Electrocatalytic Activity of Transition Metal Oxide-Carbon Composites for Oxygen Reduction in Alkaline Batteries and Fuel Cells

Calcium-doped lanthanum cobalt oxide (LCCO) of the formula, La_0.6Ca_0.4 Co O_3-x was synthesized in-house. The metal nitrates of lanthanum, calcium and cobalt were dissolved in water along with citric acid and slowly evaporated to form a sol-gel. The sol-gel was then heated in two steps to yield a nano-particulate oxide phase with the perovskite structure. A detailed description of the synthesis and structural characterization of the LCCO materials used in the present study can be found in our earlier publication.⁸ LCCO was mixed with acetylene black in various proportions shown in Table I to prepare the composite catalysts.

Each of the catalyst compositions shown in Table I was separately combined with water, isopropanol and Nafion to form an ink. A mixture of 0.008 g of catalyst, 1.8 g of water, 0.01 g of a solution of 5% Nafion ionomer, and 0.2 g of isopropanol was subjected to 40 minutes of ultrasonic agitation to form a uniformly dispersed ink. A 20 μL droplet of this catalyst ink was then placed on the clean surface of a glassy carbon electrode (5 mm diameter mounted in a rotating-disk electrode holder, Pine Instruments) and dried in air at 85°C for 15 minutes. This method allowed us to apply about 80 microgram of catalyst material on the glassy carbon disk. The coverage of the carbon disk by the catalyst layer with 80 microgram of catalyst allowed us to get reproducible results and consistent with the range of catalyst loadings used in the literature.^33,35 The Nafion content of the catalyst layer was approximately 6 wt%. At this level of Nafion content no blockage of oxygen diffusion was observed.

In three other separate preparations, 8 mg of LCCO was combined with either (a) 8 mg gold nanoparticles (nanopowder, <100 nm particle size, 99.9% trace metal basis, Sigma Aldrich), (b) 8 mg of carbon nanotubes (multi-walled >98% carbon basis, O.D. × L 6-13 nm × 2.5–20 μm, Sigma Aldrich), or (c) 8 mg of graphene (8 nm flakes Grade AO-2, Graphene Supermarket) to produce inks and catalyst layers using the procedure described above.

Electrochemical testing was conducted in three-electrode polyfluoroethylene cells. A 1 M solution of potassium hydroxide (Sigma Aldrich) was used as the electrolyte in all the tests. The cells were inert to the strongly alkaline solutions and dilute solutions of hydrogen peroxide. The electrolyte was continuously purged with high-purity oxygen or argon (UHV grade 99.999%). High-purity water (18.2 MΩ, 4 ppb total organic carbon) was used in all the experiments. A mercury-mercuric oxide (MMO) electrode served as the reference electrode (20% potassium hydroxide solution, E^o = +0.098 V) and a platinum wire was used as the counter electrode. All electrode potentials reported herein are with reference to the MMO electrode unless otherwise specified.

The oxygen reduction activity was studied by linear-sweep voltammetry (a slow sweep rate of 2 mV/s) between 0.1 V and −0.5 V while the working electrode was rotated at 400, 900, 1600 and 2500 rotations per minute (rpm). Prior to the oxygen reduction studies, the electrolyte was saturated with oxygen at 1 atm for at least 50 minutes. Before recording the polarization curves, the working electrode was cycled between 0.1 V and −0.5 V at 20 mV/s for 20 cycles to obtain an invariant voltammogram. All measurements were made with a bi-potentiostat (Autolab PGS30). A ring-disk electrode (Pine Instruments Inc.) consisting of a 5 mm diameter glassy carbon disk and 7 mm diameter platinum ring was used for studying the hydrogen peroxide intermediates produced during oxygen reduction. The ring electrode was held at +0.5 V for oxidizing the peroxide transported from the disk electrode. The collection efficiency of ring-disk combination was determined using a potassium ferricyanide solution purged with argon and the value of collection efficiency was found to be 0.20. Analysis of polarization curves focused on the dependence of the kinetic current on composition. Kinetic currents were calculated from the measured current-potential curves by correcting for mass transport effects. The kinetic current I_k is given by:

Where I_obs is the measured current and I_lim is the mass-transport limited current.

