Effect of Pyrolysis Atmosphere and Electrolyte pH on the Oxygen Reduction Activity, Stability and Spectroscopic Signature of FeN_x Moieties in Fe-N-C Catalysts

Two Fe-N-C catalysts are prepared with the sacrificial metal-organic-framework method, using ZIF-8 (Basolite Z1200, Sigma Aldrich) as support, 1,10-phenantroline (≥99%, Sigma Aldrich) as a secondary source of nitrogen and iron (II) acetate (≥99.99%, Sigma Aldrich) as the source of metal. The catalyst precursor is prepared with a weight ratio of 4/1 for ZIF-8/phenanthroline and 0.5 wt% of iron in the complete catalyst precursor. The three precursors are initially mixed using low-energy ball milling at 400 rpm for 2 h 20 min, with 5 minutes pause every 30 minutes of milling. The obtained catalyst precursor is transferred into a quartz boat and inserted in a quartz tube. The first pyrolysis is performed in flash-pyrolysis mode, pre-equilibrating the quartz tube and oven at 1050°C, then pushing the quartz boat and catalyst precursor within 1 min in the heating zone of the furnace with an outer magnet. The pyrolysis duration at 1050°C in flowing Ar is exactly 1 h. The pyrolysis is terminated by opening the split-hinge furnace, removing the quartz tube and letting it cool down at room temperature for 20 minutes. The obtained catalyst is labelled Fe_0.5-Ar. To prepare the NH₃-pyrolyzed catalyst, Fe_0.5-Ar is re-pyrolysed with the same flash-pyrolysis mode, but in flowing pure NH₃ and for only 5 minutes at 950°C. The obtained catalyst is labelled Fe_0.5-NH₃.

Activity and durability in acidic and alkaline electrolytes are obtained using a RDE set-up (Pine instruments) and either 0.1 M KOH or 0.1 M H₂SO₄ electrolytes. The three-electrode configuration involves a platinum wire immersed in a H₂-saturated electrolyte compartment, separated from the main compartment by a fritted glass, as a reversible hydrogen electrode (RHE) reference; a graphite plate as a counter electrode and a glassy carbon (GC) rotating disk (5 mm diameter, Pine Research) as a support for the active layer forming the working electrode. The ink is prepared by adding in sequence 5 mg catalyst, 54 μL Nafion (5% perfluorinated resin solution), 744 μL ethanol, 92 μL ultrapure water and sonicating for 1 hour in an ice bath. An aliquot of 7 μL of the ink is pipetted on the GC disk and dried at room temperature, resulting in a catalyst loading of 200 μg cm⁻². The Initial activity is measured in O₂-saturated electrolyte at a scan rate of 1 mV·s⁻¹ (SP-300, BioLogic Potentiostat) and at rotation rate of 1600 rpm. Due to the low scan rate and low catalyst loading, no correction for the capacitive current is needed. To evaluate the durability of the catalysts in RDE set-up, a load-cycling protocol is applied, comprising 5000 triangular cycles performed at a scan rate of 100 mV·s⁻¹ in the potential range of 0.6-1.0 V vs RHE and in N₂-saturated electrolyte. The ORR activity after the AST is measured after re-saturating the solution with O₂.

