A novel Fe–La-doped hierarchical porous silica magnetic adsorbent for phosphate removal

Abstract

Many rivers and lakes contain excess phosphate, which would cause a series of environmental problems. A novel Fe–La modified magnetic hierarchical porous silica was synthesized by an impregnation method to adsorb phosphate in aquatic ecology. Scanning electron microscopy and transmission electron microscopy suggested the structure of the adsorbent was a core–shell structure. Fourier transform infrared spectroscopy and low-angle powder X-ray diffraction indicated that adsorbents were successfully modified by Fe oxide and La oxide. After loading Fe oxide and La oxide on the material, N₂ adsorption–desorption studies showed that the specific surface area was 52.4 m² g⁻¹. Moreover, the adsorbents possessed the desired property of a magnet and easy separation from the solution by testing with a vibrating sample magnetometer. The adsorption process was highly pH-dependent and the appropriate adsorption pH range was about 3.0–6.0. And the resulting adsorption isotherms of the sorbent was better represented by the Langmuir model than the Freundlich model, and the adsorption capacity of the adsorbent was as high as 71.99 mg P per g, which was much higher than most other adsorbents (most La-doped adsorbents' adsorption capacities are under 25 mg P per g). Furthermore, adsorption kinetics studies showed that the adsorption fitted the pseudo-second order model well. Also a high selectivity of phosphate was observed in the coexisting anions system. The proposed adsorbent demonstrated a higher degree of reusability in a cycle test.

---

1. Introduction

In aquatic ecology, one of the environmental concerns caused by excess phosphate is called eutrophication. It will deplete oxygen, affect aquatic life forms, and deteriorate water quality.¹ Even though phosphate exists extensively in the natural environment, such as natural water, wastewater, sludge, igneous and sedimentary rocks,² it is not the major reason for this environmental problem. Human activities, which pour phosphate-rich waters into the aquatic environment, such as domestic sewage, agricultural use of fertilizers and industrial effluents, are the main reasons for eutrophication.³ Therefore, it is meaningful to get rid of excess phosphate from high concentration waste waters before pouring into municipal and industrial effluent water streams. The most common ways for removing phosphorus are physical, chemical and biological methods.^4–8 In general, chemical precipitation uses solid waste from other manufacture processes, so its cost maybe the lowest, while it would produce waste disposal and unwanted chemicals at the same time. Biological treatments utilize activated sludge during the nutrient removal process. It is efficient for removal of nitrite, as for phosphate is relatively low.⁹ What is more, biological treatments still need a primary biological structure and it is uneasy to carry on.¹⁰ Compared with these methods above, solid phase adsorption is considered to be more practical, economic and efficient. Therefore, the work is needed to develop an effective adsorbent material for recovery of phosphates today.

In recent years, there were considerable efforts devoted into nanostructured materials with numerous functional properties.¹¹ It had been widely recognized that the creation of hollow and mesopores was an effective way to improve the quantity of the available active center. This kind of materials has strong appeal owing to its extensive coverage in many applications, including adsorbents, luminescent, fluorescent, enzyme carriers, fuel cells, catalysts support, drug delivery, and so on.^12–17 In general, mesoporous materials are fabricated via soft templates or hard templates methods individually.^18–21 The soft-template method was carried out to create varieties of morphology of the target material. However, the pore volume and surface area are low by soft-template method,^22,23 and the quantity of the available active center is still not alotted,²⁴ and the mesostructure also has poor stability and ordering. As for the hard-templated method, it was reported to create larger surface area and higher pore volume,²⁵ which would load higher mass of oxide.²⁶ However, the hard-templated method made the composite only with single morphology. Clearly, the soft-templated and hard-templated method have their own intrinsic drawbacks and merits. Therefore, an ideal mesostructure called hollow and mesoporous material which both have larger surface area, pore volume and stable structure should be carried out by the dual templates method. Since cellulose can be expressed from enzyme rosettes into 3–5 nm diameter fibers and it later can aggregate to 20 nm diameter microfibers by biosynthesis.²⁷ And hexadecyl trimethyl ammonium bromide (CTAB) micelles could attach to nanocrystalline cellulose (NCC),²⁸ because the negatively charged NCC could attract CTAB micelles on its surface easily. Then the tetraethyl orthosilicate (TEOS) was added to fabricate the composite of silica/NCC by the sol–gel reaction. After the composite was calcined, we obtain the hierarchical porous silica nanotubes which process two kinds of size pores and large specific surface area.

Despite the material processes a high pore volume and a stable structure, it is not efficient for adsorbing phosphorus. Improving the capacity of preference adsorption is also a key for phosphorus removal. In this study, we introduced lanthanum as an inorganic material into hierarchical porous, to develop an inorganic hybrid adsorbent for anions. La is one of rare earth elements and it is considered to be an environmentally, friendly and effective adsorbent for phosphorus.^29–31 Several other rare earth elements' metal oxides also have high adsorption capacities for phosphate, but these rare earth elements are mostly costly, while La is comparatively abundant in the earth's crust and less expensive.²⁹ And that La could extract form polishing powder which was mainly used as glass polishing compounds, precipitating and decoloring agents and also used in ceramic, electronics and catalysts industries, while thousands of tons used polishing powder ultimately end up in the land fill as solid wastes. La-Containing materials have attracted considerable attention for removing phosphate recently.^29,32–40 Therefore, loading lanthanum oxide on hierarchical porous material can greatly increase the capacity of preference adsorption of phosphorus. Furthermore, we also introduce Fe oxides into mesoporous because of its strong binding and high capacity of phosphate. There are many previous researches proved that Fe-loaded materials are able to remove phosphate effectually.⁴¹ So it will be more economical if the adsorbent contains other lower cost elements such as Fe oxide.^42,43 And the La based composites with Fe oxide were prepared for the first time. So we prepare a series adsorbents which loaded different amount of Fe oxide and La oxide to research its performances and properties. On the other hand, adding Fe oxide such as Fe₃O₄ nanoparticles into composites can endow magnetic property on adsorbent, which can make composites separate from the solution more easily and conveniently than the centrifugation or filtration.

