Well-Dispersed Cellulose Nanofiber in Low Density Polyethylene Nanocomposite by Liquid-Assisted Extrusion

Two different liquid assisted processing methods: internal melt-blending (IMB) and twin-screw extrusion (TWS) were performed to fabricate polyethylene (PE)/cellulose nanofiber (CNF) nanocomposites. The nanocomposites consisted maleic anhydride-grafted PE (PEgMA) as a compatibilizer, with PE/PEgMA/CNF ratio of 97/3/0.5–5 (wt./wt./wt.), respectively. Morphological analysis exhibited that CNF was well-dispersed in nanocomposites prepared by liquid-assisted TWS. Meanwhile, a randomly oriented and agglomerated CNF was observed in the nanocomposites prepared by liquid-assisted IMB. The nanocomposites obtained from liquid-assisted TWS exhibited the best mechanical properties at 3 wt.% CNF addition with an increment in flexural strength by almost 139%, higher than that of liquid-assisted IMB. Results from this study indicated that liquid feeding of CNF assisted the homogenous dispersion of CNF in PE matrix, and the mechanical properties of the nanocomposites were affected by compounding method due to the CNF dispersion and alignment.


Introduction
Cellulose nanofiber (CNF) derived from plant receives vast attention from researchers and industries due to its superior physical properties such as high crystallinity, thermally stable, biodegradable, biocompatible and non-toxic [1][2][3][4]. Owing to its superior properties, many studies have been conducted to determine CNF potentials in various field ranging from household materials to high-tech industrial applications [5]. CNF is a promising renewable nanomaterial that can replace the use of non-renewable materials in numerous products manufacturing, mainly as a reinforcement material for the production of nanocomposite [6][7][8].

Cellulose Extraction and Nanofibrillation Process
Cellulose extraction process was carried out as described in our previous study [21], with the aim to remove lignin, hemicellulose and other extractives. The obtained cellulose fiber was then dispersed in distilled water with a concentration of 0.2% (w/v) under vigorous stirring at 60 rpm for 24 h [22]. Cellulose suspension was then passed through a high-pressure homogenizer C50 (Avestin Inc., Ottawa, Canada) for 30 passes with an operating pressure of 50 MPa to prepare a 0.2 wt.% CNF dispersed solution. The resulting CNF solution was then stored at 4 • C prior to use.

Preparation of Nanocomposites
Nanocomposites PE/PEgMA/CNF (97/3/0.5-5 wt./wt./wt.) were prepared using two different liquid assisted processing methods, which were twin screw extrusion (TWS) and internal melt-blending (IMB). The liquid-assisted TWS was conducted by a twin screw extruder (Model: IMC-160B, Imoto Machinery Co., Ltd., Kyoto, Japan) as shown in Figure 1. The liquid-assisted TWS was carried out at 50 rpm of a screw rotational speed, with a heating temperature profile of 80, 160, 160 and 160 • C, respectively for four zones of cylinder from hopper to die. Prescribed amounts of 0.2% (w/v) CNF dispersed solution, PEgMA as a dispersant, and PE as matrix were fed into the extruder cylinder and the mixture was melt-kneaded by shearing stress with proper rotation of screw, resulting in extrusion of PE/CNF nanocomposite as a strand from the die. The strand was cut to a length of approximately 10 cm and was then used for film preparation.

Cellulose Extraction and Nanofibrillation Process
Cellulose extraction process was carried out as described in our previous study [21], with the aim to remove lignin, hemicellulose and other extractives. The obtained cellulose fiber was then dispersed in distilled water with a concentration of 0.2% (w/v) under vigorous stirring at 60 rpm for 24 h [22]. Cellulose suspension was then passed through a high-pressure homogenizer C50 (Avestin Inc., Ottawa, Canada) for 30 passes with an operating pressure of 50 MPa to prepare a 0.2 wt.% CNF dispersed solution. The resulting CNF solution was then stored at 4 °C prior to use.

