Occurrence and crystal chemistry of austinite, conichalcite and zincolivenite from the Peloritani Mountains, northeastern Sicily, Italy

Abstract

A new occurrence of austinite, CaZnAsO₄(OH), conichalcite, CaCuAsO₄(OH), and zincolivenite, CuZnAsO₄(OH), is described from the Tripi mine, Peloritani Mountains, Sicily, Italy. These species have been observed in euhedral crystals in vugs of a calcite vein and were characterised using single-crystal X-ray diffraction, electron microprobe analysis and micro-Raman spectroscopy. Austinite and conichalcite have isotypic relations, both crystallising in space group P2₁2₁2₁. Unit-cell parameters of austinite are a = 7.4931(5), b = 9.0256(6), c = 5.9155(4) Å, V = 400.06(5) ų; its crystal structure was refined on the basis of 1210 unique reflections with F_o > 4σ(F_o) and 77 least-square parameters to R₁ = 0.0236. Conichalcite has unit-cell parameters a = 7.419(10), b = 9.111(11), c = 5.867(7) Å and V = 396.6(1.4) ų; the diffraction quality of its available grains was not good enough to allow a high-quality structural refinement. Chemical formulae of austinite and conichalcite are Ca_1.04(1)Zn_0.86(4)Cu_0.09(4)As_0.98(2)P_0.02(1)O₄(OH)_0.98 and Ca_0.98(1)Fe²⁺_0.02(4)Cu_0.69(10)Zn_0.30(6)As_0.97(2)P_0.03(1)O₄(OH)_0.98, respectively. The new chemical data on the austinite–conichalcite isotypic pair, coupled with previous analyses, supports a possible miscibility gap between the compositions (Zn_0.25Cu_0.75) and (Zn_0.50Cu_0.50). Zincolivenite has unit-cell parameters a = 8.4594(9), b = 8.5324(8), c = 5.9893(6) Å, V = 432.30(12) ų and space group Pnnm; its crystal structure was refined to R₁ = 0.0230 for 523 unique reflections with F_o > 4σ(F_o) and 47 least-square parameters. Its chemical composition is Cu_0.73(5)Zn_1.25(5)As_1.01(1)O₄(OH)_1.01. The refinement of the crystal structure supports the ordering of Cu and Zn in two different crystallographic sites. Micro-Raman spectra of austinite, conichalcite and zincolivenite are discussed, with a focus on the O–H stretching region where local Zn and Cu arrangements affect the position of Raman bands in zincolivenite. These arsenates probably play an environmental role in the Peloritani area, where the occurrence of high contents of some potentially toxic elements in soils and stream sediments has been reported.

Introduction

Arsenic (Z = 33) is considered one of the most worrying of the potential toxic elements (PTE), considering its prevalence, toxicity and potential for human exposure (e.g. Mitchell, 2014). Its possible dispersion in the environment and, ultimately, its bioavailability can be related to weathering processes affecting ore deposits, where As is widespread in several primary minerals (e.g. As-bearing pyrite, arsenopyrite and As-bearing sulfosalts). However, in order to understand its potential mobility, the occurrence of several attenuation processes, mainly related to the precipitation of As-minerals (both in oxidising or reducing environments) and the adsorption onto some minerals (e.g. Fe oxy-hydroxides) (e.g. Gieré et al., 2003; Drahota and Filippi, 2009; Craw and Bowell, 2014), has to be taken into account. For these reasons, the mineralogical investigation of arsenate-bearing assemblages is crucial in assessing the fate of As in the environment.

Several researchers have pointed out the occurrence of high contents of some PTE, including As, in an area of the Peloritani Mountains (northeastern Sicily, Italy) where several small base-metal ore deposits are known (e.g. Saccà et al., 2015 and references therein). For instance, De Vivo et al. (1993) stressed the presence of anomalous contents of As, Hg, Sb, Zn and Pb in stream sediments from this area, whereas Dongarrà et al. (2009) found high As concentrations in waters from mineralised areas. A geogenic source for the As contamination detected in the Peloritani Mountains was also suggested by Cangemi et al. (2021), agreeing with previous results of Cosenza et al. (2015) who reported high PTE content in soils from Fiumedinisi, Budali and Alì, related to the weathering of ore deposits and mine wastes.

In this respect, a better knowledge of the mineralogy of these deposits is mandatory. Among primary minerals, arsenopyrite and gersdorffite are known to occur (Donati et al., 1978; Triscari and Saccà, 1982; Triscari, 1985; Mauro et al., 2021) and minor As occurs in Pb/Sb sulfosalts (e.g. Triscari and Saccà, 1984) and tetrahedrite-group minerals (Mauro et al., 2021). The alteration of these species gives rise to the occurrence of supergene arsenates. La Valle (1898) reported the identification of annabergite, erythrite and possibly symplesite. Scorodite was described by Triscari and Saccà (1982). However, crystal chemical data for these arsenates are lacking and no further details on supergene assemblages was given. Recently, Mauro et al. (2022) investigated a supergene mineral belonging to the alunite supergroup from the Tripi mine (Alì, Messina Province), identifying it as the lead-iron phosphate kintoreite. The Tripi mine (latitude 38°00′50″N; 15°24′27″E), located some hundred metres SSW of the small hamlet of Alì, exploited a stratabound fine-banded ore, mainly composed of sphalerite and Ag-bearing galena, with rare pyrite and trace amounts of chalcopyrite, arsenopyrite, pyrrhotite and covellite; quartz and fluorite are gangue minerals (Saccà et al., 2007). Supergene products are smithsonite, gypsum, hydrozincite, ‘limonite’ and malachite (Saccà et al., 2007). Additional investigations of a suite of specimens sampled by one of us (DM) in the dumps of this abandoned mine allowed the identification of an interesting association of three distinct Zn-bearing arsenates, i.e. austinite, conichalcite and zincolivenite. To the best of our knowledge, these species have not been reported previously from the Sicilian ore deposits and, as regards austinite, this is the first Italian occurrence.

