Introduction

Polymer nanocomposites are a class of materials consisting of a polymeric matrix containing organic or inorganic nanofillers, and it is a steadily growing field both among academia as well as industry [1], owning the unique, versatile effect that nanofillers can have on polymeric materials. Some of the properties of polymers that can be altered based on the target application by the incorporation of nanofillers are mechanical [2], thermal [3, 4], crystallization [5,6,7,8,9], electrical [10], magnetic [11], optical [12] and biological properties [13]. Nanoclays are one important category of nanofillers that has been extensively studied during the last decades, including layered silicates and nanosized clay minerals. They possess important advantages such as low cost, abundance, relatively large aspect ratios and can be easily chemically or physically functionalized; as a result, they are already being used in the paper and ceramics industries, as well as in oil drilling [14, 15]. However, their layered structure makes exfoliation into individual sheets challenging, since strong interlayer bonds exist between the layers, forming agglomerated tactoids in nanocomposites.

Halloysite is a different type of clay nanofiller, since in contrast with the common layered aluminosilicate minerals, it possesses a nanotubular structure with high aspect ratio. Its chemical composition is similar to that of kaolinite (Al2Si2O5(OH)4·nH2O) [16]. Owning its structure, it shows a variety of unique properties that make it attractive for applications in polymeric nanocomposites, drug release and biomedical applications [17, 18]. The inner surface of the nanotubes is comprised of aluminol groups, and the outer mainly of siloxanes, with a few aluminols and silanols on the defects of the outer edges [1]. That makes HNTs relatively hydrophobic, which makes their dispersion in non-polar polymers easy, while the few hydroxyls on the surface provide with active sites for chemical functionalization [18]. The nanotubular structure also makes HNTs a great carrier for biologically active substances [17].

PCL is an aliphatic, biodegradable polyester that can be synthesized either by the ring-opening polymerization of ε-CL, or by the polycondensation of 6-hydroxyhexanoic acid. It is a semi-crystalline, hydrophobic polymer with low melting temperature (59–64 °C). PCL is a FDA-approved polymer [19] for hard and soft tissue applications, biocompatible and miscible with several polymers, easily processed and molded, properties that led to a wide range of applications, especially in tissue engineering, drug delivery and food packaging [20, 21]. Nevertheless, its poor thermal and mechanical properties prevent PCL from being used in load bearing applications; however they can be tailored with the use of nanofillers [4].

Polymeric nanocomposites containing HNTs can be synthesized with the already established methods for nanocomposite fabrication, i.e., solution casting, melt mixing and in situ polymerization [22]. In situ polymerization possesses several advantages as a technique for the fabrication of nanocomposites, such as one-pot synthesis of nanocomposites, better dispersion of the fillers compared with the other methods, improved interfacial adhesion, and consequently better properties of the final product [22,23,24,25,26]. Despite its superiority, in situ polymerization has not been studied in a great extend for the preparation of nanocomposites of polyesters with HNTs. Lahcini et al. synthesized PCL/HNTs nanocomposites via the ring-opening polymerization of ε-CL and found that the hydroxyl groups of HNTs can act as co-initiators along with tin tetrachloride [27]. Other in situ synthesized nanocomposites reinforced with HNTs that exhibited enhanced properties were based on polyethylene/ethylene methacrylate copolymer [28], polypropylene [29], polystyrene [30, 31], poly(vinyl chloride) [32], and biobased polyurethanes [33, 34].

In order to improve the interfacial adhesion of HNTs with polymeric matrices, several methods for their surface functionalization have been employed [1]. One of the most common ones is organosylilation, with several silanes containing different end groups [19]. Grafting of 3-aminopropyltriethoxysilane (APTES) on the surface of HNTs was found to improve their dispersion in PLA:PCL nanofibers [35] and in polypropylene nanocomposites prepared by injection molding [36]. Vinyltrimethoxysilane functionalization improved interfacial adhesion in unsaturated polyester nanocomposites with HNTs [37], while tetraethoxysilane grafting improved the barrier properties and decreased moisture absorption of polyimide/HNTs nanocomposites [38]. In the present study, γ-aminopropyltriethoxysilane (APTES) was used to functionalize HNTs because its amino group can interact with the ε-CL monomer during ring-opening polymerization and form covalent bonds, resulting in strong bonding between HNTs and PCL, which is expected to improve the dispersion and properties of the final nanocomposite materials [39].

The aim of this study is the synthesis of PCL nanocomposites with HNTs and HNTs–APTES with in situ ring-opening polymerization of ε-CL. The effect of the functionalization and content of HNTs on the thermal and crystallization properties of PCL were studied with a number of experimental techniques, including intrinsic viscosity measurements, scanning electron microscopy (SEM), wide-angle X-ray diffraction (WAXD), differential scanning calorimetry (DSC), polarizing light optical microscopy (PLOM) and thermogravimetric analysis (TGA).

Experimental

Materials

Halloysite nanoclay (kaolin clay) has the molecular formula Al2Si2O5(OH)4·2H2O and molecular weight 249.19 amu. Its surface area was measured equal to 53.3 m2/g with 0.371 mL/g pore volume and the approximate size of nanotubes was 30–70 nm × 3–8 μm. ε-Caprolactone (CL) monomer was of purity 99% and was further purified by distillation. Catalyst tetrabutyl titanate (TBT) and γ-aminopropyltriethoxysilane (APTES) were of analytical grade. All chemicals were supplied by Sigma-Aldrich.

