Experimental study on phosphorus removal performance from water by SW-ceramsite in a fixed-bed column Authors: Tianpeng Li, Tingting Sun, Zhan Chen, Pengpeng Jiao Published: 2025 ABSTRACT Based on the previous research on the preparation of solid waste ceramsite (SW-ceramsite), the phosphorus removal performance from aqueous solution by SW-ceramsite in a fixed-bed column was investigated. Characterization results showed that the SBET, pHpzc, pore volume, bulk density and void fraction values of pretreated SW-ceramsite were 4.18 × 104 cm2/g, 9.83, 3.05 × 103 cm3/g, 1.37 g/cm3 and 69.8%, respectively. Column experiments indicated that under optimal operating conditions of an initial pH of 5, an initial phosphorus concentration of 5 mg/L, a reaction temperature of 323 K, and an initial flowrate of 40 mL/min, the breakthrough curve (BTC) exhibited an irregular “S” shape, and the breakthrough and saturation times were 80 h and 155 h, respectively. Kinetic analysis demonstrated that compared with the Adams-Bohart model, the Yoon-Nelson model better described the phosphorus removal behavior from water by SW-ceramsite in a fixed-bed column. Grey relation analysis (GRA) results suggested that except for initial flowrate, the assumed effects of initial pH, initial concentration, and reaction temperature on the BTC were consistent with the GRA outcomes, implying that the GRA method can be used to determine the relative influence of the aforementioned factors on the phosphorus removal performance from water in a fixed-bed column packed with SW-ceramsite. Furthermore, after eight regeneration cycles, the breakthrough and saturation times of SW-ceramsite packing decreased by 30% and 12.9%, respectively, suggesting that it has a certain regenerative ability. It is widely recognized that phosphorus is an essential nutrient for ecosystems. However, its excessive discharge into water bodies poses significant threats to both environmental quality and human health1,2. The primary consequence of phosphorus pollution in aquatic systems is eutrophication, the severity of which is closely correlated with total phosphorus (TP) concentration3. Eutrophication typically begins when the TP concentration exceeds 0.02 mg/L. A TP level above 0.05 mg/L indicates mild eutrophication, while a concentration of 0.20 mg/L or higher signifies a fully eutrophic state. The underlying mechanism of phosphorus-induced eutrophication lies in its role as a key nutrient for aquatic vegetation. Elevated phosphorus levels stimulate rapid proliferation of primary producers, particularly algae, leading to algal blooms. These blooms reduce water transparency by limiting light penetration, thereby disrupting photosynthesis in submerged vegetation and altering aquatic community structure4,5. Upon senescence and decomposition, algal biomass exerts a high biochemical oxygen demand, depleting dissolved oxygen and creating hypoxic condition that further compromise aquatic life and ecosystem stability. Generally, two main types of phosphorus removal technologies are available: Biological removal and physicochemical removal6,7. Biological phosphorus removal is a cost-effective approach that relies on the assimilation activity of microorganisms such as polyphosphate-accumulating organisms (PAOs), phosphorus is subsequently eliminated through sludge wasting. Nevertheless, this strategy suffers from inherent limitations, such as long operational cycles, stringent reaction conditions, variable removal efficiency, and the generation of large quantities of excess sludge. Moreover, it often fails to guarantee consistent compliance with Class I of the Integrated Wastewater Discharge Standard of China (GB 8978 − 1996)8. Comparatively, physicochemical polishing steps are frequently required to secure stable effluent quality9. A variety of physicochemical approaches10,11, including struvite (MgNH4PO4·6H2O) precipitation, ion exchange and membrane separation, have been developed for phosphorus removal. Nevertheless, each suffers from inherent drawbacks. Struvite precipitation, for instance, demands substantial additions of magnesium and ammonium salts, inevitably raising operational costs. Adsorption, in contrast, offers the prospect of simultaneous phosphorus elimination and resource recovery. In this process, porous or high-surface-area sorbents capture HPO₄2⁻ or PO₄³⁻ ions from wastewater, the loaded phosphorus can subsequently be precipitated and removed or recovered. The efficiency of the adsorption route is governed primarily by the sorbent selected, underscoring the critical importance of high-performance adsorption materials12. Ceramsite is a novel artificial mesoporous material produced by granulating and firing clay, mudstone, slate, coal gangue, fly ash, shale, silt or other industrial solid wastes (SWs)13. Its well-developed pore network endows it with refractory, frost-resistance, chemical and biological corrosion resistance, together with good seismic performance, so it is widely applied in environmental protection, construction, landscape and horticulture14,15. Traditionally, clay ceramsite or shale ceramsite rely on valuable non-renewable natural resources, contradicting sustainable development. Replacing the above-mentioned raw materials with SWs, such as dewatered sewage sludge, coal fly ash and river sediment, simultaneously achieves resource utilization and green production16,17. Recent studies have demonstrated that SW-ceramsite is an excellent fixed-bed column packing for phosphorus removal, its performance being influenced by initial concentration, reaction temperature and initial pH18,19. However, the relative importance of these factors has rarely been systematically addressed. Fixed-bed column studies have received increasing attention because they better represent full-scale applications20. Current research focuses on the following aspects. Firstly, the amount and performance of