9
Unraveling the Active Site on Metal‐Free, Carbon‐Based Catalysts for Multifunctional Applications

Hsin‐Yi Wang1, Hongbin Yang1, Bin Liu1, and Liming Dai2

1Nanyang Technological University, School of Chemical and Biomedical Engineering, 62 Nanyang Drive, Singapore, 637459, Singapore

2Case Western Reserve University, School of Engineering, Department of Macromolecular Science and Engineering, 2100 Adelbert Road, Cleveland, OH, 44106, USA

9.1 Introduction

Currently, whole world is facing several daunting challenges on energy production and storage to sustainably support the growth and development of human society. It is thus urgent and necessary for us to fundamentally change the way we produce and consume energy to avoid catastrophic consequences, such as serious climate change caused by greenhouse gas emission from the usage of fossil fuels. Developing renewable electrochemical energy conversion and storage technologies offers promising solutions to address the global energy scarcity and the associated environmental issues [1]. Regarding the aforementioned techniques, in particular, oxygen reduction reaction (ORR) and oxygen evolution reaction (OER) are the two critical electrochemical processes participating in a wide dimension of energy storage devices and renewable energy transformations, including rechargeable metal–air batteries [2], regenerative fuel cells [3–5], water splitting cells [6], and water electrolyzers [7]. Unfortunately, the overall efficiency and performance of these green energy devices have been severely limited by the sluggish kinetics of the electrocatalytic reduction and evolution processes of molecular oxygen, which involve a relatively unfavorable four‐electron stepwise charge transfer reaction [5, 8]. Apart from ORR and OER, the electrocatalytic conversion of emitted carbon dioxide (CO2) to value‐added chemicals and fuels offers another promising alternative approach to mitigate the most urgent challenge regarding the renewable energy production and environmental remediation [9]. Key technologies based on efficient CO2 reduction reaction (CO2RR) will allow us to move away from the reliance on fossil fuels and close the carbon cycle. In addition, hydrogen evolution reaction (HER) is also critical for the development of clean and renewable energy technologies as hydrogen can be directly used as the fuel with water as the only product from combustion of H2 [10]. Based on the above information, it can be realized that nearly all of the useful electrochemical processes involved in energy conversion are based on catalytic chemical transformations, which take place at interfaces between solids and liquids or gasses. Therefore, facilitating the electrochemical reaction at the interface is essential. Several noble metals (e.g. Pt, Ir, and Ru) and transition metal (e.g. Ni, Co, Mn, and Fe) oxides/hydroxides have been widely applied as the electrocatalysts for boosting the electrocatalytic processes of ORR, OER, and HER, and some metal (e.g. Ag, Au, Cu, and Pt) complexes have also been studied as the heterogeneous catalysts for electrochemical CO2 reduction [11, 12]. However, the high cost, scarcity, detrimental environmental effects, and inferior durability of these metal‐based catalysts have hampered their widespread and large‐scale applications in renewable energy technologies.

Among the extensive research and exploration on non‐metal‐based catalysts for electrocatalysis, carbon‐based nanomaterials, such as heteroatom‐doped carbon nanotubes (CNTs) and graphene, have been proposed as a new generation of metal‐free electrocatalysts for energy conversion. Impressive experimental and theoretical results have been achieved through the molecular and/or nanoarchitecture engineering of carbon nanomaterials using various innovative strategies, such as surface functionalization, geometric structuring, and heteroatom doping [12]. Among them, being low‐cost, metal‐free catalysts for HER, OER, ORR, and CO2RR, nitrogen‐doped, carbon‐based nanomaterials have gained significant attention [10, 13–15]. For instance, very recently, metal‐free, N‐mono‐doped, carbon‐based nanomaterials were studied as efficient bifunctional electrocatalysts for both ORR and OER [16]. N‐modified carbon nanomaterials with unique architecture were demonstrated as superior HER [17] and CO2RR [15] electrocatalysts. Although metal‐free, carbon‐based catalysts with deliberate heteroatom‐doping modification into the graphitic network have been intensively studied, the detailed reaction mechanism in electrocatalysis is still equivocal. In addition, in the concrete application of carbon nanomaterials, regardless of acting as supports for other metal‐based catalysts or as metal‐free catalysts themselves, the coexistence of oxygen‐containing functional groups on the surface of carbon materials seems inevitable [13]. This is due to the fact that oxidation process is usually required to introduce oxygen‐containing groups and/or defect sites for the subsequent functionalization with other heteroatoms. Thus, the individual role of oxygen‐containing functional groups (e.g. ketonic CO group) and doped heteroatoms (e.g. nitrogen) toward the catalytic activity remains unclear. If we can identify the active site and clarify the functional group effect on electrocatalytic reaction, the obtained information and understanding should bring out various designs of multifunctional catalysts based on carbon‐based nanomaterials. Unfortunately, the identification of the active site on carbon‐based catalysts is still very challenging.

Ideally, it is desirable to monitor the electronic and chemical variations of catalysts under the operando condition to determine the critical active sites responsible to the electrocatalytic reaction [18]. However, such advanced in situ experiments are still immature and difficult to be realized in the current stage owning to the following factors: First, the chemical or electronic determination of light elements, such as carbon, oxygen, and nitrogen, usually requires the use of specific surface‐sensitive spectroscopic techniques under ultrahigh‐vacuum (UHV) environments. Under this precondition, the presence of electrolyte will greatly increase the difficulty for the design of the in‐situ‐examining platform to probe the possible active sites on the solid/electrolyte interface. Second, for a typical chemical reaction, there exist several elementary and dynamic steps, and the catalytic transformation from reactants to the final products is usually a stochastic event (the typical turnover frequency for a good catalyst is 1 per active site per second). If the reaction is monitored under operando and steady‐state conditions, the most stable intermediate with the highest activation barrier on the catalyst surface will be the only species that can be probed. The other sub‐intermediates with relatively smaller activation barriers can only show up transiently owing to their instability of the transition state, which makes it hard to depict the complete scenario of reaction process and therefore the determination of the true active sites on the catalyst surface [19].

Due to lack of valuable information from the real‐time observation via in situ experiments, most understandings of the elementary reactions, the geometry of short‐lived intermediates, and the thermodynamic energy difference between those elementary steps still rely on quantum chemical calculation, e.g. density functional theory (DFT) simulation [20]. The ex situ spectroscopic comparison, commonly employing soft X‐ray absorption spectroscopy (XAS) and X‐ray photoelectron spectroscopy (XPS) as the surface‐sensitive examining tools [19], between the pre‐ and post‐chemical states of the catalyst in a specific electrochemical reaction is the most common artifice to deduce the possible active sites. However, the simultaneous presence of contamination and/or undesired surface functional groups on the surface of catalyst induced by electrochemical activation or exposure to air during chamber‐to‐chamber transportation can disturb the interpretation on the spectroscopic fingerprints of the potential active sites. In addition, the ex situ spectroscopic comparison cannot always guarantee a meaningful trend or relevance between the chemical variation of the catalyst and the electrochemical activity. Nevertheless, such ex situ approaches are still useful in shedding light on the most possible underlying mechanisms.

In this chapter, we will give an overview of the determination of active sites on metal‐free, carbon‐based catalysts in multifunctional applications. Here, we particularly focus our discussion on HER, OER, ORR, and CO2RR using nitrogen‐doped or oxygen‐containing carbon‐based nanomaterials as the catalysts. There exist several configurations of oxygen‐ and nitrogen‐containing functional groups in a nitrogen/oxygen‐doped conjugated graphite plane [21, 22], as illustrated in Figure 9.1.

Image described caption and surrounding text.

Figure 9.1 Different forms of doped nitrogen and oxygen in nitrogen/oxygen‐functionalized carbon materials. Four featured nitrogen‐containing groups (pyrrolic‐N, pyridinic‐N, graphitic‐N, and pyridinic oxide), which are generally discussed as the active sites on nitrogen‐functionalized carbon materials, are particularly marked in the carbon matrix.

The nitrogen‐ and/or oxygen‐doping‐induced charge redistributions are usually described as the main reason for enhanced catalytic performance for carbon‐based catalysts. The configuration of nitrogen dopants influences the chemical environment of the carbon matrix and therefore will affect the electronic structure of their neighboring carbon atoms, leading to various catalytic properties. The ex situ soft XAS and/or XPS spectroscopic comparison between the pre‐ and post‐chemical states of nitrogen‐containing functional groups on the carbon‐based catalysts in a particular electrocatalytic reaction is the easiest approach to infer the possible active sites. XPS, also known as electron spectroscopy for chemical analysis (ESCA), is a highly surface‐sensitive and quantitative spectroscopic technique, which is able to provide detailed chemical information, including the elemental composition, empirical formula, valance state, and electronic state of the elements present within the shallow surface layer (top 0–10 nm usually) of the target material. XPS spectra are obtained by irradiating a material with a beam of X‐rays while simultaneously measuring the kinetic energy and number of electrons that escape from the top 0–10 nm of the analyzed material. XPS requires high‐vacuum (P ∼ 10−8 mbar) or even UHV (P < 10−9 mbar) condition, while a current area of development is the ambient pressure XPS, in which samples can be analyzed at pressures of a few tens of millibar [23]. On the other hand, XAS is usually performed by synchrotron radiation, which can provide intense and tunable X‐ray beams in the range of surface‐sensitive intensity, namely the soft X‐ray [24]. The combination of highly surface‐sensitive spectroscopic techniques and first‐principles calculations should be the efficient way to reveal insights into the reaction intermediates as well as the reaction pathways. We will discuss on how the electronic variation induced by doping nitrogen/oxygen in carbon‐based catalysts affects the performance of HER, OER, ORR, and CO2RR and illustrate the logic, which we use to deduce the final conclusion based on the spectroscopic identification. At the end, we will present another useful method to probe the active sites on carbon‐based catalysts via poisoning them with phosphate anion.

