Structural properties and heat-induced oxidation-dehydrogenation of manganoan ilvaite from Perda Niedda mine, Sardinia, Italy

An unusually Mn-rich ilvaite sample from the Perda Niedda mine in Sardinia, Italy, was studied in order to clarify the Mn distribution among the different structural sites, and to observe the structural response of the mineral upon thermally induced oxidation-dehydrogenation. The crystal structure and the chemical composition of one crystal [a = 13.014(5), b = 8.867(3), c = 5.838(4) Å, b = 90.02(4)∞] were investigated. X-ray crystal-structure refinement, performed in the Pnam space group, and electron microprobe analyses yielded the formula (Ca0.98Mn0.02)(FeFe)(Mn0.72Fe0.28) (Si2O7)O(OH). Crystal chemical details, compared to structural data from literature, led to the assumption that Mn replaces Fe, mainly at the M2 site. Annealing experiments and structure refinements were performed in the temperature range 400–690 ∞C. No phase transition was observed over the entire temperature range. Oxidation of Fe at the M1 site, with concomitant dehydrogenation, was deduced from examination of the structural adjustments occurring as the temperature was increased. A useful model to evaluate a possible OH ́ O substitution in ilvaite was obtained. et al. 1988; Ghazi-Bayat et al. 1989; Ghazi-Bayat et al. 1992; Ghazi-Bayat et al. 1993; Schmidbauer and Amthauer 1998; Amthauer et al. 1998). Although not usually considered a rock-forming mineral, ilvaite is often an abundant constituent of Fe and Zn skarns (Burt 1971; Einaudi et al. 1981). It can replace hedenbergite (Logan 2000) and, as a result of an unusual set of conditions, it was found to occur as an alteration product replacing fayalitic olivine in the Skaergaard intrusion (Naslund et al. 1983). Ilvaite probably formed during serpentinization of peridotite (Agata and Adachi 1995), and ilvaite occurring in rodingites associated with serpentinite (Lucchetti 1989) has also been reported. The chemical composition of ilvaite is often close to that of the ideal end-member, with only minor amounts of other cations such as Ti, Mg, Al, and Mn. The most common and quantitatively relevant substituent is Mn, which mainly replaces Fe at the octahedral sites but can also substitute for Ca at the seven-coordinated site (Carrozzini 1994). According to GhaziBayat et al. (1992), Mn content affects the degree of monoclinicity (b angle close to 90.0∞ when Mn = 0.19) and, therefore, the transition temperature, which decreases with increasing Mn content (Amthauer et al. 1998). However, as Takéuchi et al. (1983) pointed out, the value of the monoclinic angle is not only related to the degree of disordering. An apparent orthorhombic symmetry, in fact, can be easily simulated by the co-existence of fine components polysynthetically twinned on both (001) and (100); such a twinning would be easily generated if positional mistakes in the cation array take place during the Fe-Fe ordering process, which occurs when the crystal is cooled through the phase transition temperature (Takéuchi et al. 1994). Takéuchi et al. (1993, 1994) investigated crystals from Kamioka mine, with Mn ranging from 0.19– BONAZZI AND BINDI: STRUCTURAL PROPERTIES OF MANGANOAN ILVAITE 846 0.22 apfu (atoms per formula unit; Tochibora skarn) to 0.52 apfu (Maruyama skarn). Taking into account the findings of these authors, who cast doubt on the existence of intrinsically orthorhombic ilvaites in nature, Mn incorporation more likely should induce polysynthetic twinning, rather than cause a monoclinic to orthorhombic phase transition. A still unresolved problem concerns the ordering of Mn among the octahedral sites. Ghazi-Bayat et al. (1989) reported the results of X-ray powder diffraction and Mössbauer studies on synthetic CaFe2–xMnxFe(Si2O7)O(OH), with x = 0.00, 0.12, 0.15, 0.19, and concluded that Mn replaces Fe at both the M11 (8d) and M2 (4c) sites and