Tuning the intrinsic catalytic activities of oxygen-evolution catalysts by doping: a comprehensive review

Electrochemical water splitting produces clean hydrogen fuel as one of the pivotal alternative energies to fossil fuels in the near future. However, the anodic oxygen evolution reaction (OER) is a signi ﬁ cant bottleneck that curtails large-scale applications of electrochemical water splitting technology, owing to its sluggish reaction kinetics. In the past few decades, various methods have been proposed to improve the OER kinetics. Among them, doping is a simple and e ﬃ cient method to mold the OER kinetics of a catalyst by incorporating di ﬀ erent or hetero atoms into the host lattice. These e ﬀ orts are vital to design highly e ﬃ cient OER catalysts for real-world applications. However, the OER mechanism of a doped catalyst varies, depending on the host lattice and the dopant. This review highlights di ﬀ erent doping strategies and associated OER mechanisms of state-of-the-art catalysts, including oxides (noble metal oxides, perovskite oxides, spinel oxides, hydroxides and others), non-oxides (metal sul ﬁ des, metal selenides, metal phosphides, metal nitrides and metal carbides), and carbon-based catalysts (graphene, carbon nanotubes and others). Fundamental understanding of the doping e ﬀ ects on the OER from combined experimental and theoretical research provides guidelines for designing e ﬃ cient catalysts.


Introduction
The demand for transitioning the world's energy from fossil fuels to eco-friendly renewable energy is rapidly increasing, in order to cope with the depletion of fossil fuels and to curtail the associated CO 2 emission. 1 Currently, sunlight and wind are the main contributors to renewable energy, but they are highly intermittent. 2 To overcome this challenge, researchers have been working on new energy storage techniques that can sustain the erratic behavior of solar, wind and tidal energies. 3 Instead, we can store these excess energies in the form of hydrogen through water electrolysis (eqn (1)), 4 and the stored hydrogen is readily available as a fuel cell for energy production in a clean way without CO 2 emission (eqn (2)). 5 Hydrogen production through water electrolysis is a thermodynamically infeasible process that requires additional energy to drive this reaction, and theoretically, 1.23 V (pH ¼ 0) electric potential is required. 6 However, the oxygen evolution reaction (OER) during the water electrolysis is associated with a sluggish reaction kinetics that comprises O]O bond formation with a loss of 4 e À /4 H + , requiring an even higher electric potential than the theoretical prediction. 7 The excess potential above the theoretical prediction is known as overpotential (h), which is one of the major benchmarks for denoting OER activity, along with high current density and low Tafel slope values. 8 Over the past few decades, signicant research has been conducted to improve the OER kinetics through various strategies for economically viable hydrogen production. 9 So far, RuO 2 and IrO 2 catalysts have been proven to be two of the most efficient types of electrocatalyst for the OER in both acidic and alkaline media. However, the high cost associated with these catalysts hinders their large-scale water electrolysis application. 10 In alkaline media, researchers can replace these expensive catalysts with earth-abundant transition metal (TM)-based catalysts, 11 as they are stable in such alkaline environments, while being unstable in acidic media. 12,13 The catalyst stability is also an important factor for consideration.
The OER catalytic activity can be improved by enhancing the number of active sites on the surface of the catalyst, which is termed as the intrinsic activity of the catalyst. [14][15][16][17][18][19][20][21][22] The intrinsic activity of the OER catalyst can be augmented by introducing elemental dopants, creating vacancies, optimizing the morphology, decreasing the particle size, or via surface reconstruction during the OER process. 23 Among them, elemental doping is the most efficient way to engineer the activity of OER catalysts. 9 However, a fundamental understanding of OER mechanisms is required in catalyst design to shi its efficiency closer to the thermodynamic limit. 24 In the past few decades, several reaction mechanisms for the OER have been unveiled to fabricate efficient OER catalysts. 25 To date, the current literature on OER catalysts lacks a comprehensive review on the doping effect in the OER mechanisms of catalysts. In this review, by corroborating experimental and theoretical results, we illustrate how particular elemental doping can improve the OER efficiency of a catalyst. In the beginning, OER intermediates by doping are listed. In particular materials, the OER mechanisms are presented, including the inuence of dopant on improving the intrinsic activity through optimizing OER descriptors, such as e g orbital occupancy, metal-oxygen covalency, and structural parameters like M-O bond binding energy, length, and angle. Various types of material that have exhibited efficient doping in the OER are reviewed. Understanding of these OER mechanisms will guide the design of new efficient OER catalysts.

OER mechanisms
Generally, in the OER, four electrons and protons are generated and combine on the surface of a catalyst. Several mechanisms are proposed for alkaline and acidic electrolytes; however, the research community primarily accepts the oxide 26 and electrochemical oxide 26 mechanisms. Karsil'shchikov 27 and Wade-Hackerman 28 pathways are the two known mechanisms in an acidic medium for the OER. 28 Here, the symbol * denotes an active catalytic site on the surface of the OER catalyst. In one of the reaction steps below, 2*O / 2* + O 2(g) in both oxide mechanisms ( Fig. 1a and c) or the electrochemical oxide mechanism ( Fig. 1b and d), oxygen gas is generated. In case of recombination of two *O intermediates or generation of oxygen from the *O intermediate, oxygen evolution is impossible because of a substantial thermodynamic barrier. The oxygen gas is generated from the active catalytic site aer two steps via the formation of intermediate *O-OH, a step which is the most sluggish one out of all rate-determining steps (RDSs) in the OER.
Besides the electrocatalyst type and chemical stability, the geometry and electronic structure are the key parameters for optimization in developing efficient OER catalysts. 31 Geometrical surface area is the most fundamental and critical parameter, and it can be improved by synthesizing nanoforms of an electrocatalyst. By improving the specic surface area of the catalyst, the mass transfer rate and density of active catalytic sites can be accelerated. 32 However, nanomaterials are oen aggregated, resulting in poor stability during long-term service. 9 Tuning the electronic structure is another critical factor in improving the OER activities, and elemental doping is one of the best ways of tuning the electronic structure. 33 The tuning of the electronic structure by doping inuences the binding energy of OER intermediates, such as *O, *OH, and *O-OH. Hence, compared with other approaches for improving OER activities, doping provides an effective way to achieve large-scale activity and long-term stability under harsh applied anodic overpotentials.
The dopant is present in trace or small amounts. Doping generates ionizable species within the host material by incorporating impurities. The ionizable species could lead to the shi of the Fermi level (E F ), enabling carrier transport with superior catalytic activity. Doping can be performed through (a) gating for direct carrier injection; (b) antisites and vacancies as structural defects; and (c) photodoping through generation of transient carrier species via photo-excitation. Incorporation of foreign atoms that do not produce free carriers is not considered as doping. 34 In the following sections, we will elaborate on the inuences of doping in the following common catalytic systems, including various oxides (noble metal oxides, perovskite oxides, spinel oxides, hydroxides and other oxides), non-oxides (metal suldes, metal selenides, metal phosphides, metal nitrides, and metal carbides), and carbon-based catalysts. Fig. 2 shows the overpotential at 10 mA cm À2 and Tafel slope of the representative types of catalyst (data collected from the following sections). Lower overpotential and lower Tafel slopes are desirable for high catalytic performance. The effects of doping and its impacts on OER activity in the various prescribed electrocatalysts are nely discussed hereaer in terms of accelerating OER kinetics with less energy input. Table 1 provides a summary of some representative doped noble metal oxides and their catalytic performances. Both IrO 2 and RuO 2 are considered as benchmark catalysts for the OER in basic and acidic electrolytes. Under OER conditions, IrO 2 is more stable than RuO 2 , but both oxides are unstable at high anodic potentials. RuO 2 leaches out into the electrolyte in the form of RuO 4 at high anodic potentials. 35 Likewise, IrO 2 also dissolves in the electrolyte by forming IrO 3 at high anodic potentials. 36 However, IrO 2 shows considerably higher stability but lower OER performance, compared to RuO 2 .

Noble metal (Ir and Ru) oxides
Researchers doped RuO 2 with a small amount of Ir (Ru x -Ir 1Àx O 2 ) and observed a signicant improvement in stability without sacricing the OER performance. 37 Another way of enhancing the OER activity with lower total noble metal content is doping with earth-abundant metals, such as Cu, 38,39 Ni, 40 Co, 40,41 Mn, 42 Cr, 43,44 Ce, 45 W, 46 Mg, 47 and Zn. 48 For example, Cudoped IrO 2 (Cu 0.3 Ir 0.7 O d ) displays improved OER activity in acidic, neutral and basic media, compared to pristine IrO 2 . 39 The enhanced OER activity of Cu 0.3 Ir 0.7 O d is primarily imputed to the Jahn-Teller distortion in IrO 6 octahedra induced by Cu, as well as oxygen vacancies induced by Cu doping. Density functional theory (DFT) calculations revealed the improved OER activity mechanism by the Cu doping. With the increase in Cudoping, s and p bonding regions become broader and shi to a lower energy level because of the occupancy of s states. The partial density of states suggests that with an increase in Cu doping, the antibonding states of d xy shi up; in contrast, d z 2 shis down. The d xz and d yz orbitals were crossed by the Fermi level E F , while the d x 2 Ày 2 orbital is unchanged. Because of electrons hopping from d xz and d yz (t 2g ) to d z 2 , the e g level is partially lled as a consequence of enhanced degeneracy in Ir 5d orbitals by Cu doping. The strong interaction of e g and O 2p orbitals facilitates bonding with oxygen intermediates during s bond formation, which is more advantageous than the p bonding of t 2g orbitals. 39 Suntivch et al. postulated that the OER catalyst, specically perovskite oxide with the e g value close to unity, will exhibit superior OER performance. 62 Later, other scientists found that the e g lling values in some oxides also have a similar effect. 63 Site-specic deposition of a meager amount of Ir on facets of Ni nano-frameworks led to excellent OER activity in an acidic electrolyte and the catalyst even outperformed an Ir/C catalyst. 62 The oxygen adsorption of TMs was effectively tuned by doping with Ir. Similarly, Feng et al. reported porous IrCoNi nanocrystals as an efficient OER catalyst in highly corrosive HClO 4 electrolyte. 40 Fig. 3a shows the schematic illustration of IrCoNi porous hollow nanocrystals (PHNCs). A two-step process was followed to synthesize PHNCs. In the rst step, IrCo single nanocrystals (SNCs) were prepared in oleylamine; and in the second step, IrCoNi PHNCs were obtained from SNCs by etching with Fe 3+ in hexane/ethanol at room temperature. Fig. 3b shows a transmission electron microscopy (TEM) image of the as-synthesized IrCoNi PHNCs, presenting uniform hollow nanocrystals. The polarization curves of IrCo/C, IrCoNi/C, and IrNi/C PHNCs are compared with that of the Ir/C catalyst in Fig. 3c. The IrCoNi/C catalyst shows superior activity with an onset potential of 1.46 V and requires only 303 mV overpotential to achieve 10 mA cm À2 current density. The projected density of states (PDOS) of the d-band of Ir and IrCoNi catalysts is displayed in Fig. 3d, which shows the le side shi of the d-band center of IrCoNi from the Fermi level. The shi in the Fermi level facilitates the desired charge and discharge of OER intermediates. The schematic representation of OER reaction paths, as shown in Fig. 3e, reveals that the adsorption of OER intermediates becomes weaker when Ir is alloyed with Co and Ni, compared to pristine Ir. In another study, porous W-doped IrO 2 nanodendrites were found to exhibit superior OER performance under all pH conditions. 46 DFT calculations suggested that W doping stabilizes the IrO 2 at harsh anodic potentials by lowering the binding energies of oxygen intermediates. Qu et al. synthesized Cr-doped IrO x nanowires and evaluated their OER activity in 0.5 M H 2 SO 4 electrolyte. 43 The catalyst exhibits excellent OER activity with 25 h durability without any decay in activity. This high activity was ascribed to an increase in active site number and facile adsorption of OER intermediates via electronic structure tuning.
Like IrO 2 , the OER performance and stability of RuO 2 were also enhanced by doping. Cu-doped RuO 2 porous nano- Creating defect-containing RuO 2 via doping and dopant leaching is another way of improving the OER activity and stability in acidic medium. For example, Tian et al. doped RuO 2 with zinc and leached it out by aging in an acid, and the catalyst displayed 179 mV overpotential and 20 h stability at a current density of 10 mA cm À2 . 48 The high catalytic activity is accredited to the synergetic effect of active sites and tuning of the electronic structure. Recently, an oxygen-defect-containing Rubased catalyst was synthesized by doping with cobalt, which showed 169 mV overpotential and 50 h endurance at 10 mA cm À2 current density. 41 The doping generated a large amount of oxygen defects, following a vacancy-related lattice oxygen oxidation mechanism (LOM) rather than an adsorbate evolution reaction mechanism (AEM). Both computational and experimental studies elucidate this mechanism. All these studies revealed the effectiveness of doping Ir and Ru oxides for enabling higher activity and longer stability.

