Activation of Reduced-Graphene-Oxide Supported Pt Nanoparticles by Aligning with WO3-Nanowires toward Oxygen Reduction in Acid Medium: Diagnosis with Rotating-Ring-Disk Voltammetry and Double-Potential-Step Chronocoulometry

Activation of Reduced-Graphene-Oxide Supported Pt Nanoparticles by Aligning with WO3-Nanowires toward Oxygen Reduction in Acid Medium: Diagnosis with Rotating-Ring-Disk Voltammetry and Double-Potential-Step Chronocoulometry Iwona A. Rutkowska, 1,∗ Anna Wadas,1 Sylwia Zoladek,1 Magdalena Skunik-Nuckowska,1 Krzysztof Miecznikowski,1,∗ Enrico Negro,2 Vito Di Noto, 2,∗ Agnieszka Zlotorowicz,1,2 Piotr Zelenay,1,3,∗∗ and Pawel J. Kulesza 1,∗∗,z

Development of catalytic systems for oxygen reduction reaction (ORR), particularly with respect to potential applications in low-temperature fuel cells, [1][2][3][4][5][6][7][8][9][10][11][12][13] is still one of the most important areas of electrocatalysis. Considerable research efforts have centered on the development of Pt-free or low-Pt-content catalytic systems. 14 An ultimate goal would be to replace completely Pt with non-precious metal catalysts, e.g. by considering the metalnitrogen/carbon compounds, [15][16][17] metal oxide/oxysalts, 18,19 metalorganic-frameworks 20 or just bare or functionalized carbons. 21,22 Although these nonprecious metal alternatives could exhibit in principle activities comparable to, or sometimes even better than, those characteristic of Pt in alkaline electrolytes, most of them have been less promising both in terms activity and stability (relative to Pt-based systems) in acid environments. Indeed, the commercially available applications of fuel cells (e.g. in vehicles) are mostly based on proton exchange membranes because the practical utilization of alkaline membranes is rather limited. [23][24][25] Consequently, serious research has been focused on improving the catalytic activity of Pt-based catalysts in acid media with less Pt loading. For example, alloying Pt with transition metals has been attempted to modify electronic structure of Pt surfaces thus resulting in improved ORR activity. 26,27 In other words, there is a need of better utilization of catalytic sites and significant lowering of the noble metal loadings.
In addition to large electrochemically active surface area and the presence of highly dispersed active sites, while exhibiting long-term stability, a useful support should prevent agglomeration of catalytic centers, facilitate oxygen mass transfer and water removal, and assure good electrical conductivity at the electrocatalytic interface. Because of the high specific surface area and excellent thermal, mechanical and electrical properties, graphene-based electrocatalysts have recently been considered. Obviously, special attention has been first devoted to noble-metal-free carbon-based (e.g. heteroatom-doped, iron or cobalt modified, totally-metal-free, functionalized or derivatized) nanomaterials including graphene-type systems. [28][29][30][31][32][33][34] Different concepts of utilization, including nanostructuring, doping, admixing, preconditioning, modification or functionalization of various graphene-based systems for catalytic electroreduction of oxygen have been explored. [35][36][37][38][39][40][41][42][43][44][45][46] In the present work, we consider the chemically-reducedgraphene-oxide-supported dispersed Pt nanoparticles (loading, 30 μg cm −2 ) that have been intentionally modified with tungsten oxide nanowires (loading, 300 μg cm −2 ) as the catalytic system for the electroreduction of oxygen in acid medium (0.5 mol dm −3 H 2 SO 4 ). Reduced graphene oxides with interfacial defects, moderate hydrophilicity, low degree of organization and porosity seem to be suitable for use as supports for anchoring noble metal nanoparticles. By analogy to graphene oxide, the existence of oxygen groups in the plane of carbon atoms of reduced graphene oxide not only tends to increase the interlayer distance but also makes the layers somewhat hydrophilic. Among other important issues are such features of tungsten oxide as porosity, large population of hydroxyl groups, high Broensted acidity, as well as fast electron transfers coupled to unimpeded proton displacements. The usefulness of WO 3 nanostructures during the reduction of oxygen was demonstrated. 4 Indeed, WO 3 was found to exhibit high reactivity toward the reductive decomposition of the hydrogen peroxide intermediate. Here, we have utilized the so-called reduced graphene oxide platelets which, contrary to conventional graphene, still contains oxygen functional groups regardless of subjecting it to the chemical reduction step. 47 It is apparent from the conventional and rotating ring-disk voltammetric measurements that the system utilizing Pt nanoparticles and reduced-grapheneoxide-supports decorated with tungsten oxide nanowires behaves as the potent O 2 -reduction electrocatalytic system. Finally, the utility of J3385 double-potential-step chronocoulometry as the diagnostic tool has also been demonstrated.

