Single atom electrocatalysts supported on graphene or graphene-like carbons

Huilong Fei a, Juncai Dong b, Dongliang Chen b, Tiandou Hu b, Xidong Duan a, Imran Shakir *cd, Yu Huang *c and Xiangfeng Duan *e
aState Key Laboratory for Chemo/Biosensing and Chemometrics, College of Chemistry and Chemical Engineering, Hunan University, Changsha 410082, China
bBeijing Synchrotron Radiation Facility, Institute of High Energy Physics, Chinese Academy of Sciences, Beijing 100049, China
cSustainable Energy Technologies Centre, College of Engineering, King Saud University, Riyadh 11451, Saudi Arabia. E-mail:;
dDepartment of Materials Science and Engineering, University of California, Los Angeles, Los Angeles, CA 90095, USA
eDepartment of Chemistry and Biochemistry, University of California, Los Angeles, Los Angeles, CA 90095, USA. E-mail:

Received 11th June 2019

First published on 1st October 2019

Electrocatalysis plays an essential role in diverse electrochemical energy conversion processes that are vital for improving energy utilization efficiency and mitigating the aggravating global warming challenge. The noble metals such as platinum are generally the most frequently used electrocatalysts to drive these reactions and facilitate the relevant energy conversion processes. The high cost and scarcity of these materials pose a serious challenge for the wide-spread adoption and the sustainability of these technologies in the long run, which have motivated considerable efforts in searching for alternative electrocatalysts with reduced loading of precious metals or based entirely on earth-abundant metals. Of particular interest are graphene-supported single atom catalysts (G-SACs) that integrate the merits of heterogeneous catalysts and homogeneous catalysts, such as high activity, selectivity, stability, maximized atom utilization efficiency and easy separation from reactants/products. The graphene support features a large surface area, high conductivity and excellent (electro)-chemical stability, making it a highly attractive substrate for supporting single atom electrocatalysts for various electrochemical energy conversion processes. In this review, we highlight the recent advancements in G-SACs for electrochemical energy conversion, from the synthetic strategies and identification of the atomistic structure to electrocatalytic applications in a variety of reactions, and finally conclude with a brief prospect on future challenges and opportunities.

image file: c9cs00422j-p1.tif

Huilong Fei

Huilong Fei is a Professor at the College of Chemistry and Chemical Engineering, Hunan University, China. He received his BS degree in materials chemistry from China University of Geosciences (Wuhan) in 2011, and his PhD degree in chemistry from Rice University in 2015. He was previously a postdoc researcher in the University of California, Los Angeles, before joining Hunan University in 2018. His current research interests focus on single atom catalysts and their electrocatalytic applications.

image file: c9cs00422j-p2.tif

Juncai Dong

Juncai Dong received his PhD degree in Condensed Matter physics in 2014 from the University of Chinese Academy of Sciences. He is currently a research associate at Beijing Synchrotron Radiation Facility, Institute of High Energy Physics, Chinese Academy of Sciences. His research interests are focused on the structural characterization of various electrocatalytic materials using synchrotron-based X-ray spectroscopy.

image file: c9cs00422j-p3.tif

Tiandou Hu

Tiandou Hu is a senior scientist and director in Beijing Synchrotron Radiation Facility (BSRF) at Institute of High Energy Physics, Chinese Academy of Sciences. Over the past 30 years, he has been closely engaged in the development of synchrotron-based techniques and beamline design and construction at BSRF and High Energy Photon Source (HEPS). His current research interests are focused on the exploration of novel experimental and analytical techniques for X-ray absorption spectroscopy with particular applications in high-pressure condensed matter physics, catalytic chemistry and nanostructured materials.

image file: c9cs00422j-p4.tif

Xidong Duan

Xidong Duan is a Professor at the College of Chemistry and Chemical Engineering, Hunan University, China. His current research interests include two dimensional materials and heterostructures. He received his BS degree in Chemistry from Hunan University in 1993, his MA degree in Materials from Hunan University in 1996, and his PhD degree in Chemistry from Hunan University in 2016. He was previously a senior engineer at the Changsha Research Institute of Mining and Metallurgy before joining Hunan University.

image file: c9cs00422j-p5.tif

Imran Shakir

Imran Shakir is currently a visiting Professor in the Department of Materials Science and Engineering at UCLA and Professor at the Sustainable Energy Technologies (SET) center, King Saud University. He received his BS degree in Physics from the Punjab University of Pakistan in 2000, and his PhD degree in Physics from Sungkyunkwan University, Korea, in 2011. His current research is mainly focused on the development of functional nanomaterials and their composites with a 3D graphene framework for application in electronics, energy storage and conversion devices.

image file: c9cs00422j-p6.tif

Yu Huang

Yu Huang is currently a Professor in the Department of Materials Science and Engineering at UCLA. She received her BS degree in Chemistry from the University of Science and Technology of China in 1999, and her PhD degree in Physical Chemistry from Harvard University in 2003. Her current research interests focus on chemical and biological approaches to producing complex material structures with atomic precision and the new physical properties arising from these new materials.

image file: c9cs00422j-p7.tif

Xiangfeng Duan

Xiangfeng Duan is a Professor of Chemistry and Nanoscience at UCLA. He received his BS degree from the University of Science and Technology of China in 1997, and his PhD degree from Harvard University in 2002. He was previously a Founding Scientist and then Manager of Advanced Technology at Nanosys Inc., a nanotechnology startup founded based partly on his doctoral research. His research interests include nanoscale materials and devices and their applications in future electronic, energy and health technologies.

1. Introduction

Electrochemical energy conversion plays an essential role in addressing the global challenge in climate change and energy crisis. Consequently, extensive research efforts have been devoted to advanced electrochemical energy conversion technologies, such as water splitting, CO2 conversion, N2 reduction, fuel cells and metal–air batteries, that are critical for renewable energy harvesting, storage and efficient energy utilization.1 The performances of these devices strongly rely on the properties of the materials used in each component, particularly the catalysts employed at the electrode/electrolyte interface to decrease the kinetic barrier of the associated electrochemical processes for improving the energy conversion efficiency.2 As a result, considerable research efforts have been devoted to the development of highly active, selective, robust and cost-effective electrocatalysts.

In this regard, homogeneous and heterogeneous catalysis were developed independently as two parallel frontiers, each with its own merits and limits.3 Homogeneous catalysis is characterized by maximized atom utility efficiency, well-defined catalytic centers and coordination configuration, which allows a rational design and synthesis of high-performance catalysts based on the established structure–property correlation. However, in practical applications, homogeneous catalysis is often limited by low stability of the catalyst and difficulty in its separation from products and solvents. Although strategies have been explored to address these challenges by immobilizing the homogeneous catalysts on solid supports, none of them have thus far found use in industry mainly due to the insufficient stability of the immobilized catalysts and the extra cost imposed by the immobilization process.4 On the other hand, heterogeneous catalysts, particularly those based on nanostructured materials, have attracted increasing interest for their high stability, facile separation and recycling, and easy immobilization on electrodes. However, most heterogeneous catalysts feature a broad distribution of particle size and surface characteristics, posing serious challenges in fundamental mechanistic studies and the rational design of active sites with tailored activity and selectivity.5 Additionally, the heterogeneous catalysts typically show a limited atom utilization efficiency due to the limited surface active sites as most of the atoms are embedded in the bulk. To this end, significant research efforts have been devoted to downsizing the particle size and engineering the surface morphology to achieve more surface sites, which is particularly important in precious metal-based catalysts where cost and scarcity are of serious concern.6

Single atom catalysts (SACs) with atomically dispersed metals on a solid support represent the ultimate limit of nanoscale catalysts that bridge the gap between heterogeneous catalysts and homogeneous catalysts to deliver the combined merits of both.7–10 Due to the high surface free energy of single atoms, SACs are typically supported on substrates with high surface area and abundant anchoring sites to prevent aggregation during synthesis and catalytic processes.11,12 Typical substrates include metal surface, metal oxide, metal nitride, metal chalcogenide and nanocarbons, and SACs on these substrates have been applied to catalyze a wide range of chemical reactions, such as oxidation reactions, hydrogenation reactions, and electrocatalytic reactions.13–19 With well-dispersed SACs on a solid support, the metal–support interactions and coordination environment may be systematically varied. In this regard, the supporting substrate plays a critical role in tailoring the catalytic activity, selectivity and stability of SACs.

Graphene, a two-dimensional material consisting of a single layer of sp2-hybridized carbons arranged in a hexagonal lattice, has extraordinary physicochemical properties and it has found applications in a broad range of areas, such as electronics, sensing, polymer composites, energy storage and conversion.20 Graphene exhibits several unique features to make it an attractive supporting substrate for various nanostructured catalysts (such as nanoplates, nanoparticles or nanoclusters, nanowires)21–24 or SACs (Fig. 1). First, the atomically thin two-dimensional (2D) nature of graphene features an ultrahigh theoretical surface area (2630 m2 g−1) which is important for the uniform dispersion of metal atoms, particularly at high mass loading, and for the maximized exposure of the catalytic centers to the reactants. Moreover, the highly crystalline nature of the ultrathin graphene substrate provides an excellent platform for atomistic imaging of the exact coordination configuration of the metal centers by advanced characterization techniques, which is important for the fundamental understanding of the catalytic behavior for formulating design principles for future generations of catalysts.25 Second, graphene exhibits excellent electrical conductivity for efficient transport of electrons to or from the catalytic sites, which is particularly important for electrocatalysis involving long-range charge transport. In comparison, SACs on metal oxide-based supports typically have rather low electrical conductivity and consequently their application areas are mainly limited to the conventional heterogeneous reactions (e.g., CO oxidation, water–gas shift reactions, oxidation and hydrogenation of organic molecules) where long-range electron transport is not a necessity.13 Third, graphene exhibits extraordinary chemical and electrochemical stability over wide potential windows, and can withstand the harsh liquid electrolyte environment and large potential range that are typically encountered in electrocatalytic processes. For example, in order to achieve efficient and stable operation of water splitting, a strong acid or base electrolyte is needed to increase the ionic conductivity and minimize the pH gradient due to proton consumption and production at the cathode and anode, respectively.26 Such a requirement limits the choice of stable catalysts and excludes most of SACs supported on metals or metal oxides due to the insufficient stability in such highly corrosive media. Lastly, defect engineering and chemical doping in graphene offers an additional degree of freedom to tailor the anchoring sites for the dispersion of SACs, which provides a rich playground for exploring synthetic strategies and tuning metal–support interactions.27 In particular, pristine graphene sheets are not good supports for SACs since the adatoms on the dangling-bond free graphene sheets are highly mobile and could easily aggregate due to their low migration energy barrier.28 The calculation results suggested that the transition metal adatoms on pristine graphene had binding energies of 0.2–1.5 eV and migration barriers of 0.2–0.8 eV, indicating that the adatoms should be mobile at room temperature. To this end, the generation of defective sites in graphene provides various possible bonding configurations for metal atoms, such as single and double vacancies.29 The metal atoms can be anchored by the defective sites of graphene as a substitutional or interstitial dopant and form metal-vacancy complexes with much higher binding energies that ensure its stability, as has been demonstrated by both experiments and calculations.28–31 Besides, a wide range of heteroatoms (e.g., N, B, O, S) can be incorporated within the graphene lattice to serve as potential anchoring sites for metal atoms with diverse coordination geometries, providing a new dimension to tune the interaction between metal atoms and the coordination environment.32

image file: c9cs00422j-f1.tif
Fig. 1 Graphene as a support for nanostructured catalysts. Schematic illustration of graphene supports for different types of nanostructures, including nanoplates, nanowires, nanoparticles and single atomic metals. Image of nanoplates on graphene: reproduced with permission from Wang et al.21 Copyright 2010 American Chemical Society. Image of nanowires on graphene: reproduced with permission from Hu et al.22 Copyright 2017 Wiley-VCH. Image of nanoparticles on graphene: reproduced with permission from Kundu et al.23 Copyright 2011 American Chemical Society. Image of single atomic metals on graphene: reproduced with permission from Deng et al.33 Copyright 2015 American Association for the Advancement of Science.

Due to its intriguing physicochemical properties and versatility in chemical functionalization, the exploration of graphene as a support for SACs has attracted intense research interest and endeavors in the past few years. This review provides an overview of the recent progress in graphene-supported SACs (G-SACs) for electrochemical energy conversion processes. We will begin with recent theoretical and experimental efforts in establishing the correlation between the structure and catalytic properties of G-SACs, elaborating how the catalytic behaviors could be tuned or influenced by tailoring the electronic interaction between metal atoms and the graphene support. We will then summarize the various synthetic strategies developed to access G-SACs, the recent advancements in the identification of the atomistic structure and their applications in different types of electrochemical energy conversion processes, including oxygen reduction reaction (ORR), hydrogen evolution reaction (HER), oxygen evolution reaction (OER), CO2 reduction reaction (CO2RR), N2 reduction reaction (NRR) and so on. Finally, we briefly discuss the challenges and opportunities ahead.

2. Understanding structure–activity correlation

The activity of an electrocatalyst system can be generally improved by either increasing the number of active sites (e.g., by increasing the loading or by minimizing the size to increase the exposed active sites) or enhancing the intrinsic activity of each active site (Fig. 2).2 For catalysts based on transition metals, it has long been established that the intrinsic activity is highly dependent on the strength of the adsorbate–surface interaction and hence on the surface electronic structure (the d-band center).34 With respect to traditional heterogeneous catalysts of metal surfaces, the regulation of their electronic structure and hence catalytic activity has been realized by alloying, sizing, defect engineering, strain engineering, interface engineering, geometry optimization, etc.35–37 In the case of homogeneous catalysts with organometallic compounds, the variability in their reactivity can be generally achieved by tuning the ligand-field effect, which is governed by the central metal center as well as its chemical and structural coordination environment.38 Considering that G-SACs can be regarded as the mimics of organometallic complexes by having the metal center and neighboring coordinators accommodated in the π-conjugated graphitic carbon, the concept of electronic modulation by ligand-field effects can be transferred to some degree from the familiar paradigm of homogeneous catalysis to G-SACs. In this regard, the intrinsic activity of G-SACs can be regulated by various factors, including the metal identity, the coordination configuration of the metal center, the Lewis acidity/basicity (electron-donating/withdrawing capability) of the graphene substrate and the metal location (edge site versus basal plane site) (Fig. 2). A fundamental theoretical understanding of how these factors influence the catalytic activity along with experimental validation is central for defining the basis for the rational design of new generations of high-performance G-SACs.
image file: c9cs00422j-f2.tif
Fig. 2 Overview of factors affecting the intrinsic activity and active sites number in G-SACs.

In recent years, numerous theoretical simulations have been performed on G-SACs to explore their potential applications in electrocatalysis with a focus on establishing the correlation between the local atomic structure and the catalytic properties. Choi et al. performed a high-throughput computational search for hydrogen electrocatalysts based on porphyrin-like complexes embedded in graphene lattices (Fig. 3A).27 The regulation of the electronic structure of these motifs was achieved by varying the metal center (30 d-block transition metals), the coordination atoms (B, C, N, O, S) coordinated to the metal center, and the coordination geometry (threefold and fourfold coordination geometries at the mono-vacancy and di-vacancy sites, respectively). The consideration of these variables produced a large candidate pool (300 combinations), among which 180 combinations passed the complex stability screening test while the others would form metal aggregates. The stability screening test further suggested that C and N are the best coordination atoms and they predominantly adopt threefold (3C) and fourfold (4N) configurations, respectively. The catalytic activity of the selected combinations was then evaluated with the descriptor based on the free energy of hydrogen adsorption on the metal centers. The screening results revealed that the majority of the active and stable catalysts were related to the 3C configuration (Fig. 3B), attributed to the free orbitals of the threefold-coordinated metal center that can interact more strongly with the hydrogen 1s orbital. In another study, Xu et al. developed a theory-based universal design principle to evaluate the catalytic activity of G-SACs toward a variety of reactions, including ORR, OER and HER.39 A total of 112 G-SACs models were constructed on various transition metal atoms embedded in graphene with four types of configurations (Fig. 3C), including a mono-vacancy with three carbon atoms (SV-C3), a di-vacancy with four carbon atoms (DV-C4), four pyridine nitrogen atoms (pyridine-N4), and four pyrrole nitrogen atoms (pyrrole-N4). A complex descriptor, associated with the electronegativity of the metal atom, coordination atom type and coordination number, was identified to intimately correlate with intermediate adsorption strength and thus catalytic activities, and was used to predict highly active G-SACs with specific active centers (e.g., Fe-pyridine/pyrrole-N4 for ORR, Co-pyrrole-N4 for OER and Mn-pyrrole-N4 for HER). Similar theoretical efforts have been made to understand the influence of local chemical and structural environments of the metal centers on their catalytic activity and selectivity toward CO2RR, CO reduction, and NRR.40–42

image file: c9cs00422j-f3.tif
Fig. 3 Theoretical studies on the correlation between the structure and intrinsic catalytic activity of G-SACs. (A) Illustration of the possible transition metal-embedded graphene complexes forming the initial material pool for the search of hydrogen electrocatalysts. (B) Descriptor-based screening results of G-SACs based on activity and stability. The lower-left quadrant contains materials that passed both criteria. Reproduced with permission from Choi et al.27 Copyright 2015 Wiley-VCH. (C) Schematic of a transition metal atom embedded in graphene with different coordination environments. From left to right: mono-vacancy with three carbon atoms (SV-C3), di-vacancy with four carbon atoms (DV-C4), four pyridine nitrogen atoms (pyridine-N4), and four pyrrole nitrogen atoms (pyrrole-N4). Reproduced with permission from Xu et al.39 Copyright 2018 Nature Publishing Group.

Despite considerable progress in theoretical studies, experimentally establishing the exact structure–property correlation in G-SACs remains challenging mainly due to the difficulties in accessing G-SACs with a well-defined and controllable atomistic structure and the technical difficulties in unambiguously identifying the exact atomic structure of G-SACs. Unlike homogeneous catalysts where precise synthetic chemistry can guide the synthesis and modification of active sites, current progress in the synthesis of G-SACs primarily involves a trial-and-error approach. For instance, high-temperature pyrolysis of mixed precursors represents the most frequently used approach to metal–nitrogen–carbon (M–N–C) catalysts with atomic MNx active sites, which however usually leads to the formation of atomic metal sites mixed with metal aggregates.43 Furthermore, diverse coordination configurations could be simultaneously formed in a given M–N–C catalyst.44 These structural and compositional heterogeneities severely complicate the identification of the active site and further the correlation between the structure and catalytic properties. Nevertheless, a few experimental efforts have recently been made aiming at establishing such correlations. For example, our group recently developed a general approach to a series of atomic transition metals (Fe, Co, Ni) embedded in nitrogen-doped graphene with a common MN4C4 moiety, which can be described as the metal atom occupying the di-vacancy in the graphene lattice with coordination to four pyridinic N (Fig. 4A).25 The well-defined atomistic structure and the systematically tunable metal center in the different catalysts provided an ideal model system to correlate the specific metal center with the catalytic property. Density functional theory (DFT) calculations and electrochemical measurements suggested that the catalytic activity toward OER was strongly dependent on the metal center with the activity following the trend Ni > Co > Fe. In another example, Wang et al. synthesized a series of graphene-supported single atomic Co with varied numbers of N coordinators, as identified by EXAFS spectra (Fig. 4B and C). Experimental and theoretical explorations on applying these materials as CO2RR catalysts suggested that the decrease in the number of coordinating N atoms would result in more unoccupied 3d orbitals of Co atoms, resulting in an increase of the absorption of reaction intermediates and thus the catalytic activity.45 Moreover, the modification of the coordination configuration of the metal center in G-SACs by secondary binding sites (e.g., boron, sulfur, chlorine, carbon monoxide) in addition to the nitrogen sites has recently been employed to modulate their electronic structure and catalytic behaviors.46–50

image file: c9cs00422j-f4.tif
Fig. 4 Representative experimental studies on the correlation between the structure and intrinsic catalytic activity of G-SACs. (A) Illustration of the trend for the catalytic activity toward OER for MN4C4 moieties supported in graphene lattices. Adapted with permission from Fei et al.25 Copyright 2018 Nature Publishing Group. (B) Schematic illustration of the dependence of CO2RR catalytic behavior on coordination configurations. (C) EXAFS spectra for different atomic catalysts. Adapted with permission from Wang et al.45 Copyright 2018 Wiley-VCH. (D) Model structures of the bulk-hosted and zigzag edge-hosted FeN4. Adapted with permission from Chung et al.51 Copyright 2017 American Association for the Advancement of Science.