Perovskite oxides
Perovskite oxides are a large family of catalysts with the general formula ABO 3 , where A is either an alkaline earth metal or a rare earth metal, and B is a TM. Table 1 provides a summary of some representative doped perovskite oxides and their catalytic performances. The ideal perovskite oxide displays a cubic crystal structure. The TMs are in the oxidation states of +3, +4, or a mixture of them, governed by the metal on the A-site. 18,64 The OER activity can be tailored by doping at either the A-site, Bsite, or both sites to create oxygen vacancies in the perovskite. 18 The metal-oxygen hybridization depends upon the oxidation state of TMs, and the electronic properties of these perovskites rely on this parameter. The TMs with high electronegativity show high metal-oxygen hybridization, and the d-band center of the TM oxide moves closer to O 2p states as the electronegativity increases. However, besides the metal-oxygen hybridization, octahedral tilting, Jahn-Teller distortion, and displacement of B-site cations also play signicant roles in governing the electronic properties of perovskite oxides. 18,65,66 4.2.1 A-site doping. The activity and durability of SrRuO 3 perovskite are enhanced mainly by doping the Sr site with Na (Sr 0.95 Na 0.05 RuO 3 ), which results in a decrease in octahedral distortion and an increase in the oxidation state. 67 Na-doping results in no distortion in octahedra because of the change in the oxidation state of Ru with the inclusion of Na. DFT studies reveal that Na doping weakens the bonding between OER intermediates and Ru, allowing release of oxygen freely. 67 The increase in the oxidation state of Ru in Na-doped SrRuO 3 further improves its stability at high anodic potentials in an acidic medium. Doping of the La site of LaCoO 3 with Sr induces the formation of cobalt species with higher oxidation states, which inuence its crystal structure in such a way that O-Co-O is aligned at 180 . As a consequence, Co 3d and O 2p orbitals are mixed efficiently to enhance the intrinsic OER activity of the perovskite. 68 Grimaud et al. doped a series of lanthanides (Ln ¼ Pr, Sm, Gd and Ho) at the A-site of BaCoO 3Àd to optimize the e g lling values, thus improving the OER activity. 53 Among Ln-doped perovskites, Pr 0.5 Ba 0.5 CoO 3Àd shows superior activity. Estimated from X-ray absorption spectroscopy (XAS) and chemical titrations, the e g lling values of all Ln-doped compounds are close to unity. However, DFT calculations suggest that hybridization of Co 3d and O 2p orbitals decreases from Pr to Ho, because of the decreased Co oxidation state with an increase in electronegativity of dopants from Pr to Ho. Similarly, when the A-site of CaMnO 3 is doped with Yb (Ca 0.9 Yb 0.1 MnO 3 ), the OER activity is found to be 100 times higher than that of the pristine sample. 61 The excellent OER performance of Ca 0.9 Yb 0.1 MnO 3 is attributed to the increased Mn 3+ /Mn 4+ ratio, which provides the double exchange of electrons and optimized e g lling value of $0.81. The OER activity of double perovskites was also tuned by doping the A-site; for instance, Sr doping of the Ba site of PrBaCo 2 O 5+d improves the population of Co 4+ ions on the surface. 59 The doping enhances the ease of adsorption and deprotonation of *OOH intermediates from the catalyst surface, thus improving the OER performance of the perovskite.

B-site doping.
It is critical to tune the electronic structure of the B site to regulate the electrochemical behavior of perovskite oxides; hence, doping the B-site is the most facile way of governing the OER performance. 49,56 For instance, Shao's group demonstrated that B-site doping of SrTiO 3 with cobalt and iron elements resulted in an improved electrocatalyst for the OER. 60 Among doped perovskites, SrCo 0.9 Ti 0.1 O 3Àd displayed enhanced OER activity owing to its optimized e g lling value of 1.16, and formation of redox-active oxygen deciency. In another report, the Mn site of CaMnO 3 is doped with Nb (CaMn 0.75 Nb 0.25 O 3Àd ), which enhances the OER performance in terms of current density by onefold, and mass activity by sevenfold, compared to the undoped perovskite. 51 The signicant enhancement of electrochemical activity is imputed to the reduction of the oxidation state of Mn from +4 to +3 caused by doping with Nb 5+ , which in turn tunes the e g value close to unity.
Recently, researchers turned their attention to dope nonmetals like phosphorus, sulfur, and silicon at the B-site to stabilize the crystal structure of perovskite oxides. An important reason for doping, specically phosphorus, for improving electrocatalytic properties is that it has greater affinity towards combining with oxygen. Phosphorus doping at the B-site of SrCoO 3Àd (SC) perovskite results in tetragonal perovskite oxide SrCo 0.95 P 0.05 O 3Àd (SCP), which shows a considerable increase in OER performance in terms of activity and endurance relative to the parent compound. 58 Fig. 5a and b show linear sweep voltammetry (LSV) curves of SC and SCP before and aer the accelerated degradation test (ADT) conducted by repetition of 1000 CV cycles. The ADT results show a decrease in the activity of SC aer ADT, while SCP shows an increase in OER activity aer the ADT. The enhanced OER activity during the ADT of SCP is due to the presence of P dopant. The high-resolution TEM (HR-TEM) images presented Fig. 5c and d suggest that the decrease in the activity of SC is because of the formation of a 10-15 nm thick amorphous layer aer the ADT, which curtails the diffusion by blocking active sites, while SCP shows an only 1-2 nm thick amorphous layer (Fig. 5d). Fig. 5e shows a schematic illustration of the OER mechanism on SCP during the ADT, which indicates improved interaction between active cobalt sites and electrolyte through leaching of either Sr 2+ or P 5+ from the surface of the catalyst.
Likewise, Liu's group demonstrated the universality of P doping at the B-site of LaFeO 3Àd , LaCoO 3Àd and LaNiO 3Àd perovskites for improving the electrochemical water oxidation capacity. 55 The superior activity is attributed largely to the generation of a large amount of O 2 2À /O À species, and DFT calculation suggests a change in the oxidation state of the B cation that optimizes the e g value ($1). Zou et al. doped a series of TMs at the B-site of 6H-SrIrO 3 for improving its OER activity in an acidic medium. 52 Among them, cobalt doping was found to enhance the OER activity of 6H-SrIrO 3 signicantly. Fig. 6a presents the crystal structure of 6H-SrIrO 3 having two types of IrO 6 octahedron: one is face sharing (red), and the other one is corner-sharing isolated (olive green). Thus, the doped TMs can replace the Ir present in either the face sharing octahedra or corner-sharing octahedra. Fig. 6b displays the SEM image of Codoped 6H-SrIrO 3 having plate-like morphology and the thickness of these plates decreased aer doping. Fig. 6c shows the polarization curves of pristine and Co-doped catalysts in 0.1 M HClO 4 electrolyte. The catalyst with a 0.2 : 1 ratio of Co and Ir shows superior activity with the lowest overpotential of 235 mV at 10 mA cm À2 , and the catalytic activity is 2.5 times higher than that of the pristine catalyst. A comparison between the mass activities of pristine and Co-doped catalysts is presented in Fig. 6d, which suggests that the catalyst with the 0.2 : 1 ratio of Co and Ir has 3 times higher mass activity than 6H-SrIrO 3 . However, both catalysts show nearly the same OER kinetics with Tafel values of 51.8 mV dec À1 for the catalyst with the 0.2 : 1 ratio of Co and Ir and 58.7 mV dec À1 for the pristine sample. Computational studies suggest that in the pristine sample, a downshi of the Ir 5d band center away from the Fermi level is observed. In contrast, the O 2p band moves close to the Fermi level in the doped 6H-SrIrO 3 , which thus results in enhanced covalency and increases the OER activity. 52 Very recently, our group doped Fe on the B-site of SrCoO 3Àd and observed the formation of a tetragonal intermediate structure between perovskite and brownmillerite, with a stoichiometric formula of Sr 2 Co 1.5 Fe 0.5 O 6Àd . 57 Fig. 7a presents the OER activity of Sr-Co-Fe-oxides, which shows that Sr 2 Co 1.5 -Fe 0.5 O 6Àd exhibits superior OER activity among all doped oxides. The enhanced OER activity is ascribed to the unique stacking of octahedral (oh) and tetrahedral (th) layered units of the tetragonal structure, as shown in Fig. 7b, providing facile interactions between the electrode and electrolyte surface. The stability of the Sr 2 Co 1.5 Fe 0.5 O 6Àd catalyst was evaluated at 1.55 V for 10 h as shown in Fig. 7c, and it presents good stability with a slight increase in overpotential of 16 mV as depicted in Fig. 7d. Furthermore, theoretical calculations suggest that the enhanced covalency of Co 3d and O 2p is another prime contributor to the augmented OER performance of Sr 2 Co 1.5 -Fe 0.5 O 6Àd . This tetragonal intermediate structure presents a model for developing advanced catalysts in other systems.
4.2.3 Both A-and B-site doping, and controlled formation of oxygen vacancies. Tuning the electrocatalytic activity of perovskites is also possible by simultaneously doping both A and B sites. Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 O 3Àd (BSCF) is a well-studied dual-doped perovskite, and it displayed signicant intrinsic activity as a catalyst for the OER, which is an order magnitude higher than that of the benchmark catalyst IrO 2 . 50 The higher activity of BSCF originates from having the e g value close to unity, which is revealed by Shao-Horn's group by comparing several perovskite catalysts. 65 The doping of the Ba site with Sr and Co site with Fe in the double perovskite PrBaCo 2 O 5+d crystal structure signicantly improves the OER performance. 54 Specically, the perovskite nanobers with the composition of PrBa 0.5 Sr 0.5 Co 1.5 Fe 0.5 O 5+d displayed superior intrinsic activity that is way higher than that of the BSCF catalyst, and DFT calculations indicated that the O 2p band center was neither too close to nor too far away from the Fermi energy level. The adsorption scenarios of OER intermediates reveal that the *OH to *O conversion is the RDS, and the Fe site of Sr/Fe dual-doped PrBaCo 2 O 5+d requires a low overpotential (0.6 V) to initiate the reaction, while the Co site needs 0.7 V. 69 The best way to generate the oxygen vacancies in perovskites is doping a high valence A-site cation with a low valent dopant. For example, Sr doping at the A-site of LaCoO 3 results in ligand hole/oxygen deciency. This deciency increases with an increase in the Sr doping level. 70 The rise in oxygen vacancy leads to electroneutrality of the Co ion which improves the intermixing of Co 3d and O 2p orbitals. The evaluation of OER activity of a La 1Àx -Sr x CoO 3Àd perovskite series (x ¼ 0, 0.2, 0.4, 0.6, 0.8, and 1) suggests that with the increase in Sr content, the catalytic activity is improved. Furthermore, the authors evaluated ex situ electronic conductivity as a function of current density and found that the OER activity of the perovskite was enhanced with a large amount of Sr because of high electronic conductivity.