The aim of this paper is the crystal chemical characterisation of these three arsenates, improving the available crystallographic and spectroscopic data for these species.

Experimental

Specimen studied

The specimen studied is composed of a calcite vein with deeply oxidised pyrite crystals; in small vugs of the calcite vein, colourless to bright-green crystals can be observed (Fig. 1). Three distinct kinds of crystals were identified: Type 1, colourless bladed crystals, forming aggregates of transparent, sub-adamantine individuals, up to 1 mm in length; Type 2, bright green tabular crystals, with a vitreous lustre, up to 1 mm in size; and Type 3, pale green prismatic crystals, up to 0.2 mm, with a sub-vitreous lustre.

Figure 1. Arsenate minerals from the Tripi mine. (a) Colourless prismatic crystals of austinite associated with green conichalcite. (b) Prismatic crystals of zincolivenite. Specimens in private collection.

Chemical, spectroscopic and structural data on these three different crystal types were collected and allowed the identification of austinite (Type 1), conichalcite (Type 2) and zincolivenite (Type 3), respectively (see below).

Chemical analysis

Quantitative chemical analyses were performed at the National Museum of Prague (Czech Republic) using a Cameca SX 100 electron microprobe using the following analytical conditions: wavelength dispersive spectroscopy mode, accelerating voltage of 15 kV, beam current = 8 nA and beam size = 5 μm. The following standards (element, emission line) were used: albite (NaKα), baryte (BaLα), Bi (BiMα), celestine (SKα and SrLβ), chalcopyrite (CuKα), clinoclase (AsLα), Co (CoKα), diopside (MgKα), fluorapatite (CaKα and PKα), halite (ClKα) hematite (FeKα), LiF (FKα), rhodonite (MnKα), sanidine (AlKα, KKα and SiKα), vanadinite (PbMα and VKα) and ZnO (ZnKα). Contents of the above-listed elements not included in Table 1 were analysed quantitatively, but were consistently below the detection limit (ca. 0.03–0.08 wt.% for individual elements). Peak counting times (CT) were 20 s; CT for each background was one-half of the peak time. Matrix correction by PAP software (Pouchou and Pichoir, 1985) was applied to the data. Results of the electron microprobe analysis of the studied grains are given in Table 1.

Table 1. Electron microprobe data for arsenates from the Tripi mine.

n = number of spot analyses. *Calculated in agreement with stoichiometry. σ = estimated standard deviation; – = not detected

Micro-Raman spectroscopy

Micro-Raman spectra of the minerals studied were collected using a Horiba Jobin-Yvon XploRA Plus apparatus, equipped with a motorised x–y stage and an Olympus BX41 microscope with a 50× objective (Dipartimento di Scienze della Terra, Università di Pisa). Raman spectra were excited using a 532 nm line of a solid-state laser, attenuated to 25% (i.e. 6.25 mW) in order to avoid any potential sample damage. The minimum lateral and depth resolution was set to a few μm. The system was calibrated using the 520.6 cm^–1 Raman band of silicon before each experimental session. Spectra were collected through multiple acquisitions (3) with variable counting times, ranging from 30 to 180 s. Back-scattered radiation was analysed with a 1200 gr/mm grating monochromator. Experimental precision can be estimated at ± 2 cm^–1. Whenever necessary, Raman spectra were processed using Fityk (Wojdyr, 2010), subtracting the background and fitting the spectra to theoretical peak shapes using pseudo-Voigt functions.

X-ray crystallography

Single-crystal X-ray diffraction data were collected for austinite, conichalcite and zincolivenite using a Bruker D8 Venture diffractometer (50 kV and 1.4 mA) equipped with an air-cooled Photon III detector and microfocus MoKα radiation (Centro per l'Integrazione della Strumentazione scientifica dell'Università di Pisa, Università di Pisa, Italy). For all the three crystals studied, the detector-to-crystal distance was set to 38 mm. The diffraction quality was good enough for crystals of austinite and zincolivenite, whereas no grains of conichalcite suitable for high-quality intensity data collection were found.