Considering that the graphene substrate constitutes an integral part of the active site with covalent bonding rather than merely serving as an electrical conductor in G-SACs, its properties such as Lewis acidity/basicity and the location of the active sites in the substrate could also play important roles in determining the intrinsic activity of G-SACs. Ramaswamy et al. elucidated the relationship between the ORR catalytic activity of atomic Fe–Nx moieties and the basicity of the π-conjugated carbon support in Fe–N–C catalysts.52 It was suggested that the electron-donating or withdrawing capability of the graphene support resulting from its delocalized π-electrons could regulate the d-orbital configuration of the metal center and consequently the chemisorption strength of O2 and reaction intermediates on the active site. In another study, Zitolo et al. measured the pKa of the surface nitrogen groups in their Fe–N–C catalysts and found that the higher ORR activity of the NH3-annealed catalyst correlated well with their higher basicity compared to that of the non-annealed counterpart.43 Chung et al. obtained the atomic-scale images of the ORR-active FeN4 moiety in Fe–N–C catalysts by aberration-corrected scanning transmission electron microscopy (AC-STEM) and confirmed such structure by electron energy-loss spectroscopy (EELS) with single-atom resolution.51 Based on the quantum chemical calculation results and the observation that the dispersed Fe single atoms were preferably located at the surfaces of the graphitic domains or step edges in multilayer graphene, it was suggested that the edge-hosted FeN4 sites rather than bulk-hosted FeN4 sites primarily contribute to the observed ORR activity (Fig. 4D). The enhanced catalytic contribution of edge-hosted MNx moieties was confirmed by later studies that synthesized G-SACs with a high fraction of the edge-hosted sites by porosity engineering in the graphene matrix.53–55

3. Synthetic strategies

A prerequisite for studying G-SACs is to stabilize the single atomic metals onto the support of graphene layers. The preparation of G-SACs is challenging due to the mobility of the metal atoms and their tendency to agglomerate during synthesis and post-treatments. Pristine graphene is not a good support for G-SACs since most metal adatoms usually exhibit a rather low migration barrier and high mobility on pristine graphene even at room temperature.29 Therefore, the generation of defective sites or functionalization of graphene is generally required to create anchoring sites to stabilize metal atoms without the formation of metal aggregates. We summarize below the synthetic strategies toward G-SACs.

3.1. Pyrolysis and thermal activation

The pyrolysis and thermal activation of precursors at high temperature remain the most widely adopted method to obtain single atomic metals supported on graphene or graphene-like nanocarbons with small domains of graphene surrounding the metal centers. The pyrolysis method is simple, scalable and does not require sophisticated instrumentation, and therefore can be readily carried out across different laboratories. Furthermore, there are versatile precursors to choose from for the optimization of metal doping and nanostructure engineering that are critical in determining the catalytic activity. The precursors used for pyrolysis can be generally divided into three different categories: molecular/polymeric precursors, graphene oxide (GO) and metal organic frameworks (MOFs).
3.1.1. Molecular/polymeric precursors. Transition metal macrocycles, such as Fe and Co porphyrins or phthalocyanines, were the first studied precursors for synthesizing M–N–C catalysts through pyrolysis at high temperature (typically above 700 °C).43 It was later shown that the replacement of these expensive macrocycles was possible by pyrolyzing separate metal-, nitrogen- and carbon-containing precursors and since then different combinations of metal salts and N/C-containing precursors have been investigated, such as imidazole, pyrrole, melamine, formamide, polyaniline and polydopamine.56–61 For example, Zhao et al. developed a cascade anchoring strategy to synthesize a wide range of carbon-supported single atomic metals (e.g., Fe, Co, Cu, Mn, Ni, Mo) with metal loading up to 12 wt%.62 During the synthesis, the metal ions were firstly chelated with glucose and further anchored onto oxygen-functionalized porous carbon to isolate the metal ions (Fig. 5A). The complex was then mixed with melamine as the nitrogen source, which would release carbon nitrogen species (CNx) during high-temperature pyrolysis to bind with metal atoms to form MNx moieties. The resulting composite exhibited a uniform distribution of Fe and N elements (Fig. 5B). With systematic experimental comparisons, it was found that the chelation of metal ions, physical isolation of the chelate complex, and the binding with N-species at high temperature were critical in achieving atomic metal dispersion at high loading.
image file: c9cs00422j-f5.tif
Fig. 5 (A) The cascade anchoring strategy for the synthesis of carbon-supported single atomic metals. First, glucose as a chelating agent sequesters metal ions and binds to the O-rich carbon support. Second, the chelated metal complexes secure metal atoms via the decomposed residues at elevated temperature. Third, CNx species decomposed from melamine at higher temperature capture metal atoms to generate M–Nx moieties incorporated in the pyrolyzed carbon substrate. (B) HAADF-STEM image and EELS mapping images of Fe, N and overlaid Fe and N. Scale bar, 2 nm. Reproduced with permission from Zhao et al.62 Copyright 2019 Nature Publishing Group. (C) The preparation route to single atomic metals on nitrogen-doped graphene by a two-step mixing and annealing process with GO as the precursor. Reproduced with permission from Zhang et al.66 Copyright 2017 American Chemical Society.

The choice of precursors is not only important for the effective implantation of atomic M–Nx species into carbon but also for the structure and porosity of the carbon support. For example, Chung et al. deliberately combined precursors of polyaniline and cyanamide to synthesize a catalyst that exhibited both porous structure and high intrinsic catalytic activity.51 Precursors containing multiple heteroatoms (e.g., sulfur, boron, phosphorus) other than nitrogen could be included to introduce multiple types of dopants to modify the local coordination configuration of the metal center or tune the electronic properties of the carbon support, which could both be applied to regulate the catalytic performances.46,48,63–65

3.1.2. GO as precursors. GO is an oxidized form of graphene and it is typically prepared by the oxidation of graphite followed by solution-phase exfoliation.67 Because of the solution-based processing of GO, it has desirable manufacturability for scalable production and therefore has become one of the most popular starting materials towards graphene-based nanomaterials in a wide range of application areas, especially those where mass-production is required such as energy storage, catalysis, composite materials and coating.24,68 The surface of graphene oxide is enriched with a high density of oxygen functional groups, including carboxyl, hydroxyl, carbonyl and epoxy, and these functional groups provide handful opportunities for surface modification and functionalization. To date, research on GO has mainly focused on the restoration of GO by chemical or thermal reduction to pristine graphene alternatives (reduced graphene oxide, rGO) for applications in electronics, electrocatalysis, and energy storage.69–71 Recently, the intrinsic defects in rGO resulting from the removal of carbon-containing by-products such as CO and CO2 gas during the reduction process have been used to stabilize single atomic metals.25 The synthetic procedure typically involved simple two-step treatments (Fig. 5C).66 The first step was the mixing of metal precursor and GO in aqueous solution with the negatively charged functional groups on GO providing anchoring sites for the metal ions. After being freeze-dried to prevent the restacking of GO nanosheets, the mixture was annealed in a NH3 atmosphere at high temperature, during which GO was reduced to rGO and nitrogen dopants were introduced into the carbon lattices as the effective binding sites for the individual metal atoms. It is critical to precisely control the amount of metal precursor added in order to achieve maximized activity while preventing metal aggregation. Unlike other carbon precursors, the reduction of GO to highly conductive rGO does not require the catalyzing effects of metals for the formation of conductive graphitized carbon and therefore excessive addition of metal precursors can be avoided, which is beneficial for the achievement of exclusive or predominant atomic metal dispersion. This method has been employed to prepare a variety of atomic metals (e.g., Fe, Co, Ni, Ru, Mn, Sc) dispersed in nitrogen-doped graphene.25,66,72–75

However, the preparation of G-SACs by thermally annealing the mixture of GO and metal precursors typically involves prolonged heating treatment, which is not only time- and energy-consuming, but also could cause aggregation of metal atoms. To this end, our group recently developed a facile and rapid strategy to prepare a series of monodispersed atomic transition metals (e.g., Co, Ni, Cu) by microwave irradiation of the mixture of amine-functionalized GO and metal precursors in a conventional microwave oven.76 The microwave treatment can simultaneously induce the reduction of GO, the doping of nitrogen and the incorporation of metal atoms into graphene lattices. In addition, the microwave treatment is so rapid (∼2 s) that there is less chance for metal aggregation to occur, resulting in the predominant presence of atomic species.

3.1.3. MOFs as precursors. With rich structural diversity, high porosity, tailorable composition and versatile functionality, MOFs have been widely used as precursors to create nanostructured materials (e.g., porous nanocarbons, metal carbides, nitrides, phosphides and oxides) by high-temperature pyrolysis.77 Compared to other approaches to nanostructured materials, the pyrolysis of MOFs offers controllability of size, shape, composition and porosity and allows the integration of diverse functionalities in one step.78,79 Recently, the pyrolysis of MOFs has been demonstrated as an effective method to access G-SACs, and careful synthetic design is critical to ensure uniform distribution and spatial separation of the metal sites in order to prevent their aggregation during pyrolysis.55,80–90 For example, Yin et al. achieved the atomic dispersion of Co atoms on nitrogen-doped porous carbon by pyrolyzing Zn/Co bimetallic MOFs with a homogeneous distribution of Zn and Co in the framework.89 The addition of Zn2+ as metal nodes in the MOFs serves as a spacer to dilute Co2+ and it can be evaporated away at high temperature due to its low boiling point. By carefully controlling the molar ratio of Zn and Co, catalysts with Co predominantly in the form of single atoms could be achieved. Motivated by these results, Wu et al. synthesized nitrogen-coordinated single Mn sites supported on graphitic carbon through the pyrolysis of Mn-doped ZIF-8 precursors.91 Unlike the case of Co2+, Mn2+ cannot easily exchange with Zn2+ in ZIF-8 and therefore only a limited number of atomic Mn sites were introduced into the carbon matrix by the one-step pyrolysis treatment. In order to increase the Mn loading, the pyrolyzed product with a high degree of microporosity was further used as the host material to absorb additional metal and nitrogen sources, followed by secondary thermal activation (Fig. 6A). With such a two-step doping and adsorption strategy, the atomic Mn loading was significantly increased, resulting in excellent catalytic activity.
image file: c9cs00422j-f6.tif
Fig. 6 (A) Schematic of the synthesis of atomically dispersed MnN4 sites. In the first step, Mn-doped ZIF-8 was carbonized and leached with acid to prepare a partially graphitized carbon host with optimal nitrogen doping and microporous structures. In the second step, additional Mn and N sources were adsorbed into the carbon host by thermal activation to generate an increased density of MnN4 active sites. Reproduced with permission from Li et al.91 Copyright 2018 Nature Publishing Group. (B) Scheme of the transformation of nanoparticles to single atoms and structural characterization of Pd single atoms. (C) Energy dispersive X-ray spectroscopy elemental mapping results of Pd, N and C, showing their uniform distribution. (D) HAADF-STEM image of Pd single atoms. (E) Fourier transforms of k3-weighted Pd K-edge EXAFS experimental data for Pd nanoparticles and Pd single atoms, showing the coordinating atoms adjacent to Pd atoms before and after pyrolysis. Reproduced with permission from Wei et al.86 Copyright 2018 Nature Publishing Group.

The preparation of atomically dispersed noble metals (e.g., Ru, Pt, Au) by the strategy of converting the atomically dispersed metal nodes in situ from MOFs was hindered as these metals were rarely used as the metal nodes of MOFs. To circumvent this limitation, Wang et al. utilized the uncoordinated –NH2 group on the skeletons of the MOF structure as the Lewis base to immobilize the Ru precursor. Benefiting from the strong interaction between the lone pair electrons and the d-orbital of Ru ions, single Ru sites were prepared without obvious aggregation during high-temperature pyrolysis, while Ru clusters were found in the sample without the assistance of –NH2 groups.92 Zhao et al. found that the metal precursor need not be pre-mixed within the MOF. Rather, they developed a gas-migration strategy to access a series of metal atoms (Cu, Co, Ni) via direct thermal emission from bulk metals and the subsequent trapping on MOF-derived nitrogen-rich carbon with the assistance of NH3.84 This gas-migration strategy is general in producing different types of atomic metals and easy to scale up as it avoids the complex operating steps and sensitive conditions encountered in the conventional method. With a similar strategy, Sun and coworkers reported a metallorganic gaseous doping method to synthesize single Fe atoms by evaporating the organic iron compound ferrocene and trapping into ZIF-8 precursors placed downstream during thermal activation.93 Besides MOF-derived nanocarbon, graphene oxide was used as the trapping agent to anchor atomic Pt via the thermal emission method.94

Taking a step further, Yang et al. reported that Ni nanoparticles within MOF-derived N-doped carbon can be transformed into Ni single atoms through a thermal diffusion mechanism, during which the Ni nanoparticles could break C–C bonds to generate pores on the surface and at the same time the Ni atoms split from the nanoparticles could be bound to the N-rich defects due to the strong M–N interaction, resulting in the atomization of Ni nanoparticles.95 In another work, Li and coworkers found that noble metal nanoparticles (Pd, Pt, Au) embedded in N-doped carbon derived from ZIF-8 can be transformed into single atoms at high temperature in an inert atmosphere (Fig. 6B).86 The dynamic process of transformation was directly observed by in situ environmental TEM. Specifically, ZIF-8 nanocrystals with embedded metal nanoparticles were heated at 900 °C under an inert atmosphere for 3 h and the morphologic evolution of the nanocomposite during different heating stages was recorded by STEM. The metal nanoparticles firstly grew larger and the size distribution became less homogeneous. Subsequent heating led to the amorphization of the metal nanoparticles and the decrease in the number of nanoparticles. After extensive pyrolysis, the remaining metal nanoparticles were digested within the substrate. Characterization by energy dispersive X-ray spectroscopy analysis, AC-STEM and EXAFS confirmed that the metal atoms were atomically dispersed in the carbon matrix (Fig. 6C–E). Detailed experimental studies by TEM and DFT calculations revealed that the high temperature transformation from metal nanoparticles to metal atoms was driven by the formation of the M–N4 moiety which is more thermodynamically stable than the metal nanoparticles.

3.2. Electron/ion irradiation

High-energy electron/ion bombardment has been demonstrated as an effective method to generate defects and vacancies in graphene, which are then able to trap mobile transition metal atoms. A variety of metal atoms (e.g., Fe, Cu, Co, Pt) have been successfully incorporated into graphene lattices by this approach.29,100 For example, Wang et al. developed a two-step process to dope graphene with Fe atoms: create carbon vacancies using energetic Au or B species ejected from laser ablation and fill these vacancies with Fe atoms by a focused ion beam in the presence of Fe precursor (Fig. 7A).29 It was suggested that the defect structure was affected by the energy and the incident angle of the bombarding species and therefore the refinement of the process was important to selectively generate vacancies for hosting metal dopants. Another related study showed that focused electron beam irradiation using TEM operated at an accelerating voltage of 80 kV was able to generate vacancies for metal doping in graphene with 10 × 10 nm2 spatial control.100 Metal atoms were most frequently observed at the mono-vacancy and di-vacancy sites of graphene, and in some cases at topological defects. This method has the advantage of enabling in situ study of the structure of the metal atoms and their migration behaviors.101 However, the irradiation method lacks the capability of mass production and its application for catalysis is yet to be demonstrated.
image file: c9cs00422j-f7.tif
Fig. 7 Summary of the synthetic strategies toward G-SACs. (A) HRTEM image, simulated image and atomic model of a Pt atom trapped in a di-vacancy of graphene. The left panel shows the schematic illustration of irradiation of the graphene surface with metal ions. Reproduced with permission from Wang et al.29 Copyright 2012 American Chemical Society. (B) Schematic illustration of the ball milling approach to graphene-supported FeNx moieties. (C) Schematic illustration of Pd atoms on graphene synthesized via a two-step process of anchoring site creation and Pd ALD. Reproduced with permission from Yan et al.96 Copyright 2015 American Chemical Society. (D) Schematic of the synthesis of atomically dispersed Pt by the ultralow-temperature photochemical reaction, during which the nucleation process of isolated atoms was prevented. Reproduced with permission from Wei et al.97 Copyright 2019 Royal Society of Chemistry. (E) Schematic of the ultra-low temperature solution reduction process. The nucleation of reduced metal atoms can be significantly suppressed mainly due to the sluggish thermodynamic and kinetic process. Reproduced with permission from Huang et al.98 Copyright 2019 Nature Publishing Group. (F) Schematic illustration of surface immobilization of transition metal ions on N-doped graphene via ion adsorption. Reproduced with permission from Bi et al.99 Copyright 2018 Wiley-VCH.

3.3. Ball milling

Ball milling is a common technique in the powder production industry and it is a process where the moving balls apply their kinetic energy to the materials, reconstructing the chemical bonds of materials and producing fresh surfaces. In addition, a high-energy ball milling process has the capability to create a local high temperature up to 1000 °C and high pressure up to several GPa, and it has been demonstrated as a powerful tool for the synthesis and modification of materials, such as organics and carbon-based nanomaterials.102 For instance, ball milling has been applied for the exfoliation of graphite into graphene as well as the fragmentation and chemical functionalization of graphene.103 More recently, Deng et al. used the ball milling approach to synthesize well dispersed single atomic FeN4 moieties within the graphene matrix at a large quantity with iron phthalocyanine (FePc) and graphene nanosheets as precursors (Fig. 7B).33 For the formation mechanism, it was proposed that the outside macrocyclic structure of FePc could be destroyed by ball milling, leaving the residual isolated FeN4 center to interact with the defective sites of graphene; meanwhile the adjacent carbon atoms of FeN4 could further reconstruct to finally form the graphene-embedded FeN4 centers. The atomic Fe content was estimated to be ∼2.7 wt% by inductively coupled plasma (ICP) and further increasing the metal loading would lead to agglomeration. The generality of this method to prepare various atomic 3d transition metals (e.g., Mn, Fe, Co, Ni, Cu) within graphene lattices was demonstrated later by the same group.104

3.4. Atomic layer deposition

Atomic layer deposition (ALD) is a promising bottom-up technique to prepare catalysts with atomic precision. During the fabrication process, a substrate is alternately exposed to a series of reactive precursors, which react in a self-limiting manner to deposit materials in an atomic layer-by-layer fashion.105 As the technique is based on chemisorption, deposition only happens on reactive surface sites and therefore it is possible to realize area-selective ALD. When graphene nanosheets are used as the substrate for dispersing single atomic metals by ALD, they need to be pre-treated to introduce nucleation sites for selective deposition. Yan et al. reported the synthesis of graphene-supported Pd single atoms by the ALD technique (Fig. 7C).96 The generation of anchoring sites on perfect graphene nanosheets was achieved by firstly oxidizing graphite into graphite oxide, followed by thermal exfoliation and deoxygenation through high-temperature reduction to precisely tune the type and amount of surface oxygen functional groups. It was suggested that isolated phenols or phenol-carbonyl pairs on graphene were the primary anchoring sites for metal atom deposition during ALD. With a similar procedure, Sun et al. fabricated Pt atoms on oxygenated graphene nanosheets and found that the size, density and loading of Pt on graphene can be precisely controlled by adjusting the number of ALD cycles.106 Moreover, nitrogen-doped graphene has also been demonstrated as an effective substrate to support Pt atoms with enhanced interaction between metal and N sites.107,108 However, due to scalability and cost issues, the capability of the ALD technique to produce catalysts for commercial applications is still limited in its current format.8

3.5. Photochemical reduction

Photochemical reduction chemistry involves the absorption of photons and electronically excited states and it has been widely employed to synthesize nanocrystals.109 Recently, Wu and coworkers developed an iced-photochemical reduction method to synthesize Pt single atoms (Fig. 7D). H2PtCl6 aqueous solution was firstly frozen to yield an ice layer with a homogeneous distribution of the Pt precursor and then the frozen solution was exposed to UV irradiation to reduce the Pt precursor into Pt atoms.110 During the photoreduction process, the ice lattice confined the dispersed reactants and therefore suppressed nucleus formation. Subsequently, the melted solution containing Pt atoms was mixed with different types of substrates such as defective graphene that provided anchoring sites to stabilize Pt atoms. The same group later developed an ultralow temperature photoreduction process to synthesize Pt single atoms by firstly UV-irradiating the H2PtCl6 precursor in an antifreeze solution containing ethyl alcohol and water at −60 °C and then mixing with nitrogen-doped mesoporous carbon.97 The ultralow reaction temperature can effectively prevent the nucleation and aggregation of isolated metal atoms in solution. Wang et al. developed a modified photoreduction process, in which PtCl62− ions were firstly absorbed on nitrogen-doped porous carbon with rich absorption sites (e.g., nitrogen and oxygen heteroatoms), high porosity and abundant defects that are beneficial for the absorption of metal ions.111 The carbon with absorbed PtCl62− ions in solid state was then irradiated with UV light to realize the reduction of PtCl62− ions to Pt atoms supported on carbon. The as-prepared Pt atoms were well-dispersed on the carbon without the formation of clusters or nanoparticles. The photochemical reduction process is considered to be green and simple, but it is mainly used to synthesize noble metal atoms. Its generality for synthesizing non-noble metal atoms remains questionable considering the limited reducing capability of UV light.

