Polymer electrolyte membrane fuel cells (PEFCs) are considered as one of the best power sources to realize green and efficient energy generation for many commercial applications. PEFCs mainly rely on the use of platinium (Pt) metal because it is the most efficient catalyst for an efficient oxygen reduction reaction (ORR), in addition to its high stability under the strong PEFC acidic conditions1,2. However, Pt is a very expensive and rare. Thus, it is considered a major hinderance to the global commercialization of PEFCs1,3,4.

A widely accepted solution to reduce the cost of PEFCs is the use of a minimal amount of Pt in the electrocatalyst and thus we need to maximize the mass activity of Pt in the electrocatalyst. The major strategy of increasing the mass activity of Pt is to add a second inexpensive metal to Pt to form either a Pt core-shell5,6,7,8 or a Pt alloy in order to increase the surface area of the Pt per unit amount9,10,11,12,13,14. However, it has been pointed out that under the strong acidic environment of the PEFCs, a dissolution process of the second metal occurs resulting in a low fuel cell durability15,16,17,18,19,20. Another promising strategy towards the increase of Pt mass activity is to reduce the particle size of Pt in order to increase the surface area of the Pt nanoparticles, which can lead to an increase in the mass activityt of Pt21,22,23. The simplest technique in decreasing the size of Pt nanoparticles is to minimaze the amount of Pt-ions fed into the Pt reduction process in order to regulate the growth of the nanoparticles. Successful examples of the Pt size reduction were reported for carbon black (CB)23 and graphene oxide22. In both systems, since polar functional groups such as carboxylic and/or hydroxyl groups served as effective binding sites for the Pt ions, the nucleation step occurs homogeneously and, thus, the final Pt size was regulated by simply controlling the feed of Pt ions into the solutions.

However, the oxidized sites of the carbon support accelerate the degradation of the support material, inducing significant aggregation of Pt nanoparticles24. Therefore, novel binding sites with an effective binding of Pt ions on the carbon supports are required for the practical and the commercial use of PEFCs. Recently, we have reported a unique approach for the growth of Pt nanoparticles on non-oxidized graphitic carbon, in which polybenzimidazoles (PBIs) are strongly attached to the surface of CNTs in nanometer thickness, providing effective growth sites for Pt ions25,26,27,28,29,30. Due to the effective binding of Pt ions on PBIs via a coordination mechanism, the homogeneous growth of Pt nanoparticles was realized and thereby a homogeneous dispersion of Pt catalyst on the supporting material together with a narrow diameter distribution were achieved27. In the present study, we utilized the advantages of our effective binding technique of the Pt ion for the deposition of size-controlled Pt nanoparticles onto the non-oxidized CNT in order to increase the mass activity of Pt in PEFCs.

Poly[2,2′-(2,6-pyridine)-5,5′-bibenzimidazole] (PyPBI; Fig. 1) was used to wrap the CNTs (denoted CNT/PyPBI) because of the following three main reasons: i) this polymer efficiently solubilized the CNTs through π−π interactions, ii) it is a proton conductor material that can improve the proton conductivity of the electrocatalyst and iii) it has N-atoms that work as the binding sites for the growth of Pt ions by coordination between the PyPBI and Pt ion25. The Pt was loaded on CNT/PyPBI to prepare Pt composite (denoted CNT/PyPBI/Pt) as displayed in a schematic illustration (Fig. 1)25. For the highest loading of Pt metal, Pt ions corresponding to 50 wt% of the final composite weight was added (denoted CNT/PyPBI/Pt50). The transmission electron microscopic (TEM) image of CNT/PyPBI/Pt50 exhibits a homogeneous loading of Pt metal with diameters of 3.7 ± 0.6 nm (Fig. 2a). The decrease of the feeding amount of the Pt ions to 1/2, 1/5, 1/10 and 1/20 (the obtained composites were denoted CNT/PyPBI/Pt25, CNT/PyPBI/Pt10, CNT/PyPBI/Pt5 and CNT/PyPBI/Pt2.5, respectively) led to a gradual decrease in the Pt size to 3.6 ± 0.6, 3.3 ± 0.6, 2.7 ± 0.6 and 2.3 ± 0.5 nm for CNT/PyPBI/Pt25, CNT/PyPBI/Pt10, CNT/PyPBI/Pt5 and CNT/PyPBI/Pt2.5, respectively (Figs. 2b~2e). Importantly, the dispersions of the Pt metal on the CNTs were highly homogeneous, as observed from the TEM images (Fig. 2; left images) and the Pt size distributions for all the composites were very narrow (Fig. 2; right). Such an excellent controllability of Pt deposition is probably derived from the rapid nucleation of Pt seed on the PyPBI-wrapped CNTs through the coordination of the Pt ions with the benzimidazole unit of the PyPBI25.

