Anna
Magrasó
*a,
Charles H.
Hervoches
a,
Istaq
Ahmed‡
b,
Stephen
Hull
b,
Jonas
Nordström
c,
Anders Werner Bredvei
Skilbred
a and
Reidar
Haugsrud
a
aDepartment of Chemistry, Centre for Materials Science and Nanotechnology, University of Oslo, FERMiO, NO-0349, Oslo, Norway. E-mail: a.m.sola@smn.uio.no
bThe ISIS Facility, STFC Rutherford Appleton Laboratory, Chilton, Didcot, Oxfordshire OX11 0QX, UK
cDepartment of Applied Physics, Chalmers University of Technology, SE-412 96 Gothenburg, Sweden
First published on 11th January 2013
In situ neutron diffraction experiments of 2% Ca-doped and nominally undoped lanthanum tungstate (La28−xW4+xO54+3x/2, with x = 0.85) have been carried out under controlled pD2O and pO2 at elevated temperatures. All the diffraction patterns could be refined using an average cubic fluorite-related structure, in accordance with recent reports. The material exhibits disorder of the oxygen and the cation sublattices. Splitting of the oxygen sites around tungsten from the 32f to 96k Wyckoff position in the Fmm space group improves the model and can better represent the oxygen disorder. No phase transition was detected from room temperature up to 800 °C under any of the studied conditions. Expansion of the unit cell constants in the presence of water at intermediate and low temperatures was correlated with the formation of protonic defects. The thermal expansion coefficient for lanthanum tungstate is rather linear under all studied conditions (∼11 × 10−6 K−1). The in situ diffraction studies are correlated with dilatometry investigations and conductivity measurements.
The crystal structure and exact stoichiometry of lanthanum tungstate have been controversial. Recent investigations show that the material can be described as La28−xW4+xO54+δv2−δ, where x is the amount of tungsten sitting on a lanthanum site and, in turn, determines how many oxygen vacancies (v) are in the structure as inherent defects.1 Lanthanum tungstate was earlier referred to as La6WO12 (La/W = 6, LWO60), which according to Yoshimura et al. only exists as a high temperature phase, not being thermodynamically stable below 1740 °C.17,18 La10W2O21 (La/W = 5) is not single phase either.5,19 A systematic study in the La/W system showed that single phase materials can be prepared at 1500 °C for stoichiometries between 5.3 and 5.7,5 and it is now established that both La6WO12 and La10W2O21 are incorrect formula simplifications of the general La28−xW4+xO54+δ.1
Previously, it was reported that the ionic conductivity decreased in the order undoped > 0.5% Ca > 5% Ca when lanthanum was substituted by calcium in LWO60 (La/W = 6).2 This indicates that acceptor substitution must either decrease the number of protons incorporated, or slow down their mobility. It was speculated that the acceptor (Ca2+) could trap protons based on the observation that the enthalpy of proton mobility increased upon acceptor doping. One of the purposes of this contribution is to determine possible structural differences between a nominally undoped and a 2% Ca-doped LWO56 to rationalise the results reported in ref. 2.
Lanthanum tungstate hydrates similarly to a nominally acceptor-doped material by interaction between oxygen vacancies and water vapour forming hydroxide defects (protons):20
H2O(g) + OxO + v˙˙O = 2OH˙O | (1) |
In general, this reaction is exothermic, so protons dominate at low temperatures and oxygen vacancies become more significant at high temperatures. An expansion of the material at low temperatures will occur upon hydration, and this can be monitored by different techniques, including diffraction and dilatometry.
It is crucial to study the structural changes related to the dissolution of water and temperature, such as phase transitions, changes in the coordination environment of the cations or in the unit cell. In situneutron diffraction under controlled atmosphere is, in this respect, an invaluable tool to determine temperature- and atmosphere-induced structural changes that would occur under real conditions. We hereby report a crystallographic study from in situneutron diffraction experiments on undoped and Ca-doped lanthanum tungstate under controlled partial pressures of heavy water and oxygen (pD2O and pO2) from 800 °C to RT. We correlate the structural changes with expansion of the material from dilatometry and with its conducting properties.
A powder X-ray diffraction (XRD) pattern was also collected at room temperature on a Huber G670 diffractometer with Cu Kα1 radiation selected by a Ge(111) monochromator, 2θ range [4–100°], step size 0.005°, exposure time of 5 × 3 hours.
