Safe and efficient power generation technologies using renewable sources are becoming increasingly important for sustainable energy development, with solid oxide fuel cells (SOFCs) based on oxide-ion conductors being a typical example1,2,3,4,5. However, the high operating temperature of conventional SOFCs (higher than 700 °C) has limited their applications in a wide range of fields at low and intermediate temperatures (300–600 °C). Since the activation energy for proton (H+) conductivity is lower than that for oxide ion (O2−) conductivity, proton conductors are known to exhibit higher conductivity than oxide-ion conductors at low temperatures6,7,8. Polymeric membranes such as Nafion and solid acid proton conductors such as CsHSO4 show high conductivity near room temperature; however, they are chemically unstable and decompose at 100–200 °C9,10,11,12. In addition, materials that function at such low temperatures require expensive precious metal catalysts when used in fuel cells12,13. Some hydride (H) ion conductors show high conductivity at low temperatures14,15,16. However, they have drawbacks, such as being prone to decomposition at high temperatures or can only be synthesized by the high-pressure method15,17.

In recent years, protonic ceramic fuel cells (PCFCs), which utilize proton-conducting ceramics, have been attracting attention as a technology alternative to SOFCs7,18,19. Proton-conducting oxides used in PCFCs are not only stable on heating but also have a lower operating temperature than SOFCs, enabling a stable energy supply at a lower cost7,20,21,22,23,24,25,26. To achieve high performance of PCFCs, high proton conductivity and high proton transport numbers are required. However, since proton conduction is dominated by the anion network, a limited number of structural types exhibit high proton conductivity, such as perovskites27,28,29,30,31, layered perovskites32,33,34,35,36,37, fluorite-type oxides38,39, apatites40, brownmillerites41, fergusonites42, and La28−xW4+xO54+δ43. Therefore, it is an important and challenging task to search for proton conductors belonging to other structure types.

A conventional strategy in the search for proton-conducting materials is the acceptor doping to form ‘extrinsic oxygen vacancies’ to allow hydration and obtain higher proton conductivity than that of the parent material44,45. The doping, however, often results in compositional inhomogeneity, reaction with electrode materials, proton trapping near dopants, and instability46. An alternative strategy is the search for a stable compound containing ‘intrinsic oxygen vacancies’ \({{{{{{\rm{V}}}}}}}_{{{{{{\rm{VO}}}}}}}^{\times }\), which enables hydration and proton conduction without the acceptor doping. Here, \({{{{{{\rm{V}}}}}}}_{{{{{{\rm{VO}}}}}}}^{\times }\) represents the vacancy V at the intrinsic oxygen vacant site VO using the Kröger–Vink notation. The intrinsic oxygen vacant site can be defined using the close-packed polytypes as follows. Crystal structures of many AMX3 perovskites and perovskite-related compounds contain cubic and/or hexagonal close-packed AX3 layers where the A and M are larger and smaller cations, respectively, and X is an anion. Intrinsically oxygen-deficient AO3−δ layers are sometimes formed for the hexagonal and cubic close-packed layers (labeled h′ and c′, respectively), where δ is the amount of oxygen deficiency. Recently, perovskite-related oxides containing intrinsic oxygen vacancies have been reported to exhibit significant proton conduction without chemical doping (e.g., Ba5Er2Al2ZrO1337, Ba5In2Al2ZrO1347, BaY1/3Ga2/3O2.548, and Ba7Nb4MoO2036). Therefore, it is interesting to search for a material with a large number of intrinsic oxygen vacancies in its structure, which may show hydration and proton conduction without any chemical doping.

In this work, we searched for Ba2BMO5 compounds (B, M: cations) as candidates for proton conductors with intrinsic oxygen vacancies. The composition Ba2BMO5 (= BaB1/2M1/2O2.5) has a lower oxygen content (5/6) compared with ABO3 perovskite (1), and hence significant intrinsic oxygen vacancies are expected to exist in Ba2BMO5. To the best of our knowledge, we discovered a new material Ba2LuAlO5 through the search of Ba2BMO5 oxides (See the details in Supplementary Note 1 and Supplementary Fig. 1). Single-crystal and powder X-ray and neutron diffraction analyses reveal that Ba2LuAlO5 is a hexagonal perovskite-related oxide with highly oxygen-deficient h′ layers. In the present work, we report that the undoped Ba2LuAlO5 shows high bulk proton conductivity of 102 S cm1 at 487 °C and 1.5 × 103 S cm1 at 232 °C, which is higher than those of leading proton conductors. We also demonstrate that the hydration occurs in the h′ layer and that protons migrate mainly around cubic close-packed c layers existing at the interface of two octahedral LuO6 layers. Therefore, the search for hexagonal perovskite-related oxides with both h′ and c layers would be a strategy to develop high-performance proton conductors.

