Graphitic carbon nitride (g-C₃N₄) as an efficient metal-free Fenton-like catalyst for degrading organic pollutants: the overlooked non-photocatalytic activity

Although the classical homogeneous Fenton reaction, in which highly aggressive hydroxyl radicals (HO·) are generated from H₂O₂ by the redox cycle of Fe²⁺/Fe³⁺, has been used for the degradation of recalcitrant organic pollutants, some drawbacks hamper the wider application of this process. For example, this process only works at acidic pH (2.5–3), and Fe²⁺ ions result in formation of iron sludge, which has to be removed and hence increases the operation cost. To overcome these disadvantages, much attention has been paid to heterogeneous Fenton-like systems in recent decades (Zhu et al. 2019). Naturally, the research mainly focuses on Fe-containing solids (Wu et al. 2015; Hammedi et al. 2019; Zhu et al. 2019), and then extends to various Cu-containing materials (Dong et al. 2015; Lyu et al. 2015; Zhang et al. 2017; Su et al. 2019). Other non-iron materials, such as Ag, Al, Ce, Co, Cr, Mn, Au, and Ru, have also been investigated (Bokare & Choi 2014). Many of these solid catalysts with metals as active species show fairly high catalytic activity; however, their imperfect stability is still a major barrier impeding their commercial applications (Zhu et al. 2019). In heterogeneous Fenton-like processes, the leaching of metal ions caused by acidity and complexation is almost inevitable, resulting in a rapid deactivation of these catalysts. To avoid this deactivation, metal-free materials, especially carbon materials, have received increasing interest in recent years (Ribeiro et al. 2016). Carbon materials studied as heterogeneous Fenton-like catalysts on their own include activated carbon, carbon xerogel, carbon black, glycerol-based carbon, biochar, carbon nanotube, carbon quantum dot, graphite, and graphene. The catalytic activity was generally attributed to the basic oxygen-containing functional groups of these carbon materials. These groups activated H₂O₂, generating HO· by redox cycles, similar to the redox cycle of Fe²⁺/Fe³⁺ in the classical Fenton process (Espinosa et al. 2015; Navalon et al. 2017; Vega & Valdés 2018). Among these carbon materials, graphene, a famous 2D carbon material, has attracted the most intensive attention, mainly due to its tremendous specific surface areas, high chemical stability, superior electron mobility, abundant oxygen-containing functional groups, and easy-to-modulate active sites. Graphene (Espinosa et al. 2015), graphene oxide (GO) (Ribeiro et al. 2016; Zhao et al. 2017), and reduced graphene oxide (r-GO) (Espinosa et al. 2015) were all studied as heterogeneous metal-free catalysts for Fenton-like processes. r-GO usually exhibited the highest catalytic activity, while GO exhibited the lowest (Espinosa et al. 2015). This is because the type and the number of surface oxygen-containing functional groups markedly influenced their catalytic performance. Basic oxygen-containing functional groups were reductive and might transfer electrons to H₂O₂, generating HO·. In contrast, oxidative oxygen-containing functional groups caused invalid decomposition of H₂O₂. This invalid decomposition resulted in a decrease not only in the utilization of H₂O₂ but also in the catalytic activity. To modulate the surface functional groups of carbon materials, doping heteroatoms is an effective strategy. Nitrogen is the most widely studied heteroatom in doped graphenes as well as other carbon materials (Kong et al. 2014). Doping nitrogen on carbon materials usually enhanced their Fenton-like catalytic activity (Martin-Martinez et al. 2016; Granzier-Nakajima et al. 2019).