The leaching of Fe during short electrochemical cycling, performed either before or after the AST (AST performed in Ar-saturated electrolyte), is investigated in an on-line electrochemical scanning flow cell (SFC) directly connected to an inductively-coupled plasma mass spectrometer (ICP–MS), previously developed by us.^38–40 It is however challenging to continuously measure Fe leaching over the length of the AST, due to drift of the ICP-MS with time and need for constant recalibration. To measure ⁵⁶Fe, the ICP–MS (Perkin Elmer, NexION 350) is operated in dynamic-reaction-cell mode, using methane as the reaction gas. The cell is calibrated to both an acidic and alkaline standard solution of iron to ensure maximized detection of ⁵⁶Fe. Daily calibration of the ICP–MS is done by a four-point calibration curve (0, 0.5, 1.0, 5.0 μg·L⁻¹) of standard iron solutions prepared from Merck Centripur ICP standards (Fe(NO₃)₃, 1000 mg·L⁻¹, in 2–3% HNO₃). As an internal standard, we use ⁵⁸Co (Merck Centripur, Co(NO₃)₂, 1000 mg·L⁻¹, in 2–3% HNO₃) diluted to 50 μg·L⁻¹ in HNO₃ (0.15 mol·L⁻¹) to ensure full acidification of the electrolyte in a y- connector before its introduction in the ICP–MS. The SFC consists of a three-electrode setup using a Ag/AgCl (Metrohm, 3 M KCl) reference electrode, a graphite rod counter electrode and a GC RDE as a working electrode, on which the catalyst is drop cast. A positioning stage (Physik Instrumente, M-403.6 DG) is used to approach individual catalyst spots on the working electrode. Stability measurements are conducted in alkaline (99.99%, Suprapur, NaOH, 0.05 mol·L⁻¹) as well as in acidic (Suprapur, 0.05 mol·L⁻¹ H₂SO₄) electrolyte. The potentiostat (Gamry, Reference 600) as well as purging gases and the positioning stage is controlled by a custom LabVIEW software. The catalyst ink is prepared from the Fe-N-C catalyst, Nafion (5% perfluorinated resin solution) and water, with a mass ratio of catalyst/dry-ionomer of 4 and a catalyst concentration of 3.3 g·L⁻¹ in the liquid ink. An aliquot of 2.75 μL is deposited on the GC, resulting in a catalyst loading of 400 μg·cm⁻². Such a high loading is necessary to reach a sufficient signal-to-noise ratio in the ICP-MS measurements. This is due to the aforementioned interference of the ⁴⁰Ar¹⁶O dimers and high background noise of iron in alkaline solution. For Fe-leaching measurement before and after the AST, the latter is conducted in a separate Teflon RDE-cell containing 100 mL electrolyte, and the RDE tip is then quickly transferred from the Teflon cell to the SFC set-up, with the catalyst still wetted by electrolyte. The RDE cell used for the AST consists of four individual compartments, one each for the three electrodes and for the purging tube. The counter and reference electrodes are the same as in the SFC setup.

The pristine catalysts are characterized with ⁵⁷Fe Mössbauer spectroscopy at room temperature. To this end, ⁵⁷Fe-enriched catalysts are used, prepared identically as the ones otherwise investigated in this study, except for the use of ⁵⁷Fe acetate during their synthesis. Mössbauer spectra are measured at room temperature with a ⁵⁷Co:Rh source. The measurements are carried out in triangular velocity waveform using NaI scintillation detector for γ-rays. The velocity calibration is done with an α-Fe foil. A mass of 30 mg of ⁵⁷Fe-enriched Fe_0.5-Ar and Fe_0.5-NH₃ powders are necessary for a proper signal-to-noise resolution.

The pristine catalysts are also characterized with XAS in both ex situ and operando conditions. The XAS spectra are collected at SAMBA beamline (synchrotron SOLEIL) at the Fe K-edge using a double crystal Si 220 monochromator and a Canberra 35-elements germanium detector for operando acquisition in fluorescence mode. The catalyst ink (10 mg catalyst, 100 μL 5% Nafion solution and 50 μL ultrapure H₂O) is prepared via ultrasonication, and 50 μL is deposited on circa 3 cm² circular area of a larger conductive carbon foil, resulting in a catalyst loading of circa 1 mg cm⁻².⁴¹ The carbon foil is then inserted in a three-electrode cell, 0.1 M KOH electrolyte is added and the three electrodes are connected, using Pt-wire counter electrode and a Hg/HgO reference electrode. Note that all potentials are however reported in V vs. RHE in this work. Air is continuously bubbled in the electrolyte during the measurements. The operando XAS spectra are collected at open circuit potential (OCP), 0.2, 0.4, 0.6, 0.8 and 1.0 V vs. RHE. Ex situ spectra were collected in transmission geometry on pellets of 1 mm diameter using Teflon powder as a binder.