Results and discussion

Synthesis and characterization of Ba2LuAlO5

Ba2LuAlO5 samples were prepared by solid-state reactions (See the details in “Methods”). The average grain and pore sizes were estimated to be 3.2(2) and 0.56(5) µm, respectively (Supplementary Fig. 2). The crystal structure of as-prepared Ba2LuAlO5 sample was successfully analyzed by the hexagonal P63/mmc space group using the single-crystal X-ray diffraction (SCXRD) data (Supplementary Figs. 3 and 4; Supplementary Tables 1, 2 and 3). Details of the structure analysis are described in Supplementary Note 2. All the peaks in the X-ray powder diffraction (XRPD) pattern of as-prepared Ba2LuAlO5 were indexed to the primitive hexagonal lattice (Fig. 1a). Dry Ba2LuAlO5 samples were synthesized by annealing the as-prepared samples in vacuum at 800 °C. Rietveld refinements of the XRPD data for as-prepared Ba2LuAlO5 and neutron diffraction (ND) data for dry Ba2LuAlO5 were successfully carried out based on the hexagonal P63/mmc Ba2LuAlO5 structure (Fig. 1b, Supplementary Fig. 5a, Supplementary Table 4). The bond valence sums of all the cations and anions for the refined structures of as-prepared and dry Ba2LuAlO5 are consistent with their formal charges, demonstrating the validity of the refined crystal structures (Supplementary Tables 2 and 4).

Fig. 1: Crystal structure analysis of as-prepared Ba2LuAlO5.
figure 1

a X-ray powder diffraction profile and b Rietveld pattern of as-prepared Ba2LuAlO5 at 22 °C (Rwp = 0.084). The red crosses, dark blue lines, and black dots represent observed, calculated, and difference intensities, respectively. Green ticks denote calculated Bragg peak positions of hexagonal Ba2LuAlO5.

Hydrated (deuterated) Ba2LuAlO4.52(OD)0.96 samples were synthesized by annealing the as-prepared samples in D2O saturated air. To investigate the atomic coordinates and occupancy of proton and oxide ions, ND data of the hydrated Ba2LuAlO4.52(OD)0.96 sample were measured at 5 K. Rietveld refinements of ND data for the hydrated Ba2LuAlO4.52(OD)0.96 were successfully carried out based on the hexagonal P63/mmc Ba2LuAlO5 structure (Supplementary Fig. 5b, Supplementary Table 5). In preliminary analyses, the occupancy factors were found to be unity at all the cation sites and O1, O2, and O3 sites. The occupancy factors of Lu and Al atoms were refined to be 1 and 0, respectively, at the Lu site and 0 and 1, respectively, at the Al site, indicating complete Lu/Al occupational order. The oxygen 2b (0, 0, 1/4) site of hydrated Ba2LuAlO4.52(OD)0.96 was found to split into the 6h (x, x/2, 1/4) O3 site with 1/3 occupancy, as in the case of as-prepared and dry Ba2LuAlO5, and Ba5Er2Al2ZrO1337. Extra oxygen atoms of Ba2LuAlO4.52(OD)0.96 due to the hydration were located at the interstitial 2c (1/3, 2/3, 1/4) O4 site in the h′ layer, as observed in Ba5Er2Al2ZrO13 and Ba5Sc1.33Al2Zr1.67O13.3337,49. The reliability factor for the structural model with the O4 atom Rwp = 12.342% was much lower than that without the O4 atom Rwp = 18.352%. Further structure analyses of the ND data at 5 K on the basis of 416 models for the proton positions suggested four deuterium sites D1, D2, D3, and D4 (Supplementary Table 5), which is consistent with the probability density distribution of protons obtained by the ab initio molecular dynamics (AIMD) simulations (Supplementary Fig. 7). The average OD bond length 0.99(3) Å agrees with those from Raman spectra 0.98(4) Å (Supplementary Fig. 8) and IR spectra 0.99(4) Å (Supplementary Fig. 9). The lattice parameters a and c of hydrated Ba2LuAlO4.52(OD)0.96 are 0.26% and 0.82% larger than those of dry Ba2LuAlO5, respectively, at 5 K, which is attributable to the water incorporation. The final structure refinements yielded good fits and low reliability factors (Rwp = 9.164%, Supplementary Fig. 5b, Supplementary Table 5). The bond valence sums of all the cations and anions at all the sites are consistent with the formal charges (Supplementary Table 5). The refined occupancy factor of O4 atom was 0.957(14), from which the bulk water content x in Ba2LuAlO5 · x D2O is calculated to be x = 0.479(7), indicating the chemical formula Ba2LuAlO4.52(OD)0.96 [= Ba2LuAlO4.522(7)(OD)0.957(14) = Ba2LuAlO5 · 0.479(7) D2O = Ba2LuAlD0.957(14)O5.479(7)]. This value is consistent with that from TG measurements x = 0.50 (Supplementary Fig. 6), indicating that most of the water is incorporated in bulk Ba2LuAlO4.522(7)(OD)0.957(14). These results validate the refined crystal structure of Ba2LuAlO4.522(7)(OD)0.957(14) at 5 K.

