Introduction

Poly(ethylene terephthalate) (PET) has been established as a polymeric material of major industrial importance, due to its low cost and high performance (it exhibits high glass transition and melting temperatures), as well as good physical properties. Depending on its processing and thermal history, PET may exist both as an amorphous (transparent) and as a semicrystalline material. It has found a variety of applications, such as in textile fibers, films, bottle containers, food packaging materials, engineering plastics in automobiles, and electronics, etc. Its properties depend mainly on the degree of orientation of the polymer chains and the degree of crystallinity. PET bottles are excellent barrier materials and are widely used with soft carbonated drinks, while in recent years this application has been extended to include alcoholic beverages. The incorporation of organic/inorganic hybrids can result in materials possessing excellent stiffness, strength and gas barrier properties, with far less inorganic content than is commonly used in conventionally filled polymer composites. The enhanced properties usually exhibited by such materials can significantly expand the potential applications of PET [1]. Nanoclays have been the most studied nanofillers in PET [2–5], while nanocomposites with attapulgite [6], TiO2 [7], SiO2, [8], and calcium carbonate [9] have also been reported. PET/carbon nanotubes composites have recently been studied for their potential application in fuel cells [10, 11], flexible vapour sensors [12], conductive fibers [13] and, furthermore, composite materials with functionalized [14] and acid-treated [15] nanotubes have been prepared.

Apart from mechanical properties and processability, thermal properties and, mainly, resistance to thermal degradation can be enhanced by the addition of appropriate nanoparticles. However, in the case of PET nanocomposites different results have been reported. Yuan et al. found that amino functional groups in amino surfactants decreased the thermal stability of the organoclay and accelerated the degradation behavior of the corresponding PET nanocomposites [16]. Nanocomposites based on clay with a larger amount of hydroxyl groups on the edge of the clay platelets exhibited a much more degradation because the hydroxyl groups acted as Brønsted acidic sites and accelerated the polymer’s degradation [17]. The apparent activation energies of PET/fibrous silicates nanocomposites were evaluated using the Kissinger and Flynn–Wall–Ozawa methods [18]. It was suggested that during thermal decomposition in nitrogen the clay can slow down degradation of polymer as a mass-transport protective barrier, but the catalytic effect of the metal derivatives in the clays could accelerate the decomposition behavior of PET. The combination of these two effects determined the final thermal stability of the nanocomposite. However, in an atmosphere of air, the oxidative thermal stability of the PET nanocomposites was significantly superior to that of pure PET. However, opposite were the results by the addition of attapulgite nanoparticles, which could, however, enhance the thermal stability of PET [19]. Furthermore, thermogravimetric analysis suggested that there existed no influence by single wall nanotubes (SWNTs) upon the thermal stability of PET [20]. However, as far as we know, there are no reports about the kinetics of the thermal decomposition behavior of PET with acid-treated multi-walled carbon nanotubes (MWCNTs) nanocomposites.

In the present study, PET/MWCNTs nanocomposites were prepared by in situ polymerization. The objective was to evaluate the effect of the acid-treated MWCNTs upon the thermal degradation of PET nanocomposites, using different kinetic models.

Experimental

Materials

Dimethyl terephthalate (DMT) (99%), anhydrous 1,2-ethanediol (E.G) (99%), antimony trioxide (Sb2O3) (98%), and triphenylphosphate (TPP) (95%) were obtained from Fluka. Zinc acetate [(CH3CO2)2Zn] (99.99%) was purchased from Aldrich. The MWCNTs [21] used in this study were synthesized by chemical vapor deposition (CVD) and were supplied by Nanothinx (Patra, Greece). They had a diameter between 9 and 20 nm and a length >5 μm. All other materials and solvents used for the analytical methods were of analytical grade.

Acid-treatment of MWCNTs

The pure MWCNTs were functionalized, using the acid treatment method, to introduce reactive groups on their surface. Concentrated nitric acid and sulfuric acid were used for the chemical treatment as described in our previous study [22]. The acid treatment time was 15 min.

In situ prepared PET/MWCNTs nanocomposites

Nanocomposites of poly(ethylene terephthalate) were prepared in situ by the two-stage melt polycondensation of dimethyl terephthalate (DMT) and ethylene glycol (EG) in a glass batch reactor. Appropriate amount of filler was dispersed in EG by ultrasonic vibration (50 W) and intense stirring with a magnetic stirrer (300 rpm) for 10 min prior to polymerization. Zinc acetate was added (1 × 10−3 mol/mol DMT) and DMT at a molar ratio of DMT/EG = 1/2.2. The mixture was de-aired and purged with argon three times. Subsequently, it was heated at 190 °C for 1 h, under constant mechanical stirring (500 rpm). The methanol produced by the transesterification reaction was removed from the mixture by distillation and collected in a graduated cylinder. The temperature was raised to 230 °C and the reaction continued for a further 2 h, were the complete removal of the anticipated produced methanol took place.

