The formation of hybrid carbon nanomaterial by chemical vapor deposition: an efficient adsorbent for enhanced removal of methylene blue from aqueous solution

Many industrial wastewater effluents, specifically those from the textile industry, account for the high concentration of methylene blue in the environment. MB has a non-biodegradable aromatic structure and it is difficult to remove from discharged effluents. This contributes to high chemical oxidation demand and poses a serious risk to human health and the aquatic system (Das et al. 2016).

Various techniques have been implemented to treat the dye containing wastewater, including coagulation, ion-exchange, precipitation, membrane separation, photo degradation, biological treatment and advanced oxidation processes in conjunction with biological treatments. Among these methods, the adsorption technique is regarded as a competitive and economical process due to its simplicity, ease of operation and high efficiency without producing any harmful byproducts (Peng et al. 2015).

Various adsorbents have been studied to determine their efficiencies for the removal of MB from aqueous streams, e.g. rice husk, montmorillonite, activated carbon, carbon nanotubes (CNTs) and graphene (Altenor et al. 2009; Madrakian et al. 2011; Liu et al. 2012). Carbon nanomaterials (CNMs) have been introduced as remarkable adsorbents due to their unique features, i.e. large surface area, high reactivity and rapid adsorption capability. Despite all these extraordinary properties of CNMs, the recovery of nano-structures with minimum separation steps and minimum loss of material is a serious drawback (AlSaadi et al. 2011).

Generally, several schemes have been used to produce CNMs, e.g. arc discharge, laser ablation and chemical vapor deposition (CVD). However, CVD is the most promising and scalable approach. It is a versatile, cost-effective technique and produces high quantities of CNMs at low temperatures (Yamamoto et al. 2017). Typically, nanometer-size metal particles such as Fe, Co, and Ni are required to enable the decomposition of hydrocarbons. The CNTs produced from methane decomposition over Ni catalyst exhibited higher thermal stability than those formed using Fe catalyst (Shah & Tali 2016). Lee et al. reported that the maximum growth rate of CNTs was achieved when the Ni catalyst was used. They found that the diffusion rates of carbon were in the order of Ni > Fe > Co (Lee et al. 2002).

Hybridizing different carbon nanostructures on the surface of activated carbon (PAC) to produce material with new features appears to be a promising candidate for the removal of dyes from aqueous effluents. Unlike metallic supports, these materials retain their attachment with the substrate which eliminates the need for an additional recovery step. Moreover, the resulting hierarchical carbon combined the unique properties of CNMs with those of PAC, while maintaining chemical compatibility between these two phases. The hybridized materials have been receiving increased attention due to their unique properties; however, their application for water treatment remains to be proven (Shoukat et al. 2017).

Based on the preceding points, we used CVD method to prepare a hybrid carbon nanomaterial (HCNM) on PAC impregnated with Ni catalyst. The hybrid material was characterized using field emission scanning electron microscope (FESEM), transmission electron microscope (TEM), Raman spectroscopy, thermo-gravimetric analysis (TGA), Fourier transform infrared (FTIR), energy-dispersive X-ray spectroscopy (EDX), Brunauer–Emmett–Teller (BET), and zeta potential. We chose MB as a model dye to evaluate the adsorption performance of the HCNM. The response surface methodology (RSM) with central composite design (CCD) were used to optimize the effect of adsorption parameters including pH, HCNM dose, and contact time on the removal of MB. Moreover, kinetic and isotherm studies were investigated at the optimal removal conditions.

Nickel (II) nitrate hexahydrate (Ni (NO₃)₂·6H₂O), PAC, MB, sodium hydroxide (NaOH), and hydrochloric acid (HCl) were purchased from Sigma Aldrich, Malaysia. All chemicals were of analytical grades.

The PAC was dried overnight at 120 °C before being impregnated with nickel (II) nitrate. The catalyst precursor was mixed with 2.0 g of PAC with a solution of 1% w/v nickel (II). The mixture was sonicated for 1 h at 60 °C and then heated in an oven at 120 °C overnight to dry out. Using the CVD tubular furnace, a two-stage thermal treatment was performed; a calcination process at 350 °C for 2 h in an N₂ atmosphere followed by reduction using H₂ gas at 550 °C for 1 h. Then the impregnated substrate (Ni-PAC) was utilized for HCNM growth (Onundi et al. 2011).

The HCNM was obtained on Ni-PAC via the decomposition of methane (CH₄) in a reductive atmosphere of the tubular CVD furnace. Using a temperature increase rate of 10 °C/min, the furnace was heated to 950 °C. The flow rates of H₂ and CH₄ were adjusted at 40 mL/min to provide hydrogen to methane flow ratio of 1.0. Then, the mixture of gases was passed through the reaction tube for 20 min to initiate the growth reaction.