Preparation of Nanocomposites
Nanocomposites PE/PEgMA/CNF (97/3/0.5-5 wt./wt./wt.) were prepared using two different liquid assisted processing methods, which were twin screw extrusion (TWS) and internal meltblending (IMB). The liquid-assisted TWS was conducted by a twin screw extruder (Model: IMC-160B, Imoto Machinery Co., Ltd., Kyoto, Japan) as shown in Figure 1. The liquid-assisted TWS was carried out at 50 rpm of a screw rotational speed, with a heating temperature profile of 80, 160, 160 and 160 °C, respectively for four zones of cylinder from hopper to die. Prescribed amounts of 0.2% (w/v) CNF dispersed solution, PEgMA as a dispersant, and PE as matrix were fed into the extruder cylinder and the mixture was melt-kneaded by shearing stress with proper rotation of screw, resulting in extrusion of PE/CNF nanocomposite as a strand from the die. The strand was cut to a length of approximately 10 cm and was then used for film preparation. Liquid-assisted IMB process was conducted using a Brabender plastograph EC internal mixer (Brabender GmbH & Co. KG, Duisburg, Germany) with a mixing speed of 50 rpm for 20 min at 160 °C as shown in Figure 2. Prescribed amounts of PE, PEgMA and CNF dispersed solution were meltblended and kneaded for 20 min. Liquid-assisted IMB process was conducted using a Brabender plastograph EC internal mixer (Brabender GmbH & Co. KG, Duisburg, Germany) with a mixing speed of 50 rpm for 20 min at 160 • C as shown in Figure 2. Prescribed amounts of PE, PEgMA and CNF dispersed solution were melt-blended and kneaded for 20 min. The PE/PEgMA/CNF composites obtained from the two processing methods were placed on a molding die as described by Miyake and Imaeda [23]. For the composite prepared by twin screw extrusion, the sample was arranged in one direction, while the composite prepared by internal meltblending was arranged randomly on the mold, and both samples were hot-pressed at 160 °C under 10 atm for 5 min to obtain film specimens for following analyses (modified from Then et al. [15]). All composites were prepared in prescribed weight ratios of LDPE, PEgMA and CNF (97/3/0.5, 1, 2, 3, 4 and 5 (wt./wt./wt.)) as shown in Table 1. CNF dispersion in the nanocomposites was analyzed by examining the distribution of oxygen element using a scanning electron microscope equipped with energy dispersive spectroscopy (SEM-EDS) (JCM 6000, JEOL Ltd., Tokyo, Japan). Nanocomposite samples were firstly coated with platinum using a vacuum sputter coater prior to external morphology observation by SEM-EDS. While for internal morphology scanning of nanocomposites, an X-ray computed tomography device (X-CT) (Xradia 410 Versa, Zeiss, Oberkochen, Germany) was used at 20 kV with tube current 200 µA, scan time 10 s and scanning repetition of 1600 times exposure. Obtained X-CT images were then reconstructed using VGStudio MAX 2.2 (Volume Graphics GmbH, Heidelberg, Germany) to observe a cross-sectional image of nanocomposite samples. The PE/PEgMA/CNF composites obtained from the two processing methods were placed on a molding die as described by Miyake and Imaeda [23]. For the composite prepared by twin screw extrusion, the sample was arranged in one direction, while the composite prepared by internal melt-blending was arranged randomly on the mold, and both samples were hot-pressed at 160 • C under 10 atm for 5 min to obtain film specimens for following analyses (modified from Then et al. [15]). All composites were prepared in prescribed weight ratios of LDPE, PEgMA and CNF (97/3/0.5, 1, 2, 3, 4 and 5 (wt./wt./wt.)) as shown in Table 1.

Morphological Analysis
CNF dispersion in the nanocomposites was analyzed by examining the distribution of oxygen element using a scanning electron microscope equipped with energy dispersive spectroscopy (SEM-EDS) (JCM 6000, JEOL Ltd., Tokyo, Japan). Nanocomposite samples were firstly coated with platinum using a vacuum sputter coater prior to external morphology observation by SEM-EDS. While for internal morphology scanning of nanocomposites, an X-ray computed tomography device (X-CT) (Xradia 410 Versa, Zeiss, Oberkochen, Germany) was used at 20 kV with tube current 200 µA, scan time 10 s and scanning repetition of 1600 times exposure. Obtained X-CT images were then reconstructed using VGStudio MAX 2.2 (Volume Graphics GmbH, Heidelberg, Germany) to observe a cross-sectional image of nanocomposite samples.

