The TEM image of sample CN (Fig. 2a) shows the layered and flattened crystallites of g-C3N4 with irregular morphology produced through thermal polycondensation. The inset circle of Fig. 2a shows an HRTEM micrograph that highlights an in-plane arrangement, typical of g-C3N4 platelets. The XRD pattern (bottom of Fig. 2a) exhibits the (100) and (002) peaks, respectively at 13.1° and 27.4°, confirming that the graphitic stacking structure of g-C3N4 was formed through polycondensation of melamine and urea (Sunasee et al. 2019).
The TEM images depicted in Figs. 2b to 2e exhibit a uniform distribution of iron oxide magnetic NPs decorating the g-C3N4 sheets in the nanocomposite samples. In Fig. 2b, nearly spherical single-core Fe3O4 NPs (magnified in Fig. 2c) can be observed (sample S-CN). Figure 2d shows the formation of nanoflower-like Fe3O4 nanostructures (sample NF-CN). As expected, these nanoflowers are multicore larger structures formed by stacked smaller spherical Fe3O4 nanostructures (Gavilán et al. 2017), as shown in the inset of Fig. 2d and magnified in Fig. 2e.
Table 1 collects the values of average diameter (dTEM) and polydispersity index (s) of NPs of samples S-CN and NF-CN calculated from the size-distribution curves (Fig. 2f).
After decoration with magnetic NPs, nanocomposites preserved the typical diffraction peaks of the g-C3N4 phase. Also, the extra peaks relative to the spinel phase (indexed and shown in Fig. 2g) confirmed the presence of the Fe3O4 NPs in the graphitic g-C3N4 template structure. Scherrer's formula was utilized to determine the crystallite size of the incorporated magnetite NPs (dXRD) considering the (311) diffraction peak (Table 1). In the case of sample S-CN, the mean diameter agrees well with TEM results. However, the Fe3O4 NPs of sample NF-CN present a mean size significantly smaller than that obtained from TEM. This discrepancy is due to the fact that in nanoflowers the XRD size is related to the individual spherical NPs stacked in the flower-like structure, as shown in Fig. 2f (Gavilán et al. 2017).
Table 1
Parameters of samples characterization: mean characteristic sizes (dTEM, dXRD), polydispersity index (sTEM), BET specific surface area (SBET), pore volume (PV), average pore-size diameter (PD), band gap energy, and saturation magnetization (MS).
Sample | dTEM (nm) | sTEM | dXRD (nm) | MS (emu/g) | band gap (eV) | SBET (m2/g) | PV (cm3/g) | PD (nm) |
CN | - | - | - | - | 2.53 | 15.0 | 0.19 | 33.1 |
S-CN | 14.3 | 0.38 | 13.8 | 14.3 | 1.99 | 52.7 | 0.26 | 19.1 |
NF-CN | 180.3 | 0.44 | 9.0 | 21.9 | 1.85 | 96.6 | 0.31 | 13.8 |
The magnetic behavior of g-C3N4 samples decorated with magnetite NPs is shown in Fig. 2h. Both samples exhibit a superparamagnetic-like behavior with almost no coercivity or remanence, as seen from the hysteresis loops in low-field in the inset. The saturation magnetization (MS) of sample NF-CN is higher than that of sample S-CN and, which can be correlated to the size of the magnetic structures in the nanocomposites (Table 1). Indeed, since the mass fraction of Fe3O4 determined from ICP-OES is almost the same in both nanocomposites (25% in S-CN and 27% in NF-CN), the higher saturation magnetization of sample NF-CN arises from the larger Fe3O4 nanoflower structures (Gavilán et al. 2017) compared to the single-core smaller magnetite nanoparticles (Sousa et al. 2005) of sample S-CN. Nonetheless, using a permanent magnet, both decorated samples can be easily separated or manipulated from the liquid media (Fig. 2h), which improve their performance in pollutant removal.
Figure 3. PL spectra (3a), N2 adsorption/desorption isotherms (3b), FTIR spectra (3c), and pH-dependence of zeta potential of the samples (3d).
The nanocomposites and pristine g-C3N4 exhibit the expected optical properties of semiconductors. Sample CN shows a maximum PL peak at approximately 410 nm with a low absorption in the red-light range, which is characteristic of carbon nitride synthesized through thermal polycondensation (Zhan et al. 2017). The incorporation of iron oxide nanoparticles causes a redshift in the onset of optical absorption, resulting in a maximum PL peak at > 450 nm, suggesting an improved visible absorption of the nanocomposites. The band gaps calculated from absorption edges, as shown in Figure S1 in Supplementary Material, were around 2.53 eV, 1.99 eV, and 1.85 eV, for samples CN, S-CN and NF-CN, respectively. These results demonstrate that the incorporation of iron oxide NPs enhances the visible-light response-ability as reported elsewhere (Irfan et al. 2019).
Figure 3b shows the N2 adsorption/desorption isotherms for all samples. It can be observed a typical IUPAC type IV pattern with h3 hysteresis loops, characteristic of mesoporous structure in the form of parallel plates (Fuentez-Torres et al. 2021). The values of BET surface area (SBET), cumulative pore volume (PV) and pore size (PD) listed in Table 1 indicate the modification of the g-C3N4 template with Fe3O4 NPs sharply enhances the surface area of the nanocomposites.
