Optimization of the Urbach energy and charge carrier dynamics in g-C3N4 through strategic potassium precursor selection: Insights and challenges

Full Length Article Optimization of the Urbach energy and charge carrier dynamics in g-C 3 N 4 through strategic potassium precursor selection: Insights and challenges Ladislav Svoboda a,* , Zuzana Vilamov´ a a , Petr Praus b,c , Vlastimil Nov´ ak b , Kateˇ rina Mamulov´ a Kutl´ akov´ a a , Martin Petr d , Jiˇ rí Bedn´ aˇ r a , Vít Jochim b , Zuzana ˇ Simonov´ a a,e , Richard Dvorský a,b a Nanotechnology Centre, Centre for Energy and Environmental Technologies, VSB – Technical University of Ostrava, 17. listopadu 2172/15, 708 00 Ostrava-Poruba, Czech Republic b Department of Chemistry and Physico-Chemical Processes, VSB-Technical University of Ostrava, 17. listopadu 2172/15, 708 00 Ostrava-Poruba, Czech Republic c Institute of Environmental Technology, Centre for Energy and Environmental Technologies, VSB-Technical University of Ostrava, 17. listopadu 2172/15, 708 00 Ostrava-Poruba, Czech Republic d Regional Centre of Advanced Technologies and Materials, Faculty of Science, Palacky University, ˇ Slechtitelů 27, 783 71, Olomouc, Czech Republic e Centre for Advanced Innovation Technologies, Faculty of Materials Science and Technology, VSB – Technical University of Ostrava, 17. listopadu 2172/15, 708 00 Ostrava-Poruba, Czech Republic ARTICLE INFO Keywords: g-C 3 N 4 Urbach energy Charge carrier dynamics Photocatalysis Potassium modification Time-Correlated single photon counting ABSTRACT The previous studies on K-doped g-C 3 N 4 suggest a simple correlation between the increased potassium content and the improved photocatalytic activity due to a shift of the valence band potential to more positive values, leading to the direct oxidation of hydroxyl ions to hydroxyl radicals by holes. In this study, we investigated the influence of different potassium precursors on the properties and photocatalytic efficiency of K-modified g-C 3 N 4 materials in Rhodamine B degradation. Detailed characterization revealed that the choice of precursor significantly affects the structural and optoelectronic properties with complex correlations rather than superficial relationships between potassium content and photocatalytic activity. Notably, KCl-based samples, despite having a lower specific surface area, exhibited enhanced photocatalytic properties due to prolonged carrier lifetimes and improved charge transfer efficiency compared to KOH-based samples. Higher precursor concentrations resulted in more recombination centers, increasing Urbach energy and reducing photocatalytic activity. These findings underscore the importance of precise synthesis control, revealing that both K and Cl atoms act as charge transfer bridges, improving the material’s photocatalytic properties beyond mere surface area enhancement. In addition, even smaller amounts of Cl atoms in g-C 3 N 4 showed much higher impact on the final position of conduction and valence band potentials than K atoms. 1. Introduction The issues regarding the increasing amount of persistent chemicals still need to be adequately solved. Advanced oxidation processes (AOPs) using photocatalytically active materials should easily solve this challenge due to their ability to produce reactive oxygen species (ROS), such as hydroxyl and superoxide radicals. The commercial use of these materials depends on their harmlessness to humans and the ecosystem, the possibility of the used material separation from the environment after the cleaning process, and overall photocatalytic activity. Moreover, crucial requirements include high photocatalytic activity and energy efficiency to activate the whole process. Thus, it is possible to degrade persistent chemicals in the shortest possible time and use radiation sources such as sunlight free of cost. This can be achieved by using materials with suitable band gap positions that would produce hydroxyl radicals, among the most potent oxidizing agents after fluorine. Nanomaterials TiO 2 and ZnO achieve such requirements, as both can directly oxidize water molecules to hydroxyl radicals [1–3]. This guarantees fast photocatalytic degradation of harmful organic compounds, such as phenols, antibiotics, dyes, and their intermediates. The disadvantage, * Corresponding author at: Nanotechnology Centre, Centre for Energy and Environmental Technologies, VSB – Technical University of Ostrava, 17. listopadu 2172/ 15, 708 00 Ostrava-Poruba, Czech Republic. E-mail address: [email protected] (L. Svoboda). Contents lists available at ScienceDirect Applied Surface Science journal homepage: www.elsevier.com/locate/apsusc https://doi.org/10.1016/j.apsusc.2024.161162 Received 1 July 2024; Received in revised form 15 August 2024; Accepted 3 September 2024 Applied Surface Science 679 (2025) 161162 Available online 4 September 2024 0169-4332/© 2024 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/bync-nd/4.0/ ). however, is their wide band gap (about 3.2 eV), and their activation requires UV radiation. Moreover, nano-sized TiO 2 (d <50 nm) has been lately highly discussed regarding its possible genotoxicity based on the final report of the European Commission for Consumer Safety and Carcinogenicity, according to International Agency for Research of Cancer (IARC). Lately, IARC recently even moved TiO 2 into group 2B, possibly carcinogenic for humans by inhalation [4]. Recently, other materials, such as graphitic carbon nitride, got much attention from many scientists worldwide due to their interesting physicochemical properties, simple synthesis, and possible modifications, making this material a promising photocatalyst for the future. Due to its narrow band gap (2.7 eV), g-C 3 N 4 can utilize a visible part of the solar spectrum. It also has other precious features, such as high chemical