LaPr₃Ni₃O_9.76 as a candidate solid oxide fuel cell cathode: Role of microstructure and interface structure on electrochemical performance

One of the key tasks in SOFC development is the lowering of operational temperatures to combat material degradation and hence increase device lifetime.¹ However, at intermediate temperatures (550-750 °C), the oxygen reduction kinetics are reduced and thus negatively impact the performance of the cell. Therefore, the focus is to achieve considerable improvement in cathode oxygen reduction reaction (ORR) kinetics at intermediate temperatures and faster oxygen transport across the electrode/electrolyte interface.¹ A further shortcoming identified has been the limited participation of the triple phase boundary in terms of its area in electronic conducting solid oxide fuel cell (SOFC) electrode materials, and thus mixed ionic and electronic conducting (MIEC) materials are being studied in order to extend the involvement of the electrode in the ORR to realise better performance in SOFCs.^2–4 This outlines the importance of MIEC materials since the extended participation of the electrode in the ORR is expected to compensate for the reduced oxygen reduction kinetics at intermediate temperatures.

One approach in the development of new MIEC electrode materials for SOFCs is to focus on perovskite-related layered oxides. Recent interest has been directed toward the Ruddlesden-Popper (RP) phases. Formulated as A_n+1B_nO_3n+1 (n = 1, 2, and 3), the RP phases consist of n consecutive ABO₃ perovskite layers, alternating with AO rock-salt layers, stacking along the crystallographic c-axis.⁵ The lower-order RP phases, particularly the n = 1 phases such as La₂NiO_4+δ, Pr₂NiO_4+δ, and Nd₂NiO_4+δ,^6–8 have previously been extensively investigated, with studies considering the electronic and ionic transport as well as the performance of these oxides as SOFC electrodes. Higher-order RP phases (n = 3) have received little attention, but because of their excellent total conductivity and improved stability^6,7 in the temperature ranges of interest (550–750 °C) when compared to the n = 1 phases, there is significant potential to develop a new class of electrode.

In developing SOFC cathodes, it is important that discovery of promising new material families and optimisation of existing materials are investigated, particularly focusing on the interfacial characteristics. Oxygen ions produced at the cathode surface have to diffuse through the electrode/electrolyte interface, and depending on the interfacial resistance of the SOFC system, the cell performance could be limited. It is challenging to obtain an atomic-scale picture of the interface during operation because a non-destructive probe with high penetration power and high sensitivity, compatible with the harsh operation conditions, is needed.⁹ One approach is to construct cells for electrochemical impedance measurements and follow the measurements with post-measurement microstructural analysis. Using this information, logical steps such as modifying the electrode ink preparation, cell sintering temperatures, modifying the electrolyte surfaces, or adding an interlayer to optimise the interface structure can be systematically studied.

There are many reports in the literature highlighting routes to optimising the interface structure.^10–13 Working on ungraded and functionally graded composites of La₂NiO_4+δ and La₄Ni₃O_10−δ, Woolley and Skinner showed the significant role microstructure and electrode/electrolyte contact have on the performance of a material used as an electrode.^12,13 They reported the ASR of 0.62 Ω cm² at 700 °C for an ungraded 50:50 composite,¹³ which was further improved to 0.53 Ω cm² at 700 °C by functional grading of the 50:50 composite cathode.¹² This improved performance in both cases was ascribed to a good combination of ionic and total conductivity, being contributed by two different constituents of the composite and the resulting microstructure and improved electrode/electrolyte interface contact. Vibhu et al., understanding the importance of electrode/electrolyte contact, used a Ce_0.9Gd_0.1O_1.95 (GDC) interlayer in their study of another end member of the series, Pr₄Ni₃O_10+δ, as an SOFC cathode with yttria stabilised zirconia, ZrO₂-Y₂O₃ (YSZ), as the electrolyte.¹⁰ 