Crystal structures of dry, hydrated, and as-prepared Ba2LuAlO5

Dry Ba2LuAlO5, hydrated Ba2LuAlO5 · x D2O, and as-prepared Ba2LuAlO5 samples have a hexagonal perovskite-related structure with cubic close-packed BaO3 (c) layers ((Ba1)(O1)3 and (Ba3)(O2)3 layers) and intrinsically oxygen-deficient hexagonal close-packed BaO1+ε (h′) layer ((Ba2)(O3)(O4)ε layer) in the sequence of (ccch′)2, where ε is the occupancy factor of O4 atom (Fig. 2 and Supplementary Fig. 3). The structure of dry and as-prepared Ba2LuAlO5 consists of LuO6 octahedra, AlO4 tetrahedra, and Ba cations, whereas the structure of hydrated Ba2LuAlO5 · x D2O consists of LuO6 octahedra, AlO4 tetrahedra, Ba, interstitial oxygen O4 and D atoms. In the three samples, two LuO6 octahedra form a bioctahedron (octahedral dimer) Lu2O11 by sharing O1 atom, while two AlO4 tetrahedra form an Al2O7 dimer by sharing O3 atom.

Fig. 2: Crystal structures of dry Ba2LuAlO5 and hydrated Ba2LuAlO5 · 0.48 D2O.
figure 2

Refined crystal structures of dry a Ba2LuAlO5 and b hydrated Ba2LuAlO5 · 0.48 D2O at 5 K. Green, purple, light blue, and black balls represent Ba, Lu, Al, and O atoms, respectively. In (b), red balls denote D1 atoms, and orange balls represent D2, D3, and D4 atoms.

Ba2LuAlO5 is isostructural with β-Ba2ScAlO5 and hexagonal Ba2InAlO550,51. Hereafter, the structure type of Ba2LuAlO5 is referred to as the β-Ba2ScAlO5-type. At room temperature, the lattice parameters of as-prepared Ba2LuAlO5 (a = 5.9203(5) Å, c = 19.7448(19) Å) are larger than those of β-Ba2ScAlO5 (a = 5.79 Å, c = 19.35 Å)51 and Ba2InAlO5 (a = 5.78 Å, c = 19.62 Å)50 due to the larger ionic radius of Lu3+ cation for coordination number of 6 (0.861 Å) compared with those of Sc3+ (0.745 Å) and In3+ (0.8 Å)52. The split sites were found for O3 and Ba2 in both dry and as-prepared Ba2LuAlO5, although the split sites were not considered in β-Ba2ScAlO5 and hexagonal Ba2InAlO5 in the literature49,50. As shown later, Ba2LuAlO5 exhibits high proton conductivity as well as Ba5Er2Al2ZrO13, and they have similar crystal structures, with stacking sequences (ccch′)2 and (cccch′)2, respectively (Supplementary Note 3 and Supplementary Fig. 10). Therefore, the number of h′ layers per unit length along the c axis of as-prepared Ba2LuAlO5 (0.101292(10) h′ layers Å−1) is significantly larger than that of Ba5Er2Al2ZrO13 (0.081102(7) h′ layers Å−1). The number of oxygen vacancies per unit volume of as-prepared Ba2LuAlO5 (6.6740(13) × 10−3 Å−3) is also larger than that of Ba5Er2Al2ZrO13 (5.294(4) × 10−3 Å−3), leading to larger amount of H2O in hydrated Ba2LuAlO5·x H2O (x = 0.50) than hydrated Ba5Er2Al2ZrO13·x H2O (x = 0.27) at room temperature.

