Synergistic effects of multiwalled carbon nanotubes and core-shell particles on the toughness and physico-mechanical properties of epoxy resin: a systematic study

Here, core-shell impact modier particles (CSIMPs) and multiwalled carbon nanotubes (MWCNs) were used as reinforcing agents for improving toughness and tensile properties of epoxy resin. For this purpose, emulsion polymerization technique was exploited to fabricate poly(butyl acrylate-allyl methacrylate) core- poly(methyl methacrylate-glycidyl methacrylate) shell impact modier particles with average particle size of 407 nm. It was revealed that using a combination of the prepared CSIMPs and MWCNTs could signicantly enhance toughness and tensile properties of epoxy resin. Also, it was observed that the dominant factors in enhancing the fracture toughness of the ternary composites are crack deection/arresting as well as enlarged plastic deformation around the growing crack tip induced by the combination of rigid and soft particles. The Response Surface Methodology (RSM) with central composite design (CCD) was utilized to study the effects of the amounts of core-shell particles and multiwalled carbon nanotubes on the toughness and tensile properties of epoxy resin. The proposed quadratic models were in accordance with the experimental results with correlation coecient more than 98%. The optimum condition for maximum toughness, elastic modulus, and tensile strength were 3 % wt. MWCNT and 1.03 % wt. CSIMPs. The sample fabricated in optimal condition indicated toughness, elastic modulus, and tensile strength equal to 2.2 MPa. m 1/2 , 3014.5 MPa and 40.6 MPa, respectively.


Introduction
Epoxy resins, due to their great physico-mechanical attributes such as good chemical resistance, outstanding adhesion, and notable dimensional stability, are of paramount importance in various structural applications [1]. However, due to their high density of three-dimensional molecular structure, hardly can they resist toward growing of crack. This serious drawback limits their applications in many critical elds where a notable fracture strength or low-temperature toughness is required, such as aerospace and automotive industries [2] [3]. To deplete the severity of this issue, a plethora of different ingredients such as liquid rubbers, core-shell particles (CSIMPs), and rigid organic and inorganic ingredients were used as impact modi ers particles (IMPs) of epoxy resin [4][5][6][7].
Generally speaking, soft IMPs (i.e. liquid rubbers and CSIMPs) could endanger signi cant improvement in the toughness of epoxy resin through various mechanisms such as deboning, cavitation, and crazing [6,8]. Nonetheless, incorporation of such IMPs in epoxy formulation could result in deterioration of other desired properties of the host polymer (such as modulus, stiffness, and tensile strength) [9]. On the other hand, rigid IMPs not only could enhance the toughness of epoxy resin by crack pinning, crack de ection and de ection/bifurcation effect of growing crack, but also could boost stiffness and modulus of the host polymer [10]. However, to reach a satisfying level of toughness, it is inevitable to use a relatively high dosage of such rigid IMPs in epoxy formulation. This obligation causes dramatic increase in viscosity and subsequently inroads serious di culties in the processability of the obtained composite [11].
Regarding the mentioned limitations of IMPs and their inherent differences from the view point of toughening mechanisms, in the last decade, hybridization of different rigid and soft IMPs embedded into epoxy polymer matrix provided outstanding opportunities to the researchers for developing high performance nanocomposites with enhanced toughness and physico-mechanical properties [12][11] [13] [14] [15] [16]. The obtained ternary nanocomposite could overcome the mentioned drawbacks and cause synergistic in uences on the toughness and physico-mechanical properties of the obtained composite products.
It is no doubt that outstanding attributes of carbon nanotubes such as their high tensile strength and stiffness could remarkably enhance fracture toughness and physico-mechanical properties of conventional polymers [17]. Thus, among different rigid particles, it has been demonstrated that using carbon nanotubes as rigid llers along with other soft IMPs could culminate a notable enhancement in fracture toughness of the obtained composites. Río et al. showed that the addition of a mixture of rigid carbon nanotube and soft polystyrene-polybutadiene-poly (methyl methacrylate) triblock copolymers led to the dramatic increase in both fracture toughness and fracture energy of the epoxy resin [18]. In another attempt it was revealed that using block copolymer could ameliorate the dispersion state of carbon nanotubes and causes a synergistic effect on the toughness of epoxy resin [19]. Jojibabu et al. showed that ternary composite of triblock copolymer, modi ed carbon nanotubes, and epoxy resin possess a remarkable improved toughness in comparison with binary composites [19].
The technique of experimental design due to its valuable assistance for comprehensive understanding of the process and interactive relationships between different variables, is known as a distinguished method in the engineering elds [20]. Among different methods, response surface method, which is a collection of mathematical and statistical techniques, has been used extensively to study and optimize the effect of diverse determining factors on the various parameters [21]. To the best of current authors' knowledge, there is no single published numerical and systematic study about the synergistic effects of MWCNTs and CSIMPs on the toughness and tensile properties of epoxy resin. Thus, a systematic study of the in uence of multiwalled carbon nanotubes (MWCNTs) and CSIMPa on the mentioned attributes of epoxy resin is the main objective of this work. For this purpose, response surface methodology was used to nd an applicable approximation function for predicting and optimization of the toughness and tensile properties of composites with various amounts of CSIMPs and MWCNTs. MWCNTs with purity of >95 %, length of 5-15 µm , and outer diameter of 10-20 nm was purchased from Nutrino (Tehran, Iran). The epoxy resin was standard diglycidylether of bis-phenol A (DGEBA) (Epon828 from HEXION) with an epoxy equivalent molecular weight between 185 g/eq. and 192 g/eq. The curing agent was cycloaliphatic polyamine hardner (Epicure F205 from HEXION) with hydroxyl equivalent weight of 102-106 g/eq. Deionized water (DIW) was used in all recipes.