X-ray Diffraction Analysis
Crystallinity index (CrI) of nanocomposites was analyzed from X-ray diffraction (XRD) profiles measured with an X-ray powder diffractometer (Rigaku Corporation, Tokyo, Japan) equipped with nickel filtered Cu Kα radiation (λ = 0.1542 nm). The X-ray pattern was recorded in a 2θ range of 5-50 • at 40 kV and 25 mA as well as at a scan rate of 2 • min −1 [20,21]. The CrI was calculated based on Equation (1) where, I 002 was diffraction peak intensity from (002) plain at 2θ = 23 • , which corresponds to a specific diffraction peak of the crystalline portion. I am refers to the intensity at about 2θ = 18 • , and it refers to the amorphous region [21].

Thermal Analysis
Thermal analysis of nanocomposites was conducted using thermogravimetry (TG) and differential scanning calorimetry (DSC) analyzers to identify decomposition and melting temperatures. Samples for both analyses were weighted in a range of 10-12 mg, and each sample was placed in an aluminum pan. TG analysis was performed using a thermogravimetric analyzer EXSTAR TG/DTA 6200 (Seiko Instruments Inc., Chiba, Japan) in a temperature range of 50 to 550 • C at a heating rate of 10 • C/min under a nitrogen flow of 20 mL/min [21]. DSC analysis was performed on an EXSTAR DSC 6220 (Seiko Instruments Inc., Chiba, Japan). In a first scan, the samples were heated from 30 to 200 • C at a rate of 10 • C/min and held at 200 • C for 1 min. They were then cooled to 50 • C at a rate of 10 • C/min and held at 50 • C for 1 min. Next for the second scan, they were again heated to 200 • C at a rate of 10 • C/min and held at 200 • C for 1 min. The thermal properties of samples were determined based on the data obtained during the second heating/cooling scan. Melting temperature (T m ) was taken as the highest endothermic peak in the heating scan and cold crystallization temperature (T c ) was determined based on the exothermal peak during the cooling scan [21].

Mechanical Analysis
Tensile and flexural tests of the nanocomposite specimens were conducted using a compact tensile and compression tester IMC-18E0 (Imoto Machinery Co., Ltd., Kyoto, Japan). The tensile properties consisted of tensile strength, Young's modulus, toughness and elongation at break were evaluated from stress-strain curves measured at a crosshead speed of 5 mm/min [24]. The flexural test was performed according to ASTM D790 on rectangular standard samples with dimension size of 120 mm × 12.7 mm and 3 mm thickness. Flexural tests, flexural strength and modulus were determined from stress-strain curves, which were measured at 10 mm/min [25]. The reported tensile and flexural properties values were averaged values of five measurements for each sample.

Statistical Analysis
The mechanical properties of each nanocomposite were statistically analyzed using a one-way ANOVA in a general linear model using Statistical Analysis Software (SAS) Ver. 9.4. The mean for each measured parameter was separated statistically using Duncan Multiple Range Test (DMRT) at p < 0.05. From the micrographs in Figure 3b-g for nanocomposites prepared by liquid-assisted IMB, it was observed that PE/PEgMA/CNF nanocomposites with different CNF compositions exhibited different CNF dispersion states. As indicated in Figure 3b,c, the fibers distribution in the composites with 0.5 wt.% and 1 wt.% CNF prepared by liquid-assisted IMB seems to be homogenous. EDS images also showed the homogenous dispersion of white spots indicating the presence of oxygen, derived from CNF and PEgMA. The homogenous dispersion of CNF suggests a good wetting behavior between the CNF surface and PE/PEgMA matrix.

