Bifunctional graphene oxide–cellulose nanofibril aerogel loaded with Fe(III) for the removal of cationic dye via simultaneous adsorption and Fenton oxidation

Abstract

A highly porous cellulose nanofibril (CNF) aerogel loaded with graphene oxide–iron(III) (GO–Fe) nanocomposites was produced and used for the treatment of methylene blue (MB) in aqueous solution. The CNF aerogel serves as an adsorbent for the dye, while the GO–Fe nanocomposites play a role in the decomposition of the dye via the Fenton oxidation reaction. The aerogel exhibits rapid adsorption performance (less than 10 min) for removing MB, with a maximum adsorption capacity of 142.3 mg g⁻¹. On the side of enhancing the MB removal, the GO in the GO–CNF nanocomposite aerogel was loaded with 1 wt% of Fe(III) to perform as a catalyst for the Fenton oxidation reaction. The MB continues to decolorize by 30.4% more after 24 h of the reaction process. Moreover, by performing Fenton oxidation for adsorbent regeneration, the adsorption capacity for nanocomposite adsorption was reduced by 52.2% after five cycles of adsorption–oxidation.

---

1 Introduction

Recent promising developments of various techniques used for wastewater treatment has produced an improvement in treatment efficiency and overcome limitations compared to conventional methods. An adsorption technique, for example, in the utilization of low-cost adsorbents was seamlessly developed to overcome the high-cost production of activated carbon.^1,2 The cost of adsorbent production is not the only factor involved in developing an excellent adsorbent, there’s also adsorption performance, regeneration ability and adsorbent segregation. The massive quantities and continuous discharge of industrial effluents requires an attentive, rapid and efficient solution for effluent treatment. Thus, there are a variety of micro- to nano-sized adsorbents of different materials that were previously investigated in order to achieve rapid removal of wastewater effluents, e.g., modified activated carbon,³ nano-sized metal oxides,⁴ alumina nanoparticles,⁵ mesoporous silica,⁶ and cellulose nanofibrils (CNFs).⁷

Cellulose, a sustainable resource that can be extracted from lignocellulosic biomass, has been widely used in many forms of applications and products as it is a relatively stable polymer with high axial stiffness and high crystallinity, especially in the forms of micro- to nano-sized fibrillated cellulose.^8,9 The term CNFs refers to wood pulp defibrillated by physical means with chemical, enzymatic or physical pre-treatments.^10,11 Therefore, in order to produce CNFs, mechanical and physical disintegration is required. Various methods have been employed to defibrillate cellulose, e.g., by blender,¹² high-pressure homogenization,¹³ steam explosion¹⁴ and grinding.¹⁵ CNFs have many potential applications, such as reinforcement for composites,¹⁶ thickening agents,¹⁷ substrates for electronic devices,¹⁸ scaffolds for tissue growth,¹⁹ cationic dye adsorbent,⁷ etc.

In recent years, the research on graphene-based materials in various fields of applications has grown, blossoming mainly due to its exceptional physical and electronic characteristics.²⁰ Among them is graphene oxide, which is made of modified graphene sheets with a large number of reactive oxygen functional groups (hydroxyl, epoxy, carbonyl and carboxyl). These groups contribute to strong electrostatic interactions and the hydrophilicity of the large surface area of GO. The large surface area and great number of oxygen atoms on GO produces strong electrostatic interactions.^21–23 These factors help GO possess a high adsorption capacity against cationic dyes in aqueous solutions, comparable to that of commercial activated carbon.²⁴ However, its extremely small size and its hydrophilicity has limited its practical applications in wastewater treatment, as it requires the additional step of centrifugation for the recollection of the GO from the treated effluent.²⁵ To overcome the separation difficulties, several approaches have been adopted, including the formation of hydrogel²⁶ and aerogel,²⁷ encapsulation^28–30 and the functionalization of magnetic nanoparticles.^31,32

In our previous study, CNFs and GO proved to be efficient and promising adsorbents for the rapid removal of cationic dye.^7,24 Therefore, in the present work, GO was immobilized in a CNF aerogel and used to remove methylene blue from aqueous solution. Furthermore, in order to enhance the efficiency of the dye removal, the GO was loaded with Fe(III) ions to assist in the removal of MB via the Fenton oxidation reaction after the adsorption process of the aerogel. The adsorption performances (adsorption kinetics, maximum adsorption capacity, regeneration) of the aerogel with different amounts of GO were investigated.

