Engineering a smart multifunctional UiO-66(Zr)-based metal–organic framework for simultaneous nutrient monitoring and water remediation via spectrophotometric sensing, adsorption, Box–Behnken design, machine learning, and DFT analysis

Abstract

The development of multifunctional materials for phosphate detection and removal is crucial for combatting eutrophication and enhancing water quality. We report the preparation of a novel sulfisoxazole-functionalized zirconium-based metal–organic framework (SFX-UiO-66) through glutaraldehyde-assisted modification of NH₂-UiO-66(Zr). This framework features various active sites for effective phosphate recognition and adsorption. Characterization confirmed successful functionalization, high crystallinity, and structural stability pre- and post-phosphate uptake. The sensor enabled rapid phosphate detection (response time: 60 s) with a linear range of 0.111–0.200 ppm (R² = 0.999), a detection limit of 3.65 × 10⁻² ppm, and a quantification limit of 1.11 × 10⁻¹ ppm, while demonstrating excellent selectivity over competing anions. SFX-UiO-66 exhibited a maximum Langmuir adsorption capacity of 202.8 mg g⁻¹ under optimal conditions (pH 5, 0.02 g adsorbent, 100 min, 45 °C), aligning with the Langmuir isotherm and pseudo-second-order kinetics, indicating monolayer chemisorption mechanisms. Response surface methodology based on Box–Behnken design optimized the adsorption process. An artificial neural network (ANN) model achieved R² = 0.9815, indicating strong predictive capability for phosphate adsorption. DFT simulations revealed a significant decrease in the HOMO–LUMO energy gap post-phosphate coordination, favoring charge transfer between phosphate ions and active sites. The sensor also demonstrated excellent performance for tap and boiler water samples, with recoveries of 100.29–100.73%, consistent with ion chromatography. These results suggest that SFX-UiO-66 is an effective dual-functional technology for rapid phosphate detection and removal, with potential for environmental monitoring and sustainable water purification.

The development of multifunctional materials for phosphate detection and removal is crucial for combatting eutrophication and enhancing water quality.

1. Introduction

Phosphorus enters surface water in both organic and inorganic forms, but it is the orthophosphate ion, PO₄³⁻, that acts as the immediately bioavailable species driving primary productivity in lakes, rivers, and coastal systems.¹ Agricultural fertilizer runoff, detergent formulations, and inadequately treated municipal and industrial effluent are the dominant anthropogenic sources, and their combined loading has pushed phosphorus concentrations in many watersheds well past the levels that a stable aquatic ecosystem can absorb.² Regulatory guidance on phosphate is comparatively loose relative to toxic heavy metals: the World Health Organization treats it as an aesthetic rather than a health-based parameter, and the U.S. Environmental Protection Agency does not set an enforceable maximum contaminant level for drinking water, instead recommending that streams discharging into lakes remain below roughly 0.05 mg L⁻¹ and that in-lake concentrations stay under 0.025 mg L⁻¹ to avoid triggering algal blooms.^3,4 The absence of a hard regulatory ceiling has not made the problem smaller; it has simply shifted the burden of control onto watershed-level monitoring programs that need frequent, distributed measurements rather than occasional laboratory samples.

The ecological consequence of phosphate over-enrichment is well documented. Excess phosphorus fuels algal proliferation, and the subsequent microbial decomposition of algal biomass depletes dissolved oxygen, producing hypoxic or anoxic zones that are lethal to fish and benthic invertebrates.⁵ Cyanobacterial blooms associated with eutrophic conditions can additionally release microcystins and other toxins that compromise drinking water treatment and pose direct risks to human and animal health during recreational or occupational water contact.¹ Because phosphorus is also a finite, mined resource central to global fertilizer production, uncontrolled loss of phosphate into waterways represents a resource-recovery failure as much as a pollution problem, motivating growing interest in sensing platforms that can support both surveillance and recovery.⁶

Conventional phosphate quantification relies on the molybdenum blue colorimetric method, ion chromatography, or inductively coupled plasma techniques. Each delivers good accuracy in a controlled laboratory setting, but all three demand sample transport, trained personnel, and either continuous reagent supply or expensive multi-element hardware, none of which suit real-time, distributed monitoring of the kind watershed managers increasingly need.⁷ Portable colorimetric field kits reduce cost but typically sacrifice selectivity and long-term reagent stability, and paper-based microfluidic devices, while inexpensive, still report detection limits in the low mg L⁻¹ range, well above the concentrations associated with early-stage eutrophication risk.⁸ This mismatch between the sensitivity needed at the point of concern and the sensitivity practically achievable in the field has sustained interest in solid-state chemosensors that combine a molecular recognition element with a robust, easily regenerated support.

Porous crystalline solids, mesoporous silica, and carbon-based scaffolds have all been explored as sensor supports, but metal–organic frameworks (MOFs) occupy a distinctive niche because their pore size, surface chemistry, and metal connectivity can be adjusted independently through linker and metal-node selection.⁹ A MOF built from a rigid, chemically addressable linker offers a scaffold onto which a chromogenic recognition unit can be installed without disturbing the underlying framework topology, and the resulting composite typically inherits both the porosity of the parent MOF and the optical behavior of the grafted chromophore.

Among the zirconium-based frameworks, UiO-66(Zr) has become something of a workhorse for aqueous sensing applications precisely because it tolerates the conditions that break down many other MOF families. Its Zr₆O₄(OH)₄ secondary building unit is held together by strong Zr–O bonds, giving the framework structural integrity across a wide pH window and resistance to hydrolytic collapse that alternatives.^10,11 The amino-functionalized variant, NH₂-UiO-66(Zr), retains this stability while introducing a reactive handle on the 2-position of the terephthalate linker, and that amine has become a standard entry point for post-synthetic modification (PSM) chemistry aimed at installing new sensing functionality without resynthesizing the framework from scratch.^12,13

In this work, sulfisoxazole was selected as the chromogenic unit for PSM of NH₂-UiO-66(Zr). Sulfisoxazole carries a sulfonamide group and an isoxazole ring bearing multiple electronegative heteroatoms, both of which are plausible hydrogen-bond donor/acceptor sites for an oxyanion such as phosphate, and its extended conjugation provides the chromophoric character needed for a spectrophotometric response. Direct condensation of a primary amine with sulfisoxazole is synthetically awkward because the molecule lacks an aldehyde or ketone handle of its own, so glutaraldehyde was employed as a bifunctional spacer: one aldehyde terminus condenses with the framework amine to form an imine (Schiff base) linkage, while the second aldehyde remains available to react with a nucleophilic site on sulfisoxazole, effectively bridging the two components. This two-step grafting route offers practical advantages over direct imine coupling of a pre-functionalized aldehyde-bearing chromophore, which would require multi-step chromophore synthesis, and over click chemistry or amide coupling, which typically demand additional activating reagents and longer reaction times. The five-carbon glutaraldehyde bridge also introduces a degree of conformational freedom between the rigid MOF backbone and the sulfisoxazole recognition unit, which should assist the chromophore in adopting a binding geometry favorable for phosphate coordination without requiring the framework itself to distort.

Most MOF-based phosphate sensors reported to date operate through fluorescence quenching or turn-on mechanisms that require rare-earth doping or multi-step ligand exchange to achieve selectivity, and relatively few report a naked-eye or absorbance-based readout suitable for low-cost instrumentation.^14,15 Fewer still combine experimental sensing data with quantum-chemical validation of the binding event, and even fewer are demonstrated to retain the captured analyte, functioning simultaneously as a detection and a removal platform rather than a purely analytical tool. The present work addresses these gaps through four specific contributions: first, a two-step PSM route that installs a sulfonamide/isoxazole chromophore onto NH₂-UiO-66(Zr) via a flexible glutaraldehyde tether; second, a spectrophotometric PO₄³⁻ sensor operating at a single, mild optimum pH with a detection limit in the low-ppb-to-ppm range and a sub-minute response; third, density functional theory calculations, including frontier molecular orbital and electrostatic potential mapping, that trace the origin of the optical response to a measurable narrowing of the HOMO–LUMO gap upon complexation; and fourth, validation in real tap water and boiler water matrices with recoveries statistically consistent with ion chromatography, alongside evidence that the sensor doubles as a phosphate scavenger. The objective of this study is therefore to design, synthesize, and fully characterize SFX-UiO-66 as a dual-function material, to establish its analytical figures of merit and selectivity profile for PO₄³⁻ in aqueous media, and to rationalize its sensing mechanism through a combination of spectroscopic and computational evidence.

The novelty of this study is the first report of a multifunctional sulfisoxazole-functionalized NH₂-UiO-66(Zr) metal–organic framework (SFX-UiO-66) that can rapidly sense phosphate spectrophotometrically and simultaneously remove phosphate, all in one material. The suggested platform, in contrast to the traditional MOF-based phosphate sensors that merely focus on detection or adsorption, integrates both functionalities, but with high structural stability and rapid analytical response. A novel glutaraldehyde-assisted post-synthetic modification strategy was developed to covalently immobilize sulfisoxazole onto NH₂-UiO-66(Zr) to generate numerous sulfonamide and imine recognition sites which cooperatively engaged with phosphate through electrostatic attraction, hydrogen bonding, ligand exchange and strong inner-sphere Zr–O–P coordination. Also, this work is one of the rare studies combining two-dimensional correlation FT-IR (2D-COS) spectroscopy with density functional theory (DFT) calculations to understand the sequential adsorption mechanism and electronic interactions for phosphate recognition. Furthermore, the adsorption process was systematically optimized by the Response Surface Methodology (RSM) based on the Box–Behnken design and accurately predicted by an Artificial Neural Network (ANN) model, providing a comprehensive combination of experimental optimization and machine-learning-based prediction rarely reported for phosphate sensing materials. The combination of superior sensing performance, high adsorption capacity, fast response time, superior selectivity, successful application for real water samples, and computational validation, makes SFX-UiO-66 a new generation dual-functional MOF platform for efficient environmental monitoring and phosphate remediation.

2. Material and methods

2.1. Materials and instruments

All chemical and instruments were illustrated with details on SI Tables S1 and S2.

2.2. Synthesis of adsorbent

2.2.1. Synthesis of NH₂-UiO-66

Amine-modified zirconium-based MOF (NH₂-UiO-66) was made using a slightly different approach to solvothermal synthesis. Usually, zirconium(iv) chloride (ZrCl₄, 0.466 g) and 2-aminoterephthalic acid (NH₂-BDC, 0.332 g) get mixed with N,N-dimethylformamide (DMF, 60 mL) while stirring hard. To improve the crystal quality and control how the crystals grow, glacial acetic acid (6 mL) was added into the mix as a modulator. This clear solution goes into a 100 mL Teflon-lined stainless-steel autoclave where it will sit at 120 °C for 24 hours under solvothermal treatment Fig. 1. After it cools down to room temperature, this yellow solid is spun out by centrifuge (10 000 rpm for 10 min) and washed many times using DMF in order to get rid of any unreacted starting materials and also any trapped NH₂-BDC inside the framework. To guarantee that all high-boiling-point DMF has been removed from the pores, an exchange process with anhydrous ethanol took place over three days; every twelve hours new solvent was added. The powder of NH₂-UiO-66(Zr) was finally dried under vacuum at 80 °C for twelve hours. The structural integrity and phase purity of the as-synthesized MOF were confirmed via Powder X-ray Diffraction (PXRD) prior to further functionalization.¹⁶

Fig. 1. Schematic illustration of glutaraldehyde-bridged post-synthetic functionalization of NH₂-UiO-66(Zr) with sulfisoxazole to form the SFX-UiO-66 sensor.

2.2.2. Synthesis of SFX-UiO-66 sensor

Functionalization proceeded through a two-step glutaraldehyde-bridged Schiff base grafting sequence. In the first step, activated NH₂-UiO-66(Zr) (0.30 g) was dispersed in 30 mL absolute ethanol by sonication (20 min) and treated dropwise with 25% aqueous glutaraldehyde (2.0 mL) under continuous stirring at room temperature for 6 h, allowing the framework amine to condense with one terminal aldehyde of the crosslinker. The intermediate, designated GA-UiO-66, was isolated by centrifugation and washed with ethanol until the supernatant was free of residual aldehyde odor.^16,17 In the second step, GA-UiO-66 was redispersed in 30 mL ethanol containing sulfisoxazole (0.278 g, 1.0 mmol) and refluxed at 70 °C for 8 h to allow condensation of the remaining aldehyde with the accessible nucleophilic site of sulfisoxazole. The resulting solid, SFX-UiO-66, was recovered by centrifugation, washed sequentially with ethanol and deionized water (3 × 15 mL each) to remove unbound sulfisoxazole, and dried under vacuum at 60 °C for 12 h. A synthetic scheme illustrating the two-step grafting sequence and the resulting imine linkage is presented as Fig. 1.

2.3. Removal and batch studies of the phosphate via SFX-UiO-66 sensor

The adsorption efficiency of the SFX-UiO-66 sensor was investigated in detail, with special emphasis on the effect of different factors. The variables were pH, dose, initial concentration, contact time and temperature variation for adsorption of nutrient (phosphate). Batch adsorption studies for phosphate were carried out at Table S3. A 25 mL sample was taken from a standard phosphate solution diluted to the desired concentration.¹⁸ The adsorption isotherm was run by changing initial concentration from 5 to 282 mg L⁻¹ with interaction time of 100 min at 25 °C, pH 5 and adsorbent mass of 0.02 g in 25 mL of test volume. Also, the effect of time was investigated at 260 mg L⁻¹ for interaction time between 5 and 100 min at constant temperature, volume, pH and amount of adsorbent. The effect of temperature was studied at 280 mg L⁻¹ and 100 min contact time.¹⁹ The temperature was changed from 20 to 45 °C; the volume, pH and amount of adsorbent were maintained constant for adsorption of phosphate (Table S3). The pH values were kept constant during the experiment by adjustment with 0.10 N hydrochloric acid and/or sodium hydroxide Fig. 2. The data were also used to study the effect of temperature on the thermodynamic characteristics of adsorption reactions and to investigate the isotherm and kinetic behaviors. UV-vis spectrophotometry measurement was performed for the concentration of phosphate in solutions before and after adsorption. Then the obtained adsorption capacity (q_e, mg g⁻¹) and percentage removal (% R) were calculated by using the same eqn (1) and (2) respectively. Constant stirring was maintained in each run to ensure homogeneous mixing.²⁰

1

2

Fig. 2. SFX-UiO-66: an efficient adsorbent for phosphate removal comprehensive evaluation of key factors, isotherm, kinetics, and thermodynamics.

