Perovskite oxide LaCr 0.25 Fe 0.25 Co 0.5 O 3- δ as an efficient non-noble cathode for direct ammonia fuel cells

LaCoO


Introduction
A wide range of clean renewable energy resources are becoming growingly necessary to satisfy rising energy demands. Many of these, however, are intermittent in nature and rely heavily on external parameters such as wind and sun exposure [1]. It is therefore widely accepted that development of clean, reliable energy conversion and storage devices such as metal-air batteries, fuel cells and water splitting technologies will become progressively important in stabilising demand on the grid [2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19]. Commercial expansion of proton exchange membrane fuel cells (PEMFCs) and anion exchange membrane fuel cells (AEMFCs) however, are largely limited by the sluggish nature of the oxygen reduction reaction (ORR) and the high expense associated with Pt-based electrocatalysts [12,[20][21][22][23][24][25][26]. For perspective, the current price of electrocatalysts account for greater than 30% of the total PEMFC cost [27][28][29]. On this basis, exploration of effective electrocatalysts to assist ORR at the lowest possible cost is essential for future use of these devices [12,21,30].
Perovskite oxides have attracted much attention due to the vast flexibility in their chemical and electronic structure, allowing them to be easily exploited for the design of highly active and stable ORR electrocatalysts in alkaline media [12,13,20,21,23,24,29,42,[44][45][46][47][48][49][50][51]. Such materials have the general formula ABO 3 and can accommodate a wide variety of cations in both the A and B sublattice [29]. Here, the A-site is occupied by a rare earth metal (e.g. La, Gd, Pr or alkaline earth metal) usually with dodecahedral coordination, whilst the B-site is generally a transition metal (e.g. Mn, Cr, Fe, Ni, Co) that resides in corner-sharing octahedra [29,52]. It is widely accepted that optimum ORR activity for perovskites follow Sabatier's principle of catalysis, whereby the interaction between the catalyst and the surface adsorbed oxygen species is moderate [53]. Unfortunately, it is challenging to directly measure the adsorbate binding energy at the surface. Focus has therefore been shifted towards identifying activity descriptors which can be linked to adsorbate binding energy and govern activity [54][55][56]. By using molecular orbital theory, Suntivich et al. established a volcano-type relationship between intrinsic ORR activity and the σ * orbital (e g ) occupation of B-site ions. It was found that perovskites with near unity e g orbital occupancy such as LaMnO 3 (t 2 g 3 e g 1 ), LaCoO 3 (t 2 g 5 e g 1 ) and LaNiO 3 (t 2 g 6 e g 1 ), exhibit optimum ORR activity [23,57]. In cases where there was a lack of e g filling such as in LaCrO 3 (t 2 g 3 e g 0 ), the B-O 2 bonding was too strong and resulted in the O 2 2-/OHexchange becoming the rate determining step. Alternatively, where there was too much e g filling such as in LaFeO 3 (t 2 g 3 e g . 2 ), the interaction was too weak and the regeneration of OHions was the rate determining step. As well as e g orbital filling and metal-oxygen covalency, positioning of the D-band centre, oxidation states and oxygen vacancies have also been reported to affect ORR activity and should not be considered independently [21,[58][59][60].
More recently, strategically doping into the B-site and varying elemental composition has proven to be a promising strategy to design highly active perovskite catalysts for ORR. Sunarso 6 Ni 0.4 O 3-δ to exhibit activity similar to that of commercial Pt/C at a reduced cost, making it an efficient cathodic electrocatalyst for Li-air batteries [49].
Despite the significant body of work described above, certain parameters affecting the electrocatalytic performance of LaCoO 3 towards ORR, such as the effects of doubly doping at the B-site, have not been as widely explored. In the present work, a series of B-site substituted LaCoO 3-δ (LCO) perovskites, including LaCr 0.5 Co 0.5 O 3-δ (LCCO), LaFe 0.5 Co 0.5 O 3-δ (LFCO) and LaCr 0.25 Fe 0.25 Co 0.5 O 3-δ (LCFCO), were synthesised via a conventional combustion method and mixed with carbon black to investigate their activity towards ORR in O 2 -saturated 0.1 M KOH solution using a rotating disk electrode (RDE) technique. Thorough investigation into the structure property relationship of the electrocatalysts was conducted to study the synergetic effects of doping both Fe-and Cr at the B-site of LCO. To the best of our knowledge, this is the first report on the effects of strategically introducing two dopants into the Co B-site of LCO on oxygen reduction activity and to truly exploit the flexible nature of perovskite oxides. The effects of firing temperature of LCFCO on its electrochemical properties was also investigated. Finally, applicability of the perovskite cathode in a direct ammonia fuel cell was explored to assess the potential of the catalyst in a practical device as well as its ability to compete with expensive platinum group metals (PGMs). It was found DAFCs with LaCr 0.25 Fe 0.25 Co 0.5 O 3δ /C cathode exhibits comparable performance to those with Pt/C cathode.

Synthesis of perovskite powders
The perovskite powders were synthesised using a conventional combustion method [61]. In brief, appropriate amounts of the metal nitrate precursors were dissolved in an aqueous solution at room temperature. Citric acid was added in a molar ratio of 1:1.2:1.2 of total metal ions:citric acid:ethylene glycol. The resulting solution was stirred and heated to 120 o C. After the evaporation of water, the sample was heated to 410 o C to form an ash which was finely ground using an agate mortar and pestle. The powder was calcinated in air at 500 o C for 2 h with a heating/cooling rate of 5 o Cmin -1 before being reground and further calcinated at 1000 o C for 4 h with a heating/cooling rate of 3 o Cmin -1 to obtain the final perovskite phase. The collected LaCoO 3-δ , LaCr 0.5 Co 0.5 O 3-δ , LaFe 0.5 Co 0.5 O 3-δ and LaCr 0.25 Fe 0.25 Co 0.5 O 3-δ samples were labelled LCO, LCCO, LFCO and LCFCO respectively. The synthesis of LCFCO was further explored by adjusting the final calcinating temperature to 600, 700, 800 and 900 o C, labelled as LCFCO-600, LCFCO-700, LCFCO-800 and LCFCO-900 respectively.

