epoxy nanocomposites

Polymer 53 (2012) 6081e6088

Contents lists available at SciVerse ScienceDirect

Polymer journal homepage: www.elsevier.com/locate/polymer

Behavior of load transfer in functionalized carbon nanotube/epoxy nanocomposites Peng-Cheng Ma a, b, *, Qing-Bin Zheng a, Edith Mäder c, Jang-Kyo Kim a a

Department of Mechanical Engineering, The Hong Kong University of Science and Technology, Clear Water Bay, Hong Kong, China The Xinjiang Technical Institute of Physics and Chemistry, Chinese Academy of Sciences, Urumqi 830011, China c Leibniz-Institut für Polymerforschung Dresden e.V., Dresden 01069, Germany b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 29 March 2012 Received in revised form 29 June 2012 Accepted 28 October 2012 Available online 5 November 2012

This paper presents the correlation between the functional groups, interfacial microstructure and behavior of load transfer in carbon nanotube (CNT)/epoxy nanocomposites. Nanocomposites consisting of epoxy and CNTs with/without functionalities (amino and epoxide groups) are prepared and characterized to evaluate the CNT-matrix interactions based on strain-sensitive Raman spectroscopy. The results show that nanocomposites filled with functionalized CNTs exhibit a noticeable G0 -band shift in tension while those containing pristine CNTs have a marginal shift, suggesting a more efficient load transfer between the epoxy matrix and functionalized CNTs. An interesting observation is that the slope of the G0 -band shifts can be either positive or negative, depending on the functional groups on CNTs and the interfacial structures created between the functionalized CNTs and polymer matrix. The mechanisms behind this observation are discussed with reference to fractography and thermo-mechanical properties of nanocomposites. Ó 2012 Elsevier Ltd. All rights reserved.

Keywords: Carbon nanotubes Polymer nanocomposites Load transfer

1. Introduction With unique mechanical and transporting properties, carbon nanotubes (CNTs) have attracted much interest for the purpose of developing polymer-based nanocomposites with excellent mechanical performance and multi-functional characteristics. The excellent properties of CNTs alone, however, do not guarantee mechanically superior products because, as in most other types of composites, the mechanical properties of nanocomposites depend not only on the properties of reinforcement, but more importantly on the degree to which an applied load is transferred from the matrix phase to the reinforcement through an interface [1e4]. Therefore, many different surface treatments and functionalization techniques have been developed with varying degrees of success to improve the interfacial adhesion of CNTs with polymer resins [5e7]. The characterization of CNT-polymer interface is a challenging task because of the technical difficulties associated with the manipulation of nano-scale objects. In the field of fiber-reinforced polymers, it is well understood that the application of a mechanical load to carbon or Kevlar fiber results in shifted wavenumbers of * Corresponding author. The Xinjiang Technical Institute of Physics and Chemistry, Chinese Academy of Sciences, Urumqi 830011, China. Tel./fax: þ86 991 3677875. E-mail address: [email protected] (P.-C. Ma). 0032-3861/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.polymer.2012.10.053

Raman peaks, which are directly related to the fiber modulus. Correlating the rate of Raman shift with an applied strain leads to the evaluation of interfacial strength experienced by the stiff fibers in the composites [8]. Similar Raman shifts were also observed when CNT/polymer nanocomposites were subjected to a mechanical load [9e14]. The principle of this technique lies in the fact that when the mechanical load is transferred from polymer matrix to nanotubes, it will induce local strain to CNTs, causing a change in the CeC bond vibration which can be monitored by the Raman spectroscopy. Coupling of these Raman responses along with the mechanical deformation of matrix endows the evaluation on the interfacial interactions between the CNTs and polymer matrix. More recently, this technique was adopted to study the interfacial load transfer in graphene/polymer nanocomposites [15,16]. However, there were some discrepancies regarding experimental observations and explanation on observed Raman shift of CNT filled nanocomposites. It was reported [9] that the applied load was transferred from the matrix to pristine CNTs with a positive shift of Raman wavenumbers. Given that these CNTs were pristine ones and mostly in the form of ropes or agglomerates due to poor dispersion, it was suspected that the Raman shift obtained only reflects slippage of individual CNTs within the agglomerates [11,14], not between the CNTs and the matrix resin. For the nanocomposites containing randomly dispersed CNTs, the Raman peak of CNTs was shifted upwards under a compressive strain, whereas a slightly

