Catalyst synthesis and structuresSupplementary Fig. 1 describes the two-stage strategy for the synthesis of highly active Fe–N–C catalysts. Initially, a ZIF-8@Phen composite precursor was thermally activated to form NC carriers and the micropores in the carriers accounted for more than 90% of NC carriers (Supplementary Figs. 2 and 3, Table 1). The Fe–NC catalyst was then treated at 900 °C in an Ar atmosphere with ammonium chloride and ammonium bromide salt (NH4Cl and NH4Br, named BrCl), and the final catalyst was named as Fe–NCBrCl (Supplementary Fig. 4). Mesopores and abundant carbon defects emerged during the BrCl treatment due to the etching effect of HCl, HBr, and NH3, which are the decomposition products of NH4Cl and NH4Br (Supplementary Figs. 5–7, Supplementary Movie 1). Figure 1a demonstrated the well-dispersed atomic Fe sites (bright dots) and some holes within the Fe–NCBrCl catalyst via an aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image. The broad distribution of Fe, N and Br was further demonstrated by the energy-dispersive X-ray spectroscopy (EDS) elemental mapping and XRD (Fig. 1b, c and Supplementary Fig. 8). Figure 1d, e indicated that BrCl treatment can change the N structure and cause the production of pyridine N based on X-ray absorption near edge spectroscopy (XANES) obtained for the N K-edge and X-ray photoelectron spectroscopy (XPS) N 1s spectrum. The formation of additional edge sites is facilitated by the etching effect of NH3, which is generated from the decomposition of NH4X, thereby promoting the incorporation of pyridinic N (Supplementary Fig. 9)36. In addition, as demonstrated by the hysteresis curve, BrCl treatment inhibited the creation of superparamagnetic iron species (Fe derivative clusters) and increased the number of Fe–Nx sites (Fig. 1f, Supplementary Figs. 10 and 11). Fe K-edge energies in X-ray spectra of Fe–NC and Fe–NCBrCl were compared to ferrous and ferric standards including FePc and Fe2O3. The results revealed that Fe was mostly in the Fe3+ oxidation state in Fe–NC catalysts, whereas Fe in Fe–NCBrCl had a lower oxidation state (Fig. 1g). A first coordination shell consisting of four nitrogen atoms can faithfully duplicate the Fourier transform of the extended X-ray absorption fine structure (EXAFS) of Fe–NC and Fe–NCBrCl (Fig. 1h, i, Supplementary Table 2). The Fe–N bond length in Fe–NCBr and Fe–NCBrCl treated with NH4Br was expanded relative to that in Fe–NC. NH4Cl and NH4Br played distinct roles in the synthesis of catalysts and reaction mechanisms. NH4Cl was especially effective in fostering the development of mesopores and carbon defects, which were vital for improving mass transport in fuel cells and providing doping sites for Br. In contrast, NH4Br, while less efficient at creating mesopores and carbon defects than NH4Cl, benefits from its ability to decompose into bromine more readily, as HBr yielded Br2 with relative ease (Supplementary Note 1, Figs. 12–14, and Tables 3, 4 for details). When a Br atom was doped adjacent to the Fe–N4 site, its larger atomic radius induced spatial distortion in the Fe–N4 plane, resulting in the elongation of the Fe–N bond (Supplementary Fig. 15 and Table 5). Furthermore, high-sensitivity low-energy ion scattering spectroscopy (HS-LEIS) results revealed that 92% of the Br ions were confined to the surface region (less than 1 nm) of the Fe–NCBrCl particles, with a surface bromine content of 1.51 wt% and a bulk bromine content of 0.12 wt% (Supplementary Figs. 16–18 and Table 6).Fig. 1: Structure analysis of Fe–NC and Fe–NCBrCl catalysts.a Aberration-corrected atomic resolution HAADF-STEM micrograph of Fe–NCBrCl. b, c HAADF-STEM images and corresponding EDS mapping. d XAS spectra collected at N K-edge. e The deconvoluted high-resolution N 1s XPS. f The hysteresis curves. g Normalized Fe K-edge XANES spectra. h Fourier transform of k3-weighted EXAFS spectra. i FT-EXAFS fitting curve, gray ball represents C atom, blue ball represents N atom, red ball represents Fe atom, yellow ball represents Br atom.ORR activity, SD and TOF in RDEORR