Functionalization of nanofillers

HNTs were dehydrated at 100 °C under vacuum and functionalized with APTES in order to add amine reactive groups on their surface. Briefly, 2 g of APTES was dissolved in dry toluene, and 2 g of nanoclay was dispersed in the reaction mixture with the help of ultrasonic treatment. The reaction was refluxed at 120 °C for 20 h, while a calcium chloride tube was attached to the system. The solid phase was filtered and washed 3 times with toluene and 3 times with ethanol to ensure the removal of excess organosilane [40].

Synthesis of nanocomposites

For the synthesis of PCL, ε-CL was dried over CaH2 and purified by distillation under reduced pressure prior to use. The bulk polymerization of ε-CL was carried out in 250-mL round-bottom flask equipped with a mechanical stirrer and a vacuum apparatus. The catalyst TBT was added as a solution in toluene at a final concentration of 1 × 10−4 mol per mole of monomer. The polymerization mixture was de-gassed and purged with dry argon three times. The ring-opening polymerization (ROP) reaction was carried out for 2 h at 190 °C, followed by increasing of the reaction temperature from 210 to 240 °C over a period of 45 min. Unreacted monomer was removed through distillation by applying a high vacuum (≈ 5 Pa) slowly, to avoid excessive foaming, over a time period of 15 min. Polymerization was terminated by rapid cooling to room temperature.

For the preparation of PCL/ΗΝΤs and PCL/HNTs–APTES nanocomposites by in situ polymerization the same procedure was followed. HNTs or HNTs–APTES together with the proper amount of TBT were added in ε-CL monomer and were sonicated initially for 2 min using a tip sonication apparatus, and then in a sonic bath for 15 min. Polymerization was then carried out. Nanocomposites containing 0.5, 1 and 2.5 wt% HNTs and HNTs–APTES have been prepared.

The prepared materials were afterward hot pressed using an Otto Weber, Type PW 30 hydraulic press connected with an Omron E5AX Temperature Controller, at a temperature of 75 ± 5 °C, in order to prepare films of different thicknesses, appropriate for each type of following measurements.

Characterization techniques

TEM experiments were carried out on a JEOL 2011 TEM with a LaB6 filament and an accelerating voltage of 200 kV. The specimens were prepared by evaporating drops of HNT and HNT–APTES ethanol suspensions after sonication onto a carbon-coated lacy film supported on a 3 mm diameter, 300 mesh copper grid.

The element composition and interfacial interactions were studied using X-ray photoelectron spectroscopy (XPS). In the XPS analysis (AXIS Ultra ‘DLD’ X-ray photoelectron spectrometer) (Shimadzu Corp., Japan), a monochromatic Al Ka X-ray was used at 14 kV. A binding energy of 284.8 eV for the C 1s of aliphatic carbons was taken as the reference energy. XPS curve fitting of was accomplished using Kratos’ Vision 2 Processing software.

Fourier transform infrared spectroscopy (FTIR) spectra of all the samples were obtained using a Perkin-Elmer FTIR spectrometer, model Spectrum One. The materials were in the form of thin films with thickness of approximately 15 mm. The IR spectra of these films were obtained in absorbance mode and in the spectral region of 400–4.000 cm−1 using a resolution of 4 cm−1 and 64 co-added scans.

Wide-angle X-ray diffraction (WAXD) measurements of the samples were performed using a MiniFlex II XRD system from Rigaku Co, with CuKα radiation (λ = 0.154 nm) at the area of 5 < 2θ < 60°. Crystallinity calculations of the studied materials were performed by fitting the WAXD profiles with Gaussian–Lorentzian cross product curves and deriving the ratio of crystalline over overall fitted peaks area [41, 42]. Crystallinity of the polyesters was determined by Eq. 1:

$$ \% {\text{Crystallinity}} = \frac{{{\text{Area}}\;{\text{under}}\;{\text{crystalline}}\;{\text{peaks}}}}{{{\text{Total}}\;{\text{area}}\;{\text{under}}\;{\text{peaks}}}} \cdot 100\% $$
(1)

Nitrogen adsorption/desorption experiments at − 196 °C were performed for the determination of surface area (multi-point BET method), total pore volume (at P/Po = 0.99), and pore size distribution (BJH method using adsorption data) of the HNT and HNT–APTES samples which were previously outgassed at 250 and 80 °C, respectively, for 16 h under 6.6 × 10−9 mbar vacuum using an Automatic Volumetric Sorption Analyzer (Autosorb-1MP, Quantachrome).

Intrinsic viscosity [η] measurements were performed using an Ubbelohde viscometer at 30 °C in toluene. The sample was maintained in the solvent at room temperature for some time in order to achieve a complete solution. The solution was then filtered through a disposable membrane filter made from Teflon. Intrinsic viscosity of the samples was calculated using the Solomon–Ciuta equation:

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

where C the concentration of the solution, t the flow time and t0 the flow time of the solvent. The number-average molecular weight (\( \bar{M} \)n) of the samples was calculated from the intrinsic viscosity values using the following equation [43]:

$$ [\eta ] = 1.4 \cdot 10^{ - 4} \cdot \bar{M}n^{0.78} $$
(3)

The morphology of the prepared nanocomposites was examined using a scanning electron microscopy system (SEM) type JEOL (JMS-840) (Jeol Ltd., Tokyo) equipped with an energy-dispersive X-ray (EDX) Oxford ISIS 300 micro-analytical system (Oxford Instruments UK). The studied surfaces were coated with carbon black in order to obtain good conductivity for the electron beam. Operating conditions were: accelerating voltage 20 kV, probe current 45 nA and counting time 60 s.

