Microstructural evidence of the toughening mechanisms of polyurethane reinforced with halloysite nanotubes under high strain-rate tensile loading

In this study, we have investigated the relationship between the spherulitic morphology and the dynamic tensile response of polyurethane reinforced with Halloysite nanotubes (HNTs). The polyurethane prepolymer is partially silane end-capped and filled with only 0.8 wt.% of acid-treated Halloysite nanotubes. The resultant nanocomposite material presents a 35% higher spall strength compared to the neat polyurethane and 21% higher fracture toughness. We show evidence that the HNTs are not the toughening phase in the nanocomposite, but rather it is their influence on the resultant spherulitic structures which alters the polymer microstructure and leads to a tougher dynamic response. Microstructural characterization is performed via Scanning Electron Microscopy, Atomic Force Microscopy and Field Emission Scanning Electron Microscopy, and crystallinity examination via X-ray diffraction. The spherulitic structures present a brittle fracture character, while the interspherulitic regions are more ductile and show large deformation. The nanocomposite presents a finer and more rigid spherulitic structure, and a more energy dissipative fracture mechanism characterized by a rougher fracture surface with highly deformed interspherulitic regions.

www.nature.com/scientificreports/ HNT content of 0.8 wt.%. The partial prepolymer termination was performed in order to improve the interfacial compatibility between HNT and PU, and the selected weight fraction of HNT was low enough to maintain the transparency of the polymer as a thin film. Gas-gun spall testing results for a tensile strain rate magnitude of 10 4 s −1 showed an increase of 35% in dynamic tensile (spall) strength and 21% fracture toughness for the nanocomposite compared to the neat polyurethane while maintaining transparency. The present work is a continuation of that study. Here, the microstructural fracture mechanism of the neat PU and HNT-PU nanocomposite are explained through: (i) morphological analysis of the spherulite structure via Field Emission Scanning Electron Microscope (FESEM); (ii) failure mechanism and fracture kinetics investigation via Scanning Electron Microscope (SEM) of the spalled plane; (iii) spall plane evolution analysis via cryogenic breakage of spalled samples and SEM micrographs; (iv) crystallinity analysis via X-ray diffraction (XRD); and (v) surface topology of the spherulite/interspherulitic interface using Atomic Force Microscopy (AFM). Our characterizations show how the HNTs does not behave as a traditional toughening phase in a composite, but rather through its interaction with the PU's hard domains it can interfere with the resultant spherulitic morphology. This gives rise to a tougher PU microstructure with a finer distribution of smaller and more rigid spherulites surrounded by ductile interspherulitic regions. Further, we illustrate how and why this finer microstructure exhibits a tougher dynamic response and improved properties.

Results and discussions
The study of the microstructural fracture response was made through the analysis of recovered spalled samples, which were subjected to, and failed under, high-strain-rate tensile loads. Gas-gun spall experiments were conducted to measure the fracture toughness and spall strength of the polymeric samples. The schematic configuration of the target assembly can be seen on Fig. 1a. Spall testing provides a measure of the dynamic tensile strength of a specimen material. The geometry of an impactor and test specimen can be designed to generate a compression-expansion cycle that results in a planar dynamic tensile load within the specimen 28,29 . The wave dynamics that lead to this shock compression and subsequent isentropic expansion of the specimen material can be monitored through the motion of the rear free surface of the specimen (Fig. 1b).
Upon impact the test material achieves an equilibrium Hugoniot shock state at the peak velocity, after which the velocity decreases with the isentropic expansion of the material. The interaction of these expansion or rarefaction fans generates a tensile stress state within the target material. Figure 1b presents a simplified representation of the distance-time plot showing the wave propagations in a spall configuration, and the resultant tensile stress direction in the sample is illustrated on Fig. 1c. The spallation occurs if the tensile stress is high enough to initiate the process of nucleation growth, and coalescence of voids resulting in a failure plane (spall layer) within the   27 . From the spall plane fractographies (Fig. 2) of the recovered spalled samples it is possible to qualitatively identify a rougher fractured surface for the nanocomposites, with extensively plastic deformed regions. The progressively rougher fracture surfaces shown as one goes from Fig. 2a-c corresponds to progressively greater spall strength of the samples, going from the low strength neat PU in (a), the moderate strength nanocomposite without silane terminations in (b) and the strongest sample, the partially silane terminated nanocomposite, in (c). Greater roughness of the fracture morphology is evidence of a more energy dissipative failure mechanism. Spherulitic structures (quasi-spherical semicrystalline regions) can be seen clearly in the neat PU, and the visually flat fracture surface indicates a brittle fracture character of the spherulites.
