Preparation of carbon nanotubes/porous polyimide composites for effective adsorption of 2,4-dichlorophenol

Abstract

The novel carbon nanotubes (CNT)/porous polyimide (PI) composites were prepared via blending and were characterized by Fourier transform infrared spectroscopy (FTIR), thermogravimetric analysis (TGA), transmission electron microscopy (TEM) and nitrogen adsorption–desorption techniques. The feasibility of CNT, CNT/PI (3/1) and CNT/PI (1/1) to remove 2,4-dichlorophenol (2,4-DCP) from an aqueous solution was examined. The influence of different parameters such as pH, ionic strength, contact time and temperature has been carried out. The results showed that PI was successfully combined with CNT and formed CNT/PI composites with higher adsorption capacity and shorter equilibrium time. The adsorption capability at 298 K of CNT, CNT/PI (3/1) and CNT/PI (1/1) composites were 330.0 mg g⁻¹, 412.5 mg g⁻¹ and 506 mg g⁻¹, respectively. Their kinetic experimental data were better fitted by a pseudo-second-order model, and their adsorption isotherms could be represented by the Freundlich isotherm model perfectly. A thermodynamic study showed that the adsorption of 2,4-DCP on them was spontaneous and endothermic. Discoveries in this study suggest that the CNT/PI composites are promising adsorbents for removal of 2,4-DCP from aqueous solution.

---

1. Introduction

Chlorophenols, generated during the chlorine bleaching of pulp, in the textile industry and drinking water chloridizing disinfection, are identified as one of the common contaminants in industrial wastewater.^1,2 They are poorly degradable, suitably fat-soluble, and highly toxic, which leads to them endangering human health by food chain accumulation.³ Therefore, the rational and efficient treatment of these chlorophenol pollutants has been regarded as one of the most urgent and important global problems.

Various methods like adsorption, electrochemical oxidation, membrane filtration, solvent extraction and photo-catalytic degradation were reported for the removal of chlorophenols from wastewater.^4–7 Among these physical and chemical technologies, adsorption is regarded as the most popular one owing to its ease of operation, high efficiency, and eco-friendliness. Some adsorbents such as active carbon^8–10 and carbon nanotubes (CNT)^11,12 have been widely used for removal of chlorophenols and other noxious impurities. For example, Shaarani et al. prepared the surface modification of the activated carbon using ammonia, and the result showed that the modified activated carbon increased the adsorption capability for 2,4-dichlorophenol (2,4-DCP) from 232 to 285 mg g⁻¹.⁸ As well known, the CNT is a fascinating member of the carbon family due to the unique morphologies, large surface area, and the remarkable thermal and mechanical properties. A comparative study between CNT and powdered activated carbon (PAC) for adsorption of trihalomethanes (THMs) from water was conducted by Lu et al., and they discovered that CNT possess highly potential applications for THMs removal.¹³ Long and Yang showed that the CNT owns a significantly higher dioxin removal efficiency than the activated carbon.¹⁴ Li et al. found that CNT are good fluoride adsorbent and their fluoride removal capability is superior to activated carbon.¹⁵ In addition, metal and metal oxide nanoparticles were used to combine with CNT to apply for chlorophenols removal. Xu et al. reported simultaneous adsorption and dechlorination of 2,4-DCP by multi-walled carbon nanotubes (MWNTs)-stabilized Pd/Fe.¹⁶ Atieh's group used aluminum oxide impregnated carbon nanotubes (CNT–Al₂O₃) as adsorbent for 4-chlorophenol and phenol removal from aqueous solutions. They found that CNT–Al₂O₃ showed better adsorption efficiency than CNT due to the increase in the surface area from 155.5 m² g⁻¹ of CNT to 227.5 m² g⁻¹ of CNT–Al₂O₃.¹⁷ Nevertheless, most of these CNT-based adsorbents suffer from either low adsorption capacities or slow adsorption kinetics partially. Thus, there is an urgent need to develop novel adsorbent with rich functional groups and open pore structure for increased adsorption efficiency and enhanced adsorption rate. In our previous work, porous polyimide (PI) was synthesized and used as adsorbent for the removal of azo dye and antibiotic, which exhibits a high adsorption capacity due to its large surface area, vast micro/mesopores, and multiple amine groups.¹⁸ Therefore, we want to take the advantage of both PI and CNT by blending them into composite nanoparticles.

In this paper, for the first time, CNT were used as a porous-based support material for porous PI particles. To obtain a basic understanding of adsorption ability of the composites towards organic pollutants for developing the high-performance adsorbent, we have selected 2,4-DCP as the removal objective. As well known, 2,4-DCP is highly toxic in nature and cause severe health effect on organisms even at low concentrations, hence it is classified by US Environmental Protection Agency (EPA) as priority pollutants.¹⁹ Batch adsorption experiments were systematically performed to study effect of solution pH, ion intensity, contact time and temperature. Furthermore, the adsorption kinetics, equilibrium isotherms, and thermodynamics were also evaluated.