2 Experimental

2.1 Materials

Kenaf core (KC) powder, obtained from the Malaysia Agricultural Research and Development Institute (MARDI); sodium chlorite, 80% (Acros Organics); and glacial acetic acid, 95% (R&M Chemicals) were used for the preparation of CNFs. GO was prepared using graphite flakes (Ashbury, Inc. USA); phosphoric acid, 85% (Merck); potassium permanganate, 99.9% (Merck); hydrogen peroxide, 30% (Merck); and ferric chloride (Systerm). Stock solutions of methylene blue (MB) trihydrate (Sigma-Aldrich) were prepared and diluted accordingly. The standard calibrations of the diluted MB solutions were measured using a UV-Vis spectrophotometer (Jenway 7315 spectrophotometer) at a λ_max of 664 nm.

2.2 Preparation of GO–CNF nanocomposite aerogel

CNFs were prepared from KC wood powder according to a previous study.⁷ Briefly, KC powder was rinsed several times with distilled water to remove impurities and delignified six times through acid-chlorite bleaching to obtain holocellulose (containing a total of 1.875 g per g of sodium chlorite and 1.25 g per g of acetic acid). Defibrillation of the holocellulose was performed using a high-speed blender (Vitamix, Vita-Prep 3) at a 0.7 wt% consistency of holocellulose and with 0.1 mM NaCl as a counter ion. The defibrillation process was carried out for 30 minutes. The produced CNFs were stored at 4 °C in a refrigerator for further use.

The GO used in this study was prepared using Hummer’s method.²⁴ To start the oxidation process, a mixture of H₂SO₄ (360 mL), H₃PO₄ (40 mL), KMnO₄ (18 g) and graphite flakes (3 g) was stirred for three days. Next, the solution was mixed with ice (400 mL) and H₂O₂ (27 mL) to stop the oxidation process. The solution was washed with HCl (1 M) three times and deionized water ten times until the solution reached a pH of 4–5. The produced GO suspension had 0.01 g of ferric chloride (FeCl₃), which is the equivalent of 1.0 wt% Fe(III), added to it and was stirred for 60 minutes to produce GO–Fe nanocomposites. The desired volume of GO–Fe was added into a beaker containing 20 mL of the CNF suspension (5 wt%). The mixture was stirred at room temperature for 30 minutes using a magnetic stirrer. The mixture was lyophilized in a freeze-dryer (brand model) for 24 hours to form an aerogel. The produced aerogels were kept in a desiccator for further use. Aerogels with different percentages of GO–Fe (5, 10, 15, 20 and 30 wt%) were produced. Also, CNF and GO aerogels were also prepared as control samples using the same procedure.

2.3 Characterizations of GO–CNF nanocomposites

The nanocomposites (CNF and GO–CNF) were analyzed using transmission electron microscopy (TEM) (CM 12 Phillips). The samples were diluted with ethanol (ca. 0.01 wt%), dropped onto a copper grid and stained with uranyl acetate (3 wt%). Attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) (Perkin-Elmer Spectrum 400) characterization for the functional groups of the freeze-dried samples (GO, CNF, and GO–CNF) was performed at a resolution of 1 cm⁻¹ in the range of 4000 cm⁻¹ to 650 cm⁻¹.

