Liquid Composite Molding of Multiphase
Composites Using Resin with
Nanofibrillated Cellulose
Distribution of Particles and Effect on Composite Properties
Mikael Westin
Mechanical Engineering, masters level
2016
Luleå University of Technology
Acknowledgement
I acknowledge the financial support of the European Commission under INCOM EC FP7, Grant agreement No: 608746.
I would like to thank Dr. Yvonne Aitomäki for patiently supervising the work and supplying critical advice during the thesis work. My gratitude goes out to my examiner Prof. Kristiina Oksman for giving me the opportunity to do my master thesis at her research group and thereby giving me an opportunity to deepen my knowledge within the field of nanocomposites. Finally, my greatest gratitude to each of my colleagues and family who helped me out during the thesis work.
Abstract
Two objectives have served as the basis for the present work. The first objective was to experimentally verify the output of a Monte Carlo-based model on particle deposition in porous structures (i.e. fiber preform) during the resin transfer molding (RTM) process. This model is expected to have a positive impact on process optimization and reduce costs by enabling theoretical prediction of how particles are deposited when nanoparticle filled resin is used in RTM. The second objective was to produce a multiphase composite where the epoxy (EP) matrix makes use of the mechanical enhancement that cellulose nanofibres (CNF) can impart when incorporated into the liquid phase of the used polymer matrix.
The study was conducted by production of the CNF-filled EP matrix, and by using this to produce multiphase composites out of glass- as well as carbon fibers. The RTM was carried out in collaboration with CSI Composite Solutions and Innovations Oy (Vilppula, Finland). Characterization techniques, including Raman spectroscopy, optical and electron microscopy were used to investigate the microstructure and for assessment of the CNF distribution in the produced composites. These observations were qualitatively compared with the output from the proposed model to evaluate its applicability. EP/CNF nanocomposites (i.e. the consolidated resin) were evaluated by tensile test to investigate the influence of CNF on the mechanical properties of epoxy. Three-point bending tests (ISO 14125) was performed on the multiphase composite to evaluate the impact of CNF-inclusion in the matrix.
Obtained results indicate that the model is consistent with the process by which the CNF are deposited in RTM, as both the model and experiment show that the CNF are accumulated in the upper layers (injection side) of the preform. However, work remains to be done for the model to fully comply with specific aspects of the used reinforcement in RTM (e.g. pore size and geometry of the used fibre reinforcement), and thus predict the correct deposition profile and penetration depth of the CNF.
CONTENTS
ACKNOWLEDGEMENT ...
IIABSTRACT ...
IVLIST OF FIGURES ...
VIIILIST OF TABLES ...
IXLIST OF ABBREVIATIONS ...
X1. INTRODUCTION ... 1
1.1. Introduction to nanoparticles and CNF ... 2
1.1.1. Mechanical extraction of CNF ... 3
1.2. Dispersion of nanoparticles ... 5
1.3. Influence of nanoparticles on consolidated resin -nanocomposites ... 8
1.4. Multiphase composites... 9
1.5. Models for particle flow in RTM or porous media ... 11
1.5.1. A scattering model for estimation of particle flow in RTM ... 14
1.6. Summary of introduction and scope of work ... 16
1.7. Objectives ... 17
2. EXPERIMENTAL ... 18
2.1. Materials ... 18
2.1.1. Cellulose nanofibers ... 18
2.1.2. Epoxy resin... 18
2.1.3. CNF loaded epoxy ... 19
2.1.4. Fiber reinforcement ... 21
2.2. Composite fabrication ... 22
2.2.1. EP/CNF nanocomposites ... 22
2.2.2. Fiber reinforced EP/CNF ... 23
2.2.2.1. Glass fiber reinforced EP/CNF ... 23
2.3. Material characterization ... 25
2.3.1. Bending tests ... 25
2.3.2. Tensile tests ... 26
2.3.3. Raman spectroscopy ... 27
2.3.4. Optical microscopy ... 27
2.3.5. Scanning electron microscopy (SEM) ... 27
3. RESULTS AND DISCUSSION ... 28
3.1. Characteristics of prepared CNF/EP resin and nanocomposites ... 28
3.1.1. Viscosity of prepared resins ... 28
3.1.2. CNF dispersion in prepared resin ... 29
3.1.3. Characterization of nanocomposites ... 30
3.2. Characteristics of fiber reinforced EP/CNF ... 34
3.2.1. Raman spectroscopy ... 34
3.2.1. Optical microscopy and model comparison ... 35
3.2.1. Flexural properties ... 37
4. CONCLUSIONS ... 39
5. FURTHER WORK ... 39
APPENDIX A – CHEMICAL COMPOSITION OF PRIME 20LV/ULV ... 41
APPENDIX B– ACETONE EVAPORATION DURING EP/CNF PREPARATION ... 42
APPENDIX C – TUKEYS TEST FOR 0.5 WT% CNF AND EP ... 43
APPENDIX D – DATA FROM ALTERNATIVE CURING SCHEME ... 45
APPENDIX E – TUKEYS TEST ON FLEXURAL PROPERTIES ... 49
List of Figures
Figure 1: Schematic illustration of common nano-objects p.2
Figure 2: SEM images of pulp subjected to different passes through a disk grinder p.4
Figure 3: Fibre diameter before and after ultrafine grinding p.4
Figure 4: TEM images of nanoparticle-resin systems p.5
Figure 5: Viscosity with respect to nanoparticle concentration p.7
Figure 6: Model representation of the fiber network and injected particles p.14
Figure 7: Effect of velocity on particle flow p.15
Figure 8: Average fibre diameter with respect to grinding time p.18
Figure 9: Chemical structure of epoxy resin components p.19
Figure 10: Schematic illustration of solvent exchange procedure p.20
Figure 11: Water-acetone solvent exchange using stained color fibre p.20
Figure 12: Schematic illustration of acetone evaporation setup p.21
Figure 13: Images of used fibre reinforcement p.22
Figure 14: Mixing and casting of nanocomposites p.22
Figure 15: Schematic illustration of preform/composite reference axes p.23
Figure 16: Hydraulic press with enclosed mold p.24
Figure 17: Flow layer and RTM mold p.24
Figure 18: Setup and sample for Three-point bending test p.26
Figure 19: Setup and samples for tensile test p.26
Figure 20: Viscosity of epoxy resin with different concentrations of CNF p.28
Figure 21: Micrographs of CNF suspensions p.30
Figure 22: Graphical representation of tensile data p.31
Figure 23: Micrographs of tensile fracture surfaces p.33
Figure 24: Normalized intensities versus Raman shift p.34
Figure 25: Micrographs of a multiphase composite with stained fibres p.36
Figure 26: Flexural strength of GF/EP/CNF/carmine composite p.38
Figure 27: Flexural modulus of GF/EP/CNF/carmine composites p.38
Figure 28: Acetone weight versus evaporation time p.42
Figure 29: Acetone weight versus evaporation time, normalized p.42
Figure 30: Minitab output - nanocomposites, 0.5 wt% CNF - moduli comparison p.43 Figure 31: Minitab output - nanocomposites, 0.5 wt% CNF - strength comparison p.44 Figure 32: Minitab output - nanocomposites, 0.5 wt% CNF – strain comparison p.44
Figure 33: Stress-strain curves - nanocomposites, 1 wt% CNF p.46
Figure 34: Minitab output - nanocomposites, 1 wt% CNF - moduli comparison p.46 Figure 35: Minitab output - nanocomposites, 1 wt% CNF - strength comparison p.47 Figure 36: Minitab output - nanocomposites, 1 wt% CNF – strain comparison p.47
Figure 37: Minitab output – Epoxy/acetone - moduli comparison p.48
Figure 38: Minitab output - Epoxy/acetone - strength comparison p.48
Figure 39: Minitab output - Epoxy/acetone – strain comparison p.49
Figure 40: Minitab output – Multiphase composites – flexural strength p.50
Figure 41: ANOVA CI plot: Strength p.50
Figure 42: Minitab output – Multiphase composites – flexural moduli p.51
List of Tables
Table 1: Values on specific surface area for different type of nanoparticles p.2
Table 2: Studies on multiphase composites p.9
Table 3: Studies (models) on particle filtration in porous media p.12
Table 4: Manufacturer data on cured epoxy (Prime 20ULV) p.19