Physicochemical characterisation
X-ray Powder Diffraction (XRD) analysis was used to examine the phase and purity of the perovskite powders. Measurements were carried out at room temperature on a PANalytical X'Pert Pro diffractometer (Cu Kα source, 1.5405 Å) and collected in the 2θ range of 20-80 o with a step of 0.0167 o . Phase analysis and identification was conducted via the HighScore software. It should be noted that particle size can affect intrinsic ORR activity, therefore, to gain insight into size-activity relationship, crystallite sizes were calculated by the Scherrer equation using data extracted from the XRD patterns (Eq. S(1− 3)).
The above methods give an average value of crystalline sizes; however, real samples usually consist of grains with different domains. To evaluate the distribution of grain sizes and account for the non-coherent domains, Scanning Electron Microscopy (SEM) images were taken. SEM was used to examine the morphology of the catalyst at the surface using a Zeiss SUPRA 55-VP, equipped with an Energy-Dispersive X-ray (EDS) spectrometer for elemental analysis. Specimens were prepared prior to testing by depositing small amounts of the powder and coated with Au. Raman spectroscopy was conducted using a Renishaw InVia Raman Microscope with an exciting wavelength of 532 nm.
X-ray Photoelectron Spectroscopy (XPS) analysis was measured using a monochromated Al K α X-ray source on a Kratos Axis Ultra DLD spectrometer (Kratos Analytical, Manchester, UK). The data was collected at a take-off angle of 90 o to the surface plane and analysed using the Casa XPS package. To prevent the surfaces becoming positively charged, charge neutralization was employed and the spectra were later referenced to the C-C peak at 285.0 eV during analysis.
Textural properties were assessed via N 2 adsorption-desorption isotherms of the perovskite powders recorded at liquid N 2 temperature using a Micromeritics ASAP 2020 apparatus. Samples were degassed at 300 o C for 1 h. Specific surface areas (SSA) were determined by applying Brunauer-Emmett-Teller (BET) method.

Electrochemical characterisation
The electrochemical activity of the samples towards ORR were measured via a rotating disk electrode (RDE) technique in O 2 -saturated 0.1 M KOH solution at room temperature (ca. 20 o C). An ink was prepared by mixing of the perovskite catalyst powder and carbon black (Vulcan XC-72R) in a 5:1 mass ratio respectively. The oxide was physically mixed with high surface area carbon black (Vulcan XC-72R, specific surface area 216 m 2 g -1 ).
to facilitate electrical contact between the oxide particles and help eliminate issues relating to electronic conductivity, favouring complete utilisation of the perovskite surface [62]. The resulting powders were dispersed and sonicated in ethanol and Nafion solution (binding agent) before being pipetted onto a glassy carbon (GC) disk electrode (PINE research, AFE2M050GC) and dried at room temperature to form a thin film. The geometric surface area of the GC electrode was 0.196 cm -2 . The catalyst coated CG electrode was mounted onto a RDE shaft (AFE6MB) attached to a modulated speed regulator (PINE research, MSR) and used as the working electrode in a typical three-electrode set up contained in an electrochemical cell kit (PINE research, AKCELL2). Pt mesh and Ag/AgCl (saturated KCl) were used as the counter and reference electrodes respectively.
To assess ORR activity, electrochemical characterization was determined by recording polarisation curves on a Solartron 1470E multichannel cell test system. Tests were conducted in O 2 -saturated 0.1 M KOH solution using the potential range of − 0.6-0.1 V vs. Ag/AgCl (0.36-1.06 V RHE) at a scan rate of 10 mVs -1 and various rotation rates (100, 400, 900, 1600 rpm) of the RDE. Chronopotentiometry tests were conducted in 0.1 M KOH at room temperature under a fixed potential of − 0.4 V vs. Ag/AgCl. From analysis of the ORR polarisation curves, Koutecky-Levich (K-L) plots were exploited to calculate the electron transfer number for all electrocatalysts using the following equations [42,63]: where I is the measured current density, I K and I L are the kinetic and limiting current density respectively, ω is the angular velocity of the disk, n is the electron transfer number, F is the Faraday constant (96485 C mol -1 ), C 0 is the bulk concentration of O 2 (1.2 ×10 -6 mol cm - and V is the kinematic viscosity of the electrolyte (0.01 cm 2 s -1 ).

Electrode preparation
Carbon cloth was washed and sonicated in dilute hydrochloric acid, isopropanol and deionized water. The cathode was prepared by coating the perovskite catalyst onto the pre-treated carbon cloth with a loading of 3.4 mg oxide cm -2 . The perovskite powder was physically mixed with high surface area carbon black (Vulcan XC-72R) in a weight ratio of 1:1 to facilitate electrical contact between the oxide particles and help eliminate issues relating to electronic conductivity, favouring complete utilisation of the perovskite surface [62]. Utilising propanol as the solvent, the PiperION TP-100 ionomer was ultra-sonicated in an ice-water bath for 1 h with perovskite/C (perovskite/C:PiperIon = 4:1 wt ratio) and PTFE (10 wt%) to form a uniform dispersion. The ink was then brushed onto the carbon cloth and left to dry. The loading of the electrode was 1.23 mg oxide cm -2 . For comparison, a similar method was employed to fabricate a Pt/C electrode with a loading of 0.4 mg PGM cm -2 . PtIr(40 wt%)/C(60 wt%) powder was prepared using a method described elsewhere [64]. Utilising propanol and water as the solvent, the PiperION TP-100 ionomer was ultra-sonicated in an ice-water bath for 1 h with PtIr/C (PtIr/C:PiperIon = 4:1 wt ratio) to form a uniform dispersion. The ink was then brushed onto the pre-treated carbon cloth and a loading of 2.2 mg PGM cm -2 was obtained.