2.4 Adsorption experiments

2.4.1 Adsorption kinetics. Adsorption kinetics experiments were conducted in flasks containing 100 mL of MB solution at different concentrations. The initial pH of the dye solutions was fixed at 7, according to a previous study.⁷ The mixture was stirred at a constant speed of 250 rpm with a magnetic stirrer for 2 hours until equilibrium was achieved. Aliquots of solution (∼0.1 mL) were withdrawn at different time intervals, filtered using a nylon-membrane syringe filter (Agilent, 0.2 μm), diluted using deionized water, and analyzed using a UV-Vis spectrophotometer. The concentration of dye adsorbed at time t, q_t (mg g⁻¹), was calculated using the following equation:where C₀ and C_e are the initial and equilibrium concentrations of dye solution (mg L⁻¹), respectively. V is the volume of the solution (mL) and m is the mass of the adsorbent (g). Adsorption kinetics data was fitted to the pseudo-second order,³³ where k₂ (g mg⁻¹ min⁻¹) is the rate constant of the pseudo-second order expressed as below:

2.4.2 Adsorption isotherms. Adsorption isotherms were performed for a series of MB concentrations (50, 100, 200, 300, 400, 500 and 600 mg L⁻¹) at room temperature. A mixture of 25 mL of MB solution and 0.025 g of nanocomposite aerogel adsorbents was placed in a rotary shaker at a constant speed of 200 rpm for 2 hours. The final concentration of MB solution (C_e) at equilibrium was measured using a spectrophotometer to calculate the amount of adsorbed MB at equilibrium (q_e) using the following equation:

2.5 Simultaneous adsorption–Fenton oxidation

A similar procedure for adsorption kinetics experiments was performed with additional 40 mM H₂O₂ in the MB solution as an oxidizing agent. The mixture was stirred at a constant speed of 250 rpm with a magnetic stirrer for 24 hours or until equilibrium was achieved.

2.6 Regeneration studies

Adsorption was performed using 100 mg L⁻¹ MB for 2 hours to achieve adsorption saturation. The adsorbent was washed with deionized water to remove excessive MB. Fenton oxidation was carried out by mixing the washed adsorbent with 25 mL of 40 mM H₂O₂ for 24 hours. The adsorbent was again washed with deionized water to remove excessive H₂O₂. Regeneration of the nanocomposite adsorbents was performed for five cycles (adsorption–Fenton oxidation) using a similar procedure (100 mg L⁻¹ MB solution) at room temperature.

3 Results and discussion

3.1 Characterizations of nanocomposite

The fibrillation process on the KC powder produces aggregated CNFs with nano-diameter sizes, as shown in the TEM images (see Fig. 1(a)). Similar to the previous report on KC CNFs,⁷ the diameter of individual nanofibrils observed is in the range of 3.3–8.9 nm, with the average being 5.9 nm. On the other hand, the TEM images of the GO–CNF suspension show that a thin layer of GO was noticeably deposited on a web-like structure of CNFs (see Fig. 1(b)). The tips of the GO layer were ripped and grafted onto the CNFs, which is shown at a higher magnification in the TEM images (inset of Fig. 1(b)). In the suspension form, both GO and the CNFs interacted with each other via intermolecular hydrogen bonding that allowed a homogeneous alignment/connection between the nanocellulose and GO. Additionally, the interaction in the GO–CNF prevents the GO layers from over-stacking and retaining their active surface sites.^22,23

Fig. 1 Morphology images of the nanocomposite suspension from TEM for (a) CNFs and (b) 20% GO–Fe–CNF, and the aerogel from SEM for (c) CNFs and (d) 20% GO–Fe–CNF.

The morphology and structure of the CNF and GO–CNF aerogel nanocomposites are shown in Fig. 1(c) and (d), respectively. As shown in Fig. 1(c), the CNF aerogel exhibited a highly porous structure, which was likely formed throughout the freeze drying process. The intermolecular hydrogen bonding among the CNFs led to an entanglement of the CNFs, while the subsequent removal of water molecules resulted in the porous structure of the CNF foam.³⁴ Additionally, the SEM images of the GO–CNF nanocomposite exhibited a more porous structure, which is mainly due to the presence of GO. A three-dimensional (3-D) network was possibly formed out of these interactions between GO–CNF and GO–GO.³⁵ The additional GO has been attributed to a crosslinking effect with the CNFs due to the intermolecular hydrogen bonding between cellulose and GO.^22,36 However, the presence of Fe within the nanocomposite is hard to identify, taking into consideration the small amount of Fe(III) that was added.

