Raman spectroscopy of polymer–carbon nanotube composites
Over the past 30 years Raman spectroscopy has transformed the field of composite micromechanics and is being successfully applied to the field of nanotube-reinforced polymers. Under the effects of stress, careful study of the Raman spectrum of carbon nanotubes provides a unique insight into various physical phenomena in nanocomposites. This chapter demonstrates the power of the Raman signature of carbon nanotubes as a detector of bulk matrix defects, the occurrence of polymer phase transitions, and the nanotubes’ own orientation change with respect to applied stress, stress profiles from Raman-insensitive fibers, and dual information about improvements in the stress transfer ability and about nanotube wall structure degradation due to the surface treatment itself. Remaining challenges are described.
The interesting history of the discovery of carbon nanotube (CNT) structures, which is not well known, has recently been documented by Monthioux and Kuznetsov (2006). These remarkable materials were experimentally observed for the first time by Radushkevich and Lukyanovich (1952), whose work remains largely unknown to the scientific community. It was only much later, following the discovery of fullerenes by Kroto et al. (1985) that CNTs became an active and fascinating field of research, thanks to a celebrated report by Iijima (1991) on the synthesis of multi-walled carbon nanotubes (MWCNTs). The formation of single-walled carbon nanotubes (SWCNTs) was later reported almost simultaneously by Iijima and Ichihashi (1993) and by Bethune et al. (1993).
The structure, topology and size of carbon nanotubes are the source of their outstanding mechanical and electronic properties, and of a whole range of promising applications (Hu et al., 2006; Chou et al., 2010). The latter include their use as electron field emitters for vacuum microelectronic devices, nanoprobes at the tip of an Atomic Force Microscope (AFM), efficient supports in heterogeneous catalysis, a medium for lithium and hydrogen storage, and much more. Nanotubes have been embedded into various materials to produce composites with modified electrical conductivity, magnetic properties and optical properties. Nanotubes are also promising candidates as a mechanical reinforcement phase in composite materials. Compared to micron scale reinforcing fibers, however, nanotube reinforcement involves a number of important differences, which have consequences on the performance of the corresponding nanocomposites, as well as on the behavior and use of the nanotube itself. For example, distinctive mechanical effects arise, such as large increases in Young’s modulus and strength below a certain tube diameter (Arinstein et al., 2007; Sui and Wagner, 2009), resulting from molecular confinement: such effects are genuine nano-effects. CNTs can be spread in polymers and used either as reinforcement, or as molecular sensors, or both. Unlike micron-size reinforcement, nanotubes possess geometrical chirality, which has special significance regarding their physical (and possibly mechanical) properties. Raman spectroscopy, which beyond doubt has transformed the field of composite micromechanics since the late 1970s, has a special role to play in CNT-reinforced polymers, in many ways. This is the main focus of the present chapter.
Light scattering is the result of the relaxation of the dipole induced in a molecule by the application of the electric field that is part of an incident light beam. The scattering of light by matter is either elastic when the scattered light has the same frequency as the incident light, or inelastic when the frequency changes. In the former case the phenomenon is termed Rayleigh or Mie-Tyndall (‘classical’) scattering, and in the latter case it is termed Raman or Brillouin scattering. The intensity of classical scattering is proportional to the 4th power of the frequency of the incident light, thus, if white light is used, the blue end of the spectrum is scattered more strongly than the red end (which explains the blue color of the clear sky which arises from the Rayleigh scattering of white sunlight by the molecules in the atmosphere). It can easily be shown that the induced polarizability P of a molecule (the ease with which the electron cloud can be distorted by an applied electric field E) is the sum of three terms (Woodward, 1967; Turrell, 1996):
where v0 is the frequency of the exciting light wave, t is time, and v is the frequency of the vibrating molecule. The first term on the right-hand side of Equation 14.1 predicts that a dipole is induced (and thus scattering will occur) at the same frequency as the exciting radiation (v0), which constitutes elastic (Rayleigh) scatter. The second and third terms indicate that elastic scattering will also be accompanied by inelastic scatter, namely radiation at decreased frequencies v0 – v (known as Stokes Raman scatter) and at increased frequencies v0 + v (anti-Stokes Raman scatter). Rayleigh scatter is weak (about 10− 3 of the intensity of the incident exciting radiation), Raman scattering is much weaker (10− 2 to 10− 3 of lthe intensity of Rayleigh scattering), and the intensity of the Stokes scatter is always greater than the anti-Stokes equivalent bands. Thus, intense monochromatic light sources and sensitive detectors are required. Details of the history of the inelastic scattering effect, observed for the first time by Raman and Krishnan (1928), can be found in articles by Turrell (1996) and Woodward (1967).
A remarkable phenomenon generated a flurry of activity in the early 1980s in the area of composite micromechanics: It was observed that the frequencies of specific Raman bands of polymers possessing high structural perfection decrease upon application of a tensile stress. This stress- (or strain-) induced frequency shift was first reported by Mitra et al. (1976) for monocrystalline polydiacetylene fibers and subsequently by others (Batchelder and Bloor, 1979; Galiotis et al., 1983; Wu et al., 1989; Young, 1993, 1997). The reason why this discovery was so significant is that it became possible to obtain exceptional insight into the effect of macroscopic deformation on the molecules in the probed polymer under a variety of conditions and, in the field of composites, to monitor and quantify with so far unequalled precision the local values and variations of key mechanical parameters. Prominent among these was the mapping of the stress profile σ(x) along the length (x) of fibers such as aromatic poly(aramid) (Kevlar) or graphite and, most importantly (for composites technology), the interfacial adhesion (or shear) strength τ(x) = (r/2)(dσ/dx) (where r is the fiber radius) between the fibers and polymer matrices, as well as the fiber–fiber interactions due to increasing far-field stresses resulting in isolated but progressively interacting fiber breaks. As an example of the latter issue, micro-Raman spectroscopy was employed to map the strain along accurately positioned, individual fibers in a polymeric matrix (Grubb et al., 1995; Wagner et al., 1996; Van Den Heuvel et al., 1996). The objective was to detect possible load redistribution and sharing effects from a broken fiber onto its (still intact) near neighbors, and thereby to estimate the stress concentration factors (SCF) in these. The effect of variable inter-fiber distance upon the SCF was also examined by means of carefully prepared microcomposites, or model composites (Wagner and Steenbakkers, 1989; Jones and DiBenedetto, 1994). See also the more recent work of Kim et al. (2009). Quantitative examples of the amount of frequency shift per unit amount of strain may be found in Young (1993, 1997) for a number of rigidrod polymer fibers and non-polymeric micron-scale fibers such as carbon, silicon carbide and alumina. The magnitude of the observed decrease in frequency varies between 5 and 20 cm− 1 per percent of strain (see Table 1 in Young, 1993). The reader is also referred to Young and Eichhorn (2009) for an interesting discussion of the use of two alternative models, based on uniform stress or uniform strain, regarding the change in Raman wavenumber with applied stress or strain in polymeric fibers. The same effect and method were also applied to the study of some typical carbon and ceramic fibers, in ceramic- or metal-matrix composites (Colomban, 1999; Gouadec et al., 2001).
