On the Effect of the Nature of Carbon Nanostructures on the Activity of Bifunctional Catalysts Based on Manganese Oxide Nanowires

: Manganese oxide nanowires (MONW) combined with carbon nanostructures were synthesized using three different carbon materials, and their effect on the activity towards Oxygen Reduction Reaction (ORR) and Oxygen Evolution Reaction (OER) was investigated in alkaline electrolytes. The carbon structures were carbon nanoﬁbers (CNF), multiwall carbon nanotubes (CNT) and reduced graphene oxide (rGO). Both MONW and carbon nanostructures were characterized by X-ray diffraction, scanning and transmission electron microscopy, N 2 physisorption and X-ray photoelectron spectroscopy. The electrochemical activity was assessed in a three-electrode cell. Composite MONW/CNF showed the best activity towards ORR, and MONW/rGO exhibited the highest activity towards OER of the series. The addition of the carbon nanostructures to MONW increased the number of electrons transferred in the ORR, indicating a synergistic effect between the carbon and manganese oxide structures due to changes in the reaction pathway. The analysis of Tafel slopes and electrochemical impedance spectroscopies showed that carbons and MONW catalyze different steps of the reactions, which explains the better activity of the composites. This led us to synthesize a MONW/rGO-CNF composite, where rGO-CNF is a hybrid carbon material. Composite MONW/rGO-CNF showed an improved activity towards ORR, close to the benchmark Pt/C catalyst, and activity towards OER, close to MONW/rGO, and better than the benchmark IrO 2 catalyst. It also showed remarkable stability in challenging operation conditions.


Introduction
Bifunctional air electrodes have gained increased interest over the last years due to their relevance to sustainable energy devices, including regenerative fuel cells and metal-air batteries [1,2]. These bifunctional air electrodes are responsible for the oxygen evolution reaction (OER) and oxygen reduction reaction (ORR), which are known for their sluggish kinetic reactions due to the complex 4e − mechanisms governing them [3,4]. Noble metals such as Pt/C and IrO 2 or RuO 2 have been widely studied and are considered the most active catalysts for ORR and OER, respectively. However, these catalysts are not bifunctional: Pt shows very poor activity towards OER, and neither IrO 2 nor RuO 2 are active for the ORR [5,6]. Amongst noble metals, Pd displays high activity towards both reactions [7][8][9], but its cost is even higher than that of Pt, a reason that encourages the search for another kind of catalyst.
In the last decade, great efforts have been devoted to the search for bifunctional catalysts based on non-noble transition metals. Cobalt-based catalysts, either in the form of spinels, perovskites or mixed oxides, are reported as the most active bifunctional catalysts for ORR and OER [10][11][12][13]. However, due to the increased use of lithium-ion batteries, cobalt around 50 nm, with a fishbone morphology (Figure 1d). CNTs (Figures 1b,e) have wider diameters, ca. 60-70 nm, and are shorter than CNFs but still have lengths longer than 1 µm. CNTs are multiwall type, as can be seen in Figure 1e. In addition, graphitic carbon particles around the iron catalyst can also be observed. rGO shows the typical sheet-like morphology, as shown in Figures 1c,f, where single sheets can be observed, presenting wrinkles, which are a product of the defects that are produced during the graphene oxide thermal reduction [27,28]. Figure S1 (Supporting Information, Section S1) shows SEM and TEM micrographs of MONW. The manganese oxide nanowires have square cross-sections and widths ranging between 60 and 100 nm. Their length can reach up to several microns. TEM and SEM micrographs of the composites (Figures 1 g,h, i) show the good mixing between the phases, with several contact points between the carbon materials and MONW.  (JCPDS 75-2078) at 2θ = 26.5 • and 26.6 • , respectively, attributed to the (002) reflection. These values indicate interplanar distances of 0.337 and 0.335 nm and graphitization degrees of 80% and 100%, respectively. In the case of rGO, the interplanar distance is much higher, 0.347 nm (position of the peak at 25.65 • ), mainly due to the presence of oxygen functional groups, as can be interpreted from the 5 wt% of oxygen content in this sample (Table 1). Still, this value is far from the typical 0.8-0.9 nm layer separation distance of graphene oxide [29][30][31], and no peak at 2θ = 10 • was observed. The position of the peak agrees with others reported in the literature [32,33]. Other peaks ascribed to graphitic carbon are (100) in the three carbon materials and (004) in CNF and CNT [30]. The (004) peak is not visible in rGO due to its low number of layers. Carbons CNF and CNT also show reflection planes characteristic of the catalyst metals. In CNF, planes (111) and (200) at 44.5 • and 51.8 • , respectively, corresponding to metallic nickel (JCPDS 04-850) are also observed [34], while CNT shows peaks ascribed to (110) and (200) metallic iron (JCPDS 06-0696) [35], but also to martensite, an iron-carbon alloy (JCPDS 04-003-1419) [36].
Catalysts 2023, 13, x FOR PEER REVIEW 4 of 23 XRD analyses (Figure 2a) revealed the crystalline structure of the carbon materials. CNF and CNT show a highly graphitic structure, with the characteristic peak of graphite (JCPDS 75-2078) at 2θ = 26.5° and 26.6°, respectively, attributed to the (002) reflection. These values indicate interplanar distances of 0.337 and 0.335 nm and graphitization degrees of 80% and 100%, respectively. In the case of rGO, the interplanar distance is much higher, 0.347 nm (position of the peak at 25.65°), mainly due to the presence of oxygen functional groups, as can be interpreted from the 5 wt% of oxygen content in this sample (Table 1). Still, this value is far from the typical 0.8-0.9 nm layer separation distance of graphene oxide [29][30][31], and no peak at 2θ = 10° was observed. The position of the peak agrees with others reported in the literature [32,33]. Other peaks ascribed to graphitic carbon are (100) in the three carbon materials and (004) in CNF and CNT [30]. The (004) peak is not visible in rGO due to its low number of layers. Carbons CNF and CNT also show reflection planes characteristic of the catalyst metals. In CNF, planes (111) and (200) at 44.5° and 51.8°, respectively, corresponding to metallic nickel (JCPDS 04-850) are also observed [34], while CNT shows peaks ascribed to (110) and (200) metallic iron (JCPDS 06-0696) [35], but also to martensite, an iron-carbon alloy (JCPDS 04-003-1419) [36]. The number of layers comprising carbon nanotubes and reduced graphene oxide was determined by applying Scherrer's equation to the (002) peak. The crystallite size (in the 002 direction) of CNT is approximately 34 nm, which allows us to estimate that carbon nanotubes have 100 layers (or walls), in agreement with TEM micrographs, as the crystallite size calculated by XRD approximately corresponds to the radius of multiwall carbon nanotubes. rGO presents an average crystallite size of 15 nm, corresponding to 4.5 layers, also in good agreement with the TEM micrographs. XRD analysis of MONW is presented in Figure S2 (Section S1). The only present phase is α-MnO2, and the crystallite size obtained by applying Scherrer's equation to the main peak (211 planes at 37°) is 37 nm. The reflection planes were obtained from JCPDS (α-MnO2, 00−044−0141) and matched other works found in the literature [37]. MONW were also analyzed using X-ray photoelectron spectroscopy ( Figure S3). Mn2p spectra show two 2p3/2 peaks, at 642.0 eV, ascribed to Mn 4+ in MnO2 (≈88%) [38], and at 644.3 eV, ascribed to Mn 6+ in K2MnO4 (≈12%) [39]. The O1s orbital can be decomposed into two peaks: one at 539.5 eV, corresponding to lattice oxygen (≈65%), and another one at 531.1 eV, ascribed to adsorbed oxygen or water (≈35%) [40].