Figure 1
figure 1

Schematic drawing describing the preparation of the CNT/PyPBI/Pt composites.

A serious of CNT/PyPBI/Pt composite were prepared by changing the feeding amount of Pt salt. Chemical structure of PyPBI is presented in the dotted frame.

Figure 2
figure 2

Pt diameter and distribution analyses of the CNT/PyPBI/Pt.

The TEM images (left) and diameter histograms (right) of (a), CNT/PyPBI/Pt50, (b), CNT/PyPBI/Pt25, (c), CNT/PyPBI/Pt10, (d), CNT/PyPBI/Pt5 and (e), CNT/PyPBI/Pt2.5 revealed the diameter of Pt and density of Pt loading were decreased by decreasing the feeding amount of Pt salt with maintaining the homogeneous dispersion of Pt on the CNT/PyPBI. Scale bars, 50 nm.

Fig. 3 shows representative scanning electron microscope (SEM) and scanning transmission electron microscope (STEM) images of the CNT/PyPBI/Pt50 (Figs. 3a~3c), CNT/PyPBI/Pt10 (Figs. 3d~3f) and CNT/PyPBI/Pt2.5 (Figs. 3g~3i). Comparison of SEM and STEM image allowed us to distinguish the Pt nanoparticles in the front side and in the backside and clarified the decreasing of loading density as decreasing of the Pt feeding. As emphasized by the color contrast in Figs. 3b, 3e and 3h, we successfully visualized the coating of the PyPBI layers on the CNTs (the green color) as well as small uncoated bare surface spots based on the difference in the scattering efficiency of the PyPBI area and the bare CNT surfaces for the secondary electrons. To the best of our knowledge, this is the first example of visualizing a polymer wrapped on the CNTs by SEM measurements. Importantly, the Pt nanoparticles were observed only on the PyPBI regions, which provided clear evidence that the loading was ocuured on the PyPBI layer.

Figure 3
figure 3

Electron microscope observations of the CNT/PyPBI/Pt.

SEM (a, b, d, e, g, h) and STEM (c, f, i) images of CNT/PyPBI/Pt50 (a~c), CNT/PyPBI/Pt10 (d~f) and CNT/PyPBI/Pt2.5 (g~i).

The loading efficiency of Pt was estimated based on the final residues in the thermal gravimetric analysis (TGA) shown in (Fig. 4a), where the loaded Pt was found to be 50, 26, 9, 6 and 4 wt% for CNT/PyPBI/Pt50, CNT/PyPBI/Pt25, CNT/PyPBI/Pt10, CNT/PyPBI/Pt5 and CNT/PyPBI/Pt2.5 composites, respectively. These values correspond to ~100% of the fed Pt ions. From all above results, we concluded that the effective binding of Pt ion onto PyPBI leads the quantitative and homogeneous loading of Pt onto the CNT/PyPBI, thereby the excellent controllability was realized.

Figure 4
figure 4

TGA, XPS and electrochemical measurements of the CNT/PyPBI/Pt as well as their numerical data.

