Introduction

Many energy conversion techniques (such as metal–air batteries) have been developed with the energy consumption and climate warming [1,2,3,4]. The development of high-performance electrocatalysts is crucial to meet above techniques, focusing on facilitating kinetics and lowering the energy barrier during the electrocatalytic processes [5, 6]. Figure 1a is the mechanism diagram of metal air battery, from which we can see oxygen reduction reaction (ORR) is an important cathode reaction in the discharge processes of metal–air batteries [7]. Based on Fig. 1b, the ORR on N-doped carbon materials working principles is described for two possible pathways [8]. One is the four-electron mechanism taking place at a single site: The protons attach to the two oxygen atoms, leading to breakage of the O–OH bond and formation of OH species; the additional proton then reacts with the adsorbed OH to form H2O. The other is the 2 + 2-electron mechanism, which does not always take place at a single site; H2O2 is formed by reaction of the adsorbed OOH species with another proton, followed by reabsorption of H2O2 and its reduction by two protons to generate H2O. For metal–air batteries, ORR favors the four-electron process. Therefore, the electrocatalysts with four-electron mechanism have become the research hotpot.

Fig. 1
figure 1

Copyright 2021, Elsevier. b Schematic pathway for the ORR on N-doped carbon materials [8]. Copyright 2016, Science

a Schematic of a rechargeable zinc air flow battery [7].

Up to now, single-atom catalysts (SACs) have attracted more attention in the field of electrocatalysis. The catalytic activity and selectivity toward electrochemical reactions can be obviously increased through downsizing the metal nanoparticles into single atoms [9]. However, their high surface free energies make them easy aggregation; thus, it requires various substrates to form a stable configuration for the isolated atoms [10]. The dispersion of metal atoms in SACs not only favors the catalytic activity and selectivity during the electrocatalytic processes, but also reduces costs and increases atom utilization [11]. The catalytic performance of SACs can be enhanced via the tunable electronic structures formed by the strong interactions between the substrates and the individual atoms [12]. Unsaturated coordination environments of SACs play a critical role in the adsorption and activation of reactants, profiting to lower the energy barrier for electrochemical reactions [13].

The noble metals of platinum (Pt) have been used as ORR electrocatalysts. For example, Lai et al. [14] reported a quasi-Pt-allotrope SAC, increasing the efficiency of the ORR in virtue of the isolated Pt atoms. However, their high cost and unsatisfactory durability hinder their applications. Consequently, the development of transition metal-based catalytic systems for ORR is necessary [15]. For instance, Co single atoms dispersed on porous N-doped carbon nanotube (Co-NCNT) showed an ORR half-wave potential (E1/2) of 0.87 V, outperforming that of Pt/C (20 wt%) [16]. At the same time, the single-atom alloy SACs formed by alloying different single atoms also show good catalytic potential. Such as Xu et al. [17] synthesized FeCo alloy SAC, exhibiting an E1/2 of 0.89 V. Furthermore, the double-metal SACs (DMSACs) formed by introducing a second metal atom presented more catalytic sites, exhibiting better electrocatalytic performance [18, 19]. For instance, Han et al. [20] reported a DMSACs with Co/Ni in N-doped carbon nanotubes, delivering an E1/2 of 0.76 V and an onset potential (Eonset) of 0.88 V. Furthermore, Guo et al. [21] gave a review on the progress of approaches for the synthesis of SACs, outline the general principles and list the advantages and disadvantages of each synthesis approach. Gawande et al. [22] collected the latest advances concerning the applications of SACs in the ORR. However, there is still a lack of a comprehensive overview of the latest developments in ORR SACs. It is necessary to summarize systematically the classification and preparation methods of SACs to give the readers a clear picture and inspire more studies to exploit novel approaches.

In this review, we summarized the research progress of noble metal SACs including Pt–N–C, Pd–N–C, Pt–O–C, and Pt–C; transition metal SACs including Fe–N–C, Fe–O–C, Co–N–C, Ni–N–C, Cu–N–C, Zn–N–C, and Mn–N–C; single-atom alloy catalysts including Pt–M (Pd, Fe, Co) and Pd–M (Ag, Ir, Co, Ni); as well as double metal SACs including Pt/Ni–N–C, Fe/Ni–N–C, and Fe/Co–N–C (Fig. 2). Moreover, the synthesis approaches for SACs including wet chemistry, atomic layer deposition, and high-temperature pyrolysis were discussed. Moreover, the new research directions of SACs were proposed, such as improving the dispersion and loading of metal active centers, exploring suitable carriers and clarifying the synergistic mechanism between metal atoms in SACs.

Fig. 2
figure 2

Review scheme of SACs

Classification

Noble metal SACs

Due to the high activity and selectivity, the noble metal catalysts have been used for ORR [23, 24]. We discussed the progress of Pt–N–C and Pd–N–C in the research of ORR electrocatalysts (Tables 1 and 2).

Table 1 Comparative ORR performance of various SACs in alkali solution
Table 2 Comparative ORR performance of various SACs in acid solution

Pt–N–C

Pt–N–C has high ORR catalytic activity in alkaline medium [71]. However, the commercial value of Pt is significantly hindered by its low abundance, limited supplies. Accordingly, many researches focus on increasing Pt catalytic activity such as Eonset, E1/2, and limiting current density (JL) [72,73,74]. Recently, Pt supported on heteroatom-doped carbon, even the Pt–N–C, has been reported to improve ORR performance with long-term stability [75]. As a case, Zhao et al. [25] synthesized Pt–N–C supported on g-C3N4 nanosheets (Pt-CNHS) via the liquid-phase reaction of g-C3N4 and H2PtCl6 (Fig. 3a1-2), showing an Eonset of 3.7 V. Lai et al. [64] distributed Pt atoms on a heterogeneous substrate formed by Co particles and N-doped carbon frameworks (Pt@Co/NC; Fig. 3b1). The Pt@Co/NC delivered a mass activity is 4.2 mA μgPt−1 at 0.9 V (that of Pt/C is 0.15 mA μgPt−1) and a mass activity of 6.2 mA μgPt−1 at 0.85 V (that of Pt/C is 0.48 mA μgPt−1; Fig. 3b2). Xiang et al. [65] reported a carbon-supported Pt–N–C (Pt–N4–C) prepared using a small molecular Pt complex containing moieties as the Pt precursor. It showed high ORR performance with the activity of 0.108 mA µgPt−1 at 0.9 V (that of 20 wt% Pt/C is 0.035 mA µgPt−1). Moreover, adjusting N coordination is an effective method to improve the catalytic performance. Fan et al. [26] proposed a research model of Pt–Nx–C structure (1 ≤ x ≤ 4) fabricated by a zeolitic imidazolate framework (ZIF)-8 template (Fig. 3c1). The results showed the catalytic activity gradually increases with the decrease of x value, especially the positive Eonset (1.10 V; Fig. 3c2).