packing material; Secondly, pollutant types and removal efficiency; Thirdly, operating variables and their optimal ranges; Finally, kinetic modelling of the fixed-bed reaction21. However, the relationship between BTC and these variables remains uncertain. A fixed-bed column is a highly dynamic system that involves complex interactions among the physicochemical properties of the packing, the specific wastewater matrix, and interfacial mass-transfer processes22,23. Consequently, the fixed-bed column can be treated as a grey system: sample sizes are small and the data may not follow a normal distribution, so conventional correlation techniques are often unsuitable24. GRA, an optimization method based on grey-system theory, has proven effective in quantifying the extent to which individual factors influence operating behaviour25,26. It identifies the key variables that govern the overall trend and ranks their relative importance without requiring strict statistical assumptions. Building on our previous findings, this study prepared and characterized packing, conducted single-factor experiments to determine the optimum operating conditions and performed kinetic modelling, applied GRA to quantify the influence of each variable on the BTC, and finally evaluated the regeneration performance. Materials and methods Materials, reagents and instruments SW-ceramsite (1.0 ± 0.1 cm in diameter) was synthesized via a single-stage sintering route. Dewatered sewage sludge served as the primary precursor, supplemented by coal fly ash and river sediment, corn stalk was employed as a pore-forming agent and waste-glass cullet was employed as a binder. At ambient temperature, the raw powders, dewatered sewage sludge, river sediment and coal fly ash, were homogenized in a beaker. Subsequently, corn stalk (9 g per 100 g of mixture), glass cullet (7 g per 100 g of mixture), and 20 mL of 0.9% (w/v) poly(vinyl alcohol) (PVA) solution were introduced. Distilled water was added dropwise (~ 25 wt% of mixture) to yield a plastic paste, which was then hand-rolled into ceramsite billets. After oven-drying at 105 °C for 12 h, they were transferred to a muffle furnace, preheated to 350 °C for 30 min, and rapidly calcined at 1100 °C for 5 min. Natural cooling to room temperature afforded SW-ceramsite with a mean diameter of 10.0 ± 0.5 mm and an individual mass of 1.0 ± 0.2 g. Detailed protocols for raw-material sourcing, pretreatment, optimal mass ratios, sintering-additive origin/dosage, process flowsheet, temperature–time profile, physical-property characterization and environmental-risk assessment have been described previously by our group27. The chemical reagents, including FeCl3, PVA, HCl, H2SO4 and NaOH were purchased from Sinopharm Chemical Reagent Co., Ltd., China and used as received without further purifcation (analytical grade). Medical gauze (8 cm×8 cm) and self-locking nylon cable ties (2.5 mm×100 mm) were supplied by Caoxian Hualu Hygiene Material Co., Ltd., China and Shandong Gunuo Plastic Co., Ltd., China, respectively. Distilled water was used for all aqueous preparations. To minimize the influence of real-wastewater matrix fluctuations, especially ionic interference, on experimental reproducibility, a synthetic phosphorus-containing wastewater was adopted throughout this study. A standard phosphorus stock solution (50.0 mg/L) was prepared as follows: Analytical-grade KH₂PO₄ was dried at 110 °C for 2 h, allowed to cool in a desiccator, and exactly 0.2197 g was dissolved in distilled water. The solution was quantitatively transferred into a 1000 mL volumetric flask, acidified with 5 mL of (1 + 1) H₂SO₄ solution, and made up to the mark with distilled water. Under these conditions, phosphorus exists predominantly as orthophosphate (PO₄³⁻). Simulated wastewaters with initial phosphorus concentrations of 5–30 mg/L, covering the typical range observed in rural domestic sewage, were obtained by sequential dilution of the stock solution. The initial pH of each test solution was adjusted with 0.01 M NaOH or 0.01 M HCl as required. A UV-Vis spectrophotometer (Model: UV754N), constant-temperature convection drying oven (Model: DHG-9143 A), muffle furnace (Model: SX2-6–13), thermostatic water-bath shaker (Model: SHA-B) and peristaltic pump (Model: BT300-02), were purchased from Shanghai Instrument and Electronics Analytical Instruments Co., Ltd., Shanghai Yitian Scientific Instrument Co., Ltd., Shanghai Yifeng Electric Furnace Co., Ltd., Jintan Chengdong Xinrui Instrument Factory and Baoding Qili Constant Flow Pump Co., Ltd., China, respectively. The fixed-bed column was designed and assembled in-house, the reaction column and support frame were fabricated from commercially available polymethyl methacrylate and stainless steel, respectively. A detailed schematic of the assembly is provided in our previous work28. Pretreatment experiment SW-ceramsite (0.10 ± 0.05 cm) was crushed, rinsed with distilled water, dried at 105 °C, allowed to cool, immersed in 15% (v/v) HCl for 60 min, and then washed with distilled water until the effluent reached neutrality. The cleaned particles were re-dried at 105 °C and cooled in a desiccator for subsequent use. 1.0 mol/L FeCl3 solution was transferred into an Erlenmeyer flask, and the SW-ceramsite was added at a solid-to-liquid ratio of 1:1 (g/mL). The suspension was shaken at 25 °C for 4 h in a thermostatic water-bath, after which it was heated on a hot plate with occasional glass-rod stirring until complete evaporation of the solvent. The resulting SW-ceramsite was allowed to cool naturally, rinsed four to five times with distilled water, dried at 105 °C, transferred to a muffle furnace, and calcined at 400 °C for 3 h. After natural cooling (seen Fig. 1a), the pretreated SW-ceramsite was stored in a desiccator until further use. Fig. 1SEM images and EDS spectra of pretreated SW-ceramsite (a) and spent SW-ceramsite (b). SEM images