Photocatalytic degradation and transformation of pharmaceuticals using exfoliated metal-free g-C3N4.

Article iScience Photocatalytic degradation and transformation of pharmaceuticals using exfoliated metal-free g-C 3 N 4 Graphical abstract Highlights •Photocatalysis enables 80–98% degradation of selected pharmaceuticals within 120 min •g-C 3 N 4 thermally exfoliated for 2 h is an optimal photocatalyst under visible light •Degradation products and transformation pathways are identified by mass spectrometry Authors Petr Praus, Anna Gavlova ´, Jan Hrba ´� c, Kristina Schmidtova ´, Petr Bedna ´� r Correspondence [email protected] In brief Drugs; Chemical reaction; Catalysis; Environmental Chemical Engineering Praus et al., 2025, iScience 28, 113899 December 19, 2025 ©2025 The Author(s). Published by Elsevier Inc. https://doi.org/10.1016/j.isci.2025.113899 ll iScience Article Photocatalytic degradation and transformation of pharmaceuticals using exfoliated metal-free g-C 3 N 4 Petr Praus, 1,2,6, *Anna Gavlova ´, 1,3 Jan Hrba ´� c, 3,4 Kristina Schmidtova ´, 5 and Petr Bedna ´� r 3 1 Institute of Environmental Technology, CEET, VSB-Technical University of Ostrava, 17. Listopadu 15, Ostrava 70800, Czech Republic 2 Department of Chemistry, Faculty of Science, University of Ostrava, 30. Dubna 22, Ostrava 70103, Czech Republic 3 Department of Analytical Chemistry, Faculty of Science, Palacky University, 17. Listopadu 12, Olomouc 77900, Czech Republic 4 Institute of Chemistry, Faculty of Science, Masaryk University, Kamenice 5, Brno 62000, Czech Republic 5 Department of Chemistry and Physico-Chemical Processes, Faculty of Materials Science and Technology, VSB-Technical University of Ostrava, 17. Listopadu 15, Ostrava 70800, Czech Republic 6 Lead contact *Correspondence: petr.prau[email protected] https://doi.org/10.1016/j.isci.2025.113899 SUMMARY Pharmaceuticals are micropollutants of global concern that contribute to environmental contamination alongside other anthropogenic and natural chemical compounds. This study addresses the photocatalytic degradation of model pharmaceutical compounds ofloxacin, diclofenac, and caffeine using bulk and thermally exfoliated graphitic carbon nitride (g-C 3 N 4 ). Bulk g-C 3 N 4 was synthesized from dicyandiamide at 550◦C and exfoliated at 500 ◦C for 1-3 h in an ambient atmosphere. The structural, textural, and electronic properties of the prepared materials were evaluated. Graphitic carbon nitride exfoliated for 2 h provided the best photocatalytic degradation efficiencies (>95%) for both ofloxacin and diclofenac and approximately 80% for caffeine, determined for 120 min under irradiation at 420 nm. The pharmaceuticals were degraded, and their intermediate degradation products were investigated using liquid chromatography combined with high-resolution tandem mass spectrometry. The successful identification of the main degradation products allowed us to propose transformation pathways for the studied pharmaceuticals. INTRODUCTION Pharmaceuticals of both anthropogenic and natural origins contribute to the growing issue of micropollutant contamination worldwide. These substances can enter aquatic systems through various means, including the excretion of non-metabolized drugs by humans and animals, improper disposal of unused medications, and industrial and agricultural runoff. Urban wastewater treatment plants are significant point sources, as they often fail to completely remove these compounds during treatment processes. Pharmaceuticals have been found in surface water, groundwater, soil, and sediments. In addition to the complexity of these samples, they often exist in trace concentrations at the level of nanogrammes or micrograms per liter or kilogram in liquid or solid samples of aquatic systems, which makes their analytical determination difficult. Traditional wastewater treatment based on the activated sludge process is only partially effective in removing pharmaceuticals. 1,2 In principle, adsorption and separation processes retain pharmaceuticals; however, they do not address the ultimate disposal of the separated organic compounds. In contrast, photocatalysis enables the removal of pharmaceuticals from aqueous environments by their degradation. 3 Photocatalysis is part of advanced oxidation processes (AOPs) that can be used to treat organic pollutants in water. AOPs are usually performed using ozone, hydrogen peroxide, persulphate, peroxymonosulfate, ozone, sonolysis, and iron salts in the Fenton process. 4 Photocatalysis has also been used for various reactions, such as the reduction of CO 2 , 5 degradation of dyes 6 and other organic compounds, 7 hydrogen evolution by water splitting 8 or adding NaBH 4 , 9 fixation of nitrogen, 10 reduction of Cr(VI), 11 synthesis of hydrogen peroxide, 12 water disinfection, 13 and gene removal. 14 The application of Fe-MOF-based composites is also a promising strategy for the photocatalytic removal of environmental pollutants. 15 Most photocatalytic applications are based on TiO 2 , but graphitic carbon nitride has also been found to be a suitable photocatalyst for the degradation of various organic compounds, including pharmaceuticals. This material has been investigated over the last decade because it can be activated by visible irradiation owing to its narrow band gap (2.7 eV) and diamond-like physicochemical properties, such as mechanical, thermal, and chemical stability. 16 The popularity of g-C 3 N 4 also lies in its simple preparation via thermal synthesis iScience 28, 113899, December 19, 2025 © 2025 The Author(s). Published by Elsevier Inc. 1 This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). ll OPEN ACCESS from nitrogen-rich organic precursors, such as cyanamide, dicyandiamide, and melamine, in relatively high quantities at a low cost. However, the low specific surface area and fast recombination of photoinduced electron - hole pairs, resulting in low quantum efficiency, are drawbacks of g-C 3 N 4 . The low specific surface area of pre-prepared (bulk) g-C 3 N 4 can be increased by exfoliation using various methods, such as chemical oxidation with K 2 Cr 2 O 7 , concentrated sulfuric acid, organic liquids, sonication, ion intercalation, and thermal exfoliation. Thermal exfoliation, notable for its simplicity, ease of implementation, and environmental friendliness, is often used. In addition, the physico-chemical properties of g-C 3 N 4 can be fine-tuned by coupling with anions, metal cations, metal nanoparticles such as Au and Au/Pd, Ag, Cu, Cu/Co, or semiconductor particles such as TiO 2 , ZnO, FeS 2 , Al 2 O 3 , SnO 2 , WO 3 , Ag 3 VO 4 , BiVO 4 , BiOBr, BiIO 4 , and Cu 3 V 2 O 8 . Another approach is doping with metal and non-metal elements, such as O, S, and P. This study aimed to investigate the photocatalytic degradation of pharmaceuticals, including their transformation products. For this purpose, bulk and thermally exfoliated g-C 3 N 4 samples were used in this study. This is a simple, one-component, metal-free, and low-cost photocatalyst with potential application in water treatment technology. Moreover, its photocatalytic stability during the degradation of organic substances has been reported in the literature. 