Oxide di ff usion in innovative SOFC cathode materials

Oxide diffusion was studied in two innovative SOFC cathode materials, Ba2Co9O14 and Ca3Co4O9+d derivatives. Although oxygen diffusion was confirmed in the promising material Ba2Co9O14, it was not possible to derive accurate transport parameters because of an oxidation process at the sample surface which has still to be clarified. In contrast, oxygen diffusion in the well-known Ca3Co4O9+d thermoelectric material was improved when calcium was partly substituted with strontium, likely due to an increase of the volume of the rock salt layers in which the conduction process takes place. Although the diffusion coefficient remains low, interestingly, fast kinetics towards the oxygen molecule dissociation reaction were shown with surface exchange coefficients higher than those reported for the best cathode materials in the field. They increased with the strontium content; the Sr atoms potentially play a key role in the mechanism of oxygen molecule dissociation at the solid surface.


Introduction
With the possibility of cogeneration of electricity and heat, Solid Oxide Fuel cells are promising.First units were commercialised in Japan in 2009 and demonstrators are running in Europe.Their principle is based on a dense ceramic electrolyte which separates two compartments, one containing hydrogen, and the other containing air.Because of a difference of oxygen chemical potential between the two compartments, oxygen molecules are reduced at the cathode into oxide ions, which migrate through the electrolyte to the anode to react with hydrogen molecule and produce water.Hydrogen is usually produced by methane reforming, however internal reforming of methane is also envisaged.Typical temperatures of use range are in between 700 and 1000°C.The main electrolyte materials which are currently studied are Yttria Stabilised Zirconia (YSZ), Gadolinia Doped Ceria (GDC) and lanthanum gallates (LSGM).At the anode, a Ni-YSZ cermet containing 40% in volume of nickel is usually developed, but it is prone to carbon deposition when used with hydrocarbon fuels.It is also sensitive to sulphur poisoning and suffer from nickel agglomeration and redox stability under prolonged usage.Alternative materials such as the chromo-manganites (La 0.75 Sr 0.25 )Cr 0.5 Mn 0.5 O 3 or titanates such as lanthanum doped SrTiO 3 have been proposed.At the cathode, the strontium substituted lanthanum manganite (La 1-x Sr x MnO 3-δ ) is conventionally used.However, it displays low oxide ion conductivity and composites with YSZ were developed to allow its use at temperature lower than 800°C.To operate at lower temperature, Mixed Ionic Electronic Conductors (MIEC) are usually preferred to allow the oxygen reduction reaction to be sprayed on the whole electrode volume and not only at the electrode/electrolyte interface.As an alternative to LSM, the La 0.6 Sr 0.4 Co 0.2 Fe 0.8 O 3-δ (LSCF) perovskite has been widely studied 1 .When used with a GDC electrolyte and after optimisation of its microstructure, an area specific resistance (ASR) of only 0.13 Ω.cm 2 was obtained at 603°C for this compound 2 , the target for the cathode being 0.15 Ω.cm 2 at 700°C.These high performances are likely due to the presence of Co(III) in the structure which is known to be particularly active for the reduction of oxygen molecules into oxide ions 3 .Indeed, cobaltites are known to display high oxide ion transport parameters.Among these, the La 0.6 Sr 0.4 CoO 3-δ composition is among the best catalyst for the oxygen molecule reduction reaction with a surface exchange coefficient (k*, cm/s) of 1.1×10 -6 cm/s associated to an oxygen self-diffusion tracer (D*, cm 2 /s) coefficient of 2.0×10 -8 cm 2 /s at 680°C under 230 mbar 4 .One drawback of these materials is their high thermal expansion coefficient with a value of 23×10 -6 K -1 for this later composition which is twice the thermal expansion of YSZ (10×10 -6 K -1 ) 5 and GDC (12×10 -6 K -1 ) 6 .However, partial substitution of cobalt with iron in LSCF led to a good compromise with a TEC in the range of 14-15×10 -6 K -1 from 500 to 700°C 7 .More recently, promising performances as SOFC cathodes were shown for two innovative cobaltites, Ba 2 Co 9 O 14 and Ca 3 Co 4 O 9+δ .An ASR of 0.5 Ω.cm 2 was first measured on a symmetric cell made of a 70 