Insight into Adsorption Kinetics, Equilibrium, Thermodynamics, and Modeling of Ciprofloxacin onto Iron Ore Tailings

Abstract

To achieve the resource utilization of iron ore tailings (IOTs), two different IOTs were investigated as sustainable adsorbents for ciprofloxacin (CIP) removal from aqueous systems. Through systematic batch experiments, key adsorption parameters including initial pH, adsorbent dosage, contact time, ionic strength, and temperature were comprehensively evaluated. The results showed that CIP adsorption by IOTs remained relatively stable across a broad initial pH range (2–10), with maximum adsorption capacities of 5-IOT and 14-IOT observed at the initial pH values of 10.1 and 9.16, respectively. Competitive ion experiments revealed a gradual decrease in CIP adsorption capacity with increasing ionic strength (Na⁺, Mg²⁺, and Ca²⁺). Thermodynamic analyses indicated an inverse relationship between adsorption capacity and temperature, yielding maximum adsorption capacities (Q_max) of 16.64 mg/g (5-IOT) and 13.68 mg/g (14-IOT) at 288.15 K. Mechanistic investigations combining material characterization and adsorption modeling identified ion exchange as the predominant interaction mechanism. Notably, trace elements (Cd, Co, Cr, Cu, Fe, Ni, Pb, and Zn) were released during leaching tests, with concentrations consistently below environmental safety thresholds. A back-propagation artificial neural network (BP-ANN) with optimized architecture (8-11-1 topology) demonstrated high predictive accuracy (MSE = 0.0031, R² = 0.9907) for adsorption behavior. These findings suggested IOTs as cost-effective, environmentally compatible adsorbents for CIP remediation, offering the dual advantages of pharmaceutical pollutant removal and industrial waste valorization.

1. Introduction

Antibiotics have been frequently detected as contaminants in various environments [1,2]. Their “pseudo-persistence” and role in promoting antimicrobial resistance have raised significant concerns regarding ecosystem integrity and public health [3]. As a broad-spectrum antibiotic, ciprofloxacin (CIP) has been widely manufactured and used in the treatment of bacterial infections for both humans and animals [4,5]. Thus, CIP diffuses into the environment via multiple pathways, including wastewater discharge, pharmaceutical manufacturing effluents, and agricultural runoff containing livestock manure [6]. Indeed, CIP has been consistently ranked among the top 10 priority pharmaceuticals detected in the aquatic environment [7]. Conventional wastewater treatment plants demonstrate limited removal efficiency for such persistent contaminants [8,9], while effective advanced technologies like advanced oxidation [10] and membrane separation [11] remain energy-intensive and economically prohibitive for large-scale implementation [12]. In this context, adsorption has emerged as a promising alternative for CIP remediation due to its operational simplicity, cost-effectiveness, and adaptability to decentralized systems [13,14]. Recent research has explored various adsorbent categories: mineral-based materials (zeolites and clays) [15,16], industrial by-products such as red mud [12,17] and fly ash [18], and agricultural wastes like biochar [19]. Nevertheless, the ubiquitous detection of CIP in aquatic environments underscores the urgent need for developing novel, cost-effective sorbents with enhanced removal capacities and environmental compatibility.

Iron ore tailings (IOTs), a prominent solid waste generated from iron mining and beneficiation processes, primarily consist of quartz with secondary mineral constituents including feldspar, mica, and calcite [20,21]. The exponential growth of IOT accumulation, paralleling iron ore production rates, has raised significant environmental concerns due to land occupation and ecological risks [22,23], with current utilization rates remaining below 20% [24]. The substantial silicate content of IOTs suggests the potential for the formation of a layered silicate structure [25], which confers notable adsorption capabilities. Previous studies have demonstrated the effectiveness of IOTs as a sorbent for heavy metals [26,27], dye [28], and phosphorus [29], indicating their potential applicability for ciprofloxacin (CIP) removal. However, their capacity for antibiotic remediation, particularly for emerging contaminants in wastewater systems, remains unexplored.

Predicting and understanding relationships among different parameters in adsorption processes by modeling and simulation methods is meaningful to environmental engineering [30]. Removal efficiency is an important parameter for the adsorption process. However, conventional statistical models often prove inadequate due to the multivariate nature of adsorption systems [30,31]. The back-propagation artificial neural network (BP-ANN) has been successfully used in complex environment engineering [32,33]. The BP-ANN modeling is robust and simple [30], capable of simulating intricate multidimensional relationships through experimental data training without requiring prior mechanistic understanding [34]. Based on these advantages, a BP-ANN model was developed to characterize and predict the adsorption behavior of CIP on IOTs.