activity of different Fe–NCs was assessed in a 0.1 M H2SO4 electrolyte by rotating disk electrode (RDE, Supplementary Fig. 19). ORR voltammetry exhibits a strong ORR performance illustrated in Fig. 2a in the order of Fe–NCBrCl > Fe–NCBr ≈ Fe–NCCl > Fe–NC. With a half-wave potential (E1/2) of 0.838 VRHE and a mass activity of 4.302 A g−1 at 0.85 VRHE, the best performing Fe–NCBrCl catalyst demonstrated high activity. These values are significantly higher than that of the Fe–NC (Fig. 2b). An in-situ nitrite stripping technique was used to assess how catalyst activation affected site activity and site density (SD)37. The SD values of Fe–NC was 5.74 × 1019 sites g−1 (0.05 sites nm−2), lower than that of the Fe–NCBrCl (6.31 × 1019 sites g−1, 0.06 sites nm−2), which reflects the importance of NH4Cl and NH4Br during pyrolysis (Fig. 2c, Supplementary Figs. 20, 21, Table 7). Furthermore, the SD values of Fe–NCBrCl showed a remarkable degree of agreement with the edge iron site count (6 ± 2 × 1019 sites g−1) calculated for this catalyst using a domain size of La = 7.3 ± 2.3 nm acquired from Anthony et al.28. The TOFs on nitrite poisoning were calculated by dividing the reduction in ORR kinetic current by SD. Fe–NCBrCl at 0.80 and 0.85 VRHE had computed TOF values of 4.18 and 0.90 e site−1 s−1, respectively, which were higher than Fe–NC (Fig. 2d). This suggests that doping Br atoms increased the TOF of the electrochemically accessible Fe sites38. The results from density functional theory (DFT) calculations substantiate that Fe–N4 sites doped with halogen atoms, such as Cl and Br, exhibit an enhanced ability to activate O2 molecules. Nonetheless, the formation of Cl doping is hindered during pyrolysis due to the recalcitrance of HCl. Consequently, Br2, which arises from the decomposition of HBr, is more readily incorporated into Fe–N–C materials36 (Supplementary Note 2, Figs. 22–26, Table 8).Fig. 2: ORR activity, SD and TOF measurements of Fe–NCs.a RDE ORR curves measured in O2-saturated 0.1 M H2SO4 at 10 mV s−1 with a rotation speed of 900 rpm. The catalyst loading was 0.6 mg cm−2. b The kinetic ORR activities of Fe–NCs. c Comparison of SD values determined by nitrite stripping method for Fe–NC and Fe–NCBrCl catalysts. d Comparison of TOF at 0.80 and 0.85 VRHE for Fe–NC and Fe–NCBrCl catalysts. Error bars represent the standard deviation for three separate measurements.Performance characterization in a fuel cellFe–NC and Fe–NCBrCl served as the cathode catalysts in single-cell PEMFCs tests. Fe–NCBrCl exhibited superior performance under 250 kPaabs H2-O2 conditions, reaching a peak power density (Pmax) of 1.86 W cm−2 at 0.45 V. This value is approximately 0.6 W cm−2 higher than that of Fe–NC (Fig. 3a), which exceeded the commercial Pt/C prepared under the same conditions (0.2 mgPt cm−2, 150 kPa absolute O2 pressure). Fe–NCBrCl cathode has achieved the highest values reported for PGM-free catalysts in recent years (Supplementary Fig. 27 and Table 9)29,32,39,40,41,42,43. At 0.8 and 0.7 ViR-free, the current densities of Fe–NCBrCl were 0.781 and 3.111 A cm−2, respectively (Fig. 3b, c), surpassing the value achieved by Fe–NC (0.147 and 0.826 A cm−2 at 0.8 and 0.7 ViR-free). Figure 3d showed that the Tafel slope of Fe–NCBrCl was 84 mV dec−1, which was lower than that of Fe–NC (97 mV dec−1), reflecting the enhanced kinetic activity.Fig. 3: Performance tests with Fe–NC and Fe–NCBrCl as the cathode catalysts in a single-cell PEMFCs.a H2-O2 PEMFCs polarization curves. Cathode, ~3.5 mgcat cm−2 for Fe–NC and 0.2 mgPt cm−2 for Pt/C; anode, 0.4 mgPt cm−2 Pt/C; GORE-SELECT® membrane (15 μm thickness); 0.3 L H2 min−1 and 0.4 L O2 min−1 feed, 100% relative humidity (RH), 250 kPa absolute partial pressure H2 and O2, 80 °C, electrode area 1.21 cm2. The cell voltage and power density are not iR corrected. b The polarization curves without (solid line) and with (dotted line) iR-correction and HFR under 250 kPaabs H2-O2 PEMFCs. c Current densities at 0.8 ViR-free and 0.7 ViR-free. d Tafel plots derived from the ORR polarization curves displayed in (c).The superior activity of Fe–NCBrCl was further proven in the evaluation of PGM-free catalysts based on the DOE test protocol. As seen in Fig. 4a, Fe–NCBrCl demonstrated a Pmax of 1.29 W cm−2 at 0.45 V. Repeated tests demonstrated excellent repeatability of the high performance (Supplementary Fig. 28). Fe–NC, on the other hand, had a lower Pmax of 0.97 W cm−2. And for Fe–NCBrCl, the current density at 0.9 ViR-free was 54 mA cm−2, exceeding the DOE 2025 target of 44 mA cm−2 at 0.9 ViR-free (Fig. 4b, Supplementary Fig. 29, Table 10). In the H2–air operation (Fig. 4c), Fe–NCBrCl exhibited a Pmax of 0.88 W cm−2, which was considerably higher than most of the previously reported PGM-free catalysts (Supplementary Table 11). In addition, a measured current density of 259 mA cm−2 and 300 mA cm−2 was reached at 0.8 and 0.78 V (or 300 mA cm−2 at 0.8 ViR-free), only 41 mA cm−2 (or 0.02 V) lower than the DOE 2025 target (300 mA cm−2 at 0.8 V) (Fig. 4d). In order to better evaluate the practical application performance of Fe–NCBrCl, a three-piece PEMFCs stack with a MEA area of 25 cm2 was assembled (Supplementary Figs. 30 and 31). To mitigate mass transport limitations at high current densities, which were critical in practical situations involving lower O2 stoichiometry operation, the Fe–NCBrCl loading in the catalyst layer was set at 2.0 mg cm−2. The polarization curves of stack were acquired in an O2 atmosphere without background pressure. At 1.8 and 1.4 V (without iR-correction), the PEMFCs stack demonstrated a power of 25.3 and 36.2 W, respectively (Supplementary Fig. 32). It meant that the Fe–NCBrCl catalyst showed promise for practical applications, suggesting the feasibility of widespread application.Fig. 4: The polarization curves obtained by DOE MEA test protocol of PGM-free electrocatalyst.a H2–O2 polarization curves, acquired from OCV to 0.70 V in 25 mV steps and 0.70 V to 0.25 V in 50 mV steps, with a hold time of 45 s per point. 0.3 L H2 min−1 and 0.4 L O2 min−1 feed, 100% relative humidity (RH), 150 kPa absolute partial pressure H2 and O2, 80 °C. b H2–O2 polarization curve with iR-correction. c H2–air polarization curves. d Tafel plots derived from the ORR polarization curves displayed in (c).Mechanistic insights into the PEMFCs activity enhancementWithin the cathode catalyst layer, the spatial distribution of Fe sites and ionomer was verified by time-of-flight secondary ion mass spectrometry (ToF-SIMS). Two-dimensional images of spatial arrangement of the components on the catalytic layer surfaces corresponding to Fe–NC and Fe–NCBrCl were shown in Fig. 5a, b. S− ions (green region) indicated sulfonic groups on ionomers, whereas Fe+ ions (red region) indicated Fe sites on Fe–NCs catalysts. It was evident that the thicker ionomer coating on the Fe–NC surface resulted in a sparse dispersion of Fe+ (Fig. 5a). On the surface of the Fe–NCBrCl catalytic layer, trace amounts Br− ions can be detected (Supplementary Fig. 33). The Br− ions in Fe–NCBrCl repelled the sulfonic groups of the ionomer, causing a more even dispersion of the ionomer throughout the surface of the catalytic layer (Fig. 5b). The distribution of ionomers in the pores was further investigated by examining the depth distribution curves of Fe+ and SO3− ions. The ionomer polymer was typically 2–5 nm in size, too tiny to penetrate the core of the particles and could only be coated on their surface44,45. The pores on the surface of Fe–NC particles were mostly micropores, accounting for about 92.8% (Supplementary Table 1). Within the Fe–NC particles, the SO3− distribution intensity was essentially nonexistent (Fig. 5c). Nevertheless, Fe–NCBrCl particles formed mesopores as a result of NH4Cl and NH4Br breaking down, and the surface Br− ions on the particles aid in the uniform dispersion of ionomers. Consequently, mesopores allowed a little quantity of ionomers to enter the particles. As a result, the SO3− intensity distribution of Fe–NCBrCl particles was also found inside of it (Fig. 5c). Upon analyzing the quantitative correlation between the surface Br content