Response surface optimised photocatalytic degradation and quantitation of repurposed COVID-19 antibiotic pollutants in wastewaters; towards greenness and whiteness perspectives

The analysis was performed using a Shimadzu LC-MS/MS with a pump (model LC-20AD). The detection was performed using a triple quadrupole mass spectrometer (API 4000) equipped with an electrospray ionisation (ESI) source using multiple reaction monitoring (MRM) in positive ion mode. An autosampler was used with a sampling speed of 15 µL s^–1 (Shimadzu SIL20A) and a pump (Shimadzu LC20AD). The data acquisitions were performed using Analyst 1.6.3 software. Oasis HLB Prime cartridges (500 mg/6 mL, Waters, MA, USA) were used for sample treatment.

For characterisation of g-C₃N_4, a JEOL JEM-2010 transmission electron microscope, a Shimadzu UV-2450 spectrophotometer, a Malvern Zetasizer Nano ZS90 analyser, a Shimadzu RF5301PC spectrofluorometer, an FTIR spectrophotometer (JASCO spectrometer), a Shimadzu XRD 6100 diffractometer, a Jeol model JSM-IT100 scanning electron microscope, and a Quantachrome TouchWin™ (ver. 1.2) surface analyser were used.

A reference standard of LEVO (99.54 ± 0.67) was kindly supplied by the Sanofi pharmaceutical company (Egypt). A reference standard of ceftriaxone sodium (99.58 ± 0.56) was kindly supplied by EIPICO pharmaceutical company (Egypt). Methanol, acetonitrile (HPLC grade), formic acid, melamine (≥ 99%), a reference standard of azithromycin dehydrate (99.78 ± 0.73), and erythromycin as an internal standard (IS) were purchased from Sigma–Aldrich, USA. Sodium hydroxide and hydrochloric acid (33%) were obtained from ADWIC (Egypt). The purity of the reference standards was checked using the official British Pharmacopoeia (BP) methods (British Pharmacopoeia Commision 2013).

The separation and identification were performed using an Agilent C₁₈ column (4.6 mm × 50 mm, 4.6 µm particle size). The mobile phase consisted of 80% acetonitrile/20% (0.1% formic acid in water) in isocratic elution mode delivered at a flow rate of 0.8 mL min^–1. The column temperature was set at 30°C and the amount of sample injected was 5 µL. A mass spectrometer (API 4000) equipped with a triple quadrupole was used and the detection was acheived using positive ion electrospray ionisation (ESI+), with air as a nebuliser gas and nitrogen as an auxiliary curtain and collision gas. The gas temperature was set at 300°C. The MRM transitions for LEVO, AZI, CEF and erythromycin (IS), and mass spectrometric parameters, are listed in Table 1.

Stock standard solutions of LEVO, AZI, CEF and erythromycin (IS) were prepared in methanol to a concentration of 100 μg mL^–1. Freshly prepared working solutions for the three drugs were prepared by dilution of the stock solution using the same solvent. The calibration curves for the three drugs were prepared using concentrations of 10–500 ng mL^–1 in triplicate for each drug. A calibration curve was plotted for each drug relating the prepared concentrations versus the peak area ratio of the corresponding drug to that of the IS. The three calibration curves were plotted, and regression equations were computed. Validation of the analytical method was performed as per ICH guidelines including the parameters of linearity, range, accuracy and precision, specificity, limit of quantification (LOQ) and limit of detection (LOD) (ICH Harmonisation Tripartite Guideline 2005).

In 250-mL amber glass bottles, 200 mL of water samples were collected from different locations and stored at −15°C for 10 days till extraction and analysis. These water samples were filtered through Whatman No. 41 and 0.7-µm filters prior to the preconcentration process, at the original pH (7–7.5) using an Oasis HLB Prime cartridge (500 mg/6 mL). Loading the cartridges was done at a flow rate of 1 mL min^–1 using ultra-water samples for a sufficient contact time to allow targeted antibiotics to adsorb on the sorbent material. The sorbent was washed with 10 mL of water and then dried under vacuum and eluted with methanol (3 × 10 mL). The solvent was evaporated under nitrogen and the water samples were analysed according to the chromatographic conditions.

