1 Introduction

Polymer nanocomposites (PNC) are widely researched due to their excellent mechanical properties and rich potential for multifunctionality, with emerging and potential applications in a range of industrial sectors [1,2,3,4]. The enhancement of properties is strongly dependent on the nature of nanofillers and polymer matrices, as well as on the internal structure, i.e., dispersion and orientation, that arises from their processing history.

Elongational flow during the preparation of PNC fibres can generate higher stiffness and strength as compared to their isotropic counterparts by aligning nano-reinforcements and polymer chains along the fibre axis [5, 6]. The elongation not only facilitates the orientation of the filler and matrix that enhances the stress transfer efficiency, but can also improve the dispersion of nanofiller, and alter the crystallisation rate and the crystalline structure of polymer matrix [7, 8]. Further, nanoreinforcements with high aspect ratios are prone to the orientation effect induced by drawing during the manufacture of filaments, which is widely applied as to control modulus for unfilled polymer filaments [6].

Graphene nanoplatelets (GNP) are a widely investigated nanoreinforcement for polymers due to their combination of outstanding structural and functional properties [9,10,11,12]. Unlike carbon nanotubes (CNT), which are essentially 1-dimensional, the orientation of 2D plate-like GNP within fibres is more complicated as it has two plane axes to take into account relative to the fibre axis, resulting in variations in the orientation of GNP in different regions of fibres, which in turn influence the mechanical properties of GNP/polymer fibres; similarly, clustering of GNP resulting in local concentration differences is expected to result in corresponding local variability in mechanical properties. However, few studies have investigated the effect of orientation and dispersion of GNP on the properties of GNP reinforced PNC fibres. Nilsson et al. reported a preferential tangential orientation of GNP to the nearest filament surface in the outer layer of undrawn melt spun polypropylene (PP) strands, while in the core region, orientation was much less significant [13]. Such tangential orientation has been observed also in melt spun PP fibres containing organoclay nanoplatelets [14]. Moreover, the organoclay in turn promoted significant alignment of PP chains, even in fibres obtained with no drawing. This was attributed to the amplification or shearing effect exerted on polymer chains and reduced relaxation of oriented chains by adjacent organoclays [14]. However, the authors did not relate these structures to the fibres’ mechanical properties, which is a primary objective of the study presented here.

In present work, polyhydroxy ether bisphenol A (phenoxy) is used as the polymer matrix because thermoplastic phenoxy is a promising toughening agent for epoxy resins due to its excellent toughness and soluble nature in epoxies [15,16,17,18]. By combining both GNP and phenoxy, GNP/phenoxy fibres can potentially be used to deliver GNP into epoxy-based composites and simultaneously toughen the system to achieve complementary reinforcing effect. We will report on an investigation into the efficacy of this toughening strategy separately, and here focus our analysis on the influence of processing on the structure and mechanical properties of such melt-spun filaments in their own right.

2 Experimental Programme

Poly(hydroxy ether of bisphenol A) (phenoxy) used in this study was supplied by InChemRez® (PKHB) with an average molecular weight (Mw) of 32,000 g/mol, a glass transition temperature (Tg) of 84 °C (DSC) and a melt flow index (MFI) of 60 g/10 min at 200 °C. The nanofiller used was graphene nanoplatelets (GNP) grade M5 from XGSciences® USA, with an average diameter of 5 μm, an average thickness of around 6–8 nm, and a typical surface area of 120–150 m2/g.

Phenoxy based nanocomposites with a range of GNP contents between 5 and 20 wt% were prepared via a masterbatch dilution method with an aim of improving the dispersion of GNP and reducing the exposure time of operators to GNP powders. Prior to compounding, GNP and phenoxy were dried for 18 h at 80 °C and 75 °C, respectively.

GNP and phenoxy were melt-compounded using a Polylab PTW16 co-rotating twin-screw extruder (Thermo-Haake, Germany) at a temperature of 180 °C and a screw speed of 100 rpm to produce a 15 wt% masterbatch. A family of nanocomposites with GNP content less than 15 wt% were then prepared by diluting the masterbatch with pure phenoxy; a 20 wt% GNP/phenoxy nanocomposite was produced by further compounding the masterbatch with pristine GNP powders. Finally, all samples were processed one more time on the twin-screw extruder to improve the dispersion of GNP in the phenoxy matrix.

