**E**ff**ect of Carbon Nanostructures and Fatty Acid Treatment on the Mechanical and Thermal Performances of Flax**/**Polypropylene Composites**

**Pietro Russo 1, Libera Vitiello 1, Francesca Sbardella 2,\*, Jose I. Santos 3, Jacopo Tirillò 2, Maria Paola Bracciale 2, Iván Rivilla 2,4 and Fabrizio Sarasini 2,\***


Received: 31 January 2020; Accepted: 10 February 2020; Published: 13 February 2020

**Abstract:** Four different strategies for mitigating the highly hydrophilic nature of flax fibers were investigated with a view to increase their compatibility with apolar polypropylene. The effects of two carbon nanostructures (graphene nanoplatelets (GNPs) and carbon nanotubes (CNTs)), of a chemical modification with a fatty acid (stearic acid), and of maleated polypropylene on interfacial adhesion, mechanical properties (tensile and flexural), and thermal stability (TGA) were compared. The best performance was achieved by a synergistic combination of GNPs and maleated polypropylene, which resulted in an increase in tensile strength and modulus of 42.46% and 54.96%, respectively, compared to baseline composites. Stearation proved to be an effective strategy for increasing the compatibility with apolar matrices when performed in an ethanol solution with a 0.4 M concentration. The results demonstrate that an adequate selection of surface modification strategies leads to considerable enhancements in targeted properties.

**Keywords:** polymer matrix composites; flax fibers; surface treatments; adhesion

### **1. Introduction**

Natural-fiber-reinforced composites have received attention over the recent years because of their potential ability to replace their synthetic counterparts in an attempt to meet the new regulations that promote the use of more sustainable and recyclable materials [1,2]. The high specific mechanical properties and the carbon dioxide neutrality of natural fibers have already stimulated the replacement of glass fibers in several sectors, especially the automotive and construction ones, but usually as secondary load-bearing structures [3,4].

A step forward is their use in structural applications, but some challenges still need to be properly faced and solved [5]. The variability in physical and mechanical properties, due to their natural origin, is difficult to manage unless the fiber supply chain is carefully controlled and the manufacturing processes are optimized [6]. Goudenhooft et al. [7] recently showed that tensile properties of flax fibers are not significantly affected over time, regardless of the fiber yield and variety, and that the resulting dispersion in the specific mechanical properties is in the same range as that of glass fibers. Another significant issue is related to the processing conditions of the composites (temperature,

dwell time, pressure), which have a major impact on the final mechanical properties, especially for thermoplastic-based composites [8–10].

The last topic of considerable interest is the extent of fiber/matrix interfacial adhesion. It is well known that the mechanical properties of composites are dictated not only by the inherent properties of the constituents, but also by the fiber/matrix interface. The poor compatibility with polymer matrices (especially thermoplastics) due to their hydrophilic behavior still represents a major limitation for a wider industrial exploitation of natural fibers [11]. Several efforts to enhance the interfacial adhesion of natural fibers have been proposed, including chemical [12–16] and physical treatments [17–20], but their industrial implementations are often complicated by the large amounts of chemicals involved or the multiple processing steps required. A more recent approach deals with the grafting of nanostructures onto fiber surfaces to increase the adhesion with the polymer matrix. This strategy has been widely exploited for synthetic fibers, such as glass [21–23] and carbon fibers [24–26], but has attracted less attention in the field of natural fibers. Wang et al. [27] modified the surfaces of flax fibers by grafting TiO2 nanoparticles using a silane coupling agent. The authors reported an increase in tensile strength and interfacial strength with an epoxy matrix of 23.1% and 40.5%, respectively. Copper nanoparticles on flax fibers were found to produce significant improvements in fiber tensile modulus and strength, equal to 50% and 75%, respectively [28]. Ajith et al. [29] modified flax yarns with hydrous zirconia nanoparticles synthesized by hydrolysis of a zirconium oxychloride solution. The presence of these nanoparticles resulted in an increase in single fiber tensile strength and interfacial strength with an epoxy matrix of 85% and 65%, respectively. Lakshmanan and Chakraborty [30] synthesized and deposited silver nanoparticles on jute fibers without deteriorating the mechanical properties of the fibers. In addition, the modified fabrics exhibited good antibacterial properties. In [31], the authors reported a simple spray-coating process to deposit carbon nanotubes (CNTs) over the surfaces of ramie fibers. This coating enhanced the flexural strength and modulus of an epoxy-based composite by 38.4% and 36.8%, respectively, while a microdebonding test highlighted an increase in the interfacial shear strength of 25.7%. Sarker et al. [32] coated graphene materials, i.e., graphene oxide (GO) and graphene flakes (G), on alkali-treated jute fibers, and an interfacial shear strength enhancement of ~236% compared to untreated fibers was achieved. In [33], the authors coupled a jute fiber individualization procedure with the grafting of GO and subsequent hot pressing to get preforms that were then vacuum-infused with epoxy matrix. The graphene coating resulted in a dramatic increase in tensile modulus and strength of the jute–epoxy composites compared to untreated composites of 324% and 110%, respectively. Grafting of nanometer-sized materials can therefore be considered as an effective method for improving fiber/matrix interfacial adhesion, thus leading to the manufacturing of high-performance natural-fiber-reinforced composites. Another positive feature of this strategy is the possibility of adding functionalities to the resulting composites. Zhuang et al. [34] deposited multi-walled carbon nanotubes (MWCNTs) on the surfaces of jute fibers, and the epoxy-based composites exhibited multifunctional sensing abilities for temperature, moisture, and strain. In [35], graphene nanoplatelets (GNPs) and carbon black were used to make flax yarns electrically conductive; these were then used to fabricate stretchable strain sensors with gauge factors ranging from 1.46 up to 5.62, and a reliability for sensing strains of up to 60%.

The need to optimize the interfacial adhesion in natural-fiber composites is even more important with thermoplastic-based composites due to the non-polar nature of most of them. In particular, polypropylene (PP) is one of the most widely used polyolefins. Its low density, low price, good mechanical properties, good processability, and recyclability make it a popular material as a matrix for natural fiber composites [10,36,37]. Flax fibers currently account for about half of the natural fibers used in automotive applications, followed by kenaf and hemp [4], and the combination of PP/flax has been widely investigated in literature, highlighting the dramatic incompatibility between these two constituents. In an attempt to tailor the properties of natural fibers for their subsequent successful application in high-performance plant-fiber composites but with limited costs and environmental impact, in this work, we investigated the interfacial interactions in flax/PP composites through two

different approaches: (i) The grafting of carbon nanostructures (CNTs and GNPs) and (ii) the chemical modification with a fatty acid (stearic acid). In both cases, the addition of a maleic-anhydride-modified polypropylene (MAPP) was also used to tune the interfacial adhesion. In particular, stearic acid, a long alkyl chain fatty acid, was used to lower the hydrophilic character of flax fibers. This surface modification treatment has already been used with limited success in other studies [38–41], even though a detailed investigation on the effects of its concentration on the surface properties of flax fibers has not been reported so far. In addition, grafting of nanostructures for improving interfacial adhesion has mostly been exploited for thermoset-based composites, and scarcely with thermoplastic polymers [42]. The morphology and the thermal stability of flax/PP composites were characterized, and the impacts of the surface treatments or compatibilization with MAPP on their mechanical performance were addressed.

### **2. Materials and Methods**

### *2.1. Materials*

The composites investigated in this study are based on a polypropylene (PP) matrix (Hyosung Topilene PP J640, MFI@230 ◦C,2.16 kg: 10 g/10 min) supplied by Songhan Plastic Technology Co. Ltd. (Shanghai City, China) and a commercial 2 <sup>×</sup> 2 twill flax fabric (areal weight: 200 g/m2) commercialized without any specific sizing agent and supplied by Composites Evolution (Chesterfield, UK). The matrix was used as received or pre-modified by inclusion of 2 wt.% of a coupling agent, Polybond 3000 (maleic-anhydride-modified PP, MFI@190 ◦C, 2.16 kg: 400 g/10 min) from Chemtura (Cologne, Germany). The stearic acid (SA), ethanol, and toluene were of analytical grade and used without further purifications. Carbon nanotubes (CNTs) with average length < 1 μm, average outer diameter < 9.5 nm, and bulk density of 100 g/L were provided by Nanocyl SA (Sambreville, Belgium) with the code NC3150. The graphene nanoplatelets (GNPs) supplied by Nanesa s.r.l. (Arezzo, Italy) in the form of black powder have an average flake thickness of 40 nm corresponding to 40 stacked layers, an average particle size of 30 μm, a bulk density of 20–42 g/L, and a specific surface area (BET) > 30 m2/g.

### *2.2. Surface Treatment of Flax Fabrics with Carbon Nanostructures*

The adopted procedure involved the preparation of an aqueous dispersion of the carbonaceous filler with a concentration equal to 0.5 wt.%. In the case of carbon nanotubes, the dispersion was performed in the presence of 1% by weight of a Triton X-100 surfactant supplied by Sigma Aldrich (Milano, Italy). Layers of flax fabric, already cut to such dimensions as to be used for the preparation of the laminated samples, were immersed for 30 min at room temperature in these dispersions and pre-sonicated for 180 min at room temperature. Finally, the wet fabric layers were subjected to drying for 30 min at 80 ◦C in a ventilated oven.

### *2.3. Surface Treatment of Flax Fabrics with Stearic Acid*

Different concentrations (0.1, 0.2, 0.3, and 0.4 M) of stearic acid in toluene or ethanol were prepared. These solutions were heated at temperatures close to the boiling points of the solvents, 100 ◦C for toluene and 65 ◦C for ethanol, respectively. Once this temperature was reached, the reaction mixture, including the flax fabric, was maintained for 3 h and then washed three times with deionized water and dried at room temperature (Scheme 1). The carboxyl group (–COOH) is supposed to react with the hydroxyl groups of the fiber through an esterification reaction and, hence, the treatment should reduce the number of hydroxyl groups available for bonding with water molecules. Furthermore, the long hydrocarbon chain of stearic acid (18 carbon atoms) provides an extra protection from water due to its hydrophobic nature [43].

**Scheme 1.** Synthetic route for the modification of the flax fabric with stearic acid (SA).

### *2.4. Composite Manufacturing*

Composite samples with a symmetrical stacking sequence [0/90] and consisting of 8 plies were obtained by alternately stacking polymer films and layers of as-received or pre-treated flax fabrics, pre-conditioned in a vacuum oven at 70 ◦C for 2 h, and subsequently underwent hot compression at 210 ◦C. This last step was carried out using a Collin GmbH (Edersberg, Germany) model P400E press according to a pre-optimized pressure cycle: 2 min—0 bar, 2 min—5 bar, 2 min—15 bar, 1 min—25 bar, 2 min—35 bar, 2 min—40 bar. Finally, the cooling of the composite plates to 30 ◦C was conducted at a constant pressure of 40 bar before releasing the pressure and extracting the sample. These process conditions provided the laminates with an average thickness of 2.5 mm and with a fiber content of approximately 45% by volume. Films of PP and PPC (PP modified with coupling agent) with a thickness approximately equal to 80 μm were obtained with a Collin Teach-Line E 20-T single-head extruder and Collin CR 72T calender (Ebersberg, Germany), setting a screw speed of 55 rpm and a temperature profile from the hopper to the die equal to: 180–190–200–190–185 ◦C.

### *2.5. Characterization Techniques*

The mechanical properties of the flax fabrics before and after surface modification treatments were assessed according to ASTM D5035. Tensile tests were carried out at room temperature by means of a Zwick/Roell Z010 (Ulm, Germany) equipped with a 10 kN load cell. A gauge length of 75 mm was used for specimens with a width equal to 25 mm. Tests were performed in displacement control at a crosshead speed of 100 mm/min to ensure failure within 20 ± 3 s. At least five tests were performed for each fabric.