Raman spectroscopy (Figure 2b) was used to investigate the structural defects of the carbon nanostructures. All of them show the characteristic G band: CNF at 1570 cm −1 , CNT at 1585 cm −1 and rGO at 1593 cm −1 . The D band, associated with defects, mainly sp3 carbon atoms, is also visible in the three samples, being especially intense in CNF and rGO, with ID/IG ratios of 1.1 and 1.4, respectively, while CNT has a much smaller amount of defects  The number of layers comprising carbon nanotubes and reduced graphene oxide was determined by applying Scherrer's equation to the (002) peak. The crystallite size (in the 002 direction) of CNT is approximately 34 nm, which allows us to estimate that carbon nanotubes have 100 layers (or walls), in agreement with TEM micrographs, as the crystallite size calculated by XRD approximately corresponds to the radius of multiwall carbon nanotubes. rGO presents an average crystallite size of 15 nm, corresponding to 4.5 layers, also in good agreement with the TEM micrographs. XRD analysis of MONW is presented in Figure S2 (Section S1). The only present phase is α-MnO 2 , and the crystallite size obtained by applying Scherrer's equation to the main peak (211 planes at 37 • ) is 37 nm. The reflection planes were obtained from JCPDS (α-MnO 2 , 00−044−0141) and matched other works found in the literature [37]. MONW were also analyzed using X-ray photoelectron spectroscopy ( Figure S3). Mn2p spectra show two 2p 3/2 peaks, at 642.0 eV, ascribed to Mn 4+ in MnO 2 (≈88%) [38], and at 644.3 eV, ascribed to Mn 6+ in K 2 MnO 4 (≈12%) [39]. The O1s orbital can be decomposed into two peaks: one at 539.5 eV, corresponding to lattice oxygen (≈65%), and another one at 531.1 eV, ascribed to adsorbed oxygen or water (≈35%) [40].
Raman spectroscopy (Figure 2b) was used to investigate the structural defects of the carbon nanostructures. All of them show the characteristic G band: CNF at 1570 cm −1 , CNT at 1585 cm −1 and rGO at 1593 cm −1 . The D band, associated with defects, mainly sp3 carbon atoms, is also visible in the three samples, being especially intense in CNF and rGO, with I D /I G ratios of 1.1 and 1.4, respectively, while CNT has a much smaller amount of defects as can be inferred from its I D /I G ratio = 0.4. CNF and CNT also exhibit the 2G band at 2681 and 2700 cm −1 , respectively, whereas this band is not visible in rGO due to its low amount of layers. The intensity of the 2G band is especially high in CNT because of the high number of layers.
Elemental analysis and ICP were used to investigate the composition of the carbon materials. CNFsF are comprised of 92 wt% carbon and 7 wt% nickel, which is consistent with the catalytic decomposition of methane. An amount of 300 mg of NiCuAl 2 O 3 catalyst yielded approximately 3 g of CNF, indicating that the amount of catalyst should be less than 10 wt% in the final material. CNT, on the other hand, presents a higher amount of metal, around a quarter of the total weight of the sample. Indeed, 300 mg of iron oxide yielded ca. 1.2 g of CNT. rGO has a C:O mass ratio of 17 (atomic ratio of 22), indicating a good-but not complete-reduction of the graphene oxide.
The textural properties of the different carbon nanostructures were studied using nitrogen physisorption ( Figure 3, Table 1). All isotherms ( Figure 3a) show a hysteresis cycle-type IIb or IV according to the IUPAC classification-characteristic of small particles or crystal aggregates with mesopore-size voids between them [41]. The surface area was calculated with the Brunauer-Emmett-Teller (BET) model (Table 1), being 175 m 2 g −1 , 95 m 2 g −1 and 11 m 2 g −1 for rGO, CNF and CNT, respectively. The low surface area of CNT is explained by the graphitic carbon particles formed together with the nanotubes, the high number of walls of the tubes and the large amount of iron in this sample. The textural characteristics of MONW are displayed in Figure S4 (Section S1, in the Supplementary Materials). MONW also present a type IV adsorption isotherm, indicative of mesoporosity, with a surface area of 23 m 2 g −1 and a pore size distribution, including mesoand macropores. as can be inferred from its ID/IG ratio = 0.4. CNF and CNT also exhibit the 2G band at 2681 and 2700 cm −1 , respectively, whereas this band is not visible in rGO due to its low amount of layers. The intensity of the 2G band is especially high in CNT because of the high number of layers. Elemental analysis and ICP were used to investigate the composition of the carbon materials. CNFsF are comprised of 92 wt% carbon and 7 wt% nickel, which is consistent with the catalytic decomposition of methane. An amount of 300 mg of NiCuAl2O3 catalyst yielded approximately 3 g of CNF, indicating that the amount of catalyst should be less than 10 wt% in the final material. CNT, on the other hand, presents a higher amount of metal, around a quarter of the total weight of the sample. Indeed, 300 mg of iron oxide yielded ca. 1.2 g of CNT. rGO has a C:O mass ratio of 17 (atomic ratio of 22), indicating a good-but not complete-reduction of the graphene oxide. The textural properties of the different carbon nanostructures were studied using nitrogen physisorption ( Figure 3, Table 1). All isotherms ( Figure 3a) show a hysteresis cycle-type IIb or IV according to the IUPAC classification-characteristic of small particles or crystal aggregates with mesopore-size voids between them [41]. The surface area was calculated with the Brunauer-Emmett-Teller (BET) model (Table 1), being 175 m 2 g −1 , 95 m 2 g −1 and 11 m 2 g −1 for rGO, CNF and CNT, respectively. The low surface area of CNT is explained by the graphitic carbon particles formed together with the nanotubes, the high number of walls of the tubes and the large amount of iron in this sample. The textural characteristics of MONW are displayed in Figure S4 (Section S1, in the Supplementary Materials). MONW also present a type IV adsorption isotherm, indicative of mesoporosity, with a surface area of 23 m 2 g −1 and a pore size distribution, including meso-and macropores. Pore size distributions were obtained by applying the BJH method to the desorption branch of the isotherms, revealing that carbon nanostructures present pores in the 4-5 nm Pore size distributions were obtained by applying the BJH method to the desorption branch of the isotherms, revealing that carbon nanostructures present pores in the 4-5 nm range. Consistently with the pore size distribution and surface areas, rGO has the highest pore volume of all samples, 0.70 cm 3 g −1 , followed by 0.24 cm 3 g −1 of CNF and 0.04 cm 3 g −1 Catalysts 2023, 13, 1240 6 of 21 of CNT. CNF and rGO also display a small fraction of microporosity, with micropore areas of 17 and 29 m 2 g −1 , respectively, and micropore volumes of 0.009 and 0.017 cm 3 g −1 .

