Highly dispersed palladium nanoparticles on poly(N₁,N₃-dimethylbenzimidazolium)iodide-functionalized multiwalled carbon nanotubes for ethanol oxidation in alkaline solution

Abstract

Multi-walled carbon nanotubes (MWCNTs) have been considered as good catalyst supporting materials, and their dispersion and functionalization are important, challenging problems for high-performance composite catalysts. Here, poly(N₁,N₃-dimethylbenzimidazolium)iodide (P(DMBI)-I⁻) was successfully synthesized by methylation of polybenzimidazole (PBI) for the dispersion and functionalization of MWCNTs. The results demonstrate that the novel P(DMBI)-I⁻ exhibits a higher dispersion effect for MWCNTs than the original PBI in some typical organic solvents. MWCNTs were wrapped with a thin layer of P(DMBI)-I⁻ and achieved functionalization with π–π conjugation and cation–π interaction. Benefiting from the positive charges and imidazole rings of P(DMBI)-I⁻, the palladium nanoparticles/P(DMBI)-I⁻-functionalized MWCNTs hybrid (Pd/P(DMBI)-I⁻-f-MWCNTs) catalyst loaded with highly dispersed palladium nanoparticles was fabricated by in situ reduction. TEM and SEM images demonstrated that when the feed weight ratio of Na₂PdCl₄ and P(DMBI)-I⁻-f-MWCNTs was 6 : 1, the Pd NPs of Pd/P(DMBI)-I⁻-f-MWCNTs with particle size of ∼2.9 nm were well distributed on P(DMBI)-I⁻-f-MWCNTs in good quantity. The amount of Pd loading on the catalyst of Pd/P(DMBI)-I⁻-f-MWCNTs was about 56.43 wt%. With respect to ethanol oxidation in alkaline solution, the Pd/P(DMBI)-I⁻-f-MWCNTs exhibited higher electrochemical performance and tolerance stability, compared to commercial Pd/C and Pd/PBI-f-MWCNTs.

---

1. Introduction

As an ideal candidate for energy conversion, the direct ethanol fuel cell (DEFC) has always received considerable attention, due to its high energy conversion efficiency, low emissions and low operating temperatures.^1,2 As is well-known, the active anode catalyst is one of the key materials for the ethanol oxidation reaction in DEFC.^3,4 Although traditional Pt-based electrocatalysts have been widely studied for ethanol electro-oxidation,^5,6 the high cost and limited resources hinder its further commercial applications,⁷ and Pt-based electrocatalysts undergo deactivation/poisoning by carbon monoxide-like intermediates.^8,9 Recently, Pd-based catalysts have become a hot topic of interest because of their lower cost and greater resistance to CO in alkaline media, compared to Pt catalysts.^10–12 Nevertheless, high electrocatalytic activity is strongly dependent on the shape, size and distribution of electrocatalysts.¹³ In this case, supporting materials with high surface area are essential to disperse catalyst particles for reducing the catalyst loading and thereby improving the fuel cell performance.^14–16

Multi-walled carbon nanotubes (MWCNT) have been considered as fine heterogeneous catalyst supports, due to their unique electrical and structural properties.^17,18 Pd/MWCNTs hybrid materials have been reported as highly effective catalysts for improving DEFC performance.^18,19 However, pristine MWCNTs are chemically inert and need to be functionalized to ensure the uniform distribution and deposition of Pd nanoparticles (NPs) on their surfaces. To realize the effective combination of Pd NPs and MWCNTs, lots of research works have focused on the functionalization of MWCNTs.^20,21 The traditional functionalization method involves the chemical treatment of pristine MWCNTs with HNO₃ or H₂SO₄/HNO₃ acid solutions. For acid-treated MWCNTs, the generated functional oxygen-containing groups and defects of the MWCNTs can be used as catalyst supports for the deposition of catalyst nanoparticles. Nevertheless, the acid treatment inevitably introduces surface defects and causes structural damage to the MWCNTs. As a consequence, non-covalent functionalization methods exhibit potential advantages, in comparison to acid-treatment.²² Through π–π stacking, ion–π interactions, electrostatic and hydrogen-bond interactions, etc., the MWCNTs can be functionalized by surfactants, polymers and other capping agents.^23,24

Poly(benzimidazole) (PBI), a high-performance benzo-heterocycle polymer, has been widely used in lots of applications, including electronic and automotive components, structural resins, and DEFCs, due to its excellent mechanical properties and high thermal and chemical stability.^25–28 With respect to DEFCs, PBI can act as a proton-conducting material and be used above 100 °C under dry conditions. It is supposed to be an appropriate substitution for Nafion®, used in low-temperature DEFC systems.^29,30 It is noteworthy that on the strength of the aromatic structure, PBI was used for the exfoliation and dispersion of carbon nanotubes with π–π interactions.^31–34 Okamoto et al. reported that the hybrids of MWCNT/PBI/Pt for use in fuel-cells exhibited good electro-catalytic activity.³⁵ Fujigaya et al. further used PBIs as effective binding sites for the formation of platinum (Pt) nanoparticles and fabricated a ternary composite (CNT/PBIs/Pt) for oxygen reduction catalysts.³⁶ It was found that PBI not only acts as the wrapping polymer for the exfoliation and dispersion of MWCNTs through π–π interactions, but also as a Pt or Pd ions absorber through the coordination with the aromatic nitrogen on the PBI. Although the imidazole groups of PBI can act as active adsorption sites for metal ions for growth and dispersion, the easy conglomeration of metal nanoparticles and relatively high content of the metal nanoparticles are still problems to be solved. The main reason for the problems is the poor dissolving properties derived from the rigid main chains and strong intermolecular hydrogen bonding of pure PBI.^37,38

Based on the above analysis, we synthesized poly(N₁,N₃-dimethylbenzimidazolium)iodide (P(DMBI)-I⁻) by the alkylation reaction of the imidazole ring in primary polybenzimidazole (PBI). Through non-covalent functionalization of P(DMBI)-I⁻, the dispersion of MWCNTs was significantly improved with π–π and cation–π interactions. With the assistance of positive charges and the imidazole ring of P(DMBI)-I⁻, the Pd/P(DMBI)-I⁻-f-MWCNTs with uniformly distributed small-size Pd NPs were prepared for the first time. The obtained Pd/P(DMBI)-I⁻-f-MWCNTs exhibited superior electrocatalytic performance in ethanol oxidation in alkaline solution.

2. Experimental section

2.1 Chemicals and reagents

MWCNTs (purity > 95%, length: 10–20 μm and OD: 30–50 nm) were purchased from Chengdu Organic Chemicals Co. Ltd., (Chengdu, China). 3,3′-Diaminobenzidine (DABz), isophthalic acid (IPA), lithium hydride (LiH) and sodium tetrachloropalladate(II) (Na₂PdCl₄) were obtained from J&K® company. Sodium borohydride (NaBH₄), N-methyl-2-pyrrolidone (NMP), N,N-dimethylacetamide (DMAc), N,N-dimethylformamide (DMF), dimethyl sulfoxide (DMSO), acetone, ethanol, methanol and tetrahydrofuran were obtained from Sinopharm Chemical Reagent Co. Ltd. (Shanghai, China). 5 wt% Nafion® solution was provided by DuPont Co. Ltd. (Wilmington, DE, USA).

