Long-Term Stability of Pr₂NiO_4+δ Air Electrodes for Solid Oxide Cells against Chromium Poisoning

Ce_0.9Gd_0.1O_1.95 (GDC) was used as electrolyte material in order to avoid the detrimental formation of Pr-zirconate Pr₂Zr₂O₇ at the electrode-electrolyte interface, which is observed for zirconia-based electrolyte materials. ^26,27 A dense GDC substrate was obtained by uniaxial pressing of GDC powder (Treibacher Industrie AG, Austria) in a stainless-steel die (2.5 cm inner diameter) at 40 MPa. The green body was then pressed isostatically at 300 MPa for 15 min and subsequently sintered for 10 h at 1450 °C with heating and cooling rates of 2 K min⁻¹ and 1 K min⁻¹, respectively. This procedure yields a pellet of 20 mm diameter and 2 mm thickness with a relative density well above 95%. The electrolyte pellet was ground to a final thickness of 1.8 mm using diamond and silicon carbide grinding disks (finest grit P4000). The rather large thickness of the electrolyte substrate allows the placement of a Pt-reference electrode, which was attached with Pt-paste in a thin groove cut along the perimeter of the substrate. Moreover, geometric asymmetry due to imperfect alignment of opposite electrodes can cause artifacts in spectra of three-electrode impedance measurements but becomes less critical for thick electrolytes. ^28–30

A screen-printing paste with 65 wt-% solid loading was manufactured from PNO powder (Marion Technologies, France) with a mean particle size of 0.6 μm (d₁₀/d₅₀/d₉₀ = 0.26/0.64/1.3 μm) and a commercial ink vehicle (fuelcellmaterials, USA) using a triple roll mill (Exakt 50I, Germany). The symmetrical cell was prepared by screen printing electrodes with a diameter of 11 mm on both sides of the electrolyte substrate. The electrodes were sintered at 1100 °C for 2 h in air, resulting in layers of approximately 30 μm thickness. Current collectors made from gold grids with platinum leads were attached to both electrodes using gold paste. A schematic depiction of the symmetrical cell is shown in Fig. 1.

Zoom In Zoom Out Reset image size

The degradation study is divided into three experimental stages, adding a potential degradation source with each consecutive stage. In the first stage (stage 1) the cell was kept for ∼1000 h in dry and Cr-free atmospheres without current load (except for minor perturbation currents during impedance measurements) in order to check the intrinsic stability of the cell and to obtain a baseline against which the subsequent results of the study can be compared. In the second stage (stage 2), a constant dc load of 105 mA cm⁻² was applied to the cell for ∼1200 h (interrupted only by intermittent impedance and current-voltage measurements) in order to investigate the effect of current load on the electrode stability. In the third stage (stage 3), a chromium source was placed in the reactor and the gas stream was humidified in order to simulate air electrode poisoning for cells operated with undried ambient air for another period of ∼1000 h.

The test setup consists of an alumina sample holder, placed in an alumina tube in a horizontal tube furnace (Fig. 1). The cell was characterized via electrochemical impedance spectroscopy (EIS) and current-voltage (I–V) measurements. The temperature was monitored using an S-type thermocouple placed close to the cell. All measurements were carried out at 800 °C in 20% O₂/Ar at a flow rate of 2 L h⁻¹. SOEC and SOFC air electrode operation was simulated by applying a direct current of 100 mA using an electronic source (Toellner TOE 8952, Germany), with oxygen reduction occurring at the SOFC air electrode (i.e. SOFC cathode) and oxygen evolution at the SOEC air electrode (i.e. SOEC anode). Cr-poisoning conditions were established by placing Cr-pellets (99.999% purity) upstream in close proximity but without physical contact to the tested cell and humidifying the test gas by passing the gas stream through washing bottles filled with deionized water before entering the reactor. By thermostatting the washing bottles at 6 °C, a moisture level of 30% relative humidity (r.h.) was achieved (100% r.h. corresponds to full saturation at 25 °C with p(H₂O) = 31.7 mbar ³¹ ). During heating of the setup, the surface of the Cr-pellets is oxidized and thus the actual Cr-source in this study is chromia (Cr₂O₃), which simulates the conditions prevalent in stacks containing interconnects made from ferritic stainless steel or Cr-based alloys. For the conditions applied in this work, the maximum partial pressure of chromium (mainly H₂CrO₄ in humid and oxygen-containing atmospheres) is estimated as 3 × 10⁻⁸ bar at 800 °C ³² when assuming complete thermodynamic equilibration between the oxidized Cr-pellets and the surrounding atmosphere.

