Hydration Mechanisms and Proton Conduction in the Mixed Ionic

Article pubs.acs.org/cm

Cite This: Chem. Mater. 2018, 30, 4949−4958

Hydration Mechanisms and Proton Conduction in the Mixed Ionic− Electronic Conductors Ba4Nb2O9 and Ba4Ta2O9 Julia Wind,†,§ Richard A. Mole,‡ Dehong Yu,‡ Maxim Avdeev,‡ and Chris D. Ling*,† †

School of Chemistry, The University of Sydney, Sydney 2006, Australia Australian Centre for Neutron Scattering, ANSTO, New Illawarra Road, Lucas Heights 2234, Australia

‡

Chem. Mater. 2018.30:4949-4958. Downloaded from pubs.acs.org by UNIV OF SUNDERLAND on 10/07/18. For personal use only.

S Supporting Information *

ABSTRACT: We studied the behavior of hydrogen in the mixed ionic−electronic conductors γ-Ba4Nb2O9 and 6H-Ba4Ta2O9 using a combination of experimental (neutron diffraction and inelastic neutron scattering) and computational (ab initio molecular dynamics) methods. Although these compounds have isostructural low-temperature polymorphs, they adopt distinct forms in the high-temperature conducting regime. We show here that they also have distinct mechanisms for hydration and ionic conduction. Hydration of γ-Ba4Nb2O9 is localized to 2-D layers in the structure that contain a 1:1 ratio of isolated but adjacent NbO4 and NbO5 polyhedra. OH− and H+ ions combine with two polyhedra, respectively, to form complete layers of NbO4OH polyhedra, giving rise to a stoichiometric hydrated form γ-III-Ba4Nb2O9·1/3H2O. Protons then diffuse through these 2-D layers by “hopping” between oxygen atoms on adjacent polyhedra. In the case of 6H-Ba4Ta2O9, hydration occurs by intercalating intact water molecules into the structure up to a maximum of ∼0.375 H2O per formula unit. This explains the unusual local and longrange structural distortions in the hydrated form observed by neutron diffraction. Diffusion then occurs by water molecules moving between neighboring symmetry equivalent positions. These fundamentally different hydration and proton conduction mechanisms explain why 6H-Ba4Ta2O9 has the less well-defined and higher maximum water content, while γ-Ba4Nb2O9 has the higher proton conductivity.

1. INTRODUCTION Mixed conductors, materials that exhibit significant mobility of more than one type of charge carrier such as oxide ions, protons, and electrons, have a range of important applications, including solid oxide fuel cell membranes, electrodes, batteries, and sensors. A number of water-absorbing oxides exhibit significant proton conductivity, but many of the promising candidates (e.g., rare earth-doped BaCeO31) suffer from chemical instability in humid or CO2-containing atmospheres.2−5 It is therefore of interest to discover and characterize new candidates. Several niobium and tantalum oxides have recently been shown to exhibit promising proton or mixed conductivity with a rather high uptake of protons at moderate temperatures (e.g., BCN18 Ba3Ca1.18Nb1.82O9‑δ,6 La0.99Ca0.01NbO47,8). The main focus has been on disordered forms of these materials, which have better conductivities, using empirical engineering approaches to tune and adjust their properties. Our work focuses on chemically related systems with more ordered structures as simple model systems where we can identify crystal−chemical relationships among chemistry, structure, and properties that can be leveraged to rationally design improved versions. Ba4Nb2O9 and Ba4Ta2O9 have relatively simple chemical compositions, interesting structural features in their respective high- and low-temperature polymorphs, and show good © 2018 American Chemical Society

protonic and oxide-ionic conductivities. Previous studies focused on structural features as well as hydration behavior of these materials.9−12 As shown in Figure 1, both compounds have two basic polymorphs: a low-temperature α phase (the same for Ba4Nb2O9 and Ba4Ta2O9) and a high-temperature phase (different for Ba4Nb2O9 and Ba4Ta2O9). The low-temperature α phases have the rare Sr4Ru2O9-type,9 which is partially disordered. The main structural feature is M2O9 face-sharing octahedral dimers, aligned along the c axis and separated by voids of similar size to the dimers themselves. The disorder arises because there are four possible positions for these dimers along the c axis but only three available crystallographically distinct sites. Some deficiencies in the published Rietveld refinements of the disordered structure hint at partial ordering of the M2O9-dimers, but models including long-range ordering have not led to significantly improved fits.9 The high-temperature phases of the two compounds are completely different. The tantalum compound adopts a hexagonal 6H-perovskite structure (BaTiO3-type, space group P63/m) which on cooling below 900 K undergoes a symmetry lowering distortion to monoclinic (space group P21/c), mainly Received: March 4, 2018 Revised: July 11, 2018 Published: July 12, 2018 4949

DOI: 10.1021/acs.chemmater.8b00950 Chem. Mater. 2018, 30, 4949−4958

Article

Chemistry of Materials

Figure 1. Structures of Ba4M2O9 (M = Nb, Ta). Ba atoms are green; Ta/Nb atoms and octahedra are blue, and O atoms are red. Half-occupied O sites are represented by half-colored spheres. For γ-Ba4Nb2O9, NbO6 octahedra are drawn in blue, Nbo5 trigonal bipyramids orange, and NbO4 tetrahedra yellow.

