The crystal structure of the synthesized samples was characterized using X-ray diffraction (XRD-6100) in a diffractometer equipped with Cu Kα radiation (λ = 0.1542 nm, 40 kV, 30 mA). The high-resolution transmission electron microscope (HR-TEM), high-angle annular darkfield scanning transmission electron microscopy (HAADF-STEM), and energy-dispersive X-ray (EDX) spectrometry mapping images were obtained using a Lorentz scanning transmission electron microscope (STEM, Talos F200X) equipped with Super X-EDS system at 200 kV. The Brunauer-Emmett-Teller (BET) surface area was determined via N₂ ad-/de-sorption and analyzed using the Langmuir method (Autosorb iQ) by fitting the N₂ adsorption isotherms < 0.05 (P/P₀), and pore volume distribution was obtained using Quenched Solid Density Functional Theory (QSDFT). X-ray photoelectron spectroscopy (XPS, Thermo Fisher ESCALAB Xi+) with an Al-Kα source (hν = 1,486.68 eV) was employed to analyze the surface elemental valence states and chemical composition, with binding energies referenced to the C 1s peak at 284.8 eV. The ultraviolet-visible (UV-Vis) absorption spectra were measured on a PE Lambda950 spectrophotometer. Metal contents were quantified by inductively coupled plasma-atomic emission spectroscopy (ICP-AES, Agilent 700). X-ray absorption near-edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) analyses of the Pt L₃-edge were conducted on Table-XAFS-500, using Pt foil as reference standard. The Raman spectra were recorded using an InVia Qontor Raman Microscope with a 532 nm laser source. The electron paramagnetic resonance (EPR) measurements were performed on Bruker-A300.

The lattice strain and grain size were determined using the Williamson-Hall equation:

$$ \begin{equation} \begin{aligned} \beta_{\mathrm{hkl}} \cos \theta=(k \lambda / D)+4 \varepsilon \sin \theta \end{aligned} \end{equation} $$

where β_hkl denotes the full width at half maximum (FWHM) of the (hkl) diffraction peak, θ is the Bragg diffraction angle, k is the shape factor (k = 0.9), λ is the Cu Kα X-ray wavelength (1.5418 Å), D is the grain size (nm) and ε is the lattice strain.

All electrochemical measurements were conducted using a CHI 760e workstation in a three-electrode configuration. A graphite rod served as the counter electrode, while an Ag/AgCl (3 M KCl) electrode was used as the reference. All reference potentials were calibrated against the reversible hydrogen electrode (RHE). To prepare the catalyst ink, 4 mg of the as-synthesized catalyst was dispersed with 990 μL of isopropanol, 990 μL of DI water, and 20 μL (5 wt %) Nafion solution. Subsequently, the ink was drop-coated onto a glassy carbon disk with a diameter of 5.0 mm, ensuring Pt loading of 15 µg cm^-2. After drying at room temperature in air, the glassy carbon disk was used as the working electrode. All working electrodes were first activated by cycling the potential in N₂-saturated 0.1 M HClO₄ solution between 0 and 1.1 V at 50 mV s^-1 until a stable and reproducible CV was achieved. In O₂-saturated 0.1 M HClO₄ and 0.1 M HClO₄ + 0.1 M H₃PO₄ solution, ORR polarization curves were recorded by sweeping in the range of 0.1 to 1.1 V at 10 mV s^-1 with a rotation speed of 1,600 rpm. The kinetic current density (J_k) was calculated from the polarization curves via the Koutecky-Levich equation:

$$ \begin{equation} \begin{aligned} 1/J=1 / J_{\mathrm{K}}+1 / J_{\mathrm{L}} \end{aligned} \end{equation} $$

where J represents the measured current density, while J_K and J_L denote the kinetic and diffusion-limited current densities, respectively.

The CO stripping tests were performed to calculate the ECSA. The electrolyte was initially saturated with CO for 10 min while maintaining the working electrode potential at 0.1 V vs. RHE, then bubbled N₂ for 30 min to eliminate CO from the solution, followed by conducting 2 CV cycles from 0.01 to 1.06 V at 50 mV s^-1. The ECSA was calculated according to:

$$ \begin{equation} \begin{aligned} \text { ECSA }=\left(S_{\mathrm{CO}} / V\right) /\left(0.42 \times L_{\mathrm{pt}}\right) \end{aligned} \end{equation} $$

where S_CO, V, and L_Pt denote the integration area of CO desorption, sweep rate, and Pt loading, respectively. The mass activity and specific activity were obtained by normalizing J_K with respect to the Pt loading and ECSA on the working electrode, respectively.

The high-temperature rotating disk electrode (HT-RDE) device was similar to the traditional RDE three-electrode test, whereas the counter electrode was a Pt wire, and the reference electrode served as a RHE. The electrolyte was either 0.1 M HClO₄ + 0.1 M H₃PO₄ at 80 °C or 85 wt% H₃PO₄ at 120 °C. The cell temperature was precisely controlled using a heating magnetic stirrer with an oil bath.

