Highly selective adsorption of Hg (II) from aqueous solution by three-dimensional porous N-doped starch-based carbon

For the first time, N-doped carbon materials with 3D porous–layered skeleton structure was synthesized through a one-step co-pyrolysis method, which was fabricated by co-pyrolysis of natural corn starch and melamine using metal catalysts (Ni (II) and Mn (II)). The 3D-NC possessed a heterogeneously meso-macroporous surface with a hierarchically connected sheet structure inside. Batch adsorption experiments suggested that highly selective adsorption of Hg (II) by the 3D-NC could be completed within 90 min and had maximum adsorption capacities as high as 403.24 mg/g at 293 K, pH = 5. The adsorption mechanism for Hg (II) was carefully evaluated and followed the physical adsorption, electrostatic attraction, chelation, and ion exchange. Besides, thermodynamic study demonstrated that the Hg (II) adsorption procedure was spontaneous, endothermic, and randomness. More importantly, the 3D-NC could be regenerated and recovered well after adsorption–desorption cycles, showing a promising prospect in the remediation of Hg (II)-contaminated wastewater.


Introduction
In recent years, the development of fossil fuel extraction technology has greatly promoted the progress of manufacturing and their products such as batteries, lamps, and thermometers have become indispensable parts of daily. However, their by-products, namely, heavy metal mercury failed to be properly handled (Gavilán-García et al. 2015;Horowitz et al. 2014). In addition, the chemical treatments of nonferrous metals such as gold, silver, and copper also produce mercury as a by-product in the metallurgic industry and mercury has also been detected in ash and gases produced by burning fossil fuels (Chalkidis et al. 2020, Hylander andHerbert 2008). Consequently, the increasing industrial pollutants containing mercury have been discharged into the natural environment, leading to excessive mercury pollution in the water, which will pollute the soil, hinder seed germination, increase crop morbidity, and reduce microbial activity (Liu et al. 2021). Humans living with excessive levels of mercury are more prone to chronic, autoimmune, and neurological diseases (Bjørklund et al. 2017). Therefore, it is urgent to implement feasible and suitable approaches to remove Hg (II) ions from industrial sewage. Several treatment methods including precipitation, ion exchange, reduction, adsorption have been applied in Hg (II) removal process (Baimenov et al. 2020;Oehmen et al. 2014;Quesada et al. 2020;Tauanov et al. 2020). Among them, adsorption has the advantages of cost-effectiveness, less regeneration energy, easy operability, and fast kinetics and was regarded as the most promising method for water remediation (Yaumi et al. 2018;Zhu et al. 2015). So far, various adsorbents including activated carbons, activated alumina, zeolite, diatomite, clay have been already used in sewage purification (Diagboya et al. 2015;Duan et al. 2020;Kumari et al. 2020;Park et al. 2007;Tauanov et al. 2018). Although these adsorbents have been developed rapidly in recent years and have several advantages, their practical application has certain insurmountable limitations, such as lower adsorption capability, large consumption, difficulty in regeneration, and high cost. Therefore, it is necessary to develop new types of adsorbent materials with cost-effective performance. In recent years, N-doped carbon materials was categorized as a promising material with a degree of nitrogenization fabricated by the co-pyrolysis of graphitic carbon nitride (GCN) (as nitrogen source) and organic compounds (as carbon precursor) Yu et al. 2016). And it was equipped with a 2D or 3D structure, specific porosity, and excellent semiconductivity, which stimulated the research enthusiasm for its application in the field of catalysts and semiconductors Yang et al. 2016;Yu et al. 2016). In addition, as a porous carbonaceous material functionalized with N atoms, the N-doped carbon had prospective applications in the adsorption, and its nitrogenfunctionalized active sites not only had been proved to have excellent catalytic activity but also a had high sensitive and preferential affinity for mercury ions (Liao et al. 2019;Qu et al. 2019;Tauanov et al. 2018), which made it have potential to adsorb Hg (II) in aqueous solution.