Data were collected using φ and ω scan modes, in 0.5° slices. Intensity data were integrated and corrected for Lorentz, polarisation, background effects and absorption using the APEX4 (Bruker AXS Inc., 2022) software package. Crystal structure refinements were performed using Shelxl–2018 (Sheldrick, 2015) and neutral scattering curves were taken from the International Tables for Crystallography (Wilson, 1992). Details of data collections and structure refinements are given in Table 2. Atomic coordinates and equivalent isotropic or isotropic displacement parameters are shown in Table 3, whereas selected bond distances are reported in Table 4. Bond-valence sums, calculated using the bond parameters of Gagné and Hawthorne (2015), are given in Table 5. Crystallographic information files of austinite and zincolivenite have been deposited with the Principal Editor of Mineralogical Magazine and are available as Supplementary material (see below). In the following, details of the structure refinements are briefly reported.

Table 2. Summary of crystal data and parameters describing data collections and refinements for austinite and zincolivenite from the Tripi mine.

¹ Austinite: w = 1/[σ²(F _o²)+(0.0263P)²+0.7514P], where P = (F _o²+2F _c²)/3

² Zincolivenite: w = 1/[σ²(F _o²) +2.3444P], where P = (F _o²+2F _c²)/3

Table 3. Sites, site occupancy factors*, fractional atom coordinates and isotropic (**) or equivalent isotropic displacement parameters (Ų) for austinite and zincolivenite from the Tripi mine.

* Site occupancy: M = Zn_1.00

Table 4. Selected bond distances (Å) and angles (°) for austinite and zincolivenite from Tripi mine.

Table 5. Weighted bond-valence sums (in valence units) for austinite and zincolivenite from the Tripi mine.

Notes: left and right superscripts indicate the number of equivalent bonds involving anions and cations, respectively. The following site occupancies were used for calculating the bond-valence sums: in austinite, Ca = Ca_1.00, Zn = Zn_0.90Cu_0.10 and As = As_0.98P_0.02; in zincolivenite, Cu(1) = Cu_0.75Zn_0.25, Zn(1) = Zn_1.00 and As(1) = As_1.00.

* Corrected for H bonds.

Austinite

A total of 1621 frames was collected, with exposure time of 10 s per frame. Unit-cell parameters are a = 7.4931(5), b = 9.0256(6), c = 5.9155(4) Å, V = 400.06(5) ų, space group P2₁2₁2₁, in agreement with systematic absences and the |E ²–1| value (i.e. 0.684). The crystal structure was refined starting from the atomic coordinates given by Giuseppetti and Tadini (1988) for austinite. The following scattering curves were used: Ca at the Ca site, As at the As site, O at the O(1)–O(5) sites, and H at the H site. Taking into account the similarity in the scattering factors of Cu (Z = 29) and Zn (Z = 30), and the results of electron microprobe analysis, showing a Zn/(Zn + Cu) atomic ratio of 0.90(4), only the scattering factor of Zn was used to model the site scattering of the M site. After several cycles of isotropic refinement, the R ₁ factor converged to 0.0336, thus confirming the correctness of the structural model. A restraint on the O(5)–H bond distance was added and the displacement parameter of H was refined assuming it is 1.5 times that of U _eq of O(5). Racemic twinning was modelled, converging to a Flack parameter (Flack, 1983) of 0.17(2). After the introduction of the anisotropic displacement parameters for all atoms (but H), the R ₁ decreased to 0.0236 for 1210 unique reflections with F _o > 4σ(F _o) and 77 least-square parameters.

Conichalcite

Several grains of conichalcite were checked and all of them were polycrystalline, showing powder-like diffraction patterns. However, using a 30 × 30 × 30 μm grain, the refinement of the unit-cell parameters of one of the individuals forming the polycrystalline aggregate was possible, selecting the sharp spots superimposed on diffraction rings. On the basis of 176 reflections in the 2θ range between 7.08 and 39.37°, the following unit-cell parameters were refined: a = 7.419(10), b = 9.111(11), c = 5.867(7) Å and V = 396.6(1.4) ų. Unfortunately, the intensity data had a very high R _int value (0.238), and the crystal structure refinement converged to a high R ₁ value of 0.1101 for 342 reflections with F _o > 4σ(F _o) and 49 least-square parameters (0.2334 for all the 871 measured reflections). For this reason, structural data of conichalcite are not discussed, and only electron microprobe data are presented.

Zincolivenite

A total of 1368 frames was collected, with exposure time of 25 s per frame. Unit-cell parameters are a = 8.4594(9), b = 8.5324(8), c = 5.9893(6) Å, V = 432.30(12) ų and space group Pnnm, in accord with systematic absences and the |E ²–1| value (i.e. 0.861). The crystal structure was refined starting from the atomic coordinates given by Chukanov et al. (2007) for zincolivenite. The following scattering curves were used: As, Cu and Zn at the As(1), Cu(1) and Zn(1) sites, and O at the O(1)–O(4) sites. Owing to the similar scattering factors of Cu (Z = 29) and Zn (Z = 30), the site occupancy of the Cu(1) and Zn(1) sites was refined assuming a pure occupancy. After several cycles of isotropic refinement, the R ₁ factor converged to 0.0520, confirming the correctness of the structural model. Introducing the anisotropic displacement parameters for all the atoms decreased the R ₁ to 0.0234. The site hosting the H atom was not correctly located in the difference-Fourier maps and it was added on the basis of the atom coordinates given by Chukanov et al. (2007). A restraint on the O(2)–H bond distance was added and the displacement parameter of the H atom was restrained to be 1.5 times that of O(2). The refinement converged to R ₁ = 0.0230 for 523 unique reflections with F _o > 4σ(F _o) and 47 least-square parameters.