In this study, IOTs from two distinct sources were explored as sorbents for the removal of CIP in aqueous systems. The adsorption removal of CIP was investigated through systematic analysis of adsorption kinetics and thermodynamics. The study further evaluates critical process parameters affecting adsorption performance and assesses post-adsorption heavy metals (Cd, Co, Cr, Cu, Fe, Ni, Pb, and Zn) leaching to ensure environmental safety. Furthermore, a BP-ANN model was established to describe and predict the removal rate of CIP in different conditions. The results facilitated the prediction of CIP removal in practical applications and engineering design.

2. Materials and Methods

2.1. Materials

Iron ore tailing (IOT) samples were sampled from No.5 and No.14 tailings reservoirs, located in Hebei province, China, which were denoted as 5-IOT and 14-IOT. Before the adsorption experiment, IOT samples were crushed, homogenized into 100 mush, and dried by air. High purity (99.6%) of CIP was obtained from Shanghai Aladdin bio-Chemical (Shanghai, China). Ionic strength in the adsorption system was adjusted with NaNO₃, Ca(NO₃)₂, and Mg(NO₃)₂, which was purchased from Tianjin DAMAO Chemical (DAMAO, Tianjin, China).

2.2. Characterization Methods

X-ray fluorescence spectrometer (XRF: Thermo Scientific, ARL Advant’TM 3600, Waltham, MA, USA) was applied to analyze the chemical compositions of IOT samples. The mineralogical properties of IOT samples were elucidated by X-ray diffraction (XRD: Bruker D8 advance, Beijing China), with Cu Ka radiation at 50 KW, at scan speed = 5 min⁻¹, in a scan range of 2θ = 10–80°. The analysis of XRD was performed by HighScore Plus 3.0.5 (Trial version) with the “COD2021” database. The specific surface area of each IOT sample was measured by the MB-spot test method [35] at 25 °C. The leaching toxicity was tested following the EPA Method 1311. Briefly, 2 g of IOTs were homogenized with 50 mL of acetate buffer solution (pH = 4.93) and agitated on a rotary shaker at 180 rpm under room temperature for 18 h. The resultant mixture was subsequently analyzed for heavy metal content via an atomic absorption spectrometer. The metal in the aqueous phase was analyzed by an atomic absorption spectrometer (PERSEE TAS-990-AFG, Beijing, China) after filtered through a 0.22 μm membrane filter. The zero-point of IOT samples was measured by the previously described method [17]. The pH of the IOT sample was measured in a solution of IOT-to-water of 1:2.5 [36]. The pH values of the aqueous phase were tested by a Starter 300 pH meter (OHAUS, Shanghai, China). The U.S. EPA 9081 method was applied to the cation-exchange capacities (CEC) of IOT samples.

2.3. Adsorption Experiments

The stock solution of CIP (500 mg/L) was prepared using hyper-pure water (18 MΩ/cm) and stored at 5 °C. The adsorption experiments were carried out with 40 mL of CIP solution in a shaking incubator at the speed of 180 rpm. At the end of the experiments, the supernatant was collected by filtering with 0.22 μm filters (Biosharp BS-QT-011, Anhui, China). The concentration of CIP was measured by a Microplate reader (Thermo Scientific, Spectra Max M4, Waltham, MA, USA), with a UV plate (CORNING 3635, Corning, NY, USA) at 277 nm, following the method described previously [17,37]. Apart from the experiments related to pH, the pH for all other experiments was set at the natural pH (7.5) of the wastewater. The contact time, temperature, dosage of IOT, initial pH, and initial concentration of CIP solution were adjusted in specific experiments. The initial pH of the CIP solution was adjusted with 0.01 mol L⁻¹ HNO₃ and NaOH. The removal rate and adsorption capacity of CIP were calculated by Equations (1)–(3):

$$\mathrm{Remove}\mathrm{rate}(\%)={\displaystyle \frac{C_0-C_e}{C_0}}\times 100\mathrm{\%}$$

(1)

$$q_e={\displaystyle \frac{(C_0-C_e)\times V}{M}}$$

(2)

$$q_t={\displaystyle \frac{(C_0-C_t)\times V}{M}}$$

(3)

where C₀, Ce, and C_t (mg·L⁻¹) are the initial concentration of CIP, equilibrium concentration, and concentration at time t (h), respectively. The q_e and q_t (mg/g) are the capacities of CIP removal at equilibrium and time t, respectively. V (L) is the CIP solution volume. M (g) is the mass of the IOT.