All differential scanning calorimetry experiments were performed on a TA Instruments TMDSC (TA Q2000) combined with a cooling accessory and calibrated with indium for the heat flow and temperature measurements. All sample masses were kept close to 5 mg and the Tzero sample and reference pans were of identical mass. The samples were initially heated to 50 °C higher than the melting point and held there for 5 min in order to erase any thermal history. Afterward, the sample was quenched to − 90 °C and then reheated with 10 °C min−1. The glass transition temperature and the melting points were recorded this way. For non-isothermal crystallization, the samples were heated from room temperature up to 140 °C, held there for 5 min and cooled down to room temperatures at rates varying from 5 to 20 °C min−1. A fresh sample was used in each run.

For the isothermal crystallization experiments, temperatures of 40–46 °C were pre-selected and the experiments were performed after self-nucleation of the polyester samples, in a procedure proportional with the one described by Muller et al. [44,45,46]. The protocol can be summarized as follows: (a) heating of the sample and remaining at 140 °C or at the proper temperature for each copolymer for 5 min in order to erase any previous thermal history; (b) subsequent cooling at a rate of 10 °C min−1–40 °C or at the proper temperature for each copolymer and crystallization for 10 min, which creates a “standard” thermal history; (c) partial melting by heating at 20 °C min−1 up to a “self-nucleation temperature,” Ts; thermal conditioning at Ts for 1 min. Depending on Ts, the crystalline polyester domains will be completely molten, only self-nucleated or self-nucleated and annealed. If Ts is sufficiently high no self-nuclei or crystal fragments can remain and the sample is then in the so-called Domain I, the complete melting domain. At intermediate Ts values, the sample is almost completely molten but some small crystal fragments or crystal memory effects remain, which can act as self-nuclei during a subsequent cooling from Ts, and the sample is said to be in Domain II, the self-nucleation domain. Finally, if Ts is too low, the crystals will only be partially molten, and the remaining crystals will undergo annealing during the 5 min at Ts, while the molten crystals will be self-nucleated during the later cooling, and the sample is in Domain III, the self-nucleation and annealing domain.; (d) cooling scan from Ts at 200 °C min−1 to the crystallization temperature (Tc), where the effects of the previous thermal treatment will be reflected on crystallization; (e) heating scan at 20 °C min−1 to 120 °C or to the proper temperature for each copolymer, where the effects of the thermal history will be apparent on the melting signal. Experiments were performed to check that the sample did not crystallize during the cooling to Tc and that a full crystallization exothermic peak was recorded at Tc. In heating scans after isothermal crystallization, the standard heating rate was 20 °C min−1. If some other rate was used, this will be discussed in the specific section.

A polarizing light microscope (Nikon, Optiphot-2) equipped with a Linkam THMS 600 heating stage, a Linkam TP 91 control unit and a Jenoptic ProgRes C10Plus camera was used for PLM observations.

Thermogravimetric analysis was carried out with a Setsys 16/18 TG–DTA (Setaram). Samples (4 ± 0.2 mg) were placed in alumina crucibles and heated from ambient temperature to 620 °C at 20 °C min−1 in a 50 ml min−1 flow of N2; an empty alumina crucible was used as reference.

Results and discussion

Functionalization of HNTs

Initially, the morphology of HNTs was studied by TEM images (Fig. 1). The HNTs clearly possess a tubular structure with multilayer walls, and their lumen is open to the ends of each tube. The lumen’s diameter was calculated between 9 and 12 nm and the outer diameter 30–70 nm. Their length appears to vary, ranging from 3 to 8 μm. The tubular morphology is retained after the functionalization reaction (Fig. 1b); however, the diameter of HNTs appears to be slightly bigger, up to 110 nm. This can be attributed to the possible formation of a thin layer or oligomerized APTES on the surface of HNTs that could lead to increased external diameter [40].

Figure 1
figure 1

TEM images of a HNTs, b HNTs APTES

Full size image

Silylation is one of the most common methods to covalently functionalize HNTs, and the reaction occurs between the hydroxyl groups on the surface of the nanotubes and the silanol groups of the silane. It has been found that besides the direct bonding of APTES on the hydroxyls, oligomerization of hydrolyzed silane can also take place in the presence of physically absorbed water [47]. FTIR spectra of pure and functionalized HNTs are presented in Fig. S1. The characteristic peaks observed can be attributed to hydroxyls (3600–3700 cm−1), planar Si–O stretching (1025–1050 cm−1), symmetric Si–O bending (795 cm−1), vertical Si–O bending (690, 795 cm−1) and Al–O–Si stretching (475 cm−1). After the functionalization procedure, a few extra peaks appear on the spectrum that are caused by the N–H2 bending (1560, 1490 cm−1) and symmetric C–H stretching (2936 cm−1) of APTES.

The surface atomic percentages as determined by XPS are summarized in Table 1, and the high-resolution XPS spectra of HNTs and HNTs–APTES are presented in Fig. S2 and S3, respectively. HNTs contain oxygen, silicon and aluminum with a Si/Al ratio of 1, typical for a two layered aluminosilicate. The small amount of carbon detected is due to surface organic contamination. The success of the organosilane grafting is confirmed by the increase in Si and C atomic percentages, the emerging of peaks that correspond to nitrogen, and the increase in the Si/Al ratio [19]. The N 1s band (Fig. S3e) can be deconvoluted into 2 peaks at 407 and 405 eV that correspond to protonated amino groups and N–H bonds of free amino groups, respectively, [48]. The reaction that took place between HNTs and APTES is presented in Scheme 1.

