A new nomenclature scheme for the alluaudite supergroup

The name alluaudite was given by Damour (1847, 1848) for a Na–Mn–Fe-bearing phosphate mineral found in the Chanteloube pegmatite, Limousin, France. The mineral was named after the French naturalist François II Alluaud (1778–1866), Mayor of Limoges and director of a ceramics factory. Fisher (1957) demonstrated that two other species, varulite from the Varuträsk pegmatite, Sweden (Quensel, 1937, 1940), and hühnerkobelite from the Hühnerkobel pegmatite, Bavaria (Mason, 1942), also belong to the alluaudite group.

Alluaudite samples from the Buranga pegmatite were investigated by Thoreau (1954), and then by Moore (1971) who described the crystal structure of the mineral. Alluaudite is monoclinic, space group C2/c, a = 12.004(2), b = 12.533(4), c = 6.404(1) Å, β = 114.4(1)°, with the structural formula X(2)X(1)M(1)M(2)₂(PO₄)₃. Recently, Hatert et al. (2000) and Krivovichev et al. (2013) proposed the formula [A(2)A(2)’A(2)”₂][A(1)A(1)’A(1)”₂]M(1)M(2)₂(PO₄)₃, which takes into account the distinct crystallographic positions observed in the channels of the structure.

The alluaudite structure consists of kinked chains of edge-sharing M(1) and M(2) octahedra, stacked parallel to the {101} plane. These chains are connected by the P(1)O₄ and P(2)O₄ tetrahedra, to form layers perpendicular to the crystallographic b axis (Fig. 1). Between these layers occur two types of channels running along the c axis: channel 1 which contains the A(1), A(1)’, and A(1)” sites, and channel 2 which contains the A(2), A(2)’, and A(2)” sites (Fig. 1). Generally, only A(1) and A(2)’ are significantly occupied.

Moore & Ito (1973) described the new phosphate wyllieite in the Victory mine pegmatite, South Dakota, and found unit-cell parameters a = 11.862(2), b = 12.382(1), c = 6.354(9) Å, β = 114.52(8)°, space group P2₁/n. The unit-cell parameters of this phosphate are very close to those of alluaudite, and a single-crystal investigation by Moore & Molin-Case (1974) indicates a crystal structure topologically identical to the alluaudite structure. The only difference between these two groups of phosphates is the presence of significant Al in wyllieite, thus implying the splitting of the M(2) and X(1) sites of alluaudite into M(2a)–M(2b) and X(1a)–X(1b) in wyllieite, as well as a change of space group. The structural formula of wyllieite-type phosphates corresponds to X(2)X(1a)_0.5X(1b)_0.5M(1)M(2a)M(2b)(PO₄)₃. More recently, Ercit et al. (1986a, b) described bobfergusonite, □Na₂Mn₅Fe³⁺Al(PO₄)₆, in the crystal structure of which the adjacent octahedral chains show an ordering, thus implying a doubling of the c unit-cell parameter of wyllieites: P2₁/n, a = 12.773(2), b = 12.486(2), c = 11.038(2) Å, β = 97.15°.

Numerous arsenates also show the alluaudite structure. Caryinite, (Na,Pb)(Ca,Na)Ca(Mn,Mg)₂(AsO₄)₃, was described by Lundström (1874) in the Långban pegmatite, Sweden, and was later shown by Strunz (1960) to be isostructural with alluaudite. The other arsenates arseniopleite, badalovite, bradaczekite, calciojohillerite, camanchacaite_, canutite, erikapohlite, hatertite, johillerite, keyite, khrenovite, magnesiocanutite, magnesiohatertite, nickenichite, odanielite, paraberzeliite, yazganite, and zincobradaczekite also belong to the alluaudite group (Keller et al., 1981, 1982; Dunn & Peacor, 1987; Auernhammer et al., 1993; Cooper & Hawthorne, 1996; Krivovichev et al., 2001, 2013; Sarp & Černý, 2005; Schlüter et al., 2013; Kampf et al., 2014, 2017, 2018; Pekov et al., 2016a, b, c, d, 2018a, b).

According to the IMA-CNMNC recommendations for classification (Mills et al., 2009), mineral groups are defined on the basis of their crystal structure. Consequently, alluaudite-type arsenates and phosphates belong to the same group, which is named “alluaudite group” since alluaudite is the first described mineral (Damour, 1847, 1848). Wyllieite- and bobfergusonite-type phosphates show topologically similar structures, but with different space-groups and/or unit-cell parameters, thus implying the definition of the distinct “wyllieite” and “bobfergusonite” groups. The alluaudite supergroup consequently contains phosphates and arsenates topologically related to the alluaudite structure; these minerals are alluaudite-type phosphates and arsenates (alluaudite group: 6 phosphates and 19 arsenates), wyllieite-type phosphates (wyllieite group: 6 phosphates), bobfergusonite-type phosphates (bobfergusonite group: 3 phosphates), and manitobaite (Ercit et al., 2010).

