In recent decades, the overload discharge of refractory organics leads to water pollution, which would be threatened the sustainable development of society [1-3]. Meanwhile, the ever-growing demand on clean water has given rise to tremendous effort for developing environmentally friendly and efficient technologies for wastewater treatment. Singlet oxygen (¹O₂) is unique non-radical derivative of oxygen [4], possessing unoccupied π^* orbital and exhibiting high selectivity towards to electron-rich organic pollutants such as medicine [5, 6] and pathogenic microorganism [7].

Greater recognition of the vital significance of ¹O₂ has motivated research for more effective ¹O₂ production [4]. In general, ¹O₂ is produced via the energy transfer from a series of photosensitizers including organic dyes, fullerene derivates, porphyrins, and semiconductor quantum dots to the triplet ground state of O₂ [8]. Instead of the energy transfer, ¹O₂ can be generated via the oxidation of O₂^•- through removing the electrons in the antibonding π^* orbitals of O₂^•- in electrocatalytic and/or photocatalytic process [4, 9]. Nevertheless, most of the approaches suffer from such drawbacks as low quantum yield, poor selectivity, non-recyclability and biotoxicity, which severely restrict their potential applications.

Photoelectrocatalytic (PE) activation of O₂ has emerged as a promising and green alternative way to efficiently produce reactive oxygen species (ROS), because it can overcome the energy barrier with a high solar energy conversion efficiency [10]. In this process, O₂ is firstly reduced to O₂^•- through 1-electron reduction pathway, and then the generated O₂^•- would be simultaneously oxidized to ¹O₂ by photoinduced hole (h_vb⁺) [4, 9, 11]. The production efficiency of ¹O₂ in PE system is mainly determined by the selectivity of oxygen reduction and the oxidation ability of h_vb⁺. Graphitic carbon nitride (g-C₃N₄, denoted as CN) has attracted attention due to their special optical features and environmental friendliness [12-15]. It is a promising metal-free photocatalyst that can be widely used in environmental fields [16, 17]. Nonetheless, the photocatalytic efficiency of pure CN was relatively low. Although transition metal doped CN with variable chemical states and unoccupied orbitals appear to be more efficient to improve photocatalytic performance [18, 19], the problem of metallic ion leaching remains a significant concern for practical applications [1]. In contrast, doping non-metal elements, especially for self-doping of C atom, can not only expand the visible light absorption of CN, but also facilitate the charge transfer by forming delocalized π bonds [13, 20, 21]. However, most of the researches so far mainly focused on the direct oxidation of CN [21, 22], few examples have paid attention to non-radical ¹O₂ generated from CN. Besides, the photoelectrocatalytic mechanism on how to selectively generate ¹O₂ by designing CN-derived electrodes remain elusive.

Herein, we turn our research interest to the rational design of CN-derived electrodes (CBCN), integrated with photocatalytic and electrocatalytic process, for selectively activating oxygen to ¹O₂. The CBCN was modified by introducing delocalized π bonds and cyano group simultaneously into CN framework through thermal polymerization method. As expected, the CBCN/PE system exhibits 100% removal efficiency for electron-rich pollutants such as bisphenol A (BPA) and acetaminophen (ACT). Moreover, CBCN/PE system with reliability and wide pH range toward the degradation process exhibits great practical application prospects in wastewater treatment field.