Several authors working on related RP materials, La₃Ni₂O_7−δ and La₄Ni₃O_10−δ have explored the development of the microstructure, including preparing infiltrated composite cathodes of these oxides, and reported improvement in performance.^14–17 Choi et al. reported an ASR of 0.11 Ω cm² at 750 °C for La₃Ni₂O₇,¹⁴ and Kim et al. reported the ASR of a Sr doped n = 3 phase as ∼0.13 Ω cm² at 750 °C.¹⁵ Sharma et al. by using a spray deposition technique improved the performance for the n = 3 composition La₄Ni₃O_10−δ, reporting an ASR of 0.30 Ω cm² at 700 °C, whilst producing a composite with Ce_0.9Gd_0.1O_2-δ improved this ASR by one order of magnitude.^16,17 Earlier studies of La₄Ni₃O_10−δ as an SOFC cathode by Amow and co-workers⁷ reported an ASR of 1.0 Ω cm² at 800 °C. This significant improvement from 1.0 Ω cm² at 800 °C to 0.15 Ω cm² at 700 °C was ascribed to the improvement of the cathode microstructure and thereby the adherence of the electrode to the electrolyte pellet. A report by Sharma et al.¹⁷ in which the ASR was reported to decrease from 72.41 Ω cm² to 12.78 Ω cm² and then to 2.21 Ω cm² by changing the microstructure from isolated cauliflowers to connected round agglomerates and then to the 3-D coral-type microstructure, respectively, clearly signifies the importance of optimising the microstructure and its influence on the electrochemical performance of these materials. It, however, is also important to point out that it is possible to have comparable oxygen reduction activity even with substantial differences in the microstructure and interface morphology. Jiang, for example, reported no sizeable impact on the electrocatalytic activity by examining the thermally and electrochemically induced electrode/electrolyte interfaces in La_1−xSr_xMnO_3−δ (LSM) and La_1−xSr_xCo_1−yFe_yO_3−δ (LSCF) electrodes on YSZ and GDC electrolytes, respectively.¹⁸ 

In this regard, we have recently reported the preparation of a higher-order RP composition, La₂Pr₂Ni₃O_9.65, and presented preliminary solid oxide fuel cell cathode performance¹⁹ data. An Area Specific Resistance (ASR) of 0.34 Ω cm² at 800 °C was reported, and two important observations of poor cathode microstructure and lack of adherence to the electrolyte were noted. More importantly, this previous work highlighted the importance of electrode/electrolyte interface structure in the development of SOFC technology. Building on these previous studies as well as our previous observations of poor cathode microstructure and electrode-electrolyte contact being the reasons behind the relatively lower electrochemical performance of a related composition, La₂Pr₂Ni₃O_9.65, we report the successful preparation and optimisation of a new RP composition, LaPr₃Ni₃O_9.76, with impressive electrochemical performance.

LaPr₃Ni₃O_9.77 (L1P3N3) was synthesised using the citrate sol-gel route. Stoichiometric amounts of La(NO₃)₃·6H₂O (Sigma Aldrich, 99.0%), Pr(NO₃)₃·6H₂O (Sigma Aldrich, 99.99%), and Ni(NO₃)₂·6H₂O (Sigma Aldrich, 99.0%) were dissolved in an aqueous solution of 10% (by weight) of citric acid (Sigma Aldrich, 99.99%). The solution was heated at 250 °C, under constant stirring for three hours, until a gel was obtained. The gel, after being decomposed in air for 12 h at 600 °C, was ground, and the resultant powder was annealed for 24 h in air at 950 °C. The XRD patterns of the material were collected using a PANalytical X’Pert Pro MPD (Cu Kα source), and Rietveld refinement of the XRD data, performed with the GSAS/EXPGUI software package,^20,21 was used to confirm the phase identification and extract unit cell parameters. The XPS spectra were obtained using a Thermo Scientific K-Alpha + instrument with a monochromatic Al Kα radiation source (hν = 1486.6 eV) operating at 2 × 10⁻⁹ mbar base pressure. The X-ray source used a 6 mA emission current and 12 kV anode bias, giving an X-ray spot size of up to 400 μm². A pass energy of 200 eV was used for survey spectra and 20 eV for core levels. Avantage Data System software (ThermoFisher Scientific) was used for quantitative analysis, and binding energy values were corrected by setting the C1s peak to 284.8 eV. Post-microstructural investigations of the half-cells were achieved using a LEO Gemini 1525 FEG scanning electron microscope with an accelerating voltage of 5.0 kV.