2. Material and methods

2.1 Materials

Carbon nanotubes (CNTs) were purchased from Shenzhen Nanotech Port Co. Melamine (MA) and pyromellitic dianhydride (PMDA) were dried at 100 °C under vacuum. Dimethyl sulfoxide (DMSO) was distilled in vacuum after drying over CaH₂. Other solvents and reagents were used as received. 2,4-DCP was obtained from Aladdin, and other reagents were obtained from Sinopharm Chemical Reagent Co., all medicals were analytical grade. Ultra-pure water was employed in all the experiments.

2.2 Preparation of PI and CNT/PI composites

PI was synthesized via thermal polycondensation according to our previously used method.¹⁸ In order to obtain abundant amino-terminated porous polyimide, the molar ratio of amine and anhydride functional groups was 3 : 1. Briefly, MA (12 mmol), PMDA (6 mmol) and DMSO (30 mL) were added to 50 mL single-necked flask. The mixture was stirred 0.5 h at room temperature under inert conditions. Then, 2 mL of toluene was added, and the mixture was heated at 180 °C with a Dean-Stark apparatus for 72 h. After cooling to room temperature, the white solid was isolated and washed with acetone, tetrahydrofuran and methylene chloride for several times. Finally, the solid was dried in vacuum oven at 100 °C overnight in about 62% yield.

The composites of CNT/PI were prepared via blending. PI and CNT were first dispersed in ethanol solution, and then the mixture was stirred vigorously for 0.5 hour by ultrasound and then magnetic stirred for 12 hours to facilitate uniform dispersion of CNT and PI. The CNT/PI was collected by centrifugation and finally dried 4 hours at 50 °C in vacuum oven. The prepared CNT/PI composites were carried out using different ratios of varying CNT/PI weight ratios of 1/1 and 3/1, respectively.

2.3 Characterization

Fourier transform infrared (FT-IR) spectra in the range 4000 cm⁻¹ to 500 cm⁻¹ was measured on a FT-IR spectrometer (Vertex 70, Bruker, USA). The microstructure of CNT and CNT/PI composites was examined by a Tecnai G220 transmission electron microscope (TEM) (FEI, Netherlands). The thermogravimetric analysis (TGA) was performed using TGA (PerkinElmer Instruments) thermal analyzer system at the heating rate of 10 °C min⁻¹ under N₂ condition. Nitrogen adsorption–desorption isotherms were obtained on a Micrometrics 2020HD88 (Micromeritics Instrument Co., USA) apparatus at 77 K.

2.4 Adsorption experiment

Before adsorption, CNT and CNT/PI composites were dried in vacuum oven at 100 °C for 12 h and kept in a desiccator. The adsorbents (10 mg) were weighed accurately and added to the aqueous solutions of 2,4-DCP (25 mL) with concentrations between 20 and 1000 mg L⁻¹ at pH 6. The mixtures were shaken in an oscillator for 4 h at 180 rpm and 298 K. Each experiment was made in a centrifuge tube and the initial and final concentrations of 2,4-DCP were analyzed via a UV-vis spectrophotometer at the calibrated maximum wavelength of 284 nm. The adsorption capability (q_e) was calculated by the following equations:where C₀ is the initial concentration of 2,4-DCP solution (mg L⁻¹), C_e is the equilibrium concentration of 2,4-DCP solution (mg L⁻¹), V is the total volume of solution (mL), and m is the CNT or CNT/PI dosage (mg).

The effect of solution pH on the adsorption capability was investigated in the range of 3 to 9, where the pH of the stock solution was adjusted by adding a certain amount of 0.1 M hydrochloric acid and 0.1 M sodium hydroxide. At pH 6, the effect of ion intensity was also studied by gradually increasing KCl concentration in 2,4-DCP solution from 0.2 g L⁻¹ to 1.0 g L⁻¹. To explore the adsorption rate and dynamic characteristics, the dynamic experiments were conducted by varying the contact time from 0 min to 1440 min at three various initial 2,4-DCP concentrations (50 mg L⁻¹, 200 mg L⁻¹, 500 mg L⁻¹), respectively. To present the thermodynamic features, the isothermal adsorption experiments were done at three different temperatures (298 K, 308 K, and 318 K) over a range of initial 2,4-DCP concentration from 20 mg L⁻¹ to 1000 mg L⁻¹.

2.5 Adsorption isotherm models

The Langmuir model is obtained under the ideal assumption of a totally homogenous adsorption surface, and the adsorption is monolayer coverage, whereas the Freundlich isotherm is suitable for a highly heterogenous surface and is not restricted to the formation of a monolayer.²⁰

The Langmuir and Freundlich models isotherm equations can be expressed as follows:

where q_e is equilibrium adsorption capability of the 2,4-DCP in the adsorbed phase (mg g⁻¹), C_e is equilibrium concentration of the 2,4-DCP in the liquid phase (mg L⁻¹), q_m is maximum adsorption capability corresponding to complete monolayer converge (mg g⁻¹), K_L (L mg⁻¹) is Langmuir constant related to the adsorption energy and adsorption capability, K_F (mg^1−1/n L^1/n g⁻¹) and 1/n are Freundlich constants representing the adsorption capacity and adsorption intensity, respectively.

