Dellagiustaite: A Novel Natural Spinel Containing V 2+

: Dellagiustaite, ideally Al 2 V 2+ O 4 , is a new spinel-group mineral from Sierra de Comechingones, San Luis, Argentina, where it is found associated with hibonite (containing tubular inclusions, 5–100 µ m, of metallic vanadium), grossite, and two other unknown phases with ideal stoichiometry of Ca 2 Al 3 O 6 F and Ca 2 Al 2 SiO 7 . A very similar rock containing dellagiustaite has been found at Mt Carmel (northern Israel), where super-reduced mineral assemblages have crystallized from high- T melts trapped in corundum aggregates (micro-xenoliths) within picritic-tholeiitic lavas ejected from Cretaceous volcanoes. In the holotype, euhedral grains of dellagiustaite are found as inclusions in grossite. The empirical average chemical formula of dellagiustaite is (Al 1.09 V 2 + 0.91 V 3 + 0.87 Mg 0.08 Ti 3 + 0.04 Mn 0.01 ) Σ 3 O 4 , but it may show limited replacement of V 2+ by Mg and of V 3+ by Al. As Al is the dominant trivalent cation, the ideal formula is Al 2 V 2+ O 4 according to the current IMA rules. Dellagiustaite shows the usual space group of spinel-group minerals ( Fd 3 m , R 1 = 1.46%) with a = 8.1950(1) Å. The observed mean bond lengths < T –O> = 1.782(2) Å and < M –O> = 2.0445(9) Å, the observed site scattering ( T = 13.3 eps, M = 22.5 eps), and the chemical composition show that dellagiustaite is an inverse spinel: T tetrahedra are occupied by Al 3+ , whereas M octahedra are occupied by V 2+ and V 3+ , leading to the site assignment as T Al M (V 2 + 0.91 V 3 + 0.88 Al 3 + 0.09 Mg 0.08 Ti 3 + 0.03 Mn 0.01 )O 4 .


Introduction
A new classification has been introduced recently for the spinel supergroup [1], based on chemical information alone. It has two criteria: the cation to anion ratio (3:4) represented by the general formula AB 2 X 4 , the dominant charge and the dominant constituent (A or B). From a structural point of view, minerals of the spinel supergroup show a structure consisting on a heteropolyhedral framework of 4-fold and 6-fold polyhedra (T-sites and M-sites, respectively), the latter sharing half of the edges, (Yacanto Group), with intercalations of mafic and ultramafic rocks (San Miguel Group), and garnet gneisses, marbles and kinzigites at granulite facies [13,14]. However, the metamorphic grade is not high enough for the formation of the peculiar mineral assemblage in which dellagiustaite is found. In fact, dellagiustaite is associated with hibonite (CaAl 12 O 19 ) and grossite (CaAl 4 O 7 ), which constitute the modal composition of the rock. Hibonite crystals are remarkable, and it is common to find cm-size purple euhedral phenocrysts (Figure 1).
Grossite occurs as interstitial light violet crystals up to a few millimeters across. Other phases present are gehlenite, aluminum-rich perovskite and a new calcium aluminate with ideal formula (Ca 2 Al 3 FO 6 ) corresponding to the synthetic phase calfidine [15,16]. Hibonite and grossite crystals frequently have tubular inclusions of a V-rich phase that is isostructural with a non-stoichiometric vanadium oxide with formula approximately of V 2 O. These inclusions are related to pores running in the same direction and containing hexagonal plates of medium-crystalline graphite, which will be described in future publications. Dellagiustaite occurs in fractured hibonite and grossite crystals (Figure 2a     At Mt Carmel, dellagiustaite is associated with V-rich hibonite, grossite, krotite, Ca 2 Al 3 O 6 F and fluorite, as well as an unidentified phase with stoichiometry close to a (K,V,Mg,Na)-rich Ca-depleted hibonite. The remarkable similarity of the assemblages from the two localities is also confirmed by the presence of metallic V and V alloys as spherules, rods and dense branching structures in hibonite and grossite. Figure 5 reports the textural relations between the associated minerals at Mt Carmel. The rocks of Mt Carmel are also notable due to the significant amount of a V-Al alloy. This is the first reported in situ terrestrial Al-alloy, which probably is representative of the most reducing conditions yet documented on Earth.