Table 5: Designations for produced epoxy and nanocomposites (0.5 wt% CNF) p.23 Table 6: Viscosity of epoxy resin with different concentrations of CNF p.28 Table 7: Tensile properties of fabricated nanocomposites (0.5 wt% CNF) p.30
Table 8: Flexural properties of fabricated multiphase composites p.37
Table 9: Chemical composition of epoxy (Prime 20LV) resin p.41
Table 10: Chemical composition of hardener (Prime 20ULV) resin p.41
Table 11: Designations for epoxy and nanocomposites (1 wt% CNF) p.45
List of abbreviations
ACNT: Amino functionalized carbon nanotubes
AFM: Atomic-force microscopy or atomic force microscope
CdSe: Cadmium selenide
CF: Carbon fibre
CNC: Cellulose nanocrystal
CNF: Cellulose nanofibril or cellulose nanofibre CNT: Carbon nanotube (single-walled)
CVD: Chemical vapor deposition
DCB: Double cantilever beam
DMA: Dynamic mechanical analysis
DSC: Differential scanning calorimetry
EP: Epoxy
GF: Glass fibre
HPH: High-pressure homogenizer
ILSS: Interlaminar shear strength
LCM: Liquid composite molding
LRTM: Light resin transfer molding MFC: Microfibrillated cellulose
MWCNT: Multi-walled carbon nanotube
Nd:YAG: Neodymium-doped yttrium aluminum garnet
OMWNT: Oxidized multi-walled carbon nanotube OPEFB: Oil palm empty fruit bunch fibres
PCL: Polycaprolactone
PCNT: Pristine carbon nanotubes
PLA: Poly(lactic acid)
PSM: Poly(styrene-alt-maleic anhydride)
PU: Polyurethane
RTM: Resin transfer molding
SCA: Svenska cellulose aktiebolaget
SEM: Scanning electron microscope or Scanning electron microscopy
SSA: Specific surface area
SWCNT: Single-walled carbon nanotube
TEM: Transmission electron microscope or Transmission electron microscopy VARIM: Vacuum assisted resin infusion molding
1. Introduction
Modification of polymers through incorporation of nano- and/or microparticles has gained increased attention the past two decades (Marouf et al., 2016). Several studies have shown that by adding a second phase to the liquid state of a thermoset one may improve mechanical properties of the solid polymer such as storage modulus (Messersmith and Giannelis, 1994), shear modulus (Capadona et al., 2007), flexural modulus (Chatterjee et al., 2012) and tensile modulus and strength (Benhamou et al. (2015), to mention a few. While nanoparticles have the potential to enhance the mechanical performance of a given polymer, the nanoparticle type, its geometry, the weight percentage added and the choice of polymer will affect the nanoparticles influence on the mechanical performance. This is clear from mentioned studies where different nanoparticles and weight percentages were used. Among available nanoparticles, cellulose nanofibers1 (CNF) has a particularly interesting potential when combined with polymers, partly stemming from its mechanical properties and availability from low cost feedstock (Tang et al., 2015). It has even been shown that CNF may yield a negative cost if extracted from cellulosic waste material (Jonoobi et al., 2012). Upon consolidation of polymers with incorporated CNF, one obtains a nanocomposite. Research focusing on the manufacture of nanocomposites has seen a steady growth since the late 1990s (Oksman et al., 2016). In recent years, there has been an increased interest in using nanocomposites as matrix material in fiber reinforced composites manufactured through liquid composite molding (LCM), and hence creating multiphase composites with novel mechanical properties (da Costa and Skordos, 2012). Since nanoparticles can be incorporated in resin while keeping the viscosity below limiting values, these types of resin lend themselves to LCM techniques such as, e.g., resin transfer molding (RTM) or vacuum-assisted RTM (VARTM). Although LCM is a well-established technique for production of glass- or carbon fiber reinforced composites, studies on its use in conjunction with nanoparticle filled resins are fairly scarce in the literature. While there exist some models aimed at describing the behavior of nanoparticles in RTM (e.g. da Costa and Skordos, 2012; Frishfelds and Lundström, 2011; Hwang et al., 2011), additional models and trials are needed to understand and overcome the challenges that introducing nanoparticles into the LCM process brings. Tools are needed for predicting the behavior of nanoparticles in RTM so that one may be able to optimize the manufacturing process and predict properties of the final product.
Experimental and theoretical work on resins with nanoparticles in LCM necessitates an understanding of several subjects that by themselves are, or have been, subject to intense research. The following subsections provide a brief introduction to selected topics in LCM and reinforced resins, and reviews some results that have been achieved.
1.1. Introduction to nanoparticles and CNF
The term ”nanoparticle” is frequently used to denote materials having at least one dimension less than 100 nm in size (Šupová et al., 2011; Bittmann et al, 2009). These materials can take a
variety of forms and be of organic as well as inorganic origin (Marquis et al., 2011; Li et al., 2010). While nanoparticles can have complex shape, they may usually be categorized as being either plate-like, fiber-like (tube-like) or particle-like. Figure 1 provides a schematic illustration of the general class of nano-objects.
Figure 1 Schematic illustration of common nano-objects. Adapted from Marquis et al. (2011); shapes from ISO/TS 80004-2:2015(en) Nanotechnologies - Vocabulary - Part 2: Nano-objects. Note that "nanoparticle" can also denote a nano-object with all three dimensions smaller than 100 nm.
Nanoparticles often display high specific surface area (SSA; see Table 1), which facilitates interaction with the phase into which they are added (Šupová et al., 2011). Incorporated in a
polymer, the large surface area of nanoparticles may act as an interface for stress-transfer, enhancing the mechanical performance relative to the neat polymer. However, the large surface area may also give rise to undesired effects such viscosity increase in the monomer-particle suspension and/or agglomeration of particles, since an increase in surface area also facilitates interaction between the particles (Fiedler et al., 2006).
Table 1 Example of reported values on specific surface area (SSA) for different type
of nanoparticles. SSA values for conventional fiber reinforcement (E-glass) has also been included for comparison. All SSA values were determined experimentally.
Nanoparticle
Specific
surface area Source
m2·g-1
Cellulose nanofibrils (CNF) 153-284 Sehaqui et al. (2011a)
Silica nanoparticle 200-380 Kahlo et al. (2016)
Single walled carbon nanotube 367-662 Birch et al. (2013)
Graphene like material 1052-2274 Rhee et al. (2015)
Conventional reinforcement - -
In recent times, there has been an increased interest in using cellulose based nanoparticles to reinforce polymers. With its good mechanical properties and availability from renewable feedstock, CNF can be a competitive choice in a world where environmental concerns need to be taken into account when designing novel structural components (Dufresne, 2013; Dicker et al., 2014). Pioneering work on the isolation of CNF from wood pulp was first performed by Turbak et al. (1983). By feeding pulp fibers through a high-pressure homogenizer (HPH), Turbak and co-workers obtained a gel-like material which they referred to as “microfibrillated cellulose”. While it is known that CNF comprises a size distribution of cellulose aggregates, typically including micro-sized fibers, properly fibrillated CNF has nano-scale particles as its main component (Panthapulakkal and Sain, 2012; Chinga-Carrasco, 2011). Apart from having a large SSA (see Table 1), CNF has other properties that makes it interesting as reinforcement. Example of properties include:
A modulus of approximately 143 GPa (Šturcová, 2005);
A mean strength in the interval of 1.6-6 GPa (Saito et al., 2012); A density about 1.53 g/cm3
(Nogi et al., 2009);
Hydroxyl (-OH) groups providing strong hydrogen bonding.
