High Performance FeNC and Mn-oxide/FeNC Layers for AEMFC Cathodes

The synthesis of pure α‒, β‒, δ‒MnO₂ and α‒Mn₂O₃ phases was performed according to literature reports, and as detailed in our recent study.⁸⁵ A brief description is given here for each phase. The α‒MnO₂ phase was obtained by reducing KMnO₄ in ultrapure water (18 MΩ) and fumaric acid at room temperature. The resulting gel was filtered, washed with ultrapure water and dried.²⁵ δ‒MnO₂ was obtained via the reduction of KMnO₄ in a mixture of water, H₂SO₄ and ethanol, also at room temperature.⁸⁶ To obtain β‒MnO₂, the dried δ‒MnO₂ powder was calcined at 450 °C. For preparing α‒Mn₂O₃, γ-MnOOH was prepared in a first step, and then calcined at 550 °C for 12 h.⁸⁷ For the first step, γ-MnOOH was synthesized dissolving Mn(CH₃COO)₂·4 H₂O and KMnO₄ in deionized water, the solution refluxed for 12 h and then washed with water and dried.⁸⁸

Fe_0.5-NH₃ was prepared via the sacrificial metal-organic framework method.⁴³ Commercial ZIF-8 (Basolite^® Z1200, Sigma Aldrich), 1,10-phenantroline (≥99%, Sigma Aldrich) and iron (II) acetate (≥99.99%, Sigma Aldrich) were mixed in weight ratios of 4/1 for ZIF-8/phenanthroline and 0.5 wt% of iron in the complete catalyst precursor, using planetary ball milling (400 rpm, four cycles of 30 min milling with 5 min pause between each cycle). The catalyst precursor powder was then collected from the jar, transferred in a quartz boat and inserted in a quartz tube. For the first pyrolysis, the oven was pre-equilibrated at 1050 °C for 2 h while a continuous flow of Ar passed through the tube, with the quartz boat and catalyst precursor inside the tube but outside the heating zone. After 2 h, the quartz boat was pushed in the heating zone in three steps of 30 s ("flash pyrolysis"), with the help of an outer magnet and a quartz rod with magnet attached at one end, located inside the tube. The catalyst precursor was pyrolysed in Ar at 1050 °C for exactly 1 h, then the split-hinge oven opened and the tube removed and let to cooldown for 20 min, still under Ar flow. A second pyrolysis was then performed, following the above-mentioned procedure, but using pure NH₃ instead of Ar, an oven temperature of 950 °C and a pyrolysis duration of 5 min. The final product is collected directly from the quartz boat (Fe_0.5–NH₃).

Each Mn-oxide phase was mixed with Vulcan carbon black in a weight ratio of 1/4 by manual grinding. The Mn-oxides were mixed with Fe_0.5-NH₃ with the same procedure and weight ratio.

The crystalline structure of the four manganese oxides was verified using X-ray Powder Diffraction (XRPD) with PANAlytical X'pert diffractometer in Bragg-Brentano configuration, using a CuKα source (λ = 1.5406 Å) in a 2θ range of 5°–80° with a step size of 0.035°. The most intense peaks were compared to the literature using PANalytical X'Pert Highscore Plus (version 3.0e). To analyse the morphology of the Mn-oxides, Fe_0.5-NH₃ and α-Mn₂O₃/Fe_0.5-NH₃, scanning electron microscopy (SEM) was applied (Hitachi S4800). Specific surface area of the four Mn-oxides was determined using the Brunauer–Emmett–Teller (BET) method, cleaning previously the sample with flowing nitrogen at 200 °C for 5 h and then measuring the sorption of nitrogen at 77 K (Micromeritics ASAP 2020). Elemental distribution in the α-Mn₂O₃/Fe_0.5-NH₃ composite was analysed with SEM (ZEISS EVO HD15) coupled with Energy-dispersive X-ray spectroscopy (EDX, Oxford Instruments X-Max ^N SDD Detector).