Considering the domain segmented microstructure of PUs, we expect that the crystalline spherulitic structures are dispersed in a SS matrix 6 . Under the tensile strain rate magnitude of 10 4 s −1 experienced by the samples during the spall testing, the PU tends to present a glassy-like behaviour and to fail in a brittle manner 32 . Based on the fractography results, the higher rigidity of the spherulitic structure would lead to flat fracture, while the more ductile interfacial regions would be able to sustain deformation. The interfacial regions are mainly filled with an amorphous composition that was pushed out during the crystallization 33 . These amorphous interspherulitic areas present PU macromolecules that are predominantly constituted by soft segments 34 . As such, the deformation of these amorphous regions would delay the spallation failure by resisting the applied stress though elongation, and the fracture of highly deformed regions would eventually lead to spall plane formation. For brevity, we will refer to these highly deformed regions as macro-fibrils, acknowledging their morphological similarity to smaller fibrils commonly seen in the deformation of polymers. Figure 2d presents the flat fracture surface of the spherulite and a macro-fibril at the spalled surface of the neat PU. During the spallation event the macro-fibrils' response to the applied tensile stress increases the capacity of the polymer to plastically deform, and greater the density of macro-fibrils formed, the greater the fracture toughness. This observation is consistent with the progressively tougher response of the samples discussed earlier.
In this manner, the more energy dissipative fractured surface of the nanocomposite can be associated with the presence of smaller and more rigid spherulites, which confer a higher density of macro-fibrils formed after the brittle failure of the spherulitic structure. Analysis of the spherulite size showed that the nanocomposite presented an average spherulitic diameter of 182 ± 12 µm, whilst it was 274 ± 13 µm for the neat PU.  FESEM images (Fig. 4a,b) present the spherulitic structures of neat PU and nanocomposite. The lamellar crystal bundles can be identified radiating from the center of the spherulites 35 . Through the comparison of the spherulitic texture morphologies, we can see the presence of thicker lamellar crystal bundles for the nanocomposite. The difference in thickness of the lamellar crystal bundles is highlighted by arrows in Fig. 4. The morphology and crystallization kinetics of the spherulites are directly dependent on the thermal history experienced by the PU 36 . However, the neat PU and nanocomposite samples were submitted to the same thermal conditions, and yet a clear distinction between the lamellar morphologies is observed on the FESEM images. Considering that the content of hard segments within the nanocomposite was not significantly different than that in the neat.
PU (approximately 32 wt.%), and that the bonding of the HNTs to the hard segments has been shown to promote the micro-phase separation of the nanocomposite 37 , we expect the occurrence of spherulites with a higher concentration of crystallized hard segments in the nanocomposite. Indeed, DSC results from our previous study 27 showed that the melting point of the hard phase crystallites was approximately 15 °C higher in the nanocomposite compared to that of the neat PU.
The morphological structure of spherulites in thermoplastic PUs is such that the hard segment chains are arranged perpendicularly to the longitudinal direction of the lamellar crystal bundle structure 15 . The HNT nanotubes would likely be found in these lamellar crystal bundles, due to the bonding between the outer of surface hydroxyl groups of the HNT and the silanol groups in the silane terminations of the PU. Figure 5a presents evidence of HNTs embedded in the lamellar crystal bundles of the nanocomposite spherulitic structures. Figure 5b It is worth emphasizing that the spall failure of traditional fiber reinforced polymeric materials usually shows a decrease of the spall strength compared to the neat polymer, due to increased void nucleation at the weakly bonded filler/polymer interface during spallation 40 . Huneault et al. 41 reported the occurrence of carbon nanotube pullout at the spalled surfaces of epoxy reinforced with 1% wt. of carbon nanotubes. Their findings indicate that due to the weak adhesion at the filler/polymer interface, these regions would eventually act as nucleation sites for fracture resulting in a significant decrease in the spall strength. Lebar et al. 42,43 studied the filler/polymer  However, our results do not only show evidence of a good filler/polymer adhesion, they also indicate the ability of the HNT to be integrated into the final polymer microstructure. As shown in the FESEM images (Figs. 4b and 5a) we did not observe evidence of nanotube pullout, and the occurrence of different spherulitic morphologies suggests that the role of the HNT is to alter the final macromolecular organization of PU. This implies that HNTs in this system play a different role than a conventional reinforcement filler. Conventionally, the positive effects attributed to filler particles in a polymer are the action of the filler as a barrier to fracture growth via crack-front pinning and bowing, and fiber pull-out 44,45 . Here, we see that the HNTs affected the spherulitic structure development and therefore the final morphology of the polymer.