9.2 Electrochemical Reduction Reaction: Oxygen Reduction Reaction (ORR) and Hydrogen Evolution Reaction (HER)

Electrochemical reduction reaction, including ORR and HER, is an important half reaction in many fields of renewable energy conversion, production, and storage [15, 25, 26]. For instance, cathodic oxygen reduction acts as an essential half reaction in producing electricity and is the key kinetic‐limiting factor on the performance of fuel cells. Meanwhile, water reduction to produce hydrogen as the renewable fuel is the central idea to the area of electrocatalytic water splitting. For both ORR and HER, noble metal (e.g. platinum) has been regarded as the most active electrocatalyst; yet, it still suffers from many drawbacks, including the high cost, scarcity, and susceptibility to poisoning [5]. Therefore, the large‐scale practical use of noble metal catalysts in fuel cells, water electrolyzers, and metal–air batteries will not be realized in future and the expensive noble metal‐based catalysts for ORR and HER have to be replaced by other new class of catalysts with low cost and good performance and durability.

It is well known that to boost the slow kinetics of ORR and facilitate the widespread application of fuel cells require the development of abundant and reliable non‐noble‐metal‐based cathode (NNMC) catalysts. Over the past decades, scientists have devoted to the study of two different classes of NNMC catalysts: MeNC (Me represents a transition metal, e.g. Fe) [27] and CNx (carbon‐based nanostructures) [28]. It has been well established that the ORR active site on MeNCs should be ascribed to the metal center [3, 29, 30]. For example, there is growing consensus that planar FeN4 with Fe2+ in the low spin state of d‐orbital electronic configuration and coordinated to four pyrrolic nitrogen on the same plane in a carbon matrix is the active site for the FeNC catalyst [31]. For CNx group catalysts, they can be prepared through chemical vapor deposition (CVD) via a C and N source (e.g. CH3CN) over a metal‐oxide support (such as MgO, SiO2, or Al2O3) followed by an acid‐washing treatment to remove the exposed support. The metal center merely as the initiator for the growth of CNx shall remain encapsulated within the carbon matrix and is inaccessible to the ORR [32, 33]. Thus, the ORR activity of CNx should be attributed to a distinct factor different from that of MeNC.

Although metal‐free, carbon‐based electrocatalysts have been demonstrated as the potential alternative to replace platinum‐based catalysts for ORR and HER [18, 25, 34, 35], in which heteroatom doping is a common strategy to enhance the catalytic activity by altering the electronic structure of carbon, the detailed reaction mechanism, the role of the doped heteroatom, and the active site still remain unclear. Here, we emphasize our discussion on the recent mechanistic understanding of the nitrogen‐containing, carbon‐based electrocatalysts for the reduction reaction.

9.2.1 Oxygen Reduction Reaction (ORR)

Understanding the underlying correlation between nitrogen‐dopant configuration and electrocatalytic activity is important for the design of efficient next‐generation catalysts. However, there is still a debate on the catalytic roles of various nitrogen‐containing species, and which one should take the responsibility as the active site for ORR. First of all, the improved ORR performance of nitrogen‐modified, carbon‐based catalysts should be ascribed to the increase in electron density/electron‐donating property on the nitrogen‐doped carbon nanostructures where the extra valence electrons are donated by nitrogen. However, the total nitrogen content in the carbon‐based catalyst does not play the critical role in the ORR process; yet, the ratio between graphitic‐N and pyridine‐N does [36]. By delicately controlling the ratio of graphitic‐N to pyridine‐N in the carbon nanostructures, it is found that graphitic‐N determines the limiting current density, whereas pyridinic‐N improves the onset potential of ORR by turning the kinetics. The pyridinic‐N can introduce a lone electron pair to the carbon matrix to enhance the electron‐donating capability, which is able to weaken the OO bond of O2 molecule via the bonding of O with N and/or its adjacent C atom on the catalyst to facilitate the reduction of O2. Meanwhile, abundance of pyridinic‐N species in the catalyst tends to drive ORR through a four‐electron (4e−) reduction mechanism.

The above mechanistic understanding obtained from experiments is supported by theoretical simulation, which reveals that the active center in metal‐free catalysts can directly reduce molecular oxygen into water through a four‐electron process or a less‐effective, two‐electron pathway. The desorption of OOH(ads) to form OOH− is found to be energetically unfavorable as compared with the dissociation of OOH(ads) to form O(ads) and OH−, suggesting that O2 is mainly reduced via a 4e− reduction pathway on N‐doped graphene. Instead, the rate‐determining step (RDS) should be the removal of O(ads) from the N‐doped graphene surface. Thus, a more active catalyst should possess a structure that can facilitate the desorption of O(ads) species [37]. The DFT calculation also indicates that the doping‐induced spin redistribution is the driving force for catalytic activities of metal‐free, carbon‐based catalysts [38]. The energy calculation for each ORR step demonstrates that spontaneous oxygen reduction on nitrogen‐doped graphene through a direct four‐electron pathway is possible, which is consistent with the experimental observations. The active site for single nitrogen‐doped graphene could be the spot that possesses either high positive spin density or high positive atomic charge density. The nitrogen doping introduces asymmetric spin density and atomic charge density, making N‐doped carbon catalysts possible to exhibit high electrocatalytic activity for ORR.

The possibility for pyridinic‐N contributing to the overall catalytic activity was further confirmed by synchrotron‐based XPS analyses on NH3·H2O‐treated graphene (denoted as G‐NH3·H2O) before and after ORR [39]. XPS spectra (Figure 9.2a–d) reveal that the intensity of OH, one of the surface intermediates for ORR, increases after electrochemical reaction. The C 1s XPS spectra also show that the COH peak dramatically strengthens after ORR, suggesting that OH tends to bind on aromatic C, based on which the changes of the N 1s XPS profiles before and after ORR could be realized. In terms of the attachment of OH to the C atoms near to nitrogen, the nitrogen‐binding energy would shift upward (Figure 9.2e). Figure 9.2f illustrates the binding of OH on G‐NH3·H2O. It was reported that the attachment of OH could induce an upshift of the nitrogen‐binding energy in XPS [40]. Therefore, the observed decrease in the pyridinic‐N peak intensity and the increase in the pyrrolic‐N peak intensity after ORR could be explained. In other words, the increase in the pyrrolic‐N intensity after ORR should not be attributed to the increase in the pyrrolic‐N content, but instead of the increased content of the OH attached to the C atoms next to the pyridinic‐N. Such peak evolution of XPS spectra supports that the charge redistribution induced by pyridinic‐N will enable their neighboring C atoms as the active sites for ORR.

Image described caption and surrounding text.

Figure 9.2 XPS spectra of G‐NH3·H2O. Oxygen 1s XPS spectra (a) before and (b) after ORR. The fitted peaks are CO at 530.8 eV, C(aliphatic)OH/C(aliphatic)OC(aliphatic) at 532.0 eV, C(aromatic)OH at 533.3 eV, and chemisorbed water molecules at 535.7 eV. Carbon 1s XPS spectra (c) before and (d) after ORR. The fitted peaks are CC/CC at 284.6 eV, COH at 285.4 eV, COC at 286.5 eV, CO at 288.1 eV, and COOH at 290.8 eV. The contribution from Nafion to the carbon 1s XPS spectrum has been subtracted. (e) Nitrogen 1s XPS spectra before and after ORR. The least‐squares fitted peaks are pyridinic‐N at 398.5 eV, pyrrolic‐N at 399.8 eV, graphitic‐N at 401.2 eV, and nitrogen oxide at 403 eV. (f) Diagram of the chemical structure of OH attached to the carbon‐neighboring pyridinic nitrogen, leading to a bonding energy upshift of the pyridinic nitrogen in XPS.

Source: Xing et al. 2014 [39]. Reproduced with permission from Royal Society of Chemistry.

Image described caption and surrounding text.

Figure 9.3 Post‐ORR XPS analysis and CO2‐TPD of the N‐HOPG model catalysts. (a) N 1s XPS spectra of the N‐HOPG model catalyst before and after ORR, respectively. (b) Schematic images of the formation of pyridonic‐N by the attachment of OH to the carbon atom next to pyridinic‐N. (c) CO2‐TPD results for the HOPG model catalysts.

Source: Guo et al. 2016 [41]. Reproduced with permission from Royal Society of Chemistry.

To determine the active site and the chemical roles of pyridinic‐N, graphitic‐N, and pyrrolic‐N, a perfect model catalyst with well‐defined p‐electron conjugation of highly oriented pyrolytic graphite (HOPG), which is solely doped with one type of nitrogen, was built [41] .With the aid of such an exclusive structure design, it would be safe to compare the catalytic effect of pyridinic‐N, graphitic‐N, and pyrrolic‐N toward ORR. The key role of pyridinic‐N is revealed by XPS as shown in Figure 9.3a. According to the feature of attachment of OH onto the C atoms near to the pyridinic‐N (Figure 9.3b) together with the CO2 temperature programmed desorption (TPD) results that show strong CO2 adsorption on pyridinic‐N‐modified HOPG (Figure 9.3c), it was concluded that carbon atoms next to pyridinic‐N should be the active site for ORR. Therefore, the ORR reaction mechanism was thus proposed. Molecular oxygen would adsorb onto the C atoms next to the pyridinic‐N followed by proton‐coupled electron transfer via either of the two pathways. The first possible pathway is a four‐electron mechanism, in which a subsequent two‐proton‐coupled, two‐electron transfer breaks the OOH bond to form a water molecule. Subsequently, another proton‐coupled electron transfer causes the break of the OH bond to form another water molecule. The second pathway is a [2+2] electron mechanism, in which the H2O2 is first formed by protonation of adsorbed OOH; thereafter, the as‐formed H2O2 is further reduced to water by two protons coupled with two electrons.

How and from which direction the O2 molecule adsorbs onto the catalyst is one of the critical kinetic steps for ORR. As previously mentioned, the improved ORR catalytic performance for nitrogen‐doped, carbon‐based catalysts can be attributable to the doping‐induced charge redistribution. The nitrogen doping induces charge transfer and charge redistribution within the carbon matrix, which will alter the chemisorption mode of O2 from the usual end‐on adsorption (Pauling model) to the side‐on adsorption (Yeager model). Such transformation and the parallel diatomic O2 adsorption can effectively reduce the ORR potential and weaken the OO bond, facilitating oxygen reduction at the nitrogen‐doped CNT electrode [15].