The ORR activity and HO₂⁻ production during ORR were measured with a RRDE setup (Pine Instruments) in 0.1 M KOH, using a three-electrode configuration. The reference electrode was a Pt-wire immersed in a H₂-saturated electrolyte separated from the main compartment by a fritted glass, acting as a reversible hydrogen electrode (RHE). The counter electrode was a graphite rod immersed in the electrolyte, at a fixed distance from the working electrode. The latter was a glassy carbon tip (5.6 mm of diameter) where the catalyst is deposited. The peroxide produced at the working electrode was detected at the Pt ring. The HPRR activity was measured in a RDE setup (Pine instruments) in the same conditions (glassy carbon tip of 5 mm diameter). The inks were prepared adding in sequence 54 μl of Nafion^® (5% perfluorinated resin solution), 744 μl of ethanol and 92 μl of ultrapure water to 5 mg of FeNC, 20%-MnO_x/C or 20%-MnO_x/FeNC. The mixture was then sonicated for 1 h for homogenization. Then, 8.8 μl of the dispersion was deposited on the RRDE-GC tip, while 7 μl are drop cast on RDE-GC tip, and dried in air at room temperature, to obtain a total catalyst loading of 0.2 mg·cm⁻². For measuring the ORR activity of the Vulcan carbon black (used for Mn-oxides), an aliquot of 7.0 μl was drop cast on the RRDE-tip and 5.6 μl on the RDE-tip, in order to reach a loading of 0.16 mg cm⁻². This corresponds to the loading of Vulcan in the active layers of 20%-MnO_x/C. The electrochemical surface area of the Mn-oxides was assessed on 20%-MnO_x/C composite layers in N₂-saturated electrolyte by cycling at 10 mV mV·s⁻¹ and under 1600 rpm rotation the potential between 0.45 and 1 V vs RHE (SP-300, BioLogic Potentiostat). The lower range potential of 0.45 V was chosen to minimize Mn leaching.⁸⁵ To measure the ORR activity and selectivity, the solution was saturated with O₂ and the potential scanned in the range 0−1 V vs RHE with a scan rate of 1 mV s⁻¹, while the potential of 1.2 V vs RHE was applied to the Pt ring. To measure the activity toward HPRR, a RDE configuration was used, and the measurement performed in N₂-saturated 0.1 M KOH with 2 mM H₂O₂, scanning the potential at 1 mV·s⁻¹ in the range 0.45–1.0 V vs RHE. During ORR and HPRR measurements, the scan rate of 1 mV·s⁻¹ was sufficiently low to neglect non-Faradaic currents.

To evaluate the stability of the α-Mn₂O₃/Fe_0.5–NH₃ catalyst, a custom made polycarbonate SFC coupled with ICP-MS (Perkin Elmer, NexION 350X) is used to measure in situ the dissolution of manganese and iron during electrochemical protocols. Electrochemical measurements have been carried out in a three electrode configuration using a graphite counter electrode, Ag/AgCl reference electrode (Metrohm) and glassy carbon as working electrode. Electrochemistry is applied with a Potentiostat (Gamry, Reference 600) in 0.05 M NaOH (Merck, Certipur), reaching a controlled pH (Mettler Toledo, SevenExcellence) of 12.7, used to convert the potentials of the Ag/AgCl to the RHE.

The catalytic inks were prepared following the procedure described in Omasta et al., manually grinding the catalyst and ETFE (ethylene tetrafluoroethylene) powder ionomer,⁸⁹ and then dispersing the solids in a mixed solvent of H₂O and 1-propanol (10 vol% H₂O). All anodes were prepared from a 60% PtRu/C commercial catalyst (Johnson Matthey, 2:1 Pt:Ru mass ratio), mixed with Vulcan to reach 40 wt% PtRu/C. 150 mg of 40 wt% PtRu/C and 37.53 mg of ETFE were weighed to prepare the anode ink (ionomer/carbon ratio 0.41). Pt/C cathodes were prepared from a 60 wt% Pt/C commercial catalyst (Johnson Matthey) mixed with Vulcan to reach 40 wt% Pt/C. The catalyst and AEI mass were the same as used for the anode ink. For Fe_0.5–NH₃ cathodes, 82.59 mg of catalyst was mixed with 17.35 mg ETFE AEI (ionomer/carbon ratio 0.21). For α-Mn₂O₃/ Fe_0.5–NH₃ cathodes, 20.91 mg α-Mn₂O₃, 82.59 mg Fe_0.5–NH₃ and 17.35 mg ETFE were mixed together. The dispersion was then sonicated in an ice bath for 1 h and then sprayed on a gas diffusion layer (Toray 60, 5 wt% PTFE wet-proofing) using an airbrush (Iwata Eclipse HP CS). The obtained gas diffusion electrodes and ETFE membrane (50 μm thick in fully swollen conditions) were then soaked for 20 min in 1 M KOH, and this was repeated two times. The electrodes and membrane were combined to create Membrane Electrode Assemblies (MEAs) without prior hot-pressing in single-cell Scribner fuel cell hardware using Teflon gaskets, with the gasket thickness chosen to reach 25% compression. The AEMFC was operated using a Scribner 850e Fuel cell test system, flowing H₂/O₂ at 1.0 l min⁻¹ with a cell temperature of 60 °C. No back pressure was applied on anode and cathode sides. The break-in was performed at 0.5 V, adjusting the relative humidity (RH) at both electrodes.^34,35 AEMFC tests were carried out with three different cathodes: (a) 40 wt%Pt/C, with a loading of 0.45 mg_Pt cm⁻² (total catalyst loading of 1.125 mg cm⁻²); (b) Fe_0.5–NH₃; and c) α-Mn₂O₃/Fe_0.5–NH₃ (20 wt% α-Mn₂O₃). Both (b) and (c) had a total catalyst loading of 1.5 mg cm⁻². All AEMFC experiments were carried out using 40 wt%[PtRu]/C at the anode side (Pt:Ru 2:1 mass ratio) with a loading of 0.9 mg_PGM cm⁻².