6082

P.-C. Ma et al. / Polymer 53 (2012) 6081e6088

positive shift was obtained in tension [9]. This was contrary not only to many observations in that the Raman peaks of CNTs shifted downwards when the nanocomposites were in tension [11e13,17e 19], but also to the Raman response of composites containing carbon and Kevlar fibers [8]. In addition, there is a possibility that functionalized nanotubes may exhibit different behavior of load transfer under mechanical load, as different functionalities on CNTs create an interface with specific structure and properties. This issue has not been addressed in previous studies. This paper is part of our continuous endeavor in developing novel techniques to improve the dispersion and interfacial interactions of CNTs with polymer resins [7,20e23]. A particular emphasis in this paper was placed on studying the role of functionalities on the deformation of CNTs embedded in polymer matrix using Raman spectrometry. The mechanisms behind the different behavious of load transfer from matrix to functionalized CNTs were illustrated. 2. Experimental 2.1. Materials and nanocomposite fabrication The CNTs used in this study were basically the same as those employed previously [21,22]. They were multi-walled CNTs and synthesized by a chemical vapor deposition method (supplied by Iljin Nanotech, Korea). CNTs containing two different functionalities, i.e., amino and epoxide groups, were prepared, aiming at creating a specifically interfacial structure, a surface formed by a common boundary of reinforcing CNTs and epoxy matrix that is in contact with and maintains the bond in between for load transfer. For the functionalization of CNTs, the pristine CNTs (P-CNT) were first subjected to UV/O3 treatment in a chamber (Jelight 144AX-220 UV/Ozone Cleaning System) to oxidize and create oxygenated groups (mainly carboxylic groups) on the surface [20,24]. The attachment of amino groups on CNTs (NH2-CNT) was realized by direct coupling between the carboxylic groups on CNTs and the ethylene diamine (EDA, Aldrich) with the assistance of dicyclohexylcarbodiimide (DCC, Aldrich) and dimethylamino-pyridine (DMAP, Aldrich) as catalysts [22]. The silane functionalization of CNTs (Si-CNT) was accomplished by converting the carboxylic groups to hydroxyl ones on CNTs using lithium aluminum hydride (LiAlH4, Aldrich) followed by silanization in a glycidoxypropyltrimethoxysilane (GPTMS, Aldrich) solution [21]. Fig. 1 shows the schematics of reactions taking place during the functionalization. The nanocomposites containing different CNTs with a fixed weight content of 0.25% were made from diglycidyl ether of bisphenol A (DGEBA, Epon 828, Shell Chemical) as an epoxy monomer and m-phenylenediamine (m-PDA, SigmaeAldrich) as a curing agent. For the fabrication of CNT/epoxy nanocomposites, the pristine or functionalized CNTs were dispersed in ethanol for

1 h before adding the monomer epoxy, and the mixture was ultrasonicated in a water bath for 1 h each at 60  C and 80  C, respectively. The mixture was then outgassed at 80  C to eliminate the entrapped air and the remaining solvent. The m-PDA hardener was added into the mixture in the ratio of 14.5/100 by weight. The mixture was moulded into a flat plate and cured at 80  C for 2 h, followed by post cure at 150  C for additional 2 h. 2.2. Characterization The functional groups attached on CNTs were quantitatively analyzed using a thermogravimetric analyzer (TGA, Unix/TGA7 by Perkin Elmer). Specimens were heated from ambient to 800  C at a heating rate of 10  C/min under air flow. Field emission transmission electron microscope (TEM, 2010F, JEOL) with an energy dispersion X-ray analyzer (EDX) was used to characterize the surface morphology and chemical composition of CNTs before and after functionalization. The Raman spectra of CNTs and CNT/epoxy nanocomposites were obtained using a micro-Raman system (RM 3000, Renishaw). The excitation source was a HeeNe laser with a wavelength of 633 nm and the spectra data were collected in the backscattering mode. For the deformation of nanocomposites, the cured nanocomposite plates were cut into rectangular specimens with dimensions 50 mm long  3 mm wide  1 mm thick, which were loaded in a microtensometer consisting of a specially-made load cell in conjunction with a micrometer. A 50 lens was used to focus the laser on the specimen surface with a 2e3 mm diameter laser spot, and the laser beam was polarized using a polarizer along the direction of mechanical load to maximize the signals from CNTs parallel to this direction. To minimize stress relaxation of nanocomposites and heating effect of laser, 5 scans with 10% of full power (2 mw) were optimized for each Raman spectrum collection (time is about 60 s). The nanocomposite specimens were deformed step-by-step at intervals of about 0.1% strain by rotating the micrometer, and the strain was estimated by dividing the change in gauge length by the original length of the specimen. The Raman spectra were collected from the exactly same spot of the specimen at different strains to detect the in-situ deformation of CNTs in the nanocomposites. All selected bands were fitted and deconvoluted using a combination of Gaussian and Lorentzian functions to identify the peak position. The fractured surface of nanocomposites was examined using a scanning electron microscope (SEM, JSM-6700F, JEOL). The mechanical properties of nanocomposites were measured on a dynamic mechanical analyzer (DMA-7, Perkin Elmer). Rectangular samples with dimensions 20 mm long  3 mm wide  1 mm thick were tested in three point bending from 30  C to 170  C at a heating rate of 10  C/min and a frequency of 1.0 Hz in an argon atmosphere.