The failure mechanism of the spherulitic structure observed under high strain conditions differs from what is expected under quasi-static conditions for PU. In a quasi-static tensile testing configuration, one would expect the elongation of the soft domain matrix combined with the tilt of the hard segment lamellar domains. The spherulite structure would be deformed to an ellipsoidal shape, and upon stress release the spherical shape would be restored by retraction. Thus, the soft matrix presents an elastomeric behaviour with the hard domains acting as physical cross-link sites 7 . Under dynamic tensile loading, we have observed the brittle failure of the spherulites combined with the stretching of the interspherulitic regions. Similar to the response under quasi-static conditions, the ductility of the PU under dynamic conditions can be attributed to the soft domain stretching. However, the brittle fracture surface of the spherulites suggests that the dynamic failure of the hard segment lamellar domains is not accompanied by significant conformational motion.
The AFM topology images of the spherulite/interspherulitic interface of the unfractured samples is presented in Fig. 6. Comparing the height profiles starting from the lamellar crystal bundle and finishing in the amorphous interspherulitic matrix for both materials, the height profile for the neat PU displays a more abrupt transition between the spherulitic and amorphous regions. The vertical height difference (∆z) between hill and valley is around 900 nm for the neat PU, and 400 nm for the nanocomposite. This difference in the height profile may be related to a higher volumetric contraction of the coarse spherulitic structure in the neat PU during crystallization. This would consequently increase the void formation tendency along the boundaries and result in weaker and more brittle boundaries in the final microstructure 46 .
Furthermore, the bonding of the HNTs to the hard segments promotes the micro-phase separation of the nanocomposite, which results in a higher chain mobility of the soft segments 37 . As the interspherulitic regions of the PU are mainly formed by soft segments, the presence of a less restricted molecular motion can lead to a more ductile behaviour of these regions. Thus, the higher volume contraction, combined with a higher hard segment concentration in the interspherulitic regions, are factors that contribute to the more brittle behaviour experienced by the neat PU. In the fractography images (Fig. 7a,b) it is possible to observe a more brittle behaviour in the interspherulitic regions of the neat PU, and for the nanocomposite a more ductile behaviour with extensively deformed regions around the spherulite.
The crystallinity of the samples was studied via XRD (Fig. 7c). The neat PU and nanocomposite presented two broad diffraction peaks at approximately 20° and 45°, which can be attributed to soft segments in the PU 47 . These peaks present similar intensities on both samples but are slightly sharper for the nanocomposite, which may suggests a higher crystal regularity for the nanocomposite 48 . The peak at 29.5° was substantially higher for the nanocomposite, indicating the occurrence of more ordered crystallites, and the shift from 28.6° (neat) to 29.5° (nanocomposite) indicates a more tightly packed crystal structure in the nanocomposite. This is all evidence supporting the notion that the HNTs interact with the hard domains to help create a more ordered and stronger www.nature.com/scientificreports/ spherulitic structure. Finally, we note that the introduction of a low content of HNT did not introduce new peaks in the X-ray diffractograms, suggesting a good dispersion of the nanotubes 49 .

Conclusion
The higher values of fracture toughness and dynamic tensile strength of the nanocomposite can be mainly attributed to the following factors.
Finer, more rigid spherulites in the nanocomposite. The presence of smaller and more rigid spherulitic structure in the nanocomposite led to a higher fracture energy during the spallation process. The FESEM images showed the thicker lamellar crystal bundle structure of the nanocomposite with dispersed HNTs, while XRD results presented evidence of a more ordered and packed crystal structure, indicating a positive effect of the nanotubes on the hard phase crystallization. We believe that further studies investigating the influence of the HNTs in the nucleation and growth processes of the spherulitic structure are warranted.