The XRPD patterns shown in Fig. 1a demonstrate the formation of pure or almost pure Mn-oxide phases. All diffraction peaks could be assigned to the targeted oxide phase (Fig. S1 is available online at stacks.iop.org/JES/167/134505/mmedia), except for the α-Mn₂O₃ pattern which also shows minor peaks that can be assigned to α-MnO₂ (5% of α-MnO₂ was estimated,⁸⁵). Figure 1a also underlines the higher crystallite dimension of α-Mn₂O₃ compared to MnO₂ structures. The δ-MnO₂ phase, especially, shows broad peaks indicating small crystallite. The high crystallinity of the present α-Mn₂O₃ material may be related to the high temperature of 550 °C at which it was formed (see Experimental section), while the higher crystallinity of β-MnO₂ relative to δ-MnO₂ may also be explained from the viewpoint of the synthesis temperature, namely room temperature for δ-MnO₂ and then calcination at 450 °C to obtain β-MnO₂. Figure 1b shows the N₂ adsorption isotherms of the Mn-oxide materials, revealing the coexistence of micro and mesopores. The isotherm of δ-MnO₂ shows the highest amount of N₂ adsorbed, especially at low P/P₀ and also a strong hysteresis closing at P/P₀ of ca. 0.4. The shape of the isotherms of β-MnO₂ and α-MnO₂ are similar up to P/P₀ = 0.7, with the one for β-MnO₂ vertically translated by ca. +10 cm³·g⁻¹ due to a higher amount of N₂ adsorbed at low P/P₀. The adsorption isotherm for α-Mn₂O₃ clearly shows a much lower amount of N₂ adsorbed relative to all others, as well as little hysteresis, suggesting it is mostly non-porous. The BET specific surface areas are, in increasing order, 14 (α-Mn₂O₃), 98 (α-MnO₂), 136 (β-MnO₂) and 214 (δ-MnO₂) m²·g⁻¹. In line with the very different BET areas, SEM micrographs provide evidence of substantial morphological differences. Macroscopically, δ-MnO₂ and β-MnO₂ share a common particle morphology, with a size in the range of 300–600 nm (Figs. 2a, 2c), in line with the fact that β-MnO₂ was simply derived from δ-MnO₂ via calcination. However, the high resolution SEM micrographs show that the δ-MnO₂ surface has a higher roughness, with a sand-rose morphology and chambers of 30–60 nm, which partially collapsed in the case of the β-MnO₂ particles (Figs. 2b, 2d). The macroscopic morphology of α-MnO₂ is different, with large chunks that are several μm in size, composed of numerous well-defined rod-shaped sub-structures with ∼130–300 nm length and 30–60 nm width, and also of even smaller objects with less-defined particle-like morphology (Figs. 2e, 2f). Last, α-Mn₂O₃ has an molten-like, interconnected particle morphology with a smooth surface and an average particle dimension of ∼150 nm (Figs. 2g, 2h), in line with the higher temperature of 550 °C used for its synthesis. The BET area ranking can be reasonably well explained from the surface roughness as identified by the higher magnification SEM images. The highest BET area was measured for δ-MnO₂ (214 m²·g⁻¹) with highest surface roughness (Fig. 2b), then slightly decreased BET area of 136 m²·g⁻¹ for β-MnO₂ due to partial collapse of surface tortuosity (Fig. 2d), still a high surface area of 98 m²·g⁻¹ for α-MnO₂ with smooth but small rod substructures (Fig. 2f) and, last, drastically lowered BET area of 14 m²·g⁻¹ for α-Mn₂O₃ due to interconnected and molten particles morphology (Fig. 2h).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To further verify the apparent correlation between surface roughness and BET area, we calculated the expected size of Mn-oxide nanostructures from their measured BET areas and assuming either a slab, rod, or spherical geometry (depending on samples) for the non-porous oxide substructures assumed to contribute mainly to the surface area of the Mn-oxides (Table I). For δ-MnO₂ and β-MnO₂, the assumed slab geometry results in an estimated slab thickness of ca. 3 nm, in line with the very thin walls of the sand rose morphology observed by SEM.

For α-MnO₂, the assumed rod morphology results in ca. 9 nm diameter for the rods, which is significantly smaller than observed by SEM. Therefore, the surface of α-MnO₂ seems to result from a contribution of the observed rods but also from the smaller and less defined objects seen in the SEM images (Fig. 2f). Last, for α-Mn₂O₃, the estimated particle size of ca. 95 nm is in line with the smooth particles observed in SEM images (Fig. 2h). Due to these large morphological differences, the apparent activity that will be measured for a given loading of Mn-Oxides during RDE measurements needs to be normalized to the active area, which will be approximated by the BET area.

