Low-cost preparation method of well dispersed gold nanoparticles on reduced graphene oxide and electrocatalytic stability in PEM fuel cell

1 Introduction

The intriguing characteristics of graphene, such as unique electrical, mechanical, physical and chemical properties have conferred a merited and recognized interest for this material, mainly due to the extremely high carrier mobility, quantum Hall effect or tunable work function, making it an attractive component for the fabrication of electronic devices. Graphene and also its various related materials demonstrated also interesting advantages in electrochemical devices such as fuel cells.

A highly desirable and long awaited large-scale commercialization of fuel cells is still restricted and conditioned widely by the slow kinetics of the cathodic oxygen reduction reaction (ORR). Commonly, Pt-based nanomaterials are used as best active electrocatalysts for both anode and cathode in proton exchange membrane fuel cells (PEMFCs) (Yaldagard et al., 2013; Meier et al., 2014; Hellman and Van den Hoed, 2007; Gasteiger et al., 2005).

However in the last few years Pt nanomaterial has begun to appear more and more outdated due to some unsolvable problems to date, such as susceptibility to the crossover effect, high cost together with its limited resources. Therefore, aiming to minimize the usage of Pt, the development of alternative efficient electrocatalysts for ORR at the cathode in fuel cells has generated a great interest. Moreover, the long term stability of the ORR catalyst has been already recognized as the key to overall system durability.

To replace platinum-based catalysts, the improved catalysts for ORR must meet the following three criteria: a specific volumetric activity in comparison with Pt-based catalysts in order to meet the US department of energy (DOE)’s 2015 target, excellent mass-transport properties and high durability. While the first two criteria have demonstrated recently some relevant scientific breakthroughs at a level sufficient for practical applications, the third criteria is still a major challenge. To date, ORR cathodes have not shown yet sufficient electrochemical stability for polymer electrolyte membrane fuel cells operation.

The strong dependence for the operation conditions makes sense to include the anticipated degradation as mandatory durability test. Therefore, the durability is still a critical issue to cope with PEMFC field in order to achieve a large commercialization and a real competition with combustion engines. In this respect, the lifetime of automotive FC devices drop in performance manly due to chemical and mechanical degradation. Relevant research focused on the degradation of FC, ranging from the detailed electrochemical modeling to empirical work on accelerated ageing testing, have been recently reported.

The accurate control of the electrical transport is considered critical for the development of cathodic electrocatalysts. The graphene oxide (GO), a 2D structured material with a large surface area, is considered an alternative to graphene, due to its low cost, production scalability or good compatibility. The structural and electronic properties of graphene can be modified by chemical adsorption of metal atoms or by incorporation of metal atoms into the lattice. The surface decoration using gas molecules or metal nanoparticles can tune the band structure of graphene. Thus the presence of defects in graphene is speculated to achieve new functionalities. The graphene doping with metal atoms has received great attention mainly in respect to the transition metal atoms bound on graphene or embedded in single and double vacancies (Liu et al., 2014; Hsieh et al., 2013; Shao et al., 2010; Cheng et al., 2014). Both n- and p-type conductivities can be obtained by selecting metal dopants, and the carrier concentration can be controlled by the quantity of deposited metal. For example, several transition metals with different work functions, such as Ti, Fe, Cu, Ag, Au, and Pt, have been used for graphene band modulation (Tang et al., 2011).

A suitable ORR catalyst has to offer both catalytic activity and selectivity, but also has to able to resist in the harsh chemical environment during long-term PEMFC operational conditions, such as presence of oxidant, reactive radicals, strong acidic conditions at low pH, rapid potential fluctuations. The specific environment from the PEMFC operation demands limited alternatives for materials that can be used as ORR electrocatalysts. Only three elements with bulk nominal dissolution potential U_diss higher than the FC operating potential of ∼0.8–1.0 V (vs the standard hydrogen electrode SHE) are recommended, namely: Ir, Pt, and Au, having U_diss of 1.16, 1.18, and 1.5 V (vs. SHE), respectively. Actually, as been demonstrated, in real FC application the problem is more severe. For example, in strong acidic medium for pH < 1, the Pt dissolution potential affords an on set electrode potential of 0.65 V, which is much lower than the operating potential of an efficient FC (Stolbov and Alcántara Ortigoza, 2015).

Despite of a known less catalytic activity of Au in comparison to Pt regarding the FC application, we propose to develop an ORR electrocatalytic system containing commercial Pt/C and AuNP, whose fabrication is going to be discussed later.

The redistribution of electronic states in graphene lattice could make the metal dopant atom more active toward ORR and Au might be the best option due to its electrochemical stability in acidic reaction, i.e. typical operating conditions of FC. Taking into account that the gold is the most stable metal against dissolution, recently few attempts were performed in order to incorporate Au atoms into the carbon vacancies created in graphene layers, and investigate its performance as ORR electrocatalysts. It has been indicated that Au is strongly bound to the di-vacancy (DV) and its diffusion barriers are high (∼4 eV), which ensures the stability of the catalytic system. Au-atom dopants are incorporated into one of the most thermodynamically stable carbon vacancies created in graphene: the 5-8-5 DV, so the Au-doped graphene could be fully considered a highly stable and active electrocatalysts for ORR (Stolbov and Alcántara Ortigoza, 2015).

