New insights into evaluating catalyst activity and stability for oxygen evolution reactions in alkaline media

Abstract

The growing development of oxygen-evolution-reaction (OER) catalysts has increased the demand for robust evaluation protocols to identify promising candidates. In this work, we successfully establish an effective experimental protocol to quantify the intrinsic performance of OER catalysts in a standard thin-film rotating-(ring)-disk-electrode system. Furthermore, a new insight into the electric double layer (EDL) microstructure is gained to illustrate all experimental observations in alkaline media. The results indicate that IrO₂ has 2.6-times higher activity than NiCo₂O₄, due to the improved reaction kinetics, charge transfer and O₂-transport occurring in the EDL. However, after 10 000 cycles of cyclic voltammetry (CV), NiCo₂O₄ shows much higher stability due to continuous EDL structural modifications. In particular, it is revealed that the EDL charging process, accompanied by the electrocatalytic OER, has a strong effect on catalyst–electrolyte interfacial behaviors, which, in turn, determine the accuracy of electrochemical measurement results. It is further found that the electrochemical surface area (ECSA) obtained from traditional CV measured capacitance cannot accurately reflect the actual reaction sites of various catalysts during the OER. This discrepancy exists mainly because of the EDL capacitance, which is referred to as the inner capacitance caused by less accessible active sites. Alternatively, a novel method based on in situ electrochemical impedance spectroscopy has demonstrated improved accuracy in obtaining potential-dependent ECSAs, comparable to ex situ Brunauer–Emmett–Teller surface area measurements.

---

Introduction

The oxygen evolution reaction (OER) and hydrogen evolution reaction (HER) are pair electrochemical reactions for splitting water to generate oxygen and hydrogen, respectively. Compared with the HER, the OER suffers from major overpotential losses and thus has been the focus of extensive research efforts.^1–5 Currently, among the major limitations for commercial application of proton exchange membrane water electrolysis are the usage of noble metal catalysts (such as IrO₂ and RuO₂) and their low performance.^2,6,7 When the OER occurs in an alkaline medium, the loading of noble metal catalysts can be significantly reduced or completely removed. Over the past decade, significant progress in alkaline anion exchange membrane (AAEM) based techniques for water electrolysis has further increased the demand for more advanced OER catalysts.^6–9 Unfortunately, reliable protocols to accurately evaluate catalysts are still lacking, which poses a challenge in quantifying and comparing the catalyst active area and performance.¹⁰ Common ex situ tools used in classical heterogeneous catalysis, such as the Brunauer–Emmett–Teller (BET) method and X-ray photoelectron spectra, are not credible due to the overlapping of electric double layers (EDLs) and blockage of active sites by the reactant, product, intermediates and/or supporting inert electrolyte.^11–14 Furthermore, the EDL microstructure is constantly changed due to oxidation/dissolution of the oxide film during the OER.^11–14 Moreover, in situ methods for evaluating the active catalyst area also pose significant challenges due to (i) unstable reaction interfaces arising from the high potential and strong oxidizing conditions, (ii) a multistep electron-transfer process with various intermediates, and (iii) varied transport behaviors depending on operating conditions and catalyst materials.¹⁵

The primary goals of electrocatalytic research are to understand the interfacial behaviors of both faradaic electron-transfer reactions and non-faradaic charging processes in the EDL.¹⁶ One approach to achieving these goals is to develop state-of-the-art testing and diagnostic methods to explore the nature of the EDL. In general, major indicators for the OER should involve the redox couple potential, EDL capacitance, kinetic overpotential, Tafel slope, interfacial transport, polarization resistance, etc. When studying OER catalysts, an essential issue is how to precisely determine the electrochemical surface area (ECSA), which is currently known to be related to EDL capacitance.^17,18 Cyclic voltammetry (CV) is the most popular way to determine the EDL capacitance of catalysts. This approach assumes that EDL charging is the only process in the studied potential range and no faradaic reaction takes place.^17,19–22 However, for oxide materials, the dependence of capacitance on the applied potential and testing conditions, such as electrolyte pH and electron conductivity, is not clear, so the resulting error might be very large.¹⁷

In addition to initial catalyst activity, the stability of OER catalysts is equally important for commercial applications.^1,14,23 Nonetheless, there are three key issues related to stability that are rarely addressed: (i) how to establish standardized testing protocols that can be used for assessing the stability of various OER catalysts, (ii) how to precisely quantify the performance change after stability tests, and (iii) how to understand the fundamental relationship between catalyst activity and stability.

Since Stonehart and Ross reported their novel work in 1976,²⁴ thin film rotating disk (TF-RDE) and rotating ring-disk electrode (TF-RRDE) approaches in half cell testing have gained significant popularity and, to date, become the most widely used methods for electrocatalysis study.^25,26 Compared to full cell testing, TF-RDE/RRDE does not require sophisticated skills and equipment to evaluate the electrocatalytic properties of newly developed materials, and provides a reliable means to study reaction kinetics and mechanisms in a well-controlled environment.^26,27 In addition, the advantages of TF-RDE/RRDE include (i) quick sample evaluation, (ii) minimum catalyst quantity required, (iii) minimal influence of ohmic and mass diffusion losses, and (iv) facile electrode fabrication.

Researchers from a wide variety of backgrounds are devoting their attention to developing catalytic materials with a lower noble metal content and/or a more efficient utilization of the existing noble metal content. In alkaline media, IrO₂ and Ni/Co based oxides have emerged as the most commonly studied electrocatalysts, due to their potential for OER application.^9,21,28–30 However, there is a lack of benchmarking protocols, so the valid data in the open literature are highly scattered and performance comparison between various materials is difficult without a baseline material.^2,30–33

Herein, we report our efforts to establish a benchmarking protocol to evaluate and compare the intrinsic performance of IrO₂ and Ni/Co based materials in an alkaline medium. Based on TF-RDE observations, IrO₂ shows higher OER activity but much lower stability than NiCo₂O₄. The newly developed measurement and diagnostic procedures in this work enable the difference in electrocatalytic behavior between the two materials to be explained in terms of EDL effects and resistance to OH⁻-, electron- and O₂-transport at the reaction interface. Particularly, an in situ impedance-based method is developed to calculate the ECSA, which is related to the wide activity of OER catalysts.

Results and discussion

Reaction interface and benchmarking protocols

Our primary goal is to benchmark and compare the performance of various OER catalysts in alkaline media. A typical TF-RDE/RRDE system, shown in Fig. 1(a), was used to conduct the major electrochemical measurements. The potential calibration of the reference electrode (RE) against the reversible hydrogen electrode (RHE) was performed by CV in a H₂-saturated 1 M KOH solution at room temperature. The calibration results are shown in Fig. S1 of the ESI.† Accordingly, all measured potentials in this work are reported relative to the RHE, which allows us to directly evaluate the OER overpotential independently of electrolyte pH.³⁴

Fig. 1 Schematic illustration of the electrocatalytic OER in alkaline media. (a) TF-RDE/RRDE configuration, (b) reaction interface, (c) electrical double layer (EDL) in the OER process and (d) EDL in the charging process.

A schematic representation of the reaction interface and EDL is shown in Fig. 1(b–d). At the applied OER potential, the active catalyst layer (CL) surface has excess positive charge that can enhance the adsorption of OH⁻ ions through coulombic attraction. The major function of OER electrocatalysts is to facilitate the faradaic transition from OH⁻ ions to electrons, water and O₂ gas. The measured reaction rate is directly related to the number of active electrocatalyst sites, and the OH⁻-, electron- and O₂-transport resistances in the EDL.^11,35 The process of manufacturing the catalyst ink also has a significant impact on the electrode structure, which, in turn, determines the accessibility, stability and number of EDLs. To maximize catalyst utilization, the deposition of a thin film onto the conductive substrate is required to produce a uniformly distributed CL. Adding the Nafion ionomer to the CL can extend the reaction interface and serve as a catalyst binder.^32,36,37 Furthermore, it is found that the zeta potential changes from −6.2 mV to −25.5 mV with increasing the ionomer to IrO₂ catalyst weight ratio from 0 to 0.5. In principle, the measured zeta potential can be used to indicate the degree of electrostatic repulsion/attraction between particles in dispersion solution.³⁸ The increase of absolute zeta potential demonstrates that adding Nafion improves ink stability and prevents large particle aggregation.