Fig. 3
figure 3

a1 Illustration of the CNHS and Pt-CNHS; a2 FESEM images of Pt-CNHS [25]. Copyright 2021, Elsevier. b1 The structure of Pt@Co/NC with and without Pt, b2 compared to the mass activities of Pt@Co/NC and Pt/C [64]. Copyright 2021, American Chemical Society. c1 Schematic of fabricating Pt–N–C; c2 linear sweep voltammetry (LSV) curves of Pt–NC–1.5% samples under different calcination temperatures [26]. Copyright 2021, Elsevier.

Pd–N–C

Pd–N–C has emerged as a potential alternative to Pt–N–C for ORR due to the high selectivity and lower costs compared with Pt [76]. Different supports can significantly alter the catalytic performance of Pd catalysts [77, 78]. In this aspect, Wang et al. [79] reported Pd–N–C SACs (Fig. 4a1) derived from Pd-doped ZIFs through one-step thermolysis. The catalyst exhibited about 95% selectivity toward H2O2 (Fig. 4a2) and an Eonset of 0.8 V. Xiang et al. [27] developed a strategy to deposit Pd atoms on MnO2 nanowires and carbon nanotubes (Pd/MnO2-CNT). The Pd–N–C had an Eonset of 0.982 V. Jin et al. [66] immobilized Pd atoms on thin layers of graphitic carbon nitride with carbon black (C@C3N4) to produce the Pd–N–C. It showed that the current density increased with the Pd content increased (Fig. 4b1). The Pd–N–C showed very high selectivity for H2O2 production up to 94% (Fig. 4b2). Furthermore, one or a few layers of atoms in Pd–N–C films could minimize the use of Pd atoms and enhance their performance [80].

Fig. 4
figure 4

a1 Pd–N–C catalytic site; a2 H2O2 selectivity and electron-transfer number [79]. Copyright 2022, American Chemical Society. ORR polarization curves: b1 current density of Pd catalysts, b2 H2O2 selectivity of Pd catalysts [66]. Copyright 2019, Wiley.

Others

Other noble metals such as Au–N–C, Ag–N–C, Ir–N–C, Pt–C, and Pt–O–C can also improve ORR performance to some extent. For instance, Fortunato et al. [81] reported an Au–N–C is constructed using Au–Pd molybdenum oxide supported on graphene nanoribbons. The catalysts exhibited high catalytic activity (Eonset = 0.72 V) and selectivity for H2O2 (75%). Cartagena et al. [82] prepared the Ag–N–C by using the rotating disk slurry electrode technique. The most ORR mass activities Ag–N–C (0.1 V, 4,765,000 mA µg−1) demonstrated three times higher activity than the least active one (0.4 V, 1,637,000 mA µg−1). Xiao et al. [83] developed an Ir–N–C, which mimics homogeneous iridium porphyrins for ORR catalysis. The Ir–N–C showed a mass activity of 12.2 mA µg−1 at 0.85 V. Xu et al. [84] synthetized Pt–C with an E1/2 of 0.85 V. Roh et al. [85] synthetized Pt–O–C by controlling the hydrophilicity of carbon supports in the electrodes by functionalizing the carbon with oxygen-containing groups, showing an ECSA of 88 m2 gPt−1.

Transition metal catalysts

Transition metal ORR catalysts represent promising alternatives to noble metal catalysts owing to their high activity and high thermal stability [86,87,88]. Herein, the recent research progress of SACs, including Fe–N–C, Fe–O–C, Co–N–C, Ni–N–C, Cu–N–C, Zn–N–C, and Mn–N–C, was summarized for ORR (Tables 1 and 2).

Fe–N–C and Fe–O–C

Due to excellent adsorption capacity and good electrical conductivity, Fe has become a hot spot in monatomic catalysts [89]. Fe has a large number of filled 3d orbitals and different coordination structures, which makes it more possible to form SACs [90]. At present, in a large number of studies, four active sites of Fe SACs, Fe–Nx [91], Fe–N2 [92], Fe–N4 [93] and Fe–N6 [94], have been mainly studied. However, the traditional synthesis methods of Fe catalysts expose few active sites, and the catalytic performance is not ideal [95]. Therefore, the simple and efficient synthesis of monatomic Fe catalysts with high active sites has become the main direction.

Many methods (gas-phase fluorination-assisted method, one-pot method) to improve its ORR performance have been reported. For an instance, Meng et al. [28] prepared a Fe–Nx on single-walled carbon nanotube (SWCNT) by gas-phase fluorination-assisted method (Fig. 5a1). The reduction potential reached 0.93 V at − 3 mA cm−2, which is better than that of Pt/C (0.86 V; Fig. 5a2-3). The confined nanopores can promote the activity and stability of ORR catalysts. Therefore, Cai et al. [29] designed a nanoporous Fe/N-doped carbon (Fe-NPC) SACs through a simplified one-pot method (Fig. 5b1), exhibiting an oxygen reduction peak potential of 0.775 V, an Eonset of 0.971 V, and a JL of 6.55 mA cm−2 (Fig. 5b2-3).

