Understanding the active sites and associated reaction pathways of metal-free carbocatalysts in persulfate activation and pollutant degradation

Abstract

Environmental remediation based on persulfate (i.e., peroxomonosulfate (PMS) and peroxydisulfate (PDS))-activation-enabled radical and non-radical oxidation has captured growing interest due to the strong oxidation power, long life of reactive oxygen species, and wide pH adaptability. While metal-based catalysts suffer from environmental problems (e.g., metal ion leaching), metal-free carbocatalysts become the focus of current activities for establishing environmentally benign and efficient carbocatalyst-driven PMS/PDS-based advanced oxidation processes (AOPs). However, the complex structure and nonstoichiometry of carbocatalysts cause the decoding of their active sites and associated pathways in activating persulfates and degrading pollutants to be challenging. Moreover, the ambiguous active sites and structure–activity–performance relationships, as well as the inability to differentiate the oxidation capabilities based on different reaction pathways, present an enormous obstacle to designing and fabricating highly active and durable carbocatalysts for PMS/PDS-based AOPs. Hereby, this paper makes an effort to unravel the catalytic sites of carbocatalysts and their associated pathways in PMS/PDS-based AOPs. Apart from various non-doped active sites, we summarize the diverse active sites induced by single-doping, dual-doping, and tri-doping carbon substrates with different kinds of heteroatoms. Meanwhile, the relationships between these active sites and associated pathways in PMS/PDS-based AOPs are simultaneously analyzed before presenting the challenges and future perspectives.

---

Huawen Hu

Huawen Hu received his Ph.D. degree in philosophy from the Hong Kong Polytechnic University (PolyU) in 2015. After continuing one year of postdoctoral research at PolyU, he joined Foshan University in 2016. Prof. Hu was elected to the Guangdong Top-notch Young Talents Program of Pearl River Recruitment Program of Talents and rewarded with the Foshan Science and Technology Pioneer prize. He has over 90 peer-reviewed publications in top journals, with an h-index of 36. His research interests mainly include carbon-based functional materials for energy, catalysis, sensing, and environmental remediation applications.

Jian Zhen Ou

Jian Zhen Ou received his Ph.D. degree in Electrical and Electronic Engineering from RMIT University in 2012. Prof. Ou currently leads a research team at RMIT University for developing nanomaterial-enabled sensors. His research interests include two-dimensional materials, chemical and biological sensing, nanoscale electronics and photonics, and artificial intelligence-driven sensors. He has over 190 peer-reviewed publications in top journals including Science, Nature Materials, Nature Electronics, Nature Communications, Advanced Materials, ACS Nano, Nano Letters, and Advanced Functional Materials, with more than 17 000 citations and an h-index of 69.

---

Environmental significance

The focus of current research on the carbocatalysts constructed with a variety of carbon substrates for persulfate (i.e., peroxomonosulfate (PMS) and peroxydisulfate (PDS)) activation and environmental remediation has contributed to an explosive growth in the number of related publications over recent years. Nevertheless, the complex structure and non-stoichiometric nature of metal-free carbon materials cause the clarification of the catalytic sites and associated reaction pathways to be highly challenging. The ambiguous active sites and structure–activity–performance relationships, as well as the frustrating difficulties in discerning the oxidation abilities exhibited by different reaction pathways, call for a timely overview to elucidate the complexities and present rational research and development strategies in deeply engineering the internal structure and surface properties of metal-free carbon substrates to produce highly active and durable carbocatalysts. Herein, a comprehensive critical overview and perspectives are presented in this work, which hopefully provides the keys to the critical problems above and brings metal-free carbon materials closer to real-world environmental remediation by efficient PMS/PDS-based advanced oxidation processes.

Introduction

The increasing environmental pollution with the fast development of various industries calls for efficient and low-cost methods to remove the toxic and refractory pollutants from water radically.^1–4 The current methods involve adsorption,^5–7 Fenton oxidation,^8,9 Fenton-like oxidation,^10,11 electro-Fenton,^12,13 photo-Fenton,^14,15 electrochemical oxidation,^16,17 photocatalysis^18–21 and others.^22–26 Among these methods, Fenton-like oxidation based on peroxomonosulfate (PMS) and peroxydisulfate (PDS) has attracted significant attention due to its long life spans, strong oxidation ability, ease of storing and transporting PMS/PDS relative to liquid H₂O₂, applicability to a diversity of refractory organic pollutants, and capability to withstand a wide pH range, which is superior to the Fenton process requiring H₂O₂ as the reactant, thus causing the storage and delivery to be difficult.²⁷ Moreover, the Fenton process is limited to a strict pH range of 2.5–3.5 and involves Fe²⁺-based sludges and toxic metal leaching, thus being deleterious to the environment.^28,29

The PMS/PDS-based advanced oxidation processes (AOPs) can produce different kinds of radicals and non-radicals, especially SO₄·⁻ which possesses a longer life (30–40 μs) compared with ·OH normally generated in the Fenton process (smaller than 1 μs).²² Thus, PMS/PDS-based AOPs can be more powerful in organic pollutant degradation owing to a longer time in contact with the organic pollutants for a more sufficient reaction and thus for the better mineralization of organic pollutants.²² The inexpensive, efficient, and promising PMS/PDS-based AOPs in environmental remediation become the focus of current attention in the scientific and engineering disciplines.^30–34

PMS or PDS themselves possess inferior oxidation capability due to their low redox potentials.³⁵ An energy input or a catalyst is necessary to cleave PMS/PDS's O–O bonds to generate powerful radicals, especially SO₄·⁻ with a strong oxidation potential (2.6–3.1 V) outstripping commonly used ·OH radicals (1.8–2.7 V).³⁶ Alternatively, non-radical pathways are also widely reported through PMS/PDS's O–O activation for pollutant mineralization and for high-value-added organic product synthesis.^37–39 Many kinds of catalysts have been prepared for PMS and PDS activation to tackle persistent pollutants efficiently, normally including metal-based and metal-free ones.^40,41 Metal-based catalysts are widely demonstrated to be efficient, such as transition metal oxides,^42–44 bimetallic oxides and sulfides,^23,45 zero-valent metals,^46,47 supported metal oxides,^48–50 and supported single-atom metal catalysts.^51,52 However, the potential metal-species leaching problems, stability, and environmental concerns restrict their widespread applications. Therefore, it is highly desirable for the development of alternative catalysts to metal-based ones.⁵³

Carbon-based metal-free materials, also called carbocatalysts, have received substantial attention worldwide due to their structural tunability, superior conductivity, stability, nontoxicity, and enormous specific surface area.^29,36,54 Different kinds of carbocatalysts have been reported based on various carbon substrates (Fig. 1), such as graphene, carbon nanotubes (CNTs), fullerene, nanodiamonds (NDs), biochar (BC), carbon fiber (CF), ordered mesoporous carbon (OMC), and graphene-like carbon materials.^55–58 Despite that many kinds of carbon substrates possess unique structures and fascinating properties,^59,60 they are generally inactive as PMS/PDS activation catalysts without proper structure and surface property engineering due to their electroneutrality. Structural engineering of carbon matrix by physicochemical methods is thus essential to impart activity (Fig. 2). Typically, heteroatom doping is an effective way to break the electroneutrality of the carbon network by the regulation of charge and spin densities. On the other hand, structural defect generation is another viable method to grant activity, such as edge and vacancy defects. This paper delivered the first effort at the investigation of the active sites of carbocatalysts in PMS/PDS activation and catalytic purification applications, as well as the elucidation of the reaction pathways in PMS/PDS-based AOPs over various carbocatalysts. In the end, the current problems and future perspectives are outlined for this hot-spot area with intense research activities. The focused aspects of this timely overview are schematically shown in Fig. 3.

Fig. 1 Various carbon substrates (e.g., graphene, BC, CF, OMC, ND, and fullerene) that can be employed for synthesizing PMS/PDS activation carbocatalysts.

Fig. 2 Various carbocatalysts synthesized by doping/functionalizing/modifying a variety of carbon substrates, especially graphene and its derivatives, based on physicochemical methods.

Fig. 3 Schematic diagram for summarizing the topics reviewed.

Non-doped active sites

Many kinds of synthetic carbon nanomaterials are laden with structural defects. For example, graphene nanoribbons and quantum dots are enriched with edge defects. Even high-quality graphene sheets bear topological defects such as wrinkles, vacancies, and dislocations. These defects have been demonstrated to be active toward many kinds of catalytic processes including catalytic PMS/PDS activation and others. However, the nonstoichiometry of carbon materials makes it challenging to discern which kinds of defects dictate the catalytic activity. We put efforts into discussing different intrinsic non-doped sites in the carbon framework that may be associated with the activity toward PMS/PDS activation.

Sp² carbon network structure

Mostly, the reduction of GO and oxidized CNTs to produce RGO and thermally annealed CNTs, respectively, is resorted to recover the sp² carbon network capable of shuttling electrons and hence enhance the catalytic performance for PMS/PDS activation.⁶¹ Meanwhile, the adsorption interactions with PDS are enhanced when amorphous carbon is reformed to sp²-hybridized graphene by flash Joule heating (the PDS adsorption energies on the graphene and amorphous models were calculated to be −9.67 and −7.50 eV, respectively), as shown in Fig. 4a and b.⁶² Reduction treatments are effective in increasing the ratio of sp²- to sp³-hybridized carbon, thus bringing more active sp² carbon sites and lowering the inactive sp³ carbon sites.⁶¹ The free-flowing electrons provided by the conjugated π network can activate PMS/PDS into free radicals based on the following eqn (1) and (2).⁶³ Besides, the strong adsorption capability of graphene structure can also produce metastable complexes for non-radical pathways based on direct electron transfer (Fig. 4b).

e⁻ + S₂O₈²⁻ → SO₄·⁻ + SO₄²⁻ (1)

e⁻ + HSO₅⁻ → SO₄·⁻ + ·OH (2)

Fig. 4 (a) Mediation of the structure of sawdust-derived carbon by flash Joule heating; (b) reforming and catalysis mechanisms of engineered pyrogenic carbon by the flash Joule heating approach.⁶² (c) Schematic illustration of the relationships among graphene layer thickness on ND, sp²-to-sp³ carbon ratio, adsorption interactions with PMS, and reaction pathways. (d) Activation of PMS by the reactive edges of RGO and the corresponding reaction pathway.⁶¹

Sp²/sp³ hybrid structure

To construct the sp²/sp³ hybrid structure, Duan and coworkers modified the ND surface by thermal annealing, resulting in the formation of graphene layers with a thickness that could be tuned by adjusting the thermal annealing temperature.⁶⁴ The graphene layer number was evidenced to determine the catalytic performance and reaction pathway. With increasing the number of graphene layers on the ND surface, the O–O bond length and the carbocatalyst-to-PMS charge transport were reduced and weakened, respectively, along with declined electron density. Consequently, the activity for PMS activation and catalytic oxidation became lowered. On the other hand, the adsorption energy turned out to be more negative with increasing graphene layer thickness, indicating stronger adsorption interactions between PMS and carbocatalyst surface. This finding was also validated by the work of Fu et al.¹ who found that a positive linear relationship could be established between the direct electron transfer process-dictated catalytic efficiencies and sp²/sp³ carbon ratios. Therefore, while the moderate adsorption of PMS facilitated detaching the cleaved PMS molecules to yield SO₄·⁻ and ·OH radicals in the solution, the strong adsorption, enabled by the sp² carbon network, led to the non-radical pathway based on direct electron transfer without the generation of free radicals (Fig. 4c).

Zigzag edges

Apart from the sp² carbon network capable of electron shuttling, the zigzag edges of RGO with unconfined π electrons and electrophilic ketonic groups (C=O) were assumed to be active sites. Typically, zigzag edges decorated with ketonic groups were illustrated as the critical active sites for PMS activation toward the generation of SO₅·⁻ and SO₄·⁻ radicals (Fig. 4d),⁶¹ as well as ·OH radicals (Fig. 4b).⁶² However, the defects and structure transformation induced by excessive edges were unfavorable for achieving efficient PDS-based AOPs given that graphitic degree and conductivity were also reported to be essential in edge-enriched carbocatalysts by promoting electron transportation from pollutants to activated PDS (PDS*) based on a nonradical electron transfer pathway.³⁷

Therefore, it can be assumed that zigzag edges can be assumed as active sites for achieving radical (e.g., SO₅·⁻, SO₄·⁻, and ·OH) pathways. In contrast, graphitic structure (i.e., sp² hybridized carbon) facilitates the non-radical pathway (see Fig. 5).

Fig. 5 Schematic presentation of different reaction pathways for PMS/PDS activation, enabled by carbocatalysts' typical intrinsic structures: sp² hybridized carbon network and zigzag edges.