The results of the effect of carbon on the oxygen reduction activity of LCCO are presented in Figure 1. The kinetics of oxygen reduction on LCCO in the absence of any added carbon appeared to be limited largely by the charge transfer process even at −0.40 V, and the familiar mass-transport limited plateau current was not observed even at significantly negative electrode potentials. Acetylene black exhibited much faster charge-transfer kinetics compared to LCCO and a mass-transport limited current was observed even at −0.25 V. At similar values of electrode potential, the current observed with acetylene black was about one order of magnitude greater than that with LCCO. However, with the mixture of 9% acetylene black and 91% LCCO, the kinetic current (for example, in the potential range of −0.1 to −0.2 V) and the mass-transport limited currents were about two times greater than with acetylene black catalyst and twenty-fold more than with the "LCCO-only" catalyst. We also note that the "half-wave potentials" for carbon, LCCO and the composite LCCO-carbon catalysts are different. These differences are discussed later here in terms of the Tafel analysis of the kinetic currents.

Zoom In Zoom Out Reset image size

The doubling of the mass-transport limited current with the combination of acetylene black and LCCO suggested a doubling of the number electrons transferred in the overall oxygen reduction reaction, because the geometric area of all the electrodes tested was the same. To determine if the enhancement in catalyst performance observed with acetylene black addition to LCCO was caused by just an increase in the electrical conductivity of the catalyst layer, we also conducted experiments with various other conductive additives such as carbon nanotubes, graphene and gold nanoparticles. The results of these tests (Figure 2) show that the addition of other carbon-based additives, namely, carbon nanotubes and graphene also enhanced the observed activity of LCCO similar to the addition of acetylene black. However, among the additives tested, the gold nanoparticles despite their good electrical conductivity had the least impact on the oxygen reduction activity of LCCO. Thus, the enhancement in activity appeared to be quite specific to the carbon-based additives. Also the significant differences observed among the various types of carbon additives suggest that carbon plays a direct role in the catalysis beyond just increasing the electrical conductivity of the composite catalyst.

Zoom In Zoom Out Reset image size

As part of our effort to develop bi-functional catalysts, we also investigated the activity of the LCCO based catalysts toward oxygen evolution. We observed that the oxygen evolution activity was not influenced by the addition of carbon. If the role of carbon during oxygen reduction was merely to increase the electrical conductivity of the catalyst layer, then one should have observed a similar enhancement in oxygen evolution activity with carbon addition. Since the results of oxygen evolution studies did not support this expectation, it became clear that the carbon additives in the LCCO-carbon composites had a distinct role in enhancing the oxygen reduction activity besides the commonly perceived function of just increasing the electronic conductivity of the composite catalyst.

In further studies of the role of carbon we found that the oxygen reduction activity of the composite catalyst was dependent on the mass ratio of the acetylene black to LCCO (Figure 3). For the all the compositions that included both acetylene black and LCCO, the activity of the catalysts exceeded that of the values predicted by the weighted sum of the activity of the individual constituents (Fig. 3a). The higher than predicted values of activity for all the compositions of the carbon-containing catalysts confirmed the synergistic catalytic roles for LCCO and acetylene black. In comparing the activities of compositions 8 and 9, we found that the even 6 to 11 weight% of LCCO enhanced the activity of the composite catalyst significantly. In comparing compositions 2, 3, 4 and 5 we found that the activity of the composite catalysts scaled linearly with the amount of carbon suggesting a first order dependence on the amount of carbon for all the compositions (Figure 3b). The current also increased rapidly with small increases in the amount of LCCO, but the rise in activity was less significant at higher fractions of LCCO (Figure 3c). Thus, the dependence of activity of the composite catalysts on the amount of LCCO changed from a first-order dependence for small amounts of LCCO to a near zero-order dependence at high fractions of LCCO. Such a change in order could be because at higher fractions of LCCO in the catalyst, not all the mass of LCCO can make direct contact with the carbon material.

Zoom In Zoom Out Reset image size

The first-order dependence on carbon suggested that carbon functioned in a primary catalyst role whereby the larger the amount of carbon, the larger was the reduction current. However, the enhancement in activity with LCCO leveling off at 11% of LCCO suggested that LCCO was not the primary surface for the electrochemical reactions and that LCCO functioned as a co-catalyst that enhanced the role of carbon.