Pure phenoxy and GNP/phenoxy nanocomposite filaments were melt spun using a Bicomponent Fibers Extrusion Machine (LBS-100, Hills Incorporation, USA) equipped with a 1 mm monofilament die at a screw speed of 30 rpm and a barrel temperature ranging from 190 to 210 °C depending on the GNP content. The extrudate was then air quenched and hot drawn at varying take-up speeds ranging from 50 to 300 m/min (corresponding to a draw down ratio (DDR) from 10 to 60) before being collected on bobbins.

Scanning electron microscopy (SEM) was carried out on filaments to investigate the dispersion and orientation of GNP in phenoxy matrix. Filaments were cryo-fractured in liquid nitrogen and the cross sections were coated with platinum using a 108 auto sputter coater (Cressington Scientific Instruments, UK). Microscopy was performed on an FEI Quanta 200 (FEI Company, USA) scanning electron microscope at an accelerating voltage of 20 kV.

Transmission electron microscopy (TEM) of ultrathin sections was carried out using an FEI Tecnai G2 20 (FEI Company, USA). Ultrathin sections with a thickness of around 100 nm were prepared using a Leica UC6 ultramicrotome (Leica Microsystems, Germany) at ambient temperature and the specimens were then collected on a 300-mesh copper grid for observation.

Tensile tests of fibres were conducted on an Instron 3344 universal testing machine (Instron, USA) equipped with a 10 N load cell. Pure phenoxy and nanocomposite fibres were mounted on a relatively stiff paperboard frame using super glue. The gauge length was 20 mm, and the crosshead speed was set at 1 mm/min. Young’s modulus was calculated in the strain range of 0.05 − 0.25%. A minimum of 5 specimens were tested for each sample.

2.1 Structural Characterisation

Fig. 1
figure 1

SEM micrographs of (a) overview of the cross section of 5 wt% fibres at DDR = 10 and local magnifications at regions close to (b) fibre edge and (c) fibre core

Full size image

Figure 1 provides characteristic SEM images of the cross section of melt spun fibres containing 5 wt% GNP and clearly shows their orientation. Examples for other filler contents are provided in Fig. S1 in the Supplementary Information. The images confirm that the masterbatch dilution method has yielded good GNP dispersion, resulting in mostly thin GNP flakes in phenoxy as can be seen from Fig. 1 (b). However, as might be expected, some thick GNP stacks with significantly greater lateral dimension than the 5 μm average remain, probably due to the strong π-π stacking between GNP flakes. A typical example of such stack is shown in Fig. 1 (b) as indicated by the red arrow. From Fig. 1 (b) and (c), it can be seen that GNP flakes tend to align with their plane broadly parallel to the fibre axis, resulting in the flakes appearing as 1D lines when viewed from the transverse direction of fibres. However, significant differences in GNP orientation can be observed also in regions close to the fibre core and the edge. The normals of GNP flakes near the fibre edges are essentially perpendicular to both the fibre axis and the tangential direction of the nearest fibre surface as shown in Fig. 1 (b). This alignment reduces gradually from the edge to the core region, such that eventually the normals of GNP flakes are at any angle relative to the tangent of the nearest fibre surface, while remaining perpendicular to the fibre axis in areas close to the fibre centre, giving an essentially random nanoplatelet orientation in this region (Fig. 1 (c)). Our observations are consistent with those discussed earlier for melt spun polymer fibre systems filled with GNP and organoclay [13, 14].

Fig. 2
figure 2

TEM images of the axial sections of 5 wt% GNP/phenoxy fibres: (a) edge and (b) core region of fibres with a DDR = 10; (c) edge and (d) core region of fibres with a DDR = 60

Full size image

Given these observed orientations, it may be anticipated that “face-on” GNP is more likely to be observed in the core region of the fibre when viewed from the longitudinal direction. Importantly, this does not imply that most GNP in this region will be “face-on”. TEM images of the axial section of 5 wt% GNP/phenoxy nanocomposite fibres are presented in Fig. 2 and these show tangential orientation of GNP at the fibre edge region (Fig. 2 (a) and (c)), which is consistent with the SEM observations (Fig. 1 (b)). “Face-on” GNP is more obvious in regions close to the fibre core (Fig. 2 (b) and (d)), i.e., fewer 1D white lines are observed in the core compared to the edge region. Another feature observable in Fig. 2 is that considerable orientation of GNP is already observable at the lowest DDR (10) and is not qualitatively different from that observed at higher DDR; this is consistent with the observations of Wang et al. [6]. In what follows, we quantify the increased orientation as DDR is increased to 60.