Structural defects

The chemical structure of carbocatalysts is complex, with non-stoichiometry, thereby causing the decoding of their active sites and pathways in the PMS/PDS-AOPs to be complicated. For example, different kinds of reaction pathways coexist in the single CNT catalyst bearing carbonyl groups and structural defects (e.g., vacancies). Carbonyl groups and structural defects are demonstrated to be the active sites that are associated with the two kinds of non-radical reaction pathways; the former involves the generation of ¹O₂, while the latter facilitates the direct electron transfer from the 2,4-dichlorophenol to PDS* (see Fig. 6).⁶⁵ However, the incorporation of carbonyl into the CNT lattice is normally accompanied by inducing structural defects, making it challenging to unravel the true origin of the activity for PMS/PDS activation and to understand the related reaction pathway.

Fig. 6 The dependence of the reaction pathway for PMS/PDS activation on the type of active sites on CNTs.

Single-heteroatom doping of carbon materials

To enhance the catalytic activity of non-doped carbocatalysts such as edged/defective graphene and CNTs, the doping treatments are essential to bringing a larger number of active sites and the catalytic site with promoted activity. Specifically, the catalytic activity of unzipped MWCNTs bearing enriched carbon edges was enhanced compared with intact MWCNTs in PDS activation toward bisphenol A (BPA) oxidation. The carbon edges could be further used as the hosting sites for heteroatoms to promote the catalytic activity of unzipped MWCNTs.³⁷ As such, it is promising to employ various un-doped carbocatalysts further to fabricate doped carbocatalysts since defective sites are more reactive for the incorporation of heteroatoms, which also likely results in active sites with superior activities for PMS/PDS activation. Many kinds of heteroatoms have been explored for carbon material doping, including N, S, O, Cl, F, B, P, and I, among others. Typical examples will be presented below, not covering all kinds of heteroatoms.

O doping

The most typical example is GO laden with O groups and bearing limited electron transferring ability and hence inactivity in PMS/PDS activation.⁶⁶ The reduction of GO to yield reduced graphene oxide (RGO) can improve the catalytic activity to a certain extent (Fig. 2). The same trend can be found between oxidized CNTs and thermally reduced counterparts.⁶⁷ It is worth mentioning that GO is the most widely used precursor to prepare various activated graphene materials (e.g., reduced GO (RGO), doped RGO, graphene nanoribbons (GNRs), and graphene quantum dots (GQDs), see Fig. 2) for environmental remediation, catalysis, sensing, and energy conversion and storage applications due to its ease to prepare, unique 2D structure, and tunable O groups.^6,60,68–73

Differentiation of specific O groups in the PMS/PDS activation is challenging due to the simultaneous variation of all kinds of O groups with regulation conditions. Some methods have been reported to finely mediate the surface O groups on the carbocatalyst surface. First, thermal annealing and flash Joule heating are effective to finely tune the ratio of C=O to C–O groups given the different thermal stabilities of C=O and C–O.^62,74 Second, the use of reagents to deactivate the specific O groups; for example, hydroxyl (–OH), ketonic carbonyl (C=O), and carboxylic groups (–COOH) on the surface of CNTs can be deactivated with benzoic anhydride, phenylhydrazine, and 2-bromo-1-phenylethanone, respectively.⁶⁵ Third, the modification of carbocatalyst surface to create or bring more specific O groups (e.g., ketonic C=O group content that can be increased by modifying CNTs with glutaraldehyde⁷⁵).

To conclude, the excess O doping of the carbon network is not conducive to the catalytic activity enhancement since O breaks the sp² carbon network structure of carbon materials (e.g., graphene and CNTs) and hence lowers the electrical conductivity as well as presents a steric hindrance effect on the interactions with PMS/PDS.⁶² Meanwhile, O doping extent, location, and type should be carefully controlled to realize highly efficient O-doped carbon materials.⁷⁶

Ketonic/carbonyl (C=O) groups. The adsorption and catalytic PMS activation performance of RGO could be improved by the activation treatment of RGO under a CO₂ atmosphere at high temperatures.⁷⁷ Such activation significantly increased the specific surface area from 200 m² g⁻¹ (RGO) to more than 1200 m² g⁻¹ (activated RGO). Meanwhile, new active edges and oxygen groups (e.g., C=O and C–O) at the edges and on the surface of RGO were created. The catalytic activity toward PMS activation depended on the oxygen group content of the samples. The electron-rich ketonic groups (C=O) on the activated RGO were conjectured to be an active site for PMS activation to produce SO₄·⁻ radicals.

Despite the radical pathway mentioned above, the C=O species have been demonstrated to mainly catalyze the PMS/PDS activation toward the generation of non-radicals. For example, the electrophilic ketonic C=O species of carbocatalysts (e.g., NDs⁷⁴ and CNTs^65,75) have widely been verified to be metal-free active sites in the activation of nucleophilic PMS and PDS toward the generation of singlet oxygen (¹O₂)-based non-radicals as the active sites for organic pollutant degradation. In the ND system, electron transfer from nucleophilic PMS to electrophilic C=O for the generation of ¹O₂. The extraction of electrons from PMS leads to the generation of SO₅·⁻ radicals that were subsequently self-reacted further to yield ¹O₂, with the reaction formula shown in eqn (3)–(5).⁷⁴ The ¹O₂ formation in a CNTs/PDS system is illustrated by eqn (6)–(9)⁷⁵ where the C=O groups of CNTs were firstly attacked by S₂O₈²⁻, forming a peroxide adduct, which was followed by a series of reactions involving the dioxirane intermediate generation and decomposition. The formation of ¹O₂ was also regarded as the recombination of superoxide radical (·O₂⁻) intermediates⁷⁵ (see eqn (10)) and the reaction of ·O₂⁻ with H₂O (ref. 78) (see eqn (11)), as well as the nucleophilic addition reaction between C=O groups of carbocatalysts and PDS/PMS shown in eqn (12).⁷⁸ Such a ¹O₂ non-radical pathway is selective to electron-rich organics and can render a satisfactory AOP for actual water body treatments.

HSO₅⁻ → SO₅·⁻ + H⁺ + e⁻ (3)

SO₅·⁻ + SO₅·⁻ → S₂O₈²⁻+¹O₂ (4)

SO₅·⁻ + SO₅·⁻ → 2SO₄⁻+¹O₂ (5)

O₂·⁻ + O₂·⁻ + 2H⁺ → H₂O₂ + ¹O₂ (10)

2O₂·⁻ + 2H₂O → H₂O₂ + 2OH⁻ + ¹O₂ (11)

Alternatively, the C=O groups of CNTs were also reported to trigger another non-radical pathway (namely direct electron transfer from the 2,4-dichlorophenol (2,4-DCP) to PDS with the assistance of conductive CNTs as a bridge based on a CNTs-PDS* metastable complex) instead of the ¹O₂-based one, while CNT defects were the active sites for ¹O₂ generation.⁶⁵ The reaction pathway for the degradation of 2,4-DCP over the C=O groups and defects of CNTs is schematically illustrated in Fig. 7a.

Fig. 7 (a) Non-radical reaction mechanism underlying the degradation of 2, 4-DCP by C=O groups and defects of CNTs.⁶⁵ (b) High-resolution XPS O1s spectra of different NDs with oxygen groups engineered by thermal annealing.⁷⁴ (c) Illustration of nonradical oxidation of phenol with different CNT by the PDS/CNT system.⁷⁶

The ketonic C=O groups of NDs can be well regulated by thermal annealing at moderate temperatures in the range of 550–800 °C.⁷⁴ Such a temperature range is beneficial to maintain carbon hybridization properties and micromorphology. The higher the thermal annealing temperature over 550–800 °C, the less the C–O groups and the more C=O groups that can be found based on the quantitative analysis by XPS (Fig. 7b). Meanwhile, other structural characteristics of NDs remain almost unvaried. This finding indicates that the discreet thermal treatment can convert the relatively unstable C–O to stable C=O groups,^62,74 which provides a useful platform to investigate the role of C=O in the catalytic activation of PMS/PDS. Indeed, the ratio of C=O to C–O exhibited a positive linear relationship with the reaction rate constant (k) of the AOP with the correlation coefficient (R²) reaching 0.99, which consolidated that C=O groups play a vital role in PMS activation toward the degradation of organics, in the case of 4-chlorophenol.⁷⁴ Alternatively, the C=O group content could be increased by glutaraldehyde (GA) modification, and the grafted GA could react with amino groups of carbocatalysts to result in the new formation of C=O.⁷⁵

The catalytic effect of O-containing groups can be verified by post-characterizations such as XPS. For example, the contents of C=O and C–O groups can be monitored by XPS. If the numbers of C=O and C–O groups of the catalysts become smaller and larger after the AOP while turning to be increased and reduced after catalyst regeneration treatments, respectively, the C=O groups can be assumed to be the active sites.⁷⁹

Hydroxyl (–OH) groups. Oxygen groups could be effectively removed from the commercial CNT surface via thermal annealing, allowing the zeta potential of the surface-charged CNT to be less negative in neutral solutions and strengthening the adsorption interactions with PDS anions for the formation of metastable CNT-PDS* intermediate due to the weaker electrostatic repulsion. While the carbonyl and carboxyl groups were demonstrated to mainly contribute to the negative zeta potential,^4,76 their removal was important to make the zeta potential of the CNT less negative. Moreover, negatively charged carboxyl groups were also evidenced to increase the work function of carbocatalysts, thereby inhibiting the electron transfer process and impeding PDS activation.⁴ The adverse carboxyl groups could be removed by an ultrafast and energy-saving microwave-assisted process.⁴ Contrary to the electron-withdrawing properties of carbonyl and carboxyl groups, hydroxyl groups are electro-donating ones that can strengthen the π-conjugated system of the sp² carbon network. Interestingly, the thermally annealed CNTs/PDS oxidation system was analogous to an anodic oxidation system.⁷⁶ The adsorption of PDS onto CNTs significantly increased the oxidation potential of CNTs, while the addition of phenol dropped the potential. Nevertheless, if the oxidation potential was smaller than that for phenol oxidation (∼0.41 V), the phenol degradation reaction could not occur. Only when the oxidation potential exceeded that for phenol oxidation, phenol could be efficiently oxidized.⁷⁶ The mediation of the surface oxygen groups of commercial CNTs for the strong adsorption of PDS and formation of the metastable complex by thermal annealing is schematically illustrated in Fig. 7c.

N doping

The incorporation of N heteroatoms into the carbon frameworks has gained the most attention in developing highly efficient metal-free carbocatalysts for PMS/PDS activation since the activity of the N groups is reported to be much higher than that of the O groups.^80,81 The N heteroatom has a similar atomic radius to that of C (atom radius: N, 0.71 Å; C, 0.75 Å) and a great electronegativity difference from C (χ_C = 2.55, χ_N = 3.04), thereby making it most efficient to break the chemical inertness of sp² carbon network compared with other heteroatoms. Although N is easier to dope into the graphene lattice compared with other heteroatoms, the thermal annealing treatments with a temperature of at least 500 °C are usually required to dope N heteroatoms into carbon lattice and render the catalytic activity for PMS/PDS activation.⁸² Different N-doped carbon catalysts can be prepared by mainly two kinds of methods: one is the post-treatment of carbon substrates with N precursors. For example, N-doped graphene (NG) and N-doped CNTs can be normally prepared by post-treatments of graphene derivatives (especially GO) and CNTs with N-containing species such as urea,^82,83 ammonia,⁸⁴ melamine,⁶⁷ and ammonium nitrate.⁶⁶ The other one is the one-step in situ doping with the precursors rich in carbon and nitrogen. The precursors can be self-prepared, such as commonly used metal–organic framework (MOF) materials, N-containing polymers, and biomasses, particularly MOFs.^85–89 The metallic species originating from MOFs can be removed by thorough washing treatments with acids. If the precursors lack sufficient N, they can also be pre-modified by N-enriched polymers (such as polydopamine⁹⁰) before the in situ doping via one-step pyrolysis under an inert atmosphere.

Since graphitic or quaternary N (N-Q), pyridinic (N-6), pyrrolic (N-5), and oxidized N (N–O) can normally be simultaneously formed during the carbocatalyst preparation, it is still a current challenge to incorporate a specific N configuration into the carbon lattice. While N–O is reported to be inactive for PMS/PDS activation, we will mainly focus on the N-Q, N-6, and N-5 as the critical active sites for PMS/PDS activation. Discrepancies exist between the results from different reports involving similar N-doped carbon as the PMS/PDS activation catalysts. Therefore, to be comparable, the activity and origin of the carbocatalysts from different studies should be tested under the same metric in order to disclose the actual sources of activity.

