Phragmites australis samples were collected from a wetland in Dong Thap province, Vietnam, and cleaned with tap water to remove dirt and other impurities adhered to their surfaces. The stems were dried under sunlight for four days before being finely ground to approximately 1–2 mm sizes. The obtained biomass was rinsed with water and dried in a vacuum oven at 70°C to get a constant weight. The product was stored in a desiccator and used as raw P. australis biomass (cellulose 43.31%, hemicellulose 30.82%, and lignin 20.37%) [38].

Sodium hydroxide (NaOH, ≥99%), hydrochloric acid (HCl, 37%), potassium iodide (KI, ≥99.5%), potassium bromate (KBrO₃, 99.8%), urea (CO(NH₂)₂, >99%), and sodium chloride (NaCl, ≥99.5%) were purchased from Merck. Zinc nitrate hexahydrate (Zn(NO₃)₂·6H₂O, ≥99.0%), tert-butanol ((CH₃)₃COH, tBA, 99.5%), doxycycline hyclate (C₂₂H₂₄N₂O₈·HCl·0.5H₂O·0.5C₂H₆O, DC, 93.5% (HPLC)) and L-ascorbic acid (C₆H₈O₆, AA, 99%) were obtained from Sigma – Aldrich. All chemicals used were of analytical grade and were used as received without any further purification.

XRD patterns of ZnO, g-C₃N₄, biochar, ZCN, and ZCNB are shown in Fig. 1a. We can see that the XRD pattern of biochar has almost no diffraction peaks because of its amorphous structure [42]. The XRD pattern of ZnO presents characteristic diffraction peaks of the ZnO hexagonal wurtzite-phase structure at 2θ angles of 31.7, 34.4, 36.2, 47.5, 56.6, 62.8, 66.4, 67.9, 69.1, 72.6, and 76.9°, indexed as (100), (002), (101), (102), (110), (103), (200), (112), (201), (004), and (202) planes, respectively (JCPDS-01-075-9743). As for g-C₃N₄, the diffraction characteristic peaks at 13.1 and 27.4° are attributed to the (100) and (002) planes, as reported previously [43–45]. The ZCN sample also shows characteristic diffraction peaks for the ZnO wurtzite structure, but their intensity is lower than that of pristine ZnO. This decrease indicates the formation of g-C₃N₄/ZnO composite, and it is possible that g-C₃N₄ inhibits ZnO crystal growth, as reported in previous works [22, 46]. Similar to g-C₃N₄, ZCNB exhibits a new phase of zinc cyanamide (ZnNCN) with low diffraction intensity at 2θ angles of approximately 19.1, 27.9, and 28.6° for faces (101), (211) and (220), respectively, besides the characteristic diffraction peaks of ZnO [47]. This structure might result from condensation between the amino triazine group and the surface hydroxyl groups of ZnO to form Zn-N bonds [48].

The average crystal size of the ZnO, ZCNB, and ZCNB samples is about 30.3, 26.6, and 25.7 nm, according to the Scherrer equation at the diffraction peak (101), respectively (Table S3). The variation in grain size indicates that g-C₃N₄ and biochar can significantly inhibit the crystal growth of ZnO in the composites.