and EDS spectra of pretreated SW-ceramsite (a) and spent SW-ceramsite (b). Fabrication of packing bags Exactly 20.0 g of the aforementioned pretreated SW-ceramsite was weighed, spread evenly on a double-layer gauze pad (4 cm×4 cm), and enclosed by a self-locking nylon cable tie. The excess tie was trimmed off to yield a sealed pouch. These pouches were fabricated to: On the one hand, minimize SW-ceramsite attrition and loss during water treatment. On the other hand, facilitate the subsequent recovery and reuse of spent SW-ceramsite. And besides, reduce the liquid-level gradient and pressure drop across the column, thereby enhancing mass-transfer efficienc28. Column experiment The fixed-bed column employed in this study comprised an influent reservoir, a peristaltic pump with continuous-flow capability, the main reaction column, a packed layer of SW-ceramsite, a supporting frame, and an effluent reservoir. The reaction column had an internal diameter of 14.0 cm, a total height of 100.0 cm, and an effective volume of 4.4 × 10³ cm³. The top of the reaction column served as the combined inlet/outlet, while its bottom was fitted with a perforated distributor and a bottom inlet. The SW-ceramsite packing was confined to the middle section of the column. Driven by the peristaltic pump, the wastewater was introduced at the bottom and allowed to flow upward through the column under atmospheric pressure. Sampling commenced the moment effluent appeared at the outlet and continued at predetermined intervals. All runs were performed in triplicate under identical conditions. Regeneration experiment Spent SW-ceramsite was regenerated thermally. It was collected, transferred to a quartz crucible, and calcined at 600 °C for 60 min. After cooling to room temperature, the regenerated SW-ceramsite was re-packed and re-evaluated under identical pretreatment, bag-fabrication, and column-operation conditions. Analysis The SEM and EDS were employed to characterize the morphology and surface element composition, respectively. Prior to imaging, the specimens were sputter-coated with a thin gold layer to enhance conductivity and image quality. The BET specific surface area, SBET, was determined from N2 adsorption-desorption isotherms obtained with a fully automated micropore and surface-area analyzer. The point of zero charge, pHₚzc, was measured by mass titration. 50 mL aliquots of 0.01 M NaCl were adjusted to initial pH between 2 and 12 with 0.01 M HCl or NaOH, and the final pH was recorded after equilibration. Phosphorus concentration in the effluent was quantified spectrophotometrically with the ammonium molybdate method (GB 11893-89). Absorbance values were converted to concentrations with the calibration curve given in Eq. (1). All data processing and graphical presentations were performed with Microsoft Excel 2016 and Origin 8.0.1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$y=33.99x - 0.0252,{R^2}=0.9999$$\end{document} The BTC is an essential tool for evaluating the performance of packing in a fixed-bed column studies. It is constructed by plotting the effluent phosphorus concentration against operating time29. Breakthrough is deemed to occur once the effluent phosphorus concentration reaches the Class A discharge limit of 0.5 mg/L specified in China’s standard (GB 18918 − 2002) for primary effluents, the corresponding elapsed time is defined as the breakthrough time. Saturation is assumed when the effluent phosphorus concentration attains the Class B limit of 1.0 mg/L under the same standard, and this instant is recorded as the saturation time. Upon saturation, the column run is terminated, and the BTC is generated from the collected data. Subsequently, the Yoon-Nelson (Y-N) and Adams-Bohart (A-B) fixed-bed reaction kinetic models were applied to fit the experimental BTCs30, and their linearized forms are given in Eqs. (2) and (3), respectively.2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\ln (\frac{{{C_t}}}{{{C_0} - {C_t}}})={K_{YN}}t - {K_{YN}}\tau$$\end{document}3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\ln (\frac{{{C_t}}}{{{C_0}}})={K_{AB}}{C_0}t - {K_{AB}}{N_0}(\frac{Z}{{{U_0}}})$$\end{document} Where C0 is the initial phosphorus concentration, mg/L; Ct is the phosphorus concentration in the effluent at time t, mg/L; t is the operating time of the fixed-bed column, min; KYN is the Y-N model rate constant, min− 1; τ is the time required for 50% packing breakthrough, min; KAB is the A-B model rate constant, L/(mg·min); N0 is the saturation concentration, mg/L; Z is the height of the fixed-bed column, cm; U0 is the ratio of the volumetric flow rate to the crosssection from the beginning to the end of breakthrough; R2 is the correlation coefficient. In addition, the root-mean-square error (RMSE) and the chi-square statistic (χ²), were employed to quantify the agreement between the experimental data and the model predictions31, and their expressions are given in Eqs. (4) and (6), respectively.4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$RMSE\sqrt{\frac{1}{n-p} \sum_{i=1}^n (X_{\exp} -X_{cal})^2_i}$$\end{document}5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$X^2 = \sum_{i=1}^n \big[ \frac{(X_{\rm exp}- X_{cal})}{X_{cal}} \big]_i$$\end{document} Where Xexp is denote the empirical value using the model; Xcal is calculated value using the model; n is the number of experimental data; p is the number of the parameters within model equations. Results and discussions Characterization As shown in Fig. 1a, the surface of the pretreated SW-ceramsite is smooth, lustrous, and topographically irregular, with numerous micropores uniformly distributed across it, indicating that a well-developed porous structure. In contrast, Fig. 1b revealed that the spent SW-ceramsite is covered by a substantial layer of sediment, most surface micropores are occluded, leading to marked decreases in both specific surface area and pore