Spinel oxides
The general formula of spinel oxides is AB 2 O 4 . A and B are metal cations in general, which belong to the rst row of TMs, or group 2 or group 13 in the periodic table. Table 2 provides a summary of representative doped spinel oxides and their catalytic performances. The structure of spinel oxides is complicated, because of the presence of two crystallographic sites, namely octahedral oh and tetrahedral th sites, respectively, in a single structure. 22 Based on the arrangement of these crystallographic sites, they are classied into two types: one is spinel comprising A 2+ on the th site and B 3+ on the oh site, and the other one is inverse spinel containing A 2+ on the oh site and B 3+ on the th site. According to crystal eld theory, the TM ions present in the oh and th sites show different d-orbital splitting, i.e., t 2g e g for oh and e g t 2g for th. This difference in d-orbital splitting plays a signicant role in tuning the electronic properties of spinel oxides and optimizing OER performance. 71,72 Spinel oxides are highly stable in a basic medium with good electrical conductivity under harsh OER conditions. Mn-, Fe-, and Co-based spinel oxides are mostly reported as catalysts for the OER. 72 4.3.1 MMn 2 O 4 . Mn 3 O 4 is a normal spinel with Mn 2+ on the th site and Mn 3+ on the oh site, and its OER activity in the neutral medium is tuned by doping with various 3d-block elements such as Zn, Cu, Ni, Co, Fe, and Cr. 80 Among them, 5% Ni doping showed enhanced OER activity with better OER kinetics. Structural analysis suggests lattice distortion because Ni doping is a primary reason for improved OER activity. 80 Doping divalent metals such as Co, Zn, and Mg on the th site of Mn 3 O 4 spinel results in good electrochemical properties. Specically, Co doping is proven to be more effective in improving OER activity, and a tetragonal form of Co x Mn 3Àx O 4 shows a superior activity compared to the cubic phase. 107 Enhanced OER activity of the tetragonal phase is made possible by optimizing the binding energies of OER intermediates with the defective Co (À0.26) and Mn (À0.60) sites, and the bonding energy is much lower than those of the cubic phase sites, with suppressed Jahn-Teller distortion. In LiMn 2 O 4 , lithium is on the th site, and doping of the Mn site signicantly improves the OER activity.  78 From the mechanism, the superior OER activity is found to be linked with the reduction potential of the Mn(IV) cation, and greater reduction potential facilitates improved kinetics by directly driving surface oxidation.
4.3.2 MFe 2 O 4 . CoFe 2 O 4 is an inverse spinel, in which oneeighth of the th sites are occupied by Fe 3+ and one-half of the oh sites are occupied by Co 2+ ions. The presence of such an arrangement provides good electrical conductivity as a consequence of electron hopping between valence states of cations on oh sites. It thus facilitates good electrochemical activity and produces longer stability in electrocatalytic reactions. [109][110][111] Doping of Co and Fe sites signicantly inuences the electrocatalytic activity of CoFe 2 O 4 . For example, Ni and Mn are doped in CoFe 2 O 4 at the Fe site by stoichiometric mixing of Co 3 O 4 , Fe 2 O 3 , NiO, and MnO 2 using a solid-state synthesis. 109 The Nidoped one displayed the lowest overpotential of 260 mV, which is attributed to Co(II/III) on oh sites along with Co(III) on th sites of the inverse spinel. Similarly, Ni doping at the Co site was also evaluated by synthesizing hollow nanospheres via a hydrothermal route. The inverse spinel Ni 0.75 Co 0.25 Fe 2 O 4 showed the best OER activity, which is ascribed to the oxygen defects and cation distribution ratio on th and oh sites. 109  inverse spinels also showed improved OER activity. Doping pentavalent V into the lattice of NiFe 2 O 4 signicantly improves OER activity as well as its stability, which could be a consequence of the increased amount of Fe 2 O 3 in the host lattice, along with enriched surface porosity. 110 4.3.3 MCo 2 O 4 . Co 3 O 4 is a normal spinel, whose Co 2+ ions occupy th sites, while Co 3+ ions occupy oh sites. It has been a well-known electrocatalyst for the OER in the past few decades  lattice and observed that above or below the optimum level of Fe doping, fading of OER activity occurred. The 4.5% Fe doping caused superior activity, and the DFT study identied a considerable downshi of the Fermi energy level, originating from the optimized adsorption energy of reaction intermediates. By comparing Ni, Cu, and Li dopants, Vitanov et al. studied the effectiveness of doping in improving the OER activity of Co 3 O 4 spinel and noticed that doping with Li (Li x Co 3Àx O 4 ) produced a promising activity. 113 The raised activity of Li x -Co 3Àx O 4 spinel is ascribed to faster reaction kinetics owing to the formation of active sites with high transfer coefficients, and a shortened reaction pathway to generate oxygen. Tian et al. revealed the Cu doping effect on improving the OER activity of Co 3 O 4 nanosheets prepared on nickel foam, which outperforms the benchmark catalyst IrO 2 on nickel foam. 75 The high OER activity is found to have resulted from an increase in Co(III) species in the crystal lattice, along with highly uncovered active sites. Ding et al. demonstrated the improved OER performance of Co 3 O 4 nanosheets by creation of oxygen vacancies upon Zn doping, which even surpassed that of the Ir/C catalyst. 81 The stability and activity of Co 3 O 4 spinel could be signicantly enhanced in acidic electrolyte through Ag doping. Ag doping plays a crucial role in the growth of Co 3 O 4 nanowire arrays, improving the surface area and stability under harsh acidic conditions and OER activity by providing shortened reaction paths. 73 Ni x Co 2Àx O 4 is an inverse spinel, in which Ni(II/III) is present in 16d oh sites and Co(II/III) is present in both 8a th sites and 16d oh sites. By tuning the Ni doping level, this inverse spinel's OER activity signicantly increased, because the Ni 2+ /Ni 3+ ratio plays a key role in governing the conductivity, surface roughness, and density of active sites. For example, Zhang et al. systematically studied the inuence of an increase in the Ni dopant level to improve the overall OER performance of Ni x Co 2Àx O 4 . 112 When x ¼ 1, it shows superior OER activity. In this composition, Ni is in the +3 oxidation state. The enhanced OER activity is a result of the 0.27 eV upshiing of the valence band maximum toward the Fermi level, which creates a new hole. Such an electronic conguration favors the adsorption of OER intermediates. 112 The doping of Fe at the Ni site of NiCo 2 O 4 induces oxygen vacancies, which impacts the OER activity considerably. The Fe 0.1 Ni 0.9 Co 2 O 4 composition shows superior activity within the doped series, which is accredited to the presence of an optimum amount of oxygen defects for triggering electrocatalytic activity. 114 Huang's group reported doping Fe into the th site of the NiCo 2 O 4 lattice to improve its OER activity through a theoretical prediction. 33 As displayed in Fig. 8f, incorporation of Fe into the th site requires a lower energy (313.33 eV) compared to the oh site (313.58 eV) according to DFT calculations. Fig. 8g shows the DFT estimated H 2 O adsorptive ability where the Fe-doped one is superior to the pristine one and CoFe 2 O 4 , which is highly favorable for improving OER performance to increase active sites. Furthermore, X-ray absorption spectroscopy conrms the successful incorporation of Fe into the th site of NiCo 2 O 4 , which enhances the OER performance by increasing the population of Ni 3+ and Co 2+ ions in oh sites. Mn x Co 2Àx O 4 is also an inverse spinel, and Mn cations preferably occupy oh sites. Lankauf et al. described Mn x Co 2Àx O 4 spinel by adjusting the composition of Mn (x ¼ 0-2) using a so chemistry method, and the spinel with a moderate content of Mn shows better OER activity. 79 The high activity was attributed to the improved oxidizing ability and better electronic conductivity. Chi et al. noticed that the change in copper content in the Cu x Co 2Àx O 4 spinel could inuence its OER properties. 115 When x was increased from 0.7 to 0.9, more Cu 2+ ions displaced Co 3+ ions from oh sites. They occupy th sites, and thus the Co 3+ ion population on th sites was signicantly increased, resulting in superior OER performance.
Li and coworkers reported an increase in the OER performance of the NiCo 2 O 4 catalyst through double-exchange interaction (DEI). 72 The DEI is induced by synergistic creation of oxygen vacancies (V O ) via doping with Fe and creating a heterojunction with MoS 2 (MoS 2 /rFe-NiCo 2 O 4 ) as shown in Fig. 9a. Fig. 9b 116 The S-doping enhances the OER performance of Co 3 O 4 in terms of activity as well as stability in an alkaline medium. The formation of S-substituted CoO layers on the surface of Co 3 O 4 is a major reason for the superior activity. Theoretical calculations indicate that improving the electronic conductivity by S doping was benecial for overall electrochemical activity. Dual doping of Co 3 O 4 on both cation and anion sites with Ni and Se, respectively, resulted in a signicant increase in its OER performance. 84 The Se doping enhances the active sites on the surface of a catalyst by inducing oxygen vacancies, and Ni doping helps in generating highly active Co 4+ species.