The characteristic vibration modes assigned to the g-C3N4 phase can be observed in the pristine and nanocomposite samples, as seen in the FTIR spectra of Fig. 3c (Hu et al. 2014; Zhong et al. 2020). The bands observed between 800 and 900 cm− 1 can be attributed to tri-s-triazine, suggesting that the skeletal structure of the g-C3N4 remained unaltered in decorated samples. The characteristic vibrations of aromatic CN heterocycles found in g-C3N4 can be identified as strong bands ranging from 1700 to 1200 cm− 1. The bands observed between 3400 and 2900 cm− 1 can be assigned to N-H residual components, which corresponds to amines (primary and secondary). The nanocomposite samples exhibit additional bands related to the presence of iron oxide nanoparticles (Rubim et al. 2001). The band at 580 cm− 1 can be attributed to Fe − O vibrations in the tetrahedral sites of the spinel crystalline lattice of Fe3O4 nanoparticles. This finding confirms that the Fe3O4 nanoparticles were successfully integrated into the g-C3N4 structure.
Figure 3d exhibits the variation of the zeta potential of the samples under the different pH conditions utilized to study the adsorption-photocatalytic efficiency of the nanocomposites. For all samples, the zeta potential becomes more negative with increasing pH. In the case of CN, this behavior may be due to the substitution of ≡ C-NH2 chemical groups on the g-C3N4 surface by ≡ C-OH groups, which undergo increasing deprotonation (≡ C-O−) with pH (Zhu et al. 2015). The zeta potential of the nanocomposite samples showed a similar pH-dependent behavior, with higher absolute values of zeta potential for pH > 7, around − 30 mV. This phenomenon may be associated with the presence of magnetite nanoparticles on the g-C3N4 surface, whose hydroxyl groups (≡ FeOH) are mostly deprotonated (≡ FeO−) in strongly alkaline medium (Palomec-Garfias et al. 2018).
The kinetics of MB adsorption during the dark stage for different pH conditions is depicted in Fig. 4. As a general trend, the adsorption rate is very fast in the first minutes, then gradually decreases reaching the equilibrium within 70 min. The adjust of the kinetic results with the PSO model is shown in Fig. S2 in Supplementary Material, and Table S1 in Supplementary Material collects the obtained adjusted parameters. Most of the RW values lie in the range of 0.1 to 0.01 for all samples, indicating a highly curved adsorption curve and well-approaching equilibrium (Wu et al. 2009). Based on the pseudo-second-order rate index (k2qe), in acidic medium the MB adsorption is faster for sample CN while in neutral and alkaline media the kinetic performance is higher in the case of the nanocomposites.
To better compare the removal results obtained, a heat map with the percentage of adsorption, photodegradation and total MB removal of all nanocomposites under different pH conditions was plotted in Fig. 5, where the darker blue bars indicate the highest MB removal percentages. Concerning the equilibrium adsorption capacity, all samples exhibit a higher MB adsorption in alkaline medium (Fig. 5). This behavior may be attributed to zeta potential values, which are more negative at this pH condition, favoring electrostatic attraction between the surface sites of the nanomaterials and cationic MB species, then higher adsorption. It is noteworthy that the nanocomposites present higher adsorption capacities than the pure g-C3N4 in all pH conditions, due to their larger specific surface area and, therefore higher number of active sorption sites.
Unlike the adsorption capacity, incorporating the magnetic phase did not improve the activity of g-C3N4 towards MB photodegradation. Indeed, the photocatalytic efficiency of pure g-C3N4 is larger than that of nanocomposites under all pH conditions (Fig. 5). Also, the higher values of the rate constants obtained from the PFO model (Fig. S2) indicate that the MB photodegradation is faster for pure g-C3N4, no matter the pH of the medium (Table S1). This lower performance of the nanocomposites may result from covering a fraction of active sites of the g-C3N4 surface by the magnetic nanoparticles, which blocks visible light absorption, thereby decreasing photodegradation. It should be noted that MB molecules suffer photoinduced self-degradation, which is negligible in low pH, but sharply increases towards high pH conditions, as already reported (Soltani and Entezari 2013; Xu et al. 2015).
The performance of the samples on solutions discoloration (total MB removal) under the three pH conditions can be evaluated from the association of adsorption and photocatalytic capacities (Fig. 5). In acidic medium, the pure g-C3N4 is more efficient and can remove approximately 95% of MB molecules mainly thanks to its high photocatalytic efficiency. In alkaline medium, there is no significant difference in performance among the three nanomaterials, where the removal capacity is around 90%. However, in neutral medium, NF-CN is considerably more efficient (80.7%) than the others, which is related to its enhanced adsorption capacity in this pH region. Indeed, the decoration of the g-C3N4 with Fe3O4 nanoflowers endows the nanocomposite with more negative zeta potential and larger surface area, remarkably improving its adsorption capacity.
Depending on the photocatalyst concentration, irradiation time and lamp power, some g-C3N4-based photocatalysts can remove up to 100% of MB from water in low and high pH conditions, similar to the nanomaterials proposed here (Hassani et al. 2018; Paul et al. 2019; Sahoo et al. 2020; Ashrafi et al. 2020; Nguyen et al. 2021; Rosa et al. 2021; Waheed et al. 2023). However, sample NF-CN reaches a high adsorption capacity even in neutral medium, which is highly desirable for real-life applications in the dyeing industry (Global Effluent Guidelines requirements, 2007).