All the required materials, such as chloroform (≥99.0%), acetone (≥99.8%), carbon disulfide (≥99.90%), hydrochloric acid (≥37%), ethanol (≥99.9%), sodium hydroxide (≥97%), acetonitrile (≥99.9%), and methanol (≥99.9%) were purchased from Merck Company and sulfamethoxazole (≥98.0%), from Sigma-Aldrich. To adjust pH, solutions like HCl (0.1 M) and NaOH (0.1 M) were used. Also, the standard SMX solution with concentration of 100 mg L^-1 was used to prepare working solutions for experiments. The chemical structure of SMX (Molecular weight: 253.3 g mol⁻¹ and molecular formula: C₁₀H₁₁N₃O₃S) is shown in Figure 1. A UV-Vis spectrophotometer (Shimadzu UV-1900, Japan) was used to measure the concentration of SMX, and at each stage, spectroscopy was performed in the wavelength range of 200–400 nm. An ultrasonic device (LK-D31-1, China) was used for ultrasound-assisted extraction. Scanning electron microscope (SEM; KYKY-EM3200, China) and X-ray diffraction (XRD; Philips X’Pert Pro MPD, Netherlands) evaluated the adsorbent structure. The experiments were carried out at the Islamic University, Iraq.

ZnFe₂O₄ adsorbent were synthesized by the hydrothermal method.

For this purpose, 1,000 mL of solution was prepared by dissolving Zn(NO₃)₂.6H₂O and FeSO₄.7H₂O with a molar ratio of 1/2 (Zn/Fe) in double distilled water. The pH of the solution was adjusted to 9 by HNO₃ (0.1 N) and NaOH (0.1 N) solutions. The synthesis was continued at 80°C for 40 min under continuous air purification (flow rate = 4 L min^-1). The prepared adsorbent was separated using a magnet and washed with double distilled water until the pH of the solution reached 7. Then the ZnFe₂O₄ adsorbent was dried in an oven at 60°C for 24 h and kept for further investigation. The formation equation of ZnFe₂O₄ nanoparticles can be expressed as Eq. (1). Zn 2 + + 2 Fe 2 + + 6 OH − + 1 / 2 O 2   →   ZnFe 2 O 4 + 3 H 2 O (1)

Initially, 10 mL of SMX solution with a concentration of 250 μg L^-1 was transferred to a conical bottom test tube, and its pH was adjusted to 5. Then, 185 µL of chloroform and 535 µL of acetonitrile were injected into the aqueous solution using a glass syringe. At this stage, the test tube was placed in an ultrasonic bath for 7.5 min. The mixture was centrifuged at 5,500 rpm for 5 min. After centrifuging, solvent droplets containing the analyte were deposited in the tube. The precipitated phase was isolated entirely and in order to analyze the SMX value, it was transferred to a spectrometer, and measurements were made at 260 nm. The extraction recovery (ER) and the enrichment factor (EF) were calculated by Eqs 2, 3, respectively. E R = C s e d ∗ V s e d C 0 ∗ V a q   × 100 % = E F   ×   V s e d V a q   × 100 % (2) E F = C s e d C 0 (3)

where C ₀ is the concentration of SMX in the aqueous solution. C _sed is the concentration of SMX in the organic solution. V _aq is the volume of the aqueous solution. V _sed is the volume of the organic solution (Cruz-Vera et al., 2009; Nemati et al., 2023). A schematic of the UA-DLLME method is shown in Figure 2.

In order to preconcentration and extraction SMX from aqueous solutions, DSPME in optimal conditions was used. For this reason, 10 mL of SMX (concentration 250 μg L^-1) was prepared at pH of 5. Then, 0.024 g of ZnFe₂O₄ nanoparticles were added to it. The dispersion was performed through an ultrasonic bath for 7.5 min and centrifuged at 5,500 rpm for 5 min. Then, the adsorbent was separated from the liquid phase using a magnet and washed with 235 µL of methanol. Finally, the solvent was transferred to the cell spectrometer, and SMX was determined by UV-Vis spectrophotometry. A schematic of the UA-DSPME method is shown in Figure 3.