The functional groups of ZnO, g-C₃N₄, biochar, ZCN, and ZCNB are assessed via the FTIR spectroscopic spectra (Fig. 1b). We can see that all samples have absorption bands at around 3400 cm⁻¹ and 2350 cm⁻¹, related to the stretching vibrations of O–H in water and C–C in biochar, oxygen in the carbon dioxide molecule is adsorbed on the sample surface [21]. The absorption bands of ZnO observed at around 3400 cm⁻¹, 1640 cm⁻¹, and 498 cm⁻¹ are related to the –OH, H–O–H and Zn–O stretching vibrations, respectively [48, 49]. For biochar, in addition to the absorption bands related to the –OH and C–O vibrations as mentioned above, there are also absorption bands at 2924 cm⁻¹ and 1619 cm⁻¹, assigned to asymmetry vibrations of C–H and sp² character of C=C [37, 49]. The band at 1462 cm⁻¹ can be attributed to the stretching vibration of –COO– [50]. The region between 900–1300 cm⁻¹ and 1385 cm⁻¹ is thought to be involved with C–O vibration and the binding vibrations of O–H, C–C, and C–H [51, 52]. For g-C₃N₄, the wide absorption region between 3200 cm⁻¹ and 3600 cm⁻¹ is thought to be responsible for the stretching vibrations of N–H and O–H [49]. Furthermore, sample g-C₃N₄ has several typical bands for the stretching vibrations of the C–N heterojunctions (1636 cm⁻¹, 1459 cm⁻¹, and 1410 cm⁻¹) and high-intensity strain vibrations of tri-unit s-triazine at 813 cm⁻¹ [44, 48]. The bands at 1317 cm⁻¹ and 1240 cm⁻¹ are associated with aromatic C–N stretching [44, 48]. More importantly, the absence of absorption bands at 3000 cm⁻¹ and 2200 cm⁻¹ rules out the –C≡C– bond formation. The ZCN and ZCNB samples also have characteristic bands but with a lower intensity than that of the component materials, demonstrating the successful formation of heterojunction in the composite [22, 53]. In addition, from Fig. 1b, we also observe that both ZCN and ZCNB composite samples have a new absorption band at 2050 cm⁻¹, which is typical for the asymmetric stretching vibration of –N=C=N– [54], and this is also consistent with the XRD analysis. Xie et al. suggest that the formation of the g-C₃N₄/ZnNCN heterostructure can enhance the separation efficiency between photogenerated electrons and holes [54], which is one of the crucial factors for improving the photocatalytic activity of photocatalysts. It is possible that g-C₃N₄ uniformly coats ZnO nanoparticles, leading to blocking the order stacking pattern of g-C₃N₄ in the long-range order [20, 21].

The optical properties of ZnO, g-C₃N₄, biochar, ZCN, and ZCNB are estimated by using UV-Vis absorption spectroscopy. The absorption spectra exhibit the adsorption band covering the ultraviolet visible light region (Fig. 1c) [55]. It can be seen that the pristine ZnO has an upward-moving optical absorption edge at a wavelength of about 390 nm. The g-C₃N₄ sample has an upward-moving optical absorption edge in the wavelength region of about 445 nm. After the surface of g-C₃N₄ or both g-C₃N₄ and biochar was coated with ZnO, the optical absorption edge of the two composites ZCN and ZCNB has a significant blue shift towards the wavelength regions of 466 nm and 490 nm. This shift indicates a small crystal size of the ZCN and ZCNB nanocomposites. To calculate the band gap values, we use the Tauc plot relation [56].

where α is the absorption coefficient; h is the Planck constant; ν is the wavenumber; A is a constant, and E_bg is the energy band gap. The bandgap energy value from the absorption data was calculated by plotting (α × h × ν)² against the photon energy (E_bg = h × ν) (Fig. 1d) [22, 56]. Accordingly, the E_bg values of ZnO, g-C₃N₄, ZCN, and ZCNB are 3.18, 2.79, 2.66, and 2.53 eV, respectively (Table S3). Thus, a transition to lower energy occurs for the ZCNB nanocomposite, and its ability to absorb visible light increases [53].

Fig. 1e presents PL spectra of the obtained materials. The photoluminescence of ZnO, g-C₃N₄ and ZCN peak at 509; 438 and 473 nm, respectively. Since the PL intensity of g-C₃N₄ is significantly larger than that of ZnO, hence the photoluminescence of ZCN is determined by g-C₃N₄ and peaks at around 473 nm. The PL intensity of the ZCN composite is lower than that of the ZnO and g-C₃N₄ components, suggesting that g-C₃N₄ on the ZnO surface can inhibit electron-hole recombination [22]. The ZCNB composite exhibits a much lower absorption intensity than the others, indicating that the combination of ZnO with both g-C₃N₄ and biochar can enhance the separation efficiency of the photogenerated electron-hole pairs [53, 57].