hydration mechanism via hydroxylation of NbO4 tetrahedra and association of additional protons with existing NbO5 trigonal bipyramids. Their solid-state 93Nb NMR data added weight to this model by providing direct experimental evidence of the loss of four-coordinate Nb upon hydration. While the dehydrated structures are well-established, information about the hydration behavior and dynamics is limited to results from thermogravimetric analysis (TGA) quantification of intercalated water and impedance measurements of average dynamics. The goal of the present work was to locate the water within the structures, identify the chemical form (H+, OH−, H2O) in which the water is intercalated, and explore the ionic and/or molecular dynamics in these systems.

due to cooperative rotation of the Ta2O9 dimers and distortion of oxygen octahedra around barium atoms. The quenched γniobate adopts a unique orthorhombic structure type (space group Pca21), incorporating niobium in 4-, 5-, and 6coordinate environments.12 The structure was solved by Dunstan et al.12 in a study combining high-resolution singlecrystal neutron diffraction and solid-state 93Nb NMR. The 4and 5-coordinate niobium polyhedra lie in discrete layers in which their orientations vary systematically to form a complex superstructure. Below 950 K, both compounds in their respective high- and low-temperature forms were reported to absorb significant amounts of water.11 Both forms of Ba4Ta2O9 hydrate to an extent greater than that of the corresponding niobium phases. Specifically, α-Ba4Ta2O9 hydrates to approximately twice the extent of α-Ba4Nb2O9. This might be due to the greater polarizability of Ta5+ versus Nb5+, tentatively supported by the fact that the unit cell volume per formula unit for α-Ba4Ta2O9 is ∼0.3% larger than for α-Ba4Nb2O9. No evidence for ordered proton or hydroxide sites has been reported. The hightemperature forms absorb more water than the corresponding low-temperature forms, in agreement with their larger equivalent unit cells. However, comparing the unit cell volume per formula unit among the high-temperature phases, the volume for the tantalate is slightly smaller than that for the niobate. Conductivity measurements9,11 (impedance spectroscopy) showed that the hydrated forms exhibit significant protonic conductivity and mixed oxide ionic and electronic conductivity upon dehydration. Both low-temperature α phases show much lower conductivity than the high-temperature phases over the entire temperature range. Interestingly, the niobates show higher protonic conductivity, although less water is present in the structure. Based on the structural features in γ-Ba4Nb2O9, Dunstan et al.12 suggested a possible

2. EXPERIMENTAL DETAILS 2.1. Synthesis. The starting reagents BaCO3 (99.999%), Ta2O5 (99.999%), and Nb2O5 (99.998%) were dried for 16 h at 750 °C. Following the procedure described by Dunstan et al.,12 stoichiometric mixtures of dried powders with a total mass of 20 g were then ground together with ethanol in a planetary ball-mill and sintered twice at 1000 °C for 12 h with thorough regrinding between the individual heating steps. This yielded phase-pure samples of the respective lowtemperature α phases. To obtain the high-temperature phases, samples were annealed for 48 h at 1400 °C and then quenched to room temperature on a steel plate. Reaction progress and sample purity at room temperature were checked by powder X-ray diffraction (PXRD) on a Panalytical X’pert Pro diffractometer in Bragg− Brentano geometry using nonmonochromated Cu Kα radiation. The samples were left open to the atmosphere for hydration. 2.2. Data Collection and Analysis. Neutron powder diffraction (NPD) data were collected at the high resolution powder diffractometer ECHIDNA at the Australian Centre for Neutron Scattering (ACNS), Australian Nuclear Science and Technology Organisation (ANSTO),13 using a monochromatic neutron beam with a wavelength of 2.4395 Å. 4950


Article

Chemistry of Materials Inelastic neutron scattering (INS) data were collected at room temperature on the cold neutron time-of-flight spectrometer PELICAN at the ACNS, ANSTO.14,15 Hydrated samples were placed into annular aluminum cans with a total sample thickness of 4 mm and placed in the sample chamber under vacuum to ensure that all surface water was removed. Data were collected for 12 h per sample at an incident wavelength λ = 4.7 Å. A vanadium standard was measured for detector normalization and to determine the resolution function. The background was corrected by subtracting an empty sample can. The corrected time-of-flight spectra were then converted to S(Q,ω). The generalized density of states (GDOS) is obtained from S(Q,ω) using the equation ÄÅ ÉÑ ÅÅ Ñ ω jij ω zyzÑÑÑ ÅÅ zÑÑ g (Q , ω) = 2 S(Q , ω)ÅÅ1 − expjj− z j kBT zÑÑ ÅÅ Q (1) ÅÇ k {ÑÖ

Figure 2. Tentative hydration mechanism for γ-Ba4Nb2O9 based on forming NbO4OH polyhedra in the NbO4/NbO5 layers. O atoms are red; H atoms are white, and NbO5 trigonal bipyramids and NbO4 tetrahedra (and their corresponding NbO4OH polyhedra) are drawn in orange and yellow, respectively.

tion/hydrogenation of all NbO4/NbO5 polyhedra in the unit cell (see Figure 3).