EIS and current-voltage measurements were conducted using a frequency response analyzer (Novocontrol Alpha-A, Germany) combined with a potentiostat/galvanostat setup (Novocontrol POT/GAL 15 V/10 A). Impedance spectra were recorded under open-circuit conditions (OCV) between 10 mHz and 1 MHz (10 points per decade) using an excitation voltage of 20 mV (rms). The commercial software package WinFit v.3.4 (Novocontrol Technologies) was used for fitting the spectra with equivalent circuit models by means of complex non-linear least squares (CNLS) algorithms. I–V curves were recorded up to currents of ±100 mA with sweep rates of 1 mA s⁻¹. All electrochemical measurements were performed in four-wire mode using two-electrode (cell) and three-electrode (anode and cathode side) configurations. The three-electrode configuration was used to separate contributions of the SOFC cathode and SOEC anode by means of a Pt-reference electrode. The area-specific resistances (ASRs) were obtained by multiplying the respective resistances with the electrode area (0.95 cm²). Thus, for normalizing the ohmic resistance R₀ (mainly originating from the electrolyte), the widening of the current paths within the electrolyte between the electrodes is neglected.

XRD measurements were performed using a Bruker AXS D8 Advance Eco equipped with a Cu K_α X-ray source and a 1D-LYNXEYE XE-T detector at a step rate of 0.02° s⁻¹ with 1 s measuring time per step. The radiation source was operated at 40 kV and 25 mA. The obtained patterns were evaluated using the commercial crystallographic PDF-4+ data base and Rietveld refinement with the software package Topas. ³³

The chemical compatibility of PNO with GDC was investigated via XRD analysis of annealed powder mixtures. PNO and GDC powders were thoroughly mixed in a weight ratio of 1:1 in a roller bed mill for 24 h in ethanol. The obtained mixtures were dried and without further compaction fired in an alumina crucible at 600 °C–1200 °C for 6 h in air using a fresh powder mixture for each temperature. XRD patterns of the annealed powder mixtures were recorded at room temperature between 10° and 100° (2θ).

Figure S1 (available online at stacks.iop.org/JES/168/014509/mmedia) shows the XRD pattern of the as-received PNO powder (for Figs. S1 to S14 refer to the supplementary material). Rietveld analysis reveals the presence of small amounts of Pr₆O₁₁ (1.3 wt-%) and NiO (1.6 wt-%) secondary phases. The remaining peaks can be assigned to PNO indexed in the orthorhombic space group Fmmm, based on crystal structure data published by Chung et al. ³⁵

The results of the reactivity study of PNO and GDC between 600 °C and 1200 °C are presented in Fig. S2. After heat treatment at 600 °C, all reflections can be attributed to either PNO or GDC except for a single signal corresponding to a minor amount of NiO impurity. At 700 °C, the intensities of PNO reflections are significantly reduced compared to those of GDC and a new perovskite-type PrNiO₃ phase is detected, which indicates partial decomposition of PNO. After annealing at 800 °C, most of the PNO phase has decomposed into PrNiO₃ and weak signals of Pr₄Ni₃O_10±δ —a higher-order Ruddlesden-Popper (RP) structure—are appearing. Similarly, at 900 °C most of the PNO phase has disappeared and the decomposed phase now mainly consists of Pr₄Ni₃O_10±δ . ²⁶ Also, the mass fraction of NiO is significantly increased. After annealing at 1000 °C, a small amount of PNO can be identified in the XRD pattern while the main reflections are assigned to NiO and the ceria phase. At 1100 °C and 1200 °C, all traces of PNO or higher-order RP-phases have disappeared and the only two phases identified are NiO and the ceria phase. Since no reflections from Pr-containing secondary phases (e.g. Pr₆O₁₁) can be observed in the XRD patterns, Pr must have been incorporated into the ceria phase forming a (Ce, Pr, Gd)O_2−δ solid solution. Due to the larger ionic radius of Pr³⁺ (1.126 Å) compared to Ce⁴⁺ (0.97 Å) and Gd³⁺ (1.053 Å) in 8-fold coordination, ³⁶ a significant increase in the lattice constants of the ceria phase should be expected, which would manifest itself as a systematic shift of GDC reflections to smaller diffraction angles. Surprisingly, no such trend can be discerned in the XRD patterns of Fig. S2, indicating that the ceria-based reaction phase contains Pr mainly as tetravalent ion, whose ionic radius (0.96 Å) ³⁶ is almost identical to that of Ce⁴⁺. This is consistent with findings published by other groups, ^37,38 where even a small contraction of the ceria lattice with increasing Pr-content was observed.