Single-cell evaluation test
The test conditions for the single cell with an active area of 1 cm 2 was anode fuel: 2 mLmin -1 7 M NH 3 H 2 O at 3 bar and cathode fuel: 180 mLmin -1 CO 2 -free air fed through a humidifier at a temperature of 95 o C at 2 bar. The cell temperature was 80 and 100 o C and the corresponding polarisation curves were obtained using a Solartron 1287 A Electrochemical Station. The stability of the ammonia fuel cell was tested at fixed current density of 50 mA cm -2 at an operating temperature of 100 o C. The anodic backpressure was held at 3 bar whilst the cathodic backpressure was held at 2 bar.

Sample characterisation
The room temperature XRD patterns of the parent LCO sample as well as the singly and doubly doped counterparts are compared in Fig. 1b. Detailed lattice parameters are listed in Table 1. The XRD analysis reveals that the intense reflections of the LCO perovskite were well indexed to the rhombohedral crystal system (PDF 00-048-0123) and agrees well with literature [65,66]. The perovskite phase is successfully formed for all samples with no indications of any structural or phase changes, demonstrating that Cr and Fe are successfully introduced into the B-site.
There is a clear gradual shift in diffraction peaks to lower 2θ values on doping of Cr and/or Fe which indicates expansion of the unit cell. On introduction of Cr into the B-site of the LCO perovskite, the peaks shift to lower 2θ angles with respect to the parent structure. This can be seen with the (100) peak highlighted in Fig. 1b (right). This shift can be attributed to the larger crystal radius of Cr 3+ (0.755 Å), compared to intermediate-spin (IS) Co 3+ ions, which is presumed to be between that of low-spin (LS) (0.685 Å) and HS (0.750 Å) Co 3+ when the coordination number is 6 [42,[67][68][69]. On substitution of the B-site with Fe, there is a further obvious shift in diffraction peaks towards lower 2θ angles. This can be owed to the larger crystal radius of high-spin (HS) Fe 3+ ions (0.785 Å) compared to that of both Cr 3+ and Co 3+ , henceforth a greater expansion of the lattice can be expected to accommodate for the larger crystal radius [42,[67][68][69]. Ivanova et al. reported similar results showing lattice expansion when Fe was introduced into LaCo 1-x Fe x O 3 due to the larger ionic radius of high spin Fe 3+ ions [70]. Moreover, when the B-site is dual substituted with both Cr and Fe, the peaks are found to be situated in-between that of LFCO and LCCO. This is expected due to the lower Fe content and the smaller ionic radius of Cr 3+ in comparison. Safakas et al. showed a similar trend when the relative amount of Fe 3+ in LaFe 1-x Co x O 3-δ was decreased and a contraction effect on the unit cell was present due to the lower Fe 3+ :Co 3+ ratio [42]. These corresponding shifts reinforce the successful substitution of dopants into the B-site. Furthermore, the decrease in crystalline sizes on substitution can be related to the decrease in the unit cell volume and micro-strain [67].
Rietveld refinement of the representative LaCr 0.25 Fe 0.25 Co 0.5 O 3-δ composition was carried out by GSAS and EXPGUI [71,72]. The refined XRD pattern is shown in Fig. S1 with the lattice parameters are listed in Table 2. It has been confirmed that LCFCO exhibit a hexagonal structure with space group R-3c (167); a = 5.4786(1) Å, c = 13.2100(2) Å and V = 343.378(6) Å 3 . It should be noted that all lattice parameters are within a similar range to the parent structure and as expected, the unit cell volume of LCFCO (343.38 Å 3 ) is between that of LFCO (344.35 Å 3 ) and LCCO (341.47 Å 3 ). The refined parameters therefore support the above analysis indicating successful substitution into the B-site.
Raman spectroscopy was used to gain greater insight into the M-O bond strength for the perovskite powders. The four distinguishable bands present for LaCoO 3 in Fig. 1c coincide with those stated in literature [66,73,74]. The band that appears at the lowest Raman shift value is associated with the E g La stretching vibration [74]. There is minimal shift in this band on substitution at the B-site which indicates that the introduction of Cr and Fe dopants do not interact with the A-site but instead engage at the B-site. The second and third bands are due to the E g bending and E g quadrupole vibrations respectively. The highest energy band is attributed to the A 2 g breathing mode of the oxygen ion cage with respect to the CoO 6 octahedron [74]. These latter bands associated with CoO 6 vibrations are shown to be more sensitive to B-site substitution as expected. The A 2 g band in LCO has the lowest Raman energy (642 cm -1 ).  In the presence of B-site substitution, the A 2 g band shifts to higher Raman energies for LCCO and LFCO (681 cm -1 and 667 cm -1 respectively). The A 2 g band in LCFCO is found in-between that of the abovementioned, with a Raman energy of 678 cm -1 , revealing that the Co B-site interacts with both Cr and Fe. The absence of additional bands and the respective shifting of the A 2 g band indicates that the dopants are successfully introduced into the BO 6 octahedron, in agreeance with XRD.
