Adsorption of Nonylphenol to Multi-Walled Carbon Nanotubes : Kinetics and Isotherm Study

We explored the occurrence and distribution of nonylphenol (NP) in 13 Taiwanese source waters. From all the surveyed waters, NP was detected at a high concentration, which could be attributed to contamination by wastewater discharges. In this study, we applied modified multi-walled carbon nanotubes (MWCNTs) for removing NP from aqueous solution. The impact of a few experimental factors, i.e., pH, contact time, MWCNTs dose, and temperature on the NP removal efficiency of modified MWCNTs was studied. The maximum adsorption capacity of the MWCNTs was observed to be 1040 mg NP/g when the initial NP concentration was 2.5 mg/L, and the solution pH was 4. The adsorption process followed the Elovich kinetics and the Elovich isotherm, indicating it is multilayer adsorption. The thermodynamic analysis demonstrated the NP adsorption by MWCNTs was thermodynamically satisfactory and, for the most part, endothermic as in the case of phenol adsorption. The result of the current study demonstrated the significance of free binding sites and the pore size of MWCNTs in the NP adsorption. This paper will help to better comprehend the adsorption behavior and mechanism of alkyl phenolic compounds onto MWCNTs.


Introduction
Nonylphenol (NP), a hydrophobic contaminant, has frequently been found in surface water, groundwater, sediment, and soil [1,2].It is a significant byproduct of a non-ionic surfactant, nonylphenol polyethoxylate, which is used as an ingredient of pesticides and personal care products, lubricating additive, the catalyst in curing agent of epoxy resin, defoamer in industrial laundries and dispersant in paper industries [2][3][4].Nonyl phenol has a similar structure to natural estrogen, so it can mimic characteristic hormones by making a connection with estrogen receptors in the environment, which is highly toxic to fish, microorganisms, and aquatic plants.Additionally, the presence of NP in the human body will reduce sperm count and immunity and can cause breast and testicular cancers.Thus, NP has been listed as a priority substance in the EU Water-Framework Directive [5].As a result, NP has been replaced by alcohol polyethoxylates in the most European, Canadian, and Japanese industries [1,4].However, alcohol polyethoxylates are expensive and less efficient in general, which boosts the illegal production and usage of NP.Moreover, inadequate NP removal by wastewater treatment plants (WWTPs) results in a high NP concentration in the aquatic environment [6].
Adsorption onto sludge particulates [5], biological treatment [31], chemical treatment [32], photocatalysis [33], electrochemical degradation [34], volatilization [5], and ozonation [35] have been applied for the removal of NP from wastewater.Among the technologies, adsorption has been considered as an important option because of the unique physicochemical properties of NP such as low water solubility and a high log K OC value [3,36].
Recently, carbon nanotubes (CNTs) have received much attention as an adsorbent because of its large surface area, highly porous and hollow structure, light mass density, and strong interaction with hydrophobic organic compounds such as NP [37,38].On the other hand, spent CNTs can be readily regenerated thermally, chemically, or biologically, so it can be readily reused.
In the present investigation, NP in water samples, which were collected from source waters for 13 Taiwanese water treatment plants (WTPs), was adsorbed to modified multi-walled CNTs (MWCNTs); the kinetics and isotherm equilibrium of the adsorption process were studied.In particular, the influence of different experimental parameters, i.e., water pH, MWCNT dose, contact time, and temperature on the adsorption process.

Materials
Nonyl phenol standard was purchased from Sigma Aldrich (Saint Louis, MS, USA).Multi-walled carbon nanotubes of reagent grade were acquired from Kuang-Yuan Biochemistry Technology (New Taipei City, Taiwan).All other chemicals used in this study are of reagent grade and were also purchased from Sigma Aldrich (Saint Louis, MO, USA).N 2 of 99.99% purity was bought from Shinn Hwa Gas (Taipei, Taiwan).Table 1 summarizes the general properties of NP and MWCNT used in this study.The surface area and pore volume of the MWCNTs were 122 m 2 /g and 0.44 cm 3 /g, respectively.The purchased MWCNTs were pretreated by HNO 3 solution (65%) at 120 • C for 40 min to remove metal impurities from the surface of the MWCNTs.The treated MWCNT samples were, then, washed with deionized (DI) water to remove excess HNO 3 , which was carried out by refluxing the solution with MWCNTs in an ultrasonic cleaning bath at 80 • C for 2 h.Ultrasonication treatment also helps to remove amorphous carbons from the MWCNTs.Finally, the treated materials were filtered through a glass fiber filter (GC-50, Advantec, Taipei, Taiwan) to harvest modified MWCNTs.