The pristine g-C₃N₄ powder was synthesised in a one-step process by thermally polymerising melamine (Svoboda et al. 2019) following the literature procedure with a simple modification: 5 g of melamine was weighed and placed in a porcelain crucible, which was then covered and heated in a muffle furnace for 4 h in an air atmosphere at 550°C with a heating rate of 2°C min. The resulting, yellow-coloured product was kept at room temperature for later investigations and use.

The morphological characteristics were analysed using a transmission electron microscope, operated at anaccelerating voltage of 200 kV, and included a Gatan digital camera (Model Erlangshen ES500). A Zeta sizer set to 25°C was used to determine the zeta-potential. The optical features of the solid sample were recorded using a spectrophotometer with 1 cm quartz cells. The g-C₃N₄ photoluminescence (PL) emission spectra were captured utilising a spectrofluorometer. The structural information and functional groups were identified using an FTIR spectrophotometer in the 4000–400 cm⁻¹ range. X-Ray diffraction data were collected using a diffractometer with CuKa1 radiation (k = 1.54056 A), a 2θ range of 10–80, an operating data voltage of 40 kV, and a current of 30 mA. Energy-dispersive X-ray spectrometry (EDX) was utilised using a scanning electron microscope (SEM). The BET surface area of the sample was calculated employing a multipoint Brunauer–Emmett–Teller (BET) method using adsorption data relative pressure (P/P₀) in the range of 0.05–0.9. The particle pore size distribution was measured by the Barrett–Joyner–Halenda (BJH) method using adsorption data, both methods were performed on a Quantachrome Nova instrument running TouchWin™ software (ver. 1.2).

The treatment of the studied antibiotics was undertaken using synthesised g-C₃N₄ nanosheets under a LED lamp (35 W) which produced visible light irradiation. Screening experiments were performed under variable conditions including pH ranges, initial concentrations of the three drugs, dose of g-C₃N₄, agitation rate, and time intervals in the dark to obtain an adsorption–desorption equilibrium, where one condition was changed while the others were kept constant. Two millilitres of each suspension was taken and centrifuged to remove the photocatalyst powders before measurement. The sample was analysed using the developed LC-MS/MS procedure. The screening experiments revealed the factors that will be included in the experimental design.

Box–Behnken experimental design (BBD), three-levels and three-factors, was employed for assigning the optimum conditions for maximising the removal of the three selected antibiotics, LEVO, AZI and CEF as dependent variables, against the dose of g-C₃N₄ (0.2–1.0 g L^–1), pH (5–9) and time (30–90 min) as independent variables. Each variable was represented in three levels, high (+1), medium (0) and low (−1). Fifteen experimental points were performed including replications of the central points in randomised experimental order, as shown in Supplementary Table S2. ANOVA data analysis was performed to determine the lack of fit, and evaluate other parameters including the effects of quadratic, linear and variables’ interaction, where the model was verified by response surface methodology (RSM).

Optimum experimental conditions were applied to three types of water. The first type of water sample was a laboratory distilled water spiked with equal amounts of the three cited pharmaceuticals (500 ng mL^–1). The second one was a real water sample free from pharmaceutical pollutants (clean real water) and spiked with equal amounts of the three cited pharmaceuticals (500 ng mL^–1). The third were contaminated real water samples collected from the Nile River and industrial effluents collected from two cities named 6th of October and 10th of Ramadan. This approach allowed monitoring the degradation under realistic working conditions compared to the laboratory experiments.

Several mobile phase components with different ratios (methanol, acetonitrile, water and acidified water with formic acid) were trialled to elute symmetrical peaks with accepted tailing factors for the three drugs and IS. A C₁₈ column and a mobile phase consisting of 80% acetonitrile/20% (0.1% formic acid in water) in isocratic elution mode at a flow rate of 0.8 mL min^–1 gave optimum results as formic acid helped in the positive ionisation of the drugs. Regarding the mass spectrometric detector, the positive ion mode was selected for MRM analysis where the parent ions were selected at m/z 362.12 > 318.3 for LEVO, 555.2 > 396 for CEF, 749.61 > 83.0 for AZI and 734.55 > 158 for IS, as shown in Fig. 1a, b.

