Batch adsorption and desorption of crystal violet (CV) and basic red 9 (BR9) on multi-walled carbon nanotubes (MWCNTs) were conducted. To investigate the possible mechanisms of adsorption/desorption hysteresis, oxidized MWCNTs (O-MWCNTs) with more oxygen-containing groups were obtained by oxidizing as-purchased MWCNTs (A-MWCNTs) using nitric acid. The adsorption kinetics could be described by the pseudo-second-order model, suggesting that chemical reactions are the rate-limiting steps. The adsorption isotherms were fitted well by the Langmuir model, which suggests that, in addition to π–π interactions, chemical reactions significantly affect the adsorption. The adsorption capacity decreased in the order of CV on A-MWCNTs, BR9 on A-MWCNTs, and BR9 on O-MWCNTs, possibly because the amidation between BR9 and the surface groups of MWCNTs results in steric hindrance, which limits the adsorption of BR9 to inner grooves between CNT bundles. Adsorption/desorption hysteresis was observed for BR9 but not for CV. It was found that the π–π interaction and molecular entrapment were not responsible for the adsorption/desorption hysteresis. The hysteresis might be caused by the irreversible amide bonds between BR9 and MWCNTs. The results indicate that the steric hindrance due to the three-dimensional structure of organic compounds plays an important role in both adsorption/desorption kinetics and equilibria.
Adsorption of organic compounds on carbon nanotubes (CNTs) has been investigated intensively to understand the fate and transport of CNTs (Wang et al. 2010a; Yang et al. 2010; Apul & Karanfil 2015). The mechanisms of adsorption of organic compounds by CNTs can be divided into nonspecific interactions and conditional attractive forces. Nonspecific interactions, which are generally referred to as van der Waals interactions, are the major contributors of overall attractive forces (Apul & Karanfil 2015). π–π interactions between aromatic compounds and the surface of CNTs have been considered as the major mechanism. Other attractions include hydrogen bonding between functional groups of organic compounds and CNT surfaces and electrostatic interactions between ionic compounds and surface charges of CNTs.
Compared to the literature on adsorption, the literature on the desorption behaviors of organic compounds from CNTs is scarce, and the topic still requires further investigation (Yang & Xing 2007; Oleszczuk et al. 2009; Yang et al. 2010; Ma et al. 2011; Wu et al. 2013). The strong π–π interaction between tetra-tert-butylphthalocyanines and CNT surface has been proposed as a possible mechanism for desorption hysteresis (Wang et al. 2002). After adsorption of organic compounds, rearrangement of CNT bundles may lead to desorption hysteresis, which has been observed for the adsorption of organic compounds with polar functional groups on CNTs, such as bisphenol A, 17α-ethinyl estradiol, oxytetracycline, carbamazepine, norfloxacin, (Pan et al. 2008; Oleszczuk et al. 2009; Wang et al. 2010b), and polycyclic aromatic hydrocarbons on fullerene (Yang & Xing 2007). Wu et al. (2013) has concluded that the desorption hysteresis of aniline and 4-methylaniline on oxidized-MWCNTs is due to the irreversible amide bonding (i.e. −CONH) between the amino groups of anilines and the functional groups on MWCNTs. Desorption of organic compounds from CNTs is as important as adsorption for evaluating environmental and health risks of both CNTs and contaminants.
In general, the underlying mechanisms of desorption and desorption hysteresis from CNTs are mostly discussed by adsorption and desorption isotherms only. The adsorption and desorption kinetics should also be considered for the interpretation of desorption hysteresis. The objectives of this work were to investigate the adsorption/desorption of organic compounds from MWCNTs and to examine the mechanisms for adsorption/desorption hysteresis. Crystal violet (CV) and basic red 9 (BR9) were used to understand the influences of the structures of molecules and the functional groups on the adsorption and desorption of organic compounds on MWCNTs. Both the kinetics and the isotherms of adsorption and desorption were determined.
MATERIALS AND METHODS
Oxidation and characterization of MWCNTs
The MWCNTs (US Research Nanomaterials, Inc.) have an outer diameter of 50–80 nm and an inner diameter of 5–15 nm, according to the manufacturer. To increase the amount of surface oxygen functional groups, the as-purchased MWCNTs (A-MWCNTs) were oxidized. A-MWCNTs were dispersed in 500 mL of 3 M nitric acid and sonicated for 30 min. Then, the MWCNTs dispersion was refluxed at 120 °C for 3 h with stirring. After being cooled to room temperature, the MWCNTs were separated from nitric acid by filtration and the oxidized MWCNTs were washed by deionized water repeatedly until the pH value of the filtrate was approximately 6 (Chin et al. 2010). The as-purchased and oxidized MWCNTs are referred to as A- and O-MWCNTs, respectively, in the following text.