The goal of this paper is to revise the classification of the alluaudite supergroup, in order to homogenize the nomenclature of these phosphates and arsenates; this new nomenclature scheme has been accepted by the IMA-CNMNC.

The nomenclature of alluaudite- and wyllieite-type phosphates has been established by Moore & Ito (1979), from the examination of chemical analyses of natural samples. These analyses were interpreted taking into account the structural formula of alluaudite proposed by Moore (1971): X(2)X(1)M(1)M(2)₂(PO₄)₃. In natural samples, cations are distributed over the different crystallographic sites according to their ionic radii, with the smaller cations on M(2) and the largest cations on X(2):

According to Moore & Ito (1979), alluaudite-type phosphates are defined on the basis of the dominant cation on the M(2) site: Fe³⁺ in alluaudites, Mn in varulites, and Fe²⁺ in hagendorfites (this mineral was described by Strunz (1954) in the Hagendorf pegmatite, Bavaria). The M(1) site is generally occupied by Mn, but prefixes “ferro-” or “mag-” may be added to the root-name when Fe²⁺ or Mg are dominant on M(1), respectively.

Five alluaudite-type phosphates currently occur in the IMA-CNMNC mineral list, which follow this nomenclature scheme: alluaudite, $NaMnFe23+(PO4)3$⁠, hagendorfite, $NaCaMnFe22+(PO4)3$⁠, varulite, $NaCaMnMn22+(PO4)3$⁠, ferroalluaudite, $□NaFe2+Fe23+(PO4)3$⁠, and maghagendorfite, $NaMnMgFe22+(PO4)3$⁠. Groatite, □NaCaMn₂(PO₄)(HPO₄)₂, is a protonated alluaudite-type phosphate which does not follow the current nomenclature scheme (Cooper et al., 2009).

Mineralogists involved in the study of pegmatite phosphates know, by experience, that the nomenclature scheme suggested by Moore & Ito (1979) does not work properly. In order to demonstrate this feature, we decided to examine all chemical analyses of alluaudite-type phosphates occurring in the literature. A total of 64 analyses were selected and recalculated in order to uniformize the data, on the basis of 3 P atoms per formula unit (apfu). Then, the Fe³⁺/Fe²⁺ ratio was calculated to maintain charge balance. In some extremely oxidized samples, it was also necessary to oxidize some Mn²⁺ into Mn³⁺ to reach the 9 positive charges in the cationic groups of the formula unit. Protonated alluaudite-type phosphates were not considered in these calculations, since they contain only 7 positive charges in the cationic groups of the formula unit. Their nomenclature will be briefly discussed below.

A procedure to calculate electron-microprobe analyses of alluaudite-type phosphates is given in the Appendix, and the complete list of analyses is freely available as Supplementary Material linked to this article at https://pubs.geoscienceworld.org/eurjmin/.

Starting from these recalculated analyses, we distributed the cations on the different crystallographic sites of the structure, following the known crystal-chemical constraints mainly established from synthetic compounds (Hatert et al., 2000, 2003, 2005a; Hatert, 2008; Rondeux & Hatert, 2010). According to these constraints, cations with small effective ionic radii (Shannon, 1976) were assigned to the M(2) site (Fe³⁺, Mn³⁺, Al, Mg, Zn). When necessary, this site was filled by Fe²⁺ and then by Mn²⁺. The M(1) site was mainly filled by divalent cations Fe²⁺, Mn²⁺, and Ca, but it may also contain small amounts of Fe³⁺ and Na. The A(1) position is mainly occupied by Na, Ca, Mn²⁺ and sometimes vacancies, while the large A(2)’ site contains large monovalent cations Na, Li and K, as well as significant vacancies.

Figure 2 shows a ternary diagram where the cationic contents of the M(2) site have been plotted. From this diagram, it is clear that the majority of analyses fall in the Fe³⁺-dominant field, corresponding to alluaudites sensu stricto. The analyses of type varulites are far away from the Mn-dominant pole: they are located on the Fe³⁺-Mn side of the diagram, not far from the $Fe0.53+Mn0.5$ boundary. Analyses of type hagendorfites are also far from the Fe²⁺-dominant end-member; they are closer to the $Fe0.53+Fe0.52+$ boundary.

From this diagram, it is obvious that the nomenclature scheme of Moore & Ito (1979), which considers the cation dominance on the M(2) site as the main criteria to define the root name of alluaudite-type phosphates, is in contradiction with the analyses of type varulites and hagendorfites. A careful examination of the cationic distribution on the different crystallographic sites of these phosphates (Table 1) indicates M(2) occupancies close to $Fe0.53+Mn0.5$ for type varulite (Quensel, 1937), and close to $Fe0.53+Fe0.52+$ for type hagendorfite (Strunz, 1954), thus showing the necessity to redefine the ideal formulae of these minerals.

In the IMA-CNMNC list of minerals, hagendorfite is defined as $NaCaMnFe22+(PO4)3$⁠, and varulite as $NaCaMnMn2(PO4)3$⁠. These formulae were established by considering Ca as dominant on the A(1) site, in order to achieve charge balance. However, as shown on Fig. 3 in which the cationic contents on the A(1) site of alluaudite-type phosphates were plotted, Ca does not play a significant role, since A(1) is mainly occupied by Na in the majority of samples. This confirms again that the current nomenclature scheme is inappropriate.