More precisely, gold nanoparticles have received remarkable attention as catalytic nanoparticles due to its physical and chemical properties, such as easily controllable size distribution, long-term stability, and superior conductivity (Likai et al., 2016; Qiannan et al., 2016; Tang et al., 2018). However, there is still a significant challenge in controlling the uniformity of graphene surface decoration with metal nanoparticle and especially the strength of the attachment between the organic and the inorganic components. Despite bulk gold being chemically inert, gold nanoparticles are surprisingly catalytically effective for ORR without the drawbacks associated to platinum. Actually, creating stable AuNP catalysts has proven problematic. Previous fabrication methods have produced catalysts with more or less regulates sizes. Meanwhile, efforts to produce stable catalysts for practical use were addressed.

On the other hand, Poly-diallyldimethylammoniumchloride (PDDA) has been used as cationic polyelectrolyte, due to the ability of adsorption on the graphene surface through π−π and electrostatic interactions, acting as a stabilizer for controlled synthesis of Pt/graphene (Lei et al., 2013). The physical adsorption of PDDA, a positively charged polyelectrolyte with electron-retract ability on graphene could create net positive charge on carbon-atoms in the all-carbon graphene plane through the intermolecular charge transfer. Graphene nanosheets with negative charge can be easily functionalized by positively charged PDDA through the electrostatic interaction, which can increase the stability of graphene in solution. A key factor guaranteeing the feasibility of this system is that the resultant PDDA-functionalized/ adsorbed graphene possess enhanced electrocatalytic activities toward ORR (Wang et al., 2011; Borowiec et al., 2015). The Au loading density can be effectively adjusted by simply changing the ratio of PDDA to the Au precursor.

The proposed synthetic method is inexpensive, simple, rapid, requires low temperatures (close to room temperature). Starting material include extremely low cost material (graphite). The production of graphene oxide decorated with gold is currently achieved by graphene growth via chemical vapor deposition on a metallic surface, followed by transfer of the carbon layer on the oxide, by atomic layer and physical vapor deposition of the oxide on graphene or by carbon deposition on top of oxide surface. The mentioned methods involve some issues: they are costly, require complex procedures and can easily result in damage of the carbon network with substantial effects on the carrier mobility. Moreover, to the best of our knowledge, Au supported on reduced graphene oxide has not been reported as ORR catalyst in real-time PEMFC operation. In this paper we explored a low cost synthesis route of AuNP/rGO nanocomposites by in-situ co-reduction of GO and Au precursor in the presence of PDDA. The as-prepared sample was used in fabrication of a hybrid cathode and its electrochemical properties were investigated. The Au-based catalyst was added to the cathode layer, with the purpose of improving of electrocatalytic stability in real-time PEMFC operation conditions. The promising results may provide many potential applications for other electrochemical devices, in particular in PEMFC domain.

2 Experimental

3 Results and discussion

Au nanoparticles supported on graphene oxide nano-sheets have been chemically prepared using a simple procedure via a sequence of preparation steps based on chemical oxidation of graphite, graphene oxide functionalization, exfoliation and reduction, detailed in experimental section. Au nanoparticles were synthesized by the reduction method and adsorbed on graphene-PDDA nanohybrid sheets by electrostatic interaction. Using this strategy, AuCl₄⁻ ions could be confined in the positively charged surface of PDDA-functionalized graphene and then be in situ reduced to Au nanoparticles upon the addition of reducing agents. A controllable deposition of Au nanoparticles on graphene was performed. The structural characteristics, morphology, composition and electrocatalytic proprieties of prepared samples were evaluated through various characterization techniques.

The morphology of AuNP/rGO samples was characterized by FE-SEM technique. As it can be easily seen, the structure strongly depends on the drying method used. After simultaneous reduction of GO in the presence of PDDA and HAuCl₄, the surface of rGO nanosheets was decorated with AuNPs for both samples A and B.

As it can be seen from Fig. 1, AuNPs are distributed all over the surface of functionalized rGO sheets during both proposed synthesis protocol. In the case of lyofilization drying, the graphene layers were stacked together forming bundles with intermixed metal nanoparticles (Fig. 1a, b). Application of conventional drying in oven resulted in the structure of wrinkled features (see Fig. 1c, d).

Close

Fig. 1 The SEM images of as-prepared functionalized reduced graphene oxide decorated with gold: AuNP/rGO_A (a and b), AuNP/rGO_B (c and d).

