local environment of the W1-W4 sites, W2 and W4 are tri-coordinated, wheras W1 and W3 are fourcoordinated, and it is most likely that the OH groups occupy the W2-W4 sites. In addition, there are H acceptors in the vicinity of W2-W4 (≈2.9 Å). Therefore, the end-member formula Na 3 Mn 3 Ti 2 (Si 2 O 7 ) 2 (OH) 2 OF is proposed for janhaugite, in order to show the ionic distributions on both the X and W sites, and to follow the rules defined for end-member formula by Hawthorne (2021).
In their model, Merlino and Perchiazzi (1985), showed that the average charge of the X7 site is +3, and fixed the X7 site population to (Zr 0.33  . New crystallographic and chemical investigations have been performed on samples from the type locality, in order to determine accurately the cationic distribution in the crystal structure. These new data are presented below.

Chemical classification of WGM
The members of the wöhlerite group show a range of compositions with the main cations being Na, Ca, Fe, Mn, Zr, Ti and   Nb while the main anions are O, F and OH -. These elements are also the main components of other disilicates with similar optical properties common to alkaline rocks, e.g. rinkite-group minerals of the seidozerite supergroup (Sokolova and Cámara, 2017;Pautov et al. 2019). Most petrological studies of alkaline rocks do not utilise techniques other than chemical data to classify minerals. Therefore, we have explored the feasibility of (1) distinguishing wöhlerite-group minerals from related minerals and (2) classifying WGM down to species level solely based on EMPA data. We have used a total of 908 analyses of WGM and related minerals from our own work and the literature: Aarden and Gittins (1974); Ferguson (1978); Eggleton et al. (1979); Raade and Mladeck (1983) Before attempting a classification or discrimination of species based purely on chemistry, all data were recalculated on the basis of 18 anions. We have maintained the identification of each analytical point as given in the respective papers, i.e. we have not changed mineral identifications. The WGM typically have no substitution on the Si sites and the X sites are filled, i.e. there are no, or only limited, vacancies in fresh material. Therefore, only data where 3.9 < Si apfu < 4.1 and 7.8 < Σ X apfu < 8.2 should be treated. We allow for some variation from ideal stoichiometry due to the challenges of analyses some of these minerals. Of the 908 analytical points 258 did not fulfil these criteria and therefore the following is based on the remaining 650 analyses. Keller and Williams (1995) used three ternary plots to classify WGMs, and we present our data in two of the same diagrams (Fig. 4). Contrary to the paper by Keller and Williams (1995) our data show significant overlaps between species. For example, hainite-(Y), kochite and rosenbuschite overlap in all plots and partly overlap with låvenite. Conversely, wöhlerite forms a distinct group in Fig. 4a,c,e,f. However, Fig. 4a shows that (i) janhaugite (WGM) overlaps with rinkite-group minerals grenmarite and seidozerite and (ii) låvenite (WGM) overlaps with rosenbuschite (rinkite group, seidozerite supergroup). Figure 4b,e show (i) a strong overlap between four minerals: wöhlerite-group minerals wöhlerite and hiortdahlite and rinkite-group minerals hainite-(Y), rinkite-(Ce) and rinkite-(Y). Chakhmouradian et al. (2008) plotted data for some WGM based only on divalent cations occupying the true octahedral sites. This method suffers from the same issues as Keller and Williams (1995) with large overlaps between different species, especially if data for hiortdahlite is included. Melluso et al. (2014) suggested other graphical methods to classify WGM and related species, but also concluded that the high degree of overlap between species and endmembers of species does not make these plots suitable for species determination.