2.4. Experimental design

Box–Behnken Design (BBD) is a very common and efficient way for a large analysis of potential responses. This method is based on the construction of an accurate model that can correctly describe the observed data. The interaction of adsorbent dosage, pH and contact time as independent variables must be studied to optimize the removal of phosphate. BBD-based adsorption tests, with the help of Design-Expert software (Version 13, Stat-Ease), are an important part of this approach.²¹ The generated data includes identifiers associated with each step and descriptions of different levels of materials under consideration. One major part of this process is to relate independent variables to dependent outcomes using the second-degree quadratic model as shown in the eqn (3).

Y = β₀ + ∑β_iX_i + ∑β_iiX_i² + Σβ_ijX_iX_j 3

Here, the dependent variable is phosphate, and the independent variables are X_i and X_j. The removal method is published on the basis of 17 well-organized fully synthesized experiments. The volumes of the study were 25 mL of phosphate solution with an adequate amount of SFX-UiO-66 sensor as reported in Table S4. A conventional method to find the optimal combinations of multiple factors to maximize the responses in experiments is response surface methodology (RSM). One of its most used designs is the Box–Behnken design.²² This design uses first- and second-order models on the basis of simulated experimental data from computational modeling. 3-Level fractional factorial designs are used. An example of this is maximization of adsorption capacity when considering four parameters including time, dosage, pH etc. Therefore, three levels were defined as zero, positive and negative.

2.5. Computational details

All quantum-chemical calculations were performed with Gaussian 09 using density functional theory (DFT) at the B3LYP level. Light atoms (C, H, N, O, S) were described with the 6-31G(d) basis set, while the LANL2DZ effective core potential and basis set were applied to zirconium to reduce computational cost while retaining an adequate description of the metal center. Because full periodic treatment of the extended UiO-66 framework was computationally prohibitive, a truncated cluster model representing the local sulfisoxazole-imine binding pocket was used in place of the full crystalline lattice; this simplification captures the electronic behavior of the immediate recognition site but does not reproduce long-range framework effects, and results should be interpreted with that limitation in mind. Geometries of the free SFX ligand fragment and its PO₄³⁻ complex were fully optimized in the gas phase without symmetry constraints, and stationary points were confirmed by the absence of imaginary vibrational frequencies. Natural bond orbital (NBO) analysis was used to obtain atomic charges, and frontier molecular orbital energies (HOMO, LUMO) were extracted from the optimized structures to evaluate the HOMO–LUMO energy gap (ΔE = E_LUMO − E_HOMO) before and after complexation. Molecular electrostatic potential (ESP) surfaces were mapped onto the total electron density isosurface to visualize electrophilic and nucleophilic regions. All molecular orbital and ESP visualizations were rendered using GaussView.²³

2.6. General spectrophotometric sensing procedure

For each measurement, 5 mg of SFX-UiO-66 was dispersed in 20 mL of buffered aqueous solution at the target pH by sonication for 5 min to ensure homogeneous suspension. An aliquot of PO₄³⁻ standard was added to give the desired final concentration, and the suspension was allowed to equilibrate under gentle magnetic stirring for the pre-determined optimal contact time. The suspension was then briefly centrifuged, and the supernatant was transferred to a quartz cuvette for UV-vis absorbance recording relative to a sensor blank prepared identically but without analyte addition.²³

The effect of pH on the sensing response was evaluated across the range 2–10 using buffered PO₄³⁻ solutions at a fixed analyte concentration, with all other parameters held constant. Sensor dosage was varied between 2 and 8 mg per 20 mL to identify the amount giving maximum signal with minimum excess material. Contact time was investigated between 20 and 90 s to establish the point at which the absorbance signal plateaued. Temperature effects were examined over the range 20–45 °C to assess the thermal dependence of the binding equilibrium. In each case, the variable under study was scanned while the remaining parameters were fixed at provisional near-optimal values, and the optimized value was carried forward to subsequent experiments.

2.7. Real sample preparation and analysis

Tap water was collected locally, filtered through a 0.45 µm membrane to remove particulates, and used without further pretreatment other than pH adjustment to 6 with dilute HCl/NaOH as needed. Boiler water samples were collected from an operating industrial boiler system, filtered identically, and diluted to bring the expected phosphate concentration within the sensor's linear range. Recovery studies were carried out by the standard addition method known amounts of PO₄³⁻ standard were spiked into tap water aliquots to give nominal concentrations of 100 and 150 ppb, and the resulting solutions were analyzed by the SFX-UiO-66 sensing protocol described above. Boiler water samples were analyzed without spiking to determine native phosphate content. All real-sample results were cross-validated against ion chromatography (IC) performed on parallel, identically prepared sample aliquots.²³

3. Results and discussions

3.1. Characterization of SFX-UiO-66 sensor

3.1.1. X-ray diffraction (XRD)

Fig. 3(a) shows the XRD patterns of SFX-UiO-66 and PO₄@SFX-UiO-66. The SFX-UiO-66 sample shows a large number of sharp and well-defined diffraction peaks in the 2θ range of 5–70°, which confirms the extremely crystalline character of the UiO-66 framework following post-synthetic functionalization with sulfisoxazole. After phosphate adsorption, the PO₄@SFX-UiO-66 sample showed almost the same locations of the diffraction peaks over the whole diffraction range, suggesting the structural stability of the zirconium-based MOF framework during the adsorption. No extra diffraction peaks for crystalline phosphate molecules or zirconium phosphate phases were found, indicating that the phosphate uptake happened by surface complexation and coordination contacts rather than precipitation. However, after phosphate adsorption, small variations of the relative peak intensities and small peak broadening were observed, mainly at the intense reflections. These changes can be attributed to partial filling of the pore channels with phosphate ions and their coordination with the Zr–O clusters, imine (C N) and sulfonamide functional groups introduced during sulfisoxazole grafting. These interactions marginally weaken the long-range ordering of the framework without changing its crystal structure. The diffraction pattern before and after the phosphate adsorption is still characteristic, confirming the excellent crystallographic stability of SFX-UiO-66 and confirming the sensing/removal process by adsorption in the porous framework rather than structural degradation of the MOF.²⁴

Fig. 3. (a) XRD pattern (b) N₂ adsorption/desorption isotherm, (c) pore radius distribution, (d) pore size distribution, (e) FT-IR, (f) synchronous, and (g) asynchronous of SFX-UiO-66 and PO₄@ SFX-UiO-66.

3.1.2. N₂ adsorption/desorption isotherm

The nitrogen adsorption–desorption isotherm, pore radius and pore size distributions of SFX-UiO-66 and PO₄@SFX-UiO-66 are shown in Fig. 3(b–d). Both samples show type IV adsorption–desorption isotherms, which are typical for mesoporous materials according to the IUPAC classification, showing that the porous structure of the UiO-66 framework is maintained after phosphate adsorption. For any relative pressure (P/P₀), PO₄@SFX-UiO-66 presents a decrease of the adsorption capacity than pristine SFX-UiO-66, which indicates that the available pores are partly occupied by the phosphate ions. The BET surface area decreased from 738.04 to 688.74 m² g⁻¹ (approximate 6.7% reduction) and the cumulative pore volume from 6.14 to 5.38 cm³ g⁻¹, which further supports the successful incorporation of phosphate species in the porous framework rather than collapse of the MOF structure. In addition, the pore radius distribution (Fig. 3(c)) shifted to smaller values after adsorption, the average pore radius decreased from 2.44 to 2.08 nm, and the average particle (pore) size decreased from 2.16 to 1.85 nm. Likewise, the pore size distribution curves (Fig. 3(d)) showed similar profiles, with a slight decrease in the pore intensity after phosphate adsorption, indicating the mesoporous channels were still available, and a portion of the pore volume was filled with phosphate ions Table S5. The changes in pore distribution and the shape of the adsorption isotherm are subtle, indicating that the adsorption process is mainly based on the pore filling and surface coordination with the functional groups of SFX-UiO-66 and the Zr–O clusters, without any structural degradation of the framework. These results are in good accordance with the XRD data confirming the maintenance of the crystallinity, and FT-IR and EDX investigations confirming the incorporation of phosphates by chemical interactions rather than disintegration of the framework.²⁵

3.1.3. FT-IR

Fig. 3(e) shows the FT-IR spectra of SFX-UiO-66 and PO₄@SFX-UiO-66. Both spectra reveal the characteristic absorption bands of the functionalized UiO-66 framework, indicating the chemical stability of the MOF structure after phosphate adsorption. The broad absorption band in the region 3200–3500 cm⁻¹ is due to stretching vibrations of O–H and N–H groups arising from adsorbed water molecules, hydroxyl groups and the sulfonamide functionality of sulfisoxazole. The bands at 2920–2850 cm⁻¹ are attributed to the asymmetric and symmetric stretching vibrations of the aliphatic C–H of the glutaraldehyde linker. The band at ca. 1620–1640 cm⁻¹ is assigned to the C N stretching vibration of the Schiff-base linkage formed during post-synthetic functionalization together with the contributions of aromatic C C vibrations, confirming the effective grafting of sulfisoxazole onto the UiO-66 framework. The bands in the range 1550–1400 cm⁻¹ are characteristic and are assigned to the asymmetric and symmetric stretching vibrations of the coordinated carboxylate (–COO–) groups of the zirconium clusters, confirming the preservation of the UiO-66 coordination network. Also, the absorption region 1320–1150 cm⁻¹ is attributed to the asymmetric and symmetric stretching vibrations of O S O groups of sulfisoxazole and the peaks below 800 cm⁻¹ are due to Zr–O vibrations and deformation modes of the aromatic ring. The overall spectral profile changed little after phosphate adsorption, suggesting that the framework did not chemically break down. However, some characteristic bands of C N, N–H and S O functional groups broadened and showed little decrease in intensity, which indicated their direct participation in phosphate binding. In addition, the appearance and/or enhancement of absorption bands in the range 1000–1100 cm⁻¹ can be assigned to the asymmetric stretching vibration of P–O bonds, which is direct evidence of the effective incorporation of phosphate species into the composite. The small shifts seen to the C N and S O bands indicate that hydrogen bonding and Lewis acid–base interactions form between the phosphate oxygens and the imine nitrogen, sulfonamide groups and the exposed zirconium sites. The spectral changes, the decrease in BET surface area and pore volume, the retention of the crystallinity identified by XRD and the presence of phosphorus demonstrated by EDX analysis, all support the phosphate adsorption taking place via surface complexation, hydrogen bonding, coordination to Zr–O clusters and pore filling, while the structural integrity of the SFX-UiO-66 framework is maintained. Fig. 3(f) displays the synchronous two-dimensional correlation FT-IR (2D-COS) spectra of SFX-UiO-66 and PO₄@SFX-UiO-66. In the synchronous map, large positive auto-peaks appear along the diagonal, which represents the synchronous intensity changes of different functional groups during the phosphate adsorption. The most intense auto peaks were observed at 3300–3500 cm⁻¹, 1650–1550 cm⁻¹ and 1200–1000 cm⁻¹, which correspond to the stretching vibrations of O–H/N–H, C N and aromatic carboxylate groups and S O/P–O vibrations, respectively. The presence of strong positive cross-peaks between these sites suggests that these functional groups cooperate during the interaction with phosphate ions. Specifically, the high correlation between the N–H/O–H stretching region and C N vibration suggests that hydrogen bonding and imine coordination take place simultaneously in phosphate binding.²⁵ In the same sense, the good correlation between the C N and S O/P–O regions shows that the sulfonamide moiety and the Schiff-base linkage are the main recognition sites for phosphate ions. No competitive structural change or framework degradation upon adsorption is indicated by the lack of strong negative cross-peaks. Conversely, the synchronous spectrum supports the concerted interaction process, in which hydrogen bonding, Lewis acid–base coordination with zirconium centers and electrostatic attraction work cooperatively, while the UiO-66 framework stays structurally intact. These results are in excellent agreement with the XRD, BET, FT-IR and DFT data and demonstrate that phosphate adsorption is driven by multi-site interactions involving Zr–O clusters, imine (C N), sulfonamide (–SO₂NH–) and hydroxyl groups and not by framework collapse or precipitation.

The asynchronous 2D-COS FT-IR spectra of SFX-UiO-66 and PO₄@SFX-UiO-66 are shown in Fig. 3(g). The asynchronous map, in comparison with the asynchronous spectrum, resolves the sequential spectral fluctuations during phosphate adsorption, and identifies the sequence of the functional groups responding to the adsorption process. The presence of different positive and negative cross peaks in spectral regions 3300–3500 cm⁻¹, 1650–1550 cm⁻¹ and 1200–900 cm⁻¹ suggests that the adsorption is not a single simultaneous process, but is achieved through a number of sequential interactions. Asynchronous correlations in the O–H/N–H stretching region and the C N stretching band indicate the first response of hydroxyl and sulfonamide groups through creation of hydrogen bonds with phosphate oxygens and subsequent reorganization of the structure around the imine linkage. Also, the presence of significant asynchronous cross-peaks between the C N region and the S O/P–O vibration region indicates interaction of the phosphate ions with the sulfonamide/imine functionality after the initial hydrogen-bonding contact. The intense asynchronous features below 1200 cm⁻¹, which are the characteristic P–O stretching vibrations, show the growing incorporation of phosphate into the porous framework and the formation of stable coordination complexes with the zirconium clusters. The absence of random, irregularly distributed asynchronous signals suggests that these spectrum shifts are caused by well-defined chemical interactions and not by framework deterioration or structural collapse.