2.2 Materials synthesis

2.2.1 Preparation of PBI. PBI was first synthesized from DABz (1.56 g) and IPA (2.01 g) by melt polycondensation in polyphosphoric acid (PPA) under nitrogen atmosphere, as reported in literature.^39,40 Typically, DABz (1.56 g) was mixed into the melted PPA at 200 °C in N₂ atmosphere, with refluxing for 4 h. Afterwards, IPA was added to the mixture solution for polymerization in N₂ atmosphere at 200 °C for 36 h, and then at 210 °C for another 24 h. After the reaction, the product mixture was poured into distilled water to precipitate the PBI product. The precipitated PBI was milled into a powder and then washed with 50 mL of NaOH aqueous solution (1 mol L⁻¹) and 50 mL of deionized water, five times. After removing the residual PPA and NaOH, the PBI was finally dried under vacuum at 100 °C for 24 h.

2.2.2 Synthesis of P(DMBI)-I⁻. P(DMBI)-I⁻ was synthesized by a method previously reported by Pearce et al.⁴¹ PBI powder (6.72 g) was first dissolved in 100 mL of NMP with magnetic stirring under nitrogen atmosphere at 100 °C. After cooling to room temperature, LiH (0.06 g) was then put into the above PBI/NMP solution and heating was continued up to 100 °C. After 15 h of reaction, the obtained mixture solution was again cooled to room temperature. Methyl iodide (10.0 g) was added to the mixture solution with 6 h of reaction at 60 °C. Afterward, the raw brown-yellow products were collected by vacuum filtration and dried. To further enhance the degree of methylation, the raw products were dissolved in 100 mL of DMSO, followed by the addition of methyl iodide (12.0 g). The mixture solution was reacted at 90 °C for 15 h with magnetic stirring under nitrogen atmosphere. The reacted mixture solution was poured into 1 L of deionized water and the brown product of P(DMBI)-I⁻ was precipitated and collected by vacuum filtration. Finally, the P(DMBI)-I⁻ was purified by Soxhlet extraction in acetone and vacuum dried at 90 °C for 24 h.

2.2.3 P(DMBI)-I⁻-functionalized MWCNTs (P(DMBI)-I⁻-f-MWCNTs). 0.01 g of P(DMBI)-I⁻ was first dissolved in 10 mL of DMAc and formed a homogeneous solution. Next, 0.01 g of MWCNTs was added into the above solution and sonicated for 6 h in an ice water bath (100 W, 40 kHz). The redundant MWCNTs were then removed by centrifugal separation at 3000 rpm for 10 min. The P(DMBI)-I⁻-f-MWCNTs were sequentially collected from the supernatant dispersion by repeated centrifugal separation (15 000 rpm, 60 min) and washing with DMAc at least 5 times. Finally, the P(DMBI)-I⁻-f-MWCNTs product was vacuum dried at 80 °C for 12 hours. For comparison, PBI-f-MWCNTs were obtained by the same process.

2.2.4 Preparation of Pd/P(DMBI)-I⁻-f-MWCNTs composites. 10 mg of P(DMBI)-I⁻/MWCNTs were dispersed in 10 mL of DMAc with sonication for 30 minutes. After that, Na₂PdCl₄/DMAc solution (1 mol L⁻¹) was added into above dispersion for 12 h of reaction at room temperature. Next, 10 mL of NaBH₄/DMAc (5 mg mL⁻¹) were slowly added into the dispersion by a peristaltic pump, then, the mixture was further stirred for 30 minutes at room temperature. The sample was collected by centrifugation and washed with deionized water and ethanol. After vacuum drying at 80 °C for 12 h, Pd/P(DMBI)-I⁻-f-MWCNTs were finally obtained. For comparison, Pd/MWCNTs and Pd/PBI-f-MWCNTs were prepared by the same process.

2.2.5 Instruments and characterization. Transmission electron microscopy (TEM) images were obtained by a JEOL JEM-2100 instrument. The Raman spectra of the samples were measured using inVia Reflex laser confocal Raman microspectroscopy at an excitation wavelength of 633 nm. Scanning electron microscopy (SEM) was conducted using Hitachi S4800 cold field emission scanning electron microscopes. X-ray diffraction (XRD) characterization was carried out by a Rigaku D/max-2550. The X-ray photoelectron spectroscopy (XPS) measurements were performed by Thermo Scientific ESCALAB 250Xi using Al Kα (1486.6 eV) excitation. The ¹H NMR spectra were recorded at room temperature on the Varian 400 MHz INOVA instrument. Fourier transform infrared (FTIR) spectra of samples were obtained using a Nicolet 8700 FTIR spectrometer. The TGA analyses were carried out on a Perkin-Elmer TGA2050 apparatus. The TGA curves of MWCNTs and functionalized MWCNTs were obtained at a heating rate of 20 °C min⁻¹ from 30 to 700 °C, under nitrogen atmosphere. UV-vis absorption spectra were obtained using a Perkin-Elmer Lambda35 UV/vis spectrometer.

The electrochemical experiments were carried out using a conventional three-electrode cell using a CHI 660E electrochemistry workstation at room temperature. Pt wire and a saturated calomel electrode (SCE) were used as the counter electrode and reference electrode, respectively. To prepare the catalyst ink, the prepared catalyst was suspended by sonication in an isopropanol/DI water mixture (1 : 4 v/v) with a 20 wt% Nafion solution (5 wt%, IonPower, Inc.). The catalysts with a Pd loading of 28 μg cm⁻² were then dropped onto the polished and cleaned glassy carbon electrode (GCE, 3 mm in diameter) surface and dried. The electrolyte solutions were purged with N₂ gas before use for about 1 h. Cyclic voltammetry was performed in 1 M NaOH, and a solution of 1 M NaOH with 1 M C₂H₅OH at a scan rate of 50 mV s⁻¹. Chronoamperometry was carried out at −0.2 V for 3600 s.

3. Results and discussion

3.1 Preparation of P(DMBI)-I⁻

According to the method reported by Hu et al., P(DMBI)-I⁻ was synthesized by the alkylation reaction of the imidazole ring in the polymer chains of polybenzimidazole (PBI).⁴¹ As shown in Fig. 1a, the entire reaction was divided into two steps. Firstly, the PBI lithium salt (PBI–Li) was prepared by breaking the hydrogen bonds of N–H groups in the structure of PBI chains through the reaction between LiH and PBI in NMP solvent under the protection of nitrogen. The P(DMBI)-I⁻ was finally obtained by electrophilic reaction of PBI–Li with excess CH₃I in dimethylsulfoxide (DMSO).