Savinova et al.³⁴ have recently reported that this co-catalyst role of the perovskite oxide is due to the decomposition of hydrogen peroxide produced on the surface of carbon. For the first time their studies drew direct attention to higher catalytic activity of mixtures of carbon and with lanthanum cobalt oxide and strontium-doped lanthanum manganese oxide compared to that of the individual constituents. They concluded that the oxygen reduction reaction on carbon containing oxide composite cathodes must be considered as coupled reactions where the individual contributions cannot always be separated. They have recommended further validation of this mechanism by rotating-ring disk methods. We also recognize that in the study of cobalt containing non-platinum catalysts reported by Atanassov et al.,^35,36 suggest a dual-site synergistic process. Here, the authors identify the cobalt oxide on the nanoparticle to be the site for the destruction of peroxide.

Therefore, in the present study we have focused on understanding the mechanism of interaction between the oxides and carbon in greater detail by extensive analysis of ring-disk results.

In addition to LCCO, we have observed similar enhancements in oxygen reduction activity when lanthanum calcium cobalt manganese oxide (La_0.6Ca_0.4Co_0.5 Mn_0.5 O_3-x, LCCM), and nickel cobalt oxide spinel are combined with carbon (Figure 4). We find that nickel cobalt oxide when combined with carbon was at least and order of magnitude more active compared to LCCO and LCCM. It is therefore clear that the chemical constitution of the transition metal oxide plays a crucial role in the observed kinetics of oxygen reduction. Thus, our results also indicated that the degree of enhancement in activity varied with the type of transition metal oxide (Figure 4) and the type of carbon additive (Figure 2). Therefore, while studying the oxygen reduction activity of transition metal oxides mixed with carbon it is important not to ascribe the observed activity differences to just the electrochemical processes on the transition metal oxide.

Zoom In Zoom Out Reset image size

We describe below our efforts to confirm the mechanism by which the combination of carbon and transition metal oxide gave rise to the enhanced oxygen reduction activity.

The oxygen reduction reaction in alkaline media is known to occur by three familiar pathways⁴ represented by the following chemical reactions:

• (1)  
  Direct four-electron pathway

  
• (2)  
  "Series" pathway with two consecutive electrochemical steps

  

  
• (3)  
  Electron transfer followed by decomposition of hydrogen peroxide

  

  

The formation of hydroperoxide (Eqs. 4 and 6) can be monitored using the rotating-ring disk electrode arrangement wherein the hydroperoxide produced on the rotating disk electrode is convectively transported to the ring electrode and detected by an oxidation current.^37–39 The ratio of the disk and ring currents can be used to diagnose the reaction pathway.⁴⁰ Therefore, the disk electrode was coated with the composite catalyst of interest, and the hydroperoxide formed at the disk electrode was oxidized at the platinum ring electrode. The disk and ring currents were used to determine the percentage of the current involved in the production of hydroperoxide and the number of electrons transferred per mole of oxygen (Figure 5).

Zoom In Zoom Out Reset image size

On LCCO, the formation of hydroperoxide accounted for approximately 5% of the oxygen reduction current at the disk. The number of electrons transferred was calculated to be 3.9. Consequently, it was concluded that the oxygen reduction reaction on LCCO occurred by either the direct four-electron pathway (Eq. 3) or by the "series" pathway with a slow two-electron transfer step followed by a rapid two-electron reduction of the hydroperoxide (Eqs. 4 and 5), or the slow electrochemical formation of hydroperoxide followed by rapid decomposition to oxygen (Eqs. 6 and 7). For all these situations hydroperoxide should be detected only in insignificant amounts at the ring electrode and the number of electrons transferred will approximate to four. Further identification of the exact pathway could be carried out only with the following additional experiments.

To determine which of the "four-electron" pathways were operative in the reduction of oxygen on LCCO, we measured the currents for the direct reduction of 0.1 M hydroperoxide on the LCCO-coated disk electrode to estimate the contribution of the electrochemical reduction reaction in Eq. 4, and also measured the decomposition rate of hydroperoxide in contact with LCCO. When LCCO powder was added to 0.1M hydrogen peroxide solution a rapid decrease in the hydrogen peroxide oxidation current at the glassy carbon disk electrode suggested a decrease in the concentration of hydroperoxide and confirmed that LCCO rapidly decomposed the hydroperoxide anion to produce oxygen. Thus, we were able to confirm that the process of decomposition according to Eq. 7 readily occurred on LCCO. Further, when a disk coated with 80 μg of LCCO was polarized in 0.1 M solution of hydroperoxide (Figure 6a) the reduction current observed was similar to the current observed with oxygen saturated solutions (Figure 5a). This result is consistent with the occurrence of two-electron electro-reduction of oxygen to hydroperoxide (Eq. 6) followed by rapid decomposition of the hydroperoxide to oxygen (Eq. 7). However, if there was a small current contribution on LCCO by the direct four-electron reduction of oxygen (Eq. 3), it could not be ruled out.