not preferentially at one site (M2) as previously found by Haga and Takéuchi (1976). According to Cesena et al. (1995), their Mössbauer data obtained for synthetic Mn-bearing ilvaite allowed an unambiguous assignment of Mn to the M11 site. By contrast, accurate structural data reported by Carrozzini (1994) showed a good linear relationship between the Mn content and . However, a partially disordered distribution of Mn over all the cation sites was proposed by this author for Mn-rich crystals (up to 0.58 apfu) from Oridda (Sardinia, Italy). Ilvaites with higher Mn contents have also been described (Plimer and Ashley 1978; Meinert, 1987; Logan 2000) but no crystallographic data were reported. The present study was undertaken to provide a crystalchemical characterization of an ilvaite sample from Perda Niedda mine (Sardinia, Italy) exhibiting an unusually high manganese content (up to 0.73 apfu). In addition, the structural variations induced by heating the sample in air were examined, and compared to those previously reported for heated ilvaite having a chemical composition close to that of the endmember (Bonazzi and Bindi 1999). In accordance with the previous finding, ilvaite easily undergoes the heat-induced ilvaite Æ oxy-ilvaite transformation, involving both hydrogen loss and iron oxidation. Because of charge balance requirements, oxidation involves only Fe located at the M1 site (Bonazzi and Bindi 1999). Keeping this in mind, as well as the crystal chemical differences between Fe and Mn, it should also be possible to obtain evidence of the manganese ordering, if any, between the M1 and M2 sites. EXPERIMENTAL METHODS Six single crystals of manganoan ilvaite from a skarn at the Perda Niedda mine, Sardinia, Italy (sample no. 10297/509, Mineralogical Museum of the University of Florence) were selected and unit-cell parameters were determined by means of least-squares refinement of the setting angles of 25 reflections (16 < J < 25∞) measured with a CAD4 single-crystal diffractometer (Table 1). Subsequently, the same crystals were used for chemical analysis. Chemical compositions were determined using a JEOL JXA 8600 electron microprobe operating at 15 kV and 10 nA and equipped with four wavelengthdispersive spectrometers. Table 2 reports the chemical data and the atomic proportions calculated on the basis of six cations. On the basis of its Mn content, the crystal labeled PN6 was selected for the heat treatments and the structural study. For this purpose, it was subsequently removed from the resin and the unit-cell parameters were determined again. Intensity data were collected in the J-range 2–35∞, using graphitemonochromatized MoKa radiation. Data were subsequently corrected for Lorentz-polarization and absorption effects (North et al. 1968). The crystal was annealed in air for 48 h at selected temperatures ranging from 400 to 700 ∞C using a magnetic release furnace that allows rapid cooling to room temperature. After each heat treatment, determination of unit-cell parameters and intensity data collection was repeated (experimental conditions are given in Table 3). As the annealing temperature increased, reflections became broader and weaker; after the annealing at 700 ∞C, the intensities were no longer collected. As shown in Table 3, the b angle value for the untreated PN6 crystal was found to be 90.00∞ within the limits of experimental error. Only five 0kl reflections with k + l = 2n + 1 were observed (i.e., 041, 061, 032, 052, 054), showing [Fo/s(Fo)] £ 5.0. After heating at 400 ∞C, no systematic absence violations were found. Structure refinements were performed using the SHELXL93 program (Sheldrick 1993), with weighting scheme w = k/ sFo. Scattering factors and correction factors were taken from the International Tables for X-ray Crystallography, vol. IV (Ibers and Hamilton 1974). As for PN6-RT, least squares were run in both monoclinic (Robs = 2.79%) and