The cathode catalyst ink was prepared by dispersing the catalyst powder and PTFE into a mixed solvent of water and isopropanol with a 1:1 ratio. The ink was then sprayed onto a gas diffusion electrode (GDE) with a microporous layer (MPL) using pneumatic spray, covering an active electrode area of 4 cm² (2 cm × 2 cm). For the low Pt loaded cathodic CLs, the Pt loading was controlled at 0.3 mg cm^-2, and the thickness of CLs was maintained at 40-50 μm. In this case, the TANAKA-PtCoC catalyst with a Pt content of 46.9 wt% was selected as the control sample. For ultra-low Pt loaded cathodic CLs with Pt loading of 0.1 mg cm^-2, the Pt content of the as-synthesized catalysts was reduced and the commercial PtCoC catalyst with a Pt content of 10 wt% was used as the control sample to maintain the thickness of CLs at an appropriate 40-50 μm to prevent "acid flooding". JM Pt/C was employed as the anode catalyst, with a Pt loading of 0.1 mg cm^-2, a catalyst layer thickness of ~20 μm, and a PTFE content of 10 wt%.

The PBI membrane was immersed in 85 wt% H₃PO₄ to obtain the PA-doped PBI membrane with PA content of 400 wt%. The membrane was then sandwiched between the cathode and anode GDEs and hot-pressed at 160 °C under 1 MPa for 5 min to fabricate the membrane electrode assembly (MEA).

The performance of HT-PEMFC was tested at 160 °C; dry H₂/air (H₂/O₂) was used as the anode/cathode inlet gases with the flow rate of 100/400 (100/200) sccm and without back pressure. The MEA was first activated at a current density of 0.5 A cm^-2. Subsequently, polarization curves were recorded by adjusting the cell current and monitoring the resulting voltage. Electrochemical impedance spectroscopy (EIS) measurements were conducted by sweeping frequencies from 20 kHz to 0.1 Hz under a DC voltage of 0.6 V. Cyclic voltammetry (CV) was performed at 160 °C with dry H₂/N₂ as the anode/cathode supplied gases and the scan was conducted from 0 to 1.2 V at a sweep rate of 50 mV s^-1.

TEM images [Figure 2A and B] of the STGPtCo/EC electrocatalyst reveal that the ultrafine PtCo nanoparticles are uniformly distributed on the carbon support and surrounded by amorphous CELs. A high-resolution transmission electron microscopy (HR-TEM) image shows lattice spacings of 0.223 nm and 0.188 nm, corresponding to the (111) and (002) planes of PtCo, respectively [Figure 2C]. Energy-dispersive spectroscopy (EDS) mapping shows the co-localization of Pt, Co, and S elements within the particles of STG-PtCo/EC, indicating that S derived from the STG is incorporated into the CELs during thermal treatment [Figure 2D]. Moreover, EDS spectra [Supplementary Figure 1] further confirm the presence of S element, corroborating successful S-doping in the CELs. Notably, varying the carbon chain length of the additives enables precise control over the size of the PtCo nanoparticles and the thickness of the CELs. Using STG (C₂), MHA (C₆), and MUA (C₁₁) as additives yields PtCo nanoparticles with average sizes of 2.82 ± 0.58 nm, 3.62 ± 0.73 nm, and 4.82 ± 1.00 nm, and corresponding CEL thicknesses of 0.4 ± 0.2 nm, 1.0 ± 0.4 nm, and 2.3 ± 1.2 nm, respectively [Figure 2E and F, Supplementary Figures 2 and 3]. In contrast, the additive-free PtCo/EC exhibits significantly larger PtCo nanoparticles (9.98 ± 7.92 nm) with pronounced size nonuniformity and severe particle sintering, and no discernible CELs surround the particles [Supplementary Figure 4]. A HR-TEM image of PtCo/EC reveals lattice spacings of 0.370 nm, 0.270 nm, and 0.216 nm, corresponding to the (001), (110), and (111) planes of the face-centered tetragonal (fct) PtCo alloy, respectively [Supplementary Figures 5 and 6]^[26]. Notably, the introduction of additives increases the lattice spacing of the PtCo (111) plane, with the spacings being 0.218 nm in MHA/MUA-PtCo/EC samples and 0.223 nm in STG-PtCo/EC samples, indicating a lattice expansion of approximately 1% for MHA/MUA-PtCo/EC, whereas STG-PtCo/EC exhibits a more pronounced lattice expansion of nearly 3%. We attribute this phenomenon to the metal-support interaction (MSI) between the PtCo nanoparticles and the S-doped CELs.

Raman spectroscopy combined with N₂ adsorption-desorption analysis was employed to elucidate the effects of S-doping on the defect structure and specific surface area of the carbon support. The two peaks at approximately 1,330 and 1,590 cm^-1 in the Raman spectra [Figure 2G] correspond to the D band (associated with structural defects) and the G band (related to the crystallinity of sp²-hybridized carbon), respectively. STG-PtCo/EC exhibits the highest peak intensity ratio (I_D/I_G), indicating that S-doping introduced by STG leads to the highest defect density^[27]. Nitrogen ad-/de-sorption isotherms indicate that all samples exhibit type IV isotherms with H4 hysteresis loops, suggesting the presence of both micropores and mesopores^[28,29]. Compared with the additivefree PtCo/EC, the BET surface areas of the additive-assisted samples decrease markedly (PtCo/EC: 977.7 m² g^-1 > STG-PtCo/EC: 728.3 m² g^-1 > MHA-PtCo/EC: 395.8 m² g^-1 > MUA-PtCo/EC: 318.0 m² g^-1), primarily because the CELs cover the intrinsic micropores and mesopores of the carbon support. Notably, STG-PtCo/EC exhibits a higher porevolume distribution at ~0.68 nm, which is attributed to the formation of carbon defects within the CELs induced by S-doping [Supplementary Figure 7]. Additionally, EPR characterization in Figure 2H detects unpaired electrons associated with carbon defects. The STG-PtCo/EC exhibits the strongest signal intensity at g = 2.0032, indicating the highest defect density in the S-doped CELs^[29].