Morphological Analysis of Nanocomposite
It can be observed that CNF was agglomerated, and homogenous dispersion was failed to be achieved when more than 1 wt.% CNF was incorporated in the PE matrix, as presented in Figure 3d-g. Interaction between CNFs happens due to surface forces to produce agglomerates, which are difficult to be re-nanofibrillated during the liquid-assisted IMB fabrication since the added shearing stress was not enough to break the agglomerates. This is supported by the results of Cruz and Viana [26] that the input process during melt-blending couldn't break all the aggregate fillers, thus resulting in poor fillers dispersion. Herrera et al. [10] also reported that a high amount of CNF suspension may result in boiling and lead to fast evaporation of water, which consequently creating agglomeration of the CNF.
As for PE/PEgMA/CNF nanocomposites prepared by the liquid-assisted TWS, the homogenous dispersion of CNF was observed in the nanocomposites reinforced with 0.5, 1, 2 and 3 wt.% CNF as shown in Figure 3h-k. No agglomeration of CNF was detected from their EDS images. This proves that CNF can be dispersed homogeneously by the liquid-assisted TWS. Heterogeneous dispersion, however, was observed in the nanocomposites reinforced with 4 and 5 wt.% CNF as shown in Figure 3l,m, in which the SEM-EDS images showed unevenly distributed white spots, supporting the aggregation of CNF. Based on these results, the homogenous dispersion of higher amount of CNF was achieved by the liquid-assisted TWS, since the shearing stress during the processing must be enough to suppress and/or re-nanofibrillate the CNF aggregation. The presence of the second vent site at the end of the extruder also helps in preventing fast evaporation during composite processing and this reduces the potential of CNF agglomeration in the PE matrix.
Based on the overall morphological analysis, it was found that liquid feeding favored the CNF dispersion and distribution, as agglomeration was only observed in the PE/PEgMA nanocomposites consisted of 4-5% CNF for liquid-assisted TWS, and 2-5% CNF for liquid-assisted IMB. A similar report was also reported by Lo Re et al. [27]. They discovered that the liquid feeding can prevent agglomeration of fibers, as this method can prevent the formation of strong fiber-fiber interactions. Based on their findings, the fibers were found to be well-distributed in the PCL matrix for composites prepared by liquid feeding, whereas for composites prepared by dry feeding method, the fibers were found to be agglomerated and poorly distributed in the PCL matrix. Additionally, this method can also prevent the initial agglomeration caused by drying prior to compounding.