Nitrogen adsorption and desorption on the MWCNTs were conducted at 77 K in an accelerated surface area and porosimetry system (ASAP, ASAP2010, Micrometrics Inc.). The surface area and the pore size distribution were then obtained by analyzing the nitrogen adsorption–desorption isotherms using Brunauer-Emmett-Teller (BET) and Barrett-Joyner-Halenda (BJH) models. Fourier transform infrared spectroscopy (FTIR, JASCO, FT/IR-4600) was used to characterize the surface groups on the MWCNTs.
Batch adsorption and desorption
The adsorption/desorption kinetics and isotherms were studied by batch experiments in screw cap centrifuge tubes with Teflon liners. CV (Sigma-Aldrich) and BR9 (Sigma-Aldrich) were dissolved in a background solution containing 0.01 M CaCl2 (Merck) and 200 mg/L NaN3 (as a biocide, Merck) in deionized water. The pH values of the suspensions were adjusted to 8.0 ± 0.2 by 0.01 M HCl and 0.01 M NaOH. MWCNTs (10 mg) were dispersed in 6.5 ml CV or BR9 solution. For adsorption kinetic study, the initial concentration of CV or BR9 was 100 mg/L. The samples were taken at 5, 10, 20, 30 40, 60, 120, and 240 min. For adsorption/desorption isotherms, the initial concentrations were 20, 50, 100, 150, 200, 150, 300, 350 and 400 mg/L. All experiments were conducted in a reciprocal shaker with a rotation speed of 120 rpm at room temperature (25 ± 1 °C). The experiments were shielded from light to avoid photolysis of CV and BR9, and blank experiments without addition of MWCNTs were also conducted to confirm that the concentration decrease was solely due to adsorption on MWCNTs.
Desorption experiments were conducted immediately after adsorption reached equilibrium. For the desorption kinetic study, the supernatant was removed and displaced by the same volume of background solution at the same time interval as the adsorption kinetic study. The tubes were resealed and shaken. The MWCNTs suspension was centrifuged at 6,000 rpm for 10 min before withdrawing the supernatant. The concentrations of CV and BR9 in the supernatant were determined by UV-Vis spectrophotometer (Jasco, V-650). The absorbance bands of CV and BR9 are 590 and 540 nm, respectively. All adsorption and desorption experiments were at least duplicated.
Here, qe and qt are the amounts of adsorbate on adsorbent (mg/g) at equilibrium and at time t (min), respectively, and k1 is the pseudo-first-order rate constant (1/min). k2 is the pseudo-second-order rate constant (g/mg·min). α (mg/g·min) and β (g/mg) are constants for the Elovich model. ki is the rate constant (mg/g·min0.5) for the intraparticle diffusion model.
Here, KL is the parameter of the Langmuir model and qmax is the maximum amount of adsorption. KF and n are the parameters for the Freundlich model. K1 and K2 are the coefficients of the Temkin model.
RESULTS AND DISCUSSION
Characterization of MWCNTs
Table 1 gives the micropore area, external surface area, BET surface area and micropore volume obtained by analyzing the adsorption/desorption of nitrogen on the MWCNTs at 77 K. The BET surface area of A-MWCNTs increases from 87.98 m2/g to 97.69 m2/g after HNO3 treatment. The external surface area, which is mainly attributed to groove and interstitial sites of CNT bundles (Agnihotri et al. 2005), increases from 73.89 m2/g to 82.69 m2/g (11.9%). The increase in external surface area may be caused by the damage of surface structure of A-MWCNTs after HNO3 treatment. The pore size distributions of A- and O-MWCNTs are given in Figure S1 (available with the online version of this paper).
Figure 1 shows the FTIR spectra of A-MWCNTs and O-MWCNTs. Bands can be observed at approximately 1,500 cm−1 and 2,400 cm−1 on both A-MWCNTs and O-MWCNTs. The bands at 1,550–1,650 cm−1 represent the –C = O from the acidic group, and those at 2,300–2,400 re-present the –OH stretch from strongly H-bonded –COOH (Salam & Burk 2008; Kuo 2009). Additionally, these two bands of O-MWCNTs are distinctly larger than those of A-MWCNTs, which suggests that HNO3 treatment increases the oxygen-containing functional groups on the surface of A-MWCNTs. Previous work has also shown that the amount of oxygen-containing groups on CNTs can be increased by the same oxidation process (Chin et al. 2010; Chi et al. 2016).