Export to PPT

The surface functional groups of AuNP_rGO samples were investigated by FTIR spectroscopy (Fig. 2). The reduction of GO was depicted base on the weak peaks of —OH, C⚌O and O—H bond vibrations situated at 3400, 1720, and 1630 cm⁻¹ respectively. In the case of the functionalized and reduced graphene oxide, the intensity of these peaks decreased dramatically or even disappeared.

Close

Fig. 2 FT-TIR spectra of graphene oxide and AuNP/rGO nanocomposites.

Export to PPT

The functionalization of GO with PDDA is confirmed in spectra of AuNP/rGO nanocomposites for both samples, by appearance of supplementary peaks at 1463 and 850 cm⁻¹, assigned to the N—C bond from characteristic bands of PDDA.

The change of chemical composition during the reduction of GO and functionalization with PDDA and Au nanoparticles was investigated by X-ray photoelectron spectroscopic (XPS). Fig. 3 presents the XPS survey spectra for reduced graphene oxide (rGO) and AuNP/rGO nanocomposites. Main peaks for carbon, oxygen, nitrogen, chlorine and gold are indicated. The appearance of Cl 2p peaks in spectra of AuNP/rGO nanocomposites confirmed the functionalization of rGO with PDDA while Au 4f peak indicated presence of the gold.

Close

Fig. 3 XPS survey spectra for rGO and AuNP/rGO nanocomposites.

Export to PPT

High-resolution spectra were scanned (for atomic concentration calculations) and deconvoluted for detailed analysis of possible chemical bonds (Fig. 4a–l). The atomic and weight concentrations of the elements present in the samples are given in the Table 1. AuNP/rGO_A and AuNP/rGO_B samples exhibit similar surface composition. As expected, the O/C atomic ratio decreased as an effect of the NaBH₄ reduction, which was accompanied by the appearance of N 1s and Cl 2p peaks. As it is shown in Table 1, the significant amount of oxygen was found in the AuNP/rGO nanocomposites (∼9 at%) and in rGO sample (∼12 at%). The origin of oxygen could be addressed to efficiency of oxidation step. It is also worth to notice that equal amount of nitrogen and chlorine should be expected from the chemical formula of PDDA – (C₈H₁₆ClN)_n, nevertheless the N/Cl ratio of 8.43 and 6.23 was obtained for AuNP/rGO_A and AuNP/rGO_B sample, respectively. The reduced amount of chlorine compared to initial PDDA could be explained due to the reaction of Cl⁻ (from PDDA) with H⁺ during reduction process (Kaur et al., 2013). Deconvoluted high-resolution spectra for chlorine in Cl 2p region (Fig. 5g and h) showed Cl 2p_3/2 and Cl 2p_1/2 orbit split doublet located at 197.8 eV and 199.4 eV respectively which could be assigned to chlorine in PDDA (Marinoiu et al., 2017). Low intensity peaks at 201.2 eV and 202.7 eV indicates chlorine bounded to oxygen (NIST, 2003). Both assumptions suggest that applied method B provides less chlorine losses thus preserving unaltered PDDA during rGO functionalization process.

Close

Fig. 4 The high-resolution XPS spectra of AuNP/rGO nanocomposites and rGO.

Export to PPT

Close

Fig. 4 The high-resolution XPS spectra of AuNP/rGO nanocomposites and rGO.

Export to PPT

Close

Fig. 5 Raman spectra for rGO and AuNP/rGO samples.

Export to PPT

In contrary to chlorine, the amount of nitrogen in AuNP/rGO samples is much larger of that expected from the PDDA formula (theoretical N/C ratio is 1.02). Namely, the C/N ratio calculated from XPS spectra is 5.55% and 5.32% for AuNP/rGO_A and AuNP/rGO_B respectively. These values are similar to reported C/N value of 3.8% for PDDA – carbon nanotubes (CNTs) composite and 13% for pure PDDA in (Wang et al., 2011). Similar effect of nitrogen enrichment was also noticed for PDDA – CNTs composite where the amount of nitrogen increased up to 1.8–14 times (Zhang et al., 2013; Yang et al., 2005) compared to the theoretical composition of PDDA.

Deconvolution of nitrogen high-resolution N 1s spectra for AuNP/rGO nanocomposite samples revealed presence of one main peak and two lower intensity peaks in the spectra. Lowest intensity peak at approximately 404 eV can be assigned to nitrogen-oxygen bonds. The position of main nitrogen N 1s peak at ∼403 eV detected for both AuNP/rGO samples (Fig. 7) is typical for pure PDDA (403 eV (Yang et al., 2005) and 403.25 eV) (Kaur et al., 2013), and carbon allotropes functionalized with PDDA (403.25 eV (Kaur et al., 2013) and 403 eV (Yang et al., 2005). Moreover, the lower intensity peak at ∼400 eV observed for functionalized samples were also attributed to the PDDA as was previously reported (Dong et al., 2016; Marinoiu, 2017). Appearance of this peak could be due to an uncharged side product (Yang et al., 2005) when PDDA lose part of Cl⁺ (Kaur et al., 2013). This assumption coincides with effect of reduced chlorine amount as was discussed in the text above. Calculated ratio of these two peaks (at 403 eV and 400 eV) is 2.2 and 9.6 for the sample AuNP/rGO_A and AuNP/rGO_B samples respectively. These values fit well with other reported results (from 1.8 and 4.6 up to 13.3) (Yang et al., 2005; Dong et al., 2016; Marinoiu, 2017). On another hand, 9.6 value (AuNP/rGO_B sample) indicates that in this sample most of nitrogen is bound to chlorine. This assumption is in agreement with larger amount of chlorine found on the surface of AuNP/rGO_B sample (Table 1). The difference could arise due to loss of chlorine and increased amount of N.