3.6. Solution-phase synthesis

The rapid nucleation and growth of solid-state products during solution-phase synthesis make the formation of ultrafine nanoclusters or metal single atoms challenging. Recent research efforts suggest that metal single atoms can be obtained by solution-phase synthesis provided that the kinetics of nucleus formation could be significantly reduced or strong metal–support interaction could be introduced. Wu et al. developed a low-temperature (−60 °C) solution synthesis of single atomic Co supported on nitrogen doped mesoporous carbon through the liquid-phase reduction of CoCl2 with hydrazine hydrate using a water/alcohol mixed solvent having a low freezing point (Fig. 7E).98 The low reaction temperature was found to be critical in suppressing nucleus formation to ensure atomic metal dispersion. Zhang et al. immobilized single atomic Pt onto nitrogen-doped carbon nanocages by simply dispersing the nanocarbon into H2PtCl6 aqueous solution, followed by filtration, washing and mild drying.112 Combined experimental and theoretical investigations suggested that the micropore trapping and nitrogen anchoring from the nanocarbon substrate were beneficial for trapping the metal ion and the derived Pt single atoms. Such impregnation–adsorption method was employed to synthesize atomically dispersed Ni and Cu by immersing nitrogen-enriched graphene in the corresponding solution of metal precursors, followed by mild post-treatments (Fig. 7F).99,113 Besides nitrogen dopants, graphene functionalized with nitrile groups and carboxy groups was employed to anchor metal atoms (Cu and Fe) as the nitrile groups and carboxylated groups could serve as strong and selective ligands for the metal ions.114 The solution-phase synthetic method is simple, general and potentially suitable for mass production.

4. Identification of the atomistic structure by synchrotron X-ray absorption spectroscopy

Recent years have witnessed much progress in the synergistic efforts of theoretical and experimental studies in the rational design of high-performance electrocatalysts.2 At the foundation of such design lies the catalyst's atomistic and electronic structure that dictates the reaction mechanism and intrinsic activity trends. However, the unambiguous identification of the atomistic structure of the active sites in G-SACs represents a serious challenge largely due to the non-crystallographic ordering of the metal atoms and the heterogeneity in structure and composition.43 Several characterization techniques, such as STEM, Mössbauer spectroscopy, X-ray photoelectron spectroscopy (XPS) and electron paramagnetic resonance (EPR) spectroscopy, have been used to probe the atomistic or electronic structures of SACs.8,13,46 Recently, there has been a blossoming interest in synchrotron X-ray absorption spectroscopy (XAS), which can provide valuable information about the coordination environment and the chemical state of the probed atom in an element-selective way. Moreover, with the capability of hard X-rays to penetrate the electrode through the catalyst/electrolyte interface, XAS allows the study of the dynamic process of electrochemical reactions under in situ/operando conditions, which is essential for establishing a more precise structure–activity relationship and thus a fundamental understanding of the key factors governing these reactions. The rapidly increasing number of publications on these subjects have proved XAS to be a well-suited and powerful technique for characterizing G-SACs.25,43,44,46,115 Nonetheless, unambiguously extracting the exact atomistic and electronic structure by XAS remains neither trivial nor straightforward and the complementary use of different techniques is often needed. Here we focus our discussion on recent efforts in the employment of XAS for identifying the atomistic and electronic structure of the active sites in G-SACs. By examining the advantages and disadvantages of different XAS analysis methods, we describe a global strategy for the structural identification by XAS that can reduce loose ends to a minimum, likely serving as a general guide for future studies facing the structural characterization of G-SACs.

4.1. Brief introduction to XAS basics

XAS measures the X-ray absorption coefficient of a material as a function of X-ray photon energy (Fig. 8A). The incident X-ray photon energy scans through a specific element threshold to excite inner-shell electrons to the unoccupied continuum level (Fig. 8B), resulting in a sharp rise at certain energies (the absorption edge) and a series of oscillatory structures that contain the unique signature of a given material. XAS can be roughly divided into two regions (Fig. 8A), the X-ray absorption near edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) for their different origins, and thus they will be treated differently with an emphasis on different information that can be provided. Specifically, in the frame of the single/multiple scattering approach (Fig. 8C), the EXAFS oscillations obtained by data reduction can be interpreted by the well-known standard EXAFS equation expressed as the sum of a series of sine-function-like scattering signals from different coordination shells.116 By modeling and least-squares curve fitting of the experimental EXAFS data, local structure such as interatomic distance and coordination number around the probed atoms can be determined. On the other hand, the XANES region contains much chemical information and is generally not as readily interpreted as EXAFS. The edge energy position exhibits sensitive dependence on the formal oxidation state of the probed atom (chemical shift), which can be quantitatively identified using reference compounds with known oxidation states. The XANES spectral shape and edge energy position are also sensitive to ligand types and the three-dimensional coordination environment around the central absorbing atom. Shoulder features on the edge of the XANES spectrum are sometimes introduced for square-planar configuration of the ligand atoms. The pre-edge features are sensitive to the coordination symmetry and the formal oxidation state of the probed atom. For instance, the pre-edge peak for 3d transition metals can result from quadrupolar allowed transition (1s–3d) or from 3d–4p orbital mixing induced by a noncentrosymmetric chemical environment where the peak intensity is directly proportional to the coordination deviation of the absorbers from centrosymmetry. In general, XANES contains more abundant structural information and can be described qualitatively in terms of coordination chemistry, molecular orbitals, band structure and multiple scattering. Interested readers are referred to more detailed reviews and books on XAS theory and its application in catalysts.116–118
image file: c9cs00422j-f8.tif
Fig. 8 Brief introduction to XAS basics. (A) The XAS spectrum consisting of absorption edge and a series of oscillatory structures. The conventional division between XANES and EXAFS regions is indicated by different colors. The inset shows the magnified XANES region. (B) X-ray absorption edges nomenclature and their relationship with the corresponding excitation of core electrons. (C) The schematic illustration of the single (left) and multiple (right) scattering process of the excited photoelectrons. The EXAFS oscillations can be regarded as originating from the constructive and destructive interferences between the scattered photoelectron waves near neighbor atoms (denoted as S1 and S2) and the outgoing photoelectron waves at the absorbing atom (denoted as O).

4.2. Probing local structure by EXAFS

EXAFS has been extensively employed to decipher the local bonding environment in various materials, attributed mostly to the fact that the Fourier transform (FT) of the EXAFS spectra enables the separation of backscattering atoms directly by a path-length criterion.116 Specifically, a qualitative estimate of the coordination shells surrounding the absorbing atom can be made according to the radial locations of the EXAFS peak distributions in the FT modulus. Further quantitative structural information (e.g., bond length, coordination number and bond disorder) can be obtained by a least-squares curve fit of the EXAFS data to the standard EXAFS equation. However, a necessary step of EXAFS analysis is to pre-select and test the identity of the backscattering atom through trial-and-error fit; this process is a black box and usually suffers from large uncertainty and artificiality. Moreover, the experimental XAS data of G-SACs often have a low signal-to-noise ratio and a short k-range, which can make the pre-selection of atom type more difficult. For FT, while the identification of the backscattering atoms is relatively straightforward if the coordination shells constitute only one type of atoms at one distance (as shown in Fig. 9A and C), the analysis is substantially complicated with different types of atoms at one distance (where the FT is characterized by a single peak as shown in Fig. 9B and D).119 In contrast to FT, the wavelet transform (WT) of EXAFS allows one to resolve the backscattering atoms in both k- and R-spaces that provide the elemental nature and radial distance resolutions, respectively. From Fig. 9E and F, it can be appreciated that the type of backscattering atoms indistinguishable in the FT peak can be readily distinguished based on the k locations of the intensity maxima in the WT distribution, without any fitting process required. Therefore, the R and k space information given by WT can greatly ease the discrimination between heavier (i.e., metal atoms) and lighter backscattering atoms (i.e., C, N and O) even if they overlap at the same distance from the central atom, which is the frequently encountered situation for atomic structural investigation in G-SACs.
image file: c9cs00422j-f9.tif
Fig. 9 Comparison between EXAFS FT and WT. (A and B) Schematic view of two different coordination configurations for one type of backscattering atom (denoted as S1) located at two distances (so-called two shells) and two types of backscattering atoms (denoted as S1 and S2) overlapping at one distance (so-called one shell) around the central absorbing atom (denoted as O). (C–F) The simulated EXAFS-FT moduli (C and D) and EXAFS-WT distributions (E and F) corresponding to the two coordination configurations as shown in (A and B), respectively.

We recently synthesized a series of monodispersed atomic transition metals of Co, Ni and Cu embedded in nitrogen-doped graphene.76 The EXAFS-FT profiles feature a single main peak at ∼1.6 Å and another minor peak at ∼2.1 Å (Fig. 10A). While the main peak at ∼1.6 Å can be readily ascribed to the bonding between metal and nearest-neighbor light atoms, the assignment of the minor peak at ∼2.1 Å is rather difficult mainly because in this radial range the peak arising from next nearest-neighbor light atoms overlaps substantially with the peak from the metal–metal bond. The metal–metal peak is regarded as an indicator of the presence of a metal-derived crystalline structure. Nevertheless, the WT of the weighted EXAFS signals clearly reveals a single intensity maximum at ∼4.0 Å−1 that can be assigned to the metal-N/O/C contributions (Fig. 10B), in marked contrast to the intensity maxima at ∼7.0 Å−1 resulting from the metal–metal contributions, confirming the atomic dispersion of the metals in the samples. Another example is concerning the comparison of the EXAFS signals for Co-NG with Co atomically dispersed on N-doped graphene and Co-G without N doping.72 The FT moduli of EXAFS signals for both Co-NG and Co-G show only one single main peak at ∼1.6 Å in R-space (Fig. 10C), and the WT distributions correspond to one intensity maximum at 3.4 Å−1 for Co-NG and one at 3.2 Å−1 for Co-G (Fig. 10D). Theoretical calculations indicate that the location of the WT maximum is largely affected by the type of backscattering atoms (i.e., 3.2 Å−1 for Co–C path, 3.5 Å−1 for Co–N path, 4.3 Å−1 for Co–O path, and 6.8 Å−1 for Co–Co path). As a result, the WT maximum A at 3.2 Å−1 for Co-G can be associated with the Co–C path, and 3.4 Å−1 for the Co–N path within Co-NG. A least-squares curve fitting analysis was carried out for the first coordination shell of Co to confirm the validity of the above WT-EXAFS interpretation. Thus, combined WT and FT is highly useful for the discrimination of the single atomic dispersion in G-SACs during EXAFS analysis of G-SACs.

image file: c9cs00422j-f10.tif
Fig. 10 EXAFS-FT and EXAFS-WT analysis of the coordination environment for 3d transition metal atoms in M–N–C catalysts. (A) FT magnitudes of the K-edge phase-uncorrected EXAFS signals of Ni-NG-MW and Cu-NG-MW along with bulk metals. The vertical dashed line highlights the position of the M–M peaks. (B) WT for the k2-weighted EXAFS signals of Ni-NG-MW, Cu-NG-MW and bulk metals with optimum resolutions at ∼2.0 Å. Reproduced with permission from Fei et al.76 Copyright 2018 Wiley-VCH. (C) FT magnitudes of the Co K-edge phase-uncorrected EXAFS signal in R space. (D) WT in both k- and R-spaces for Co-NG and Co-G. Reproduced with permission from Fei et al.72 Copyright 2015 Nature Publishing Group.

In the past several years, WT has been widely used to characterize many different types of materials. Besides G-SACs, WT is also suitable for SACs supported on non-carbon substrates.120 WT has been used to preclude the presence of a metal cluster, and it can also be used to detect the existence of weak metal–metal interaction in sub-nano clusters.121 Therefore, complementary use of FT and WT is strongly recommended in future EXAFS analysis.

4.3. Electronic and 3D geometric structure by XANES

In the energy region of XANES, the single scattering approximation breaks down and many processes contribute in a non-negligible way, including the many-body effect, multiple scattering effect, polarization effect and so on, which make the signal at the relevant energies very sensitive to the 3D atomic, electronic and magnetic structures. Therefore, XANES is highly valuable in resolving the electronic and 3D geometric structure in G-SACs. For 3d transition metals, two significant pre-edge features can often be observed. The peak at lower energy is due to quadrupolar-allowed 1s–3d transition or 3d–4p orbital hybridization and its intensity is directly proportional to the coordination deviation of the photo-absorbers from centrosymmetry, while the peak at higher energy arises from a 1s → 4pz shakedown transition and acts as a fingerprint of a square-planar configuration with high D4h symmetry. For example, Zitolo et al. revealed slight differences in the XANES spectra of a pyrolyzed Fe–N–C catalyst and non-pyrolyzed Fe(II) phthalocyanine (Fe(II)Pc) (Fig. 11A), which suggests a valence state of +2 for the Fe sites in Fe–N–C.43 However, the pre-edge feature at 7118 eV shows a strong intensity for Fe(II)Pc that has a square-planar configuration but is absent in the pyrolyzed Fe–N–C, thus revealing a broken D4h symmetry with axial ligands. Similar results for the two pre-edge features were reported in a Co–N–C catalyst (Fig. 11B).25 Yang et al. also used XANES to study the oxidation state and coordination structure of atomic Ni catalysts with and without the addition of a sulfur precursor (denoted as A-Ni-NG and A-Ni-NSG).46 Interestingly, a comparison of the inflection point of the second-derivative spectrum of Ni K-edge XANES suggests that the oxidation state of the Ni atoms in A-Ni-NG and A-Ni-NSG was +1 with a 3d9, S = 1/2 electronic configuration, which could result from the lower electronegativity of pyridinic N and S relative to the pyrrolic N in Ni(II)Pc (Fig. 11C). Furthermore, as compared to Ni(II)Pc, the increased intensity in the pre-edge peak A and the reduced intensity in the pre-edge peak B for A-Ni-NG and A-Ni-NSG both indicate that the Ni atoms possess distorted D4h symmetry.
image file: c9cs00422j-f11.tif
Fig. 11 Structural identification of single metal moieties in various M–N–C catalysts by XANES spectroscopy. (A–C) The experimental K-edge XANES spectra of Fe–N–C, Co–N–C and Ni–N–C and reference samples. The inset in (A) and (C) shows a magnification of the pre-edge feature. The inset in (B) shows the first derivative curves. Reproduced with permission from Zitolo et al.43 Copyright 2015 Nature Publishing Group. Reproduced with permission from Fei et al.25 Copyright 2018 Nature Publishing Group. Reproduced with permission from Yang et al.46 Copyright 2018 Nature Publishing Group. (D and E) Comparison between the K-edge XANES experimental spectra (black dashed lines) of Fe–N–C and the theoretical spectra calculated with the depicted porphyrinic structures (solid red lines). Reproduced with permission from Zitolo et al.43 Copyright 2015 Nature Publishing Group. (F–H) Comparison between the K-edge XANES experimental spectra (black dashed lines) of Co–N–C and the theoretical spectra calculated with the depicted porphyrinic structures (solid red lines). In (G) and (H), two defective sites derived from the porphyrinic moiety by subtracting one or two nitrogen atoms are shown. Reproduced with permission from Zitolo et al.44 Copyright 2018 Nature Publishing Group. (I) Comparison between the K-edge XANES experimental spectrum of Co–N–C (solid red line) and the theoretical spectrum (black dotted line) calculated with the depicted pyridinic moiety. Reproduced with permission from Liu et al.122 Copyright 2016 The Royal Society of Chemistry.

Besides the above qualitative analysis, XANES calculations and fits can be used to better identify the atomistic structure in G-SACs. XANES calculation has been used to resolve the structural configuration based upon a profile/shape comparison between the experimental and theoretical spectra, which includes the peak positions and their intensity relationship. This is based on the fingerprint of XANES to similar structural configurations despite possible slight differences in the bond distances and/or angles. On the other hand, XANES fitting is a method to extract quantitative structural information via an iterative minimization of the differences between the experimental XANES spectrum and many theoretical ones, which proceeds via varying iteratively the structure parameters (e.g., bond distances and bond angles) based on an initial given geometry until the convergence criteria are reached.123,124 Therefore, XANES fitting can be considered as a refinement of the moiety used for XANES calculation and can also be used to confirm or discard a structural model. Zitolo et al. performed a thorough XANES fitting investigation of the active site structure in their single atomic Fe–N–C catalyst.43 They initially excluded a FeN4 moiety enclosed in a graphene sheet and a FeN2+2 moiety bridging two graphene sheets, due to obvious discrepancies between the experimental and calculated spectrum. Then, they examined a porphyrin-like FeN4C12 moiety and the agreement was improved. Substantial improvement over the whole energy range was obtained by including one O2 in the end-on configuration. The inclusion of a second O2 along the axial direction resulted in an excellent agreement between experimental and theoretical spectra (Fig. 11D). Besides, the same FeN4C12 moiety with O2 adsorbed side-on was also found to correctly reproduce the experimental spectrum (Fig. 11E). In a following study, Zitolo et al. reported the identification of the atomistic structure in a Co–N–C catalyst.44 Similar to the Fe–N–C catalyst,43 the porphyrin-like CoN4C12 moiety with and without an end-on adsorbed O2 was found to produce an excellent agreement between experimental and theoretical spectra (Fig. 11F). Additionally, two defective moieties, derived from the porphyrinic CoN4C12 moiety by subtracting one or two nitrogen atoms, were also suggested as viable candidates (Fig. 11G and H). In another study, Liu et al. studied the structure of an atomic Co catalyst. Different from the porphyrinic moieties, their XANES calculation shows that the main features of the experimental spectrum could be reproduced by a pyridinic CoN4C8 moiety with the addition of two O2 in end-on mode (Fig. 11I).122

4.4. Global XAS analysis strategy and application

XAS studies on G-SACs (and other types of SACs) were usually separated into two independent processes of EXAFS and XANES, with an emphasis on atomic radial distribution and electronic structure, respectively. However, it is worth emphasizing that both EXAFS and XANES have a common structural origin and thus the derived structural information should be consistent with each other. Based on this fact, it is strongly suggested to routinely make complementary uses of multiple XAS analysis approaches (i.e., EXAFS fitting, XANES simulation, XANES fitting, EXAFS calculation) in a global way, as schematically shown in Fig. 12A. In particular, qualitative EXAFS-WT comparison can be first used to identify the atomic distribution; then, quantitative EXAFS-FT fitting is used to extract interatomic distances and coordination number, which are crucial parameters to distinguish potential structural motifs and can facilitate the modeling of the atomic structure used for XANES calculation. As compared to the radial atomic distribution detected by EXAFS, one of the greatest strengths of XANES is its high sensitivity to the spatial arrangement of the coordinating atoms. Detailed XANES analysis usually requires extensive theoretical calculation with various well-defined model structures. Also, EXAFS simulation can be used to assist the structural identification. Just as triangulation, the ultimate goal is to achieve a global consistency between EXAFS and XANES. Therefore, the as-described global XAS analysis strategy can allow the analyzers to efficiently and properly identify the structure of SACs with minimized uncertainty and artificiality.
image file: c9cs00422j-f12.tif
Fig. 12 Schematic presentation of the global XAS analysis strategy and its application to the structural identification of the atomistic sites in Ni-NHGF catalyst. (A) Global XAS analysis strategy consisting of various EXAFS and XANES analysis approaches. (B) FT magnitudes of the experimental K-edge EXAFS signals of Ni-NHGF along with reference samples (solid lines). The dashed lines represent calculated spectra based on a divacancy-based NiN4C4 moiety enclosed in the graphene lattice. (C) WT for the k3-weighted EXAFS signals of Ni-NHGF. (D and E) Ni K-edge EXAFS analysis of Ni-NHGF in k (D) and R (E) spaces. Curves from top to bottom are the two-body backscattering signals included in the fit and the total signal (red line) superimposed on the experimental signal (black line). The inset shows the structural model of a NiN4C4 moiety. (F) The experimental Ni K-edge XANES spectra and the first derivative curves (insets) of Ni-NHGF and reference samples. (G) Comparison between the experimental K-edge XANES spectra of Ni-NHGF and the theoretical spectra calculated with the depicted structure. (H) The best fit between the theoretical (solid red lines) and experimental XANES spectra (black dotted lines) for Ni-NHGF. Reproduced with permission from Fei et al.25 Copyright 2018 Nature Publishing Group.