The previously proposed methods for classification do not enable a satisfactory determination at species level from only chemical data. At first glance, the plots by Keller and Williams (1995) do seem to create some distinct groups and it may be possible to separate some species based on them. However, the data presented by Keller and Williams (1995) has in a sense already been filtered as they only presented WGM data. Therefore, these plots may help identify some WGM, but only when it is already known that the chemical data is actually from a WGM. Conversely, they fail when the mineral has not been identified, at least, to a group level. On the basis of the available chemical data it is not possible to make a graphic interpretation to identify WGM or distinguish them from related minerals solely based on chemical data. However, the plots may work if additional methods are applied, for example X-ray diffraction (XRD), to determine if the mineral is a member of the wöhlerite group or another chemically related group e.g. the seidozerite supergroup.
As the graphical methods do not enable distinction of WGM form related mineral groups, let alone identification at species level we have used the same chemical data to establish a workflow for treating chemical data. It must be stressed that the flow below requires the sequential removal of data so that the next step in the flow are criteria to be applied on the remaining data after samples have been removed by the previous step (Fig. 5).
After these steps a total of 122 data points remain corresponding to götzenite, hainite-(Y), kochite and rosenbuschite. In addition, one data point given as hiortdahlite and the Zr-Ti-cuspidine of Sharygin et al. (1996a) remains. The method above provides a good separation of WGM from seidozerite-group minerals.

Solid solution in WGM
The literature often refers to the WGM having a flexible structure resulting in large degree of solid solution (e.g. Perchiazzi et al., 2000;Mitchell and Belton, 2004;Chakhmouradian et al., 2008). Regardless of the diverse composition of the WGM, solid solutions are not as extensive as the chemical data may suggest. There seems to be a high degree of solid solution between cuspidine and niocalite, as well as between cuspidine and baghdadite, however, the solid solution between niocalite and baghdadite is limited (Fig. 4f). In baghdadite, the two Zr sites are edge sharing, resulting in a highly distorted site. If niobium completely replaces Zr, two Nb polyhedra would be edge sharing, which is highly unlikely to happen. Furthermore, a complete replacement of Zr by Nb is not possible as the additional two charges cannot be balanced becasue all anions in baghdadite are already oxygen. A full solid solution would be the coupled substitution 2Zr 4+ + O 2-↔ Ca 2+ + Nb 5+ + F -, but this is not a simple substitution as baghdadite and niocalite are not isostructural, and it would lead to a major structural change (Fig. 2). However, it is likely that there is a limited solid solution between the two species as indicated by Casillas et al. (2008). Keller and Williams (1995) suggested a degree of solid solution between niocalite and wöhlerite, but most of these data have more than 8.2 apfu X site cations and were removed by the above data processing. Wöhlerite and niocalite are not isostructural, therefore solid solution between the minerals not only requires substitution between Ca, Na, Zr and F, but also a change of the position of Nb in the structure between X5 and X6 (Tables 1,2; Fig. 3). Sharygin et al. (1996a) investigated a series of cuspidine and götzenite minerals from Pian di Celle in Italy and suggested a partial solid solution between cuspidine and an end-member with the composition NaCa 6 Zr(Si 2 O 7 ) 2 OF 3 . From powder XRD and Raman spectroscopy Sharygin et al. (1996a) showed this phase to be structurally more closely related to hiortdahlite than cuspidine, and in fact this composition corresponds to the recently approved mineral moxuanxueite (Qu et al., 2020).
Extended solid solution has been documented between låvenite and normandite (Perchiazzi et al., 2000) as Zr in låvenite and Ti in normandite both occupy the X1 site and hence can replace each other with no additional structural modifications. Similarly, the Mn/Fe ratio differs in låvenite and as both these cations occupy the X3 site, it is likely that Fe-equivalent species may be found of both minerals.
In summary, solid solution in the WGM is controlled by the crystal structure and occurs where no major structural modifications are required, e.g. between låvenite and normandite, or between cuspidine, baghdadite and niocalite. Therefore, it is not uncommon to find several different WGM in the same rocks as small chemical changes favour the formation of multiple species rather than creating solid-solution series.