17 The photocatalysts were synthesized from dicyandiamide and characterized by elemental analysis, electron microscopy, X-ray diffraction, physisorption of nitrogen, and infrared, UV-Vis, photoluminescence, and electron paramagnetic resonance spectroscopies. In photocatalytic experiments, the common and widespread pharmaceuticals ofloxacin, diclofenac, and caffeine, selected as model compounds, were degraded under visible light irradiation at 420 nm. The degradation rates and efficiencies were evaluated, and degradation pathways were suggested based on the identification of intermediary degradation products using liquid chromatography combined with high-resolution tandem mass spectrometry. The novelty of this study lies in the investigation of the transformation pathways of selected pharmaceuticals, such as ofloxacin, diclofenac, and caffeine. Such transformation investigations are important for possible water treatment applications, but are often omitted from reported studies. An important feature is that pure g-C 3 N 4 can be an effective and suitable photocatalyst, which can be simply tuned through thermal exfoliation to achieve results comparable to those of other complex multi-component photocatalytic systems. Moreover, unlike other photocatalysts, such as TiO 2 , g-C 3 N 4 operates under visible light irradiation. RESULTS AND DISCUSSION The g-C 3 N 4 bulk and exfoliated materials were synthesized from dicyandiamide in air, and their physico-chemical properties were extensively characterized. The photocatalytic properties were studied by the degradation of selected pharmaceuticals (ofloxacin, diclofenac, and caffeine), and the degradation products were analyzed by liquid chromatography. Transformation pathways were proposed for all three pharmaceuticals studied. Elemental analysis The synthesized bulk and exfoliated g-C 3 N 4 samples were analyzed to determine their elemental compositions (Table 1). The C, H, and N contents were determined directly, while the O content was calculated by difference to reach a total of 100%. The C and N contents were similar in the exfoliated g-C 3 N 4 samples but were generally lower than those in bulk g-C 3 N 4 because these elements were released during the exfoliation process, indicating the formation of both nitrogen and carbon defects (vacancies). The oxygen and hydrogen contents were higher in the exfoliated materials owing to their partial oxidation in air and the formation of -OH groups, as shown in the Fourier transform infrared (FTIR) spectra (Figure 4). Characterization by electron microscopy The morphologies of the g-C 3 N 4 materials were investigated by scanning electron microscopy (SEM). Micrographs of the Bulk and TEX3 samples (samples with the lowest and highest surface areas, respectively) are presented in Figure 1. Their morphologies were very similar, that is, large particles composed of aggregated flakes. The results of the SEM-EDS analysis are summarized in Table S1. The nitrogen, carbon, and oxygen contents of g-C 3 N 4 , as well as the C/N ratios, do not correspond with the results of the elemental analysis given in Table 1. The EDS results are not realistic for several reasons, namely a small amount of the analyzed materials, the impossibility to determine hydrogen, and a low sensitivity of EDS to light elements (C, N). The data can also be distorted due to the use of carbon tape to affix the powder samples. Therefore, we consider the results of conventional elemental analysis, which uses at least 0.5 g of each sample, to be reliable and realistic. However, a useful piece of information from the SEM-EDS is that the synthesized g-C 3 N 4 materials were not contaminated by other elements. The transmission electron microscopy (TEM) micrographs are shown in Figure 1. The presence of worm-like structures is typical of exfoliated g-C 3 N 4 materials. One can also see large flat flakes with different thicknesses and agglomerations. The worm-like structures were probably formed by the wrapping and deformation of flat flakes due to prolonged exposure to high temperatures. Characterization by X-Ray diffraction The structures of the synthesized bulk and exfoliated g-C 3 N 4 were studied using X-ray diffraction (XRD) (Figure 2). Two typical main diffraction peaks, (002) and (100), were observed (JCPDS 87–1526, supplemental information). The (002) and (100) diffractions can be ascribed to the interlayer stacking of the g-C 3 N 4 planes and the in-plane ordering of the nitrogen-linked heptazine Table 1. Elemental composition of bulk and exfoliated g-C 3 N 4 Material C (wt. %) H (wt. %) N (wt. %) C/N O (wt. %) Bulk 35.11 1.53 62.40 0.563 0.96 TEX1 34.35 1.85 61.40 0.559 2.40 TEX2 34.45 1.80 61.73 0.558 2.02 TEX3 34.25 1.83 61.33 0.558 2.59 2 iScience 28, 113899, December 19, 2025 iScience Article ll OPEN ACCESS units, respectively. The selected characteristics of the diffraction peaks are listed in Table 2. The d(002) spacing of bulk g-C 3 N 4 decreased slightly after exfoliation, while the crystallite size L(002) increased. These structural changes can be explained by the formation of expanded crystallites owing to the thermal treatment of bulk g-C 3 N 4 . Characterization by X-Ray photoelectron spectroscopy X-ray photoelectron spectroscopy (XPS) was performed to investigate the surface composition of graphitic carbon nitride materials, specifically focusing on the bulk and TEX3 samples. The elemental compositions are summarized in Table S2. The C 1s spectra, shown in Figure 3, were deconvoluted into two main peaks with comparable binding energies: 285.6 and 288.4 eV for bulk and 285.8 and 288.3 eV for TEX3. These peaks comprised overlapping contributions from C–C (284.8 eV), C–N (∼286 eV), and C–O (286–287 eV) bonds. The peaks at 288.4 and 288.3 eV can be attributed to sp 2 -hybridized carbon atoms in N–C=N environments. 18 The contributions from C–O and C–C bonding may arise from oxidative alterations and/or surface contamination by adventitious carbon. The C 1s spectra of the Bulk and TEX3 films exhibited differences, indicating a lower surface C content in TEX3 (Table S2) owing to the release of CO 2 during thermal exfoliation. The N 1s spectra (Figure 3) were fitted with four components located at 398.8, 400.0, 401.4, and 404.2 eV. The peaks at 398.8 eV and 400.0 eV correspond to sp 2 -hybridized nitrogen Figure 1. Characterization by electron microscopy SEM (BSE+SE) micrographs of (A) Bulk and (B) TEX3 graphitic carbon nitride. TEM micrographs of (C) Bulk and (D) TEX3 graphitic carbon nitride. SEM analysis: the acceleration voltage was 30 keV. The samples were gold-sputtered before analysis to ensure adequate electron conductivity. TEM analysis: an accelerating voltage of 200 kV was applied. Ethanol dispersions were placed on copper grids with holey carbon films. (N 2 C) and tertiary nitrogen (N 3 C), respectively, which are commonly associated with triazine units and amine-type linkages. 