Lately, we have conducted such a global and comprehensive XAS analysis toward the unambiguous identification of the MN4C4 (M = Fe, Co, Ni) moiety implanted in the graphene lattice.25 Taking the Ni case for instance, we firstly performed EXAFS-FT and EXAFS-WT analyses to confirm the exclusive atomic dispersion of Ni atoms. We found that the EXAFS-FT profile features the co-presence of a Gaussian-like main peak at 1.44 Å and a minor satellite peak at 2.01 Å (Fig. 12B). While the minor satellite peak overlaps partially with the Ni–Ni peak at 2.18 Å for bulk Ni, EXAFS-WT analysis detects only one intensity maximum at approximately 4.0 Å−1 (Fig. 12C), which can be assigned to the M–N/O/C contributions. On the basis of the EXAFS-WT result, three backscattering paths of Ni–N, Ni–O and Ni–C were used to model the EXAFS signal by EXAFS curve fitting. The first coordination sphere of Ni was revealed to be a square-pyramidal configuration for the Ni–N/O bonding (Fig. 12D and E), with a second coordination sphere of about four C atoms, indicating a NiN4C4 moiety with one oxygen atom in the axial direction. On the other hand, the XANES spectrum indicates a valence state of +2 for Ni (Fig. 12F). Moreover, the rather weak intensity for the pre-edge peak (Fig. 12F), which is due to a 1s–4pz shakedown transition characteristic for a square-planar configuration with high D4h symmetry, indicates a broken D4h symmetry and calls for axial ligands, which is in accordance with the EXAFS fitting results. Then, XANES calculation based on various possible NiNxCy moieties was applied to resolve the atomic-site structure. The porphyrin-based NiN4C12 moiety and a pyridinic-N-based NiN4C8 moiety were excluded due to the large discrepancy observed between the simulated spectrum and the experimental spectrum. Another series of XANES simulations by embedding the NiN4C4 moieties in the 2D graphene lattice were examined, with substantially improved agreement. With the addition of one end-on dioxygen molecule in the axial position of the Ni centre, excellent agreement was obtained (Fig. 12G). XANES fitting based on the as-determined NiN4C4 moiety reveals that the calculated spectra match excellently with the experimental spectra (Fig. 12H), and the bond metrics are in good agreement with those determined by EXAFS analysis and DFT prediction. Together, these XANES analyses clearly demonstrate single atomic metal centers with four pyridinic N atoms and one adsorbed O atom in the first coordination sphere and four C atoms in the second coordination sphere, which is fully consistent with the EXAFS results. Lastly, EXAFS simulations based on such divacancy-based NiN4C4 moiety show that the simulated EXAFS spectra (dashed lines in Fig. 12B) are nearly identical to the experimental spectra. Therefore, the combination of EXAFS-WT, EXAFS-FT and XANES analyses on Ni-NHGF has achieved a global consistency and unambiguously identified the divacancy-based NiN4C4 moiety in the graphene lattices.

4.5. Monitoring dynamic processes by in situ/operando XAS

The real-time observation of the reacting single-atom sites via in situ/operando studies could provide in-depth insights into the dynamic changes in the atomic and electronic structures of the catalysts as a function of the applied potential, the complex reaction kinetics and mechanisms as well as the stability over time or potential. Benefiting from the ability of X-ray photons to penetrate matter at the micrometer or even millimeter scale (depending on the photon energy) and the suitability of XAS to study the geometric and electronic structure of the catalysts, XAS holds a place of choice for catalytic in situ/operando studies of G-SACs.125–128 Zitolo et al. studied the atomic and electronic structures of pyrolyzed Co–N–C and Fe–N–C catalysts during the ORR using operando XANES.44 Upon increasing the potential, the main absorption edge for Co–N–C showed insignificant changes across the whole ORR potential range (Fig. 13A), suggesting that the Co moieties did not change the structure. In contrast, Fe–N–C exhibited a gradual shift toward higher energy (Fig. 13B), which indicates an increase of oxidation state and structural changes due to the reorganization of the N (or C) ligands and/or spin state. Besides, the authors used comparative operando XANES spectra collected in an O2-free or O2-saturated electrolyte to detect the bonding of the active sites (Fig. 13C and D). While slight spectral changes were observed for Fe–N–C, large variation was detected in the white line peak region for Co–N–C. Analysis of the differential ΔμO2–N2 spectra of Co–N–C at 0.8 V revealed an active-site structure with four N atoms at 1.95 Å with or without an oxygen molecule adsorbed end-on at 2.22 Å. As oxygen comes exclusively from water activation in a N2-saturated electrolyte, these results suggest the formation of a Co–O bond only in the O2-saturated electrolyte for Co–N–C and the formation of a Fe–O bond probably in both O2- and N2-saturated electrolytes for Fe–N–C. However, the exact structural change induced by the redox switch remains unclear and will be of great interest in the future.
image file: c9cs00422j-f13.tif
Fig. 13 Operando XANES studies during the ORR process. (A and B) Metal K-edge XANES spectra taken in 0.5 M H2SO4 aqueous solution saturated with N2 at ORR potentials for Co–N–C (Co0.5) (A) and Fe–N–C (Fe0.5) (B), respectively. The insets show differential Δμ XANES spectra relative to the spectrum recorded at the lowest potential of 0.2 V. (C and D) Metal K-edge XANES spectra for Co–N–C (Co0.5) (C) and Fe–N–C (Fe0.5) (D) measured as a function of the saturating gas (O2 or N2) at either 0.2 V or 0.8 V versus RHE. The insets are Δμ spectra obtained by subtracting the normalized XANES spectra recorded in an O2-saturated electrolyte from that recorded in a N2-saturated electrolyte at a fixed potential. Reproduced with permission from Zitolo et al.44 Copyright 2017 Nature Publishing Group.

Yang et al. reported an A-Ni-NG catalyst with atomically dispersed monovalent Ni(I) atoms for CO2RR.46 They applied both the operando Ni K-edge XANES and EXAFS spectra to track the adsorption and activation of CO2 on A-Ni-NG. The absorption edge position exhibited a positive shift of ∼0.4 V under open-circuit voltage in CO2-saturated KHCO3 solution when compared to the Ar-saturated electrolyte (Fig. 14A), indicating a slight oxidation of Ni sites that was interpreted as delocalization of the unpaired electron in the 3dx2y2 orbital and spontaneous charge transfer from Ni(I) to the carbon 2p orbital in CO2 to form a CO2δ species. The formation of the Ni–C bond due to CO2 absorption also caused a slight increase of the intensity for the first nearest-neighboring major peak at approximately 1.45 Å in EXAFS-FT (Fig. 14B). However, when a CO2 reduction potential of −0.7 V versus RHE was applied, the Ni oxidation state was reduced based on the observation of a shift of the absorption edge back to lower energy, which was ascribed to the recovery of low-oxidation-state Ni after one cycle of CO2 reduction. Simultaneously, the major peak dominated by the Ni–N bond showed a shift of ∼0.04 Å to longer lengths, ascribed to an elongation of the Ni–N bond that resulted from the displacement of Ni atoms out of the graphene plane due to the electron transfer from the Ni atom to CO2.

image file: c9cs00422j-f14.tif
Fig. 14 Operando X-ray absorption spectroscopy of A-Ni-NG during the CO2 reduction process. (A) Normalized operando Ni-edge XANES spectra for A-Ni-NG at various biases (applied voltage versus RHE) in 0.5 M KHCO3 aqueous solution at room temperature in an atmosphere of Ar or CO2. The inset shows the enlarged Ni K-edge XANES spectra. (B) EXAFS-FT spectra (without phase correction) of A-Ni-NG under open-circuit voltage in Ar (OCV-Ar) and CO2 (OCV-CO2), and at −0.7 V (versus RHE), in which an expanded Ni–N bond was detected. Reproduced with permission from Yang et al.46 Copyright 2018 Nature Publishing Group.

5. Applications for electrochemical energy conversion

Electrochemical energy conversion has been considered as one of the most promising routes for the construction of a sustainable and environmental-friendly energy cycle.129 The development of high-performance and inexpensive electrocatalysts is of vital importance to increase the reaction rate, efficiency, selectivity and stability of energy conversion technologies, such as water splitting, fuel cells, CO2 conversion, N2 reduction and metal–air batteries. As stated above, the ultrahigh surface area, superior electrical conductivity, and excellent (electro-)chemical stability make G-SACs highly attractive as electrocatalysts for energy conversion applications. Below, we will summarize the recent progress in the applications of G-SACs in ORR, HER, OER, CO2RR, NRR and so on.

5.1. Oxygen reduction reaction

The ORR proceeds through either a 2-electron pathway (O2 + 2H+ + 2e → H2O2 in acid) or a 4-electron pathway (O2 + 4H+ + 4e → 2H2O in acid). The 4-electron pathway is desired for electricity generation and it is important in several energy-conversion technologies, such as fuel cells and metal–air batteries, while H2O2 generated by the 2-electron pathway is a valuable chemical used in disinfection, pulp bleaching and the oxidation of various chemicals. The electrochemical production of H2O2 is a preferred method compared to the traditional anthraquinone redox process as it is environmentally friendly, safe, and easily adopted on a small scale.130 Therefore, considerable research efforts have been devoted to developing catalysts that can selectively catalyze either the 2-electron pathway or the 4-electron pathway.

Pt-Based materials remain the most active catalysts toward the 4-electron process, but their scarcity and high cost call for alternative catalysts based on non-precious transition metals or nanocarbons, among which M–N–C (M = Fe or Co) catalysts with atomic MNx moieties embedded in graphene represent an attractive alternative.131 M–N–C catalysts were generally prepared by the pyrolysis method and the optimizations in the synthetic conditions (e.g., precursor combinations, pyrolysis parameters) have resulted in a large number of high-performance ORR catalysts.53,62,64,65,82,93,131–135 Zitolo et al. synthesized a Fe–N–C catalyst by pyrolyzing the Zn(II) zeolitic imidazolate framework, Fe(II) acetate and 1,10-phenanthroline, and the resulting composite was quasi-free of crystallographic Fe structures, as demonstrated by 57Fe Mössbauer spectroscopy and EXAFS-FT analysis.43 They found that keeping the metal loading sufficiently low and pyrolysis duration short was critical in minimizing the formation of crystalline Fe species. When tested in acidic conditions with a rotating ring-disk electrode (RRDE) and in a polymer electrolyte membrane (PEM) fuel cell, the optimized catalysts generated less than 2% H2O2 at potentials of interest for PEM, suggesting a 4-electron dominated reduction process. Combined EXAFS and XANES analyses revealed the atomistic structure of Fe to be a porphyrinic planar architecture with a FeN4C12 core integrated in a disordered graphene matrix and with O2 absorbed in the side-on or end-on configuration. Adopting a similar synthetic approach, the same group then synthesized a Co–N–C catalyst free of cobalt particles and it exhibited considerably higher ORR activity in acidic medium than metal-free N–C catalyst, suggesting that the atomic Co moieties were the active sites.44 The selectivity for water formation of the Co–N–C catalyst was found to be lower than the Fe–N–C counterpart. XANES, EXAFS, magnetic susceptibility measurements and DFT analysis were applied to identify the atomistic structure and electronic state of the Co sites in Co–N–C. Interestingly, other than the porphyrinic CoN4C12 moiety as found in Fe–N–C, two defective cobalt porphyrinic moieties (CoN3C10 and CoN2C5) were also suggested as viable candidates in Co–N–C. DFT calculation showed that these cobalt moieties likely bind oxygen intermediates too weakly to allow efficient O2 reduction. Based on these findings, it was proposed that the modifications of the carbon matrix in Co–N–C with nitrogen or other heteroatoms resulting in a stronger O2 binding energy at Co centers should improve their ORR activity.

It is important to note that not all single atomic metal sites in M–N–C catalysts can participate in the ORR process as the reaction takes place at the multiple-phase boundary and metal sites that are buried within the carbon matrix remain inactive. To improve the utilization efficiency of active sites, a high degree of mesoporosity and macroporosity can be introduced into the M–N–C catalysts. For example, Chung et al. synthesized a porous Fe–N–C catalyst with polyaniline and cyanamide as nitrogen precursors.51 The cyanamide with low decomposition temperature served to form macropores, which would provide better accessibility to the catalytic sites and a more open framework for improving the ionomer distribution within the catalyst layers. The atomic dispersion of FeN4 active sites and the coordination between metal and nitrogen atoms were evidenced by HAADF-STEM and EELS but without verification by XAFS analysis. Significantly, the hydrogen (H2)-air fuel cell fabricated using the Fe–N–C catalysts exhibited comparable performance to a Pt cathode under automotive-application conditions. In another work, Wan et al. fabricated a concave-shaped Fe–N–C catalyst with rich mesoporosity, external surface area and dense Fe–N4 moieties by pyrolyzing mesoporous SiO2-coated ZIF-8.85 The high porosity in this catalyst can afford a large number of active sites at the triple-phase boundary of carbon (e), the ionomer (H+) and O2 during ORR tests and at the same time facilitate efficient mass transport (Fig. 15A). The EXAFS-FT of the catalyst exhibited a main peak at 1.5 Å−1 in R space and the EXAFS-WT showed only one intensity maximum at ∼4.0 Å−1 in k space that was assigned to Fe–N(O), suggesting the mononuclear Fe centers free of crystalline structures (Fig. 15B and C). The structure of the Fe center was identified as the FeN4C8 moiety with two O2 molecules adsorbed in end-on mode by XANES analysis (Fig. 15D). When tested in a PEM fuel cell under 2.5 bar H2–O2, this porous Fe–N–C catalyst exhibited a current density of 560 mA cm−2 at 0.8 ViR-free that is comparable to Pt/C and a peak power density of 1.18 W cm−2 at 0.47 V (Fig. 15E). The control sample without additional mesopores displayed inferior activity and its voltage dropped more rapidly at higher current densities when compared to the porous counterpart, suggesting its poor mass transport. When tested under the DOE testing protocol at 1 bar H2–O2, the porous Fe–N–C catalyst exhibited a current density of 0.047 A cm−2 at 0.88 ViR-free, higher than the DOE target of 0.044 A cm−2 (Fig. 15F).

image file: c9cs00422j-f15.tif
Fig. 15 Fe–N–C electrocatalysts toward ORR with the 4-electron pathway. (A) A schematic illustration of active sites at the triple-phase boundary in the concave-shaped Fe–N–C catalyst-based electrodes. (B) k3-Weighted EXAFS-FT with optimum resolutions at 2.0 Å. (C) k3-Weighted EXAFS-WT with optimum resolutions at 2.0 Å of the concave-shaped Fe–N–C catalyst. (D) A comparison between an experimental XANES spectrum and a theoretical spectrum, for which calculations are based on the predicted structure of the FeN4C8 moiety with two O2 molecules adsorbed in end-on mode. (E) Polarization and power density curves of the optimal catalyst and control catalysts at 2.5 bar H2–O2. (F) Tafel plot for determination of the activity of the optimal catalyst at 0.9 ViR-free and volumetric activity at 0.8 ViR-free measured under 1 bar H2–O2. Reproduced with permission from Wan et al.85 Copyright 2019 Nature Publishing Group.

In addition to Co and Fe single atoms, other transition metal atoms (e.g., Zn, Mn, Cu, Ir, Pt) supported on defective carbon have also been demonstrated as excellent ORR electrocatalysts with the 4-electron pathway that have comparable or even superior performance compared to commercial Pt/C catalysts.80,91,136–138

The unique electronic properties and isolated atomic sites in SACs can suppress reactions requiring metal ensemble sites and therefore offer opportunities to tune the selectivity. For example, Choi and co-workers achieved selective O2 reduction toward H2O2 production on atomically dispersed Pt catalyst.130 This catalyst was synthesized via a simple wet-impregnation method with S-doped zeolite-templated carbon composed of 3D graphene nanoribbons as the substrate, where the S content can be tuned by adjusting the annealing atmosphere (H2S/He and pure He) during the post-treatment of the carbon substrate. As revealed by TEM and EXAFS-FT (Fig. 16A and B), Pt (5 wt%) was in a highly dispersed atomic form in the sample with high S content (up to 17 wt%) (HSC), while the formation of Pt clusters was observed in the sample with low S content (4 wt%) (LSC) and non-doped zeolite-templated carbon (ZTC), indicating that the S-moieties can efficiently stabilize the Pt atoms. The active site structure of Pt/HSC was proposed to be Pt atom ligated by two thiophenes and two thiolates based on EXAFS fitting (Fig. 16C), and the oxidation state of Pt was determined to be +2 by XPS and XANES. Electrocatalytic measurements under acidic media showed that the Pt/HSC catalyst predominantly produced H2O2 (96% selectivity, n = 2.1) via a 2-electron ORR process, in contrast to Pt/LSC (60% selectivity, n = 2.9) and Pt/ZTC (28% selectivity, n = 3.5) catalysts mediating mixed 2-electron and 4-electron pathways (Fig. 16D and E). DFT calculations implied that the 2-electron pathway is kinetically more favored than the 4-electron pathway on the S-coordinated Pt atoms.

image file: c9cs00422j-f16.tif
Fig. 16 Structure and ORR activity of Pt/ZTC, Pt/LSC and Pt/HSC catalysts. (A) Atomic resolution HAADF-STEM image of Pt/HSC catalyst. (B) Fourier transforms of k3-weighted Pt LIII-edge EXAFS, confirming that Pt species in Pt/HSC exhibited only Pt–S coordination (CN = 3.8) without significant Pt–Pt coordination and the formation of Pt–Pt bonding in Pt/ZTC and Pt/LSC. (C) Proposed atomistic structure of Pt/HSC with possible thiophene- and thiolate-like functional groups at graphene edge sites, as well as coordinated Pt species. (D) ORR activities of the catalysts measured in an O2-saturated 0.1 M HClO4 electrolyte with 900 rpm rotation. (E) H2O2 production selectivity estimated by RRDE experiments (Pt ring potential: 1.2 VRHE). Reproduced with permission from Choi et al.130 Copyright 2016 Nature Publishing Group.

5.2. Hydrogen evolution reaction

Hydrogen (H2) is the cleanest fuel available and is an attractive substitute for fossil fuels. Electrochemical reduction of water through hydrogen evolution is a sustainable approach to H2 and it has promised to replace the traditional production method of steam-reformed methane that is energy demanding and environmentally unfriendly.139 The most effective catalysts for HER are based on precious Pt nanomaterials with low overpotential, small Tafel slope and high exchange current density. Due to the high cost and scarcity of Pt, decreasing the size of Pt nanoparticles to the limit of single atoms is highly desirable to maximize their efficiency by making use of nearly all Pt atoms.140–142 Cheng et al. fabricated single Pt atoms and clusters on nitrogen-doped graphene using the ALD technique and studied their catalytic performance toward HER.108 During the ALD process, the Pt precursor reacted with the nitrogen-doped graphene, where the nitrogen dopants provided a strong interaction between the deposited species and the support (Fig. 17A). The resulting catalysts exhibited negligible onset overpotentials with higher activity than the commercial Pt/C catalyst (Fig. 17B). Moreover, when normalized to the Pt loading, the mass activity of the optimized catalyst was 37 times higher than that of the Pt/C catalyst. The enhancement in performance was ascribed to the partially unoccupied density of states in the 5d orbital of the Pt atom. However, it is worth noting that there was the co-presence of Pt atoms and clusters in the catalysts, as observed by STEM images, and thus it was difficult to deconvolute the catalytic contributions from each. In another work, Liu et al. suggested that the introduction of curvature into the graphitic carbon can enhance the HER activity of the supported Pt atoms (Fig. 17C).141 Pt single atoms distributed on nanosized onion-like carbon (Pt1/OLC) were synthesized by ALD and their atomic deposition on multishell fullerene structures was characterized by STEM and EXAFS-FT (Fig. 17D and E). Electrochemical studies suggested that the Pt1/OLC with only 0.27 wt% of Pt exhibited a low overpotential of ∼38 mV at 10 mA cm−2, which is comparable to a commercial 20 wt% Pt/C catalyst and superior to a 5 wt% Pt/C catalyst (Fig. 17F). Additionally, Pt atoms supported on two-dimensional graphene (Pt1/graphene) prepared by the same ALD procedure displayed inferior activity compared to Pt1/OLC. When normalizing the current by Pt loading at the same overpotential of 38 mV, the Pt1/OLC catalyst exhibited a mass activity of 7.4 A mg−1Pt that is ∼2.5 times and ∼43 times higher than those of the Pt1/graphene catalyst and the commercial Pt/C catalyst, respectively (Fig. 17G). DFT calculation revealed that the enhanced catalytic activity of Pt1/OLC could be ascribed to a tip-enhanced local electric field at the Pt site on the curved support and the high local proton concentration near the catalyst surface. This interesting study highlights the critical roles of the local structure and morphology of the substrates in modulating the catalytic performances of the supported metal atoms and provides a new avenue to the rational design of high-performance SACs.
image file: c9cs00422j-f17.tif
Fig. 17 (A) Schematic illustration of the Pt ALD mechanism on nitrogen-doped graphene nanosheets. (B) HER polarization curves of ALD-derived Pt catalysts. Reproduced with permission from Cheng et al.108 Copyright 2016 Nature Publishing Group. (C) The schematic illustrates that the catalytically active particles were reduced in size to a single-atom form and the dimensionality of the catalyst support was reduced by using quasi-0D OLC. (D) The HAADF-STEM image of Pt1/OLC. (E) The Pt L3 edge EXAFS-FT spectra of Pt1/OLC (red) along with several control samples for comparison. (F) Polarization curves of Pt1/OLC in comparison with Pt1/graphene and Pt/C in 0.5 M H2SO4. (G) The mass activity of Pt1/OLC normalized to the Pt loading at an overpotential of 38 mV compared to the reference catalysts. Reproduced with permission from Liu et al.141 Copyright 2019 Nature Publishing Group.