Although nitrogen‐doped, carbon‐based nanomaterials have been demonstrated as efficient ORR electrocatalysts, there are still some concerns about the possible contribution of metal impurities to the ORR activity [22, 42]. Through the poisoning experiment, it is possible for us to rule out such possibility. For instance, the observed CO‐independent ORR activity of nitrogen‐doped, carbon‐based catalysts indicates that the enhancement can be safely ascribed to the nitrogen‐doping effect [15]. Meanwhile, the edge carbon atoms on nitrogen‐doped carbon nanostructures could be more active than those sitting on the basal plane for ORR. The configuration of doped nitrogen determines the chemical environment and thus influences the electronic structure of the neighboring C atoms, especially for the C atoms at the edge, leading to different catalytic properties. The doped nitrogen atoms near the edge could provide stronger chemical reactivity owing to its stronger electron‐withdrawing capability, which enables the positively charged C atoms with stronger oxygen adsorption affinity [43, 44] and hence higher catalytic activity toward ORR. Ball‐milled graphite and CNTs with more exposed edges were thus prepared, which showed significantly enhanced ORR activities [43].

In addition to foreign atom doping, enhanced ORR activity could also be realized by physical adsorption of positively charged polyelectrolyte (i.e. poly‐diallyldimethylammonium chloride, PDDA) [45] onto pure carbon framework. Because of the strong electron‐withdrawing ability, the physically adsorbed PDDA could create net positive charges for carbon atoms in the carbon matrix via intermolecular charge transfer. Thus, enhanced ORR activity in carbon‐based catalysts can be accomplished by either doping‐induced intramolecular charge transfer or physical/chemical‐adsorption‐induced intermolecular charge transfer.

9.2.2 Hydrogen Evolution Reaction (HER)

HER is a relatively simpler reduction process in which Volmer–Heyrovsky, Volmer–Tafel mechanism, or the combination of both [46] will take place. Thanks to those well‐established models, DFT calculation together with observation from experiments offers the most common approach to conclude the reaction mechanism as well as the active site. Graphitic‐carbon nitride (C3N4) coupled with nitrogen‐doped graphene (NG) as a metal‐free hybrid catalyst (C3N4@NG) shows extraordinary HER activity with comparable overpotential and Tafel slope to the well‐developed noble metal catalysts (i.e. Pt) [17]. By carefully controlling the composition, it is deduced that C3N4 acts as the active hydrogen adsorption site, while the highly conductive NG facilitates the electron transfer for proton reduction. Experimental observations in combination with DFT calculations further disclose that the enhanced HER electrocatalytic performance of the C3N4@NG originates not only from its increased electrical conductivity because of incorporated conductive NG but also from the complex electronic interaction between C3N4 and NG to synergistically promote HER by facilitating the adsorption of intermediates as well as reducing the activation barrier. In addition, it is also proposed that the HER reaction pathway on C3N4@NG should be dependent on potential: at low overpotential, the Volmer–Heyrovsky mechanism with a rate‐limiting step of electrochemical H2 desorption is most probable, whereas at high overpotential, the reaction is dominated by the Volmer–Tafel mechanism.

Five heteroatoms (boron, nitrogen, oxygen, sulfur, and phosphorus) chemically substituting either the edge or the central (graphitic) carbon atoms in the graphene matrix yield 15 different doping configurations with a total of 72 possible HER active sites (Figure 9.4) [10].

Image described caption and surrounding text.

Figure 9.4 Hydrogen adsorption and reaction mechanism on various graphene models. (a) ΔGH* for different models. The values of ΔGH* on graphitic‐type doping models are labeled by solid bars, whereas those on edge‐doping models are labeled by shaded bars (except for th‐S, which is represented by shaded bars due to the inability to construct a graphitic doping model). (b) Reaction pathways on edge (marked by dark grey circle) and non‐edge carbons (marked by light circle) within the py‐N model. The edge carbon possesses a ΔGH* of −0.45 eV, which should lead to a higher activity as compared with the non‐edge carbon in the same model (0.97 eV). However, the free energy change toward the next H2 recombination step of the reaction on the edge carbon is 2.92 eV, indicating that it is difficult for this step to proceed and therefore that the overall reaction pathway on the edge carbon is unfavorable. For the non‐edge site, such a recombination step is limited by the Heyrovsky route and is, therefore, not shown. Free energy diagram for HER follows the Volmer–Heyrovsky pathway (c) and the Volmer–Tafel pathway (d) on various graphene models. (e) Tafel slopes obtained from theoretical computation (filled symbols) and experimental measurements (open symbols) on various graphene models/samples.

Source: Jiao et al. 2016 [10]. Reproduced with permission from AAA.

The change in adsorption energy of H* – the surface intermediate (ΔGH*) – was employed as a descriptor, for which a smaller value of ΔGH* represents a higher activity of the site for HER. The results indicate that the most active sites are the carbon atoms rather than the heteroatoms on the non‐edge sites (Figure 9.4a), and in some cases, the carbon at the edge of specific models may exhibit lower value of ΔGH* without consideration of the high free energy change in the second step of H2 formation (Figure 9.4b). Volmer–Heyrovsky (Figure 9.4c) and Volmer–Tafel (Figure 9.4d) pathways were applied to reveal the reaction mechanism. First, the energy‐favorable Volmer–Heyrovsky pathway is considered as the dominant mechanism for carbon‐based catalyst, and the most significant free energy change for all models results from the first Volmer step, which is, therefore, assigned as the RDS for HER at the equilibrium potential. The assignment of Volmer step as the RDS is quite different from other metallic electrocatalysts such as MoS2 (42 mV dec−1 adopting Heyrovsky step as the RDS) and Pt (30 mV dec−1 adopting Tafel step as the RDS) [47]. These differences can be attributed to the weak hydrogen adsorption on graphene surfaces. The observed Tafel slopes of around 120 mV dec−1 for all graphene samples (Figure 9.4e) are in good agreement with such assignment. A more complex model with bi‐heteroatom doping was also discussed. It is concluded that the X‐doped graphene with optimum electrochemical active surface could, in principle, be synthesized by multiple‐element doping or by introducing structural defects, or their possible combinations to exhibit comparable performance to the state‐of‐the‐art Pt catalyst.

9.3 Electrochemical Oxidation: Oxygen Evolution Reaction (OER)

Overall water splitting through either photoelectrochemical or electrochemical approach provides a practical and environmentally friendly method to produce molecular hydrogen (H2) [48]. However, the efficiency of water splitting is greatly limited by the considerable overpotential requirement of OER due to its unfavorable four‐electron transfer process [49]. Expensive noble metal catalysts such as iridium and ruthenium oxides are typically used to facilitate the water oxidation process [50]. However, to sustainably utilize water electrolysis as the green energy generator, it is important to explore other efficient and affordable OER electrocatalysts made of earth‐abundant elements.

Among the various candidates, carbon‐based nanomaterials, which can be presented in the form of multidimensional nanoarchitectures with tunable electronic and surface reaction properties, have been studied as metal‐free OER electrocatalysts [13, 51–53]. However, as compared with the ORR studies, the detailed discussion on the active site toward OER based on DFT calculations or experiments was less reported [18, 54]. Generally, it is believed that the OER mechanism on pure carbon catalysts is sensitive to the catalyst structure including defects as well as functional groups like ketonic CO [13]. According to a recent theoretical simulation, it is predicted that the near‐edge armchair carbon around doped nitrogen in graphene or the adjacent C atoms close to pyridinic‐N and/or quaternary‐N would favor the water oxidation reaction [51]. Here, we separately discuss the catalytic effect of the two functional groups on carbon‐based nanomaterials, namely the oxygen‐containing species and the nitrogen‐containing species, toward the OER.

9.3.1 Oxygen Functional Group‐Induced Active Site

As oxidation of the nanocarbon substrate is generally needed to introduce oxygen functional groups, electronic charges, and defect sites for subsequent functionalization, dispersion, or doping of heteroatoms, the presentence of oxygen‐containing groups on the surface of nanocarbon materials becomes very common. Therefore, fundamental understanding of the effect of pure oxygen functional groups on carbon nanomaterials will provide the primary knowledge on the water catalytic reaction. Zhao et al. studied multiwalled carbon nanotubes (MWCNT) as the OER catalysts [13]. To systematically compare the effect of oxygen functional groups, weak, medium, and strong oxidizing agents (namely three different acidic solutions) were applied to introduce different levels of surface functional groups on MWCNTs. Hydrothermal annealing to remove unstable surface species and defects and electrochemical activation were subsequently carried out to further enhance the OER performance. The results indicated that upon mild surface oxidation, followed by subsequent hydrothermal annealing and electrochemical activation, MWCNTs could be converted to effective water oxidation catalysts.

The change of the surface chemical composition of oxidized MWCNTs (o‐MWCNTs) was monitored by XPS. The O 1s XPS spectra (Figure 9.5a) show that the surface of oxidized CNTs (O‐CNTs) is mainly composed of the ketonic CO group, while the CO group is ascribed to the epoxide and/or hydroxyl groups. After oxidizing the raw CNTs in piranha solution, both intensities for CO and CO increase. It is worth noting that the ratio of CO to CO also increases after oxidation, which is in line with the trend of the OER performance. Following a subsequent hydrothermal treatment, owing to enhanced conductivity and removal of unstable oxygen groups and residuals to uncover more active sites, further enhancement in OER activity is observed. The final electrochemical activation process further dramatically introduces more CO and CO functional groups on the surface of CNTs, leading to the dramatic enhancement in OER activity. The XPS observation suggests that the ketonic CO sites should make substantial contribution to the enhanced OER activity.

Image described caption and surrounding text.