Hydroxides
Hydroxides, oxyhydroxides, and layered double hydroxides (LDHs) of oxides of TMs, such as Ni, Co, Fe, Mn, and V, are highly active in the electrocatalytic oxygen evolution reaction. 86,117,118 They exhibit typical layered stacking conformation with proton sandwiching between layers, and the metal is centered in the octahedron surrounded by oxygen anions at the corners as MO 6 . 117 Some representative hydroxides and their catalytic performances are listed in Table 2.
Fe doping of TM hydroxides is most widely reported in the literature for enhancing their OER performance. 119 Specically, Fe doping in Ni(OH) 2 (ref. 94 and 120-123) and Co(OH) 2 (ref. 89 and 124-126) has been widely studied. In 1901, Edison, 127 for the rst time, noticed an increase in the OER activity of Ni(OH) 2 because of Fe impurity in the electrolyte in a Ni-based alkaline battery. 120,128 Because of oxygen evolution, the capacitance and cycle life of the battery were decreased considerably by the incorporation of Fe into the Ni(OH) 2 electrode. Later, researchers revealed that the OER activity of Ni(OH) 2 shortened the voltage window of the battery owing to the presence of Fe in the electrolyte, and they also found that the presence of a discrete amount of Fe (<1%) considerably contributed to augmentation of OER activity. 129 Corrigan demonstrated that the presence of 0.01% Fe impurity in electrolytes substantially decreased the discharge capacitance. 130 Boettcher's group studied the role of Fe doping in enhancing the OER performance of Ni(OH) 2 thin lms and provided more insights on catalytic reactions at the electrode/electrolyte interface. 122,[124][125][126] They witnessed an increase in conductivity of thin lms with systematically increasing the doping amount from 5 to 25%. Still, the CV curves show drastic changes, which signify the change in surface reconstruction with varying dopant levels. 122 Klaus et al. studied the Fe inclusion by aging the thin lms in an electrolyte with or without Fe impurities for 6 days. 131 The sample aged in the electrolyte with Fe impurity shows an increase in OER activity gradually from the rst day to the 6 th day. In contrast, the other one without Fe shows a decrease in activity gradually upon aging, and the enhanced OER activity is attributed to the increase in conductivity. Landon et al. noticed a rise in the coordination number of Fe in the OER environment using an in situ extended X-ray absorption ne structure (EXAFS) method. 132 They proposed that forming a new phase (NiFe 2 O 4 ) through reconstruction increased the coordination number of Fe. Nevertheless, the formation of NiFe 2 O 4 is rejected by other researchers with the aid of in situ Raman spectroscopy. 133 They did not notice any signals for NiFe 2 O 4 formation in the in situ Raman spectra (700 cm À1 ), even aer aging in KOH electrolyte with Fe at high OER potentials. In contrast, they observed two peaks which are characteristic of the NiOOH phase at 555 and 475 cm À1 , even in the presence of a high percentage of Fe.
Recently, Zhao's group studied the effect of doping Co and Fe into the lattice of Ni(OH) 2 for improving the OER performance, and they conducted electron energy-loss spectroscopy (EELS) characterization and Hubbard corrected density functional theory (DFT + U) calculations to understand the activity trend. 90 The Fe-doped sample shows superior OER activity to Co and Fe co-doped or Co-doped samples. 90 EELS analysis suggests electronic interactions between Ni, Co, and Fe dopants even aer surface reconstruction during the OER process. A DFT + U study reveals that the d-band center of the Fe-doped sample is much closer to the Fermi level than that of the others, enhancing the binding energy values of OER intermediates to an optimal value. Wang et al. demonstrated an enhancement in OER activity of Ni(OH) 2 by co-doping Mn and F elements. 91 A considerable increase in oxygen deciencies was found aer doping, which reduced the energy barrier for the key OER of *OH / *O. Ma's group doped a single atom of W 6+ into the lattice of Ni(OH) 2 and studied its electrocatalytic oxygen evolution behavior. 96 Interestingly, W doping signicantly enhances the OER activity compared to the pristine sample, and the superior activity is ascribed to the boosting of carrier migration at the interface of the electrode and electrolyte. Furthermore, a DFT study reveals that low spin W 6+ (d 0 ) stabilizes the O* and thus helps in the formation of O-O.
The augmentation of OER activity was also reported for the Co(OH) 2 catalyst by Fe doping. Boettcher's group reported the effect of Fe doping on the OER activity of electrodeposited lms of Co(OH) 2 . 124 They observed an enhancement of OER performance by 100 fold when the stoichiometric ratio of Co and Fe was 1 : 0.6-0.7. The reason for the superior OER activity is increased Co 3+ population and improved thin-lm conductivity with Fe doping. Unlike Ni(OH) 2 95 XRD patterns, as shown in Fig. 10a, display the complete disappearance of diffraction peaks of FeSe which was converted into Se-doped FeOOH. However, the electrochemical oxidation did not disturb the nanosheet morphology as revealed by SEM and TEM images in Fig. 10b and c, respectively. Fig. 10d shows the polarization curves of the as-prepared catalysts, and among them, FeOOH(Se)/IF delivers industrial scale 500 mA cm À2 current density at 348 mV overpotential. Tafel plots of these catalysts are presented in the inset of Fig. 10d, which shows that the FeOOH(Se)/IF catalyst possesses OER kinetics with the lowest Tafel slope value of 54 mV dec À1 . As a unary Fe-based catalyst, FeOOH(Se)/IF shows comparative activity and even outperforms Ni and Co-based OER catalysts as observed in Fig. 10e. Furthermore, DFT calculations revealed a considerable decrease in the energy barrier for the RDS from 3.20 eV to 1.45 eV aer Se inclusion, which accounts for the superior OER activity.

Other oxides
TM oxides, such as NiO, CoO, and MnO 2 , are widely reported as catalysts for the OER owing to their high durability and low cost. 5 Some other representative oxides and their catalytic performances are listed in Table 2. The electrochemical activity of these oxides was further improved by tuning their electronic structure through the inclusion of various dopants. For instance, lithium-doped NiO (Li x Ni 1Àx O) has shown unprecedented OER performance and even outperformed non-noble benchmark perovskite catalysts, such as BSCF and LaNiO 3 . 134 Zhang et al. studied the role of lithium incorporation into the lattice of NiO. 134 Fig. 11a shows the BET surface area normalized LSV curves of pristine and Li-doped NiO catalysts, and among them, the catalyst with a 1 : 1 ratio of Li and Ni shows 11 fold higher activity than pristine NiO. The inset of Fig. 11a presents cyclic voltammogram (CV) curves, which further reveals a positive shi in redox potential with an increase in Li content. A combinatorial study of the electronic structure of Li x Ni 1Àx O was conducted using X-ray absorption spectroscopy and synchrotron-based photoemission spectroscopy. The schematic illustration of covalency of Ni 3d-O 2p from DFT calculations is shown in Fig. 11b, which reveals the formation of a new hole state near the Fermi level, and the increased covalency of Ni 3d-O 2p that ensured higher OER activity. Enhancement in the OER activity of the NiO catalyst with Fe doping was rst observed by Corrigan in 1987, 130 and Fe doping into the NiO thin lm was found to be affected by the impurities present in KOH electrolyte. 135 Recently, Nocera et al. identied the essential role of Fe doping in improving the electrocatalytic activity of Ni oxides. 136 5-10% doping of Fe increased the population of Ni 4+ ions, as revealed by coulometric titrations. Consequently, O and Ni K-edge spectra indicated an increase in hybridization between Ni 3d and O 2p orbitals. Both experimental and computational studies showed that Fe 3+ is a Lewis  acid with pK a ¼ 2.2, and is the most acidic among TMs. Hence, the incorporation of Fe 3+ into Ni oxides helps in enhancing the acidity of OH x moieties coordinated with Ni, which increases the population of Ni 4+ ions by lowering the Ni 3+ /Ni 4+ reduction potential. 136 Ce-doped NiO x deposited on a Au substrate shows signicantly higher OER performance than pristine NiO x . 97 The enhanced OER activity is accredited to the modication of the local environment favorable for binding of OER intermediates.
A nickel-doped rock-salt-structured OER catalyst, Ni x -Co 1Àx O, was synthesized using a solid-state method. 102 The inclusion of a discrete amount of Ni into the lattice of CoO stabilizes the crystal structure and tunes its e g value. Within the doped series, the Ni 0.3 Co 0.7 O catalyst displays the highest OER activity with an e g value of $1.3. Huo et al. reported the improved OER performance of CoO nanoowers by doping with Zn. 106 A DFT study conrmed that OER active sites were increased by Zn doping, and the high surface area of nano-owers was another contributor to the improved catalytic activity. Like in NiO, Fe doping of CoO could also greatly enhance its electrocatalytic performance by increasing the number of OER active sites. 101 Ye et al. prepared V-, Fe-, Co-, and Ni-doped MnO 2 ultrathin nanosheets on carbon ber paper (CFP) using an electrodeposition method as shown in Fig. 12a, and studied their OER activity in a basic medium. 104 Fig. 12b and c show the doped MnO 2 ultrathin nanosheet/CFP composite electrode at low and high magnications, which shows very dense nanoclusters with 5 nm thick nanosheets. Fig. 12d shows the LSV curves of CFP, pure MnO 2 /CFP, metalion-doped MnO 2 ultrathin nanosheet/CFP, and IrO 2 /CFP composite electrodes in 1 M KOH with a scan rate of 5 mV s À1 .
The authors noticed that doping with these TM cations vastly enhances MnO 2 conductivity, resulting in higher catalytic activity compared to the pristine sample. The electrochemical stability of doped and pristine MnO 2 compared with IrO 2 evaluated at 20 mA cm À2 is shown in Fig. 12e. Doped MnO 2 ultrathin nanosheet/CFP shows signicant stability compared to the pristine sample with a slight increase in overpotential.
Yang et al. intercalated 3d block metal cations such as Zn 2+ , Cu 2+ , Ni 2+ , Co 2+ , and Fe 3+ into MnO 2 nanosheets to enhance OER activity. 103 Among them, the inclusion of Ni 2+ ions was more effective in enhancing the OER activity of MnO 2 nanosheets. The increased activity is associated with strengthening the Mn-O bond perpendicular to layered chains to promote the catalysis between two Mn sites. Oxygen vacancy formation in MnO 2 nanosheets by Co doping is demonstrated by Zhao et al. 99 Co doping enhances the OER activity of MnO 2 considerably by enhancing its conductivity and decreasing the energy barrier for adsorption of OER intermediates on the catalyst surface. Interestingly, a small amount of cobalt doping (3.47%) could activate the OER activity of ZnO. Theoretical studies suggest that Co activates inherently inactive adjacent Zn sites by increasing conductivity, which act as OER active sites for adsorption of OER intermediates. 100 Co doping also activates the OER activity in CuO. Sun et al. demonstrated a multifold enhancement of OER activity in a basic electrolyte aer doping with Co on a Cu foam substrate, 98 which was related to a change in the electronic structure of CuO by Co doping. Meyer et al. reported the fabrication of an OER catalyst by modifying uorine-doped tin oxide (FTO) by Co 2+ single-atom doping. 137 Xray photoelectron spectroscopy (XPS) and TEM reveal that doping Co 2+ into FTO by replacing Sn 4+ results in oxygen vacancies. The as-modied FTO electrode shows water oxidation activity at pH 7.2 with a turnover frequency of 29 000 per cobalt site, at a current density of 0.16 mA cm À2 during an endurance test for 2 h. Furthermore, the current density was enhanced by 5 fold by utilizing nanoporous FTO in place of planar FTO. Wang et al. reported Ba doping of Y 2 Ru 2 O 7 pyrochlore oxide on the yttrium position (YBRO) as an efficient catalyst for the OER in an acid electrolyte. 105 Among the doped samples, YBRO-0.15 shows superior activity, and the OER activity decreases in the following sequence: YBRO-0.15 > YBRO-0.1 > YBRO-0.25 > YBRO-0.4 > YRO > RuO 2 . The enhanced activity is attributed to the oxygen vacancy-rich surface because of partial replacement of Ba 2+ with Y 3+ and simultaneous formation of multivalent Ru (Ru 5+ /Ru 4+ ) species.