3.1.4. EDX analysis

The elemental composition of the as-synthesized SFX-UiO-66 was analyzed by energy dispersive X-ray spectroscopy (EDX). The corresponding spectrum is shown in Fig. 4(a). The EDX spectrum revealed unique distinctive peaks of C, N, O, S, and Zr, which confirmed the effective synthesis of the sulfisoxazole-functionalized UiO-66 framework. Strong zirconium peaks at ∼2.0 and 16–18 keV are assigned to the Zr6-oxo clusters that are the inorganic secondary building components of the UiO-66 framework, confirming that the zirconium-based MOF remains intact after post-synthetic alteration. The main components of the organic terephthalate linker and coordinated carboxylate groups were carbon (25.4%) and oxygen (42.64%), and the relatively high nitrogen content (14.21%) confirmed the presence of amino and imine functional groups introduced by the sulfisoxazole grafting. In addition, the identification of sulfur (0.96%) is a direct evidence for the incorporation of the sulfisoxazole molecules onto the MOF surface, since sulfur does not exist in pristine NH₂-UiO-66. The low amount of sulfur relative to carbon and oxygen is anticipated, because sulfur is only present in terminal sulfonamide groups of grafted ligand and not throughout the entire framework.²⁶ Importantly, no further elemental signals attributable to metallic or inorganic impurities were seen, indicating the synthesized material was pure and free of unwanted secondary phases. The elemental analysis is in good agreement with the predicted chemical structure and complements with the FT-IR data verifying the synthesis of Schiff-base, the XRD analysis showing the maintenance of crystallinity and the BET measurements revealing the successful functionalization without pore collapse. Thus, the EDX data strongly support the effective immobilization of sulfisoxazole to the UiO-66 framework without any loss of the structural integrity and chemical purity of the produced SFX-UiO-66 material. The EDX spectrum of PO₄@SFX-UiO-66 Fig. 4(b) verifies the effective adsorption of phosphate on the functionalized MOF framework, as evidenced by the distinctive signals of C, N, O, S, Zr, and P. The elemental composition was 25.64 wt% C, 14.70 wt% N, 39.88 wt% O, 0.86 wt% S, 16.84 wt% Zr, and 2.08 wt% P, and the emergence of the phosphorus peak was a direct proof for phosphate incorporation, which was not observed in the pristine SFX-UiO-66 sample. The distinctive zirconium peaks did not alter after adsorption, showing that the Zr₆-oxo secondary building blocks and the overall UiO-66 framework were structurally intact during adsorption. The elemental percentages showed slight variations in comparison with the pristine material, mainly a decrease in the oxygen and sulfur contents, indicating that phosphate ions occupied the available active sites by interacting with the Zr–O clusters, imine (C N) linkages and sulfonamide (–SO₂NH–) groups of the grafted sulfisoxazole molecules. Furthermore, the lack of other elemental indications indicates that there were no impurity phases or zirconium phosphate precipitates generated, suggesting that phosphate removal happened through surface complexation and pore constriction, rather than precipitation. The EDX results agree well with the XRD, BET, FT-IR and 2D-COS results, which collectively indicate the synergistic mechanism of phosphate adsorption via Lewis acid–base coordination, hydrogen bonding, electrostatic interactions and pore filling with the retention of the crystallinity and chemical stability of the SFX-UiO-66 framework.

Fig. 4. EDX analysis of (a) SFX-UiO-66, (b) PO₄@ SFX-UiO-66.

3.1.5. SEM analysis

Fig. 5(a–d) show the surface morphology and elemental distribution of SFX-UiO-66 studied by SEM with EDX elemental mapping. According to the SEM micrograph, the synthesized SFX-UiO-66 consists of uniformly distributed and densely packed tiny particles with rough and porous surface morphologies that offer many exposed active sites for phosphate adsorption. No substantial particle aggregation or structural collapse was detected, showing that the post-synthetic functionalization with sulfisoxazole maintained the structural integrity of the UiO-66 framework. The related elemental mapping images show the homogenous distribution of C, N, O, Zr, and S throughout the entire material without any localized enrichment, demonstrating the effective immobilization of sulfisoxazole on the MOF surface. The elemental composition determined by mapping analysis (ca. 26.1 wt% C, 14.0 wt% N, 41.0 wt% O, 17.8 wt% Zr, and 1.1 wt% S) accords with the EDX results and further confirms the effective functionalization of UiO-66 with sulfur-containing sulfisoxazole molecules. Moreover, the 3D surface topography images of grayscale, thermal-color, and blue-tone (Fig. 5(b–d)) show that the surface has relatively homogeneous features with uniformly distributed protrusions and valleys, suggesting uniform surface roughness and the absence of significant defects or phase segregation.²⁷

Fig. 5. (a) SEM mapping of SFX-UiO-66, (b) grayscale, (c) thermal color, (d) blue tones, and (e) SEM mapping of PO₄@SFX-UiO-66, (f) grayscale, (g) thermal color, (h) blue tones.

Fig. 5(e–h) presents the SEM images and EDX elemental mapping of PO₄@SFX-UiO-66 following phosphate adsorption, which were used for analysis of the surface shape and elemental distribution. The general morphology of the material is well preserved after phosphate adsorption, which is shown in the SEM image. The particles are tightly agglomerated and seem to have a rough and porous surface with no obvious collapse of structure or excessive agglomeration, which shows the excellent stability of the functionalized UiO-66 framework during the process of adsorption. The elemental mapping images indicate the consistent distribution of C, N, O, Zr, S and P throughout the composite, showing the homogeneous adsorption of phosphate ions over the entire surface instead of local accumulation. The elemental composition determined by mapping analysis, which includes approximately 25.8 wt% C, 13.7 wt% N, 40.84 wt% O, 16.94 wt% Zr, 0.92 wt% S and 1.8 wt% P, matches well with the EDX study and validates the effective incorporation of phosphate into the SFX-UiO-66 framework. The formation of phosphorus, together with the minor reduction of the sulfur content relative to the pristine material shows that phosphate ions occupied the active adsorption sites via interaction with the Zr–O clusters, imine (C N) links, and sulfonamide (–SO₂NH–) groups. Moreover, the three-dimensional surface topography images in grayscale, thermal-color, and blue-tone (Fig. 5(f–h)) show a homogeneous surface with only slight variations in the surface roughness after adsorption, suggesting partial occupation of the pores by the phosphate species without altering the overall porous architecture.²⁷

3.2. Batch experiments

3.2.1. Adsorption factors

3.2.1.1. Determination of pH_zpc

The point of zero charge (pH_ZPC) of SFX-UiO-66 was determined to be 6.6 (Fig. 6(a)), which indicates that the adsorbent surface is positively charged for pH values less than 6.6 and negatively charged at pH values greater than this point. Maximum phosphate adsorption was seen at pH 5, which is lower than pH_ZPC, indicating the very favorable circumstances for phosphate uptake on the positively charged surface. At this pH, phosphate is mostly in the monovalent H₂PO₄⁻ form, which can be easily attracted to the protonated amino groups of NH₂-UiO-66 and the protonated sulfonamide and heterocyclic nitrogen functionalities introduced by sulfisoxazole through strong electrostatic interactions. This initial attraction promotes the diffusion of phosphate ions towards the active adsorption sites, which then form stable inner-sphere complexes with the coordinatively unsaturated Zr(iv) centres through the formation of strong Zr–O–P bonds via ligand-exchange reactions involving surface hydroxyl groups and coordinated water molecules.²⁸ Furthermore, the hydrogen bonding between the phosphate oxygen atoms and the numerous amino (–NH₂) and sulfonamide (–SO₂NH–) groups further stabilize the adsorbed phosphate species, while the polar sulfonyl (S O) groups improve dipole–ion interactions, enhancing the overall adsorption affinity. In contrast, the surface of SFX-UiO-66 is negatively charged at pH values above 6.6, leading to electrostatic repulsion with phosphate anions, stronger competition of hydroxide ions for zirconium active sites, and reduced accessibility of adsorption sites, thereby causing a substantial reduction in phosphate removal efficiency. Therefore, the pHZPC value of 6.6 is a good indicator of the maximum adsorption at pH 5, where the combined effects of appropriate phosphate speciation, positive surface charge, electrostatic attraction, Zr–O–P complexation, ligand exchange, H-bonding and dipole–ion interactions contribute to the highest phosphate adsorption efficiency.

Fig. 6. (a) Determination of pHzpc, (b) effect of pH, (c) effect of adsorbent dose, (d) effect of initial concentration, (e) effect of contact time, and (f) effect of temperature of adsorption of PO₄ onto SFX-UiO-66 sensor.

3.2.1.2. Effect of pH

The influence of solution pH on phosphate adsorption by SFX-UiO-66 was examined in the pH range of 2–10 and the findings are shown in Fig. 6(b). The adsorption capacity increased markedly from 98.5 mg g⁻¹ at pH 2 to a maximum value of 178 mg g⁻¹ at pH 5, followed by a gradual decline to 121 mg g⁻¹ at pH 10. The lower adsorption capacity observed at highly acidic conditions (pH 2–3) can be attributed to the excessive protonation of phosphate species and the strong competition of abundant H⁺ ions with phosphate for the available adsorption sites, which suppresses phosphate uptake despite the positively charged adsorbent surface. As the pH increased to 5, the adsorption capacity achieved its optimum since this pH is below the pH_ZPC (6.6) of SFX-UiO-66, resulting in a positively charged surface due to protonation of the amino groups of NH₂-UiO-66 and the sulfonamide functions of sulfisoxazole. At the same time, phosphate is mainly present as H₂PO₄⁻, which has a significant affinity to the positively charged surface through electrostatic interactions. Such initial attraction is followed by formation of stable inner-sphere Zr–O–P complexes (ligand-exchange interactions with surface hydroxyl groups and coordinated water molecules) with the coordinatively unsaturated Zr(iv) centers.²⁸ Hydrogen bonding between phosphate oxygen atoms and amino (–NH₂) and sulfonamide (–SO₂NH–) groups also gives further stability. Dipole–ion interactions are also given by the extremely polar sulfonyl (S O) groups of sulfisoxazole. Above pH 6.6, the adsorption capacity declined gradually due of the rising negative charge on the surface of SFX-UiO-66, which caused an electrostatic repulsion effect on phosphate anions (mostly HPO₄²⁻ and, at higher pH, PO₄³⁻). In addition, hydroxide ions competed more and more with phosphate ions for zirconium coordination sites, hence diminishing the production of Zr–O–P complexes and the number of accessible active sites. The increased negative charge and the bigger hydration shell of HPO₄²⁻ further limited its diffusion and access to the porous structure. Thus, the highest adsorption at pH 5 indicates that the phosphate removal by SFX-UiO-66 is controlled by the synergistic effects of favorable phosphate speciation, positive surface charge, electrostatic attraction, zirconium coordination, ligand exchange, hydrogen bonding, and dipole–ion interactions, resulting in the maximum adsorption efficiency under mildly acidic conditions.

3.2.1.3. Effect of adsorbent dose

The effect of adsorbent dosage on the phosphate adsorption performance of SFX-UiO-66 was studied by changing the mass of adsorbent from 0.02 to 0.50 g at constant starting phosphate concentration, solution volume, pH, contact time and temperature, as shown in Fig. 6(c). The results showed that the phosphate removal efficiency increased dramatically from 56.0% for 0.02 g adsorbent to 97.0% for 0.50 g adsorbent, and the adsorption capacity (q_e) fell significantly from 140.0 mg g⁻¹ to 9.7 mg g⁻¹. The enhanced removal efficiency is attributed to the gradual increase of available adsorption sites, coordinatively unsaturated Zr(iv) centers, amino (–NH₂) groups and sulfisoxazole functional groups (–SO₂NH– and heterocyclic nitrogen atoms) which offer more active sites for phosphate adsorption. The larger dosage of adsorbent also provides a larger external surface area and pore volume, which is favorable for the diffusion of phosphate into the porous framework and enhances the opportunity of interactions through electrostatic attraction, Zr–O–P inner-sphere complexation, ligand exchange, hydrogen bonding and dipole–ion interactions. However, the constant decline of adsorption capacity with the increase of dosage is mainly due to the fixed number of phosphate ions in solution leading to unbalanced usage of adsorption sites. At higher adsorbent doses, a large fraction of active sites remains unoccupied because phosphate ions are not available to saturate all adsorption centers. Also, at high adsorbent concentrations, particle aggregation can partially block the surface area and overlap some adsorption sites, leading to a decrease in adsorption capacity per unit weight. Therefore, the highest phosphate removal efficiency (97%) was achieved with 0.50 g, but the optimum adsorbent dosage for maximizing adsorption capacity was 0.02 g, giving the highest q_e value (140 mg g⁻¹) due to the more efficient utilization of the available active adsorption sites. The results demonstrated a normal reverse correlation between adsorption capacity and adsorbent dose, and it was confirmed that the phosphate removal by SFX-UiO-66 was governed by the balance between the availability of active adsorption sites and the adsorbate-to-adsorbent ratio.²⁸

3.2.1.4. Effect of initial concentration

The effect of initial phosphate concentration (5–282 mg L⁻¹) on the adsorption performance of SFX-UiO-66 was studied with constant adsorbent dosage (0.02 g), solution pH (5), contact time (100 min), temperature (25 °C) and solution volume (25 mL) as shown in Fig. 6(d). The adsorption capacity (q_e) gradually increased from 5.68 mg g⁻¹ at an initial concentration of 5 mg L⁻¹ to a maximum value of 202.83 mg g⁻¹ at 282 mg L⁻¹, while the phosphate removal efficiency fell gradually from 96.40% to 57.47% in the same concentration range. The increased adsorption capacity is due to the larger concentration gradient between the bulk solution and the adsorbent surface as the principal driving factor for mass transfer which enhances the phosphate diffusion towards the active adsorption sites. As the phosphate concentration increases, more phosphate ions clash with the external surface and infiltrate into the porous structure of SFX-UiO-66, which leads to higher usage of the numerous zirconium coordination sites, amino groups and sulfisoxazole functional groups. The adsorption occurs through a combination of electrostatic attraction, inner-sphere Zr–O–P complexation, ligand-exchange reactions, hydrogen bonding, dipole–ion interactions, and pore filling, which causes a significant increase in the amount of phosphate adsorbed per unit mass of adsorbent. On the other hand, the slow reduction in removal efficiency with increased phosphate concentration is mostly due to the slow saturation of the limited number of active adsorption sites on SFX-UiO-66. The occupation of a large part of these sites results in an increased residual phosphate content in the solution and a decreased removal percentage. Moreover, the rivalry among phosphate ions for the available adsorption sites increases, and the diffusion into the pores that become increasingly occupied slows down, which leads to a decrease in the removal effectiveness at high concentrations. Although it is a decline, the continuous increase in adsorption capacity indicates the very high affinity and adsorption potential of SFX-UiO-66 for phosphate, proving that the porous sulfisoxazole-functionalized UiO-66 framework has many available adsorption sites that can hold a large amount of phosphate before the system reaches equilibrium saturation.²⁹