Graphitic carbon nitride (g-C₃N₄) is another famous 2D carbon material, which has a similar structure to graphene but containing nitrogen. Mainly as a photocatalytic material, g-C₃N₄ has attracted increasing interest in recent years (Mamba & Mishra 2016; Ong et al. 2016; Kumar et al. 2018). g-C₃N₄ contains not only plenty of nitrogen-containing but also some oxygen-containing functional groups. This material is easy to synthesize on a large scale with various cheap raw materials, and also easy to modify by various strategies (Reddy et al. 2019). Curiously, it seems that this material has never been studied as a Fenton-like catalyst. In fact, in many studies dealing with g-C₃N₄ as photo-Fenton catalysts or as supports of Fenton-like catalysts, controlled experiments of g-C₃N₄ as Fenton-like catalysts were always carried out to illustrate the low activity of g-C₃N₄ (Table S1, Supplementary Information). In this study, we using rhodamine B (RhB) as a simulated organic pollutant to evaluate the Fenton-like catalytic performance of g-C₃N₄ over a wide pH range. We found that even the unmodified g-C₃N₄ synthesized by conventional urea pyrolysis was still an efficient heterogeneous Fenton-like catalyst.

Urea was purchased from Tianjin Fengchuan Chemical Reagent Co., Ltd (Tianjin, China). RhB was purchased from Sinopharm Chemical Reagent Co., Ltd (Shanghai, China). H₂O₂ (30%), HCl (36%), and NaOH were purchased from Shanghai Wokai Biotechnology Co., Ltd (Shanghai, China). All these reagents are of analytical grade.

g-C₃N₄ was synthesized by the conventional urea pyrolysis (Liu et al. 2011). Urea in a crucible with a cover was dried at 80 °C for 1 h. Then the crucible was put in a muffle furnace, heated to 580 °C with the heating rate of about 3 °C min⁻¹, and remained at 580 °C for 3 h to complete the reaction. The yellow product obtained was washed in turn with 0.1 mol L⁻¹ HCl, deionized water, and ethanol until the filtrate was neutral. Lastly, the wet sample was dried in a vacuum freeze dryer for 10 h.

X-ray diffraction (XRD) data were collected on a Bruker D8 Advance diffractometer with Cu K_α radiation. Fourier transform infrared (FTIR) spectra were obtained on a Bruker Tenser II infrared spectrophotometer using the KBr disc method. Scanning electron microscopy (SEM) images were recorded on a Hitachi S2400 scanning electron microscope. X-ray photoelectron spectroscopy (XPS) measurements were conducted on a Thermo Scientific-Escalb 250XI electron spectrometer. The specific surface area was obtained according to the Brunauer–Emmett–Teller (BET) method from the nitrogen adsorption at liquid nitrogen temperature using a Qutantachrome NOVA 2000e system.

To evaluate the catalytic performance of the g-C₃N₄, 100 mL RhB solution with a concentration of 20 mg L⁻¹ was used as simulated wastewater. The degradation of RhB was carried out in a glass reactor of 250 mL immersed in a constant temperature water bath with magnetic stirring. The initial solution pH was adjusted using a small amount of 0.1 mol L⁻¹ HCl or 0.1 mol L⁻¹ NaOH solution. After adding 20 mg g-C₃N₄ into the RhB solution, the mixture was stirred for 30 min to achieve adsorption–desorption equilibrium. Then the reaction was started by adding a certain amount of H₂O₂ solution. All reactions were carried out at 50 °C, pH 10, and 100 mmol L⁻¹ H₂O₂ unless noted otherwise. At predetermined time intervals, a 4 mL reaction solution was taken out. After immediate centrifugation, the concentration of RhB was analyzed at 554 nm using an INESA 722G spectrophotometer. The UV-vis absorption spectra of the reaction solution were recorded by using a Shimadzu UV-1780 spectrophotometer.

The electron paramagnetic resonance (EPR) spectra were recorded on an EPR spectrometer (EMX Plus, Bruker, Germany). 5,5-dimethyl-1-pyrroline N-oxide (DMPO) was used as a radical trapping agent.