Doping non-oxide catalysts
Non-oxide catalysts include metal suldes, metal selenides, metal phosphides, metal nitrides, and metal carbides. These catalysts are pre-catalysts, and the actual catalysts are the oxides of the corresponding cations generated on the surface during electrochemical water oxidation.

Metal suldes
Transition metal suldes are widely explored in electrocatalysis due to their high stability in acidic and alkaline electrolytes and abundant redox chemistry. 138 Some doped metal suldes and their catalytic performances are listed in Table 3. Like in metal oxides and hydroxides, Fe doping is effective in signicantly improving the OER activity of TM suldes. 139,140 For example,  150 Zn, 160 Ce, 147 and Al 146 is also found to enhance the OER activity of metal suldes. For instance, Lin et al. fabricated Ni-doped FeS nanosheets as an OER catalyst in alkaline electrolyte through a one-pot synthesis approach. 152 The Ni-doped FeS nanosheets showed signicant OER activity compared to commercial IrO 2 with 228 mV overpotential at 10 mA cm À2 current density, by the consequent doping effect of exposing more active sites. Similarly, Kuila et al. synthesized binder-free self-standing Ni-doped FeS OER catalysts on Ni foam via an electrodeposition method. 178 The Ni doping amount is controlled by varying the electrodeposition time, and the electrode fabricated by electrodeposition at 0.9 V for 30 min shows superior activity among them. The change in the electronic structure and charge carrier density via the incorporation of Ni into the lattice domain of FeS signicantly improves its electrocatalytic activity. Likewise, Wen and coworkers incorporated Ce into the lattice of Ni 3 S 2 via a one-step electrodeposition method. 179 Doping with Ce preserves the original structure and morphology. Still, it promotes the OER activity by regulating the active sites and chemical states on the surface of Ni 3 S 2 . Yang et al. signicantly enhanced the electrocatalytic ability of WS 2 by incorporating nearly 1% Co into its lattice. 148 The overpotential was decreased from 492 to 303 mV aer incorporation of Co, and the improved activity was ascribed to the presence of Co(II/ III) species that improved the binding of reactants and facilitated the formation of the *OOH intermediate. Darr and coworkers optimized the OER activity of bimetallic cobaltnickel suldes by doping with various TMs, such as Ag, Fe, Mn, Cr, V and Ti. 150 Among them, Mn-incorporated cobalt-nickel sulde (Ni 1.29 Co 1.49 Mn 0.22 S 4 ) shows enhanced activity through tuning of the electronic structure of the host by Mn(III) cations. Zn-doped Ni 3 S 2 nanoarrays on nickel foam grown by Sun et al. have better OER activity than their pristine counterpart, with excellent stability for 20 h at 300 mV overpotential. 160 The extended endurance and comparable activity were caused by modulation of the electronic structure of the Ni 3 S 2 catalyst through Zn incorporation.
Recently, researchers reported incorporating non-metals, such as N, P, and F, into the lattice of metal suldes to improve their electrocatalytic capability. 149,151,[153][154][155][156][157][158][159]180 N doping of a NiS 2 catalyst using NH 3 $H 2 O as a precursor considerably increased its water oxidation ability by overcoming the intrinsic limits of reaction kinetics through optimizing d-states. 153 3D Ndoped Ni 3 S 2 nanoakes grown on Ni foam show superior durability for 20 h at 1.7 V with $93% retention of their initial activity. 154 The experimental results indicate that N doping plays a key role in the quick discharging of OER intermediates and improved intrinsic activity. 155 Dou et al. found that a Co 9 S 8 / graphene hybrid, doped with N using NH 3 plasma etching, showed a signicant improvement in its OER performance compared with a pristine sample. 151 The synergetic effect of Ndoping and surface etching considerably tuned the electronic structure of the Co 9 S 8 /graphene hybrid and thus enhanced the number of active sites on the surface. He et al. synthesized 3D self-supported and P-doped Ni 3 S 2 on Ni foam by a one-step hydrothermal process by simultaneous phosphorization and sulfuration. 157 The as-fabricated electrode exhibits enhanced OER activity with an overpotential of 256 mV at 10 mA cm À2 . The inclusion of P anions optimizes the electronic structure and provides facile adsorption of reactants and thus enables a faster mass transfer. Similarly, P-doped bimetallic nickel-iron sulde nanosheets on Ni foam were synthesized via a one-pot process. 159 The incorporated P considerably enhances the electrochemical surface area and electrical conductivity by optimizing the adsorption of OER intermediates. F-doped CoS 2 on Ni foam was synthesized using NH 4 F as a uorine source by a hydrothermal process. 180 The F dopant signicantly enhances the wettability of the catalyst, and thus facilitates the maximum interaction with electrolyte in both alkaline and neutral media.

