Introduction

The continuous release of pharmaceutical residues into aquatic environments has become a critical environmental issue, primarily due to their persistence, biological activity, and insufficient removal by conventional wastewater treatment processes [1,2]. Among the various classes of pharmaceuticals, fluoroquinolone antibiotics and particularly ciprofloxacin (CIP), are consistently reported in municipal wastewater, hospital effluents, and surface waters worldwide. This widespread occurrence is largely attributed to the extensive clinical use of CIP and its inherent resistance to biodegradation [3]. Even at low concentrations, CIP has been shown to induce ecotoxicological effects, interfere with microbial ecosystems, and contribute to the development of antimicrobial resistance, underscoring the necessity for more effective treatment strategies [4, 5].

The limited removal of ciprofloxacin in conventional activated sludge systems can be primarily ascribed to its high chemical stability and poor biodegradability [4]. As a result, increasing research attention has been directed toward advanced treatment technologies that are capable of achieving complete mineralization rather than merely transforming the parent compound into potentially persistent by-products [6]. In this context, advanced oxidation processes (AOPs) have emerged as a particularly attractive option because they rely on the
in-situ generation of highly reactive oxygen species (ROS), such as hydroxyl (·OH) and superoxide (·O₂⁻) radicals, which can non-selectively oxidize recalcitrant organic contaminants [7].

Among AOP-based techniques, heterogeneous photocatalysis has been widely investigated for the degradation of persistent pharmaceutical pollutants through light-induced redox reactions. However, commonly used photocatalysts such as TiO₂ and ZnO suffer from intrinsic limitations, including restricted spectral response, rapid recombination of photogenerated charge carriers, and low quantum efficiency under practical operating conditions [8]. In recent years, graphitic carbon nitride (g-C₃N₄) has attracted considerable interest as a metal-free semiconductor owing to its favorable band structure, chemical stability, and potential for visible-light-driven photocatalysis [9,10]. Despite these advantages, the photocatalytic performance of pristine g-C₃N₄ remains constrained by fast electron hole recombination and a relatively low specific surface area.

To address these limitations, coupling g-C₃N₄ with magnetic iron oxide (Fe₃O₄) nanoparticles has been proposed as an effective strategy. The incorporation of Fe₃O₄ can promote interfacial charge transfer, suppress recombination of charge carriers, and simultaneously enable rapid magnetic separation of the photocatalyst after treatment, thereby improving both catalytic efficiency and practical applicability [11,12].

On this basis, the present study focuses on: (I) the synthesis and characterization of a magnetically recoverable g-C₃N₄/Fe₃O₄ nanocomposite, (II) the systematic evaluation of its photocatalytic performance for ciprofloxacin degradation under UV-C irradiation, and (III) the optimization of key operational parameters using response surface methodology (RSM) to determine conditions leading to maximum COD (chemical oxygen demand) removal.

Literature Review

The environmental occurrence and persistence of antibiotics, particularly fluoroquinolones, have been widely reported in recent years. Among them, ciprofloxacin is frequently detected in wastewater effluents and surface waters and is regarded as a compound of significant ecological concern due to its chemical recalcitrance and biological activity [1-3]. Because conventional wastewater treatment systems are often unable to achieve complete ciprofloxacin removal, increasing attention has been directed toward advanced treatment strategies capable of promoting complete mineralization rather than partial transformation [4].

Photocatalytic advanced oxidation processes (AOPs) represent one of the most extensively studied approaches for antibiotic abatement, as they rely on the generation of ROS that can effectively degrade recalcitrant organic pollutants [6]. Early investigations predominantly focused on metal oxide photocatalysts such as TiO₂ and ZnO; however, practical limitations related to restricted spectral response, rapid charge carrier recombination, and narrow operational pH windows have motivated the exploration of alternative semiconductor materials [9].