The chemical composition of HCNM was analyzed using EDX. The BET surface area was determined using surface area analyzer. The morphology of HCNM was investigated using FESEM and TEM. The HCNM was characterized using Raman spectroscopy (Renishaw 2000-Spectrometer). TGA was conducted in oxygen with a heating rate of 10 °C min⁻¹. FTIR spectra were recorded by IR-21 FTIR. Ultraviolet-visible (UV-vis) measurements were conducted using UV-2300 spectrophotometer. The surface charge of HCNM suspension was conducted using a Zetasizer (Malvern).

The RSM/CCD method using Design Expert V7.0 software was selected to optimize the effects of the adsorption parameters; solution pH (x₁: 3–11), HCNM dose (x₂: 5–20 mg), and contact time (x₃: 20–120 min) on the process responses; removal percentage (RV %) and HCNM adsorption capacity (Q, mg·g⁻¹). The F-value and P-value of analysis of variance (ANOVA) were used to determine the significance of the proposed model (Dutta et al. 2011). A two-factor level CCD with 15 experimental runs was conducted. To achieve the highest performance, the desired responses were defined as ‘maximize’. The actual parameters used in the design, the RV % and Q are listed in Tables S1 and S2 (Supporting Information, available with the online version of this paper).

A convenient expression of the adsorption rate is provided by kinetic models. In the kinetic study, the optimum removal conditions suggested by the optimization study were employed in the following models: pseudo-first-order, pseudo-second-order and intraparticle diffusion. The optimum dose of HCNM was shaken at room temperature into 50 mL of a MB solution of 50 mg·L⁻¹ at constant agitation speed of 180 rpm. The applied contact times were (5, 10, 15, 20, 25, 30, 40, 50, 60, 90, 120, 150, 180, and 1,440 min (24 h)). A known volume of the solution was centrifuged at 4,000 rpm for 10 min. The MB concentration in the supernatant was measured by UV-vis at absorbance wavelength of 665 nm.

The surface morphology of the HCNM were characterized using BET, FESEM, TEM and EDX. Table 1 summarizes the BET surface area of PAC, Ni-PAC and HCNM. The PAC catalyzed the production of HCNM and enhanced the surface area from about 100 to about 165 m²·g⁻¹. The surface area of Ni-PAC was reduced due to the blockage of the PAC pores by the impregnated catalyst. Song et al. results show that activated carbon possesses good activity to catalyze the formation of CNTs which likely is associated with its porous structures (Song et al. 2010). The total pore volume of HCNM is 0.29 cm³·g⁻¹, where it was only 0.09 cm³·g⁻¹ for PAC. Given this difference, improvement of HCNM sorption capacity is expected. Allaedini et al., reported that the enhancement in the adsorbent surface area was attributed to the small particle size of the impregnated catalyst (Allaedini et al. 2015).

FESEM and TEM images were taken for the Ni-PAC (Figure 1(a) and 1(b)). The carbon support has irregular and rough pores with well dispersed Ni particles. The TEM image of Figure 1(b) shows that catalyst was trapped successfully into PAC. The large surface area and high porosity of PAC prevent the agglomeration of the catalyst particles and provide a good catalyst distribution. The images of HCNM in Figure 1(c) and 1(d) show short CNTs embedded in the abundant porous structure of PAC. The HCNM has microstructures with groove-like features, and very similar to the surface conditions of the PAC support.

The EDX result in Figure S1 (Supporting Information, available with the online version of this paper) for HCNM confirms the successful growth of the HCNM. Most of the surface (98.6%) was carbon, but other impurities were also present, including Si and Fe (<2.0%). The Ni content had decreased from 2.0% to 0.9% after the growth.

Figure 2 shows the Raman spectra, TGA and FTIR curves. The Raman spectra for PAC, Ni-PAC and HCNM in Figure 2(a) show the D and G peaks with high intensity near 1,360 cm⁻¹ and 1,590 cm⁻¹, respectively. The peak near 1,360 cm⁻¹ is the D band which is a defect-induced mode originated from the distorted hexagonal sp³ carbon network (Lee et al. 2013). The peak near 1,590 cm⁻¹ is the G band, which is reflected a typical graphite mode of C = C bonds. The extent of the carbon-containing defects of the HCNM can be evaluated by intensity of the D band to that of G band (I_D/I_G). The I_D/I_G ratio for HCNM was of 0.93, which was similar to a previously reported value (Izadi et al. 2011). The appearance of unusually shaped 2D peaks at 2,500 − 3,200 cm⁻¹ could have been due to effects of strain on the structure (Hernandez et al. 2013).

