Synthesis, adsorption and molecular simulation study of methylamine-modified hyper-cross-linked resins for efficient removal of citric acid from aqueous solution

A series of methylamine-modified hyper-cross-linked resins were fabricated from chloromethylated polystrene-co-divinylbenzene by two continuous reactions (Friedel-Crafts alkylation and amination). The BET surface area and pore volume of the as-prepared resins took a positive correlation to the reaction time and temperature during alkylation reaction while lessened during amination process. When running batch adsorption experiments for adsorption of citric acid, the methylamine-modified resin named HM-65-2 showed higher adsorption capacity of 136.3 mg/g and selectivity of 6.98 (citric/glucose) than the precursor resins. The pseudo-second-order rate model fitted better than the pseudo-first-order model, implying the adsorption sites distributed on the resins surface tended to be heterogeneous. Subsequently, the interactions between citric acid and the resin were investigated by means of molecular simulation. Simulation result showed the addition of nitrogen-containing groups significantly enhanced the adsorption performance of citric acid. Lastly, the dynamic column experiments were performed to obtain the suitable operating conditions for the citric acid adsorption.

microspheres 14 have been tried for recovery of citric acid from aqueous solution, but the low adsorption capacity and complex desorption process limited their further application. Related literature reports were summarized in Table 1, which some adsorption amounts data were calculated based on the batch adsorption data at 298 K. It can be seen that the adsorption capacity of citric acid by ion exchange resins were generally higher than that of these non-ionic adsorbents, but excessive adsorption amounts usually causes more difficulties in desorption process. Therefore, it is beneficial to find a suitable adsorbent with moderate adsorption capacity of citric acid. Fortunately, the hyper-cross-linked poly(styrene-co-divinylbenzene) (PS-DVB) resins, as a typical adsorbent, possessing abundant microporous or mesoporous structure, have attracted substantial attention for the potential separation of organic components owe to its outstanding thermal stability, moderate surface polarity, and high specific surface area.
The hyper-cross-linked polystyrene resins proposed by Davankov et al. 15 are usually synthesized from linear polystyrene or low cross-linked polystyrene by adding bi-functional/multifunctional cross-linking organic reagents such as tetrachloromethane(CCl4), dichloroxylene(DCX) 16 , trifunctional tris-(chloromethyl)-mesitylene(TCMM) and the Friedel-Crafts catalysts including anhydrous zinc chloride, iron (III) chloride and stannic (IV) chloride 17 . In order to avoid using certain ether carcinogens, the hyper-cross-linked resins can also be prepared from macroporous low cross-linked chloromethylated polystyrene by Friedel-Crafts alkylation reactions. Hence, the obtained hyper-cross-linked networks consist of an intensive bridging of strongly solvated polymer chains with conformationally rigid links, leading a major shift of their pore diameter distribution from predominate mesopores to mesopores-micropores bimodal distribution. This also resulted a notable amplify of the Brunauer-Emmet-Teller (BET) surface area and pore volume. Because of these significant changes, the hyper-cross-linked polystyrene resins display excellent adsorption capacities towards non-polar and weakly polar compounds from aqueous solutions. However, the weakness of low capacity to polar substances restrains the applications of hyper-cross-linked resins. In order to enhance the adsorption capacity of polar compounds, the resins can be introduced into some polar comonomers 18 . Moreover, on the purpose of the chemical modification introduced into the resin surfaces, the functionalized hyper-cross-linked resins can also be prepared from the chloromethylated polystyrene according to two continuous procedures. The first produce is that the chloromethyl groups react with another neighboring phenyl ring of the polystyrene by formation of an equivalent number of diphenylmethane-type rigid polymer bridges during a typical Friedel-Crafts reaction, and the second produce is the reaction between the residual chloromethyl groups of the hyper-cross-linked resins and then the functional groups introduced. The residual chloromethyl content of the polymer obtained after the first produce reaction determines the resin pore structure and the amount of subsequent functional-group modification. Thus, it is feasible to adjust the hyper-cross-linked resin morphology and polarity by ordinarily governing the corresponding Friedel-Crafts reaction parameters. Although numerous references have shown that the resins modified with polar functional groups (i.e., amino group) exhibit predominant adsorption capacity to certain weak acids, such as p-aminobenzoic acid 19 , salicylic acid 20 , phenol 21 , and 2-naphthol 22 due to the high BET surface area providing more active sites, there are relatively few reports concerning the adsorption of citric acid on the methylamine-modified hyper-cross-linked resins.