The morphologies of the prepared perovskite powders were characterised by SEM. The images in Fig. 1d-g reveal globular grains are formed for all compositions which are relatively uniform in size. It should also be noted that an obvious reduction in grain size is observable as Fe is introduced into the sample. This has been reported in literature where the presence of Fe reduces the extent of particle agglomeration [70]. EDS was used to explore the different elemental compositions and a detailed analysis is reported in Table S1. Elemental mapping of the LCFCO perovskite is also provided to demonstrate visual distribution of the elements against the scale bar given in Fig. 1h. The elements of La, Fe, Cr and Co are homogenously dispersed with no apparent areas of agglomeration, this is in agreeance with the formation of a single-phase perovskite with no bulk impurities.
The nature and surface composition of the samples were studied by XPS. The spectrum of LCFCO ( Fig. 2a-d), LCO, LCCO and LFCO (Fig. S2, S3 and S4 respectively) are provided. Since the data can be fitted well to the La 2 O 3 spectra, the La species present at the A-sites are reasonably assumed to be in the + 3 oxidation state as expected, satisfying the ABO 3 composition. The deconvoluted spectra can be fitted to peaks arising from La 4d 5/2 and La 4d 3/2 [75]. The Co 2p 3/2 peaks of all samples can be fitted well to the Co 3 O 4 spectra [76]. It is stated in literature that for Co 3 O 4 , where Co(II/III) exists in the + 2 and + 3 oxidation states, the Co 2p 3/2 peak corresponding to binding energies (BE) around 780 eV can be assigned to octahedral Co 3+ , whereas the Co 2p 3/2 signal at higher BE, at around 781 eV, together with the shake-up satellite signals are indicative of surface Co 2+ [77][78][79]. The results therefore suggest the presence of Co in both the + 2 and + 3 oxidation states in all samples. Furthermore, the Co 2p 3/2 peak observed at higher BE may be closely related to the CoO species where Co is in the + 2 oxidation state [76]. The peak with lowest BE in all spectra can therefore be associated with Co in the +3 oxidation state whilst the latter two peaks at higher BE may be assigned to the + 2 oxidation state. The deconvoluted Cr 2p 3/2 spectra in LCCO and LCFCO are characteristic of Cr 6+ and Cr 3+ species [77]. The fractional percentage of Cr in the +3 oxidation state is calculated to be around 56.9% in LCFCO, which is similar to that of LCCO (61.9%) (Table S2). This demonstrates that the fraction of Cr 3+ species within the sample remains relatively unchanged in the presence of additional dopants. Furthermore, the Fe 2p 3/2 and 2p 1/2 spectra can be fitted to peaks corresponding to Fe 2 O 3 where Fe is in the + 3 oxidation state. On introduction of Cr into LCO, there is a clear shift of the Co 2p 3/2 peak position towards higher binding energies from 779.7 eV to 780.4 eV, as well as an obvious increase in satellite signal intensity, both of which are indications of strong interactions between Cr and Co [77]. There is also an increase in binding energy of the Co 2p 3/2 profile when Fe is introduced into LCO to form LFCO (779.9 eV), however, this is not as pronounced, indicating Cr may interact more strongly with the Co centre. When both Fe and Cr are introduced, LCFCO also demonstrates a shift in Co 2p 3/2 peak position towards a binding energy of 780.0 eV. These results demonstrate that in the presence of both dopants, the surface species do not change and there are no further variations in oxidation states. Through slight changes in binding energy however, it can be Space group R-3c (167); a = 5.4786(1) Å, c = 13.2100(2) Å, V = 343.378(6) Å 3 , Z=6. R wp = 9.84%, R p = 7.84%, χ 2 = 2.495. deduced that these ions interact at the B-site which may lead to synergistic effects. The surface oxygen species at the surface of the perovskite powders were also investigated by XPS (Fig. 2e). The spectra can be fitted into four subpeaks that can be assigned to the lattice oxygen species (O 2-), highly oxidative oxygen species (O 2 2-/O -), surface-adsorbed oxygen species or hydroxyl groups (OH -) and surface adsorbed water (H 2 O) labelled as O1, O2, O3 and O4 respectively [48,80]. The relative concentration of each oxygen-containing species is listed in Table S3. According to literature, the highly oxidative oxygen species (O 2 2-/O -) is closely related to surface oxygen vacancies and defects [48,[80][81][82]. Compared to LCO, incorporation of Cr in LCCO results in an increase in the O2 peak, indicating an enhancement in surface oxygen vacancies. When Fe is introduced however, this is much less pronounced. LCFCO also shows a relative increase in surface oxygen vacancy ratio.
Since LCFCO is a new perovskite, its synthesis methods have not been widely explored. The synthesis method of LCFCO was therefore studied to assess the limits and properties of the perovskite at varying calcinating temperatures. LCFCO was prepared at different temperatures of 600, 700, 800 and 900 o C labelled as LCFCO-600, LCFCO-700, LCFCO-800 and LCFCO-900 respectively. The corresponding XRD analysis for these powders is shown in Fig. 3. For comparison, the parent LCFCO (LCFCO prepared at 1000 o C) perovskite sample is also displayed. The major differences found by XRD analysis between the samples are: (i) LCFCO-700 has broader peaks compared to its counterparts indicating smaller particle size and (ii) sample LCFCO-600 contains a secondary phase which is reflected by an additional reflection at a 2θ value of 28.4 o belonging to the (040) plane of La 2 CoO 4 (PDF 04-008-9324) whereas samples LCFCO-700, LCFCO-800 and LCFCO-900 are single phase. Nevertheless, the results imply that the perovskite only becomes single phase at temperatures above 700 o C. For this reason, only LCFCO-700, LCFCO-800 and LCFCO-900 are considered as potential ORR catalysts.