3.2.1.5. Effect of contact time

Fig. 6(e) presents the influence of contact time on the adsorption of phosphate onto SFX-UiO-66 for 5–100 min under the ideal circumstances of initial phosphate concentration of 280 mg L⁻¹, adsorbent dose of 0.02 g, pH 5, 25 °C and solution volume of 25 mL. The adsorption capacity (q_t) increased rapidly in the first stage of the adsorption process, from 24.91 mg g⁻¹ at 5 min to 119.70 mg g⁻¹ at 30 min, and then increased slowly until equilibrium was reached at about 100 min, when the adsorption capacity reached 203.80 mg g⁻¹. The rapid adsorption in the first 30–40 min is attributed to the presence of abundant vacant active adsorption sites on the external surface of SFX-UiO-66, such as coordinatively unsaturated Zr(iv) centers, protonated amino groups and sulfisoxazole functional groups, which can easily interact with phosphate ions through electrostatic attraction, hydrogen bonding, ligand exchange and inner-sphere Zr–O–P complexation. At this stage, a significant concentration gradient between the bulk solution and adsorbent surface provides a strong driving force for mass transfer, such that phosphate ions can rapidly diffuse from the solution to the external adsorption sites. The adsorption rate decreases with the lowering of the concentration gradient and increase of the occupied active sites during the adsorption. In the later stage, phosphate molecules diffuse into the internal microporous and mesoporous structure of SFX-UiO-66, and the adsorption proceeds by intraparticle diffusion and pore filling until the adsorption equilibrium is reached. The slow approach to equilibrium is typical of the incremental saturation of the available adsorption sites and the rising diffusion resistance within the porous framework. The equilibrium time of 100 min indicates that SFX-UiO-66 has rapid adsorption kinetics with a high adsorption capacity. This indicates the excellent accessibility of its porous structure and the strong affinity of its zirconium clusters and sulfisoxazole functional groups towards phosphate ions. The two-stage adsorption behavior including an initial rapid surface adsorption and subsequent slower intraparticle diffusion and pore-filling indicates that the overall adsorption process is affected by both surface reaction and internal diffusion, which is in accordance with the excellent adsorption performance of the SFX-UiO-66 adsorbent.²⁹

3.2.1.6. Effect of temperature

The effect of temperature on the adsorption of phosphate by SFX-UiO-66 was investigated over the range of 20–45 °C under the optimum conditions of 280 mg L⁻¹ initial phosphate concentration, 0.02 g adsorbent dosage, pH 5, 100 min contact time, and 25 mL solution volume, as illustrated in Fig. 6(f). The phosphate removal efficiency increased progressively from 46.6% at 20 °C to 97.5% at 45 °C, indicating that elevated temperatures significantly enhance the adsorption process. This behavior suggests that phosphate adsorption onto SFX-UiO-66 is predominantly endothermic, where increasing temperature provides additional thermal energy to overcome the activation energy barrier for adsorption and accelerates the diffusion of phosphate ions from the bulk solution to the adsorbent surface. Higher temperatures also increase the mobility of phosphate ions and reduce the thickness of the boundary layer surrounding the adsorbent particles, thereby improving external mass transfer and facilitating faster intraparticle diffusion into the porous UiO-66 framework. Moreover, the increased thermal energy provides a higher accessibility of the coordinatively unsaturated Zr(iv) active sites and functional groups (amino, –NH₂, sulfonamide, –SO₂NH–, and sulfonyl, S O) introduced by sulfisoxazole, which further accelerates the electrostatic attraction, ligand-exchange reactions, and the formation of stable Zr–O–P inner-sphere complexes. The better diffusion into the interconnected pore network also improves the pores utilization and phosphate loading of the adsorbent. The continuous enhancement in phosphate removal with increasing temperature indicates that the adsorption process is not diffusion-limited under the investigated conditions but is thermodynamically favored at higher temperatures. These results are compatible with an endothermic and spontaneous adsorption mechanism, in which the increase in temperature increases the affinity of SFX-UiO-66 for phosphate ions and resulting in better adsorption performance. Such behavior is typically associated with a positive enthalpy change (ΔH° > 0), while the thermodynamic parameters obtained from the Van't Hoff analysis are expected to confirm the spontaneous nature of adsorption through negative Gibbs free energy (ΔG° < 0) and increased randomness at the solid–liquid interface reflected by a positive entropy change (ΔS° > 0).³⁰

3.2.2. Adsorption isotherm

The equilibrium adsorption data of phosphate onto SFX-UiO-66 fitted best with the Langmuir isotherm model, indicating monolayer adsorption on a homogeneous surface Tables S6 and S7. Each active site accommodates one phosphate ion with minimal interactions between adsorbed species. The maximum adsorption capacity (q_m,_exp) was 202.8 mg g⁻¹, reflecting the excellent ability of SFX-UiO-66 to adsorb phosphate ions. This capacity arises from the high specific surface area, porous structure, coordinatively unsaturated Zr(iv) centers, and functionalities like amino (–NH₂) groups and sulfonamide (–SO₂NH–) from sulfisoxazole. Active sites enhance phosphate adsorption through mechanisms including electrostatic attraction, ligand exchange, hydrogen bonding, and stable Zr–O–P coordination complexes, improving adsorption affinity Table S8. The Langmuir affinity constant (K_L = 0.095 L mg⁻¹) indicates a good interaction between phosphate and the adsorbent. The dimensionless separation factor (R_L = 0.08) suggests favorable phosphate adsorption on SFX-UiO-66, highlighting its effectiveness even at high concentrations. Overall, the high capacity and selectivity for phosphate ions stem from the uniform distribution of active sites and multifunctional surface chemistry Fig. 7(a).³¹

Fig. 7. (a) Adsorption isotherm, (b) adsorption kinetic model, and (c) IPD model of PO₄ onto SFX-UiO-66 sensor.

To investigate phosphate adsorption on SFX-UiO-66, the Freundlich isotherm model was used to assess surface heterogeneity. A high Freundlich constant (K_F = 66.66) indicates a strong affinity for phosphate ions and significant adsorption capacity. The intensity parameter (n = 4.06) exceeds one, suggesting highly favorable adsorption efficiency within the tested concentration range. Values of n between 2 and 10 reflect favorable interactions and effective contact between adsorbate and adsorbent. The high n value implies multiple interaction mechanisms, including electrostatic attraction, hydrogen bonding, ligand exchange, and the creation of stable Zr–O–P coordination complexes with phosphate ions. Additionally, the Freundlich model's suitability suggests an energetically heterogeneous surface with various active sites, including Zr(iv) clusters and amino and sulfonamide groups due to sulfisoxazole functionalization. Varied adsorption energies and pathways enhance overall efficacy. Though the Langmuir model fits better for monolayer adsorption, the Freundlich constants indicate that surface heterogeneity contributes significantly to phosphate affinity and adsorption capacity.³²

The Dubinin–Radushkevich (D–R) isotherm model was used to study phosphate adsorption on SFX-UiO-66, distinguishing between physical and chemical processes. The experimental adsorption capacity matched the theoretical D–R capacity (Q_DR = 180.64 mg g⁻¹), indicating excellent phosphate ion adsorption ability. The mean free adsorption energy (E_a = 45.88 kJ mol⁻¹) calculated from the D–R constant (K_DR = 6.26 × 10⁻⁶ mol² kJ⁻²) suggested that chemisorption dominates, as E_a >16 kJ mol⁻¹ indicates strong chemical bonding. This supports the proposed adsorption mechanism involving stable inner-sphere Zr–O–P coordination complexes between phosphate and Zr(iv) centers, along with ligand-exchange reactions displacing hydroxyls and water. Additional factors enhancing adsorption include hydrogen bonding with amino and sulfonamide groups, electrostatic attraction at pH < pH_ZPC, and dipole–ion interactions from sulfisoxazole functional groups. Therefore, the D–R model confirms that phosphate adsorption on SFX-UiO-66 is primarily driven by strong chemical interactions, highlighting the adsorbent's high affinity, stability, and effective phosphate removal.³³

The Temkin isotherm model was applied to study phosphate adsorption on SFX-UiO-66 and the adsorbate–adsorbent interaction's impact on heat of adsorption. The constants were b_T = 77.12 J mol⁻¹ and K_T = 3.93 L mol⁻¹, indicating a strong binding affinity of phosphate ions to SFX-UiO-66's active sites, showing favorable adsorption equilibrium. The model assumes heat of adsorption varies linearly with surface coverage from phosphate interactions, reflecting SFX-UiO-66's multifunctional surface chemistry. Adsorption occurs at the most active sites like Zr(iv) centers and amino groups, with lower-energy sites showing decreased heat of adsorption. The Temkin model confirms strong adsorbate–adsorbent interactions during phosphate uptake via electrostatic attraction and other mechanisms. Results suggest phosphate adsorption is not constant but decreases with increasing surface coverage, demonstrating the heterogeneous nature of the functionalized adsorbent and strong affinity toward phosphate ions.³⁴

The Jossens isotherm model examined phosphate adsorption onto SFX-UiO-66, revealing constants K = 83.14 and J = 0.86. The strong K value indicates robust interactions between phosphate ions and SFX-UiO-66's active sites, showcasing its effective adsorption capabilities. The J value, below one, signifies a favorable adsorption process and moderate energetic heterogeneity on the adsorbent surface, characterized by numerous sites with varied binding energies. This behavior reflects SFX-UiO-66's structural features, including coordinatively unsaturated Zr(iv) clusters and various functionalities from sulfisoxazole, which enable diverse adsorption mechanisms like electrostatic attraction, ligand exchange, and hydrogen bonding. The Jossens model indicates that adsorption energy is not uniformly distributed but varies with phosphate occupancy across active sites, suggesting the existence of both high-energy and lower-energy adsorption centers. As a result, the Jossens parameters affirm that phosphate adsorption onto SFX-UiO-66 is highly favorable, driven by strong chemical interactions on a multifunctional, energetically heterogeneous surface, contributing to its high adsorption capacity and efficiency in phosphate removal.³⁴

The Tóth isotherm model elucidated the equilibrium adsorption of phosphate onto SFX-UiO-66, accounting for surface heterogeneity. The parameters were q_m = 319.42 mg g⁻¹, K_T = 0.84 L mg⁻¹, and t = 0.33. The high theoretical adsorption capacity indicates the significant potential of SFX-UiO-66 for phosphate ions and highlights the accessible active sites in the framework. The Tóth affinity constant (K_T) suggests strong interactions between phosphate ions and the adsorbent, while the heterogeneity parameter confirms that the SFX-UiO-66 surface is not homogeneous. This heterogeneity arises from coordinatively unsaturated Zr(iv) centers and the functional groups introduced by sulfisoxazole, which provide varied adsorption sites. Phosphate adsorption involves electrostatic attraction, ligand exchange, hydrogen bonding, pore filling, and stable Zr–O–P coordination complexes. The Tóth model effectively describes the occupancy of high-energy adsorption sites at low phosphate concentrations and lower-energy sites at higher coverages, showing SFX-UiO-66's multifunctional and energetically heterogeneous adsorption surface.³⁴

3.2.3. Adsorption kinetic

The pseudo-first-order (PFO) kinetic model was used to evaluate phosphate adsorption on SFX-UiO-66. The low-rate constant (K₁ = 0.024 × 10⁻² min⁻¹) suggests that the PFO model is not a good fit for the experimental data. This model assumes the adsorption rate relates to unoccupied sites and typically corresponds to reversible physisorption, driven by diffusion and weak interactions. However, phosphate adsorption on SFX-UiO-66 is much stronger Fig. 7(b).³⁵ It begins with electrostatic attraction between the positively charged surface at pH 5 and negatively charged phosphate, followed by stable Zr–O–P coordination complexes formed through ligand-exchange at Zr(iv) centers. Stabilization is also enhanced by hydrogen bonding and dipole–ion interactions. The small K₁ value indicates physical adsorption is not the rate-controlling mechanism. D–R isotherm analysis revealed a mean adsorption energy of 45.88 kJ mol⁻¹, confirming that phosphate adsorption is mainly governed by chemisorption. Overall, the pseudo-second-order model better describes the adsorption kinetics, highlighting the significance of chemical interactions in phosphate adsorption Table S9.