Once CNTs were discovered, it was natural, and only a matter of time, until the strain-induced Raman band shift just described for micron size fibers possessing high structural perfection would be investigated for CNTs.
The Raman spectra of carbon nanotubes are shown in Figs 14.1 (a) and 14.1 (b). They are not greatly different from that of a high modulus graphite fiber. However, a key Raman feature of SWNTs is the radial breathing mode (RBM) cluster in the 160 to 300 cm− 1 region, associated with a symmetric movement of all carbon atoms in the radial direction. Both theoretical and experimental investigations, mainly by Young and colleagues (Lucas and Young, 2004a, 2004b, 2007a, 2007b), have shown that the frequency of the RBM is inversely proportional to the diameter of an individual nanotube. However, low-frequency Raman modes have also been observed in MWNTs samples, although their origin has not been clearly identified. Calculations suggested that van der Waals interactions between concentric tubes of different diameters could lead to such low frequency breathing modes, the effect of the interactions being to upshift the Raman modes compared to those of individual tubes (Stephan et al., 2002).
14.1 (a) Raman spectra of SWNTs in air at T = 295 K using three excitation wavelengths. The changes in intensity, shape, and position of the spectra features are due to diameter-selective Raman scattering. (b) Comparison of Raman spectra of SWNTs prepared by the CVD and AD methods, and of MWNTs prepared by the CVD method, in air at T = 295 K. From Lourie et al. (1999).
The D Raman band of carbon nanotubes observed between 1250 and 1450 cm− 1 has a linear dependence on the laser excitation energy (Lourie et al., 1999). This band is activated in the first order scattering process of sp2 carbons by the presence of in-plane substitutional hetero-atoms, vacancies, grain boundaries or other defects (D stands for ‘defects’) and by finite size effects, all of which lower the crystalline symmetry of the quasi-infinite lattice. The G′ (or D*) band is the second order overtone of the D band. The locations of D and G′ Raman bands of carbon nanotubes depend linearly on the laser excitation energy and this dispersion relation of the D and G′ band of nanotubes is similar to that in other sp2 carbons, though with some distinct characteristic behavior that is specific to nanotubes. A shift of the D and G′ bands of carbon nanotubes has been observed in different cases. As seen in Fig. 14.1 (b), the presence of a strong D band in the spectrum of MWNTs reflects the much greater density of defects in these tubes, compared to SWNTs. Also, SWNTs prepared by the arc discharge (AD) method are of better quality (less defects) than those prepared by chemical vapor deposition (CVD).
Raman spectroscopy can provide unique information about vibrational and electronic properties of CNTs. It can also be used to identify materials through the characteristic vibrations of certain structures. Because the Raman intensity of a vibration or phonon in a crystal depends on the relative directions of the crystal axis and the electric wave polarization of the incident and scattered light, it may also be used to determine the orientation of CNTs in polymer matrices or within CNT bundles. Reviews of Raman spectroscopy as a sensitive probe of SWNT properties are given in Zhao and Wagner (2004) and Dresselhaus et al. (2005). Raman spectroscopy has been used to determine the diameter of SWNTs, the diameter distribution of SWNT bundles and the structural properties of nanotubes (Dresselhaus et al., 2002a). The unique 1D molecular nature of SWNTs makes the resonance Raman technique an extremely useful and accurate method for the identification of the diameter and chirality of an individual SWNT. If the (n, m) vector is known, the dependence of all the features of the spectra on the diameter, chiral angle, laser excitation energy and other parameters can be worked out in detail. Therefore, the spectrum of SWNT bundles can be interpreted and the effect of nanotube–nanotube interactions can be deduced (Dresselhaus et al., 2002b). Raman spectroscopy has been used to identify the structure of individual SWNTs (Souza Filho et al., 2001), to investigate the diameter distributions of bulk samples (Rao et al., 1997), to study the transfer of stress to nanotubes during the deformation of nanocomposites (Cooper et al., 2001), and to follow the effect of stress upon the electronic structure of SWNTs (Lucas and Young, 2007b). Raman spectroscopy was also used to non-invasively determine the ratio of metallic SWNTs to semiconducting SWNTs (Jorio et al., 2005). Selected applications of the Raman technique specifically dealing with CNT-based nanocomposites will be described in the next section.
There is wide-ranging conviction that carbon nanotubes are promising reinforcement materials for a new class of nanocomposites with much higher stiffness, strength and toughness, owing to the great strength of the sp2 bonds in the graphite structure of CNTs, since all basal planes run approximately parallel to the CNT axis. In comparison, however, commercial graphite fibers contain various types of structural defects and misaligned planes, and by contrast, the quasi-perfect structure of CNTs opens a new route to super-strong nanofibers. Nonetheless, the potentially outstanding mechanical properties of CNTs will be of little engineering value unless nanotubes can either form macroscopic cables (Atkinson et al., 2007; Koziol et al., 2007), or can be incorporated into a matrix to form a (preferably unidirectional) composite. A large amount of research has appeared in the literature over the past decade, dealing with possible significant improvements in the mechanical and electrical properties of CNT-reinforced polymers. In the present chapter we only focus on those works which include a significant contribution of Raman spectroscopy. We show that when CNTs are embedded in polymers, Raman spectroscopy may achieve a number of unique functions, including as a detector of CNT orientation, as a molecular sensor of structural defects and of second phase materials (such as micron-size fiber reinforcement), as a sensor of the CNTs’ own chirality and diameter, or as a quantitative probe of CNT–polymer adhesion strength, and more. This is now reviewed in sequence.