In this study, a series of novel La–Fe-doped hierarchical porous silica magnetic adsorbents were obtained by the impregnation method. The characteristics of the adsorbent were analyzed by the vibrating sample magnetometer (VSM), the scanning electron microscopy (SEM), the transmission electron microscope (TEM), the Fourier transform infrared spectroscopy (FTIR) and so on. A series of batch experiments was performed to study the adsorption isotherm, adsorption kinetics, effects of pH and co-anion and regeneration of the adsorbents. The recyclability tests were also carried out to evaluate the reusability of the adsorbent. Through these examinations and experiments, the mechanisms of phosphate removal would be better understand.

2. Materials and methods

2.1. Materials

All reagents and solvents, unless otherwise stated, were of analytical standard, purchased from Sinopharm Chemical Reagent Co., Ltd. and used without further purification. Degreasing cotton provided by Sinopharm Chemical Regent Co. Ltd. (Shanghai, China) was used as a raw material for producing nanocrystalline cellulose (NCC). All other chemicals, including tetraethoxysilane (TEOS, >95%), sulphuric acid (H₂SO₄, 95–98%), cetyltrimethyl ammonium bromide (CTAB, 98%), ethanol (>99.7%) and aqueous ammonia (25% w/w) were purchased from Everbright Chemical Inc. (Nanjing, China). All solutions were prepared using DI water.

2.2. Preparation of nanocrystalline cellulose (NCC)

The Beck-Candanedo's method was employed to prepare NCC.⁴⁴ The process as follows: degreasing cotton (5 g) and H₂SO₄ (250 mL, 64 wt%) were added into a 500 mL three-neck flask including a magnetic stirrer. Then this mixture was hydrolyzed in water bath at 45 °C for 30 min under constant stirring. 4000 mL deionized water was added into suspension to stop hydrolysis. After that, the suspension was centrifuged to remove excess sulphuric acid several times until the suspension became turbid. In order to make the suspension pH reduce to neutral, it was packed into a dialysis bag to dialyze in pure water for several days.

2.3. Fabrication of hierarchical porous material

A typical coating procedure was performed as follows: 30 mg CTAB and 400 μL aqueous ammonia were initially added into the mixture of H₂O and ethanol which hold in a 250 mL three-neck flask. The amounts of H₂O and ethanol were 40 mL and 20 mL respectively. Then 4 mL NCC dispersion (0.5 wt%) was charged in the solution. The mixture was stirred at the speed of 250 rpm under 30 °C water bath. After stirring for 10 min, 100 μL TEOS was slowly added into the flask. After two hours' reaction, the suspension was subjected to centrifugation (12 000 rpm) to separate the NCC particles which coated with silica. After that, pure water and ethanol was added into the centrifugal tubes to disperse the sediment. And the centrifugal tubes was subjected another centrifugation. In order to remove the excess chemicals and byproducts, the separation step was repeated for three times. Then the composites of silica/NCC was obtain after the sediment was dried at 50 °C for hours. The hierarchical porous materials were finally obtained through using a muffle furnace to calcine the composites of silica/NCC at 600 °C for 6 h to remove the CTAB and NCC. The organic moieties of silica/NCC composites were released as water and CO₂. For convenience the resultant microporous and mesoporous nanotube was denoted as MNT.

2.4. Synthesis of adsorbents

La and Fe oxides were then loaded into MNT using the ethanol evaporation method. 0.5 g MNT was charged into 100 mL of ethanol solution in which contained a certain amounts of La(NO₃)₃·6H₂O and Fe(NO₃)₃·12H₂O. The solution was stirred at 60 °C for 20 h, and then dried at 80 °C to remove the ethanol completely. Then the product was calcined at 700 °C for 6 h. These obtained samples were denoted as MNT-x, where x refers to the mole ratios of Fe/Si in the initial impregnation solution, and the mole ratios of La/Si was 1/5 in all initial impregnation solution. Herein, four different La–Fe-impregnated MNT samples, MNT-0, MNT-1/10, MNT-1/5 and MNT-1/2.5, were prepared, corresponding to the theoretical Fe/Si mole ratios of 0, 1/10, 1/5, and 1/2.5, respectively.

2.5. Analytical methods

FT-IR spectra (4000–400 cm⁻¹) were collected on a Nicolet NEXUS-470 FT-IR apparatus (U.S.A.) using KBr disks. Transmission electron microscopy (TEM) was performed using a JEOL JEM-2100 (HR) at an accelerating voltage of 200 kV. The morphologies of materials were obtained by scanning electron microscope (SEM, JEOL, JSM-7001F). Low-angle powder X-ray diffraction (XRD) were test by D8 ADVANCE Germany Bruker. The surface area was determined by N₂ adsorption N₂ adsorption–desorption tests on to the sorbent using a D-35614 Assiar Instruments (Pfeiffer vacuum, Germany). Samples were outgassed for 6 hours at 100 °C (parent silica) or for 6 h at 60 °C (functionalized sorbents) prior to the N₂ adsorption analysis, which was carried out at liquid nitrogen temperature (−196 °C). Surface area was obtained by a multi-point analysis of the volume of nitrogen adsorbed as a function of relative pressure.