The adsorption kinetics of phosphate onto SFX-UiO-66 was studied using the pseudo-second-order (PSO) kinetic model, which effectively describes the process as primarily controlled by chemisorption. The rate constant (K₂ = 6.35 × 10⁻⁵ g mg⁻¹ min⁻¹) and equilibrium capacity (q_e,cal = 203.64 mg g⁻¹) closely matched experimental values (q_e,exp = 203.80 mg g⁻¹), validating the PSO model. This correlation indicates that the adsorption rate is influenced by the chemical activity of available sites rather than external mass transfer. Phosphate ions are initially attracted to the positively charged surface of SFX-UiO-66 (pH 5 < pH_ZPC), forming stable Zr–O–P coordination complexes through ligand-exchange reactions. Additionally, hydrogen bonding and dipole–ion interactions further stabilize phosphate adsorption within the porous UiO-66 framework. The high adsorption energy from the Dubinin–Radushkevich model (E = 45.88 kJ mol⁻¹) signifies that phosphate adsorption is driven by strong chemical interactions. Consequently, chemisorption involving electron exchange at active sites is the key factor for the adsorbent's high capacity and rapid performance, confirmed by the PSO model.³⁶

The intraparticle diffusion (Weber–Morris) model assesses whether phosphate adsorption in SFX-UiO-66 is limited by diffusion within its pores. The intraparticle diffusion rate constant (K_i = 21.83 mg g⁻¹ min^−1/2) suggests fast diffusion due to the porous framework's accessibility. The significant intercept (X = 15.64 mg g⁻¹) indicates that intraparticle diffusion is one of several processes influencing phosphate adsorption. This positive intercept signals a notable boundary-layer effect, with external film diffusion and surface adsorption also contributing to the overall adsorption rate. Phosphate ions quickly transfer from the solution to the SFX-UiO-66 surface by film diffusion and are attracted to the positively charged surface (at pH 5 < pH_ZPC = 6.6) via electrostatic forces. They then diffuse through pore channels, where adsorption involves ligand exchange with Zr–OH groups, forming Zr–O–P complexes, hydrogen bonding, and dipole–ion interactions. As adsorption progresses and high-energy sites fill, diffusion resistance increases, reducing the adsorption rate until equilibrium is reached. The analysis reveals that phosphate adsorption is a multi-step process influenced by chemical interactions, consistent with the pseudo-second-order kinetic model and high adsorption energy from the Dubinin–Radushkevich isotherm.³⁷

To study phosphate adsorption kinetics on SFX-UiO-66, the Elovich kinetic model was used to assess chemisorption effects on a heterogeneous surface. The Elovich parameters were calculated as α = 0.11 mg g⁻¹ min⁻¹ and β = 49.13 g mg⁻¹. The initial adsorption rate constant (α) reflects rapid phosphate sorption due to readily accessible active sites like Zr(iv) centers and amino (–NH₂) groups. Electrostatic attraction and ligand exchange facilitate quick phosphate adsorption. The high desorption constant (β = 49.13 g mg⁻¹) indicates a slow decrease in adsorption rate with surface coverage, owing to the occupation of favorable sites and increased activation energy for further adsorption. This behavior suggests varying adsorption energies on the surface. As phosphate adsorption progresses, intraparticle diffusion and pore filling become more significant, with hydrogen bonding and dipole–ion interactions stabilizing adsorbed phosphate. The Elovich model highlights the dominance of chemisorption on SFX-UiO-66, corroborated by the pseudo-second-order model, high adsorption energy from the Dubinin–Radushkevich isotherm (E = 45.88 kJ mol⁻¹), and mechanisms like electrostatic attraction and the formation of stable Zr–O–P complexes.³⁸

3.2.3.1. Diffusion mechanism

The Weber–Morris intraparticle diffusion model was used to analyze phosphate adsorption on SFX-UiO-66. The multilinear plot indicated three diffusion stages, meaning intraparticle diffusion isn't the only rate-control mechanism. The highest diffusion rate constant (K_i1 = 37.35 mg g⁻¹ min^−1/2, R² = 0.999) occurred in Stage I, reflecting rapid transport of phosphate ions from the bulk solution to the external surface of SFX-UiO-66 via film diffusion due to a high concentration gradient and available active sites like protonated amino groups and coordinatively unsaturated Zr(iv) centers Fig. 7(c). In Stage II, the diffusion rate dropped to K_i2 = 21.82 mg g⁻¹ min^−1/2 (R² = 0.996), indicating that phosphate ions diffused into the internal microporous structure, with adsorption also influenced by intraparticle diffusion and stable Zr–O–P complexes, ligand-exchange with Zr–OH groups, hydrogen bonding, and dipole–ion interactions. As adsorption sites filled, diffusion resistance increased, slowing the adsorption rate. In Stage III, the lowest diffusion rate constant (K_i3 = 5.80 mg g⁻¹ min^−1/2, R² = 0.982) indicated equilibrium, with occupied high-energy sites leading to slow diffusion into smaller pores. This stage showed a significant diffusion rate drop, suggesting pore saturation and limited adsorption sites. Overall, the multistep adsorption mechanism indicated control by external film diffusion, intraparticle diffusion, pore filling, and surface chemisorption rather than solely intraparticle diffusion, supported by good pseudo-second-order fitting and high Dubinin–Radushkevich energy (E = 45.88 kJ mol⁻¹) Table S10.³⁷

3.2.4. Adsorption thermodynamics

The thermodynamic Van't Hoff and Arrhenius models were used to investigate the phosphorus adsorption on SFX-UiO-66. The results of the models showed that the adsorption is a spontaneous, endothermic and chemically initiated process. The Van't Hoff plot showed a good linearity (Fig. 8(a)) and the thermodynamic parameters were obtained. The Gibbs free energy change (ΔG°) was negative for all temperatures studied, ranging from −0.68 kJ mol⁻¹ at 293 K to −9.04 kJ mol⁻¹ at 318 K, indicating that the phosphate adsorption is spontaneous and more favorable at higher temperatures. This result is consistent with the temperature study (experimental) where the phosphate removal efficiency increased from 46.6% (20 °C) to 97.5% (45 °C) showing that higher temperatures improved phosphate uptake. The positive value of ΔH° (97.32 kJ mol⁻¹) indicates that the adsorption process is endothermic, meaning that thermal energy promotes the diffusion of phosphate into the porous framework and improves the formation of strong chemical interactions between phosphate ions and the active adsorption sites of SFX-UiO-66. Moreover, the very high value of ΔH° (>40 kJ mol⁻¹) is characteristic of chemisorption, in agreement with the adsorption process proposed by the kinetic and isotherm studies. The positive value of entropy change (ΔS° = 334.47 J mol⁻¹ K⁻¹) indicates the increase in disorder at the solid–solution interface during adsorption which may be due to the displacement of the coordinated water molecules and hydroxyl groups from the zirconium clusters by phosphate ions via ligand-exchange reactions along with the release of the hydrated water molecules around phosphate into the bulk solution. The activation energy (E_a) calculated from the Arrhenius plot (Fig. 8(b)) was 42.3 kJ mol⁻¹ which shows that the adsorption of the phosphate onto SFX-UiO-66 needs to overcome an activation barrier and further confirms the fact that the adsorption process is mainly governed by chemical interactions rather than simple physical adsorption.³⁹ The calculated activation energy is consistent with the high value of the adsorption energy calculated from the Dubinin–Radushkevich isotherm (E = 45.88 kJ mol⁻¹) and the high quality of the fit of the pseudo-second-order kinetic model, which suggests that the adsorption mechanism is mainly controlled by the formation of stable inner-sphere Zr–O–P coordination complexes via ligand exchange at the Zr(iv) sites that are not involved in the coordination with the organic linker, combined with electrostatic attraction between the positively charged adsorbent surface (pH 5 < pH_ZPC = 6.6) and phosphate anions, hydrogen bonding involving amino and sulfonamide groups, dipole–ion interactions generated from the sulfisoxazole functionalities, and pore filling within the highly porous UiO-66 framework. The thermodynamic parameters collectively reveal that the phosphate adsorption on SFX-UiO-66 is a spontaneous, endothermic and chemically controlled process with the increasing feasibility at higher temperatures, which manifests the excellent affinity and practical applicability of the adsorbent for the phosphate removal from aqueous environments Table S11.

Fig. 8. (a) Van't Hoff, (b) Arrhenius model of adsorption of PO₄ onto SFX-UiO-66 sensor.

3.2.5. Mechanism of interaction

The great adsorption performance of phosphate adsorption on sulfisoxazole-functionalized NH₂-UiO-66(Zr) (SFX-UiO-66) can be ascribed to the simultaneous play of several complementary mechanisms. The pH-dependent adsorption study revealed that the maximum phosphate removal was achieved at pH 5, and the point of zero charge (pH_ZPC) of SFX-UiO-66 was 6.6. Since the solution pH is lower than pH_ZPC, the adsorbent surface is positively charged due to the protonation of amino (–NH₂) groups of NH₂-UiO-66 and the amino and heterocyclic nitrogen atoms of sulfisoxazole. Here, the phosphate is mainly present as H₂PO₄⁻ and the initial driving force for adsorption is the electrostatic attraction between the negatively charged phosphate and the positively charged adsorption sites.³⁹ Then, phosphate forms very stable inner-sphere complexes with the Zr₆-oxo clusters of the UiO-66 framework via direct coordination of the phosphate oxygen atoms to coordinatively unsaturated Zr(iv) centers, resulting in strong Zr–O–P bonds. As Zr(iv) as a hard Lewis acid has an extraordinary affinity with the hard Lewis base oxygen atoms of phosphate. Simultaneously, ligand-exchange reactions take place, where phosphate ions displace terminal Zr–OH groups and coordinated H₂O molecules. These ligand-exchange reactions further stabilize the adsorbed species through monodentate, bidentate, or even bridging phosphate coordination geometries, depending on the available zirconium sites. Hydrogen bonding is also important for phosphate immobilization. The amino groups (–NH₂) of NH₂-UiO-66, sulfonamide group (–SO₂NH–), and nitrogen atoms of the isoxazole ring donate or accept hydrogen bonds with the hydroxyl and oxygen atoms of H₂PO₄⁻, which improves adsorption. The sulfisoxazole molecules improve the adsorption through various polar functional groups such as sulfonyl (S O), sulfonamide (–SO₂NH–), amino (–NH₂), and heterocyclic nitrogen atoms that form strong dipole–ion interactions with phosphate oxyanions and increase the hydrophilicity of the framework, facilitating the diffusion of phosphate to the active sites. Furthermore, protonated functional groups can electrostatically form weak outer-sphere complexes with phosphate prior to conversion to stronger inner-sphere zirconium-phosphate complexes. The aromatic benzene and isoxazole rings of sulfisoxazole also indirectly contribute by modifying the electronic environment around the functional groups, enhancing the polarization of the local environment and reinforcing the intermolecular interactions. The π–π interaction is not expected because phosphate does not have aromatic electrons. The nitrogen adsorption–desorption results further support another key mechanism, which is pore-filling. After adsorption, the BET surface area decreased from 738.04 to 688.74 m² g⁻¹, the pore volume decreased from 6.14 to 5.38 cm³ g⁻¹ and the average pore radius decreased from 2.44 to 2.08 nm, which confirms that phosphate molecules penetrated into and filled the internal porous network of SFX-UiO-66. These reductions are interpreted as the partial blockage of micropores and mesopores by the adsorbed phosphate species and confirm that adsorption did take place within the framework and not only on the external surface. Phosphate was transported from the bulk solution to interconnected porous channels by capillary diffusion and intraparticle diffusion until equilibrium was achieved. Besides, there can also be some ion-exchange by replacement of the weakly bound counterions or the protonated water molecules attached to the functional groups. van der Waals interactions might play a role in some extra stabilization after stronger chemical interactions have taken place, but their contribution is significantly less than that of coordination bonding. The overall adsorption process is governed by electrostatic attraction, Zr–O–P inner-sphere complexation, ligand exchange, hydrogen bonding, dipole–ion interactions from sulfisoxazole functionalization, pore filling and intraparticle diffusion, which work cooperatively to impart the remarkable affinity and selectivity of SFX-UiO-66 towards phosphate ions.

3.2.6. Effect of salinity

The effect of salinity on the adsorption performance of SFX-UiO-66 on phosphate removal was studied with different dissolved salts at a concentration of 50 ppm in solutions containing 150 ppm phosphate ions. The results obtained showed that the presence of dissolved salts significantly affected the adsorption efficiency due to competitive ion interactions, ionic strength effects and the change in electrostatic attraction between phosphate ions and the active adsorption sites of the adsorbent. The highest removal efficiencies are presented in Fig. 9. The lack of added salts led to a removal efficiency of 98.2% for phosphate, which suggests that the SFX-UiO-66 shows a strong intrinsic affinity to phosphate ions under non-competitive conditions. The presence of monovalent salts such as NaCl and KCl resulted in only a slight decrease in removal efficiency with phosphate removal dropping to 94.6 and 91.3%, respectively.³⁰ This relatively small drop indicates the weaker competition of monovalent chloride ions with phosphate for the active adsorption sites of the adsorbent. However, the presence of divalent salts such as CaCl₂ and MgCl₂ resulted in a greater reduction in the adsorption efficacy, reducing phosphate removal to 84.8 and 79.6% respectively. The stronger repression by divalent cations is likely due to their higher ionic strength and stronger binding to the negatively charged functional groups on the membrane surface, which in turn partially blocks the adsorption sites and compresses the electrical double layer.

Fig. 9. Effect of interference ion son adsorption of PO₄ onto SFX-UiO-66 sensor.

3.2.7. Reusability

A systematic study on the regeneration and reusability of the SFX-UiO-66 for the removal of phosphate under optimum adsorption conditions was performed. This was done using individual phosphate solutions with concentration of 100 mg L⁻¹, contact time of 100 min, adsorbent dosage of 0.02 g and solution volume of 25 mL. The saturated SFX-UiO-66 was regenerated through 0.1 mol L⁻¹ NaOH solution after each adsorption cycle to desorb the phosphate ions adsorbed on the active adsorption sites. The regenerated membrane was washed several times with deionized water until the pH became neutral and dried before reusing in the next adsorption cycle. The results showed that the SFX-UiO-66 still maintained high removal efficiency after five cycles of adsorption–desorption, and the phosphate removal efficiency was slightly decreased from 97.4 to 90.6% Fig. S2(a). The slight decrease in adsorption performance after multiple cycles can be attributed to partial occupation or deactivation of some active sites and a minor loss of material during the regeneration process. However, the membrane still exhibited excellent adsorption efficiency, which proved its high structural stability, chemical durability and strong regeneration capability. The XRD patterns of the regenerated SFX-UiO-66 after phosphate removal are shown in Fig. S2(b). It illustrates that the characteristic diffraction peaks of the membrane almost did not change after regeneration, which proves that the crystalline structure and structural integrity of the SFX-UiO-66 were well maintained in the adsorption–desorption cycles. Furthermore, no obvious peak shift and no extra diffraction peaks were observed, which further demonstrates the good stability of the polymeric matrix and SFX-UiO-66 active sites during the regeneration process. In conclusion, these results demonstrated the excellent reusability, regeneration efficiency and long-term operational stability of the SFX-UiO-66, indicating the great potential of this adsorbent for practical wastewater treatment and sustainable nutrient-removal applications.⁴⁰

3.2.8. Comparison with other adsorbents

Table S12 presents the estimation of extreme sorption capacities for the comparison with the values mentioned in the works. This is a step towards achieving such broad understanding of the effectiveness of the sorbent, despite the inherent difficulties to establish direct correlations in findings due to differing experimental conditions. Another important advantage of SFX-UiO-66 is that it is a fast interaction sorbent. The very high binding capacity to phosphate shows that this material could be used for wastewater management to remove this pollutant from water.

