Snail shell-derived hydroxyapatite on ant clay: a sustainable solution for phosphate sequestration and circular agricultural reuse

Abstract

In this study, a sustainable adsorbent comprising hydroxyapatite derived from snail shells and clay collected from ant nests was developed and used for the efficient removal of phosphate (PO₄³⁻) from aqueous solutions. The synthesized adsorbent was characterized using different characterization techniques like FTIR spectroscopy, XRD, SEM, TGA, PZC, and BET analysis to confirm its successful synthesis, morphology, thermal stability, surface charges, and surface area, respectively. The adsorbent was systematically investigated for phosphate adsorption by optimizing key experimental parameters, comprising contact time (25 min), adsorbent dosage (25 mg), initial phosphate concentration (5 mg L⁻¹), solution pH (5), and temperature (40 °C). Furthermore, the adsorption kinetics followed the pseudo-second-order model (R² = 0.98), implying that the adsorption rate was controlled by the availability of active sites. The equilibrium data were best fitted to the Langmuir isotherm model (R² = 0.98), with the maximum adsorption capacity of 97.18 mg g⁻¹. The thermodynamic analysis confirmed the endothermic (ΔH° > 0) and spontaneous (ΔG° < 0) nature of the adsorption process. Real sample analysis validated the practical application, with 84.1% removal of PO₄³⁻. Exploratory machine-learning modelling of 84 experimental observations showed that SVR-RBF most accurately predicted phosphate-removal efficiency (cross-validated R² = 0.789 ± 0.127), while Extra Trees regression provided the strongest prediction of adsorption capacity (cross-validated R² = 0.894 ± 0.210). The modelled response trends were consistent with the experimentally identified optimum conditions, supporting the use of data-driven methods for adsorption-process interpretation and future optimization. Additionally, the PO₄³⁻-loaded adsorbent was evaluated as a fertilizer for wheat cultivation; it induced an enhanced growth performance compared to pristine ant clay, highlighting its potential for sustainable agricultural applications. Thus, this study established the adsorbent's efficacy in PO₄³⁻ removal and its suitability for improving growth parameters, including germination rate, shoot/root development, and chlorophyll/protein content, resulting in improved phosphorus bioavailability.

In this study, a sustainable adsorbent comprising hydroxyapatite derived from snail shells and clay collected from ant nests was developed and used for the efficient removal of phosphate (PO₄³⁻) from aqueous solutions.

1. Introduction

Phosphate (PO₄³⁻) is present in small amounts, mostly as inorganic phosphates, organic phosphates, oligophosphates and polyphosphates, in aqueous media. It is important for the development of many living organisms as a nutrient and is necessary for many biochemical processes, but when its concentration increases beyond a certain limit, it causes serious issues. The permissible level of PO₄³⁻ is 1 µg L⁻¹ in water according to the United States Environmental Protection Agency (USEPA).¹ When the concentration of PO₄³⁻ exceeds this limit in water, it can enhance water eutrophication and degrade water quality by causing toxic algal blooms and reducing the concentration of dissolved oxygen, thus resulting in the deterioration of aquatic life.² Eutrophication also increases the growth of zooplankton and phytoplankton and is therefore responsible for the extinction of many species of fish and coral reefs. Consequently, this undesirable phenomenon of eutrophication can be controlled by removing excess phosphorus from water bodies.³

There are a number of different techniques available for PO₄³⁻ removal from aqueous media, i.e., reverse osmosis,⁴ electrodialysis,⁵ ion-exchange process,⁶ and chemical precipitation.⁷ These techniques have their own advantages and limitations, such as high running costs, the generation of secondary pollutants, and complexity. Among these processes, adsorption is a more promising, relatively simple, and widely recognized process for the removal of PO₄³⁻ due to its technical and economic viability, ease of management, high performance, and rapid operation. Selecting appropriate adsorbents is an important consideration that is based on technical skills, ecological considerations, and performance. For this purpose, the adsorption of phosphate (PO₄³⁻) has been investigated using a variety of adsorbents, including biosorbents, modified adsorbents, clay-based materials, and organic adsorbents.⁸ Different studies have been reported the use of natural adsorbents, i.e., red mud,⁹ slag obtained from a blast furnace,¹⁰ and zeolites¹¹ for PO₄³⁻ removal from aqueous media. Recent studies have emphasized the potential of waste-derived and metal-functionalized materials as sustainable adsorbents for water treatment,¹² with calcium and iron containing biochar-based materials demonstrating effective contaminant removal through metal-mediated interactions and surface functional groups.^13,14

Different types of clay, including montmorillonite, bentonite, illite, and chlorite, are examples of adsorbents with high adsorption capacities. Recent studies have proposed that anthill clay (AC) is a fascinating clay structure formed by certain species of ants, which use organic waste and clay to build AC, allowing it to absorb materials from the surroundings. Therefore, anthill clay, a locally available natural mineral, has been reported to be an eco-friendly, effective, and low-cost adsorbent for phosphate removal.¹⁵ In the current study, anthill clay was further used as a supporting material for hydroxyapatite to develop an efficient adsorbent for phosphate sequestration.

Formicidae-associated clay (anthill clay) is a naturally occurring material composed of sand and other mineral particles collected and bound together by ants. It is generally classified into two types, Type I and Type II. Type I has a relatively smaller diameter than Type II. AC comprises metal oxides such as SiO₂, Al₂O₃, and Fe₂O₃, which make it a suitable adsorbent for the removal of heavy metal ions, cationic and anionic dyes, and inorganic ions from wastewater.¹⁶Formicidae-associated clay is a natural product found in the nests and foraging trails of ants. It is a mixture of minerals and organic matter with a significant content of decomposed vegetation, insect fragments, and ant castings. Its biochemical composition features a wide range of cell-wall-degrading enzymes from bacteria, fungi, and higher eukaryotes, including cellulases, proteases, lipases, chitinases, and lignin oxidoreductases.¹⁷

The biodegradation processes of complex organic polymers and their oxidation products contribute to the formation of mineral elements in clay. Thus, clay may have a wide range of applications in environmental biotechnology because of its chemical and physical properties derived from biogenic and mineral components.

The microbial composition and extracellular enzymes in clay are influenced by natural conditions and organic matter content. These factors regulate humic substance formation and the availability of surface functional groups for physicochemical interactions. The mineralogical composition, porosity, and surface area of Formicidae-associated clay determine its capacity to accumulate and retain substances. Thus, this clay represents an inexpensive, environmentally friendly, and promising adsorbent for removing pollutants from water.¹⁸ Although its efficacy as an adsorbent for particular contaminants¹⁹ depends on several factors, including the type of pollutant, the composition of the AC, and the surrounding environment, AC may also be able to adsorb certain pollutant chemicals.²⁰

Alternatively, hydroxyapatite (HAP) (Ca₁₀(PO₄)₆(OH)₂) is a mineral that belongs to the apatite family and has gained significant attention in the scientific community as a superior adsorbent for wastewater treatment.²¹ The chemical composition of HAP includes calcium (Ca²⁺), PO₄³⁻, and hydroxyl (OH⁻) ions, and the surface of HAP has both positive and negative charges, which facilitates the formation of bonds with other atoms.

HAP can be synthesized from a number of calcium-rich natural resources, including eggshells, seashells, crab shells, and naturally occurring minerals.^22,23 The use of such biogenic resources has been gaining significant attention because they provide abundant and inexpensive calcium precursors, as well as offering a sustainable approach for the utilization of waste materials. Among these sources, snail shells are predominantly composed of calcium carbonate (CaCO₃), accounting for approximately 95% of their composition, along with minor amounts of chitin, chitosan, and other inorganic constituents. This high calcium carbonate content makes snail shells an important precursor for HAP synthesis. The successful synthesis of HAP from Cornu aspersum (garden snail) shells has been reported using a combination of calcination and wet-chemical precipitation methods. These results exhibited the potential of snail-shell waste as a sustainable calcium source for the preparation of HAP. HAP is considered a highly efficient adsorbent due to its biocompatibility,²⁴ fast kinetics, and high efficiency. This is due to its unique characteristics, which include low water solubility, strong adsorption affinity for numerous contaminants, and thermal stability.²⁵ However, nearly all pure HAP materials have relatively limited adsorption capacities due to their low specific surface area. Therefore, combining HAP with a suitable substrate is important for improving its characteristics and providing a suitable alternative.²⁶

Machine learning (ML) approaches were also employed in this study to further investigate and predict the phosphate adsorption behavior of the developed adsorbent. Unlike conventional adsorption models, ML techniques can capture complex relationships between experimental variables and adsorption performance. Therefore, the application of ML models provides a predictive and comparative tool for identifying the effects of key experimental parameters and assessing the reliability of the adsorption process.

Consequently, the present study aimed to develop a cost-effective and environmentally friendly AC-HAP composite for PO₄³⁻ removal from aqueous solutions. The specific objectives were to synthesize and characterize snail-shell-derived HAP and the AC-HAP composite. The prepared material was used as an adsorbent, and the effects of contact time, pH, adsorbent dosage, initial phosphate concentration, and temperature on phosphate removal were investigated. The optimized data from the adsorption studies were further evaluated through adsorption kinetic, isotherm, thermodynamic, and regeneration studies, and machine-learning regression algorithms were applied to predict PO₄³⁻ removal efficiency and adsorption capacity. AC-HAP was also used for PO₄³⁻ removal from real wastewater, and the potential reuse of phosphate-loaded AC-HAP as a phosphorus source for agricultural applications was assessed.