The interaction of adsorbed OH⁻ ions with the charged interface forms an EDL, which is the location where the electrocatalytic OER occurs.¹⁶ Its thickness is generally on the order of several nanometers and a potential difference is generated across this thin layer.¹⁶ As illustrated in Fig. 1(c and d), the EDL is commonly known to consist of three layers: an inner Helmholtz plane (IHP), an outer Helmholtz plane and a diffusion layer.^11,16,39 To maintain the rate of the OER, OH⁻ ions should be supplied continuously to the IHP where electrons are transferred to the solid catalyst surface and the O₂ produced diffuses into the bulk electrolyte solution. The transport resistance within the EDL should be minimized in the design of a high performance catalyst to allow efficient supply and removal of reactants and products. The EDL acts as a dielectric capacitor, which has the ability to accumulate aqueous OH⁻ ions when an electric force is applied. It is worth pointing out that the capacitive current increases with the potential scan rate and cannot be ignored, especially at low faradaic reaction rates.¹⁶ Compared with pure capacitor charging processes, the presence of the OER complicates the EDL behavior because of uncertainties in O₂ coverage, metal surface oxidation, reaction intermediate (MOH, MO and MOOH) adsorption, EDL microstructure reconstruction, etc.

To characterize the activity, reaction mechanisms and long-term stability of OER catalysts, the proposed benchmarking protocol includes typical experiments using RDE, RRDE, and alkaline aqueous electrolyzer systems. The typical testing methods employed, and their corresponding experimental parameters, are presented in Fig. 2. For each method, proper design of experimental parameters is essential to producing accurate and reliable results. In this work, all measurements were conducted in 1 M KOH due to its high OH⁻ activity.⁴⁰ Before the TF-RDE/RRDE characterization, the material durability was evaluated using a gas diffusion electrode in an alkaline aqueous electrolyzer system, as shown in Fig. S2.† During 1000 hours of continuous operation at 10/50 mA cm⁻², the measured cell voltage, shown in Fig. 3, remained stable, with minimum loss detected for both the IrO₂ and NiCo₂O₄ catalysts.

Fig. 2 Benchmark protocol for evaluating OER electrocatalysts in alkaline media.

Fig. 3 Durability tests of catalysts in alkaline electrolysis cells using (a) IrO₂ and (b) NiCo₂O₄ as OER catalysts (2.5 mg cm⁻², ionomer/catalyst = 1/2, by weight). Testing conditions: 10/50 mA cm⁻², 1 M KOH, magnetic stirring (200 rpm) and room temperature.

Electrocatalyst activity

Experimental tests were performed using a TF-RDE to quantitatively study the electrocatalytic behaviors of IrO₂ and NiCo₂O₄ in alkaline media. In cyclic voltammetry, the electrode potential is swept linearly between a lower and an upper potential, and the voltammograms obtained are often referred to as the unique electrochemical “fingerprint” spectra of various materials.¹⁶ In our measurements, an upper potential of 1.4 V was selected to prevent the OER and catalyst dissolution, while a lower potential of 0.4 V was selected to prevent any unexpected reduction reactions, such as the HER. A scan rate of 50 mV s⁻¹ was selected to both efficiently avoid the limitation of mass transport in the EDL and reduce the time consumption for conducting such measurements. Since this is a non-steady state technique, the electrodes should be immobile and sit in the unstirred solution to maintain a controlled double layer interface. In this case, species diffusion is the only contributor to mass transport.¹⁶

The current densities as a function of potential are plotted in Fig. 4. The detected current mainly results from (i) EDL capacitive current due to the charging/discharging process and (ii) faradaic pseudocapacitive current due to the solid-state redox reactions.^41,42 In a potential window of 0.8–1.0 V, the CV shape of each studied material is close to a rectangular shape, confirming the EDL capacitive behavior. The overall CV curve of IrO₂ appears more symmetric than that of NiCo₂O₄, indicating more reversible electrochemical behavior both in the EDL charging process and faradaic redox reaction due to lower polarization, ohmic and mass transport resistance.

Fig. 4 CV curves at a scan rate of 50 mV s⁻¹ in the potential range of 0.4–1.4 V.

Another important feature observed from CV is related to the pseudocapacitive behavior due to faradaic redox reactions of active elements. The characteristic curve of IrO₂ exhibits broad oxidation peaks at around 0.65 V and 1.05 V, corresponding to the redox couples of Ir^III/Ir^IV and Ir^IV/Ir^V (eqn (1)), respectively. In the case of Ni–Co oxide, only one well-defined oxidation/reduction peak is observed at ca. 1.30 V, which is attributed to the redox reactions of Co^III/Co^IV as well as M^II/M^III transitions (where M represents Ni or Co) at the reaction interface,^43,44 as described in eqn (2) and (3), respectively. The potentials of these redox peaks are so close that a broader overlapping peak appears in the CV pattern.^43,44 Furthermore, it is found that the current density of NiCo₂O₄ increases drastically when the redox reactions occur. This observation points to a transition from a non-conducting species of Ni^II (or Co^II) to a conducting species of Ni^III (or Co^III). The formation of Ni^II (or Co^II) thus suppresses the EDL capacitive behavior, as reported by Costentin and coworkers.⁴⁵

IrO_a(OH)_b + δOH⁻ ↔ IrO_a(OH)_(b+δ) + δe⁻ (1)

CoOOH + OH⁻ ↔ CoO₂ + H₂O + e⁻ (2)

NiCo₂O₄ + OH⁻ + H₂O ↔ NiOOH + 2CoOOH + e⁻ (3)

To further identify the pseudocapacitive characteristics of NiCo₂O₄, additional CV measurements were performed in an extended potential window of 0.8–1.45 V with increasing scan rates. Fig. 5 shows the relationship of current density (j) and potential (E) in the forms of j, j/ν^0.5 and j/ν and their alterations with the scan rate (ν). It is assumed that j ∝ ν⁰, ν^0.5 and ν¹ can qualitatively indicate the kinetic control by catalytic reaction, diffusion-controlled redox reaction and EDL charging capacitance, respectively.⁴⁵ At the CV positive/negative peak position, the relationship of j ∝ ν^0.5, in Fig. 5(b), is confirmed at a relatively high scan rate, substantiating the pseudocapacitive behavior of NiCo₂O₄. In the first few CV scans (5–100 mV s⁻¹), the increased value of j/ν^0.5 with the scan rate can be ascribed to the increasing oxyhydroxide coverage and EDL reconstruction.^1,46–48 However, Fig. 5(b) reveals that the pseudocapacitance decreases by 75% when the scan rate increases from 5 mV s⁻¹ to 300 mV s⁻¹, confirming the slow charge transfer process in the bulk material of NiCo₂O₄. At high scan rates, there is shorter time to respond to the rapid potential perturbation. For NiCo₂O₄, the high charge transfer resistance increases the difficulty for the redox reaction of solid-state catalysts to happen. As a result, only the outer surface is accessible for the reaction. In addition, the observed redox peak potential shifts with increasing scan rate; this demonstrates that polarization and transport resistance exist in the EDL during the potential scanning process.

Fig. 5 CV curves of NiCo₂O₄ in the 0.8–1.45 V potential window at various scan rates (ν) in 1 M KOH. The current density (j) is normalized by (a) ν⁰, (b) ν^0.5 and (c) ν¹, respectively.

To quantify the OER activity, potential-controlled LSV tests were conducted in the TF-RDE system. The characteristics of the LSV curve depend on several key parameters, including the applied potential window, the scan rate, and the rotating speed of the working electrode (WE). Measurements began at 1.3 V, which is much lower than the onset potential of the OER. The final potential was set to 1.65 V, where mass transport is not yet a key factor for determining the reaction rate. Note that the potential scan rate (2 mV s⁻¹) is much lower than that of the above fingerprint CV, since the non-faradaic capacitance current has a linear relationship with the scan rate. Therefore, the EDLs are closer to steady-state conditions, which limits the contribution of the EDL capacitive current. The electrode rotating speed was set to 1600 rpm, as reported in most literature studies, which facilitates O₂ removal and OH⁻ transport in EDLs.^8,19,30 To characterize the influence of the WE substrate, the bared disk electrode was also examined using the same technique. As shown in Fig. 6(a), the current density of the bare RDE is negligible compared to that of the catalyst-loaded electrodes.