Mechanical Properties of Nanocomposite
The increment in mechanical properties was observed when CNF was incorporated in the PE matrix by both compounding methods as shown in Table 2. It was discovered that PE/PEgMA/CNF nanocomposites prepared by liquid-assisted TWS exhibited higher tensile strength, Young's modulus, flexural strength and flexural modulus compared to PE/PEgMA/CNF nanocomposites prepared by liquid-assisted IMB. This outcome was highly related to compounding methods. Thus, this suggests that the existential state of CNF in the polymer matrix significantly reflected in the mechanical properties. For nanocomposites prepared by liquid-assisted IMB method, the maximum mechanical increment was recorded when 1 wt.% CNF was incorporated in the PE matrix, as the values of tensile strength, Young's modulus, flexural strength, and flexural modulus increased to 10.5, 145, 11.3 and 230.1 MPa, respectively. When the CNF composition exceeded 1 wt.%, the mechanical properties decreased along with the increase in CNF composition. Nevertheless, at 2 and 3 wt.% of CNF, the values were higher than those of PE/PEgMA as reference. A similar trend was also recorded for the nanocomposites prepared by liquid-assisted TWS processing. Increment in mechanical properties was observed along with the increase in CNF composition from 0.5 to 3 wt.%, reaching the highest mechanical properties at 12.1, 166, 16.5 and 273 MPa for tensile strength, Young's modulus, flexural strength, and flexural modulus, respectively. It was also noticeable that by reinforcing with 3 wt.% CNF by liquid-assisted TWS, the mechanical properties of nanocomposites were considerably improved by more than 50% higher than PE/PEgMA.
In general, mechanical properties of composite increases by increasing the fiber loading in the matrix [28,29]. However, monotonous reduction in the overall mechanical properties was recorded when CNF composition was over 1 and 3 wt.% for liquid-assisted IMB and TWS, respectively. According to Mittal et al. [30], the reduction in mechanical properties occur after reaching a threshold limit and this is mainly due to poor mechanical interlocking, which may degrade load transfer between fibers and matrix. In addition to fiber loading, it could be seen from this study that the reduction in mechanical properties was also highly related to the agglomeration of CNF in the PE matrix during processing. As observed in the SEM-EDS images, agglomeration was observed when CNF composition was more than 1 and 3 wt.% for liquid-assisted IMB and TWS, respectively (Figure 3). This explains the reduction in mechanical properties. CNF agglomeration may also reduce the formation of a percolating network of CNF within the polymer matrix, hence reduce the mechanical properties of nanocomposites [7,19,31].
Findings from this study also confirmed that liquid feeding promoted better dispersion of CNF in the polymer matrix which contributed to improved mechanical properties of PE/PEgMA/CNF nanocomposites. These findings are in accordance with the findings from other studies as tabulated in Table 3. According to Lo Re [27], liquid feeding provided milder shear stress during compounding, thus the CNF length and L/D ratio can be preserved and indirectly resulting in better mechanical properties of composites. Homogenous dispersion of CNF observed in SEM-EDS images is due to the usage of PEgMA as a compatibilizer, resulting in effective adherence between CNF surface and matrix PE. The good dispersion of CNF and wetting situation at the interface may also generate the desired stress transfer between PE/PEgMA matrix and CNF throughout the stress loading, resulting in excellent tensile and flexural strength of the nanocomposite. This has been reported by Essabir et al. [35], in which good compatibility between CNF and polymer matrix resulted in better adhesion and provided better stress transfer. Due to the reactive interaction of hydrogen bonding between hydroxyl groups on CNF surface and MA moieties in PEgMA, CNF is able to be bound with PEgMA [35]. PEgMA is also compatible with the PE matrix due to the internal hydrophobic PE sequences, resulting in enhancing adhesion between the CNF surface and PE matrix.
It is interesting to note that nanocomposites prepared by liquid-assisted TWS recorded higher mechanical properties than nanocomposites prepared by liquid-assisted IMB. The use of liquid-assisted TWS allowed higher concentration of CNF to be dispersed homogeneously in the nanocomposite, which affected mainly the flexural properties of nanocomposites. For instance, flexural strength and modulus of the nanocomposite reinforced by 3% CNF showed the maximum increment of 139% and 195% compared to corresponding nanocomposites prepared by liquid-assisted IMB and PE/PEgMA as a reference, respectively. There would be two main reasons, which led to the better mechanical properties of the nanocomposite prepared by TWS: (i) homogeneous dispersion of CNF avoiding agglomeration, and (ii) aligned orientation of CNF in the nanocomposite [36]. To confirm the orientation of CNF in the PE matrix, an X-CT scan of nanocomposites was conducted, and the image is as shown in Figure 4. Although finer CNF portions less than 1 µm were difficult to detect by the X-CT analysis, the observed morphology of the pre-nanofibrillated portions suggests the same morphology of the finer portions.
Polymers 2020, 11, x FOR PEER REVIEW 11 of 16 CNF surface and MA moieties in PEgMA, CNF is able to be bound with PEgMA [35]. PEgMA is also compatible with the PE matrix due to the internal hydrophobic PE sequences, resulting in enhancing adhesion between the CNF surface and PE matrix. It is interesting to note that nanocomposites prepared by liquid-assisted TWS recorded higher mechanical properties than nanocomposites prepared by liquid-assisted IMB. The use of liquidassisted TWS allowed higher concentration of CNF to be dispersed homogeneously in the nanocomposite, which affected mainly the flexural properties of nanocomposites. For instance, flexural strength and modulus of the nanocomposite reinforced by 3% CNF showed the maximum increment of 139% and 195% compared to corresponding nanocomposites prepared by liquidassisted IMB and PE/PEgMA as a reference, respectively. There would be two main reasons, which led to the better mechanical properties of the nanocomposite prepared by TWS: (i) homogeneous dispersion of CNF avoiding agglomeration, and (ii) aligned orientation of CNF in the nanocomposite [36]. To confirm the orientation of CNF in the PE matrix, an X-CT scan of nanocomposites was conducted, and the image is as shown in Figure 4. Although finer CNF portions less than 1 µm were difficult to detect by the X-CT analysis, the observed morphology of the pre-nanofibrillated portions suggests the same morphology of the finer portions. In reports by Maniruzzaman et al. [37]; Cruz and Viana [26]; Hietala et al. [38]; Salleh et al. [39]; Khanam and AlMaadeed [40], it has been well-understood that the alignment of large size (microscale onwards) natural fibers in biocomposites is very much dependent on the melt-processing method. Nevertheless, the effect of processing methods on the alignment of fibers in the polymer matrix has not been studied for nano-scale fibers. Based on the X-CT analysis results in Figure 4, the effect of liquid-assisted TWS processing on CNF alignment is similar to the previous findings using microfiber, in which the liquid-assisted TWS processing has caused the fiber to be well-aligned to the same direction. In the case of liquid-assisted IMB, random alignment and dispersion were observed. Agglomeration of CNF was also occurred, and these have contributed to the lower mechanical In reports by Maniruzzaman et al. [37]; Cruz and Viana [26]; Hietala et al. [38]; Salleh et al. [39]; Khanam and AlMaadeed [40], it has been well-understood that the alignment of large size (micro-scale onwards) natural fibers in biocomposites is very much dependent on the melt-processing method. Nevertheless, the effect of processing methods on the alignment of fibers in the polymer matrix has not been studied for nano-scale fibers. Based on the X-CT analysis results in Figure 4, the effect of liquid-assisted TWS processing on CNF alignment is similar to the previous findings using microfiber, in which the liquid-assisted TWS processing has caused the fiber to be well-aligned to the same direction. In the case of liquid-assisted IMB, random alignment and dispersion were observed. Agglomeration of CNF was also occurred, and these have contributed to the lower mechanical properties of the IMB nanocomposites.
The main reason which contributed to the differences between TWS and IMB is the processing mechanism. According to Drobny [41], IMB involves kneading for the preparation of non-continuous homogenous polymer blend. It does not involve pressurized system and extrusion, therefore it contributes to the non-uniform alignment of CNF in the polymer matrix. Due to this, CNF tends to get entangled with each other when they are subjected to the liquid-assisted IMB. Meanwhile, TWS consists of two rotating screws that can continuously melt mixing and dispersive mixing both of the CNF and polymer at high shear force. It is operated in a pressurized system with dies that extrude the nanocomposite strands. The presence of flight around the TWS screw also helps in aligning the CNF. This has caused liquid-assisted TWS process to provide a continuous homogenous polymer blend.