Carbon C 1s spectra for PDDA functionalized rGO was deconvoluted using 7 peaks. These peaks indicate that part of carbon atoms are bound to oxygen or/and hydrogen, though some carbon-nitrogen bonds cannot be excluded also (Fig. 5a and b) considering probable loss of some chlorine from PDDA. The peak at lowest binding energy (284.3 eV) was assigned to C—C bonds (sp² configuration) – the binding energy value is close or identical to reported in (Marinoiu et al., 2017, 2019; Bertoti et al., 2015; Bayev et al., 2018; Vinayan et al., 2016) for C—C sp² bonds. This peak was fitted using 20:80 Gauss-Lorentz function and asymmetrical shape due to conductive nature of sp² carbon (Kahlert et al., 2014.). Peak at ∼284.9 eV was assigned to C—C bonds (sp³ configuration) – similar value of binding energy is reported in (Bertoti et al., 2015; Bayev et al., 2018; Marinoiu et al., 2019; Vinayan et al., 2016). It should be noted that binding energy values of 284.8 eV and 285.2 eV for C—C and C—H bonds respectively was reported in (Bertoti et al., 2015) for graphene oxide and pristine graphene respectively, therefore this peak most likely represents both C—C (sp³) and C—H bonds. Calculation of sp³/sp² ratio showed similar values (4.6 and 3.67 for NP/rGO_A and AuNP/rGO_B respectively) for both samples. The difference could arise due to chlorine losses and increase of N⁰ content – this effect could provide possibility for part of graphene to restack into graphite form (Wang et al., 2011) as was noticed in composite preparation part.

The rest of fitted C 1s spectra for PDDA functionalized rGO samples consist of the most intense peak (∼30%) at 285.7 eV, two lower intensity peaks at approximately 286.5 eV and 287.2 eV and two low intensity (less than 6%) peaks at 288.1 eV and 289.5 eV. The most intense peak at 285.7 eV was assigned to C—O—C bonds – similar binding energy value (285.5 eV) was found on surface of the C₃N film (Lazauskas et al., 2016) for these bonds. In addition, almost the same binding energy value was reported for C—N sp² bonds (285.8 eV) in N doped graphene (Lin et al., 2010) while in our case C-N bonds could originate from PDDA due to partial chlorine loss. Therefore, it is adequate to state that both components contribute to this peak, and C—O—C bonds could be dominant contributor since the ratio of O/N atomic concentrations is about 2.0 for both AuNP/rGO_A and AuNP/rGO_B samples. Fitted peaks at approximately 286.5 eV, 287.2 eV, 288.1 eV and 289.5 eV were assigned to C—O/C—OH, C⚌O/C—N sp³, O⚌C—O/C—N—C and O⚌C—OH bonds respectively. Similar binding energy values were reported for these bounds: C—O/C—OH (286.3 eV and 286.4 eV) (Marinoiu et al., 2017); C⚌O (287.3 eV and 287.7 eV) (Borowiec et al., 2015; Bertoti et al., 2015) and C—N sp³ (287.1 eV) (Bertoti et al., 2015); O⚌C—O (288.3 eV) (Bayev et al., 2018) and C—N—C (282.0 eV) (Lazauskas et al., 2016); O⚌C—OH (289.3 eV and 289.6 eV) (Borowiec et al., 2015; Bertoti et al., 2015). These peaks indicate considerable amount of carbon – oxygen bonds most likely due to atmospheric contaminants or/and PDDA surface oxidation or some residuals of graphene oxide.

In deconvoluted high-resolution C 1s spectra of rGO sample (Fig. 5k) one can see that the fitted peak at 284.4 eV dominates (60%). The asymmetrical shape of this peak is typical for C—C sp² bonds in graphene (Kahlert et al., 77, 2014.). The low intensity peak at 290.9 eV could be assigned to π-π shake up satellite (Bertoti et al., 2015). The majority of carbon-oxygen bonds most likely represents adsorbed atmospheric contaminants or/and some residuals of graphene oxide, while low sp³/sp² ratio value (0.2) indicates high efficiency of applied reduction process.