orthorhombic symmetry (Robs = 2.59%). The atomic arrangement obtained by reTABLE 1. Unit-cell parameters for selected crystals of ilvaite from Perda Niedda a (Å) b (Å) c (Å) b (∞) V (Å3) PN1-RT 12.988(7) 8.829(2) 5.844(3) 90.19(3) 670.1(6) PN2-RT 13.000(8) 8.835(5) 5.837(8) 90.05(7) 670.4(7) PN3-RT 12.995(3) 8.839(2) 5.840(1) 90.15(2) 670.8(4) PN4-RT 13.008(5) 8.844(1) 5.847(2) 90.20(2) 672.7(4) PN5-RT 13.017(6) 8.857(1) 5.846(2) 90.26(2) 674.0(5) PN6-RT 13.014(5) 8.867(3) 5.838(4) 90.02(4) 673.7(5) PN6-400 13.015(6) 8.868(1) 5.842(2) 90.01(2) 674.3(5) PN6-500 13.008(7) 8.865(2) 5.845(3) 90.00(2) 674.0(6) PN6-600 12.993(7) 8.864(2) 5.837(2) 90.00(2) 672.2(6) PN6-650 12.983(8) 8.868(3) 5.822(2) 90.00(2) 670.3(6) PN6-675 12.992(6) 8.869(2) 5.820(1) 90.00(2) 670.6(5) PN6-690 13.015(6) 8.873(2) 5.816(1) 90.00(2) 671.6(5) PN6-700 13.032(5) 8.877(2) 5.812(1) 90.00(2) 672.4(6) TABLE 2. Chemical composition (wt%) and atomic proportions of the crystals from Perda Niedda PN1-RT PN2-RT PN3-RT PN4-RT PN5-RT PN6-RT SiO2 29.22 29.37 29.11 29.27 29.06 29.31 TiO2 – 0.09 0.12 – – – Fe2O3 19.66 19.36 19.46 19.59 20.03 19.56 Al2O3 – 0.08 0.26 0.26 0.02 – FeO 27.13 26.21 26.06 25.73 24.13 22.49 CaO 12.18 13.17 13.10 13.33 13.27 13.47 MnO 9.61 9.52 9.33 9.43 11.00 12.86 MgO 0.07 0.06 0.08 0.12 0.06 0.03 Total 97.87 97.86 97.52 97.73 97.57 97.72 Si4+ 1.993 1.997 1.985 1.989 1.983 1.994 Ti4+ – 0.004 0.006 – – – Fe 1.009 0.990 0.998 1.002 1.028 1.001 Al3+ – 0.006 0.021 0.020 0.002 – Fe2+ 1.547 1.490 1.486 1.463 1.377 1.279 Ca 0.890 0.959 0.957 0.971 0.970 0.982 Mn2+ 0.555 0.548 0.539 0.543 0.635 0.741 Mg2+ 0.006 0.006 0.008 0.012 0.005 0.003 S cations 6.000 6.000 6.000 6.000 6.000 6.000 Notes: Partition between FeO and Fe2O3 was calculated in order to ensure charge neutrality. H2O was not determined. BONAZZI AND BINDI: STRUCTURAL PROPERTIES OF MANGANOAN ILVAITE 847 finement in the monoclinic space group (P21/a) clearly showed orthorhombic symmetry, with Fe and Fe completely disordered between two independent 4e positions (M11 and M12); subsequent structure refinements were performed in space group Pnam. Keeping in mind that the apparent orthorhombic symmetry, as discussed above, can be reasonably related to the presence of twinned ordered domains, in the following discussion we will refer to the orthorhombic structure. Details of refinements are given in Table 3. Fractional coordinates and isotropic equivalent displacement parameters are given in Table 4. RESULTS AND DISCUSSION Description of the structure The atomic arrangement of ilvaite is formed of octahedral ribbons linked by Si2O7 groups and Ca polyhedra. Hydrogen bonds, directed approximately along the a axis, also connect neighboring ribbons. These ribbons run parallel to the c axis and consist of edge-sharing double chains of M1 octahedra, with the larger M2 octahedra alternatively attached above and below by means of four shared edges. Bond distances and distortion parameters are given in Table 5. Effects of the Mn ́ Fe substitution Before heating, the PN6 crystal shows a structural arrangement quite similar to that of ilvaites exhibiting a very low monoclinicity (Beran and Bittner 1974; Takéuchi et al. 1994). The value of the distance (2.410 Å) and the refined site-scattering of the A site (close to 20.0 electrons) suggest that, in spite of the high Mn content, no extensive Ca ́ Mn substitution occurs. The value of the distance (2.081 Å) is quite similar to that observed for the grand mean in Mn-poor ilvaites (i.e., 2.078–2.081 Å, according to Beran and Bittner 1974; Finger and Hazen 1987; Carrozzini 1994; Bonazzi and Bindi 1999), thus suggesting that no remarkable Mn substitution occurs at the M1 site. On the contrary, the value of the distance (2.207 Å) is unusually high when compared to the corresponding value for Mn-free ilvaites (2.187–2.188 Å for crystals from Seriphos having Mn = 0.02 apfu; Finger and Hazen 1987; Carrozzini 