Figure 9.5 Identification of the catalytic active site in echo‐MWCNTs. (a) High‐resolution XPS O 1s spectra obtained from raw MWCNTs, oxidized MWCNTs (o‐MWCNTs), hydrothermally treated o‐MWCNTs (ho‐MWCNTs), and electrochemically activated MWCNTs (echo‐MWCNTs), respectively. (b–d) Energy profiles of the four‐electron water oxidation mechanism. The energy corrections due to the working voltage potential of 1.53 V and the pH of 13 were taken into account. Water oxidation on preoxidized graphene cluster model with (b) both lactone and ketone groups and (c,d) ketone groups only.

Source: Lu et al. 2015 [13]. Reproduced with permission from American Chemical Society.

The mechanism of electrocatalytic oxidation of water over ketonic CO groups at o‐MWCNTs in alkaline media was explained using first‐principles calculations, in which the possible intermediates are a typical four‐electron oxidation reaction involving four elementary steps with ideal coupled electron–proton process. To corroborate with the results of functional group characterizations by XPS, two types of oxidized graphene clusters were used. The first one contains one lactone and one ketone group, while the second one contains only two ketone groups. The DFT calculations (Figure 9.5b–d) show that water oxidation reaction can only take place at the carbon sites near the oxidized functional groups, and water oxidation on o‐MWCNTs with ketonic CO functional groups is thermodynamically more feasible. Another interesting observation is that the intermediates, OH and OOH, do not adsorb at the same carbon atop adsorption site, indicating multi‐active sites of adjacent C atoms. The presence of the electron‐withdrawing ketonic CO group can lower the electron density of the carbon atoms at the meta‐position, whereas the carbon atoms at the ortho‐ and para‐positions are less affected. As a result, the OH species first adsorb at the meta‐position, and the branching reaction leads to either an ortho‐OOH or a para‐OOH intermediate.

Liu et al. discussed the detailed catalytic role of the ketonic CO in CNTs for OER based on a mild oxygen plasma treatment, which only moderately functionalizes the CNT surface without damaging the bulk structure [52]. The ex situ XPS examination before and after OER as well as the electrochemical impedance spectroscopy (EIS) measurements obtained under operando condition confirmed the essential role of ketonic CO for water oxidation. The O 1s XPS spectra (Figure 9.6a) verify that with more ketonic CO introduced onto the CNT surface, a much enhanced OER performance can thus be realized. This result delivers a rough picture of a positive correlation between the ketonic CO group and the electrocatalytic activity. The C 1s XPS spectra of O‐CNTs acquired after applying different anodic potential are shown in Figure 9.6b. Below the thermodynamic potential for water oxidation (at 1.2 V versus reversible hydrogen electrode (RHE)), the recorded current is ascribed to the double‐layer capacitance effect. The faradaic adsorption of oxygenated intermediates emerges at a relatively higher applied potential (1.4 V versus RHE), while both intermediates adsorption and their electrochemical conversion to molecular oxygen take place when the applied potential is further increased to 1.6 V versus RHE. The C 1s XPS spectra can be deconvoluted into three peaks: CC (284.6 eV), CO (285.4 eV), and CO (286.5 eV), in which the key CO fraction can be assigned to the C adsorbates bonding (COH*, CO*, or COOH*) as well as the original CO group on CNTs. The peak intensity related to CO increases from 18.4% at 1.2 V versus RHE to 31.4% at 1.4 V versus RHE, evidencing an accumulation of intermediates on O‐CNTs. Further increase in applied bias slightly reduces the CO content, which is likely due to the rapid consumption of intermediates for oxygen evolution.

Image described caption and surrounding text.

Figure 9.6 Ex situ XPS and in situ EIS measurements. (a) Deconvolution of the XPS O 1s spectra after different treatments, where P‐CNTs, R‐CNTs, and O‐CNTs represent raw CNTs, thermally reduced CNTs, and plasma O‐CNTs, respectively. (b) Ex situ C 1s XPS spectra of O‐CNTs acquired after applying different anodic potential for 15 min. (c) Nyquist plots of O‐CNTs, P‐CNTs, and R‐CNTs obtained at 1.625 V versus RHE (inset shows the electrical equivalent circuit) and (d) variation in the surface coverage of intermediates on CNTs.

Source: Li et al. 2017 [52]. Reproduced with permission from John Wiley & Sons.

The EIS analyses (Figure 9.6c) in the potential range from 1.35 to 1.65 V were conducted to estimate the surface coverage of intermediates (θ) through the Nyquist data fitting. The adsorption capacitance of intermediates, Cφ, is defined as the derivative of coverage of intermediates (θ) with potential, and thus, the surface coverage of the intermediates (θ) can be obtained by integrating Cφ within a window of applied potential. As shown in Figure 9.6d, O‐CNTs possess the largest surface coverage of the intermediates, which is in line with the best OER performance of O‐CNTs as compared with the pristine CNTs (P‐CNTs) and thermally reduced CNTs (R‐CNTs). For the first time, the combination of XPS and EIS results experimentally witnessed and verified the enhanced adsorption of oxygenated intermediates on surface O‐CNTs, which supports the conclusion that the ketonic CO groups indeed facilitate the chemical adsorption of water intermediates.

Image described caption and surrounding text.

Proposed OER pathway on O‐CNTs.

Source: Li et al. 2017 [52]. Reproduced with permission from John Wiley & Sons.

The DFT calculations on heteroatom‐doped carbon materials have demonstrated that the carbon atoms next to the dopants generally serve as the active sites in electrocatalysis due to the doping‐induced charge redistribution in the π‐conjugated system. In particular, an electron‐withdrawing group, such as ketonic CO, can attract electrons from its adjacent carbon atoms, rendering them with p‐type characteristics. The partial positive charge on the neighboring C atoms grants them with strong affinity to adsorb water intermediates. Therefore, a possible reaction pathway was proposed by Liu et al., in Scheme 9.1, to explain the effect of oxygen functional groups on OER. The main active sites for OER would originate from the carbon atoms near the CO group. The highly electronegative oxygen‐containing groups could induce positive charges to the adjacent carbon atoms, facilitating adsorption of key intermediates during OER.

9.3.2 Nitrogen Functional Group‐Induced Active Site

Nitrogen‐doped carbon nanomaterials can be prepared by pyrolysis of nitrogen‐rich polymers, which have exhibited outstanding OER activities in alkaline media even exceeding those of the traditional noble metal electrocatalysts [51]. The pyrolysis process could make the heteroatom dopants, such as nitrogen, escape from the carbon matrix due to surface rearrangement and/or recrystallization. Therefore, carefully controlling the pyrolysis in terms of temperature and duration would obtain different concentrations of nitrogen‐doped carbon nanostructures, making it possible to investigate the effect of nitrogen dopant on OER activity. It has been revealed that with simultaneous increase in pyridinic‐N and quaternary‐N functional groups, the OER activity of the carbon‐based catalysts can be improved. These results strongly suggest that nitrogen‐related species would be responsible for the OER activity, even though the clear individual role of pyridinic‐N and quaternary‐N is still missing. Detailed electrochemical and spectroscopic studies together with DFT calculations should be helpful to discern the active site as well as to distinguish the specific role played by different nitrogen species.

Additionally, both nitrogen and oxygen functional groups could be simultaneously introduced into the carbon‐based catalysts to further improve the catalytic activity [53]. The thus formed dual active sites, nitrogen‐containing groups and/or oxygen‐containing groups, would provide numerous catalytic centers with possible synergistic effect resulted from the electron‐withdrawing features of the nitrogen‐ and oxygen‐containing species.

9.4 Bifunctional ORR and OER Electrocatalyst

Over the past few years, there has been an increasing number of reports on efficient carbon‐based bifunctional electrocatalysts such as ORR–OER and ORR–HER [14, 16, 55]. Among them, ORR–OER‐coupled catalytic reactions are most interesting [56]. With an increasing demand on the development of renewable energy and electrified transportation of smart grids, new systems for energy storage will become much more critical in future. The secondary energy storages, such as lithium‐ion batteries (LIBs), have been considered the most promising candidate for next‐generation electronics owning to their long cycling durability and acceptable high efficiency [57]. However, the limited energy density (normalized by the total weight of the cell) together with the high cost of the electrode materials still holds up their long‐term application [58]. Alternatively, metal–air batteries are proposed as another potential large‐scale electricity storage and energy production technology by substituting the sophisticated intercalation reaction of LIBs with a fast electrocatalytic redox reaction of metal–oxygen couple, in which a metal anode with high energy density is combined with an air cathode having open structure to draw the active reactant (i.e. oxygen gas) [2, 59]. During discharging and charging, the ORR and OER are the thermodynamic bottlenecks on the air electrodes, which limit the overall cell efficiency. Thus, an efficient, durable, affordable, and, most importantly, bifunctional (ORR–OER) electrocatalyst is the key for the wide‐range application of metal–air battery technology. Currently, the ORR and OER still rely on traditional noble metals and metal oxides. Although these metal‐based catalysts exhibit good catalytic activity, they often simultaneously suffer from multiple drawbacks, including high cost, poor stability, etc.

Dai et al. reported a new type of high surface area mesoporous carbon foam co‐doped with nitrogen and phosphorus as superior bifunctional electrocatalyst for both ORR and OER [55]. In this work, DFT calculations were used to gain insights about the bifunctional electrocatalytic activity. Following this work, soft XAS was applied for comparison between the pre‐ and post‐chemical states of the bifunctional nitrogen‐doped, carbon‐based electrocatalyst [16], which showed distinctive roles of various nitrogen‐containing functional groups and concluded the two different functioning sites for ORR and OER.

