Occurrence

The new mineral was found at the Torrecillas mine, Salar Grande, Iquique Province, Tarapacá Region, Chile (20°58′36″S, 70°08′31″W). Torrecillas Hill, on which the Torrecillas mine is located, is composed of four different rock units. The Coastal Range Batholith (mainly gabbros) extends from the seashore to the Pan-American Road along the base of Torrecillas Hill. At the foot of Torrecillas Hill is a small area of contact metamorphic rocks in which garnet crystals occur in metamorphosed shales. Higher on the hill, the rocks are predominantly porphyritic andesitic lavas of the Jurassic La Negra Formation (García, Reference García1967; Buchelt and Tellez, Reference Buchelt, Tellez, Bahlburg, Breitkreuz and Giese1988). The Torrecillas deposit, in which the new minerals were found, consists of two main veins rich in secondary arsenic and copper minerals that intersect metamorphosed marine shales and lavas. These mineralised veins are genetically related to the aforementioned porphyritic andesitic lavas of the Jurassic La Negra Formation. More information on the geology and mineralogy of the area is provided by Gutiérrez (Reference Gutiérrez1975).

The rare secondary chlorides, arsenates and arsenites (and associated sulfates) have been found at three main sites on the hill: an upper pit measuring ~8 m long and 3 m deep, a lower pit ~100 m from the upper pit and measuring ~5 m long and 3 m deep, and a mine shaft adjacent to the lower pit and lower on the hill. Naalasite was found in January 2016 by one of the authors (GM) in the lower pit. The new mineral is a secondary alteration phase occurring in association with anhydrite, juansilvaite, magnesiokoritnigite and a lavendulan-like phase. The new mineral is very rare, having been found only on a few specimens.

The secondary assemblages at the Torrecillas deposit are interpreted as principally having formed from the oxidation of native arsenic and other As-bearing primary phases, followed by later alteration by saline fluids derived from evaporating meteoric water under hyperarid conditions (Cameron et al., Reference Cameron, Leybourne and Palacios2007); however, considering the proximity of the Torrecillas deposit to the Pacific Ocean, it seems possible that the frequent dense coastal camanchaca fogs have also played a role in the alteration of the veins and the formation of the secondary minerals, particularly in the recent past, since the exhumation of the deposit well above sea level on Torrecillas Hill.

Physical and optical properties

Naalasite occurs in tightly intergrown aggregates and druses of equant crystals, up to 0.15 mm in diameter (Fig. 1). While it was not possible to measure crystal forms, the Bravais–Friedel–Donnay–Harker principle (Donnay and Harker, Reference Donnay and Harker1937) predicts that the {101} and {001} are most prominent. Equal development of these forms is consistent with the crystal morphology observed visually (Fig. 2). Crystals are light pink and transparent, with vitreous lustre and white streak. The mineral does not fluoresce in long- or short-wave ultraviolet light. The Mohs hardness is ~3½ based on scratch tests. Crystals are brittle with irregular fracture and no cleavage. The density measured by flotation in a mixture of methylene iodide and toluene is 3.19(2) g⋅cm^–3. The calculated density is 3.235 g⋅cm^–3 for the empirical formula and the unit-cell refined from powder X-ray diffraction (PXRD) data; 3.162 for the ideal formula and single-crystal cell. Note that the single-crystal cell was determined on a crystal close in composition to the Al end-member, whereas the PXRD was measured on material closer in composition to the crystals analysed by electron microprobe. The mineral dissolves slowly in dilute HCl at room temperature. Optically, naalasite is uniaxial (+) with ω = 1.630(3) and ɛ = 1.660(3) measured in white light. No pleochroism was observed.

Figure 1. Pink naalasite crystals with colourless crystals of anhydrite on holotype specimen 76282; the field of view is 0.68 mm across.

Figure 2. Crystal drawing of naalasite; clinographic projection. The crystal drawing was made using SHAPE (v7.4) (Shape Software, Kingsport, Tennessee, USA).

Raman spectroscopy

Raman spectroscopy was done on a Horiba XploRa PLUS micro-Raman spectrometer using an incident wavelength of 532 nm, laser slit of 100 μm, 1800 gr/mm diffraction grating and a 100× (0.9 NA) objective. The spectrum recorded from 4000 to 60 cm^–1 is compared to that of nafeasite in Fig. 3. The features, in the 3600–2800 cm^–1 region, are related to O–H stretching. The bands between ~1000 and ~700 cm^–1 are probably due to the ν₃ and ν₁ stretching modes of AsO₃OH groups. The bands between ~550 and ~350 cm^–1 are probably due to ν₄ and ν₂ bending modes of AsO₃OH groups. The bands at lower wavenumbers are attributable to various lattice vibrations.

Figure 3. Raman spectra of naalasite and nafeasite recorded with a 532 nm laser.