The mechanical properties of the composites were investigated in quasi-static tensile and flexural tests with a Zwick/Roell Z010. For tensile measurements, a gauge length of 60 mm and a cross-head speed of 5 mm/min were set in accordance with ASTM D3039, while flexural tests were conducted in a three-point bending configuration with a span of 76 mm and a speed of loading equal to 5 mm/min as per ASTM D790. Five tests were carried out for each composite formulation.

FT-IR spectra were carried out with a Bruker Vertex 70 spectrometer (Bruker Optik GmbH, Ettlingen, Germany) equipped with a single reflection Diamond ATR (Attenuated Total Reflectance) cell. The ATR-FTIR spectrum was recorded with a 3 cm−<sup>1</sup> spectral resolution in the mid-infrared range (350–4000 cm<sup>−</sup>1) using 256 scans.

CPMAS (Cross Polarization/Magic Angle Spinning) NMR spectra were recorded on a 9.4T (400 MHz) Bruker (Billerica, USA) system equipped with a 4 mm MASDVT Double Resonance HX MAS probe. Larmor frequencies were 400.17 MHz and 100.63 MHz for 1H and 13C nuclei, respectively. Chemical shifts were calibrated indirectly with glycin, with a carbonyl peak at 176 ppm. The sample rotation frequency was 10 kHz and the relaxation delay was 5 s. The number of scans was 4096. Polarization transfer was achieved with RAMP cross-polarization (ramp on the proton channel) with a contact time of 5 ms. High-power SPINAL 64 heteronuclear proton decoupling was applied during acquisition.

X-ray diffraction (XRD) analysis was performed with a diffractometer X'Pert PRO by Philips (Malvern, UK) (CuKα radiation = 1.54060 Å; 40 kV and 40 mA) at room temperature. XRD patterns were collected in the range of 2θ = 10◦–80◦ with a step size of 0.02◦ scan and a time per step of 3 s.

Surface wettability tests were performed measuring the contact angles of water droplets on the twill fabric surface using an optical analyzer (OCA15Pro, DataPhysics Instruments, Filderstadt, Germany). The static sessile method with a droplet volume of 3 μL was selected and a Milli-Q ultrapure water was used as the testing liquid. A minimum of ten droplets localized on different areas of the flax fabric samples were analyzed. Contact angle values were determined by drop shape analysis using the DataPhysics SCA 20 software module.

A scanning electron microscope, FEG Mira3 (Tescan, Brno, Czech Republic), was used to analyze the morphologies of neat and pre-treated flax fabrics as well as those of fractured surfaces of composite laminates. Prior to observation, the fracture surfaces were sputter-coated with gold.

### **3. Results and Discussion**

### *3.1. Characterization of Composites Reinforced with Flax Fabrics Decorated with Carbon Nanostructures*

In an attempt to reduce the use of chemicals and to make the process easy and industrially scalable, a simple dip-coating method was used to decorate the flax fibers. SEM micrographs were taken to investigate the morphologies of coated and uncoated fabrics (Figures 1–3). Comparing the surfaces of the untreated (Figure 1) and treated flax fabrics with CNTs (Figure 2), it is possible to observe the formation of an interconnected MWCNT network (white arrows in Figure 2c) on the fiber surface, though the dispersion was not uniform with the presence of agglomerates (Figure 2d). Untreated flax fibers (Figure 1) showed an almost smooth and featureless surface, with the presence of some impurities because no pre-treatment was applied.

**Figure 1.** SEM micrographs showing untreated (**a**) flax yarn and (**b**) a close-up view of a single flax fibre.

The same conclusions hold for flax fabrics decorated with GNPs (Figure 3). The distribution is not completely uniform over the fiber surface, and agglomerates can be easily detected (white arrows in Figure 3c,d). These results indicate that the bonding between CNTs, GNPs, and flax fibers may only result from weak van der Waals forces with no covalent bonds. In principle, natural fibers exhibit the unique feature of having a set of hydroxyl groups in cellulose that are reactive and available for potential interactions with host nanostructures, coupled with mechanical interlocking with the rough grooves that characterize the surfaces of natural fibers. This feature was used by Sarker et al. [32] for decorating jute fibers with a uniform layer of graphene oxide (GO) by exploiting the oxygen functional groups of GO and the effects produced by an alkali pre-treatment that removed the cementing layer and exposed the hydroxyl groups of cellulose. When less-reactive graphene flakes were used, these were not fixed on the jute fiber surface. Wang et al. [31] observed a uniform deposition of CNTs on ramie fibers, but also, in this case, the fibers were subjected to an alkali pre-treatment and CNTs were prepared in a solution containing a silane coupling agent and a dispersant (polyvinylpyrrolidone). The results highlighted the positive role played by the silane coupling agent, which was essential in order to avoid

agglomeration and to promote the formation of Si–O–C covalent bonds between the silane molecule and the hydroxyl groups of ramie fibers.

**Figure 2.** SEM micrographs showing the flax fabrics decorated with carbon nanotubes (CNTs) at different magnifications.

**Figure 3.** SEM micrographs showing the flax fabrics decorated with graphene nanoplatelets (GNPs) at different magnifications.

In the present work, to keep the number of processing steps and the amounts of chemicals at a minimum, no pre-treatments were used, and this resulted in a certain degree of agglomeration of both carbon nanostructures, which were not functionalized.

The convenient dip-coating method used did not lead to a significant reduction in the mechanical properties of the fabrics if one considers the natural variability in the mechanical responses of natural fibers, as can be inferred from the tensile tests on flax fabrics (Figure 4), thus excluding any degradation effects of the cellulose and of the cell wall materials.

**Figure 4.** Breaking forces of untreated and surface-modified flax fabrics.

GO and graphene flakes significantly increased both the tensile strength and the Young's modulus of single jute fibers [32]. This better mechanical performance was ascribed by the authors to a kind of healing effect played by these nanostructures, in that they were able to remove stress concentrations on the fiber surfaces due to their homogeneous and uniform coating. A similar increase in tensile strength was reported in [29] on single flax fibers, again attributed to the removal of surface defects, even though, in this case, the distribution of zirconia particles was not sufficiently homogeneous.

Despite the non-optimal distribution of carbon nanostructures on the flax fabrics, the surfacemodified fabrics were used for manufacturing composites based on the PP matrix. To tailor the interfacial adhesion, a standard coupling agent (MAPP) was used to explore any synergistic effects. The mechanical properties of tension and bending are summarized in Table 1, along with the percentage of variation in comparison with untreated PP/Flax composites. The incorporation of a polypropylene-grafted maleic anhydride (PPC/Flax) improved both the flexural and tensile properties due to an increased interfacial adhesion between the flax fibers and the PP matrix.

**Table 1.** Summary of the tensile and flexural properties of the composite materials based on polypropylene (PP) and modified flax fabrics (YM = Young's modulus; TS = Tensile strength; FM = Flexural modulus; FS = Flexural strength; percentage variation is in reference to the corresponding property of PP/Flax).


The maleic anhydride polar groups create covalent and hydrogen bonds with the flax fiber surfaces, while the polypropylene chains of MAPP form compatible blends with the bulk PP matrix through

co-crystallization [37]. The improved interfacial adhesion can be readily confirmed by comparing the fracture surfaces of untreated (Figure 5) and compatibilized (Figure 6) composites.

**Figure 5.** SEM micrographs of the fracture surfaces of PP/Flax composites at different magnifications.

**Figure 6.** SEM micrographs of the fracture surface of PPC (PP modified with coupling agent)/Flax composites at different magnifications.

In non-compatibilized composites, flax yarns are completely pulled out from the matrix and interfacial debonding turns out to be the dominant failure mechanism, thus suggesting a rather low fiber/matrix adhesion. Clear gaps at the interfaces between the fibers and the PP matrix can be easily observed (Figure 5d), along with the clean surfaces of the flax fibers (Figure 5c) with no matrix residues.

On the contrary, a lower degree of fiber pull-out and no gaps were found between the fibers and the matrix (Figure 6b–d), with a large amount of matrix adhering to the flax surface (Figure 6d) and fiber fractures (Figure 6b).

The presence of CNTs was not beneficial, with resulting composites that exhibited reduced flexural and tensile properties to a great extent compared to untreated composites. This behavior can be ascribed to the agglomerations of CNTs which acted as stress concentrations and to their poor compatibility with the PP matrix [44,45]. Figure 7 shows the fracture surfaces of such composites, where carbon nanotubes are entangled and defects along the flax fibers can be observed (Figure 7d). The fibers are still scarcely covered by the matrix after pull-out (Figure 7b,d), and in the grooves created by the pulled-out fibers (Figure 7c), CNTs clusters can easily be seen, thus confirming the occurrence of weak van der Waals bonds between the CNTs and the flax fibers. Due to the poor mechanical results and fiber/matrix adhesion, composites with CNTs were not investigated further. In fact, it is often suggested that the existence of hydrogen bonds or other dipole–dipole interactions between maleic anhydride and modified nanotubes with carbonyl and carboxyl groups increases CNT dispersion and the properties of PP-based nanocomposites [46,47].

**Figure 7.** SEM micrographs of the fracture surfaces of PP/Flax\_CNT composites at different magnifications.

On the other hand, the decrease in mechanical properties caused by the presence of GNPs was lower compared to CNTs. The fracture surfaces of the resulting composites (Figure 8) showed the same features as those of CNT-reinforced composites, but with a slightly better interfacial compatibility, confirmed by layers of matrix material pulled out together with the flax fibers (Figure 8b,c), where GNPs are well embedded in the polymer matrix (white arrows in Figure 8c). The relative chemical inertness of GNPs [48] did not allow the exploitation of their full potential; therefore, these composites were modified with the MAPP coupling agent.

**Figure 8.** SEMmicrographs of the fracture surfaces of PP/Flax\_GNP composites at different magnifications.

These composites (PPC/Flax\_GNP) outperformed the tensile and flexural behaviors of untreated PP/Flax composites, showing values of tensile and flexural strength even higher than those of PPC/Flax composites. As already found for GNPs in polypropylene matrices [48,49], MAPP is able to increase the chemical compatibility between PP and the polar groups of GNPs, providing a better anchoring of the GNPs into the PP matrix, which results in the enhanced adhesion between them and in a higher constraint of the polymer chains. These effects are visible in the fracture surfaces of the corresponding composites (Figure 9). In this case, it is much more difficult to differentiate the flax fibers from the PP matrix, and the extent of fiber pull-out is significantly reduced. In addition, the fibers are, in many cases, completely covered by the matrix reinforced with GNPs (Figure 9d). It is interesting to note that the tensile and flexural moduli of these composites are higher than those reported in literature for flax/PP composites [10,50,51] when considering similar fiber volume fractions, even higher than low-twisted and unidirectional MAPP-treated flax yarns in a polypropylene matrix, for which a tensile modulus of 9.26 ± 0.4 GPa was reported [52]. In addition, the tensile and flexural strength values are comparable with those of similar unidirectional composites [52]. Figure 10 shows the thermal stability (TGA) curves of the composites, and the results point out that the presence of carbon nanostructures did not markedly affect the degradation profile in comparison with untreated flax/PP composites.

While the first weight loss around 340–360 ◦C is due to the degradation of flax fibers and, in particular, of the cellulose [53]; the second significant weight loss (380–470 ◦C) is related to the degradation of PP [44]. The increased decomposition temperatures (inset of Figure 10) of the composites can be ascribed to the physical–chemical absorption of the decomposed products [54]. The physical absorption of PP molecules on the carbon nanostructures induces a delay in their volatilization, but the decomposition temperature of PPC/Flax\_GNP composites is the highest among the other formulations. This significant increase cannot be due only to physical absorption, but also to a chemical absorption, thus confirming the higher level of interfacial adhesion.