Fig. 5
figure 5

a1 Schematic of the SAFe-SWCNT; a2, a3 electrocatalytic performance of the SAFe-SWCNT and Pt/C-Ir/C [28]. Copyright 2021, Elsevier. b1 The diagram of Fe-NPC sample; b2 CV curves and b3 LSV curves of the samples [29]. Copyright 2022, Elsevier. c1 Illustration of Fe–N–C/N-OMC catalyst; c2 CV curves and ORR; c3 LSV of the samples in 0.1 M KOH electrolyte [30]. Copyright 2021, Elsevier. d1 Schematic of the Fe-ZIF-8NC and NaCl based on the salt-assisted emitting approach; d2 the corrected LSV at 1600 rpm; d3 E1/2 and current density at 0.85 V [31]. Copyright 2021, Elsevier. e1 The structure and ORR mechanism of Fe–O–C; e2 SEM: e3 LSV curves of Fe–O–C; e4 chronoamperometric test curves of Fe–O–C [34]. Copyright 2018, Elsevier.

Furthermore, the inaccessibility of Fe–N–C sites and the low Fe loading are important reasons, which limit the overall ORR activity. To solve the influence of this factor, Han et al. [30] synthesized Fe–N–C (Fig. 5c1) by using KIT-6, Fe(II)-Phen, and 2-methylimidazole, displaying an Eonset of 0.93 V and JL of 8.14 mA cm−2 higher than that of Pt/C (0.84 V, 6.57 mA cm−2; Fig. 5c2-3), approximately 93 mV. Cao et al. [31] developed a molten-salt-assisted thermal emitting approach, taking ZIF-8 template to transform bulk ferric(III) oxide powder into a Fe–N–C for ORR (Fig. 5d1). The Fe–N–C presented an E1/2 of 0.896 V (Fig. 5d2) and an ORR Tafel slope of 42 mV dec−1 (Fig. 5d3). Han et al. [32] synthesized a Fe–N–C by incorporating Fe-Phen complexes into the nanocages during the in situ growth of ZIF-8, followed by pyrolysis at 900 °C under inert atmosphere exhibiting an E1/2 of 0.91 V. Chen et al. [33] synthesized Fe–N–C via ultrasonic plasma engineering and direct carbonization in a precursor solution containing metal phthalocyanine and aniline. It exhibited the ORR Eonset and E1/2 of Fe–N–C equaled 0.92 V and 0.80 V. Moreover, Fe–O–C has attracted extensive attention because it can selectively achieve ORR of 4e by adjusting the metal central atom or non-metallic coordination environment [96]. Guo et al. [34] constructed a graphene-based Fe–O–C via a facile graphene-assisted wet chemical route (Fig. 5e1). Its morphology is shown in Fig. 5e2; the formation of Fe–O–C bond between Fe and graphene can boost 4e path, exhibiting an Eonset of 1.00 V (Fig. 5e3) and a good durability (Fig. 5e4).

Co–N–C

Due to the inevitable side reaction of Fe-based materials, namely the Fenton reaction, it leads to poor stability and rapid degradation [97]. Therefore, Fe-free catalysts have gradually attracted attention. Co–N–C has attracted extensive exploration due to their high ORR inherent activity and high stability [98,99,100].

To simplify synthesis steps and improve catalytic performance, Xiong et al. [35] prepared Co–N4–C derived from a Chlorella precursor via a straightforward pyrolysis strategy assisted by the high polarization force of a NaCl/KCl eutectic system (Fig. 6a1), exhibiting an Eonset of 0.81 V and an E1/2 up to 0.83 V (Fig. 6a2-3). Xu et al. [36] synthesized Co–N4–C atomically dispersed in the porous shell via integrating core–shell electrospun. The Co–N–C exhibited superior ORR activity with E1/2 of 0.84 V and JL of 5.22 mA cm−2. Xu et al. [37] synthesized a single Co atom embedded in the sandwich peak porous N-doped double carbon structure (NMCS-rGO-Co; Fig. 6b1-2). The Co–N–C exhibited an E1/2 of 0.72 V (Fig. 6b3). Yang et al. [38] proposed a catalyst consisting of high-purity pyrrole N anchored Co single-atom (CoN4). The Co–N–C exhibited an E1/2 of 0.84 V at 10 mA cm−2.

Fig. 6
figure 6

a1 The preparation scheme for Co–N/C-SAC; a2 CV and a3 LSV of the samples [35]. Copyright 2021, Elsevier. b1 Illustration of NMCS-rGO-Co; b2 TEM images for the NMCS-rGO-Co; b3 LSV curves [37]. Copyright 2022, Elsevier.

NiNC

Ni–N–C has been considered promising catalyst for various batteries because of good dynamics, high conductivity, and chemical stability. The local environment of single atom is sensitive to the supports, location, and surrounding heteroatoms. Such as a boosted ORR is achieved with the decrement in coordinated Ni atoms for Ni–N–C. Inspired by this, Luo et al. [39] regulated the coordination of Ni–N–C by a secondary single-atom modulation strategy (Fig. 7a1). The obtained Ni–N–C displayed an E1/2 of 0.854 V ((Fig. 7a2). Qiu et al. [40] developed Ni–N–C with a loading of ≈ 23 wt% by the chemical vapor deposition (CVD) process and chemical etching (Fig. 7b1). The Ni–N–C exhibited an Eonset of 0.984 V, an E1/2 of 0.859 V ( (Fig. 7b2). Zhao et al. [41] synthesized Ni–N–C supported on N-doped carbon nanotube (Ni–N-CNT) with a loading of 20.3 wt% via a one-pot pyrolysis method by employing Ni acetylacetonate and dicyandiamide precursors (Fig. 7c1). The Ni–N–C exhibited an Eonset of 0.88 V and an E1/2 of 0.75 V (Fig. 7c1).

Fig. 7
figure 7

a1 Scheme of the all-solid-sate ZABs; a2 LSV curves of the samples [39]. Copyright 2021, Elsevier. b1 SEM images of N-doped graphene/porous Ni; b2 ORR LSV of different samples [40]. Copyright 2019, Wiley. c1 Diagram of the forming procedures for Ni SAC; c2 the ORR curves on Ni–N-CNT SAC electrocatalysts [41]. Copyright 2018, American Chemical Society.