To get a better idea of size-activity relationship, it is important to explore crystal sizes of the LCFCO based samples. Table 3 shows the crystallite domains for the LCFCO-700, LCFCO-800, LCFCO-900 and parent LCFCO perovskite samples calculated from both the Scherrer equation and Williamson-Hall relationship. The data reveals that there is an obvious reduction in crystalline size as calcinating temperature is reduced. These results are consistent with literature [29]. The methods above give an average value of crystalline domain size of the grains; however, real samples tend to have grains with different domains. To assess the distribution of different grain sizes and explore the morphologies of the prepared perovskite powders, SEM characterisation was used. The images in Fig. 3b-d reveal an obvious reduction in grain size as calcinating temperature is reduced, with LCFCO-700 showing a distinctively small grain size. This coincides with the analysis obtained from XRD.
The measured specific surface areas (SSA) of the synthesised perovskite powders ranged from 1.0 to 5.2 m 2 g -1 as shown in Table 3. Such SSA values are typical for perovskites due to the relatively high temperature treatment required to form single phase perovskite oxides   [29,42]. There is a clear increase in SSA as calcining temperature is reduced, which is in accordance to literature [29]. The results can be rationalised due to higher calcining temperatures resulting in the agglomeration of particles and consequently smaller exposed surface areas. The development of synthesis methods leading to reduced particle sizes, and therefore an increased SSA, could allow for the same ORR performance to be obtained at a lower cathode loading. This indicates that a smaller cathode layer thickness can be utilised and for reduced voltage losses due to mass transport resistances [42,83].

Evaluation of oxygen reduction reaction
The electrocatalytic performance of the LCO, LFCO, LCCO and LCFCO perovskite samples towards ORR were evaluated in O 2 -saturated 0.1 M KOH solution between − 0.6 and 0.1 V vs. Ag/AgCl (0.36-1.06 V vs. RHE) at a scan rate of 10 mVs -1 . The potential range was chosen to avoid possible irreversible reduction of the perovskite catalysts by avoiding more cathodic potentials being applied. For comparison, the polarisation curve for the Pt/C electrocatalyst was also collected under similar conditions. Fig. 4a illustrates the comparative current density (normalised to the geometric surface area of the electrode disk) vs. potential for the electrocatalysts with a rotation rate of the RDE equal to 1600 rpm. Furthermore, polarisation curves for the electrocatalysts over rotation rates of 100, 400, 900 and 1600 rpm are shown in Fig. 4b and Figs. S5ac. Over the entire potential range, the current density at the same potential increases upon doping at the B-site, indicating an increase in ORR activity. Though, it is difficult to draw definite conclusions without further analysis since the reduction reaction depends on several factors. An overview of the intrinsic ORR activity for electrocatalysts can be inferred from their onset potentials [20,42]. The onset potentials E onset of the perovskites were determined by intersection of the current baseline and a tangent line drawn from the polarisation curve at the end of the kinetic control region (Fig. S6a) (Fig. 4d). This value is similar to that of PGM-based electrocatalysts reported in literature, whereby ORR onset potentials more positive than 0.93 V vs. RHE have been recorded for both Pt/C (20 wt%) and Pd/C (20 wt%) in alkaline medium [20,84,85]. Notably, the onset potential of the LCFCO electrocatalyst can be deduced as being heavily influenced by the presence of Fe, since LFCO shows a similar positive shift in onset potential compared to undoped LCO. Introduction of Cr in LCCO shows a much less pronounced shift. A similar positive shift in ORR onset potential when Fe is introduced into electrocatalysts has been reported in literature [84,85].
One of the most common methods for evaluating catalyst performance is to compare half-wave potentials E 1/2 . This is particularly useful as it allows for PGM-free electrocatalysts to be compared with Pt catalysts in terms of their half-wave potential [86]. The E 1/2 values of the perovskites were determined by the potential recorded at the midpoint of the polarisation curves; the point at which the reduction current begins rapidly increasing and saturation in the limiting region (Fig. S6b). On this basis, the half-wave potential of the electrocatalysts follows the descending order LFCO > LCFCO > LCCO > LCO with E 1/2 values of 0.87 > 0.86 > 0.71 > 0.68 V vs. RHE. Despite the E 1/2 value for LFCO being slightly greater than that of LCFCO, the change is minimal, and both show a substantial difference compared to the undoped LCO. The half-wave potential value of LCFCO is comparative to Pt/C and other PGM-free electrocatalysts reported in literature, which fall in the range between 0.53 and 0.87 V vs. RHE [84,[86][87][88][89][90][91]. Furthermore, Pt/C in this study showed a slightly lower half-wave potential of 0.85 V vs. RHE (Fig. 4e) [84]. This illustrates the great potential of LCFCO as a PGM-free electrocatalyst for ORR.
To gain further insight into the ORR kinetics, Tafel plots were constructed and fitted for the perovskite electrocatalysts (Fig. 4 f). The Tafel slope for LCFCO (93 mVdec -1 ) is smaller than LFCO, LCCO and LCO (99, 101 and 133 mVdec -1 respectively), indicating that the incorporation of the Cr and Fe dopants into LCO induces faster electrokinetics.