2.6. Adsorption experiments

A batch of various concentration phosphate solutions was obtained by adding different amount of anhydrous potassium dihydrogen phosphate (KH₂PO₄) into pure water. Adsorption processes were performed as follows: the adsorbent dosage added into the adsorption performance test was 2 g L⁻¹. Then the centrifuge tubes were placed in a thermostatic shaker bath at 25 °C for certain time. The last step of the adsorption procedure was that the solution was filtrated by using 0.45 μm membrane syringe filter, and the filtrate was analyzed for adsorption efficiency.

The adsorption kinetics experiment was performed at initial phosphate concentration of 100 mg L⁻¹, and the initial solution pH value was adjusted at 5.0. The dosage of phosphate solution was 10 mL each centrifuge tube. During the experiment, each sample was taken out at a certain time interval then to filtrate for determining phosphate concentrations. The phosphate concentrations were detected by ICP-OES. The amount of phosphate adsorbed at time t (q_t) was calculated by eqn (1),

where C₀ (mg L⁻¹) and C_t (mg L⁻¹) stand for the initial and final phosphate concentrations in solution before and after time t, respectively. V (L) is the volume of solution and W (g) is the mass of adsorbent.

And the kinetic data were fitted by pseudo-first-order and pseudo-second-order models, which are described in eqn (2) and (3):^45,46

Pseudo first-order:

q_t = q_e − q_e^−K₁t (2)

Pseudo second-order:

where q_t (mg g⁻¹) and q_e (mg g⁻¹) are the amount of phosphate adsorbed at time t, and at equilibrium, respectively. K₁ (L min⁻¹) and K₂ (g mg⁻¹ min⁻¹) are the rate constants of the pseudo-first-order and pseudo-second-order models, respectively.

In the experiment of pH effect, the initial phosphate concentration was 100 mg L⁻¹ and the dosage of phosphate solution was 10 mL each centrifuge tube. The initial pH of each samples were adjusted by adding amount of HCl and/or NH₃·H₂O. The centrifuge tubes were placed in a shaker for 24 h; after that, the solutions were taken out and filtrated for the measurement of phosphate concentrations.

The adsorption isotherm experiments were performed at initial pH of 5.0. The initial concentration of phosphate solution was varied from 10 to 100 mg L⁻¹. Then 20 mg adsorbent was charged into 10 mL phosphate solutions of different initial concentrations. Other process were the same as the experiment for pH effect. The final equilibrium adsorption capacity (q_e mg g⁻¹) was calculated by eqn (4),

where C₀ (mg L⁻¹) and C_e (mg L⁻¹) respectively, stand for the initial and equilibrium concentrations in phosphate solution. V (L) is the volume of solution and W (g) is the weight of adsorbent.

The equilibrium data can be fitted by the Langmuir and Freundlich isotherm models by eqn (5) and (6), respectively.^47,48

Langmuir model:

Freundlich model:

q_e = K_FC_e^1/n (6)

where C_e (mg L⁻¹) represents the equilibrium phosphate concentration in solution; q_e (mg g⁻¹) is the equilibrium phosphate adsorption capacity; q_m (mg g⁻¹) represents the maximum adsorption capacity of the materials and K_F (mg g⁻¹) is the sorbent adsorption capacity direction constant in the Freundlich isotherm model; K_L (L mg⁻¹) is the affinity constant in Langmuir isotherm model and 1/n indicates favorable adsorption conditions in Freundlich isotherm model.

In the study of competitive factors on adsorption, 0.1 M NaF, NaCl, Na₂SO₄, NaHCO₃ and NaNO₃ were respectively added into solutions that had phosphate concentration of 100 mg L⁻¹. The initial solution pH was 5.0; the dosage of phosphate solution was 10 mL each centrifuge tube. Other process were the same as the experiment for pH effect.

In the study of the ions release, 1 g MNT-x (x = 1/2.5, 1/5, 1/10) were added into 500 mL DI water separately, then each beaker was shook in 28 °C water bath kettle for 24 h. The Fe and La concentrations in water were detected by ICP-OES.

In the study of regeneration, the used MNT-1/5 was separated from solution by centrifugation or magnetic separation and then washed by pure water for five times. 0.1 M NaOH was chosen as elution solution. The iron oxides would dissolve in solution. Then the used adsorbent was reloaded Fe and La oxides on its surface as previous method. After that, the adsorption experiment was the same as previous procedure. The regeneration experiment was conducted for five cycles to evaluate the adsorbent reusability.

3. Results and discussion

3.1. Structural and morphological characterization

The SEM images of MNT and MNT-x (x = 1/2.5, 1/5, 1/10) were presented in the Fig. 1. The TEM images of MNT and MNT-x (x = 0, 1/2.5, 1/5, 1/10) were presented in the Fig. 2. The SEM and TEM images exhibited an obvious core–shell structure. From TEM images, the channel made by NCC template could be clearly observed, and the outer diameters of nanotubes were ca. 120 nm, which was consistent with the diameters of core–shell structured. The thickness of nanotube structure was determined to be ca. 8.0 nm. The average diameter of the inner core was ca. 10 nm. The image also revealed that the silica layer maintained its shape well even after the step of calcination at 700 °C, it may due to the covalently connection by Si–O bonds, so it did not collapse. The agglomerates of silica nanotubes were viewed in the TEM image as well. And there were some dispersive and dense black spots could be seen. We evaluated that these spots were caused by Fe and La oxides particles which stick into the pores of silica nanotubes. To further confirm components of adsorbents, EDX was used to detect the atomic species, and the O, Si, La and Fe elements were detected, and the Si, La, Fe weight ratio of MNT-x were shown in Table 1.

Fig. 1 Representative SEM images of the top surface of MNT (a), MNT-1/10 (b), MNT-1/5 (c), and MNT-1/2.5 (d).