Traditional method to prepare N-doped carbon material usually adopted a GCN as the template with separate steps, which would result in unnecessary energy loss due to the repeated steps of temperature rise and fall, the instability of materials because of the complexity of multi-procedure, and potential hazards caused by high temperature annealing calcination. However, on the theoretical side, the synthesis of GCN could be completed by the carbonization of polysaccharides such as starch, cellulose, and resin and the polymerization of the nitrogen-rich precursors such as urea, thiourea, and melamine at a certain temperature (Ong et al. 2016). Moreover, the metal catalysts such as Ni (II) and Mn (II) had the ability to moderate the formation temperature of the GCN, gasify the crystal structure of GCN to developed porosity, and promote the transmutation of GCN to the N-doped carbon (Demir et al. 2015;Huang et al. 2011;Kim et al. 2001;Ong et al. 2016;Oya et al. 1995;Wang et al. 2008aWang et al. , 2021Wang et al. , 2020Xie et al. 2021). These properties made it possible to synthesize carbon nitride materials in one-step co-pyrolysis method.
Starch, as a low-cost and biological carbonaceous polysaccharide in grains, had a considerable crystallinity structure, which was an excellent candidate for constructing carbon skeleton (Chen et al. 2020b). In addition, the foaming reaction of starch granules occurred during carbonization at high temperature (Qu et al. 2020;Shi et al. 2021;Song et al. 2022), which was conducive to form three-dimensional structure by vertical extension of starch-based carbon sheet. Melamine was a common raw material for the synthesis of N-doped carbon material and its abundant N atoms had the ability to attack atomic lattice of carbon and replace carbon atoms in carbon structure (Guo et al. 2016;Lin et al. 2012;Sun et al. 2006;Wood et al. 2014;Xing et al. 2015;Zhang et al. 2016), which had a positive effect on maintaining carbon skeleton (Gao et al. 2021). At the same time, the N-containing groups had a considerable affinity for Hg(II), which had a high capability in selective adsorption performance (Liao et al. 2019;Qu et al. 2019).
Hence, for the first time, a new-typed N-doped starchbased carbon material was synthesized through one-step copyrolysis of corn starch and melamine catalyzed by metal catalysts (Ni (II) and Mn (II)). The product (3D-NC) showed 3D porous skeleton of starch-based carbon and the one-step synthesis route tactfully realize one-step transformation from raw materials (starch, melamine) to the N-doped carbon, namely, the GCN was firstly synthesized at a lower temperature (550 °C) and followed by heating up to the point (900 °C) at which the GCN completely decomposed to the N-doped carbon (Wang et al. 2012).
On the whole, the three-dimensional porous N-doped starch-based carbon (3D-NC) was initially constructed through simple, reliable, and cost-effective one-step synthesis routine and it performed excellent adsorption capacity and highly selective adsorption for Hg(II)-containing sewage. In order to investigate the effectiveness of the 3D-NC, experiments were conducted by altering the conditions including pH, adsorption time, and temperature. Differential pulse voltammetry (DPV) was used to reflect the changes in the concentration of Hg (II) before and after adsorption. Its relevant characteristics in morphology and performance were evaluated by scanning electron microscopy(SEM), thermogravimetric analysis(TGA), X-ray diffraction(XRD), X-ray photoelectron spectroscopy(XPS), Fourier transform infrared spectroscopy(FTIR), and zeta potential. The adsorption models were established to investigate the adsorption mechanism. We envisioned that the synthetic route of the 3D-NC could provide inspiration for the preparation of costeffective materials from starch-based biological sources and broaden the path for water pollution control.

Preparation of the 3D-NC
The 3D-NC was synthesized through the one-step copyrolysis and catalyzed by Ni (II) and Mn (II) (Park et al. 2021;Tan et al. 2012;Wang et al. 2020). At first, 2 g corn starch powder was mixed with deionized water (200 mL) and stirred at ambient temperature for 30 min. Afterward, 40 mL Ni (II) solution (0.1 M) and 40 mL Mn (II) solution (0.2 M) was simultaneously poured into the above starch solution and mixed to form a uniform system. Then, the pH of the homogeneous solution was adjusted by the diluted ammonia solution to 7. After that, the solution continued stirring at room temperature for 1 h followed by centrifugal washing (6000 rpm/10 min) to get the mixed sediment. Afterward, the sediment was dried at 80 °C for 24 h to obtain dry powder. Next, the dry powder containing catalytic ions was mixed with melamine and fully ground. The mixed powder was carbonized in a tubular furnace at 550 °C (ramped at a rate of 5 °C/min) for 2 h under N 2 flow. Then, the temperature was further raised to 900 ℃ under the same conditions and kept for 3 h. Finally, the furnace was turned off to attain ambient temperature under the nitrogen atmosphere and the 3D-NC was prepared. The obtained samples were treated with centrifugal washing by diluted hydrochloric acid and deionized water until pH to 7 for removing the catalysts, and the products were dried overnight at 50 °C. The stepwise procedure for the fabrication of of the 3D-NC was shown in Scheme 1.