In the present work, six kinds of the hyper-cross-linked resins were synthesized from the commercial available macroporous cross-linked chloromethylated polystyrene resins by regulating the Friedel-Crafts reaction time (1.0, 2.0 and 4.0 h) and reaction temperature (65 and 85 °C). Then these hyper-cross-linked resins were chemically modified by superfluous methylamine reagent to obtain the methylamine-modified resins. Thereafter the adsorption capacity and selectivity of the resins toward citric acid/glucose were evaluated by the batch adsorption experiments, among which one named HM-65-2 was selected as the proper resin to adsorb citric acid from aqueous solution. Afterwards, the interaction energy between citric acid and resins (unmodified or methylamine-modified resins) were analyse by means of density functional theory (DFT) based on B3LYP mode. The result of theoretical computation proved methylamine-modified resins superiority in adsorbing citric acid due to stronger interaction effect, which provides the fundamental information for the future applications.

Experimental Section
Materials. Macroporous low cross-linked chloromethylated polystyrene used as a reaction precursor was purchased from Nankai University Chemical Plant (Tianjin, China). Anhydrous iron (III) chloride, 1, 2-dichloroethane (DCE), methylamine and ethanol were all analytical reagents and used without further purification. Glucose and citric acid (purity ≥ 99%) purchased from Merck (Darmstadt, Germany) were employed as adsorbates in this work.
Synthesis of methylamine-modified hyper-cross-linked polystyrene resins. As shown in Fig. 1, the methylamine-modified hyper-cross-linked polystyrene resins were prepared by two continuous steps and Name of adsorbents Type of adsorbents www.nature.com/scientificreports www.nature.com/scientificreports/ these steps were derived from ref. 23  Characterization of the resins. The Brunauer-Emmett-Teller (BET) surface area and pore volume of the resins prepared were determined by the N 2 adsorption-desorption isotherms using a Micromeritics ASAP 2020 Physisorption Analyzer (Micromeritics, the United States) with the temperature at 77 K and the results were calculated with the relative pressure (P/P 0 ) at 0.05~0.35 according to the adsorption isotherms. The elemental analysis of resin was performed by a VarioEL Elemental Analysis System and the chlorine content of the resins was received as: The Fourier-transformed infrared (FT-IR) spectra of the resins were collected on a Nicolet 510 P FT-IR instrument (Nicolet, the United States) via the potassium bromide technique with the wavenumbers ranged from 500 to 4000 cm −1 .
Theoretical computation. Theoretical computation were carried out with the Materials Studio software 7.0 (BIOVIA, CA) on a P4-2.6 G computer in this part. The assumed complexes between the citric acid and the simplified structure of resins (including unmodified and methylamine-modified resins) were optimized by means of density functional theory (DFT) based on B3LYP mode. The interaction energy (ΔE, kJ/mol) of different system was introduced according to Eq. (2) where E CA (kJ/mol) is the total energy of citric acid molecular structure, E resin (kJ/mol) is the total energy of resins structure and E Total is the total energy of CA-resin structure. Generally, the larger ΔE value means higher energy barrier as well as weaker affinity between the dsorbate and adsorbent 24 .