Fig. 1.

(a) Product ion spectra of [M + H]⁺ of the positive ion ESI-MS/MS spectra: LEVO 362.12 > 318.3 (m/z), AZI 749.61 > 83.0 (m/z), CEF 555.2 > 396 (m/z) and IS 734.55 > 158 (m/z); (b) multiple reaction monitoring (MRM) chromatograms of the three drugs and IS.

The validation parameters are presented in Table 2. The linearity range was 10–5000 ng mL^–1 for AZI, CEF and LEVO as shown in Supplementary Fig. S1. The accuracy of the method was confirmed by calculating the recovery % of each drug using the corresponding regression equations. The inter- and intra-day precision scored %RSD <2 units. The method was specific as it could determine the three drugs simultaneously without interference in the laboratory prepared mixtures. The LOD of the drugs was ~3.5 ng mL^–1 and LOQ was ~9 ng mL^–1, which prove the high sensitivity of the method.

Oasis Prime HLB cartridges are preferred over HLB cartridges for SPE because they eliminate the conditioning step, and thus reduce solvent consumption and sample preparation time. They were also reported to have higher extraction recoveries (40%) than HLB cartridges that are reported to be commonly used for sample preconcentration of pharmaceuticals in complex environmental matrices (Elbalkiny et al. 2019; Yehia et al. 2019), even though the sorbent is the same in both cartridges (N-vinylpyrrolidone and divinylbenzene copolymer). As a result, Oasis Prime HLB cartridges at the original sample pH were chosen as they enabled a greener and better recovery of the selected drugs using a single solvent elution with pure methanol after spiking all water samples with 100 ng of each drug standard. The analysis of laboratory-spiked distilled water yielded the recovery of cartridges, and it was found to be 86.3, 83.8 and 89.1% for LEVO, AZI and CEF, respectively. The contaminated real water samples showed the presence of one or two of the cited drugs in each sample as shown in Table 3, while the clean real samples didn’t show any traces of the three drugs.

Based on the above characterisation results, the bandgap of g-C₃N₄ is relatively narrow, so it can generate electron–hole pairs under visible light irradiation. The photoexcited electrons in the conduction band (CB) of g-C₃N₄ can react with O₂ in air to produce active superoxide radicals (·O₂⁻), where this generated species can react with electrons to form H₂O₂. H₂O₂ then reacts with electrons to form hydroxyl radicals (OH). O₂⁻, OH and H₂O₂ generated by g-C₃N₄ are involved in converting antibiotic pollutants into carbon dioxide and water (Dong et al. 2022), as shown in Supplementary Fig. S5.

The performance of the photocatalytic activity of g-C₃N₄ depends on different variables. Different pH ranges, initial concentrations of the three drugs, dose of g-C₃N₄, agitation rate and time intervals were studied. After the exploratory runs, it was found that the initial concentration of the studied antibiotics and the agitation rate had insignificant effects on the degradation rate. Meanwhile, the pH of the solution, dose of g-C₃N_4, and time intervals played a remarkable role in the photocatalytic reactions. As such, these three variables were included in the optimisation study that was designed via response surface methodology.

The quadratic statistical models were suggested with the highest order polynomial based on the experimental design results and statistical analysis performed on the three responses (highest F value and lowest P value). The model results obtained by experimental design are shown in Eqns 1–3:

where A is the pH, B is the dose of g-C₃N₄ in g L^–1 and C is the time in minutes. The sign indicates the effect of the variable. A positive sign indicates a direct effect for B, C, AB and BC, while a negative sign indicates an inverse effect for A, AC, A², B² and C².