2.4. Adsorption Analysis

2.4.1. Kinetic Characterization of CIP Removal

Aiming to explore the adsorption rate, and describe the adsorption process, the pseudo-first-order and the pseudo-second-order were used to fit experimental data. The pseudo-first-order kinetic model was generally expressed as follows:

$$q_t=q_e\left(1-e^{-k_{1}t}\right)$$

(4)

where k_l (h⁻¹) is the rate constant of pseudo-first-order equation

The pseudo-second-order kinetic model was linearized and was expressed as follows:

$${\displaystyle \frac{t}{q_t}}={\displaystyle \frac{1}{k_2{q}_e^2}}+{\displaystyle \frac{1}{q_e}}t$$

(5)

where k₂ (g (mg·h)⁻¹) is the second-order rate constant of adsorption.

2.4.2. Isotherms of CIP Removal

Langmuir, Freundlich, and Dubin–Radushkevich isotherm models were fitted to better explore the mechanism of CIP adsorption by IOTs. The Langmuir equation is shown below:

$$q_e={\displaystyle \frac{Q_{mas}{K}_L{C}_e}{1+K_L{C}_e}}$$

(6)

which can be rearranged in a linear form [38,39,40]:

$${\displaystyle \frac{C_{\mathrm{e}}}{q_e}}={\displaystyle \frac{1}{Q_{\max }}}·C_e+{\displaystyle \frac{1}{K_L{Q}_{\max }}}$$

(7)

where Q_max is the maximum CIP adsorption capacity (mg·g⁻¹), and K_L is the Langmuir affinity constant (L·mg⁻¹), related to the energy of adsorption. The equilibrium parameter R_L was defined as:

$$R_L={\displaystyle \frac{1}{1+K_L{C}_0}}$$

(8)

The Freundlich equation is expressed as follows:

$$q_e=q_F{C}_e^{1/n}$$

(9)

which can be rearranged in a linear form as below [41]:

$$\log q_e=\log K_F+{\displaystyle \frac{1}{n}}\log C_e$$

(10)

where K_F (mg·g⁻¹) is the Freundlich adsorption constant, and 1/n is the Freundlich adsorption intensity parameter (dimensionless).

The Dubin–Radushkevich (D-R) equation is expressed as:

$$\ln q_e=\ln Q_{\max \text{-}D}-K_{\mathrm{D}}{\epsilon }^2$$

(11)

$$\epsilon =RT\ln (1+{\displaystyle \frac{1}{C_e}})$$

(12)

where K_D (mol² kJ⁻²) is the D-R adsorption constant, ε is the Polanyi potential, and Q_max-D is the maximum CIP adsorption capacity based on the D-R model. Also, the mean energy of adsorption (E) can be expressed as:

$$E={\displaystyle \frac{1}{\sqrt{2K_D}}}$$

(13)

2.4.3. Thermodynamic Characterization of CIP Removal

The effect of temperature on CIP removal progress was studied at different temperatures: 288 K, 298 K, and 308 K. The data were used to evaluate the thermodynamic parameters by the Gibbs model, following Equations (14)–(16):

$$\Delta G^0 =\Delta H^0-T\Delta S^0$$

(14)

$$\Delta G^0=-RT\ln K^0$$

(15)

$$\ln K^0={\displaystyle \frac{\Delta S^0}{R}}-{\displaystyle \frac{\Delta H^0}{RT}}$$

(16)

where $\Delta G^0$ is Gibbs adsorption free energy (KJ·mol⁻¹), $\Delta H^0$ is standard enthalpy change (KJ/mol), $\Delta S^0$ is entropy change (J·(mol·K)⁻¹), and $K^0$ is defined as an adsorption equilibrium constant. $R$ is the universal gas constant and equal to 8.314 J·(mol·K)⁻¹. T is the thermodynamic temperature (°K)

2.5. BP-ANN Model and Statistical Analysis

The main parameters, including contact time, temperature, IOT dosage, ionic strength, initial pH, and concentration of CIP solution, were represented as 6 input variables. The whole data were normalized into [−1, 1] and randomized into three parts, training (75%), testing (5%), and validating (20%), respectively. The detailed parameter of BP-ANN was listed in Supplemental Table S1. The performance of the ANN model was analyzed by the mean square error (MSE) and the coefficient of determination (R²). The training parameter for BP-ANN was optimized and set as Table S1.