**Figure 9.** SEM micrographs of the fracture surfaces of PPC/Flax\_GNP composites at different magnifications.

**Figure 10.** Thermograms for the different PP/flax composites.

### *3.2. Characterization of Composites Reinforced with Flax Fabrics Treated with Stearic Acid*

Organic acids, and in particular fatty acids, are extensively used in surface treatments for particulate mineral fillers. The resulting modification causes the filler surface to become hydrophobic, thus reducing the moisture adsorption during storage and improving the incorporation of polar mineral fillers in non-polar polymer matrix melts, with reduced melt viscosity and associated enhanced dispersion [55]. These commercially available fatty acids are generally sourced from plant or animal

sources and contain mixtures of mainly even-carbon-number acids. The use of stearic acid as a surface modification treatment for natural fibers has been investigated in other studies dealing with PP. This treatment is usually performed via a solution process, in which the stearic acid is dissolved in a suitable solvent, via a vapor phase [56], or by dry-blending [39]. In the first case, different solvents have been suggested in literature, including from acetone [38], toluene [57], and ethanol [58]. During the modification, it is expected that the carboxyl group of the stearic acid reacts with the hydroxyl groups of the natural fibers, but the OH groups of the different solvents, characterized by different reactivity, could also be involved. This explains why it was decided to investigate the effects of two different solvents, toluene and ethanol.

At first, all of the treated flax samples were analyzed with the FT-IR and CPMAS NMR spectroscopies to understand the chemical structure. In the infrared spectra (Figure 11), the adsorption bands at about 2916 and 2848 cm−<sup>1</sup> are attributed to the asymmetric (νas(CH2)) and symmetric (νs(CH2)) methylene vibration, while the carbonyl absorption of the carboxylic acid dimer (νC=O) for stearic acid appeared clearly at 1703 cm−1. This last band at 1703 cm−<sup>1</sup> is a strong stretching vibrational mode of modified cellulose, which can be attributed to the ester –C=O moieties present. These are formed by esterification between –CO2H in stearic acid and –OH in modified cellulose, indicating that stearic acid undergoes a chemical reaction with cellulose. The main bands between 815 and 1469 cm−<sup>1</sup> were attributed to the δOH, νC–O, deformation bands of (–CH2–)n, and the out-of-plane vibration bands of O–H of stearic acid dimer [59,60]. The bands at 3024–3650 cm−<sup>1</sup> correspond to the stretching vibrations of the OH of cellulose, which is the main element of flax created via β1-4 linked D-glucose. The corresponding vibrational bands of C=O and OH are gradually affected as the concentration of SA increases from 0.1 to 0.4 M. This fact indicates the strong intermolecular hydrogen bond interactions between cellulose and SA. The interactions between SA and cellulose were further studied by solid-state 13C CP/MAS NMR (Figure 11c (ethanol) and d (toluene)) at 0.4 M. The spectrum exhibited some characteristic peaks at 181.3 and 181.0 ppm, corresponding to C from the –CO– group to free stearic acid (Figure 11c.1,d.1) and the ester group (Figure 11d.3,c.3), respectively, 104.2, 88.3, 74.4, and 71.5 ppm, corresponding to C1, C4, C5, and (C3, C2), respectively, and the peak for C6 at 64.3 ppm. These peaks could be assigned to the cellulose. In addition, 32.3 (c.3 and d.3) and 32.14 (c.1 and d.1), 24.7 and 14.5 ppm could be assigned to the aliphatic chain to SA [61,62]. The slight shift of the resonance corresponding to the group –CO– with respect to the same group –CO– of the stearic acid, together with the broadening of the signals corresponding to the –CH2– groups of the aliphatic chain of the stearic acid, as well as a slight chemical shift, would indicate the binding or formation of ester groups in the flax fabric after treatment. These data could indicate that, in both cases, using toluene or ethanol as solvents and at two different temperatures, the cellulose that makes up the flax fabric is functionalized with stearic acid through its OH groups.

An assessment of the wettability of the flax fabric surfaces was performed on the flax fabrics treated with stearic acid + ethanol (0.1–0.4 M) and stearic acid + toluene (0.1–0.4 M). In Figure 12, the contact angles for the samples treated with 0.4 M of stearic acid are reported, both in toluene (b) and in ethanol (c). The images were taken from the videos at a fixed time of 10 s after water contact with the fabrics, so that all of the treated fabrics could be compared with each other. The values of the contact angles at different concentrations of stearic acid for the fabrics treated in ethanol and in toluene are reported in Table 2.

**Figure 11.** FTIR and NMR spectra for flax modified by stearic acid. (**a**) (Ethanol) and (**b**) (toluene) FTIR spectra ([SA] = 0.1, 0.2, 0.3, and 0.4 M). 13C CP/MAS NMR spectrum of stearic acid **1** (SA), flax fabric (blue) **2,** and, in purple, the flax fabric treated with 0.4 M of SA in ethanol (**c**) and in toluene (**d**).

**Figure 12.** (**a**) Plot of contact angle vs. stearic acid concentration. Picture of a drop of water on flax fabric treated with 0.4 M stearic acid in toluene (**b**) or ethanol (**c**).

**Table 2.** Contact angles for flax fabrics treated with stearic acid in different solvents.


The results shown in Table 2 demonstrate that the surface modification carried out in ethanol reaches higher values in terms of contact angle, up to 128.8◦ relative to a 0.4 M concentration of stearic acid, thus obtaining highly hydrophobic surfaces; likewise, it is evident that the synthesis carried out in toluene as a solvent led to lower contact angle values, leaving the surface of the flax mostly hydrophilic [63]. The differences in hydrophobicity shown in Figure 12 and Table 2 by the flax fabric samples treated in different solvents may be due to different phenomena. The reaction between an acid and an alcohol, which is known as Fischer–Speier esterification [64], produces an ester by refluxing a carboxylic acid and an alcohol in the presence of an acid catalyst. In our case, in none of the reactions was an acid used as a catalyst, but EtOH was able to yield H<sup>+</sup> to the reaction media and then to act as catalyst and reaction solvent at the same time. In fact, the pKa of EtOH is 15.9, while that of toluene is 43. This fact makes EtOH a stronger acid with respect to toluene. In addition, this reaction, which takes place through cationic type intermediates, is most favored in polar media such as EtOH [65]. On the other hand, EtOH itself could give esterification processes, whereby not only SA, but also esters would be formed with the OH groups of the cellulose. This fact could significantly reduce the number of OH groups present in the sample, increasing hydrophobicity due to the loss of OH groups that can interact with water via H-bonds. Finally, the relatively high temperature at which the reaction is carried out with toluene, 100 ◦C, could cause some degradation of the molecular structure of the cellulose of flax, making it less reactive. The high level of hydrophobicity reached with a treatment performed in ethanol at 0.4 M allowed us to select this treatment condition for the modification of flax fabrics to be used as reinforcement in the PP matrix.

To determine the structure and dispersion of stearic acid on flax fabrics before composite manufacturing, the morphologies of the resulting treated fabrics were investigated by SEM (Figure 13).

**Figure 13.** SEM micrographs showing the flax fabrics modified with 0.4 M SA at different magnifications.

The fibers appear to be covered by a thin layer of stearic acid with some micro-sized waxy protrusions (white arrows in Figure 13d), indicating a quite uniform distribution of stearic acid on the flax fiber surface. Figure 14 shows the XRD spectra of pure stearic acid and modified flax fabrics. The main characteristic peaks of the untreated flax fabric were located at 2θ = 14.7◦, 16.4◦, 22.6◦, and 34.5◦, which can be assigned to cellulose I [66], for planes 110 , (110), (200), and (004), respectively. In the surface-modified flax fabric, additional peaks located 21.6◦ and 24.0◦ can be clearly seen, which correspond to the interplanar spacings of stearic acid, thus suggesting that the stearic acid exists in its

crystal form in the modified fabric. These values can be assigned to the stearic acid monoclinic C-form, which is in line with the crystallized form obtained from solution [67].

**Figure 14.** XRD patterns of pure stearic acid (SA), untreated flax fabric (Neat flax), and flax fabric modified with 0.4 M SA (Flax–SA).

The effect of stearic acid treatment on the thermal stability of flax fabrics was investigated by thermogravimetric analysis, and the corresponding thermograms are reported in Figure 15. The thermogram of the untreated flax fabric showed the typical three peaks in the derivative curve. The first mass loss, at about 60–120 ◦C, is due to the release of water, a shoulder at about 240–280 ◦C is ascribed to the decomposition of the non-cellulosic components such as pectin and hemicellulose, and the third mass-loss peak, at about 340–360 ◦C, is due to the cellulose degradation [68]. A one-step mass loss of pure stearic acid was observed, with an onset weight loss temperature higher than 230 ◦C, which is thus compatible with the manufacturing process of PP-based composites. The thermal stability of the modified flax fabric was not significantly affected, with the exception of an additional decomposition step at a lower temperature (>230 ◦C) due to the vaporization of stearic acid [69], which again confirms the successful deposition of stearic acid on the flax fabric. In addition, the stearic acid treatment did not degrade the mechanical properties of the modified flax fabric, as can be inferred from the results included in Figure 4.

The reduced polar character of the modified flax fabrics resulted in composites characterized not only by higher moduli compared to composites reinforced with GNPs, but also by lower strength values, which, in any case, are higher than those exhibited by PP/Flax composites (Table 1). It is worth noting the synergistic effect on the modulus and, to a lesser extent, on the strength played by the further use of MAPP, an effect already observed by Spoljaric et al. in microcrystalline cellulose composites [57]. An improved interfacial bond strength can be the reason for this behavior. Zafeiropoulos et al. [70] reported a slight increase in stress transfer efficiency in flax fibers treated with stearic acid in the vapor phase after 36 h treatment with a PP matrix. They ascribed this improvement to the inter-entanglement of the stearic acid chains with the PP chains. The same authors also reported the development of a transcrystalline layer in stearic-acid-treated cellulose fibers [71], which was suggested to increase the interfacial adhesion, as assessed by fiber fragmentation tests.

**Figure 15.** (**a**) Thermal stability (TGA) thermograms and (**b**) derivative thermogravimetry (DTG) curves of flax fabrics before and after the treatment with stearic acid (SA).

In the present work, a better fiber/matrix adhesion was induced by the surface treatment between the hydrophobic chains of stearic acid and the polypropylene matrix and between the hydrophilic carboxyl group of the stearic acid and the flax fibers. This was confirmed by SEM analysis of the fracture surfaces of the resulting composites, reported in Figures 16 and 17 for non-compatibilized (PP/Flax\_SA) and compatibilized (PPC/Flax\_SA) composites with stearic-acid-treated flax fibers, respectively. A strong interfacial adhesion was found in PP/Flax\_SA composites, which increased after the addition of MAPP. It is possible to note the presence of a significant number of stearic acid plate-like crystals on the flax fiber surface (for instance, the white arrows in Figure 16c,d). These are supposed to create a rough surface on the flax surface, thus promoting mechanical interlocking and hindrance to polymer chain mobility, which supports the significant increase in stiffness. Flax fibers can be hardly seen in Figure 17a,b, as they are densely covered with matrix residues. The dramatic increases in the moduli were not accompanied by similar increases in strength. It is speculated that the large amount of stearic acid deposited on the fiber surface might have acted, at stresses higher than those needed to evaluate the tensile and flexural moduli, more as a lubricant than as a compatibilizer. Stearic acid is, in fact, commonly used as a processing aid to increase homogeneity and processability [39].

**Figure 16.** SEM micrographs of the fracture surfaces of PP/Flax\_SA composites at different magnifications.

**Figure 17.** SEM micrographs of the fracture surfaces of PPC/Flax\_SA composites at different magnifications.

This consideration suggests not only the potential of stearic acid, but also the need to optimize its content in order to balance these opposing effects.

Compared to untreated PP/Flax composites, the treatment with stearic acid did not modify the overall degradation profile (Figure 18), even though a lower onset temperature of thermal instability occurred in composites with stearic acid. A slight shift toward higher temperatures for the two peak mass-loss temperatures was detected after the incorporation of stearic acid. The mass-loss rate for compatibilized systems was slightly reduced, thus confirming the higher level of interfacial adhesion.