Table 1 Surface atomic percentages determined by XPS for HNTs and HNTs–APTES
Full size table
Scheme 1
scheme 1

Schematic presentation of the functionalization reaction of HNTs with APTES

Full size image

X-ray diffraction patterns of pure and functionalized HNTs are presented in Fig. S4. The characteristic peak of the (001) crystal plane of HNTs appears at 11.56° with a lattice spacing of d = 7.68 Å (determined by Bragg’s law) that shifts to 11.18° with d = 7.94 Å after functionalization. This indicates the HNT is type halloysite-7Å, i.e., the dehydrated form [18, 47, 49]. Moreover, HNTs exhibit diffraction peaks at 2θ = 19.7° και 24.16° that result from the crystal planes (110) and (002), respectively, and shift to 2θ = 19.16° and 23.78° [50]. The organosilylation of HNTs did not affect the crystalline structure; however, it led to increase in the basal spacing of HNTs.

The effect of the organosilylation on the surface area and pore volume of HNTs was studied by nitrogen adsorption/desorption experiments with the multi-point BET method. The nitrogen adsorption–desorption isotherms and the respective pore size distribution curves of HNTs and HNTs–APTES are shown in Figs. S5 and S6, respectively, while the surface area, total pore volume, and average pore size data are presented in Table 2. HNTs have a surface area value of 53.5 m2/g, which is reduced to 47.2 m2/g after functionalization. This can be attributed to the presence of silane molecules and the changes in the stacking of the nanotubes [19]. The isotherm curves of Fig. S5 are characterized as type IIb which is typical for natural clays, possessing very small hysteresis loops that reveal the presence of large mesopores [47].

Table 2 Porosity characteristics of HNTs and HNTs–APTES
Full size table

TGA thermograms of HNTs and HNTs–APTES are presented in Fig. S7. The first step of weight loss occurs at 87 and 104 °C for HNTs and HNTs–APTES, respectively, and is caused by loss of free interlayer water, about 3% in both cases [51]. Pure HNTs lose 14.3% of their weight between 400 and 500 °C that can be attributed to the decarboxylation of the aluminol groups that are present on the surface of the nanotubes [18, 19]. HNTs–APTES lose 16.5% of their weight in the same temperature region and that 2.2% of extra weight loss can be attributed to the degradation of the organosilane. The organic content of HNTs–APTES clearly reduces their thermal stability, as expected.

Synthesis and characterization of PCL nanocomposites

Synthesis of PCL nanocomposites via in situ polymerization

The ROP of CL is effectively catalyzed by metal alkoxides, such as TBT, with the first step of the reaction mechanism being the complexation with the monomer at the carboxyl oxygen, followed by the insertion of the lactone into the metal–oxygen bond [52]. Molecular weight values, as determined by IV measurements, are presented in Table 3. The presence of the nanofillers in the polymerization mixture resulted in lower molecular weight compared to neat PCL, which has a \( \bar{M} \)n value of 63435, while all nanocomposites display lower \( \bar{M} \)n, ranging from 39591 to 61357. The effect of the nanofillers on the \( \bar{M} \)n is more pronounced as the nanotube concentration in the polymeric matrix increases. The hydroxyl groups of HNTs and the amino groups of HNTs–APTES could act as co-initiators for the ROP of CL that would lead to a higher concentration of chain-ends and therefore lower molecular weight values [27, 39]. In the case of HNTs–APTES, higher \( \bar{M} \)n values were obtained in comparison with unfunctionalized HNTs. This could be attributed to the fact that every 3 of their surface hydroxyls were used as binding sites for the organosilane, making the number of the –NH2 groups smaller than the number of free hydroxyls in unfunctionalized HNTs. It has also been suggested by Vassiliou et al. [43] that this reduction in the \( \bar{M} \)n could be caused by the trapping of the active catalyst species in heterogeneous exchange reaction between the TBT and the nanofiller’s surface silanol groups. However, all synthesized polymers exhibited satisfactory IV and \( \bar{M} \)n values that rendered them suitable for further characterization and confirmed the synthesis parameters as viable.

Table 3 Molecular weights determined by viscometry
Full size table

Morphological characterization

Morphological observation of the nanocomposites was conducted by SEM (Figs. 2, 3). The dispersion of nanofillers is one of the most important factors that affect the final performance and properties of nanocomposites, since aggregates can act as premature failure points and result in low mechanical properties. HNTs appear to be well dispersed in the polymeric matrix, with no obvious aggregates, in all HNT contents. However, a number of individual nanotubes are sticking out of the surface and appear to be not fully embedded into the PCL matrix. Functionalization of inorganic nanofillers with organic compounds enhances their chemical affinity with polymeric matrices, resulting in stronger interactions, homogenous dispersion and improved mechanical properties [53]. Compared to unfunctionalized nanotubes, HNTs–APTES (Fig. 3) appears to be better entrapped into PCL, especially in the lowest concentrations 0.5, since less nanotubes are sticking out of the surface, suggesting the formation of a stronger interface between the components of the system. In both cases, dispersion appeared to be satisfactory, with no visible aggregates.