3.3. Sensing process

3.3.1. Optimization of sensing conditions

The pH-dependent response profile rose sharply between pH 2 and 6, reached a maximum at pH 6, and declined at more alkaline pH Fig. 10(a). This trend tracks the protonation behavior of both the sensor and the analyte: at low pH, competitive protonation of the sulfonamide and imine nitrogen sites reduces their availability for phosphate coordination, while above pH 7 the increasing deprotonation of PO₄³⁻ toward HPO₄²⁻ and PO₄³⁻ forms, combined with partial deprotonation of the sulfonamide N–H itself, weakens the hydrogen-bonding interaction that anchors the complex. A pH of 6 was therefore adopted as optimal, balancing sufficient protonation of the recognition sites against retention of the anion's binding-competent speciation.

Fig. 10. Optimization of the spectrophotometric sensing response of SFX-UiO-66 toward PO₄³⁻: (a) effect of solution pH, (b) effect of sensor dosage, (c) effect of contact time, and (d) effect of temperature. Error bars represent the standard deviation of triplicate measurements.

Sensor dosage studies showed the absorbance signal increasing up to approximately 5 mg per 20 mL before plateauing, reflecting saturation of available phosphate in solution relative to the number of accessible binding sites; 5 mg was selected as the working dosage since higher loadings offered no further gain while increasing turbidity and light scattering background Fig. 10(b).

Contact time studies showed the signal reaching a stable plateau within 60 s, a response speed consistent with the confined, short-diffusion-path pore geometry, 60 s was adopted as the standard equilibration time for all subsequent measurements Fig. 10(c).

Temperature had a comparatively modest effect on signal intensity across the 20–40 °C range tested, with a broad optimum near room temperature and a gradual decline at higher temperature attributable to weakening of the non-covalent hydrogen-bonding component of the interaction; ambient temperature (approximately 25 °C) was used throughout the remaining experiments Fig. 10(d).

3.3.2. Spectrophotometric detection performance

Sequential addition of PO₄³⁻ to a fixed dispersion of SFX-UiO-66 produced a concentration-dependent increase in UV-vis absorbance Fig. 11(a), with no accompanying shift in the position of the diagnostic absorption band, indicating a purely hypochromic-type response governed by the extent of complexation rather than by a change in the electronic transition itself Fig. 11(b). The absorbance increase tracked phosphate concentration linearly across the 0.111–0.20 ppm range, and the well-defined isosbestic behavior observed across the titration series is consistent with a single, dominant equilibrium between the free sensor and its 1 : 1 complex with PO₄³⁻, rather than a mixture of competing binding modes Fig. 11(c).

Fig. 11. (a) UV-vis absorption spectra of SFX-UiO-66 in the presence of increasing PO₄³⁻ concentrations (0.0–1.50 ppm), (b) full-range plot of (A₀–A) versus [PO₄³⁻], (c) linear calibration plot of (A₀–A) versus [PO₄³⁻].

3.3.3. Calibration curve, LOD, and LOQ

The calibration relationship constructed from the titration series gave a linear regression of absorbance against PO₄³⁻ concentration across 0.111–0.20 ppm with a correlation coefficient of R² = 0.999 (Table 1), confirming excellent adherence to Beer–Lambert behavior within the tested range. The limit of detection (3σ/S) and limit of quantification (10σ/S) were calculated according to IUPAC recommendations from the standard deviation of the blank response (σ) and the calibration slope (S), giving LOD = 3.65 × 10⁻² ppm and LOQ = 1.11 × 10⁻¹ ppm. This detection limit falls comfortably below the 0.05 mg L⁻¹ (50 ppb) threshold commonly cited for stream discharge into eutrophication-sensitive water bodies,⁴ indicating that SFX-UiO-66 possesses the sensitivity needed for early-warning environmental monitoring rather than only gross-contamination screening.

Table 1. Analytical performance parameters of SFX-UiO-66 for PO₄³⁻ detection.

Parameter	Value
Linear range (ppm)	0.111–0.20
Regression equation	0.582 × [PO43−] + 0.0011
Correlation coefficient (R2)	0.999
LOD (ppm)	3.65 × 10−2
LOQ (ppm)	1.11 × 10−1
Optimal pH	6
Response time (s)	60
Sensor dosage (mg/20 mL)	5

3.3.4. Selectivity and interference studies

Selectivity was assessed by exposing SFX-UiO-66 to a fixed PO₄³⁻ concentration in the presence of a substantial excess of common co-existing ions, including Cl⁻, NO₃⁻, SO₄²⁻, HCO₃⁻, and the cations Na⁺, K⁺, Ca²⁺, and Mg²⁺ typical of tap and boiler feed waters Fig. 12. Signal recovery in the presence of these potential interferents remained more than 80%, within acceptable tolerance limits, indicating negligible cross-reactivity.⁴¹ This selectivity is attributable to a combination of factors operating together: electronically, none of the tested monatomic or simple oxyanion interferents present the tetrahedral, multiply hydrogen-bond-accepting geometry that PO₄³⁻ offers to the sulfonamide/imine recognition pocket; geometrically, the confined pore environment favors ions whose size and shape match the accessible channel dimension after grafting; and in terms of pH speciation, PO₄³⁻ at the optimal working pH of 6 exists predominantly as H₂PO₄⁻/HPO₄²⁻, forms that retain multiple hydrogen-bond-accepting oxygens not available to simple monovalent anions such as Cl⁻ or NO₃⁻ under the same conditions.

Fig. 12. Selectivity of SFX-UiO-66 toward PO₄³⁻ in the presence of common interfering species. Error bars represent the standard deviation of triplicate measurements.

3.3.5. Sensing mechanism

Bringing together the spectroscopic, 2D-COS, and computational evidence, the sensing event proceeds through a small number of interlocking steps. The octahedral SFX-UiO-66 particle presents a moderately constricted, mesopore-containing channel network that allows PO₄³⁻ to diffuse into proximity with the grafted sulfisoxazole units within seconds rather than minutes, consistent with the 60 s response time.⁴¹ The flexible glutaraldehyde tether, distinct from a rigid direct-coupling linkage, allows the sulfisoxazole head group some rotational freedom to orient its recognition face toward the approaching anion without requiring reorganization of the zirconium framework itself. Initial engagement occurs through hydrogen bonding between the electrophilic sulfonamide N–H (NBO charge +0.438) and the phosphate oxygens, the first event resolved by asynchronous 2D-COS analysis; this is followed by a secondary, Lewis acid–base-type coordination involving the imine nitrogen and the sulfonyl/isoxazole oxygens, which together stabilize the bound complex and account for the smaller, delayed shift observed in the imine C N band. At the optimal pH of 6, PO₄³⁻ is present predominantly as H₂PO₄⁻/HPO₄²⁻, retaining sufficient negative charge and multiple accessible oxygen lone pairs to sustain this multi-point interaction while avoiding the competitive protonation of the sensor's own recognition sites that suppresses response at lower pH. The resulting complex redistributes electron density from the sensor's conjugated donor framework toward the bound-anion environment, sharply narrowing the HOMO–LUMO gap (4.01 → 1.33 eV) and producing the concentration-dependent absorbance enhancement that constitutes the analytical signal. Because the bound phosphate remains physically retained within the pore/surface environment of the composite rather than being released back to solution, the same interaction responsible for the optical signal simultaneously withdraws PO₄³⁻ from the surrounding water, giving SFX-UiO-66 its combined sensing and removal character.

3.3.6. Real sample analysis

No native phosphate was detected in the untreated tap water sample by either method, and recoveries following spiking at 100 and 150 ppb fell between 100.29% and 100.73% for SFX-UiO-66, closely matching the 100.18–100.33% recovery obtained by IC on the same spiked aliquots. The two boiler water samples, analyzed without spiking, gave phosphate concentrations of 3.57 ± 0.27 and 2.95 ± 0.26 ppb by the proposed sensor against 3.48 ± 0.18 and 2.88 ± 0.18 ppb by IC. Relative standard deviations for the proposed method less than 2% were modestly higher than those obtained by IC, as expected for a solid-dispersion absorbance measurement relative to a chromatographic separation, but the two methods agreed within their combined uncertainty in every case tested, with no evidence of a systematic matrix-dependent bias between tap water and the more chemically complex boiler water Table 2. These results indicate that SFX-UiO-66 can be applied directly to both potable-grade and industrial process water without requiring extensive sample pretreatment beyond filtration and pH adjustment, supporting its practical utility for routine phosphate surveillance.⁴²

Table 2. Recovery of PO₄³⁻ from real water samples using SFX-UiO-66 compared with ion chromatography (IC).

Sample	Added (ppm)	Found, proposed method (ppm) ± SD	Recovery, proposed (%)	Found, IC (ppm) ± SD	Recovery, IC (%)
Tap water	—	ND	—	ND	—
	0.10	0.1007 ± 0.25	100.73	0.1003 ± 0.15	100.33
	0.15	0.1504 ± 0.25	100.29	0.1503 ± 0.15	100.18
Boiler water 1	—	3.57 ± 0.27	—	3.48 ± 0.18	—
Boiler water 2	—	2.95 ± 0.26	—	2.88 ± 0.18	—

3.4. Statistical analysis

3.4.1. ANOVA

For further optimization of the adsorption of phosphate onto SFX-UiO-66, Response Surface Methodology (RSM) based on a Box–Behnken Design (BBD) was used. The analysis of variance (ANOVA) indicated that the developed quadratic model was highly significant (F = 52.12, p < 0.0001), which means that the adsorption behavior was well explained by the selected experimental variables. Among the examined factors, contact time (B) was the most influential on phosphate adsorption (F = 315.12, p < 0.0001), followed by adsorbent dosage (C) (F = 31.40, p = 0.0008) and solution pH (A) (F = 7.61, p = 0.0281), showing that all three main factors had a significant impact on the adsorption efficiency. Among the interaction factors, only the interaction between contact time and adsorbent dose (BC) was statistically significant (F = 11.04, p = 0.0127) meaning that the effect of contact time on phosphate removal depends on the amount of SFX-UiO-66 used. However, interaction between pH and contact time (AB) (p = 0.1723) and pH and adsorbent dosage (AC) (p = 0.3133) were not statistically significant, which means that these parameters were mostly independent in the studied experimental range. The quadratic terms A² (F = 28.35, p = 0.0011) and B² (F = 65.74, p < 0.0001) were very significant, confirming the existence of nonlinear effects of pH and contact time on the phosphate adsorption Table 3. However, C² (p = 0.6496) was insignificant, indicating that the influence of adsorbent dosage was almost linear in the studied range. The high value of the coefficient of determination (R² = 0.9853) further proved the statistical adequacy of the model, meaning that 98.53% of the phosphate adsorption variation was described by the proposed quadratic model. The anticipated R² (0.7647) was found to be closely aligned with the adjusted R² (0.9664), demonstrating a reasonable predictive power and acceptable model agreement, but a moderate difference was noted due to the constraints of the experimental design. Additionally, the low coefficient of variation (C.V. = 9.80%) suggests good experimental precision and reproducibility. The high Adeq Precision value (24.48) is substantially greater than the recommended minimum value of 4, confirming an adequate signal-to-noise ratio and showing that the model can be reliably used to navigate the design space and predict phosphate adsorption. These statistical parameters collectively validate the robustness, statistical reliability, and high suitability of the developed RSM model for optimization of phosphate adsorption onto SFX-UiO-66 where contact time is the most influential process variable followed by adsorbent dosage and solution pH.

Table 3. ANOVA for the RSM model of phosphate adsorption.

Source	Sum of squares	df	Mean squares	F-value	P-value
Model	47 783.78	9	5309.31	52.12	<0.0001	Significant
A-pH	775.59	1	775.59	7.61	0.0281
B-Time	32 100.71	1	32 100.71	315.12	<0.0001
C-Dose	3198.80	1	3198.80	31.40	0.0008
AB	235.32	1	235.32	2.31	0.1723
AC	120.23	1	120.23	1.18	0.3133
BC	1124.93	1	1124.93	11.04	0.0127
A 2	2887.79	1	2887.79	28.35	0.0011
B 2	6696.90	1	6696.90	65.74	<0.0001
C 2	22.93	1	22.93	0.2251	0.6496
Residual	713.08	7	101.87
Lack of fit	713.08	3	237.69
Pure error	0.0000	4	0.0000
Cor total	48 496.86	16
Std. dev.	10.09
Mean	102.94
C.V.%	9.80
R 2	0.9853
Adjusted R2	0.9664
Predicted R2	0.7647
Adeq precision	24.4785
PRESS	11 409.28
−2 log likelihood	111.76
BIC	140.09
AICc	168.43

The regression model fitting compared linear, two factor interaction (2FI), quadratic, and cubic models for optimizing phosphate adsorption onto SFX-UiO-66 using Response Surface Methodology (RSM). The quadratic model was chosen as the best fit due to the lowest sequential p-value (0.0002), showing significant improvement over linear and 2FI models. The linear model was significant (p = 0.0004) but had poor prediction accuracy (adjusted R² = 0.6848, predicted R² = 0.5669) and failed to represent the nonlinear behavior accurately. The 2FI model was insignificant (p = 0.7222) with low predictive performance (adjusted R² = 0.6390, projected R² = 0.2691). The quadratic model, however, demonstrated enhanced performance (adjusted R² = 0.9664, predicted R² = 0.7647), with good agreement between experimental and predicted values. While the cubic model had a corrected R² of 1.0000, it was aliased, indicating insufficient data for independent cubic coefficient estimation. Thus, the quadratic regression model was selected for its statistical significance, predictive ability, and avoidance of overfitting, demonstrating robustness in predicting phosphate adsorption onto SFX-UiO-66 Table 4.

Table 4. Model fitting assesses linear, interaction, quadratic, and cubic regression models to optimize phosphate adsorption on SFX-UiO-66 sensor using response surface methodology.