19 The component at 401.4 eV is indicative of N atoms involved in C–N–H bonding configurations. 19 Peaks at 404.2 and 404.3 eV may originate from π–π* (HOMO–LUMO) transitions 18 or, alternatively, from surface charging effects. 19 No significant spectral differences in the N 1s region were observed between the Bulk and TEX3. In the O 1s spectra (Figure 3), the distinction between C–O and N–O, or C=O and N=O bonds, could not be reliably resolved. The broad peak was deconvoluted into two components: one at 532.4 eV, associated with C=O (or N=O) species, and another at 533.7 and 533.8 eV, attributed to C–O (or N–O) bonding environments. Because carbon and nitrogen have similar electronegativity values, the oxygen-binding site cannot be conclusively identified. The lower O surface content in TEX3 can also be explained by the release of CO 2 . Textural properties by nitrogen physisorption The textural properties of the bulk and exfoliated g-C 3 N 4 samples were investigated through the physisorption of nitrogen. The adsorption–desorption isotherms are presented in Figure 4. All isotherms exhibited hysteresis loops, indicating the presence of mesopores in the materials. The pore size distributions were calculated according to the BJH model applied to the adsorption branch of the nitrogen adsorption-desorption isotherm. From the distribution curves, it is evident that mesopores dominate in the exfoliated samples (Figure 4). It can be concluded that the exfoliation of bulk g-C 3 N 4 is accompanied by mesopore formation, leading to an increase in the specific surface area. The specific surface area calculated using the BET method, surface area of the micropores, and pore volumes for the bulk and exfoliated g-C 3 N 4 samples are summarized in Table 3. The S BET increased with exfoliation time. A higher specific surface area of the catalyst provides more active sites for the adsorption of reactant molecules. The pore structure of semiconductors affects their photocatalytic activity owing to the larger contact interface and mass transfer. 20 Moreover, there is a quantitative iScience 28, 113899, December 19, 2025 3 iScience Article ll OPEN ACCESS nonlinear relationship showing that the overall reaction rate increases with an increase in S BET and the lifetime. 21 Characterization by Fourier transform infrared spectroscopy Fourier transform infrared (FTIR) spectra of bulk and exfoliated gC 3 N 4 are demonstrated in Figure 4. Typical vibration bands for the N-H (3300-3000 cm −1 ), C-N, and C=N (1600-800 cm −1 ) groups are prominent. 22 The medium band at 809 cm −1 was attributed to the breathing mode of the triazine units, and the spectral bands around 3400 cm −1 were attributed to O-H stretching vibrations. Their origin can be explained by the adsorbed water and the presence of -OH groups in the g-C 3 N 4 structures. The formation of -OH groups was also indicated by the higher oxygen and hydrogen content in the exfoliated materials (Table 1). Characterization by UV-Vis spectroscopy The UV-Vis absorption spectra of the samples are shown in Figure 4. The spectra contain a single broad band peaking at ca 400 nm and are similar for both bulk and exfoliated g-C 3 N 4 . The band-gap energies were calculated according to Tauc’s approach and are 2.74, 2.80, 2.80, and 2.82 eV for Bulk, TEX1, TEX2, and TEX3, respectively. For further details, see Figure S1 in the supplemental information. The slight increase in the bandgap energies (blue shift) of the exfoliated materials can be explained by the quantum confinement effect due to the decrease in the g-C 3 N 4 particle size in terms of cracking of large g-C 3 N 4 bulk structures into smaller nanosheets. 23,24 Table 2. XRD characteristics of bulk and exfoliated g-C 3 N 4 Material 2θ (deg) FWHM (deg) L(002) (nm) d(002) (nm) Bulk 27.33 1.30 6.11 0.326 TEX1 27.73 1.08 7.36 0.321 TEX2 27.79 1.15 6.91 0.321 TEX3 27.72 1.10 7.22 0.322 Note: The 2θ values correspond to the (002) diffraction. Figure 2. Characterization by X-ray diffraction XRD patterns of bulk and exfoliated g-C 3 N 4 using a Cu tube operated at 40 kV and 30 mA. Characterization by photoluminescence spectroscopy Photoluminescence (PL) spectra (Figure 5) were recorded to understand the changes in the electron-hole recombination processes within the g-C 3 N 4 materials following thermal exfoliation. The broad PL band with a maximum at approximately 481 nm corresponds to the σ*-LP and π*-LP transmissions (LP denotes the nitrogen lone electron pair), and is predominantly attributed to the π–π* electron transitions observed in bulk g-C 3 N 4 . 23,25 There is also a broad band between 500 and 550 nm, which is attributed to electron transitions between defect levels and the LP, and electron transitions between defect levels and levels related to oxygen atoms in g-C 3 N 4 , respectively. 26 As shown in Figure 5, the maximum PL intensity first increased in the order Bulk < TEX1 < TEX2, for which the maximum PL intensity was observed and discussed later in discussion, followed by a decrease in PL intensity from TEX2 to TEX3. This can be explained by non-radiative electron transitions due to the formation of g-C 3 N 4 structural defects. Together with the PL intensity changes, the emission bands were blue-shifted from 481 nm (Bulk) to 465 nm (TEX1) and to 461 nm (TEX2). However, the differences between TEX2 and TEX3 were minimal. The blue shift is explained in the previous section. PL decay curves were measured to calculate the lifetimes of the photoinduced electrons and holes, and were found to follow a triple-exponential function; the results are presented in Table 4and Figure S2. The PL lifetimes were consistent with the PL intensity, considering the first dominant PL decay component. The second ‘‘stronger’’ component also agrees with the PL intensity. The third component was similar for all materials and had only approximately 5–7% of the total amplitude owing to the relatively high lifetimes (∼20 ns). Therefore, the average lifetimes (τ avg2 ) were calculated for the two main components. The lifetimes of the three components (τ avg3 ) were calculated for comparison (Table 4). In general, the longer lifetimes of exfoliated g-C 3 N 4 are indicative of more efficient separation of photoinduced electrons and holes, thereby increasing the likelihood of their participation in photocatalytic reactions. The charge separation has been confirmed by electrochemical measurements. 17,27 The increased PL intensities and prolonged lifetimes of TEX2 and TEX3 can be attributed to the higher number of photoinduced electrons and holes, which is in agreement with the results of the photocatalytic and EPR experiments discussed later in discussion. Considering the highest PL intensities and long PL lifetimes of TEX2 and TEX3, it can be concluded that these materials are promising photocatalysts for the degradation of pharmaceuticals. 4 iScience 28, 113899, December 19, 2025 iScience Article ll OPEN ACCESS The role of vacancies in g-C 3 N 4 is important in terms of its optical and electronic properties. 