There have been significant developments in recent years in the application of SEM and X-ray imaging to the characterization of cement-based materials. Because the flux of backscattered electrons (BSE) produced by the incident electron beam in the SEM is proportional to the average atomic number of a phase, in a BSE image of unhydrated cement particles, the C 4AF phase, having the highest average atomic number, shows up as the brightest phase and is easily distinguishable from the other phases. Unfortunately, the other major phases, although exhibiting some contrast differences (i.e., the C3S being brighter than the others), cannot be reliably distinguished based solely on a BSE image. However, the X-rays emitted when the incoming electron beam interacts with the specimen surface provide valuable chemical information on the underlying phases, to supplement the BSE image. Typically, X-ray images are collected either as a binary dot map image (where all pixels exhibiting an X-ray intensity in a specific energy window that is greater than a preset threshold are set to a value of 255, for example) or as a continuous-valued X-ray image in which each pixel is assigned a 0-255 value based on the X-rays counts within some preset energy window. Using the X-ray techniques, resolution is typically limited to ~1 µm, which is smaller than the size of nearly all cement particles and fortuitously provides a scale identical to that commonly used in the NIST cement hydration model. Several examples of applying these combined SEM techniques to cement-based materials can be found in recent literature. Scrivener 22 has collected X-ray dot map images for silicon and aluminum for several cements and processed them to determine the proportions of silicates and interstitial (aluminate) phases present overall and present on the surfaces of the cement particles. These results have been applied to interpreting the hydration characteristics of these cements as quantified by calorimetric heat-release measurements. Stutzman 23 has collected X-ray map images for calcium, silicon, aluminum, and iron for cement clinkers (prior to grinding) and computed the area fractions of each of the four major clinker phases. For example, image areas containing calcium and aluminum, but no iron, were assigned to be the C3A phase. Bonen and Diamond24 have quantified the effects of grinding techniques on cement particle size, shape, and phase distribution, using SEM and X-ray analysis to classify the predominant phase found in each cement particle. By applying these techniques to each pixel in an image, as opposed to each particle, Bentz and Stutzman25have been able to map each pixel to its component phase. To perform this analysis, the cement of interest is first dispersed in a low-viscosity epoxy, which is subsequently cured. A polished surface is then prepared and viewed by SEM. For this application, in addition to the BSE image, X-ray images are collected for calcium, silicon, aluminum, iron, and sulfur. The uniqueness of this mapping is shown in a false-color image in which the calcium, silicon, and aluminum X-ray signals are assigned to the red, green, and blue color channels, respectively, as shown in Fig. B1 for Cement 115. The figure clearly shows the presence of two levels of yellow/ orange, distinguishing the C 3S phase from the C2 S phase, due to a different calcium/silicon (red/green) ratio. Red areas indicate the presence of either gypsum or free lime (CaO), with the sulfur X-ray image being used to distinguish between the two. Magenta areas indicate the presence of both calcium and aluminum, with the BSE or iron X-ray images then being used to distinguish C4AF from C3A. Although, in the current study, no attempt is made to distinguish the alkali sulfates from the gypsum, X-ray images of potassium and sodium could be collected easily and used for this purpose.
Because of the inherent noise in the X-ray images, after an initial segmentation is performed, a type of median filtering is applied to smooth the image. Here, each nonporosity pixel is replaced by the majority solid phase present in a limited neighborhood (e.g., 3 x 3) centered at the pixel. The image produced by this median filtering process can then be analyzed to determine phase area fractions, the perimeter fraction of each phase in contact with porosity, and the two-dimensional correlation function for any individual or combination of phases. This quantitative spatial information then can be utilized in reconstructing a three-dimensional representation of the cement of interest.
Fig. B1. False-color two-dimensional image of cement particles for Cement 115 with X-ray signal to color mappings of calcium to red, silicon to green, and aluminum to blue.