The desire to eliminate the reliance on the use of Pt has stimulated the search for HER catalysts based on earth-abundant metals, which however usually suffer from insufficient chemical stability in strong acidic or basic electrolytes. The strong coordination of the metal atoms to the support can greatly enhance their (electro-)chemical stability. By acid etching of a porous Ni foam pre-coated with CVD-grown graphene layers, Qiu et al. fabricated nanoporous graphene-supported Ni atoms along with small amounts of 1–2 nm clusters, as observed by STEM.143 The resulting catalyst showed superior HER performance with a low onset overpotential of ∼50 mV and cycling stability, which were ascribed to the charge transfer between the Ni atoms and the neighboring C atoms. Yao and co-workers prepared well-dispersed atomic Ni (1.5 wt%) in a graphitized carbon matrix by carbonization of a Ni-based MOF in an inert atmosphere followed by electrochemical activation.144 Examination of different regions by STEM suggested that about 78% of the Ni species were present as well-separated single atoms with the rest as nanoclusters. This catalyst exhibited overpotentials of 34 mV, 48 mV, and 112 mV at a current density of 10 mA cm−2, 20 mA cm−2 and 100 mA cm−2, respectively, a low Tafel slope of 41 mV dec−1, and stable operation for over 25 h in a chronoamperometric test. The same group later synthesized another HER catalyst of atomically dispersed Ni on defective graphene (A-Ni@DG, 1.24 wt% Ni) by annealing the mixture of Ni2+ precursor and defective graphene, followed by acid-leaching to remove the redundant Ni-derived nanoparticles (Fig. 18A).145 The A-Ni@DG catalyst displayed a fragmentized and porous 2D structure without any obvious nanoparticles and the Ni atoms were uniformly distributed on the graphene layers (Fig. 18B). The Ni atom was suggested to occupy the di-vacancy of the graphene lattices, as revealed by the HAADF-STEM image (Fig. 18C). The Ni species were predominantly present as atomically dispersed Ni, as evidenced by the absence of an appreciable Ni–Ni peak in the EXAFS-FT spectra. The HER activity of A-Ni@DG, with an overpotential at 10 mA cm−2 as low as 70 mV and a Tafel slope of 31 mV dec−1 in 0.5 M H2SO4, was highly comparable to that of the commercial Pt/C catalyst (Fig. 18D).

image file: c9cs00422j-f18.tif
Fig. 18 Representative examples of HER catalysts based on graphene-supported atomic Ni and Co catalysts. (A) The fabrication schematic of Ni atoms on defective graphene. (B and C) STEM-EDS elemental mapping (B) and HAADF-STEM image (C) of the atomic Ni catalyst. (D) HER polarization curves of the atomic Ni catalysts along with other control samples. Reproduced with permission from Zhang et al.145 Copyright 2018 Elsevier. (E) HAADF-STEM image of the single Co atoms on nitrogen-doped graphene in the Co-NG catalyst. (F) HER polarization curves of the Co-NG catalyst along with control samples. (G) Tafel plots of the Co-NG catalyst and control samples. Reproduced with permission from Fei et al.72 Copyright 2015 Nature Publishing Group.

In addition to Ni, another first-row transition metal, Co, has also been atomically dispersed in graphene and repeatedly explored as a HER electrocatalyst. Fei et al. achieved atomic Co (0.57 at%) embedded in nitrogen-doped graphene (Co-NG) that is active and stable in both acidic and alkaline conditions.72 The catalyst was synthesized by thermally annealing GO and Co precursor in a NH3 atmosphere. The Co metals were atomically dispersed in the graphene substrate (Fig. 18E), and further confirmed by EXAFS-FT and EXAFS-WT analyses. The co-existence of Co and N in the catalysts was found to be critical in achieving high HER activity (Fig. 18F), indicating that the active sites were associated with Co and N. Impressively, the onset overpotential of Co-NG in acid was as low as 30 mV and the overpotential required to deliver 10 mA cm−2 was 147 mV. The Tafel slope of Co-NG was determined to be 82 mV dec−1 (Fig. 18G). Other than GO, molecular and polymeric precursors have been used to prepare atomic Co catalysts toward HER via the pyrolysis strategy.146–148 Liang et al. obtained a CoNx/C catalyst by pyrolyzing Co–N4 macrocycles or cobalt/o-phenylenediamine composites with silica colloids as hard templates.147 This catalyst, with only 0.14 wt% Co, exhibited a low overpotential of 133 mV at 10 mA cm−2, high turnover frequencies and high stability. Though XAFS characterization was not provided in this study, high-resolution TEM and XPS analyses, along with acid leaching and thiocyanate ion poisoning experiments, suggested that the catalytic activity originated from the atomically dispersed CoNx centers. Wang et al. synthesized self-supported and porous carbon fiber foams embedded with the CoNx complex by pyrolyzing the Co2+-absorbed polyaniline precursor.146 The annealing temperature and the porous structure were suggested to play critical roles in the formation and activity of the atomic Co sites. XAFS analyses suggested that annealing below 850 °C resulted in the predominant formation of atomic Co, while metal nanoparticles would form when annealed at 950 °C. The optimal catalyst annealed at 750 °C with 0.22 at% Co showed excellent HER activity (138 mV at 10 mA cm−2; 212 mV at 100 mA cm−2) and long-term stability. Besides, single Mo and W atoms supported on nitrogen-doped graphene layers have recently been demonstrated as highly efficient HER catalysts.149–151Table 1 summarizes different HER catalysts based on single metal atoms supported on graphene layers or graphitized carbons.

Table 1 Single atom catalysts supported on graphene layers or graphitized carbon for electrocatalytic water splitting
Catalyst Synthetic method Metal content [wt%] Catalyst loading [mg cm−2] Electrolyte η 10 [mV] Tafel slope [mV decade−1] TOF [s−1] at specific η Ref.
Hydrogen evolution reaction
ALDPt/NGNs ALD 2.1 by ICP 0.05 0.5 M H2SO4 ∼40 29 N/A 108
Pt1/OLC ALD 0.27 by ICP 0.51 0.5 M H2SO4 38 36 40.78 at 100 mV 141
Pt1/graphene ALD 0.33 by ICP 0.51 0.5 M H2SO4 ∼60 42 16.67 at 100 mV 141
Pt1/MC Photochemical reduction 2.6 by ICP 0.38 0.5 M H2SO4 25 26 N/A 110
Pt1/NPC Photochemical reduction 3.8 by ICP 0.1 0.5 M H2SO4 29 28 2.93 at 25 mV 111
Pt1/NMC Photochemical reduction 2.54 by ICP 0.39 0.5 M H2SO4 29 26 N/A 97
Pt SAs/DG Thermal emission 2.10 by ICP 1.0 0.5 M H2SO4 23 25 N/A 94
Pt1/hNCNC Impregnation-adsorption 2.92 by UV-vis 0.1 0.5 M H2SO4 15 24 7.67 at 20 mV 112
Pt@PCM Pyrolysis 0.53 by ICP N/A 0.5 M H2SO4 105 64 43.6 at 500 mV 142
Co-SAS/HOPNC Pyrolysis 0.49 by ICP 0.6 0.5 M H2SO4 137 52 0.41 at 100 mV 60
Co-NG Pyrolysis 2.54 by ICP 0.28 0.5 M H2SO4 147 82 0.10 at 100 mV 72
Co-NG-MW Microwave-assisted pyrolysis 1.1 by ICP 0.1 0.5 M H2SO4 175 80 0.385 at 100 mV 76
Co1/PCN Pyrolysis 0.3 by ICP 0.5 1 M KOH 89 52 5.98 at 100 mV 126
PANICo750A Pyrolysis 1.05 by ICP N/A 0.5 M H2SO4 138 55 N/A 146
CoNx/C Pyrolysis 0.14 by ICP 2.0 0.5 M H2SO4 133 57 0.39 at 100 mV 147
Co-SAC Pyrolysis 1.39 by XPS 0.2 0.5 M H2SO4 230 99 N/A 148
Ni-Doped graphene CVD N/A N/A 0.5 M H2SO4 ∼180 45 0.8 at 300 mV 143
A-Ni-C Pyrolysis 1.5 by ICP 0.28 0.5 M H2SO4 34 41 N/A 144
A-Ni@DG Pyrolysis 1.24 by ICP 0.26 0.5 M H2SO4 70 31 5.7 at 100 mV 145
W-SA Pyrolysis 1.21 by ICP 0.2 0.1 M KOH 85 53 6.35 at 120 mV 150
Mo1N1C2 Pyrolysis 1.32 by ICP 0.41 0.1 M KOH 132 90 0.465 at 100 mV 149
Oxygen evolution reaction
Ni-NHGF Pyrolysis 0.25 by ICP 0.27 1 M KOH 331 63 0.72 at 300 mV 25
S|NiNx-PC/EG Pyrolysis 0.2 by ICP 0.15 1 M KOH 280 45 10.9 at 300 mV 48
A-Ni@DG Pyrolysis 1.24 by ICP 0.26 1 M KOH 270 47 13.4 at 300 mV 145
S,N-Fe/N/C-CNT Pyrolysis 3.3 by XPS 0.6 0.1 M KOH 370 82 N/A 57
Fe–N4 SAs/NPC Pyrolysis 1.96 by ICP 0.51 1 M KOH 430 95 N/A 151
Mn-NG Pyrolysis 0.026 by ICP N/A 1 M KOH 337 55 N/A 74

5.3. Oxygen evolution reaction

OER is an important process in several energy-conversion technologies, such as water electrolyzers, metal–air batteries and CO2 conversion. The intrinsically sluggish kinetics of the OER, which involves multiple proton-coupled electron-transfer steps, have stimulated the exploration of highly active and stable electrocatalysts.5,57 Precious-metal-based electrocatalysts (Ru and Ir) possess high activity, but their high cost and scarcity have hindered their large-scale practical application. As alternative catalysts based on earth-abundant elements, Fe-, Co- and Ni-based oxides and derivatives have been intensively studied,152 and their atomic counterparts supported on graphene have recently been demonstrated as efficient OER catalysts (Table 1). For example, the A-Ni@DG catalyst discussed in the HER section was also highly active toward OER with a small overpotential of 270 mV to deliver 10 mA cm−2 and a Tafel slope of 47 mV dec−1 in 1 M KOH.145 Chen et al. developed an OER catalyst composed of atomic Fe–N species (verified by HAADF-STEM and EXAFS-FT) on N and S co-doped hierarchical carbon layers, prepared by pyrolysis of CNTs coated with 2,2-bipyridine and Fe precursor.57 The resulting catalyst exhibited an overpotential of 370 mV at a current density of 10 mA cm−2 and a Tafel slope of 82 mV dec−1 in 0.1 M KOH solution. Guo et al. reported boron-doped Co–N–C active sites dispersed in porous carbon sheets by a soft template self-assembly pyrolysis method.47 The atomically dispersed boron-doped Co–N–C sites were evidenced by HAADF-STEM, XPS, XANES and EXAFS characterization. The catalyst required 430 mV to deliver 10 mA cm−2 in 0.1 M KOH. DFT calculations suggested that the introduction of boron gives an electron-deficient site that can strengthen the interaction with oxygenated species, facilitating the reaction kinetics. Hou et al. reported S- and N-coordinated Ni atoms dispersed in porous carbon nanosheets as a highly active and stable OER catalyst with a low overpotential of 1.51 V at a current density of 10 mA cm−2 and a Tafel slope of 45 mV dec−1 in alkaline media, making it superior to the benchmark Ir/C catalysts.48 XANES and EXAFS studies revealed that the Ni atom adopted a distorted Ni–[N/S]4 bonding configuration that deviated from the ideal square planar geometry and the coordination numbers of Ni–N and Ni–S were estimated by EXAFS fitting results to be 2.8(2) and 0.8(3), respectively. Theoretical studies based on such a coordination environment suggested that the high catalytic activity of the S- and N-coordinated Ni atoms resulted from the optimized density-of-states distribution and the enhanced electron transfer ability of the active sites. Our group recently reported a series of atomic transition metals (Fe, Co, Ni) in nitrogen-doped graphene by pyrolyzing GO and metal precursor in a NH3 atmosphere and investigated the influence of the metal center in determining the OER activity.25 The uniform distribution of the metal atoms in the graphene matrix was evidenced by HAADF-STEM (Fig. 19A) and EXAFS-FT/WT (Fig. 12B and C). DFT calculation results suggested that the OER activity and reaction pathways were strongly dependent on the number of d electrons (Nd) of the metal with Fe (Nd = 6) and Co (Nd = 7) following the single-site mechanism and Ni (Nd = 8) following the dual-site mechanism (Fig. 19B). Electrochemical measurements in a 1 M KOH electrolyte confirmed the metal-dependent OER activity in different atomic catalysts following the trend Ni > Co > Fe. Specifically, the atomic Ni catalyst exhibited an onset potential of 1.43 V and an overpotential at 10 mA cm−2 of 331 mV, much smaller than those of atomic Co (402 mV) and atomic Fe (488 mV) (Fig. 19C). The Tafel slope for atomic Ni was determined to be 63 mV dec−1, smaller than those of atomic Co (80 mV dec−1), atomic Fe (164 mV dec−1) and metal-free catalysts (175 mV dec−1) (Fig. 19D). Stability tests by both electrochemical measurements and structural characterization of XAFS suggested that the metal atom and its coordinating matrix were robust enough to withstand the OER operation condition. Atomic Mn in nitrogen-doped graphene has recently been demonstrated to exhibit superior OER activity with intrinsic activity close to that of nature's catalyst CaMn4O5 in photosystem II and much higher than those of previously reported Mn-based OER catalysts.74
image file: c9cs00422j-f19.tif
Fig. 19 Structural and electrocatalytic characterization of atomic metal catalysts toward OER. (A) Atomic structure characterization of Ni, Fe, and Co single atoms supported on nitrogen-graphene by HAADF-STEM. (B) Proposed reaction scheme with the intermediates having the optimized geometry of the single-site (top) and dual-site mechanisms (bottom) toward OER. (C) OER polarization curves of different atomic catalysts along with control samples. (D) Tafel plots for different atomic metal catalysts and metal-free nitrogen-doped graphene. Reproduced with permission from Fei et al.25 Copyright 2018 Nature Publishing Group.

5.4. CO2 reduction reaction

Electrochemical reduction of CO2 to value-added molecules such as carbon monoxide (CO) and hydrocarbons is an effective strategy to remit the environmental and energy issues. However, significant overpotentials are typically needed to overcome the activation barrier of CO2 reduction due to its sluggish kinetics; moreover, there are wide distributions of products (e.g., CO, HCHO, CH3OH, CH4) and competitive side reactions such as HER.153 Therefore, developing highly active, selective and stable electrocatalysts is of paramount significance to realize CO2RR for practical applications. Many types of materials, including metals, metal oxides, carbon-based materials, and molecular compounds, have been shown to be effective CO2RR electrocatalysts and their performances have been greatly enhanced by optimizing the size, structure, and composition.154–156

Recently, first-row transition metals (e.g., Zn, Mn, Fe, Co, Ni, Cu, Sn) atomically dispersed in graphene layers have been explored as CO2RR electrocatalysts in aqueous solution with CO as the dominant product and Ni-based ones are the most widely studied among different atomic metal catalysts (Table 2).46,73,99,157–172 For example, using the pyrolysis method with Ni-pentaethylenehexamine and GO as precursors, Su et al. prepared atomic Ni embedded in nitrogen-doped graphene that exhibited a high activity for CO formation in neutral solution with faradaic efficiency exceeding 90% at −0.7 V to −0.9 V.157 The competitive HER was greatly suppressed, in strong contrast to the metal-free nitrogen-doped graphene and Ni metal electrode. The NiNx moiety was proposed to be the active center based on the EXAFS-FT analysis as well as from the observation that the replacement of Ni by Cu resulted in significantly lower CO2RR activity and that the faradaic efficiency of CO did not decrease after lowering the concentration of the NiNx moiety. With a similar synthetic strategy, Jiang et al. reported another atomic Ni catalyst that presented a high CO selectivity of 95% at an overpotential of 0.55 V and long-term stability over 20 hours’ continuous electrolysis in 0.5 M KHCO3.73 The atomic dispersion of Ni and a mixed Ni–C/Ni–N coordination environment within the catalyst were suggested by XANES and EXAFS analyses. The active sites were proposed to be associated with the NiNx moiety by the control experiments on Ni-coordinated graphene with no nitrogen incorporation, nitrogen-doped graphene and Ni3N catalysts, all exhibiting significantly lower CO2RR activity and CO2-to-CO selectivity. The critical role of the Ni atoms in forming the active centers was confirmed by the catalytic evaluation on a variety of other atomic transition metals (e.g., Fe, Co, Mn, Cu) synthesized by the same approach, which showed dramatically decreased selectivity. The use of GO as the pyrolysis precursor to prepare CO2RR-active atomic catalysts was recently extended to the Fe–N–C catalyst by Tour and coworkers.173 They found that the CO2RR activity of Fe–N–C was highly dependent on the pyrolysis temperature and the sample prepared at 750 °C had predominant atomic Fe dispersion (evidenced by EXAFS-FT) that resulted in the optimal activity of a maximum faradaic efficiency of 80% toward CO formation at −0.6 V. Moreover, the formation of Fe3O4 and metallic Fe nanoparticles resulting from the increase in the Fe loading would lead to lower selectivity (<10%) toward CO production, consistent with the study by Huan et al.174