Figure 2
figure 2

SEM micrographs of a PCL/HNTs 0.5%, b PCL/HNTs 1.0%, c PCL/HNTs 2.5 wt% in different magnifications

Full size image
Figure 3
figure 3

SEM micrographs of a PCL/HNTs APTES 0.5%, b PCL/HNTs APTES 1.0%, c PCL/HNTs APTES 2.5 wt% in different magnifications

Full size image

Structural characterization

FTIR spectra of PCL and the nanocomposites are presented in Fig. 4a, b. The characteristic peaks of PCL can be observed with no noticeable differences in the presence of the nanofillers. The strongest bands correspond to asymmetric CH2 stretching at 2942 cm−1, symmetric CH2 stretching at 2865 cm−1, carbonyl stretching at 1728 cm−1, asymmetric COC stretching at 1243 cm−1, and symmetric COC stretching at 1173 cm−1. Due to the small HNT content, filler peaks cannot be detected in the spectra. However, alterations on the peak intensities and ratios before and after HNT incorporation were noticed. Therefore, carbonyl indices of PCL and its nanocomposites were calculated by dividing the integral normalized intensity of the C=O peak at 1728 cm−1 with the sum of the normalized integral intensities of the –CH2– peaks at 1470 and 1398 cm−1, and the results are presented in Fig. 4c. The number of carbonyls increased with increasing HNT content, with a more pronounced increase for the nanocomposites with HNT–APTES. This increase could be due to the increased amount of carboxyl end groups, as the molecular weight decreases after the incorporation of nanofillers, while in the HNTs–APTES nanocomposites, the possible formation of amide bonds between the amino groups of the modified nanofiller and the carbonyl of the lactone may contribute in the increased amount of C=O groups. The possible resulting structure of the PCL/HNTs–APTES nanocomposites is presented in Scheme 2.

Figure 4
figure 4

FTIR spectra of a PCL/HNTs, b PCL/HNTs–APTES and c carbonyl indices of PCL and its nanocomposites

Full size image
Scheme 2
scheme 2

Schematic presentation of in situ synthesis of PCL/ΗΝΤs and PCL/HNTs–APTES nanocomposites

Full size image

WAXD diffraction patterns of PCL and its nanocomposite are presented in Fig. 5, and the crystallinity values calculated by Eq. 1 in Table 4. PCL exhibits crystalline peaks at 2θ = 21.54°, 22.12° and 23.8° that correspond to the (110), (111) and (200) planes of the orthorhombic unit cell [54]. Also, a small peak of PCL at about 2θ = 17° can be noticed [55,56,57,58], and its intensity is diminished in the presence of the nanofillers.

Figure 5
figure 5

WAXD patterns of a PCL/HNTs nanocomposites and b PCL/HNTs APTES nanocomposites (inset: X-ray diffractograms of the respective nanofillers)

Full size image
Table 4 Xc (%) of the polyesters as calculated by WAXD
Full size table

After the incorporation of the pristine and functionalized HNTs into PCL, the diffraction peaks did not shift, implying that the crystal structure of the polymeric matrix was not altered by the nanofillers. Additionally, the peaks of pure HNTs and HNTs–APTES did not appear in the patterns of the nanocomposites, which is an indication of good dispersion and absence of aggregates [59]. However, it is also possible that the diffraction peaks of HNTs could not be detected due to the small amount of fillers in the polymer.

The WAXD profiles were fitted with Gaussian–Lorentzian cross product curves and the results are presented in Fig. S8. The crystallinity Xc % of the polymers was calculated by obtaining the ratio of the area of crystalline peaks over the overall area under the diffractogram, and the values are presented in Table 4. There are no major differences between the crystallinity of the matrix and the nanocomposites, as expected due to the small amount of filler.

Thermal characterization

The thermal properties of PCL and nanocomposites were evaluated by DSC. Like most nanofillers, HNTs can accelerate the crystallization behavior of PCL, which is generally a slow-crystallizing polyester, by acting as heterogeneous nucleation points [60, 61]. As it can be seen in Fig. S9, the glass transition temperature (Tg) of the matrix and the composites was observed at − 63 °C, while the melting temperature (Tm) was observed at 56 °C. The presence of the fillers did not seem to affect the thermal characteristics (Tg, Tm) of the matrix during the initial assessment. The crystallization behavior is very important aspect of their nature as it can determine their final degree of crystallinity and the morphology of the materials after processing [62]. Several studied have found that silicon-based nanofillers like Polyhedral Oligomeric Silsesquioxanes (POSS) enhance the isothermal melt crystallization of PCL in low concentrations [63,64,65], as well as carbon-based nanofillers [66]. In the following, we will evaluate the crystallization behavior of PCL and nanocomposites under isothermal and dynamic conditions.

Isothermal crystallization: Avrami theory

The samples were crystallized at different temperatures, namely 40, 42, 44 and 46 °C and the heat flow curves with increasing time were obtained, as shown in Fig. S10 only for pure PCL, for brevity reasons. Moreover, the relative degree of crystallinity was obtained from Eq. 4:

$$ X(t) = \frac{{\int_{0}^{t} {({\text{d}}H_{\text{c}} /{\text{d}}t){\text{d}}t} }}{{\int_{0}^{\infty } {({\text{d}}H_{\text{c}} /{\text{d}}t){\text{d}}t} }} $$
(4)

where dHc denotes the measured enthalpy of crystallization during an infinitesimal time interval dt.

The evolution of the relative degree of crystallinity versus time can be seen in Fig. 6, for pure PCL and for the PCL/HNTs–APTES 1% sample, at different crystallization temperatures. It is obvious that the crystallization time increases with increasing temperature, indicating that nucleation is a limited parameter of crystallization, while the presence of the functionalized fillers in the nanocomposite sample enabled faster crystallization, as it was expected. The half-time of crystallization (t1/2) was obtained from the curves in Fig. 6, and the results are presented in Fig. 7. All nanocomposites exhibit smaller t1/2 than the matrix, but it is also important to notice that the t1/2 of the nanocomposites containing the functionalized filler accelerated the crystallization of PCL, as a result of the enhanced interactions with the matrix.

Figure 6
figure 6

Evolution of relative degree of crystallinity as a function of time for a PCL and b PCL/HNT–APTES 1.0% under isothermal conditions

Full size image
Figure 7
figure 7

Half-time of crystallization versus the crystallization temperature for each sample

Full size image

A number of macrokinetic models have been proposed in literature [62] to describe the effect of the fillers on isothermal crystallization, but the theory proposed by Avrami [67,68,69] is most probably the most used one. According to the modified Avrami method, the relative degree of crystallinity can be calculated from Eq. 5:

$$ X(t) = 1 - \exp ( - kt^{n} )\quad {\text{or}}\quad X(t) = 1 - \exp [ - (Kt)^{n} ] $$
(5)

where n is the Avrami exponent, a function of the nucleation process and k is the growth function, which is dependent on nucleation and crystal growth. Since the units of k are a function of n, the Avrami equation can be written in the composite Avrami form using k = Kn. In this work, the nonlinear curve fitting procedure was followed, based on the Levenberg–Marquardt method and the results are summarized in Table 5. The guidelines presented from Lorenzo et al. on the correct use of Avrami equation to fit the isothermal crystallization data, were carefully followed [70]. It can be seen that for all samples, n was between 2.5 and 4, indicating 3-dimensional nucleation [43]. Moreover, the growth function decreased with increasing Tc and increased with increasing filler content, consistent with the results from the t1/2.

Table 5 Avrami parameters of PCL and nanocomposites
Full size table
Lauritzen–Hoffman analysis

The kinetic data from isothermal crystallization has also been analyzed with Lauritzen–Hoffman secondary nucleation theory [71]. According to this, the growth rate G is given as a function of temperature Tc by Eq. 6:

$$ G = G_{0} \exp \left[ { - \frac{U^*}{{R(T_{\rm c} - T_{\infty } )}}} \right]\exp \left[ {\frac{{K_{\rm g} }}{{T_{\rm c} (\Delta T)f}}} \right] $$
(6)

where the first exponential term contains the contribution of the diffusion process to the growth rate and G0 is the pre-exponential factor, while the second exponential term represents the contribution of the nucleation process. Moreover, the term U* denoted the activation energy of the molecular diffusion across the interfacial boundary between crystalline and melt state and it is usually set at 1500 cal mol−1. Also, T∞ = Tg − 30 K, and it is the temperature below which the diffusion stops, ΔΤ denotes the degree of undercooling (ΔΤ = T 0m  − Tc), f is a correction factor which is close to unity at high temperatures and is given as f = 2Tc/(T 0m  + Tc), while Kg is the nucleation parameter. In the current work, the equilibrium melting point was set equal to 78 °C for PCL, while the Tg = − 66 °C. The nucleation parameter Kg can be obtained by the double logarithmic transformation of Eq. 7:

$$ \ln (G) + \frac{{U^{*} }}{{R(T_{\text{c}} - T_{\infty } )}} = \ln (G_{0} ) - \frac{{K_{\text{g}} }}{{T_{\text{c}} (\Delta T)f}} $$
(7)

and after plotting the left-hand side of Eq. 7 with respect to 1/Tc(ΔΤ)f a straight line should appear with a slope equal to Kg.

The common empirical approximation of the spherulite growth rate is inversely proportional to the crystallization half-time G ≈ 1/t1/2 has been used here for the evaluation of the nucleation parameter [72]. The t1/2 values presented earlier have been used for this reason and the Lauritzen–Hoffman type plots for PCL and nanocomposites are presented in Fig. 8. In addition, for a secondary or heterogeneous nucleation, Kg can be obtained from Eq. 8:

$$ K_{\text{g}} = \frac{{jb_{0} \sigma \sigma_{\text{e}} T_{\text{m}}^{0} }}{{k_{\text{B}} \left( {\Delta h_{\text{f}} } \right)}} $$
(8)

where j is a constant equal to 2 for regime II and equal to 4 for regime I and III, also σ, σe are the side surface (lateral) and fold surface (end) free energies which represent the work needed for the creation of a new surface, Δhfρc = ΔHf is the enthalpy of melting per unit volume, b0 is the single layer thickness, and kB is the Boltzmann constant (kB = 1.38·10−23 J K−1). The values used here were a0 = 4.5 Å, b0 = 4.1 Å, and U* = 1500 cal mol−1 which correspond to PCL and were also adopted for all materials. ΔHf for PCL was taken equal to 136 J g−1, while the unit cell density was taken 1.175 g cm−3. Moreover, it was supposed that regime II holds at this temperature range and thus j = 2. According to Thomas and Staveley, the lateral surface free energy can be obtained by \( \sigma = \alpha \Delta h_{f} \sqrt {a_{0} b_{0} } \) where the a0 and b0 factors are the molecular width and the molecular layer thickness, respectively, and α was derived to be empirically equal to 0.1 as this is a commonly accepted value for flexible polyesters [72]. Finally, the work of chain folding can be obtained from the fold surface free energy as \( q = 2\sigma_{\text{e}} \alpha_{0} b_{0} \).