Composition

Analyses (six points on two crystals) were performed at Caltech on a JEOL JXA-iHP200F field-emission electron microprobe in WDS mode. Analytical conditions were 15 kV accelerating voltage, 10 nA beam current and 10 μm beam diameter. A time-dependent intensity correction was applied to Na. Insufficient material is available for the determination of H₂O, so it is calculated based on the structure. Analytical data are given in Table 1.

Table 1. Analytical data (in wt.%) for naalasite.

* Based on the structure; S.D. – standard deviation.

The empirical formula (based on 2 As and 9 O atoms per formula unit) is Na_0.92Al_0.61Fe³⁺_0.39As₂O₉H_4.07. The simplified structural formula is Na(Al,Fe³⁺)(AsO₃OH)₂⋅H₂O. The idealised formula is NaAl(AsO₃OH)₂⋅H₂O, which requires Na₂O 8.91, Al₂O₃ 14.66, As₂O₅ 66.08, H₂O 10.36, total 100 wt.%. The Gladstone–Dale compatibility (Mandarino, Reference Mandarino2007) 1 – (K_p/K_c) is 0.011 (superior) for the empirical formula using the powder cell and –0.068 (fair) for the ideal (Al end-member) formula using the single-crystal cell. Presumably, the optical properties were determined on a crystal closer in composition to those analysed by electron microprobe than to that on which the structure was determined.

X-ray crystallography and structure refinement

Powder X-ray diffraction data for the holotype were recorded using a Rigaku R-Axis Rapid II curved imaging plate microdiffractometer with monochromatised MoKα radiation. A Gandolfi-like motion on the φ and ω axes was used to randomise the sample. Observed d-values and intensities were derived by profile fitting using JADE Pro software (Materials Data, Inc.). The powder data are presented in Supplementary Table S1. This table is deposited with the Principal Editors of Mineralogical Magazine and is available as Supplementary material (see below). The unit-cell parameters refined from the powder data using JADE Pro with whole pattern fitting are a = 8.494(4), c = 26.430(13) Å and V = 1651.5(4) ų.

Single-crystal X-ray studies were done on the same diffractometer and radiation noted above. The occurrence of naalasite crystals in tightly intergrown clusters made the selection of a crystal fragment suitable for structure study difficult. Even the best fragment exhibited significant mosaicity; however, it yielded a reasonable structure solution and refinement.

The Rigaku CrystalClear software package was used for processing the structure data, including the application of an empirical absorption correction using the multi-scan method with ABSCOR (Higashi, Reference Higashi2001). The structure was solved using the intrinsic-phasing algorithm of SHELXT (Sheldrick, Reference Sheldrick2015a). SHELXL-2016 (Sheldrick, Reference Sheldrick2015b) was used for the refinement of the structure. Considering the similarity of the naalasite and nafeasite PXRDs (Fig. 4), it was anticipated that, like nafeasite, naalasite would be monoclinic with space group C2. However, the unit cell of naalasite proved to be metrically hexagonal and its structure was successfully solved in the trigonal space group R32. Indeed, the structures of naalasite and nafeasite are based on virtually identical frameworks of corner-linked alternating octahedra and tetrahedra (Fig. 5) and their cell directions and lengths can be related as follows: [100]_Nfa = [–⅓ +⅓ –⅔]_Naa; [010]_Nfa = [100]_Naa; [001]_Nfa = [210]_Naa; however, the cells cannot be rigorously transformed into one another, as evidenced by the fact that there is not a simple relationship between their volumes. That also explains why efforts to obtain structure solutions for naalasite in C2 and nafeasite in R32 were both unsuccessful.

Figure 4. Powder X-ray diffraction patterns of naalasite and nafeasite. The lines calculated from the structures are shown as red sticks.

Figure 5. 3D frameworks of corner-sharing octahedra and tetrahedra in the structures of naalasite and nafeasite. Note that both structures are noncentrosymmetric and the original atom coordinates reported for nafeasite (Kampf et al., Reference Kampf, Schluter, Malcherek, Paulenz, Pohl, Ma, Dini and Molina Donoso2022) have been inverted to correspond to the enantiomorph matching that of naalasite. The unit cell outlines are shown with dashed lines. The structure drawings were made using ATOMS (v6.5) (Shape Software, Kingsport, Tennessee, USA).

All framework atoms in the naalasite structure were refined anisotropically at full occupancies. Although the chemical analyses indicated significant substitution of Fe³⁺ for Al, the structure refinement indicated the octahedrally-coordinated cation sites (Al1 and Al2) to be essentially fully occupied by only Al. A scanning electron microscope survey using energy-dispersive spectroscopy of naalasite crystals confirmed that some crystals are near the pure Al end-member.