N-Q configuration. For the N-Q configuration, its high electronegativity enables the attraction of electrons from the adjacent C, making the N-Q and adjacent C bear negative and positive charges, respectively. Thus, the adjacent C is prone to adsorb negatively charged ionized PMS (HSO₅⁻, and SO₅²⁻) and PDS (HS₂O₈⁻, and S₂O₈²⁻).^67,82,91 Consequently, once the electron-rich N-Q sites offer electrons to PMS/PDS, causing the PMS/PDS's O–O bonds to be split into SO₄·⁻, ·OH, and ·O₂⁻ radicals, whereas the electron-deficient sites (e.g., the C adjacent to N) oxidize PMS/PDS into non-radicals, such as ¹O₂.⁹² Meanwhile, ·O₂⁻ radicals can also be formed by converting dissolved O₂ in the solution in the presence of a carbocatalyst and PDS/PMS.⁸¹

The creation of the N-Q configuration can be achieved by regulating the calcination temperature.⁹³ The elevation of the annealing temperatures from 550 to 1000 °C resulted in higher N-Q contents from 18.83 to 62.81% in N-doped CNTs prepared by the initial oxidation of CNTs with mixed acids and then thermal treatments in an NH₃ atmosphere, as shown in Fig. 8a.⁹³ Even though the high-temperature thermal treatment can break C–N bonds and lower the doped N content, the N-Q proportion can be increased due to the higher thermal stability compared with other N configurations (i.e., N-6, N-5, and N–O).^67,85,93,94 The porous architecture engineering can maximize the N-Q exposure and enhance the adsorption performance and catalytic activity,⁹² while enhanced graphitization at high temperatures contributes to the faster electron transfer over the sp² carbon network.⁸⁵

Fig. 8 (a) Schematic illustration of the preparation of N-doped CNT by a two-step process. (b) The linear plot of the reaction rate constant (k_obs) as a function of the N-Q atomic content based on the results of examining the N-Q contents in different samples.⁹⁵ (c and d) High-resolution XPS N1s (c) and C1s (d) spectra of N-doped carbon nanosheets (NCN-900) before and after spent to catalyze PMS activation and BPA degradation.⁹⁶

Besides verification based on theoretical calculation, the role of different N configurations can be experimentally discriminated by the relationship between the specific N configuration content and the apparent rate constant (k) of the oxidation reaction of organics. For example, only N-Q was able to establish a positive linear correlation with k (R² = 0.953, see Fig. 8b), revealing that N-Q was the probable active site.⁹⁵ N-Q and its neighboring C serving as the active sites can also be revealed by the high-resolution XPS N1s and C1s spectra, where their corresponding deconvoluted peaks will be shifted after being used for catalytic peroxide reactions. Specifically, the spent N-doped carbon nanosheets exhibited a deconvoluted peak assigned to the N-Q configuration that shifted to lower binding energy while N-6 and N-5 were almost unchanged (Fig. 8c). On the other hand, the C–N bond that shifted downward by 0.17 eV was noted in the high-resolution XPS C1s spectra after the use of the N-doped carbon nanosheets in the AOPs (Fig. 8d).⁹⁶

As early as 2016, Wang et al. prepared high-quality NG by pyrolyzing a mixture of glucose (as a C source), urea (as an N source), and ferric chloride (as both a catalyst and a template). The controlled pyrolysis process yielded the O content to be in the range of 2–4 at% while generating 0.5–1.8 at% N content in an NG sample. Phenol could be efficiently degraded by PMS in the presence of the NG. The electron paramagnetic resonance (EPR) and kinetic studies revealed that the PMS activation by the NG resulted in ·OH and SO₄·⁻ radicals which played a critical role in phenol degradation.⁹⁷ The authors presumed that N-Q, N-5, and N-6 all contributed to the PMS activation, with N-Q as the dominant one due to the high charge and asymmetric spin densities generated by N-Q.⁹⁷ However, the use of ferric chloride as the catalyst likely left Fe-based metal species to be confined in the NG lattice, thus changing the electronic structure and activity origin of NG.

To enhance the N heteroatom concentration and optimize the N configuration for PMS activation, Chen et al. prepared NG by the thermal annealing of a mixture of GO and urea.⁸² The thermal annealing temperature of 600 °C resulted in the NG with a high N heteroatom concentration (16.0 wt%). The N doping effect was demonstrated to be more significant than the specific surface area impact on the catalytic activity for PMS activation to degrade sulfacetamide (SAM). The SAM degradation mainly proceeded by the SO₄·⁻ radical pathway that contributed at least 3 times more to that by the non-radical pathway (Fig. 9a). DFT calculation revealed that N-Q was conjectured to be the primary active site for PMS adsorption and SAM degradation. Even higher N contents (23.78 wt% and 31.93 at%) were reported in N-doped GO⁹⁸ and N-doped porous carbon⁷⁸ catalysts that were prepared by pyrolysis of a mixture of GO and melamine under an air atmosphere at a lower temperature (i.e., 350 °C) for 30 min and pyrolysis of a mixture of cyanuric acid, L-ascorbic acid and melamine at a moderate temperature (i.e., 550 °C) under N₂ atmosphere for 4 h, respectively.

Fig. 9 (a) Mechanism of PMS activation and SAM degradation on NG600.⁸² (b and c) theoretical models showing the adsorption of H₂O (b) and PS + H₂O (c) on N-SWCNTs (the grey, white, blue, red, and yellow atoms are C, H, N, O, and S atoms, respectively); (d and e) the O–O bond length (l_O–O) and adsorption energy of H₂O (d) and PS (e) under different conditions: free molecule, adsorption alone on N-SWCNTs, and co-adsorbed on N-SWCNTs; (f) proposed mechanism of PS activation over N-SWCNTs.⁶⁷

A new activation mechanism was proposed for the PDS/N-SWCNTs system, where H₂O molecules played an essential role in varying the adsorption energy, as well as H–O and O–O bond lengths of H₂O and PDS, respectively. The presence of H₂O resulted in a much-stretched bond length and significantly increased adsorption energy (Fig. 9b–e). PDS promoted the dissociation of H₂O to produce –OH groups followed by ·OH radical generation while PDS directly extracted electrons from the N-SWCNTs matrix to yield SO₄²⁻ (Fig. 9f).⁶⁷

Apart from the radical pathway achieved by N-Q configurations of N-doped carbonaceous materials, the non-radical pathways are also frequently reported on metal-free carbocatalysts with giant specific surface areas, such as N-doped CNTs^67,93 and N-doped biochar,⁹⁴ since N-Q induces positive charges onto the adjacent carbon atoms which further strongly interact with PMS and PDS anions via electrostatic attraction interactions to form N-doped carbon surface-bound metastable PDS*/PMS* species with the O–O bonds being stretched.^67,93–95 Electrons can then be transferred from organic pollutants to N-doped-carbon-surface-confined-PDS*/PMS* and release PDS/PMS via electron transfer to fulfill the redox cycle. The degradation capability based on such a non-radical-mediated pathway is usually selective to the type of organics and dependent upon the ionization potential (IP) of organics and the metastable oxidation state of the carbocatalyst with electrons abstracted by PMS.^86,95 Only the organics with an IP value of smaller than 9.0 eV can be oxidized by the N-doped carbonaceous nanosphere/PMS system.⁸⁶ Generally, the organics bearing more electron-donating groups have smaller IP values whereas the electron-withdrawing groups lead to an opposite trend,⁹⁵ indicating that the organics with more electron-donating groups (such as –OH) can be easier to degrade.

Such a surface-confined reactive complex-based non-radical pathway can be well evidenced by many tactics. First, the electrochemical impedance spectroscopy (EIS) that shows the charge transfer properties can reflect the electron-transfer non-radical pathway. The lower charge-transfer resistance and the greater conductivity imply the faster electron transfer kinetics. Second, linear scan voltammetry (LSV) can be employed to elucidate the surface-confined non-radical pathway. The addition of PMS/PDS significantly increases the current on the carbocatalyst electrode during LSV, indicating the interactions between PMS/PDS and carbocatalysts to yield metastable reactive complexes. The further injection of organics induces another notable elevation of current, revealing that a current flow is generated on the carbocatalyst surface from organic pollutants to the surface-confined nonradicals. Third, the enhanced adsorption of organics onto carbocatalysts is proven to be a decisive factor in the surface-confined non-radical pathway since more organics can be involved in the direct charge transfer process.⁹⁴ Fourth, Raman spectroscopy can be harnessed to manifest the direct electron transfer pathway by analyzing the carbocatalyst surface-confined PDS/PMS anions.⁹⁹ Fifth, pre-mixing PMS/PDS and carbocatalysts for various durations before adding organics is also an effective way to verify the direct electron transfer pathway. If the addition of organics at different durations results in similar degradation efficiencies, the degradation process is governed by the mediated electron transfer mechanism.⁸⁵ Sixth, adding an electron trapper (e.g., potassium iodide and potassium dichromate) can block the electron transfer between the surface-confined active complex and organics and thus substantially depress the catalytic oxidation efficiency.¹⁰⁰ Seventh, adding a reagent (e.g., NaClO₄) that is capable of inhibiting the surface interactions between PMS/PDS and carbocatalyst but not able to disturb the radical pathway is an effective way to consolidate the electron-transfer non-radical pathway.¹⁰¹ Last, the formation of metastable carbocatalyst-PMS*/PDS* complexes with high redox potential can also be revealed by the correlation between PMS/PDS adsorption amount and open circuit potential. The good linear relationship between the open circuit potential and k suggests that the electron transfer mediated by the metastable complex plays a vital role in the AOPs.⁹⁰

Besides the non-radical pathway based on the N-Q doped carbocatalyst surface-confined PMS*/PDS* complexes, the ¹O₂-based non-radical pathway is also widely reported as the non-radical generated for the PMS/PDS-carbocatalyst system.^85,96 The active sites for ¹O₂ generation can be the electron-deficient carbon adjacent to the electron-rich N-Q reported with the highest electron density (Fig. 10a–c), where PMS/PDS can be adsorbed and oxidized by electron transfer. By contrast, electron-rich N-Q with the highest electron density can reduce PMS/PDS to yield a small amount of SO₄·⁻ and ·OH radicals.⁹⁶ The ¹O₂ detection can be realized by ¹O₂ quenching experiments using ¹O₂ scavengers, such as β-carotene,⁸⁵L-histidine,⁷⁸ furfuryl alcohol,⁸⁶ and sodium azide (NaN₃).⁷⁴ The EPR/electron spin resonance (ESR) spectroscopy is also frequently employed to detect different kinds of reactive oxygen species (ROS), including SO₄·⁻, ·OH, ·O₂⁻, and ¹O₂. Furthermore, the replacement of H₂O with D₂O can lead to a prolonged lifetime of ¹O₂. Consequently, the organic degradation efficiency is promoted if the AOP is dictated by the ¹O₂-based non-radical pathway.⁸⁵ A comparison in the degradation of BPA (with electron-donating groups) and diphenhydramine (without electron-donating groups) that are sensitive and immune to ¹O₂, respectively, can also be adopted to detect the role of the ¹O₂ in the AOPs.⁹⁶ If the ¹O₂ is the primary active site, BPA can be effectively degraded while ineffective to diphenhydramine.

Fig. 10 (a) Structural models and corresponding 2D charge distribution images of un-doped carbon and N-doped counterparts bearing various N doping configurations including N-6, N-6 with the vacancy (N-6/V), N-5, N-5 with the vacancy (N-5/V), and N-6 (from left to right). (b and c) Bader charges of different N configurations (b) and carbon atoms neighboring N-Q (c). Blue and brown colors denote N and C atoms, respectively.⁹⁶

The N-Q-related activity and reaction pathway also depend on the type of superoxides^67,84 and carbocatalysts' adsorption performance.¹⁰² For example, PDS was reported to be more difficult to activate due to its symmetric structure with a more stable O–O bond, in contrast to the asymmetric structure of PMS that was more ready to be adsorbed and polarized by the positively charged carbon adjacent to the doped N.⁸⁴ Besides, while the N-Q configuration enabled N-CNTs to catalyze the PMS activation through a non-radical pathway, it served as a bridge to drive the electron transfer from water molecules to PDS for ·OH radical generation.⁶⁷ On the other hand, once the simultaneous addition of PDS and N-doped carbon xerogels into an acid orange 7 (AO7) solution, adsorption promoted catalytic AO7 degradation, whereas the pre-adsorption of AO7 declined the catalytic performance due to the competing adsorption between PDS and organics (especially those bearing sulfonic groups) onto the same active sites.¹⁰²

N-Q cooperated with N-6. While N-Q induces the generation of high local positive charge densities on the adjacent carbon, facilitating the adsorption of PMS/PDS anions, N-6 shows a high adsorption affinity to some organics (e.g., tetracycline (TC)), with the adsorption energy of 0.16 eV and the shortest distance (2.49 Å) from TC to the model catalyst substrate based on DFT calculation shown in Fig. 11a–d.⁷⁹ Such dual N configurations' cooperation effects result in the direct electron transfer pathway for the activation of PMS and PDS to mineralize TC.^78,79 XPS can be employed to monitor the activity of different N configurations. Specifically, the N-5 content of the carbocatalyst increased after use, while those of N-Q and N-6 decreased.^78,79 Correspondingly, the contents of N-Q and N-6 were increased to a value near the fresh catalyst after restoring the used carbocatalyst. The N-Q and N-6 also coexisted in an N-doped GO catalyst for the cooperative catalytic activation of PDS for phenol and oxalic acid degradation.⁹⁸ The N-6 vanished after the use of the catalyst, whereas N-Q remained. As a result, part of the catalytic activity was lost. Alternatively, N-6 and N-Q could cooperate to catalyze the oxidative polymerization of BPA based on PDS to turn waste into treasure via the first formation of BPA phenoxyl radicals and then oxidative polymerization to yield a valuable polymer under alkaline conditions. While N-6 was reported to serve as the adsorption site for BPA, N-Q was assumed as the catalytic site for PS moderate activation, giving rise to metastable O–O peroxo that was extended moderately, instead of being decomposed to SO₄·⁻. As a result, the high reactivity of the O–O peroxo was achieved, thus allowing the H to be abstracted from BPA to form BPA phenoxyl radicals.¹⁰³