2.6 Kinetics study

Adsorption kinetics is one of the main indicators reflecting the adsorbent efficiency and determining potential application in a large extent. The pseudo-first-order and pseudo-second-order dynamic model were used to describe the dynamic behavior and explore the adsorption mechanism.¹² Expression of these kinetics models are as follows:

q_t = q_e(1 − e^−k₁t) (4)

where q_t is adsorption capability of the 2,4-DCP in the adsorbed phase vary with time (mg g⁻¹), q_e is equilibrium adsorption capability of the 2,4-DCP in the adsorbed phase (mg g⁻¹), t is adsorption time, k₁ (min⁻¹) is the rate constant of pseudo-first-order model, k₂ (g mg⁻¹ min⁻¹) is the rate constant of pseudo-second-order model.

2.7 Thermodynamics study

The thermodynamic parameters such as change in standard Gibbs free energy (ΔG), enthalpy (ΔH) and entropy (ΔS) were determined in this study. It should be considered whether the adsorption process can occur spontaneously in the actual process.¹² The ΔG is the basic standard to judge spontaneity. The apparent equilibrium constant (K₀) can be expressed as follows:where C_ad,e is the quantity of 2,4-DCP (mg) adsorbed on the adsorbent per liter of the solution at equilibrium and C_e is the equilibrium concentration (mg L⁻¹) of 2,4-DCP in solution. The value of K₀ can be obtained in the lowest experimental 2,4-DCP concentration.^21,22 The K₀ value is used in the following equation to determine the Gibbs free energy of adsorption.

ΔG = −RT ln K₀ (7)

ΔG = ΔH − TΔS (8)

where ΔG is the Gibbs free energy (kJ mol⁻¹), ΔH is the enthalpy (kJ mol⁻¹), ΔS is the entropy (J mol⁻¹ k⁻¹).

3. Results and discussions

3.1 Characterization of adsorbents

Fig. 1 shows the TEM images of CNT, PI, CNT/PI (1/1) and CNT/PI (3/1). Obviously, the clear observation of CNT could be achieved in Fig. 1a. The CNT monomer consists of snake-like nanotubes with diameters of 10–20 nm. From Fig. 1b, it shows that the PI is exclusively made up of nanoparticles and appears to be highly agglomerated. The combination of the PI and CNTs can be observed in Fig. 1c and d. The PI particles are wrapped by the CNTs bundles, which results in the formation of CNT/PI composites. Aggregation of CNT/PI composites will result in closed interstitial spaces in its aggregates, while CNTs with snake-like nanotubes cannot form closed interstitial spaces in their aggregates due to their length.^23,24 These interstitial spaces are the pores of CNT/PI composites that are possibly available for adsorption.

Fig. 1 TEM images of CNT (a), PI (b), CNT/PI (1/1) (c) and CNT/PI (3/1) (d).

Fig. 2 shows the FT-IR spectra of PI, CNT, and CNT/PI composites. The spectrum of CNT does not exhibit any visible peaks in the range of 4000 cm⁻¹ to 500 cm⁻¹ owing to no groups on the surface of CNT. In the FT-IR spectrum of PI, the three characteristic absorptions bands of polyimide can be observed at 1730 cm⁻¹ (C=O symmetrical stretching), 1350 cm⁻¹ (C–N stretching), and 737 cm⁻¹ (C=O bending), revealing the formation of imide bond.²⁵ The distinct bands 1545 and 1480 cm⁻¹ found in the spectra of polymer are assigned to the quadrant and semicircle stretching of the triazine ring, respectively, suggesting the successful incorporation of the melamine into the polymer.²⁶ The band at 980 cm⁻¹ corresponds to C–N stretching of the triazine ring.²⁷ The band around 2922 cm⁻¹ is distributed to the terminal amine groups of triazine in resonance structure.²⁸ The strong band around 3417 cm⁻¹ is assigned to the terminal amine groups since the raw material MA was used in excess.²⁹ Meanwhile, all of these adsorption peaks appear in the FT-IR spectrum of CNT/PI composites. This observation suggested that the PI was successfully combined with CNT and formed the CNT/PI composites.³⁰

Fig. 2 FT-IR spectra of CNT (a), PI (b) and CNT/PI (1/1) (c).

The TGA curves of CNT, PI, and CNT/PI composites are shown in Fig. 3. From Fig. 3a, the weight loss is found to be 5.7 wt% at 800 °C, which might due to the decomposition of impurities, suggesting that both the thermal stability and purity of CNT is high. In the TGA curve of PI and the CNT/PI composites, the weight does not reduce dramatically until 450 °C due to the excellent thermal stability of the polyimide polymers.³¹ And the weight losses between 200 °C and 450 °C should be ascribed to the degradation of PI. An additional weight loss occurs at above 450 °C due to the further condensation of carbonaceous surface species.³² Finally, the residue yields of CNT, PI, CNT/PI (3/1) and CNT/PI (1/1) are about 94.3 wt%, 78.1 wt%, 52.1 wt% and 14.4 wt%, respectively, which are related to the fixed carbon content of materials. According to the data, the weight of PI in CNT/PI (3/1) and CNT/PI (1/1) composites are estimated to be about 25 wt% and 50 wt%, respectively.