The FTIR spectrum of the CNF nanocomposite (see Fig. 2), shows all of the characteristic peaks for cellulose (3341, 2915, 1640 and 1160 cm⁻¹), which are attributed to O–H stretching, aliphatic alkyl stretching, adsorbed water and ether linkages of pyranose, respectively.⁷ The IR spectrum of GO shows characteristic peaks at 1391 and 1737 cm⁻¹, which are due to C–O–H deformation and C=O stretching of COOH groups, respectively. The absorption bands at 1072 and 1249 cm⁻¹ are due to epoxy ring deformation and C–O stretching mixed with C–OH bending, respectively.²⁴ Both the GO–CNF and the GO–Fe–CNF nanocomposites have introduced peaks at 1419, 1621 and 1719 cm⁻¹, which can be attributed to the vibration of O–H, intramolecular hydrogen bonds and the surface functionality of GO and CNF, respectively.²²

Fig. 2 Characterizations of the nanocomposites (CNF, 20% GO–CNF and 20% GO–Fe–CNF) using FT-IR spectra.

A 100 g weight was placed on top of the GO, CNF and GO–CNF aerogels to investigate their compression properties. GO aerogel, a low-density material, is likely to have low mechanical properties and can be compressed easily under a 100 g mass (see Fig. 3). Meanwhile, the GO–Fe–CNF nanocomposite shows good mechanical properties and maintained its aerogel structure before and after the compression (see Fig. 3). The good mechanical properties of the CNF aerogel (flexibility and ductility) are attributed to the strong intermolecular hydrogen bonds and the highly porous structure of the CNFs during the drying process.³⁴ Therefore, the CNFs have constructed a good mechanical structure with the incorporation of GO to prevent the deformation of the aerogel.

Fig. 3 (a) Compression of the GO, CNF and 20% GO–Fe–CNF nanocomposite aerogels using a 100 g mass and (b) comparisons after the aerogels had been compressed.

3.2 Rapid adsorption of GO–CNF nanocomposite

GO possesses high surface functionality (hydroxyl, carboxyl, carbonyl, etc.) and a large specific area, which contribute to high adsorption capacity and rapid adsorption.²⁴ As the GO suspension formed into GO aerogel via freeze drying, the adsorption capacity was reduced by approximately 50% to 384 mg g⁻¹, which can be attributed to the decrease of the total exposed surface of the GO (see Fig. S1†).²⁸ The rapid adsorption kinetics of the CNF aerogel for the removal of MB were acquired, and the adsorption equilibrium was achieved in less than 5 minutes (see Fig. 4(b)). The CNF aerogel has a higher adsorption rate, k₂, compared to the GO suspension, 0.0218 and 0.0007 g mg⁻¹ min⁻¹, respectively (see Table S1†). The efficiency of CNFs to rapidly uptake cationic dye has been reported in our earlier work.⁷

Fig. 4 Adsorption kinetics of MB (600 mg L⁻¹) fitted to the pseudo-second order for (a) GO and CNF nanocomposite aerogels, and (b) at different wt% of GO–Fe on the CNF nanocomposites. (c) Images taken of MB solution (100 mg L⁻¹) before and after 30 minutes of adsorption using 20% GO–Fe–CNF.

In order to obtain a better adsorption capability, the GO suspension was used as a surface modifier on the CNFs to form a good nanocomposite adsorbent. By adding 5 wt% of GO on the CNF nanocomposite, its adsorption capacity was increased by 16%, i.e., from 100.3 to 116.5 mg g⁻¹. The high specific surface area and large amount of chemical functional groups of both GO and the CNFs are the main reasons for the rapid adsorption performances of the nanocomposites. In addition, as can be seen in Fig. 1(d), the pore structure of the CNF nanocomposite was undisrupted by the addition of GO.

