4.1.1.1 Electrochemical Process and Catalytic Mechanism of the ORR

The electrochemical ORR is the reduction of oxygen molecules that are dissolved in aqueous/organic solutions or solids, from 0 oxidation state to −1 or −2 oxidation state through either two 2‐electron or one 4‐electron processes. Generally, the ORR involves the transfer of multiple electrons and has different initial and final potentials and products in different electrolytes (acidic or basic). As a result of all of these factors, the ORR is an extremely complex reaction.

If the ORR takes place in an aqueous acidic electrolyte and is catalyzed by Pt/C catalysts (the most common configuration for a low‐temperature proton‐exchange fuel cell), every oxygen molecule (O₂) accepts four electrons and combines with four protons (hydrogen cations, H⁺) to form water (two H₂O molecules). This process can be described as follows:

4.1

where E₀ represents the relative potential to the standard hydrogen electrode (SHE).

If a catalyst is absent or not active enough, the ORR can also go through two 2‐electron processes in acidic electrolytes. Here, hydrogen peroxide is the intermediate product of the first two‐electron process and water is the final product of the second two‐electron process. These two steps can be described as follows:

4.2

4.3

However, in basic electrolyte, similar four‐electron and two‐electron processes still occur, with OH⁻ being the final reduction product. The four‐electron ORR process in basic electrolytes can be described as follows:

4.4

For the two‐electron ORR processes in basic electrolyte, it is generally believed that an oxygen molecule first receives two electrons to produce HO₂²⁻ and then another two electrons to produce OH⁻, as shown in the following equations:

4.5

4.6

Apart from this, in some conditions, the ORR may also undergo a possible one‐electron process to produce free radicals [1]. However, this is not commonly observed in typical energy conversion and storage devices. Therefore, we will not discuss this aspect in detail in this chapter.

If the aqueous electrolyte is replaced by an organic electrolyte (e.g. in the cathode of a lithium–air battery), similar 4‐electron or two 2‐electron pathways still exist, but with lithium ions (Li⁺) incorporated:

4.7

4.8

And in certain aprotic electrolytes, oxygen molecules can also receive one electron to form superoxides:

4.9

For a fuel cell or lithium–air/oxygen battery, it is required that the electrode reactions proceed relatively easily. Consequently, a larger reaction current is desirable at a lower overpotential for these reactions, especially for the naturally sluggish ORR processes. The intrinsic catalytic activity of various metals (regardless of the effect of their microstructure, morphology, or mass diffusion effects) can be described and reflected by the adsorption energy of oxygen molecules on their surface. A volcano plot can be drawn versus their catalytic activity (Figure 4.1) [2].

As illustrated, Pt, Pd, and other noble metals exhibit the highest intrinsic catalytic activity toward the ORR. The transition metal species (Ni, Co, Fe, Co, etc.) tend to bind too strongly with oxygen and form oxides, whereas Au and Ag hardly adsorb oxygen. As a result, these metals exhibit inferior ORR catalytic activity compared with Pt and Pd. Activity improvements still exist for these noble metals, and this can be achieved by alloying Pt with metals that bind oxygen more strongly (e.g. Fe, Co, Ni, Cr, etc.) to form an optimized structure (e.g. a single‐atom layered alloy outside a Pt core). Theoretical calculations have shown that the adoption energy of these alloys toward oxygen can be slightly improved compared with Pt (i.e. from 1.57 eV for pure Pt to 1.89, 2.00, and 2.06 eV for Pt/Ni, Pt/Co, and Pt/Fe alloys, respectively). Due to these improvements, their ORR catalytic activity can be improved above that of pure Pt [2].

For the metal‐free catalysts, it is generally believed that their catalytic activity originates from the nonuniform distribution of charges on their surface (e.g. caused by heteroatom dopants in carbon frameworks), which can adsorb oxygen molecules by static electric forces. Similar to their metallic counterparts, the activity of these metal‐free catalysts can also be determined by their adsorption energies toward oxygen, as illustrated in Figure 4.2 [3].

According to the calculations, the adsorption energies of these metal‐free catalysts toward oxygen are far lower than those of metallic catalysts at present. However, similar to the metal‐based catalysts, these metal‐free catalysts can be improved by simultaneously doping with multiple elements. These metal‐free catalysts are also much lower in price, making them quite attractive for practical applications.

Compared with ORR behavior in ideal conditions mentioned above, the actual ORR process is largely related to the catalysts' surrounding environment. Taking the ORR process in aqueous electrolyte as an example, the overall reaction pathway is as follows: (i) dissolved oxygen molecules diffuse from the electrolyte to the catalyst surface, (ii) oxygen adsorbs on the catalysts surface, (iii) electron relocation from the electrode to the adsorbed oxygen molecules, (iv) desorption of the reduction products from the catalyst surface, and (v) diffusion of these products from the catalyst surface to the electrolyte (Figure 4.3). Consequently, the actual ORR process involves not only energy conversion but also mass transfer. The role of ORR catalysts is the promotion of one or several of these steps, in order to facilitate the kinetics of the overall reaction. Based on these considerations, an ideal ORR catalyst should possess the following characteristics: appropriate adsorption strength toward oxygen molecules, high electronic conductivity, suitable pore structures, and good stability in electrochemical environments. By far, the most commonly employed commercial catalysts are Pt, Pd, and other platinum group metals. These metals are often prepared as nanoparticles and loaded on highly conductive substrates (e.g. carbon black) in order to most effectively utilize their catalytic surface. For example, the catalytic activity of Pt‐based catalysts is largely related to their geometry and crystal facets (e.g. 110 facets are preferred in acidic electrolytes, but 100 and 111 facets are not) [4]. In addition, choosing suitable particle size is also important in exposing these facets to maximize the adsorption of oxygen molecules. This requires the particle size to be as small as possible, while maintaining good stability. For other metal or metal‐free catalysts, these principles are also applicable to their design and synthesis.

