The impact of anion ordering on octahedra distortion and phase transitions in SrTaO 2 N and BaTaO 2 N

Synopsis Anion (O/N) ordering was observed in BaTaO 2 N and SrTaO 2 N. A further Ta-O/N octahedra displacement (rotation about the c axis) distortion was observed in SrTaO 2 N. Abstract In this work, we synthesized BaTaO 2 N and SrTaO 2 N using a two-step high temperature solid state reaction method and analysed the structural distortions, relative to the ideal cubic perovskite structure, according to group theory. From a complete distortion analysis/refinement using high resolution neutron diffraction data in the temperature range 8 K to 613 K, we identified tetragonal structures for BaTaO 2 N [ P 4/ mmm (No 123)] and SrTaO 2 N [ I 4/ mcm (No 140)]. In contrast to an anion-disordered cubic perovskite ( 𝑃𝑃 3 � 𝑃 No 221) with Ta at the cell center, both systems show a site preference for oxygen anions in the two opposite corners (along the c axis) of the Ta-O/N octahedra rather than the four square corners in the ab plane ( Γ 3+ occupancy distortion), which induces a tetragonal elongation of the unit cell with the c axis being longer than the a axis. A further Ta-O/N octahedra displacement [ R 5-(a,0,0), rotation about the c axis] distortion was observed in SrTaO 2 N. This distortion mode is accompanied by an increased unit cell distortion that decreases as the temperature increases. Ultimately a second-order phase transition caused by the loss of the R 5-(a,0,0) mode was observed at 400 – 450 K.