**Figure 18.** (**a**) TGA thermograms and (**b**) derivative thermogravimetry (DTG) curves of stearic-acid-treated flax fibers in a non-compatibilized (PP) or compatibilized (PPC) matrix.

### **4. Conclusions**

Four different surface modification treatments, including grafting of GNPs and CNTs, stearation, and incorporation of maleated polypropylene, were developed and applied on flax fibers to produce high-performance polypropylene-based composites. The grafting of carbon nanostructures by a simple and cost-effective dip-coating process was implemented in order to try to limit the amounts of chemicals and the number of processing steps. This resulted in a non-optimal distribution of carbon nanostructures on the fiber surface. GNPs were found to be much more effective than CNTs, leading to composites with an increased Young's modulus and tensile strength of 54.96% and 42.46% compared to

the reference ones when combined with maleated PP. These results are comparable to those obtained for unidirectional PP/Flax composites developed in other studies. The stearation treatment was optimized in terms of the solvent type and the amount of stearic acid. A 0.4 M concentration of stearic acid in ethanol provided the highest reduction in the polarity of flax fibers without altering their degradation profile and mechanical properties. The higher compatibility with apolar PP resulted in enhanced mechanical properties in tension by 102.96% and 3.39% for modulus and strength, respectively, and in bending by 61.81% and 13.97% for modulus and strength, respectively, compared to baseline composites; these were further improved by the addition of maleated polypropylene. These simple treatments, potentially prone to further optimization, can represent a step toward producing natural fiber composites with mechanical profiles compatible with semi-structural applications.

**Author Contributions:** All authors have read and agreed to the published version of the manuscript. Conceptualization, F.S. (Francesca Sbardella), F.S. (Fabrizio Sarasini), and I.R.; methodology, F.S. (Francesca Sbardella), P.R., and J.T.; validation, F.S. (Francesca Sbardella), F.S. (Fabrizio Sarasini), and I.R.; formal analysis, F.S. (Francesca Sbardella), F.S. (Fabrizio Sarasini), and J.T.; investigation, J.T., I.R., F.S. (Francesca Sbardella), L.V., M.P.B., and J.I.S.; data curation, F.S. (Francesca Sbardella), I.R., and F.S. (Fabrizio Sarasini); writing—original draft preparation, F.S. (Francesca Sbardella), F.S. (Fabrizio Sarasini), and I.R.; writing—review and editing, P.R., J.T., and M.P.B.; visualization, F.S. (Francesca Sbardella), I.R., F.S. (Fabrizio Sarasini); supervision, F.S. (Fabrizio Sarasini), P.R., and J.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

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

### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Modeling the Sti**ff**ness of Coupled and Uncoupled Recycled Cotton Fibers Reinforced Polypropylene Composites**

**Albert Serra 1, Quim Tarrés 1,2, Miquel-Àngel Chamorro 3, Jordi Soler 3, Pere Mutjé 1,2, Francesc X. Espinach 4,\* and Fabiola Vilaseca <sup>5</sup>**


Received: 3 October 2019; Accepted: 17 October 2019; Published: 21 October 2019

**Abstract:** The stiffness of a composite material is mainly affected by the nature of its phases and its contents, the dispersion of the reinforcement, as well as the morphology and mean orientation of such reinforcement. In this paper, recovered dyed cotton fibers from textile industry were used as reinforcement for a polypropylene matrix. The specific dye seems to decrease the hydrophilicity of the fibers and to increase its chemical compatibility with the matrix. The results showed a linear evolution of the Young's moduli of the composites against the reinforcement contents, although the slope of the regression line was found to be lower than that for other natural strand reinforced polypropylene composites. This was blamed on a growing difficulty to disperse the reinforcements when its content increased. The micromechanics analysis returned a value for the intrinsic Young's modulus of the cotton fibers that doubled previously published values. The use of two different micromechanics models allowed evaluating the impact of the morphology of the fibers on the Young's modulus of a composite.

**Keywords:** recycled cotton fibers; stiffness; micromechanics; Young's modulus

### **1. Introduction**

The use of fibrous industrial byproducts as reinforcement for polymer-based composites has increasingly been attracting the attention of researchers. The use of byproducts is in line with the principles of green chemistry and the actual demands of the society for greener materials and more environmentally friendly products [1–3]. The literature shows the opportunity to use agroforestry wastes such as prunings, used paper fibers, or textile byproducts [4–8]. These studies reveal how the nature of the reinforcements has a high impact on the mechanical properties of its composites. In this sense, artificial fibers like glass fibers, aramids, or carbon fibers show the highest strengthening and stiffening abilities [9,10]. Natural fiber strands and wood fibers also show notable capabilities as polyolefin reinforcements. Nonetheless, strands like jute or hemp showed higher stiffening potential than wood fibers [11–15]. In this sense, cotton strands have been used as polyolefin reinforcement successfully [16–18]. While some of the studies have used raw cotton as reinforcing fibers, a vast

majority prefer to use recycled fibers from the textile industry [7,13,19,20], however, the number of published studies are still limited.

Cotton fibers recovered from the textile industry have some advantages, such as low cost and availability, but also some apparent drawbacks, since usually these fibers contain textile dyes [7,20]. Additionally, there is a large quantity of discarded textiles that are directly landfilled [21]. Moreover, landfilled textiles contribute to the formation of 'leachate' that can contaminate ground and underground waters [22]. Thus, the use of such textiles as composite reinforcement can contribute to widen the value chain of the textile sector on the one hand, and to decrease landfilling and contamination on the other hand.

Cotton fibers are almost 100% cellulosic fibers, and thus have a high presence of hydroxyl groups in their surface, and a high potential to create hydrogen bonds under favorable conditions. Therefore, they tend to aggregate, making their individualization and dispersion on a polymeric phase difficult [8]. In addition, cotton fibers are hydrophilic, while the vast majority of polymeric matrices are hydrophobic.

A previous study revealed that the presence of dyes diminished the hydrophilicity of cotton fibers, allowing the obtaining of better interphases without using any coupling agent [7,20]. The same dyes eased the dispersion of the fibers without any treatment. Nonetheless, the tensile strength of the composites reinforced with dyed cotton fibers were lower than those obtained with other natural fiber strands. Some authors claim that the same dyes hindered the action of coupling agents [7], although these researchers did not publish any results concerning the stiffness of the composites. According to the literature, the intrinsic Young's modulus of cotton fibers is found between 5 and 13 GPa, however, this value seems too low compared with the values obtained for other strands [11,12,23–28].

This paper examines the Young's modulus of cotton fiber (CF) reinforced polypropylene (PP) composites. CFs were recovered from a yarning process where all the fibers with lengths below 10mm were discarded. The byproduct has the shape of cotton dyed flocks that must be individualized prior to its use as reinforcement [20]. Composite materials adding CF percentages ranging from 20 to 50 wt% were formulated. Two batches of every formulation were prepared, one with 6 wt% of coupling agent added, and the other without. The composites were mold injected to obtain the standard specimens, and later on tested under tensile conditions. The Young's moduli of the materials were evaluated and discussed. The Young's moduli of the composites were not coherent with the intrinsic Young's modulus for CFs found in the literature. Therefore, a micromechanical analysis was proposed to analyze the properties of CF. First, the Hirsch model provided a value for the intrinsic Young's modulus of CF that doubled those on the literature [29]. Then, the efficiency factors allowed discussing a possible poor dispersion of the fibers at high reinforcement contents. Finally, the Tsai and Pagano model in combination with Halpin and Tsai equations [30,31] allowed incorporating the morphology of the reinforcements to back-calculate a theoretical Young's modulus for the composites. The paper actualizes the value of the intrinsic Young's modulus of cotton fibers, and proposes a series of composites that reuse textile byproducts, and thus avoids their landfilling or incineration.

### **2. Materials and methods**

### *2.1. Materials*

The cotton fibers (CF) used as reinforcement were recovered from cotton flock residues. These cotton flocks are textile industry byproducts and are composed of entangled cotton fibers with lengths too short for spinning. The flocks were previously treated with a reactive dye and were kindly supplied by Fontfilva S. L. (Olot, Girona, Spain).

A polypropylene (PP) Isplen PP090 62M by Repsol-YPF (Tarragona, Spain) was kindly supplied by its producer and used as the polymeric matrix. The use of a coupling agent was proposed in order to prevent chemical incompatibilities between the hydrophilic reinforcements and the hydrophobic matrix. Epolene G3015 polypropylene functionalized with maleic anhydride (MAPP) by Eastman Chemical Products (San Roque, Spain) was purchased for this purpose. This reactive has an acid number of 15 mg KOH/g and a Mn of 24800.

Decalin (decahydronaphthalene) was acquired from Fischer Scientific (Madrid, Spain) and had a 190 ◦C boiling point and 97% purity. This reagent was used to dissolve the PP matrix in the fiber extraction from composites.

Figure 1 shows the workflow for the research, from the production of cotton flocks by the textile industry to the measurement and evaluation of the mechanical properties.

**Figure 1.** Workflow of the research, including the production of cotton flock byproducts, composite mixing and material testing.

### *2.2. Cotton Flocks Treatment and Composites Preparation*

The cotton residues were passed through a blade mill in order to individualize the entangled fibers. The mill was provided with a 1mm sieve to obtain cotton fibers able to attain a good dispersion. These CFs were mixed with the PP in a Brabender Plastograph kinetic mixer by Brabender® (Duisburg, Germany). The coupled composites added a 6 wt% of MAPP at the same time than the other phases. The process took 10 min, at 185 ◦C and at a speed of 80 rpm. Coupled and uncoupled composites with CF contents ranging from 20 to 50 wt% were prepared. The obtained blends were cut down to 8 mm pellets able to be mold injected. These pellets were stored for 24 h in an oven at 80 ◦C to eliminate the humidity.

### *2.3. Composite and Standard Specimen Preparation*

The composite pellets were injection molded in the shape of standard dog bone specimens, in agreement with ASTM D638. The injection molding machine was a Meteor-40 by Mateu & Solé (Barcelona, Spain). The machine has three heating areas that were operated at 175 ◦C, 175 ◦C, and 190 ◦C, corresponding to the highest to the nozzle. The injection pressure was 120 kg/cm−<sup>2</sup> and the maintaining pressure was 37.5 kg/cm<sup>−</sup>2. A steel mold with a cavity in the shape of the standard specimen was used, and at least ten specimens for every one of the composite formulations were obtained.

### *2.4. Mechanical Test*

Prior to any mechanical test, the specimens were stored in a conditioning chamber by Dycometal. The stabilization of the specimens took 48 h, and the conditions were at 23 ◦C with 50% relative humidity.

The specimens were placed in the gauges of an Instron TM 1122 universal testing machine. The machine was fitted with a 5 Kn load cell. The test was performed in agreement with ASTM D790 standard. An extensometer MFA2 was used to measure the strains with addequate precision. The measurements were the result of testing at least 5 samples for every composite formulation.

### *2.5. Morphologic Analysis of the Reinforcements*

Some micromechanics models use the morphology of the reinforcements as input. As soon as the literature accepts that the morphology of such reinforcements changes noticeably during composite preparation, the study was performed to reinforcements extracted from the polymeric matrix. The extraction was obtained by matrix solubilization using a Soxhlet apparatus and using Decalin as a solvent. Composite material pieces approximately 10 × 10mm were placed inside a cellulose filter into the Soxhlet equipment. The process lasted 24 h until the matrix was totally dissolved. Then, the fibers were rinsed with acetone and distilled water.

The morphology of the fibers was measured in a FS-300 Kajanni analyzer. The equipment measured from 2500 to 3000 fibers and returned a fiber length distribution, mean length, and diameter and the percentage of fines (fibers shorter than 70 μm).