Oxygen spectra in O 1s region was deconvoluted using 4 curves for AuNP/rGO_A, AuNP/rGO_B and rGO samples as shown in Fig. 5e, f, l respectively. The lowest intensity peak at ∼534.5 eV was observed in the spectra of all samples and was attributed to O—H bonds in water (NIST, 2003). As the intensity of this peak is smallest for the sample AuNP/rGO_B, it could be stated that it adsorbed less moisture compared to the other samples. The peak at 531 V was attributed to O⚌C/O—C⚌O/O⚌C—OH bonds – position of this peak coincides with reported positions for O = C (530.6 eV, 531.0 eV and 531.4 eV) (Bertoti et al., 2015; Marinoiu et al., 2019; Vinayan et al., 2016) O—C⚌O (530.6 eV and 531.2 eV) (Bertoti et al., 2015; Jankauskaite et al., 2016), O⚌C—OH (530.8 eV) (Song et al., 2016) bonds. This peak is dominant in O 1s spectra for AuNP/rGO samples and is lower intensity for rGO sample. In contrary, peak at 533.2 eV is dominant for rGO sample and is less pronounced for AuNP/rGO samples. This peak was attributed to C—O—C and C—OH bonds (Bertoti et al., 2015; Wang et al., July 2014). Third peak at 532.1 eV was assigned to C—O bonds (Jankauskaite et al., 2016). Noticeable, intensity of C—O—C/C—OH and C—O peaks for sample AuNP/rGO_B correlates with lower intensity of O-H peak when compared with these peaks for AuNP/rGO_A sample. The difference of intensity of these three peaks in rGO and AuNP/rGO most likely arise due to different surface adsorption properties of the samples thus providing different amount of adsorbed atmospheric contaminants.

Only two peaks were detected in Au 4f spectra (Fig. 5i and j) of AuNP/rGO samples. Deconvolution of Au 4f region showed two asymmetrical peaks best fitted using 80:20 Lorenz-Gauss function. Position of these peaks at 84.0 eV and 87.7 eV are in good agreement with known orbit-split doublet positions of Au 4f_7/5 and Au 4f_5/2 peaks (Wang et al., 2011) for Au⁰. Position and shape of these peaks confirms the presence of metallic gold on the surface of functionalized rGO samples and suggests its successful decoration with gold nanoparticles. As shown in Au 4f spectrum of AuNP/rGO, only doublets of Au 4f5/2 and Au 4f7/2 were observed, indicating absence of Au²⁺ and Au⁴⁺ oxidation states. It is worth noting that gold nanoparticles existed only in metallic form despite the existence of a large amount of oxygen-containing functional groups in PDDA-functionalized rGO.

The Raman spectra of rGO and AuNP/rGO samples were analyzed. The Raman spectra provided in Fig. 6 indicate the presence of all characteristic bands of few layer graphene with structural defects. The AuNP/rGO composites have similar characteristic Raman peaks, which corresponds to the D and G bands from GO sheets. Three strong bands appeared at 1344–1347 cm⁻¹ (D band – associated with the presence of structural defects), 1588–1591 cm⁻¹ (G band – characteristic of all sp² – based materials), and a D-peak overtone band at about 2675–2680 cm⁻¹ (2D). Additionally, to these bands, in the case of nanocomposite samples, the D + G combination band was observed around  ∼2919 cm⁻¹ that is induced by disorder in the structure (Antony et al., 2015). The position and intensity of the peaks was determined by approximation using Lorentzian profile and summarized in Table 2.

Close

Fig. 6 Nitrogen adsorption-desorption isotherms and the corresponding pore size distribution for undoped rGO and AuNP/rGO.

Export to PPT

The 2D band intensity is generally very sensitive to the number of layers in the graphene structure. The relative intensity ratio of the D and G bands (I_G/I_D) may be associated with the number of defects within the graphene sheets and the in plane sp² crystallite size L_a (Hafiz et al., 2014). Generally, the I_D/I_G ratio increases when more defects are brought into GO. Due to further functionalization of rGO, the D band intensity increased dramatically and I_D/I_G ratios (Table 2) corresponding to AuNP/rGO composites were much higher (1.02 and 0.85 for sample A and B respectively) compared with rGO (0.51). The results illustrate that the loading amount of Au NPs has strong influence on the ratio, implying that the nucleation of gold nanoparticles at rGO surfaces introduces additional defects into structure.

Heteroatom-doped graphenes are well known for its largely exposed surface that interacts, therefore, evaluating the surface area is a crucial step in understanding and optimizing their performance. Currently, the most reliable method to evaluate the surface area is the classical BET method, which derives from nitrogen sorption isotherms collected at 77 K (Comparison of Average Number of Stacked Layers).