9.4.1 Density Functional Theory (DFT) Calculation Approach

DFT calculation offers a powerful tool to study electrocatalysis. As shown in the DFT calculation results (Figure 9.7), co‐doping of N and P onto graphene gives a wide range of dopant configurations, e.g. isolated N dopant, isolated P dopant, and N,P‐coupled dopants (in which N and P are close to each other). Thus, to simulate the ORR/OER catalytic activities, all types of doping structures have to be considered for the DFT calculation. Moreover, to further reveal the effect of doping sites, the doping positions in each of the structures need to be changed with respect to the graphene edge. So far, without any further advanced kinetic consideration, it is generally accepted that lower overpotential corresponds to better catalytic activity. Figure 9.7a,b presents the volcano plots using overpotential as the descriptor for various reaction sites on N,P‐co‐doped graphene in alkaline media. Particularly, the distance of the doping sites from the graphene edge seems critical for the adsorption energy of the intermediates and, therefore, the overall catalytic activity. Note that the N,P‐coupled doping gives the best OER and ORR performance, indicating the importance of simultaneous presence of N and P in the matrix of graphene. Figure 9.7c–f shows the most possible active structure of N,P‐co‐doped graphene as well as the elementary reactions of OER, where the active sites are located at the edge of graphene. Similar phenomena are also observed for ORR; yet, the most active site is the N‐dopant edge site. Figure 9.7g,h displays the schematic energy profiles for the OER and ORR pathways, respectively.

Image described caption and surrounding text.

Figure 9.7 Mechanistic study of bifunctionality for ORR and OER. (a) ORR and (b) OER volcano plots of overpotential η versus adsorption energy of O* and the difference between the adsorption energy of O* and OH*, respectively, for N‐doped, P‐doped, and N,P‐doped graphene. (c) Initial structure and structures after the adsorption of (d) hydroxyl OH*, (e) oxyl O*, and (f) peroxyl OOH* intermediates on N‐ and P‐coupled graphene. O*, OH*, and OOH* are adsorbed intermediates. The overpotentials of the best catalysts predicted theoretically for ORR (Pt) [60] and OER (RuO2) [61] are also plotted in (a) and (b), respectively. Inset in (b) shows the details of the volcano top in the main panel. Schematic energy profiles for the (g) OER pathway and the (h) ORR pathway on N,P‐co‐doped graphene in alkaline media.

Source: Zhang et al. 2016 [14]. Reproduced with permission from John Wiley & Sons.

9.4.2 Soft X‐ray Absorption Spectroscopy (XAS) Approach

Liu et al. reported a three‐dimensional (3D), nitrogen‐doped graphene nanoribbon network (N‐GRW) bifunctional electrocatalyst based on a pyrolysis method, which exhibited superb electrocatalytic activities for both ORR and OER in alkaline electrolyte [16]. The unique 3D nanoarchitecture provides high densities of the ORR and OER active sites and facilitates the electrolyte diffusion and electron transport. It is noteworthy that, for the first time, it was experimentally demonstrated and observed by spectroscopic approaches that the electron‐donating quaternary‐N sites were responsible for ORR, whereas the electron‐withdrawing pyridinic‐N moieties served as the active sites for OER. Mott–Schottky experiment was first conducted to identify the doping states of the N‐doped graphene catalysts. N‐doped graphene has been shown both theoretically and experimentally to be either p type or n type. In general, n‐type behavior is usually found in quaternary/pyrrolic‐N doping, whereas p‐type behavior can be ascribed to the pyridinic‐N doping. Because N‐GRW simultaneously possesses quaternary‐N and pyridinic‐N in its structure, a special coexistence of the n‐type and p‐type domain is therefore present.

The surface‐sensitive soft X‐ray absorption near‐edge structure (XANES) spectroscopic measurements on carbon and nitrogen edges were carried out before and after the electrochemical reaction to deduce the active sites for OER and ORR on N‐GRW. Figure 9.8 displays the evolution of carbon and nitrogen K‐edge XANES spectra of N‐GRW before and after ORR and OER. The peaks of carbon and nitrogen K‐edge XANES spectra were assigned according to the literature database as indicated in the caption of Figure 9.8a. The peak intensity at 287.7 eV assignable to πCOC, CN in the C K‐edge XANES spectra was enhanced after both ORR and OER, suggesting adsorption of intermediate species (O*) on carbon atoms, which is consistent with the newly emerged peak relating to adsorption of OOH* intermediate at 289.6 eV.

Image described caption and surrounding text.

Figure 9.8 Electronic characteristics and ORR/OER active sites of N‐doped graphene catalysts. (a,b) Carbon and nitrogen K‐edge XANES spectra of N‐GRW catalyst, acquired under ultrahigh‐vacuum, pristine (black line), after ORR (light grey line), and after OER (thick grey line). In carbon K‐edge XANES spectra, A: defects, B: π*CC, C: π*COH, D: π*COC, CN, E: π*CO, COOH, F: σ*CC. (c) Schematic diagram of ORR and OER occurring at different active sites on the n‐ and p‐type domain of the N‐GRW catalyst.

Source: Yang et al. 2016 [16]. Reproduced with permission from Science Advances.

The variation of the nitrogen K‐edge XANES spectra is informative for probing the nitrogen‐containing group effect toward the electrocatalytic activity. After ORR, the nitrogen K‐edge XANES spectrum of N‐GRW as given in Figure 9.8b displays a newly emerging peak as the shoulder of graphitic‐N at the lower energy side at 401 eV, which is contributed by the distortion of the carbon matrix induced by the O* and/or OOH* adsorption (both of them are critical intermediates for ORR) on the neighboring C atoms of graphitic‐N. On the other hand, the intensity and position of the pyridinic‐N peak at 398.0 eV remain nearly unchanged. The quaternary‐N atoms are able to donate electrons into the π‐conjugated system, leading to increased nucleophile strength for the adjacent C atoms to enhance the O2 adsorption owing to high densities of O lone pair electrons of O2, hence promoting the ORR efficiency. A similar scenario can also be applied to the neighboring C atoms of pyrrolic‐N. These results indicate that the quaternary‐N with electron‐donating feature (n‐type doping) is responsible for the ORR activity on N‐GRW.

Interestingly, the changes of the nitrogen K‐edge XANES spectra give a totally different pattern after OER. The width at half maximum of the pyridinic‐N peak (∼398.0 eV) increases from 0.8 to 1.15 eV, together with the appearance of another new peak as the shoulder at relatively higher energy side of pyridinic‐N. On the contrary, graphitic‐N and pyrrolic‐N remain nearly the same before and after OER. Because the lone pair electrons of N involve in the resonance to delocalize electrons, the N atoms of the pyridinic‐N become more electron deficient. As a result, the electrons of adjacent C atoms would be pulled out toward pyridinic‐N, leading to an enhanced adsorption affinity for the OER intermediates. In other words, the p‐type region of N‐GRW can accept electrons from the adsorbed moieties (e.g. OH−) and thus facilitate the very first elementary step of OER (OH− → OHads + e−). The observed influences as well as the individual electrochemical role of the graphitic‐N and pyridinic‐N toward ORR and OER can thus be understood based on the doping‐induced charge redistribution effect as illustrated in Figure 9.8c. As discussed above, this should be the first experimental evidence to distinguish the role of different nitrogen species in nitrogen‐doped carbon nanomaterials toward different electrocatalytic reactions.

9.5 CO2 Reduction Reaction (CO2RR)

Carbon dioxide (CO2) emission resulted from combustion of fossil fuels greatly threatens the human society and has become the most urgent concern at the global level [62]. It has been predicted that in the coming few decades, ∼500 Gt of CO2 will be produced if we do not divert our dependence on petrochemical energy [63]. Fortunately, recent researches have demonstrated that CO2 can be electrochemically converted back into fuels or value‐added chemicals [9]. This approach not only helps to address the issue of controlling the CO2 concentration in atmosphere but also offers an alternative strategy for renewable fuels or chemical production. In recent years, increasing efforts are being devoted on the electrochemical CO2RR [64].

Currently, there are still many issues to be addressed in electrochemical CO2RR such as the high overpotential, poor product selectivity, and low faradaic efficiency [65]. Several metal‐based catalysts have been studied to boost the electrochemical CO2 reduction kinetics, and among them, copper or copper‐oxide‐derived catalysts are very interesting and they are capable to promote hydrocarbon products, mostly methane, ethylene, and ethanol, through a multiple‐step electron transfer pathway [66] with acceptable efficiencies [67]. However, the overpotentials to produce hydrocarbons and oxygenates are typically high, and the selectivity and durability still pose the problem [68]. Thus, development of new catalysts is urgently needed to improve the product selectivity and efficiency, while simultaneously lowering the overpotential.

Electrochemical CO2RR has been successfully demonstrated on metal‐free, carbon‐based [69–71] or nitrogen‐based catalysts [72, 73] with good catalytic activity, high faradaic efficiency, decent durability, and/or specific product selectivity. However, determination of the reaction mechanism and the active site on such carbon‐based catalysts is still very challenging. At this current stage, we still rely on experimental observation (pre‐/post‐ spectroscopic comparison) and theoretical simulation to deduce the possible reaction mechanism and active site. Next, we are going to present some examples of nitrogen‐doped, carbon‐based electrocatalysts to illustrate the idea.

9.5.1 Selective Conversion of CO2 to CO

Through the pyrolysis of electrospun nanofibers of polyacrylonitrile (PAN) polymer, a nitrogen‐modified, metal‐free carbon nanofiber (CNF) electrocatalyst was thus synthesized for conversion of CO2 into CO [69]. In the experiment, 1‐ethyl‐3‐methylimidazolium tetrafluoroborate (EMIM‐BF4) ionic liquid was chosen as the electrolyte owing to its high CO2 solubility [74] and low overpotential dropping for CO2 reduction [64]. Inside the CNF structure, pyridinic‐N, quaternary‐N, and N‐oxide species were identified as the main nitrogen‐containing groups by XPS (Figure 9.9). After electrochemical CO2RR, the peak intensity of nitrogen oxide decreases with simultaneous emerge of additional peak of pyridonic‐N. The quantitative elemental analysis indicates that nitrogen oxides are converted into pyridonic‐N. However, such a transformation does not affect the overall CO2RR activity, suggesting that nitrogen oxides should not take part in the CO2 conversion process. More interestingly, the quantitative elemental analysis shows that the amount of pyridinic‐N and quaternary‐N remains approximately constant before and after CO2RR. Let us consider two possible scenarios for pyridinic‐N as the active site. First, during the redox reaction, protonation of pyridinic‐N would generate a quaternary‐N‐like structure. Second, pyridinic‐N could weakly bind with molecular CO2 in a way similar to the CO2 reduction process on pyridine, leading to conversion of pyridinic‐N into pyridonic‐N‐like species. Both of the two scenarios should lead to the reduction in the pyridinic‐N intensity. However, spectroscopic studies give a totally different observation, suggesting that direct participation of nitrogen species in CO2RR is unlikely. Instead, the positively charged C atoms, induced by the strong electron affinity of the neighboring nitrogen dopants, would take the responsibility as the active site. The CO2 intermediates could adsorb onto partially reduced C atoms and reoxidize these active sites to their initial state during the reduction of themselves to form the final products. Thus, a mechanism for the reduction of CO2 to CO could be proposed, which involves the redox process by the adsorbed intermediate complex (1‐ethyl‐3‐methylimidazolium–CO2) and the release of CO gas.

