Alternative Supports for Electrocatalysis of the Oxygen Evolution Reaction in Alkaline Media

Abstract

The anodic stability of tungsten carbide (WC) and iron oxide with a spinel structure (Fe₃O₄) were compared against similar data for nanostructured, boron-doped diamond (BDD), and the benchmark Vulcan XC72 carbon, in view of their eventual application as alternative supports for the anion exchange membrane electrolyzer anode. To this end, metal oxide composites were prepared by the in situ autocombustion (ISAC) method, and the anodic behavior of materials (composites as well as supports alone) was investigated in 1 M NaOH electrolyte by the rotating ring–disc electrode method, which enables the separation oxygen reduction reaction and materials’ degradation currents. Among all supports, BDD has proven to be the most stable, while Vulcan XC72 is the least stable under the anodic polarization, with Fe₃O₄ and WC demonstrating intermediate behavior. The Co₃O₄-BDD, -Fe₃O₄, -WC, and -Vulcan composites prepared by the ISAC method were then tested as catalysts of the oxygen evolution reaction. The Co₃O₄-BDD and Co₃O₄-Fe₃O₄ composites appear to be competitive electrocatalysts for the OER in alkaline medium, showing activity comparable to the literature and higher support stability towards oxidation, either in cyclic voltammetry or chronoamperometry stability tests. On the contrary, WC- and Vulcan-based composites are prone to degradation.

1. Introduction

One of the major challenges of sustainable development is the replacement of fossil fuels by renewable energy. The intermittency of renewable energies (solar, wind) poses an issue of energy storage. Hydrogen produced by water electrolysis is seen as a promising green fuel allowing the storage of renewable energies. Two reactions occur in parallel during water electrolysis: hydrogen production at the cathode and oxygen at the anode, the latter known as the oxygen evolution reaction (OER, Equation (1)). However, the oxygen evolution reaction is a sluggish kinetics reaction responsible for the low electrochemical performance of the electrolyzer. One of the challenges in deploying this technology is to improve the OER kinetics and develop innovative and efficient materials without noble metals, as reviewed in [1]. Transition metal oxides with a spinel structure are promising electrocatalysts for the OER in alkaline media, because they are (i) easy to synthesize, (ii) inexpensive, and (iii) their properties can be widely modified by changing their elemental composition.

4HO⁻ → O₂ + 4e⁻ + 2H₂O

(1)

Due to its high activity and stability, Co₃O₄ oxide with spinel structure is extensively studied for the OER [2,3,4]. However, because of its low electronic conductivity, it is often either mixed with or supported on carbon black, containing a high content of sp² carbon and giving it high electronic conductivity [5,6,7]. The addition of carbon black allows the improvement of the apparent electrocatalytic activity of the oxide nanoparticles; specifically, it increases the electronic conductivity of the catalyst thanks to the improvement of the surface utilization of the oxide nanoparticles due to the increase in the number of accessible active sites [6,7]. However, carbon black is oxidized under the OER condition (see Carbon Oxidation Reaction (COR) (Equation (2))), which prevents their use in water electrolysis systems. To this end, we have recently investigated the stability of carbon materials in alkaline media [4].

C + 4HO⁻ → CO₂ + 4e⁻ + 2H₂O

(2)

In this work, we investigate nanostructured, boron-doped diamond (BDD), tungsten carbide (WC), and iron oxide with a spinel structure (Fe₃O₄) as alternative catalyst supports for preparing Co₃O₄-based OER composites.

BDD consists of sp³ carbon with a negligible contribution of sp² carbon [8], unlike the Vulcan XC72 carbon from Cabot, containing a high content of sp² carbon, used as a benchmark. There is considerable interest in using BDD as an electrode material (see the Table S1 in Supplementary Information). Electrodes based on BDD have been investigated in the literature [9]. It was concluded that BDD offers a wide electrochemical window, high anodic stability [9], chemical inertness [9], low capacitance, and a fast electrochemical response [10]. It is known to withstand corrosive and high temperature/pressure environments [10]. Even though the BDD is more expensive than the Vulcan XC72 carbon, we recently observed that the apparent lifetime for the BDD is more than an order of magnitude higher than the lifetime of Vulcan XC72 by applying Faraday’s law, as explained in [5].

Tungsten carbide (WC) is a combination of carbon and tungsten atoms with the high strength and rigidity of a covalent compound, the high melting point of an ionic crystal, and the electromagnetism of a transition metal. This ceramic has been studied in chemical and electrochemical catalysis since the discovery of its Pt-like character, as reported by Levy and Boudart [11]. Tungsten carbide’s low electrical resistivity of about 0.2 µΩ·m is comparable with that of some metals (e.g., vanadium). Several publications have reviewed WC as an electrocatalyst, but misunderstandings about its stability persist. For example, Zhu et al. show that WC can be used as a stable carrier in fuel cells, where it has shown superior stability and conductivity to commercial carbon products, such as Vulcan [12]. Supported metal catalysts on WC have shown stability and high activity towards HER [13,14] and other reactions, such as methanol oxidation or OER [15,16]. Weidman et al. demonstrated that the potential stability window of WC is much wider in acidic than alkaline electrolytes [17]. Finally, Göhl et al. recently warned the users of WC regarding the degradation of WC in acid in the potential interval relevant to fuel and electrolysis cells [18].

Fe₃O₄ is a rather inexpensive and highly conductive material. Although most transition metal oxides are semiconductors, Fe₃O₄ (magnetite) has low electrical resistivity 0.3 mΩ m [19], which is significantly lower than that of Fe₂O₃ (~kΩ m). This is ascribed to electron exchange between the Fe(II) and Fe(III) ions in Fe₃O₄. Recently, Fe₃O₄-core CoFe₂O₄-shell nanoparticles showed high stability and high OER activity in alkaline medium per Co unit mass, suggesting synergetic interaction between the core and the shell [20,21].