To investigate the role of additives in enhancing particle sintering resistance and forming S-doped CELs, UV-Vis spectroscopy and XPS analyses were conducted. Upon the addition of additives to the H₂PtCl₆ aqueous solution, the solution exhibits a discoloration from light yellow to orange, indicating an altered coordination environment of Pt [Supplementary Figure 8]. The UV-Vis spectral variations confirm that sulfhydryl and carboxyl groups in the additives coordinate with Pt(IV) and Co(II), respectively, forming a bimetallic complex network (detailed analysis is shown in Supplementary Figure 9).

Pt 4f XPS spectra of H₂PtCl₆-CoCl₂/EC reveal four characteristic peaks at 72.7/76.1 eV and 75.3/78.3 eV, corresponding to Pt(II) and Pt(IV) species, respectively. As the additive chain length decreases, the Pt(IV) signals progressively attenuate while the Pt(II) component increases, suggesting partial substitution of Cl ligands by sulfhydryl groups, thereby enhancing the electron density around Pt [Figure 2I]. Co 2p XPS spectra exhibit a positive binding-energy shift as the additive chain length decreases, which is attributed to the electron-withdrawing carboxyl groups coordinating with Co(II) [Figure 2J]. This bimetallic complex network can reduce the nucleation and growth energy barrier of the PtCo alloy. During high-temperature annealing (900 °C, 3 h), the network decomposes, enabling Pt and Co atoms to nucleate into PtCo clusters with a 1:1 stoichiometry [Supplementary Table 1], while the additive carbon chains carbonize to form CELs around the PtCo clusters. Meanwhile, a minor fraction of S atoms from the sulfhydryl groups is incorporated into the CELs, generating carbon defects [Supplementary Table 2]. Subsequent low-temperature annealing (600 °C, 6 h) promotes the rearrangement of Pt and Co atoms to form ordered PtCo alloys. The physical confinement by CELs and the chemical interaction between Pt and S prevent the sintering of PtCo into larger particles during annealing^[1,30].

X-ray diffraction (XRD) patterns exhibit an ordered crystal structure matching the PtCo alloy, adopting a tetragonal structure in the P4/mmm space group [Figure 3A]. Notably, as the additive chain length decreases, the XRD characteristic peaks progressively broaden, accompanied by a negative shift of the (111) diffraction peak, indicating that shorter-chain additives lead to the formation of smaller PtCo particles along with lattice tensile strain^[31-33]. The average particle sizes, calculated using the Williamson-Hall equation based on the FWHM of the XRD peaks^[34-36], follow the order: PtCo/EC (9.24 nm) > MUA-PtCo/EC (6.47 nm) > MHA-PtCo/EC (4.43 nm) > STG-PtCo/EC (3.05 nm), which aligns well with the particle size distributions obtained from STEM analysis [Figure 3B]. The corresponding ECSAs follow the order: PtCoC (18.6 m² g_Pt^-1) < MUA-PtCo/EC (51.4 m² g_Pt^-1) < MHA-PtCo/EC (77.7 m² g_Pt^-1) < STG-PtCo/EC (82.4 m² g_Pt^-1). The lattice strains (ε), calculated from the slopes of the Williamson-Hall plots, are determined to be 0.39% for STG-PtCo/EC, higher than those of MHA-PtCo/EC (0.23%) and MUA-PtCo/EC (0.17%) [Figure 3C]. This enhanced strain is attributed to the higher density of S-induced carbon defects in STG-PtCo/EC, which strengthens the MSI between the PtCo alloys and the CELs, thereby leading to more pronounced lattice expansion^[37,38].

XPS is employed to investigate the electronic structure and surface chemical states of the catalysts. The XPS survey spectra of STG-PtCo/EC reveal the presence of C, Pt, Co, S, and O signals, which is consistent with the expected composition [Supplementary Figure 10]. The C 1s XPS spectra reveal a peak associated with the C-S bonds at 285.7 eV for additive-assisted PtCo/EC, whose intensity progressively increases with decreasing additive chain length, indicating that S from the additives is thermally converted and doped into the CELs [Supplementary Figure 11A]^[39,40]. The Co 2p XPS spectra show an increase in metallic Co (Co⁰) content with shorter additive chain lengths, indicating that shorter chain additives facilitate the reduction of Co²⁺ to Co⁰ during thermal treatment [Supplementary Figure 11B, Supplementary Table 3]. The Pt 4f spectra [Figure 3D] reveal peaks corresponding to metallic Pt (Pt⁰) and oxidized Pt (Pt²⁺) species^[35,41]. Compared to commercial PtCoC, the binding energies of Pt in PtCo/EC, MHA-PtCo/EC, and MUA-PtCo/EC exhibit only a slight positive shift of ~0.04 eV, whereas the STG-PtCo/EC sample shows a pronounced positive shift of 0.17 eV. This substantial shift indicates that S-induced carbon defects in the CELs enhance the MSI, thereby promoting electron transfer from Pt to adjacent S atoms within the CELs and subsequently redistributing to the surrounding C matrix and neighboring Co atoms.

