Crystal Chemistry of Stanﬁeldite, Ca 7 M 2 Mg 9 (PO 4 ) 12 ( M = Ca, Mg, Fe 2 + ), a Structural Base of Ca 3 Mg 3 (PO 4 ) 4 Phosphors

: Stanﬁeldite, natural Ca-Mg-phosphate, is a typical constituent of phosphate-phosphide assemblages in pallasite and mesosiderite meteorites. The synthetic analogue of stanﬁeldite is used as a crystal matrix of luminophores and frequently encountered in phosphate bioceramics. However, the crystal structure of natural stanﬁeldite has never been reported in detail, and the data available so far relate to its synthetic counterpart. We herein provide the results of a study of stanﬁeldite from the Brahin meteorite (main group pallasite). The empirical formula of the mineral is Ca 8.04 Mg 9.25 Fe 0.72 Mn 0.07 P 11.97 O 48 . Its crystal structure has been solved and reﬁned to R 1 = 0.034. Stanﬁeldite from Brahin is monoclinic, C 2 / c , a 22.7973(4), b 9.9833(2), c 17.0522(3) Å, β 99.954(2) ◦ , V 3822.5(1)Å 3 . The general formula of the mineral can be expressed as Ca 7 M 2 Mg 7 (PO 4 ) 12 ( Z = 4), where the M = Ca, Mg, Fe 2 + . Stanﬁeldite from Brahin and a majority of other meteorites correspond to a composition with an intermediate Ca ≈ Mg occupancy of the M 5A site, leading to the overall formula ~Ca 7 (CaMg)Mg 9 (PO 4 ) 12 ≡ Ca 4 Mg 5 (PO 4 ) 6 . The mineral from the Lunar sample “rusty rock” 66095 approaches the M = Mg end member, Ca 7 Mg 2 Mg 9 (PO 4 ) 12 . In lieu of any supporting analytical data, there is no evidence that the phosphor base with the formula Ca 3 Mg 3 (PO 4 ) 4 does exist.


Introduction
It is known that the speciation of chemical elements in meteoritic substance significantly differs from their speciation in contemporary terrestrial lithosphere [1]. Concerning phosphorus, the main geochemical factors governing the diversity of terrestrial phosphorus-bearing minerals are (1) highly oxidative conditions typical of the present Earth and (2) the aquatic environment, which dramatically multiplies the number of possible pathways for phosphate geosynthesis. Contrary to Earth, the reductive and (in general) water-free conditions that accompanied the formation and early evolution of celestial bodies determined the limited number of meteoritic phosphorus-bearing minerals [2].