G′ Raman wavenumber shifts (as mentioned earlier, the G′ band is sometimes termed D*) are observed when SWNTs are dispersed in liquids, the extent of shifting compared with the initial G′ wavenumber in air being dependent on the nature of the liquid. A correlation is found between the cohesive energy density (CED, or, more loosely, the surface tension) of a medium and the G′ wavenumber of the dispersed CNT (Wood et al., 2000). To corroborate this, further research was done by using a diamond anvil cell (DAC) to apply hydrostatic stress while the Raman response of both the G and G′ peaks was monitored in situ. A striking similarity was indeed observed between the hydrostatic pressure experiments, on the one hand, and the molecular pressure experiments with different media, on the other, regarding the degree of Raman wavenumber shift (Wood et al. 1999, 2000). The fact that both sets of data agree with each other demonstrates the sensitivity of nanotubes to molecular pressure from the surrounding medium, as measured either by CED or by mechanical pressure using a DAC. A similar study was performed by Cooper et al. (2001), where SWNT material was pressurized in a diamond anvil cell and it was found that the G′ Raman band shifts to a higher wavenumber with increasing pressure, with initial Raman shift of 23 cm− 1/GPa.
Furthermore, Raman spectroscopy has been used to probe the interaction between polymers and nanotubes in CNT-based composites. Generally, such interaction is reflected by a peak shift or a peak width change. In the field of fiber composite materials, it has been known for more than two decades that the application of a mechanical strain to fibers (in air, thus without any polymer matrix) such as carbon or Kevlar results in shifted frequencies of the Raman bands (usually the G′ band), which are directly related to the interatomic force constants. A similar effect is observed when carbon nanotubes are embedded in polymers. Correlating such shifts with the applied strain, through a calibration procedure, leads to the determination of local stress profiles in the embedded fibers. In particular, a tensile strain transferred from the polymer matrix to SWNTs results in a downward shift of the G′ Raman wavenumber of the nanotubes (Cooper et al., 2001). The stress transfer issue is further discussed in Section 14.5.6.
An empirical linear relationship exists between the SWNT G′ wavenumber shift and the applied elastic strain (Wood et al., 2001). If the nanotube G′ wavenumber difference between zero strain and the applied strain (ε) is defined as the Raman wavenumber shift Δwn, then the empirical slope m of the wavenumber–strain relation is
A mechanical stress–strain curve can be recorded simultaneously with the Raman measurement. In the elastic regime of a tensile test, Hooke’s law of linear elasticity states that the stress σ and the strain ε in the polymer are related by
Thus, in parallel to the standard mechanical stress–strain curve for the matrix material, the Raman signal captured from the embedded SWNT under strain permits the construction of a spectroscopic stress–strain curve, using the stress calculated by Equation 14.4. Experiments show that m varies when nanotubes are embedded in different polymer matrices (Wood et al., 2001; Zhao et al., 2001b; Frogley et al., 2002), and that, for a given matrix, m is temperature dependent. Figure 14.2 shows the Raman wavenumber–strain response of SWNTs embedded in polyurethane acrylate (PUA) at 298 K (room temperature) and at 235 K. The initial part of both data sets is approximately linear within the elastic strain region (up to ~ 1.5%). The values of m at both temperatures are 467 cm− 1/ε and 909 cm− 1/ε, respectively (Wood et al., 2001). As the tensile strain increases, the wavenumber stabilizes at a plateau value of about 2622 cm− 1 where it becomes insensitive to increasing strain, for both temperatures. The temperature dependence of the m values of the SWNTs in PUA can be attributed to the fact that Young’s modulus (E) of the polymer is a temperature-dependent parameter, E(T), in other words, Equation 14.4 becomes:
14.2 The wavenumber–strain response of SWNTs embedded in PUA. Symbols: data at 298 K (room temperature) and data at 235 K. From Zhao and Wagner (2004).
In Figure 14.3, the solid lines represent the mechanical stress–strain curves of PUA at 235 K and 298 K, and the symbols are the spectroscopic stress–strain signatures obtained from the Raman data in Fig. 14.2, using the values of m(T), E(T) at the corresponding temperatures. Within the linear region, the mechanical and spectroscopic curves are in perfect agreement, and beyond about 1.5%, they deviate from each other, likely because of local matrix yielding beyond which the CNT strain does not follow the matrix strain anymore. The yield stress σy at which the mechanical stress–strain curve becomes non-linear is also indicated in Fig. 14.3.
14.3 Independently obtained mechanical and spectroscopic stress–strain curves. The latter was determined from the Raman shift data presented in Fig. 14.2. Symbols: data at 298 K (room temperature) and data at 235 K. σy is the yield stress of the polymer according to the mechanical stress–strain curve. From Zhao and Wagner (2004).
To demonstrate that embedded nanotubes can accurately map stress fields, a simple classical elasticity problem with a known analytical solution is first selected, namely, that of a circular hole in a plate under uniaxial tension (Dally and Riley, 1985). Referring to Fig. 14.4 and Equation 14.6, the solution shows that the maximum value of the σyy stress component occurs at the boundary of the hole (along the x axis, at the ends of the diameter perpendicular to the direction of applied tensile stress) and that, at that point, it is three times larger than the applied stress:
where a is the hole radius and σ0 is the applied stress. From an experimental viewpoint, the first issue that needs to be addressed concerns the orientation of the nanotubes within the (polymer) plate. When the tubes are randomly oriented in the plate, in the case of a uniaxial stress, as in simple mechanical tension, Poisson contraction occurs in the transverse direction. Thus, some nanotubes will be under compression while others will be under tension, resulting in a mixed signal. Aligning the nanotubes is thus preferable – although generally more complicated in practice – to measure the specific components of the complex strain or stress distributions. A simple shear flow method can be developed to orient nanotubes in a polymer such as polyurethane acrylate (PUA). The elastic strain dependence of the G′ wavenumber shift was indeed measured with the loading parallel and perpendicular to the flow direction. A significant difference between the two cases was found, indicating that this orientation method is effective (Wood et al., 2001). However, regardless of whether SWNTs are or are not oriented in PUA, the stress distribution in the plate around the circular hole under uniaxial tension can be mapped, as shown by Zhao et al. (2002). Figure 14.5 shows the normalized matrix stress data (calculated from the experimental Raman shift) along the x axis, starting at the edge of a circular hole, under three increasingly higher levels of applied stress. The experimental results fit the linear elastic solution (Equation 14.6) very well, showing that SWNT sensors indeed provide accurate quantitative stress field distribution information at a stress discontinuity (Zhao et al., 2001c).