To measure the specific surface area of carbon in the catalysts, sorption isotherms of N₂ are measured in liquid nitrogen (77 K) with a Micromeritics, ASAP 2020 instrument. The sample is previously cleaned at 200°C for 5 h in flowing nitrogen. The specific surface area is determined by the multipoint Brunauer-Emmett-Teller (BET) method.

The ORR activity before and after the AST is shown in Figure 1 in linear- and semi-logarithmic scales, for Fe_0.5-Ar (Figs. 1a–1b) and Fe_0.5-NH₃ (Figs. 1c–1d). In each sub-figure, the measurements performed in alkaline electrolyte are shown as purple curves while those performed in acidic electrolyte are plotted in gray color. The curves before and after AST are identified with filled and open symbols, respectively. Before the AST, Fe_0.5-Ar shows a similar activity in both electrolytes (Fig. 1b, curves with filled symbols), with ORR mass activity at 0.9 V_RHE of 0.35 and 0.25 A·g⁻¹ in alkaline and acidic electrolytes, respectively. A slight difference is visible only at low-potential, with a less defined diffusion-limited current density in acidic medium (Figures 1a, filled gray symbols).

Zoom In Zoom Out Reset image size

A distinct behavior appears for Fe_0.5-NH₃, characterized with a much higher initial activity in alkaline vs. acidic electrolyte (Figures 1c–1d, filled purple vs. filled gray symbols), with ORR mass activity at 0.9 V vs. RHE of 4.6 and 0.5 A·g⁻¹, respectively. The former is among the highest reported value of the ORR activity in alkaline medium for PGM-free catalysts.^42–46 Its lower initial activity in acidic than in alkaline electrolyte can mostly be assigned to a fast protonation of highly-basic nitrogen groups followed by anion-adsorption on positively-charged [NH]⁺ groups, which had been previously shown, on other NH₃-pyrolyzed Fe-N-C catalysts, to divide the activity by a factor 5 to 10.^27,45,47 High initial activity of NH₃-pyrolyzed Fe-N-C catalysts in acidic liquid-electrolyte could previously be achieved only with high catalyst loading and optimized Nafion/catalyst ratio.⁴⁷ It is anticipated that these two parameters allow a short time protection of the highly basic N-groups from the liquid acidic electrolyte. With the present experimental conditions (low scan rate and low catalyst loading), the high ORR activity of Fe_0.5-NH₃ cannot be captured in acidic liquid-electrolyte because the highly basic N-groups were likely protonated and charge-neutralized even before the first polarization curve was recorded. The surface state of Fe_0.5-NH₃ in acidic medium is then similar to that of Fe_0.5-Ar, explaining similar initial ORR activities in acid (compare filled gray symbols in Fig. 1b and Fig. 1d). The slightly higher initial ORR mass activity in acid of Fe_0.5-NH₃ vs. Fe_0.5-Ar (0.50 vs 0.25 A·g⁻¹ at 0.9 V vs. RHE, see Table I) can be explained by its higher BET specific area, 970 vs. 635 m²g⁻¹.

A second possibility to explain the lower initial activity of Fe_0.5-NH₃ in acid vs. alkaline electrolyte is that (at least some) FeN_x moieties after NH₃ pyrolysis are intrinsically different from those after Ar-pyrolysis and, while being more active for ORR (regardless of which pH), would also be less stable in acidic medium. It might be that such highly-active FeN_x moieties were leached in acidic medium even before completing the acquisition of the first polarization curve. Evidence for increased Fe leaching with the Fe_0.5-NH₃ catalyst in acidic medium is reported in the section entitled electrochemical stability and operando iron leaching study. We also note that the contribution of N-groups (not binding Fe) to the overall initial ORR activity of those two catalysts can be neglected. Two reference materials prepared similarly as Fe_0.5-Ar and Fe_0.5-NH₃ but without addition of Fe acetate were studied in RDE and their initial ORR activity in 0.1 M KOH at 0.9 V vs. RHE is 0.03 and 0.57 A·g⁻¹, respectively. This is circa 10 times lower than the activity of corresponding Fe-NC materials, 0.35 and 4.60 A·g⁻¹ (Table I). The difference would be even larger in acidic medium.