Materials and Methods
A piece of the Brahin pallasite was polished and coated with a carbon film for electron microprobe study. SEM imaging ( Figure 1) and microprobe analysis for the main elements were conducted by means of a CamScan 4 scanning electron microscope (SEM) (Cambridge, U.K.) equipped with a LINK AN1000 energy-dispersive analyzer (LINK Analytical, California). The following standards were used: chlorapatite (Ca-K, P-K), enstatite (Mg-K), hematite (Fe-K). The analysis was carried out at 20 kV acceleration voltage, 0.8 nA beam current, 1 μm estimated beam diameter and 60 s live acquisition time per spot. The check-up for minor constituents was performed with a Microspec WDX-2 wavelength-dispersive X-ray spectrometer (Microspec Corporation, California) attached to the same SEM. The Mn content was determined using Mn-Kα line (MnCO3 standard) at 20 kV and 15 nA, whereas the contents of Ni, Co, Na, K and Si were found to lie below the detection limit (less than 0.05 wt.%).
For the purposes of the X-ray structural study, the grain of stanfieldite was extracted from the section and crushed into a few fragments, which were examined under a polarizing microscope in the immersion oil. Several optically homogeneous grains were checked using a Rigaku Oxford diffraction Xcalibur single-crystal diffractometer equipped with a fine-focus sealed tube and graphite monochromator (MoKα, 50 kV, 40 mA). It was found that all checked fragments are optically irresolvable intergrowths, each of them being composed of two or more domains misoriented within 5-10°. The best selected two-domain grain (0.15 × 0.10 × 0.10 mm) was glued onto a plastic loop and subjected to further data collection. A hemisphere of reciprocal space was collected up to 70° at room temperature, and the details are provided in Table 1. Subsequent data processing routines (integration, scaling and SHELX files setup) were performed by means of a CrysAlisPro software (Rigaku Oxford diffraction) [44]. The crystal structure has been solved using an intrinsic phasing approach and refined by means of a SHELX-2018 set of programs [45] incorporated into the Olex2 operation environment [46]. The details of structure refinement are given in Table 1 and in the crystallographic information file (CIF) attached to the Supplementary Materials (S1). The X-ray powder diffraction pattern (Table 2) was obtained with a Rigaku R-AXIS Rapid II difractometer (Rigaku Corporation, Tokyo, Japan) equipped with a curved (semi-cylindrical) imaging plate. A ~150 μm ball was prepared from the stanfieldite powder mixed with an epoxy resin and was picked onto a glass fiber. The image acquisition conditions were: CoKα-radiation, rotating anode with microfocus optics, 40 kV, 15 mA, Debye-Scherrer geometry, r = 127.4 mm, exposure 30 min. The imaging plate was calibrated against Si standard. The image-to-profile data conversion was performed with an osc2xrd program [47]. The unit-cell parameters and occupancies of Mg1-Mg5

Materials and Methods
A piece of the Brahin pallasite was polished and coated with a carbon film for electron microprobe study. SEM imaging ( Figure 1) and microprobe analysis for the main elements were conducted by means of a CamScan 4 scanning electron microscope (SEM) (Cambridge, UK) equipped with a LINK AN1000 energy-dispersive analyzer (LINK Analytical, CA, USA). The following standards were used: chlorapatite (Ca-K, P-K), enstatite (Mg-K), hematite (Fe-K). The analysis was carried out at 20 kV acceleration voltage, 0.8 nA beam current, 1 µm estimated beam diameter and 60 s live acquisition time per spot. The check-up for minor constituents was performed with a Microspec WDX-2 wavelength-dispersive X-ray spectrometer (Microspec Corporation, CA, USA) attached to the same SEM. The Mn content was determined using Mn-Kα line (MnCO 3 standard) at 20 kV and 15 nA, whereas the contents of Ni, Co, Na, K and Si were found to lie below the detection limit (less than 0.05 wt.%).
For the purposes of the X-ray structural study, the grain of stanfieldite was extracted from the section and crushed into a few fragments, which were examined under a polarizing microscope in the immersion oil. Several optically homogeneous grains were checked using a Rigaku Oxford diffraction Xcalibur single-crystal diffractometer equipped with a fine-focus sealed tube and graphite monochromator (MoKα, 50 kV, 40 mA). It was found that all checked fragments are optically irresolvable intergrowths, each of them being composed of two or more domains misoriented within 5-10 • . The best selected two-domain grain (0.15 × 0.10 × 0.10 mm) was glued onto a plastic loop and subjected to further data collection. A hemisphere of reciprocal space was collected up to 70 • at room temperature, and the details are provided in Table 1. Subsequent data processing routines (integration, scaling and SHELX files setup) were performed by means of a CrysAlisPro software (Rigaku Oxford diffraction) [44]. The crystal structure has been solved using an intrinsic phasing approach and refined by means of a SHELX-2018 set of programs [45] incorporated into the Olex2 operation environment [46]. The details of structure refinement are given in Table 1 and in the crystallographic information file (CIF) attached to the Supplementary Materials (S1). The X-ray powder diffraction pattern ( Table 2) was obtained with a Rigaku R-AXIS Rapid II difractometer (Rigaku Corporation, Tokyo, Japan) equipped with a curved (semi-cylindrical) imaging plate. A~150 µm ball was prepared from the stanfieldite powder mixed with an epoxy resin and was picked onto a glass fiber. The image acquisition conditions were: CoKα-radiation, rotating anode with microfocus optics, 40 kV, 15 mA, Debye-Scherrer geometry, r = 127.4 mm, exposure 30 min. The imaging plate was calibrated against Si standard. The image-to-profile data conversion was performed with an osc2xrd program [47]. The unit-cell parameters and occupancies of Mg1-Mg5 sites were refined by the Rietveld method (Table 1, Figure 2) using Bruker TOPAS v. 5.0 software (Bruker Inc., Wisconsin). The occupancies at the M5A site were fixed at the values determined by single-crystal refinement. The atomic coordinates were not refined but were fixed according to single-crystal data. The XPRD pattern ( Table 2)  The micro-Raman spectrum was recorded from a random powder sample using a Horiba Jobin-Yvon LabRam 800 instrument (HORIBA Jobin Yvon GmbH, Bensheim, Germany), equipped with a 50× confocal objective. The instrument was operated with a 514 nm Ar + laser at a 1 nm lateral resolution and 2 cm −1 spectral resolution. The optics were preliminarily calibrated using a Si reflection standard.