Fig. 1. Schematics of reactions for CNT functionalization.

P.-C. Ma et al. / Polymer 53 (2012) 6081e6088

Fig. 2. Raman spectra of CNTs showing that the intensity of D-band increases due to the attachment of organic groups on functionalized CNTs, and this becomes more pronounced in Si-CNT.

3. Results and discussion 3.1. Surface functionalities on CNTs Fig. 2 shows the Raman spectra obtained for different CNT samples along with corresponding intensity ratios. The Raman wavenumbers of CNTs correspond to three major peaks [11,14]: i) Dband at around 1340 cm1 representing the residual, ill-structured graphite with sp3 carbon arising mainly from the impurities and defects on the surface; ii) G-band at around 1570 cm1 corresponding to the perfect sp2 graphene layer of CNTs; and iii) G0 -band at around 2660 cm1 assigned to the second overtone of the D-band. The intensity ratio of D to G band, ID/IG, has been used to quantify the structural integrity of CNTs, and the higher value of this ratio is, the more defects and impurities present on CNTs. Compared with the ID/ IG value of 0.82 for P-CNT, the ratios for NH2-CNT and Si-CNT increased to 1.12 and 1.58, respectively, confirming the attachment of functional groups on CNT surface. The higher intensity ratio for SiCNT than NH2-CNT reflects a higher degree of flawed sp3 carbon in Si-CNT. It should be noted that the P-CNT exhibited an inherently high ID/IG ratio of 0.82 due to the existence of impurities and defects arising either from the fabrication or purification after production, which was confirmed in our previous studies [21]. Significant Raman shifts were also observed in functionalized CNTs (Inset in Fig. 2), depending on the heterogeneous atoms attached on CNTs: For NH2CNT, upshifts were found for all three distinctive Raman peaks, whereas an upshift in G band along with downshifts in D and G0 bands were observed in Si-CNT. These shifts are attributed to the modification of the electronic structure of CNTs with different functionalities: amino groups were attached on NH2-CNT and the nitrogen atoms in these groups contain lone-pair electrons that can function as the electron donor, resulting in a charge transfer between the electrons and walls of CNTs [25]. In contrast, the shifts observed in Si-CNT arise from the oxygen atoms in Si-CNT [26,27].