Cu–N–C

Cu–N–C is commonly used SACs with catalytic properties due to hydrogenation, dehydrogenation, and oxidation [101]. Significantly, the ORR inherent of Cu–N–C is compared to Pt/C and the method of synthesizing catalyst is facile. Thus, this novel catalyst is a promising alternative ORR replacing Co/C catalyst.

For improving ORR of Cu–N–C catalysts, Ji et al. [42] fabricated Cu–N–C by using 2,6-diaminopyridine C/N source and silica template (Fig. 8a1). The Cu–N–C delivered high ORR activity with an E1/2 of 0.83 V and a JL of 5.25 mA cm−2 (Fig. 8a2). Combination with Cu and polymer is also a method to improve the ORR performance. Ma et al. [43] fabricated Cu–N–C with the designed coordination structure by an ionic exchange strategy. The obtained Cu–N–C-900 exhibited an E1/2 of 0.87 V and Tafel slope of 79 mV dec−1 (Fig. 8b1–2). Han et al. [44] reported the single-atom Cu supported on graphene with a loading of 5.4 wt% by a confined self-initiated dispersing protocol (Cu/G; Fig. 8c1). The E1/2 and JL of the optimized Cu–N–C were 0.85 V and 3.787 mA cm−2, respectively (Fig. 8a2).

Fig. 8
figure 8

a1 Illustration of the Cu–N–C; a2 ORR LSV of Cu–N–C samples [42]. Copyright 2021, American Chemical Society. b1 E1/2 and current density of Cu–N–C and Pt/C; b2 Tafel plots of the samples [43]. Copyright 2020, Wiley. c1 Schematic conventional and confined self-initiated dispersing protocols; c2 ORR LSV of the samples [44]. Copyright 2019, Elsevier.

Zn–N–C

Due to the high volatility of metal Zn, Zn-based materials have less progress for the transition metal catalysts [102, 103]. But, the Zn-based precursors are applied in producing porosity via pyrolysis for the Zn-free SACs [104,105,106,107]. Moreover, Zn–N–C catalysts were active for various electrochemical reactions.

Jiang et al. [45] reported the catalyst with Zn sites supported on N and S co-doped biomass graphene (A-Zn@NSG; Fig. 9a1). The Zn–N–C exhibited an E1/2 of 0.805 V in 0.1 M HClO4 and 0.905 V in 0.1 M KOH (Fig. 9a2-3). Wang et al. [46] synthesized single Zn atoms dispersed on N-doped hierarchically porous carbon (Zn–N–C) by employing apples, egg whites, and ZnCl2 raw materials (Fig. 9b1–2). The catalyst exhibited an Eonset = 0.85 V, E1/2 = 0.78 V, and JL = 4.71 mA cm−2 in 0.1 M KOH electrolyte solution (Fig. 9b3-4).

Fig. 9
figure 9

a1 Preparation diagram of A-Zn@NSG; a2 LSV curves for A-Zn@NSG, A-Zn@NG, NSG, and Pt/C (20%); a3 LSV curves of A-Zn@NSG and Pt/C (20%) [45]. Copyright 2021, American Chemical Society. b1 HAADF-STEM image of Zn–N–C; b2 HRTEM image of Zn–N–C; ORR LSV of Zn–N–C–X and Fe–N–C–X catalysts recorded in O2-saturated 0.1 m KOH b3 and 0.1 m HClO4 solution b4 [46]. Copyright 2019, Wiley. c1 Diagram for the Zn–B/N–C synthesis; c2 LSV curves of different samples; c3 LSV curves at different rotation rates [48]. Copyright 2020, Wiley.

Owing to the high volatility of the metal Zn at high temperature, the Zn loading in Zn–N–C catalysts still remains a formidable challenge [108, 109]. Li et al. [47] prepared a Zn–N–C catalyst with a Zn loading of 9.33 wt% by adopting a low annealing rate of 1 °C min−1. The Zn–N–C catalyst largely maintained its Zn–Nx active sites, and the E1/2 decayed by only 19.88 mV after 1000 CV cycles. Wang et al. [48] synthesized Zn–N–C catalyst via a facial one-step strategy, giving a Zn loading of 2.4 wt% (Fig. 9c1). The Zn–N–C showed an E1/2 of 0.886 V and an average number of 4.08 (Fig. 9c2-3).

Mn–N–C

Mn–N–C is used SACs with high potential due to cheap, non-toxic, and widely sourced [110, 111]. Mn considered as an active metal for ORR electrocatalysis compared to zirconium [112, 113].

Mn–N–C is generally low active for the ORR under alkaline conditions. To solve this problem, Lin et al. [49] synthesized the Mn–N–C supported on the carbon frameworks (Fig. 10a1), displaying an E1/2 of 0.87 V (Fig. 10a2). Inspired by the catalytic activity and selectivity of enzymes, Yan et al. [50] designed ultrathin carbon nanosheet-supported Mn–N4–C (PT-MnN4) by using a bio-mimicking strategy (Fig. 10b1). The Mn–N–C showed an Eonset of 0.95 V, an E1/2 of 0.88 V, and a JL of 5.80 mA cm−2 (Fig. 10b2-3).

Fig. 10
figure 10

a1 The synthesis route of the Mn–N–C; a2 LSV curves of the Mn–N–C SAC, N/C, and 20 wt% Pt/C [49]. Copyright 2020, American Chemical Society. b1 Illustration of pyrrole-type (PT)-MnN4; b2 LSV curves of PT-MnN4, NC, and Pt/C; b3 E1/2 and kinetic current density of PT-MnN4, NC, and Pt/C [50]. Copyright 2021, Wiley.

