- © 2014 The Mineralogical Society
The new mineral oskarssonite (IMA2012-088), with ideal formula AlF3, was found in August 2009 at the surface of fumaroles on the Eldfell volcano, Heimaey Island, Iceland (GPS coordinates 63°25′58.9″N 20°14′50.3″W). It occurs as sub-micron-sized crystals forming a white powder in association with anhydrite, bassanite, gypsum, jarosite, anatase, hematite, opal, ralstonite, jakobssonite and meniaylovite. Chemical analyses by energy-dispersive spectrometry with a scanning electron-microscope produced the following mean elemental composition: Al, 31.70; F, 58.41; O, 9.22; total 99.33 wt.%. The empirical chemical formula is AlF2.6(OH)0.5 which suggests partial substitution of F by OH. Oskarssonite is rhombohedral, space group R3İc, with ah = 4.9817(4) Å, c = 12.387(1) Å, Vuc = 266.23(5) Å3, Z = 6. The five strongest lines in the powder diffraction diagram [d in Å (I) (hkl)] are as follows: 3.54 (100) (012), 2.131 (13) (113), 1.771 (20) (024), 1.59 (15) (116), 1.574 (10) (122). Rietveld refinement confirms the identity of oskarssonite with the synthetic rhombohedral form of AlF3. Its structure can be described as a rhombohedral deformation of the idealized cubic perovskite-type octahedral framework of corner-sharing AlF6 groups. Oskarssonite appears in the surface part of the fumaroles where fluorides are abundant. At greater depths (below 10 cm) sulfates dominate among the fumarolic minerals. In accordance with its occurrence, we surmise that oskarssonite forms in the later stages of the fumarolic activity in an environment poor in alkalies and Mg. Ralstonite (NaxMgxAl1–xF3(H2O)y), which, unlike oskarssonite, contains Na and Mg as important constituents, dominated in the first-formed fumaroles, but now, 41 years after the eruption of Eldfell, is only a minor phase. The new mineral is named after the Icelandic volcanologist Niels Oskarsson.
In the summer of 2009 our team undertook an expedition to the Eldfell volcano on the Heimaey Island in the Vestmannaeyjar archipelago, Iceland, in order to investigate the fumaroles, which at that time had been active for 36 years after the eruption of the volcano. A previous investigation in the years closely following the eruption revealed rich mineralogy with several new minerals in fumaroles developed on hawaiite scoria host rock (Jakobsson et al., 2008). The descriptions of three of them have already been published (Balić-Žunić et al., 2009, 2012; Mitolo et al., 2013) and more are under preparation. Our purpose was to record the long-term development of the fumaroles. Although the extent of the fumarolic field and the peak temperatures decreased with time, the existing fumaroles proved to be still active, with temperatures at a depth of ~50–60 cm exceeding 400°C. Surprisingly, we observed that the relative abundances of the previously recorded minerals changed significantly and even noted some new ones. This work describes one of those. It is the low-temperature form of AlF3 which so far has not been fully characterized from natural occurrences. To the best of our knowledge, the only records of its probable occurrence are from Hekla after the 1970 eruption (Oskarsson, 1981) and from Mont Erebus, Antarctica (Rosenberg, 1988).
The new mineral and its name have been approved by the IMA Commission on New Minerals Nomenclature and Classification (CNMNC) (No. 2012-088). The mineral is named after Niels Oskarsson (1944–present), an Icelandic volcanologist, in acknowledgement of his pioneering work on Icelandic fumaroles.
The holotype is kept in the mineral collection of the Icelandic Institute of Natural History, Gardabaer, Iceland, under the sample number NI 24489. The sample is temporarily on loan at the Natural History Museum of Denmark in Copenhagen due to on-going analysis on several other new minerals. Other samples, in which oskarssonite has been identified, are kept in the mineral collection of the Natural History Museum of Denmark and have the following numbers: 2012.112, 2012.113 and 2012.114.
Mineral paragenesis and physical properties
The samples with the new mineral were collected on August 7, 2009 at GPS coordinates 63°25′58.9″N 20°14′50.3″W. Oskarssonite occurs as a primary phase in recently deposited encrustations at a measured temperature of 90°C. It forms a white to greyish-yellow fine powder close to the surface of fumarole vents exhausting through hawaiite scoria, mixed with anhydrite, bassanite, gypsum, jarosite, anatase, hematite, opal, ralstonite, jakobssonite, meniaylovite, and four other fluorides which also appear to be new minerals and are presently being investigated by our group. Macroscopically, it appears as a white powder (Fig. 1). Investigation by scanning electron microscopy (SEM) reveals massive aggregates of sub-micron crystals (Fig. 2). Oskarssonite and other fluoride minerals have been collected only from the upper 10 cm at this locality. At greater depths, a completely different association of fumarolic minerals occurs, dominated by sulfates.
