Theoretical Study of Oxygen Reduction Reaction Catalysts: From Pt to Non-precious Metal Catalysts

Abstract

Fuel cells are regarded as one of the most promising candidates for stationary and mobile power generation due to their high energy yield and low environmental impact of hydrogen oxidation. The oxygen reduction reaction (ORR) at cathode is a very complex process and plays a crucial role during operation of the PEM fuel cells. However, its mechanism and the nature of intermediates involved remain vague. This chapter focuses on the recent theoretical modeling studies of ORR catalysts for PEMFC. Recent theoretical investigations on oxygen reduction electrocatalysts, such as Pt-based catalysts, non-Pt metal catalysts (Pd, Ir, CuCl), and non-precious metal catalysts (transitional metal macrocyclic complexes, conductive polymer materials, and carbon-based materials), are reviewed. The oxygen reduction mechanisms catalyzed by these catalysts are discussed based on the results.

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

11.1 Introduction

Fuel cells are electrochemical devices that directly convert the chemical energy of a fuel into electrical energy by electrochemical reactions. Since their invention in the middle of the nineteenth century, fuel cells have been identified as the most promising energy conversion technology and have been successfully used in many specific areas, including space exploration. Fuel cells offer many advantages such as high efficiency and reliability, minimum impact to environment, unique operating characteristics, and so on.

There exist a variety of fuel cells. For practical reasons, fuel cells are classified by the type of electrolyte employed. The following names and abbreviations are frequently used in publications: alkaline fuel cells (AFC), molten carbonate fuel cells (MCFC), phosphoric acid fuel cells (PAFC), solid oxide fuel cells (SOFC), and proton exchange membrane fuel cells (PEMFC). Among different types of fuel cells under development today, the PEMFC, also called polymer electrolyte membrane fuel cells (PEFC), is considered as a potential future power source due to its unique characteristics [1–3]. The PEMFC consists of an anode where hydrogen oxidation takes place, a cathode where oxygen reduction occurs, and an electrolyte membrane that permits the transfer of protons from anode to cathode. PEMFC operates at low temperature that allows rapid start-up. Furthermore, with the absence of corrosive cell constituents, the use of the exotic materials required in other fuel cell types is not required [4].

Although PEMFC has many advantages compared with other kinds of fuel cells, there are several barriers challenging PEMFC’s commercialization. These barriers are performance, cost, and durability. PEMFCs are sensitive to poisoning by trace levels of contaminants including CO, sulfur species, and ammonia. The high overpotential from oxygen reduction at cathode contributes significantly to performance losses [5]. Under PEMFC’s operating environment, platinum can react with water and oxygen forming a surface oxide layer that inhibits the catalysis of the oxygen reduction reaction (ORR) especially at high potential [6, 7]. Thus development of more efficient ORR electrocatalysts and reducing their cost are the major focuses for advancing PEMFC technology.

ORR is a very complex process. It involves multi-electrons, various intermediates, and many possible pathways [8–10]. Different experimental characterization methods have been employed to study the ORR mechanism, such as ultraviolet photoemission spectroscopy (UPS) [11], X-ray photoemission spectroscopy (XPS) [12, 13], electron energy loss spectroscopy (EELS) [14], near-edge X-ray absorption spectroscopy (NEXAFS), and Auger spectroscopy (AES) [15, 16]. Despite all these studies, the nature of intermediates involved in the reduction steps remains vague. First-principle theoretical modeling is a good tool complementing state-of-the-art experimental techniques [17]. It provides important insights regarding electronic structure, geometries of reaction intermediates, and adsorption energies at metal/gas or metal/liquid interfaces. Theoretical modeling is becoming an indispensible tool in electrochemical study [18].

This chapter focuses on the theoretical modeling studies of ORR catalysts for PEMFC. Theoretical methods, such as density functional theory (DFT) and ab initio molecular dynamics (AIMD) simulation, are presented. Current understanding of ORR mechanism in acidic medium is briefly discussed. Recent theoretical investigations on oxygen reduction electrocatalysts, such as Pt-based catalysts, non-Pt metal catalysts (Pd, Ir, CuCl), and non-precious metal catalysts (transitional metal macrocyclic complexes, conductive polymer materials, and carbon-based materials), are reviewed. The oxygen reduction mechanisms catalyzed by these catalysts are discussed based on the results.

11.2 Computational Methods

11.2.1 Density Functional Theory Methods

DFT has become increasingly the method of choice for the solution of large systems, not only because of its sufficiently high accuracy but also of its computational simplicity. DFT reduces the calculation of system of many interacting particles to the solution of single-particle Hartree-type equations. Figure 11.1 shows the usage of DFT method from the year of 1980, as compared with the application of Hartree–Fock method.

Fig. 11.1

Number of publications that employ DFT each year (from 1980 to September 23, 2011) based on search from Web of Science database (http://apps.webofknowledge.com) using “density,” “functional,” and “theory” as the keywords. This is compared with a similar search for keywords “Hartree” and “Fock,” using Web of Science database

In DFT, a many-electron (N) system is considered as a fictitious system of N noninteracting electrons which is described by a single determinant wave function with N “orbitals” ϕ _i . The electron density of this fictitious state can be obtained exactly by solving a set of one-electron Schrödinger equations (Kohn–Sham orbital) [19]. The Kohn–Sham self-consistent equation can be described as

$$\left[{-\frac{1}{2}{\nabla^2} + {V_{\rm{eff}}} (r)} \right] {\varphi_i} = {\varepsilon_i} {\varphi_i}$$

(11.1)

The effective external potential, V _eff(r), is described as

$${V_{\rm{eff}}} (r) = V(r) + \int {\frac{{\rho ({r}^{\prime})}} {{\left| {r-{r}^{\prime}}\right|}}} {\rm d} r\prime + {V_{\rm{XC}}} (r)$$

(11.2)

where V(r) is the external potential, $\int {\frac{{\rho ({r}^{\prime})}} {{|r-{r}^{\prime}|}}} {\rm d}{r}^{\prime}$ is Hartree potential, and V _XC(r) is the exchange–correlation potential. The ground-state energy is given by

$$E = \sum\limits_i^N {{\varepsilon_i} - \frac{1}{2}} \iint {\frac{{\rho (r)\rho ({r}^{\prime})}} {{|r-{r}^{\prime}|}}} {\rm d} r {\rm d} {r}^{\prime} + {E_{\rm{XC}}} [\rho] - \int {{V_{\rm{XC}}} (r) \rho (r) {\rm d} r}$$

(11.3)

In spite of the appearance of simplicity, the Kohn–Sham equation is in principle “exact” provided that the exact exchange–correlation energy E _XC is used. In other words, the only error in Eq. (11.3) is due to the approximate nature of E _XC.