Unless otherwise stated, each group of run and analysis were in triplicate, and the results are expressed as the average “±” standard deviation with an error bar. Fisher’s test for analysis of variance (ANOVA) was conducted in statistical analysis. The regression of linear or nonlinear in different models, which were carried out in Origin 2021 (OriginLab, Corp., USA, education version). The BP-ANN was calculated by the neural network toolbox in MATLAB R2016 (MathWorks, Inc., Natick, MA, USA, Trial version).

3. Result and Discussion

3.1. The Characterization of IOT

The XRF results showed the chemical composition in the iron ore tailing (IOT) samples was similar, in which SiO₂ was the most abundant constituent, accounting for over 38% of the total mass (Figure 1). Other major constituents in the IOT samples included CaO, MgO, and Fe₂O₃, similar to previous tailing studies [20,42]. Furthermore, the combined content of silica (SiO₂) and calcium oxide (CaO) exceeds 50% in both types of IOT samples. Specifically, in 5-IOT, magnesium oxide (MgO) accounts for 28.5%, while iron oxide (Fe₂O₃) comprises 7.8%. In contrast, in 14-IOT, calcium oxide (MgO) accounts for 11.9%, while iron oxide (Fe₂O₃) comprises 13.0%. These differences in chemical composition, particularly the varying proportions of CaO and Fe₂O₃, may result in the distinct cation exchange capacity and pH values observed in the two IOT samples [43,44]. The mineral composition of the two IOT samples was observed by XRD. The results showed that calcite, phlogopite (potassium magnesium aluminum silicate hydroxide), dolomite, and labradorite (sodium calcium aluminum silicate) were the main mineral phases in 5-IOT (Figure 1A). Dolomite, hornblende (magnesium calcium aluminum iron silicate), and chlorite (magnesium nickel aluminum silicate hydroxide) were the main mineral phases in 14-IOT (Figure 1B). In addition, the specific surface area (SSA) of 5-IOT and 14-IOT were 24.62 ± 0.55 and 22.13 ± 1.32 m²/g, respectively. The pH of these two kinds of IOT was slightly alkaline. The cation exchange capacity (CEC) of 5-IOT (83 mmol (+)·kg⁻¹) was higher than that of 14-IOT (50 mmol (+)·kg⁻¹). The pH_pzc of 5-IOT and 14-IOT are approximately 8.49 and 8.35, according to the delta pH = 0 in different initial pH values (Figure S1).

3.2. Batch Experiment Results

3.2.1. Effect of pH

The bench experiments were conducted with the initial pH (pH_i) ranging from 1 to 11. Temperature, adsorbent dosage, contact time, and initial CIP concentration were maintained at 25 °C, 2.5 g/L, 48 h, and 30 mg/L, respectively. The pH was not adjusted during the entire adsorption process. Results showed that the pH_i in the range of 2–10 had little effect on the equilibrium pH (Figure 2A). This phenomenon might be attributed to calcite and aluminosilicates in IOTs, which could provide buffering capacity during the CIP adsorption process [45]. Meanwhile, the optimal adsorption capacity of CIP onto IOT was observed across a wide range of pH_i from 2 to 10 (Figure 2B), and similar tendencies were found in other tailings [46], indicating the stability of IOTs in aqueous environments with varying pH levels.