28 The electronic properties are closely related to the photocatalytic activity. Defect engineering often leads to alterations in the conduction and valence band positions, enhancing the photocatalytic performance. Incorporating various vacancy types into g-C 3 N 4 has been found to promote electron accumulation and facilitate surface charge transfer, creating active sites for electrochemical reactions. 29 This Figure 3. Characterization by X-ray photoelectron spectroscopy XPS spectra of (A–C) Bulk and (D–F) TEX3 g-C 3 N 4 using Mg Kα radiation (hν = 1253.6 eV) generated at 12 kV and 10 mA. Figure 4. Textural and spectral characterization (A) Nitrogen adsorption and desorption isotherms at 77 K for bulk and exfoliated g-C 3 N 4 samples. (B) The nitrogen adsorption–desorption data were processed according to the classical BET theory (for p/p 0 ≈ 0.05–0.25). Distribution of pore sizes calculated using the BJH model. (C) FTIR spectra of bulk and exfoliated g-C 3 N 4 with a resolution of 2 cm −1 . The KBr method was employed. (D) UV-Vis absorption spectra of bulk and exfoliated g-C 3 N 4 recorded using UV-Vis diffuse reflectance spectroscopy. iScience 28, 113899, December 19, 2025 5 iScience Article ll OPEN ACCESS defect-induced enhancement contributes to the improved photocatalytic efficiency of g-C 3 N 4 . The formation of complex vacancies in bulk and exfoliated g-C 3 N 4 has been reported. 30 Reactive oxygen species and their study by electron paramagnetic resonance spectroscopy Photocatalysis generally involves the use of a semiconductor photocatalyst that absorbs light energy and generates electron-hole pairs. Photocatalytic degradation is based on reactions involving photoinduced electrons and holes that participate in redox reactions. As photocatalytic reactions occur in the presence of oxygen and water, the photocatalytic process is accompanied by the formation of reactive oxygen species (ROS), such as the superoxide anion radical (⋅O 2− ), its protonated form, hydroperoxyl radical (HO 2⋅ ), hydrogen peroxide, singlet oxygen ( 1 O 2 ), and hydroxyl radical (⋅OH). 31 The main reactions leading to their formation and transformations are given later in discussion (Equations 1, 2, 3, 4, 5, and 6): g-C3N4+hυ → e−(g-C3N4)+h+(g-C3N4)(Equation 1) e−+O2→O⋅− 2(Equation 2) O⋅− 2+H+⇆HO⋅ 2(pK =4:8)(Equation 3) 2 HO⋅ 2→H2O2+O2(Equation 4) H2O2+e−→OH⋅+OH−(Equation 5) H2O2+hv→2 OH⋅(Equation 6) In addition to ROS, holes can also react with organic substances as oxidising species 32,33 ; therefore, the overall photocatalytic process can be written as follows: ROS +h++organic substances →degradation products +CO2+H2O (Equation 7) The important role of hydrogen peroxide in the photocatalytic degradation of organic compounds has been discussed in the literature. 34,35 As hydrogen peroxide is a stable substance, the stability of g-C 3 N 4 was tested by treating TEX2 with 10–30% H 2 O 2 solutions for 5 h, followed by structural (FTIR, XRD, and XPS) analyses, as shown in Figures S3–S5. No g-C 3 N 4 instability was observed. This is in line with the findings of many studies, which have reported the photocatalytic stability of gC 3 N 4 used for the degradation of organic substances, for example. 36 The generation of superoxide and hydroxyl radicals upon the irradiation of the g-C 3 N 4 samples was studied by spin trapping using 3,4-dihydro-2,3-dimethyl-2H-pyrrole 1-oxide (DMPO). DMPO forms a stable spin adduct with hydroxyl radicals (DMPO-OH) featuring a four-line electron paramagnetic resonance (EPR) spectrum with 1:2:2:1 line intensities. The DMPO adduct with superoxide radical anion (DMPO–OOH) features a 12-line spectrum 37 that is usually not fully resolved owing to broadening effects, merging the features into sextets or quartets. Moreover, the adduct is unstable and may transform into DMPO–OH after some time period. Therefore, EPR experiments were performed in a mixed DMSO/water solvent, where the stability of the primary photogenerated superoxide radical anion is increased, as reported in. 38 EPR spectra were acquired consecutively at 30 s intervals for bulk g-C 3 N 4 and exfoliated samples TEX1-3 under constant intensity irradiation (Figures S6–S9). The control experiment was performed by irradiating the DMPO solution in the absence of g-C 3 N 4 (Figure S9). The results show that in the absence of g-C 3 N 4 , no spin adducts were formed. In contrast, both bulk and exfoliated g-C 3 N 4 samples produced the same primary radical or a mixture of primary radicals, as indicated by the identical EPR spectra of the DMPO spin adducts, differing only in intensity. The kinetics of spin adduct formation were markedly different between the bulk Table 3. Textural characteristics of bulk and exfoliated g-C 3 N 4 Material S BET (m 2 g −1 ) S m (m 2 g −1 ) V p (cm 3 g −1 ) Bulk 10 0.48 0.063 TEX1 58 5.5 0.21 TEX2 167 15 0.66 TEX3 183 17 0.88 Note: S m is the surface area of the micropores and V p is the pore volume. Figure 5. Characterization by photoluminescence spectroscopy Photoluminescence spectra of bulk and exfoliated g-C 3 N 4 using a 450 W Xe arc lamp. Table 4. Fitting parameters of PL decay curves of bulk and exfoliated g-C 3 N 4 Material B 1 (%) τ 1 (ns) B 2 (%) τ 2 (ns) B 3 (%) τ 3 (ns) τ avg3 (ns) τ avg2 (ns) Bulk 61.1 1.14 33.6 4.02 5.3 19.4 8.5 3.0 TEX1 53.5 1.45 40.6 4.44 5.9 18.9 8.2 3.5 TEX2 51.9 1.74 41.4 4.95 6.7 19.8 8.8 4.0 TEX3 49.7 1.71 42.8 5.04 7.5 19.4 9.1 4.1 6 iScience 28, 113899, December 19, 2025 iScience Article ll OPEN ACCESS and exfoliated samples (Figure 6). At a given irradiation intensity, the initial rates of spin adduct formation were determined to be 0.14, 0.70, 1.35, and 0.78 μmol L −1 s −1 ⋅mg −1 for the Bulk, TEX1, TEX2, and TEX3 samples, respectively. In the 5-min EPR experiment, a gradual build-up of the EPR signal was observed, while exfoliated g-C 3 N 4 provided a sharp increase in the EPR intensity followed by decay, indicating the photocatalytic decomposition of the DMPO spin adduct present in the EPR capillary. The fastest build-up and decay of the EPR signal was observed for TEX2 (maximum EPR intensity was observed at 30 s), while TEX1 provides maximum at 90 s and TEX3 at 60 s. It can be concluded that superoxide anion radicals were the primary reactive oxygen species produced by both bulk and exfoliated g-C 3 N 4 samples. The EPR experiments also indicated that TEX2 was the most efficient photocatalyst. Photocatalytic degradation of pharmaceuticals The bulk and exfoliated g-C 3 N 4 samples were used as photocatalysts for the degradation of the model pharmaceuticals ofloxacin, diclofenac, and caffeine. In the dark, suspensions of g-C 3 N 4 and the pharmaceutical were first stirred for 60 min to reach adsorption-desorption equilibrium. The suspensions were then stirred under 420 nm irradiation for 120 min. The kinetic curves are presented in Figure 7. The selected pharmaceuticals were phototactically degraded by their oxidation according to the reaction (7). The degradation of all pharmaceuticals followed first-order kinetics: −dc dt =kc (Equation 8) and Figure 6. Characterization by EPR spectroscopy (A) EPR spectra obtained from the DMPO spin trapping experiment