- E. Schweizerbart'sche Verlagsbuchhandlung. D-70176 Stuttgart
The crystal structure of isolueshite, a recently described member of the perovskite group from the Khibina complex (Russia), was refined by single-crystal methods in the space group Pm3̄3m [a = 3.909(1) Å] to R = 0.031 and wR = 0.061. The previously proposed cubic symmetry of isolueshite was confirmed. In contrast to the known perovskite-group minerals crystallizing with the cubic symmetry, isolueshite shows a disordered arrangement of oxygen atoms in a Pm3̄3m-type cell. In isolueshite, the oxygen atoms are disordered from their “ideal” site at 3c (1/2, 0, 1/2) over four 12h sites (x, 0, 1/2) with a refined x of 0.579(5). The refinement was based on the structural formula (Na0.75La0.19Ca0.06)∑1.00(Nb0.50Ti0.50)∑1.00O3.00 closely corresponding to the empirical formula of the mineral. The presence of hydroxy1 groups in the structure of isolueshite was confirmed by infrared spectroscopy (absorption lines at 3450 cm−1 and 1080 cm−1). Structural formulae of the mineral accounting for the presence of (OH)−1 groups are given. A synthetic compound of the composition (Na0.75La0.25) (Nb0.50Ti0.50)O3 was prepared using the ceramic technique (final heating at 1200°C), and shown not to crystallize with cubic symmetry. The crystal structure of this compound is best refined (Rietveld method) in orthorhombic symmetry [Cmcm, a = 7.7841(9), b = 7.8045(2), c = 7.7831 (9) Å]. This structure is derived from the ideal perovskite lattice (Pm3m) by rotation of the (Nb,Ti)O6 octahedra about two tetrad axes of the cubic subcell (tilt system a0b+c). Our data and, in particular, the presence of oxygen disorder (incongruent octahedral tilting) in the structure of isolueshite, suggest that stabilization of the cubic symmetry of this mineral may be controlled by thermodynamic or kinetic factors (crystallization temperature, pressure or cooling rate), rather than by the isomorphic substitutions involving La and Ti. Hence, isolueshite should probably be considered a quenched or “frozen” polymorph of NaNb03.
The new mineral isolueshite, ideally (Na,La,Ca) (Nb,Ti)O3, has been recently described from a pegmatite vein at the Khibina alkaline complex, Kola Peninsula, Russia (Chakhmouradian et al., 1997). Isolueshite belongs to the perovskite group, and is dimorphous with orthorhombic lueshite which is a relatively common accessory mineral in carbonatites (Safiannikoff, 1959; Chakhmouradian & Mitchell, 1997, 1998). The cubic symmetry (space group Pm3̄3m) of isolueshite has been proposed on the basis of the following observations (Chakhmouradian et al., 1997):
the mineral is optically isotropic;
in contrast to orthorhombic lueshite, loparite and some other perovskite-group minerals, the X-ray powder diffraction (XRD) pattern shows no superlattice lines indicative of lower-than-cubic symmetry;
single-crystal XRD studies by oscillating and de Jong-Bouman methods support the cubic symmetry of isolueshite, as no superlattice reflections were observed.
Isolueshite crystals are complexly-zoned and show progressive enrichment in LREE and Ti towards the rim. The complete intra- and intergranular compositional range of this mineral can be adequately characterized in terms of two perovskite-type end-members NaNbO3 and Na0.5La0.5TiO3, a lanthanian analogue of loparite-(Ce). At room temperature, both of these endmembers have distorted perovskite structures derived from the cubic parent by octahedral rotation, and in NaNbO3, by displacement of cations from their ideal positions (Sakowski-Cowley et al., 1969; Mitchell & Chakhmouradian, 1998). Therefore it is expected that the members of the solid solution series between NaNbO3 and Na0.5La0.5TiO3 also have structures departing from the ideal cubic parent.
The present work was initiated to study the structure of isolueshite and, perhaps, explain the apparent discrepancy between the mineralogical observations and logical expectations. In this study, we describe and compare the crystal structures of naturally-occurring isolueshite, and a synthetic compound of nearly analogous composition. As the synthetic compound proved not to have the ideal cubic structure, we refer to it as a compositional, rather than structural, analogue.
