Crystal Chemistry of an Erythrite-Köttigite Solid Solution (Co3–xZnx) (AsO4)2·8H2O

A wide compositional range, covering about 90% of an expected erythrite-köttigite substitutional solid solution with extreme compositions of (Co2.84Mg0.14Zn0.02) (AsO4)2·8H2O and (Zn2.74Co0.27) (AsO4)2·8H2O, was revealed in a suite of samples from a polymetallic ore deposit in Miedzianka, SW Poland. Members of the solid solution series were examined by means of Electron Probe Microanalysis (EPMA), Scanning Electron Microscopy (SEM)/Energy-Dispersive Spectrometer (EDS), X-ray single-crystal and powder diffraction, and Raman spectroscopy. Metal cations were randomly distributed between two special octahedral sites in the erythrite–köttigite structure. In response to Co ↔ Zn substitutions, small but significant changes in bond distances (particularly in [AsO4] tetrahedra), rotation, and distortion of co-ordination polyhedra were observed. Two sub-series of dominant cationic substitutions (Co-Mg-Ni and Co-Fe-Zn) were noted within the arsenate series of vivianite-group minerals linked by erythrite. The paragenetic sequence erythrite → Zn-rich erythrite → Co-rich köttigite → köttigite reflects the evolution of the solution’s pH towards increased acidity and a relative increase in the concentration of Zn ions following precipitation of erythrite.


Introduction
Erythrite and köttigite belong to the arsenate series of vivianite-group minerals with the general formula of M3(AsO4)2·8 H2O, where M represents divalent cations of Co (erythrite), Cu (babánekite), Fe (parasymplesite), Mg (hörnesite), Ni (annabergite), and Zn (köttigite). Minerals of the arsenate series occur in the oxide zones of numerous polymetallic deposits. Erythrite is by far the most common of these minerals.
Members of the arsenate series are isostructural (space group C2/m), as revealed by the structure refinement of annabergite [1] and its Mg-rich variety cabrerite [2] as well as babánekite [3], erythrite [1, 4,5], köttigite [6], köttigite-parasymplesite intermediate [7], and parasymplesite [8]. The latter was originally referred to as symplesite [8], which is at present the name for a triclinic dimorph of parasymplesite. Metals in these minerals occupy two types of octahedral sites: insular M1 and double edge-sharing M2 (Figure 1). Metals in the M1 octahedra are coordinated by two oxygens and four water molecules [M1O2(H2O)4], while those in the M2 octahedra are coordinated by four oxygens and two water molecules [M2O4(H2O)2]. The structure is layered parallel to the (010) mirror plane. The layers consist of octahedra linked by [AsO4] tetrahedra and hydrogen bonds. The adjacent layers are held together by a network of hydrogen bonds. According to some investigators [4,6], transition metals and Mg are randomly distributed over both octahedral sites. Others suggest that Co, Ni, and Zn cations tend to occupy larger M1 sites, whereas Mg and Fe 2+ prefer M2 octahedra [1,7]. While reports on the occurrence of limited solid solutions within the arsenate series minerals are frequent [5,[9][10][11][12], no complete or extended solid solutions have been observed as yet in natural samples. Experiments involving syntheses showed a complete solid solution for the erythriteannabergite binary [13][14][15] and a continuous solid solution for the erythrite-annabergite-hörnesite ternary up to the experimental limits [13].
No crystal chemical constraints hinder the unlimited solid solution between erythrite and köttigite. The similarity of the effective ionic radii ( [6] Co 2+ = 74.5 pm, [6] Zn 2+ = 74 pm [17]) enables easy CoZn-1 exchange in octahedral sites of the vivianite-type structure. Therefore, the limited compositional range between these two minerals reported in the literature (Table 1) may reflect an insufficient number of analysed samples.
In this paper, we report on the chemical composition, structure refinement, and results of Raman spectroscopy of an extended erythrite-köttigite [(Co1-xZnx)3(AsO4)2·8H2O] solid solution from a polymetallic ore deposit in Miedzianka, Sudetes Mts, SW Poland. It is not often that an extended solid solution occurs within a single locality. Miedzianka is unique in this respect, as two other extended solid solutions (77.5 mol% cornwallite-pseudomalachite and 75 mol% kipushite-philipsburgite) have been observed among base-metal arsenates and phosphates in samples from this locality [18].