In order to lower the surface energy, fine metal particles tend to spontaneously aggregate into larger ones. Simultaneously, it is difficult to form sufficient conductive networks with these particles due to vast interphase resistance. Therefore, these metal particles (i.e. active sites) have to be loaded onto various conductive substrates to achieve high dispersion and conductivity for practical applications. As a result, the final microstructure and the overall performance of these composite catalysts are heavily determined by such conductive substrates. For metal‐free catalysts, the catalytic activity originates from their surface chemistry (i.e. active sites in the carbon frameworks). Hence, the chemical composition and microstructure of these catalysts also directly influence their ORR activity.

Based on this, carbon materials not only act as highly conductive substrates in ORR catalysts but also significantly affect (i) the effectiveness of active site exposure, (ii) the ease of reactant diffusion to active sites, and (iii) the electrochemical stability of the active sites. All these factors strongly contribute to the overall performance of these ORR catalysts.

4.1.1.2 Applications of ORR and ORR Catalysis

The ORR commonly exists in applications related to energy conversion and storage, such as fuel cells or metal–air batteries. In these devices, the ORR is the main cathodic reaction, and its efficiency largely determines their performance. In this regard, developing highly active ORR catalysts is critical for achieving high‐performance devices that are suitable for commercialization and practical applications.

A fuel cell is a high‐efficiency and low emission energy generation device that directly converts the chemical energy in fuels (hydrogen, methanol, etc.) into electricity [5, 6]. Taking a low‐temperature fuel cell, for example, it is composed of a cathode, anode, electrolyte (commonly aqueous acidic/basic solutions), proton/anion exchange membrane, and circuit/gas/heat‐controlling systems (Figure 4.4) [7]. In an acidic electrolyte, hydrogen is catalytically oxidized to produce hydrogen cations (i.e. protons, H⁺), which then travel through the proton exchange membrane to the cathode side of the cell. The ORR takes place at the cathode side of the cell, and the product combines the protons with oxygen to yield water. As only protons, rather than electrons, travel through the proton exchange membrane, the electrons have to pass through the outer circuit to form electrical current. If the electrolytes were basic, then the ORR product at the cathode is OH⁻ anions, which again pass through the anion exchange membrane to the anode to combine with hydrogen to yield water.

Despite the different respective onset potentials for the cathodic/anodic reactions in acidic/basic electrolytes, the actual output voltage of the whole cell, which is dependent on the potential difference between the anode and the cathode, is the same (1.23 V) regardless of electrolyte type. Here, the overall reaction can be described as the combination of hydrogen and oxygen to yield water as follows:

4.10

Depending on operating temperatures, fuel cells can be sorted into high‐temperature fuel cells (operated over 600 °C) and medium/low‐temperature fuel cells (operated below 200 °C) (Figure 4.5) [8]. High‐temperature fuel cells are primarily solid oxide fuel cells (SOFCs) and molten carbonate fuel cells (MCFCs), which generally use hydrogen, natural gas, hydrocarbons, or even coal as fuel. Due to their very high operating temperature (600–1000 °C), the electrode reactions can be carried out spontaneously and continuously even without the need of catalysts. However, they require large volumes and complex structures, which limit them to stationary applications, e.g. in a power grid.

Low‐temperature fuel cells often operate in the temperature range between room temperature to 200 °C. This includes polymeric electrolyte membrane/proton exchange membrane fuel cells (PEMFCs), direct methanol fuel cells (DMFCs), phosphoric acid fuel cells (PAFCs), and alkaline fuel cells (AFCs)/anion exchange membrane fuel cells (AEMFCs). Apart from their low operating temperatures, they have the advantages of smaller required volumes and higher energy density, which make them suitable for portable energy supply for applications like electric vehicles and mobile devices. However, in these types of devices, the ORR cannot occur spontaneously at a satisfactory rate for high power output, and as a result, catalysts are often inevitable in order to accelerate the electrode reactions.

Another important application of electrochemical ORR is metal–air batteries, using metals such as aluminum, magnesium, zinc, or lithium as the anode and air/oxygen as the cathode. The most widely studied example is the rechargeable lithium–air battery. Due to lithium's extremely high capacity, this type of battery could theoretically yield a very high energy density (up to 5200 Wh kg⁻¹), giving it a great potential for high density and portable energy storage applications. As a result, they have attracted huge attention recently.

In a lithium–air battery, lithium metal is directly used as the anode‐active material, whereas air or oxygen is used as the cathode‐active material. During discharge, lithium is oxidized and loses electrons to form lithium ions (Li⁺), which pass through the electrolyte to the cathode to combine with reduced oxygen to form various lithium oxides. The electrons then travel though the outer circuit and generate a current [9]. The discharge process can be generally described by the following equation:

4.11

Some researchers also believe that lithium oxide, rather than peroxide, may be partially produced:

4.12

When aprotic solvent electrolytes are used, a pathway to lithium superoxide is also possible, as shown in Equation 4.9.

Because lithium metal is directly used in lithium–air batteries, only nonaqueous electrolytes can be in direct contact with the lithium surface. However, there are more electrolyte choices available for the cathode side. According to these differences in electrolytes, lithium–air batteries can be divided into three main types [9]: (i) with lithium immersed in an electrolyte of dissolved lithium salts, in which condition, a solid electrolyte interphase is simultaneously or artificially formed on the surface of the lithium metal, preventing further corrosion of the lithium and providing a diffusion pathway for lithium ions (Figure 4.6a,b); (ii) with a solid‐phase electrolyte in between the anode and the cathode, which provides pathways for lithium ions and prevents internal short circuiting (Figure 4.6c); and (iii) with nonaqueous electrolytes in the anode and aqueous electrolyte in the cathode, employing a hydrophobic membrane between these two electrolytes for lithium transportation (Figure 4.6d). Despite the structural differences among these different lithium–air batteries, their cathodes are all highly air‐penetrable electrodes on which the ORR is carried out.

