Automated Linescan Analysis for CMP Modeling

Many of the process steps used in semiconductor chip manufacturing require planar (smooth) surfaces on the wafer to ensure correct pattern printing and generation of multilevel interconnections in the chips during manufacturing. Chemical-mechanical polishing/planarization (CMP) is the primary process used to achieve these surface planarity requirements. Modeling of CMP processes allows users to detect and fix large surface planarity variations (hotspots) in the layout prior to manufacturing. Fixing hotspots before tape-out may significantly reduce turnaround time and the cost of manufacturing. Creating an accurate CMP model that takes into account complicated chemical and mechanical polishing mechanisms is challenging. Measured data analysis and extraction of erosion and dishing data from profile linescans from test chips are important steps in CMP model building. Measured linescans are often tilted and noisy, which makes the extraction of erosion and dishing data more difficult. The development and implementation of algorithms used to perform automated linescan analysis may significantly reduce CMP model building time and improve the accuracy of the models. In this work, an automated linescan analysis (ALSA) tool is presented that performs automated linescan delineation, test pattern separation, and automatic extraction of erosion and dishing values from linescan data.

Application of Machine Learning-Based Electrochemical Deposition Models to CMP Modeling

Chemical mechanical polishing/planarization (CMP) is the primary process used for modern integrated circuits (IC) manufacturing. Modeling of the post-CMP surface profile is critical for detecting planarity hotspots prior to manufacturing and avoiding fatal failures of chips. Electrochemical deposition (ECD) is a key process for the void-free filling of interconnection wires and vias in modern chips. Large surface topography variations generated after ECD affect the post-CMP surface profile. In this paper, several machine learning approaches are used to model surface profiles after ECD that are used as input to CMP models. Different combinations of deep neural networks, long-shortterm-memory (LSTM) recurrent networks, convolutional neural networks and XGBoost algorithms are investigated and compared. The model based on the XGBoost library showed superior performance and accuracy

Some Mechanical Models of Chemical-Mechanical Polishing Processes

Chemical-mechanical polishing (CMP)-a perspective technology in fabrication of micro-and nanoelectronics elements, devices and systems. The development of models of CMP processes remains to be the actual problem. It is pointed out that known CMP models do not account for the features of chemical and mechanical mechanisms of interaction of active fluid and particles with a polished surface as well as an interaction of a viscoelastic pad with the surface. A description of the elementary acts of such interaction are absent in the available models. On the base of the analytical review of the current state of the theory and problems of (CMP) modeling some approaches were suggested to the problem accounting for the complex of the phenomena of different scales determining the polishing rate such as diffusion of slurry into the surface layer and restriction of time of chemical treatment of the surface by a rough pad being under the action of a mechanical load. A model of the CMP process was developed. Within the framework of this model a dependence of the polishing rate on the loading parameters was derived. The dependence generalizes the empirical Preston law.