Fig. 11 (a–d) Local adsorption configurations of TC on un-doped graphene (a), and N-Q (b), N-6 (c), N-5 (d) -doped graphene based on DFT calculations.⁷⁹ (e) Contents of N-5, N-6, and N-Q in five N-doped carbon samples. (f) Catalytic mechanism of an N-doped carbonaceous catalyst/PMS system. The gray, white, red, blue, and yellow balls represent to C, H, O, N, and S atoms, respectively.¹⁰⁴

N-5/N-6 configuration. For N-6 and N-5, their lone pair of electrons allows them to be the Lewis basic sites, where the electrons can be transferred to PMS/PDS molecules, breaking PMS/PDS's O–O bonds and ensuingly generating SO₄·⁻ radicals. For example, the thermal reaction between GO and ammonium nitrate at 350 °C for 1 h rendered N doping and meanwhile reduction of GO, leading to the generation of an NG catalyst for PMS activation to degrade phenol, where N-Q was not the dominant active site while N-5 and N-6 played critical roles in activating the free-flowing π electrons of sp² carbon through conjugating with their lone-pair electrons.⁶⁶ To reduce the N-Q contents while obtaining a higher ratio of N-5 and N-6, the thermal annealing temperature should not be too high. This is because N-Q is more thermally stable, and meanwhile, N-Q is prone to be imbedded into the sp² carbon network at high temperatures. For example, the doping of glucose-derived carbon with melamine at a low annealing temperature of 550 °C resulted in a higher content of N-5 and N-6 and fewer N-Q configurations (Fig. 11e).¹⁰⁴

Instead of activating the free-flowing π electrons of sp² carbon within the carbon matrix, the N-5 configuration bearing lone-pair electrons was reported as the catalytic site of an N-doped carbonaceous catalyst to attract electrophilic PMS and impel the electron transfer to PMS.¹⁰⁴ Such N-5 doping could transform a radical pathway of the un-doped carbon counterpart to a non-radical pathway based primarily on surface-confined-metastable-complex-mediated electron transfer, with ¹O₂ also making some contributions (Fig. 11f). The AOPs based on the non-radical pathway were selective, and the IP values of various aromatic compounds bearing different substituents determined the degradation performance.

N-5/N-6 configuration coupled with defects. While introducing N configurations, structural defects (e.g., vacancies) are also created around the N configurations, which cooperate with N dopants to alter the electronic structure of N-doped carbon catalysts and enhance the adsorption and catalytic degradation of organics. For example, PMS activation was distinctly enhanced by the vacancies created around the N-5 and N-6 albeit with still inferior catalytic activity compared with the N-Q based on experimental studies and DFT computation with the adsorption energies and O–O bond lengths as the descriptors.⁹⁶ The more active the catalytic sites, the larger the adsorption energy and the longer the O–O bond could be.

Amine groups. Amine groups can be introduced by a milder hydrothermal reaction since more severe high-temperature pyrolysis treatments lead to the carbocatalysts bearing only N-5, N-6, and N-Q.⁸⁰ The amine groups at highly acidic conditions can thus be quaternized to bear positive charges, facilitating the strong adsorption of PMS/PDS anions via electrostatic attraction. Such adsorption interactions are favorable for the electron-transfer-based non-radical pathway under acidic conditions. Indeed, the non-radical pathway mainly governed the AOPs with an aminated N-doped graphene hydrogel as a carbocatalyst.

Other N-involved active sites. To enhance the catalytic performance of N-doped carbonaceous materials, one can further resort to activation treatments that can not only significantly increase the specific surface area but also bring highly active sites for PMS/PDS activation. Specifically, CO₂ activation was coupled with N doping for porous NG preparation, and the structural vacancies with two N heteroatoms were deemed to be highly active for PMS activation toward phenol degradation based on experimental studies and DFT calculation.⁸⁴ The sequence of CO₂ activation and N doping was also investigated, and only the CO₂ activation that proceeded in the first step followed by N doping could result in the generation of a favorable configuration (namely, active vacancies with two N heteroatoms), as schematically shown in Fig. 12. Such active sites facilitated the adsorption and activation of PMS to form a surface metastable complex with a prolonged O–O bond, rendering a non-radical pathway for phenol oxidation.

Fig. 12 Schematic representation of the fabrication process of (a) N-PRGO (first CO₂ activation and then N doping) and (b) P-NRGO (first N doping and then CO₂ activation).⁸⁴

S doping

The S heteroatom possesses a larger radius than C (1.03 vs. 0.75 Å), thus being favorable for the generation of structural defects in carbon crystal lattices such as edge sites. More importantly, S doping can result in S-containing groups (e.g., thiophene-S) which alter the electronic structure of the sp² carbon and hence break the chemical inertness.

Thiophene-S. S doping of the carbon matrix was realized by pyrolysis of self-prepared polythiophene in the presence of KOH as an activator.¹⁰⁵ It was found that high-temperature thermal annealing enabled the removal of acidic functional groups, most of which (e.g., oxidized S and carboxyl groups) were assumed to have an electron-withdrawing capability. Nevertheless, the high-temperature pyrolysis process also helped to elevate the ratio of thiophene-S (–C-S-C–) to some oxidized S. The thiophene-S played a critical role in enhancing the catalytic performance since its insertion into the sp² carbon network imparted the adjacent carbon atoms with high spin density and thus facilitated charge localization. As a result, the chemical inertness of sp² carbon was broken, along with improving the catalytic performance in PMS activation. Later, the same group prepared S-doped carbon using a potential waste, poly(phenylene sulphide), as a precursor based on the aforementioned method albeit with somewhat simplification.¹⁰⁶ The more systematic later work revealed that the adsorption of organics played a vital role in PDS activation based on the result that a good linearity could be established between the equilibrium adsorption amount and the PDS oxidation reaction constant. The quenching tests revealed that surface-confined radical and non-radical pathways were both involved in the efficient degradation of phenolic compounds. Besides, they found that more Cl substitution in the phenolic compounds resulted in enhanced hydrophobicity and π conjugation and, consequently, more efficient degradation.

I doping

I₃⁻ and I₅⁻. The electronegativity of I (χ_I = 2.66) is slightly higher than that of C (χ_C = 2.55) but much lower than that of N (χ_N = 3.04). The atom radius of I (1.33 Å) is notably larger than that of C (0.75 Å) and N (0.71 Å). Like the S doping, the I doping is likely to enhance the spin densities and create active sites in the conjugated π system. The I-doped carbocatalyst could be prepared by pyrolyzing the mixture of oxidized carbon fiber (CF) and NH₄I,¹⁰⁷ which was subsequently employed to activate PMS for acid red 1 (AR1) removal. Such an I-doped CF carbocatalyst was tolerant to aqueous solutions with a wide pH range and resistant to a typical dye auxiliary (i.e., NaCl) that was recognized as an effective radical capturing agent. The existence of I with different valent states (i.e., I₃⁻ and I₅⁻) was assumed to be the origin of the activity in PMS activation. The reaction between I₃⁻ and PMS resulted in the formation of HO₂· and O₂·⁻ radicals, which could be further converted into non-radical ROS (i.e., ¹O₂) via electron transfer. On the other hand, the persistent free radicals might donate electrons for the reduction of I₅⁻ to I₃⁻. The transformation between I₃⁻ and I₅⁻ was regarded as the critical step in determining the catalytic process.

B doping

Unlike other heteroatoms such as N, S, and I, the electronegativity of B (χ_B = 2.04) is lower than that of C (χ_C = 2.55), enabling the adjacent C to be negatively charged. In this case, doped B is positively charged, rendering the B itself as the potential active site to adsorb and activate PDS/PMS for AOPs based on radical and non-radical pathways. In addition, B possesses great strength, good chemical resistance, and high hardness, making it promising in the design and development of durable carbocatalysts for AOPs. The doping of B into different CNTs was reported.⁹⁹ The relationship of the wall thickness of different types of CNTs (including SWCNTs, DWCNTs, and MWCNTs) was studied as a function of catalytic activity. The larger the thickness, the higher the catalytic activity was demonstrated. The B doping not only induced a charge redistribution of the adjacent C but also triggered the formation of active edges with B atoms, affording a satisfactory catalytic activity. Radicals (including ·OH, SO₄·⁻, and O₂·⁻) and non-radicals (¹O₂ and B-CNTs-PMS*) were proven to coexist in the B-CNT/PMS system.

P doping

P-doped carbon materials are also electron-rich since P possesses the same number of valence electrons as N.¹⁰⁸ However, the lower electronegativity of P (χ_P = 2.19) relative to C (χ_C = 2.55) and N (χ_C = 3.04) allows P to be partially positively charged and hence work as the active sites, which is unlike the case of N that makes the adjacent C positively charged due to N's higher electronegativity (3.04 vs. 2.55). On the other hand, the much larger atomic radius of P compared with C and N renders P capable of distorting the underlying carbon structure to a larger extent when compared with the case of N doping, thus generating more structural defects. In addition, P has the additional vacant 3d orbitals and the valence electrons in the third shell relative to N, which might impart unique catalysis performance to P-doped carbon. A P-doped carbon fiber material was reported to catalyze PMS activation for sulfamethoxazole degradation and highly toxic Cr(VI) reduction to Cr(III).¹⁰⁹ The incorporation of P was able to enhance the electron density of carbon fiber, facilitating electron transportation. Once electron transfer to PMS occurred during the catalytic processes, SO₄·⁻ and ·OH free radicals, as well as ¹O₂ non-radical species, were generated for organic pollutant degradation. Meanwhile, electrons could also be transferred to inorganic Cr(VI) for its reduction to Cr(III).

Other single heteroatom doping

The heteroatoms other than commonly-used ones, including O and N (especially N), are rarely reported as single heteroatoms to dope carbon matrix. Although some single heteroatoms were employed for doping carbon matrix, the resulting activities are not satisfactory (e.g., single B-, P-, or S-doped graphene that was proved to have limited catalytic activity for PMS activation for phenol removal^110–112 albeit with some reports that disclosed single B-, P-, or S-doped carbon materials possessed decent activities in PMS/PDS activation). It is therefore highly challenging to explore non-N dopants for carbon materials. Alternatively, the use of non-N heteroatom to co-dope carbon matrix together with N becomes the focus of current activities to alter the electronic structure of the carbon matrix to a significant extent, enhance the activity and stability of each catalytic site, and increase the number of active sites.

Dual-heteroatom doping of carbon materials

While N, P- or N, I-dual-doped graphene is not satisfactorily active in PMS activation,⁶⁶ N, S-dual-doped graphene shows good activity for phenol and BPA degradation.^110,113 Thus, not all the co-doping treatments of N-doped carbocatalysts can further enhance the catalytic activity in PDS/PMS activation. The proper location of the dual-doped heteroatom sites is essential for obtaining a synergistic effect from N and other heteroatoms. For example, the electron-deficient B connected to electron-rich N will counteract the positive effect of spin and charge redistribution enabled by N, and the N-imparted breaking of the chemical inertness of sp² carbon will thus be invalid in the presence of adjacent B.

N, O dual-doping

Doping N into a carbon lattice is usually accompanied by the incorporation of O groups and by the coexistence of O groups in the carbon substrate. Recently, N, O-dual-doped carbocatalysts were synthesized by a pyrolysis method (Fig. 13a) for the activation of PDS toward TC degradation. The N-6 was demonstrated to be the adsorption sites while C=O groups were verified to be catalytically active sites. The cooperation between N-6 and C=O results in a fast catalytic reaction kinetics and TC degradation efficiency, with the k value calculated to be as high as 1.50 min⁻¹ which was the largest one compared with the data before the work.¹¹⁴ The dual function of these adsorption and catalytic sites shortened the migration distance between the ¹O₂ generated on C=O and TC molecules adsorbed, as schematically shown in Fig. 13b. Such efficient structural features of N, O-dual-doped carbocatalyst could even allow a continuous flow test with a flow rate of 4 mL min⁻¹ to be stable and efficient in TC removal for 4 h. More recently, N, O-dual-doped CNTs were employed as a carbocatalyst for PMS activation to degrade acetaminophen. N-6, N-Q, and C=O were reported as the primary active sites. N-6 reinforced the electron population in the sp² carbon framework, resulting in the formation of surface-bound radicals to initiate a non-radical pathway (without producing radicals to the solution). By contrast, N-Q induced the generation of a carbocatalyst-PMS* reactive complex that facilitated the non-radical pathway based on direct electron transfer. The lone-pair electrons of O in the C=O group might partially participate in electron transfer for PMS activation, whereas electron-donating –OH groups were conjectured to be unfavorable for the electron-transfer oxidative steps.