Characterizations
Size and external morphology of the prepared latex particles were investigated by scanning electron microscopy (SEM) with MIRA3 instrument from Tescan (The Czech Republic). Also, SEM images of the fracture surfaces of the epoxy specimens were used to assess the toughening mechanisms. The internal morphology of the prepared latex particles was studied by transmission electron microscopy (TEM) with CEM 902A ZEISS with an accelerating voltage of 80 keV (Oberkochen, Germany). Tensile properties of the dog-bone-shaped specimens were obtained by a universal testing machine according to ASTM D-638. For each sample, the data reported are the average of ve measurements. Fracture energy (G IC ) and fracture toughness (K IC ) were determined using a single-edge notch bend (SENB) specimen with dimensions of 6 ×12 × 52.8 mm. The tests were performed at a rate of 10 mm/min.

Synthesis of the impact modi er particles
Soft impact modi er particles with core-shell morphology was prepared through emulsion polymerization technique in two stages. At the rst step, elastomeric cores of poly (BA-ALMA) were prepared through batchwise emulsion polymerization. For this purpose, 160 g DIW, 1 g Emulsogen APS 100, 30 g BA, and 0.16 g ALMA were premixed and charged into a 250 mL, four-necked round bottom glass reactor equipped with re ux condenser, thermocouple, crescent Te on-steel mechanical stirrer, and feeding inlets. The content of the reactor was deoxygenated by purging with nitrogen for 15 min. The temperature was raised to 80 °C and the polymerization reaction was started by the addition of solution of initiator (0.04 g APS dissolved at 10 mL water). The conversion of the polymerization reaction was followed by gravimetric analysis of samples withdrawn from the reactor at regular time intervals. After 120 min, the nal conversion of the polymerization reaction was above 99%.
In the second stage, copolymerization of MMA and GMA was occurred on the surface of the prepared elastomeric cores. For that, 200 g of the prepared latex, 9 g MMA, 1 g GMA, and 0.1 g TBPB charged into the reactor. The content of the reactor was mixed for 30 min at room temperature. Then, 10 ml of the aqueous solution of 0.02 g EDTA, 0.01 g FeSO 4 . 7H 2 O, and 0.1 g SFS was added into the reactor. The redox emulsion copolymerization was proceeded for 3 h at 30 °C to reach conversion of >98 %. Then, the prepared latex was ltered through a 53 mm sieve and the ltrate was freezing for 12 h and freeze-drying for another 12 h.

