Contribution of the Surface Treatment of Nanofibrillated Cellulose on the Properties of Bio-Based Epoxy Nanocomposites Intended for Flexible Electronics

The growing interest in materials derived from biomass has generated a multitude of solutions for the development of new sustainable materials with low environmental impact. We report here, for the first time, a strategy to obtain bio-based nanocomposites from epoxidized linseed oil (ELO), itaconic acid (IA), and surface-treated nanofibrillated cellulose (NC). The effect of nanofibrillated cellulose functionalized with silane (NC/S) and then grafted with methacrylic acid (NC/SM) on the properties of the resulted bio-based epoxy systems was thoroughly investigated. The differential scanning calorimetry (DSC) results showed that the addition of NCs did not influence the curing process and had a slight impact on the maximum peak temperature. Moreover, the NCs improved the onset degradation temperature of the epoxy-based nanocomposites by more than 30 °C, regardless of their treatment. The most important effect on the mechanical properties of bio-based epoxy nanocomposites, i.e., an increase in the storage modulus by more than 60% at room temperature was observed in the case of NC/SM addition. Therefore, NC’s treatment with silane and methacrylic acid improved the epoxy–nanofiber interface and led to a very good dispersion of the NC/SM in the epoxy network, as observed by the SEM investigation. The dielectric results proved the suitability of the obtained bio-based epoxy/NCs materials as substitutes for petroleum-based thermosets in the fabrication of flexible electronic devices.


Introduction
Epoxy resins are one of the most widely used thermosetting polymers, which become irreversibly hardened when cured by heating and/or by the addition of crosslinking agents [1,2]. Today, epoxy resins account for approximately 70% of the thermosets market, excluding the polyurethane thermosets [3]. Endowed with high tensile and compression strengths, excellent adhesiveness to a variety of substrates, good electrical insulation properties, high corrosion and chemical resistances, and superior thermal stability, epoxy resins have found multiple uses in the coatings and adhesives industries, electronics, civil engineering, and as matrices for high-performance composites [4]. Unfortunately, nowadays, 90% of the epoxy resins produced around the world are still based on diglycidyl ether of bisphenol A (DGEBA), which results from the polycondensation of bisphenol epoxy groups from the ELO's structure. The dynamic mechanical results (DMA) revealed that the thermosets based on ELO/IA and ESO/IA possessed higher flexibility as compared to the system based on DGEBA/IA, exhibiting much lower T g values of 37 • C and 13 • C as compared to the 63 • C recorded for the DGEBA/IA thermoset. The higher T g value obtained for the DGEBA/IA system was attributed to the rigid aromatic rings from the DGEBA's structure and its higher crosslinking density [20].
The room for improvement in the properties of epoxy resins can be filled with nanocellulose fibers. Cellulose, a polysaccharide with straight chains of D-glucose connected by β-1,4-glycosidic linkages, represents the most abundant terrestrial biopolymers in the world and is a source of nanocellulose [24,25]. Due to its versatility regarding the geometry, from (nano)fibers to (nano)particles, the multitude of isolation sources, and superior properties in terms of renewability, biocompatibility, biodegradability and lowcost, cellulose has been an attractive material choice for developing new bio-based products for a variety of applications [26,27]. As a consequence of its nanodimensions, nanocellulose (NC) possesses a high surface area, which may allow for the formation of stronger interactions with the polymer matrix. The valuable properties of nanocellulose, such as its all-around availability, renewability, biodegradability, and good mechanical properties, come to strengthen those of the epoxy resins. NC has been intensively studied as a reinforcing agent in bio-based polymers such as poly(lactic acid) [28,29], polyhydroxyalkanoates [30,31], poly(butylene succinate) [32], and starch [33,34], leading to fully-biobased materials with improved mechanical properties and, in some cases, with better thermal properties. However, a smaller number of studies were dedicated to the reinforcement of common epoxy resins with cellulose nanocrystals or nanofibers [17,[35][36][37][38]. One problem in the case of nanocomposites based on hydrophobic polymer matrices and hydrophilic fillers is the weak interaction at the interface that leads to the poor dispersion of the nanofiller [37]. This drawback can be corrected via the chemical modification of NC using the multitude of reactive hydroxyl groups from its surface [37,39]. In particular, cellulose nanofibers extracted from filter-paper were grafted with polyethyleneimine (PEI) to improve the interfacial adhesion between NC and the bisphenol A type epoxy resin [37]. The increased interfacial crosslinking density in these nanocomposites was reflected by the strong increase in tensile strength and modulus. A similar effect was observed for a bisphenol A type epoxy resin reinforced with cellulose nanocrystals (CNCs) grafted with lauric acid [39]. Good mechanical properties were obtained for the nanocomposites with slightly modified CNCs (degrees of substitution of 0.2 and 0.8). These nanocomposites also exhibited a slight change in flexibility [39]. In another attempt, microfibrillated cellulose was functionalized with triethoxy(3-glycidyloxypropyl)silane (GPS) and used as a reinforcing agent in a bisphenol A type epoxy resin [40]. A strong increase in the impact strength was obtained for the epoxy resin containing GPS-treated NC as fillers as compared to the one recorded for the epoxy matrix containing untreated NC. GPS-treated NC also led to epoxy nanocomposites, showing better tensile strength and modulus [40]. Triblock copolymers surfactants containing poly(ethylene oxide) and poly(propylene oxide) blocks were also assessed for strengthening the NC-bisphenol A epoxy interface [41].
Only a few works deal with bio-based epoxy resins reinforced with NC or cellulose fibers [42][43][44]. Shibata et al. prepared bio-based epoxy resins from ESO by using TA as a curing agent [42]. The composites obtained from these bio-based epoxy systems and 5-11 wt% cellulose fibers (CF) with submicron width showed improved mechanical properties; an increase in tensile strength by 50% and a 3-fold increase in modulus along with a drastic reduction in the elongation at break were observed for the composite with 11 wt% CF [42]. Weaker improvements in the mechanical properties were reported for bio-based epoxy composites containing similar concentration of CF but based on sorbitol glycidyl ether and TA [43]. A terpene-maleic ester type epoxy resin reinforced with 0.5-8 wt% cellulose nanowhiskers (CNW) showed improved mechanical properties [44]. The addition of only 1 wt% CNW led to more than 2-fold increase in the Young's modulus and a 50% increase in the tensile strength of the terpene-maleic ester epoxy resin. However, the elongation at break of the same composite suffered a 12-fold reduction compared to the unmodified epoxy [44], which characterizes a very stiff material. Therefore, in this study, an epoxidized vegetable oil and a bio-based crosslinker were used to obtain the epoxy resin. Epoxy resins were obtained from ELO and IA using small amounts of grafted and ungrafted NCs for balancing the stiffness and flexibility of these systems. For improving the interface properties, NC was surface-modified firstly by treatment with a silane and then by grafting polymerization with methacrylic acid. To the best of our knowledge, bio-based epoxy nanocomposites consisting of ELO resin, IA as a cross-linking agent, and NCs as filler have not yet been reported. The structural, morphological, thermo-mechanical, and electrical properties of the resulting nanocomposites were determined for assessing the performances of the new fully-bio-based nanomaterials. The epoxy nanocomposites with balanced mechanical properties obtained in this study are designed for engineering applications, and especially for flexible electronic devices.

