Ferri-taramite, a new member of the amphibole supergroup, from the Jakobsberg Mn–Fe deposit, Värmland, Sweden

. Ferri-taramite (IMA CNMNC 2021-046), ideally A Na B (CaNa) C (Mg 3 Fe 3 + 2 )(Si 6 Al 2 )O 22 W (OH) 2 , occurs in skarn from the Jakobsberg manganese mine, Värmland, Sweden. Associated minerals are celsian, phlogopite, aegirine-augite, andradite, hancockite, melanotekite, microcline (var. hyalophane), calcite, baryte, prehnite, macedonite and oxyplumboroméite. Conditions of formation, close to peak metamorphism (at circa 650 ◦ C and 0.4 GPa), include silica undersaturation, a slightly peralkaline character and relatively high oxygen fugacities. Ferri-taramite forms poikiloblastic crystals up to 5 mm and is dark brownish black with a yellowish grey streak. The amphibole is brittle with an uneven to splintery fracture. 26 % for 2722 reﬂections with I o > 2 σ(I) The Mn 2 + and Fe 2 + ions show preference for the M 1 and octahedrally coordinated sites, whereas Fe 3 + is strongly ordered at M 2. The A -group cations, K and Na, are split over two subsites, A(m) and A (2), respectively.

Chemical analyses corresponding to potassic-ferri-taramite have been reported in intergranular spaces of an orthopyroxenite xenolith from the Udachnaya-East kimberlite pipe, Daldyn kimberlite field, Siberian platform (Rezvukhin et al., 2020), and interpreted as metasomatic, similar to conditions in the Ilmenogorsky miaskite complex in the South Urals, Russia (Makagonov et al., 2018).
The present paper is the full scientific characterization of the C Fe 3+ analogue of taramite. Its name, ferri-taramite, is thus dictated by the amphibole nomenclature in force (Hawthorne et al., 2012). The mineral was approved prior to publication by IMA CNMNC (no. 2021-046). The material used here, a skarn sample from the Jakobsberg manganese mine in Sweden, was once examined by Flink (1914), who noted the unusual character of the amphibole and described it as a "strange hornblende". No chemical data were reported for the sample at the time. The holotype material is deposited in the type mineral collection of the Department of Geosciences, Swedish Museum of Natural History, P.O. Box 50007, 10405 Stockholm, Sweden, under collection number GEO-NRM #19221254 (specimen and sections).

Physical and optical properties
Ferri-taramite has a short prismatic habitus, along [001]. The colour of the crystals is dark brownish black (greenish yel- low to brown in splinters), with a yellowish grey streak. The lustre is vitreous and semi-transparent, and no fluorescence is observed under UV light. The amphibole is brittle with an uneven to splintery fracture. Cleavage parallel to {110} is good. Hardness (Mohs) is ∼ 6. Vickers hardness numbers were obtained on a polished section of ferri-taramite using a micro-hardness tester loaded with a 200 g weight, giving a mean value of 919 (9.0 GPa) for a range of 733-1190 from eight indentations. The density was not measured because of frequent inclusions of heavier minerals (e.g. macedonite); a calculated value is 3.227(5) g cm −3 based on the unit-cell volume and atom contents. Ferri-taramite is slightly attracted by a neodymium magnet. It is insoluble in HCl and HNO 3 .

Chemical composition
The chemical composition (Table 1) was primarily determined using an FEI Quanta 650 field-emission scanning electron microscope, fitted with an 80 mm 2 X-Max N Oxford Instruments EDS detector (20 kV, beam size >1 µm, working distance 10 mm). The beam current was calibrated on Co metal, with the instrument calibrated against metal and mineral standards for each element. The number of point analyses is 10. F and Cl were below detection. Contents of major, minor and other trace elements (Table 2) were also determined with laser-ablation ICP-MS analysis (an NWR 213 nm laser coupled to an Agilent 8800 MS/MS) at the University of Gothenburg, with 16 analysed spots. Element oxide quan-tification was performed by normalizing to 100 wt %, assuming 2 wt % H 2 O and the Fe 2+ / Fe 3+ ratio from Mössbauer data. We used a multi-standard approach, with measuring calibration factors for each major element from a wide spectrum of known standard materials: NIST SRM 610, ATHO-G, BCR-2G, BHVO-2G, GSC-1G, GSD-1G, Kakanui amphibole, NKT-G, T1-G and TB-1G (all data from the Geo-ReM database, except for Kakanui amphibole, which is from unpublished electron microprobe data). Consistency of calibration factors is generally better than 4 %, confirming an excellent accuracy when compared with EDS data (Table 1). NIST SRM610 was used as the only primary standard for trace element quantification. The elements P, Cl, Br, Cs, W, Tl, Ta, Bi, Th and U were found to be at or below the level of detection.
The empirical formula of ferri-taramite derived from EDS data and calculated on the basis of 24 ( 5 X-ray diffraction data and structure refinement