The X-ray absorption near edge structure (XANES) and EXAFS analyses were performed on STG-PtCo/EC and related samples to elucidate the valence states and coordination environments of Pt. The white-line intensity of the Pt L₃-edge XANES for STG-PtCo/EC lies between that of Pt foil and PtO₂, yet closer to Pt foil, indicating that the valence state of Pt in STG-PtCo/EC falls between 0 and +4, and more closely aligned with the metallic state. The higher intensity of STG-PtCo/EC compared to PtCo/EC suggests that the introduction of STG promotes electron transfer from Pt to Co and defect-rich S-doped CELs, resulting in more unoccupied Pt 5d orbitals [Figure 3E]^[26,42]. The k³-weighted EXAFS spectra at the Pt L₃-edge show a prominent peak for Pt foil at 2.55 Å [Figure 3F], corresponding to the first-shell Pt-Pt coordination. The peaks for the PtCo alloys shift to shorter radial distances (~2.39 Å) for both STG-PtCo/EC and PtCo/EC, indicating Pt-Co/Pt bonding resulting from alloy-induced atomic rearrangements. Notably, the peak position for STG-PtCo/EC exhibits an additional increase in radial distance of approximately 0.05 Å compared with PtCo/EC, suggesting lattice expansion caused by stronger MSI^[43]. This variation is further corroborated by the wavelet transform analysis [Figure 3G]

Collectively, these results indicate that the strong MSI between the S-doped CELs and PtCo alloys induces tensile lattice strain, which upshifts the Pt d-band center and alters the electronic structure, thereby optimizing the adsorption strength of oxygenated intermediates and enhancing ORR activity^[44,45].

The ORR performance of the synthesized catalysts was initially evaluated in 0.1 M HClO₄ at 25 °C using a RDE. Polarization curves reveal that the half-wave potential (E_1/2) for commercial PtCoC, PtCo/EC, STG-PtCo/EC, MHA-PtCo/EC and MUA-PtCo/EC catalysts are 0.910, 0.807, 0.908, 0.904 and 0.900 V (vs. RHE), respectively [Figure 4A, Supplementary Figure 12]. Meanwhile, STG-PtCo/EC exhibits the lowest Tafel slope (67.6 mV dec^-1), the highest mass activity (1.389 A mg_Pt^-1) and specific activity (1.685 mA cm_Pt^-2) at 0.85 V (vs. RHE) [Figure 4B and C]. These results indicate that the strong MSI between the alloys and the S-doped CELs induces lattice microstrain and modulates the Pt electronic structure, thereby effectively accelerating the ORR kinetics, while the reduced particle size ensures high Pt atomic utilization.

To evaluate the resistance of the catalysts to PA poisoning, the ORR performance was compared in O₂-saturated 0.1 M HClO₄ and 0.1 M HClO₄ + 0.1 M H₃PO₄ solutions at room temperature. The linear sweep voltammetry (LSV) curves reveal that, compared with the HClO₄ system, all catalysts exhibit a negative shift in E_1/2 in the PA-containing electrolyte, accompanied by a decrease in diffusion-limited current density (J_L). The negative shifts of E_1/2 follow the order: MUA-PtCo/EC (37 mV) < MHA-PtCo/EC (51 mV) < STG-PtCo/EC (56 mV) < PtCo/EC (87 mV) < commercial PtCoC (98 mV), indicating that the additive-assisted PtCo/EC catalysts possess resistance to PA poisoning [Figure 4D].

CV curves confirm the poisoning effect of phosphate anions on Pt sites [Supplementary Figure 13]. The CV curves in the PA-containing electrolyte exhibit a new peak at approximately 0.28 V, corresponding to the adsorption of phosphate anions on the Pt surface, resulting in surface reconstruction and introducing additional positive charges on Pt due to the shift in the potential of zero charge (pzc)^[46,47]. The peak appearing at 0.4-0.7 V corresponds to the adsorption/desorption of phosphate anions^[48]. Phosphate adsorption markedly decreases the peak area related to oxide formation/reduction region (Pt-O(H), 0.8-1.05 V vs. RHE).

The decrease in ECSA was evaluated through CO-stripping measurements [Supplementary Figure 14]. By establishing the correlation among the E_1/2 decay, ECSA retention, and MA retention, it is evident that MUA-PtCo/EC exhibits the best resistance to PA poisoning at room temperature, which can be attributed to its relatively thicker CELs that physically hinder the contact between PA molecules and the PtCo surface [Supplementary Figure 15].

All catalysts exhibit increased Tafel slopes in the PA-containing electrolyte, indicating that phosphate anions inhibit the ORR kinetics [Supplementary Figure 16]^[49]. Although MUA-PtCo/EC exhibits the highest ECSA retention, the excessively thick CELs impede mass diffusion, hindering O₂ access to the active sites, resulting in low specific activity (0.369 mA cm_Pt^-2) and slower reaction kinetics (Tafel slope = 104.2 mV dec^-1). Whereas STG-PtCo/EC maintains the highest mass activity (0.356 A mg_Pt^-1 @ 0.85 V) and specific activity (0.625 mA cm_Pt^-2 @ 0.85 V) [Figure 4E, Supplementary Figures 17 and 18], along with the smallest Tafel slope (82.9 mV dec^-1) in the HClO₄ + H₃PO₄ mixed electrolyte at room temperature, demonstrating excellent ORR performance [Figure 4F].

The i-t curves at a constant potential of 0.7 V reflect the stability of the catalysts in the PA-containing system [Supplementary Figure 19]. The ORR polarization curves before and after the i-t test indicate that STG-PtCo/EC exhibits the smallest ΔE_1/2 of 79 mV [Supplementary Figure 20]. The excellent stability of STG-PtCo/EC in the PA-containing system can be ascribed to the incorporation of S into the CELs, which introduces negatively charged carbon defects. These defects electrostatically repel phosphate anions, thereby suppressing their adsorption on Pt active sites or facilitating their rapid desorption.

Compared to room-temperature evaluations, assessing the phosphate tolerance of catalysts at high temperatures is more critical, as it better reflects their practical applicability in HT-PEMFCs. It is well established that the dissociation of PA is an endothermic process; therefore, elevated temperatures promote its dissociation, leading to an increased concentration of phosphate anions, which may further enhance the coverage of Pt active sites. The relationship between ORR current density, temperature, and the accessible active sites is given by

(1) $$ \begin{equation} \begin{aligned} j=n F K c_{\mathrm{O}_{2}}\left(1-\theta_{\mathrm{ad}}\right)^{\chi} \exp (-\beta F E / R T) \exp \left(-\gamma \Delta G_{\mathrm{ad}} / R T\right) \end{aligned} \end{equation} $$

where j is the current density; n, F, K, and R represent the number of electrons, Faraday constant, reaction rate constant, and gas constant, respectively. χ denotes the exponential factor; β and γ are symmetry factors; c_O2 is the O₂ concentration in the solution; θ_ad is the total surface coverage (including OH_ad and adsorbed phosphate species), E is the applied potential, T is the temperature, and ΔG_ad is the Gibbs free energy of adsorption for the oxygen-containing intermediate^[50].

This equation indicates that the elevated temperatures favor the enhancement of ORR kinetics in the absence of specifically adsorbed anions. However, in PA-containing media, increasing temperature substantially raises the concentration of phosphate anions, thereby intensifying the total surface coverage (θ_ad) on the catalyst surface. Moreover, elevated temperatures decrease the solubility of O₂ (c_O2) in the electrolyte. These combined effects are detrimental to the ORR. Therefore, it is more meaningful to evaluate catalytic performance in high-temperature, PA-containing solutions.

The ORR performance of the synthesized catalysts in 0.1 M HClO₄ + 0.1 M H₃PO₄ solution was evaluated at 80 °C using a high-temperature RDE setup [Supplementary Figure 21]. The LSV curves demonstrate that STG-PtCo/EC exhibits the highest ORR activity [Supplementary Figure 22]. Compared with the LSV curves at 25 °C, elevating the operating temperature to 80 °C results in positive shifts in E_1/2 of 26, 17, and 6 mV for STG-PtCo/EC, MHA-PtCo/EC, and commercial PtCoC, respectively, indicating enhanced ORR activity. In contrast, MUA-PtCo/EC exhibits a negative shift of 4 mV, suggesting a decline in its ORR activity at elevated temperature [Figure 4G].

The CV curves elucidate the underlying causes of these changes [Supplementary Figure 23]. Upon increasing the temperature to 80 °C, the peak area corresponding to phosphate anion adsorption in the range of 0.4-0.7 V rises, indicating enhanced PA dissociation and a higher phosphate coverage on the catalyst surface. Concurrently, the O-reduction peak area in the range of 0.8-1.05 V increases, suggesting accelerated reduction of oxygenated intermediates, which is favorable for ORR. For MUA-PtCo/EC, the phosphate adsorption peak area increases by 64.8%, whereas the O-reduction peak area increases by only 10.2%, indicating that mere physical blocking of PA by the thick CELs is insufficient to effectively suppress phosphate poisoning of active sites. Consequently, a significant loss of accessible active sites occurs, leading to an overall decline in ORR activity despite the intrinsic catalytic enhancement. In contrast, STG-PtCo/EC exhibits only an 8.9% increase in phosphate adsorption peak area but a 45.4% increase in O-reduction peak area, demonstrating superior phosphate tolerance. This performance is attributed to the incorporation of S into the CELs, which introduces abundant negatively charged carbon defects that electrostatically repel phosphate anions, thereby suppressing their adsorption or facilitating their desorption from Pt sites.

The Tafel plots reveal that elevating the temperature to 80 °C reduces the Tafel slopes of all as-synthesized catalysts, indicating enhanced ORR kinetics at higher temperatures. Notably, STG-PtCo/EC exhibits the smallest Tafel slope (79.9 mV dec^-1), confirming its superior ORR activity in the high-temperature, PA-containing electrolyte [Figure 4H].

Further comparison with the ORR polarization curves obtained at 25 °C in 0.1 M HClO₄ reveals that the negative shifts in E_1/2 follow the order: STG-PtCo/EC (29 mV) < MHA-PtCo/EC (34 mV) < MUA-PtCo/EC (48 mV) < commercial PtCoC (92 mV), demonstrating the superior resistance of STG-PtCo/EC to PA poisoning at higher temperatures [Figure 4I and Supplementary Figure 24]. CO-stripping voltammograms were used to evaluate the ECSA retention, providing insight into the poisoning effects of the catalysts in high-temperature conditions [Supplementary Figure 25]. All catalysts exhibit a negative shift in the CO oxidation peak at 80 °C, which is attributed to the accelerated oxidation of CO_ad to CO₂ at elevated temperatures, allowing the reaction to initiate at lower potentials. The CO-stripping area retention serves as an indicator for the preservation of accessible electrochemically active surface sites. For STG-PtCo/EC, MHA-PtCo/EC, MUA-PtCo/EC, and commercial PtCoC, the CO-stripping area retentions are 76.3%, 65.5%, 51.3%, and 47.7%, respectively, exhibiting an inverse correlation with the shifts in E_1/2. This suggests that the loss of accessible active sites plays a dominant role in the observed ORR performance degradation [Figure 4J]. The outstanding ORR performance of STG-PtCo/EC arises from the synergistic effects of CELs providing a physical barrier against PA and negatively charged carbon defects electrostatically repelling phosphate anions, thereby minimizing surface phosphate coverage.

STG-PtCo/EC exhibits a maximum mass activity of 0.695 A mg_Pt^-1 @ 0.85 V at 80 °C, which is 3.58 times of commercial PtCoC (0.194 A mg_Pt^-1). After 10,000 accelerated durability test (ADT) cycles at 80 °C in 0.1 M HClO₄ + 0.1 M H₃PO₄, it exhibits only 14 mV negative shift in E_1/2 and 21.8% mass activity loss, demonstrating robust high-temperature durability [Figure 4K and L, Supplementary Figure 26].

After the ADT test, TEM characterization was performed on the catalysts. The results reveal that commercial PtCoC undergoes severe particle agglomeration, leading to a significant increase in the average particle size [Supplementary Figure 27A-C]. In contrast, no pronounced agglomeration is observed for the additive-assisted catalysts, whose particle size distribution and lattice fringe spacing remain nearly unchanged before and after testing [Supplementary Figure 27D-L]. These findings indicate that the CELs effectively suppress PtCo particle dissolution, migration, and aggregation during electrochemical operation through physical confinement, thereby enhancing both thermodynamic and dynamic stability. Since no particle agglomeration is observed, the performance degradation of these catalysts is mainly attributed to the cumulative coverage of phosphate anions on the catalyst surface.

To further investigate the practical applicability of the catalysts, HT-RDE tests were conducted in 85 wt% H₃PO₄ at 120 °C. The higher temperature and concentrated PA environment result in an extremely low oxygen solubility in the electrolyte and impose significant limitations on proton transport. The LSV curves reveal a substantial decline in ORR activity for all catalysts at 25 °C, which is attributed to the predominance of the vehicle mechanism for proton conduction, resulting in sluggish proton transport. When the temperature rises to 120 °C, proton conduction is dominated by the Grotthuss mechanism, where protons are transferred by hopping between adjacent sites through a hydrogen-bond network formed among free PA molecules, resulting in a significant enhancement of proton transport rates^[51]. Consequently, the ORR performance of all catalysts is significantly improved, accompanied by the appearance of phosphate anion adsorption/desorption peaks and oxygen reduction peaks in the CV curves [Supplementary Figure 28].

The ORR polarization curves reveal that STG-PtCo/EC maintains the highest performance in 85 wt% H₃PO₄ at 120 °C, with the E_1/2 of 0.942 V, surpassing those of MHA-PtCo/EC (0.932 V), MUA-PtCo/EC (0.909 V), and commercial PtCoC (0.853 V). Notably, a kink is observed in the LSV curves between 0.7-0.8 V, attributed to the slow O₂ diffusion in the high-concentration PA solution, which cannot meet the oxygen consumption at the electrode surface^[52]. Furthermore, STG-PtCo/EC exhibits the highest diffusion limited current density (j_L) of 0.450 mA cm^-2, exceeding those of MHA-PtCo/EC (0.293 mA cm^-2), MUA-PtCo/EC (0.288 mA cm^-2), and commercial PtCoC (0.309 mA cm^-2) [Figure 5A and B]. This superior performance is attributed to the S-doped CELs with optimal thickness, which not only physically block direct contact between PA and Pt sites but also electrostatically repel phosphate anions via negatively charged carbon defects, thereby effectively reducing surface phosphate coverage without hindering O₂ access to active sites.

The growth in CO-stripping peak area and Δj_m @ 0.9 V at 120 °C highlights the superior phosphate tolerance of STG-PtCo/EC. Its CO-stripping peak area increases by 29.2%, substantially higher than that of other samples, indicating that the presence of negatively charged carbon defects preserved more unoccupied active sites despite the elevated phosphate anion concentration at higher temperatures [Figure 5C and Supplementary Figure 29].

To simulate the adsorption behavior of phosphate anions on catalysts under HT-PEMFC operating conditions, the synthesized catalysts were treated with 85 wt% H₃PO₄ solution at 160 °C for 18 h under an Ar atmosphere. XRD patterns before and after PA treatment exhibit no discernible changes [Supplementary Figure 30], indicating that the high-temperature PA treatment did not alter the crystal structure of the catalysts. The P 2p XPS spectra exhibit characteristic phosphorus peaks within the range of 132-138 eV for all samples after PA treatment [Supplementary Figure 31], indicating interactions between PA and the electrocatalyst surfaces. Notably, compared with PtCo/EC and commercial PtCoC, the additive-assisted PtCo/EC samples with CELs exhibit a markedly attenuated P signal intensity, confirming that the CELs effectively act as a physical barrier, limiting the interaction of PA with the PtCo alloy surface. Furthermore, the Pt 4f XPS spectra reveal that, after treatment with 85 wt% H₃PO₄, the peaks of STG-PtCo/EC remain essentially unchanged, whereas those of MHA-PtCo/EC, MUA-PtCo/EC, PtCo/EC, and commercial PtCoC exhibit positive shifts of 0.10, 0.19, 0.28, and 0.35 eV, respectively [Figure 5D and E, Supplementary Figure 32]. These results suggest that STG-PtCo/EC exhibits the weakest Pt-PA interaction, which can be attributed to the synergistic effects of the physical shielding imparted by the CELs and the electrostatic repulsion arising from negatively charged carbon defects^[18].

To investigate the anti-PA performance of the catalysts, we chose H₂PO₄^- as the representative spectator species, because it is the predominant phosphate anion under operating conditions and the principal contributor to Pt catalyst poisoning. The adsorption behavior of H₂PO₄^- was simulated on the PtCo (111) facet, which represents the catalytically active surface. H₂PO₄^- adsorbs onto the Pt surface through two O atoms, which is consistent with previous reports [Supplementary Figure 33]^[53]. The adsorption energies of H₂PO₄^- on both STG-PtCo/EC and PtCo/EC surfaces are negative, indicating that the adsorption process is spontaneous. Notably, the adsorption energy on STG-PtCo/EC (111) is -2.201 eV (smaller in magnitude), suggesting a weaker interaction and thus superior phosphate tolerance. In contrast, PtCo/EC (111) exhibits a stronger interaction, with an adsorption energy of -3.759 eV (larger in magnitude), which accounts for its more severe susceptibility to phosphate poisoning. The results indicate that the S-doped CELs substantially weaken H₂PO₄^- adsorption on the PtCo surface, thereby mitigating active-site poisoning [Figure 5F]. The differential charge density plot [Figure 5G] reflects an enhanced MSI between the S-doped CELs and PtCo alloys. Electrons from Pt atoms partially transfer to adjacent S atoms within the CELs and are subsequently redistributed to the surrounding C matrix and neighboring Co atoms. This electronic redistribution weakens the adsorption strength of phosphate species and promotes ORR kinetics. Figure 5H illustrates the anti-PA mechanism of the STG-PtCo/EC electrocatalyst, wherein the CELs physically isolate the PtCo alloys from PA, while the negatively charged carbon defects electrostatically repel phosphate anions, thereby suppressing their adsorption or facilitating their desorption on Pt sites.

To investigate the performance of the catalysts in HT-PEMFCs, MEAs (2 × 2 cm²) were fabricated using the synthesized electrocatalysts as cathodes, commercial Pt/C as anodes, and PA-doped PBI membranes as proton exchange membranes. To eliminate the effects on the anode side, the Pt loading at the anode was set in excess (0.3 mg_Pt cm^-2), while the cathode loading was controlled at 0.3 mg_Pt cm^-2. As shown in the H₂-air fuel cell polarization curves [Figure 6A], STG-PtCo/EC achieves a peak power density of 613 mW cm^-2, outperforming PtCo/EC (405 mW cm^-2) and commercial PtCoC (411 mW cm^-2). Under H₂-O₂ conditions, STG-PtCo/EC delivers a peak power density of 908 mW cm^-2, markedly higher than that of PtCo/EC (582 mW cm^-2) and commercial PtCoC (801 mW cm^-2) [Supplementary Figure 34]. CV curves [Figure 6B] reveal a pronounced reduction in the H-desorption peak at 0.1-0.4 V for PtCo/EC relative to STG-PtCo/EC, along with additional peaks at 0.5-0.85 V attributable to phosphate anion adsorption/desorption^[48]. These results indicate that phosphate anions extensively occupy the active sites of PtCo/EC, whereas STG-PtCo/EC demonstrates enhanced resistance to PA poisoning.

When the cathode Pt loading is further reduced to 0.1 mg_Pt cm^-2, the PTFE content in the CL significantly influences MEA performance^[3]. For STG-PtCo/EC, the optimal performance is achieved at 5 wt% PTFE in the cathode CL [Supplementary Figure 35]. Conversely, the absence of PTFE leads to performance decay due to the excessive affinity of the CL for PA, causing "acid flooding" and a reduction of three-phase interfaces^[25,64]. The H₂-air fuel cell polarization curves in Figure 6C demonstrate that STG-PtCo/EC delivers a peak power density of 355 mW cm^-2, significantly exceeding PtCo/EC (170 mW cm^-2) and commercial PtCoC (156 mW cm^-2). At 0.6 V, STG-PtCo/EC exhibits lower anode impedance (Ra, high-frequency region) and cathode impedance (Rc, low-frequency region) than PtCo/EC and PtCoC [Figure 6D and E, Supplementary Figure 36], indicating that the S-doped CELs not only protect the active sites from PA poisoning but also optimize PA distribution within the MEA, preventing PA loss at the anode CL due to excessive PA affinity at the cathode CL^[24,65]. STG-PtCo/EC and PtCo/EC show lower ohmic resistance than commercial PtCoC, attributed to the superior conductivity of the Ketjenblack EC-300J carbon support. The ultra-low-Pt-loading MEA with STG-PtCo/EC exhibits a remarkably high Pt utilization, achieving a cathode Pt power density of 3.53 kW g_Pt^-1, which reaches the advanced level of current low-Pt HT-PEMFCs [Figure 6F]. Additionally, a series of PtM (M = Fe, Ni, Cu, Mn, Zn) catalysts synthesized via the STG-assisted impregnation method [Supplementary Figures 37-39, Supplementary Table 4] exhibit similarly superior performance in ultra-low-Pt-loading MEAs [Supplementary Figure 40]. These results demonstrate the versatility of the STG-assisted impregnation method in synthesizing PA-resistant PtM/C catalysts, offering a promising strategy for the development of advanced electrocatalysts for HT-PEMFCs.

Durability testing demonstrates that the ultra-low-Pt-loading MEA with STG-PtCo/EC exhibits a voltage decay rate of 0.32 mV h^-1 over 100 h at a constant current of 200 mA cm^-2 [Figure 6G]. Post-test analysis shows an increase in ohmic resistance, likely due to PA migration from the PBI membrane to the CL^[66]. Meanwhile, the cathode impedance slightly decreased, indicating improved interfacial kinetics due to a more uniform PA distribution within the cathode CL of STG-PtCo/EC, which mitigates acid flooding and active-site poisoning while constructing rapid proton transport pathway. Conversely, a marked increase in anode impedance is observed, possibly attributed to PA poisoning of the commercial Pt/C anode catalyst, slowing the hydrogen oxidation reaction (HOR) [Supplementary Figure 41]^[24].

Contact angle measurements of PA on the electrode surface reveal that both STG-PtCo/EC and PtCo/EC with 20wt% PTFE in the CL exhibit repellent behavior toward PA [Supplementary Figure 42A and B]. PtCo/EC with 0 or 5 wt% PTFE exhibits excessive affinity for PA [Figure 6H1 and Supplementary Figure 42C]. Similarly, STG-PtCo/EC without PTFE also shows excessive PA affinity [Supplementary Figure 42D], highlighting the critical role of binders in regulating PA distribution within the CL of HT-PEMFCs. However, STG-PtCo/EC with only 5 wt% PTFE achieves a balanced PA affinity [Figure 6H2]. Figure 6I illustrates the impact of the three scenarios on cell performance. (i) Excessive PTFE increases the acid-repellent character of the CL but covers active sites, resulting in the loss of three-phase interfaces [Figure 6I1]; (ii) Strong PA affinity of the catalyst leads to active site poisoning and, in severe cases, “acid flooding” of the CL [Figure 6I2]; and (iii) The catalyst’s moderate PA-repellent property simultaneously prevents excessive PTFE coverage on Pt sites and protects the active sites from PA poisoning [Figure 6I3]. These observations indicate that the S-doped CELs optimize PA distribution around the PtCo alloys and construct efficient proton transport channels, ultimately leading to excellent fuel cell performance.

EDS spectra of CL cross-section corroborate these findings: for PtCo/EC with 5 wt% PTFE, relatively strong P and O signals indicate significant PA adsorption on the CL surface [Supplementary Figure 43A], whereas the signals are markedly weaker for STG-PtCo/EC, confirming reduced PA adsorption [Supplementary Figure 43B]. At 20wt% PTFE, PTFE aggregation extensively covers the catalyst surface, leading to a loss of active sites [Supplementary Figure 43C], whereas the absence of PTFE results in “acid flooding” [Supplementary Figure 43D]. The SEM images of the CL reveal that at a PTFE content of 20 wt%, pronounced agglomerates are formed, which obstruct the pore channels and cover the active sites, whereas no such phenomenon is observed at PTFE contents of 5 wt% and 0 wt% [Supplementary Figure 44]. Therefore, STG-PtCo/EC achieves excellent performance as a cathode electrocatalyst for ultra-low-Pt-loading HT-PEMFCs, owing to its high activity and phosphate tolerance, reduced PTFE binder requirements, and optimized PA distribution within the CL.