In this context, graphitic carbon nitride (g-C₃N₄) has gained increasing interest as a metal-free photocatalyst with favorable chemical stability and visible-light responsiveness [13]. Nevertheless, its practical photocatalytic efficiency remains limited by fast electron–hole recombination. To overcome this drawback, hybridization with magnetic Fe₃O₄ nanoparticles has been proposed, leading to the development of g-C₃N₄/Fe₃O₄ heterostructures. Such composites have been reported to exhibit improved charge separation, enhanced ROS generation, increased adsorption capacity, and convenient magnetic recovery after treatment [11, 14].

Recent studies have demonstrated that g-C₃N₄/Fe₃O₄-based photocatalysts are effective for the degradation of dyes, pharmaceuticals, and antibiotics under both UV and visible irradiation, often outperforming the individual components [15]. With respect to ciprofloxacin, g-C₃N₄-based binary and ternary systems incorporating metal oxides or plasmonic materials have shown enhanced degradation and mineralization efficiencies, which have been attributed to improved light harvesting, interfacial charge transfer, and suppressed electron–hole recombination [5,16].

Despite these advances, photocatalytic performance remains highly sensitive to operational parameters such as photon flux, catalyst loading, solution pH, and pollutant concentration. Statistical optimization tools, including RSM coupled with CCD, are therefore increasingly employed to quantify variable interactions and establish predictive models for COD and TOC (total organic carbon) removal [17-19]. However, substantial discrepancies in reported optimal conditions persist across studies, photon fluence is often insufficiently described, and the assessment of transformation product toxicity is frequently neglected. Addressing these limitations through systematic optimization and mechanistic analysis is essential for advancing photocatalytic processes toward reliable scale-up and practical implementation.

Materials and Methods

Chemicals and Materials:

All chemicals used in this study were of analytical grade and were used as received without further purification. Ciprofloxacin hydrochloride (C₁₇H₁₈FN₃O₃, purity ≥98%) was purchased from Sigma-Aldrich (USA)
(Fig. 1). Sodium hydroxide (NaOH) and hydrochloric acid (HCl), employed for pH adjustment during the photocatalytic experiments, were also obtained from Sigma-Aldrich. UV-C lamps with nominal power outputs of 10, 20, and 40 W (λ = 254 nm) were supplied by Philips (Germany) and served as irradiation sources throughout the study.

Graphitic carbon nitride (g-C₃N₄) was synthesized by thermal polymerization of melamine following a conventional procedure. Fe₃O₄ nanoparticles were prepared via a co-precipitation method based on previously reported protocols. The g-C₃N₄/Fe₃O₄ magnetic nanocomposite was subsequently synthesized using a solid-state coupling strategy, with a post-calcination step applied to facilitate intimate interfacial contact, thereby promoting heterojunction formation between the two components. All synthesis procedures were carried out under carefully controlled conditions to ensure consistency and reproducibility.

Fig. 1. Chemical structure of ciprofloxacin

Characterization:

The morphology and microstructural characteristics of the g-C₃N₄/Fe₃O₄ nanocomposite were examined using scanning electron microscopy (SEM, Hitachi S-4160). The crystalline structure and phase composition were analyzed using X-ray diffraction (XRD) on a Philips X’Pert PW 3040/60 diffractometer equipped with 
Cu Kα radiation (λ = 1.5406 Å). XRD measurements were performed to identify the characteristic diffraction peaks of g-C₃N₄ and Fe₃O₄ and to confirm the successful formation of the composite material. Representative SEM micrographs and XRD patterns of the synthesized nanocomposite are shown in Fig. 2.

Experimental Design and Optimization:

Process optimization was performed using RSM combined with a CCD to systematically evaluate both the individual and interactive effects of key operating parameters on photocatalytic performance. Based on preliminary experiments and relevant literature, three independent variables were selected: initial ciprofloxacin concentration (A, 5-120 mgL⁻¹), catalyst dosage (B, 0.05-0.25 gL⁻¹ of g-C₃N₄/Fe₃O₄), and solution pH (C, 4-10).