Effect of the Carbon Nanostructure on the Catalytic Activity towards the Oxygen Reduction and Evolution Reactions
Linear sweep voltammetries ( Figure 4) performed in the ranges 1.2-0.2 V for ORR and 1.1-1.8 V vs. RHE for OER were carried out to investigate the catalytic activity of MONW/C composites, alongside evaluating their reversibility towards ORR/OER. Table 2 collects the main electrochemical parameters derived from the LSV curves, such as the onset potential for both ORR and OER (E ORR, onset being the potential at −0.1 mA cm −2 current density and E OER, onset the potential at 1 mA cm −2 current density), half-wave potential for the ORR (E 1/2 ), the limiting current density (j d ), the number of exchanged electrons for the ORR (n), the potential for the OER at 10 mA cm −2 (E 10 ) and the potential gap between ORR and OER.  Upon combination of MONW with the different carbon nanostructures (Figure 4b), there is a clear improvement in the catalytic activity of the individual manganese oxide nanowires, as can be ascertained from both the parameters displayed in Table 2 and the curves of Figure S6 in the Supporting Information (Section S2.1), comparing each MONW/C composite with its individual components. This highlights the importance of adding carbon for an enhanced activity for the oxygen evolution reaction. On the other hand, for the ORR, there is only a synergistic effect between Mn and C in the case of using CNF. Neither the combination of MONW with rGO nor with CNT provides a more active catalyst, which could be ascribed to a worsened interaction between these carbonaceous materials with the metal oxide active phase. As previously explained, CNT possesses a highly graphitic character and a low surface area, which could hinder the carbon-metal interaction. On the other hand, in the case of MONW/rGO, its low performance for the ORR might be explained by the large difference in surface area between carbon and manganese oxide phases. As the surface area of reduced graphene oxide is approximately seven times greater than that of the manganese oxide nanowires, most of the exposed area of the catalyst is rGO and not MONW. The poor mixing between the two materials can be seen in the SEM micrograph displayed in (Figure 1i).
In general, the electrochemical parameters shown in Table 2 reveal a positive shift of the onset potential for the ORR between 70 and 150 mV and between 30 and 90 mV for the half-wave potential for the MONW/C composites with respect to the individual carbon nanostructures.
Comparing the activity of the composite catalysts (MONW/C: CNF, CNT and rGO in Figure 4b), both MONW/CNF and MONW/rGO show an excellent OER activity, and MONW/CNF also exhibits an outstanding ORR performance. The MONW/CNT composite is the least active one for both ORR/OER, probably due to a worsened interaction between Mn and C due to the lower surface area of CNTs, the low amount of defects, and   material. In a previous work [24], we showed that acid-leached carbon nanofibers also hold a considerable, yet slightly lower, catalytic activity towards ORR and non-negligible activity towards OER. As was discussed in Section 2.1, CNF, although still presenting a small amount of nickel (<2 wt%), has a large amount of structural defects, which have been reported as active sites [42]. This is consistent with what has been reported in the literature regarding acid-leached carbon nanofibers showing activity for ORR and OER [43,44]. CNT also shows a decent OER activity due to the presence of Fe; however, its performance towards the ORR is the lowest of the series, which could be due to the low amount of defects (acting as active sites for ORR) since this sample is the most graphitic one. Acid-leached CNTs were tested to assess the activity of the carbon nanostructure ( Figure S5). It can be observed that the activity towards ORR barely reduces, while the activity towards OER shows a more significant decrease in activity, with the onset overpotential increasing 66 mV. The mild effect of leaching on the activity of CNT can be explained by the encapsulation of Fe particles in a graphitic matrix under a high number of layers (as explained by XRD and seen in TEM), which does not allow them to act as catalytic active sites. This is why other iron-over-carbon catalysts in the literature show better activities towards OER [45] or ORR [46]. rGO shows the lowest ORR/OER performance on its own. On the other hand, the catalytic activity of MONW without any carbon material is good for the ORR but not for the OER, highlighting the need for a carbon nanostructure to maximize the catalytic activity of Mn oxide sites by enhancing the electrical conductivity. The onset potentials for the ORR follow the trend CNT < rGO < CNF < MONW, and for the OER, the trend is rGO > CNT > MONW > CNF.
Upon combination of MONW with the different carbon nanostructures (Figure 4b), there is a clear improvement in the catalytic activity of the individual manganese oxide nanowires, as can be ascertained from both the parameters displayed in Table 2 and the curves of Figure S6 in the Supporting Information (Section S2.1), comparing each MONW/C composite with its individual components. This highlights the importance of adding carbon for an enhanced activity for the oxygen evolution reaction. On the other hand, for the ORR, there is only a synergistic effect between Mn and C in the case of using CNF. Neither the combination of MONW with rGO nor with CNT provides a more active catalyst, which could be ascribed to a worsened interaction between these carbonaceous materials with the metal oxide active phase. As previously explained, CNT possesses a highly graphitic character and a low surface area, which could hinder the carbon-metal interaction. On the other hand, in the case of MONW/rGO, its low performance for the ORR might be explained by the large difference in surface area between carbon and manganese oxide phases. As the surface area of reduced graphene oxide is approximately seven times greater than that of the manganese oxide nanowires, most of the exposed area of the catalyst is rGO and not MONW. The poor mixing between the two materials can be seen in the SEM micrograph displayed in (Figure 1i).
In general, the electrochemical parameters shown in Table 2 reveal a positive shift of the onset potential for the ORR between 70 and 150 mV and between 30 and 90 mV for the half-wave potential for the MONW/C composites with respect to the individual carbon nanostructures.
Comparing the activity of the composite catalysts (MONW/C: CNF, CNT and rGO in Figure 4b), both MONW/CNF and MONW/rGO show an excellent OER activity, and MONW/CNF also exhibits an outstanding ORR performance. The MONW/CNT composite is the least active one for both ORR/OER, probably due to a worsened interaction between Mn and C due to the lower surface area of CNTs, the low amount of defects, and we cannot discard some effect due to the encapsulation of the iron particles with the carbon matrix.
The synergistic effect between C and MONW for the OER is also reflected in the improvement of the ∆E value, around 70 to 200 mV lower for the MONW/C composites than for the individual components. This reveals that adding carbon to manganese-based catalysts is a good strategy to improve the reversibility of systems operating on ORR/OER. Both MONW/rGO and MONW/CNF show the lowest values of ∆E (1.01 and 1.06 V, respectively), mainly due to their superior performance towards OER (lower E 10 ). The composite MONW/CNT has the highest ∆E of all the composites, as it presents the lowest activities towards both ORR and OER.