As shown in Fig. 3, XRPD patterns of Ba2LuAlO5 after annealing at 400 °C for 24 h in dry O2, dry 5% H2 in N2, and wet air remain very similar to that of the as-prepared sample, indicating its high chemical stability. Supplementary Fig. 11 shows the XRPD patterns of Ba2LuAlO5 after annealing in dry CO2 at 400 and 500 °C for 24 h, also indicating the high phase stability of Ba2LuAlO5, although a small amount of impurity Lu2O3 appeared during the CO2 annealing (weight fraction of Lu2O3 in the sample during CO2 annealing at 400 °C: 0.0057).

Fig. 3: High chemical stability of Ba2LuAlO5.
figure 3

X-ray powder diffraction patterns of Ba2LuAlO5 at room temperature. a As-prepared sample and samples after annealing at 400 °C in b dry O2, c dry 5% H2 in N2, d wet air [H2O vapor pressure of 0.021 atm] (100 mL min1) for 24 h. There are no additional peaks in these XRPD patterns after annealing, which indicates the high chemical stability of Ba2LuAlO5 in dry O2, dry 5% H2 in N2, and wet air.

High proton conduction in Ba2LuAlO5

The UV-vis spectrum of Ba2LuAlO5 powders showed a wide optical band gap of Eg = 3.99 eV, indicating that Ba2LuAlO5 is an electronic insulator (Supplementary Fig. 12), as supported by DFT calculations (Supplementary Fig. 13). Figure 4a shows oxygen partial pressure P(O2) dependencies of total DC electrical conductivities σtotal of Ba2LuAlO5 measured at 400 °C under dry atmosphere σ(dry) and wet condition σ(H2O). In the dry atmosphere, the slope of log(σtot) versus log(P(O2)) has a positive value in the P(O2) range from 1 to 10−5 atm, indicating p-type conduction. At low P(O2) range from approximately 10−5 to 10−25 atm, the conductivity σ(dry) is almost independent of P(O2), demonstrating electrolyte domain with negligible electronic conduction. Under the wet condition at 400 °C, σ(H2O) is almost independent of P(O2), and the ion conduction is dominant over the entire P(O2) range from 1 to 10−21 atm. The σ(H2O) value in the electrolyte domain is 35 times higher than that under the dry condition σ(dry). These observations strongly suggest that Ba2LuAlO5 is a proton conductor in the wet state. The proton conduction in Ba2LuAlO5 is further supported by the isotope effect on σtotal. We measured σtotal of Ba2LuAlO5 in D2O saturated air σ(D2O) and H2O saturated air σ(H2O) atmospheres. The conductivity ratio σ(H2O)/σ(D2O) is in the range of 1.2–1.6 (Fig. 4b), close to the expected value of 1.414 for the Grotthuss mechanism of proton transport53. Similar conductivity ratio values have been reported in known proton conductors24,54,55.

Fig. 4: Proton conduction in Ba2LuAlO5.
figure 4

a Oxygen partial pressure P(O2) dependencies of total DC electrical conductivity σtotal under dry atmospheres σ(dry) and under H2O saturated gas σ(H2O). b Temperature dependence of the σH+/σD+ ratio in N2 flow where the σH+ and σD+ are the proton and deuteron conductivities, respectively, in N2 flow. σH+ and σD+ were estimated by the equations σH+ = σ(H2O) − σ(dry) and σD+ = σ(D2O) − σ(dry), respectively, where the σ(H2O) and σ(D2O) are DC electrical conductivities σtotal under H2O and D2O saturated N2 flow, respectively. c Arrhenius plots of σ(dry) and σ(H2O) under N2 gas flow.

Figure 4c shows the Arrhenius plots of total DC electrical conductivity σtotal of Ba2LuAlO5 in dry N2 gas flow σ(dry) and wet N2 gas flow σ(H2O). In the whole temperature range, σ(H2O) is higher than σ(dry) (e.g., 63 times higher at 300 °C). The σ(H2O) increases with increasing temperature from 300 to 500 °C while decreasing beyond that. This behavior is typical of known proton conductors, including Ba5Er2Al2ZrO13, which is ascribed to thermal dehydration at high temperatures37. Indeed, thermal dehydration of Ba2LuAlO5 upon heating is observed in TG measurements (Supplementary Fig. 6). We can calculate proton transport number tH+ and proton conductivity σH+ assuming tH+ = σH+ / σ(H2O) and σH+ = σ(H2O) − σ(dry). The obtained tH+ is close to unity over the whole temperature range, indicating pure proton conduction (Supplementary Fig. 14).