The pronounced reactivity between PNO and GDC at temperatures above 1000 °C has also been reported by other authors ^39,40 and raises some questions about the usefulness of PNO as an electrode material for SOFC and SOEC applications. Also from the viewpoint of thermomechanical compatibility, such reactions may be detrimental to cell integrity by inducing electrode delamination. ⁴¹ However, it should be pointed out that temperatures of 1000 °C and above are only relevant for electrode sintering during cell preparation (with sintering times of usually 2 h in contrast to 6 h applied in our reactivity study), whereas the operating temperature of the cell is typically below 800 °C. Moreover, the large particle contact area in the PNO-GDC powder mixture used in our reactivity study clearly promotes interdiffusion processes at higher temperatures. On the one hand this might limit the applicability of PNO in form of a composite with GDC. However, pure PNO electrodes with a comparatively small contact area to the electrolyte or barrier layer might still be viable at operating temperatures around 800 °C, even if PNO partly decomposes into electrocatalytically active components Pr₄Ni₃O_10±δ and Pr₆O₁₁, which may be tolerable or even beneficial since it may further enhance the electrode activity. On the other hand, even if there was some diffusion of Pr into the GDC phase during cell preparation, the electrode might still maintain its functionality, as Pr-doped ceria is a mixed ionic-electronic conductor with significant catalytic activity for the oxygen exchange reaction ⁴² and has been proposed as active component in composite electrode functional layers. ^43,44 This is also consistent with findings reported by Laguna-Bercero et al., ³⁸ who applied a PNO-GDC composite as barrier layer between the PNO electrode and YSZ electrolyte in microtubular cells and observed even a slight improvement of the cell performance at 800 °C with time. Finally, it should be mentioned that compatibility issues can also be resolved by using other fabrication techniques such as the infiltration process, where a porous electrolyte backbone is impregnated with a PNO precursor solution, which would allow the sintering of the impregnated electrode layers at lower temperatures. ^45,46 Also, using modified barrier layers—such as GDC co-doped with Pr—has been shown to mitigate compatibility issues for screen-printed PNO electrodes. ^47,48

Figure 2 depicts OCV impedance spectra of (a) the SOEC anode, (b) the SOFC cathode and (c) the symmetrical cell PNO∣GDC∣PNO at 800 °C. The spectra were recorded during all stages of the long-term study and fitted by CNLS regression analysis. By inspection of the Nyquist plots in Fig. 2, up to three semicircles can be observed in a single spectrum, corresponding to three different processes with distinct time constants. Thus, the applied equivalent circuit (see inset in Fig. 2a) is composed of an inductance L₀, which is attributed to the experimental setup, an ohmic resistance R₀, which is mainly ascribed to the electrolyte, and three R∣∣CPE elements (resistance R in parallel with a constant phase element (CPE) ⁴⁹ ). In Fig. 2, the inductance of the setup as well as the electrolyte resistance have been subtracted from the curves. Corresponding spectra that include the electrolyte resistance are shown in Fig. S3.