The number of exchanged electrons for the ORR was assessed by applying the Koutecky-Levich (K-L) approach in the diffusional limit of the LSVs performed at different rotation speeds (400, 625, 900, 1600 and 2500 rpm). Further details can be found in the Supporting information (Section S2.3, Figures S7 and S8). The K-L plots for both carbon nanostructures and composites (MONW/C, C: CNF, CNT, rGO) are shown in Figure 5, and all the electrochemical parameters obtained from the LSVs are gathered in Table 2. CNF, CNT, rGO and MONW exhibit the exchange of ca. 3 electrons in the diffusional limit of ORR (2.9, 2.9, 3.0 and 2.7, respectively), indicating that oxygen is reduced to both hydroxide and peroxide. Despite not having the highest number of electrons transferred, CNF is more active than the other materials in the diffusional limit, as is corroborated by the intercept on the Y-axis at a lower value (higher kinetic current at 0.2 V vs. RHE). The superior performance of nanofibers can be explained by their physical-chemical properties: the presence of nickel from the catalyst [24], their relatively high surface area and the amount of structural defects acting as active sites [42]. Two paths have been described in the literature for oxygen reduction over manganese catalysts [47]. The first mechanism is the four-electron reaction that yields OH -. Since the surface of metal oxide electrodes in a strongly alkaline medium is covered by adsorbed OH species [48,49], the first step involves the displacement of OH to adsorb one O2 molecule with the acceptance of one electron (Equation (1)). The adsorbed OO is then protonated (Equation (2)), and after that, a hydroxide anion is released, and a single oxygen atom remains adsorbed (Equation (3)). Finally, the adsorbed oxygen is protonated to obtain a hydroxyl group adsorbed to the manganese oxide (Equation (4)).
Oxygen can also be reduced to peroxide in a two-electron pathway. The first two steps are the same as in the four-electron mechanism (Equations (1) and (2)). The adsorbed peroxyl then desorbs as hydroperoxide and is replaced by hydroxide (Equation (5)). This happens because the energy is not enough to break the O=O bond. MONW/C composites exhibit a higher number of exchanged electrons than the individual carbon materials. The largest synergistic effect in terms of transferred electrons is observed in the composite MONW/CNT, as it exhibits 3.8 transferred electrons. This number, however, must be analyzed cautiously, as the LSV at 1600 rpm (Figure 4b) suggests the diffusional limit has not yet been reached at 0.2 V. In the case of MONW/CNF and MONW/rGO, the number of electrons is slightly lower, with 3.4, meaning 70% of hydroxide formation. However, the high intercept of the MONW/rGO curve with the Y-axis (Figure 5b, blue triangles) indicates a low intrinsic current density at 0.2 V vs. RHE, only 8.4 mA cm −2 , which is due to the poor interaction between the carbon phase and the metallic phases, as previously explained.
Two paths have been described in the literature for oxygen reduction over manganese catalysts [47]. The first mechanism is the four-electron reaction that yields OH − . Since the surface of metal oxide electrodes in a strongly alkaline medium is covered by adsorbed OH species [48,49], the first step involves the displacement of OH to adsorb one O 2 molecule with the acceptance of one electron (Equation (1)). The adsorbed OO is then protonated (Equation (2)), and after that, a hydroxide anion is released, and a single oxygen atom remains adsorbed (Equation (3)). Finally, the adsorbed oxygen is protonated to obtain a hydroxyl group adsorbed to the manganese oxide (Equation (4)).
Oxygen can also be reduced to peroxide in a two-electron pathway. The first two steps are the same as in the four-electron mechanism (Equations (1) and (2)). The adsorbed peroxyl then desorbs as hydroperoxide and is replaced by hydroxide (Equation (5)). This happens because the energy is not enough to break the O=O bond.
The peroxide can degrade non-electrochemically to water and oxygen (Equation (6)) or react electrochemically with another adsorbed OH group to yield adsorbed oxygen and 2 hydroxide ions (Equation (7)) and then reduce to hydroxide following equation Equation (4), completing a "2 + 2" mechanism. The difference between the four-electron and the 2+2 electron pathways is that the first one is dissociative, and the latter is associative.
In both mechanisms, the rate-determining step is the electrochemical adsorption of oxygen. In fact, in a previous work [24], we showed that the ability of manganese oxide to adsorb oxygen plays a key role in the ORR kinetics. The four-electron and the two-electron pathways compete, as the operation potentials are lower than the standard reduction potentials for the two mechanisms (1.23 V vs. RHE and 0.77 V vs. RHE, respectively). As a result, the number of electrons is between 2 and 4 (2.7 in MONW at 0.2 V vs. RHE). By operating at a less negative potential, i.e., calculating the number of exchanges by the Koutecky-Levich method at 0.4 V vs. RHE, it increases to 3.7 ( Figure S9, Section S2.3). This happens because the overpotential is too low for the 2-electron pathway. As the standard potential for peroxide reduction in alkaline electrolyte is 1.71 V vs. RHE, the potentials at which the number of electrons are calculated (either 0.2 or 0.4 V vs. RHE) should imply a rapid reduction of the peroxide if it re-adsorbs. Consequently, the amount of hydroperoxide that reduces to hydroxide does not depend on the potential-but, rather, on the rates at which hydroperoxide degrades into oxygen (Equation (6)) and at which it re-adsorbs on the catalyst.
The activity of the catalysts was also assessed by analyzing Tafel slopes for both ORR and OER. Further details can be obtained in the Supporting Information. To obtain the Tafel slopes in ORR, the intrinsic current density was estimated using the Koutecky-Levich method (Equation S1 in Supporting information). The effective overpotential (η eff ) was calculated according to Equation (8): where E is the measured potential, E 0 is the standard potential at the electrolyte concentration (1.23 V vs. RHE), I is the measured current, and R is the electrolyte resistance, calculated using Newman's equation [35]. Figure 6 shows the Tafel slopes obtained for ORR ( Figure 6a) and OER (Figure 6b) for the MONW/C composites, and Figure S10 (Section S2. 4 in the Supporting Information) shows the Tafel slopes for the individual components, carbon nanostructures and MONW. The Tafel parameters for all the catalysts are collected in Table 3.
atalysts 2023, 13, x FOR PEER REVIEW 11 Tafel slopes also provide information about the mechanism of the reactions and closely related to the kinetics [50], especially to the rate-determining steps of ORR OER. As previously stated, in Mn-based catalysts, the rate-determining step for OR assumed to be the hydroxide-displacement oxygen adsorption. In fact, the obtained mV dec −1 in MONW is close to the theoretical 120 mV dec −1 calculated as the Tafel s when the first electron transfer is the rate-limiting step [51]. In contrast, carbon mate studied in this work show ORR Tafel slopes closer to 60 than to 120 mV dec −1 (espec CNF, Figure S10), suggesting that the rate-limiting step is the protonation of adsorbe rather than the first electron transfer [51,52]. This way, catalyst MONW/CNF, for exam has two phases that are able to catalyze different steps of the reaction pathway, explai its better performance than the separate phases. In the case of OER, the reverse of the electron ORR pathway was considered as the predominant mechanism (inverse of E tions (1)-(4)). The Tafel slope of MONW (73 mV dec −1 ) suggests that the third step (inv of Equation (2)) is the rate-limiting step on this material [53]. In rGO, on the other h the Tafel slope of 131 mV dec −1 suggests that the first step, i.e., the protonation of an sorbed hydroxyl group (inverse of Equation (4)), which has a theoretical Tafel slope o mV dec −1 [54]. As in the case of ORR, the different phases show the ability to cata different steps of the reaction pathway.