Others

Other transition metals can also have ORR performance. As a case, Liu et al. [114] reported Sn–N–C with an E1/2 of 0.87 V and excellent durability. Luo et al. [115] used a metal–organic framework as an anchoring matrix obtained pyrolyzed Cr–N–C. It exhibited an E1/2 of 0.773 V and showed superb stability. Xue et al. [116] reported a Mo–N–C exhibiting excellent activity and selectivity toward electrochemical reduction reaction with favorable limiting potential of 0.40 V.

Single-atom alloy SACs

Single-atom alloys with the characteristics of unique electronic structures and well-defined active sites demonstrated the promising potential to replace traditional metal catalysts [117]. The advantages of decoupling the dissociation and reaction sites in electrocatalysis are expected to tailor the adsorption properties of reaction species, thus promoting electrocatalytic behaviors [118]. We discussed the progress of Pt–M (Pd, Co, Ni) and Pd–M (Ag, Ir, Co, Ni) in this section (Tables 1 and 2).

Pt–M (Pd, Co, Ni)

To diminish the Pt loading, manufacturing Pt-based alloy catalysts (Pt–M) has made great progress, wherein M can tailor the electronic and geometric structures to improve the ORR activity [119]. Zhang et al. [120] synthesized Pt–Pd SACs on nitrogen-doped carbon nanotubes by ALD technology (Fig. 11a1). The catalyst showed a mass activity of 0.91 mA gPt−1 and good durability (Fig. 11a2). Cheng et al. [67] reported a Pt–Co SAC by pyrolysis method (Fig. 11b1). The catalysts exhibited an E1/2 of 0.85 V (Fig. 11b2) and Tafel slope of 82 mV dec−1 in 0.1 M HClO4 solution. Lee et al. [68] presented a strategy of active Pt–Ni SACs via fluidized bed reactor ALD (FBR-ALD; Fig. 11c1). The catalyst exhibited a mass activity of 0.57 mA μgPt−1 (Fig. 11c2).

Fig. 11
figure 11

a1 The schematic for the synthetic of Pd–Pd SACs; a2 LSV curves of the octahedral Pt/Pd SACs and durability test [120]. Copyright 2019, American Chemical Society. b1 The schematic for the synthetic of Pt1Con/N-GCNT; b2 LSV curves [67]. Copyright 2022, Elsevier. c1 Schematic of Pd–Ni SACs; c2 mass activity of alloy catalysts [68]. Copyright 2022, Elsevier.

Pd–M (Ag, Ir, Co, Ni)

Pd-based alloy catalysts have been regarded as potential substitutes for Pt catalysts. Both Pd and Pt are platinum group metals adjacent in the periodic table and share similar electronic structures and properties. However, Pd is more cost-effective than Pt [121]. For a case, Betancourt et al. [52] synthesized Pd–Ag alloy catalyst by using under potentially deposited (UPD; Fig. 12a1). With Pd modifying Ag/C, the ORR E1/2 was shifted to more positive values, and the overpotential is significantly reduced along with an overall activity increment observed as an increase in Pd (Fig. 12a2). Kim et al. [53] prepared a Pd–Ir alloy catalyst via the polyol process. The E1/2 of catalysts can reach 0.626 V (Fig. 12b1), and Tafel slop can reach 176 V dec−1 (Fig. 12b2). Wei et al. [122] reported that the metal only grows and spreads out in two dimensions, the freestanding and self-stabilized Pd–Co SAL generated in the angstrom-sized interlayer space of layered minerals (Fig. 12c1). Pd–Co SAL presented a good ORR performance (Fig. 12c2) and a mass activity of 0.257 mA µgPd−1 (Fig. 12c3). Sahoo et al. [51] reported an ultrathin twisty Pd–Ni alloy catalyst (Fig. 12d1). The catalysts exhibited a low reaction overpotential with an E1/2 = 0.95 V (Fig. 12d1).

Fig. 12
figure 12

a1 The schematic for Pd–Ag alloy catalysts; a2 LSV curves [52]. Copyright 2020, American Chemical Society. b1 I–V curves for the ORR activity and b2 Tafel slope of Pd–Ir-based ternary alloy catalysts [53]. Copyright 2022, Elsevier. c1 Scheme for Pd–Co alloy; c2 LSV and c3 specific mass activities [122]. Copyright 2022, Elsevier. d1 Schematic of Pd–Ni alloy; d2 LSV curves of Pd–Ni alloy [51]. Copyright 2022, American Chemical Society.

Others

Noble metal-free alloy catalysts have always been a research hotpot due to the rich resources. However, it is still a challenging to develop noble metal-free alloy catalysts. Nam et al. [123] synthesized a Cu–Fe alloy SACs, exhibiting a good ORR activity compared to that of Pt–C electrocatalysts. Niu et al. [124] reported Co–Ni alloy catalyst synthesized from metal precursors and dicyandiamide nanofibers formed by freeze-drying pretreatment. The catalyst exhibited an Eonset of 0.9 V, an E1/2 of 0.84 V, and a JL of 6.4 mA cm−2.

Double metal SACs

For solving the defects of SACs, multiple atomic active sites and high metal loading have been gradually designed and constructed in recent years [125,126,127]. Double-metal SACs (DMSACs) were born. Recent research proved that the SAC activity can be enhanced by adding a second metal atom, pointing out the research direction of DMSACs [128]. DMSACs composed of bonded two metal atoms and coordination of metal pairs with N atoms in carbon framework, opposite to M–N–C SACs. Such configuration could improve the electrocatalytic activities of metal centers (Tables 1 and 2) [129, 130].

Pt/Ni–N–C

Although Pt-SAC has some advantages, the shortage of resources reduces the commercial value of Pt [131]. Therefore, researchers began to adopt a diatomic approach, aiming to reduce the content of Pt while maintaining or increase its ORR performance. Pt/Ni–N–C is a promising ORR catalyst due to that the alloying Pt with Ni can reduce the Pt content and increase its intrinsic activity [132]. Liu et al. [69] designed Pt/Ni–N–C by using a low-temperature impregnation–reduction method (Fig. 13a1), exhibiting more satisfactory ORR performance, a remarkable cathodic peak (Fig. 13a2), more positive E1/2, and incremental JL (Fig. 13a3). Wan et al. [133] prepared a Pt/Ni–N–C showing a high mass activity of 3.25 ± 0.14 mA µgPt−1 at 0.9 V. Li et al. [70] synthesized a Pt/Ni–N–C with a nanoflower structure by controlling the surface-active agents (Fig. 13b1), exhibiting the E1/2 of 0.930 V (Fig. 13b2-3).