The mechanical properties of produced CNFs, and any polymer which display effect of CNF inclusion, will depended on undertaken processing steps during CNF extraction, e.g., pretreatment of raw material, fibrillation method and the origin of processed fibers (Takagi and Asano, 2007; Iwamoto et al., 2007). This fact motivates a brief review of CNF extraction.
1.1.1. Mechanical extraction of CNF
Currently, the field of CNF production is at a stage where patents, as well as pilot plants for CNF production, exists. However, this would not be a reality without the many works focusing on its extraction. Mechanical means of extracting CNF from lignocellulosic sources have been a key topic in several works (e.g. Turbak et al., 1983; Taniguchi and Okamura, 1998; Bruce et al., 2005; Uetani and Yano, 2010 Jonoobi et al., 2012). By exposing, e.g., kraft pulp to repeated impact and/or shear, cellulose fibers are gradually separated (i.e. fibrillated) and nanofibrils are obtained. Examples of mechanical fibrillation include the use of a high-pressure homogenizer by Bruce et al. (2005) to obtain CNF from Swede root in order to produce composite materials with different binders, and the use of a PFI mill (mechanical beating) by Sehaqui et al. (2011b)
to produce CNF composites of dual-scale structure (micro- and nanoscale). One of the early strategies for CNF extraction is the use of ultrafine grinding. In this technique, shear force is used to break down cellulose pulp to nanofibrils (Hietala and Oksman, 2014). Production of CNF through ultrafine grinding, using the Supermasscolloider (Masuko Sangyo Co. Ltd., Japan), was introduced by Taniguchi and Okamura (1998) in an article on the production of films from different sources of natural fibers subjected to grinding2. It has been shown that
ultrafine grinding is less energy demanding than other methods (e.g. homogenization or microfluidization) of obtaining CNF (Spence et al., 2011).
In general, ultrafine grinding requires the CNF source to be feed multiple times through the grinder to yield CNF of uniform fibril bundles (Hietala and Oksman, 2014). However, with specific preparations methods it is possible to obtain them with a single pass through the grinder (Abe et al., 2007). The change in fibril morphology following repeated passes through an ultrafine grinder has been studied by Iwamoto et al. (2007). It was demonstrated that ultrafine grinding rapidly gave rise to fibrils with transverse dimensions on the nanometer scale but that several passes was needed to achieve uniform fibril appearance (Fig. 2).
Figure 2 FE-SEM images of pulp subjected to different passes through a disk grinder. Left: 1 pass;
middle: 3 passes; right: 5 passes. Images taken from Iwamoto et al. (2007).
A more recent study on the diameter distribution of CNF, using ultrafine grinding, has been performed by Panthapulakkal and Sain (2012). Using UTHSCSA ImageTool (University of Texas, San Antonio, USA) on SEM and TEM images, it was found that ultrafine grinding shifted more than 60 % of the fibrils into a diameter range of 1-60 nm, with the majority of fibrils shifted into a diameter range of 20-40 nm (Fig. 3).
Figure 3 Fiber diameter before and after ultrafine grinding. Image taken from Panthapulakkal
and Sain (2011).
A similar result on diameter distribution was obtained by Jonoobi et al. (2012). Analyzing TEM images of sludge subjected to ultrafine grinding, it was found that more than 60 % of the sludge
fibrils obtained a diameter in the range of 10-40 nm following grinding. A closer look at the sludge diameters reveals that the majority of sludge fibrils had diameters in the range of 20-40 nm in this study as well.
Knowing the size distribution of produced CNF is important from a manufacturing perspective since the size of the nanoparticles is known to influence the viscosity of the liquid which they are incorporated into (Zhao et al. 2009; Shaffer et al., 1998). An important observation to make is that conventional, ultrafine grinding yields CNF in an aqueous state, this becomes particularly important when considering CNF inclusion in a polymer, since polymers, in general, are non-polar which complicates CNF inclusion (Matos Ruiz et al., 2000).
1.2. Dispersion of nanoparticles
When incorporating nanoparticles into a polymer matrix, obtaining a good dispersion of the particles has been highlighted as an essential step for efficient transfer of their mechanical and physical properties to the resulting composite (Oksman et al., 2016; Zheng et al., 2003). Ideally, good dispersion entails a uniform distribution of individual nanoparticles, well wetted by the polymer and sufficiently separated to minimize agglomeration. If insufficient wetting or particle separation is present, agglomerated nanoparticles may act as defects rather than an enhancing phase (Taylor, 2010). In practice, it is difficult to completely avoid agglomerates, and “well-dispersed” nanoparticle-matrix systems may very well include some degree of agglomeration (Müller et al., 2004), as can be seen from the images of Fig. 4.
Figure 4 TEM images of “well-dispersed” nanoparticle-resin systems. Left: Cadmium
selenide/poly(styrene-alt-maleic anhydride) nanocomposite, adapted from Liu et al. (2003). Right: Multi/polymer composite, adapted from Coiai et al. (2015).
Agglomeration and nanoparticle-matrix incompatibility are two recurring challenges when incorporating nanoparticles into a given matrix (Hooshmand et al., 2014; Eyholzer et al., 2010; Bittmann et al., 2009).
nanoparticles can be surface treated (chemically modified) or mixed with a suitable compatabiliser (Taylor, 2010). For example, Gojny et al. (2003) demonstrated that amino-functionalization of MWCNTs improved dispersion by reducing agglomeration of the nanoparticles, owing to steric and electrostatic repulsion of added surface groups. Another example of chemical modification is found in Siqueira et al. (2008), where N-Octadecyl isocyanate was used to facilitate dispersion of CNF in dichloromethane, used for further dispersion in a PCL. It was found that PCL/CNF nanocomposites could not be produced from untreated CNF due too poor dispersibility in the carrying medium (dichloromethane). However, surface treated CNF was readily dispersible and gave rise to composites with increased modulus relative the neat PCL for weight fractions of above 3 wt%.
Controlling the dispersion and obtaining a homogenous distribution is a fundamentally challenging task due to the fact that phase separation as well as nanoparticle agglomeration is energetically favorable for the nanoparticle-monomer system, since nanoparticles minimizes its surface energy by forming agglomerates (Smith and Bedrov, 2009). Nevertheless, dispersion sufficient for the resulting nanocomposite to display enhanced properties relative the neat polymer has been reported in numerous of studies. In the case of well-compatabilised nanoparticle-resin systems, low shear mixing may be sufficient for nanoparticle dispersion. For systems with a higher tendency for agglomeration, high-power dispersions methods such as calendaring, high-shear mixing, sonication or a combination of two or more techniques have successfully been used for particle dispersion (Taylor, 2010). For example, in an early work by Messersmith and Giannelis (1994), silicate-epoxy nanocomposites were prepared using sonication to disperse nanoparticles in liquid resin, and by subsequent casting of the resulting mixture. By comparing DMA results from cured, unmodified epoxy and a nanocomposite containing 4 vol% silicate, it was found that storage modulus increased about 57% (from 1.55 GPa to 2.44 at 40°C) in the glassy region and 350% (from 11 MPa to 50 MPa at 150 °C) in the rubbery region. Bondioli et al. (2005) prepared EP/SiO2 nanocomposites by dispersing different
weight fractions of silica nanoparticles using high-sheer mixing. It was found that methanol based SiO2 increased the tensile modulus relative neat EP for silica loadings above 1 wt%. In a
more recent study, PU/CNF nanocomposites were prepared by Benhamou et al. (2015) by dispersing different concentrations of CNF in PU using sonication for 20 min. It was found that tensile strength as well as tensile modulus was significantly higher than that of neat PU for CNF loadings above 5 wt%. Good dispersion using sonication has also been achieved by Fiedler et al. (2006). In this study, of carbon nanotubes, weight fractions as low as 0.1 wt% gave rise to increased fracture toughness relative neat epoxy. Also, the authors suggested that dispersion would be further facilitated by using acetone as a viscosity lowering agent for the nanoparticles.
concentrations while a steep increase was observed for concentrations beyond 0.7 vol% of CNT. The viscosity increase following this concentration was identified as an entanglement transition.