Zoom In Zoom Out Reset image size

In order to check the internal consistency of the EIS measurements, impedance spectra of the cell were compared with the sum of the SOEC and SOFC contributions (both measured vs. the same Pt-reference electrode) and found to be in reasonable agreement (Fig. S4). Low-frequency inductive loops start to appear in impedance spectra of the SOEC anode in the final stages of the degradation study (see Fig. 2a), which are artifacts caused by the reference electrode due to the increasing functional asymmetry between both electrodes of the symmetrical cell. ⁵⁰ No such artifacts are visible in the cell spectra recorded without a Pt-reference.

Figure 3a shows results from EIS analysis as area-specific resistances of the ohmic component as well as electrode polarization contributions of the SOEC anode, the SOFC cathode and the entire cell under increasingly harsh conditions.

Zoom In Zoom Out Reset image size

Stage 1 was conducted in order to investigate the intrinsic stability of the pristine PNO electrodes. Initially, both electrodes showed a low polarization resistance of approximately 0.1 Ω cm², which slightly increased during the first 200 h. After that, the ASR values remained stable below 0.15 Ω cm², except for an increase near the end of the first period due to a cooling-reheating cycle necessary for the transfer of the symmetrical cell to another test rig in preparation for the following poisoning experiments.

During stage 2, the influence of electrode polarization was investigated by applying a constant current load of 105 mA cm⁻² with intermittent EIS and I–V measurements. In this stage, it is more difficult to get a clear picture from the ASR trends in Fig. 3a due to a major shutdown shortly after the beginning of stage 2, caused by a malfunction of the tube furnace. After reheating, the electrodes showed a significantly higher ASR but were found to slowly recover, except for a step-wise increase at the end of the segment, which was again caused by a cooling-reheating cycle required to introduce the chromium source. I–V curves were recorded up to current densities of ±105 mA cm⁻² in SOEC and SOFC mode (blue curves in Fig. 3b) which essentially confirm these trends. Thus it appears that, up to this point, cell degradation was not so much caused by the current load but more by thermal cycling between ambient and operating temperature.

In the last segment of the long-term study (stage 3), the actual Cr-poisoning experiment was conducted. Following the introduction of chromium and humidity in the reactor, a slow but steady increase in the ASR of the SOFC cathode up to a final value of 0.5 Ω cm² was observed. In contrast, the ASR of the SOEC anode remained at a constant value of 0.2 Ω cm² (see Fig. 3a). A similar trend is observed for the I–V curves in Fig. 3b, confirming that the SOFC cathode is much more affected by Cr-poisoning than the SOEC anode.

EIS measurements were employed in order to separate the ohmic resistance of the electrolyte from the polarization resistance of the electrodes. However, if a model for the interpretation of the electrode impedance is available, more details about the Cr-poisoning mechanism can be learned. Looking at Bode-plots of the imaginary part of the impedance data (Fig. S5), peak frequencies of the impedance arcs (i.e. single electrode processes) can be readily determined. The most dominant process has its peak frequency at 10¹−10² Hz, while a second process is located in the high-frequency region of ∼10⁴ Hz. In many spectra a third process can be observed at very low frequencies of ∼1 Hz. The corresponding values of the area-specific capacitances obtained by CNLS fitting of cell spectra are ∼10⁻⁴ F cm⁻², ∼10⁻² F cm⁻² and 10¹−10² F cm⁻² for the high-, mid- and low-frequency process, respectively (capacitances have been calculated from CPE parameters according to Fleig ⁵¹ ).

Based on results reported for rare-earth nickelates, ^52–54 the high-frequency arc is assigned to the charge transfer of oxygen ions across the electrode-electrolyte interface and the double layer capacitance at the interface. The main contribution in the mid-frequency region corresponds to the actual electrode reaction, comprising—for an SOFC cathode—a complex series of elementary reaction steps such as adsorption of molecular oxygen on the electrode surface, oxygen dissociation, reduction and incorporation into the crystal lattice as well as oxygen bulk diffusion towards the electrolyte. ^54,55 For the SOEC anode the same elementary steps are expected to proceed in reverse order. In both cases, the low-frequency arc may be connected with concentration polarization caused by gas diffusion or gas conversion. ^56–58 Its size stays constant throughout the whole degradation study (∼10 mΩ cm² for the complete cell) and, because of its small contribution, is soon masked by the mid-frequency arc strongly expanding in stage 3. Hence the low-frequency process can be regarded to be of minor importance for the conclusions drawn in this study.