Since the exact exchange–correlation term is unknown, several different schemes have been developed to obtain approximation of the E _XC. There are several types of approximate functionals such as local density approximation (LDA) and generalized gradient approximation (GGA) [20]. The LDA is very successful for many systems of interest, especially those with uniform electronic density distributions such as those in bulk metals, but LDA fails in systems where weak molecular bonds exist [21]. To account for the inhomogeneity of the electron density, a nonlocal correlation is often added to the exchange energy; this is the so-called generalized gradient approximation (GGA). There are several exchange–correlation functionals such as P86 [22], PW91 [23, 24], BPW91 [25], LYP [26], BLYP [20, 26], B3LYP [25–27], and PBE [28]. In general, as the exchange–correlation functional contains both exchange and Coulomb correlation terms, the DFT methods provide better results than that of a single determinant Hartree–Fock (HF) methods which do not have Coulomb correlation term [5].

The selection of functional directly controls the accuracy of the DFT methods. A study of ORR by hybrid density functional (HDFT) method and MP2 method has been reported by Albu et al. [29, 30]. They evaluated a large number of HDFT methods toward calculation of potential-dependent activation energies for uncatalyzed and Pt-catalyzed oxygen reduction and hydroperoxyl oxidation reactions. In the HDFT methods, the one-parameter hybrid Fock–Kohn–Sham operator is written as

$$ F={F^{\mathrm{ H}}}+\left( {\frac{X}{100 }} \right){F^{\mathrm{ H}\mathrm{ FE}}}+\left[ {1-\left( {\frac{X}{100 }} \right)} \right]({F^{\mathrm{ SE}}}+{F^{\mathrm{ GCE}}})+{F^{\mathrm{ C}}} $$

(11.4)

where F ^H is the Hartree operator (i.e., the non-exchange part of the Hartree–Fock operator), F ^HFE is the Hartree–Fock (HF) exchange operator, X is the fraction of Hartree–Fock exchange, F ^SE is the Dirac–Slater local density functional for exchange, F ^GCE is the gradient correction for the exchange functional, and F ^C is the total correlation functional including both local and gradient-corrected parts.

Table 11.1 lists results obtained from MP2/6-31G(d,p) and various HDFT methods for uncatalyzed ORR. Table 11.2 contains results obtained with MP2/6-31G(d,p) and a subset of HDFT methods for Pt-catalyzed ORRs. In the tables, HDFT methods are named by the gradient-correlated exchanged functional used, followed by 1 (indicating a one-parameter method), followed by the correlation functional used, and finally by the value of X separated by a dash. For example, mPW1B95-50 represents a one-parameter hybrid DFT method based on the modified Perdew–Wang (mPW) gradient-corrected exchange functional [31] and B95 gradient-corrected correlation functional [32], and with a Hartree–Fock exchange contribution of 50 % (X = 50).

Table 11.1 Comparison of calculated reversible potential for O₂ + 4H⁺ + 4e⁻ → 2H₂O reaction, electrochemical potentials (V) and activation energies (eV) for O₂ + H₃O⁺ + e⁻ → OOH + H₂O reaction, and the average relative error to the MP2/6-31G(d,p) results [29]

Table 11.2 Electrochemical potentials and activation energies (eV) for Pt–O₂ + H₃O⁺ + e⁻ → Pt–OOH + H₂O reaction and the average relative error to the MP2/6-31G(d,p) results [29]

The results indicate that the inclusion of diffuse functions for oxygen by using the 6-31+G(d,p) basis set is unnecessary and gives inferior results. The HDFT methods with low HF exchange contributions show typically larger errors in comparison with the results obtained with the MP2 method. These errors are a consequence of a systematic underestimation of the transition-state energy for oxygen reduction to hydroperoxyl and hence its reverse hydroperoxyl oxidation reaction. Among the 25 HDFT methods investigated, the best results were obtained using the HDFT method based on Becke’s exchange functional, B95 correlation functional, and a Hartree–Fock exchange contribution of 50 % in conjunction with the 6-31G(d,p) basis set [B1B95-50/6-31G(d,p)]. Results with similar accuracy were obtained using the HDFT method based on Becke’s exchange functional, PW91 correlation functional, and a Hartree–Fock exchange contribution of 50 % in conjunction with the 6-31G(d,p) basis set [B1PW91-50/6-31G(d,p)]. Both of these methods also give very accurate gas-phase bond energies for the oxygenated species. Based on the calculated results, these two methods are recommended for study of electrochemical ORR.

11.2.2 Ab Initio Molecular Dynamics Method

Recently, ab initio or the first-principle molecular dynamics (AIMD), which is based on Car and Parrinello’s technique [33], has become a valuable method for study condensed matter structure and dynamics, in particular liquids, surfaces, and clusters. The basic idea underlying AIMD method is to compute the forces acting on the nuclei from electronic structure that are calculated “on the fly” as the molecular dynamics trajectory is generated [34]. In this way, the electronic variables are not integrated out beforehand, but are considered as active degrees of freedom. Thus, the Car and Parrinello (CP) algorithm overcomes limitations of standard empirical-potential approaches employed in classical molecular dynamics and provides direct information about the electronic structure. Figure 11.2 depictures the number of publications related to AIMD method.

Fig. 11.2

Publication and citation analysis. Squares: number of publications with usage of AIMD each year from 1991 to September 28, 2011, which is based on Web of Science database (http://apps.webofknowledge.com) using “ab initio” and “molecular dynamics” (or synonym such as “first-principle MD” and “Car–Parrinello simulations”) as the keywords. Diamonds: number of publications which cite the 1985 paper by Car and Parrinello [33]

The CP method makes use of the following classical Lagrangian

$${L_{\rm{CP}}} = \sum\limits_i {\frac{1}{2}} \mathop{\mu}\nolimits_i \langle {\psi_i} \vert {\psi_i} \rangle + \frac{1}{2} \sum\limits_I {{M_I}} \mathop{{R_I^2}}\limits^{.}-E({\psi_0},\;\vec{R}) + {\rm constraints}$$

(11.5)

to generate trajectories for the nuclear and electronic degrees of freedom via the coupled set of equations of motion:

$${M_I} {{\ddot{R}}_I} (t) = - \frac{\partial } {{\partial {R_I}}} \langle {\psi_0} \vert H \vert {\psi_0} \rangle + \frac{\partial }{{\partial {R_I}}} \left\{ {\rm{constraints}} \right\}$$

(11.6)

$$\mathop{\mu}\nolimits_i {{\ddot{\psi}}_i} (t) = - \frac{\partial}{{\partial \psi_i^{*}}} \langle {\psi_0} \vert H \vert {\psi_0} \rangle + \frac{\partial }{{\partial \psi_i^{*}}} \left\{ {\rm{constraints}} \right\}$$

(11.7)

where $\vert {\psi_0}\rangle = {\rm det} \left\{ {{\psi_i}} \right\}$ is the total wave function; M _I and R _I are atom I’s nuclear mass and position, respectively; μ _i are the “fictitious electron masses” or inertia parameters assigned to the orbital degrees of freedom; and E is the energy functional of the system. For a given nuclear geometry, the ground-state electronic wave function can be obtained by standard techniques such as steepest descent or other higher order methods:

$$ {{\dot{\psi}}_i}(t)=-\frac{1}{2}\frac{{\delta E}}{{\delta \psi }}+\sum\limits_j {{\Lambda_{ij }}} \psi (t)=-\frac{1}{2}H{\psi_i}(t)+\sum\limits_j {{\Lambda_{ij }}} \psi (t) $$

(11.8)

where ${\Lambda_{ij }}$ are Lagrange multipliers which impose orthonormality constraints on the electronic orbitals. Thus, once the ground-state wave function has been determined, the dynamical behavior of a system can be studied at a given time; the force acting on the nuclei is given by Eq. (11.6). In the dynamical simulation, both the electronic and nuclear (ionic) degrees of freedom can propagate concurrently [35].