Fig. 13
figure 13

a1 Illustration of PtM NCs (M = Fe, Co, Ni); a2 CV curves and a3 ORR LSV of PtM/Cs [69]. Copyright 2019, American Chemical Society. b1 Diagram of the Pt–Ni PND; b2 CV curves and b3 ORR LSV of Pt–Co GND, Pt–Ni PND, and Pt/C [70]. Copyright 2021, Elsevier.

Fe/Ni–N–C

Both Fe and Ni elements are ideal SACs, which have excellent electrocatalytic performance [134]. With the rise of DMSACs, the preparation of DMSACs with high ORR performance by combining Fe and Ni and anchoring on the same matrix has attracted the interest of researchers [135].

Bai et al. [54] reported a host–guest method to load the Fe–Ni-mIm (guest) on the surface of ZIF-8 (host), and the resulting ZIF-8@Fe–Ni(mIm)X precursors can be converted to Fe/Ni–N–C after carbonization (Fig. 14a1). The Fe/Ni–N–C exhibited an E1/2 = 0.91 V and a loss of 6 mV after 5000 cycles (Fig. 14a2-3). Liu et al. [55] synthesized a dual metal (Fe, Ni)-N-doped carbon (Fe/Ni–N–C) by the pyrolysis of Fe, Ni co-doped ZIF-8 exhibiting the E1/2 of 0.79 V in HClO4 and 0.86 V in NaOH electrolyte. Wang et al. [56] presented an elaborate method for preparing Fe/Ni–N–C in N-doped carbon via introducing Fe/Ni Hofmann-type MOFs precursors. As-obtained Fe/Ni–N–C showed an ORR potential gap of 0.76 V.

Fig. 14
figure 14

a1 Illustration of FeNi SAs/N–C; a2 E1/2 and Eonset of the samples a3 LSV curves of FeNi SAs/N–C [54]. Copyright 2022, Elsevier. b1 Schematic of Fe/Ni(1:3)-NG; b2 LSV curves of the different samples; b3 LSV curves of Fe/Ni(1:3)-NG [57]. Copyright 2021, Elsevier.

In addition, graphene is also a popular material for electrocatalysis [136]. To anchor metal ions by the trapping ability of graphene, Ma et al. [57] synthetized the Fe/Ni–N–C supported on the N-doped graphene (Fe/Ni(1:3)-NG; Fig. 14b1), exhibiting good ORR performance and excellent stability (E1/2 = 0.842 V; Fig. 14b2-3). Sirirak et al. [137] synthesized Fe/Ni–N–C (Fe/Ni-NG) supported on graphite oxide via thermal annealing of GO-mixed melamine under N2 atmosphere. The Eonset for Fe/Ni-NG is 0.61 mV positive shift compared with NG alone (0.48 V). Its peak potential is 0.12–0.34 V. Zhang et al. [58] reported a novel oxygen electrode catalyst based on bimetal Fe–Ni atoms anchored on N-doped graphene-like carbon (Fe/Ni–N–C). The ORR Eonset and E1/2 of Fe/Ni–N–C were about 0.90 V and 0.81 V.

Fe/Co–N–C

It has been found that Fe/Co–N–C showed tremendous promise for ORR to replace noble metal catalysts due to stability and catalytic activity [138].

Wu et al. [59] synthesized a Fe/Co–N–C in that Fe and Co atoms are stabilized separately on 2D carbon nanosheets via coordination with N and S heteroatoms to form a Fe/Co–N–C. The Fe/Co–N–C exhibited outstanding electrocatalytic activities of ORR with an E1/2 of 0.86 V, outperforming the Fe and Co SACs. Liu et al. [60] synthesized Fe/Co–N–C via the micro-mesopore confinement synthetic strategy. The catalyst had superhigh specific surface area and abundant defective structure and displayed excellent ORR performance in 0.1 M KOH (Eonset = 1.05 V, E1/2 = 0.845 V). Kumar et al. [61] prepared Fe/Co–N–C on phthalocyanine-modified multiwalled carbon nanotubes (MWCNTs). The Fe/Co–N–C exhibited an Eonset of 0.93 V and an E1/2 of 0.86 V (Fig. 15a1-2). Yu et al. [62] via high-temperature calcination pyrolysis of Zn and Co dual-metal sites of ZIF-coated electrospun polyacrylonitrile fibers synthesized the Fe, Co, and N co-doped catalyst (Fe/Co–N–C), exhibiting an Eonset of 0.99 V and an E1/2 of 0.83 V in 0.1 M KOH solution. Zhou et al. [63] synthesized an efficient Fe/Co–N–C based on N-doped carbon by pyrolysis of Fe/Co and 10-phenanthroline complexes (Fe/Co-phen) supported on ZIF-8 (Fig. 15b1). The Fe/Co–N–C exhibited an Eonset of 0.97 V, an E1/2 of 0.86 V, and a Tafel slope of 74 mV dec−1 (Fig. 15b2-3).

Fig. 15
figure 15

a1 Illustration of FeCo-N-C; a2 ORR curves recorded in O2-saturated 0.1 M KOH [61]. Copyright 2021, American Chemical Society. b1 Illustration of FeCo-NC; b2 The ORR polarization profiles of the samples; b3 Tafel curves of FeCo-NC and Pt/C [63]. Copyright 2021, Wiley.

