(Bi,Sr) (Fe1−x,Mx)O3−δ (M = Co, Ni and Mn) Cathode Materials with Mixed Electro-Ionic Conductivity

(Bi,Sr)FeO3−δ (BSF) cathode materials doped with either Co, Ni or Mn are synthesized by an ethylene diamine tetra-acetic acid (EDTA)-citrate complexing method, and the effects of the doping level on the mixed electronic-ionic conductivity at various temperatures are studied up to 800 °C. The phase purity and solid solution limit are investigated by X-ray diffraction (XRD). The ionic conductivity is measured by the four-probe direct current (DC) method, the valence state of Fe and Mn by X-ray photoelectron spectroscopy (XPS), and the oxygen non-stoichiometry by differential thermo-gravimetric analysis (TGA). The doped ferrites show interesting electronic conductivity dependent on the testing temperature, implying two conductive mechanisms, either controlled by double exchange at lower temperatures or small polaron (electron-oxygen vacancy) conduction at temperatures greater than 400 °C. The results of Co-doped BSF (S50C20) show the best mixed conductivity among the ferrites, and this is used to assemble cells. The cell with a S50C20 cathode in the region of 600–800 °C is improved by 15% in maximum power density greater than the cell with La0.6Sr0.4Co0.2Fe0.8O3−δ (LSCF) due to the balanced contribution from oxygen ions, vacancies and electrons.


Introduction
Solid oxide fuel cells (SOFCs) are power generation devices that directly convert fuels into electricity via electrochemical reactions through clean and highly efficient operations. To improve the reliability and stability of the cells, lowering the operating temperature to an intermediate-temperature range (below 800 • C) [1,2] is one of the main targets in the recent research efforts. These advantages of lower-temperature operation, including the ease of material selection, reduction in total cost and accelerating the start-up process, are generally recognized. At present, the operation of SOFCs at 650 • C to 800 • C is currently feasible for a SOFC using ultrathin yttria-doped zirconia [3]. However, a lower operating temperature results in sluggish oxygen reduction reaction (ORR) at the cathode and apparently increases the resistance of the electrolyte/cathode, potentially degrading the performance of SOFCs [4,5].
High performance cathodes based on La-perovskite (LaMO 3 ) materials have been studied, and it was recently reported that cobalt-containing (La,Sr)CoO 3 perovskites exhibit the best activity of electro-catalytic ORR at intermediate temperatures (600 • C-800 • C). Since cobalt is a good catalyst for reduction oxygen molecules to ions and flexible redox behavior via valence transition [4], the cobalt containing perovskites, e.g., La 0.6 Sr 0.4 Fe 0.8 Co 0.2 O 3−δ (LSFC6482), have a good electronic conductivity, However, they have very low ionic conductivity and the high coefficient of thermal expansion (CTE) of the LSFC cathode mismatches to that of zirconia-based electrolyte, which is also unfavorable to thermal cycling of the cells [6][7][8][9][10]. Besides, an interfacial reaction might occur in various (a symmetry of Pm-3m), except the S10 sample, which shows a (110) c peak with a very minor (104) r intensity belonging to a rhombohedral structure in a symmetry of R3C. No other secondary phases are found over the composition range. The evidence reveals the substitution of Bi 3+ by Sr 2+ and Fe 3+ by Co 2+ are complete, resulting in the formation of an oxygen vacancy and multiple valence states at the Fe-site. These will be discussed in Section 3. Table 1. Abbreviations and properties of Bi 1−x Sr x FeO 3 (x = 0.1 to 0.5) and Bi 0.5 Sr 0.5 Co y Fe 1−y O 3 (y = 0 to 0.2). intensity belonging to a rhombohedral structure in a symmetry of R3C. No other secondary phases are found over the composition range. The evidence reveals the substitution of Bi 3+ by Sr 2+ and Fe 3+ by Co 2+ are complete, resulting in the formation of an oxygen vacancy and multiple valence states at the Fe-site. These will be discussed in Section 3.  The ferrite samples were densified to a relative density better than 95.0% T.D. (Table 1), and then put to conductivity measurement. The total conductivities of Bi1−xSrxFeO3 and Bi0.5Sr0.5CoyFe1−yO3 are shown in Figure 2. The data depict that the electrical conductivity increases with increasing temperature. The S50 sample has the best electrical conductivity among the Sr-doped bismuth ferrites, reaching a maximum value of 1.5 S·cm −1 at 800 °C. In the S50Cy ferrites co-doped with cobalt, S50C20 shows the best electrical conductivity of 5.4 S·cm −1 at 800 °C.