Fig. 6 Evaluation of catalyst activity. (a) LSV curves at a scan rate of 2 mV s⁻¹ and (b) IR-corrected Tafel plots. Symbols – raw data; lines – linear fit to the data.

By integrating LSV results within the measured potential window,⁴⁹ it was found that the IrO₂ activity is around 2.6-times higher than that of NiCo₂O₄. At a current density of 10 mA cm⁻², at which performance comparisons of a wide variety of catalysts have been commonly reported in the literature,^9,12,19 the overpotentials without ohmic resistance correction were 300 and 350 mV, for IrO₂ and NiCo₂O₄, respectively. These values agree with previously reported results.^19,37 The specific and mass activity of IrO₂ (NiCo₂O₄) at 1.6 V were about 39.0 (15.1) mA cm⁻² and 162.5 (62.9) A g⁻¹, respectively. It is important to note that the mass activity is strongly related to catalyst utilization. Our research further demonstrates that effective utilization of active species decreases as the catalyst loading increases (Fig. S3†). IrO₂ with 0.08 mg cm⁻² loading had around 3 times the mass activity of the same material with 0.72 mg cm⁻² loading. It is highly possible that the combined effects of mass diffusion and EDL resulted in lower catalyst utilization at higher loading. Loading for evaluating novel OER catalysts typically ranges from 0.2 to 0.3 mg cm⁻², as reported in the literature.^50–52 Therefore, a loading of 0.24 mg cm⁻² was selected for all TF-RDE and TF-RRDE measurements in this work to compare the performance of different catalysts.

To gain insights into the kinetics of the OER processes of the various catalysts, LSV performance curves were plotted as the IR-free potential vs. log(current density) as shown in Fig. 5(b). Two distinct linear trends can be observed in the lower and higher overpotential regions. The slope is referred to as the Tafel slope, which can be calculated according to the following equation:

η = b × log(j) + a (4)

where η is the IR-free potential, b is the Tafel slope, j is the current density and a is a constant.

To determine the appropriate potential region for Tafel fitting, it is important to ensure that the current densities result from faradaic kinetic reactions. Both high potentials, at which oxygen bubbles block the active sites, and low potentials, at which the non-faradaic capacitive current results in an obvious contribution to the detected current, should be avoided.³⁰ Prior to fitting, the potentials of each studied sample were corrected for ohmic resistance, which was obtained using EIS at the open circuit potential (OCP, 0 V vs. the RHE). In the TF-RDE system, the major contribution of ohmic resistance came from the bulk electrolyte solution and the distance between the RE and the WE. Nevertheless, a slight difference due to catalyst electrical resistance was observed between IrO₂ and NiCo₂O₄ catalysts, as shown in Fig. S4.† The fitted ohmic resistances are 0.951 and 1.002 Ω cm² for IrO₂ and NiCo₂O₄, respectively, suggesting a lower electrical resistance of IrO₂. Actually, IrO₂ possesses metallic conductivity with an electrical resistivity of 6 × 10⁻⁵ Ω cm (for the bulk single crystal at 300 K), which is much smaller than that of the transition metal oxide, NiCo₂O₄ (∼0.016 Ω cm).^42,53

It is well known that the OER is a 4-electron transfer reaction, which involves a sequence of steps and many intermediates, such as MO, MOOH or physisorbed peroxide species.^1,15,54 Over the past several decades, various possible mechanisms have been proposed with notable early work. One of the most famous mechanisms is Krasil'shchikov's Path, which is described by eqn (5)–(8) with the corresponding Tafel slopes.⁵⁵

M + OH⁻ ↔ MOH + e⁻, b = 120 mV dec⁻¹ (5)

MOH + OH⁻ ↔ MO⁻ + H₂O, b = 60 mV dec⁻¹ (6)

MO⁻ → MO + e⁻, b = 45 mV dec⁻¹ (7)

2MO → 2M + O₂, b = 19 mV dec⁻¹ (8)

Based on the fitted Tafel slope, the rate-determining step (RDS) can be deduced reasonably. As shown in Fig. 6(b), at the near onset potential for the OER, the curves deviate from linearity due to the EDL charging current. The major difference between the two Tafel slopes appears in the low current density region, where the values are 48 and 59 mV dec⁻¹ for IrO₂ and NiCo₂O₄, corresponding to the RDS shown in eqn (7) and (6), respectively. The lower Tafel slope of IrO₂ indicates reduced kinetic overpotential losses. On the one hand, this can be attributed to the increase of bond strength for OH⁻ adsorption on IrO₂, which accelerates the rate of the first electron reaction step shown in eqn (5) and, accordingly, improves electrocatalytic kinetics.¹ On the other hand, the change in Tafel slopes can reflect the concentration of active sites and their contribution. For IrO₂, the increase of active sites and their contribution may be the second reason for the reduced Tafel slope.

At higher current densities, the Tafel slopes changed to 80 mV dec⁻¹ for both materials. This increase of Tafel slope can be attributed to four mechanisms: (i) increased mass transport resistances of OH⁻, electrons and/or O₂, (ii) change in the RDS, (iii) adsorption of the reaction intermediates, and (iv) change of active sites and their contribution due to EDL reconstruction. The produced O₂ bubbles blocking the available reaction interface will produce extra mass transport resistance, which prevents OH⁻ adsorption and electron transport in the EDLs. To efficiently remove O₂ bubbles and improve mass transport, a common practice is to increase the rotating speed of the WE. The data in Fig. S5(a)† indicate that the OER performance was slightly enhanced by increasing the rotating speed. Upon further analysis (results shown in Fig. S5(b)†), the Tafel slopes in the low (2–10 mA cm⁻²) and high (30–55 mA cm⁻²) current density regions decreased slightly as the rotating speed increased, except for that at 500 rpm at high current density. This confirms that, at a baseline rotating speed of 1600 rpm, the mass transport resistance caused by O₂ bubbles was minimized.

ECSA evaluation via CV

Accurately quantifying the ECSA is important for determining catalyst activity, since the morphology-dependent OER predominantly occurs at the active surface. For an ideal compact oxide catalyst without a surface hydroxide phase, the surface area can be estimated from low-temperature N₂ adsorption–desorption isotherms. Using the BET method, the results show that the surface area of IrO₂ (32.5 mg cm⁻²) is 1.3 times higher than that of NiCo₂O₄ (24.7 mg cm⁻²). However, BET results cannot directly be used to determine the ECSA available for the OER.¹² The discrepancy is often attributed to the blocking of pores, adsorbing of H₂O molecules and hydrated OH⁻ anions, and EDL reconstruction at the reaction interface during the OER.^12,17

Alternatively, it is commonly accepted that the ECSA has a linear relationship with the EDL capacitance caused by the interfacial charging process.¹⁸ For OER catalysts, the standard CV technique is often used in the literature to obtain surface capacitance.^19,37,56,57 Note that it is significantly important to separate the solid-state pseudocapacitive and EDL capacitive contributions, namely, no faradaic oxidation/reduction reaction exists for studying the EDL capacitive behavior. In this case, the measured current is ascribed only to the surface-controlled capacitive behaviour, which is independent of the potential, as illustrated in Fig. 1(d). The EDL potential range (0.8–1.0 V) within which the CV curve exhibits a rectangular shape was determined from the “fingerprint” spectra in Fig. 4. Representative CV data for both catalysts are shown as a function of scan rate in Fig. S6.† This capacitive response is closely related to the EDL active area, which can be quantified as an impedance for electron transfer and OH⁻ adsorption. The measured currents of IrO₂ at the center of the applied potential window remained constant in each positive/negative scan segment. This reveals that the EDL charging process is sufficiently fast and stable. In contrast, a continuous increase of NiCo₂O₄ current was observed in a positive CV scan especially at a higher scan rate, confirming that this material cannot quickly respond to the applied potential to form a steady EDL for surface charge accumulation. The possible reasons are related to the slow charge transport in the less accessible EDLs and/or the continuously varied EDL microstructure during the voltammetry process. Based on the CV method, the EDL capacitance (C) can be calculated from the following the equation,^19,58

where E₁ and E₂ are the cutoff potentials in the CV curves, i is the instantaneous EDL current, is the total voltammetric area obtained by integration of positive and negative sweeps in each CV curve, and ν is the potential scan rate.