The textural properties of the ZnO, g-C₃N₄, biochar, ZCN, and ZCNB samples were studied from the nitrogen adsorption/desorption isotherms at 77 K (Fig. 1f and Table S4). Their N₂ isotherms belong to type IV with an H3 hysteresis loop, according to the International Union of Pure and Applied Chemistry (IUPAC), and they indicate that the obtained samples possess a mesoporous structure [58, 59]. The H3 hysteresis loop is assigned to the non-rigid materials with plate-like particles, forming slit-shaped pores [60]. The individual components of the composite have a specific surface area (S_BET) as follows: 5.5 m²·g⁻¹ for ZnO, 22.8 m²·g⁻¹ for biochar, and 36.8 m²·g⁻¹ for g-C₃N₄ (Table S2). Meanwhile, their two- and three-component composites exhibit a higher specific surface area: 28.4 m²·g⁻¹ for ZCN and 25.1 m²·g⁻¹ for ZCNB. These values demonstrate that g-C₃N₄ and biochar significantly improve the surface characteristics of ZnO, and the resulting composites have a capillary structure. Therefore, it can be concluded that the dispersion of ZnO onto g-C₃N₄ and biochar increases the surface area, which is beneficial for the surface reactions of the adsorbents. In general, an increased specific surface area can provide more potential active sites. Therefore, the large surface area, volume, and pore diameter of ZCNB could be one of the reasons for higher catalytic activity since the large surface area helps to increase the mass transfer rate, leading to an improved photocatalytic activity [61]. Furthermore, g-C₃N₄ in the composite not only enhances the surface area and reduces the particle size of ZnO but also forms heterojunctions. This structure is beneficial for improving the efficiency of migration of the photogenic charged particles, providing more photocatalytic sites and further improves the photodegradation efficiency [58].

The morphology of the ZnO, g-C₃N₄, biochar, ZCN, and ZCNB samples can be seen on SEM and TEM images. Fig. 2a and 2b show that ZnO exists as agglomerated spherical particles with an average diameter of about 30–35 nm. Fig. 2c displays the structure of g-C₃N₄ as thin flake-like particles with diameters ranging from a few hundred nanometres to several micrometres, overlapping the slits. Fig. 2d shows the structure of biochar in the plate form with a smooth surface. Fig. 2e and 2f display the SEM and TEM images of the ZCN composite and indicate that this material exists in the form of spherical particles with a ZnO particle size of about 30–35 nm. Fig. 2g and 2h show the SEM and TEM images of the ZCNB composite and reveal that ZCNB consists of both nodular and plate-shaped particles with a ZnO grain size of about 20–30 nm. The grains are more even than those of ZnO itself, and they seem to be successfully attached to g-C₃N₄ and biochar surfaces. Finally, Fig. 2i shows a high-resolution TEM (HRTEM) with lattice fringes with distances d(002) (2.6 Å) and d(101) (2.4 Å) for the ZnO wurtzite in ZCNB composite [62, 63].

The EDX elemental mapping of a selected area on the ZCNB composite sample is shown in Fig. 3. Carbon and nitrogen are densely distributed in the ZCNB lattice (Fig. 3d and 3e). Meanwhile, O and Zn are uniformly distributed on the surface of g-C₃N₄ and biochar in the ZCNB composite (Fig. 3f and 3g). The existence of C, N, O, and Zn on ZCNB and the merged image of C, N, O, and Zn (Fig. 3c) confirm the spatial distribution of the elements in the ZCNB composite structure. The EDX analysis further indicates the presence of C, N, O, and Zn in ZCNB (Fig. 3b). The atomic composition of the surface is as follows: C 29.17, N 36.69, O 16.9, and Zn 17.24% (the inset of Fig. 3b). These results demonstrate that ZnO nanoparticles are highly dispersed on g-C₃N₄ and biochar.