Water Sample Preparation
Raw water samples were collected at the source water for each of 13 WTPs in Taiwan (2009)(2010)(2011)(2012)(2013)(2014)(2015).Produced water samples were also collected from the 13 plants.As soon as they were collected, the water samples were filtered with a cellulose acetate membrane filter (0.22-0.45 µm, Merck, Taipei, Taiwan).After filtration, samples were acidified to pH 4.0 using 2 N sulfuric acid and kept at 4 • C.

Instrumentation and Techniques
The surface morphology of the modified MWCNTs was obtained by a scanning electron microscope (SEM, FEI Quanta 200 Environmental Scanning Electron Microscope, Hillsboro, OR, USA) and a transmission electron microscope (TEM, JEM-1400 Transmission Electron Microscope, Peabody, MA, USA).The zeta potential of the MWCNTs during modification at different pHs (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11) was measured by a zeta analyzer (Brookhaven BI-90 Plus Particle Size Analyzer, Champaign, IL, USA).Using the BET surface area analyzer (Beckman Coulter SA3100 Surface Area Analyzer, Indianapolis, IN, USA), the surface area and pore size of the modified MWCNTs were analyzed.Nonyl phenol in water samples was extracted by solid phase extraction (SPE) and subsequently quantified by high-performance liquid chromatography-mass spectrometry (HPLC-MS) [40].The SPE was carried out according to the method proposed by Lin and Tsai [41].Briefly, the Oasis HLB SPE cartridge (500 mg, 6 mL, Waters, Milford, MA, USA) which was first pretreated with methanol and deionized water of 6 mL each.Then, a water sample of 400 mL with 13 C 6 -sulfamethazine (use as a surrogate) added were loaded to the cartridge at 3-6 mL/min.After a sample loaded, the cartridge was rinsed with 6 mL DI water and dried with N 2 gas.Then, analytes were eluted with 4 mL methanol and 4 mL methanol-diethyl ether (1:1 ratio, 50:50, v/v).The eluates were collected to pass for drying with N 2 gas and reconstituted in 0.4 mL of 25% aqueous methanol.Finally, the obtained liquid was filtered through a 0.45 µm PVDF membrane before the HPLC-MS/MS analysis.

Batch Adsorption Experiments
In the present investigation, all experiments were carried in a batch mode.In the adsorption isotherm experiments, 10 mL of 2.5 mg/L NP was introduced into a centrifuge tube containing different concentrations (0.5-10 mg/L) of adsorbate.The tube was, then, tightly closed with a Teflon-lined screw cap and shaken at 150 rpm at 25 ± 1 • C over 24 h to reach equilibrium.To study the pH dependency of the adsorption, pH (2-10) of the mixture was adjusted by adding 0.1 M HCl/NaOH solution.Once equilibrium was reached, the adsorbent (i.e., MWCNTs) was separated from the solution using a 0.2-µm membrane filter, and the residual NP concentration was analyzed.For the case of kinetic study, samples were collected after 0.5, 1, 2, 4, 7, 8, 10, 12 and 14 h of reaction, and residual NP concentration of each sample was analyzed.All the experiments were performed three times to determine averages and standard deviations; in fact, the calculated c.v. for each set of experiments was within ±3%.

NP Concentrations of the Source and Treated Water Samples
As shown in Figure 1, water samples were collected at various locations covering the whole area of Taiwan, including Ban-shin (A), Bao-shan (B), Bei-dou (C), Carp Lake (D), Chan-shin (E), Chen-chin Lake (F), Dong-shin (G), Feng-shan (H), Feng-yuan (I), Kao-tan (J), and Kin-men (K), Nan-hua (L) and Shin-shan (M).All of them are important source waters of Taiwan, so each of them is withdrawn to a nearby WTP.Raw water and produced water samples were collected 2 to 19 times in 2009-2015.The NP concentration of each source water is provided in Supporting Information (Table S2).Different unit processes, namely, rapid filtration, pre-chlorination, coagulation/sedimentation, pre-ozonation, post-chlorination, granular activated carbon, ultrafiltration, and low-pressure reverse osmosis, are employed in combination at each of the WTPs (Table S2).
The result shows that NP concentration was in the range from below detection limit to several hundred ng/L for most of the treatment plants (Figure 1).The maximum NP concentration of 986 ng/L was found at the source water of the Site-M, which consists of conventional processes such as pre-chlorination, coagulation/sedimentation, rapid filtration, and post-chlorination.While the range of NP concentrations of source water was found from 46.1 to 986 ng/L, that of produced water (after conventional treatments) was from below detection limit to 117 ng/L, respectively (as shown in Table S2).It was observed that the NP concentration was not varying significantly at the two sites, i.e., Sites-F and -E.On the other hand, the water samples collected from the Sites-I and -L did not show NP; the concentration was below the detection limit.The high NP contents of the raw water samples collected from Sites-M, -B, -E and -H were attributed to discharge from municipal WWTPs and industrial parks [42].