The sample was inserted into the humidity cell24 (see Fig. S1†) for the in situNPD experiment and datasets were collected in both dry and humid atmospheres. First, each sample was heated in a dry Ar atmosphere until 800 °C in a vanadium shielded furnace, and when no change/shift on the cell parameter was observed, it was considered that equilibrium was achieved. The neutron data were then collected for 1–2 hours from 800 to 150 °C (ΔT = 50 °C) after equilibration at each temperature. After the dry runs, the sample was heated and equilibrated at 800 °C under humid conditions, following a similar procedure as for the dry runs. The humidity (pD2O) was controlled by bubbling the carrier gas through a D2O solution at a controlled temperature of 60 °C (pD2O = 0.18 atm), where D2O was used instead of H2O to decrease incoherent scattering. A heating cord wrapped with aluminium foil was used around the flowing hose to avoid undesirable condensation.
Additional diffractograms were collected at intermediate pD2O for Ca-LWO56 every 100 °C between 200 and 800 °C. The different humidity levels were established by changing the temperature of the D2O bath from 40, 60 and 80 °C, which yielded pD2Os of 0.065, 0.18 and 0.44 atm as calculated from the pH2O at saturation temperature using the Antoine equation and a pH2O/pD2O correction factor reported by Jakll and Alexander Van Hook.25
The backscattered C-bank data were analysed by the Rietveld method26 as implemented in the GSAS software package.27
Fig. 1 Schematic representation of the (a) averaged high symmetry crystal structure from ref. 5, (b) a relaxed local structure modeled by DFT from ref. 1, and (c) averaged structure obtained by replacing O2a and O2b from ref. 1 with the relaxed oxygen position according to the octahedra model,5 by DFT. |
The determination of the exact location of the oxygen sites is more troublesome since tungsten is 6-fold coordinated and forms octahedra that are disordered.1 The oxygen vacancies in lanthanum tungstate are essentially not ordered, as opposed to the tungstates with a smaller rare-earth, e.g. Y6WO12 and Ho6WO12,29 where the vacancies stay in well defined positions in the structure. In the average high symmetry description of lanthanum tungstate,5 oxygen can be located around the 16e (x, x, x) Wyckoff position: La1 is coordinated with O1a (x = 0.1376) and O1b (x = 0.8677). These oxygens are easy to locate, the refinement renders reasonable thermal factors, and form relatively symmetric LaO8 cubes. W(4b), on the other hand, is coordinated with O2a (x = 0.4005) and O2b (x = 0.5995), forms [WO8] “cubes” with oxygen positions that exhibit partial occupancy and high thermal factors. Since the coordination of tungsten with oxygen at the local scale is octahedral and not cubic, the high thermal factors of O2a and O2b are likely to reflect disorder. Fig. 1c shows the averaged structure obtained by replacing O2a and O2b from ref. 5 (Fig. 1a) with the relaxed oxygen positions forming octahedra as according to the DFT model1 (Fig. 1b). Accordingly, 3 equivalent O positions are generated around each cube corner with no preference for any of the orientations (Fig. 1c).
The relative distortion from the cubic cell has been reported to be 2.5 × 10−4,1 a deviation that is too small to be seen by the NPD data from POLARIS. Indeed, refinements using the tetragonal and rhombohedral space groups reported in ref. 1 did not improve the agreement factor. Therefore, deviation from the cubic structure in the present dataset will not be considered further.
Fig. 2 Raman spectrum of La28−xW4+xO54+3x/2 (La/W = 5.6, x = 0.85). Inset: zoom showing a weak band at ∼670 cm−1. |
Overall, the Raman spectrum is in agreement with the model shown in Fig. 1b;1 tungsten can be described locally as forming distorted octahedra, also consistent with average W–O distances.
The Rietveld refinements were performed using F3m and Fmm space groups. The atomic positions by the two models are illustrated in Table 1. They both give a similar agreement factor: F3m: χ2 = 2.032, Rp = 0.0303, Rwp = 0.0181 for 52 variables; Fmm: χ2 = 2.225, Rp = 0.0312, Rwp = 0.019 for 47 variables. From these refinements it is evident that the positions La1(4a), La2(24g), W1(4b) and O(16e) (x ∼ 0.13 and x ∼ 0.87) in F3m and Fmm are essentially correct, and there are minimal differences by using one or the other space group. Decreasing the average symmetry further, as suggested from synchrotron studies,1 does not improve the model significantly. However, the description of the oxygen sublattice in the structure needs a slightly different approach to be represented more accurately. In the Fmm space group, it is possible to split the oxygen positions (32f) into three equivalent positions (96k) that are similar to the oxygen positions predicted by the “average” DFT structure shown in Fig. 1c. We have, therefore, refined our patterns using the Fmm space group with a split oxygen position, for which further reasoning will be discussed in the coming section. The final refined parameters are shown in Table 2 and Table 3.