According to the ANOVA of the model in Table 4, the F values were found to be 35.7, 26.4 and 30.5, and the P values were found to be 0.0005, 0.0011 and 0.0008 for the removal percentages of LEVO, AZI and CEF, respectively, which implies that the model is significant. The model coefficients (R²) were found to be 0.9847, 0.9794 and 0.9821, for the three responses, respectively. This indicates that the quadratic model explained 98.5, 97.9 and 98.2% of the total variances of the three responses. The R² and adjusted R² values were close to unity, which display the good agreement between experimental and predicted values. The values of the predicted R² were found to be in reasonable agreement with the values of the adjusted R², which supports the high significance of the model. High adequate precision values (>4) indicate that the model has a high signal-to-noise ratio as shown in Table 4.

The P and F values were calculated for each independent variable and its interaction. The calculated F value of the square term of pH had the largest effect on the removal percentages of the three drugs, followed by the effect of time. However, the dose of nanoparticles showed the minimum effect for the removal of the three drugs. Lack of fit was found to be insignificant for the three responses. The ANOVA of the quadratic model is shown in Supplementary Tables S4, S5 and S6.

The simultaneous effect of pH and dose of nanoparticles at constant time is shown in Fig. 4a. The pH of the reaction solution governs the surface charge on the photocatalyst. The surface of the prepared photocatalyst with pH values higher than 6.8 will be saturated by negative charges and at lower pH has a positive charge caused by H⁺ protons. Photodegradation efficiency was analysed at different pH values 5, 7 and 9 under visible light irradiation. LEVO and CEF contain a variety of functional groups with strongly acidic (carboxylic/sulfonic) or basic (amine/aminothiazole) properties, so they will be present in three different ionic forms, mainly depending on the pH. At acidic pH 5, they are positively charged, while AZI will exist as zwitterion in the pH range of 3.5−7.7, which explains the minimum interaction between the drug molecules and particles. The percentage removal of the three drugs LEVO, AZI and CEF increases until it reaches a maximum point at pH 7, where the molecules of the three drugs are uncharged. On this basis, it is preferable to keep the pH neutral to avoid repulsions that may cause lesser adsorption to the catalyst surface, which diminishes photodegradation efficiency (Shandilya et al. 2018). By increasing the pH of the solution to 9, AZI, CEF and LEVO molecules acquire a negative charge, which causes repulsion and a decrease in interaction with particles and thus their removal; another factor that can play a major role in low antibiotic degradation is the decomposition of hydroxyl radicals in alkaline media (Fauzi et al. 2018).

Fig. 4.

3D response surface contour plots showing the simultaneous effects on the removal % of LEVO, AZI and CEF. (a) Effects of pH and the dose of nanoparticles; (b) response surface contour plots showing the effects of pH and time; (c) response surface contour plots showing the effects of time and dose of nanoparticles.

On the other hand, the removal of the three drugs is maximised at the middle level of the dose of nanoparticles (0.6 g L^–1), whereas it slightly decreases at lower and higher doses as shown in Fig. 4a. Additionally, the mutual effect of pH and time at a constant dose of nanoparticles is shown in Fig. 4b, demonstrating that the removal of the three drugs increases with time at pH 7 as mentioned above. The removal % reaches its maximum at 60 min, but no further significant increase is observed against time. Fig. 4c shows the effect of time and dose of nanoparticles at constant pH which shows that the removal of the three drugs is greatly affected by increasing time (from 60 to 90 min) with a lower effect of nanoparticle dose.

Upon applying the optimum conditions, the removal percentage of the studied drugs was maximised as it reached ~97% in only 60 min, so no further modification such as element doping was required to accelerate photocatalytic performance. In addition, the higher surface area of the synthesised g-C₃N₄ provides more sites for interaction with the studied drugs (Zhu et al. 2015).

The validation of the RSM model was performed by measuring triplicate experiments in selected optimisation conditions as shown in Supplementary Fig. S6, where the optimised parameters obtained from the statistical analysis showing the highest desirability index (0.886) are listed in Table 5. By comparing actual percentage removal versus predicted percentage removal in the laboratory distilled water spiked with 500 ng mL^–1 of each of LEVO, AZI and CEF, an error of 0.18, 1.2 and 0.46% was observed for the removal of each drug, respectively. Thus, the second-order polynomial regression equation was confirmed as an estimate of the accurate percentage of the removal from water samples.