Metal selenides
Some doped metal selenides and their performances are listed in Table 3. The electrocatalytic behavior of metal selenides depends on the number of exposed edge sites. 21 Edge sites are step-like apertures or defects generated at the basal plane because of doping. 181 Besides the edge sites, the electrical conductivity and intrinsic activity of metal selenides are enhanced by doping; specically, doping with TMs signicantly improves the OER activity of metal selenides. 169,171,173,174 For example, Zare et al. designed a self-supported electrocatalyst on Ni foam by doping Fe into the lattice domain of CoSe 2 , which only requires 256 mV overpotential to attain 100 mA cm À2 current density. 161 Here, Fe improves the intrinsic electrical conductivity, and thus lowers the energy barrier for the OER. Zhao and coworkers improved the OER activity of CoSe 2 nanobelts by simultaneous Fe doping and Co vacancy creation, as shown in Fig. 13a. 182 LSV curves of CoSe 2 , CoSe 2 -D Fe , CoSe 2 -D Fe -V Se , and CoSe 2 -D Fe -V Co catalysts, normalized by electrochemical double-layer capacitance, are shown in Fig. 13b, and suggest that CoSe 2 -D Fe -V Co has superior OER activity among the catalysts. Fig. 13c displays the FeCoOOH with Co vacancies and the OER mechanism occurring on its surface, which involves three OER intermediates (*OH, O*, and *OOH) and four protoncoupled electron transfer steps. The Gibbs free energy proles of this catalyst show different catalytic mechanisms for Co and Fe sites. The transformation of OH* to O* potentially limits the activity of Co active sites owing to the large Gibbs free energy difference (DG 2 ). In contrast, the conversion of O* to *OOH is the RDS (DG 3 ) in case of Fe sites, due to the high free energy difference. Theoretical simulations identied that the cobalt dimer (Co 2 ) is the most active site for the OER, and is present adjacent to the Co vacancy near the surface Fe site. The combinatorial effect of Co vacancy and Fe doping signicantly intensies the OER performance by reducing the peak intensity near the Fermi level, so the binding of *OH with active sites is weakened. Zou et al. found that the minimal amount of Fe (7.4%) inclusion into the NiSe nanorod/nanosheet lattice signicantly improved its OER activity. 170 The Fe-doped NiSe only required 291 mV overpotential to achieve 1000 mA cm À2 current density. This superior activity is accredited to the increased electrical conductivity, active sites, and mass transfer. Xu et al. fabricated hierarchical Fe-doped Ni 3 Se 4 on Ni foam using a hydrothermal method, and NiFe LDH was used as a precursor material. 172 The doped sample shows superior OER activity than IrO 2 and pristine catalysts, because of its higher electrical conductivity. Mohanty et al. reported an increase in OER activity of nickel selenide, owing to the synergetic effect of Fe doping along with the creation of Se vacancies by chemical reduction with NaBH 4 . 176 Computational studies suggest an increase in DOS near the Fermi level because of Fe doping and Se vacancies, which helps in decreasing the energy barrier for OER intermediates.
Besides Fe, Co was also used as a dopant to improve the OER performance of nickel selenides, 165 and Co-doped NiSe nano-owers grown on Ni foam exhibited considerable activity compared to their pristine counterpart. Here, Co doping generates more OER active sites via tuning the electronic structure of NiSe. Wang et al. showed MOF-derived Co-doped nickel selenide on Ni foam and analyzed its OER activity in alkaline electrolyte. 164 The as-designed catalyst exhibits an overpotential of 275 mV at 30 mA cm À2 , and the high activity is ascribed to an increase in electrical conductivity because of Co doping. Zhang et al. co-doped Ni and Fe into the matrix of CoSe 2 , and noticed that the catalyst with a stoichiometric formula of Ni 0.04 Fe 0.16 Co 0.8 Se 2 has remarkable OER activity among the prepared catalysts. 175 Ni doping does not alter the OER mechanism but reduces the energy barrier for the RDS and keeps Co-active sites intact, and Fe doping improves the charge transfer kinetics by transferring the active sites from Co to Fe. FeSe 2 was doped with various transition elements (Co, Ni, Mn, and Cu) by Wei et al. 168 Among these dopants, Cu is found to be more effective and shows enhanced OER activity. The theoretical and experimental results reveal that Cu doping optimizes the electronic structure of the FeSe 2 lattice to enable faster electron transfer and thus boosts the reaction kinetics of the OER. In cobalt selenide, the Cu doping results in a phase transformation from orthorhombic CoSe 2 to monoclinic Co 3 Se 4 , causing improved OER performance of cobalt selenide. 167 The electronic structure change in cobalt selenide by Cu doping is evident from X-ray photoelectron spectroscopy and theoretical simulations. DFT calculations further reveal that because of structural transformation from orthorhombic CoSe 2 to monoclinic Co 3 Se 4 , and the Gibbs free energy for the RDS was drastically decreased from 2.64 to 1.90 eV.
The minimal amount of Ag (1%) doping into the lattice domain of CoSe 2 nanobelts signicantly increases its electrical conductivity without much loss of active sites (Co). 162 Ag-doped CoSe 2 nanobelts exhibit high OER activity relative to pristine CoSe 2 nanobelts, because of high electrical conductivity and active site retention. Incorporation of low-valence Zn into the CoSe 2 lattice domain triggers subtle changes in the lattice and increases the number of active sites for the OER. 177 Theoretical simulations suggest that Zn inclusion decreases the kinetic energy barrier for the adsorption of OER intermediates by encouraging adsorption on Co sites and neighboring Zn sites. Boron doping into the lattice of ternary pyrite selenide (Fe 5 -Co 4 Ni 20 Se 36 ) decreased its overpotential from 543 to 279.8 mV at 10 mA cm À2 , with the Tafel slope lowered from 161 to 59.5 mV dec À1 . 163 Fig. 14a and b show the schematic representation of NiSe 2 and Fe 5 Co 4 Ni 20 Se 36 B x crystal structures, and the doped B atoms are incorporated into the interstitial positions and some B atoms even bond with metals (Ni, Co, and Fe). The proposed OER mechanisms on pristine and doped pyrite are shown in Fig. 14c and d, and reveal that B doping facilitates an easy discharge of *OOH from the active sites, as a consequence of lower Gibbs free energy. The PDOS reveals that B doping signicantly reduces the total d electrons around metals, enhancing the interaction between them and thus weakening the bond between *OOH and active sites. Recently, anion doping, i.e., doping on the Se site, was also found to be bene-cial in enhancing the OER activity. For example, Xu et al. reported doping of sulfur on the Se site, enhancing the OER activity of bimetallic cobalt iron selenide Co 0.75 Fe 0.25 (S 0.2 Se 0.8 ) 2 as compared to non-doped Co 0.75 Fe 0.25 S 2 and Co 0.75 Fe 0.25 Se 2 , since charge transfer and electrochemical surface area were increased by doping. 166

Metal phosphides
In the past ve years, metal phosphides have emerged as one of the most promising types of catalyst for water oxidation, even though the rst metal phosphide catalyst was reported as early as in 1990. 183 Some doped metal phosphides and their performances are listed in Table 4. Incorporation of a second element into the matrix of a metal phosphide could enhance the OER performance, in terms of both activity and stability. 184,185 For instance, Chen et al. doped Fe into the matrix of CoP ultrathin nanosheets grown on Ni foam, which exhibited a very low overpotential of 67 mV at 10 mA cm À2 in 1 M KOH. 186 The excellent OER activity is attributed to the interference of Fe/Co/ Ni in the low current density region. Post OER characterization revealed the complete absence of P because of the formation of CoOOH active species during the OER. Baek and coworkers synthesized Fe-doped porous CoP polyhedrons from MOFs by varying the Fe/Co ratio. 187 Among them, Co 0.68 Fe 0.32 P had better OER activity over its pristine counterparts, i.e., CoP and FeP. Fe doping accelerated the discharge of OER intermediates by optimizing their adsorption energy and stabilized the P species during contact with air. Peng's group initially conducted theoretical simulations to predict the Fe inuence on the water oxidation capability of Ni 2 P by constructing partially oxidized surfaces of Ni (111), Ni 2 P (111), and Fe-doped Ni 2 P (111) models. 188 The Gibbs free energy studies reveal that the Fedoped sample requires an energy 0.14 eV lower than that of Ni 2 P for the RDS (O* to *OOH). Experimental results were in accordance with the theoretical calculations, and they observed the formation of Fe, P co-doped g-NiOOH during the electrochemical water oxidation reaction from the Fe-doped Ni 2 P. Fe, P co-doped g-NiOOH is a good catalyst for improved OER activity, and Fe doping primarily enhances the active sites for the OER. Xiang et al. tuned the electronic structure of vanadium phosphide by including Fe into its lattice domain, which demonstrated remarkable OER activity. 189 Among all the samples, the sample with a 1 : 3 ratio of Fe/V has the highest activity. XPS analysis identies increased valency of V because of the incorporation of Fe, which optimizes the adsorption of *OH on active sites. Fe doping of bimetallic phosphides was also found to signicantly contribute to improving electrochemical water oxidation.
Like monometallic phosphides, bimetallic phosphides can also convert into their respective oxyhydroxides at high anodic potentials. Fe-doped NiCoP converts into Fe-doped NiOOH and CoOOH, and Fe doping improves the OER activity by accelerating the discharge of OER intermediates. 190,191 Simultaneously, the conductive metallic phosphide support improves OER kinetics by promoting the proton transfer. Tsang and coworkers studied the comparative doping effects of TMs (Co, Fe, Mn, and Mo) on the OER activity of NiP. 192 Among them, the Fe doped sample was found to have more prominent activity since Fe is able to tune the electronic structure of Ni to interact optimally with OER intermediates.
Mn doping of cobalt phosphide and nickel phosphide improves their catalytic activity for electrochemical water oxidation. In cobalt phosphide, Mn doping enhances the electrochemical surface area by increasing the number of exposed active sites. 194 The optimal Mn/Ni ratio of 0.053 is found to provide superior activity, as the Ni 2 P electronic structure is optimized. 195 Al doping was also used to enhance the electrochemical surface area of cobalt phosphide by dissolution during the OER. Al-doped cobalt phosphide shows a fourfold increase in electrochemical surface area compared to the pristine sample by the formation of Al/Co spinel oxide and partial leaching of Al. 193  Along with doping of metals, doping of non-metals such as N, O, and S can also enhance the electrocatalytic activity of phosphide catalysts for the OER. 196,[198][199][200] N doping in the anionic site (P) of NiCoP improved its OER performance, with better activity compared to NiCoP and NiCoN. The high activity was induced by the formation of abundant oxygen vacancies, improving the charge transfer during oxygen generation. 196 Likewise, oxygen and sulfur were also doped on the anionic sites to increase the OER activity of metallic phosphides. Oxygen doping of cobalt phosphide generates abundant defects and electroactive sites on the surface by optimizing the adsorption energy of reactants. 199 In the case of bimetallic nickel-iron phosphide, the OER activity is enhanced by oxygen doping on account of an efficient decrease in the energy barrier of OER intermediates and products. 197 Electronegative sulfur doping in the anionic site of cobalt phosphide signicantly improved its OER activity, owing to the tuning of the electronic structure and increased number of active sites. 200