Catalytic Activity towards the Oxygen Reduction and Evolution Reactions for a Compos Material between Mn Oxide and Hybrid Carbon Nanostructure
Taking into account that the composite catalyst based on CNF (MOWN/CNF) the most active towards ORR while the one based on rGO (MONW/rGO) was the active towards OER, hybrid carbon materials composed of rGO and CNF were inv gated, aimed to take advantage of both carbon nanostructures towards an enha MONW/C catalyst.
Two synthetic strategies were evaluated to obtain the hybrid carbon material. first strategy consisted of mechanical mixing of the two carbon nanostructures CNF rGO (25 wt% CNF and 25 wt% rGO) with 50 wt% MONW. This process yielded a mat  The superior performance of MONW/CNF for the ORR is also evident in its low Tafel slope, 53 mV dec −1 , while MONW/CNT and MONW/rGO exhibit approximately double that value (ca. 100 mV dec −1 ). The synergistic effect between CNF and MONW can be appreciated in the composite having a lower Tafel slope than the materials on their own. In the case of OER, the three composites have similar Tafel slopes, ranging from 57 to 70 mV dec −1 . However, the MONW/rGO composite has an exceptional exchange current density (j 0 ) of 1.0 × 10 −5 mA cm −2 , one order of magnitude higher than the 1.4 × 10 −6 mA cm −2 of MONW/CNF and the 3.3 × 10 −6 mA cm −2 corresponding to MONW/CNT. When analyzing the Tafel slopes of the carbon materials, it is possible to observe that the exchange current density of rGO for OER is 4.3 × 10 −4 mA cm −2 , a higher value than the rest of the catalysts. This suggests that reduced graphene oxide provides the largest surface where the OH − anions can adsorb and then be oxidized over the manganese oxide nanowires.
Tafel slopes also provide information about the mechanism of the reactions and are closely related to the kinetics [50], especially to the rate-determining steps of ORR and OER. As previously stated, in Mn-based catalysts, the rate-determining step for ORR is assumed to be the hydroxide-displacement oxygen adsorption. In fact, the obtained 136 mV dec −1 in MONW is close to the theoretical 120 mV dec −1 calculated as the Tafel slope when the first electron transfer is the rate-limiting step [51]. In contrast, carbon materials studied in this work show ORR Tafel slopes closer to 60 than to 120 mV dec −1 (especially CNF, Figure S10), suggesting that the rate-limiting step is the protonation of adsorbed O 2 − rather than the first electron transfer [51,52]. This way, catalyst MONW/CNF, for example, has two phases that are able to catalyze different steps of the reaction pathway, explaining its better performance than the separate phases. In the case of OER, the reverse of the four-electron ORR pathway was considered as the predominant mechanism (inverse of Equations (1)-(4)). The Tafel slope of MONW (73 mV dec −1 ) suggests that the third step (inverse of Equation (2)) is the rate-limiting step on this material [53]. In rGO, on the other hand, the Tafel slope of 131 mV dec −1 suggests that the first step, i.e., the protonation of an adsorbed hydroxyl group (inverse of Equation (4)), which has a theoretical Tafel slope of 120 mV dec −1 [54]. As in the case of ORR, the different phases show the ability to catalyze different steps of the reaction pathway.

Catalytic Activity towards the Oxygen Reduction and Evolution Reactions for a Composite Material between Mn Oxide and Hybrid Carbon Nanostructure
Taking into account that the composite catalyst based on CNF (MOWN/CNF) was the most active towards ORR while the one based on rGO (MONW/rGO) was the most active towards OER, hybrid carbon materials composed of rGO and CNF were investigated, aimed to take advantage of both carbon nanostructures towards an enhanced MONW/C catalyst.
Two synthetic strategies were evaluated to obtain the hybrid carbon material. The first strategy consisted of mechanical mixing of the two carbon nanostructures CNF and rGO (25 wt% CNF and 25 wt% rGO) with 50 wt% MONW. This process yielded a material named MONW/mm[rGO-CNF] that, just like the three original composites, was comprised of 50% carbon material and 50% manganese oxide nanowires and tested in the same conditions as the others. However, this strategy was not successful since the catalytic activity for ORR and OER was not remarkable, as can be seen in the Supporting Information ( Figure S11, Section S2.5 in Supporting Information).
The second strategy consisted of synthesizing a hybrid carbon material by growing carbon nanofibers in an rGO matrix (rGO-CNF) in order to obtain an intimate mixture at the nanoscale of the two carbon phases. To do so, nickel nanoparticles, acting as CNF growth sites, were deposited on rGO, which was subsequently treated in a CH 4 atmosphere, following a similar method to the one reported in a previous work [55]. TEM images of this hybrid carbon are shown in Figure 7, where the presence of irregular nanofilaments is advised together with an amorphous carbon phase from rGO. At high magnification (Figure 7b), the graphitic layers corresponding to the nanofilaments can be clearly observed, which follow a platelet structure with an irregular surface. This hybrid rGO-CNF presents a BET surface area of 380 m 2 g −1 , much higher than the individual counterparts (95 m 2 g −1 for CNF and 175 m 2 g −1 for rGO, see Table 1), as a result of the separation of graphenic layers from rGO due to the intercalation of nanofilaments. Figure 7c shows a SEM image of the composite MONW/CNF-rGO, where a good contact between the carbon and MnO 2 phases is visible, with manganese oxide nanowires inserted between the carbon nanofibers.
To produce the catalyst, the rGO-CNF hybrid material was mixed in a planetary mill with manganese oxide nanowires in a 1:1 mass ratio following the same procedure as the other carbon materials. The composite material was named as MONW/rGO-CNF. Figure 8 shows the LSVs of the hybrid carbon composited with manganese oxide, compared to MONW/rGO and MONW/CNF, and with benchmark catalysts for ORR and OER: Pt supported on carbon, and IrO 2 , respectively.