The single-point current at 0.898 V for each scan rate is reported in Fig. 7(a and b). According to eqn (9), the EDL capacitance can be estimated by linearly fitting the slopes of the scan rate data below 75 mV s⁻¹. Taking the average of the absolute slope values from the anodic and cathodic sweeps, the resulting EDL capacitance values are 12.5 and 0.076 mF for IrO₂ and NiCo₂O₄, respectively. Moreover, it can be observed that the deviation from linearity increases with the scan rate, which can be ascribed to the inner capacitance (C_i) caused by slow OH⁻ penetration into the less accessible areas (such as micro-pores, corners, cracks and grain boundaries) and/or slow electron transfer. These capacitances cannot be fully measured at high scan rates, because the OH⁻ diffusion and/or electron transport timescales are slower than the potential changing rate.¹⁸ Furthermore, in comparison to IrO₂, NiCo₂O₄ shows more deviation from linearity at high scan rates. The pore structure of each catalyst sample was characterized using low-temperature N₂ adsorption/desorption measurements. As shown in Table S1,† NiCo₂O₄ has a much lower total pore volume (0.0767 cm³ g⁻¹) and smaller average pore diameter (18 nm) than IrO₂ (pore volume, 0.248 cm³ g⁻¹, and diameter, 30.2 nm). For these nanoscale pores, Knudson diffusion is the dominant transport mechanism and the diffusivity decreases with decreasing pore diameter.⁵⁹ Therefore, NiCo₂O₄ should have increased OH⁻ diffusion resistance compared to IrO₂. Furthermore, the lower electron conductivity shown in Fig. S4† suggests that NiCo₂O₄ with increased inner capacitance would require more time for each of the charging/discharging processes to reach a steady state. Unfortunately, this single-point current method does not provide quantitative measures of inner capacitance.

Fig. 7 Evaluation of catalyst capacitance using a CV method. Anodic (black solid square symbols) charging current and cathodic (red solid circular symbols) discharging current of (a) IrO₂ and (b) NiCo₂O₄ at 0.898 V plotted as a function of scan rate, and plots of calculated capacitance as a function of scan rate (l/C vs. ν^1/2) for (c) IrO₂ and (d) NiCo₂O₄. Symbols – raw data; lines – linear fit to the data.

Alternatively, the EDL capacitance can also be obtained by integrating the CV curve over the entire potential range for each scan rate using eqn (9). In order to quantify the total capacitance (C_t), the outer capacitance (C_o), and their difference, C_i, we have developed a thorough analysis procedure (refer to the ESI†). Compared with the above single-point current method, this procedure has several advantages, including the following: (i) the integration method utilizes the entire range of CV data, (ii) a single capacitance is obtained for each scan rate, and (iii) C is measured as a function of scan rate. In principle, because of the low EDL resistance of IrO₂, the linear fit should be avoided at low scan rates due to potential overlapping adsorption of OH⁻ ions in the same position as well as the adsorption occurring at the inner and less accessible sites for the OER. In contrast, only the experimental data at low scan rates were taken into account for NiCo₂O₄. At high scan rates, distortion of the CV curve for EDL measurement occurs due to high charge transport resistances.

As shown in Fig. 7(c and d), the calculated C_t values are 16.0 and 0.826 mF for IrO₂ and NiCo₂O₄, respectively. It is noteworthy that, using the integration method, the resulting C_t of NiCo₂O₄ is over 10 times larger than the capacitance obtained using the single-point current method. In addition, as the scan rate approaches infinity, it is assumed that OH⁻ ions can diffuse only to the outer accessible active area during the charging/discharging process, so the inner and less accessible active sites are excluded.^18,60 The values of C_o obtained by extrapolating the integrated C as a function of ν from the plot of C vs. ν^−1/2 are 6.71 and 0.0409 mF for IrO₂ and NiCo₂O₄, respectively, as given in Fig. S7.† The ability to transfer OH⁻ ions and electrons in the EDL can be expressed by the ratio of C_o/C_t.⁶⁰ This ratio is around 44% for IrO₂ and 5% for NiCo₂O₄. This evaluation is consistent with the fact that NiCo₂O₄ has a much lower pore volume, smaller pore size and higher electron resistance, all of which are expected to increase the EDL resistance during the charging/discharging process.

The capacitance obtained is then compared to the reference areal capacitance C_s so that the ECSA can be calculated as

ECSA = C/mC_s (10)

where m is the loaded catalyst weight (0.047 mg). The same C_s value (i.e. 0.4 F m⁻²) can be used by assuming that the studied catalysts are equally conductive.^19,32,34 Therefore, only reliable measurements of capacitance can result in an accurate ECSA. Using the CV integration method, the ECSAs of IrO₂ and NiCo₂O₄ are found to be 852 and 43.9 m² g⁻¹, respectively, based on total capacitance. These results suggest that the enhanced OER activity of IrO₂ can be partially attributed to the significantly increased ECSA. In the case of the Ni–Co material, the computed ECSA is comparable to the BET surface area, confirming the improved validity of the developed CV integration method. However, it is found that the estimated ECSA of IrO₂ is over 25-times more than the BET surface area.

In the case of IrO₂, the much higher ECSA obtained by the CV integration method can be attributed to the interaction between the evaluation method and the nature of the material. This implies over-prediction of total capacitance and/or under-prediction of standard capacitance, C_s. The integration method provides an estimate of total capacitance (C_t), which also includes the capacitance due to OH⁻ ion adsorption and intercalation in the inner and less accessible active sites. In this case, the capacitance is a result of the steady charging process and cannot be used to represent real active sites that are available at the fast rate of the OER. Furthermore, the standard capacitance, C_s, used to estimate the ECSA depends strongly on the applied materials and has a range of 0.2–1.3 F m⁻² in 1 M OH⁻ solution, as reported in the literature.¹⁹ Using the same standard capacitance C_s implies that the catalyst materials have similar EDL behaviors for charge transfer. Based on our study, IrO₂ and NiCo₂O₄ have distinctly different EDL capacitive and pseudocapacitive behaviors and, therefore, should use different standard capacitances (C_s) for determining their ECSA. If alternative values of C (C_o, 6.71 mF and C_s, 1.3 F m⁻²) are used, the ECSA of IrO₂ can be estimated to be 110 m² g⁻¹, which is comparable to the BET measurement results. It is worth noting that the ECSA obtained from CV might change with the OER potential due to EDL reconstruction and/or gas bubble effects. In addition, the adsorption of reaction intermediates can further influence the actual active area during the OER.

ECSA evaluation via EIS

In order to address potential issues using the CV method, an in situ EIS method was developed to evaluate the ECSA. EIS is a useful analytical diagnostic tool, which is capable of evaluating various resistances associated with charge-transfer processes in the EDL.^16,61 In an EIS test, the DC potential and AC perturbation are two key parameters, which determine the rate of the OER and EDL dynamic behavior, respectively.¹⁶ Based on the LSV results, four DC potentials (including 1.50, 1.55, 1.60 and 1.65 V) with an AC perturbation of 10 mV were selected for EIS testing and the Nyquist plots are shown in Fig. 8.

Fig. 8 EIS spectra of IrO₂ and NiCo₂O₄ at four applied DC potentials. Symbols – raw data; lines – linear fit to the data. The inset is the equivalent circuit.

A full semi-circular loop was observed for each study case, indicating that no mass transport limitation occurred in the studied potential range.⁶¹ With the increase of applied DC potential, a decreased semi-circular diameter was observed, corresponding to an increase in the overpotential to drive interfacial OER processes. The intercept of high-frequency spectra and x-axis at zero imaginary impedance gives an estimate of the ohmic resistance, mainly arising from the bulk electrolyte solution. Meanwhile, the ohmic resistance increases slightly with the applied potential. This is caused by the generation of gas bubbles, which impedes the transport of OH⁻. In all Nyquist plots of NiCo₂O₄, the impedance at high frequency shows an increase at 45°, which corresponds to the effective OH⁻ transport resistance.²⁵ Such a response suggests that the OH⁻ adsorption of NiCo₂O₄ is a limiting factor for the OER, which is consistent with the study of the RDS based on Tafel slopes.