stability, environmental friendliness, and its not-toxicity character, which makes it an almost perfect candidate for practical application in AOPs [3,5–8]. The main drawback of g-C 3 N 4 is that the shifted valence band generates holes with less oxidizing power, unable to directly produce hydroxyl radicals like TiO 2 and ZnO. Recently, a band gap engineering of g-C 3 N 4 by alkali metal doping has been seen as a feasible way to overcome the drawback of achieving faster photocatalytic degradation and high mineralization ability. However, there is a significant discrepancy among scientists regarding the band gap modification of g-C 3 N 4 by potassium, and there are also several different suggestions on how potassium influences photocatalytic activity. The ability to shift the position of the valence band by simple potassium doping would be desirable and much appreciated for different practical applications in AOPs. The following works provided such bandgap engineering possibilities [9–12]. These papers showed information based on XPS measurements that the valence band potential (VBP) shifted significantly compared to pristine g-C 3 N 4 . Hu et al. [9] published the significant shift of VBP from +1.56 eV to +2.21 eV for a sample with 2.3 wt% of K. Zhang et al. [11] reported the shift from +1.56 eV to +2.21 eV at 2.3 wt% of K according to inductively coupled plasma (ICP) spectroscopy. Values reported by Zhang et al. and by Hu et al. are identical. The further comparison of those papers revealed that both contained the same results, including kinetic data, despite the temperature during the synthesis of K-modified g-C 3 N 4 , which differed by 30 ◦C. Even such a small difference in preparation temperature should result in different material properties and kinetic rates [13]. Xiong et al. [12] also reported the shift of VBP from +1.58 eV to +1.86 eV. In this case, the amount of potassium should be 5 wt%. based on a theoretical calculation comparing the amount of used potassium precursors and the weight of the final g-C 3 N 4 material. Further, other papers also reported that VBP was shifted to more positive values. These works dealt with tuning the band gap of g-C 3 N 4 by using a mixture of KCl and NaCl with a ratio of 1:1 to obtain K-Na co-modified g-C 3 N 4 by a so-called molten salt method. Zhao et al. [14] used KCl-NaCl as a solvent, and the shift of VBP was said to be from +1.55 eV to +2.25 eV, and the amount of K and Na was reported to be 1.88 wt% and 0.95 wt%, respectively. However, the actual melting temperature of KCl-NaCl (1:1) reported by other papers and literature should be as high as 656 ◦C according to the literature [15], 685 ◦C [16], 667 ◦C [17], and 645 ◦C (the eutectic point) at 44 wt% NaCl in the mixture of KCl and NaCl [18]. Zhao et al. [14] reported that the melting temperature of KCl:NaCl (1:1) was observed to be at 476 ◦C, and thus, this mixture was used as the solvent during K-Na co-modified g-C 3 N 4 synthesis at 520 ◦C. However, this temperature is still more than 100 ◦C below the required temperature to obtain the melted salt solution of KCl-NaCl. Similarly, such a strange temperature value regarding the melting point of KCl:NaCl (1:1) can be found in another paper published three years later where authors reported that VBP shifted from +1.62 eV to +2.26 eV and the amount of K and Na were reported to be 1.7 wt% and 1.0 wt%, respectively [10]. So far, several papers have concluded that VBP can be easily shifted to more positive values by controlling the potassium content in the final K-modified g-C 3 N 4 materials. Moreover, other papers also dealt with the K modification of g-C 3 N 4 but they did not observe such a significant VBP shift. Wang et al. [19] prepared KBr-modified g-C 3 N 4 by using different mass fractions of KBr. In this case, the VBP of pristine g-C 3 N 4 was reported to be +1.65 eV. After doping (7, 10, 13 wt% of KBr relative to gC 3 N 4 ), the VBP of materials was reported to have even less oxidative power +1.57, +1.59, and +1.58 eV, respectively. Yu et al. [20] published the preparation of K-modified g-C 3 N 4 by a one-step KOH-assisted route using different nitrogen-rich precursors, such as melamine, thiourea, or urea. Here, the VB XPS measurement revealed that the VBP did not change. Instead of a VBP shift, a conduction band (CB) position changed due to introduced surface vacancies. The presence of potassium in the final material or its influence was not investigated or considered in their work. Based on the literature, the impact of potassium precursors on the band gap engineering of g-C 3 N 4 has not been fully clarified, and we might question some of the published data. This study aims to delve deeper into the proposed potassium-based band gap engineering phenomena. We will achieve this by preparing K-modified g-C 3 N 4 materials, thoroughly characterizing them to obtain detailed physicochemical and photocatalytic properties, and determining the influence of potassium precursors on the optoelectronic properties of K-modified g-C 3 N 4 materials. This includes a focus on the Urbach energy (E U ), a crucial parameter for optimizing a range of applications, such as transistors [21,22], light-emitting diodes [23], photodetectors [24], and thin-film transistors [25]. During electron excitation, if a material contains intra-band gap states, the density of electronic states in the valence and conduction band tails into the energy band gap is known as the Urbach tail [26]. The Urbach energy, i.e., the width of the band tail of localized states in the bandgap, is used to determine the degree of the energetic disorder of band edges caused by a structural disorder (defects) of the lattice [27]. It was reported that the Cu-doped ZnO photocatalyst and TiO 1.7 ultra-thin films with higher Urbach energy had higher degradation activity [28,29]. However, an