The HPRR activity of the four Mn-oxide polymorphs mixed either with Vulcan (Fig. 3a) or Fe_0.5-NH₃ (Fig. 3b) was investigated by RDE in deaerated 0.1 M KOH with 2 mM HO₂⁻. The polarization curves seen in Figs. 3a–3b show a cathodic branch that can mainly be assigned to HPRR (Eq. 1), and an anodic branch that can be assigned to the hydrogen peroxide oxidation reaction (HPOR) (Eq. 2). In addition, a minor fraction of the cathodic current might also be related to the ORR, with O₂ being produced in situ via non-electrochemical disproportionation of peroxide on the Mn-oxide surface (Eq. 3).

$$\begin{eqnarray}&&\begin{array}{l}{{{\rm{HO}}}_{2}}^{-}+{{\rm{H}}}_{2}{\rm{O}}+2{{\rm{e}}}^{-}\to 3{{\rm{OH}}}^{-};{{\rm{E}}}^{0}\sim {\rm{1.73}}\,{\rm{V}}\,{\rm{vs}}\,{\rm{RHE}}\\ \,{\rm{at}}\,{\rm{pH}}\,13\end{array}\end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray}&&\begin{array}{l}{{{\rm{HO}}}_{2}}^{-}+{{\rm{OH}}}^{-}\to {{\rm{O}}}_{2}+{{\rm{H}}}_{2}{\rm{O}}+{{\rm{2e}}}^{-};{{\rm{E}}}^{0}\sim 0.72\,{\rm{V}}\,{\rm{vs}}\,{\rm{RHE}}\\ \,{\rm{at}}\,{\rm{pH}}\,13\end{array}\end{eqnarray} \tag{ 2 }$$

$$\begin{eqnarray}&&2\,{{{\rm{HO}}}_{2}}^{-}\to 2{{\rm{OH}}}^{-}+{{\rm{O}}}_{2}\end{eqnarray} \tag{ 3 }$$

Zoom In Zoom Out Reset image size

Considering the electrochemical reactions, it is important to note that reaction 2 is not the reverse of reaction 1, and the HPRR and HPOR have therefore very different equilibrium potentials, at pH 13 (see the equations above).⁹⁰ This expresses that, in the entire potential region between 0.72 and 1.73 V vs RHE, HO₂⁻ should be highly unstable and could, theoretically, be at the same time electro-oxidized to O₂ and electro-reduced to OH⁻. In practice, however, a net oxidation current is generally observed at potentials above 0.8–0.9 V, depending on the electrocatalyst investigated, and assigned to HPOR.^83,91 The oxidation reaction (2) may be kinetically much faster than the reduction reaction (1), explaining why a net positive current is observed on all types of catalysts in that potential region. A net reduction current is observed for MnO_x/Vulcan only when the HPOR becomes thermodynamically unfavored, i.e. at potentials lower than 0.83–0.85 V vs RHE (Fig. 3a). At those conditions, peroxide anions at the electrode surface can no longer be oxidized, and become available for HPRR. No clear redox process associated with Mn-oxides is observed near the open circuit potential of 0.83–0.85 V vs RHE, except for δ-MnO₂, as shown by cyclic voltammetry in a N₂-saturated electrolyte free of peroxide (Fig. S2). Therefore, the experimental values of open circuit potential (OCP) should be close to the equilibrium potential of reaction (2). With 2 mM HO₂⁻ in the electrolyte and assuming an activity of 1 for O₂ (i.e. assuming that reaction 3 can saturate with O₂ in a very thin layer of electrolyte near the catalyst surface), one can calculate from the Nernst equation that the equilibrium potential for reaction (2) should be 0.801 V vs RHE, assuming a pH of 13. This is close to the experimentally observed values of OCP, ranging from 0.83 to 0.85 V vs RHE for δ-MnO₂, β-MnO₂, α-MnO₂ and α-Mn₂O₃, in the order of increasing OCP values. Looking at the cathodic branch, assigned mainly to HPRR, the current density reaches different values of plateau depending on the Mn-oxide considered (Fig. 3a), which is in contradiction with a same diffusion-limited current density expected for the two-electron HPRR. During the HPRR, the current density plateau is highest for α-Mn₂O₃, followed by β-MnO₂ and δ-MnO₂, and lowest for α-MnO₂. It was shown by Ryabova et al. that this limiting current density can be assigned entirely to a rate-determining chemical step in the HPRR on Mn₃O₄ and MnOOH (with the limiting current density showing no variation with the rotation rate), while β-MnO₂ was in an intermediate situation in which the limiting current density slightly varied with rotation rate, but much less than expected from the Levich equation.⁸³ In contrast, the limiting current density of HPRR on α-Mn₂O₃ obeyed the Levich equation and increased proportionally with the square root of the electrode rotation rate. The authors concluded that for MnOOH and Mn₃O₄ (as well as β-MnO₂, partially) the rate of the HPRR at low potential is not controlled by the diffusion of HO₂⁻ species toward the electrode, but by a slow chemical step. The latter was proposed to be the chemical dissociation of the O–O bond of an adsorbed HO₂⁻ on a Mn^III site. Our experimental results are in line with this, showing that other MnO₂ allotropes behave similarly to β-MnO₂ (Fig. 3a). In the anodic branch (where the HPOR dominates), in contrast, no such effect was reported by Ryabova et al., and the HPOR was limited by peroxide diffusion at potentials above 1 V vs RHE for both α-Mn₂O₃ and β-MnO₂.⁸³ A similar observation is made here for oxidation current densities at 1 V vs RHE on the present set of Mn-oxides mixed with Vulcan, though the anodic branches at this potential have not yet reached their plateau. This may be due to the lower OH⁻ concentration used in our work compared to Ref. 83, 0.1 and 1.0 M respectively. As one can see from reaction (2), the HPOR involves OH⁻ as a reactant, and this could therefore impact the HPOR reaction rate.