To evaluate the suitability of WC and Fe₃O₄ as alternative supports for the transition metal oxide-based electrocatalysts in OER condition, we studied their anodic behavior in the potential interval 1.53–2.03 V vs. RHE in NaOH 1 M and compared it to the similar measurements published for Vulcan XC72 and BDD [22]. Since at high anodic potentials the electrochemical degradation of materials (such as COR for carbons) occurs in parallel with OER, the rotating ring–disc electrode (RRDE) was used to separate these two contributions. Previously, we evaluated the limits of applicability of the RRDE during OER in alkaline media for determining the amount of oxygen [5,23]. It has been shown that the RRDE is a suitable approach for separating the ORR and the support oxidation reaction in the absence of vigorous oxygen bubble formation.

Furthermore, Co₃O₄-WC, -Fe₃O₄, -BDD, and -Vulcan XC72 composites were prepared by an in situ autocombustion method described in our previous publications [6,7]. Once synthesized, the materials were analyzed using X-ray diffraction (XRD), thermogravimetric analysis (TGA), transmission electron microscopy (TEM), electron energy loss spectroscopy (EELS), X-ray photoelectron spectroscopy (XPS), and low-temperature nitrogen adsorption (analyzed in the framework of Brunauer–Emmett–Teller approximation, BET method). These analyses are used to define the structure of the synthesized metal oxide, the composition of the composite (metal oxide support), and the active surface of materials. Then these materials were electrochemically studied using the RRDE to determine their OER activity and the oxygen faradaic efficiency. Stability tests were performed at 1.66 and 1.58 V vs. RHE to assess the tolerance of composites to degradation.

2. Materials and Methods

2.1. Materials

A 5 mg mL⁻¹ suspension of BDD in deionized water was purchased from Adama Nanotechnology. The morphological and structural characteristics of mesoporous BDD are summarized in

Table 1. Morphological, structural, and electrochemical characteristics of Fe₃O₄ commercial, WC commercial, Vulcan XC72, BDD, Co₃O₄AC, and ISAC composites.

Support or Composite	SBET/m2 g−1	SBJH/ m2 g−1	VBJH/cm3 g−1	Smicroa/m2 g−1	Vmicro a/cm3 g−1	Mean Pore Size/nm	DDRX b/nm	Q c/ C g−1
WC commercial	4	4.9	0.010	-	-	8.5	-	-
Fe3O4 commercial	36	43.4	0.045	-	-	4.2	-	-
Vulcan XC-72	194	117	0.3	118	59	11	-	-
BDD	181	184	0.4	-	-	8.5	-	-
16Co3O4ISACBDD	43.1	42.5	0.1	4	0.0003	-	21	24.40
29Co3O4ISACVulcan	80	46.7	0.33	34.8	0.0146	28.5	9	4.80
26Co3O4ISACFe3O4	7	11.4	0.04	10	-	14	23	10.00
32Co3O4ISACWCcalc	19	20.6	0.052	21.3	-	10.1	22	-
Co3O4AC	1.47	0.56	0.0067	1.18	0.0006	-	24	0.01

S_BET: BET surface area. S_BJH: BJH surface area, V_BJH: BJH volume and mean mesopore size were measured by B.J.H method. B.J.H. and BET methods are applied to the N₂ adsorption branch of isotherm. ^a V_micro and S_micro: volume and surface of micropores determined from the t-plot of the N₂ adsorption branch of the isotherm. ^b The Scherrer equation with the shape factor K = 0.89 was used to estimate the average size of the Co₃O₄ crystallite from the XRD data. ^c Determined from electrochemistry, see Section 3.2.2 for details.

, with a BET specific surface area of 181 m² g⁻¹, a BJH pore volume of 0.4 cm³ g⁻¹, and a BJH adsorption mean pore width (“mean pore size” in Table 1) of 8.5 nm. The morphological and structural characteristics of BDD have been extensively studied in the literature [22,24,25,26]. The specifications of BDD used in this work are summarized in Table S1 in the Supplementary Information.

WC powder was purchased from Merck. WC consists of nanoparticles with an irregular and spiky menhir-like shape [22,24,25,26]. It is a mesoporous material with a low specific surface (4 m² g⁻¹ by BET), a low BJH volume of pores (0.010 cm³ g⁻¹) compared to BDD nanoparticles, and a mean pore size of 8.5 nm (Table 1).

Fe₃O₄ (magnetite) has a cubic inverse spinel group structure, which consists of a cubic close packed array of oxide ions where all the Fe(II) ions occupy half of the octahedral sites, and the Fe(III) are split evenly across the remaining octahedral and the tetrahedral sites. Fe₃O₄ was purchased from Merck. It is a mesoporous material with a lower porosity than BDD nanoparticles. Its specific surface area is 36 m² g⁻¹, its BJH volume of pores is 0.045 cm³ g⁻¹, and its mean pore size is 4 nm (see Table 1).

Vulcan XC72 is a mesoporous carbon black purchased from Cabot Company, with a similar porosity than BDD. Its specific surface, determined by the BET method, is 192 m² g⁻¹. Its BJH volume of pores is 0.3 cm³ g⁻¹ and the mean pore size is 11 nm (Table 1). Vulcan XC72 nanoparticles are spherical and have a regular shape [7].

The unsupported Co₃O₄ was synthesized by autocombustion (AC), as detailed in [22] and the following Equation (3):

3Co(NO₃)₂ + Ψ NH₂CH₂COOH → Co₃O₄ + 2Ψ CO₂ + 5Ψ/2H₂O + (6 + Ψ)/2N₂ + (7 − 9/4Ψ)O₂

(3)

To prepare the composites within the in situ autocombustion (ISAC) route, an appropriate amount of support material (either BDD or WC, Fe₃O₄, Vulcan XC72) was added depending on the desired composition. Only the WC composite was calcined at 300 °C for 30 min, as indicated by the addition of “calc” at the end of the composite name, to ensure the formation of a spinel phase. The nomenclature used to designate the composites is explained in [6,7,22] and summarized in Scheme 1. The composite designation begins with the weight % of Co₃O₄ in the final material (determined by TGA), then the active phase (Co₃O₄ in this work), the synthesis (AC or ISAC), and finally the support material (BDD or WC or Fe₃O₄ or Vulcan). The unsupported Co₃O₄ synthesized by the autocombustion is named Co₃O₄AC. High-purity water (18.2 MΩ cm, <1 ppb TOC, ELGA Purelab, UK) was used to prepare all aqueous solutions in this study.

