Department of Materials, Imperial College London, Prince Consort Road, London SW7 2BP, UK

In-situ neutron diffraction study of cathode/electrolyte interactions under electrical load and elevated temperature

F. Tonus and S. J. Skinner

Department of Materials, Imperial College London, Prince Consort Road, London SW7 2BP, UK

Abstract

Fuel cells are proposed as a future energy conversion technology that will reduce greenhouse gas emissions at the point of operation due to their ability to produce electrical energy from non-hydrocarbon fuel sources. The Solid Oxide Fuel Cell (SOFC) is amongst the most efficient fuel cell types, however, due to the high cell operating temperature cation diffusion occurs between the different components of the cell, resulting in rapid degradation of the power output. In this paper we investigate cation migration between the promising intermediate temperature-SOFC cathode La1-xSrxCo1-yFeyO3- (LSCF) and a fluorite type electrolyte Ce1-xPrxO2- (CPO). The crystallographic structure evolution and degradation of the materials were studied by neutron diffraction in-situ under pseudo-operating conditions, i.e. at 600 ºC under air and under electrical polarisation. The lattice parameter and cation occupancy evolution were analysed by Rietveld refinement as a function of time and applied potential. The materials were found to be stable, as no impurity formation, lattice parameter or site occupancy evolution was observed during the experiment. However La migration prior to the experiment from LSCF to CPO was observed as well as B-site vacancies in LSCF.

Keywords: SOFC, in-situ neutron diffraction, LSCF, CPO, polarisation

Introduction

In order to mitigate the effects of climate change, reduction in carbon-based emissions is necessary and significant research efforts have been directed towards low emission power generation. As such research on fuel cells and particularly on Solid Oxide Fuel Cells (SOFCs), has been extensively developed as the only waste produced is water, assuming a non-hydrocarbon fuel is utilised. However due to high operating temperatures (≈800-1000 ºC), cells are subject to issues such as delamination1,2, cation segregation3,4,5 and interdiffusion of cations6,7, which can cause rapid aging of the cells. In order to decrease these effects, lower operating temperatures (500-700 ºC) are targeted and so new materials with enhanced performance at these temperatures are required. However the phenomenon of cation diffusion through the interface is still not very well understood, particularly during operation. Yang et al.8 showed, for example, the effect of temperature on the degradation of the material: after annealing an yttria stabilised zirconia (YSZ)/ LaSrxMnO3 (LSM) symmetrical cell at 1400 ºC for 48 h they highlighted by scanning electron microscopy (SEM) and X-ray diffraction (XRD) the formation at the interface of a passive phase composed of cations originating from both YSZ and LSM. Harvey et al.9 showed additionally by secondary ion mass spectrometry (SIMS) the influence of temperature on the diffusion of La and Mn from a LSM thin film into Ba1-xSrxCo1-yFeyO3-δ (BSCF). Backhaus-Ricoult et al.10 and Huber et al.11 on the other hand highlighted the influence of the potential between the electrodes on the cation migration. They showed in-situ by X-ray photoelectron spectroscopy (XPS) and Scanning X-ray Photoelectron Microscopy (SPEM) or SIMS at high temperature (600-650 ºC) under vacuum the migration of cations at the surface between either LSM or La1-xSrxCr1-yMnyO3-δ (LSCrM) and YSZ depending on the polarisation. Backhaus-Ricoult et al. showed first the migration of Sr and Mn from LSM to YSZ under cathodic polarisation which was reversible for Mn when applying an anodic bias whilst Huber et al. showed the same results between LSCrM and YSZ.

Whilst these studies provide valuable information on the aging mechanism of SOFCs, they were not performed under realistic operating conditions and concern only high temperature SOFC materials when conceivably the future lies in intermediate temperature SOFC (IT-SOFC) materials, particularly for mass market combined heat and power operations. In this study we analyse the interaction between a promising IT-SOFC cathode (La1 xSrxCo1 yFeyO3 : LSCF) and electrolyte (Ce1-xPrxO2-: CPO) by in-situ neutron diffraction under operating conditions (under air, under polarisation at 600 ºC).