As discussed in the previous section, the different Mn-oxides possess very different BET areas, α-Mn₂O₃ in particular having a lower BET area than the other materials. To assess their intrinsic activity for HPRR/HPOR, we normalized the geometric current density by the surface area of the oxides in the active layer (from the known loading and the known BET area of each oxide). Since Vulcan has negligible activity for HPRR/HPOR (Fig. 3a, black diamonds), this normalization should correctly represent the intrinsic specific surface activity of the Mn-oxides in a three-electrode environment where the liquid electrolyte should wet the entire surface of all of the catalysts. Figure 3c highlights the superior intrinsic activity of α-Mn₂O₃ compared to the three MnO₂ allotropes. In this graph, α-Mn₂O₃ not only reaches higher reduction currents at low potential (due to non-limitation by a slow chemical step) but also higher oxidation and reduction currents near OCP, i.e. faster intrinsic electrochemical kinetics. To quantify the latter, we fitted the polarization curves in the region close to OCP (−30/+50 mV) with a straight line and report the derivative dj/dE (units of mA cm_oxide⁻² V⁻¹) in Fig. 3d. The figure shows a seven-fold higher intrinsic activity for α-Mn₂O₃ compared to β- and α-MnO₂, and fourteen-fold higher intrinsic activity compared to δ-MnO₂. These results are in line with the five-fold higher kinetic rate constants k₂ and k₃ for Mn₂O₃ vs β-MnO₂ reported in Ref.83, and describing the kinetics of reaction (2). In conclusion, α-Mn₂O₃ is shown to be intrinsically more active toward HPRR and HPOR than the other Mn-oxides. Furthermore, our recent study showed that Mn₂O₃ is also more stable than the other Mn-oxides in the ORR region, even after normalization for the oxide area.⁸⁵ Therefore, both from activity and stability viewpoints, α-Mn₂O₃ is clearly identified as the most promising Mn-oxide material for scavenging peroxide species produced during ORR. The full potential of α-Mn₂O₃ vs other Mn-oxides is, however, challenged by its generally lower BET area, related to the high temperature calcination usually required to form this phase.^92–94