Stanfieldite: A Complete Structure-Composition Dataset
The X-ray examination of Brahin stanfieldite was carried out by two different methods. Both single-crystal and Rietveld refinements of the unit cell had led to almost identical parameters (deviation between the unit-cell volumes is 0.04%, Table 1). The refined Mg/Fe occupancies were also well converged (  (Table 3) are herein calculated with a good confidence.
Fuchs [9], in 1969, could reliably determine the unit-cell metrics, but misrecognized the space group of the mineral (Table 4), perhaps due to the same pseudo-twinning of the crystals [40,41] which we observed on our studied stanfieldite.  Dickens and Brown [40] have synthesized the synthetic, Fe-free analogue of stanfieldite and thoroughly described its crystal structure. However, the latter authors did not perform independent determination of the chemical composition of synthesized material-as one will see, this is an essential requirement in view of the widely varying composition of at least one structural site of stanfieldite. Steele and Olsen [41] have reported the preliminary results of structural examination of natural stanfieldite from the Imilac pallasite. They gave the analytical chemical formula of the mineral, but did not provide full structural data, confining the results to unit cell metrics, average bond lengths and selected site occupancies (Tables 3 and 4). As a consequence, no complete structure-composition dataset for stanfieldite is available so far, and the data provided herein are the first report of that type.

General Features of Stanfieldite Structure and Its Formula
The crystal structure of stanfieldite is a complex framework composed of 10 metal sites and 6 phosphate groups (Table 3, Figure 3, Supplementary Table S1). Dickens and Brown [40] gave the detailed description of each site in the structure of synthetic analogue of the mineral, and the present paragraph aims to overview stanfieldite structure and highlight its features. The most interesting one is a pseudo-hexagonal character of the framework which can be best viewed via the arrangement of [PO 4 ] tetrahedra along the [10-2] axis ( Figure 3A). In principle, stanfieldite, being presented by the oversimplified formula M 3 (PO 4 ) 2 (M = Mg, Ca; Z = 24), can be regarded as a derivative of the well-known glaserite structure type, K 3 Na(SO 4 ) 2 [49,50]. Dickens and Brown [40] discuss the relationships between stanfieldite and glaserite-related phosphates belonging to αand β-Ca 3 (PO 4 ) 2 structural types. The latter is known as a basement of whitlockite-group mineral structures [51], two of which, merrillite and ferromerrillite, are of fundamental importance in the mineralogy of meteorites [3,4]. In view of the common and intimate association of stanfieldite and merrillite in pallasite meteorites (Figure 1), these relationships could be of particular interest. However, contrary to Dickens and Brown [40], we would not overestimate the similarity of stanfieldite and merrillite structures. The unusual face-sharing of adjacent [MO 6 ] octahedron and [PO 4 ] tetrahedron characteristic of merrillite [3] does not occur in stanfieldite structure.
The refinement of occupancies of four Ca-sites in the stanfieldite structure showed no evidence for either Mg or Fe substitution. However, refinement of four Mg sites using both single-crystal and Rietveld methods concordantly leads to a partial substitution of Mg for Fe, with iron being preferentially concentrated in Mg1 (tetrahedral) and Mg5 (octahedral) positions ( Table 3). The tetrahedral coordination of Mg1 is highly unusual; however, it is sometimes encountered in mineral structures such as åkermanite, Ca 2 MgSi 2 O 7 (melilite structure type), and spinel. The M5A site allows mixed occupancy by Ca, Mg and Fe, and thus will be discussed in the next section. Based on the structural data, the overall formula of stanfieldite can be written as Ca 7 M 2 Mg 9 (PO 4 ) 7 , where M = Ca, Mg or Fe 2+ . Crystals 2020, 10, x 8 of 14