Image described caption and surrounding text.

Figure 9.9 Evolution of nitrogen atomic structure in CNFs. (a) Deconvoluted N 1s spectra for CNFs before and (b) after electrochemical CO2RR. (c,d) show the corresponding atomic structure deduced from the XPS analysis.

Source: Kumar et al. 2013 [69]. Reproduced with permission from Springer Nature.

9.5.2 CO2 Reduction to Multiple Products

Nitrogen‐doped graphene quantum dots (NGQDs) are able to catalyze the electrochemical reduction of CO2 into multi‐carbon hydrocarbons and oxygenates [70]. To study the functional active site, the NGQD samples were examined by ex situ post‐reaction XPS (Figure 9.10). After electrochemical CO2RR, it reveals that the relative content of pyridinic‐N (at 398.5 eV) decreases from 65% to 38% and another N component (pyrrolic‐N or pyridonic‐N) at 400.0 eV increases from 20% to 50%, whereas the percentage of graphitic‐N (at 401.2 eV) remains nearly unchanged. The variation of the nitrogen‐containing species could be possibly ascribed to the adsorption of CO2 onto the pyridinic‐N site (here, pyridinic‐N is considered the potential active site for a rational discussion), which causes an upshift of binding energy of pyridinic‐N to a value similar to that of the pyrrolic‐N. To further explore the active site, N‐doped reduced graphite oxides (NRGOs) with less exposed edges (dominated by the basal planes) but with similar compositions of nitrogen‐containing species were prepared for comparison. It is found that the structure of the catalyst plays the most important role in determining the overall activity toward CO2 reduction. However, unfortunately, the detailed mechanism of the conversion of CO2 over NGQDs to multi‐carbon hydrocarbons and oxygenates as well as the active site remains elusive. The combination of highly surface‐sensitive operando spectroscopic measurement together with first‐principles calculation should provide insights into the reaction intermediates to disclose the reaction pathways as well as the active site.

Image described caption and surrounding text.

Figure 9.10 The active N site for CO2 adsorption. (a) Post‐CO2 reduction XPS analysis showing the change of N configuration concentration. (b) A schematic of CO2 adsorption onto the pyridinic‐N site.

Source: Wu et al. 2016 [70]. Reproduced with permission from Springer Nature.

9.5.3 Selectively Reduction of CO2 to Formate

It is important to stabilize the key intermediates during the CO2RR to enhance the product selectivity. Thomas' group has successfully demonstrated that using polyethylenimine (PEI), a polymer with amine functional groups that is commonly used as the CO2 absorbent [75], as the cocatalyst together with nitrogen‐doped CNTs, the critical CO2RR intermediate, CO2·−, can be stabilized on the surface of CNTs for the further reduction and protonation, leading to a selective transformation of CO2 to formate [71]. The combined use of N‐doped CNTs and PEI results in a significant reduction in the overpotential and enhanced faradaic efficiency for selective CO2 reduction to formate in an electrochemical cell. First, the high‐resolution N 1s XPS spectrum (Figure 9.11a) shows that N2 plasma treatment produces two dominant nitrogen‐containing species, pyridinic‐N and pyrrolic‐N, on CNTs. The N atoms are negatively charged due to their strong electron‐withdrawing capability in the graphene‐π electronic structure [15]. In the proposed CO2RR mechanism (Figure 9.11b), CO2 is first reduced to CO2·− at the nitrogen‐binding site [73] coordinated by PEI with H‐bond interaction effect. The stabilized CO2·− is protonated and reduced to HCO3−. Further protonation is followed by a rapid second electron transfer reduction to give formate as the final product. Unfortunately, all of the above‐proposed mechanism is just a hypothesis. In addition, it is not guaranteed that the reaction of the intermediates will take place on nitrogen atoms. Further theoretical and experimental evidences are needed to support this hypothesis.

Image described caption and surrounding text.

Figure 9.11 (a) Deconvoluted N 1s spectrum for N‐doped CNTs, which elucidates the existence of four main nitrogen species: pyridinic N (B.E. ∼ 398.9 eV), pyrrolic N (B.E. ∼ 400.1 eV), quaternary N (B.E. ∼ 401.5 eV), and nitrogen oxide (B.E. ∼ 402.2 eV). (b) Proposed mechanism for CO2 reduction at the PEI‐functionalized, nitrogen‐doped CNTs.

Source: Zhang et al. 2014 [71]. Reproduced with permission from American Chemical Society.

9.6 Identification of Possible Active Site by Poisoning

Identification of possible active site on metal‐free, carbon‐based electrocatalyst through poisoning could be another effective and promising approach [76]. Here, we use nitrogen‐doped carbon (CNx) nanostructured ORR catalyst as an example to illustrate the idea. So far, there is still limited information about the ORR active site on CNx catalysts. A correlation of the ORR activity with the pyridinic‐N content has been proposed, but it is not clear if the pyridinic‐N species are the actual active site or just function as the “indicator” to illustrate the neighboring carbon active sites. Several possible ORR active sites on CNx have been suggested, including pyridinic‐N [33, 34], quaternary‐N [77], and C atoms adjacent to pyridinic‐N [41]. The latter proposal is consistent with the pyridinic‐N sites acting as the “indicator.” For metal‐centered ORR catalysts, it is convenient to determine the possible active sites by selectively poisoning them with CO [78], CN− [79], or H2S [30]. Unfortunately, CNx catalysts are inert to the aforementioned poisoning molecules, which make it challenging to identify the active site on CNx catalysts via the similar approach [80]. Thus, finding a useful marking molecule to passivate the active site on CNx catalysts is a priority to realize this probing method. Umit S. Ozkan's group identified [76] a possible probe molecule, phosphate anion, which could poison CNx toward ORR. With the aid of spectroscopic techniques, i.e. XPS, it is possible to identify the active site on metal‐free, carbon‐based electrocatalysts.

9.6.1 Electrochemical Testing

ORR performance was evaluated before and after exposing the catalyst‐coated electrode into 0.1 M phosphoric acid H3PO4. The rate constant (k) and selectivity between H2O2 and H2O formation were estimated by measuring the total number of electrons passed during ORR using Koutecky−Levich equation or rotating ring disk electrode (RRDE) method. In addition, Tafel analysis was performed to obtain information on the possible RDS for ORR. First, infrared (IR) spectrum confirms the presence of adsorbed H2PO4− species on soaked CNx (0.1 M H3PO4). ORR polarization curves exhibited obviously reduced activity in 0.1 M H3PO4 electrolyte than in 0.1 M HClO4. The performance of H3PO4‐exposed CNx cannot be recovered even after transferring the electrode back to the pure HClO4 electrolyte (Figure 9.12a). The similar trend of decreasing performance of soaked CNx was also observed (Figure 9.12b). The rate constant (k) for phosphate‐anion‐poisoned CNx is evidently decreased as compared with the non‐poisoned sample, indicating that the phosphate anion can hinder the activity of CNx by blocking the active site. However, the ORR selectivity (n) was found to be close to 4 for both samples, indicating a complete oxygen reduction process to water. In addition, the Tafel slope analysis disclosed quite similar numbers in both cases. These observations suggest that H3PO4 soaking does not alter the ORR pathway or the RDS.

Image described caption and surrounding text.

Figure 9.12 (a) ORR polarization curves of CNx catalyst in 0.1 M H3PO4 and 0.1 M HClO4 before and after exposing the catalyst‐coated electrode to 0.1 M H3PO4. (O2‐saturated 0.1 M HClO4, 1600 rpm, 10 mV s−1, and 800 μg catalyst cm−2). (b) ORR polarization curves of CNx catalyst before and after soaking in 0.1 M H3PO4.

Source: Mamtani et al. 2016 [76]. Reproduced with permission from American Chemical Society.

9.6.2 XPS Measurement

The variation of chemical nature of the surface species and nitrogen‐containing groups before and after H3PO4 soaking was studied using XPS. A comparison of the P 2p, O 1s, and N 1s XPS spectra between pristine CNx and soaked CNx is shown in Figure 9.13.

Image described caption and surrounding text.

Figure 9.13 P 2p, O 1s, and N 1s XPS spectra for CNx catalyst before and after soaking in 0.1 M H3PO4.

Source: Mamtani et al. 2016 [76]. Reproduced with permission from American Chemical Society.