The core-level XPS of Zn 2p, O 1s, C 1s and N 1s of ZCNB are presented in Fig. 4. The Zn 2p spectrum at 1024.7 eV and 1047.8 eV are acribed Zn 2p_3/2 and Zn 2p_1/2 of hexagonal structure (Fig. 4a) [48]. O 1s spectrum of Fig. 4b presents the peaks at 524.5 eV and 526.3 eV contributed to O²⁻ in ZnO, while peaks at 533.5 eV, 534.5 eV and 535.4 eV belong to O²⁻ in oxigen deffect in ZnO or the OH and O2 adsorbed on the catalyst surface, respectively [48, 64]. The presence of O₂ in defected region make it form O^{2− ·} and improve catalytic ativity (Fig. 4b) [64]. C 1s spectrum peaks at 281.6 eV, 287.5 eV that assigned to C–N [65] and 290.6 eV corresponding to bonding energy of C in C–(N)₃ [66, 67] or C=O [68] (Fig. 4c). The core-level XPS of N 1s is deconvulated into three peaks at 397.2 eV, 399.1 eV and 407.6 eV, which are assigned to the sp² hybridized nitrogen C=N–C, tertiary N in N–C₃ and N atoms in amino moieties, respectively (Fig. 4d) [64, 65, 69].

For primary tests, the effect of the different catalyst loadings (0.05, 0.10, 0.15, and 0.20 grams) in 200 mL of 25 mg.L⁻¹ DC at pH 5 on DC degradation is depicted in Fig. S1a. DC decomposition efficiency increased from 90.9 to 99.8% corresponding to an increase in the amount of ZCNB from 0.05 to 0.20 grams. This may be due to the presence of more active sites with increasing ZCNB dosage and thus facilitating the generation of oxidation agents [70]. The sample at the amount of ZCNB 0.10 gram showed a fairly high DC decomposition efficiency (98.9%), and therefore we chose this amount of catalyst as the appropriate condition for the further experiments.

The photocatalytic activity of pristine ZnO, pristine g-C₃N₄, ZCN, and ZCNB was studied via doxycycline photodegradation under visible light irradiation (λ > 420 nm) (Fig. 5a). When no catalysts are applied, no photolysis of DC is observed after 2 h of visible light irradiation, demonstrating that DC is stable under visible light [40, 71]. Similarly, ZnO does not favour DC decomposition under visible light radiation. Like previously reported results [72, 73], pristine g-C₃N₄ exhibits good performance in photodegradation under visible light in the present work. The ZCN composite has a much higher photocatalytic activity than pristine g-C₃N₄ for DC degradation, with an efficiency of 69.74% after 120 min. Meanwhile, the ZCNB photocatalyst is the most active in DC degradation, with an efficiency of 98.9%.

To better understand the reaction kinetics of DC degradation catalyzed by different photocatalysts, we fitted the experimental data with the Langmuir-Hinshelwood first-order model (Eq. 3). As can be seen from Fig. 5b, the photocatalytic degradation curves in all cases are consistent with the apparent first-order kinetic model with high determination coefficients (R²>0.95). Furthermore, a significant difference in the photocatalytic activity of ZCNB from the others implies that the incorporation of ZnO with g-C₃N₄ and biochar substantially improves the photocatalytic activity of ZnO, and this composite has the highest degradation rate constant for DC photodegradation. The rate constant for ZnO, g-C₃N₄, ZCN, and ZCNB is about 2.39×10⁻⁴, 0.007, 0.016, and 0.069 min⁻¹, respectively. The ZCNB composite is about 9.31 and 4.18 times more effective than pristine g-C₃N₄ and ZCN. A comparison of the first-order apparent degradation rate constant of the present catalyst with the literature is presented in Table S5. The data show that the catalytic activity of the ZCNB composite is relatively high compared with what was reported in previous papers [6, 39, 72–74].