with TEX2 g-C 3 N 4 . (B) The time development of the normalized EPR signal induced by the irradiation of bulk and exfoliated g-C 3 N 4 . The acquisition time of each spectrum was 30 s with a modulation amplitude of 0.2 mT and a microwave attenuation factor of 10 dB. Figure 7. Photocatalytic degradation of pharmaceuticals Kinetic curves of the photocatalytic degradation of (A) ofloxacin, (B) diclofenac, and (C) caffeine using Bulk, TEX1, TEX2, and TEX3 photocatalysts and (D) energy diagram. An LED source of 420 nm and an intensity of 13.5 mW⋅cm −2 was used. iScience 28, 113899, December 19, 2025 7 iScience Article ll OPEN ACCESS c=c0exp(− kt)(Equation 9) where k is the rate constant, and c and c 0 are the concentrations of pharmaceuticals at time t = t and t = 0, respectively. The rate constants are listed in Table 5. Their values increased in the sequence Bulk < TEX1 < TEX3 < TEX2. The degradation efficiency decreased in the following order: ofloxacin > diclofenac > caffeine. After 120 min, more than 95% of ofloxacin and diclofenac and approximately 80% of caffeine were decomposed. The photocatalytic results were compared with those reported in the literature for the different photocatalysts listed in Table S3. The experimental conditions under which the reported results were obtained varied in terms of reactor geometry, irradiation intensity and energy, reaction time, pharmaceutical concentration, photocatalyst mass, and photocatalyst composition. Therefore, it is impossible to compare them accurately; however, these studies suggest that simple thermally exfoliated g-C 3 N 4 provides comparable photocatalytic results. Photocatalytic processes usually obey the LangmuirHinshelwood mechanism, which assumes that two molecules adsorb on neighboring sites of a photocatalyst and then react with each other. Thermal exfoliation increased the specific surface area of g-C 3 N 4 from 10 to 184 m 2 g -1 (Table 3), leading to the formation of a higher number of reaction sites on the material surface and an increase in the electron–hole lifetime (Table 4). In addition, the PL intensity indicates a higher number of electrons and holes produced by the exfoliated g-C 3 N 4 . The number of reaction sites, charge carriers, and their lifetimes are important parameters for suitable photocatalysts and, in turn, for efficient photocatalytic degradation. A comparison of all three parameters (Figure S10), TEX2 was identified as the optimum photocatalyst. The different degradation rates of individual pharmaceuticals can be discussed in terms of their electrochemical redox potentials. The values of 0.90 V for ofloxacin, 39 0.66 V for diclofenac, 40 and 1.50 V for caffeine 41 (measured at pH 7 vs. the Ag/AgCl reference electrode) are available in the literature. Recently, graphitic carbon nitride was investigated to have the valence band potentials (E VB ) of 1.54–1.65 V and the conduction band potential (E CB ) from −1.10 to −1.19 V 42 ; therefore, photoinduced electrons are able to react with oxygen dissolved in water, resulting in the formation of superoxide radicals with a redox potential of −0.33 V. 43 These potentials agree well with the calculated medians of E VB and E CB from the literature data of 1.63 V and −1.09 V (n = 98), respectively, 44 which are displayed in Figure 7. The thermodynamic data indicate that oxidation by reactive oxygen species (ROS) is feasible for all pharmaceuticals. Similarly, the photoinduced holes oxidized ofloxacin and diclofenac. However, the involvement of holes in the oxidation of caffeine is likely limited because the oxidation potential of caffeine (E o = 1.50 V) Table 5. Rate constants (k) of photocatalytic degradation of pharmaceuticals Substance Bulk k ×10 −3 (min −1 ) TEX1 k ×10 −3 (min −1 ) TEX2 k ×10 −3 (min −1 ) TEX3 k ×10 −3 (min −1 ) Ofloxacin 4.67 ± 0.96 15.5 ± 0.6 27.5 ± 2.3 26.2 ± 1.0 Diclofenac 1.30 ± 0.06 4.54 ± 0.08 28.7 ± 2.6 20.4 ± 1.3 Caffeine 0.453 ± 0.027 4.45 ± 0.11 12.6 ± 0.9 10.5 ± 0.5 Figure 8. Analysis of degradation products HPLC chromatograms of (A) ofloxacin, (B) diclofenac, and (C) caffeine during the photocatalytic degradation using TEX2 at 420 nm. The column of 150 mm length, 4.6 mm i.d. (2.6 μm core-shell particles) and the mobile phase of 0.1% formic acid in water and a mixture of acetonitrile and methanol (85:15, v/v) in a ratio of 40:60 (v/v) were used. 8 iScience 28, 113899, December 19, 2025 iScience Article ll OPEN ACCESS Photocatalytic Performance of g-C 3 N 4 by Thermal Exfoliation. Bull. Chem. React. Eng. Catal. 19, 442–454. https://doi.org/10.9767/bcrec.20189. 28. Chen, M., Sun, M., Cao, X., Wang, H., Xia, L., Jiang, W., Huang, M., He, L., Zhao, X., and Zhou, Y. (2024). Progress in preparation, identification and photocatalytic application of defective g-C 3 N 4 . Coord. Chem. Rev. 510, 215849. https://doi.org/10.1016/j.ccr.2024.215849. 29. Wang, L., Wang, K., He, T., Zhao, Y., Song, H., and Wang, H. (2020). Graphitic Carbon Nitride-Based Photocatalytic Materials: Preparation Strategy and Application. ACS Sustain. Chem. Eng. 8, 16048–16085. https://doi.org/10.1021/acssuschemeng.0c05246. 30. Praus, P., � Reha ´� ckova ´, L., � Cı´� zek, J., Smy´ kalova ´, A., Ko� stejn, M., Pavlovsky´ , J., Filip Edelmannova ´, M., and Ko� cı´, K. (2022). Synthesis of vacant graphitic carbon nitride in argon atmosphere and its utilization for photocatalytic hydrogen generation. Sci. Rep. 12, 13622. https://doi.org/10. 1038/s41598-022-17940-3. 31. Nosaka, Y., and Nosaka, A.Y. (2017). Generation and Detection of Reactive Oxygen Species in Photocatalysis. Chem. Rev. 117, 11302–11336. https://doi.org/10.1021/acs.chemrev.7b00161. 32. Zheng, Q., Xu, E., Park, E., Chen, H., and Shuai, D. (2019). Looking at the overlooked hole oxidation: Photocatalytic transformation of organic contaminants on graphitic carbon nitride under visible light irradiation. Appl. Catal. B Environ. 240, 262–269. https://doi.org/10.1016/j.apcatb.2018. 09.012. 33. Xiong, L., and Tang, J. (2021). Strategies and Challenges on Selectivity of Photocatalytic Oxidation of Organic Substances. Adv. Energy Mater. 11, 2003216. https://doi.org/10.1002/aenm.202003216. 34. Svoboda, L., Praus, P., Lima, M.J., Sampaio, M.J., Maty´ sek, D., Ritz, M., Dvorsky´ , R., Faria, J.L., and Silva, C.G. (2018). Graphitic carbon nitride nanosheets as highly efficient photocatalysts for phenol degradation under high-power visible LED irradiation. Mater. Res. Bull. 100, 322–332. https://doi.org/10.1016/j.materresbull.2017.12.049. 35. Li, S., Dong, G., Hailili, R., Yang, L., Li, Y., Wang, F., Zeng, Y., and Wang, C. (2016). Effective photocatalytic H 2 O 2 production under visible light irradiation at g-C 3 N 4 modulated by carbon vacancies. Appl. Catal. B Environ. 190, 26–35. https://doi.org/10.1016/j.apcatb.2016.03.004. 36. Yang, J., Zhang, X., Xie, C., Long, J., Wang, Y., Wei, L., and Yang, X. (2021). Preparation of g-C 3 N 4 with High Specific Surface Area and Photocatalytic Stability. J. Electron. Mater. 50, 1067–1074. https://doi.org/10. 1007/s11664-020-08654-1. 37. Cle´ ment, J.