Experimental and analytical methods
For the X-ray diffraction (XRD) data collection, we used an isometric crystal of isolueshite measuring 0.4 × 0.4 × 0.4 mm. The crystal was first examined by Laue, oscillating and Weissenberg methods using CuKα radiation. Careful analysis of the obtained photographs revealed no indication of lower-than-cubic symmetry or twinning, in agreement with the data of Chakhmouradian et al. (1997). Intensities were collected on an automated four-circle diffractometer Syntex P21. Integral intensities were measured by ω-2𝛉 scans and then converted to structure factors by applying the Lorentz and polarization correction corrections (software package AREN: Andrianov, 1986). Absorption correction was done with the method DIFABS (Walker & Stewart, 1983). The structure of isolueshite was refined using the program SHELXL-93 (Sheldrick, 1993). Initially, the structure was refined on the basis of the ideal perovskite lattice ABO3 with A (Na,La,Ca), B (Nb,Ti) and O in Wyckoff positions 1a, 1b and 3c, respectively. This refinement converged to R1 = 0.052 and wR2 = 0.143. A large thermal parameter for the oxygen atom obtained in the initial refinement (Uiso = 0.067 Å2) suggested that either the symmetry of the structure was lower than cubic or that the oxygen atoms were displaced from their special positions in a disordered manner. As the single-crystal study of isolueshite did not reveal any evidence of reduced symmetry, we tested an alternative structural model in which oxygen atoms were disordered. It is noteworthy that a disordered distribution of oxygen atoms has been described in a number of perovskite-type phases (see below).
The disorder of oxygen atoms in isolueshite was approximated by placing them in twelve-fold positions 12h (x, 0, 1/2), and thus allowing the x coordinate of the oxygen atom to vary (site occupancy was fixed at the ideal value of 0.25). The structural refinement based on this model gave improved R1 and wR2: 0.031 and 0.061, respectively, for the observed data. Note that due to statistical reasons, wR2 is normally two-three times larger than R1 (Sheldrick, 1993). The structural and thermal parameters, of isolueshite were refined by the full-matrix least squares method. Details of the data collection, intensity correction and structure refinement are given in Table 1.
A compositional analogue of isolueshite was synthesized from stoichiometric amounts of Na2CO3, La2O3, Nb2O5 and TiO2. (high purity grade). The oven-dried reagents were mixed, ground in an agate mortar and heated in air for 24 h initially at 1000°C to avoid the loss of Na. After regrinding, the sample was heated for 48 h at 1200°C and then rapidly cooled in air to room temperature. The synthesized compound was a yellowish white polycrystalline aggregate with the size of individual crystals less than 20 μm. As no large single crystals were available, the crystal structure of the compound was refined from an XRD powder pattern using the Rietveld full-profile refinement method. The powder pattern was obtained on a Philips 3710 diffractometer, and then analyzed by the Rietveld method using the program FULLPROF (Rodriguez-Carvajal, 1990). The details of the data collection and structure refinement are given in Table 2.
Crystal structure of isolueshite and synthetic (Na0.75La0.25)(Nb0.50Ti0.50)03
The refined structure of isolueshite has the A- and B-site cations in their “ideal” positions (1a and 1b, respectively), and the oxygen atoms at 12h. For the oxygen atoms, the refinement gave x = 0.579 (5) and Uiso = 0.029(3) Å2. The significance of anisotropic parameters for (Nb,Ti) and oxygen atoms was examined using the Hirshfeld's (1976) test on the bond rigidity based on estimation of a difference between the mean square amplitudes of oscillations of atoms B (i.e. Nb and Ti) and O parallel to the B-O bond, ΔBO = Z2(BO) - Z2(OB). It is considered that thermal ellipsoids for B and O reflect the character of atom motions properly if ΔBO is lower than 10−4 Å2. For the (Nb,Ti)-O bond parallel to  in the structure of isolueshite, U22(Nb,Ti) = 0.0128 Å2 and U22(O) = 0.013 Å2, which gives ΔBO = 5 · 10−6 Å2. The obtained value is obviously much lower than the limiting parameter of 10−4 Å2.
