Nuclear energy is regarded as one of the most effective solutions to solve the shortage of fossil energy and environmental problems because of its high energy density and without-producing greenhouse gas [1]. As a long half-life nuclide, radionuclide uranium is mainly produced in spent fuel reprocessing process. Its high chemical toxicity and radioactivity result in a threat to human health and environmental safety, which places higher demands on environmental pollution remediation and prevention [2]. Thus, how to safely and efficiently remove U(Ⅵ) from wastewater has become a concern of the public and scientific community. Until now, there are many technologies which have been developed for treatment and removal of U(Ⅵ) including adsorption, ion exchange and chemical reduction [3]. Among them, adsorption technology has been popularized and applied for its cost-effectiveness and simple operation [4, 5]. To be specific, adsorption materials such as layered double hydroxides (LDHs) [6] and metal organic frameworks (MOFs) [7] have been used to efficiently remove uranium in wastewater. However, the recovery and reuse of LDHs and the high cost of MOFs limit their further application. Therefore, polymeric carbon nitride (PCN) with its easier-functionalization and unique electronic property has attracted more attention as a uranium adsorption material in recent years [8].

Graphitic C₃N₄ (g-C₃N₄) is one of the most stable allotropes of polymeric carbon nitride with simple modification and abundant electron density, which has been widely used in adsorption in the past [9]. Generally, g-C₃N₄ is used to adsorb radionuclides by adding or releasing adsorption sites by compounding with other materials or grafting functional groups, and the directional chemical structure modification of g-C₃N₄ can increase active sites so to bring special characteristics [10-12]. Vinu et al. found that adding additional nitrogen-rich units to the g-C₃N₄ framework to increase the N/C ratio in carbon nitride from 1.3 to 1.6–1.8 could expand the conjugate network and electron density. The precursor aminoguanidine hydrochloride was used to synthesize mesoporous carbon nitride (MCN-4) with high nitrogen content through the hard template method. MCN-4 had a lower band gap and a larger specific surface area, and it could be a highly active catalyst in friedel–crafts acylation of benzene using hexanoyl chloride as the acylating agent [13-15]. Sasanka et al. used the precursor dicyandiamide and the template SBA-15 to synthesize mesoporous carbon nitride nanodomains (MDY) with g-C₃N_4.7 stoichiometry by the hard template method, showing excellent performance in the photocatalytic hydrogen precipitation [16]. Liu et al. used PVP and NH₄Cl to synthesize N-doped carbon (NCG) with high nitrogen content through the one-pot pyrolysis of melamine, showing excellent adsorption performance for methylene blue [17]. Furthermore, Wu et al. [18] and Kumar et al. [19] used melamine pretreatment to synthesize C₃N₅ with a new stoichiometry, which also showed excellent adsorption photoreduction activity. However, these nitrogen-rich carbon nitrides research seem to be more focused on the field of catalysis or removal of dye pollutants, the research of nitrogen-rich carbon nitrides for radionuclide uranium absorption is still quite scarce.

Today, some researchers have found that nitrogen-rich materials or nitrogen-containing functional groups have a good capture effect on radionuclides, such as uranium. Wang et al. prepared N-doped porous magnetic materials (N-doped Fe/Fe₃C@C) by hydrothermal method using dopamine as the N source, which improved their adsorption properties on uranium and recyclability [20]. The same group also synthesized fabricate cotton-like graphitic carbon nitride@layered double hydroxides nanocomposites (g-C₃N₄@Ni-Mg-Al-LDH) by a mild hydrothermal process. The synergistic effect of carbon nitride and LDH enhanced the capture of UO₂²⁺ [21]. Wang et al. prepared metal-organic framework (UiO-66-AO) appended amidoxime by post synthetic modification method for extraction of uranium from seawater. It demonstrated very fast adsorption kinetics and high adsorption capacity of uranyl ion, also suggesting multiple amidoxime ligands can chelate U(VI) ions [22]. Furthermore, they found the placement of amidoxime chelating groups on MXene surface by diazonium salt grafting can enhance the selectivity for uranyl ions, and improve their stability in aqueous solution [23]. Qiu et al. fabricated amidoxime/carbon nitride (AO/g-C₃N₄) for capturing U(VI) from nuclear wastewater and seawater. It exhibited the high uranium extraction capacity from natural seawater and excellent selectivity [24]. Chen et al. report a sunlight-driven photocatalysis-assisted extraction (SUPER) method by utilizing a bifunctional carbon nitride material, CN550. Compared with the pure physicochemical adsorption (PA) method, a tenfold higher uranium extraction capacity [25].