The bulk conductivity (σbulk) of Ba2LuAlO5 was investigated in dry N2 and in H2O saturated N2. Figure 5b, c, Supplementary Figs. 15 and 16 show typical AC impedance spectra of Ba2LuAlO5. The σbulk and σgb of Ba2LuAlO5 were obtained by the equivalent circuit analysis (Fig. 5a; Supplementary Figs. 15a, b, 16a, b, and 17; Supplementary Note 4). As shown in Fig. 5a, the bulk conductivity in H2O saturated N2 (vapor pressure of 0.021 atm) σbulk(H2O) is higher than that in dry N2 σbulk(dry) in the whole temperature range, indicating bulk proton conduction. For example, the σbulk(H2O) at 400 °C is 130 times higher than σbulk(dry).

Fig. 5: High bulk conductivity of Ba2LuAlO5.
figure 5

a Arrhenius plots of bulk conductivities under dry N2 gas flow σbulk(dry) and under wet N2 gas flow σbulk(H2O) [P(H2O) = 0.021 atm]. b, c Complex impedance plots of Ba2LuAlO5 recorded in wet N2 gas flow at b 46.6 °C and c 487 °C. Blue numbers in panels (b) and (c) stand for the frequencies (Hz) at the points of light blue diamonds.

Figure 6a compares the proton conductivities σH of Ba2LuAlO5 and other proton conductors where the definition of σH is described for each data in the caption of Fig. 6. Ba2LuAlO5 exhibits σH value as high as 10−2 S cm−1 at 487 °C and 1.5 × 103 S cm1 at 232 °C. The conductivity is 4.3 times higher at 200 °C and 2.0 times higher at 400 °C than those of the cubic perovskite-type BaZr0.8Y0.2O2.956. The σH of Ba2LuAlO5 is even higher than that of Ba5Er2Al2ZrO13, a structurally-related oxide with intrinsically oxygen-deficient h′ layers37. The activation energy for proton conductivity in Ba2LuAlO5 is estimated to be 0.36 eV below 200 °C, which is lower than that of cubic-perovskite BaZr0.8Y0.2O3−δ (0.47 eV) and BaZr0.4Sc0.6O2.7 (0.44 eV)24. The proton diffusion coefficient D of Ba2LuAlO5 is calculated by the Nernst–Einstein equation using the proton concentration determined by TG measurements (see details in “Methods”). As shown in Fig. 6b, Ba2LuAlO5 shows higher D values than the acceptor-doped perovskite-type BaZr0.8Y0.2O2.9 and BaZr0.4Sc0.6O2.724,46. Diffusion coefficient D and activation energy for D extracted from the AIMD simulations are in good agreement with the experimental values (Supplementary Fig. 18), supporting the high D values. The acceptor doping is known to make the proton-dopant association, leading to the high apparent activation energy for D and low proton conductivity at low temperatures46. In sharp contrast, the present Ba2LuAlO5 has intrinsic oxygen vacancies without chemical doping, which leads to low activation energy and high proton conductivity at low temperatures.

Fig. 6: Higher proton conductivity and diffusion coefficient of Ba2LuAlO5 compared with other leading proton conductors.
figure 6

a Comparison of proton conductivities σH of Ba2LuAlO5 with other proton conductors. Shown are bulk conductivity of the present Ba2LuAlO5 in wet N2 and BZYO (BaZr0.8Y0.2O2.9) in wet air56, bulk conductivity of BCO (BaCe0.9Y0.1O2.95) in wet air56, total AC conductivity of BZSO (BaZr0.4Sc0.6O2.9) in wet Ar24, DC proton conductivity σH+ of Ba5Er2Al2ZrO1337, DC proton conductivity σH+ of BIAO (Ba5In2Al2ZrO13)47, bulk conductivity of LBGO (La0.8Ba1.2GaO3.9) in wet air69, DC conductivity of LPO (La0.95Sr0.05PO3.975) in wet O270, total AC conductivity of LMO (La5.4MoO11.1) in wet N239, bulk conductivity of LGS (La3Ga5.06Si0.94O14) in wet air71. b Arrhenius plots of proton diffusion coefficients D of Ba2LuAlO5, BZYO (BaZr0.8Y0.2O2.9)46 and BZSO (BaZr0.4Sc0.6O2.7)24.