Currently, the most frequently used commercial ORR catalysts are Pt or Pt alloy (Pt/Pd, Pt/Ru, etc.) nanoparticles loaded on highly conductive carbon black. However, due to the high cost and poor recyclability of Pt‐based catalysts, it is difficult to produce a commercially viable device for widespread use on these catalysts. In a PEMFC, the ORR process is much more sluggish compared with the anode reactions. Consequently, catalyst loading at the cathode side of a PEMFC largely overweighs the loading on the anode side. However, in a metal–air battery, all of the catalyst is used to accelerate the cathodic ORR, whereas its anode does not require any catalysts. In this regard, if the specific activity of these catalysts could be improved or they are replaced altogether by cheaper catalysts, then fuel cells and metal–air batteries could be significantly more viable from the economic point of view. According to the above discussions, because the actual performance of a catalyst is largely dependent on the microstructure and chemistry of the carbon (substrates), the recently emerged novel carbon structures and materials thus provide a number of possibilities for the development of inexpensive catalysts with high activity.

4.1.2.1 Composition of Metal Nanoparticles and Carbons

To prepare Pt/C composite catalysts, two main routes have been developed, namely the impregnation method and colloidal method. In the impregnation method, carbon substrates are dispersed in the solution containing the metal precursors (e.g. metal salts). Thereafter, reducing agents (e.g. sodium borohydride or hydrazine) are added to reduce the metal ions into the metal. The metal forms nanoparticles that attach to the surface of the carbon substrates to produce the composite catalysts. The metal precursor can also be soaked in the pores of the carbons by evaporating the solvents in the suspension containing carbon and metal precursor before the reduction. Following this, the metal nanoparticles can be obtained inside the pores of the carbons by reducing the metal precursors, e.g. by heating the mixture in reductive atmospheres. This method is especially suitable when loading metal nanoparticles on highly porous carbon substrates, which can provide a large surface for depositing nanoparticles. Based on this impregnation method, a modified method, called incipient wetness impregnation, has been developed specifically to suit the synthesis of catalyst particles inside the pores of carbon support. This method uses the same amount of precursor solution as the carbon pore volume, thus avoiding possible catalyst deposition outside of the pores. In the colloidal method, metal nanoparticles are first prepared to form a colloidal suspension in a solvent and carbons are then mixed with this colloidal suspension to adsorb the metal particles on its surface. The colloidal method can provide catalyst particles with uniform and small size, especially when the particles are stabilized by proper surfactants in the solution.

Apart from these two methods, other techniques, like electrochemical deposition and the scattering method, have been studied for the preparation of Pt/C catalysts. In electrochemical deposition, a potential is applied to the carbon substrates in order to reduce the metal species and form metal nanoparticles on the carbon surface, whereas in the scattering method, Pt metal is used as a target material and carbon as a substrate in order to load fine Pt nanoparticles on the carbon surface.

For newly emerged ORR catalysts based on non‐noble metals and their compounds, they can be composited with carbon through a variety of methods. This includes the hydrothermal method, which converts metal precursors into metal (oxides) of different structures and directly loads them onto the carbon surface. Other techniques explored involve the direct introduction of metal precursors with carbon precursors, followed by simultaneous carbonization and metal (oxide) formation. These will be discussed in detail in the following sections.

4.1.2.2 Conventional Carbons: Carbon Black and Graphite

Carbon black and natural graphite are typical conventional carbon materials used as catalyst substrates. They possess advantages such as low cost and high electrical conductivity, making them the most frequently used support for ORR catalysts. Carbon black powder is usually obtained by pyrolyzing hydrocarbons (e.g. natural gas, gasoline, and coal) in an oxygen‐deficient environment [11]. Microscopically, it is an onion‐like particle composed of stacked graphite microlayers (Figure 4.7a). The particle size of carbon black is in the range of several to tens of nanometers. Due to the lack of long‐range order in the structure of carbon black, they are generally regarded as amorphous carbon.

Based on synthesis method, carbon black can be categorized into furnace black, channel black, and acetylene black. Carbon black prepared by different methods may have greatly different morphologies and properties, as summarized in Table 4.1.

Among these carbon black materials, Vulcan XC‐72 produced by Cabot is the most widely used catalyst support, commanding about 80% of the current market share [15]. Other carbon blacks, such as Black Pearl and Ketjen Black have also been widely used as catalysts supports. In order to maximize the utilization of expensive Pt, Pt is commonly synthesized into particles of 3–5 nm and then loaded onto the carbon black substrates (Figure 4.7b). However, these highly active metal particles can not only catalyze the reduction of oxygen but also accelerate the electrochemical oxidation of the amorphous carbon substrates, especially in corrosive electrolytes [21, 22].

Pristine carbon black can withstand air at over 400 °C without being notably oxidized. However, once it is loaded with Pt nanoparticles, oxidation can occur at temperatures as low as 125 °C. At higher temperatures, oxidation of carbon black becomes more rapid upon higher Pt loadings (Figure 4.8). Once oxidation takes place, the microstructure of the carbon black can be rapidly destroyed, which leads to severe detachment or aggregation of the catalyst nanoparticles and thus loss of catalytic activity. Compared with the oxidation in air, electrochemical corrosion of carbon takes place more easily in the corrosive electrolyte environments of fuel cells or metal–air batteries. This can be ascribed to two main factors: (i) the oxidation potential of carbon is as low as 0.2 V versus SHE, which means electrochemical oxidation of carbon may initiate at this low potential (being far lower than the ORR equilibrium potential of 1.23 V) [15, 24] and (ii) the water present in the electrolyte can also enhance the oxidation of carbon, combining with oxidized carbon to form either carbon monoxide or carbon dioxide in the medium‐ and low‐temperature fuel cells, respectively [25, 26].