Compared to ordinary quantum chemical approaches, the AIMD free-energy calculations have significant advantage as entropic effects and anharmonicity are explicitly included in the calculation. In contrast, a normal static DFT procedure requires that the vibrational entropy be added via harmonic frequency calculations and the computational demands may be quite substantial for large molecular systems. Furthermore, the harmonic approximation may not even be valid in situations where weak interactions are dominant.

11.3 ORR in Acidic Medium

The ORR is a multi-electron reaction that includes a number of elementary reactions. Yeager proposed two pathways for the ORR in acidic medium [36]: (1) a “direct” four-electron pathway where O₂ is reduced directly to water without involvement of hydrogen peroxide (H₂O₂), O₂ + 4H⁺ + 4e⁻ → 2H₂O and (2) a “series” pathway in which O₂ is reduced to H₂O₂, O₂ + 2H⁺ + 2e⁻ → H₂O₂, followed by its further reduction to H₂O, H₂O₂ + 2H⁺ + 2e⁻ → 2H₂O. Adzic et al. proposed a parallel pathway, with the direct and series mechanisms occur simultaneously, with the direct pathway as the dominant one [37].

For transition metal catalysts, two-electron reduction was reported for less active metals such as Au and Hg. For the most active catalyst, Pt and Pt-based alloys, four-electron reduction steps are generally believed. However, the detailed mechanism and reduction pathways are not clear and much debate remains. Even for the first electron transfer step, different views still exist [38–40]. Examples of plausible first steps include the following: (1) splitting of the O–O bond upon oxygen adsorption on two Pt sites (S) in a bridge configuration, O₂ + 2S → O* + O*; (2) formation of the superoxide anion, O₂ + 2S + e⁻ → O₂ ⁻; and (3) simultaneous electron and proton transfer, O₂ + 2S + (H⁺ + e⁻) → OOH.

O₂ reduction pathways were affected to a great extent by the O₂ adsorption modes on catalyst surfaces. There are three adsorption models for molecular oxygen adsorption (Fig. 11.3):

Fig. 11.3

Models of O₂ adsorption on electrode surfaces

• 1. O₂ interacts with a single substrate atom (Griffith model) [41] forming a bond between its σ-orbitals and the empty ${d_{{{z^2}}}}$ orbitals of the metal surface atom and by forming a σ-backbond from the partially filled d _xy or d _yz metal orbitals to the antibonding π* orbitals of O₂.

• 2. End-on adsorption through a single σ-type bond (Pauling model), in which the σ-orbital of O₂ donates electron to the ${d_{{{z^2}}}}$ orbital on the metal.

• 3. Bridge model, with O₂ bonds with two sites, which was proposed by Yeager [42] principally for the reaction on platinum group metals.

In order to study the oxygen reduction mechanism, it is crucial to know the nature and coverage of adsorbed reaction intermediates. However, there is no simple adequate spectroscopic method for identifying adsorbed intermediates. Computational studies can provide insights regarding intermediates, their geometries, and energies. Anderson and Albu [30] carried out ab initio studies of reversible potentials and activation energies for uncatalyzed oxygen reduction to water and the reverse oxidation reaction. The electrode was modeled by a noninteracting electron-donor molecule with a chosen ionization potential (IP). When the reactant reached a point on the reaction path where its electron affinity (EA) matched the donor’s IP, an electron transfer was assumed occurring, and the donor’s IP or reactant’s EA was identified with the electrode potential. To locate transition state, potential surface scan is deployed [43]. Once the transition state is located, the energy difference between the transition-state structure and reactants energy is taken as the activation energy. They calculated the activation energy of the following one-electron transfer steps:

$$ {{\mathrm{ O}}_2} (\mathrm{ g}) + {{\mathrm{ H}}^{+}} (\mathrm{ aq}) + {{\mathrm{ e}}^{-}} ({\mathit U}) \to \mathrm{ H} {{\mathrm{ O}}_2}^{\bullet } (\mathrm{ aq})$$

(11.9)

$$\mathrm{ H} {{\mathrm{ O}}_2}^{\bullet } (\mathrm{ aq}) + {{\mathrm{ H}}^{+}} (\mathrm{ aq}) + {{\mathrm{ e}}^{-}} ({\mathit U}) \to {{\mathrm{ H}}_2} {{\mathrm{ O}}_2} (\mathrm{ aq})$$

(11.10)

$${{\mathrm{ H}}_2} {{\mathrm{ O}}_2} (\mathrm{ aq}) + {{\mathrm{ H}}^{+}} (\mathrm{ aq}) + {{\mathrm{ e}}^{-}} ({\mathit U}) \to \mathrm{ H} {{\mathrm{ O}}^{\bullet }} (\mathrm{ g}) + {{\mathrm{ H}}_2} \mathrm{ O} (\mathrm{ aq})$$

(11.11)

$$\mathrm{ H} {{\mathrm{ O}}^{\bullet }} (\mathrm{ g}) + {{\mathrm{ H}}^{+}} (\mathrm{ aq}) + {{\mathrm{ e}}^{-}} ({\mathit U}) \to {{\mathrm{ H}}_2} \mathrm{ O} (\mathrm{ aq})$$

(11.12)

The MP2/6-31G** method is used. The H⁺(aq) was modeled with H₃O⁺(H₂O)₂ cluster. The electron transfer was assumed to occur when the electron affinity of the reaction complex equaled the ionization potential, IP, of the electrode.

They studied activation energies for the four-electron transfer steps at potentials range between 0 and 2 V (vs. standard hydrogen electrode, SHE). The results showed that as the electrode potential increases, the activation barriers increase as well. The third electron and proton transfer step (Eq. 11.11), i.e., the H₂O₂ reduction, has the highest activation energy. Over the potential range they considered, activation energies for the reduction reactions decrease in the sequence: third step (Eq. 11.11) > first step (Eq. 11.9) > second step (Eq. 11.10) > fourth step (Eq. 11.12). From the calculation results, it is evident that an efficient four-electron reduction catalyst must not liberate hydrogen peroxide and it must activate the first and third reduction steps without deactivating the other two steps. Complete dissociation of HOOH on the electrode surface should lead to good activity, based on the low activation energy calculated for HO^• reduction.