Source	Sum of squares	df	Mean square	Sequential p-value	Adjusted R2	Predicted R2
Linear	12 421.76	9	1380.20	0.0004	0.6848	0.5669
2FI	10 941.28	6	1823.55	0.7222	0.6390	0.2691
Quadratic	713.08	3	237.69	0.0002	0.9664	0.7647	Suggested
Cubic	0.0000	0			1.0000		Aliased

The Box–Behnken Design (BBD) within Response Surface Methodology (RSM) successfully represented phosphate adsorption onto SFX-UiO-66. The regression equation reveals the relationship between adsorption capacity (Q_e) and three independent variables: solution pH, contact time, and adsorbent dosage, along with their interactions. Positive coefficients for pH (+17.8145) and contact time (+3.81423) show that increasing these variables enhances adsorption capacity, while the negative coefficient for adsorbent dosage (−19.2845) indicates that higher dosages reduce capacity due to underutilized adsorption sites. The interaction between pH and dosage (AC, +5.71094) demonstrates a synergistic effect, whereas the interactions of pH with time (AB, −0.0403684) and time with dosage (BC, −1.47105) exhibit antagonistic effects, slightly diminishing phosphate adsorption when both are increased. Negative quadratic coefficients for pH² (−1.6368), time² (−0.0176759), and dosage² (−40.5165) confirm a nonlinear adsorption process, indicating capacity increases to an optimal point before declining. The quadratic regression model effectively predicts adsorption performance and optimizes conditions for phosphate removal by SFX-UiO-66 eqn (4).

Q_e = −32.7206 + 17.8145 × pH + 3.81423 × time + −19.2845 × dose + −0.0403684 × pH time + 5.71094 × pH dose − 1.47105 × time dose − 1.6368 × pH² − 0.0176759 × time² − 40.5165 × dose² 4

The adsorption capacity of phosphate onto SFX-UiO-66 was analyzed using a quadratic polynomial equation from the Box–Behnken Design (BBD), where A, B, and C represent solution pH, contact time, and adsorbent dosage. Contact time (+63.345) significantly enhances phosphate adsorption, while pH (−9.84625) and dosage (−19.9963) negatively impact it beyond optimal values due to electrostatic effects and underutilization of active sites. Interaction terms show a synergistic effect between pH and dose (AC) (+5.4825) but antagonistic effects between pH and time (AB) (−7.67) and time and dose (BC) (−16.77), indicating that increasing these factors together slightly inhibits their overall effect on phosphate adsorption. The negative quadratic coefficients for pH² (−26.1888), time² (−39.8812), and dose² (−2.33375) highlight the non-linear adsorption process and optimal conditions beyond which capacity decreases. In summary, the quadratic regression equation accurately expresses the combined effects of pH, contact time, and dosage on phosphate adsorption by SFX-UiO-66, aiding in performance prediction and optimization within the experimental domain eqn (5).

Q_e = 135.13 + −9.84625 × A + 63.345 × B + −19.9963 × C − 7.67 × AB + 5.4825 × AC − 16.77 × BC − 26.1888 × A² − 39.8812 × B² − 2.33375 × C² 5

3.4.2. Response surface analysis and modelling of experimental designs

Fig. 13(a) displays the three-dimensional response surface, contour, and desirability plots of solution pH (A) and contact time (B) on phosphate adsorption by SFX-UiO-66. The adsorption capacity increases with contact time, peaking at ∼80–100 min, as time allows phosphate ions to diffuse into the adsorbent's porous structure. Initially, the capacity improves with pH, reaching a maximum around pH 5, then declines at higher pH levels. Below the pH_ZPC of the adsorbent (6.6), the surface is positively charged, attracting H₂PO₄⁻ ions. As pH increases, the surface charge becomes less positive and eventually negative, leading to electrostatic repulsion with phosphate anions and competition from hydroxide ions, reducing phosphate uptake. The contour plot indicates a statistically less significant interaction between pH and contact time than the main contact time effect, according to ANOVA. The desirability plot shows maximum values near pH 5 and contact time of 100 min, optimal for phosphate adsorption. Thus, contact time is critical, while maintaining pH around 5 enhances SFX-UiO-66's adsorption capacity.

Fig. 13. 3D interaction, contour, and desirability: (a) communication among pH and time, (b) dose and pH, and (c) interaction between dose and time.

Fig. 13(b) displays the three-dimensional response surface and contours showing the concurrent impact of solution pH (A) and adsorbent dose (C) on phosphate adsorption onto SFX-UiO-66, with contact time held constant. The response surface indicates that phosphate adsorption capacity increases as solution pH rises from acidic to about pH 5, then declines at higher pH. This trend is linked to the pH-dependent surface charge of SFX-UiO-66, which attracts phosphate species (H₂PO₄⁻) below its pH_ZPC (6.6). Above this pH, negative surface charge leads to repulsion from phosphate anions and competition from hydroxide ions, reducing adsorption capacity. Conversely, while increasing adsorbent dosage decreases adsorption capacity (mg g⁻¹), overall phosphate removal efficiency improves due to more available adsorption sites. The reduction in capacity stems from the fixed phosphate amount and potential particle aggregation at higher dosages, limiting effective surface area. The contour plot shows a moderate interaction effect between pH and dosage, although ANOVA indicated this interaction was statistically insignificant (AC, p = 0.3133). The desirability plot identifies an optimal pH of about 5 and low dosages (≈0.02 g) for maximum adsorption capacity, driven by favorable phosphate species, electrostatic attraction, ligand exchange, and formation of stable Zr–O–P complexes, confirming SFX-UiO-66's effectiveness in mildly acidic conditions with minimal dosage.

In Fig. 13(c), the three-dimensional response surface, contour, and desirability plots illustrate the interaction between adsorbent dosage (A) and contact time (B) on phosphate adsorption by SFX-UiO-66 at a mid-level pH. The adsorption capacity increased significantly with contact time, peaking around 100 min before decreasing with rising adsorbent dosage from 0.02 to 0.50 g. Longer contact time allows more phosphate ions to diffuse into the SFX-UiO-66 micropores and occupy Zr(iv) centers through various interactions. Conversely, increased adsorbent dosage leads to limited phosphate availability, resulting in incomplete use of adsorption sites, while particle aggregation can block pore openings and reduce effective surface area. The contour plot's circular lines confirm the interaction between contact time and adsorbent dosage, supported by an ANOVA finding (p = 0.0127) indicating significant interaction. The desirability plot shows that maximum values (∼1.0) occur at a lengthy contact time (∼100 min) and low adsorbent dosage (∼0.02 g), optimizing adsorption capacity. Increased contact duration enhances adsorption equilibrium and the use of SFX-UiO-66's active sites, while excessive adsorbent ultimately decreases the mass-based adsorption capacity. Thus, response surface analysis indicates that optimal phosphate adsorption occurs with minimal adsorbent dosage and adequate contact time for efficient site utilization.

3.4.3. Model adequacy checking

A normal probability plot of externally studentized residuals in Fig. 14(a) checks the normality assumption for the quadratic Response Surface Methodology (RSM) model of phosphate adsorption onto SFX-UiO-66. The residuals align closely with the straight reference line, indicating an approximately normal distribution without major outliers. This suggests random, independent experimental errors and no serious violations of normality during model development. Most residuals cluster around zero, parallel to the normal probability line, revealing that the regression model effectively captures the relationship between operating variables and phosphate adsorption capacity, with no bias or systematic errors. The fit between experimental residuals and theoretical normal distribution affirms the assumptions of the ANOVA analysis, validating the statistical significance of the RSM model for predicting and optimizing phosphate adsorption on SFX-UiO-66.

Fig. 14. (a) Normal probability plot of residuals for phosphate adsorption model, (b) predicted versus actual values for phosphate adsorption model, (c) Box–Cox plot for power transforms of phosphate adsorption model, (d) residuals versus predicted values for phosphate adsorption model, (e) perturbation plot for phosphate adsorption model, and (f) cube plot for phosphate adsorption model.

The quadratic Response Surface Methodology (RSM) model for phosphate adsorption on SFX-UiO-66 was validated using a projected vs. actual plot to assess its predictive accuracy. The projected adsorption capacities closely aligned with the actual values, as indicated by their proximity to the 45° reference line, demonstrating the model's effectiveness in explaining the relationship between independent variables (pH, contact time, and adsorbent dosage) and phosphate adsorption capacity. A few data points deviated slightly, reflecting minor prediction errors, which are typical in experimental research and do not greatly affect the model's overall performance. The high coefficient of determination (R² = 0.9853), adjusted R² (0.9664), and predicted R² (0.7647) from ANOVA analysis confirm the model's strong predictive capabilities. The good agreement between observed and predicted responses indicates no systematic bias or significant overfitting, allowing for reliable predictions of phosphate adsorption under various conditions. Thus, the plot validated the robustness and predictive performance of the quadratic RSM model for optimizing phosphate adsorption Fig. 14(b).

To test the need for a transformation of the response variable in the quadratic RSM model for phosphate adsorption onto SFX-UiO-66, the Box–Cox plot was applied. The best Box–Cox transformation parameter (lambda) is around zero, while the currently selected transformation (lambda = 1) falls within the 95% confidence interval. This suggests no significant improvement in model performance by transforming the response variable, as the adsorption capacity can be evaluated in its original scale. The smooth minimum in the Box–Cox curve further indicates that residual variance is nearly constant. Additionally, the small residual sum of squares at the optimal λ shows that the quadratic model effectively stabilizes variance and minimizes prediction errors without requiring transformations. These findings align with the normal probability plot of residuals, a high coefficient of determination (R² = 0.9853), and satisfactory adjusted and predicted R² values. The results confirm the statistical significance, assumption satisfaction, and predictive capability of the developed RSM model Fig. 14(c).

The adequacy of the quadratic Response Surface Methodology (RSM) model was assessed using the externally studentized residuals versus predicted values plot shown in Fig. 14(d). The residuals scatter randomly around the zero-reference line, indicating constant variance without any apparent trend or funnel shape. This suggests the model accurately reflects the relationship between operating variables and phosphate adsorption capacity, with no evidence of model misspecification or heteroscedasticity. All external studentized residuals are within the control limits (±4.82), indicating no statistically significant outliers. The symmetrical distribution of residuals around zero shows that model predictions are unbiased and errors are random. These findings align with the normal probability plot, the projected vs. actual plot, and the Box–Cox analysis, confirming that regression assumptions are satisfied. Therefore, the residuals vs. predicted values plot supports the robustness and predictive ability of the quadratic RSM model for optimizing phosphate adsorption onto SFX-UiO-66 Fig. 14(d).

The perturbation plot in Fig. 14(e) illustrates the effects of solution pH (A), contact time (B), and adsorbent dosage (C) on phosphate adsorption capacity (q_e) onto SFX-UiO-66. The steepest curve indicates that contact time has the greatest influence on adsorption, improving significantly with longer durations due to slow phosphate diffusion and site occupation until equilibrium is reached. In contrast, the pH demonstrates a parabolic shape: adsorption capacity increases as pH approaches around 5, then decreases at higher pH values due to pH-dependent surface charges. Below pH_ZPC (6.6), the positively charged adsorbent attracts phosphate ions, while competition with hydroxide ions rises at higher pH. Adsorbent dosage (C) has the least impact, causing a slight decrease in q_e as the dosage increases because of reduced available sites and potential particle aggregation. The plot shows that contact duration is most sensitive to changes, followed by solution pH, with adsorbent dosage exerting the least influence. ANOVA results affirm the findings, with contact time having the highest F-value (315.12).⁴³

The cube plot presented in Fig. 14(f) shows the global view of the combined effects of solution pH (A), contact time (B) and adsorbent dosage (C), on the adsorption capacity of phosphate on SFX-UiO-66 by showing the predicted response at the eight corner points of the experimental design space. The maximum anticipated adsorption capacity (189.84 mg g⁻¹) was found at the combination of low pH (2), high contact time (100 min) and low adsorbent dose (0.02 g), demonstrating that long term adsorption duration along with low adsorbent dosage greatly improves phosphate uptake. In contrast, the lowest adsorption capabilities (−3.15 to 14.27 mg g⁻¹) were anticipated at short contact time (5 min) regardless of pH and dosage, suggesting that insufficient contact time substantially inhibits phosphate diffusion and adsorption equilibrium.⁴³ The adsorption capacity decreased significantly at the higher adsorbent dosage (0.50 g) with predicted values from 3.46 to 105.34 mg g⁻¹, indicating that the increase of the adsorbent dosage reduces the adsorption capacity on a mass basis due to the incomplete occupation of the available adsorption sites. Moreover, increasing the solution pH from 2 to 10 also resulted in a decrease in adsorption capacity under most operating conditions, which is attributed to the reduction of positive surface charge of SFX-UiO-66 above its pH_ZPC (6.6) and the subsequent electrostatic repulsion between the negatively charged adsorbent surface and phosphate anions, as well as the enhanced competition from hydroxide ions for the coordinatively unsaturated Zr(iv) active sites. Overall, the cube plot shows that contact time is the most positively influencing factor on phosphate adsorption, while increased pH and high adsorbent dosage negatively affect the adsorption capacity, consistent with the ANOVA and perturbation analyses. These results further confirm that maximum phosphate adsorption by SFX-UiO-66 is achieved under conditions of low adsorbent dosage, long contact time, and mildly acidic pH, where electrostatic attraction, ligand exchange, hydrogen bonding, pore filling, and the formation of stable Zr–O–P inner-sphere complexes collectively promote efficient phosphate removal.⁴⁴

3.4.4. Validation of models and the desirability approach

The Response Surface Methodology (RSM) model underwent desire function analysis to identify optimal parameters for maximum phosphate adsorption onto SFX-UiO-66. The overall desirability was 0.991, indicating strong agreement between the model and experimental conditions. The best settings were determined as pH 5.0, a contact time of 100 min, and an adsorbent dosage of 0.02 g, with an expected adsorption capacity of 197.34 mg g⁻¹. The optimum pH aligns with the zero-point charge (6.6) of SFX-UiO-66, facilitating electrostatic attraction to phosphate (H₂PO₄⁻) and the formation of stable Zr–O–P complexes. The 100 min contact time ensures effective diffusion of phosphate ions to establish adsorption equilibrium.⁴⁴ The low dosage of 0.02 g maximizes adsorption capacity by efficiently utilizing active sites. The high desirability indicates the RSM model's successful prediction of optimum conditions and confirms the effectiveness of SFX-UiO-66 as an adsorbent for phosphate removal from water Fig. 15(a).