followed by integration over Q. A suitable multiplicative scaling factor was applied to allow for comparison between experimental and simulated GDOS. All data manipulations used the Large Array Manipulation Program (LAMP).16 2.3. Ab Initio Simulations. Ab initio density functional theory (DFT) calculations were performed using the Vienna ab initio simulation package (VASP).17,18 The generalized gradient approximation (GGA) in the scheme of Perdew−Burke−Enzerhof (PBE)19,20 was employed. Projector augmented wave (PAW) pseudopotentials with a plain wave cutoff energy of 400 eV (determined by oxygen) were used throughout all simulations. The Brillouin zone was sampled via automatic k-mesh generation using a Γ-centered grid for hexagonal unit cells and the Monkhorst−Pack grid for all other symmetries. Single point energy (SPE) and geometry optimizations (GO) on the single hexagonal Ba4Ta2O9 unit cell were performed with a 2 × 2 × 1 k-mesh. Only one single k-point was used for calculations on larger supercells as well as ab initio molecular dynamics (AIMD) simulations. Total energies were converged within the self-consistent scheme with a tolerance of 10−6 eV. Structures were optimized by allowing ions to relax to their minimum energy position using the conjugate gradient algorithm.19 Lattice parameters were fixed to their experimentally determined values.9,11 In each step, all forces were relaxed below 0.01 eV Å−1. These optimized structures were used as starting configurations for subsequent AIMD simulations with a less dense FFT grid (75% of the original) and a fixed time step of 1 fs. AIMD simulations were run at temperatures of 600, 1000, and 1250 K for the fully hydrated, partially hydrated, and dehydrated structures, respectively. At each temperature, an initial equilibration simulation of approximately 1000 fs was performed using the canonical ensemble (NVT), where a constant temperature was achieved through a simple velocity scaling scheme as implemented in VASP. Temperature stability was carefully monitored throughout; Table S3 of the Supporting Information file gives values for target T, average T, and T variation for the runs. The obtained equilibrated velocity distribution was subsequently used as a starting configuration for an NVE (constant energy) simulation. NVE AIMD simulations were typically run for at least 12 000 fs for samples based on Ta and at least 30 000 fs for Nb-based samples. Trajectories were viewed within LAMP16 and analyzed using the programs nMOLDYN21 and MDANSE.22 This included the calculation of mean square displacement (MSD) parameters as well as the GDOS from the velocity autocorrelation function.

Figure 3. Structure of fully hydrated (a) and partially hydrated (b) γBa4Nb2O9 (cf. Figure 2 for atom color schemes).

A fully ordered model for the hydrated version can be constructed as follows. In dehydrated γ-Ba4Nb2O9, there are two crystallographically distinct tetrahedrally coordinated Nb sites, Nb3 and Nb7 (atom sites are labeled following the notation used by Dunstan et al.12). The additional oxygen atom was added to the corresponding Nb coordination sphere exactly opposite to the oxygen with the longest Nb−O distance (O34 and O36 for Nb3 and Nb7, respectively), i.e., ÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷◊ Onew1 = Nb3 + O34Nb3 (2) ÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷◊ Onew2 = Nb7 + O36Nb7 (3) followed by the addition of one proton to each new oxygen according to ÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷◊ H1 = Onew1 + nO34Nb3 (4) ÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷÷◊ H2 = Onew2 + nO36Nb7 (5) where the factor n accounts for normalization and rescaling of the Nb−O distance to obtain an O−H distance value of 1 Å. Similarly, one more proton was added to one selected oxygen on each of the two crystallographically distinct NbO 5 polyhedra (H3 to O39 around Nb5, H4 to O44 around Nb9). The obtained fully hydrated model of γ-IIIBa4Nb2O9·1/3H2O is shown in Figure 3a. Although powder samples readily hydrate, attempts to fully hydrate γ-Ba4Nb2O9 crystals were unsuccessful. Hydration of the crystal surface was clearly observed; however, it was not possible to fully hydrate even small crystals (of suitable size for single crystal XRD experiments) without significantly degrading their quality due to the large change in unit cell volume. The next obvious step was therefore to attempt to solve the hydrated crystal structure from NPD data. Rietveld refinements against high-resolution NPD data collected on

3. RESULTS AND DISCUSSION 3.1. Ba4Nb2O9. The highly ordered superstructure of γBa4Nb2O9 suggests a hydration mechanism involving hydroxylation of NbO4 and hydrogenation of NbO5 polyhedra to NbO4OH, as illustrated in Figure 2. According to previously reported TGA results11 (see above), the fully hydrated structure absorbs 1/3 of a water molecule per formula unit, i.e., γ-III-Ba4Nb2O9·1/3H2O. This corresponds to hydroxyla4951


Article

Chemistry of Materials ECHIDNA for γ-III-Ba4Nb2O9·1/3H2O using the dehydrated model (a) and a fully ordered version of a hydrated model (b), as shown in Figure 4. Refined fit parameters were almost identical (see Table S1 of the Supporting Information file), although a small improvement was seen for the hydrated model.