In a subsequent study, Anderson and Albu [44] employed a similar method to study the catalytic effect of Pt on ORR. A single Pt atom is used to model the system. They calculated activation energies of the following four reduction steps:

$$ \mathrm{ Pt}-{{\mathrm{ O}}_2}+{{\mathrm{ H}}^{+}}(\mathrm{ aq})+{{\mathrm{ e}}^{-}}(\it U)\to \mathrm{ Pt}-\mathrm{ OOH} $$

(11.13)

$$ \mathrm{ Pt}-\mathrm{ OOH}+{{\mathrm{ H}}^{+}}(\mathrm{ aq})+{{\mathrm{ e}}^{-}}(\it U)\to \mathrm{ Pt}-\mathrm{ OHOH} $$

(11.14)

$$ \mathrm{ Pt}-\mathrm{ OH}\mathrm{ O}\mathrm{ H}+{{\mathrm{ H}}^{+}}(\mathrm{ aq})+{{\mathrm{ e}}^{-}}(\it U)\to \mathrm{ Pt}-\mathrm{ OH}+{{\mathrm{ H}}_2}\mathrm{ O} $$

(11.15)

$$ \mathrm{ Pt}-\mathrm{ O}\mathrm{ H}+{{\mathrm{ H}}^{+}}(\mathrm{ aq})+{{\mathrm{ e}}^{-}}(\it U)\to \mathrm{ Pt}-\mathrm{ O}{{\mathrm{ H}}_2} $$

(11.16)

It is evident that the Pt atom has a significant catalytic effect on the most difficult step of the ORR, i.e., H₂O₂ → HO^• + H₂O. Activation energies drop about 1 eV over the 0–2 V potential range studied. Activation energies for the other reduction steps are also substantially reduced. The activation energies for HO^• reduction to H₂O increase when HO^• is bonded to the Pt atom and bonding to Pt decreases the activation energies for OOH^• reduction to H₂O₂. In all cases, the activation energies are predicted to increase with increasing potential (Fig. 11.4).

Fig. 11.4

Activation energy for the four steps of oxygen reduction to water as a function of electrode potential, U. Heavy lines connect points with species undergoing reduction bonded to a Pt atom. Dotted lines connect points with no bonding to the Pt. The same key applies to both sets of curves [44]

Sidik and Anderson further studied the O₂ electroreduction when bonded to Pt dual sites [45]. A Pt dimmer, Pt₂, with the bulk distance of 2.775 Å, was used to provide one- and twofold bonding sites for coordinating O₂, O, OOH, and OH. The results suggest that the O₂ on dual sites does not dissociate before the first electron transfer and the product for this step, OOH, easily dissociates with a small 0.06 eV activation barrier to form O and OH. The first electron transfer step has the highest activation barrier (0.60 eV at 1.23 V), which is close to the experimental value, and is predicted to be the rate-determining step.

Li and Balbuena [46] used DFT method with B3PW91 functional, LANL2DZ effective core pseudopotentials for Pt, and 6-311G(d) basis set for O and H to study the first electron transfer step:

$$ \mathrm{ P}{{\mathrm{ t}}_5}-{{\mathrm{ O}}_2}+{{\mathrm{ H}}^{+}}(\mathrm{ aq})+{{\mathrm{ e}}^{-}}\to \mathrm{ P}{{\mathrm{ t}}_5}-\mathrm{ OOH} $$

(11.17)

The solvent influence on the reaction is considered with H₃O⁺(H₂O)₂ ion cluster model. It is found that the electron transfer process takes place gradually as the hydronium ion gets close to the adsorbed oxygen. The activation energies show a strong dependence on the O_ads⋯H⋯O_water distance between the adsorbed oxygen and the proton water. The proton transfer reaction is not involved in the rate-determining step if the O_ads···H···O_water distance is smaller than a certain threshold, which depends on the degree of proton solvation. Negatively charging the cluster/adsorbate complex causes a sharp decrease in the activation barrier.

Jinnouchi and Okazaki [47] performed AIMD studies of the first electron transfer reaction with one oxygen molecule, 1 hydronium ion, 9 water molecules, and 12 Pt atoms at 350 K. They proposed that the first reaction step would be rapid oxygen adsorption on the catalyst induced from the strong attractive force between the oxygen molecule and the platinum surface. The adsorbed water molecules and the hydronium ion hydrated the adsorbed oxygen atoms, and proton transfer through the constructed hydrogen bonds frequently occurred. When the conformation of these species satisfied certain conditions, the oxygen dissociation with the proton transfer reaction was induced, and three OH were generated on the platinum surface. The authors suggested that the generation of ordered cluster structure that provides the effective path for ions and water molecules is the most important guiding principle for designing high ion conductive electrolytes.

11.4 ORR Catalyzed by Metal Catalysts

11.4.1 Platinum

Platinum has long been known as the best electrocatalyst for oxygen reduction in acidic medium [48]. The first step of catalyzing oxygen reduction involves chemisorption of oxygen molecule on the electrocatalyst surface. The adsorbate’s structure, bonding type, and energy are key elements in understanding the effects of adsorption on the reaction kinetics. Theoretical studies of chemisorption of oxygen molecule can provide information about the nature of bonding at the surface, bond strength, geometry, and site preferences of adsorbates. Furthermore, theoretical calculations have advantage that the study can be performed for situations not realizable experimentally and can eliminate concerns about the effects of contaminants or other unknown variables.

Li and Balbuena [49] performed DFT calculation to study the interactions of oxygen with platinum clusters. They firstly calculated the ground-state structures of Pt _n clusters (n = 4–6). The results indicated that the structures for Pt _n clusters are nonplanar and most of them exhibit Jahn–Teller distortions. For atomic oxygen, a bridge-site adsorption is found as the most stable structure for Pt₃O. The adsorption energy for atomic oxygen shows a strong dependence on the metal cluster size and geometry. The bridge site is found to be the most favorable site for the adsorption of O₂ on Pt _n . In the dimer complexes Pt₂O₂, they found that the electron transferred from the Pt’s s orbitals to the oxygen p orbitals, whereas in Pt _n O₂ (n > 2), the electron transferred from the Pt’s d and p orbitals. The potential energy surface for the adsorption of oxygen shows a double well feature. The first potential energy minimum occurs at adsorption distances relatively far from the metal cluster. As the molecule gets closer to the cluster, the second minimum, which is more stable than the first one, appears corresponding to adsorbed state. A bond-order conservation approximation analysis illustrates that the dissociation activation energies depend on the Pt cluster size. The activation energies are reduced as the cluster’s ionization potential (IP) decreases, which facilitates the charge transfer to the adsorbate. A similar effect on dissociation is observed upon application of an electric field along the direction of increase of the dipole moment of the adsorbate–cluster complex.

Similarly, Wang et al. [50] used B3LYP functional to study the adsorption and decomposition of the radical OOH and H₂O₂ on Pt clusters (Pt₃, Pt₄, Pt₆, and Pt₁₀). The potential energy surface for the complete reduction of oxygen in acidic medium has been calculated (Fig. 11.5). They found that the OOH radical has a very strong adsorption on Pt clusters. Both thermodynamics and kinetics indicate that OOH readily decomposes into O and OH before a second electron transfer takes place. Thus, although a one-end adsorbed hydrogen peroxide H₂O₂, arising from the reduction of adsorbed OOH, has been located on Pt₃ and Pt₁₀, a favorable pathway for the second electron transfer should occur on the coadsorbed O and OH species, suggesting that O₂ reduction on a Pt surface may proceed via a parallel pathway: the direct (no H₂O₂ generated as intermediate) and series (H₂O₂ generated) occurring simultaneously, with the direct as the dominant step.