Adsorption experiments
The adsorption capacity of 3D-NC (10 mg) was evaluated in the 50 mL Hg (II) aqueous solutions with pH, time, and temperature control system on a thermostatic water bath stirring at 300 r/min. Differential pulse voltammetry (DPV) was used to determine the changed electrical signals of Hg (II) solutions before and after adsorption. DPV, as a fast, sensitive, reproducible, and low-cost detection method, has been confirmed to be capable of rapidly and accurately detecting the concentration of heavy metals and organic compounds (Estrada-Aldrete et al. 2020;Sun et al. 2018;Xu et al. 2008). In this article, the used electrochemical detection equipment was a traditional three-electrode system, and its electromotive force (E) range was set as − 0.2 to 0.5 V. The concentration of Hg(II) solutions was reflected by the electrical signal determined by DPV and the difference in electrical signal before and after adsorption could represent the change of Hg(II) solutions concentration. Meanwhile, all these test results are further verified by atomic absorption spectrometry. All the measurements were carried out in triplicate and the standard deviation was kept within 5%.The adsorption capacity ( q e , mg/g) of the 3D-NC and the removal rate (%) of Hg (II) were calculated by the following formulas: where V and m represented the volume of solution (mL) and the dosage of the adsorbent (g), respectively. And C 0 and C e were defined as the initial concentration (mg/L) of Hg (II), and the equilibrium concentration (mg/L) of Hg (II), respectively.
Firstly, the optimal adsorption of pH (pH = 4, 5, 6, 7) was determined and equilibrium adsorption time on Hg (II) (t = 0, 10, 20, 30, 45, 60, 90, 120,150 min) was then investigated at 293 K. Afterward, the influence of concentration of Hg (II) solution on adsorption efficiency was evaluated in the range of 10 mg to 100 mg/L. Finally, the influence of temperature (T = 293 K, 303 K, 313 K, 323 K, 333 K) on the adsorption capacity of the 3D-NC was summarized. Adsorption kinetics models, adsorption isotherm equations, and thermodynamic parameters were investigated to analyze the adsorptive property of the 3D-NC for Hg (II) ions in an aqueous solution.
Scheme 1 Schematic illustration for the synthesis of the 3D-NC

Regeneration and selective adsorption of the 3D-NC
All of the sorbents used in the experiments were collected for further assessing the regeneration of the 3D-NC. The adsorbent products were desorbed by being stirred in the pre-prepared dilute hydrochloric acid (0.05 M) for 2 h. After desorption, the absorbents were centrifuged and washed with deionized water, and then, the separated powders were dried at 50 ℃ overnight for cycle adsorption experiment. The 10 mg regenerated 3D-NC was added into Hg (II) solution (50 mL, 100 mg/L) at pH = 5, T = 293 K for 2 h. The regeneration of the 3D-NC was performed by X-ray photoelectron spectroscopy (XPS) and Differential pulse voltammetry (DPV). The selective adsorption property of the 3D-NC was verified by applying the adsorbent to a mixed ions system to compare its adsorption capacity for different heavy metal ions. The mixture ion stock solution (50 mL) contained coexisting ions of Hg (II), Cd (II), Co (II), Cu (II), and Pb (II) and the concentration of each ion was 100 mg/L. The detection measurement followed as the aforementioned method.