SrO-Doping
The temperature dependence of the electric conductivity of the ferrites is noted changing at about 500 °C-527 °C, as shown by the dot curve in Figure 2a. The slopes of the best fitting lines are calculated to reveal the activation energy of the conductivity. The energies and transition temperatures of all SxCy are shown in Table 2. There are no such transitions in the S50C10 and S50C20 samples. The Ea(LT) of Sx samples is in the range of 0.4 eV to 0.7 eV, but Ea(HT) is in the range of 0.21 eV to 0.48 eV. The SrO-doped BSFs possibly act as a double exchange mechanism at a lower temperature range, and change to a vacancy control mechanism as the temperature goes higher than Ttr. For Co-doped Note the (110) c peak of S10 sample after de-convoluted to two. The smaller one may belong to (104) r of a rhombohedral structure.
The ferrite samples were densified to a relative density better than 95.0% T.D. (Table 1), and then put to conductivity measurement. The total conductivities of Bi 1−x Sr x FeO 3 and Bi 0.5 Sr 0.5 Co y Fe 1−y O 3 are shown in Figure 2. The data depict that the electrical conductivity increases with increasing temperature. The S50 sample has the best electrical conductivity among the Sr-doped bismuth ferrites, reaching a maximum value of 1.5 S·cm −1 at 800 • C. In the S50Cy ferrites co-doped with cobalt, S50C20 shows the best electrical conductivity of 5.4 S·cm −1 at 800 • C.
The temperature dependence of the electric conductivity of the ferrites is noted changing at about 500 • C-527 • C, as shown by the dot curve in Figure 2a. The slopes of the best fitting lines are calculated to reveal the activation energy of the conductivity. The energies and transition temperatures of all SxCy are shown in Table 2. There are no such transitions in the S50C10 and S50C20 samples. The E a (LT) of Sx samples is in the range of 0.4 eV to 0.7 eV, but E a (HT) is in the range of 0.21 eV to 0.48 eV. The SrO-doped BSFs possibly act as a double exchange mechanism at a lower temperature range, and change to a vacancy control mechanism as the temperature goes higher than T tr . For Co-doped cases, the electronic conduction in S50C10 and S50C20 still retains the double exchange mechanism up to 800 • C. Later, XPS/TGA results will be provided to discuss the possible conductive mechanisms.
Materials 2016, 9,922 4 of 17 cases, the electronic conduction in S50C10 and S50C20 still retains the double exchange mechanism up to 800 °C. Later, XPS/TGA results will be provided to discuss the possible conductive mechanisms.
(a) (b)   Figure 2b shows the temperature dependence of the ionic conductivity of all SxCy ferrites. The ionic conductivity increases with testing temperature, and displays a single slope through the testing temperature range. The correspondent activation energies show a similar value of 1.24 ± 0.06 eV for all SxCy ferrites, lower than that of 1.32-1.34 eV for LSCF6428 [11]. In the series of Bi1−xSrxFeO3, the best ionic conductivity is 0.032 S·cm −1 among all samples at 800 °C which is apparently higher than that (2.2 × 10 −3 -4 × 10 −3 S·cm −1 ) of LSCF82xx at the same temperature [11].
The ionic conductivity and the CTE of SxCy shown in Table 1 and Figure 2b are plotted and compared with that of LSCF in Figure 3. The σi values of the Bi-based perovskites are in general greater than that of the La-perovskites. Besides, the Bi-ferrites show two-stage increment of ionic conductivity. The first is increased by Sr-doping, and the second is by additional Co-doping. Based on the CTE match to that of yttria-doped ZO2, three S30My series (M = Co, Ni and Mn) possibly showing a CTE of ca. 12 × 10 −6 K −1 were selected for further investigation.   Figure 2b shows the temperature dependence of the ionic conductivity of all SxCy ferrites. The ionic conductivity increases with testing temperature, and displays a single slope through the testing temperature range. The correspondent activation energies show a similar value of 1.24 ± 0.06 eV for all SxCy ferrites, lower than that of 1.32-1.34 eV for LSCF6428 [11]. In the series of Bi 1−x Sr x FeO 3 , the best ionic conductivity is 0.032 S·cm −1 among all samples at 800 • C which is apparently higher than that (2.2 × 10 −3 -4 × 10 −3 S·cm −1 ) of LSCF82xx at the same temperature [11].
The ionic conductivity and the CTE of SxCy shown in Table 1 and Figure 2b are plotted and compared with that of LSCF in Figure 3. The σ i values of the Bi-based perovskites are in general greater than that of the La-perovskites. Besides, the Bi-ferrites show two-stage increment of ionic conductivity. The first is increased by Sr-doping, and the second is by additional Co-doping. Based on the CTE match to that of yttria-doped ZO 2 , three S30My series (M = Co, Ni and Mn) possibly showing a CTE of ca. 12 × 10 −6 K −1 were selected for further investigation. . Ionic conductivity at 800 °C of (Bi,Sr)FeO3 and (Bi,Sr)(CoFe)O3 plotted against CTE (room to 800 °C) and the data (dot line) of La-perovskites [11]. Note the numbers following LSCF are the proportion of the La, Sr, Co, Fe elements.

S30 with Co, Ni or Mn Doping
The solubility of Co, Ni and Mn performs differently in Bi0.7Sr0.3FeO3 (S30). A previous report by Chen [27] showed that only 4% of Co can form a solid solution in S30Cy ferrite as the Sr-content was reduced to 30% (i.e., S30C4). The same testing procedure of Ni and Mn was conducted and determined by XRD, as shown in Figure 4. After calcining at 900 °C, a secondary phase (the impurity phase) Bi2Fe4O9 was found in Ni-doped BSFs, i.e., S30N4 and S30N5, indicating that a 4 at % doping level of Ni was over the solubility limit of the S30Ny. Therefore, the maximum Ni soluble content in BSF was 3 at %. Similar to previous procedure, Mn was also doping in BSF, 50 at % Mn-doped S30 still appeared in a single perovskite phase (Figure 4b). The result is similar to that reported by Baek et al. [20].
The electrical conductivity of the S30My is shown in Figure 5. The samples were sintered at 1000 °C for 4 h, and achieved 92.5%, 91.6% and 95.1% relative density (R.D.) for S30, S30N3 and S30M50, respectively. The S30 data in the Figure appear to be well matched to that of the same composition reported by Niu et al. [15]. The electrical conductivity of their (Bi0.7Sr0.3)FeO3−δ at 800 °C was 1.2 S·cm −1 , which was almost the same as our data.  . Ionic conductivity at 800 • C of (Bi,Sr)FeO 3 and (Bi,Sr)(CoFe)O 3 plotted against CTE (room to 800 • C) and the data (dot line) of La-perovskites [11]. Note the numbers following LSCF are the proportion of the La, Sr, Co, Fe elements.