The COD values were determined to assess the mineralisation of DC over ZCNB (Fig. 5c). It is clear that the COD gradually decreases with time, indicating a gradual decomposition of DC. The reduction of COD up to 93.52% is achieved after 120 min of illumination with the ZCNB composite. Therefore, it can be concluded that the DC is decomposed into carbon dioxide over the ZCNB almost completely.

Since photocatalytic degradation involves photo-induced electron/hole pairs (e⁻/h⁺), superoxide radicals (•O₂⁻) and hydroxyl radicals (•OH), several scavengers were used to study the effect of free radicals on DC degradation (Fig. 5d). For these experiments, 10 mM of potassium iodide (KI), potassium bromate (KBrO₃), tert-butanol (t-BA), and ascorbic acid (AA) was added to the DC solution to capture h⁺, e⁻, •OH, and •O₂⁻, respectively. Fig. 5d shows that the degradation efficiency drops to 33.61, 58.58, 80.14, and 71.88% upon the addition of KI, KBrO₃, tBA, and AA, respectively. First, tBA is considered a scavenger for •OH [72], it is considered that t-BA presents a strong interaction with the •OH through an electron transfer process [75].

If the •OH radicals play a crucial role in the DC degradation, the reaction rate is expected to decrease significantly. As shown in Fig. 5d, adding an excess amount of t-BA to the reaction mixture significantly suppresses the DC degradation (by ~ 80.1%) compared with the mixtures without the scavenger. Secondly, the addition of AA drops the decomposition efficiency by around 71.9% compared with the absence of scavenger. Like benzoquinone, AA exhibits ability to capture •O₂⁻ by an electron transfer mechanism [76],

Thirdly, the BrO₃⁻ ion captures photogenerated electrons and then reduces to bromide (Br⁻) [77],

and the addition of the KBrO₃ makes the degradation efficiency dropping around 58.58% compared with the absence of scavenger.

Finally, the iodide ion is a common scavenger that reacts with vacancy-band holes [78].

The addition of the KI reduces the degradation efficiency only to 33.62%. Therefore, it can be concluded that the formation of •OH and •O₂⁻ play an essential role in the photocatalytic degradation of DC.

The electrons and holes generated in the photocatalyst react with the water and oxygen molecules adsorbed on the catalyst surface and generate •OH and •O₂⁻ that react to decompose the DC molecules in solutions [44,45].

The valence-band energy potential (E_VB) and conduction-band energy potential (E_CB) of a semiconductor can be calculated as follows [22]:

where χ is the absolute electronegativity of the semiconductor, and the values of χ were 4.74 and 5.6 eV for g-C₃N₄ and ZnO [22]. E_e is the energy of free electrons versus hydrogen (4.5 eV), and E_bg is the bandgap value of the semiconductor. The calculated values of E_VB and E_CB are 2.69 and –0.49 eV for ZnO and 1.635 and −1.155 eV for g-C₃N₄. Because of the different positions of valence-band and conduction-band potentials between ZnO and g-C₃N₄, type II heterojunction can occur with coupling effects [79]. The ZnO semiconductor cannot be excited under visible-light illumination since it has a large band gap (3.18 eV). Pristine g-C₃N₄, with its narrow bandgap energy (2.79 eV), is excited by visible light to create photo-generated electron-hole pairs (Eq. 13). Since the CB edge potential (−1.155 eV) of g-C₃N₄ is more negative than that of ZnO (−0.49 eV), the photoinduced electrons in the conduction band of g-C₃N₄ transfer directly to the conduction band of ZnO and subsequently to the surface of ZCNB [80]. On the other hand, biochar on the ZCNB composite surface can also accept electrons on the conduction band of ZnO or g-C₃N₄ [77, 78]. The photosensitive electrons on the catalyst surface react with O₂ dissolved in the solution and create •O₂⁻ (Eq. 14). These radicals continue to react with adsorbed H₂O on the surface of the catalyst or H⁺ ions present in the solution to form •OH (Eq. 15–18), and the resulting •OH decomposes DC [22] (Eq. 19). On the opposite side, since the VB edge potential of ZnO (2.69 eV) is more positive than the potential of the •OH/OH⁻ pair (1.99 eV) [81], the holes in the valence band cannot react with the OH⁻ ion to form active •OH, but they can instead directly oxidize DC molecules (Eq. 20 and 21) [23, 82]. Based on the literature [22, 82–84] and the aforementioned results, we propose a possible visible light-driven photocatalytic mechanism at the interface of the ZCNB heterojunction. Briefly, visible light excites g-C₃N₄ to form photoinduced electron/hole pairs. Then these photogenerated electrons transferred directly to the conduction band of ZnO and the surface of ZCNB lead the promotion of separation of photogenerated electron-hole pairs. These free electrons will react with oxygen to form •O₂⁻, followed by the reactions of these radicals with H₂O or H⁺ to form •OH. This process is illustrated as shown in Fig. 6.

