Facile Fabrication of Novel NiFe₂O₄@Carbon Composites for Enhanced Adsorption of Emergent Antibiotics

Abstract

Water purification is becoming one of the most pertinent environmental issues throughout the world. Among common types of water pollution involving heavy metals, pharmaceutical drugs, textile dyes, personal care products, and other persistent organic pollutants, the pollution of antibiotic drugs is increasingly emerging due to their adverse effects on microorganisms, aquatic animals, and human health. Therefore, the treatment of such contaminants is very necessary to reduce the concentration of antibiotic pollutants to permissible levels prior to discharge. Herein, we report the use of NiFe₂O₄@C composites from a bimetallic-based metal-organic framework Ni-MIL-88B(Fe) for removal of ciprofloxacin (CFX) and tetracycline (TCC). The effect of production temperatures (600–900 °C), solution pH (2–10), NiFe₂O₄@C dose (0.05–0.2 g/L), concentration of antibiotics (10–60 mg/L), and uptake time (0–480 min) was investigated systematically. Response surface methodology and central composite design were applied for quadratic models to discover optimum conditions of antibiotic adsorption. With high coefficients of determination (R² = 0.9640–0.9713), the proposed models were significant statistically. Under proposed optimum conditions, the adsorption capacity for CFX and TCC were found at 256.244, and 105.38 mg/g, respectively. Recyclability study was employed and found that NiFe₂O₄@C-900 could be reused for up to three cycles, offering the potential of this composite as a good adsorbent for removal of emergent antibiotics.

1. Introduction

Global industrialization not only accompanies drastically increasing demands for clean water, but it can also trigger many potential pollutions related to water resources. Therefore, water purification is currently one of the most pertinent environmental issues throughout the world. There are common types of water pollution caused by heavy metals, pharmaceutical drugs, textile dyes, personal care products, or other persistent organic pollutants [1,2,3]. Among them, the pollution of pharmaceutical drugs, including antibiotics, is alarmingly emerging due to their adverse effects on microorganisms, aquatic animals, and human health through bioaccumulation in food chains [4]. Tetracycline (TCC, molecular formula: C₂₂H₂₄N₂O₈, molar weight: 444.435 g/mol) acts as a broad-spectrum antibacterial antibiotic [5]. It has been widely consumed in livestock industries and human therapies mainly due to low-cost production [6]. It was estimated that TCC accounts for one third of the total production, and is ranked second in the consumption of worldwide antibiotics [7]. Meanwhile, ciprofloxacin (CFX, molecular formula: C₁₇H₁₈FN₃O₃, molar weight: 331.346 g/mol) presents as one of the most widely used second-generation quinolones [8]. It was reported that millions of prescriptions were consumed for antibacterial purposes, eye irritations, and bone infections in the US [9]. The occurrence of TCC and CFX pollution can originate from the wastewater discharge of pharmaceutical factories, or residuals of aquafarming activities. In hospital wastewaters, an amount of these pharmaceuticals can exist through untreated urine excretion as a result of partial metabolism in human bodies [10]. They directly enter and accumulate in water sources, enhancing the antibiotics-resistant activities of bacteria, and imposing great threats on human health and safety [11]. More importantly, both TCC and CFX exhibit a difficult degree of biodegradability in aquatic systems [12,13]. Hence, adoption of treatment technologies is wholly necessary to reduce the concentration of TCC and CFX antibiotics to permissible levels prior to discharge.

There are a variety of physical and chemical techniques applied for the treatment of TCC and CFX antibiotics. For example, filtration and infiltration processes using granulation, activated alumina, and activated carbon for water purification have been addressed [14,15]. However, many studies indicated that adsorption acted as a feasible approach [16]. Indeed, the main advantages of adsorption such as simplicity, outstanding recyclability, and high efficiency of removal of these pharmaceuticals and others, were demonstrated [17,18]. Various materials such as metal-organic frameworks [19], ferrites [20], composite carbon nanotubes [21], activated carbons [22], and functionalized zeolites [23] have been studied as promising adsorbents. Among the materials mentioned above, ferrites such as Fe₃O₄, NiFe₂O₄, CoFe₂O₄, and MnFe₂O₄ possess a special magnetic property, which enables easy separation from the solutions. Previous literatures have addressed the use of NiFe₂O₄ ferrites in wastewater remediation [24,25,26]. However, the adsorption efficiency of single ferrites is relatively low, and its recyclability seems unstable. Porous carbons have superior advantages, i.e., their structure often obtains high porosity, and surface area, along with an amount of functional groups, which accelerates the efficiency of antibiotics removal [27,28]. Attachment of magnetic components (e.g., NiFe₂O₄) on porous carbon structure is highly expected to reach out higher results of uptake effectiveness and separation. Thus, fabrication of such magnetic porous carbons has recently attracted great attention [29,30,31]. Several methods such as coprecipitation and impregnation have been widely applied, but many disadvantages (e.g., poor dispersion of ferrite nanoparticles in carbon matrixes or layers due to magnetic agglomeration) were observed [32,33,34]. These problems can be solved by using metal-organic frameworks (MOFs) as self-sacrificial templates through pyrolysis to synthesize magnetic carbons. After the carbonization process, the crystalline structure of MOFs including both metal clusters and ligands is decomposed; leaving a carbon skeleton embedded with metal oxides [35,36]. This strategy was successfully applied to form magnetic composite carbons in our previous studies [37,38,39].