volume. EDS spectra further showed that the pretreated SW-ceramsite contains Mg, Al, Ca, and Fe at weight percentages of 2.25%, 3.76%, 14.60%, and 2.79%, respectively (seen Fig. 1a). Although these elements remain detectable on the spent SW-ceramsite (seen Fig. 1b), but their contents notable decline by 52.89%, 52.66%, 55.41%, and 37.28%, respectively, confirming that their participation in the phosphorus removal process within the fixed-bed column packed with SW-ceramsite. Table 1 showed that the pretreated SW-ceramsite exhibited a SBET of 4.18 × 104 cm2/g, a pore volume of 3.05 × 103 cm3/g, a bulk density of 1.37 g/cm3 and a void fraction of 69.8%. These values confirm its large surface area, well-developed porosity, and rough surface morphology. The pHpzc value of the pretreated SW-ceramsite was 9.83, indicating a high affinity for anionic species in solution32. After saturation, the above-mentioned parameters of spent SW-ceramsite declined significantly. In particular, the pHpzc, SBET, pore volume and void fraction decreased by 3.56%, 39.95%, 28.85%, and 16.62%, respectively (Table 1). The pronounced reductions in specific surface area and pore volume are attributed to extensive sediment deposition on the external surface and within the pore network, consistent with the SEM image in Fig. 1b. Table 1Textural properties of samples.SamplepHpzcSBET(cm2/g)Pore volume(cm3/g)Bulk density(g/cm3)Void fraction(%)Pretreated SW-ceramsite9.834.18 × 1043.05 × 1031.3769.8Spent SW-ceramsite9.482.51 × 1042.17 × 1032.2258.2 Table 1Textural properties of samples.SamplepHpzcSBET(cm2/g)Pore volume(cm3/g)Bulk density(g/cm3)Void fraction(%)Pretreated SW-ceramsite9.834.18 × 1043.05 × 1031.3769.8Spent SW-ceramsite9.482.51 × 1042.17 × 1032.2258.2 Textural properties of samples. Influence of initial pH on BTC The effect of initial pH on the BTC was examined by changing the initial pH (3, 5, 7 and 9), and the results were observed in Fig. 2a. According to the findings from the figure, elevating the initial pH produced progressively steeper BTCs, with both breakthrough and saturation points shifting to shorter operating times. When the initial pH rose from 5 to 9, breakthrough time declined from 80 h to 55 h, and saturation time decreased from 155 h to 90 h, decreased by 31.25% and 41.94%, respectively, demonstrating that phosphorus removal is more efficient under acidic conditions. Notably, performance at initial pH 5 marginally exceeded that at initial pH 3. Therefore, considering both treatment efficiency and practical engineering costs, initial pH 5 was selected as the optimum influent pH for subsequent phosphorus removal from aqueous solution in a fixed-bed column packed with SW-ceramsite. Fig. 2Influence of initial pH (a) on BTCs of the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column with initial concentration = 5 mg/L, reaction temperature = 323 K and initial flowrate = 40 mL/min. Influence of initial concentration (b) on BTCs of the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column with initial pH = 5, constantreaction temperature = 323 K and initial flowrate = 40 mL/min. Influence of reaction temperature (c) on BTCs of the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column with initial concentration = 5 mg/L, initial pH = 5 and initial flowrate = 40 mL/min. Influence of initial flowrate (d) on BTCs of the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column with initial pH = 5, initial concentration = 5 mg/L, and reaction temperature = 323 K. Influence of initial pH (a) on BTCs of the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column with initial concentration = 5 mg/L, reaction temperature = 323 K and initial flowrate = 40 mL/min. Influence of initial concentration (b) on BTCs of the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column with initial pH = 5, constantreaction temperature = 323 K and initial flowrate = 40 mL/min. Influence of reaction temperature (c) on BTCs of the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column with initial concentration = 5 mg/L, initial pH = 5 and initial flowrate = 40 mL/min. Influence of initial flowrate (d) on BTCs of the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column with initial pH = 5, initial concentration = 5 mg/L, and reaction temperature = 323 K. The ability of SW-ceramsite to remove phosphorus from water is mainly attributable to, a well-developed pore structure (Fig. 1a) and high BET specific surface area (Table 1), abundant surface functional groups33,34, and a mineral phase rich in reactive oxides35, such as MgO, CaO, Al₂O₃, Fe2O3. The SW-ceramsite adsorb aqueous HPO₄²⁻/PO₄³⁻ onto its external or internal surfaces, where the anions subsequently precipitate with Ca²⁺, Al³⁺, Fe³⁺ or Mg²⁺ to form Ca₃(PO₄)₂, AlPO₄, FePO₄, Mg₃(PO₄)₂ etc., as shown in Eqs. (6)–(9). Initial pH governs both the speciation of phosphate and the surface charge of the SW-ceramsite36, as well as the activity and supersaturation of the participating ions37. At initial pH < 5, phosphorus is present predominantly as H₂PO₄⁻. Fe²⁺/Fe³⁺ and PO₄³⁻ remain below supersaturation, so precipitation is suppressed. Between initial pH 5 and 7, HPO₄²⁻ dominates, whereas PO₄³⁻ becomes the main species above initial pH 7. Under these conditions FePO₄ or Fe₃(PO₄)₂ forms readily, rapidly depleting dissolved Fe and halting further crystallization. Excessive acidity or alkalinity lowers phosphorus removal. Strong acid promotes dissolution of Fe-bearing phases, such as Eqs. (10)–(11), releasing Ca²⁺, Al³⁺, Fe²⁺/Fe³⁺ before they can react with phosphate. Conversely, high initial pH precipitates these cations as Al(OH)₃ or Fe(OH)₃, as shown in Eqs. (12)–(13), suppressing the reactions in Eqs. (6)–(9), and thereby impeding phosphorus removal.6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F{e^{3+}}+P{O_4}^{{3 - }} \to FeP{O_4} \downarrow$$\end{document}7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A{l^{3+}}+P{O_4}^{{3 - }} \to AlP{O_4} \downarrow$$\end{document}8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F{e^{2+}}+P{O_4}^{{3 - }} \to F{e_3}{(P{O_4})_2} \downarrow$$\end{document}9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C{a^{2+}}+P{O_4}^{{3 - }} \to C{a_3}{(P{O_4})_2} \downarrow$$\end{document}10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F{e_2}{O_3}+6{H^+} \to 2F{e^{3+}}+3{H_2}O$$\end{document}11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$FeO+2{H^+} \to F{e^{2+}}+{H_2}O$$\end{document}12\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F{e^{3+}}+O{H^ - } \to Fe{(OH)_3} \downarrow$$\end{document}13\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A{l^{3+}}+O{H^ - } \to Al{(OH)_3} \downarrow$$\end{document} Influence of initial concentration on BTC The effect of initial concentration on the BTC was examined by changing the initial phosphorus concentration (5, 10, 15 and 30 mg/L), while all other parameters were held constant, and the results were shown in Fig. 2b. It can be clearly seen from the figure that lowering the initial concentration produced progressively flatter BTCs and markedly postponed both breakthrough and saturation. As the initial concentration decreased from 30 to 5 mg/L, the breakthrough and saturation times increased from 50 h to 85 h to 80 h and 155 h, respectively, increased by 60% and 82.35%, respectively. The extended service life at 5 mg/L indicates more efficient utilization of the sorption capacity. Consequently, 5 mg/L was selected as the optimum initial concentration for subsequent phosphorus removal from aqueous solution in a fixed-bed column packed with SW-ceramsite. The main reason may be attributed to the higher initial concentration gradient, which results in faster transport. Phosphorus in wastewater can be indirectly removed through co-precipitation and the enmeshment of Fe(OH)2 or Fe(OH)3. Fe2+ or Fe3+ ions, either on the surface of SW-ceramsite or dissolved in the solution, readily react with PO43− to form precipitate such as Fe3(PO4)2 or FePO438. Additionally, at high initial phosphorus concentrations, available adsorption sites are rapidly occupied39. These findings suggests that the novel SW-ceramsite can serve as an effective packing for phosphorus removal from aqueous solution in a fixed-bed column. Hence, It can be inferred that this new type of fixed-bed column packed with SW-ceramsite may well suited for treating low-concentration phosphorus-containing wastewater. Influence of reaction temperature on BTC The influence of reaction temperature (313, 323, 333 and 343 K) on the BTC was examined, while all other parameters were held constant, and the results were summarized in Fig. 2c. The figure indicated that raising the reaction temperature progressively flattened the BTC and shifted both the breakthrough and saturation points to the right. When the reaction temperature increased from 323 to 333 K, the breakthrough and saturation times rose from 80 to 85 h and 155 to 160 h, respectively, and only increased by 6.25% and 3.23%, respectively. Such minor increments indicate that reaction temperature elevation within the studied range does not significantly enhance phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column. The limited improvement can be explained by two competing factors. On the one hand, higher reaction temperature intensifies the surface reactivity of SW-ceramsite and increases the collision frequency between PO₄³⁻/HPO₄²⁻ and cations (such as Mg²⁺, Ca²⁺, Fe²⁺/Fe³⁺), thus promoting both adsorption and precipitation process. On the other hand, exothermic adsorption and complexation reactions become less favorable thermodynamically as reaction temperature rises, while side reactions and structural changes in the medium may also occur40. Considering the marginal performance gain and the additional energy cost, 323 K was selected as the optimum reaction temperature for subsequent phosphorus removal from aqueous solution in a fixed-bed column packed with SW-ceramsite. Influence of initial flowrate on BTC The effect of initial flowrate on the BTC was examined by changing the influence flow rate (30, 40, 50 and 60 mL/min), and the results were listed in Fig. 2d. It can be clearly seen from the figure that an increase in initial flowrate produced a steeper BTC and shifted both the breakthrough and saturation points to the left. Raising the initial flowrate from 30 to 60 mL/min shortened the breakthrough time from 85 to 60 h and the saturation time from 160 to 125 h, marking reductions of over 29.41% and 21.88%, respectively. The phenomenon is attributed to shorter residence time and intensified competition for adsorption sites41. At elevated initial flowrate, PO₄³⁻/HPO₄²⁻ ions have insufficient contact time to establish equilibrium. Consequently, the solution exits the column before the SW-ceramsite packing is fully utilized, leading to earlier breakthrough42. These findings demonstrate that SW-ceramsite employed as a fixed-bed column packing exhibits high efficacy for phosphorus removal from aqueous solution. Taking treatment efficiency and effluent-quality compliance into account, an initial flowrate of 30 mL/min was selected as the optimum operating condition for subsequent phosphorus removal from aqueous solution in a fixed-bed column packed with SW-ceramsite. In summary, Fig. 2 indicates that the optimal operating conditions for phosphorus removal from aqueous solution in a fixed-bed column packed with SW-ceramsite are as follows: initial pH value is 5, initial concentration is 5 mg/L, reaction temperature is 323 K, and initial flowrate is 40 mL/min. Under these operating conditions, breakthrough and saturation times occurred at 80 h and 155 h, respectively. Considering both packing preparation and wastewater-treatment costs, this phosphorus removal performance from aqueous solution in the fixed-bed