During the second stage, under an argon atmosphere, the polycondensation catalyst (Sb2O3, 250 ppm) and TPP (0.03 wt% DMT) as thermal stabilizer, dispersed in a small amount of EG, were added in the mixture. TPP prevents side reactions such as etherification and thermal decomposition. Afterwards, vacuum was applied (~5 Pa) slowly, over a period of 15 min, to avoid excessive foaming and to minimize oligomer sublimation, which is a potential problem during the melt polycondensation. The temperature was raised to 280 °C and the reaction ensued for 2 h. When the polycondensation reaction was completed, the reaction tube was broken to remove the product out of the flask. All polyester samples, after the glass particles were removed with a grinder, were grounded in a mill, sieved, washed with methanol, and dried at 110 °C for 12 h.

Different amounts of MWCNTs were added at the beginning of the transesterification reaction. PET/MWCNT nanocomposites were prepared containing 0.25, 0.5, 1.0, and 2.0 wt% MWCNTs (PET-0.25, PET-0.5, PET-1, and PET-2, respectively).

Intrinsic viscosity measurements

Intrinsic viscosity measurements on the isolated polymers were performed using an Ubbelohde viscometer cap. Ic at 25 °C in phenol/tetrachloroethane 60/40 w/w at a solution concentration of 1 wt%. The solutions were filtered through a disposable membrane filter 0.2 μm (Teflon) prior to the measurement. The intrinsic viscosities [η] were calculated using the Solomon–Ciutâ equation [23]:

$$ \left[ \eta \right] = {\frac{{\sqrt {2\; \times \;({\frac{t}{{t_{0} }}} - \ln {\frac{t}{{t_{0} }}} - 1)} }}{C}} $$
(1)

where C is the concentration of the solution, t the flow time of the solution and t 0 the flow time of the solvent. Some samples were insoluble in the above solvent mixture. In these samples the insoluble part was removed by filtration and washed extensively first with the solvent mixture and secondly with acetone to remove the solvent. After that it was dried to constant weight by heating at 100 °C under vacuum, in order to measure the insoluble content [24].

Transmission electron microscopy

Electron diffraction (ED) and transmission electron microscopy (TEM) observations were made on ultra thin film samples of the various nanocomposites prepared by an ultra-microtome. These thin films were deposited on copper grids. ED patterns and TEM micrographs were obtained using a JEOL 120 CX microscope operating at 120 kV. In order to avoid the destruction of the films after exposure to the electron irradiation, an adequate sample preparation is required, thus, the thin films were coated with carbon black.

Thermogravimetric analysis

Thermogravimetric analysis (TG) was carried out with a SETARAM SETSYS TG-DTA 1,750 °C. Samples (5.5 ± 0.2 mg) were placed in alumina crucibles. An empty alumina crucible was used as reference. Samples were heated from ambient temperature to 600 °C in a 50 ml/min flow of N2. Heating rates of 5, 10, 15, and 20 °C/min were used and continuous recordings of sample temperature, sample weight, its first derivative, and heat flow were taken.

Results and discussion

Characterization of the prepared materials

The MWCNTs used in this study were functionalized using chemical oxidation with a mixture of sulfuric and nitric acids [22]. XPS study of the oxidized nanotubes has revealed that the surface is covered with carboxylic (–COOH), carbonyl (−C=O), and hydroxyl (–OH) groups in the approximate proportions of 4:2:1, respectively. These groups increase the chemical reactivity and specificity of the otherwise relatively inert carbon surface. The concentration of the surface acidic groups after the acid treatment was 2.8 meq g−1 [22]. The inorganic impurities of the Al–Fe catalyst used in the production of the nanotubes are solubilized by the mixture of inorganic acids and their concentration is reduced with the treatment. The consequences of this oxidative treatment, except for the formation of reactive groups on, mainly, the nanotube’s endcaps and at defect sites along the sidewalls, are, simultaneously, the length reduction of the nanotubes into smaller fragments. After 15 min treatment the length of the MWCNTs ranged between 5 and 10 μm.

PET/MWCNTs nanocomposites were prepared in situ during polymerization of PET by the two-stage melt polycondensation method of dimethyl terephthalate (DMT) and ethylene glycol (EG) in a glass batch reactor in the presence of MWCNTs (Fig. 1).