increase in Urbach energy does not always mean an improvement in photocatalytic properties. In the case of TiO 2 thin film [30], higher photocatalytic activity was observed for the sample with the lowest Urbach energy value. These findings point out that it is essential to know to what extent the structural defects formed are still beneficial and function as active centers and when they become more of a problem and the defects formed serve as recombination centers. Moreover, we also focused on using different potassium precursors, taking into note the anion part of the precursors used, unlike other authors who studied somewhat different nitrogen-rich precursors in their experiments or various alkaline salts. Thus, this work fills another scientific gap and helps better understand the overlooked mechanism underlying the band gap formation and charge carrier dynamics, leading to enhanced photocatalytic activity of K-modified g-C 3 N 4 materials. 2. Experimental 2.1. Reagents and materials Dicyandiamide (99.5 %) and Rhodamine B (Rh B) (98+%) were purchased from Acros Organics (Belgium). Potassium dichromate (99.5 %) and potassium hydroxide (85+%, G.R.) were purchased from Penta (Czech Republic). Potassium chloride (100 %, G.R.) and isopropyl alcohol (IPA) (G.R., ISO reagent) were purchased from Lach-Ner, s.r.o. Terepthalic acid (99 %) was purchased from MACH CHEMIK´ ALIE, s.r.o. (Czech Republic). Triethanolamine (TEOA) (95+%) was purchased from VWR. LUDOX® AS-30 colloidal silica and p-benzoquinone (p-BQ) were purchased from Merck (Germany). Zinc oxide (ZnO, nZ-BOCH 01) was purchased from company Bochemie a.s. (Czech Republic). All chemicals in this study were used as received without further purification. Deionized water was used for all prepared solutions. L. Svoboda et al. Applied Surface Science 679 (2025) 161162 2 2.2. Preparation of materials All materials used in this work were synthesized via thermal polycondensation of a powdered mixture of dicyandiamide and one of the selected potassium precursors (KOH or KCl). In a typical process, dicyandiamide (2 g) was dissolved in 10 ml deionized water under constant stirring. Then 10 ml of the potassium precursor solution with different concentrations (0.02, 0.04, 0.06, and 0.08 mol L -1 ) was added. In the next step, the suspension was heated to 100 ◦C and the obtained powder was dried in an oven at 80 ◦C and grounded in an agate mortar to obtain a fine powder. Such prepared powder was further annealed at 520 ◦C (5 ◦C min −1 ) for 2 h. The obtained materials were denoted as KOH(x)–CN and KCl(x)–CN, where x =molar concentration of the used potassium precursor during synthesis. Pristine g-C 3 N 4 was prepared according to the protocol described above, but without the potassium precursors. 2.3. Materials characterization X-ray powder diffraction (XRPD) analysis was recorded by a Bruker D8 Advance diffractometer in reflection mode in a symmetrical BraggBrentano arrangement. A fast position sensitive detector (VÅNTEC 1) and radiation CoK α (λ =0.1789 nm, U=35 kV, I=25 mA) were used. The phase composition was evaluated using database PDF 2 Release 2020 (International Centre for Diffraction Data). The textural properties of samples were determined by using nitrogen physisorption at 77 K using a 3Flex apparatus (Micromeritics, USA). Before to the physisorption analysis each sample was degassed under a vacuum of 0.6 bar at a temperature of 350 ◦C for 24 h to remove physisorbed moisture. Then nitrogen adsorption–desorption isotherms at 77 K were obtained in a relative pressure range p/p 0 ~ 1⋅10 -4 -0.99. Measured nitrogen adsorption–desorption data were processed using the standard BET theory (for p/p 0 ~ 0.05–0.3). From measured data, the specific surface area (A BET (m 2 g −1 )) and the net pore volume (V net (mm 3 liq g −1 )) were evaluated. The morphology, size, and structure of pristine g-C 3 N 4 and K-modified g-C 3 N 4 were investigated by a scanning electron microscope (SEM; JSM-7200F, Jeol) and a transmission electron microscope (TEM; 2200FS, Jeol) equipped with an FEG cathode and a large-angle silicon drift detector (SDD-EDX; Centurio, Jeol). An energy dispersive fluorescence (XRF) spectrometer SPECTRO XEPOS (SPECTRO Analytical Instruments GmbH, Germany) equipped with a 50 W Pd X-ray tube was used for the determination of doped potassium and chloride concentration in final samples. Fourier transform infrared (FT-IR) spectra were measured by ATR technique by using a Fourier Transform Infrared Spectrometer Nicolet iS50 (Thermo Nicolet, USA) in the range 4000 – 400 cm −1 . The X-ray photoelectron spectroscopy (XPS) measurements were performed using a Nexsa G2 XPS system (Thermo Fisher Scientific) with a monochromatic source (Al-K α ) and a photon energy of 1486.7 eV. All the spectra were measured in a vacuum of 1.2 x 10 -7 Pa and at the room temperature of 20 ◦C. The analyzed area on each sample was spot of 200 µm in diameter. The survey spectra were measured with a pass energy of 150.00 eV with 1.0 eV step while for the high-resolution spectra were used the pass energy of 30.00 eV with 0.1 eV step. Charge compensation was used for all the measurements. The spectra were evaluated with an Avantage 6.5.1 (Thermo Fisher Scientific) software. The prepared powders’ diffuse reflectance was obtained using a UV–VIS diffuse reflectance spectrometer UV-2600 (IRS-2600Plus, Shimadzu, Japan) at room temperature in the 220 – 700 nm range. The spectra were transformed to equivalent Kubelka-Munk units. A spectrometer FLS1000 (Edinburgh Instruments Ltd., UK) was used to obtain the steady-state photoluminescence spectra. The spectrometer was equipped with a 450 W ozone-free xenon arc lamp and a PMT-900 detector. An excitation wavelength was set at 385 nm and emission spectra were recorded in a range from 400 to 600 nm. The Mott-Schottky recordings were