Characterization of 3D-NC
As shown in Fig. 1, the 3D-NC had a three-dimensional fluffy network sheet structure and its structure presented a connected skeleton structure. In addition, its surface was porous and the interior was connected. The nitrogen adsorption-desorption isotherm in Fig. 2A revealed the surface proprieties of the 3D-NC. According to International Union of Pure and Applied Chemist (IUPAC) classification (Rouquerol et al. 2014), the gas adsorption type of the 3D-NC was conformed to the type IV isotherm and a hysteresis loop could be observed, which illustrated that the mesopores emerged in the 3D-NC (Lu et al. 2014;Tabatabaei Shirazani et al. 2020). Moreover, the occurrence of the hysteresis loop mapped the capillary condensation of the adsorbate in the mesopores. From the above characteristics, the solid-gas adsorption of the 3D-NC conformed to the hypothesis from the Kelvin equation, namely, the gas adsorption began with the monolayer adsorption on the surface and was followed by multilayer adsorption and was finished by capillary condensation (Horikawa et al. 2011). The specific surface area of the 3D-NC was calculated as 20.113 m 2 /g.
In the BJH analysis of the 3D-NC (Fig. 2B), the materials exhibited not only the mesopore structure in the 2-50 nm but also micropore structure in > 50 nm. In addition, the BJH method gave a calculation of the total pore volumes of 0.059 cm 3 /g. In consequence, combined with the SEM result, the mesoporous/macropore structure of the 3D-NC had been successfully molded and in the adsorption process, the macropores served as ions channels, which was beneficial to diffusion and transfer of adsorbates. Furthermore, the active sites and functional groups hid in the gaps of the mesopores, which facilitated to improve adsorption performance of the materails (Chen et al. 2020a;Lu et al. 2014).
In this section, the 3D-NC without catalyzation by Ni (II) and Mn (II) (the 3D-NC without Ni (II) and Mn (II)) were prepared as comparison to verify the effect of metal ions on porosity enhancement. The relevant differences between the 3D-NC and the 3D-NC without Ni (II) and Mn (II) were shown in Table 1. The results showed that the 3D-NC had a larger specific surface area than the 3D-NC without Ni (II) and Mn (II), which implied that the 3D-NC had a higher porosity. The differential results indicated that the catalyzation of Ni (II) and Mn (II) caused an enhancement of the surface proprieties of the materials.
TGA in the range of 30-700 ℃ of the 3D-NC was used to verify the stability of the introduced N atoms. In order to further eliminate the interference of metal catalyst, the TGA of 3D-NC without Ni (II) and Mn (II) were collected and the corresponding results were depicted in Fig. 2C. Different from the pyrolysis curve of corn starch, the 3D-NC showed a horizontal stable curve and maintained structural stability up to 500 ℃. Interestingly, the 3D-NC without Ni (II) and Mn (II) exhibited higher structural stability than the 3D-NC. Combined with the results from Table 1, it could be inferred that although the function of Ni (II) and Mn (II) could further improve the porosity of 3D-NC, its higher porosity also meant that more inner surfaces of the materials were exposed to heat flow, which led to the decrease of structural stability.
The pH value of solution was an important factor in the adsorption process and it not only affected the protonation/ deprotonation of function groups but also adjusted the surface charge of the adsorbents (Nguyen Van et al. 2020). Hence, zeta potential and the point of zero charge (PZC) were used to evaluate the adsorption efficiency of the materials (Umh and Kim 2014). As shown in Fig. 2D, the PZC of the 3D-NC was about 1.6 and the surface of the materials was negatively charged when pH > 1.6, which could attract positively charged Hg (II). Therefore, the electrostatic interaction was a considerable adsorption force during the adsorption process, which significantly affected adsorption capacity of the 3D-NC on cationic Hg (II) in the aqueous solution.