wt% Ba 2 Co 9 O 14 -30 wt% GDC composite electrode, deposited on a GDC  and 770 mW/cm 2 on a cell made of a LSGM electrolyte with a Ba 2 Co 9 O 14 -samarium doped ceria composite cathode at 800 and 850°C, respectively 3 .This compound exhibits a three dimension structure based on a [Co 8 O 8 ] blocks sandwiched in between [Ba 2 CoO 6 ] layers 9,10 (Fig. 1a).With cobalt atoms at difference valences, it displays high electronic conductivity with a value of 240 S.cm -1 at 650°C but as other cobaltites, its thermal expansion coefficient is high (22×10 -6 K -1 ) which supposes the use of a composite at least for mechanical compatibility between the electrode and the electrolyte.After optimisation of its composition and thickness, an ASR of only 0.08 Ω.cm 2 at 700°C was recently measured on symmetric cell made of a 50 wt% Ba 2 Co 9 O 14 -50 wt% GDC composite electrode with a GDC electrolyte 11 .In contrast, a thermal expansion coefficient of only 10×10 -6 K -1 was measured for the well-known Ca 3 Co 4 O 9+δ thermoelectric compound 12 whose structure is a misfit composed of CdI 2 -type [CoO 2 ] layers sandwiched in between [Ca 2 CoO 3-δ ] rock salt slabs 13,14 (Fig 1b).With an electronic conductivity of 100 S.cm -1 at 700°C and oxygen vacancies in the rock salt layers, Mixed Ionic Electronic Conductivity was expected for this compound.When tested in symmetrical cell with a GDC electrolyte, ASR of 4.94 Ω.cm 2 for the pure compound and 1 Ω.cm 2 for a 70 wt% Ca 3 Co 4 O 9+δ -30 wt% GDC composite electrode were first measured 15,16 .After optimisation of composition and thickness it was decreased to 0.5 Ω.cm 2 at 700°C for a 50 wt% Ca 3 Co 4 O 9+δ -50 wt% GDC composite electrode with a thickness of 21 µm, and further decrease to 0.35 Ω.cm 2 when calcium was partly substituted with 10% of strontium 17 .These good performances were confirmed by Samson et al. 18 , who reported a polarization resistance of 0.64 Ω.cm 2 at 600°C for a 50 vol% Ca More recently, oxygen diffusion in this material was confirmed by combing 18 O/ 16 O isotope exchange, electrical conductivity relaxation and pulse isotopic exchange techniques 20 .Despite low diffusion with coefficient of 2.7×10 -10 cm 2 .s - at 700°C, this study showed high kinetics toward oxygen dissociation with a surface exchange coefficient as high as 1.6×10 -7 cm.s -1 .
In the present paper, oxygen diffusion in these two innovative cathodes was investigated with an emphasise put on the impact of calcium partial substitution with strontium on the Ca 3 Co 4 O 9+δ oxygen transport properties.2011] was used for refinement of the unit cell parameters 21 .An Accupyc II 1340 Micromeritic pychnometer was used to measure the powder density.The absolute oxygen stoichiometry of the sample was determined by iodometric titration 22 .Thermal analyses were conducted under air, using a TG/DTA Setaram Setsys apparatus.40 mg of the sample powder was introduced into a Pt crucible, and the experiment was carried out at a heating rate of 5°C/min under flowing air (0.4 L/h) from room temperature to 950°C.A blank experiment with an empty crucible was performed to correct the data for effects of buoyancy.Thermogravimetry was also carried out for evaluation of the oxygen stoichiometry of the sample as a function of pO 2 .In this case, a Setaram 92-1750 apparatus was used.The experiment was carried out using 50 mg of powder in a Pt crucible under flowing O 2 /N 2 mix (5 L/h), with three successive oxygen partial pressures (100, 200 and 400 mbar), and temperature ranging from 600 to 750°C (50°C step) for each pressure.For this set-up, the buoyancy was negligible and the data were directly used to derive the thermodynamic factor at each temperature.Dense ceramics, 3 mm thick, were prepared by Spark Plasma Sintering at the IdF Platform in Thiais (France).A 20 mm (diameter) carbon die was used.The samples were sintered within 15 minutes with a 2 minutes dwell at 850°C, under a 50 Mpa (8.8 kN) pressure.To remove carbon residues the pellets were annealed for 12 h at 800°C.A Hitachi S-3400N was used to characterise the samples' surface.For Isotopic Exchange and Electrochemical Conductivity Relaxation, samples surfaces were roughly polished in a first step before polishing down to 1/4 th of micron with successive SiC papers and diamond pastes (Ra~10-20nm) in a second step.