Sampling Site
Thirty-seven samples for this study were collected from dumps around the historic mining town of Miedzianka (known as Kupferberg prior to 1945) (50°52′41.06″ N; 15°56′34.91″ E; 504 m above sea level) in the Sudetes Mts, SW Poland. The hydrothermal vein-type Miedzianka polymetallic deposit was mined for copper, silver, lead, arsenic, and pyrite from the beginning of the fourteenth century until 1925 and for uranium between 1948 and 1952. Miedzianka is a type locality for uranophane.
The oxidative supergene alteration zone in the Miedzianka ore deposit extends to a depth of 178 m below ground level [19]. Two principal secondary supergene assemblages have been distinguished [20]. The most common assemblage consists of abundant chrysocolla and Cu-arsenates ± malachite and Cu-phosphates. The second assemblage consists of Cu-sulphates, mostly brochantite, langite, and devilline. In total, 38 secondary supergene minerals have been observed in old dumps and in museum specimens [18][19][20][21]. Arsenates of the vivianite group are represented by abundant erythrite, less abundant köttigite, and rare annabergite.

Sample Description
The examined minerals occur as radial aggregates or crystal clusters, either directly on the altered (chloritised) amphibolite or on the chrysocolla coating of host rocks. Similar to the erythritehörnesite solid solution [9], the colour of members of the erythrite-köttigite series changes with increasing Zn content from dark pink or crimson erythrite to the pale pink and pinkish white

Scanning Electron Microscopy (SEM)
Mineral morphology and spatial relationships between observed minerals were examined using a Philips XL 30 ESEM/TMP scanning electron microscope coupled with an energy-dispersive spectrometer (EDS; EDAX type: Sapphire) at the Institute of Earth Sciences, University of Silesia (Sosnowiec, Poland). Operating conditions were: accelerating voltage 15 kV; working distance ca 10 mm; counting time 40 s. The same conditions were applied for back-scattered electron (BSE) imaging.

Electron Probe Microanalysis (EPMA)
The chemical composition of the investigated minerals was determined at the Inter-Institute Laboratory of the Microanalysis of Minerals and Synthetic Materials (Faculty of Geology, University of Warsaw, Warszawa, Poland) using a CAMECA SX-100 microprobe operated in wavelengthdispersive (WDS) mode at an accelerating voltage of 15 kV with a beam current of 10 nA. Counting time for each element was 40 s at the peak position and 20 s for background. Numerous test analyses with varying beam currents, probe diameters, and counting times were performed in order to optimise the experimental conditions. A beam diameter of 1-2 μm diameter caused significant damage to the sample, whereas a value of 5 μm and greater was in most cases too wide to collect Xrays from a single homogeneous grain. The optimal beam diameter was 3 μm, representing a compromise between the precision and accuracy of the measurements and the availability of suitable spots for analysis. However, the 3-μm beam caused partial dehydration of the köttigite-rich compositions, whereas erythrite was not affected. Lα lines were measured for As; Kα lines were measured for other elements. The following standards were used: GaAs for As; YPO4 for Y, CoO for Co, NiO for Ni, chalcopyrite for Cu, hematite for Fe, sphalerite for Zn, barite for S, diopside for Si, Ca, and Mg. Mn and Cr occurred below the detection limit of the microprobe. In total, 656 electron probe microanalyses were performed.