Results and Discussion
The schematic representations of the surface modification of NC and the preparation route of the cured epoxy nanocomposites are shown in Figure 1a,b. epoxy composites containing similar concentration of CF but based on sorbitol glycidyl ether and TA [43]. A terpene-maleic ester type epoxy resin reinforced with 0.5-8 wt% cellulose nanowhiskers (CNW) showed improved mechanical properties [44]. The addition of only 1 wt% CNW led to more than 2-fold increase in the Young's modulus and a 50% increase in the tensile strength of the terpene-maleic ester epoxy resin. However, the elongation at break of the same composite suffered a 12-fold reduction compared to the unmodified epoxy [44], which characterizes a very stiff material. Therefore, in this study, an epoxidized vegetable oil and a bio-based crosslinker were used to obtain the epoxy resin. Epoxy resins were obtained from ELO and IA using small amounts of grafted and ungrafted NCs for balancing the stiffness and flexibility of these systems. For improving the interface properties, NC was surface-modified firstly by treatment with a silane and then by grafting polymerization with methacrylic acid. To the best of our knowledge, bio-based epoxy nanocomposites consisting of ELO resin, IA as a cross-linking agent, and NCs as filler have not yet been reported. The structural, morphological, thermo-mechanical, and electrical properties of the resulting nanocomposites were determined for assessing the performances of the new fully-bio-based nanomaterials. The epoxy nanocomposites with balanced mechanical properties obtained in this study are designed for engineering applications, and especially for flexible electronic devices.

Results and Discussion
The schematic representations of the surface modification of NC and the preparation route of the cured epoxy nanocomposites are shown in Figure 1a