An equivalent circuit is often introduced to study the impedance contribution from the faradaic process and EDL charging. The equivalent circuit we used is LR_Ω (R_ctC₁) (R₁C₂), where L, R_Ω, R_ct and C₁ symbolize inductance, ohmic resistance, charge transfer resistance and EDL capacitance, respectively. The (R₁C₂) circuit represents the diffusion/adsorption of reaction intermediates due to slow diffusion through the reaction interface in a porous material.⁶² The intermediates, such as MOH, can be oxidized to MO during OER, resulting in the (R₁C₂) impedance in the low-frequency region in the semi-circle. The capacitive components are frequently replaced with constant phase elements (CPEs) to model depressed semi-circles arising from heterogeneities and porous surfaces.⁶³ The CPE behavior may also be ascribed to a variation of properties in the direction normal to the catalyst surface. Such variability may involve changes in oxide conductivity and/or microstructure reconstruction.

In general, the impedance of a CPE can be expressed as

Z_CPE = 1/[C_f(iw)ⁿ] (11)

where C_f is a frequency-independent constant, w is the angular frequency of the alternating current signal, i is the square root of −1, and n is a factor ranging from 0 to 1. When n = 1, the CPE describes a pure capacity, and the parameter C_f has units of capacitance; otherwise, C_f has units of F s⁽ⁿ⁻¹⁾ cm⁻². Note that it is inaccurate to equate the parameter C_f with the interfacial capacitance in most cases.^63,64

As shown in Fig. 8, the fitted results from the selected equivalent circuit coincide well with the experimental data. The fit values for each circuit element/parameter are given in Table S2 and S3 (refer to the ESI†). This reveals that the resistances (including R_Ω, R_ct and R₁) of IrO₂ are lower than those of NiCo₂O₄, which results in the enhanced OER performance of the IrO₂ catalyst.

The EDL capacitance can be derived from the values fit to the impedance measurements. Combining electron transfer and OH⁻ transport, the OER occurs on an oxidized catalyst surface that OH⁻ can easily access.⁶⁵ According to Chassaing et al.⁶⁶ and Bidóia et al.,⁶⁷ the CPE behavior associated with a surface distribution of time constants requires the contribution of ohmic and polarization resistances. Therefore, the EDL capacitance can be calculated using the equation^63,64

Fig. 9 shows the dependence of EDL capacitance on the applied potential, which reflects the actual surface area available at different OER rates. The capacitances of IrO₂ and NiCo₂O₄ ranged between 1.77–1.51 and 1.17–0.70 mF, respectively. The calculated capacitance values of both samples decrease with increasing potential, which can be primarily attributed to increased O₂ coverage on the CL surface. Compared to IrO₂, NiCo₂O₄ exhibits a steeper decrease of EDL capacitance with increasing applied potential, confirming the increased EDL resistance to O₂, electron and OH⁻ transport. For comparison, the impedance results measured at the OCP (Fig. S4†) were also used to calculate the values of C (Table S4†). In the absence of the OER, the contribution of (R₁C₂) at the OCP can be neglected when fitting the data to the equivalent circuit. It is found that both the OCP capacitances, i.e. IrO₂ 5.82 μF and NiCo₂O₄ 3.41 μF, are much smaller (<1/300) than those obtained at other potentials at which the OER is occurring. The faradaic impedance (R_ct) approaches infinity without the OER. Therefore, the corresponding AC current measured is extremely small, which significantly limits the accuracy of this method.¹⁶

Fig. 9 EDL capacitance of the OER catalysts calculated from EIS data as a function of applied DC potentials.

According to the LSV results in Fig. 6(a), the lowest potential tested in EIS (i.e. 1.5 V) is close to the onset of the OER, and the transport resistance caused by O₂ gas bubbles is low. The ECSAs derived at this potential based on eqn (10) are 29.0 and 19.1 m² g⁻¹ for IrO₂ and NiCo₂O₄, respectively, when C_s is equal to 1.3 F m⁻². Compared with the above CV-based techniques (Table S4†), it is clear that the ECSA results from EIS are more comparable to the corresponding BET surface area. Despite the uncertainties caused by C_s, the collective results confirmed that the newly developed EIS approach provides an effective and reliable way to calculate the EDL capacitance, which can be used for a wide variety of catalysts. This is because the capacitance obtained by EIS is from a realistic operating potential and accounts for active surface areas and their alterations due to EDL reconstruction during the OER, as compared to voltammetry capacitance which may be significantly affected by the EDL properties. For IrO₂ with low EDL resistance, the measured EDL capacitance possibly includes the overlapping adsorption of OH⁻ and the adsorption at the inner and less accessible active sites. In contrast, for the catalyst with high EDL resistance, real voltammetry capacitance due to EDL charging can be detected only at relatively low potential scan rates, which makes the measured data less accurate due to uncertainties arising from the inclusion of charging capacitance at the less accessible sites for the OER. Moreover, in the presence of the OER, the EDL capacitance possibly varies with EDL reconstruction, redox reaction of bulk catalysts, O₂ gas bubble coverage, and adsorption of reaction intermediates. It is worth specially pointing out that the formation of Ni³⁺ (or Co³⁺) or higher oxidation states at high OER potentials can improve the material capacitive behaviors due to the increased charge conductivity compared with the non-conducting Ni²⁺ (or Co²⁺).

Mass transport in the EDL via a TF-RRDE approach

During the OER, continuous charge supply to or removal from the EDL is required to sustain a practical reaction rate. The charges involve electrons and OH⁻, and their amount determines the level of charge accumulation in the EDL. It has been reported that, when increasingly charged, the EDL becomes more active, resulting in the acceleration of reaction kinetics.^47,48 In addition to charge transport, the O₂ transport efficiency needs to be quantified, especially at relatively high current densities, to fully characterize the reaction rate and efficiency of a catalyst. To address these issues, a TF-RRDE evaluation method is developed using pair reactions of OER-ORR, as illustrated in Fig. S8.† Prior to the OER study, the collection efficiency (N) of our RRDE design was obtained by reducing Fe(SCN)₆³⁻ at the disk electrode and oxidizing its product at the Pt-ring. The collection efficiency was measured to be 37.3% for both TF-RRDE loaded IrO₂ or NiCo₂O₄ catalysts (Fig. S9†).

Before RRDE voltammetry measurements, the electrolyte (1 M KOH) was purged with N₂ for over 30 min to get rid of any dissolved O₂. Then, using the rotating LSV technique, the disk potential was stepped from 1.3 V to 1.6 V at 2 mV s⁻¹. The ring was held at 0.4 V, which has sufficient overpotential to rapidly reduce all collected oxygen. After subtracting the background current, the corrected currents at the Pt-ring electrode for both materials were recorded throughout the measurement, as shown in Fig. 10(a). The beginning of the ORR current on the ring coincides with the onset of the OER on the disk, indicating successful measurement of O₂ generation. With the increase of disk potential, the ring current increases until the collected oxygen becomes saturated. This is an indication of a significant O₂ bubble effect and that most of the O₂ escaped from the disk electrode to the electrolyte solution without any response at the ring.

Fig. 10 TF-RRDE evaluation in 1 M KOH. (a) RRDE voltammetry at the disk electrode (at a rotating rate of 1600 rpm and a scan rate of 2 mV s⁻¹) and ring electrode (holding at 0.40 V vs. the RHE). The inset illustrates the RRDE configuration. (b) EDL capacitive current ratio and (c) O₂ transport efficiency as a function of disk current.

In the LSV measurement, the disk current is the sum of the double-layer capacitive current and the OER faradaic current. In the absence of the OER, the ratio of the EDL capacitive current to the total disk current in Fig. 10(b) is equal to 100%, and the ring current response is equal to 0. The capacitive current as a function of scan rate is generally independent of the applied potential. In contrast, the OER current increases with the applied potential. This can explain why the EDL current ratio decreases with increasing current and with the OER rate. It is worth noting that the calculated EDL current ratio is most accurate at low current at which most generated oxygen can be detected by the ring electrode. It is further observed that the EDL current ratio of IrO₂ is higher than that of NiCo₂O₄. This indicates that the collective current of IrO₂ has an increased contribution from the EDL current due to its lower EDL resistance, which agrees with the above CV and EIS results. These findings reveal that the EDL current resulting from the charging/discharging process has an impact on LSV data due to the variation of EDL capacitance. Therefore, measurement methods using a low scan rate or steady-phase polarization are recommended for characterizing catalyst performance.