As we known, creating large delocalized π bonds in the framework of CN and introducing cyano group could improve the photocatalytic ability of pure CN [13, 23, 24]. An environmental begin and facile strategy of adjusting polymerization of CN precursors was proposed to fabricate carbon bridged carbon nitride (CBCN) containing π bonds and cyano functional group. As expected, the delocalized π bonds in CBCN was formed by substituting bridge N atom with C atom, and the cyano groups originated from the incomplete condensation of dicyandiamide (Fig. 1a). The crystalline structure of CN and CBCN were probed by X-ray diffraction (XRD). Both CN and CBCN have two peaks at 27.5° and 13.0° (Fig. 1b), corresponding to the interlayer stacking (002) and in-plane ordering of heptazine units (100) [25, 26]. The layer stacking peak of CBCN was positively shifted 0.3° due to the increased stacking density of conjugated layers [13]. Besides, CBCN exhibited typical layered structure of g-C₃N₄ with more pores on its surface, possibly due to formation of NH₃ and CO₂ during the incomplete thermal decomposition of ammonium citrate (Fig. S1 in Supporting information) [21]. The specific surface area of CBCN was 15 m²/g, larger than that of CN (8 m²/g). Adsorption isotherms (Fig. S2 in Supporting information) and pore size distribution (Fig. S3 in Supporting information) indicated that CN and CBCN mainly contained mesopores. The pore volume of CBCN (0.090 cm³/g) compared to CN (0.047 cm³/g) also demonstrated the enhanced porosity of CBCN. Typical Fourier transform infrared (FTIR) bands of cyano groups (-C≡N) centered at 2177 cm^-1 was only obtained for CBCN (Fig. S4 in Supporting information) [24]. Further information about the structure of the catalysts was explored by X-ray photoelectron spectroscopy (XPS) and X-ray absorption near edge structure (XANES) measurements. As exhibited in Fig. 1c, the C 1s spectrum can be deconvolved into three peaks with binding energies of 288.1 eV, 286.4 eV and 284.9 eV ascribed to C1 (N=C-N₂ in the framework of CN), C2 (C-NH_x on the edges of heptazine units) and C3 (C-C/C=C) [27, 28]. The peak area ratio for C3 and C1 was calculated to be 0.19 and 0.35 for CN and CBCN, respectively. Moreover, XPS N 1s spectrum of CBCN was separated into four peaks with binding energies of 398.6 eV, 399.6 eV, 401.1 eV and 404.6 eV ascribed to N1 in the heptazine units (sp² structure of C-N=C), N2 (sp³ bridging N of N-C3), N3 (–NH_x) and N4 (π-π^* excitations) (Fig. 1d) [24]. The peak area ratios of N2 and N1 was 0.54 and 0.42 respectively for CN and CBCN. These conclusions confirmed the replacement of bridging N with C after carbon doped [13, 20]. In addition, comparing to CN, N1, N2 peaks of CBCN shift to lower binding energy, which due to the existence cyano groups whose N 1s binding energy are intermediate between those of N1 and N2 [21, 23, 24]. These observations were further verified by the normalized N 1s and C 1s K edge XANES spectra. As revealed in Fig. 1e, a decrease of N=C-N₂ as well as an increase of C-C/C=C and C-NH_x from CN to CBCN was observed, which was consistent with the C 1s XPS results. Besides, as shown in Fig. 1f, the C-N=C and N-C3 peaks of CBCN slightly shift to lower a binding energy and N-C3 showed weaker peak than CN.

The valence band maximum (VBM) and conduction band minimum (CBM) of CN and CBCN could be calculated from bandgap energies and Mott–Schottky curves (Fig. S5 in Supporting information). As shown in Fig. 2a, the conduction band potential of CBCN (-0.3 V vs. SHE) suggested the photogenerated electron (e_cb^-) could reduce O₂ to produce O₂^•- (-0.16 V vs. SHE) [29]. Moreover, CBCN exhibited 0.45 V positive valence band potential than CN, indicating the stronger oxidation ability of photogenerated hole (h_vb⁺) that can oxidize O₂^•- to produce ¹O₂. In order to verify the catalytic mechanism for selectively generating ¹O₂ through photoelectrochemical reduction of O₂, electron paramagnetic resonance (EPR) measurement was carried out for investigating the formation of active oxygen species (ROS) such as ^•OH, O₂^•- and ¹O₂. Obviously, as shown in Fig. 2b, the main ROS with CBCN in photoelectrocatalytic (PE) process was ¹O₂. The EPR intensity of DMPO-^•OH was relatively weak compared to TEMP-¹O₂. Almost no O₂^•- species were detected in PE process both for CBCN and CN, possibly due to that O₂^•- was simultaneously transferred to ¹O₂ [30]. Besides, the EPR intensity of TEMP-¹O₂ with CBCN was much stronger than CN, indicating the enhanced oxidation ability of h_vb⁺ by the substitution of N atoms with bridged C atoms and cyano group. The synergistic effect of photocatalytic- and electrocatalytic- reduction of oxygen was summarized in Figs. 2c and d. Interesting, the EPR intensity of TEMP-¹O₂ in PE was almost 6–9 times higher than in sole electrocatalytic (E) and photocatalytic process (P) with CBCN. In PE system, O₂ was firstly reduced to O₂^•- by electrons (e^-) from external circuit in E system and/or e_cb^- in P system through single electron pathway, and then h_vb⁺ oxidized O₂^•- to form ¹O₂ (Eq. 1) [4, 9]. That is why the formation of¹O₂ is more efficient in PE system. To further confirm the generation pathway of ¹O₂ in PE process, the ethylenediaminetetraacetic acid disodium salt (EDTA-2Na) and potassium dichromate (K₂Cr₂O₇) were respectively used as scavengers of h_vb⁺ and e_cb^- (Fig. 2c). The intensity of TEMP-¹O₂ with CBCN almost remained the same after the addition of K₂Cr₂O₇, indicating that e_cb^- had little effect in the formation of ¹O₂. That is to say, the formation of O₂^•- in PE process was mainly contributed by e^- from external circuit in E system. However, the ¹O₂ was obviously inhibited once using EDTA-2Na to capture h_vb⁺. Note that, comparing to CBCN, h _vb⁺ had a less impact on the generation of ¹O₂ for CN (Fig. S7 in Supporting information). This observation indicated that the delocalized π bonds and cyano groups in CBCN greatly favor the formation of ¹O₂. In the sole E process, the generation of ¹O₂ was originated from the oxidation of O₂^•- with surface ^•OH (^•OH_sur), which acted as surface trapped holes (Eq. 2) [9]. The low generation efficiency of ^•OH in E system resulted in the weak intensity of TEMP-¹O₂.