14.5 Normalized stress along the x axis (refer to Fig. 14.4) from the edge of a circular hole, based on the G′ peak shift of SWNTs in a UV cured urethane-acrylate polymer. Applied loads (σ0.) of 4, 6, and 8 MPa were employed. The solid line is the linear elastic solution given by Equation 14.6. Adapted from Figure 2 in Zhao et al. (2001b), with permission from the American Institute of Physics.
However, it is not always straightforward to induce shear flow and orient nanotubes in a polymer. An alternative is to use plane polarized Raman measurements. In this case, the measured G′ Raman peak intensity from a single nanotube is high when the polarization direction is parallel to the nanotube axis, and low when the polarization direction is perpendicular to it (Saito et al., 1998; Duesberg et al., 2000; Gommans et al., 2000; Wood et al., 2001; Zhao et al., 2002). This implies that randomly oriented SWNTs in a polymer can perform as strain sensors if polarized Raman spectroscopy is used – as an alternative to using unpolarized Raman spectroscopy with oriented SWNTs – provided that a sufficiently high density of nanotubes are present in the polarization direction. The advantage of this technique is that, when nanotubes are used as strain sensors, randomly oriented nanotubes combined with polarized Raman may be used to detect the stress or strain distribution in any direction in the polymer matrix. The G′ peak position of SWNTs embedded in PUA was measured as a function of tensile strain, for both uniaxial and random tube orientations (Frogley et al., 2002). The results showed that the polarized Raman can indeed ‘select’ those nanotubes that are lying in one direction.
Further convincing evidence for the exploitation of nanotubes as mechanical sensors, again using polarized Raman, was demonstrated for complex stress state systems by monitoring the stress in the matrix in the vicinity of a fiber end (Zhao et al., 2001a, 2001c) and of a single fiber break (Zhao and Wagner, 2003).
The discussion thus far shows that SWNTs indeed perform as sensors to detect the elastic stress or strain in a polymer matrix. Such a microscale Raman sensing technique has important practical importance, for example, to measure the matrix stress distribution in the vicinity of fibers to detect or predict the onset of failure in composite materials. As already mentioned, Raman spectroscopy has been successfully used since the early 1980s to monitor the deformation of specific fibers such as aramid, SiC and carbon, since specific Raman bands of these fibers are indeed sensitive to the applied strain (Mitra et al., 1976; Batchelder and Bloor, 1979; Galiotis et al., 1983; Robinson et al., 1987). Correlating these Raman shifts with the applied strain leads to an evaluation of the stress distributions in the fibers at a micron scale, and provides a mean to deduce the interfacial shear stress (Galiotis and Batchelder, 1988; Galiotis et al., 1984; Nielsen and Pyrz, 1999; Wagner et al., 2000). In reality, the micro-Raman technique cannot be universally applied, since some fibers do not have strain-sensitive Raman peaks, glass fibers being the best-known example. In other words, it is impossible to perform in-situ measurements of the stress (or strain) distribution in glass fibers. Moreover, since most polymers do not have strain sensitive Raman bands, it is practically impossible to use the Raman spectra of those materials to detect stress (or strain) distributions around discontinuities in polymers. Such measurements become possible, however, by dispersing SWNT sensors in Raman insensitive polymer matrices, or in the sizing layer of Raman insensitive glass fibers. These two cases are now discussed.
The stress field in a polymer matrix in the vicinity of a single glass fiber was mapped on the micron scale by using the strain response of the Raman spectrum of SWNTs embedded in the matrix (Zhao et al., 2001a). A stress concentration zone was observed around the fiber end. Referring to Fig. 14.6, the applied stress value is recovered at a radial distance of 5 or 6 fiber radii from the fiber end (along A–B), whereas axially the stress returns back to the applied value only after about 2 fiber radii (along C–A). Moreover, again for a single glass fiber embedded in a polymer, the tangential thermal residual stress in the vicinity of the fiber was found by Raman spectroscopy to be in satisfactory agreement with a standard two-phase concentric cylinder model (Wagner, 1996; Wagner and Nairn, 1997; Zhao et al., 2001c).
14.6 Strain mapping around a glass fiber: all the measurements were performed with the Raman polarization direction parallel to the fiber axis. From Zhao et al. (2001c).
As an alternative to spreading SWNTs in the matrix around a (Raman-insensitive) glass fiber, Sureeyatanapas and Young (2009) adopted a different approach by preparing a model glass-fiber/epoxy composite with SWNTs incorporated as a strain sensor on the fiber surface. The SWNTs were distributed along the fiber surface either by dispersing them in an amino-silane coupling agent or coating with an epoxy resin solution containing the SWNTs. Point-by-point mapping of the fiber strain in single fiber fragmentation tests was then performed and the interfacial shear stress distribution along the fiber length could be determined. The behavior was found to be consistent with the classical shear-lag model. The effects of SWNT type and preparation procedure on the sensitivity of the technique were evaluated and optimized from single fiber deformation tests. Such SWNT-containing coatings can be used in composites to follow fiber deformation and stress transfer between the matrix and reinforcing glass fibers. This technology has considerable potential for a number of composite systems for which it is not possible to use Raman spectroscopy to follow fiber deformation.
Significant G′ Raman shifts are measured when a nanotube polymer composite is probed at different temperatures under no mechanical tension. Lourie and Wagner (1998a) found that the Raman bands of SWNT shift to a higher wavenumber (~ 15 cm− 1/% strain) when embedded in a thermally cured epoxy resin and cooled to room temperature, due to thermal contraction of the resin. Zhao et al. (2001c) have observed irregularities (or discontinuities) in the wavenumber–temperature plot for both amorphous bisphenol A polycarbonate (PC) and polyurethane acrylate (PUA), in which SWNTs had been embedded. The Raman results were compared with dynamic mechanical thermal analysis (DMTA) data for both polymers, and the sources of the discontinuities were investigated. The concordance found between the DMTA data and the Raman spectral response shows that these irregularities reflect basic polymer phase transitions, namely, glass-transition temperatures and secondary transitions, which may thus be sensed by nanotubes. In other words, this confirms that the Raman spectral response of carbon nanotubes embedded in polymers is sensitive to polymer transitions. In a recent study, De la Vega et al. (2009) used SWNTs to monitor internal stresses developing during the curing process of thermoset materials. In-situ Raman spectroscopy was used to identify chemical and thermal-induced stresses by following the changes in the G′ band versus time and temperature.