In order to characterize the active-site structure in Fe_0.5-NH₃ and Fe_0.5-Ar, we first resorted to ex situ spectroscopic characterization. The ex situ ⁵⁷Fe Mössbauer spectra could not reveal any significant difference between the two catalysts (Figures 2a–2b and supporting Table S1). They were fitted with two doublets D1 and D2, each one having similar Mössbauer parameters for both catalysts and being present in similar ratio. These doublets are assigned to atomically-dispersed FeN_x moieties.¹⁸ Similar Mössbauer spectra imply that the local Fe coordination and site geometry, up to two coordination spheres, are similar in both catalysts. Further identification of the active-site structure was performed using XAS. Figure 2c shows the ex situ X-ray absorption near-edge structure (XANES) spectra of Fe_0.5-NH₃ (circle symbols) and Fe_0.5-Ar (solid curve), that are characteristic for Fe-N-C catalysts free of metallic particles.¹⁸ The absence of a strong signal at ca 2.2 Å (Fe-Fe bond distance in metallic and metal-nitride particles, uncorrected for phase shift) in the Fourier transform (FT) of the EXAFS spectra of both Fe_0.5-Ar and Fe_0.5-NH₃ indicates that both catalysts are free of Fe-based particles, or that the amount of particles is below the XAS detection limit, which is approximately 3% relative to the total amount of Fe in the catalysts (Figure 2d). The absence in the XANES spectra of the pre-edge peak at 7118 eV, characteristic for unpyrolyzed Fe(II) phthalocyanine (FeN₄ square-planar structure with identical Fe-N bond distances), reveals a broken D_4h symmetry of the FeN_x moieties. This may be due to structural disorder or existence of ferric moieties with an axial ligand such as O₂. The latter case is particularly possible when recording ex situ spectra of catalysts in their resting state in air environment. The white line intensity is slightly stronger for Fe_0.5-Ar than Fe_0.5-NH₃. This may be interpreted as a higher coordination number in the first coordination sphere surrounding Fe (either N, C or O atoms). This hypothesis is supported by the stronger signal of the first peak (at ca 1.4 Å) in the FT-EXAFS spectra for Fe_0.5-Ar vs. Fe_0.5-NH₃ (Figure 2d). Such ex situ changes may be assigned to a stronger FeN₄-O₂ interaction ex situ for Fe_0.5-Ar, possibly due to a higher average oxidation state of Fe in Fe_0.5-Ar vs. Fe_0.5-NH₃. In the latter, a lower oxidation state of Fe may be expected due to the presence of Lewis-base (highly basic) nitrogen groups, if some of those groups are directly involved in Fe cations ligation.

Zoom In Zoom Out Reset image size

In order to investigate whether the small differences observed ex situ with XANES and FT-EXAFS spectra of both catalysts remained, disappeared or were exacerbated during the ORR in alkaline medium, we performed operando XAS. The XANES and EXAFS spectra were recorded in alkaline electrolyte between 0.2 and 1.0 V vs. RHE, covering all regions of the RDE polarization curves (Supporting Figure S1). Figure 3 shows the operando XANES spectra (left-handside) and FT-EXAFS spectra (right-handside) for Fe_0.5-Ar (top) and Fe_0.5-NH₃ (bottom). For Fe_0.5-Ar, both the operando XANES and FT-EXAFS spectra reveal large changes with the electrochemical potential (Figures 3a–3b). The change of the XANES spectra with potential observed here for Fe_0.5-Ar in alkaline medium is similar to that reported by us for the same catalyst in acidic medium.⁴¹ The magnitude of change is however ca twice smaller in alkaline vs. acidic medium for Fe_0.5-Ar, as can be seen by comparing the Δμ spectra (compare the inset of Fig. 3a in this work with the inset of Fig. 4b in Ref. 41). In the ORR potential region, the FeN₄-moieties in Fe_0.5-Ar undergo a structural change, as revealed by the complex change and existence of isobestic points at 7130.5 and 7156.4 eV in the set of XANES spectra.