FT-IR Spectra of Pristine Materials and Epoxy/NC Nanocomposites
FT-IR was used to investigate the changes induced by the addition of differently treated NCs on the structural properties of cured epoxy nanocomposites ( Figure 2). The   [20,45]. The bands at 1700 and 1713 cm −1 are related to the C=O stretching, and that at 1630 cm −1 to the C=C stretching vibrations from IA's chemical structure [46]. condensation, and hydrolysis are all possible [48]. Thus, several changes were detected in the close-up spectra of the thermally cured neat epoxy system (E) and epoxy nanocomposites (Figure 2c,d).
The bands assigned to the stretching vibrations of OH groups from 3500 cm −1 increased in intensity in the cured samples, indicating that the carboxyl groups of itaconic acid determined the opening of the epoxy cycles with the formation of new OH groups [49]. For comparing these signals, the intensity of the band was normalized, using the C-H stretching vibrations at 2830-2900 cm −1 as an internal reference. A new band at 1719 cm −1 appeared after curing, which could be assigned to a newly formed ester C=O bond ( Figure 2c) [20,22].  The spectrum of neat ELO shows absorption bands at around 3500 cm −1 , which correspond to the stretching vibrations of the O-H groups, a band assigned to the carbonyl (C=O) stretching from the ester functionality (1736 cm −1 ), bands associated with the -CH asymmetric and symmetric bending vibrations from the CH 2 and CH 3 groups (1464 and 1385 cm −1 ), and the oxirane C-O twin bands at 823 cm −1 and 844 cm −1 [20,47] (Figure 2).
According to literature, the reactions between the epoxy groups from resins and the carboxyl groups of dicarboxylic acids are quite complex, and esterification, etherification, condensation, and hydrolysis are all possible [48]. Thus, several changes were detected in the close-up spectra of the thermally cured neat epoxy system (E) and epoxy nanocomposites (Figure 2c,d).
The bands assigned to the stretching vibrations of OH groups from 3500 cm −1 increased in intensity in the cured samples, indicating that the carboxyl groups of itaconic acid determined the opening of the epoxy cycles with the formation of new OH groups [49]. For comparing these signals, the intensity of the band was normalized, using the C-H stretching vibrations at 2830-2900 cm −1 as an internal reference. A new band at 1719 cm −1 appeared after curing, which could be assigned to a newly formed ester C=O bond (Figure 2c) [20,22].
Moreover, the disappearance of the bands attributed to the specific signals of oxirane groups (823, 844, and 629 cm −1 ) from the cured epoxy matrix and epoxy nanocomposites' spectra are a clear demonstration of the formation of ELO-based 3D networks following the crosslinking with IA ( Figure 2d). Interestingly, the band at 1100 cm −1 assigned to the C-O stretching vibrations of O=C-O from ELO splits into two new bands (1092, 1067 cm −1 ) after the curing reaction as a consequence of the formation of new types of covalent C-O-C bonds. The newly emerged bands are typical of the ether C-O-C bands formed in accordance with the epoxy-acid curing mechanism [47]. These findings are valid for the neat epoxy system and the epoxy/NCs nanocomposites, regardless of the NCs type.
All the above observations are solid proof of the efficiency of the bio-derived IA as a crosslinker for the ELO matrix and its nanocomposites with NCs and confirm the occurrence of the curing process.

TMDSC Analysis of the Obtained Epoxy/NC Nanocomposites
The suitability of IA as a crosslinking agent, together with the influence of nanocellulose addition upon the curing behavior of the ELO bio-resin, was assessed by MDSC. Figure 3a-d show the thermal behavior of the epoxy resins after each step of the curing protocol and the main parameters i.e., the glass transition temperature (T g ), the maximum peak temperature (T P ), and corresponding enthalpies (∆H).  Further, the addition of NCs did not significantly affect the curing process, which proceeded similarly to that of the neat epoxy matrix but influenced the exothermic peak. Therefore, a slight decrease in the TP with about 6 °C was noticed in the case of the epoxy nanocomposite containing unmodified NC against the neat epoxy matrix, especially in the last step of the curing reaction, which suggests a higher reactivity of the E/NC nanocomposite. This may be due to the hydroxyl groups from the NC's surface, which could accelerate the curing reaction between the epoxy groups and IA through the opening of the oxirane cycle and the generation of new -OH groups that further interact with the IA structure [22]. This assumption is also supported by the higher value calculated for the curing enthalpy of the E/NC samples, because the formation of a higher number of bonds (even H bonds) will lead to an increased exothermal effect. Opposite behavior was observed for the nanocomposites containing grafted NCs (Figure 3c,d). In this case, the All the DSC curves displayed a single exothermic peak corresponding to the ringopening reaction between the epoxy functions of ELO and the carboxyl groups of IA (Figure 3a-d). The T P of the DSC exothermic curves can provide important information regarding the efficiency of the bio-derived IA as a curing agent, being oftentimes taken as an indicator for the reactivity of the constituents in a curing reaction, a lower peak corresponding to a higher reactivity of the components [22]. Moreover, both the neat epoxy matrix and its nanocomposites containing NCs displayed a single T g , and the T g values decreased with increasing the curing temperature (Figure 3a-d).
From the DSC curves of all studied samples, one can observe that the applied thermal curing protocol was effective, leading first to the decrease and then to the almost complete disappearance of the T P peak (Figure 3a-d). This behavior denotes the consumption of the reactive groups involved in the crosslinking mechanism and is strong evidence of a complete curing reaction. T P is clearly dependent on the increase in both curing temperature and time and it is accompanied by a dramatic decrease in the corresponding enthalpies values (Figure 3a-d). The values of the T P belonging to the neat epoxy matrix are comparable with already reported literature data concerning the crosslinking of different vegetable oils with C5 and C6 dicarboxylic acids [20,48]. T P values in the range 155-158 • C were reported by Di Mauro et al. who studied the reactivity of linseed and soybean epoxidized vegetable oils in the presence of IA. In this case, ethanol was employed as a solvent for the ELO resin and the cure protocol involved higher temperatures (150 • C and 180 • C) [20]. A T P value of~178 • C, closer to those of the samples in this study was found by Ding et al. for the ELO/adipic acid epoxy system in the presence of 4-N,Ndimethylaminopyridine (DMAP) as an accelerator [48].
Further, the addition of NCs did not significantly affect the curing process, which proceeded similarly to that of the neat epoxy matrix but influenced the exothermic peak. Therefore, a slight decrease in the T P with about 6 • C was noticed in the case of the epoxy nanocomposite containing unmodified NC against the neat epoxy matrix, especially in the last step of the curing reaction, which suggests a higher reactivity of the E/NC nanocomposite. This may be due to the hydroxyl groups from the NC's surface, which could accelerate the curing reaction between the epoxy groups and IA through the opening of the oxirane cycle and the generation of new -OH groups that further interact with the IA structure [22]. This assumption is also supported by the higher value calculated for the curing enthalpy of the E/NC samples, because the formation of a higher number of bonds (even H bonds) will lead to an increased exothermal effect. Opposite behavior was observed for the nanocomposites containing grafted NCs (Figure 3c,d). In this case, the number of free hydroxyl groups was reduced as a result of NCs chemical grafting, and the above-mentioned mechanism was hindered. Indeed, the addition of chemically modified NCs in the epoxy matrix results in a small increase (with less than 2 • C) of the T P value during the curing process (first and third stage of curing) as against neat sample. Similarly to the T P variation, a slight reduction in the T g values was found for the epoxy nanocomposites containing grafted NC [50]. Thompson et al. attributed the reduction in the T g of an epoxy composite containing 1 wt% cellulose nanocrystals modified with methacryloyl chloride (M/CNCs) to the fact that M/CNCs can act as a physical barrier within the epoxy matrix, reducing the curing efficiency of the crosslinking system or to the improved dispersion of the M/CNCs inside the matrix, which could have further hindered the curing process [51].