column packed with SW-ceramsite was much better than that of achieved in the research of Meina et al.43. Mathematical modeling of BTC To better understand the operational stability of the fixed-bed column packed with SW-ceramsite for the removal of phosphorus from aqueous solution, Y-N and A-B fixed-bed reaction-kinetic models were used to fit the BTC data linearly (seen Fig. 2). The Y-N model, a relatively simple theoretical approach, describes the breakthrough behavior of phosphorus on SW-ceramsite. A straight line was obtained when ln[Ct/(C0-Ct)] was plotted versus t, as detailed in Fig. 3. The A-B model, which is based on surface-adsorption theory and assumes that equilibrium is not attained instantaneously, predicts the relationship between initial concentration and operating time during continuous adsorption in a fixed-bed column. Here, a straight line was obtained by plotting ln(Ct/C0) against t, as shown in Fig. 3. The corresponding model parameters, such as KYN, KAB, τ, N0, and R2, together with the RMSE and χ2, were calculated, and the results were illustrated in Table 2. Table 2Kinetic parameters for the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column.Y-N modelA-B modelKYN×10 − 2(min− 1)τ(min)R2RMSEχ2×10 − 3KAB×10 − 3[L/(mg·min)]N0(mg/L)R2RMSEχ2×10 − 3Initial pH35.02144.810.9350.0150.249.7297.190.9250.0250.6853.52172.930.9230.0270.796.7360.290.9140.0361.4273.43146.840.8490.10112.106.4312.870.8300.12017.3593.67119.910.8630.0878.776.8257.430.8400.11014.40Initial concentration(mg/L)53.52172.930.9230.0270.796.7360.290.9140.0361.42103.41179.610.8140.13622.723.3740.480.8050.14526.12153.18181.130.8210.12920.272.11095.540.8140.13622.72304.89142.400.8470.10312.531.61739.850.8450.10513.05Reaction temperature(K)3133.06161.920.7770.17338.525.8340.260.7560.19449.783233.52172.930.9230.0270.796.7360.290.9140.0361.423333.76169.890.8910.0593.917.1356.950.8810.0695.403433.65176.930.9070.0432.046.9371.100.9010.0492.66initial flowrate(mL/min)303.27180.420.9180.0321.126.2283.160.9080.0421.94403.52172.930.9230.0270.796.7360.290.9140.0361.42503.41156.520.8710.0797.176.4413.650.8560.09410.32603.48144.350.8310.11917.046.5461.040.8120.13823.45 Table 2Kinetic parameters for the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column.Y-N modelA-B modelKYN×10 − 2(min− 1)τ(min)R2RMSEχ2×10 − 3KAB×10 − 3[L/(mg·min)]N0(mg/L)R2RMSEχ2×10 − 3Initial pH35.02144.810.9350.0150.249.7297.190.9250.0250.6853.52172.930.9230.0270.796.7360.290.9140.0361.4273.43146.840.8490.10112.106.4312.870.8300.12017.3593.67119.910.8630.0878.776.8257.430.8400.11014.40Initial concentration(mg/L)53.52172.930.9230.0270.796.7360.290.9140.0361.42103.41179.610.8140.13622.723.3740.480.8050.14526.12153.18181.130.8210.12920.272.11095.540.8140.13622.72304.89142.400.8470.10312.531.61739.850.8450.10513.05Reaction temperature(K)3133.06161.920.7770.17338.525.8340.260.7560.19449.783233.52172.930.9230.0270.796.7360.290.9140.0361.423333.76169.890.8910.0593.917.1356.950.8810.0695.403433.65176.930.9070.0432.046.9371.100.9010.0492.66initial flowrate(mL/min)303.27180.420.9180.0321.126.2283.160.9080.0421.94403.52172.930.9230.0270.796.7360.290.9140.0361.42503.41156.520.8710.0797.176.4413.650.8560.09410.32603.48144.350.8310.11917.046.5461.040.8120.13823.45 Kinetic parameters for the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column. Initial concentration (mg/L) Reaction temperature (K) initial flowrate (mL/min) As shown in Fig. 3a; Table 2, the KYN decreased with increasing initial pH, indicating that a higher initial pH led to a lower reaction rate. By contrast, the τ value remained essentially unchanged. R² gradually decreased, whereas RMSE and χ² increased, implying that linearity of the Y-N model fit became progressively poorer. According to the findings from Fig. 3b and c; Table 2, it was found that variations of initial concentration or reaction temperature had no significant effect on the Y-N model parameters. Finally, it was also found from Fig. 3d; Table 2 that raising the initial flowrate had little influence on KYN, but τ decreased markedly, R² followed the same downward trend, while RMSE and χ² rose steadily, again indicating a gradual loss of linear correlation. As shown in Fig. 4a; Table 2, KAB decreased with increasing initial pH, indicating that a higher initial pH lowered the reaction rate. The N0 value remained essentially unchanged over the initial pH range from 3 to 9. R² gradually declined, whereas RMSE and χ² increased, implying that the linearity of the A-B model fit deteriorated. As seen from Fig. 4b; Table 2, it was obtained that as the initial concentration increased, the KAB value decreased. When the initial concentration was increased from 5 to 30 mg/L, the N0 value increased from 360.29 to 1739.85 mg/L, increased by nearly 4.83 times, demonstrating that the fixed-bed column packed with SW-ceramsite has an enhanced removal capacity, and the packing layer is more resistant to breakthrough44. According to Fig. 4c; Table 2, both KAB and N0 rose with increasing reaction temperature. Finally, according to the results obtained by Fig. 4d; Table 2, varying the initial flowrate had little effect on KAB, whereas N0 increased steadily. In conclusion, according to Figs. 3 and 4; Table 2, on the basis of R², RMSE and χ², the Y-N model describes the phosphorus removal process in a fixed-bed column packed with SW-ceramsite more accurately than that of the A-B model. This implies that the removal rate is directly proportional to both the residual capacity of the SW-ceramsite packing and the amount of phosphorus already adsorbed45. Fig. 3Linearized fitting results of Y-N fixed-bed reaction kinetic model for the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column. Linearized fitting results of Y-N fixed-bed reaction kinetic model for the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column. Fig. 4Linearized fitting results of A-B fixed-bed reaction kinetic