Structure Refinement
Single-crystal X-ray refinement was performed at room temperature with a single crystal (0.18 mm × 0.09 mm × 0.07 mm) mounted on a quartz capillary. The single-crystal composition was 1.92 Zn and 1.08 Co atoms per formula unit (apfu) as determined by SEM/EDS. A four-circle SuperNova X-ray diffractometer with a microfocus X-ray tube, optimised multi-layer optics for MoKα (λ = 0.71073 Å) radiation, and an Atlas CCD detector were used. Control of the measurement procedure and data reduction were performed by CrysAlis Pro software (version 1.171.38.41q; Rigaku Oxford Diffraction, 2015); the same program was used to determine and refine the lattice parameters. For integration of the collected data and correction of Lorentzian and polarisation effects, the CrysAlis RED program was used (version 1.171.38.41q; Rigaku Oxford Diffraction, 2015). Experimental details are summarised in Table 2.  The köttigite-erythrite structure was solved using direct methods with the SHELXS-2013 program [22] and refined by the full-matrix least-squares method on all F 2 data using the SHELXL program (version 2018/3) [22]. Structure refinement, including anisotropic atom displacement parameters, was carried out with neutral-atom scattering factors. All hydrogen atoms were found using the difference Fourier synthesis after a number of cycles of anisotropic refinement. Then their isotropic temperature factors, equal to 1.5 times the value of the equivalent displacement parameters, were assigned and suitable restraints used (DFIX).

X-Ray Powder Diffraction (XRD)
Powdered samples weighing 50 mg were examined to determine unit cell parameters using a fully automated X-ray PANalytical X'Pert PRO PW 3040/60 X-ray diffractometer equipped with CuKα1 source radiation (λ = 1.540598 Å) operated at 45 kV and 30 mA at the Institute of Earth Sciences, University of Silesia (Sosnowiec, Poland). Graphite-monochromated CuKα radiation was applied. Samples were placed on the surface of a silicon plate (zero background holder). Measurement parameters were: 10-70° or 5-65° 2θ range, time limit 2500 s or 3000 s, step size 0.02° 2θ. Quantitative analysis of the collected data was carried out using X'Pert HighScore Plus software with the newest PDF-4+ database.

Raman Spectroscopy
Raman spectra were recorded within the range 120-4000 cm -1 using a WITec alpha 300R confocal Raman microscope equipped with an air-cooled solid-state laser (λ = 532 nm) and a CCD camera. The excitation laser radiation was introduced into the microscope through a single-mode optical fibre 50 μm in diameter. An air Olympus MPLAN (100×/0.90NA) objective was used. Raman scattered light was focused onto a multi-mode fibre (100 μm in diameter) and onto a monochromator with a 600 line/mm grating. Raman spectra were accumulated by 10 scans with integration time of 10 s and a resolution of 3 cm -1 . The monochromator was calibrated using the Raman scattering line of a silicon plate (520.7 cm -1 ). To avoid possible polarisation of scattered light, all analysed crystals were oriented along (010) cleavage planes. Baseline correction and cosmic ray removal were performed using WITec Four Plus software. Peak fitting was done using GRAMS software. Spectra were fitted using a Voigt function with preservation of a minimum number of components.