2.2. Material Characterisation

The synthesized materials were characterized by X-ray diffractometer (XRD), transmission electron microscopy (TEM), N₂-physisorption, and thermogravimetric analysis (TGA) to investigate the phase purity, the morphology, and the specific surface area (S_BET) using the Brunauer, Emmett, and Teller theory and by the Barrett, Joyner, and Halenda method for the volume (V_BJH) and the surface area (S_BJH) of the pores and the weight percentage of Co₃O₄ in the composites, respectively. Details of the equipment used and the conditions for analysis are given in [22]. Moreover, a Gatan spectrometer and an ultrahigh vacuum spectrometer (equipped with a RESOLVE 120 MCD5 hemispherical electron analyzer, using the Al Kα line (1486.6 eV) of a dual anode X-ray source as incident radiation, and recording the survey and high-resolution spectra in constant pass energy mode at 100 and 20 eV respectively) were used to determine the presence of cobalt element in the supported and the unsupported spinel (and the presence of iron element in the 26Co₃O₄ISACFe₃O₄ composite). Spectral maps were produced using a 30 mm² EDAX Octane 9 silicon detector mounted on a Zeiss GeminiSEM 500 scanning electron microscope (SEM). A working distance of 13 mm enabled both the microanalysis (EDX) and backscattered electron (BSD) detectors to be used for electron imaging. A working voltage (EHT) of 12 kV was used to make the K lines of Fe and Co visible.

2.3. Electrochemical Measurements

The electrocatalytic OER activity of the Co₃O₄-supported composites was evaluated using a thin-film approach. The preparation of electrode and the conditions of the electrochemical measurements are described in [22,27]. The following equation was used to determine the oxygen faradaic efficiencies (F.E.):

F.E.(%) = 100 × (4I_R/n_ORR)/(N × I_D)

(4)

where I_D and I_R are the disc and the ring currents, n_ORR = 2 for a gold ring [28], and N is the collection factor. N factor determination is estimated using the well-known reversible ferro-, ferricyanide redox couple, as explained in [22], where N = 0.22 at 1600 rpm.

The potential was iR-corrected after the measurements using the estimated cell resistance, as described in [22].

Aging tests were performed by either staying at a constant potential of 1.66 V vs. RHE for 3 h or at 1.58 V vs. RHE for 30 min at 10 mV s⁻¹, depending on the composite studied.

2.4. The Methodology for Measuring the OER Activity and Selectivity

While numerous publications report on the activities and behavior of transition metal oxide catalysts in the OER, for the same catalyst composition, the reported activity values often show significant variation. Thus, the following questions can be asked: why the data are so different, and is the OER activity properly determined? The OER activity is conventionally determined in liquid electrolyte three-electrode electrochemical cell using a thin-film rotating disc (RDE) approach, which was proposed by Schmidt et al. for carbon-supported Pt [29] and later extended by Suntivich et al. [30] to transition metal oxides. It should be noted, however, that the assessment of electrocatalytic activity by non-planar thin-film RDE is by no means trivial, since it is based on several assumptions, among them are (i) homogeneity of the film (ii) and the catalyst effectiveness factor equal to unity (100%) (for more information the reader is referred to [31] and references therein). The latter is only valid if the catalyst is conductive, and either current or potential gradients are absent in the catalyst film. Since many of transition metal oxides are semiconductors, the conductivity of the oxide film may not be sufficient for ensuring 100% catalyst utilization. Even for conductive materials (e.g., carbon-supported oxide nanoparticles) for thick catalyst layers, potential and current gradients may arise (subject to the catalyst film thickness and hence loading on the RDE) in the film compromising activity evaluation.

A convenient approach to check whether the electrocatalytic activity is correctly determined is to study the influence of the catalyst loading on the measured current. For example, Figure 1A shows OER current measured at 1.55 V vs. RHE versus the oxide loading on the GC disc for the 29Co₃O₄ISACVulcan composite. It appears clearly that the disc current is proportional to the oxide loading up to 100 µg cm⁻² and then shows high scatter either levelling off or decaying. The plot confirms that catalytic activity cannot be correctly determined at high catalyst loadings. In Figure S1 in the Supplementary Information, the OER current measured at 1.55 V vs. RHE versus the oxide loading on the GC disc for the 16Co₃O₄ISACBDD, 32Co₃O₄ISACWC, and 26Co₃O₄ISACFe₃O₄ composites in NaOH media are presented. The disc currents are at least proportional to the Co₃O₄ loading up to 40 µg cm⁻². Considering this, in what follows, we used the oxide loadings below 40 µg_oxide cm⁻².

Another important characteristic related to the OER studies is the faradaic efficiency (F.E.), since the measured current does not necessarily correspond to the OER alone but may include electrochemical degradation currents of the support or the metal oxide itself. The faradaic efficiency of the OER (or O₂%) can be studied, for example, by the RRDE approach. The RRDE method is widely used in electrochemistry for the detection of electroactive reaction intermediates, since the current measured at the ring gives an indication of the amount of species formed at the disc. However, the RRDE technique is applicable for the detection of oxygen produced at the disc only under specific conditions, notably in the absence of oxygen bubble formation. We estimated a critical disk current density, i.e., 450 µA cm⁻², at which the oxygen bubble formation starts to affect the measured OER efficiency and explained that this critical current density value likely depends on the microstructure and porosity of the IrO₂ supported on glassy carbon layer [23]. Thus, the ring of the RRDE can be used as an O₂ sensor only when the rate of the O₂ formation is not too high, which requires the application of low catalyst loadings and avoiding high electrode potentials. For example, Figure 1B represents the OER faradaic efficiency (F.E.) versus time for the 29Co₃O₄ISACVulcan sample measured at a constant potential of 1.58 V vs. RHE (with E_Ring = 0.3 V vs. RHE) at two selected oxide loadings (2 and 15 µg_oxide cm⁻²). One can see that the faradaic efficiency is close to 100% for the lower loading of oxide but decays down to 70% at 15 µg cm⁻² loading. Note that the noise in the transient is due to bubble formation, the latter escaping the ring and resulting in the underestimation of the F.E. Figure 1C is associated to Figure 1B and represents disc currents versus time. It confirms that the underestimation of the F.E. at 15 µg cm⁻² oxide loading is due to the highest disc current, hence the high rate of O₂ production. Then a critical disk current may be estimated, in this work, at around 200 µA cm⁻² from the Figure 1C.