The O₂ transport efficiency calculated after deducting the capacitive current at the disk electrode is shown in Fig. 10(c). The disk current ranges from 0.15 to 1.50 mA to avoid noise at low current and mass transport limitations at high current. In this current range, the OER current has a dominant contribution from the detected disk current as compared to the EDL capacitive current. As shown in eqn S13,† the O₂ transport efficiency is obtained from the ratio of the ring current to the OER current obtained from the disk. The O₂ transport efficiency can be used to quantify the amount of oxygen escaping from the disk electrode to the bulk electrolyte solution without being reduced at the ring. Compared with NiCo₂O₄, a higher O₂ transport efficiency was observed for IrO₂ at a given current, confirming that IrO₂ with an improved pore structure is more efficient in removing the produced O₂.

The faradaic process may involve an OER through two-electron oxidation to form an HO₂⁻ intermediate. To assess this effect, the ring potential (0.4 V) was changed to 1.4 V for oxidizing HO₂⁻. As shown in Fig. S10,† for these two catalysts, no significant ring current could be detected when the OER occurred at the disk electrode. This confirms that both electrocatalytic OERs of IrO₂ and NiCo₂O₄ are 4-electron transfer processes.

Catalyst EDL stability

To evaluate the stability characteristics of the EDL, repetitive potential cycling tests were carried out in an O₂-saturated environment. Fig. 11(a and b) show the CV curves recorded at 100 mV s⁻¹ in an EDL potential window of 0.8–1.0 V. For both of the two materials, it can be observed that obvious changes of the CV shape occur during the first 500 or so cycles. This may be due to the losses of catalyst materials and/or the irreversible chemical and structural restructuring of the reaction interface. Throughout the 10 000 cycles, the charging/discharging curves of IrO₂ appear to be relatively symmetric around the zero current density axis. For Ni–Co oxides, the symmetry of the CV curves improves with increased scanning cycles, as demonstrating by the reversible improvement of the charging behavior. This can be attributed to the formation of a hydrous oxidation layer at the reaction interface during the CV, which can reduce the EDL resistance to charge migration.⁴⁷

Fig. 11 Stability evaluation in 1 M KOH. Transition of CV curves for (a) IrO₂ and (b) NiCo₂O₄ at a scan rate of 100 mV s⁻¹ during 10 000 cycles, and performance changes of (c) IrO₂ and (d) NiCo₂O₄ after 10 000 cycles.

After 10 000 test cycles, measurements of LSV and EIS at 1.5 V were performed to assess performance changes. The change of catalyst activity can be quantified as the difference between the integrated LSV results before and after stability tests. The major parameters indicative of catalyst activity are the Tafel slope, ECSA, and polarization resistance. The Tafel slope is obtained by fitting IR-free LSV data, and the ECSA and reaction resistance can be quantified from EIS results. In our equivalent circuit model, the total polarization resistance, R_P, is given as the sum of the charge transfer resistance, R_ct, and the reaction intermediate adsorption resistance, R₁. Based on the fitted EIS results of each material, the polarization resistance is found to be close to the diameter of the intersections between the semi-circle and the real axis at the low and high frequencies in the Nyquist plot. The change of LSV, ECSA, and R_P for both of the two materials is presented in Fig. 11(c and d).

The results obtained after cycling confirm that IrO₂ suffers from around 10% activity loss, while the NiCo₂O₄ activity is increased by 45%. This indicates that the latter provides superior long-term durability to the EDL in the alkaline medium. Further analysis reveals that Tafel slopes, shown in Fig. S11(a),† have no significant change after stability tests. In contrast, major alterations related to the ECSA and R_P were obtained from EIS results in Fig. S11(b).† The decay of IrO₂ activity can be attributed to a potential-induced transformation from a stable low-valence oxide, i.e. Ir^III/Ir^IV, into an unstable higher-valence oxide, i.e. Ir^IV/Ir^V.⁵⁸ For NiCo₂O₄, the hydrous oxidized films in the EDL can continue to grow with time due to hydrolysis reactions, accompanied by the oxidation of active elements, until the reaction reaches a balance.^1,46–48 These EDL structural modifications could result in the (i) increase of the catalyst layer roughness with a larger ECSA (ca. 5%) exposed to the electrolyte solution, (ii) reduction of the electron conductivity, i.e., R_Ω reduced from 5.2 to 4.95 Ω cm⁻², and (iii) adsorption of OH⁻ anions.¹

Conclusions

In this study, we developed an advanced benchmarking method to quantify the activity and stability of various OER catalysts. The parameters for each electrochemical method were carefully selected to produce reliable results for characterizing catalyst properties during the electrocatalytic OER. Moreover, the newly developed EDL microstructure illustrates the significance of interfacial charging, pseudocapacitive and transport behaviors. The collective results confirm that IrO₂ exhibited 2.6-times higher OER activity than Ni–Co oxide, which was ascribed to an increased ECSA and decreased resistances for electron, OH⁻ and O₂ transport in the EDL. To obtain reliable ECSAs of the catalysts, three traditional CV methods were compared with a newly developed in situ EIS method. It was found that based on the EIS approach, the ECSAs obtained as a function of applied OER potential are more reliable than those from CV methods and have the best agreement with the BET surface area. In addition, repetitive potential cycling results indicated the superior EDL stability of NiCo₂O₄, resulting from the continuous growth of hydrous oxidized films. Based on these findings, future work in designing improved OER catalysts should focus on increasing the availability and stability of the EDL for long-term OER applications.

Experimental section

Chemical reagents

Potassium hydroxide (KOH, ≥90 wt%), anhydrous ethanol (≥95 wt%), 5 wt% Nafion® solution, and nickel–cobalt oxides (NiCo₂O₄, ≥99.99%) were purchased from Sigma-Aldrich. Iridium dioxide (IrO₂, ≥99.99%) was obtained from Alfa Aesar. Solutions were prepared using ultrapure deionized water (18 MΩ cm resistivity) supplied by an in-house water purification system from Thermo Scientific Barnstead Nanopure.

Measurement equipment

Electrochemical measurements in a half cell were conducted with a CHI 750E bipotentiostat at room temperature and ambient pressure. A standard three-electrode system was constructed for the TF-RDE/RRDE measurements and the major components were purchased from Pine Instrument Company. The three electrodes include an Hg/HgO RE (filled with 1 M KOH), a Pt wire (7 mm OD × 65 mm long) CE, and a WE with a rotating function. After the deposition of a thin-film catalyst layer, the working electrode was affixed to a modulated speed rotator (MSR). For TF-RDE measurements, the working electrode was an E3 series RDE (OD = 5 mm) with a glassy carbon (GC) disk. TF-RRDE tests used an E7R9 series RRDE with a sub-millimeter gap between the GC disk (OD = 5.61 mm) and the Pt ring (OD = 7.92 mm; ID = 6.25 mm). The electrolyte solution of 1 M KOH was added to a solid PTFE cell (300 mL) to prevent any corrosion. The pH of the 1 M KOH solution was measured to be 13.93 with an HI98128 pH/Temperature Tester (Hanna Instruments). After degassing in a vacuum at 200 °C for 2 h, the materials were tested using low-temperature N₂ adsorption–desorption isotherm measurements at −196 °C on a MicroMeritics TriStar 3000 Porosimeter. Based on the collected isotherms, the Brunauer–Emmett–Teller (BET) method was used to estimate the specific surface area in the relative pressure range between 0.01 and 0.30. The average pore diameter and total pore volume were assessed by applying the Barrett–Joyner–Halenda (BJH) mode to the desorption branch of the isotherm. The zeta potential of the catalyst ink was obtained using a Malvern Zetasizer Nano ZS90 at room temperature.