The adsorption kinetics of CV on A-MWCNTs and BR9 on A-MWCNTs and O-MWCNTs are shown in Figure 2. The amounts of adsorption increase significantly in the first 10 min and then slow down. There were no significant changes in the adsorption amounts of dyes after 50 min for CV on A-MWCNTs and for BR9 on O-MWCNTs. The adsorption of BR9 on A-MWCNTs increased slowly until approximately 120 min. Therefore, all of the adsorption isotherm experiments were conducted for 120 min.
The adsorption of CV by A-MWCNTs or BR9 by A- and O-MWCNTs reached equilibrium much faster than reported in the literature (Oleszczuk et al. 2009), and the two-stage adsorption (i.e. fast initial adsorption on outer surfaces and slow diffusion to interlayers and micropores) was not observed. The pores in MWCNTs used in this study are mainly composed of mesopores, as shown in Figure S1(a), so that the slow diffusion of dye into interlayers and micropores was apparently negligible. In addition, CV and BR9 molecules are approximately 1 nm (estimated by bond lengths), which are larger than the main size of micropores of A-MWCNTs and O-MWCNTs (0.69 nm and 0.68 nm, respectively, as shown in Figure S1(b)). This means that most of the micropores are not available for CV or BR9. Therefore, the adsorption process for dyes on both A-MWCNTs and O-MWCNTs can be considered as fast adsorption.
The kinetic data were also fitted by the pseudo-first-order, pseudo-second-order, Elovich, and intraparticle diffusion models. The fitted kinetic parameters of all four models are shown in Table 2. The R2 values for the pseudo-first-order, Elovich, and intraparticle diffusion models are less than 0.95, which means that the adsorptions of CV and BR9 onto MWCNTs do not follow these three models. The regression coefficients of CV on A-MWCNTs, BR9 on A-MWCNTs, and BR9 on O-MWCNTs fitted by the pseudo-second-order model are 1.000, 0.998, and 0.999, respectively. The comparison of experimental data and fitting results of four models are shown in Figure S2 (available with the online version of this paper). The results agree with the literature (Sabna et al. 2016) and suggest that the interaction between CV, as well as BR9, and the surface of MWCNTs is the rate-limiting step (Ho & McKay 1999).
Figure 3 shows that the adsorption capacity of CV on A-MWCNTs is the highest. The adsorption capacity of BR9 on A-MWCNTs is higher than that on O-MWCNTs. The Langmuir, Freundlich, and Temkin models have been used to fit the experimental data. Table 3 shows the fitted adsorption isotherm parameters and regression coefficients. Although all three models show good correlations to the experimental results, the Langmuir model has the highest R2 values (i.e. 0.999 for CV on A-MWCNTs, 0.994 for BR9 on A-MWCNTs, and 0.981 for BR9 on O-MWCNTs). Similar fitting results were found in the adsorption of CV on nanocables (Liu et al. 2015) and on magnetic-modified MWCNTs (Madrakian et al. 2011).
According to Table 3, the adsorption capacities of CV on A-MWCNTs, BR9 on A-MWCNTs, and BR9 on O-MWCNTs are 57.80 mg/g, 55.55 mg/g, and 48.78 mg/g, respectively. The hydrophobic force, which can be described by log Kow, is another potential interaction affecting adsorption. Organic compounds with higher log Kow are more hydrophobic and may have higher tendencies to be adsorbed by MWCNTs. The log Kow values of CV and BR9 are 0.51 and −0.21, respectively (Tsai et al. 1991). Thus, the adsorption amount of CV on A-MWCNTs is greater than those of BR9 on both A- and O-MWCNTs.
The π–π interaction has been considered as the major mechanism for the adsorption of aromatic compounds on CNTs (Lin & Xing 2008). In this study, CV and BR9 both contain three aromatic rings; however, the strength of the π–π interaction between them and MWCNTs may not be as strong, due to their 3D structures. Aromatic compounds with planar structure, such as polycyclic aromatic hydrocarbons, can match the array of hexagonal carbon rings on the surface of MWCNTs well (Gotovac et al. 2007). However, with their three-dimensional structures, the three benzene rings of CV and BR9 cannot lie flat on the surfaces of MWCNTs. Hence, the π–π interaction between adsorbate and adsorbent may be relatively weak in this study as compared to the literature (Gotovac et al. 2007; Lin & Xing 2008).