1994). The observed in PN6-RT appears even higher than those found in Mn-rich ilvaites from Oridda ( = 2.199–2.201, MnM2 = 0.36–0.42 apfu; Carrozzini 1994) and Maruyama ( = 2.201 Å MnM2 = 0.50 apfu; Takéuchi et al. 1993). According to the model proposed by Carrozzini (1994), the distance (2.207 Å) observed here would be consistent with an Mn replacing Fe content of 0.59 apfu, a value much lower than that predicted on the basis of the analytical data (MnM2 ~ 0.74tot – 0.02A – 0.00M1 ~ 0.72 apfu). The model proposed appears in fact to underestimate the amount of manganese entering the M2 site, especially in the case of ilvaites with extremely high Mn contents, such as the Maruyama crystal (0.41 instead of 0.50 as proposed by Takéuchi et al. 1994). On the basis of the fact that , independently of the total Mn content, is quite invariable in most of the ilvaite structures previously published, we can assume that only negligible amounts of Mn replace Fe at M1. Therefore, the structural data available in the literature (references in Fig. 1) were used to calculate a new regression line assuming MnM2 = Mnoct = Mntot – MnA. As discussed below, the crystals from the Tochibora skarn (Takéuchi et al. 1993) likely exhibit a more disordered Mn distribution, so they were not included in the model. The following linear equation was obtained: = 2.1869(3) + 0.0266(9) Mn (apfu) (r = 0.992). The intercept ( = 2.187 Å) is identical to the value resulting from the model proposed by Carrozzini (1994), whereas the theoretical value for differs slightly in the two models (2.214 instead of 2.221 Å). It is worth noting that the grand mean distances (2.080 Å for all the crystals) for the ilvaite material from Oridda, for which a disordered model had been proposed (Carrozzini 1994), agrees perfectly with the values found for M1 sites randomly occupied by Fe and Fe alone. On the contrary, the crystals from Tochibora skarn examined by Takéuchi et al. (1993) exhibit slightly greater values for . According to the new model, the predicted MnM2 content for the ilvaites from Oridda (0.45, 0.53, and 0.53 for the three crystals studied, respectively) is consistent with an intermediate solid solution between ilvaite and a hypothetical CaFeMnFe(Si2O7)O(OH) end-member. In the sample from Perda Niedda, solid solution extends toward a higher manganoan component (up to 70%). As previously reported by Carrozzini (1994), a linear relationship can be observed between the Mn content and the b parameter, while a and c are not as markedly affected by the incorporation of this element (Fig. 2). The reason is that the replacement of Fe by Mn in the distorted (2 + 4) M2-octahedron causes a significant lengthening of the shortest M2-O1 and M2-O6 distances, which are both directed along the b axis. By plotting b against the octahedral Mn content a linear trend was obtained for data reported here and from the literature, with the exception of the synthetic crystals examined by GhaziBayat et al. (1989) and the crystals from Tochibora skarn studied by Takéuchi et al. (1993). For the synthetic crystals, a possible reason for this feature could be the disordered distribution of Mn between M11 and M2 inferred on the basis of Mössbauer experiments (Ghazi-Bayat et al. 1989). A minor Mn disordering could be hypothesized for the Tochibora crystals as well. The resulting regression line, b = 8.799(2) + 0.081(5) Mn(apfu) (r = 0.967), agrees fairly well with the model [b = 8.799(3) + 0.092(6) Mn(apfu)] previously proposed by Carrozzini (1994). TABLE 3. Experimental details of intensity data collections and structure refinements PN6-RT PN6-400 PN6-500 PN6-600 PN6-650 PN6-675 PN6-690 space group Pnam Pnam Pnam Pnam Pnam Pnam Pnam scan mode w w w w w w w scan width (∞/min) 2.2 2.2 2.2 2.2 2.5 2.5 2.5 scan speed (∞/min) 4.12 4.12 4.12 4.12 2.75 2.75 2.06 no.coll.refl. 6390 6392 6401 6372 3346 3333 3346 no.ind.refl. 1598 160