2. Materials and methods

2.1. Chemical reagents and equipment

All the chemicals used in this study were of high-purity analytical grade and used as received, without purification. Potassium dihydrogen phosphate (KH₂PO₄, 99%), ammonium molybdate ((NH₄)₂MoO₄, 98%) and ascorbic acid (C₆H₈O₆, 99.5%) were purchased from DaeJung. Hydrochloric acid (HCl, 37%), sodium chloride (NaCl, 99.9%), and sulfuric acid (H₂SO₄, 97%) were purchased from Alfa Aesar. All experimental procedures were conducted using ultrapure distilled water. A standard stock solution of PO₄³⁻ (1000 mg L⁻¹) was prepared by dissolving 1 g of anhydrous potassium dihydrogen phosphate (KH₂PO₄) in 1 L of ultrapure water. Subsequent working solutions with varying PO₄³⁻ concentrations (typically 5–100 mg L⁻¹) were obtained through precise dilution of the stock solution using volumetric glassware. Working PO₄³⁻ solutions of varying concentrations were prepared by diluting a stock solution. The solutions of ammonium molybdate ((NH₄)₂MoO₄) and ascorbic acid (C₆H₈O₆) were prepared, and the water bath was set at 60 °C for complexation.

2.2. Characterization techniques

AC-HAP was characterized using different analytical techniques. Functional groups were identified by FTIR spectroscopy of the samples using a Bruker Alpha 2 over the wavenumber range of 4000–500 cm⁻¹. The crystal structure and phase composition of AC-HAP were characterized by X-ray diffraction (XRD). XRD patterns were recorded using AXRD-LPD at an operating voltage and a current of 40 kV and 40 mA, respectively. The morphology of the prepared sample was analyzed using an ZEISS EVO 15 SEM. The pore volume, surface area, pore size, and pore diameter were analyzed using a BET V-S 280 0P, and thermal stability was analyzed using a TGA-HQG-4.

2.3. Synthesis of adsorbents

2.3.1. Sample collection and pretreatment

Snail shells of Cornu aspersum were collected from COMSATS University Islamabad, Abbottabad Campus (Khyber Pakhtunkhwa), Pakistan, and subsequently washed and cleaned with distilled water to remove sand and particulate contaminants and dried in an oven for 24 h at 100 °C. The obtained shell material was ground into a powder using a ball mill. The produced powder was calcined in a muffle furnace at 900 °C for 3 h to obtain calcium oxide (CaO) from calcium carbonate (CaCO₃). After that, 250 mL of distilled water was added to 14.02 g of CaO powder to prepare a 1 M calcium hydroxide solution. The mixture was stirred for 2 h to obtain a homogeneous suspension for the next step.

2.3.2. Preparation of HAP from snail shells

HAP was synthesized using a combination of wet-chemical precipitation and calcination. A 0.6 M phosphoric acid (H₃PO₄) solution (249.5 mL) was added dropwise at a rate of approximately 15–20 drops min⁻¹ to the Ca(OH)₂ suspension at room temperature under constant stirring at 300 rpm. The amount of H₃PO₄ was calculated to maintain a theoretical Ca/P molar ratio of 1.67, corresponding to the stoichiometric composition of hydroxyapatite (HAP), thereby promoting HAP formation. The pH of the reaction mixture was maintained at approximately 8 by the gradual addition of ammonium hydroxide (NH₄OH). The reaction was continued for approximately 2 h under constant stirring, followed by aging to facilitate complete precipitation. After stirring was completed, a gelatinous white precipitate was obtained and allowed to settle overnight. The precipitate was then washed thoroughly with distilled water, filtered, and dried at 200 °C for 24 h to remove the remaining moisture. The obtained HAP was subsequently sintered in a muffle furnace at 900 °C over a period of 4 h to improve its crystallinity and structural stability.

2.3.3. Collection of AC samples and pretreatment

AC was collected from multiple locations within COMSATS University Islamabad, Abbottabad Campus (Khyber Pakhtunkhwa), Pakistan. Representative and uncontaminated samples of AC were carefully selected for the study. To minimize external contamination, clean, non-reactive stainless-steel trowels and plastic or glass containers were used during sample collection, and the outer layer of each sample was discarded. The collected AC was thoroughly washed with distilled water to eliminate soluble salts, organic matter, and other surface impurities. The clay was vigorously stirred in distilled water and subsequently soaked for 24 h to facilitate the dissolution of soluble contaminants. Samples were washed repeatedly until the supernatant became clear and free from visible impurities. The resulting clay suspension was filtered using filter paper with a pore size of 2 µm to recover the clay particles. The filtered clay was then dried in a hot-air oven at 105 °C for 24 h. After drying, the sample was mechanically pulverized and passed through a 200-mesh standard sieve to obtain a uniform particle size. Finally, the obtained AC samples were stored in airtight containers under ambient conditions to maintain their physicochemical stability before being used in adsorption experiments.

2.3.4. Modification of AC with HAP

To prepare the AC-HAP composite, HAP and AC were mixed at a mass ratio of 1 : 3. Specifically, 10 g of HAP and 30 g of AC were dispersed in 100 mL of distilled water and stirred continuously at 300 rpm for 24 h to ensure uniform mixing and effective surface modification. The resulting suspension was thoroughly washed with distilled water to remove any unreacted species, followed by filtration. The AC-HAP sample was dried at 100 °C for 12 h and then ground using a mortar and pestle to obtain a fine and homogenous powder. A schematic diagram of the preparation of AC-HAP is shown in Fig. 1.

Fig. 1. Synthesis of HAP and AC-HAP.

2.4. Phosphate (PO₄³⁻) determination

A series of standard phosphate (PO₄³⁻) solutions with concentrations ranging from 20 to 60 mg L⁻¹ was prepared in 50 mL volumetric flasks. To each flask, 2 mL of ascorbic acid solution, 2 mL of ammonium molybdate solution, and 0.2 mL of sulfuric acid were added. The reaction mixture was thoroughly mixed and incubated in a water bath at 60 °C for 30 min to allow the formation of the phosphomolybdenum blue complex. The absorbance of the resulting solution was measured using a UV-visible spectrophotometer at the maximum absorption wavelength (λ_max) of 862 nm.

2.5. Point of zero charge (PZC)

The point of zero charge (PZC) is the pH at which the net surface charge of the adsorbent is zero. It can be determined using either the potentiometric titration method or the salt addition method. At the PZC, the adsorbent surface does not become electrically neutral; rather, it possesses equal amounts of positive and negative surface charges, resulting in a net zero surface charge. In the salt addition method, 20 mL of 0.1 mol L⁻¹ NaCl was transferred into a series of conical flasks, and the required pH values were attained with the help of 0.1 M (NaOH) and 0.1 M (HCl). The initial pH of each solution was adjusted to values ranging from 2 to 12 using 0.1 M NaOH or 0.1 M HCl. Subsequently, 15 mg of the biosorbent was added to each flask, and the suspensions were shaken on an orbital shaker for 24 h to attain equilibrium. The change in pH was plotted against the initial pH values. The PZC was determined from the point at which ΔpH = 0, corresponding to the pH at which the initial and final pH values are equal.

2.6. Machine-learning-assisted interpretation of phosphate adsorption

The experimental adsorption dataset was rearranged into a long format for exploratory machine-learning analysis. In total, there were 84 observations, of which 42 measurements were obtained for AC and 42 measurements for AC-HAP. Each observation was result of an individual experiment to assess the effect of contact time, initial phosphate concentration, adsorbent dosage, solution pH, and temperature. For all experiments, the ratio of AC to HAP was maintained as 1 : 3. All experiments were conducted at 300 rpm; consequently, stirring speed was retained as experimental metadata. Because the original experiments followed a one-factor-at-a-time design, the models were constructed to interpolate the measured response profiles rather than to estimate multivariable interaction effects.

Each observation was thus represented by the varying factors, the normalized position of the experimental condition within the tested range and the HAP : AC ratio. The response variables were phosphate-removal efficiency (%) and equilibrium adsorption capacity, q_e (mg g⁻¹). All adsorption-capacity values were converted to a common unit before modelling. Seven regression algorithms were benchmarked: ordinary linear regression, second-order polynomial ridge regression, radial-basis-function support-vector regression (SVR-RBF), k-nearest-neighbors regression, random forest, Extra Trees, and gradient boosting. Model ranking was based on repeated five-fold cross-validation with five repetitions. Predictive performance was quantified using the coefficient of determination (R²), mean absolute error (MAE), and root-mean-square error (RMSE). A separate seven-fold cross-validation was used to generate out-of-fold predictions for parity and residual analyses. Permutation importance was calculated for the best-performing model for each response to evaluate the decrease in predictive accuracy caused by disrupting each encoded feature group.

2.7. Leakage-free cross validation and model interpretation

As the experimental dataset was prepared through one-factor-at-a-time experiments, the observations belonging to the same experimental curve are not completely independent. Hence, random k-fold cross validation may distribute neighboring observations from the same experimental curve between the training and validation sets, resulting in overly optimistic estimates of predictive performance. To minimize this potential information leakage, an additional leave-one-experimental-curve-out (LOCO) grouped cross-validation procedure was performed. The dataset contains 84 observations from five experimental factors evaluated for two adsorbents (AC and AC-HAP), resulting in 10 factor–adsorbent experimental curves. In each LOCO iteration, all observations belonging to one complete experimental curve were removed from model training and used exclusively for validation. Thus, observations from the same experimental curve could not occur at the same time in the training and validation datasets. The final R², MAE, and RMSE values were calculated from the pooled out-of-fold predictions across all 84 observations. Also, permutation importance was interpreted cautiously because the study factor descriptor represents the identity of the experimental series rather than a physical operating variable.