The M5A Site, A Key to A Flexibility of Stanfieldite Composition
This cation site (which was previously referred to as Ca5 [40]) deserves a special discussion as it determines the variability of stanfieldite composition which, in turn, has led to misinterpretations of the chemical formula of the mineral and its synthetic analogues. The central atom resides in the general 8f position and coordinates to six oxygen atoms to form a highly distorted octahedron ( Figure 3C), with the bond distances varying from 2.1 to almost 2.5 Å (Table 5).

The M5A Site, A Key to A Flexibility of Stanfieldite Composition
This cation site (which was previously referred to as Ca5 [40]) deserves a special discussion as it determines the variability of stanfieldite composition which, in turn, has led to misinterpretations of the chemical formula of the mineral and its synthetic analogues. The central atom resides in the general 8f position and coordinates to six oxygen atoms to form a highly distorted octahedron ( Figure 3C), with the bond distances varying from 2.1 to almost 2.5 Å (Table 5). The M5A octahedra form paired clusters in the structure via corner-linking by phosphate tetrahedra ( Figure 3C). Based on previous reports [40,41,52] and our data (Table 5), M5A may accommodate Ca, Mg, Fe 2+ and Ni in different proportions, with the total occupancy equal to unity. Natural stanfieldite is a Mg-dominant mineral, and the refinement of M5A occupancy leads to a dominance of Mg over Fe 2+ as well ( Table 5). The latter is supported by calculation of the bond-valence sum, which is almost identical to that of synthetic analogue of stanfieldite ( Table 3). The variability of M5A occupancy substantiates the existence of solid solution between hypothetical Mg and Ca end members.
The former would have the composition corresponding to Ca 7 Mg 2 Mg 9 (PO 4 ) 12 . The latter member would correspond to Ca 7 Ca 2 Mg 9 (PO 4 ) 12 , that is equal to Ca 3 Mg 3 (PO 4 ) 4 . The intermediate composition having Ca = Mg in M5A results in a formula Ca 7 (CaMg)Mg 9 (PO 4 ) 12 , or, in a simplified form, Ca 4 Mg 5 (PO 4 ) 4 . One can see that the latter perfectly fulfils the ideal composition of stanfieldite proposed by Fuchs [9]. It is noteworthy that stanfieldite from Brahin described herein, the previously reported mineral from Imilac [41] and the synthetic analogue of stanfieldite [40] have M5A occupancies almost equally shared between Ca and (Mg + Fe) ( Table 5). This could lead to the assumption that the ordering between Ca and Σ(Mg, Fe) might exist in the M5A site. However, neither our observations nor previously reported data reveal superstructure reflections which would evidence the Ca/Mg ordering. In this respect, an overview of reported compositions of stanfieldite-like minerals and compounds would be of special interest. We have collected the chemically relevant data which are gathered in Table 6 and plotted in Figure 4. It can be seen that the overwhelming majority of stanfieldite compositions fall within the range corresponding to Ca ≈ (Mg + Fe) in the M5A site. Therefore, the above assumption on the possible Ca/Mg ordering, albeit speculative, has a statistically substantiated basis. The M5A octahedra form paired clusters in the structure via corner-linking by phosphate tetrahedra ( Figure 3C). Based on previous reports [40,41,52] and our data (Table 5), M5A may accommodate Ca, Mg, Fe 2+ and Ni in different proportions, with the total occupancy equal to unity. Natural stanfieldite is a Mg-dominant mineral, and the refinement of M5A occupancy leads to a dominance of Mg over Fe 2+ as well ( Table 5). The latter is supported by calculation of the bond-valence sum, which is almost identical to that of synthetic analogue of stanfieldite (Table 3). The variability of M5A occupancy