All chemicals and reagents were of analytical grade and used as received unless otherwise stated. CNTs-COOH was purchased from Times nano, Chengdu Organic Chemicals Co., Ltd., China. APTMS was purchased from Shanghai Macklin Biochemical Co., Ltd., China. MPTMS was purchased from Qufu Wanda Chemical Co., Ltd., China. Mercuric nitrate (Hg(NO₃)₂·1/2H₂O) were purchased from Sinopharm Chemical Reagent Co., Ltd., China.

Bifunctional PSQ/CNTs magnetic composites were synthesized by sol-gel method. The first step is magnetization; prepared magnetic Fe₃O₄ was embedded in the original material to give CNTs-COOH@Fe₃O₄, and then CNTs-COOH@Fe₃O₄ was coupled with APTMS via amidation to give the intermediate CNTs-APTMS@Fe₃O₄ with siloxane introduced on the surface. In the final step, CNTs-APTMS@Fe₃O₄ was decorated with variable proportions of APTMS and MPTMS to give the bifunctional PSQ/CNTs magnetic composites. The reaction steps as shown in Figure 1 and the monofunctional PSQ/CNTs magnetic composites were also prepared in the same way for comparison. The specific experimental methods and dosages were described in Ref. Xu et al. (2021). The structures and surface morphologies of PSQ/CNTs magnetic composites were confirmed by FT-IR, SEM, VSM, XRD, XPS, and BET analysis, and all the characterization analyses were also detailed explicated in Ref. Xu et al. (2021).

For measuring the adsorption properties of the composites to different metal ions, 10 mg composite was added into 40 ml 100 mg L⁻¹ Hg(NO₃)₂·1/2H₂O, AgNO₃, Pb(NO₃)₂, Cu(NO₃)₂·3H₂O, and Ni(NO₃)₂·6H₂O, respectively. At the same time, the unmodified CNTs-COOH@Fe₃O₄ and the intermediate CNTs-APTMS@Fe₃O₄ were also measured in the same way. Then the mixtures were shaken at 25°C for 24 h. The initial and equilibrium concentrations of metal ions were measured by atomic absorption spectrometry (AAS). The adsorption capacity was calculated by the equation as follows: q e = ( C 0 − C e ) V W (1) where q _e (mmol g⁻¹) is the equilibrium adsorption capacity; C ₀ and C _e (mmol L⁻¹) represent the initial and equilibrium concentration of metal ion, respectively; V (L) is the volume of solution; and W (g) is the weight of the composite adsorbent.

The effect of pH on the uptake of composites for Hg(II) was performed by the following method: 10 mg composite was added into 40 ml 100 mg L⁻¹ Hg(II) solution, the different solution pH values were adjusted with dilute HNO₃ and NaOH aqueous, after that the mixtures shaken at 25°C for 24 h. The initial and equilibrium concentrations were measured by AAS and the adsorption capacities were calculated by Equation 1 as well.

A total of 10 mg composite was added into 40 ml Hg(NO₃)₂·1/2H₂O solution with different initial concentrations, adjusted to optimal pH, and the mixtures were then shaken at 25°C for 24 h. Simultaneously, the same method was used to determine the adsorption capacities of the composites at 15 and 35°C and calculated with Equation 1.

A total of 10 mg composite was added into 100 ml 50 mg L⁻¹ Hg(II) solution, adjusted to optimal pH, and then shaken at 25°C. The concentrations of Hg(II) were determined at regular time intervals and calculated by the following equation as follows: q e = ∑ ( C 0 − C t ) V t ​ W (2) where q _e (mmol g⁻¹) is the equilibrium adsorption capacity; C ₀ and C _t (mmol L⁻¹) represent the initial and time t concentration of metal ion, respectively; V _t (L) is the volume of solution at time t; and W (g) is the weight of the composite adsorbent.

A total of 10 mg composite was added into 40 ml solution with binary metal ions in the same concentrations (100 mg L⁻¹) and shaken at 25°C for 24 h. The concentrations of the two metal ions were determined by AAS, respectively, calculated with Equation 1, and then the adsorption selective coefficient (α) was defined as formula (3): α = The adsorption capacity of Hg  ( II )  on adsorbent  The adsorption capacity of coexisting metal ion on adsorbent  (3)

A total of 10 mg composite was added into the Hg(II) solution under the optimal conditions and shaken thermostatically for 24 h. It was filtered and then the adsorbent with metal ions anchored was placed in the oven at 60°C until completely dry. The mechanism of adsorption process was inferred by analyzing the changes of the elements in XPS spectra before and after adsorption.