Here, we firstly reported the removal of uranium using the nitrogen-rich carbon nitride polymer with C₃N₅ stoichiometry and explored its adsorption mechanism. Compared with g-C₃N₄, the heptazine unit is connected through an azo bridge in the C₃N₅ structure, which increases the conjugated system of the material, and its high electron density of states enhances the electrostatic bonding ability. The results show that C₃N₅ can capture uranyl ions by a combination of physical adsorption with chemical adsorption, also have high kinetic characteristics for uranyl ions, which only costs 10 min to adsorb 100 mg/g U(VI). Besides, it can achieve complete adsorption balance in 60 min with maximum adsorption capacity 207 mg/g. In addition, it shows high selectivity in presence of excess of competitive ions and good recyclability. Combining theoretical and experimental results, this work provides a method to improve the adsorption performance of carbon-nitrogen polymers on uranium by nitrogen-enrichment treatment and sheds light on the design and development of nitrogen-rich carbon nitride polymer for application in the separation and enrichment of uranium.

As illustrated in Fig. S1 (Supporting information), C₃N₅ is synthesized by melamine thermal polymerization and deamination (see Supporting information for experimental details). In order to explore the morphology of the prepared C₃N₅, we characterized SEM and TEM images. As shown in Fig. 1a and Fig. S2 (Supporting information), the prepared C₃N₅ has a porous overall structure, which is different from the traditional g-C₃N₄ block structure. And the TEM of C₃N₅ (Fig. 1b) further demonstrates the characteristics of the material being stacked in layers. The wider diffraction ring is seen in the selected area electron diffraction (SARD) of C₃N₅ corresponds to the 002 plane of the material, and it shows that the material is amorphous (Fig. 1c).

In order to research the crystal mode information, we performed XRD characterization on the prepared C₃N₅ and g-C₃N₄. The difference in the structure of phases and changes in the crystalline property of the C₃N₅ precursor are shown in Fig. S3 (Supporting information) [26]. As shown in Fig. 2a, the XRD spectrum of g-C₃N₄ shows two peaks at around 12.9° and 27.3° indexed to the (100) and (002) planes, which the former corresponds to the in-planar structural packing of the motif, and the latter can be ascribed to the interplanar stacking of the sheets [27, 28]. The XRD spectrum of C₃N₅ only shows one peak at 27.5° corresponded to (002) plane, and there is no peak corresponding to (100) plane. It is proved that the distortion of the C₃N₅ grid structure also means that the bridging of the azo bond to the heptazine unit makes the nanochannel distance between heptazine larger [29]. These XRD results clearly support the distinct structure of C₃N₅ possessing azo linkage. The slight increase in the 2θ value and d spacing can be explained by the repulsion between the electron-rich π-conjugated C₃N₅ sheets [30]. Additionally, the changes in the functional groups on the surface of the original g-C₃N₄ and C₃N₅ were further studied by FT-IR spectroscopy (Fig. 2b). The changes in the preparation of C₃N₅ precursors are shown in Fig. S4 (Supporting information). The peaks of C₃N₅ and g-C₃N₄ at 809 cm⁻¹ are all attributed to the characteristic peaks of the triazine ring, which is consistent with the literature reports [27]. The peaks at 1617, 1413 and 1245 cm⁻¹ are basically the same as g-C₃N₄, indicating that the basic building blocks are still triazine units [31]. However, the difference between C₃N₅ and g-C₃N₄ infrared spectra at 887, 1324 and 1554 cm⁻¹ indicate that C₃N₅ and g-C₃N₄ have completely different grid structures [19]. As for C₃N₅, since the azo bond is forbidden to vibrate, no new vibration peaks have been found due to azo. We also tested the N₂ sorption/desorption isotherm of C₃N₅ and g-C₃N₄ (Fig. S5 in Supporting information), there is no significant difference in the specific surface area of C₃N₅ and g-C₃N₄, which are 7.04 m²/g and 7.3 m²/g, respectively. Moreover, it is tested that the bonding performance and electronic structure of C and N in the material by measuring the electron energy loss spectrum (EELS).

