Figure 10: Plastic spacers allow direct access of chlorides to the reinforcement
Chapter 5 Microstructure of the spacer-concrete interface
5.2.3 Scanning electron microscopy
BSE images were collected using a Camscan Apollo 300 field emission SEM operated at 10 kV accelerating voltage and 10 mm working distance. Fifty images were collected per sample for those containing plastic and cementitious spacers. For samples with steel spacer, the number of images collected per sample was forty one and thirty six for those conditioned at 50°C and at 20°C, 55% RH respectively. The decrease in number of images captured is due to the small size of steel wire (5 mm) spacer which limits the available area for observation.
general viewing of the microstructure. The pixel spacing and field of view of each magnification used is tabulated in Table 5.2 . In order to ensure random and unbiased sampling, the microscope stage was moved at equally spaced distance along the interface of the spacer (around 1 mm increments) ensuring both bulk paste and spacer are imaged. To isolate the effect of aggregate particles on the microstructure, images were selected such that they were located at least 50 µm away from the nearest aggregate particle to avoid sampling the aggregate ITZ. If an image contains aggregate particle close to the spacer interface, the image will be replaced by another image within the surrounding interface. Also, areas near the sample edge were not imaged to avoid the probability of sampling saw- damaged areas.
Table 5.2 Pixel spacing and field of view of images digitised to 2560 x 2048 pixels, captured at various magnifications.
Magnification Field of view (µm) Pixel spacing (µm)
50x 2400 x 1920 0.940
100x 1200 x 960 0.470
500x 240 x 192 0.094
Acquiring high-quality images is vital for accurate segmentation of features and subsequent quantification steps. Brightness and contrast settings of the SEM can have a substantial effect on the appearance of the microstructure, specifically the apparent porosity. Therefore, the brightness and contrast settings should be calibrated prior to image capture to optimise the image brightness histogram and to ensure consistent results. An optimised brightness histogram fully utilises the entire dynamic range of available grey scale (0-255) where 0 represent black pixels and 255 represent white pixels. Image detail will be lost in the low and high regions of the grey scale if the contrast setting was too high, thus creating a distorted histogram with artificial peaks at both ends. Overly contrasted images cannot be improved by image processing, as this will only compress the histogram, but not remove the artificial peaks [Russ, 2002]. For this study, the brightness and contrast settings of the SEM were adjusted by trial and error to achieve the optimum BSE image and histogram as shown in Figure 5.3. The optimum settings were then applied to all subsequent images captured for the particular sample to ensure a faithful reproduction of grey values.
Figure 5.3 Example of BSE image and greyscale histogram obtained when the
brightness and contrast settings of the SEM are set at optimum. The sample is C10-28d-20°C, 50%RH-50:Co. Image was captured at 500x magnification giving a field of view of 240 x 198 µm.
5.2.4 Image analysis
Spacer and aggregate segmentation
A significant challenge for accurate quantitative microscopy is in the segmentation of the phases of interest, which for this study are the spacer, aggregates, pores and unreacted cement. However, automated detection and segmentation is often unreliable [Wong and Buenfeld, 2006a]. Therefore, in this thesis, a careful manual tracing operation was applied to
Capillary pores
Hydration products
Unreacted cement
measured as part of the original pore structure for the purpose of this study. The procedure requires about 2 minutes per image to perform, and is shown in Figure 5.4 and Figure 5.5.
Although relatively time consuming, it is necessary for accuracy and statistical significance.
Once the spacer and aggregate boundaries are accurately marked, subsequent image operations can be automated as described in the following section.
(a) (b)
(c) (d)
Figure 5.4 Segmentation of cementitious spacer: (a) original BSE image at 500x magnification; (b) spacer boundary is manually traced with a white line; (c)
thresholding, particle detection and hole filling to produce spacer binary mask; (d) final image with the spacer removed (field of view: 240 x 192 µm, sample is C10-28d-20°C, 55% RH-50: C1).
Cementitious spacer Cementitious spacer
Cementitious spacer Cementitious spacer
(a) (b)
(c) (d)
Figure 5.5 Segmentation of plastic spacer: (a) original BSE image at 500x magnification; (b) spacer boundary is manually traced with a white line; (c)
thresholding, particle detection and hole filling to produce spacer binary mask; (d) final image with the spacer removed (field of view: 240 x 192 µm, sample is C1028d50°C -50: P1).
Pore segmentation
The most important characteristic of the microstructure that is to be investigated is the porosity. Herein, the term porosity includes the capillary pores and microcracks. This can be measured using image analysis by carrying out a greyscale thresholding to segment the pores and cracks. For thresholding, the lower threshold value can be set to zero because the pores and cracks are the darkest phase in a BSE image. However, the determination of the upper
Bond crack due to sample preparation artefact
Plastic spacer Plastic spacer
BSE image. This represents a critical point where a small increment in the threshold value will cause a sudden increase in the segmented area, indicating that the pore-solid boundary is reached. The inflection point is obtained from the intersection of two best-fit lines in the cumulative brightness histogram as shown in Figure 5.6 (a, c, e). It can be seen that the obtained upper threshold values varied slightly from one image to another. This is because the overflow method compensates for small and unavoidable fluctuations in beam conditions, sample surface roughness and brightness/contrast settings during image capture. As such, the method provides a consistent and reliable means of segmenting porosity.
(a) Cumulative grey scale histogram for C10-28d-50°C-50:P1
(b) Pores segmented at threshold level 108. Porosity of paste = 43.7%
(c) Cumulative grey scale histogram for C10-28-50°C-50:C1
(d) Pores segmented at threshold level 114. Porosity of paste = 19.9%
0
(e) Cumulative grey scale histogram for C10-28-20°C, 55%RH-50:S1
(f) Pores segmented at threshold level 107. Porosity of paste = 13.4%
Figure 5.6 Application of the overflow method to determine the upper threshold level for pore segmentation.
Unreacted cement segmentation
In a BSE image, the unreacted cement particles appear the brightest and are highly contrasted from other phases that exist in the microstructure. As such, the brightness histogram tends to show a peak at the far right end that represents the unreacted cement. Therefore, segmentation of the unreacted cement can be carried out by selecting the minimum grey value between the peaks for hydration products and unreacted cement as the lower threshold value (Figure 5.7). The exact location of the minima is determined from the first derivative of the brightness histogram. The upper threshold level for unreacted cement is set at 255.