(a), TGA curves of CNT/PyPBI/Pt50 (gray), CNT/PyPBI/Pt25 (purple), CNT/PyPBI/Pt10 (blue), CNT/PyPBI/Pt5 (orange) and CNT/PyPBI/Pt2.5 (green). TGA curve of CNT (dotted line) is also displayed as a reference. (b), XPS narrow scans of Pt 4f for CNT/PyPBI/Pt50 (gray), CNT/PyPBI/Pt25 (purple), CNT/PyPBI/Pt10 (blue), CNT/PyPBI/Pt5 (orange) and CNT/PyPBI/Pt2.5 (green). (c), The relation between the Pt loaded on the polymer-wrapped CNT composites and the binding energy of Pt 4f. (d), CV of CNT/PyPBI/Pt50 (gray), CNT/PyPBI/Pt25 (purple), CNT/PyPBI/Pt10 (blue), CNT/PyPBI/Pt5 (orange) and CNT/PyPBI/Pt2.5 (green). (e), Plot of ECSA values as a function of the Pt loading. (f), LSV of the CNT/PyPBI/Pt50 (gray), CNT/PyPBI/Pt25 (purple), CNT/PyPBI/Pt10 (blue), CNT/PyPBI/Pt5 (orange) and CNT/PyPBI/Pt2.5 (green). (g), list of particle size, ECSA, TSA and utilization efficiency for the CNT/PyPBI/Pt.

The X-ray photoelectron spectroscopy was used to confirm the composite structure of the electrocatalyst and to collect information about the bonding between N-atoms and Pt ions. As observed from the XPS measurements, the N 1s peak at 400 eV31 due to PyPBI together with the Pt 4f doublet peaks at 71.1 and 74.4 eV attributable to Pt 4f7/2 and Pt 4f5/2, respectively32, are observed in all the prepared composites (Supplementary, Fig. S1). The nitrogen/carbon ratio determined from XPS was recorded to be ~5%, indicating a thin layer of PyPBI on the surface of the CNTs. Together with the decrease in the intensity of the Pt 4f peaks due to the lowering of the Pt loading amount in the composites (Fig. 4b), we recognized an upshift in the Pt 4f peaks with the decrease in Pt size as plotted in Fig. 4c. A similar shift of the Pt 4f peak towards a higher binding energy was also reported for small Pt clusters and this upshift has been attributed to the increase in the number of Pt atoms with low coordination numbers22,33,34.

The electrochemically active surface areas (ECSAs) of Pt nanoparticles in the composites were evaluated by cyclic voltammetry (CV) measurements of loading the same amount of Pt on GC. However, the catalyst layer with the same mass of Pt was unstable on GC during the ORR measurements because the rotation of the GC electrode strongly affected the diffusion process due to the differences of the film thicknesses of the catalysts. Therefore, it was important to measure the ORR of the catalysts with the same film thickness (the same amount of composites). As shown in Fig. 4d, all the composites exhibited clear peaks due to hydrogen adsorption and desorption at 0–0.3 V (vs. RHE). The ECSAs were determined using the hydrogen desorption peak. The determined ECSA values were plotted as a function of the amount of Pt loaded onto the CNT/PyPBI composites (Fig. 4e). As observed, the decrease in the Pt diameter led to a dramatic increase in the ECSA. The highest ECSA value was recorded to be 114.5 m2/gPt for the CNT/PyPBI/Pt2.5 composite, which is comparable to that of Pt of a similar size on graphene oxide22 and CB35, meaning the PyPBI layer has served as a growth site for Pt similar to other oxidized carbon materials. In order to evaluate the Pt utilization efficiency of the prepared electrocatalysts, the ratio of the ECSA to the calculated theoretical surface area (TSA) of Pt was evaluated. The obtained Pt utilization efficiency values are summarized in Fig. 4g. Interestingly, the reduction of the Pt size has led to an increase in the Pt utilization efficiency. The value reached as high as 94% for the smallest Pt particles in the CNT/PyPBI/Pt2.5 electrocatalyst. In addition, the increase in the utilization efficiency revealed a decrease in the contact area of Pt with the PyPBI layers and also has indicated an increase in the exposed surface of Pt as the Pt diameters decreased.

Watanabe et al. pointed out that, when the Pt nanoparticles are too close to each other (roughly below 1.8 nm for oxygen gas and 2.0 nm for air), a poor diffusion of oxygen is obtained since not all the Pt surface areas are expected to be available for the oxygen reduction reaction (ORR). Consequently, not only the size of Pt, but also the Pt inter-particle distance plays an important role for the increase of the Pt mass activity23. In our composites, the distributions of the Pt inter-particle distances measured from the TEM images have shifted to higher values (>18 nm) when the loading amounts of Pt were decreased (Supplementary, Fig. S2). Indeed, it was obvious from the ORR (Fig. 4f) that the current density per unit weight of Pt increases with the decrease in the diameter of the Pt nanoparticles.