As shown in Fig. 2c, two symmetrical peaks at 286.6 eV and 293.1 eV were appeared in the normalized EELS spectra of C₃N₅ and g-C₃N₄ due to C K-edge loss, corresponding to the 1s-π* and 1s-σ* electronic transitions of the sp² hybrid carbon trigonally bonding with N in the s-heptazine nucleus [32, 33]. The C K-edge intensity of π* belonging to C₃N₅ and the peak area ratio of π*/σ* of C₃N₅ are slightly higher, indicating that the superposition of the π orbitals of the bridged azo group in C₃N₅ structure and the heptazine sequence leads to increased conjugation, and such an increased conjugation also leads to higher electron density in C₃N₅ than in g-C₃N₄ [34]. This is also demonstrated by the calculation of the surface charge density distribution. In contrast to the electron density dispersion at the bridge linkage of g-C₃N₄, the electron density at the bridge linkage of C₃N₅ is more concentrated due to the azo. Meanwhile, the N K edge peaks of C₃N₅ and g-C₃N₄ are located at 397.8 eV and 406.3 eV (Fig. 2d), which are attributed to the sp² hybrid nitrogen and bridging nitrogen in the heptazine unit [35]. The electronic states of N of the two are basically similar, but the peak area of C₃N₅ was increased slightly relative to g-C₃N₄, which might be because of the extra nitrogen in the framework. Since the electronic transition caused by the azo motif makes a greater contribution to the 1s-σ* transition, the relative peak intensity of the π* signal at the N K edge of C₃N₅ is reduced [14]. The N(C)₃ of g-C₃N₄ passes through the azo to form a lone pair on the azo, which suppresses the relative strength of the π* and σ* signals.

Moreover, XPS spectra were obtained to further confirm the functional information on pristine g-C₃N₄ and C₃N₅. From Fig. S6 (Supporting information), the XPS elemental survey scan of g-C₃N₄ and C₃N₅ shows peaks corresponding to C 1s, N 1s, O 1s and C 1s, N 1s, Na 1s, O 1s, respectively. The presence of Na 1s of C₃N₅ comes from intercalated Na⁺ ions in the supramolecular cavity of the C₃N₅ motif. As shown in Fig. 3a, the high-resolution of spectrum for C 1s of g-C₃N₄ can be deconvolved into three components at binding energies of 284.8 eV, 286.4 eV and 288.15 eV, corresponding to C-C coordination of surface adventitious carbon, C-NH_x (x = 1, 2) on the edges of heptazine units and the sp² hybrid N=C-N in the stronger tri-s-triazine ring [36, 37]. The spectrum for N 1s of g-C₃N₄ (Fig. 3b) includes four peaks at binding energies of 396.61 eV, 399.83 eV, 401.12 eV and 404.45 eV, corresponding to the N (C-N=C) bonded to two sp² hybrid carbons in the g-C₃N₄ framework, the N (C-N(C)-C) connected to the three sp² hybrid carbons in the tri-s-triazine ring, the residual N-H bond at the edge of g-C₃N₄, and the charged effect caused by π excitation, respectively [38, 39]. The high-resolution of spectrum for C 1s of C₃N₅ (Fig. 3c) is deconvolved into three components at binding energies of 284.8 eV, 286.39 eV and 287.9 eV, corresponding to the sp³ hybrid carbons, C-NH_x (x = 1, 2) on the edges of heptazine units and sp² hybrid carbons, respectively [36]. The sp³ carbon corresponds to the exogenous carbon, turbostratic carbons and the edge carbon in the scaffold of the C₃N₅ framework, while the stronger sp² carbon corresponds to the N=C-N in the C₃N₅ framework [18]. In addition, the difference for C 1s of C₃N₅ is slightly lower (0.14 eV) at the sp² hybrid carbon N=C-N bond at 288.01 eV compared to g-C₃N₄, indicating that the charge density on the nitrogen-containing heterocycle of C₃N₅ is greater. Zeta potential test results also provide support for this view (Fig. S7b in Supporting information). The spectra for N 1s of C₃N₅ (Fig. 3d) can be divided into two components at binding energies of 398.55 eV and 400.13 eV. The peak at 398.55 eV could be ascribed to C-N=C and C-N(C)-C in the aromatic ring structure of C₃N₅, while the other peak at 400.13 eV corresponds to bridged C-N=NnullC and residual C-NH₂ [19]. There is an obvious bridging C-N=NnullC peak instead of N-(C)₃ in N 1s region of C₃N₅, which shows that the connection mode of each unit of C₃N₅ is azo bond connection instead of N-(C)₃ bond connection. Since the precursor MH of C₃N₅ have undergone the reaction of melem and hydrazine hydrate to convert amino groups into hydrazine groups, there are fewer residual amino groups in the N 1s of C₃N₅, which are separated to 400.13 eV. Furthermore, the N on the aromatic ring and the bridging N account for 60.83% and 39.17%, respectively, and their ratio is 3:2, which indicates that the two heptazine units are connected through one azo. The XRD, FT-IR and XPS results reveal that the C₃N₅ with a heptazine structure was successfully synthesized.