The most intriguing result obtained in the present study is splitting of the “ideal” oxygen position at 3c (1/2, 0, 1/2) into four 12h sites with coordinates (x, 0, 1/2) (Table 3). The stoichiometry of perovskite-type compounds ABO3 does not allow for a complete occupancy of this position, and, generally, only 25% of the 12h site can be occupied by oxygen. A disordered arrangement of oxygen atoms in perovskite-type compounds was first described by Iyer & Smith (1967). These authors studied a single crystal of La1/3TaO3 and noted that peaks obtained for the oxygen atoms using Fourier synthesis were toroidal in shape and had a density minimum at the special positions. When the structure of La1/3TaO3 was refined with the oxygen atoms fixed to the special positions, i.e. complete ordering of oxygen was assumed, no meaningful thermal parameters for these atoms could be obtained (Iyer & Smith, 1967). Subsequently, anion disorder has been described in some other perovskites, e.g. K2Nd2Ti3O10 (Amow & Greedan, 1998), and CsPbCl3 (Ahtee et al., 1980). Splitting of the x coordinate of ligands (O, F, Cl) in the ideal perovskite structure (Pm3̄3m) is normally observed close to a point of phase transition to a lower symmetry (Tsire'son et al., 1991). In other words, structures with a disordered arrangement of anions can be interpreted as transitional between the ideal undistorted highly symmetrical structures and those generated by rotation of BO6 octahedra in the lattice. This explains why anion disordering in perovskites is also termed incoherent octahedral tilting (e.g. Amow & Greedan, 1998). As many perovskite-type compounds crystallize as metastable cubic structures, incoherent octahedral tilting is probably much more common than has been realized. However, the low scattering factors of oxygen and fluorine complicate detection and quantitative characterization of such structural phenomena. Atomic coordinates, thermal displacement parameters and selected interatomic distances for isolueshite are given in Table 3. The crystal structure depicted in terms of thermal displacement ellipsoids is shown in Fig. 1.
The XRD pattern of this compound shows several low-intensity peaks which cannot be indexed on a cubic cell. Complete indexing can be done on a tetragonal cell (I4/mcm, a ≈ ap and c ≈ 2ap). Transition of the cubic perovskite-type structure into such tetragonal derivative involves rotation of the BO6, octahedra about one of the tetrad axes of the pseudocubic subcell. The corresponding octahedral tilt is classified as a°a°c− in Glazer's (1972) notation. However, the refinement of the XRD pattern in the tetragonal model gives an unacceptable high isotropic thermal parameter for O2 (Table 4). Consequently, a number of alternative structural models were tried, which would involve two octahedral tilts in the perovskite-type lattice. These included the space groups (corresponding tilt systems are given in parentheses) Imma (a°b−b−), I2/m (a°b−b−c−) and Cmcm (a°b+c-). We also attempted to refine the XRD pattern in the space group Pnma (a+b−b−), as most orthorhombic perovskite-type compounds, including naturally-occurring perovskite, neighborite, loparite-(Ce) and latrappite, crystallize in this space group (Hu et al., 1992; Thomas, 1996; Woodward, 1997a; Chakhmouradian & Mitchell, 1997; Mitchell et al., 1998). However, the refinement based on a Pnma cell did not converge. Initial atomic coordinates pertaining to each of the examined models, were taken from the work of Woodward (1997b). The best correspondence between the observed and calculated XRD data was achieved with the space group Cmcm (a ≈ b ≈ c ≈ 2ap) (Table 4). However, diffraction lines characteristic of Cmcm, but absent in I4/mcm (e.g. at d ≈ 3.5 and 2.6 Å), are indistinguishable from the background and cannot be used to assign the space group unequivocally. The indexing using both space groups is given in Table 5.