The most remarkable result is that the catalyst MONW/rGO-CNF presents an improved activity for the ORR compared to CNF-and rGO-supported MONW. The number of exchanged electrons is higher, 3.7 (see Table 4), and the onset and half-wave potentials are more than 50 mV more positive than the other MONW/C catalysts. observed, which follow a platelet structure with an irregular surface. This hybrid rGO-CNF presents a BET surface area of 380 m 2 g −1 , much higher than the individual counterparts (95 m 2 g −1 for CNF and 175 m 2 g −1 for rGO, see Table 1), as a result of the separation of graphenic layers from rGO due to the intercalation of nanofilaments. Figure 7c shows a SEM image of the composite MONW/CNF-rGO, where a good contact between the carbon and MnO2 phases is visible, with manganese oxide nanowires inserted between the carbon nanofibers. To produce the catalyst, the rGO-CNF hybrid material was mixed in a planetary mill with manganese oxide nanowires in a 1:1 mass ratio following the same procedure as the other carbon materials. The composite material was named as MONW/rGO-CNF. Figure  8 shows the LSVs of the hybrid carbon composited with manganese oxide, compared to MONW/rGO and MONW/CNF, and with benchmark catalysts for ORR and OER: Pt supported on carbon, and IrO2, respectively. The most remarkable result is that the catalyst MONW/rGO-CNF presents an improved activity for the ORR compared to CNF-and rGO-supported MONW. The number of exchanged electrons is higher, 3.7 (see Table 4), and the onset and half-wave potentials are more than 50 mV more positive than the other MONW/C catalysts.    Table 4 also compares our catalyst with some similar materials based on manganese and carbon found in the literature. It is possible to see that our catalyst MONW/rGO-CNF has an interesting activity. Some materials, like those in [25,57,58], show reversibility gaps up to 0.13 V lower, but these differences must be considered with caution, as their results are reported corrected by IR, unlike in this work. It is interesting to note that the material by He et al. [25] has a relatively high activity, even though it uses carbon nanotubes, suggesting that the CNT used in this study have low activity due to their high number of layers and low surface area, rather than the cause being the structure of CNT. The work by Wahab et al. [58], uses reduced graphene oxide and contrary to our findings, their material has an extraordinary activity for ORR, but lower activity than our MONW/rGO-CNF material. The referenced work uses a metal-organic framework (MOF) with manganese supported over reduced graphene oxide, which may result in different interactions between the carbon and metallic phases. Moreover, the MOF as a catalyst with rGO showed extremely limited activity towards ORR, which reinforces our idea that it is a combination of graphitic carbon phases and manganese phases responsible for the high activity of the catalysts rather than the carbon phase just working as a conductive additive.
The good ORR activity of the composite reveals a synergistic effect that must comprise the three phases: the two carbons and the manganese oxide, which can be derived from the analysis of the separated phases (see Figure S6d). As stated before, oxygen reduction on rGO involves the transfer of 2.9 electrons, indicating partial reduction to peroxide. However, some works have shown that peroxyl can adsorb over the edges of graphene layers [38]. Therefore, and most probably, the superior number of electrons exchanged in MONW/rGO-CNF is explained by a 2+2 mechanism. We propose that the reduction of hydroperoxide (Equation (9)) can take place between two adsorbed species (Equations (9) and (10)).
Mn(OH) ads + C(OOH) ads + 2e − Mn(O) ads + 2OH − (10) In this case, the hydroxyperoxide anion electrochemically adsorbs over a carbon edge site, either on rGO or CNF. It is crucial for this mechanism to take place to ensure good contact between carbon and manganese phases. A possible explanation for the superior performance of MOWN/rGO-CNF over that of the rest of the carbon composites is that graphene oxide provides a large surface area where peroxyl can adsorb, and CNF improves the contact between the carbon and manganese phases. In this regard, Poux et al. [23] synthesized composites containing La-Sr-Mn perovskites and pyrolitic carbon and reached the conclusion that carbon enhances the activity of the perovskites in two different ways: first, it improves the electrical conductivity from the current collector to the outside layers of the perovskite, and second, it provides a large area where oxygen can adsorb and begin to reduce. Despite these two effects are definitively important, they are not enough by themselves to explain the increase in electrons exchanged during ORR. Our proposed explanation can deal with this observation.
Even though composite MONW/rGO-CNF has a slightly lower activity for OER than MONW/rGO, especially at high current densities, it holds especially well against the benchmark catalyst, IrO 2 . They have an almost identical onset potential, but the composite has a 20 mV lower E 10 , which may be a consequence of the good conductivity of carbon materials in comparison to metallic oxides. Therefore, composite MONW/rGO-CNF has a good compromise in activity towards ORR and OER, and consequently, it exhibits the lowest value of reversibility gap amongst the tested catalysts, with 0.99 V (see Table 3).
The activity of the MONW/rGO-CNF hybrid catalyst was also assessed, analyzing Tafel slopes for both ORR and OER, compared to the benchmark catalysts, as shown in Figure 9 and Table 5.   The Analysis of Ta Tafel slopes for ORR also reveal the superior behavior of the benchmark catalyst. Even though it is not the catalyst with the lowest slope, it has a superior exchange current density (j0), as is evident from the position of the data in the graph. Pt/C has a j0 of 4.2 × 10 −4 mA cm −2 , two orders of magnitude higher than the MONW/rGO-CNF composite, 6.3 × 10 −6 mA cm −2 , which has a comparable Tafel slope. None of our carbon-manganese catalysts can match the high activity of the well-dispersed, high surface-area Pt/C catalyst towards ORR. In the case of the OER, the benchmark material has the lowest Tafel slope (half that of MONW/rGO-CNF); however, composite MONW/rGO ( Figure 6) has a higher j0, explaining its better activity. Although the ORR activity of the MONW/rGO-CNF is significantly lower than that of Pt/C, its good performance (bifunctionality) for both ORR and OER makes it an interesting choice due to the lower cost and abundance of manganese compared to Pt.
The low ORR Tafel slope of hybrid carbon rGO-CNF of only 42 mV dec −1 ( Figure  S10a) is indicative that, in this material, the second electron transfer limits the kinetics of the reaction. In this case, it is possible to see that the different phases have the ability to catalyze the different steps of the reaction and yield the most active of the synthesized Tafel slopes for ORR also reveal the superior behavior of the benchmark catalyst. Even though it is not the catalyst with the lowest slope, it has a superior exchange current density (j 0 ), as is evident from the position of the data in the graph. Pt/C has a j 0 of 4.2 × 10 −4 mA cm −2 , two orders of magnitude higher than the MONW/rGO-CNF composite, 6.3 × 10 −6 mA cm −2 , which has a comparable Tafel slope. None of our carbonmanganese catalysts can match the high activity of the well-dispersed, high surface-area Pt/C catalyst towards ORR. In the case of the OER, the benchmark material has the lowest Tafel slope (half that of MONW/rGO-CNF); however, composite MONW/rGO ( Figure 6) has a higher j 0 , explaining its better activity. Although the ORR activity of the MONW/rGO-CNF is significantly lower than that of Pt/C, its good performance (bifunctionality) for both ORR and OER makes it an interesting choice due to the lower cost and abundance of manganese compared to Pt.
The low ORR Tafel slope of hybrid carbon rGO-CNF of only 42 mV dec −1 ( Figure S10a) is indicative that, in this material, the second electron transfer limits the kinetics of the reaction. In this case, it is possible to see that the different phases have the ability to catalyze the different steps of the reaction and yield the most active of the synthesized catalysts.
The hypothesis of different phases catalyzing different steps of the reaction was studied through EIS ( Figure 10). Figure 10a shows the Nyquist plot for ORR for MONW. A large transfer resistance can be observed over a single semicircle. The equivalent circuit is visible inside the plot. By contrast, the Nyquist plot of ORR for the MONW/rGO-CNF catalyst can be modeled by a series of two RQ elements (resistance in parallel with a constant phase element). Despite the presence of two resistances, their lower values result in a lower total charge transfer resistance, in agreement with the higher activity observed for this catalyst. More interestingly, the presence of a series of two RQ elements suggests the transfer of two electrons or, at least, two processes with similar rates, indicating that there is no single rate-determining step for ORR for this catalyst, contrary to what happens in the MONW catalyst. The large difference in the admittance of the constant phase elements in the circuit in Figure 10b suggests that the two different electron transfers could take place over different surfaces, in agreement with our hypothesis.