Fig. 1
figure 1

Synthetic route to PET preparation as well as its nanocomposites with MWCNTs, through the two-step melt polycondensation

Full size image

At the first stage of transesterification oligomers are prepared, while methanol is removed as by-product. In order to prepare high molecular weight polyesters, the polycondensation of these oligomers ensued at higher temperature, with the simultaneous application of high vacuum. The intrinsic viscosity of PET that was prepared was 0.57 dL/g. The presence of the MWCNTs in the polymerization mixture affected the reaction, leading to variations of the final intrinsic viscosities of the nanocomposites prepared. As shown in Table 1, the intrinsic viscosity increased at the MWCNT’s concentration of 0.25 wt%, and, then, gradually decreased at higher concentrations.

Table 1 Intrinsic viscosity [η] of the prepared PET/MWCNTs nanocomposites
Full size table

Such a behavior was also observed in previous studies by the addition of SiO2 nanoparticles, as in the case of poly(ε-caprolactone) [25], prepared by ring opening polymerization, and in poly(butylene succinate) [26], prepared by the two-stage melt polycondensation method. SiO2 nanoparticles contain surface silanol groups (≡Si–OH) which, due to their weak acidic nature, can react with the hydroxyl end groups of aliphatic polyesters. It seems plausible that the surface carboxylic groups of acid-treated MWCNTs could also interact with the hydroxyl end groups of PET. In similar nanocomposites containing SiO2, which can also act as multifunctional additives, it was found that in small amounts SiO2 nanoparticles could act as chain extender, increasing the molecular weight of the polymer, while at higher contents, due to the extended degree of the aforementioned reaction, branched and crosslinked macromolecules are formed, leading to reduce apparent molecular weights [27, 28]. It is well known that the hydrodynamic dimensions of branched polymers in solution are smaller compared to those of linear polymers with the same molecular weight [29]. Intrinsic viscosity measurements calculate the molecular weight of polymers based on their hydrodynamic size in solution. Therefore, branched polymers appear as having lower molecular weights in comparison to their actual one. This may explain the gradual decrease that observed in molecular weight of the PET samples with concentrations of MWCNTs nanoparticles greater than 0.25 wt%. Another possible contributor to the aforementioned variation of molecular weights was the formation of extensive cross-linkings, resulting in the appearance of an insoluble fraction, as in the case of the sample containing 2 wt% MWCNTs (Table 1).

From TEM micrographs (Fig. 2), it was observed that the dispersion of the MWCNTs into the PET matrix was homogeneous, while at higher concentrations of MWCNTs some small aggregates were also formed. This was due to the extremely high specific area of MWCNTs and the strong interparticle interactions that take place. MWCNTs also influenced the thermal properties of the prepared nanocomposites. From the DSC thermograms the melting point temperatures were calculated. As seen from Table 1, the melting temperatures shifted to lower values by increasing the MWCNTs content. The slight decrease in the melting point might result from the specific interaction between PET and the MWCNTs, since the homogeneous dispersion of MWCNTs affected the formation of PET crystals. It should be mentioned that the formed crystals were of lower perfection compared to those of neat PET and, probably, this was due to the heterogeneous nucleation induced by the MWNTs [30, 31]. Furthermore, the samples containing higher MWCNTs content contained branched and crosslinked macromolecules, which also contributed to the exhibited lower melting points.

Fig. 2
figure 2

TEM micrograph of PET/MWCNT nanocomposite containing 1 wt% MWCNTs

Full size image

Thermogravimetric analysis

Thermal degradation of PET containing different MWCNTs contents was studied by determining its mass loss during heating. In Fig. 3, the mass loss (TG%) and the first derivative of mass loss (DTG) curves of all the studied samples are presented, at a heating rate of 10 °C/min. From the thermogravimetric curves it is obvious that PET and the samples having different MWCNTs content exhibited good thermostability, since no remarkable mass loss occurred up to 320 °C (<0.5%). As seen from the peak of the first derivative, the temperature at which PET decomposition rate was highest was at T p = 436.4 °C, for a heating rate of 10 °C/min. The shape of the mass curve was the same in all samples. In nanocomposites containing up to 1 wt% MWCNTs this temperature was gradually shifted to higher values, up to 439.3 ± 0.5°C, which was an indication that MWCNTs could enhance the thermal stability of the matrix. However, in the sample containing 2 wt% MWCNTs this temperature was shifted to lower values. Since the addition of MWCNTs, as well as that of several other nanoparticles, increased the thermal stability of polymers [32–34], this reduction should be attributed to the presence of extensively branched and crosslinked macromolecules that this sample probably contains.