substantiates the existence of solid solution between hypothetical Mg and Ca end members. The former would have the composition corresponding to Ca7Mg2Mg9(PO4)12. The latter member would correspond to Ca7Ca2Mg9(PO4)12, that is equal to Ca3Mg3(PO4)4. The intermediate composition having Ca = Mg in M5A results in a formula Ca7(CaMg)Mg9(PO4)12, or, in a simplified form, Ca4Mg5(PO4)4. One can see that the latter perfectly fulfils the ideal composition of stanfieldite proposed by Fuchs [9]. It is noteworthy that stanfieldite from Brahin described herein, the previously reported mineral from Imilac [41] and the synthetic analogue of stanfieldite [40] have M5A occupancies almost equally shared between Ca and (Mg + Fe) ( Table 5). This could lead to the assumption that the ordering between Ca and Σ(Mg, Fe) might exist in the M5A site. However, neither our observations nor previously reported data reveal superstructure reflections which would evidence the Ca/Mg ordering. In this respect, an overview of reported compositions of stanfieldite-like minerals and compounds would be of special interest. We have collected the chemically relevant data which are gathered in Table 6 and plotted in Figure 4. It can be seen that the overwhelming majority of stanfieldite compositions fall within the range corresponding to Ca ≈ (Mg + Fe) in the M5A site. Therefore, the above assumption on the possible Ca/Mg ordering, albeit speculative, has a statistically substantiated basis.  Table 6.  4 . The blue dots and labels mark the compositions of particular interest which are discussed in the paper. References and source data are given in Table 6. The next interesting point is a significant departure of total cationic sums of many analyses from the ideal value requiring 18 cations per formula unit. These departures are readily revealed by the shifts of corresponding analytical points from the linear fit in Figure 4. At present, we have no explanation for the observed departures. They could imply the existence of analytical errors in the reported microprobe data. On the other hand, these shifts might mean the occurrence of vacancies in cationic sites of stanfieldite structure, and then they deserve a special investigation.
Although the majority of reported data fall within the central area of the plot in Figure 4, there are a few points showing significant prevalence of (Mg + Fe) sum over total Ca. These include one analysis from the Eagle Station pallasite [7] and the mineral found in the Lunar sample 66095 returned by the Apollo 16 mission [16]. These two analyses approach the Ca 7 Mg 2 Mg 9 (PO 4 ) 12 end-member of the M5A solid solution. At the opposite extreme of the plot, there is a single point approaching hypothetical Ca 3 M 3 (PO 4 ) 4 composition. This analysis, along with two more listed in Table 6, relate to a stanfieldite-like phosphate described from the ancient slags found in Tyrol [19]. The main feature of this compound is wide variations both in Ca/(Fe + Mg) and Fe/Mg ratios, up to nearly Fe-dominant compositions. Schneider, with co-authors [19], has provided Raman spectrum for this phosphate, but in the absence of Raman spectra for genuine stanfieldite, the comparison was not possible. We herein provide the Raman spectrum of stanfieldite from the Brahin meteorite ( Figure 5). A comparison of this spectrum with that reported by Schneider with co-authors [19] shows that the latter can represent a poorly crystallized Fe-dominant analogue of stanfieldite.
Crystals 2020, 10, x 11 of 14 not possible. We herein provide the Raman spectrum of stanfieldite from the Brahin meteorite ( Figure 5). A comparison of this spectrum with that reported by Schneider with co-authors [19] shows that the latter can represent a poorly crystallized Fe-dominant analogue of stanfieldite.