However, graphene sheets tend to stack on top of each other due to weak but extensive van der Waals interactions between their surfaces, which may appear quite likely during the chemical synthesis. Fig. 6 shows the nitrogen adsorption‐desorption isotherms and the corresponding pore size distribution curves of rGO and AuNP/rGO composites. Lower pressure region (quite similar to Type II) and intermediate flat region in the isotherms correspond to the formation of monolayer. At higher pressures the slope presents increased uptake of adsorbate as pores become filled, inflection point typically occurs near completion of the first monolayer followed by multilayer. The curves of studied samples are type IV (according to IUPAC classification), describing the adsorption behavior of porous adsorbents and indicating the presence of mesopores mainly, micropores and some macropores, together with hysteresis behavior between the adsorption and the desorption branch. A hysteresis appeared, with the shape of loops in-between type H3, assigned to the process of capillary condensation of nitrogen, and type H4, relative to the presence of pores with the shape of thin slots formed between two surfaces, indicating that the cleft‐like pores were formed by the aggregation of platelets‐like particles. The pores in an efficient ORR catalyst layer must act two complementary roles, namely the primary pores work as reaction volume, while secondary pores play the gas channel role in the porous structure.

Pore size distribution data was obtained from desorption branches of isotherms using Barrett-Joyner-Halenda (BJH) method and are illustrated as insets in Fig. 6. The results of surface area, pore volume, and pore radius are listed in Table 3. The BET specific surface area decreased from 245 m² g⁻¹ for rGO to 87 and 52 m² g⁻¹ for AuNP/rGO_A and AuNP/rGO_B, respectively and may be attributed to reduced porosity due to functionalization with Au nanoparticles. The measured surface area of functionalized graphene samples is consistent with the stacking structure and agglomerated morphology of the reduced graphene sheets.

BJH pore radius of 1.967 nm corresponding to rGO was close to the pore radius of gold doped graphene AuNP/rGO_A and AuNP/rGO_B (1.567 nm and 1.875 nm respectively), while cumulative pore volume (2.214 cm³ g⁻¹) is much bigger than doped graphene (0.112 cm³ g⁻¹ and 0.083 cm³ g⁻¹), indicating that a large number of mesopores have been clogged during doping process. The mentioned radii implied a hierarchical interconnected porous framework in doped graphene structure. It is known that different pore type displayed different roles in the PEMFC electrochemical performance (Avcioglu et al., 2015; Marinoiu et al., 2015). Their existence could offer sufficient space to allow access of the reactants to the catalytic sites and would accelerate the kinetic process of ion diffusion. Thus, such hierarchical porous nanomaterials, which facilitated solid electrolyte membrane penetration and ion diffusion, could be ideal for the high-performance electrode for PEMFC.

To evaluate the redox behavior of the AuNP/rGO samples, the cyclic voltammetry measurements using carbon paste modified electrodes (as described in experimental section) were performed. The electrochemical characterization of the electrodes modified with AuNP/rGO was undertaken in order to evaluate their electrochemical efficiency by using the ferro/ferricyanide redox probe (Marinoiu et al.) and the electroactive areas, respectively. For this purpose, parallel experiments with electrodes made of bare carbon paste and modified graphene oxide electrodes were performed. In addition, for a better evaluation of these modified electrodes, similar experiments with glassy carbon as working electrodes were performed. In all cases, the experimental results showed well defined and reproducible anodic and cathodic peaks related to Fe(CN)₆³⁻/Fe(CN)₆⁴⁻ redox couple with a quasi reversible behavior. The peak separation potential ΔEp (Epa − Epc) was observed between 180 and 325 mV for all electrodes. A comparison of electrodes CV measurements is presented in Fig. 7.

Close

Fig. 7 Cyclic voltammograms for 1.0 mM K₃Fe(CN)₆ in 0.5 M KCl solution at v = 20 mV s⁻¹.

Export to PPT

Analysis of the cyclic voltammograms revealed that all modified electrodes exhibit an enhancement of the current response when compared to classical glassy carbon which translates in better electrocatalytic activity for both oxidation and reduction processes. Moreover, comparing the performance of rGO modified electrode with the AuNP/rGO the positive effect of the metal was obvious. The best results were obtained for AuNP/rGO_A modified electrode for which the current was 3.94 times higher comparing with the CPE. Also, the anodic and cathodic peak potentials are shifted by around −100 mV for AuNP/rGO_A and −80 mV for AuNP/rGO_B, comparing with the potential of CPE. Electrochemical data and areas of the used electrodes are presented in Table 4.

An increase in the peak currents and a decrease in the separation between the peak potentials (ΔEp) of 20 mV s⁻¹ were observed for all modified electrodes in comparison to the bare carbon paste electrode, indicating that the electron transfer reaction was kinetically and thermodynamically favored at the used electrode surface (see Table 4). The electroactive areas of carbon-paste modified electrodes was calculated according to Randles–Sevcik equation using the slope of the I_a versus square root of the scan rates plot for the studied redox probe, namely 1.0 mM K₃[Fe(CN)₆] solution in 0.50 M KCl electrolyte: $$I_a=2.69·{10}^5·n^{3/2}·D^{1/2}·A·c·v^{1/2}$$ where I_a is the anodic peak current, n – number of transferred electrons, A –electroactive area of the electrode, D is the diffusion coefficient, ν is the scan rate and c is the concentration of potassium ferrocyanide solution.