Classification
The general formula for WGM can be expressed as X 8 (Si 2 O 7 ) 2 W 4 , without any further distinction among the X and W sites. As shown in Fig. 2, a specific chemical element will have a different preferential X site in different WGM. For instance, high-field-strength elements (HFSE) such as Ti, Zr and Nb, are hosted on crystallographic sites located in different columns of the walls and on different cationic X sites. Consequently, there is no crystal-chemical feature to assign elements to specific X sites. As a rule, the topological representation of the cationic walls has been made by drawing the projection of the wall along the crystallographic c axis. The labelling of the sites always starts with X1, which is the smallest site belonging to outer columns of the wall. The site siting in the outer column and connected to X1 though edge-sharing is labelled X2. Afterwards the sites are labelled in succession, going from the outer to the inner columns of the wall, and according to the symmetry of the mineral. The crystal structure of hiortdahlite is characterised by two cationic walls, where wall I contains the HFSE-dominated site located in the outer column (Fig. 3). The anionic W sites correspond to the ligands that are not bonded to the disilicate groups. W1 and W2 are the anionic sites bonded to X1 and X2, respectively. Depending of the symmetry of the mineral, W3 and W4 are either belonging to the same outer column as W1 and W2, or to the second outer column of the wall. The W anionic sites occurring between the two inner columns are the symmetrical equivalents of the W sites defined previously.
The classification of WGM is based on the occupancy of the structural X and W sites, and the different combination of chemical elements on these sites are regarded as different mineral species. Considering only chemical data is in some cases insufficient to correctly identify WGM at a species level. Consequently, a structural refinement is needed for a complete characterisation. The sequential application of the dominant-valency and the dominant-constituent rules is suitable for the determination of the end-member formula (Hatert and Burke, 2008;Bosi et al., 2019b). However, due to the possible occurrence of heterovalent substitutions on both the cationic and anionic sites, in some cases this method may fail to provide an end-member formula fulfilling the end-member definition (Hawthorne, 2021;Hawthorne et al., 2021). Therefore, we suggest using the site-total charge (STC) approach to define the end-member formula (Bosi et al., 2019a(Bosi et al., , 2019b. The relative position of the cation or anion inside the structural walls is not a valid criterion to define a new species, as it would lead to a proliferation of the number of species. A mineral phase showing the same chemical composition as previously described species but with a different cationic ordering must be considered as an analogue to that species. Polytypism is likely to occur among WGM (Merlino and Perchiazzi, 1988), and polytypes are not considered as different mineral species (Nickel and Grice, 1998). The polytypes must be described according to the nomenclature proposed in the IMA-CNMNC guidelines (Guinier et al., 1984;Nickel, 1993;Nickel and Grice, 1998).

Name, prefixes and suffixes
All the wöhlerite-group minerals have a distinct name, with no prefix or suffix. Due to the relatively large number of structural sites that are able to host the same cation (e.g. Ca 2+ ), we strongly discourage the use of a compositional suffix (e.g. calcio-, ferro-), as this is unclear to which structural site these prefixes will be related. For the same reason, we also discourage the use of prefix referring to the composition of anionic W sites. In the case of a species characterised by one crystallographic site dominated by Y or REE, we recommend using a new rootname as well as the Levinson suffix (Bayliss and Levinson, 1988).

Hiortdahlite end-member formula
In order to define the correct end-member formula of hiortdahlite new chemical data was collected and a crystal structure refinement was performed on a sample from the type locality, Langodden, Langesundsfjord, Norway (samples located in the NHM Oslo collections, catalogue number KNR 24099). The chemical data were acquired using the CAMECA SX100 electron microprobe housed at the Department of Geosciences, University of Oslo. The instrument was operated with a beam current of 15 nA and an acceleration voltage of 15kV, creating a 10 μm spot. The following natural and synthetic standards were used: albite (Na), zircon (Zr), wollastonite (Ca and Si), pyrophanite (Ti and Mn), REE orthophosphate (Y, La, Ce and Nd; Jarosewich and Boatner, 1991), MgO (Mg), Fe metal (Fe), Nb metal (Nb) and fluorite (F). The intensity data were corrected for inter-element overlaps and matrix effects using the PAP routine (Pouchou and Pichoir 1984). The chemical data are compared with those published by Andersen et al. (2010) on type-locality material (Table 3).