the fully ordered hydrated model using a coarse optimization algorithm (conjugate gradient algorithm) were attempted, but convergence could not be achieved. Furthermore, the real structure is almost certainly partially disordered and thus very difficult to solve definitively. The approach we adopted was thus to investigate the dynamics in these materials by means of AIMD simulations at elevated temperatures (to overcome potential energy barriers and maximize the number of diffusion events in these computationally expensive calculations) and then validate those simulations against experimental INS data. This includes simulations on fully hydrated γ-III-Ba4Nb2O9·1/3H2O at 600 K (just below the first dehydration step), partially hydrated γ-IIBa4Nb2O9·1/6 H2O at 1000 K, and dehydrated γ-Ba4Nb2O9 at 1250 K. Note that the 600 K simulation can be legitimately compared to room-temperature INS data because the hydrated species cannot “escape” the simulated unit cell, even at temperatures where the real materials dehydrate. At the same time, high-temperature INS data are less useful than roomtemperature data in this case because the sample stoichiometry is uncertain (noting that the dehydration temperatures are affected by a vacuum9), while the width of the excitations are broadened, making them more difficult to usefully interpret. Note that we did not convolute the simulated density of states with the experimental energy resolution because the resolution for the simulations (0.2 meV across all energy transfers) is significantly better than for the experimental data, especially at higher energy transfers. The instrument used in the INS experiment (PELICAN) has an energy resolution 8 meV at an energy transfer of 60 and 16 meV at an energy transfer of 100 meV. The instrumental resolution function is given in Figure S1 of the Supporting Information file. Starting Models. Due to the large superstructure, all simulations were reasonably performed within one unit cell (approximately 12.11 × 24.96 × 21.15 Å). Lattice parameters were fixed to their experimental values at the relevant temperatures. As a starting model for γ-III-Ba4Nb2O9·1/3H2O, a fully ordered hydrated model was used, i.e., full hydroxylation and hydration of all NbO4 and NbO5 polyhedra, respectively, according to the mechanism illustrated in Figure 2 and described above. On partial dehydration, the structure loses half the amount of water, resulting in γ-II-Ba4Nb2O9·1/6 H2O. There are two possibilities for partial dehydration: either complete dehydration of alternate layers or partial dehydration of each layer (see Figure 3b). Selected SPE calculations for these configurations yield very similar energies with complete dehydration of alternate layers slightly favored (−0.05 eV/f.u. on average). However, as the typical error in calculated energies from such DFT simulations is ∼10−20 meV, this difference at 0 K is clearly not significant at the elevated temperatures of interest. For the dehydrated model, the structure published by Dunstan et al.12 was used. To allow for structural relaxation and equilibration, an initial AIMD simulation using an NVT ensemble was performed (∼1000 fs with 1 fs time steps). Subsequent simulations of 30 ps (1 fs time step) were performed using an NVE ensemble with the equilibrated structures and velocity distributions as the starting configuration. Dynamics. Figure 5a shows the GDOS measured on PELICAN at room temperature using 4.7 Å neutrons compared to the GDOS as extracted from our ab initio MD simulation of the fully hydrated structure. The two curves show good agreement, with all the sharp peaks (15, 25, 35, and 46

Figure 4. Rietveld refinements of the dehydrated model (a) and the hydrated (b) model of γ-Ba4Nb2O9 against NPD data (ECHIDNA, λ = 2.4395 Å).

The fact that a completely ordered hydrated model and the dehydrated model both give reasonable fits to the data suggests that a structural solution from powder samples may not be feasible. Note that incoherent scattering from 1H in the samples was not found to cause significant problems with the background. Attempts to prepare deuterated samples to further minimize this background and to exploit the different neutron scattering lengths of D vs H were plagued by the extremely ready exchange with atmospheric water, even when the sample was sealed in vanadium cans for measurement. This resulted in samples with uncontrolled H/D ratios, data from which were ambiguous and therefore less informative than the naturally hydrated samples. Future experiments on neutron beamlines with in situ gas-flow setups might solve this problem by measuring under a humid deuterated atmosphere. A computational approach faces similar problems. The large number of plausible hydrated starting models results in a large number of local minima on the potential energy surface and thus a high risk of falling into a false minimum in the course of a structural geometry optimization (GO). GO starting from 4952


Article

Chemistry of Materials

and so can be confidently ascribed to O−H modes. This discrepancy therefore leads us to the feasible scenario that the O−H modes in γ-III-Ba4Nb2O9·1/3H2O are very anharmonic, which makes sense in terms of their chemical environment (Nb−O−H). This would cause strain in the system, normally driving toward a structural phase transition, consistent with the experimentally observed dehydration between the experimental (298 K) and simulated (600 K) temperatures. The partial pair distribution functions for γ-IIIBa4Nb2O9·1/3H2O extracted from the simulations are shown in Figure 6. The very sharp peak in PDFO−H at r = 1 Å

Figure 5. (a) Experimental (4.7 Å neutrons) vs simulated GDOS at room temperature for hydrated γ-III-Ba4Nb2O9·1/3H2O. (b) Partial contributions to the simulated GDOS at room temperature.

meV) accounted for. Differences in spectral weight between the measured and simulated spectra can be accounted for by variation in the energy-dependent resolution function. This gave us the confidence to further interrogate our simulations. Contributions to the GDOS at low energy transfers arise from barium; while at intermediate energy transfers, oxygen contributions dominate with hydrogen contributions becoming more substantial at even higher energy transfers (Figure 5b). We first note that for γ-III-Ba4Nb2O9·1/3H2O, in Figure 5 we see excellent agreement between experimental and simulated GDOS peaks from 0 to 140 meV, with one exception: a broad experimental peak at 70 meV is reproduced but appears at ∼85 meV in the simulated data. As already noted above, a 10 meV discrepancy is close to both the instrumental energy resolution (∼12 meV at these energies) and the typical error range for such DFT simulations. However, with the rest of the energy range so well fit, the discrepancy appears to be real and could potentially be considered in terms of anharmonicity, recalling that the experimental data were collected at 298 K (to ensure that the sample remained hydrated) while the simulations were carried out at 600 K (to maximize dynamics within the computationally feasible simulation time frame). The simulated peak appears in both the oxygen and hydrogen partials

Figure 6. (a) Partial pair distribution functions for γ-IIIBa4Nb2O9·1/3H2O and (b) PDFNb−O for hydrated, partially hydrated, and dehydrated γ-Ba4Nb2O9.

indicates that protons stay in the vicinity of oxygen atoms throughout the simulation and proton conduction through the structure occurs via transfer between neighboring oxygen atoms. At lower temperatures (fully hydrated structure), the PDF for Nb−O shows two peaks due to different Nb coordination environments: the peak at lower distances (d = 1.89 Å) corresponds to Nb in 5-fold coordination, while the smaller shoulder at d = 2.11 Å arises from octahedrally coordinated Nb atoms. This is in agreement with the effective ionic radii for Nb5+ in different coordination environments23 and Nb−O distances reported by Dunstan et al.12 At elevated temperatures (partially hydrated and dehydrated structures), the Nb−O peak broadens due to increased thermal vibrations (Figure 6b). We analyzed atomic motion through mean square displacement (MSD) parameters, i.e., the average displacement of a particle from its initial position within a given time interval t. 4953


Article

Chemistry of Materials Figure 7 shows the calculated MSDs for hydrated (600 K), partially hydrated (1000 K) and dehydrated (1273 K)

structures. In all simulations, MSDs for both metal cations (Ba2+ and Nb5+) plateau at very low values and stay constant over all larger time intervals. This indicates highly localized motion corresponding to thermal vibration around their respective equilibrium positions, as expected for these heavy ions. Oxygen shows quite different behavior. Average oxygen MSD curves initially increase linearly (diffusion) but plateau at larger time intervals. Such behavior is indicative of localized diffusion. A closer inspection of oxygen MSD curves at 1273 K reveals that this local diffusivity can be attributed to oxygens that are part of NbOx units, while no diffusion is observed for oxygens within Nb2O9 dimers. The three-dimensional clouds in Figure 8c show the space sampled by individual oxygen atoms over the course of the entire AIMD simulation at 1273 K. These ellipsoids elucidate the local diffusion behavior as onset of rotations of NbO4 and NbO5 polyhedra and highly localized motion (thermal vibrations) of oxygens that form part of Nb2O9 dimers. No long-range oxygen diffusivity was observed on the time scale of the simulation. Figures 8a and b show the space occupied by H atoms over the course of the AIMD simulation. This points to proton conduction occurring mainly within the A−B plane of NbO5 polyhedra. (Although the simulations show incidences of protons moving into the Nb2O9 dimer layers, they move back after a small number of steps, and we do not observe proton diffusion within the dimer layer itself.) However, the unusual shape of the hydrogen MSD curve in Figure 7a indicates that dynamic equilibrium might not yet have been reached, something corroborated by inspection of MSDs for individual oxygen atoms within the structure, which reveals large differences in their corresponding MSD curves even for crystallographically equivalent oxygen atoms. Longer simulation times would afford more detailed and more reliable insights into proton and oxygen conduction mechanisms in γBa4Nb2O9, but unfortunately, the large cell size means that these are not feasible at the AIMD level. Different approaches (e.g., classical MD) are necessary to further elucidate diffusion mechanisms, especially long-range diffusivity, in γ-Ba4Nb2O9.

Figure 7. MSDs for γ-Ba4Nb2O9 (a) at variable temperatures. Solid lines correspond to MSDs for hydrated γ-III-Ba4Nb2O9·1/3H2O at 600 K, dotted lines for partially hydrated γ-II-Ba4Nb2O9·1/6H2O at 1000K and dashed lines for dehydrated γ-Ba4Nb2O9 at 1273 K. (b) Oxygen MSDs for different coordination environments in γ-Ba4Nb2O9 at 1273 K.

Figure 8. Volumes (white clouds) visited by (a and b) hydrogen atoms in fully hydrated γ-III-Ba4Nb2O9·1/3H2O and (c) oxygen atoms in dehydrated γ-Ba4Nb2O9 (1273 K) over the course of 30 ps (1 fs time step) AIMD simulations. 4954


Article

Chemistry of Materials Nevertheless, our simulations thus far clearly support our model hydration mechanism. 3.2. Ba4Ta2O9. While for γ-Ba4Nb2O9, the structure suggested a clear hydration mechanism, there are no obvious locations for such sites in 6H-Ba4Ta2O9.11 For simplicity, in our initial structural analysis, we consider high-temperature 6H-Ba4Ta2O9 in the hexagonal space group P63/m (176), containing two formula units per unit cell. The largest voids in the structure are located between Ta2O9 dimers and BaO6 octahedra, i.e., close to the crystallographic 6g site (1/2,1/2,1/2). Including an interstitial oxygen, iO, at the 6g site (see Figure 9a), P63/m symmetry generates a total

Table 1. Calculated Energy Differences (with Respect to the Lowest Energy Configuration) For Different Proton Positions in 6H-Ba4Ta2O9·OHa position

SPE (eV/f.u.)

optimized (eV/f.u.)