In this study, we converted Ni-MIL-88B(Fe) as a template into novel nickel ferrites-embedded porous carbons (NiFe₂O₄@C). Various temperatures at 600, 700, 800, and 900 °C were investigated. The composite materials were characterized and applied as adsorbents for the adsorption of TCC and CFX antibiotics in waters. The effect of contact time, antibiotics concentration, solution pH, and NiFe₂O₄@C dosage was surveyed. Adsorption kinetics and isotherms for antibiotic uptake on NiFe₂O₄@C were also performed herein. To the best of our knowledge, there was still no report on the use of NiFe₂O₄@C adsorbents derived from Ni-MIL-88B(Fe) for TCC and CFX antibiotics removal.

2. Experimental

2.1. Chemicals

In this study, tetracycline hydrochloride, ciprofloxacin, 1,4–benzene–dicarboxylic acid (H₂BDC), iron(III) nitrate nonahydrate (Fe(NO₃)₃∙9H₂O), nickel(II) chloride hexahydrate (NiCl₂∙6H₂O), N,N–dimethylformamide (DMF), triethylamine, methanol (CH₃OH), and acetonitrile (CH₃CN) were commercially purchased from Sigma–Aldrich. These chemicals were used without any need of purification. Double-distilled water was used to dilute the aqueous solutions.

2.2. Instrumentation

The D8 Advance Bruker powder diffractometer was used to record the X-ray powder diffraction (XRD, Hitachi Inc., Krefeld, Germany) profiles using Cu–Kα beams as excitation sources. The S4800 instrument (JEOL, Tokyo, Japan) was implemented to capture the scanning electron microscope (SEM) images with magnification of 7000× using an accelerating voltage source (15 kV). The N₂ adsorption/desorption isotherm measurements were recorded on the Micromeritics 2020 volumetric adsorption analyzer system (Micromeritics Inc., Norcross, GA, USA). The FT-IR spectra were recorded on the Nicolet 6700 spectrophotometer (Thermo Fischer Scientific Inc., Waltham, MA, USA). The UV–Vis spectrophotometer (Shimadzu, Kyoto, Japan) was used to determine the concentrations of TCC and CFX antibiotics at 273 and 272 nm in wavelength, respectively.

2.3. Synthesis of Ni-MIL-88B(Fe) Material

The Ni-MIL-88B(Fe) material could be fabricated by solvothermal method, referring to a previous study with some modifications of amount of salts and solvents [40]. In detail, the synthesis process is summarized in Figure 1. As a typical procedure, 0.3 g H₂BDC, 0.4309 g Fe(NO₃)₃·9H₂O, and 0.1268 g NiCl₂·6H₂O were each dissolved completely in separate beakers containing DMF (40 mL) under stirring for 15 min. All solutions were then poured into a larger beaker, and CH₃CN (40 mL) was carefully added under stirring for the next 30 min. The homogeneous mixture was divided and transferred into Teflon-lined autoclaves with appropriate volume. All autoclaves were loaded in a heating oven at 100 °C for 15 h. After cooling down the oven, the suspension was extracted, washed with DMF (three times) and methanol (three times) to make sure that all residual metals and impurities were removed. The solid was separated from the solution by centrifugation. Finally, all bold red yellow crystals were dried by vacuum system and stored for usage as template precursor for NiFe₂O₄@C synthesis.

2.4. Synthesis of NiFe₂O₄@C Materials from Ni-MIL-88B(Fe)

As-produced Ni-MIL-88B(Fe) was used as a template to synthesize NiFe₂O₄@C. As a typical procedure, 2.0 g Ni-MIL-88B(Fe) were heated (ramping rate of 3 °C/min) at various temperatures, 600, 700, 800, and 900 °C, from the room temperature for 4 h. To make sure that degasification of oxygen was complete, nitrogen (100 mL/min) was flowed continuously throughout the samples for 1 h before pyrolysis. After pyrolysis, the system temperature was allowed to reduce gradually to room temperature beneath nitrogen flow (100 mL/min). The pyrolysis products were collected and labeled as NiFe₂O₄@C-x, where x denotes the pyrolysis temperature (NiFe₂O₄@C-600, NiFe₂O₄@C-700, NiFe₂O₄@C-800, and NiFe₂O₄@C-900).

