Chemical mechanical polishing (CMP) has become one of the most important process stages in the fabrication of advanced integrated circuits (IC). The CMP pattern effect strongly influences the planarization of the chip surface morphology after CMP, degrading the performance and the yield of the circuits. In this paper, we introduce a method to predict the post-CMP surface morphology with a convolutional neural network (CNN)-based CMP model. Then, CNN-based, density step height (DSH)-based, and common neural-network-based CMP models are built to compare the accuracy of the predictions. The test chips are designed and taped out and the predictions of the three models are compared with experimental results measured by an atomic force profiler (AFP) and scanning electron microscope (SEM). The results show that CNN-based CMP models have better accuracy by taking advantage of the CNN networks to extract features from images instead of the traditional equivalent pattern parameters. The effective planarization length (EPL) is introduced and defined to make better predictions with real-time CMP models and in dummy filling tasks. Experiments are designed to show a method to solve the EPL.
In advanced integrated circuit manufacturing processes, the quality of chemical mechanical flattening is a key factor affecting chip performance and yield. Therefore, it has become increasingly important to develop an accurate predictive model for the chip surface topography after chemical mechanical flattening. In the modeling process, the noise problem of atomic force microscopy measurement data is relatively serious. To solve this problem, the noise characteristics of atomic force microscope measurement data for chip surface topography in this field are studied and discussed in this paper. It is found that the noise present in such problems is mainly triggered by the vibration and tilt of the probe. Two types of noise, low-frequency and high-frequency, are presented in the time domain. In order to solve the noise problem in this modeling data, this paper analyzes the spectral characteristics of the measurement data using Fourier transform, and a wavelet-Fourier transform composite noise reduction process is proposed. The algorithm is applied to the noise reduction of the chip surface data of 32 nm copper interconnect process. The noise reduction results were compared with scanning electron microscope photographs to verify the effectiveness of the noise reduction.
As a planarization technique, chemical mechanical polishing (CMP) continues to suffer from pattern effects that result in large variations in material thickness, which can influence circuit performance and yield. Therefore, tools for predicting post-CMP chip morphology based on the layout-dependent effect (LDE) have become increasingly critical and widely utilized for design verification and manufacturing development. In order to characterize the impact of patterns on polishing, such models often require the extraction of graphic parameters. However, existing extraction algorithms provide a limited description of the interaction effect between layout patterns. To address this problem, we calculate the average density as a density correction and innovatively use a one-dimensional line contact deformation profile as a weighting function. To verify our hypothesis, the density correction method is applied to a density step-height-based high-K metal gate-CMP prediction model. The surface prediction results before and after optimization are compared with the silicon data. The results show a reduction in mean squared error (MSE) of 40.1% and 35.2% in oxide and Al height predictions, respectively, compared with the preoptimization results, confirming that the optimization method can improve the prediction accuracy of the model.