In addion, we applied ZCNB composites to decompose other types of pollutants under visible light illumination. As presented in Fig. S1b, ZCNB exhibited an excellent degradation efficiency within 120 min for ciprofloxacin (98.2%), amoxicillin (90.3%), methylene blue (95.2%), rhodamine B (99.9%), and methyl orange (82.4%). These results show that composites have great potential for applications in treating organic pollutants in wastewater.

Inorganic anions can trap oxidants that play a prominent role in photocatalysis, reducing the decomposition ability of the photocatalytic treatment system. For this reason, this study incorporated 1.0 mM of ions commonly present in natural waters namely Cl⁻, PO₄³⁻, CO₃²⁻, NO₃⁻ and SO₄²⁻ into the reactor (anion/DC molar ratio of 20.5) as coexisting anions to show their impact on the DC degradation rate. The reactions were conducted under selected conditions with a pH of 5.0, a catalyst amount of 0.1 g, and 200 mL of DC solution at a concentration of 25 mg L⁻¹. The results are illustrated in Fig. S1c. It is observed that all these anions have the effect of reducing DC decomposition efficiency. Adsorption of these anions onto the active sites of the catalyst will reduce the available surface sites for DC adsorption, as will the reaction of the anions with oxidizing agents to produce oxygen species less strong oxidation, are the main reasons for the smaller DC removal efficiency in the presence of these anions [85]. Among the combined anions, phosphate and nitrate showed a slight decreasing effect on the degradation rate, reducing the removal efficiency from 98.9 to 90.1 and 93.5%, respectively. In contrast, the addition of carbonate, chloride and sulfate resulted in a significant decrease in DC degradation efficiency to 55.6, 62.5 and 73.9%, respectively. It seems that the stronger competition of carbonate, chloride and sulfate anions with DC for adsorption on the ZCNB surface and the scavenging of oxidative radicals by these ions is responsible for their destructive effects on the ZCNB surface. with effective DC removal [86]. The quenching effect of carbonate, chloride and sulfate anions can be described by equations (22)–(24).

For the purposes of practical applications, it is necessary to evaluate the long-term stability of the photocatalyst during the reaction. The ZCNB catalyst was reused four times. After each test, the catalyst was separated by using centrifugation; then, it was eluted several times with methanol [87] to ensure complete removal of residual DC, and finally was dried at 105°C for 24 h. The photocatalytic degradation efficiency of ZCNB decreases from 98.9 to 90.1% after six cycles (Fig. 7a). The XRD patterns of the ZCNB samples before and after the fourth test remain almost unchanged; therefore, it can be concluded that ZCNB is stable in the photocatalytic degradation reactions (Fig. 7b).

The shrimp aquaculture wastewater was collected from a shrimp aquaculture pool at Mekong delta. The characteristics of shrimp farming wastewater samples are shown in Table S6. It was used to test DC degradation over the present catalyst (Fig. 7c). Around 90% of DC is degraded after 120 minutes. In addition, ZCNB also showed a certain degradation effect on other pollution (Table S6). The TOC and COD removal efficiency reached values of 73.1% and 87.1%, respectively, showing that ZCNB is a promising material in practical wastewater treatment.