Three non-framework sites were identified. Two were assigned initially as partially occupied Na sites (Na1 and Na2) and one as a fully occupied H₂O site (OW). The Na1 site is split across a (x,0,0) special position at an Na1–Na1 distance of 0.75 Å. The coordination geometry of the Na1 site and its proximity to other Na1 sites surrounding a three-fold axis (Fig. 6) led us to assign the site joint occupancy by Na and O (OW1), such that an O (H₂O group) occupying the Na1 site is coordinated to an Na occupying another Na1 site. The distances (all < 2.97 Å) between the individual components of the split Na1/OW1 site (with 6-fold multiplicity) dictate that the site can at most accommodate a total of 1 Na atom (an Na atom could reside at only one of the six at any given time), so the Na site occupancy is limited to at most 0.1667. Combining that with the occupancy of the Na2 site yields 0.91 Na apfu, agreeing almost exactly with the empirical formula. The site can also accommodate up to 2 O (H₂O) atoms (which would be at distances of 2.82 to 2.96 Å from the Na and from each other) for a maximum O occupancy of ⅓; however, without a water determination, there is no reason to assume that the O occupancy is not less than ⅓. If we set the Na occupancy at the Na1/OW1 site at ⅙ and refine the O occupancy, the resulting site occupancy is Na_0.1667O_0.23(2), which, when combined with the fully occupied OW site that is coordinated to Na2, provides 1.13 H₂O pfu.

Figure 6. Coordination of disordered Na1/OW1 site viewed down the three-fold c axis.

Difference-Fourier failed to locate possible H-atom sites. Because of uncertainties in the hydrogen bonding and the ‘flexibility’ in the Na–O distances indicated by the prolate ellipsoid for the Na1/OW1 site, a rigorous analysis of the bond-valence sums for the channel was not attempted. The data collection and refinement details are given in Table 2, atom coordinates and displacement parameters in Table 3, selected bond distances in Table 4 and a bond-valence analysis in Table 5. The complete structure is shown in Fig. 7. The crystallographic information file has been deposited with the Principal Editor of Mineralogical Magazine and is available as Supplementary material (see below)

Table 2. Data collection and structure refinement details for naalasite.

R _int = Σ|F _o²–F _o²(mean)|/Σ[F _o²]. GoF = S = {Σ[w(F _o²–F _c²)²]/(n–p)}^1/2. R ₁ = Σ||F _o|–|F _c||/Σ|F _o|. wR ₂ = {Σ[w(F _o²–F _c²)²]/Σ[w(F _o²)²]}^1/2; w = 1/[σ²(F _o²)+(aP)²+bP] where a is 0.0292, b is 9.6773 and P is [2F _c²+Max(F _o²,0)]/3.

Table 3. Atom coordinates and displacement parameters (Ų) for naalasite.

Table 4. Selected bond distances (Å) for naalasite.

Table 5. Bond-valence analysis for naalasite. Values are expressed in valence units (vu).

Bond valences are based on refined site occupancies. Bond-valence parameters are from Gagné and Hawthorne (Reference Gagné and Hawthorne2015). Hydrogen-bond valence is based on the OH⋅⋅⋅O2 bond length from Ferraris and Ivaldi (Reference Ferraris and Ivaldi1988); the negative value indicates donated bond valence.

Figure 7. The structure of naalasite viewed along [100]. The unit cell outlines are shown with dashed lines.

Description of the structure

The structures of naalasite (space group R32) and nafeasite (space group C2) are based on topologically equivalent loose 3D frameworks of alternating AsO₃OH tetrahedra and (Al,Fe³⁺)O₆ octahedra (Fig. 5). Each of the octahedra shares all six of its vertices with AsO₃OH tetrahedra and each AsO₃OH tetrahedron shares its three O vertices with an octahedron, the remaining OH vertex being unshared. The resulting frameworks consist of layers of mutually unconnected AsO₃OH-tetrahedra and layers of mutually unconnected octahedra, alternating along [001] in naalasite and [100] in nafeasite. Relatively large interstices within the frameworks accommodate partially occupied Na and H₂O sites. Hydrogen bonding clearly plays a significant role in the stabilities of both structures; however, the complexity of the disordered interstitial portion of the structures made it impossible to devise a comprehensive and unambiguous hydrogen-bonding scheme.

Arsenate mineral structures based on simple frameworks of alternating octahedra and tetrahedra are uncommon. The only ones noted by Majzlan et al. (Reference Majzlan, Drahota, Filippi, Bowell, Alpers, Jamieson, Kirk Nordstrom and Majzlan2014) were parascorodite, Fe(H₂O)₂AsO₄; the isostructural minerals scorodite, Fe(H₂O)₂AsO₄, mansfieldite, Al(H₂O)₂AsO₄, and yanomamite, In(H₂O)₂AsO₄; and berzeliite, (NaCa₂)Mg₂(AsO₄)₃, and manganberzeliite, (NaCa₂)Mn₂(AsO₄)₃, which have the garnet structure. None of these structures is very similar to those of naalasite and nafeasite; however, the naalasite and nafeasite frameworks are topologically equivalent to those of synthetic RbAl(AsO₃OH)₂ and CsFe³⁺(AsO₃OH)₂, which, like naalasite, crystallise in space group R32 (Schwendtner and Kolitsch, Reference Schwendtner and Kolitsch2018).