Oxygen reduction on silver catalysts electrodeposited on various nanocarbon supports

In this work, Ag particles were electrodeposited onto nitrogen-doped graphene oxide, graphene, multi-walled carbon nanotube (MWCNT), and Vulcan carbon XC-72R supports by varying the upper potential limit. The surface morphology of the resulting Ag-based catalysts was examined by scanning electron microscopy. The electrochemical oxygen reduction reaction (ORR) was tested in alkaline media employing the rotating disk electrode method. The variation of the upper potential limit influenced the size of silver nanoparticles and their number density on the substrate surface. All the Ag-based electrocatalysts studied in this work showed remarkable ORR activity in terms of half-wave potentials. The ORR results combined with hydrogen peroxide reduction results prove that all Ag catalysts tested are suitable for both reactions. Ag/NGO2 catalyst possesses the highest mass activity for ORR, which indicates a relationship between the Ag loading and electrocatalytic activity. The electroreduction of oxygen on all the electrodeposited silver catalysts follows a four-electron pathway in alkaline environment. These materials are promising alternatives for Pt/C catalyst to be used as alkaline membrane fuel cell cathodes.


Introduction
As energy consumption grows, more sustainable green energy is needed. Renewable energy sources are increasingly employed and low-temperature fuel cells have been taken into serious consideration [1]. For example, anion exchange membrane fuel cells (AEMFCs) use hydrogen gas as a fuel and produce electricity, water vapor, and no CO 2 . The main obstacle in using the AEMFC technology is its high cost due to Pt catalyst materials, which are currently employed for oxygen reduction reaction (ORR) on the cathode side of the fuel cell [2][3][4][5][6]. Thus novel, cheaper, more stable, and highly active catalysts for replacement of Pt/C have been developed during the past decade.
A good substitute to platinum is silver, which has shown in alkaline media comparable ORR performance [4,[7][8][9][10][11]. The ORR occurs via 4-electron pathway on both electrode materials. Although compared to Pt, silver shows slightly lower activity, it still is an attractive material because of its significantly lower cost [12]. The ORR on silver-based catalyst materials proceeds either through a 2e − + 2e − reduction, where hydrogen peroxide intermediate forms or via direct four-electron pathway or by both pathways simultaneously [7]. In fuel cells a complete four-electron reduction of O 2 is preferred because the formation of peroxide decreases the fuel cell efficiency and can degrade the cathode catalyst material.
The ORR studies on low-index single-crystal Ag electrodes have shown that this is a structure-sensitive reaction [11,13]. When examining the ORR kinetics on single-crystal silver surfaces in alkaline solution Blizanac et al. determined that the ORR activation energy depends on the structure and decrease in the order of Ag(100) > Ag(111) > Ag(110) [13]. They also stated that serial 4e − pathway was the most likely pathway via which the ORR on Ag(hkl) surfaces would proceed. Structuresensitivity has been also studied on differently shaped Ag nanocatalysts, for example, nanowires, nanocubes, nanoplatelets, nanodecahedra, and nanosheets [14][15][16][17]. Lee et al. showed that due to the decreased amount of Ag(111) planes on Ag nanocubes the specific activity (SA) for O 2 reduction increased [14]. It has been also demonstrated that on silver nanocubes with (100) facets OH species adsorb stronger than on silver nanodecahedra with (111) facets and therefore on Ag decahedra the ORR proceeds likely via direct 4-electron pathway and on Ag nanocubes via series pathway [15].
