“Structural characterisation of semiconductors by computer methods of image analysis.”

M.HERNÁNDEZ FENOLLOSA; D. CUESTA-FRAU; L.C.DAMONTE; M.A. SATORRE AZNAR

San Diego, USA

Conferencia; SPIE 2005, Optics & Photonics 2005, San Diego, USA; 2005

Analysis of microscopic images for automatic particle detection and extraction is a field of growing interest in many scientific fields such as biology, medicine and physics. In this paper we present a method to analyse microscopic images of semiconductors in order to, in a non-supervised way, obtain the main characteristics of the sample under test: growing regions, grain sizes, dendrite morphology and homogenisation. In particular, nanocrystalline semiconductors with dimension less than 100 nm represent a relatively new class of materials. Their short-range structures are essentially the same as bulk semiconductors but their optical and electronic properties are dramatically different. The images are obtained by scanning electron microscopy (SEM) and processed by the computer methods presented. Traditionally these tasks have been performed manually, which is time-consuming and subjective in contrast to our computer analysis. The images acquired are first pre-processed in order to improve the signal-to-noise ratio and therefore the detection rate. Images are filtered by a weighted-median filter, and contrast is enhanced using histogram equalisation. Then, images are thresholded using a binarization algorithm in such a way growing regions are segmented. This segmentation is based on the different grey levels due to different sample height of the growing areas. Next, resulting image is further processed to eliminate the resulting holes and spots of the previous stage, and this image is used to compute the percentage of such growing areas. Finally, using pattern recognition techniques (contour following and raster to vector transformation), single crystals are extracted to obtain their characteristics.