Results and discussion

Chemical data

Chemical formulae of austinite, conichalcite and zincolivenite were recalculated on the basis of 3 cations per formula unit (pfu), assuming the presence of the amount of H₂O required to achieve the electrostatic balance.

The total of conichalcite is low, even after the addition of calculated H₂O; this is probably due to the small size of the polished grain embedded in epoxy. Such low totals were observed previously by other authors during the examination of this species (e.g. Braithwaite et al., 2009). However, atom ratios agree with the ideal formula of this mineral.

Austinite and conichalcite and their solid-solution series

The chemical formula of austinite from the Tripi mine is Ca_1.04(1)Zn_0.86(4)Cu_0.09(4)As_0.98(2)P_0.02(1)O₄(OH)_0.98, whereas that of conichalcite is Ca_0.98(1)Fe²⁺_0.02(4)Cu_0.69(10)Zn_0.30(6)As_0.97(2)P_0.03(1)O₄(OH)_0.98. With respect to the ideal formula, austinite has minor Cu-to-Zn and P-to-As substitutions, and a slight Ca excess; on the contrary, in conichalcite the replacement of Zn by Cu is significant. Minor Fe, assumed as Fe²⁺, may partially replace Ca²⁺ or Zn²⁺. However, owing to the oxidised nature of the assemblage studied, the occurrence of Fe³⁺ cannot be discarded. The relationship between the Zn and Cu contents, in atoms pfu, in austinite and conichalcite is shown in Fig. 2. Austinite has a Cu/(Cu + Zn) atomic ratio ranging between 0.03 and 0.23, whereas conichalcite from Tripi has the same ratio in the range 0.58–0.77.

Figure 2. Relationship between Cu and Zn (in atoms per formula unit) in austinite (grey diamonds), conichalcite (green diamonds) and zincolivenite (light green diamonds) from the Tripi mine. Red circles indicate compositional data for austinite and conichalcite after Staples (1935), Radcliffe and Simmons (1971), Gunter (1977), Jambor et al. (1980), Giuseppetti and Tadini (1988), Henderson et al. (2008), Sakai et al. (2009) and Ðorđević et al. (2016).

Further analytical data from literature for the isotypic series between austinite and conichalcite analysis is also plotted in Fig. 2. Type material of the former was Cu free (Staples, 1935) and other occurrences of austinite are usually Cu poor. For instance, Giuseppetti and Tadini (1988) reported the replacement of Zn by 0.03 Cu atoms and 0.01 Fe atoms pfu for a sample from Kamareza (Greece), whereas Clark et al. (1997) gave the composition Ca(Zn_0.95Cu_0.03Mg_0.02)AsO₄(OH) for a sample from Sonora (Mexico). Gunter (1977) provided several analyses of members of the austinite–conichalcite series, with Cu/(Cu + Zn) atomic ratios in the range 0.47–1.00. Lead-bearing conichalcite described by Jambor et al. (1980) has a Cu/(Cu + Zn) ratio in the interval 0.69–0.91, and similarly Taggart and Foord (1980), on the basis of electron dispersive spectroscopy data, found ratios of 0.11 for Cu-rich austinite and 0.60 for conichalcite. Henderson et al. (2008) refined a sample of conichalcite from the RRUFF project with only 0.01 Zn pfu. Other specimens of conichalcite from RRUFF have up to 0.30 Zn apfu (see http://rruff.info – Lafuente et al., 2015). Conichalcite studied by Sakai et al. (2009) has no Zn, and Cu was partly replaced by very small amounts of Mg (up to 0.13 atoms pfu in a sample from Fukushima, Japan). Finally, conichalcite studied by Ðorđević et al. (2016) has a Cu/(Cu + Zn) ratio of 0.97.

It is interesting to note that type material of the Pb-analogue of austinite, arsendescloizite PbZn(AsO₄)(OH), contains no Cu (Keller and Dunn, 1982), whereas the P-analogue of conichalcite, hermannroseite, ideally CaCu(PO₄)(OH), has a very small Zn content (Schlüter et al., 2011).

Notwithstanding the hypothesis of Radcliffe and Simmons (1971) who suggested a complete solid solution between austinite and conichalcite, the evidence is meagre. Indeed, Fig. 2 suggests the possible existence of a miscibility gap in the compositional region having Cu/(Cu + Zn) in the range 0.25–0.50, i.e. (Zn_0.25Cu_0.75)–(Zn_0.50Cu_0.50). This could be due to the Jahn-Teller effect shown by Cu²⁺ (Burns and Hawthorne, 1996); indeed, it is well-known that Cu²⁺ and Zn often tend to be partitioned in different structural sites, characterised by a different degree of distortion, e.g. in paratacamite and herbertsmithite (Braithwaite et al., 2004), in rosasite (Perchiazzi, 2006), in ktenasite (Mellini and Merlino, 1978) and in zincolivenite (Chukanov et al., 2007). The finding of both austinite and conichalcite in the same small vugs of a calcite vein from the Tripi mine is in keeping with this hypothesis; indeed, two distinct phases crystallised instead of a single phase with intermediate composition.