2.8. Ammonium, nitrate, and potassium (NPK) analysis protocol

Wastewater samples were collected from agricultural fields and sludge sites in Abbottabad and homogenized by shaking at 250 rpm for a period of 30 min at room temperature. The samples were then centrifuged twice at 5000 rpm, and the supernatant was collected for analysis. Nitrate nitrogen (NO₃⁻–N) was evaluating through a salicylic acid colorimetric method, with absorbance measured at 410 nm using KNO₃ standards.²⁷ For ammonium analysis, 2 mL of wastewater was mixed with 5 mL of 0.5 M K₂SO₄ and shaken on a shaker at 250 rpm for 1 h at room temperature to facilitate extraction. The mixture was then centrifuged three times at 5000 rpm, and the supernatant was collected for further analysis. Ammonium–nitrogen (NH₄⁺–N) was subsequently determined colorimetrically using the salicylate method, with absorbance measured at 650 nm and quantification performed using NH₄Cl standards.²⁸ The wastewater sample was shaken on a shaker at 250 rpm for 90 min to ensure thorough extraction, followed by centrifugation twice at 6000 rpm. The supernatant was then collected for analysis. Soluble potassium (K⁺) and sodium (Na⁺) concentrations were quantified using flame atomic absorption spectrophotometry, with measurements performed at 766.5 nm for K⁺ and 589.0 nm for Na⁺. Calibration was conducted using KCl and NaCl standard solutions, respectively.²⁹ For total volatile fatty acids (TVFA) and alkalinity, the samples were centrifuged at 4500 rpm for 5 min, and the resulting supernatant was titrated using 0.1 N H₂SO₄. For total reducing sugars (TRS), the samples were centrifuged at 5000 rpm, and the supernatant was collected for analysis. An aliquot of the supernatant was mixed with dinitrosalicylic acid reagent and heated in a boiling water bath for 5 min. After cooling to room temperature, absorbance was measured at 540 nm using a spectrophotometer. TRS concentration was quantified using a standard calibration curve prepared with glucose solutions.

2.9. Adsorption experiments

To determine the optimal conditions for efficient PO₄³⁻ removal from aqueous media, the effects of various adsorption parameters were systematically investigated. The equilibrium time was determined by adding 15 mg of AC-HAP to 150 mL of 30 mg L⁻¹ of PO₄³⁻ solution in a conical flask. The adsorption process was monitored at time intervals ranging from 5 to 40 min. The effect of solution pH was evaluated by adding 15 mg of AC-HAP to 20 mL of 30 mg L⁻¹ PO₄³⁻ solution, with pH adjusted from 2 to 9. The equilibrium study was performed by adding 15 mg of AC-HAP to 20 mL of PO₄³⁻ solutions having initial concentrations ranging from 5 to 45 mg L⁻¹. The suspensions were stirred at 298 K for 35 min to attain adsorption equilibrium. The effect of temperature was investigated by conducting adsorption experiments at varying temperatures between 20 °C and 50 °C. In addition, the influence of adsorbent dosage was evaluated by varying the amount of AC-HAP from 10 to 45 mg while maintaining a PO₄³⁻ concentration of 30 mg L⁻¹. After adsorption, the residual phosphate concentration was determined using a UV-visible spectrophotometer at the maximum absorption wavelength (λ_max) of 862 nm. The PO₄³⁻ removal efficiency (%) and adsorption capacity (mg g⁻¹) were calculated using eqn (1) and (2), respectively.

1

2

here, C_i (mg L⁻¹) and C_e (mg L⁻¹) are the initial and equilibrium phosphate concentrations, respectively; q_e (mg g⁻¹) is the equilibrium adsorption capacity; V (L) is the volume of the solution; and M (g) is the mass of the adsorbent. To investigate the adsorption kinetics, the experimental data were fitted using pseudo-first-order and pseudo-second-order kinetic models. The linear form of the pseudo-first-order kinetic model is given in eqn (3).

ln(q_e − q_t) = lnq_e − k₁t 3

here, q_e (mg g⁻¹) and q_t (mg g⁻¹) are the adsorption capacity at equilibrium and at time t, respectively. k₁ (min⁻¹) is the pseudo-first-order adsorption rate constant. q_t can be calculated using eqn (4).

4

here, C₀ is the initial concentration and C_t is the final concentration at time t.

The pseudo-second-order kinetic model can be expressed in linearized form, as shown in eqn (5).

5

here, q_t (mg g⁻¹) is the adsorption capacity at time t (min) and K₂ (g mg⁻¹ min⁻¹) is the pseudo-second-order rate constant. The linear forms of the intraparticle diffusion and the Elovich kinetic models are given in eqn (6) and (7), respectively.

q_t = k_pt^1/2 + C 6

7

here, k_p (mg g⁻¹ min^−1/2) represents the intraparticle diffusion rate constant and C is the intercept; with the same units as q_t (mg g⁻¹). a_E (mg g⁻¹ min⁻¹) and b_E (g mg⁻¹) are the initial adsorption rate and desorption constant, respectively.

The linearized form of the Langmuir adsorption isotherm is given in eqn (8).

8

here, C_e (mg L⁻¹) is equilibrium concentration of adsorbate in solution, q_e (mg g⁻¹) is equilibrium adsorption capacity, q_m (mg g⁻¹) represents the maximum adsorption capacity of the adsorbent, and K_L (L mg⁻¹) is the Langmuir adsorption constant related to the affinity of the adsorbent for the adsorbate.

The linear form of the Freundlich isotherm is given in eqn (9).

9

here, C_e (mg L⁻¹) is the equilibrium concentration of the adsorbate in solution and q_e (mg ·g⁻¹) is the adsorption capacity at equilibrium. The Freundlich constant K_F ((mg g⁻¹) (L mg⁻¹))^1/n is an indicator of adsorption capacity, and the dimensionless parameter 1/n serves as a heterogeneity factor, indicating the adsorption intensity.

The linear form of the Temkin isotherm is given in eqn (10).

q_e = B_Tln(A_T) = B_TlnC_e 10

here, B_T (kJ mol⁻¹) is the Temkin constant related to the heat of adsorption and A_T (L g⁻¹) is the Temkin equilibrium binding constant corresponding to the maximum binding energy.

The thermodynamic parameters were evaluated using the van't Hoff plot. The Gibbs free energy change (ΔG°) (kJ mol⁻¹), enthalpy change (ΔH°) (kJ mol⁻¹), and entropy change (ΔS°) (kJ mol⁻¹ K⁻¹) are related by the van't Hoff equation, as shown in eqn (11).

11

here, K_d is the equilibrium constant, R is the universal gas constant and T (K) is the absolute temperature. The Gibbs free energy change (ΔG°) was calculated using eqn (12).

ΔG = ΔH − TΔS 12

The physicochemical characteristics of raw wastewater including TYFA and alkalinity were calculated using eqn (13) and (14), respectively. All experiments were performed in triplicate, and the results are presented as the mean ± standard deviation.

13

14

here, Con A is the volume (mL) of 0.1 N H₂SO₄ required to lower the sample pH to 5.0, Con B is the volume (mL) of 0.1 N H₂SO₄ required to further reduce the pH to 4.4, and EF is the extracted sample volume (mL).

The ammonia–ammonium equilibrium relationship given in eqn (15) was used as reported in the literature.³⁰ The concentration of free ammonia nitrogen (FAN) was calculated from the measured total ammonia nitrogen (TAN), pH, and temperature.

15

3. Results and discussion

3.1. Characterization results

3.1.1. Point of zero charge (PZC)

To understand the mechanism of adsorption behavior on the adsorbents, the PZC is an important parameter. The surface of an adsorbent is positively charged at pH values below its PZC and negatively charged at pH values above its PZC. The PZC of AC and AC-HAP are shown in Fig. 2. The PZC of AC was determined to be approximately 4, whereas that of AC-HAP was between 7 and 8. Since PO₄³⁻ is a negatively charged species, electrostatic attraction between phosphate ions and the surface of the adsorbent is expected to be more favorable at pH values below the PZC, where the surface is positively charged.³¹

Fig. 2. PZC of AC and AC-HAP.

3.1.2. Fourier transform infrared (FTIR) analysis

The surface functional groups of the adsorbents were characterized by FTIR spectroscopy over the spectral range of 4000–500 cm⁻¹, as these groups provide active adsorption sites for phosphate removal. The FTIR spectra of AC, HAP, and AC-HAP before and after PO₄³⁻ adsorption are presented in Fig. 3(a) and (b), respectively. The broad absorption band at 3400–3500 cm⁻¹ was assigned to the O–H stretching vibration of surface hydroxyl groups and physically adsorbed water molecules. The distinct absorption band observed at approximately 1000–1050 cm⁻¹ corresponds to Si–O–Si and Si–O–Al asymmetric stretching vibrations, confirming the aluminosilicate framework of AC.³² The band at 700 cm⁻¹ is attributed to Si–O symmetric stretching or quartz-related vibrations, whereas the band at 769 cm⁻¹ corresponds to Al–Mg–OH vibrations from clay sheets, along with metal-OH vibrations in clays that frequently appear below 900 cm⁻¹.³³

Fig. 3. (a and b) FTIR spectra of adsorbents before and after loading PO₄³⁻, respectively, (c) XRD patterns of AC, HAP, and AC-HAP, and (d) TGA curves of snail shell.