6083

The attachment of functional groups on CNTs was further confirmed by employing surface analysis techniques and studying the thermal stability of samples. Table 1 summarizes the results arising from FT-IR and XPS characterization [21,22]. IR wavenumbers representing the characteristic adsorption of functional groups, such as eCH2e, e(C]O)eNHe, eCeNe and epoxide structures were identified in functionalized CNTs. The XPS results showed the presence of nitrogen and silicon atoms in functionalized CNTs, and the Si-CNT exhibited a higher concentration of heterogeneous atoms than that of NH2-CNT (11.06% vs. 9.17%), suggesting a higher degree of functionalization. Thermogravimetric analysis was performed to shed an insight into the analysis of functional groups on CNTs, as shown in Fig. 3. The P-CNT showed an excellent thermal stability (Curve A) with 97% weight retention along with a weight loss plateau before catastrophic decomposition at 570  C. The NH2-CNT exhibited a lower thermal stability (Curve B), as a noticeable weight loss was found at 400  C. The weight loss for these CNTs was 14% at the temperature of 570  C, suggesting that the organic component in NH2-CNT was about 10%. In sharp contrast, the Si-CNT showed a continuous weight loss as temperature increased (Curve C). The total weight loss of Si-CNT before catastrophic decomposition was 22.6%, which was much higher than the corresponding losses in P-CNT and NH2-CNT. The reason for this observation is that the silanization of CNTs involves hydrolysis of silane and anchoring of silane resulting from the reactions between the hydroxyl groups on CNTs and methoxide groups in silane (see Fig. 1), thus leading to a higher degree of functionalization. It is interesting to note, however, that Si-CNT exhibited better thermal stability than the other two CNTs at further elevated temperatures above 650  C. This phenomenon may be attributed to the physical barrier effect of silicon oxide: when silane is pyrolyzed in air flow, a protective silicon oxide layer is formed on CNTs, which effectively impedes the propagation of decomposition reactions of CNTs [28]. This hypothesis was verified by the fact that a significant yield, i.e., 9.6% residual, was noticed for SiCNT, which was much higher than its counterparts, P-CNT and NH2-CNT, with 3.8% and 0.7% yields, respectively. Functionalization also led to changes in morphology of CNTs. The high resolution TEM were employed to study the surface structure of CNT before and after functionalization, and Fig. 4 shows the typical TEM images. The P-CNT exhibited a relatively clean surface with multi-layer of graphene (Fig. 4A). The sound, layered structure of NH2-CNT (Fig. 4B) indicates that there was little damage to CNT surface due to the amidization. Some amorphous materials with a thickness of 1e3 nm were attached on the wall of Si-CNT (Fig. 4C). The detection of silicon (Spot A in Fig. 4C) by EDX (Fig. 4D) confirms that these amorphous materials were derived from the silane molecules. In light of different surface morphologies for CNTs with different surface functionalities, it can be concluded that the silanization of CNTs resulted in wrapping of silane molecules on CNTs, a reflection of a higher degree of functionalization than that of amidization. The results are in good agreement with those from Raman and TGA.

Table 1 Summary on the functional groups and elemental composition of CNTs. Parameter

CNT P-CNT

NH2-CNT

Si-CNT

Functionalization agent Major functional groups (by FT-IR, wavenumbers and assignation)

/ 3419 and 1058 cm1 (eOH from H2O)

Glycidoxypropyl-trimethoxy silane 2914 cm1 (eCH2e from silane), 793 cm1 (epoxy group)

Element composition (by XPS, atomic %)

C ¼ 98.34, O ¼ 1.32

Ethylene diamine 2911 cm1 (eCH2e from EDA), 1539 and 1258 cm1 (-(C]O)eNHe), 1080 cm1 (CeN stretching) C ¼ 90.83, O ¼ 6.64, N ¼ 2.53

C ¼ 88.51, O ¼ 9.38, Si ¼ 1.68

6084

P.-C. Ma et al. / Polymer 53 (2012) 6081e6088

100 90

A: P-CNT B: NH2-CNT C: Si-CNT

80

Weight (%)

70

A

B C

60 50 40 30 20 10 0 0

100

200

300 400 500 Temperature (oC)

600

700

800

Fig. 3. Thermal stability of CNTs before and after functionalization.

3.2. Load transfer in CNT/polymer nanocomposites Fig. 5 shows the Raman G0 -band shifts of CNTs in nanocomposites as a function of tensile strain. The G0 -band of nanocomposites filled with P-CNT scattered in a narrow range (Fig. 5A), indicating very small change in atomic distance of -CeCe bond and poor load

transfer from the matrix to CNTs. This observation matches closely with previous reports that multi-walled CNTs without functionalization exhibited marginal capabilities of transferring mechanical or thermal load from the polymer matrix [13,17,18]. In contrast, the Raman wavenumbers in the functionalized CNT/epoxy nanocomposites shifted significantly with increasing strain, initially at a steep rate followed by a rather lower rate at high tensile strains (B and C in Fig. 5). An interesting observation is that the slope of G0 -band as a function of strain varied either positively or negatively depending on the type of functional groups created on CNTs: The NH2-CNT/epoxy nanocomposites showed a G0 -band upshift (Fig. 5B), whereas the G0 -band of the nanocomposites containing Si-CNT shifted downwards (Fig. 5C). Three important features can be identified from these data: i) the shape of G0 -band shift as a function of strain illustrates the type of deformation of functionalized CNTs in the polymer matrix, which can be either a compression mode (with a positive slope) or a tension mode (with a negative slope) when the matrix is subjected to uniaxial tension; ii) the value of the slope is a measure of the efficiency of load transfer to CNTs; and iii) there are different load-transfer mechanisms when the nanocomposites are under different loading situations. In light of the report that the shift of the G0 -band is proportional to the stress applied on CNTs [12,17], it can be said that the higher is the slope as in the nanocomposites containing functionalized CNTs (B and C in Fig. 5), the more efficient is the load transfer from the matrix to CNTs. Micromechanical interlocking, van der Waals interaction and chemical bonding are