Others

With the deepening of the research on DMSACs, more and more elements are introduced [139]. As a case, Parkash et al. [140] used CTAB reduction method to prepare carrier-free, ultralow Pt content nanostructured Pt/Cu atoms (NPS) with different Pt contents. The Pt/Cu–N–C exhibited an Eonset = 0.98 V. Xu et al. [141] reported Fe/Zn–N–C supported on porous N-doped carbon frameworks. It achieved an E1/2 of 0.78 V, 0.85 V, and 0.72 V in 0.1 M HClO4, 0.1 M KOH, and 0.1 M phosphate buffer saline solutions, respectively. Gharibiab et al. [142] synthesized Cu/Fe–N–C with an Eonset of 0.86 V and a JL of 6.34 mA cm−2 in 0.1 M HClO4.

Synthetic strategies

The method of accurately preparing SACs is expanding [143, 144]. But when the metal particles are reduced to the single-atom level, the specific surface area and surface free energy will increase sharply, making it easy for single atoms to form large clusters through coupling in the process of preparation and reaction, resulting in reduced catalytic activity [145, 146]. Therefore, the synthesis of SACs remains a huge challenge. In this section, we summarized several synthesis methods and discussed their advantages and disadvantages [147].

Wet chemistry

As a kind of common methods for catalyst preparations, wet chemistry includes impregnation, co-precipitation, and solvothermal method [148]. The wet chemistry methods were used to prepared SACs by adsorbing metallic ionics in solution and then attaching atoms on the supports. Such methods are employed to synthesize SACs via adjusting the metal selecting and loading [149]. It aims to stabilize the atom species on a suitable supporter via chemical coordination, inhibiting aggregation during the preparation process [150].

Zhang et al. [151] employed an impregnation method to prepare the Pd SACs. Because the interaction between defect sites and functional groups, the Pd2+ ions accumulated around the defect sites on the carbon surface of the Pd precursor solutions (Fig. 16a). It improved the in-plane and inter-lamellar separation/transfer of the carrier, improving the ORR efficiency. Zou et al. [152] synthesized ethylenediaminetetraacetic acid-M/CeO2 catalysts (EDTA, M = Ni or Co, CCI; Fig. 16b1) via a chelation coupled wetness impregnation. The catalysts possess rich oxygen vacancies and interfacial metal active sites (Fig. 16b3). Yi et al. [153] employed the co-precipitation method with Pd2+ ions and an ionic liquid of ReO4 salt to load single-atom Pt on NBF-ReS2, realizing high Pd atom utilization efficiency. Qin et al. [154] synthesized Aux–NiMn2O4 by a modified solvothermal method (Fig. 16c1). The result showed that the Au atoms dispersed on NiMn2O4 lattice and occupied the position of Mn, Ni, or O (Fig. 16c2-3), which improved the electrocatalytic performance of ORR. Lu et al. [155] fabricated Co-MOF SACs based on Co2+ source and trimeric acid (H3BTC) through a solvothermal method (Fig. 16d). More catalytic active sites were provided for high ORR performance.

Fig. 16
figure 16

a Schematic of the Pd/carbon cathode architecture [151]. Copyright 2019, American Chemical Society. b1 Ni/CeO2(CCI) Concept Schematic; b2 Ni 2p spectra of Ni/CeO2(IWI) and Ni/CeO2(CCI) [152]. Copyright 2022, Elsevier. c1 The synthesis procedure of Aux-NiMn2O4; c2 HAADF-STEM image and c3 EDS mapping of Au0.12-NiMn2O4 [154]. Copyright 2022, Elsevier. d Schematic diagram of Co-SAC synthesis [155]. Copyright 2018, Elsevier.

Wet chemistry has the following advantages. (1) The loading components are only loaded on the surface of the carrier in most cases, which can effectively improve the utilization rate of atoms and reduce the dosage [156]. Therefore, this method is of great significance for Pt-based and other noble metal catalysts and can effectively reduce the amount of noble metals. (2) Different supports can be selected to provide the required physical properties for the catalyst, such as specific surface and pore radius [157]. Therefore, wet impregnation method is a simple and convenient method. Although wet impregnation has the above advantages, there are still some disadvantages that are difficult to avoid [158]. It is easy to cause the migration of active components during drying [159].

Atomic layer deposition

Researchers found that atomic deposition layer (ALD) can effectively avoid migration of active sites via depositing atoms on the carrier in the atmosphere of vapor phase. As a kind of gas-phase deposition strategy, ALD can deposit single atoms layer by layer on the surface of supports [160, 161]. This is an effective strategy to control SACs based on atomic scaling.

As a case, Kim et al. [162] prepared Pt3Ti SACs via the coupled of ALD with thermal reductive annealing. This affinity ensured that Pt and TiO2 were in close contact due to the strong metal support interaction between Pt and TiO2. As-synthesized Pt3Ti SACs showed excellent specific activity 5.3-fold higher than Pt/C. Jiao et al. [163] fabricated MoS2-NTA with different nanotube diameters and wall thicknesses by a sacrificial strategy of anodic aluminum oxide template via ALD. Then, Pt atoms were fixed on the wall of Ti3C2-supported MoS2-NTA (Fig. 17a1). The abundant defects on Pt/MoS2-NTA/Ti3C2 (Fig. 17a2) and individual Pt atoms on the MoS2-NTA surface (Fig. 17a3) are beneficial for enhancing the catalytic efficiency. Wang et al. [164] reported Fe SACs supported on the substrates of multiwalled carbon nanotubes, SiO2, and TiO2 with the Fe loading of > 1.5 wt% via ALD (Fig. 17b1), as shown in Fig. 17b2-3.

Fig. 17
figure 17

a1 The synthesis processes of Pt/MoS2-NTA/Ti3C2; a2 HRTEM image of the single MoS2 nanotube in Pt/MoS2-NTA/Ti3C2; the inset: partial enlargement displaying the MoS2 interlayer spacing; a3 the HAADF-STEM images of Pt/MoS2-NTA [163]. Copyright 2022, Elsevier. b1 Scheme for Fe ALD half-reactions; HAADF-STEM images of b2 10c-Fe/MWCNTs and b3 15c-Fe/TiO2 samples [164]. Copyright 2020, American Chemical Society.