S30 with Co, Ni or Mn Doping
The solubility of Co, Ni and Mn performs differently in Bi 0.7 Sr 0.3 FeO 3 (S30). A previous report by Chen [27] showed that only 4% of Co can form a solid solution in S30Cy ferrite as the Sr-content was reduced to 30% (i.e., S30C4). The same testing procedure of Ni and Mn was conducted and determined by XRD, as shown in Figure 4. After calcining at 900 • C, a secondary phase (the impurity phase) Bi 2 Fe 4 O 9 was found in Ni-doped BSFs, i.e., S30N4 and S30N5, indicating that a 4 at % doping level of Ni was over the solubility limit of the S30Ny. Therefore, the maximum Ni soluble content in BSF was 3 at %. Similar to previous procedure, Mn was also doping in BSF, 50 at % Mn-doped S30 still appeared in a single perovskite phase (Figure 4b). The result is similar to that reported by Baek et al. [20].
The electrical conductivity of the S30My is shown in Figure 5. The samples were sintered at 1000 • C for 4 h, and achieved 92.5%, 91.6% and 95.1% relative density (R.D.) for S30, S30N3 and S30M50, respectively. The S30 data in the Figure appear to be well matched to that of the same composition reported by Niu et al. [15]. The electrical conductivity of their (Bi 0.7 Sr 0.3 )FeO 3−δ at 800 • C was 1.2 S·cm −1 , which was almost the same as our data. . Ionic conductivity at 800 °C of (Bi,Sr)FeO3 and (Bi,Sr)(CoFe)O3 plotted against CTE (room to 800 °C) and the data (dot line) of La-perovskites [11]. Note the numbers following LSCF are the proportion of the La, Sr, Co, Fe elements.

S30 with Co, Ni or Mn Doping
The solubility of Co, Ni and Mn performs differently in Bi0.7Sr0.3FeO3 (S30). A previous report by Chen [27] showed that only 4% of Co can form a solid solution in S30Cy ferrite as the Sr-content was reduced to 30% (i.e., S30C4). The same testing procedure of Ni and Mn was conducted and determined by XRD, as shown in Figure 4. After calcining at 900 °C, a secondary phase (the impurity phase) Bi2Fe4O9 was found in Ni-doped BSFs, i.e., S30N4 and S30N5, indicating that a 4 at % doping level of Ni was over the solubility limit of the S30Ny. Therefore, the maximum Ni soluble content in BSF was 3 at %. Similar to previous procedure, Mn was also doping in BSF, 50 at % Mn-doped S30 still appeared in a single perovskite phase ( Figure 4b). The result is similar to that reported by Baek et al. [20].
The electrical conductivity of the S30My is shown in Figure 5. The samples were sintered at 1000 °C for 4 h, and achieved 92.5%, 91.6% and 95.1% relative density (R.D.) for S30, S30N3 and S30M50, respectively. The S30 data in the Figure appear to be well matched to that of the same composition reported by Niu et al. [15]. The electrical conductivity of their (Bi0.7Sr0.3)FeO3−δ at 800 °C was 1.2 S·cm −1 , which was almost the same as our data.   All samples in Figure 5 showed a single perovskite phase. The conductivity of three doped S30 series displayed similar conductive behavior. The temperature dependence of the electric conductivity of all ferrites appeared two activation energy levels (Ea) either at low or high temperature separated by a transition temperature (Ttr), as summarized in Table 3. The samples had approximately two Ea(LT) values in the lower temperature region, either 0.55 ± 0.04 eV or 0.40 ± 0.03 eV ( Table 3). The conduction behavior was possibly controlled by a double-exchange mechanism (M n+ -O 2− M (n+1)+ ) along the Fe(M)O6 octahedral chains in the perovskite structure.
As the temperature increased to be greater than Ttr, the Ea(HT) either decreased to 0.25-0.36 eV or increased to 1.09 eV·mol −1 . The Ea(HT) of three samples, S30, S30N3 and S30C4, at lower range possibly indicated that the oxygen site of the double-exchange mechanism replaced by the oxygen vacancy (M-VO-M), therefore, the other mechanism dominated. A small polaron ( conductivity through the exchange of electrons by the neighboring oxygen vacancies. On the other hand, the apparent increase in the conductivity of S30M10 and S30M50 at higher temperature was noted, showing the highest Ea(HT) among the Bi-ferrites. This might be caused by the multiple valence states of Mn and/or by the introduction of more oxygen vacancies than BSF and BSFN03. The evidence from the XPA and TGA results in the following Sections 2.3 and 2.4 being presented to support the interpretations. Table 3. Activation energies at LT and HT, and the temperature of turning-point of doped S30.

Valence State Analysis
To investigate the effects of different Ar + sputtering periods (0-180 s) on the valence states of Fe in S30 were first investigated and the spectra are shown in Figure A1. The binding energy of Fe on the B-site of perovskite and the double perovskite structure have been reported and summarized in previous references [28][29][30][31][32]. Eb of Fe 3+ 2p3/2 was around 709.5 eV-710.2 eV, while the Eb of Fe 4+ 2p3/2 was from 711.2 eV to 712.8 eV. The peak in Figure A1 appearing at 709 eV-713 eV contained more All samples in Figure 5 showed a single perovskite phase. The conductivity of three doped S30 series displayed similar conductive behavior. The temperature dependence of the electric conductivity of all ferrites appeared two activation energy levels (E a ) either at low or high temperature separated by a transition temperature (T tr ), as summarized in Table 3. The samples had approximately two E a (LT) values in the lower temperature region, either 0.55 ± 0.04 eV or 0.40 ± 0.03 eV ( Table 3). The conduction behavior was possibly controlled by a double-exchange mechanism (M n+ -O 2− M (n+1)+ ) along the Fe(M)O 6 octahedral chains in the perovskite structure.
As the temperature increased to be greater than T tr , the E a (HT) either decreased to 0.25-0.36 eV or increased to 1.09 eV·mol −1 . The E a (HT) of three samples, S30, S30N3 and S30C4, at lower range possibly indicated that the oxygen site of the double-exchange mechanism replaced by the oxygen vacancy (M-V O -M), therefore, the other mechanism dominated. A small polaron (e − : V 2+ O ) offered conductivity through the exchange of electrons by the neighboring oxygen vacancies. On the other hand, the apparent increase in the conductivity of S30M10 and S30M50 at higher temperature was noted, showing the highest E a (HT) among the Bi-ferrites. This might be caused by the multiple valence states of Mn and/or by the introduction of more oxygen vacancies than BSF and BSFN03. The evidence from the XPA and TGA results in the following Sections 2.3 and 2.4 being presented to support the interpretations.