Furthermore, we have examined the effect of the Pt inter-particle distance and the diameter of the Pt nanoparticles on the Pt-dispersion durability of the composites. The CNT/PyPBI/Pt50 and CNT/PyPBI/Pt2.5 composites were measured by an in situ SEM/STEM dual monitoring system which is equipped with a gas injection system while heating at 200°C. In these experiments, the composites were heated under vacuum at 3.3 × 10−5 Pa, then atmospheric air (RH = 42%, temp; 24°C) was introduced until the pressure reached 1.2 × 10−2 Pa (for pressure data, Supplementary, Fig. S3) to accelerate the particle activity36. The Pt nanoparticles were found to be quite stable at 200°C under high vacuum conditions for both composites but when the air was introduced, the migration of the Pt on the surfaces of the CNTs gradually started to form bigger particles by Pt fusion on the surface of the CNTs (Supplementary, Movies 1 and 2). Such phenomena are in sharp contrast to that of Pt on CB, in which the Pt is inserted into the CB support and then a Pt agglomeration process occurs inside the carbon36. This difference in the Pt behaviour on the CNT and CB supports would be due to the higher stability of the CNTs towards the catalytic oxidation of carbon by Pt in air.

In order to compare the difference in the Pt-dispersion stability of the CNT/PyPBI/Pt50 and CNT/PyPBI/Pt2.5 composites, snapshots of the two composites were extracted from videos in the pressure ranges of 4 × 10−5 ~ 9 × 10−4 Pa (Fig. 5a) and 1 × 10−3 ~ 5 × 10−3 Pa (Fig. 5b). Notably, the Pt in the CNT/PyPBI/Pt50 started to move from around 6 × 10−4 Pa (Fig. 5a), whereas the Pt in the CNT/PyPBI/Pt5 was stable even below 3 × 10−3 Pa. Furthermore, a number of small Pt nanoparticles still remained in the CNT/PyPBI/Pt5, probably due to the lower density of the Pt nanoparticles on the CNTs (Fig. 5b), which lowered the chance of aggregation. This finding has provided a practical and useful strategy to suppress Pt aggregation caused by migration during fuel cell operations.

Figure 5
figure 5

In situ stability test of Pt dispersion under the electron microscope measurements.

(a), (b), Snapshots extracted from the videos (Supplementary, Movies S1 and 2) of the CNT/PyPBI/Pt50 (upper a and b) and CNT/PyPBI/Pt5 (lower a and b) in the pressure ranges of (a), from 4.28 × 10−5 to 8.79 × 10−4 Pa and (b), from 1.52 × 10−3 to 5.04 × 10−3 Pa. Scale bars, 50 nm.

A series of membrane-electrode assemblies (MEAs) using CNT/PyPBI/Pt50, CNT/PyPBI/Pt10 or CNT/PyPBI/Pt2.5 for both the anode and cathode (denoted MEA50, MEA10 and MEA2.5, respectively) were fabricated in order to investigate the fuel cell performance carried out under dry conditions at 120°C. The same amounts of the composites were loaded on the electrocatalyst layers to prepare the electrode catalyst with similar thickness (~10 μm), which means that different amounts of Pt corresponding to 0.50, 0.09 and 0.04 mg/cm2 of Pt were loaded for MEA50, MEA10 and MEA2.5, respectively. The small Pt loading values of MEA10 and MEA2.5 (0.09 and 0.04 mg/cm2) were lower than those of current PEFCs, in which the Pt amounts are 0.4–1.0 mg/cm237. The I-V curves together with their power density curves for the MEA50 (gray), MEA10 (blue) and MEA2.5 (green) are shown in Fig. 6a. The I-V curves were re-plotted based on the total amounts of Pt in the anode and cathode to compare the Pt mass activity of the MEAs (Fig. 6b). Importantly, the mass activity of MEA2.5, was around 8 times greater than that of MEA50. Fine control of the Pt size strongly indicates a correlation of Pt nanoparticle size with the feed of Pt ions and further improvements in the mass activity of the MEA is expected38,39; however, several reports have pointed out that smaller-sized Pt particles (<2 nm) show a lower ORR activity and these are not suitable for the fuel catalysts1,40,41,42.