We have demonstrated that Ba2LuAlO5 shows a high level of proton conductivity without chemical doping. This suggests that water is incorporated in the oxygen-deficient h′ layers, leading to the formation of proton carriers. TG analysis of undoped Ba2LuAlO5 indeed indicated a water uptake upon cooling (Supplementary Figs. 6, 19, and 20; Supplementary Table 6; Supplementary Note 5). Refined occupancy factor of oxygen atoms in the Rietveld analysis of hydrated (deuterated) Ba2LuAlO4.52(OD)0.96 also indicated the water uptake and presence of hydroxide ions OD (Fig. 2b). The water content x in Ba2LuAlO5·x H2O is estimated to be 0.50 at 100 °C by TG measurements, which is larger than that for Ba5Er2Al2ZrO13·x H2O (x = 0.27) at the same temperature. The concentration of water relative to the number of available oxygen vacancies for Ba2LuAlO5·x H2O (x = 0.50) is relatively low, 50%, compared with other proton conductors (Supplementary Fig. 21a). Meanwhile, the number of H2O per unit volume of Ba2LuAlO5 (3.3370(7) × 1021 cm3) is relatively high as shown in Supplementary Fig. 21b. The high proton diffusion coefficient and large water concentration in Ba2LuAlO5 may account for the high proton conductivity.

Proton diffusion mechanism through ab initio molecular dynamics simulations of Ba2LuAlO5

To gain more insight into the high proton conductivity in Ba2LuAlO5, ab initio molecular dynamics (AIMD) simulations were performed for Ba2LuAlO5 · 0.125 H2O using a 2×2×2 supercell (Ba32Lu16Al16O82H4). The initial structural model was made by locating two water molecules in two different h′ layers of the 2×2×2 supercell. The calculated mean square displacement (MSD) of protons is much larger than those of the other constituent atoms (Fig. 7a), indicating that proton conduction is dominant. As shown in Fig. 7b, the MSDs of protons along the a and b directions are larger than that along the c direction, showing the dominant two-dimensional diffusion in the ab-plane. The trajectory and density distribution of hydrogen atoms (Fig. 8, and Supplementary Figs. 7b and 22) indicate that two inserted water molecules dissociate into O4 atoms in the h′ layer and protons, which is consistent with the refined crystal structure (Supplementary Fig. 7a). As shown in Fig. 8 and Supplementary Fig. 22, there are two types of protons of four inserted protons: (i) two mobile protons around the (Ba1)(O1)3 c layer at the interface of two LuO6 octahedra and (ii) two trapped protons around the h′ layer. The trapped protons in the h′ layer move around an interstitial O4 oxygen atom formed by the hydration but do not migrate across the lattice. In contrast, the protons around the (Ba1)(O1)3 c layer move across the lattice. Therefore, the crystal structure of Ba2LuAlO5 consists of (1) proton conducting c layers and (2) oxygen-deficient h′ layers that incorporate the extra oxygen atoms due to hydration. Such structural character of Ba2LuAlO5 enables high proton conductivity.

Fig. 7: Mean square displacements of Ba2LuAlO5 · 0.125 H2O.
figure 7

AIMD-simulated mean square displacements (MSDs) of a constituent atoms and b protons along each direction in Ba2LuAlO5 · 0.125 H2O at 1473 K. In panel (b), red, blue, and black curves stand for the MSDs of protons along a, b, and c axes, respectively.

Fig. 8: Probability density distribution of protons in Ba2LuAlO5 · 0.125 H2O.
figure 8

Yellow isosurface of the probability density distribution of protons at 2.0 × 10−4 Å−3, which were obtained by the AIMD simulations at 1473 K. Probability density distributions a without and b with the polyhedra of AlO4 and LuO6.


In conclusion, we have demonstrated high proton conductivity as high as 10−2 S cm−1 at 487 °C and 1.5 × 103 S cm1 at 232 °C and high proton transport number >0.92 in the range 300–800 °C in Ba2LuAlO5. Structural analysis revealed that Ba2LuAlO5 is a hexagonal perovskite-related oxide having oxygen-deficient BaO h′ layers. The large amount of intrinsic oxygen vacancies in Ba2LuAlO5 allows higher water uptake of x = 0.50 in Ba2LuAlO5·xH2O than other typical perovskite and perovskite-related proton conductors, resulting in the high proton conductivity. AIMD simulations for Ba2LuAlO5 · 0.125 H2O have shown that protons migrate mainly near the interface of two LuO6 octahedral layers, in contrast to the case of the hexagonal perovskite-related proton conductor Ba7Nb4MoO20 where protons migrate in the oxygen-deficient c′ layer36,57. By modifying the chemical composition of Ba2LuAlO5, further improvement in conductivity could be expected58. For example, the hexagonal perovskite-related oxide Ba2InAlO5 is expected to show high conductivity as it is isostructural with Ba2LuAlO5. The present guidelines for the material design open new avenues for the development of high-performance proton conductors.


