5.6 Frequency sweeps
6.2.1 Characterisation of CNT dispersion
6.2.1.1 Section preparation methods
Rectangular bar specimens of PC-MWCNT 2205 with dimensions 60 mm × 5 mm × 2 mm were manufactured by compression moulding at 250◦C. Thin sections of the specimens were prepared using the microtome to obtain sections with thickness of 2 µm and 4 µm. The rectangular specimen was cut to a length of 30 mm and embedded in a circular mould filled with EpoFix epoxy and left to cure overnight at room temperature.
Microtoming was performed with two instruments. The first instrument was a Reichert Jung 2050 SuperCut rotary microtome using a tungsten carbide knife. The cutting speed was fixed at 6 mm s−1 and the cutting surface was perpendicular to the length of the specimen. Slices were placed on microscope slides and manually uncurled using a soft brush. Microtoming was also performed using an MC PowerTome employing a glass knife. The block face of the
cutting surface (similar to the rotary microtome) was trimmed prior to sectioning. A boat containing distilled water, attached to the cutting edge of the glass knife, is used to float the sections away from the knife edge and to collect the specimens. The sections are transferred to a microscope slide using a tool consisting of an eyelash and placed on a gently heated stage (∼50 ◦C) to evaporate the water. Whilst heating, the eyelash tool is used to uncurl the specimens. Although attempts were made to uncoil the curled sections, the section’s edge that made the initial contact with the knife’s cutting edge remained curled. Optical investigations of such areas were omitted.
6.2.1.2 OM investigation
OM was conducted using a Olympus BX-51 microscope in transmission with a 10× lens. A digital camera, Q Imaging Fast 1396, was used to record the images. The Q Capture software was used to control exposure time. Light intensity was fixed for all images. Each micrograph represents an area of ∼1.26 mm2.
6.2.1.3 OM micrographs of thin sections
As PC is translucent in thin sections, it will appear bright, providing an obvious contrast to the CNT agglomerates that appear black. For the purpose of this technique, only CNT agglomerates above ∼1 µm in diameter are detectable due to resolution limit of OM.
Fig. 6.1 shows representative optical micrographs of 2 µm thick sections obtained using the tungsten carbide knife. A wide variety of defects are visible. The bright areas suggest tearing of the matrix, and the hazy or blurry parts are attributed to a non-uniform surface.
The tungsten carbide knife was considered unsuitable for the preparation of slices as it was difficult to differentiate between defects and CNT agglomerates.
Fig. 6.2 shows optical micrographs of 2 µm and 4 µm sections obtained using a glass knife.
Knife scratches are the dominant defect and the surfaces are still not perfectly planar. Com-paring Fig. 6.2 with Fig. 6.1 there is, however, an improvement in the quality of sections.
Hence, sections obtained with the glass knife are used for further analysis in this work.
(a) (b)
Figure 6.1: Optical micrographs of PC-MWCNT 2205 sections of thickness 2 µm, obtained with a tungsten carbide knife
(a) (b)
Figure 6.2: Optical micrographs of PC-MWCNT 2205 sections of thickness (a) 2 µm and (b) 4 µm, obtained with a glass knife
6.2.1.4 Attempts to produce an improved specimen surface
A procedure was attempted to improve the surface. A slice is sandwiched between two slide covers and placed on a heated Linkam stage that was mounted on a microscope. A top cover enclosed the device to reduce the loss of heat from the stage to the surrounding environment. A window seal on the top cover allowed real-time observation of the heat treatment of the specimen. The sandwich assembly was heated to 170 ◦C (above Tg) at a rate of 30◦C min−1. The temperature was selected to allow the PC to flow while minimising possible re-agglomeration of CNTs. At 160◦C, a small weight of 5 g was placed on the slide covers and the specimen was observed over 15 minutes.
Fig. 6.3 illustrates micrographs of microtomed 2 µm sections of PC-MWCNT 2205 at various times during the heat treatment. The images are blurred because of the uneven surface and due to the temperature of the air in the device. Striations formed on the sections as seen in Fig. 6.3(b) - (d), hence the procedure was found to be unsuitable to produce planar sections.
(a) (b)
(c) (d)
Figure 6.3: Optical micrographs showing the heat treatment of a 2 µm section of PC-MWCNT 2205 at 10x magnification. At 170◦C, a weight of 5 g was placed on the section and the images were obtained at several intervals showing (a) the specimen just after application of the load (t = 0). The formation of striations appearing at (b) t = 3 minutes, and remaining
after (c) t = 6 minutes and (d) t = 10 minutes
6.2.1.5 Image analysis
The optical micrographs were analysed with image processing software, ImageJ (Rasband, 1997). An image processing procedure for CNT agglomeration from OM micrographs was
developed by Pegel et al. (2011) that consisted of the following steps: background correc-tion, binarising the image and identifying an appropriate threshold to quantify the CNT agglomerates. Both Pegel and Kasaliwal successfully applied the procedure to analyse low concentrated PC-MWCNT (1 wt%) and obtained estimates of CNT agglomeration areas.
