The calcioferrite group approved and kingsmountite redefined

The four phosphate minerals calcioferrite (Blum, 1858), montgomeryite (Larsen, 1940), kingsmountite (Dunn et al., 1979) and zodacite (Dunn et al., 1988) have been reported to have in common the same general formula, Ca₄AB₄(PO₄)₆(OH)₄·12H₂O and similar powder X-ray diffraction (PXRD) patterns, with C-centred monoclinic unit cells, a ~ 10.1, b ~ 24.1, c ~ 6.3 Å and β ~ 91.2°. A and B are divalent and trivalent cations, respectively, with A²⁺ and B³⁺ = Mg²⁺ and Fe³⁺ for calcioferrite, Mg²⁺ and Al³⁺ for montgomeryite, Fe²⁺ and Al³⁺ for kingsmountite, and Mn²⁺ and Fe³⁺ for zodacite. The minerals have been described in the literature as forming the calcioferrite group (Lafuente et al., 2014) or the montgomeryite group (Dunn et al., 1988), but neither the group nor the group name had previously been approved by the International Mineralogical Association’s Commission on New Minerals, Nomenclature and Classification (CNMNC). Following the recent approvals by the CNMNC of fanfaniite, Ca₄MnAl₄(PO₄)₆(OH)₄·12H₂O (Grey et al., 2018), the Mn-analogue of montgomeryite, and the related mineral “aniyunwiyaite”, Ca₃Mn²⁺Fe²⁺Al₄(PO₄)₆(OH)₄·12H₂O (IMA2018-054), the need to formalise a group for these minerals became clear. In the course of establishing the group, we re-examined the holotype kingsmountite specimen and found that the species corresponds to “aniyunwiyaite”, which is herein discredited.

Calcioferrite was the first of the minerals in this group to be described (Blum, 1858). Blum reported chemical analyses (wt%): Fe₂O₃ 24.34, Al₂O₃ 2.90, CaO 14.81, MgO 2.65, P₂O₅ 34.01, H₂O 20.56, Σ99.27 and gave a measured density of 2.52 g cm⁻³. On the basis of the Blum (1858) analyses, Palache et al. (1951) proposed the formula Ca₃Fe₃(PO₄)₄(OH)₃·8H₂O for calcioferrite. The second mineral to be described, montgomeryite, was assigned the ideal formula Ca₄Al₅(PO₄)₆(OH)₅·11H₂O by Larsen (1940). It was only when the crystal structure of montgomeryite was determined by Moore & Araki (1974) that the correct formula Ca₄MgAl₄(PO₄)₆(OH)₄·12H₂O was proposed. Mead & Mrose (1968) had previously suggested that calcioferrite and montgomeryite are isostructural, but the correct ideal formula for calcioferrite, Ca₄MgFe₄(PO₄)₆(OH)₄·12H₂O, was not presented until 1983 by Dunn et al. Kingsmountite, ideally Ca₄FeAl₄(PO₄)₆(OH)₄·12H₂O, from the Foote mine, North Carolina, was identified and named by Dunn et al. (1979) as the Fe²⁺ analogue of montgomeryite. Those authors noted that individual crystals were extremely small and imperfect, and that a single-crystal study could not be undertaken. They noted, however, the close similarity of the PXRD pattern to that of montgomeryite and they refined C-centred monoclinic unit-cell parameters. In their 1983 paper, Dunn et al. reported that chemical analysis of a kingsmountite-like mineral from the Hagendorf pegmatite gave a formula corresponding ideally to Ca₄MnFe₄(PO₄)₆(OH)₄·12H₂O. Dunn et al. (1988) subsequently described a mineral from Mangualde, Portugal, with composition similar to the Hagendorf mineral and named it zodacite. They confirmed the unit-cell symmetry and metrics using single-crystal methods and refined the cell parameters from the PXRD data.