Fmm | F3m1 | ||
---|---|---|---|
Atom(Wyckoff) | Position (x y z) | Atom(Wyckoff) | Position (x y z) |
La1(4a) | (0 0 0) | La1(4a) | (0 0 0) |
La2(24d) | (0 0.25 0.25) | La2(24g) | (−0.004(6) 0.25 0.25) |
W1(4b) | (0.5 0.5 0.5) | W1(4b) | (0.5 0.5 0.5) |
O1(32f) | (x x x); x = 0.1336(1) | O1a (16e), O1b (16e) | (x x x); x = 0.1376(8), (x' x' x'); x' = 0.8677(8) |
O2(32f) | (y y y); y = 0.4032(2) | O2a (16e), O2b (16e) | (y y y); y = 0.4005(6), (y' y' y'); y' = 0.5995(6) |
Fractional atomic coordinates and isotropic or equivalent isotropic displacement parameters (Å2)a | |||||
---|---|---|---|---|---|
Atom | x | y | z | Ueq | Occ. |
a χ2 = 1.716, Rp = 0.031, Rwp = 0.017, Rexp = 0.013 for 50 parameters. | |||||
La1 | 0.0 | 0.0 | 0.0 | 0.01954 | 1.0 |
La2 | 0.0 | 0.25 | 0.25 | 0.0325 | 0.95 |
W1 | 0.5 | 0.5 | 0.5 | 0.01204 | 1.0 |
O1 | 0.13369(9) | 0.13369(9) | 0.13369(9) | 0.02022 | 1.0 |
O2 | 0.4357(3) | 0.3878(2) | 0.3878(2) | 0.0183 | 0.199(5) |
W2 | 0.0 | 0.25 | 0.25 | 0.0325 | 0.05 |
Atomic displacement parameters (Å2)a | ||||||
---|---|---|---|---|---|---|
Atom | U11 | U22 | U33 | U12 | U13 | U23 |
La1 | 0.0195(7) | 0.0195(7) | 0.0195(7) | 0.0 | 0.0 | 0.0 |
La2 | 0.0074(7) | 0.0451(8) | 0.0451(8) | 0.0 | 0.0 | −0.0353(7) |
W1 | 0.0120(9) | 0.0120(9) | 0.0120(9) | 0.0 | 0.0 | 0.0 |
O1 | 0.0202(4) | 0.0202(4) | 0.0202(4) | 0.0031(4) | 0.0031(4) | 0.0031(4) |
O2 | 0.037(3) | 0.009(1) | 0.009(1) | −0.015(1) | −0.015(1) | −0.0010(9) |
W2 | 0.0074(7) | 0.0451(8) | 0.0451(8) | 0.0 | 0.0 | −0.0353(7) |
Rietveld refinements corresponding to the Ca-LWO56 followed the same approach. Placing calcium in different cation sites did not make a significant difference in the refinement parameters as the doping concentration is only 2%. Moreover, the neutron coherent scattering lengths for W (4.86) and Ca (4.7) are very similar. On these bases it was not possible to determine the location of Ca.
Atoms | Interatomic distances (Å) |
---|---|
La1—O1( × 8) | 2.588(1) |
La2—O1( × 4) | 2.369(2) |
La2—O2( × 4) | 2.294(3) |
La2—O2( × 8) | 2.872(3) |
W1—O2( × 24) | 1.913(3) |
The refinement of the anisotropic atomic displacement parameters (ADPs) using the “simple model” (O2 in the 32f position in the Fmm space group) is shown as supplementary material (Fig. S3†). The refinement yields very large ADPs for O2 oxygens at the 32f position (∼0.4 0.4 0.4), which are those surrounding tungsten in the cube-like symmetry description (“simple model”). Moreover, the refined ADPs for O2 have negative mean-square atomic displacements. Such physically unrealistic results suggest a strong local structural disorder, as reported for other materials.33 The refined ADPs of oxygen bonded to tungsten in this model show that the oxygen presents high anisotropy around the “corners of the cube”, which are represented as cuboids (see Fig. S5a†). An additional inconsistency we observed in the “simple model” is that the ADPs do not extrapolate towards zero at T = 0 K when drawing a median line through the O(2) Uij (Fig. S3†), which indicates that the position of the O2 atoms is either displaced or split (static disorder).