TGA Analysis of Neat Epoxy System and Epoxy/NC Nanocomposites
The thermal stability of the neat epoxy matrix and its nanocomposites containing NCs was assessed by TGA analysis and depicted in Figure 4, where the integrated (TGA) and derivative (DTG) forms are shown. The related thermal parameters i.e., the temperatures at 5% weight loss (T d,5% ) and at the maximum degradation rate (T d,max ), the weight loss at temperature of 200 • C (WL 200 • C ), and the residue at 750 • C (R 750 ) are summarized in Table 1.
The neat epoxy system and its nanocomposites containing NCs present similar degradation behavior, with two degradation processes being noticed in the thermal degradation profiles. No separate degradation stage of NCs was observed in the TGA or DTG thermograms of the epoxy nanocomposites, regardless of the NCs type. The first degradation stage from 150 • C to 300 • C is probably due to the labile C-O type bonds from the molecular structures with higher mobility not being embedded within the epoxy network [47,48,52].
The main degradation step, which is placed between 330 • C and 480 • C, results from the chain cleavage of the ester groups in the ELO epoxy resin structure [53,54]. The breakage of the ester groups formed through the ring-opening reactions between the epoxy functional groups and the carboxyl ones occurred above 480 • C, and this is more evident in the matrix without NCs (Figure 4)   On the other hand, the Td,max values of the nanocomposites were quite close to the one observed for the neat epoxy matrix. Therefore, regardless of the surface treatment,  The presence of NC and its treatment influenced the thermal stability of the nanocomposites, especially at temperatures below 300 • C. All nanocomposites, regardless of the NCs type, exhibited an important improvement in the thermal stability, T d,5%, increasing with up to~44 • C as compared with the neat epoxy matrix (Figure 4, Table 1). Additionally, small mass loss values at 200 • C were obtained for the nanocomposites as against the epoxy matrix without NCs (Table 1). In the case of ungrafted NC, the enhancement of the thermal stability can be explained by the active hydroxyl groups on the surface of NC that can be involved in the crosslinking reactions along with the carboxyl groups of IA, increasing the crosslinking efficiency. For the NC/SM, the poly(methacrylic acid) grafts present on its surface could react with the epoxy system during the cross-linking process, with the formation of a more complicated and stable network (Scheme 1) [51]. In both cases, a strong interfacial adhesion was formed between the NCs and the epoxy network [51]. Therefore, this high thermal stability of nanocomposites can be attributed to effective interactions between NCs and the epoxy matrix [55]. On the other hand, the Td,max values of the nanocomposites were quite close to the one observed for the neat epoxy matrix. Therefore, regardless of the surface treatment, the NC did not influence the Td,max, i.e., the temperature corresponding to the decomposition of the polymer with the highest rate. With regards to the residual content, a very Improved thermal properties were also reported by Pandurangan et al., when cellulose nanofibers derived from banana fibers were incorporated in low amounts (<3 wt%) into a commercially petroleum-based epoxy resin [56]. The authors explained the positive results through the nanolevel dispersion of the CNF particles inside the epoxy polymer matrix.
On the other hand, the T d,max values of the nanocomposites were quite close to the one observed for the neat epoxy matrix. Therefore, regardless of the surface treatment, the NC did not influence the T d,max , i.e., the temperature corresponding to the decomposition of the polymer with the highest rate. With regards to the residual content, a very small amount of char residue was obtained for the neat epoxy, and slightly higher amounts were detected for the nanocomposites due to the presence of NCs and the additional reactions discussed above.
The above-discussed results showed that the addition of NCs improved the thermal stability of the epoxy-based materials, thus increasing the chances of using vegetable oilbased epoxy systems in fields that require resistance to high temperatures. From Table 1 it is evident that, among all samples, the E/NC/SM nanocomposite exhibited the highest thermal stability.