XRD and FTIR were used to verify the crystal structure and surface characteristics of the synthesized g-C₃N₄ product. The results were in good accordance with the data reported previously (Figure 1) (Liu et al. 2011; Wang et al. 2019). The strong XRD peak at 27.5° was indicative of the graphitic stacking structure, which was indexed as (002). Another obvious peak at 13.1° was attributed to the in-planar repeating unit, which was indexed as (100). The FTIR spectrum of the product demonstrated the graphitic structures further. The wide band between 3,000 and 3,500 cm⁻¹ was attributed to N–H stretching vibration modes, indicating that uncondensed amino groups still existed in the product. The strong bands at 1,630, 1,569, 1,454, and 1,395 cm⁻¹ were assigned to stretching vibration modes of heptazine-derived repeating units. The sharp band at 808 cm⁻¹ represented the out-of-plane bending vibration of heptazine rings. The bands at 1,313 and 1,229 cm⁻¹ corresponded to stretching vibration of connected units of N–(C)₃ or C–NH–C. The SEM image of the synthesized product presented a sheet structure with a thin thickness, which was similar to the morphology of g-C₃N₄ synthesized using similar methods (Lu et al. 2016; Yue et al. 2016; Yang et al. 2017). The nitrogen adsorption–desorption isotherm of the synthesized product showed the type IV isotherm. The specific surface area was calculated according to BET theory as 77 m² g⁻¹, which was also near the values of g-C₃N₄ synthesized using similar methods (Liu et al. 2011; Sudhaik et al. 2018).

The initial pH value of wastewater usually strongly affects the degradation of organic pollutants in Fenton or Fenton-like processes, whether in homogeneous or heterogeneous systems. Catalysts usually have high activity only in a relatively narrow pH range. In most cases, the optimum pH range is weak acidity, and for some catalysts, the optimum pH can be extended to neutrality. Few catalysts have activity in the alkaline range. The Fenton-like catalytic activity of the g-C₃N₄ was also strongly influenced by the initial pH value, as shown in Figure 2. The g-C₃N₄ had the lowest catalytic activity at pH 4, and when pH was less than 4 the catalytic activity increased slightly. However, when the pH was greater than 4, the catalytic activity of the g-C₃N₄ obviously increased with the pH increase. That is to say, the catalytic activity of the g-C₃N₄ was the highest under alkaline conditions (around pH 10). To eliminate the influence of the non-catalytic degradation of H₂O₂, a controlled experiment was carried out in the same pH range. The results showed that without catalytic activation H₂O₂ itself is indeed a rather weak oxidant, although the non-catalytic degradation of RhB increased slightly with the increase of pH. For example, at pH 10, the non-catalytic degradation was only 27% in 140 min, and the catalytic degradation of RhB reached 92%. This clearly proved that the g-C₃N₄ had a high heterogeneous Fenton-like catalytic activity. On the other hand, we can get the rate constants by fitting the experimental data with the first-order kinetic equation. The rate constant was 0.00209 min⁻¹ for the non-catalytic degradation, but it was 0.0184 min⁻¹ for the catalytic degradation, which increased about nine times. This rate constant also showed that the Fenton-like catalytic activity of g-C₃N₄ was fairly good, especially when considering the dosage of g-C₃N₄ as low as 0.2 g L⁻¹.

It should be pointed out that our experiments showed that g-C₃N₄ had only lower Fenton catalytic activity under weak acidic to neutral condition. This condition was usually the optimized pH range for Fenton and photo-Fenton studies. Furthermore, the natural pH of dye solutions commonly used as pollutant models is usually weak acidic or neutral if the pH value is not adjusted. This may be one of the reasons for the low activity of g-C₃N₄ as Fenton catalysts observed in the controlled experiments reported in the literature (Table S1).

At room temperature, the Fenton catalytic activity of the g-C₃N₄ was undistinguished, which may be another reason for its neglect with regard to catalytic performance. As shown in Figure 3, at 30 °C the degradation of RhB was only about 40% in 140 min. Increasing the temperature can expectedly enhance the catalytic activity of the g-C₃N₄. At 60 °C the degradation of RhB could reach 96% in 140 min. At 70 °C the degradation in 20 min reached 83%. Correspondingly, the rate constant of the first-order reaction increased from 0.00300 min⁻¹ at 30 °C to 0.0706 min⁻¹ at 70 °C, which was about 23 times higher. The rate constants at different temperatures agreed well with the Arrhenius equation (Table S2, Figure S1, Supplementary Information), and the apparent activation energy of degradation calculated was 65.6 kJ mol⁻¹.