and the Pmax, a parabolic relationship was observed. The Pmax attained its apex at an equimolar ratio of NH4Br to NH4Cl (1:1) (Supplementary Note 3, Figs. 34–36, Table 12).Fig. 5: The interface effect of MEA.a, b The secondary ion two-dimensional imaging from ToF-SIMS, a Fe–NC, b Fe–NCBrCl. The green is S−, the red is Fe+. c Integral SO3− distribution in cathode catalyst layer shows intensity gradients from both Fe–NC (blue) and Fe–NCBrCl (red) catalysts. d Relationship between Rtotal and absolute gas pressure obtained by the limiting current method. Insets were Rnp and Rp. e An Arrhenius plot of i0 obtained from the Nyquist plots shown in Supplementary Fig. 37. f Nyquist plots for PEMFCs of indicated cathode catalysts at the current density of 1.0 A cm–2, the inset shows the equivalent circuit model, Rct and Rmt of Fe–NC and Fe–NCBrCl obtained by EIS fitting. Error bars represent the standard deviation of the fitting results.By using the limiting current method46,47, the oxygen transport behavior of membrane electrodes was typically studied (Supplementary Fig. 37). Traditionally, Rtotal was the total of the pressure-dependent transfer impedance Rp and the pressure-independent transfer impedance Rnp48. The oxygen molecules diffusing through the mesopores in the gas diffusion layer (GDL) and the cathode catalytic layer (CCL) were the primary source of the resistance of oxygen diffusion, or Rp. On the other hand, oxygen was transported in the micropores of the catalyst aggregate and the ionomer layer, which caused the resistance to oxygen diffusion, or Rnp49,50,51. Rnp somewhat increased as a result of the ionomer seeping into the interior of Fe–NCBrCl particles (Fig. 5d). Lower Rp will arise from improved oxygen transport in the catalytic layer due to the mesoporous Fe–NCBrCl. It was established that the oxygen transport limitation in CCL was predominantly caused by the diffusion resistance of O2 in the mesopores of the thick cathode catalytic layer, when comparing Rp and Rnp. The effect of cell temperature on the charge transfer of Fe–NC and Fe–NCBrCl was also examined. As the temperature rose, the high-frequency limit shifted on the real axis in a negative direction (Supplementary Fig. 38). It indicated that the ohmic resistance, that is, the resistance of the membrane, decreased as the temperature of the cell increased. The charge-transfer mechanism was responsible for the depressed ellipse that was seen in all charts in the high-frequency range. The charge transfer resistance Rct was estimated by fitting these impedance measurements to the equivalent circuit (Supplementary Table 13). By replacing the constant phase angle element (CPE) with the double layer capacitance of the Randles equivalent circuit, this equivalent circuit was created. A porous electrode causes a depressed semicircle, which is frequently described by the CPE. The exchange current i0 was calculated from Rct according to the Eq. (1)52:$${i}_{0}=\left(\frac{{RT}}{{nF}}\right)\frac{1}{{R}_{{ct}}}$$
(1)
The temperature dependency of i0 might be assessed. In Supplementary Fig. 39, the Arrhenius plot of i0 was displayed52. In addition, good linearity was noted. For Fe–NCBrCl, the apparent activation energy (Ea) of charge transfer was determined to be 8.56 kJ mol−1 based on the angle of the plot (Fig. 5e). Compared to the value of 10.56 kJ mol−1 for Fe–NC, this value showed a faster charge transfer for Fe–NCBrCl, which was compatible with the lower Tafel slope and lower resistivity (Supplementary Fig. 40). Electrochemical impedance spectroscopy Nyquist plots at 1.0 A cm−2 and a Randles model simulation in Fig. 5f further supported the quicker mass and charge transfer. As a result of a richer triple-phase boundary (TPB), faster interfacial charge transfer, better combination of ionomers and Fe active sites, and a more porous catalytic layer, fuel cells assembled with Fe–NCBrCl demonstrated an extremely high levels of operational performance.
A Fe-NC electrocatalyst boosted by trace bromide ions with high performance in proton exchange membrane fuel cells