2.5. Adsorption Experiments

All the adsorption experiments were carried out under batch mode at room temperature, except for further indication. Each of the adsorbents, including Ni-MIL-88B(Fe) and NiFe₂O₄@C-x (0.05–0.2 g/L), were weighed, and loaded into 250-mL beakers containing TCC or CFX solutions (50 mL) at various concentrations (5–60 mg/L), and pH (2–10). After adsorption for 480 min, the antibiotic solution was sampled, and measured by UV–Vis spectroscopy to determine the concentration. For kinetic experiments, the aliquot (2 mL) was sampled at various interval periods (0, 5, 10, 20, 30, 60, 90, 120, 180, 240, 360, and 480 min). The percentage of removal (H, %) and uptake capacity (q_e, mg/g) were defined by Equations (1) and (2).

$$H\left(\%\right)=\frac{C_0-C_f}{C_0}\times 100$$

(1)

$$q_e=\frac{C_0-C_f}{W}\times V$$

(2)

where, C₀ (mg/L) and C_f (mg/L) denote the initial and finial concentrations, respectively; W (g) and V (L) denote the mass of adsorbent and volume of the solution, respectively.

2.6. Experimental Design for Optimization

After analyzing the most influential effect of a range of factors, three main factors involving antibiotic concentration, sorbent mass, and pH were selected to discover the optimum adsorption of antibiotic ciprofloxacin (CFX) and antibiotic tetracycline (TCC). The response surface methodology was applied for statistical analysis, and proposed solutions to reach the highest adsorption capacity. In detail, based on the Box–Behnken design,

Table 1. Box–Behnken designs for main factors with value range.

No	Independent Factor	Unit	Code	Value Range
				TCC			CFX
				–1	0	+1	–1	0	+1
1	Antibiotic concentration	mg/L	x1	10	20	30	20	30	40
2	Adsorbent dosage	g/L	x2	0.05	0.1	0.15	0.05	0.1	0.15
3	pH	–	x3	2	3	4	3	4	5

summaries the list of factors and value ranges for adsorption models of TCC and CFX on adsorbent.

3. Results and Discussion

3.1. Characterization

3.1.1. XRD Analysis

Figure 2 presents the results of crystalline structural analysis of Ni-MIL-88B(Fe), NiFe₂O₄@C-600, NiFe₂O₄@C-700, NiFe₂O₄@C-800, and NiFe₂O₄@C-900 materials through X-ray diffraction patterns. For Ni-MIL-88B(Fe), the main peaks appear at positions 10°, 16.8°, 17.7°, 18.7°, 20.1°, and 22°, which corresponded to that of past literature [41]. This material has a high crystallinity, indicating that it has been successfully synthesized by solvothermal method. Meanwhile, the crystal profile of pyrolysis materials as NiFe₂O₄@C-x (x = 600, 700, 800, 900) were completely different from that of the precursor, offering the event that Ni-MIL-88B(Fe) was decomposed under the effect of thermal energy. However, many typical peaks at 18.7° (111), 30.4° (220), 35.8° (311), 37.4° (222), 44.9° (400), 51.6° (311), 57.4° (511), and 63.0° (440) supported the evidence of single-phase cubic spinel NiFe₂O₄ in composite structure of NiFe₂O₄@C-x. This finding fitted very well with the XRD standard profile of pure NiFe₂O₄ (JCPDS PDF No. 10-0325), as well as many reported documents [42,43]. Moreover, graphitic carbon was found due to the presence of a broad peak between 20 and 30°. This crystallite phase may be formed by the destruction of ligand structure containing aromatic rings, converting them into carbon layers or matrix. A previous study reported the same phenomenon caused by carbonizing metal-organic framework under nitrogen to break down the coordination bonds, and form ionic bonds as of NiFe₂O₄ [44]. The existence of the magnetic component aids the separation of the sorbent materials while the porous carbon component may enhance the adsorption of antibiotic drugs.

