Abstract

The array of SU-8 photoresist pillars (10 ´ 10 ´ 50 µm) on a copper substrate was patterned by X-ray lithography using a synchrotron source. Flat cobalt films can then be sputter-deposited on the chemically stable and mechanically hardened SU-8 pillars in spite of geometry distortion and pattern collapse in some regions. Since any deviations from the pattern affect magnetic properties of the structure, images of these micropillars obtained from an optical microscope and a scanning electron microscope (SEM) were inspected using five different edge detection algorithms. The Canny and Laplacian of Gaussian (LoG) operators can detect all cross sections of micropillars, but noise points and lines were also included. The outputs from image processing by the Prewitt and Sobel operators yielded lower noise, but suffered from discontinuity, especially in the case of tilt angled SEM images. The Roberts cross edge detector had a weakest response to edges of the pattern under test. The 3D nature of the micropillars is an origin of low-contrast edge and background noise in the images. 

Key words:    X-ray lithography, SU-8 photoresist, high-aspect-ratio microstructure, image processing, edge detection.