The adsorption capacity of GO was reduced significantly after loading with 1 wt% of Fe(III), and was 28% lower than that of the pure GO suspension (see Fig. S1†). Moreover, a similar amount of Fe content was added onto the CNFs and resulted in a much lower MB uptake (5.4 mg g⁻¹) (see Fig. 4(a)). The reduced adsorption capacity for both GO and the CNFs can be attributed to the reduction of reactive sites due to them being partially occupied by the Fe ions via electrostatic interaction.³⁷

In order to identify the optimum parameters for the GO–Fe–CNF nanocomposite, the amount of GO–Fe added onto the CNFs was increased accordingly (5, 10, 15, 20 and 30 wt%) (see Fig. 4(b)). The adsorption capacity of the GO–Fe is much higher compared to the CNFs, therefore, it is expected that the increasing amount of GO–Fe in the CNFs will boost their adsorption capacity. 30% GO–Fe in the CNFs increased the adsorption capacity to 143 mg g⁻¹. However, the increment of GO–Fe content in the nanocomposite has prolonged the adsorption equilibrium time and decreased the adsorption rate, k₂, from 0.0091 to 0.0036 g mg⁻¹ min⁻¹. This could be due to the fact that the GO sheets overlay each other at high concentration, whereby lower GO content is favored for better interaction with CNFs for preventing over stacking of the GO layers.^22,23

3.3 Adsorption isotherm

Briefly, the Langmuir derivation assumes monolayer adsorption between the adsorbate molecules and adsorbent medium, while the Freundlich model explains the heterogeneous adsorption behavior of the adsorption mechanism. The Langmuir isotherm model can be expressed as:where Q₀ is the maximum adsorption capacity per unit mass of adsorbent (mg g⁻¹) and b is a constant related to the adsorption energy (L mg⁻¹).³⁸ To determine whether adsorption is “favorable” or “unfavorable”, a dimensionless constant separation factor or equilibrium parameter, R_L, was calculated using the following equation:where b is the Langmuir constant (L mg⁻¹) and C_m is the highest initial MB concentration (mg L⁻¹). The value of R_L indicates whether the type of isotherm is irreversible (R_L = 0), favorable (0 < R_L < 1), linear (R_L = 1) or unfavorable (R_L > 1).

Meanwhile, the Freundlich isotherm model is expressed as:

where K_F and 1/n_F are the Freundlich constants, with K_F representing the relative adsorption capacity of the adsorbent and n_F representing the degree of dependence of adsorption on the equilibrium concentration of MB.³⁹ The experimental data were plotted for the non-linearized Langmuir and Freundlich models (see Fig. 5), and the calculated isotherm constants are summarized in Table 1. The adsorption data plotted for the CNF and GO–CNF nanocomposites were well fitted by the Langmuir model (correlation coefficient, r² ∼ 0.99) and represent monolayer adsorption, which is similar to the adsorption mechanism of GO and CNF suspension.^7,24 The maximum adsorptions of the CNF, 20% GO–CNF and 20% GO–Fe–CNF nanocomposites acquired from the Langmuir model were 97.1, 142.3 and 126.6 mg g⁻¹, respectively. The adsorption performances of the GO–CNF nanocomposites were much lower than for the GO suspension, which is in agreement with recent studies for adsorption of MB for most of the GO-substrates or GO nanocomposites (see Table 2). It can be explained by the interactions of GO with the support materials and the reduction of active surface sites for the uptake of adsorbate, which is difficult to avoid.

Fig. 5 Non-linearized adsorption isotherms of the GO–Fe–CNF nanocomposite compared to those of GO–CNF and CNF nanocomposites (temperature: 20 °C; adsorbate dose: 50–600 mg L⁻¹).