XRD pattern of corn starch and the 3D-NC were presented in Fig. 2E. It could be seen that the corn starch exhibited A-type crystalline pattern with its strong reflections at 2θ of 15.0°, 17.0°, 18.0°, and 23.0° (Zhang et al. 2012). After reaction, the A-type crystalline disappeared, just leaving two crystalline peak in 3D-NC pattern. The 3D-NC exhibited crystal structures with two broad bumps in 2θ = 26° and 2θ = 43°, corresponding to (002) and (100) planes and showing graphitization (Chen et al. 2020a;Kim et al. 2014;Luan et al. 2022). The diffraction peak at 26° was ascribed to amorphous carbons formed by the π-conjugation of non-regular aromatic carbon rings and the peak at 43° was attributed to the occurrence of graphitic structure, which usually formed in the preparation of GCN (Chen et al. 2020a;Kim et al. 2014;Pang et al. 2018). This compassion indicated that the pyrolysis further dehydrated the original sugar ring in corn starch to form an aromatic ring structure similar to graphite, which was the base crystalline phase of GCN and also proved that the one-step method to prepare N-doped carbon did not require the synthesis of the GCN precursor in another step. Generally, the 3D-NC was amorphous N-doped carbon that was composed of phenyl ring units with a degree of graphitization. Fig. 2F shows the FTIR spectrum of the 3D-NC. The broad peak located in 3440 cm −1 was ascribed to the stretching vibrations of -NH/-OH (Qu et al. 2019). The peaks around 1620 cm −1 were related to the stretching vibration of C = O/C = N (Gao et al. 2021). The characteristic peaks presented at 1417 cm −1 were associated with C-N stretching vibration and the peaks at 1070 cm −1 belonged to C = O/C = N stretching vibrations (Jin et al. 2019;Liao et al. 2019). These phenomena demonstrated that the 3D-NC had functional hydroxyl, carboxyl, and nitrogen groups, which were expected to serve as active adsorption sites during the adsorption process (Zhu et al. 2015) XPS was used to analyze the surface functional groups and element states of the 3D-NC, and the relevant spectra were shown in Fig. 3. In the full spectrum of the 3D-NC, the C1s, O1s, and N1s peaks were observed at 284.1 eV, 532.8 eV, and 400.2 eV, respectively. To further estimate the binding states of surface functional groups, the high resolution spectrum of the above elements were deconvoluted and the relevant results were presented. According to the high precision spectrum of C1s, the peaks at 284.80 eV, 286.33 eV, and 288.79 eV were attributed to C-C, C-O/C-N, and C = O/C = N, respectively (Chen et al. 2020a;Liao et al. 2019;Yue et al. 2018). The N1s XPS spectrum revealed the overlapping peaks of C-N-H / C-N-C at 400.01 eV (Liao et al. 2019). Similarly, the O1s peak was deconvoluted into O-H at 533.40 eV and the overlapping

Optimization of experimental conditions
The optimal pH value of the adsorption was investigated by controlling for a single variable. Figure 4A demonstrates the curve of adsorption efficiency changed with pH values under the same initial mercury ions concentration (100 mg/L) and adsorption temperature (293 K). As shown in Fig. 4A, the removal rate (%) of Hg (II) at different pH values had distinct differences, which implied that pH played a significant role in Hg (II) adsorption and the maximum adsorption capacity was achieved at pH = 5. Combined with the analysis of zeta potential, there was an electrostatic interaction between the 3D-NC and cationic Hg (II). It could be explained that when the pH of the solution exceeded the PZC of the materials (PZC = 1.6), the functional groups (-COOH, -OH, -NH 2 , etc.) of the materials were deprotonated and presented a negatively charged surface, which could attract the positively charged Hg (II) (Ghiorghita et al. 2020, Sarmah andKarak 2020). Moreover, with the increasing of the pH values, the deprotonation was strengthened and resulted in the increase of negative charge on the surface of materials, which further reinforced the electrostatic attraction toward Hg (II) and improved the adsorption efficiency (Park et al. 2021;Seo et al. 2017). However, after achieving the optimal adsorption performance at pH = 5, the removal rate of Hg (II) showed a downward trend. It could be inferred that as the solution tend to alkaline environment, the content of Na + and K + increased, which led to electrostatic shielding effect between the ionized functional groups and the Hg (II) (Moharrami and Motamedi 2020, Zhu et al. 2015). Fig. 4B depicts the relation between absorption efficiency and absorption time at pH = 5. The adsorption equilibrium was achieved at 90 min, where the adsorption efficiency reached 403.24 mg/g. Within the initial 30 min, the curve raised steeply and in the state of 30-60 min, the slope of the curve approached zero. After 90 min, the curve no longer had a distinct variation and showed a flat line. This observation indicated that the 3D-NC could maximize its effectiveness in less than 90 min and had a promising application prospect. And the adsorption effect of similar materials was summarized in Table S1 (Supplementary Material) as comparison Figure 4C summarizes the influence of temperature (293 K, 303 K, 313 K, 323 K, and 333 K) on the adsorption capacity of the 3D-NC. With the increasing temperature from 293 to 333 K, the removal rate of Hg (II) increased from 80.06 to 92.48%. The elevated ambient temperature provided more energy, which enhanced the mobility of mercury molecules (Ghorai et al. 2013). In addition, the swelling effect of the internal structure of adsorbent occurred in the temperature-elevated environment and was conducive to infiltrate the adsorbate molecule into the structure, further improving adsorption efficiency of the adsorbent (Wang et al. 2008b).