Experimental

Due to the high conductivity of La0.6Sr0.4Co0.2Fe0.8O3- (LSCF6428) leading to resistances beyond the measurement range of the in-situ stage, only the Ce0.9Pr0.1O2- (CPO) and 50:50% LSCF/CPO mixture were studied. CPO was synthesised using a sol gel route described elsewhere12 and LSCF was supplied by Praxair Surface Technologies (Lot #03-P6171DM) (99.9%). One of the most promising IT-SOFC electrolytes is currently Ce0.9Gd0.1O2-δ (CGO), however it was not possible to carry out the experiment on this material due to the high Gd neutron absorption cross-section (11.5 barns for Pr vs. 49700 barns for Gd). As some studies previously showed that CGO and CPO had a similar crystallographic structure13 and close transport properties (σT(CPO) × 10 ≈ σT(CGO))14,15,16,17, the experiment was performed on CPO as an analogue for CGO. Whilst Ce1-xSmxO2- (CSO) would be a better electrical analogue this phase also suffers from high neutron absorption cross-section (5922 barns) and hence the best option structurally was viewed as CPO. Sintering of the samples was performed under the same conditions for each pellet. 13mm diameter pellets were pressed uniaxially at a pressure of 5 tons then isostatically pressed at 300 MPa to ensure the maximum green density was achieved. The samples were then sintered for 1h at 1200 ºC to minimise the diffusion of cations from one phase to the other during the synthesis stage. Platinum paste was deposited on both faces of the pellet and annealed for 30 min at 900 ºC to ensure a good contact with the electrodes. Phase purity of the prepared samples was verified using X-ray diffraction (Bruker D2 Phaser, CuKα, λ = 1.5406Å).

CPO and LSCF/CPO stability under polarisation in an air atmosphere was then studied in-situ by time-of-flight neutron diffraction (Figure 1). The experiment was carried out at the spallation neutron source (ISIS) at the Rutherford Appleton Laboratory, Oxfordshire, UK on the POLARIS diffractometer. The collected data were refined using the Rietveld method with the FullProf Suite18. Both the CPO and LSCF/CPO samples were heated at 600 ºC in a quartz ampoule under flowing air. The samples were kept at this temperature during the experiment, and neutron diffraction data were collected every hour, applying different potentials to the material sequentially. The potential was applied using a Keithley 2400 source meter limited to a maximum current of 3 A. Voltages of +0.5, +1.0 and +1.5 V followed by -0.5, -1.0 and -1.5 V were applied sequentially to the CPO sample while collecting diffraction data, and the material was held under these conditions for between 4 and 6 hours at each voltage. Afterwards the experiment was repeated on the mixture of LSCF/CPO, applying a voltage from +0.5 to +1.5 V and -0.5 to 1.5 V, and finally increasing the potential to -5 V and +5 V for 1 hour at each potential at the end of the experiment to test the material stability in extreme polarisation conditions although being far from normal operating conditions.

Figure 1: Schematic diagram of the experimental set-up used for in-situ combined structural and electrical measurements

Results and discussions

The fit of the refined structural model obtained from neutron diffraction data of LSCF6428 recorded at room temperature is shown in Figure 2 (rhombohedral symmetry, R-3c space group). The refinement of the occupancies of the elements on the A- and B-sites showed that they correspond to the expected formula within the refinement error (Table 1). The oxygen stoichiometry was found to be 3.00 (compared to the 2.98 found by Kuhn et al. in the same atmosphere and at the same temperature19), meaning an average oxidation state of +3.4 on the B-site, close to values obtained in previous studies20.

Figure 2: Rietveld refinement of the structural model of La0.6Sr0.4Co0.2Fe0.8O3 from neutron diffraction data collected at room temperature under a flowing air atmosphere. The background originates from the quartz sample holder. Rwp = 2.55, RBragg = 3.21, Chi2 = 2.25.

Table 1: Calculated cation occupancies in LSCF6428 at room temperature.

La / Sr / Co / Fe
Theoretical occupancy (fraction) / 0.60 / 0.40 / 0.20 / 0.80
Refined occupancy (fraction) / 0.59 (2) / 0.41 (2) / 0.188 (3) / 0.812 (3)

As mentioned in the experimental section, no significant voltage could be applied to the LSCF material due to the limitation of the potentiostat and so data collection focussed on the Ce0.9Pr0.1O2-δ and CPO/LSCF composite materials. For the first refinement of data collected at room temperature, the occupancy between Ce and Pr was refined to Ce0.9Pr0.1O2-δ, however the standard deviation was found to be very large (±0.1) due to the fact that the coherent scattering length of Ce and Pr are very close (0.484×10-12 and 0.45810-12 cm, respectively). Due to this similarity in the scattering length no refinement of the Ce/Pr occupancy in Ce0.9Pr0.1O2-δ was undertaken for all subsequent refinements, with the model using the fixed nominal ratio of 0.9:0.1. The evolution of the cubic lattice parameter (Fm-3m) as a function of temperature and voltage is displayed in Figure 3 and the evolution of the oxygen stoichiometry in Figure 4.

Figure 3: Evolution of the a lattice parameter of Ce0.9Pr0.1O2-δ as a function of time and voltage derived from the Rietveld refinement of neutron diffraction data recorded in-situ at 600 ºC.