Thus, we conclude that to compare the anodic behavior of various materials, i.e., OER activity and F.E., it is essential, in addition to avoid high electrode potentials, to optimize the loading of material on the GC disk of an R(R)DE. Firstly, the loading must not be too high to ensure a correct determination of the electrocatalytic activity (shown in Figure 1A) and the formation of bubbles of oxygen during the OER, thus avoiding errors in the measurement of the F.E. (shown in Figure 1B) [22]. Secondly, the loading must be high enough to cover the surface of the disc (in our case the GC) surface and avoid the contribution of the GC to the total measured current, as discussed in more detail in Section 3.2.1.

3. Results and Discussion

3.1. Morphology and Structure of Synthesized Materials

The formation of a crystalline Co₃O₄ spinel phase (ICDD card n°042-1467) designated by a “#” symbol both for the unsupported oxide synthesized with the AC method and for ISAC composites was evidenced by XRD in Figure 2. Moreover, Fe₃O₄ oxide (COD 1010369) is designated by a “§” symbol, WC by a “$” symbol, BDD by a “£” symbol, and Vulcan XC72 by a “&” symbol. We can notice, in the XRD pattern of 16Co₃O₄ISACBDD and the 32Co₃O₄ISACWCcalc composites, respectively, a small amount of the CoB phase (ICDD card n°75-1066) at 2θ = 49.3° designated with the @ symbol and a small amount of the tungsten trioxide WO₃ phase (COD 2311041) at 2θ = 23.6–24.4° designated with the “/” symbol (Figure 1). Hatel et al. showed that tungsten trioxide can be formed by oxidation and/or the thermal treatment of WC [32]. The size of the spinel oxide crystals is listed in Table 1 (applying the Scherrer equation with the shape factor K = 0.89).

In Table 1, one may notice similar crystal size of Co₃O₄ in the 16Co₃O₄ISACBDD, 26Co₃O₄ISACFe₃O₄, 32Co₃O₄ISACWCcalc, and Co₃O₄AC samples (around 22 nm) and a twice smaller one (around 9 nm) in the 29Co₃O₄ISACVulcan composite. Previously in [6,7], we investigated LaMnO₃ perovskite oxide–carbon composites prepared by the ISAC method and concluded that the specific surface area, pore size, and size distribution of carbon supports affect nucleation and the growth of oxide nanoparticles, with Vulcan XC72 offering the best characteristics among studied carbon materials. In this work, it seems also that among the studied materials, the porosity of Vulcan XC72 is best suited for the nucleation and growth of Co₃O₄ nanoparticles resulting in the smallest XRD derived average particle size.

Figure 3 shows TEM images of composites and bare supports. Note that in Figure 3, different magnifications are deliberately used to better observe the support and the oxide components, whose particle size varies in the composites studied. Vulcan XC72 carbon consists of spherical nanoparticles of around 25 nm (Figure 3C). BDD nanoparticles look like irregular spiky menhir (see Figure 3I and images in [24,25]). Fe₃O₄ comprises spherical nanoparticles of 15–30 nm (Figure 3L). WC material comprises menhir-like shape structures of a few hundred nanometers (Figure 3E,F). In the unsupported Co₃O₄AC sample, 15–50 nm size spherical nanoparticles are strongly agglomerated (Figure 2C of [22]), while in the Vulcan-supported sample (Figure 3A,B), BDD-supported sample (Figure 3G,H), Fe₃O₄-supported sample (Figure 3J,K), and WC-supported sample (Figure 3D), spherical nanoparticles are observed on the support surface with a smaller size around 5–10 nm. We have to notice that for the 26Co₃O₄ISACFe₃O₄ composite it is difficult to discern Fe₃O₄ and Co₃O₄ particles, since both crystallize in a similar spinel structure. However, based on the images of bare Fe₃O₄ support, one may hypothesize that the size of Co₃O₄ is around 5–10 nm; whereas, the size of Fe₃O₄ is higher, i.e., 15–30 nm. EDS elemental mapping and the corresponding spectra help us identify the presence of iron, cobalt, and oxygen in the 26Co₃O₄ISACFe₃O₄ composite (see Figure S2 in the Supplementary Information). EDS mapping shows that oxygen is distributed homogeneously throughout the composite. The iron and cobalt elements appear separately but in the presence of oxygen. The atomic composition of the oxygen and metals is consistent with the formation of spinel oxides (i.e., Co₃O₄ and Fe₃O₄). EDS mapping overlay shows an enrichment of Co at the surface of the composite, indicating that the nanoparticles of Co₃O₄ spinel recover the nanoparticles of Fe₃O₄ spinel during the ISAC synthesis. This is consistent with previous studies in which metal oxide nanoparticles were dispersed within the pores and at the surface of the support [7,22]. It should be noted that traces of carbon are observed in the EDS spectra, which may be due to carbonaceous residues formed during the ISAC process, as explained in our previous papers [6,7]. From the XRD, TEM, and EDS elemental data (Figure 2 and Figure 3 and Figure S2 in Supplementary Information), we conclude that the ISAC route is well suited to support Co₃O₄ nano-sized particles on Fe₃O₄ and WC. This agrees with Co₃O₄ supported on BDD and Vulcan XC72 carbon [22]. The fact that for BDD-, Fe₃O₄-, and WC-based composites, the TEM-size of Co₃O₄ nanoparticles is substantially smaller than the DRX-size can be explained by the existence of a fraction of larger particles in these composites resulting in narrowing of the XRD peaks. The formation of the unsupported Co₃O₄ spinel and supported composites was confirmed by XPS and EELS in Figure S3 and discussed in the Supplementary Information.