Experimental
In case of calcium cobaltites, to avoid evolution of the surface roughness, samples were annealed for 72 h at 900°C before the second polishing in order to release residual strains in the material.
The 18 O/ 16 O Isotope Exchange Depth Profiling technique was applied to all samples to derive the oxygen self-diffusion tracer (D*, cm 2 /s) and surface exchange coefficients (k*, cm/s).The diffusion time was chosen to get diffusion lengths significantly smaller than the sample dimensions to consider a 1D diffusion model perpendicular to the surface of the dense ceramic into a semi-infinite medium.In these conditions, assuming first order surface exchange reaction and constant 18 O concentration during the exchanges, the solution for a semi-infinite solid to the Fick's 2 nd law of diffusion is given by the following equation 23 : where C bg is the natural 18 O ratio, C g the 18 O ratio in the enriched exchange atmosphere, x the distance from the surface of the specimen and t the time of the isotope exchange.Exchange conditions are reported in Table 1.Experiments were carried out in between 650 and 750°C under a pressure close to 210 mbar with a pre-annealing under normal oxygen (search grade oxygen 99.995% ALPHAGAZ) to reach thermodynamic equilibrium prior to the exchange under labelled oxygen.For both annealings, under the 18 O enriched atmosphere or preannealing, the samples were annealed for the required time, and quenched.To determine the O content in the oxygen gas used for the exchange, a silicon wafer was oxidised overnight at 1000°C under 200 mbar.A ToF-SIMS machine (ION-TOF GmbH, 25 keV Bi + LMIG analysis gun, DSCS sputter gun) was used to measure the 18 O diffusion profiles using the same experimental protocol as described in ref 20.Both depth analysis from the sample surface (depth profile mode) 24 and edge analysis of the cut sample (line scan mode) 25 were performed and profiles merged.Depth of craters, after SIMS depth profiling, were measured with a KLA Tencor Alpha-step IQ profilometer.The same apparatus was used to measure the samples' roughness.Electrical Conductivity Relaxation was also carried out on strontium doped Ca 3 Co 4 O 9+δ .For this purpose, plane-parallel samples (17 mm in length and 1 to 2.4 mm in width) were cut from the strontium substituted dense ceramics using a diamond saw.Electrical conductivity was measured using the 4 points configuration while changing the oxygen concentration of the surrounding atmosphere as shown in ref 20.A gold paste was used to paint gold contacts which were sintered at 750 °C for 1h at a rate of 100 °C/h before experiment.The distance between the two internal electrodes was about 14 mm.Once the sample was at equilibrium at given temperature, the pO 2 was step-wise changed and the conductivity was monitored as a function of time.The variation of pO 2 was performed by an abrupt switch among air and mixtures of O 2 /Ar with oxygen partial pressures of 0.05 atm and 0.01 atm, respectively, at a 5L/h flow rate.In addition, the temperature-dependent electrical conductivity was measured under air with 5L/h flow rate at a very slow speed of 25 °C/h.To characterise the nature of the atoms at the uppermost surface of the sample, Low-Energy Ion Scattering (LEIS) was also applied to a dense ceramic (Ca 1-x Sr x ) 3 Co 4 O 9+δ with x=0.1 composition which had been heated at 800°C in air during 12 hours prior to the analysis.The analysis was performed using a Qtac 100 (ION-TOF GmbH) apparatus.Before analysis, the sample was in situ submitted to an atomic oxygen treatment (2-3 × 10 -4 mbar) to remove possible organic surface contamination.The study was carried out with a 4 He + (3keV) beam on a 1mm² analysed area.