Metal nitrides and carbides
Metal nitrides belong to an interstitial compound family, which contains nitrogen atoms on the interstitial sites of parent compounds. The presence of N atoms on the interstitial sites increases the distance between metal atoms; thus, it tunes the d band structure. 20 Some doped metal nitrides and carbides and their performances are listed in Table 4. Recent studies on the electrochemical activity of metal nitrides suggest that they are highly active for the OER owing to their high electrical conductivity. 204,205 However, the OER activity and stability were further improved by tuning their electronic structure by doping with TMs and non-metals. For instance, Mn-doped CoN nanowires grown on carbon cloth exhibit good OER activity in alkaline and neutral electrolytes. 203 Among the doped catalysts, Mn 0.15 Co 0.85 N shows superior activity, and Mn doping considerably increases the charge transfer kinetics and facilitates the optimized discharge of OER intermediates. Zhou et al. investigated the doping effect of various metal atoms, such as Zn, Ni, Fe, and Mn, on the OER activity of cobalt nitride nanosheets. 202 The Fe doped one has better OER activity, and XPS analysis suggests more Co 4+ OER active species because of the doping effect of Lewis acid Fe 3+ .
Cobalt doping into the lattice of bimetallic MoNi nitride nanowires shows a considerable increase in their OER activity by reduced charge transfer resistance and augmentation of the electrochemically active surface area. 201 The dual modulation in the electronic structure of Mo and Ni was induced by cobalt doping that improved the electrical conductivity for highly active OER sites. Liang et al. theoretically studied the doping effect of rst-row TMs (Sc to Zn) for improving the OER activity of Al/GaN. 213 The inclusion of transition metals from Sc to Fe resulted in low OER activity because they bonded to OER intermediates too rmly and resulted in high overpotentials. Meanwhile, Cu and Zn also displayed no or low OER activity, as they attached to the intermediates very weakly. Only Ni doping had signicant OER activity because it could switch from a high spin state to a low spin state and stabilized the OOH adsorbate.
Tong and coworkers synthesized P and Co dual-doped vanadium nitride nanowires using a combined hydrothermal and annealing approach. 214 The dual-doped vanadium nitrate catalyst showed better activity and durability for the OER, compared to the single-doped or pristine sample, or IrO 2 . Dual doping of Co and P synergistically tunes the number of active sites on the catalyst surface by moving the d states of vanadium closer to the Fermi level.
Zhang and coworkers demonstrated an increase in the OER performance of bi-metallic nitrides by co-doping with phosphorus and uorine. 204 Fig. 15a shows the schematic representation of the fabrication process for P and F co-doped nickel-cobalt nitride (PF/Ni 1.5 Co 1.5 N) nanorods using an ionic liquid. The PF/Ni 1.5 Co 1.5 N prepared without using any ionic liquid results in a lamellar structure, as displayed in Fig. 15b. The SEM and elemental mapping analysis results of PF/ Ni 1.5 Co 1.5 N nanorods are presented in Fig. 15c and d, which indicates the uniform distribution of N and F dopants all over the catalyst. Fig. 15e and f show the IR uncompensated LSV curves and comparison of h 10 values (potentials required to reach 10 mA cm À2 ) for PF/Co x Ni 3Àx N catalysts. Electrochemical studies reveal that the PF/Ni 1.5 Co 1.5 N catalyst shows augmented OER activity as well as facile reaction kinetics among the prepared catalysts. DFT calculations suggest that co-doping of P and F atoms into the lattice of the Ni 1.5 Co 1.5 N catalyst considerably decreases the free energy barrier for OER intermediates, thus helps in improving the electrocatalytic performance.
Recently, researchers found that metal carbides, specically transition metal carbides, can be applied as electrocatalysts for the OER owing to their high stability. 215 Some doped metal carbides and their performances are listed in Table 4. The activity of TM carbides was improved by including one or more foreign dopants into their lattice domain. Yan et al. improved the OER activity of Ni 3 C by doping 2% atomic iron, which yields an overpotential of 275 mV at 10 mA cm À2 . 211 The optimized doping of Fe tuned the electronic structure of Ni and thus improved the number of active sites on the surface. Similarly, Jia et al. tuned the water oxidation ability of Ni 3 C via incorporation of a small amount of Co into it using a MOF-derived method. 209 The as-fabricated catalyst had better activity than the IrO 2 catalyst in the alkaline electrolyte, and the improved activity was associated with the active surface sites through optimization of electronic properties. In general, the Mo 2 C catalyst is inactive for the OER, but doping 10% cobalt into the Mo 2 C catalyst was found to increase its OER performance through signicant enhancement in the density of active sites of the catalyst. 208 Zhao et al. optimized the electronic structure of Co 3 C by doping with low electronegativity vanadium element. 212 Specically, V 0.28 Co 2.72 C shows superior activity by forming local negative Co centers through the intake of electrons from vanadium dopants. Xu et al. doped TMs such as Mn, Co, and Ni into the matrix of Fe 3 C and evaluated its OER performance. 210 They noticed that Co-doped iron carbide had enhanced activity, and the Gibbs free energy was estimated in the different steps in the OER process. The free energy proles of the Co and Fe sites were compared, and the Gibbs free energy required for the Co site (2.35 eV) was lower than that for the Fe site (2.75 eV), for the conversion of *O to *OOH (RDS). From the comparison, Fe has a stronger affinity towards *O compared to Co, hence doping of Co into the Fe 3 C facilitates the discharge of intermediates, thus lowering the potential.
Doping of metal carbides on the anionic site also improves their activity for electrocatalytic oxygen evolution. Generally, WC is a poor OER catalyst; however, when the carbon site is doped with nitrogen, it shows even better OER performance than IrO 2 in an acid electrolyte. 216 The N doping helps in accelerating the discharge of intermediates and thus improves activity, but its poor stability is still a concern. Dual doping of N and B on the anionic site of Co 3 C considerably increases its water oxidation ability compared to the pristine or single-doped sample, or IrO 2 . The synergetic effect of N and B dopants improves both activity and stability by boosting the charge transfer kinetics of intermediates. 207 Likewise, Lee and coworkers synthesized dual-doped N and B into the matrix of Mo 2 C embedded in the B, N doped carbon network (B,N:Mo 2 -C@BCN), which shows a considerable increase in OER activity compared to the mono-doped or pristine sample, or IrO 2 . 206 The signicant improvement in electrocatalytic performance is accredited to the formation of active B-and N-doped molybdenum oxide at high anodic potentials.

Tuning carbon-based catalysts
Although metal-based catalysts are employed in electrocatalytic OER applications widely because of their promising capabilities to deliver efficient catalytic activity by variable oxidation states, 217 these catalysts suffer from relatively poor endurance, low selectivity, susceptibility towards potential gas poisoning, and hazardous environmental effects. In 2009, the rst carbonbased catalyst was reported by Dai's group in fuel cells as an alternative to the platinum catalyst. 218 Since then, carbon-based catalysts, particularly those doped with various heteroatoms or metals, have received extensive attention and tremendous progress has been made in electrocatalysis. 219,220 Table 5 provides a summary of some representative doped carbonbased electrocatalysts and their catalytic performances.

Graphene
Graphene is a thin layer of single atoms packed with sp 2 hybridized carbon atoms in a hexagonal lattice. It is considered either a zero-bandgap semiconductor or conductor because of its symmetric band structure. 244 Hence, pristine graphene shows low or no catalytic activity; however, the unique structural exibility of graphene allows extensive functionalization via doping to improve catalytic activity. The dopants could interrupt the sp 2 hybridized network and induce the formation of sp 3 in the lattice, resulting in an electronic density change that imparts catalytic activity to it. 244 For instance, doping of N into the lattice of graphene disrupts the charge distribution and spin density of neighboring carbon atoms and, consequently, imparts OER activity. Yang et al. conducted a study to gain mechanistic insights into N-doped graphene in the OER. 230 They identied the formation of electron-withdrawing pyridinic N functionalities, which is the key reason for the superior OER performance of N-doped graphene. Xia's group studied the active sites in N-doped graphene nanoribbons combined with other dopants, such as B, Cl, P, and S, using DFT calculations. 245 Gibbs free energy analysis for the OER on doped graphene nanoribbons suggests that the conversion of *O to *OOH is the RDS. The co-doping shows better activity than mono-doping, and the synergetic effect of co-doping is mostly related to the distance between them. In another report, N doping was done by using polyaniline and graphene oxide as precursors, which produced nanoporous layered and N-doped graphene that exhibits enhanced OER activity. 246 Nearly 51% quaternary N plays a signicant role in the considerable increase in OER activity. Wang et al. demonstrated a signicant improvement in the OER performance of graphene by P doping. 241 They found that the stronger oxygen adsorbing ability of P, which was present at the edges of graphene, played a crucial part in stabilizing OER intermediates.
Besides single atom doping of graphene, dual-and multiatom doping were also found to deliver signicant results in improving OER activity by their synergistic effects. Zhao and Xia studied several combinations of dual doping such as N-N, N-P, N-S, N-B, and N-Cl in optimizing the OER activity of graphene using rst-principles modeling. 247 They suggested that optimizing the distance between dopants, selecting the right combination of dopants, and efficient utilization of edge effect are the vital factors in designing a highly efficient electrocatalyst. For example, the co-doping of P and N into the graphene lattice results in graphene formation with P, N population at the edges, as a consequence of the large atomic size of P that always moves away from the planar surface. This co-doped catalyst exhibits a signicant improvement in OER performance, and the elevated activity is attributed to the synergetic effect of both dopants, the high oxidation ability of the P-site that initially adsorbs O 2 , and the N-site that stabilizes the OER intermediates and promotes desorption. 248 Qiao's group synthesized N and S dual-doped graphene microwires using an ionic liquid assisted method, which showed an overpotential of 310 mV at 10 mA cm À2 , stability for 25 h and 95% faradaic efficiency. 237 Moreover, this electrode was employed as an efficient cathode for Zn-air batteries, and this superior OER activity is attributed to the numerous active sites created by C-N-S spin density. Furthermore, Zhang and Dai demonstrated tri-doping of graphene with N, P and F atoms, yielding a multifunctional electrocatalyst. 226 The N and P doping roles in improving OER performance were elaborated. Still, it is inconclusive how F inclusion into the graphene lattice contributes to the electrocatalytic activity. Li et al. demonstrated a signicant improvement of the OER performance of graphene by including single nickel atoms. 235 The single-atom doped graphene displayed 50 h of long-term stability at a very high current density of 115 mA cm À2 . This unusual activity is accredited to the bonding between single nickel atoms and oxygen sites of graphene. Chen et al. doped Ni single atoms on N-doped nano-porous graphene via a chemical vapor deposition (CVD) method, followed by chemical etching. 234 The dual-doped catalyst showed superior OER activity to other catalysts, and even outperformed the IrO 2 catalyst. The DFT calculations suggest that Ni, N co-doping signicantly decreased the stability difference between O* and *OH intermediates by 0.3 eV.