Chemical Composition
Representative EPMA data for the erythrite-köttigite series are presented in Table 3. The analysed samples contain three sets of chemical compositions ( Figure 6). One set of EPMA data plots linearly (R 2 = 0.9812) and continuously along the 3.00 apfu tie line in the Co vs. Zn diagram ( Figure  6a), covering about 90% of the expected erythrite-köttigite solid solution. A compositional gap of 0.26 apfu (9%) extends towards the köttigite end-member. The Zn-poor compositions are relatively enriched in Mg (Table 3). Two other compositional datasets of erythrite with a more complex pattern of isomorphous substitutions (Figure 6b) include Zn-free erythrite and erythrites with Co > 1.75 apfu which display scattered data points in the Co vs. Zn diagram below the Co-Zn regression line ( Figure  6b) due to substitutions of Mg, Ni, and Zn for Co (Figure 7).   Elevated contents of Mg were found in numerous erythrite analyses with Mg > Zn (Table 3, Figure 7). The highest Mg content in a single analysis (No. 27 in Table 3) was 4.04 wt% MgO (0.58 apfu; 29 mol% hörnesite); however, in most compositions the hörnesite mole fraction was <0.20. One group of analyses corresponded to a limited erythrite-hörnesite binary with up to 22 mol% hörnesite, whereas most analyses represented a limited erythrite-hörnesite-köttigite ternary (Figure 7). With increasing Zn content, the amount of Mg decreased to 1 mol% hörnesite, or even less.
Some minor elements reported in Table 3 were considered to be impurities. Copper, ranging from 0.4 to 1.5 wt% CuO, was detected in two samples, most probably due to contamination with cooccurring pseudomalachite. Silicon is most probably an impurity element from host rocks and, as such, was not considered in crystal chemical calculations. Calcium occurred persistently in almost all analyses, particularly in erythrite-rich compositions with a maximum content of 0.82 wt% CaO. However, due to its erratic content and incompatibility with the erythrite-köttigite structure, it was also considered an impurity and excluded from crystal chemical calculations.
The content of water was not determined. The ideal number of eight H2O molecules was assumed in crystal-chemical formulae. However, Zn-rich analyses (Zn >> Co) display totals too high to account for the eight H2O molecules in köttigite with an ideal formula (Table 3). Test analyses indicated a high vacuum and a small electron beam size as causes of the partial dehydration of köttigite resulting in elevated totals ( Figure 5). Based on the evaluation of EPMA data obtained during this study as well as on theoretical considerations, we found that the water molecule content in the köttigite formula unit decreased during partial dehydration from eight to (8 -n), where n is an even number of water molecules. Inspection of the data in Table 3 shows that analyses with Zn >> Co are characterised by totals close to or even exceeding those for n = 4, i.e., they are close to the stoichiometry of arsenohopeite (Zn3(AsO4)2 4H2O). The decrease in H2O does not change the metalto-arsenate ion ratio.
The small beam size and high vacuum did not affect the erythrite-rich (Co >> Zn) analyses ( Table  3). The different behaviour of erythrite and köttigite under the same analytical conditions can be explained by considering the electronic structure of Co and Zn in relation to the crystal field stabilisation energy (CFSE) of the octahedral sites. Cobalt is an open-shell d 7 transition metal, whereas Zn is a closed-shell d 10 transition metal. The electronic configuration of Co favours octahedral coordination, as suggested by a CFSE value of approximately 0.8. The CFSE value for octahedral Zn is 0.0, meaning that there is no difference in CFSE between octahedral and tetrahedral sites. Therefore, the structure with octahedral Zn is relatively unstable. This structural instability may explain the apparent susceptibility of the köttigite to both electron beam damage and high-vacuum dehydration.