Experimental
Chemical reagents were analytical grade materials. Tungsten(VI) oxide nanowires and 5% Nafion-1100 solution were purchased from Aldrich. Platinum black was obtained from Alfa Aesar. Sulfuric acid was from POCh (Gliwice, Poland).
All solutions were prepared using doubly-distilled and subsequently deionized (Millipore Milli-Q) water. They were deoxygenated by bubbling with high purity nitrogen. Measurements were made at room temperature (22 ± 2 • C).
All electrochemical measurements (including voltammetry and double-potential-step chronocoulometry) were performed using a CH Instruments (Austin, TX, USA) Model 760D workstation in three electrodes configuration. The glassy carbon working electrode was utilized in a form of the disk of geometric area, 0.071 cm 2 . The reference electrode was the K 2 SO 4 -saturated Hg 2 SO 4 electrode, and the carbon rod was used as the counter electrode. All potentials reported here were recalculated and, unless otherwise stated, expressed vs. Reversible Hydrogen Electrode (RHE).
The rotating ring-disk electrode (RRDE) voltammetric experiments were conducted via variable speed rotator (Pine Instruments, USA). RRDE assembly included a glassy carbon disk (diameter 5.61 mm) and a platinum ring (inner and outer diameters were 6.25 and 7.92 mm, respectively). The collection efficiency, 0.39, was determined by usual means from the ratios of ring and disk currents obtained by performing six independent RRDE measurements at the rotation rate of 1600 rpm in the deoxygenated solution of 5 mmol dm −3 K 3 [Fe(CN) 6 ] and 0.01 mol dm −3 K 2 SO 4 . Prior to all electrochemical experiments the working electrode was polished with aqueous alumina slurries (grain size, 0.05 μm) on a Buehler polishing cloth.
Reduced graphene oxide (rGO) was obtained in the course of the hydrazine reduction method in an analogous manner as described before. 47 In brief, 10 ml of 50% hydrazine water solution was added to 100 ml of 0.5 wt% graphene oxide (GO) water dispersion. The mixture was heated up to 100 • C and kept under stirring for 2 h. 47 After reduction, the product was filtered using polyethersulfone filter of the 0.8 μm pore size. To finalize preparation of the reduced graphene oxide (rGO) suspension, 50 mg of the resulting material was subjected to ultrasonic bath in 5 cm 3 of deionized water. Modification of glassy carbon electrode was achieved by deposition of 2 μdm 3 of the above suspension on the electrode followed by drying in air. Loading of rGO was equal to 300 μg cm −2 .
Suspensions (inks) of platinum black nanoparticles were prepared by dispersing 2.1 mg of the Pt commercial samples through sonication for 120 min in 2.0 cm 3 of distilled water to obtain a homogenous mixture. Later, 2 μdm 3 of the ink was placed onto the film surface to yield the noble metal loading of 30 μg cm −2 .
WO 3 nanowires were prepared by dispersing 1.0 mg of the material in 0.1 cm 3 of deionized water using an ultrasonic bath for 1 h. Later 2.0 μdm 3 of aqueous suspension of WO 3 nanowires was used to cover catalytic films. Loading of WO 3 nanowires was equal 300 μg cm −2 .
To prepare a hybrid system, first rGO was dropped as ink onto the glassy carbon electrode surface, followed by introduction of Pt nanoparticles and, at the end, by covered with the ink of tungsten oxide nanowires.
As a rule, the films were over-coated and stabilized with ultrathin layers of Nafion polyelectrolyte by depositing 1 μdm 3 of the Nafion solution (prepared by introducing 5%mass of the commercial Nafion solution into ethanol at the 1 to 10 volumetric ratio).