Valence State Analysis
To investigate the effects of different Ar + sputtering periods (0-180 s) on the valence states of Fe in S30 were first investigated and the spectra are shown in Figure A1. The binding energy of Fe on the B-site of perovskite and the double perovskite structure have been reported and summarized in previous references [28][29][30][31][32]. E b of Fe 3+ 2p 3/2 was around 709.5 eV-710.2 eV, while the E b of Fe 4+ 2p 3/2 was from 711.2 eV to 712.8 eV. The peak in Figure A1 appearing at 709 eV-713 eV contained more than one emission, indicating the possible co-existence of Fe 3+ and Fe 4+ . No Fe 2+ signals in the range of 708.8-710.8 eV were indexed [33].
Note that the Fe 2p 3/2 peak was significantly shifting to a lower binding energy after sputtering for 180 s. According to the report [34], both Fe 3 O 4 and Fe x O samples of newly made and stabilized in air were covered with a thin layer of Fe 2 O 3 , The Ar + bombardment could remove the covered oxide layer in 5 s and reveal the original states within the samples. Similarly, when the samples were exposed to air, the proportion of oxygen states changed, as shown in Figure A2. According to the electric neutrality, the valence state of Fe would increase. Thus, a part of Fe 3+ on BSF surface would oxidize to Fe 4+ . After slightly sputtering for 5 s at an acceleration voltage of 3 keV, the oxidized layer on BSF surface was removed.
In this study, the samples showed similar Fe 2p 3/2 spectra after etching for 5 s, 20 s and even for as long as 80 s. For a longer period than or equal to 180 s, the Fe 2p 3/2 peak shifted to a lower binding energy and the spectra indicated a reduction of Fe 4+ to Fe 3+ . Thus, it is concluded that a limited sputtering (3 keV for 5 s) can remove the surface oxide layer on the BSF surface. Therefore, all samples in the following XPS study will go through sputtering for 5 s before the analysis. Figure 6a shows the XPS spectra of Fe 2p 3 [33]. Note that the Fe 2p3/2 peak was significantly shifting to a lower binding energy after sputtering for 180 s. According to the report [34], both Fe3O4 and FexO samples of newly made and stabilized in air were covered with a thin layer of Fe2O3, The Ar + bombardment could remove the covered oxide layer in 5 s and reveal the original states within the samples. Similarly, when the samples were exposed to air, the proportion of oxygen states changed, as shown in Figure A2. According to the electric neutrality, the valence state of Fe would increase. Thus, a part of Fe 3+ on BSF surface would oxidize to Fe 4+ . After slightly sputtering for 5 s at an acceleration voltage of 3 keV, the oxidized layer on BSF surface was removed.
In this study, the samples showed similar Fe 2p3/2 spectra after etching for 5 s, 20 s and even for as long as 80 s. For a longer period than or equal to 180 s, the Fe 2p3/2 peak shifted to a lower binding energy and the spectra indicated a reduction of Fe 4+ to Fe 3+ . Thus, it is concluded that a limited sputtering (3 keV for 5 s) can remove the surface oxide layer on the BSF surface. Therefore, all samples in the following XPS study will go through sputtering for 5 s before the analysis. The XPS spectra of Mn 2p of the polished BSFMn0.5 surface following Ar + etching for 5 s are shown in Figure 6b. Mn 2+ , Mn 3+ and Mn 4+ were indexed in the spectrum of BSFMn0.5. The average valence state of Mn ions in BSFMn0.5 is 2.65. These results are similar to the work reported by Chuang et al. [35] and the presence of Mn 2+ increased the oxygen vacancy concentration. As the valence state of Bi and Sr in A-site are constant, i.e., trivalent and divalent, respectively, the oxygen nonstoichiometry is mainly controlled by the valence state of the cations in the B-site. Thus, the oxygen stoichiometry of the S30, S30M50 and S30C4 equilibrium in the atmosphere and at room temperature are 2.95, 2.88, and 2.97, respectively. The S30M50 shows the greatest non-stoichiometry (Δδ = 0.12) of oxygen among the samples at room temperature.
The oxygen non-stoichiometry (Δδ) of the sample while undergoing the thermo-gravimetric analysis (TGA) test to high temperature was calculated from the mass loss, as shown below: The XPS spectra of Mn 2p of the polished BSFMn0.5 surface following Ar + etching for 5 s are shown in Figure 6b. Mn 2+ , Mn 3+ and Mn 4+ were indexed in the spectrum of BSFMn0.5. The average valence state of Mn ions in BSFMn0.5 is 2.65. These results are similar to the work reported by Chuang et al. [35] and the presence of Mn 2+ increased the oxygen vacancy concentration. As the valence state of Bi and Sr in A-site are constant, i.e., trivalent and divalent, respectively, the oxygen non-stoichiometry is mainly controlled by the valence state of the cations in the B-site. Thus, the oxygen stoichiometry of the S30, S30M50 and S30C4 equilibrium in the atmosphere and at room temperature are 2.95, 2.88, and 2.97, respectively. The S30M50 shows the greatest non-stoichiometry (∆δ = 0.12) of oxygen among the samples at room temperature.
The oxygen non-stoichiometry (∆δ) of the sample while undergoing the thermo-gravimetric analysis (TGA) test to high temperature was calculated from the mass loss, as shown below: where M s is the molar mass of the stoichiometric oxide (g·mol −1 ) of BSFMx, m s is the sample mass of the stoichiometric oxide of BSFMx, and M O is the molar mass of oxygen (g·mol −1 ), and ∆m is the measured mass loss in the sample. We assume that the sample will not lose any of the cations (not vaporizing), but the oxygen may release and create a new oxygen vacancy when it is heated to 800 • C. The TGA curves of S30, S30C4 and S30M50 are shown in Figure 7. The oxygen losses from room temperature to 800 • C are 0.13%, 0.13% and 0.20%, respectively, and their correspondent ∆δ from room temperature to 800 • C are 0.022, 0.022 and 0.034. Besides, the loss (as the details shown in Figure A3) was nearly complete when the temperature was greater than 400 • C. S30M50 produced a greater oxygen vacancy than S30 and S30C4 at high temperatures (400-800 • C). Table 4 summarizes the TGA results of S30, S30C4, S30M50 and LSCF6428. Clearly, LSCF6428 experienced much greater mass loss, implying the LSCF6428 generated greater oxygen vacancy. The ionic conductivity of LSCF6428 (ca. 1.0-1 × 10 −2 S·cm −1 at 800 • C) [26,36] is higher than that (4 × 10 −3 S·cm −1 at 800 • C) of the S30. Therefore, S50C20 was selected as the cathode to assemble SOFCs, and compared to the results from LSCF6428. where Ms is the molar mass of the stoichiometric oxide (g·mol −1 ) of BSFMx, ms is the sample mass of the stoichiometric oxide of BSFMx, and MO is the molar mass of oxygen (g·mol −1 ), and Δm is the measured mass loss in the sample. We assume that the sample will not lose any of the cations (not vaporizing), but the oxygen may release and create a new oxygen vacancy when it is heated to 800 °C. The TGA curves of S30, S30C4 and S30M50 are shown in Figure 7. The oxygen losses from room temperature to 800 °C are 0.13%, 0.13% and 0.20%, respectively, and their correspondent Δδ from room temperature to 800 °C are 0.022, 0.022 and 0.034. Besides, the loss (as the details shown in Figure  A3) was nearly complete when the temperature was greater than 400 °C. S30M50 produced a greater oxygen vacancy than S30 and S30C4 at high temperatures (400-800 °C). Table 4 summarizes the TGA results of S30, S30C4, S30M50 and LSCF6428. Clearly, LSCF6428 experienced much greater mass loss, implying the LSCF6428 generated greater oxygen vacancy. The ionic conductivity of LSCF6428 (ca. 1.0-1 × 10 −2 S·cm −1 at 800 °C) [26,36] is higher than that (4 × 10 −3 S·cm −1 at 800 °C) of the S30. Therefore, S50C20 was selected as the cathode to assemble SOFCs, and compared to the results from LSCF6428.