Batch adsorption experiments. The loading and selectivity of the resins prepared for citric acid and glucose adsorption were evaluated to screen a suitable adsorbent used in the following experiments. In a typical adsorption experiment, 2 g of the tailor-made resins were weighed accurately in a series of flasks with a stopper and then mixed with 10 mL of solution containing 100 g/L citric acid and 10 g/L glucose. Then the flasks were shaken in a water-bath thermostatic oscillator (the agitation speed was set to be 200 rpm) for 4 h at 298 K to guarantee the equilibrium of adsorption process. Subsequently, the resins were filtered and the equilibrium concentration (C e (g/L)) of citric acid and glucose in the solution were determined and the equilibrium adsorption capacity (q e (mg/g)) and selectivity coefficient (α) for the citric acid and glucose were obtained as the following equations 25 : www.nature.com/scientificreports www.nature.com/scientificreports/ where i, j represent the citric acid and glucose adsorbates, q e is equilibrium adsorption capacity (mg/g), C 0 and C e are the initial and the equilibrium concentration of adsobate (g/L), V is the volume of the solution (mL) and W is the weight of the resins (g).
Comparison of the results of q e and α, one of the resins was selected as the proper resin to separate citric acid and glucose. The citric acid adsorption isotherms on the selected resin were performed at three different temperatures (298, 308 and 318 K). 2 g of the resins was completely dispersed in 10 mL of citric acid aqueous solution with the concentration range from 15 to 150 g/L. Then the solutions were continuously shaken in a thermostatic oscillator with identical rotational speed (200 rpm) for 4 h at a desired temperature to ensure the adsorption equilibrium.
The kinetic adsorption of citric acid on the selected resin was performed by analyzing the uptakes on the resin until the equilibrium was reached. 80 g of the resins and 400 ml of citric acid solution with the initial concentration (100 g/L) were quickly introduced into a 500 mL round-bottomed flask. The mixture was continuously stirred at 298 K and 0.5 mL of solution was sampled at different time intervals. The concentration of the residual solution was determined and the adsorption capacity at the contact time t was calculated as 19 : where q t (mg/g) and C t (g/L) are the adsorption capacity and the concentration at contact time t (min).
Column breakthrough experiments. The column breakthrough experiments were performed to better reflect the adsorption behavior of citric acid on the resin. In a typical dynamic experiment, 120 g of wet resins were packed in a glass column. The concentration of citric acid from the effluent was dynamically recorded until it reached the initial concentration. The breakthrough point (C t /C 0 = 0.05) and saturated point (C t /C 0 = 0.95) was measured and the breakthrough capacities (q b , mg/g) and saturated capacities (q s , mg/g) of citric acid on the resins were estimated by the following equations 26 :  Table S2. These results implied that the -CH 2 Cl groups were consumed and transformed to rigid cross-linking bridges after the Friedel-Crafts reaction. After the amino reaction, there appeared a strong and wide vibrational band at 3417 cm −1 representing the N-H stretching from the secondary amino in the amino-modified resins 29 , which confirms the successful uploading of methylamine on the hyper-cross-linked resins. Fig. S3(a,c) display the N 2 adsorption-desorption isotherms of unmodified resins reacted at 65 °C, while Fig. S3(b,d) were the ones reacted at 85 °C. Comparing with the precursor resins, the N 2 uptake of the obtained hyper-cross-linked resins are extremely larger at the same relative pressure, indicating Friedel-Crafts reaction contributes to the porous addition. At a relative pressure below 0.05, the N 2 uptakes increased sharply with increasing of the relative pressure, and at a relative pressure above 0.95, there was an increase of the N 2 uptakes as well. According to the IUPAC classification 30 , these isotherms nearly closed to type-II, which demonstrated that the micropores were predominant and the macopores were also existent. Meanwhile, visible hysteresis loops of the desorption isotherms indicated the existence of a few mesopores. However, after the amination reaction, the decrease of mesopores led to the reduce of the BET surface area of the resins (Table S2), which was in agreement with the conclusion reported in ref. 31 .