2.2 Mechanical Properties

The mean Young’s modulus of the fibres prepared is plotted against GNP content in Fig. 3; in this and subsequent figures, error bars represent 95% confidence intervals on the mean. The weight fraction was converted to volume fraction using

$$V = \frac{{{\rho _m}\,{\omega _f}}}{{{\rho _m}\,{\omega _f} + {\rho _f}\,{\omega _m}}}$$
(1)

where V is the volume fraction of GNP, ρf is the density of GNP (g/cm3), ρm is the density of the matrix (g/cm3), and ωf and ωm are the weight fractions of GNP and phenoxy matrix, respectively. We assumed that ρf = 2.2 g/cm3 and that for our phenoxy matrix, ρm = 1.18 g/cm3. As expected, Young’s modulus of fibres increases with both the GNP content and the DDR. The former is consistent with the results of others [19, 20] and we note that Young’s modulus is a low-strain property averaged over the full gauge length of the sample: so, whilst we expect some stochastic distribution of local levels of nanoparticle concentration along a filament, any localised increase in stiffness in regions of higher nanoparticle concentration will be counterbalanced by the lower stiffness of regions with lower nanoparticle concentration. Accordingly, we do not expect modulus to be greatly influenced at moderate degrees of nanoparticle clustering.

Fig. 3
figure 3

Fitting of Young’s modulus vs. GNP volume fraction data to Eq. (2)

Full size image

The lines passing through the data arise from a least-squares fit of the simple rule of mixtures (RoM) [21]:

$${E_c} = (1 - {V_f})\,{E_m} + {V_f}\,{E_f}$$
(2)

,

where Ec and Em are the Young’s modulus (GPa) of the nanocomposite and the matrix, respectively; Ef and Vf are the effective modulus and the volume fraction of the GNP nanofiller, respectively. It is important to note that the effective modulus, Ef, depends on the Young’s modulus of the nanofiller, Ey, f and their orientation, as captured by the Krenchel orientation factor, h0 and a length factor, hl, such that

$${E_f} = {\eta _0}{\eta _l}{E_{y,f}}$$
(3)

where h0 ≤ 1 and hl ≤ 1 and therefore Ef ≤ Ef, y. Further, and somewhat counter-intuitively, it has been reported that Ef is strongly dependent on Em in graphene reinforced PNCs, Ef being higher when the polymer matrix is more rigid [22]; we interpret this as being a consequence of the efficiency of stress transfer between the matrix and the particles being dependent of the stiffness of the matrix. As such, Ef for GNP in relatively stiffer polymers can be a few orders of magnitude higher than that in thermoplastic elastomers [23,24,25].

Table 1 Best fit parameters from regression on data for Fig. 3 including 95% confidence intervals
Full size table

The fitting parameters for the lines in Fig. 3 including 95% confidence intervals are provided in Table 1. We note that the values for Ef obtained here are broadly similar to those reported for GNP/PP and GNP/epoxy resin systems (30–60 GPa) [12] and, as expected, fitted values of Em are essentially constant.

Young et al. [12]. combined a simple rule of mixtures (cf. Equation 2) with shear-lag theory to yield

$${E_f} \approx \,\frac{{{\eta _0}\,{s^2}\,t}}{{12\,(1 + \nu )\,T}}\,{E_m},$$
(4)

,

where the Krenchel orientation factor, η0 is 1 for perfectly aligned GNP and 8/15 for 3D randomly oriented nanoplatelets [26, 27]; s and t are the aspect ratio and thickness (m) of the nanofiller, respectively, and T is the thickness (m) of a layer of polymer matrix that is assumed to surround the nanofiller; ν is the Poisson’s ratio of phenoxy. On manipulation, substitution of Eq. (4) in Eq. (2) yields,

$${E_c} \approx \left( {(1 - {V_f}) + \,\frac{{{\eta _0}\,{s^2}\,t}}{{12\,(1 + \nu )\,T}}\,{V_f}} \right){E_m}$$
(5)

.