Fig. 15. (a) Numerical optimization ramp plot for phosphate adsorption onto SFX-UiO-66, (b) desirability function plot for optimization of phosphate adsorption onto SFX-UiO-66, (c) desirability and response profile plots for optimization of PO₄ adsorption onto SFX-UiO-66, and (d) three-dimensional desirability surface for optimization of phosphate adsorption onto SFX-UiO-66.

The desirability analysis confirmed the optimization results from the Response Surface Methodology (RSM) model for phosphate adsorption onto SFX-UiO-66. The optimized operating variables showed desirability values of 0.999999 for pH, 0.999999 for contact time, and 1.000 for adsorbent dosage, indicating proximity to optimum values. Under these conditions, the predicted adsorption capacity (q_e) had a desirability value of 0.9651, indicating good alignment with the desired target.⁴⁵ The overall combined desirability was 0.9912, close to the ideal value of unity, affirming the RSM model's robustness and predictive capability. The conditions pH 5, 100 min contact time, and 0.02 g adsorbent dosage maximize adsorption capacity by promoting favorable phosphate speciation, strong electrostatic attraction, efficient diffusion, and stable Zr–O–P coordination complexes. Thus, the analysis validates that the optimization model is reliable and provides accurate conditions for maximizing phosphate adsorption by SFX-UiO-66, highlighting its potential for phosphate removal from water Fig. 15(b).

The ramp desirability plots in Fig. 15(c) illustrates the influence of independent variables on phosphate adsorption and identify optimum operating conditions from the Response Surface Methodology (RSM) model. The maximum desirability occurs at a solution pH of roughly 5, a contact time of 100 min, and an adsorbent dosage of 0.02 g, corresponding to a predicted adsorption capacity of about 197.34 mg g⁻¹. The pH profile follows a parabolic trend, indicating that phosphate adsorption increases with pH until 5, after which capacity declines due to the positively charged surface of SFX-UiO-66, which attracts H₂PO₄⁻ ions but competes with hydroxide ions at higher pH levels. The contact time profile shows that both desirability and adsorption capacity rise until 100 min, highlighting the time needed for phosphate ions to diffuse and form stable interactions. However, increasing adsorbent dosage leads to decreased adsorption capacity due to excess available sites resulting in diminished utilization. Predicted adsorption capacity curves display narrow confidence intervals, indicating high precision and confidence in the model predictions. Overall, the optimized conditions (pH 5, contact time 100 min, adsorbent dosage 0.02 g) maximize phosphate adsorption performance, confirming the robustness of the quadratic RSM model for optimizing phosphate removal.

Fig. 15(d) shows the three-dimensional desirability surface reflecting the effects of solution pH (A) and contact time (B) on phosphate adsorption over SFX-UiO-66 at an optimal dosage of 0.02 g. Desirability increases with contact time, peaking at about 0.991 at pH 5 and 100 min, which is deemed the optimal zone for phosphate adsorption as per the Response Surface Methodology (RSM) model. The gradual increase suggests that adsorption equilibrium is reached with phosphate ions occupying active sites in the SFX-UiO-66 structure. The highest desirability at pH 5 is attributed to the adsorbent's favorable surface chemistry, with the surface being positively charged, strongly adsorbing dominant phosphate species (H₂PO₄⁻) via electrostatic interactions. This interaction stabilizes through ligand-exchange reactions and the formation of stable Zr–O–P complexes, alongside hydrogen bonding from amino and sulfonamide functional groups. Conversely, lower contact times or suboptimal pH values diminish desirability due to insufficient adsorption site utilization. The smooth response surface indicates model stability and confirms the RSM model's predictive accuracy for optimal phosphate removal using SFX-UiO-66.⁴⁴

3.5. Machine learning

3.5.1. ANN model

Fig. 16(a) regression analysis of the experimental (target) and ANN-predicted adsorption capabilities of phosphate (PO₄³⁻) adsorption on the SFX-UiO-66 sensor. The developed artificial neural network showed a good predictive capability for training, validation, testing and combined datasets indicating that the model was able to capture the complex nonlinear relationship between the operational variables and phosphate adsorption efficiency. The regression equation Output = 0.99 × target + 0.88 and the correlation coefficient R = 0.99991 obtained for the training dataset showed an almost perfect correlation, which means that the predicted values were very close to the experimental results, and the model learned the adsorption behavior well during training. The regression equation Output = 0.86 × target + 23 obtained a high correlation coefficient (R = 0.96958) with the validation dataset, indicating good generalization ability. This confirmed the strong predictive capability of the ANN on previously unseen data without significant overfitting. The regression equation Output = 1.40 × target − 32 also demonstrated a very good correlation coefficient (R = 0.99522) on the testing dataset, which further validates the robustness and reliability of the trained network despite the short amount of testing samples. When all datasets were merged, the total correlation coefficient was R = 0.98153 with the regression equation Output = 0.97 target + 4.4, which showed a good linear agreement between the anticipated and experimental values over the whole adsorption range. In addition, the regression lines for all datasets were very close to the ideal Y = T line and the data points were densely grouped around the fitted lines, indicating small prediction errors and good model accuracy. The little differences seen between the validation and test datasets are due to the natural variability of experimental adsorption data and not due to any shortcomings of the model. In general, the results revealed that the ANN model had a good fitting accuracy, strong generalization ability and high predictive reliability, which made it a powerful computational tool to accurately predict the phosphate adsorption onto the SFX-UiO-66 sensor and to facilitate the process optimization while reducing the extensive experimental investigations.

Fig. 16. (a) Comparison of the predicted and target values of ANN model (b) ANN performance plot for adsorption of PO₄ onto SFX-UiO-66 sensor.

Fig. 16(b) shows the mean squared error (MSE) performance of the artificial neural network (ANN) during the training process for predicting phosphate (PO₄³⁻) adsorption on the SFX-UiO-66 sensor.⁴⁶ The learning efficiency and generalization capacity of the created model were evaluated by monitoring the MSE curves of training, validation and testing datasets throughout seven training epochs. As training proceeded, the training error (blue curve) was reduced continually, suggesting that the ANN successfully learnt the nonlinear relationship between input variables and phosphate adsorption performance. The minimum validation error (593.9153) is attained at epoch 4, which is considered as the ideal training point and picked as the best validation performance. After this epoch, the validation (green curve) and testing (red curve) errors become almost constant, while the training error continues to decrease considerably. This suggests that further training may not improve the predictive performance on unseen data, and may even worsen the performance due to overfitting. Thus, the ANN automatically picked the network parameters corresponding to epoch 4 where the trade-offs between learning accuracy and generalization were best. The validation and testing error curves are in good agreement which further confirms the robustness, stability and dependability of the constructed ANN model. The fast convergence of the error curves, low minimum validation MSE, and no significant divergence between validation and testing datasets generally confirm the successful training of the ANN model and its excellent predictive ability for accurate estimation of phosphate adsorption onto the SFX-UiO-66 sensor under different operating conditions.

Table 5 outlines the prediction performance of the artificial neural network (ANN) model that was created for the adsorption of phosphate (PO₄³⁻) onto the SFX-UiO-66 sensor, with the mean squared error (MSE) and coefficient of determination (R²) used as assessment metrics. The training data set showed a very low MSE of 0.7013 and an excellent R² value of 0.9999, showing that the ANN learned the complex nonlinear relationship between the adsorption parameters and phosphate removal accurately with very little prediction error. The model showed an MSE of 593.9153 and R² value of 0.9696 for the validation dataset, indicating the high prediction power and adequate generalization of the model when used for unseen data. The testing dataset also yielded an MSE of 265.3029 and a R² value of 0.9952, which confirms the resilience and dependability of the trained ANN in predicting independent experimental findings.⁴⁶ The validation MSE was larger than the training and testing errors. However, the high R² values for all datasets (>0.96) indicate that the model had outstanding predictive accuracy without substantial overfitting. The lower MSE in testing than in validation also shows the network built can represent well the independent testing samples. The very high R² values and relatively low prediction errors in general confirm that the ANN model has excellent fitting accuracy, strong generalization ability, and high computational reliability in the prediction of phosphate adsorption onto SFX-UiO-66 sensor, which makes it a valuable tool for adsorption modeling and process optimization while minimizing the need for extensive experimental investigations.

Table 5. Artificial Neural Network (ANN) model performance for adsorption of PO₄ onto SFX-UiO-66 sensor.

Parmeter	Training	Validation	Testing
MSE	0.7013	593.9153	265.3029
R 2	0.9999	0.9696	0.9952

3.6. DFT-based computational validation

Geometry optimization of the free sensor fragment and its PO₄³⁻ complex at the B3LYP/6-31G(d)/LANL2DZ level produced stable, imaginary-frequency-free structures for both states Fig. 17(a). NBO analysis of the SFX-GA fragment placed the largest positive partial charge on the sulfonamide hydrogen (+0.438) and the sulfonyl sulfur (+1.767), while the two sulfonyl oxygens (−0.773, −0.795), the isoxazole ring oxygen (−0.346), and both nitrogen atoms (−0.792 sulfonamide, −0.119 isoxazole) carried substantial negative charge (Table 6). This charge distribution identifies the sulfonamide N–H as the most electrophilic hydrogen available for donor–acceptor interaction with the electron-rich oxygens of an incoming PO₄³⁻ ion, consistent with the asynchronous 2D-COS finding that the sulfonamide site responds first to phosphate exposure.

Fig. 17. (a) Optimized molecular geometries, (b) frontier molecular orbital (FMO) analysis, and (c) molecular electrostatic potential (MESP) maps of the SFX-GA, and PO₄³⁻-complex.

Table 6. NBO atomic charges of the SFX-GA fragment.

Atom	Assignment	NBO charge
S	Sulfonyl	+1.767
O	Isoxazole ring	−0.346
O	Sulfonyl	−0.773
O	Sulfonyl	−0.795
N	Sulfonamide	−0.792
N	Isoxazole ring	−0.119
H	Sulfonamide	+0.438

Frontier molecular orbital analysis of the free ligand placed the HOMO at −0.2434 a.u. and the LUMO at −0.0961 a.u., giving an energy gap of 0.1473 a.u. (4.01 eV). Upon PO₄³⁻ complexation, both frontier orbitals dropped in energy (HOMO = −0.2743 a.u., LUMO = −0.2254 a.u.), and the gap contracted sharply to 0.0489 a.u. (1.33 eV), a reduction of roughly two-thirds relative to the free ligand Fig. 17(b). Spatial plots of the frontier orbitals showed the HOMO of the free ligand localized principally over the isoxazole/sulfonamide conjugated backbone, while in the complex the LUMO redistributed toward the region proximal to the bound phosphate oxygens, a pattern consistent with a metal-to-ligand/ligand-to-ligand charge transfer process in which electron density shifts from the sensor's donor framework toward the electron-accepting environment created by the coordinated oxyanion. This substantial gap narrowing provides a direct quantum-chemical rationale for the experimentally observed absorbance enhancement: a smaller HOMO–LUMO gap corresponds to a lower-energy, longer-wavelength-accessible electronic transition, and the resulting increase in transition probability within the visible/near-UV window manifests as the concentration-dependent absorbance growth.

Molecular electrostatic potential mapping of the free ligand showed the expected electron-deficient (blue) region concentrated around the sulfonamide N–H and sulfonyl sulfur, and electron-rich (red) regions localized over the sulfonyl and isoxazole oxygens Fig. 17(c). This polarity pattern is geometrically and electronically complementary to the tetrahedral, oxygen-rich PO₄³⁻ ion: the anion's oxygen lone pairs are positioned to engage the ligand's electron-deficient N–H, while any residual positive character on the phosphorus center is oriented toward the ligand's electron-rich oxygen sites, together supporting a combined hydrogen-bonding and Lewis acid–base binding mode rather than a single-point electrostatic contact.

4. Conclusion

Herein, we describe the successful synthesis of a new sulfisoxazole-functionalized NH₂-UiO-66(Zr) (SFX-UiO-66) via a glutaraldehyde-assisted post-synthetic modification strategy and its use as an efficient dual-functional platform for simultaneous spectrophotometric sensing and adsorption of phosphate ions from aqueous media. The detailed examination of the results obtained from XRD, FT-IR, 2D correlation FT-IR, N₂ adsorption–desorption, SEM and EDX proves that sulfisoxazole can be immobilized effectively, the UiO-66 framework is retained and the structural stability is high after phosphate absorption. The developed sensor showed fast detection of phosphate within 60 s, linear response in the range of 0.111–0.200 ppm (R² = 0.999), low detection limit of 3.65 × 10⁻² ppm, excellent selectivity over interfering anions and accurate determination of phosphate in real water samples with recoveries of 100.29–100.73%, showing its practical analytical applicability. Meanwhile, SFX-UiO-66 showed the greatest adsorption performance with the maximal Langmuir adsorption capacity of 202.8 mg g⁻¹ and the adsorption process followed Langmuir isotherm and pseudo-second-order kinetic models, supporting monolayer chemisorption. The adsorption process was dominated by the combined effect of electrostatic attraction, hydrogen bonding, ligand exchange, pore filling and strong Zr–O–P coordination. The Response Surface Methodology (RSM) and the Artificial Neural Network (ANN) model were used to find the optimal operating conditions for process optimization, and the ANN model successfully predicted phosphate adsorption with an overall R² of 0.9815, demonstrating high predictive performance. Moreover, density functional theory (DFT) calculations revealed significant electronic interactions between phosphate ions and the sulfisoxazole functionalized active sites through a large decrease in the HOMO–LUMO energy gap that theoretically supported the sensing mechanism observed in the experimental results. The synergy of sophisticated material design, fast sensing, high adsorption efficacy, machine-learning prediction, statistical optimization, and quantum-chemical calculation demonstrates the potential of SFX-UiO-66 as a new-generation multifunctional MOF for environmental phosphate sensing and remediation. Future practical environmental applications will be developed further with emphasis on long-term operating stability, continuous-flow applications, regeneration in genuine wastewater conditions and large-scale implementation.

Conflicts of interest

There are no conflicts of interest to declare.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Supplementary information (SI) is available. See DOI: https://doi.org/10.1039/d6ra07114g.