The HAP spectrum exhibited the characteristic phosphate vibrations, including ν₃ stretching bands at 1030–1090 cm⁻¹, a ν₁ stretching band near 960 cm⁻¹, and ν₄ bending bands at approximately 602 and 565 cm⁻¹, together with a broad hydroxyl stretching band around 3400 cm⁻¹. The successful incorporation of HAP onto AC was confirmed by the coexistence of the characteristic phosphate bands at 1030–1090, 960, and 565–602 cm⁻¹ with the silicate bands of AC.³⁴ In addition, carbonate bands were observed at 1394–1453 and 868 cm⁻¹, further confirming the successful preparation of AC-HAP. A slight shift in the peak positions indicated PO₄³⁻ adsorption on the composite material.³⁵

The intensity and shape of the phosphate peaks (around 1040, 960, 600, and 560 cm⁻¹) shown in Fig. 3(a) and (b) appear to change and increase in intensity compared to the before adsorption spectrum. The changes in the phosphate-related peaks after adsorption suggest interactions between phosphate species and the AC–HAP surface. However, it is important to note that HAP itself contains structural phosphate groups, and thus FTIR analysis alone cannot definitively distinguish between the intrinsic phosphate groups of HAP and the externally adsorbed phosphate species. Therefore, ligand exchange is considered a possible adsorption mechanism rather than being conclusively established by FTIR.

3.1.3. XRD and TGA analyses

The crystalline structure and phase composition of AC and AC-HAP were examined by XRD analysis, as shown in Fig. 3(c). The XRD pattern of AC revealed the presence of hematite (Fe₂O₃), kaolinite (Al₂Si₂O₅(OH)₄), and a significant amount of quartz (SiO₂), as well as clay minerals, with quartz exhibiting higher peak intensities at 2θ = 26.2°, 39.3°, 50.9°, and 63.1° due to its abundance in the sample.³⁶ The sharp peak at approximately 12.4° (001) corresponds to the characteristic basal reflection for kaolinite. The HAP (Ca₁₀(PO₄)₆(OH)₂) diffraction peaks emerged at 26.2° (002), 27.1° (102), 29.4° (210), 31.8° (211), 32.6° (300), 34.0° (202), 46.9° (222), and 48.3° (320), corresponding to the crystalline planes of HAP (JCPDS 09-0432).³⁷ Notably, the peak at 28.1° overlapped with the quartz signal at 28.2°, indicating a combined contribution from both SiO₂ and HAP, as shown in Fig. 3(c). The coexistence of diffraction peaks corresponding to both quartz-rich AC and HAP confirms the successful formation of the AC-HAP composite.

Fig. 3(d) shows the TGA curves illustrating the thermal behaviour of the snail shell powder before calcination and the snail shell-derived HAP as a function of temperature. The thermal stability of the snail shell powder is relatively high up to about 600 °C, showing a gradual weight loss from about 100% to 90%. This initial weight loss is owing to the loss of moisture and volatile/organic materials adsorbed on the natural snail shell. A significant weight loss was observed in the temperature range of 650 °C to 800 °C, with the weight decreasing from 90% to 47%. This can be attributed to the decomposition of calcium carbonate (CaCO₃), the main inorganic component of the snail shell, resulting in the release of CO₂ as a major thermal event. The curve is almost constant after about 800 °C, which indicates the formation of a thermally stable inorganic residue mainly consisting of CaO.³⁸ The snail shell-derived HAP exhibited a much more gradual weight loss across the temperature range. The absence of the sharp decomposition event seen in the original snail shell powder suggests that the original carbonate-rich structure was significantly altered during processing. In contrast, the snail shell-derived HAP sample exhibited relatively high thermal stability with about 76% residue remaining after heating to 1000 °C. Additional weight loss occurred gradually, which can be a result of the removal of adsorbed/bound water or residual volatile (organic, and/or carbonate) species.³⁹

3.1.4. BET analysis

The textural properties of AC and AC-HAP were evaluated by N₂ adsorption–desorption analysis using the BET method, and the corresponding adsorption–desorption isotherms are presented in Fig. 4. The pore size distribution and pore volume were calculated using the Barrett–Joyner–Halenda (BJH) method from the adsorption branch of the isotherm, and the results are summarized in Table 1. The BET analysis showed that the surface area of AC was 10 m² g⁻¹, while the surface area of AC-HAP was 12 m² g⁻¹. The BJH analysis revealed average pore radii of 194 Å and 179 Å for AC and AC-HAP, respectively. Both samples exhibited typical type IV adsorption–desorption isotherms, with H3 hysteresis loops, characteristic of mesoporous materials comprising slit-shaped pores formed by the aggregation of plate-like particles.⁴⁰ The incorporation of HAP slightly increased the BET surface area from 10 to 12 m² g⁻¹, accompanied by an increase in pore volume from 0.035 to 0.057 cm³ g⁻¹. The slight increase in surface area, as well as the higher pore volume, after modification is expected to increase adsorption capacity. Furthermore, the presence of HAP introduces Ca-rich active sites capable of interacting with PO₄³⁻ ions. Consequently, the enhanced phosphate removal performance of AC-HAP can be attributed to the combined effects of its improved textural properties and the chemical functionality imparted by HAP.

Fig. 4. BET analysis of AC and AC-HAP.

Table 1. Surface area, pore size, and pore volume of AC and AC-HAP.

Sample	Surface area (m2 g−1)	Pore size (Å)	Pore volume (cm3 g−1)
AC	10	194	0.035
AC-HAP	12	179	0.057

3.1.5. SEM analysis

The surface morphology of AC and the AC-HAP composite was investigated by SEM, and representative micrographs of the samples are shown in Fig. 5. Prior to modification, AC exhibited a compact and plate-like structure with some aggregations (Fig. 5(a) and (c)). Nevertheless, when AC was functionalized with HAP, the surface of AC-HAP became rough and irregular, suggesting that the deposition of HAP may provide additional accessible adsorption sites (Fig. 5(b) and (d)). HAP modification reduced the compact aggregation of the clay particles and produced a rougher, more heterogeneous surface with irregularly distributed fine particles, thereby increasing the surface roughness and porosity, as also supported by a previously reported study.⁴¹ These observations are in good agreement with the BET analysis, which revealed a minor increase in surface area and an increase in pore volume following HAP incorporation, confirming the successful fabrication of the AC-HAP composite.

Fig. 5. SEM images of AC (a and c), and AC-HAP (b and d).

3.2. Optimization studies

3.2.1. Effect of contact time

Contact time is an important parameter that governs the adsorption kinetics of phosphate onto the adsorbent surface. The effect of contact time on the removal efficiency of PO₄³⁻ using both AC and AC-HAP was investigated at intervals of 5–40 min, as depicted in Fig. 6(a). The adsorption curves obtained this study exhibited typical two-phase behavior, with an initial rapid uptake stage and a subsequent asymptotic approach to equilibrium.⁴² This kinetic trend is due to the initially abundant availability of surface-active sites, which enable the capture of PO₄³⁻ in the early stages. In the case of both adsorbents, the adsorption efficiency was observed to increase over time until reaching the optimum contact times of 30 min for AC and 25 min for AC-HAP. As adsorption proceeds, the number of available active sites decreases, and diffusion resistance increases, leading to a gradual reduction in the adsorption rate until equilibrium is reached. Quantitative analysis showed that AC and AC-HAP had maximum PO₄³⁻ removal efficiencies of 82.5% and 94.5%, respectively, with the modified adsorbent exhibiting a higher capacity and faster kinetics. The enhanced adsorption performance of AC-HAP can be attributed to the incorporation of HAP, which provides additional hydroxyl groups and calcium-rich active sites capable of interacting with PO₄³⁻ as high-affinity coordination sites through electrostatic attraction, surface complexation, and ion exchange.

Fig. 6. (a) Effect of contact time on the adsorption of PO₄³⁻ by AC and AC-HAP. (b) Pseudo-first-order kinetic model, (c) pseudo-second-order kinetic model, (d) Elovich model, and (e) intraparticle diffusion model.

The adsorption kinetics were evaluated using kinetic models, including the pseudo-first-order, pseudo-second-order, and Elovich models, as shown in Fig. 6(b)–(d), respectively, while the intraparticle diffusion to elucidate the adsorption mechanism is shown in Fig. 6(e).

Among these kinetic models, the pseudo-second-order model shown in Fig. 6(c) provides the best fit to the experimental data, indicating the highest adjusted R² value of 0.98, which is consistent with the data but does not by itself confirm a chemisorption-controlled mechanism. The calculated equilibrium adsorption capacity value (q_e = 86.21 mg g⁻¹) from the pseudo-second-order model also matches well with the experimental value (q_e = 87.91 mg g⁻¹), further supporting its suitability for describing phosphate adsorption onto AC-HAP.⁴³ The Elovich model also showed a good correlation coefficient (R² = 0.94), suggesting that adsorption occurred on a heterogeneous surface through chemical interactions between PO₄³⁻ and the active functional groups of AC-HAP.⁴⁴ The results are consistent with the heterogeneous morphology observed in the SEM images. The intraparticle diffusion model exhibited a relatively good correlation coefficient (R² = 0.92), with an intraparticle diffusion rate constant (k_p) of 8.66 mg g⁻¹ min^−1/2. However, the regression line did not pass through the origin, and the relatively high intercept (C = 21.56 mg g⁻¹) showed that intraparticle diffusion was not the sole rate-controlling step.⁴⁵ Both intraparticle diffusion and boundary-layer diffusion contributed to the overall adsorption process. The corresponding kinetic parameters for all the evaluated models are summarized in Table 2.

Table 2. Parameters of the kinetic models.