Results and discussion
Characterization of powder and ceramics.Synthesised powders and dense ceramics purity was confirmed by X-ray diffraction.
As shown in Fig. 2, a good pattern matching was obtained for both type of ceramics.Corresponding unit-cell parameters are given in   As shown by Ehora et al. 9 , Ba 2 Co 9 O 14 decomposed into CoO and a cubic BaCoO ~2 form at 1030°C.To check the stability of the calcium cobaltites, TGA and DTA analyses were carried out on the two strontium doped compositions.Their thermograms are compared to the signals measured for the non-doped compound in Fig. 3.They exhibit similar behavior.The first transition observed at 130°C for the pure compound disappears when strontium is introduced in the structure, the second transition at 550°C, corresponding to the beginning of a mass loss is maintained at the same temperature but the intensity of the associated endothermic peak on heating decreases with the increase of strontium content.All compounds decomposed before 950°C, the temperature of decomposition being slightly lowered when strontium content increases.This decomposition process is reversible but it is worth noting that the recombination happens at a lower temperature when substitution rate increases.Because of lack of stability at high temperature, it was not possible to sinter these compounds by conventional sintering but dense ceramics were obtained by spark plasma sintering with relative densities higher than 96% as confirmed by SEM (see Fig. 4).The normalised 18 O/( 16 O+ 18 O) isotope concentration in function of penetration depth are given in Fig. 5 for the three studied temperatures.Both data recorded in depth-profile and line-scan mode are reported.A good junction between the two sets of data was observed.However, whatever the temperature of exchange, a high oxygen content was measured at the surface of the samples in the first microns depth, the thickness of this high oxygen content layer increasing with the temperature.It was therefore not possible to fit these data to the diffusion equation [1].Taken into account only the data far from the surface, the transport parameters given in Table 3 were derived.
Table 3. Tentative Ba2Co9O14 transport parameters derived from normalised 18 O isotope profile far from the sample edges.
T (°C) k* (cm.s -1 ) D* (cm 2 .s - ) 648 1.3×10 -10 2.5×10 -12 700 4.0×10 -10 5.0×10 -12 750 2.5×10 -9 6.0×10 -12 Surprisingly, despite very good electrochemical performances, rather low oxygen diffusion and kinetics toward oxygen molecule dissociation were shown.However, one must be careful with these results.The increase of 18 O at the sample surface is not clearly understood.It is likely due to an oxidation process when cooling the sample but this hypothesis still has to be confirmed.Data obtained for both compositions in the 650-750°C range were fitted to the theoretical model of diffusion.In these conditions, only the diffusion coefficient could be derived with precision, reflecting a D chem /k chem characteristic length much smaller than the size of the sample and suggesting much higher surface exchange kinetics for these materials than chemical diffusion 28 .Corresponding D chem values are given in Table 4 shown in Fig. 8. Dchem values obtained from the air↔0.05atm and air↔0.01atm are close and similarly to what was obtained for the pure material 20 , moreover the chemical diffusion coefficients are independent of whether the relaxation is a reduction or an oxidation process.As shown in Fig. 8, as expected the diffusion process is thermally activated and activation energies of 1.03±0.03eVand 0.98±0.04eVwere measured for the x=0.1 and x=0.2 compositions, respectively, slightly lower than the activation energy displayed by the pure compound (1.2±0.02eV)indicating easier diffusion for the strontium doped compound which also exhibit higher oxygen diffusion coefficients.This journal is © The Royal Society of Chemistry 2012 1.35×10 -7 1.34×10 -7 1.31×10 -7 675 1.92×10 -7 1.91×10 -7 1.85×10 -7 1.95×10 -7 700 2.58×10 -7 2.51×10 -7 2.52×10 -7 2.57×10 -7 725 3.47×10 -7 3.61×10 -7 3.44×10 -7 3.53×10 -7 750 4.68×10 -7 4.95×10 -7 4.73×10 -7 4.76×10 -7 (Ca1-xSrx)3Co4O9+δ x=0.20 650 3.09×10 -7 3.17×10 -7 3.04×10 -7 3.25×10 -7 675 4.62×10 -7 4.54×10 -7 4.57×10 -7 4.58×10 -7 700 5.99×10 -7 6.40×10 -7 5.87×10 -7 6.37×10 -7 725 8.41×10 -7 8.50×10 -7 8.00×10 -7 8.24×10 -7 750 1.04×10 -6 1.14×10 -6 0.96×10 -6 1.08×10 -6  This increase of oxygen diffusion coefficient with substitution ratio was confirmed by 18 O/ 16 O isotope exchange.The normalized 18 O diffusion profile corresponding to the (Ca 1-x Sr x ) 3 Co 4 O 9+δ with x=0.1 composition, exchanged at 649°C during 567 minutes under 211 mbar dry oxygen is shown in Fig. 9. Again a good junction was obtained between data collected in depth profile mode and line scan mode.However in this case, as shown for the parent compound 20 , a small decrease of the 18 O content was noticed at the sample surface, attributed to a possible carbonation or hydroxylation when back to ambient atmosphere.In insert, the same plot in decimal logarithm scale has been drawn to compare experimental and theoretical profiles far in depth.As previously noticed for the pure compound, a « tail » can be seen, suggesting most likely an anisotropic diffusion 20 .Nevertheless, as shown in Fig. 9, a good fit between the calculated and experimental data was obtained.The derived k* and D* parameters are given in Table 5.Their evolution as a function of temperature is given on an Arrhenius plot in Fig. 10.The tracer diffusion can be correlated to chemical diffusion via the thermodynamic factor γ according to the following equation.