### **3. Results and Discussion**

### *3.1. Young's Modulus of the Composites*

Table 1 shows the Young's moduli of the coupled and uncoupled composites (*Et <sup>C</sup>*) reinforced with CF contents ranging from 20 to 50 wt%. The table also shows the tensile strength of the composites (σ*<sup>t</sup> <sup>C</sup>*), the percentage of reinforcement in weight (*WF*), and its volume fraction (*VF*).


**Table 1.** Young's modulus and tensile strength of the cotton fiber (CF)/polypropylene (PP) composites.

It was found that the use of a coupling agent had a low effect on the Young's modulus of the composites. In fact, an ANOVA analysis (at 95% confidence rate) reveals that the differences between the Young's moduli of the composites with the same percentage of reinforcement, despite adding or not adding a coupling agent, were not statistically relevant. This result was expected as it has been reported previously in the literature [5,32]. The same materials revealed totally different behaviors in the case of the tensile strength, where the presence of MAPP considerably increased the strength of the materials [7,20]. Thus, while the coupling agent has a noticeable effect on the tensile strength of the composites, its impact is not statistically relevant in the case of the Young's modulus. Some authors prefer to state that the strength of the interphase between the matrix and the reinforcements has a limited impact on the stiffness of the composites [6,11]. Other authors prefer to justify the differences on the methods used to evaluate the strength and the modulus. While the strength is measured at the maximum strain, where the interphase has been fully put to test, the Young's modulus is measured at low strains [32].

The Young's modulus of a semi-aligned short fiber reinforced composite is mainly affected by the properties of the phases and its contents, the morphology of the reinforcement, its mean orientation, and its grade of dispersion. In the case of a correctly dispersed reinforcement, the increase of the Young's modulus against reinforcement content was expected to be linear (Figure 2) [32].

**Figure 2.** Young's modulus of the coupled and uncoupled CF-PP composites against reinforcement content.

Both, coupled and uncoupled composites showed a linear evolution of its Young's moduli against CF content. Thus, a proper dispersion of the reinforcement was assumed. Nonetheless, the higher the percentage of reinforcement the harder it becomes to obtain a good or proper dispersion. In the case of the coupled composite at 50 wt% of CF, the Young's modulus seems to start to decrease under the regression line. Besides, CF incorporates a textile dye that seems to increase the strength of the interphase and ease the dispersion at low reinforcement rates. Nonetheless, it was impossible to corroborate this evolution due to the impossibility of preparing materials with higher reinforcement contents that are still able to be mold injected. From now on, and due to the equivalence between coupled and uncoupled CF-based composites, the analysis will be referring to the coupled materials.

The literature shows multiple studies on the evolution of the Young's modulus against the fiber contents. We have chosen stone groundwood fibers (SGW), commonly used for papermaking, hemp strands (HS), as a byproduct of agroforestry, old newspaper fibers (ONPF), as recycled fibers, and glass fibers (GF) as an industrial commodity and the most commonly used reinforcement [11,14,33]. Table 2 shows the Young's moduli of SGW, HS, ONPF and GF reinforced PP composites.


**Table 2.** Young's moduli of stone groundwood, hemp strands, and glass fiber reinforced polypropylene coupled composites.

The Young's moduli of natural fiber reinforced polypropylene composites are similar, with slight advantages for those reinforced with strands, especially at high reinforcement contents. Cotton fibers showed Young's moduli as superior to SGW and ONPF, and in line with the other strands. Nonetheless, cotton fibers are recycled and a byproduct of the textile industry, while hemp strands can be considered virgin materials. ONPF are recycled fibers that come from the disintegration of used newspaper. The Young's moduli of ONPF and SGW based composites are very similar, showing that recovering the fibers from the paper had little effect on the stiffening potential of the reinforcements. Moreover, CF showed higher Young's moduli than ONPF based composites. All the natural-based composites showed a linear evolution of their Young's moduli against fiber contents, but different slopes on their regression lines.

On the other hand, GF-based materials showed noticeably higher Young's moduli than natural fiber-based composites. At the same reinforcement contents, Young's modulus of CF-based composites is noticeably lower than GF-based ones. It was necessary to increase 20 wt% the amount of CF to obtain a Young's moduli similar to GF.

### *3.2. Neat Contribution of the Reinforcements*

Attending to the above-mentioned parameters that affect the Young's modulus of a composite, the differences must be related with the morphology of the reinforcements, its mean orientation, or the intrinsic properties of the phases. The modified rule of mixtures (RoM) for the Young's modulus summarizes all these parameters (Equation (1)):

$$E\_t^C = \eta\_{t^c} E\_t^F \cdot V^F + \left(1 - V^F\right) \cdot E\_t^M \tag{1}$$

where *Et <sup>C</sup>*, *Et <sup>F</sup>*, and *Et <sup>M</sup>* are the Young's moduli of the composite, reinforcement, and matrix, respectively. *VF* represents the reinforcement volume fraction, and η*<sup>e</sup>* is a modulus efficiency factor that equalizes the contribution of the reinforcements to the Young's modulus of the composite. This efficiency factor is seldom presented as a length efficiency factor times an orientation efficiency factor (η*e*= η*l*· η*o*). At the exception of the intrinsic Young's modulus of the reinforcements and the modulus efficiency factor, the rest of the values can be easily obtained during the tensile test of the composites. Clearly, the RoM can only be used if the Young's modulus of the composite evolves linearly against reinforcement content.

In any case, the neat contribution of the reinforcements to the Young's modulus of the composite is represented by η*e*·*Et <sup>F</sup>* in the RoM. Thus, the RoM can be rearranged to account for such neat contribution as:

$$
\eta\_{\epsilon'} \cdot E\_t^F = \frac{E\_t^C - \left(1 - V^F\right) \cdot E\_t^M}{V^F} \tag{2}
$$

Then, if the neat contribution is represented against the reinforcement volume fraction, a regression line is obtained, and the slope of such a line has been referred to in the literature as a fiber tensile modulus factor (FTMF) [6,32]. This factor can be used as a measure of the stiffening capabilities of a reinforcement. Figure 3 shows the FTMF for different fibers as polypropylene reinforcement.

The FTMF of CF was between HS and SGW. This value ensures good stiffening abilities for CF as PP reinforcement because the literature shows possible applications for materials with similar FTMF for building or product design purposes [34,35]. Moreover, some researchers used an ONPF-based composite to substitute a GF-based one [36]. On the other hand, GF showed higher stiffening capabilities than the rest of the reinforcements. This is not a surprise, having in account that GF is a man-made material with more stable intrinsic properties and a regular morphology. The FTMF of the reinforcements shows a similar behavior than the Young's moduli of its composites. Thus, the differences between such moduli seem to be focused on the neat contribution of the fibers, specifically, the intrinsic Young's modulus of the reinforcement and the modulus efficiency factor. In order to analyze such differences, the researchers propose a micromechanics analysis.

**Figure 3.** Neat contribution of the reinforcements to the Young's modulus of the polymers.

### *3.3. Micromechanics Analysis of the Young's Modulus*

The RoM (Equation (1)) shows two unknowns that coincide with the neat contribution of the fibers: η*e*·*Et <sup>F</sup>*. While it is possible to measure the intrinsic Young's modulus of the fibers, and more so in the case of the strands, some authors defend the use of micromechanics methods as an alternative [11,37,38]. In addition, a high number of experiments are necessary due to the foreseeable standard deviations of the mechanical properties of natural fiber reinforcements. Thus, the Hirsh model was proposed as a means to evaluate the intrinsic Young's modulus of CF.

$$E\_t^C = \beta \cdot (E\_t^F \cdot V^F + E\_t^M \{1 - V^F\} + (1 - \beta) \frac{E\_t^F \cdot E\_t^M}{E\_t^F \cdot V^F + E\_t^m (1 - V^F)} \tag{3}$$

where β is a parameter that modules the stress transference between both phases of the composite material. In the case of semi-aligned short fibers reinforced composites β has a value of 0.4 [14]. Table 3 shows the micromechanical parameters obtained after the analysis.


**Table 3.** Micromechanics of the Young's moduli of CF reinforced polypropylene coupled composites.

The mean intrinsic Young's modulus of CF was found to be 27.87 ± 2.63 GPa, similar to HS, with a value of 26.8 GPa [11]. This coincidence agrees with the already similarities found in the neat contributions of such fibers (Figure 3). Nonetheless, the computed intrinsic Young's modulus of CF

contrasts heavily with the neatly inferior values found in the literature. Some authors place this intrinsic Young's modulus in the range from 5 to 13 GPa [26–28]. Using such values with the RoM is not possible to reach the obtained experimental values without using modulus efficiency factors outside the usual range.

On the other hand, the value for CF is noticeably higher than the value obtained for SGW and ONPF, 21.2 ± 1.9 and 22.8 ± 1.8 GPa, respectively [14,39]. This also agrees with the neat contributions of such fibers. Similarly, GF showed an intrinsic Young's modulus of 76 GPa, justifying the differences obtained in the Young's modulus of its composites [11,40].

Consequently, the intrinsic Young's moduli of the different reinforcements affected heavily the Young's moduli of its composites. Nonetheless, CF showed a higher intrinsic Young's modulus than HS, but HS-based composites showed higher Young's moduli, at the same reinforcement contents than CF-based composites. Thus, the differences are expected to be found in the modulus efficiency factor (Table 3).

The values for the modulus efficiency factor were obtained by using all the experimental data (Table 1) and the mean intrinsic Young's modulus of CF (Table 3). The mean value was found to be 0.47 ± 0.03. The value is inside the usual range of values, between 0.45 and 0.56 for such factor [11,12,14,40]. Nonetheless, it is worth noting that the obtained value is in the lower half bound of the expected values. Thus, presumably, CF based composites have not taken advantage of the stiffening capabilities of CF. Particular values decrease when the CF contents increase (Table 3). The composite with a 20 wt% of CF exhibits a modulus efficiency factor higher than the other CF-based composites, and also a higher intrinsic Young's modulus, indicating a higher yield on the stiffening capabilities of CF. The reasons must be found on the mean orientation of the fibers, the morphology of the reinforcements or its dispersion.

In the case of HS the modulus efficiency factor was evaluated at 0.50 ± 0.02. This value is higher than CF and can compensate the difference between the intrinsic Young's moduli of the reinforcements and justify the higher moduli of the HS-based composites. In the case of ONPF, the value of η*<sup>e</sup>* was evaluated at 0.49 ± 0.04, a value similar to HS. Finally, SGW showed the highest values for η*e*, with a mean of 0.56 ± 0.02.

In order to find the impact of the morphology and the mean orientation of CF, the morphology and orientation efficiency factors were computed. The length efficiency factor was calculated according to Cox-Krenchel's model (Equations (4) and (5)) [41]:

$$\eta\_l = 1 - \frac{\tanh\left(\frac{\mu \cdot L^F}{2}\right)}{\left(\frac{\mu \cdot L^F}{2}\right)}\tag{4}$$

with

$$\mu = \frac{1}{r^F} \sqrt{\frac{E\_t^M}{E\_t^F \cdot (1 - \nu) \cdot Ln \left(\sqrt{\pi / 4 \cdot V^F}\right)}}\tag{5}$$

where *L<sup>F</sup>* and *r<sup>F</sup>* are the reinforcement mean weighed length and radius, respectively. The Poisson's ratio of the matrix is represented by ν and μ is a coefficient of the stress concentration rate at the end of the fibers. The Poisson ratio was 0.36, as found in the literature [22]. The orientation factor η*<sup>o</sup>* was obtained from η*<sup>o</sup>* = η*l*/η*e*. Table 3 shows the obtained values.

