*3.1. Adsorbent Characterization*

TEM images confirmed the nanoscaled structure of all the prepared nanofibril samples. Specifically, LCNF has an average width of 25.3 ± 6.7 nm (Figure 1A). Compared with LCNF, the CLCNF samples have well-individualized structures with a homogeneous size distribution (Figure 1B–D). This was more directly reflected in the width of CLCNF samples. As shown in Figure 1E, the average width for CLCNF-1, CLCNF-2 and CLCNF-3 was 7.9 ± 1.7, 5.5 ± 1.0 and 4.6 ± 0.8 nm, respectively. These results suggest that the cationization of bagasse could facilitate the mechanical disintegration process, thus resulting in nanofibers with small and homogeneous widths. Such a positive effect for mechanical disintegration is especially remarkable when upgrading the amount of GTAC in DES. Additionally, the precursor of LCNF needs to be mechanically pretreated for avoiding

the blockage of the chamber for the high-pressure homogenizer. However, this is not the case with the cationized bagasse fibers, which can readily pass through the chamber of the high-pressure homogenizer without the occurrence of clogging. It is presumed that the charged groups on the cationized bagasse fibers can create electrostatic repulsion between fibirls and thus enhance the penetration of water into the fibers to create osmotic pressure. Undoubtedly, such effects would promote the mechanical defibrillation process of fibers [26,35].

**Figure 1.** TEM images of LCNF (**A**), CLCNF-1 (**B**), CLCNF-2 (**C**), CLCNF-3 (**D**) and their corresponding widths (**E**). Scale bar is 100 nm.

The suspension transmittance is one of the main methods to indirectly evaluate the degree of fibrillation [36]. The light transmittance measurement results of the nanofiber suspensions are shown in Figure 2A. It was shown that the percentage transmittance in the UV–visible range of LCNF suspension was the lowest because of its huge fiber size. In contrast, the percentage transmittance of the CLCNF suspensions was significantly higher than the LCNF suspension and the CLCNF-3 suspension exhibited the highest transmittance. This is because the light transmittance of the suspension is related to the light-scattering phenomenon, and the light scattering is proportional to the cross section area of the particles [37]. A higher transmittance means a smaller fiber size and a higher degree of fibrillation. This result corroborates what was observed in the TEM images and it is also consistent with the visual observations of the nanofiber suspensions in Figure 2B.

**Figure 2.** Light transmittance (**A**) and photograph (**B**) of LCNF and various CLCNF suspensions having a concentration of 0.1 wt%.

In order to confirm the successful cationic modification of bagasse, a Fourier transform infrared (FTIR) experiment was carried out. Figure 3 illustrates the FTIR spectra of bagasse, LCNF and various CLCNF samples. It was shown that the new peak at 1484 cm−<sup>1</sup> corresponding to the C-N stretching vibration appeared in all CLCNF samples but not in the

other samples. And the epoxy ether vibrations that belong to GTAC at 1261 cm−<sup>1</sup> were not present in all CLCNF samples, indicating the epoxide ring opening reaction of bagasse with GTAC [38]. This confirms the successful introduction of quaternary ammonium groups on bagasse. Another significant chemical structure change is the disappearance of the characteristic peak at 1733 cm−<sup>1</sup> attributed to the C=O bond after DES treatment. This is because highly alkaline conditions of DES caused the breakage of naturally occurring ester bonds in bagasse, specifically the acetyl group of hemicellulose and the ester linkage of the carboxylic groups in the ferulic and p-coumaric acids of lignin/hemicellulose [32,39]. In addition, the broadening of the hydroxyl peak around 3300 cm−<sup>1</sup> indicates that chemical modification altered the hydrogen bonding pattern of bagasse and made the bagasse more susceptible to moisture uptake from air. This will be further confirmed in the results of the thermogravimetric analysis later.