In the next step, the Mn-oxides were mixed with Fe_0.5-NH₃, in the same weight ratio as was used for Vulcan, i.e. 20 wt% Mn-oxide on Fe_0.5-NH₃. A SEM image of Fe_0.5-NH₃ is shown in Fig. S3a, showing an alveolar structure in that region of the catalyst. Due to the commensurate sizes of the Fe_0.5-NH₃ and MnO_x particles materials (Fig. 2), distinguishing the two phases is difficult in SEM images of the composite catalysts. As an example, Figs. S3b–S3c show SEM images of α-Mn₂O₃/Fe_0.5-NH₃, where some particles, due to their different morphology can likely be assigned to α-Mn₂O₃. Table SI reports the elemental contents determined by SEM-EDX on different regions of α-Mn₂O₃/Fe_0.5-NH₃, revealing areas with Mn and Fe weight contents close to those expected for a homogeneous distribution of Mn₂O₃ on Fe_0.5-NH₃, as well as areas with lower-than-expected Mn contents. The HPRR and HPOR polarization curves of Fe_0.5-NH₃ and of the four composite materials are shown in Fig. 3b, confirming the previously discussed feature of cathode limiting current density but also revealing a new feature with respect to trends in the OCP values. The Fe_0.5-NH₃ catalyst alone, while showing a much higher OCP value (ca. 0.94 V vs RHE) compared to all MnO_x/Vulcan, has poor HPRR and HPOR kinetics, as visible from its lower dj/dE near OCP and lower absolute value of cathodic current density at low potential (Figs. 3a–3b). The poor HPRR and HPOR activity of Fe_0.5-NH₃ is a general phenomenon for FeNC catalysts, observed in both alkaline and acid electrolytes.^95–101 The maximum HPOR and HPRR current densities for Fe_0.5-NH₃ in the potential range 0.5–1.0 V vs RHE are ca. 0.2 and 0.4 mA cm⁻², respectively (Fig. 3b). These values are significantly increased for the MnO_x/Fe_0.5-NH₃ composites. In the cathodic region, the limiting current density increases in the same order than for the MnO_x/Vulcan catalysts, i.e. α-MnO₂ < δ-MnO₂ = β-MnO₂ < α-Mn₂O₃, demonstrating the role of Mn-oxides in increasing the HPRR activity of MnO_x/Fe_0.5-NH₃ composites. In the anodic region of 0.95–1.0 V vs RHE, the HPOR current density is also higher for all MnO_x/Fe_0.5-NH₃ composites relative to Fe_0.5-NH₃ (Fig. 3b). The composites MnO_x/Fe_0.5-NH₃ also show higher OCP values than the corresponding MnO_x/Vulcan layers, with OCP values of 0.86 and 0.91 V vs RHE for α-MnO₂/Fe_0.5-NH₃ and α-Mn₂O₃/Fe_0.5-NH₃, respectively. This shift of OCP toward higher values seems driven by the even higher OCP of Fe_0.5-NH₃ (0.94 V vs RHE). In fact, due to the negligible activity of Vulcan in MnO_x/Vulcan layers, and the comparable loading of Fe_0.5-NH₃ in the pure Fe_0.5-NH₃ layer and in the composite MnO_x/Fe_0.5-NH₃ layers (0.20 and 0.16 mg cm⁻², respectively), one may expect the polarization curves of MnO_x/Fe_0.5-NH₃ layers to behave, in first approximation, as the linear superposition of the polarization curves of MnO_x/Vulcan and Fe_0.5-NH₃.

Overall, the results confirm the higher HPRR activity of α-Mn₂O₃/Fe_0.5-NH₃ relative to Fe_0.5-NH₃ itself, at potentials of 0.50–0.85 V vs RHE. At the cathode of an AEMFC, peroxide may be produced in this potential region due to high current density and relatively low selectivity of metal-free N-doped sites (along with some peroxide produced from FeN_x sites).

The demonstration of the concept of scavenging HO₂⁻ formed during ORR with the addition of Mn-oxides to Fe_0.5-NH₃ requires RRDE measurements in O₂-saturated electrolyte. Figures 4a–4c shows, for reference, the ORR selectivity, activity, and Tafel plots of MnO_x/Vulcan layers, respectively. The ORR onset of MnO_x/Vulcan layers is ca. 0.9 V vs RHE for α-Mn₂O₃, α- and β-MnO₂ and ca. 20 mV lower for δ-MnO₂. The ORR polarization curves show otherwise a similar shape and reach comparable O₂-diffusion-limited current densities (Fig. 4b). The latter are close to the value expected for an overall apparent four-electron ORR, correlated by low% HO₂⁻ detected at the ring (Fig. 4a). A closer detail at the% HO₂⁻ reveals fine differences, with decreasing% HO₂⁻ in the order α-MnO₂, δ-MnO_2, β-MnO₂, α-Mn₂O₃. This is in the exact same order as the absolute value of limiting current density observed on MnOx/C layers in the HPRR measurement (Fig. 3a). This suggests that ORR on these MnO_x proceeds via a 2e + 2e pathway (subsequent 2 electron reduction on same MnO_x sites) and/or via 2 electron pathway to peroxide followed by catalytic HO₂⁻ decomposition (reaction 3), which could also lead to apparent overall near four-electron ORR.

Zoom In Zoom Out Reset image size

Then, the ORR selectivity, polarization curves and Tafel plots were recorded for Fe_0.5-NH₃ and the four MnO_x/Fe_0.5-NH₃ composite layers (Figs. 5a–5c). Fe_0.5-NH₃ alone shows high ORR activity of ca. 1 mA·cm⁻² at 0.9 V vs RHE (Fig. 5c), i.e. a mass activity of ca. 5 A·g⁻¹, similar to our recent study on Fe_0.5-NH₃ in alkaline electrolyte.⁴³ The ORR activity (in mA per cm² geometric area of electrode) at 0.9 V of the four MnO_x/Fe_0.5-NH₃ composite layers is ca. 30% lower (Fig. 5c), as expected from the slightly lower loading of Fe_0.5-NH₃ in the composite cathodes(only 0.16 mg_FeNC cm⁻², while the reference Fe_0.5-NH₃ layer has 0.2 mg_FeNC cm⁻²). Last but not least, Fig. 5a demonstrates higher selectivity for all MnO_x/Fe_0.5-NH₃ composite layers relative to Fe_0.5-NH₃. This result can be interpreted as a bifunctional effect of the composite layers, the FeN_x sites mainly driving the ORR to water, with minute amount of peroxide produced (either on some FeN_x sites, or on the N-C surface), and the MnO_x particles mainly acting as electrochemical scavengers of the minute amount of peroxide produced.