Raman spectroscopy combined with mechanical testing provides a way to probe the alignment of SWNTs in composites. Zhao et al. (2001c) prepared specimens using a flow orientation method designed to align SWNTs in the matrix. The matrix was a UV curable urethane acrylate. Raman spectra obtained for specimens cut both parallel and perpendicular to the (average) tube direction were found to be significantly different, as a function of mechanical strain. The Raman shift–strain response for samples loaded perpendicular to the flow direction suggested that nanotube reorientation was achieved upon straining the polymer beyond its yield point. Frogley et al. (2002, 2003) and Frogley and Wagner (2002) then performed a thorough study of nanotube alignment in polymers using polarized Raman spectroscopy, and compared a large amount of experimental results with existing models such as those by Saito et al. (1998), Gommans et al. (2000), and Hwang et al. (2000). Specimens with a good degree of nanotube alignment could be prepared by a simple shear-flow technique. For single-walled carbon nanotubes excited with polarized light of wavelength 632.8 nm, the Raman scattering is resonant so that for a single nanotube, or for tubes oriented perfectly along one direction, the total intensity of the Raman modes is found to vary as cos4(θ) (Frogley et al., 2002, 2003), where θ is the angle between the nanotube and the loading directions. Kao and Young (2005) and Cooper et al. (2001) performed similar studies with randomly oriented SWNTs. Polarized Raman spectroscopy has also been used by other authors to probe the orientation of CNT bundles prepared by an electrophoretic method (Poulin et al., 2002). In another case, nanotubes were oriented into bundles by applying an electric field between a carbon fiber and an ultrasonicated SWNT/N, N-dimethylformamide (DMF) suspension, and polarized Raman spectroscopy was then used to quantify the alignment of nanotubes as a function of the angle θ between the fiber and the polarizer (Gommans et al. 2000). The crystallite orientation and the SWNT alignment in melt-blended SWNT–PP composites fibers have been studied using X-ray diffraction and polarized Raman spectroscopy, which showed that there is an orientation effect in the drawn SWNT–PP fiber (Bhattacharyya et al. 2003). A combination of solvent casting and melt mixing was used to disperse SWNTs in poly(methyl methacrylate) (PMMA), and polarized Raman spectroscopy was again used to demonstrate the alignment of nanotubes in the PMMA (Haggenmueller et al. 2000).
The previous sections clearly demonstrate that carbon nanotubes embedded in composite fibers allow these to be used as strain or stress sensors by measuring the shift of the G′ Raman band under mechanical load. Surface functionalization of the nanotubes with carboxylic groups, which provides better tube–polymer adhesion and thus stress transfer, leads to substantial improvements of the fiber as a Raman strain sensor. This opens the way for the inclusion of such fibers in structural composites, to probe their sensing ability in practical applications.
However, the following examples show that surface functionalization is unfortunately also a double-edged sword, which at the same time brings about improved tube dispersion in the polymer mass but also degrades the structural integrity of the tube walls, the consequence of which is a decrease in mechanical properties of the nanotubes. Indeed, carboxylation by acid nitric treatments can degrade the wall structure of single-walled carbon nanotubes (Liu et al., 1998; Monthioux et al., 2001; Hu et al., 2003) and thus reduce their Young’s modulus and tensile strength. The defect density can be quantified by Raman spectroscopy, as in our recent investigation (Lachman et al., 2009) of the strain-induced shift of the G′ Raman band of single-walled carbon nanotubes in polyvinyl alcohol (PVA)–nanotube composite fibers (Fig. 14.7). The Raman D band (D = disorder) located at 1320 cm− 1 (Fig. 14.1) originates from amorphous carbon and structural defects; the G band (G = graphite) at 1570 cm− 1 is related to graphite structures, and stems from tangential shearing mode of the carbon atoms (Coleman et al., 2004). The G′ band at 2640 cm− 1 is an overtone of the D band. The ratio of integrated intensities of the D and G bands, ID/IG, can be used to estimate the density of defects in the CNT structure: the larger the value of the ID/IG ratio, the higher the defect density (Lachman et al., 2009, and references therein). Indeed, the integrated peak intensity ratio, ID/IG (based on the area under the peaks), of pristine SWNT–PVA was found to be 0.16, whereas that of COOH-SWNT–PVA is 0.73 (Lachman et al., 2009). Such a large difference confirms that COOH-SWNTs contain significantly more defects than pristine SWNTs. In other words, carboxylation by nitric acid treatment has significantly degraded the wall structure of single-wall carbon nanotubes. Moreover, referring to Fig. 14.8, upon straining, the ID/IG ratio of the COOH-based specimen is seen to progressively decrease down to a constant value, whereas this ratio remains constant for the pristine tube-based specimen. This leads to a lower but constant, large difference between the ID/IG ratio at higher strains. This CNT wall structure degradation effect is most likely the main cause of the lack of improvement in the mechanical properties of the PVA fibers. In such conditions, it is possible that, in spite of a better chemical adhesion to the matrix, the weakening or shortening of carboxylated nanotubes produces no improvement in the mechanical properties of the nanocomposites.
14.7 Strain-induced shifts of the G′ band for carboxylated and pristine single-walled nanotubes in PVA. At low strain (ε < ~ 0.01), in the elastic regime, the data can be approximately fitted by linear relationships. At larger strain, in the plastic regime, the Raman response becomes insensitive to strain, likely due to weakening of the interfacial adhesion as the polymer chains possibly slide at the nanotube interface. (From Lachman et al. (2009).)
14.8 Effect of applied strain on the integrated peak intensity ratio, ID/IG, for pristine SWNT–PVA and COOH-SWNT–PVA fibers. The intensity ratio calculation is based on the measurement of the areas under the D and G peaks. (From Lachman et al. (2009).)