Table 1 shows the cationic distributions in several alluaudite-type phosphates, taking into account the nomenclature scheme of Moore & Ito (1979). As shown on this table, trying to derive end-member formulae by simply considering the site dominant cation would lead to unbalanced formulae in several cases (identified by double quotes), e.g. the compositions of type varulite, type hagendorfite, as well as for two compositions reported by Moore & Ito (1979). If we consider the 64 analyses from the literature, it appears that 44% of these analyses do not allow the straightforward derivation of a charge-balanced formula, thus explaining the difficulties of some authors to identify phosphates of the alluaudite group. Among these analyses, 48% are alluaudites, 3% are hagendorfites, 3% are ferroalluaudites, and 2% are unknown end-members, while varulites are absent.

Finally, it is surprising to see that the analysis of “hühnerkobelite”, described by Mason (1942) and later discredited by Moore & Ito (1973), leads to the charge-balanced end-member formula $NaCaMnFe22+(PO4)3$ (Table 1), which is now attributed to hagendorfite. As shown above, however, type hagendorfite does not contain high amounts of Ca (Strunz, 1954).

To solve this issue, we suggest to start from alluaudite, which is the only near-end-member frequently observed in Nature. Alluaudite occurs in many pegmatites, and is frequently of secondary origin. As observed by Fransolet et al. (2004) in the Kibingo pegmatite, Rwanda, a primary weakly-oxidized hagendorfite may be oxidized into secondary alluaudite, from the substitution mechanism: Na⁺ + Fe²⁺ = □ + Fe³⁺. This oxidation is clearly visible in thin sections, where green hagendorfite is oxidized, along fractures, into yellow alluaudite (Fig. 4). During this oxidation mechanism, phosphates show chemical compositions plotted on Fig. 5, which clearly indicates positive correlations between the Fe³⁺ contents and the number of vacancies per formula unit. The slopes of these correlations are close to 1.0, thus confirming the substitution mechanism defined above.

Starting from the end-member formula of alluaudite, □$NaMnFe23+(PO4)3$⁠, and applying the oxidation mechanism Na⁺ + Fe²⁺ = □ + Fe³⁺, it is then possible to redefine the end-member composition of primary hagendorfite as NaNaMn(Fe²⁺Fe³⁺)(PO₄)₃. This formula seems to contradict the dominant-constituent rule, since a mixed occupancy of the M(2) site by 50% Fe²⁺ and 50% Fe³⁺ is considered here. However, as recommended by Hatert & Burke (2008), valency-imposed double site-occupancy may be required to establish charge-balanced end-member formulae. We consequently suggest to apply, for the nomenclature of alluaudite-type phosphates and arsenates, a nomenclature that takes into account this double site occupancy.

It is important to underline that such a nomenclature scheme is consistent with the nomenclature of the wyllieite group, in which the M(2) site is split into two crystallographically independent positions, M(2a) and M(2b). But unlike the wylliteite group, the two M(2) sites are not symmetrically distinct in the alluaudite group, even if the “M(2a)” and “M(2b)” sites names are also used in this proposal. This “artificial splitting” is only applied, here, for nomenclature purposes.

As mentioned above, the end-member formula of varulite, accepted by the CNMNC-IMA, is NaCaMn(MnMn)(PO₄)₃, which does not correspond to the observed chemical compositions. If we allow a valency-imposed double site occupancy on M(2), and if we take into account the measured composition of type varulite (Table 2; Figs. 2, 3), we can redefine the end-member formula of varulite as NaNaMn(MnFe³⁺)(PO₄)_3.

Table 2 shows the cationic distributions in selected analyses from the literature, considering this new nomenclature scheme. This table shows that the majority of tentative end-member formulae are now charge-balanced, except the formulae of type alluaudite, type hühnerkobelite, and of the sample from Dyke Lode (Moore & Ito, 1973). This is due to the fact that some heterovalent substitutions occur in these samples, which modify their compositions. In type alluaudite and in type hühnerkobelite, the substitution mechanism Na⁺ + Fe²⁺ = □ + Fe³⁺ occurs, responsible for the presence of significant Na on A(2)’ and Fe²⁺ on M(2a) in type alluaudite, and of significant vacancies on A(2)’ and Fe³⁺ on M(2b) in type varulite (Table 2).

Even if the end-member formulae of these type samples are not straightforwardly derived in the new nomenclature scheme, we recommend keeping the end-member formula of alluaudite as □NaMn(Fe³⁺Fe³⁺)(PO₄)₃, since this composition is extremely common in pegmatites and corresponds to the ideal composition of alluaudite from the Buranga pegmatite, which was investigated in detail (Table 2).