The RSM is a method based on statistical and mathematical techniques that can be used to examine the relationships between variables and responses and analyze interaction effects (Xu et al., 2021). In addition, the RSM provides a mathematical model for the researcher to study the effects of independent variables (Tao et al., 2022). The central composite design (CCD) framework is used to design of matrix. This approach optimizes the number of experiments as well as provides a platform to statistically validate the range of independent process variables. This statistical design also provides a data-driven model based on the analysis of variance (ANOVA) method. The identified model can be a linear or quadratic or cubic regression model, which will be determined based on the interaction between the variables and their statistical significance. In many studies performed by RSM, the quadratic regression model was the most significant multivariate model (Shojaei et al., 2021b; Boublia et al., 2023). The multivariate model, which involves linear and non-linear terms, is given in Eq. (4). Y = β 0 + ∑ i k β i x i + ∑ i k β i i x i 2 + ∑ i < j k ∑ j k β i j x i x j + e (4)

This equation includes linear sentences (x _i ), interactions (x _i x _j ), and quadratic variables (x _i ² ). Y is the extraction recovery and e is the random error. The coefficients β ₀ , β _i , β _ij , and β _ii also show the constant term, the linear, quadratic, and interactive coefficients, respectively (Pourabadeh et al., 2020).

CCD is a typical RSM experimental design framework. Each factor in CCD is examined at five levels (−α, −1, 0, +1, +α) and includes axial points, factorial points, and center points. Factorial points are named high and low levels and are marked with +1 codes for high and −1 for low levels. Axial points are at a distance from the center. Also, there is more than one central point for estimating experimental error and determining data reproducibility. This study evaluated the effect of parameters to investigate the increase in SMX extraction and process optimization at five levels and six replications at the central points of the design. Design-expert software v10 (trail) was used to design experiments and data processing. Table 1 shows the independent variables and their values.

SMX drug was measured to evaluate the UA-DLLME and UA-DSPME methods in studying real-time samples with different matrices of tap water, wastewater of hospital, and urine. For this reason, certain amounts of drugs to the samples were added, and the extraction was performed according to the proposed methods in optimal conditions. The results were reported for three replications. In order to evaluate urine samples from five healthy men (25–30 years old) with consent, urine samples were collected without medication and stored at −10°C. Urine samples were diluted to reduce the effect of matrix (5 mL of double distilled water was added to 5 mL of urine). Further, several compounds at lower pH may become insoluble for extraction from hospital wastewater and urine samples. As a result, it is necessary to separate the sediments from the solution using a centrifuge before adding the drug. For this reason, the hospital wastewater and urine samples were centrifuged (3,500 rpm) for 10 min.

SEM and X-ray diffraction were applied to evaluate the structure and size of synthesized ZnFe₂O₄ nanoparticles. An SEM image (see Figure 4) was taken to study the surface structure and size of ZnFe₂O₄ nanoparticles. As shown in Figure 4A, ZnFe₂O₄ nanoparticles are seen as granular and separate structures, and their size is less than 100 nm. The peak position and relative intensity observed in the XRD pattern of ZnFe₂O₄ nanoparticles are shown in Figure 4B. The XRD pattern of the ZnFe₂O₄ adsorbent corresponded to the standard card (JCPDS 22–1,012) related to the ferrite. The plates 220, 311, 400, 422, 511, and 440 in the diffraction pattern show the reason for the formation of spinel cubic phase with the space group Fd-3m (Navgare et al., 2020). In this study, the absence of additional plates indicated the formation of a pure phase of the zinc ferrite nanostructure. The crystal sizes of the ZnFe₂O₄ nanoparticles were calculated using Debye–Scherrer equation [Eq. (5)]. D = K λ β Cos θ (5)

In this equation, θ (degree), D (nm), λ (nm), K, and β (radians) are the Bragg’s angle, the crystal size, the X-ray wavelength, the shape factor, and full width at half maximum (FWHM) respectively (Egbosiuba et al., 2021). The crystal size of the nanostructures was determined by the Debye–Scherrer equation and found to be 15 nm. The surface area is an important parameter in determining the morphology of the absorbent. The higher the surface area of the ZnFe₂O₄ adsorbent, the more active surfaces are available for the adsorption of pollutants on the adsorbent. Therefore, BET analysis of ZnFe₂O₄ nanoparticles was performed. The total pore volume, the surface area, and the mean pore diameter was 0.16 cm³ g⁻¹, 80.06 m² g⁻¹, and 9.7 nm, respectively.

Relevant tests were performed according to the points defined in the RSM scheme. The results of SMX extraction efficiency tests are presented in Table 2. In the next step, the response surface methodology analyzed the data from different tests. Also, the regression coefficients were estimated, and the ANOVA Table was obtained for each of the answers shown in Table 3.