Fig. 13 (a) Schematic illustration for the preparation of an N, O-dual-doped carbon; (b) schematic representation for designing and creating the dual reaction sites on the N, O dual-doped carbocatalyst: adsorption site for TC (translucent blue background color) and catalytic activation site for PDS (translucent red background color).¹¹⁴ (c) Schematic representation of the N–C-900 fabrication process; (d) schematic illustration of PDS activation and PCA degradation mechanism. N–C and PCA refer to N-doped carbon and p-chloroaniline, respectively.⁸⁹

Judging from the linear relations between both the N-Q and C–O contents and reaction rate constants (k), Liu and co-workers proposed that N-Q and C–O were the active sites of N, O-dual-doped porous carbon prepared using a MOF (i.e., ZIF-8) as the precursor (Fig. 13c) for PDS activation to generate comparatively fewer radicals and dominant non-radicals (including ¹O₂ and carbocatalyst surface-confined PDS* metastable complex), as illustrated in Fig. 13d.⁸⁹ However, their results showed that after PDS-based AOPs, C–O groups decreased from 24.7% to 8.8%, and the subsequent thermal annealing at 900 °C for 1 h remarkably elevated the C–O percentage from 8.8% to 30%. We presume that the catalytic oxidation process is likely to oxidize the carbocatalysts to increase the C–O content, while thermal annealing at such a high temperature (namely, 900 °C) is prone to graphitize the N, O-dual-doped carbon network, resulting in O group removal and C–O content decrease. Therefore, future work can be conducted to probe the mechanism of such unexpected and interesting findings further (e.g., to find whether the C–O decrease is due to its conversion into C=O).

N, B dual-doping

To improve the catalytic activity of N-doped carbonaceous materials in PMS/PDS activation toward pollutant degradation, Sun and co-workers co-doped a small amount of B (B₂O₃, 0.10 wt%) into NG to enhance the catalytic activity of NG, whereas a further increase in the co-doped B content to 0.25 wt% depressed the catalytic activity.⁶⁶ We believe that the large content of B co-doping is more ready to yield B–N bonds with inferior catalytic activity toward PMS activation. The origin of the B, N-dual-doped graphene was assumed to be the activated sp² carbon and the heteroatom-modified edge defects of graphene. The poorer stability in the recycling and reusability tests indicated that the reaction intermediates located at the edge defects blocked the catalytic sites and lowered the catalytic activity. This finding could also imply that the heteroatom-doped edge defects of graphene might be more critical in activating PMS compared with activated sp² carbon. However, clarification of the specific catalytic sites was not provided, suggesting that more systematic work is needed to unravel the specific active sites among the obscure N configurations and N configurations coupled with other heteroatoms.

Chen et al. conducted a systematic study on B, N-dual-doped graphene as a PMS activator for SAM degradation.¹¹² The methods used to prepare the doped graphene with dual heteroatoms were demonstrated to affect the structure and activity of the resulting co-doped graphene in PMS/PDS activation. For example, if B was co-doped into NG via a two-step process that yielded B, N-dual-doped graphene (BNG), favorable dopant configurations could be formed (e.g., the coupled N-6 with BC3 (B–C–C–C)), as shown in Fig. 14a. The B–C–C–C–(N-6) coupling configuration provided more positive carbon atoms to intimately interact with PMS for the activation of the latter. By contrast, a one-step doping process produced a B, N-dual-doped graphene catalyst with the generation of adverse h-BN composition, where electron-rich N (electron donor) and electron-deficient B (electron acceptor) resulted in the neutralization of the B–N bonds and hence lowered the catalytic performance.¹¹² It is also worth mentioning that B co-doping into NG could turn the non-radical pathways of NG/PMS into both radical and non-radical pathways of BNG/PMS. Radical quenching tests and DFT calculations revealed that single N-Q doping of graphene rendered the strongest adsorption energy, which contributed to the generation of surface-confined PMS-based active species to realize the non-radical pathway in the degradation of SAM. With B and N dual-doping, adsorption energy toward PMS became significantly lowered, implying that the reaction intermediates of PMS/BNG were more easily detached from the BNG surfaces as the radicals (SO₄·⁻ and ·OH).

Fig. 14 (a) Molecular model of the active site on BNG prepared by two-step method (4 refers to bonding configuration of B–C–C–C–N-6).¹¹² (b) Molecular structure model of SNG; (c–g) electrostatic potential mapping from charge density matrix for undoped model graphene (c), SG (d), NG (e), SNG (f), and SSNG (g).¹¹⁰

Most recently, B, N-dual-doped porous carbon was reported as the PMS activation catalyst for tetracycline hydrochloride (TCH) degradation through non-radical pathways based on ¹O₂ and electron transfer from TCH to surface-bound PMS*.¹¹⁵ Such a B, N-dual-doped carbon was fabricated by a two-step method: N-doped carbon was first prepared via the pyrolysis of ZIF-67 at 950 °C for 5 h followed by acid washing to remove Co residuals, and then B co-doping proceeded with boracic acid as a B-containing precursor by thermal annealing at 1100 °C for 2 h. The synergistic coupling effect of electron-deficient B (χ_B = 2.04) and electron-rich N (χ_N = 3.04) is assumed to promote the catalytic performance of carbon matrix (χ_C = 2.55) for PMS activation and TCH degradation. However, the compensation of the charge redistribution of the carbon network caused by N and B likely occurred due to such a coupling. Moreover, the concrete catalytic sites originating from B, N-dual-doping were not elucidated.

N, S dual-doping

To further enhance the activity of N-induced active sites for PMS/PDS activation, the S heteroatom was also considered as the promising heteroatom to work together with N to tune the electronic structure of carbon substrate, synergistically promoting the catalytic activity. Specifically, while N could regulate the charge density of the neighboring carbon atoms, S was capable of modulating their spin densities.³⁷ The chemical inertness of the RGO network can be efficiently broken by S and N dual-doping, and the resulting S, N dual-doped RGO (SNG) was demonstrated to be superior to GO, RGO, S-doped RGO (SG), and NG.¹¹⁰ The first-order apparent rate constant was calculated to be 0.043 ± 0.002 min⁻¹ for the AOP using SNG as the PMS activation catalyst, 4.5, 19.7, 22.8, 86.6-fold larger than that over NG, SG, RGO, and GO, respectively. The similar electronegativity between S and C (2.58 vs. 2.55) enabled the S doping to decrease the energy difference of carbocatalysts between the lowest unoccupied molecular orbital (LUMO) and highest occupied molecular orbital (HOMO).¹¹⁶ When S was doped along the zigzag edges of carbocatalysts (e.g., graphene-like materials), a relatively high spin and charge densities can be imparted to adjacent C atoms, yielding more active sites for carbocatalytic reactions.¹¹⁶ Simultaneously, N with a higher electronegativity (3.04) versus C (2.55) can facilitate electron attraction from adjacent C, thus allowing the C to bear a high charge density. Therefore, proper dual-doping of S and N can profoundly activate the carbocatalysts to produce active sites with higher activity stemming from the redistribution of spin and charge densities. For instance, S doping into carbon nitride was reported to enhance the positive charge density of the C (adjacent to N) further and hence the catalytic and photocatalytic activity for PMS activation toward bisphenol A degradation.¹¹⁷ The S, N dual-doping of wood-shavings-derived biochar resulted in a combination of N-6 and thiophenic S as the active sites for PMS activation to degrade methylene blue.¹⁰¹

The atomic radius of S is much larger than C (1.03 Å vs. 0.75 Å), making it difficult to incorporate S into the graphene network. As a result, a relatively low S doping level can be achieved (e.g., 0.69 at% in an SNG catalyst).¹¹⁰ The N doping increased the charge densities of the adjacent carbon (C1, C2, and C3, as marked in the molecular structure of SNG shown in Fig. 14b), while S co-doping further increased the charge density of C2 from 0.31 to 0.48. Significantly breaking the chemical inertness of pure graphene by heteroatom doping-induced electron redistribution can be well reflected by the electrostatic potential mapping shown in Fig. 14c–g. SNG exhibits the most uneven electron distribution over the graphene charge density matrix, rendering its greatest activity in PMS activation for SO₄·⁻ and ·OH radical generation. The authors believed that the positively charged carbon atoms were the active sites to adsorb HSO₅⁻ and cleave the O–O bond for PMS activation to generate radicals. Although an appropriate amount of S co-doping boosted the catalytic activity of NG, excess S doping was proven to be unfavorable for catalytic PMS activation since the charge redistribution system was disrupted. The theoretical calculation revealed the charge densities of C1, C2, and C3 decreased for SNG when excess S was incorporated, which could also be embodied by the electrostatic potential mapping where the positive charge area was significantly reduced when one more S was incorporated into SNG (Fig. 14f and g).

Using almond shells as the starting material for the fabrication of porous carbon support, Anfar and co-workers reported a new approach for the preparation of N, S-dual-doped porous carbon.¹¹⁸ The presence of high-concentration –COCl– and –CCl– groups in the porous carbon matrix was favorable for surface functionalization, rendering high contents of N and S heteroatoms within the porous carbon framework and hence resulting in high activity in PDS activation. Nevertheless, the single-N-doped porous carbon was found to be superior to N, S-dual-doped porous carbon in PDS activation, which might be due to the co-doped S likely interrupting the spin and charge-redistributed network of the N-doped porous carbon.

Additional S heteroatom co-doping can turn the non-radical pathway of NG into a radical (SO₄·⁻ and ·OH) pathway due to the lower adsorption energies of S, N-dual-doped graphene than those of NG.¹¹⁹ Interestingly, the KOH or ZnCl₂ activation of S, N-dual-doped graphene enabled the generation of the carbocatalyst exhibiting an enhanced free-radical (SO₄·⁻, ·OH, and ·O₂⁻) pathway (particularly KOH activation that could increase the heteroatom ratio and intensify the generation of SO₄·⁻ and ·OH). However, this was not the case when using the CO₂ atmosphere as an activator. The CO₂ atmosphere activation resulted in activated S, N-dual-doped graphene exhibiting a primary non-radical ¹O₂ pathway coupled with the O₂·⁻ radical-based pathway. The KOH, ZnCl₂, or CO₂-based activation of S, N-dual-doped graphene significantly increased the specific surface area and enabled the formation of structural defects with delocalized electrons that were presumed to be associated with O₂·⁻ formation. Further, the recombination of O₂·⁻ species was also likely to occur to generate ¹O₂ species.

N, P dual-doping

The doping of the P atom with a large radius compared with C (1.06 vs. 0.75 Å) is a grand challenge to reach a high doped P content. Nevertheless, once the P heteroatom is doped into a carbon matrix, the sp² carbon structure can be largely distorted, inducing the formation of structure defects that might be beneficial for the adsorption and catalytic reactions with PDS/PMS.¹⁰⁰ Considering that single P doping might not be capable of significantly breaking the chemical inertness of the carbon matrix due to the limited doping content, co-doping with P and N is promising for obtaining highly active carbocatalysts in PMS/PDS activation. More importantly, the synergistic impact of N and P on the adjacent C can maximize the local charge density to boost the catalytic activity for PMS/PDS activation toward pollutant degradation. For example, N, P-dual-doped biochar catalyst was reported to be superior to single N- and P-doped biochar and undoped biochar counterparts for PMS activation toward Rhodamine B degradation, with the N-6, N-5, and P–O–C and O–P–C bonds presumed as the active sites.¹⁰⁰ Interestingly, despite investigating the N, P-dual-doped core/shell porous carbon as the PMS catalyst for phenol degradation (see the synthetic diagram in Fig. 15a), N-6 and C=O were reported as the dominant catalytic sites while doped P played an assistant role in inducing the formation of several active sites and further regulating the surface charge density for enhanced electron transfer (see the metal-free carbocatalysis mechanism in Fig. 15b).¹²⁰ The quenching tests and EPR measurements revealed that ¹O₂ was the key ROS while surface-bound SO₄·⁻ radicals were auxiliary ROS. The non-radical pathway based on electron transfer also made some contributions to catalytic purification reactions.¹²⁰

Fig. 15 a and b) Synthesis of N, P-dual-doped core/shell porous carbon as metal-free carbocatalysts for PMS activation toward phenol degradation: (a) schematic illustration of the synthesis of N, P-dual-doped core-shell porous carbon by pyrolyzing ZIF-8 and phytic acid, followed by acid etching; (b) mechanism of PMS activation on the N, P-dual-doped core/shell porous carbon.¹²⁰

N, F dual-doping

F possesses a much higher electronegativity (χ_F = 3.98) than C (χ_C = 2.55), even higher than N (χ_N = 3.04), making it highly effective in breaking the electroneutrality of the sp² carbon network once F is effectively doped into the C lattice. The lone-pair electrons of F can also enhance the conjugation with the π electrons and thus improve the electron transfer rate of the carbocatalyst.¹²¹ The expanded insertion space imparted via F doping enables the creation of defects and disorder within the carbon structure. As a result, more active sites can be produced, along with improved electron transfer.¹²² The combination of N and F can be a promising way to engineer the internal structure of the carbon matrix based on the synergistic effects generated between N and F. Using ZIF-8 as a precursor, Huang and co-workers first prepared N-doped porous carbon via the thermal treatment at 900 °C under N₂ for 8 h, which was subsequently treated by an NH₄F solution at 80 °C for 12 h. As a result, N, F-dual-doped porous carbon with a polyhedral morphology was produced.¹²¹ The further F co-doping of N-doped carbon allowed the formation of more mesopores, along with the tuning of the electronic structure of N-doped carbon to enhance the electron transfer and ¹O₂ generation. Thus, the radical-free pathway based on ¹O₂ and electron transfer occurred over such an F, N-dual-doped porous carbon catalyst.