model for the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column. Linearized fitting results of A-B fixed-bed reaction kinetic model for the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column. Grey relational analysis The BTC describes how the phosphorus concentration in the effluent evolves with operating time when an aqueous solution is treated in a fixed-bed column packed with SW-ceramsite46. It is obtained by plotting the ratio of effluent to initial concentrations (Ct/C0) against time t47,48. In this study, the GRA method was employed to quantify the influence of four operating variables, included initial pH, initial concentration, reaction temperature and initial flowrate, on the BTC. Taking initial pH as an example, the step-by-step correlation analysis between the BTC and initial pH was carried out as follows: Step 1. Establish the evaluation index matrix Based on the experimental data of the BTCs listed in Fig. 2, Ct/C0 with different operating times (20, 40, 60, and 80 h) at initial pH values of 3, 5, 7, and 9 were formed into a matrix according to a 4 × 4 sequence as the following Eq. (14).14\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(X')=\left[ {\begin{array}{*{20}{c}} {0.0010}&{0.0036}&{0.0252}&{0.0666} \\ {0.0034}&{0.0070}&{0.0200}&{0.0994} \\ {0.0078}&{0.0398}&{0.0866}&{0.1284} \\ {0.0308}&{0.0704}&{0.1166}&{0.1672} \end{array}} \right]$$\end{document} Step 2. Normalize the index data In this study, the initialization method was employed to normalize the index data, that is: X = X’ij/X’i1(i = 1, 2, 3, ···; j = 2, 3, 4, ···), and the calculation results were shown in Eq. (15).15\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(X)=\left[ {\begin{array}{*{20}{c}} 1&{3.6000}&{25.2000}&{66.6000} \\ 1&{2.0600}&{5.8800}&{29.2400} \\ 1&{5.1000}&{11.1000}&{16.4600} \\ 1&{2.2900}&{3.7900}&{5.4300} \end{array}} \right]$$\end{document} Step 3. Determine the reference data column The reference data column was a standard idealized data, and its reference value can be selected based on the optimal value. The average of each column in matrix (X) was taken as the reference series in this study, as shown in Eq. (16).16\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${X_0}=\left\{ {\begin{array}{*{20}{c}} 1&{3.2625}&{11.4925}&{29.4325} \end{array}} \right\}$$\end{document} Step 4. Assess the absolute difference between target metric sequence and reference sequence That is, |x0(k)-xi(k)| (k = 1, ···, m; i = 1, ···, n), where n is the number of evaluated objects). The absolute difference calculated by |X0-Xi| was shown in Eq. (17).17\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(\Delta )=\left[ {\begin{array}{*{20}{c}} 0&{0.3375}&{13.7075}&{37.1675} \\ 0&{1.2025}&{5.6125}&{0.1925} \\ 0&{1.8375}&{0.3925}&{12.9725} \\ 0&{0.9725}&{7.7025}&{24.0025} \end{array}} \right]$$\end{document} Step 5. Find the extreme value The minimum and maximum values obtained in this study were presented in Eqs. (18) and (19), respectively.18\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathop {\hbox{min} }\limits_{{i=1}}^{n} \mathop {\hbox{min} }\limits_{{k=1}}^{m} \left| {{x_0}(k) - {x_i}(k)} \right|=0$$\end{document}19\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathop {\hbox{max} }\limits_{{i=1}}^{n} \mathop {\hbox{max} }\limits_{{k=1}}^{m} \left| {{x_0}(k) - {x_i}(k)} \right|=37.1675$$\end{document} Step 6. Calculate the correlation coefficient According to Eq. (20), the correlation number between each comparison sequence and the corresponding element in the reference sequence was calculated, and the results were depicted in Table 3. Table 3Calculation results of GRA correlation coefficient and its comprehensive evaluation.Relation coefficientComprehensive evaluation20 min40 min60 min80 minRelation degreeAssumptionGRAInitial pH310.98220.57550.33330.7228WeakWeak510.93930.76800.98970.9243MediumStrongest710.91000.97930.58890.8696StrongStrong910.95030.70700.43640.7734StrongestMediumInitial concentration(mg/L)510.72900.56460.34430.6595WeakMedium1010.70360.36700.33330.6010MediumWeak1510.86380.56430.30410.6831StrongStrong3010.90230.84660.38220.7828StrongestStrongestReaction temperature(K)31310.69770.60180.33330.6582WeakWeak32310.84500.84100.51420.8001MediumMedium33310.89860.98680.42660.8280StrongStrong34310.87950.68670.77560.8355StrongestStrongestInitial flowrate(mL/min)3010.78620.56920.84940.8319WeakStrong4010.79620.63340.33330.6907MediumMedium5010.99030.90220.56260.8638StrongStrongest6010.65030.44900.48890.6471StrongestWeak 20\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\zeta (k)=\frac{{\mathop {\hbox{min} }\limits_{i} \mathop {\hbox{min} }\limits_{k} \left| {{x_0}(k) - {x_i}(k)} \right|+\rho \mathop {\hbox{max} }\limits_{i} \mathop {\hbox{max} }\limits_{k} \left| {{x_0}(k) - {x_i}(k)} \right|}}{{\left| {{x_0}(k) - {x_i}(k)} \right|+\rho \mathop {\hbox{max} }\limits_{i} \mathop {\hbox{max} }\limits_{k} \left| {{x_0}(k) - {x_i}(k)} \right|}}$$\end{document} Table 3Calculation results of GRA correlation coefficient and its comprehensive evaluation.Relation coefficientComprehensive evaluation20 min40 min60 min80 minRelation degreeAssumptionGRAInitial pH310.98220.57550.33330.7228WeakWeak510.93930.76800.98970.9243MediumStrongest710.91000.97930.58890.8696StrongStrong910.95030.70700.43640.7734StrongestMediumInitial concentration(mg/L)510.72900.56460.34430.6595WeakMedium1010.70360.36700.33330.6010MediumWeak1510.86380.56430.30410.6831StrongStrong3010.90230.84660.38220.7828StrongestStrongestReaction temperature(K)31310.69770.60180.33330.6582WeakWeak32310.84500.84100.51420.8001MediumMedium33310.89860.98680.42660.8280StrongStrong34310.87950.68670.77560.8355StrongestStrongestInitial flowrate(mL/min)3010.78620.56920.84940.8319WeakStrong4010.79620.63340.33330.6907MediumMedium5010.99030.90220.56260.8638StrongStrongest6010.65030.44900.48890.6471StrongestWeak Calculation results of GRA correlation coefficient and its