In order to suppress the oxidation of the carbon substrates, graphitic carbon materials with higher crystallinity have been employed. However, graphite generally does not provide sufficient surface area to load high amount of Pt nanoparticles. To solve this problem, graphite has been finely ground to expose greater surface area. For example, after high‐energy ball milling for 50 h, the specific surface area of graphite increased significantly from 7 to 580 m² g⁻¹ [27]. However, this top–down treatment also simultaneously introduces a large number of defects into the graphite, which was evidenced by Raman spectroscopy. Figure 4.9 compares the Raman spectra of milled graphite for different mill times where an increase in the D peak indicates a higher degree of defects. This shows that simple ball milling severely compromised the desired high crystallinity of the graphite, and thus, it is not suitable for the synthesis of carbon substrates.

Another route to synthesizing highly crystalline carbon is to graphitize amorphous carbons (e.g. carbon black) at extremely high temperatures to cause the rearrangement of carbon atoms with higher order. For example, by heating XC‐72 carbon black in high‐vacuum or inert atmosphere to over 2000 °C, its crystallinity can be greatly enhanced. This is supported by an increased I_g:I_d ratio in the Raman spectra, which indicates the formation of long‐range graphitic structures. However, the rearrangement of carbon atoms also causes the disappearance of pores, especially micropores and mesopores, which are also a form of defect. Commonly, a loss of surface area up to 60% is expected to occur for carbons after such treatment [28, 29].

For these conventional carbon materials, it is extremely hard to simultaneously achieve both large surface area and high crystallinity. With the purpose of achieving higher nanoparticle loadings on carbons with higher graphitization for better overall catalytic performance, a number of novel carbon materials have been developed and employed for ORR catalysis applications.

4.1.2.3 Carbon Nanomaterials as Supports for ORR Catalysts

Carbon nanomaterials include carbons on which at least one dimension is in the nanoscale. For example, if all the three dimensions of a carbon material are in the nanoscale (e.g. carbon quantum dots or fullerenes), they can be regarded as 0‐dimensional nanocarbon. If one dimension of the materials is in the nanoscale (e.g. graphene), they can be called as two‐dimensional nanocarbons. If two dimensions of the materials are in nanoscale (e.g. carbon nanotubes (CNTs) or carbon nanofibers (CNFs)), then they belong to the group of one‐dimensional nanocarbons. Some research also categorizes carbons with nanopores into this type of nanocarbons; however, we will discuss them in a separate section.

Carbon nanomaterial formation can be regarded as the scissoring, bending, stacking, and/or reassembling of two‐dimensional graphene nanosheets, which is constructed of sp²‐hybridized carbon atoms (Figure 4.10) [30, 31]. For example, directly bending, rolling, and stitching of a ribbon‐like graphene sheet would result in a CNT, whereas stacking of single‐layered graphene would result in multilayered graphene or graphite, if thick enough.

The 0‐dimensional fullerenes are usually very small in size (<1 nm). Consequently, they cannot be used to support metal nanoparticles or effectively form a highly conductive network. As a result, the most commonly used carbon nanomaterials to support Pt nanoparticles are the one‐dimensional CNTs/CNFs or two‐dimensional graphene.

CNTs can be divided into single‐walled carbon nanotubes (SWCNTs) with a typical diameter less than 1 nm and multiwalled carbon nanotubes (MWCNTs) with a diameter up to tens of nanometers. The length of CNTs may vary from tens of nanometers to several millimeters [32]. If structural defects exist on the walls of the CNTs (e.g. pentagonal rings instead of hexagonal rings), distortion of the tube walls may occur. This may lead to the formation of a tube end closure (i.e. a carbon nanohorn) or twisting of tubes (i.e. a carbon nanocoil) [31, 33].

Due to their unique one‐dimensional and high‐crystalline characteristics, CNTs have very unique properties compared with conventional carbon materials, including high electrical conductivity (up to 10⁴ versus 10 S m⁻¹ for carbon black), very few surface defects to permit electrochemical corrosion, and good mechanical properties to maintain the structural stability during electrochemical processes. However, the very highly crystallized but passive surface of CNTs also has its disadvantages, namely limited deposition sites on which to anchor the metal nanoparticles or other active species for the ORR. To deal with this, artificial defects have to be intentionally introduced onto the surface of CNTs as anchoring sites for metal particle deposition and for better particle dispersion as well. A common method to achieve this is by surface oxidation of the CNTs in order to graft oxygen‐containing functional groups on their surface. For instance, soaking CNTs in nitric acid can graft carboxyl groups (COOH) on their surface [34], whereas harsher acidic treatments, such as refluxing CNTs in a mixture of sulfuric and nitric acids at 140 °C for several hours, will result in surface decoration with multiple functional groups (i.e. hydroxyl (OH), carboxyl (COOH), and carbonyl (CO) groups with an approximate molar ratio of c. 4 : 2 : 1) [35, 36]. Small particle size (2–5 nm) and very good particle dispersion have been achieved for Pt/Ru alloy nanoparticles when deposited on surface‐functionalized CNTs with a mass loading of 20 wt.% (Figure 4.11) [37].

Other than treating CNTs with strong and oxidative acids, there are also reports using alternative techniques to realize surface modification of CNTs. For example, after pre‐oxidation of CNTs using dilute nitride acid (5 M), further treatment with acetic acid to graft additional oxygen‐containing groups has been tested [38]. Compared with the treatments with harsh acids, this method tends to retain the smooth surface and structural integrity of the CNTs (Figure 4.12). Further, Pt nanoparticles with small size and high dispersion can still be deposited on CNTs treated in this way [38].