Adsorption kinetics with the 3D-NC
In order to further illustrate the adsorption mechanism of the 3D-NC on Hg (II). The pseudo-first-order model, pseudosecond-order model, and intra-particle diffusion model were used to evaluate the experimental data. The adsorption kinetic curves were presented in Fig. 5A, Fig. 5B, and Fig. S1 (Supplementary Material).
The models of the above equations were as follows: where Q e (mg/g) and Q t (mg/g) were equilibrium adsorption capacity of the 3D-NC and adsorption capacity of it at contact time t, respectively.k 1 (min −1 ) and k 2 (g/(mg·min)) represented the kinetic adsorption rate constants. k 3 (mg/g·min −1/2 ) describe the rate constants of intra-particle diffusion model and C is the boundary layer thickness. Compared to the less correlation coefficient (R 2 = 0.9667) of the pseudo-first-order model, the pseudo-second-order model with an excellent correlation coefficient (R 2 = 0.9997) was more suitable for the adsorption data. In addition, the calculated equilibrium adsorption value (Q e,c = 403.29 mg/g) of the pseudo-second-order model was more consistent with the reality (Q e,e = 403.24 mg/g) than pseudo-first-order model. The relevant details were summarized in Table 2. Hence, the pseudo-second-order model could better describe the adsorption process and the model indicated that surface electron exchange occurred in the interaction between the metal ions and adsorbent, and it usually occurred as a limiting step (Akpotu et al. 2023(Akpotu et al. , 2022Wang et al. 2019).
Intra-particle diffusion models indicated that when the plot q t versus t 1/2 presented a straight line, the intra-particle diffusion process occurred in the adsorption process. And if the straight line passed through the origin, the intra-particle diffusion process operated as a controlling step. However, Figure S1 presents multi-linear plots and there were more than one process working in the adsorption process (Sarmah and Karak 2020). Figure S1 shows that two linear regions operated in the adsorption process, which indicated that two successive adsorption stages occurred in the adsorption process. In the first phase, particles differed from external aqueous solution to the boundary layer of the 3D-NC and this stage was mainly controlled by the external mass transfer diffusion (Nag et al. 2016). The second phase was the particle internal diffusion. In this part, the particles passed through pores of 3D-NC and were anchored by its binding sites. As could be seen from Fig. S1 and Table 2, the slope of the first phase (k 3-1 = 34.58) was much higher than that of the second segment (k 3-2 = 1.65), so the first phase was the limiting step. In addition, none of the plots passing through the origin indicated that not only intra-particle diffusion occurred during adsorption but also that other mechanisms simultaneously existed (Karaçetin et al. 2014).

Adsorption isotherm experiments with the 3D-NC
Langmuir, Freundlich, and D-R isotherm models were used to evaluate the interactive behavior between adsorbates and the adsorbents. Figure 5C-E represents the fitting curves for the above models and the related details were tabulated in Table 3.
Langmuir and Freundlich models could be presented as the following formulas: where C e (mg/L) was the concentration of solution at equilibrium. Q e (mg/g) represented the equilibrium adsorption capacity. Q m (mg/g) and b (L/mg) represented the maximum adsorption capacity of the 3D-NC and the constant values in the Langmuir model, respectively. n and K f represented the constant values of Freundlich equation. D-R model was presented as the following formulas (Zou et al. 2021): where Q e and Q m were equilibrium adsorption capacity and saturated monolayer adsorption capacity, respectively. K d and represented the average adsorption free energy coefficient and the polanyi potential energy of the surface, respectively. R (8.314 J /mol•K) was the gas constant. T (K) was the temperature. C e (mg/g) and E (kJ/mol) respectively stood for the concentration of mercury ions at adsorption equilibrium and adsorption free energy.