Due to the minute size of the crystals, it was not possible to quantify the basic mineral properties like hardness, indices of refraction, etc. Synthetic AlF3 is uniaxial positive with no = 1.3765 and nE = 1.3770 (5893 Å) (Staritzky and Asprey, 1957). The mean index of refraction is reported to increase with the OH content (Rosenberg, 2006). The density, calculated from the X-ray diffraction (XRD) data and the empirical chemical formula is 3.184 gcm−3.
Five point chemical analyses were obtained by SEM energy-dispersive spectroscopy. The instrument used was an S 360 Cambridge SEM coupled with an Oxford-Link Ge ISIS energy-dispersive spectrometer equipped with a Super Atmosphere thin window as this allows better detection of light elements. An energy-dispersive spectrometer was chosen for quantitative analyses of the material in preference to a wavelength-dispersive spectrometer, due to the better specimen stability provided by lower beam currents. The investigated sample consisted of an aggregate of submicron crystals (Fig. 2). As we had to measure uneven surfaces, a “non-critical” working distance was utilized (Ruste, 1979; Acquafredda and Paglionico, 2004). The sample was sputtered with a 30 nm thick carbon film before analysis. The operating conditions were 15 kV accelerating potential, 500 pA probe current, 3 μm beam diameter, 2500 counts per second as the average count rate across the whole spectrum and typical counting times of 100 s. The X-ray intensities were converted to wt.% elements by ZAF4/FLS quantitative analysis software provided by Oxford-Link Analytical (UK). The procedure allows the sample peak heights and areas to be compared with standards, and it takes the relative contribution of the matrix and partial or complete overlaps of peaks produced by different elements into account. Analytical data are given in Table 1. All the measured amounts of O have been attributed to OH.
The empirical formula (based on Al = 1 a.p.f.u.) is: AlF2.62(OH)0.49. The ideal pure fluoride, AlF3, requires: Al = 32.13 wt.%, F = 67.87 wt.%.
In a study of the system AlF3–Al2O3–H2O–HF (Rosenberg, 2006) Al(F,OH)3, corresponding to oskarssonite, was synthesized as one of the phases at 450–700°C and 2 kbar. It showed compositions reaching AlF2.45(OH)0.55. This is quite close to the empirical composition obtained in the present investigation. Unfortunately, Rosenberg has chosen an unusual way of reporting the XRD data for this phase: as a difference in angle for the 024 reflection (in Rosenberg's notation 220, because he uses the rhombohedral assignment of crystal axes) and a non-specified reflection of the standard used. It is therefore not possible for us to compare our diffraction data with his to check which hydration grade they suggest. It also seems that his and our data are contradictory, because we obtained a larger d024 than observed for pure AlF3, whereas Rosenberg reports decreasing values with increasing hydration. In any case, the conclusion from the latter work is that the hydration grade increases with decreasing temperature in the AlF3–Al2O3–SiO2–H2O system, whereas it increases with increasing temperature in the AlF3–Al2O3–H2O–HF system. The reason for the former behaviour is that HF, released through hydration, is consumed by the formation of gaseous SiF4 (Rosenberg, 1992). As the fumarolic system at Eldfell hosts abundant SiO2, we can expect similar conditions, and the occurrence of oskarssonite at the coldest upper part of fumaroles suggests the maximum possible hydration, again in accordance with the chemical analysis.
Diffraction properties and crystal structure
X-ray powder diffraction data (Table 2, Fig. 3) were collected on a Bruker-AXS D8 diffractometer in Bragg-Brentano geometry with a primary-beam Ge(111) monochromator, using CuKα1 radiation (λ = 1.5406 Å) and a fixed divergence slit. The diagram matches the data of synthetic rhombohedral AlF3 (Hoppe and Kissel, 1984) well. The purest powder sample, also used for the Rietveld analysis, contained small amounts of opal, bassanite, anatase and hematite. The former three are admixed with oskarssonite as fine powders, whereas the latter occurs as relatively large black metallic crystals up to mm size dispersed through the sample. For the data in Table 2, the positions and intensities of individual diffraction maxima were determined using the Bruker-AXS program EVA. Attribution of the maxima and indexing were done in accordance with the results of the Rietveld refinement (program Topas 4.1, Bruker-AXS product). Unit-cell data are as follows: crystal system – rhombohedral (hexagonal setting); space group – R3İc, ah = 4.9817(4) Å, c = 12.387(1) Å, V = 266.23(5) Å3, Z = 6.
Rietveld refinement data are given in Table 3. Refinement confirms the identity of oskarssonite with the rhombohedral form of AlF3 (Daniel et al., 1990). The structure can be described as a rhombohedral deformation of the idealized cubic perovskite-type octahedral framework of corner-sharing AlF6 groups. The Al coordination octahedra are very regular and the deformation may be described as their rigid-body rotations around one of the <111> directions of the idealized cubic structure, decreasing the Al–F–Al bond angles from 180° to around 160°. The unit cell contains only two independent sites, Al (Wyckoff position 6b, coordinates 0, 0, 0) and F (Wyckoff position 18e, coordinates x, 0, 1/4). Results of the oskarssonite chemical analyses suggest partial substitution of OH for F in the structure. It has larger ah, smaller c, and a larger unit-cell volume than the pure synthetic AlF3 (Table 4) but the difference in bond length Al–F(OH) is marginal, being 1.797(3) Å in synthetic AlF3 (Daniel et al., 1990) compared to 1.791(1) Å in oskarssonite.