Carbon nanotubes (CNTs)
CNTs are 1D structural allotropes of carbon that consist of hexagons of carbons arranged in a concentric fashion. CNTs possess good electrical-thermal properties due to their unique hollow geometry and conjugated long-range p electron structure. 249 However, CNTs in their pristine form exhibit poor electrocatalytic properties. To improve the electrocatalytic properties, CNTs need to be functionalized with several dopants. 250 This functionalization can be done through two methodologies, i.e., covalent and non-covalent methods. The covalent functionalization proceeds via forming a covalent bond between the functional group and CNTs, which could disrupt the unique structure of CNTs. In the non-covalent method, the functionalization happens through physical processes such as electrostatic, hydrogen bonding, p stacking and van der Waals interactions, which leaves the CNT structure undisturbed. 249 For example, Tian et al. doped nitrogen into CNTs and synthesized highly active CNT@NCNT for the OER as a catalyst. 251 The superior OER activity is ascribed to the highly exposed integrated N atoms present on the surface with the combined mechanical stability of the undisrupted inner walls of CNTs. Likewise, the Kallio group doped nitrogen into multiwalled CNTs (MWCNTs) and observed a considerable increase in OER performance aer doping, compared with pristine MWCNTs. 221 The signicant OER activity of N-doped MWCNTs is induced by pyridinic N functionalities that are electronwithdrawing. Wang's group designed a generalized approach to make highly efficient hollow N-doped CNTs as a catalyst for the OER from a MOF. The remarkable electrochemical activity is attributed to the synergistic effects from interconnected Ndoped CNT structures and chemical compositions with high mechanical stability. 229 Qiao's group synthesized dual-doped CNTs with nitrogen and sulfur inspired by the features of polydopamine (PDA), as depicted in Fig. 16a. 236 N and S doped CNTs were prepared through a two-step "gra-and-pyrolyze" route using PDA. Initially, PDA was deposited on the surface of multiwalled CNTs and sulfur was incorporated at room temperature via chemical graing, and subsequent pyrolysis at 700 C resulted in a N and S uniformly co-doped carbon framework with enhanced Sdoping efficiency. Even aer doping and pyrolysis, the nanotube morphology of CNTs was preserved, as shown in Fig. 16b. The OER performance of pristine and doped CNT catalysts is displayed in Fig. 16c, which shows that the N, S-CNT catalyst exhibited superior OER activity among them with an overpotential of 360 mV and a Tafel slope of 56 mV dec À1 (Fig. 16d). Theoretical studies show a considerable increase in spin-charge density difference among carbon atoms with doping; these carbon atoms act as electrocatalytically active sites. As a consequence, the apparent OER activity improved with dual doping compared to single doping. Liu's group has extensively studied the doping effect of oxygen on the OER performance of CNTs by synthesis via two different methods, i.e., oxygen plasma oxidation (O-CNTs) and thermal reduction (R-CNTs). 238 O-CNTs exhibit enhanced OER activity over R-CNTs and pristine CNTs, and DFT calculations reveal that both O-CNTs and R-CNTs have the same active sites, i.e., carbon near C]O. The high-water oxidation ability of O-CNTs was thought to be possible due to the presence of a high density of active sites, with high E a,app at the specic reaction temperature.
Lu et al. oxidized the surface of MWCNTs through hydrothermal annealing and electrochemical activation, which results in a highly active OER catalyst. 239 The superior OER activity originated from the formation of carbonyl functionalities on the outer wall, which altered the electronic structure of neighboring carbon atoms and facilitated adsorption of OER intermediates. Suib et al. fabricated a highly stable metal-free catalyst at high anodic potentials by successive two-step doping of sulfur into the MWCNTs, displaying two orders of magnitude higher turnover frequency (TOF) compared to the Ir/ C catalyst. 243 The enhanced OER activity was caused by the inclusion of heterocyclic sulfur into the hexagonal rings of CNTs. Wu and co-workers developed a single-atom doped catalyst for the water oxidation reaction by coating CNTs with Fe salt and 2,2 bipyridine, followed by pyrolysis and acid leaching, yielding FeN x species on the S and N co-doped CNT catalyst. 242 This catalyst exhibits superior OER activity in alkaline medium, as a result of high exposure of Fe atoms and synergetic effects of S and N doping.

Other carbon-based catalysts
In addition to graphene and CNTs, other forms of carbon-based catalysts were also reported for OER activity. Graphitic carbon nitride (g-C 3 N 4 ) has a C-N structure with graphite like sp 2 bonding, which exhibits semiconducting behavior due to pconjugation. 252 Like graphene and CNTs, g-C 3 N 4 is also inactive in its pristine form, and tuning its electronic structure by doping non-metal or metal elements signicantly improves its water oxidation ability. 233,240 For instance, Huang et al. dualdoped N and B into 3D porous g-C 3 N 4 and observed superior OER activity in alkaline solution, compared to N-doped or N and P co-doped 3D g-C 3 N 4 . 253 It was found that the N and B co-doped g-C 3 N 4 has a high population of pyrrolic and pyridinic nitrogen for enhanced OER. 222 Based on DFT calculations, Jiang and coworkers proposed doping single transition elements such as Pt, Pd, Co, Ni, and Cu into carbon nitride (CN) that could lead to a cost-efficient, highly efficient OER catalyst. 254 Specically, the electrocatalysts of non-noble metal Co/Ni@CN and Co-O@CN exhibit excellent activity and function with extremely low overpotential. Qiao's group conducted the theoretical study and experimental validation on cobalt inclusion on g-C 3 N 4 for OER activity in alkaline medium. 223 The combinative evaluation of theoretical and experimental methods reveals that metal-nitrogen M-N 2 coordination in the matrix of g-C 3 N 4 is the origin of high electrocatalytic activity compared to the pristine sample. Song et al. studied the effect of dual metal doping in enhancing the OER activity, and doped Ni and Fe metals in tri-s-triazine units of g-C 3 N 4 enveloped by CNTs. 232 Synchrotron-based X-ray absorption with combination of atomic microscopy indicated that the electronic conguration of both Fe and Ni was recon-gured, and the valence state of Ni was increased, and thus more OER active sites were created compared to single doping. Sun and coworkers dual-doped Ni and Fe into the framework of polymeric carbon nitride (Ni, Fe@PCN), in which superior activity was found. 233 Computational studies suggest that the enhanced OER activity of NiFeOH@PCN is a consequence of the vastly reduced energy barrier (0.10 or 0.22 eV for U 1 4 1.58 V) through synergetic contribution of adjacent Ni and Fe atoms.
In 2013, Zhao et al. reported N-doped carbon nanomaterials as metal-free OER catalysts other than graphene and CNTs for the rst time. 227 In alkaline electrolyte, the catalyst exhibits 380 mV overpotential at a current density of 10 mA cm À2 , which is comparable to iridium and cobalt oxide catalysts. The pyridinic and quaternary nitrogen species shorten the diffusion path of OER intermediates. The N-doped carbon hollow cubes (NCHCs) for the OER were synthesized from biomass lysine using a NaCl-template method by Yang and coworkers. 228 The NCHCs show comparable OER activity and high endurance compared with the RuO 2 /C catalyst in alkaline medium. Synergetic effects of N-doping and high specic surface area were found. Ozkan et al. demonstrated the OER activity of Ndoped carbon nanostructures (CN x ) in an acidic medium, which showed signicantly higher OER performance than the Ir/C catalyst. 255 XPS results suggested the presence of pyridinic N species for the enhanced OER activity of the CN x catalyst. Zhao's group demonstrated a novel strategy to fabricate a highly efficient OER catalyst by synthesizing an ultra-thin Ndoped holey carbon layer (HCL) on graphene through a twostep process as displayed in Fig. 17a. 231 Fig. 17b presents the HR-TEM image of Zn@G-hydro annealed at 900 C, which exhibits nanoholes with an approximate diameter of 2 nm, as a result of Zn evaporation (N-HC@G-900). The nano-holes are not observed when zinc gluconate is not used in the synthesis (N-C@G-900). N-HC@G-900 displayed enhanced OER activity, as observed in Fig. 17c, because of pyridinic N moieties at the edges of the HCL, and the sandwiched graphene sheet provided mechanical support.
The combination of N and transition metal doping into the carbon materials effectively improved the OER performance and stability by creating more active sites. 224,225,256 For instance, Mu et al. co-doped Co and N into the matrix of graphene-like porous carbon nanosheets (Co@N-PGCS), which possess a high specic surface area of 1716 cm 2 g À1 . 256 The Co@N-PGCS catalyst showed comparable activity over the state-of-the-art catalyst IrO 2 by a synergetic effect of Co and N co-doping, and the porous surfaces open more diffusion channels. In another report, Tong and coworkers designed an efficient self-standing OER catalyst by dual doping Co and N into a graphite foam matrix (Co-N/GF). 224 The Co-N/GF catalyst prepared at 700 C could deliver signicant OER activity, compared to the catalyst prepared at other temperatures in alkaline electrolyte. Initially, Co-N species were partly oxidized to CoOOH/Co-N. At higher anodic potentials, CoOOH was further oxidized to the most active CoO 2 , which considerably promoted the oxidation of OH À species. Wang's group, for the rst time, fabricated a Co/N codoped carbon composite for water oxidation from metal-polyphenol (Tanin) crystals, which are highly abundant in nature. 225 Co/N co-doping in a carbon composite was found to produce a synergistic effect with a high specic surface area (180 m 2 g À1 ), producing excellent OER activity in the alkaline electrolyte. Lee et al. demonstrated the synthesis of a tri-doped carbon catalyst for acidic OER. 257 Here, they doped N, O, and TMs, such as Fe, Co, and Ni, onto the carbon, and they noticed that N/Fe/O and N/Ni/O doped carbon catalysts showed high OER activity, and N/Co/O displayed better stability.

Summary and concluding remarks
Searching for eco-friendly renewable energy that can replace carbon-based fossil fuels for future fuel is a prime goal of the research community. Electrochemical oxygen evolution plays a crucial role in energy storage by converting renewable energies, such as solar, wind, and tidal forms, into storable chemical fuel of hydrogen. Hence, the fundamental priority is to fabricate efficient OER catalysts that can be applied in the real world. This review summarized the research on the tuning of OER catalyst properties by doping. Doping of foreign elements can signicantly alter the electrical conductivity and binding ability of OER intermediates by optimizing the host electronic structure. The considerable changes in the OER activity of metal oxides, non-metal oxides, and carbon-based catalysts via doping were reviewed. Experimental and theoretical studies were discussed to explain the OER mechanisms. Among the catalysts discussed above, doping in hydroxides, suldes, selenides, and phosphides has been shown to be relatively more effective compared to the others discussed in this paper. Overall, this review outlines the systematic understanding of experimental and theoretical effects of doping on OER catalysts, which is associated with several challenges involving energy processes such as carbon dioxide reduction, nitrogen reduction, and production of hydrogen peroxide.
Considering the studies reported in this paper, an efficient OER catalyst should have optimum adsorption ability for OER intermediates, neither too weak nor too strong, and possess a good electrical conductivity. The optimized e g lling value ($1.2) can be obtained by tuning the electronic structure of TM ions present in metal oxides, and optimizing the covalency of metal and oxygen orbitals, including other metal ions along with TM ions. An optimized surface area can be achieved by preparing nanoforms. Furthermore, the catalyst should possess excellent resistance under extreme OER conditions with high anodic potentials and a good redox center that can lose or gain electrons.
In future, the development of advanced characterization techniques, such as in situ XPS, Fourier-transform infrared spectroscopy (FT-IR), X-ray absorption spectroscopy, XRD, etc., will provide more insights into OER intermediates on the surface of a catalyst. To design a highly stable catalyst for industrial-scale applications, it is essential to have the metric of long-term stability without dropping its initial activity evidently. Several catalysts display high activity, but a few of them can reach long-term endurance. Besides, large surface area and porosity are the other vital parameters in producing catalysts for large-scale use, which could provide high activity with a small catalyst. Most of the catalysts described in this paper were synthesized on a lab scale, while mass production of the same type of catalyst will be challenging in terms of cost effectiveness to fulll the practical needs. Interdisciplinary collaborative research efforts are needed in science and engineering towards widespread applications of fuel cells.

Conflicts of interest
There are no conicts to declare.