The commercially available La_0.8Sr_0.2Ga_0.8Mg_0.2O_3−δ (LSGM) electrolyte powder, purchased from Fuel Cell Materials, USA, was pressed (0.04 MPa) into 13 mm diameter disks using a uniaxial press, followed by the cold isostatic press (300 MPa), and then sintering at 1450 °C for 8 h. The thickness of the LSGM pellets after sintering and polishing with silicon carbide paper (1200 grit) was in the 700-800 µm range. An L1P3N3 electrode ink was prepared by mixing with an organic ink vehicle supplied by Fuel Cell Materials, USA, in a 2:1 weight ratio to prepare symmetrical cells by a screen-printing technique. Two types of electrode slurries were prepared, one by homogeneously mixing the as-synthesised powders with an ink vehicle (ink A) and another by ball milling the powders for 24 h before mixing with an ink vehicle and then triple roll milling the prepared ink to have uniform particle size (>3 µm) (ink B).

Symmetrical cells of the configuration L1P3N3∥LSGM∥L1P3N3 were prepared by screen-printing ink A on both sides of polished LSGM pellets (sample 1) and ink B on both sides of as-sintered LSGM electrolyte pellets (sample 2), polished LSGM pellets (sample 3), and polished and calcined (1000 °C for 6 h) LSGM pellets (sample 4), all with an active area of 0.5 cm². Sample 5 was prepared in the same way as sample 3 except for the modification of adding an LSGM ink interlayer, which was screen-printed on the LSGM pellet which was then annealed for 1 h at 1000 °C before the L1P3N3 electrode could be screen-printed, leading to the cell configuration of L1P3N3$|$LSGMInk$|$LSGM$|$LSGMInk$|$L1P3N3. In all the samples described above, cells were sintered at 1150 °C for 1 h to ensure good electrode adherence. Since the L1P3N3 decomposes at temperatures above 1000 °C, slow cooling (0.5 °C/min) to 950 °C from 1150 °C under oxygen was required, which leads to the re-formation of L1P3N3, and to make sure that L1P3N3 is fully recovered, the cell was kept at 950 °C for 6 h. Sample 6 was prepared in the same way as that of sample 5, except that the cell sintering was performed at 1050 °C, and sample 7 was sintered for 2.5 h at 1050 °C under ambient conditions. A summary of the cell preparation is presented in Table I. A Pt mesh was used as the current collector, and impedance measurement of the half-cells was performed using a Solartron 1260 impedance gain/phase analyzer in combination with a Solartron 1287 electrochemical interface in the 0.1–10⁶ Hz frequency range with a perturbation amplitude of 10 mV. DRTtools was used to interpret the impedance spectroscopy data via distribution of relaxation time to identify the processes at low and high frequencies.²² 