The BET surface area of the precursor resin was only 31.88 m 2 /g, while abundant porous structure were produced after the Friedel-Crafts reaction, causing a significant increase of the BET surface area and pore volume. This increase may be assigned to contribute from the formation of the methylene cross-linking bridges in the Friedel-Crafts reaction. Longer reaction time, as well as higher reaction temperature, would bring about more www.nature.com/scientificreports www.nature.com/scientificreports/ cross-linking bridges 32 . While methylamine-modified resins exhibited significantly decrease of the BET surface area and pore volume, which may be from the partial hole collapse during the polar modification process 22,31 . Furthermore, this phenomenon could also be explained with an increased modified solvent when uploaded with methylamine 33 . During the amination process, the original mesoporous structures in the resin were filled with methylamine molecules and formed some closed pore structures. The combined effect of pore collapse and modifiers made the surface areas of the final methylamine-modified resins were smaller than that of the hyper-cross-linked resins. Figure 3(a,b) reveal the adsorption capacity (q e , mg/g) and selectivity of citric acid/glucose (α) onto the unmodified resins and methylamine-modified resins. The adsorption capacity of citric acid is proportional to the Friedel-Crafts reaction time and temperature for the unmodified resins, indicating more favorable binding sites were formed as the increase of reaction time and temperature. Whereas the adsorption selectivity of citric acid and glucose onto the unmodified resins appears were insensitive to the reaction time and temperature. After the methylamine modification, the q e and α value of the modified resins exhibit sharp increase comparing to the unmodified resins, and the reaction temperature and time also have effect for this increase. Although the BET surface area possessing numerous adsorption sits significantly reduced after methylamine-modified, the introduced amino groups would provide other stronger interactions, such as hydrophobic interaction, electrostatic interaction or hydrogen bonding 34 . These stronger interactions are sufficient to enhance the adsorption capacity and selectivity of citric acid. Comparing the results depicted in Fig. 3, the methylamine-modified resins named HM-65-2, occurring Friedel-Crafts reaction at 65 °C during 2 h, showed higher adsorption capacity of 136.3 mg/g and selectivity of 6.98 (citric/glucose) than others. Consequently the HM-65-2 resin was selected to complete the other adsorption experiments for removing citric acid from aqueous solutions.

Comparison of resin adsorption properties.
Furthermore, a compare experiment of commercial weak basic anion-exchange resin (Amberlite FPA53) was carried out to illuminate the superiority of the tailor-made methylamine-modified resin (more experimental details can see in Supplementary materials). Results shown in Fig. S4 indicated that the adsorption capacity of citric acid on the weak anion-exchange resin reached the maximum value (378.1 mg/g), which was higher than that of HM-65-2 resin (136.3 mg/g). But the desorption rates (<40%) were lower than that of HM-65-2 www.nature.com/scientificreports www.nature.com/scientificreports/ resin (>80%) when only using hot water (65 °C) as an eluent. This enormous difference can be attributed to the stronger acid-base interaction between the ion-exchange resin and citric acid. Higher desorption rate presented the superiority of the tailor-made HM-65-2 resin, that is, a benign separation routine may be developed to purify citric acid from its fermentation broth by using only hot water as an eluent in future. Figure 4 presents complexes of the CA-unmodified resin and CA-methylamine-modified resin. It could be seen that the carboxyl of citric acid was biased toward the nitrogen-containing group of the resin after methylamine modification while there was not of the  www.nature.com/scientificreports www.nature.com/scientificreports/ unmodified resin. Calculated interaction energy shown in Table 2 also proved this. The interaction energy of CA-methylamine-modified resin was smaller than that of CA-unmodified resin, including that the citric acid was preferentially adsorbed onto the methylamine-modified resin and this preference could attribute to the amino group with stronger electrostatic attraction.

Results of molecular simulation.