Adsorption kinetics	Parameters	Values
Pseudo-first-order	K 1 (min−1)	0.202
	q e (mg g−1)	219.83
	R 2	0.78
Pseudo-second-order	q e (mg g−1)	86.21
	K 2 (g mg−1 min−1)	0.1394 × 10−4
	R 2	0.98
Intraparticle diffusion	k p (mg g−1 min−1/2)	8.66
	C (mg g−1)	21.56
	R 2	0.92
Elovich	b E (mg g−1 min−1)	30.47
	a E (g mg−1)	0.057
	R 2	0.94

3.2.2. Dosage study

To determine the optimal adsorbent dosage, varying amounts (5–45 mg) of AC and AC-HAP were added to a 30 mg L⁻¹ PO₄³⁻ solution, and the corresponding removal efficiencies were evaluated. As shown in Fig. 7(b), the removal efficiency of PO₄³⁻ increased to a maximum of 92.8% at an adsorbent dosage of 25 mg, whereas AC-HAP achieved a higher removal efficiency of 98.1% at the same dosage. The optimum dosage for both AC and AC-HAP was therefore determined to be 25 mg, beyond which no significant improvement in removal efficiency was observed. The enhanced performance of AC-HAP can be attributed to the incorporation of HAP, which provides additional calcium-rich active sites and –OH groups on the adsorbent surface that enhance phosphate adsorption. The plateau observed at higher dosages is due to the saturation of available adsorption sites, limiting further PO₄³⁻ uptake regardless of the addition of additional adsorbent.³⁵

Fig. 7. (a) Effect of pH of the PO₄³⁻ solution on the adsorption process. (b) Effect of the dosage of AC and AC-HAP. (c) Effect of temperature on the PO₄³⁻ adsorption process by AC and AC-HAP. (d) van't Hoff plot for AC-HAP.

3.2.3. pH study

The effect of pH on the phosphate adsorption process was explored because pH significantly influences both the surface charge of the adsorbent and the speciation of phosphate ions, thereby affecting the removal efficiency. The adsorption of PO₄³⁻ by both AC and AC-HAP was strongly dependent on the solution pH, as shown in Fig. 7(a). In the case of AC, the maximum phosphate removal of 84.5% was obtained at pH 2. In contrast, the phosphate removal efficiency by AC-HAP increased as the pH increased from 3 to 5, reaching a maximum removal efficiency of 95.5% at pH 5. Beyond the optimum pH, the adsorption efficiency gradually decreased. The enhanced adsorption behavior can be explained by considering the surface charges of the adsorbent and the aqueous speciation of phosphate.

Since the pH_pzc of AC-HAP was determined to be 7.5, the adsorbent surface is positively charged, favouring electrostatic attraction of the dominant species H₂PO₄⁻. Besides electrostatic interactions, the calcium-rich active sites and surface –OH groups of HAP can also promote phosphate adsorption though surface complexation and ligand exchange, respectively. However, at low pH values, excessive protonation of the adsorbent surface, together with partial dissolution of HAP, may reduce the number of active sites for the adsorption of phosphate. At pH above the pH_pzc, the adsorbent surface is negatively charged, and the dominant phosphate species are HPO₄²⁻ and PO₄³⁻, resulting in electrostatic repulsion and competition from OH⁻ ions for the available adsorption sites.⁴⁶ In contrast, the maximum phosphate removal by AC was found at pH 2, where the pH_pzc was approximately 4, resulting in a positively charged adsorbent surface. The predominant phosphate species, H₂PO₄⁻, was electrostatically attracted to the protonated surface hydroxyl groups of AC, leading to enhanced adsorption. As the pH increased above the pHpzc, the AC surface gradually became negatively charged, while phosphate increasingly existed as HPO₄²⁻. The resulting electrostatic repulsion between the negatively charged adsorbent surface and phosphate ions, together with competition from OH⁻ ions for adsorption sites, led to a decrease in phosphate removal efficiency.

3.2.4. Thermodynamic study

Thermodynamic studies are important for understanding the influence of temperature on the feasibility, spontaneity and energetic nature of the adsorption process. The influence of temperature on the removal of PO₄³⁻ by AC and AC-HAP is presented in Fig. 7(c). It was observed that the adsorption efficiency increased with increasing temperature for both adsorbents, indicating that elevated temperatures favored PO₄³⁻ removal. These results suggest that the adsorption process is endothermic, as higher temperatures provide sufficient kinetic energy for PO₄³⁻ ions to diffuse more effectively towards the adsorbent surface and interact with the active sites. For AC-HAP, adsorption increased steadily up to 40 °C, beyond which only a slight improvement was observed, indicating that adsorption equilibrium was reached due to the progressive occupation of the available active sites.

The thermodynamic parameters were determined from the van't Hoff plot, as shown in Fig. 7(d). The positive value of ΔH° confirms that the adsorption process is endothermic, while the positive ΔS° reflects an increase in randomness at the solid–liquid interface during PO₄³⁻ adsorption. The positive ΔS° value can be attributed to the displacement of water molecules and structural rearrangement at the adsorbent surface. Furthermore, the negative ΔG° values at all investigated temperatures demonstrate that the adsorption process is spontaneous, while the increasingly negative ΔG° values with increasing temperature indicate that higher temperatures enhance the spontaneity of phosphate adsorption. The calculated thermodynamic parameters (ΔH°, ΔG°, and ΔS°) are summarized in Table 3.

Table 3. Thermodynamic parameters for AC-HAP.

Temperature (K)	ΔG° (kJ mol−1)	ΔH° (kJ mol−1)	ΔS° (kJ mol−1 K−1)
293	−20.75	70.22	0.310
298	−22.30
303	−23.85
308	−25.40
313	−26.96
318	−28.36
323	−29.88
328	−31.36

3.2.5. Equilibrium study

The initial concentration of PO₄³⁻ is an important parameter influencing the adsorption performance of adsorbents. The effect of initial PO₄³⁻ concentration was investigated over the concentration range of 5–45 mg L⁻¹ at room temperature using both AC and AC-HAP, as shown in Fig. 8(a). The respective equilibrium adsorption capacities are presented in Fig. 8(b).

Fig. 8. (a) Effect of the initial concentration of PO₄³⁻ solution on percentage PO₄³⁻ removal, (b) corresponding adsorption capacity (q_e), (c) Freundlich isotherm, (d) Langmuir isotherm and (e) Temkin isotherm.

The results revealed that the percentage removal of phosphate decreased with increasing initial phosphate concentration for both adsorbents. This decline can be attributed to the fixed amount of adsorbent used during the experiments, which provides a limited number of active adsorption sites. As the initial phosphate concentration increased, these active sites gradually became occupied, leading to increased competition among phosphate ions and consequently reducing the percentage removal. In contrast, the adsorption capacity (q_e) increased with increasing phosphate concentration because the higher concentration gradient provided a greater driving force for mass transfer, allowing more phosphate ions to be adsorbed per unit mass of adsorbent until equilibrium was reached.

Throughout the investigated concentration range, AC-HAP exhibited superior adsorption performance compared with AC. The maximum phosphate removal efficiencies of AC-HAP and AC were approximately 97% and 84%, respectively, while their corresponding maximum adsorption capacities were 87.9 mg g⁻¹ and 56.0 mg g⁻¹. The enhanced adsorption performance of AC-HAP is attributed to the successful incorporation of HAP, which introduces additional calcium-rich active sites and surface hydroxyl groups that facilitate phosphate adsorption through electrostatic attraction, surface complexation, and ligand exchange. Furthermore, the slightly higher surface area, larger pore volume, and rougher surface morphology of AC-HAP improve the accessibility of phosphate ions to the available adsorption sites, resulting in enhanced adsorption efficiency.

To understand the adsorption mechanism, the concentration study data were further analyzed by applying different isotherm models, including the Freundlich isotherm, Langmuir isotherm and Temkin isotherm, as shown in Fig. 8(c)–(e), respectively. The Langmuir isotherm model assumes monolayer adsorption onto homogeneous surface sites, where each active site accommodates only one adsorbate molecule without interactions between adjacent adsorbed species. In contrast, the Freundlich isotherm describes multilayer adsorption on heterogeneous surfaces, where the heterogeneity factor (1/n) serves as an indicator of adsorption intensity and surface energy distribution for PO₄³⁻ adsorption onto AC-HAP. The Langmuir isotherm (Fig. 8(d)) exhibited the best fit to the experimental data, with an adjusted R² value of 0.98. However, given the confirmed heterogeneous structure of AC-HAP, this fit was used to estimate the maximum adsorption capacity for comparison purposes rather than as evidence of a homogeneous, monolayer surface.⁸ The theoretical maximum adsorption capacity (q_m) calculated from the Langmuir model was 97.18 mg g⁻¹, demonstrating the high adsorption affinity of the AC-HAP composite for phosphate ions. In addition, the Langmuir adsorption constant (K_L = 0.881 L mg⁻¹) reflects a strong adsorbent–adsorbate interaction and high adsorption affinity. The Freundlich model (Fig. 8(c)) also provided a satisfactory description of the adsorption equilibrium, with an adjusted R² value of 0.95. The Freundlich adsorption intensity parameter (n = 2.44) was greater than unity, indicating favorable adsorption, while the value of 1/n = 0.409 suggested moderate surface heterogeneity and a non-uniform distribution of adsorption sites. The high Freundlich constant (K_F) further reflects the strong adsorption capacity of the composite. Although the Freundlich model suggests the expected heterogeneity of the AC-HAP surface, the Langmuir model provided a relatively higher R² and was therefore retained for estimating the maximum adsorption capacity (q_max). This statistical preference does not indicate that adsorption occurs strictly as a homogeneous monolayer, given the confirmed heterogeneous structure of the composite.

