Interstitial Oxide Ion Order and Conductivity in La1.64Ca0.36Ga3O7.32 Melilite**

Solid oxide fuel cells (SOFCs) are a major candidate technology for clean energy conversion because of their high efficiency and fuel flexibility.1 The development of intermediate-temperature (500–750 °C) SOFCs requires electrolytes with high oxide ion conductivity (exceeding 10−2 S cm−1 assuming an electrolyte thickness of 15 μm1). This conductivity, in turn, necessitates enhanced understanding of the mechanisms of oxide ion charge carrier creation and mobility at an atomic level. The charge carriers are most commonly oxygen vacancies in fluorites2, 3 and perovskites.3, 4 There are fewer examples of interstitial-oxygen-based conductors such as the apatites5, 6 and La2Mo2O9-based materials,7–9 so information on how these excess anion defects are accommodated and the factors controlling their mobility is important.


Experimental Section
The A 2 B 3 O 7 melilite structure, where A is a large eight-coordinated cation (A = Ca, Sr, Na, Ba, La) and B is small cation (B = Mg, Zn, Al, Ga, Si, Ge, …) in tetrahedral coordination, consists of anionic layers of five-membered rings of totally (four neighbouring tetrahedra) and partially condensed (three neighbouring tetrahedral with a terminal oxygen) BO 4 tetrahedra, separated by parallel sheets of A cations located over the five-ring centres along [001] (Figure S1.1).
After regrinding, the powders were uniaxially pressed into pellets and fired at 1400 °C for samples with x < 0.5 and 1350 °C for samples with x > 0.5 for 12 h at 5 °C/min heating and cooling rates. To compensate the Ga volatilization during the reaction which results in the form of the second phase Cadoped LaGaO 3 , a slight excess of Ga 2 O 3 (~0-1.7 mol% for x ≤ 0.6; ~2-3.3mol% for 0.6 < x ≤ 0.65; ~5mol% for x > 0.65) was added into the initial starting mixture. Attempts to obtain single phase materials at x = 0.65 by adding excess Ga 2 O 3 (e.g. ~ 2 mol% and ~ 3.3 mol%) failed to remove the second phase Ca-doped LaGaO 3 . Slightly reducing the La/Ca ratio to x = 0.64 with ~ 2.3 mol% extra Ga 2 O 3 (i.e. initial composition La 1.64 Ca 0. 36 Ga 3.07 ) produced a nearly single phase material La 1.64 Ca 0.36 Ga 3 O 7.32 with a very small Ga 2 O 3 impurity. To avoid reacting with the alumina, the pellets were put on platinum foil during the preparation. The phase purity was checked by powder X-ray diffraction data with a Panalytical X'pert Pro Multi-Purpose X-ray diffractometer (Co Kα 1 radiation λ = 1.78901Å). Silicon was added as an internal standard during the laboratory XRD experiments in order to refine the variation of the cell parameter with composition. Time-of-flight (TOF) neutron powder diffraction (NPD) data of La 1.64 Ca 0.36 Ga 3 O 7.32 sample were collected from ambient temperature to 800 °C on the HRPD diffractometer at ISIS. After the variable temperature diffraction, the sample was offline cooled to ambient temperature and another ambient temperature TOF data was collected from comparison with that for the as-made sample. Rietveld refinement was carried out using the GSAS package 1 . Compositional analysis was carried out by using the EDAX analyzer on a JEOL 2000FX transmission electron microscope operated at 200 KV. Ac Impedance Spectroscopy (IS) measurements in air from 200 °C to 850 °C were performed with a Solartron 1255B Frequency Response Analyzer, a Solartron 1296 dielectric interface and a Solartron 1287 electrochemical interface over the 10 -2 -10 6 Hz frequency range. Prior to the measurement, platinum paste was painted on the opposite faces of the pellets and fired at 800 °C for 30 minutes in air. 3

O 7+0.5x
The parent material LaCaGa 3 O 7 crystallizes in a tetragonal melilite structure in m P 1 2 4 − (a = 7.9553(2) Å, c = 5.2727(2) Å). With the substitution of Ca 2+ by La 3+ , the tetragonal phase extends to La 1.5 Ca 0.5 Ga 3 O 7.25 . The cell parameters of La 1+x Ca 1-x Ga 3 O 7+0.5x obey Vegard's law for 0 < x < 0.5. The La 3+ substitution expanded the a-axis and shortened the c-axis. The volume increased with increasing La 3+ content. Each (hkl) (h and k ≠ 0) reflection is broadened at x = 0.525 and splits into two distinct peaks at x > 0.525, which could be explained by a lower symmetry with inequivalent a-axis and b-axis, i.e. an orthorhombic cell (a and b ~ 2 a tetra , c = c tetra ) with a sub-group Cmm2 of m P 1 2 4 . The XRD patterns ( Figure S2.1) suggested that the as-made samples x > 0.5 actually contained both tetragonal and orthorhombic phases and the orthorhombic phase dominates in the samples x ≥ 0.55. The cell parameters for orthorhombic and tetragonal La 1+x Ca 1-x Ga 3 O 7+0.5x , shown in Figure S2.2, become constant after x = 0.65 which suggests that the solid solution terminates at x = 0.65. The XRD pattern of the as-made La 1.64 Ca 0.36 Ga 3 O 7.32 showed a pure orthorhombic phase with no apparent residual tetragonal phase in the sample. EDS (energy-dispersive spectroscopy) elemental analysis in the TEM investigation showed a homogenous cation ratio La 1.62 Ca 0.36 Ga 3 for La 1.64 Ca 0.36 Ga 3 O 7.32 against La 1.01 Ca 0.93 Ga 3 from the EDS analysis for the parent material LaCaGa 3 O 7 .