Thin-film catalyst layer (CL) preparation

A high-quality CL can be obtained by casting catalyst–ionomer–ethanol-containing inks onto a cleaned glassy carbon disk electrode. Prior to depositing the catalyst–ionomer film, the disk electrode was polished with a 50 nm Al₂O₃ particle suspension on a rayon micro-cloth disk in a “Fig. 8” pattern for 2–3 minutes. The ink in our work was prepared by mixing 5 mg of catalyst with a Nafion ionomer solution (5 wt%) and 4 mL ethanol. The major roles of the Nafion ionomer were to improve the catalyst dispersion in the resulting ink and enhance the interfacial adhesion between the CL and the disk substrate.³⁶ The mass ratio of ionomer to catalyst was 1/2 in each sample. To form a uniform dispersion, the ink was sonicated for 30 min in a water bath and the temperature controlled to less than 35 °C. The resulting ink was then deposited onto the entire surface of the disk electrode as droplets using a 2–20 μL adjustable micro-pipette and dried in air at ambient temperature. The coating/drying cycles were carried out several times until the target 0.24 mg cm⁻² loading was reached.

Measurement procedure for the TF-RDE

The TF-RDE was mounted onto the MSR assembly with the electrode surface facing downward in the 1 M KOH solution. In the case of stability tests, O₂ was used instead of N₂ for purging the electrolyte solution throughout the process. Detailed parameters for each electrochemical technique, such as CV, LSV and EIS, are given in Fig. 2. The quiet time in each test was set to 10 s before beginning a new measurement. CV curves were obtained under static conditions (rotating speed = 0). During the LSV and EIS processes, a standard electrode rotating speed of 1600 rpm was used in order to facilitate mass transport and O₂ removal from the reaction interface. The Nyquist impedance diagrams were obtained using the CHI program.

For accelerated stability evaluation, 10 000 potential cycles from 0.8–1.0 V were performed in the O₂-purged 1 M KOH solution at a scan rate of 100 mV s⁻¹. Subsequently, the performance degradation was investigated by comparing the LSV and EIS results obtained before and after stability tests.

Measurement procedure of the TF-RRDE

As shown schematically in Fig. S8,† the RRDE consists of two working electrodes, i.e. an inner GC disk electrode and an outer Pt-ring electrode, in which the potentials can be controlled individually with the bipotentiostatic CHI 750E. Typical TF-RRDE voltammetry experiments are conducted by scanning the potential on the disk electrode via LSV, while keeping the Pt-ring potential constant at a value at which an intermediate/product of interest can be oxidized or reduced. Major experimental parameters are listed in Fig. 2.

Depending on the ring potential controlled with the bipotentiostatic instrument, an oxygen reduction reaction (ORR) or peroxide oxidizing reaction would occur at the Pt-ring electrode. In our work, a newly developed analysis method, described in the ESI,† was used to examine the contribution of the EDL charging current and O₂ transport efficiency (ε) during the OER.

In addition, the collection efficiency (N) of the RRDE was obtained using a N₂-saturated solution of 10 mM K₃Fe(CN)₆³⁻ and 1 M KOH mixture. The details of the measurement and calculation procedures are provided in the ESI.†

Extended stability measurements using a gas diffusion electrode (GDE)

The extended stability tests were carried out in a traditional alkaline water electrolyser, including an OER WE and a Pt gauze CE, as illustrated in Fig. S2 (refer to the ESI†). In preparing the WE, the ink with the catalyst and Nafion ionomer was sonicated for 30 min. The dispersed ink was then coated dropwise on a wet-proofed carbon paper, which is frequently used as the substrate of GDEs in fuel cells due to its high mechanical strength, gas transport, electrical conductivity and stability.⁶⁸ The fabricated GDE was dried under an infrared lamp, and subsequently cut into a 1 × 1 cm² active area with 2.5 mg cm⁻² catalyst loading. The catalyst stability was validated by current-controlled (10/50 mA cm⁻²) electrolysis tests for 1000 h in the stirred 1 M KOH at room temperature.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

The authors would like to thank Prof. Peggy O'Day and Dr Ajith Pattammattel for their support in low-temperature N₂ adsorption–desorption isotherm measurements and Prof. Ashlie Martini for reviewing our work. This work was supported by the GREENPower Program (IIID 2015-09) from the Commission on Higher Education – Philippine California Advanced Research Institutes (CHED-PCARI) of the Republic of the Philippines, “Fundamental Research Funds for the Central Universities” from Wuhan University of Technology, and University of California, Merced.