As expected, there is no residual metal observed on both pure CNx and H3PO4‐soaked CNx catalysts. Additionally, no phosphorus with corresponding oxygen signal of PO and POH on the surface of pristine CNx can be detected. For nitrogen 1s XPS spectra, three types of nitrogen functional groups are present, namely pyridinic‐N (∼398.3 eV), quaternary‐N (400.7–400.8 eV), and pyridinic nitrogen oxide, N+O− (402.4–402.5 eV), for both pristine CNx and H3PO4‐soaked CNx samples. The N 1s distribution of the nitrogen‐containing groups for soaked CNx samples varies with the concentration of H3PO4. With an increase in the concentration of H3PO4, the content of nitrogen‐containing species reduces [76]. It is interesting to note that the decrease in ORR performance is in line with the reduction of the pyridinic‐N content. XPS results indeed provide some valuable information on poisoning of the CNx catalyst by H2PO4−. It seems reasonable to deduce that the pyridinic‐N is responsible for the overall ORR activity. If this is the case, pyridinic‐N species will be protonated to pyridinic‐NH as a result of H3PO4 soaking where the proton (H+) is stabilized by the presence of a neighboring H2PO4− anion with negative charge, which will result in a decrease in the ORR activity. It should be noted that pyridinic‐NH possesses similar planar geometry as the quaternary‐N species bonded with three C atoms. Thus, XPS cannot distinguish their difference. Therefore, it can be observed that with the decrease in pyridinic‐N fraction, the quaternary‐N fraction increases, while the summation of the two fractions remains the same. These results support the pyridinic‐N species as the indicator for the neighboring C active sites. Owing to the high electronic affinity of the pyridinic‐N groups, the electron density around the adjacent C centers will be redistributed and these C atoms become more electronegative, making themselves as the Lewis base for O2 adsorption. After poisoning the pyridinic‐N sites with H3PO4 to form pyridinic‐NH, protonation of pyridinic‐N would alter the electronic interaction with the adjacent C atoms. As a result, the adjacent C atoms become less electronegative due to reduced Lewis acidity of the protonated pyridinic‐N sites, which thus decreases the ORR activity. The aforementioned two types of active site models are consistent with the fact that the ORR activity decreases linearly with a decrease in the pyridinic‐N content as a result of H3PO4 soaking. Finally, there still exists a minor issue of the phosphate‐anion‐poisoning method that the coordinating H2PO4− ligand around the pyridinic‐NH would act as a barrier to prevent the reactants from accessing the active sites.

9.7 Summary

Electrochemical reduction reaction on metal‐free, carbon‐based catalysts is important in the field of catalysis. Massive efforts have been devoted to studying the relationship between surface functional groups and the catalytic activity. With the aid of model catalysts, the ORR active sites on N‐doped carbon catalysts are determined as the carbon atoms with Lewis basicity next to the pyridinic‐N, where the Lewis basicity induced by the neighboring electron‐withdrawing pyridinic‐N is believed to facilitate the O2 adsorption. Compared with ORR, HER on carbon‐based catalysts has been systematically studied by well‐considered DFT calculations on representative models, which leads to a two‐step experimentally achievable strategy, concerning intrinsic electronic structure and extrinsic physicochemical characteristics.

The amount of ketonic CO groups is in line with the trend of OER activity, and the critical role of ketonic species has been verified by DFT calculation, ex situ XPS, and in situ EIS measurements. Due to the p‐type‐doping (electron‐withdrawing) feature, the CO groups are able to alter the electronic density around the adjacent carbon atoms, which facilitates the adsorption of oxygenated intermediates and therefore promotes the overall OER efficiency. In addition, nitrogen‐doped carbon nanomaterials also exhibit superb OER performances; some can even exceed the metal‐based catalyst. The nitrogen‐containing functional groups, e.g. pyridinic‐N and quaternary‐N species, in the carbon matrix formed by nitrogen doping are anticipated to be responsible for the enhanced catalytic activity. The C atoms adjacent to the nitrogen atoms are positively charged owing to the electron‐withdrawing property of nitrogen in a graphene π‐electron system, which should enhance the adsorption of oxygenate intermediates (e.g. OH−).

The active sites on bifunctional, metal‐free, carbon‐based catalysts were first deduced by the DFT calculation, which highlights the essential roles of N,P co‐doping and graphene edge effects for the high electrocatalytic activity. The multiple independent electrocatalytic roles were further studied using soft XAS technique, which reveals that the ORR and OER catalytic sites are associated with different N species, graphitic‐N/pyrrolic‐N for ORR, and pyridinic‐N for OER, respectively. The thus gained information from the combination of DFT and XAS studies should lead to the design and development of various bifunctional catalysts from other non‐metal, heteroatom‐doped carbon nanomaterials.

Electrochemical CO2 reduction to CO or multi‐carbon hydrocarbons and oxygenates have been demonstrated on nitrogen‐doped carbon nanomaterials. In addition, it is also possible to selectively convert CO2 into other desirable products by stabilizing the key intermediates. However, among those reactions, the reaction mechanisms and active sites are still proposed based on hypothesis. In many cases, the spectroscopic observations on catalyst before/after the CO2RR even could not provide any useful information. The combination of operando surface‐sensitive spectroscopy and carefully designed DFT calculation should provide a valuable tool to study the CO2RR.

The poisoning approach offers a straightforward method to identify the active site on carbon‐based catalysts under operando condition, which has been widely applied on metal‐based catalysts as various molecules have strong interactions with metal atoms. Interaction of phosphoric acid with surface functional groups on CNx catalysts was discovered. It was found that phosphate‐anion‐adsorbed CNx catalyst showed much reduced ORR activity, but the ORR pathway remained the same. Furthermore, it was revealed that pyridinic‐N would be converted to pyridinic‐NH through protonation by H3PO4, which eventually impairs the ORR activity of phosphate‐anion‐poisoned CNx catalyst.

Although it seems that the active site and reaction mechanism on metal‐free, carbon‐based catalysts toward various electrochemical reactions have been successfully deduced by a combination of spectroscopic experiments and DFT calculations, we are still aware about the existence of some contradictions. For instance, the roles of electron‐withdrawing/donating features of pyridinic‐N/pyrrolic‐N for the neighboring carbon atoms during ORR are still elusive. Despite that the carbon atoms with Lewis basicity induced by pyridinic‐N were proposed as the active site for the reduction reaction of oxygen with the aid of the model catalyst study [41], the same carbon atoms were also found to be responsible for electrochemical oxidation of water based on spectroscopic observations [16]. Even though many comprehensive explanations toward the mechanism of ORR and the roles of pyridinic‐N and pyrrolic‐N have been proposed, it is still not clear which electronic affinity, electron‐withdrawing (pyridinic‐N) or electron‐donating (pyrrolic‐N), truly benefits the interaction between the intermediates and the catalysts during the reduction process. The combination of the ideal model catalyst as a studying platform as well as the development of in situ spectroscopic techniques with good time‐resolved resolution (<ps) should offer the most effective way to clarify the underlying stepwise elementary reaction mechanisms during electrochemical reaction. The insights thus obtained would contribute as the stepping‐stone toward the qualitative and quantitative identification of the critical active sites on the universal heteroatom‐modified, carbon‐based catalysts.