As the pH_i increased from 1 to 2, the CIP adsorption capacity of the two IOTs increased remarkably (Figure 2B). The maximum adsorption capacity by 5-IOT was observed at pH_i 10.1, with a capacity of 9.09 mg/g. The maximum adsorption of 14-IOT to CIP occurred at pH_i = 9.16, with a capacity of 6.42 mg/g. At pH values higher than that, the adsorption capacity of both IOTs decreased sharply (Figure 2B). Since the pH shifting was caused by the buffering effect of IOTs after their addition, considering the equilibrium pH and the ionization state of CIP at equilibrium could provide a better understanding of the adsorption behavior. It should be noticed that CIP had two dissociable functional groups: an amino group (pK_a1 = 6.10) and a carboxyl group (pK_a2 = 8.70). When the pH_e value was below 6.10, the amino group of CIPs protonated to form CIPH²⁺. Meanwhile, the main charge of the surface of the IOTs was positive. Consequently, the electrostatic repulsion between CIPH²⁺ and the positively charged IOT surface reduced adsorption capacity under acidic pH_e conditions. A similar phenomenon occurred where the negative charge of IOTs repelled CIP− from their surface, further decreasing the adsorption capacity (Figure 2C). Similar trends of reduced CIP adsorption under extremely acidic or alkaline pH conditions have been reported for red mud [12,17] and clay minerals [16]. The optimal CIP adsorption capacity on IOTs occurred at pH_e values between pK_a1 and pK_a2. Notably, maximum CIP adsorption correlated with the pH of different IOTs (Figure 2D), indicating that the surface of IOTs without any charge may decrease the impediment between CIP and IOT, thus favoring adsorption and increasing the CIP adsorption capacity. Based on these findings, the IOTs are suitable adsorbents for CIP removal across a wide pH range, especially in weakly acidic or neutral wastewater.

3.2.2. Effect of Sorbent Dosage

Different dosages of IOT were evaluated under the following experimental conditions: CIP concentration of 30 mg/L, initial pH of 7.4, adsorption temperature of 25 °C, and contact time of 48 h. As shown in Figure 3, as the IOT dosage increased from 0.5 g/L to 7.5 g/L, the removal efficiencies of 5-IOT and 14-IOT increased from 42.02% to 92.26%, and from 37.58% to 84.60%, respectively. This observed trend aligns with expectations, as higher sorbent dosages provide more adsorption/exchange sites, thereby enhancing the removal rate [47,48]. Meanwhile, the adsorption capacity of IOT exhibited a sharp decrease when the sorbent dosage was increased from 0.5 g/L to 3.5 g/L. This phenomenon was mainly attributed to the non-saturation of the adsorption sites during the adsorption process [49]. When the IOT dosage reached 6.5 g/L, the adsorption sites achieved equilibrium with CIP, resulting in stabilization of the adsorption process. Due to the fact that CIP concentration can rarely reach 30 mg/L [50], our findings demonstrate that a sorbent dosage of 6.5 g/L remains effective for most practical applications.

3.2.3. Effect of Competitive Cation

Given the widespread presence of cations in natural surface water and industrial wastewater, the effects of cation types (Na⁺, Mg²⁺, and Ca²⁺) and competitive cation concentrations on CIP adsorption were systematically investigated. The experiments were conducted under the following conditions, contact time of 48 h, adsorbent dosage of 2.5 g/L, initial CIP concentration of 30 mg/L, initial pH of 7.4, and temperature of 25 °C. Figure 4 illustrates that the CIP adsorption capacity of both IOT samples exhibited a marked decline with increasing cation concentrations. This observed inverse correlation between adsorption capacity and ionic strength suggested that non-specific adsorption and outer-sphere complexation mechanisms were negligible within the tested ionic strength range [51]. Specifically, when Na⁺, Ca²⁺, and Mg²⁺ concentrations increased from 0.01 to 0.5 mol/L, the adsorption capacity of IOT-5 reduced by 57.69%, 65.81%, and 80.7% respectively, compared to the control group. Similarly, 14-IOT showed reductions of 48.33%, 56.46%, and 74.96% under the same conditions. Without consideration of the unstable outer sphere and non-specific adsorption, the cations in water contend with CIP for adsorption onto IOT, and the status became ferocious with increasing concentration of cations, which decreased the CIP adsorption capacity when cations existed. Divalent cations (Ca²⁺ and Mg²⁺) exhibited stronger interference than monovalent Na⁺, likely attributable to their higher charge density and enhanced affinity for IOT adsorption sites [52]. Meanwhile, the higher adsorption affinity depended partly on smaller radii of hydrated ions. The hydrated radii of cations follow the order Mg²⁺ (72 pm) < Ca²⁺ (99 pm) < Na⁺ (102 pm). Smaller hydrated radii facilitate a closer approach to adsorption sites, thereby strengthening interactions. This explains why Mg²⁺ demonstrated the greatest inhibitory effect despite its lower valence than Ca²⁺ [53]. Consequently, lower cation concentrations favor CIP adsorption on IOTs. Under equivalent cation concentrations, the inhibitory effects on CIP removal followed the sequence Ca²⁺ > Mg²⁺ > Na⁺, aligning with the combined influence of ionic charge and hydration characteristics.