This method is employed in this work to determine the degree of dispersion using OM DOM that is expressed as (Kasaliwal et al., 2009)
DOM= 1 −
f ACNT
Ao
VCNT (6.1)
where f is the packing density of CNTs (f = 0.25 (Kasaliwal et al., 2009)), ACNT is the agglomerate area, Ao is the total observed area and VCNT is the CNT volume fraction in percentage.
6.2.1.6 Image processing procedure
The image processing procedure used in this work is based on Pegel et al. (2011). The first step is background correction of an OM image (converted to an 8-bit image) to reduce the effects of illumination gradient, typically found on micrographs. A mean filter is applied to a duplicate of the original image using a radius (in pixels) corresponding to the largest agglomerate in the image. This smooths the duplicated image by substituting each pixel with the surrounding mean. The filtered image is then subtracted from the original image to remove the brighter pixels in the original image. The same mean filter is applied to the new image and divided by the original image to remove any illumination gradient.
The corrected image is then binarised in the second step. Pegel and co-workers’ (2011) de-veloped a feature-based algorithm that determines the relationship between the threshold value and the quantity of objects, yielding a histogram. This was pursued because the grey-scale based algorithms, typically used to determined threshold values, were not sufficient to identify small CNT agglomerates due to the weak distinction of the agglomerates from the background (Pegel et al., 2011). Therefore, the threshold value for binarisation is obtained using the triangle algorithm, whereby a line is constructed between the minimum and max-imum value on the histogram (Wu et al., 2008). The furthest normal distance between this line and the histogram identifies the threshold value.
Lastly, a morphological filter is applied to the binarised image to perform an erosion process followed by a dilation process. An erosion process removes the pixels from the edges of a black object, each pixel in the image with the smallest pixel value in that pixels neighborhood, employing a 3×3 square structuring element. The dilation process is the opposite of the erosion, replacing of each pixel in the image with the highest pixel value found surrounding it.
The open morphology operation helps to remove isolated pixels and smooths the agglomerate boundaries.
6.2.1.7 Quantification of CNT dispersion
Figs. 6.4(a) and (b) illustrate the original image of a PC-MWCNT (3 wt%) section and the filtered image respectively. The feature based algorithm applied to Fig. 6.4(b) produced the histogram, shown in Fig. 6.5, yielding the grey scale threshold value of 116.
Pegel’s binarisation method was found to be unsuitable for the current work as the feature-based algorithm picked up the less circular shapes associated to the dominant scratch defects, shown in Fig. 6.4(c). The agglomerate area fraction, Aag for Fig. 6.4(c), with the defects visible, is 4.9% for a 2µm thick section. When visible defects were excluded by selecting regions of interest for the quantification of CNT dispersion levels, Aag is estimated to be 4.6%. A higher value of Aagindicates a lower level of CNT dispersion. There is a relationship between the section thickness and Aag, hence the CNT agglomerate volume fraction can be obtained by determining Aag for of range section thickness. Due to the technical difficulties and user dependence of separating defects from CNT agglomerates in the OM micrographs, no further evaluations using this method was performed to quantify CNT dispersions of these systems.
Since ACNT could not be evaluated, DOM cannot be determined using the current micro-graphs. It should be noted that Pegel and co-workers highlighted that the error for DOM increases with high CNT agglomerate volume fractions. Furthermore, the uncertainty of the appropriate packing density f value for CNT agglomerates contributes to this error. Hence, obtaining a reliable estimate of agglomerate volume fraction for percolated PC-MWCNT sys-tems is not trivial. Although a quantitative measure could not be achieved for PC-MWCNT (3 wt%), the image can be qualitatively inferred as having poor dispersion due to the high density of CNT agglomerates visible.
(a) (b)
(c)
Figure 6.4: The background correction procedure applied on the (a) original image of PC-MWCNT (3 wt%) with thickness of 2 µm to obtain the (b) new filtered image with minimised illumination gradient. A feature based algorithm (Pegel et al., 2011) is applied to (b) to yield the (c) binarised image with artifacts still visible as indicated by the red ellipses
In the present case, image processing and analysis fell short owing to damage during specimen section preparation. There are considerable practical challenges in producing thin, flat and artifact-free specimen sections of saturated PC-MWCNT for microscopy. The technique requires a high level of skill and a reasonable number of sections, in order to represent the bulk structure of the specimens. OM sample preparation technique is destructive and laborious, and highlights the difficulties of implementing OM in commercial processes or as a service-life inspection of CNT filled thermoplastics.
There is a demand for non-invasive optical and imaging processing methods to characterise
Figure 6.5: An illustration of the triangle algorithm, described in Section 6.2.1.6, to de-termine the threshold value for Fig.6.4(c)
CNT dispersion in the matrix. Recently, techniques based on optical diffraction and diffu-sion mechanisms such as optical coherence tomography, oscillatory photon correlation spec-troscopy and Fraunhofer wavefront correlation have been developed and applied on nanoclay and MWCNT nanocomposites (Broughton et al., 2013). They developed these methods to be used in production lines and as an inspection tool during nanocomposites’ product lifetimes.