As a second example, taken from Sui et al. (2009), the mechanical properties of electrospun fibers made of PMMA containing surface modified nanotubes generally fall below those of fibers with pristine nanotubes, sometimes below those of pure polymer fibers. We show that covalent functionalization produces defects in the graphene structure, leading to mechanical weakening of the nanotube and, therefore, of the nanocomposite. To demonstrate the presence of defects in the functionalized MWCNTs, Raman and transmission electron microscopy (TEM) characterization were performed. As indicated in Fig. 14.9, the values of ID/IG for the functionalized MWCNTs are found to be larger than for pristine MWCNTs, the largest ratio being observed for f-MWCNTs. The same trend is observed by using the ratio of intensities of the D and G′ bands, ID/IG′. Thus, chemical surface modification indeed again causes defects in the CNT structure.
14.9 Raman spectra of pristine and functionalized MWCNTs. The ratio ID/IG of integrated intensities of the D and G peaks reflects the amount of defects in the CNT structure, with larger ratio values corresponding to higher defect densities. (From Sui et al. (2009), with permission from the American Institute of Physics.)
So far, it has not been possible to calibrate the shift of a Raman peak position with an applied strain in CNTs, as is done with individual graphite fibers in air. In composites it is clear, however, that the larger the Raman peak shift, the larger is the strain carried by the nanotubes. Schadler et al. (1998) measured the peak shift for the sensitive G′ band of nanotubes (2700 cm− 1), a strong peak for MWNT, with no epoxy peak overlap in this region. Figure 14.10 shows the G′ Raman peak shift in tension and compression. Under a 1% compressive strain the peak is observed to shift upwards by 7 cm− 1, whereas only a slightly positive shift appears when the composites are under tension. The shifts are thought to arise from the strain transferred from the matrix to nanotubes. The different Raman responses in tension and compression are most likely due to the structure of MWNTs. Indeed, Schadler et al. (1998) believe that, under tension, the outer layer of the MWNT is loaded, but load is not effectively transferred to the inner layers due to the relatively weak bonding between the nanotube layers. Instead, the inner tubes may slip with respect to the outer tubes. Since the Raman signal is averaged over the whole MWNT, the result is only an insignificant Raman peak shift. Under compression, however, the load transfer to the inner layers of the MWNT occurs through buckling and the bent sections of the nanotubes. Slippage of nanotubes layers in compression is prevented because of the seamless structure of the tubes and of the geometrical constraint the outer layer imposes on the inner layers (Schadler et al. 1998). Similar experiments have been performed by Ajayan et al. (2000) with SWNTs, but the G′ Raman results showed that under compression there is almost no shift and only a small downward shift under tension. Ajayan et al. (2000) observed that the applied compressive stress is transferred into buckling, bending or twisting of the nanotube network without introducing important local deformations that can be monitored by Raman. Since SWNTs tend to form ropes, the small Raman shift under tension is due to the fact that the individual SWNTs are slipping within the ropes and decrease the load required to deform the ropes (Ajayan et al., 2000). In both cases, the tensile and compressive strains are applied uniaxially on the samples, though the nanotubes are dispersed randomly in the polymer matrix (Schadler et al., 1998; Ajayan et al., 2000). When the nanotube composite samples are under tension, nanotubes lying along the tensile direction are under tension, but those in the perpendicular direction are under compression because of Poisson’s contraction. The reverse is true when the sample is under compression. Since the Raman laser spot size is ca. 1–5 μm thick, and nanotube diameters are several nanometres in size, the Raman signal is averaged over nanotubes in all directions, but Ajayan et al. (2000) and Schadler et al. (1998) do not consider this in detail.
14.10 Nanotube Raman peak shift as a function of applied strain showing the large shift in the Raman peak (second order peak at 2700 cm− 1) in compression compared to tensile loading. (From Schadler et al. (1998), with permission from the American Institute of Physics.)
The mechanical properties of a composite depend not only on the properties of the fiber and the matrix, but also on the quality of the interface between these. In composite material research a single fiber composite test is often used to quantify the level of interfacial adhesion. The single fiber fragmentation technique is a known, practically convenient and reproducible method often used to characterize the fiber–matrix adhesion or interfacial properties. However, it also is a complex test which involves shear yielding of the matrix, interfacial debonding and transverse matrix cracking (Kelly and Tyson, 1965; Wagner and Eitan, 1990, 1993; Eitan and Wagner, 1991; Detassis et al., 1996). The occurrence of these additional damage events during the fragmentation test makes the conventional data reduction technique (based on the constant shear model) problematic. Thus, a determination of the stress distribution around the fiber break in the matrix is necessary. This is illustrated in an experiment in which two-dimensional stress profiles around a glass fiber break were mapped using SWNT sensors randomly dispersed in a PUA matrix, by means of polarized Raman spectroscopy. A contour map showing the distribution of the stress concentration factor around a fiber break could be produced (Zhao and Wagner, 2003). The stress concentration (defined as Kc = local stress/applied stress) reached a maximum value just near the glass fiber break (Kc = 1.42) and decreased radially and longitudinally away from the break point, as expected (Eitan and Wagner, 1991; Wagner and Eitan, 1993).
A further experiment demonstrating the sensing power of Raman microscopy was performed by combining the Raman sensitivity of both a carbon fiber and carbon nanotubes, the latter being dispersed in the polymer around the former. The basic idea of this experiment was to simultaneously detect the stress distributions in the fiber and in the matrix using Raman spectroscopy, and to compare the results with existing stress transfer models. A continuous high modulus carbon fiber (HMCF)–PUA composite with nanotubes dispersed in the matrix was chosen. The G′ Raman band of HMCF shifts linearly under an applied strain. The strains in the fiber and the matrix were measured simultaneously under several applied stress levels, parallel to the fiber direction. Figure 14.11 presents the distributions of stress in the carbon fiber and in the matrix, between two break locations measured at an applied stress of 10 MPa. Polynomial curves were fitted through the experimental data points. As seen, the stress profiles in the matrix and the fiber, measured simultaneously for the first time, complement each other well (Zhao and Wagner, 2003).
14.11 Stress distributions in the HMCF (a) and the PUA matrix along the fiber edge (b), measured simultaneously by microRaman spectroscopy. The distributions are mirror images of each other. The applied stress level was 10 MPa. The solid line in (a) is a polynomial fit to the data. In (b), the line is the mirror image of the fit in (a), scaled accordingly. (From Zhao and Wagner (2003).)