The terminal functional groups of CV and BR9 molecules are –N-(CH3)2 and −NH2, respectively. Both –N(CH3)2 and −NH2 are electron-donating groups which can enhance the density of π-electrons and increase the strength of π–π interactions. The adsorption capacity of CV is higher than that of BR9, which suggests that the electron-donor-acceptor reaction does not significantly influence the adsorption of CV and BR9 by MWCNTs. Furthermore, the pH values were adjusted to 8 to avoid protonation of the −NH2 on BR9 so that the electrostatic interaction might not come into play.
Wu et al. (2013) observed an irreversible amidation reaction between the amino groups of anilines and the carboxyl/lactonic groups (−COOH/−COOR) on the oxidized MWCNTs surfaces. BR9 contains two amino groups and one imino group, which could be considered as three aniline molecules connected to a carbon. From the FTIR spectra (Figure 1), there are carboxyl groups on the surfaces of both A-MWCNTs and O-MWCNTs. Therefore, the amidation may occur between the amino groups of BR9 and the carboxyl groups on MWCNTs. The amine groups of CV are tertiary: that is, there is no hydrogen bound to the nitrogen. There can thus be no amidation between CV and MWCNTs. Because only one of the three benzene rings can lie on the surface of the MWCNTs, and amidation prevents further interaction between BR9 and the MWCNTs, the adsorption is thus monolayer. Consequently, the Langmuir model describes the adsorption isotherm well.
The mesopores of A-MWCNTs and O-MWCNTs are mostly of diameters of 2–3 nm (Figure S1(a)), and the size of BR9 is approximately 1 nm (estimated by the bond lengths). Amidation between BR9 and the carboxyl groups on MWCNTs might result in longer bond lengths than π–π interaction and lead to adjustment of BR9 for better adsorption position (Jin et al. 2015; Yu et al. 2017), which may result in loose packing of BR9 on MWCNTs and limit the transport of BR9 into the inner adsorption sites. Amidation cannot occur between CV and MWCNTs, because CV does not contain amino groups, and the adsorption of CV by MWCNTs may be only due to the π–π interaction. Without steric hindrance, CV might diffuse into the inner space. As a result, the adsorption capacity of CV is higher than that of BR9. Compared to A-MWCNTs, O-MWCNTs contain more carboxyl groups, which may provide more active sites for BR9, and the adsorption capacity may be higher. However, because O-MWCNTs contain more carboxyl groups, those BR9 molecules quickly adsorbed on the pore openings of the O-MWCNTs in the beginning stage may limit diffusion to deeper adsorption sites. Therefore, the adsorption of BR9 on O-MWCNTs reaches equilibrium faster than that on A-MWCNTs (Figure 2(b)), while the adsorption capacity of BR9 on O-MWCNTs is lower than that on A-MWCNTs.
Desorption of CV and BR9 from MWCNTs
The desorption kinetics of CV from A-MWCNT and BR9 from both A- and O-MWCNTs are illustrated in Figure S3 (available online). It requires approximately 100 min to reach desorption equilibrium for all three systems. Additionally, it is observed that the desorption of BR9 from O-MWCNTs reaches equilibrium slower than that from A-MWCNTs.
Desorption isotherms of CV on A-MWCNTs as well as BR9 on A-MWCNTs and O-MWCNTs are provided in Figure 4. No desorption hysteresis was observed for CV on A-MWCNTs, since the desorption and adsorption isotherms overlapped, while significant desorption hysteresis was observed for BR9 from both A-MWCNTs and O-MWCNTs.
Desorption hysteresis may be divided into true hysteresis and artificial hysteresis. Artificial hysteresis is caused by some external factors such as loss of adsorbent or adsorbate, degradation of adsorbate, and insufficient time for adsorption or desorption equilibrium. In this study, the suspension was adequately centrifuged after shaking, before analyzing the concentrations of CV and BR9, in order to avoid the loss of the MWCNTs. The losses of CV and BR9 were less than 3%, as confirmed by the blank experiments. All of the batch experiments were shielded from light to avoid the potential photodegradation of CV and BR9, and the biodegradation was controlled by adding 200 mg/L NaN3 into the background solution. Additionally, the kinetic experiments showed that the adsorption or desorption reached equilibriums within 120 min. Therefore, the de-sorption hysteresis observed in this study should be considered as true hysteresis.
Organic compounds with higher affinity to CNTs would be more likely to rearrange the bundles of CNTs and cause more significant desorption hysteresis (Ma et al. 2011). CV has higher affinity to the MWCNTs than does BR9; however, no adsorption/desorption hysteresis is observed. In addition, ultrasonication can significantly loosen the bundles of CNTs (Saleh et al. 2008). If adsorbed molecules were entrapped between bundled MWCNTs, ultrasonic treatment should increase the desorption. There was no significant change between desorption before and after ultrasonic treatment in all three conditions (Figure 5). Hence, the molecular entrapment made no contribution to the desorption hysteresis of BR9 on MWCNTs.