3.2.4. Effect of Contact Time

Equilibrium time is a critical parameter in practical applications. The effect of contact time was investigated under the following conditions, adsorbent dosage of 2.5 g/L, initial pH of 7.4, and temperature of 25 °C. Adsorption capacities of both IOT samples for CIP increased rapidly within the first 30 min, followed by a gradual increase, and then reached equilibrium at 48 h. Meanwhile, the adsorption capacity of CIP increased significantly with higher initial concentrations, as elevated concentrations provided a larger mass of CIP capable of occupying more adsorption sites on IOTs [49]. To elucidate the adsorption mechanism, the interparticle diffusion model was applied to identify rate-limiting steps. The adsorption process was divided into three phases according to the fitting curves, as illustrated in Figure S2. The steepest slope in the first phase indicated rapid initial adsorption, while the subsequent two phases represent the primary rate-controlled steps.

3.3. Adsorption Kinetic

The adsorption data of 5-IOT and 14-IOT at different concentrations of CIP were fitted to the pseudo-first-order kinetic model and the pseudo-second-order kinetic models to characterize CIP removal behavior (Figure 5). The pseudo-second-order model exhibited superior fitting (R² > 0.999) across all tested CIP concentrations (

Table 1. Parameters in kinetics model for ciprofloxacin adsorption onto iron ore tailings.

Initial Concentration (mg/L)		The Pseudo-First-Order Kinetic Equation			The Pseudo-Second-Order Kinetic Equation
	Qa	Qe	k1	R2	Qe	K2	R2
	(mg/g)	(mg/g)	(1/h)		(mg/g)	(g/(mg·h))
5-IOT
50	12.26	11.83	5.52	0.9965	12.27	0.54	0.9999
30	8.13	7.78	4.57	0.9920	8.15	0.50	0.9999
20	6.60	6.45	4.74	0.9932	6.64	1.38	0.9999
14-IOT
50	9.80	9.58	4.40	0.9955	9.83	0.88	0.9999
30	6.27	5.93	4.24	0.9872	6.30	0.48	0.9998
20	5.35	5.13	4.67	0.9953	5.35	0.77	0.9998

), with theoretical equilibrium adsorption capacities (Q_e) closely matching experimental values (Q_a). This aligns with previous studies demonstrating the effectiveness of the pseudo-second-order model for describing CIP adsorption kinetics on materials such as fly ash and red mud [12,17]. It is also observed from Table 1 that rate constant k₂ decreased as the initial CIP concentration increased. This trend can be attributed to reduced competition for adsorption sites at lower concentrations. At higher concentrations, intensified competition among CIP molecules for limited sites led to accelerated sorption rates [54].

3.4. Adsorption Isotherms

The Freundlich model, Langmuir model, and D-R model were employed to analyze the relationship between CIP adsorption capacity and equilibrium concentration (Figure 6). As shown in Figure 6, the adsorption isotherms of CIP on IOTs were determined at 298 K, 308 K, and 318 K, with the corresponding isotherm parameters summarized in

Table 2. Parameters in isotherms model for ciprofloxacin adsorption onto iron ore tailing.

T (K)	Freundlich Model			Langmuir Model			Dubin–Radushkevich Model
	n	KF	R2	Qmax	KL	RL	R2	Qmax	Kd	E	R2
5-IOT
308.15	0.526	0.136	0.9986	15.610	0.178	0.208–0.922	0.9412	0.00032	0.003	12.127	0.9977
298.15	0.598	0.181	0.9533	16.077	0.215	0.209–0.916	0.8560	0.00047	0.004	11.180	0.9582
288.15	0.594	0.22	0.9812	16.639	0.160	0.198–0.900	0.9792	0.00047	0.004	10.660	0.9898
14-IOT
308.15	0.356	0.058	0.9912	11.876	0.139	0.091–0.590	0.9425	0.00009	0.023	14.744	0.9771
298.15	0.387	0.081	0.9920	12.484	0.137	0.092–0.594	0.9425	0.00012	0.027	13.609	0.9816
288.15	0.405	0.114	0.9978	13.680	0.104	0.118–0.659	0.8721	0.00015	0.032	12.500	0.9716