Fig. 11.5

Potential energy surface profile for the oxygen reduction reaction at the standard hydrogen electrode potential scale: the proton was modeled by two shells of water molecules, H⁺OH₂(H₂O)₃(H₂O)₆, and the data in parentheses are Gibbs free energies [50]

As discussed above, the oxygen adsorption and reduction processes are often simulated either on small Pt clusters or flat surfaces. However, both experimental measurements [51, 52] and computational calculations [53, 54] indicate that nanosized electrocatalysts show a considerably different catalytic activity from extended flat surfaces. These investigations would suggest that effects observed with particle size reduction go well beyond the increase in surface area and involve fundamental physical and chemical changes in the reaction steps. Han and his coworkers [55] studied explicitly Pt nanoparticles with 1 and 2 nm sizes and compared their chemical adsorption properties to those of an extended flat Pt(111) surface. As atomic oxygen (O) and hydroxyl group (OH) are two species of considerable importance [56], they focused on effect of particle size and Pt coordination on the chemisorption energies of O and OH.

Based on their calculations, the chemisorption energies of O and OH on the Pt(111) surface vary considerably from on that of 1 and 2 nm particles. With only sites near the center of the (111) facet on the 2 nm particle display similar adsorption energies to that of the Pt(111) bulk surface. While they find that on the bulk surface the fcc site is the most stable adsorption site, in agreement with previous work [57–59], oxygen adsorption is strongest at bridge sites on the edges of the 2 and 1 nm particles. For both particles, the maximal adsorption energy is larger than on the bulk surface. Depending on the location of sites, the adsorption energies at the fcc and hcp sites of the nanoparticles can be larger or small than that of the bulk surface value, with adsorption getting stronger as the site location getting closer to the edge of the surface.

While, in general, the adsorption energy on the 1 nm particle on equivalent sites is higher than on the 2 nm particle, this trend is reversed for the fcc (111) site. This peculiarity is attributed to the significant curvature on the surface of 1 nm particle in order to reduce the total surface area of the particle. This causes the distance between nearest neighbor Pt atoms to be longer on the surface of the 1 nm particle (2.85 Å) than on the 2 nm particle (2.7 Å). This bond elongation prevents the effective interaction between oxygen and the three Pt atoms defining the site.

The effect of particle size on adsorption energy and catalytic activity is complex. It depends on the precise role that the adsorbed species play in the catalytic process and on the change in adsorption of other relevant species. Watanabe et al. [60] argued that particle size has no effect on the rate of the ORR; however, the mean particle spacing on the support is the more relevant variable though his findings have been disputed by Giordanoa et al. [61] and Takasu et al. [62]. Similarly, Yano et al. [63] studied the ORR catalytic activity and the ratio of peroxide (H₂O₂) formation on Pt particles with 1.6, 2.6, and 4.8 nm in diameter. By NMR analysis of ¹⁹⁵Pt located at surfaces of the Pt nanoparticles, they found that the electronic structure of Pt does not change with particle size, from which they concluded that the ORR activity does not change with Pt particle size. Mayrhofer et al. [64], on the other hand, argued that the rate of the ORR decreases with decreasing in particle size as stronger OH adsorption blocks oxygen adsorption sites. Based on the analysis above, the authors concluded that there might be no general relationship between the particle size and the catalytic activity. Nevertheless, clever engineering of nanostructure of catalyst materials to optimize the adsorption strength and increase number of adsorption sites for the intermediate species involved could, in principle, be used to fabricate better electrocatalysts. Certainly such an approach would require a better understanding of reaction mechanism and effect of nanostructure on adsorption of various intermediate species.

11.4.2 Platinum-Based Metal Catalysts

The polarization in a PEMFC comes mainly from the slow cathodic ORR. Reducing the Pt loading (particularly in the cathode catalyst layer) without compromising fuel cell performance is a strategy to meet the cost reduction requirements for PEMFC commercialization [65, 66]. In other words, the design of novel catalysts requires not only reducing the amount of Pt used but also enhancing catalytic activity and stability [67]. Several approaches have been developed to address reduction of Pt loading, including (1) improving nanostructures to increase the surface-to-volume ratio of catalyst materials [68]; (2) using an alloying technique to incorporate non-precious metals into the nanostructures; and (3) texturing nanostructures, e.g., replacing the core atoms in Pt nanoparticles with a non-precious metal, resulting in a Pt-skin [69, 70] or Pt monolayer catalysts [71–73].

Several Pt-based bimetallic systems have been reported showing good ORR activity. For instance, alloying Pt with transition metals such as Co [74, 75], Ni [76–78], Fe [79], Mn [79], Cr [80–82], and V [83] is highlighted as a promising approach toward improving ORR electrocatalysis in acidic solutions. A number of explanations have been proposed for the observed improvement in activity by the addition of a less noble metal to Pt [84]. These include the lower oxidation state of the Pt which can suppress Pt oxide formation, a shortening of Pt–Pt interatomic distance and therefore a more favorable adsorption of O₂ (geometric effect), an increased 5d-orbital vacancies (electronic effect), and formation of a thin Pt skin on the surface of the alloy [85] which has unusual electronic properties. Using the conventional Pt/C catalyst as the baseline, the ORR activities of PtM alloys (M = Ni, V, Co, and Fe) follow the order: Pt/C < PtNi/C < PtV/C < PtCo/C < PtFe/C [86–88].

Wang et al. [89] carried out first-principle DFT calculations to investigate how a subsurface transition metal M (M = Ni, Co, or Fe) affects the energetics and mechanisms of ORR on the outermost mono-surface layer of Pt/M(111). With the calculation, they illustrated that the subsurface Ni, Co, and Fe could downshift the d-band center of the Pt surface and thus weaken the binding of chemical species to the Pt/M(111) surface. Moreover, the subsurface Ni, Co, and Fe could modify the heat of reaction and activation energy of various elementary reactions of ORR. The results revealed that, due to the influence of the subsurface Ni, Co, or Fe, ORR would adopt a hydrogen peroxide dissociation pathway. The activation energy for the rate-determining O₂ protonation reaction is 0.15, 0.17, and 0.16 eV on Pt/Ni(111), Pt/Co(111), and Pt/Fe(111), respectively. In contrast, the activation energy for the same reaction on a pure Pt(111) surface is 0.79 eV. The theoretical study explained why the subsurface Ni, Co, and Fe could lead to multifold enhancement in catalytic activity of ORR on the Pt/M(111) catalysts.