As oxidant, the concentration of H₂O₂ strongly affected the degradation of RhB (Figure 4, Table S3). When the concentration of H₂O₂ was 40 mmol L⁻¹, the degradation of RhB was only about 43.1% in 140 min, and when the concentration was 100 mmol L⁻¹, the degradation reached 93.4% in the same time. Correspondingly, the rate constant of the first-order reaction increased from 0.00380 to 0.0184 min⁻¹, which was about four times higher. Further increasing the concentration of H₂O₂ had no contribution to improving the degradation.

It can be calculated that the concentration of H₂O₂ should be 2.92 mmol L⁻¹ for mineralizing RhB of 20 mg L⁻¹ (corresponding to 0.04 mmol L⁻¹). Thus, an H₂O₂ concentration of 100 mmol L⁻¹ is 34 times the theoretical amount, indicating relatively lower utilization of H₂O₂ in this system. To achieve higher degradation at low H₂O₂ concentration, a longer time or higher temperature is needed. It also shows that g-C₃N₄ as a Fenton-like catalyst still should be improved.

As a metal-free material, it is expected that g-C₃N₄ exhibits better stability because there is no leaching problem of metal ions. Its stability was tested by separating the catalyst from the Fenton-like system after the reaction and repeatedly using it. No significant decrease in catalytic activity was observed after five cycles as shown in Figure 5, indicating that the g-C₃N₄ had good repeatability and long-term stability.

The reactive oxidizing species of HO· was confirmed by EPR, as shown in Figure 6. Using DMPO as the radical trapping agent, the characteristic EPR spectrum of the DMPO-OH spin adduct was recorded in the g-C₃N₄/H₂O₂ system. However, without g-C₃N₄ the spectrum was very weak. These results clearly demonstrated that g-C₃N₄ activated H₂O₂, generating HO·. That HO· was the major reactive oxidizing species was further supported by a scavenging experiment of free radicals, in which the addition of isopropanol, an HO· scavenger, significantly decreased the degradation of RhB (Figure 7). Highly active HO· can quickly destruct the structure of the xanthene-conjugated chromophore (554 nm) of the RhB molecule, as shown in the UV–vis spectra of the reaction solution (Figure 8) (Lee et al. 2016). The superoxide radicals and non-catalytic oxidation of H₂O₂ may also contribute to the degradation of the dye because the addition of isopropanol did not completely inhibit the degradation (Figure 7).

The reducibility of active sites was supported by a preliminary experiment. After oxidation by HNO₃, the g-C₃N₄ almost completely lost its Fenton-like catalytic activity (Figure 7). In order to identify the specific active group, we compared the structure of g-C₃N₄ before and after HNO₃ oxidation by XRD, FTIR, and XPS characterization. The results are shown in Figure 9, Figure S2, and Table S4. After HNO₃ oxidation, the main peak (002) of the g-C₃N₄ XRD pattern showed only a decrease of the intensity and a slight shift towards higher angles (Figure S2), reflecting a possible disorder introduced in the graphitic planes by the oxidation process and a reduction of the inter-planar stacking distance (Pisanu et al. 2018). Comparing the FTIR spectra of the g-C₃N₄ before and after HNO₃ oxidation, we cannot observe the structural change resulting from the oxidation (Figure S2). Although C=O and C–O groups are commonly formed in the process of g-C₃N₄ oxidation, it is difficult to detect them by the conventional FTIR technique, because these groups are usually present only on the surface of g-C₃N₄. The adsorption band of C=O vibration should appear at around 1,720 cm⁻¹; nevertheless there was no detectable absorption peak before and after the HNO₃ oxidation. The very weak peak at 1,080 cm⁻¹, which was attributed to C–N vibration, did not change observably after the oxidation. It is generally believed that the C–O bond and C–N bond have similar bond force constants; the weak C–O absorption can hardly cause the observable change of the 1,080 cm⁻¹ peak (Li et al. 2015; Yang et al. 2017; Pisanu et al. 2018).