The length efficiency factor remained almost the same for all the composite formulations, with a mean value of 0.89 ± 0.01. Usually this factor decreases when the percentage of reinforcement increases [5]. This is due to the changes in the mean length of the reinforcements during compounding, when reinforcements are exposed to attrition phenomena and tend to break. There is a decrease in the mean length of such reinforcements as the reinforcement content increases [32]. Thus, these changes are expected to affect the Young's moduli of the composites. In the case of CF based composites, the impact of the morphology of the fibers seems to little impact the Young's modulus, although the reinforcements decreased their mean length from 293 to 185 μm [7]. This hypothesis will be put to test later on by applying a different micromechanics model.

On the other hand, the orientation efficiency factor clearly changed with the amount of reinforcement (Table 3). This factor showed a mean value of 0.53 ± 0.04 and ranged from 0.58 to 0.49. Usually, the orientation efficiency factor is more stable than the length efficiency factor, because the mean orientation of the fibers is heavily impacted by the geometry of the injection mold and the parameters used during the mold injection [42,43].

Fukuda and Kawata [44] studied the tensile modulus of short fiber reinforced thermoplastics, and the orientation of the fibers inside the composites. The authors proposed different fiber distributions, but based on the literature, a rectangular distribution (square packing) renders adequate results for short fiber semi-aligned reinforced composites [32,45]. The authors present an equation that computes the orientation efficiency factor from a mean orientation angle (α):

$$\eta\_{\vartheta} = \frac{\sin(\alpha)}{\alpha} \Big( \frac{3-\nu}{4} \frac{\sin(\alpha)}{\alpha} + \frac{1-\nu}{4} \frac{\sin(3\alpha)}{3\alpha} \Big) \tag{6}$$

Equation (6) was used to compute the mean orientation angles of the reinforcements (Table 3). The mean orientation angle was found to be 53.3 ± 3.3◦. This value is in line with other natural fiber reinforced composites, thought in the upper bounds, meaning that the reinforcements are les oriented than the expected. It must be stressed that the orientation decreased with the amount of reinforcement, from 48.8◦ for the composite containing 20 wt% of CF to 56.2◦ for the composite containing 50 wt%. In other cases, it was found that the composites with higher fiber contents showed higher orientations [11]. In these studies, it was assumed that the shorter fibers were easily aligned than the longer ones. Nonetheless, the RoM does not incorporate a factor taking in account the dispersion of the fibers, and as has been previously commented, the authors suspect that the dispersion of the composites at high-reinforcement contents was improvable.

### *3.4. E*ff*ect of the Morphology of the Reinforcements*

While the rule of mixtures (Equation (1)) and Hirsch's equation (Equation (3)) are elegant models that allow predicting the Young's modulus of a composite from a variety of parameters, they do not incorporate any morphological property of the reinforcements. The morphology of the reinforcements is known to greatly impact the mechanical properties of a composite, especially the ratio between its mean length and diameter, known as aspect ratio [12,46,47]. Thus, the authors propose using the Tsai and Pagano model (Equation (7)) in combination with Halpin and Tsai equations (Equations (8)–(10)) to evaluate a theoretical Young's modulus of the composites.

In agreement with Tsai and Pagano model, the stiffness in the fiber direction is given by:

$$E\_t^C = \frac{3}{8}E^{11} + \frac{5}{8}E^{22} \tag{7}$$

where, *E*<sup>11</sup> and *E*<sup>22</sup> are the longitudinal and transversal elastic modulus, calculated by the Halpin–Tsai equations [11]:

$$E^{11} = \frac{1 + 2\left(L^F / 2r^F\right) \cdot \lambda\_I V^F}{1 - \lambda\_I V^F} E\_t^M \tag{8}$$

with

$$\lambda\_I = \frac{\left(E\_t^F / E\_t^M\right) - 1}{\left(E\_t^F / E\_t^M\right) + 2\left(L^F / 2r^F\right)}\tag{9}$$

and:

$$E^{22} = \frac{1 + 2\lambda\_t V^F}{1 - \lambda\_t V^F} E\_t^M \tag{10}$$

with

$$\lambda\_t = \frac{\left(E\_t^F / E\_t^M\right) - 1}{\left(E\_t^F / E\_t^M\right) + 2} \tag{11}$$

Table 4 presents the obtained values. The values obtained by using the micromechanics model show a good alignment with the experimental values, especially at reinforcement contents higher than 20 wt%. In these cases, the error goes to a maximum of 5.6%, which is assumable when modeling a mechanical property of a natural fiber reinforcement composite. In fact, the standard deviation of the experimental values is higher than this maximum of 5.6% (Table 1). The higher errors found for the composite with a 20 wt% of CF can be explained by the intrinsic Young's modulus derived from these composites (Table 2). The value of such a parameter is 11.5% higher than the mean value. If the 31.48 GPa value is used as input for the Tsai and Pagano model, the theoretical Young's modulus increases to 3.02 GPa. This value is more similar to the value obtained experimentally. Moreover, the authors blame the differences between the experimental and theoretical values to deviations from a totally linear evolution of the Young's moduli against reinforcement contents—due to possible auto entanglements between the CF—and thus showing a slightly worse dispersion of the CF.

**Table 4.** Theoretical Young's moduli of the composites computed by using the Tsai and Pagano model in combination with Halpin andTsai equations.


The obtained values were plotted against the experimental ones to find the grade of correlation between both values (Figure 4).

**Figure 4.** Correlation between the experimental Young's moduli of the composites and the computed ones by using the Tsai and Pagano model in combination with Halpin and Tsai equations: (**A**) Unweighted correlation; (**B**) correlation line adding the condition of such line going through the origin.

The regression lines for the uncoupled and coupled composites using the experimental and theoretic values (Figure 4A) show 1.05 slopes, with a high correlation. This slope shows a high coincidence between the computes and experimental values. A second regression line was proposed, passing thorough the origin (Figure 4B). In this case the slopes were found to be 0.99 and 0.98 for the uncoupled and coupled composite, respectively. As mentioned above, these results prove the accuracy of the values predicted from the model. The values also prove the impact of the morphology of the reinforcements in the Young's moduli of the polymers, as Tsai and Pagano model uses these values as the input.

### **4. Conclusions**

Composite materials with CF as reinforcement and PP as the matrix were formulated with 20 to 50 wt% CF contents. Two batches were prepared, one including a 6 wt% of coupling agent and another without.

The Young's moduli of the composites were little impacted by the presence of coupling agents and thus by the strength of the interphase. Thus, for applications where stiffness is paramount, uncoupled composites can be used with the same response under load as the coupled composites. On the other hand, in the case of semi-structural uses, the coupled composites ensure a higher tensile strength and similar deformations under load than the uncoupled ones.

The Young's moduli of the composites were similar to those obtained for natural strands reinforced composites and higher than wood fibers reinforced composites. The presence of a textile dye in the CF decreased its hydrophilicity but also seemed to increase the difficulty in obtaining a good dispersion of the fibers inside the composite, and the decrease in the quality of the dispersion decreased the stiffening yield of CF.

A value of the intrinsic Young's modulus of CF was obtained by using Hirsh's model. The mean value of the modulus was found to be 27.87 GPa, lower than other strands. Nonetheless, this value doubles the values previously published. The value obtained for the CF in a composite with 20 wt% of reinforcement reached 31.48 GPa, a value more proper to other natural strands like hemp.

The micromechanics properties of the Young's modulus of the composites showed the effect of the orientation and the morphology of the fibers. Nonetheless, the authors found some deviations from a linear behavior of the Young's modulus against CF contents. The authors assume that the dispersion of the fibers can be improved to increase the Young's moduli of the composites with high rates of reinforcement. Nonetheless, further research is needed to prove this point.

This paper shows the opportunity of recovering textile cotton fibers, which are useless for the textile industry, to obtain composite materials able to replace glass fiber reinforced materials. In doing so, the dumping and incinerating of such fibers is avoided and the value chain of the textile industry is widened.

**Author Contributions:** A.S. performed the experiments and wrote the first version of the manuscript; M.-À.C. and J.S. conceived and designed the experiments; F.X.E. and Q.T. performed the calculus and represented the data; F.V. and P.M. guided the project. All the authors contributed to writing and correcting the document.

**Funding:** The authors wish to acknowledge the financial support of the Càtedra de Processos Industrials Sostenibles of the University of Girona.

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

### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **E**ff**ect of Graphene Oxide Coating on Natural Fiber Composite for Multilayered Ballistic Armor**

### **Ulisses Oliveira Costa, Lucio Fabio Cassiano Nascimento \*, Julianna Magalhães Garcia, Sergio Neves Monteiro, Fernanda Santos da Luz, Wagner Anacleto Pinheiro and Fabio da Costa Garcia Filho**

Department of Materials Science, Military Institute of Engineering-IME, Rio de Janeiro 22290270, Brazil **\*** Correspondence: lucio\_coppe@yahoo.com.br; Tel.: +55-(21)-98500-7084

Received: 3 July 2019; Accepted: 12 August 2019; Published: 16 August 2019

**Abstract:** Composites with sustainable natural fibers are currently experiencing remarkably diversified applications, including in engineering industries, owing to their lower cost and density as well as ease in processing. Among the natural fibers, the fiber extracted from the leaves of the Amazonian curaua plant (*Ananas erectifolius*) is a promising strong candidate to replace synthetic fibers, such as aramid (Kevlar™), in multilayered armor system (MAS) intended for ballistic protection against level III high velocity ammunition. Another remarkable material, the graphene oxide is attracting considerable attention for its properties, especially as coating to improve the interfacial adhesion in polymer composites. Thus, the present work investigates the performance of graphene oxide coated curaua fiber (GOCF) reinforced epoxy composite, as a front ceramic MAS second layer in ballistic test against level III 7.62 mm ammunition. Not only GOCF composite with 30 vol% fibers attended the standard ballistic requirement with 27.4 ± 0.3 mm of indentation comparable performance to Kevlar™ 24 ± 7 mm with same thickness, but also remained intact, which was not the case of non-coated curaua fiber similar composite. Mechanisms of ceramic fragments capture, curaua fibrils separation, curaua fiber pullout, composite delamination, curaua fiber braking, and epoxy matrix rupture were for the first time discussed as a favorable combination in a MAS second layer to effectively dissipate the projectile impact energy.

**Keywords:** curaua fibers; graphene oxide coating; epoxy composites; ballistic performance

### **1. Introduction**

In recent decades, the increasing efficiency of ballistic armors has emerged as a relevant factor in personal and vehicular security, for both civilian and military protection. The search for lighter and stronger armor materials has been increasing in proportion to the escalating power and sophistication in firearms development [1]. Research works are showing that polymer composites reinforced with natural lignocellulosic fibers (NLFs) present ballistic efficiency in multilayered armor systems (MAS), with front ceramic, comparable to synthetic aramid fabric, such as Kevlar™ [1–17]. In general, NLF composites have the advantage of environmental sustainability in association with cost-effectiveness, lower density, and easy fabrication as compared to synthetic fibers composites [18–21].

Together with ballistic protection, recent works on nano and micro cellulose [22–27], are also disclosing special applications for NLFs. Among the several papers on ballistic application of NFL composites for MAS second layer stands those using curaua fibers (CF) [1,6,7,10,11,17]. This fiber, native of the Amazonian region, is extracted from the leaves of a plant, *Ananas erectifolius*, sharing the pineapple family. It has attracted considerable interest as polymer composite reinforcement [28–34] owing to relatively lower density (0.96 g/cm3) in comparison to glass (2.58 g/cm3) and aramid (1.44 g/cm3) synthetic fibers [35]. In consequence, the CF specific tensile strength (~2.2·GPa.cm3/g) is higher than that of glass (~1.4 Gpa·cm3/g) and close to that of aramid (~2.8·Gpa.cm3/g) fibers.