Mineral Description and Physical Properties
Dellagiustaite exhibits idiomorphic morphology (Figure 2a) with frequent {111} forms; the crystals (up to 200 µm) are euhedral and/or subhedral and commonly overgrow tubular inclusions of a non-stoichiometric vanadium oxide.
Dellagiustaite crystals are black, opaque, show metallic luster and black streak, and are not fluorescent. Mohs hardness is ca 6.5-7, analogous to spinel. Tenacity is brittle and cleavage is not observed. Parting is irregular and conchoidal and fracture is uneven and splintery. Density could not be determined due to the frequent presence of vanadium inclusions in the crystals. Calculated density is 4.6 g·cm −3 (from ideal formula and unit-cell parameters from X-ray single-crystal data).
In reflected light, the mineral is light grey and isotropic. In the holotype, no evidence of growth zonation is observed. Reflectance measurements were performed in air using an MPM-200 Zeiss microphotometer equipped with an MSP-20 system processor on a Zeiss Axioplan ore microscope. Filament temperature was approximately 3350 K. An interference filter was adjusted, in turn, to select four wavelengths for measurement (471.1, 548.3, 586.6, and 652.3 nm). Readings were taken for specimen and standard (SiC) maintained under the same focus conditions. The diameter of the circular measured area was 0.07 mm. Measurements of reflectance (in %) are 14.1 (471.1 nm), 13.8 (548.3 nm), 13.6 (586.6 nm), and 13.7 (652.3 nm).

Chemical Data
The chemical composition of dellagiustaite was determined using a JEOL 8200 Super Probe electron microprobe analyzer (EMPA) operating in X-ray wavelength-dispersive mode (WDS-EMPA) at the Department of Earth Sciences, University of Milan, Milan, Italy. Operating conditions included an accelerating voltage of 15 kV, a beam current of 5 nA, and a beam diameter of 1 µm, peak counting-time 30 s and 10 s in both peak sides background. Corrections of the raw data were made with the Φ Z procedures [18]. Spectral lines, analyzing crystals and standards used were:  Table 1. Elements sought but not detected in dellagiustaite were Si, Fe and Cr. VO/V 2 O 3 ratio was calculated to obtain 3 atoms per formula unit (apfu). Ti was considered to be all as Ti 3+ because the presence of V 2+ implies that all the Ti must be trivalent at reduced conditions. Associated minerals were also analyzed and the results of representative analyses are reported in Table 2.
Samples from Mt Carmel were analyzed at the CCFS/GEMOC, Earth and Planetary Sciences, Macquarie University, Sydney, Australia with a Zeiss EVO MA15 scanning electron microscope operating in X-ray energy-dispersive mode (EDS) using an accelerating voltage of 15 keV and a beam current of 1 nA. The SEM-EDS employs a standardless analysis technique, with peak positions recalibrated when necessary by analysis of a copper grid. The results are closely comparable to WDS-EMPA analysis for major elements and for minor elements at levels >1 wt.% (cf. [12]). The analytical results for dellagiustaite in Mt Carmel are given in Table 3 25 wt.% and VO 39.75 wt.%. However, as discussed below, the V 3+ abundance is important for cation ordering and other features. If a crystal with composition V 3+ > Al were to be found, it would represent a successive member of the spinel subgroup, however with a new root-name. Dellagiustaite can be slightly zoned due to limited V 2+ −1 Mg substitution in the crystals from the type locality (Table 1), whereas significant variation is observed in samples from Israel (Table 3).     Along with dellagiustaite, other spinel grains in samples from the same locality can occur and they show zonation from V-rich cores (close in composition to dellagiustaite), and increasing Mg and Al towards the rim of the crystals, suggesting that substitution of V 3+ −1 Mg is more pronounced than V 3+ −1 Al (Table 2). In Sierra de Comechingones, hibonite is rich in V 3+ (ca. 0.21-0.24 apfu), while grossite is depleted in V 3+ (<0.03 apfu).
At Mt Carmel, dellagiustaite is generally richer in Mg and is highly zoned with rims being richer in Mg and poorer in V 2+ , while V 3+ and Al maintain an almost constant ratio. Smaller grains are richer in V 2+ , as are micrometer-sized inclusions included in hibonite and grossite.