Structure Refinement
The topology of the refined Co-rich köttigite (Zn/Co = 1.78) structure (Figures 1 and 8) is exactly the same as that determined by previous investigators for all vivianite-type arsenates [1][2][3][4][5][6][7][8]. Atomic coordinates, equivalent isotropic displacement parameters, selected inter-atomic distances and angles in the examined crystal are presented in Tables 4 and 5. Anisotropic displacement atom parameters together with bond lengths and angles not included in Tables 4 and 5 are given in  supplementary Tables S1 and S2. As in all vivianite-type structures of arsenates and phosphates, the average M2-O bond length (2.103 Å) is shorter than the average M1-O distance (2.122 Å). The <M1-O> average bond length in erythrite with 0.11 apfu Zn (2.122 Å) is greater than that in köttigite with 0.42 apfu Co (2.115 Å) ( Table 6), whereas the average <M2-O> distance is smaller in erythrite (2.088 Å) than in köttigite (2.100 Å). Inter-atomic distances in Co-köttigite (64 mol% köttigite, 36 mol% erythrite) do not follow this trend. The <M1-O> distance in Co-köttigite is the same as in erythrite, whereas the <M2-O> distance is slightly greater (by 0.003 Å) than that in köttigite (Table 6). Regularity is, however, observed for changes in <As-O> distances ( Table 6). With increases in Zn, all <As-O> distances are shortened on average from 1.710 Å in erythrite to 1.691 Å in Co-erythrite to 1.682 Å in köttigite. This bond shortening is associated with the slight increase in the <O-As-O> angle from 109.36° in erythrite to 109.41° in Co-köttigite.    Symmetry: (i) −x + 1/2, −y + 1/2, −z + 2; (ii) x − 1/2, −y + 1/2, z; (iii) x + 1/2, −y + 1/2, z. D-donor; Aacceptor. The refinement of the Co and Zn occupancies in the M1 (0 0 0) and M2 [(0.5, y, 1), (0.5, -y, 1)] octahedra did not reveal any significant differences between these sites. The calculated value of the total scattering power (S) for both M1 and M2 sites is 28.92 epfu (electrons per formula unit). The calculated average bond lengths (based on values of ionic radii in [17]) are 2.140 Å for Zn-O, 2.145 Å for Co-O, and 2.142 Å for the presented model of the refined structure with Zn/Co of 1.8. The differences in bond lengths are not large enough to counter the statistically random distribution of Zn and Co between the M1 and M2 sites. This observation is in accord with conclusions in [4,6] that the transition metals are randomly distributed over both octahedral sites in köttigite and erythrite. However, according to Yoshiasa et al. [7], Zn ions prefer larger M1 sites, whereas Fe 2+ occupies M2 octahedra in köttigite-parasymplesite intermediate composition. Antao and Dhaliwal [5] suggest that minor Ni and Zn may preferentially occupy M1 sites and that Fe 2+ occupies the M2 site in erythrite, with site occupancy factors for the M1 and M2 sites of 1.028 and 0.912, respectively. In the erythriteköttigite solid solution, the close similarity of [6] Co and [6] Zn ionic radii enables the randomisation of site occupancy upon Co ↔ Zn substitution.
Yoshiasa et al. [7] noted a significantly greater M2-O2 distance (2.112 Å), i.e., between the metal ion and oxygens of the O2-O2 sharing edge, than that between the metal ion and oxygens O5 in the M2-coordinating H2O molecules (M2-O5 = 2.094 Å) in the intermediary of the parasymplesite (54 mol%)-köttigite (46 mol%) series. They explained the difference in bond length (0.018 Å) as the result of cation repulsion in the edge-sharing octahedra. A similar but smaller difference in the bond distances of M2-O2 and M2-O5 (0.006 Å) can be inferred from the data in [6] for köttigite. However, this is not the case in the examined Co-rich köttigite. The M2-O2 distance is shorter than the M2-O5 distance by 0.015 Å (Table 4). In this respect, the Co-rich köttigite is similar to Zn-bearing erythrite [5] with M2-O2 < M2-O5, although in the latter the difference between bonds is much greater (0.039 Å).