Zincolivenite

The chemical formula of zincolivenite from the Tripi mine is Cu_0.73(5)Zn_1.25(5)As_1.01(1)O₄(OH)_1.01. The Cu/(Cu + Zn) atomic ratio ranges between 0.33 and 0.41 (Fig. 2). This occurrence is richer in Zn than type material from the Lavrion deposit (Greece), having composition Cu_0.94Zn_1.03Fe_0.02[(AsO₄)_0.98(PO₄)_0.02](OH)_0.98(H₂O)_0.10, it is close to a sample from Dal'negorsk (Russia), with composition Cu_0.58Mg_0.07Zn_1.36(AsO₄)(OH)_0.98O_0.02 (Chukanov et al., 2007). According to these authors, the chemical composition of zincolivenite should span the range between Cu_0.5Zn_1.5(AsO₄)(OH) and Cu_1.5Zn_0.5(AsO₄)(OH). Zinc-rich zincolivenite from the Tripi mine also has a composition similar to the specimen reported by Šejkora et al. (2008) from Krupka, northern Bohemia (Czech Republic), with composition (Zn_1.27Cu_0.62Al_0.08Pb_0.01Ca_0.01)[(AsO₄)_0.98(PO₄)_0.01(SO₄)_0.01](OH)_1.09. Other quantitative data on zincolivenite are given by Gołębiowska et al. (2006), Braithwaite et al. (2009) and Andersen and Moulding (2009).

Micro-Raman spectroscopy

Austinite and conichalcite

The Raman spectra of austinite and conichalcite are shown in Fig. 3. The observed bands are at similar positions and only minor differences, in the lower wavenumber region, were observed.

Figure 3. Raman spectra of austinite (a,b), conichalcite (c,d) and zincolivenite (e,f) in the region between 100–1200 cm^–1 and 3000–4000 cm^–1.

In the region between 100 and 1200 cm^–1, the strongest Raman band occurs at 828 and 831 cm^–1 for austinite and conichalcite, respectively, with two additional weaker bands at 784 and 805 cm^–1 in both Raman spectra. These bands can be attributed to the ν₃ and ν₁ modes of AsO₄ groups, in agreement with Reddy et al. (2005) and Ðorđević et al. (2016). It is worth noting that, in this spectral region, Martens et al. (2003b) also reported strong bands at 859 and 889 cm^–1. However these bands have not been observed in the sample from the Tripi mine or in isotypic phases like cobaltaustinite CaCo(AsO₄)(OH) and nickelaustinite CaNi(AsO₄)(OH). Finally, the occurrence of these bands do not agree with the infrared spectral features described by Sumin de Portilla (1974). In conclusion, the spectrum of austinite reported by Martens et al. (2003b) is not representative of this species. In the sample from the Tripi mine, two very weak bands occur at 920 and 975 cm^–1 and at 924 and 975 cm^–1 in the Raman spectra of austinite and conichalcite, respectively, and they could be attributed to the O–H deformation modes (e.g. Martens et al., 2003b; Reddy et al., 2005).

In the spectrum of austinite, bands at 381, 430 and 466 cm^–1 can be related to the ν₄ bending modes of AsO₄ groups; these bands were observed at 394, 424 and 469 cm^–1 in the spectrum of conichalcite. Weak bands at 502 cm^–1 for austinite and at 502 and 518 cm^–1 in conichalcite may be interpreted as due to ν₄ modes, although these band could be due to M–O stretching vibrations, in agreement with Martens et al. (2003b) who proposed such a possible interpretation for bands at 487 and 495 cm^–1 observed in cobaltaustinite and nickelaustinite, respectively. The ν₂ bending modes of AsO₄ groups is possibly related to the bands at 328 and 347 cm^–1 in the spectrum of austinite and at 330 and 353 cm^–1 in that of conichalcite. Raman bands at lower wavenumbers, i.e., 134, 173, 216 and 281 cm^–1 in austinite and 144, 188, 218 and 283 cm^–1 in conichalcite, can be interpreted either as lattice modes or M–O vibrations.