Zoom In Zoom Out Reset image size

On acetylene black, the rate of generation of hydroperoxide was at least an order of magnitude higher than that on LCCO (Figure 5b). The hydroperoxide measurements on the ring electrode showed that almost 95% of the reduction current was directed through the two-electron pathway (Eq. 4). The number of electrons transferred in the reaction of 2.1 to 2.5 was consistent with Eq. 4. This number of electrons being slightly higher than 2 suggested a minor amount of decomposition of hydrogen peroxide on acetylene black or a small contribution from subsequent reduction steps through electron transfer.

On the composite catalysts consisting of acetylene black and LCCO, unlike the acetylene black electrode, the hydroperoxide generation was just 5% of the total reduction current at the disk electrode. Thus, the number of electrons transferred was calculated to be 3.9 suggesting a net four-electron process. This value of approximately 4 for the number of electrons was consistent with the mass transport-limited disk currents being twice as much as that on acetylene black electrode. Since we had independently confirmed the near quantitative generation of hydroperoxide on acetylene black (Figure 5b) and the rapid decomposition of hydroperoxide on LCCO (Figures 6a and 6b), it was reasonable to expect that in the composite catalyst, the hydroperoxide formed at the carbon electrode would be rapidly decomposed by the LCCO. In the steady state, the electro-reduction to hydroperoxide on carbon and the subsequent decomposition to oxygen on LCCO would lead to a net observation of four electrons transferred per oxygen molecule, readily seen by adding Eqs. 6 and 7.

The generation of hydroperoxide was studied for varying amounts of LCCO and acetylene black (Figure 7).

Zoom In Zoom Out Reset image size

When the amount of LCCO was large as compared to the amount of acetylene black (compositions 2, 3, 4, 5 and 6 in Table I) almost all the hydroperoxide produced on the surface of acetylene black is decomposed, and a very small amount of free hydroperoxide was detected at the ring electrode. However, when the carbon-LCCO composite was made of a very large fraction of carbon as in compositions 7,8 and 9, the hydroperoxide was not completely decomposed allowing significant amounts of hydroperoxide to be detected at the ring electrode. Thus, the calculated value for the number of electrons transferred varied between 3 and 4 when the LCCO concentration was low.

Based on the results presented in Figure 5 and Figure 6b it was certain that LCCO produced almost no hydroperoxide during oxygen reduction and the acetylene black catalyst produced hydroperoxide almost quantitatively. We have estimated the expected production of the amount of hydroperoxide on the composite catalyst using a weighted sum of the independent contributions of LCCO and acetylene black to the production of hydroperoxide. We have compared (Figure 8) the calculated results of hydroperoxide production with the experimental results (Figures 3a and 7a).

Zoom In Zoom Out Reset image size

If the LCCO and carbon acted independently the amount of hydroperoxide detected on the LCCO- carbon composites would have to be considerably larger than that experimentally observed (Figure 8). Therefore, the decreased amount of hydroperoxide detected at the ring when both acetylene black and LCCO are present is simply the result of the decomposition by LCCO of the hydroperoxide produced on acetylene black. Thus, LCCO acted synergistically with acetylene black in catalyzing the electro-reduction of oxygen and the rapid decomposition of the hydroperoxide produced on the acetylene black when sufficient LCCO was present, and thus yielded the net four electrons transferred per oxygen molecule.