Introduction
Mixed anion materials, in particular oxynitrides, are attracting considerable attention for their novel properties.SrTaO 2 N and BaTaO 2 N are two typical perovskite oxynitrides that have been widely studied due to their promising properties that include, semiconductivity (Kim et al., 2004), dielectric relaxation (Kim et al., 2004, Sun et al., 2014, Hinuma et al., 2012, Masubuchi, 2013) and photocatalytic activity (Higashi et al., 2008, Higashi et al., 2013, Higashi et al., 2015, Oehler & Ebbinghaus, 2016, Ueda et al., 2015).Evidently the distortion of the octahedra in perovskite oxynitrides is a key parameter in understanding their interesting properties.Thus, a careful and systematic study of the Ta-O/N octahedra distortion in SrTaO 2 N and BaTaO 2 N is likely to enhance our understanding of these fascinating materials.
SrTaO 2 N and BaTaO 2 N are also good structural models to study O/N anion ordering in perovskites and probe the structural distortions that arise from such ordering.BaTaO 2 N has been reported to adopt an ideal cubic perovskite structure (3 �  No 221) with the O/N anions completely disordered in the anion octahedra (Kim et al., 2004, Pors et al., 1988) (Figure 1).The situation of SrTaO 2 N is more complex, and several structural models have been reported: I4/mcm (No 140) (Günther et al., 2000, Zhang et al., 2011, Kim et al., 2004, Clarke et al., 2002), I112/m (No 12) (Yang et al., 2011) and Fmmm (No 69) (Clark et al., 2013).Wolff and Dronskowski investigated the structure of SrTaO 2 N and BaTaO 2 N using first-principle and molecular dynamics calculations based on density functional theory (DFT) and suggested that for both oxynitrides the space group should be orthorhombic in Pmc2 1 (No 26) (Wolff & Dronskowski, 2008).The major difference among the various structural models described for SrTaO 2 N is the nature of the O/N ordering and the distortion (bond angles/distance anisotropy) of the Ta-O/N octahedra.For example, Günther et al. reported a completely ordered structural model with site s1 (Figure 1) occupied by N and the other anion sites occupied by O (Günther et al., 2000).Meanwhile, Clark et al. described another ordering pattern with site s1 (Figure 1) mainly occupied by O (~95%) and the other anion sites randomly occupied by O/N (Clark et al., 2013).Other works have suggested only partial O/N ordering (~75% O at site s1) (Clarke et al., 2002, Yang et al., 2011, Zhang et al., 2011) in the Ta-O/N octahedra.All the models reported to date gave similar distortion of the Ta-O/N octahedra.
The local structure of the Ta-O/N octahedra in SrTaO 2 N and BaTaO 2 N was also studied by DFT calculations (Fang et al., 2003, Page et al., 2007, Wolff & Dronskowski, 2008, Ravel et al., 2006), extended X-ray absorption fine structure (EXAFS) (Ravel et al., 2006), and neutron pair distribution function (PDF) analysis (Page et al., 2007).These confirmed the cis-TaO 4 N 2 configuration in both SrTaO 2 N and BaTaO 2 N. Thus, the possible N/O ordering patterns (Attfield, 2013, Yang et al., 2011, Zhang et al., 2011) in a long range average structure are built upon a cis-TaO 4 N 2 configuration and there should be more oxygen at the s1 site relative to the other sites illustrated in Figure 1.
The structures of SrTaO 2 N or BaTaO 2 N can be described as distortions from the ideal cubic perovskite structure (Figure 1), in which the O/N occupancy and atomic displacements are the distortion parameters (degrees of freedom).All possible distorted structures (complete base) can be derived using group theory analysis, from the parent structure, as long as the wave vector(s) is known (Campbell et al., 2006).The distorted structure is caused by the action of one or more distortion modes, which have physical meanings in terms of the atomic interactions.Describing complex structures in terms of distortion modes has a significant advantage in that, under some circumstances, it can reduce the number of structural parameters and this can increase the reliability of the refined structural model (an application of Occam's razor to crystallographic analysis).It has previously been demonstrated that the structure analysis based on the symmetry-mode refinements is an effective and powerful method and has been successful used in the analysis of complex oxides including WO 3 and Bi 2 Sn 2 O 7 (Kerman et al., 2012, Lewis et al., 2016), although this approach does not appear to have been used in systems with mixed anions.Thus, a structure analysis of SrTaO 2 N and BaTaO 2 N based on the distortion refinement is helpful for more reliable structure determination of them.Using =group theory, Talanov et al. have analysed all the 261 possible low symmetry structures caused by anion ordering in perovskite up to 2 × 2 × 2 super cell (four vectors considered: X, M, R, and Γ) and summarized the reported experiment results (Talanov et al., 2016).They supposed that the SrMO 2 N (M = Ta, Nb) compounds adopt A 1a B 1d X 1c X 2e 2 anion ordering.However, they did not analyse the complete distortions (ordering and displacement) for oxynitride perovskite by structure refinements, which would give more details on the nature and magnitude of the distortion modes.In the present work the temperature-dependent structural distortions of SrTaO 2 N and BaTaO 2 N have been studied using high resolution neutron diffraction data measured at fine temperature intervals between 8 K and ~600 K.The structures and phase transitions are described based on the distortion mode refinements.
Polycrystalline samples of BaTaO 2 N and SrTaO 2 N were synthesized in two steps.Initially a 1:2 (mole ratio) of Ta 2 O 5 (Aithaca, 99.99%) and SrCO 3 (Aldrich, >99.9%) or BaCO 3 (Aithaca, 99.999%), all dried at 400 °C for 10 h before use, was weighed and ground.The well mixed powder was then calcined at 1100 °C for 36 h with an intermediate regrinding.After the calcination, the powder was confirmed as a single phase of Sr 2 Ta 2 O 7 or Ba 2 Ta 2 O 7 from powder X-ray diffraction (XRD) analysis.
The precursor Sr 2 Ta 2 O 7 or Ba 2 Ta 2 O 7 was then put into an alumina boat and nitrided in a tube (squartz) furnace under a 300 ml/min NH 3 flow.The nitridation was conducted at 1000 °C for 24 h with an intermediate regrinding.After the nitridation, the white precursor changed to a bright orange (SrTaO 2 N) or red brown (BaTaO 2 N) product.The phases of samples prepared were determined from Rietveld refinement using synchrotron X-ray diffraction to be a major (~99% wt.) SrTaO 2 N or BaTaO 2 N phase and a minor (~1% wt.) Ta 3 N 5 impurity phase.Time-of-flight (TOF) powder neutron diffraction (ND) data of BaTaO 2 N (~7 g) and SrTaO 2 N (~7 g) samples were collected on the high resolution powder diffraction (HRPD) diffractometer at the ISIS facility of the Rutherford Appleton Laboratory (UK).The data collected using the backscattering (b1) and 90° (b2) detectors were analysed.The temperature range of interest, from 8 to 613 K, necessitated the use of both cryostat and furnace.The powdered samples were lightly packed into either an aluminium can of slab geometry for measurements in the cryostat (4.2 -300 K), or a thin-walled 11 mm diameter vanadium can for analysis in the furnace (first at room temperature, followed by measurements from 373 to 613 K).The data were acquired for 2 -4 h at RT (13°C, 286 K), 418 and 8K and 30 min/scan for the other temperatures.
The structure of BaTaO 2 N and SrTaO 2 N was investigated by Rietveld refinements against the b1 and b2 ND data) using TOPAS Academic (TA) (Coelho et al., 2011, Evans, 2010).For the refinements the cubic disordered perovskite structure model was selected as parent structure (undistorted) and symmetry representation analysis was applied using ISODISTORT (Campbell et al., 2006) according to the observed supercell reflections, to build and test the occupancy and displacement distortions against the RT and 300 K data.After the distortion was confirmed, a batch refinement using selected distortion models against the data at all temperatures was conducted to investigate the temperature dependent distortion transitions.
Room temperature oxygen and nitrogen K-edge X-ray absorption near edge spectra (XANES) of BaTaO 2 N and SrTaO 2 N were collected at the soft X-ray spectroscopy beamline of Australian Synchrotron.Finely ground powder was pressed onto a conducting carbon tape to form a homogenous layer.Due to the very small penetration depth of the selected soft X-rays, a careful thickness optimisation was necessary to ensure sure that there is no signal from the tape substrate.The tape was placed on the gold-coated disc.The spectra were collected with a vacuum lower than 10 -9 mbar.The X-ray absorption spectra (530 -600 eV for oxygen K-edge and 385 -435 eV for nitrogen K-edge) were collected in a total fluorescence yield (TFY) mode with 0.1eV/step and the default collection angle (55°) was selected.For each measurement, the spectra of reference (MnO for O K-edge and BN for N K-edge) was simultaneously measured to calibrate the energy.
Absorption/reflection spectra of BaTaO 2 N and SrTaO 2 N were measured using a CARY5000 UV-Vis-NIR spectrophotometer equipped with a Harrick Praying Mantis solid-state reflection accessory.The spectra were collected from 200 -1500 nm with BaSO 4 as reference.