Fig. 3 TGA for CNT (a), CNT/PI (3/1) (b), CNT/PI (1/1) (c) and PI (d).

The pore structure and surface area of CNT, PI and CNT/PI composites were investigated by the N₂ adsorption–desorption isotherms at 77 K, and their porosity properties including Brunauer–Emmett–Teller (BET) surface areas and pore volume were listed in the Table 1. As indicated in Fig. 4a. CNT, PI and CNT/PI composites show a type II isotherm with a small H3 hysteresis loop in the P/P₀ range of 0.7 to 1.0. The H3 hysteresis loop expresses the aggregation of adsorbents containing slit shaped pores. Pore size distributions of the adsorbents, calculated via the Barrett–Joyner–Halenda (BJH) method based on desorption curves, was presented in the Fig. 4b. The peaks centered at ca. 4 and 52 nm correspond to the inner cavities of CNT and PI, respectively. For CNT/PI composites, there were two major peaks located in the size range of 3–4 nm and 30–36 nm, which should be attributed to the pores resulting from the aggregation of CNT and PI. According to Table 1, the CNT (223 m² g⁻¹; 0.113 cm³ g⁻¹) shows a lower surface area and pore volume than CNT/PI (3/1) composites (261 m² g⁻¹; 0.227 cm³ g⁻¹) and CNT/PI (1/1) composites (340 m² g⁻¹; 0.300 cm³ g⁻¹). The higher BET specific surface areas and pore volume of the CNT/PI composites are the positive factors contributing to excellent adsorption performance.

Table 1 The BET surface area and pore volume of CNT, PI and CNT/PI composites

Pore properties	CNT	CNT/PI (3/1)	CNT/PI (1/1)	PI
SBET/m^2 g^−1	223	261	340	566
V/cm^3 g^−1	0.113	0.227	0.300	0.628

---

Fig. 4 (a) Nitrogen adsorption–desorption isotherms and (b) BJH pore size distribution for CNT, PI and CNT/PI composites.

3.2 Adsorption evaluation

3.2.1 Effect of pH. One of the most important factors which affect the adsorption of pollutes from aqueous solution is pH because the variation in pH influences the surface charge of the adsorbent materials, consequently alters their adsorption capability. The effect of pH on the removal of 2,4-DCP was investigated in Fig. 5. The removal efficiency of 2,4-DCP in this study was evaluated at pH from 3 to 9 in order to emulate the real wastewaters. The pH effect was researched using the adsorbent material dosage of 10 mg, temperature of 298 K, contacting time of 4 hours and concentration of 50 mg L⁻¹. It can be seen that adsorption capability for 2,4-DCP removal decreases with rise in pH value. Increased pH generally leads to rising ionization and hydrophilicity, which weakens adsorption capacity of CNT towards organic pollutes.³³ This trend is more obvious for CNT/PI composites compared with CNTs, which was attributed to the types and ionic state of the vast amine groups on the surface of the PI. Given the 2,4-DCP is a weak acid (pK_a = 7.85), at low pH, the adsorbed 2,4-DCP was mostly in the nonionized form, so there was no electrical repulsion occurred between the 2,4-DCP and the adsorbents. The 2,4-DCP could be adsorbed onto CNT and CNT/PI composites surface through π–π interaction, which is attributed to the dispersion interaction between the π electrons in the aromatic ring of 2,4-DCP and the carbonaceous adsorbent graphite layers. The scheme of adsorption mechanism was shown in Fig. 6. Moreover, the hydrogen bonding interaction was also formed between 2,4-DCP molecule and amide groups of CNT/PI composites, which was led to higher adsorption efficiency. When pH value was beyond 7, the adsorption capability decreased sharply due to the repulsive force prevailing at a higher pH value.³⁴ Considered that most of the 2,4-DCP molecules and adsorbents were present in their negative ions at high pH value, it gives rise to the repulsion force between negatively charged surface of the absorbents and 2,4-DCP.^35–37 And this phenomenon was also observed in the adsorption behavior of PLAC and PLAC/Mn composite towards 2,4-DCP.¹⁰ The adsorption capacity of CNT/PI (1/1) composites was higher than that of CNT/PI (3/1) composites and CNT for 2,4-DCP removal at all pH.

Fig. 5 Effect of pH on the adsorption of 2,4-DCP onto CNT, CNT/PI (3/1) and CNT/PI (1/1).

Fig. 6 Adsorption mechanism of 2,4-DCP on the CNT/PI composites.

3.2.2 Effect of ion intensity. At pH 6.0, the effect of ionic strength was also investigated by gradually adding KCl concentration in 2,4-DCP solution from 0.2 g L⁻¹ to 1 g L⁻¹. From Fig. 7, it is clear that no obviously change occurs by adding KCl. The pH-dependent but ionic strength independent adsorption suggests that non covalent bonds interactions, such as electrostatic attraction, hydrogen bonding, or hydrophobic attraction, are responsible for the adsorption. So the ions can't compete with functional groups of 2,4-DCP molecules onto the absorbent. This also indicates that the CNT and CNT/PI composites can be used for the removal of chlorophenols from salt-containing wastewater.