3.1.2. FT–IR Analysis

The FT-IR spectra presented in Figure 3 show the moderate difference between original Ni-MIL-88B(Fe) and NiFe₂O₄@C-x materials. Overall, there are several peaks that still remained, but others are absent from the original structure. In detail, a strong absorption band at the wavenumber 3500–3750 cm⁻¹ was typical for the –OH functional groups in all samples [45]. Another absorption band at 1600–1631 cm⁻¹ was maintained, and characteristic of the vibration of C=O bonds, regardless of a minor shift of wavenumber from 1600 in Ni-MIL-88B(Fe) to 1631 cm⁻¹ in NiFe₂O₄@C-x [46]. The strong emerging absorption band at 1392 cm⁻¹ characterized the symmetrical and asymmetrical oscillations of carboxylate groups (–COO) of ligand on Ni-MIL-88B(Fe), but it was not observed on NiFe₂O₄@C-x series [47]. This finding provided the information of breaking out the carboxylate ligands under pyrolysis. Clear evidence of the presence of C–H at 2926, 2864, and 750 cm⁻¹ was presented in NiFe₂O₄@C-x materials [48]. These chemical bonds were typical on graphical carbon with high hydrophobicity. The peaks shifting from 558 to 568 cm⁻¹ were observed since the temperature of pyrolysis increased and it was attributable to the existence of Fe–O [49,50]. This proof consolidated the successful conversion of coordination bonds of Ni-MIL-88B(Fe) into ionic bonds of NiFe₂O₄@C-x. It is possible that carbon acts a strong reductant to transform Fe(III) into Fe(II) species. The previous study explained the same event occurred on MIL-53(Fe) or Fe(BDC) (BDC = 1,4-benzene-dicarboxylic acid) material [51].

3.1.3. Morphological Analysis

Morphological characteristics of original Ni-MIL-88B(Fe) and NiFe₂O₄@C-x materials by scanning electron microscopy (SEM) analysis. Figure 4A shows that Ni-MIL-88B(Fe) particles have a relatively uniform size distribution with mainly hexagonal shape. This observation was well consistent with that of a previous study [40]. When carbonizing Ni-MIL-88B(Fe) material at different temperatures, the morphology of the pyrolysis materials changes markedly as shown in Figure 4B–E. The results indicate that the morphology of NiFe₂O₄@C-x became relatively amorphous, and their surface has more defects at higher calcination temperature. This finding was explained due to the effect of thermal energy, causing the destruction of hexagonal structure, and strong rearrangement of metal atoms under magnetic field.

3.1.4. Nitrogen Adsorption/Desorption Isotherm Analysis

Ni-MIL-88B(Fe) and NiFe₂O₄@C-900 were selected as representatives to characterize the nitrogen adsorption and desorption isotherms as shown in Figure 5. The results depict that the isotherms resemble between mixed Type II and Type IV curves with the presence of hysteresis, indicating that the structures of Ni-MIL-88B(Fe) and NiFe₂O₄@C-900 obtained mesoporous and microporous properties. In addition, BET surface area and pore volume of NiFe₂O₄@C-900 were measured, at 68.3 m²/g, and 0.1541 cm³/g, respectively, which were considerably lower than those of original Ni-MIL-88B(Fe), at 268 m²/g, and 0.14 cm³/g, respectively. This finding reconfirmed that the pyrolysis process causes the collapse of crystalline structure, and agglomeration of magnetic NiFe₂O₄ nanoparticles during pyrolysis at high temperature (900 °C). However, NiFe₂O₄@C-900 with a sufficiently large surface area and diverse surface groups can attain good adsorption efficiency results.

3.2. Adsorption Study

3.2.1. Effect of Pyrolysis Temperature

Calcination condition strongly affects the structure and adsorption properties of materials. Thus, the composite materials produced at various temperature were compared in adsorption capacity to CFX and TCC. Figure 6 shows that NiFe₂O₄@C-900 achieved the highest adsorption capacity in both CFX and TCC antibiotics, 80.0 mg/g and 65.5 mg/g, respectively. This can be explained as follows. At higher pyrolysis temperatures, the structures of NiFe₂O₄@C-x are more defective, higher degree of graphitic carbons, and larger number of functional groups. Therefore, the adsorption capacity of NiFe₂O₄@C-x was higher at the higher produced temperatures.

Herein, Ni-MIL-88B(Fe) attained a relatively good adsorption capacity to TCC antibiotic (63.7 mg/g), which was very close to that of NiFe₂O₄@C-900. However, this material exhibited some disadvantages such as poor stability in aqueous solvent (particularly at low pH), difficult recovery after treatment, and low recyclability. As a result, Ni-MIL-88B(Fe) was not selected to perform further investigations.

3.2.2. Effect of Solution pH

Surface charge of magnetic composites and ionization of antibiotic molecules can be controlled by pH value. Thus, the influence of solution pH 2–10 on CFX and TCC adsorption capacity was investigated in this work. Figure 7 shows the highest adsorption capacity of CFX and TCC on NiFe₂O₄@C-x series at pH 4 and pH 3, respectively, except for the case of CFX adsorption by NiFe₂O₄@C-600, where optimum pH was found at 8.0. NiFe₂O₄@C-900 exhibited the highest CFX and TCC adsorption capacities, at 100.0 mg/g and 118.7 mg/g, respectively. These optimum pH values found herein were consistent with some previous studies [12,52]. Therefore, the CFX solution at pH 4, and TCC solution at pH 3 were selected to investigate for the effect of adsorbent dose.