As we know, the aqueous speciation of U(VI) in an aqueous solution is greatly affected by solution pH (Fig. S7a in Supporting information). When pH < 4, U(Ⅵ) exists as free UO₂²⁺ cations, with the increase of pH, it is dominated by multi-nuclear hydroxide complex cations, such as (UO₂)₂(OH)₂²⁺, (UO₂)₃(OH)₅⁺, which may affect the adsorption behavior. Hence, the adsorption experiments of U(Ⅵ) on g-C₃N₄ and C₃N₅ under different pH (2–5.5) were carried out (Fig. 4a). It can be seen that with the increase of pH, the removal ratio of U(VI) by C₃N₅ gradually increases from 5% to a maximum of 99% at pH 4, but decreases as the pH continues to go up. As for g-C₃N₄, no matter how the pH changes, it has almost no adsorption performance for U(VI). This phenomenon does not only come from the change of species, but also be affected by the surface potential of the material. C₃N₅ has a lower surface potential than g-C₃N₄ due to the increase in electron density, which are −18.76 mV and 23.67 mV at pH 4, respectively (Fig. S7b). Thus, the adsorption performance of C₃N₅ is better than that of g-C₃N₄ during the experimental pH range because of the electrostatic interaction force. Taking account of the better adsorption performance and U(VI) hydrolysis caused by excessively high pH, the subsequent adsorption experiments were carried out at a pH value of 4.

To better clarify the performance differences between C₃N₅ and g-C₃N₄ in adsorption of U(Ⅵ), adsorption kinetics experiments and adsorption isotherm studies were being carried out. As depicted in Fig. 4b, the uranium capture by g-C₃N₄ is sluggish and almost inefficient, with the removal efficiency of uranium achieving 4% after 8 h. While the adsorption equilibrium time of C₃N₅ is relatively faster, the removal efficiency of uranium can reach more than 75% in 10 min, and the adsorption equilibrium can be reached at near 30 min (Fig. S8 in Supporting information). The sorption kinetics curve of C₃N₅ and g-C₃N₄ towards U(Ⅵ) can be well fitted to the pseudo-second-order model (Fig. S9 and Table S1 in Supporting information), which means that the pseudo-second-order kinetic model is more suitable to describe the adsorption of U(Ⅵ) on C₃N₅ and g-C₃N₄, and further shows that the process is controlled by chemical adsorption. The adsorption isotherm studies were determined by adjusting the initial concentration of U(Ⅵ) from 10 ppm to 500 ppm at a constant pH of 4.0 ± 0.1. As shown in Fig. 4c, the maximum adsorption capacity of C₃N₅ to U(VI) is 207 mg/g, while the g-C₃N₄ to U(VI) is about 50 mg/g. Obviously, the adsorption performance of C₃N₅ for U(VI) is much higher than that of g-C₃N₄, which is about 4.1 times that of g-C₃N₄. The sorption isotherms curve of C₃N₅ and g-C₃N₄ towards U(Ⅵ) can be well fitted to the Langmuir model and Freundlich model, respectively (Fig. S10 and Table S2 in Supporting information). These results reveal the monolayer adsorption of U(Ⅵ) on C₃N₅ and multilayer adsorption reaction on g-C₃N₄.