Consequently, we recommend modifying the end-member formulae of alluaudite-type phosphates as reported in Table 3. In this new nomenclature, the root-name is defined by the M(2) content, and a prefix is added depending on the M(1) content. Three situations occur:

1. 0.5 < ^M(2)M²⁺ < 1.5 (M²⁺ mainly Fe²⁺): hagendorfites,

2. 0.5 < ^M(2)M²⁺ < 1.5 (M²⁺ mainly Mn²⁺): varulites,

3. ^M(2)M²⁺ < 0.5 (M³⁺ mainly Fe³⁺): alluaudites.

The application of the new nomenclature scheme to all chemical analyses from the literature shows that it allows one to derive straightforwardly a charge-balanced end-member formula for 72% of the samples, which constitutes a significant improvement compared to the previous nomenclature. For the other samples, it is necessary to identify the substitution mechanisms, in order to find the end-member composition. From a general point of view, four different types of substitution mechanisms occur in alluaudite-type phosphates:

1. S1: Na⁺ + M²⁺ = □ + Fe³⁺ (oxidation),

2. S2: □ + 2Fe³⁺ = M²⁺ + 2 M²⁺ (Na constant),

3. S3: Na⁺ + Na⁺ = □ + Mn²⁺ (Fe constant),

4. S4: Na⁺ + Fe³⁺ = M²⁺ + M²⁺ (□ constant).

The oxidation mechanism S1 was demonstrated above (Fig. 5), and Fig. 6 shows a negative correlation between Fe³⁺ + □ and the M²⁺ content, thus confirming substitution mechanism S2.

A careful examination of the chemical analyses shows that when a tentative end-member formula (derived by mere consideration of the site-dominant constituent) is not charge-balanced, this is due to the influence of one of these supplementary substitution mechanisms. The most common mechanism is S3, which explains how the number of Na atoms may decrease on the A sites, where they are replaced by Mn and vacancies. Several examples of this type are shown in Table 2:

1. The alluaudite type sample (Damour, 1848) corresponds to hagendorfite, since 0.720 M²⁺ (dominantly Fe²⁺) occur on M(2). However, the substitution mechanism S3 renders vacancies dominant on A(2)’, which, upon mere consideration of the site-dominant constitutent, would lead to the unbalanced tentative end-member formula “□NaMnFe²⁺Fe³⁺(PO₄)₃”. This mineral has to be considered as a Na-poor hagendorfite.

2. The sample from the Buranga pegmatite, investigated by Héreng (1989), is an alluaudite, since it contains less than 0.298 M²⁺ cations on M(2), where Fe³⁺ is dominant. However, the A(1) site is dominantly occupied by Mn, due to substitution mechanism S3, which again offers the pitfall of an unbalanced tentative formula, “□MnMn$Fe23+$(PO₄)₃”. This sample has to be considered as a Mn-rich alluaudite.

3. A sample from La Fregeneda pegmatite, Spain (Roda et al., 1996), contains 0.586 Fe²⁺ on M(2), and therefore corresponds to a hagendorfite. The M(1) site is predominantly occupied by Fe²⁺, so it is a “ferrohagendorfite” (not IMA-CNMNC approved). However, vacancies are dominant on A(2)’ due to the substitution mechanism S3, thus leading to the unbalanced tentative formula “□NaFe²⁺(Fe²⁺Fe³⁺)(PO₄)₃”. This sample has to be considered as a Na-poor “ferrohagendorfite”.

4. Another sample from La Fregeneda contains 0.703 divalent cations on M(2), among which Mn²⁺ is clearly dominant: it is a varulite. However, the S3 substitution mechanism provokes the dominance of vacancies on A(2)’, leading to an unbalanced tentative formula, “□NaMn(MnFe³⁺)(PO₄)₃”. This sample has to be considered as a Na-poor varulite.

5. The sample from Skrumpetorp is characterized by a very high Mn³⁺ content, reaching 0.484 Mn³⁺ on M(2). In that case, a cationic distribution considering a splitting of the M(2) site seems to indicate a formula □NaMn(Mn³⁺Fe³⁺)(PO₄)₃. However, two cations of the same valence state occur on the same site, and this cannot be considered as a valency-imposed double site-occupancy. The “artificial splitting” of M(2) cannot be applied here, and since M³⁺ cations reach 1.667 apfu on M(2), with Fe³⁺ dominant, we have simply a Mn³⁺-rich alluaudite.

These examples illustrate how to use the new nomenclature scheme to determine an alluaudite-type phosphate. Taking into account the multiple substitution mechanisms, it is now possible to identify all samples. Starting from the 64 analyses from the literature, we found 33% hagendorfites, 31% alluaudites, 11% varulites, 1% ferroalluaudites, and 24% of new species that will be described below. Maghagendorfite is absent from the samples, since the cationic distribution used here considers a preference of Mg for the M(2) site, not for M(1), as experimentally demonstrated by Rondeux & Hatert (2010). Since the cationic distribution of maghagendorfite is unlikely, we recommend considering this species as “Questionable”.

