Na4Fe2+Fe3+(PO4)3, a new synthetic NASICON-type phosphate

This paper reports the crystal structure of tetrasodium diiron tris(phosphate), Na4Fe2+Fe3+(PO4)3, which has been synthesized hydrothermally at 773 K and 0.1 GPa. The crystal structure has been refined in the space group R c and is identical to that of γ-NASICON. The heteropolyhedral framework is based on a regular alternation, in three dimensions, of corner-sharing PO4 tetrahedra and FeO6 octahedra, constituting so-called ‘lantern units’ stacked along the c axis. The Na+ cations are distributed over two crystallographic sites: the six-coordinated Na1 site which lies between two ‘lantern units’, and the eight-coordinated Na2 site which lies at the same z value as the P site.


Comment
In the natural geological environment of granitic pegmatites, Na-Fe-Mn-bearing phosphates play important geochemical and petrological roles. The alluaudite group of minerals, with an idealized chemical composition Na 2 (Mn,Fe 2+ ) 2 Fe 3+ (PO 4 ) 3 , constitutes a good example of primary phosphates which can be used as geothermometer, or to constrain the oxygen fugacity which prevailed in granitic pegmatites (Hatert et al., 2006). In order to better understand the crystallization conditions of iron-rich alluaudites in pegmatites, we decided to investigate the Na-Fe 2+ -Fe 3+ (+PO 4 ) ternary system by hydrothermal methods (Hatert & Fransolet, 2006). These experiments produced several new phosphates, which crystallized in the Na-rich part of the system and were investigated by single-crystal X-ray diffraction techniques (Hatert, 2007a,b). Starting from the composition Na 4 Fe 2+ Fe 3+ (PO 4 ) 3 , the hydrothermal synthesis at 500°C and 0.1 GPa produced large pink crystals; their crystal structure is reported herein.
The crystal structure of the title compound has been refined in space group R3c and corresponds to that of centrosymmetric γ-NASICON-type phosphates (Sljukic et al., 1969;Masquelier et al., 2000). The heteropolyhedral framework is based on the regular alternation, in the three dimensions, of corner-sharing PO 4 tetrahedra (P-O = 1.533-1.538 Å) and FeO 6 octahedra (Fe-O = 2.010-2.130 Å; Fig. 1), and shows the stacking, along the c direction, of the so-called 'lantern units' (Masquelier et al., 2000). The monovalent Na + cations are distributed over two crystallographic sites: the 6-coordinated Na1 site (Na1-O = 2.402 Å) which lies between two 'lantern units', and the 8-coordinated Na2 site (Na2-O = 2.487-2.990 Å) which lies at the same z value as the P atom (Fig. 2).
Bond-valence sums were calculated for each ion using the parameters of Brown & Altermatt (1985). The P1 bond-valence sum is 4.99, and the O-atom bond-valence sums are within the normal acceptable range (1.93-2.10). The bond-valence sums also confirm that the Na1 and Na2 sites are filled by Na (0.98-1.19), and that the Fe1 site contains an equal amount of Fe 2+ and Fe 3+ (2.54).
A comparison with the crystal structure of γ-Na 3 Fe 2 (PO 4 ) 3 (Masquelier et al., 2000) shows that the two phosphates are isostructural. However, the amount of Na in the title compound reaches 4 atoms per formula unit, and is higher than the Na-content of any other known NASICON-type phosphate. This high Na-content is necessary to maintain charge balance, since the Fe1 site is occupied by 50% Fe 2+ and 50% Fe 3+ . Na atoms are located on the same positions as in γ-Na 3 Fe 2 (PO 4 ) 3 , but the two Na1 and Na2 sites are completely filled by Na atoms in the title compound, whereas their occupancy factors are 0.85 (3) (Na1) and 0.72 (3) (Na2) in γ-Na 3 Fe 2 (PO 4 ) 3 (Masquelier et al., 2000). Masquelier et al. (2000) demonstrated that the size of the Na1 cavity is increased as this site is depopulated, and that this increase has a direct influence on the value of the c unit-cell parameter. This hypothesis is corroborated by our structural data, which show that a full occupancy of the Na1 site in Na 4 Fe 2+ Fe 3+ (PO 4 ) 3 induces small values for the Na1-O distances supplementary materials sup-2 and for the c parameter (2.402 and 21.280 Å, respectively), while a partial occupancy of this site in γ-Na 3 Fe 2 (PO 4 ) 3 induces larger values for these parameters (2.500 and 21.808 Å, respectively; Masquelier et al., 2000). The presence of significant amounts of Fe 2+ in the Fe1 site of the title compound also induces a significant increase of the Fe1-O average bond length (2.070 Å), when compared to the Fe1-O bond length reported for γ-Na 3 Fe 2 (PO 4 ) 3 (2.002 Å; Masquelier et al., 2000). This increase does not affect the c unit-cell parameter, but induces a significant increase of the a parameter, from 8.727 Å in γ-Na 3 Fe 2 (PO 4 ) 3 (Masquelier et al., 2000), to 8.954 Å in the title compound.

Experimental
The title compound was synthesized under hydrothermal conditions. The starting material was prepared by mixing NaH 2 PO 4 .H 2 O, Na 2 HPO 4 .2H 2 O, FeO and Fe 2 O 3 in proportions 2:1:1:1/2. About 25 mg of the homogenized mixture was sealed into a gold tube with an outer diameter of 2 mm and a length of 25 mm, containing 2 mg of distilled water. The gold capsule was then inserted in a Tuttle-type pressure vessel (Tuttle, 1949) and maintained at a temperature of 500°C and a pressure of 0.1 GPa. After 3 d, the sample still in the gold tube in the autoclave, was quenched to room temperature in a stream of cold air. The synthesized phosphates consisted of large pink crystals of the title compound, associated with colourless crystals of maricite, NaFe 2+ (PO 4 ), and with isometric black crystals of Na 7 Fe 4 (PO 4 ) 6 .
A chemical analysis of the title compound has been performed with a CAMEBAX SX-50 electron microprobe (

Data collection
Bruker P4 diffractometer 284 reflections with I > 2σ(I) Refinement. Refinement of F 2 against ALL reflections. The weighted R-factor wR and goodness of fit S are based on F 2 , conventional R-factors R are based on F, with F set to zero for negative F 2 . The threshold expression of F 2 > σ(F 2 ) is used only for calculating Rfactors(gt) etc. and is not relevant to the choice of reflections for refinement. R-factors based on F 2 are statistically about twice as large as those based on F, and R-factors based on ALL data will be even larger.