Synthesis and characterization

Ba2LuAlO5 sample was prepared by a high-temperature solid-state reaction method. The starting materials of BaCO3 (99.9% purity), Lu2O3 (99.9% purity), and Al2O3 (99.9% purity) at a molar ratio of Ba:Lu:Al = 2:1:1 were mixed and ground as dried powders and as ethanol slurries for 1 h in an agate mortar, and then calcined in air at 1000 °C for 10 h to remove carbonates. The calcined materials were crushed and ground with the agate mortar, uniaxially pressed into pellets at 62–150 MPa, and then sintered in air at 1600 °C for 10 h. Parts of the sintered pellets were crushed with a tungsten carbide crusher and ground with the agate mortar into powders. Inductively coupled plasma atomic emission spectroscopy (ICP-AES) data of the Ba2LuAlO5 powders indicated that the chemical composition Ba:Lu:Al = 1.92(11):1.07(11):1.02(2) is in good agreement with the ratio of the nominal composition, where the numbers in the parentheses are the standard deviations. The microstructure of the as-prepared Ba2LuAlO5 was observed using a scanning electron microscope (SEM, Keyence VE-8800).

As-prepared pellets (approximately 5 mm in diameter, 2–3 mm in height) of Ba2LuAlO5 were annealed at 400 °C for 24 h under dry O2, dry 5% H2 in N2, and wet air to investigate its phase stability. The as-prepared pellets of Ba2LuAlO5 were also annealed under dry CO2 flow (100 mL min1) at 400 and 500 °C for 24 h to investigate its phase stability. The annealed pellets were crushed and ground into powders, and their XRPD data were measured.

Thermogravimetric (TG) measurements of as-prepared Ba2LuAlO5 were carried out with NETZSCH STA 449 F3 Jupiter. The sample was heated to 800 °C at the heating rate of 10 °C min1 in dry N2 flow (vapor partial pressure P(H2O) < 105 atm) and kept for 4 h at 800 °C in order to remove water, and then the atmosphere was switched to wet N2 flow (P(H2O) = 0.021 atm) and kept for 4 h. In the cooling process, the sample was kept for 4 h at 700, 600, 500, 400, 300, 200, and 100 °C to reach equilibrium after cooling at the cooling rate of 10 °C min1. Raman spectrum of Ba2LuAlO5 was collected using NRS-4100 (JASCO Co.) with an excitation wavelength of 532 nm in static air. The IR spectrum of Ba2LuAlO5 was collected using DR PRO 410MX (JASCO Co.) in dry N2. Ultraviolet-visible (UV-vis) diffuse reflectance spectrum of Ba2LuAlO5 was measured in static air at room temperature between 200 and 700 nm using a JASCO V-670 scanning double-beam spectrometer. The optical direct bandgap Eg was estimated using the Kubelka–Munk equation and a Tauc plot.

Diffraction experiments and structural analysis

A part of the as-prepared Ba2LuAlO5 pellet was crushed, and a single crystal of Ba2LuAlO5 with a size of 5 × 5 × 5 μm was picked up. Single-crystal X-ray diffraction (SCXRD) data of the as-prepared crystal were measured using a Rigaku XtaLAB Pro diffractometer (Mo Kα radiation) at 20 °C. The crystal structure was determined by the charge flipping method using SuperFlip59 followed by least-square refinement with SHELX (ver. 2018/3)60. The bond valence sum at each site was calculated using the bond-valence parameter reported in the literature61,62. Difference Fourier maps (FoFc) were calculated using WinGX (ver. 2018.3)63. The refined crystal structures, probability density distribution of protons, and difference Fourier maps were depicted with VESTA 364. The X-ray powder diffraction (XRPD) data of the as-prepared Ba2LuAlO5 powders were measured at 22 °C by a laboratory-based X-ray diffractometer (Bruker AXS D8 Advance) with Cu Kα radiation at 40 kV and 40 mA (step scanning mode, 0.02° per step, counting time: 20 s per step, and 2θ range: 4–120°). Rietveld refinement was performed using the computer program Z-code65 and crystallographic parameters obtained in the SCXRD analysis as initial parameters.