In the region between 3000 and 4000 cm^–1, a relatively strong band at 3265 cm^–1 was observed in both austinite and conichalcite, with a shoulder at 3330 cm^–1. These bands are related to the O–H stretching modes. Whereas in the olivenite–adamite series the band shift in the O–H stretching region can allow prediction of the local cation arrangements around the (OH) group (see below), this seems to be not applicable to the austinite–conichalcite isotypic pair. In these compounds, the (OH) group is bonded to one Ca and two M cations (M = Zn or Cu in austinite and conichalcite, respectively). It is worth noting that the M–(OH) bonds are the shortest ones in the M-centred octahedra; the replacement of Zn²⁺ by Cu²⁺, the latter being responsible for the Jahn-Teller effect distorting the coordination polyhedron (Burns and Hawthorne, 1996), did not significantly affect such short distances. For this reason, the O⋅⋅⋅O distance of the H-bonds in austinite and conichalcite may be not so different to give rise to significant Raman shifts, i.e. 2.721–2.723 Å in austinite (Giuseppetti and Tadini, 1988; Clark et al., 1997) and 2.741 Å in conichalcite (Sakai et al., 2009). However, Henderson et al. (2008) reported a shorter O⋅⋅⋅O distance for conichalcite, i.e. 2.678 Å, stating that, in agreement with the relations between O⋅⋅⋅O and the wavenumber of O–H stretching modes (Libowitzky, 1999), conichalcite should show the smallest O–H wavenumber in the adelite series. In this respect, they cited the band at 3158 cm^–1 observed by Martens et al. (2003b); a similar band position was reported by Reddy et al. (2005) at 3148 cm^–1. This band position corresponds, according to Libowitzky (1999), to an O⋅⋅⋅O distance of ca. 2.69 Å, in agreement with the distance observed by Henderson et al. (2008). This seems to suggest that a systematic study of the spectral features of the O–H stretching region on well-characterised specimens of the austinite–conichalcite isotypic series is mandatory to understand the actual relations between their crystal chemistry and band positions.

Zincolivenite

The Raman spectrum of zincolivenite is shown in Fig. 3. The region between 100 and 1200 cm^–1 is characterised by two strong bands at 850 and 876 cm^–1, with a weak shoulder at 924 cm^–1.The strong bands can be interpreted as due to the ν₃ and ν₁ stretching modes of AsO₄ groups, in agreement with Martens et al. (2003a) and Makreski et al. (2013). The weak shoulder at 924 cm^–1 could be due to O–H deformation modes, as in the austinite–conichalcite pair (see above).

Bands at 487, 502 and 520 cm^–1 can be interpreted as due to the ν₄ bending modes of AsO₄ groups, whereas the ν₂ modes can be represented by the band at 414 cm^–1. Makreski et al. (2013) interpreted the bands at 325 and 380 cm^–1 in the Raman spectrum of adamite as due to the vibrations of M²⁺O₅ unit; in agreement with this interpretation, a similar attribution can be proposed for the bands at 320 and 369 cm^–1 observed in zincolivenite from the Tripi mine. Finally, bands at 135, 175, 202, 222 and 274 cm^–1 are probably due to M–O modes and lattice vibrations.

In the 3000–4000 cm^–1 region, three sharp bands were observed at 3492, 3514 and 3540 cm^–1. These three bands can be interpreted as due to O–H stretching modes and they can be related to different local Cu–Zn distributions, in agreement with the interpretation of Chisholm (1985), based on previous results of Braithwaite (1983). Taking into account the electron microprobe data of zincolivenite from the Tripi mine, the site occupancy (Cu_0.75Zn_0.25) and Zn_1.00 can be proposed for the Cu(1) and Zn(1) sites, respectively. As the (OH) group is bonded to one Zn(1) and two Cu(1) sites, the following cation distributions around the (OH) group can be observed [elements are listed according to the occupancy of sites Zn(1)Cu(1)Cu(1)]: (i) ZnCuCu; (ii) ZnCuZn; and (iii) ZnZnZn. In agreement with Chisholm (1985), these three different local arrangements should correspond to bands at 3490, 3510 and 3540 cm^–1, comparable with the values observed in the Sicilian sample studied in this work. Moreover, considering the relative proportions of Zn and Cu, configuration (i) is more abundant than configurations (ii) and (iii). Through the fitting of the O–H stretching bands using Fityk (Wojdyr, 2010), a ratio 75:17:8 for the three possible arrangements has been obtained. This ratio can be taken into account qualitatively to indicate that ZnCuCu is the most probable configuration, followed by the configurations ZnCuZn and ZnZnZn, in agreement with the electron microprobe data and the Cu and Zn preferential partitioning at Cu(1) and Zn(1) sites, respectively.

Crystal structure description

Austinite and conichalcite

Austinite and conichalcite belong to the adelite group within the descloizite supergroup. In agreement with previous authors (Qurashi and Barnes, 1963; Giuseppetti and Tadini, 1988; Clark et al., 1997; Henderson et al., 2008; Sakai et al., 2009; Ðorđević et al., 2016), their crystal structure can be described as formed by columns of edge-sharing M-centred octahedra running along c; every chain is connected to four adjacent octahedral chains through corner-sharing As-centred tetrahedra. In the cavities resulting from this heteropolyhedral framework, eight-fold coordinated Ca²⁺ cations are hosted (Fig. 4). Following Eby and Hawthorne (1993), the structure of the isotypic pair austinite–conichalcite could be described as a M = M–T framework, i.e. edge-sharing octahedral chains linked by tetrahedral corners.

Figure 4. Crystal structures of austinite (a) and zincolivenite (b) seen down c. Arsenic-, Cu- and Zn-centred polyhedra are shown in violet, blue and grey, respectively. Circles represent Ca (light blue), O (red) and H (pink) atoms. Thick red lines and dashed black lines are, respectively, O–H and H⋅⋅⋅O bonds. Unit cells are shown as blue dotted lines. Drawn with CrystalMaker ^® software.