Some doubt remains concerning the correct space group for these minerals. Moore & Araki (1974) refined the structure of montgomeryite from Fairfield, Utah, in space group C2/c. The structure is based on zig-zag 7 Å chains of corner-connected B-centred octahedra along [101] that are interconnected via PO₄ tetrahedra forming [B₂(PO₄)₃(OH)₂] layers parallel to {010}, shown in projection in Fig. 1. Channels along [001] in the layers are occupied by Ca (site Ca2 in Fig. 1). The layers are connected via [001] edge-shared chains of Ca1O₈ polyhedra and AO₆ octahedra (A = Mg for montgomeryite). In the C2/c model, the A site is disordered with only 50% occupancy (Moore & Araki, 1974). Subsequently, Fanfani et al. (1976) reported that the diffraction patterns for montgomeryite from the same locality displayed weak additional reflections that violated the c-glide extinction condition. They refined the structure in C2. In the lower-symmetry space group, the A site splits into two independent sites and Fanfani et al. (1976) proposed that Mg was ordered in one of the two sites. It is possible, however, that the appearance of the very weak forbidden reflections was due to multiple diffraction, and any ordering is short-range. In their precession study of zodacite, Dunn et al. (1988) did not report any reflections violating the c-glide, and proposed that the space group is either C2/c or Cc. In a recent study on calcioferrite from the Moculta quarry, South Australia, Lafuente et al. (2014) obtained an excellent single-crystal refinement (R = 3.9%), with location of H atoms, based on the space group C2/c.

The two new minerals, fanfaniite (Grey et al., 2018, 2019) and “aniyunwiyaite” (IMA2018-054) are both from the Foote lithium mine, North Carolina. Single-crystal refinements were conducted on both minerals. Fanfaniite refined satisfactorily in space group C2/c, but “aniyunwiyaite” has triclinic symmetry with a doubling of a. The superstructure along [100] was found to be due to ordering of octahedrally coordinated Mn²⁺ in one of the four independent eight-coordinated Ca sites in the triclinic structure.

The “aniyunwiyaite” proposal was based on the characterization of specimens found in small dissolution cavities and along thin solution fractures in a partially altered spodumene-bearing pegmatite boulder on the eastern dump at the Foote mine, North Carolina (35°12′40″ N, 81°21′20″ W). The mineral occurs as hemispherical radial aggregates, ~0.5–1 mm in diameter, consisting of transparent blades with widths 20–100 μm, lengths up to 500 μm and thickness generally < 5 μm (Fig. 2a). The blades are flattened on {010}, elongate on [001] and exhibit the forms {010}, {100} and {20$1¯$}. They are flexible and elastic and have good cleavage on {010}. The measured indices of refraction, electron-microprobe analyses, density, unit-cell parameters and PXRD lines for the mineral are compared with those for kingsmountite from the same locality (Dunn et al., 1979) in Table 1.

The measured properties for the two minerals are almost identical. In addition to the C-centred monoclinic unit-cell parameters for kingsmountite in Table 1, we give the primitive cell parameters obtained by applying the transformation matrix (1 0 0, $0.5¯$ 0.5 0, 0 0 1) for ease of comparison with the triclinic cell for “aniyunwiyaite”. It is seen that there is a very close match of the two sets of cell parameters, and a for “aniyunwiyaite” is double that reported for kingsmountite (Dunn et al., 1979), i.e. the “aniyunwiyaite” structure is a superstructure of the montgomeryite-type structure with a doubling of a. However, the monoclinic cell for kingsmountite was assumed and not verified by single-crystal studies (Dunn et al., 1979). To check this we obtained the holotype specimen of kingsmountite.

In contrast to the well-developed crystals of “aniyunwiyaite”, the type specimen of kingsmountite comprises hemispherules of poorly developed blades (Fig. 2b). Despite numerous attempts, we were not successful in locating crystals of type kingsmountite suitable for a single-crystal data collection. Dunn et al. (1979) had originally noted that even the smallest elongate fibres that they could extract from the spherules gave uninterpretable single-crystal patterns. We then collected a PXRD pattern on ground spherules of type kingsmountite and conducted Rietveld refinements (Rodriguez-Carvajal, 1990) using atomic coordinates obtained from single-crystal refinements of both “aniyunwiyaite” (triclinic P$1¯$⁠) and fanfaniite, (monoclinic, C2/c). In both refinements, the coordinates were fixed at the single-crystal values and the same profile parameters were refined. The triclinic model gave a much better fit to the PXRD, with R_wp = 6.3%, R_Bragg = 8.9%, than the monoclinic model, with R_wp = 9.1%, R_Bragg = 15.2%. The powder pattern fitted by the triclinic model is shown in Fig. 3.