Performance of Cells
Standard half cells Ni+8YSZ/8YSZ/20SDC with different cathodes were made for comparison. One cell consisted of a porous cathode layer of 7-15 µm, a porous 5 µm SDC layer as a barrier, and a dense 15 µm electrolyte layer closely sintered on a 300 µm anode. The cathode was either made by S50C20, or LSCF6428 for comparison. The interfaces between the layers were intact. No cracks were observed. Figure 8 is the output voltage, current density and power density of the cells. The open circuit voltage (OCV) of the made cells show a value close to 1.1 V, implying the electrolyte is leakage-free. The maximal power density of the cells is 353 mW·cm −2 and 406 mW·cm −2 , respectively, at 800 °C. The power density of the cell with S50C20 cathode is 15% higher than that of the cell consisting of LSCF.

Performance of Cells
Standard half cells Ni+8YSZ/8YSZ/20SDC with different cathodes were made for comparison. One cell consisted of a porous cathode layer of 7-15 µm, a porous 5 µm SDC layer as a barrier, and a dense 15 µm electrolyte layer closely sintered on a 300 µm anode. The cathode was either made by S50C20, or LSCF6428 for comparison. The interfaces between the layers were intact. No cracks were observed. Figure 8 is the output voltage, current density and power density of the cells. The open circuit voltage (OCV) of the made cells show a value close to 1.1 V, implying the electrolyte is leakage-free. The maximal power density of the cells is 353 mW·cm −2 and 406 mW·cm −2 , respectively, at 800 • C. The power density of the cell with S50C20 cathode is 15% higher than that of the cell consisting of LSCF.

Discussion
A substitution of trivalent cation (e.g., Bi 3+ ) on the A-site of perovskite by divalent cation (e.g., Sr 2+ ) induces two possible reactions in the materials. One is the increase in the valence state of B-site cations, and the other is the formation of oxygen vacancy. If the B-site cations are transition metals (M), the change in valence state creates M n+ /M (n+1)+ couples, which act as the hopping sites for electrons via neighboring oxygen, as shown in Equation (2). A three-dimensional MO6 octahedral network [36] in the perovskite structure acts as the double exchange media and induces electronic conduction, causing valence exchange along the MO6 octahedron, as shown in Equation (2) [37]. The electrical conductivity of S30 ( Figure 2a) and XPS results (Figures 5 and 6) depicting the valence states of Fe (4 + and 3 + ) in nearly equal quantity at the B-sites are mostly favored as hopping sites (e.g., Fe 3+ /Fe 4+ and Mn 3+ /Mn 4+ ) and generates electron conduction at low temperatures.
The ionic conductivity of the perovskite/fluorite structure occurs in more complex way at higher temperature via the vacancy mechanism of oxygen ion diffusion. Two factors, the concentration of the oxygen vacancy   O V  and the mobility µ of the oxygen vacancy are effective and affect the oxygen ionic conductivity in the perovskite structure.
However, there are several factors influencing the ionic mobility, basically, to which the ions block transport of the oxygen in the perovskite structure.
Several structural factors have been proposed to explain the relationship of ionic transportation behavior with the oxygen vacancies, etc. These factors can be explained by a schematic diagram in Figure 9. Goldschmidt et al. [38] defined a tolerance factor, S, to reveal the stability of the perovskite structure, as shown below:

Discussion
A substitution of trivalent cation (e.g., Bi 3+ ) on the A-site of perovskite by divalent cation (e.g., Sr 2+ ) induces two possible reactions in the materials. One is the increase in the valence state of B-site cations, and the other is the formation of oxygen vacancy. If the B-site cations are transition metals (M), the change in valence state creates M n+ /M (n+1)+ couples, which act as the hopping sites for electrons via neighboring oxygen, as shown in Equation (2). A three-dimensional MO 6 octahedral network [36] in the perovskite structure acts as the double exchange media and induces electronic conduction, causing valence exchange along the MO 6 octahedron, as shown in Equation (2) [37]. The electrical conductivity of S30 (Figure 2a) and XPS results (Figures 5 and 6) depicting the valence states of Fe (4 + and 3 + ) in nearly equal quantity at the B-sites are mostly favored as hopping sites (e.g., Fe 3+ /Fe 4+ and Mn 3+ /Mn 4+ ) and generates electron conduction at low temperatures.
The ionic conductivity of the perovskite/fluorite structure occurs in more complex way at higher temperature via the vacancy mechanism of oxygen ion diffusion. Two factors, the concentration of the oxygen vacancy V .. O and the mobility µ o of the oxygen vacancy are effective and affect the oxygen ionic conductivity in the perovskite structure.
However, there are several factors influencing the ionic mobility, basically, to which the ions block transport of the oxygen in the perovskite structure.
Several structural factors have been proposed to explain the relationship of ionic transportation behavior with the oxygen vacancies, etc. These factors can be explained by a schematic diagram in Figure 9.

Discussion
A substitution of trivalent cation (e.g., Bi 3+ ) on the A-site of perovskite by divalent cation (e.g., Sr 2+ ) induces two possible reactions in the materials. One is the increase in the valence state of B-site cations, and the other is the formation of oxygen vacancy. If the B-site cations are transition metals (M), the change in valence state creates M n+ /M (n+1)+ couples, which act as the hopping sites for electrons via neighboring oxygen, as shown in Equation (2). A three-dimensional MO6 octahedral network [36] in the perovskite structure acts as the double exchange media and induces electronic conduction, causing valence exchange along the MO6 octahedron, as shown in Equation (2) [37]. The electrical conductivity of S30 (Figure 2a) and XPS results (Figures 5 and 6) depicting the valence states of Fe (4 + and 3 + ) in nearly equal quantity at the B-sites are mostly favored as hopping sites (e.g., Fe 3+ /Fe 4+ and Mn 3+ /Mn 4+ ) and generates electron conduction at low temperatures.
The ionic conductivity of the perovskite/fluorite structure occurs in more complex way at higher temperature via the vacancy mechanism of oxygen ion diffusion. Two factors, the concentration of the oxygen vacancy   O V  and the mobility µ of the oxygen vacancy are effective and affect the oxygen ionic conductivity in the perovskite structure.
However, there are several factors influencing the ionic mobility, basically, to which the ions block transport of the oxygen in the perovskite structure.
Several structural factors have been proposed to explain the relationship of ionic transportation behavior with the oxygen vacancies, etc. These factors can be explained by a schematic diagram in Figure 9. Goldschmidt et al. [38] defined a tolerance factor, S, to reveal the stability of the perovskite structure, as shown below: Goldschmidt et al. [38] defined a tolerance factor, S, to reveal the stability of the perovskite structure, as shown below: where r A and r B are the mean ionic radii for A-and B-site cations, and r O is the radius of an oxygen ion. For a stable perovskite structure, the tolerance factor S is between 0.77 and 1.0. When S = 1.0, a perfect cubic structure is achieved, and the B-O-B bond angle is 180 • ; when S < 1.0, the lattice structure tends to change from cubic to rhombohedral and to orthorhombic. Table 5 shows the ionic radii of various cations of a specified coordination number (CN) referenced from Shannon [39]. Since the ionic radius of Bi 3+ of CN = 12 is not reported in the literature, one extrapolated value was calculated from a linear relationship between the reported ionic radii and CNs, as shown in Figure 10. An ionic size 1.45 Å of Bi 3+ is obtained, which is slightly greater than 1.37 Å calculated from X-ray data in our previous work [27].
If the value 1.45 Å of Bi 3+ (CN = 12) was used to calculate the tolerance factor of LSCF6428, S30, S30N3 and S30M50, the results are shown in Table 6. Since the S values of all samples are less and close to 1.0, the result indicates all BSFM samples have a stable perovskite structure.  Figure 10; c Referenced from Chen [27]. Table 6. Calculated tolerance factor, critical radius and specific free volume of LSCF6428, S30, S30N3, and S30M50 cathode materials.