Looking at the impedance data shown in Fig. 2 and Fig. S5, it is obvious that Cr-poisoning is mainly affecting the mid-frequency arc. This can be expected, since for mixed-conducting electrodes this frequency region is linked to the electrode process including several surface-based elementary reactions (gas adsorption, reduction/oxidation steps, etc.) which will be adversely affected by deposition of Cr-species from the gas phase due to blocking of active surface sites. The high-frequency arc stays almost perfectly constant in the cell spectra (Figs. 2c and S5c), indicating that the corresponding charge transfer process at the electrode-electrolyte interface is not severely affected by Cr-poisoning. This is somewhat surprising, as results from post-test analyses show the presence of chromium at the interface (see below). It should be mentioned that, contrary to the cell spectra, opposite trends in the high-frequency arc are visible in single electrode spectra (Figs. 2a, 2b and S5a, S5b). However, this is certainly an artifact caused by the reference electrode, since it is very unlikely to see these changes almost perfectly cancel each other out in their algebraic sum (i.e. the cell spectrum, see Fig. 2c).

A pronounced discrepancy between degradation rates of differently polarized electrodes on a symmetrical cell has also been reported for La₂NiO_4+δ (LNO) in our previous publication. ¹² Similar to findings in this work, the SOFC cathode was found to be much stronger deactivated than the SOEC anode. Possible explanations have been put forward ¹² and, briefly summarized, propose the more reducing conditions at the SOFC cathode side under polarization (i.e. passage of a current), which might promote Cr-deposition through the reduction—(electro)chemically and/or caused by the lower oxygen activity—of volatile hexavalent chromium species from the gas phase to Cr(III)-containing phases on the electrode surface. Another reason, not incompatible with the first one, can be the inward-directed oxygen flow at the SOFC cathode under current load, accelerating the transport of Cr-species into the porous electrode structure, contrary to the oppositely directed flow at the SOEC anode side. A speculation put forward in our previous work ¹² was the interpretation of the observed discrepancy as an artifact, caused by a shadowing effect in the test rig due to the asymmetric design of the sample holder, where top and bottom electrodes might have been differently exposed to the Cr-containing gas stream. This can now be excluded as a possible explanation, since in this work the position of the SOFC cathode and SOEC anode was reversed.

Finally, it should be mentioned that the increase in polarization resistance observed for PNO in this work is significantly lower than that of LNO, ¹² despite almost identical ambient and operating conditions. This suggests a higher resilience of PNO to Cr-poisoning than its La-analogue. For LNO, post-test analysis revealed surface contamination with chromium throughout the entire porous electrode structure and the presence of a La(Ni, Cr)O₃ reaction phase, which was also identified in a previous work dealing with Cr/Si-poisoning of densely sintered LNO samples. ¹³ Extensive post-test investigations have also been performed in the current study by means of SEM and TEM analyses, which will be discussed in the following sections.

In Figs. S6 and S7, SEM-BSE cross section images of the SOEC anode and SOFC cathode with corresponding EDXS analyses are shown. Both PNO electrodes have a thickness of approximately 30 μm and seem to adhere well to the electrolyte. No distinct deposits of Cr or any other contaminants could be identified by EDXS area analysis. For a more detailed examination, EDXS elemental mappings of both electrodes were recorded. Since the detection of chromium by EDXS is complicated by a strong overlap of Cr-lines with those of lighter rare-earth elements and oxygen, chromium maps were generated using the higher-resolution WDXS technique.