Xu et al. [57] performed self-consistent periodic DFT calculations (GGA–PW91) to study the adsorption of O and O₂ and the dissociation of O₂ on the (111) facets of ordered Pt₃Co and Pt₃Fe alloys and on monolayer Pt skins covering these two alloys. They also investigated explicitly the strain effect by a 2 % compression of Pt(111). They discovered that the Co atoms on the Pt₃Co(111) surface allow O₂ to dissociate more easily than on Pt(111) [the lowest activation energy on Pt₃Co(111) is 0.24 eV/O₂, compared to 0.77 eV/O₂ on Pt(111)] and also bind O and O₂ more strongly (−4.29 eV/O vs. −3.88 eV/O, −0.92 eV/O₂ vs. −0.62 eV/O₂). While for monolayer Pt-skin catalysts, the authors showed that although these catalysts are less reactive than pure Pt for O₂ dissociation, they are more active toward the ORR because they are less poisoned by O and facilitate the activation of O and O-containing intermediates in bond-making elementary reaction steps in the ORR.

The authors also revealed that there is a linear relationship between atomic oxygen binding energy and the oxygen dissociation barrier of the transition metals and alloys (Fig. 11.6). The more strongly a material binds atomic oxygen, the more effective it will be in dissociating molecular oxygen. So instead of conducting a complicated and expensive transition-state study, a more affordable atomic binding study can be used to screen and design better oxygen reduction catalysts.

Fig. 11.6

Binding energies of the transition states of O₂ dissociation ($E_{\mathrm{ b}}^{\mathrm{ TS}}$) vs. binding energies of the atomic final states with respect to gas-phase O₂ ($E_{\mathrm{ b}}^{\mathrm{ FS}}$) on the (111) facets of several fcc transition metals and alloys [57]

Mavrikakis et al. [90] performed spin-polarized DFT calculations to study the ORR catalytic activity of platinum monolayers supported on close-packed transition metal surfaces [(Au(111), Pt(111), Pd(111), and Ir(111)]. To estimate the relative activity of these catalysts at two different cell potentials (1.23 and 0.80 V), both the free-energy changes and activation energy barriers for elementary reactions are calculated. They observed that OH removal is the rate-limiting step for ORR at the potential of 1.23 V. At a cell potential of 0.80 V, the proton/electron transfer steps become easier, and Pt*/Pd shows the highest activity among the surfaces examined, even higher than that of pure Pt(111). The activity of the other two surfaces, namely, Pt*/Au and Pt*/Ir, is lower due to the difficulty in OH removal and in O–O bond scission, respectively. Their findings provide additional insights into the ORR mechanism for these supported platinum monolayer surfaces.

11.4.3 Other Metal Catalysts

Palladium-alloy materials have been recently introduced as a promising ORR cathode electrocatalyst for replacing Pt [91–98]. Pd alloys are considerably less expensive than Pt. Besides, experimental studies indicate that they have high methanol tolerance for direct methanol fuel cells (DMFCs) in which the methanol crossover to the cathode significantly decreases the cell’s efficiency [99, 100].

Shao et al. [101] measured the ORR activity of Pd monolayers supported on Ru(0001), Rh(111), Ir(111), Pt(111), and Au(111) using the rotating disk electrode technique and correlated the findings with the d-band center of Pd calculated from DFT method. They observed the activity of Pd monolayers supported on Ru(0001), Rh(111), Ir(111), Pt(111), and Au(111) increased in the following order: Pd/Ru(0001) < Pd/Ir(111) < Pd/Rh(111) < Pd/Au(111) < Pd/Pt(111). A volcano-type dependence of activity on the energy of the d-band center of Pd monolayers was found, with Pd/Pt(111) at the top of the curve. They also synthesized a Pd₂Co/C alloy, and the results showed that the non-Pt alloy’s ORR activity is comparable to that of commercial Pt-based catalysts. The kinetics of the ORR catalyzed by this electrocatalyst predominantly involves a four-electron reduction with the first electron transfer being the rate-determining step. The downshift of the d-band center of the Pd “skin,” which constitutes the alloy surface due to the strong surface segregation of Pd at elevated temperatures, contributed to its high ORR activity. Additionally, the catalyst showed very high methanol tolerance, retaining very high catalytic activity for the ORR at high concentrations of methanol.

Very recently, Erikat et al. [102] used periodic self-consistent DFT calculations to study the adsorption and the reaction barrier for the dissociation of O₂ on Ir(100) surface. Dissociative adsorption is found to be energetically more favorable than molecular adsorption. Among all the cases studied, parallel approaches (Prl1 and Prl2) on a hollow site, with the same adsorption energy of −3.93 eV, are found to be the most energetically preferred sites of adsorptions. Hybridization between p-O₂ and d-metal orbitals is responsible for the dissociative adsorption. By using the nudge elastic band method (NEB), they found that the dissociation occurs immediately and very early along the dissociation path with a small activation barrier (0.26 eV).

Through DFT slab calculations, the adsorption and dissociation of O₂ on CuCl(111) surface have been systematically studied by Zhang et al. [103]. Different modes of atomic O and molecular O₂ adsorption on CuCl(111) surface and possible dissociation pathways are examined; vibrational frequency and Mulliken charge are also calculated. Their results showed that the favorable adsorption occurs at hollow site for O atom and molecular O₂ lying flatly on the surface with one O atom on top of Cu atom is the most stable adsorption configuration. The O–O stretching vibrational frequencies are significantly red-shifted, and the charges are transferred from CuCl to oxygen. Upon O₂ adsorption, adsorbed species on CuCl(111) surface exhibit the characteristic of the superoxo (O₂ ⁻), which contributes to improving the catalytic activity of CuCl. Meanwhile, O₂ dissociation into atomic O could also occur; however, it needs to overcome a very large activation barrier.

11.5 Non-precious Metal Catalysts

At the current technology stage, Pt-based electrocatalysts are the most practical materials in terms of both activity and stability, although their performance is still insufficient and needs further improvement. The major drawback of these Pt-based catalysts is the limited availability and high cost, contributing to the excessive production costs of fuel cell systems.

The search for inexpensive, high-performance electrocatalysts has taken two different approaches. The first approach is to reduce catalyst usage through increasing Pt utilization in the catalyst layers. This can be achieved by alloying Pt with inexpensive metals (Co, Fe, etc.) and/or by utilizing unique support materials for thin-layer Pt nanoparticle deposition. In the last two decades, the required Pt loading has been reduced significantly to approximately 0.4 mg cm⁻² [104]. Unfortunately, the increasing of Pt price during the same time has totally offset the Pt loading reduction, rendering the efforts toward Pt loading reduction ineffective. Therefore, this approach may not be a viable solution for cost reduction of PEMFC. The other approach is to develop non-precious metal-based electrocatalysts. Although the performance of the non-precious metal catalysts (carbon-supported Fe– and/or Co–N catalysts) is still inferior to Pt-based catalysts in terms of both activity and stability, in the last few years, significant progress has been made in the development of non-precious metal catalysts making it one of the most active and promising research areas [105].

Non-precious metal catalyst research covers a broad range of materials. The most promising catalysts investigated thus far are carbon-supported M–N _x /C materials (M = Co, Fe, Ni, Mn, etc.) formed by pyrolysis of a variety of metal, nitrogen, and carbon precursor materials [106]. Other non-precious metal electrocatalyst materials investigated include non-pyrolyzed transition metal macrocycles [107–122], conductive polymer-based complexes (pyrolyzed and non-pyrolyzed) [123–140], transition metal chalcogenides [141–148], metal oxide/carbide/nitride materials [149–166], as well as carbon-based materials [167–179]. The advances of these types of materials can be found in Chaps. 7–10 and 12–15 of this book.