Hibonite at Mt Carmel is rich in V and is zoned to rims that are depleted in Ca and richer in V, K and Na (Table 3). This phase, closely related to hibonite, is presently under study. Grossite at Mt Carmel is very similar in composition to that of Sierra de Comechingones (Table 3). Krotite (Table 4) and Ca 2 Al 2 O 5 are also observed at Mt Carmel but not observed at Sierra de Comechingones yet.
In both localities, a high-Al perovskite phase is present as dendritic crystals (Tables 2 and 4). The substitution of Ti 4+ by V 3+ and Al can be coupled with some H, involving contemporaneous vacancies at A sites. Further study is necessary but is hampered by the small size of these crystals.

X-ray Crystallography
X-ray diffraction data for dellagiustaite were obtained by Debye-Scherrer geometry using an Oxford Diffraction Xcalibur diffractometer at the Department of Earth Sciences, University of Milan, Milan, Italy, operating at 50 kV and 30 mA, with a monochromatized MoKα radiation and equipped with a CCD detector at 80 mm from the sample position. The one-dimensional pattern was refined by the Le Bail method using JANA2006 [19]. Indexed d values and relative peak heights above background are given in Table 5. Refined cell parameters are a = 8.1877(10) Å, V = 548.89(6) Å 3 .
A small dellagiustaite fragment was extracted from the polished section. X-ray single-crystal intensity data were collected using the same instruments and conditions. A combination of ω/ϕ scans, with step scan 1 • and exposure time 30 s per frame at low theta angles, and duplicating counting time at high-theta angles, was used to maximize redundancy and data coverage. A second data collection was performed on a V-rich spinel grain coming from another Argentinian sample corresponding to the spinel composition #7 reported in Table 2, collecting images at 20 s per frame. Crystal data and details of the intensity data collection and refinement are reported in Table 6.
The structure of dellagiustaite was refined using SHELX-97 [20] starting from the atomic coordinates of magnetite. Scattering curves for fully ionized chemical species were used at cation sites; neutral vs. ionized scattering curves were used at the anion sites. This strategy allowed accounting for the ionization of anions. Scattering curves were taken from the International Tables for X-ray Crystallography [21]. The Fourier difference map did not reveal any maximum over 0.28 e − Å 3 . Refined atom coordinates and equivalent isotropic displacement parameters are reported in Table 7. Selected interatomic distances and bond angles are given in Table 8. Crystallographic information files and lists of observed and calculated structure factors for both crystals are available as Supplementary Materials (Files: dellagiustaite.cif/.fcf and 12794B.cif/.fcf).    (7) The anisotropic displacement parameters refined for the M sites showed a particular behavior, depicted as a prolate thermal ellipsoid. Therefore, several tests were undertaken to explore a possible decrease of symmetry but none reported a better fit for solving this feature. The prolate character of ellipsoid was reduced (although not removed) in the V-rich spinel and therefore it could be related to the V content.

Description of the Crystal Structure
Site populations for dellagiustaite and V-rich spinel have been derived from the unit formula (for V-rich spinel we used the analysis #7 from Table 2) and are reported in Table 9. The refined <M-O> distance of 2.045 Å is somewhat longer than the expected 2.038 Å, possibly due to the presence of V 2+ at the octahedral sites, as estimated by WDS-EMPA; in fact, the values for V 2+ -O in octahedral coordination were taken from the VO structure [22], whereas the other bond distances are taken from [23] except for Ti 3+ -O, which was taken from the Ti 2 O 3 structure [24]. The V 2+ -O value of 2.129(4) Å, obtained statistically in [25], is certainly too high and leads to a higher disagreement (calculated M-O mean bond length using such value would be 2.069 Å). Sutton et al. [8] found an almost zero intensity for the pre-edge peak in V 2+ glasses by XANES and concluded that it is likely that V 2+ occurs in a highly symmetrical octahedral site in these glasses. Therefore, we have kept both V 3+ and V 2+ in octahedral coordination when calculating the distribution of cations among the sites. Overall, the agreement in terms of site scattering (in electrons per site, eps) with the chemical analyses is within 3% and, for mean bond lengths, the difference is about 0.006 Å. While the V-rich spinel shows an intermediate degree of inversion (i = 0.53), dellagiustaite is fully inverted.