SAED and HRTEM analysis
SAED results showed that, although the main diffraction features were consistent with the crystal structure proposed by XRD (space group Cmm2, a =11.42 Å, b = 11.23 Å and c = 5.25 Å), extra weak reflections were observed in the diffraction patterns along certain crystal axes. These extra spots can be indexed simply by doubling the c axis of the proposed unit cell. The reciprocal lattice reconstruction indicates a c-axis doubled body-centered supercell based on the above Cmm2 cell with no other systematic absences observable, suggesting possible space groups Imm2, Im11, and even in triclinic cell I1. Figure S3.2 shows the simulated SAED patterns for La 1.64 Ca 0.36 Ga 3 O 7.32 3 O 7.32

in average models
High resolution neutron HRPD (TOF) data from both backscattering (NPD168) and 90 degree (NPD90) detectors were used for crystal structure refinement in GSAS. Refinements were carried out in the following sequence ( Figure S5.1), in order to explore the possible space groups suggested by SAED.

Orthorhombic cell, Cmm2 space group
The Rietveld refinement plots of data in the Cmm2 cell (see Section 3 and Figure S5.1) are shown in Figure S5.2 (R wp /R p = 4.81/5.47%, χ 2 = 9.31). As can be seen in Figure S5.2, the fit is poor, especially in the high d-spacing range (~ 2.4 Å). No significant improvement on the fitting was achieved by adding a tetragonal phase in the refinement, indicating that there is no minor quenched tetragonal phase.
Excess weak reflections could be examined as shown in the inset of Figure S5.2a, which requires a larger supercell. Thus, the following refinements were carried out considering the c-axis doubled supercell as discussed above (Section 3 and Figure S5   The followed Rietveld refinement in the c-doubled supercell in Imm2 still gave poor fitting ( Figure   S5.3; R wp /R p = 4.12/5.37%, χ 2 = 6.84). The fit cannot be improved by switching to the Stephen's anisotropic broadening function 2 or by adding the tetragonal phase into the refinement, which suggested the investigation of lower symmetry for La 1.64 Ca 0.36 Ga 3 O 7.32 . The Rietveld refinement in the lower symmetry monoclinic space group Im11 is significantly improved as shown in Figure S5.4 compared with Figure S5.3 and converged to R wp /R p = 3.37/4.96%, and χ 2 = 4.69. Three distinct interstitial sites were revealed and the total population was constrained to be 2.56 for charge balance as listed in Table S5.2 and Figure S5

Triclinic cell with space group of P1
The optimized refinement in triclinic P1 (R wp /R p = 3.21/4.60%, and χ 2 = 4.28, Figure S5.6) exhibited four distinct interstitial sites with total population of 1.28 as listed in Table S5.4. Compared with the crystal structure in monoclinic symmetry Im11 (Section 5.2), all the interstitial oxides are incorporated into GaO 4 tetrahedra with terminal oxygens. Four of the A-sites show cation ordering and are occupied by La only -the occupancy of La was set to be one if it refined to be greater than 0.98.

Simulated annealing (SA) approach
To validate the structure model (P1) from the least-squares Rietveld refinement, simulated annealing (SA) analysis was performed. Peak shape and background parameters were obtained by Le Bail refinement of the NPD data from the 90º and backscattering banks of detectors of the HRPD instrument using TOPAS 3 , with a single main melilite phase. The standard simulated annealing macro implemented within TOPAS was run using a fixed framework model from the GSAS Rietveld refinement, with eight possible interstitial oxygen sites specified. The occupancies of these sites were allowed to vary during the minimization, along with the La and Ca occupancies of the split A sites.
Penalty functions were specified to bias the total interstitial oxygen and La/Ca occupancies towards previously obtained values constrained by the chemical composition. The SA result (R wp = 6.78 %, R p = 7.72 %, χ 2 = 4.32, Figure S5.2) is in good agreement with refined structure ( Figure S5.1 and Table   S5.2) and confirmed the reliability of the proposed structural model regarding the interstitial occupancies and A site cation ordering.  In this split-site model, the relaxations around interstitials were modelled by introducing atomic displacements as shown in Table S7.1-3. Ga3a1, O11a, O11b, and La3a were displaced to Ga3a1 L , O11a L , O11b L , and La3a S around O13b, including O4a in the local defect structure. Similarly, Ga3a2 L , O11c L , O11d L , La/Ca3b L , La/Ca4b L , and O4b are involved in the local defect structure around O13a, Ga3b1 L , O12a L , O12b L , La/Ca3a L , La/Ca4a L , and O3a around O14a, Ga3b2 L , O12c L , O12d L , La/Ca4b S , and O3b around O14b, respectively. This models the cases where an interstitial defect is present and where it is absent.  a Occupancies of mixed La/Ca sited were constrained to be unity, the ADPs were refined to have the same values. b ADPs for related bulk-and defect-structure atoms were constrained to be the same, and the total occupancy was constrained to be unity, e.g. Occ.(La3a) + Occ.(La3a L ) + Occ.(La3a S ) = 1. c Occupancies of O3a/O3b, and O4a/O4b pairs were constrained to be unity, and the ADPs were refined to have the same values, respectively. d The interstitial oxide content was fixed according to the charge-balanced nominal composition of 1.28 in the unit cell. 1,2,3,4 Occupancies were constrained to have the same values as the defect interstitials with which they are associated.

Bulk structure Defect structure Bond
Angle (