An improvement of the “simple model” can be achieved by splitting the O2 position into three equivalent positions as explained in Section 4.2, i.e. the oxygen occupying the atomic position 96k (∼0.44 0.39 0.39) with an occupancy of ∼0.2 rather than the 32f (∼0.4 0.4 0.4) with occupancy ∼0.54. Analysis of the evolution of the anisotropic ADPs (Uij) with temperature for LWO56 under wet Ar in the improved “split model” is shown in Fig. 3. The ADPs of O2 can now be fitted with a median line passing through zero at T = 0 K, which shows that the “split model” is an improvement of the former model. The ADPs are still relatively large and some slightly negative mean-square atomic displacements remain, but the evolution of Uij with temperature indicates that the ADPs in the “split model” are essentially due to localized dynamic (vibrational) disorder of the O2 oxygens around W.
Fig. 3 Evolution of Uij anisotropic displacement parameters with temperature in wet Ar in the “split model” with O2 in 96k position. Dashed lines are guide to the eye. |
This model has been independently suggested by DFT calculations,1 which support that the split of the oxygen positions is a reasonable representation of the disordered oxygen positions around tungsten (see comparison in Fig. S6†). Information about the atomic pair distribution functions (PDF) using the total scattering method34 could overcome the problem of negative mean-square atomic displacements obtained by the Rietveld method and would clearly be valuable in this highly disordered material. This is, however, out of the scope of the present contribution, and will be treated separately.
In a similar fashion, the ADPs for La2 (0 0.25 0.25) are rather large, with strong anisotropy in the La2–W1 direction, and do not extrapolate towards zero at T = 0 K. Since the site contains a mixed La/W occupancy (∼95%/5%), it is conceivable that tungsten (much smaller than lanthanum) occupies a slightly displaced position. This is consistent with both the large anisotropic ADP of the La2 site and the weak Raman band at ∼670 cm−1 that indicates the presence of W–O–O–W bonds.31 The crystal structural model from DFT calculations1 shows that the La2 cations are not perfectly aligned with each other, which is also in line with our study. We have tried to include splitting the La2 site in our model, as suggested recently by Scherb et al.,35 but the refinements are unstable and tend to diverge. It is reasonable that the La2 site may be split, but the limitations due to the high disorder of the material along with the medium resolution of the POLARIS diffractometer make it difficult to confirm this unambiguously.
The evolution of W–O distances with temperature in dry and wet Ar is displayed in Fig. 4. It shows that the apparent W–O distance is decreasing with increasing temperature, which is clearly unphysical. This artificial bond shortening is due to the disorder of the oxygen positions around tungsten. The circular oscillating movements (libration) of oxygen atoms around W increase with temperature, which make the interatomic distances appear shorter on average. The bond length was therefore corrected using the simple rigid bond model correction proposed by Downs et al.36 after which the bond length between W and O increases with increasing temperature, as expected. The La1–O and La2–O distances also increase with temperature (shown as supplementary material, Table S3†).
Fig. 4 Evolution of W–O distances with temperature in dry and wet argon atmosphere (circles), and the distance corrected by the simple rigid bond correction model (triangles).36 |
The final refinements for both LWO56 and Ca-LWO56 at all temperatures will be performed using the Fmm space group with split oxygen position (96k). The refinement of the XRD and NPD patterns obtained at room temperature using this model are shown in Fig. 5, and the atomic coordinates and displacement parameters of LWO56 in Table 2.
Fig. 5 Final Rietveld fits of the (a) XRD data and (b) neutron diffraction data, at room temperature. |
Fig. 6 Representative NPD patterns collected in dry Ar for undoped LWO56 at different temperatures. |
It is interesting to investigate whether water affects the structure to any significant degree. Representative NPD patterns for both LWO56 and Ca-LWO56 in D2O–Ar and nominally dry Ar at 500 °C are shown as supplementary information (Fig. S7†). The water vapour did not introduce any measurable symmetry distortions for any of the compositions studied here, and the Rietveld refinement confirms that the structure remains as a cubic fluorite-type Fmm (on average) at all measured temperatures, pD2O and doping. Atomic positions and selected bond lengths and angles from the final Rietveld refinement at room temperature for LWO56 are presented in Table 2, and additional temperatures are available as supplementary material (Tables S4–S9).