The catalytic materials (films) were first pre-treated in the deaerated 0.5 mol dm −3 H 2 SO 4 electrolyte by subjecting them to repetitive potential cycling (in the range from 0.04 to 1.04 V vs. RHE) at 10 mV s −1 for 30 min. Before the actual electrocatalytic experiments were performed, the electrodes were conditioned by potential cycling (10 full potential cycles at 10 mV s −1 ) in the potential range from 1.04 to 0.04 V (vs. RHE) in the oxygen-saturated 0.5 mol dm −3 H 2 SO 4 . Before each representative voltammogram was recorded, the working electrode was kept for 20 s ("quiet time") at the starting potential.

Physicochemical identity of graphene-supported Pt and WO 3 nanostructures.-
The graphene-based catalytic materials were characterized using Transmission Electron Microscopy (TEM). It is apparent from Fig. 1 that, while the reduced graphene oxide (rGO) deposits are in a form of platelet nanostructures (Fig. 1A), the platinum nanoparticles introduced onto rGO (Fig. 1B) are largely dispersed but in some cases Pt agglomerates are formed. On mechanistic grounds, the attachment of Pt may take place at the rGO "defect" sites including surface polar groups. It is commonly accepted 36 that partially reduced graphene oxide, rGO contains various carbon-oxygen groups (hydroxyl, epoxy, carbonyl, carboxyl), in addition to the large population of water molecules still remaining in the reduced samples. Furthermore, the Raman spectra, which were reported earlier, 47 showed that intensity of the G band at 1580 cm −1 (reflecting graphitic structures of carbon) is rather low and the intensity the D band near 1350 cm −1 (originating from the amorphous structures of carbon) is relatively high. In other words, this result implies presence of interfacial defects and rather low degree of organization of the graphitic structure in the rGO material used here. Most likely WO 3 nanostructures (illustrated in Inset to Fig. 2) are attached mainly to the edges of graphene thus preventing its stacking and creating space for the flux of oxygen and the reaction products. 36  Figure 2 shows cyclic voltammetric behavior of WO 3 nanowires investigated as deposit on glassy carbon. Judging from the transmission electron micrograph of WO 3 material (Inset to Fig. 2), while diameters of the nanowires are on level of 20 nm, their lengths are in submicrometer range. Two dominanting sets of the system's voltammetric peaks appearing at potentials lower than 0.3 V reflect redox transitions of tungsten oxide consistent with the formation of two partially reduced forms of WO 3 , namely hydrogen tungsten oxide bronzes of the type H x1 WO 3 and H x2 WO 3 . 4 The electrochemical behavior of the latter nonstoichimetric forms is characterized by fast and reversible redox transitions. The fact, that the pair of peaks at about −0.2 V is not exactly symmetrical around the zero current axis and the reduction peak current appears relatively higher the respective oxidation one, reflects most likely contribution from the parallel irreversible generation of lower tungsten oxides, WO 3-y , accompanied by sorption of hydrogen. 4

Reduction of O 2 at Pt nanoparticles deposited onto rGO-based
catalysts.-The rGO-supported Pt nanoparticles are obviously less active (larger sizes and lower electrochemically active surface area) than conventional Vulcan-supported Pt during electroreduction of oxygen under conditions of the RRDE voltammetric diagnostic experiments at the comparable loadings (30 μg cm −2 ). On the other hand, the rGOsupported Pt (Fig. 3A) exhibits appreciable electrocatalytic activity during reduction of oxygen in acid medium. Apparently, when acting as nanostructured support, rGO facilitates dispersion of Pt catalytic sites presumably due to its anchoring capabilities through interactions with the rGO interfacial groups. As expected from the voltammetric response recorded in the deoxygenated supporting electrolyte (dashed line in Fig. 3A), while oxidation of Pt to PtO is apparent at potentials higher than 0.7 V, the peaks existing at potentials lower than 0.35 V should attributed, as expected for Pt, to hydrogen adsorption/desorption phenomena. On the other hand, these peaks seem to be less defined and more poorly developed (in comparison to what is typically observed on clean polycrystalline platinum) 1-4 most likely due to the existence of strong interactions between immobilized Pt nanostructures and surface groups existing within rGO pores.
