Differential Scanning Calorimetry (DSC) heating and cooling curves of neat PP matrix and MWCNT/PP nanocomposites are displayed in Figure 1.

A slight increase of the crystallization temperature is observed with the addition of carbon nanotubes (Table 1). The measurements were repeated three times giving an error of ±0.5 °C corroborating that the crystallization temperature, T_c, augments up to 2.3 °C.

To confirm the effect of MWCNTs on the crystallization process, isothermal DSC measurements were carried out. The results obtained with the nanocomposite which contains 1 wt.% MWCNT are presented in Figure 2. A nucleating effect of the MWCNTs is noticed by a reduction of the crystallization time, this reduction becoming more evident as the temperature of the isothermal measurement is increased.

This is compatible with a lower viscosity, which favors the mobility of the polymer chains towards the crystallization surface and thus accelerates the crystallization process [32]. On the other hand, the evaluated crystallization degree of pure PP and the nanocomposites, also included in Table 1, reveals that the crystallization degree varies between 38% and 40% aleatory for all the samples, except for 8 wt.% MWCNT nanocomposite, which shows a lower value of 34%. This decrease of the crystallization degree observed at the highest MWCNT concentration is probably a consequence of the steric hindrance, which reduces the chain ability to crystallize, induced by the considerable amount of carbon nanotubes [4].

Many papers have been published on changes in the crystallization kinetics and nucleating effect of carbon nanotubes in semi-crystalline polymers in general [33] and PPs in particular [4,28,30,34,35,36]. The recent reviews of Bikiaris [36] and Grady [33] offer an overview of the effect of nanotubes in PP crystallization. The nucleation mechanism of polypropylene on the surface of SWCNTs and MWCNTs has been explained assuming that PP macromolecules are adsorbed and partly wrapped around the CNT, allowing to form nuclei with crystallographic c-axis perpendicular to the axis of CNTs [37]. Actually, dispersions reached at the nanoscale will favor wrapping of PP, which is compatible with the observed higher crystallization increase with the better dispersion, reported in the literature [14,38,39]. Actually, the degree of interaction between polymer and CNT plays the most important role, since weak CNT-polymer interaction reduces polymer-wrapping around CNTs. This has been pointed out by Zeng et al. [40] for the case of MWCNT/Polyoxymethylene nanocomposites, demonstrating the effect of the surface energy of CNT and the wettability of polymer on the crystallization degree. In the case of MWCNT/PP nanocomposites, the very low polarity of pure PP is an impediment for any interaction with carbon nanotubes. To avoid this and to favor dispersion of MWCNTs in PP matrixes using melt-mixing procedure, a double strategy has been developed in recent years. MWCNTs have been functionalized attaching oxygen containing groups, such as carboxyl, carbonyl or hydroxyl groups [6,7,12] and, on the other hand, coupling agents consisting of graft or block copolymers (i.e., maleic anhydride grafted polypropylene) have been used as a third component [12,14,41].

Within this context, the little effect of MWCNTs on the crystallization process of PP, as is noticed in Figure 1, Figure 2, Table 1, probably reflects the small degree of interaction between the nanofiller and the polymer in our nanocomposites. This is a logical consequence of using commercial components, such as non-modified MWCNTs and PP, which are based on the leitmotiv of this work. In any case, our polypropylene presumably contains a nucleating agent, and it could be expected that the effect of MWCNTs on the crystallization process would be more evident in a PP free of this kind of agents [33,42].

The study of the nanocomposites in the molten state is of crucial importance, because, as has been remarked specially by the group of Alig and Pötschke [16,17,18,19,25,26,43,44,45,46], a rearrangement of carbon nanotubes/polymer chains network can take place above T_m (or at approximately T_g + 100 °C in amorphous polymers), leading to the so called secondary agglomeration [15]. Besides its scientific interest, this thermally driven process is very relevant from a practical point of view, because brings about a striking improvement of the electrical conductivity of the nanocomposite.