obtained by using a Metrohm Autolab PGSTAT302 potentiometer with an AC frequency of 300 Hz and an amplitude of 10 mV. A glassy carbon electrode was used as a working electrode, and Ag/AgCl (3 M KCl) and a Pt sheet served as reference and counter electrodes, respectively. Time-resolved fluorescence decay spectra were obtained by using a time-correlated single photon counting (TCSPC) setup at a photoluminescence spectrometer FLS1000 (Edinburgh Instruments Ltd., UK). Samples were excited by monochromatic picosecond pulsed diode laser (EPL-375) and the fluorescence emissions were monitored at their maximum emission wavelengths. The obtained decay curves were analysed and fitted with a three-exponential model by using data analysing software FAST (v3.5.0) and Fluoracle (v2.21.0). The intensity average lifetime ( τ avg ) of photogenerated charge carriers was calculated by using equation (1) [31]: τ avg =∑ n i=1 ai τ 2 i/∑ n i=1 ai τ i(1) where a i is the pre-exponential factor and τ i is the lifetime. 2.4. The photocatalytic activity measurements The photocatalytic activities were studied by the decomposition of Rh B under visible light irradiation using one high-power 10 W LED chip with an emission maximum of 416 nm. For the photocatalytic experiments, 50 mg of a photocatalyst was dispersed in an aqueous solution of Rh B (5 ppm, 50 ml). Before irradiation, the suspension was magnetically stirred in the dark to achieve adsorption–desorption equilibrium. Then, 1.3 ml aliquots were sampled at a specific interval and centrifuged to remove particles before absorbance measurements. The UV–VIS spectra were measured using a UV–VIS spectrophotometer (SHIMADZU UV-1601, Japan) ranging from 390 to 640 nm with a maximum absorbance of Rh B at 554 nm. The Rh B concentration was calculated using the Lambert-Beer equation. Photolysis was measured under the same conditions without the presence of photocatalysts. The radical trapping experiments were carried out under the same conditions in the presence of scavengers. TEOA was used as a hole scavenger (h + ), IPA for free hydroxyl radicals (•OH), and p-BQ for superoxide radical anions (•O 2 – ) scavenging test. To evaluate the stability of the K-modified g-C 3 N 4 , the most active photocatalyst, KOH(0.04)–CN, was selected as a model example. After each photocatalytic test, the photocatalyst was separated from the solution by centrifugation, washed with deionized water, and subsequently dried at 60 ◦C. Such a recycled photocatalyst was used without any additional treatment. The photocatalytic performance was evaluated within four subsequent cycles under the same reaction conditions. 3. Results and discussion 3.1. Microstructural analysis The XRD analysis investigated the crystalline and chemical structures of prepared pristine g-C 3 N 4 and K-modified g-C 3 N 4 . The obtained XRD patterns are depicted in Fig. 1. Pristine g-C 3 N 4 has two diffraction peaks located at 2θ =14.999◦, which correspond to an in-planar ordering of tri-s-triazine units (100) and 31.967◦, which is indexed as (002), and is attributed to the interlayer stacking of conjugated aromatic layers [32]. As shown on Fig. 1, the diffraction peak positions did not change. This means that the main structure of g-C 3 N 4 was preserved and corresponds to the heptazine-based g-C 3 N 4 structure [33]. Furthermore, the XRD patterns were carefully fitted by applying deconvolution on raw data (Fig. S1) to study the shapes and intensities of (100) and (002) diffraction peaks. The intensities of these two diffraction peaks decreased with the increasing potassium precursor content. This suggests that both potassium precursors affected the L. Svoboda et al. Applied Surface Science 679 (2025) 161162 3 formation of g-C 3 N 4 , and the graphitic C 3 N 4 repeat motif was restricted and disrupted in its regularity of synthesis. These results agree with other work [34] where even higher potassium content, up to 7 wt%, was used. The K-modified g-C 3 N 4 was prepared from a mixture of melamine and potassium fluoride dihydrate. However, different results were observed regarding the position of the diffraction peak (002). When KF and melamine [35], KOH and urea [20], KBr and thiourea [19], KCl and melamine [36], and KCl and melamine/urea mixtures [37] were used as precursors, the (002) diffraction peak shifted to higher 2θ values. This was attributed to the formation of defects, such as structural distortion and introduction C≡N groups, and that the surface of prepared K-doped g-C 3 N 4 samples underwent the flattening process of the undulated single layers. Some other papers reported the shift of the diffraction peak (002) to lower 2θ values. This shift would suggest that the ion intercalation into the g-C 3 N 4 interlayer has occurred. Examples of such reported shifts can be found in the following papers, where KOH and dicyandiamide [9,11], KCl-NaCl and melamine [10,14], NaBr (or KBr) and thiourea [12], and KCl and melamine [38], were used. Our case had no significant peak shift, although other measurements suggest that doping occurred. More significant changes were observed in crystallite sizes (Table S1). Crystallite sizes of all samples were calculated from the most intensive reflections (002) using the Scherrer equation (2) [39]: LC=K•λ β•cosθ(2) where K=0.89 is the dimensionless shape factor, λ is the Co K α radiation wavelength (nm), β is full width at half maximum (FWHM) of the g-C 3 N 4 (002) reflection (◦), and θ is a position of the reflection (◦). The crystallite sizes of g-C 3 N 4 samples obtained from dicyandiamide and KOH are smaller than those obtained using KCl, where the crystallite sizes increased with the increasing concentration of KCl. The possible explanation for this tendency might be that the polycondensation route for dicyandiamide to form g-C 3 N 4 differs in the presence of KOH and