DMA Analysis of Neat Epoxy System and Epoxy/NC Nanocomposites
The E' vs temperature curves of all materials are resembling those of amorphous thermoset polymers, presenting three stages: a glassy plateau, an α transition region (or "glass transition region") where the E' drops significantly as a consequence of the increase in the chain segments mobility near the glass transition temperature, and a rubbery plateau ( Figure 5).

DMA Analysis of Neat Epoxy System and Epoxy/NC Nanocomposites
The E' vs temperature curves of all materials are resembling those of amorphous thermoset polymers, presenting three stages: a glassy plateau, an α transition region (or "glass transition region") where the E' drops significantly as a consequence of the increase in the chain segments mobility near the glass transition temperature, and a rubbery plateau ( Figure 5).  Table 2, in the glassy domain (E'−25°C and E'0°C), the lowest E' was recorded for the E/NC nanocomposite. The reduction in the E' modulus for the E/NC nanocomposite as compared to the neat epoxy may be attributed to an insufficient interfacial adhesion between the hydrophilic NC and the hydrophobic epoxy matrix, which did not allow for an efficient load transfer from the matrix to the reinforcing agent.  As indicated in Table 2, in the glassy domain (E' −25 • C and E' 0 • C ), the lowest E' was recorded for the E/NC nanocomposite. The reduction in the E' modulus for the E/NC nanocomposite as compared to the neat epoxy may be attributed to an insufficient interfacial adhesion between the hydrophilic NC and the hydrophobic epoxy matrix, which did not allow for an efficient load transfer from the matrix to the reinforcing agent. On the contrary, the incorporation of surface-modified CNs had a strong reinforcing effect in both the glassy and the glass transition regions (Table 2). Thus, an improvement in the E' of 7.4% and 31%, respectively, was recorded at −25 • C for the E/NC/S and E/NC/SM nanocomposites as compared to the neat epoxy, showing that the surfacemodified NCs increased the epoxy matrix's ability to undergo mechanical stress with recoverable viscoelastic deformation [57].

As indicated in
The increase in the E' following the addition of modified NCs may be an expression of the good compatibility between the surface-modified NCs and the epoxy matrix due to the favorable interactions established between the polymethacrylate grafts of the modified NC and the epoxy matrix, as previously demonstrated by the thermal analysis results. Thus, the good adhesion at the interface between the nanofiller and the epoxy matrix facilitated a good load transfer from the epoxy matrix to the rigid NCs and led to nanocomposites with enhanced homogeneity and superior stiffness.
At room temperature, an increase in the E' 25 • C of around 44%, 30%, and 60% was obtained for the E/NC, E/NC/S, and E/NC/SM nanocomposites over the neat epoxy matrix, while at higher temperatures (i.e., at 50 • C) the increase was even more significant, being 232%, 190%, and 98%. In both cases, the higher E' values can be ascribed to the NCs which act as physical hindrances, restricting the free movement of the epoxy matrix chain segments. Again, the nanocomposite containing NC/SM stands out as compared with the other nanocomposites due to its higher stiffness. In this case, we might assume that NC/SM has an improved distribution in the epoxy matrix which leads to an improved toughening mechanism [58]. For verifying this assumption, light transmittance measurements were performed, and the results confirmed the better dispersion of NC/SM nanofibers in the epoxy matrix as compared with the nanocomposite containing unmodified or silane modified NC (Figure 6b). The light transmittance of polymer composites depends on many factors, such as the size and concentration of the filler, the refractive index of the components, the composite thickness, surface roughness, and the dispersion of the filler particles in the polymer matrix. Our nanocomposites containing unmodified NC, and NC bearing poly(methacrylic acid) grafts showed a high transparency, only 10% lower than that of the neat epoxy system. Similar results were obtained by Lee et al. [59] for millimeter thick cellulose nanofiber/epoxy laminates and by Fan et al. [60] for a biobased epoxy resin containing SiO 2 -hybridized nanocellulose.   The strong increase in the storage modulus of the E/NC/SM nanocomposite as compared to the neat epoxy matrix could indicate a poorer flexibility or even brittleness [61,62]. However, the E/NC/SM nanocomposite film showed good flexibility when subjecting to repeated manual flexures, returning to its initial form in about 70 s (Figure 7).
As one can observe from Table 2, the presence of ungrafted and silane-grafted NCs determined a slight increase in the T g values as compared with the neat epoxy matrix, which correlated well with the increase in the calculated crosslink density [48]. Thus, a higher crosslink density of the nanocomposites generates a decrease in the free volume within the neat epoxy system, restricting the free movement of the polymer chain segments in these nanocomposites, which leads to an increase of in the T g values [51].  The magnitude of tan δ peak was considerably reduced, by up to 42%, for the nanocomposites with ungrafted and grafted NCs as compared to the neat epoxy matrix ( Table 2). This is consistent with a less viscous behaviour and improved thermomechanical stability for the obtained nanocomposites due to the addition of NCs.