The refined atomic coordinates, isotropic thermal parameters and unit-cell dimensions for synthetic (Na0.75La0.25)(Nb0.50Ti0.50)03 are reported in Table 6. The table also includes selected interatomic distances for this compound. From Table 6, it is obvious that Na and La are distributed over two types of A-sites. In our model, the A2 site is somewhat larger than A1, although both sites provide a nearly ten-fold coordination for the large cation. Cation occupancies for Na and La refined from the XRD data demonstrate no preference in accommodation of either Na or La at any of the available A-sites. Thus, (Na0.75La0.25)(Nb0.50Ti0.50)O3 shows no cation ordering, in agreement with the absence of superlattice peaks on the XRD pattern of this compound. The BO6 polyhedron is distorted, with the mean (Nb,Ti)-O distance 1.969 Å. This value approaches the sum of radii for the corresponding ions (1.972 Å), assuming 1/2Nb5+ and 1/2Ti4+ at the B-site (radii from Shannon, 1976).
Other perovskite-type phases which crystallize in the space group Cmcm, include NaNbO3 (T1 phase), NaTaO3 and SrZrO3 (Ahtee et al., 1972, 1978; Ahtee & Darlington, 1980). In contrast to (Na0.75La0.25)(Nb0.50Ti0.50)O3, the Cmcm modifications of these compounds are stable at high temperature, and occur in a series of phase transitions between a less symmetric room-temperature form involving three octahedral tilts (Pbma or Pnma) and a single-tilt tetragonal structure (P4mbm or I4/mcm) (Sakowski-Cowley et al., 1969; Ahtee et al., 1978; Ahtee & Darlington, 1980). It is unlikely that a simple perovskite-type compound with one type of A cations adapts the Cmcm structure characterized by two non-equivalent A-sites. The perovskite synthesized in the present study contains two types of A cations differing in size and charge; consequently, crystallization of this compound in the space group Cmcm does not violate Pauling's rule of parsimony (Woodward, 1997a).
Infrared spectroscopy and role of water in the composition of isolueshite
An infrared transmission spectrum of isolueshite (Fig. 2) was recorded from a powder in the range 400-4000 cm−1 with a spectrophotometer Specord 75 IR. The powder sample was pressed into a KBr pellet at 70 kbar pressure. The spectrum shows a strong absorption band at 630 cm−1 corresponding to the v1 stretching vibration of (Nb,Ti)-O bonds (Last, 1957; Pilipenko et al., 1971). A similar absorption band is observed in the spectra of perovskite-type minerals and synthetic compounds (Pilipenko et al., 1971; Kaleveld et al., 1973; Sych et al., 1973), but its frequency range varies depending on the Nb/Ti ratio. In the spectrum of isolueshite, the band is relatively symmetric, and does not show splitting indicative of reduction in local symmetry of the (Ti,Nb)O6 octahedra (Kaleveld et al., 1973; Sych et al., 1973). A broad absorption band at approximately 3450 cm-1 corresponds to the stretching mode of O-H bonds. This band is present in hydrated perovskite-type compounds such as HNbWO6, but is not observed in anhydrous phases such as NbWO5.5 (Bhat & Gopalakrishnan, 1986). The spectrum of isolueshite also includes a weak absorption band at 1080 cm-1 [bending vibrations of the (Nb,Ti)-OH bond], but does not show absorption at 1620-1640 cm-1 (bending mode of water), suggesting that the mineral contains hydroxyl groups, but no molecular H2O. Note that molecular water sorbed by the KBr pellet would produce absorption lines at different frequencies.
Hydrogen-doped (“protonated” compounds are well known among synthetic perovskites. These have the general formula BO3-(OH)x or HxBO3 (B = Re, W, Mo, Nb, Ta), and are structurally related to ReO3 (Bhat & Gopalakrishnan, 1986; Birtill & Dickens, 1978; Dickens & Weller, 1983; Wiseman & Dickens, 1973). The compound HNbO3 (Fourquet et al., 1983) is of direct relevance to the subject of the present study. In this compound, the protons are accommodated in the vacant A-sites within the framework of NbO6 octahedra, creating hydroxyl bonding with the oxygen atoms (Fourquet et al., 1983). Theoretically, the amount of protons in the A-site (x) is limited by the number of oxygen atoms available for bonding, relative size of the B-site cation, and by the lowest valence exhibited by this cation. The maximum x is found in hydroxides of relatively large-size trivalent cations, such as Sc(OH)3 (Christensen et al., 1967). The high valence and small radius of Nb5+ probably limits the amount of protons (x) in the structure of Nb-dominant perovskite-group minerals to 1.