KCl. According to the previous report [40], dicyandiamide is condensed into melamine at around 350 ◦C, and tri-s-triazine units are formed at around 390 ◦C. The final condensation of these units and their polymerization into g-C 3 N 4 form is achieved at 520 ◦C. It is important to note that both KOH and KCl differ significantly in their melting temperatures. Potassium hydroxide melts around 406 ◦C [41] and KCl at 776 ◦C [42]. This means that the melted KOH could easily disrupt the nucleation and growth of g-C 3 N 4 during polycondensation. Thus, it is easy to understand that the crystallite sizes of g-C 3 N 4 samples prepared from KOH and dicyandiamide are smaller within the same preparation conditions. Moreover, for the KCl precursor, the crystallite size increases with the increasing KCl concentration, suggesting that the higher amounts of KCl crystals promote the nucleation and growth of g-C 3 N 4 . Further, the sample denoted as KCl(0.04)–CN also exhibited the diffraction peak located around 2θ =33.1◦, which corresponds to KCl indexed as (200) (Fig. 1B). The same diffraction peak was observed for the samples KCl(0.06)–CN and KCl(0.08)–CN, where the new diffraction peak at 2θ =47.5◦appeared and was attributed to the KCl structure (220). This suggests that during the synthesis, at the higher concentration of KCl precursor, the mixture of g-C 3 N 4 and crystalline KCl was formed. The diffraction peaks of pure KOH were not observed. The FTIR spectra of K-modified g-C 3 N 4 are very similar to that of pristine g-C 3 N 4 (Fig. 2). The absorption band located at 801 cm −1 is typical for the tri-s-triazine rings of g-C 3 N 4 . Several other vibration modes are attributed to aromatic CN heterocycles, and the absorption bands at 1315 cm −1 , 1548 cm −1 , 1616 cm −1 , and 3124 cm −1 correspond to C-N, C – – N, and N–H groups. The FTIR spectra of K-modified g-C 3 N 4 show the same bands as the pristine g-C 3 N 4 except tiny absorption bands located at 2177 cm −1 and 2188 cm −1 corresponding to newly formed C≡N groups (Fig. S2). The presence of C≡N groups agrees with previously reported works [19,20,34,38]. The FTIR absorbance values indicate that the chemical structure was not significantly affected by the presence of KOH or KCl at selected concentrations during the synthesis. However, potassium precursors formed a defect, such as the C≡N groups, thus decreasing the polymerization degree of K-modified g-C 3 N 4 samples, which XRD also confirmed. High-resolution S/TEM characterizations with element mapping were used to obtain more extensive information about the morphological structure. The TEM images of K-modified g-C 3 N 4 showed an anisotropic 2D flake-like morphology, where flakes were packed randomly, resulting in a mesopores structure (Fig. 3). The TEM image of pristine gC 3 N 4 was like K-modified g-C 3 N 4 materials (Fig. S3), suggesting that the presence of K and Cl atoms does not influence the morphology on a microscale. The SEM images demonstrated that pristine g-C 3 N 4 and the Kmodified g-C 3 N 4 materials were remarkably similar, both consisting of larger irregular particles that were covered by other smaller particles (Fig. S4). Furthermore, the elemental mapping of KOH(0.04)–CN (Fig. 3A) and KCl(0.04)–CN (Fig. 3B) was conducted, confirming a uniform distribution of all elements (C, N, K, and Cl). In Fig. S5, clear lattice fringes were observed in the higher magnification for pristine gC 3 N 4 , KOH(0.04)–CN, and KCl(0.04)–CN samples, further confirming the presence of crystalline regions in the studied materials. Fig. 1. XRD patterns of pristine g-C 3 N 4 , (A) KOH(X)–CN, and (B) KCl(X)–CN samples. L. Svoboda et al. Applied Surface Science 679 (2025) 161162 4 3.2. Chemical and textural analysis The amount of potassium and chloride in the final samples was investigated by XRF, and additional quantitative analysis was also performed from the obtained XPS data (Table 1). Additional elements, such as C, O, and N from XPS data, are summarized in Table S2. As shown in Table 1, the concentration of both elements increases with the increasing KOH (or KCl) content used during the synthesis. Table 1 also shows that the difference in the potassium content between KOH(X)–CN and KCl(X)–CN of the same potassium precursor concentration is almost negligible. However, a higher content of potassium was observed for the KCl(X)–CN samples. Moreover, a higher potassium content compared to Cl was observed for all the KCl(X)–CN samples. This suggests that the element Cl was not doped into the g-C 3 N 4 crystal lattice at all. Its loss might be due to gas release during the polycondensation thermal process, which would agree with other reports [43,44]. Based on the XPS analysis, the C 1 s peaks in both samples (KOH (0.04)–CN and KCl(0.08)–CN) display similar components and trends, with a small variation in percentage quantifications. Deconvolution of both spectra reveals C – C, C – O, C – – O, and O – – C – O components, as Fig. 2. FTIR spectra of pristine g-C 3 N 4 (A) KOH(X)–CN, and (B) KCl(X)–CN samples. Fig. 3. TEM images of (A) KOH(0.04)–CN and (B) KCl(0.04)–CN with element mapping. Table 1 The concentration of potassium and chloride determined by XRF and XPS, specific surface area (A BET ), and net pore volume (V net ) of pristine g-C 3 N 4 and K-modified gC 3 N 4 samples. Samples K XRF [wt.%] Cl XRF [wt.%] K 2pXPS [wt.%] Cl 2pXPS [wt.