Oxygen transport in strontium doped
when σ e >>σ i [2]   where Δln‫‬O ଶ and Δlnܿ are the corresponding variations in the oxygen partial pressure and oxygen content, and σ e and σ i , are the electronic and ionic conductivity, respectively.
For sake of comparison, the evolution of D ୈ * =D chem /γ is also reported in Fig. 10.Values are in good agreement.The slightly higher values observed for D ୈ * compared to the tracer diffusion coefficient values can be explained by a lack of accuracy in the determination of the thermodynamic coefficient because of small oxygen stoichiometry variations (see value of the thermodynamic coefficient in Table 6).An increase of the diffusion coefficient with the strontium content is confirmed and may be explained by the increase of the rock salt layers volume which may facilitate the oxygen mobility within these layers.However, with a value in the range of 10 -10 cm 2 .s - it is still low which justify the addition of gadolinia doped ceria to improve electrochemical performances [15][16][17][18][19] .With surface exchange coefficient of 3.0×10 -6 cm.s -1 and 4.0×10 -6 cm.s -1 at 700°C for both x=0.1 and x=0.2 compositions, these materials exhibit very fast kinetics for the oxygen reduction reaction compared to the best cathode materials in the field since values of only 1.3×10 -7 , 1.1×10 -6 and 1.3×10 -7 were reported for La 2 NiO 4+δ at 700°C 29 , La 0.6 Sr 0.4 CoO 3-δ at 680°C 4 and GdBaCo 2 O 5+δ at 686°C 30 , respectively.To go further in the understanding of oxygen transport in these materials, the x=0.1 composition was characterised by LEIS.The initial spectrum corresponding to an estimated removal of 0.2 monolayer is compared to the one measured after an estimated removal of 3 monolayers in Fig. 11.Despite a bulk contamination with sodium which was also observed at the surface of powders and ceramics of the parent compound 20 , mainly strontium and calcium were initially observed at the uppermost sample surface.No cobalt was initially observed, just a sub-surface signal at low energy side from the surface peak 31 in the background indicating cobalt atoms are beneath the first measured atomic layer.After ionic sputtering, cobalt was evidenced at the outermost surface and an increase of calcium was also noticed.Although, one must be careful with this technique since strontium may also segregate at the sample surface, the same behaviour was observed for other compounds.For SmCoO 3 , Fullarton et al. also showed that samarium was preferentially at the surface with only 5% of cobalt at the outmost surface and a segregation of strontium at the surface was evidenced for strontium doped compounds 32 .Similarly, Viitanen et al. showed the absence of cobalt and iron in the outermost atomic layer of a La 0.6 Sr 0.4 Co 0.2 Fe 0.8 O 3 membrane 33 .We also showed calcium oxide at the uppermost surface of the Ca 3 Co 4 O 9 parent compound.Here, strontium seems to be mainly at the sample surface and its presence may partly explain the increase of kinetics towards the oxygen dissociation reaction when strontium content increases.