Most recently, the deformation micromechanics of SWNTs embedded in a polymeric matrix was investigated (Kao and Young, 2010) through the use of Raman spectroscopy. The degree of interfacial adhesion in nanotube–epoxy composites was evaluated through cyclic deformation of the specimens and was found to be dependent on the maximum loading strain and the numbers of deformation cycles. A hysteresis loop was observed from the mismatch between the loading and unloading results and the loop area was estimated to evaluate the energy dissipated in the composites. This was employed in an approximate model to quantify the extent of interface damage, using assumptions about the nanotube distribution in the composite. In other words, in this case, the Raman technique did not lead to a measurement of the stress but an energy-based approach was used to monitor stress transfer.
The use of Raman spectroscopy to produce the full interface shear strength profile along a tube in a single nanotube pull-out or a fragmentation test, similar to the one produced in a micromechanical test using a micron-size carbon or Kevlar fiber, remains an elusive challenge. If the interfacial shear strength in CNT-based nanocomposites is to be evaluated by a combination of, say, Raman measurement and a model for the interfacial strength based on a force balance – in traditional micron-size fibers this is done using the Kelly–Tyson model (Kelly and Tyson, 1965) – the fact that tubes are hollow becomes crucial. This was examined and discussed by Wagner (2002). The interrelation between the tube–matrix interfacial (adhesion) shear strength τNT, the tensile strength σNT(lc) of a tube fragment (of length lc), the (empirically measured) critical length lc, and the inner and outer diameters dNT and DNT, is given by
and the nanotube wall thickness is h = (DNT – dNT)/2. It is easily seen that high values of the interfacial shear strength (compared to those in current advanced fiber-based polymer composites) are in principle attainable. Defects in the hexagonal structure of a nanotube, which technically is a ‘perfect’ material, are expected to strongly reduce its strength and the model predicts that, as a consequence, large variability should be experimentally observed in either the interfacial strength or the critical length of apparently identical nanotubes. The presence or absence of such defects in the structure of the CNT could in principle be detectable by Raman spectroscopy through the ID/IG ratio, and an experimental correlation with the interfacial shear strength be found via Equation 14.7. However, at the present time, such a correlative experiment seems beyond reach. The CNT critical length and interfacial strength could be measured by either nanofragmentation or nanopull-out testing. However, the occurrence of the CNT fragmentation phenomenon has been observed only occasionally, with great difficulty (Wagner et al., 1998; Lourie et al., 1998; Lourie and Wagner, 1999). In addition, Lourie and Wagner (1998b) provide evidence of significant polymer–nanotube wetting and interfacial adhesion, see Fig. 14.12. Unique single nanotube pull-out experiments were conducted by Barber et al. (2003, 2004, 2006), following a preliminary attempt to experimentally ‘drag out’ a nanotube from a polymer matrix by means of an AFM tip (Cooper et al., 2002). The pull-out model of Gou et al. (2005) predicts the effects of temperature and of the number of walls on the interfacial shear strength. See also an interesting paper by Lau (2003). Interestingly, very high values (hundreds of MPa) of the CNT–matrix interface strength are attainable in principle, based on the models by Wagner (2002) and Zhang and Wang (2005), as well as on numerical simulations (Liao and Li, 2001). The effect of chirality on the interfacial strength is also of interest (Zheng et al., 2008) but a challenging correlation with Raman spectroscopy has not been attempted.
14.12 TEM image of aligned single-walled carbon nanotube ropes bridging an elliptical hole in a polymer film. The ropes tend to orient parallel to each other and to the short axis of the elliptical hole. The image provides strong evidence of significant polymer–nanotube wetting and interfacial adhesion. (From Lourie and Wagner (1998b), with permission from the American Institute of Physics.)
Recent developments in the application of Raman spectroscopy to carbon nanotube-based composite materials have been reviewed. This technique may be used to distinguish carbon nanotubes from other carbon materials or polymers, as well as to sort nanotubes by diameter or length (Zhao and Wagner, 2004). Raman spectroscopy has also been used to check the dispersion of nanotube in polymers, evaluate nanotube–matrix interactions, and detect polymer phase transitions. The Raman spectra of nanotubes can also be utilized to quantify the strain or stress transferred to nanotubes from the surrounding environment. Since the small amounts of embedded nanotubes required to make the polymer Raman sensitive to strain do not affect the mechanical properties of the embedding matrix, these nanoscale tubes may be used to investigate local stresses and strains in polymer materials, a result that is so far inaccessible by other methods. Polarized Raman can be used to detect the orientation of nanotubes in polymer matrices. Oriented SWNT sensors can be used to detect stress fields around a circular hole in a plate, demonstrating the applicability of the technique for the mapping of the stress distribution in the vicinity of discontinuities.
A polarized Raman technique may be used to detect the stress or strain in a matrix using randomly dispersed SWNTs, based on the fact that the intensity of the signal arising from SWNTs is a function of the polarization direction. This method is more convenient than the tube orientation technique since it is not always possible to orient SWNTs inside a polymer.
Finally, the sensing capability of nanotubes clearly makes it possible to investigate and settle various fracture mechanics problems, including so far unsolved ones. For example, current studies of stress transfer and fiber–fiber interactions in fiber reinforced composites are all based on theoretical models (such as shear-lag models) which involve a ‘characteristic distance’ in the matrix, radially from the fiber–matrix interface (one version of this is the so-called inverse Cox parameter, β− 1). This fundamental parameter, which represents a ‘zone of influence’ around a fiber break or end, is currently difficult to measure directly. The relatively simple measurement of such characteristic distance would be an immediate outcome of the mapping of the stress or strain field using nanotubes as Raman sensors in composites.
This author wishes to acknowledge support from the Israel Ministry of Industry, Trade and Labor, through the Nanotubes Empowerment Solutions (NES) Magnet research consortium, and from the G. M. J. Schmidt Minerva Centre of Supramolecular Architectures. This research was also made possible in part by the generosity of the Harold Perlman family. Thanks are due to Dr Shigeo Maruyama (The University of Tokyo) and Dr Hagai Cohen (Weizmann Institute of Science) for providing the SWNT sample prepared by CVD, the Raman spectrum of which is shown in Fig. 14.1 (b). Thanks are gratefully extended to my postdoctoral fellow Dr XiaoMeng Sui for producing the Raman spectra shown in that figure. H. D. Wagner is the recipient of the Livio Norzi Professorial Chair in Materials Science.
