Multi-Scale Thickness and Roughness Characterization of Thin Silicon-On-Insulator Films

Thickness uniformity of the silicon top layer is extremely important for the next generation of CMOS electronics based on Fully-Depleted Silicon-On-Insulator (FD-SOI) wafers. It requires accurate thickness characterization over a wide spatial bandwidth, from transistor to wafer scales [10 nm-300 mm]. We propose a data processing method using one-dimensional Power Spectral Density functions to describe both roughness and thickness variations of thin layers over the bandwidth of interest. The study of the correlation between thickness and roughness of the silicon top layer of SOI wafers leads to the distinction of three spectral domains: high-frequencies, where thickness variations are larger than roughness variations, mid-frequencies, where thickness and roughness variations are comparable, and low-frequencies, where the top layer conformably follows the buried oxide roughness. Furthermore, the spectral analysis of thickness variations, points out fractal behaviors over three regions (scaling exponent 0.3, 1.2 and 0.2 respectively). These regions provide the spectral foot-prints of fracture propagation [1.10−5−2.10−3 μm−1], surface self-diffusion [0.3−1 μm−1] and sacrificial oxidation [1−10 μm−1] respectively, all phenomena involved in fabrication of SOI wafers by the ion-cut process. Additionally, over the mid-frequencies region [0.03−1 μm−1], we observe a "bump" which results from the depth and size-distribution of micro-cracks generated during the ion-implantation step.