Crystal structure refinement
A single-crystal X-ray diffraction (SC-XRD) study was done on a 146×213×342 µm fragment (Fig. 4) using a Rigaku Oxford Diffraction XtaLAB Synergy diffractometer, equipped with a PhotonJet (Mo) X-ray source operating at 50 kV and 1 mA, with monochromatized MoKα radiation and equipped with a HyPix detector working at 62 mm from the crystal. A combination of ω scans at different values of φ, χ and θ positions, with a step scan of 0.5 • and exposure time of 2 s per frame, was used to maximize redundancy and data coverage. High-resolution data were collected (up to 0.46 Å).  The crystal structure of ferri-taramite was refined starting from the atom coordinates of sample Q99-3 of Oberti et al. (2007a) (named "alumino-magnesiotaramite" in the paper). Scattering curves for fully ionized chemical species were used at sites where chemical substitutions occur; neutral vs. ionized scattering curves were used at the T (2) and anion sites. Scattering at T (1) sites was fixed at the 54 : 46 Si 4+ : Al 3+ ratio as deduced form application of the equation of Oberti et al. (2007b) and in agreement with chemistry. Scattering curves for neutral and fully ionized atoms were taken from International Tables for Crystallography (Wilson, 1992). Splitting of the electron density at site A was observed and added to the model as A(2) and A(m) sites. Full-matrix least-squares refinement on F 2 yielded R obs = 2.26 % (2722 reflections with I o >2σ (I )) and R all = 2.54 % (2983 reflections) using SHELXL (Sheldrick, 2015). See Table 4 for experimental details. Refined atom coordinates, site scattering and equivalent isotropic displacement parameters are reported in Table 5. Selected interatomic distances and bond angles are given in Table 6. Site populations for ferri-taramite have been derived from the unit formula and the results of the structure refinement (Table 7). There is an excellent agreement between the refined values of site scattering (ss, electrons per formula unit) and mean bond lengths (mbl's, Å) and those calculated based on the proposed site populations. A crystallographic information file (CIF) containing observed structure factors has been deposited in the Supplement.  assigned to C Al-O stretching, whereas the one at 669 cm −1 corresponds the bending of Si-O-Si bridges (see Waeselmann et al., 2020). The strong peak at 536 cm −1 (with a shoulder at ca. 500 cm −1 ) is tentatively assigned to stretching vibrations of Fe 3+ -O octahedra and the one at 370 cm −1 to Mg-O vibrations.

Fourier-transform infrared (FTIR) spectroscopy
Polarized single-crystal FTIR spectra were collected in the range 2000-10000 cm −1 at a spectral resolution of 2 cm −1 on a 50×100 µm raster, with a Bruker Vertex 70 spectrometer attached to a Bruker Hyperion 2000 IR microscope. Measurements were done from a 58 µm thick doubly polished singlecrystal fragment oriented parallel (010). The spectra (Fig. 6) show   Figure 5. Raman spectrum of ferri-taramite, obtained with a 514 nm laser. (Hawthorne and Della Ventura, 2007). The interatomic distance from the H atom to the O(7) receptor of 2.67 Å is in good agreement with the main band at 3700 cm −1 , based on the known empirical correlations (Libowitsky, 1999). The shoulder band at ca. 3670 cm −1 is possibly related to the partial vacancies at the A site. Local A Na-MnMgMg-SiAl arrangements as a substitution of Mg by Mn 2+ at M(1,3) is expected to cause a band shift of at least −15 cm −1 (see Reece et al., 2002;Hawthorne and Della Ventura, 2007), which may contribute to the observed spectrum. Figure 6. Polarized single-crystal FTIR spectra of ferri-taramite. Sample thickness is 58 µm.

Optical absorption spectroscopy
Polarized optical absorption spectra (Fig. 7) were measured on an X-Z crystal section of ferri-taramite (same as used for FTIR). The spectra were recorded in the visible range (350-800 nm) with an AvaSpec-ULS2048x16 spectrometer attached to a Zeiss Axiotron UV microscope. A xenon lamp (75 W) was the light source, and Zeiss Ultrafluar 10× lenses were used as the objective and condenser. The size of the aperture was 30 µm in diameter on the sample. A Glan-Thompson calcite prism served as the polarizer. The spectra show a significant absorption edge towards the  (6) (10) UV, most pronounced in the Z direction. A set of absorption bands caused by spin-forbidden electronic d-d transitions are superimposed on the background. Narrow bands located at ca. 24 500 and 23 000 cm −1 mark field-independent transitions in Mn 2+ and Fe 3+ , respectively. Broader bands at higher wavelengths, ca. 14 000 cm −1 , could be related to Fe 2+ → Fe 3+ or Fe 2+ → Ti 4+ charge transfer between neighbouring amphibole octahedra (Burns, 1993).

Crystal structure and chemistry
Site populations (Hawthorne et al., 1995) were calculated based on present knowledge of amphibole crystal chemistry. The results are reported in Table 7. Here we describe the different cation and anion groups. T sites. The observed T (1)-O distance is in perfect agreement with the calculated values using the equations of Oberti et al. (2007b; see Tables 6 and 7) and the chemical analyses. Therefore, the 54 : 46 Si 4+ : Al 3+ ratio was fixed at  the last stages of refinement. The limited presence of Al in the T (2) sites assumed in Table 7 (2 % per site) is observed in high-temperature high-T Al amphiboles, in agreement with peak conditions for its crystallization (see paragenesis section).
C-group cation sites. On the basis of the observed mean bond lengths, for C Mn 2+ the following (strong) site preference is inferred: M(1) = M(3) M(2). This is similar to what is expected for C Fe 2+ according to present knowledge of amphibole crystal chemistry (see the review in Oberti et al., 2007b), and it is related to the structural relaxation of M(1) and M(3) sites as a consequence of the size reduction of the adjacent M(2) sites. As in other taramitic amphibole samples (Oberti et al., 2007a), the observed mean octahedral distances are larger than the calculated values, although in Table 7. Site occupancies for ferri-taramite with site scattering (ss) and mean bond lengths (mbl's).

Site
Site population (apfu) ss (epfu) mbl (Å) this case the differences are within 0.002 Å, probably related to the fact that the presence of Fe 3+ in M(2) does not allow for a strong relaxation of this site. In fact, the observed mean octahedral bond length for M(2) site is larger than the calculated value (Table 7). At any rate, the very good agreement between observed and calculated site scattering (a remarkable one) is obtained considering some disordering of Mn 2+ and Fe 2+ among the C sites, probably due to the relatively high T of crystallization in the amphibolite facies (see the following). B-group cation sites. The M(4) site shows the typical y/b coordinate of a Ca and/or Na populated site. No evident residual electron density has been found around the M(4) sites in agreement with the absence of Mg and negligible Mn 2+ at the B-group sites, based on the chemical analyses.

Refined Calculated Refined Calculated
A-group cation sites. The A cations are preferentially ordered, with K at the A(m) site (0.066 apfu of K) and Na (0.873 apfu of Na and minor 0.06 apfu of K) at the A(2) subsite (0.57 and 10.87 epfu, respectively). Both K and Na contents agree reasonably well with the chemical analysis (Na 0.79 K 0.16 Pb 0.01 ). The main preference for A(2) (see Figs. 9 and 1 of Oberti et al., 2007a, for comparison) is also observed in ferro-taramite and taramite related to the preferred local arrangement M(4) Ca-O(3) OH-A(2) Na, although higher occupancy of A(m) sites would be expected due to the local arrangement M(4) Na-O(3) OH-A(m) Na. This might be due to the absence of O(3) F, responsible for ordering at the A(m) sites observed in fluoro-taramite (Oberti et al., 2007a). When the A site is dominantly populated by K, like in potassic-ferro-taramite, strong ordering of K at the A(m) site is also observed even in the absence of F at O(3) (Oberti et al., 2008).
W anion sites. Despite a relatively high Ti content of 0.06 apfu, there is no indication of deprotonation and existence of an oxy-component ( W O 2− ) in ferri-taramite, which is consistent with the metamorphic origin of ferri-taramite (see Oberti et al., 2015b).  (201) and centred at 0, 1/2, 0 for ferri-taramite, showing the distribution of the electron density within the A cavity. The b axis is horizontal, and contours are drawn with steps of 1 e Å −1 starting from 1.