In Fig. 4a a SEM-BSE image of a cross section of the PNO SOEC anode with corresponding elemental maps of Pr, Ni, Cr, Ce and Gd are shown. Bright spots in the Ni-map indicate NiO particles, already present as a secondary phase in the raw material (see Fig. S1). Elemental maps of Pr, Ce and Gd do not display any irregularities. Pr seems to be distributed rather homogeneously throughout the entire electrode layer. Ce and Gd are found only in the electrolyte and no signs of cation interdiffusion can be detected at this magnification. According to the chromium map, Cr-levels within the SOEC anode are below the detection limit of WDXS (intensities in the Cr-map of Fig. 4a have been strongly amplified and correspond to the noise level, as can be seen from the low contrast between electrolyte and electrode). However, in the Cr-map of the PNO SOFC cathode (Fig. 4b), there is clear evidence of Cr-deposits, which are mostly concentrated near the electrode surface. Similar to the performance degradation of LNO electrodes in our previous Cr-poisoning study, ¹² the increase in ASR of the SOFC cathode might thus be correlated with the presence of Cr-compounds within the electrode structure.

Zoom In Zoom Out Reset image size

The impact of Cr-poisoning on the electrode microstructure was assessed on the basis of polished cross sections of the electrode layers before and after the long-term study (Fig. S8). No significant differences in the microstructures can be observed and thus the degradation behavior cannot be correlated with particle coarsening or other changes in electrode morphology.

Semi-quantitative analyses by SEM-EDXS were able to confirm the presence of Cr-species in the SOFC cathode layer but cannot provide specific information on the type of Cr-compounds. Therefore, comprehensive STEM investigations were conducted. FIB samples were prepared for both electrodes from sites near the electrode surface and at the electrode-electrolyte interface.

A detailed analysis of the porous SOFC cathode close to the surface by means of EELS elemental mapping of Pr, Ni, O, and Cr is presented in Fig. 5. The Pr-map shows accumulations of praseodymium next to pores, which is not surprising considering that PNO is thermodynamically unstable at 800 °C in air and expels Pr-oxides, ^59,60 thereby transforming into higher-order RP-phases (e.g. Pr₄Ni₃O_10±δ ) with lower Pr:Ni-ratio. It should be mentioned that these decomposition products have themselves catalytic properties for sustaining the electrode reaction. Pr-oxides enhance the oxygen reduction kinetics while Pr₄Ni₃O_10±δ does possess significant oxygen exchange activity and high electronic conductivity ^61,62 making it suitable for SOFC electrode applications. ⁶³ Decomposition might indeed enhance the electrode performance since the phase mixture has been reported to show even faster oxygen exchange rates than the parent PNO phase. ^60,61

Zoom In Zoom Out Reset image size

A closer look at the Cr-map in Fig. 5 reveals that Cr is absent from areas with high Pr-content, which suggests that Pr-oxides have a low affinity for chromium. This is in contrast to La-oxide phases which readily form Cr-containing compounds at temperatures of 700 °C and 800 °C. ^11,17 The presence of Cr is also confirmed by the EEL spectrum of the spot marked in Fig. 5a, where the Cr-L edge can be clearly identified. As evident from the Ni-map, the distribution of Ni is mostly homogeneous except for a small isolated NiO grain. The same electrode section was analyzed using EDXS spot analysis (see Fig. S9 for details), which reveals the presence of decomposition products of PNO as well as impurity elements Pt, Ir and Si. Chromium is confirmed on some sites but a precise phase analysis is not possible due to the small size of the Cr-deposits.

A similar approach was used to study the decomposition and reaction phases close to the electrode-electrolyte interface of the SOFC cathode (see Figs. S10 and S11 for details). Again, Cr-depositions were confirmed by EELS but no phase identification was possible. Isolated Si-impurities were detected in conjunction with elevated levels of Pr and O and almost no Ni, which suggests the formation of a Pr-silicate phase ^64,65 by analogy with findings reported for LNO. ¹³ Silicon is originating from impurities in PNO, since traces of Si were detected in the as-delivered powder by X-ray fluorescence spectroscopy. However, due to its sporadic occurrence, silicon is unlikely to play a major role in the degradation behavior observed in this study.