As discussed, the GNP in the outer layer of the fibre exhibits higher orientation than that in the core region, with the plane of GNP at the fibre edge being approximately parallel to the nearest fibre surface (tangential orientation). Accordingly, we expect the local orientation factor, η0, to differ in these two regions, such that the values reported in Table 1 are the weighted average of these local values over the cross-sectional area of the filament. On this basis, we used a semi-quantitative approach guided by simple image processing to estimate the area of the fibre sheath layer and the entire fibre from SEM images of the fibre’s cross section; example micrographs of 5 wt% GNP/phenoxy nanocomposite fibres with different DDRs are presented in Fig. 4. These show an approximate boundary between the sheath and core layer of the fibre, indicated by a yellow dashed line, which we used to calculate the area of the two regions and hence compute an area ratio, AR, representing the fraction of the fibre cross-section that is sheath.

Fig. 4
figure 4

SEM micrographs of the cross section of 5 wt% GNP/phenoxy nanocomposite fibres with a DDR of (a) 10, (b) 20, (c) 30, (d) 40 and (e) 60 for the calculation of the sheath/total areal ratio

Full size image

Figure 5 shows the values of Ef obtained from linear regression on the data in Fig. 3, as given in Table 1, plotted against the area ratio obtained from micrographs. We observe a linear dependence of Ef on the area ratio (r2 = 0.899), as would be expected from Eq. 2 if the area ratio is proportional to η0.

Fig. 5
figure 5

Ef plotted against Area Ratio

Full size image

To investigate this further, we estimated η0 for our most oriented fibre (DDR = 60; AR = 0.87) by applying a simple rule of mixtures to the Krenchel factor such that

$$\begin{array}{c}{\eta _0} = \left( {1 - AR} \right){\eta ^{random}} + AR\,{\eta ^{aligned}}\\= \left( {1 - AR} \right)\frac{8}{{15}} + AR\end{array}$$
(6)

This yields η0 = 0.94 for filaments produced at 300 m/min. Substituting this value into Eq. (4) and assuming the aspect ratio, s = 740 (mean particle size of the GNP used assumed to be 5.2 μm [12], cf. 5 μm specified by the supplier [28]; thickness specified in data sheet as 6–8 nm [28]) and Poisson’s ratio ν = 0.38 [29], we obtain an estimate of the dimensionless ratio t/T as 5.4 × 10− 4, which is consistent with the literature [12].

Subsequently, we assumed that this parameter (t/T) was constant for all samples, allowing us to estimate the value of η0 for all other samples the values of Ef given in Table 1.

Fig. 6
figure 6

The calculated η0 as a function of Area Ratio

Full size image

The obtained η0 and Area Ratio are compared in Fig. 6. The red line represents a linear regression on the data (r2 = 0.837). Given that our treatment neglects the inevitable transition region between the highly aligned GNP in the sheath and its random orientation, this clear correlation between η0 and Area Ratio is reassuring and indicates a strong dependence of Young’s modulus on the internal structure of fibres.

3 Conclusions

We have presented quantitative analysis of the cross-sectional profile of the oriented structure of GNP in polymer nanocomposite fibres prepared via melt spinning and demonstrated good correlation with the mechanical property of these fibres. The morphology of GNP/phenoxy fibres where the orientation of GNP in the outer layer (tangential orientation) of the fibre is more regular than that in the fibre core is confirmed by both SEM and TEM. The Young’s modulus of fibres increases with the DDR and theoretical analysis on the modulus data using a simple RoM suggests an increased Ef of GNP, which is also found to be linearly dependent on the fibre sheath area ratio obtained from SEM micrographs. This relationship has been further investigated to reveal via an analysis combining both simple RoM and shear-lag theory that the area ratio is proportional to η0. This finding indicates a variation in the reinforcing efficiency of GNP in the fibre core and sheath layer, such that the increased area ratio is mainly responsible for the increased Ef of GNP and thus Ec of nanocomposite fibres. We consider that these findings should be widely applicable to other 2D nanoplatelet reinforced PNC fibre systems.