ANOVA is a technique used to assess the significance of differences between groups or treatments in a study. In the context of the provided information, ANOVA can be applied to evaluate the relationship between variables and a response (Promoppatum and Yao, 2020). The coefficient of determination (R ²) expresses the proportion of total changes in responses that can be determined by the variables. For the UA-DLLME method, R ² was calculated to be 0.9977, indicating that 99.77% of the changes in the response can be determined by the variables. In the UA-DSPME method, the R ² value was even higher at 0.9998, suggesting that the variables explain 99.98% of the response variation. The adjusted coefficient of determination (R²-adj) considers the number of parameters in the design. The R²-adj for the UA-DLLME method was reported as 0.9956, which considers the complexity of the model. The R²-adj for the UA-DSPME method was 0.9997, reflecting its high explanatory power. To evaluate the model fit, the lack-of-fit test was conducted. The lack-of-fit test examines whether the model successfully represents the data at points outside the regression model’s domain. It is used to detect any shortcomings in the model’s fit. The article mentions the lack-of-fit test as a means of identifying areas where the model fails to capture the data accurately (Almeida et al., 2017; Bhowmik et al., 2018). Regarding precision, adequacy precision values were provided for each method. The UA-DLLME method achieved an Adeq-precision of 66.66, while the UA-DSPME method had a value of 318.75. Higher Adeq-precision values indicate better reliability of the regression models. Therefore, the results showed that the model significantly predicts the extraction under different variable conditions. The quadratic model was performed on the test data, and Eqs 6, 7 were obtained as encoded representations for the predicted extraction recovery of UA-DLLME and UA-DSPME methods, respectively. % ER  UA − DLLME = + 92.41 + 2.02 * A + 8.55 * B − 9.48 * C + 2.36 * D − 0.27 * AB − 0.17 * AC + 0.44 * AD + 0.28 * BC − 0.42 * BD − 0.08 * CD − 7.68 * A 2 − 4.22 * B 2 − 3.85 * C 2 − 4.14 * D 2 (6) % ER  UA − DSPME = + 71.15 + 10.41 * A + 9.60 * B − 9.17 * C + 7.63 * D − 0.37 * AB − 0.17 * AC + 0.17 * AD − 0.14 * BC − 0.20 * BD + 0.09 * CD − 2.29 * A 2 − 7.15 * B 2 − 3.77 * C 2 − 2.29 * D 2 (7)

The experimental data versus predicted data plots and the normal plot of residuals demonstrate good agreement and normal distribution of the data (Figure 5). To assess the normal distribution of residuals, Figures 5A, B have been utilized. These plots show how well the predicted values align with the actual experimental data, indicating the model’s accuracy. The normal probability graphs in Figures 5C, D are used to check the normality of the data. The closeness of the data to the straight line in these graphs showed a normal distribution for errors with zero mean and constant variance.

In the UA-DLLME process, the extraction solvent was selected from solvents that were insoluble in water and had a higher density than water. Also, they could extract the desired compound. In this regard, carbon tetrachloride (CCl₄), carbon disulfide (CS₂), and chloroform (CHCl₃) were tested as extraction solvents. Based on the results, a clear cloud solution was not obtained using CS₂ and CCl₄ as extraction solvent. It indicates that these solvents cannot disperse appropriately in the aqueous phase and may not have a good extraction ability. However, with CHCl₃, high extraction efficiencies were obtained for SMX. Therefore, CHCl₃ was selected as the optimal extraction solvent in later stages.

For the sample preparation process in the UA-DLLME method, it should be noted that the type of disperser solvent is important for the preconcentration of the material. The basis for selecting the disperser solvent is its solubility in the organic phase and the aqueous phase. For this purpose, acetonitrile (ACN), methanol (ME), acetone (AC), and ethanol (ET) were selected as disperser solvents because of their chemical and physical properties. According to the results, the highest recovery rate was achieved while using acetonitrile as a disperser solvent. So, acetonitrile was selected as disperser solvent for further experiment analysis.

The eluent solvent plays an important role in the UA-DSPME method because the efficiency is affected by the proper dispersion of the ZnFe₂O₄ adsorbent in the solution. Various solvents such as ACN, ME, ET, and AC were tested. The results (Figure 6) show that methanol as an eluent increases extraction and adsorption efficiency. So, methanol was selected as the eluent in the UA-DSPME method.

In order to accelerate the process in this study, centrifugation is used. Also, centrifugation is essential as the precipitating force of the extracting solvent containing the analyte. Hence, the effect of centrifuge speed in the range of 1,000–4000 rpm was investigated. According to Figure 7, the extraction efficiency of SMX increases with increasing centrifuge speed and remains constant at 3,500 rpm for UA-DLLME and UA-DSPME. The low extraction efficiency at low velocities occurs because the phase separation in the UA-DLLME method and the adsorbent settling in the UA-DSPME method are not entirely done. Therefore, 3,500 rpm was chosen as the optimal centrifuge speed to complete the separation of phases.