As a further check on the crystallography of type kingsmountite, we applied electron diffraction (ED). A ground spherule was sonicated and the crystals dispersed on a carbon grid for examination in a transmission electron microscope. Single blades were readily located, lying flat on the grid and giving ED patterns corresponding to a*c*. The ED patterns showed weak superlattice reflections corresponding to a = 20 Å, consistent with the cell of “aniyunwiyaite”. An example ED pattern is shown in Fig. 4.

The results obtained for type kingsmountite are consistent with it having the crystal structure of “aniyunwiyaite”. This information, coupled with the near-identical nature of the chemistry and optical properties leaves no doubt that “aniyunwiyaite” is kingsmountite and “aniyunwiyaite” is discredited. The discreditation of “aniyunwiyaite” and redefinition of kingsmountite have been approved by the IMA CNMNC (proposal 19-B). The crystallographic details in the following section, which were obtained for “aniyunwiyaite” are then those for kingsmountite.

The PXRD data for kingsmountite were obtained on a Rigaku R-Axis Rapid II curved–imaging-plate microdiffractometer utilising monochromatised MoKα radiation. A Gandolfi-like motion on the φ and ω axes was used to randomize the sample. Observed d values and intensities were derived by profile fitting using JADE 2010 software (Materials Data Inc.). The unit-cell parameters refined from the powder data using JADE 2010 with whole-pattern fitting are a = 20.080(12), b = 13.188(12), c = 6.258(12) Å, α = 89.065(19), β = 91.285(13), γ = 112.20(2)° and V = 1534(3) Å³. The indexed PXRD pattern is given in Table S1 of the Supplementary Material, linked to this article and freely available at https://pubs.geoscienceworld.org/eurjmin/.

A single-crystal data collection was made at the macromolecular beam line MX2 of the Australian Synchrotron. Data were collected using monochromatic radiation with a wavelength of 0.7107 Å. The crystal was maintained at 100 K in an open-flow nitrogen cryostream. Further data collection details are given in Table 2.

A structural model was obtained in space group P$1¯$ using SHELXT (Sheldrick, 2015) and refined using JANA2006 (Petříček et al., 2014). It was found to conform broadly to the monoclinic montgomeryite-type structure (Moore & Araki, 1974), but in the lower symmetry structure there are four crystallographically independent Ca sites, while there are only two independent Ca sites in the monoclinic structure. However, it was also noted that one of the four Ca sites is six- rather than eight-coordinated and has bond-distances consistent with occupancy by Mn²⁺ (Mn4 in Fig. 5). In addition, the half-occupied edge-sharing chains of interlayer octahedra in the monoclinic structure were found to be replaced by isolated octahedra with almost full occupancy (A in Fig. 5) in the triclinic crystal structure. Site-occupancy refinements, coupled with difference-Fourier maps, showed that the A site was only 85% occupied with 15% occupancy of an adjacent, edge-shared octahedron (A′ in Table 3), and that the Mn4 site was 82% occupied with 18% occupation of an adjacent site displaced by 0.83 Å, which corresponds to a Ca site in montgomeryite. There was found to be the same (82/18%) partitioning of one of the H₂O molecules coordinated to Mn4 (Ow5/Ow5A in Table 3) to satisfy the coordination requirements of the adjacent Ca4. Minor (9%) partitioning of the Ca site Ca3 occurred into an adjacent site displaced by 0.86 Å. Mn was allocated to this site (Mn3 in Table 3). Although the electron microprobe (EMP) analyses showed a small (4%) deficiency in Al, site-occupancy refinements and bond valence sums (BVS, Gagné & Hawthorne, 2015) for the five independent Al sites gave no clear evidence for either vacancies or Fe³⁺ substitution for Al. The BVS for the A site indicated that this site was predominantly occupied by Fe²⁺. Minor Mg from the EMP analysis and minor Mn were also included in the A site and the Fe occupancy was refined. The BVS for the minority A′ site was consistent with it being occupied by Mn²⁺ (Table 3).