Adsorption isotherms. Figure 5 shows the adsorption isotherms of citric acid on the HM-65-2 resins from aqueous solution with temperature of 298, 308 and 318 K. It could be clearly observed that the adsorption capacity took a positive correlation to the equilibrium concentration and a negative correlation to the temperature, implying that the adsorption was an exothermic process. Furthermore, two typical isotherm models, the Langmuir and Freundlich models were used to analyze the experimental data. Two equations are described by linear as: where C e is the equilibrium concentration in the solution (g/L), q e is the equilibrium adsorption capacity of citric acid onto the resins (mg/g), and a [(mg L)/g 2 ], b (mg/g), K F [(mg/g)(L/mg) 1/n ] as well as n are the characteristic constants.  Table 2. Calculated interaction energy of (a) the CA-unmodified resin and (b) CA-methylamine-modified resin. www.nature.com/scientificreports www.nature.com/scientificreports/ The fitting curves are shown in Fig. 5 and the corresponding parameters a, b, K F and n, as well as the correlation coefficients R 2 by the linear method are summarized in Table 3. It reveals that both Langmuir and Freundlich equations are suitable for prediction of the adsorption data since R 2 > 0.970. The Freundlich isotherm appears to be more appropriate than the Langmuir isotherm due to the higher R 2 . The larger characteristic constants (n > 1) indicating the preferential nature of this adsorption process infer that the adsorption sites distributed on the resins surface tends to be heterogeneous because of the polar modification 35 . Furthermore, in the Langmuir model, b is related to the bonding energy coefficient 36 . As the temperature increases, the b value becomes smaller, indicating the force between the adsorbent and the adsorbate decreases, which testified its exothermic.
Some important thermodynamic basic parameters such as the Gibbs free energy change (ΔG, kJ/mol), adsorption enthalpy (ΔH, kJ/mol), and adsorption entropy (ΔS, J.mol −1 .K −1 ) are obtained according to the following expressions where R is the universal gas constant (8.314 J.mol −1 .K −1 ), T is the solution temperature (K), and K C , the thermodynamic equilibrium constant, can be determined by plotting ln (C s /C e ) versus C s and extrapolating C s to zero 37 .
The ΔG for the adsorption of citric acid on the HM-65-2 resins can be attained from substitution calculation and the ΔH and ΔS values figured out according to the slope and intercept of plotting ΔG versus T are tabulated in Table 4. It is evident that ΔH (−7.792 kJ/mol) is negative ignoring the influence of temperature, suggesting that citric acid adsorption on HM-65-2 resin is exothermic process and the adsorption is not solely simple physical process while it may involves a weakly bonding force formation 38 . The negative ΔG at all temperature are resulted from the spontaneous nature of the adsorption process on resin polar modified and the positive ΔS (20.35 J.mol −1 .K −1 ) indicates the increase slightly of system randomness at the solid-solution interface during adsorption 39 .
Adsorption kinetics. Figure 6 displays the kinetic curves of citric acid adsorption on the HM-65-2 resin from aqueous solution at 298, 308 and 318 K. The samples had a relatively fast adsorption rate in the initial minutes followed by a raised plateau, implying that the methylamine-modified resins display an excellent kinetic property for citric acid and suggesting that the process likely happens through intermolecular forces 40 . Meanwhile, the required time for the equilibrium was gradually shorter with increasing temperature, suggesting that a higher temperature induces a faster adsorption rate. This phenomenon might owe to the increasing mobility of citric acid and the number of molecules with sufficient energy under higher temperature, and they were more easily undergo an interaction with the active sites at the resin surfaces.
In order to study the mechanism of the adsorption kinetic process, three commonly adsorptions kinetic models, the pseudo-first-order model, pseudo-second-order model and Weber-Morris model were used to fit the kinetic data (seen Fig. 7), equations are shown as follows: where K 1 (min −1 ), K 2 (g.mg −1 .min −1 ) and K 3 (mg.g −1 .min −1/2 ) are the constants of pseudo-first-order, pseudo-second-order and Weber-Morris models, respectively. According to the corresponding model parameters   www.nature.com/scientificreports www.nature.com/scientificreports/ and the correlation coefficient (R 2 ) shown in Tables 5 and 6, the adsorption kinetic data could be fitted well by the pseudo-second-order model, and the calculated equilibrium adsorption capacities are relatively close to the real experimental results, suggesting the chemisorption would be the rate-control mechanism 41 .