Characterization of Adsorbent
The physical properties of MWCNTs, such as specific surface area, pore size distribution and pore volume, were modified via acidification and summarized in Table 2.The acidification increased the surface area, average pore diameter and pore volume of the MWCNTs.Probably, amorphous carbons were removed from the structure by the acidification to open the pores of MWCNTs.While the S BET of the modified MWCNTs was comparable with those reported by others [43][44][45], the pore volume was much higher; 0.81 cm 3 /g was obtained (Table 2).HNO 3 treatment reduced metals on the surface of MWCNTs and created new acidic sites, which dominate the roughness of the MWCNTs [46]. Figure 2a gives the SEM (left) and TEM (right) images of raw MWCNTs, respectively.It can be observed that raw MWCNTs shows typical tubular structure morphology, with a length of 5-15 nm. Figure 2b shows the morphology of the modified MWCNTs.The result shows the tubular structure morphology of the modified MWCNTs that has no particle on its surface.The TEM image presents the tubular structure of the raw MWCNTs (right of Figure 2a) and the modified MWCNTs samples (right of Figure 2b).These TEM images confirm the successful removal of metals from the MWCNT surface, which is consistent with the SEM observation.Figure 3 shows the zeta potential of the modified MWCNTs as a function of pH which gives the point of zero-charge (ZPC); the ZPC was observed at pH 2.3 (=pH zpc ).The zeta potential of the modified MWCNTs was negative in the neutral pH range (Figure 3), indicating that the surface of the MWCNTs was negatively charged.In comparison, the zeta potential of the modified MWCNTs was positive at acidic pHs.

Effects of pH and MWCNTs Concentration on NP Removal
Figure 4a clearly shows the influence of pH on the NP removal.At alkaline pHs, NP adsorption decreased.The obtained results are in line with previously obtained ones that deprotonation of organic molecules (phenolic) took place and promoted diffusions [49].Considering pH zpc of the modified MWCNTs (i.e., 2.3) and pK a of NP (i.e., 10.7; [39]), at neutral pHs (e.g., 4-10), the adsorption of NP onto the MWCNTs is favored due to the weak electrostatic interaction and H-bond formation between the negatively charged MWCNTs and NP. Figure 4b demonstrates the effect of the modified MWCNTs dosage (0.1 to 10 mg/L) on NP adsorption at pH 4. As the MWCNTs concentration increased, the NP adsorption increased to some extent possibly due to increased free binding sites.As shown in Figure 4b, the highest adsorption capacity of the modified MWCNTs for NP was estimated at 1040 mg/g at pH 4 and 25 • C. In any case, the adsorbent mass higher than 0.5 mg/L did not further improve the adsorption.Therefore, the modified MWCNTs of 0.5 mg/L was used in the following batch experiments.