Figure 8
figure 8

Lauritzen–Hoffman plots for a PCL/HNTs and b PCL/HNTs–APTES nanocomposites

Full size image

The calculated values of Kg can be found in Table 6, and it can be seen that they are lower than the ones of PCL for all nanocomposite samples. This is an indication that the foreign particles within PCL reduce the critical nucleus size needed for crystal growth and the free energy that acts against primary nucleation. Moreover, the lower values of the work of chain folding q for the nanocomposites indicate that the folding of macromolecular chains proceeds easier (especially in the samples filled with HNTs–APTES), therefore facilitating the growth of crystals. The results are very close to previously published literature works [73, 74].

Table 6 Results from the Lauritzen–Hoffman analysis for PCL and nanocomposites
Full size table
Non-isothermal Crystallization

The dynamic crystallization of PCL from the melt in the presence of untreated and functionalized HNTs was studied in detail. Figure 9a shows the characteristic DSC scans for the crystallization of PCL in a range of cooling rates (5–20 °C min−1). As expected, the crystallization peaks moved to lower temperatures with increasing cooling rate, while the variation of the peaks for PCL and nanocomposites is presented in Fig. 9b. It can be seen that the differences are quite small and the crystallization of the nanocomposites took place 2–3 °C higher than the matrix. Moreover, the functionalized nanofillers crystallized at slightly higher temperatures than their unfunctionalized counterparts as a result of enhanced interactions with the matrix and optimum dispersion, which are known to affect the crystallization procedure. The nucleating effect of nanofillers on PCL has been observed by several research groups [74,75,76,77,78].

Figure 9
figure 9

a DSC cooling curves of non-isothermal crystallization of PCL, b peak temperatures for non-isothermal crystallization of PCL and nanocomposites

Full size image
Nucleation Activity

The nucleation activity of the HNTs within the PCL was calculated with the theory proposed by Dobreva and Gutzow [79, 80]. In this method, the nucleation activity (φ) is a factor which describes the work of three-dimensional nucleation and it decreases with the addition of a filler particle within the matrix. If the filler is extremely active, then φ ≈ 0, while if it is inert then φ ≈ 1. The nucleation activity is calculated by the ratio:

$$ \phi = \frac{{B^{*} }}{B} $$
(9)

where the parameter Β can be obtained from the slope of Eq. 10, after plotting ln(β) versus the inverse squared degree of supercooling 1/ΔT 2p (ΔTp = Tm − Tp):

$$ \ln \beta = {\text{Const}} - \frac{B}{{\Delta T_{\text{p}}^{2} }} $$
(10)

For the case of heterogeneous nucleated by using a nucleating agent, Eq. 10 is transformed to the following:

$$ \ln \beta = {\text{Const}} - \frac{{B^{*} }}{{\Delta T_{\text{p}}^{2} }} $$
(11)

The plots of ln β versus 1/ΔT 2p can be seen in Fig. 10a for all the nanocomposites under study. Straight lines were obtained for every samples, while from the slope of the linear fits the values of Β and B* can be calculated. Then, the nucleation activity can be obtained from the B*/B ratio and the results are presented in Fig. 10b. As it was described earlier, the lower the B*/B ratio, the higher the nucleation activity, while values close to 1 indicate low nucleation activity. It is obvious from Fig. 10b that the unfunctionalized HNTs do not affect heavily the crystallization of PCL, being almost inert during the crystallization process (φ ≈ 0.95). On the other hand, the functionalized HNTs display higher nucleation activity with the sample filled with 1 wt% HNTs–APTES being more active, probably as a result of optimum dispersion and stronger interactions.

Figure 10
figure 10

a Dobreva plots for PCL and nanocomposites, b calculated nucleation activity of the nanocomposites

Full size image
Effective activation energy

The calculation of the effective activation energy can provide insights on the crystallization behavior of polymer nanocomposites [81]. A number of models have been proposed in literature, but the most used one is the one proposed by Kissinger [81]:

$$ \ln \left( {\frac{\beta }{{T_{P}^{2} }}} \right) = {\text{Const}} .- \frac{{\Delta {\rm E}}}{{RT_{\text{P}} }} $$
(12)

where β is the cooling rate, TP is the peak temperature and R is the universal gas constant. Vyazovkin and Sbirrazzuoli have shown that the negative values of the logarithm of the cooling rate β, which is present in the Kissinger equation, represent a mathematically invalid procedure; therefore, this theory cannot be used in non-isothermal crystallization from the melt [81]. Another issue of the method is that it assumes that the processes described are modeled using single-step kinetics; however, polymer crystallization is a quite complex procedure; thus, the use of multiple heating/cooling rates is recommended [82]. An isoconversional method is therefore suggested for the calculation of the effective activation energy during non-isothermal crystallization. The differential isoconversional method of Friedman [83] and the advanced integral method of Sbirrazzuoli and Vyazovkin [84] are the most popular ones and in this work, the method of Friedman was applied to the data from the dynamic crystallization of the nanocomposites. In this method, the activation energy takes different values for different degrees of crystallinity:

$$ \ln \left( {\frac{{{\text{d}}X}}{{{\text{d}}t}}} \right)_{X,i} = {\text{Const}} .- \frac{{\Delta E_{X} }}{{RT_{X,i} }} $$
(13)

where dX/dt is the instantaneous crystallization rate as a function of time at a given conversion X, ΔΕΧ is the effective activation energy at a given degree of conversion, TX,i is the set of temperatures related to a given conversion X at different cooling rates and the i subscript refers to different cooling rates that were used during the experimental procedure. The instantaneous crystallization rate is obtained by differentiating the X(t) function obtained from the integration of the experimentally measured crystallization, with respect to time. Furthermore, the values of dX/dt at a specific X are correlated to the corresponding crystallization temperature at this X(i.e., TX), by selecting appropriate degrees of crystallinity (i.e., from 5 to 95%). Then, by plotting the left-hand side of Eq. 13 with respect to 1/TX, a straight line must be obtained with a slope equal to ΔEX/R.