References

1. R. L. Doyle and M. E. G. Lyons, in Photoelectrochemical Solar Fuel Production: From Basic Principles to Advanced Devices, ed. S. Giménez and J. Bisquert, Springer International Publishing, Cham, 2016, pp. 41–104,  DOI:10.1007/978-3-319-29641-8_2.
2. M. S. Burke, L. J. Enman, A. S. Batchellor, S. Zou and S. W. Boettcher, Chem. Mater., 2015, 27, 7549–7558.
3. B. A. Pinaud, J. D. Benck, L. C. Seitz, A. J. Forman, Z. Chen, T. G. Deutsch, B. D. James, K. N. Baum, G. N. Baum, S. Ardo, H. Wang, E. Miller and T. F. Jaramillo, Energy Environ. Sci., 2013, 6, 1983–2002.
4. E. M. Miner and M. Dincă, Nat. Energy, 2016, 1, 16186.
5. Y. Chen, F. Mojica, G. Li and P.-Y. A. Chuang, Int. J. Energy Res., 2017, 1–9,  DOI:10.1002/er.3806.
6. J. R. Varcoe, P. Atanassov, D. R. Dekel, A. M. Herring, M. A. Hickner, P. A. Kohl, A. R. Kucernak, W. E. Mustain, K. Nijmeijer, K. Scott, T. Xu and L. Zhuang, Energy Environ. Sci., 2014, 7, 3135–3191.
7. C. C. Pavel, F. Cecconi, C. Emiliani, S. Santiccioli, A. Scaffidi, S. Catanorchi and M. Comotti, Angew. Chem., Int. Ed., 2014, 53, 1378–1381.
8. A. T. Swesi, J. Masud and M. Nath, Energy Environ. Sci., 2016, 9, 1771–1782.
9. X. Li, X. Hao, A. Abudula and G. Guan, J. Mater. Chem. A, 2016, 4, 11973–12000.
10. A. Grimaud, A. Demortière, M. Saubanère, W. Dachraoui, M. Duchamp, M.-L. Doublet and J.-M. Tarascon, Nat. Energy, 2016, 2, 16189.
11. J. H. Bae, J.-H. Han and T. D. Chung, Phys. Chem. Chem. Phys., 2012, 14, 448–463.
12. A. Ganassin, A. Maljusch, V. Colic, L. Spanier, K. Brandl, W. Schuhmann and A. Bandarenka, ACS Catal., 2016, 6, 3017–3024.
13. L. Trotochaud and S. W. Boettcher, Scr. Mater., 2014, 74, 25–32.
14. C. Spoeri, J. T. H. Kwan, A. Bonakdarpour, D. Wilkinson and P. Strasser, Angew. Chem., 2017, 56, 5994–6021.
15. H. Dau, C. Limberg, T. Reier, M. Risch, S. Roggan and P. Strasser, ChemCatChem, 2010, 2, 724–761.
16. A. J. Bard and L. R. Faulkner, Electrochemical Methods. Fundamentals and Applications, John Wiley & Sons, Inc., 2nd edn, 2001.
17. S. Trasatti and O. A. Petrii, J. Electroanal. Chem., 1992, 327, 353–376.
18. S. Ardizzone, G. Fregonara and S. Trasatti, Electrochim. Acta, 1990, 35, 263–267.
19. C. C. L. McCrory, S. Jung, J. C. Peters and T. F. Jaramillo, J. Am. Chem. Soc., 2013, 135, 16977–16987.
20. C. C. McCrory, S. Jung, I. M. Ferrer, S. M. Chatman, J. C. Peters and T. F. Jaramillo, J. Am. Chem. Soc., 2015, 137, 4347–4357.
21. R. Frydendal, E. A. Paoli, B. P. Knudsen, B. Wickman, P. Malacrida, I. E. L. Stephens and I. Chorkendorff, ChemElectroChem, 2014, 1, 2075–2081.
22. S. Zhao, Y. Wang, J. Dong, C.-T. He, H. Yin, P. An, K. Zhao, X. Zhang, C. Gao, L. Zhang, J. Lv, J. Wang, J. Zhang, A. M. Khattak, N. A. Khan, Z. Wei, J. Zhang, S. Liu, H. Zhao and Z. Tang, Nat. Energy, 2016, 1, 16184.
23. E. Fabbri, A. Habereder, K. Waltar, R. Kotz and T. J. Schmidt, Catal. Sci. Technol., 2014, 4, 3800–3821.
24. P. Stonehart and P. N. Ross, Electrochim. Acta, 1976, 21, 441–445.
25. K. Shinozaki, J. W. Zack, S. Pylypenko, B. S. Pivovar and S. S. Kocha, J. Electrochem. Soc., 2015, 162, F1384–F1396.
26. M. Shao, Q. Chang, J.-P. Dodelet and R. Chenitz, Chem. Rev., 2016, 116, 3594–3657.
27. S.-J. Lee, S.-I. Pyun, S.-K. Lee and S.-J. L. Kang, Isr. J. Chem., 2008, 48, 215–228.
28. F.-X. Ma, L. Yu, C.-Y. Xu and X. W. Lou, Energy Environ. Sci., 2016, 9, 862–866.
29. X. Gao, H. Zhang, Q. Li, X. Yu, Z. Hong, X. Zhang, C. Liang and Z. Lin, Angew. Chem., Int. Ed., 2016, 55, 6290–6294.
30. L. Wang, C. Lin, D. Huang, F. Zhang, M. Wang and J. Jin, ACS Appl. Mater. Interfaces, 2014, 6, 10172–10180.
31. X. Lu, W.-L. Yim, B. H. R. Suryanto and C. Zhao, J. Am. Chem. Soc., 2015, 137, 2901–2907.
32. J. T. Mefford, X. Rong, A. M. Abakumov, W. G. Hardin, S. Dai, A. M. Kolpak, K. P. Johnston and K. J. Stevenson, Nat. Commun., 2016, 7, 11053.
33. B. You, N. Jiang, M. Sheng, S. Gul, J. Yano and Y. Sun, Chem. Mater., 2015, 27, 7636–7642.
34. A. Grimaud, O. Diaz-Morales, B. Han, W. T. Hong, Y.-L. Lee, L. Giordano, K. A. Stoerzinger, M. T. M. Koper and Y. Shao-Horn, Nat. Chem., 2017, 9, 457–465.
35. B. Bladergroen, H. Su, S. Pasupathi and V. Linkov, in Electrolysis, ed. V. Linkov, InTech, 2012, pp. 45–60,  DOI:10.5772/52947.
36. H.-Y. Jung and J.-H. Choi, J. Solid State Electrochem., 2012, 16, 1571–1576.
37. S. Jung, C. C. L. McCrory, I. M. Ferrer, J. C. Peters and T. F. Jaramillo, J. Mater. Chem. A, 2016, 4, 3068–3076.
38. R. J. Hunter, Zeta potential in colloid science: principles and applications, Academic press, 2013.
39. V. R. Stamenkovic, D. Strmcnik, P. P. Lopes and N. M. Markovic, Nat. Mater., 2017, 16, 57–69.
40. W. M. Haynes, CRC Handbook of Chemistry and Physics, 94th edn, CRC Press, 2013.
41. J. Wang, J. Polleux, J. Lim and B. Dunn, J. Phys. Chem. C, 2007, 111, 14925–14931.
42. V. Augustyn, P. Simon and B. Dunn, Energy Environ. Sci., 2014, 7, 1597–1614.
43. S. Khalid, C. Cao, L. Wang and Y. Zhu, Sci. Rep., 2016, 6, 22699.
44. W. Zhou, D. Kong, X. Jia, C. Ding, C. Cheng and G. Wen, J. Mater. Chem. A, 2014, 2, 6310–6315.
45. C. Costentin, T. R. Porter and J.-M. Savéant, J. Am. Chem. Soc., 2016, 138, 5615–5622.
46. K. K. Lian and V. I. Birss, J. Electrochem. Soc., 1991, 138, 2885–2890.
47. F. M. Mulder, B. M. H. Weninger, J. Middelkoop, F. G. B. Ooms and H. Schreuders, Energy Environ. Sci., 2017, 10, 756–764.
48. N. Weidler, J. Schuch, F. Knaus, P. Stenner, S. Hoch, A. Maljusch, R. Schäfer, B. Kaiser and W. Jaegermann, J. Phys. Chem. C, 2017, 121, 6455–6463.
49. G. Li, H. Yu, D. Yang, J. Chi, X. Wang, S. Sun, Z. Shao and B. Yi, J. Power Sources, 2016, 325, 15–24.
50. B. Zhang, X. Zheng, O. Voznyy, R. Comin, M. Bajdich, M. García-Melchor, L. Han, J. Xu, M. Liu, L. Zheng, F. P. García de Arquer, C. T. Dinh, F. Fan, M. Yuan, E. Yassitepe, N. Chen, T. Regier, P. Liu, Y. Li, P. De Luna, A. Janmohamed, H. L. Xin, H. Yang, A. Vojvodic and E. H. Sargent, Science, 2016, 352, 333–337.
51. R. A. Rincon, J. Masa, S. Mehrpour, F. Tietz and W. Schuhmann, Chem. Commun., 2014, 50, 14760–14762.
52. M. Gong, Y. Li, H. Wang, Y. Liang, J. Z. Wu, J. Zhou, J. Wang, T. Regier, F. Wei and H. Dai, J. Am. Chem. Soc., 2013, 135, 8452–8455.
53. D. B. Rogers, R. D. Shannon, A. W. Sleight and J. L. Gillson, Inorg. Chem., 1969, 8, 841–849.
54. I. C. Man, H.-Y. Su, F. Calle-Vallejo, H. A. Hansen, J. I. Martínez, N. G. Inoglu, J. Kitchin, T. F. Jaramillo, J. K. Nørskov and J. Rossmeisl, ChemCatChem, 2011, 3, 1159–1165.
55. Y. Matsumoto and E. Sato, Mater. Chem. Phys., 1986, 14, 397–426.
56. H. F. Liang, F. Meng, M. Caban-Acevedo, L. S. Li, A. Forticaux, L. C. Xiu, Z. C. Wang and S. Jin, Nano Lett., 2015, 15, 1421–1427.
57. F. Song and X. L. Hu, Nat. Commun., 2014, 5, 4477.
58. G. Li, H. Yu, X. Wang, S. Sun, Y. Li, Z. Shao and B. Yi, Phys. Chem. Chem. Phys., 2013, 15, 2858–2866.
59. J. C. Liu and J. Wei, Chem. Eng. Sci., 2014, 111, 1–14.
60. J. Gaudet, A. C. Tavares, S. Trasatti and D. Guay, Chem. Mater., 2005, 17, 1570–1579.
61. X. Yuan, C. Song, H. Wang and J. Zhang, in Electrochemical Impedance Spectroscopy in PEM Fuel Cells, Springer, London, 2010, ch. 5, pp. 193–262,  DOI:10.1007/978-1-84882-846-9_5.
62. E. Rasten, G. Hagen and R. Tunold, Electrochim. Acta, 2003, 48, 3945–3952.
63. M. E. Orazem and B. Tribollet, in Electrochemical Impedance Spectroscopy, John Wiley & Sons, Inc., 2008, pp. 233–263,  DOI:10.1002/9780470381588.ch13.
64. M. E. Orazem, I. Frateur, B. Tribollet, V. Vivier, S. Marcelin, N. Pébère, A. L. Bunge, E. A. White, D. P. Riemer and M. Musiani, J. Electrochem. Soc., 2013, 160, C215–C225.
65. I. Katsounaros, S. Cherevko, A. R. Zeradjanin and K. J. J. Mayrhofer, Angew. Chem., Int. Ed., 2014, 53, 102–121.
66. E. Chassaing, B. Sapoval, G. Daccord and R. Lenormand, J. Electroanal. Chem. Interfacial Electrochem., 1990, 279, 67–78.
67. E. D. Bidóia, L. O. S. Bulhões and R. C. Rocha-Filho, Electrochim. Acta, 1994, 39, 763–769.
68. J. Park, H. Oh, T. Ha, Y. I. Lee and K. Min, Appl. Energy, 2015, 155, 866–880.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c7se00337d

---