Zoom In Zoom Out Reset image size

Then, the stability of the composite catalyst α-Mn₂O₃/Fe_0.5-NH₃ was evaluated in operando SFC-ICP/MS. α-Mn₂O₃ was selected for the composite catalyst due its higher oxide-surface-normalized HPRR activity, as shown in this study and previously, as well as its slightly higher stability among this set of four oxides, as recently reported by Speck et al.⁸⁵

Figure 6a (middle panel) in the region from 0 to ca. 500 s shows similarly low Mn dissolution rates (<ca. 0.1 ng·cm⁻²·s⁻¹) for the composite α-Mn₂O₃/Fe_0.5-NH₃ catalyst in both Ar- and O₂-saturated electrolytes when the potential is cycled for the first time between 1 and 0.6 V vs RHE. This is due to the relative thermodynamic stability of α-Mn₂O₃ in that restricted potential range. In contrast, when the potential is cycled between 1 and 0 V (time between 500 and 1500 s in Fig. 6a, middle panel), the Mn leaching rates increased when compared to the 1–0.6 V cycling and a higher Mn leaching rate is observed in O₂-saturated electrolyte vs Ar-saturated electrolyte. This trend is similar to that observed for α-Mn₂O₃/C.⁸⁵ An accelerated stress test protocol showed that the overall dissolution of manganese in O₂-saturated electrolyte is higher than in Ar-saturated electrolyte (Fig. 6b, middle panel), which is explained by the presence of hydrogen peroxide produced in O₂-saturated solution, as already reported by us.⁸⁵ Concerning Fe dissolution, the catalyst does not show evident dissolution peaks during the first slow scan between 1 and 0 V in O₂-saturated electrolyte (Fig. 6b, bottom panel). The dissolution rate of Fe is 0.05–0.08 ng cm⁻²·s⁻¹, comparable to the Fe leaching rate observed for Fe_0.5-NH₃ and previously reported by us.⁴³ Fig. 6b (middle panel) shows that the Mn dissolution rate from α-Mn₂O₃/Fe_0.5-NH₃ is comparable to the dissolution rate previously reported for the same Mn-oxide supported on Vulcan,⁸⁵ with maximum leaching rates of 0.15–0.30 ng·cm⁻²·s⁻¹ in O₂-saturated electrolyte during the slow-scan from 1 to 0 V vs RHE.

Zoom In Zoom Out Reset image size

Finally, Fe_0.5-NH₃ and α-Mn₂O₃/Fe_0.5-NH₃ were evaluated as cathode catalysts in operating AEMFCs and compared to Pt/C (Fig. 7a). The initial activity at 0.9 V of the PGM-free cathodes is comparable with that of Pt/C. Fe_0.5-NH₃ shows a higher current density at 0.9 V (65 mA·cm⁻²) than the composite catalyst, which shows 52 mA·cm⁻², almost identical to Pt/C (55 mA·cm⁻²) (Fig. 7b). At lower cell voltages, the PGM-free cathodes do not perform as well as the Pt/C cathode, due to mass transport issues. The latter may be assigned to a more difficult path for oxygen to reach the active centers in such cathodes (possibly due to the microporous nature of the carbon in Fe_0.5-NH₃), but also to local O₂ starvation due to the much lower site density in Fe_0.5-NH₃ (ca. 1.6 wt% Fe only)⁴³ compared to Pt/C (40 wt% Pt). Nevertheless, the absolute performance obtained with these PGM-free cathodes is promising: the Fe_0.5-NH₃ and α-Mn₂O₃/Fe_0.5-NH₃ cathodes are able to support a current density of 1.43 A·cm⁻² and 1.25 A·cm⁻² at 0.6 V, respectively. This corresponds to 59 and 52% of the current density at 0.6 V obtained with 0.45 mg_Pt·cm⁻² (2.41 A·cm⁻²). The peak power density is 1.04 W·cm⁻² for Fe_0.5-NH₃ and 0.98 W·cm⁻² for the α-Mn₂O₃/Fe_0.5-NH₃ composite, compared to 1.53 W·cm⁻² for Pt/C (Fig. 7c). The higher current density at low potential (below 0.4 V) obtained using the composite α-Mn₂O₃/Fe_0.5-NH₃ cathode relative to Fe_0.5-NH₃ can be attributed either to the effect of the Mn-oxide helping in reducing peroxide into water at those potentials, or to the slightly lower loading of Fe_0.5-NH₃ in the composite cathode (80% the loading compared to that in the Fe_0.5-NH₃ layer), helping the mass-transport of oxygen.