Unit-Cell Dimensions
The similarity of the [6] Co 2+ and [6] Zn 2+ crystal radii (88.5 and 88.0 pm, respectively [17]) suggests that the unit-cell dimensions of isostructural erythrite and köttigite are similar, with the former's volume being slightly larger. Data in Table 7 confirm this supposition. The unit-cell volume decreases from 632.7 Å 3 in erythrite (Co/(Co+Zn) = 0.996) to 630.2 Å 3 in köttigite from Schneeberg (Co/(Co+Zn) = 0.15) ( Table 7). However, these changes are not linear, and some intermediary minerals are characterised by unit cells (Table 7) larger than expected for a solid solution conforming to Vegard's law. Perhaps subordinate cations (Mg and Ni) and distortion of the coordination polyhedra (discussed later) disrupt the linear trend. The β angle is the most sensitive to Co ↔ Zn substitutions, increasing linearly by 0.13° from erythrite (ER2) to köttigite ( Table 7). The c parameter is practically constant in all samples, whereas changes in the a and b parameters equal approximately 0.03 Å. Table 7. Unit-cell dimensions calculated from X-ray powder diffraction data for erythrite from Bou Azer, Morocco [5], köttigite from Schneeberg, Germany [6], and samples from Miedzianka.
The low-frequency region (i.e., <1000 cm -1 ) in the Raman spectra of the arsenate series (Figure 9) of the vivianite group minerals consists of numerous overlapping bands related to stretching and bending modes of the major structural units [AsO4] and [MO6] (M = Co, Cu, Ni, Mg, Zn, Fe) as well as of librational modes of H2O [14]. The two highest-intensity bands at 854 and 794 cm -1 in the Raman spectrum of erythrite ( Figure 9) correspond to the asymmetric (ν3) and symmetric (ν1) vibrations of [AsO4], respectively [10,14,26]. With increasing substitution of Zn 2+ for Co 2+ , the ν3 and ν1 vibration bands shift towards higher wavenumbers up to 861 and 803 cm -1 in 90 mol% köttigite (Figure 9). This upshift is caused by the contraction of chemical bonds as Co is being replaced by Zn. Otherwise, it would be difficult to explain the upshift of Raman bands, since the atomic mass of Zn (65.38) is higher than that of Co (58.93). In fact, the <As-O> bonds become shorter with increasing Zn content ( Table  6). The shortening of the As-O distances is associated with the O1-As-O2 angle adjustment, from 105.5° in erythrite [5] to 106.6° in köttigite [6]. The Raman spectrum below 500 cm -1 is complex due to the multitude of overlapping bands which are difficult to assign to particular vibrational modes ( Figure 10). It is, however, possible to distinguish a region of ν4 and ν2 deformation modes of [AsO4] tetrahedra in the range 500-400 cm -1 and 400-250 cm -1 from the lattice modes below 250 cm -1 (Figure 10). This observation is in agreement with previously published data for erythrite [10,14,26]. The occurrence of a HAsO4 group in erythrite may activate a deformational type of vibration at 374 and 344 cm -1 [32]. Substitution of Co by Zn causes these bands to shift to 378 and 336 cm -1 , respectively ( Figure 10). The octahedral Co should activate stretching vibrations of Co-O expressed as a band at ~690 cm -1 in the Raman spectrum. However, there is no such band in the erythrite spectrum, possibly due to its downshift to ~450 cm -1 , caused by the Jahn-Teller distortion of octahedra and a resulting overlap with the deformational modes of arsenate tetrahedra. Two structurally different Co environments may induce band asymmetry at around 450 cm -1 , which can be deconvoluted into three bands at 458, 444, and 428 cm -1 . A similar effect may induce deformational modes assigned to bands at about 226 cm -1 ( Figure 10). All of these band assignments are supported by intensity changes in bands related to the Co ↔ Zn substitution. The band at 445 cm -1 in the Raman spectrum of köttigite is characterised by much lower intensity than the equivalent band at 444 cm -1 in erythrite ( Figure 10). The opposite is observed for a band at 226 cm -1 in erythrite, which, upon an increase in Zn content, shifts to 216 cm -1 in köttigite. This downshift is associated with an increase in intensity ( Figure 10). Bands centred at 393, 374, 263, and 248 cm -1 in the Raman spectrum of erythrite (Co >> Zn) are observed at 378, 262, and 245 cm -1 in köttigite (Zn >> Co) ( Figure 10).
Factor group analysis of erythrite from Miedzianka predicts 12 active internal modes of structural water in accord with the previously reported data [10,14,26], of which five are typical of  hydroxyl stretching vibrations (3800-2900 cm -1 ) ( Figure 11); the other seven occur within the range 1800-100 cm -1 . The relationship between band positions caused by the O-H stretching vibrations in the Raman spectra of the erythrite-köttigite solid solution and the donor-acceptor hydrogen bond length was determined using Libowitzky's equation [33]. The Raman spectrum of erythrite in the region of the O-H stretching vibrations consists of a narrow band at 3453 cm -1 corresponding to dH...O ≈ 1.98 Å and four broader bands at 3326, 3192, 3050, and 2924 cm -1 corresponding to dH...O of 1.89, 1.78, 1.72, and 1.68 Å, respectively ( Figure 11). In köttigite, the equivalent band positions are at 3427, 3382, 3195, 3043, and 2909 cm -1 ( Figure 11) corresponding to dH...O ≈ 1.95, 1.91, 1.79, 1.72, and 1.68 Å, respectively. The Raman spectrum of Co-rich köttigite examined in this study show intermediate band positions between the two end members. The intensity of the band at 3432 cm -1 in the Co-rich köttigite spectrum is much lower than those of the equivalent bands in the spectra of erythrite and köttigite and is also lower than the intensity of a band at 3368 cm -1 . The opposite is observed in the Raman spectrum of köttigite, in which the band at 3427 cm -1 is a very pronounced feature ( Figure  11).