Figure 5 Experimental (○) and theoretical (—) values of viscosity with respect to nanoparticle
(CNT) concentration (Shaffer et al., 1998). Theoretical values were obtained by fitting data to the Schulz–Blaschke equation (Schulz and Blaschke, 1941).
Although lowering the viscosity for a nanoparticle-polymer system using solvents has been suggested as a possible way to facilitate dispersion of nanoparticles, care should be taken to minimize the amount of solvent, since residuals of it might have a detrimental effect on the mechanical performance of the polymer (Fiedler et al., 2006; Loos et al., 2008). Preparing epoxy with different weight fractions of residual acetone, Loos et al. (2008) observed a decrease in tensile strength and strain (relative to the neat epoxy) for residual acetone content above 10 wt%. Nevertheless, the authors pointed out that the positive gain in processing due to lowered viscosity may very well be sufficiently beneficial to balance the detrimental effects of residual solvent.
Another issue that arises in the context of nanoparticle dispersion is how to assess the quality of a performed dispersion. Since the length scale of nanoparticles is typically less than that of visible light (~400-700 nm), evaluation of the dispersion quality may not be feasible using optical microscopy only. However, optical microscopy permits one to conveniently and quickly determine whether micrometers agglomerates are present in the suspension. Characterization techniques such as TEM or AFM can readily detect nanoparticles, but may be impractical due to extensive preparation times or morphology changes due to sample preparation (Taylor, 2010). In general, several techniques are required to assess the quality of a dispersion, and each technique has its own merits and drawbacks.
1.3. Influence of nanoparticles on consolidated resin -nanocomposites
Polymer matrix nanocomposites can be defined as polymer-particle systems where the particle have at least one dimension on the nanometer scale, i.e. smaller than 100 nm (Siqueira et al., 2010). An additional requirement that might be imposed on the polymer-particle system to be categorized as a nanocomposite is that it has physical or mechanical properties not obtainable by solely one of its constituents (Oksman and Sain, 2005). Although the purpose of adding a particular nanoparticle to a polymer matrix may vary from one study to the other, their inclusion has in many cases served the purpose of overcoming drawbacks of the used matrix, e.g., reducing the brittleness of epoxy through incorporation of carbon nanotubes (Domun et al., 2015).
High strength and stiffness are usually desirable of a matrix material; achieving this by incorporation of nanocellulose in a polymer matrix has been the subject of several studies. For example, in a study by Benhamou et al. (2015), the reinforcing effect of CNF and CNC, respectively, on polyurethane was studied. Nanocomposites were prepared by adding different amounts of CNC or CNF to PU, and by consolidation through casting/evaporation. The study showed that the PU/CNF nanocomposites had significantly higher tensile strength than those prepared from CNC. By applying a number of characterization techniques, including DMA, DSC and tensile tests, and by analyzing the nanoparticle-PU interaction based on solubility parameters, it was inferred that CNF had higher thermodynamic compatibility with PU (than CNC with PU), resulting in stronger interactions, adhesion and miscibility of CNF and PU. This result was attributed to interaction with molecules on the CNF surface, leading to better interaction with the PU matrix. An important result from this study is that the higher aspect ratio and flexibility of CNF (compared to CNC) give rise to increased tensile strength, underlining the importance of nanoparticle geometry. However, in contrast to the study by Okubo et al (2009) (where PLA was used as matrix) a statistical increase in modulus and strength, respectively, were not observed for CNF fractions below 5 wt%, implying that choice of polymer matrix also plays an important role when evaluating the effect of CNF as reinforcement.
Gabr et al. (2010) observed that addition of low quantities of CNF to epoxy did not result in a statistical increase in tensile properties. This is in accordance with other studies where low levels of CNF have been used (e.g. Carvelli et al., 2016). However, other studies have shown that low concentration of CNF can indeed give rise to an increases in tensile properties of the investigated matrix (e.g. Okubo et al, 2009). To identify whether this discrepancy (with respect to tensile properties) are due to differences in used materials (e.g. CNF extraction, CNF aspect ratio distribution etc.) or used process parameters is a challenging task, due to the large variability from on study to the other.
result is interesting as it indicates that CNF nanocomposites may be feasible to use as matrix material in so called “multiphase” composites.
1.4. Multiphase composites
In recent years, nanoparticle incorporation in fiber reinforced composites has become an interesting option for, e.g., the aerospace industry due to the potential of nanoparticles to improve composite performance (da Costa and Skordos, 2012; Domingues et al., 2012). Although a majority of the studies concerned with fiber reinforced composites have utilized regular resin, with no inclusion of nanoparticles, several works have investigated their influence on resulting composites. Table 2 Provides a selection of performed studies.
Table 2 Selection of studies focused on “multiphase” composites. “Nanoparticle content” is the amount
for which the “best” properties were achieved (i.e. properties under the “Enhanced or added properties“-column).
Jawaid and Khalil (2011) describes hybrid composites as systems where a reinforcing material is incorporated in a matrix blend, or where two or more reinforcing phases are added to a single matrix (or a combination of both cases). These kind of systems have also been referred to as “multiphase” or “multi-phase” composites (Chen et al., 2015; Garg et al., 2015; Shokrieh et al, 2014). The term "multiphase" is adopted in this paper since it reflects the fact that several materials have been incorporated in the composite structure. However, other designations are in use and several studies on "multiphase" composites do not refer to these type of systems by any generic name (see Table 2).
Concerning how nanoparticles have been incorporated into multiphase composites, dispersion into the matrix (or conversely) before combining this suspension with additional reinforcement, is a typical procedure. Apart from the study by Thakre et al. (2006), all the studies mentioned in
Author(s) Denotation Nanofiller; other reinforcement Nanofiller content Matrix Manufacturing technique Enhanced or added properties
Thakre et al., 2006 - Func. SWCNT, CF 0.1 wt% Epoxy VARTM ILSS Fan et al., 2008 Hybrid OMWNT; GF 1 wt% Epoxy VARTM ILSS Karapappas et al., 2009 - MWCNTS; CF 0.5 wt% Epoxy Hand lay-up/
Vacuum bagging Fracture energy Gabr et al., 2009 - MFC; CF 2 wt% Epoxy Hand lay-up Fracture toughness Morales et al., 2010 Hybrid Carbon nanofiber; GF 2 wt% Polyester Light RTM
Flexural and tensile strength; resistivity Okubo et al., 2009 Hybrid MFC; bambo fiber 1 wt% PLA Injection
moulding Strain energy Domingues et al., 2012 - CNT; GF 0.3 wt% Epoxy VARTM Interlaminar toughenss
Shokrieh et al., 2014
Three-phase MWCNTS; CF 1 wt% Epoxy Hand lay-up
Reduction of residual stress Garg et al., 2015 Multi-phase ACNT, PCNT; GF 0.1 wt% Epoxy VARIM Tensile modulus,
flexural moduls Saba et al., 2016 Hybrid OPEFB; Kenaf fibre 3 wt% Epoxy Hand lay-up Tensile and
Table 2 have utilized dispersion of nanoparticles into the matrix before combining this with the additional reinforcement. For example, Morales et al. (2010) utilized a high speed mixer to obtain Polyester/carbon fiber suspensions of different concentrations. Obtained suspensions were not further processed before using them in LRTM. While manufactured composites showed increased tensile and flexural strength for concentrations as low as 0.5 wt% carbon nanofiber, the highest increase was found for a nanofiber load of 2 wt%. Similar, Shokrieh et al. (2014) used high a speed mixer to disperse MWCNTs in epoxy, but with the addition of sonication to reduce carbon tube aggregates.