References

  1. 1 (a) Schlogl, R. (2010). ChemSusChem 3: 209.(b) Centi, G. and Perathoner, S. (2010). ChemSusChem 3: 195.
  2. 2 Cheng, F.Y. and Chen, J. (2012). Chem. Soc. Rev. 41: 2172.
  3. 3 Lefèvre, M., Dodelet, J.P., and Bertrand, P. (2000). J. Phys. Chem. B 104: 11238.
  4. 4 Vielstch, W., Lamm, A., and Gasteiger, H.A. (2003). Handbook of Fuel Cells: Fundamentals, Technology and Applications.
  5. 5 Debe, M.K. (2012). Nature 486: 43.
  6. 6 (a) Kudo, A. and Miseki, Y. (2009). Chem. Soc. Rev. 38: 253.(b) Walter, M.G., Warren, E.L., McKone, J.R. et al. (2010). Chem. Rev. 110: 6446.
  7. 7 (a) Millet, P., Ngameni, R., Grigoriev, S.A. et al. (2010). Int. J. Hydrogen Energy 35: 5043.(b) Lee, Y., Suntivich, J., May, K.J. et al. (2012). J. Phys. Chem. Lett. 3: 399.
  8. 8 Piana, M., Catanorchi, S., and Gasteiger, H.A. (2008). Proton Exchange Membrane Fuel Cells 8, Pts 1 and 2, vol. 16 (ed. T. Fuller, K. Shinohara, V. Ramani, et al.), 2045.
  9. 9 (a) Benson, E.E., Kubiak, C.P., Sathrum, A.J., and Smieja, J.M. (2009). Chem. Soc. Rev. 38: 89.(b) Whipple, D.T. and Kenis, P.J.A. (2010). J. Phys. Chem. Lett. 1: 3451.
  10. 10 Jiao, Y., Zheng, Y., Davey, K., and Qiao, S.Z. (2016). Nat. Energy 1: 16130.
  11. 11 Jiao, Y., Zheng, Y., Jaroniec, M.T., and Qiao, S.Z. (2015). Chem. Soc. Rev. 44: 2060.
  12. 12 Montoya, J.H., Seitz, L.C., Chakthranont, P. et al. (2017). Nat. Mater. 16: 70.
  13. 13 Lu, X.Y., Yim, W.L., Suryanto, B.H.R., and Zhao, C. (2015). J. Am. Chem. Soc. 137: 2901.
  14. 14 Zhang, J.T., Qu, L.T., Shi, G.Q. et al. (2016). Angew. Chem. Int. Ed. 55: 2230.
  15. 15 Gong, K.P., Du, F., Xia, Z.H. et al. (2009). Science 323: 760.
  16. 16 Yang, H.B., Miao, J.W., Hung, S.F. et al. (2016). Sci. Adv. 2: e1501122.
  17. 17 Zheng, Y., Jiao, Y., Zhu, Y.H. et al. (2014). Nat. Commun. 5: 3783.
  18. 18 Liu, X. and Dai, L.M. (2016). Nat. Rev. Mater. 1: 16064.
  19. 19 Nilsson, A., LaRue, J., Öberg, H. et al. (2017). Chem. Phys. Lett. 675: 145.
  20. 20 Hammer, B. and Norskov, J.K. (2000). Impact of surface science on catalysis. In: Advances in Catalysis, vol. 45 (ed. B.C. Gates and H. Knozinger), 71.
  21. 21 Dai, L.M., Xue, Y.H., Qu, L.T. et al. (2015). Chem. Rev. 115: 4823.
  22. 22 Masa, J., Xia, W., Muhler, M., and Schuhmann, W. (2015). Angew. Chem. Int. Ed. 54: 10102.
  23. 23 Knop‐Gericke, A., Kleimenov, E., Havecker, M. et al. (2009). Advances in Catalysis, vol. 52 (ed. B.C. Gates and H. Knozinger), 213.
  24. 24 Nilsson, A. and Pettersson, L.G.M. (2004). Surf. Sci. Rep. 55: 49.
  25. 25 Shui, J., Wang, M., Du, F., and Dai, L. (2015). Sci. Adv. 1: e1400129.
  26. 26 (a) Walter, M.G., Warren, E.L., McKone, J.R. et al. (2010). Chem. Rev. 110: 6446.(b) Li, Y.G., Wang, H.L., Xie, L.M. et al. (2011). J. Am. Chem. Soc. 133: 7296.(c) Suntivich, J., May, K.J., Gasteiger, H.A. et al. (2011). Science 334: 1383.
  27. 27 (a) Lefevre, M., Proietti, E., Jaouen, F., and Dodelet, J.P. (2009). Science 324: 71.(b) Proietti, E., Jaouen, F., Lefevre, M. et al. (2011). Nat. Commun. 2: 416.
  28. 28 (a) Matter, P.H., Wang, E., Arias, M. et al. (2007). J. Mol. Catal. A: Chem. 264: 73.(b) Zhang, S.M., Liu, B., and Chen, S.L. (2013). Phys. Chem. Chem. Phys. 15: 18482.
  29. 29 (a) Tylus, U., Jia, Q.Y., Strickland, K. et al. (2014). J. Phys. Chem. C 118: 8999.(b) Jaouen, F., Lefèvre, M., Dodelet, J.‐P., and Cai, M. (2006). J. Phys. Chem. B 110: 5553.
  30. 30 Singh, D., Mamtani, K., Bruening, C.R. et al. (2014). ACS Catal. 4: 3454.
  31. 31 Zitolo, A., Goellner, V., Armel, V. et al. (2015). Nat. Mater. 14: 937.
  32. 32 Wang, S.Y., Zhang, L.P., Xia, Z.H. et al. (2012). Angew. Chem. Int. Ed. 51: 4209.
  33. 33 Rao, C.V., Cabrera, C.R., and Ishikawa, Y. (2010). J. Phys. Chem. Lett. 1: 2622.
  34. 34 Qu, L.T., Liu, Y., Baek, J.B., and Dai, L.M. (2010). ACS Nano 4: 1321.
  35. 35 (a) Yu, D.S., Nagelli, E., Du, F., and Dai, L.M. (2010). J. Phys. Chem. Lett. 1: 2165.(b) Hu, C.G. and Dai, L.M. (2016). Angew. Chem. Int. Ed. 55: 11736.
  36. 36 Lai, L.F., Potts, J.R., Zhan, D. et al. (2012). Energy Environ. Sci. 5: 7936.
  37. 37 Yu, L., Pan, X.L., Cao, X.M. et al. (2011). J. Catal. 282: 183.
  38. 38 Zhang, L.P. and Xia, Z.H. (2011). J. Phys. Chem. C 115: 11170.
  39. 39 Xing, T., Zheng, Y., Li, L.H. et al. (2014). ACS Nano 8: 6856.
  40. 40 (a) Kelemen, S.R., Gorbaty, M.L., and Kwiatek, P.J. (1994). Energy & Fuels 8: 896.(b) Friebel, J. and Kopsel, R.F.W. (1999). Fuel 78: 923.
  41. 41 Guo, D.H., Shibuya, R., Akiba, C. et al. (2016). Science 351: 361.
  42. 42 (a) Masa, J., Zhao, A.Q., Xia, W. et al. (2013). Electrochem. Commun. 34: 113.(b) Zhang, J.T. and Dai, L.M. (2015). ACS Catal. 5: 7244.
  43. 43 Shen, A.L., Zou, Y.Q., Wang, Q. et al. (2014). Angew. Chem. Int. Ed. 53: 10804.
  44. 44 Su, D.S., Zhang, J., Frank, B. et al. (2010). ChemSusChem 3: 169.
  45. 45 Wang, S.Y., Yu, D.S., and Dai, L.M. (2011). JACS 133: 5182.
  46. 46 Conway, B.E. and Tilak, B.V. (2002). Electrochim. Acta 47: 3571.
  47. 47 Chen, Z.B., Cummins, D., Reinecke, B.N. et al. (2011). Nano Lett. 11: 4168.
  48. 48 (a) Lewis, N.S. and Nocera, D.G. (2006). Proc. Natl. Acad. Sci. U.S.A. 103: 15729.(b) Yeo, B.S. and Bell, A.T. (2011). J. Am. Chem. Soc. 133: 5587.(c) Wang, H.Y., Hung, S.F., Chen, H.Y. et al. (2016). J. Am. Chem. Soc. 138: 36.
  49. 49 Koper, M.T.M. (2011). J. Electroanal. Chem. 660: 254.
  50. 50 (a) Lee, Y., Suntivich, J., May, K.J. et al. (2012). J. Phys. Chem. Lett. 3: 399.(b) Lyons, M.E.G. and Floquet, S. (2011). Phys. Chem. Chem. Phys. 13: 5314.(c) Reier, T., Oezaslan, M., and Strasser, P. (2012). ACS Catal. 2: 1765.
  51. 51 Zhao, Y., Nakamura, R., Kamiya, K. et al. (2013). Nat. Commun. 4: 2390.
  52. 52 Li, L.Q., Yang, H.B., Miao, J.W. et al. (2017). ACS Energy Lett. 2: 294.
  53. 53 (a) Chen, S., Duan, J.J., Jaroniec, M., and Qiao, S.Z. (2014). Adv. Mater. 26: 2925.(b) Silva, R., Voiry, D., Chhowalla, M., and Asefa, T. (2013). J. Am. Chem. Soc. 135: 7823.
  54. 54 Dai, L.M. (2013). Acc. Chem. Res. 46: 31.
  55. 55 Zhang, J.T., Zhao, Z.H., Xia, Z.H., and Dai, L.M. (2015). Nat. Nanotechnol. 10: 444.
  56. 56 (a) Katsounaros, I., Cherevko, S., Zeradjanin, A.R., and Mayrhofer, K.J.J. (2014). Angew. Chem. Int. Ed. 53: 102.(b) Fujishima, A. and Honda, K. (1972). Nature 238: 37.(c) Lu, Y.C., Xu, Z.C., Gasteiger, H.A. et al. (2010). J. Am. Chem. Soc. 132: 12170.
  57. 57 Ibrahim, H., Ilinca, A., and Perron, J. (2008). Renewable Sustainable Energy Rev. 12: 1221.
  58. 58 Goodenough, J.B. and Kim, Y. (2010). Chem. Mater. 22: 587.
  59. 59 (a) Girishkumar, G., McCloskey, B., Luntz, A.C. et al. (2010). J. Phys. Chem. Lett. 1: 2193.(b) Lee, J.S., Kim, S.T., Cao, R. et al. (2011). Adv. Energy Mater. 1: 34.(c) Li, Y.G. and Dai, H.J. (2014). Chem. Soc. Rev. 43: 5257.
  60. 60 Norskov, J.K., Rossmeisl, J., Logadottir, A. et al. (2004). J. Phys. Chem. B 108: 17886.
  61. 61 Man, I.C., Su, H.Y., Calle‐Vallejo, F. et al. (2011). ChemCatChem 3: 1159.
  62. 62 Karl, T.R. and Trenberth, K.E. (2003). Science 302: 1719.
  63. 63 Davis, S.J., Caldeira, K., and Matthews, H.D. (2010). Science 329: 1330.
  64. 64 Rosen, B.A., Salehi‐Khojin, A., Thorson, M.R. et al. (2011). Science 334: 643.
  65. 65 Wang, W., Wang, S.P., Ma, X.B., and Gong, J.L. (2011). Chem. Soc. Rev. 40: 3703.
  66. 66 Kortlever, R., Shen, J., Schouten, K.J.P. et al. (2015). J. Phys. Chem. Lett. 6: 4073.
  67. 67 (a) Kuhl, K.P., Cave, E.R., Abram, D.N., and Jaramillo, T.F. (2012). Energy Environ. Sci. 5: 7050.(b) Li, C.W. and Kanan, M.W. (2012). J. Am. Chem. Soc. 134: 7231.
  68. 68 (a) Reske, R., Mistry, H., Behafarid, F. et al. (2014). J. Am. Chem. Soc. 136: 6978.(b) Hori, Y., Murata, A., and Takahashi, R. (1989). J. Chem. Soc. Faraday Trans. 1 85: 2309.
  69. 69 Kumar, B., Asadi, M., Pisasale, D. et al. (2013). Nat. Commun. 4: 2819.
  70. 70 Wu, J.J., Ma, S.C., Sun, J. et al. (2016). Nat. Commun. 7: 13869.
  71. 71 Zhang, S., Kang, P., Ubnoske, S. et al. (2014). J. Am. Chem. Soc. 136: 7845.
  72. 72 (a) Morris, A.J., McGibbon, R.T., and Bocarsly, A.B. (2011). ChemSusChem 4: 191.(b) Tornow, C.E., Thorson, M.R., Ma, S. et al. (2012). J. Am. Chem. Soc. 134: 19520.
  73. 73 Cole, E.B., Lakkaraju, P.S., Rampulla, D.M. et al. (2010). J. Am. Chem. Soc. 132: 11539.
  74. 74 Blanchard, L.A., Gu, Z.Y., and Brennecke, J.F. (2001). J. Phys. Chem. B 105: 2437.
  75. 75 Sanz, R., Calleja, G., Arencibia, A., and Sanz‐Perez, E.S. (2013). J. Mater. Chem. A 1: 1956.
  76. 76 Mamtani, K., Jain, D., Zemlyanov, D. et al. (2016). ACS Catal. 6: 7249.
  77. 77 Geng, D.S., Chen, Y., Chen, Y.G. et al. (2011). Energy Environ. Sci. 4: 760.
  78. 78 Bae, I.T. and Scherson, D.A. (1998). J. Phys. Chem. B 102: 2519.
  79. 79 von Deak, D., Singh, D., Biddinger, E.J. et al. (2012). J. Catal. 285: 145.
  80. 80 von Deak, D., Singh, D., King, J.C., and Ozkan, U.S. (2012). Appl. Catal., B 113: 126.
..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.
Reset