Nano Ic Chemical Mechanical Polishing Process Modeling And Simulation And Design For Manufacturing Technology Research

The semiconductor industry is one of the industries which have the most rapid growth rate in history. The great improvements of the semiconductor industry depend on continuously technology scaling down, which provides integrated circuit with better performance and exponentially increasing integration capability. As IC technology enters the nanometer regime, more and more large process variations have seriously deteriorated the yield. Yield loss problem has become a fateful and critical problem for nano-VLSI design. IC design is now entering into a new era of DFM/DFY (Design for Manufacturability and Design for Yield). Chemical mechanical polishing (CMP) process is recognized to suffer from pattern dependent problems which cause severe thickness variations of interconnects and the dielectric. The thickness variations can lead to variations in interconnect capacitance and resistance, leading to the performance uncertainty of the circuit. Thus, CMP process has become one of the most dominant factors that influence the chip performance and yield.In order to address the critical yield problem caused by CMP process mentioned above, predictive and accurate CMP models are highly desirable for better undersanding of the process, aiding in the optimization and development of the process and providing thickness variation information for designers. On the other hand, layout optimization for better planarization is also needed which reduces manufacturing variation and enhances the yield of the IC product. This dissertation mainly studies the above mentioned two parts, and the major contributions are listed below.A general rough-pad model is proposed for the chemical mechanical polishing process. The proposed rough-pad model has several advantages over existing models. First, general height distribution functions and autocorrelation functions are used to describe the pad surface, which are easier to obtain than pad asperity height and curvature distributions in existing models. Thus, the proposed model may be used as a CMP pad design tool for improving dishing and erosion. Second, the spectral representation technique and nonlinear transformation method used in the model allow rough-pad surfaces with general pad surface height distributions and autocorrelation functions. Thus, no assumption is made on the surface geometry and statistics of the pad. Third, a conjugate-gradient iteration scheme combined with the fast Fourier transform technique is used to solve the resulting wafer-pad rough-contact problems to fully take into account the bulk deformation of the pad and the interactions among neighboring asperities, which further reduces the simulation complexity. Model predictions are in good agreement with the experimental data in the existing literature. Based on the proposed model, the effects of CMP process parameters and underlying pattern geometries on dishing and erosion can be evaluated.To tackle the difficult balance between efficiency and accuracy in existing methods, a Fully Polynomial Time Approximation Scheme (FPTAS) for CMP dummy fill is proposed. The proposed work first formulates the dummy fill problem as a Covering Linear Program (CLP), and then presents an efficient FPTAS for such a problem. The proposed FPTAS approximates the optimal dummy fill solution within 1+εfactor in time O(ε-2m2 log m). Thus, it allows continuous trade-off between computational efficiency and solution quality. The proposed algorithm can be applied to solve both the traditional density-driven problem and the problem considering fill-induced coupling capacitance impact. Experimental results show that the new FPTAS algorithm is highly practical. Moreover, based on the approximation algorithm, we also propose a new greedy iterative algorithm. Experimental results show that it is more efficient than the approximation algorithm with only slight solution degradations and it achieves high quality solutions more efficiently than previous Monte-Carlo based heuristic methods.