Adsorption Kinetics
Figure 5 shows the kinetics of NP adsorption onto the modified MWCNTs.The obtained result shows that the equilibrium is established in 14 h (Figure 4a).In the beginning, the rate of adsorption was high probably due to a large number of free surface sites available on the MWCNTs.Afterward, the adsorption rate became slower and finally reached the equilibrium state.Thus, it could be concluded that free binding sites and affinity between the modified MWCNT and NP might be responsible for the NP adsorption, similar to the result from our previous study in which NP was adsorbed to clays and soil materials [50].In fact, the kinetics profile from the current NP adsorption study is characterized with two linear regions; they are the liquid film diffusion layer and the intra-particle diffusion one [51].The rate of NP adsorption in the liquid film diffusion layer is expressed as Equation (1): where S 0 is the particle surface area per unit particle volume, and k f is the rate constant.The intra-particle diffusion model assumes that the adsorption density, q t , varies proportionally with t 1 2 , as shown in Equation ( 2), where I and k id is the intercept and the intra-particle diffusion rate constant, respectively.The boundary layer effect is indicated by the initial high slope, and the intra-particle diffusion or pore diffusion is demonstrated by the less steep slope (Figure 5).The NP adsorption capacity increases with the increase in the adsorbent surface area (S 0 ) [45].Moreover, the increase in the surface area of the adsorbent strongly correlates with the equilibrium adsorption capacity [49].This kind of relationship has already been described in the literature [52].It suggests that the pore-filling mechanism is mostly responsible for the adsorption.Generally, small-size organic compounds are responsible for pore-filling [43,53].In addition, the octanol-water partitioning coefficient (log K ow ), an indicator of hydrophobicity of an organic compound, plays an important role in the adsorption [1].Therefore, the compound with a high log K ow (i.e., log K ow > 4) should exhibit a strong adsorption affinity towards the active binding sites of the adsorbent [1].Since the log K ow of NP is 4.8-5.3(Table 1), NP seems to have a strong adsorption affinity.
Table 3 lists the kinetics models that were applied for the NP adsorption to the modified MWCNTs.The coefficients of determination for the kinetics models are in the order of Elovich (R 2 = 0.96) > pseudo-first order (0.93) > pseudo-second order (0.46).Pseudo-first-order log(q e − q t ) = logq e − k 1 2.303 t q e (mg/g) 2.1 × 10 1 0.93 k 1 (/min) 2.7 × 10 −1 Pseudo-second-order Note: q e and q t are the amount of NP adsorbed at equilibrium at time t (mg/g), respectively.k 1 and k 2 are the rate constant of the pseudo-first-order, pseudo-second order, respectively.a is the Elovich constant and α is the initial sorption rate constant for the Elovich model (g/mg/min).D 1 and D 2 are the 1st and 2nd stage concentrations (mg/L) in the DEM model, while Dk 1 , and Dk 2 are the 1st and 2nd stage rate constants (1/s).Lastly, m a is dosage of an adsorbent in the DEM model (g/L).
The calculated q e value was not in agreement with the experimental data, indicating that the pseudo-second-order kinetic model may not explain the adsorption of NP onto the modified MWCNTs.Preferably, the experimental adsorption capacity data well fit to the Elovich model, which implies that the NP adsorption is governed by chemisorption processes [54].

Equilibrium Adsorption Isotherm
To examine the effect of the initial NP concentration on the adsorption capacity of the modified MWCNTs, the adsorption isotherm study was carried out.Table 4 shows the adsorption isotherm models applied for the adsorption of NP onto MWCNTs.The Langmuir [55], Freundlich [56], Jovanovic [57], Elovich [54] and Temkin [58] models were respectively employed to fit the adsorption data, and the result is summarized (Table 4).In short, Freundlich, Jovanovic, Temkin, and Elovich models fitted the data better than the Langmuir.The maximum NP adsorption capacity (q max , mg/g) of the modified MWCNTs estimated using the Langmuir model at pH 2 and 293 K was 1.11 × 10 4 mg/g.On the other hand, a much higher q max value (i.e., 25,000 mg/g) was estimated by Elovich isotherm.Furthermore, the calculated adsorption data matched well with the observed ones.The current result demonstrated that the adsorption of NP onto the modified MWCNTs is multilayer (chemisorption) rather than Langmuir-mono-layer.

Thermodynamic Aspects
Temperature affects the adsorption capacity of an adsorbent.It was observed that the amount of NP adsorbed increased as temperature increased, indicating an endothermic process.To evaluate the spontaneity of the adsorption process, the thermodynamic parameters, i.e., changes in the Gibb's free energy (∆G • , kJ/mol), the enthalpy (∆H • , kJ/mol), and the entropy (∆S • , kJ/mol/K) of the process were calculated by Equations (3)-( 5): ln where R is the universal gas constant (8.314 × 10 −3 kJ/mol•K), and T is the absolute temperature (K).The K L (L/g), equilibrium constant, is calculated from the Langmuir constant, q m [55].The value of ∆G • was calculated from Equation (3), while ∆H • and ∆S • were estimated from the slop and intercept of the graph (straight line) of ((1/T) vs. ln K) (Figure S2).The values of various thermodynamic parameters for the NP adsorption onto the modified MWCNTs are presented in Table 5.

Figure 1 .
Figure 1.(a) Sampling Locations and (b) NP concentration of source waters.The highest, second highest, and the third highest NP concentrations were detected at (M) Shin-shan, (F) Chen-chin Lake, and (C) Bei-dou, respectively.

Figure 2 .
Figure 2. SEM (left) and TEM (right) images of the MWCNTs (a) before and (b) after modification.

Figure 3 .
Figure 3. Zeta potential of modified MWCNTs for different pHs.

Table 2 .
Specific surface area and pore diameter of MWCNTs.

Table 3 .
Kinetic models and parameters for adsorption of NP to MWCNTs.

Table 4 .
Parameters of NP adsorption based on various models.

Table 5 .
Values of various thermodynamic parameters for adsorption of NP to MWCNTs.