Zoom In Zoom Out Reset image size

These isobestic points support the existence of at least two Fe-site geometries, whose fraction switches gradually from 0 to 1 across the potential range of 0.2 to 1.0 V vs. RHE. This observation has also been reported, but only in acidic electrolyte hitherto, for other Fe-N-C materials^19,48,49 and interpreted as a change from in-plane FeN₄ to out-of-plane FeN₄ configuration as a function of potential. The operando FT-EXAFS (Figure 3b) also support a structural change, with a decreasing signal intensity at 1.4 Å with decreasing potential. This can be interpreted as the presence of oxygen adsorbates (O₂, OH and H₂O) strongly adsorbed on FeN₄ sites at high potential, and their absence or elongated Fe-O bond distance at low potential. Those spectroscopic changes were reversible, similar spectra being recorded at a given potential, when scanning down and then up the potential from 1.0 to 0.2 V and then back up to 1.0 V.

The operando XANES spectra of Fe_0.5-NH₃ in alkaline electrolyte reveal much smaller changes with electrochemical potential, as compared to changes observed for Fe_0.5-Ar in the same electrolyte (compare Figure 3c vs. Figure 3a). The Δμ spectra (inset of Fig. 3c) reveals a trend that is however comparable to the Δμ signals observed with Fe_0.5-Ar in alkaline electrolyte (inset of Fig. 3a), but with ca twice lower magnitude of change. The smaller spectral changes with electrochemical potential observed for Fe_0.5-NH₃ vs. Fe_0.5-Ar may appear at first counterintuitive, since the former has a higher BET area and expectedly a higher exposure of Fe-sites to the top surface and to the electrolyte. The smaller spectral changes with potential observed for Fe_0.5-NH₃ must therefore be assigned to a distinct environment of Fe-sites in Fe_0.5-NH₃ compared to Fe_0.5-Ar, in line with their much different ORR activities in alkaline electrolyte (Figure 1).

In line with the operando XANES spectra, the corresponding FT-EXAFS spectra of Fe_0.5-NH₃ are almost unchanged with potential (Fig. 3d). The trend of the signal intensity at 1.4 Å with electrochemical potential is the same as for Fe_0.5-Ar (decreasing signal with decreasing potential), but the magnitude of the change is also much smaller. It is in fact only significant at the lowest potential studied, namely 0.2 V vs. RHE.

Coming back to the initial question whether the small differences observed ex situ with XANES and FT-EXAFS spectra of both catalysts remained or disappeared in operando, the direct comparison of the XANES and EXAFS spectra of both catalysts at low potential (0.2 V vs. RHE, Figure S2) reveals that the spectroscopic signatures are almost identical. Thus, the two catalysts differ in ex situ conditions or in situ at high electrochemical potential, but the differences become smaller as the potential is lowered, and become negligible at 0.2 V vs. RHE. This supports the hypothesis that, in ex situ conditions, Fe is in a higher oxidation state in Fe_0.5-Ar than in Fe_0.5-NH₃, and binds more O₂ or oxygen adsorbates. In operando conditions, the difference progressively vanishes as the electrochemical potential is reduced (ferric moieties turning into ferrous moieties in Fe_0.5-Ar, becoming then in a similar state as Fe_0.5-NH₃).

The stability of Fe_0.5-Ar and Fe_0.5-NH₃ catalysts was studied in both acidic and alkaline electrolytes. Table I summarizes the activity observed before and after the 5000 load-cycle AST. Starting with Fe_0.5-Ar in alkaline medium, an activity increase from 0.07 to 0.11 mA cm⁻² at 0.9 V vs. RHE can be seen after 5000 load-cycles in N₂-saturated solution (purple curves in Figs. 1a–1b). Similarly, also Fe_0.5-NH₃ shows high stability in alkaline electrolyte, with very slight decay in activity at 0.9 V vs. RHE (from 0.92 to 0.83 mA·cm⁻², see purple curves in Figs. 1c–1d). In acidic medium, no activity loss was observed for Fe_0.5-Ar, but a significant loss observed for Fe_0.5-NH₃, with a relative decrease of about 24% (Table I).