Fig. 4. TEM images showing the surface morphology of different CNTs (A: P-CNT; B: NH2-CNT; C: Si-CNT; D: EDX of Si-CNT).

P.-C. Ma et al. / Polymer 53 (2012) 6081e6088

A

20

B

50

Stage III Slope=3.0 cm-1/strain%

40 -1

-1

ΔG' (cm )

10 ΔG' (cm )

6085

0 -1

Slope=0.34cm /strain% -10

30

Stage II Slope=20.7 cm-1/strain%

20

Stage I Slope=38.1 cm-1/strain%

10 0

-20 0

0.5

1

Strain (%)

1.5

0

2

0.5

1

1.5

2

Strain (%)

C

0

-1

ΔG' (cm )

Stage I Slope=-25.8cm-1/strain% -10

Stage II Slope=-4.3cm-1/strain%

-20 0

0.5

1 Strain (%)

1.5

2

Fig. 5. G0 -band shift (DG0 ) of different CNTs embedded in an epoxy matrix in tension (A: P-CNT; B: NH2-CNT; C: Si-CNT).

known to be three main interfacial mechanisms between the polymer matrix and CNTs [11,14]. The non-linear Raman shifts observed in Fig. 5B and C are a reflection of different bonding mechanisms: a linear wavenumberestrain relationship with a high slope corresponds to the load transfer due to covalent bonding (Stage I in Fig. 5B and C), whereas a low slope at high strains (>0.5% strain in the nanocomposites containing Si-CNT, and >1.1% strain in the

nanocomposites containing NH2-CNT) suggests a weak mechanical or physical interactions between CNTs and polymer matrix (Stage III in Fig. 5B and stage II in Fig. 5C). Whereas the intermediate stage observed in Fig. 5B (Stage II) suggested the combination of covalent and non-covalent bonding, or a transition of load transfer from chemical bonding to physical interactions. In summary, major findings from Fig. 5 are that the initial G0 -band slope (<0.5% strain) of the

Fig. 6. Schematic of interfacial molecular structures between functionalized CNTs (A: NH2-CNT; B: Si-CNT) and epoxy matrix, and C): Deformation of CNTs embedded in the matrix in tension.

6086

P.-C. Ma et al. / Polymer 53 (2012) 6081e6088

Table 2 Mechanical properties of bulk materials with similar structure to interface. Property

Epoxy (cured by EDA)

Silicone

Tensile modulus (GPa) Tensile strength (MPa) Elongation at break (%)

2.7e2.9 20e39 1.8e3.2

(6.5e47)  103 0.3e2.5 >100

nanocomposites containing NH2-CNT was higher than those containing Si-CNT, i.e. 38.1 vs. 25.8 cm1/strain%, and the strain corresponding to the obvious shift of G0 -band was higher, i.e. 1.1% vs 0.5%. It follows then that NH2-CNT exhibited a better load transfer capability than Si-CNT in polymer matrix. It was shown that the shifts of Raman bands greatly depended on the deformation of CNTs embedded in the polymer matrix, and the G0 -band would shift downwards when the CNTs were under tension, and vice versa [14]. In view of this, it can be concluded that the different deformation behavious of CNTs in response to the same load applied to the nanocomposites originate from the functionalities on CNT surface, which may result in different structures when reacted with precursors of epoxy matrix (either DGEBA or m-PDA). Because the CNTs embedded in the matrix have largely different dimensions and are random dispersed in all directions, they can be loaded in both tension and compression, or even a combination of the two depending on their orientation relative to the loading direction although a uniaxial tension is applied. The randomness of CNT dimensions and orientations have another implication in that the Raman signals can represent only an average of numerous, randomly-dispersed CNTs. However, the averaged Raman signals can be used as representative to show the deformation of CNTs that are parallel to the tensile direction [11]. In this study, the polarized laser was arranged parallel to the tensile direction of the samples so that the in-situ response of CNTs can be detected from the exactly same spot on the nanocomposites. Previous studies indicated that there was almost no CNT shift in Raman spectroscopy when the nanotube axis was perpendicular to the polarization direction of laser [14,29]. In light of the above findings and discussions, we are proposing probable molecular structures of the interfaces between the functionalized CNTs and the epoxy matrix, as schematically shown in Fig. 6. The amino groups attached on NH2-CNT can react easily with the epoxide groups of DGEBA (Fig. 6A), creating an interfacial structure similar to that of matrix. Whereas the interface between the matrix and Si-CNT is characteristic of eSieOeSie structure and ring-opening reaction between the epoxide groups on Si-CNT and amine hardener (m-PDA) (Fig. 6B). The resultant molecular structures affect the response of CNTs embedded in the matrix under mechanical load (Fig. 6C): an absent of interface or poor bonding between the matrix and P-CNT is expected, as the outermost layer of nanotubes is chemically inert and smooth, and these CNTs slip easily from the matrix when subjected to a tensile load. The interaction

between NH2-CNT and the matrix is enhanced due to the functionalization, and CNTs can deform together with the matrix without slippage. However, the similarity in molecular structure of the interface material to the matrix can lead to CNTs under compression because of the Poisson contract of matrix (the Poission ratio of epoxy is about 0.4), resulting in an upshift of CNT G0 -band [9,16]. In contrast, the Si-CNT is modified with a higher degree of functionalization, creating a thicker interface region than that of NH2-CNT. Under the same tensile load, the Si-CNT is deformed/sheared along the loading direction due to the non-uniform thickness of silane on CNT surface (Fig. 4C) and the difference in elastic properties of matrix and interface, resulting in elongation of eCeCe bonds and consequent downshift of the Raman G0 -band [30]. It is likely that that the mechanical properties of the interface may govern the deformation behavior [31]. It is expected that the interface created between NH2-CNT and epoxy is similar to that of the matrix material, whereas the corresponding interface between Si-CNT and epoxy may be similar to a silicone (eSieOeSie structure). While it is impossible to measure them, the properties of bulk materials with similar structures to that of interfaces offer an insight for comparison. Table 2 summarizes the mechanical properties of these bulk materials [32e35]. The tensile modulus of epoxy (cured by EDA) is above 2.5 GPa, which is much higher than that of silicone. The lower mechanical properties of the interface may lead to an easier debonding of Si-CNT from the matrix, which is consistent with the lower strain of about 0.5% corresponding to the initially high slope of the G0 -band shift found in the nanocomposites containing Si-CNT (Fig. 5C). The aforementioned hypotheses were partly supported by the SEM images taken from the fracture surfaces of the nanocomposites, as shown in Fig. 7. Long CNTs were pulled out from the matrix in the nanocomposites containing P-CNT (Fig. 7A), suggesting a weak CNT-matrix interaction, which is quite agreeable with the results in Fig. 5A and the corresponding discussions. In contrast, the pull-out CNTs were very short and accompanied by the deformation of neighbouring matrix in the nanocomposites containing functionalized CNTs (Fig. 7B and C), confirming the strong covalent interactions between these CNTs and the matrix. The functionalization brought an additional benefit of CNT dispersion: P-CNTs were severely agglomerated in the matrix, whereas these agglomerates were almost absent in the samples containing functionalized CNTs. A relevant and important question is whether the excellent mechanical properties of CNTs can be best utilized in nanocomposites through a designed filler-matrix interface. The answer seems to be positive according to the DMA results. Fig. 8 shows the storage modulus and tan d of samples filled with different CNTs, which representing the elastic and dissipated portion of energy in nanocomposites, respectively. Compared with neat epoxy, the nanocomposites filled with P-CNT showed a slight decrease in storage modulus in the glassy region. In contrast, those containing

Fig. 7. Typical SEM images taken from fracture surface of nanocomposites containing different CNTs (A: P-CNT; B: NH2-CNT; C: Si-CNT).