The point of zero charges (pH_pzc) of the ZnFe₂O₄ adsorbent was determined by charging 9 Erlenmeyer flasks (100 mL) with 50 mL of 0.01 M NaCl solution. In this process, their initial pH was adjusted to 2–10 using 0.1 M NaOH and 0.1 M HCl solutions. Then, 0.1 g of adsorbent was weighed and added to each Erlenmeyer. The mixture was stirred on a shaker at 100 rpm at 25°C for 24 h. Afterward, the pH_f of the solutions was determined with a pH meter. The difference between the pH_f and the pH_i (∆pH = pH_f-pH_i) was plotted versus the pH_i. The point where the graph intersected the X-axis was reported as the pH_pzc. This point shows where the sum of the negative surface charge balances the sum of the positive surface charge. This value for the synthetic adsorbent was about 6.1 (Figure 8). Therefore, at pH < 6.1 and pH > 6.1, the ZnFe₂O₄ adsorbent surface has a positive charge and a negative charge, respectively.

The reusability of ZnFe₂O₄ nanoparticles is an important aspect of operational and environmental objectives. The ability to reuse and regenerate the adsorbent is critical for economic feasibility. The reversible adsorption process allows for adsorbent regeneration. To study the reusability, the adsorbent was placed in 5 mL of methanol for 5 min after each extraction. After the desired time, the mixture was centrifuged (3,500 rpm) for 5 min. After that, the ZnFe₂O₄ nanoparticles was separated from the solution using a magnet and washed with double distilled water. The ZnFe₂O₄ nanoparticles was dried in an oven (90°C) for 1 h. The reusability experiments of ZnFe₂O₄ nanoparticles in this study showed that the adsorbent can be effectively used several times for the extraction of SMX (Figure 9).

Many cations and anions with different concentrations in natural water samples can affect the extraction of the pollutants and cause errors (negative or positive) in their measurement. Hence, it is highly significant to study the effect of various ions and their interference. In this research, SMX is extracted from aqueous samples containing 250 μg L^-1 of SMX in the presence of various anions and cations, and determined the degree of interference of these ions. It should be noted that an interfering ion refers to an ion that causes a ±5% change in the absorption signal of the analyte. The results of interference effect are presented in Table 4. The results showed that most investigated anions and cations did not significantly interfere with SMX extraction and measurement.

Based on the above results, the optimal conditions for extracting of SMX from environmental water and biological samples using Design-expert software v10 were predicted (Table 5). The maximum extraction recovery under the optimal conditions (0.024 g of adsorbent, 7.5 min of ultrasonication time, 235 µL of eluent volume, pH = 5, 535 µL of disperser solvent volume, and 185 µL of extraction solvent volume) was estimated. The extraction recovery of SMX was determined as 94.11% for the UA-DLLME method and 93.63% for the UA-DSPME method. The RSD (N = 3) for the UA-DLLME and UA-DSPME methods were obtained as 1.6% and 2.2%, respectively, confirming the model’s high accuracy in predicting the results.

Parameters such as preconcentration factor (PF), percent relative standard deviation (RSD%), enrichment factor (EF), limit of detection (LOD), and linear range (LR) were calculated to demonstrate the validity of the methods (Table 6). Under optimal conditions, the LR for the UA-DLLME and UA-DSPME methods were determined as 10–800 μg L^-1 and 20–1,200 μg L^-1, respectively. The LOD for the UA-DLLME and UA-DSPME methods were obtained as 3 μg L^-1 and 6 μg L^-1, respectively, indicating the high sensitivity of the measurement methods. The repeatability was also assessed by calculating the RSD (N = 5), resulting in 1.9% and 2.3% for the UA-DLLME and UA-DSPME methods, respectively. EF represents the ratio of analyte concentration in the organic solution to its initial concentration in the aqueous solution. EF values of 108 and 95 were obtained for the UA-DLLME and UA-DSPME methods, respectively. The PF, indicating the ratio of the analyte concentration in the analyzed solution to its concentration in the initial solution, was 54.05 for UA-DLLME and 42.55 for UA-DSPME. The LOD obtained when UA-DLLME was applied was lower than that obtained when UA-DSPME was used. The UA-DLLME has advantages, such as high enrichment factor, high preconcentration factor, and low LOD and low RSD values. The UA-DSPME and UA-DLLME methods represent inexpensive alternatives to conventional SPE and LLE extraction methods.