**Figure 3.** FTIR spectra of bagasse, LCNF and various CLCNF.

The specific number of quaternary ammonium groups grafted to bagasse was measured by elemental analysis (based on the weight percentages of N) and was also confirmed by polyelectrolyte titration. The results of elemental analysis and polyelectrolyte titration are listed in Table 1. The result shows that the quaternary ammonium group contents and surface charge densities were 0.97–1.76 mmol/g and 0.87–1.85 meq/g, respectively. The contents of the quaternary ammonium group in the same sample varied slightly due to the analyzing methods [40]. These results show that the increase of GTAC content in the reaction mixture improved the reactivity. The charged group content is a key performance indicator for nanofiber adsorbents in water purification applications, and it may directly affect the purification effect of contaminants [41]. CLCNF-3 was selected as the nanofiber with the highest cationic group content in this study. This was a fairly high value when compared with previous reports. In the literature, cationized CNF with an ammonium content of 0.134 mmol/g were obtained using 3-chloro-2-hydroxypropyl trimethylammonium chloride as the cationization agent [42]. Cationic CNF with a quaternary ammonium group content of 1.2 mmol/g were obtained from cellulose pulp etherified with a quaternary ammonium salt in water [15]. Cationic wood nanofiber with a cationic group content around 1.5 mmol/g were obtained using four different aqueous solvents containing TEAOH with different carbamides and GTAC as the cationization agent [32]. Surface quaternized cellulose nanofibrils with a trimethylammonium chloride content of 2.31 mmol/g were obtained from wood pulp [41]. In addition, LCNF were detected to contain trace amounts of N elements, which may be due to incomplete washing of TEAOH during centrifugal washing. Zeta-potential and surface charge densities both reflected the charge on the nanofiber surface. The zeta-potential and surface charge density were detected as negative for LCNF, but positive for CLCNF. A negative zeta-potential corresponds to a negatively charged

surface, and vice versa. The results again confirmed that the presence of GTAC successfully introduced positively charged quaternary ammonium groups to the fibers.

**Sample N (%) Cationic Group Content (mmol/g) Surface Charge Density (meq/g) Zeta-Potential (mV) Nanofibers Yield (%)** LCNF 0.08 0 −0.11 ± 0.02 −13.6 ± 0.3 8.49 CLCNF-1 1.36 0.97 0.87 ± 0.02 26.7 ± 0.9 76.49 CLCNF-2 1.49 1.06 0.97 ± 0.01 35.7 ± 0.9 88.59 CLCNF-3 2.46 1.76 1.85 ± 0.06 44.7 ± 0.8 92.50

**Table 1.** Nitrogen content, cationic group content, surface charge density, zeta-potential and nanofibers yield of LCNF and various CLCNF.

In this study, nanofibers were separated and quantified by means of centrifugation. After centrifuging the nanofiber suspensions, the larger-sized fibers will sink to the bottom and the smaller-sized fibers will remain suspended in the upper layer. The nanofiber yield refers to the percentage of small size fibers with diameters in nanoscale in the total fiber weight [43]. It showed that the nanofiber yield of LCNF was extremely low at only 8.49%. The increase in the amount of GTAC caused the value of the nanofiber yield to increase linearly to at least 76.49%, with a maximum value of 92.50%. This again confirms the results observed in the TEM images.

Zeta-potential is an important and useful parameter to describe the electric potential in the solid/liquid interfacial layer of a material in aqueous solution [44]. The zeta-potential as a function of pH for LCNF and CLCNF-3 water suspensions at consistencies of 0.1 wt% was shown in Figure 4. In this way, more information on the surface charge states of representative samples LCNF and CLCNF-3 was obtained. In the studied pH range (3–11), the zeta-potential of CLCNF-3 was always positive, while LCNF was always negative. The absolute value of the zeta-potential of LCNF became larger as the pH in water increased, with values ranging from −6.49 to −19.20 mV. This was attributed to the negatively charged surface groups (i.e., hydroxyl groups) on LCNF and the interactions of protons and sodium ions with them at different pH. The zeta-potential of CLCNF-3 fluctuates in the range of 21.17 to 44.73 mV. These values were always positively attributed to the presence of quaternary ammonium functional groups on the surface of the nanofibers. The zeta-potential of CLCNF-3 was maximum at neutral pH, and a pH that is too high or too low will decrease the zeta-potential. This is similar to the phenomenon observed for another cationized cellulose nanofiber [42]. The state of the nanofiber surface charge will most likely affect its adsorption performance for anionic contaminants in water [41,45].