N, Si dual-doping

Inspired by the finding in a previous report¹²³ that Si could enhance the activity of superoxide dismutase and catalase in plants, Duan et al. dual-doped nanocarbon with N and Si (an environmentally friendly and abundant element on the earth). The Si, N dual-doped configuration shown in Fig. 16a could exclusively activate PMS to produce non-radical ROS, that is ¹O₂, via the recombination of ·O₂⁻ radicals. The redox reactions between the O=Si(δ⁺)=2(N-6) configurations were beneficial for PMS activation to produce ¹O₂. Nevertheless, the produced ¹O₂ exhibited no resistance to some anions (e.g., Cl⁻ and HCO₃⁻), contrary to the traditional ¹O₂ produced by the PMS's self-decomposition. This finding might be ascribed to the competing reactions between background anions and ·O₂⁻ radicals.²⁷

Fig. 16 (a) Proposed catalytic mechanism of the Si-N/C@PMS system.²⁷ (b) Schematic illustration of synthesizing N, S, P-tri-doped carbon shell; (c) mechanism and pathways of PMS activation over the tri-doped carbon.¹²⁵

N, Se dual-doping

N was also coupled with Se to dual-dope biochar with a hierarchically porous structure for PMS activation toward phenol degradation.¹²⁴ Efficient adsorption and catalytic processes were simultaneously realized over such a Se, N-dual-doped biochar catalyst. While N enhanced the electron transfer of the biochar matrix, Se further concentrated the electron density by its lone-pairs electron. As a result, the chemical inertness of the biochar matrix can be significantly broken, resulting in high performance in PMS activation and phenol degradation. The adjacent carbon neighboring N and Se was regarded as the catalytic sites with net positive charges to strongly interact with PMS, resulting in the formation of metastable N, Se-dual-doped carbon/PMS* complex with a high potential to expedite phenol oxidation based on the non-radical pathway dominated by the direct-electron-transfer bridge of N, Se-dual-doped carbon. The good antioxidation properties of Se could prevent the carbocatalyst from oxidation, thus making the carbocatalyst stable and durable.

The most disturbing questions concerning the heteroatom-doped carbon materials as metal-free carbocatalysts lie in how to precisely implant specific active sites into the carbon network with desired configurations, locations, and contents. Existing methods usually rely on pyrolysis, making the doping configuration random (especially for multi-heteroatom doping). Even worse, the nonstoichiometry of carbon substrates causes a more complex situation. Therefore, precise doping with ease is a challenging yet worthwhile research direction to unveil the mystery of the activity origin of carbocatalysts for future rational research and development.

Tri-heteroatom doping of carbon materials

Tri-heteroatom doping of carbonaceous materials is rarely reported likely due to the difficulties in clarifying the structure and active sites. Moreover, tri-doping may result in depressed catalytic activity (e.g., the activity arising from co-doping-induced redistributed charge and spin densities of the carbon network is likely to be lost once a third type of heteroatom is incorporated) compared with single- and dual-heteroatom doping. Nevertheless, a few efforts have been made to tri-dope carbon materials with multiple heteroatoms for PMS/PDS activation.

N, S, P-tri-doping

N, S, and P tri-doping of porous carbon was implemented using ZIF-67 as the sacrificial template and poly(cyclotriphosphazene-co-4,40-sulfonyldiphenol) as a precursor (Fig. 16b). ZIF-67 could also provide abundant N sources for the C shell and avoid excessive P doping by consuming excessive P content via Co–P nanoparticle formation under pyrolysis.¹²⁵ The controlled P doping helped avoid the catalytic activity decline attributed to excessive P doping. The higher electronegativity of N (χ_N = 3.04) and S (χ_S = 2.58) relative to C (χ_C = 2.55) indicates that N and S are p-type dopants. By contrast, the electronegativity of P (χ_P = 2.19) is lower than C (χ_C = 2.55), rendering P as an n-type dopant. If the n-type dopant is at a high concentration, the compensation effect between p-type and n-type dopants can be generated, thus causing the catalytic efficiency to be significantly depressed. Thus, it is important to control the P doping to an appropriate extent in the tri-doped porous carbon. As such, the multiple heteroatoms (i.e., N, S, and P) worked synergistically to break the electroneutrality and chemical inertness of sp² carbon in a pronounced manner. Consequently, the efficient degradation of BPA over the N, S, and P tri-doped carbon in the presence of PMS was achieved through the radical pathway, including SO₄·⁻, ·OH, and O₂·⁻. The reaction mechanism and pathway in the case of the tri-doped carbon/PMS system are provided in Fig. 16c.

Conclusion and perspectives

This paper presents a timely overview of the exploration of metal-free carbocatalysts for the AOPs based on PMS/PDS activation toward pollutant degradation, which are undergoing an intense study worldwide. We mainly focus on the active sites of various carbocatalysts and their catalytic reaction pathways for pollutant degradation in PMS/PDS-AOPs (mainly including radical and non-radical pathways). Most pristine carbonaceous materials are found to be inactive toward PMS/PDS activation, thus requiring effective methods to alter the electronic and physicochemical properties, as well as the textures and morphologies, in order to bring intrinsically active sites. The balance between the active sites and auxiliary sites also needs to be taken into account (e.g., the overabundant active site incorporation is usually accompanied by destroying the sp² carbon network, thus depressing the electron-transfer capability and being unfavorable for efficient PMS/PDS activation).

More attention can be paid to raising the intrinsic activity of each active site. To reach this goal, one can sufficiently exploit different kinds of configurations of carbocatalysts. To better clarify the active site configuration and reaction pathway, researchers are suggested to couple in situ and operando characterization techniques with theoretical computations based on systematical studies on PMS/PDS activation and pollutant degradation processes. Although much progress has been achieved in designing and developing various carbocatalysts for efficient PMS/PDS activation and pollutant mineralization, many critical issues remain, which are provided below together with our perspectives for future rational design and fabrication of high-performance carbocatalysts for large-scale applications.

1) The unsatisfactory stability of carbocatalysts should be resolved

Currently, most carbocatalysts exhibit unsatisfactory stability in PMS/PDS activation,^{75,82,84,86,89,98} which can be attributed to a change in the surface charge and spin densities after the AOP process, the consumption of the active functional groups (e.g., the conversion of active N-Q and N-6 to N-5 and N–O), and the blockage of the active sites by the PMS/PDS and reaction intermediates.^75,84 The design and creation of stable active sites on carbocatalysts that can withstand harsh reaction conditions and prevent the blocking effects from PMS/PDS and reaction intermediates are worthwhile to be considered.

Post-treatments such as thermal annealing, UV irradiation, and solvent elution are reported to remove the contaminants that block the active sites of carbocatalysts and hence regenerate the active sites.^65,94 The direct-electron-transfer-based non-radical pathway for organics oxidation is also believed to be an effective way to promote the durability of the catalysts in AOPs because carbocatalysts are prone to be oxidized and passivated while mineralizing organic pollutants in the ROS pathway (e.g., the active N species of the carbocatalysts can be oxidized to –NO, –NO₂, and –NO₃⁻ species after AOPs, along with the introduction of O groups⁸⁰). On the contrary, Li et al. believed that the radical pathways were more likely to occur in solutions compared with the non-radical pathway taking place on the catalyst surface, especially for the non-radical pathway involving surface-confined metastable complexes.¹¹⁹ Their results provide a hint that the design and development of catalysts for the AOPs based on the radical pathway with the free radicals yielded in the bulk solution might be more appropriate to avoid the direct contact between the radicals and carbocatalysts to enhance durability.¹¹⁹ Alternatively, the use of carbocatalysts in a specific environment can also improve their stabilities. For example, an acidic reaction environment can provide proper H⁺ that can firmly combine with amine groups to stabilize the latter against being oxidized.⁸⁰ The doping of carbon materials with heteroatoms possessing antioxidative properties (e.g., Se and B (ref. 99 and 124)) is also capable of preventing the active doped sites of carbocatalysts from oxidation during AOPs.

Although some tactics have been reported to restore and enhance the activities and stability of carbocatalysts, respectively, the effective methods are still not many, especially for improving the oxidation resistance of carbocatalysts during AOPs. The studies on improving the stability and durability of carbocatalysts are still in their infancy, suggesting that many key problems still remain regarding how to promote the stability of carbocatalysts, especially the stability of their internal active sites.

2) Cautious attention should be paid to trace metal residues in many reported carbocatalysts

Many studies prepared the carbocatalysts involved the metal-based catalysts/activators/precursors, leaving metal residues albeit with trace quantities, which might be the active sites for PMS/PDS activation. Therefore, discreetly monitoring the trace metal residues should be taken into account. For example, the deliberate addition of the given metal species in a trace quantity into the AOPs can be considered to rule out the possible contribution from the metal residue determined by inductively coupled plasma mass spectrometry (ICP-MS).

3) Tri-doping or quaternary-doping of carbon lattice with multiple heteroatoms is a promising way to create active and durable sites

As time goes by, more attempts should be considered other than single-heteroatom doping and dual-heteroatom doping for the development of more powerful metal-free carbocatalysts bearing both high activity and stability. Thus, tri-doping and quaternary-doping with multiple heteroatoms have become the focus of current attention albeit with prohibitive challenges since multiple heteroatoms can not only work synergistically to promote the catalytic site activity and durability but also may adversely affect each other to depress the catalytic activity (such as excessive n-type dopants (e.g., B and P) lowered the catalytic efficiency in p-type dopants (e.g., N and S)-incorporated carbon catalysts due to the compensation effects between n-type and p-type doping). Moreover, it is difficult to disclose the synergistic role of multiple heteroatoms in tri-doped and quaternary-doped carbon. The optimal doping with multiple heteroatoms is highly desirable to maximize the charge and spin density redistribution over the inert carbon network. Apart from the heteroatom doping location, the proper selection of ternary or quaternary heteroatoms to co-dope the carbon lattice is also essential. Considering that it is the easiest to dope N heteroatom into carbonaceous materials due to its similar atomic radius to C and much higher electronegativity than C, its combination with other heteroatoms might make it easier to yield stable and active co-doped configurations. For example, considering that N and S can be selectively incorporated into the edge sites of unzipped CNTs,³⁷ one can further consider embedding other heteroatoms such as F to regulate the charge and spin density of carbon atoms adjacent to N and S further. Meanwhile, the in-plane sp² carbon network should be kept unaffected. As a result, more active sites can be obtained while maintaining electron-transferrable sp² carbon sites, promoting catalytic activity. When high-temperature carbonization is applied to the preparation of carbocatalysts, their structures can be stabilized to yield stable and durable carbocatalysts. Furthermore, suggestions have been provided to address the durability issues of carbocatalysts (please refer to our perspective in the first place: “1) the unsatisfactory stability of carbocatalysts should be resolved”. Nevertheless, research work on tri- and quaternary-doping treatments of carbon substrates is still rare, suggesting that grand challenges remain and more efforts need to be invested in the future.

4) Accurate doping location in the carbon network is challenging yet worthwhile to put effort into

The current doping treatments are generally random based on pyrolysis and cannot accomplish accurate doping at a desired location. Future work is thus necessary to establish the methods for controllable doping at wanted locations in carbon planes and at edges. However, the nonstoichiometry and complex surface chemistry of most carbon-based materials make it an enormous challenge to realize the aforementioned goal. New kinds of precursors can be tested via trial and error to explore a viable way to obtain active carbon materials with clear active sites imparted via dopant engineering. Certainly, the theoretical calculation studies are suggested to be combined with experimental trials and advanced characterization techniques (e.g., in situ and operando characterizations) to facilitate the accurate doping and engineering of carbon materials.