Brittle interspherulitic regions in the neat PU. The neat PU fractographies present coarse spherulites
with some voids and cracks propagating in the interspherulitic regions. Considering the macro-fibril stretching of the soft segment matrix during tensile loading, the presence of brittle interspherulitic interfaces would accelerate the spall fracture. AFM height profiles indicate a greater height transition between the spherulite and interspherulitic regions for the neat PU, which could be related to a higher volume contraction of the coarse spherulites in the neat PU during crystallization. These results reinforce the possibility that the neat PU present weaker spherulitic boundaries, as the increase in the volume contraction can increase the possibility for void nucleation to occur along the interspherulitic regions. Also, taking into consideration the lower degree of microphase separation in the neat PU compared to the nanocomposite, the higher concentration of hard segments in the interspherulitic regions of the neat PU can interfere with the mobility of the soft segments and reduce the overall ductility of these regions.
Higher interspherulitic area in the nanocomposite. A higher density of macro-fibrils is formed for the nanocomposite, as can be seen in the fractographies. The presence of finer spherulites confers to the nanocomposite a higher density of interspherulitic boundaries. Taking into account that these regions are mainly constituted by a soft segment matrix and that these areas sustain the highest degree of plastic deformation dur- Sample preparation. During the synthesis of the nanocomposites, the HNTs were added into the liquid pre-polymer at a weight fraction of 0.8% and were dispersed via a sonication process. Next, the partial silane termination occurred for an aminosilane content equivalent to weight fraction of 0.6% of the pre-polymer weight. As a result of the partial termination process, monodentate urea linkages and trimethoxysilane terminations were created in the PU's prepolymer 50 . A small content of aminosilane was adopted to inhibit a substantial increase in the viscosity of the HNT/prepolymer solution. Additionally, the same curative was chosen for the neat PU and nanocomposite. The neat PU and nanocomposites were annealed for 22 h at 120 °C in a metallic mold. During the post-curing stage, the hydrolysis of the silane terminations can potentially produce silanol terminations in the prepolymer end groups, which can possibly react via condensation with surface hydroxyl groups of HNT 26 . Thus, the adhesion between the PU and nanotubes is promoted. The detailed synthesis and post-curing process can be found in our previous work 27 .
Spall testing. The testing was performed in a 64-mm smooth-bore single-stage light gas gun at the Impact Research Lab facility at Carleton University. Acrylic flyer plates (3.2 mm thick) were used to induce spall in the neat PU and nanocomposite, according to the plate rigidity and shock Hugoniot similarity requirements for plate impact configurations 41 . The resultant strain rate in the samples were controlled by varying target specimen thicknesses. The back-face velocity histories of the samples were measured with a two-channel photonic Doppler velocimeter (PDV) 51,52 . A summary of the experimental results can be seen in Table 1. The acoustic properties of the polymers were determined using an Olympus 45MG ultrasonic thickness gage coupled with a delay line transducer at a frequency of 10 MHz. The sound speeds were measured to be 1.98 km/s and 2.03 km/s for the neat PU and nanocomposite, respectively.
SEM and FESEM analysis. SEM fractography analysis of the recovered samples after the spall tests were recorded using a Tescan Vegal microscope. The HNTs were identified by using INCA Energy-dispersive X-ray spectroscopy analysis together with SEM. The morphology of the spherulitic structure was analyzed via FESEM imaging of the spall plane surface using a JSM-7500F FESEM (JEOL). The samples used for the SEM and FESEM were sputter coated with a 10 nm thick layer of gold. The average spherulitic diameters were found by using an image analysis program Fiji (ImageJ) based on the measurement of 30 spherulites for each material 53 . The standard deviations of the measurements were 12 μm for the neat PU and 13 μm for the nanocomposite.

AFM analysis.
Top surfaces (the sides of the sample open to air during polymerization and annealing) of neat PU and nanocomposites specimens were analyzed via AFM (Dimension 3100, Veeco Corp. Santa Barbara, CA, USA) in tapping mode with an aluminum-coated silicon tip (f 0 = 278 kHz, r < 10 nm) installed. The largest scan range of the instrument was 100 μm × 100 μm. The surface topology of the interface between the spherulitic structure and the interspherulitic region was analyzed using the height images and surface profiles 54 . The AFM results analysis was performed using the WSxM software 55 .
XRD examination. The analysis was performed with a Rigaku Ultima IV Diffractometer (CuKα source with λ = 1.54184 Å) with a scintillation counter and diffracted beam monochromator, using a scanning range (2θ) of 10°-90° with a 0.02° step size at a scan rate of 0.5°/min.