Different concentrations of eluents were prepared according to the acid environment that Hg(II) located. A total of 10 mg adsorbent with Hg(II) anchored was added into 40 ml eluent (0.1 mol L⁻¹ HNO₃, 1, 2, 3, 4, and 5% thiourea in 0.1 mol L⁻¹ HNO₃, respectively). It was then shaken at 25°C and desorption for 24 h, and the metal ion concentrations of eluents after desorption were measured by AAS, to determine the best eluent. Repeat adsorption-desorption three times, and judge the regeneration performances of the composites.

The adsorption properties of different materials for metal ions are shown in Figure 2A. As you can see from the figure, the PSQ/CNTs magnetic composites showed excellent adsorption capacities for Hg(II) rather than Ag(I), Pb(II), Cu(II), and Ni(II). Compared with pristine CNTs-COOH@Fe₃O₄ and intermediate CNTs-APTMS@Fe₃O₄, PSQ/CNTs magnetic composites displayed more remarkable adsorption performance to Hg(II), and the adsorption capacities increased 0.3–0.9 mmol g⁻¹ to different extent. Further, the bifunctional PSQ/CNTs magnetic composites expressed relatively higher adsorption capacities than monofunctional composites, especially the composite A@Fe₃O₄. As a whole, the bifunctional PSQ/CNTs magnetic composites showed superior adsorption abilities to Hg(II) and the systematic exploration would be further studied in the following.

Figure 2B shows the effect of solution pH values on different composites for adsorbing Hg(II); the experimental pH values varied from 1.0 to 4.5. Obviously, the adsorption capacities of composites to Hg(II) were greatly affected by the solution pH values. To be specific, the monofunctional composite A@Fe₃O₄ showed consistent change with pH variation, and this can be interpreted as H⁺ tends to compete with Hg²⁺ to form -NH₃ ⁺ at relatively low pH values (Wang et al., 2017; Wang et al., 2019), the adsorbent surface is positively charged and electrostatic repulsion occurs with Hg²⁺, thus the adsorption capacity is low at highly acidic solution. With the growth of pH value, the concentration of H⁺ decreases and the protonation of -NH₂ weakens, and thus the adsorption amount of composite for Hg(II) increased spontaneously. For M@Fe₃O₄, the effect of solution pH was mainly divided into two parts, and the adsorption capacity reached the maximum at pH 4.5. For bifunctional composites, the adsorption capacities greatly affected by the thiol group before pH 1.5 and mainly affected by the amino group after pH 1.5, and finally reached the maximum at pH 4.5.

The adsorption isotherms of the composites for Hg(II) were carried out with initial concentration at 50, 75, 100, 125, and 150 mg L⁻¹ at pH 4.5 and different temperatures. The adsorption capacities variation of PSQ/CNTs magnetic composites with initial concentration are shown in Figure 3A, and the bifunctional composites can basically achieve complete adsorption at relatively low initial concentrations; the monofunctional composites have less adsorption, especially the composite A@Fe₃O₄. As the initial concentrations increase, the equilibrium adsorption capacity of all composites increased.

The equilibrium adsorption isotherm plays an important part in investigating the adsorption mechanism and adsorption capacity (Li et al., 2019; Hu et al., 2021). Four types of adsorption model, Langmuir, Freundlich, Tempkin, and Hill models, were used to fit the adsorption process to further explore the thermodynamic adsorption for Hg(II). The Langmuir model deems that adsorption occurs on a mono-layer of uniform surface with no interaction between the adsorbents (Zhao et al., 2019; Ghodsi et al., 2021); differently, the Freundlich mode deems that adsorption occurs on multi-layers of heterogeneous surfaces. The Temkin model reflects that the reaction between the adsorbate and the adsorbent linearly reduces the heat of adsorption. The Hill model showed the relation of different species on the homogeneous surfaces and assumed that one adsorption site can capture n ₂ ions. Where n ₂ > 1 said positive cooperativity, n ₂ = 1 proved non-cooperative and n ₂ < 1 showed negative cooperativity between the binding (Saadi et al., 2015). The equations of the four models are as follows, respectively: C e q e =   C e q max +   1 q max K L (4) q e =  K F ⋅ C e 1 n , (5) q e = B T ⋅ ln ( K m C e ) , (6) q e = n 2 N M C e n 2 C e n 2 + C 1 / 2 n 2 (7) where C _e (mmol L⁻¹) is the equilibrium solution concentration; q _e (mmol g⁻¹) is the equilibrium adsorption capacity; q _max (mmol g⁻¹) is the maximum adsorption capacity; K _L is the Langmuir constant; and K _F and n are the Freundlich constants related to adsorption capacity and strength, respectively. K _m was the constant of the Temkin model and B _T was related to the heat of adsorption. n ₂ , N _M, and C _1/2 were the number of the adsorbed ions at each site, the density of the receptor sites, and semi saturated concentration, respectively.

The fitting of the composite 2A-M@Fe₃O₄ at different temperatures to the four models are shown in Figures 3C–F, respectively, and the related parameters of all bifunctional PSQ/CNTs magnetic composites are calculated and shown in Tables 1, 2. From the fitting curves in figure and parameters in the table, we can find that the fitted related coefficient R_L ² > R_F ² > R_T ^{2 >} R_H ², that is, the Langmuir model is more suitable to describe the thermodynamic process of composites to Hg(II) adsorption, and thus, the adsorption occurs on a mono-layer with no interaction.

Further, thermodynamic parameters of the bifunctional PSQ/CNTs magnetic composites to Hg(II) were also calculated by Equations 8, 9, including Gibbs free energy (ΔG), entropy change (ΔS), and enthalpy change (ΔH) (Niu et al., 2014): ln K L = Δ S R − Δ H RT (8) Δ G = Δ H − T Δ S (9) where K _L is the Langmuir constant, R is the gas constant (R = 8.314 J mol⁻¹ K⁻¹), and T (K) is the experimental temperature.

Plotting ln K _L against 1/T to give fitting straight-lines of bifunctional PSQ/CNTs magnetic composites in Figure 3B, the slope and intercept are -ΔH/R and ΔS/R, respectively. The adsorption thermodynamic parameters ΔG, ΔS, and ΔH were calculated in Table 3, in which the values are ΔG < 0, ΔS > 0, and ΔH > 0, indicating that the adsorption process of bifunctional PSQ/CNTs magnetic composites to Hg(II) is a spontaneous, increased confusion and endothermic reaction process. That is to say, the whole adsorption process is a spontaneous behavior and we can promote the metal ions adhered on the surface of composites to enhance adsorption capacities by increasing temperatures. Compared with the adsorbents of the same type reported in the literature listed in Table 4, the adsorbents prepared in this paper show superior adsorption performance.

Adsorption kinetics is a common method to characterize the adsorption efficiency and indicate the adsorption type of solute in the adsorption process (Yang et al., 2014; Wei et al., 2021). Figure 4A is the adsorption rate of all prepared PSQ/CNTs magnetic composites for Hg(II) at 25°C, and the adsorption rate fell in the following order: A-M@Fe₃O₄ > A-2M@Fe₃O₄ > M@Fe₃O₄ > A@Fe₃O₄ > 2A-M@Fe₃O₄. Obviously, the whole adsorption process for Hg(II) consists of two distinct parts, the first hour is the rapid adsorption stage to give the main adsorption capacity, the second stage is relatively smooth and slow to give the final equilibrium adsorption capacity. As can be seen from the figure, the adsorption process reached adsorption equilibrium in about 11–13 h.

For better exploring the adsorption process of PSQ/CNTs magnetic composites to Hg(II), the kinetics data were fitted into the pseudo-first-order and pseudo-second-order models (Wang et al., 2019; Wang L. et al., 2020) as follows, respectively: ln ( q e − q t ) = lnq e − k 1 t (10) t q t = 1 k 2 q e 2 + t q e (11) where q _e and q _t (mmol g⁻¹) are the adsorption capacity for Hg(II) at equilibrium and time t (h), respectively; k ₁ (h⁻¹) and k ₂ (g mmol⁻¹h⁻¹) are the rate constant of pseudo-first-order and pseudo-second-order models.