As well as improved onset potential and half-wave potential, the LCFCO electrocatalyst also shows an enhanced current density at the same potential over the same range compared to the undoped and singly doped counterparts (Fig. 5a). The enhanced ORR activity cannot therefore be merely described by differences in E onset and E 1/2 . Fig. 5b shows Koutecky-Levich (K-L) plots for the electrocatalysts at an applied potential of − 0.55 V vs. Ag/AgCl (0.41 V vs. RHE) using polarisation data obtained in Fig. 4b  that the presence of Cr plays a key role in increasing the number of electrons involved in the ORR mechanism and therefore play a key role in the improved current density of LCFCO compared to undoped LCO. This may be partly attributed to the oxygen species present on the surface of the perovskite [85]. Oxygen vacancy formation at the perovskite surface has been reported to promote oxygen adsorption and charge transfer [21,42,82]. Several studies have reported the beneficial effects of surface oxygen vacancies on the ORR activity of perovskites as well as selectivity to the 4ereduction pathway [21,42,82,[92][93][94][95]. XPS analysis in Fig. 2e reveals that the mass percent of O 2 2-/O -(O2) is higher in both LCCO and LCFCO than in LCO. As mentioned above, literature shows that this highly oxidative oxygen species (O 2 2-/O -) is closely related to surface oxygen vacancies and defects [48,[80][81][82]. This can therefore be used as an indication of oxygen vacancy content present within the samples studied. The respective ratio of O2/(O1 +O3 +O4) was subsequently calculated to give an estimated oxygen vacancy ratio and is visually displayed in Fig. 5d [48,80,81]. On introduction of Fe, LFCO displays a lower area ratio (9%), which is indicative of a lower oxygen vacancy ratio, and minimal change in current density at the same potential over the same potential range compared to LCO. Similar trends in oxygen vacancy formation have been observed in literature [21,42,96].  [81]. It is therefore deduced that the presence of the Cr introduced more oxygen defects, in turn enhancing OER activity and the interaction of adsorbed oxygen-containing species. It can subsequently be assumed that the presence of Cr doping can lead to oxygen vacancies which increase ORR activity and selectively favour the 4ereduction pathway. Apart from electrocatalytic activity, stability is another important factor for advanced electrocatalysts. The long-term stability of the LCFCO was confirmed by chronopotentiometry tests in O 2 -saturated 0.1 M KOH solution at a fixed potential of − 0.4 V vs. Ag/AgCl with a RDE rotation rate of 1600 rpm. As shown in Fig. 6a, LCO shows a decrease in stability over the testing duration. However, with the doped perovskites and particularly the LCFCO electrocatalyst, there is an initial slight decrease during testing which then stabilises over the duration. The relative current density of LCFCO reaches around 94.5% throughout the testing period, indicating that the electrocatalyst remains stable over an extended period. Notably, the stability of LCFCO is slightly higher than that of Pt/C over the same testing duration and conditions (Fig. 6b). Similar stability of Pt/C (20 wt%) electrocatalysts has been observed in literature, where a retention of 93.5% was observed after 10000 s in O 2saturated 0.1 M KOH solution with a RDE rotation rate of 1600 rpm [85]. The results indicate that LCFCO is a highly active and stable electrocatalyst towards ORR.
As mentioned in previous studies, ORR on perovskite oxides in alkaline media is complicated and the reaction depends on several factors such as intrinsic activity, conductivity and surface adsorption properties, all of which should not be considered independently [20,42]. Changes in electronic structure can affect the rate determining steps and the overall ORR activity. Suntivich et al. recently proposed that the electronic configuration and e g orbital filling at the B-site of the perovskite can effectively be used as an activity descriptor for ORR [23,57]. It was determined that a maximum ORR activity was observed when the e g -filling was close to 1. This concept has been applied to pioneer successful perovskites electrocatalysts towards ORR activity such as LaMn 0. 3 [48,49,[97][98][99].
Since the exact spin states of the transition metals in LaCr 0.25 Fe 0.25-Co 0.5 O 3-δ have not been determined prior to this study, an approximation of the e g filling based on literature and XPS data provided in Table S2 has been made. Given that the B-site metal exhibits different valence states, the electronic filling of the e g orbital becomes complex. It is often widely accepted that Co 2+ (d 7 ) ions exist in high spin states (e g = 2), whereas the spin state of Co 3+ in LaCoO 3 has proven to be more difficult to determine. It is often accepted that Co 3+ exists in an intermediate-spin (IS) state at temperatures above 70-90 K, rather than the Hund's predicted high-spin (HS) state (t 2 g 4 e g 2 ) [66,100,101]. It is due to this phenomenon that Co 3+ in LaCoO 3 is assumed to have unique electronic properties where the e g occupancy is ~1. Although there is still extensive debate over Co spin state, Co 3+ is assumed to exist in an intermediate spin state (e g ~1) for the simplicity of the calculations in this study [100][101][102][103][104]. Based on literature, Fe 3+ tends to favour high spin configuration (t 2 g 3 e g 2 ), leading to an e g filling of 2. Furthermore, XPS data also reveals that Cr in LCCO and LCFCO exists in the + 6 and + 3 oxidation state, with the latter being dominant. For e g calculations, the effect of Cr 6+ are neglected since there are no d electron orbitals and only Cr 3+ (t 2 g 3 e g 0 ) is considered, leading to an e g filling of 0. The average filling of electrons in the e g orbital (n‾) can be calculated based on Eq. (S4) [48]. From the calculations, it can be seen that the e g filling of LCO was around e g = 1.5 and approaches closer to 1 when doped by both Cr and Fe in LCFCO where e g = 1.16. The near unity e g -filling found in LCFCO may also partially explain the enhanced ORR activity compared to its undoped and singly doped counterparts. Furthermore, the electrocatalytic performance of the LCFCO-700, LCFCO-800 and LCFCO-900 perovskite samples towards ORR were evaluated in O 2 -saturated 0.1 M KOH solution between − 0.6 and 0.1 V vs. Ag/AgCl (0.36-1.06 V vs. RHE) at a scan rate of 10 mVs -1 (Fig. 7a and Figs. S7a-b). The potential range was chosen to avoid possible irreversible reduction of the perovskite catalysts by avoiding more cathodic potentials being applied. For comparison, the polarisation curves for the parent LCFCO perovskite sample and Pt/C electrocatalyst are also displayed in Fig. 7b and c respectively. These improvements in performance may be attributed to the enhancement in the number of exposed active sites due to a higher SSA. Similar trends have been seen by Retuerto et al. who found that perovskite catalysts 700-LaNiO 3 and 800-LaNiO 3 , which were calcinated at 700 and 800 o C respectively, showed more positive E onset potentials than 900-LaNiO 3 and 1000-LaNiO 3 , which were calcinated at 900 and 1000 o C respectively [29]. Moreover, Fig. 7c shows that LCFCO-700 has similar enhancement in E onset and E 1/2 potentials compared to commercial Pt/C. Furthermore, the current density value is remarkably similar to the commercial Pt/C electrocatalyst, illustrating that LCFCO-700 is a promising candidate for ORR. Amongst AEMFCs, direct ammonia fuel cells (DAFCs) have becoming increasingly popular due to the fuel's high energy density, large-scale global production, extensive existing infrastructure, and low cost per unit energy [1,[31][32][33][34][35][36]38]. These systems employ electrocatalysts at the anodic and cathodic site that can assist the ammonia oxidation reaction (AOR) and oxygen reduction reaction (ORR) respectively. The choice of a cathode catalyst for DAFCs has been directed by: (i) the desire to reduce use of PGMs and (ii) a minimal impact on ORR activity in the presence of ammonia [105]. The latter stems from the acknowledgement that ammonia crossover to the cathode is common in existing in DAFCs based on polymeric membrane electrolytes [1,105]. This study has shown how the first directive is satisfied; avoiding reliance on PGMs and making use of a cheaper, more readily available LCFCO-700 perovskite oxide that shows comparable ORR activity to commercial Pt/C. To satisfy the second requirement, that being a minimal impact on ORR activity in the presence of ammonia, the LCFCO-700 perovskite oxide was tested in O 2 -saturated 0.1 M KOH solution with and without the presence of 0.1 M NH 3 to mimic the effects of ammonia fuel cross over in a fuel cell. Fig. 7d shows that on introduction of 0.1 M NH 3 , there is negligible change in ORR activity for the LCFCO-700 perovskite oxide, indicating that the ORR activity is not affected by the presence of a small amount of ammonia. It can therefore be reasonably assumed that there will be minimal negative effect of ammonia cross-over if LFCCO-700 were to be employed as a cathode in a DAFC, making it a suitable and promising non-noble cathode candidate.

Performance of direct ammonia fuel cells
Due to its promising activity towards ORR and minimal change in the presence of ammonia, LCFCO-700 was considered as a suitable cathode in a DAFC. To assess the performance of the perovskite oxide, a catalyst ink was prepared and brushed onto carbon cloth to form an electrode. PtIr/C was used as an anode and PiperION-A20-HCO3 TP-85 was utilized as the membrane to form a membrane electrode assembly (MEA). Fig. 8a presents the polarisation curve recorded for the corresponding DAFC employing an anode feed consisting of 7 M NH 4 OH + 1 M KOH and a cathode feed of CO 2 -free air at different operating temperatures. For comparison, a fuel cell employing a Pt/C cathode was also investigated under similar conditions ( Fig. 8b and Fig. 8c).
The effects of operating temperature on fuel cell performance were investigated and are shown in Fig. 8a. The data shows that as temperature increases, there is an obvious increase in overall performance of the DAFC. When temperature was increased from 20 to 40 o C in the DAFC employing LCFCO-700 as a cathode, a notably increase in OCV was observed from 0.3 to 0.6 V. However, as temperature increased above 60 o C, there was minimal change in OCV, with values of 0.69, 0.71 and 0.72 V being obtained at 60, 80 and 100 o C respectively. There is also a substantial increase in current density as operating temperature is increased, with a maximum current density of 379 mAcm -2 and peak power density of 34 mWcm -2 being obtained at 100 o C. This positive increase in DAFC performance with an increase in operating temperature has commonly been observed in literature [106,107]. Ishiyama et al. reported the effects of temperature on a DAFC employing PGM-based PtRu/C at both the anode and cathode site [107]. At an operating temperature of 50 o C, an OCV and maximum current density of around 0.45 V and 40 mAcm -2 was respectively obtained when using 5 M NH 4 OH + 1 M KOH as the anodic fuel and air as the cathode stream. A corresponding PPD of 3 mWcm -2 was obtained. Under similar operating temperatures of 60 o C, the DAFC shown in Fig. 8b, which utilises similar anodic/cathodic oxidants (7 M NH 4 OH + 1 M KOH/air), demonstrates a maximum current density and PPD of 145 mAcm -2 and 11 mWcm -2 respectively. This reinforces the feasibility of LCFCO-700 to replace PGM-based cathodes in DAFCs, showing their ability to obtain enhanced performances. Furthermore, when the operating temperature was increased in 80 o C in the aforementioned study, Ishiyama et al. observed that the OCV and maximum current density increased to around 0.55 V and 48 mAcm -2 respectively. A corresponding PPD of 4.5 mWcm -2 was obtained for the cell. At the same operating temperature, a maximum current density and PPD of 317 mAcm -2 and 30 mWcm -2 were respectively obtained in this study. These results illustrate the promising performance of LCFCO-700 as an active and promising cathode under various operating conditions. Moreover, a closer comparison between Pt/C and LCFCO-700 as cathodes in the DAFC at operating temperatures of 80 and 100 o C are shown in Fig. 8b and Fig. 8c. It should be noted that different loadings were used between the PGM and non-PGM cathodes however, given the high expense of Pt/C as a cathode material, along with the cheap and easy synthesis nature of the LCFCO-700 perovskite, the latter is a promising candidate to replace Pt/C. This would significantly reduce cost of devices such as fuel cells and batteries which rely heavily on PGMs to assist ORR. Nevertheless, the two DAFCs demonstrate very similar performances and power densities and are investigated in further detail below.