As most NLFs applied in polymer composite [36–41], the curaua fiber also displays low interfacial shear strength, associated with poor fiber adhesion, while reinforcing a polymer matrix. This is due to their amorphous hemicellulose and lignin that act as natural hydrophilic wax adsorbing water on the fiber surface. Consequently, a weak bonding is expected to exist between the surface of the curaua fiber and the hydrophobic polymer such as polyester [29] and epoxy [37]. This affects the composite performance as MAS second layer for ballistic protection. Indeed, the impact of a high velocity projectile against a MAS with curaua composite results in different fracture mechanisms including delamination and matrix cracking pattern as well as fiber rupture and pullout [7,10,11]. Some of these mechanisms are essential for impact energy. However, others like delamination can impair the integrity of the composite target after a first ballistic shooting. This causes loss of its ability to protect against serial shootings as required by the standard [42].

In spite of the comparable ballistic performance to a same thickness Kevlar™ laminate as MAS second layer, the integrity of a NLF composite is always questionable. Lower amounts, usually less than 30 vol%, of fiber were found to result in composite shattering [4,5,8,9,11–17]. Even a 30 vol% NLF composite may be split by delamination, i.e., decohesion between fiber and matrix, which allows easy perforation of the projectile in case of a second shooting. Surface modification of NLFs has extensively been applied to improve the fiber matrix adherence [43,44]. This will be an effective way to prevent delamination.

Since the rise of graphene [45], it has increasingly been studied and investigated for possible technological applications. In particular, graphene has attracted a considerable attention for its superior performance as composite reinforcement owing to outstanding mechanical properties [46]. The direct oxidation of graphite is considered as an alternative route for producing substantial quantities of another remarkable material, the graphene oxide (GO). Studies conducted on the properties of GO revealed good chemical reactivity and easy handling owing to its intrinsic functional groups in association with amphiphilic behavior [47,48]. Among the several methods reported, to improve NLF composite adhesion and prevent lamination, only few have today been dedicated to graphene or graphene oxide coating [45,49,50].

To the knowledge of the authors of the present work, GO has not yet been applied as a coating onto NLFs to improve interfacial shear strength with respect to a polymer composite for armor application. More specifically, as a novel method to provide efficient fiber/matrix interface for impact energy dissipation. Therefore, the objective of this work is, for the first time, to investigate the ballistic performance of 30 vol% graphene oxide coated curaua fiber (GOCF) reinforced epoxy composite, as a MAS second layer against the treat of level III [42] high velocity projectile. In addition to the comparison of GOCF with both non-coated 30 vol% CF epoxy composite and same thickness Kevlar™, this work also investigates the integrity condition of these composites.

### **2. Materials and Methods**

Curaua fibers, shown in Figure 1a, were supplied by the University of Pará (UFPA), Belém, Brazil. The polymer used as matrix was a commercially available epoxy resin, bisphenol A diglycidyl ether type (DGEBA), hardened with triethylene tetramine (TETA), using the stoichiometric ratio of 13 parts of hardener per 100 parts of resin, fabricated by Dow Chemical, São Paulo Brazil, and distributed by Resinpoxy Ltda (Rio de Janeiro, Rio de Janeiro).

Curaua fibers were used in two main conditions, namely: as-received, non-coated fibers (CF), and graphene oxide coated fibers (GOCF). Initially the as-received fibers were subjected to a mechanical treatment using a hard bristle brush for cleaning, separation, and fiber alignment. Then fibers were cut into 150 mm in length and placed in an oven at 80 ◦C for 24 h until the fiber weight remained stable. This corresponds to the as-received CF used to produce plain composite plates.

**Figure 1.** General macroscopic aspect of curaua fibers: (**a**) curaua fibers (CF); (**b**) graphene oxide coated fibers (GOCF); (**c**) their 30 vol% epoxy composites.

The GO used in this work was produced by the Hummers Offeman method, modified by Rourke et al. [47]. The CFs, that have already passed the brush and drying stages, were then immersed in a 0.56 mg/mL GO solution corresponding to 0.1% of weight of the fiber and kept under agitation for 30 min in a universal mechanical shaker, in order to guarantee and optimize the contact of the GO with the fiber. Thereafter the CF soaked with GO dispersion were placed in an oven at 80 ◦C for 24 h, obtaining at the end the GOCF. Raman spectroscopy analysis was conducted in a model NTEGRA Spectra equipment to certify the existence of GO layers on the fiber surfaces.

To fabricate the composite plates, a metal mold with dimensions of 150 × 120 × 12 mm was used. The plates were processed in a SKAY hydraulic press by applying a load of 5 tons for 24 h. For the CF, the density of 0.92 g/cm<sup>3</sup> [31] was used as the initial reference and 1.11 g/cm3 for the epoxy resin [35]. The percentages of both CF and GOCF studied in this work was 30 vol%. Figure 1 shows the general macroscopic aspect of (a) CF, (b) GOCF, (c) and their corresponding epoxy composites.

Interfacial shear strength tests were performed to investigate the influence of GO coating onto curaua fiber in curaua-epoxy composites. For this, the method described by Kelly and Tyson [51] was used. The measured parameters were the critical length and the interfacial shear strength. Tensile tests of the individual fiber were carried out according to ASTM D 3822-01 standard [52]. The test used a support (frame) made of paper and plaster, in order to keep the fiber stretched and firmly attached to facilitate the positioning in the grips of the model 3365 Instron equipment. A 25 KN load cell and a strain rate of 5 mm/min were used to perform each individual fiber specimen tensile test. Ten specimens for each test condition were used for both CF and GOCF, with a gage length of 40 mm. The fiber diameter was measured by an optical microscope Olympus BX53M. Before starting the test, the paper is cut to avoid interference in the tensile results.

Ballistic tests were carried out to investigate both CF and GOFC composites capacity of dissipating kinetic energy of a high velocity projectile in a MAS. The MAS used in this work consist, of a front layer of ceramic, an intermediate layer made from both the CF and GOCF epoxy composites, Figure 2. The MAS is placed over a 50 mm thick clay witness (CORFIX™), which has a similar consistency as a human body. The ballistic test system is illustrated in Figure 3. The objective is to obtain the measurement of the trauma, also known as backface signature (indentation) caused by the impact of the 7.62 mm caliber ammunition on the MAS target. According to the NIJ 0101.04 standard [42] a ballistic armor will be effective if the indentation caused in the clay witness is equal to or less than 44 mm. Measurements were performed with a Q4X Banner digital laser sensor. The tests were carried out at the Brazilian Army Assessment Center (CAEx), Rio de Janeiro.

**Figure 2.** Multilayer armor system (MAS) mounted: (**a**) MAS with CF composite and (**b**) MAS with GOCF composite.

**Figure 3.** System used for ballistic tests: (**a**) Shooting support frame filled with clay witness; (**b**) MAS target ahead of the clay witness; (**c**) scheme of the system used for ballistic tests [42].

Microscopic analyses of the curaua fibers and fractured surface of the investigated composites were performed by scanning electron microscopy (SEM) in a model Quanta FEG 250 Fei microscope operating with secondary electrons between 5 and 10 KV. The energy dispersive spectroscopy (EDS) analyses were performed using a Bruker Nano GmbH XFlash 630M detector.

The FTIR technique was used to investigate the possible influences of GO on the functional groups of the curaua fibers, in an IR-Prestige-21 model spectrometer from Shimadzu, using the transmittance method with the KBr insert technique. For all samples, the same mass quantities of 2 mg of fiber and 110 mg of KBr were used.

For the analysis by thermogravimetry (TGA), the curauá fibers in CF, GOCF, and its composites were comminuted and placed in aluminum crucible of the TA Instruments, model Q 500 analyzer. Samples were subjected to a heating rate of 10◦/min, starting at 30 ◦C up to 700 ◦C.

The thickness estimation of GO coating was obtained by atomic force microscope in a model Park systems XE7 atomic Force Microscope.

### **3. Results and Discussions**

The Raman spectra of GO is shown in Figure 4. The intensity ratio of the D and G bands (ID/IG) revealed structural defects and the indication of disorder. The (ID/IG) ratio was calculated as 1.032:1, in accordance with previous authors [47,48]. Besides, a broad and shifted to higher wavenumber of 2D band was seen at 2720 cm−<sup>1</sup> for GO in Figure 4. 2D band can be used to determine the layers of graphene (monolayer, double layer or multilayer) as it is highly sensitive to stacking of graphene layers. Thus, the location of 2D band confirms that the produced GO was multilayer. A monolayer graphene is normally observed at 2679 cm−<sup>1</sup> from the spectra. In addition, the shifted location of 2D band, because of the presence of oxygen-containing functional groups, prevents the graphene layer to stack [49].

**Figure 4.** Raman spectra of GO colloid solution.

The main absorption bands of the CF fiber spectrum can be seen as: 3379 cm<sup>−</sup>1, which is related to the elongation of OH groups present in cellulose and water. The 2916 cm−<sup>1</sup> band can be attributed to the symmetrical and asymmetrical stretching (C–H) of the aliphatic chain, 1736 cm−<sup>1</sup> corresponding to the acid elongation vibration (C=O); 1430 cm−<sup>1</sup> (aliphatic C–H vibration) and 1110 cm−<sup>1</sup> from the elongation vibration of the ether groups. Other bands refer to the existence of high content of oxygen functional groups on GOCF surface, such as (–C–O–C) and (–C–OOH) [53]. Chemical treatments or modifications of major fiber surface groups (–OH) can be very valuable in detecting and confirming the type of new bond established on the fiber surface and the interaction with the polymer in the case of fiber reinforced polymers [31].

With GO coating, even at low concentrations, several changes in the spectra can be seen in Figure 5. The relative intensities between some bands have changed, suggesting that the GO molecule may have linked to the functional groups such as those mentioned above, reducing almost all intensities of the spectrum. In addition, the absorption band at 1649 cm−<sup>1</sup> may refer to vibrations of the present GO skeletal ring [54].

**Figure 5.** FTIR spectrum of CF and GOCF fibers.

The light band that can be seen at 1560 cm−<sup>1</sup> can be attributed to the vibrations of benzene rings present in GO [55]. In addition, with the cure of GO curaua fibers, the absorption bands at 833 cm−<sup>1</sup> (C–H out of plane for *p*-hydroxyphenyl units) reduced the intensity [56], suggesting that the GO caused changes in the CF fiber functional groups, such as the hydroxyl and carboxyl groups of GO sheets react with the hydroxyl groups of CF, resulting in better wettability between CF and epoxy matrix [49].

The onset of the degradation step was observed at approximately 64 to 150 ◦C in both CF and GOCF. This effect may indicate the evaporation of moisture absorbed by the fibers. The main mass degradation step was observed starting at 293 ◦C for CF and 300 ◦C for GOCF fibers as can be seen in Figure 6. According to some studies, this indicates the stages of hemicellulose, cellulose, and lignin degradation, respectively [33,57–59]. The residue generated by CF fibers was 15% and by GOCF it was 14%.

In the differential thermal analysis (DTA) curves as shown in Figure 7, for the CF fibers, three stages were observed: the first one was between 250 and 300 ◦C, referring to the decomposition of the hemicellulose, with maximum degradation rate at 272 ◦C. The second process occurred between 293 and 350 ◦C, with a maximum degradation rate around 327 ◦C, which may be related to decomposition of cellulose. Lignin decomposition occurred in the third stage, between 400 and 450 ◦C, with a maximum degradation rate of around 422 ◦C. However, a distinct behavior was presented by the GOCF fibers. Their degradation was shifted to higher temperatures and this effect may indicate an increase in the thermal stability of the fibers [33,57–59].

**Figure 6.** Thermogravimetry analysis (TGA) curves of CF and GOCF fibers.

**Figure 7.** DTA curves of CF and GOCF fibers.

The degradation ratio of the different fibers, CF and GOCF, in Figures 6 and 7 indicate that by the presence of graphene oxide increases the thermal stability of the temperature at 7 ◦C, with the onset around 300 ◦C. The effect may be due to the formation of an insulation to heat propagation by the GO coating which retarded degradation and improved thermal stability [50].

Figure 8 shows SEM images of both curaua fibers investigated: (a) as-received non-coated (CF) and (b) graphene oxide coated (GOCF). Average diameter measurements conducted in 10 fibers for each case revealed values of 54.2 ± 14.3 μm for the CF and 51.1 ± 12.0 μm for the GOCF. As expected, these values are practically the same within the standard deviation. This indicates that the graphene oxide coating did not affect the fiber diameter. In fact, the GO coating was estimated to be approximately 10 nm. This would correspond to a negligible increase of less than 10−<sup>3</sup> vol% in the composite volume fraction of curaua fiber. Since the thinner GO coating cannot affect the curaua fiber strength.

With higher magnification, Figure 9 illustrates the different surface aspects of CF and GOCF. One should notice the uniform smoother surface of the GOFC, Figure 9b, because of the graphene oxide coating as compared with the rougher CF surface in Figure 9a. The GO sheets, prepared by the modified Hummers method, form a stable and homogeneous suspension and exhibit a typical transparent wavy aspect, when coated on the fibers as shown in Figure 9b.

**Figure 8.** Scanning electron microscopy (SEM) micrographs of both investigated curaua fibers: (**a**) non-coated CF; (**b**) GOCF.

**Figure 9.** SEM surface images of both fibers: (**a**) CF; (**b**) GOCF.

With higher magnification, one notes that the CF fiber is not very stable under the electron beam, with cracks opening on its surface as can be seen in the Figure 9a and indicated by a white arrow. On the other hand, GOCF fiber is more thermally stable, not reacting with the heat generated by the electron beam during the image acquisition, which corroborates the TGA results.

Through the EDS analysis in Figure 10 it was possible to identify the elements present on the surface of both CF and GOCF fibers. For the CF, only carbon and oxygen were identified, the other peaks of the spectrum refer to the copper used in the covering of the fibers. For GOCF fibers, besides carbon and oxygen, it was identified phosphorus and sulfur, which are residues of the reagents used for the production of GO. It can be noted that for CF fibers the C/O ratio is 0.91 decreasing to 0.14 for GOCF fibers, due to the presence of oxygen-containing functional groups in GO [50].

**Figure 10.** EDS pattern of CF and GOCF fibers: (**a**) CF; (**b**) GOCF.

Table 1 presents tensile test results for both curaua fibers, CF and GOCF, with regard to the ultimate stress (σ*f*), total strain (ε), and Young's modulus (**E**). Regarding this table, it is important to mention that the values obtained for these properties agree with those reported in the literature [29,31].


**Table 1.** Mechanical properties of CF and GOCF.

One may infer from the results in Table 1 that the GO coating caused an increase of 47.8% in the Young's modulus of the fibers, corroborating to other authors [49,50]. On the other hand, in maximum tension there was a 71.9% reduction showing a more brittle but more rigid behavior of GOCF compared to CF, possibly because of the relatively low amount of GO used, forming a very thin film on the surface. In addition, it will be shown that, as reinforcement of epoxy matrix composite, this coating is associated with relevant differences in terms of fiber/matrix adherence.

Figure 11 shows the pullout curves, based on the Kelly and Tyson method [51], the curve has three levels, corresponding to the failure mechanisms that occurs in the composite, the first one, for short embedded lengths, refers to the level where only the fiber pullout occurs. In the second stage, there are pullout and fiber rupture; at this stage the length of the fiber in the composite has already reached but not exceeded the critical length. However, when the critical length is exceeded, as the case of the third stage, the failure mechanism of the composite is only by rupture of the fibers, i.e., there are no longer the pullout mechanism. Thus, the critical length for the system fiber/matrix is defined by the maximum value associated with the first stage of the curve [30]. The value of the fiber critical length was calculated as *lc* = 2 mm for the CF/epoxy *lc* = 1 mm for the GOCF/epoxy. These values are much lower than that of *lc* = 10.2 mm, reported for curaua fiber/polyester [30]. The GOCF/epoxy critical length is sensibly lower than non-coated CF/epoxy. Consequently, the interfacial shear strength of the GOCF/epoxy, τ*<sup>c</sup>* = 27.5 MPa is more than 50% higher than that of the CF/epoxy, τ*<sup>c</sup>* = 18.2 MPa. One may infer that for the same embedded length the CF fiber pullout voltage is greater than GOCF fiber, however, this behavior is due to the fact that the GOCF composite is now a new system with a new fiber/matrix interface [30]. Therefore, for each system, there is a strength and a certain critical length. As the fibers had their tensile strength affected by the GO coating, it is expected that the fiber/matrix system strength also presents similar behavior.

**Figure 11.** Pullout stress of both curaua fibers, CF and GOCF, versus epoxy embedded length curves.

In the present work, for the first time, ballistic tests were carried out to measure the trauma on the witness clay in MAS target with a second layer epoxy matrix composite reinforced with 30 vol% of curaua fiber both CF and GOCF. In none of the MAS tested, there was complete perforation of the 7.62 mm projectile and the indentation of the clay witness was less than 44 mm, a value considered to be non-lethal to humans by the standard [42]. The results obtained are presented in the Table 2 and visualized in Figure 8. They were also compared with other MAS using distinct fibers, as well as with a same thickness laminate of Kevlar™, as a second layer. The limit value established by the standard is shown as an upper dashed horizontal line, in Figure 12. These results were found to be in good agreement with other authors [4,15] and relatively better than those by Braga et al. [7].

In Figure 12, one should note a slight increase in the value of the indentation in the clay witness caused by the 7.62 mm projectile impact against a MAS target with GOCF epoxy composite as a second layer. Figure 13 illustrates the aspect of both MASs, with CF and GOCF composites, before and after the ballistic test. The integrity, an essential factor for practical applications, is shown to be better than the GOCF in comparison to MAS with CF epoxy composite as a second layer. Indeed, in this latter, the plate fractured into two large pieces as can be seen in Figure 13b. By contrast, MAS target with GOCF composite remained relatively intact in Figure 13d.

**Table 2.** Depth of indentation of MAS with natural fibers composites and same thickness Kevlar™ for comparison.


PW—Present work.

**Figure 12.** Depth indentation in clay witness of the reinforced composites with 30 vol%.

**Figure 13.** View of MAS target before (**a**,**c**) and after (**b**,**d**) the ballistic test: with second layer of (**a**,**b**) 30 vol% CF; (**c**,**d**) 30 vol% GOCF.

The smaller hexagonal ceramic tiles, front MAS layer in Figure 13a,c are completely destroyed, Figure 13b,d upon the projectile impact. In an actual armor vest, these tiles compose a mosaic to allow multiple shootings in which a single tile is hit at a time without compromising the armor protection. Figure 14 shows by SEM the ruptured surface of a tile ceramic totally destroyed. This rupture occurs by intergranular fracture absorbing most of the kinetic energy of the projectile. The magnified image in Figure 14b, displays in detail an intergranular microcrack associated with this mechanism of fracture, similar to what was verified by other authors [15,34].

**Figure 14.** Surface of fracture of the ceramic tablets: (**a**) 3000×; (**b**) 10,000×.

Another important participation of the composite plate as MAS second layer is the capture of ceramic fragments resulting from the shattered front ceramic, Figure 14, which corresponds to a significant amount of the absorbed impact energy [60]. Figure 15 illustrates the capture of ceramic fragments by curaua fibrils that compose each curaua fiber in the epoxy composite. In this figure it is important to note not only the extensive incrustation of microfragments covering the fibrils but also effective fibrils separation. Indeed, as shown in Figures 15 and 16 like most LNFs a curaua fiber is composed of well-adhered fibrils that split apart when subjected to an applied stress [29]. The shock wave resulting from the projectile impact in the present ballistic tests, Figure 3, in addition to complete shatter the front ceramic, Figure 14, also caused separation of fibrils clearly shown in Figure 15. Therefore, for the first time, it is reported a whole view of the mechanisms responsible for dissipating the remaining energy, after the projectile impact against the front ceramic, by the curaua fiber composite as MAS second layer. The indentation results in Table 2, indicate that these mechanisms are responsible for a ballistic performance comparable to Kevlar™ laminate, which is a much stronger material. While the Kevlar™ mechanisms of energy absorption, as MAS second layer, is basically the capture of fragments [60], the curaua fiber composite is associated with several mechanisms with distinct participation of the GO coating. The combination of the following mechanisms makes both CF and GOCF epoxy composites in Table 2 as effective as Kevlar™.

**Figure 15.** Curaua fiber covered with ceramic fragments.

**Figure 16.** Fiber breaking of the GOCF composite fracture surfaces.

Capture of fragments, Figure 15, the same mechanism first shown in Kevlar™ [26] and later reported for curaua fiber [11,17] and non-woven curaua fabric [7] polymer composites. Apparently, this capture of fragments is not affected by the GO coating.

Fibrils separation, also illustrated in Figure 16, is a specific mechanism for stress-subjected curaua fibers [29], which contributes to dissipate energy by generating free surface area between fibrils. Observed evidences suggest that GO coating makes difficult the fibril separation and has, comparatively, a reduced dissipated energy. This separation in plain curaua fibers (CFs) might disclose individual nano and micro cellulose chains with special behavior [22–27].

Fiber pullout shown in Figure 17 in which a hole left in one site of the fracture surface was caused by a curaua fiber pullout. The insert with higher magnification revels a remaining attached fibril separated from the pulled fiber. In this case, energy is dissipated by the created hole/pulled-out fiber-free surface. No evidence of pullout was found in the GOCF composites, which also indicates a reduced impact energy absorption.

**Figure 17.** Fiber pullout of the CF composites.

Composite delamination, Figure 13b, which is a macro mechanism of energy dissipation involving the creation of relatively large free surface area associated with the extensive separation between curaua fiber/epoxy matrix. As aforementioned, delamination impairs the integrity of the 30 vol% CF composites despite the dissipated impact energy. In contrast, delamination is not effective in the

30 vol% GOCF. In this case, integrity is maintained as required by the standard for testing armor vests [42].

Fiber breaking, depicted in Figure 16, is a general mechanism common to natural and synthetic fibers, including the aramid fibers in Kevlar™ [61]. In principle, fiber breaking is an alternative to its pullout. In other words, a matrix well-adhered fiber will break instead of pulled-out. This is the case of GOCF composites in which the graphene oxide coating, Figure 9b, is expected to improve the curaua fiber adhesion to the epoxy matrix. Therefore, no pullout occurs in the GOCF fibers that comparatively dissipates more energy by breaking. It is interesting to observe in Figure 16 the rupture of an intact as well as a fibrils split curaua fibers, both indicated by corresponding arrows.

Matrix rupture exemplified in Figure 18 by a flat epoxy broken surface (right side) around a well-adhered GOCF fiber (left side). This is a specific mechanism for brittle polymer composites that undergo extensive matrix rupture upon a ballistic impact. A significant amount of energy is dissipated but enough well-adhered fibers, like in the present case of 30 vol% of GOCF, is important to avoid loss of integrity as shown in Figure 13.

**Figure 18.** Matrix rupture of the GOCF Composite.

As a final remark, it is worth reminding that the combination of energy dissipation mechanisms guarantees to a 30 vol% curaua fiber (plain or graphene oxide coated) reinforced epoxy composite as MAS second layer, an acceptable ballistic performance, Table 2, similar to that of a Kevlar™ laminate with same thickness. This performance, given by the standard backface signature less than 44 mm [42], is slightly superior in the GOCF composites, Table 2 and Figure 12, owing to the better fiber/matrix adhesion provided by the GO coating, in some of the aforementioned mechanisms. On the other hand, this better adhesion supports the 30 vol% GOCF integrity, which is essential for MAS in armor vest.

### **4. Conclusions**


**Author Contributions:** All authors contributed equally to this manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors thank the support to this investigation by the Brazilian agencies: CNPq, CAPES and FAPERJ; and UFPA for supplying the mallow fibers.

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

### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