Conclusions
Although oxygen diffusion was confirmed in the promising Ba 2 Co 9 O 14 , it was not possible to derive accurate transport parameters because of an oxidation process at the sample surface which has still to be clarified.In contrast, for Ca 3 Co 4 O 9+δ derivatives, it was shown that oxygen diffusion was improved when calcium was partly substituted with strontium likely due to an increase of the volume of the rock salt layers of the structure which may help for oxygen mobility.Although diffusion coefficient remains low, interestingly fast kinetics towards oxygen dissociation reaction were shown with surface exchange coefficient higher than those reported for the best cathode materials in the field.The surface exchange coefficient is increased when the strontium content increases which may be correlated to the presence of a higher amount of strontium at the sample surface which may play a key role in the mechanism of the oxygen molecule dissociation at the surface of the solid.Faraday Discussions
(5.6963(8) Å, 28.924(6) Å)9 and Sun et al.(5.6958(4)Å, 28.909(4) Å)10 , the small difference observed between the powders and the ceramics, on one hand, and these authors, on the other hand, being likely due to small difference in the oxygen content due to different thermal histories.For the refinement of (Ca 1-x Sr x ) 3 Co 4 O 9+δ , the phases were considered as a mixture of two phases, one corresponding to the [CoO 2 ] layers and the second one corresponding to the [Ca 2 CoO 3-δ ] rock salt layers, with common a, c and β cell parameters, constrained to be the same, and different b H and RS , refined independently.The evolution of the unit cell parameters with the strontium content is in good agreement with the results reported by Wang et al.26 , who showed almost no variation of the b H unit cell characterising the [CoO 2 ] layers periodicity and an increase of the b RS unit cell with strontium content in good agreement with the higher radius of Sr 2+ (1.18 Å) compared to Ca 2+ (1.00 Å)27 .

Fig. 4 .
Fig. 4. SEM images of Ba2Co9O14 (a), and (Ca1-xSrx)3Co4O9+δ with x=0.10(b) and 0.20 (c) showing dense ceramics. 18O/ 16 O isotope exchange in Ba 2 Co 9 O 14The normalised18 O/( 16 O+ 18 O) isotope concentration in function of penetration depth are given in Fig.5for the three studied temperatures.Both data recorded in depth-profile and line-scan mode are reported.A good junction between the two sets of data was

Fig. 5 .
Fig. 5. Normalised 18 O diffusion profile measured on Ba2Co9O14 dense ceramics after annealing under labelled oxygen at 650°C, 700°C and 750°C for about 46 hours and corresponding fit (in red) taking into account the data far from the surface.

Ca 3 Co 4 O 9 Fig. 6 .
Fig. 6.Evolution of the electrical conductivity of (Ca1-xSrx)3Co4O9+δ with x=0.10 composition at thermodynamic equilibrium under different oxygen pressures (a) and comparison of composition x=0 (CCO in black), x=0.10 (CSCO10 in red) and x=0.20 (CSCO 20 in blue) electrical conductivities in function of temperature under air and 0.05 atm (b).

Fig. 7 .
Fig. 7. Evolution of the electrical conductivity of (Ca1-xSrx)3Co4O9+δ with x=0.10 composition at 750°C with surrounding atmosphere changes showing the reversibility of oxidation and reduction processes.

Fig. 11 . 4
Fig. 11.4 He+ (3keV) LEIS spectra recorded on a dense ceramics of (Ca1-xSrx)3Co4O9+δ with x=0.1 after annealing at 800°C for 12 hours prior to the analysis.The first spectrum (in black) corresponds to an estimated removal of 0.2 monolayer and the last spectrum (in red) to an expected removal of around 3 monolayers.

Page 1 of 8 Faraday Discussions Faraday Discussions Accepted Manuscript This
journal is © The Royal Society of Chemistry 2012 electrolyte 8 .The good electrochemical performances of this compound were later confirmed by Li et al. who obtained ASR of only 0.133 Ω.cm 2 and 0.068 Ω.cm 2 at 750 and 850 °C and powder density of 450 mW/cm 2 3 Co 4 O 9+δ -50 vol% GDC composite electrode.Using a samarium doped ceria electrolyte, Zou et al. 19 measured a maximum power density of 430 mW.cm -2 at 700°C for a 70 wt% Ca 2.9 Bi 0.1 Co 4 O 9+δ -30 wt% Ce 1.8 Sm 0.2 O 1.95 │ Ce 1.8 Sm 0.2 O 1.95 │Ce 1.8 Sm 0.2 O 1.95 -Ni button cell.

Table 2 .
For Ba 2 Co 9 O 14 , they are in good agreement with the parameters reported by Ehora et al.