In austinite, M is zinc. This element is octahedrally coordinated by four O^2– and two (OH)^– groups. Bond distances range between 1.936 and 2.234 Å, with average <Zn–O> distance of 2.101 Å, to be compared with 2.104 and 2.106 Å observed by Giuseppetti and Tadini (1988) and Clark et al. (1997), respectively. The two shortest Zn–O distances are formed with (OH) groups (1.936 and 1.951 Å, respectively), in agreement with previous studies (Giuseppetti and Tadini, 1988; Clark et al., 1997). The difference between the longest and shortest Zn–O distance is 0.297 Å, to be compared with 0.311 and 0.300 Å given by Giuseppetti and Tadini (1988) and Clark et al. (1997); this difference is increased in conichalcite, e.g. 0.476, 0.470 and 0.503–0.504 Å for conichalcite from Arizona, Japan and Chile, respectively (Henderson et al., 2008; Sakai et al., 2009; Ðorđević et al., 2016). This is due to the Jahn-Teller effect shown by Cu²⁺ (Burns and Hawthorne, 1996). Along the columns, Zn–Zn distance is 2.967 Å, close to 2.970 Å reported for the sample from Kamareza studied by Giuseppetti and Tadini (1988), whereas it is shorter than that observed by Clark et al. (1997) for the sample from Sonora, i.e. 2.977 Å. This is probably due to the minor replacement of Zn by Cu observed in the sample from Tripi. In conichalcite, indeed, the Cu–Cu distance along the c direction is 2.92–2.93 Å (Qurashi and Barnes, 1963; Sakai et al., 2009). According to Clark et al. (1997), the value observed by Giuseppetti and Tadini (1988) could indicate some Cu substitution, in agreement with data of austinite from the Tripi mine. The bond-valence sum at the Zn site, calculated assuming the site population (Zn_0.90Cu_0.10), is 2.11 valence units (vu) (Table 5).

Arsenic is tetrahedrally coordinated, with As–O bonds ranging between 1.681 and 1.697 Å. The average <As–O> bond distance is 1.688 Å, comparable to 1.692 and 1.694 Å of Giuseppetti and Tadini (1988) and Clark et al. (1997), respectively, and fitting with the grand value of 1.688 Å given by Majzlan et al. (2014). The bond-valence sum at this site, calculated with the site occupancy (As_0.98P_0.02), is 4.94 vu (Table 5), in agreement with the occurrence of As⁵⁺.

Calcium is coordinated by seven O^2- and one (OH)^- group in a distorted square antiprism. Calcium–O distances range between 2.309 and 2.657 Å, the shortest distance being represented by the Ca–(OH) bond. These distances can be compared with the results of Giuseppetti and Tadini (1988) and Clark et al. (1997), who reported Ca–O distances in the range 2.315–2.656 Å and 2.313–2.661 Å, respectively. The CaO₇(OH) polyhedron shares edges with two AsO₄ tetrahedra and three edges with Zn-centred octahedra; the edges shared with AsO₄ groups are the shortest ones in the Ca-centred polyhedron. The bond-valence sum at the Ca site is 1.85 vu (Table 5).

Oxygen atoms are three- or four-fold coordinated by Ca, Zn and As atoms. Those atoms hosted at O(1), O(2) and O(4) sites are bonded to these three cations, achieving bond-valence sums ranging between 1.87 and 2.02 vu, in agreement with the occurrence of O^2- anions. On the contrary, oxygen atoms at O(2) and O(5) are three-fold coordinated. The former is bonded to two Ca and one As atom, with a bond-valence sum of 1.67 vu, whereas the latter is bonded to one Ca and two Zn atoms, with a bond-valence sum of 1.40 vu. The undersaturation of O(2) and O(5) indicates their involvement in the O(5)–H⋅⋅⋅O(2) hydrogen bond. The O(5)⋅⋅⋅O(2) distance is 2.719(5) Å, to be compared with 2.721 and 2.723 Å given by Giuseppetti and Tadini (1988) and Clark et al. (1997); using relationships between bond distance and bond strength in H bonds from Ferraris and Ivaldi (1988), such a distance corresponds to a bond strength of 0.22 vu. These values agree with the O⋅⋅⋅O distances estimated on the basis of the wavenumbers of O–H stretching modes observed in the Raman spectrum of austinite following Libowitzky (1999), i.e. 2.726 and 2.755 Å, respectively.

Zincolivenite

Zincolivenite belongs to the olivenite group, a series of orthorhombic and monoclinic arsenates and phosphates whose crystal structures are related to that of andalusite, Al₂SiO₄O. The crystal structure of zincolivenite can be described as formed by chains of edge-sharing distorted Cu(1)O₄(OH)₂ octahedra running along c. These octahedral chains are connected by isolated As(1)O₄ tetrahedra through corner-sharing, giving rise to channels where dimers of edge-sharing Zn(1)O₄(OH) trigonal bipyramids are hosted.