comprehensive evaluation. Initial concentration (mg/L) Reaction temperature (K) Initial flowrate (mL/min) Where k = 1, ···, m. ρ is the resolution coefficient, ranging from 0 to 1. If ρ is smaller, it indicates that there are greater differences between correlation coefficients and clearer divisions. In this study, ρ is set at 0.5. Step 7. Calculation of grey relational degree According to Eq. (21), the correlation between initial pH and BTC data was calculated, and the results were detailed in Table 3.21\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${r_i}=\frac{1}{m}\sum\limits_{{k=1}}^{m} {{\zeta _i}(k)}$$\end{document} Subsequently, following the aforementioned procedures and drawing upon the experimental data of BTCs presented in Fig. 2, we calculated the correlation between initial concentrations (5, 10, 15, and 30 mg/L), reaction temperatures (313, 323, 333, and 343 K), and initial flowrate (30, 40, 50, and 60 mL/min) with different operating times (20, 40, 60, and 80 h) in terms of BTC data, and the findings were illustrated in Table 3. It can be clearly seen from Table 3 that the Ct/C0 ratio first rose and then fell as the initial pH increased. This is because either very low or very high initial pH alters the concentrations of Mg²⁺, Ca²⁺, Fe³⁺ and Fe²⁺ in the system, which in turn affects the Ct/C0 profile. The Ct/C0 value increases continuously with the increase of initial concentration, showing that higher initial concentrations improve removal efficiency and strengthen the correlation between Ct/C0 ratio and operating time. A similar upward trend was observed with reaction temperature. The higher the reaction temperature, the higher the Ct/C0 ratio, again indicating better phosphorus removal and a stronger Ct/C0 correlation. By contrast, as the initial flowrate rose, the Ct/C0 ratio decreased, then increased, and finally decreased once more. These fluctuations arise because initial flowrate governs the mass-transfer rate within the system, thereby influencing the relationship between Ct/C0 and operating time. In brief, the assumed effects of initial concentration and reaction temperature on the BTC agreed well with the GRA results. Similarly, the assumed initial pH effect agreed reasonably well, whereas the initial flowrate assumption showed the poorest agreement. Overall, the GRA findings provide a sound basis for explaining how these variables influence factors on phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column, elucidating its operating mechanism and facilitating wider application. Regeneration performance When a fixed-bed column packed with SW-ceramsite is used to treated phosphorus-containing wastewater, regeneration and recyclability are critical parameters for its industrial application. Therefore, under the optimum operating conditions of an initial pH was 5, an initial phosphorus concentration was 5 mg/L, a reaction temperature was 323 K, and an initial flowrate was 40 mL/min, we therefore examined how repeated regeneration affects on the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column, with breakthrough and saturation times were used as evaluation indicator, and the results are presented in Fig. 5. Both breakthrough and saturation times gradually decreased as the number of regeneration cycles rose. The main reason is the small particle diameter of SW-ceramsite packing, which causes unavoidable losses during elution48. Specifically, after eight regeneration cycles, the breakthrough time fell from 80 h to 55 h, decreased by about 30%. The saturation time showed a similar tendency, which reduced from 155 h to 135 h, only reduced by 12.9%. These results demonstrate that SW-ceramsite retains a strong regeneration capacity and is well suited for use as a fixed-bed column packing. Fig. 5Effect of regeneration cycle on the phosphorus removal performance from aqueous solution by SW-ceramsite in a fixed-bed column under the optimal operating conditions of initial pH = 5, initial concentration = 5 mg/L, reaction temperature = 323 K, and initial flowrate = 40 mL/min. Effect of regeneration cycle on the phosphorus removal performance from aqueous solution by SW-ceramsite in a fixed-bed column under the optimal operating conditions of initial pH = 5, initial concentration = 5 mg/L, reaction temperature = 323 K, and initial flowrate = 40 mL/min. Conclusion The SW-ceramsite can be served as a fixed-bed column packing to remove phosphorus from aqueous solution, with notable effect. This removal process is influenced by many factors, in particular, with the decreasing in initial pH value, initial concentration and initial flowrate, or increasing in reaction temperature, the breakthrough curve gradually becomes flatten, along with the breakthrough and saturation points shift rightward, and both breakthrough and saturation times increase. Furthermore, according to R2, RMSE and χ2 values shows that the removal process is described better by the Y-N model than by the A-B fixed-bed reaction kinetic model. The impact of the influencing factors on the BTC of the phosphorus removal from aqueous solution by SW-ceramsite in a fixed-bed column can be interpreted as follows: The assumed effects of initial concentration and reaction temperature agree best with the GRA results, the effect of initial pH shows moderate agreement, and the effect of initial flowrate shows the poorest agreement. These findings not only clarify the operating mechanism of the fixed-bed column packed with SW-ceramsite, but also accelerate its scale-up to full-scale application. SW-ceramiste has been proven to be a cost-effective, recyclable fixed-bed packing for phosphorus remove from aqueous solution, which not only mitigates phosphorus pollution and its threats to the water environment and human health, but also provides a practical route for the comprehensive, resource-oriented reuse of SW-ceramiste in water-treatment applications.