Figure 4: Evolution of the oxygen occupancy of Ce0.9Pr0.1O2-δ as a function of time and voltage derived from the Rietveld refinement of neutron diffraction data recorded in-situ at 600 ºC.

In Figure 3, the evolution of the lattice parameter remains constant and stays within the error bar whatever the voltage applied to the material was. Atkinson et al.21 calculated an increase of 0.4% of the lattice for CGO under reducing atmosphere at 700 ºC leading to a loss of 0.028 oxygen ions per formula unit. This would mean an increase of the lattice parameter from 5.434 Å to 5.456 Å, a significant variation from the value achieved at the maximum applied potential, 5.436 Å, if the maximum of the error bar is considered. If we assume that CPO is behaving in the same way as the analogous CGO, this means that no significant change in oxygen stoichiometry occurs under the conditions applied in our experiment. To confirm this reasoning, oxygen occupancy has been refined and results of the refinement show an oxygen stoichiometry of 2.00 all through the experiment (Figure 4), in agreement with Bishop et al. who found a value of 1.994 at the same pO2 and temperature17. This would mean that the oxidation state of Pr is +4, rather than +3.

Resistance of the sample has been measured at different periods during the experiment. An average resistance of R  2000  was observed for CPO through a thickness of 0.63 mm, leading to a conductivity of  = 810-4 S.cm-1. This is slightly lower than that found by Bishop et al. ( = 410 3 S.cm-1)17 in the same conditions, however this could be explained by the fact that in our experiment we probably have fewer oxygen vacancies (Ostoich = 2.00 vs 1.994) although their value remains within our error bar.

The experiment was repeated on the LSCF/CPO mixture to investigate if there were any significant interactions between the two phases under these conditions. Figures 5 and 6 show the lattice parameter evolution of CPO and LSCF respectively. No evolution can be observed as the fluctuation of the values of the lattice parameters stays within the error bar. Hardy et al.22 performed a similar experiment on the same materials but on a different configuration (fuel cell stack instead of powder mixture) using X-ray diffraction on the cathode. They observed an increase in lattice parameters over 60 hours at 0.8 V at a higher temperature (750 ºC) that remains within our error bar, so their observed evolution could not be seen under the experimental conditions used here. In addition to the absence of lattice parameter evolution, no impurity formation could be detected in the diagrams, even when the highest voltages (±5.0 V) were applied to the sample. However we can note a difference in the value of the lattice parameter of CPO when in the mixture or when alone. Indeed we can observe a significant increase of the lattice parameter of CPO at 600 ºC from 5.434(2) Å in the pure phase to 5.445(3) Å in the mixture. This increase has been observed previously in similar materials6,23, and was attributed to the substitution of the Ce4+ (0.97 Å) or Pr4+ (0.96 Å) cations by the larger La3+ (1.16 Å) cation coming from the cathode during the sintering of the material at high temperature.

Figure 7 shows the resistance evolution of the mixture during the experiment. The resistance recorded remained constant, except when a higher voltage (±5.0 V) was applied where the resistance was found to increase by 2% from 3.14 to 3.20 Ω. However this increase was so small that no crystallographic modification was observed that correlated with this change.

Figure 5: Evolution of a lattice parameter of CPO from CPO/LSCF mixture as a function of time and voltage from Rietveld refinement.

Figure 6: Evolution of (a) a lattice parameter and (b) c lattice parameter of LSCF from CPO/LSCF mixture as a function of time and voltage from Rietveld refinement.

Figure 7: Evolution of resistance of CPO/LSCF mixture as a function of time and voltage.