Fig. 2 Representative TEM image of MNT (a), MNT-1/10 (b), MNT-1/5 (c), and MNT-1/2.5 (d), respectively under different magnifications.

Table 1 Structure characteristics of the MNT and MNT-x (x = 1/10, 1/5, and 1/2.5) samples

Sample	SBET (m^2 g^−1)	d (nm)	Vtotal (cm^3 g^−1)	La content^a (wt%)	Fe content^a (wt%)
a The mean value of La and Fe content (wt%) was calculated from EDX by averaging 3 portions of each sample.
MNT	324.1	3.4	0.28	0	0
MNT-1/10	52.4	6.4	0.08	34.27	4.52
MNT-1/5	49.9	7.3	0.09	30.64	8.07
MNT-1/2.5	27.8	7.3	0.11	27.36	17.24

---

The low-angle powder X-ray diffraction patterns of MNT, MNT-1/10, MNT-1/5, and MNT-1/2.5 were presented in Fig. 3. From the XRD pattern, some discernible differences could be found in the textural properties between these samples. Well-resolved peaks of MNT, MNT-1/10, MNT-1/5, and MNT-1/2.5 were appeared at 0.535° and 1.65°. It revealed that there are two kind diameter pores exist in these samples. Compared to the XRD patterns of MNT, the MNT-1/10's, MNT-1/5's and MNT-1/2.5's reflection intensity decreases at 0.535° and 1.65°. And after impregnating larger doses of La oxides, the peaks of MNT-1/2.5 were lower than MNT-1/5 and MNT-1/10, which suggesting the declining periodicity of the hierarchical porous structure. That may be caused by the stress on the siliceous framework which anchoring Fe and La oxides.^49,50

Fig. 3 Low-angle powder X-ray diffraction of MNT, MNT-1/10, MNT-1/5, and MNT-1/2.5 respectively.

The Brunauer–Emmett–Teller (BET) was used to characterize the MNT and MNT-x (x = 1/2.5, 1/5, 1/10) and the nitrogen adsorption–desorption isothermal plot of the mesoporous silica nanotubes was showed in Fig. 4. The adsorption–desorption isothermal plot would be ascribed to the type IV, and the mesopores shape was cylindrical. This was ascribed to the capillary condensation of nitrogen in the confined mesopores, and there was an apparent hysteresis loop at relative pressures (P/P₀) between 0.6 and 1.0, indicating that mesopores apparently exist. After calcinating the MNT, the BET surface area of silica mesoporous nanotube increased to 324 m² g⁻¹, and the total pore volume was up to 0.276 cm³ g⁻¹. Contrasted with the mesoporous silica spheres that made by Wang et al.,¹³ its BET specific surface area was 204.6 m² g⁻¹ and total pore volume was 0.264 cm³ g⁻¹. It revealed that the NCC and CTAB contributed a significant effect on the total pore volume and the surface area of the material. As showed in the Table 1, after loading Fe and La oxides on MNT's surface, the BET specific surface area decreased with the increased amount of Fe. Once loading Fe and La, the BET specific surface area reduced sharply from 324 m² g⁻¹ to 52.4 m² g⁻¹. Owing to the modification, many mesopores would be blocked by the Fe and La oxides.

Fig. 4 N₂ adsorption–desorption isotherms for the MNT (a), MNT-1/10 (b), MNT-1/5 (c), and MNT-1/2.5 (d).

Fig. 5 showed the BJH pore size distribution plots which were contingent on the adsorption branches. The plot revealed that two average diameters of bimodal sharp peaks were found at ca. 2 nm and 10 nm. The latter sharp peak which correspond to tubular cores was originated from the NCC template. The former sharp peak was originated from the template of CTAB, which sticked on the NCC template.⁵¹ And this value also near to the mesopores size (ca. 2 nm) of silica nanotubes, which was created by Zhang et al.,⁵² or by Scheel et al.,¹⁴ and they both utilize the CTAB only. We deduced that the employment of the CTAB made this silica material obtained the ca. 2 nm mesopores. Meanwhile, V_total is reduced with the increased of the amount of La and Fe oxides (Table 1). The variation in structure characteristics could be attributed to the impregnation of La and Fe oxides inside the mesopores, which narrowed the pore sizes and reduced the pore volumes. So, the results suggest that La and Fe oxides was successfully doped into the mesopores of MNT.

Fig. 5 Pore size distributions for the MNT (a), MNT-1/2.5 (b), MNT-1/5 (c), and MNT-1/10 (d).

3.2. FT-IR analysis

The FT-IR, which served to characterize the surface silanols and hydrogen-bonded hydroxyl groups was shown in Fig. 6. The adsorption bands at 1080 cm⁻¹ corresponded to the Si–O–Si asymmetric stretching vibrations. The broad band 3420 cm⁻¹ in the hydroxyl region could be assigned to the Si–OH stretching vibration. And in the spectra of MNT-1/2.5 (a), MNT-1/5 (b) and MNT-1/10 (c), there was a small peak at 1384 cm⁻¹, which was due to remaining nitrate after calcination (La(NO₃)₃ and Fe(NO₃)₃ was used as La and Fe sources).

Fig. 6 FT-IR spectra for the MNT-1/2.5 (a), MNT-1/5 (b), MNT-1/10 (c) and MNT (d), respectively.

3.3. Magnetic property studies

The magnetic properties of MNT-x (x = 1/2.5, 1/5, 1/10) which loaded different amount of Fe oxides were studied by vibrating sample magnetometer (VSM) at 300 K. From the Fig. 7(a), the hysteresis loops revealed that the MNT-x displayed a low coercivity with no distinct hysteresis, so it confirmed the existence of superparamagnetism in the hybrid systems. Owing to the superparamagnetism, the adsorbent exhibited a sensitive response to a magnetic field. After removing the magnetic field, the adsorbent could re-disperse quickly by slight shaking. The reversible process was a superiority that could be applied in many fields. Furthermore, with loading larger doses of Fe oxide, the magnetization value of the MNT-x increased from 5.1 to 15.9 to 26.8 emu g⁻¹, and it could further confirm that more Fe oxides coated on MNT.