3.2.3. Effect of Composite Dose

Adsorbent dosage is optimized to reach lower amounts of used materials, but higher level of antibiotic treatment. In this study, composite dosage was set between 0.05 and 0.2 g/L. The adsorption capacity and removal efficiency were measured to totally assess the effect of adsorbent dosage. Figure 8 indicates that composite dose has strong effects on adsorption of CFX and TCC. Overall, the dose increased from 0.05 to 0.2 g/L, the removal efficiency would gradually increase, but uptake capacity would gradually decrease in most cases. These findings are because higher dose of materials provided more adsorptive sites to improve the percentage of antibiotic removal. However, the use of higher amount of composite could lead to a decrease of uptake capacity based on Equation (2). The same phenomenon of adsorption of CFX and TCC on various composites was found in the previous studies [53,54].

In particular, the adsorbents produced at higher temperatures gave higher adsorption of both CFX and TCC. Indeed, take dose of 0.1 g/L as an example, the percentages of CFX and TCC removal by NiFe₂O₄@C-600 obtained at only 32.3% and 19%, respectively, but those by NiFe₂O₄@C-900 reached the highest values, at 72.7% and 68.4%, respectively. At the same trend, uptake capacity of CFX and TCC by NiFe₂O₄@C-600 obtained at only 80.7 mg/g and 42.7 mg/g, respectively, but those by NiFe₂O₄@C-900 reached the highest values, at 181.7 mg/g and 154 mg/g, respectively. This was due to pyrolysis conditions’ influence on pore structure and surface area of composites. Finally, we considered several factors of material mass as well as adsorption efficiency to select a dose of 0.1 g/L for further investigations.

3.2.4. Effect of Contact Time

Figure 9 shows the amount of CFX and TCC antibiotics adsorbed on NiFe₂O₄@C-x composite against contact time. In general, the adsorption occurred rapidly at the initial periods (0–30 min) due to the availability of adsorptive sites. Most of the kinetic adsorption curves reached a threshold of equilibrium at about 180 min. Adsorption CFX and TCC antibiotics by NiFe₂O₄@C-x was unconducive after this moment because desorption was initiated. This event was because the water molecules would enhance the occupation of pores of NiFe₂O₄@C-x, pushing the antibiotic molecules away from the adsorbents. Moreover, the immersion of composite adsorbent in aqueous solutions reduced their stability. Based on such observations, optimum time of 180 min was selected to explore the effect of further parameters such as antibiotic concentration.

According to Figure 9, NiFe₂O₄@C-900 gave the highest removal percentage and adsorption capacity. Therefore, this material was used to calculate the kinetic constants from pseudo first-order, pseudo second-order, Elovich, and Bangham.

Table 2. A summary of kinetic constants for adsorption of CFX and TCC on NiFe₂O₄@C-900.

Models	Parameter	Unit	TCC	CFX
Pseudo first–order	k1	min–1	0.2557	0.1555
	Q1	mg/g	149.47	151.88
	R2	–	0.9917	0.9785
Pseudo second–order	k2	g/(mg·min)	0.0057	0.00205
	Q2	mg/g	151.99	157.64
	R2	–	0.9951	0.9854
Elovich	α	mg/(g.min)	7.7 × 1010	1.54 × 105
	β	g/mg	0.187	0.094
	R2	–	0.9975	0.9787
Bangham	kB	mL/(g/L)	126.36	108.12
	αB	–	0.0367	0.0719
	R2	–	0.9975	0.9773

compares the fitness of kinetic models by coefficient of determination (R²). With R² values greater than 0.9, adsorption kinetics described the experimental data well. TCC adsorption was followed by both Elovich and Bangham, while CFX adsorption was followed by pseudo second–order equation.

3.2.5. Effect of Initial Concentration

Figure 10 shows the amount of CFX and TCC antibiotics adsorbed on NiFe₂O₄@C-x composite against initial concentration. In general, the adsorption capacity would increase with increasing concentration of antibiotics. The composite adsorbent produced at higher temperature gave better results of uptake capacity. NiFe₂O₄@C-900 gave the highest adsorption capacity for CFX (389 mg/g) and TCC (354 mg/g) at initial concentration of 60 mg/L. At the same trend, this material gave the highest percentage of removal for CFX (75.2%) and TCC (76.7%) at initial concentration of 30 mg/L and 20 mg/L, respectively.

To gain insight into the adsorption equilibrium, the isotherm equations including Langmuir, Freundlich, and Temkin were fitted using their nonlinear forms.