Dellagiustaite is therefore an inverse spinel. Besides being V 3+ -rich, it cannot be classified as a "vanadium coulsonite" because that would require Al < V 3+ . In addition, coulsonite is a normal spinel. We are looking carefully through our samples in search for spinel grains with higher V-contents, which would require disordering of V 3+ over Tand M-sites and would represent a new spinel species with a new root-name. Table 9. Site occupancies (site scattering (s.s.) and electrons per site (eps)) and mean bond-lengths (m.b.l., Å) for dellagiustaite and V-rich spinel. The behavior of prolate thermal ellipsoids of the M site would seem to show that the spinel structure has some difficulty in hosting V cations with two valences, when V is a dominant component. In fact, Bosi et al. [26] found a limited accommodation of V 4+ in the MgAl 2 O 4 -Mg 2 V 4+ O 4 series and concluded that the distribution of V valences in spinels does not exclusively reflect oxygen fugacity, but also depends on the activity and solubility of all chemical components in the crystallization environment. Bosi et al. [26] also concluded that the inverse cation distribution can influence the incorporation of V 4+ , being favored in inverse spinels. Apparently, this is also observed for V 2+ in our samples.
On the other hand, Sutton et al. [8] observed a slight compositional dependence of the oxidation state of V with the composition of glasses, with more highly reduced V in low-Ca glasses. The zoning observed in spinel ( Table 2) seems to support this hypothesis as zoning encompasses an oxidation of V as well as the crystallization of more Ca-rich phases due to the fractional crystallization of more Al-rich phases like hibonite.
It is worth noting that T Al M (V 2+ V 3+ )O 4 is a synthetic compound that has been reported to show a charge ordering (CO) phase transition with rhombohedral distortion along [111] of the cubic cell (α decreases from 60 • to ca. 59 • below 427 • C) [27]. The CO structure is converted at P > 23 GPa to the cubic structure, which does not revert back to the rhombohedral form on decompression [28]. In the rhombohedral phase, three distinct type of sites are observed for vanadium: V1, V2 and V3. Molecular groups made of six V3-sites and one V2-sites form heptamers with short V-V distances (V2-V3 = 2.8238(2) Å and V3-V3 = 2.6740(1) Å) and antibonding distances (V3-V3 = 3.0850(1)Å), while V1-V3 are also in a non-bonding configuration of 3.0001(2) Å [28]. Recently, this model has been discussed proposing alternatively the presence of trimers (V 9+ 3 ) and tetramers (V 8+ 4 ) rather than V 17+ 7 heptamers, by allowing the shift of V2 atoms along [111] in a polar R3m model obtained by PDF using X-ray diffraction powder data [29]. The trimers were also considered by Talanov et al. [30] in their density functional theory study of the crystal chemical analysis of V-O bond lengths of In dellagiustaite, we observed a cubic structure at room T, with no evidence of superlattice reflections at 1 2 [111]. In the cubic model the V-V distances are indeed all equivalent (2.89738(5) Å in this work; 2.92 Å in [27]). However, the elongated thermal ellipsoids along [111] in the mixed tetrahedra-octahedra layers, and along (100) in the octahedra layers (Figure 6c-e) are an indication of static disorder at the M-sites, possibly indicating a sort of frustrated charge ordering. Our single-crystal data allowed an accurate determination of the atomic displacement parameters (adp); this is not possible with the previous experimental data, all of which are from powdered synthetic run products.
In dellagiustaite, charge ordering possibly could be frustrated at room T due to the presence of minor amounts of Al at the M sites. In fact, it has been described that doping of AlV 2 O 4 by Cr 3+ frustrates the CO with the generation of microdomains (as small as 10-20 nm) observed by electron diffraction [31]. It could be worthwhile studying dellagiustaite by electron diffraction to test for the presence of CO domains.  (111) showing the disorder of V atoms within the plane, related to statistical configurations of three V-bonded and three anti-bonded groups. Thermal ellipsoids plotted at 95% probability with Vesta 3.0 [32]. Al in four-fold coordination (blue navy) and vanadium (2+ and 3+) in six-fold coordination (red). Green = O. There are three symmetrically different V sites in the rhombohedral phase but only one in the cubic. The single crystal data on dellagiustaite shows the adp values of V atoms compatible with static disorder (domains?) or dynamic disorder, in the cubic phase. The disorder resembles the low-T configuration.