5) The relationship between the specific active sites and the reaction pathway should be clarified

The controversial conclusion regarding the same kinds of active site configurations that result in different reaction pathways necessitates systematic work on similar carbocatalysts under the same AOP conditions. For example, carbon materials' zigzag edges decorated with C=O groups were reported to catalyze PMS/PDS activation to yield free radicals, while non-radicals generated by the same active oxygen groups were also reported to be the dominant species (such as ¹O₂ and active metastable complexes for direct electron transfer). The controversy will hamper the development of the appropriate carbocatalysts for oriented applications. The impact of the type of carbon matrix should be carefully evaluated because distinct carbon matrices possess different chemical bonding microenvironments, textures, and morphologies (e.g., curved graphene vs. planar graphene, and wrinkled graphene vs. smooth graphene). The subtle difference might lead to a giant discrepancy in catalytic performance. Therefore, future work is suggested to unveil the impact of the subtle difference in the structure on the overall catalytic performance of the carbocatalysts with similar structures and surface properties.

6) The critical point of adsorption interactions that determine the surface-confined metastable-complex-based non-radical pathways deserves to be disclosed

It has been widely reported that the strong adsorption interactions between the carbocatalysts and PMS/PDS result in surface-confined metastable PMS*/PDS* complexes for pollutant degradation based on direct-electron-transfer pathways. By contrast, the weak adsorption interactions lead to separating the PDS/PMS intermediates from the carbocatalyst surface into free radicals. However, the critical point of adsorption energies in this regard is neglected, thus suggesting that systematic studies need to be conducted, with the assistance of advanced characterization and theoretical calculations, to determine the critical points of the adsorption energy, at which the transition of the free radical pathway to the surface-confined metastable complex-based direct electron pathway occurs and vice versa.

7) The high cost of some carbon materials should be addressed to bring the carbocatalysts closer to real-world applications

Given that some popular carbon materials, such as graphene, CNTs, and nanodiamonds, are still costly and far away from real-world practical applications, one should put efforts into slashing the material production costs. Specifically, GO is one of the most widely used precursors for the production of carbocatalysts. Although its fabrication starts from low-cost earth-abundant graphite, the harsh oxidation and purification conditions lead to high production costs coupled with unsafety issues. To address this, one should consider the optimized structure of GO-based carbocatalysts with high catalytic activity. For example, the sp²-to-sp³ ratio is closely related to the GO-based carbocatalyst activity. The overabundant sp³ sites induced by enriched carbon edges restrain the catalytic activity since the electron transportation capability imparted via the sp²-hybridized carbon network is weakened.³⁷ Therefore, a balance should be struck between catalytically active sites based on sp³ carbon and electron-transferring sites contributed by sp² carbon. In this regard, unlike the harsh and corrosive oxidation conditions in producing GO, it is probable to mildly oxidize graphite to produce an active carbocatalysts for PMS/PDS activation. The costs generated from such mild oxidation can be reduced, resulting in an overall low production cost. Alternatively, the traditional GO can be replaced by low-cost carbon materials, such as biocarbon (i.e., biochar). Biochar can be produced from renewable biomass wastes, and a “two bird with one stone” effect can be generated. The treatment of biomass wastes can be resolved, and they can meanwhile be turned into treasure for the production of carbocatalysts. Although biochar production usually relies on high-temperature pyrolysis, causing a high energy cost, low-temperature pyrolysis might be a promising way to produce a carbon network enriched with abundant active sites. Besides, the combination with low-cost carbon supports such as graphite plates can remedy the unsatisfactory electron transportation ability of the biochar produced via low-temperature pyrolysis.

Author contributions

Huawen Hu, and Dongchu Chen: conceptualization, writing – original draft, writing – review & editing, formal analysis, project administration, and funding acquisition. Yaoheng Liang, Wenyi Wang, Yinlei Lin, Xuejun Xu, Xiaowen Wang, Kun Wang, and Yuyuan Zhang: resources, and writing – review & editing. Jian Zhen Ou: conceptualization, funding acquisition, formal analysis, project administration, and writing – review & editing.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

The authors thank the financial support from the Key-Area Special Project of the Guangdong Provincial Education Department (2020ZDZX2066), the Innovation Team Project for Colleges and Universities of Guangdong Province (2023KCXTD030), the Key Project of Guangdong Basic and Applied Basic Research Foundation (2020B1515120081), the National Natural Science Foundation of China (51702050), the Guangdong Basic and Applied Basic Research Foundation (2022A1515140104, 2020A1515111082), and the Academic Foundation for Students at Foshan University (xsjj202306kjb07).