Fig. 1 (a) Schematic diagram for the preparation of P(DMBI)-I⁻; (b) ¹H NMR spectroscopy of as-prepared P(DMBI)-I⁻ and (c) FT-IR spectra of PBI and P(DMBI)-I⁻.

¹H NMR was first used to characterize the P(DMBI)-I⁻. From Fig. 1b, compared to pure PBI, there are new chemical shifts of δ 4.2 and δ 4.13 ppm in ¹H NMR spectroscopy of P(DMBI)-I⁻, attributed to the methyl groups bonded to the nitrogen atoms of the imidazole ring. Additionally, the peak for the N–H bond of PBI located at δ 13.3 ppm disappeared, also demonstrating the breaking of the hydrogen bonds of the N–H groups of PBI with the methylation reaction. Calculated from the integral area, the degree of methylation of the as-prepared P(DMBI)-I⁻ was ∼95.3%. FT-IR spectra were also used for characterizing the structural change of the PBI chains. From Fig. 1c, the peak at ∼1630 cm⁻¹ for PBI was attributed to the characteristic adsorption peak of C=N/C=C groups. The stretching vibration band for the C=C of the benzene ring was exhibited at ∼1450 cm⁻¹. Compared to PBI, the stretching vibration peak for N–H groups at around ∼3200 cm⁻¹ disappeared in P(DMBI)-I⁻, which is ascribed to the replacement of the protons of N–H groups by methyl groups during the methylation reaction. The new peak for the C–N groups in the structure of P(DMBI)-I⁻ appeared at ∼1216 cm⁻¹. These results demonstrated the successful methylation of PBI, which is consistent with ¹H NMR spectroscopy.⁴²

3.2 P(DMBI)-I⁻-assisted dispersion and functionalization of MWCNTs

As shown in Table 1, the P(DMBI)-I⁻ exhibited good solubility after methylation and iodine ionization. For low or high boiling-point polar solvents, the solubility of P(DMBI)-I⁻ was obviously improved, compared to pure PBI; this is mainly attributed to the breaking of hydrogen bonds and the positive charge of the molecular chains of P(DMBI)-I⁻.

Table 1 Comparison of the solubilities of PBI and P(DMBI)-I⁻ in typical organic solvents^a

Product	Ethanol	Acetone	THF	DMF	DMAc	DMSO	NMP
a +++: totally dissolved at R.T., ++: totally dissolved at 80 °C, +: partial dissolution at 80 °C, −: insoluble (initial concentration: 1 mg mL^−1).
PBI	−	−	−	++	++	++	++
P(DMBI)-I^−	+	+	+	+++	+++	+++	+++

---

The stabilization and exfoliation of MWCNTs were carried out by bath sonication of the dispersion mixture of MWCNTs in typical polarity organic solvents, e.g., DMF, DMAc, NMP and DMSO. From Fig. 2a, the MWCNTs were not well dispersed in the pure solvents and almost formed granular dispersions. After the addition of PBI and P(DMBI)-I⁻, the colors of both MWCNTs-based dispersions changed from clear and transparent to dark and black; this proved that the PBI and P(DMBI)-I⁻ both exhibited good dispersibility of MWCNTs in organic solvents (Fig. 2b and c). For further studying the differences in dispersing properties between PBI and P(DMBI)-I⁻, the concentrations of MWCNTs for MWCNTs/PBI and MWCNTs/P(DMBI)-I⁻ dispersions were calculated from the absorbance at 500 nm with the Lambert–Beer law by UV-vis analysis.^43,44 The concentration of MWCNTs (C_MWCNTs, mg mL⁻¹) was calculated using A = εCl, where A is the absorbance of PBI- and P(DMBI)-I⁻ aided, stabilized MWCNTs dispersion at the wavelength of 500 nm; ε is the extinction coefficient, which is related to the absorbance (value: 47.04 mL mg⁻¹ cm⁻¹); l (value: 10⁻² m) is the path length.⁴⁵ As shown in Fig. 2d, the C_MWCNTs for MWCNTs/PBI and MWCNTs/P(DMBI)-I⁻ dispersions significantly improved with the existence of PBI and P(DMBI)-I⁻. It is worth noting that the C_MWCNTs of MWCNTs/P(DMBI)-I⁻ is higher than the C_MWCNTs of MWCNTs/PBI in all organic solvents. The C_MWCNTs of MWCNTs/P(DMBI)-I⁻ in NMP and DMF reached ∼0.45 and ∼0.365 mg mL⁻¹, nearly 5 and 4 times higher than the C_MWCNTs for MWCNTs/PBI in NMP and DMF, respectively. As aromatic polymers, there are strong interactions between polymer chains and MWCNTs. The PBI and P(DMBI)-I⁻ can be absorbed on the MWCNTs through π–π conjugation with MWCNTs, and then can achieve good dispersion of MWCNTs. However, with respect to the PBI-assisted dispersion system, the strong intermolecular hydrogen-bond interaction in PBI chains resulted in the inhomogeneous adsorption of PBI on the tube surfaces of MWCNTs, and the dispersed MWCNTs easily formed agglomerations. After methylation and iodine ionization, the P(DMBI)-I⁻ was also absorbed on the MWCNTs by cation–π interactions, due to the positive charge of the molecular chains, while the reduced rigidity and enhanced electrostatic repulsion of the molecular chain significantly improved the stability of the MWCNTs/P(DMBI)-I⁻ dispersion.⁴⁶

Fig. 2 (a) The obtained MWCNTs dispersion after centrifugal separation (initial concentration of MWCNTs was 1 mg mL⁻¹). (b) The obtained MWCNTs/PBI dispersion with PBI-assisted exfoliation and dispersion after centrifugal separation (initial concentrations of MWCNTs and PBI were all 1 mg mL⁻¹). (c) The obtained MWCNTs/P(DMBI)-I⁻ dispersion with PBI-assisted exfoliation and dispersion after centrifugal separation (initial concentrations of MWCNTs and P(DMBI)-I⁻ were all 1 mg mL⁻¹). (d) The columnar analysis diagram of C_MWCNTs for pure solvents, PBI and P(DMBI)-I⁻ dispersion liquid mediums.

In fact, during the process of P(DMBI)-I⁻-assisted dispersion and exfoliation of MWCNTs, the tube surfaces of MWCNTs were simultaneously non-covalently functionalized by π–π conjugation and cation–π interaction with P(DMBI)-I⁻. TEM images were used to characterize the P(DMBI)-I⁻-functionalized MWCNTs. From Fig. 3a, there are abundant windings and aggregations in the original MWCNTs. After exfoliation and dispersion by PBI and P(DMBI)-I⁻, the phenomena of windings and aggregations were significantly alleviated, as shown in Fig. 3b and c. From the high-magnification TEM images inset, the tube structures of MWCNTs can be clearly identified, with inner and outer diameters along the length. Nevertheless, compared to the smooth and clean surface of MWCNTs, the surfaces became rough and coated by thin polymer layers after the functionalization with PBI and P(DMBI)-I⁻. It is worth noting that the P(DMBI)-I⁻-dispersed MWCNTs exhibited a nearly monodispersed state and exhibited better dispersed state, compared to that of PBI-dispersed MWCNTs.