The results from the calculation of E for PCL and nanocomposites can be seen in Fig. 11. It should be noted that the negative E values for PCL and nanocomposites indicate that the crystallization of polymers is a spontaneous process and that the rate of crystallization increased with decreasing temperature [84] and this observation has been confirmed for several thermoplastics in the literature [5, 6, 47, 85,86,87]. For neat PCL, the activation energy is higher than that of the nanocomposites, while it can be seen that E keeps increasing with increasing degree of crystallinity, indicating that the phenomenon is more demanding energetically at higher Xc. Interestingly, the curves of the activation energy between the two sets of fillers display some differences between them. The activation energy of the nanocomposites with the functionalized filler present a sudden drop at around 50% crystallinity and from Xc ≈ 60% they start to increase again, while the E values do not change significantly with increasing filler content. It should be stated that the effective activation energy here has the meaning of the temperature coefficient of the crystallization rate [88]. This dependence of E on Xc for the PCL/HNTs–APTES samples is an indication of a complex crystallization mechanism, different than the neat polymer and the unfunctionalized nanocomposites.

Figure 11
figure 11

Effective activation energy as a function of relative degree of crystallinity for a PCL/HNT and b PCL/HNTs–APTES samples

Full size image
Polarizing light microscopy

Polarized light optical microscopy was utilized in order to observe the spherulitic morphology of PCL and nanocomposites during isothermal crystallization. From the images presented in Table 7, it can be seen that the nucleation density of the nanocomposite samples is much higher than the matrix, as a result of the fillers acting as heterogeneous nucleating agents during crystallization. Moreover, the size of the spherulites decreases significantly after the introduction of the functionalized filler, as a result of enhanced crystallization rates and homogeneous distribution of the filler.

Table 7 Spherulites of PCL and its nanocomposites grown at 42, 46 and 48 °C (magnification 40×)
Full size table
Thermogravimetric analysis

Thermal stability of the nanocomposites was studied by TGA, with the resulting curves presented in Fig. 12, and the characteristic temperatures of degradation presented in Table 8, as well as the % inorganic content at 600 °C. HNTs did not affect the thermal stability of PCL, and the recorded Tmax values were ranging from 427 to 431 °C, with insignificant variations. The free hydroxyls on the inner surface of HNTs and the subsequent moisture they can entrap prevent the otherwise thermally stable nanotubes from decelerating the thermal decomposition of PCL [60]. In the presence of HNTs–APTES, a small decrease in Tmax was observed, of about 5 °C for all filler contents. Organosilane-modified nanofillers cannot improve the thermal stability of polymers, probably due to the thermally unstable organic functional group of the silane. Functionalized clay minerals can catalyze polymer degradation due to the presence of Lewis acidic sites on their edges after the degradation of the modification agent [89, 90]. Additionally, the free amino groups of HNTs–APTES can cause aminolytic decomposition of PCL [91]. In all cases, degradation occurs in a single step, between 300 and 450 °C, suggesting no changes in the decomposition mechanism were caused by the presence of the nanofillers. Regarding the mass residue at 600 °C, the inorganic content of the nanocomposites was very close to the initial amounts of nanofillers used. The thermal stability of PCL can be either enhanced or reduced by HNTs. When increase in thermal stability was reported, it was attributed to the inherent excellent thermal stability of the nanotubes, and their action as mass and heat transport barriers [92, 93]. Decrease in thermal stability has also been reported, and was caused by the presence of water molecules inside the lumens of HNTs [94].

Figure 12
figure 12

TGA curves of PCL and its nanocomposites with HNTs and HNTs–APTES

Full size image
Table 8 Characteristic temperatures of thermal degradation and residue of PCL and its nanocomposites
Full size table

Conclusions

Nanocomposites of PCL with HNTs and functionalized HNTs–APTES were successfully synthesized with the in situ ring-opening polymerization method. Both nanofillers were found able to act as co-initiators, as evidenced by the decrease in the IV and \( \bar{M} \)n values of the nanocomposites. SEM micrographs indicated good dispersion of the fillers since there were no visible aggregates. The presence of the nanofillers enhanced slightly the crystallization rates of the matrix, with the functionalized nanotubes being more effective toward the nucleation of PCL, as a result of homogeneous dispersion and stronger interfacial interactions. From Lauritzen–Hoffman analysis, it was found that the fillers reduced the critical nucleus size needed for crystal growth and the free energy that acts against primary nucleation, while the folding of the macromolecular chains proceeds easier in the case of the PCL/HNTs–APTES samples. The thermal stability of PCL was not reduced by HNTs, but by HNTs–APTES which were found capable of catalyzing the polymer’s degradation. Solid residues values at 600 °C were very close to the initial amount of nanofillers added.