CLCNF-3 is predicted to be the most outstanding adsorbent due to it containing the highest surface charge density of all samples. Therefore, the rheological properties of CLCNF-3 were investigated. The rheological properties were obtained from the viscosity measurements of CLCNF-3 suspension and are shown in Figure 5. The results show that the viscosity of the CLCNF-3 suspension decreases with increasing shear rate, suggesting the typical shear thinning behavior of the nanofiber suspension [40]. This indicates that the CLCNF-3 suspension is a pseudoplastic fluid and its diffusion in water was unimpeded. This would facilitate the capture of contaminants in the wastewater by CLCNF-3. The literature reports that some samples containing the naturally hydrophobic non-derivated lignin were difficult to disperse in water [46,47]. This would have raised concerns about the diffusivity of CLCNF-3 in water affecting water-based applications but did not occur.

**Figure 4.** Zeta-potential of LCNF and CLCNF-3 water suspensions at consistencies of 0.1 wt% at different pH values.

**Figure 5.** Viscosities of CLCNF-3 water suspensions at consistencies of 0.5 wt%.

The crystal structure and the crystallinity index (*CrI*) of bagasse, LCNF and various CLCNF was investigated by X-ray diffraction (Figure 6). All samples exhibit a sharp high peak around 2θ = 22.2◦ in XRD graphs (Figure 6A), which corresponds the (0 0 2) lattice plane of cellulose I. The results indicate that all samples contain a large amount of native cellulose I, and the cellulose I crystal structure was basically retained after modification [48]. The *CrI* of various samples are shown in Figure 6B. In general, mechanical disintegration led to the destruction of the crystalline region of cellulose and the depolymerization of cellulose, resulting in a decrease in crystallinity [38]. Chemical treatment can dissolve the amorphous fraction of cellulose [49], hemicellulose [38] and lignin [50], which in turn may lead to an increase in crystallinity. The *CrI* values of bagasse and LCNF were 50.6% and 48.8%, respectively. It may be due to the combined effect of mechanical defibrillation and chemical treatment that the *CrI* of LCNF is slightly lower than that of bagasse. On the other hand, the *CrI* values of all CLCNF samples were significantly lower compared to that of bagasse, and CLCNF-3 showed the lowest *CrI* value at 26.5%. This is mainly due to the fact that the charged quaternary ammonium groups on CLCNF made a strong electrostatic repulsion between fibers making the mechanical disintegration effect amplified. Secondly, the quaternary ammonium groups on CLCNF were detected as an amorphous region that negatively affects *CrI* [40].

**Figure 6.** X-ray diffractogram (**A**) and calculated crystallinity indices (**B**) of bagasse, LCNF and various CLCNF.

The TGA and DTG curves are presented in Figure 7, and they were used to assess the effect of DES treatment and cationic modification on the thermal degradation behavior. Bagasse exhibited a typical thermal degradation profile of lignocellulose. Its DTG curve shows two main peaks, a shoulder peak at 200~310 ◦C attributed to hemicellulose decomposition, and the most prominent peak at 310~395 ◦C, which is attributable to cellulose decomposition [51]. The TGA curves of the LCNF and bagasse almost overlap, indicating that they have similar thermal degradation behaviors. A different thermal degradation behavior was observed for CLCNF compared to bagasse. CLCNF lost more mass in the low temperature region from 30 to 160 ◦C. Bagasse lost about 8.8% of mass in water form and CLCNF lost 10.6% or more. This was attributed to the grafting of CLCNF with extremely hydrophilic quaternary ammonium groups. At the same time, the onset decomposition temperature (Tonset) and the maximum decomposition temperature (Tmax) of CLCNF were significantly lower than that of bagasse. The Tonset of bagasse was 313.8 ◦C, and the Tonset of CLCNF was reduced by 6.3 to 24.5 ◦C in comparison. The Tmax of bagasse was 356.1 ◦C, and the Tmax of CLCNF was reduced by 20.6 to 37.8 ◦C in comparison. This is attributed to the fact that the quaternary ammonium groups on CLCNF contain a volatile component (NCH2(CH3)2), making it unstable and thus susceptible to degradation [32]. Secondly, the decrease in *CrI* also negatively affected the degradation temperature [52]. Clearly, DES treatment has little effect on the pyrolysis behavior of bagasse, but cationic modification greatly reduced its thermal stability.

**Figure 7.** TGA (**A**) and DTG (**B**) curves of bagasse, LCNF and various CLCNF.