Experimental design and fabrication of toughened specimens
The CCD of RSM was used to assess the effects of the contents of MWCNTs and CSIMPs as independent factors on the fracture toughness and tensile properties of epoxy composites. Table 1 shows the ranges and levels of these factors. Total number of 13 experiments was designed by CCD algorithm which consists of 4 cube points, 4 axial points and 5 replications of the central points in cube to check the reproducibility variance.  Table 2 illustrates the details of the designed experimental conditions proposed by CCD. To nd the best tted model, the step-wise model tting by Minitab 14 software was used. Equation 1, which is a secondorder empirical quadratic polynomial model, shows the behavior of the system Where Y is the predicted physico-mechanical response, bo is the constant, bi is the linear effect of the factor xi (i=1, 2 and 3), bii is the quadratic effect of the factor xi and bij is the linear interaction effect between the input factors, xi and xj.
To evaluate the effect of CSIMPs and MWCNTs on the physico-mechanical properties of epoxy resin, specimens with different contents of these reinforcing llers were fabricated according to the Table 2. For this purpose, de nite amounts of MWCNTs and CSIMPs were added to 100 g of epoxy resin and stirred to form a uniform mixture. Then, to make sure that the llers were uniformly distributed in the epoxy matrix, the mixture was processed by a three-roll milling (EXAKT 80E, Exakt Technologies, Germany) for 20 min via a three-roll grinder. The three-roll milling was used in gap mode with a 5 mm distance con guration between the center and apron roller. The speed of the apron roller was 260 rpm and the roller speed ratio was 6:2:1. Then, 58 g Epicure F205 as a curing agent was added to the mixture and stirred with a glass rod. Finally, the prepared mixtures were degassed by a centrifugal deaerator for 10 min at 3000 rpm. The mixtures were poured into silicon molds and a two-step curing procedure was carried out; 2 h at 40 °C then 4 h at 120 °C using a ramp rate of 10 °C/min.

Preparation of impact modi er particles with core-shell morphology
The capability of emulsion polymerization in preparing colloidal particles with disparate morphologies, physical, and chemical attributes makes it a unique and enticing polymerization method [22]. Here, thanks to these features, impact modi er particles with elastomeric core and epoxy-functionalized plastic shell was prepared successfully. A schematic illustration of the preparation steps of the CSIMPs is shown in Figure 1.
It is well known that using monomer such as ALMA, which has unsaturated bonds with different reactivity, could signi cantly enhance grafting e ciency, gel content, and structural stability of CSIMPs [23]. Linked to these rationales, in this work ALMA was chosen as a gra inking monomer in the formulation of the prepared elastomeric cores of CSIMPs. During the polymerization process, the colloidal stability of the growing latex particles was supplied by using a polymerizable anionic surfactant (Emulsogen APS 100). It well known that accumulations of surfactants in epoxy matrix could act as softening agents and render changes in the toughness and other physico-mechanical properties of epoxy polymers [24]. Here, it is suggested that through using a surfmer in emulsion polymerization formulation, the migration and accumulation of surfactants were hindered effectively.
To fabricate the plastic shell, redox emulsion copolymerization of MMA and GMA was occurred at the interfacial portion of elastomeric cores and aqueous medium. It was suggested that using a swelling time for dissolving monomers in poly (BA-ALMA) colloidal particles as well as proceeding the radical formation reactions at the interfacial region, could effectively minimize the probability of micellar and homogeneous nucleation [25]. In other words, it was surmised that exploiting a combination of oil soluble and water soluble ingredients as couple of redox initiator could guarantee that the copolymerization of MMA and GMA was mainly occurred at the surface region of elastomeric particles. Thus, by considering the water solubility of TBPB (<1 g/L) and SFS (~600 g/L), the initiation step of the polymerization probably occurs at the surface of monomer-swollen poly particles. Also, iron (II) as co-catalyst and EDTA as a sequestering agent were used to increase the e ciency of the redox polymerization system and to prevent radical formation in an uncontrolled way, respectively [26]. Figure 2 shows the electron microscopy results of the prepared CSIMPs and manifest that employed synthesis method could successfully fabricate CSIMPs with average particle sized of 407 nm.

Modeling and Optimization using Response surface methodology
The results in Table 2 were used to t the polynomial models representing the Young's modulus (E), tensile strength (S) and fracture toughness (K IC ) of the toughened specimens (responses) as a function of the contents of MWCNTs (wt.%; X 1 ) and CSIMPs (wt.%; X 2 ). The following equations express the overall predictive model in terms of the coded variables: Y Young' s modulus = 1882.9 + 664.2 X 1 -345.7 X 2 + 164. 6   The importance and statistical signi cance of the quadratic models was evaluated by the analysis of variance (ANOVA), which has been presented in Table 3. The ANOVA includes some statistic factors such as lack of t, R 2 , and adjusted R 2 . The ANOVA results (Table 2) indicated that the models were highly signi cant, as P-value for the models was 0.000. The high R 2 values of the models demonstrated that they were capable of accurately predicting the mechanical properties of the prepared epoxy samples in the studied range. The insigni cance "lack of t" with P-value of higher than 0.05 indicated that in the models, LOF is meaningless and the models have appropriate accuracy for predicting of the results. All terms in the regression models are not equally important. The signi cance of each coe cient was determined by P values (probability), which are listed in Table 3. The probability values less than 0.05 call for the rejection of the null hypothesis indicating that the particular term signi cantly affects the mechanical properties of the coatings.
Page 8/18 Table 3. Analysis of variance for the models and estimated regression coefficients The obtained response surface and contour plots provide a procedure to predict the mechanical properties of the prepared epoxy specimens reinforced with various amounts of MWCNTs and CSIMPs.
As it can be seen obviously in Figure 3-a, by increasing the content of MWCNTs, due to their intrinsic rigidity and signi cantly high Young's modulus, the E values of the fabricated epoxy composites increased remarkably [27]. Also, it can be seen that addition of CSIMPs could not effectively alter the Young's modulus of the host polymer up to 2.2 %wt. The plausible explanation for this observation could be due to the relatively low dosage of the used CSIMPs and also their ability for formation of covalent bonds with epoxy matrix [28]. The latter phenomena could effectively form spatial hindrance for polymer chains and increase the crosslink density of the nal composite [29]. However, this effect could partially be compensated by the elastomeric core of the CSIMPs and result in the observed plateau region of modulus as a function of the content of CSIMPs. However, at higher levels of CSIMPs (> 2.2 % wt.) an augmentation in modulus of the prepared composites was observed which could be attributed to the increasing in the content of the glassy functional shell of the CSIMPs in the composites structures [30].
As it can be seen clearly from the obtained contour plot, the measured modulus of the fabricated ternary composites could be explained by the cited rationales for individual effects of MWCNTs and CSIMPs.
It is well known that introducing MWCNTs in epoxy matrix could exert signi cant in uence over tensile strength of the obtained composite [31]. This in uence could be seen as a remarkable amelioration in tensile strength by increasing the content of MWCNTs in Figure 3-b. Also, it is seen that the addition of CSIMPs causes an improvement in tensile strength of the host polymer. This observation could be related to two pivotal traits of CSIMPs; i) the ability of the plastic shell for creation of covalent bonds with epoxy matrix, ii) crosslinked attribute of the core which could guarantee the structural stability of CSIMPs [32].
The obtained S values of ternary composites revealed that the contributional effect of CSIMPs on the obtained results is more prominent than MWCNTs. In other words, at a constant concentration of MWCNTs, the addition of CSIMPs render a signi cant decreasing in tensile strength of the obtained nanocomposite, which could be attributed to the elastomeric portion of CSIMPs.
The in uence of the contents of MWCNTs and CSIMPs on the fracture toughness of epoxy resin is shown in Figure 3-c. It can be seen clearly that both of the added ingredients could signi cantly increase the K IC of epoxy resin, however, as it is expected, the in uence of CSIMPs is more prominent than MWCNTs. Moreover, it is observed that for the binary composites of MWCNTs/epoxy the measured fracture toughness is not directly proportional to the amounts of MWCNTs. This observation is probably due to the aggregation of MWCNTs which could be occurred at higher concentration of this material [33]. It is well-known that due to the enormous speci c surface area of MWCNTs, they have strong tendency for creating clusters within polymer matrix. The formed clusters of MWCNTs could decrease the interfacial surface area between epoxy resin and rigid llers, which consequently can reduce their interaction with growing crack and decrease the toughening performance of MWCNTs. An enticing additional insight is the synergistic toughening effects of MWCNTs and CSIMPs. It can be seen conspicuously that the ternary composite with 3 %wt. MWCNTs and 3 %wt. CSIMPs possesses the highest value of facture toughness. This observation shows that the employed rigid and soft llers could exert signi cant positive in uence over fracture toughness of epoxy resin. It is substantiated that the dispersion state of MWCNTs could be improved signi cantly by the addition of block copolymers as IMPs, here it is suggested that the added CSIMPs could decrease the aggregated structures of MWCNTs and ameliorate their dispersion in epoxy matrix [18].
Optimization of effectual factors was mainly done to de ne the optimum formulation that caused the achievement of maximum E, S, and K IC .

Fracture surface
To assess the effect of MWCNTs and CSIMPs on the fracture toughness of the ternary composite, the fracture surface of epoxy polymer, binary and optimized ternary composites were comparatively examined by SEM. As it can be seen in Figure 4-a, the neat epoxy sample exhibits a typical fracture surface for un lled epoxy resins, i.e. smooth fracture surface with the ne line patterns which indicates generally uninterrupted crack propagation and low fracture toughness [34]. On the other hand, for binary epoxy composite with 1.5 %wt. MWCNTs (Figure 4-b), a rough fracture surface indicating a ductile fracture of the composite is observed. It has been reported that addition of MWCNTs to epoxy matrix could signi cantly increase the fracture toughness of host polymer through introducing restrictions to crack propagation and increasing the amount of plastic deformation [35]. In other words, the tip of the propagating crack tips could not break the strong MWCNTs and consequently they force to arrest or change their direction. Also, the high magni cation SEM fractograph of this specimen (Figure 4-c) shows a plethora of pulled out nanotubes along with curved pattern which indicate a great deal of di culty in crack initiation and propagation within the matrix of MWCNT containing samples in comparison with bared epoxy matrix [36]. Therefore, through the process of pulling out of MWCNTs from epoxy matrix, a great deal of dissipation of energy was occurred which cause a remarkable amelioration in fracture toughness of the prepared nanocomposites. Figure 4-d shows the fractograph of binary composite of epoxy/CSIMPs. Rough fracture surface as well as spherical holes which are related to the deboning of CSIMPs and subsequent plastic deformation of matrix, indicate a ductile fracture of the prepared specimen [37]. It is surmised that stress-activated shear yielding in the high stress region around CSIMPs leads to the deformation of their rubbery portion and cavitation phenomenon, which in turn could signi cantly boost toughness of epoxy resin. In addition, as it can be seen obviously, the sizes of the holes are considerably bigger that the measured size of CSIMPs in SEM images ( Figure 2). This observation is probably due to massive plastic deformation of the matrix around the AIMPs [38].
Fractographs of toughened samples with CNT and CSIMPs conspicuously show plastic deformation, crack de ection and pull out MWCNTs (Figure 4-e, f). This observation surmised that combination of crack de ection and inhibition of MWCNTs along with cavitation of CSIMPs could have synergistic toughening effects on epoxy matrix and enhance its toughness remarkably.

Conclusion
In this study, for the rst time CCD/RSM design was adopted to determine the optimum condition for tensile properties and fracture toughness of epoxy/MWCNT/CSIMPs ternary composite. The CCD/RSM prepared the su cient statistical data to t a quadratic model in terms of the %wt. MWCNTs and %wt. CSIMPs in epoxy resin, on the tensile and toughness properties. The signi cance of the proposed models was indicated by the p-value of Regression <0.0001 and P-value of LOF>0.05. Analysis of variance showed good coe cient of determination values, R 2 , (>98) for the studied properties. The optimum condition for maximum Young's modulus, tensile strength, and toughness of the epoxy composite obtained from RSM were 3 %wt. MWCNTs and 1.03 %wt. CSIMPs in epoxy of matrix. The sample prepared in optimal conditions indicated module= 3014 MPa, tensile strength= 40.57 MPa, and K IC = 2.2 MPa.m 1/2 . The obtained results indicated that the optimized ternary composites possess a good balance between Young's modulus, tensile strength, and toughness, which could not be achieved by other binary ones. The SEM images of the fractographes substantiated that combination of MWCNTs and CSIMPs could merge their toughening mechanisms and cause a synergistic toughening effect on the obtained ternary composite.  Response surface and Contour plots of tensile properties and toughness of epoxy composites