P.-C. Ma et al. / Polymer 53 (2012) 6081e6088

A Storage Modulus (Pa)

that in the nanocomposites containing Si-CNT. However, the tan d was close to each other when the temperature was near the glass transition temperatures (Tg, as determined from the maxima of tan d in Fig. 8B), and an increased Tg was observed for Si-CNT nanocomposites as compared with those containing NH2-CNT. This is due to the higher degree of functionalization in Si-CNT, which inhibited the movement of matrix chains, especially at elevated temperatures.

3.00E+09 A: Neat epoxy B: P-CNT C: NH2-CNT D: Si-CNT

2.50E+09 2.00E+09 1.50E+09

D B

1.00E+09

4. Conclusions

A

5.00E+08

C

0.00E+00 30

50

70

90 110 Temperature (oC)

130

150

170

B 0.5 A: Neat epoxy

C

B

B: P-CNT

0.4

C: NH2-CNT D: Si-CNT

0.3

Tan δ

A

0.2 D

0.1

0 30

6087

50

Fig. 8. Thermo-mechanical nanocomposites.

70

90 110 Temperature (oC)

properties

(A:

Storage

130

modulus;

150

B:

170

Tan

d) of

functionalized CNTs exhibited much increased moduli in the whole temperature range studied (Fig. 8A). The elastic storage modulus of the nanocomposites filled with P-CNT, Si-CNT and NH2-CNT was 1.72, 1.93 and 2.59 GPa, respectively, at ambient temperature. These observations suggested that the interfacial interactions between functionalized CNTs and epoxy matrix were enhanced significantly, and particularly NH2-CNT demonstrating the highest improvement among the three as evidenced by the most pronounced increase in storage modulus in the glass region. However, the storage modulus of the nanocomposites filled with Si-CNT outperformed that of NH2-CNT at elevated temperatures possibly due to the higher degree of functionalization in Si-CNT, which discouraged the movement of matrix chains along CNTs after cure reaction [36]. The tan d value, a ratio of the loss modulus to the corresponding storage modulus, is an important parameter to study the energy loss due to the filler-matrix interactions, thus offering an indirect way to evaluate the load transfer behavior in CNT filled nanocomposites. The tan d for the neat epoxy was consistently lower than those of the nanocomposites containing CNTs at the glassy region. Broader glass transition peaks and shoulders were noted for the nanocomposites containing functionalized CNTs (Fig. 8B). These phenomena can be attributed to the facts that: i) the functionalized CNTs promoted the cross-linking reactions of epoxy and hardener, effectively discouraging the movement of molecular chains [36]; and ii) the covalent bonds enhanced the interactions between the functionalized CNTs and epoxy matrix, resulting in an increase in energy dissipation in the nanocomposites [37]. The consistently higher value of tan d found in NH2-CNT based nanocomposites implied the stronger interactions between CNTs and matrix than

The behavior of load transfer in CNT/epoxy nanocomposites was studied by examining the in-situ deformation of CNTs using Raman spectroscopy. Functionalization of CNTs offers an effective way to promote the load transfer from the polymer matrix to CNTs through the interface. Major findings arising from this study are highlighted as follows: 1) The behavior of load transfer from the polymer matrix to CNTs is dominated by the nature of functionalities on CNTs, which determined the molecular structures and the properties of the interface formed in-between. 2) Functionalized CNTs displayed more prominent Raman G0 -band shifts with steeper slopes in the linear wavenumberestrain relationship than that of pristine CNTs. 3) Functionalized CNTs showed bi-linear Raman G-band shifts as a reflection of two different bonding mechanisms at different stages of tensile loading: the higher slope at low strains corresponds to the interfacial load transfer due to covalent bonding, whereas the lower slope at high strains corresponds to mechanical or physical bonding. 4) The slope of G0 -band shift as a function of tensile strain was either positive or negative, depending on the type of functional groups on CNTs: the NH2-CNT/epoxy nanocomposites showed a G0 -band upshift, whereas the nanocomposites containing SiCNT had a G0 -band downshift. 5) An efficient load transfer from matrix to functionalized CNTs corresponds to enhanced properties of nanocomposites, as confirmed by the thermo-mechanical testing of nanocomposites. Acknowledgments This project was supported by the Research Grant Council of Hong Kong SAR (Project No. 614505) and Finetex-HKUST R&D Centre. Part of research was conducted when PC Ma was supported by the One Hundred Talent Program of Chinese Academy of Science (CAS) and the Alexander von Humboldt (AvH) Foundation. References [1] Wong P, Paramsothy M, Xu XJ, Ren Y, Li S, Liao K. Polymer 2003;44:7757e64. [2] Velasco-Santos C, Martinez-Hernandez AL, Castano VM. Compos Interfaces 2005;11:567e86. [3] Zheng QB, Xue QZ, Yan KY, Gao XL, Li Q, Hao LZ. Polymer 2008;49:800e8. [4] Ganesan Y, Lou J. Jom-J Min Met Mater Soc 2009;61:32e7. [5] Hirsch A. Angew Chem Int Ed 2002;41:1853e9. [6] Tasis D, Tagmatarchis N, Bianco A, Prato M. Chem Rev 2006;106:1105e36. [7] Ma PC, Siddiqui NA, Marom G, Kim JK. Compos A 2010;41:1345e67. [8] Kim JK, Mai YW. Engineered interfaces in fiber reinforced composites. Oxford: Elsevier; 1998 [Chapter 2]. [9] Schadler LS, Giannaris SC, Ajayan PM. Appl Phys Lett 1998;73:3842e4. [10] Hadjiev VG, Iliev MN, Arepalli S, Nikolaev P, Files BS. Appl Phys Lett 2001;78: 3193e5. [11] Zhao Q, Wagner HD. Phil Trans R Soc Lond A 2004;362:2407e24. [12] Ruan SL, Gao P. Polymer 2006;47:1604e11. [13] Roy D, Bhattacharyya S, Rachamim A, Plati A, Saboungi ML. J Appl Phys 2010; 107:043501. 1-6. [14] Gao Y, Li LY, Tan PH, Liu LQ, Zhang Z. Chin Sci Bull 2010;55:3978e88.

