Quantitative and non‐destructive defect metrology for beyond Si semiconductors

Abstract

Electron channeling contrast imaging (ECCI) is a powerful scanning electron microscopy (SEM) technique for the visualization and analysis of crystalline defects like dislocations and stacking faults. Distortion of the crystal lattice of a material due to the presence of such defects causes the variation of the backscattered electron intensity, allowing their visualization. ECCI has been demonstrated to be a fast and robust method for assessing the density and Burgers vector of different defect types in various materials with reliability comparable to that of transmission electron microscopy [1, 2]. However, in contrast to TEM, ECCI can be used for the non‐destructive investigation of large areas. This fact makes the technique particularly interesting for the semiconductor industry, where defect metrology techniques for the non‐destructive analysis of (Si)Ge and III/V compounds with dislocation densities below 10 5 cm ‐2 , are crucial to support CMOS scaling beyond the 10nm node. In order to analyze such lowly defective samples, areas exceeding about 200 x 200µm 2 in size need to be examined to ensure proper statistics. For this purpose we acquire a set of tiles that can be stitched into a single image, thereby leading to an image resolution allowing for the detection of single threading dislocations. The FEI software application MAPS, dedicated to the automated acquisition of high resolution images from large areas, is used to record a set of ECC images that are further processed and examined for the presence of defects. It is important to note that for proper ECC imaging (i.e. maximum channeling contrast at a defect site), the sample needs to be oriented close to the Bragg condition. This is facilitated by tilting and rotating the sample according to the electron channeling pattern (Fig. 1) acquired by scanning the sample at low magnification. The investigated area of the tilted specimen is maintained in focus during the acquisition through the interpolation of the settings. Imaging conditions such as accelerating voltage and beam current were fine‐tuned in advance using a sample with higher defect density (Fig. 2). A retractable below‐the‐lens backscatter electron (BSE) detector is used to record the individual ECC images. Using the above described procedure we analyzed in detail the density and distribution of threading dislocations in blanket SiGe layers of different defect densities. Our results reveal that in case of dedicated strain relaxed buffer layers the surface appears basically defect free over several tens of micrometers, only locally one can observe individual defects and pile‐ups of threadings reaching the layer surface. Results are verified by defect decoration using chemical etching followed by optical etch pit detection. Our work demonstrates that ECCI in combination with automated image acquisition provides quantitative information on defect density and distribution on systems foreseen for future semiconductor devices.