Table 1 also provides values of the BET surface area, the surface and area of mesopores and micropores, and the mean pore size for the supports and the synthesized composites. Note that the specific surface area determined by the BET method corresponds to the total surface area of the metal oxide and the supports accessible to N₂ adsorption and thus does not show direct correlation with the size of metal oxide crystallites. ISAC synthesis allows the development of high specific surface composites of several tens m² g⁻¹. However, we have to note that if the specific surface area of the support is low, the specific surface area of the corresponding composite is also low. For example, for the 26Co₃O₄ISACFe₃O₄ composite, which has a specific surface of 7 m² g⁻¹, the specific surface area of Fe₃O₄ material support is 36 m² g⁻¹. Indeed, in a previous paper [7], we suggested that the morphology and the porous structure of carbon materials influence the growth of the metal oxide nanoparticles in the composites prepared by in situ autocombustion. Further experiments should be carried out to improve the synthesis and increase the specific surface area of Co₃O₄ composite based on Fe₃O₄.

3.2. Electrochemical Characterisation

3.2.1. Anodic Behavior of Vulcan XC72, BDD, WC, and Fe₃O₄

To investigate the anodic stability of WC and Fe₃O₄ and thus evaluate their suitability as support for the Co₃O₄ OER catalysts, we used RRDE and applied successive potential steps of 1.53–2.03 V vs. RHE in increments of 50 or 100 mV to the disc for 15 min and then measured the resulting current transients (Figure 4). By measuring the current of the disk (Figure 4A) and the current of the ring (Figure 4B), the OER current faradaic efficiency (F.E. calculated using Equation (4) and plotted in Figure 4C) can be calculated. Then the OER currents can be separated from the electrochemical support degradation currents shown in Figure 4D. The results obtained were then compared with the recently published data [22] for Vulcan XC72 carbon and BDD supports plotted in Figure 4 for convenience. Note that for Vulcan XC72, which produced the highest current at the disc (Figure 4A), a lower loading (91 μg cm⁻²) was used compared to other materials (183.4 µg cm⁻²). Indeed, in our previous publication, we have shown that for BDD (which is much more stable against anodic degradation than conventional carbon black materials), the utilization of low loading results in an overestimation of degradation currents, since the GC support (the reader is referred to [22] for further details), unless fully screened by the thin layer, may also contribute to the measured currents (see dotted curves of GC in Figure 4). Therefore, the loading of BDD was optimized at 183.4 µg cm⁻², and a similar loading was also used for Fe₃O₄ and WC materials.

Figure 4A shows that disc currents increase with potential for all the studied materials with the highest anodic currents detected for Vulcan XC72 and the lowest for the BDD and WC powders, which is consistent with the chemical and electrochemical inertness previously reported in the literature [9,10,17,20,21]. One should keep in mind however that high anodic currents do not necessarily reflect fast material’s degradation, since the OER in alkaline electrolytes may occur on the studied materials even without Co₃O₄. This is reflected by the ring currents (Figure 4B) and the faradaic efficiencies (Figure 4C) calculated from the disc and ring currents (see Equation (4)). The ring currents (Figure 4B) increase with potential for Vulcan XC72 in all the potential interval. For WC, Fe₃O₄, and BDD materials the ring currents increase from 1.53 to 1.83 V vs. RHE and decrease from 1.83 to 2.03 V vs. RHE. The highest ring currents are detected for Vulcan XC72 and the lowest one for the WC and BDD powders, similarly to the disc currents (Figure 4A). Then depending on the ratio of ring and disc currents, the faradaic efficiencies are calculated (Figure 4C) and increase from 1.58 V to 1.73 V vs. RHE for all materials, reaching ~85–95% for Vulcan XC72, 60% for WC, ~70% for BDD, and 80% for Fe₃O₄. At a potential greater than 1.73 V, the OER efficiencies decrease.

Since the disk currents (Figure 4A) is the sum of the OER and the support degradation currents, the latter can be calculated from the overall disk current as j_Disk (1–FE). Figure 4D represents the support degradation currents normalized to the support mass (the support degradation currents normalized to the geometric electrode area are presented in Figure S4 in the Supplementary Information). One may notice (i) degradation currents are observed on all the supports used in this study and (ii) lower mass-normalized support degradation currents on BDD, WC, and Fe₃O₄ compared to Vulcan XC72 in Figure 4D. For example, at 1.58 V vs. RHE the mass-normalized support degradation currents equal 500 mA g⁻¹ for Vulcan XC72 against 65 mA g⁻¹ for WC, 35 mA g⁻¹ for Fe₃O₄, and 5 mA g⁻¹ for BDD. Note also that the support degradation currents may be overvalued, since it is not possible to exclude the contribution of the surface of GC to the measured RRDE transients (see dotted curves of GC in Figure 4 and Figure S4 in the Supplementary Information). One may notice that for Fe₃O₄, which is thermodynamically unstable in the whole interval of investigated potentials (see Pourbaix diagram of iron), degradation kinetics is slower at low potential and increases at higher potentials (Figure 4D and Figure S4 in the Supplementary Information). The observed slow degradation kinetics agrees with the electrochemical inertness of Fe₃O₄ reported in the literature [20,21] and may be attributed to the formation of a thin passivating layer of γ-Fe₂O₃ on the surface of Fe₃O₄ nanoparticles [33] and reconstruction of the Fe₃O₄ surface [34]. Note that the thickness of the γ-Fe₂O₃ layer may be particle size-dependent as reported by Baaziz et al. [35]. At higher potentials, Fe₃O₄ could be oxidized into soluble species containing iron in a higher oxidation state (e.g., FeO₄²⁻) (see Pourbaix diagram of iron).