The results show an OCV of 0.71 V was achieved when LCFCO was employed as the cathode whereas a lower OCV of 0.60 V was obtained when Pt/C was used. The latter value is common for a DAFC employing Pt/C as a cathode material and is therefore expected [31]. Moreover, a maximum current density of 317 mAcm -2 and a peak power density (PPD) of 30 mWcm -2 was achieved at an operating temperature of 80 o C when LCFCO was employed as the cathode. Remarkably, this performance was very similar to the DAFC employing Pt/C as the cathode, which obtained a maximum current density of 285 mAcm -2 and a peak power density of 32 mWcm -2 under similar operating conditions. When the temperature was increased to 100 o C, an observable enhancement in performance for both fuel cells was observed. Again, the performance was extremely similar, with the DAFC employing LCFCO as the cathode obtaining a maximum current density of 379 mWcm -2 and a peak power density of 34 mWcm -2 , and the DAFC employing Pt/C obtaining a maximum current density of 345 mWcm -2 and a peak power density of 40 mWcm -2 .
Under closer observation, there is a greater potential drop in the low current density regime of the LCFCO-700-based DAFC compared to that of the Pt/C. This region is dominated by the activation overpotential (η a ), which is closely related to: (i) surface conversions preceding the electron transfer such as adsorption on the electrode surface and chemical reactions which are inherit to the catalyst, (ii) electron transfer at the electrode surface of the electrode-electrolyte interface and/or (iii) surface conversions following the electron transfer such as desorption from the electrode surface and chemical reactions which are inherit to the catalyst. Based on these contributing factors, the larger activation overpotential may be due to the surface conversions that occur on Pt compared to LCFCO-700. This is inherent to the catalyst employed and indicates that the Pt surface may facilitate more facile conversion mechanisms and thus lower the activation overpotential associated with the reaction. Furthermore, the electron transfer step plays a key role in this type of overpotential [108]. Since perovskite oxides have inherently lower electrical conductivity than that of platinum, this explanation is plausible and reinforces the need for good electrical contact between the oxide particles [21].
A list of comparative DAFC performances are shown in Table S4. It should be recognised that key studies demonstrated high DAFC performances employ pure O 2 as the oxidant [31,106]. In this study however, CO 2 -free air (20% oxygen) is used as the cathodic oxidant instead of pure O 2 which provides a much safer fuel, particularly for applications such as transport where consumer safety is of upmost importance [1]. Nevertheless, a relatively high current density and high peak power density is still obtained when air is used despite the oxygen content being much lower than pure oxygen. Remarkably, the results in this study are the highest reported for DAFCs employing air as the cathode stream.
Furthermore, the stability of the DAFCs based on LCFCO-700 and Pt/ C cathode catalysts were tested in a preliminary durability test at a continuous current density of 50 mA cm -2 (Fig. 8d). The results show that the two cells both demonstrate a steady decrease in voltage over the testing duration and follow the same overall trend, implying comparable nature between the two DAFCs. The fluctuations recorded in the data are owed to ammonia cross over from the anode and cathode, which subsequently lowers voltage over time [1109]. This was visually observed in the outlet of the cathode stream and may partially be owed to employment of a thin polymeric membrane [1].
The results demonstrate that the LCFCO-700 perovskite oxide is an excellent, cheap catalyst for ORR and shows practical applicability in fuel cell devices.

Conclusions
A series of LaCoO 3-δ (LCO) based perovskite oxides including LaCr 0.5 Co 0.5 O 3-δ (LCCO), LaFe 0.5 Co 0.5 O 3-δ (LFCO) and LaCr 0.25 Fe 0.25-Co 0.5 O 3-δ (LCFCO) were prepared by a simple sol gel method and their oxygen reduction reaction (ORR) activity was studied in 0.1 M KOH solution at room temperature using a rotating disk electrode (RDE) technique. The highest and lowest ORR activities amongst the electrocatalysts were exhibited by LCFCO and LCO respectively, with the performance of LCFCO being comparable to commercial Pt/C at a much lower cost. The presence of both Cr and Fe were found to play vital roles in enhancing activity, influencing structural and electronic properties respectively. The observed changes in ORR activity of the perovskite electrocatalysts upon doping of Cr was associated with induced changes in the surface oxygen vacancy formation and an increase in the number of transferred electrons to 4, signifying a gradual change in ORR pathway. When introducing Fe into the B-site of the perovskite oxide, a clear enhancement of onset potential E onset was observed in order of LCFCO > LFCO > LCCO > LCO with values of 0.97 > 0.95 > 0.83 > 0.80 V vs. RHE respectively. The half-wave potential of LCFCO was also superior (0.86 V vs. RHE) compared to undoped LCO (0.68 V vs. RHE). Furthermore, the effects of calcining temperatures were explored and LCFCO fired at 700 o C (LCFCO-700) was found to give superior activity, with a notable shift in E onset to more positive potentials and the ability to obtain excellent performance at a lower synthesis temperature. Remarkably, when LCFCO-700 was employed in an ammonia-air fuel cell, an OCV of 0.72 V and a maximum current density of ~320 mAcm -2 was achieved, which are comparable to the DAFCs using Pt/C as the cathode, indicating LaCr 0.25 Fe 0.25 Co 0.5 O 3δ /C is a promising non-noble cathode for low temperature fuel cells.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
Data will be made available on request.