Figures 4B,C and Table 5 show the fitting results and the specific kinetic parameters of bifunctional PSQ/CNTs magnetic composites to the two models. As can be seen from the fitting lines and data, the pseudo-second-order model has higher correlation coefficient than the pseudo-first-order model (R₂ ² > R₁ ²). That is, the pseudo-second-order model is more suitable to describe the adsorption process for Hg(II), thus, the limiting step of adsorption rate may be chemical adsorption involving valency forces through the sharing or exchange of electrons between PSQ/CNTs and Hg(II) (Qu et al., 2009).

The adsorption selectivity is an important factor to evaluate the adsorption properties of adsorbents (Mudasir et al., 2020; Wang et al., 2020a), thus, the adsorption selectivity of the representative composite A-2M@Fe₃O₄ for Hg(II) with common coexisting metal ions was carried out. The experiments were operated at optimal pH (pH = 4.5) and the results in Table 6 suggested that the composite tend to adsorb Hg(II) rather than coexisting metal ions, especially ions Pb(II), Ni(II), Cu(II), and Cd(II). The selectivity coefficient is relatively low when Ag(I) coexists with Hg(II), and this can be interpreted by the hard-soft acid-base theory (HSAB). Hg(II) and Ag(I) are classified as soft ions and can form high affinity with functional groups containing nitrogen and sulfur, thus the coexistence of Ag(I) has a great influence on the adsorption of Hg(II). As a whole, the bifunctional PSQ/CNTs magnetic composites can potentially adsorb and separate Hg(II) in coexisting metal ions system.

The binding energy changes of elements in XPS scan before and after adsorption were analyzed to infer the adsorption mechanism for Hg(II). Figure 5A shows the wide scan XPS spectra of bifunctional PSQ/CNTs magnetic composites after Hg(II) adsorption, the binding energy peaks of Hg_4f and Hg_4d are clearly displayed in the curves, and the peak of Hg_4p is easily found as well. The binding energy peaks of Hg_4f at 99.8 and 103.6 eV in Figure 5D are vested to Hg_{4f 5/2} and Hg_{4f 7/2} (Guo et al., 2020; Xia et al., 2021), respectively. Further, the main high-resolution peaks of Hg_{4f 7/2} contains two different valence state peaks Hg²⁺ and Hg⁰ (Wang et al., 2020b), which located at 99.6 and 100.5 eV, respectively, indicating that the redox reaction may occur in the process of adsorbing Hg(II). The element binding energy changes of the composite A-M@Fe₃O₄ before and after adsorption are displayed in Figures 5B,C. In the spectra of N_1s, the binding energy shifts from 397.3 to 397.5 eV for -NH₂ and the relative peak area increased, the binding energy shifts from 398.1 to 398.4 eV for -CONH-, the binding energy shifts from 399.2 to 400.1 eV for the protonated -NH₃ ⁺. The changes of binding energy and peak area indicate that chelation or ion exchange may occur during the process of adsorbing Hg(II). Similarly, in the spectra of S_2p, the binding energy of -SH shifts from 162.6 to 161.8 eV, indicating that chelation or ion exchange may be carried out. Meanwhile, the binding energy of oxidized S appeared at around 166.5 eV, which can be inferred that redox reaction may took place between -SH and Hg(II), and corresponds to the reduction of Hg²⁺. In short, chelation or ion exchange may primarily occurred in the whole adsorption process, and certainly, limited redox reaction took place as well.

The composite A-M@Fe₃O₄ was selected to explore the elution and regeneration performance for Hg(II), the desorption rate of A-M@Fe₃O₄ in a series of 0.1 mol L⁻¹ HCl solution with different concentrations of thiourea eluents shown in Figure 6. Obviously, the desorption rate reached maximum in 0.1 mol L⁻¹ HCl solution with 3% thiourea eluents; thus, the following elution experiments were taking place in it and the adsorption rate varied with regeneration times shown in Figure 6 inset. The adsorption rate gradually decreased with the increasing of regeneration frequency and the adsorption rate still achieved 78% after repeat elution-regeneration three times. That is to say, the bifunctional PSQ/CNTs magnetic composites have favorable regenerability and can be expected to be a kind of potential economic adsorbent.