Figures 1 to 3 provide two-dimensional and three-dimensional images of the raw microtomography data sets and of the results of subtracting the original data set from that obtained after 1 d of hydration. In the two-dimensional images in Figure 1, the specimen is seen to contain four readily distinguishable phases based on differences in density (or more correctly x-ray absorption): dark air voids and empty pores (within LWA particles), dark grey LWA particles and water-filled pores within the LWA, light grey sand particles, and bright cement paste. Some empty pores are observed within the LWA particles even in the image set obtained immediately after mixing, suggesting that either these pores were not saturated by the saturation procedure employed in this study or that they had emptied during the initial mixing and casting of the specimen. In carefully comparing the original 2-D slice to that obtained after 1 d or 2 d of hydration, for several of the (circled) LWA particles, one can notice a general darkening and the appearance of new (empty) pores within the particles in the latter images. These correspond to initially water-filled pores that were emptied to supply internal curing water to the surrounding cement paste during its hydration. These emptied pores are highlighted in light grey in the lower right image of Figure 1 and in Figures 2 and 3. All three of these images were obtained by subtracting the original image set from that obtained after 1 d of hydration. The sizes and shapes of many of the “individual” pores (or porous regions) within the LWA particles are readily observed, both in their two-dimensional (Figures 1 and 3) and in their three-dimensional (Figure 2) forms. Most of these empty pores were created during the first day of hydration, with little further change being observed between the 1 d and 2 d hydrated image sets.
To perform a more detailed quantitative analysis of the volume of water moving from the saturated LWA to the hydrating cement paste, each three-dimensional image set was processed and analyzed as follows. First, to reduce the random noise present in the three-dimensional image sets, each image was processed using a median filter. In applying a
Figure 1: Two-dimensional slices (13 mm x 13 mm) from 3-D microtomographic data sets corresponding to: upper left- mortar immediately after mixing, upper right – mortar after about 1 d of hydration, lower left – mortar after about 2 d of hydration, and lower right – subtracted image of 1 d - original slices. In the subtracted image, lightest grey indicates regions of drying and darkest grey indicates regions of wetting [11].
median filter, each voxel (image element) within the specimen volume is replaced by the median (greylevel) value for all voxels within a fixed size cube centered on the voxel being considered. Median filters remove noise by smoothing the data, but while preserving small details and sharp edges [17]. For this study, 3 voxel x 3 voxel x 3 voxel and 5 voxel x 5 voxel x 5 voxel cubes were investigated. After filtering, the greylevel histogram (a plot of the number of voxels containing each greylevel intensity value vs. greylevel intensity) was determined for each (original and processed) data set. Representative histograms are provided
Figure 2: Three-dimensional image of original data set subtracted from that obtained after 1 d of hydration. Light grey volumes indicate regions where the LWA particles have lost water (to the surrounding hydrating cement paste). 3-D volume is 4.6 mm x 4.6 mm x 4.7 mm [11].
Figure 3: Two-dimensional image (4.6 mm x 4.6 mm) of a portion of the original mortar microstructure with the locations of the evacuated water (in light grey) superimposed [11].
in Figure 4 for the data set obtained after 2 d of hydration. Peaks in a greylevel histogram usually indicate individual phases within a microstructure and the sharpening of the peaks in the median-filtered image sets relative to the original data set is clearly observed in Figure 4. Based on its greater separation into individual peaks, the 5 voxel x 5 voxel x 5 voxel median-filtered data sets were selected for further analysis. In Figure 4, the “peaks” corresponding to each of the four detectable phases listed above are labeled with their corresponding name (pores, LWA, sand, and paste).
Figure 4: Greylevel histograms for three-dimensional image set obtained after 2 d of hydration, subjected to different size (three-dimensional) median filters.