We first investigate the effect of time on the dynamic viscoelasticity at T = 180 °C, to detect any change in the MWCNTs-PP chains interactions. The evolution of the elastic modulus (G’) with time in the molten state of MWCNTs nanocomposites with PP and polycarbonate (PC) respective matrixes, has been reported in the literature [17,20,21], revealing an enhancement of G’ for microscale dispersions [17] or a reduction of G’ for well-dispersed nanocomposites [20,21]. The increase of the elastic modulus is reported to be a consequence of the formation of a combined MWCNTs-polymer network. Our microscopy (see Figure 10) and thermal results indicate an “initial aggregation” state or microscale dispersion, so dynamic viscoelastic results can help to analyze the eventual evolution of this primary structure. The most significant result is presented in Figure 3, which shows the effect of time on the loss tangent (tan δ) plots as a function of frequency (ω).

An example of the effect of nanoparticles/polymer chain interactions on tan δ versus frequency plots is shown in the literature for phenoxy/nanoclay nanocomposites [47]: Instead of a continuous increase of tan δ as frequency tends to zero (due to the mobility of the whole chain in the flow viscoelastic zone), a tan δ maximum (tan δ)_Max appears at low frequencies. This maximum marks a limit in the mobility of the polymer chains, owing to the topological effect of the nanoparticles. Our results in Figure 3 constitute an example of the presence of a (tan δ)_Max relaxation for MWCNT/PP nanocomposites, in contrasts with neat polymers that present a continuous increase of tan δ. A progressive diminution of the height of (tan δ)_Max and a shift to lower frequencies is observed, as time increases with different sequential sweeps. The shift of the maximum signifies that the reduction of the mobility is noticed at higher frequencies (equivalent to shorter times) as time increases, which denotes more effective topological interactions between MWCNTs and PP chains.

After reaching stable values of the dynamic viscoelastic functions at T = 180 °C, the variation of the elastic (G’) and loss modulus (G’’) with frequency for the different compositions is analyzed, to evaluate the rheological percolation threshold of our thermally treated nanocomposites.

The results shown in Figure 4 indicate that as the concentration of MWCNTs increases, in particular above 2 wt.% MWCNTs, the elastic modulus tends to level off. This is considered a symptom of the interactions between nanoparticles and polymer chains that bring about a percolated network, which suppresses flow. The statistical percolation theory [48] predicts the dependence of a particular property (X) on filler concentration (Φ as a scaling law: X = X₀ (Φ − Φ_c)^t, where Φ_c is the percolation threshold and t is an adjustable parameter. This statistical percolation theory is applied considering the values of G’ taken at the lowest values of the frequency (0.0628 rad/s) divided by the lowest value of the elastic modulus for neat PP, G’(0.0628 rad/s)/G₀ value. The experimental data fitted to the scaling law of the percolation theory are presented in Figure 5. The obtained percolation threshold is Φ_c = 1, and the exponent t = 1.5.

PVT measurements offer complementary information to dynamic viscoelastic tests in the analysis of polymer chain dynamics. For instance, the lattice-hole model developed by Simha and Somcynsky [49] introduces the concept of a temperature-dependent hole fraction, h, related with free volume, which can be obtained from PVT measurements and has been correlated with transport properties [50,51]. Certainly, the eventual effect of nanoparticles on hole fraction or on free volume is a subject of interests, but very few paper refer to PVT measurements of nanocomposites. The efforts of Utracki [52,53,54,55,56,57,58] have been mostly directed to adapt and apply the PVT equations of state to nanocomposites obtained from dispersions of organoclays, in amorphous and semicrystalline polymer matrixes. The existence of a percolation threshold for the constitution of a nanoparticle/polymer chain network is not treated in these works, probably because PVT measurements are not able to detect this physical fact. Significantly enough, in Reference [55] Utracki remarks that dynamic viscoelastic measurements reveal the formation of structures of interacting nanoclay platelets in a 5% nanoclay/Polyamide 6 nanocomposite, but not for 2% concentration. However, their PVT results do not show any qualitative change between 2% and 5% nanoclay compositions.