Single-crystal X-ray data were collected at room temperature with monochromated MoKα radiation (λ = 0.71703 Å -50 kV and 1 mA) on a Rigaku Synergy-S diffractometer equipped with a HyPix-6000He detector housed at NHM Oslo. The instrument has Kappa geometry and both data collection and subsequent data reduction, together with face-based absorption corrections were carried out using the Rigaku CrysAlisPro software. The details of the data collection and refinement are provided in Table 3. The initial structure solution in space group P 1 was determined by the charge flipping method using the Superflip algorithm (Palatinus and Chapuis, 2007), and the structural model was subsequently refined on the basis of F 2 with the Jana2006 software (Petříček et al., 2014). All atoms were refined with anisotropic thermal parameters. The details of the refinement are provided in Table 4 and the atoms coordinates, anisotropic thermal parameters and detailed bond distances are provided in the Supplementary tables S1, S2 and S3. Free refinement of the site-scattering factors showed that the X3 (Na), X4 (Ca) and X6 (Ca) sites were fully occupied, and that X1 (Zr), X2 (Ca) and X5 (Ca) sites were slightly deviating from a full occupancy. To extract accurate site-scattering factors, mixed occupancies Zr + Ca (X7) and Ca + Na (X8) were refined. The empirical formula is recalculated on the basis of (O+F) = 18 apfu. The site-scattering factors and the established cationic distribution in the crystal structure are provided in Table 5 and the  bond-valence table is provided in Table 6. The crystallographic information file has been deposited with the Principal Editor of Mineralogical Magazine and is available as Supplementary material. As shown by Merlino and Perchiazzi (1985), the main chemical substitutions occur on the X7 and X8 sites and the bondvalence analysis indicate a mixed O 2-+ Foccupancy on the anionic W1 site (Table 7). The refinement of the site scattering for the Ca1 (X2) and Ca3 (X5) sites shows a deficit and an excess of electronic density, respectively. Therefore, considering the average bond distances around these sites, Ca + Na and Ca + Y + REE have also been refined on these sites, respectively (Table 5). All Nb, Ti and Hf has been attributed to the Zr site (X1), which was then filled up with Zr. The remaining Zr was attributed to the X7 sites, along with all the Mn and Fe content. The X7 site was then filled with Ca. Refinement of the site scattering in the X8 position indicates an excess of electron density and therefore Ca was assumed to replace Na. The cationic distribution proposed is in excellent agreement with the different structural parameters and the bond-valence analysis (Tables 5, 6). Bond-valence sums show that the W2, W3 and W4 sites are populated by Fand that a substitution O 2-↔ Fis occurring on the W1 site (Table 6). Taking into account previous work and the new data presented here, the end-member formula of hiortdahlite is Na 2 Ca 4 (Ca 0.5 Zr 0.5 )Zr(Si 2 O 7 ) 2 OF 3 , with a constrained mixed occupancy of Ca 0.5 Zr 0.5 (M 2+ 0.5 M 4+ 0.5 ) Σ3+ , in order to obtain a chargebalanced formula.
The refinement provided in this work is slightly different than the one proposed by Biagioni et al. (2012) on a mineral phase structurally related to hiortdahlite ( Table 7). The incorporation of Ti 4+ on the larger X7 site (<X7-O> = 2.233Å) instead of the smaller X1 site (<X1-O> = 2.082Å) is unlikely considering the ideal bond distance for octahedrally coordinated Ti (2.005Å, Shannon (1976); Table 5, S3). The partitioning of Y and REE between the X5 and Na sites (X3) is not clear, however, we have not detected a refined site-scattering factor higher than 11 epfu for the X3 site in our investigations.