X-ray diffraction patterns and XPS spectra, for phase identification and for elemental composition, respectively, both are provided in the supplementary material. The cathode performance obtained by impedance spectroscopy measurements of all of the samples is presented as Nyquist plots, along with the fitting results, in Fig. 1. The equivalent circuit adopted for samples 1-4 is LR(CPE₁R₁), where L is the inductance, R represents the ohmic resistance, and (CPE₁R₁) represents the constant phase element and resistance of the processes occurring at the cathode. In the case of samples 5-7, the Nyquist plot shows two separate responses (high and low frequency responses) at low temperatures, but only one response can be seen at high temperatures (Table II). Accordingly, the equivalent circuit adopted for low temperatures is LR(CPE₁R₁) (CPE₂R₂), whereas the equivalent circuit LR(CPE₁R₁) was used for high temperatures. Since it is difficult to separate and compare the high, medium, and low frequency responses in Nyquist plots above, DRTtools²² was used to interpret the impedance spectroscopy data in terms of distribution of relaxation times for sample 6 (Fig. 2). Similar behavior was observed in samples 5 and 7. A clear observation of a very small contribution arising from the high-frequency response compared to the low-frequency response can be seen. High-frequency contributions in such materials are proposed to stem from charge transfer of oxygen ions between the cathode and the electrolyte, and the low-frequency contribution stems from the molecular oxygen dissociation process.^23–26 The minor contributions as revealed by DRT analysis could not be identified and will need further experimental work such as impedance measurements under varying pO₂. Nevertheless, it is noteworthy and signifies the importance of adding an LSGM interlayer and ball milling the electrode material in order to reduce the interfacial resistance contribution to overall polarisation resistance. It becomes clear when we compare these results to our previous investigations of the La₂Pr₂Ni₃O_9.65 composition, wherein an interlayer was not used; it was observed that the poor adherence of La₂Pr₂Ni₃O_9.65 to the electrolyte pellet was responsible for the charge transfer process being the rate-limiting factor for the ORR and thus the major contribution to overall cell polarisation resistance.¹⁹ The contribution of interfacial resistance to overall polarisation resistance is significantly reduced in the present case by adding the LSGM interlayer and ball-milling the electrode material. As we will see later in post-testing microstructural SEM images, the electrode/electrolyte interface improvement by adding an interlayer is indeed substantial and therefore the enhancement in cathode performance.

The normalised ASR obtained for sample 1 at a temperature of 700 °C was found to be 0.61 Ω cm², with an activation energy of 0.63 eV (Fig. 3). This improved to 0.53 Ω cm² by simply ball milling the cathode material and then triple roll milling the ink (sample 2), which further increased in terms of performance, 0.17 Ω cm², when the LSGM pellet used was polished (sample 3). This is interesting as intuitively one would expect better performance when the pellet used was not polished, providing contact points for sintering, but we observe otherwise. One possible explanation could be that total effective contact area of the electrode ink layer with the electrolyte was reduced when compared to the polished smooth pellet, because of the local roughness observed in the un-polished pellets. The next change was made by sintering the LSGM pellet again after polishing, and it was observed that it did not have a significant impact on the performance of the material, presenting an ASR of 0.18 Ω cm² at 700 °C. As one would expect, the polarisation resistance of samples 2-4 showed similar thermal activation behavior, with almost the same activation energy in all cases (Fig. 3). The thermal behavior of sample 6 was further analyzed, and it was found that activation energy for the high frequency response (1.18 eV) was greater than the low frequency response (0.52 eV) (Fig. 4). Similar behavior was observed in samples 5 and 7.

The microstructure of the material showed considerable improvement when ball milled and triple-roll milled. The uniform particle size (1-3 µm) also resulted in better contact with the electrolyte pellet, which was later reflected in the half-cell performance. In both the top and cross-sectional view of the cell, sample 1 shows considerable coarsening (>10 µm) when compared to samples 2-4 (Figs. 5 and 6). Since the ink used for samples 2-6 was the same (ink B), the top-view of the cathode after impedance measurements looks similar in these samples. Although the microstructure shows significant improvement from sample 1, two important things that stand out in samples 2-6 are the slight densification (>5 µm) in certain areas (highlighted in white outlines in Fig. 5) and the regions with poor particle interconnectivity (Fig. 5).

Densification is expected to impact the adherence of the L1P3N3 electrode to the LSGM pellet negatively and therefore the interface structure. This also explains the inferior performance of sample 1 because the substantial sintering of L1P3N3 heavily increases the interfacial resistance and thus poor cathode performance. An interesting observation, however, is the better performance of samples 3 and 4 over sample 2 when the SEM image suggests relatively better electrode-electrolyte contact to the naked eye in the latter case. This, we believe, is because the un-polished LSGM pellet used in sample 2 presents an uneven surface and thus reduces the effective electrode-electrolyte contact at the interface (Fig. 6). The top-view of sample 7 shows almost no particle growth due to cell sintering and also presents better particle interconnectivity when compared to samples 1-6, which partly explains its impressive performance (Fig. 5).