Such two‐step oxidation method not only well maintains the structure of CNTs but also helps obtain highly dispersed Pt particles on the CNT surface. At a very high Pt loading of 0.42 mg cm⁻², the Pt nanoparticle size can still be controlled to less than 5 nm with a uniform distribution on the CNT surface (Figure 4.12) [38]. Due to these advantages, this Pt/Ru/CNT composite catalyst achieves improved performance compared with conventional Pt/C.

Apart from the noble metals such as Pt, Ru, and Pd, other non‐noble metals or their compounds have also been recently used in order to develop more cost‐effective ORR catalysts. The ORR catalytic mechanism of these materials has also been studied in detail, and a number of new synthesis methods have also been developed.

For example, CNTs were slightly oxidized at low temperatures, and oxygen‐containing functional groups were formed on their surface without compromising their structures. Thereafter, the as‐treated CNTs were hydrothermally treated together with cobalt acetate and ammonia to finally form a CoO/CNT composite for ORR catalysis (Figure 4.13) [39]. Using this method, simultaneous N doping by ammonia and uniform distribution of CoO nanoparticles of very small size (∼5 nm) were achieved on the resultant CNT material. In terms of performance, this material showed comparative activity to commercial Pt/C catalysts in basic electrolyte. The origin of this excellent activity, as summarized by the author, can be attributed to the covalent bonding between the N‐doped CNTs and the CoO particles anchored on their surface (CoNC bonds). This bonding configuration significantly enhances the electron transition from the CNT surface to the CoO and thus facilitates the electrochemical ORR process.

Non‐noble metal nanoparticles can also be formed during the formation of carbon nanomaterials. For instance, cobalt ferrocyanide (Co₂Fe(CN)₆) was ball milled with graphene nanosheets and then annealed at 600 °C in an inert atmosphere (Figure 4.14a). During this process, the graphene sheets curved to form CNTs, which wrapped the Co/Fe nanoparticles inside to form a capsule‐like hybrid material (Figure 4.14b–g) [40]. Compared with pure graphene, a significant improvement in ORR catalytic activity was achieved on this Co/Fe/CNT/graphene hybrid material. By simply changing the mass ratio of the two precursors (ferrocyanide and graphene), the CNT wall number could also be tuned. This enabled the researchers to study the effect of wall number on ORR mechanism and performance. The results show that with increasing wall numbers, the overall ORR activity of the hybrid decreases and that the optimum wall number is 1–3. A Fe/CNT composite with a similar capsule‐like structure can also be fabricated by directly annealing suitable precursors in an inert gas. For example, by pyrolyzing sodium azide (NaN₃) and ferrocene (Fe(C₅H₅)₂), a bamboo‐like CNT encapsulating iron (oxide) nanoparticles can be obtained, also showing stable ORR activity [41]. In spite of enhanced ORR performance to comparative analogs, the catalytic activities of these materials are still inferior to commercial Pt/C. This may be attributed to the capsule structure, which prevents the active species from effectively contacting the reactant oxygen species in the electrolyte.

Some studies have shown that some organic polymers are also active for ORR catalysis and thus can be used as active materials and loaded on CNTs. One such polymer is poly(diallyldimethylammonium chloride) (PDDA), which can provide a positively charged surface when they are grafted onto CNTs and electrostatically adsorb the oxygen molecules in the electrolyte (Figure 4.15a) [42]. Despite the relatively poor ORR activity compared with that of conventional Pt/C, in terms of ORR onset potential and catalytic current, they possess superior ORR stability, far outperforming Pt/C, and show very good reaction selectivity.

Generally, CNTs have unique advantages of high electrical conductivity and structural stability. In order to best utilize these characteristics, future studies may be necessary to develop more facile and cost‐effective methods to achieve uniform nanoparticle loading without compromising their graphitic structures (i.e. maintaining its high conductivity and resistance against electrochemical corrosion). Attention should also be paid to the electron interface transition between the metal nanoparticles and the CNT substrates, which may contribute to high impedance if sufficient coupling is not achieved. This electron interface transition may be enhanced by creating strong interactions between these two species, like by the formation of covalent bonds between them.

The other typical and commonly applied one‐dimensional carbon nanomaterials are CNFs. Highly graphitized CNFs can be obtained by catalyst‐assisted (e.g. transition metals) chemical vapor deposition at 700–1200 K. The most frequently used precursors for CNFs are carbon‐containing gases, such as methane, ethane, and carbon monoxide [43, 44]. Another type of CNF is composed of amorphous carbon, which can be derived from the carbonization of precursor fibers (e.g. polyacrylonitrile (PAN) nanofibers) [45]. According to the stacking orientation of the graphite platelets in CNFs, they can be divided into three types: platelet, ribbon, and herringbone structure, as shown in Figure 4.16 [46]. The diameters of CNF can range from ten to several hundred nanometers based on the synthesis method.

Rather than loading the metal (compound) nanoparticles on the surface of the CNFs, as in the case of CNTs, these particles can also be homogeneously distributed throughout the CNF by directly incorporating the metal precursor to the CNF precursors (i.e. the polymer nanofibers) and then carbonizing the composite precursor fiber. This method has been used to obtain Fe‐containing CNFs. Specifically, polyvinylpyrrolidone (PVP) was chosen as the CNF precursor. It was then dissolved in ethanol and ferrous acetylacetonate (Fe(acac)₂) was added as the iron precursor. The composite precursor fibers were fabricated by electrospinning, which laid over each other to form an interconnected network (Figure 4.17a). This composite precursor (PVP/Fe(acac)₂) was then carbonized and converted into the Fe/CNFs composite to be used for ORR catalysis [47].