Fig. 7 Effect of ion intensity on the adsorption of 2,4-DCP by using CNT (a), CNT/PI (3/1) (b) and CNT/PI (1/1) (c).

3.2.3 Effect of contact time. The effect of contact time on the adsorption capacity of 2,4-DCP at different initial concentrations (50 mg L⁻¹, 200 mg L⁻¹, and 500 mg L⁻¹) were shown in Fig. 8. It is found that 2,4-DCP adsorption is rapid at all the investigated initial concentrations. Especially for CNT/PI (1/1) composite, it can be observed that 30 min is sufficient for the 2,4-DCP adsorption to reach adsorption equilibrium at a low initial concentration (50 mg L⁻¹). This may be attributed to the fact that the CNT/PI composites are dispersed uniformly in low concentration solution and vast amine groups of PI improves the hydrogen bonding interaction between CNT/PI and 2,4-DCP molecule. When the initial concentrations were up to 200 mg L⁻¹ and 500 mg L⁻¹, the adsorption equilibrium time were about 60 min and 90 min for CNT/PI (1/1) composite, respectively. These results indicates that the adsorption of 2,4-DCP from the aqueous solution is dependent on its initial concentration. When the adsorption on the exterior surface reached saturation, the 2,4-DCP diffused into the pores of the adsorbents and was adsorbed on the interior surface, which lead to a relatively longtime.³⁸ For the CNT and CNT/PI (3/1) composites, it also shows an increase in removal efficiency with an increase in adsorption contact time. And the adsorption equilibrium time was 90 min, 120 min and 180 min for CNT, and 60 min, 90 min, and 120 min for CNT/PI (3/1) at concentration 50 mg L⁻¹, 200 mg L⁻¹ and 500 mg L⁻¹, respectively. On the other hand, to attain the same adsorption capacity at the same initial concentration, the order of equilibrium contact time is CNT/PI (1/1) < CNT/PI (3/1) < CNT. This observation is attributed to the vast amine groups and pores of CNT/PI composites.

Fig. 8 Effect of contact time on the adsorption of 2,4-DCP at different initial concentrations by using CNT (a), CNT/PI (3/1) (b) and CNT/PI (1/1) (c).

3.2.4 Effect of temperature. The temperature is considered as a critical factor which may affect the adsorption process greatly. Generally, high temperature can increase the rate of diffusion of the adsorbate molecules in the solution to the external and internal surface of the adsorbents, and may change the equilibrium adsorption capacity of the adsorbents for a particular adsorbate.^39,40 The effect of temperature on adsorption capability of the absorbents towards 2,4-DCP was carried out by varying temperatures from 298 K to 318 K. The results are presented in Fig. 9. It is significant that the adsorption capacity increases as the temperature rises. For example, when 2,4-DCP concentration is 1000 mg L⁻¹, the adsorption capacity of 2,4-DCP onto CNT/PI (1/1) composites increase from 506 mg g⁻¹ to 534 mg g⁻¹, and 567 mg g⁻¹ by increasing the temperature from 298 K to 308 K, and 318 K, respectively. This results suggested that adsorption and removal of 2,4-DCP from aqueous solution is an exothermic process, which will be discussed in detail at the Thermodynamic section.

Fig. 9 Effect of temperature on the adsorption of 2,4-DCP by using CNT (a), CNT/PI (3/1) (b) and CNT/PI (1/1) (c).

3.3 Adsorption kinetics, isotherm and thermodynamics study

Adsorption kinetics provides information about the mechanism of adsorption, which is important for efficiency of the adsorption process. Adsorption kinetics can be elucidated by pseudo-first-order and pseudo-second-order models. The adsorption kinetics by using the nonlinear method at three different initial concentrations of 2,4-DCP on the absorbents are shown in Fig. 10. The values of the two kinetic model constants (k₁, k₂), the theoretically calculated equilibrium adsorption capacities (q_e,calc,1, q_e,calc,2), along with the error function values (R²) are listed in Table 2. From the result, it is clearly seen that the equilibrium adsorption capacities q_e,calc,2 from the pseudo-second-order model are much close to the experimental data (q_e,exp). For CNT/PI (1/1) composite, the R² values (0.9987 to 0.9992) for pseudo-second-order equation are closer to unity than those (0.9965 to 0.9986) for pseudo-first-order equation, so it is concluded that pseudo-second-order kinetics model fitted the experiments better. In addition, the equilibrium adsorption capability of the 2,4-DCP (q_e) represented adsorption ability of the absorbent, followed an order CNT/PI (1/1) > CNT/PI (3/1) > CNT, which was consistent with the above study in Section 3.2.

Fig. 10 Nonlinear fitting of pseudo-first-order and pseudo-second-order models for 2,4-DCP adsorption on CNT (a), CNT/PI (3/1) (b) and CNT/PI (1/1) (c).