The difficulty of applying equations of state to hybrid materials impedes us to determine the hole-fraction and free volume of our nanocomposites. However, the obtained thermodynamic coefficients, such as the thermal expansion coefficient and the compressibility coefficient (not shown), do not reveal the existence of any percolation threshold. In Figure 6, the eventual effect of the nanotube concentration on specific volume and thermal expansion coefficient is analyzed. To the difference of oscillatory flow results presented in Figure 4, which show a qualitative change in G’ curves (from convex to concave) when passing from 2 wt.% to 3 wt.% MWCNT concentration, PVT data do not follow any particular trend or alteration. This reflects the inability of this technique to detect percolation, because no particular difference is noticed in PVT results below and above the rheological percolation threshold determined by oscillatory flow experiments (Figure 5).

From these results it can be concluded that the very local polymer chain motions involved in the creation of free volume as temperature increases, are not affected, as least in our case, by the presence of MWCNTs. This is actually an expected result if one considers the effect of frequency on the dynamic viscoelastic functions, in Figure 3, Figure 4. At high frequencies (i.e., above 10² rad/s), the effect of MWCNTs is dramatically reduced and the percolation threshold cannot be detected, because the hindering effect of MWCNTs only affects to the motion of the whole chain (observed only in the terminal zone, at low frequencies).

Following the modus operandi established by Alig and Pötschke (see for instance Reference [15]), we first investigate the effect of time on electrical conductivity, at T > T_m, to correlate this result with the constitution of the secondary agglomeration, mentioned in Section 2.2. As an example, the electrical conductivity variation with time in the quiescent state at T = 180 °C for 2% MWCNT/PP nanocomposite is presented in Figure 7. This temperature is sufficiently high to facilitate the motion of the polymer chains in the terminal viscoelastic zone. The observed conductivity increase is very similar to the behavior depicted by Alig et al. [25,26], confirming the rearrangement of MWCNTs in the matrix to form conductive aggregates. The mechanism for electrical conduction, implies that within the agglomerates the distances between the nanotubes are sufficiently short to allow electron tunneling, and also implies the formation of a conductive network of interconnected agglomerates [21].

We remark that in the first annealing, which is initiated taking the molded sample (see Experimental Section), the conductivity augments seven orders of magnitude, from 10⁻¹¹ S/cm to 10⁻⁴ S/cm. However, in the second annealing, which is carried out after the annealed sample has been cooled from 180 °C to 50 °C at a rate of 1 °C/min, the reached final conductivity is the same, but the initial conductivity is five orders of magnitudes (10⁻⁶ S/cm face to 10⁻¹¹ S/cm) higher than in the first annealing. Considering the results presented in Figure 21 of reference [15], our results of Figure 7 indicate that the starting point of the second annealing is a more agglomerated dispersion than that we have at t = 0 in the first annealing. This can be explained assuming that the secondary aggregation state, reached at the end of the first annealing, is partially preserved on cooling, in particular during crystallization. The effect of annealing time on the electrical conductivity is particularly dramatic for 2% MWCNT/PP nanocomposite, probably because this composition is relatively close to the rheological percolation threshold and in the limit of the electrical percolation threshold in the molten state, as it is seen below. Actually, for a concentration of 4 wt.% MWCNT the effect of annealing is less severe, since the observed increase of the electrical conductivity is only o approximately two orders of magnitude.

The analysis of the electrical conductivity as a function of frequency for different MWCNTs concentrations in the molten state, at T = 160 °C (not presented), shows a practically frequency independent DC conductivity (except for pure PP and 1% MWCNT/PP nanocomposite) and a power law frequency dependency, σ(f) = A υ^S, which stands for AC conductivity, above a certain critical frequency value. The evolution of the conductivity of the nanocomposites taken at a frequency of 20 Hz, which corresponds to the constant conductivity zone for the samples that have DC conductivity, on cooling from the melt, is displayed in Figure 8.

We first remark that during crystallization, at approximately 125 °C–130 °C, an abrupt conductivity decrease is observed, similarly to that reported in the literature [59]. In a recent paper about the increase of electrical conductivity of MWCNT/Polyurethane (PUR) nanocomposites during crystallization, we summarize literature results of different nanocomposites (Table 1 of reference [59]). Two hypotheses are contemplated for the case of a conductivity decrease during crystallization: (a) Reduction of the ionic conductivity of the nanocomposite due to a reduction of the amorphous phase of the matrix, considered by Alig et al. [26] and (b) Destruction of the conductive pathways, because conducting nanoparticles are expelled from the crystalline phase, contemplated by Pang et al. [60]. The observation of Figure 8 leads to remark that the conductivity decrease is much more severe in the case of 2% MWCNT concentration than in the rest of the sample: A reduction of two orders of magnitude face to less than one. To investigate this result, which is linked to the proximity of 2% MWCNT concentration to the rheological percolation limit (see above), in the continuation we analyze the electrical percolation threshold in the molten state, at T = 160 °C, and in the solid state, at T = 50 °C. The obtained results are helpful to disclose the origin of the observed conductivity changes during crystallization.