Conclusions
A wide compositional range, covering about 90% of the expected erythrite-köttigite substitutional solid solution, was revealed in a suite of samples from a single polymetallic deposit, thus proving the miscibility between Co and Zn cations. This solid solution was possible due to combinations of favourable conditions, i.e., the availability of essential cations, the absence of crystal chemical constraints, and a suitable Eh-pH range. Perhaps it would not have been possible to directly prove the occurrence of such an extended solid-solution series within a single locality if not for the large numbers of representative samples (37) and of performed electron microprobe analyses (656). The absence of the köttigite end member may have resulted from changes in the geochemical (a) (b) environment that precluded its crystallisation. Instead of köttigite, unspecified hydrous Zn-arsenate crystallised as the last mineral in the examined assemblage (Figure 4d). In addition to the unlimited erythrite-köttigite binary, erythrite also displays a more complex pattern of isomorphous substitutions, including limited erythrite-hörnesite, erythrite-annabergite, erythrite-hörnesite-köttigite, and erythrite-hörnesite-annabergite solid-solution series. Unlike erythrite, köttigite lacks significant substitutions of Zn with Mg and Ni. This observation, combined with the data in Table 1, leads to the conclusion that there are two subseries of dominant cationic substitutions within the arsenate series linked by erythrite ( Figure 12). Substitutions between these subseries are possible but rather limited. Whether these two subseries are an artefact of a limited amount of data or are genuine is an open question that requires further investigation. The ternary solid solution between erythrite, hörnesite, and annabergite appears to have been enabled by their overlapping Eh-pH stability fields [24,25]. This external factor explains the preferred incorporation of Mg and Ni into erythrite in the examined samples.
The single-crystal refinement of köttigite and Co-rich köttigite, as performed by Hill [6] and as part of the present study, suggests structural disorder in the erythrite-köttigite solid solution, i.e., the random distribution of Co and Zn ions over two special octahedral sites. This is in contrast with the parasymplesite-köttigite solid solution, which displays cations partitioned between octahedral sites [7]. Therefore, there are two structural types of solid solutions within the arsenate series of the vivianite group minerals: disordered and ordered.
Small but significant changes in bond distances, rotation, and distortion of coordination polyhedra were observed in response to the Co ↔ Zn substitutions ( Figure 13). Raman spectra are extremely sensitive to these changes and can be used to distinguish between both end members and intermediary compositions in the erythrite-köttigite solid-solution series. The structure of köttigite is less stable in a vacuum, as well as more susceptible to electron beaminduced dehydration, than erythrite, due to the instability of Zn ions in the octahedral coordination.
The paragenetic sequence within the extended erythrite-köttigite solid solution reflects the chemical evolution of supergene fluids towards higher acidity associated with a gradual increase in Zn(II) activity related to the exhaustion of Co(II) due to the early precipitation of erythrite.
are thanked for their help in samples collection. We want to thank Rafał Juroszek for the separation of crystals to single-crystal diffraction measurements. This manuscript has benefitted from the reviews by two Minerals anonymous reviewers.

Conflicts of Interest:
The authors declare no conflict of interest.