Crystallinity and Thermal Properties of Nanocomposite
The crystallinity index (CrI) of PE/PEgMA/CNF nanocomposites and PE/PEgMA are summarized in Table 4. It was observed that all PE/PEgMA/CNF nanocomposites recorded higher CrI values than 37.2% for PE/PEgMA as reference. For nanocomposites prepared by liquid-assisted IMB, the CrI value was approximately between 38.0-43.9%, while for nanocomposites prepared by liquid-assisted TWS, the CrI value was evaluated to be around 38.1-44.4%.
These results indicate that the incorporated CNF essentially increased CrI of the PE matrix. According to Shalwan and Yousif [42] and Essabir et al. [43], the CrI of composite changed by self-nucleation effect (homogenous nucleation) and/or by a nucleating agent (heterogeneous nucleation). Hence, in this case, it is considered that CNF acted as a nucleating agent and produced seed crystals, resulting in the promotion of secondary crystallization and finally increased the CrI value of composites [40].
The CrI results were also in accordance with thermal properties measured by TG and DSC as listed in Table 2. The T d20% of the nanocomposites prepared by liquid-assisted IMB was in a range of 444-450 • C, while the T d50% was in a range of 464-467 • C, which were lower than 443-459 and 464-478 • C for T d20% and T d50% of the nanocomposites prepared by TWS, respectively, as shown in Figure 5. It was also noticeable that the highest T d50% and T d20% values were observed on 0.5% and 3% CNF of PE/PEgMA/CNF nanocomposites for liquid-assisted IMB and TWS processing methods, respectively. Similar result of T max was also observed for the nanocomposites prepared by these two different methods. In the case of nanocomposites prepared by liquid-assisted TWS, the T max was around 473-476 • C, which was higher than 472 • C of PE/PEgMA, whereas the nanocomposites prepared by liquid-assisted IMB recorded much lower range, which was around 468-471 • C.  In addition to mechanical properties, the highest thermal properties of PE/PEgMA/3% CNF nanocomposite prepared by liquid-assisted TWS processing is also considered to be attributed by homogenous dispersion and less agglomeration of CNF in PE matrix (Figure 3), because the homogeneously dispersed CNF restrained the thermal mobility of PE sequences. On the other hand, agglomerated CNF reduced the thermal properties of composites because of its poorer thermal stability (Td50% = 353 °C) than PE.
The crystallization temperature (Tc) and melting temperature (Tm) were determined by DSC measurement. It was revealed that both Tc and Tm of PE/PEgMA/CNF nanocomposites showed nearly the same values with the neat PE/PEgMA. This is attributed by the low CNF loading, in which the amount was insufficient to give effects on the melting behavior of PE/PEgMA composites. Similar to Tm, the increment of Tc for all PE/PEgMA/CNF nanocomposite is quite low since PE is an elastomeric copolymer [44]. A similar finding has also been reported by Maia et al. [44] that incorporates CNF in the PE matrix and Pollanen et al. [45] that incorporates microcrystalline cellulose (MCC) in the HDPE matrix. Maia et al. [44] revealed that there was no significant differences of Tm and Tc between neat PE and PE composites incorporated with CNF in a range of 1, 3, 5, 10, 20 and 30 wt.%, but such differences can be seen in PE incorporated with 90 wt.% CNF composites. In the case of Pollanen et al. [45], the value of Tm and Tc remained significant even after 5, 10, 20 and 40 wt.% MCC were added in the HDPE.
By comparing the nanocomposites prepared by different compounding method, the Tc values of PE/PEgMA/CNF nanocomposites prepared by liquid-assisted IMB were slightly higher than the corresponding nanocomposites prepared by liquid-assisted TWS. This suggests that the agglomerates of CNF contributed to the crystallization of PE. On the other hand, the Tm and CrI values increased with increase in CNF content, and both values of PE/PEgMA/CNF nanocomposites prepared by liquid-assisted TWS were higher than corresponding nanocomposites prepared by liquid-assisted IMB, suggesting that the more the CNF surface area, the higher the Tm and CrI values showed. Thus, it was recognized that the compounding method of CNF had also clear effects on the thermal properties of the nanocomposites. In addition to mechanical properties, the highest thermal properties of PE/PEgMA/3% CNF nanocomposite prepared by liquid-assisted TWS processing is also considered to be attributed by homogenous dispersion and less agglomeration of CNF in PE matrix (Figure 3), because the homogeneously dispersed CNF restrained the thermal mobility of PE sequences. On the other hand, agglomerated CNF reduced the thermal properties of composites because of its poorer thermal stability (T d50% = 353 • C) than PE.