The intensity of generated ^•OH in PE process was relatively stronger than in sole E and P systems for CBCN. The intensity of DMPO-^•OH was almost vanished in the presence of K₂Cr₂O₇ in PE system, while remained the same with the addition of EDTA-2Na as the scavenger of h_vb⁺. This phenomenon indicated that e_cb^- was active sites for reducing electrochemical generated H₂O₂ to generate ^•OH radicals in PE process. In fact, H₂O₂ was formed via 2-electron reduction pathway of oxygen reduction reaction (ORR). In P system, the adsorbed O₂ on the surface of the catalyst was reduced by e_cb^- to generate O₂^•-, then, O₂^•- react with H⁺ and e_cb^- to generate ^•OH (Eq. 3) [9]. Different with PE and P systems, in E process, almost no ^•OH was obtained, confirming that there are no active sites in CBCN for decomposing H₂O₂.

The delocalized π bonds and cyano group can change the electronic structure and modify the band gap of CN, thus increasing the light absorption ability and charge separation efficiency. As shown in Fig. 3a, CBCN would adsorb more visible light than CN to generate more photogenerated electron-hole pairs. In addition, 70 nm redshift in the adsorption edge was obtained for CBCN. According to the Tauc plots, as shown in Fig. 3b, the band gaps of CBCN and CN can be obtained by the linear extrapolation of a straight line with the baseline [23]. The band gaps of CBCN and CN were determined to be 2.30 and 2.65 eV, respectively. The charge separation efficiency was further clarified by photoluminescence (PL) spectra measurements. As shown in Fig. 3c, the PL intensity of CBCN was much lower than CN, indicating the improved separation of photoexcited charge carriers [31]. The observation indicated that cyano groups can introduce intraband states into the band gap of CN and carbon species in bridged carbon can act as electron sink, inhibiting the recombination of photogenerated electron-hole pairs [21].

The internal electron transport capability was characterized by electrochemical impedance spectroscopy. Fig. 3d presented that the Nyquist plots diameter of CBCN was much smaller than that of CN, indicating that the conductivity and internal electron transmission capacity were enhanced due to the formation of delocalized big π bonds [13]. Fig. 3e demonstrated that the photocurrent response of CBCN was 3 times higher than CN under the same conditions, suggesting the fabricated CBCN with delocalized π bonds and cyano group exhibited excellent photocatalysis performance and can be a promising electrophotocatalyst for wastewater treatment. The selectivity and activity of oxygen reduction reaction were investigated by rotating ring-disk electrode (RRDE) measurement. The linear sweep voltammetry (LSV) results suggested that the improved ORR activity was obtained for CBCN with 300 mV positive shift of the onset potential, but decreased the 2e^- ORR selectivity with a declined ring current (Fig. 3f). The H₂O₂ selectivity (%H₂O₂) for CBCN and CN at -0.21 V (vs. SHE) were respectively 55% and 66%. The slight decrease in %H₂O₂ via 2e^- ORR with CBCN was possibly due to the decreased concentration of pyrrolic-N during the substation of C atom for bridge N in CN [32].