From the fitting equations under the same conditions, the correlation coefficient (R 2 ) was analyzed. It was concluded that the correlation coefficient of the Freundlich model (R 2 = 0.9824) was more considerable than the value of the Langmuir model (R 2 = 0.9569), which implied that Langmuir model failed to properly describe the adsorption process of the 3D-NC toward Hg (II). On the basis of the assumptions of the Langmuir model, adsorption occurred as monolayer adsorption on homogeneous and energetically uniform surfaces, while Freundlich adsorption model assumed that the adsorption took place on heterogeneous surfaces by multilayer adsorption with nonhomogeneous distribution of adsorption heat and affinities (Bello et al. 2018;Gao et al. 2021). As a result, the 3D-NC preferred the Freundlich model and multilayer adsorption occurred during the Hg (II) adsorption. Furthermore, the evidence from Table 3 demonstrated that the values of 1/n (0.5251) was in the range of 0.1 to 1, which indicated that a non-linearity tendency for Hg (II) adsorption occurred (Akpotu et al. 2023). According to the D-R isotherm, when the free energy E < 8 kJ/mol, the adsorption mechanism belonged to physical adsorption; when the free energy E was in the range of 8 to 16 kJ/mol, the ion exchange played a primary role; when E > 16 kJ/mol, adsorption was chemical adsorption (Chen et al. 2013;Yu et al. 2013). The adsorption free energy E of the 3D-NC was calculated to14.25 kJ/mol at 293 K, which was in the range of 8 to 16 kJ/mol, further implying that the ion exchange was involved for the adsorption of Hg (II) onto the 3D-NC.

Thermodynamic parameters
In the adsorption process, the Gibbs free energy ( ΔG • , kJ/ mol), standard enthalpy change ( ΔH • , kJ/mol), and standard entropy change ( ΔS • , J/mol·K) were the important thermodynamic parameters to judge the spontaneous reaction of adsorption . Therefore, the Van't Hoff equation was carried out to illustrate the ΔG • , ΔH • , and ΔS • , and the results were shown in Fig. 5F and Table 4. The relevant equations were as follows (Liu 2009, Liu andLiu 2008): where R (8.314 J /mol•K), T (K), and m (g) were the gas constant, the temperature, and the dosage of adsorbent, respectively. K c was the equilibrium constant. As shown in Table 4, the positive values of ΔH • (22.43 kJ/ mol) demonstrated that the Hg (II) adsorption by 3D-NC was endothermic process and the elevation of temperature was beneficial for the adsorption (Hao et al. 2021;Jia et al. 2017). The positive ΔS • (87.03. J/(mol•K) reflected the increase of reaction disorder in the adsorption process (Chen et al. 2020a;You et al. 2019) and the ΔG • values (-7.31 kJ/ mol) were negative, demonstrating that the adsorption was spontaneous (Liu andLiu 2008, Zhang et al. 2018). As a whole, the Hg (II) adsorption on the 3D-NC was spontaneous, endothermic, and random. Fig. 6A and Fig. 6B show the surface element analysis and adsorption performance of the 3D-NC after five adsorption-desorption cycles, respectively. The results indicated that after five adsorption-desorption cycles, the 3D-NC was still able to retain its surface functional groups and demonstrated excellent adsorption capacity, and the removal rate of Hg (II) only decreased by 10%. And the decline in the removal rate of Hg (II) was majorly caused by the inevitable mass loss in regeneration of the 3D-NC. Generally, the 3D-NC had considerable recyclability in the application for removal of Hg (II). Figure 6C represents the result from selective adsorption experiment and indicated that compared to other metal ions, the 3D-NC had higher remove adsorption capacity for Hg (II) as high as 67.10% which other metal ions could not achieve and the result confirmed that its N-contained structure had highly selective adsorption for Hg (II). The highly selective adsorption mechanism of 3D-NC for Hg (II) was attributed to its abundant N-containing groups in its structure which had a considerable affinity for Hg (II) (Liao et al. 2019;Qu et al. 2019;Tauanov et al. 2018). There was a description related to hard-soft acid base (HSAB) concept, namely, soft acids were prone to combine with soft bases, hard acids preferred to work with hard bases (Fayazi 2020;Manohar et al. 2002). And in the adsorption procedure, N-doped groups in 3D-NC of soft base could more easily interact with metal-hydroxyl species of soft acid, thus 3D-NC presented adsorption affinity for Hg (II). In addition, when the N-doped groups served as softer bases, it were more inclined to bind to soft acid Hg (II) ion than other metals, which was attributed to that the electronic configuration of Hg (II) made itself perform the preference for sp hybrid to form coordinate complexes with N-donor ligands than other atoms (Fayazi 2020, Meyer andNockemann 2003). In addition, the application of mercury ion removal in actual sewage sample was investigated (Supplementary Material)