Stability Studies
The durability and resistance to degradation of the MONW/CNF, MONW/rGO and MONW/rGO-CNF catalysts were assessed in an endurance test at ±10 mA cm −2 for 20 h in a gas-diffusion electrode in 6 M KOH (Figure 11). Catalyst MONW/rGO-CNF shows remarkable stability, with the potential for OER increasing an average of 1.9 mV per hour of operation and the overpotential for ORR increasing less than 1 mV per hour. The overpotential for ORR (corrected by the ohmic potential loss, which was obtained from electrochemical impedance spectroscopy, see Figure S13) is around 450 mV and for OER, ca.

Stability Studies
The durability and resistance to degradation of the MONW/CNF, MONW/rGO and MONW/rGO-CNF catalysts were assessed in an endurance test at ±10 mA cm −2 for 20 h in a gas-diffusion electrode in 6 M KOH ( Figure 11). Catalyst MONW/rGO-CNF shows remarkable stability, with the potential for OER increasing an average of 1.9 mV per hour of operation and the overpotential for ORR increasing less than 1 mV per hour. The overpotential for ORR (corrected by the ohmic potential loss, which was obtained from electrochemical impedance spectroscopy, see Figure S13) is around 450 mV and for OER, ca. 330 mV in the first cycle, resulting in a potential gap between ORR and OER at 10 mA cm −2 (IR free) of ca. 780 mV. This is consistent with what was observed during the LSVs (Table 3), as the observed onset potential for OER was lower than for ORR in those conditions. The stability of the catalyst towards OER is especially remarkable since it is considerably better than the benchmark IrO 2 catalyst. Gebreslase et al. [61] showed that after only 5 h of operation at 10 mA cm −2 in 1 M KOH, the IrO 2 catalyst began to quickly lose activity. for OER and ORR. Catalyst MONW/rGO has a good performance at the beginning, especially for OER, but quickly starts to deactivate, incrementing its overpotential at a rate of 2.1 mV h −1 for ORR and 11.3 mV h −1 for OER. By the end of the 20 h, the difference of potentials in this catalyst is almost 1100 mV.

Catalysts Synthesis
Manganese oxide nanowires (MONW) were synthesized following a procedure described elsewhere [24]. Briefly, 100 mL of an aqueous solution with 1.26 g of potassium permanganate (99%, Sigma-Aldrich, St. Louis, MO, USA) and 2.8 mL of concentrated hydrochloric acid (37% wt., Fluka) was stirred at room temperature for 30 min and then transferred to an autoclave at 120 °C for 12 h. A brown precipitate formed, which was washed with 2 L of deionized water (DI) and dried overnight at 70 °C.
Carbon nanofibers (CNF) and nanotubes (CNT) were obtained by catalytic decomposition of methane, varying the synthesis catalyst and temperature. The catalyst for CNF was based on NiCuAl2O3, as described by Sebastián et al. [62], and for CNT, Fe2O3 nanoparticles (<50 nm) obtained from Sigma-Aldrich. The catalysts were reduced with a 10 mL min −1 flow of hydrogen at 550 °C for CNF and at 700 °C for CNT. Subsequently, the temperature was raised under a flow of nitrogen up to 700 °C for CNF and 900 °C for CNT, after which a flow of 50 mL min −1 of methane was passed through the reactor for 3 h.
Reduced graphene oxide was synthesized by the Tour's method [63]. Briefly, 120 mL of concentrated sulphuric acid (98% Labkem) and 13.3 mL of phosphoric acid (85% Panreac) were mixed in a flask. After cooling, 1 g of graphite (Sigma-Aldrich) was added to the mixture under stirring. An amount of 6 g of KMnO4 (99%, Sigma-Aldrich) was added slowly under stirring. The flask was then kept at 50 °C under stirring for 18 h. After that period, the mixture was poured over 135 mL of DI water ice, obtaining a brownish solution that was stirred for 30 min at 95 °C and then left to cool to room temperature. A solution of H2O2 (30% wt., 8 mL) was poured drop by drop, and then the mixture was placed in an ultrasound bath for 30 min to exfoliate the graphite oxide layers. Finally, the graphene oxide (GO) dispersion was washed and centrifuged repeatedly until a neutral pH was achieved, and the collected solid was dried overnight at 65 °C. The graphene oxide was reduced at 800 °C in a nitrogen atmosphere for 2 h to obtain reduced graphene oxide (rGO).  Figure S12 (Section S2.6 in Supporting Information) shows the endurance test for MONW/CNF and MONW/rGO, alongside the comparison with MONW/rGO-CNF. Catalyst MONW/CNF is also significantly stable, showing an increase in overpotential of 1.8 mV h −1 for ORR and 4.2 mV h −1 for OER, but exhibits a wider gap between the potentials for OER and ORR. Catalyst MONW/rGO has a good performance at the beginning, especially for OER, but quickly starts to deactivate, incrementing its overpotential at a rate of 2.1 mV h −1 for ORR and 11.3 mV h −1 for OER. By the end of the 20 h, the difference of potentials in this catalyst is almost 1100 mV.

Catalysts Synthesis
Manganese oxide nanowires (MONW) were synthesized following a procedure described elsewhere [24]. Briefly, 100 mL of an aqueous solution with 1.26 g of potassium permanganate (99%, Sigma-Aldrich, St. Louis, MO, USA) and 2.8 mL of concentrated hydrochloric acid (37% wt., Fluka) was stirred at room temperature for 30 min and then transferred to an autoclave at 120 • C for 12 h. A brown precipitate formed, which was washed with 2 L of deionized water (DI) and dried overnight at 70 • C.
Carbon nanofibers (CNF) and nanotubes (CNT) were obtained by catalytic decomposition of methane, varying the synthesis catalyst and temperature. The catalyst for CNF was based on NiCuAl 2 O 3 , as described by Sebastián et al. [62], and for CNT, Fe 2 O 3 nanoparticles (<50 nm) obtained from Sigma-Aldrich. The catalysts were reduced with a 10 mL min −1 flow of hydrogen at 550 • C for CNF and at 700 • C for CNT. Subsequently, the temperature was raised under a flow of nitrogen up to 700 • C for CNF and 900 • C for CNT, after which a flow of 50 mL min −1 of methane was passed through the reactor for 3 h.