The variation of the lattice parameter as a function of temperature in dry and wet Ar is shown in Fig. 7a. First, the lattice parameters increase with increasing temperature in a rather linear fashion, which supports the finding that the material does not exhibit any measurable phase transition over the studied temperature range. Being a cubic system, the expansion is isotropic along the three directions of the unit cell. It is important from a technological point of view that a material with a practical application exhibits a linear isotropic thermal expansion coefficient, to make thermal compatibility with the surrounding materials easier. Second, there is a small but significant decrease of the lattice parameter upon Ca substitution (in dry Ar), which indicates that some substitution has occurred, in accordance with the smaller ionic radius of Ca2+ (1.12 Å) compared to La3+ (1.16 Å) in VIII coordination.37 Third, a measurable cell expansion upon wetting is detected for both the doped and undoped LWO56, and can be associated with the incorporation of protons in the material. The differences between wet and dry conditions become smaller at higher temperatures which reflect that the concentration of protons in the structure of LWO56 and Ca-LWO56 gradually decreases as a consequence of the exothermic nature of the hydration reaction in eqn (1).20
Fig. 7 (a) Evolution of lattice parameter with temperature in dry and wet argon for both LWO56 and Ca-LWO56. (b) Additional tests with different D2O bath temperatures (variation in pH2O) for Ca-LWO56. |
It is important in the study of proton conductors to determine how the structure changes as a function of pH2O, and correlate this behaviour with its functional properties. This can be assessed in our experiment by following the variation in the lattice parameters with temperature under different pD2O for the Ca-LWO56, as displayed in Fig. 7b. The lattice parameters increase with increasing pD2O, which supports that the material incorporates more protons with increasing humidity, as expected from eqn (1). There are no detectable differences between pD2O of 0.18 atm (TD2O = 60 °C) and 0.44 atm (TD2O = 80 °C), which indicates that the material is saturated with water at these pressures, i.e. no more protons can be further incorporated. These results are in very good agreement with recent thermogravimetric measurements, which showed saturation of protonic defects at pH2O ∼0.2 atm.38
Fig. 8 Representative temperature dependence of the linear thermal expansion for LWO56 upon cooling from 1000 to 50 °C in (a) flowing air and (b) 5% H2/Ar (harmix). |
The thermal expansion coefficient can be calculated from both the dilatometric studies and the evolution of the lattice parameters from the neutron diffraction experiment, which are summarized in Table 4. The overall thermal and chemical expansion coefficient of LWO56 is 10.9 × 10−6 K−1 and 11.9 × 10−6 K−1 in wet and dry Ar, respectively, while the values are 11.0 × 10−6 K−1 in wet air and wet 5% H2/Ar, and increases slightly under dry conditions: ∼11.5 × 10−6 K−1. This confirms that the expansion coefficients are very consistent within the different techniques used, essentially constant throughout the studied temperature range, independent of the oxygen partial pressure, and somewhat dependent on the water vapour partial pressure, all in line with earlier reports.1–3,5,38
Atmosphere | From neutron diffraction (cooling cycle) | From dilatometry (cooling cycle) | ||||
---|---|---|---|---|---|---|
150–500 °C | 600–800 °C | Overall | 100–550 °C | 650–900 °C | Overall | |
Wet Ar | 11.0 | 10.9 | 10.9 | — | — | — |
Dry Ar | 10.9 | 14.0 | 11.9 | — | — | — |
Wet air | — | — | — | 10.8 | 11.7 | 11.0 |
Dry air | — | — | — | 10.7 | 12.7 | 11.4 |
Wet 5% H2/Ar | — | — | — | 10.8 | 11.7 | 11.0 |
Dry 5% H2/Ar | — | — | — | 10.8 | 12.9 | 11.6 |
Fig. 9 Arrhenius representation of the grain interior conductivity vs. temperature for LWO56 and Ca-LWO56, and compared to literature data. |
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c3ta00497j |
‡ Now working at Volvo Group Trucks Technology, Advanced Technology & Research, Gothenburg, Sweden. |
This journal is © The Royal Society of Chemistry 2013 |