Although the ASR of 0.17 Ω cm² measured at 700 °C can be considered a good performance, the interface structure as revealed by SEM (S2-4, Fig. 6) in the post-test microstructural analysis showed room to further optimise the contact area. In the case of higher-order RP phases, achieving a good adherence between electrolyte and electrode has been a problem, primarily because of the decomposition of these materials at temperatures above ∼1050 °C and therefore the employment of low temperatures for sintering the cells. By sintering the cells under an oxygen atmosphere, it was possible to avoid the temperature restriction imposed by the phase stability, but it was observed that the interface could still be improved. Adding an interlayer of LSGM ink before screen-printing the electrode ink onto the electrolyte pellet was therefore conceived as a route to improve the adherence of the electrode material to the electrolyte pellet. An LSGM interlayer of ∼10 µm thickness, as used in sample 5, resulted in reducing the ASR to 0.08 Ω cm² at 700 °C. This significant improvement has been assigned mainly to the improved interface adherence. This is very clear when we examine the post-test microstructural SEM images of sample 5. The contact between the electrode and electrolyte has improved considerably and therefore the cathode performance (Fig. 6). Particle densification, however, can still be observed in sample 5, and reducing the sintering temperature to 1050 °C under oxygen atmosphere still showed particle coarsening (S6, Fig. 5). This is because of the elongated periods of time (∼11-14 h) needed for the cell sintering procedure under oxygen atmosphere conditions. Therefore, an attempt was made to reduce the cell sintering time to avoid particle agglomeration and also to avoid the complex process of sintering the cell under an oxygen atmosphere. Sintering the cell at 1050 °C for only 2.5 h under ambient atmosphere reduced the electrode particle coarsening significantly (S7, Figs. 5 and 7). This is because the time for sintering the cell was reduced substantially in sample 7, and this is also reproduced in the performance of the material, presenting an ASR of 0.08 Ω cm² at 700 °C. The post-test microstructural SEM images (Fig. 5) in sample 7 show that there is almost no particle coarsening, improved inter-particle connectivity, and very good levels of porosity in the electrode ink. This is one of the reasons behind its superior performance, apart from the observation that the electrode adheres to the electrolyte very strongly, thereby reducing the interfacial polarisation significantly in sample 7 (Fig. 7). It is notable that this performance is a significant improvement to the one reported by Sharma et al. for La₄Ni₃O_10−δ using the spray deposition technique and showing an ASR of 0.15 Ω cm² at 700 °C.¹⁷ 

A new higher-order RP phase composition LaPr₃Ni₃O_9.76 was synthesised by a sol-gel route and tested for IT-SOFC cathode properties. Microstructure and interface optimisation was the focus of this work, which was first achieved by ball-milling the electrode material and triple-roll milling the ink prepared, and the best performance achieved was 0.17 Ω cm² at 700 °C. However, the electrode adherence was still observed to be limited and thus an LSGM ink interlayer between the electrode and electrolyte was used to improve the interface contact, which produced the best performance of 0.08 Ω cm² at 700 °C. This significant improvement has been ascribed to the improved inter-particle connectivity, reduced densification of electrode material, and improved interface structure. It is to be noted that this is the best performance ever reported for n = 3 Ruddlesden-Popper phase materials with LSGM as an electrolyte.

See supplementary material for the powder X-ray diffraction data and X-ray photoelectron spectroscopy data of the L1P3N3 composition.

The authors would like to thank the Imperial College President’s Ph.D. Scholarships and EPSRC Advanced Characterisation of Materials Centre for Doctoral Training (No. EP/L015277/1) for funding this research.