Fig. 3 TEM images of (a) MWCNTs, (b) PBI- and (c) P(DMBI)-I⁻-dispersed MWCNTs (inset of (a)–(c): high-magnification TEM images).

For further characterizing the P(DMBI)-I⁻-dispersed MWCNTs, Raman spectroscopy was also applied. As shown in Fig. 4, the peaks of 1314 and 1570 cm⁻¹ were ascribed to the D and G bands of MWCNTs, which originated from the defects and disorder induced modes and the in-plane E_2g zone-center mode. The second prominent peak, the 2D band of MWCNTs from a double-resonance process, is located at 2612 cm⁻¹. After non-covalent functionalization, there was an evident blue shift of the 2D band from 2612 cm⁻¹ for MWCNTs to 2630 and 2648 cm⁻¹ for PBI- and P(DMBI)-I⁻-functionalized MWCNTs, respectively, which may be attributed to axial deformation with the squeezing action of PBI- and P(DMBI)-I⁻ on the surfaces of the MWCNTs. Notably, the blue shift of the 2D band for the P(DMBI)-I⁻-functionalized MWCNTs was larger than the blue shift for the PBI-functionalized MWCNTs. It demonstrated that the adhesion strength of P(DMBI)-I⁻-f-MWCNTs was higher than PBI-f-MWCNTs, due to π–π conjugation and cation–π interactions between MWCNTs and P(DMBI)-I⁻.⁴⁷

Fig. 4 Raman spectra for MWCNTs, PBI-functionalized MWCNTs and P(DMBI)-I⁻-functionalized MWCNTs.

After functionalization, P(DMBI)-I⁻-functionalized MWCNTs exhibited favorable dispersibility in some organic solvents. Here, DMAc was chosen for further growth and coating of Pd NPs on the surfaces of MWCNTs, based on the low toxicity and reducing capacity of DMAc. Firstly, the sedimentation experiments of PBI/MWCNTs and P(DMBI)-I⁻/MWCNTs were performed in DMAc. From the inset photograph of Fig. 5a, the MWCNTs can be well dispersed in P(DMBI)-I⁻/DMAc without any precipitation after standing for one month. However, there are obvious sediments in the PBI/MWCNTs/DMAc dispersion. Through high-speed centrifugal separation (20 000 rpm, 30 min), the P(DMBI)-I⁻-f-MWCNTs were obtained and quantitatively analyzed. The P(DMBI)-I⁻ decomposed primarily by the step at around ∼300 °C, corresponding to the decomposition of the side methyl groups. From Fig. 5b, calculating from the weight loss at around 700 °C in nitrogen, the adsorbed P(DMBI)-I⁻ and PBI of P(DMBI)-I⁻-f-MWCNTs and PBI-f-MWCNTs accounts for ∼11.29 and ∼4.97 wt%, respectively. It is clear that the amount of P(DMBI)-I⁻ from P(DMBI)-I⁻-f-MWCNTs is significantly higher than the PBI from PBI-f-MWCNTs, due to the dual action of π–π conjugation and cation–π interaction between P(DMBI)-I⁻ and MWCNTs.

Fig. 5 (a) TGA curves for MWCNTs, PBI-f-MWCNTs, and P(DMBI)-I⁻-f-MWCNTs (inset: the photograph of the MWCNTs dispersion in DMAc of PBI and P(DMBI)-I⁻ after standing for one month). (b) TGA curves for MWCNTs, P(DMBI)-I⁻ and P(DMBI)-I⁻-f-MWCNTs.

3.3 Preparation of Pd/P(DMBI)-I⁻-f-MWCNTs hybrids

For P(DMBI)-I⁻, the original coordinating atoms and groups of the imidazole ring from PBI were displaced by methyl groups. The polymer chains for P(DMBI)-I⁻ cannot form coordination compounds by making full use of the coordination effect between the imidazole ring and metal ions, but there is a very large number of positive charges in the polymer skeleton of P(DMBI)-I⁻ after methylation. The metal ions can be attracted through electrostatic interaction and reduction, for the fabrication of metal nanoparticles/MWCNTs hybrids. As shown in Fig. 6, using Na₂PdCl₄ and NaBH₄ as precursor and reductant, respectively, Pd/P(DMBI)-I⁻-f-MWCNTs were prepared by electrostatic adsorption and chemical reduction.

Fig. 6 Schematic diagram for the preparation of Pd/P(DMBI)-I⁻-f-MWCNTs.

TEM images were first used to characterize the Pd/P(DMBI)-I⁻-f-MWCNTs hybrids with different feed weight ratios of Na₂PdCl₄ and P(DMBI)-I⁻-f-MWCNTs. As shown in Fig. 7a, e and i, the Pd NPs with average particle sizes of ∼2.8 nm were successfully decorated and sparsely distributed on the surfaces of MWNTs when the feed weight ratio of Na₂PdCl₄ and P(DMBI)-I⁻-f-MWCNTs was 2 : 1. From Fig. 7b, f and j, when the ratio of Na₂PdCl₄ and P(DMBI)-I⁻-f-MWCNTs was 6 : 1, the Pd NPs were well distributed on the surfaces of MWCNTs in good quantity. The particle sizes were concentrated mainly at ∼2.9 nm. As the ratio continued to increase, agglomeration began to occur. As shown in Fig. 7c, g, k, d, h and i, there are some obvious aggregations of Pd NPs on the surfaces of MWCNTs when the feed weight ratios of Na₂PdCl₄ and MWNT/P(DMBI)-I⁻ are 10 : 1 and 16 : 1. It is worth noting that the particle sizes of Pd NPs are all relatively small, mainly concentrated in the range from 2 to 4.5 nm. Actually, the imidazole groups of P(DMBI)-I⁻ on the surfaces of MWCNTs provide lots of active sites for nucleation and crystal growth of Pd NPs. With the electrostatic interaction between PdCl₄²⁻ and the electropositive polymer skeleton of P(DMBI)-I⁻, the precursor of Pd NPs can be uniformly adsorbed onto the surfaces of MWCNTs. By then, the Pd/P(DMBI)-I⁻-f-MWCNTs hybrids were successfully prepared and well distributed on the surfaces of MWCNTs after reduction.

Fig. 7 TEM images and histograms of Pd/P(DMBI)-I⁻-f-MWCNTs obtained from different feed weight ratios of Na₂PdCl₄ and P(DMBI)-I⁻-f-MWCNTs: (a), (e) and (i) for 2 : 1, (b), (f) and (j) for 6 : 1, (c), (g) and (k) for 10 : 1 and (d), (h) and (l) for 16 : 1.