%] A BET [m 2 g −1 ] V net [mm 3 lig g −1 ] g-C 3 N 4 −——————— −——————— −——————— −——————— 9.1 37 KOH(0.02)–CN 1.33 −——————— 1.37 −——————— 7.7 43 KOH(0.04)–CN 2.41 −——————— 2.17 −——————— 9.1 49 KOH(0.06)–CN 3.20 −——————— 2.59 −——————— 8.8 57 KOH(0.08)–CN 3.99 −——————— 3.02 −——————— 9.3 64 KCl(0.02)–CN 1.44 0.05 1.46 −——————— 6.6 35 KCl(0.04)–CN 2.70 0.06 2.19 −——————— 6.0 31 KCl(0.06)–CN 3.59 0.26 3.62 0.31 4.5 33 KCl(0.08)–CN 4.14 0.45 3.08 1.02 5.5 33 L. Svoboda et al. Applied Surface Science 679 (2025) 161162 5 detailed in Figure S6. The presence of the K 2p peak in both samples is clear indication that K was successfully doped into g-C 3 N 4 . Furthermore, the sample KCl(0.08)–CN’s N 1 s spectrum shows three main components: pyridinic N, pyrrolic N, and graphitic N, along with a small peak attributed to π - π interactions, referring positive charge localization in heterocycles (Fig. S6c). The Brunauer-Emmett-Teller analysis revealed that the selection of the potassium precursor can have a significant impact on the specific surface area (A BET ) and the net pore volume (V net ) of the final material (Table 1). The A BET values of pristine g-C 3 N 4 and KOH(X)–CN samples remained almost unchanged except for KOH(0.02)–CN. However, the pore volume increased with the increasing content of KOH. Different results were observed for the samples prepared in the presence of KCl. In this case, A BET and V net decreased compared with pristine g-C 3 N 4 . 3.3. Optical and electrochemical properties In optical absorption, an electron gets excited from the top of the valence band into the bottom of the conduction band. The energy difference between these two bands is the energy band gap (E bg ) of a material. Fig. 4 demonstrates the Kubelka-Munk absorption spectra of all the prepared samples. To correctly estimate the E bg values, the Tauc equation was used: ( α h ν )n=A(h υ −Ebg)(3) where α is the absorption coefficient, h ν is the incident photon energy, A is an energy-independent constant, and n depends on the type of transition. Heptazine-based g-C 3 N 4 was found to have an indirect transition [45]. Thus, the Tauc plot was constructed, where a linear portion of the (ah ν ) 1/2 vs. h ν curve is fitted with a straight line and extrapolated to an xaxis to intersect it (Fig. 4 inset). The obtained E bg values of all the samples are summarized in Table S3. Pristine g-C 3 N 4 shows typical strong absorption about 449 nm (2.76 eV). The increasing content of the potassium precursors slightly decreased the band gap energy values from 2.76 eV for pristine g-C 3 N 4 to 2.74 eV for KOH(0.08)–CN, and 2.72 eV for KCl(0.08)–CN, respectively. Such a decrease was more apparent for the samples prepared in the presence of the KCl precursor. Furthermore, the obtained UV–VIS spectra were used to calculate the Urbach energy. The E U values were calculated using equation (4) [46]. α ( ν ) = α 0exp(h υ EU )(4) where α ( ν ) is the absorption coefficient, α 0 is a constant, and E U is the Urbach energy. The values of the E U , in meV, can be derived from the slope of the linear region of the ln( α ) vs. h ν curves (Fig. 4c,d). The Urbach energy values for the KOH(X)–CN samples with the higher potassium content were increasing in the range of 66 – 75 meV and of 73 – 90 meV for KCl(X)–CN samples (Table S3). Here, we observe that the Urbach energy is higher for both Kmodified g-C 3 N 4 samples than that of pristine g-C 3 N 4 . The higher E U value suggests more defects and enhancement of localized state density in the K-modified g-C 3 N 4 samples. Localized defect states are another reason for the narrowing of the band gap in K-modified g-C 3 N 4 samples. In Fig. 5, we can see that the E U values of the KOH(0.02)–CN, KOH (0.04)–CN, and KOH(0.06)–CN samples show a linear trend except for the last sample KOH(0.08)–CN, while the KCl(X)–CN samples show the two-step increasing trend. Similar trends are also evident in the absorbance spectra in the 350 – 400 nm region in Fig. 4. The increasing number of midgap states in the K-modified g-C 3 N 4 samples can be seen in the absorption spectrum as the Urbach tail, which is more apparent in the case of the KCl(X)–CN Fig. 4. UV–VIS absorption spectra of pristine g-C 3 N 4 , (A) KOH(X)–CN and (B) KCl(X)–CN samples. Inset: Tauc plot and the estimated band gap values of the corresponding samples. Urbach energy values of pristine g-C 3 N 4 , (C) KOH(X)–CN, and (D) KCl(X)–CN samples. L. Svoboda et al. Applied Surface Science 679 (2025) 161162 6 samples (Fig. 4b). To gain a deeper understanding of the effect of the selected precursors on the electronic properties of the prepared samples, the valence band energy (E VB ) and the conduction band energy (E CB ) of each sample were studied using the XPS and Mott-Schottky measurements. The XPS spectra of all the samples are depicted in Fig. 6. The valence band XPS values (E VB, XPS ) of all the KOH(X)–CN samples are little bit more positive relative to pristine g-C 3 N 4 (see Tab. S3). The corresponding conduction band energies (E CB ) were calculated by using the XPS data (E VB,XPS ) combined with the UV–VIS DRS ones (E bg ). The obtained results indicate that the presence of potassium precursors influenced the electronic structure, strengthened the lightharvesting ability of the KOH(X)–CN samples, and increased oxidation power of photogenerated holes. However, an opposite effect was observed in the case of the KCl(X)–CN samples, where the E VB became less positive. This trend is more evident for the KCl concentration starting from 0.04 mol L -1 . The increasing content of KCl also shifted the conduction band edge potential towards more negative values. A deviation from the standard redox potential of the O2/•O− 2 pair suggests that the conduction band electrons have an increased ability to form superoxide radicals (•O− 2). To confirm the valence band XPS results, the Mott-Schottky curves (Fig. S7) were also measured. The conduction band energy (E CB, MottSchottky ) values were further used to determine E VB by applying E bg values as well. The data from Table S3 were used to construct energy diagrams for all the samples (Fig. S8). The Mott-Schottky method entails the use of a bias voltage to modify the width of the space-charge interfacial double layer and, consequently, the capacitance value when the solution side of the interface can be approximated by the Helmholtz model of the double layer [47]. Equation (5) of the Mott-Schottky formula elucidates the correlation between the applied potential and the space charge capacitance. 