Table 1 Calculated non-linear isotherm models (Langmuir and Freundlich) for the CNF, 20% GO–CNF and 20% GO–Fe–CNF nanocomposites^a

Adsorbent	Langmuir model				Freundlich model
	Q0	b	RL	r^2	KF	n	r^2
a Q0 (mg g^−1) = maximum adsorption capacity; b (L mg^−1) = constant related to the adsorption energy; RL = equilibrium parameter; KF ((mg g^−1)(L mg^−1)^1/n) = relative adsorption capacity; n = degree of dependence of adsorption; r^2 = coefficient correlation.
CNF	97.1	0.15	0.011	0.994	2.65	6.01	0.901
20% GO–CNF	142.3	0.42	0.004	0.997	50.3	6.86	0.904
20% GO–Fe–CNF	126.6	0.25	0.006	0.997	9.15	7.02	0.885

---

Table 2 Adsorption performances of various GO-substrates for MB removal in recent studies

Adsorbent	Maximum adsorption capacity, Q0 (mg g^−1)	Reference
GO	752.1	This work
GO–hydrogel	7.9	Tiwari et al. 2013 [ref. 26]
GO–polyethersulfone	62.5	Zhang et al. 2013 [ref. 30]
GO–magnetic chitosan	180.8	Fan et al. 2013 [ref. 31]
20% GO–CNF	142.3	This work
20% GO–Fe–CNF	126.6	This work

---

3.4 Adsorption with Fenton oxidation

The kinetics of the simultaneous adsorption and Fenton oxidation of MB is plotted in Fig. 6. Since both nanocomposites (20% GO–CNF and 20% GO–Fe–CNF) show rapid adsorption used for MB removal, at the initial stage of the simultaneous adsorption–Fenton oxidation process, the performance of the adsorbents are consistent with the single adsorption process (see Fig. S2†). As soon as the 20% GO–Fe–CNF achieved adsorption equilibrium at ∼30 min, the concentration of MB started to decrease gradually as H₂O₂ was introduced (see Fig. 6). On the contrary, without Fe(III) loading, no Fenton oxidation occurred on the 20% GO–CNF nanocomposite. Additionally, the adsorption performance of the 20% GO–Fe–CNF was increased by 62.4–86.7% with the assistance of Fenton oxidation. The catalytic activity of the 20% GO–Fe–CNF in the presence of H₂O₂ towards MB suggested that the hydroxyl radical ˙OH slowly reduced the MB concentration.⁴⁰

Fig. 6 Fenton oxidation assisted adsorption of GO–Fe–CNF nanocomposite: rapid adsorption phase and Fenton oxidation phase (temperature: 20 °C; adsorbate dose: 200 mg L⁻¹; H₂O₂: 40 mM).

Based on the catalytic activity of the 20% GO–Fe–CNF on MB, the regeneration of the nanocomposite adsorbent was performed with five cycles of MB adsorption and Fenton oxidation with H₂O₂ for 24 hours (see Fig. 7). The regeneration results of the 20% GO–Fe–CNF nanocomposite show that its adsorption efficiency was decreased after three cycles of the adsorption–oxidation process. The reduction in the adsorption and oxidation performances may be due to the repetitive washing and drying processes, which contributed to the loss of adsorbent and Fe(III) ions. Contrary to the 20% GO–CNF, the adsorption performances were reduced drastically after the second cycle of adsorption. The desorption performance of the 20% GO–CNF was lower compared to that of neat CNFs obtained from our previous study, as a result of strong interactions between MB and GO–CNF.⁷ Previous studies on adsorbents loaded with Fe (activated carbon, montmorillonite, etc.) also demonstrated adsorption–oxidation behavior with good adsorption–regeneration cycles of the adsorbent.^37,40,41

Fig. 7 Regeneration of the GO–CNF and GO–Fe–CNF nanocomposites by desorption (20% GO–CNF) and adsorption–oxidation cycles (20% GO–Fe–CNF) (temperature: 20 °C; adsorbate dose: 100 mg L⁻¹; H₂O₂: 40 mM).