Silver nanoparticle (AgNP) size effect plays an important role in Ag electrocatalyst activity towards the ORR and therefore it has received a lot of attention [17][18][19][20][21][22][23]. The topic has proved itself to be quite intriguing since size effects have been observed even for larger particles. In a study by Lee et al., it was demonstrated that on silver nanoparticles sized 136.1 nm the electron transfer number (n) was 3.16, while on 98.7 nm AgNPs the n value was 3.27 [17]. In another study, it was concluded that on smaller AgNPs (4.1 nm) the ORR proceeds via both 2e − + 2e − and four-electron pathways at the same time and on larger ones (174 nm) via 4-electron pathway [18]. In a study conducted by Alia et al., it was reported that decreasing the silver nanowire diameter increases the specific activity for oxygen reduction [19]. Ag nanoparticles showed the opposite results, where the SA value grew with increasing the particle size, and with the increase of the diameters of both silver nanowires and nanoparticles, mass activities (MA) decreased in both cases. Lu et al. studied the ORR on 0.7 nm Ag nanoclusters and 3.3 nm silver nanoparticles, and the results showed that SA was 5 times higher on the nanoclusters as compared to nanoparticles [20]. In a study conducted by Ohyama et al., it was concluded that for larger particles than 10 nm the size effect had an insignificant impact on SA, however the size effect with particles smaller than 3 nm noticeably affected the SA values because of the quantum size effects and a large amount of coordinatively unsaturated silver atoms [21].
For better efficiency of the catalyst materials, the studies show that silver loading for a complete 4e − ORR process should be higher than 10 wt% [24][25][26][27]. Guo et al. suggested that the reason for the lower than 4 electrons transferred per O 2 molecule during the ORR process on 10 wt% Ag/C is that the reaction proceeds partially on a carbon support due to low amount of silver [27]. The n value is affected by the metal loading as well as by distinct reaction mechanism and the mean crystallite size. The Ag loading of 40 wt% compared to 10 and 20 wt% Ag on multi-wall carbon nanotubes (MWCNT) gave the highest ORR current densities [28]. Larger loading also provides better stability to the electrocatalyst. In a study where the Ag loading was increased from 20 to 60 wt%, the decrease in the n value from 4 to 3.7 was reported [29]. Taking previous work into account, it is clear that in order to prepare electrocatalytically active Ag/C catalysts for ORR in alkaline media the loading of Ag should be at least 20 wt% [30].
Catalyst morphology is strongly affected by substrate materials and as a result many carbon and also non-carbon supports have been tested. Several high-surface-area carbon-based support materials have been used in the preparation of Ag-based catalysts [31]. Carbon nanotubes are one of the most frequently used support materials [32,33]. AgNPs have been deposited onto MWCNTs using sputterdeposition [34]. In addition to MWCNTs, both graphene [35] and nitrogen-doped graphene (NGO) have received a lot of attention as substrates for silver catalysts [36][37][38]. Other N-doped carbon supports have also been employed [39] and Vulcan carbon XC-72R has been used as a reference carbon support [30]. Mesoporosity of the support material plays a role in the enhanced electrocatalytic activity of Ag catalysts for ORR [40].
In previous studies, we have electrochemically deposited silver onto a bare glassy carbon electrode both in aqueous and non-aqueous solutions [41,42]. In this work, we prepared Ag catalysts via electrodeposition in aqueous media using a simple potential cycling procedure, where the upper potential limit was varied to better control the deposition of Ag particles in both size and uniform distribution on carbon supports. In order to study the effect of support materials on the surface morphology of the silver catalyst, Ag particles were deposited onto four nanocarbon supports (NGO, graphene, MWCNT, and Vulcan XC-72R). The electroreduction of oxygen in alkaline media on these Ag-based catalysts was tested using a rotating disk electrode.