The oxidation behaviour of the adsorbent was investigated using TGA. Figure 2(b) shows that PAC, Ni-PAC and HCNM exhibited single step degradation. The oxidation onset points of PAC, Ni-PAC and HCNM occurred at about 430, 490 and 510 °C, respectively. High thermal stability of HCNM at temperatures higher than 800 °C is observed through its high oxidation onset point. The weight residue for PAC sample at the end of the test was about 7 wt%, which contributed to the impurities in the substrate. This percentage increased to 21 wt% for the Ni-PAC sample after catalyst impregnation. However, the remaining weight in the HCNM sample above 650 °C was about 14 wt%. This reduction in the residual percentage was attributed to the growth of the new nanostructures on the catalyst surface, which led to a variation in the overall composition (Deshmukh et al. 2010).

FTIR analysis of the HCNM was used to identify the groups responsible for the adsorption of MB. Figure 2(c) shows FTIR spectra of the HCNM-MB sample. The bands at 1,400 to 1,735 cm⁻¹ were assigned to the C = O stretch mode of carboxylic acid and carbonyl moieties. The bands for HCNM and HCNM-MB at 3,500 to 3,700 cm⁻¹ were due to hydroxyl groups. Before the adsorption of MB, the HCNM presented the symmetric stretching vibration bands of C = O groups at 1,420 cm⁻¹, and the band at 1,580 cm⁻¹ was assigned to C = C skeletal stretching. The band at about 1,110 cm⁻¹ represented the C–O vibration of various oxygen-containing groups. The peaks at 2,922 and 2,855 cm⁻¹ in HCNM were related to the symmetric alkane stretching of C—H bond, which become weaker after the adsorption of MB (Wang et al. 2015). The FTIR spectrum of free MB exhibited its ring stretch at 1,604 cm⁻¹, the symmetric stretch of C-N at 1,398 cm⁻¹, and symmetric deformation of -CH3 at 1,354 cm⁻¹. The heterocycle bending vibrations of the C–H groups of the unsaturated dimethyl-amino groups is described by the peak at 879 cm⁻¹. The peaks found in the HCNM-MB in the range of 1,200–1,500 cm⁻¹ are in accordance with the peaks that originated from the spectra of MB. The ring stretching band at 1,604 cm⁻¹ in the MB- free spectrum probably was shifted to 1,715 cm⁻¹, and the symmetric stretch of C–N in MB at 1,398 cm⁻¹ became weaker after its adsorption onto the HCNM (Manilo et al. 2016).

Zeta potential is the potential difference between the stationary fluid layer attached to the adsorbent and the dispersion medium (Kyzas & Matis 2015). The HCNM exhibited a low and positive value (+9.64 mV), which suggested that the adsorbent surface had a positive charge. The pH of the zero point of charge (pH_zpc) for HCNM was determined by pH drift test to be 9.3. This result indicated that the adsorbent surface was positively charged at a solution pH of <9.3. Under the working conditions of pH11, the adsorbent can be considered negatively charged. This means that, in neutral to alkali media, the electrostatic interactions between the negative adsorbent and cationic MB lead to more attraction of dye molecules to the adsorbent (Fu et al. 2015).

Figure S2(a) and S2(b) (available with the online version of this paper) present the theoretical and the experimental values for RV % and Q, respectively. The predicted values were quite close to the experimental values. The optimum values for MB removal using the obtained model were 11, 15 mg and 120 min for pH, adsorbent dose and reaction time, respectively. At these optimum values, a set of experiments were conducted using the procedure discussed in the experiment section. The observed removal efficiency of MB and the adsorbent capacity were about 92% and 153 mg·g⁻¹, which were close to the RSM analysis results (94% and 161 mg·g⁻¹) with a mean error <5%.

The removal efficiency % and the adsorption capacity of HCNM over different combinations of independent variables were presented by the RSM plot as a function of two independent parameters (Figure 3). The pH and adsorbent dose were the most influential variables in MB adsorption. The removal efficiency improved as the pH was increased, reaching its maximum value at pH = 11 (Figure 3(a) and 3(b)). The pH can affect the surface charge of the adsorbent, the degree of pollutants ionization, the dissociation of functional groups on the adsorbent as well as the structure of the dye molecule. Hence, the higher removal of MB that was observed at pH11 can be explained by the deprotonation of some functional groups resulting in more negatively charged surface (pH > pH_pzc). Also, more dissociation of the MB molecules (pKa = 3.8 for MB) occurred at high pH values which enhanced the electrostatic attraction between the HCNM and the MB molecules. On the other hand, in acidic solutions, the competitive effects of excess H⁺ and the electrostatic repulsion between the cationic dye would result in a decrease in the removal efficiency of the dye (Ai et al. 2011).