In order to define new possible end-member compositions, we decided to establish theoretically all possible formula types for alluaudite-type phosphates. Starting from the formula of alluaudite, $□NaMnFe23+(PO4)3$⁠, we defined formulae of type 1, $□NaM2+Fe23+(PO4)3$ (Fig. 7). We consider here Fe³⁺ as the main trivalent cation occurring on M(2); however, some analyses from the literature show that Mn³⁺ may play a significant role on that site (Table 2). In wyllieites, Al also occurs on one of the two M(2) positions; this will be discussed in detail below. From type formula 1, it is then possible to define the formulae $□NaM2+Fe23+(PO4)3$ = alluaudites, and $□NaM2+Mn23+(PO4)3$ = a new root-name called ROOT1 (Fig. 8).

Starting from alluaudites, and applying the oxidation mechanism S1, we obtain type 2 with the theoretical formula $Na2M22+Fe3+(PO4)3$ (Fig. 7). From this general formula we deduce three root-names: hagendorfites = Na₂M²⁺(Fe²⁺Fe³⁺)(PO₄)₃, varulites = Na₂M²⁺(Mn²⁺Fe³⁺)(PO₄)₃, and a new root-name called ROOT2 = Na₂M²⁺(MgFe³⁺)(PO₄)₃ (Fig. 8).

Starting from alluaudites and applying substitution mechanism S2, or starting from formula type 2 and applying substitution mechanism S4, we obtain formula type 3, $NaM42+4(PO4)3$ (Fig. 7). Taking into account the crystal-chemical constraints existing in alluaudite-type phosphates, it is possible to define three root-names from this formula type: $NaCaM2+Fe22+(PO4)3=ROOT3$⁠, NaCaM²⁺Mg₂(PO₄)₃ = ROOT4, and $NaCaM2+Mn22+(PO4)3=ROOT5$ (Fig. 8).

Finally, starting from formula type 3 and applying the substitution mechanism S1, or starting from formula type 2 and applying the substitution mechanism S3, we obtain formula type 4, $□M32+Fe3+(PO4)3$ (Fig. 7). The possible root-names are □CaM²⁺(Fe²⁺Fe³⁺)(PO₄)₃ = hagendorfites-Ca, □CaM²⁺(Mn²⁺Fe³⁺)(PO₄)₃ = varulites-Ca, and □CaM²⁺(MgFe³⁺)(PO₄)₃ = ROOT2-Ca (Fig. 8). Note that in those cases, the M(2) contents are identical to those of formulae type 2, but the two Na cations on the A sites are replaced by Ca²⁺ and a vacancy. Therefore, the names of type 2 phosphates are used with the addition of a chemical suffix to designate that Ca is dominant on A(1). This suffix is written without parentheses since A cations in the alluaudite structure occur in channels, and can therefore be considered as extra-framework cations (Hatert et al., 2013).

Nineteen arsenates belong to the alluaudite group; they are listed in Table 4. Among these species, four show a double occupancy of the M(2) site: caryinite, hatertite, badalovite and magnesiocanutite, thus implying the use of the new nomenclature scheme. Alluaudite-type arsenates are found in exotic geochemical environments, such as volcanic fumaroles or Pb–Zn deposits. For this reason, their crystal chemistry is very different from that of alluaudite-type phosphates, with divalent cations Zn, Cd, and Cu²⁺ playing significant roles.

The formulae of alluaudite-type arsenates are most frequently of type 3, $NaM42+(AsO4)3$⁠, but significant amounts of hydrogen, of water molecules, or of vacancies may occur in these minerals (Table 4). Three different mechanisms can be identified, to explain the presence of vacancies: (i) an oxidation mechanism producing Fe³⁺ on M(2) and vacancies on the A sites, as observed in yazganite; (ii) the replacement of Na⁺ by (0.5 Cu²⁺ + 0.5 □), as observed in keyite and erikapohlite; (iii) the presence of H atoms in the channels, producing the protonated arsenates o’danielite, canutite, magnesiocanutite and camanchacaite. Hatertite and badalovite are characterized by formulae of type 2, $Na2M22+Fe3+(AsO4)3$⁠, while khrenovite has the ideal formula $Na3Fe23+(PO4)3$ (Table 4).

In these minerals, cationic distributions in the channels are governed by the presence of water molecules and hydrogen atoms, as underlined by Krivovichev et al. (2013) and Kampf et al. (2014). In unprotonated alluaudite-type arsenates, the A(1) and A(2)’ sites are generally occupied but, if vacancies occur, they are preferentially located on A(2)’, as observed in yazganite, keyite and erikapohlite (Table 4). The resulting free space is filled by water molecules in channel 2 of these minerals; it seems that half occupancy of A(2)’ is enough to accommodate one water molecule. In protonated arsenates, channel 1 is occupied by the H atoms of the [AsO₃(OH)] and [AsO₂(OH)₂] groups, thus explaining the presence of vacancies on the A(1) sites of o’danielite, canutite, magnesiocanutite and camanchacaite (Table 4).