-L., Ferre´ , N., Siri, D., Karoui, H., Rockenbauer, A., and Tordo, P. (2005). Assignment of the EPR Spectrum of 5,5-Dimethyl-1-pyrroline N-Oxide (DMPO) Superoxide Spin Adduct. J. Org. Chem. 70, 1198– 1203. https://doi.org/10.1021/jo048518z. 38. Dvoranova ´, D., Barbierikova ´, Z., Mazu´ r, M., Garcı´a-Lo´ pez, E.I., Marcı`, G., Lu� spai, K., and Brezova ´, V. (2019). EPR investigations of polymeric and H 2 O 2 -modified C 3 N 4 -based photocatalysts. J. Photochem. Photobiol. A. Chem. 375, 100–113. https://doi.org/10.1016/j.jphotochem.2019.02.006. 39. Liu, T., Xue, Q., Jia, J., Liu, F., Zou, S., Tang, R., Chen, T., Li, J., and Qian, Y. (2019). New insights into the effect of pH on the mechanism of ofloxacin electrochemical detection in aqueous solution. Phys. Chem. Chem. Phys. 21, 16282–16287. https://doi.org/10.1039/C9CP03486B. 40. Cid-Cero ´n, M.M., Guzma ´n-Herna ´ndez, D.S., Ramı´rez-Silva, M.T., Galano, A., Romero-Romo, M., and Palomar-Pardave ´, M. (2016). New insights on the kinetics and mechanisms of the electrochemical oxidation of diclofenac in neutral aqueous medium. Electrochim. Acta 199, 92–98. https:// doi.org/10.1016/j.electacta.2016.03.094. 41. Mersal, G.A.M. (2012). Experimental and Computational Studies on the Electrochemical Oxidation of Caffeine at Pseudo Carbon Paste Electrode and Its Voltammetric Determination in Different Real Samples. Methods 5, 520–529. https://doi.org/10.1007/s12161-011-9269-2. 42. Praus, P., Smy´ kalova ´, A., � Skuta, R., Ko� stejn, M., Pavlovsky´, J., Tokarsky´ , J., Foniok, K., Filip Edelmannova ´, M., and Ko� cı´, K. (2024). Graphitic C 3 N 4 and Ti 3 C 2 nanocomposites for the enhanced photocatalytic degradation of organic compounds and the evolution of hydrogen under visible irradiation. J. Photochem. Photobiol. A. Chem. 447, 115260. https://doi.org/10. 1016/j.jphotochem.2023.115260. 43. Wood, P.M. (1988). The potential diagram for oxygen at pH 7. Biochem. J. 253, 287–289. https://doi.org/10.1042/bj2530287. 44. Praus, P. (2021). On electronegativity of graphitic carbon nitride. Carbon 172, 729–732. https://doi.org/10.1016/j.carbon.2020.10.074. 45. Elhalil, A., Elmoubarki, R., Farnane, M., Machrouhi, A., Sadiq, M., Mahjoubi, F.Z., Qourzal, S., and Barka, N. (2018). Photocatalytic degradation of caffeine as a model pharmaceutical pollutant on Mg doped ZnOAl 2 O 3 heterostructure. Environ. Nanotechnol. Monit. Manag. 10, 63–72. https://doi.org/10.1016/j.enmm.2018.02.002. 46. Fernandes, T.A., Mendo, S.G., Ferreira, L.P., Neng, N.R., Oliveira, M.C., Gil, A., Carvalho, M.D., Monteiro, O.C., Nogueira, J.M.F., and Calhorda, M.J. (2021). Photocatalytic degradation of acetaminophen and caffeine using magnetite–hematite combined nanoparticles: kinetics and mechanisms. Environ. Sci. Pollut. Res. Int. 28, 17228–17243. https://doi.org/ 10.1007/s11356-020-12016-z. 47. Su, Q., Li, J., Yuan, H., Wang, B., Wang, Y., Li, Y., and Xing, Y. (2022). Visible-light-driven photocatalytic degradation of ofloxacin by g-C 3 N 4 / NH 2 -MIL-88B(Fe) heterostructure: Mechanisms, DFT calculation, degradation pathway and toxicity evolution. Chem. Eng. J. 427, 131594. https://doi.org/10.1016/j.cej.2021.131594. 48. Shang, Q., Liu, X., Zhang, M., Zhang, P., Ling, Y., Cui, G., Liu, W., Shi, X., Yue, J., and Tang, B. (2022). Photocatalytic degradation of ofloxacin antibiotic wastewater using TS-1/C 3 N 4 composite photocatalyst: Reaction performance optimisation and estimation of wastewater component synergistic effect by artificial neural network and genetic algorithm. Chem. Eng. J. 443, 136354. https://doi.org/10.1016/j.cej.2022. 136354. 49. Zhang, D., Qi, J., Ji, H., Li, S., Chen, L., Huang, T., Xu, C., Chen, X., and Liu, W. (2020). Photocatalytic degradation of ofloxacin by perovskite-type NaNbO 3 nanorods modified g-C 3 N 4 heterojunction under simulated solar light: Theoretical calculation, ofloxacin degradation pathways and toxicity evolution. Chem. Eng. J. 400, 125918. https://doi.org/10.1016/j.cej.2020. 125918. 50. Xia, B., Yao, J.-j., Han, C.-x., Zhang, Z., Chen, X.-Y., and Fang, Y.-J. (2018). Degradation of ofloxacin by UVA-LED/TiO 2 nanotube arrays photocatalytic fuel cells. Chem. Pap. 72, 359–368. https://doi.org/10.1007/ s11696-017-0285-6. 51. Chen, R., Ding, S., Fu, N., and Ren, X. (2023). Preparation of a g-C 3 N 4 / Ag 3 PO 4 composite Z-type photocatalyst and photocatalytic degradation of Ofloxacin: Degradation performance, reaction mechanism, degradation pathway and toxicity evaluation. J. Environ. Chem. Eng. 11, 109440. https://doi.org/10.1016/j.jece.2023.109440. 52. Jimenez-Villarin, J., Serra-Clusellas, A., Martı´nez, C., Conesa, A., GarciaMontan˜ o, J., and Moyano, E. (2016). Liquid chromatography coupled to tandem and high resolution mass spectrometry for the characterisation of ofloxacin transformation products after titanium dioxide photocatalysis. J. Chromatogr. A. 1443, 201–210. https://doi.org/10.1016/j.chroma.2016. 03.063. 53. Md Fauzi, M.A.F., Razali, M.H., Osman, M.U., and Mohd Azam, B. (2022). Synthesis and characterisation of TiO 2 /g-C 3 N 4 as photocatalyst for photodegradation of dyes, phenol and caffeine. Adv. Mater. Process. Technol. 8, 4395–4415. https://doi.org/10.1080/2374068X. 2022.2076976. 54. Tahir, M.B., Sagir, M., and Shahzad, K. (2019). Removal of acetylsalicylate and methyl-theobromine from aqueous environment using nano-photocatalyst WO 3 -TiO 2 @g-C 3 N 4 composite. J. Hazard. Mater. 363, 205–213. https://doi.org/10.1016/j.jhazmat.2018.09.055. 55. Ndabankulu, V.O., Maddila, S., and Jonnalagadda, S.B. (2019). Ozone facilitated degradation of caffeine using Ce-TiO 2 catalyst. J. Environ. Sci. Health B 54, 138–146. https://doi.org/10.1080/03601234.2018. 1530549. iScience 28, 113899, December 19, 2025 15 iScience Article ll OPEN ACCESS 56. Li-Chin, C., Chin-Hsiang, L., Sing-Wei, H., Yu-Cheng, W., and Yi-Chin, H. (2011). Photocatalytic Degradation Mechanism and Kinetics of Caffeine in Aqueous Suspension of Nano-TiO 2 . Adv. Mater. Res. 214, 97–102. https://doi.org/10.4028/www.scientific. net/AMR.214.97. 57. Li, M., Mei, Q., Han, D., Wei, B., An, Z., Cao, H., Xie, J., and He, M. (2021). The roles of HO⋅, ClO⋅ and BrO⋅ radicals in caffeine degradation: A theoretical study. Sci. Total Environ. 768, 144733. https://doi.org/10.1016/j.scitotenv.2020.144733. 58. Liu, X., Li, F., Liu, Y., Li, P., Chen, L., Li, B., Qian, T., and Liu, W. (2022). Degradation of diclofenac in a photosensitization-like photocatalysis process using palladium quantum dots deposited graphite carbon nitride under solar light. J. Environ. Chem. Eng. 10, 107545. https://doi.org/10.1016/ j.jece.2022.107545. 59. Mombelli, E., and Pandard, P. (2021). Evaluation of the OECD QSAR toolbox automatic workflow for the prediction of the acute toxicity of organic chemicals to fathead minnow. Regul. Toxicol. Pharmacol. 122, 104893. https://doi.org/10.1016/j.yrtph.2021.104893. 