Fig. 3
figure 3

a Mass loss (%) versus temperature and b first derivative of mass loss (DTG) versus temperature, at a heating rate β = 10 °C/min for all the studied samples

Full size image

In order to analyze more thoroughly the degradation mechanism of the PET nanocomposites with MWCNTs, it was important that the kinetic parameters (activation energy E and pre-exponential factor A) and conversion function f(α) were evaluated. The relationship between the kinetic parameters and conversion (α) can be found using the mass curves recorded in the TG dynamic thermograms. The thermogravimetric curves of PET containing 1 wt% MWCNTs, heated in N2 at different heating rates, are shown in Figs. 4 and 5, at temperatures above the melting point up to 550 °C. It is clear from the DTG plots that the peak temperature, T p, shifted to higher values at increasing heating rates.

Fig. 4
figure 4

Mass loss (%) curves at different heating rates β of PET containing 1 wt% MWCNTs. 1 β = 5 °C/min, 2 β = 10 °C/min, 3 β = 15 °C/min, 4 β = 20 °C/min

Full size image
Fig. 5
figure 5

Derivative mass loss curves at different heating rates β of PET containing 1 wt% MWCNTs. 1 β = 5 °C/min, 2 β = 10 °C/min, 3 β = 15 °C/min, 4 β = 20 °C/min

Full size image

The activation energy of degradation of the studied polyester was estimated using the Ozawa, Flynn, and Wall (OFW), Friedman and Kissinger’s methods. The isoconversional method of Ozawa, Flynn, and Wall (OFW) [32, 33] is, in fact, a ‘‘model-free’’ method which assumes that the conversion function f(α) does not change with the alteration of the heating rate for all values of the degree of the conversion α. It involves the measuring of the temperatures corresponding to fixed values of α from experiments at different heating rates β. Therefore, plotting ln(β) against 1/T according to Eq. 2 in the form of

$$ { \ln }(\beta ) = { \ln }[Af(\alpha )/({\text{d}}\alpha /{\text{d}}T)] - E/RT $$
(2)

should result in straight lines, the slope of which is directly proportional to the activation energy (−E/R). If the determined activation energy is the same for the various values of α, the existence of a single-step reaction can be concluded with certainty. On the contrary, a change in E with increasing degree of conversion is an indication of a complex reaction mechanism that invalidates the separation of variables involved in the OFW analysis [34]. These complications are serious, especially in the case where the total reaction involves competitive reaction mechanisms.

In Fig. 6, the straight lines obtained by fitting the experimental data are presented, for the sample containing 0.5 wt% MWCNTs, with a correlation coefficient greater than 0.999. These lines are nearly parallel (0.1 < α<0.9), thus indicating the applicability of the Ozawa method to our system in the conversion range presented. In Fig. 7 the dependence of the activation energy (E) versus conversion α can be seen, for all the studied samples.

Fig. 6
figure 6

OFW plots at different degree of the conversion α for PET containing 0.5 wt% MWCNTs

Full size image
Fig. 7
figure 7

Dependence of the Activation Energy (E) on the degree of the conversion (α) of the mass loss, as calculated with OFW method for the different samples (lines drawn for eye guidance)

Full size image

As observed in Fig. 7, the activation energies of the nanocomposites containing up to 1 wt% MWCNTs were higher compared to pure PET. This is an indication that the addition of MWCNTs enhanced the thermal stability of PET. Such behavior is very common in nanocomposites and such thermal stabilization was also reported by the addition of SiO2 nanoparticles or nanoclays [35–40]. However, the sample containing 2 wt% MWCNTs, even though it had the highest content of nanotubes and, thus, the highest prospective thermal stabilization, it had on the contrary the lowest. In all mass conversions the activation rate values were even lower than pure PET. This abnormal behavior should be attributed to the formation of extended branched and crosslinked macromolecules, which reduce thermal stability, in the presence of the MWCNTs.

Friedman [41] proposed the application of the logarithm of the conversion rate dα/dt as a function of the reciprocal temperature, in the form of

$$ { \ln }({\text{d}}\alpha /{\text{d}}T) = { \ln }(A/\beta ) + { \ln }\;f(\alpha ) - E/RT \, $$
(3)

It is obvious from Eq. 2 that if the function f(α) is constant for a particular value of α, then the sum ln(f(α)) + ln A/β is also constant. By plotting ln(dα/dT) i against 1/T i, the value of the −E/R for a given value of α can be directly obtained. Using this equation, it is possible to obtain values for E over a wide range of conversions. The results are presented in Fig. 8 and have great similarities with those of Fig. 7. Both methods revealed the same trend of activation energies for the whole conversion range studied. However, comparing the results of the application of the two methods, we observed that the values calculated by the Friedman method were slightly higher than those of the Ozawa method.