4 Conclusions

In conclusion, the produced aerogel nanocomposite composed of GO and CNFs demonstrated a rapid adsorption of MB, and the adsorption capacity was improved compared to their stand-alone forms. By loading the surface active sites of GO with Fe(III), the removal of MB via an adsorption–Fenton oxidation process was twice as effective than the purely adsorption process. Although it takes a longer time to complete the oxidation phase at a high concentration of MB, this process offered the advantage of regenerating the nanocomposite adsorbent. Compared with the desorption and oxidation processes, the GO and CNF nanocomposites can be regenerated for five cycle adsorption–oxidation processes. Despite the fact that the interactions of GO with the supported material reduced the active surface sites for the uptake of the adsorbate, the recovery of GO after the adsorption is more desirable in wastewater treatment, especially for industry applications.

Acknowledgements

This research was supported by research grants from UKM (DIP-2014-013 & DIP-2015-009) and the Ministry of Higher Education (LRGS/TD/2012/USM-UKM/PT/04).

References

1. A. Bhatnagar, W. Hogland, M. Marques and M. Sillanpää, Chem. Eng. J., 2013, 219, 499–511.
2. M. Rafatullah, O. Sulaiman, R. Hashim and A. Ahmad, J. Hazard. Mater., 2010, 177, 70–80.
3. A. Asfaram, M. Ghaedi, S. Hajati, M. Rezaeinejad, A. Goudarzi and M. K. Purkait, J. Taiwan Inst. Chem. Eng., 2015, 53, 80–91.
4. M. Hua, S. Zhang, B. Pan, W. Zhang, L. Lv and Q. Zhang, J. Hazard. Mater., 2015, 211–212, 317–331.
5. S. Banerjee, R. K. Gautam, A. Jaiswal, M. C. Chattopadhyaya and Y. C. Sharma, RSC Adv., 2015, 5, 14425–14440.
6. A. Shahat, M. R. Awual, M. A. Khaleque, M. Z. Alam, M. Naushad and A. M. S. Chowdhury, Chem. Eng. J., 2015, 273, 286–295.
7. C. H. Chan, C. H. Chia, S. Zakaria, M. S. Sajab and S. X. Chin, RSC Adv., 2015, 5, 18204–18212.
8. S. J. Eichhorn, A. Dufresne, M. Aranguren, N. E. Marcovich, J. R. Capadona, S. J. Rowan, C. Weder, W. Thielemans, M. Roman, S. Renneckar, W. Gindl, S. Veigel, J. Keckes, H. Yano, K. Abe, M. Nogi, A. N. Nakagaito, A. Mangalam, J. Simonsen, A. S. Benight, A. Bismarck, L. A. Berglund and T. Peijs, J. Mater. Sci., 2010, 45, 1–33.
9. H. P. S. A. Khalil, A. H. Bhat and A. F. I. Yusra, Carbohydr. Polym., 2012, 87, 963–979.
10. G. Siqueira, J. Bras and A. Dufresne, Polym., 2010, 2, 728–765.
11. D. Klemm, F. Kramer, S. Moritz, T. Lindström, M. Ankerfors, D. Gray and A. Dorris, Angew. Chem., Int. Ed., 2011, 50, 5438–5466.
12. T. Saito, Y. Nishiyama, J.-L. Putaux, M. Vignon and A. Isogai, Biomacromolecules, 2006, 7, 1687–1691.
13. M. Pääkkö, M. Ankerfors, H. Kosonen, A. Nykänen, S. Ahola, M. Österberg, J. Ruokolainen, J. Laine, P. T. Larsson, O. Ikkala and T. Lindström, Biomacromolecules, 2007, 8, 1934–1941.