Copper is in a distorted octahedral coordination. Bond distances range between 2.006 and 2.359 Å, with an average value of 2.124 Å. These values can be compared with those given by Chukanov et al. (2007) for type zincolivenite from Lavrion, Greece and by Toman (1978) for a sample from Tsumeb, Namibia. The former has an average <Cu–O> distance of 2.132 Å (range 1.986–2.409 Å) and the latter 2.12 Å (range 1.99–2.37 Å). It is interesting to observe that the difference between the longest and shortest Cu–O distances is 0.35, 0.38 and 0.42 Å for the samples from Tripi, Tsumeb, and Lavrion, respectively; this increasing difference is directly related to the Cu content. Indeed, the Cu/(Cu + Zn) atomic ratios of these three studied samples are 0.37, 0.42 and 0.48, respectively. The higher the Cu content, the higher is the distortion of the Cu-centred Cu(1) octahedron. This is in keeping with the Jahn-Teller effect shown by Cu²⁺ (Burns and Hawthorne, 1996). In the Tripi mine, the Cu(1) site is probably a mixed (Cu,Zn) site, with an idealised site occupancy (Cu_0.75Zn_0.25), based on electron microprobe data. The bond-valence sum at this site, calculated using this site occupancy, is 1.96 vu. (Table 5).

Zinc is hosted at the Zn(1) site. In zincolivenite from the Tripi mine, the average <Zn–O> is 2.030 Å, with bond distances ranging between 1.999 and 2.083 Å. These features are close to those observed in the Greek and Namibian specimens by Chukanov et al. (2007) and Tomin (1978), respectively. In the former, the average distance is 2.031 Å (range 1.998–2.074 Å), and in the latter 2.04 Å (range 2.00–2.08 Å). The bond-valence sum at the Zn(1) site is 2.03 vu. (Table 5).

Arsenic is tetrahedrally coordinated, with As–O bonds ranging between 1.679 and 1.694 Å. Average <As–O> bond distance is 1.684 Å, comparable to 1.690 and 1.689 Å of Chukanov et al. (2007) and Tomin (1978). The bond-valence sum is 5.04 vu. (Table 5).

Among the four symmetry-independent O atoms, three have bond-valence sums ranging between 1.88 and 2.04 vu, whereas the fourth, namely the oxygen atom at the O(2) site, shows a strong undersaturation, i.e. 1.19 vu. This latter atom is involved in a H bond with O(4), with the O(2)–H⋅⋅⋅O(4) distance of 2.883(4) Å, corresponding to a bond strength of 0.16 vu. (according to Ferraris and Ivaldi, 1988); taking into account the bifurcated nature of this H bond (as, for instance, in adamite or the related phosphate libethenite – e.g. Cordsen, 1978; Jinnouchi et al., 2016), the bond valence at the O(4) site is increased by 0.08 vu, up to 1.96 vu, in agreement with the occurrence of O^2– anions. The observed O⋅⋅⋅O distance agrees with the bond lengths calculated using the relation of Libowitzky (1999), i.e. between 2.88 and 2.97 Å.

Conclusions

The crystal chemical characterisation of austinite, conichalcite and zincolivenite is a further step in the knowledge of the supergene phases formed through the alteration of base-metal ore deposits of the Peloritani Mountains, after the study on the alunite-supergroup mineral kintoreite (Mauro et al., 2022). In particular, the identification of these three arsenates gives interesting insights for the speciation, transport and dispersion of As in this sector of Sicily, where, as discussed briefly in the Introduction, high contents of this toxic element (as well as other PTE) have been reported in stream sediments and soils since the beginning of the 1990s (e.g. De Vivo et al., 1993; Cosenza et al., 2015).

This study also provides some interesting crystal-chemical and physical data for the arsenates studied. Some unexpected spectral features described by Martens et al. (2003b) in the Raman spectrum of austinite (i.e. the occurrence of strong bands at 889 and 859 cm^–1) are not supported by the current study, where the same grain used for single-crystal X-ray diffraction study and electron microprobe analysis was used for the collection of micro-Raman spectra. Some discrepancies occur also in the O–H stretching region, suggesting the opportunity to perform a systematic study on well-characterised specimens of the austinite–conichalcite isotypic series to describe the actual relations between crystal-chemistry and the band position of the O–H stretching modes. Moreover, a critical examination of the austinite–conichalcite isotypic pair suggests the existence of a miscibility gap between these two species, in disagreement with the hypothesis of Radcliffe and Simmons (1971). This gap would not be surprising, considering the Jahn-Teller effect shown by Cu²⁺ cations and the Zn/Cu different partitioning observed in several mineral species. One of them is represented by zincolivenite. The refinement of the crystal structure of Zn-rich zincolivenite from the Tripi mine, with a Cu/(Cu + Zn) atomic ratio different from the samples studied previously by Toman (1978) and Chukanov et al. (2007), has shown that, between Cu(1) and Zn(1), only the former seems to be affected by geometrical changes related to the different chemistry. Majzlan et al. (2023) stated that anomalous X-ray diffraction would be required to prove cation ordering in zincolivenite. This is correct, but it should be noted that examination of the geometrical features of the Cu(1) and Zn(1) sites in the available refinements of zincolivenite supports the presence of Cu and Zn ordering; the study of other specimens, possibly Cu-rich, may be able to support or refute this statement.