Zoom In Zoom Out Reset image size

Table SII gives an overview of AEMFC performance previously reported in H₂ and O₂ conditions with PGM-free cathodes, indicating three key metrics corresponding to pure kinetic control (current density at 0.9 V), mixed kinetic and mass-transport control (current density at 0.6 V) and mainly mass-transport control (peak power density) of the cell performance. From Table SII, it can be seen that the current densities reached at 0.9 V with the present Fe_0.5-NH₃ and α-Mn₂O₃/Fe_0.5-NH₃ cathodes (52–65 mA·cm⁻²) are significantly above those reported for PGM-free metal oxides (≤15 mA·cm⁻²)^102–106 and other FeNC catalysts with atomically-dispersed FeN_x sites (≤25 mA·cm⁻²),^107–111 with the exception of the material labelled Fe-S-Phen/CNT in Ref. 112 and that resulted in a current density of ca. 120 mA·cm⁻². The latter was however reached at a cell temperature of 80 °C, significantly above 60 °C used in the present study. The current density at 0.6 V is an important operating point of fuel cells, often resulting in a balanced compromise between energy efficiency (proportional to the ratio of operating to thermodynamic cell voltage) and power density. The current density obtained with Fe_0.5-NH₃ and α-Mn₂O₃/Fe_0.5-NH₃ cathodes (1430 and 1250 mA·cm⁻²) are above those reported for other FeNC cathodes (190–977 mA·cm⁻²),^{107,108,110–112} with two exceptions. The first, is a FeCoPc/C material in Ref., 109 that achieved 1670 mA·cm⁻² at 0.6 V, though this was obtained at a higher cell temperature (80 °C vs 60 °C here) and with a thinner AEM (15 μm, vs 50 μm here). Both factors are known to significantly improve the cell performance, due to lower Ohmic resistance but also improved water management. The second exception was by our team using a similar FeNC material to Fe_0.5-NH₃,¹¹³ where a current density of 1830 mA·cm⁻² was achieved at 65 °C and 0.6 V. However, the data reported there was taken with a much higher IEC, thinner AEM based on high density polyethylene.¹¹⁴

The current density obtained with Fe_0.5-NH₃ and α-Mn₂O₃/Fe_0.5-NH₃ cathodes at 0.6 V (1430 and 1250 mA·cm⁻²) are also above those reported recently for CoO and Co₃O₄ (320 and 520 mA cm⁻²),^105,106 and comparable to those reported for CoO_x or cobalt ferrite (1330 and 1560 mA cm⁻²)¹⁰³ and MnCo₂O₄ spinel (970 mA cm⁻² at 60 °C).¹⁰⁴ The CoO_x and cobalt ferrite were tested in the same cell housing and with the same FC test bench as in the present study, although at slightly higher cell temperatures of 65 and 70 °C (Table SII). Thus, in spite of a much higher ORR activity at 0.9 V for the present cathodes based on Fe_0.5-NH₃ vs those based on cobalt-rich oxides, similar current densities are hitherto observed at 0.6 V. This reveals that mass-transport must be further optimized in Fe_0.5-NH₃ cathodes, in order to improve the catalyst utilization at high current density. This could be achieved with improved water management (thinner membrane, optimized cell and O₂ gas temperature and RH) or improved layer structure (pore size distribution, ionomer content and distribution, optimized catalyst loading). Finally, the peak power densities achieved with PGM-free cathodes are also reported in Table SII. A graph of peak power density vs current density at 0.6 V of the data shown in Table SII would reveal a straight proportional relationship, such that a separate discussion is not necessary for the peak power density. A similar linear relationship is also visible in Fig. 7b of Ref. 108

Overall, this overview of recent AEMFC data with PGM-free cathodes highlights the kinetic advantage of the present Fe_0.5-NH₃ and α-Mn₂O₃/Fe_0.5-NH₃ cathodes over other PGM-free cathodes, and its state-of-the-art power density. Compared to the 2022 milestone of initial AEMFC performance recently defined by the U.S. Department of Energy (≥ 1000 mA cm⁻² at 0.65 V on H₂/O₂, T ≥80 °C and total PGM loading ≤0.2 mg cm⁻², see Table II in Ref. 8), the cell performance achieved at 60 °C and at 0.65 V in the present work (1110 mA cm⁻²) already slightly exceeds this milestone with a PGM-free cathode, but with a higher PGM loading than targeted in the MEA (i.e. at the anode). Improved cathode and cell performance by operating to higher temperature and with optimized conditions should, in the near future, leverage the reduction of the PGM loading at the anode while maintaining the same overall cell performance. The investigation of the durability of FeNC layers during AEMFC operation and their optimization for operation with air (CO₂-free) feed are the next important steps for increased maturity of FeNC-based AEMFC cathodes.