 The CCD matrix consisted of 20 experimental runs, including factorial points, axial points, and replicated center points, and was constructed according to standard RSM design principles. COD removal efficiency was chosen as the response variable (Y) to reflect the extent of mineralization achieved during the photocatalytic process. The experimental results were subsequently fitted to a second-order polynomial model, as expressed in Eq. (1):

Where: Where Y is the response variable (COD removal efficiency), xᵢ and xⱼ denote the coded independent variables, β₀ is the intercept, βᵢ is the linear coefficient, βᵢⱼ is the interaction coefficient, βᵢᵢ is the quadratic coefficient, and K is the number of independent variables (K = 3).

 Batch Photocatalytic Experiments:

Photocatalytic degradation experiments were performed in a custom-built batch reactor fabricated from UV-transparent Plexiglas. The reactor consisted of a cylindrical reaction chamber with an effective working volume of 500 mL. To ensure stable thermal conditions during irradiation, the main chamber was surrounded by a secondary water-jacketed compartment (3 L capacity) connected to a recirculating water bath, maintaining the reaction temperature at 27 ± 2 °C.

UV-C irradiation was provided by vertically mounted low-pressure mercury lamps with nominal power outputs of 10, 20, and 40 W (λ = 254 nm). To enhance photon utilization within the reactor, all internal surfaces were lined with aluminum foil to promote light reflection. Continuous agitation was provided by a magnetic stirrer throughout the experiments to minimize external mass-transfer limitations during the photocatalytic process.

All photocatalytic experiments were performed according to the 20-run central composite design described in the previous section. Each batch experiment was conducted for 60 min, and all experimental conditions were repeated in triplicate to ensure reproducibility.

A fresh ciprofloxacin stock solution was prepared daily using deionized water and stored at 4 °C in the dark. The desired initial concentrations were obtained by appropriate dilution of the stock solution prior to each run. For each experimental run, the required volume of ciprofloxacin solution and the prescribed amount of g-C₃N₄/Fe₃O₄ nanocomposite were introduced into the reactor, and the solution pH was adjusted using standardized 1.0 M NaOH or 1.0 M HCl solutions. Before initiating UV-C irradiation, the suspension was magnetically stirred in the dark for 30 min to establish adsorption–desorption equilibrium between ciprofloxacin and the photocatalyst surface.

At the end of the irradiation period, the photocatalyst was rapidly separated from the treated solution by magnetic decantation and, when necessary, further assisted by centrifugation at 3000 rpm for 10 min. The collected supernatant was subsequently filtered through a 0.22 µm PTFE syringe filter (Schleicher & Schuell, Germany) to remove any remaining suspended particles.

COD was determined using a HACH DR-5000 UV–Vis spectrophotometer following Standard Method 5220D (APHA, 2005). The COD removal efficiency (Y, %) was calculated according to Eq. (2):

Where:

C₀ and C_e are initial and final COD concentrations.

 Statistical Analysis:

The experimental data were analyzed using RSM via multiple regression modeling implemented in Design-Expert software (version 7.0). The adequacy of the fitted second-order polynomial models was evaluated by analysis of variance (ANOVA), including Fisher’s
F-test, lack-of-fit assessment, and significance testing of individual model coefficients. Statistical significance was defined at a 95% confidence level (p < 0.05).

The predictive performance and reliability of the developed models were further evaluated using the coefficient of determination (R²), adjusted R², predicted R², and adequate precision. These statistical indicators were used to verify the robustness of the regression models and ensure that the RSM provided reliable predictions within the studied experimental domain.

 Results

Model Fitting and Statistical Analysis:

The experimental and predicted COD removal efficiencies for the 20 experimental runs used to develop the regression models are summarized in Table 2. To account for differences in photon flux, separate CCD-based RSM models were developed for each UV-C lamp power, thereby enabling evaluation of the effects of operating parameters on COD removal efficiency under varying irradiation intensities.

Several regression models were examined, and a second-order polynomial model was identified as the most appropriate for predicting COD removal efficiency as a function of ciprofloxacin concentration (A),
g-C₃N₄/Fe₃O₄ dosage (B), and solution pH (C). The model incorporates linear (A, B, C), interaction (AB, AC, BC), and quadratic (A², B², C²) terms.