Composition
where rA and rB are the mean ionic radii for A-and B-site cations, and rO is the radius of an oxygen ion. For a stable perovskite structure, the tolerance factor S is between 0.77 and 1.0. When S = 1.0, a perfect cubic structure is achieved, and the B-O-B bond angle is 180°; when S < 1.0, the lattice structure tends to change from cubic to rhombohedral and to orthorhombic. Table 5 shows the ionic radii of various cations of a specified coordination number (CN) referenced from Shannon [39]. Since the ionic radius of Bi 3+ of CN = 12 is not reported in the literature, one extrapolated value was calculated from a linear relationship between the reported ionic radii and CNs, as shown in Figure 10. An ionic size 1.45 Å of Bi 3+ is obtained, which is slightly greater than 1.37 Å calculated from X-ray data in our previous work [27].
If the value 1.45 Å of Bi 3+ (CN = 12) was used to calculate the tolerance factor of LSCF6428, S30, S30N3 and S30M50, the results are shown in Table 6. Since the S values of all samples are less and close to 1.0, the result indicates all BSFM samples have a stable perovskite structure.  Figure 10; c Referenced from Chen [27].  Figure 10. Ionic radius of Bi 3+ with different coordination number (the solid points were obtained from Shannon [39] while the holly-square point is calculated from a linear relationship). Two data points (Atou [40] and Chen [27]) are plotted as well. Figure 10. Ionic radius of Bi 3+ with different coordination number (the solid points were obtained from Shannon [39] while the holly-square point is calculated from a linear relationship). Two data points (Atou [40] and Chen [27]) are plotted as well.
Other than the tolerance factor, Richter et al. [36] also proposed another parameter, critical radius r cr , to describe the open space of the triangle (Figure 9) allowing oxygen ion diffusion in the perovskite structure, and it could be calculated by the following equation: where a 0 is the pseudo-perovskite lattice parameter, and a 0 can be derived experimentally or as below [41] if the structure is nearly cubic: In other words, the r cr describes the maximum size of the gap for a mobile oxygen ion to pass through the "saddle-point" between the triangle surround by two A-site cations and one B-site cation. The critical radii r cr of four samples are calculated and shown in Table 5. As all critical radii are close to 0.90 Å and far smaller than the oxygen ions radius (1.40 Å in six-fold coordination or 1.36 Å in three-fold coordination in the saddle-point), the oxygen ions need a strong thermal vibration of the lattice to assist the migration [36].
Sammells et al. [42] introduced another parameter, lattice free volume V f , which was defined as equal to the difference in the unit cell volume (V = a 3 0 ) and the total volume of the constituent ions. Moreover, Hayashi et al. [43] modified the V f and proposed the other parameter, the specific free volume (V sf ) to describe the same diffusion phenomena: They used the parameters, V f and V sf , as the bases to compare the free space in the perovskite structure of different compositions. The larger the V sf , the lower the driving force and activation energy of the migration. Based on the calculated results shown in Table 5, the difference between the two structural factors, the tolerance factor and critical radius are not significant, having less than 1% difference. However, a V sf valued of S30 and S30M50 being 10% more than that of LSCF6428 provides more free space for mobile oxygen diffusing in the perovskite structure. Therefore, the structural advantage of S50C20 offering a lower activation energy (E a ) of 1.2 eV and a higher mobility (µ 0 ) of oxygen migration are expected. The power density output of the cell with the S50C20 cathode is better than that with LSCM6328.

Preparation of Fuel Cell
SOFC was prepared by the following procedures. In brief, 8 mol % yttria stabilized zirconia (8YSZ, Tosoh Co., Tokyo, Japan), NiO (F grade, INCO Ltd., Toronto, ON, Canada), and mesophase carbon powder (MCMB, National Chung-Shan Institute of Science & Technology, Taoyuan, Taiwan) in a mass ratio of 45:45:10 were used to prepare three separated slurries in a solid loading of 28 vol %. The slurries contained de-ionized water as a liquid carrier and Darvan C (PMAA-N, Vanderbilt Co. Ltd., Norwalk, CT, USA) as dispersant. After ball-milling for 24 h, the slurries in a specified composition of the anode, were mixed for an additional 1 h, and then dried to powder form. The anode powder was pressed into the pellets of 20 mm in diameter and 1.0 mm in thickness. The pellets were calcined at 1000 • C for 2 h in air before spin-coating of the 8YSZ layer.
The slurry for spin-coating of 8YSZ electrolyte layer was prepared by mixing YSZ powder (TZ-8Y, Tosoh, Tokyo, Japan) with ethyl cellulose (E0266, TCI, Tokyo, Japan) and terpineol (regent grade, ACROS, NJ, USA) in a mass ratio of 30:2:68. The mixture was ball-milled for 24 h until the slurry was stable and homogeneous. Spin-coating of the YSZ slurry on the anode was conducted in three steps: (1) a few drops of the slurry were coated onto a spinning anode pellet in a turning rate increasing from 0 to 6000 rpm; (2) the spinning speed was kept at 6000 rpm for 60 s; and (3) after coating, the film and the substrate were co-calcined at 500 • C for 10 min. In some cases, the coating was repeated up to three times. A thick coating, made by three cycles, for example, was capable of forming a ca. 15 µm thickness 8YSZ electrolyte layer after sintering. Finally, the coated pellets (half cells) were sintered at 1400 • C for 4 h. A layer of cathode with the pastes of LSFC6482 was screen-printed on a half cell. The S50C20 cathode in an area of 1.00 cm diameter, and a thickness of ca. 10 µm was made. After drying and binder-burn-out of the printed cathode layer, the cell was sintered at 1000 • C for 2 h at a heating rate of 5 • C·min −1 .

Characterization
The crystal structure and phases identification of the calcined powder were determined by X-ray diffractometry (XRD, TTRAX 3, Rigaku Ltd., Tokyo, Japan). The total electrical conductivity of the sintered sample in a disk shape and the ionic conductivity of the rectangular samples were measured from 150 • C to 800 • C by a two-probe or four-probe DC technique (Figure 11), respectively [26]. The valence states of Fe and Mn in Bi 0.7 Sr 0.3 Fe 1−x M x O 3−δ were investigated by X-ray photoelectron spectroscopy (XPS) (Thermo Scientific, Theta Probe, Waltham, MA, USA). Spectra were recorded before and after a step of bombardment cleaning using Ar + ion at 3 keV for a selected period (0 s-180 s). Thermo-gravimetric analysis (TGA) was conducted for estimating the oxygen non-stoichiometry of the ferrite samples using a TA Q600 instrument (SDT Q600, TA Waters LLC, New Castle, DE, USA). The samples underwent through careful drying and sintering steps before the XPS and TGA tests.