Trace elements
Ferri-taramite from Jakobsberg is low in most trace elements. Increased levels, compared to the continental-crust average, are however seen for Be, As, Sb and Pb, which can be considered a Långban-type signature reflecting the general enrichment of these elements in the skarn (Holtstam and Mansfeld, 2001). There is also a significant concentration of the transition elements Sc, Cr, Co, Ni and Zn. Global comparisons are hampered by the fact that comprehensive trace element data for amphiboles of an unmistakable non-igneous origin are very rare. Cr and Co concentrations are slightly enriched compared to an amphibolite-grade hornblende from the Njurundukan mafic complex, Baikal, Russia (samples 487, 499, 23 and 54/3 in Skublov and Drugova, 2003). Concentrations for other elements available for comparison, such as Y, La, Ce, Lu, Hf and Ta, are significantly lower in ferritaramite. This effect can possibly be explained (in the case of REE) by the coeval crystallization of ferri-taramite with hancockite that has an epidote-type structure. Amphibole in a magmatic-influenced skarn can show considerably higher Sr, Rb, Sn and Ta contents but is generally lower in the period-4 transition elements (e.g. Žáček, 2007) than seen for ferritaramite. The partitioning behaviour of trace elements among major skarn-forming minerals (where there is no melt phase present) is, however, essentially unexplored.
In the case of rare hjalmarite, B Mn > B Ca (Holtstam et al., 2019b). The rock with ferri-taramite formed during regional metamorphism as is the case for the above-mentioned amphiboles in general. In a possible scenario, precipitates of Fe-Mn (hydroxy-)oxides, Ca-Mg-Ba-Pb carbonates and siliceous material of hydrothermal origin were mixed with aluminous detritus and reacted to eventually form a skarn assemblage of amphibole, mica, feldspar, clinopyroxene and Pb silicates. The alkali metals have probably originally been derived from metasomatic alteration of the felsic volcanic units (see Magnusson, 1930). Conditions for ferri-taramite formation, close to peak-metamorphism (at circa 650 • C and 0.4 GPa), demonstrably included silica undersaturation, a slightly peralkaline character and relatively high oxygen fugacities of the system. Obviously, hydrous alteration has affected the Jakobsberg skarn at some point, but the amphibole shows little sign of breakdown (limited prehnite formation).
Katophorite, also a rare species (Oberti et al., 2015a), can be considered an intermediate amphibole composition between richterite and taramite, related by the exchange C Mg 5 + T Si 8 → C (Mg 4 Al) + T (AlSi 7 ) → C (Mg 3 Al 2 ) + T (Al 2 Si 6 ) (or with C Fe 3+ instead of C Al, for the "ferri-" members). It is however not known from the Långban-type deposits. Although the proper bulk compositions could likely be achieved in the rocks present, katophoritic amphiboles seem to belong to a highpressure domain (Harlow and Sorensen, 2005;Pirard and Hermann, 2015;Oberti et al., 2015a). Ferri-taramite must be a rare composition since we have found very few verified analyses elsewhere in the literature (see Schumacher, 2007). In addition, the present specimen is the first example of a taramitic amphibole from a skarn, not associated with rock types of magmatic origin. The closest composition to ferri-taramite has been reported from eclogites in the Aktyuz area, northern Kyrgyz Tian Shan, Kyrgyzstan (Takasu and Orozbaev, 2009), but without a direct Fe 2+ / Fe 3+ determination and with very similar Fe 3+ and Al V I content, it is difficult to discern if these samples are ferri-taramite or taramite. In the Ilmenogorsky complex in the South Urals, analysis no. 49 of Nikandrov et al. (2000) can be also classified as ferri-taramite, although it is richer in Fe 3+ than our sample from the type locality. Similar compositions are reported by Makagonov et al. (2018) from the same locality, but Ca at the B site of the amphibole is 1.49 apfu, and the amphibole is therefore very close to magnesio-hastingsite in composition.
Data availability. A CIF file is deposited in the Supplement below.
Author contributions. DH and AK measured optical and physical properties. DH and HS obtained and interpreted spectroscopic data. AK and TZ performed the chemical analyses. FC collected singlecrystal X-ray diffraction data and executed the structural refinement. DH and FC wrote the manuscript, in consultation with all authors.