Synthesis of GO and NGO
Graphene oxide (GO) was prepared using a modified Hummers method [43] and the GO material was doped with nitrogen to synthesize N-doped GO (NGO), which was used as a substrate [37,43]. Briefly, a known amount of graphite was added into a glass flask, to which concentrated sulfuric acid (H 2 SO 4 ) was added and the mixture was sonicated for 1 h. Then the temperature of the mixture was raised to 35 °C and held for 18 h. In the following step, the mixture was cooled down in an ice bath and deionized water was added to it. In a consecutive step, hydrogen peroxide was added and after that, the suspension was washed with hydrochloric acid. The resulting mixture was washed with copious amount of Milli-Q water (Millipore, Inc.) and dried at 75 °C. After that a known amount of GO powder was weighed into a beaker to which Milli-Q water and ethanol were added and the mixture was sonicated. While sonicated polyvinylpyrrolidone (PVP) and dicyandiamide (DCDA, Aldrich) were added, as described previously [43]. The resulting mixture was then sonicated for 2 h and dried at 75 °C. Dry powder was collected and pyrolyzed at 800 °C for 2 h under N 2 flow. After that the furnace was cooled down and the resulting N-doped graphene oxide material was gathered into a vial.

Modification of the electrodes
Before the electrodeposition, 20 μl of suspensions (1 mg of carbon material in 1 ml ethanol) of NGO, graphene (Gr, Strem Chemicals), Vulcan carbon XC-72R (C, Cabot Corp.), and multi-walled carbon nanotubes (MWCNT, diameter 30 ± 15 nm, length 5-20 μm, purity > 95%, NanoLab Inc., USA) were drop-casted onto glassy carbon (GC) electrodes. MWCNTs were acid-treated before use according to a procedure reported previously [44]. Electrodeposition of Ag was carried out in aqueous 0.1 M KNO 3 solution containing 0.01 mM AgNO 3 by cycling between the deposition potential of -0.6 V and different upper potential (0.3 and 0.4 V) for 100 cycles with scan rate of 100 mV s −1 . The resulting materials were designated according to the upper potential value. For example, for NGO-supported Ag catalysts when it was 0.3 V the catalyst was named Ag/NGO1 and when the upper potential was 0.4 V then Ag/NGO2. Other Ag-based electrocatalysts were named in the same way.

Electrochemical measurements
The electrochemical measurements were carried out in a standard three-electrode glass cell. A Pt wire was used as a counter electrode, which was separated from the working solution by a glass frit. Saturated calomel electrode (SCE) was used as a reference electrode separated from the working solution by Luggin capillary. Potentials were applied using Autolab PGSTAT30 potentiostat/galvanostat (Metrohm Autolab B.V., The Netherlands), which was controlled by General Purpose Electrochemical System (GPES) software. To carry out the rotating disk electrode (RDE) measurements EDI101 rotator and CTV101 speed control unit (Radiometer) were used and electrode rotation rate (ω) was varied from 360 to 4600 rpm. A potential scan rate (v) of 10 mV s −1 was used for oxygen reduction and peroxide reduction measurements. All electrochemical experiments were conducted at room temperature (23 ± 1 °C).
Most of the electrochemical experiments were carried out in 0.1 M KOH solution, which was made up by dissolving KOH pellets (puriss p.a., Sigma-Aldrich) in Milli-Q water. To conduct the oxygen reduction measurements the electrolyte solution was saturated with O 2 (99.999%, Linde Gas). The solution was deaerated with Ar (99.999% Linde Gas) for some other experiments, including cyclic voltammetry (CV) and Ag stripping experiments (see Supplementary material). For peroxide reduction experiments, hydrogen peroxide (p.a. 30%, Lach-Ner) was added to 0.1 M KOH solution (final concentration 1 mM HO 2 − ). Glassy carbon disk electrodes with 5 mm diameter were cut from GC rods (GC-20SS, Tokai Carbon, Japan). The GC disks were mounted into Teflon holders and polished on a 4000 grit emery paper and thereafter with alumina slurries (1 and 0.3 μm, Buehler). After polishing the electrodes were sonicated in both Milli-Q water and 2-propanol in an ultrasonic bath for 3 min. Bulk polycrystalline Ag (99.95%, Alfa Aesar) electrodes (geometric area 0.2 cm 2 ) were used for comparison.

Surface characterization of the Ag catalysts
For the surface morphology characterization, a scanning electron microscope (SEM, Helios™ Nanolab 600, FEI) with an accelerating voltage of 10 kV and a working distance of 4 mm was used. Samples for SEM were prepared in the same way as the electrodes for electrochemical experiments and finally removing the modified GC disks from Teflon holders. On all of these SEM images quasi-spherical Ag particles can be observed. The average particle size of the Ag-based catalyst materials varies depending on the deposition procedure used and the size distribution is rather large. The Ag particles are randomly distributed on the carbon support. In our previous work where the silver nanoparticles were electrodeposited onto bare GC disks using different deposition times yielded smaller nanoparticles [41,42]. During the electrodeposition the upper potential limit was varied and by using a more positive potential smaller silver nanoparticles were deposited and their number density on the substrate surface was also lower. In addition to the deposition environment and controlling the deposition potential and the time (or the number of cycles and the scan rate), the carbon support plays a great role in determining the Ag nanoparticle size and also the distribution of the nanoparticles [37]. From Fig. 1 it can be seen that on the MWCNT surface the silver nanoparticles are the smallest and the distribution of particles is the widest. In this case the upper potential limit of 0.3 V vs SCE was applied.