The Pd/P(DMBI)-I⁻-f-MWCNTs were further characterized by HRTEM and XRD. As shown in Fig. 8a, the crystal lattice of Pd is continuous, in the same direction and 3–4 nm in grain size. Meanwhile, the HRTEM image clearly shows close contact between the Pd nanoparticles and the MWCNTs surface. It demonstrated that the precursor had already grown into Pd NPs with good crystal structure on the P(DMBI)-I⁻ layer, followed by penetration into the P(DMBI)-I⁻ layer to form a close contact structure with the MWCNT surfaces. From Fig. 8b, the diffraction peak of the XRD pattern for Pd/P(DMBI)-I⁻-f-MWCNTs at 25.9° is attributed to the C (002) reflection of MWCNTs.⁴⁸ The XRD pattern also shows the characteristic peaks of the face-centered cubic Pd lattice at 2θ = 40.1°, 46.7°, 68.1° and 81.1°, corresponding to Pd (111), Pd (200), Pd (220) and Pd (311), respectively.⁴⁹ The grain sizes were calculated by the Debye–Scherrer formula, according to the XRD spectra, to be about ∼4.2 nm for Pd NPs, which is consistent with the results of the TEM images.

Fig. 8 (a) HRTEM image and (b) XRD pattern of Pd/P(DMBI)-I⁻-f-MWCNTs, with the feed weight ratio of Na₂PdCl₄ and P(DMBI)-I⁻-f-MWCNTs of 6 : 1.

Obviously, the P(DMBI)-I⁻ can well facilitate the formation and distribution of Pd NPs after methylation. For comparing and analyzing the influence of P(DMBI)-I⁻ and PBI on the formation and distribution of Pd NPs, the Pd/MWCNTs and Pd/PBI-f-MWCNTs were also prepared with the same feed weight ratio. As shown in Fig. 9, regarding Pd/MWCNTs, the Pd NPs were unequally distributed on the surfaces of MWCNTs. Due to a lack of functional groups, the Pd NPs barely grew and deposited defect sites of un-modified pure MWCNTs on the structure, consequently resulting in lots of agglomeration. After functionalization by PBI, the distribution and agglomeration of Pd NPs on the surfaces of MWCNTs were obviously improved. However, because of the poor solubility and strong hydrogen-bonding interactions between molecular chains of PBI, the Pd NPs were not well dispersed and grown on the surfaces of MWCNTs. The Pd NPs preferentially deposited on the sites of the imidazole rings of PBI adsorbed on the surfaces of MWCNTs through π–π stacking between PBI and MWCNTs. Once the precursor is in excess, the Pd NPs tend to aggregate, induced by high surface energy. Obviously, compared to the PBI, the P(DMBI)-I⁻ is more beneficial to the surface growth and distribution of Pd NPs. In addition, from the XRD pattern of Pd/MWNT, the intensities of diffraction peaks for Pd (220) and Pd (311) are very low, due to a lower quantity of Pd NPs.⁵⁰ SEM images were also used for comparing the dispersion and growth of Pd NPs in Pd/MWCNTs, Pd/PBI-f-MWCNTs and Pd/P(DMBI)-I⁻-f-MWCNTs. From Fig. 10, the Pd NPs were nearly dominated by agglomeration in Pd/MWCNTs and Pd/P(DMBI)-I⁻-f-MWCNTs. Regarding Pd/P(DMBI)-I⁻-f-MWCNTs, the Pd NPs were uniformly distributed on the surfaces of MWCNTs without aggregates, which is consistent with the TEM images.

Fig. 9 TEM images of (a and b) Pd/MWCNTs (6 : 1) and (d and e) Pd/PBI-f-MWCNTs-6 : 1 with different magnifications. XRD patterns of (c) Pd/MWCNTs-6 : 1 and (f) Pd/PBI-f-MWCNTs-6 : 1.

Fig. 10 FE-SEM images of (a) Pd/MWCNTs-6 : 1, (b) Pd/PBI-f-MWCNTs-6 : 1 and (c) Pd/P(DMBI)-I⁻-f-MWCNTs-6 : 1.

XPS was utilized to characterize the samples of Pd/P(DMBI)-I⁻-f-MWCNTs. As depicted in Fig. 11, the elements of C_1s, Pd_3d and N_1s exist in all samples of Pd/P(DMBI)-I⁻-f-MWCNTs, with different feed ratios. When the feed weight ratio of Na₂PdCl₄ and P(DMBI)-I⁻-f-MWCNTs was 2 : 1, the ratio of C_1s/Pd_3d for Pd/P(DMBI)-I⁻-f-MWCNTs was ∼21.03; the ratio of C_1s/Pd_3d increased with the increase in the feed amount of Na₂PdCl₄. Evidently, the ratio of C_1s/Pd_3d reached ∼11.37 as the feed weight ratio increased to 6 : 1. However, when the feed weight ratio of Na₂PdCl₄ and P(DMBI)-I⁻-f-MWCNTs increased to 10 : 1 and 16 : 1, the ratios of C_1s/Pd_3d were ∼4.58 and ∼4.61, respectively, which were nearly the same. This indicated that once the precursors of Pd NPs are in excess with respect to the surface active sites, the Pd NPs begin to stack and form aggregations with interactions of Pd NPs during the formation of Pd NPs.⁵¹ Once the amounts of Pd precursors are in excess, the loading amounts of Pd no longer increase. From the TGA curves (Fig. 12a) for P(DMBI)-I⁻-f-MWCNTs and Pd/P(DMBI)-I⁻-f-MWCNTs with different feed ratios, the residue of P(DMBI)-I⁻-f-MWCNTs was ∼1.72 wt% when it was heated to 850 °C in an oxygen atmosphere. The loading contents of Pd were calculated to be ∼20.48 wt%, ∼56.43 wt%, ∼64.38 wt% and ∼67.58 wt% in Pd/P(DMBI)-I⁻-f-MWCNTs with feed ratios of 2 : 1, 6 : 1, 10 : 1 and 16 : 1, respectively. It is obvious that with increasing amounts of Na₂PdCl₄, the loading amounts of Pd no longer increased after the feed ratio exceeded 10 : 1. This is in accordance with XPS analysis. Additionally, the loading amount of Pd NPs also accounted for ∼54.03 wt% in Pd/PBI-f-MWCNTs-6 : 1, determined from Fig. 12b. Although the loading amount of Pd NPs for Pd/PBI-f-MWCNTs-6 : 1 is close to that of Pd NPs for Pd/P(DMBI)-I⁻-f-MWCNTs, the Pd NPs for Pd/P(DMBI)-I⁻-f-MWCNTs exhibited better dispersion and decoration on the surfaces of MWCNTs. On the other hand, the P(DMBI)-I⁻ was beneficial to the growth and distribution of Pd NPs. In Fig. 11, the peaks of 340.96 and 335.7 eV were identified to be the Pd_3d3/2 and Pd_3d5/2 of elemental palladium. It demonstrated that the Pd²⁺ ions were successfully reduced to Pd⁰.⁵² Through analyzing the integral areas, the proportion of valence states were 40 and 60% for Pd_3d3/2 and Pd_3d5/2, respectively, consistent with the literature report.⁴⁸

Fig. 11 XPS spectroscopy of Pd/P(DMBI)-I⁻-f-MWCNTs with the feed weight ratios of Na₂PdCl₄ and P(DMBI)-I⁻-f-MWCNTs: (a and b) 2 : 1, (c and d) 6 : 1, (e and f) 10 : 1, (g and h) 16 : 1.