Although matrix dispersion is frequently employed, nanoparticles have also been incorporated in multiphase composites by attaching them directly to the fiber reinforcement. For example, Thakre et al. (2006) incorporated SWNCTs by dispersing them in ethanol and spraying the nanotube-suspension onto the CF fabric before performing VRTM. Upon comparison with composites sprayed with ethanol only, an increase in interlaminar shear strength was observed. Yet, this was only observed for functionalized SWCNTs, non-treated nanotubes did not display any enhancing effect on the ILSS.
Ávila et al. (2013) incorporated MWCNTs by growing these directly onto CF fibers through the use of thermal vapor deposition (CVD). Multiphase composites obtained by combining this reinforcement with epoxy (in a laminate structure) showed an increase in strength and crack bridging relative the case without carbon nanotubes.
Evidently, different types of carbon nanotubes have been widely employed in the manufacture of multiphase composites. Another example of this is provided by Fan et al. (2007). Fan and co-workers found that the incorporation of oxidized multi-walled carbon nanotubes (OMWNTs) in the epoxy matrix improved the interlaminar shear strength (ILSS) of GF reinforced epoxy composites. Also, it was thought that the ILSS was further improved by inducing a preferential orientation of the OMWNTs in the thickness direction. This orientation was achieved using a flow layer and performing VARTM through-the-thickness of the fiber.
Fewer studies have been found where CNF is used to produce multiphase composites. However, there are some examples. In a study by Gabr et al. (2010a), the effect of CNF on the mechanical and thermal performance of carbon fiber (CF) composites was investigated. CNF was dispersed in epoxy in fractions of 0.5, 1, 2 wt%, respectively, through a solvent-exchange/evaporation procedure using ethanol. Composites were thereafter manufactured by hand lay-up of plain woven carbon fibers, using the different EP/CNF-combinations as matrices. Resulting composites had fiber volume fractions of approximately 50%. Using uniaxial tensile test, DMA and fracture toughness test (DCB, mode I) the influence of CNF on manufactured composites could be compared with that of a reference sample (0 wt% CNF. When compared with the reference sample, it was found that the tensile strengths and Young’s moduli of manufactured composites were statistically unchanged with increasing CNF content. However, double cantilever beam (DCB) tests showed a 80% increase (from 0.28 to 0.50 kJ/m2) in initiation
the addition of 2 wt% CNF. Furthermore, with an addition of 2 wt% CNF, DMA showed that the glass transition temperature was increased by approximately 12 °C. The authors suggested that the absence of increased tensile properties was a consequence of too few bonding sites being formed between the CNF and CF at investigated fractions of CNF. The increase in fracture toughness properties was attributed to CNF bringing about a toughening effect on the matrix, enhancing the energy required for crack-propagation. Increase in glass transition temperature was explained by an increase in cross-link density with increasing content of CNF. Although a statistical increase in tensile properties was not found for the EP/CNF nanocomposite (for low concentrations of CNF), the multiphase composite obtained enhanced mechanical performance. This type of behavior indicates that the mechanical performance of multiphase composites is not necessarily reflected by the nanoparticle matrix. Although experiments will provide direct results of any effect of nanoparticle incorporation, this approach might be too time consuming for new type of materials. Hence, modelling is required to be able to assess the changes that nanoparticles are subjected to combined with additional reinforcement.
1.5. Models for particle flow in RTM or porous media
Resin transfer molding is a closed mold technique, commonly used for the manufacture of fiber reinforced composites. In this manufacturing technique, liquid resin is injected into a dry fiber preform through an applied pressure gradient. Since production takes place in a closed mold, exposure to volatile gasses can be avoided. Furthermore, the RTM technique enables good control of fiber orientation and can be used to produce composites in short cycle times compared to other manufacturing techniques (Agarwal et al., 2006). The technique can yield composites with a fiber volume fraction up to approximately 60% (Hussain et al., 2006; Turner et al., 2006).
Table 3 Selected work on models dealing with particle filtration in porous media.
Year Author(s) Assumed
particle type Model type
1996 Destephen and Choi Spherical,
micro-sized
Monte-Carlo based (FORTRAN)
1999 Erdal et al. Spherical Continuum mechanics
(Finite element)
2006 Chohra et al. Spherical Continuum/random
2006 Elgafy and Lafdi Spherical Continuum mechanics
(Finite element)
2007 Elgafy and Lafdi Spherical Continuum mechanics
(Finite element)
2007 Lefevre et al. Spherical Continuum mechanics
(Finite element)
2009 Lefevre et al. Spherical Continuum mechanics
(Finite element) 2011 Frishfelds and Lundström Spherical/variable
size
Continuum mechanics (Finite element, ANSYS)
2011 Steggall-Murphy et al. Spherical Continuum mechanics s
2011 Hwang et al. spherical Continuum mechanics
(Finite element)
2012 da Costa and Skordos spherical Analytical/Numerical
2012 Lundström and Frishfelds Spherical Continuum mechanics
(Finite element, ANSYS)
2014 Sas and Erdal Spherical
Continuum mechanics (Finite element, COMSOL Multiphysics)
2014 Haji et al. spherical Continuum mechanics
(Finite element)
2015 Haji and Saouab Spherical Continuum mechanics
(Finite element)
Erdal et al. (1999), implemented a process model intended for analysis and simulation of RTM using particle filled resin and continuous-fiber preforms. Using a finite difference scheme, Darcy flow (isotropic and isothermal) was coupled to a multidimensional, transient particle filtration formulation. Using this model, 1-D and 2-D simulations, respectively, was used to relate the processing conditions to the particle-concentration distribution. According to this model, the particle filtration characteristics should be insensitive to injection flow rate and predominantly governed by the filtration coefficient (an empirically determined quantity) and the permeability. Unfortunately, the presented model was not experimentally validated in their paper, making the study of limited value for actual processing of composites. However, the model of Erdal et al. (1999) was later expanded upon and experimentally validated in studies by Levere et al. (2007; 2009).
In the first study (Levere et al., 2007), a routine to accommodate particle retention3 in the preform was added to the model. By addition of this routine, change in porosity (permeability) and hence resin particle-flow, due to particle retention, could be taken into account. The expanded model was used to simulate 1-D injection of a composite part and the simulation was supplemented experimentally by manufacturing a composite part under 1-D injection of particle filled resin. As predicted by the model of Erdal et al. (1999), the experiments verified that the particle retention strongly decreases along the part length (away from the injection point). In addition, lower concentration in the vicinity of the flow front (attributed to retention) and particle deposit near the outlet (attributed to liquid depletion) were found and predicted due to the added routine. In the later study (Levere et al., 2009), the model was further expanded to enable extraction of results with respect to position and time. Although providing good agreement with performed experiments, no data for the case of nano-sized particles were provided, only for micro-sized ones. Furthermore, results were obtained considering particle contents in the range of 21.5- 40.9 vol%.
In a work by Da Costa and Skordos (2012), an analytical and numerical model was formulated for through-the-thickness injection of nanoparticle loaded thermosets. Their analytical model is based on the 1-dimensional solution of Darcy’s problem and incorporates filtration kinetics and conservation of nanoparticles. The analytical solution takes into account the permeability as well as viscosity and porosity as a function of local nanoparticle loading. Following validation of the models through experiments work it was concluded that their analytical model is an accurate and efficient estimate for through-the-thickness infusion, while the numerical model was necessary to account for in-plane flow. It should be pointed out that the proposed model provides numerical data in the form of length (i.e. depth in composite) versus filtration for a given set of output data ,e.g., flow front, pressure, and particle concentration. Thus, in-plane infusion essentially becomes an extension of through-the thickness infusion, i.e., the resin will traverse a longer distance of material in the former case. Apparently, this kind of model does not provide a 2-D geometry of the composite.
For a model similar to the one by Erdal et al (1999), and later Levere et al. (2007; 2009), Chohra et al. (2006) added stochastic methods to enable prediction of the particle profile when performing VARTM through-the-thickness. The particle distribution when performing LCM through-the-thickness has been particularly in mind in for the model by Aitomäki et al. (2016).
1.5.1. A scattering model for estimation of particle flow in RTM
In the work by Aitomäki et al. (2016), a model for assessment of nanoparticle distribution following RTM with CNF filled resin is presented. Based on simple scattering of photons, the model (implemented in Matlab 2014b) utilizes a Monte-Carlo technique to simulate the transport of nanoparticles. The proposed model provides a 2-D representation of the fiber network to be impregnated, taking into account fiber-fiber as well as inter-bundle (channel) distances (Fig. 6).
The main purpose of the model is to provide a helpful tool that can be set to match the geometry and distribution typically obtained for CNF. This model has the potential to predict the resulting nanofiber distribution and provide the user with a quick assessment of where the nanofibers will end up following RTM with CNF filled resin.
Figure 6 Model representation of the fiber network and injected particles at different
magnifications. a) Fiber (black ellipses) bundles separated by channels; b) Individual fibers with filtered nanoparticles (blue); c) Fiber surface with attached nanoparticle. Note that the nanofiber size in a) is greatly exaggerated to enable representation of spatial distribution. Figure adapted from Aitomäki et al. (2016).
represented by ellipses). One advantage of using ellipses is that the fibers can be modified to reflect the hydrodynamic appearance of the fibers at increased fluid velocities (Fig. 7).
Figure 7 Effect of velocity on particle flow characterized by change in aspect ratio of fibers in the
network. Image reproduced from Aitomäki et al. (2016).
Inter-bundle channels are obtained by cutting of the bundles at well-defined image positions (see Fig. 7). The fiber dimensions, bundle and inter-bundle sizes can easily be varied in the program, and works as input parameters for the model together with the number of fibers. Furthermore, since the length and diameter of the simulated nanofibers are obtained from a probability density function (PDF), the particular choice of PDF works as an input parameter for the model. Currently, the model assumes a normal distribution for the nanofibers4.
At the beginning of a simulation, each nanofiber is assigned a directional vector and will subsequently move (in the z-direction) as dictated by associated vector. In addition, the initial orientation of each nanofibre is randomly chosen before the nanofibre is randomly positioned along a specified length at the upper part (injection side) of the fibre network.
The model assumes that the nanoparticles travel with the resin, which in turn is assumed to have a Darcy flow. The velocity of this flow can be written as
𝜐𝑖 = − 𝐾𝑖𝑗
𝜇 𝜕𝑝
𝜕𝑥𝑗, (1)
where (Kij) the permeability tensor, (μ) the dynamic viscosity and (𝜕𝑝/𝜕𝑥𝑗) the pressure change
at position (𝑥𝑗). For the 1-D case, assuming a constant pressure throughout the flow path (𝐿),
Eq. (1) is reduced to 𝑢𝑖𝑛𝑡= − 𝐾 𝜖𝜇 ∆𝑝 𝐿 , (2)
where (𝜖) is the porosity and (𝑢𝑖𝑛𝑡) will be the particle velocity in the z-direction (Aitomäki et
al., 2016).
From performed simulations with the model, it is indicated that used fiber volume fraction will have a dominant effect on the penetration depth of the nanofibers; although size dependent, bundle penetration reaches no more than 40 μm for homogenous nanofibers of 75 nm diameter and 3000 nm in length. It is observed that inter-bundle channels will allow the nanofibers to
penetrate deeper into the fiber structure. The model suggests that RTM should be performed through the thickness of the preform if the influence of nanoparticles are to be displayed by the resulting composite, since the majority of nanoparticles will get filtered out in the initial part of the composite (using in-plane infusion, most of the nanofibers would end up in the trimmed part of the composite).
1.6. Summary of introduction and scope of work
The literature shows that it is possible to enhance the physical and mechanical properties relative the neat polymer by incorporation of nanoparticles. Different particles, e.g., CNF, SWCNT, TiO2 to mention a few, and various polymer types have been used to achieve promising results.
In particular, CNF offers a competitive alternative among available nanoparticles due to its mechanical properties and availability from low cost feedstock.
While studies focused on multiphase composites are readily available from the literature, a large variation in preparation techniques and process parameters is observed. The variation in manufacturing will result in differences in mechanical performance of the final composite part, making direct comparison between studies challenging. However, good results (i.e. an increase in mechanical properties relative traditional, bi-phase composites) have been achieved in more than a few studies.
RTM is a common technique for consolidation of high performance composites. Models pertaining to the flow of nanoparticle filled resins in porous structures, with particular emphases on RTM, have been found, though relatively few. These models are to a large extent based on continuum mechanics and implemented in software utilizing finite elements techniques. Similar to the simple scattering model proposed by Aitomäki et al (2016), several of the works, utilizes a Darcy flow to describe the movement of nanoparticles (which will be governed by the resin flow) and predicts that particles will be filtered out at an early stage when infused in fibrous structures. Although, particles may penetrate into fiber bundles, it is predicted that particles are likely to be filtered out when encountering narrower structures, and that they preferentially will move along wider paths (e.g. inter bundle channels) in a fibrous preform.
While findings from the literature provides a promising outlook on the possibility to manufacture multiphase composites and model the behavior of particle filled resins in RTM, some questions are left unanswered; e.g., how CNF networks may enhance the mechanical structure of multiphase composites and how the particular way of incorporating nanoparticles may alter (enhance or degrade nanoparticles) the final properties of a composite. However, results from the literature that should be highlighted are as follows:
Addition of nanoparticles to any pre-polymer will most likely increase the viscosity. Hence, care should be taken to not exceed feasible processing conditions;
Multiphase composites with epoxy, glass fiber and CNF have been manufactured to display enhanced mechanical properties relative composites without nanoparticles. Still, no studies exactly matching CNF/Epoxy/GF through RTM was found;
Models for prediction of nanoparticle distribution following RTM with particle filled resins are relatively few in the literature;
From proposed model it is indicated that nanoparticles encountering a narrow porous structure will be filtered out at an early stage of the resin the flow path. Particles travelling through wider pores are most likely to travel a longer distance within a given preform.
1.7. Objectives
The primary objective of the present study is to
use experimental methods to verify a scattering model designed for prediction
of the spatial distribution of cellulose nanofibers in a composite fabricated by RTM using a CNF loaded resin.
The secondary objective is to
evaluate how the addition of the CNF affects the mechanical performance of
the resulting multiphase composite.
Two hypothesis have been adopted in this study. The first hypothesis is that CNF will accumulate in the upper part of the composite due to filtering in the fiber preform. The second hypothesis is that this accumulation of CNF will give rise to increased bending properties of the composite
2. Experimental
2.1. Materials
2.1.1. Cellulose nanofibers
Cellulose nanofibers (CNF) isolated from Birch kraft pulp (SCA, Sweden) were used in the present study. Nanofibers were obtained via ultrafine grinding by repeatedly feeding an aqueous suspension of the pulp through a Supermasscolloider (MKZA10-20J, Masuko Sangyo Co., Ltd., Japan) for 100 min. The solid content of fibers (1.3 wt%) in the resulting CNF suspension was taken as the average of five measurements with a Mettler HG53 Halogen Moisture Analyzer (Mettler Toledo, United States). Typical size distribution of fibers following ultrafine grinding are presented in Fig. 8
Figure 8 Average fiber diameter with respect to grinding time. Fiber areas were
estimated through AFM imaging using a MultiMode AFM (Bruker, USA). Data kindly provided by Noël, M. (2015).