Figure 6
figure 6

MEA single cell measurements using CNT/PyPBI/Pt as the electrocatalyst.

(a), I-V curves and power density curves for MEA50 (gray), MEA10 (blue) and MEA2.5 (green) plotted as a function of the current density. (b), I-V curves for MEA50 (gray), MEA10 (blue) and MEA2.5 (green) plotted as a function of Pt mass activity.

In order to shed some light on the stability of the electrocatalysts, the durability of the MEA10 was compared to that of the MEA50. Figs. 7a and 7b show the potential curves of the MEA50 and MEA10 as a function of cycling, respectively. In order to avoid a complex graph, only representative curves of the cycling were selected for display. The durability test was terminated after 60,000 potential cyclings. The potential at 200 mA mgPt−1 for both the MEA50 and MEA10 as a function of cycling is displayed in Fig. 7c. As observed, the degradation rates of both MEAs were almost the same, indicating a good stability of the low Pt loading composites. A more detailed study using half-cell measurements is now undergoing.

Figure 7
figure 7

Durability test of MEA.

(a). (b), Polarization curves of a durability test using MEA50 (a) and MEA10 (b). The polarization curves were measured after every 1,000 cycles. For convenience, the data were plotted every 10,000 cycles. (c). Plots of the cell voltage at 200 mA mgPt−1 for the MEA50 and MEA10.

In conclusion, this is the first example of fine control of Pt nanoparticle diameter on polymer-wrapped carbon supporting materials. The results showed that the ECSA values were dramatically increased and the utilization efficiency of the Pt was also increased. To compare the mass activity of Pt, we fabricated different MEAs using the prepared electrocatalysts with different Pt diameters. The I-V measurements at 120°C under the non-humidified condition revealed that the MEA using the smallest Pt (diameter = 2.3 nm) showed a mass activity about 8 times higher than that of the MEA using the largest Pt particles (diameter = 3.7 nm) due to the significant increase in the surface area of Pt as well as their utilization efficiency. Such a fine control of the ‘Pt size/size reduction and density’ in a simple approach for pristine CNTs is highly useful for the development of a next generation fuel cell electrocatalyst, in which in which the reduction of the Pt amount is highly demanded.



N,N-Dimethylacetamide (DMAc), ethylene glycol (EG) and hydrogen hexachloroplatinate hexahydrate (H2PtCl6·6H2O) were purchased from Wako Pure Chemical, Ltd. and used as received. Multi-walled carbon nanotubes (ca. 20 nm diameter) were kindly donated by the Nikkiso Co. Poly[2,2′-(2,6-pyridine)-5,5′-bibenzimidazole] (PyPBI) and poly[2,2′-(2,6-phenyl)-5,5′-bibenzimidazole] (PBI) were synthesized according to the literature43. CNTs wrapped with PyPBI (denoted CNT/PyPBI) were prepared according to the method described in our previous report25.


The size and distribution of the Pt nanoparticles were determined using TEM images (JEOL JEM-2010, acceleration voltage = 120 kV). A copper grid with a carbon support (Okenshoji) was used for the TEM observations. Samples were dispersed in 2-propanol in an ultrasonic bath for 10 min and a drop of the suspension was placed on a copper grid coated with a carbon film. The copper grid was then dried in a vacuum overnight. The TGA diagrams were obtained using a TGA-50 (Shimadzu) instrument operated in conditioned air (20 mL min−1) at a heating rate of 10°C min−1. The XRD and XPS spectra were measured using a Smart-Lab (Rigaku) and AXIS-ULTRADLD (SIMADZU), respectively. The binding energy in the XPS spectra was calibrated using the C 1s peak that appeared at 284.5 eV. Simultaneous SEM and STEM (SEM/STEM dual monitoring system operated at an acceleration voltage of 300 kV) observations were carried out using an HF-3300 (Hitachi High Technologies) microscope equipped with a gas injection system. In these experiments, the composites were heated up 200°C under vacuum at 3.3 × 10−5 Pa, then atmospheric air (RH = 42%, temp; 24°C) was introduced until the pressure reached 1.2 × 10−2 Pa (for pressure data, Supplementary, Fig. S3)