O1-2-5 O1-1-0 O2-3-0 O2-3-2 O3-1-1 O3-2-6 iO-1-1 iO-3-3

2.04 2.88 2.69 2.67 2.17 2.18 0.00 0.00

0.27 0.48 1.41 1.37 0.62 1.30 0.00 0.00

a

The typical error in calculated energies from such DFT simulations is ∼10−20 meV.

2. One of the three oxygen atoms on top of polyhedra close to iO, either (a) H side or (b) opposite H side; 3. One of the three oxygen atoms in the center of polyhedral, either (a) H side or (b) opposite H side; and 4. One of the three oxygen atoms on the bottom of polyhedra away from iO, either (a) H side or (b) opposite H side. The same procedure as described above (spheres of protons around corresponding oxygen atoms, see Figure 9b) was performed to construct a potential energy surface with slightly lower resolution (only 15 points per proton-sphere around each oxygen). Interestingly, the results from SPE calculations suggested that the second proton is most stable when attached to the interstitial O−H group, forming a discrete water molecule (Figure 9c). Calculated energies for different proton positions (including a second proton) are summarized in the Supporting Information file (Table S2). To test for the stability of the water molecule within the structure, an AIMD simulation at 1273 K was performed. The high temperature was chosen to increase the total energy of the system and thus provide sufficient energy to move out of possible false minima. The simulation shows the entire water molecule diffusing (intact) from its initial position, centered at (1/2,1/2,1/2), to a symmetry equivalent position. A representative example of this motion is shown in an animation extracted from the simulation, uploaded as Supporting Information. This result, combined with the fact that the intercalated water does not dissociate even at these high (simulated) temperatures (far above the dehydration temperature) further support the hypothesis that intact water molecules are intercalated into 6H-Ba4Ta2O9 rather than H+ and OH−, as for γ-Ba4Nb2O9. Dynamics. For subsequent AIMD simulations, the unit cell was doubled in the a and b directions. To obtain the experimental value of 0.375 H2O per formula unit, a total of three water molecules had to be included in the 2 × 2 × 1 cell. This gives a total of 24 possible positions for a water molecule in the simulation cell, in 24 3 = 2024 possible arrangements. Most of these arrangements are equivalent because symmetry reduces the number of crystallographically unique input structures to 25. SPE and GO calculations were performed to test the stability of these arrangements. Additional tests were carried out for water molecules in different orientations. No obvious preference for any arrangement was found, indicating disorder of the water within the 6H-Ba4Ta2O9 structure. GO of hydrated structures leads to a slight tilting/shifting of the

Figure 9. Crystal structure of 6H-Ba4Ta2O9 (Ba atoms, green; Ta atoms and polyhedra, blue; O atoms, red; H atoms, white) including (a) one single interstitial oxygen at the 6g site (1/2,1/2,1/2). (b) Possible proton sites at a distance of 1 Å from each oxygen (the large H atom corresponds to the energetically most stable H position). (c) An intercalated H2O molecule.

of four crystallographically distinct oxygen atoms in the unit cell: O1 in the middle of two face sharing octahedra, O2 at the bottom of the face sharing octahedra (away from iO), O3 at the top of face sharing octahedra (close to iO), and iO. Based on this, we constructed a potential energy surface for possible proton positions within the structure. As illustrated in Figure 9, protons were placed on a sphere of radius r = 1 Å (typical O− H bond distance24) around each of the crystallographically distinct oxygen atoms within the structure (18 points on each sphere around each oxygen atom), giving 4 × 18 = 72 configurations. SPEs were calculated for each of the 72 configurations. The two lowest-energy configurations for each oxygen atom were then run through DFT GO. Results are summarized in Table 1. In general, the configurations with the proton attached to the interstitial oxygen show the lowest energy (−244 vs −247 eV in the nonrelaxed structure). The least stable configurations were found to be those with the proton positioned furthest away from the interstitial oxygen iO. GO leads to slight relaxations (twists) of the Ta2O9 polyhedra, while the O−H distance stays close to 1 Å (0.991 Å). Based on these results, we now assume that one proton is attached to the interstitial oxygen (larger hydrogen atom in Figure 9b), according to configuration iO-1-1. The resulting possible proton positions for the second proton in the structure are attached to oxygen atoms as follows: 1. iO;

( )

4955


Article

Chemistry of Materials Ta2O9 dimers, creating more space for the water molecules. This is consistent with the unusual distortion of those dimers observed at the transition from the hexagonal to the monoclinic structure (∼900 K) reported in earlier diffraction studies.9 Figure 10 shows the GDOS measured on PELICAN at room temperature using 4.7 Å neutrons compared to the GDOS as

Figure 11. Volumes (white clouds) visited by (a) water molecules and (b) oxygen atoms in partially hydrated 6H-Ba4Ta2O9 over the course of 30 ps (1 fs steps) AIMD simulations at 873 K.

the entire AIMD simulation. Despite the high simulation temperatures, the H2O does not dissociate into H+ and OH− at any time during the simulation. No evidence for long-range diffusion was observed in any of our AIMD simulations (Figure 11b). All MSDs saturate at low values within the first few fs of the simulation.

Figure 10. (a) Experimental (4.7 Å neutrons) vs simulated GDOS for hydrated 6H-Ba4Ta2O9 at room temperature. (b) Partial contributions to the simulated GDOS at room temperature.