To meet the need of PEMFC, increasing volumetric density of non-precious metal catalysts is required; however, without knowledge of active-site structure, it is very challenging to increase the active-site density. There are many challenges in determining the active-site structures. There is lack of adequate material characterization techniques. The active sites may be scarce. Furthermore, different active sites may contribute to the catalytic activity of the materials [180]. There are different reviews regarding the nature of active sites. One school believes that metal is at the active sites, while the other believes that metal merely catalyzes the formation of active sites from carbon, nitrogen, and perhaps oxygen atoms [181–184]. Some researchers believe that the ORR activity is originated from the formation of graphitic nitrogen and/or pyridinic nitrogen functional groups. This claim has been supported in the literature with ORR activity of metal-free N/C catalyst materials [185, 186]. This further demonstrates the convoluted understanding regarding the exact nature of the catalytically active moieties with respect to ORR. First-principle modeling is a valuable tool in exploring active-site structures and providing insights into the reaction mechanism.

11.5.1 Transition Metal Macrocycle Catalysts

Transition metal macrocycle compounds have been employed as ORR catalysts since the first study by Jasinski in 1964 [187]. The activity of these complexes has been found to be directly related to the metal ion center and encompassing ligand structure [105]. Simple cobalt-based complexes (i.e., cobalt phthalocyanines or cobalt porphyrins) have catalytic activity toward the reduction of oxygen by a two-electron process to produce H₂O₂, whereas a four-electron reduction process forming H₂O is commonly observed for Fe-based complexes. Although the activity of unpyrolyzed transition metal macrocycle catalysts is inferior to its pyrolyzed counterpart, the major benefit of studying non-pyrolyzed macrocyclic catalyst is for fundamental understanding, as the active sites of these catalysts are well known, which allows a direct link between the catalyst structure and its ORR activity. Thus, theoretical studies have been conducted to understand the relationship between the transition metal macrocyclic complex structures and their electrocatalytic properties.

Sun et al. [188] performed DFT calculation to study the electrochemical reduction mechanism of oxygen on four metal macrocyclic complexes, iron phthalocyanines (FePc), iron porphyrin (FeP), cobalt phthalocyanine (CoPc), and cobalt porphyrin (CoP). Based on their calculation results, for the four metal macrocyclic complexes, O₂ will not directly be cleaved without the cooperation of hydrogen. In the reduction process, with FePc or FeP as catalyst, O₂ is reduced to H₂O (4e⁻ transfer) without the formation of intermediate H₂O₂, while on CoPc or CoP, O₂ is reduced to H₂O₂ (2e⁻ transfer). The authors attributed the four-electron reduction capability of Fe complexes to their high-lying HOMO of these complexes (see Fig. 11.7), which make the electron donation to oxygen easier; as a result, more electron are transferred to O–O antibonding orbital, and O–O bond is readily broke.

Fig. 11.7

Energy levels of MPc and MP (M = Fe, Co, α = spin-down orbital, β = spin-up orbital) [188]

Shi and Zhang [189] carried out DFT calculations to study the dioxygen-binding abilities on the following iron and cobalt systems: CoPc (cobalt phthalocyanine), CoF₁₆Pc (cobalt hexadecafluorophthalocyanine), CoMeOPc (cobalt octamethoxyphthalocyanine), CoTSPc (cobalt tetrasulfophthalocyanine), CoTNPPc (cobalt tetraneopentoxyphthalocyanine), CoP (cobalt porphyrin), CoTPP (cobalt tetraphenylporphyrin), CoTPFPP (cobalt tetrakis(pentafluorophenyl)porphyrin), FePc, FeF₁₆Pc, FeMeOPc, FeTSPc, FeTPyPz (iron-2,3-pyridinoporphyrazine) FeP, FeTPP, and FeTPFPP. The ionization potential of the above metal macrocyclic complexes was also calculated.

Their calculations demonstrated that dioxygen-binding abilities of the transition metal macrocyclic complexes are determined by central metal, ligand, and substituents. For cobalt phthalocyanine systems, electron-donating substituents increase its dioxygen-binding ability, and electron-withdrawing substituents decrease the dioxygen-binding ability of the systems. In general, the transition metal macrocyclic systems’ ORR catalytic activity is related to its ionization potential and dioxygen-binding ability. Higher ionization potential and larger dioxygen-binding energy are associated with better catalytic activity. For porphyrin systems, cobalt derivatives have higher ionization potential and higher reactivity. For phthalocyanine systems, iron derivatives have good ionization potential and large oxygen-binding energy.

Similarly, Li et al. [190] investigated the interaction between dioxygen and various iron macrocyclic complexes (Fig. 11.8) by means of first-principle calculations. Their results indicated that the macrocyclic ligands without aromaticity were better electron-donor ligands, facilitating the backbonding from iron to O₂ and activated O₂ more significantly. Furthermore, the activation of O₂ was closely related to the energy of the HOMO of iron macrocyclic complexes. The smaller the energy gap between the HOMO of iron macrocyclic complexes and the LUMO of dioxygen, the more reactive the adsorbed O₂ became, with a longer O–O bond distance and a shorter Fe–O bond length. Based on their study, the authors proposed that FeTMTAA is a potential catalyst in oxidation reactions using O₂ as oxidant. FeAcacen, FeSalen1, and FeSalen2 are also promising catalysts for oxidation reactions. Furthermore, by modifying the structure of iron macrocyclic complexes through substituent groups and axial ligand, the ORR catalytic efficiency could be tuned.

Fig. 11.8

The structures for the deoxy-systems (FeX) under study. The maroon balls are iron atoms, the blue balls are nitrogen atoms, the red balls are oxygen atoms, the green-black balls are carbon atoms, and the turquoise balls are hydrogen atoms. R _Fe–N is the distance (Å) between the Fe and the coordinated nitrogen atom, and R _Fe–O is the distance between the Fe and the coordinated oxygen atom. Data out of the parentheses are for the triplets, in the parentheses for the quintets [190]

11.5.2 Conductive Polymer-Based Complex Catalysts

Conductive polymers such as polypyrrole (PPy), polyaniline (PAni), and polythiophen (PTh) have been the subject of much research owing to their wide applications in biosensors, electrochemistry, and electrocatalysis [191, 192]. Recently, conductive polymers have been also investigated as ORR electrocatalysis in three different ways: (1) utilizing conductive polymers as ORR electrocatalysts on their own, (2) incorporating non-precious metal complexes into the conductive polymer matrix, and (3) employing conductive polymers as a nitrogen/carbon precursor material for pyrolyzed M–N _x /C catalysts [105].