The microscopic areas presented in Table 4 were calculated for n = 1 and D = 6.68 × 10⁻⁶ cm² s⁻¹ using the slope of the Ip–ν^1/2 relation. Analysis of the data revealed the positive effect of the gold component in the modified electrodes. Comparing the non-metallic modified graphene oxide (GO) electrode (entry 3, Table 4) with both AuNP/rGO electrodes, the superior performance of the latter was observed. The determined values of electroactive surface (entries 4 and 5) suggest that these materials could be used as electrocatalysts for ORR reaction in fuel cells. However, comparison between the two AuNP/rGO electrodes (A (entry 4) and B (entry 5)) show that the preparation of the composite is crucial for the resulting electrochemical efficiency.

The effect of varying the voltammetric scan rate is very useful because by increasing the scan rate, the current of the adsorption peak increases in magnitude. Therefore the AuNP/rGO electrodes were tested for Fe(CN)₆³⁻/Fe(CN)₆⁴⁻ redox couple at scan rate from 20 to 120 mV/s (Figs. 8 and 9). As the scan rate was increased, the current for both anodic and cathodic peak also increased and both peak potentials remained almost unchanged over the chosen scan rate range. These cyclic voltammograms were used to monitor the variation of the voltammetric response as a function of scan rate (v). The plot of the peak height versus square-root of the scan rate is illustrated in Fig. 8(b) and Fig. 9(b). The shape of the voltammograms and their variation with the scanning speed is similar for both samples, also the variation of the potential (oxidation and reduction) has a linear dependence on speed. Both, the anodic and the cathodic peak currents reveal linear response at all scan rates with very good correlation factors, R² = 0.992 for the oxidation peak and R² = 0.999 for the reduction peak. This is indicative of a diffusion-controlled electrochemical process.

Close

Fig. 8 Cyclic voltammograms for 1.0 mM K₃Fe(CN)₆ in 0.5 M KCl solution for AuNP/rGO_A electrode, v = 20–120 mV s⁻¹ (a) plot I vs. v^1/2 (b).

Export to PPT

Close

Fig. 9 Cyclic voltammograms for 1.0 mM K₃Fe(CN)₆ in 0.5 M KCl solution for AuNP/rGO_B electrode, v = 20–120 mV s⁻¹ (a) plot I vs. v^1/2 (b).

Export to PPT

In conclusion, the electrochemical performance for ferri/ferro redox process is higher for Au/rGO_A modified electrode, its sensitivity at higher scanning rates showing well defined peaks. Thus, benefiting from unique features, the developed Au/rGO_A material indicated good electrochemical performances as possible electrodes. The challenge of achieving an electrochemical stability is considered especially difficult because, for ORR catalysts, the role of support should be combined with the role of catalyst. So, while the microporosity is required to obtain high catalytic activity and a high mesoporosity is required for an efficient transport of ORR related species (O₂, H₂O and protons) in the catalyst layer, moreover, a nanostructured catalyst with high stability in acid media is necessary. In this work, we explored the use of AuNP/rGO as co-catalyst in creating of an improved ORR cathode which includes: Pt/C catalyst that assures the ORR catalytic activity, and AuNP/rGO that allows a high electrochemical stability of cathode catalytic system.

Electrical performance of developed cathode was performed by polarization curve (Fig. 10). In order to evaluate the power density, a fuel cell voltage of 0.6 V is recommended, because it provides an accurate balance between energy efficiency and power density for well performing electrocatalysts. The membrane electrode assembly (MEA) using the developed cathode and tested in PEM fuel cell, produced a peak power density of 0.65 W cm⁻² and a power density of 0.59 W cm⁻² at 0.6 V, the latter comparable with that of a state-of-the-art platinum-based cathode having a loading of 0.3 mg_Pt cm⁻² (Peng et al., 1765; Proietti et al., 2011).

Close

Fig. 10 Polarization curve plot and Power density plot of a hydrogen-air PEMFC including developed cathode (with AuNP/rGO); Cell runs with H₂/air at 1 bar anode and cathode back pressure and 70 °C fuel cell temperature.

Export to PPT

In order to validate the previously ex-situ CV results, AuNP/rGO material was involved in fabrication of a hybrid cathode, as follows: commercial Pt/C 0.2 mg_Pt cm⁻² loading was deposited on Nafion membrane and prepared AuNP/rGO with 0.2 mg_Au cm⁻² loading was deposited on carbon paper. These two layers form a novel ORR catalytic system which was evaluated in comparison to cathode containing no Au/rGO. In actual study, the Pt loading was established at 0.2 mg_Pt cm⁻² for anode of both developed fuel cells. The cathode catalyst layer was evaluated in two configurations: (i) commercial Pt/C catalyst with 0.4 mg_Pt cm⁻² loading (test 1); developed ORR catalytic system, including commercial Pt/C 0.2 mg_Pt cm⁻² and prepared Au/rGO Pt/rGO with 0.2 mg_Pt cm⁻² loading (test 2).