Relation to other minerals
Oskarssonite corresponds to the stable low-temperature form of AlF3. At 450°C it transforms reversibly to the cubic ReO3 structure type (Daniel et al., 1990). This is a displacive phase transition where both phases contain a perovskite-like framework of octahedra which in the low-temperature phase is rhombohedrally distorted from the ideal cubic arrangement. Through rotation of octahedra around one of the <111> sets of cubic directions, it becomes the c axis in the hexagonal setting of oskarssonite. Several other metastable synthetic forms of AlF3 have been reported and structurally confirmed. The orthorhombic Cmcm polymorph with a slightly distorted hexagonal tungsten bronze (HTB) structure type (Le Bail et al., 1988); the tetragonal P4/nmm modification with a complex octahedral framework (Le Bail et al., 1992; Herron et al., 1995); plus a cubic Fd3m form with a pyrochlore-type framework of octahedra (Herron et al., 1995) are known. Furthermore, a possible tetragonal P4/mbm polymorph was discovered by Herron et al. (1995), having a tetragonal tungsten bronze (TTB) framework (Le Bail and Calvayrac, 2006) but still lacking a completely satisfactory structure confirmation. All metastable synthetic forms crystallize under specific conditions and from specific precursors, transforming irreversibly to the high- or low-temperature polymorphs with perovskite-related octahedral frameworks on heating (Herron et al., 1995). It is therefore not very likely that they will be found as natural products. Also, due to the fast reversible transition, it is not expected that the stable high-temperature cubic form will be found as a mineral and therefore it seems that oskarssonite is the only probable mineral form of AlF3, at least under a low-pressure regime.
The relationship between oskarssonite and ralstonite is quite interesting. The crystal structure of ralstonite is based on the pyrochlore-type framework of AlF6 octahedra where some Al is substituted by Mg, the valence-balance being compensated by additional Na atoms in the framework cavities and additional H2O molecules that coordinate Na (Effenberger and Kluger, 1984). Thus, the ralstonite octahedral framework seems to be stabilized by other cations. Although Al(OH,F)3 phases and even practically pure AlF3 with the latter framework have been synthesized (Herron et al., 1995), the pure unsubstituted form seems to be metastable (see above). One can expect that ralstonite forms readily in fumarolic environments rich in additional elements (Mg, Na) and possibly also at a higher H2O activity, whereas the formation of oskarssonite needs a different chemistry with Al dominant over other possible constituents of ralstonite. This is supported by the geological evidence at Eldfell where ralstonite is dominant in the initial fumarolic stage and in the “fresh fumaroles” (Jakobsson et al., 2008). During recent sampling of old fumaroles ralstonite occurred as a minor fluoride phase whereas a conspicuous concentration of oskarssonite was found at the very top of the fumaroles. The presence of opal, cristobalite, anatase and corundum in present-day fumaroles also suggests a ‘simple’ chemistry, represented by phases which can be regarded as final products after leaching of the majority of Na, K, Mg, Ca and Fe in the host rock.
Among the known hydrated phases of AlF3 there is so far only one accepted mineral, rosenbergite, AlF2.5(OH)0.5(H2O)3. However, another hydrated form with ralstonite structural characteristics has been reported in a number of papers (Rosenberg, 2006 and references therein) but so far not approved as a valid mineral species. Its composition is given as AlF3–y(OH)y(H2O)n with y varying between 1.3 and 1.9 and n between 0.9 and 1 according to the synthesis study of Rosenberg (2006). The latter phase in association with ralstonite and jakobssonite has also been indicated by an XRD diagram of a fumarolic sample formed after the Hekla 1992 eruption. This suspected new mineral was labelled HS by Jakobsson et al. (2008), and was later given the CNMNC number UM2008-II-? (Smith and Nickel, 2007). It has so far not been observed in samples collected on Eldfell. Rosenbergite, however, is observed in the samples collected on Eldfell in 2009 at the very top of the fumaroles (determined by XRD). It forms rare, small stalagmitic aggregates, which suggests it forms due to a reaction with atmospheric humidity.
The authors thank Dora Balić Žunić for help in collecting the samples and Helene Almind for help in conducting X-ray diffraction measurements. This work has been supported financially by the Danish Agency for Science, Technology and Innovation. The suggestions for improving the article by the editor Peter Williams, Thomas Armbruster and the other, anonymous, reviewer are much appreciated.
- Manuscript received 19 November 2013.
- Manuscript Accepted for publication 19 January 2014.