Adsorption mechanism
The adsorption mechanism of Hg (II) onto the 3D-NC was elaborated in this section. Firstly, its porous structure had potential to diffuse adsorbates to its inner and the adsorbates could be further fixed by functional groups, finishing the adsorption procedure. Moreover, combining with the zeta potential analysis, it could be inferred that electrostatic attraction force also participated in adsorption for Hg (II)  -11.40 ions. According to the adsorption kinetics and adsorption isotherm analysis, the electron exchange or ion exchange played a principal role in the adsorption of Hg (II), which ensured the adsorption stability of the materials. In order to further investigate the role of functional groups in adsorption in more detail, FTIR and XPS were applied to perform the comparison of the 3D-NC before and after adsorption. As could be seen from the FTIR spectra in Fig. 7A, the peaks of -NH/-OH and C-N had shifted from 3440 cm −1 to 3447 cm −1 , from 1417 cm −1 to 1428 cm −1 , respectively, and the C = O/C = N peak at 1620 cm −1 and C = O/C = N peak at 1070 cm −1 had transformed into 1638 cm −1 and 1098 cm −1 , respectively. The above characteristic indicated that the nitrogen-containing groups (-NH, -NH 2 ) and oxygen-containing groups (-OH, -COOH) participated in the adsorption of Hg (II), and the shifts in the FTIR spectra might be caused by the change in the electron distribution due to the electrostatic attraction or ion exchange reaction between the functional groups and Hg (II) (Hu et al. 2018).
Similarly, the full spectrogram of XPS was presented in Fig. 7B, and a unique characteristic peak of Hg4f occurred at 100.0 eV, which indicated that the Hg (II) was effectively trapped by the materials. To further illustrate the surface elements change, the high resolution spectrum of C1s, O1s, N1s, and Hg4f were analyzed (Fig. 7C-F). After adsorption of Hg (II), the high resolution spectrum of Hg4f showed a typical doublet of Hg4f 7/2 at 100.96 eV and Hg4f 5/2 at 104.99 eV. In the high resolution spectrum of N1s, a metal nitride peak at 398.18 eV appeared due to the introduction of Hg (II), which was inferred that the C-N-H bonds exhibited adsorption affinity for mercury ions and formed N-Hg bonds (Liao et al. 2019;Niu et al. 2016;Qu et al. 2019;Song et al. 2016). In the spectra of O1s, the distinct decrease of O-H at 533.34 eV could be observed after adsorption of Hg (II), which implied that the -OH and the -COOH groups could chelated Hg (II) and it was inferred that the formation of chelate rings reduced the free O-H bonds. After the adsorption of Hg (II), in the high resolution spectrum of C1s, the binding energy of the C-O/C-N and C = O/C = N peaks had 0.1 eV increase, which was considered to be caused by the reduction of electron cloud density due to the transfer of lone pair electrons after the adsorption of mercury ions by functional groups around C atoms (Kong et al. 2017). Therefore, combined with the XPS and FTIR results, the synergistic interaction of N, O functional groups supported the chelate or electrostatic adsorption of Hg (II) onto the 3D-NC.

Conclusion
In conclusion, the starch-based 3D porous-layered skeleton of N-doped carbon (3D-NC) was successfully prepared through one-step co-pyrolysis with metal catalyst Ni (II) and Mn (II). It exhibited not only abundant porous structures but also specific adsorption capacity for Hg (II) as high as 403.24 mg/g and finished within 90 min. In addition, the cyclic adsorption experiment and selective adsorption experiment also demonstrated that the 3D-NC had considerable recyclability and highly selective adsorption performance for Hg (II). Through a series of investigations of adsorption behavior, the adsorption process was spontaneous and endothermic, and adsorption force including physical adsorption, electrostatic attraction, ion exchange, and chelation contributed to the excellent removal capacity of Hg (II). On the whole, the remarkable adsorbent 3D-NC had a promising application in water purification and it was expected to promote the development of preparation of starch based biomaterials for multifunctional applications.