Regarding the WC material, even if the disc currents remain low (Figure 4A), the low and stable F.E. of the OER (Figure 4C) results in substantial currents of degradation. At a potential less than 1.63 V vs. RHE, WC degrades faster than Fe₃O₄, and at potential less than 1.93 V vs. RHE (Figure 4D), WC degrades faster than BDD. This agrees with the paper of Weidman et al. stating that WC is prone to dissolution in alkaline pH at OER potential [36]. Moreover, Göhl et al. observed that WC showed significant corrosion within the potential window for the oxygen evolution reaction, especially at oxidative potentials higher than 0.80 V vs. RHE. After reaching this threshold potential, the surface oxidizes continuously, and only partial passivation is achieved, even in acid electrolyte [18]. BDD shows the lowest degradation currents at potentials less than 1.93 V vs. RHE, where it degrades slightly more than WC. Anyway, BDD, Fe₃O₄, and WC show lower degradation currents compared to Vulcan XC72, with BDD being the most stable among the material supports.

3.2.2. Electrochemical Properties of Composite Materials

Cyclic voltammetry (CV) in a supporting electrolyte under nitrogen atmosphere is a convenient tool for in situ exploring of the interfacial properties of electrocatalytic materials. In Figure 5, iR-corrected CVs and chronoamperograms of unsupported Co₃O₄AC and ISAC composites are presented. Figure 5B shows an enlargement of the current of the disc in Figure 5A. Note that the disc currents of Co₃O₄AC and the Vulcan XC72, BDD, Fe₃O₄, and WC supports are shown in Figure S5 in the Supplementary Information for comparison. As outlined in previous publications [37,38,39,40,41], the redox peak observed around 1.45 V vs. RHE in Figure 5B can be assigned to the Co³⁺/Co⁴⁺ redox couple. It is interesting to note that the Co²⁺/Co³⁺ redox peak around 1.15 V vs. RHE is clearly observed only for the 26Co₃O₄ISACFe₃O₄ composite. Neither the Co²⁺/Co³⁺ nor Co³⁺/Co⁴⁺ redox couple is observed for 32Co₃O₄ISACWC composite, and one can note that the shape of the CV significantly differs from those of the other composites. Indeed, the pseudocapacitive behavior is not observed above 1.2 V vs. RHE, and oxidative currents are detected on both the positive and the negative CV scans, which agrees with the oxidation of WC at anodic potential, as previously reported by Göhl et al. [18] and Weidman et al. [36].

The charge corresponding to the redox peaks of the Co³⁺/Co⁴⁺ couple increases in the following series: Co₃O₄AC < 29Co₃O₄ISACVulcan < 26Co₃O₄ISACFe₃O₄ < 16Co₃O₄ISACBDD. The charge of the cathodic peak was used (due to difficulties to separate the anodic OER from pseudocapacitive currents), estimated to 0.01, 24.4, 10.0, and 4.8 C g⁻¹_oxide for Co₃O₄AC, 16Co₃O₄ISACBDD, 26Co₃O₄ISACFe₃O₄, and 29Co₃O₄ISACVulcan, respectively, and summarized in Table 1. The increase in the charge of the Co³⁺/Co⁴⁺ redox peaks observed for the ISAC composites, compared to the unsupported Co₃O₄, suggests that the addition of support increases the accessible sites of Co₃O₄ nanoparticles on the surface of ISAC composites. Previously, we explained this trend by a superior dispersion, an inferior agglomeration area, and an improvement of the charge transport between the Co₃O₄ particles and the current collector [22].

3.2.3. Electrochemical Activity of Composite Materials

The RRDE method was used to investigate the electrochemical activity of Co₃O₄ composites in Figure 5, to separate the support degradation and the OER currents. In Figure 5C,D, one can observe the ring currents and the oxygen faradaic efficiency, respectively, of ISAC composites associated to the disc currents shown in Figure 5A,B. The current of OER observed at the disc (Figure 5A) is supported by the appearance of the current of ORR at the ring (Figure 5C). We observe that the currents of the ring decrease in the following series: 29Co₃O₄ISACVulcan > 26Co₃O₄ISACFe₃O₄ > 16Co₃O₄ISACBDD > 32Co₃O₄ISACWCcalc > Co₃O₄AC, suggesting decrease of the OER activity in this series. Meanwhile, for Co₃O₄AC, both the currents of the disc and ring are much lower, confirming its lower OER activity. The oxygen F.E. can be calculated with reasonable accuracy at potentials above ca. 1.46 and 1.52 V (Figure 5D) for the Co₃O₄ composites and Co₃O₄AC, respectively. The OER “onsets” (Figure 5D) increase similarly to the ring currents, i.e., 29Co₃O₄ISACVulcan > 26Co₃O₄ISACFe₃O₄ > 16Co₃O₄ISACBDD > 32Co₃O₄ISACWCcalc > Co₃O₄AC. The 32Co₃O₄ISACWCcalc composite shows disc currents higher than those of 26Co₃O₄ISACFe₃O₄ and 16Co₃O₄ISACBDD but much lower ring currents and hence a lower faradaic efficiency. This apparent discrepancy can be attributed to oxidation of WC in the OER potential interval, as previously observed [18,36]. This observation also highlights the fact that it may not be appropriate to access the OER activity from disc currents alone.