Taking into account the application of Au/rGO_A material as ORR electrode for PEMFC, the in-situ electrochemical accelerated stress tests (ASTs) at potentials above 0.0 V_SHE were carried out in order to accelerate the degradation of catalysts aiming to assess their durability. The state of the art data highlights that the potential cycling may cause the remarkable most relevant changes of the catalyst structure. Overall, the carbon corrosion at cathode side must be understood as a structural breakdown of support from ORR catalyst, which results in the separation or/and agglomeration of Pt particles. Thus, the catalyst involvement in accelerated test was focused on potential cycling between 0.0 and 1.2 V_SHE, which theoretically cause a fast catalyst aging, in contrast to a constant potential keeping test.

Figs. 11 and 12 present the cyclic voltammogrames corresponding to catalysts exposure to 50 cycles between 0.02 and 1.2 V potential cycling for 2 cases: (i) commercial Pt/C and (ii) developed catalytic system Pt/C – AuNP/rGO respectively. During the forward voltage sweep in the corresponding CV measurement (0.05–0.4 V), the protons permeates through the membrane, then are adsorbed on the metal surface from the cathode side. In the backward voltage sweep the generated H⁺ is then adsorbed and oxidized on the Pt surface. The hydrogen desorption peak at around 0.2 V during the forward scan decreased during AST mainly in FC operation condition corresponding to 1 bar pressure. Moreover, the CV results showed that the oxidation and reduction peaks corresponding to hydrogen adsorption/desorption (0.025–0.3 V) decreased differently during AST, indicating that a decrease of electrochemical surface area occurs for both cases, more visible for commercial catalyst.

Close

Fig. 11 Cyclic voltammetry of commercial Pt/C for 50 cycles in following operation conditions: temperature 70 °C; scan rate 50 mV/s; atmospheric pressure (a) and 1 bar (b).

Export to PPT

Close

Fig. 12 Cyclic voltammetry of developed ORR catalytic system for 50 cycles in following operation conditions: temperature 70 °C; scan rate 50 mV/s; atmospheric pressure (a) and 1 bar (b).

Export to PPT

Some observations appear, namely:

• –

  the shapes of curves are fairly similar for the two samples demonstrating that they generate well-defined adsorption/desorption hydrogen regions. Although, it seems numerically easy to evaluate the electrochemical surface area, the validity of this area approach is questionable as it assumes that the contributions from the platinum and gold are independent;

• –

  the MEA including AuNP/rGO as ORR catalyst showed a good electro-stability in PEMFC durability tests.

Moreover, it is easily to notice the change in hydrogen adsorption/desorption on commercial Pt single catalyst both for hydrogen oxidation and hydrogen reduction (Fig. 10). The lower PEMFC performance in the first case was attributed to the mass transport limitations and carbon corrosion caused by carbon support.

As regards the developed hybrid catalyst, insignificant changes were noticed for reduction signal, suggesting the potential of using Au/rGO as ORR catalyst for future AST durability investigation. Such behavior may be assigned to:

1. long-term stability and superior conductivity of Au based catalyst fulfilling in this case the role of co-catalyst;

2. hydrophilic nature of graphene oxide support, which comports like a water reservoir leading to a better hydration of membrane. Thus a lower ohmic resistance and higher proton conductivity are expected. The detailed long-term durability tests of PEMFC with Au/rGO catalyst and the degradation studies in different PEMFC conditions are ongoing studies.

4 Conclusions

In this research, graphene oxide nanosheets were functionalized with cationic PDDA and decorated with gold nanoparticles. Au nanoparticles were synthesized by the reduction method and then adsorbed on graphene-PDDA nanohybrid sheets by electrostatic interaction. The electrochemical measurements revealed an enhanced performance of the electrode after addition of AuNP/rGO modified electrode. Sharp and well-resolved peaks are observed when CV experiments were utilized. The anodic and cathodic peaks are higher for graphene oxide modified carbon paste electrodes with the higher peak current observed for gold doped reduced graphene oxide. After CV durability testing, the performance loss between the Pt/C and developed electrode based on Pt/C – Au/rGO were observed to be smaller for hybrid electrode. In summary, a novel AuNP/rGO catalyst with a graphene structure was successfully prepared by a low-cost synthesis. The addition of this material to Pt-based catalyst exhibited a high electrocatalytic stability for the ORR in an acidic medium. The Au–graphene electrocatalytic system can be potentially further developed for possible applications as cathodic ORR electrode in PEMFC.