In the present work, we have also considered (Fig. 3B) the rGOsupported Pt nanoparticles further decorated with WO 3 nanorods (300 μg cm −2 ). Comparison of the background-subtracted voltammetric responses recorded in the oxygen saturated 0.5 mol dm −3 H 2 SO 4 ) shows clearly that WO 3 -modified rGO-supported Pt nanoparticles exhibits relatively the highest electrocatalytic currents (curve c in Fig.  3C), relative to those recorded for the same loading of platinum deposited on the WO 3 -free rGO support and the simple bare glassy carbon electrode substrate.
It is also apparent from the rotating ring-disk (RRDE) experiments (Fig. 4) that the disk (background-subtracted) currents have occurred to be higher during the reduction of oxygen at the WO 3 -modified rGO-supported Pt nanoparticles (Fig. 4B) relative to the performance of pristine Pt nanoparticles deposited directly onto rGO (Fig. 4A). As expected, the ring currents which permit monitoring of the formation of the undesirable hydrogen peroxide intermediate are lower in a case of the WO 3 -decorated system (Fig. 4B') relative to the performance of the bare rGO-supported Pt (Fig. 4A'). Fig. 5A illustrates the respective dependencies of limiting currents on the square root of rotation rate. The negative deviation from linearity (which is indicative of kinetic limitations) is much more pronounced for the the WO 3 -free system. Also the Koutecky-Levich reciprocal plots (I disk −1 = (nFA k s C OXYGEN ) −1 + I LEVICH −1 , where k s stands for heterogeneous rate constant and other parameters have usual significance) are characterized by higher intercepts in the latter case (Fig. 5B)   where I ring and I disk are the ring and disk currents, respectively, and N is the collection efficiency (equal to 0.39) and plotted against the applied potential (Fig. 5C). The results clearly show that the production of H 2 O 2 is lower in a case of the system utilizing tungsten oxide nanowires, particularly at potentials lower than 0.4 V. The overall number of electrons exchanged per O 2 molecule (n) can be a measure of the effectiveness of the process and has the mechanistic significance: namely it is a measure to what extent the reaction proceeds effectively according to the direct 4-electron pathway to water (O 2 + 4H + + 4e− → 2H 2 O) rather than the 2-electron one (O 2 + 2H + + 2e− → H 2 O 2 ) with hydrogen peroxide intermediate. Assuming that there are no other limitations, the n value was calculated and expressed as a function of the applied potential using the RRDE voltammetric data of Figs. 5A, 5A', 5B and 5B' and the equation given below: The corresponding number of transferred electrons (n) per oxygen molecule (Fig. 5D involved in the oxygen reduction was obviously higher (and closer to the ideal value of n = 4) in the case of the system utilizing WO 3 -decorated Pt nanoparticles supported onto rGO. The differences between the determined n values became apparent at potentials lower than 0.4 V (Fig. 5D). For example, the n values (determined at 0.1 V) were equal to 3.90 (±0.07) and 3.92 (±0.07) for the WO 3 -free and WO 3 -containing systems, respectively. The calculations were based on the highly reproducible (six independent) RRDE experiments at 1600 rpm; thus the standard deviations of n were on the level ±0.07.