We have determined that the potentiostatic mode is the most appropriate method for estimating the faradaic efficiency of the OER using the RRDE technique. This is since the fact that the slow diffusion of oxygen in the pores of the catalytic layer can result in a delay in the detection of oxygen at the ring, which can compromise the reliability of the faradaic efficiency measurement [23]. Thus, RRDE transients at a constant disc potential of 1.58 V vs. RHE for the unsupported Co₃O₄AC and for Co₃O₄ composites are compared in Figure 6 (note that a suitable catalyst loading was used to decrease current and avoid oxygen bubble formation). As demonstrated in Figure 6A,B, the currents of the disc and the ring of the Co₃O₄AC sample are relatively low. This is in agreement with the low cobalt redox peak charge. However, the currents (disc and the ring) and the F.E. increase for composites significantly (see Figure 6C, which is coherent with Figure 5D). The faradaic efficiency for the 16Co₃O₄ISACBDD, 26Co₃O₄ISACFe₃O₄, and 29Co₃O₄ISACVulcan composites with values close to 90–98% is the highest (this difference could be considered within the accuracy of the measurements), suggesting that these composites are stable against oxidative degradation, at least for RRDE transients at the disk potential of 1.58 V vs. RHE and/or under short-term polarization conditions. High faradaic efficiency observed for 26Co₃O₄ISACFe₃O₄ and 29Co₃O₄ISACVulcan may seem surprising considering the Fe₃O₄ and Vulcan XC72 degradation currents discussed in Section 3.2.1. This related inconsistency may be explained by (i) the different conditions such as the polarization duration, i.e., only 10 min in Figure 6 (longer polarization duration, under industrially relevant electrolysis conditions, are required to better evaluate the electrocatalyst’s potential for electrolysis applications), (ii) the slow degradation kinetics observed for the Fe₃O₄ and Vulcan XC72 supports at 1.58 V vs. RHE (see Figure S4 in Supporting Information, for Fe₃O₄ support equal to about 10 µA cm⁻² and for Vulcan XC72 support of 35 µA cm⁻²), and (iii) a “protecting effect” of Co₃O₄ nanoparticles deposited on the Fe₃O₄ and Vulcan supports. Regarding the latter, in several papers, authors reported an improved carbon durability due to the presence of oxide in the OER oxide–carbon composite [5,42,43]. For example, we studied a composite OER catalyst consisting of a perovskite oxide mixed with carbon (Sibunit-152, Vulcan XC-72R, and acetylene black) and discovered that the oxide component protected the carbon counterpart [5]. In [42], the authors show that the addition of Ni perovskite to the nitrogen-doped carbon nanotubes (N-CNTs) improves the galvanostatic stability for the OER by almost two orders of magnitude. The nanoscale geometries of the perovskites and the CNTs enhance the number of metal–support and charge transfer interactions and thus the activity. Finally, in [43], the authors explained that, in the literature, high oxygen evolution reaction (OER) active electrocatalysts often exhibit improved OER durability in the presence of carbon oxidation reaction (COR). The coincidence of activity and durability in OER electrocatalysts is theoretically understood as preferential depolarization in galvanostatic situations. For a system involving multiple independent and competitive reactions at constant-current conditions, the overpotential is most dominantly determined by the most facile reaction, meaning that the most facile reaction is responsible for a dominant portion of the overall current. Therefore, higher OER activity improves durability by mitigating the current responsible for COR. This coincidence was then proven experimentally by comparing two catalysts with the same chemical identity of different dimensions (5 and 100 nm). The smaller-dimension catalyst, having a greater number of active sites per fixed mass, was found to be more durable due to the decrease in carbon corrosion.

For 32Co₃O₄ISACWCcalc composite the disc and ring currents are much smaller, suggesting a lower OER activity. Moreover, we can note that in Figure 6C the faradaic efficiency value is low at short times (~20%, which is similar of the one of Figure 5D) and then increases with time and reaches a maximum of 70%, suggesting a transition phenomenon with slow kinetics possibly due to the WC degradation. Moreover, we have to underline that the faradaic efficiency of all materials may be underestimated due to the involvement of the underlying GC surface (cf. Section 3.2.1). This explains the low faradaic efficiency of the unsupported Co₃O₄ sample.

Table 2. Summary of the Co₃O₄ oxygen evolution activities obtained in this work and compared with literature data.

Material	Designation (Oxide: Support wt Ratio)	Oxide Loading/µg cm−2geo	Media	ECSA */m2 g−1	Tafel Slope/ mV dec−1	Mass-Normalized OER Activity at 1.58 V RHE/ A g−1oxide	Ref.
Co3O4AC	Unsupported Co3O4 nanoparticles	15	1 M NaOH	~5 × 10−3 *	119	1.3	this work and [22]
16Co3O4ISACBDD	BDD-supported Co3O4 nanoparticles (16:84)	15	1 M NaOH	13 *	65	14.2 [12.6]	this work and [22]
29Co3O4ISACVulcan	Vulcan XC72 carbon-supported Co3O4 nanoparticles (29:71)	15	1 M NaOH	2.4 *	59	55.7 [54.6]	this work and [22]
26Co3O4ISACFe3O4	Fe3O4-supported Co3O4 nanoparticles (26:74)	15	1 M NaOH	4.9 *	48	24.8 [23.3]	this work
32Co3O4ISACWCcalc	WC-supported Co3O4 nanoparticles (32:68)	15	1 M NaOH	-	134	19.7 [13.4]	this work
Fe3O4@CoFe2O4	Core-shell Fe3O4@CoFe2O4 nanoparticles	1.43	0.1 M NaOH	-	63 ± 2	28.8	[18]
16Co3O4ISACSibunit152	Sibunit-supported Co3O4 nanoparticles (16:84)	15	1 M NaOH	24 *	80	28.1 [18.5]	[22]
Co3O4	Acetylene Black- supported mesoporous Co3O4	380	1 M NaOH		161	3.5	[2]
Co3O4 (a)	Carbon fiber paper-supportedCo3O4 nanoparticles	300	1 M KOH		61.8	11	[37]
Co3O4 (b)	Co3O4 impregnated on the graphene	127.4	0.1 M KOH		147	19.6	[3]
Co3O4 (c)	Co3O4 post coated on carbon cloth	8100	1M KOH		90	0.4	[4]
Co3O4(111) (d)	Co3O4 film electrodeposited on Au(111)	15	0.1 M NaOH		65	-	[38]

* This estimation value has been calculated based on the charge of the Co⁴⁺/Co³⁺ reduction. In the square bracket: activity corrected of the support degradation. Data are given at 25 °C except for (a) at RT, (c) unknown, and (d) at RT, and potentials are iR-corrected except for (a), (b), and (c), where potentials are not iR-corrected.

shows the OER activities, normalized to the mass obtained at 1.58 V vs. RHE, of the synthesized materials of this work and of the materials selected from the literature for comparison. Table 2 includes also the corresponding Tafel slopes (for this work, the reader can refer to our previous publication [22] for the procedure used to determine the Tafel slopes). Note that the activity and the Tafel slopes in Table 2 are not consistent with each other. This may be due to the divergent experimental conditions (loading of catalyst, thickness of the catalytic layer, addition of conductive additive, electrolyte, etc.) and to the material morphology or structure (e.g., crystallography of the surface, defects, etc.) of the nanoparticles of Co₃O₄. For example, in this work, the activity per mass of the composites is similar to that of a core-shell Fe₃O₄@CoFe₂O₄ nanoparticles [20] and higher than the activity of mesoporous Co₃O₄ supported in acetylene black [2], the Co₃O₄ impregnated on the graphene [3] and Co₃O₄ coated on carbon cloth [4]. One may notice that contrary to this work, in many publications, the activity of OER is only determined using the disk current, and so potential support and oxide degradation (if any) are neglected of the activity of OER.