To further understand the nature of the sorption process, the adsorption of U(VI) by C₃N₅ sorbent at different temperatures was performed, and the results were shown in Fig. 4d. It can be seen that the adsorption capacity of C₃N₅ to U(VI) gradually increases with the increase of temperature, indicating that higher temperature is more beneficial to the adsorption. It is known that the dependence of adsorption on temperature is related to changes in thermodynamic parameters, which are standard enthalpy (ΔH₀), standard entropy (ΔS₀), standard free energy (ΔG₀), these can be calculated by the following equations:

(1)

(2)

where K_d is the distribution coefficient (mL/g), R is the gas constant (8.314 J mol⁻¹ K⁻¹) and T is the absolute temperature (K). The results were shown in inset of Figs. 5a and d and Table S3 (Supporting information). The positive value of ΔH₀ (21.04 kJ/mol) indicates that the adsorption process is an endothermic reaction, and the negative value of ΔG₀ indicates that the adsorption process can proceed spontaneously. Besides, the more negative values of ΔG₀ at higher temperature further confirm that the higher temperature promotes the progress of the adsorption reaction.

In order to investigate the practical application potential of the material, the selective experiment and cyclic experiment of C₃N₅ to uranium were carried out. The U(VI) solution was added with excessive amounts of metal ions in the form of nitrates, such as La³⁺, Nd³⁺, Sm³⁺, Gd³⁺, Yb³⁺, Sr²⁺, Ni²⁺, Co²⁺ and Zn²⁺ ions, whose concentration shows 10 times than that of U(VI) [40]. The removal rate of U(VI) by C₃N₅ can reach 57.4% in the system with excess metal ions (Fig. 4e) while the removal rate of other ions is at a very low level. Compared with the removal ratio of 63.2% in the blank experiment (Fig. S11 in Supporting information), there was no significant change. It indicates that competing ions have no obvious inhibitory effect on the uptake of uranium and also shows that C₃N₅ has good selectivity to U(VI). The reason for the selectivity may be on the electronegativity of different metal ions. The solution-phase electronegativity is a description of the strength of the electron absorption process of metal ions in aqueous solution, and the order of change of electronegativity of metal ions might be Sr²⁺ < Co²⁺ < Ni²⁺ < Zn²⁺ < La³⁺ < Nd³⁺ < Gd³⁺ < Sm³⁺ < Yb³⁺ < UO₂²⁺ [41, 42]. In addition, C₃N₅ exhibits high consistency in the continuous adsorption process (Fig. 4f). It can be seen that the U(VI) sorption by the regenerated C₃N₅ does not obviously decrease even after 5 cycles (> 85%), which suggests excellent reusability of the sorbent.

Furthermore, to theoretically investigate the structure, adsorption energy, differential charge, and charge density for C₃N₅ and g-C₃N₄. The g-C₃N₄ and C₃N₅ structures were optimized for cell expansion of 2 × 2 × 1 size, and 2 × 2 × 1 supercell structures were formed by cutting along the (001) crystal plane (Fig. S12 in Supporting information). We analyzed the distribution of the charge density surface of g-C₃N₄ and C₃N₅ at adsorption position Ⅱ, which was shown in Figs. 5b, c, e and f. The isosurface of the electron density difference was plotted at a value of 0.0037 electrons/ų, and the yellow and green isosurfaces representing the charge accumulation and depletion regions, respectively. It can be clearly seen that compared with the g-C₃N₄, there is obvious charge transfer in the C₃N₅, and the latter one shows higher adsorption energy in Tables S4 and S5 (Supporting information). Besides, the average bond length of U-N of C₃N₅-U is generally smaller than that of g-C₃N₄-U, indicating that C₃N₅ can form stronger interactions with uranyl ions than g-C₃N₄. By comparing the adsorption energies and U-N distances of g-C₃N₄ and C₃N₅ at three different adsorption positions (Figs. S13 and S14 in Supporting information), we found that C₃N₅ always exhibits a stronger adsorption capacity for uranyl ions regardless of the adsorption position. In addition, the apparent charge transfer differences at the Ⅲ adsorption positions may result in C₃N₅, possessing more adsorption sites than g-C₃N₄. All in all, the DFT calculation results are in accordance with the above experimental results.