Additional details on the isolation process of CNF using ultrafine grinding can be found in, e.g., the work by Nair et al. (2014).
2.1.2. Epoxy resin
epoxy monomers (Gurit, 2006; Pascault et al., 2002). A list of the chemical species present in Prime 20 LV, and their percentage, is provided in Appendix A.
Figure 9 The chemical structure of the main components of the used epoxy system. Left: Araldite
B: (main component of the epoxy resin). Right: 2-Methylpentanediamine (main component of the hardener). Systematic names of main components can be found in Appendix A.
The main component of the hardener is 2-Methylpentanediamine (Fig. 9, right), however, according to the manufacturer, the hardener may also contain a substantial percentage of Isophorone diamine5 (Gurit, 2010). Crosslinking using diamines usually involves polyaddition, analyzing the reaction paths for the hardening process is challenging due to the complex reaction kinetics that take place (Pascault et al., 2002). The main focus with respect to the epoxy is its properties as cured (with and without the addition of CNF). Data on the mechanical properties of the cured epoxy (as provided by the manufacturer) is provided in Table 4.
Table 4 Mechanical properties of the cured epoxy (24 h at 21°C
+ 16 h at 50°C). Data reproduced from (Gurit, 2015)
Tensile strength 71.2 MPa
Tensile modulus 2.98 GPa
Strain to failure 6.28 %
Cured density 1.14 g·cm-3
2.1.3. CNF loaded epoxy
2.1.3.1. Solvent exchange procedure
The water content of the prepared fiber suspensions was reduced and replaced with acetone before any dispersion of CNF in epoxy resin. In this study the generic solvent exchange procedure is as follows: 1200 ml of aqueous CNF was divided into six centrifuge bottles and centrifuged at 8000 rpm, at an average radius of 86 mm, for 10 minutes using an Avanti J-25I
centrifuge6 (Beckman Coulter Inc., USA). Carefully pouring out the supernatant (i.e., water with less than 0.1 wt% of suspended nanofibers) of each bottle and replacing this volume by an equal amount of technical grade acetone (WVR international S.A.S, France), the CNF was redispersed by using an Ultra-Turrax disperser (T25 digital ULTRA-TURRAX®, IKA-Werke GmbH & Co) for 20 min at 1.5·104 rpm. The centrifugation-redispersion procedure was repeated five
times after the initial centrifugation resulting in a suspension with minimal amounts of residual water. A schematic illustration of the procedure is provided in Fig. 10.
Figure 10 Schematic illustration of the generic water-acetone solvent exchange procedure. The
“×6” indicate that the process was performed using six bottles in each centrifugation step (corresponding to the maximum number of bottles that the rotor could accommodate). Redispersion-centrifugation was performed 5+1 times (as described in the above text).
In addition to preparation of acetone/CNF suspensions, stained fibers were also prepared by adding a coloring agent (Carmine) to 1200 ml of aqueous CNF. The stained nanofibers were prepared according to the generic solvent exchange procedure, with the addition that Carmine was mixed with the aqueous CNF before the initial centrifugation step. This mixing was done for 20 min at 1.5·104 rpm (Fig. 11).
Figure 11 Water-acetone solvent exchange using with carmine stained fibers.
When preparing acetone/CNF/carmine suspension, each redispersion-centrifugation step gradually removed non-bound carmine, resulting in almost transparent supernatant following the last centrifugation step. Following the last centrifugation step, prepared suspensions were all diluted to be in the range of 1.3-1.5 wt% CNF.
2.1.3.2. Dispersion of CNF in epoxy
CNF loaded epoxy resin of different volumes and with different CNF content was prepared using stained as well as non-stained fibers. The preparation time for a given batch generally increased with added amount of CNF and initial volume of liquid (i.e. EP and acetone). However, the same preparation procedure was applied for each batch; this procedure was as follows. Solvent exchanged CNF was added to epoxy resin (in a glass beaker of known weight) in such quantity that the final resin system (EP, hardener and CNF) would obtain a solid content of either 1 wt% or 0.5 wt% CNF. The resulting suspension was thoroughly mixed using an Ultra-Turrax disperser (T25 digital ULTRA-TURRAX®, IKA-Werke GmbH & Co) for 20 min at 1.5×104 rpm. Negligible amounts of the suspension were lost during the mixing process, and
it was observed that between 13-27 % of the acetone evaporates during this stage of the process. After mixing, the suspension was degassed for 10 min using a vacuum desiccator with approximately – 1 bar of applied pressure. Any visible air bubbles remaining in the suspension were removed by gently stirring the mixture for approximately 10 min using a spatula. Heat aided evaporation was thereafter initiated by placing the beaker on a hot plate stirrer (VMS-C7, WVR international, USA). The temperature was set to 70 °C and the acetone was left to evaporate under gentle stirring while monitoring the weight of the suspension (Fig. 12). Depending on batch, this part of process would take between 5-24 h.
Figure 12 Schematic illustration of the evaporation setup
Under the assumption that only acetone had been evaporated it was found that it was possible to reduce the acetone content to less than 1 wt% using heat aided evaporation. On the other hand, for batches containing larger amounts of CNF (1 wt% intended for the final product) and/or of larger volume, the fraction of residual acetone will be between 5-7 wt% even after 48 hours of heat aided evaporation. Following acetone evaporation, the prepared resin was either covered and stored at 5 °C until addition of hardener or directly mixed with hardener, cured and prepared for evaluation of mechanical properties. Graphs of processing time and residual amount of acetone are provided in Appendix B for two representative batches of prepared resin.
2.1.4. Fiber reinforcement
Cross-ply laminates were manufactured using either glass or carbon fiber reinforcement (Fig.
450 gm-2 or 600 gm-2. The carbon reinforcement (Tenax®, Hexcel Tianjin Composite Material Co Ltd, China) have a density of 1.76 g/cm3, tensile strength and elastic modulus are 3950 MPa and 238 GPa, respectively. The thickness of the carbon fabric is 0.26 mm.
Figure 13 Images of used fiber reinforcement. Left: Biaxial glass fiber. Middle left: individual GF.
Middle right: Carbon fiber. Right: individual CF.
2.2. Composite fabrication
2.2.1. EP/CNF nanocomposites
Nanocomposites (i.e. CNF reinforced EP) were obtained by casting CNF loaded epoxy resin mixed with hardener (19/100 parts hardener by resin weight). Following mixing, the resin system was degassed and stirred for 10 min and carefully poured into a metal mold (Fig. 14). Curing was performed according to one of the recommend curing schemes for the PRIME™ 20ULV resin system: 24 h at 21 °C followed by 16 h at 50 °C. Nanocomposites were prepared according the cases given in Table 5.
Figure 14 Mixing and casting of nanocomposites: EP/CNF with added hardener (left), degassing (middle
Table 5 Manufactured nanocomposites and reference samples of epoxy. Designation CNF content (wt%) Comment
EP 0 Regular EP
EP/acetone 0 Epoxy treated with acetone (acetone evaporated) EP/CNF-0.5 0.5 Nanocomposite
EP/CNF-0.5/c. 0.5 Nanocomposite with carmine stained fibers
EP/CNF-1 1 Nanocomposite
2.2.2. Fiber reinforced EP/CNF
Glass fiber and carbon fiber composites, respectively, were manufactured by CSI Composite Solutions and Innovations Oy (Vilppula, Finland). All fiber reinforced composites were manufactured using RTM. Reference axes and infusion directions are illustrated in Fig. 15.