Hiortdahlite and moxuanxueite are the only approved WGM with a structure containing two topologically independent octahedral walls. The ideal compositions of the walls are given by The chemical compositions of the walls are similar, although in the first wall the Zr site is in the outer columns, while in the second wall the X7 site is in the central columns. Note that both walls in hiortdahlite are topologically and chemically unique among the WGM ( Table 2).

Discreditation of marianoite
Marianoite was discovered from the silicocarbonatite Prairie Lake complex, Ontario, Canada (Chakhmouradian et al., 2008), and was considered the Nb-analogue of wöhlerite. Its simplified formula is Na 2 Ca 4 (Nb,Zr) 2 (Si 2 O 7 ) 2 (O,F) 4 . Marianoite was described as monoclinic, P2 1 , with a = 10.846, b = 10.226, c = 7.273 Å and β = 109.33°. The highest Nb content reported by Chakhmouradian et al. (2008) for marianoite is 1.019 apfu, which is roughly 0.3 apfu more than in wöhlerite from the Langesundsfjord (Sunde et al, 2018). The approval of marianoite as a valid mineral species was based on the assumption that both Zr and Nb are disordered on the two smallest octahedral sites [average bond lengths: 2.031 (X6) and 2.080 Å (X1)] occurring in the structure. As a result of the similar X-ray and neutron scattering characteristics of Zr and Nb, it is not possible to solve the ordering issue between these two chemical elements by using standard diffraction methods. Following the description of marianoite, Merlino and Mellini (2009) published a discussion arguing that in wöhlerite and marianoite there is an ordering of Zr and Nb, with Nb preferentially occupying the smallest site (X6) and Zr the second smallest octahedra (X1) of the structure. The same authors have proposed to solve this question through anomalous scattering using synchrotron radiation sources that will allow Zr and Nb to be distinguished. Bellezza et al. (2012) and Biagioni et al. (2012) have used an ordered approach in their refinements. Readers are referred to Merlino and Mellini (2009) and Chakhmouradian and Mitchell (2009) for more information on that discussion. Following the classification system proposed herein for WGM, wöhlerite and marianoite are equivalent. If one considers a complete cationic ordering between Zr 4+ and Nb 5+ , the resulting endmember formula for both wöhlerite and marianoite is Na 2 Ca 4 X1 (Zr) X6 (Nb)(Si 2 O 7 ) 2 O 3 F. The maximum Nb content reported for marianoite is 1.02 apfu (associated with 0.85 Zr pfu) (Chakhmouradian et al., 2008), which is not enough to achieve Nb > Zr on both X1 and X6 sites and then define a  'Hiortdahlite II' is not an approved mineral species, although Merlino and Perchiazzi (1987) stated that the name 'hiortdahlite II' was approved by the IMA Commission on New Minerals and Mineral Names (merged with CNMNC in 2006), a subsequent new mineral proposal was never submitted. Hiortdahlite II was described by Aarden and Gittins (1974) in samples from the Kipawa River, Kipawa alkaline complex, Quebec, Canada. Roda Robles et al. (2001) reported hiortdahlite II from the Ilímaussaq alkaline complex, Greenland, Tamazeght complex, Morocco, and Iles de Los, Guinea, based on chemical analyses and powder X-ray diffraction. We have analysed material from the same locality in the Ilímaussaq complex, and all of the supposed hiortdahlite II crystals have a unit-cell setting and a crystal structure identical to those of hiortdahlite. Therefore, it may be questionable that hiortdahlite II exists at these localities. This further emphasises the need for full crystal structure refinement to correctly identify WGM at a species level.