Fig. 12 (a) TGA curves for P(DMBI)-I⁻-f-MWCNTs and Pd/P(DMBI)-I⁻-f-MWCNTs with different feed ratios. (b) TGA curves for PBI-f-MWCNTs and Pd/PBI-f-MWCNTs-6 : 1.

3.4 Electrocatalytic performance of Pd/P(DMBI)-I⁻-f-MWCNTs

The P(DMBI)-I⁻ not only acts as the wrapping and functionalizing polymer for MWCNTs through π–π and cation–π interactions, but also as a PdCl₄²⁻ ion absorber, through coordination with the aromatic nitrogen and electrostatic absorption. To demonstrate the capability of Pd/P(DMBI)-I⁻-f-MWCNTs as an electrocatalyst for ethanol oxidation in alkaline solution, the electrochemical activities were examined. The electrochemical active surface area (EASA) is an important index to provide a wealth of information about the active sites in a catalyst, and to assess the conductive pathways available to transfer electrons from and to the catalyst surface.⁵³ Fig. 13a shows the CVs of the commercial Pd/C, Pd/PBI-f-MWCNTs and Pd/P(DMBI)-I⁻-f-MWCNTs in 1 M NaOH solution without ethanol at 50 mV s⁻¹. The hydrogen adsorption–desorption zone ranges from −1.0 V to −0.7 V. The reduction of palladium oxide was used to determine the EASA of the catalysts, which was calculated using the following equation:⁵⁴
where s is the proportionality constant used to relate charge to area, and l is the catalyst loading in g. A charge value of 4.05 (C m⁻²) is the charge required for the reduction of the PdO monolayer. The EASA of the Pd/PBI-f-MWCNTs and Pd/P(DMBI)-I⁻-f-MWCNTs were calculated to be 55.78 m² g⁻¹ and 77.45 m² g⁻¹, which are both much greater than that of the commercial Pd/C (47.24 m² g⁻¹).⁵⁵ This demonstrated that the PBI and P(DMBI)-I⁻ can both effectively disperse MWCNTs and result in the homogeneous loading of the Pd NPs on the MWCNTs. In addition, the loading amounts of Pd and EASA values of Pd/PBI-f-MWCNTs and Pd/P(DMBI)-I⁻-f-MWCNTs with different feed weight ratios are illustrated in Table 2 (Fig. S2 and S3†). Clearly, the loading amounts of Pd and EASA values increase with increasing feed weight ratios, and more remarkably, when the feed weight ratio exceeded 6 : 1, the EASA of Pd/P(DMBI)-I⁻-f-MWCNTs was obviously higher than the EASA of Pd/PBI-f-MWCNTs. On the other hand, the loading amounts of Pd for Pd/P(DMBI)-I⁻-f-MWCNTs were slightly higher than that of Pd/PBI-f-MWCNTs for the same feed weight ratios. Most importantly, the better distribution of Pd NPs of Pd/P(DMBI)-I⁻-f-MWCNTs delivered higher EASA, compared to the Pd/P(DMBI)-I⁻-f-MWCNTs. It may be attributed to the higher amounts of P(DMBI)-I⁻ of P(DMBI)-I⁻-f-MWCNTs providing more active sites for growth and distribution, compared to the PBI of PBI-f-MWCNTs.

Fig. 13 (a) CVs of commercial Pd/C, Pd/PBI-f-MWCNTs (6 : 1) and Pd/P(DMBI)-I⁻-f-MWCNTs (6 : 1) in 1 M NaOH solution at 50 mV s⁻¹. (b) CVs of commercial Pd/C, Pd/PBI-f-MWCNTs (6 : 1) and Pd/P(DMBI)-I⁻-f-MWCNTs (6 : 1). (c) Chronoamperograms of commercial Pd/C, Pd/PBI-f-MWCNTs (6 : 1) and Pd/P(DMBI)-I⁻-f-MWCNTs (6 : 1) in 1 M NaOH solution containing 1 M C₂H₅OH at 50 mV s⁻¹.

Table 2 The loading amounts of Pd and EASA of Pd/PBI-f-MWCNTs and Pd/P(DMBI)-I⁻-f-MWCNTs with different feed weight ratios

Sample	Pd loading amount (wt%)	EASA (m^2 g^−1)
Pd/PBI-f-MWCNTs-2 : 1	22.07	24.01
Pd/PBI-f-MWCNTs-6 : 1	54.03	55.78
Pd/PBI-f-MWCNTs-10 : 1	58.56	47.03
Pd/PBI-f-MWCNTs-16 : 1	65.92	33.16
Pd/P(DMBI)-I^−-f-MWCNTs-2 : 1	24.48	40.54
Pd/P(DMBI)-I^−-f-MWCNTs-6 : 1	56.43	77.45
Pd/P(DMBI)-I^−-f-MWCNTs-10 : 1	64.38	65.41
Pd/P(DMBI)-I^−-f-MWCNTs-16 : 1	67.58	45.78

---

For further investigating the electrocatalytic performance, CV studies of commercial Pd/C, Pd/PBI-f-MWCNTs and Pd/P(DMBI)-I⁻-f-MWCNTs in 1 M NaOH solution containing 1 M C₂H₅OH were conducted. As shown in Fig. 13b, there are two characteristic peaks, and anodic and cathodic peaks are all clearly observed, indicating that the Pd/C, Pd/PBI-f-MWCNTs (6 : 1) and Pd/P(DMBI)-I⁻-f-MWCNTs (6 : 1) exhibit similar catalytic-activity toward ethanol oxidation in alkaline solution. The anodic and cathodic peaks were centered at around −0.2 and −0.4 V, respectively.⁵⁶ The forward scan indicates the ethanol oxidation and the backward scan indicates the removal of the residual carbonaceous species. The intermediates generated in the forward scan probably diffuse into the bulk electrolyte and are fully oxidized at high potentials. The sharp increase also indicates that the peak is probably due to the oxidation of fresh ethanol, immediately after the reduction of PdO. Above all, the lower onset potential and higher peak current density are standard for evaluating the catalytic performance of catalysts. The forward peak current density of Pd/P(DMBI)-I⁻-f-MWCNTs was 55.98 mA cm⁻², which was much higher than that of commercial Pd/C (26.28 mA cm⁻²) and Pd/PBI-f-MWCNTs (47.74 mA cm⁻²) catalysts. The ratio of I_f/I_b can be used to investigate the catalyst tolerance to carbonaceous species accumulation.⁵⁷ The I_f/I_b for Pd/P(DMBI)-I⁻-f-MWCNTs (0.769) was higher than that of Pd/C (0.617) and the Pd/PBI-f-MWCNTs (0.764) catalyst. Obviously, the Pd/P(DMBI)-I⁻-f-MWCNTs exhibited highest activity for ethanol oxidation. It is quite clear that the good dispersion and small particle size of Pd NPs of Pd/P(DMBI)-I⁻-f-MWCNTs played a key role in electrocatalytic performance.