Figure 15 Schematic illustration of preform/composite reference axes and
resin injection directions.
2.2.2.1. Glass fiber reinforced EP/CNF
Resin with carmine stained CNF was used for fabrication of GF reinforced EP/CNF. Upon mixing the hardener, the epoxy system obtained a viscosity about 393 mPas (at 25°C), making it suitable for processing by RTM. The epoxy system was infused in a preform of biaxial glass fiber laid up as [0/90]4, where the outer layers had basis weights of 450 gm-2 and the inner layers
Figure 16 Hydraulic press with enclosed mold (left), clamped outlet (upper right) and sample
from finished EP/CNF/carmine composite (lower right).
Following addition of hardener and 5 min of mixing, the resin system was degassed for 20 min. Before infusion, the mold had been clamped and heated to 35°C in a hydraulic press (LinpuTM,
Metecno, Finland) equipped with heating elements (Fig. 16). A pressure of 2 bar was applied to the resin inlet of the mold and a vacuum pressure of 1 bar was applied to the resin outlet. After 15 min, the outlet was clamped and the inlet pressure increased to 5 bar for 90 min. The part was thereafter cured for 10 hours at 35°C, then post cured at 70°C for 10 hours. Following demolding the composite was trimmed and samples for mechanical and optical characterization were prepared.
2.2.2.2. Carbon fiber reinforced EP/CNF
Through-thickness infusion: Carbon fiber reinforced EP/CNF was manufactured in the same manner as the GF-reinforced composite, apart from the following differences. Five plies of CF fabric was stacked as [0/90/0̅]s (Fig. 17). A pressure differential of 3 bar was applied for 12 min before clamping the outlet and increasing the inlet pressure. The part was thereafter cured for 16 hours at 35°C before post-curing was performed.
Figure 17 Flow layer (left), CF preform (middle) and the enclosed mold (right) placed in a
In-plane infusion: Using the same mold as when performing through-thickness infusion, two additional layers of fabric were used to compensate for removal of the flow-mesh. In-plane infusion was performed using carbon fiber fabric laid up as [0/90/0̅]s (i.e. symmetrical lay-up of seven plies). For In-plane infusion, the resin infiltrated the preform in less than one minute using the same pressure parameters as earlier. The same curing cycle was applied for this part as well.
2.3. Material characterization
2.3.1. Bending tests
Three-point flexural tests were performed on a Instron 4411 (Instron, USA) according to ISO 141257. Samples having average dimensions of 1.8 mm in thickness, 60.2 mm length and 14.9 mm in width were used. Following ISO14125 the cross-head speed was 1.18 mm·min−1 while the span was 17.8 mm. Flexural stress (𝜎𝑓) was calculated according to
𝜎𝑓 = 3𝐹𝐿
2𝑏ℎ2 , (3)
where (𝐹) is the load, (𝐿) the span length, (𝑏) the sample width and (ℎ) the sample thickness. Flexural modulus (𝐸𝑓) was calculated according to
𝐸𝑓 = 𝐿 3 4𝑏ℎ3(
Δ𝐹
Δ𝑠), (4)
where (Δ𝐹/Δ𝑠) where obtained by linear regression in a deflection (𝑠) region corresponding to 0.05%-0.25% for each sample. Flexural modulus and strength, respectively, are reported as the mean value of at least five samples in each direction
Samples with their length (𝑙) machined parallel to the x-axis (see Fig. 15) are designated with an appended “x” and samples with the length in the y direction with a “y”. Samples tested in compression are those which have their surface facing the loading member during flexural test, and samples tested in tension have their surface facing the support members during testing. Samples tested in compression are designated with a “C” and tension with a “T”. An image of the testing setup is provided in Fig. 18.
Figure 18 Three-point bending test of GF/EP/CNF sample (left). Sample before and after
flexural test (right).
2.3.2. Tensile tests
Tensile tests were exclusively performed for nanocomposites (i.e. EP/CNF or EP/CNF/carmine) and reference EP samples. Mechanical properties of produced nanocomposites and reference EP were characterized using a Universal testing machine (Shimadzu Autograph AG-X, Shimadzu Corporation, Japan) equipped with a load cell of 5 kN (Fig. 19, right). Tensile samples (Fig. 19, middle) were prepared by machining rectangular samples (on average: 15 mm × 170 mm) from the cast nanocomposites sheets. The thickness of each sample was determined using a flat anvil type thickness gage (Mitutoyo 547-400S, Mitutoyo Corporation, Japan).
Figure 19 Universal testing machine (left), prepared tensile samples (right) and clamped sample
(right). Samples in middle figure include EP (from left to right), EP/CNF-0.5wt% and EP/CNF/carmine-0.5wt%.
DVE 201, Shimadzu Corporation, Japan). The elastic modulus was calculated from the initial part of the stress-strain curve (deemed to be in the linear region) and the yield strength was calculated using a 0.1% offset criterion. Strength, modulus and yield strength are presented in terms of averages values and the standard error of the mean from at least 5 samples from each nanocomposite or reference EP.
2.3.3. Raman spectroscopy
Dispersive Raman spectroscopy was performed using a SENTERRA Raman microscope (Bruker, USA). Using an excitation source of 795 nm (Nd:YAG), Raman spectra were obtained by illuminating the cross sections of highly polished CF/EP/CNF and EP samples, respectively, at multiple sites. All measurements were performed using a 10 mW output and 1s x 10 co-additions. Obtained spectra were compared with known spectra of cellulose in order to locate the presence of CNF in investigated samples.
2.3.4. Optical microscopy
Optical microscopy was employed for estimation of fiber and fiber bundle size, concentration and distribution of CNF in GF/EP/CNF/carmine samples and dispersion of CNF in prepared EP-resins. The CNF dispersion in prepared EP resin was examined using a polarizing microscope (LV100POL, Nikon, Japan). Fiber bundled size and CNF distribution were examined under a stereomicroscope (SMZ800, Nikon, Japan). The glass fibre and bundles sizes were measured from the polished cross-section of the composite. Distribution of CNF and its concentration in the manufactured composite were estimated using NIS-elements software (Nikon Instruments, Japan) and fibre volume fractions obtained through the RTM process.
2.3.5. Scanning electron microscopy (SEM)
3. Results and discussion
3.1. Characteristics of prepared CNF/EP resin and nanocomposites
3.1.1. Viscosity of prepared resins
Table 6 provides the resulting viscosity of liquid EP resin, with and without hardener added, following incorporation of 0.5 wt% and 1 wt% CNF, respectively. Graphical representation of the data is provided in Fig. 20.
Table 6 Viscosity of epoxy resin (Prime 20 LV) with different concentrations of CNF (with
and without hardener). Uncertainties are given as the standard error of the mean based on five measurements. The reported temperature under: ”With hardener” is the average temperature following 6 min after addition of hardener.
Case
Without hardener With hardener
Viscosity Temperature Viscosity Temperature
(mPa·s) (°C) (mPa·s) (°C)
0 wt% CNF 954 ± 3 21.3 219 22.8
0.5 wt% CNF 996 ± 17 23.0 393 25.2
1 wt% CNF 2734 ± 129 21.8 1145 24.6
Figure 20 Viscosity of epoxy resin (Prime 20 LV) with different concentrations of CNF
(with and without hardener). Given error bars are ± two times the standard error of the mean (SE) for each case of CNF content.
The resin viscosity plays a crucial role in RTM since lower viscosity facilitates resin infiltration 0 500 1000 1500 2000 2500 3000 0 wt% CNF 0.5 wt% CNF 1 wt% CNF Viscosity (mPa s)