Shi et al. [193] employed DFT and experimental characterization to investigate active-site structures of unpyrolyzed cobalt–polypyrrole (Co–PPy) composites. They specifically examined four types of the active-site structures and studied the stability of each structure. They also examined oxygen adsorption on these structures with both end-on and side-on configurations and calculated the redox potential of each structure (Fig. 11.9). The theoretical simulations indicated that various active-site structures could contribute to the ORR activity of non-heat-treated Co–PPy catalysts and both Co(III) and Co(II) could play the role of an active center. The active site’s oxygen binding strength (end-on or side-on) dictates its dominant reaction pathway (2e⁻ or 4e⁻). With calculated redox potential of each active site and its oxygen binding strength, each active site’s contribution to 2e⁻ or 4e⁻ pathway at a defined potential can be deduced. Base on their results, the studied active sites contributed more to 2e⁻ pathway at low potential (0.3 V vs. RHE) than that at high potential (0.6 V vs. RHE), which is consistent with experimental observed change of hydrogen peroxide formation at different potentials. The calculations illustrated that more than one type of active sites contribute to the ORR activity observed with untreated cobalt–polypyrrole catalysts. Type A structures give similar potential dependence hydroperoxide formation trend as that observed from experiment. They postulated that for untreated Co–PPy composites, the presence of the type A structures is likely.

Fig. 11.9

Optimized structures of four proposed active sites [193]. Blue: Co, dark blue: N, gray: C, white: H, red: O

Another insight provided by their study is that the side-on oxygen adsorption requires more space than the end-on adsorption configuration. This may be one of the reasons why cobalt porphyrin and phthalocyanine systems cannot form stable side-on adducts which generally lead to 4e⁻ transfer products. The destruction of the ordered structure of such a macrocyclic catalyst during the heat treatment is likely to increase the number of sites that facilitate the side-on oxygen adsorption, and this leads to an increase in the number of electrons transferred in the ORR.

Similarly, the interaction between the oxygen molecule and Co–(n)pyrrole model clusters (n = 4, 6) was recently studied by Dipojono et al. [194, 195]. The stable adsorption site of the O₂ molecule on Co–(4)pyrrole is found to be in a side-on configuration, while for the case of Co–(6)pyrrole cluster, the O₂ molecule is slightly deviated from the side-on configuration. The O–O bonds of the O₂/Co–(4)pyrrole and the O2/Co–(6)pyrrole systems have elongated by 11.84 and 9.86 %, respectively. The elongation mechanism of O₂ on Co–(n)pyrrole is induced by the interaction between the cobalt d-orbitals and the O₂ antibonding π* orbital, which results in a charge transfer from the cobalt atom toward the O₂ molecule. In Co–(4)pyrrole, the elongation of the O₂ bond is larger than that of Co–(6)pyrrole since a complete side-on configuration has more symmetric overlapping between the cobalt d-orbitals and the O₂ antibonding orbital. Furthermore, the dissociation energy is affected by the amount of the charge transferred from Co–(n)pyrrole clusters to the O₂ molecule in the transition state.

11.5.3 Carbon-Based Materials

Carbon-based materials with large surface area such as carbon nanotubes, nanofibers, and graphene have recently received attention as potential metal-free catalyst materials for ORR [167–179]. In particular, nitrogen (N)-doped carbon materials have been shown to yield high ORR activity with four-electron transfer mechanism [167, 169]. Matter et al. [173] reported that the N-doped carbon nanofibers have high ORR catalytic activity and the N-doped nanotubes and graphene [170] were also recently reported to have high activity for ORR in the absence of transition metal atoms. The pyridinic N site (nitrogen atoms with lone pair electrons) has been generally considered as catalytic active site for ORR due to delocalizing of π electrons to pyridinic N, and existence of a large amount of pyridinic nitrogen is usually observed with highly active carbon materials [169, 170, 172]. On the contrary, a recent experiment suggested that the graphitic N (nitrogen bonded to three sp² carbon atoms) was the key for ORR activity rather than pyridinic N [196]. It is, in fact, difficult to distinguish the edge vs. pyridinic N experimentally [197], as graphitic N at edge can become a pyridinic N.

In order to elucidate the active sites for N-doped graphene materials, Jung et al. [198] performed the spin-polarized DFT calculations of graphene nanoribbon edges. They observed that the nitrogen doping increases the activity of graphene edges by enhancing the first electron transfer and giving dominate four-electron reduction products, which are the two most important aspects for improving ORR selectivity and performance. The outermost graphitic nitrogen site among others yields the lowest barrier for the rate-limiting first electron transfer as well as the highest selectivity toward the four-electron reduction pathway and hence is proposed to be the primary active site. Furthermore, the proposed catalytic cycle (Fig. 11.10) involves a ring opening of the cyclic C–N bond at the edge of graphene which converts the graphitic N to pyridinic N. This interconversion of graphitic to pyridinic N may reconcile the controversy regarding pyridinic or graphitic nitrogen as the ORR active sites for N-doped graphene materials.

Fig. 11.10

The proposed ORR catalytic cycle [198]

Functionalized graphitic materials with late transition metals were recently studied by Rossmeisl et al. [199]. They showed that graphitic materials with active sites composed of four nitrogen atoms and transition metal atoms belonging to groups 7–9 in the periodic table are active toward ORR. Spin analyses suggested that the oxidation state of those elements in the active sites is, in general, +2. Moreover, their results showed that transition metals do not have intrinsic catalytic activities, since their adsorption behavior can be severely altered by changes in the local geometry of the active site, the chemical nature of the nearest neighbors, and the oxidation states. Furthermore, the catalytic performance of ORR can be improved by stabilizing the adsorption of OOH with respect to that of OH, which can be achieved by carefully engineering the active sites.

Yu et al. [200] performed periodic DFT calculations to study the ORR catalyzed by N-graphene. The study took the solvent, surface adsorbates, and surface coverage into consideration. They observed that it is essential to include water effect in constructing a reliable free-energy reaction profile, especially for O₂ adsorption which is significantly enhanced by the polarization of O₂ in the presence of water due to hydrogen bond formation. Without water, O₂ cannot be adsorbed on the N-graphene surface. The formation of OOH⁻ is found to be energetically unfavored compared with the dissociation reaction OOH_(ads) → O_(ads) + OH⁻. The result suggests that O₂ is mainly reduced via a “4e⁻ reduction” pathway on N-doped graphene. The rate-determining step is the removal of O_(ads) from the N-graphene surface. Thus, they suggested that a catalyst with a high efficiency in O_(ads) removing should have a high activity.

11.6 Conclusions

Past few years or so have witnessed a big increase of theoretical modeling studies in electrocatalyst materials. These studies are enabled by better computational algorithms and fast computational facilities and fueled by the demand of environmental friendly and sustainable energy resources. Theoretical modeling complementing experimental study has shed light on detailed surface phenomena, formation of intermediates, and the activation energies related to elementary reaction steps and provided better understanding of ORR mechanisms. Methodologies that employ structure activity relationship such as adsorption energy, activation energies, and d-band centers have shown encourage results in identifying better electrocatalysts. They are invaluable tools to screen for better electrocatalysts. On the other hand, due to its limitation with simple model and small-sized systems, first-principle methods have difficulty in providing quantitative numbers for detailed reaction steps. The major issue with theoretical study of ORR in PEMFC is how to model the interface effectively. Better models and methodologies need to be developed to study electrochemical systems. More theoretical studies are needed to provide better understanding of ORR and to help design novel electrocatalysts.