ALD can precisely control the growth of atoms in a single metal on the carrier, so such technology has been used to study the growth mechanism and the influence during the formation processes of isolated metal atoms. However, although ALD method has been used, the shortcomings of time-consuming and high requirements for material properties still exist [165].

High-temperature pyrolysis

For reducing the time, high-temperature pyrolysis method has attracted the attention of researchers. It is a method to generate target SACs by heating functional organic structure with suitable metal elements at high temperature, displaying the advantages of simple operation, rapid response, and easy adjustment [166]. At present, high-temperature pyrolysis has been widely used [167, 168].

For instance, Chen et al. [169] synthesized Fe/Co–N–C by a water-regulated and bioinspired one-step pyrolysis method at 800 °C (Fig. 18a1-2). Through pyrolysis, single atoms were coated on carbon nanotubes, so that the large specific surface area and porous structure of the catalyst template can be fully utilized. It made the catalyst have higher activity and stable. Ding et al. [170] reported the preparation of thermally stable metal oxide–supported Pt SAC by flame spray pyrolysis (Fig. 18b1). The results revealed that flame spray pyrolysis favored the formation of tetragonal-monoclinic phase of ZrO2 with improved redox property, thus leading to enhanced catalytic activity in high-temperature applications (Fig. 18b2).

Fig. 18
figure 18

a1 Schematic of FeCo-CNTs/NHC; a2 SEM of FeCo-CNTs/NHC-800 [169]. Copyright 2022, Elsevier. b1 Schematic of Pt SAC; b2 XRD pattern of Pt SAC [170]. Copyright 2021, Elsevier. c1 Illustration for the preparation of Fe SAC-MIL101-T; c2 LSV curves; c3 normalized I–T curves of Fe SAC-MIL101-T [171]. Copyright 2021, Wiley. d1 The synthesis strategy of N-rich Fe-NC; d2 LSV curves; d3 Tafel plots of N-rich Fe-NC [172]. Copyright 2021, Elsevier.

In addition, taking MOF as a precursor and then pyrolysis is also a common synthetic method. Xie et al. [171] synthesized Fe SAC via pyrolysis a mesoporous cage architecture (Fig. 18c1). The catalysts exhibited an E1/2 of 0.94 V (Fig. 18c2) and good durability (Fig. 18c3). Yaengthip et al. [173] prepared N-doped reduced graphene oxide by pyrolysis with urea as N2 source. Then, by controlling the temperature, different N contents of SACs are achieved, which showed higher ORR activity and they had outstanding durability and stability after running within saturated oxygen for over 2000 cycles. Xu et al. [174] developed a rapid and solvent-free method to produce Fe-doped ZIF-8 under microwave irradiation. After rational pyrolysis, Fe–N–C was obtained. The catalysts exhibited an Eonset of 0.884 V and an E1/2 of 0.782 V. Chen et al. [172] employed urea as a dispersant and as an additional nitrogen source to synthesize ZIF-8-derived Fe–N–C (Fig. 18d1). The catalyst showed an Eonset of 0.91 V (Fig. 18d2) and JL of 5.38 mA cm−2 (Fig. 18d3).

A series of SACs prepared by the high-temperature pyrolysis has been reported. However, it is limited by the choice of the metal precursor ligands and carrier, being accessible to residual during the process of pyrolysis. Moreover, isolated atoms with high surface free energy are easy to agglomeration at high temperature, reducing the catalytic activity.

Others

Furthermore, there are many other ways to synthesize SACs such as chemical vapor deposition (CVD), Lewis acid, and ball milling. For example, Liu et al. [175] synthesized Fe–N–C with FeN4 sites using CVD technology. Zhong et al. [176] prepared the single-atom Fe on wood-based porous carbon via Lewis acid pretreatment and carbonization. Lewis acid FeCl3 pretreatment not only produced abundant microchannels but also introduced Fe–N species, enhancing the ORR performance and durability. Jin et al. [177] reported a rapid and straightforward method to construct K2PtCl4@NC-M by the spherical grinding of Pt precursors and N-doped carbon carriers. Many Pt atoms are dispersed on NC through ball milling conducive to improving ORR performance. However, most of these methods are still in the initial experimental stage, and further research is still necessary.

Conclusion and outlook

In summary, we presented a comprehensive study on the progress of SACs for ORR, including noble metal SACs, transition metal SACs, single-atom alloy SACs, and double metal SACs, as well as discussed several common synthesis methods. For noble metal SACs, their electrocatalytic activity is higher than that of transition metal SACs, but the limited resources and poor durability lead to difficult application widely. About transition metal SACs, they have sufficient resources and excellent stability, but the issues of low catalyst performance and low metal loading still need to be solved. The single-atom alloy SACs have unique electronic structures and well-defined active sites, but complex synthesis methods and mechanisms still need to be improved and explored. About double metal SACs, they possess higher metal loading and better performance compared to transition metal SACs. But the metal atom aggregation is still a big challenge. In order to accurately preparing SACs, some synthesis methods are often used. In terms of wet chemistry, it can inhibit metal aggregation during the preparation process and effectively improve the utilization rate of atoms. But it is easy to cause the migration of active site during drying. ALD can effectively avoid metal atom aggregation and control the metal atoms on the carrier. However, it also has the shortcoming of time-consuming and high requirements for material properties. In addition, high-temperature pyrolysis is a method with the advantages of simple operation, rapid response, and easy adjustment. But its high surface free energy of the isolated atoms is easy to agglomerate at high temperature.

To obtain the high-performance SACs based on the recent progresses, the following research directions should be considered: (1) improving the dispersion and loading of metal active centers by regulating the coordination structure of metal atoms, such as S and P participating in metal atom coordination; (2) exploring suitable carriers involving coordination controllability of metal active centers, abundance of raw materials, and structural stability; (3) combining the advantages of different synthesis methods while avoiding their disadvantages to achieve the efficient synthesis of SACs; (4) further clarifying the synergistic mechanism between metal atoms in SACs containing two and/or more metal active centers to guide the active center regulation.