Table 2 Pseudo-first-order and pseudo-second-order kinetic parameters for the adsorption process

Material	Pseudo-first-order model					Pseudo-second-order model
	C0 (mg L^−1)	qe,exp (mg g^−1)	k1 (min^−1)	qe,calc,1 (mg g^−1)	R^2	k2 × 10^−3 (g mg^−1 min^−1)	qe,calc,2 (mg g^−1)	R^2
CNT	50	47.62	0.02486	49.23	0.9868	0.83	47.17	0.9901
	200	113.69	0.02326	112.74	0.9331	0.35	113.39	0.9679
	500	240.20	0.0695	252.08	0.9591	0.26	245.21	0.9690
CNT/PI (3/1)	50	63.49	0.02892	64.49	0.9571	0.76	63.71	0.9856
	200	121.41	0.02966	127.62	0.9704	0.40	120.80	0.9781
	500	253.97	0.0695	261.02	0.9969	3.5	250.00	0.9712
CNT/PI (1/1)	50	69.05	0.08695	70.43	0.9965	4.9	69.69	0.9987
	200	166.47	0.05487	149.61	0.9699	0.73	166.11	0.9853
	500	384.92	0.02606	394.92	0.9986	0.1	390.16	0.9992

---

Generally, adsorption process on a porous adsorbent may be controlled by three stages: (1) the adsorbates move from the bulk solution to the external surface of the adsorbents, which was called external diffusion; (2) the intraparticle diffusion, diffusion of the adsorbate into the internal sites; (3) adsorption of adsorbates onto the external and inner surface of adsorbents by ion exchange, precipitation, complexation and so on. In this study, the 2,4-DCP adsorption rate is controlled by external or intraparticle diffusion, or both.

The evaluation of the maximum adsorption capacity of the adsorbents, CNTs, CNT/PI (3/1) and CNT/PI (1/1) were conducted under selected conditions. The Langmuir and Freundlich isotherms in formula (2) and (3) were used to fit the experimental information. The equilibrium curves were described in Fig. 11 and the related parameters were summarized in Table 3. Through a comparison between R² values of two isotherms, it is found that Freundlich isotherm model provides better mathematical coincidence to experimental data, which enlightens the adsorbent surface is heterogeneous. The indicators of absorption capacity (the constant K_F and n) followed an order CNT/PI (1/1) > CNT/PI (3/1) > CNT, which was consistent with the above study. In addition, the constant K_F of CNT/PI (1/1) has a value of 8.8718 mg^1−1/n L^1/n g⁻¹ for 298 K, 11.997 mg^1−1/n L^1/n g⁻¹ for 308 K, and 15.963 mg^1−1/n L^1/n g⁻¹ for 318 K, respectively. The constant K_F interprets again that the adsorption of 2,4-DCP onto the adsorbents are favored at high temperatures.

Fig. 11 Langmuir adsorption model of 2,4-DCP for CNT (a), CNT/PI (3/1) (c) and CNT/PI (1/1) (e); Freundlich adsorption model of 2,4-DCP for CNT (b), CNT/PI (3/1) (d) and CNT/PI (1/1) (f).

Table 3 Adsorption isotherm constants for 2,4-DCP adsorption onto CNT, CNT/PI (3/1) and CNT/PI (1/1)

Material	Langmuir				Freundlich
	T (K)	Qm (mg g^−1)	KL × 10^−3 (L mg^−1)	R^2	KF (mg^1−1/n L^1/n g^−1)	n	R^2
CNT	298	450	2.1	0.6606	1.9080	0.5753	0.9588
	308	500	3.3	0.9053	2.0207	0.6152	0.9905
	318	588	4.2	0.9110	2.2267	0.6199	0.9798
CNT/PI (3/1)	298	529	3.2	0.7095	2.7837	0.4426	0.9193
	308	559	4.5	0.8766	2.7992	0.4699	0.9475
	318	621	5.4	0.8794	2.8130	0.5337	0.9249
CNT/PI (1/1)	298	667	3.3	0.7694	8.8718	1.6478	0.9447
	308	719	3.4	0.7422	11.997	1.7369	0.9201
	318	758	3.7	0.7271	15.963	1.8319	0.8924

---

Thermodynamic consideration of an adsorption process can give very valuable insight into the nature of the adsorption process such as its spontaneity, randomness, endothermicity, or exothermicity and so on.³⁰ The values of ΔG of three adsorbents were obtained from formula (6) and (7), and listed in Table 4. The negative values of ΔG confirm the practicability of the process and autonomous adsorption of 2,4-DCP on adsorbents.^41,42 Moreover, the decrease in negative value of ΔG with the increase in temperature indicates that the adsorption process becomes more profitable at higher temperatures, which was consistent with the above study.

Table 4 Thermodynamics parameters for the adsorption process

Material	T (K)	ΔG (kJ mol^−1)	ΔH (kJ mol^−1)	ΔS (J mol^−1 k^−1)
CNT	298	−0.88	13.45	48.0
	308	−1.27
	318	−1.84
CNT/PI (3/1)	298	−3.32	13.23	56.0
	308	−4.30
	318	−4.44
CNT/PI (1/1)	298	−2.41	18.04	68.5
	308	−2.98
	318	−3.78

---

Table 4 also summarizes the results of ΔH and ΔS. The positive value of ΔH (13.45 kJ mol⁻¹, 13.23 kJ mol⁻¹ and 18.04 kJ mol⁻¹ for the CNT, CNT/PI (3/1) and CNT/PI (1/1), respectively) indicates the endothermic nature of the process, confirming that the adsorption is endothermic and the adsorption is favored in high temperature.⁴³ The entropy changes are 48.0 J mol⁻¹ K⁻¹, 56.0 J mol⁻¹ K⁻¹ and 68.5 J mol⁻¹ K⁻¹, for CNT, CNT/PI (3/1) and CNT/PI (1/1), respectively, indicating the fact that the adsorption process is irreversible and favored sorption.