Table 2 Single atom catalysts supported on graphene layers or graphitized carbon for electrocatalytic CO2 reduction
Catalyst Synthetic method Metal content [wt%] Catalyst loading [mg cm−2] Electrolyte Potential Current density [mA cm−2] Main product (faradaic efficiency) Ref.
A-Ni-NSG Pyrolysis 2.8 by ICP 0.4 0.5 M KHCO3 −0.72 V vs. RHE 23.5 CO (94%) 46
Ni–N–C Pyrolysis 0.24 by ICP 0.5 0.5 M KHCO3 −0.77 V vs. RHE ∼7.5 CO (92%) 61
Ni-NG Pyrolysis 2 by XPS 1 0.5 M KHCO3 −0.74 V vs. RHE 11 CO (95%) 73
Ni2+@NG Ion-adsorption 0.8 by ICP 0.3 0.5 M KHCO3 −0.68 V vs. RHE 10.2 CO (92%) 99
Ni–N–Gr Pyrolysis 2.2 by ICP 0.3 0.1 M KHCO3 −0.8 V vs. RHE ∼1.5 CO (90%) 157
Ni SAs/N–C Pyrolysis 1.53 by ICP 0.1 0.5 M KHCO3 −1.0 V vs. RHE 10.5 CO (70.3%) 158
Ni–N4–C Pyrolysis 1.41 by ICP 0.2 0.5 M KHCO3 −0.81 V vs. RHE 28.6 CO (99%) 159
NiN-GS Pyrolysis ∼5 by XPS 1 0.5 M KHCO3 −0.76 V vs. RHE 21.3 CO (96%) 160
C-Zn1Ni4 ZIF-8 Pyrolysis 5.44 by ICP 2 0.5 M KHCO3 −1.03 V vs. RHE 71.5 CO (93.5%) 161
H-CP Solid-state diffusion N/A 3.5 0.5 M KHCO3 −1.2 V vs. RHE 60.11 CO (90.8%) 166
Ni SAs/NCNTs Pyrolysis 6.63 by ICP 0.8 0.5 M KHCO3 −1 V vs. RHE 57.1 CO (95%) 167
Ni–N-MEGO Pyrolysis 6.7 by ICP 0.5 0.5 M KHCO3 −0.7 V vs. RHE 29.1 CO (92%) 170
Ni-NCB Pyrolysis 0.27 by ICP 0.2 0.5 M KHCO3 −0.68 V vs. RHE 6.8 CO (99%) 171
Ni–N–C Pyrolysis 1.8 by XPS 0.75 0.1 M KHCO3 −0.84 V vs. RHE 15 CO (81%) 175
Ni–N–C Pyrolysis 1.73 by XPS 0.76 0.1 M KHCO3 −0.75 V vs. RHE 11.2 CO (80%) 176
Co–N2 Pyrolysis 0.25 by ICP 0.8 0.5 M KHCO3 −0.63 V vs. RHE 18.4 CO (94%) 45
Co1–N4 Pyrolysis 0.6 by ICP N/A 0.1 M KHCO3 −0.8 V vs. RHE ∼15 CO (82%) 177
Co–N5/HNPCSs Pyrolysis 3.54 by ICP N/A 0.2 M NaHCO3 −0.79 V vs. RHE 10.2 CO (99.3%) 178
Fe3+–N–C Pyrolysis 2.8 by ICP 0.6 0.5 M KHCO3 −0.47 V vs. RHE ∼21 CO (∼95%) 128
Fe–N–C Pyrolysis 0.5 by XPS 0.6 0.1 M KHCO3 −0.68 V vs. RHE 2.8 CO (93%) 163
Fe/NG-750 Pyrolysis 1.25 by ICP 0.32 0.1 M KHCO3 −0.6 V vs. RHE ∼3 CO (80%) 173
Fe0.5d Pyrolysis 1.5 by XAS 1 0.1 M KHCO3 −0.6 V vs. RHE 4.5 CO (91%) 174
ZnNx/C Pyrolysis 0.1 by ICP 1 0.5 M KHCO3 −0.43 V vs. RHE 4.8 CO (95%) 164
Zn–N-G-800 Pyrolysis ∼1.6 by ICP 2 0.5 M KHCO3 −0.5 V vs. RHE 4.1 CO (91%) 172
(Cl,N)-Mn/G Pyrolysis 0.05 by ICP 0.5 0.5 M KHCO3 −0.6 V vs. RHE 13.4 CO (97%) 168
Sn1δ+-NG Pyrolysis 0.82 by ICP 0.24 0.25 M KHCO3 −1.6 V vs. SCE 11.7 Formate (75.1%) 165

In addition to GO, MOF and molecular/polymeric precursors have also been frequently applied in the pyrolysis method for the preparation of M–N–C catalysts toward CO2RR.46,158–161,175 With handful choices of molecular precursors for the pyrolysis synthesis of G-SACs, multiple dopants can be simultaneously introduced in the carbon substrate to tune the electronic structure of the metal center and thus the CO2RR activity. Yang et al. prepared atomically dispersed Ni on nitrogen-doped graphene (A-Ni-NG) and sulfur-, nitrogen-codoped graphene (A-Ni-NSG) by pyrolyzing a mixture of amino acid (L-alanine or L-cysteine), melamine and Ni acetate.46 The resulting composite possessed a two-dimensional structure, and the Ni existed as isolated atoms on the graphene substrate with no Ni–Ni bonding, confirmed by the HAADF-STEM image and FT-EXAFS spectra (Fig. 20A). When tested as CO2RR catalysts (Fig. 20B), A-Ni-NSG showed an earlier onset potential than A-Ni-NG and both of them delivered considerably higher current density than control samples of nitrogen-doped graphene (NG) and Ni nanoparticle decorated NG (Ni-NG). The enhancement from sulfur doping was attributed to the distorted coordination geometry around the Ni center in A-Ni-NSG owing to the non-centrosymmetric ligand strength, which was expected to promote better adsorption of reactants and intermediates on the catalyst surface. The two single Ni atom catalysts maintain over 80% selectivity toward CO production, even at very negative applied potentials (Fig. 20C). Moreover, the A-Ni-NSG catalyst could deliver stable performance, retaining 98% of the initial current after 100 hours operation (Fig. 20D). The active site was identified as a monovalent Ni atomic center with a d9 electronic configuration based on operando X-ray absorption and photoelectron spectroscopy measurements.

image file: c9cs00422j-f20.tif
Fig. 20 Structural and CO2RR catalytic characterization of the atomic Ni catalyst supported on N- and S-codoped graphene. (A) FT-EXAFS spectra of the atomic Ni catalyst along with reference samples. (B) Polarization curves in CO2-saturated 0.5 M KHCO3 solution. (C) CO faradaic efficiency at various applied potentials. (D) Current–time response at constant potential. Reproduced with permission from Yang et al.46 Copyright 2018 Nature Publishing Group.

Though Co–N–C catalysts with atomic CoNx moieties were most frequently investigated as HER electrocatalysts and studies focusing on elucidating the correlation between the type of metal center and CO2RR activity of M–N–C catalysts have suggested that Co–N–C catalysts generally exhibit low faradaic efficiency for CO formation compared to Ni-based and Fe-based counterparts,176 it was indicated by recent reports that their selectivity toward CO formation and H2 generation can be tuned by altering the coordination environment of the Co center.45,163,177,178 For example, Wang et al. synthesized a series of atomic Co catalysts by pyrolyzing Co/Zn MOFs and the nitrogen coordination number of the Co center was regulated by the variation of pyrolysis temperature. Specifically, an increase in pyrolysis temperature would lead to a decrease in N coordinators due to the volatile nature of C–N fragments, generating three different types of CoNx moieties (i.e., Co–N2, Co–N3 and Co–N4), as determined by EXAFS fitting (Fig. 4C). As shown by the linear sweep voltammetry (Fig. 21A), the overall current density was highest in Co–N2 and it reached 18.1 mA cm−2 at −0.63 V, which was significantly higher than those for Co–N3, Co–N4 and Co nanoparticles. Analysis of product distribution by gas chromatography and 1H nuclear magnetic resonance confirmed that CO and H2 were the main products with no other gas products or liquid products detected for all catalysts studied. The CO faradaic efficiency for Co–N2 reached up to 95% at −0.68 V, higher than the maximum value for Co–N3 (63% at −0.53 V), Co–N4 (below 5% at −0.83 V to −1.03 V) and Co nanoparticles (below 7% across all applied potentials) (Fig. 21B). Experimental and DFT calculations suggested that the high performance of the Co–N2 catalyst could be ascribed to its low coordination number that facilitated the activation of CO2 to the CO2˙ intermediate.

image file: c9cs00422j-f21.tif
Fig. 21 CO2RR activity and selectivity of Co–N2, Co–N3, Co–N4 and Co NPs. (A) Linear sweep voltammetry of the different catalysts evaluated in 0.5 M KHCO3 with pure carbon paper as background. (B) CO faradaic efficiencies at different applied potentials of the different catalysts. Reproduced with permission from Wang et al.45 Copyright 2018 Wiley-VCH.

5.5. N2 reduction reaction

NH3 is one of the most highly produced chemicals that is widely used in agriculture and industry applications.179 Currently, NH3 is primarily synthesized by the Haber–Bosch process with heterogeneous Fe-based catalysts at high temperature and pressure, accounting for ∼1.5% of global energy consumption and generating 300 million metric tons of fossil fuel-derived CO2 annually.180 The electrochemical reduction of N2, ideally under ambient conditions with H2O as the hydrogen source, has been proposed as a sustainable alternative for NH3 synthesis.181 However, the development of the electrochemical approach has been impeded by the lack of efficient electrocatalysts with high activity and selectivity over hydrogen generation. Owing to the recent research efforts, some electrocatalysts for N2 reduction have been developed, including noble metals, nonprecious metal-based materials, and nanocarbons.180–182 Recently, Geng et al. reported that Ru single atoms (0.18 wt%) on nitrogen-doped carbon (denoted as Ru SAs/N–C) can work as an efficient electrocatalyst toward electrochemical N2 reduction.183 The catalyst was synthesized by pyrolyzing the Ru-containing derivative of zeolitic imidazolate frameworks (Fig. 22A). XANES and EXAFS studies suggested that the Ru was atomically dispersed in the carbon matrix with coordination to N atoms and the valence state of Ru species was ≈+3. The electrochemical measurements were performed in a three-electrode system with a two-compartment cell in 0.05 M H2SO4. The partial current for NH3 production (determined by the indophenol blue method) of Ru SAs/N–C was significantly higher than that of the control sample Ru NPs/N–C at all applied potentials (Fig. 22B). At −0.2 V versus RHE, Ru SAs/N–C achieved a faradaic efficiency of 29.6% for NH3 production, which was about 2-fold higher than that of Ru NPs/N–C (Fig. 22C). When normalized by the weight of catalysts, Ru SAs/N–C exhibited a yield rate of 120.9 μgNH3 mg−1 h−1 at −0.2 V (Fig. 22D). It should be cautioned that without an elaborated calibration and control, the accurate quantification of NH3 production could be plagued by possible contamination in the laboratory environment, purging gases, electrolyte solvents or the catalysts N2, electrolytes or the catalysts. To this end, the authors conducted a series of control and confirmatory experiments, including open circuit control measurements, argon control measurements and 15N2 reduction experiments, to verify that the NH3 production over Ru SAs/N–C is indeed originated from N2 electrochemical reduction. Electrochemical stability tests revealed that Ru SAs/N–C experienced less than 7% decay of the yield rate for NH3 reduction during a 12 h potentiostatic test at −0.2 V (Fig. 22E). In another work, Mo single metals (evidenced by EXAFS-FT) supported on N-doped carbons were reported as efficient and stable electrocatalysts toward N2 reduction with high NH3 production yield and faradaic efficiency.184
image file: c9cs00422j-f22.tif
Fig. 22 Synthetic route and NRR electrocatalytic characterization of Ru SAs/N–C. (A) Scheme of the synthetic procedure for Ru SAs/N–C. (B) Current densities for NH3 production of Ru SAs/N–C and Ru NPs/N–C. (C) Faradaic efficiency and (D) yield rate of NH3 production at different applied potentials on Ru SAs/N–C and Ru NPs/N–C. (E) 12 h durability test for Ru SAs/N–C toward N2 electrochemical reduction at −0.2 V versus RHE. Reproduced with permission from Geng et al.183 Copyright 2018 Wiley-VCH.

5.6. Other electrochemical reactions

Bao and co-workers reported the use of G-SACs as an alternative counter electrode to Pt electrode for dye-sensitized solar cells (DSSCs).104 A series of single atomic metals (Mn, Fe, Co, Ni and Cu) in the form of MN4 moieties confined in the basal plane of graphene were synthesized by ball milling of metal phthalocyanine and graphene nanosheets, and the atomic dispersion of the metals was confirmed by HAADF-STEM and EXAFS-FT. Among the different metals, CoN4 can work as a highly active and stable catalyst facilitating the interconversion of the redox couple I/I3. Further, a solar cell device utilizing CoN4 as the counter electrode was fabricated which displayed a higher power conversion efficiency than the Pt counterpart. DFT calculations suggested that the resulting superior catalytic properties of CoN4 could be ascribed to the desired adsorption energy of iodine on the single Co sites.

Sun and co-workers employed the ALD technique to synthesize single atoms and sub-nanometer clusters of Pt on the surfaces of defective graphene nanosheets and studied their catalytic behaviors toward methanol oxidation reaction (MOR), which represents the key reaction of the direct methanol fuel cell technology.106 The morphology, size, density and loading of Pt on graphene could be well regulated by tuning the number of ALD cycles, which allowed the investigation of the effects of different Pt species on catalytic properties. Catalytic measurements revealed that these catalysts had much higher activity for methanol oxidation and superior tolerance than commercial Pt/C and the excellent catalytic performance was attributed to the low-coordination and partially unoccupied densities of states of Pt atoms’ 5d orbitals.

6. Conclusion and outlook

The development of efficient and cost-effective catalysts is of great interest and importance for clean energy applications. Benefiting from the unique physicochemical properties of the graphene substrate, G-SACs represent a unique class of single atom catalysts compared to those supported on metal oxide-based substrates for catalyzing electrochemical energy conversion processes including hydrogen evolution, oxygen reduction, oxygen evolution, CO2 reduction, N2 reduction and methanol oxidation. These G-SACs have been prepared by a variety of strategies including electron/ion irradiation, atomic layer deposition, ball milling, photochemical reduction, solution-phase synthesis, and pyrolysis using different types of precursors including molecules, polymers, MOFs and GO. In addition, considerable research efforts have been devoted to establishing the correlation between the structure and catalytic properties of G-SACs by applying the concept of ligand-field effects adapted from coordination chemistry since single atoms embedded in graphene can be considered as the mimics of organometallic complexes with the metal center and neighboring coordinators accommodated in the π-conjugated carbon matrix. This aspect has been demonstrated by several experimental and theoretical studies where the regulations in the identity of the metal center and coordinators, coordination configuration, the electronic structure of the carbon substrate and the location of the active site have resulted in the effective tuning of the catalytic behaviors of G-SACs. The foundation for the establishment of the structure-to-activity linkage in G-SACs lies in the identification of the active site structure. Among the various characterization techniques, XAS analyses comprising EXAFS and XANES are powerful tools for determining the geometric and electronic structure of the active sites, such as the oxidation state of the metal center, bond length, bond disorder, coordination number and geometry. Despite the excitement and achievements brought by the research on G-SACs for electrochemical energy conversion to date, considerable challenges and opportunities remain from both fundamental and applied research perspectives.

From the outset, the achievement of high metal loading in G-SACs with exclusive single atomic dispersion remains the most critical challenge for the field. The formation of strong interactions between the metal center and the neighboring coordinators typically involves the processes of bond breaking and reforming, which require the input of high energy during catalyst preparation (e.g., pyrolysis or ball milling). These activation processes could simultaneously cause the metal atoms to diffuse and agglomerate due to their large surface free energy and therefore the amounts of metals added during preparation need to be sufficiently low to avoid aggregation. Thus far, the metal contents in most studies are typically below 5 wt%, which greatly limits the number of active sites and performance of G-SACs to meet practical applications. Besides the nominal metal loading, another important consideration is that the single metal atoms should be preferentially exposed at the surface so as to actively participate in catalytic reaction. To increase the metal loading of G-SACs, defect engineering of graphene could potentially be explored to introduce additional doping sites for both metal atoms and heteroatoms. Alternatively, modification of existing synthetic strategies and development of novel methods that can minimize the exposure of precursor materials to the high-energy environment could minimize aggregation while increasing atomic metal loading.

Second, despite some pioneering attempts, it remains difficult to correlate the exact structure configuration with the catalytic properties, which further precludes the establishment of design principles in optimizing the intrinsic activity of G-SACs. The reasons behind this difficulty are multifold. Firstly, the heterogeneous structure with the co-presence of atomic sites and metal aggregates, and the nonuniform distribution of different types of atomic sites severely complicate the identification of active sites as well as the origin of catalytic activity. Secondly, there is a lack of efficient synthetic methods that can systematically modulate the metal center and its coordination environment in a tunable and predictable manner. For most methods developed for synthesizing G-SACs to date, any given method can usually generate only one or a few specified structures and it remains unclear how the adopted structures are correlated with the synthetic conditions. In addition, most studies of G-SACs have focused on atomic MNx moieties, while other heteroatoms (e.g., B, O, S) are less investigated either as coordinators to the metal center or as dopants in the carbon matrix. The incorporation of these multiple dopants could provide new avenues to tune the electronic structure and therefore the intrinsic activity of the metal center. Additionally, the comparison of the intrinsic activity among different types of atomic sites prepared by different methods is not straightforward considering that several structural characteristics (e.g., electrical conductivity, porosity, surface area and wettability) beyond the active site structure can influence the catalytic activity of G-SACs and are highly sensitive to the synthetic routes. Therefore, the development of a universal method to synthesize G-SACs with monodispersed and systematically tailorable active sites is highly desirable for unraveling the fundamental correlation between local atomic structures and catalytic properties.

Lastly, there is little experimental insight into the metal–reactant and metal–support interactions in G-SACs under electrocatalytic conditions and a limited understanding of the mechanistic details of the reactions on the catalyst surface. As large differences could exist in the geometric and electronic structure of the active metal during electrochemical reaction from that derived under ex situ conditions, in situ/operando studies will play important roles in properly unveiling the real active site and the dynamic process for various catalytic processes. The precise understanding of the reaction mechanism is pivotal for the rational design of next generation of high-performing G-SACs for practical technological applications. Despite many challenges ahead, we have witnessed an enormous increase of research efforts in the relevant areas in the past few years. The continued efforts in synthetic design and structural/mechanistic studies will surely help accelerate the development of G-SACs with high activity, selectivity and stability for application in a wide range of electrochemical reactions as well as other types of catalytic processes.

Conflicts of interest

There are no conflicts to declare.


X. D. acknowledges support from the National Science Foundation (Grant No. 1800580). Y. H. acknowledges support from the Office of Naval Research (grant N000141812155). J. D. acknowledges support from the National Natural Science Foundation of China (Grant No. 11605225) and Youth Innovation Promotion Association CAS.