Reduced graphene oxide was synthesized by the Tour's method [63]. Briefly, 120 mL of concentrated sulphuric acid (98% Labkem) and 13.3 mL of phosphoric acid (85% Panreac) were mixed in a flask. After cooling, 1 g of graphite (Sigma-Aldrich) was added to the mixture under stirring. An amount of 6 g of KMnO 4 (99%, Sigma-Aldrich) was added slowly under stirring. The flask was then kept at 50 • C under stirring for 18 h. After that period, the mixture was poured over 135 mL of DI water ice, obtaining a brownish solution that was stirred for 30 min at 95 • C and then left to cool to room temperature. A solution of H 2 O 2 (30% wt., 8 mL) was poured drop by drop, and then the mixture was placed in an ultrasound bath for 30 min to exfoliate the graphite oxide layers. Finally, the graphene oxide (GO) dispersion was washed and centrifuged repeatedly until a neutral pH was achieved, and the collected solid was dried overnight at 65 • C. The graphene oxide was reduced at 800 • C in a nitrogen atmosphere for 2 h to obtain reduced graphene oxide (rGO).
The carbon materials and MONW were mixed in a planetary ball mill in a 1:1 mass ratio, with a rotation speed of 100 rpm and using ethanol as a dispersing agent. The obtained composites were named MONW/C, where C is CNF, CNT or rGO.
A hybrid carbon material (rGO-CNF) was also synthesized by growing nanofibers in reduced graphene oxide. To do this, 0.7 g of GO was dispersed in 140 mL of water, and a solution of 1.48 g of Ni(NO 3 ) 2 ·6H 2 O in 68 mL of water was added dropwise. The pH of the mixture was then raised to 5 with NaOH, and a solution with 0.38 g of NaBH 4 in 204 mL of water was slowly added. The solid was then filtered, washed with abundant water and dried overnight at 60 • C. This Ni oxide on GO material was introduced in a tubular quartz reactor, reduced in H 2 at 550 • C for one hour and then CH 4 was passed at 700 • C for one hour for the growth of CNF by catalytic decomposition on Ni particles.

Physical-Chemical Characterization
MONW and carbon materials were characterized using solid-state characterization techniques. The crystalline structures of the materials were investigated using powder X-Ray Diffraction (XRD). The equipment used was a Bruker AXS D8 Advance diffractometer (Billerica, MA, USA), which has a Cu Kα radiation source and a θ-θ configuration. The interplanar distances were obtained using Bragg's equation, and the number of graphitic planes was obtained by dividing the 002 crystal size (obtained from Scherrer's equation) by the interplanar distance. The determination of the crystallographic phases was carried out through Rietveld refinement. The morphology of MONW and carbon nanostructures was investigated by both transmission (TEM) and scanning (SEM) electron microscopy. TEM images were acquired with a Tecnai F30 microscope (Thermofisher, Waltham, Massachusetts, Estados Unidos operated at 300 kV. The samples were ultrasonically dispersed in ethanol for 15 min and then placed in a Cu carbon grid. SEM images were obtained with a Hitachi 3400-N microscope (Hitachi Ltd., Tokyo, Japan) with the powdered samples directly placed on a carbon tape. Textural features of MONW were assessed through nitrogen physisorption at −196 • C in Micromeritics ASAP 2020 (Norcross, GA, USA). The amount of metals in the samples was determined by inductively coupled plasma combined with atomic emission spectrometry (ICP-AES) using an Xpectroblue-EOP-TI FMT26 spectrometer (SPECTRO Analytical Instruments, Kleve, Germany). In addition, the carbon and hydrogen content of carbon materials was determined by elemental analysis (EA) in a Thermo Flash 1112 analyzer (Thermoscientific, Waltham, MA, USA).

Electrochemical Characterization
All the electrochemical tests were carri, ed out in a three-electrode cell connected to an Autolab potentiostat/galvanostat (Metrohm 302N, Metrohm Hispania, Madrid, Spain). The activity of the electrocatalysts was assessed through linear sweep voltammetries (LSVs) in a 0.1 M KOH electrolyte, using a reversible hydrogen electrode as the reference electrode and a graphite rod as the counter electrode. The working electrode was a glassy carbon rotating disk electrode (RDE, 5 mm diameter), on top of which a fine layer of the catalyst was deposited. To do so, an ink was prepared dispersing 5 mg of the catalyst and 10 µL of Nafion (10% wt., Sigma-Aldrich, St. Louis, MO, USA) in 490 µL of ethanol under sonication and then deposited over the RDE in order to achieve a loading of 1 mg cm −2 . The LSVs to evaluate the ORR were performed in the range 1.2-0.2 V vs. RHE at a sweeping rate of 5 mV s −1 in oxygen-saturated electrolyte and at different rotation speeds (400, 625, 900, 1600 and 2500 rpm). Prior to the ORR LSVs, blank LSVs in nitrogen-saturated electrolyte were complete in the same potential range and used as a baseline. OER tests were carried out at 1600 rpm in the 1.1-1.8 V vs. RHE potential range at 5 mV s −1 of scan rate. Electrochemical impedance spectroscopy (EIS) tests were executed at the potential at which current had a magnitude of 1 mA cm −2 for both ORR (E j = 1) and OER, with a sine wave of 10 mV of amplitude and in the range between 10 kHz and 100 mHz, acquiring ten points per decade. Endurance tests were carried out over a gas diffusion electrode of 0.5 cm 2 of exposed area in the strong basic electrolyte (6 M KOH). An ink of the same characteristics as the one of the RDE was sprayed over carbon paper, achieving a coverage of ca. 800 µg cm −2 . The reference electrode was a Hg|HgO electrode in 1 M KOH. Consecutive cycles (OER/ORR) of 2.5 h for each reaction at a current density of ±10 mA cm −2 were run.

Conclusions
Three different carbon nanostructures (carbon nanofibers, carbon nanotubes and reduced graphene oxide) were synthesized and mixed with manganese oxide nanowires to obtain bifunctional oxygen electrocatalysts. The type of carbon employed in the catalyst had an important effect on the activity, being the catalyst synthesized with CNF the most active towards ORR and the catalyst with rGO the most active towards OER, even surpassing the performance of the benchmark IrO 2 catalyst. Catalysts with CNT showed poor activity for both reactions, which was ascribed to their low surface area and lack of structural defects. The Koutecky-Levich plots show that the MONW/C catalysts have an increased number of electrons transferred in comparison to the phases on their own, indicating a synergistic effect that goes beyond just increasing conductivity or enhancing surface area but also changes the pathway of the reaction. The analysis of Tafel plots and electrochemical impedance spectroscopy allowed us to conclude that carbon and manganese oxide phases can catalyze different steps of the reaction pathways, and thus, by mixing them, a catalyst with lower energetic barriers for the reactions could be obtained. With this information in mind, we synthesized a hybrid rGO-CNF material and mixed it with MONW, which showed increased activity towards ORR in comparison to MONW/CNF and a closer activity towards OER than that of catalyst MONW/rGO. The hybrid MONW/rGO-CNF catalyst was tested in demanding conditions in a gas diffusion electrode (6 M KOH, at ±10 mA cm −2 ) and showed remarkable stability for 20 h of operation, maintaining less than 0.77 V of difference between the potentials for OER and ORR (IR corrected) under the mentioned conditions. This work proves that the structure of carbon materials used as an additive in manganese bifunctional oxygen catalysts plays a key role in both the mechanisms of the reactions and the performance of the catalysts and that this can be exploited to rationally design catalysts with enhanced activity and/or bifunctionality.