Results and discussion
Structure of BaTaO 2 N. The structure of BaTaO 2 N has previously been reported to be an undistorted cubic perovskite (Pors et al., 1988).No peaks indicative of supercell symmetry were observed in the current neutron diffraction data.The Rietveld refinement using the undistorted 3 �  model against the HRPD data gave R wp factors of: 3.51% (overall), 4.46% (b1), 2.80% (b2) for the 8 K data and 3.64% (overall), 4.67% (b1), 2.83% (b2) for the 300 K data.A distortion analysis was undertaken and the resulting models tested by refinement against the HRPD data to test the possibility of small distortions of the perovskite structure from ordering of the N and O anions.Since no supercell reflections were observed in the diffraction data, the distortion vector should be k 12 (0 0 0) (Γ) and considering the distortion vector five distortion irreducible representations (IR) are possible (Campbell et al., 2006): Γ 3 + ordering distortion (τ 5 in Kovalev notation) (Talanov et al., 2016).The refined BaTaO 2 N structure is shown in Figure 3 and the structural parameters are summarized in Table 1.
It is illustrative to consider the refined O/N occupancies.The negligible difference between the refined O/N occupancy at 300 K and 8 K is consistent with our expectation that the anions do not have sufficient energy to hop between different sites at low temperatures, and that the anion distribution is frozen at a blocking temperature during the high temperature synthesis.Consequently, the O/N occupancy was fixed to the optimal fit obtained against the 300 K data for all the refinements against data measured at other temperatures.Compared with the random occupancy (O/N = 0.67/0.33)at both sites, the optimised tetragonal P4/mmm model in BaTaO 2 N shows that there is more oxygen    The temperature dependence of the refined unit cell parameters (a, c, V) are plotted in Figure 4.All the cell parameters increase with increasing temperature and the cell parameters can be described by a single term Einstein-type expression.The cell volume V against temperature (T) is well produced by an expression of the type: where T is temperature, V and V 0 are the cell volume at T and 0 K, C is a constant, and θ is the empirical saturation temperature.The fitted parameters are V 0 = 69.243(1)Å 3 , C = 2.05(1) × 10 -5 K -1 and θ = 243(5) K.The difference between a and c (c -a = c p -a p , where the subscript p refers to the equivalent perovskite cell) can be considered as the degree of tetragonal distortion.This remains essentially constant as the temperature is increased from 9 to 300 K demonstrating the tetragonal unit cell distortion is a consequence of O/N ordering, which does not change over the temperature range investigated.Å at 145 °C (418 K), 13 °C (286 K), 8K respectively.The δ parameter is expected to be positive, i.e. c p /a p > 1 for the same reasons described above for BaTaO 2 N.    The thermal expansion of the tetragonal cell is highly anisotropic.At all temperatures (c p /a p ) > 1 as evident from Figure 9c although the individual lattice parameters expand at different rates.The final reversal is associated with the proposed I4/mcm to P4/mmm transition.Between ~ 450 and 600 K the distortion parameter δ = c/2 -a/√2 = c p -a p is essentially constant.This is equivalent to the spontaneous strains described in other studies of perovskites (Tan et al., 2012).Around 550 K there is a sharp increase in this and it appears that the introduction of tilting results in a greater contraction of the a-parameter, and the c-parameter is only weakly temperature dependent.The introduction of outof-phase tilting (3 �  → 4/ transition) in perovskites such as CaTiO 3 (Kennedy et al., 1999) and SrZrO 3 (Howard et al., 2000) results in cell metrics (c p /a p ) > 1, due to a contraction in the ab plane.A similar effect appears in SrTaO 2 N at T < 450 K, where the out-of-phase Ta-O/N octahedra rotation distortion caused an enhanced contraction in ab plane.Therefore, a strong correlation between the distortion and tetragonal cell distort (δ) was observed when the distortion is larger than 0.20, as shown in the Figure 9e.
The nature of the continuous phase transition can be established by examining the temperature dependence of the order parameter (Q), e.g., Q = A(T c -T) n with n = ½ or ¼ for a second order or tricritical transition, respectively.(Salje, 1993) In the present case the distortion amplitude of R 5 -mode is a suitable order parameter.A plot of the (R 5 -) 2 as a function of temperature is linear at high temperatures showing the transition to be second order with a transition temperature = 488(5) K.The value of R 5 -plateaus as T approaches 0 K due to saturation effects, and to model this, the distortion parameter (D) was fitted to expression of the type D 2 = D 1 +D 2 θcoth(θ/T).The fit was relatively insensitive to the precise value of θ which was tried and fixed at 130 K, which is comparable to the values observed in other perovskites (Tan et al., 2012).The results of this are illustrated in Figure 9d.eV and E g (i) = 1.84 eV.The band gap of SrTaO 2 N is estimated to be appreciably larger E g (d) = 1.99 eV and E g (i) = 2.09 eV.Kim et al. have previously estimated the band gap of BaTaO 2 N and SrTaO 2 N to be 1.8eV and 2.1eV respectively (Kim et al., 2004), which agree with our estimates of the indirect