Conditions of Formation
The genesis of dellagiustaite at Sierra de Comechingones is difficult to decipher, mainly because of the missing information about the exact location of the outcrop and the related rocks. Furthermore, the paragenesis is rather uncommon: the rock is mostly composed of Al 2 O 3 and CaO, with minor other components (mostly V 2 O 3 and even lesser amounts of TiO 2 , Na 2 O and F). The almost complete absence of SiO 2 also makes the paragenesis rather unusual. The dominant mineral phases are hibonite and grossite. Hibonite is a constituent of some refractory Calcium-Aluminum Inclusions (CAIs) in carbonaceous meteorites, commonly accompanied by grossite and spinel, crystallized early in the evolution of the solar nebula. Grossite has been found in very few localities but is also a relatively frequent mineral phase in CAIs. However, other aspects are even more puzzling. In particular, the presence of almost pure vanadium alloys with requires extremely reducing conditions (∆ IW, −9, where IW = iron-wustite buffer).
The fact that very similar rocks have been recently described from Mt Carmel (northern Israel) makes its origin even more interesting. The rocks in Israel contain super-reduced mineral assemblages, which crystallized from high-T melts trapped in corundum aggregates (micro-xenoliths) ejected from Cretaceous volcanoes of picritic tholeiite composition [9,10]. In these rocks, highly reduced minerals have been described, such as moissanite (up to 4 mm long; larger than those reported in [33]), tistarite, khamrabaevite, and gupeiite along with native vanadium, and have been interpreted as the result of interaction between deep-seated magmas and CH 4 ± H 2 fluids in volcanic plumbing systems, producing local ultra-reducing conditions [9][10][11][12]. The Argentinian rocks containing dellagiustaite are very similar to type-DF pockets in Mt Carmel, although, in the dellagiustaite-bearing rocks, the vanadium alloy is included in both hibonite and grossite and has tubular-cylindrical shapes. In addition, corroded laths of corundum included in hibonite have been observed in some Mt Carmel material, but not in the Argentinian samples; the rock containing dellagiustaite may represent a more calcic composition.
Hibonite and grossite are rather rare minerals in terrestrial occurrences. Hibonite originally was described in the Esiva alluvial deposits (Madagascar [34]), probably derived from nearby deposits of thorianite bearing skarns in granulite belts of Madagascar and Tanzania [35]. In the Madagascar occurrences, hibonite replaces corundum and spinel in corundum + spinel + scapolite assemblages. Another occurrence of hibonite has been reported in Ca-Mg-Al granulitic xenoliths in basanites of the Kwa Nthuku volcano in the northwestern part of the Chyulu Hills, a volcanic field at the eastern flank of the Kenya rift some 150-200 km east of the rift axis, about 50 km NE of Mt. Kilimanjaro. In the Ca-Al granulites, hibonite (±spinel) are the earliest, possibly igneous, minerals in the crystallization sequence [36]. Hibonite has been also reported as tabular crystals up to 3 cm across in calcitic marbles from the Tashelga-Malzaskaya region in Siberia [37]. In such localities, metamorphic hibonites are notably high in Fe 3+ , and coexist with V-rich minerals (goldmanite, tashelgite, and mukhinite). Hibonite has also been found in the alluvia of Myanmar, Mandalay Region, which has been used as a gemstone [38]. Grossite has been found on Earth only in the type locality, the Hatrurim Formation; the rest of the observed occurrences are in meteorites. Krotite previously has been observed only in Calcium-Aluminum Inclusions in meteorites [39].
If the origin of the type material is similar to the Mt Carmel rocks, a highly plausible locality is in one of the outcrops of the Neogene Volcanic Belt of San Luis (SLNVB) in the Pampean flat-slab segment, in particular in Sierra del Morro (33 • 10 S, 66 • 24 W), in the Conlara Valley, which is closer to Sierra de Comechingones than to Sierras de San Luis. These volcanic materials are pyroclastic trachyandesites-trachydacites with shoshonitic to high K calc-alkaline character consistent with a subduction-related continental margin setting [40].