  1. V. R. Stamenkovic, D. Strmcnik, P. P. Lopes and N. M. Markovic, Nat. Mater., 2016, 16, 57–69 CrossRef PubMed.
  2. Z. W. Seh, J. Kibsgaard, C. F. Dickens, I. Chorkendorff, J. K. Nørskov and T. F. Jaramillo, Science, 2017, 355, eaad4998 CrossRef PubMed.
  3. C. Copéret, M. Chabanas, R. Petroff Saint-Arroman and J.-M. Basset, Angew. Chem., Int. Ed., 2003, 42, 156–181 CrossRef PubMed.
  4. S. Hübner, J. G. de Vries and V. Farina, Adv. Synth. Catal., 2016, 358, 3–25 CrossRef.
  5. S. Zhao, Y. Wang, J. Dong, C.-T. He, H. Yin, P. An, K. Zhao, X. Zhang, C. Gao, L. Zhang, J. Lv, J. Wang, J. Zhang, A. M. Khattak, N. A. Khan, Z. Wei, J. Zhang, S. Liu, H. Zhao and Z. Tang, Nat. Energy, 2016, 1, 16184 CrossRef CAS.
  6. B. C. Gates, Chem. Rev., 1995, 95, 511–522 CrossRef CAS.
  7. X.-F. Yang, A. Wang, B. Qiao, J. Li, J. Liu and T. Zhang, Acc. Chem. Res., 2013, 46, 1740–1748 CrossRef CAS PubMed.
  8. J. Liu, ACS Catal., 2017, 7, 34–59 CrossRef CAS.
  9. A. Wang, J. Li and T. Zhang, Nat. Rev. Chem., 2018, 2, 65–81 CrossRef CAS.
  10. X. Cui, W. Li, P. Ryabchuk, K. Junge and M. Beller, Nat. Catal., 2018, 1, 385–397 CrossRef CAS.
  11. R. Qin, P. Liu, G. Fu and N. Zheng, Small Methods, 2017, 2, 1700286 CrossRef.
  12. M. Hu, J. Zhang, W. Zhu, Z. Chen, X. Gao, X. Du, J. Wan, K. Zhou, C. Chen and Y. Li, Nano Res., 2018, 11, 905–912 CrossRef CAS.
  13. H. Zhang, G. Liu, L. Shi and J. Ye, Adv. Energy Mater., 2017, 8, 1701343 CrossRef.
  14. C. Zhu, S. Fu, Q. Shi, D. Du and Y. Lin, Angew. Chem., Int. Ed., 2017, 56, 13944–13960 CrossRef CAS PubMed.
  15. M. Dhiman and V. Polshettiwar, ChemCatChem, 2017, 10, 881–906 CrossRef.
  16. B. Bayatsarmadi, Y. Zheng, A. Vasileff and S.-Z. Qiao, Small, 2017, 13, 1700191 CrossRef PubMed.
  17. Y. Peng, B. Lu and S. Chen, Adv. Mater., 2018, 30, 1801995 CrossRef PubMed.
  18. B. Long, Y. Tang and J. J. N. R. Li, Nano Res., 2016, 9, 3868–3880 CrossRef CAS.
  19. J. Liang, Q. Yu, X. Yang, T. Zhang and J. Li, Nano Res., 2018, 11, 1599–1611 CrossRef CAS.
  20. K. S. Novoselov, V. I. Fal’ko, L. Colombo, P. R. Gellert, M. G. Schwab and K. Kim, Nature, 2012, 490, 192–200 CrossRef CAS PubMed.
  21. H. Wang, H. S. Casalongue, Y. Liang and H. Dai, J. Am. Chem. Soc., 2010, 132, 7472–7477 CrossRef CAS PubMed.
  22. S. Hu, T. Han, C. Lin, W. Xiang, Y. Zhao, P. Gao, F. Du, X. Li and Y. Sun, Adv. Funct. Mater., 2017, 27, 1700041 CrossRef.
  23. P. Kundu, C. Nethravathi, P. A. Deshpande, M. Rajamathi, G. Madras and N. Ravishankar, Chem. Mater., 2011, 23, 2772–2780 CrossRef CAS.
  24. M. Liu, R. Zhang and W. Chen, Chem. Rev., 2014, 114, 5117–5160 CrossRef CAS PubMed.
  25. H. Fei, J. Dong, Y. Feng, C. S. Allen, C. Wan, B. Volosskiy, M. Li, Z. Zhao, Y. Wang, H. Sun, P. An, W. Chen, Z. Guo, C. Lee, D. Chen, I. Shakir, M. Liu, T. Hu, Y. Li, A. I. Kirkland, X. Duan and Y. Huang, Nat. Catal., 2018, 1, 63–72 CrossRef CAS.
  26. J. Luo, D. A. Vermaas, D. Bi, A. Hagfeldt, W. A. Smith and M. Grätzel, Adv. Energy Mater., 2016, 6, 1600100 CrossRef.
  27. W. I. Choi, B. C. Wood, E. Schwegler and T. Ogitsu, Adv. Energy Mater., 2015, 5, 1501423 CrossRef.
  28. A. V. Krasheninnikov, P. O. Lehtinen, A. S. Foster, P. Pyykkö and R. M. Nieminen, Phys. Rev. Lett., 2009, 102, 126807 CrossRef CAS PubMed.
  29. H. Wang, Q. Wang, Y. Cheng, K. Li, Y. Yao, Q. Zhang, C. Dong, P. Wang, U. Schwingenschlögl, W. Yang and X. X. Zhang, Nano Lett., 2012, 12, 141–144 CrossRef CAS PubMed.
  30. A. W. Robertson, B. Montanari, K. He, J. Kim, C. S. Allen, Y. A. Wu, J. Olivier, J. Neethling, N. Harrison, A. I. Kirkland and J. H. Warner, Nano Lett., 2013, 13, 1468–1475 CrossRef CAS PubMed.
  31. J. Zhao, Q. Deng, S. M. Avdoshenko, L. Fu, J. Eckert and M. H. Rümmeli, PNAS, 2014, 111, 15641 CrossRef CAS PubMed.
  32. J. Duan, S. Chen, M. Jaroniec and S. Z. Qiao, ACS Catal., 2015, 5, 5207–5234 CrossRef CAS.
  33. D. Deng, X. Chen, L. Yu, X. Wu, Q. Liu, Y. Liu, H. Yang, H. Tian, Y. Hu, P. Du, R. Si, J. Wang, X. Cui, H. Li, J. Xiao, T. Xu, J. Deng, F. Yang, P. N. Duchesne, P. Zhang, J. Zhou, L. Sun, J. Li, X. Pan and X. Bao, Sci. Adv., 2015, 1, e1500462 CrossRef PubMed.
  34. F. Vines, J. R. B. Gomes and F. Illas, Chem. Soc. Rev., 2014, 43, 4922–4939 RSC.
  35. P. Strasser, S. Koh, T. Anniyev, J. Greeley, K. More, C. Yu, Z. Liu, S. Kaya, D. Nordlund, H. Ogasawara, M. F. Toney and A. Nilsson, Nat. Chem., 2010, 2, 454–460 CrossRef CAS PubMed.
  36. G. Chen, C. Xu, X. Huang, J. Ye, L. Gu, G. Li, Z. Tang, B. Wu, H. Yang, Z. Zhao, Z. Zhou, G. Fu and N. Zheng, Nat. Mater., 2016, 15, 564–569 CrossRef CAS PubMed.
  37. V. R. Stamenkovic, B. S. Mun, M. Arenz, K. J. J. Mayrhofer, C. A. Lucas, G. Wang, P. N. Ross and N. M. Markovic, Nat. Mater., 2007, 6, 241–247 CrossRef CAS PubMed.
  38. D. J. Gorin, B. D. Sherry and F. D. Toste, Chem. Rev., 2008, 108, 3351–3378 CrossRef CAS PubMed.
  39. H. Xu, D. Cheng, D. Cao and X. C. Zeng, Nat. Catal., 2018, 1, 339–348 CrossRef CAS.
  40. V. Tripkovic, M. Vanin, M. Karamad, M. E. Björketun, K. W. Jacobsen, K. S. Thygesen and J. Rossmeisl, J. Phys. Chem. C, 2013, 117, 9187–9195 CrossRef CAS.
  41. S. Back, J. Lim, N.-Y. Kim, Y.-H. Kim and Y. Jung, Chem. Sci., 2017, 8, 1090–1096 RSC.
  42. C. Choi, S. Back, N.-Y. Kim, J. Lim, Y.-H. Kim and Y. Jung, ACS Catal., 2018, 8, 7517–7525 CrossRef CAS.
  43. A. Zitolo, V. Goellner, V. Armel, M.-T. Sougrati, T. Mineva, L. Stievano, E. Fonda and F. Jaouen, Nat. Mater., 2015, 14, 937–942 CrossRef CAS PubMed.
  44. A. Zitolo, N. Ranjbar-Sahraie, T. Mineva, J. Li, Q. Jia, S. Stamatin, G. F. Harrington, S. M. Lyth, P. Krtil, S. Mukerjee, E. Fonda and F. Jaouen, Nat. Commun., 2017, 8, 957 CrossRef PubMed.
  45. X. Wang, Z. Chen, X. Zhao, T. Yao, W. Chen, R. You, C. Zhao, G. Wu, J. Wang, W. Huang, J. Yang, X. Hong, S. Wei, Y. Wu and Y. Li, Angew. Chem., Int. Ed., 2018, 57, 1944–1948 CrossRef CAS PubMed.
  46. H. B. Yang, S.-F. Hung, S. Liu, K. Yuan, S. Miao, L. Zhang, X. Huang, H.-Y. Wang, W. Cai, R. Chen, J. Gao, X. Yang, W. Chen, Y. Huang, H. M. Chen, C. M. Li, T. Zhang and B. Liu, Nat. Energy, 2018, 3, 140–147 CrossRef CAS.
  47. Y. Guo, P. Yuan, J. Zhang, Y. Hu, I. S. Amiinu, X. Wang, J. Zhou, H. Xia, Z. Song, Q. Xu and S. Mu, ACS Nano, 2018, 12, 1894–1901 CrossRef CAS PubMed.
  48. Y. Hou, M. Qiu, M. G. Kim, P. Liu, G. Nam, T. Zhang, X. Zhuang, B. Yang, J. Cho, M. Chen, C. Yuan, L. Lei and X. Feng, Nat. Commun., 2019, 10, 1392 CrossRef PubMed.
  49. Y. Han, Y. Wang, R. Xu, W. Chen, L. Zheng, A. Han, Y. Zhu, J. Zhang, H. Zhang, J. Luo, C. Chen, Q. Peng, D. Wang and Y. Li, Energy Environ. Sci., 2018, 11, 2348–2352 RSC.
  50. H. C. Kwon, M. Kim, J. P. Grote, S. J. Cho, M. W. Chung, H. Kim, D. H. Won, A. R. Zeradjanin, K. J. J. Mayrhofer, M. Choi, H. Kim and C. H. Choi, J. Am. Chem. Soc., 2018, 140, 16198–16205 CrossRef CAS PubMed.
  51. H. T. Chung, D. A. Cullen, D. Higgins, B. T. Sneed, E. F. Holby, K. L. More and P. Zelenay, Science, 2017, 357, 479 CrossRef CAS PubMed.
  52. N. Ramaswamy, U. Tylus, Q. Jia and S. Mukerjee, J. Am. Chem. Soc., 2013, 135, 15443–15449 CrossRef CAS PubMed.
  53. X. Fu, N. Li, B. Ren, G. Jiang, Y. Liu, F. M. Hassan, D. Su, J. Zhu, L. Yang, Z. Bai, Z. P. Cano, A. Yu and Z. Chen, Adv. Energy Mater., 2019, 9, 1803737 CrossRef.
  54. Y. Wang, D. Adekoya, J. Sun, T. Tang, H. Qiu, L. Xu, S. Zhang and Y. Hou, Adv. Funct. Mater., 2018, 29, 1807485 CrossRef.
  55. R. Jiang, L. Li, T. Sheng, G. Hu, Y. Chen and L. Wang, J. Am. Chem. Soc., 2018, 140, 11594–11598 CrossRef CAS PubMed.
  56. K. Liu, H. Zhong, F. Meng, X. Zhang, J. Yan and Q. Jiang, Mater. Chem. Front., 2017, 1, 2155–2173 RSC.
  57. P. Chen, T. Zhou, L. Xing, K. Xu, Y. Tong, H. Xie, L. Zhang, W. Yan, W. Chu, C. Wu and Y. Xie, Angew. Chem., Int. Ed., 2017, 56, 610–614 CrossRef CAS PubMed.
  58. D. Liu, C. Wu, S. Chen, S. Ding, Y. Xie, C. Wang, T. Wang, Y. A. Haleem, Z. ur Rehman, Y. Sang, Q. Liu, X. Zheng, Y. Wang, B. Ge, H. Xu and L. Song, Nano Res., 2018, 11, 2217–2228 CrossRef CAS.
  59. G. Zhang, Y. Jia, C. Zhang, X. Xiong, K. Sun, R. Chen, W. Chen, Y. Kuang, L. Zheng, H. Tang, W. Liu, J. Liu, X. Sun, W.-F. Lin and H. Dai, Energy Environ. Sci., 2019, 12, 1317–1325 RSC.
  60. T. Sun, S. Zhao, W. Chen, D. Zhai, J. Dong, Y. Wang, S. Zhang, A. Han, L. Gu, R. Yu, X. Wen, H. Ren, L. Xu, C. Chen, Q. Peng, D. Wang and Y. Li, Proc. Natl. Acad. Sci. U. S. A., 2018, 115, 12692 CrossRef CAS PubMed.
  61. X.-M. Hu, H. H. Hval, E. T. Bjerglund, K. J. Dalgaard, M. R. Madsen, M.-M. Pohl, E. Welter, P. Lamagni, K. B. Buhl, M. Bremholm, M. Beller, S. U. Pedersen, T. Skrydstrup and K. Daasbjerg, ACS Catal., 2018, 8, 6255–6264 CrossRef CAS.
  62. L. Zhao, Y. Zhang, L.-B. Huang, X.-Z. Liu, Q.-H. Zhang, C. He, Z.-Y. Wu, L.-J. Zhang, J. Wu, W. Yang, L. Gu, J.-S. Hu and L.-J. Wan, Nat. Commun., 2019, 10, 1278 CrossRef PubMed.
  63. K. Yuan, S. Sfaelou, M. Qiu, D. Lützenkirchen-Hecht, X. Zhuang, Y. Chen, C. Yuan, X. Feng and U. Scherf, ACS Energy Lett., 2018, 3, 252–260 CrossRef CAS.
  64. Y. Chen, S. Ji, S. Zhao, W. Chen, J. Dong, W.-C. Cheong, R. Shen, X. Wen, L. Zheng, A. I. Rykov, S. Cai, H. Tang, Z. Zhuang, C. Chen, Q. Peng, D. Wang and Y. Li, Nat. Commun., 2018, 9, 5422 CrossRef PubMed.
  65. H. Shen, E. Gracia-Espino, J. Ma, K. Zang, J. Luo, L. Wang, S. Gao, X. Mamat, G. Hu, T. Wagberg and S. Guo, Angew. Chem., Int. Ed., 2017, 56, 13800–13804 CrossRef CAS PubMed.
  66. C. Zhang, J. Sha, H. Fei, M. Liu, S. Yazdi, J. Zhang, Q. Zhong, X. Zou, N. Zhao, H. Yu, Z. Jiang, E. Ringe, B. I. Yakobson, J. Dong, D. Chen and J. M. Tour, ACS Nano, 2017, 11, 6930–6941 CrossRef CAS PubMed.
  67. D. C. Marcano, D. V. Kosynkin, J. M. Berlin, A. Sinitskii, Z. Sun, A. Slesarev, L. B. Alemany, W. Lu and J. M. Tour, ACS Nano, 2010, 4, 4806–4814 CrossRef CAS PubMed.
  68. K. Chen, S. Song, F. Liu and D. Xue, Chem. Soc. Rev., 2015, 44, 6230–6257 RSC.
  69. L. Dai, Acc. Chem. Res., 2013, 46, 31–42 CrossRef CAS PubMed.
  70. I. K. Moon, J. Lee, R. S. Ruoff and H. Lee, Nat. Commun., 2010, 1, 73 CrossRef PubMed.
  71. A. Bagri, C. Mattevi, M. Acik, Y. J. Chabal, M. Chhowalla and V. B. Shenoy, Nat. Chem., 2010, 2, 581–587 CrossRef CAS PubMed.
  72. H. Fei, J. Dong, M. J. Arellano-Jiménez, G. Ye, N. Dong Kim, E. L. G. Samuel, Z. Peng, Z. Zhu, F. Qin, J. Bao, M. J. Yacaman, P. M. Ajayan, D. Chen and J. M. Tour, Nat. Commun., 2015, 6, 8668 CrossRef CAS PubMed.
  73. K. Jiang, S. Siahrostami, T. Zheng, Y. Hu, S. Hwang, E. Stavitski, Y. Peng, J. J. Dynes, M. Gangishetty, D. Su, K. Attenkofer and H. Wang, Energy Environ. Sci., 2018, 11, 893–903 RSC.
  74. J. Guan, Z. Duan, F. Zhang, S. D. Kelly, R. Si, M. Dupuis, Q. Huang, J. Q. Chen, C. Tang and C. Li, Nat. Catal., 2018, 1, 870–877 CrossRef CAS.
  75. X. Wen, Z. Duan, L. Bai and J. Guan, J. Power Sources, 2019, 431, 265–273 CrossRef CAS.
  76. H. Fei, J. Dong, C. Wan, Z. Zhao, X. Xu, Z. Lin, Y. Wang, H. Liu, K. Zang, J. Luo, S. Zhao, W. Hu, W. Yan, I. Shakir, Y. Huang and X. Duan, Adv. Mater., 2018, 30, 1802146 CrossRef PubMed.
  77. B. Y. Guan, X. Y. Yu, H. B. Wu and X. W. Lou, Adv. Mater., 2017, 29, 1703614 CrossRef PubMed.
  78. S. Dang, Q.-L. Zhu and Q. Xu, Nat. Rev. Mater., 2017, 3, 17075 CrossRef.
  79. J. Liu, D. Zhu, C. Guo, A. Vasileff and S.-Z. Qiao, Adv. Energy Mater., 2017, 7, 1700518 CrossRef.
  80. M. Xiao, J. Zhu, G. Li, N. Li, S. Li, Z. P. Cano, L. Ma, P. X. Cui, P. Xu, G. Jiang, H. Jin, S. Wang, T. Wu, J. Lu, A. Yu, D. Su and Z. Chen, Angew. Chem., 2019, 131, 9742 CrossRef.
  81. J. Han, X. Meng, L. Lu, J. Bian, Z. Li and C. Sun, Adv. Funct. Mater., 2019 DOI:10.1002/adfm.201808872.
  82. J. Wu, H. Zhou, Q. Li, M. Chen, J. Wan, N. Zhang, L. Xiong, S. Li, B. Y. Xia, G. Feng, M. Liu and L. Huang, Adv. Energy Mater., 2019, 9, 1900149 CrossRef.
  83. T. Sun, L. Xu, D. Wang and Y. Li, Nano Res., 2019, 12, 2067 CrossRef CAS.
  84. Y. Qu, Z. Li, W. Chen, Y. Lin, T. Yuan, Z. Yang, C. Zhao, J. Wang, C. Zhao, X. Wang, F. Zhou, Z. Zhuang, Y. Wu and Y. Li, Nat. Catal., 2018, 1, 781–786 CrossRef CAS.
  85. X. Wan, X. Liu, Y. Li, R. Yu, L. Zheng, W. Yan, H. Wang, M. Xu and J. Shui, Nat. Catal., 2019, 2, 259–268 CrossRef CAS.
  86. S. Wei, A. Li, J.-C. Liu, Z. Li, W. Chen, Y. Gong, Q. Zhang, W.-C. Cheong, Y. Wang, L. Zheng, H. Xiao, C. Chen, D. Wang, Q. Peng, L. Gu, X. Han, J. Li and Y. Li, Nat. Nanotechnol., 2018, 13, 856–861 CrossRef CAS PubMed.
  87. Y. Chen, S. Ji, Y. Wang, J. Dong, W. Chen, Z. Li, R. Shen, L. Zheng, Z. Zhuang, D. Wang and Y. Li, Angew. Chem., 2017, 129, 7041–7045 CrossRef.
  88. M. Xiao, J. Zhu, L. Ma, Z. Jin, J. Ge, X. Deng, Y. Hou, Q. He, J. Li, Q. Jia, S. Mukerjee, R. Yang, Z. Jiang, D. Su, C. Liu and W. Xing, ACS Catal., 2018, 8, 2824–2832 CrossRef CAS.
  89. P. Yin, T. Yao, Y. Wu, L. Zheng, Y. Lin, W. Liu, H. Ju, J. Zhu, X. Hong, Z. Deng, G. Zhou, S. Wei and Y. Li, Angew. Chem., Int. Ed., 2016, 55, 10800–10805 CrossRef CAS PubMed.
  90. L. Jiao and H.-L. Jiang, Chem, 2019, 5, 786–804 CAS.
  91. J. Li, M. Chen, D. A. Cullen, S. Hwang, M. Wang, B. Li, K. Liu, S. Karakalos, M. Lucero, H. Zhang, C. Lei, H. Xu, G. E. Sterbinsky, Z. Feng, D. Su, K. L. More, G. Wang, Z. Wang and G. Wu, Nat. Catal., 2018, 1, 935–945 CrossRef CAS.
  92. X. Wang, W. Chen, L. Zhang, T. Yao, W. Liu, Y. Lin, H. Ju, J. Dong, L. Zheng, W. Yan, X. Zheng, Z. Li, X. Wang, J. Yang, D. He, Y. Wang, Z. Deng, Y. Wu and Y. Li, J. Am. Chem. Soc., 2017, 139, 9419–9422 CrossRef CAS PubMed.
  93. Y. Deng, B. Chi, J. Li, G. Wang, L. Zheng, X. Shi, Z. Cui, L. Du, S. Liao, K. Zang, J. Luo, Y. Hu and X. Sun, Adv. Energy Mater., 2019, 9, 1802856 CrossRef.
  94. Y. Qu, B. Chen, Z. Li, X. Duan, L. Wang, Y. Lin, T. Yuan, F. Zhou, Y. Hu, Z. Yang, C. Zhao, J. Wang, C. Zhao, Y. Hu, G. Wu, Q. Zhang, Q. Xu, B. Liu, P. Gao, R. You, W. Huang, L. Zheng, L. Gu, Y. Wu and Y. Li, J. Am. Chem. Soc., 2019, 141, 4505–4509 CrossRef CAS PubMed.