Table 3. A summary of isotherm constants for adsorption of CFX, and TCC on NiFe₂O₄@C-900.

Models	Parameter	Unit	TCC	CFX
Langmuir	kL	L/mg	0.0281	0.0542
	Qm	mg/g	827.34	737.42
	R2	–	0.9218	0.9499
Freundlich	kF	(mg/g)/(mg/L)1/n	42.43	75.16
	1/n	–	0.623	0.506
	R2	–	0.9040	0.8711
Temkin	kT	L/mg	0.4035	0.4549
	BT	–	150.81	174.56
	R2	–	0.8542	0.9714

lists the isotherm constants for adsorption of CFX and TCC on NiFe₂O₄@C-900 material. Based on R² values, the best fitting order of CFX adsorption isotherm models would be: Temkin > Langmuir > Freundlich, and that of TCC adsorption isotherm models would be: Langmuir > Langmuir > Temkin. As a result, adsorption of CFX and TCC on NiFe₂O₄@C-900 followed by Temkin and Langmuir, respectively. With a range of values R_L and 1/n between zero and one, the adsorption of antibiotics in this study offered a favorable process. More importantly, the values of maximum adsorption capacity of CFX and TCC identified from Langmuir were very high, 737.42 mg/g and 827.34 mg/g, respectively. It is therefore proposed that NiFe₂O₄@C-900 can be a good adsorbent to adsorb CFX and TCC from water.

3.3. Adsorption Optimization Using Response Surface Methodology

3.3.1. Model Design

From results obtained from investigating the effect of each factor, the optimization model was designed by selecting the powerful factors such as antibiotic concentration, adsorbent dosage, and pH. Based on the widely used central composite design, the range of experimental values are shown in Table 1. There are twenty experimental runs for each antibiotic adsorption model, giving twenty experimental results. The corresponding twenty predicted results were listed in

Table 4. Experimental and predicted adsorption capacity vales for the adsorption models of CFX and TCC on NiFe₂O₄@C-900.

No	CFX					TCC
	Factors			Actual (mg/g)	Predicted (mg/g)	Factors			Actual (mg/g)	Predicted (mg/g)
	x1	x2	x3			x1	x2	x3
1	20	0.05	3	35	33.21	10	0.05	2	49	47.49
2	40	0.05	3	115	124.09	30	0.05	2	83	89.67
3	20	0.15	3	98	94.53	10	0.15	2	53	57.35
4	40	0.15	3	195	207.05	30	0.15	2	74	77.02
5	20	0.05	5	92	95.65	10	0.05	4	55	53.16
6	40	0.05	5	169	188.17	30	0.05	4	90	86.83
7	20	0.15	5	79	85.61	10	0.15	4	86	80.51
8	40	0.15	5	182	199.63	30	0.15	4	89	91.69
9	13.2	0.1	4	81	85.68	3.2	0.1	3	29	32.24
10	46.8	0.1	4	285	258.11	36.8	0.1	3	82	77.10
11	30	0.016	4	150	139.74	20	0.016	3	86	86.48
12	30	0.184	4	213	201.06	20	0.184	3	101	102.71
13	30	0.1	2.3	35	33.07	20	0.1	1.3	75	68.12
14	30	0.1	5.7	100	79.59	20	0.1	4.7	80	85.21
15	30	0.1	4	196	205.80	20	0.1	3	102	102.71
16	30	0.1	4	221	205.80	20	0.1	3	105	102.71
17	30	0.1	4	210	205.80	20	0.1	3	103	102.71
18	30	0.1	4	188	205.80	20	0.1	3	102	102.71
19	30	0.1	4	218	205.80	20	0.1	3	101	102.71
20	30	0.1	4	198	205.80	20	0.1	3	103	102.71

. Overall, the adsorption capacity of CFX antibiotic reached the highest value at 285 mg/g (entry 10), and the lowest at 33.21 mg/g (entry 1). Meanwhile, TCC absorption capacity was the highest value of 105 mg/g (entry 16), and the lowest value of was 29 mg/g (entry 9).

The relationship between the antibiotics adsorption capacity on NiFe₂O₄@C-900 and independent variables x₁, x₂, and x₃ generated from Design-Expert 11 software expressed in the actual equation is as follows:

Q_TCC (mg/g) = −120.61 + 9.89 x₁ + 320.26 x₂ + 55.84 x₃ − 11.25 x₁ x₂ − 0.21 x₁ x₃ + 87.50 x₂ x₃ − 0.17 x₁² − 1421.04 x₂² − 9.21 x₃²,

(3)

Q_CFX (mg/g) = −1148.96 + 11.09 x₁ + 2473.45 x₂ + 470.99 x₃ + 10.75 x₁ x₂ + 0.04 x₁ x₃ –357.50 x₂ x₃ − 0.12 x₁² − 5006.86 x₂² − 52.82 x₃²,

(4)

where x₁ is the antibiotic concentration (mg/L), x₂ is the adsorbent mass (g/L), and x₃ is the pH value. From Equations (3) and (4), all factors had positive, and great effects on adsorption capacity, which means that the value of response would increase with the increase of factors.

