Morphological classification of nanoceramic aggregates

Aluminum silicate nanoaggregates grown at near-room temperature on an organic template under a variety of experimental conditions have been imaged by transmission electron microscopy. Images have been automatically classified by an algorithm based on “spectrum enhancement”, multivariate statistics and supervised optimization. Spectrum enhancement consists of subtracting, in the log scale, a known function of wavenumber from the angle averaged power spectral density of the image. Enhanced spectra of each image, after polynomial interpolation, have been regarded as morphological descriptors and as such submitted to principal components analysis nested with a multiobjective parameter optimization algorithm. The latter has maximized pairwise discrimination between classes of materials. The role of the organic template and of a reaction parameter on aggregate morphology has been assessed at two magnification scales. Classification results have also been related to crystal structure data derived from selected area electron diffraction patterns.