14. B. Deepa, E. Abraham, B. M. Cherian, A. Bismarck, J. J. Blaker, L. A. Pothan, A. L. Leao, S. F. D. Souza and M. Kottaisamy, Bioresour. Technol., 2011, 102, 1988–1997.
15. Q. Q. Wang, J. Y. Zhu, R. Gleisner, T. A. Kuster, U. Baxa and S. E. McNeil, Cellulose, 2012, 19, 1631–1643.
16. Q. Cheng, S. Wang and T. G. Rials, Composites, Part A, 2009, 40, 218–224.
17. K. Dimic-Misic, P. A. C. Gane and J. Paltakari, Ind. Eng. Chem. Res., 2013, 52, 16066–16083.
18. G. Zheng, Y. Cui, E. Karabulut, L. Wågberg, H. Zhu and L. Hu, MRS Bull., 2013, 38, 320–325.
19. N. Ninan, M. Muthiah, I.-K. Park, A. Elain, S. Thomas and Y. Grohens, Carbohydr. Polym., 2013, 98, 877–885.
20. S. Basu and P. Bhattacharyya, Sens. Actuators, B, 2012, 173, 1–21.
21. S.-T. Yang, S. Chen, Y. Chang, A. Cao, Y. Liu and H. Wang, J. Colloid Interface Sci., 2011, 359, 24–29.
22. Y. Liu, J. Zhou, E. Zhu, J. Tang, X. Liu and W. Tang, J. Mater. Chem. C, 2015, 3, 1011–1017.
23. B. Wang, W. Lou, X. Wang and J. Hao, J. Mater. Chem., 2012, 22, 12859–12866.
24. C. H. Chia, N. F. Razali, M. S. Sajab, S. Zakaria, N. M. Huang and H. N. Lim, Sains Malays., 2013, 6, 819–826.
25. X. Zhou and X. Yang, Carbon, 2012, 50, 4566–4572.
26. J. N. Tiwari, K. Mahesh, N. H. Le, K. C. Kemp, R. Timilsina, R. N. Tiwari and K. S. Kim, Carbon, 2013, 56, 173–182.
27. X. Mi, G. Huang, W. Xie, W. Wang, Y. Liu and J. Gao, Carbon, 2012, 50, 4856–4864.
28. B. Yu, J. Xu, J.-H. Liu, S.-T. Yang, J. Luo, Q. Zhou, J. Wan, R. Liao, H. Wang and Y. Liu, J. Environ. Chem. Eng., 2013, 1, 1044–1050.
29. W. M. Algothmi, N. M. Bandaru, Y. Yu, J. G. Shapter and A. V. Ellis, J. Colloid Interface Sci., 2013, 397, 32–38.
30. X. Zhang, C. Cheng, J. Zhao, L. Ma, S. Sun and C. Zhao, Chem. Eng. J., 2013, 215–216, 72–81.
31. L. Fan, C. Luo, M. Sun, X. Li and H. Qiu, Colloids Surf., B, 2013, 103, 523–529.
32. L. Fan, C. Luo, M. Sun, X. Li, F. Lu and H. Qiu, Bioresour. Technol., 2012, 114, 703–706.
33. Y. S. Ho and G. Mckay, Process Biochem., 1999, 34, 451–465.
34. W. Chen, H. Yu, Q. Li, Y. Liu and J. Li, Soft Matter, 2011, 7, 10360–10368.
35. L. Yao, Y. Lu, Y. Wang and L. Hu, Carbon, 2014, 69, 552–562.
36. S. W. Chook, C. H. Chia, S. Zakaria, M. K. Ayob, N. M. Huang, H. M. Neoh and R. Jamal, RSC Adv., 2015, 5, 26263–26268.
37. L. Wang, Y. Yao, Z. Zhang, L. Sun, W. Lu, W. Chen and H. Chen, Chem. Eng. J., 2014, 251, 348–354.
38. I. Langmuir, J. Am. Chem. Soc., 1916, 39, 2221–2295.
39. H. M. F. Freundlich, J. Phys. Chem., 1906, 57, 385–470.
40. Y. Yao, L. Wang, L. Sun, S. Zhu, Z. Huang, Y. Mao, W. Lu and W. Chen, Chem. Eng. Sci., 2013, 101, 424–431.
41. L. Guz, G. Curutchet, R. M. T. Sánchez and R. Candal, J. Environ. Chem. Eng., 2014, 2, 2344–2351.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra26193g

---