By tuning the synthesis parameters (i.e. time, humidity, and the aging period in air), the structures of the final products can also be changed. For example, if the precursor was aged in air for 24 h with a humidity of 70% and then carbonized, a mat‐like network with individual fibers overlapping each other can be obtained (FeN/C NMs, Figure 4.17b,c), whereas direct carbonization of the precursor fibers without aging would lead to the formation of an interconnected structure with the fibers fused together at the connecting points (FeN/C NNs) [47]. In this material, the iron species exist in the forms of both metal and metal‐oxide nanoparticles (i.e. Fe and Fe₃O₄), which evenly distribute on the surface and inside of the CNFs due to the homogeneous distribution of the iron precursors. Better ORR catalytic performance was observed for FeN/C NNs compared with FeN/C NMs, which was explained to be due interconnectedness of the fused points among the CNFs, permitting improved electron transport throughout the network.

A similar electrospinning concept can be transferred to the fabrication of Co/CNF composites, using PAN as the CNF precursor and cobalt acetate as the Co precursor. By carbonizing the composite fibers containing these two components, a composite CNF loading a mixture of Co and its oxides (CoO and Co₃O₄) can be obtained with most of the metal species located near or on the surface of the CNFs (Figure 4.18) [48]. Similar to the previous case, the Co content in this material can be tuned by simply changing the ratio of the two precursors, and by increasing the Co content, ORR activity also improves. Even though its activity is not as good as Pt/C, enhanced ORR stability is achieved, as is the case for many other non‐noble metal ORR catalysts.

Compared with CNTs, CNFs are more versatile from the aspects of their synthesis methods, which make it easier to load active metal species on them. However, these Pt‐free catalysts are still inferior to commercial Pt/C catalysts. The main reason for this is the intrinsically low ORR activity of these transition metal species. Additionally, these amorphous CNFs have poor conductivity, much lower than CNTs and carbon black. Therefore, finding a solution to improve the crystallinity of the CNFs, while maintaining their synthesis versatility, may be helpful to further improve their ORR catalytic performance. Apart from this, other research attention should be focused on decreasing the metal (oxides) particle size being loaded on the CNFs. Improving from the current tens of nanometers to a more desirable size, say 3–5 nm, could help to better utilize the metal species.

Graphene is a new member in the family of carbon materials, which has drawn tremendous attention since its discovery in 2004. To prepare graphene, a variety of technologies have been developed. It can be prepared by mechanical methods, like the mechanical exfoliation of natural graphite, or by electrochemical/chemical methods. Some of these include the electrochemical/sono‐assisted exfoliation of graphite in solvents [49–54], sono‐assisted exfoliation of chemically oxidized graphite (i.e. graphite oxide) to prepare graphene oxide (GO) followed by reduction [55], and the thermal expansion technique whereby rapid heating of oxidized graphite exfoliates the graphene layers [56]. High‐quality and highly crystalline graphene can also be prepared by chemical vapor deposition of carbon‐containing gases on transition metal substrates (e.g. Cu or Ni) [57, 58]. Among the graphene materials produced by these methods, the synthesis route via reduction of GO is especially suitable for catalyst construction because it has abundant surface functional groups that can act as anchoring sites for active species. It is also a relatively cost‐effective method that is scalable for mass production.

Chloroplatinic acid (H₂PtCl₆) is the typical precursor for the preparation of Pt nanoparticles. To obtain a H₂PtCl₆/graphene composite precursor, H₂PtCl₆ was dissolved in acetone and mixed with graphene powders produced by the thermal expansion method. After heating this composite in hydrogen at 300 °C, the H₂PtCl₆ decomposed into Pt nanoparticles, which homogeneously distributed on the graphene surface. The Pt formed nanoparticles in the size range of 1.5–3.5 nm at a loading of 20 wt.% (Figure 4.19) [59]. Compared with commercial Pt/C catalysts, the Pt particles on the Pt/graphene composite are much smaller. As a result, this material has a larger electrochemical active surface available to adsorb oxygen (108 versus 72 m² g⁻¹), affording better ORR performance. Further, the rich surface functional groups are able to strongly adsorb the Pt nanoparticles on the graphene surface, which help improve the stability of the catalyst. After long‐term testing (5000 cycles), the particle size of Pt on the Pt/graphene composite is still smaller than that of the Pt/C catalyst (5.5 versus 6.9 nm). At a significantly higher Pt loading of 60 wt.%, monodispersed Pt nanoparticles can still be obtained on the graphene surface, with only a slightly larger particle size of c. 1.87 nm [60]. Besides Pt, this method is also applicable for loading other noble metal nanoparticles with very small sizes, such as Pd or Pt/Ru alloys [60].

The colloid method can also be used to load metal nanoparticles on graphene. For example, Pt/Fe alloy nanoparticles (diameter of c. 7 nm) were first synthesized by the solvothermal treatment of platinum(II)acetylacetonate (Pt(acac)₂) and iron carbonyl (Fe(CO)₅) at 220 °C. A homogeneous suspension of graphene in tetrahydrofuran was then mixed with a suspension of the Pt/Fe alloy and dissolved in a mixture of tetrahydrofuran and ethanol. The resultant mixture was ultrasonicated and centrifuged in order to obtain the Pt/Fe/graphene composite ORR catalyst [61]. The same process was used to prepare Pt/Fe nanoparticles loaded on carbon black as a comparative analog material. In acidic electrolyte, both catalysts exhibit superior activity to commercial Pt/C for ORR. However, the graphene composite had improved catalytic activity over the carbon black composite. Additionally, this Pt/Fe/graphene composite catalyst also possesses extremely good electrochemical stability in acidic electrolyte and shows negligible performance attenuation after 10 000 cycles. Using a similar strategy, Co metal nanoparticles (diameter of c. 10 nm) were synthesized and their surface was spontaneously oxidized by air to form a Co/CoO core shell structure. A suspension of these particles was ultrasonically blended with graphene in order to obtain a Co/CoO/graphene composite catalyst for the ORR [62]. However, this catalyst does not have comparable ORR activity to commercial Pt/C, despite its good stability.