6088

P.-C. Ma et al. / Polymer 53 (2012) 6081e6088

[15] Gong L, Kinloch IA, Young RJ, Riaz I, Jalil R, Novoselov KS. Adv Mater 2010;22: 2694e7. [16] Srivastava I, Mehta RJ, Yu ZZ, Schadler L, Koratkar N. Appl Phys Lett 2011;98: 063102e5. [17] Wang S, Liang R, Wang B, Zhang C. Chem Phys Lett 2008;457:371e5. [18] Mu M, Osswald S, Gogotsi Y, Winey KI. Nanotechnology 2009;20:335703e11. [19] Li J, Gao Y, Ma W, Liu L, Zhang Z, Niu Z, et al. Nanoscale 2011;3:3731e6. [20] Sham ML, Li J, Ma PC, Kim JK. J Compos Mater 2009;43:1537e64. [21] Ma PC, Kim JK, Tang BZ. Carbon 2006;44:3232e8. [22] Ma PC, Mo SY, Tang BZ, Kim JK. Carbon 2010;48:1824e34. [23] Ma PC, Kim JK, Tang BZ. Compos Sci Technol 2007;67:2965e72. [24] Simmons JM, Nichols BM, Baker SE, Marcus MS, Castellini OM, Lee CS, et al. J Phys Chem B 2006;110:7113e8. [25] Klinke C, Chen J, Afzali A, Avouris P. Nano Lett 2005;5:555e8.

[26] Zhou W, Vavro J, Nemes NM, Fischer JE, Borondics F, Kamaras K, et al. Phys Rev B 2005;71:205423. 1-7. [27] Kang D, Park N, Ko JH, Bae E, Park W. Nanotechnology 2005;16:1048e52. [28] Berlin T, Epron F. Maters Sci Eng B 2005;119:105e18. [29] Frogley MD, Zhao Q, Wagner HD. Phys Rev B 2002;65:113413. 1-4. [30] Liu ZF, Zhang J, Gao B. Chem Commun 2009;45:6902e18. [31] Yang S, Yu S, Kyoung W, Han DS, Cho M. Polymer 2012;53:623e33. [32] Myung IH, Chung IJ, Lee JR. J Korean Ind Eng Chem 1999;10:990e5. [33] Xu DM, Zhang KD, Zhu XL. J Appl Polym Sci 2006;101:3902e6. [34] Material properties database. http://www.makeitfrom.com. [accessed in March, 2012]. [35] Bayley GM, Hedenqvist M, Mallon PE. Polymer 2011;52:4061e72. [36] Hwang GL, Shieh YT, Hwang KC. Adv Funct Mater 2004;14:487e91. [37] Sun L, Gibson RF, Gordaninejad F, Suhr J. Compos Sci Technol 2009;69:2392e409.