16 iScience 28, 113899, December 19, 2025 iScience Article ll OPEN ACCESS STAR★METHODS KEY RESOURCES TABLE EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS Most of the experiments and tests were performed at VSB-Technical University of Ostrava. EPR and XPS experiments were performed at University of Pardubice and the Czech Academy of Science, respectively. All the experiments involve no laboratory animals. METHOD DETAILS Synthesis of g-C 3 N 4 Bulk g-C 3 N 4 was synthesized by direct heating of 5 g of DCDA placed in a ceramic crucible with a lid in a muffle furnace and heated at a rate of 3 ◦C⋅min −1 from room temperature to 550 ◦C, the total heating time was 4 h. The crucible was then transferred out of the furnace and cooled to room temperature. The as-prepared bulk g-C 3 N 4 was ground into a fine powder using a laboratory mill. Bulk g-C 3 N 4 was exfoliated by heating a thin layer spread on a ceramic plate (diameter 8 cm, 50 mL) in a muffle furnace for 1–3 h (initial heating rate of 10 ◦C⋅min −1 , final temperature 500 ◦C). The ceramic plate containing the product was then cooled to ambient temperature. The exfoliated materials were labeled TEX1, TEX2, and TEX3 according to the exfoliation time (1, 2, and 3 h). The stability of the TEX photocatalysts toward hydrogen peroxide was tested using TEX2 and 10–30% H 2 O 2 solutions. One gram of TEX2 was placed in a 100 mL glass autoclave together with 50 mL of H 2 O 2 solution and stirred for 5 min. After stirring, the autoclave was placed in an oven at 150 ◦C for 5 h. The resulting TEX2 was filtered, washed several times with deionised water, and dried overnight at 105 ◦C. The stability of TEX3 during the photocatalytic process was tested over 3 cycles of the photocatalytic degradation of ofloxacin. The material was filtered after every cycle (120 min), washed with distilled water, and dried at 105◦C in a laboratory drier. Elemental analysis The C, H, and N contents in the g-C 3 N 4 materials were determined using a Flash 2000 elemental analyser (ThermoFisher Scientific, Waltham, MA, USA). The oxygen content was estimated as the remaining fraction after subtracting the measured C, H, and N content from 100%. Scanning and transmission electron microscopy Scanning electron microscopy and energy-dispersive X-ray spectroscopy (EDS) analyses of the g-C 3 N 4 materials were performed using a Tescan Vega microscope (Brno, Czech Republic) equipped with a tungsten cathode. SEM micrographs were obtained using secondary electron (SE) and backscattered electron (BSE) modes with an acceleration voltage of 30 keV. The samples were goldsputtered before analysis to ensure adequate electron conductivity. REAGENT or RESOURCE SOURCE IDENTIFIER Chemicals, peptides, and recombinant proteins Ofloxacin Merck 33703 Diclofenac Merck SML3086 Caffeine Merck C0100000 Dicyandiamide (DCDA) Merck D76609 Acetonitrile Supelco 1.00030 Formic acid Supelco 1.00029 4-hydroxy-2,2,6,6-tetramethyl-1piperidinyloxyl (TEMPOL) Aldrich S332216 Dimethyl sulfoxide (DMSO) Merck 472301 5,5-dimethyl-1-pyrroline N-oxide (DMPO) Merck 92688 Methanol Merck 646377 Software and algorithms QSAR Toolbox OECD and ECHA 4.6 iScience 28, 113899, December 19, 2025 e1 iScience Article ll OPEN ACCESS Transmission electron microscopy was performed using a JEOL 2100 microscope equipped with a LaB 6 electron gun. An accelerating voltage of 200 kV was applied. TEM micrographs were recorded using a Tengra camera (EMSIS). For the TEM analysis, the samples were dispersed in ethanol and sonicated for 5 min. One drop of this solution was placed on a copper grid with a holey carbon film and dried at room temperature. X-Ray diffraction analysis X-ray diffraction patterns were recorded using a Rigaku SmartLab diffractometer (Rigaku, Tokyo, Japan) equipped with a D/tex Ultra 250 detector. The X-ray irradiation source was a Cu tube (CuK α , λ 1 = 0.154059 nm, λ 2 = 0.154441 nm) operated at 40 kV and 30 mA. The incident and diffracted beam optics were equipped with 5 deg Soller slits, and the incident slits were set to 1 mm. The powder samples were gently ground using an agate mortar, pressed by a microscope glass in a rotational sample holder, and measured in reflection mode (Bragg-Brentano geometry). The samples were rotated (30 rpm) to eliminate the effects of the preferred orientation. The XRD patterns were collected in the 2θ range of 5–90 deg with a step size of 0.01 deg and a speed of 0.5 min −1 . Measured XRD patterns were evaluated using PDXL 2 software (version 2.4.2.0, Rigaku, Tokyo, Japan) and compared with a database PDF-2, release 2015 (ICDD, Newton Square, USA). A crystallite size L was obtained from Scherrer’s equation B(2θ)=Kλ Lcosθ where B(2θ) is a broadening B(2θ) (in radians) at a half-maximum intensity (FWHM) of a diffraction band, λ is the wavelength of X-rays, θ is Bragg’s angle, and K is a constant equal to 0.94 for cubic or 0.89 for spherical crystallites (the value K = 0.90 was used in this work). X-ray photoelectron spectroscopy X-ray photoelectron spectroscopy was performed using a spectrometer ESCA 3400 (Kratos Analytical Ltd, UK) with a base pressure in an analysis chamber of 5.0 ×10 −7 Pa. The powdered materials were placed on conductive carbon tape and analyzed. The electrons were excited using Mg K α radiation (hν = 1253.6 eV) generated at 12 kV and 10 mA, respectively. For all spectra, the Shirley background was subtracted. Peaks ascribed to sp 2 hybridised nitrogen (C=N-C) were set to 398.8 eV for charge correction. Physisorption of nitrogen The specific surface area was measured by physisorption of nitrogen at 77 K using a 3Flex apparatus (Micromeritics, USA). Before the analysis, each material was degassed under vacuum (approximately 0.6 bar at 105 ◦C for 120 h (5 days)) to release physisorbed water and organic residuals from the pores. After this pre-treatment, the nitrogen adsorption-desorption isotherms of all materials were measured at 77 K in a relative pressure range of p/p 0 ∼10 −9 -0.99. The nitrogen adsorption–desorption data were treated according to the classical BET theory (for p/p 0 ≈ 0.05–0.25). The Barrett-Joiner-Halenda (BJH) model was used to calculate the pore size distributions of the materials. The Carbon Black STSA standard isotherm (typically used for carbon-based materials) was used for comparison using the Faas correction and assuming a cylindrical-pore geometry of mesopores and macropores. Fourier transform infrared spectroscopy Fourier transform infrared spectroscopy was performed using a Nicolet iS50 device (Thermo Scientific, Waltham, MA, USA) with the KBr pellet technique. The powder sample was mixed and homogenised with KBr (approximately 200 mg) and pressed at a pressure of 20 MPa to obtain a transmission pellet. The prepared pellet was placed in the holder of a transmission attachment, and the FTIR spectra were collected in the wavenumber range of 500–4000 cm −1 with a resolution of 2 cm −1 . Each spectrum consisted of at least 64 scans, each lasting 1 s. Before each measurement, the background was collected to eliminate the effects of the apparatus and environment. UV-Vis spectroscopy UV-Vis diffuse reflectance spectroscopy was performed in the range of 300–1000 nm using a Specord 250 (Analytik Jena, Germany) instrument equipped with an integrating sphere (Analytik Jena, type 820-60139-P). The absorbance data