Conclusions
The crystallization temperature (T c ) and melting temperature (T m ) were determined by DSC measurement. It was revealed that both T c and T m of PE/PEgMA/CNF nanocomposites showed nearly the same values with the neat PE/PEgMA. This is attributed by the low CNF loading, in which the amount was insufficient to give effects on the melting behavior of PE/PEgMA composites. Similar to T m , the increment of T c for all PE/PEgMA/CNF nanocomposite is quite low since PE is an elastomeric copolymer [44]. A similar finding has also been reported by Maia et al. [44] that incorporates CNF in the PE matrix and Pollanen et al. [45] that incorporates microcrystalline cellulose (MCC) in the HDPE matrix. Maia et al. [44] revealed that there was no significant differences of T m and T c between neat PE and PE composites incorporated with CNF in a range of 1, 3, 5, 10, 20 and 30 wt.%, but such differences can be seen in PE incorporated with 90 wt.% CNF composites. In the case of Pollanen et al. [45], the value of T m and T c remained significant even after 5, 10, 20 and 40 wt.% MCC were added in the HDPE.
By comparing the nanocomposites prepared by different compounding method, the T c values of PE/PEgMA/CNF nanocomposites prepared by liquid-assisted IMB were slightly higher than the corresponding nanocomposites prepared by liquid-assisted TWS. This suggests that the agglomerates of CNF contributed to the crystallization of PE. On the other hand, the T m and CrI values increased with increase in CNF content, and both values of PE/PEgMA/CNF nanocomposites prepared by liquid-assisted TWS were higher than corresponding nanocomposites prepared by liquid-assisted IMB, suggesting that the more the CNF surface area, the higher the T m and CrI values showed. Thus, it was recognized that the compounding method of CNF had also clear effects on the thermal properties of the nanocomposites.

Conclusions
This study demonstrated that CNF can be incorporated in polyolefin such as PE for reinforcement material by liquid feeding of CNF. Results obtained from this study revealed that liquid-assisted TWS contributed to better CNF orientation and dispersion as compared to liquid-assisted IMB. This contributed to significant increment of mechanical properties, up to 195% for flexural strength at CNF composition of 3 wt.%. It was also observed that random and heterogeneous CNF dispersion in nanocomposites led to lower thermal and mechanical properties.
As a conclusion, liquid-assisted TWS is an excellent approach to ensure homogenous dispersion of CNF in the polymer matrix which affected positively on the mechanical, crystallinity and thermal properties of nanocomposites.

Conflicts of Interest:
The authors declare no conflict of interest.