4. CONCLUSIONS The combination of INS experimental data and AIMD simulations used in this study have allowed us to identify very different mechanisms of hydration for the quenched hightemperature phases of Ba4Nb2O9 and Ba4Ta2O9. In the case of γ-Ba4Nb2O9, water is taken up as OH− and H+ ions, which combine with NbO4 and NbO5 polyhedra respectively to form NbO4OH polyhedra. A clear and sharp peak in the partial O− H PDF of AIMD simulations indicates that protons then move through the structure by hopping between oxygen atoms of adjacent NbO4OH polyhedra. H ellipsoids (volumes explored by H atoms over the course of the simulation) indicate that protons predominantly move within the layers formed by these polyhedra. At higher temperatures, when γ-Ba4Nb2O9 becomes a predominantly oxide-ionic conductor, AIMD simulations of the dehydrated structure show that oxygen diffusion occurs via rotation of NbO4 and NbO5 units and oxygen exchange between them, while oxygen atoms in the layers of Nb2O9

extracted from our AIMD simulation of the fully hydrated structure. The two curves show excellent agreement. The feature at low energy transfers (∼7 meV) comes from Ba, while at larger energy transfers, the oxygen contributions dominate. As for γ-Ba4Nb2O9, we can compare a high-temperature simulation (which provide better statistics on diffusion events) to room-temperature INS data (which provide narrower line widths and remove uncertainly about the stoichiometry) because hydrated species are effectively trapped in the simulated unit cell. The only significant visual difference is the presence of a sharp peak at ∼80 meV in the simulated GDOS. This could be an artifact of the periodic boundary conditions used and could be revisited by simulating a larger unit cell, if computational resources permit in the future. Figure 11a shows the space occupied by intercalated water molecules in partially hydrated 6H-Ba4Ta2O9 at 1000 K over 4956


Article

Chemistry of Materials Funding

dimers vibrate only around their equilibrium positions. While the local mechanisms of proton and oxygen diffusion in γBa4Nb2O9 are now clear, a future study using classical MD to access longer simulation times would be worthwhile to gain insight into the long-range processes that result in bulk conductivity. In the case of 6H-Ba4Ta2O9, our results points toward the intercalation of intact H2O molecules into the lattice. The nature of this hydration mechanism means that the water content is not stoichiometrically constrained like it is in the niobate. Although no diffusive behavior was observed at simulation temperatures lower than the experimental dehydration temperatures, an AIMD simulation at 1273 K showed concerted diffusion of a water molecule from its initial position to a neighboring symmetry equivalent position in the structure. This further supports the hypothesis that the tantalate truly hydrates rather than hydroxylates like the niobate. This surprising behavior, for what is essentially a close-packed oxide structure, explains the unusual symmetry distortions previously observed for 6H-Ba4Ta2O9 at room-temperature under ambient conditions (i.e., in the hydrated form). A future detailed IR spectroscopy study under controlled atmospheres would help to corroborate the presence of molecular water in the bulk lattice of this compound; note that a key challenge here will be to control the conditions so that the sample hydrates internally but does not adsorb any surface water, to which these measurements will be disproportionally sensitive. Finally, a future variable-temperature 1H/17O solid-state NMR study could provide insights by probing the link between oxygen and proton dynamics. These fundamentally different hydration mechanisms (hydroxylation vs hydration) explain why the niobate shows higher proton conductivity despite its lower water uptake (1/3 H2O per f.u., compared to 0.375 H2O per f.u. in the tantalate). Noting that the higher water uptake by the tantalite appears to be related to its higher polarizability, a promising avenue for future work aimed at optimizing mixed conductive behavior in these systems would be to explore hydration behavior in the γtype phase space of the Ba4Nb2−xTaxO9 solid solution.25

■

This work was supported by the Australian Research Council (DP150102863) and the Australian Institute of Nuclear Science and Technology (PGRA scheme). Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This research was undertaken with the assistance of resources and services from the National Computational Infrastructure (NCI), which is supported by the Australian Government. The authors thank Gordon J. Kearley of the Australian Nuclear Science and Technology Organisation (ANSTO) for helpful discussions.