Chemical analysis on the 'type' material gave the formula (Na 1.70 Ca 4.02 Mn 0.04 Fe 0.02 Mg 0.02 Al 0.02 Y 0.24 REE 0.08 Zr 1.16 Nb 0.04 Ti 0.02 ) Σ7.36 (Si 2.05 O 7 ) 2 O 0.82 OH 0.36 F 2.68 , and therefore hiortdahlite II was interpreted as a cationic-deficient analogue of hiortdahlite (Aarden and Gittins, 1974). However, recent chemical analyses performed on the type material of hiortdahlite (Andersen et al., 2010) indicate roughly the same amount of Zr per unit formula than in the material described by Aarden and Gittins (1974). Note that hiortdahlite II contains up 0.24 Y apfu, while in hiortdahlite from Langodden, the Y content is below 0.05 apfu. The total REE content is also slightly larger in hiortdahlite II than in hiortdahlite.
Hiortdahlite II is reported as triclinic, P 1, with a = 10.95, b = 10.31, c = 7.29 Å, α = 90.19, β = 109.02 and γ = 90.05° (Aarden and Gittins, 1974). The crystal structure refinement was performed on samples from Kipawa, and gave a structural model based on two independent topological octahedral walls (Merlino and Perchiazzi, 1987). The wall-I in hiortdahlite II has the same chemical composition as the wall-I in hiortdahlite, though the cationic distribution is not strictly equivalent (Fig. 3). The main difference between species is observed in the topology of wall-II, with a composition of [ X5 Ca X6 Ca X7 (Zr 0.5 Ca 0.5 ) X8 Na] and [ X5 Y X6 Ca X7 Ca X8 (Ca 0.5 Na 0.5 )] in hiortdahlite and hiortdahlite II, respectively. However, the difference between the crystal chemical formula provided by Merlino and Perchiazzi (1987) and the chemical data reported by Aarden and Gittins (1974) is significant, for instance 1.76 Y apfu is reported in the structure while the chemical data indicate 0.32 Y + REE apfu.
Consequently, the presence of a Y-dominant site in the structure of hiortdahlite II is questionable, and new investigations must be performed on hiortdahlite II material to explore if it is a polytype of hiortdahlite or a distinct and valid mineral species. A new mineral proposal would still be required to be submitted to the IMA-CNMNC.

Conclusions
The general formula of the wöhlerite-group minerals is given by X 8 (Si 2 O 7 ) 2 W 4 , where X denotes the cations occurring in the polyhedra building the four-column wall, and where W denotes the anionic sites that are not bonded to the disilicate groups. The crystal structure of WGM is characterised by 'octahedral walls' made of four columns of edge-sharing X sites. The symmetry of the different species can vary from monoclinic to triclinic, according to the cationic ordering on the X sites and the relative position of the disilicate groups. Distinction between the mineral species is made based on the dominant elements at the X and W sites, and different combinations of X and W constituents should be regarded as separate mineral species.
In addition to the classification scheme, the following changes have been approved by the IMA-CNMNC: (i) the end-member formula of hiortdahlite has changed to Na 2 Ca 4 (Ca 0.5 Zr 0.5 )Zr (Si 2 O 7 ) 2 OF 3 , with a valency-imposed double-site occupancy of (Ca 2+ 0.5 Zr 4+ 0.5 ) Σ3+ on the X7 site, and (ii) marianoite is discredited, as it is structurally and chemically equivalent to wöhlerite.
The chemical variation in WGM results in the formation of individual species rather than solid-solution series. The reason being that despite similar compositions between many of the members they are not isostructural, therefore, heterovalent substitutions typically require a complete reorder of the structure. Such reorders appear energetically unfavourable compared to the formation of another species, commonly resulting in rocks containing several WGM or even seidozerite supergroup minerals. Furthermore, the co-existence of different WGM and seidozeritesupergroup minerals in the same rock makes it a challenge for petrologists to identify minerals on a species level. We proposed a discrimination flow-chart for separating various WGM from chemically related species.