3.4 Comparison with other adsorbents

The adsorption capability of 2,4-DCP by CNT/PI composites and other adsorbents are concluded in Table 5. As shown in Fig. 9, the maximum adsorption capability of CNT/PI (1/1) composites is found to be 565.5 mg g⁻¹ at 318 K. Allowing for the fact that adsorption process is always performed at room temperature, we use adsorption capacity values 506 mg g⁻¹ at 298 K to compare with those obtained by the reported adsorbents (Table 5). It is obvious that the adsorption capacity of CNT/PI (1/1) composites is much higher than other adsorbents, indicating the excellent adsorption performance of CNT/PI composites towards 2,4-DCP. This investigation suggested that π–π interaction, hydrogen bonding interactions and pore-filling may contribute to the high adsorption capacity. The high adsorption capacity in this study revealed that CNT/PI composites is a promising adsorbent for removing 2,4-DCP from aqueous solutions.

Table 5 Comparison of the adsorption capacities for 2,4-DCP onto various adsorbents

Adsorbent	pH	Dosage (g L^−1)	T (°C)	Qm (mg g^−1)	Reference
CNT/PI (1/1)	6	0.4	25	506	This work
Active carbon modified by amide	Original	2	30	285.71	8
Cattail fiber-based activated carbon	Initial	1	20	142.85	9
Mn-Modified activated carbon	—	0.5	25	188.73	10
Coir pith carbon	2	2	35	19.12	44
Maize cob carbon	—	—	30	17.94	45
Oil palm empty fruit bunch carbon	—	—	30	27.25	46

---

4. Conclusion

In summary, the novel CNT/PI composites were prepared by blending CNT and abundant amino-terminated porous PI in ethanol solution and acted as an effective adsorbent for the adsorption of 2,4-DCP. The CNT/PI composites was characterized by TEM, FT-IR, TGA and nitrogen adsorption–desorption techniques, and the results indicated that PI was successfully combined with CNT and formed the CNT/PI composites, consequently improved its adsorption capability of 2,4-DCP removal from aqueous solution. The adsorption capability at 298 K was 330.0 mg g⁻¹, 412.5 mg g⁻¹ and 506 mg g⁻¹ for CNT, CNT/PI (3/1) and CNT/PI (1/1), respectively. Results of batch adsorption experiment showed that the 2,4-DCP removal efficiency was regulated by the solution pH, contact time, temperature, and initial 2,4-DCP concentration but was independent of ion intensity. For both CNT and CNT/PI composites, the adsorption process was well represented using pseudo-second-order kinetics model. Freundlich isotherm model fitted well on the experimental data. Thermodynamic constants including ΔG, ΔH and ΔS revealed the spontaneous, endothermic, and entropy-driven nature of 2,4-DCP adsorption on the adsorbents. Further tests are still required to explore the performance of CNT/PI composites for simultaneous removal of other organic pollutes from aqueous solutions.

Acknowledgements

This work was financially supported by the National Scientific Foundation of China (21477118, 51103139), the Key Program of Natural Science Foundation of Hubei Province (2014CFA530).