  95. J. Yang, Z. Qiu, C. Zhao, W. Wei, W. Chen, Z. Li, Y. Qu, J. Dong, J. Luo, Z. Li and Y. Wu, Angew. Chem., Int. Ed., 2018, 57, 14095–14100 CrossRef CAS PubMed.
  96. H. Yan, H. Cheng, H. Yi, Y. Lin, T. Yao, C. Wang, J. Li, S. Wei and J. Lu, J. Am. Chem. Soc., 2015, 137, 10484–10487 CrossRef CAS PubMed.
  97. H. Wei, H. Wu, K. Huang, B. Ge, J. Ma, J. Lang, D. Zu, M. Lei, Y. Yao, W. Guo and H. Wu, Chem. Sci., 2019, 10, 2830–2836 RSC.
  98. K. Huang, L. Zhang, T. Xu, H. Wei, R. Zhang, X. Zhang, B. Ge, M. Lei, J.-Y. Ma, L.-M. Liu and H. Wu, Nat. Commun., 2019, 10, 606 CrossRef PubMed.
  99. W. Bi, X. Li, R. You, M. Chen, R. Yuan, W. Huang, X. Wu, W. Chu, C. Wu and Y. Xie, Adv. Mater., 2018, 30, 1706617 CrossRef PubMed.
  100. E. Kano, A. Hashimoto, T. Kaneko, N. Tajima, T. Ohno and M. Takeguchi, Nanoscale, 2016, 8, 529–535 RSC.
  101. Z. He, K. He, A. W. Robertson, A. I. Kirkland, D. Kim, J. Ihm, E. Yoon, G.-D. Lee and J. H. Warner, Nano Lett., 2014, 14, 3766–3772 CrossRef CAS PubMed.
  102. S. L. James, C. J. Adams, C. Bolm, D. Braga, P. Collier, T. Friscic, F. Grepioni, K. D. M. Harris, G. Hyett, W. Jones, A. Krebs, J. Mack, L. Maini, A. G. Orpen, I. P. Parkin, W. C. Shearouse, J. W. Steed and D. C. Waddell, Chem. Soc. Rev., 2012, 41, 413–447 RSC.
  103. M. Yi and Z. Shen, J. Mater. Chem. A, 2015, 3, 11700–11715 RSC.
  104. X. Cui, J. Xiao, Y. Wu, P. Du, R. Si, H. Yang, H. Tian, J. Li, W.-H. Zhang, D. Deng and X. Bao, Angew. Chem., Int. Ed., 2016, 55, 6708–6712 CrossRef CAS PubMed.
  105. J. Lu, J. W. Elam and P. C. Stair, Acc. Chem. Res., 2013, 46, 1806–1815 CrossRef CAS PubMed.
  106. S. Sun, G. Zhang, N. Gauquelin, N. Chen, J. Zhou, S. Yang, W. Chen, X. Meng, D. Geng, M. N. Banis, R. Li, S. Ye, S. Knights, G. A. Botton, T.-K. Sham and X. Sun, Sci. Rep., 2013, 3, 1775 CrossRef.
  107. S. Stambula, N. Gauquelin, M. Bugnet, S. Gorantla, S. Turner, S. Sun, J. Liu, G. Zhang, X. Sun and G. A. Botton, J. Phys. Chem. C, 2014, 118, 3890–3900 CrossRef CAS.
  108. N. Cheng, S. Stambula, D. Wang, M. N. Banis, J. Liu, A. Riese, B. Xiao, R. Li, T.-K. Sham, L.-M. Liu, G. A. Botton and X. Sun, Nat. Commun., 2016, 7, 13638 CrossRef CAS PubMed.
  109. S. Linic, U. Aslam, C. Boerigter and M. Morabito, Nat. Mater., 2015, 14, 567 CrossRef CAS PubMed.
  110. H. Wei, K. Huang, D. Wang, R. Zhang, B. Ge, J. Ma, B. Wen, S. Zhang, Q. Li, M. Lei, C. Zhang, J. Irawan, L.-M. Liu and H. Wu, Nat. Commun., 2017, 8, 1490 CrossRef PubMed.
  111. T. Li, J. Liu, Y. Song and F. Wang, ACS Catal., 2018, 8, 8450–8458 CrossRef CAS.
  112. Z. Zhang, Y. Chen, L. Zhou, C. Chen, Z. Han, B. Zhang, Q. Wu, L. Yang, L. Du, Y. Bu, P. Wang, X. Wang, H. Yang and Z. Hu, Nat. Commun., 2019, 10, 1657 CrossRef PubMed.
  113. Q. Lai, Y. Zhao, J. Zhu, Y. Liang, J. He and J. Chen, ChemElectroChem, 2018, 5, 1822–1826 CrossRef CAS.
  114. A. Bakandritsos, R. G. Kadam, P. Kumar, G. Zoppellaro, M. Medved’, J. Tuček, T. Montini, O. Tomanec, P. Andrýsková, B. Drahoš, R. S. Varma, M. Otyepka, M. B. Gawande, P. Fornasiero and R. Zbořil, Adv. Mater., 2019, 31, 1900323 CrossRef PubMed.
  115. G. Malta, S. A. Kondrat, S. J. Freakley, C. J. Davies, L. Lu, S. Dawson, A. Thetford, E. K. Gibson, D. J. Morgan, W. Jones, P. P. Wells, P. Johnston, C. R. A. Catlow, C. J. Kiely and G. J. Hutchings, Science, 2017, 355, 1399 CrossRef CAS PubMed.
  116. J. J. Rehr and R. C. Albers, Rev. Mod. Phys., 2000, 72, 621–654 CrossRef CAS.
  117. S. Bordiga, E. Groppo, G. Agostini, J. A. van Bokhoven and C. Lamberti, Chem. Rev., 2013, 113, 1736–1850 CrossRef CAS PubMed.
  118. D. C. Koningsberger and R. Prins, X-ray absorption: principles, applications, techniques of EXAFS, SEXAFS, and XANES, John Wiley and Sons, New York, NY, 1988 Search PubMed.
  119. H. Funke, A. C. Scheinost and M. Chukalina, Phys. Rev. B: Condens. Matter Mater. Phys., 2005, 71, 094110 CrossRef.
  120. Y. Chen, S. Ji, W. Sun, W. Chen, J. Dong, J. Wen, J. Zhang, Z. Li, L. Zheng, C. Chen, Q. Peng, D. Wang and Y. Li, J. Am. Chem. Soc., 2018, 140, 7407–7410 CrossRef CAS PubMed.
  121. J. Yang, B. Chen, X. Liu, W. Liu, Z. Li, J. Dong, W. Chen, W. Yan, T. Yao, X. Duan, Y. Wu and Y. Li, Angew. Chem., Int. Ed., 2018, 57, 9495–9500 CrossRef CAS PubMed.
  122. W. Liu, L. Zhang, W. Yan, X. Liu, X. Yang, S. Miao, W. Wang, A. Wang and T. Zhang, Chem. Sci., 2016, 7, 5758–5764 RSC.
  123. M. Benfatto, S. Della Longa and C. R. Natoli, J. Synchrotron Radiat., 2003, 10, 51–57 CrossRef CAS PubMed.
  124. K. Hayakawa, K. Hatada, S. Della Longa, P. D’Angelo and M. Benfatto, AIP Conf. Proc., 2007, 882, 111–113 CrossRef CAS.
  125. X. Li, X. Yang, J. Zhang, Y. Huang and B. Liu, ACS Catal., 2019, 9, 2521–2531 CrossRef CAS.
  126. L. Cao, Q. Luo, W. Liu, Y. Lin, X. Liu, Y. Cao, W. Zhang, Y. Wu, J. Yang, T. Yao and S. Wei, Nat. Catal., 2019, 2, 134–141 CrossRef CAS.
  127. B. Lassalle-Kaiser, S. Gul, J. Kern, V. K. Yachandra and J. Yano, J. Electron Spectrosc. Relat. Phenom., 2017, 221, 18–27 CrossRef CAS PubMed.
  128. J. Gu, C.-S. Hsu, L. Bai, H. M. Chen and X. Hu, Science, 2019, 364, 1091 CrossRef CAS PubMed.
  129. G. B. B. Varadwaj and V. O. Nyamori, Nano Res., 2016, 9, 3598–3621 CrossRef CAS.
  130. C. H. Choi, M. Kim, H. C. Kwon, S. J. Cho, S. Yun, H.-T. Kim, K. J. J. Mayrhofer, H. Kim and M. Choi, Nat. Commun., 2016, 7, 10922 CrossRef CAS PubMed.
  131. T. Sun, B. Tian, J. Lu and C. Su, J. Mater. Chem. A, 2017, 5, 18933–18950 RSC.
  132. A. Serov, K. Artyushkova and P. Atanassov, Adv. Energy Mater., 2014, 4, 1301735 CrossRef.
  133. G. Wu, Front. Energy, 2017, 11, 286–298 CrossRef.
  134. L. Yang, D. Cheng, H. Xu, X. Zeng, X. Wan, J. Shui, Z. Xiang and D. Cao, PNAS, 2018, 115, 6626–6631 CrossRef CAS PubMed.
  135. Y. He, S. Hwang, D. A. Cullen, M. A. Uddin, L. Langhorst, B. Li, S. Karakalos, A. J. Kropf, E. C. Wegener, J. Sokolowski, M. Chen, D. Myers, D. Su, K. L. More, G. Wang, S. Litster and G. Wu, Energy Environ. Sci., 2018, 12, 250–260 RSC.
  136. J. Liu, M. Jiao, B. Mei, Y. Tong, Y. Li, M. Ruan, P. Song, G. Sun, L. Jiang, Y. Wang, Z. Jiang, L. Gu, Z. Zhou and W. Xu, Angew. Chem., Int. Ed., 2018, 58, 1163–1167 CrossRef PubMed.
  137. F. Li, G.-F. Han, H.-J. Noh, S.-J. Kim, Y. Lu, H. Y. Jeong, Z. Fu and J.-B. Baek, Energy Environ. Sci., 2018, 11, 2263–2269 RSC.
  138. J. Li, S. Chen, N. Yang, M. Deng, S. Ibraheem, J. Deng, J. Li, L. Li and Z. Wei, Angew. Chem., Int. Ed., 2019, 58, 7035–7039 CrossRef CAS PubMed.
  139. J. D. Holladay, J. Hu, D. L. King and Y. Wang, Catal. Today, 2009, 139, 244–260 CrossRef CAS.
  140. J. K. Nørskov, T. Bligaard, A. Logadottir, J. R. Kitchin, J. G. Chen, S. Pandelov and U. Stimming, J. Electrochem. Soc., 2005, 152, J23–J26 CrossRef.
  141. D. Liu, X. Li, S. Chen, H. Yan, C. Wang, C. Wu, Y. A. Haleem, S. Duan, J. Lu, B. Ge, P. M. Ajayan, Y. Luo, J. Jiang and L. Song, Nat. Energy, 2019, 4, 512–518 CrossRef CAS.
  142. H. Zhang, P. An, W. Zhou, B. Y. Guan, P. Zhang, J. Dong and X. W. Lou, Sci. Adv., 2018, 4, eaao6657 CrossRef PubMed.
  143. H. J. Qiu, Y. Ito, W. Cong, Y. Tan, P. Liu, A. Hirata, T. Fujita, Z. Tang and M. Chen, Angew. Chem., Int. Ed., 2015, 54, 14031–14035 CrossRef CAS PubMed.
  144. L. Fan, P. F. Liu, X. Yan, L. Gu, Z. Z. Yang, H. G. Yang, S. Qiu and X. Yao, Nat. Commun., 2016, 7, 10667 CrossRef CAS PubMed.
  145. L. Zhang, Y. Jia, G. Gao, X. Yan, N. Chen, J. Chen, M. T. Soo, B. Wood, D. Yang, A. Du and X. Yao, Chem, 2018, 4, 285–297 CAS.
  146. Z.-L. Wang, X.-F. Hao, Z. Jiang, X.-P. Sun, D. Xu, J. Wang, H.-X. Zhong, F.-L. Meng and X.-B. Zhang, J. Am. Chem. Soc., 2015, 137, 15070–15073 CrossRef CAS PubMed.
  147. H.-W. Liang, S. Brüller, R. Dong, J. Zhang, X. Feng and K. Müllen, Nat. Commun., 2015, 6, 7992 CrossRef CAS PubMed.
  148. M. D. Hossain, Z. Liu, M. Zhuang, X. Yan, G.-L. Xu, C. A. Gadre, A. Tyagi, I. H. Abidi, C.-J. Sun, H. Wong, A. Guda, Y. Hao, X. Pan, K. Amine and Z. Luo, Adv. Energy Mater., 2019, 9, 1803689 CrossRef.
  149. W. Chen, J. Pei, C.-T. He, J. Wan, H. Ren, Y. Zhu, Y. Wang, J. Dong, S. Tian, W.-C. Cheong, S. Lu, L. Zheng, X. Zheng, W. Yan, Z. Zhuang, C. Chen, Q. Peng, D. Wang and Y. Li, Angew. Chem., 2017, 129, 16302–16306 CrossRef.
  150. C. Wenxing, P. Jiajing, H. Chun-Ting, W. Jiawei, R. Hanlin, W. Yu, D. Juncai, W. Konglin, C. Weng-Chon, M. Junjie, Z. Xusheng, Y. Wensheng, Z. Zhongbin, C. Chen, P. Qing, W. Dingsheng and L. Yadong, Adv. Mater., 2018, 30, 1800396 CrossRef PubMed.
  151. Y. Pan, S. Liu, K. Sun, X. Chen, B. Wang, K. Wu, X. Cao, W.-C. Cheong, R. Shen, A. Han, Z. Chen, L. Zheng, J. Luo, Y. Lin, Y. Liu, D. Wang, Q. Peng, Q. Zhang, C. Chen and Y. Li, Angew. Chem., 2018, 57, 8614–8618 CrossRef CAS PubMed.
  152. P. Du and R. Eisenberg, Energy Environ. Sci., 2012, 5, 6012–6021 RSC.
  153. B. Kumar, J. P. Brian, V. Atla, S. Kumari, K. A. Bertram, R. T. White and J. M. Spurgeon, Catal. Today, 2016, 270, 19–30 CrossRef CAS.
  154. L. Zhang, Z.-J. Zhao and J. Gong, Angew. Chem., Int. Ed., 2017, 56, 11326–11353 CrossRef CAS PubMed.
  155. X. Su, X.-F. Yang, Y. Huang, B. Liu and T. Zhang, Acc. Chem. Res., 2019, 52, 656–664 CrossRef CAS PubMed.
  156. A. S. Varela, W. Ju, A. Bagger, P. Franco, J. Rossmeisl and P. Strasser, ACS Catal., 2019, 9, 7270–7284 CrossRef CAS.
  157. P. Su, K. Iwase, S. Nakanishi, K. Hashimoto and K. Kamiya, Small, 2016, 12, 6083–6089 CrossRef CAS PubMed.
  158. C. Zhao, X. Dai, T. Yao, W. Chen, X. Wang, J. Wang, J. Yang, S. Wei, Y. Wu and Y. Li, J. Am. Chem. Soc., 2017, 139, 8078–8081 CrossRef CAS PubMed.
  159. X. Li, W. Bi, M. Chen, Y. Sun, H. Ju, W. Yan, J. Zhu, X. Wu, W. Chu, C. Wu and Y. Xie, J. Am. Chem. Soc., 2017, 139, 14889–14892 CrossRef CAS PubMed.
  160. K. Jiang, S. Siahrostami, A. J. Akey, Y. Li, Z. Lu, J. Lattimer, Y. Hu, C. Stokes, M. Gangishetty, G. Chen, Y. Zhou, W. Hill, W.-B. Cai, D. Bell, K. Chan, J. K. Nørskov, Y. Cui and H. Wang, Chem, 2017, 3, 950–960 CAS.
  161. C. Yan, H. Li, Y. Ye, H. Wu, F. Cai, R. Si, J. Xiao, S. Miao, S. Xie, F. Yang, Y. Li, G. Wang and X. Bao, Energy Environ. Sci., 2018, 11, 1204–1210 RSC.
  162. A. S. Varela, N. Ranjbar Sahraie, J. Steinberg, W. Ju, H.-S. Oh and P. Strasser, Angew. Chem., Int. Ed., 2015, 54, 10758–10762 CrossRef CAS PubMed.
  163. F. Pan, H. Zhang, K. Liu, D. Cullen, K. More, M. Wang, Z. Feng, G. Wang, G. Wu and Y. Li, ACS Catal., 2018, 8, 3116–3122 CrossRef CAS.
  164. F. Yang, P. Song, X. Liu, B. Mei, W. Xing, Z. Jiang, L. Gu and W. Xu, Angew. Chem., Int. Ed., 2018, 57, 12303–12307 CrossRef CAS PubMed.
  165. X. Zu, X. Li, W. Liu, Y. Sun, J. Xu, T. Yao, W. Yan, S. Gao, C. Wang, S. Wei and Y. Xie, Adv. Mater., 2019, 31, 1808135 CrossRef PubMed.
  166. C. Zhao, Y. Wang, Z. Li, W. Chen, Q. Xu, D. He, D. Xi, Q. Zhang, T. Yuan, Y. Qu, J. Yang, F. Zhou, Z. Yang, X. Wang, J. Wang, J. Luo, Y. Li, H. Duan, Y. Wu and Y. Li, Joule, 2019, 3, 584–594 CrossRef CAS.
  167. P. Lu, Y. Yang, J. Yao, M. Wang, S. Dipazir, M. Yuan, J. Zhang, X. Wang, Z. Xie and G. Zhang, Appl. Catal., B, 2019, 241, 113–119 CrossRef CAS.
  168. B. Zhang, J. Zhang, J. Shi, D. Tan, L. Liu, F. Zhang, C. Lu, Z. Su, X. Tan, X. Cheng, B. Han, L. Zheng and J. Zhang, Nat. Commun., 2019, 10, 2980 CrossRef PubMed.
  169. Y. Cheng, S. Yang, S. P. Jiang and S. Wang, Small Methods, 2019, 3, 1800440 CrossRef.
  170. Y. Cheng, S. Zhao, H. Li, S. He, J.-P. Veder, B. Johannessen, J. Xiao, S. Lu, J. Pan, M. F. Chisholm, S.-Z. Yang, C. Liu, J. G. Chen and S. P. Jiang, Appl. Catal., B, 2019, 243, 294–303 CrossRef CAS.
  171. T. Zheng, K. Jiang, N. Ta, Y. Hu, J. Zeng, J. Liu and H. Wang, Joule, 2019, 3, 265–278 CrossRef CAS.
  172. Z. Chen, K. Mou, S. Yao and L. Liu, ChemSusChem, 2018, 11, 2944–2952 CrossRef CAS PubMed.
  173. C. Zhang, S. Yang, J. Wu, M. Liu, S. Yazdi, M. Ren, J. Sha, J. Zhong, K. Nie, A. S. Jalilov, Z. Li, H. Li, B. I. Yakobson, Q. Wu, E. Ringe, H. Xu, P. M. Ajayan and J. M. Tour, Adv. Energy Mater., 2018, 8, 1703487 CrossRef.
  174. T. N. Huan, N. Ranjbar, G. Rousse, M. Sougrati, A. Zitolo, V. Mougel, F. Jaouen and M. Fontecave, ACS Catal., 2017, 7, 1520–1525 CrossRef CAS.
  175. T. Möller, W. Ju, A. Bagger, X. Wang, F. Luo, T. N. Thanh, A. Varela, J. Rossmeisl and P. Strasser, Energy Environ. Sci., 2019, 12, 640–647 RSC.
  176. W. Ju, A. Bagger, G.-P. Hao, A. S. Varela, I. Sinev, V. Bon, B. Roldan Cuenya, S. Kaskel, J. Rossmeisl and P. Strasser, Nat. Commun., 2017, 8, 944 CrossRef PubMed.
  177. Z. Geng, Y. Cao, W. Chen, X. Kong, Y. Liu, T. Yao and Y. Lin, Appl. Catal., B, 2019, 240, 234–240 CrossRef CAS.
  178. Y. Pan, R. Lin, Y. Chen, S. Liu, W. Zhu, X. Cao, W. Chen, K. Wu, W.-C. Cheong, Y. Wang, L. Zheng, J. Luo, Y. Lin, Y. Liu, C. Liu, J. Li, Q. Lu, X. Chen, D. Wang, Q. Peng, C. Chen and Y. Li, J. Am. Chem. Soc., 2018, 140, 4218–4221 CrossRef CAS PubMed.
  179. S. L. Foster, S. I. P. Bakovic, R. D. Duda, S. Maheshwari, R. D. Milton, S. D. Minteer, M. J. Janik, J. N. Renner and L. F. Greenlee, Nat. Catal., 2018, 1, 490–500 CrossRef.
  180. C. Xiaoyang, T. Cheng and Z. Qiang, Adv. Energy Mater., 2018, 8, 1800369 CrossRef.
  181. W. Qiu, X.-Y. Xie, J. Qiu, W.-H. Fang, R. Liang, X. Ren, X. Ji, G. Cui, A. M. Asiri, G. Cui, B. Tang and X. Sun, Nat. Commun., 2018, 9, 3485 CrossRef PubMed.
  182. C. Guo, J. Ran, A. Vasileff and S.-Z. Qiao, Energy Environ. Sci., 2018, 11, 45–56 RSC.
  183. Z. Geng, Y. Liu, X. Kong, P. Li, K. Li, Z. Liu, J. Du, M. Shu, R. Si and J. Zeng, Adv. Mater., 2018, 30, 1803498 CrossRef PubMed.
  184. L. Han, X. Liu, J. Chen, R. Lin, H. Liu, F. Lü, S. Bak, Z. Liang, S. Zhao, E. Stavitski, J. Luo, R. R. Adzic and H. L. Xin, Angew. Chem., Int. Ed., 2019, 58, 2321–2325 CrossRef CAS PubMed.

This journal is © The Royal Society of Chemistry 2019