. Based on correlation coefficients (R² > 0.99), both the Freundlich and D-R models provided significantly better fits to the experimental data than the Langmuir model (p < 0.05, Figure S3). This result indicated that the adsorption sites on IOT surfaces may exhibit a non-uniform distribution [55]. The Freundlich exponent n values (0.1 < n < 1) further indicated favorable adsorption conditions and implied that CIP adsorption on IOTs likely involved chemical processes such as ion exchange or surface partitioning [56]. Furthermore, 5-IOT has higher K_F and n than 14-IOT at the identified temperature, indicating that the 5-IOT had larger adsorption capacity and higher affinity for CIP. These trends align with the established interpretation of K_F (indicating adsorption capacity) and n (reflecting surface heterogeneity) in Freundlich modeling [57]. The D–R model analysis revealed a mean free energy E value range of 10.66 to 14.74 KJ/mol for IOTs, which falls within the 8–16 kJ/mol threshold characteristic of chemical adsorption mechanisms dominated by ion exchange [55].

Regarding the coefficients in the Langmuir model, the equilibrium parameter R_L for CIP adsorption on IOTs fell within the range 0 < R_L < 1, indicating that CIP adsorption onto IOT was a favorable process [57]. The adsorption constants (K_F, K_D, and K_L) in the three models decreased with increasing temperature, as did the adsorption capacity, suggesting that CIP adsorption onto IOTs may be an exothermic process [57]. The CIP adsorption capacities were compared with other low-cost and readily available sorbents from previous studies (

Table 3. The adsorption capacities compared to other absorbents.

	T (K)	Qmax (mg/g)	Reference
5-IOT	288	16.64 (predicted)	This study
14-IOT	288	13.68 (predicted)	This study
Modified coal fly ash	313	1.157	[58]
Red mud	293	6.13	[17]
Kaolinite	308	7.95	[59]
Goethite	295	19.88	[60]
Montmorillonite	310	397.6	[61]

). The results demonstrated that IOTs exhibited comparable adsorption capacity, making them a cost-effective alternative for CIP removal.

3.5. Adsorption Thermodynamics

Adsorption thermodynamics is a critical aspect of studying IOTs for practical application. It provides important parameters such as feasibility, spontaneity, and the thermal nature of the adsorption reaction [62].

Table 4. Thermodynamic parameters for ciprofloxacin adsorption onto iron ore tailing.

T (K)	ΔG (KJ/mol)	ΔH (KJ/mol)	ΔS (J/(mol·K))
5-IOT
308.15	−27.462	−17.665	31.941
298.15	−27.272
288.15	−26.829
14-IOT
308.15	−25.253	−25.041	0.714
298.15	−25.268
288.15	−25.240

summarizes the adsorption thermodynamic parameters. The negative values of ΔG⁰ indicated the feasibility and the spontaneous nature of CIP adsorption onto IOTs. As the temperature of the system increased, the magnitude of ΔG⁰ became more negative, proving that higher temperatures enhance the thermodynamic favorability of the process. The negative ΔH⁰ value confirms the reaction is exothermic, a finding consistent with adsorption isotherm results. Additionally, the positive ΔS⁰ value suggested increased system disorder as CIP adsorbs onto the IOT surface, with the process being enthalpy-driven rather than entropy-driven.

3.6. The Potential Mechanism of CIP Removal by IOTs

The difference in CIP adsorption capacity between 5-IOT and 14-IOT (Figure S3) was likely attributable to determinants of the adsorption process, which has been considered a primary mechanism for environmental contaminants removal by industrial solid waste [17,63]. Oxide components such as MnO, Fe₂O₃, CaO, MgO, and Al₂O₃ in IOTs are known to contribute to the removal of contaminants like dyes [28] and phosphate [64]. However, despite 5-IOT and 14-IOT having similar total oxide content (Al + Fe + Ca + Mg + Mn-oxides), their differing CIP adsorption capacities suggested that oxide complexation may not be the primary mechanism for CIP adsorption. As indicated by adsorption kinetics and isotherms predictions, chemical reactions might be involved in the adsorption process. Consequently, the cation exchange capacity (ECE) of both IOTs was evaluated. Notably, higher CIP adsorption capacity correlated with greater CEC (Figure S4), indicating that the ion exchange is a dominant mechanism during the CIP adsorption process. This aligns with findings for CIP removal by red mud and clay minerals, where cation exchange was identified as the main mechanism [16,17]. Meanwhile, the addition of competitive cations to the solution significantly reduced CIP adsorption capacity, demonstrating strong competition between cations and exchangeable ions in the sorbent—a key indicator of ion exchange.