Notes and references

1. Q. Fu, H. Yang, J. Yu, N. Li, Y. Tong, S. Wei, Z. Hao, J. Wang and G. Ouyang, Environ. Sci.: Nano, 2022, 9, 1748–1758.
2. H. Peng, W. Xiong, Z. Yang, J. Tong, Y. Xiang, Z. Zhang and Z. Xu, Environ. Sci.: Nano, 2024, 11, 216–228.
3. X. Jiang, W. Yan, Z. Xiong and L. Zhao, Environ. Sci.: Nano, 2022, 9, 2939–2953.
4. Q. Fu, H. Zhong, Y. Hou, J. Yu, H. Yang, N. Li, Y. Tong, S. Wei, J. Wang and G. Ouyang, Sci. Total Environ., 2023, 902, 166121.
5. Y. Liang, F. Yuan, X. Xu, X. Wang, H. Hu and J. Z. Ou, Carbon, 2023, 214, 118354.
6. Y. Liang, X. Xu, F. Yuan, Y. Lin, Y. Xu, Y. Zhang, D. Chen, W. Wang, H. Hu and J. Z. Ou, Carbon, 2023, 205, 40–53.
7. Y. Zhang, J. Zhang, K. Chen, S. Shen, H. Hu, M. Chang, D. Chen, Y. Wu, H. Yuan and Y. Wang, Resour., Conserv. Recycl., 2023, 190, 106821.
8. S. Zeng and E. Kan, Chemosphere, 2022, 306, 135554.
9. Y. Sun, P. Zhou, P. Zhang, S. Meng, C. Zhou, Y. Liu, H. Zhang, Z. Xiong, X. Duan and B. Lai, Chem. Eng. J., 2022, 450, 138423.
10. X. Mao, M. Wang, J. Li, M. Zhang, C. Dong, H. Lei, Y. He, M. Zhang, Z. Ge, R. Shen, H. Han, J. Hu and G. Wu, Appl. Catal., B, 2023, 331, 122698.
11. Z. An, Y. Hu, D. Zhang, H. Zhou, J. Zhan and M. Wu, J. Environ. Chem. Eng., 2022, 10, 107867.
12. S. Midassi, A. Bedoui and N. Bensalah, Chemosphere, 2020, 260, 127558.
13. P. Dong, H. Wang, W. Liu, S. Wang, Y. Wang, J. Zhang, F. Lin, Y. Wang, C. Zhao, X. Duan, S. Wang and H. Sun, J. Hazard. Mater., 2021, 401, 123423.
14. G. Jiang, X. Wang, B. Zhu, J. Gong, F. Liu, Y. Wang and C. Zhao, Carbon, 2022, 190, 32–46.
15. J.-B. Tarkwa, N. Oturan, E. Acayanka, S. Laminsi and M. A. Oturan, Environ. Chem. Lett., 2018, 17, 473–479.
16. Y. De Luna and N. Bensalah, Curr. Opin. Electrochem., 2022, 32, 100900.
17. H. Long, K. Wen, C. Liu, X. Liu and H. Hu, Molecules, 2023, 28, 5798.
18. Y. Zhang, K. Chen, J. Zhang, K. Huang, Y. Liang, H. Hu, X. Xu, D. Chen, M. Chang and Y. Wang, J. Environ. Chem. Eng., 2023, 11, 109358.
19. Y. Xu, Y. Liang, Q. He, R. Xu, D. Chen, X. Xu and H. Hu, Bull. Mater. Sci., 2022, 46, 6.
20. Y. Xu, Y. Liang, Z. Yuai, H. Long, Q. He, K. Guo, Y. Zhang, D. Chen, X. Xu and H. Hu, Ceram. Int., 2022, 48, 24677–24686.
21. X. Wang, D. Chen, M. Zhang and H. Hu, Polymer, 2019, 11, 1860.
22. Y. Xu, Y. Yu, Y. Yang, T. Sun, S. Dong, H. Yang, Y. Liu, X. Fan and C. Song, Sep. Purif. Technol., 2021, 276, 119328.
23. J. Ye, J. Dai, D. Yang, C. Li, Y. Yan and Y. Wang, Chem. Eng. J., 2021, 418, 129383.
24. L. Lin, Y. Wen, L. Li, Y. Tan, P. Yang, Y. Liang, Y. Xu, H. Hu and Y. Xu, Nanomaterials, 2022, 12, 3339.
25. H. Hu, J. H. Xin, H. Hu, X. Wang and Y. Kong, Appl. Catal., A, 2015, 492, 1–9.
26. H. Hu, J. H. Xin, H. Hu, X. Wang, D. Miao and Y. Liu, J. Mater. Chem. A, 2015, 3, 11157–11182.
27. W. Duan, J. He, Z. Wei, Z. Dai and C. Feng, Environ. Sci.: Nano, 2020, 7, 2982–2994.
28. Z. Yang, X. Duan, J. Wang, Y. Li, X. Fan, F. Zhang, G. Zhang and W. Peng, ACS Sustainable Chem. Eng., 2020, 8, 4236–4243.
29. X. Duan, H. Sun and S. Wang, Acc. Chem. Res., 2018, 51, 678–687.
30. K. Tian, L. Hu, L. Li, Q. Zheng, Y. Xin and G. Zhang, Chin. Chem. Lett., 2022, 33, 4461–4477.
31. A. Hassani, J. Scaria, F. Ghanbari and P. V. Nidheesh, Environ. Res., 2023, 217, 114789.
32. J. Ji, X. Yuan, Y. Zhao, L. Jiang and H. Wang, J. Hazard. Mater., 2022, 430, 128375.
33. B. Wang and Y. Wang, Sci. Total Environ., 2022, 831, 154906.
34. J. Peng, Z. Wang, S. Wang, J. Liu, Y. Zhang, B. Wang, Z. Gong, M. Wang, H. Dong, J. Shi, H. Liu, G. Yan, G. Liu, S. Gao and Z. Cao, Chem. Eng. J., 2021, 409, 128176.
35. X. Pan, J. Chen, N. Wu, Y. Qi, X. Xu, J. Ge, X. Wang, C. Li, R. Qu, V. K. Sharma and Z. Wang, Water Res., 2018, 143, 176–187.
36. H. Liu, P. Sun, M. Feng, H. Liu, S. Yang, L. Wang and Z. Wang, Appl. Catal., B, 2016, 187, 1–10.
37. Q. Yang, Y. Chen, X. Duan, S. Zhou, Y. Niu, H. Sun, L. Zhi and S. Wang, Appl. Catal., B, 2020, 276, 119146.
38. J. Li, S. Zhao, L. Zhang, S. P. Jiang, S. Z. Yang, S. Wang, H. Sun, B. Johannessen and S. Liu, Small, 2021, 17, e2004579.
39. J. Li, F. Li, Q. Yang, S. Wang, H. Sun, Q. Yang, J. Tang and S. Liu, Carbon, 2021, 182, 715–724.
40. M. Kohantorabi, G. Moussavi and S. Giannakis, Chem. Eng. J., 2021, 411, 127957.
41. X. Duan, W. Li, Z. Ao, J. Kang, W. Tian, H. Zhang, S.-H. Ho, H. Sun and S. Wang, J. Mater. Chem. A, 2019, 7, 23904–23913.
42. D. Jia, K. Hanna, G. Mailhot and M. Brigante, Molecules, 2021, 26, 5748.
43. X. Zhou, A. Jawad, M. Luo, C. Luo, T. Zhang, H. Wang, J. Wang, S. Wang, Z. Chen and Z. Chen, Appl. Catal., B, 2021, 286, 119914.
44. W. Sang, Z. Li, M. Huang, X. Wu, D. Li, L. Mei and J. Cui, Chem. Eng. J., 2020, 383, 123057.
45. X. Zhou, C. Luo, M. Luo, Q. Wang, J. Wang, Z. Liao, Z. Chen and Z. Chen, Chem. Eng. J., 2020, 381, 122587.
46. Z. Wang, W. Qiu, S. Pang, Y. Gao, Y. Zhou, Y. Cao and J. Jiang, Water Res., 2020, 172, 115504.
47. S. Mu, X. Chen, Y. Luo and J. Zhang, Process Saf. Environ. Prot., 2022, 168, 883–891.
48. Z. Bai, S. Gao, H. Yu, X. Liu and J. Tian, Water Res., 2022, 222, 118928.
49. J. Li, L. Yang, B. Lai, C. Liu, Y. He, G. Yao and N. Li, Chem. Eng. J., 2021, 414, 128674.
50. X. Zhang, M. Feng, R. Qu, H. Liu, L. Wang and Z. Wang, Chem. Eng. J., 2016, 301, 1–11.
51. H.-Q. Zhao, J.-S. Song, P. Lu and Y. Mu, Chem. Eng. J., 2023, 456, 141045.
52. M. Li, J. Song, W. Han, K. L. Yeung, S. Zhou and C.-H. Mo, npj Clean Water, 2023, 6, 37.
53. S. Liu, C. Lai, B. Li, X. Liu, X. Zhou, C. Zhang, L. Qin, L. Li, M. Zhang, H. Yi, Y. Fu, H. Yan and L. Chen, Chem. Eng. J., 2022, 427, 131655.
54. J. Yu, H. Feng, L. Tang, Y. Pang, G. Zeng, Y. Lu, H. Dong, J. Wang, Y. Liu, C. Feng, J. Wang, B. Peng and S. Ye, Prog. Mater. Sci., 2020, 111, 100654.
55. X.-F. Huang, J.-X. Yuan, X. Zhao, S.-Q. Li, L.-H. Yang, W.-Y. Li, K.-M. Peng, C. Cai and J. Liu, Chem. Eng. J., 2023, 465, 142801.
56. P.-F. Xiao, L. An and D.-D. Wu, New Carbon Mater., 2020, 35, 667–683.
57. Y. Gao, Q. Wang, G. Ji and A. Li, Chem. Eng. J., 2022, 429, 132387.
58. J. Wei, L. Yin, R. Qu, X. Pan and Z. Wang, Environ. Sci.: Water Res. Technol., 2022, 8, 116–126.
59. H. Hu, W. Wen and J. Z. Ou, Environ. Sci.: Nano, 2022, 9, 3226–3276.
60. H. Hu, J. Z. Ou, X. Xu, Y. Lin, Y. Zhang, H. Zhao, D. Chen, M. He, Y. Huang and L. Deng, Chem. Eng. J., 2021, 425, 130587.
61. H. Sun, S. Liu, G. Zhou, H. M. Ang, M. O. Tadé and S. Wang, ACS Appl. Mater. Interfaces, 2012, 4, 5466–5471.
62. T. Teng, X. Wu, Y. Lu, F. Yu, C. Jia, L. Sun, L. Lin, Z. He, J. Gao, S. Zhang and X. Zhu, Adv. Mater. Technol., 2023, 8, 2300236.
63. W. Tian, S. Chen, H. Zhang, H. Wang and S. Wang, Curr. Opin. Chem. Eng., 2022, 37, 100838.
64. X. Duan, Z. Ao, H. Zhang, M. Saunders, H. Sun, Z. Shao and S. Wang, Appl. Catal., B, 2018, 222, 176–181.
65. X. Cheng, H. Guo, Y. Zhang, G. V. Korshin and B. Yang, Water Res., 2019, 157, 406–414.
66. H. Sun, Y. Wang, S. Liu, L. Ge, L. Wang, Z. Zhu and S. Wang, Chem. Commun., 2013, 49, 9914.
67. X. Duan, Z. Ao, H. Sun, L. Zhou, G. Wang and S. Wang, Chem. Commun., 2015, 51, 15249–15252.
68. H. Hu, H. Quan, B. Zhong, Z. Li, Y. Huang, X. Wang, M. Zhang and D. Chen, ACS Appl. Nano Mater., 2018, 1, 6502–6513.
69. H. Hu, X. Wang, K. I. Lee, K. Ma, H. Hu and J. H. Xin, Sci. Rep., 2016, 6, 31815.
70. H. Hu, J. Xin, H. Hu, X. Wang and X. Lu, Molecules, 2014, 19, 7459–7479.
71. H. Hu, J. H. Xin and H. Hu, J. Mater. Chem. A, 2014, 2, 11319–11333.
72. H. Hu, C. C. K. Allan, J. Li, Y. Kong, X. Wang, J. H. Xin and H. Hu, Nano Res., 2014, 7, 418–433.
73. H. Hu, X. Wang, D. Miao, Y. Wang, C. Lai, Y. Guo, W. Wang, J. H. Xin and H. Hu, Chem. Commun., 2015, 51, 16699–16702.
74. P. Shao, J. Tian, F. Yang, X. Duan, S. Gao, W. Shi, X. Luo, F. Cui, S. Luo and S. Wang, Adv. Funct. Mater., 2018, 28, 1705295.
75. X. Cheng, H. Guo, Y. Zhang, X. Wu and Y. Liu, Water Res., 2017, 113, 80–88.
76. W. Ren, L. Xiong, G. Nie, H. Zhang, X. Duan and S. Wang, Environ. Sci. Technol., 2019, 54, 1267–1275.
77. W. Peng, S. Liu, H. Sun, Y. Yao, L. Zhi and S. Wang, J. Mater. Chem. A, 2013, 1, 5854.
78. S. He, X. Zhang, X. Fang, X. Zhang, S. Zhong, G. Zhang, W. Xu, T. Gu and L. Shi, J. Taiwan Inst. Chem. Eng., 2023, 152, 105196.
79. C. Fu, G. Sun, C. Wang, B. Wei, G. Ran and Q. Song, Environ. Res., 2022, 206, 112242.
80. X. Yongsheng, L. Xintong, H. Hongwei, S. Yuexiao, X. Qing and P. Wenchao, J. Hazard. Mater., 2021, 401, 123742.
81. H. Yin, F. Yao, Z. Pi, Y. Zhong, L. He, K. Hou, J. Fu, S. Chen, Z. Tao, D. Wang, X. Li and Q. Yang, J. Colloid Interface Sci., 2021, 586, 551–562.
82. X. Chen, W.-D. Oh, Z.-T. Hu, Y.-M. Sun, R. D. Webster, S.-Z. Li and T.-T. Lim, Appl. Catal., B, 2018, 225, 243–257.
83. X. Duan, S. Indrawirawan, J. Kang, W. Tian, H. Zhang, X. Duan, X. Zhou, H. Sun and S. Wang, Catal. Today, 2020, 355, 319–324.
84. D. Wu, W. Song, L. Chen, X. Duan, Q. Xia, X. Fan, Y. Li, F. Zhang, W. Peng and S. Wang, J. Hazard. Mater., 2020, 381, 121010.
85. M. Zhang, R. Luo, C. Wang, W. Zhang, X. Yan, X. Sun, L. Wang and J. Li, J. Mater. Chem. A, 2019, 7, 12547–12555.
86. P. Hu, H. Su, Z. Chen, C. Yu, Q. Li, B. Zhou, P. J. J. Alvarez and M. Long, Environ. Sci. Technol., 2017, 51, 11288–11296.
87. Y. Shi, D. Feng, S. Ahmad, L. Liu and J. Tang, Chem. Eng. J., 2023, 454, 140244.
88. L. Wang, D. Luo, J. Yang and C. Wang, J. Cleaner Prod., 2022, 375, 134118.
89. Y. Liu, W. Miao, X. Fang, Y. Tang, D. Wu and S. Mao, Chem. Eng. J., 2020, 380, 122584.
90. S. Cai, X. Zuo, H. Zhao, S. Yang, R. Chen, L. Chen, R. Zhang, D. Ding and T. Cai, J. Mater. Chem. A, 2022, 10, 9171–9183.
91. X. Duan, Z. Ao, H. Sun, S. Indrawirawan, Y. Wang, J. Kang, F. Liang, Z. H. Zhu and S. Wang, ACS Appl. Mater. Interfaces, 2015, 7, 4169–4178.
92. X. Li, X. Yan, X. Hu, R. Feng, M. Zhou and L. Wang, J. Colloid Interface Sci., 2021, 591, 184–192.
93. M. Zhang, C. Han, W. Chen, W. Luo, Y. Cao, G. Qian, X. Zhou, X. Duan, S. Wang and X. Duan, Green Energy Environ., 2020, 5, 444–452.
94. S. Zhu, X. Huang, F. Ma, L. Wang, X. Duan and S. Wang, Environ. Sci. Technol., 2018, 52, 8649–8658.
95. L. Liang, X. Yue, Y. Wang, Y. Wu, S. Dong, J. Feng, Y. Pan and J. Sun, J. Colloid Interface Sci., 2021, 604, 660–669.
96. Y. Gao, Z. Chen, Y. Zhu, T. Li and C. Hu, Environ. Sci. Technol., 2019, 54, 1232–1241.
97. C. Wang, J. Kang, H. Sun, H. M. Ang, M. O. Tadé and S. Wang, Carbon, 2016, 102, 279–287.
98. M. Pedrosa, G. Drazic, P. B. Tavares, J. L. Figueiredo and A. M. T. Silva, Chem. Eng. J., 2019, 369, 223–232.
99. X. Li, D. Liang, C. Wang and Y. Li, Chemosphere, 2021, 275, 130058.
100. F. El Ouadrhiri, R. H. Althomali, A. Adachi, E. Abdu Musad Saleh, K. Husain, A. Lhassani, I. Hassan, M. Mostafa Moharam, A. F. Kassem, M. Chaouch, M. Ali Oturan and A. Lahkimi, J. Saudi Chem. Soc., 2023, 27, 101648.
101. W.-D. Oh, J. R. J. Zaeni, G. Lisak, K.-Y. A. Lin, K.-H. Leong and Z.-Y. Choong, Chemosphere, 2021, 277, 130313.
102. L. Liang, X. Yue, S. Dong, J. Feng, J. Sun, Y. Pan and M. Zhou, Sep. Purif. Technol., 2021, 272, 118923.
103. C. Wang, Y. Liu, F. Han, Y. Han, T. Liu, H. Ren and X. Han, Phys. Chem. Chem. Phys., 2023, 25, 13716–13727.
104. Y. Zhang, H. Pan, M. Murugananthan, P. Sun, D. D. Dionysiou, K. Zhang, A. Khan and Y. Zhang, Carbon, 2020, 156, 399–409.
105. Y. Guo, Z. Zeng, Y. Li, Z. Huang and Y. Cui, Catal. Today, 2018, 307, 12–19.
106. Y. Guo, Z. Zeng, Y. Liu, Z. Huang, Y. Cui and J. Yang, J. Mater. Chem. A, 2018, 6, 4055–4067.
107. X. Liu, Y. Chen, Y. Yao, Q. Bai and Z. Wu, Catal. Sci. Technol., 2018, 8, 5482–5489.
108. M. A. Patel, F. Luo, M. R. Khoshi, E. Rabie, Q. Zhang, C. R. Flach, R. Mendelsohn, E. Garfunkel, M. Szostak and H. He, ACS Nano, 2016, 10, 2305–2315.
109. X. Liu, L. Rao, Y. Yao and H. Chen, Chemosphere, 2020, 246, 125783.
110. X. Duan, K. O'Donnell, H. Sun, Y. Wang and S. Wang, Small, 2015, 11, 3036–3044.
111. X. Duan, S. Indrawirawan, H. Sun and S. Wang, Catal. Today, 2015, 249, 184–191.
112. X. Chen, X. Duan, W.-D. Oh, P.-H. Zhang, C.-T. Guan, Y.-A. Zhu and T.-T. Lim, Appl. Catal., B, 2019, 253, 419–432.
113. W. Ma, N. Wang, Y. Du, P. Xu, B. Sun, L. Zhang and K.-Y. A. Lin, ACS Sustainable Chem. Eng., 2018, 7, 2718–2727.
114. S. Zuo, S. Zhu, J. Wang, W. Liu and J. Wang, Appl. Catal., B, 2022, 301, 120783.
115. J. Xie, L. Chen, X. Luo, L. Huang, S. Li and X. Gong, Sep. Purif. Technol., 2022, 281, 119887.
116. K.-Y. A. Lin, Z.-Y. Zhang and T. Wi-Afedzi, J. Water Process Eng., 2018, 24, 83–89.
117. K.-Y. A. Lin and Z.-Y. Zhang, Chem. Eng. J., 2017, 313, 1320–1327.
118. Z. Anfar, A. A. El Fakir, M. Zbair, Z. Hafidi, A. Amedlous, M. Majdoub, S. Farsad, A. Amjlef, A. Jada and N. El Alem, Chem. Eng. J., 2021, 405, 126660.
119. X. Li, J. Wang, X. Duan, Y. Li, X. Fan, G. Zhang, F. Zhang and W. Peng, ACS Catal., 2021, 11, 4848–4861.
120. X. Li, L. Ye, Z. Ye, S. Xie, Y. Qiu, F. Liao, C. Lin and M. Liu, Sep. Purif. Technol., 2021, 276, 119286.
121. L. Huang, X. Gong, J. Xie, L. Zhang and X. Luo, Chem. Eng. Sci., 2023, 280, 118979.
122. Q. He, J. Jiang, J. Zhu, Z. Pan, C. Li, M. Yu, J. Key and P. K. Shen, J. Power Sources, 2020, 448, 227568.
123. Q. Shi, Z. Bao, Z. Zhu, Y. He, Q. Qian and J. Yu, Phytochemistry, 2005, 66, 1551–1559.
124. K. Zhang, X. Min, T. Zhang, M. Xie, M. Si, L. Chai and Y. Shi, J. Hazard. Mater., 2021, 413, 125294.
125. W. Ma, N. Wang, T. Tong, L. Zhang, K.-Y. A. Lin, X. Han and Y. Du, Carbon, 2018, 137, 291–303.

---