To assess the compatibility of each model, statistical analysis such as ANOVA (analysis of variance) discloses the significance based on the model terms such as sum of squares, degrees of freedom, mean of squares, F- and p-values, as well as fitness statistics. In detail, the statistically analytical results are summarized in

Table 5. ANOVA data for the adsorption models of CFX and TCC on NiFe₂O₄@C-900.

Sources	Sum of Squares	Freedom Degree	Mean Square	F-Value	Prob. > F	Remarks
CFX model	87,334.80	9	9703.87	29.74	<0.0001	Significant
x1	35,888.21	1	35,888.21	109.98	<0.0001	–
x2	4538.20	1	4538.20	13.91	0.0039	–
x3	2596.73	1	2596.73	7.96	0.0181	–
x1 x2	231.13	1	231.13	0.7083	0.4197	–
x1 x3	1.13	1	1.13	0.0034	0.9543	–
x2 x3	2556.13	1	2556.13	7.83	0.0188	–
(x1)2	2070.67	1	2070.67	6.35	0.0304	–
(x2)2	2257.94	1	2257.94	6.92	0.0251	–
(x3)2	40,210.20	1	40,210.20	123.22	<0.0001	–
Residuals	3263.20	10	326.32	–	–	–
Lack of Fit	2394.37	5	478.87	2.76	0.1451	Not significant
Pure Error	868.83	5	173.77	–	–	–
Core Total	90,598.00	19	–	–	–	–
Fitness statistics: standard deviation (SD = 18.06), mean = 153.0, coefficient of variation (CV = 11.81%), coefficient of determination (R2 = 0.9640), adequate precision (AP = 17.6179).
TCC model	8412.03	9	934.67	37.57	<0.0001	Significant
x1	2429.05	1	2429.05	97.64	<0.0001
x2	184.72	1	184.72	7.43	0.0214
x3	352.76	1	352.76	14.18	0.0037
x1 x2	253.12	1	253.12	10.18	0.0097
x1 x3	36.12	1	36.12	1.45	0.2559
x2 x3	153.13	1	153.13	6.16	0.0325
(x1)2	4158.80	1	4158.80	167.18	<0.0001
(x2)2	181.88	1	181.88	7.31	0.0222
(x3)2	1222.28	1	1222.28	49.13	<0.0001
Residual	248.77	10	24.88
Lack of fit	239.43	5	47.89	25.65	0.0014	Significant
Pure error	9.33	5	1.87
Core total	8660.80	19
Fitness statistics: standard deviation (SD = 4.99), mean = 82.40, coefficient of variation (CV = 6.05%), coefficient of determination (R2 = 0.9713), adequate precision (AP = 19.9834).

.

According to Table 5, CFX and TCC antibiotic treatment models exhibit F-values of 29.74 and 37.57, respectively, corresponding to p-values less than 0.0001. This means that both models presented a chance of 0.01% that F-values could occur due to noise. As a result, both design models were statistically significant at a reliability level of 95%. More importantly, all factors which obtained p-value less than 0.0001, were statistically significant at the same reliability level. As such, the models have been successfully designed with the statistical significance of all factors [55].

Some fitness statistics such as coefficient of determination (R²) and adequate precision could be used to diagnose the compatibility of models. Obviously, R² values for both models (0.9640–0.9713) were very close to 1.0, suggesting the outstanding fitness properties. In addition, both models had adequate precision (AP) values between 17.6179 and 19.9834. A model with AP value greater than 4.0 offers a high degree of goodness and compatibility. Here, AP values were both greater than 4.0, indicating the high compatibility of proposed models, providing the appropriate signals for the quadratic models [56]. The coefficient of variation (CV%) is a statistical measure of data dispersion level in a collection of data relative to the mean. As observed, both indicators (6.05–11.81%) were lower than 15%, suggesting that the dispersion of experimental data of proposed models should be acceptable.

Diagnostic plots can be used to evaluate whether a model is compatible with the experimental results. One of the most common ones is predicted versus actual plots as shown in Figure 11A,B. It was obvious that model data acquired a good correlation since the data points were concentrated in straight lines. Moreover, plots in Figure 11C,D again supported the evidence of random distribution of data points without any trends or scatters.