Effect of operating parameters on breakthrough curves
Effect of feed flow rate. Three feed flow rates (0.5, 1 and 2 mL/min) were performed while fixing the initial citric acid concentration of 100 g/L, the bed height of 38.5 cm and the bed inner diameter of 2.4 cm. The breakthrough curves with C t /C 0 against time t shown in Fig. 8(a) become steeper at a higher flow rate. As shown in Table S3, the more adsorption capacity would be achieved at a slower flow rate. It could be explained by the  www.nature.com/scientificreports www.nature.com/scientificreports/ decrease of the preserved time at high flow rate. This effect caused by higher perturbation was easier to weaken the interaction between adsorbate molecules and adsorbent particles, resulting in the decrease of adsorption 42,43 . However, considering the lower rate often accompanies excessive adsorption time and unnecessary instrumental spoilage, 1 mL/min will be more suitable from the practical operation point of view.
Effect of initial concentration. The effects of initial concentrations on the citric acid adsorption behaviors were investigated at the feed concentration of 33, 50 and 100 g/L, respectively. The flow rate was remained at    Effect of bed height. The citric acid breakthrough curves with bed heights of 38.5, 30 and 25 cm are described in Fig. 8(c) (the flow rate was 1.0 mL/min, bed inner diameter was 2.4 cm and feed concentration was 100 g/L). The retention time was reduced corresponding to the reduction of column height, whereas the loading capacities were almost identical (see Table S3) and it could be interpreted by the positive proportion between the retention time and column height. Reduction of the column height led to fewer adsorbent sits available loaded in the direction along the mobile phase moving, which shorten the retention of the adsorbate on the resins 44 . However, the loading capacity in terms of adsorption amount per mass of resins (mg/g) would be nearly unchanged.
Effect of bed inner diameter. The citric acid breakthrough curves with bed inner diameters of 2 and 2.4 cm are presented in Fig. 8(d) with the bed height to be 38.5 cm, the feed flow rate to be 1 mL/min and the initial concentration to be 100 g/L, and the related characteristic parameters for each experimental breakthrough curve were given in Table S3. It was clear that the breakthrough time (t b ) increased with increasing D c , which may owe to the fact that a larger number of resins existed along the cross sectional area of the column at a wider diameter column which was beneficial to the contact of citric acid molecules to the resins, resulting in a decrease of the mass transfer zone 44 . Analyzing comprehensively these above results based on Table S3, the flowrate of 1 mL/min, the initial concentration of 100 g/L, the height of 38.5 cm, and the diameter of 2.4 cm would be the feasible manipulative parameters.

conclusions
During this study, we successfully tailored the pore textural property and surface functionality of the polymeric adsorbent to enhance its adsorption performance to citric acid through controlling the related parameters of the two continuous reactions (Friedel-Crafts alkylation and amino modification). The samples synthesized were characterized with N 2 adsorption-desorption, FT-IR and analysis of residual chlorine content. Thereafter, the screened resin (named HM-65-2) was evaluated for adsorption of citric acid according to its adsorption capacity and adsorption selectivity. Both the Langmuir and Freundlich models correlate the equilibrium data well and the calculated thermodynamic parameters indicated that the adsorption process was a spontaneous and exothermic process. The adsorption kinetics of citric acid on the resin could be well correlated by the Pseudo-second-order model and the column dynamic adsorption experiments show that the elution flow rate, injection concentration, column height and inner diameter have significant effects on the adsorption performance of citric acid. Meanwhile, the force mechanism was explored from the molecular simulation perspective. It can be concluded that the adsorption behaviors of citric acid molecules on the methylamine-modified resins would be stronger.
Due to the unique properties in terms of proper adsorption capacities and high selectivity, HM-65-2 resins could be an attractive candidate in citric acid purification and further employed in the simulated moving bed separation process.