For isolueshite, a value of x cannot be estimated directly from the microprobe analyses, as all analysis totals are within the analytical error to 100 wt.% (Chakhmouradian et al., 1997). However, structural formulae of the mineral calculated to 6 negative charges systematically show some excess of cations at the B-site, which may indicate a smaller total negative charge. By fixing the cation totals at the B-site at a constant value, we will be able to obtain a rough estimate of x for isolueshite. The structural formula of the mineral calculated on the basis of one B-site cation, are given in Table 7.
Discussion and conclusions
NaNbO3 is one of the most structurally complex compounds in the perovskite family. The room-temperature form of NaNbO3 (P phase) shows two types of octahedral tilting combined with displacement of Na and Nb from the geometrical centers of the polyhedra (Sakowski-Cowley et al., 1969). This distortion results in an orthorhombic structure with the space group Pbma, a = 5.566, b = 15.520 and c = 5.506 Å. The Pbma structure is unstable at high temperatures and undergoes a series of phase transitions leading to a two-tilt structure Cmcm (> 520°C), one-tilt P4/mbm (> 575°C), and eventually, to an undistorted cubic structure stable above 640°C (Lefkowitz et al., 1966; Ahtee & Glazer, 1976). A similar “stabilizing” effect can be achieved by introducing other cations into the structure. The latter mechanism referred to as stabilized polymorphism, is relatively common in a number of compounds, including ZrO2 and TiO2 (Smirnova & Belov, 1969; Grunin et al., 1983). The effect of stabilizing polymorphism is not well understood for the members of perovskite structural family. It has been established experimentally that the introduction of 0.1 apfu of Th or 0.5 apfu of K into NaNbO3 gives a rise to the undistorted cubic structure (Ahtee & Glazer, 1976; Labeau & Joubert, 1978).
The present study confirms that isolueshite has a cubic perovskite-type structure, as suggested by Chakhmouradian et al. (1997). Previously, it has been proposed that, in contrast to lueshite (essentially NaNbO3 the cubic symmetry of isolueshite is stabilized by accommodation of LREE and Ti in the A- and B-sites, respectively. However, the orthorhombic symmetry of synthetic (Na0.75La0.25)(Nb0.50Ti0.50)03 clearly demonstrates that incorporation of LREE and Ti alone cannot account for the stabilization of the Pm3m lattice. This may indicate that other minor elements present in the composition of isolueshite contribute to maintaining the stability of its cubic structure. Indeed, the chemistry of this mineral is far more complex than the simplified formula used in the structural refinement and synthesis. As shown in the present study, this complexity arises not only from cationic substitutions at the A and B-sites, but also from the O2- ← (OH)1- substitution at the anion site. However, contributions of individual elements into the combined stabilization effect are very difficult to assess for such a compositionally complex phase as isolueshite. Alternatively, the transitional, anion-disordered character of the isolueshite structure may suggest that thermodynamic or kinetic factors such as crystallization temperature, pressure or cooling rate played a more important role in stabilizing the cubic symmetry. In this case, isolueshite should be considered as a quenched (“frozen”), rather than stabilized polymorph of NaNbO3. Such interpretation is supported by the structural data obtained in the present study, i.e. the presence of incipient octahedral tilting in the lattice of isolueshite.
We gratefully acknowledge the help of I.I. Bannova and V.S. Fundamensky with the single-crystal data collection. We also wish to thank Doctors R. Angel, W.V. Maresch, P.M. Woodward, A.I. Becerro, and an anonymous reviewer for stimulating discussion, and many constructive comments that have helped to improve the initial version of the manuscript. This work was partly supported by the Natural Sciences and Engineering Research Council of Canada and Lakehead University (RHM, ARC).
- Received 30 June 1998.
- Modified version received 30 April 1999.
- Accepted 17 January 2000.