Figure 3(c) and 3(d) show that, at constant pH values, the adsorbent capacity and its removal efficiency were increased as the amount of adsorbent increased. The significant effect of the HCNM dosage on both responses was confirmed by the high F-values (98.22 and 64.63) presented in Tables S3 and S4, respectively (available with the online version of this paper). The adsorption enhancement could be due to the availability of more adsorption sites. Conversely, at low dosages, the removal percentage decreased significantly because longer time is required for the adsorption process to reach equilibrium due to the reduction of reactive sites with respect to the dye molecules (Krishni et al. 2014). The effect of the interaction between the pH and the adsorbent dosage (x₁ x₂) was also valuable on the adsorbent capacity since it exhibited an F-value of 19.43.

Kinetic analyses were conducted at the optimal removal conditions using pseudo-first-order and pseudo-second-order, as well as intraparticle diffusion. Table S5 (available online) summarizes the related equations and the characteristic parameters. In the linear form of the pseudo-first-order model, q_e and q_t are the amounts of the adsorbed MB at equilibrium and at time t, respectively, and K₁ is the adsorption rate constant. The pseudo-second-order rate constant of adsorption K₂ and q_e were determined from the slope and intercept of plots of t/q_t versus t (Figure 4). The removal rate is designated by the square root of time in the intraparticle diffusion (ID). The values of q_e and K₁ in Table S5 were obtained from the intercept and slope of the plot of ln (q_e-q_t) versus t (Figure S3(a)) (Figure S3 is available online). The values of the ID rate constant, K_d and the thickness of the boundary layer in terms of c values were depicted from the slope and intercept of the plot of q_t vs. t^0.5 (Figure S3(b)). As seen from Figure S3(b) the ID plot was not linear over whole time range and did not pass through the origin. This indicates the participation of ID in the adsorption process but it was not the only rate-controlling step (Zhang et al. 2015).

Freundlich isotherm constants are presented by K_F and n and determined from the intercept and slope of the plot of ln q_e versus ln C_e (Figure S4(a), available online). The distribution coefficient, K_F, represents the amount of MB adsorbed by HCNM per a unit equilibrium concentration. The heterogeneity of the adsorbent surface and the adsorption favourability were evaluated by the magnitude of 1/n. As this value approaches zero, it indicates that the adsorbent surface is becoming more heterogeneous (Tahir et al. 2017). The Temkin isotherm is characterized by uniform distribution of the binding energies. In the equation, B and K_T were determined from the slope and intercept of the linearized plot of q_e vs. ln C_e (Figure S4(b)). B (dimensionless) = RT/b, where b is the Temkin constant related to the heat of adsorption (J·mol⁻¹), R is the universal gas constant, T is the absolute temperature in Kelvin, and K_T is the Temkin equilibrium binding constant.

The results in Table S6 show the adsorption favourability of MB onto HCNM confirmed by the constants of Langmuir and Freundlich isotherms (R_L = 0.031 and n = 2.9). The respectful MB removal demonstrated by Freundlich model assumes the presence of non-equivalent binding sites in the heterogenous surface energy system. A similar result was obtained for MB removal onto low cost bio-waste (Krishni et al. 2014). However, Langmuir isotherm provided the best fit (Figure 5) with the highest R² at maximum adsorption capacity of 250 mg·g⁻¹. This suggests the occurrence of monolayer adsorption of MB at the homogeneous sites of the HCNM surface.

In this study, a novel sorbent and a simple pathway was used to synthesize HCNM with multiscale structure. The as-synthesized HCNM exhibited good graphitic structure and excellent sorption capacity. In alkaline medium, at pH of 11, excellent removal of MB was achieved. The RSM/CCD design effectively optimized the adsorption parameters such as pH, adsorbent dose and contact time. Based on the optimization studies, the adsorption of MB was highly dependent on the pH of solution, and the adsorbent surface charge. The adsorption kinetics and isotherms studies demonstrated that the adsorption process obeyed the pseudo-second-order kinetics and Langmuir isotherm model, respectively. The as-synthesized HCNM exhibited maximum adsorption capacity of 250 mg·g⁻¹. The produced HCNM can be used directly in adsorption processes without removing the substrate from the collected product, unlike other CNMs prepared on metal substrates. The main strength of MB adsorption is the electrostatic interaction, while the π-π stacking interaction between the aromatic backbone of the dye and the graphene planes of the HCNM might also contribute to the whole interaction. The results of the present investigation indicate that HCNM is a potential adsorbent for methylene blue removal from aqueous solutions.

The authors express their thanks to the University of Malaya UMRG (RP017A-13AET) for providing a Postgraduate Research Grant (PG243-2015B) to fund this research.