Figure 1 .
Figure 1.Representation of the perovskite structure of ATaO 2 N. The mixed occupancy of O (red) and N (blue) is represented by the shading.

and Γ 5 -.
The model Γ 3 + (one internal parameter) gave a better refinement result against our ND data with R wp factors: 3.46% (overall), 4.41% (b1), 2.76% (b2) for 8 K data and 3.58% (overall), 4.60% (b1), 2.77% (b2) for 300 K data.The overall R wp factors of the other models are: 8K 3.50% (should be noted that the Γ 4 -, and Γ 5 -models give similar quality refinements to that of the Γ 3 + model but required more internal parameters (12 in Γ 4 -and 3 in Γ 5 -).The refined profiles at 300 K are shown in Figure 2, and demonstrate this model provides an excellent fit to the data.The variable temperature (VT, 9 to 300 K) neutron diffraction data of BaTaO 2 N are shown in Figure S1 of the supporting information (SI).The distorted structural model can be described with an O/N ordering mode and the corresponding unit cell distortion from cubic to tetragonal in space group P4/mmm (No 123).Consideration of the most general order parameter direction of the two-dimensional IR Γ 3 + , which has two O/N-ordering modes rather than just one and results in Pmmm symmetry, did not result in any improvement to the refinement.Nor did simultaneously invoking the other available gamma-point IR modes improve the refinement.Thus, it is concluded that the structure of BaTaO 2 N is tetragonal with one occupancy distortion mode from the O/N ordering.This is in agreement with the recent results of Talanov et al., who described a "pure"