To explain the adsorption mechanism, the FTIR spectra and XPS were conducted for pristine and U(VI)-adsorbed samples (g-C₃N₄ and C₃N₅) (Fig. 6). Obviously, no peak of U 4f and characteristic stretching bond of UO₂²⁺ were found in the g-C₃N₄-U (uranium-loaded g-C₃N₄ after sorption) sample (Figs. S15 and S16 in Supporting information). As shown in Figs. 6a and b, the XPS spectra for C 1s and N 1s of g-C₃N₄ before and after adsorption have almost no shape change and shift, which proves that there is no interaction between g-C₃N₄ and U(VI). On the contrary, a new peak at 933 cm⁻¹ on C₃N₅-U (uranium-loaded C₃N₅ after sorption) of FTIR spectra, which is a characteristic stretching bond of UO₂²⁺, confirms the fact of U(VI) adsorption, while there is no change in the characteristic peak of the triazine, indicating that the basic framework of the material is well maintained (Fig. 6f). As shown in Fig. S17 (Supporting information), the XPS elemental survey scan of C₃N₅-U shows a new peak corresponding to U 4f, which also explains the adsorption of U(VI).

As shown in Fig. 6c, compared with the C 1s spectra in the two samples, it is observed that the peak located at 288.01 eV for C₃N₅ become wider and moved 0.29 eV to higher binding energy after loading with uranium for C₃N₅-U. Meanwhile, the N 1s spectra peak locates at 398.55 eV and 400.13 eV for C₃N₅ moves 0.31 and 0.3 eV to higher binding energy after loading with uranium for C₃N₅-U, respectively (Fig. 6d). These indicate the decrease of electron density on the bridged azo and heptazine rings in the C₃N₅ skeleton after loading of U(VI). The high-resolution of U 4f_5/2 and U 4f_7/2 spectra contains several single peaks at 392.9, 390.9 eV and 384.74, 382.05, 379.96 eV, respectively (Fig. 6e). The peak at 390.9 and 379.96 eV correspond to U(IV), which shows the part of U(VI) is reduced to U(IV) during the adsorption process, when the relative ratio of U(IV) was calculated to be 13.2% in samples [43]. Besides, the uranium adsorbed on C₃N₅ exists mainly in the UO₃·2H₂O and (UO₂)₄O(OH)₆(H₂O)₅ form of U(VI), indicates that the uranyl ion forms coordination with negative oxygen and hydroxide ions in solution in the presence of C₃N₅ (Fig. S18 in Supporting information). The main mechanism for the formation of these complexes may be from the electron-rich property of C₃N₅ trapping the hydrogen ions in solution and increasing the concentration of hydroxide ions in the system, leading to the complexation of excess hydroxide ions with UO₂²⁺. The results testify that strong chemical interaction and electron transfer happened between uranyl and the bridged azo on C₃N₅, which indicated that the excess electrons on the C₃N₅ backbone are transferred to the surface adsorbed uranyl during the adsorption process to reduce it from U(VI) to U(IV). On the other hand, many adsorption processes are related to the surface potential of material, the lower surface potential of C₃N₅ and change in electron density before and after adsorption suggest that electrostatic adsorption is also a possible mechanism.

In conclusion, we synthesized a new carbon nitride polymer with C₃N₅ stoichiometric for the removal of uranium. The C₃N₅ backbone composed of azo-bridged heptazine units which had a strong conjugation system and high electron density for the generation of additional lone pairs of electrons. Compared to conventional carbon nitrogen polymers (g-C₃N₄), the nitrogen-rich C₃N₅ had several advantages including high adsorption capacities and rapid kinetics. Meanwhile, C₃N₅ exhibited excellent selectivity for the UO₂²⁺ ion against hard ions such as Gd³⁺, Yb³⁺, Ni²⁺ and Sr²⁺. In addition, it can be recycled by a low-cost and environmentally friendly approach. The results of spectral analysis and theoretical calculations elucidated the mechanism of nitrogen enrichment for enhancing the physicochemical properties and the ability to capture uranium in carbo nitride polymers. Therefore, this work indicated that C₃N₅ has great promise and application in the separation and enrichment of uranyl ions, and also provides a reference for the systematic investigation of the adsorption ability of nitrogen-rich carbon nitrogen polymers on uranyl ions.

Declaration of competing interest

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.

Acknowledgments

This work was supported by National Natural Science Foundation of China (Nos. U2167221, 21976148, 21906133, 21902129); the Basic Scientific Research Project of China (No. JCKY2018404C008); the National key research and development Project of China (No. 2016YFC1402500); the Project of State Key Laboratory of Environment-friendly Energy Materials, Southwest University of Science and Technology (No. 18ZXHK04); the Long Shan Talent Project (Nos. 18LZX304, 18LZXT04).

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cclet.2022.03.097.