XPS is a surface-sensitive technique that easily detects changes of the catalyst surface groups (Figure 9). The XPS spectrum of pristine g-C₃N₄ presented characteristics of the common g-C₃N₄. The survey scan of the XPS spectrum showed that the synthesized product was composed primarily of carbon and nitrogen. The very weak O1s peak may come from adsorbed oxygen, which was commonly observed in XPS spectra of g-C₃N₄. The high-resolution spectrum of C1s showed the deconvoluted two peaks at 288.0 and 284.8 eV. The major peak at 288.0 eV corresponded to sp2-bonded carbon (N–C=N), and the weaker one at 284.8 eV was attributed to graphitic carbon, which usually appears on the XPS characterization for g-C₃N₄. The high-resolution spectrum of N1s showed the deconvoluted three peaks at 398.6, 400.2, and 401.3 eV. The main peak at 398.6 eV could be attributed to sp2 hybrid nitrogen (C–N=C). The peak at 400.2 eV was ascribed to tertiary nitrogen (N–(C)₃). The peak at 401.3 eV corresponded to amino groups with a hydrogen atom (C–N–H). After HNO₃ oxidation, an obvious enhanced O1s peak appeared in the XPS survey scan of the oxidized g-C₃N₄, indicating the oxidation resulted in formation of oxygen-containing groups. The high-resolution spectrum of O1s showed the deconvoluted three peaks at 531.8, 532.7, and 533.8 eV, being attributed to oxygen of O=C, O–C and O–H, respectively. Accordingly, two new C1s peaks at 286.1 and 289.0 eV attributed to carbon of O–C and O=C–O, respectively, can be deconvoluted from the C1s high-resolution spectrum of the oxidized g-C₃N₄ (Li et al. 2015; Yang et al. 2017; Pisanu et al. 2018). However, after oxidation no new peaks appeared in the high-resolution spectrum of N1s, but the surface content of nitrogen decreased obviously and the N/C atomic ratio decreased from 1.32 (very close to the theoretical ratio 1.33) to 0.90 (Table S4), indicating that some of the nitrogen atoms were lost during the HNO₃ oxidation. Nevertheless, most of the nitrogen-containing groups remained, and the relative content did not change significantly.

Given the above discussion, it can be concluded that the oxidation process only introduced a certain amount of oxygen-containing groups but did not completely destroy the basic result of g-C₃N₄. This may indicate that the electron-rich triazine group was the active site for activating hydrogen peroxide. The HNO₃ oxidation, which introduced electron-withdrawing oxygen-containing groups, especially the carboxyl group, significantly changed the surface electron distribution of g-C₃N₄, resulting in catalyst deactivation. In addition, oxidative oxygen-containing functional groups may cause invalid decomposition of H₂O₂, resulting in a decrease not only in the utilization of H₂O₂ but also in the catalytic activity.

The photocatalytic performance of g-C₃N₄ has attracted extensive research, but its Fenton-like catalytic performance has been neglected. In this paper, it has been proved for the first time that g-C₃N₄ is an efficient and stable heterogeneous Fenton-like catalyst, and its highest activity occurs in the alkaline condition of about pH 10. Like other carbon materials, it may activate H₂O₂ at the Lewis base site to produce HO·, thus degrading organic pollutants. Of course, the study of the mechanism is insufficient, and more evidence is needed for the identification of active groups. Also, the reaction activity is not high at room temperature and the consumption of H₂O₂ is relatively high. However, because g-C₃N₄ can be modified by various strategies and this material is easy to synthesize on a large scale with various cheap raw materials, g-C₃N₄ should be a potential candidate as excellent heterogeneous metal-free Fenton-like catalysts.

We are grateful to the National Natural Science Foundation of China (21802125) and the Henan Science and Technology Project (172102210490) for the financial support.

There are no conflicts to declare.

The Supplementary Material for this paper is available online at https://dx.doi.org/10.2166/wst.2020.129.