The interface between the electrode and the electrolyte is of special interest for SOFC/SOEC applications and was therefore more closely inspected. Figure S12a shows a STEM-HAADF cross section image of a single PNO grain in the SOFC cathode layer in direct contact with the GDC electrolyte. Contrast variations in the HAADF image indicate a heterogeneous chemical composition across the interface to the electrolyte. EDXS elemental distribution maps are presented in Fig. S12b and show that Pr, Ni, and O are homogeneously distributed within the PNO grain, with no indication of Pr-oxide exclusions. The Cr-map confirms the presence of chromium at the interface, where Cr-deposits are located at the surface of the back side of the sectioned PNO grain.

From the elemental maps in Fig. S12b it is evident that significant diffusion of Pr into the electrolyte material has occurred, while no such interdiffusion is observed for Ni. This is not surprising, considering the results from the PNO-GDC compatibility study (see above). A line scan across the interface (Fig. 6) shows that the cation interdiffusion zone is approximately 160 nm wide with small amounts of Cr present at its boundary. Pr is highly concentrated in this region, whereas Ni remains stationary in the PNO phase. The interdiffusion zone consists mainly of Pr-doped ceria (PDC)—with a strong gradient in the Pr:Ce-ratio from pure GDC to almost pure Pr-oxide—sandwiched between the GDC electrolyte and a PNO grain with pronounced Pr-deficiency ("Pr_1.8NiO₄"). While the PNO structure may sustain such high A-site deficit, ⁶⁶ it is more likely to be a region in transition from the PNO parent phase to higher-order RP-phases, as has been observed by HR-STEM in the form of stacking faults at several locations. An example of the coalescence of single perovskite layers to double, triple etc. layers can be seen in Fig. S11.

Zoom In Zoom Out Reset image size

At high oxygen partial pressure, Pr-doped ceria is a mixed ionic-electronic conductor with its ionic conductivity close to that of GDC, ^41,43,67 so that charge transfer of oxide ions between the electrode and the electrolyte may still be sustained. However, from the standpoint of thermomechanical compatibility, the formation of such interdiffusion zones may be expected to weaken the electrode-electrolyte adhesion and may lead to delamination in the long term.

STEM investigations of the SOEC anode, both at the surface and close to the electrode-electrolyte boundary, clearly demonstrate the absence of chromium from the electrode layer (see Figs. S13 and S14) and none of the contaminants found at the SOFC cathode side (Pt, Ir and Si) were detected. Again, Pr-rich and Ni-depleted regions, often located at the pore surface, ⁶⁰ indicate Pr-oxide secondary phases resulting from the decomposition of PNO at higher temperatures. The Pr/Ni-overlay in Fig. S13c shows the precipitation of PrO_x from the Pr-nickelate phase more clearly. Similar to the SOFC cathode side, strong diffusion of Pr into the GDC electrolyte was observed for the SOEC anode at the electrolyte interface (Fig. S14). The electrolyte surface appears to be covered with a dense PrO_x-layer and a thin Pr-GDC interdiffusion zone is visible along the boundary (see Figs. S14b and S14c). The contact region shows clear signs of fracture and disintegration, which could be a result of the interdiffusion process, weakening the structural integrity of the PNO grains directly connected to the GDC substrate. However, it might as well be an artifact caused by FIB lamella preparation which is quite a delicate procedure for porous structures. If it was indeed a true representation of the interface quality during operation, it would be expected to result in almost complete cell failure as opposed to the comparatively low level of electrode polarization observed in this study.

Interestingly, while Cr-deposition in the SOFC cathode structure has been unambiguously established, no clear indication of a chemical reaction of Cr with PNO or any of the secondary phases could be found. This is quite surprising considering the high affinity of the La-analog compound LNO for Cr-species. ^12,13,17 The presented results suggest that chromium is deposited as Cr-oxide with low chemical affinity towards the underlying PNO electrode material. This could then be responsible for the performance degradation of the SOFC cathode, where even a thin layer of inactive material can significantly reduce the oxygen exchange activity by partially blocking the electrode surface.

To be sure, other potential sources for cell degradation have been identified as well (partial decomposition of PNO, cation interdiffusion and grain fracture at the electrode-electrolyte interface), but they occurred on both electrode sides and thus cannot account for the observed differences in performance degradation. Moreover, since significant degradation was observed only after the introduction of chromium into the reactor, a clear correlation with Cr-poisoning can be established.