More methods have been developed to load non‐noble metals and their compounds onto the surface of graphene, and the most common one is the “one‐pot” hydrothermal treatment of GO and metal salts to convert them into graphene and metal oxides/hydroxides. For instance, by hydrothermally treating a suspension containing GO, cobalt acetate, and ammonia, a composite catalyst composed of reduced GO and cobalt oxide (Co₃O₄/graphene) can be obtained with a Co₃O₄ nanoparticle size of 4–8 nm (Figure 4.20a,b) [63]. In this process, the ammonia also dopes the graphene with nitrogen (N content of c. 4%), and the N dopants within the graphene network form covalent bonds with the Co₃O₄ particles, effectively anchoring them on graphene surface. This strong covalent bonding is able to facilitate electron transfer and thus significantly improves the material's ORR performance. Moreover, the ORR performance of this material further improves with higher concentrations of basic electrolyte.

Similarly, dual‐metal oxides (e.g. MnCo₂O₄) can also be synthesized and loaded on graphene by the hydrothermal method (Figure 4.20c,d) [64]. The obtained oxide particles (c. 5 nm) are homogeneously dispersed on the surface of graphene. In MnCo₂O₄, Co³⁺ ions in the Co₃O₄ crystal are replaced by Mn³⁺ ions, and as a result, the ORR activity of these oxide nanoparticles is improved. Interestingly, this dual‐metal oxide can directly reduce oxygen into OH⁻ in basic electrolyte through a four‐electron pathway that is similar to that of Pt.

During the hydrothermal treatment, the functional groups on the GO surface are partially removed, not only resulting in the reduction of GO but also leading to the reassembly of the graphene sheets due to the π–π interactions between them. In the case that the initial concentration of GO is high enough, they may go through a self‐assembly process to form bulky units (i.e. graphene hydrogel) after being reduced. During this process, the hetero particles in the suspension may be trapped inside the three‐dimensional network of the hydrogel to form a composite with the graphene. The solvents inside the hydrogel can be removed afterward by other techniques (e.g. by cyro/freeze‐drying) without destroying its three‐dimensional structure and finally yield a highly porous and lightweight graphene‐based composite aerogel.

This unique property of graphene has been utilized to prepare a Fe₃O₄/graphene composite aerogel as a free‐standing, highly porous, and bulky ORR catalyst [65]. A hydrogel containing Fe₃O₄/graphene was prepared by the hydrothermal treatment of an aqueous solution containing iron(II)acetate, GO, and pyrrole (Figure 4.21a). Water in the gel was then removed by freeze‐drying of the hydrogel to yield the Fe₃O₄/graphene aerogel. Interconnected macropores of several microns can be observed on the aerogel, whose walls are composed of graphene (Figure 4.21b,c). On the graphene walls, Fe₃O₄ particles of about 100 nm can be found to be evenly dispersed and act as the ORR‐active sites. Compared with Pt/C catalysts, this Fe₃O₄/graphene composite is still inferior, especially with respect to ORR onset potential and conversion efficiency (i.e. relatively low electron transfer numbers). Similar concepts for fabricating bulky graphene assemblies may also be transferred to the synthesis of materials with other metal species and may also find roles for other electrochemical/structural applications.

4.1.2.4 Porous Carbons as Catalyst Supports for ORR

Apart from conventional carbon and carbon nanomaterials, another very important group of carbon materials is porous carbons. According to the difference in pore size, porous carbons can be divided into three categories: microporous carbon (pore size < 2 nm), mesoporous carbon (2 nm < pore size < 50 nm), and macroporous carbon (pore size > 50 nm). The synthesis of porous carbon is also very versatile, and the most frequently adopted method for controllable synthesis is the templating method. This method utilizes “templates” with different structures to tailor the pore structures of the obtained carbons and the templates are then removed to expose the carbon pores.

Template Synthesis of Porous Carbons

Porous carbons can be synthesized using two types of templates: a soft template and a hard template. Hard templates are materials with certain “rigidity,” which does not significantly change its structure or morphology during the synthesis of the porous carbon. Hard templates that are usually employed include porous silica, porous alumina, silica microspheres, polystyrene (PS)/poly(methyl methacrylate) (PMMA) spheres, and many bio‐derived templates. Usually, hard templates are stable during the synthesis of carbons (especially the various oxide‐based templates), which means they may need to be removed by subsequent treatments in order to expose the pores on the carbons.

By choosing hard templates with different pore structures, a porous carbon with a reversed structure can be obtained. For example, zeolite templates will result in carbons with micropores, whereas the mesoporous template SBA‐15 with tubular pores (p6mm structure) is frequently used to synthesize mesoporous carbon composed of parallel carbon nanorods (CMK‐3). If silica or other polymer microspheres are chosen as the template, the obtained carbon would then show sphere‐like pores with interconnected necks. When these spheres are orderly assembled (i.e. by sedimentation in solvents), then the obtained carbon would show a reversed opal structure with honeycomb‐like ordered pores. By simply changing the sphere diameter, the pore size of the carbons can be tuned (Figure 4.22a) [66]. Silica templates can also be used to introduce pores into graphene. For example, a silica colloidal suspension (12 nm) was blended with GO solution and then dried to form a silica/GO composite with silica nanospheres among the GO sheets. Thereafter, the silica was removed by etching with hydrofluoric acid (HF), forming a mesoporous graphene with the same pore size as the silica particle diameter (Figure 4.22b) [67].