Figure 5 presents the greylevel histograms for the original, 1 d, and 2 d hydration image sets, after applying the 5 voxel x 5 voxel x 5 voxel median filter to each. The differences in the histograms as a function of hydration time should now correspond to the changes occurring within the specimen microstructure (mainly water movement from the LWA to the surrounding cement paste). The effects of the water movement are three-fold: 1) an increase in the number of voxels detected as empty pores, 2) a decrease in the number of voxels detected as (water-filled) LWA, and 3) an increase in the brightness of the voxels containing hydrating cement paste (due to the density and x-ray absorption increase accompanying the water imbibition into the hydrating porous microstructure) as indicated by the three greylevel histogram curves crossing one another at a greylevel value of about 5000 in Figure 5. Here, the first of these will be used to estimate the cumulative volume fraction of internal curing water that has moved from the LWA reservoirs to the surrounding paste, corresponding to the volume of created empty pores. To do this, the integral of the greylevel histogram is computed for the darkest (lowest intensity) subset of greylevels, specifically between 0 and an upper limit that defines the greylevel boundary between empty pores (both air voids and empty pores within the LWA) and water-filled pores/LWA particles. For the analysis conducted here, this upper limit was set both at 3500 (corresponding to the presence of a slight shoulder in the greylevel histograms in Figure 5) and at 3900 (corresponding to the point where the 1 d and original histograms cross each other indicating an increase in empty pores and a decrease in water-filled ones). For example, the number of voxels with intensities less than 3900 was determined for each median-filtered image set and divided by the total number of voxels within the three-dimensional specimen volume to obtain a volume fraction.
Fig. 5: Greylevel histograms for three-dimensional image sets (median filtered) obtained immediately after mixing, after 1 d, and after 2 d of hydration.
In Figure 6, these volume fractions are plotted against hydration time and compared to more conventional measures of hydration such as cumulative heat release, degree of hydration based on LOI, and chemical shrinkage (in units of milliliter of imbibed water per milliliter of mortar). The initial volume fraction of voxels with greylevel intensities less than 3900 (about 2.8 %) is seen to be in reasonable agreement with the measured air void content of the fresh mortar (2.1 %), considering that the image-based measurement also includes some empty pores within the LWA in addition to the air voids. The shape of the curves for the volume of water removed from the LWA during the first day of internal curing is seen to closely follow the shape of the curves for the degree of hydration of the specimen as determined by the three different analytical methods. This suggests that for this LWA system, the needed curing water is readily available to the hydrating cement paste during the first few days of hydration. A more direct comparison of the volume fraction of “emptied LWA” voxels to the volume fraction of imbibed water (chemical shrinkage) of an equivalent saturated cement paste is provided in Figure 7. Because the volume fraction of “emptied” voxels is only a lower bound on the volume of water moving from the LWA to the hydrating cement paste, due to the emptying of water-filled pores that are smaller than the resolution limits of the tomographic data sets, etc., in Figure 7, the emptied voxels volume fraction vs. time has been scaled to match exactly the 24 h measured chemical shrinkage data. With this scaling, Figure 7 shows a virtual one-to-one agreement between the measured volume of imbibed water and the scaled measured volume of emptied pores during the first 36 h of the experiment.
Fig 6: Estimated curing water supplied vs. time in comparison to other conventional measurements of hydration.
Fig 7: Comparison of measured chemical shrinkage volume fraction to scaled (see text) emptied LWA pore volume fraction vs. curing time.
It should be noted that while the individual pores being emptied within the LWA particles are readily identified (e.g., in Figure 1), individual “locations” within the hydrating cement paste to which this water is moving could not be reliably detected in this experiment. At these early ages, the water from the LWA readily moves distances of 5 mm or more from its source [5, 6, 8] so that the internal curing water is more or less homogeneously distributed throughout the three-dimensional microstructure. In this study, the mortar was proportioned with enough internal curing water to satisfy the hydration demand only during the first 48 h of hydration, as it was envisioned to freeze the specimen after 2 d and hopefully observe water/ice movement from the hydrating cement paste to the air voids and (now) empty pores within the LWA particles. Unfortunately, we were only able to achieve a specimen temperature of about -8oC during this freezing, such that the water within the porous hydrating cement paste may not have frozen due to supercooling effects.
Next: Summary Up: Main Previous: Experimental