![]() Here, the depth estimation task is constrained with a single image available per scene during the inference phase. Traditionally, these techniques relied on stereo pairs of input images, but more recently the subfield of monocular depth estimation has emerged. ĭepth estimation from 2D images has been studied thoroughly in the field of computer vision and is nowadays applied to robotics, autonomous driving, medical imaging and many other scene understanding tasks. It is expected that 3D metrology will become crucial in the semiconductor industry’s quest to keep up with the requirements of Moore’s Law. In practice, detailed metrology that provides the true 3D geometry of this structure is desired for various reasons. However, the obtained SEM images are a two-dimensional (2D) representation of the electron interactions with the surface. Currently, SEM is the fastest way of measurement that provides local geometry information. ![]() ![]() These measurements are important for controlling the fabrication process, which enables yield optimization of a produced wafer. In the semiconductor industry, critical dimension scanning electron microscopes (CD-SEMs) are used to measure the spatial lateral dimensions of structures on a microchip. An experiment on a dense line space dataset yields a mean relative error smaller than 1%. To the extent of our knowledge, we are the first to achieve accurate depth estimation results on real experimental data, by combining data from SEM and scatterometry measurements. Additionally, it is shown that this method is well suited for other important semiconductor metrics, such as top critical dimension (CD), bottom CD and sidewall angle. We have tested this method first on a synthetic contact hole dataset, where a mean relative error smaller than 6.2% is achieved at realistic noise levels. This step employs scatterometry data to address the ground-truth scarcity problem. The training procedure includes a weakly supervised domain adaptation step, which is further referred to as pixel-wise fine-tuning. We demonstrate that the proposed neural network architecture, together with a tailored training procedure, leads to accurate depth predictions. In this work, we present a method that is able to produce depth maps from synthetic and experimental SEM images. The main objective of this work is to investigate whether sufficient 3D information is present in a single SEM image for accurate surface reconstruction of the device topology. In the semiconductor industry, critical dimension scanning electron microscopes (CD-SEMs) are predominantly used for 2D imaging at a local scale. Unfortunately, present metrology tools do not offer a practical solution. To support the ongoing size reduction in integrated circuits, the need for accurate depth measurements of on-chip structures becomes increasingly important.
0 Comments
Leave a Reply. |
Details
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |