Protonated phosphonic acid electrodes for high power heavy-duty vehicle fuel cells

State-of-the-art automotive fuel cells that operate at about 80 °C require large radiators and air intakes to avoid overheating. High-temperature fuel cells that operate above 100 °C under anhydrous conditions provide an ideal solution for heat rejection in heavy-duty vehicle applications. Here we report protonated phosphonic acid electrodes that remarkably improve the performance of high-temperature polymer electrolyte membrane fuel cells. The protonated phosphonic acids comprise tetrafluorostyrene-phosphonic acid and perfluorosulfonic acid polymers, where a perfluorosulfonic acid proton is transferred to the phosphonic acid to enhance the anhydrous proton conduction of fuel cell electrodes. By using this material in fuel cell electrodes, we obtained a fuel cell exhibiting a rated power density of 780 mW cm–2 at 160 °C, with minimal degradation during 2,500 h of operation and 700 thermal cycles from 40 to 160 °C under load. High-temperature polymer electrolyte membrane fuel cells are promising for heavy-duty vehicle applications, but strides in performance are needed to improve their commercial viability. Here it is demonstrated that protonating phosphonic acid electrodes greatly enhances power density and durability.

H ydrogen fuel cells are attractive devices for automotive applications with benefits such as extended driving range, swift refuelling time and clean exhausts 1 . Although passenger fuel cell electric vehicles have been successfully launched, further technological innovations are needed for the next-generation fuel cell platform to evolve for heavy-duty vehicles (HDVs) [2][3][4] . One of the difficulties encountered in HDV operation is adequate heat rejection. As current low-temperature polymer electrolyte membrane fuel cells (LT-PEMFCs) operate at ~80 °C, the waste heat needs to be rejected across a 40 °C temperature difference 5 . Heat rejection in HDVs is facilitated by high cell voltage at rated power (≥0.76 V) 6 , which translates to lower power densities (<0.45 W cm -2 ). To achieve high power without the heat rejection issue, the operating temperature of the fuel cell stacks must increase to the engine coolant temperature (100 °C) and ideally up to 160 °C. At the higher operating temperature, the cost of fuel cell systems can be reduced by downsizing the fuel cell cooling system and eliminating a large humidifier or complex temperature/humidity controller unit 7 . Furthermore, enhanced CO tolerance would permit the use of reformed methanol-water mixtures that typically contain 2% CO (ref. 8 ), which would further simplify the fuelling process and mitigate the complicated and energy-intensive liquefaction of pure hydrogen transportation. If pure hydrogen fuel is replaced by methanol, the volume of the fuel, fuel tanks and reformer used to store and convert the methanol into hydrogen becomes approximately 30% of the volume-equivalent of compressed hydrogen, substantial increasing the payload space of HDVs 9 .
However, increasing the operating temperature for LT-PEMFCs has overwhelming challenges because perfluorosulfonic acid (PFSA) electrolytes require adequate hydration, which is difficult when the cell operates at >100 °C due to high water partial vapour pressure 10 . Therefore, extensive research efforts to develop polymer electrolytes for high-temperature polymer electrolyte membrane fuel cells (HT-PEMFCs) have been undertaken over the past decade. HT-PEMFCs typically use phosphoric acid-doped polybenzimidazole (PA-PBI) [11][12][13] . However, PA-PBI-based fuel cells are challenging to operate below 140 °C or during frequent startup/ shutdown 14 without suffering a loss of phosphoric acid, which makes HT-PEMFCs unfavourable for automotive applications. We previously reported that HT-PEMFCs based on quaternary ammonium biphosphate ion-pair coordination (ion-pair HT-PEMFCs) exhibited excellent phosphoric acid retention in the temperature range 80-200 °C, but the performance of the ion-pair HT-PEMFCs was poor [15][16][17] . Therefore, further performance improvement is required for the ion-pair HT-PEMFCs to be commercially viable for HDV applications 18 (Table 1).
Here we report on a protonated phosphonated ionomer that increases proton conductivity more than an order of magnitude compared with a non-protonated phosphonated ionomer. We show experimental and theoretical evidence of the protonation of phosphonic acids that is distinct from the hydrogen bonding of phosphonic acids 18 . Based on this concept, we designed protonated phosphonic acid electrodes that enable remarkable power density and are well suited for HDV fuel cells. from a PFSA that has a stronger acidic moiety [19][20][21] . For example, the pK a of pentafluorophenylphosphonic acid (PFPA) decreases from 1.3 to −0.4 (Extended Data Fig. 1a) when a proton from the PFSA (pK a = −14) is transferred to the phosphonic acid (Fig. 1a). Density functional theory (DFT) calculations indicate that the proton transfer from perfluoroethanesulfonic acid to PFPA is a spontaneous process (Δ r G = −4.7 kJ mol -1 ) with a small kinetic barrier of 5.0 kJ mol -1 (Fig. 1b). To probe the protonation of phosphonic acid, we prepared a composite ionomer by blending poly (2,3,5,6-tetrafluorostyrene-4-phosphonic acid) (PWN) and Nafion (Fig. 1c). The nature of the interactions of the composite ionomer was investigated by 31 P NMR spectroscopy (Fig. 1d). As the content of Nafion increased, the phosphorus peak broadened and four distinctive peaks evolved 22 . The 31 P NMR peaks were assigned on the basis of calculations of the change in the 31 P NMR chemical shift of PFPA when it is coordinated to perfluoroethanesulfonic acid (Extended Data Fig. 1b,c). The DFT calculations show that the 31 P NMR signal of protonated PWN exhibits an upfield shift of −1.9 ppm (peak 1) when PWN is coordinated to one sulfonic acid equivalent. In this case, the coordination is realized by the phosphonic oxygen of PWN and the SO 3 H group of Nafion. When additional hydrogen bonds form between the phosphonic POH groups of PWN and the sulfonic oxygen atoms in the sulfonic acids of Nafion, the 31 P NMR signal of PWN shifts downfield to +1.6 and +2.2 ppm (peaks P3 and P4).
Next, we investigated the Gibbs free energy for anhydride formation from the protonated phosphonic acid. It is known that anhydride formation lowers the proton conductivity of phosphonic acids 23 . First-principles calculations at the MP2/6-31G(d) level of theory 24 further show that anhydride formation from protonated PFPA is 56.5 kJ mol -1 more endergonic than from non-protonated PFPA at 160 °C (Fig. 2a). The higher Gibbs free energy for anhydride formation from the protonated PFPA may be attributed to the smaller pK a , which makes protonation of this acid energetically more difficult, as shown by the linear correlation between the pK a of various phosphonic acids 25 and the Gibbs free energy for acid anhydride formation (Fig. 2b).

Fuel cell performance enhancement of protonated electrodes
First, we examined the impact of the ion-exchange capacity (IEC) of the PWN on fuel cell performance. The membrane electrode assemblies (MEAs) prepared with the Nafion-PWN-1.8 and Nafion-PWN-3.0 ionomers outperformed the MEA using the Nafion-PWN-0.9 ionomer (Fig. 3a). Electrochemical impedance spectroscopy (EIS) analysis indicated that the Ohmic, charge transfer and mass transport resistance of the Nafion-PWN-0.9 electrode are substantially higher than those of the other electrodes (Extended Data Fig. 2). Figure 3b shows the proton conductivity of the PWN and Nafion-PWN composites in N-methyl-2-pyrrolidone (NMP) as a function of the IEC of PWN. The conductivity of Nafion was 2.0 mS cm -1 . The conductivity of PWN increased from 0.2 to 2.9 mS cm -1 as the IEC of the phosphonated polymers increased from 0 to 3.0 mequiv. g -1 . The non-zero proton conductivity of PWN-0 is probably due to residual water (~0.1%). When the IEC of the PWN was 1.3 mequiv. g -1 or below, the proton conductivity of the composite polymers was lower than the average value of the individual Nafion and PWN. In contrast, the proton conductivity of the composites exceeded the average value of the individual components when the IEC of PWN was >1.3 mequiv. g -1 . This result indicates that adding PFSA to PWN with low IEC does not increase proton conductivity, although adding PFSA to PWN with high IEC does effectively increase proton conductivity. The same conductivity behaviour was observed in dimethylsulfoxide (DMSO; Extended Data Fig. 3), although the conductivity of the ionomers in DMSO was higher because of different polarities. This conductivity behaviour suggests that only protonated phosphonic acids (corresponding to P1 of the 31 P NMR spectra in Fig. 1d) contribute to the increased proton conductivity while hydrogen-bonded phosphonic acids (corresponding to P3 and P4 in the 31 P NMR spectra) play a minor role. This is because the P1 interaction (P=O···H-O-S) enhances the acidity of the phosphonic acid, whereas the P3 and P4 interactions (P-O-H···O=S) limit the proton mobility of the phosphonic acid. This result explains the improvement in proton conductivity for Nafion-PWN-1.8 and Nafion-PWN-3.0 compared with Nafion-PWN-0.9, in which most phosphonic acid groups are hydrogen-bonded rather than protonated due to the lower IEC of the phosphonic acid.
Next, we compared the fuel cell performance using Nafion, non-protonated and protonated electrodes (Fig. 3c). We used the Nafion-PWN-1.8 ionomer with a Nafion content of 0.4 for the protonated electrode as this electrode showed the highest fuel cell performance (Extended Data Fig. 4). We achieved substantial performance improvement when the electrode binder was changed from Nafion to non-protonated and protonated ionomers, which exhibited peak power densities (PPDs) of 0.96, 1.30 and 1.67 W cm -2 , respectively. The EIS analysis indicated that the primary reason for the improved performance with the protonated electrode is the lower charge transfer resistance (R ct ; Extended Data Fig. 5a). The EIS analysis also suggested a slightly lower mass transport resistance (R mt ) of the protonated electrode (0.031 Ω cm 2 ) compared with the non-protonated electrode (0.035 Ω cm 2 ; Extended Data Fig. 5b,c), possibly due to its enhanced hydrophobicity ( Supplementary Fig. 1).
To understand the low charge transport resistance of the protonated electrode, we further measured the proton conductivity of the Nafion-PWN-1.8 ionomer. The conductivity of the ionomer increased as the content of PWN increased, but deviated from the average value (blue line), which confirms the enhancement of proton conductivity by protonation (Fig. 3d). Note that the highest conductivity deviation and the highest fuel cell performance (Extended Data Fig. 4) were observed at a Nafion content of ~0.35, at which the highest degree of protonated phosphonic acid was formed through P=O···H-O-S interactions (P1), as shown in the 31 P NMR spectra in Fig. 1d. When compared with the conductivity of the non-protonated PWN (grey line), the conductivity of the Nafion-PWN-1.8 composite was notably higher. For example, at an IEC of 0.9 mequiv. g -1 , the conductivity of the protonated phosphonic acid was 0.9 mS cm -1 , 50% higher than that of the non-protonated phosphonic acid (0.6 mS cm -1 ).
Other possible factors that may affect the electrode charge transport are related to catalyst poisoning by the ionomers. To investi-gate catalyst poisoning effects, we first evaluated the difference in the electrochemical surface areas (ECSAs) of the protonated and non-protonated electrodes using small molecules that form the ionomers (Extended Data Fig. 6a). The ECSA of the non-protonated electrode (3.6 m 2 g -1 ) was 88% of that of the protonated electrode (4.1 m 2 g -1 ). Half-cell tests using phosphoric acid or pentafluorophenylphosphonic acid (PPA) suggested that the decrease in the ECSA of the non-protonated electrode was not due to phosphoric δ (ppm)  Nafion (0.9 mequiv. g -1 )   Fig. 2 | Anhydride formation from phosphonic acids. a, Gibbs free-energy diagrams for anhydride formation at 160 °C from protonated PFPA (blue) and non-protonated PFPA (red). b, Correlation between the pK a of various phosphonic acids and the Gibbs free energy of phosphonic acid anhydride formation. The error bar of ±0.2 pK a units was determined as an r.m.s. error in the fit of the experimental pK a values to the DFT-calculated difference in the electronic energy of the protonated and unprotonated forms of an acid.
acid but to the undesirable adsorption of fluorophenyl groups (Extended Data Fig. 6b-d). The ECSA ratio of the non-protonated and protonated electrodes calculated from the results of the half-cell tests and equivalent concentration of the ionomer fragments in the electrode was 0.89, which is in good agreement with the ECSA ratio for the fuel cell electrodes.
Next, we investigated the impact of catalyst poisoning by phosphoric acid and phenyl groups on the kinetics of the hydrogen oxidation reaction (HOR; Extended Data Fig. 6e). The mass activity of the PtRu/C nanoparticles was influenced mostly by the phenyl group; the addition of 0.1 mM of PPA reduced the HOR mass activity by about 10%. Further increasing PPA caused additional HOR mass activity loss, but the impact was less notable compared with the loss during the addition of the first 0.1 mM PPA. These results indicate that the loss of HOR kinetic performance is primarily due to fluorophenyl adsorption 26,27 . The difference in HOR mass activity between the protonated and non-protonated PtRu/C anodes was estimated to be ~13%. Next, we considered the effect of phosphoric acid poisoning 28,29 on the cathode oxygen reduction reaction (ORR; Extended Data Fig. 6f). The phosphate anion concentrations in the protonated and non-protonated phosphonic acid cathodes were estimated to be 0.797 and 0.809 M, respectively (Supplementary Table 1). The difference in the ORR mass activity of Pt/C between equivalent phosphoric acid concentrations in the protonated and non-protonated cathodes was estimated to be very small because the mass activity was immediately reduced with the introduction of phosphoric acid to 1 mM and then remained steady until the maximum tested concentration of 1 M phosphoric acid. No negative impact of phenyl adsorption on the Pt catalyst was observed. This section corroborates that the increase in the kinetic performance of the protonated electrode is mainly due to the higher proton conductivity of the protonated phosphonic acids, followed by less HOR poisoning with the protonated phosphonic acid ionomer. The influence of the protonated ionomer on the ORR kinetics and mass transport is trivial.
Finally, we investigated the effect of operating temperature on fuel cell performance (Fig. 3e). As expected, the fuel cell performance, in terms of the PPD of the MEA, increased from 0.53 to 2.01 W cm -2 as the operating temperature increased from 80 to 200 °C. The change in the high-frequency resistance (HFR) of the cell between 80 and 200 °C was relatively small ( Supplementary  Fig. 2), primarily because the proton conductivity of the ion-pair membrane has little dependence on temperature 15 . Considering that the PPDs of the SnP 2 O 7 -based fuel cell 30 and the non-protonated HT-PEMFC 17 at 200 °C are 0.71 and 1.50 W cm -2 , respectively, the performance improvement shown with the protonated MEA is remarkable.
To explain the notable performance improvement with temperature, we measured the proton conductivity of the film of the ionomer cast from DMSO solution ( Supplementary Fig. 3). The conductivity of the phosphonated ionomers is known to be sensitive to humidity and phosphonic acid concentration, that is, the IEC 17 . For example, the proton conductivity of PWN-1.8 exposed to 35% RH at room temperature was 0.06 mS cm -1 at 80 °C, whereas the conductivity of the same polymer under anhydrous conditions was only 5 × 10 -4 mS cm -1 . Because of substantial phosphoric acid redistribution 31 , the ionomer conductivity in fuel cell electrodes is better estimated with a phosphoric acid-doped ionomer. Figure 3f shows that the proton conductivity of the protonated PWN (phosphoric acid-doped) membrane was more than an order of magnitude higher than that of the non-protonated phosphonic acid (phosphoric acid-doped) membrane. This result suggests that when phosphoric acids are introduced into the protonated electrode, the less acidic phosphoric acids start to associate with the protonated phosphonic acid cluster rather than excluding the phosphonated group from the acid cluster because adding more phosphoric acids increases the cluster interaction energy 32 . The proton conductivity of the protonated ionomer increased slightly with temperature (for example, from 3.3 mS cm -1 at 80 °C to 6.3 mS cm -1 at 200 °C), suggesting that enhanced catalyst activity is the most critical contributor to the high performance at 200 °C.

Fuel cell performance comparison
We compared the performance of the HT-PEMFC using a protonated ionomer (Nafion-PWN-1.8) with a commercial PA-PBI HT-PEMFC, LT-PEMFC and anion-exchange membrane fuel cell (AEMFC) under H 2 /air conditions. The rated power density of these three fuel cell systems was calculated at their optimum operating temperatures (see the Methods for calculation details). The commercial PA-PBI HT-PEMFC showed a rated power density of 0.50 W cm -2 with an HFR of 0.064 Ω cm 2 at 0.43 V and 160 °C (Fig. 4a). The non-protonated HT-PEMFC exhibited a similar rated power density with a higher HFR (0.097 Ω cm 2 ). In contrast, the rated power density of the protonated HT-PEMFC was 0.78 W cm -2 with a HFR of 0.048 Ω cm 2 at 160 °C. The kinetic performance of the protonated HT-PEMFC was higher than that of the PA-PBI HT-PEMFC at high cell voltages, although the difference was small. By comparison, the rated power density of the commercial LT-PEMFC was 0.41 W cm -2 with a HFR of 0.043 Ω cm 2 at 0.76 V and 80 °C under fully hydrated conditions (Fig. 4b), which is ~53% of the protonated HT-PEMFC at 160 °C. The rated power density of the LT-PEMFC at 100 °C increased to 0.78 W cm -2 at 100% inlet RH, but decreased to 0.29 W cm -2 at 40% inlet RH (Extended Data Fig. 7). Besides the rated power, the current density of the protonated HT-PEMFC reached 2.7 A cm -2 under anhydrous conditions, which may have a cost benefit over the LT-PEMFC, which needs complicated and expensive bipolar plates and a microporous layer for a lower current density under fully hydrated conditions. The AEMFC performed similarly to the LT-PEMFC, but with a higher HFR (0.62 Ω cm 2 ) at 80 °C under fully hydrated conditions 33 . The kinetic performance of the LT-PEMFC and AEMFC was substantially higher than that of the protonated HT-PEMFC (for example, ~0.6 W cm -2 for the LT-PEMFC versus 0.2 W cm -2 for the protonated HT-PEMFC at 0.7 V) due to the detrimental effects of phosphate chemisorption on ORR catalysts, as well as lower HOR kinetics by undesirable fluorophenyl adsorption.
We also compared the effect of the cathode Pt loading on the protonated HT-PEMFC performance using three commercial high-surface-area carbon-supported Pt catalysts (HiSPEC 9100 (Pt 60%), TEC10E40E (Pt 40%) and TEC10E20E (Pt 20%); Extended Data Fig. 8). At a given catalyst loading, the Pt/C catalyst with a higher Pt content showed higher performance, suggesting that the use of a Pt catalyst with a high Pt-to-carbon ratio is beneficial for the protonated electrode. The effect of the reactant gas flow rate on performance is shown in Supplementary Fig. 4. As expected, fuel cell performance increased as the gas flow rate increased. With a flow rate of 2,000 cm 3 min -1 , the PPD of the protonated HT-PEMFC reached 1 W cm -2 at 160 °C. Lastly, the scalability of the MEA was investigated. We obtained an identical cell performance using a medium-size MEA (25 cm 2 ; Supplementary Fig. 5). The identical fuel cell performance using the larger cell suggests that minimal performance loss can be expected with a stack of large-area cells (≥100 cm 2 ), although other technical issues, such as non-uniform reactant supply, temperature and electrode-membrane contact, still still need to be addressed in the development of larger fuel cell stacks. Figure 5 portrays the fuel cell paradigm shift that emerges from the protonated phosphonic acid ionomer. The rated power density of state-of-the-art LT-PEMFCs using advanced catalysts (orange bars) 34-36 is ~0.4 W cm -2 at 80 °C. When compared with these state-of-the-art LT-PEMFCs, the rated power density of the protonated HT-PEMFC (blue bar) at 160 °C is approximately twofold higher. Although the rated power density of the LT-PEMFCs could be increased to ~1.0 W cm -2 at 95 °C, the durability of the LT-PEMFCs at this operating temperature is a concern 36 . The rated power density of the protonated HT-PEMFC also increased as the temperature increased, and achieved the LT-PEMFC benchmark performance at 200 °C. AEMFCs (dark-blue bars) have a similar rated power density to the LT-PEMFCs at 80 °C, but lower values at 95 °C (refs. 33,37,38 ). PA-PBI HT-PEMFCs (dark-grey bars) have a rated power density of 0.42-0.53 W cm -2 at 160 °C (refs. 13,[39][40][41] ). However, the rated power density did not increase much at 200 °C due to possible evaporation of the phosphoric acid. At 120 °C, all fuel cells suffered from relatively low performance (Extended Data Fig. 9).

Durability of the protonated Ht-PEMFC
We evaluated the durability of the protonated HT-PEMFC under three operating conditions: constant current density at 80 and 160 °C, and thermal cycling between 40 and 160 °C at 0.5 V. The durability of HT-PEMFCs at 80 °C is critical to the rapid startup of fuel cell vehicles. The performance of the protonated HT-PEMFC at 80 °C and 0% inlet RH was reasonably high (PPD = 0.35 W cm -2 and HFR ≈ 0.15 Ω cm 2 ; Extended Data Fig. 10). The low-temperature durability of the protonated HT-PEMFC was evaluated for 200 h at a constant current density of 0.2 A cm -2 and under a high H 2 /air stoichiometry of 72:30.
We used such a high stoichiometry because the performance of PA-PBI HT-PEMFC rapidly decreases at high stoichiometry 42,43 . In addition, a high-stoichiometry capability of HT-PEMFCs is desirable for stack cooling, diluted hydrogen from methanol reforming and flow tolerance. Under these conditions, no cell voltage loss or HFR growth was observed for the protonated HT-PEMFC (Fig. 6a). In contrast, the commercial PA-PBI HT-PEMFC exhibited rapid cell voltage decay (10 mV h -1 ) accompanied by an increase in the HFR, and the cell stopped working after 40 h of operation, probably due to the loss of phosphoric acid over time.
The durability of the protonated HT-PEMFC was also evaluated at 160 °C at a constant current density of 0.6 A cm -2 and H 2 /air stoichiometry of 24:10 (Fig. 6b). In a previous study, we demonstrated a non-protonated HT-PEMFC with a stable performance for ~550 h We further evaluated the durability of the protonated HT-PEMFC using a thermal cycling AST protocol to investigate the impact of thermal stress during the fuel cell startup/shutdown stage 46 as well as the impact of dynamic current generation 47 . In this AST protocol, we applied deep thermal cycling (40-160 °C) at 0.5 V to mimic the temperature change during the startup/shutdown of HT-PEMFCs ( Supplementary Fig. 7a). Compared with our previous AST protocol (thermal cycling of 80-160 °C at a constant current density 0.15 A cm -2 ) 17 , this test protocol was more rigorous. The results show that during the first 150 cycles, the current density at 160 °C increased from 0.77 to 1.05 A cm -2 (Fig. 6c). This behaviour is probably due to cell break-in, when phosphoric acid redistribution occurs to enhance the catalyst activity. After 150 thermal cycles, the cell stabilized until the test was finished. The current density decay rate during thermal cycles 150 to 700, calculated from the average value of ten consecutive current densities, was 9.7 µA cm -2 cycle -1 . The corresponding current density decay after 10,000 startup/shutdown cycles was <100 mA cm -2 . No notable change in HFR was observed after the first 150 thermal cycles. For comparison, a com-mercial PA-PBI HT-PEMFC was subjected to the same AST and showed rapidly degrading behaviour during the first seven cycles and the cell became inoperable (Supplementary Fig. 7b), confirming that the PA-PBI HT-PEMFC is difficult to run with frequent startup/shutdown cycles under load. Titrations revealed that the PA-PBI membrane lost 58% of its initial phosphoric acid after seven cycles, whereas the ion-pair membrane only exhibited a negligible 7% acid loss after 700 cycles. This reaffirms the superior phosphoric acid retention capability of the protonated HT-PEMFC, which also hold great promise for use with reformate gases accompanied by trace water vapour in anode flows. More thorough research on fuel cell durability under dynamic drive cycles, including startup/shutdown, large voltage swings and freeze-thaw operations, may be further required for commercial fuel cells.

Conclusions
The excellent performance and durability of the protonated HT-PEMFC presents opportunities in HDV fuel cell applications that require high device power and robustness. Our material platform combined with the use of a protonated phosphonic acid ionomer enables ion-pair HT-PEMFCs that operate not only with the same overall thermal balance of the internal combustion engine, but also generate substantially higher rated power than state-of-the-art LT-PEMFCs, making it well-suited for HDV fuel cell applications. Furthermore, the low degradation rate of the fuel cells under steady-state and thermal cycling conditions is promising for meeting the HDV lifetime target 48 , that is, 12 years and 1.6 million kilometres of operation. Therefore, this study fills several research gaps in the development of HT-PEMFCs for HDV applications. Additional work to address the remaining challenges for the full consideration of protonated HT-PEMFCs for HDV applications includes improving relatively low performance at high cell voltages, durability under a dynamic driving cycle, partial humidification as well as practical/technical challenges with scaling up the MEA to a full-size fuel cell stack.

Materials.
The PA-PBI membrane/PTFE-bonded electrode MEAs (Celtec P1100) produced by the polyphosphoric acid process were supplied by BASF Fuel Cells. For the LT-PEMFCs, Nafion membranes (Nafion NR-211, 25.4 µm thickness) and Nafion D2020 dispersions were purchased from Ion Power. For the anode and cathode, 35.5% Pt on high-surface-area carbon (0.1 mg Pt cm -2 , Tanaka, TEC10E40E) was used. Carbon paper gas diffusion layers (GDLs; SGL 29BC) were used. We also used a Gore MEA for comparison.
For the fabrication of the protonated and non-protonated HT-PEMFC MEAs, commercial 60% Pt/C (HiSPEC 9100) and 75% PtRu/C (Pt/Ru ratio = 2:1, HiSPEC 12100) were purchased from Alfa Aesar and Johnson Matthey, respectively. The Pt loading on the anode was 0.5 mg Pt cm -2 . The Pt loading on the cathode was in the range 0.1-0.7 mg Pt cm -2 . We also used Tanaka Pt/C catalysts for the study of the ORR catalyst: 35.5% Pt on high-surface-area carbon (TEC10E40E) and 19.4% Pt on high-surface-area carbon (TEC10E20E). The GDLs for the HT-PEMFC electrodes were CeTech W1S1009.
Preparation of QAPOH-PA ion-pair membranes. The QAPOH membranes were synthesized by an irreversible Diels−Alder reaction as described by Hibbs et al. 49,50 .
First, brominated alkyl ketone-functionalized Diels-Alder poly(phenylene) (DAPP) was synthesized by the reaction of DAPP with bromohexanoyl chloride, followed by removal of the ketone group using triethylsilane and trifluoroacetic acid. The bromide polymer was cast into films from chloroform solution. These films were aminated by immersion in trimethylamine solution (45% w/w in water) for 48 h. The quaternized membranes were thoroughly washed with deionized water to remove residual trimethylamine. The membranes with bromide counter ions were converted into the hydroxide form by immersing in 0.5 M NaOH at 80 °C, followed by rinsing with deionized water at room temperature. After blotting excess water, the hydroxide form of the quaternized membranes were soaked in 85 wt% aqueous phosphoric acid solution for 50 h at room temperature. The QAPOH-PA membranes were used after removing the excess phosphoric acid from the membrane surface by blot-drying. The areal amount of phosphoric acid in the QAPOH-PA after the imbibing process was 0.06 mmol cm -2 .
Synthesis of the phosphonated polymer PWN. PWN was synthesized by phosphonation of poly(pentafluorostyrene) (PFS) 51 . For PWN-1.8, PFS (100 g) was dispersed in 400 ml dimethylacetamide and tris(trimethylsilyl) phosphite (200 g, 670 mmol) was added slowly. The reaction solution was then heated to 160 °C and magnetically stirred overnight. After completion of the reaction, the warm mixture was precipitated in 2 l water and the precipitate collected by filtration. The resulting white powder was refluxed in water three times for 30 min each, changing the water each time, followed by boiling in 2 wt% phosphoric acid solution.
Washing with water until neutral and drying at 140 °C yielded the phosphonated polymer PWN-1.8 with a 66% degree of phosphonation (yield: 99%). The degree of phosphonation was controlled by the amount of tris(trimethylsilyl) phosphite. 1 H NMR spectroscopy. 1 H NMR spectra were recorded using a Bruker Avance 500 spectrometer (500 MHz) in [D 6 ]DMSO. The chemical shifts of the 1 H NMR spectra were referenced to tetramethylsilane at 0 ppm as internal reference. Titration. The IECs of the PWN samples were determined by acid-base titration according to the following procedure. All samples were dried at 100 °C for 12 h before titration to obtain a dry mass. Each sample (H + form) was immersed in 1.0 M NaCl solution and stirred at room temperature for 24 h. Then, the solution was titrated with 0.5 M NaOH solution using two to three drops of a methyl orange aqueous solution (0.1%) as the indicator. The IEC was calculated from the dry mass and the amount of NaOH used in the titration.

DFT calculations.
The pK a value of pentafluorophenylphosphonic acid was calculated from a linear regression fit to the deprotonation energies: where E − A is the DFT-calculated electronic energy of the deprotonated acid and E HA is the DFT-calculated energy of the protonated acid 25 . Also, in the equation, a is a slope of the line, and b is the y intercept. As explained in our previous work 17 , we derived values of a and b using a dataset of nine experimental pK a values. For each of these acids, we performed a full geometry optimization for both HA and the anion Ausing the solvation model based on density (SMD) 52 with water as solvent at the M06L/6-311++G(d,p) 53 level of theory using Gaussian 09 (revision C.01) 54 . Linear fitting of the data in Extended Data Fig. 1 resulted in a line fitted by the following equation: with R 2 = 0.90 and a root-mean-square error of 0.2 pK a units. ΔE = ( E − A − E HA ) for pentafluorophenylphosphonic acid was calculated to be 273.34 kcal mol -1 , resulting in pK a = 1.3 ± 0.2.
for protonated pentafluorophenylphosphonic acid was calculated to be 257.92 kcal mol -1 , resulting in pK a = -0.4 ± 0.2. 31 P NMR chemical shifts were calculated using a gauge-including atomic orbital (GIAO) method, as implemented in Gaussian 09 (revision C.01) 54 . In this case, the M062X/6-311+G(2d,p) level 55 of theory was used. The geometries of pentafluorophenylphosphonic acid and pentafluorophenylphosphonic acid coordinated to an increasing number of Nafion fragments were first optimized, after which the 31 P NMR chemical shifts were calculated.
The change in the Gibbs free energy upon anhydride formation was calculated at the MP2/6-31 G(d) level of theory, as implemented in the Gaussian 09 (revision C.01) 25,51,52 . In all cases, the structures of the acids and anhydrides were optimized and the change in the Gibbs free energy was calculated by performing a frequency analysis at 160 °C.
Proton conductivity. The solution ionic conductivity of the PWN and Nafion-PWN composite ionomers was measured using a custom liquid cell with 1-cm-diameter stainless-steel electrodes separated by 1 cm and encased in polypropylene casing. The Nafion solution was prepared by the direct dissolution method using the proton form Nafion 212 (Nafion water content: 6 wt%) 56 . All samples were prepared at a concentration of 2.5 or 5 wt% in anhydrous DMSO or NMP to mitigate solvent effects. The solution conductivity was measured by AC impedance spectroscopy (Solartron 1260 gain phase analyser) over a frequency range of 1 Hz to 1 MHz.
For the in-plane film conductivity measurements of the PWN and Nafion-PWN composite membranes, we prepared the membranes by solution casting. First, PWN polymers were dissolved in DMSO at 5 wt% by sonication at 40 °C. For the composite membranes, a commercial Nafion dispersion (D520, 5 wt%) (Ion Power) was added at the desired blend ratio. The clear composite solution was poured onto a glass Petri dish and placed on a hot plate at 80 °C. After slow evaporation of the solvent overnight, the Petri dish with the membrane was placed in a vacuum oven at 80 °C overnight to evaporate the residual solvent completely. PWN films were fabricated by directly pouring the DMSO solution onto a Petri dish and drying at 80 °C overnight in a convection oven. After cooling to room temperature, we detached the membrane from the glass substrate in deionized water. The membranes were stored in water. Membrane conductivity was measured in a small window cell with an opening length of 0.5 cm. Membranes were placed between two platinum-coated electrodes and clamped tight. The window cell was placed in a convection oven, and the oven temperature was slowly increased to 120 °C over 30 min. The membrane's impedance was measured by AC impedance spectroscopy (Solartron 1260 gain phase analyser). The temperature was then increased to 240 °C, and the conductivity was measured during cooling at 20 °C intervals until the temperature reached 80 °C. + F + C). Samples for transmission electron microscopy (TEM) analysis were prepared by microtoming epoxy-embedded small sections (1 cm × 0.5 cm) of the electrodes. The epoxy resin was prepared using a 1:1 (wt/wt) mixture of trimethylolpropane triglycidyl ether resin (Sigma-Aldrich) and 4,4′-me thylenebis(2-methylcyclohexylamine) hardener (Sigma-Aldrich), and the sections embedded during overnight polymerization at 60 °C. The thin (~100 nm) electrode cross-sections were placed on 200 mesh Cu/Pd grids. A Talos 200 kV transmission electron microscope (Thermo Fisher) with four Super-X silicon drift detectors for energy-dispersive spectroscopy (EDS) was used for TEM as well as for high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) imaging and EDS elemental mapping. The EDS mapping was performed at ×5K and ×79K magnifications, with a 1,000 μs dwell time for one cycle and an electron dose of 2.34 × 10 4 enm -2 . Map processing, elemental analysis and visualization were completed using ESPIRIT 1.9 analytical software (Bruker).

MEA fabrication.
HT-PEMFC MEAs were fabricated from catalyst inks containing Pt/C or PtRu/C catalysts and single or composite ionomer dispersions (solid content: 5 wt%). For the PWN and Nafion ionomer dispersions, we used NMP as dispersing agent. For the anode of the ion-pair HT-PEMFC, a PtRu/C catalyst (Pt/ Ru ratio = 2:1; HiSPEC 12100) was used to minimize adverse phenyl adsorption of the ionomer (Supplementary Fig. 8). The composite ionomer dispersions were prepared by mixing the PWN and Nafion ionomer dispersions in different ratios to obtain Nafion contents of 0, 0.3, 0.4, 0.5, 0.6 and 1.0. The catalyst ink was sonicated in an ultrasonic bath for 1 h to achieve a uniform dispersion and painted onto the GDLs (W1S1009, CeTech) by hand until the Pt loading reached 0.5 and 0.7 mg cm -2 for the anode and cathode, respectively (confirmed by X-ray fluorescence (XRF)). After the hand painting, the gas diffusion electrodes (GDEs) were placed on a vacuum plate for 10 min at 70 °C to remove residual dispersion agent within the electrodes. The catalyst-coated GDEs were sandwiched with a QAPOH-PA PEM (35 µm thick). The active area of each MEA was 5 or 25 cm 2 . HAADF-STEM images and the corresponding EDS elemental maps (Pt + F + C) of the GDEs indicated that the protonated ionomers were uniformly distributed within the electrodes with greater ionomer distribution on the catalyst nanoparticles ( Supplementary Fig. 9).
As a control, a Nafion-based LT-PEMFC MEA was fabricated with a Nafion 211 membrane and Nafion D2020 (20 wt%) ionomer (Ion Power). The ionomer-to-carbon ratio by mass was 0.9. The Pt/C catalyst and D2020 ionomer were dispersed in a mixture of deionized H 2 O and 1-propanol (4:3, volume ratio) by stirring at 700 r.p.m. for 4 h, and then in an ultrasonic bath for 20 min. For the cathode, the Pt/C catalyst was deposited on the membrane by ultrasonic spray-coating. For the anode, Pt/VC (Vulcan XC-72 carbon, TEC10V20E) was used for all experiments and was also spray-coated onto the membrane after the cathode had been spray-coated. The Pt loading was fixed to around 0.1 mg Pt cm -2 (±10%) for both the cathode and anode (confirmed by XRF). The active area of the catalyst-coated membrane was 5 cm 2 . For the LT-PEMFC MEAs, SGL 29BC carbon paper GDLs were used for all MEA tests. Polyurethane gaskets were used for Nafion-based LT-PEMFCs, and reinforced PTFE gaskets were used for HT-PEMFCs.
Fuel cell performance and durability. The fuel cell performance of the MEAs was measured using a fuel cell test station (Fuel Cell Technologies). For the HT-PEMFCs, the polarization curves and HFRs of the MEAs were obtained at temperatures ranging from 80 to 200 °C. H 2 and air (or O 2 ) were both supplied at a rate of 500 cm 3 min -1 , unless noted otherwise in the manuscript. The cell current density and HFR were measured every minute without external humidification. For the Nafion-based LT-PEMFC control MEA, the polarization curves and HFRs were obtained at temperatures ranging from 80 to 100 °C. H 2 and air were both supplied at a rate of 500 cm 3 min -1 .
Steady-state H 2 /air durability tests were performed at 80 and 160 °C. At 80 °C, the cell voltage and HFR were measured for 200 h at a constant current density of 0.2 A cm -2 . The H 2 /air stoichiometry was 72:30 and the backpressure was 148 kPa. At 160 °C, the cell voltage and HFR were measured for 2,000 h at a constant current density of 0.6 A cm -2 . A high gas flow of 500 cm 3 min -1 was used for both H 2 and air to accelerate fuel cell degradation 13 . The H 2 /air stoichiometry was 24:10 with the same backpressure of 148 kPa. The thermal cycling AST involved deep triangle thermal cycles from 40 to 160 °C with a ramp of 15 °C min -1 under anhydrous conditions to simulate cold startup cycles. A constant voltage of 0.5 V was applied, and the current density was monitored as a function of time. The HFR was measured when the cell temperature reached either 40 or 160 °C.
Rated power calculation. The heat rejection requirement has been expressed as a constraint for which a nominal 90 kilowatt-electric (kW e ) fuel cell stack should have waste heat Q/ΔT less than 1.45 kW °C -1 , where ΔT is the initial temperature difference between the stack coolant outlet temperature (T c ) and the ambient temperature (T a ), and Q is defined as the stack power (90 kW e ) × (1.25 V -voltage at rated power)/(voltage at rated power) (ref. 6 ). The rated power was calculated to meet the Q/ΔT = 1.45 kW °C -1 target at the cell voltage, that is, 77.6/(22.1 + T).
Electrochemical impedance spectroscopy. EIS of the HT-PEMFC was performed using a Biologic SP-200 device after measuring the polarization curve and HFR. The spectra were recorded by sweeping frequencies over a range from 1 MHz to 0.1 Hz at a DC voltage of 0.8 and 0.6 V and a constant current density of 2.0 A cm -2 . The experimental spectra were fitted to equivalent circuits by employing EC-Lab software. The equivalent circuit applied here consisted of Ohmic resistance (R Ohm ) in series with two parallel constant phase elements (CPEs), namely CPE ct /R ct for charge transfer resistance and CPE mt /R mt for mass transport resistance 57,58 .
Cyclic voltammetry and ECSA calculation. To measure the ECSA, cyclic voltammetry (CV) was conducted using the Biologic SP-200 instrument with a single cell. The CV was performed in the voltage range 0.05-1.2 V at a scan rate of 50 mV s -1 and a cell temperature of 80 °C. H 2 and N 2 were provided to the anode and cathode at flow rates of 500 cm 3 min -1 , respectively. The ECSA was determined from the cyclovoltammograms using Eq. (3), where Q H is the charge integration of the hydrogen underpotential deposition region, the anodic peak between 0.05 and 0.4 V, q H is the unit charge of 210 μC cm -2 for hydrogen desorption on a polycrystalline Pt surface and m is the Pt loading 59 : Contact angle measurement. The hydrophobicity of the GDEs was examined by measuring the contact angle of a deionized water droplet using a Ramé-Hart Model 500 goniometer.
where n is the number of moles of electrons transferred in the half reaction, F is the Faraday constant (C mol −1 ), A is the electrode area (cm 2 ), D is the diffusion coefficient (cm 2 s −1 ), ω is the angular rotation rate of the electrode (rad s −1 ), υ is the kinematic viscosity (cm 2 s −1 ), C is the analyte concentration (mol cm −3 ). The ORR activity current (I k ) was calculated from the relation: where I is the baseline-subtracted and IR-compensated LSV current at 0.9 V.
HOR activity measurement and calculation. LSV was conducted in a cell saturated with H 2 gas in the voltage range of -0.05 to 0.50 V at a scan rate of 10 mV s -1 and a rotating speed of 2,500 r.p.m. The exchange current density (j 0 ) for the H 2 reaction was determined using the approximate Butler-Volmer equation, j = j 0 (hF/RT), in a linear region of the LSV curve (−0.010 to 0.010 V) after IR compensation was performed for the LSV curve. The HOR mass activity was determined by dividing j 0 by the Pt loading in the electrode.
Calculation of the concentration of phosphoric acid and phenyl groups. After testing the initial performance of the HT-PEMFCs, the cell was cooled to room temperature and the anode and cathode GDEs were carefully separated from the membrane. The phosphoric acid concentration in the electrodes was determined by acidbase titration: 5-cm 2 electrode samples were titrated with a 0.05 M NaOH solution using methyl orange as indicator. The phosphoric acid concentration in the electrodes was calculated from the following equation: where V NaOH is the volume of NaOH solution (cm 3 ), C NaOH is the concentration of NaOH (M), and V electrode is the volume of the electrode (cm 3 ). The change in the weight of the electrodes before and after the titration (ΔW gravimetric )was also measured to estimate the amount of water within the electrodes: Water concentration = { ΔWgravimetric − (V NaOH × C NaOH × M W,PA ) } /MW,water V electrode (7) where M W,PA and M W,water is the molecular weight of phosphoric acid and water, respectively. The phenyl group concentration was calculated from the catalyst ink composition and the chemical structure of the ionomer.

Statistics and reproducibility.
To ensure that representative image and elemental mapping data were obtained, we repeated the experiments as detailed below. TEM images for particle size distribution were collected at a magnification of ×190000. Usually six or seven images, or 200 particles, were recorded.
For TEM images at ×280 magnification, just two or three images were collected to provide an overview.
Elemental EDS mapping at ×5000 magnification, three images per catalyst layer were collected.
For elemental EDS maps at ×79000 magnification, usually two images were collected per catalyst layer.

Data availability
The data supporting the findings of this study are available within the paper, Extended Data and Supplementary Information. Source data are provided with this paper.