Preparation of Fuel Cell
SOFC was prepared by the following procedures. In brief, 8 mol % yttria stabilized zirconia (8YSZ, Tosoh Co., Tokyo, Japan), NiO (F grade, INCO Ltd., Toronto, ON, Canada), and mesophase carbon powder (MCMB, National Chung-Shan Institute of Science & Technology, Taoyuan, Taiwan) in a mass ratio of 45:45:10 were used to prepare three separated slurries in a solid loading of 28 vol %. The slurries contained de-ionized water as a liquid carrier and Darvan C (PMAA-N, Vanderbilt Co. Ltd., Norwalk, CT, USA) as dispersant. After ball-milling for 24 h, the slurries in a specified composition of the anode, were mixed for an additional 1 h, and then dried to powder form. The anode powder was pressed into the pellets of 20 mm in diameter and 1.0 mm in thickness. The pellets were calcined at 1000 °C for 2 h in air before spin-coating of the 8YSZ layer.
The slurry for spin-coating of 8YSZ electrolyte layer was prepared by mixing YSZ powder (TZ-8Y, Tosoh, Tokyo, Japan) with ethyl cellulose (E0266, TCI, Tokyo, Japan) and terpineol (regent grade, ACROS, NJ, USA) in a mass ratio of 30:2:68. The mixture was ball-milled for 24 h until the slurry was stable and homogeneous. Spin-coating of the YSZ slurry on the anode was conducted in three steps: (1) a few drops of the slurry were coated onto a spinning anode pellet in a turning rate increasing from 0 to 6000 rpm; (2) the spinning speed was kept at 6000 rpm for 60 s; and (3) after coating, the film and the substrate were co-calcined at 500 °C for 10 min. In some cases, the coating was repeated up to three times. A thick coating, made by three cycles, for example, was capable of forming a ca. 15 µm thickness 8YSZ electrolyte layer after sintering. Finally, the coated pellets (half cells) were sintered at 1400 °C for 4 h. A layer of cathode with the pastes of LSFC6482 was screen-printed on a half cell. The S50C20 cathode in an area of 1.00 cm diameter, and a thickness of ca. 10 µm was made. After drying and binder-burn-out of the printed cathode layer, the cell was sintered at 1000 °C for 2 h at a heating rate of 5 °C·min −1 .

Characterization
The crystal structure and phases identification of the calcined powder were determined by X-ray diffractometry (XRD, TTRAX 3, Rigaku Ltd., Tokyo, Japan). The total electrical conductivity of the sintered sample in a disk shape and the ionic conductivity of the rectangular samples were measured from 150 °C to 800 °C by a two-probe or four-probe DC technique (Figure 11), respectively [26]. The valence states of Fe and Mn in Bi0.7Sr0.3Fe1−xMxO3−δ were investigated by X-ray photoelectron spectroscopy (XPS) (Thermo Scientific, Theta Probe, Waltham, MA, USA). Spectra were recorded before and after a step of bombardment cleaning using Ar + ion at 3 keV for a selected period (0 s-180 s). Thermo-gravimetric analysis (TGA) was conducted for estimating the oxygen non-stoichiometry of the ferrite samples using a TA Q600 instrument (SDT Q600, TA Waters LLC, New Castle, DE, USA). The samples underwent through careful drying and sintering steps before the XPS and TGA tests.

Conclusions
Single-phase doped (Bi,Sr)FeO 3−δ materials were synthesized and displayed interesting mixed electronic and ionic conductivities at temperatures up to 800 • C. The electric conductivity of SxCy (x = 0-30, y = 0-5) shows two-stage temperature dependence, but the others (x = 40-50, y = 10-20) show only one-stage dependence. However, for the ionic conductivity of Sx and SxCy samples, the dependence follows only one E a 1.24 ± 0.06 eV to the entire temperature region 300-800 • C, which is lower than that (1.32-1.34 eV) of LSCF cathodes [11].
The solubility limit of Ni in S30 is 0.03, but there is no limit for Mn up to 0.50 to keep the single perovskite phase. Three samples, doped S30, S30N3 and S30M50, show CTE values ranging between 12.8 × 10 −6 and 13.7 × 10 −6 K −1 , which are close to that of zirconia-or ceria-based electrolyte.
The results of the valence state of Fe by XPS show the fraction of Fe 3+ and Fe 4+ are in a ratio of 0.47 vs. 0.53 in S30M50 at room temperature. Similar XPS analysis on the Mn shows divalent and trivalent Mn ions appear in nearly equal quantity (0.46 vs. 0.44). Both B-site ions in equal quantity are favored for the double-exchange conduction of electrons along the Fe(Mn)O 6 octahedral chains in the perovskite structure at lower temperature. At temperatures greater than 400 • C, the oxygen vacancies in BSF samples increase by more than a few percentages, and result in not only an interruption of the double-exchange mechanism, but also enhancement of a small-polaron conduction by the help in neighboring oxygen vacancies.
One of the structural factors, specific free volume (V sf ) for BSFM materials, offers 10% more free space than LSCF6428 and may result in a better diffusive mobility of oxygen/vacancy in BSF materials at higher temperature.
All doped BSFs show similar two-stage temperature dependence of conductivity: when lower than the transition temperature (T tr ), all samples show similar double-exchange conductive mechanism with an activation energy of 0.54-0.59 eV·mol −1 or 0.37-0.44 eV·mol −1 . At temperatures higher than T tr , the conductive mechanism relies on a few percent concentration of oxygen vacancies, offering the exchange media to electrons to hop between small polarons (e − : V 2+ O ). This allows the cells with Co-doped S50C20 to reach a maximum power density of 406 mW·cm −2 , with 15% being better than the cell with LSCF6428 testing at 800 • C.