Reduction of O 2 under chronocoulometric conditions.-An
alternative and a very useful mode of recording the electrochemical responses is to integrate the current and to report charge passed as a function of time. 48,49 Chronocoulometry offers important advantages including good signal-to-noise ratio because the act of integration smooths random noise on the current transients. In comparison to chronoamperometry, by integrating the response, it is possible to separate surface phenomena (double layer-charging, surface electrochemistry) more readily from bulk electrochemical responses. When it comes to the oxygen reduction, the double-potential-step chronocoulometry permits, in principle, estimation of the number of electrons (n) involved in the oxygen reduction and diagnosis of formation of the hydrogen peroxide intermediate. The problem lies in understanding of the reaction kinetics and mechanisms in order to define properly experimental parameters (pulse length, starting and final potentials). Figure 6A illustrates a representative (conventional) response for the double-potential-step experiment performed on oxygen reduction using rGO-supported Pt nanoparticles (the same catalyst as for the Fig. 5 RRDE investigations). Similar responses have been recorded for bare Pt nanoparticles as well as for the WO 3 -decorated system (for simplicity not shown here). In the diagnostic experiment, the potential is shifted, or stepped, from the initial potential (e.g. 1.1 V), where insignificant electrolysis (at least in the O 2 -containing H 2 SO 4 solution) takes place, to the final potential of 0.3 V that is sufficiently negative to enforce a diffusion-limited current. By plotting charge (Q) versus square root of time (t 1/2 ), the plot of the "diffusional" charge is linear (in Fig. 6B, at the t 1/2 values larger than 0.15). As expected, 48 the total charge Q vs. t 1/2 has not passed through the origin, because additional contributions to Q arising from double-layer charging and from the electroreduction of any surface species (e.g. PtO, oxygencontaining groups on rGO, etc.) including molecules (e.g. CO) that might be adsorbed at the time of application of the starting potential. The charges devoted to those processes are passed very quickly when compared to the slow accumulation of the diffusional component. Consequently, they are included in the integrated Cottrell equation by adding two time-independent terms: Q = 2nFπ 1/2 r 2 D app 1/2 C 0 /t 1/2 + Q dl + nFA 0 [3] where Q dl is the capacitive charge, nFA 0 quantifies the faradaic component given to the reduction of the surface excess, and 0 stands for surface concentration (in mol cm −2 ), of adsorbed species. The intercept of Q vs. t 1/2 is therefore equal to Q dl + nFA 0 . Thus chronocoulometry permits us is to comment on the surface phenomena occurring at the electrocatalytic interface as well. Simple examination of the approximate values of nFA 0 (obtained by considering intercepts of the Q-t 1/2 plots in Fig. 6B and in the Inset) implies that, regardless the presence or absence of oxygen, they are comparable for both forward and reciprocal steps, respectively. In other words, the present system is well-behaved (no degradation on the time scale of the experiment) and permits pursuing further diagnostic experiments.
More detailed study of the surface processes occurring at the electrocatalytic interface during reduction of oxygen will be a subject of our next communication.
The systems which can be studied using the rotating disk voltammetric methodology (including oxygen reduction) are usually wellsuited for chronocoulometry because the diffusional region can be clearly established. The slopes (Q/t 1/2 ) of the diffusional (linear) portions of the Q vs. t 1/2 plots (Fig. 7B) can have diagnostic meaning by using the equation as follows: Q/t 1/2 = 2nFπ 1/2 r 2 D O2 1/2 C O2 [4] where r, D O2 and C O2 stand for the radius, diffusion coefficient, and concentration of redox centers (here oxygen), as well as n is a number electrons (for oxygen ideally n = 4), and F is Faraday constant (96500 C mol −1 ). In practice, there is a need to correct the Q/t 1/2 slope determined in the presence of oxygen by subtracting the blank slope determined from the background measurement performed in the oxygen-free electrolyte (Fig. 7A). For the oxygen saturated 0.5 mol dm −3 H 2 SO 4 at 20 • C, where D O2 = 1.4 * 10 −5 cm 2 s −1 , and C O2 = 1.1 * 10 −6 mol cm −3 (1.1 mM), 50 the theoretical value of Q/t 1/2 has been calculated to be equal to 1.18 * 10 −4 C s −1/2 . Using the data, namely the background (electrolyte response) corrected slope of the liner portion of Fig. 7 (for WO 3 -free rGO-supported platinum nanoparticles), the value of Q/t 1/2 (slope) equal to 1.10 (±0.013) * 10 −4 C s −1/2 has been obtained. This parameter can be determined with high accuracy (standard deviation, ±0.013 * 10 −4 C s −1/2 ) provided that the correlation coefficient (linear fit) is on the level 0.9999. Based on Equation 4, simple comparison of the latter value to the theoretical one of 1.18 * 10 −4 C s −1/2 implies that most likely ca. 3.7 electrons, instead of the theoretical 4 electrons, have been involved in the oxygen reduction under chronocoulometric conditions (Fig. 7) with the potential step down to 0.3 V. Using the same approach to comment on the performance of the WO 3 -decorated system (graphs are for simplicity not shown here), the value of Q/t 1/2 equal to 1.14(±0.014) * 10 −4 C s −1/2 and n close to 3.9 have been determined. On the whole, these values are consistent with the RRDE results (Fig. 5D).