The specific surface area (SSA) of both IOTs was described in Section 3.1. Using their CIP adsorption capacities (12.26 and 9.80 mg/g, as detailed in Section 3.3), the surface area occupied per CIP molecule was calculated to be 110.53 Ų and 124.25 Ų, respectively. These values were significantly higher than the cationic amine area range reported for CIP molecules adsorbed on surfaces in either a vertical orientation (17 Ų) [65] or on polar surfaces (80 Ų) [66], suggesting that charge density rather than SSA limited CIP adsorption on IOTs. This was another strong indication of the cation exchange dominating the whole CIP adsorption process mechanism, instead of surface complexation, which also validated the results of the isotherm models.

3.7. Leaching Toxicity After Adsorption

To assess the environmental safety of IOTs as a CIP sorbent, leaching toxicity tests were conducted by analyzing trace elements (Cd, Co, Cr, Cu, Fe, Ni, Pb, and Zn) in solutions after adsorption under varying initial pH conditions (pH_i = 1.06, 3.9, and 11.15), and an IOT dosage of 2.5 g/L. As shown in Table S2, only Fe was detected in the solution of both IOT after 48 h equilibrium, with concentrations far below the 0.5 mg/L threshold specified in the Standards for Drinking Water Quality [67]. Furthermore, compliance with the HJ-557 2010 protocol [68] confirmed the absence of detectable levels of these metals. These results validate the low leaching risks of IOTs across a broad pH range, ensuring their environmental safety as a sorbent.

3.8. Modeling by IOT with Back-Propagation Artificial Neural Network (ANN)

Predicting and understanding adsorption parameters is critical for environmental engineering applications [30]. However, adsorption processes are inherently complex due to nonlinear relationships between input parameters and outputs, making statistical modeling challenging [69,70]. Computational intelligence models, such as artificial neural networks (ANNs), offer superior flexibility in handling nonlinearity and incomplete datasets compared to traditional statistical methods [71]. Given the higher CIP adsorption capacity of 5-IOT, a backpropagation artificial neural network (BP-ANN) model was developed. The number of hidden layer neurons significantly influenced model stability and accuracy [31,48]. After optimization (Table S3), an 8-11-1 architecture achieved the best performance with a mean squared error (MSE) of 0.0031 and R² of 0.9907. Although minor data scattering was observed in Figure 7D, the model demonstrated excellent agreement with validation and testing datasets. The BP-ANN effectively simulated batch adsorption experiments (Figure 7), and training convergence was achieved within 10 epochs (Figure S5). Prior studies emphasize the importance of optimized ANN architectures for adsorption simulations [72]. For instance, Chowdhury and Saha reported that a 3-13-1 structure better modeled methylene blue adsorption on rice husks [73], while Ahmad et al. demonstrated the superior performance of a “3-6-1” structure [31]. These findings align with our results, underscoring the capability of well-structured BP-ANN models to simulate adsorption processes.

4. Conclusions

This study comprehensively evaluated two iron ore tailings (5-IOT and 14-IOT) as low-cost sorbents for ciprofloxacin (CIP) removal from aqueous solutions. Both exhibited maximum adsorption capacities (16.64 mg/g and 13.68 mg/g at 288.15 K, respectively), with optimal performance at initial pH 10.1 (5-IOT) and 9.16 (14-IOT). Adsorption capacity decreased with rising temperature and ionic strength (Na⁺, Mg²⁺, and Ca²⁺), highlighting the need to control these factors in practical applications. Minimal leaching of trace elements confirmed the environmental safety of IOTs. A BP-ANN model (8-11-1 architecture) successfully predicted CIP removal efficiency, demonstrating its utility for process optimization. Overall, IOTs represent an effective, economical, and eco-friendly solution for mitigating CIP contamination in aquatic systems.