1 C2=2 ε r ε 0eND(V−VFB −kT e)(5) where ε r is the dielectric constant, ε 0 is the dielectric constant of vacuum, ND is the dopant density for an n-type semiconductor, T is the temperature, K is the Boltzmann constant and e is the elementary electron charge. To determine flat band potentials (VFB), intercepts of extrapolation lines with the potential axis in the Mott-Schottky plot were calculated. This plot is a function of the inverse of measured capacitance squared on the potential. In the case of n-type semiconductors, the flat band potential is considered to be almost identical to the conduction band edge potential (VCB) as per equation (6) [48,49]. VCB ≈VFB(NHE)=VFB +ΔV−0.059 • (7−pH0.1 M KCl)(6) where VFB(NHE)is the flat band potential recalculated against NHE electrode at pH=7, ΔV is the potential of Ag/AgCl reference electrode vs. NHE. The valence band edge potential (VVB)was calculated using equation (7) [48,49]. VVB =VCB +Ebg e(7) Upon interaction with photocatalysts, light induces the formation of holes in the valence band and transfers electrons to the conduction band. The chemical reactions leading to the formation of •O− 2 and •OH can be attributed to previous literature (equations (8)–(11). OH−+h+⟶•OH (8) H2O+h+⟶•OH (9) O2+e−⟶•O− 2(10) H++ • O− 2⟶•HO2(11) Fig. 5. Dependence of Urbach energy on potassium concentration in prepared samples. Fig. 6. XPS valence band spectra of pristine g-C 3 N 4 , (A) KOH(X)–CN, and (B) KCl(X)–CN samples. L. Svoboda et al. Applied Surface Science 679 (2025) 161162 7 Figure S8 displays the CB and VB edge potentials, and the band gap values. Notably, the CB edge potential energy is consistently above −0.33 V for all the materials, implying that they can form superoxide radicals. This ability was more or less uniform across all the KOH(X)–CN samples. Conversely, the valence band edge potential position is lower than the standard ⋅OH/OH – redox potential, ruling out the possibility of hydroxyl radical formation through the abovementioned mechanism. This is especially true for the KCl(X)–CN samples. Comparing the data obtained by the XPS and Mott-Schottky measurements, we can conclude that they agree and that both analyses proved different trends in the band alignment depending on the potassium precursor used and its concentration during the synthesis. The precursors’ choice and their concentration play a significant role in the band gap structure. It can be deduced from the measured and the calculated values of E VB that the increasing potassium content in g-C 3 N 4 does not simply lead to such high positive oxidation potential values for E VB that photogenerated holes would be able to form hydroxyl radicals, and thus, the obtained results show a discrepancy with other published results [9–11,14]. According to the obtained E VB values, pristine g-C 3 N 4 showed the value of E VB to be +1.88 eV, about 0.32 eV higher than reported [9]. The higher E VB value for pristine g-C 3 N 4 reported in this work is in good agreement with other published papers which also dealt with the synthesis of g-C 3 N 4 by using dicyandiamide as a precursor: +1.91 eV [50], +1.99 eV [51], +1.89 eV [52]. It is important to note that the use of different precursors, such as urea and KBr, can lead even to an opposite effect, i.e., the shift of E VB to less positive oxidation potential [19], and in the case of melamine and KOH, no change in the position of E VB was observed based on XPS measurements [20]. The final electronic structure of the resulting gC3N4 is influenced not only by the potassium content in the g-C3N4 structure but mainly by the selection of potassium precursor, which significantly impacts the Urbach tails and the band gap structure. The theoretical calculation of the partial density states of g-C 3 N 4 materials that were modified by alkaline-earth metal barium confirmed that the minimum of CB and a maximum of VB of such materials are formed solely only by C 2p and N 2p orbitals [53]. The same can be expected in the case of the potassium-modified g-C 3 N 4 samples. These findings highlight the complexity and diversity of modified g-C 3 N 4 materials, which although all can be denoted as K-modified g-C 3 N 4 materials, their properties can be opposed. The photoluminescence spectroscopy was used to examine the steady-state emission spectra. From Fig. 7, it can be seen that pristine gC 3 N 4 exhibited an intense emission band with a maximum at 449 nm, while the increased amount of potassium precursor shifted the emission band maximum to higher wavelengths. The observed redshift is in good agreement with the UV–VIS results. All the K-modified g-C 3 N 4 samples also showed a progressive loss of fluorescence signal with the increasing amount of the potassium precursor leading to the increased content of potassium and C≡N group (strong electron-withdrawing group) (Fig. S2). This indicates that the presence of both, potassium and C≡N groups, is beneficial for separating of photogenerated charge carriers, reducing the probability of radiative electron-hole recombination. 