Further, chronoamperometric (CA) measurements were performed in 1 M NaOH solution containing 1 M C₂H₅OH, for a duration of 3600 s. As shown in Fig. 13c, all of Pd/C, Pd/PBI-f-MWCNTs and Pd/P(DMBI)-I⁻-f-MWCNTs exhibited a similar trend. All the polarization currents decay rapidly during the initial period and a pseudo-steady state is gradually achieved. The decrease in current is caused by the Pd surface poisoning induced by the CO_ads species.⁵⁸ It is obvious that Pd/P(DMBI)-I⁻-f-MWCNTs exhibits the lowest decay speed and is able to maintain the highest stable current, due to enough active Pd sites and the even particle distribution on the MWCNTs surface. Compared to Pd/P(DMBI)-I⁻-f-MWCNTs, a faster degradation is observed in Pd/PBI-f-MWCNTs, which is attributed to the clear agglomeration of Pd NPs observed in the TEM and SEM images of Pd/PBI-f-MWCNTs-6 : 1, probably deteriorating the tolerance stability in CA tests, due to the decreased active sites in the aggregated Pd NPs.

4. Conclusions

In conclusion, through methylation of PBI, P(DMBI)-I⁻ was successfully synthesized and used for dispersion and functionalization of MWCNTs. The P(DMBI)-I⁻ exhibited higher dispersing performance for MWCNTs. The concentration of P(DMBI)-I⁻-assisted dispersed MWCNTs reached ∼0.45 and ∼0.365 mg mL⁻¹ in NMP and DMF, nearly 5 and 4 times higher, respectively, than the concentration for PBI-assisted dispersed MWCNTs. With π–π conjugation and cation–π interactions between P(DMBI)-I⁻ and MWCNTs, P(DMBI)-I⁻ was wrapped on the surfaces of MWCNTs and achieved the functionalization of the MWCNTs. On that premise, the P(DMBI)-I⁻ is more beneficial to highly efficient and homogeneous growth and distribution of the Pd NPs onto the MWCNTs, compared to the original PBI. Under the optimum ratio of Na₂PdCl₄ and P(DMBI)-I⁻-f-MWCNTs of 6 : 1, the Pd NPs in Pd/P(DMBI)-I⁻-f-MWCNTs are well distributed on the surfaces of MWCNTs with particle sizes of mainly ∼2.9 nm. Electrocatalytic tests of ethanol oxidation in alkaline solution demonstrated that the electrochemical performance and tolerance stability of Pd/P(DMBI)-I⁻-f-MWCNTs are superior to commercial Pd/C and Pd/PBI-f-MWCNTs in the same feed ratio, due to the good distribution and small particle size of Pd NPs. We therefore believe that P(DMBI)-I⁻, as a derivative of PBI, can be further applied in the dispersion of different carbon-based materials and metal nanoparticles for fabrication of more novel composites for catalyst and electrode material.

Acknowledgements

This research was supported by Leading Talent Program of Shanghai, Sailing Program of Shanghai Science and Technology Commission (15YF1404700), Startup Fund for New Talent, Shanghai University of Electric Power (K-2014-046), National Nature Science Foundation of China (No. 21271010, 21671133, and 21604051) and Science and Technology Commission of Shanghai Municipality (No. 14JC1402500 and 14DZ2261000). Additionally, the study was also sponsored by “Chenguang Program” supported by Shanghai Education Development Foundation and Shanghai Municipal Education Commission (14CG55).