Atkinson, K.R., Hawkins, S.C., Huynh, C., Skourtis, C., Dai, J., Zhang, M., Fang, S., Zakhidov, A.A., Lee, S.B., Aliev, A.E., Williams, C.D., Baughman, R.H. Multifunctional carbon nanotube yarns and transparent sheets: fabrication properties and applications. Phys. B: Cond. Matter. 2007; 394(2):339–343.
Bhattacharyya, A.R., Sreekumar, T.V., Liu, T., Kumar, S., Ericson, L.M., Hauge, R.H., Smalley, R.E. Crystallization and orientation studies in polypropylene/single wall carbon nanotube composite. Polymer. 2003; 44:2373–2377.
Coleman, J.N., Cadek, M., Blake, R., Nicolosi, V., Ryan, K.P., Belton, C., Fonseca, A., Nagy, J.B., Gun’ko, Y.K., Blau, W.J. High performance nanotube-reinforced plastics: understanding the mechanism of strength increase. Adv. Funct. Mater. 2004; 14(8):791–798.
Detassis, M., Pegoretti, A., Migliaresi, C., Wagner, H.D. Experimental evaluation of residual stresses in single fiber composites by means of the fragmentation test. J. Mater. Sci. 1996; 31:2385–2392.
Galiotis, C., Young, R.J., Batchelder, D.N. A resonance Raman-spectroscopic study of the strength of the bonding between an epoxy-resin and a polydiacetylene fiber. J. Mater. Sci. Lett. 1983; 2:263–266.
Gou, J., Liang, Z.Y., Zhang, C., Wang, B. Computational analysis of effect of singlewalled carbon nanotube rope on molecular interaction and load transfer of nanocomposites. Composites B: Engineering. 2005; 36:524–533.
Gouadec, G., Karlin, S., Wua, J., Parlier, M., Colomban, Ph. Physical chemistry and mechanical imaging of ceramic-fibre-reinforced ceramic-or metal-matrix composites. Compos. Sci. Tech. 2001; 61:383–388.
Hwang, J., Gommans, H.H., Ugawa, A., Tashiro, H., Haggenmueller, R., Winey, K.I., Fischer, J.E., Tanner, D.B., Rinzler, A.G. Polarized spectroscopy of aligned single-wall carbon nanotubes. Phys. Rev. B. 2000; 62(20):R13310.
Jorio, A., Santos, A.P., Ribeiro, H.B., Fantini, C., Souza, M., Vieira, J.P.M., Furtado, C.A., Jiang, J., Saito, R., Balzano, L., Resasco, D.E., Pimenta, M.A. Quantifying carbon-nanotube species with resonance Raman scattering. Phys. Rev. B. 2005; 72:075207.
Kao, C.C., Young, R.J., Angular dependence upon deformation of SWNT/epoxy composites using polarized Raman spectroscopy (2005). CNT-Polymer Composites International Conference. Hamburg, Germany, 2005.
Liu, J., Rinzler, A.G., Dai, H., Hafner, J.H., Bradley, R.K., Boul, P.J., Lu, A., Iverson, T., Shelimov, K., Huffman, C.B., Rodriguez-Macias, F., Shon, Y.-S., Lee, T.R., Colbert, D.T., Smalley, R.E. Fullerene pipes. Science. 1998; 280:1253–1256.
Monthioux, M., Smith, B.W., Burteaux, B., Claye, A., Fischer, J.E., Luzzi, D.E. Sensitivity of single-wall carbon nanotubes to chemical processing: an electron microscopy investigation. Carbon. 2001; 39:1251–1272.
Nielsen, A.S., Pyrz, R. Study of the influence of thermal history on the load transfer efficiency and fibre failure in carbon/polypropylene microcomposites using Raman spectroscopy. Compos. Interf. 1999; 6:467–482.
Rao, A.M., Richter, E., Bandow, S., Chase, B., Eklund, P.C., Williams, K.A., Fang, S., Subbaswamy, K.R., Menon, M., Thess, A., Smalley, R.E., Dresselhaus, G., Dresselhaus, M.S. Diameter-selective Raman scattering from vibrational modes in carbon nanotubes. Science. 1997; 275:187–191.
Souza Filho, A.G., Jorio, A., Hafner, J.H., Lieber, C.M., Saito, R., Pimenta, M.A., Dresselhaus, G., Dresselhaus, M.S. Electronic transition energy Eii for an isolated single-wall carbon nanotube obtained by anti-Stokes/Stokes resonant Raman intensity ratio. Phys. Rev. B. 2001; 63:241404.
Stephan, C., Nguyen, T.P., Lahr, B., Blau, W., Lefrant, S., Chauvet, O. Raman spectroscopy and conductivity measurements on polymer-multiwalled carbon nanotubes composites. J. Mater. Res. 2002; 17:396–400.
Wagner, H.D., Amer, M.S., Schadler, L.S. Residual compression stress profile in high modulus carbon fiber embedded in isotactic polypropylene by micro-Raman spectroscopy. Appl. Compos. Mater. 2000; 7:209–217.
Wagner, H.D., Nairn, J.A. Residual thermal stresses in three concentric transversely isotropic cylinders: application to thermoplastic matrix composites containing a transcrystalline interphase. Compos. Sci. Technol. 1997; 57:1289–1302.
Wood, J.R., Frogley, M.D., Meurs, E.R., Prins, A.D., Peijs, T., Dunstan, D.J., Wagner, H.D. Mechanical response of carbon nanotubes under molecular and macroscopic pressures. J. Phys. Chem. B. 1999; 103:10388–10392.
Wu, G., Tashiro, K., Kobayashi, M. Vibrational spectroscopic study on molecular deformation of polydiacetylene single crystals: stress and temperature dependences of Young’s modulus. Macromol. 1989; 22:188–196.
Young, R.J., Eichhorn, S.J. Raman applications in synthetic and natural polymer fibres and their composites. In: Amer M.S., ed. Raman Spectroscopy for Soft Matter Applications. New York: John Wiley & Sons Inc; 2009:63–94.