Oxygen reduction reaction studies
Before the ORR studies the electrodes were pre-conditioned by cycling in a potential range from − 1.2 to 0 V versus SCE in Ar-saturated 0.1 M KOH solution for 10 potential cycles with scan rate of 100 mV s −1 . In this potential range, the silver catalyst surface is not yet oxidized and the rearrangement of silver nanoparticles on all of the carbon supports is avoided. The background currents were measured and subtracted from the corresponding oxygen reduction currents. The ORR measurements were performed in O 2 -saturated 0.1 M KOH solution using the RDE method. The ORR polarization curves on all silver catalysts were recorded only by anodic scans. It can be seen in Figs. 3a and 4a that all Ag-based catalysts show only single-wave polarization curves with more positive ORR onset and half-wave potentials (E 1/2 ) than bulk polycrystalline silver electrode (see also Figure S1 in Supplementary material). From all the Ag-based catalyst materials studied Ag/NGO1 possessed the highest ORR electrocatalytic activity (see Table 1). It is evident that a more positive upper potential limit leads to lower E 1/2 values. In a control experiment it was found that by selecting the upper potential limit of 0.5 V the Ag nanoparticles start to dissolve and therefore less amount of silver is present on the carbon supports. The dissolution of Ag occurs when the upper potential limit is close to the silver redox potential (Ag + + e − ↔ Ag 0 ) of 0.7996 V vs. standard hydrogen electrode (SHE) [45]. For further analysis of RDE results of the synthesized Ag-based electrocatalysts, the Tafel plots of ORR were constructed (Figs. 3b and 4b). It can be seen that Tafel plots of polycrystalline silver, Ag/NGO1, and Ag/MWCNT1 electrodes are almost parallel. The Tafel slope values at low overpotentials are − 92, − 60, − 95, and − 93 mV for Ag/Gr1,  [34,37,41,42,[46][47][48][49][50]. Blizanac et al. reported Tafel plots for ORR on Ag(hkl) where a change in slope values between − 60 and − 120 mV was observed. At lower current densities the Tafel slope was close to − 80 mV and at higher current densities around − 120 mV [13]. Also, in a work where the ORR was investigated on Ag nanocubes, the Tafel slope values were − 60 and − 120 mV at low and high overpotentials, respectively. This change in slope was explained to come from the changes in the adsorption properties of the oxygen intermediates [51]. Changes in the Tafel slope values can also originate from the differences in the Ag loadings on the electrodes, where small Ag loading gives changing slope values, and when critical Ag loading is reached the slope becomes linear [39]. In addition Tafel slopes might be affected by incorrect solution resistance correction [52]. As the Tafel slopes determined in this work are close to the values reported in the literature, it can be concluded that the rate-determining step for the ORR is the slow transfer of the first electron to the adsorbed O 2 molecule [34,37,41,42,[46][47][48][49][50].
In Fig. 5a a typical set of ORR polarization curves for Ag/NGO1 catalyst recorded at different rotation rates is presented. The RDE data was analyzed using the Koutecky-Levich (K-L) equation [53]: where j is the ORR current density measured at a specific potential, j k and j d are kinetic and diffusion-limited current densities, respectively, n is the electron transfer number during the ORR, F is the Faraday constant (96,485 C mol −1 ), k is the heterogeneous rate constant for O 2 reduction at a specific potential (cm s −1 ), C b O 2 is the concentration of O 2 (1.2 × 10 -6 mol cm −3 ) [54], Do 2 is the diffusion coefficient of O 2 (1.9 × 10 -5 cm 2 s −1 ) [54], v is the kinematic viscosity of the solution (0.01 cm 2 s −1 ) [55], and ω is the electrode rotation rate (rad s −1 ).
The electron transfer number n was determined from the K-L plots of Ag/NGO1 (Fig. 5b). The value of n was approximately four for all Ag-based catalysts in this study. From this result, it can be concluded that the ORR proceeds via a 4-electron transfer pathway, thus hydrogen peroxide formation is either almost completely avoided, or the formed hydrogen peroxide is immediately further electrochemically reduced. However, the K-L analysis does not allow differentiating between the direct 4-electron and 2e − + 2e − pathways, because it yields an overall n value. Guo et al. and Fazil et al. have reported the n value close to four for 40 wt% Ag/CNT catalyst for oxygen reduction in an alkaline environment [27,28]. For Ag nanowires [19], through plasma jet treatment prepared silver nanoparticles [46] and Ag nanocubes [14] the n values close to 4 have been reported as well. Silver catalysts deposited onto carbon paper have shown similar n values [56]. In other works where AgNPs were electrodeposited onto bare GC substrate, and the deposition time was varied, the values of n grew with deposition time because of the higher mass of silver on the electrode surface [41,42]. Ag catalysts prepared through heteropolytungstate-assisted fabrication and (1) deposition of Ag NPs on reduced graphene oxide (rGO) substrates gave n values close to four as well [50].
To compare the intrinsic electrocatalytic activity of Ag-based catalysts, mass activities (MA) for ORR were calculated using the following equation: where I k is the kinetic current determined at − 0.2 V vs SCE and m is the mass of Ag on the electrode surface calculated from the Ag dissolution experiments. From Table 1 it can be seen that the highest MA values were obtained for Ag/NGO catalysts. Ag/C1 possessed the lowest MA value (0.009 mA μg −1 ), which was closely followed by Ag/C2 and Ag/MWCNT1 catalyst materials (for both MA = 0.010 mA μg −1 ). This is apparently related to large Ag particle sizes of these electrocatalysts, and in Figs. 1 and 2 the Ag/NGO catalysts had overall the smallest particles with most uniform distribution. The higher mass activity of Ag/NGO catalysts may also be related to synergistic effect between NGO and Ag-particles, and some nitrogen groups make it easier for oxygen molecules to adsorb thus making the reduction of oxygen more efficient [57]. Similar MA values have been reported previously [21,30]. In both cases the mass of Ag on the electrode surface was calculated from the Ag dissolution experiments (see Supplementary material).
(2) MA = I k ∕m Durability of Ag-based catalysts is of utmost importance for their application as cathodes in AEMFCs. Figure 6 shows a comparison of the ORR polarization curves recorded during the stability test. From Fig. 3 it can be seen that Ag/NGO1 has the most positive half-wave and onset potential for ORR, thus the stability test was conducted with the Ag/NGO1 catalyst. For comparison a stability test was also conducted with Ag/C1 catalyst. For stability experiments, a 1000 potential cycles between − 1.3 and 0 V vs. SCE at a scan rate of 10 mV s −1 were carried out in O 2 -saturated 0.1 M KOH solution. The E 1/2 value for Ag/ NGO1 shifted negative only by 20 mV and for Ag/C1 only 22 mV and the diffusion-limited current density for both catalysts stayed almost unchanged, which allows us to conclude high stability of these Ag-based catalysts.