We attribute this reduced activity following AST in N₂-saturated electrolyte to a loss of a fraction of the active centers, i.e. FeN₄ or only Fe cations from FeN₄ moieties, from the nitrogen-doped carbon network. Since the AST was performed in N₂-saturated electrolyte, no ORR occurred during the AST and we can exclude a degradation or deactivation due to hydrogen peroxide or reactive oxygen species formed from peroxide and FeN_x sites, which was recently demonstrated to cause a main deactivation of Fe-N-C catalysts in acidic medium.³⁴ To show this loss of iron in situ, we conducted SFC-ICP–MS measurements. Figure 4 summarizes the operando Fe leaching measurements where we applied the same potential-scan protocol (upper plot) to Fe_0.5-Ar and Fe_0.5-NH₃ in oxygen-saturated electrolyte. The electrode was first contacted by the SFC (corresponding time marked with * in the graph, t ∼ 250 s) at open circuit potential (OCP). The applied electrochemical potential protocol is then 20 cyclic voltammograms (CVs) in the range 1.0 to 0.6 V vs. RHE at a scan rate of 100 mV·s⁻¹, a potential range typically occurring during the ORR in fuel cell devices. Another chronoamperometry at 1.0 V was recorded for 200 s, before applying one CV at a low scan rate of 2 mV·s⁻¹ from 1.0 V down to 0.0 V vs. RHE, and back up to 1.0 V vs. RHE to identify in more detail the potential-dependence of Fe dissolution

Zoom In Zoom Out Reset image size

For each catalyst considered separately, Figure 4 clearly identifies higher Fe dissolution rates in acid than in alkaline electrolyte, especially during the slow CV. We first discuss the transient release of Fe occurring when the SFC contacts the electrode with the electrolyte (time marked with an asterisk), leading to an initial loss of iron. For Fe_0.5-Ar, this initial loss is similar in both electrolytes (Fig. 4a, middle panel), while for Fe_0.5-NH₃ the dissolution rate in acid medium is more than double that in alkaline electrolyte (Fig. 4b, middle panel). Comparing the contact dissolution in acidic medium, the peak dissolution rates are ca 1.0 and 0.4 ng_Fe·cm⁻²·s⁻¹ for Fe_0.5-NH₃ and Fe_0.5-Ar, respectively. After 400 s at OCP in acid medium, the Fe dissolution rate became < 0.15 ng_Fe·cm⁻²·s⁻¹ for Fe_0.5-Ar but remained significant (∼ 1.0 ng_Fe·cm⁻²·s⁻¹) and quite constant for Fe_0.5-NH₃. The cumulative Fe dissolution of Fe_0.5-NH₃ at that stage is however restricted to ca 0.5 μg_Fe·cm⁻² (Fig. 4b, lower panel), much lower than the total amount of Fe in the catalyst layer (ca 2.5 wt% Fe in Fe_0.5-NH₃, leading to ca 10 μg_Fe·cm⁻²). It is therefore unlikely that the lower activity measured for Fe_0.5-NH₃ in acid vs. alkaline originates from a very fast dissolution of iron. In alkaline medium, the curves of Fe dissolution rate vs. time of both catalysts are nearly superimposed (from immersion to OCP hold, time 250 to 850 s on x-axis), with a peak value of Fe dissolution rate of ca 0.5 ng_Fe·cm⁻²·s⁻¹, quickly decreasing (only ca 0.1 ng_Fe·cm⁻²·s⁻¹ before starting the fast CVs). Therefore, a trend is observed that Fe_0.5-Ar is more stable than Fe_0.5-NH₃, and that alkaline environment leads to lower dissolution rate than acidic medium.