were evaluated using Aspect UV software. Band gap energies were calculated using from Tauc’s plots as Ahν=C(hν−Eg)p where A is the absorbance, hν is the energy of the incident photons, E g is the bandgap energy, C is a constant, and p is a power quotient depending on the type of electron transition (p = ½ was used in this study). Photoluminescence spectroscopy Photoluminescence spectroscopy measurements were performed on an FLS980 fluorescence spectrometer (Edinburgh Instruments) with double monochromators on both the excitation and emission sides, equipped with an R928P photomultiplier in a e2 iScience 28, 113899, December 19, 2025 iScience Article ll OPEN ACCESS thermoelectrically cooled housing (Hamamatsu Photonics). A 450 W xenon arc lamp served as the excitation source for the steadystate spectra. Spectral correction curves were obtained from the Edinburgh Instruments. The powder samples were mounted onto a front-face quartz sample holder. For time-resolved measurements, an EPL-375 ps pulsed diode laser (λ em = 372 nm) with a pulse width of 66.5 ps, repetition rate of 20 MHz, and average power of 75 μW (Edinburgh Instruments) was used in conjunction with a time-correlated single-photon counting (TCSPC) system. Obtained PL decay curves were fitted using a triple exponential function I(t)=∑3 i=1Biexp(−t τi);where ∑3 i=1Bi=1 In this expression, τ i represents the decay time constant, and B i represents the normalised amplitude of each component. The amplitude-weighted average decay lifetime τ avg of the entire fluorescence decay process was calculated as follows: τavg = τ2 iBi ∑τiBi Electron paramagnetic resonance spectroscopy Electron paramagnetic resonance spectra were acquired using a Miniscope MS300 X-band spectrometer (Magnettech, Germany). A 50 μL aliquot of the mixture containing 1 mg⋅mL −1 of g-C 3 N 4 sample and 0.05 mmol⋅L −1 DMPO in DMSO-water mixed solvent 4:1 (v/v) was briefly sonicated and then aspired into a capillary. The EPR spectra were measured in the dark and then directly irradiated in the EPR cavity. Ten spectra were recorded in sequence, with an acquisition time of 30 s, modulation amplitude of 0.2 mT, and microwave attenuation factor set at 10 dB. A Lightningcure LC8 (Hamamatsu, Japan), featuring a high-intensity mercury-xenon lamp set at 100% intensity, served as the light source, and the far-UV part of the spectrum was removed using an A9616-05 filter (Hamamatsu, Japan). The EPR spectra were quantitatively assessed by converting the EPR double integrals to spin concentration values. For this purpose, the EPR spectrum of TEMPOL standard stable radical TEMPOL was recorded (in the absence of illumination under otherwise identical instrument settings), baseline-corrected, and double-integrated to obtain the conversion factor. The obtained spin concentrations were normalised per unit catalyst mass and used to calculate the initial rates of spin adduct formation. Photocatalytic experiments All chemicals used were of analytical reagent grade (pro analysi, purity 99.0–99.8%). Deionised water with a pH of 6.5–6.9 and conductivity of <0.1 μS⋅cm -1 (AQUAL 29, Czech Republic) was used throughout the study. Photocatalytic degradation experiments were performed in suspensions containing 10 mg of bulk or exfoliated g-C 3 N 4 photocatalysts in 150 mL aqueous solutions of ofloxacin (c = 10 mg⋅L −1 ), diclofenac, or caffeine (c = 20 mg⋅L −1 ). The suspensions were stirred in the dark for 60 min to reach adsorption-desorption equilibrium and then irradiated with an LED source (420 nm, intensity of 13.5 mW cm −2 ) for 120 min. The temperature of the reaction mixture was maintained at 20 ◦C. Aliquots (3 mL) were collected at regular intervals and filtered using Chromafil GF/RC-20/25 syringe filters (pore size 0.2–1.0 μm). The filtrates were analyzed using a high-performance liquid chromatography (HPLC) Nexera XR chromatograph (Shimadzu, Kyoto, Japan) equipped with a photodiode array (PDA) detector. An analytical Kinetex Biphenyl 100 A ˚ column (150 mm length, 4.6 mm i.d., 2.6 μm core-shell particles; Phenomenex, Torrance, CA, USA) was used for HPLC analysis. HPLC separations were performed in isocratic mode, with a mobile phase consisting of 0.1% formic acid in ultrapure water and a mixture of acetonitrile and methanol (85:15, v/v) in a ratio of 40:60 (v/v). The analysis time was 5 min, the flow rate was 1 mL⋅min −1 and the injection volume was 10 μL. Identification of intermediary transformation products and prediction their toxicity To identify intermediary transformation products, liquid chromatography combined with high-resolution tandem mass spectrometry was employed. For this purpose, 10 mL aliquots were collected before and after the termination of the photocatalysis process and filtered. Blank (demineralised water in contact with g-C 3 N 4 particles for 1 h) and standard samples (stock solutions of the pharmaceuticals diluted with demineralised water at appropriate concentrations) were prepared for comparative purposes. HPLC/HRTMS analysis was performed using an Acquity HPLC system coupled to a high-resolution tandem mass spectrometer (Synapt G2-S; Waters, Massachusetts, USA). The same Kinetex Biphenyl 100 A ˚ column (150 mm length, 4.6 mm i.d., 2.6 μm core-shell particles; Phenomenex, Torrance, CA, USA) was used for chromatographic separation, along with a mobile phase consisting of ultrapure water (A) and acetonitrile (B), both acidified with 0.1% formic acid. A gradient method was applied with a profile as follows: start with 5% B, followed by an increase to 90% B in 1 min, held for 5 min, decreased to 5% B in 20 s, and then held for 2 min and 40 s. The injection volume was 10 μL, and the flow rate was 0.5 mL min −1 . The parameters of the mass spectrometer were as follows: spray voltage of +2.5 kV, sample cone voltage of 18 V, source temperature of 100 ◦C, desolvation temperature of 250◦C, desolvation gas flow of 600 L/h, analyser in V-mode, scan range of 50–1200 (Da), scan time of 0.2 s, and interscan delay of 0.02 s. MS/ MS data were obtained via fragmentation experiments using a trap-collision cell. A collision energy of 25 eV was used. The instrument was calibrated using a sodium formate solution (0.5 mmol L −1 , dissolved in 90:10 2-propanol/water, v/v). Leucine-enkephalin iScience 28, 113899, December 19, 2025 e3 iScience Article ll OPEN ACCESS (50 pg μL −1 , dissolved in 50:50 acetonitrile/water + 0.1% formic acid, v/v/v) was used for lock mass correction using a reference electrospray probe during an HPLC/HRTMS run. Data collection was performed using MassLynx software (version 4.1; Waters, Milford, MA, USA). The quantitative structure–activity relationship method was used to predict the ecotoxicity of the identified TPs. The freely available software QSAR Toolbox by OECD and ECHA was used for this purpose, implying a method previously described in the literature. 59 QUANTIFICATION AND STATISTICAL ANALYSIS Confidence intervals were calculated for n = 3–5 replicates at the significance level α = 0.05. e4 iScience 28, 113899, December 19, 2025 iScience Article ll OPEN ACCESS