No nomenclature guidelines actually exist for alluaudite-type arsenates; for this reason, new names were given for each species. According to the CNMNC guidelines (Hatert et al., 2013), we recommend preserving the existing historical names, reported in Table 4. These names must be considered as root-names; they are defined from their contents on both M(1) and M(2) sites. If new arsenates would be found, belonging to the alluaudite group, they could receive a new root-name if their occupancies on M(1) and M(2) are different from those of known species (Table 4). However, as has recently been done for magnesiocanutite, zincobradaczekite and magnesiohatertite, a prefix may also be used to designate the cationic content on the M(1) or M(2) sites. If the M(1) and M(2) occupancies are identical, but if the cations occurring in the channels are different, we recommend keeping the existing root-name, and to add a chemical suffix without parentheses, to describe the cationic content of the A(1) and A(2)’ extra-framework sites. A maximal number of two suffixes would be sufficient.

A careful examination of the site populations for alluaudite-type arsenates, given in Table 4, shows that some end-member formulae have to be significantly modified, compared to those occurring in the IMA-CNMNC mineral list. The main modifications are given in Table 5, but all CNMNC formulae have to be modified according to Table 4.

The double site-occupancy is applied for mineral with a mixed valence on M(2); however, no artificial splitting is used when M²⁺ cations are dominant on this site. The valency-imposed double site-occupancy is therefore only applicable for hatertite, badalovite, and magnesiohatertite (Table 3). In arseniopleite, the A(1) site is mainly occupied by Ca, not by Na (Table 3).

The most problematic issue in alluaudite-type arsenates concerns the recently described calciojohillerite (Pekov et al., 2016c), which shows a M(2)–M(1) distribution similar to that of nickenichite, but with slightly different cationic distributions between the sites occurring in channel 1. Indeed, nickenichite is the only alluaudite-type arsenate in which both A(1) and A(1)’ sites are significantly occupied. The resulting ideal formula of nickenichite becomes Na(Ca_0.5Cu_0.5)MgMg₂(AsO₄)₃, while calciojohillerite corresponds to NaCaMgMg₂(AsO₄)₃. In this special case, we recommend keeping both mineral species, but in the future, we suggest to avoid defining alluaudite-type minerals from slight modifications of the cationic distributions among the different crystallographic positions located in the channels. We recommend considering, for nomenclature purpose, the cations A(1) + A(1)’ + A(1)” (channel 1) as a whole, as well as the cations A(2) + A(2)’ + A(2)” (channel 2).

Table 6 gives a summary of the classification of alluaudite-type arsenates, distributed according to their M(2) contents.

Phosphates of the wyllieite group also occur in granitic pegmatites, but they generally crystallize as primary phases (Vignola, 2018), or under high-temperature hydrothermal conditions, by reaction between Fe–Mn–phosphates and albite (Fig. 9). For this reason, these phosphates contain significant amounts of aluminium. The presence of this element induces a splitting of the M(2) and X(1) sites of alluaudite into M(2a)–M(2b) and X(1a)–X(1b), and provokes a change of space group. For this reason, wyllieite-type phosphates form their own mineral group within the alluaudite supergroup.

Wyllieite is the first species described in the group (Moore & Ito, 1973); however, its name was changed to ferrowyllieite by Moore & Ito (1979) who established nomenclature guidelines for the group. According to this nomenclature scheme, and to recent crystal-chemical investigations (Hatert et al., 2005b, 2006, 2010, 2016), the M(2a) site of the structure is systematically occupied by Al; the root-name consequently depends on the M(2b) content: Fe²⁺ in wyllieites, Fe³⁺ in rosemaryites, and Mg in qingheiites. The M(1) site is dominantly occupied by Mn or Fe²⁺; in the latter case, the prefix “ferro-” is added to the mineral name. The six valid mineral species in the wyllieite group, as well as their end-member formulae, are reported in Table 7.

Cation occupancies, established from X-ray crystal-structure determinations, indicate that the X(1a) site is generally occupied by Mn, while the X(1b) site contains Na. If we consider both sites together, for nomenclature purposes, we realize that X(1a+b) systematically contains significant amounts of Mn, in contradiction with the ideal end-member formulae which consider this site as predominantly occupied by Na. In rosemaryite and ferrorosemaryite, the presence of significant amounts of vacancies on the X(1a+b) sites indicates the occurrence of the substitution mechanism Na⁺ + Na⁺ = □ + Mn²⁺, previously observed in alluaudites.

This feature is also due to the detailed cationic distributions established from structural data: different cation grouping would generally provoke a dominant occupancy of Na on the X(1a+b) sites. Cationic distributions established from chemical data, by grouping cations according to their ionic radii, are also shown on Table 7; they better fit with the end-member formulae. The significant amounts of vacancies on the X(2) site of qingheiite-(Fe²⁺) are due to the presence of Fe³⁺ on the M(2b) site of the structure, according to the substitution mechanism: ^X2Na⁺ + ^M2bMg²⁺ = ^X2□ + ^M2bFe³⁺. And the significant amounts of Mn on the X(1a + b) sites of qingheiite are due to the presence of significant amounts of Fe²⁺ on the M(2a) site, according to the substitution mechanism: ^X1(a+b)Na⁺ + ^M2a(Al,Fe)³⁺ = ^X1(a+b)Mn²⁺ + ^M2aFe²⁺ (Table 7). These small differences between the ideal end-member formulae and the observed occupancies are not significant for nomenclature purpose, since the nomenclature of the wyllieite group is based on the contents of the M(2a), M(2b), and M(1) sites.