In Table 2, the Tafel slope values vary from ~48 to 160 mV dec⁻¹. For ISAC composites, the smaller Tafel slope (~50 mV dec⁻¹) corroborates ohmic limitations in the film and lack of mass transport, as explained in our previous work [22]. Additionally, the highest Tafel slopes, observed for 32Co₃O₄ISACWCcalc and Co₃O₄AC, result from the lack of electronic conductivity due to ohmic limitation in the film, as well as with mesoporous Co₃O₄ supported in acetylene black and Co₃O₄ impregnated on graphene [2,3]. We can note that the high value of Tafel slope observed for 32Co₃O₄ISACWCcalc (134 mV dec⁻¹) is consistent with the degradation of WC.

Table 2 also includes the electrochemical active surface area (ECSA). ECSA values were estimated considering the cathodic peak of cobalt at 1.45 V vs. RHE of cyclic voltammograms in Figure 5B, the mean surface density of Co cations estimated from crystallographic data of Co₃O₄ spinel of 1.27 × 10¹⁵ cm⁻² (see Table S2 in the Supplementary Information), and the electron charge of 1.6 × 10⁻¹⁹ C. The ECSA calculation is explained in [22].

3.2.4. Stability Tests of ISAC Composites

Considering the previous results, a three-hour stability test of 29Co₃O₄ISACVulcan, 26Co₃O₄ISACFe₃O₄, and 16Co₃O₄ISACBDD was performed under a high constant potential of 1.66 V vs. RHE. Figure 7 shows CVs measured on the disc (black curves) and after 3 h of polarization (red curves) of 29Co₃O₄ISACVulcan (A), 26Co₃O₄ISACFe₃O₄ (B), and 16Co₃O₄ISACBDD (C). In Figure S6 in the Supplementary Information, the corresponding ring currents are presented. First, we observe a good reproducibility of the disc and ring currents for Fe₃O₄ (Figure 7B and Figure S6B) and BDD-based composites (Figure 7C and Figure S6C). As mentioned above, the observed stability of the Fe₃O₄ composite may be tentatively assigned to a “protecting effect” (see Section 3.2.3) of Co₃O₄ or by the slow degradation kinetics of the Fe₃O₄ supports observed at potentials less than 1.68 V vs. RHE (see Figure S4 in the Supporting Information, for Fe₃O₄ support equal to about 60 µA cm⁻²). Finally, we observed a significant loss in both the disc and ring currents (Figure 7A and Figure S6A) for the Vulcan XC72-based composite after three hours of the polarization test at 1.66 V vs. RHE. It shows that (i) the degradation of the Vulcan XC72 carbon support at 1.66 vs. RHE, and (ii) there is no “protecting effect” of Co₃O₄ oxide in these experimental conditions (i.e., 3 h at 1.66 V vs. RHE).

As WC degradation had previously been observed in this study, a simple 30 min stability test was performed on the 32Co₃O₄ISACWCcalc composite at a lower potential of 1.58 V vs. RHE. Figure S7 in the Supplementary Information shows a significant loss in the disc and ring currents, confirming the degradation of the WC-based composite in OER conditions.

However, long-term durability studies under industrially relevant electrolysis conditions, such as higher temperatures and the use of polymer rather than liquid electrolytes, are required to better evaluate the electrocatalyst’s potential for electrolysis applications.

4. Conclusions

To conclude, tungsten carbide (WC) and iron oxide with spinel structure (Fe₃O₄) were investigated as possible substitutes for carbon support in transition metal oxide composites for the oxygen evolution reaction and compared with recent results obtained for BDD and Vulcan XC72. To do so, we prepared composites of WC and Fe₃O₄, with Co₃O₄ (active for the OER) by the in situ autocombustion (ISAC) route. We observed that (i) the nanoparticles of Co₃O₄, of ca. 5–10 nm, are well supported on WC and Fe₃O₄, and (ii) the deposition of Co₃O₄ nanoparticles on support significantly increases the number of accessible sites of spinel surface and the increase in the charge of the Co⁴⁺/Co³⁺ redox peak.

The stability of support materials (alone) and the corresponding Co₃O₄-based composites and their OER activity were studied with the rotating ring–disc electrode (RRDE), using the ring electrode as an oxygen sensor. We demonstrated the importance of using RRDE method to discriminate between the OER and the electrochemical degradation currents.

Investigation of the degradation of support at anodic potential confirms that the stability of Fe₃O₄ support against oxidation is better than the one of Vulcan XC72 carbon at potential below 1.68 V vs. RHE, but lower than the one of BDD. In contrast, WC support does not seem to show much advantage over Vulcan XC72 in terms of its anodic stability until the applied potential is increased up to ~1.9 V vs. RHE.

A significant improvement in the mass-normalized OER activity of the ISAC composites is demonstrated compared to the unsupported Co₃O₄. Then the Co₃O₄ alternative support composites show OER activity comparable or even superior to the literature data for Co₃O₄.

Moreover, three-hour stability tests at 1.66 V vs. RHE confirm the higher stability of BDD- and Fe₃O₄-based ISAC composites; whereas, the composites based on Vulcan XC72 carbon and WC (for this latter, a 30 min stability test at 1.58 V vs. RHE was performed) suffers from support degradation.