The subsequent fast 20 CV scans increased the Fe release rate in acidic medium, especially for Fe_0.5-NH₃ (increasing to 1.5 ng_Fe·cm⁻²·s⁻¹), while it had no impact on the Fe dissolution rate in alkaline medium. Note that, due to the high scan rate used during the 20 CVs, the effect of scanning up or down the potential cannot be distinguished, and only a lump Fe dissolution rate is observed.

The time-resolved Fe dissolution rate during the subsequent slow potential scan then allows us identifying at which potential Fe is dissolved. In sulfuric acid, the onset of Fe dissolution while scanning the potential from 1 V down to 0 V occurs at ca 0.75 V vs. RHE, and the peak of dissolution rate occurs at ca 0.2–0.3 V vs. RHE, for both catalysts. The intensity of the peak of Fe dissolution is however 10 times higher for Fe_0.5-NH₃ vs. Fe_0.5-Ar. At potentials E < 0.2 V vs. RHE, the Fe dissolution rate decreases for both catalysts, and remains very low also during the positive-going scan from 0.0 V to 1.0 V vs. RHE. For Fe_0.5-NH₃, the cumulative Fe amount leached after the slow CV reaches ca 2 μg·cm⁻², representing about 50% of the total Fe content initially present (Fig. 4b, lower panel). Regarding the Fe release rate during the slow CV in alkaline electrolyte, there is no significant effect of the electrochemical potential in the negative-going branch of the scan, while reverting the scan direction from 0.0 V and upwards resulted in increased Fe dissolution rate for Fe_0.5-NH₃ but unmodified Fe dissolution rate for Fe_0.5-Ar. These experiments were repeated multiple times, and showed reproducible trends.

While Figure 4 informs on the electrochemical conditions in which Fe is dissolved, Figure 5 quantitatively shows how much Fe from the catalysts was dissolved as a function of time in the SFC-ICP-MS protocol before the AST. The y-axis shows the %Fe remaining in the catalyst relative to the initial Fe content. The cumulative dissolved Fe content was obtained from the integral of the curves shown in the lower panels of Figure 4 while the total Fe content in each electrode was derived from i) the fixed Fe-N-C catalyst loading value and the exact geometric area investigated by SFC-ICP-MS (verified each time by a microscope) and ii) the knowledge of the initial Fe content in each catalyst. The latter were measured by ICP-MS on the catalyst powders to be 1.45 wt% for Fe_0.5-Ar and 1.57 wt% for Fe_0.5-NH₃. Figure 5 shows that the absolute Fe dissolution is restricted for Fe_0.5-Ar (at both pH) and for Fe_0.5-NH₃ at high pH (5 to 10% relative Fe content is dissolved after 20 fast CVs and a slow scan) while for Fe_0.5-NH₃ at acidic pH, more than 50% of the initial Fe content present in the active layer was dissolved after the same time.

Zoom In Zoom Out Reset image size

These time-resolved Fe dissolution data reveal that the Fe-based sites in Fe_0.5-NH₃ are less stable in acidic medium than those present in Fe_0.5-Ar, while in alkaline medium the stability of Fe_0.5-NH₃ is as good, or even better, than that of Fe_0.5-Ar. While the data might be interpreted by assuming that a much higher fraction of all FeN_x sites are exposed to the electrolyte in Fe_0.5-NH₃ than in Fe_0.5-Ar, this assumption should have resulted in a slightly increased Fe dissolution for Fe_0.5-NH₃ in alkaline electrolyte compared to that for Fe_0.5-Ar in the same electrolyte. This is however not observed. The operando XAS data are also not in support of an increased fraction of FeN_x sites being exposed to the electrolyte in Fe_0.5-NH₃ (smaller magnitude of change for the XANES and EXAFS spectra with potential than for Fe_0.5-Ar). Thus, the electrolyte-exposed FeN_x sites in Fe_0.5-NH₃ seem to be intrinsically less stable in acidic medium than those in Fe_0.5-Ar.