Refinement with anisotropic displacement parameters for all atoms except the partially occupied sites A′, Mn3, Ca4 and Ow5A converged at R_obs = 0.059 for 4351 reflections with I > 3σ(I). Further details of the refinement are given in Table 2. The refined coordinates, equivalent isotropic displacement parameters and BVS values are reported in Table 3. Selected interatomic distances are reported in Table 4. The H atoms were not unambiguously located in difference-Fourier maps, but BVS values in Table 3 show clearly the presence of four hydroxyl ions, Oh1–Oh4, and 12 independent H₂O molecules, Ow1–Ow12. The Crystallography Information File (CIF), including reflection data, is available online as part of the Supplementary Material.

A [001] projection of the structure is shown in Fig. 5. As for montgomeryite, it is built from (010) heteropolyhedral slabs of corner-connected octahedra and tetrahedra, with composition [Al₂(PO₄)₃(OH)₂]⁵⁻. Eight-coordinated Ca-centred polyhedra, Ca1O₆(H₂O)₂ and Ca2O₆(H₂O)₂, occupy cavities within the slabs, while eight-coordinated Ca3O₄(H₂O)₄, octahedrally coordinated AO₂(H₂O)₄ (A = dominant Fe²⁺) and octahedrally coordinated Mn4O₄(H₂O)₂ occupy sites on the surface of the slabs. The Ca3O₄(H₂O)₄ polyhedra share edges via coordinated H₂O (Ow11 and Ow12) to form chains parallel to [001]. The A-centred octahedra are isolated, while the Mn4O₄(H₂O)₂ octahedra occur in pairs, as for Ca3, but are too far apart for edge-sharing as occurs for the Ca3-centred polyhedra.

Kingsmountite is closely related, both compositionally and structurally, to the minerals calcioferrite, montgomeryite, zodacite and fanfaniite, which have the general formula Ca₄AB₄(PO₄)₆(OH)₄·12H₂O and C-centred monoclinic unit cells. If the unit cell for the C-centred monoclinic minerals is reduced to a primitive cell, the cell parameters are a ~ 10.0, b ~ 13.1, c ~ 6.2 Å, α ~ 89.5, β ~ 91.5, γ ~ 112.5°. A comparison with the cell parameters for triclinic kingsmountite shows that the latter has a doubled a cell parameter relative to the montgomeryite-type cell, with the other parameters almost the same for both minerals.

The triclinic kingsmountite superstructure of the monoclinic Ca₄AB₄(PO₄)₆(OH)₄·12H₂O structures results both from ordering of Mn in one of the Ca sites, and from ordering of the A cations. As shown in Fig. 5, along the lengths of the heteropolyhedral slabs, isolated A-centred octahedra alternate in position from one side of the slabs to the other, whereas in the C2/c structure (Fig. 1), edge-shared chains of half-occupied A-centred octahedra span the interlayer region between the slabs. The partially occupied sites A′, Mn3 and Ca4 in kingsmountite represent local disordered regions of the C2/c-type structure.

The minerals calcioferrite, montgomeryite, kingsmountite, zodacite and fanfaniite meet the requirements for formation of a mineral group according to the definition given by Mills et al. (2009) that: A mineral group consists of two or more minerals with the same or essentially the same structure, and composed of chemically similar elements. Based on the recommendation by Mills et al. (2009) that the group name be that of the first mineral to be adequately characterized, the group name is calcioferrite (Blum, 1858). The general formula for the group members except for kingsmountite is Ca₄AB₄(PO₄)₆(OH)₄·12H₂O. In kingsmountite, Ca₄ is replaced by Ca₃Mn. The group is subdivided into two subgroups based on the dominant trivalent B cation. The calcioferrite subgroup has B = Fe³⁺ and the montgomeryite subgroup has B = Al³⁺. Within each subgroup the different root-name minerals are distinguished by the dominant divalent A cation.

This study was funded by the John Jago Trelawney Endowment to the Mineral Sciences Department of the Natural History Museum of Los Angeles County. The research was undertaken in part using the MX2 beamline at the Australian Synchrotron, part of ANSTO, and made use of the Australian Cancer Research Foundation (ACRF) detector. The authors acknowledge the use of facilities within the Monash Centre for Electron Microscopy. We acknowledge also the Diffraction Laboratory at CSIRO Mineral Resources, Clayton, for use of their Panalytical Empyrean powder XRD diffractometer. Thanks to Cameron Davidson for EMP sample preparation.