Dry Ba2LuAlO5 pellets were prepared by heating the as-prepared Ba2LuAlO5 pellets at 800 °C for 30 min in a vacuum quartz tube. The sample in the quartz tube was cooled in vacuum down to 300 °C, and then the quartz tube containing the pellets was sealed at this temperature. Hydrated (deuterated) Ba2LuAlO5 · x D2O pellets were synthesized by heating the as-prepared Ba2LuAlO5 pellets at 800 °C for 30 min in dry N2, cooled down to 200 °C at the cooling rate of 10 °C min1 in D2O/He flow (water vapor pressure P(D2O) = 0.021 atm) and then kept at 200 °C for 2 h in the D2O/He flow. Neutron-diffraction data of the dry Ba2LuAlO5 and hydrated Ba2LuAlO5 · x D2O pellets were measured at 5 K with a fixed-wavelength neutron diffractometer HERMES66 at the JRR-3 research reactor of JAEA, Tokai, Japan (wavelength = 1.34171(5) Å). The collected data were analyzed by the Rietveld method with the program RIETAN-FP67.

Electrical conductivity measurements

The DC electrical conductivity of Ba2LuAlO5 was measured using sintered pellets (approximately 4.5 mm in diameter, 10–12 mm in height, relative density of 73%). Pt paste dissolved in ethanol and Pt wires were attached to both sides of the sintered pellets and heated at 500 °C for 1 h to remove the ethanol in the Pt paste. DC electrical conductivity of the Ba2LuAlO5 pellet was measured between 300 and 800 °C by a DC 4-probe method under N2 flow in dry (P(H2O) < 105 atm) and wet (P(H2O) = 0.021 atm and P(D2O) = 0.021 atm) atmospheres (200 ml min−1). Oxygen partial pressure P(O2) dependencies of the DC electrical conductivity of Ba2LuAlO5 were investigated at 400 °C in the dry and wet (P(H2O) = 0.021 atm) atmospheres. P(O2) was controlled by O2/N2 or H2/N2 gas mixture and monitored by a YSZ oxygen sensor placed at the outlet of the apparatus.

The AC impedance spectra of Ba2LuAlO5 were collected in dry N2 and wet N2 in the temperature range of 50–500 °C. We used two sintered pellets (9.4 mm in diameter, 4.5 mm in thickness, and a relative density of 78%; and 9.9 mm in diameter, 2.6 mm in thickness, and a relative density of 59%) for the AC impedance measurements. The impedance spectra were recorded using a Solartron 1260 impedance analyzer in the frequency range of 10 MHz to 1 Hz with an applied alternating voltage of 1 V. The equivalent circuit analysis was carried out to extract the bulk conductivity σbulk at each temperature using the ZView software (Scribner Associates, Inc.).

Density functional theory calculations and ab initio molecular dynamics simulations

Density functional theory (DFT) simulations of hexagonal Ba2LuAlO5 (1×1×1 cell) were carried out using the Vienna Ab initio Simulation Package (VASP)68. We used projector augmented-wave (PAW) potentials for Ba, Lu, Al, and O atoms; the plane-wave basis sets with a cutoff of 500 eV, and the Perdew–Burke–Ernzerhof (PBE) GGA functionals. A 5×5×2 k-point mesh was used in the Monkhorst–Pack scheme. In self-consistent cycles, the total energy was minimized until the energy convergence was less than 107 eV. Lattice parameters and atomic coordinates of Ba2LuAlO5 were optimized for 27 models with different atomic configurations in the space groups Pm and P21/m with a convergence condition of 0.001 eV Å1. The crystal parameters refined using the SCXRD data were used as initial parameters in the DFT structure optimizations.

AIMD simulations of (Ba2LuAlO5 · 0.125 H2O)16 (= Ba32Lu16Al16O80 · 2 H2O) were performed at 873, 1073, 1273, and 1473 K by VASP using a 2 × 2 × 1 supercell of Ba2LuAlO5 to investigate the proton migration. The initial structural model was made by locating two water molecules in two different h′ layers of the supercell. The geometry-optimized structure was heated from 0 K to the target temperature at a rate of 1 K fs1. The system was further equilibrated for 5 ps, and the production trajectory was accumulated for the canonical (NVT) ensemble using a Nosé thermostat for ~200 ps with the time step of 1 fs. The cutoff energy was set to 300 eV, and the reciprocal space integration was performed only at the Γ-point.