To separate, preconcentration, and measure SMX from various samples, including deionized water, wastewater of hospital, tap water, and urine, experiments were conducted for UA-DLLME and UA-DSPME methods. The preconcentration and extraction of SMX from real samples were done by adding 250 μg L^-1 of SMX to the samples in three repeated measurements according to the proposed methods (Section 2.3 and Section 2.4) in optimal conditions, and the amount of recovery SMX was determined. Table 7 presents the results of UA-DLLME and UA-DSPME methods for SMX extraction and measurement in real samples. The results of Table 7 showed that these methods were suitable for use in real samples.

Each process parameter (keeping other parameters constant) has a different effect on the overall extraction of SMX. However, the interaction effect of the two parameters has a significant effect on the extraction. Therefore, surface response plots were drawn to study the contour trends and effects of different parameters on the overall extraction efficiency (Figure 10). The parameters are plotted on the x and y-axes in these three-dimensional diagrams, and the response is plotted on the z-axis.

Figure 10A shows the effect of two parameters of extraction solvent volume and disperser solvent volume in the UA-DLLME method on SMX extraction. The volume of the disperser solvent is of high importance in the UA-DLLME method. Also, different volumes of acetonitrile (100–700 μL) were tested to determine the optimal disperser solvent volume. Initially, with increasing disperser solvent volume, the recycling efficiency also increased, but at volumes above 535 μL, the recycling efficiency decreased. The results showed that the highest recovery was observed in 535 μL of acetonitrile.

Different amounts of chloroform (50–250 μL) were investigated to determine the effect of extraction solvent volume on the SMX extraction. According to Figure 10A, the SMX extraction was maximum in the chloroform volume of 185 μL, and then the SMX extraction decreased with the increase of the chloroform volume. The reason is that the concentration factor drops since the volume of the precipitated phase increases. So, 185 μL of chloroform was chosen as the optimal volume in the UA-DLLME method.

Figure 10B presents the effect of disperser solvent volume and ultrasonication on the SMX extraction process in the UA-DLLME method. In the effect of ultrasonic duration (2–10 min), the highest amount of recovery was obtained when the solution of the cloudy was placed in an ultrasonic bath for 7.5 min, and as time increased, the amount of recovery decreased. Only quantities of analyte dissolved during extraction can be analyzed and quantified in the UA-DLLME sample preparation process. Therefore, the reason for the reduction in recycling, if the extraction time is increased, is that some of the analytes dissolved in the extraction solvent can be separated from it and dissolved in the disperser solvent.

Figure 10C illustrates the changes in SMX extraction due to changes in the adsorbent amount and eluent volume in the UA-DSPME method. Different volumes of methanol (100–300 μL) were investigated to determine the effect of eluent volume on SMX extraction in the UA-DSPME method. According to Figure 10C, the recycling efficiency increased when the eluent volume rose to 235 μL. However, at volumes above 235 μL, the recycling efficiency decreased. Therefore, 235 μL of methanol was chosen to obtain high efficiency in analyte extraction.

Figure 10 depicts the effect of pH and eluent volume on the SMX extraction. The pH changes the ions or molecules, which affects the extraction efficiency. For this reason, the amount of SMX extraction in the pH range of 3–11 was investigated. Based on the results, the maximum extraction was achieved at pH 5, and with increasing pH, the recycling efficiency decreased. The adsorbent surface is negatively charged because hydroxide ions in the solution at alkaline pH values go up, and SMX is expelled from the adsorbent surface with a negative charge. Therefore, it reduces the extraction rate.

The efficiency of UA-DLLME and UA-DSPME methods for SMX preconcentration was compared with other previous methods and the results are summarized in Table 8. As observed in Table 8, the proposed method was comparable or better than many previous methods in terms of various analytical parameters such as LR and LOD. Another important parameter is the extraction time. The extraction time was more than 7.5 min in previous studies on SMX extraction, and it shows the high speed of the methods in the present study. In many of the studied methods in the literature, devices such as HPLC or MS, which are selective analytical instruments with high sensitivity, have been utilized. On the other hand, the high cost of using these devices and the lack of accessibility in developing countries make their widespread adoption unfeasible. Hence, the current approach using spectrophotometry, which offers high sensitivity, repeatability, speed, and simplicity, is deemed suitable for preconcentration of SMX from aqueous and biological samples. Furthermore, the results showed that the use of the CCD matrix leads to a reduction in the number of experiments required. Additionally, this approach can reduce the associated time and costs involved in the investigation.