The Temkin isotherm (Fig. 8(e)) illustrates adsorption on heterogeneous surfaces, predicting a linear decrease in the heat of adsorption with surface coverage.⁴⁷ Unlike the Langmuir model, the Temkin model assumes that the heat of adsorption decreases linearly with increasing surface coverage due to interactions between the adsorbed species. The Temkin model produced an adjusted R² value of 0.92, indicating a reasonable fit to the experimental data, although it was lower than those obtained for the Langmuir and Freundlich isotherms. This suggests that while adsorbate–adsorbate interactions may occur during phosphate adsorption, they are not the dominant mechanism governing the adsorption process. The Temkin equilibrium binding constant (K_T = 18.59 L g⁻¹) indicates a relatively strong affinity between phosphate ions and the AC-HAP composite. The Temkin constant (B_T = 16.54 mg g⁻¹), which is related to the variation in adsorption energy, suggests that the heat of adsorption decreases gradually as surface coverage increases. This behavior is expected because the most energetically favorable adsorption sites are occupied first, followed by adsorption onto sites with progressively lower energies.⁴⁸

The better fit of the Langmuir model can be attributed to the presence of abundant Ca²⁺ active sites on the HAP surface, where phosphate ions are adsorbed through ligand exchange and electrostatic interactions, resulting in the formation of a relatively uniform adsorption layer. Meanwhile, the contribution of the clay component introduces surface heterogeneity, consistent with the reasonable fit of the Freundlich model. Instead of indicating strictly monolayer adsorption, this suggests that adsorption on AC-HAP likely involves a combination of processes, with the Langmuir model serving as a useful approximation for estimating the maximum capacity rather than confirming true surface homogeneity. The values of all the parameters calculated from the various isotherms are provided in Table 4.

Table 4. Parameters of adsorption isotherms for the adsorption of PO₄³⁻ onto AC-HAP.

Models	Parameters	AC-HAP
Langmuir isotherm	q m (mg g−1)	97.18
	K L (L mg−1)	0.881
	R 2	0.98
Freundlich isotherm	K F ((mg g−1) (L mg−1))1/2	3.9 × 104
	n	2.44
	R 2	0.95
Temkin isotherm	K T (L g−1)	18.59
	B T (J mol−1)	16.54
	R 2	0.92

3.3. Proposed adsorption mechanism

AC-HAP possesses surface-exposed Ca²⁺ ions that facilitate PO₄³⁻ adsorption through ligand exchange and surface complexation. In this mechanism, PO₄³⁻ replaces surface hydroxyl groups, forming stable Ca–PO₄ complexes and releasing OH⁻, where indicates a surface-bound species.

Ca(OH)₂ + H₃PO₄ → Ca–O–PO₃H + 2H₂O 16

It was observed that the percentage removal was lower under strongly acidic conditions, due to the high concentration of H⁺ ions that compete with PO₄³⁻ ions and partially protonate the adsorbent surface, thereby reducing electrostatic attraction. In addition, under acidic conditions, phosphate exits as H₂PO₄⁻ rather than PO₄³⁻, and partial dissolution of the HAP phase may further decrease the availability of surface Ca²⁺ adsorption sites.

In alkaline media, OH⁻ ions compete with PO₄³⁻ ions for Ca²⁺ sites on HAP, forming Ca(OH)₂ on the surface. Resulting decrease in phosphate adsorption. At pH 5–6, the maximum adsorption was therefore observed, where phosphate species interacted most effectively with the active sites of the AC-HAP composite.

The possible adsorption mechanism between AC-HAP and PO₄³⁻ ions involves electrostatic attraction between positively charged Ca²⁺ sites on HAP and negatively charged PO₄³⁻ species,⁴⁹ followed by ligand exchange to form stable calcium–phosphate complexes. Besides these interactions, weak van der Waals interactions and hydrogen bonding are also involved in the adsorption mechanism. In addition, Si–OH and Al–OH groups present in the structure of clay may also serve as active sites that facilitate phosphate adsorption on AC, as follows:

Si–OH + HPO₄²⁻ → Si–O–PO₃H⁻ + OH⁻ 17

Al–OH + HPO₄²⁻ → Al–O–PO₃H⁻ + OH⁻ 18

The Langmuir isotherm provided the best fit to the equilibrium data; however, given the heterogeneous structure of AC-HAP, this fit was used primarily to estimate the maximum adsorption capacity rather than to imply a truly homogeneous surface. The pseudo-second-order kinetic model indicated that the adsorption rate was influenced by interactions between phosphate ions and the available adsorption sites. These findings suggest that phosphate adsorption onto AC-HAP involves multiple interaction mechanisms involving electrostatic interactions, hydrogen bonding, and other weak intermolecular forces, while surface complexation with Ca²⁺ sites also enhances the adsorption affinity as shown in Fig. 9(a) and (b). Thermodynamic analysis revealed the endothermic and spontaneous nature of the adsorption process, while the positive entropy change indicates that the adsorption of PO₄³⁻ onto AC-HAP is thermodynamically favorable.

Fig. 9. (a) Interaction of phosphate ions with AC-HAP adsorbent. (b) Proposed mechanism for the adsorption of PO₄³⁻ on the AC-HAP adsorbent.

3.4. Real sample analysis and regeneration study

For practical evaluation, PO₄³⁻-contaminated wastewater samples were collected from agricultural fields and sludge sites in Abbottabad. The samples were filtered to eliminate suspended solids, and the initial PO₄³⁻ concentration in the filtrate was quantified via UV-vis spectroscopy at λ_max = 862 nm. The filtered wastewater was subsequently treated with the optimum adsorbent dosage. After treatment, the residual PO₄³⁻ concentration was measured at the same wavelength, and the removal efficiencies were calculated. The results demonstrated PO₄³⁻ removal efficiencies of 65.13% for AC and 84.10% for AC-HAP, highlighting the superior performance of the composite adsorbent. The approximately 19% improvement in phosphate removal by AC-HAP compared with pristine AC can be attributed to electrostatic interactions, hydrogen bonding, physical adsorption, and surface complexation between phosphate species and the Ca²⁺-, Si–OH-, and Al–OH-containing active sites of the composite. These synergistic interactions significantly improve the adsorption affinity and phosphate uptake capacity of AC-HAP, highlighting its potential as an efficient and sustainable adsorbent for practical phosphate removal from wastewater.

In addition, physicochemical characteristics of the raw wastewater sample were determined, as presented in Table 5. In this study, total solids (TSS), volatile solids (VSS), pH, electrical conductivity (EC), and total and soluble chemical oxygen demand (COD) were determined according to APHA Standard Methods.⁵⁰ Ammonium nitrogen, nitrate nitrogen, and potassium concentrations were determined following the protocols described in a previous study.⁵¹ Total reducing sugars (TRS) were determined using the DNS (3,5-dinitrosalicylic acid) method, with glucose as the standard.⁵² Total volatile fatty acids (TVFAs) and alkalinity were determined using the Nordmann method and calculated according to eqn (13) and (14), respectively, as described previously.⁵³ The adsorption performance of AC-HAP was further compared with those of previously reported adsorbents, as shown in Table 6.

Table 5. Physicochemical characteristics of the raw wastewater sample used in this study.

Parameter	Unit	Mean ± SD
pH	—	7.09 ± 0.56
Total volatile fatty acids (TVFAs)	g L−1	0.41 ± 0.12
Chemical oxygen demand (COD)	g L−1	5.78 ± 1.46
Total reducing sugars (TRS)	mM	24.85 ± 12.72
Phosphate removal (PO43−–P)	mg L−1	4.26 ± 0.56
Free ammonia nitrogen (FAN)	mg L−1	69.61 ± 32.03
Ammonium nitrogen (NH4+–N)	mg L−1	3.16 ± 0.48
Nitrate nitrogen (NO3−–N)	mg L−1	4.78 ± 0.86
Potassium (K+)	mg L−1	111.65 ± 65.55
Sodium (Na+)	mg L−1	28.61 ± 6.97
Conductivity	mS cm−1	15.54 ± 3.45
Total suspended solids (TSS)	mg L−1	320 ± 48
Volatile suspended solids (VSS)	mg L−1	250 ± 35
Total alkalinity	mg CaCO3 L−1	1260 ± 50

Table 6. Comparison of the PO₄³⁻ removal performance of different adsorbents with that of the proposed AC-HAP adsorbent.

Adsorbents	Adsorption capacity, mg g−1	Reference
Magnetic amorphous lanthanum silicate alginate hydrogel beads (MALS-B)	40.14	54
La–Fe-incorporated chitosan beads (LaFe@CS)	52.0	55
Lanthanum cross-linked polyvinyl alcohol/alginate/palygorskite (LPAP) composite hydrogel beads	85.35	56
Magnesite-modified seaweed hydrochar–chitosan composite hydrogel beads	73.41	57
Exfoliated Moroccan lava clay (ghassoul)	72.31	58
Mg-modified biochar/yttrium alginate	67.20	59
La-doped bentonite/Cs composite	23.52	60
NH-SA-ZrBT	63.90	61
Mg Al-modified biochars	56.12	62
l-Histidine-modified montmorillonite	81.96	63
AC-HAP	97.61	This study

In order to evaluate the reusability and structural stability of the adsorbents, regeneration studies were performed on both AC and AC-HAP over six successive adsorption–desorption cycles using NaOH as the regenerant (Fig. 10). The phosphate-loaded AC and AC-HAP were separated through rinsing with deionized water to remove unbound or loosely adsorbed ions and then treated with 0.5 M NaOH solution for 1 h under constant stirring to desorb the bound phosphate and regenerate the active sites. The regenerated adsorbent was subsequently rinsed repeatedly with deionized water until the rinsate reached a neutral pH, dried at 80 °C, and reused for the next adsorption cycle. This adsorption–desorption process was repeated for six consecutive cycles, and the removal efficiency (%) was recorded after each cycle to assess the stability and reusability of AC and AC-HAP under identical regeneration conditions. Both AC and AC-HAP showed a gradual decline in phosphate removal efficiency with increasing cycle number, which is expected due to progressive saturation and incomplete desorption of the active sites; however, the two materials exhibited markedly different behavior, as shown in Fig. 10.

