These clusters (void or inorganic phase) can act as either reservoirs for drug delivery, inhibitors in self-healing coatings or mechanical reinforcement in polymers. From the authors’ perspective this system is of great importance since it represents a model for protective coatings. The release of SrCrO 4 inhibitor from epoxies is not well understood and the discovery of clusters of particles and the possibility of voids created through the dissolution of the cluster represents a significant scientific breakthrough for coatings and other applications as indicated earlier. We have devised metrics that can be used to characterize the clusters and ultimately to control their design. While there are a few reports on the application of CT to the study of micrometer-sized inorganic phase distributions within polymers,[11] this is largely an unexplored field. In this letter we combine X-ray CT and a new SEM-based sectioning technique called serial block face scanning electron microscopy (SBFSEM)[12] to investigate the 3D distribution of inorganic particles (SrCrO 4) in a polymer (epoxy-polyamine matrix). The combination of these two techniques provides spatial information covering four orders of magnitude in scale from 50 nm to 500 µm.

CT volume reconstructions of the sample showed three distinguishable regions comprising the epoxy resin which included regions of low (green) and normal (blue) density volumes (LDE and NDE respectively) and SrCrO 4 particles (red) as shown in the full section in Figure 1 (a). The least-square optimised segmentation approach used here relies on differentiating the individual components on the basis of their linear absorption coefficients, which is derived from their composition. The label “normal” has been used for voxels that match the linear absorption coefficient for the pure epoxy. The LDE is about one-third, implying that it has a lower density. It should be noted that the sample is a long slither, and its outline is of irregular shape. The 3D distribution of the SrCrO 4 particles and the LDE are revealed more clearly when the NDE component is removed (Figure 1 (b)). However, the image of the total volume of the either the LDE or the SrCrO 4 particles does not give any indication of the distribution of structures within this phase. To investigate these structures, a 3D clustering algorithm, consisting of target phase binarisation and nearest neighbour labelling [13] was developed for the voxels. The CT process divides the total volume of the sample into unit sub-volumes called voxels, which form a 3D lattice. A cluster is defined as a group of neighbouring voxels of the same type of composition. It is important to note that the clustering algorithm does not provide any information of how many SrCrO 4 particles are involved in cluster formation, it only counts the voxels. For the