However, soft templates are usually molecules with hydrophilic/hydrophobic terminals (i.e. surfactants; e.g. the tri‐block copolymer P123/F127). These molecules form micelles in the solvent through self‐assembly and bind with the carbon precursor molecules (e.g. low‐molecular‐weight phenolic resin) through hydrogen bonding or electrostatic interactions to form a micelle/resin precursor in the solvent (Figure 4.23). This composite precursor then goes through an evaporation‐induced self‐assembling (EISA) process in ambient atmosphere followed by heating–solidification in air or an assembly–solidification process during the hydrothermal treatment [68–71].

In some cases, hard templates and soft templates can be used together, in order to prepare carbon materials with special pore structures. For example, PMMA microspheres were assembled into a hexagonally ordered structure (Figure 4.24a), and silica was then cast into the voids among the spheres using tetraethyl orthosilicate (TEOS) as the precursor. The PMMA template was then removed during the calcination of the silica, which leads to the formation of porous silica with honeycomb‐like pores (Figure 4.24b). Subsequently, a solution containing a soft template (P123) and carbon precursor (phenolic resin) was impregnated into the spherical pores of the porous silica and then converted to carbon after solidification and carbonization at high temperatures. After removal of the silica template, carbon nanospheres with ordered mesopores were obtained (Figure 4.24c–e) [72].

Another example is the simultaneous use of PS nanospheres and F127 as hard and soft templates, respectively, to prepare carbon materials with both macropores and mesopores [73]. In this method, PS spheres were orderly packed by sedimentation in an aqueous suspension to form an opal structure as the hard template. An ethanol solution containing phenolic resin (i.e. the carbon precursor) and F127 (i.e. the soft template) was then impregnated into the voids among the PS spheres under vacuum. Following, a resin/F127 composite with mesostructure, which existed in the void of the PS spheres, was formed by the room temperature EISA process during the removal of ethanol. The templates were then thermally decomposed during the carbonization of the resin, resulting in a hierarchical porous carbon with ordered tubular mesopores on the walls of the interconnected spherical macropores (Figure 4.24f).

Apart from these commonly used template materials, people have also used other materials, such as bio‐derived or other natural material with fine structures as the template to synthesize mesoporous carbons [74, 75]. These newly emerged template materials not only result in a variety of new carbon materials with various pore structures but also help in the development of more environment‐friendly technologies to synthesize cost‐effective porous carbons.

Porous Carbons as Catalyst Supports for ORR

Mesoporous carbon with dual pore sizes (CMK‐5) was first utilized as a supporting material for Pt nanoparticles for ORR [76]. Similar to CMK‐3, CMK‐5 is also composed of hexagonally ordered carbon nanorods. However, these are hollow‐structured like a tube. As a result, this type of carbon usually gives a much larger specific surface area than other mesoporous carbons, making them very suitable for catalyst support applications. To utilize this unique property, the impregnation–reduction method was applied to prepare the Pt/CMK‐5 composite catalyst for the ORR. Specifically, chloroplatinic acid precursor was dissolved in solvent and impregnated into the pores of CMK‐5, followed by the removal of solvents and reduction of precursor into Pt nanoparticles. The resultant material comprises smaller Pt particle size (<5 nm) than commercial Pt/C catalysts. As a result, a higher specific activity toward ORR catalysis was achieved on this Pt/CMK‐5 catalyst. Another typical mesoporous carbon (CMK‐3) has also been used for this purpose, which exhibits superior ORR activity to Pt/C as well. This indicates the huge potential of mesoporous carbons for ORR applications due to their unique pore structures [77].

For mesoporous carbon prepared by hard templates, metal (compound) nanoparticles are often loaded inside their pores by posttreatment methods after the carbons are synthesized. In contrast, it is more versatile and convenient to prepare the mesoporous carbon/metal composite through a soft template method. For example, during the soft template synthesis of mesoporous carbons, as illustrated in Figure 4.23 [68], metal salt can be used as the metal precursor and added into the system to involve in the self‐assembling process and to form a metal‐containing resin with ordered meso‐structures. After the subsequent carbonization, ordered mesoporous carbon with metal (compound) nanoparticles embedded in their framework can be obtained.

For example, in the dual‐template synthesis of a hierarchical porous carbon, iron chlorides (metal precursor) and acetylacetone (acac, used as a chelating agent for better dispersion of the metal ions) were dissolved in ethanol and then added to the precursor solutions (i.e. phenolic resin as the carbon precursor and F127 as the soft template). This mixture precursor solution was then impregnated in the void of an orderly packed PS hard template, which then went through a room temperature EISA process, solidification at 100 °C, and carbonization at a high temperature. This process yielded a hierarchical porous carbon with both macropores and mesopores (Figure 4.25) [78].

Interestingly, in this material, the iron species not only act as the ORR‐active species but also perform as the catalyst for the growth of CNTs during the carbonization of the resin precursor (Figure 4.26). These CNTs thus enhance electron transport during the ORR process and greatly improve the ORR activity of the material. The ORR performance of this material is closely comparable with that of commercial Pt/C with respect to ORR onset potential, catalytic current, and reaction efficiency [78].

Two major advantages exist for mesoporous carbons employed as ORR catalysts, especially for carbons with ordered mesopores. These are as follows: (i) their large specific surface area provides sufficient sites that contribute to the homogeneous anchoring of the metal (compound) nanoparticles and (ii) the uniform and ordered mesopores can provide continuous and effective mass diffusion for the reactants to travel from the electrolyte to the surface active sites. In order to improve the intrinsically low conductivity of mesoporous carbons, due to their relatively low graphitization degree, it would be necessary to incorporate various carbon nanomaterials with the porous carbons to form interconnected networks for facile electron transportation, which is essential for electrochemical processes [73, 78].