Fig. 7 Magnetic hysteresis loop (a) of MNT-1/2.5, MNT-1/5, and MNT-1/10 respectively, and pictures (b) of particle suspension in water and separation using a magnet.

3.4. Effect of initial pH on the adsorption

The effect of the phosphate solutions' pH on phosphate adsorption of MNT-1/5 was investigated, as shown in Fig. 8. The investigative pH range was from 2.0 to 10.0, and the process of adsorbing phosphate was intensely dependent on the solution's pH value. From the Fig. 8, a high fluctuating of phosphate capacities was observed, between 51.80 mg P per g and 65.4 mg P per g, and it suggesting that MNT-1/5 could be employed in a wide pH range. In addition, when the pH was increased from 2.0 to 6.0, the phosphate adsorption capacities declined from 65.4 mg P per g to 51.8 mg P per g; when the pH was increased from 6.0 to 10.0, the phosphate adsorption capacities sharply declined 64.7%, from 51.8 mg P per g to 18.3 mg P per g. As we known that the ionic species of phosphate existed in solution strongly depended on pH.⁵³ The different ionic forms of phosphate in solution depending on pH were showed as eqn (7):where pK₁ = 2.12, pK₂ = 7.20, and pK₃ = 12.33, respectively.

Fig. 8 Effect of pH on the phosphate adsorption on MNT-1/5.

At pH ≤ 2.0, the predominant species of phosphate is H₃PO₄ which was weakly adsorbed to the sites of the materials, besides highly acidic solution affects the stability of the sorbents. Based on the eqn (7), when the initial pH was in the range of 2.13–7.20, the main ionic species of phosphate were H₂PO₄⁻ and HPO₄²⁻, and when 2.13 < pH < 5.34, the adsorption capacity of MNT-1/5 changed slightly. The high adsorption capacity possibly was result of that protonated adsorption sites had electrostatic attraction to H₂PO₄⁻ and HPO₄²⁻. When 5.34 < pH < 7.20, the phosphate adsorption capacity was decreasing. At this time the adsorption mechanism can be explained according to the ligand-exchange mechanism.⁵⁴ And at high pH value (8.0–10.0), adsorbent is unfavorable for the adsorption of phosphate, and it causes a remarkable decrease as shown in Fig. 8. Addition, at pH 2, usually metal oxides, and especially iron oxides are soluble. Hence most likely the removal is a result of precipitation of phosphate with dissolved iron when the pH at 2. Therefore, the appropriate solution pH range for adsorption of phosphate should be 3.0–6.0.

3.5. Adsorption kinetics studies

Fig. 9 showed the adsorption kinetic study by using MNT-x (x = 0, 1/10, 1/5, and 1/2.5) in an initial phosphate concentration solution of 100 mg P per L. From the Fig. 9, it showed that MNT-1/2.5's adsorption capacity (q_t) reached to 12.77 mg P per g in the first hour. Then, the adsorption capacities increased continuously and reached equilibrium in 60 h, and the q_t was 50.13 mg P per g. In order to further understand the process of adsorption, the kinetic curves were fitted by the pseudo-first-order model and pseudo-second-order model, and their correlation coefficients (R²) and corresponding parameters are listed in Table 2. As it shown, the pseudo-first-order model and pseudo-second-order model fitted the kinetic curves both well. However, the pseudo-second-order model had the higher relative correlation coefficient than the pseudo-first-order model, indicated that the phosphate adsorption process of MNT-x was more likely governed by chemisorption. In addition, even though the MNT-1/2.5 had the biggest adsorption capacity, its initial velocity of adsorbing phosphate was a litter lower than MNT-0, MNT-1/10 and MNT-1/5. The MNT-0, MNT-1/10 and MNT-1/5 had the same condition with MNT-1/2.5. That is to say, with the higher amount of loading Fe, the adsorbent had the higher adsorption capacity, but its initial velocity of adsorbing phosphate would be lower.

Fig. 9 Kinetic data, pseudo-first-order kinetic modelling (a) and pseudo-second-order kinetic modelling (b) for the adsorption of phosphate on MNT-0, MNT-1/10, MNT-1/5 and MNT-1/2.5.

Table 2 Parameters of the pseudo-first-order and pseudo-second-order

Samples	Pseudo-first-order			Pseudo-second-order
	q (mg g^−1)	KL (L mg^−1)	R^2	q (mg g^−1)	KL	R^2
MNT-0	36.74	0.013	0.950	34.16	0.009	0.988
MNT-1/10	43.87	0.007	0.949	40.47	0.004	0.977
MNT-1/5	50.13	0.004	0.975	45.20	0.003	0.994
MNT-1/2.5	62.86	0.002	0.982	54.98	0.002	0.988