3.4. Photocatalytic activity In our study, samples were tested under a single wavelength (416 nm) irradiation to avoid strong light absorption by Rh B (450 nm <λ < 600 nm), thus eliminating the known photosensitisation effect [54]. This means that the decomposition of Rh B by g-C 3 N 4 should only be attributed to photocatalysis and not to the photosensitization process. The results of the Rh B degradation in an aqueous solution under visible light irradiation are shown in Fig. 8a and summarized in Table S4. The rate constant of the Rh B degradation was calculated by using a pseudo first-order kinetic equation (12): ln(c0 ct)=kobst(12) where c 0 and c t are the concentration of Rh B before and after irradiation, t is the irradiation time, and k obs is the observed pseudo first-order rate constant. All the KOH(X)–CN and KCl(X)–CN samples exhibited higher photocatalytic activity than the pristine g-C 3 N 4 sample. However, the increase in degradation rate was mainly higher for the KOH(X)– CN samples compared to the KCl(X)–CN ones. A noticeable decrease in the kinetic rate for the KOH(0.06)–CN sample was observed. To confirm the reproducibility of this drop in the photocatalytic activity, a new KOH(0.06)–CN sample, denoted as KOH (0.06)–CN-2, was prepared and tested in the degradation of Rh B. The obtained kinetic data confirmed that this decrease in photocatalytic activity is reproducible and thus related to the amount of potassium precursor used during the synthesis (Fig. S9). It is not a measurement or preparation error. It was found that the KOH(0.02)–CN and KOH(0.04)– CN samples had the fastest kinetic rates. In addition to the enhanced photocatalytic activity, the K-modified gC 3 N 4 materials also showed good stability and reusability in the four subsequent cycles of the Rh B photodegradation. After each cycle, the KOH(0.04)–CN material was separated and reused without additional treatment. As presented in Fig. 8b, the degradation efficiency remained stable, confirming its good stability and recyclability. The slight decrease in the photocatalytic activity is attributed to the inevitable loss of the photocatalyst mass during the recycling process. Besides, the FTIR spectra of the KOH(0.04)–CN sample before and after the fourth photocatalytic cycle did not differ (Fig. S10). 3.4.1. Mechanism of the photocatalytic degradation In general, the increase in photocatalytic activity can be so far Fig. 7. Photoluminescence spectra of pristine g-C 3 N 4 , (A) KOH(X)–CN, and (B) KCl(X)–CN samples (λ exc =385 nm). L. Svoboda et al. Applied Surface Science 679 (2025) 161162 8 related to several factors: a) improved electron-hole separation efficiency according to steady-state photoluminescence spectra, where the electron-hole radiation recombination probability was significantly suppressed, and b) better harvesting of used visible light irradiation due to an enhancement of localized state density, which also causes a narrowing of the band gap in photocatalysts. Given that TCSPC measurements provide more information than the steady-state emission spectra and that the fluorescence lifetime is typically independent of the sample concentration [55], we measured the fluorescence decays of all samples. In Fig. S11, the fluorescence decay curves of all samples under 375 nm excitation wavelength are depicted with instrument response function (IRF) and three-exponential fit curves. The obtained decay curves do not follow the single-exponential decay, which indicates the presence of the shorterand longer-lived excited states [56,57]. The obtained data from TCSPC are summarized in Table S5. To gain a better understanding of the resulting photocatalytic behavior of K-modified samples, the construction of the plots Fig. 9 a) and b) was necessary. For both kinds of samples, a similar dependence of Rh B removal on the specific surface area of the material can be observed (Fig. 9a). For KOH(0.02)–CN, although there was a decrease in the A BET value compared to the pristine g-C 3 N 4 , its activity significantly increased. The difference in the concentration of C≡N groups between both materials is negligible (see Fig. S2a). This leads us to conclude that C≡N groups are not responsible for the enhanced photocatalytic activity. The increasing content of potassium precursors introduced more C≡N groups in g-C 3 N 4 , which resulted in higher Urbach energy (Fig. 5). This became apparent as a narrowing of the band gap. However, the average carrier lifetime is reduced by the increased amount of K and C≡N groups (Fig. 9b). As the number of defects increased, their negative effect on the resulting Rh B degradation became more apparent. Although the A BET of the KOH(0.04)–CN sample increased, the efficiency of Rh B degradation increased significantly less compared to KOH (0.02)–CN. For the KOH(0.06)–CN sample, it can be observed that the shortened lifetime and the accompanying decrease in A BET led to a much more significant decrease in photocatalytic activity in the series of the KOH-based samples. However, the A BET value of the KOH(0.08)–CN sample rose again and was higher than that of the KOH(0.02)–CN and the KOH(0.04)–CN samples, its photocatalytic activity was found to be lower. This suggests that higher KOH precursor concentrations led to more recombination centers, resulting in decreased lifetime and, consequently, photocatalytic activity. A similar dependence can be observed for the KCl(X)–CN sample. However, the significant difference between the KOHand KCl-based materials is that the average lifetime of KCl(X)–CN samples was not as significantly reduced as for the KOH(X)– CN (Fig. 9b). Furthermore, in the case of the KCl(0.06)–CN sample, the A BET and the average lifetime decreased. This had the same effect on the observed drop of photocatalytic activity as in the case of the KOH(0.06)–CN sample in its KOH(X)–CN series. Two critical questions should be Fig. 8. (A) Photocatalytic activity of pristine g-C 3 N 4 , and K-modified g-C 3 N 4 samples (λ LED =416 nm), m photocatalyst =50 mg, c Rh B =5 ppm, V=50 ml. (B) The stability and reusability tests of KOH(0.04)–CN material in the four subsequent cycles of Rh B photodegradation. Fig. 9. (A) Plots of A BET and Rh B removal related to individual materials. (B) Plot of K concentration, Urbach energy, and average lifetime related to individual materials. L. Svoboda et al. Applied Surface Science 679 (2025) 161162 9