[
O/N = 0.87(2)/0.13(2)]at the two opposite corners of the Ta-O/N octahedra (along the c axis) and more nitrogen [O/N = 0.56(1)/0.44(1)]at the four square sites parallel to the ab plane.Thus, there are more cis-Ta-N bonds parallel to the ab plane than in either the ac or bc planes.A consequence of the O/N ordering is the slight difference observed between the a and c axis [~0.0037(1)Å at 9 K and ~0.0041(1) Å at 300 K].However, it is impossible to determine which axis (a or c) is longer from current diffraction data.The overall R wp factors of a > c model at 8K and 300K are 3.47% and 3.59%, which are insufficient to distinguish this from the a < c model described above.DFT calculations have shown the average cis-Ta-N bond distance is less than the average Ta-O bond distance (2.01 versus 2.12 Å)(Wolff & Dronskowski, 2008) reflecting the stronger Ta-N π bonds compared to the Ta-O π bonds.Since there are more Ta-N bonds in the ab plane, it is expected that the a axis should be shorter than the c axis for the tetragonal BaTaO 2 N unit cell and this assumption was used in developing the structural models.

Figure 2 .
Figure2.Observed, calculated and difference profiles of ND data for BaTaO 2 N at 9 K and 300 K using Γ 3 +

Figure 3 .
Figure 3. Representation of the crystal structure of BaTaO 2 N. The large green spheres represent the Ba cations.The occupancy of O (red) and N (blue) is shown as a fraction of the small spheres in the TaO 4 N 2 octahedron.

Figure 4 .-,
Figure 4. Temperature dependence of the tetragonal unit cell parameters and equivalent unit cell volume for BaTaO 2 N estimated from Rietveld refinements of PND data.

Figure 5 .
Figure 5. Neutron diffraction patterns of SrTaO 2 N at 8K (left, light green experiment data, red: simulated from the cubic perovskite model) and normalizes intensities of peaks marked versus temperature (right).

Figure 6 .
Figure 6.Illustration of the Ta-O/N octahedra displacement distortions of different R type IRs.For each of these, the direction of the distortion is arbitrary to the principle directions of the unit cell.

Figure 7 .Figure 8 .
Figure 7. R wp factors of the Rietveld refinements of SrTaO 2 N ND data using different distortion models.(a) (b) different R IRs with all the distortion modes refined; (c) six different structural models of R 5 -.
/N1 site is 0.5 O and 0.5 N.All the other sites are fully occupied.

Figure 9 .
Figure 9. Temperature dependence of the appropriately scaled unit cell parameter a p , c p , and volume (a&b), cell distortion (c), Ta-O/N octahedra displacement distortion (d) and the correlation between distortion and cell distortion (e).
band gap.The band gap of BaTaO 2 N and SrTaO 2 N should be dominated by the orbital splitting of Ta-O/N octahedra.Since the Ta -O/N bond lengths in SrTaO 2 N are shorter than those in BaTaO 2 N, SrTaO 2 N is expected to have a larger effective crystal field than BaTaO 2 N resulting in a larger orbital splitting (band gap).

Figure S2
Figure S2Neutron diffraction patterns of SrTaO 2 N at 300 K and 13 °C.

Figure S3
Figure S3Variable temperature neutron diffraction patterns of SrTaO 2 N.

Table 1
Refined structural parameters for BaTaO 2 N models is given as cif files in the SI.It is reasonable to conclude that around 450 K there is a structural transition from I4/mcm to P4/mmm.Additional measurements such as heat capacity are required to obtain an accurate estimate of the transition temperature.

Table 2
Refined structure parameters of SrTaO 2 N