We recommend keeping the current nomenclature scheme for wyllieite-type phosphates, with three root-names wyllieite, rosemaryite, and qingheiite. In order to uniformize the nomenclature of the group, we suggest to re-name “qingheiite-(Fe²⁺)” in “ferroqingheiite”. If new species would be found in the group, with different M(1) occupancies, a prefix would be added to the corresponding root-name. New occupancies of the X sites would necessitate the addition of chemical suffixes without parentheses, as recommended for alluaudite-type phosphates and arsenates (see above).

An important remark concerns the distinction between alluaudite- and wyllieite-type phosphates. Indeed, both structure types are topologically identical, the only difference being the space-group types. But the accurate determination of space group in these minerals is tricky. To demonstrate this feature, we collected X-ray diffraction data on a single crystal of alluaudite from the Buranga pegmatite, Rwanda (sample RGM-9156, Rigaku Xcalibur four-circle diffractometer, EOS detector, MoKα radiation). A reciprocal-space image of the hk0* plane (Fig. 10) clearly shows weak 120-type diffraction spots, forbidden in the alluaudite C2/c space group. Moreover, both structure refinements in the C2/c or P2₁/n space groups give excellent structural models, with R₁ factors of 2.34% in C2/c and 2.64% in P2₁/n. This very difficult space-group choice forces us to redefine the wyllieite group as phosphates with an alluaudite topology, containing between 0.5 and 1.5 Al per formula unit on the M(2a+b) sites, and/or unambiguously with space-group type P2₁/n. An Al-free phosphate with the alluaudite topology, impossible to refine satisfactorily in C2/c, but refining well in P2₁/n, would then belong to the wyllieite group.

Recently, the new mineral species fupingqiuite, $(Na,Mn2+,□)2Mn22+Fe3+(PO4)3$⁠, was described by Yang et al. (2017) as a wyllieite-type phosphate, since its space-group type was P2₁/n. However, the authors did not provide the results of the structure refinement in C2/c, thus preventing an unambiguous space-group determination. Moreover, the composition of this species corresponds to the ideal composition of redefined varulite; consequently, varulite and fupingqiuite are topologically similar polymorphs, and cannot be considered as separate mineral species (Nickel & Grice, 1998). We recommend consequently discrediting fupingqiuite.

Finally, the bobfergusonite group contains three mineral species: bobfergusonite, □Na₂Mn₅Fe³⁺Al(PO₄)₆ (Ercit et al., 1986a, b), ferrobobfergusonite, $□Na2Fe52+Fe3+Al(PO4)6$ (Yong et al., 2017), and zhanghuifenite, Na₃Mn₄Mg₂Al(PO₄)₆ (Yang et al., 2016). Their crystal structures are based on the alluaudite topology, but an ordering of the adjacent octahedral chains provokes a doubling of the c unit-cell parameter, compared to wyllieites: space-group P2₁/n, a = 12.773(2), b = 12.486(2), c = 11.038(2) Å, β = 97.15° (for bobfergusonite). Manitobaite, $Na16Mn252+Al8(PO4)30$_, is a separate species with an even larger unit-cell: a = 13.452(2), b = 12.515(2), c = 26.661(3) Å, β = 101.58(1)° (space group Pc; Ercit et al., 2010). Nomenclature guidelines for the bobfergusonite group cannot be defined here, due to the few mineral species of this group known to date. However, some of these species may appear as topologically similar polymorphs of alluaudite- or wyllieite-type phosphates.

Due to the different geochemical environments in which minerals of the alluaudite supergroup occur, and to their complex crystal chemistry, it is very difficult to establish a unique nomenclature for the whole supergroup. However, the use of a nomenclature based on a double occupancy of the M(2) site is necessary, for alluaudite-type phosphates and arsenates, in order to better describe their chemical complexity. This new nomenclature scheme is consistent with the nomenclature of wyllieites, and allows one to easily establish charge-balanced end-member formulae for these phosphates.

A summary of the IMA-CNMNC-approved formulae of alluaudite-supergroup minerals is given in Table 8.

Many thanks to Stuart Mills, who handled the CNMNC proposal, as well as to all CNMNC members for their constructive comments on the first version of this paper.

Alluaudite-type phosphates generally contain Fe²⁺ and Fe³⁺, and it is necessary to recalculate the contents of these cations in the analyses.

Calculation basis is 3 P atoms per formula unit (apfu). Then, the Fe³⁺/Fe²⁺ ratio is calculated to reach nine positive charges in the cationic groups of the formula unit, maintaining charge balance. If a significant deficit of positive charges is still observed when all Fe is considered as Fe³⁺, it means that we have an extremely oxidized alluaudite. In that case, it is recommended considering all iron as Fe³⁺, and to calculate the Mn²⁺/Mn³⁺ ratio to reach nine positive charges in the cationic groups.