The optimised conditions were applied to clean real water and compared to the results of the laboratory distilled water samples (both spiked with equal amounts of the three cited pharmaceuticals, 500 ng mL^–1), and it was observed that the efficiency of photodegradation in real samples showed a slight decrease than that in distilled water, due to the matrix effect of environmental parameters such as the presence of pollutants other than our studied drug or microorganisms that might affect their removal, as shown in Fig. 5. The photodegradation of AZI showed the highest matrix effect of 6.20%, while LEVO and CEF showed similar matrix effects of 4.11 and 4.01%, respectively.

Fig. 5.

The calculated average photodegradation % of LEVO, AZI and CEF in spiked distilled water, spiked clean real water and contaminated real wastewater using g-C₃N₄.

Due to the high efficiency of the process by g-C₃N₄ (>90%) and the low matrix effect for all drugs, the proposed photodegradation procedure was applied to real water samples containing the three cited drugs as shown in Fig. 5. The results of the analysis after photodegradation with g-C₃N₄ using the adopted chromatographic method are shown in Table 3, where the obtained results were found to be similar to the results of the clean real water samples.

Complex green analytical procedure index (ComplexGAPI) was chosen for the assessment as it provides an evaluation of the entire methodology in addition to the synthesis of the proposed g-C₃N₄ particles. ComplexGAPI is considered a complementary tool to the previously developed GAPI metric (Płotka-Wasylka 2018; Płotka-Wasylka and Wojnowski 2021; Saleh et al. 2023). One hexagonal field was added to the original GAPI graph which evaluates several components of the pre-analysis processes and its conditions, the number of rules correlated to green economy, health and safety hazards of chemicals and instrumentation used in the synthesis process, purity of the end products, and calculation of its E factor. The original five pictograms of the GAPI tool are subdivided into smaller sub-pictograms to assess 15 aspects related to the analytical method including the type of method, the sample (collection – preservation – transport – storage – preparation), reagents, instrumentation and waste treatment (El-Naem and Saleh 2021; Kayali et al. 2023; Mannaa et al. 2023).

The ComplexGAPI showed a green profile as it scored 2 red sub-hexagons only related to the high temperature applied during synthesis of g-C₃N₄ which consumes a high amount energy. Only two red sub-pictograms were scored due to the high energy output of the analytical method (LC-MS/MS) and the non-treated waste of acetonitrile and formic acid with an NFPA score 2 or 3, produced from the method. Applying extraction using Oasis Prime HLB cartridges scored a yellow pictogram due to lower solvent consumption (10–100 mL) and miniaturised procedure. Moreover, the E factor was less than 0.1 because the analysis was conducted without any prior preparation, confirming the recommended method’s overall environmental friendliness as shown in Fig. 6a. The scores of the pictograms and hexagon are shown in Supplementary Table S7.

Fig. 6.

(a) ComplexGAPI metrics (greenness tool) and (b) RGB 12 algorithms (whiteness tool) for the proposed method.

Analytical procedures are considered sustainable when they consider not just their environmental impact but also their economic cost, validity and efficiency. An effective tool for assessing the sustainability of analytical techniques is the RGB 12 algorithm (Morgan et al. 2023). According to White Analytical Chemistry (WAC), RGB 12 includes three principles (red, green and blue) of assessment. Each principle was subdivided into four parameters, for example red is subdivided into R1, R2, R3 and R4, where a score from 0–100 is assigned to each parameter, with 0 being the worst, and 100 means that the method is adapted to this parameter. The red principle (R1–R4), green principle (G1–G4) and blue principle (B1–B4) relate to analytical performance, environmental impact and economic criteria, respectively. The general accordance of the method with all three principles is expressed by an overall quantitative principle known as ‘whiteness’ (Nowak et al. 2021; Elbalkiny et al. 2022; Marzouk et al. 2023).

The proposed method showed an excellent balance between the three aspects of WAC, scoring an average white score of 80.1 as shown in Fig. 6b. It is worth mentioning that, through the performed WAC evaluation, all colours and principles were equally important, to maintain the idea of sustainability. This reflects the method’s high analytical efficiency, green profile and effective cost.