In the above experiments, the pulse time has been adjusted to the experimental conditions and the nature of the catalytic system just to obtain well defined linear (diffusional) portion permitting precise determination of the Q/t 1/2 slope. In general, to avoid too excessive formation of H 2 O 2 intermediate, fairly short pulses (here 0.5 s) shall be used. But the potential step cannot be too short to avoid kinetic limitations and related non-linearity of the Q to t 1/2 response. In other words, the experimental conditions have to be adjusted to meet the criterion of high correlation coefficient (possibly on the level 0.9999) for the linear portion of Anson (Q to t 1/2 ) plot.
Application of longer pulses (e.g. 5 s) allows us to get insight into formation of the hydrogen peroxide intermediate. Under such condi- tions, more excessive electrolysis (oxygen reduction within the diffusional flux) is possible during the forward step thus eventually leading to some formation of H 2 O 2 in the vicinity of the electrode surface. Furthermore, application of longer pulses (reduction) permits better control of background currents (it should be remembered that the responses originating from the hydrogen peroxide intermediate could be low). Immediate switching to the reverse oxidation step permits estimation of the percent of hydrogen peroxide formation. Here the maximum charges (Fig. 8) characteristic of both processes (2-electron oxidation of H 2 O 2 , and the approximately 4-electron reduction of O 2 ) should be considered. They are estimated upon assumption that diffusion coefficients of both oxygen and hydrogen peroxide are identical and by subtracting contributions originating from the material's surface processes (appearing in Fig. 8 at t 1/2 values lower than 0.2 s 1/2 ). It should also be remembered that different numbers of electrons are involved in both processes mentioned above. Consequently, we have found that, for the system of WO 3 -free rGO-supported Pt nanoparticles, approximately 6-7% of H 2 O 2 is produced. When a similar approach has been applied to the WO 3 -decorated rGO-supported Pt nanoparticles (for simplicity not shown here), the amounts of hydrogen peroxide have been estimated on the level 4-5%. The approach is approximate but fast and simple. Although comparison to the RRDE data (Fig. 5C) is not straightforward because of different experimental conditions, the results of chronocoulometric estimations are consistent with our calculations based on RRDE voltammetric observations.

Conclusions
This study clearly demonstrates that the chemically-reduced graphene-oxide, particularly when decorated with tungsten oxide nanowires, acts as a robust and activating support for dispersed Pt nanoparticles during electrocatalytic reduction of oxygen in acid medium (0.5 mol dm −3 H 2 SO 4 ). For the same low loading of catalytic Pt nanoparticles (30 μg cm −2 ), decoration of the electrocatalytic interface with WO 3 results in the formation of lower amounts of the undesirable H 2 O 2 intermediate. Moreover the onset potential for the oxygen reduction has been the most positive in a case of the system utilizing WO 3 -decorated reduced graphene oxide. Synergistic effects and activating interactions between catalytic metal nanoparticles, tungsten oxide and nanostructured graphene supports cannot be excluded here with respect to lowering the dissociation activation energy for molecular O 2 through accelerating the charge transfer from metal in the presence of graphene and by reducing stability of the H 2 O 2 intermediate species. By reviewing the RRDE parameters obtained here (such as current densities, reduction potentials, percent of the hydrogen peroxide formation) it can be stated that the electrocatalytic behavior of the WO 3 -modified low-Pt-loading system is comparable, if not better, relative to the performance of other "cheap" Pt-free catalysts. 53,54 We have also demonstrated the usefulness of the double-potentialstep chronocoulometry, particularly of the charge vs. square root of time (so called Anson) plots as the fast and reliable diagnostic tool for evaluation of the effectiveness of the oxygen reduction. This work parallels other recent attempts to use chronocoulometry in the oxygen reduction research. 51,52 In the present work, we have demonstrated that, by proper adjustment of pulse times and other experimental parameters within the double-potential-step experiment, mechanistic considerations permitting estimation of the number of electrons involved and the percent of formation of the H 2 O 2 intermediate are feasible.