3.3.2. Optimization of Process Parameters

The influence of influential factors on the adsorption capacity of TCC and CFX antibiotics by NiFe₂O₄@C-900 can be depicted by three-dimensional surface response plots as shown in Figure 12. Herein, two of three factors would be changed while another was kept at a central level or zero. Overall, the grid regions that described adsorption capacity were converged, indicating that the model could be optimized in the range of investigated values of factors. In detail, for response surfaces of the CFX adsorption model in Figure 12A,C,E, the adsorption capacity would be optimized at around pH 4. This finding was again compatible with surveyed results in the previous sections. The great improvement could be obtained if the CFX concentrations were set to high values (30–40 mg/L). NiFe₂O₄@C-900 adsorbent dose (0.05–0.15 g/L), however, showed a moderately influential effect, but an interactive correlation with other factors.

Meanwhile, for response surfaces of the TCC adsorption model, the response regions could be highly converged, which are highlighted with red color in Figure 12B,D,F. Blue color regions showed lower values of adsorption capacity. Based on optimum surfaces, the TCC concentrations were suggested from 20 to 30 mg/L to obtain higher magnitude of adsorption capacity. The conditions at around pH 3 would facilitate the adsorption of TCC, which was fitted with surveyed results in the previous sections. As similar as the case of CFX adsorption model, the effect of NiFe₂O₄@C-900 adsorbent dose (0.05–0.15 g/L) was moderate. However, TCC adsorption capacities were higher at lower NiFe₂O₄@C-900 dose which is described by Equation (2). Finally, the optimum solutions for adsorption of CFX were proposed as follows, concentration of 40.0 mg/L, adsorbent weight of 0.148 g/L, and pH 3.97. Meanwhile, the optimum solutions for adsorption of TCC were proposed as follows, concentration of 23.93 mg/L, adsorbent weight of 0.115 g/L, and pH 3.6. The desirability values were found at 0.885 and 1.0000 to obtain the highest capacities for CFX and TCC at 256.244, and 105.38 mg/g, respectively.

Model verification was used to repeat the proposed conditions, and attain the statistical errors of experimental capacity values. As expected, the verified errors were lower than 5.0%, suggesting the good compatibility of the quadratic models proposed in this study.

3.4. Recyclability Study

The research on material recyclability are an important strategy to demonstrate the potential effect of sustainable adsorbents on environmental remediation. This experiment was carried out under the following conditions: NiFe₂O₄@C-900 material weight of 0.15 g/L, CFX antibiotic concentration of 40 mg/L, time of 180 min, and pH 4. HCl solvent (0.1 mol/L in ethanol) was used as an effective eluent by referring to the study reported by [57]. Indeed, HCl in aqueous solution disassociates a species of Cl⁻. As a result, it can complete the adsorption with CFX, enhancing the effectiveness of desorption of CFX from NiFe₂O₄@C-900 material. Accordingly, Figure 13 shows that NFO NiFe₂O₄@C-900 could be reused at least three times with a moderate change (~25%) in adsorption capacity, from 201.75 (1st run) to 149.0 mg/g (3rd run). However, there was a considerable decrease in adsorption (98.5 mg/g) for 4th run with 51.2%. Thereby, NiFe₂O₄@C-900 may be a potential material for antibiotic adsorption.

4. Conclusions

In this work, a series of NiFe₂O₄@C-x composites were successfully produced from a bimetallic-based metal-organic framework Ni-MIL-88B(Fe). Production temperatures (600–900 °C) had a strong effect on the structure and adsorption properties of CFX and TCC antibiotics. The characteristics indicated the existence of NiFe₂O₄ nanoparticles in the carbon matrix/layer. The NiFe₂O₄@C-x composites possessed a range of functional groups. For adsorption results, the composite produced at 900 °C obtained higher adsorption capacities for both CFX and TCC. The optimum pH investigations were achieved at pH 4 for CFX and pH 3 for TCC. The other optimum conditions of NiFe₂O₄@C-900 dose, contact time, and concentration were found, at 0.1 g/L, 180 min, and 20–30 mg/L, respectively. Response surface methodology was applied for quadratic models with very high coefficient of determination and adequate precision, suggesting high compatibility with experimental data. For CFX adsorption, the experimental conditions were identified at concentration of 40.0 mg/L, adsorbent weight of 0.148 g/L, and pH 3.97. Additionally, for TCC adsorption, the experimental conditions were identified at concentration of 23.93 mg/L, adsorbent weight of 0.115 g/L, and pH 3.6. The desirability values were found, at 0.885 and 1.0000, respectively. Under such proposed optimum conditions, the experimental adsorption capacities for CFX and TCC were found at 256.244 and 105.38 mg/g, respectively. Recyclability study indicated NiFe₂O₄@C-900 could be reused up to three cycles, offering the potential of this composite adsorbent for removal of emergent antibiotics.