Cations are then distributed among the crystallographic sites, according to their ionic radii:

• M(2a): Fe³⁺, and then successively Mn³⁺, Al, Mg, and eventually Zn, Fe²⁺ or Mn²⁺ to fill the site (1 apfu).

• M(2b): Eventually the remaining Fe³⁺, Mn³⁺, Mg or Zn. If necessary, the site is then filled by Fe²⁺ first, then by Mn²⁺.

• M(1): In rare analyses of very oxidized alluaudites, small amounts of Fe³⁺ or Mn³⁺ may be attributed to that site. Then, Mg, Fe²⁺ and Mn²⁺ are successively added here. If the site is not yet filled, Ca and then Na are added.

• A(1): Remainings of Mn²⁺ are located here, as well as Ca and Na. If the site is not yet filled, K or Li may also be added here. Generally, this site is filled, but in a few analyses, small amounts of vacancies may occur here.

• A(2)’: Remainings of Na, as well as K and Li, are located on that site. Generally, significant amounts of vacancies also occur on that site.

Once the cations are distributed, it is necessary to determine the name of the mineral.

The root-name is based on the content of M(2), and a prefix may be added, depending on the content of the M(1) site. Generally, large A sites are occupied by Na and/or vacancies, and the contents of these constituents may vary significantly, since they are very sensitive to the complex substitution mechanisms affecting alluaudites. So, generally, the contents of A(1) and A(2)’ sites are not considered for nomenclature purposes. However, if the A(1) site of hagendorfites, varulites and root2 is mainly occupied by divalent cations (^A(1)M²⁺ > 0.5), a suffix is added to the name to designate this cation.

Mineral names of alluaudite-type phosphates may be determined according to the following scheme:

1. ^M(2)M²⁺ ≤ 0.5 apfu:

   • 1.a: ^M(2)M³⁺ dominantly Fe³⁺: Alluaudites

     • 1.a.a: ^M(1)M²⁺ dominantly Mn²⁺: Alluaudite

     • 1.a.b: ^M(1)M²⁺ dominantly Fe²⁺: Ferroalluaudite

   • 1.b: ^M(2)M³⁺ dominantly Mn³⁺: Root1 s.l.

     • 1.b.a: ^M(1)M²⁺ dominantly Mn²⁺: Root1

     • 1.b.b: ^M(1)M²⁺ dominantly Fe²⁺: Ferroroot1

2. 0.5 ≤ ^M(2)M^{2+ ≤} 1.5 apfu, and ^A(1)M²⁺ ≤ 0.5 apfu:

   • 2.a. ^M(2)M³⁺ dominantly Fe³⁺, and ^M(2)M²⁺ dominantly Fe²⁺: Hagendorfites

     • 2.a.a: ^M(1)M²⁺ dominantly Mn²⁺: Hagendorfite

     • 2.a.b: ^M(1)M²⁺ dominantly Fe²⁺: Ferrohagendorfite

   • 2.b. ^M(2)M³⁺ dominantly Fe³⁺, and ^M(2)M²⁺ dominantly Mn²⁺: Varulites

     • 2.b.a: ^M(1)M²⁺ dominantly Mn²⁺: Varulite

     • 2.b.b: ^M(1)M²⁺dominantly Fe²⁺: Ferrovarulite

   • 2.c. ^M(2)M³⁺ dominantly Fe³⁺, and ^M(2)M²⁺ dominantly Mg²⁺: Root2 s.l.

     • 2.c.a: ^M(1)M²⁺ dominantly Mn²⁺: Root2

     • 2.c.b: ^M(1)M²⁺ dominantly Fe²⁺: Ferroroot2

3. 0.5 ≤ ^M(2)M^{2+ ≤} 1.5 apfu, and ^A(1)M²⁺ ≥ 0.5 apfu:

   • 3.a. ^A(1)M²⁺ dominantly Ca: Add a suffix “-Ca” to the names mentioned above (point 2).

   • 3.b. ^A(1)M²⁺ dominantly Mn: Add a suffix “-Mn” to the names mentioned above (point 2).

4. ^M(2)M^{2+ ≥} 1.5 apfu, and ^A(1)M²⁺ dominantly Ca:

   • 4.a: ^M(2)M²⁺ dominantly Fe²⁺: Root3 s.l.

     • 4.a.a: ^M(1)M²⁺ dominantly Mn²⁺: Root3

     • 4.a.b: ^M(1)M²⁺ dominantly Fe²⁺: Ferroroot3

   • 4.b: ^M(2)M²⁺ dominantly Mg: Root4 s.l.

     • 4.b.a: ^M(1)M²⁺ dominantly Mn²⁺: Root4

     • 4.b.b: ^M(1)M²⁺ dominantly Fe²⁺: Ferroroot4

   • 4.c: ^M(2)M²⁺ dominantly Mn: Root5 s.l.

     • 4.c.a: ^M(1)M²⁺ dominantly Mn²⁺: Root5

     • 4.c.b: ^M(1)M²⁺ dominantly Fe²⁺: Ferroroot5