Fig. 8
figure 8

Dependence of the Activation Energy (E) on the degree of the conversion (α) of the mass loss, as calculated with Friedman’s method for the different samples (lines drawn for eye guidance)

Full size image

The difference in the E value calculated by the two methods can be explained by a systematic error due to improper integration. The method of Friedman employs instantaneous rate values, being therefore very sensitive to experimental noise. In the Ozawa method the equation used is derived assuming constant activation energy, introducing systematic error in the estimation of E in the case that E varies with α, error that can be estimated by comparison with the Friedman results [42].

For comparative purpose, we calculated the activation energy calculated using the Kissinger method [43], which has been used in the literature in order to determine activation energy from plots of the logarithm of the heating rate versus the inverse of the temperature at the maximum reaction rate in constant heating rate experiments. The activation energy can be determined by Kissinger method without a precise knowledge of the reaction mechanism, using the following equation:

$$ \ln \left( {\beta /T_{\text{p}}^{2} } \right) = - E/RT_{\text{p}} + {\text{const}} . $$
(4)

where T p is the temperature corresponding to the inflection point of the thermal degradation curves, which correspond to the maximum reaction rate. The values of E, for all the studied samples, were obtained from the slope of the ln(β/T 2p ) versus 1/T p plot, with a correlation coefficient greater than 0.995 (Fig. 9).

Fig. 9
figure 9

Kissinger plots of PET/MWCNTs nanocomposites

Full size image

In order to determine the conversion function f(α), we used a method referred to as the ‘‘model fitting method’’ [44, 45]. This method that does not assume the knowledge of E and f(α) in advance, was applied simultaneously on the experimental data taken at the heating rates β = 5, 10, 15, and 20 °C/min. For the fitting 16 different kinetic models were used. In Figs. 10 and 11 the results of this fitting for PET and PET containing 0.5 wt% MWCNTs are presented exemplarily. The form of the conversion function, obtained by fitting, was mechanism of nth-order auto-catalysis that is described by the equation \( f\left( a \right) = \left( {1 - \alpha } \right)^{\text{n}} \left( {1 + K_{\text{cat}} X} \right) \). As can be seen, the degradation mechanism in the PET/MWCNTs nanocomposites remained the same, as in the case of pure PET. Thus, the addition of the MWCNTs didn’t affect the degradation mechanism, but only the activation energies and the decomposition rate. The correlation coefficient was higher than 0.9999. The calculated parameters are shown in Table 2 for all the samples. The values of the activation energy lie between the limits of the calculated values from the Ozawa and Friedman methods and are higher than the calculated ones using Kissinger’s method. The fitting to the experimental data was very good for all the range of α. When the value of the activation energy used for fitting was the one calculated by Kissinger’s method, the results weren’t better compared to the aforementioned ones.

Fig. 10
figure 10

Fitting and experimental mass (%) curves for all the different heating rates β for PET. 1 β = 5 °C/min, 2 β = 10 °C/min, 3 β = 15 °C/min, 4 β = 20 °C/min

Full size image
Fig. 11
figure 11

Fitting and experimental mass loss (%) curves for all the different heating rates β for PET containing 0.5 wt% MWCNTs. 1 β = 5 °C/min, 2 β = 10 °C/min, 3 β = 15 °C/min, 4 β = 20 °C/min

Full size image
Table 2 Calculated values of activation energy, pre-exponential factor, and exponent n for all studied samples
Full size table

Conclusions

The addition of MWCNTs in situ during polymerization of PET affected most of its properties. MWCNTs acted as multifunctional agents and, as their amount increased, PET macromolecules of lower molecular weight were prepared. This was because branched and crosslinked macromolecules were formed. Thermal decomposition was also improved by the incorporation of MWCNTs into the PET matrix, since the decomposition temperature was shifted to higher values for nanocomposites containing up to 1 wt% MWCNTs. The activation energies for all values of α were determined with the isoconversional methods of Ozawa, Flynn, Wall, and Friedman and it was found that all nanocomposites, except the one containing 2 wt% MWCNTs, exhibited higher activation energies than pure PET. The decomposition mechanism of PET and its nanocomposites with acid-treated MWCNTs was nth-order auto-catalysis and could be described by the equation \( f\left( a \right) = \left( {1 - \alpha } \right)^{\text{n}} \left( {1 + K_{\text{cat}} X} \right) \).