Synthesis of CNT/PyPBI/Pt electrocatalysts

Pt was deposited on the CNT/PyPBI using EG as a reducing agent44. Typically, CNT/PyPBI (10 mg) in a 60 vol% aqueous EG solution (60 mL) was dispersed by sonication to which H2PtCl6·6H2O in a 60 vol% aqueous EG solution (1.15 mM; 40 mL) was added. The resultant mixture was stirred for 4 h at room temperature, then refluxed at 140°C for 8 h. The obtained material was filtered using a 0.2 μm Millipore membrane, washed several times with water and finally dried under vacuum. The amount (volume) of the aqueous H2PtCl6·6H2O solution added to the solution was changed in order to optimize the amounts of the Pt loaded on the CNT/PyPBI.

Preparation of phosphoric acid doped-PBI membrane

LiBr (100 mg) was dissolved in DMAc (10 mL) in a 50 mL glass bottle to which PBI (200 mg) was added. The bottle was sealed under N2 and magnetically stirred at 50°C for 1 day, then the solvent was evaporated to obtain a solution (PBI content = ~4 wt%). The resultant PBI polymer solution was carefully cast on a glass plate using a film applicator (Elcometer 3600, 50 μm strip width), then gradually heated to 120°C to evaporate DMAc and held at 120°C for another 5 h to completely remove the solvent. The obtained PBI film was peeled off and immersed in hot distilled water for 30 min several times to remove the LiBr. Finally, the PBI membrane was doped in a 85-wt% phosphoric acid solution for 5 days to provide a membrane film with a 30-μm thickness. The amount of H3PO4 loaded on the membrane after doping was calculated by the weight change from the original membrane.

Electrochemical measurements

Electrochemical measurements were performed using a rotating ring disk electrode (RRDE-3, Bioanalytical Systems, Inc.) with a conventional three-electrode configuration at room temperature. A glassy carbon electrode (GC) with a geometric surface area of 0.282 cm2 was used as the working electrode. A Pt wire and Ag/AgCl were used as the counter and reference electrodes, respectively. The Ag/AgCl reference electrode was calibrated against the reference hydrogen electrode (RHE) potential in 0.1 M HClO4. All work potentials are indicated against the RHE. The potential of the sample electrode was controlled by a potentiostat (Model DY2323, ALS). The catalyst suspension was prepared as follows. CNT/PyPBI/Pt (1.0 mg) was ultrasonically dispersed in an aqueous EG solution (2.0 mL) to provide a homogeneous suspension. A portion of the composite suspension was then cast onto the GC electrode to load 14 μg cm−2 of Pt for CV measurements and 40 μL of the suspension was cast on a GC electrode for the ORR measurements. Finally, the casted GC electrodes were air dried. The CVs of the electrocatalysts were carried out at a scan rate of 50 mV/sec in N2-saturated 0.1 M HClO4 solutions to determine the ECSA values.

Fuel cell testing

The electrocatalyst composite was ultrasonically dispersed in an aqueous EG solution, then deposited on a gas diffusion layer (GDL) (SIGRACET gas diffusion media GDL 25 BC, SGL Carbon Group) by vacuum filtration. The GDL was used as a filter to produce a gas diffusion electrode (GDE). The H3PO4 doped-PBI membrane was then sandwiched between two GDEs to fabricate a membrane-electrode-assembly (MEA). The obtained MEA was hot-pressed at 120°C and 5 MPa for 30 sec. The performance of the MEA was measured at 120°C using a computer-controlled fuel cell test system (890e, Scribner Associate, Inc.). Polarization curves were recorded at atmospheric pressure under a dry hydrogen flow (100 mL/min) and air flow (200 mL/min) as the anode and cathode gases, respectively.

Durability testing

Selected MEAs were subject to an accelerated durability test by the protocol provided by the Fuel Cell Commercialization of Japan (FCCJ)45. Typically, the potential sweeps were cycled between 1.0 and 1.5 V at 120°C under non-humidified conditions in order to accelerate the oxidation process. The scanning rate was 0.5 V/s. The H2 and N2 gases were fed to the anode and the cathode, respectively. The I-V curves were recorded every 1,000 cycles after switching the cathode gas from N2 to air. The potential was plotted as a function of cycling at open circuit potential (OCV). The durability test for the examined MEAs was terminated when the potential became less than 50% of the initial OCV.