Strong π–π interaction has been proposed to represent the mechanism of immobilization of tetra-tert-butylphthalocyanines on CNT surfaces and to lead to desorption hysteresis (Wang et al. 2002). Both CV and BR9 might not be able to lie flat on the MWCNTs surface, and the π–π interaction would not be as strong as reported in the literature (Gotovac et al. 2007; Lin & Xing 2008) due to their 3-D structures. However, desorption hysteresis was only observed for BR9. Therefore, π–π interaction might not be the reason for the desorption hysteresis observed here.
It has been discussed that the adsorptions of BR9 on A-MWCNTs and O-MWCNTs were mainly due to the irreversible amidation between the amino groups of BR9 and the carboxyl groups on MWCNTs. Such amidation could not occur between CV and MWCNTs because CV has no amino groups. During desorption, some BR9 molecules are fixed by irreversible amidation, and the BR9 molecules adsorbed via π–π interaction might desorb and diffuse to the solution. Without irreversible amidation, all CV molecules easily desorb from MWCNTs: thus, the adsorption/desorption hysteresis was not observed.
The desorption HI value has been applied to compare the degree of desorption hysteresis (Huang et al. 1998). HI is defined as the ratio of the excessive adsorbed amount after desorption to the adsorbed amount during adsorption with respect to the specified equilibrium concentration. Higher HI values indicate more desorption hysteresis, and no desorption hysteresis occurs when HI = 0. The HI values of three conditions in this study at different equilibrium concentrations were calculated and listed in Table 4. The HI values for CV on MWCNTs are less than 0.05, which means that the difference between desorption and adsorption equilibria is within 5%. Thus, no adsorption/desorption hysteresis is considered to exist for adsorption of CV on A-MWCNTs. The HI values of BR9 on O-MWCNTs are higher than those of BR9 on A-MWCNTs. This is because there are more carboxyl groups on O-MWCNTs than on A-MWCNTs, as indicated by FTIR analysis, which results in more irreversible amidation for BR9 on O-MWCNTs than on A-MWCNTs. Similar results have been observed for the adsorption/desorption hysteresis of aniline and 4-methylaniline on MWCNTs (Wu et al. 2013).
The HI value decreases with increasing equilibrium concentration in all three systems. For example, adsorption/desorption hysteresis of BR9 on O-MWCNTs is the greatest, and the HI values are 1.36, 0.75, and 0.22 when the equilibrium concentrations are 10 mg/L, 20 mg/L, and 50 mg/L, respectively. As mentioned previously, amidation might limit the access of BR9 to the inner grooves. At high concentration, the inner grooves might be quickly blocked, and most of the adsorption occurs on the exterior surface. Those BR9 molecules adsorbed via π–π interaction on the exterior surface can thus desorb easily. As the concentration decreases, the blockage of the inner surface due to amidation may also be limited in the beginning stage of adsorption. Thus, more BR9 molecules may transport into the pores and may be difficult to desorb if those pores are blocked in the later stage of the adsorption.
The adsorption and desorption of CV and BR9 on MWCNTs were conducted to investigate adsorption and desorption mechanisms. The adsorption kinetics could be described by the pseudo-second-order model, suggesting that the interaction between CV, as well as BR9, and the surface of MWCNTs is the rate-limiting step. The adsorption and desorption of CV and BR9 follow the Langmuir model; however, the mechanisms differ. The π–π interaction dominates the adsorption and desorption of CV on A-MWCNTs, and the additional irreversible bonding and steric hindrance significantly affect the transports and equilibria of adsorption and desorption of BR9 on both A- and O-MWCNTs. The adsorption/desorption kinetics also support the conclusions of adsorption/desorption isotherms and hysteresis. Amidation leads to slower adsorption equilibrium, lower adsorption capacity, and more significant adsorption/desorption hysteresis. These results show that the structural and chemical properties of organic compounds play significant roles in their interactions with CNTs. These findings may also provide additional information with respect to the fate and transport of both carbonaceous nanomaterials and organic pollutants in aquatic environments. Solution pH and ionic strength may also affect the adsorption due to dissociation of adsorbates and functional groups on MWCNTs, which may affect the hydration bonding, steric hindrance, and electrostatic interactions. The effects of solution pH and ionic strength may be considered in the future to better explore the adsorption/desorption phenomena.