Fig. 10. Regeneration study of AC and AC-HAP.

For cycles 1–5, AC-HAP consistently outperformed AC, retaining 73–95% removal efficiency compared to 70–82% for AC. These results suggest that the HAP coating not only improves the initial interactions with phosphate but also regenerates more effectively under NaOH treatment because NaOH efficiently reverses the electrostatic and ligand–exchange interactions between phosphate and the HAP surface. A considerable decline in AC-HAP performance was observed in cycle 6, with the removal efficiency dropping sharply from 73% to 52%, thereby falling below that of AC (68%) for the first time. This behavior suggests a shift from a chemically limited to a structurally limited regeneration regime. Repeated treatment with NaOH likely promoted progressive dissolution or restructuring of the HAP coating on the composite surface, thereby compromising its structural integrity during repeated use. Meanwhile, AC retained considerable stability even after six adsorption–desorption cycles, as the carbon matrix of clay remained structurally unaffected by NaOH treatment, while the HAP coating on AC-HAP was progressively consumed through repeated alkaline treatment.

3.5. Model benchmarking

The regression algorithms differed substantially in their ability to reproduce the phosphate-removal response, as shown in Fig. 11. SVR-RBF exhibited the best overall performance, with a repeated-cross-validation R² of 0.789 ± 0.127, an MAE of 3.65 and an RMSE of 5.19. Polynomial ridge regression produced a nearly identical mean R² (0.788 ± 0.096), but its MAE and RMSE were slightly higher. Extra Trees, gradient boosting, k-nearest neighbors, and random forest showed moderate predictive accuracy, whereas ordinary linear regression performed poorly (R² = 0.089; RMSE = 10.93 mg g⁻¹).

Fig. 11. Comparative performance of the seven regression algorithms for the prediction of phosphate-removal efficiency and equilibrium adsorption capacity. (a and b) Mean R² and (c and d) RMSE obtained from repeated five-fold cross-validation with five repetitions. Error bars represent the standard deviation across validation folds.

Adsorption capacity was predicted most accurately using ensemble tree-based models.⁶⁴ Extra Trees regression ranked first, with a repeated-cross-validation R² of 0.894 ± 0.210, an MAE of 3.69 mg g⁻¹, and an RMSE of 4.98 mg g⁻¹, as presented in Tables 7 and 8. Random forest yielded a slightly lower mean R² (0.850 ± 0.067) but showed less variation among validation folds, indicating a comparatively stable performance. SVR-RBF and gradient boosting also captured the principal trends, whereas linear regression yielded a negative mean cross-validation R² (−0.014), showing that a constant mean prediction was on average, more reliable than the linear model. Together, these results demonstrate that the relationship between the encoded experimental conditions and adsorption capacity was strongly nonlinear.⁶⁵

Table 7. Full regression metrics for phosphate-removal efficiency.

Model	Repeated-CV R2	MAE (%)	RMSE (%)
SVR-RBF	0.789 ± 0.127	3.648	5.194
Polynomial ridge	0.788 ± 0.096	4.139	5.288
Extra trees	0.665 ± 0.206	4.377	6.546
Gradient boosting	0.638 ± 0.191	5.193	6.871
k-nearest neighbors	0.616 ± 0.234	5.070	7.061
Random forest	0.534 ± 0.256	5.899	7.784
Linear regression	0.089 ± 0.475	8.722	10.925

Table 8. Full regression metrics for equilibrium adsorption capacity.

Model	Repeated-CV R2	MAE (mg g−1)	RMSE (mg g−1)
Extra Trees	0.894 ± 0.210	3.690	4.979
Random forest	0.850 ± 0.067	5.298	7.335
SVR-RBF	0.812 ± 0.147	4.431	8.023
Gradient boosting	0.763 ± 0.236	6.489	8.520
Polynomial ridge	0.624 ± 0.198	7.854	11.325
k-nearest neighbors	0.602 ± 0.248	7.334	11.947
Linear regression	−0.014 ± 0.272	13.966	19.567

3.6. Best diagnostic model

The out-of-fold predictions were used only for diagnostic visualization and were not treated as an independent external test set. For removal efficiency, the SVR-RBF parity plot showed close agreement with the 1 : 1 line across most of the experimental range, giving an out-of-fold R² of 0.869, an MAE of 3.15, and an RMSE of 4.55 (Fig. 12(a)). The Extra Trees model predicted adsorption capacity with an out-of-fold R² of 0.962, an MAE of 3.06 mg g⁻¹, and an RMSE of 3.99 mg g⁻¹ (Fig. 12(b)). The residuals of both models were distributed above and below zero without a uniform directional bias (Fig. 12(c) and (d)); nevertheless, several larger errors occurred in regions where the measured response changed sharply. For this reason, the repeated-cross-validation statistics provide a more conservative estimate of model performance.

Fig. 12. Diagnostic evaluation of the best-performing models. Experimental versus seven-fold out-of-fold predicted values for (a) phosphate-removal efficiency using SVR-RBF and (b) adsorption capacity using Extra Trees regression. The dashed diagonal line denotes ideal agreement. Residual distributions for (c) SVR-RBF and (d) Extra Trees calculated as the difference between the experimental and out-of-fold predicted values.

3.7. Permutation importance

Permutation analysis identified the categorical study-factor descriptor as the most influential encoded feature for both response variables, followed by the normalized factor level, as shown in Fig. 13. For removal efficiency, the mean permutation scores were 9.61 ± 1.34 for study-factor identity, 8.72 ± 2.52 for normalized factor level, and 3.88 ± 0.68 for the HAP : AC ratio. For adsorption capacity, the corresponding scores were 23.34 ± 3.92, 18.96 ± 5.24, and 2.86 ± 0.74. These results show that the model relied primarily on distinguishing the five experimental series and locating each condition within its tested range. The contribution of the HAP : AC ratio indicates that adsorbent composition helped the models distinguish between AC and AC-HAP responses. However, this analysis does not constitute a direct ranking of pH, dosage, temperature, concentration, and contact time because factor identity was represented as a grouped categorical variable and only two composition states were available.

Fig. 13. Cross-validated permutation importance for the prediction of (a) phosphate-removal efficiency using SVR-RBF and (b) equilibrium adsorption capacity using Extra Trees regression. Higher values indicate a greater deterioration in predictive performance after random permutation of the corresponding encoded feature.

3.8. Model-smoothed experimental response profiles

The best-performing removal model was subsequently used to generate smooth, within-range response profiles for the five one-factor experiments, as shown in Fig. 14(a)–(e). These curves are presented as model-assisted interpolations of the measured observations rather than response surfaces or evidence of interactions among simultaneously varied operating parameters.

Fig. 14. Model-smoothed experimental response profiles for phosphate removal by AC and AC-HAP as a function of (a) initial phosphate concentration, (b) adsorbent dosage, (c) contact time, (d) solution pH, and (e) temperature. Symbols denote experimental observations, and solid curves denote within-domain model predictions from the SVR-RBF model. The plots interpolate separate one-factor-at-a-time experiments and should not be interpreted as multivariable interaction surfaces.

3.8.1. Initial phosphate concentration

Increasing the initial concentration from 5 to 45 mg L⁻¹ decreased the removal efficiency from 89.47% to 50.29% for AC and from 98.42% to 78.22% for AC-HAP, as show in Fig. 14(a). In parallel, adsorption capacity generally increased because the larger concentration gradient supplied more phosphate ions to the available adsorption sites. For AC-HAP, the adsorption capacity increased from 12.30 to 87.99 mg g⁻¹, whereas AC reached a maximum adsorption capacity of approximately 63.45 mg g⁻¹ at 35 mg L⁻¹ before declining at a higher concentration. The smaller decrease in removal efficiency for AC-HAP indicates greater resistance to increased phosphate loading.

3.8.2. Adsorbent dosage

Removal efficiency increased as the dosage was increased from 5 to 25 mg and subsequently approached a plateau. At 25 mg, AC and AC-HAP achieved removal efficiencies of 95.83% and 98.11%, respectively, as shown in Fig. 14(b). The simultaneous decrease in adsorption capacity per unit mass at higher dosages is consistent with reduced fractional utilization of the available sites. The combined experimental and modelled profiles therefore identify 25 mg as a practical dosage, beyond which additional adsorbent provides little improvement in removal.

3.8.3. Contact time

AC-HAP showed faster uptake than AC, increasing from 52.85% removal at 5 min to 94.47% at 25 min and remaining near 94–96% thereafter, as shown in Fig. 14(c). AC required approximately 30 min to reach 82.50% removal. The rapid initial stage followed by a plateau is consistent with the occupation of readily accessible sites and subsequent approach to adsorption equilibrium. Accordingly, 25 min was selected as the practical contact time for AC-HAP.

3.8.4. Solution pH

The AC-HAP response was distinctly nonlinear, with a maximum removal efficiency of 95.53% and a corresponding adsorption capacity of 71.64 mg g⁻¹ at pH 5. Performance declined under neutral-to-alkaline conditions, whereas AC varied within a narrower range of approximately 79–85%, as shown in Fig. 14(d). The mildly acidic optimum is consistent with favorable attraction between protonated surface sites and anionic phosphate species below the point of zero charge of the modified adsorbent.

3.8.5. Temperature

Removal by AC-HAP increased from 64.25% at 20 °C to 97.32% at 40 °C and then decreased slightly at 45–50 °C. The corresponding adsorption capacity reached 72.99 mg g⁻¹ at 40 °C. AC showed a more gradual increase from 54.61% at 20 °C to 85.18% at 50 °C, as shown in Fig. 14(e). The positive temperature dependence up to the optimum agrees with the endothermic character derived from the thermodynamic analysis, whereas the plateau above 40 °C indicates that further heating is not operationally advantageous for AC-HAP.