Ca3Mg3(PO4)4 Phosphors: Do They Exist?
In this section, we would like to clarify the mistake caused by the incorrect database assertion of primary structural data on the synthetic analogue of stanfieldite reported by Dickens and Brown [40]. In the title of their article, Dickens and Brown report the formula Ca7Mg9(Ca,Mg)2(PO4)12, with Ca = Mg in the Ca5 site, leading to a bulk one Ca4Mg5(PO4)6. It looks obvious that the mistake was introduced in the stage of structural data transfer from the article tables to the ICSD database. The Ca5 site, equally occupied by Ca and Mg [40], was erroneously assigned to be fully occupied by Ca. The latter had led to a wrong formula, Ca3Mg3(PO4)4, which still appears in the ICSD database [53] (ICSD code 23642). Moreover, the calculated X-ray powder diffraction pattern has been further included into the ICDD (JCPDS) database under the reference number JCPDS-ICDD 73-1182 ( Figure  4). There is a substantial interest to the family of luminophores (phosphors) based on the stanfieldite structure [35][36][37][38][39]. It is erroneous that the mistake in the chemical formula caused by the incorrect primary data transfer has passed first to the ICDD database and then to the papers devoted to a study of these phosphor materials [35][36][37][38][39]. Unfortunately, neither of the published articles does contain quantitative chemical data on synthesized phosphors. Thus, in lieu of any evidence supporting the existence of Ca3Mg3(PO4)4, one can state that these phosphors are in fact stanfieldite-based, Ca4Mg5(PO4)4 compounds.

Supplementary Materials:
The following are available online at www.mdpi.com/xxx/S1: Supplementary crystallographic data for stanfieldite from the Brahin meteorite in Crystallographic Information File (CIF) format. Alternatively, CCDC reference number 1998335 contains the same data, and it can be obtained free of charge from the Cambridge Crystallographic Data Centre at www.ccdc.cam.ac.uk. S2, supplementary crystallographic tables S1-S3.

Ca 3 Mg 3 (PO 4 ) 4 Phosphors: Do They Exist?
In this section, we would like to clarify the mistake caused by the incorrect database assertion of primary structural data on the synthetic analogue of stanfieldite reported by Dickens and Brown [40]. In the title of their article, Dickens and Brown report the formula Ca 7 Mg 9 (Ca,Mg) 2 (PO 4 ) 12 , with Ca = Mg in the Ca5 site, leading to a bulk one Ca 4 Mg 5 (PO 4 ) 6 . It looks obvious that the mistake was introduced in the stage of structural data transfer from the article tables to the ICSD database. The Ca5 site, equally occupied by Ca and Mg [40], was erroneously assigned to be fully occupied by Ca. The latter had led to a wrong formula, Ca 3 Mg 3 (PO 4 ) 4 , which still appears in the ICSD database [53] (ICSD code 23642). Moreover, the calculated X-ray powder diffraction pattern has been further included into the ICDD (JCPDS) database under the reference number JCPDS-ICDD 73-1182 ( Figure 4). There is a substantial interest to the family of luminophores (phosphors) based on the stanfieldite structure [35][36][37][38][39]. It is erroneous that the mistake in the chemical formula caused by the incorrect primary data transfer has passed first to the ICDD database and then to the papers devoted to a study of these phosphor materials [35][36][37][38][39]. Unfortunately, neither of the published articles does contain quantitative chemical data on synthesized phosphors. Thus, in lieu of any evidence supporting the existence of Ca 3 Mg 3 (PO 4 ) 4 , one can state that these phosphors are in fact stanfieldite-based, Ca 4 Mg 5 (PO 4 ) 4 compounds.