SEM Analysis of Neat Epoxy System and Epoxy/NC Nanocomposites
The SEM images of cryo-fractured samples from the cured neat epoxy matrix and its nanocomposites with NCs are shown in Figure 8. The magnitude of tan δ peak was considerably reduced, by up to 42%, for the nanocomposites with ungrafted and grafted NCs as compared to the neat epoxy matrix ( Table 2). This is consistent with a less viscous behaviour and improved thermomechanical stability for the obtained nanocomposites due to the addition of NCs.

SEM Analysis of Neat Epoxy System and Epoxy/NC Nanocomposites
The SEM images of cryo-fractured samples from the cured neat epoxy matrix and its nanocomposites with NCs are shown in Figure 8. The microstructure of the cured neat epoxy matrix shows a brittle behavior in liquid nitrogen with a completely smooth and flat surface and evidence of branched river-like fracture propagation (Figure 8) [38]. Contrarily, the epoxy nanocomposites showed remarkably different fracture microstructures (Figure 8). Thus, the fracture surface of all nanocomposites was rough, with a leaf-like uniform pattern consisting of both The microstructure of the cured neat epoxy matrix shows a brittle behavior in liquid nitrogen with a completely smooth and flat surface and evidence of branched river-like fracture propagation (Figure 8) [38]. Contrarily, the epoxy nanocomposites showed remarkably different fracture microstructures (Figure 8). Thus, the fracture surface of all nanocomposites was rough, with a leaf-like uniform pattern consisting of both crack-deflections and stress-whitening areas, with the latter being responsible for the restriction of fracture growth through the epoxy matrix (Figure 8) [38]. No pulled out NCs or voids were detected on the nanocomposite surface micrographs, revealing a good interfacial adhesion with the epoxy matrix.
Nevertheless, a distinct morphology was noticed in the case of E/NC/SM nanocomposite, probably due to the presence of poly(metacrylic acid) on the NCs surface which led to an increased interface volume and compatibility with the epoxy matrix [63]. Additionally, this sample showed a large number of stress-whitening zones due to the plastic deformation of the epoxy matrix around the SM grafted NCs [38]. Indeed, the nanocomposite containing NC/SM showed the best thermal and thermo-mechanical properties, thus confirming this observation. Moreover, few individual fractured NC ends (indicated by the yellow arrows) could be localized mainly on the surface of the E/NC/S and E/NC/SM nanocomposites, indicating that most of the NCs were embedded within the cross-linked network of the epoxy matrix and almost fully covered by the epoxy polymer chains (Figure 8).
The higher magnification SEM images of the fractured samples confirm the previous observations. More than that, a homogenous distribution of the grafted NCs was noticed in the epoxy matrix due to the good miscibility between the surface-treated nanofibrillated cellulose and the epoxy matrix ( Figure 9).

BDS Analysis of Neat Epoxy System and Epoxy/NC Nanocomposites
Generally, epoxy resins-based materials have widespread use in many applications because of their favorable mechanical properties, excellent chemical resistance, and especially good electrical insulation. Considering this last characteristic, the obtained bio-based nanocomposites were investigated through BDS in order to assess their suitability as potential candidates for replacing the petroleum-based thermosets in electronic

BDS Analysis of Neat Epoxy System and Epoxy/NC Nanocomposites
Generally, epoxy resins-based materials have widespread use in many applications because of their favorable mechanical properties, excellent chemical resistance, and especially good electrical insulation. Considering this last characteristic, the obtained bio-based nanocomposites were investigated through BDS in order to assess their suitability as potential candidates for replacing the petroleum-based thermosets in electronic applications.
The dielectric spectra of the neat bio-based epoxy system and epoxy nanocomposites containing differently treated NCs are shown in Figure 10a,b. The ε r decreased with frequency for all the samples in the studied range. The ε r values are mainly determined by the dielectric response of bio-based epoxy. The presence of NC, the epoxy-NC interface, and the silane treatment seem to have a small influence on the dielectric activity of the bio-epoxy material as previously observed for common epoxy nanocomposites [64]. The decrease was steeper for frequencies less than 1 Hz, mainly due to the electrode polarization, but a very limited influence of direct-current (dc) conductivity can also be possible. Although the difference between the values of the real permittivity of the nanocomposites is small, a lower ε r value was observed for the E/NC/S compared with the E reference and the other nanocomposites, regardless of the frequency. The lower permittivity of this nanocomposite as compared to the unfilled resin is unexpected and may be explained by the silane treatment of NC which prevents the absorption of water and reduces the number of hydrophilic OH on the surface of nanocellulose. A similar behavior was not observed in the case of E/NC/SM due to the polymethacrylic acid lateral chains and their hydrophilic carboxyl groups. It is worth remarking that the ε r 's values obtained in the case of bio-based epoxy and its nanocomposites are very close to those reported for both common and bio-based epoxy resins in a similar frequency range [65][66][67].  The dielectric loss (εr″) values are higher for all the samples at low frequencies due to electrode polarization (Figure 10b). A relaxation peak was observed for all the samples at about 10 2 Hz. This peak corresponds to the α-relaxation at the glass-to-rubber transition [68,69]. Interfacial polarization due to the accumulation of charges at the polymer-NC interface has a small influence on the variation of εr″ with frequency, probably because it is hidden by the low space charge level. Similar to the εr′, slightly smaller εr″ values were obtained for the E/NC/S nanocomposite as compared to the E reference and the other nanocomposites, showing slightly lower losses regardless of the frequency. A very small shoulder may be seen at high frequencies (10 5 -10 6 Hz), and it may be attributed to the β-relaxation of the dipoles associated with the polar groups of the bio-epoxy resin [68].
The small influence of the NCs, regardless of the surface treatment, on the dielectric properties of the bio-based epoxy matrix correlated with the improved thermal and mechanical properties. This shows that these nanocomposites may be recommended as Figure 10. Dielectric spectra of bio-based epoxy and its nanocomposites: real part of the complex relative permittivity(ε r ) (a) and imaginary part of the complex relative permittivity (ε r ) (b).
The dielectric loss (ε r ") values are higher for all the samples at low frequencies due to electrode polarization (Figure 10b). A relaxation peak was observed for all the samples at about 10 2 Hz. This peak corresponds to the α-relaxation at the glass-torubber transition [68,69]. Interfacial polarization due to the accumulation of charges at the polymer-NC interface has a small influence on the variation of ε r with frequency, probably because it is hidden by the low space charge level. Similar to the ε r , slightly smaller ε r values were obtained for the E/NC/S nanocomposite as compared to the E reference and the other nanocomposites, showing slightly lower losses regardless of the frequency. A very small shoulder may be seen at high frequencies (10 5 -10 6 Hz), and it may be attributed to the β-relaxation of the dipoles associated with the polar groups of the bio-epoxy resin [68].
The small influence of the NCs, regardless of the surface treatment, on the dielectric properties of the bio-based epoxy matrix correlated with the improved thermal and mechanical properties. This shows that these nanocomposites may be recommended as suitable materials for flexible electronic devices.

Isolation and Surface Treatment of NC
According to our previous report, NC was obtained from microcrystalline cellulose (MCC) via a high-pressure mechanical treatment [70]. Briefly, a 2 wt% MCC suspension in distilled water was passed 12 times at a pressure of 200 MPa through an LM20 type microfluidizer (Microfluidics, Westwood, MA, USA) with the formation of a cellulose gel. Further, the obtained cellulose gel was frozen, then freeze-dried, and finally milled using an ultra-centrifugal mill (ZM 200, Retsch GmbH & Co., Düsseldorf, Germany) at a speed of 6000 min −1 , when a nanofibrillated cellulose (NC) powder was obtained. The resulting NC was firstly treated with a 5 wt% γ-methacryloyloxypropyltrimethoxysilane (S) water solution at 50 • C, under strong magnetic stirring, for 4 h for generating active sites on its surface and then subjected to graft polymerization with methacrylic acid (M). The graft polymerization reaction was carried out using acetone as solvent and AIBN as initiator by mixing all components together and allowing them to react under reflux for 4 h. The silane modified (NC/S) and methacrylic acid grafted (NC/SM) nanocellulose were further washed with distilled water, freeze-dried, and milled ( Figure 1a). The results from the structural analysis of NC/S and NC/SM allowed us to conclude that both the silane and graft polymerization treatments were highly efficient [70].

Preparation of Epoxy System and Epoxy Nanocomposites
All formulations were prepared by maintaining a stoichiometric ratio of 1:1 between the epoxy groups of ELO and the acid groups of IA, while the weight ratio between THF and IA was set at 3:1. The content of NC, NC/S, and NC/SM, respectively, was maintained constant at 0.75 wt% in all the epoxy nanocomposites. For the preparation of the ELObased nanocomposites, the IA was first solubilized in THF under magnetic stirring at 50 • C until a transparent solution was obtained. Simultaneously, the desired amount of unmodified or surface-modified NCs was dispersed in the ELO resin and homogenized at 80 • C for 30 min under vigorous stirring. Afterwards, the THF/IA solution was poured over the ELO/NC mixture, and the stirring continued for another 30 min. Subsequently, the resulting nanocomposites were poured into silicon molds, degassed under vacuum at ambient temperature, and cured in an oven at 90 • C for 90 min, 120 • C for 60 min, and 140 • C for another 60 min. A neat epoxy system, without NCs, was prepared following the same protocol and served as a control. The schematic illustration of the nanocomposites' preparation protocol is displayed in Figure 1b while the samples' code names, together with the curing conditions, are summarized in Table 3. It is important to state that both the neat epoxy and the nanocomposites were transparent before and after curing, indicating a good macroscopic homogeneity (Figure 6a). The FT-IR absorbance spectra of pristine IA and ELO, the neat epoxy system and epoxy nanocomposites obtained following the curing protocol were recorded using a Jasco FTIR 6300 instrument (JASCO Int. Co., Ltd., Tokyo, Japan) in attenuated total reflectance (ATR) mode over a wavenumber range from 4000 to 400 cm −1 . The resolution and number of scanning cycles were 4 cm −1 and 32, respectively.

Temperature-Modulated Differential Scanning Calorimetry (TMDSC)
TMDSC analysis was employed for monitoring the efficiency of the curing protocol used in the preparation of the neat epoxy system and epoxy nanocomposites and was conducted by using a TA Instruments DSC Q2000 equipment (TA Instruments Inc., New Castle, DE, USA) under helium flow (25 mL/min), with the underlying rate of 10 • C/min, amplitude of 0.8 • C/min, and a period of 30 s as follows: equilibrating at −85 • C; isothermal for 5 min; heating from −85 • C to 305 • C.

Thermogravimetric Analysis (TGA)
The thermal stability of the cured neat epoxy matrix and epoxy nanocomposites was evaluated using TA Q5000 equipment (TA Instruments Inc., New Castle, DE, USA). The samples were analyzed in nitrogen at a flow rate of 50 mL min −1 during heating from room temperature to 750 • C at 10 • C min −1 . All samples were run in duplicate to verify the reproducibility, and the average values, along with the standard deviation, were reported.

Dynamic Mechanical Analysis (DMA)
The viscoelastic behavior of cured epoxy nanocomposites was assessed using a TA Instruments DMA Q800 (TA Instruments Inc., New Castle, DE, USA). The tests were performed in multi-frequency strain mode (tension clamp) by heating the bar samples with dimensions of 13 × 7 × 1.2 mm 3 (length × width × thickness) from −30 to 90 • C at a heating rate of 3 • C min −1 and an oscillation frequency of 1.0 Hz.
The crosslink density (ν) was determined in accordance with the Flory theory of rubber elasticity [71], using Equation (1): where E is the storage modulus of the neat epoxy system and epoxy nanocomposites, respectively, in the rubbery region at T = (T g + 50) in Kelvin, and R is the ideal gas constant.

Scanning Electron Microscopy (SEM)
The cured epoxy nanocomposites were fractured in liquid nitrogen and then sputtercoated (Q150R Plus, Quorum, SXE, Lewes, UK) with a 5 nm layer of gold. The surface of fractured blends was visualized with a Hitachi TM4000Plus II Benchtop microscope working in a low vacuum at 10 kV accelerating voltage and a working distance of about 7 mm. The SEM images were taken using backscattered electron detector (BSE).

Broadband Dielectric Spectroscopy (BDS)
BDS measurements were carried out with a Novocontrol Alpha-A Analyzer equipped with an Active Sample Cell ZGS over a frequency range from 5 × 10 −2 to 1 × 10 6 Hz. The real (εr ) and imaginary (εr ) parts of the complex relative permittivity were determined for each sample at the temperature of 30 • C. The samples were maintained for 30 min at this temperature before measurements.

Conclusions
Bio-based epoxy nanocomposites consisting of ELO, surface-modified NCs as fillers, and IA as cross-linking agent were developed for the first time. The overall results showed that NCs can be used for tailoring the mechanical properties of ELO-based systems. DSC results revealed that the addition of NCs into the bio-based epoxy network did not alter the curing mechanism and had a slight impact on the maximum peak temperature. Even at low content, NCs exhibited a favorable effect in terms of both thermal stability and mechanical behavior. As compared with the neat epoxy, the onset degradation temperature of the epoxybased nanocomposites increased by more than 30 • C, regardless of their treatment. The nanocomposite containing NC/SM showed the best mechanical performance, exhibiting an increase in the storage modulus by more than 60% at room temperature. This indicates a better dispersion of the nanofibers and a higher adhesion between the epoxy system and the poly(methacrylic acid) grafts, which resulted in a more homogenous structure, as supported by the SEM results. A small influence of NCs on the dielectric activity of the bio-epoxy nanocomposites was observed from the dielectric measurements, thus showing the suitability of the obtained bio-based epoxy/NCs for replacing the petroleum-based thermosets in electronic applications. Therefore, the developed bio-based nanocomposites may provide a new perspective on using them as coatings or flexible films for different engineering applications.

Data Availability Statement:
The data presented in this study are available on request from the corresponding authors.