Hydrogen peroxide reduction studies
As Ag-based electrocatalysts can also be used as hydrogen peroxide detectors [58], the hydrogen peroxide reduction measurements were carried out as well. In Fig. 7 a typical set of RDE polarization curves for peroxide reduction is presented. The RDE polarization data were further analyzed using the K-L equation, where the diffusion coefficient for HO 2 − anions is 1.36 × 10 -5 cm 2 s −1 [59]. The resulting n value is two for all measured Ag/NGO, Ag/C, Ag/Gr, and Ag/MWCNT catalysts. This shows that the electroreduction of hydrogen peroxide is efficient on all the Ag catalysts studied. It can be thus reasoned that the Agcatalyst materials prepared in this work catalyze the ORR through a 4-electron pathway, however it does not allow one to conclude that the ORR proceeds only via direct 4e − pathway. The assumption that the electroreduction of oxygen on the studied Ag-based catalysts proceeds simultaneously via both 4-electron and 2 + 2 electron pathways is more realistic. A similar conclusion has also been made in a work where silver nanoparticles were electrodeposited onto a bare GC substrate [42].

Conclusion
The Ag-based electrocatalysts were prepared by electrodeposition of silver onto four different carbon substrates by cycling between the deposition potential of − 0.6 V versus SCE and two different upper potential limits (0.3 and 0.4 V). The study of the oxygen reduction activity of these Ag catalysts was carried out in 0.1 M KOH solution. The amount of deposited silver and the size of Ag particles were the biggest when the upper potential was lower (0.3 V). The RDE results indicate the influence of the upper potential limit on the ORR activity in terms of the onset and half-wave potentials. The O 2 reduction results as well as hydrogen peroxide reduction results show that all Agbased catalysts studied in this work are efficient electrocatalysts for both reactions in alkaline media. It was found that Ag/NGO2 possessed the biggest mass activity, thus being the most ORR-active catalyst. This indicates about the effect of catalyst loading and carbon substrate on the electrocatalytic activity for ORR. The ORR proceeds via 4-electron pathway on all Ag-based catalysts and Tafel analysis reveals that the rate-determining step is the slow transfer of the first electron to the O 2 molecule.