In order to further investigate the structural modifications of the material, cationic occupancies of both the A- and B-sites in the LSCF phase were refined. As the sites were considered fully occupied, they were refined according to the relationship Laoccup + Sroccup = 1 and Cooccup + Feoccup = 1. La and Co occupancy evolution is shown in Figures 8 and 9 respectively. From Figure 8 it is apparent that the La occupancy does not evolve with changes in the applied voltage but is lower than expected (0.28 instead of 0.6) even at room temperature, suggesting that the evolution of the material occurred during its synthesis. On the other hand, this suggests that Sr occupancy is higher than expected (0.72 instead of 0.4) as Laoccup + Sroccup = 1. To understand what is really happening on this site, as for example Sr occupancy increase cannot be rationally explained, as there is no other source of Sr apart from LSCF in the experiment, we look at the average coherent scattering length of the site. La, having a higher coherent scattering length than Sr (0.824 vs 0.702 10-12 cm), would give a theoretical average for the site of 0.775 10-12 cm, assuming a La:Sr ratio of 60:40. If the calculated ratio is considered this results in the average scattering of the site being lower (0.73610-12 cm). This difference in the scattering factor could be explained by the formation of La and/or Sr vacancies induced by the migration of La and/or Sr to the electrolyte. This value would correspond to the formation of 8% La vacancies if only La vacancies are considered, 14% Sr vacancies if only Sr vacancies are considered, or any value in between if both La and Sr vacancies are considered. As La and Sr are randomly distributed (no superlattice reflections are observed) it is not possible to calculate the number of La and Sr vacancies separately. This result is in agreement with the observation from Figure 5 where the increase of the lattice parameters of CPO was explained by a migration of La cations during the sintering of the material, as the vacancies are already present at room temperature before the experiment. We can also notice that Ce and Pr coherent scattering lengths (0.484 and 0.458 10-12 cm respectively) are lower than those of La and Sr. So it is also possible that the decrease of the average of the coherent scattering length would be explained by a substitution of La and/or Sr by Ce and/or Pr. Due to the similarity in these scattering lengths it is not possible to calculate the distribution of cation vacancies and substitutions in the material.

Figure 8: Evolution of La occupancy of LSCF in CPO/LSCF mixture as a function of time and voltage. Room temperature data for the La occupancy is shown in blue.

Figure 9: Evolution of Co occupancy of LSCF in CPO/LSCF mixture as a function of time and voltage. Room temperature data for the Co occupancy is shown in blue.

Figure 10: Evolution of oxygen stoichiometry of LSCF (black) and CPO (blue) in CPO/LSCF mixture as a function of time and voltage recorded at 600oC. Room temperature data for the oxygen stoichiometry are shown in green.

As shown in Figure 9, the occupancy of Co does not change during the experiment when applying different voltages. Although close to the expected value of 0.2, the Co occupancy seems slightly higher than expected with values around 0.22 although some data points’ error bars cover the theoretical value of 0.2. The room temperature value follows this trend, confirming that evolution of the material occurred during its synthesis. Similarly to La occupancy evolution, a decrease of the average scattering length in the B-site (from the expected value of 0.80610-12 cm to the 0.792×10-12 cm observed value) is observed as Co occupancy is increasing. Indeed Co scattering length is lower than that of Fe (0.249 vs 0.945 ×10-12 cm respectively) and thus the decrease of the average scattering length of the B-site implies the formation of vacancies. If we estimate these vacancies, we get 28% vacancies if considering only Co vacancies and 1.8% vacancies if considering only Fe vacancies. These values are very different because Fe scattering length is about 4 times higher than that of Co, so it may be more relevant to keep the value of 2.3% vacancies in the B-site if we consider equal amount of Co and Fe vacancies.

Oxygen occupancy in LSCF and CPO has also been refined, and the oxygen stoichiometry evolution is displayed in Figure 10. It shows no significant evolution during the experiment for both LSCF and CPO, however, if the oxygen stoichiometry for LSCF is assumed to be 3.00 during the whole experiment, the oxygen stoichiometry of CPO is significantly lower than the value of 2.00 found for the pure CPO material, with an average value of 1.92. If all the Pr4+ were reduced to Pr3+, the material would have an oxygen stoichiometry of 1.95, meaning that the site needs to be further reduced to explain the 1.92 value. This result correlates well with the earlier results that showed that some Ce4+/Pr4+ species were substituted by La3+. If we take the extreme case were all the Pr4+ is reduced to Pr3+ and all the A-site vacancies in LSCF are La vacancies and substitute Ce from CPO, then we would have the formula La0.04Ce0.86Pr0.1O1.93, with an oxygen stoichiometry of 1.93, close to the 1.92 we find in our refinements. Therefore the most likely formulae are La0.04Ce0.86Pr0.1O1.93 and (La0.57Sr0.4)(Co0.2Fe0.8)0.98O3 However there were insufficient data to confirm whether Co and/or Fe was migrating to CPO or to the interface where it could form a new phase undetected by neutron diffraction. These phases were formed during the synthesis of the material, and were stable during the experiment whatever voltage was applied to the material.

Conclusion

The collection of in-situ neutron diffraction data at high temperature and under electrical polarisation in flowing air has enabled us to directly observe the behaviour of an IT-SOFC cathode and electrolyte material under simulated operating conditions. Our results showed no significant interaction between the two materials when a potential was applied to the sample, even under extreme conditions (+/-5 V), however cation migration was observed to have occurred during the synthesis of the sample. These results demonstrate the good compatibility between CPO and LSCF under operating conditions, although extreme care should be taken when synthesising the materials as the lowest temperature and shortest annealing time should be selected to avoid cation diffusion.