3.9. Scope of the model

The present analysis demonstrates that nonlinear regression models can reproduce the principal adsorption trends contained in the original dataset. Their predictive scope is nevertheless restricted to interpolation within the experimentally investigated one-factor series. Interactions such as pH × dosage, concentration × temperature, or contact time × HAP content were not sampled and therefore cannot be inferred from the current models. In addition, the composition variable comprised only unmodified AC and AC-HAP prepared at a fixed 1 : 3 ratio, while stirring speed was fixed at 300 rpm. A subsequent factorial or response-surface design with prospectively measured validation experiments will be required before the framework can be used for genuine multivariable optimization.

The removal-efficiency model shows a substantial reduction in generalization performance when an entire experimental curve is withheld (pooled R² = 0.259 versus random-CV R² = 0.789), confirming that the random split was optimistic for cross-series prediction. In contrast, the Extra Trees adsorption-capacity model remains comparatively strong under curve-wise validation (pooled R² = 0.858), although its MAE and RMSE increase relative to random CV. Therefore, the original random-CV statistics should be described as within-series interpolation performance rather than evidence of fully independent generalization.

The following tables show the prediction errors obtained when each complete experimental curve was held out once. Individual-curve R² values should be interpreted cautiously when a curve has little response variation; the pooled LOCO metrics in Table 9 are the primary grouped-validation statistics. LOCO diagnostics for removal efficiency by SVR-RBF and adsorption capacity using Extra Tree regression are shown in Tables 10 and 11, respectively.

Table 9. Comparison of the original random repeated five-fold CV with leakage-free LOCO grouped validation.

Target/model	Validation	R 2	R 2 SD	MAE	RMSE
Removal efficiency/SVR-RBF	Random repeated 5-fold CV	0.789	0.127	3.648%	5.194%
Removal efficiency/SVR-RBF	LOCO grouped CV	0.259	—	8.461%	10.812%
Adsorption capacity/Extra Trees	Random repeated 5-fold CV	0.894	0.210	3.690 mg g−1	4.979 mg g−1
Adsorption capacity/Extra Trees	LOCO grouped CV	0.858	—	5.616 mg g−1	7.679 mg g−1

Table 10. LOCO diagnostics for removal efficiency (SVR-RBF).

Held-out curve	N	R 2	MAE (%)	RMSE (%)
Contact time – AC	8	−0.459	14.024	16.895
Contact time – AC-HAP	8	0.402	10.474	12.152
Dosage – AC	9	−3.836	11.435	11.784
Dosage – AC-HAP	9	0.073	6.194	6.787
Initial concentration – AC	9	−0.387	11.667	14.798
Initial concentration – AC-HAP	9	−1.161	6.165	9.076
Temperature – AC	7	0.459	5.709	8.172
Temperature – AC-HAP	7	0.645	5.644	6.492
pH – AC	9	−8.300	3.969	5.211
pH – AC-HAP	9	−0.681	8.930	10.385

Table 11. LOCO diagnostics for adsorption capacity (Extra Trees).

Held-out curve	N	R 2	MAE (mg g−1)	RMSE (mg g−1)
Contact time – AC	8	0.231	7.253	9.202
Contact time – AC-HAP	8	0.720	4.755	6.236
Dosage – AC	9	0.984	2.671	3.855
Dosage – AC-HAP	9	0.983	2.671	3.855
Initial concentration – AC	9	0.741	7.022	9.084
Initial concentration – AC-HAP	9	0.788	7.451	11.677
Temperature – AC	7	0.578	4.403	5.412
Temperature – AC-HAP	7	0.409	5.662	6.281
pH – AC	9	−17.930	4.982	5.576
pH – AC-HAP	9	−1.977	9.116	10.365

The high permutation importance of the study-factor descriptor should not be interpreted as evidence that one operational variable is intrinsically more important than another. Rather, it reflects the ability of the model to distinguish between the separate one-factor experimental series. Because pH, dosage, contact time, initial concentration, and temperature were not varied simultaneously, the present permutation analysis does not provide a causal interpretation or physical ranking of these operational variables.

3.10. Effect of PO₄³⁻-absorbed AC-HAP on wheat seed germination and growth

Two experimental systems were compared: the control group, which consisted of untreated soil, and the test group, consisting of soil that was amended with phosphate-loaded AC-HAP. The parameters used to monitor growth, such as the rate of germination, plant height, and the density of leaves, were measured at 8 day intervals. The findings revealed much better growth in the AC-HAP-treated group, with improvements across all measured parameters. There are a few mechanisms that may account for this improved performance. The phosphate-impregnated AC-HAP composite acts as an efficient slow-release phosphorus source, which releases bioavailable PO₄³⁻ in a slow and controlled manner as a result of the desorption of phosphate and the dissolution of minerals. The high surface area and porosity of the material enhance the soil aeration and water-holding capacity, while the calcium ions in HAP maintain the optimal soil pH, as well as inhibit phosphorus fixation. The combined effects of sustained phosphorus availability, improved soil conditions, and reduced environmental stressors explain the observed 25–30% increase in germination rate, 40% increase in plant height, and 35% increase in leaf density compared to the control. These results highlight the dual benefit of AC-HAP as a useful wastewater treatment medium and a soil amendment, thereby contributing to the idea of a circular economy in the management of water and nutrients. This multifunctional application may improve the efficiency of phosphorus utilization compared with traditional fertilizers, contributing to the recorded increase in the plant growth parameters (Fig. 15(a) and (b)). The comparison of growth, morphological, and biochemical parameters between the control and treatment groups following fertilizer application is presented in Table 12. The experiment proves that spent AC-HAP recovered from wastewater treatment can be successfully used as a high-performance fertilizer and has economic and environmental advantages due to the recovery of resources and valorization of waste.

Fig. 15. (a) Effect of PO₄³⁻-loaded AC-HAP on wheat germination and growth. (b) Scheme for the comparison of biochemical parameters between the control and treatment groups.

Table 12. Comparison of growth, morphological, and biochemical parameters between the control and treatment groups following fertilizer application.

Parameter	Control group	Treatment group	Effect
Seed germination (%)	90%	95%	5% increase; fertilizer showed no toxic effects, enhanced by phosphorus availability
Germination initiation (days)	8	6	Germination began 2 days earlier in the treatment group
Plant height (cm)	12.44 (days 10–30), 22 (days 30–60)	14.8 (days 10–30), 23 (days 30–60)	Initially, faster growth; after 30 days, growth slowed due to phosphorus depletion
Shoot length (cm)	22 ± 3	23 ± 4	1 cm increase; due to phosphorus aiding metabolic processes like cell elongation
Root length (cm) and no. of branches	16 ± 3; roots: 3 ± 2	20 ± 2; roots: 4 ± 1	4 cm increase; improved root branching and nutrient uptake due to phosphorus availability
Leaf length (cm) and no. of leaves	10 ± 3; leaves: 3 ± 1	13 ± 2; leaves: 4 ± 1	3 cm increase; attributed to phosphorus supporting cell division and chlorophyll production
Leaf color	Pale green, yellowish lower leaves	Uniformly green, vibrant	Chlorosis in control; adequate chlorophyll in treatment due to phosphorus
Fresh weight (g)	0.43 ± 0.06	0.59 ± 0.09	Increase biomass; linked to phosphorus-enhanced nutrient uptake and metabolic activity
Dry weight (g)	0.09 ± 0.01	0.14 ± 0.02	Improved structural integrity and growth due to phosphorus
Chlorophyll content (SPAD)	19.16 ± 3.6	28.23 ± 3.8	47.34% increase; higher phosphorus content facilitated photosynthesis and chlorophyll biosynthesis
Carotenoid content (mg g−1)	0.6 ± 0.2	0.8 ± 0.4	33.33% increase; linked to enhanced photosynthesis and IPP production
Protein content (mg g−1)	0.7 ± 0.3	1.0 ± 0.4	42.86% increase; phosphorus supported nucleotide and protein synthesis

4. Conclusion

In this study, HAP produced from snail shells was effectively integrated with AC to create an effective adsorbent for PO₄³⁻ removal in aqueous solutions. The successful modification of AC with HAP was verified through FTIR, SEM, BET, TGA, PZC, and XRD characterization. Kinetic data revealed that adsorption followed a pseudo-second-order model, while isotherm analysis showed that the Langmuir model provided the best statistical fit. The thermodynamic examination of the adsorption procedure confirmed that it was an endothermic and spontaneous process. Exploratory machine-learning analysis complemented the conventional adsorption models by demonstrating that nonlinear SVR-RBF and Extra Trees algorithms accurately interpolated phosphate-removal efficiency and adsorption capacity, respectively, while independently reproducing the experimentally identified favorable conditions for AC-HAP, namely, 25 mg dosage, approximately 25 min contact time, pH 5, and 40 °C. The composite material was found to exhibit an improved adsorption capacity, i.e., 97.18 mg g⁻¹, because of the synergistic effects of the high affinity of HAP for PO₄³⁻ and the porous nature and anion-retention characteristics of the clay. Also, the spent material, which retained adsorbed phosphorus that could be gradually released, was utilized as a fertilizer, which led to a substantial enhancement in plant growth. Growth parameters, such as germination rate, shoot/root development, and chlorophyll/protein content, evidenced an increase in phosphorus bioavailability. These results suggest the dual application of AC-HAP in wastewater treatment and cultivation. Further research should optimize fertilizer formulations and assess their long-term agronomic and soil health impacts. Consequently, this study underscores the potential of bio-waste-derived materials for eco-friendly pollution control and agricultural productivity.

Conflicts of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

Data are provided within the manuscript.