---

3.6. Adsorption isotherms studies

Fig. 10 showed the phosphate adsorption equilibriums of MNT-x (x = 0, 1/10, 1/5, and 1/2.5) fitted by the Langmuir isotherm and Freundlich isotherm at various initial concentrations. Each corresponding isotherm parameter was summarized in Table 3. The MNT-0's phosphate adsorption capacity was measured to be 46.35 mg P per g. In contrast, the MNT-x (x = 1/10, 1/5, and 1/2.5) samples showed different adsorption capacities, and the adsorption capacities was proportional to the x. The Langmuir model and Freundlich model were both suitable for describing the phosphorus adsorption isotherms. While the Langmuir model gave a better fit than the Freundlich model on the basis of their correlation coefficients (R² > 0.95, Table 3), showing that the adsorption process took place on the homogeneous surface of the adsorbent and it was a monolayer adsorption. And the maximum adsorption capacities of the adsorbent which estimated by the Langmuir model (q_m, Table 3) were continuously increased with the increasing of the x from 1/10 to 1/2.5. The Langmuir parameter q_m for MNT-1/2.5 was 71.9923 mg P per g, which was extraordinary greater than MNT-1/5 (57.97 mg P per g) and MNT-1/10 (37.37 mg P per g). This would be ascribed to more Fe oxides doped into the MNT-1/2.5, so MNT-1/2.5 could provide more active sites for adsorbing phosphate. In particular, MNT-1/2.5 showed superior maximum adsorption capacity (q_m = 71.9923 mg P per g) when compared with other La-loaded mesoporous silica adsorbents in the document, as shown in Table 4.^55–58 This was probably owing to the great surface area of the adsorbent MNT-x, and combined with its hierarchical porous structure; it largely ensures the high accessibility of Fe and La active sites to adsorb phosphate anions. Due to the higher Fe loaded on MNT-1/5 (8.07%) than MNT-1/10 (4.25%, Table 1), under the same conditions, MNT-1/5's adsorption capacity (57.97 mg P per g) was greater than MNT-1/10's (37.37 mg P per g). In addition, the molar ratio of adsorbed phosphate versus Fe, which has been commonly employed to evaluate the efficiency of Fe usage,^55,59 was only 4.17 in MNT-1/2.5. This value was much lower than 7.18 for MNT-1/5, indicating that MNT-1/2.5 had a low efficiency of Fe usage. We deduced that this matter was caused by mesoporous blockage, which had proven by the BET analysis of MNT-1/2.5, so a great deal of Fe and La inside the channels was unable to contact phosphate during adsorption.

Fig. 10 Equilibrium data, Langmuir modelling (a) and Freundlich modelling (b) for the adsorption of phosphate on MNT-x (x = 0, 1/10, 1/5 and 1/2.5).

Table 3 Parameters of the Langmuir and Freundlich isotherm

Samples	Langmuir			Freundlich
	q (mg g^−1)	KL (L mg^−1)	R^2	n	KF	R^2
MNT-0	46.35	0.47	0.987	21.1	0.18	0.815
MNT-1/10	37.37	1.63	0.945	21.7	0.13	0.794
MNT-1/5	57.97	0.17	0.948	16.3	0.28	0.922
MNT-1/2.5	71.99	0.08	0.989	12.5	0.38	0.959

---

Table 4 Comparison of phosphate adsorption capacities between MNT-1/5 and other lanthanum-modified mesoporous adsorbents in the literature

Adsorbents	pH	Temperature (°C)	Adsorption capacity (mg P per g)	Ref.
La40MCM-41	—	24	23.8	55
La-Doped MCM-41	—	24	7.2	56
La-Coordinated amino-functionalized MCM-41	7.0	24	17.7	57
La-Doped mesoporous SiO2	7.3	—	23.1	58
MNT-1/5	5.5	24	57.9	Present work

---

3.7. Effect of co-existing anions on the adsorption

Fig. 11 showed the effects of coexisting anions on the phosphate adsorption capacity of MNT-1/5, including Cl⁻, F⁻, NO³⁻, CO₃²⁻, and SO₄²⁻. Compared with the phosphate adsorption capacity (q_e = 55.78 mg P per g) of MNT-1/5 in the absence of any influential anions, 0.01 M coexisting anion in the solution only slightly affect its adsorption capacity. But CO₃²⁻ had the most significant impact, it could be explained from the above experiment (the effect of initial solution pH). When Cl⁻, F⁻, NO³⁻, and SO₄²⁻ were introduced separately, the initial solution's pH before adsorption progress was approximately at 5.3. However, after added 0.01 M CO₃²⁻, the pH was up to 10.4. Due to the appearance of a large amount of OH⁻, which would compete with PO₄³⁻ for combining the adsorption active sites and preventing the ligand-exchange mechanism, so it lead to a lower phosphate uptake significantly. Nevertheless, the decreasing degree of q_e was much less than preview experiment under the same pH which showed in Fig. 8. This might be owing to the form of La₂(CO₃)₃ and Fe₂(CO₃)₃, which had been approved that forms had an obvious function of binding phosphate. As a result, the Fig. 11 indicated that MNT-1/5 possessed a high selectivity toward phosphate anions over these selected competitive anions except CO₃²⁻.

Fig. 11 Effect of co-anions on the phosphate adsorption on MNT-1/5.

3.8. Ion release tests and the analysis of environmental safety

The Fe and La ions release from the prepared nanocomposite after 24 h were showed in Fig. 12. As a whole, the amounts of Fe and La ions which released in DI water were very low. And its concentration was basically in direct proportion to La and Fe weight ratio of MNT-x. The amounts of the released ions were less than 1% of the MNT-x (x = 1/10, 1/5 and 1/2.5). It might indicate that La and Fe oxides strongly doped on MNT. As for environmental safety, silicon substrate nanomaterials have good biocompatibility and it is non-toxic. And as everyone knows, the metal oxides nanoparticles would resulted in uncertainties regarding their environmental impacts. The LD₅₀ of E. coli after exposure to the La₂O₃ and Fe₂O₃ were 456.9 mg L⁻¹ and 638.3 mg L⁻¹ respectively.⁶⁰ While the amount of metal oxides nanoparticles which released by the 1 g MNT-x (x = 1/10, 1/5 and 1/2.5) in 500 mL DI water were all under 2 mg. This value was far below the value of LD₅₀. So we could evaluate the Fe–La modified adsorbent has low cytotoxicity (Schemes 1 and 2).

Fig. 12 Ions release of MNT-x (x = 1/10, 1/5 and 1/2.5).

Scheme 1 Schematic diagram of the Fe–La modified magnetic hierarchical porous silica adsorbent adsorbing phosphorus and recycling from solution.

Scheme 2 Synthesis procedures of the Fe–La modified magnetic hierarchical porous silica adsorbent.

3.9. Recyclability tests

The recyclability of our as-prepared adsorbent MNT-1/5 was also explored, shown in Fig. 13. The regenerated MNT-1/5 showed a high phosphate adsorption of 50.10 mg P per g for the fifth cycle, which capacity is very close to the fresh MNT-1/5 (55.78 mg P per g). After five recyclability tests, the regeneration efficiency was still higher than 90%. It indicated the excellent reusability of the adsorbent, and the adsorbent could be served as supports in manufacturing highly efficient phosphate adsorbents.

Fig. 13 Regeneration of MNT-1/5 over 5 cycles. Error bars represent the standard error of the mean for measurements carried out in triplicate.

4. Conclusion

A novel Fe–La modified magnetic hierarchical porous silica adsorbent was successfully developed by impregnation method. The adsorbent was a core–shell structure with hierarchical pores. The FT-IR, low-angle powder X-ray diffraction and N₂ adsorption isotherms analyzed that the adsorbents were successfully modified by Fe and La and the adsorbents processed high surface area. The adsorption kinetics experiment showed that the adsorption equilibrium more fitted by pseudo-second-order model; about 50% of equilibrium adsorption capacity could be reached in the first 2 h. And the Langmuir model gave a better fit for describing the phosphorus adsorption isotherms. The appropriate solution pH range for adsorption of phosphate was 3.0–6.0. And the MNT-1/2.5's maximum adsorption capacities was as high as 71.99 mg P per g, which exceed many reported adsorbents. The presence of Cl⁻, F⁻, NO³⁻ and SO₄²⁻ almost had no influence on the phosphate adsorption, while the existence of CO₃²⁻ performed some influence on the phosphate removal. In addition, the regeneration experiments confirmed that the novel adsorbent had good reusability for the phosphate removal. These adsorption experiments revealed that the uptake of phosphate involved a complicated mechanism including an ion exchange between the Fe and La ion with phosphate in the solution, and we should research it further in the future. In general, it could be concluded that the Fe–La modified magnetic hierarchical porous silica adsorbent could be a promising material in the phosphate removal.

References

1. W. Chouyyok, R. J. Wiacek, K. Pattamakomsan, T. Sangvanich, R. M. Grudzien, G. E. Fryxell and W. Yantasee, Environ. Sci. Technol., 2010, 44, 3073–3078.
2. O. Akpor, M. Momba and J. Okonkwo, Pak. J. Biol. Sci., 2007, 10, 4008–4014.
3. L.-l. Bao, L. Dong, X.-k. Li, R.-x. Huang, J. Zhang, L. Yang and G.-q. Xia, J. Environ. Sci., 2007, 19, 391–395.
4. M. A. Ramos, W. Yan, X.-q. Li, B. E. Koel and W.-x. Zhang, J. Phys. Chem. C, 2009, 113, 14591–14594.
5. W. Yan, M. A. Ramos, B. E. Koel and W.-x. Zhang, J. Phys. Chem. C, 2012, 116, 5303–5311.
6. W. Yan, R. Vasic, A. I. Frenkel and B. E. Koel, Environ. Sci. Technol., 2012, 46, 7018–7026.
7. J. Cao, D. Elliott and W.-x. Zhang, J. Nanopart. Res., 2005, 7, 499–506.
8. X.-q. Li, D. W. Elliott and W.-x. Zhang, Crit. Rev. Solid State Mater. Sci., 2006, 31, 111–122.
9. S. Yeoman, T. Stephenson, J. Lester and R. Perry, Environ. Pollut., 1988, 49, 183–233.
10. C. P. Leo, W. Chai, A. W. Mohammad, Y. Qi, A. Hoedley and S.-P. Chai, Water Sci. Technol., 2011, 64, 199–205.
11. F. Caruso, Adv. Mater., 2001, 13, 11–22.
12. S. Gruber, A. Gottschlich, H. Scheel, C. Aresipathi, I. Bannat, C. Zollfrank and M. Wark, Phys. Status Solidi B, 2010, 247, 2401–2411.
13. F. Wang, Y. Tang, B. Zhang, B. Chen and Y. Wang, J. Colloid Interface Sci., 2012, 386, 129–134.
14. H. Scheel, C. Zollfrank and P. Greil, J. Mater. Res., 2009, 24, 1709–1715.
15. D.-J. Guo and S.-K. Cui, J. Colloid Interface Sci., 2009, 340, 53–57.
16. Q.-G. Xiao, X. Tao, J.-P. Zhang and J.-F. Chen, J. Mol. Catal. B: Enzym., 2006, 42, 14–19.
17. J.-F. Chen, H.-M. Ding, J.-X. Wang and L. Shao, Biomaterials, 2004, 25, 723–727.
18. F. Caruso, R. A. Caruso and H. Möhwald, Science, 1998, 282, 1111–1114.
19. N. Kato, T. Ishii and S. Koumoto, Langmuir, 2010, 26, 14334–14344.
20. Y. Zhao, L. N. Lin, Y. Lu, S. F. Chen, L. Dong and S. H. Yu, Adv. Mater., 2010, 22, 5255–5259.
21. S. O. Obare, N. R. Jana and C. J. Murphy, Nano Lett., 2001, 1, 601–603.
22. H.-Q. Li, R.-L. Liu, D.-Y. Zhao and Y.-Y. Xia, Carbon, 2007, 45, 2628–2635.
23. Y. Lv, F. Zhang, Y. Dou, Y. Zhai, J. Wang, H. Liu, Y. Xia, B. Tu and D. Zhao, J. Mater. Chem., 2012, 22, 93–99.
24. J. Hu, M. Noked, E. Gillette, Z. Gui and S. B. Lee, Carbon, 2015, 93, 903–914.
25. K. Xia, Q. Gao, J. Jiang and J. Hu, Carbon, 2008, 46, 1718–1726.
26. K. P. Prasad, D. S. Dhawale, S. Joseph, C. Anand, M. A. Wahab, A. Mano, C. Sathish, V. V. Balasubramanian, T. Sivakumar and A. Vinu, Microporous Mesoporous Mater., 2013, 172, 77–86.
27. L. A. Lucia and M. A. Hubbe, BioResources, 2008, 3, 668–669.
28. Y. Habibi, L. A. Lucia and O. J. Rojas, Chem. Rev., 2010, 110, 3479–3500.
29. E. W. Shin, K. Karthikeyan and M. A. Tshabalala, Environ. Sci. Technol., 2005, 39, 6273–6279.
30. S. Tokunaga, S. A. Wasay and S.-W. Park, Water Sci. Technol., 1997, 35, 71–78.
31. S. Tokunaga, S. Yokoyama and S. Wasay, Water Environ. Res., 1999, 71, 299–306.
32. P. Melnyk, J. Norman and M. Wasserlauf, in Proc. Rare Earth Res. Conf., NTIS, Springfield, Cleveland, OH, USA, 11th edn, 1974, pp. 4–13.
33. H. L. Recht, M. Ghassemi and E. V. Kleber, Precipitation of phosphates from water and waste water using lanthanum salts, Proceedings of the 5th International Water Pollution Research, Pergamon, 1970, pp. 1–17.
34. J. Liu, Q. Zhou, J. Chen, L. Zhang and N. Chang, Chem. Eng. J., 2013, 215, 859–867.
35. S. Tian, P. Jiang, P. Ning and Y. Su, Chem. Eng. J., 2009, 151, 141–148.
36. B. M. Spears, S. Meis, A. Anderson and M. Kellou, Sci. Total Environ., 2013, 442, 103–110.
37. F. Haghseresht, S. Wang and D. Do, Appl. Clay Sci., 2009, 46, 369–375.
38. K. Reitzel, F. Ø. Andersen, S. Egemose and H. S. Jensen, Water Res., 2013, 47, 2787–2796.
39. M. Zamparas, A. Gianni, P. Stathi, Y. Deligiannakis and I. Zacharias, Appl. Clay Sci., 2012, 62, 101–106.
40. M. Zamparas, M. Drosos, Y. Georgiou, Y. Deligiannakis and I. Zacharias, Chem. Eng. J., 2013, 225, 43–51.
41. X.-P. Liao, Y. Ding, B. Wang and B. Shi, Ind. Eng. Chem. Res., 2006, 45, 3896–3901.
42. Y.-M. Zheng, S.-F. Lim and J. P. Chen, J. Colloid Interface Sci., 2009, 338, 22–29.
43. S.-F. Lim, Y.-M. Zheng and J. P. Chen, Langmuir, 2009, 25, 4973–4978.
44. S. Beck-Candanedo, M. Roman and D. G. Gray, Biomacromolecules, 2005, 6, 1048–1054.
45. Y. S. Ho and G. McKay, Process Biochem., 1999, 34, 451–465.
46. Y.-S. Ho, J. Hazard. Mater., 2006, 136, 681–689.
47. H. Ou, Q. Chen, J. Pan, Y. Zhang, Y. Huang and X. Qi, J. Hazard. Mater., 2015, 289, 28–37.
48. X. Zheng, C. Wang, J. Dai, W. Shi and Y. Yan, J. Mater. Chem. A, 2015, 3, 10327–10335.
49. S. Shylesh and A. Singh, J. Catal., 2004, 228, 333–346.
50. S. M. Rivera-Jiménez and A. J. Hernández-Maldonado, Microporous Mesoporous Mater., 2008, 116, 246–252.
51. Y. Le, M. Pu and J.-f. Chen, J. Non-Cryst. Solids, 2007, 353, 164–169.
52. Y. Zhang, X. Liu and J. Huang, ACS Appl. Mater. Interfaces, 2011, 3, 3272–3275.
53. D. Perrin, Buffers for pH and metal ion control, Springer Science & Business Media, 2012.
54. F. Haghseresht, S. Wang and D. D. Do, Appl. Clay Sci., 2009, 46, 369–375.
55. J. Yang, L. Zhou, L. Zhao, H. Zhang, J. Yin, G. Wei, K. Qian, Y. Wang and C. Yu, J. Mater. Chem., 2011, 21, 2489–2494.
56. J. Zhang, Z. Shen, W. Shan, Z. Chen, Z. Mei, Y. Lei and W. Wang, J. Environ. Sci., 2010, 22, 507–511.
57. J. Zhang, Z. Shen, W. Shan, Z. Mei and W. Wang, J. Hazard. Mater., 2011, 186, 76–83.
58. E. Ou, J. Zhou, S. Mao, J. Wang, F. Xia and L. Min, Colloids Surf., A, 2007, 308, 47–53.
59. J. Yang, P. Yuan, H.-Y. Chen, J. Zou, Z. Yuan and C. Yu, J. Mater. Chem., 2012, 22, 9983–9990.
60. X. Hu, S. Cook, P. Wang and H.-m. Hwang, Sci. Total Environ., 2009, 407, 3070–3072.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra17279b

---