References

1. S. Abdullah, S. K. Kamarudin, U. A. Hasran, M. S. Masdar and W. R. W. Daud, J. Power Sources, 2014, 262, 401–406.
2. S. Rousseau, C. Coutanceau, C. Lamy and J.-M. Léger, J. Power Sources, 2006, 158, 18–24.
3. M. S. Ahmed and S. Jeon, ACS Catal., 2014, 4, 1830–1837.
4. O. Savadogo and F. R. Valera, ECS Trans., 2006, 1, 247–254.
5. S. Fu, C. Zhu, D. Du and Y. Lin, ACS Appl. Mater. Interfaces, 2015, 7, 13842–13848.
6. M. C. Figueiredo, J. Solla-Gullón, F. J. Vidal-Iglesias, M. Nisula, J. M. Feliu and T. Kallio, Electrochem. Commun., 2015, 55, 47–50.
7. S. M. Alia, K. Jensen, C. Contreras, F. Garzon, B. Pivovar and Y. Yan, ACS Catal., 2013, 3, 358–362.
8. D. R. Godoi, H. M. Villullas, F.-C. Zhu, Y.-X. Jiang, S.-G. Sun, J. Guo, L. Sun and R. Chen, J. Power Sources, 2016, 311, 81–90.
9. S. Jongsomjit, P. Prapainainar and K. Sombatmankhong, Solid State Ionics, 2016, 288, 147–153.
10. Z. Bai, P. Xu, S. Chao, H. Yan, Q. Cui, L. Niu, L. Yang and J. Qiao, Catal. Sci. Technol., 2013, 3, 2843.
11. M. Obradović, Z. Stančić, U. Lačnjevac, V. Radmilović, A. Gavrilović-Wohlmuther, V. Radmilović and S. L. Gojković, Appl. Catal., B, 2016, 189, 110–118.
12. Y. Liu, L. Wei, Y. Hu, X. Huang, J. Wang, J. Li, X. Hu and N. Zhuang, J. Alloys Compd., 2016, 656, 452–457.
13. A. Chen and C. Ostrom, Chem. Rev., 2015, 115, 11999–12044.
14. L. An, Z. H. Chai, L. Zeng, P. Tan and T. S. Zhao, Int. J. Hydrogen Energy, 2013, 38, 14067–14075.
15. R. Krishna, D. M. Fernandes, J. Ventura, C. Freire and E. Titus, Int. J. Hydrogen Energy, 2016, 41, 11811–11822.
16. J. Li, P. Zhou, F. Li, J. Ma, Y. Liu, X. Zhang, H. Huo, J. Jin and J. Ma, J. Power Sources, 2016, 302, 343–351.
17. E. Antolini, Appl. Catal., B, 2009, 88, 1–24.
18. Y. Wang, Q. He, H. Wei, J. Guo, K. Ding, Q. Wang, Z. Wang, S. Wei and Z. Guo, Electrochim. Acta, 2015, 184, 452–465.
19. A. Brouzgou, A. Podias and P. Tsiakaras, J. Appl. Electrochem., 2012, 43, 119–136.
20. D.-j. Guo and H.-l. Li, Electrochem. Commun., 2004, 6, 999–1003.
21. P. Singh, M. V. Kulkarni, S. P. Gokhale, S. H. Chikkali and C. V. Kulkarni, Appl. Surf. Sci., 2012, 258, 3405–3409.
22. C. Wang, H. Li, J. Zhao, Y. Zhu, W. Z. Yuan and Y. Zhang, Int. J. Hydrogen Energy, 2013, 38, 13230–13237.
23. Z.-P. Sun, X.-G. Zhang, Y.-Y. Liang and H.-L. Li, Electrochem. Commun., 2009, 11, 557–561.
24. N. M. Meman, B. Zarenezhad, A. Rashidi, Z. Hajjar and E. Esmaeili, J. Ind. Eng. Chem., 2015, 22, 179–184.
25. A. Eguizábal, M. Sgroi, D. Pullini, E. Ferain and M. P. Pina, J. Membr. Sci., 2014, 454, 243–252.
26. E. van de Ven, A. Chairuna, G. Merle, S. P. Benito, Z. Borneman and K. Nijmeijer, J. Power Sources, 2013, 222, 202–209.
27. C.-L. Sun, J.-S. Tang, N. Brazeau, J.-J. Wu, S. Ntais, C.-W. Yin, H.-L. Chou and E. A. Baranova, Electrochim. Acta, 2015, 162, 282–289.
28. T. Fujigaya, S. Hirata and N. Nakashima, J. Mater. Chem. A, 2014, 2, 3888–3893.
29. M. Z. F. Kamarudin, S. K. Kamarudin, M. S. Masdar and W. R. W. Daud, Int. J. Hydrogen Energy, 2013, 38, 9438–9453.
30. H.-L. Lin, Y.-C. Chou, T. L. Yu and S.-W. Lai, Int. J. Hydrogen Energy, 2012, 37, 383–392.
31. L.-c. Jheng, C.-y. Huang and S. L.-c. Hsu, Int. J. Hydrogen Energy, 2013, 38, 1524–1534.
32. Y. Zhang, Y. Wang, J. Yu, L. Chen, J. Zhu and Z. Hu, RSC Adv., 2014, 4, 35904–35913.
33. M. R. Berber, I. H. Hafez, T. Fujigaya and N. Nakashima, Sci. Rep., 2015, 5, 16711–16722.
34. Z.-F. Li, L. Xin, F. Yang, Y. Liu, Y. Liu, H. Zhang, L. Stanciu and J. Xie, Nano Energy, 2015, 16, 281–292.
35. M. Okamoto, T. Fujigaya and N. Nakashima, Small, 2009, 5, 735–740.
36. T. Fujigaya and N. Nakashima, Adv. Mater., 2013, 25, 1666–1681.
37. S. Yu and B. C. Benicewicz, Macromolecules, 2009, 42, 8640–8648.
38. J. Jouanneau, R. Mercier, L. Gonon and G. Gebel, Macromolecules, 2007, 40, 983–990.
39. A. Su, Y. M. Ferng, J. Hou and T. L. Yu, Int. J. Hydrogen Energy, 2012, 37, 7710–7718.
40. Q. Li, J. O. Jensen, R. F. Savinell and N. J. Bjerrum, Prog. Polym. Sci., 2009, 34, 449–477.
41. M. Hu, E. M. Pearce and T. K. Kwei, J. Polym. Sci., Part A: Polym. Chem., 1993, 31, 553–561.
42. K. Jiang, J. Chang, H. Wang, S. Brimaud, W. Xing, R. J. Behm and W. B. Cai, ACS Appl. Mater. Interfaces, 2016, 8, 7133–7138.
43. S. Attal, R. Thiruvengadathan and O. Regev, Anal. Chem., 2006, 78, 8098–8104.
44. Z. F. Li, G. H. Luo, W. P. Zhou, F. Wei, R. Xiang and Y. P. Liu, Nanotechnology, 2006, 17, 3692.
45. W. Yuan, J. Che and M. B. Chan-Park, Chem. Mater., 2011, 23, 4149–4157.
46. T. Luo, O. David, Y. Gendel and M. Wessling, J. Power Sources, 2016, 312, 45–54.
47. J. J. Linares, T. A. Rocha, S. Zignani, V. A. Paganin and E. R. Gonzalez, Int. J. Hydrogen Energy, 2013, 38, 620–630.
48. Q. Liu, J. Fan, Y. Min, T. Wu, Y. Lin and Q. Xu, J. Mater. Chem. A, 2016, 4, 4929–4933.
49. E. D. Sultanova, V. V. Salnikov, R. K. Mukhitova, Y. F. Zuev, Y. N. Osin, L. Y. Zakharova, A. Y. Ziganshina and A. I. Konovalov, Chem. Commun., 2015, 51, 13317–13320.
50. K. Qu, L. Wu, J. Ren and X. Qu, ACS Appl. Mater. Interfaces, 2012, 4, 5001–5009.
51. C. Hu, Y. Guo, J. Wang, L. Yang, Z. Yang, Z. Bai, J. Zhang, K. Wang and K. Jiang, ACS Appl. Mater. Interfaces, 2012, 4, 4461–4464.
52. W. Hong, J. Wang and E. Wang, ACS Appl. Mater. Interfaces, 2014, 6, 9481–9487.
53. S.-S. Li, J.-J. Lv, Y.-Y. Hu, J.-N. Zheng, J.-R. Chen, A.-J. Wang and J.-J. Feng, J. Power Sources, 2014, 247, 213–218.
54. S. C. Lai and M. T. Koper, Phys. Chem. Chem. Phys., 2009, 11, 10446–10456.
55. K. Lee, S. W. Kang, S. U. Lee, K. H. Park, Y. W. Lee and S. W. Han, ACS Appl. Mater. Interfaces, 2012, 4, 4208–4214.
56. Z. X. Liang, T. S. Zhao, J. B. Xu and L. D. Zhu, Electrochim. Acta, 2009, 54, 2203–2208.
57. T. Maiyalagan and K. Scott, J. Power Sources, 2010, 195, 5246–5251.
58. H. Qiu, X. Dong, B. Sana, T. Peng, D. Paramelle, P. Chen and S. Lim, ACS Appl. Mater. Interfaces, 2013, 5, 782–787.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra19256d

‡ These authors contributed equally to this work.

---