LCCO, by itself, was found to be a relatively poor catalyst for the oxygen reduction reaction compared to the composite of LCCO and acetylene black. Acetylene black was not a spectator that simply enhanced electrical conductivity, but in fact had a primary role in determining the catalytic activity of the composite. Thus, LCCO may be termed a co-catalyst as it enhanced the observed currents by decomposing the hydroperoxide species produced on acetylene black. This role of LCCO is similar to that of silver and manganese dioxide (commonly termed peroxide decomposers) used widely with carbon-based electrodes in alkaline fuel cells and primary metal-air batteries. The findings presented above require that the analysis of the electrocatalytic activity of composites of LCCO and carbon do not ignore the primary function of LCCO as a peroxide decomposer.

We have determined that many other transition metal oxide perovskites and spinels behave in a similar manner to LCCO. In the other examples of transition metal oxides the enhanced activity of composite catalysts correlated well with the reduced amount of hydroperoxide detected at the ring electrode, as in the case of LCCO. Consequently, the differences in the observed electrocatalytic activity of various transition metal oxides when combined with carbon, although apparently a four-electron transfer process, depends significantly on the ability of the transition metal oxide to decompose the hydroperoxide generated by the electro-reduction of oxygen on the carbon constituent of the composite catalyst (Eqs. 6 and 7). It is quite likely that the ability of the transition metal oxide to decompose hydroperoxide (Eq. 7) is related to the ability of the oxide surface to support a direct four-electron transfer (Eq. 3). However, we must make a distinction between these two net "four electron" pathways while interpreting the observed activity of transition metal oxide-carbon composites.

When the acetylene black in the composite catalyst was substituted by carbon nanotubes or graphene, a similar enhancement in catalytic activity was observed (Figure 2). However, substitution of acetylene black with gold nanoparticles did not produce such an enhancement. While carbon-based materials are known for their ability to perform efficient two electron reduction to hydroperoxide, gold is generally a poor catalyst for oxygen reduction.^41,42 Consequently, gold does not substitute the primary catalyst role of carbon in the production of hydroperoxide necessary for producing the enhancement in activity in conjunction with LCCO.

To further verify that LCCO-carbon composite catalyst operates through the pathway involving steps of electrochemical generation of the hydroperoxide and the decomposition of hydroperoxide to oxygen, we have analyzed the relationship between the disk and ring currents using the formalism similar to that developed by Damjanovic and Bockris.^43,44 We have performed the analysis of the oxygen reduction process that involves the following three main steps (also depicted in Figure 9):

Zoom In Zoom Out Reset image size

Step 1: Oxygen is electrochemically reduced to hydroperoxide by a two-electron process.

Step 2: The hydroperoxide generated on the carbon surface is decomposed on the surface of LCCO to produce half a mole of oxygen that will contribute to the electrochemical step of reduction.

Not all the hydroperoxide may be decomposed as this will depend on the kinetics of the decomposition reaction and the amount of LCCO present. We associate this reaction with a heterogeneous rate constant k_p.

Step 3: The hydroperoxide that is not decomposed to oxygen diffuses to the bulk and is detected on the ring electrode

The current observed at the disk electrode, I_disk, results from the current for the reduction of oxygen to hydroperoxide from Step1, I₁, and the current generated from the oxygen produced by decomposition of hydroperoxide from the Step 2, I₂.

The rate of generation of hydroperoxide in the volume of the boundary layer, I₃ is the difference between the rate at which hydroperoxide is produced at the disk and the rate at which it is decomposed by the transition metal oxide. Consequently, the current in Step 3 is given by,

In the steady state, the rate of diffusion of hydroperoxide across the boundary layer of the disk electrode is given by,

where A_disk is the area of the electrode where hydroperoxide is produced, D_HO2- is the diffusion coefficient for hydroperoxide, and C_HO2- is the concentration of hydroperoxide at the surface of the disk, δ is the boundary layer thickness at the rotating electrode, n is the number of electrons in the oxidation of hydroperoxide, and F is the Faraday constant.

The value of I₂ is determined by the rate of decomposition of hydroperoxide to oxygen and can be expressed in terms of the rate constant for decomposition by,

Where A_i is the area of the perovskite participating in the decomposition process and k_p is the heterogeneous rate constant for decomposition with the unit cm s⁻¹. Thus A_i will be determined by the mass of catalyst and its dispersion on the carbon.

The current observed at the ring, I_ring, is obtained from the mass transport of un-decomposed hydroperoxide across the boundary layer into the bulk and also the collection efficiency N at the ring electrode.

From equations 10 and 11 we obtain:

Therefore, by combining equations 8, 9, 12 and 13 we obtain,

For a rotating disk electrode,

where υ is the kinematic viscosity of the electrolyte, and ω is the rotation rate.

Therefore, from equations 14 and 15,

Therefore, a plot of I_disk/I_ring vs. ω^−1/2 is expected to be linear. The slope of this line will depend on the heterogeneous rate constant for the transition metal oxide and the effective interfacial area involved in decomposition as determined by the amount of the decomposer (transition metal oxide). Thus, the slope will also depend on the dispersion of the transition metal oxide to the carbon catalyst, as determined by A_i. However, the heterogeneous rate constant k_p will be specific to the type of oxide material and its surface area.

We find that plot of I_disk/I_ring vs. ω^−1/2 could be fitted with a line for various values of electrode potential (Figure 10a), and the slope of this line at any particular potential increases in proportion to the amount of LCCO for a fixed amount of carbon (Figure 10b) and is consistent with Eq. 16.

Zoom In Zoom Out Reset image size

Table II lists the value of the rate constant for decomposition of peroxide by LCCO as calculated from the slope of the lines in Figure 10b and Eq. 16. The calculations used the following in values for the constants : the diffusion coefficient of hydroperoxide D_HO2- of 1.65 × 10⁻⁵ cm² s⁻¹, kinematic viscosity of the electrolyte, υ = 0.0095 cm² s⁻¹, the collection efficiency, N = 0.2, A_disk = 0.1925 cm². The interfacial contact area, A_i, for each sample was calculated from the mass of LCCO and previously measured values of electrochemically active surface area of 10 m²/g reported in our earlier publication.⁸ The rate constant values were in the range of 0.0009 to 0.0025 cm s⁻¹. These values of rate constants are in close agreement with those for platinum surfaces in alkaline media.⁴⁵

The Tafel slopes were determined from the polarization results for various catalyst compositions (Figure 11).

Zoom In Zoom Out Reset image size

When just LCCO was present in the catalyst layer, the Tafel slope was 94 mV/decade. With just acetylene black in the catalyst layer, the Tafel slope was 60 mV/decade. These values of Tafel slope are consistent with those reported in the literature.³³ For the composite catalysts with a higher fraction of LCCO than acetylene black, the Tafel slope was in the range of 80–95 mV/decade consistent with that observed with LCCO. However, with compositions rich in acetylene black, the values were closer to 60 mV/decade. The apparent exchange current density calculated from the intercept of the Tafel line was one order of magnitude lower for pure LCCO compared to acetylene black, suggesting that the direct reduction of oxygen to peroxide was significantly slower on the LCCO compared to carbon.

The results and analysis of the electro-reduction of oxygen in aqueous alkaline media establish that the composite catalysts prepared by mixing electrically conductive LCCO with acetylene black exhibit two to ten times higher electrocatalytic activity compared to catalysts consisting of just LCCO or carbon. We establish that the apparent synergistic effect arises from the acetylene black serving the role of the primary electrocatalyst for oxygen reduction and LCCO acting as a co-catalyst. Specifically, the surface of acetylene black and other conductive carbon additives supports the electro-generation of hydroperoxide anion while the transition metal oxide rapidly decomposes the hydroperoxide anion to oxygen to enhance the total current. We conclude that the role of carbon in the composite catalyst is substantially different from the commonly perceived additive that increases the electrical conductivity of the catalyst layer. We have verified the well-defined roles of the transition metal oxide and carbon by the analysis of the results of rotating-ring disk experiments and determined the heterogeneous rate constants for the decomposition of hydrperoxide on the surface of LCCO. We find that LCCO is about as active as platinum for decomposing hydroperoxide. We can anticipate these rate constants to vary with the composition of the oxide, and also accordingly the observed activity for oxygen reduction. We expect our findings to be broadly relevant to understanding and design of transition metal oxide based catalysts for oxygen reduction in alkaline media and specifically applicable to the development of air cathodes for rechargeable metal-air batteries.

The research presented here was supported by the Department of Energy, ARPA-E GRIDS Program (DE-AR0000136), the Loker Hydrocarbon Research Institute and the University of Southern Califiornia, Los Angeles, CA. At the Jet Propulsion Laboratory was under a contract with the National Aeronautics and Space Administration, California Institute of Technology, with funding from the ARPA-E Grids program.