References

1. M. Czaplicka, Sci. Total Environ., 2004, 322(1), 21–39.
2. F. W. Shaarani, Desalination, 2010, 255(1), 159–164.
3. Y. B. Lin, B. Fugetsu, N. Terui and S. Tanaka, J. Hazard. Mater., 2005, 120(1), 237–241.
4. S. H. Lin and C. C. Lo, Water Res., 1997, 31(8), 2050–2056.
5. V. K. Gupta, M. R. Ganjali, P. Norouzi, H. Khani, A. Nayak and S. Agarwal, Crit. Rev. Anal. Chem., 2011, 41(4), 282–313.
6. R. N. Goyal, V. K. Gupta and S. Chatterjee, Sens. Actuators, B, 2010, 149(1), 252–258.
7. R. Jain, V. K. Gupta, N. Jadon and K. Radhapyari, Anal. Biochem., 2010, 407(1), 79–88.
8. F. W. Shaarani and B. H. Hameed, Chem. Eng. J., 2011, 169(1), 180–185.
9. L. Ren, J. Zhang, Y. Li and C. Zhang, Chem. Eng. J., 2011, 168(2), 553–561.
10. L. Wang, J. Zhang, R. Zhao, C. Zhang, C. Li and Y. Li, Desalination, 2011, 266(1), 175–181.
11. M. Kragulj, J. Tričković, Á. Kukovecz, B. Jović, J. Molnar, S. Rončević and B. Dalmacija, RSC Adv., 2015, 5(32), 24920–24929.
12. M. A. Salam, M. Mokhtar, S. N. Basahel, S. A. Al-Thabaiti and A. Y. Obaid, J. Alloys Compd., 2010, 500(1), 87–92.
13. C. Lu, Y. L. Chung and K. F. Chang, Water Res., 2005, 39(6), 1183–1189.
14. R. Q. Long and R. T. Yang, J. Am. Chem. Soc., 2001, 123(9), 2058–2059.
15. Y. H. Li, S. Wang, J. Wei, X. Zhang, J. Wei, C. Xu, Z. Luan and D. Wu, Mater. Res. Bull., 2003, 38(3), 469–476.
16. J. Xu, X. Lv, J. Li, Y. Li, L. Shen, H. Zhou and X. Xu, J. Hazard. Mater., 2012, 225, 36–45.
17. H. A. Asmaly, T. A. Saleh, T. Laoui, V. K. Gupta and M. A. Atieh, J. Mol. Liq., 2015, 206, 176–182.
18. Y. Wang, Q. Gao, Q. You, G. Liao, H. Xia and D. Wang, React. Funct. Polym., 2016, 103, 9–16.
19. J. Jensen, Chlorophenols in the Terrestrial Environment, Springer, New York, 1996, pp. 25–51.
20. B. Özkaya, J. Hazard. Mater., 2006, 129(1), 158–163.
21. R. Han, W. Zou, W. Yu, S. Cheng, Y. Wang and J. Shi, J. Hazard. Mater., 2007, 141(1), 156–162.
22. D. Duranoğlu, İ. G. B. Kaya, U. Beker and B. F. Şenkal, Chem. Eng. J., 2012, 181, 103–112.
23. K. Yang and B. Xing, Chem. Rev., 2010, 110(10), 5989–6008.
24. J. Ding, Q. Gao, X. S. Li, W. Huang, Z. G. Shi and Y. Q. Feng, J. Sep. Sci., 2011, 34(18), 2498–2504.
25. J. Peter, B. Kosmala and M. Bleha, Desalination, 2009, 245(1), 516–526.
26. Y. Luo, B. Li, L. Liang and B. Tan, Chem. Commun., 2011, 47(27), 7704–7706.
27. A. Navarro, M. Fernandez, J. J. López, J. Tomkinson and G. Kearley, Phys. Rev. B: Condens. Matter Mater. Phys., 1997, 241, 247–254.
28. Y. Cheng, Q. Feng, X. Ren, M. Yin, Y. Zhou and Z. Xue, Colloids Surf., A, 2015, 485, 125–135.
29. M. Sayın, M. Can, M. İmamoğlu and M. Arslan, React. Funct. Polym., 2015, 88, 31–38.
30. S. Wang, Y. Y. Zhai, Q. Gao, W. J. Luo, H. Xia and C. G. Zhou, J. Chem. Eng. Data, 2013, 59(1), 39–51.
31. Q. R. Fang, Z. B. Zhuang, S. Gu, R. B. Kaspar, J. Zheng and J. H. Wang, Nat. Commun., 2014, 5, 4503.
32. J. P. Wang, Y. Z. Chen, H. M. Feng, S. J. Zhang and H. Q. Yu, J. Colloid Interface Sci., 2007, 313(1), 80–85.
33. F. A. Banat, B. Al-Bashir, S. Al-Asheh and O. Hayajneh, Environ. Pollut., 2000, 107(3), 391–398.
34. B. Pan and B. Xing, Environ. Sci. Technol., 2008, 42(24), 9005–9013.
35. K. A. Halhouli, N. A. Darwish and N. M. Al-Dhoon, Sep. Sci. Technol., 1995, 30(17), 3313–3324.
36. M. Ghaedi, A. Hassanzadeh and S. N. Kokhdan, J. Chem. Eng. Data, 2011, 56(5), 2511–2520.
37. R. Han, J. Zhang, W. Zou, J. Shi and H. Liu, J. Hazard. Mater., 2005, 125(1), 266–271.
38. S. Hydari, H. Sharififard, M. Nabavinia and M. rezaParvizi, Chem. Eng. J., 2012, 193, 276–282.
39. K. D. Samar, J. Eng., 2012, 18, 63–77.
40. S. G. Wang, X. W. Liu, W. X. Gong, W. Nie, B. Y. Gao and Q. Y. Yue, J. Chem. Technol. Biotechnol., 2007, 82(8), 698–704.
41. H. B. Senturk, D. Ozdes and C. Duran, Desalination, 2010, 252(1), 81–87.
42. H. Bian, M. Li, Q. Yu, Z. Chen, J. Tian and H. Liang, Int. J. Biol. Macromol., 2006, 39(4), 291–297.
43. X. F. Sun, S. G. Wang, W. Cheng, M. H. Fan, B. H. Tian, B. Y. Gao and X. M. Li, J. Hazard. Mater., 2011, 189(1), 27–33.
44. C. Namasivayam and D. Kavitha, Sep. Sci. Technol., 2005, 39(6), 1407–1425.
45. M. Sathishkumar, A. R. Binupriya, D. Kavitha, R. Jayabalan, J. G. Choi and S. E. Yun, Chem. Eng. J., 2009, 147(2), 265–271.
46. A. Zahangir, S. A. Muyibi and J. Toramae, J. Environ. Sci., 2007, 19(6), 674–677.

---