Model adequacy and predictive performance were evaluated using the coefficient of determination (R²), adjusted R², predicted R², F-values, and lack-of-fit tests. The R² values were 0.8371, 0.8372, and 0.882 for the 10, 20, and 40 W systems, respectively, with close agreement between adjusted and predicted R² values, indicating satisfactory predictive reliability. In all cases, lack-of-fit p-values exceeded 0.05, confirming the statistical validity of the models. The corresponding
F-values were 5.71, 5.71, and 8.31, with p-values of 0.0059, 0.0059, and 0.0014, demonstrating the statistical significance of the regression models (Tables 3-5).

Model coefficients were estimated using multiple regression analysis, and the resulting second-order polynomial equations including only statistically significant terms (p < 0.05) are presented in Eqs. (3-5). The close agreement between experimental and predicted COD removal efficiencies across all experimental runs confirms that the developed models adequately describe the photocatalytic behavior of the
g-C₃N₄/Fe₃O₄ nanocomposite under UV-C irradiation.

Table 2. Central composite design (CCD) and observed responses

Effect of Various Parameters on COD Removal:

The ANOVA results for the 10, 20, and 40 W UV-C lamps are summarized in Tables 3-5, respectively. Contour plots were employed to visualize the interactions between two independent variables while maintaining the third variable at its central level, thereby facilitating a clearer interpretation of parameter interdependencies.

Effect of CIP Concentration: Figure 3 illustrates the combined influence of ciprofloxacin concentration and
g-C₃N₄/Fe₃O₄ dosage at pH 7. COD removal efficiency increased with increasing CIP concentration up to approximately 72 mgL⁻¹. Beyond this concentration, a plateau was observed, which can be attributed to saturation of active sites on the catalyst surface and limited photon availability to sustain further photocatalytic reactions.

Effect of g-C₃N₄/Fe₃O Dosage: Increasing the catalyst dosage enhanced COD removal efficiency until an optimum value of 0.23 gL⁻¹ was reached. Further increases in catalyst loading resulted in a decline in removal efficiency, primarily due to increased light scattering and shielding effects, which reduced effective photon penetration and limited uniform irradiation of the catalyst surface.

Effect of Solution pH: The photocatalytic performance was highly dependent on solution pH (Figs 4-5). Maximum COD removal was achieved within the pH range of 5-9, with an optimum near pH ~7. This behavior can be attributed to the surface charge of g-C₃N₄/Fe₃O₄ and the acid–base behavior of ciprofloxacin, affecting adsorption and photocatalytic degradation.

Figure 5 further highlights the interactive effect of catalyst dosage and solution pH on COD removal efficiency, indicating that optimal performance results from a balance between catalyst loading and solution conditions.

Model Validation: Numerical optimization using the desirability function predicted maximum COD removal efficiencies of 82.3%, 90.8%, and 98.5% for the 10, 20, and 40 W UV-C lamps, respectively, at a CIP concentration of 8.05 mg L⁻¹, catalyst dosage of 0.23 gL⁻¹, and pH~7.18. Experimental validation under these conditions yielded COD removal efficiencies of 82.2%, 85.2%, and 87.1%, respectively.

Despite the larger deviation observed for the 40 W system, experimental trends generally followed the predicted values, confirming the practical reliability of the developed models (Table 6). The observed deviation at higher lamp power may be attributed to localized heating or non-uniform light distribution, potentially affecting photon utilization efficiency.

Table 3. Analysis of variance (ANOVA) for the fitted second-order polynomial model of COD removal under UV-C irradiation (10 W lamp)

Table 4. Analysis of variance (ANOVA) for the fitted second-order polynomial model of COD removal under UV-C irradiation (20 W lamp)

 Table 5. Analysis of variance (ANOVA) for the fitted second-order polynomial model of COD removal under UV-C irradiation (40 W lamp)

Fig. 3. Effect of CIP concentration and g-C₃N₄/Fe₃O₄ dosage on COD removal efficiency

Fig. 4. Effect of CIP concentration and solution pH on COD removal efficiency

Fig. 5. Effect of g-C₃N₄/Fe₃O₄ dosage and solution pH on COD removal efficiency

Table 6. Optimization results of independent variables and corresponding predicted and experimental responses