■

(1) Iwahara, H. Proton Conduction in Sintered Oxides Based on BaCeO3. J. Electrochem. Soc. 1988, 135 (2), 529. (2) Fabbri, E.; Pergolesi, D.; Traversa, E. Materials challenges toward proton-conducting oxide fuel cells: a critical review. Chem. Soc. Rev. 2010, 39 (11), 4355−69. (3) Bhide, S. V. Stability of BaCeO3-Based Proton Conductors in Water-Containing Atmospheres. J. Electrochem. Soc. 1999, 146 (6), 2038. (4) Tanner, C. W. Instability of BaCeO3 in H2O-Containing Atmospheres. J. Electrochem. Soc. 1996, 143 (4), 1386. (5) Malavasi, L.; Fisher, C. A.; Islam, M. S. Oxide-ion and proton conducting electrolyte materials for clean energy applications: structural and mechanistic features. Chem. Soc. Rev. 2010, 39 (11), 4370−87. (6) Du, Y. Galvanic cell measurements on a fast proton conducting complex perovskite electrolyte. Solid State Ionics 1996, 91 (1−2), 85− 91. (7) Haugsrud, R.; Norby, T. Proton conduction in rare-earth orthoniobates and ortho-tantalates. Nat. Mater. 2006, 5 (3), 193−196. (8) Haugsrud, R.; Norby, T. High-temperature proton conductivity in acceptor-doped LaNbO4. Solid State Ionics 2006, 177 (13−14), 1129−1135. (9) Ling, C. D.; Avdeev, M.; Kutteh, R.; Kharton, V. V.; Yaremchenko, A. A.; Fialkova, S.; Sharma, N.; Macquart, R. B.; Hoelzel, M.; Gutmann, M. Structures, Phase Transitions, Hydration, and Ionic Conductivity of Ba4Nb2O9. Chem. Mater. 2009, 21 (16), 3853−3864. (10) Bezjak, J.; Abakumov, A. M.; Rečnik, A.; Maček Kržmanc, M.; Jančar, B.; Suvorov, D. The local structure and composition of Ba4Nb2O9-based oxycarbonates. J. Solid State Chem. 2010, 183 (8), 1823−1828. (11) Ling, C. D.; Avdeev, M.; Kharton, V. V.; Yaremchenko, A. A.; Macquart, R. B.; Hoelzel, M. Structures, Phase Transitions, Hydration, and Ionic Conductivity of Ba4Ta2O9. Chem. Mater. 2010, 22 (2), 532−540. (12) Dunstan, M. T.; Blanc, F.; Avdeev, M.; McIntyre, G. J.; Grey, C. P.; Ling, C. D. Long-Range-Ordered Coexistence of 4-, 5-, and 6Coordinate Niobium in the Mixed Ionic-Electronic Conductor γBa4Nb2O9. Chem. Mater. 2013, 25 (15), 3154−3161. (13) Liss, K.-D.; Hunter, B.; Hagen, M.; Noakes, T.; Kennedy, S. Echidnathe new high-resolution powder diffractometer being built at OPAL. Phys. B 2006, 385−386, 1010−1012. (14) Yu, D.; Mole, R.; Noakes, T.; Kennedy, S.; Robinson, R. Pelican - a Time of Flight Cold Neutron Polarization Analysis Spectrometer at OPAL. J. Phys. Soc. Jpn. 2013, 82 (Suppl.A), SA027. (15) Yu, D.; Mole, R. A.; Kearley, G. J. Performance test on PELICAN − a multi-purpose time of flight cold neutron spectrometer. EPJ Web Conf. 2015, 83, 03019. (16) Richard, D.; Ferrand, M.; Kearley, G. J. Analysis and visualisation of neutron-scattering data. J. Neutron Res. 1996, 4 (1), 33−39.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.chemmater.8b00950. Refinement details for γ-Ba4Nb2O9, calculated energies for different proton positions in 6H-Ba4Ta2O9·H2O, temperature parameters for the dynamics simulations, and the resolution function for the inelastic neutron spectrometer (PDF) Animation showing the simulated motion of water molecules in 6H-Ba4Ta2O9·H2O (AVI)

■

REFERENCES

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. ORCID

Maxim Avdeev: 0000-0003-2366-5809 Chris D. Ling: 0000-0003-2205-3106 Present Address §

J.W.: Department of Chemistry, University of Oslo, Blindern, P.O. Box 1033, 0315 Oslo, Norway. 4957


Article

Chemistry of Materials (17) Kresse, G.; Furthmüller, J. Efficient iterative schemes forab initiototal-energy calculations using a plane-wave basis set. Phys. Rev. B: Condens. Matter Mater. Phys. 1996, 54 (16), 11169−11186. (18) Kresse, G.; Joubert, D. From ultrasoft pseudopotentials to the projector augmented-wave method. Phys. Rev. B: Condens. Matter Mater. Phys. 1999, 59 (3), 1758−1775. (19) Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made Simple. Phys. Rev. Lett. 1996, 77 (18), 3865− 3868. (20) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1997, 78, 1396. (21) Rog, T.; Murzyn, K.; Hinsen, K.; Kneller, G. R. nMoldyn: a program package for a neutron scattering oriented analysis of molecular dynamics simulations. J. Comput. Chem. 2003, 24 (5), 657−67. (22) Goret, G.; Aoun, B.; Pellegrini, E. MDANSE: An Interactive Analysis Environment for Molecular Dynamics Simulations. J. Chem. Inf. Model. 2017, 57 (1), 1−5. (23) Shannon, R. D. Revised effective ionic radii and systematic studies of interatomic distances in halides and chalcogenides. Acta Crystallogr., Sect. A: Cryst. Phys., Diffr., Theor. Gen. Crystallogr. 1976, 32 (5), 751−767. (24) Kato, K.; Toyoura, K.; Nakamura, A.; Matsunaga, K. Proton Channels along Oxygen Octahedral Chains in La3NbO7. J. Phys. Chem. C 2014, 118 (18), 9377−9384. (25) Dunstan, M. T.; Southon, P. D.; Kepert, C. J.; Hester, J.; Kimpton, J. A.; Ling, C. D. Phase diagram, chemical stability and physical properties of the solid-solution Ba4Nb2−xTaxO9. J. Solid State Chem. 2011, 184 (10), 2648−2654.

4958


Hydration Mechanisms and Proton Conduction in the Mixed Ionic

Recommend Documents