The effect of MWCNTs concentration on the electrical conductivity of the nanocomposites at respective temperatures of 160 °C and 50 °C is presented in Figure 9. The corresponding fittings to the equation of the theory of percolation, X = X₀ (Φ − Φ_c)^t, (see section 2.2), taking X as the electrical conductivity (σ’), give the following values for the adjusting parameters: Φ_c = 1.2 and t = 2.7 for T = 160 °C and Φ_c = 2 and t = 1.8 for T = 50 °C. Therefore, the percolation threshold is significantly higher after nanocomposites have crystallized.

In particular, the concentration of 2% MWCNTs is above the percolation threshold estimated in the molten state (Φ_c = 1.2%), but is just in the percolation threshold (Φ_c = 2%) estimated in the solid state. In view of these results, we can consider that the percolating network, associated to the secondary agglomeration reached in the molten state for 2% MWCNT/PP nanocomposite, is rather unstable and is destroyed during crystallization process, because of crystals growth. However, for concentrations clearly above the percolation threshold in the molten state, Φ_c = 2%, such as 3% up to 8% MWCNT shown in Figure 8, the constituted network is sufficiently strong not to be destroyed during crystallization and the conductivity decrease is much more smaller, approximately one order of magnitude. This reduction may be due to a reduction of the ionic conductivity of the nanocomposite, compatible with a reduction of the amorphous phase of the polymer matrix, as proposed by Alig et al. [60]. These results are different, but not contradictory, with those obtained with MWCNT/PUR nanocomposites [60], which show a conductivity decrease during crystallization for Φ < Φ_c, only compatible with a ionic conductivity reduction, but a conductivity increase for Φ > Φ_c, that is explained by a reorganization of the conductivity network inside the aggregates, brought about by crystal growth.

Our electrical conductivity results come to confirm that maintaining the nanocomposites for a certain time at temperatures at which the polymer chain mobility is allowed (T > T_m for semicrystalline polymers) brings about high electrical conductivities, with respect to the observed values in polymer nanocomposites, although a reduction of approximately one order of magnitude should be expected during solidification process. From a practical point of view, this constitutes a reliable strategy to obtain semiconductor polymers, face to complex elaboration procedures, based on modifying the natural lack of interaction between carbon nanotubes and polymer chains, which do not offer necessarily better results.

Considering the scope and the aims of the paper, a significant characteristic of the filler and the polymer matrix used in this work is that both are non-modified, which indicates that the simplest commercially available MWCNTs and polypropylene are employed in the elaboration of the nanocomposites.

The investigated isotactic polypropylene (Moplen EP340K; M_w = 440,000 g/mol, M_w/M_n = 4.4) was purchased from Basell Polyolefines Company (Ferrara, Italy), whereas Multiwall Carbon Nanotubes (MWCNT), which have specified diameters of D = 30–50 nm, lengths between 10 and 20 μm and purity greater than 95%, were supplied by Cheap Tubes Inc. (Brattleboro, VT, USA).

A very single procedure was used to disperse the MWCNTs in the iPP matrix. Before the melt-mixing process, polymer powder was prepared from pellets using a Mill Retsch® ZM 200 (Retsch Technology, Düsseldorf, Germany) and both, polymer powder and MWCNTs, were properly dried (PP 80 °C at vacuum, 2 h; MWNT 100 °C, 2 h). The polymer and MWCNTs were stirred to obtain a homogeneous mixture. The powder mixture was blended in a Haake Mini-Lab twin-screw extruder (Thermo Electron Corp., Hamburg, Germany). The mixing was processed at T = 180 °C and 100 rpm using a counter-rotating screw configuration for 10 min. Extrudate was cooled at room temperature and then pelletized.