As investigated above, the selective formation of ¹O₂ via photoelectrocatalytic oxygen reduction can be successfully achieved with CBCN in this work. Moreover, ¹O₂ is a more selective oxidant (1.1 V vs. SHE) than ^•OH, which can preferentially oxidize electron-rich organic pollutants [33]. Therefore, we further evaluated the role of ¹O₂ by estimating the degradation efficiency of BPA, ACT and 3-chlorophenol (3-CP) with different charge density. According to Fig. 4a, the apparent rate constants (k_obs) of removing BPA, ACT and 3-CP could be fitted with pseudo-first-order kinetic model. And the value of k_obs for BPA was 0.140 min^-1, which is 7.8 and 2.2 times higher than 3-CP (0.018 min^-1) and ACT (0.063 min^-1), respectively. Compared to the organics with electron-donating groups such as BPA, ACT and 3-CP, the degradation rate of nitrobenzene (NB) without electron-donating was obviously lower (0.0057 min^-1) (Fig. S9 in Supporting information). The generated ¹O₂ in CBCN/PE system has strong electrophilic ability, so BPA with highest charge density exhibited significant degradation, moderate and lowest degradation was respectively obtained for ACT and 3-CP. To identify the role of different active spices (¹O₂, ^•OH, h_vb⁺) on the removal of BPA, ACT and 3-CP, quenching experiments were designed. L-histidine and isopropanol (IPA) were used as scavengers to capture ¹O₂ and ^•OH respectively. As exhibited in Fig. 4b, the degradation efficiency of BPA was decreased from 100% to 90.5% and 9.4% after the addition of IPA and L-histidine, respectively. The addition of IPA and L-histidine resulted the ACT removal efficiency decreased to 68.1% and 29.7%. Note that, the degradation efficiency of 3-CP was greatly decreased to 15.5% in the presence of IPA. Besides, the degradation of efficiency of BPA, ACT and 3-CP was respectively decreased to 17.2%, 38.5% and 70.8% after the addition of EDTA-2Na as h_vb⁺ scavenger. This phenomenon reveals that ¹O₂ and h_vb⁺ played significant role for BPA and ACT removal, while ^•OH was relatively responsible for 3-CP removal. In addition, the PE degradation process under N₂ atmosphere was carried out to exclude the oxidation contribution of ROS. As shown in Fig. 4b, the degradation efficiency of BPA, 3-CP, ACT was only 12.6%, 11.3% and 8.4%. In order to further clarify the contribution of h_vb⁺, the electrosorption of BPA (7.1%), ACT (6.7%) and 3-CP (5.6%) with electrode was investigated. All the observation revealed that the contribution of h_vb⁺ on the degradation of organic pollutants can be ignored. The role of h_vb⁺ was acted for the generation of ¹O₂.

As we known, actual wastewater had variable pH values, and thus, it is necessary to determine the effect of pH values on the degradation efficiencies. As shown in Fig. 4c, obviously, the BPA removal rate constant remained as high as 0.14–0.13 min^-1 in the wide pH range of 3–11. This phenomenon was attributed that h_vb⁺ was active sites for oxidizing electro/photochemical generated O₂^•- to produce ¹O₂ for BPA removal. The generation route of ¹O₂ was not restricted by pH value. As shown in Fig. 4d, the degradation efficiency of BPA was nearly maintained at the level of fresh sample after five consecutive runs, indicating that this fabricated CBCN cathode exhibited good stability. Based on above results, the CBCN/PE system exhibited great practical application prospects in wastewater treatment field.

In summary, we have constructed a metal free CBCN/PE system for selective generation of ¹O₂ via oxygen reduction, and then applied for efficient degradation of electron-rich organic pollutants. The CBCN was modified by introducing delocalized π bonds and cyano group simultaneously into CN framework through thermal polymerization method. The delocalized π bonds and cyano group in CBCN change the electronic structure of CN for both enhancing the photocatalytic and electrocatalytic activity. In CBCN/PE system, O₂ was firstly reduced to O₂^•- via 1-electron pathway, and then the produced O₂^•- would be simultaneously oxidized by h_vb⁺. As expected, the CBCN/PE system exhibits 100% removal efficiency for electron-rich pollutants such as bisphenol A (BPA) and acetaminophen (ACT). This study supplies a general strategy for the spontaneous formation of abundant ¹O₂ via PE oxygen activation. Hence, we expect this green, simple and economic strategy to prepare nonmetal CBCN/PE system with excellent capability could have broad application in water treatment filed.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

The authors acknowledge funding from the National Natural Science Foundation of China (Nos. 22076142, 21677106, 22076140), National Key Basic Research Program of China (No. 2017YFA0403402), National Natural Science Foundation of China (No. U1932119), the Science & Technology Commission of Shanghai Municipality (No. 14DZ2261100), the Fundamental Research Funds for the Central Universities.

Supplementary material associated with this article can be found, in the online version, at 10.1016/j.cclet.2021.05.066.