Research On Optical Lithography Simulation And DFM For Nano-scale Circuits

VLSI chip manufacturing technology has been advanving along with the Moore's Law. The Critical Dimension of state-of-the-art layout already achieves 32nm. However, the systematical variations are increasing. Due to the fact that systematical variation could be simulated by model, it is necessary to research on some manufacturing related critical processes, in order to correct or compensate effects caused by systematical variations through models. In addition, for some complex SoC designs, it is becoming more and more difficult to improve the yield only through optical proximity correction from the manufacturing side. In fact, it is urgent to consider the manufacturing effect early in the design phase, this is what we ailed design for manufacturing (DFM). DFM is a bridge connecting both design and manufacturing. Based on this interface, some problems really affecting yield are being solved. Under this background, this thesis is exploring two problems of how to get the more accurate phenomenal model and how to solve the DFM problems. Next is a summarization of main research contents and innovations:Phenomenal model calibration system based on Bessel-like sampling functionsA model, for the ultra-deep sub-micron optical lithography, should consider all sub-distortion effects, or give a series of calibration parameters which could affect these effects. However, if we use method of using first-principle to simulation the effects, it will make the simulation time longer, which can not be used in the full-chip layout simulation. So we use the experience model or black-box model to characterize this kind of effects. There're already some researches finding that the suggested effects could still be modeled using a single TCC matrix. In theory, we can consider the effects by modifying the TCC matrix coefficients directly. However, the TCC matrix is large in size, it is difficult to modify the coefficients directly. So in this thesis, we proposed a new flow. That is, first we use the Bessel-like kernel to sample the original TCC matrix, then we get a smaller size matrix BTCC. Then we select the diagonal coefficients of matrix to optimize the matrix using the genetic algorithms. Experiments results show that this kind of model calibration method could simulate the distortion related effects much faster and accurate.Inverse mask synthesis based on initial SRAF insertionThe traditional optical proximity correction technology is based on the heuristic iteration algorithms. It moves the fragmented segments guided by EPE related cost function until getting convergent. However, in the 32nm technology node or beyond, this kind of segment interation related method is already very complex, and can not achieve high accuracy. Then we need to go back to the pixel based inverse mask synthesis technology which inverse the layout directly. In this thesis, we proposed a new two dimentional discrete cosine transformation based pixel mapping technique, and proposed an initial SRAF insertion based flow. We use this kind of new generated layout to feed into the inverse engine. For the initial SRAF insertion lacation determination, we use the single kernel method to place the SRAF in the constructure places. After the experiments, we can find that the layout optimizd by the intial SARF inserted initial layout could have less complex features.Design for Manufacturing based on Autonomous OPCThe existing OPC engine could only moderately correct the pattern, and could not make some modifications or movement in some locations. In this thesis, we novelly proposed a new autonomous OPC paradigm. OPC engine could use the new suggested data structure to autonomously modify some lacations originally should be considered in the placement or routing phase. It is not only differenct from traditional hotspots alleviation in the design phase through prediction, but also different from the existing OPC engine which only have limited capability. The initial experiments show that, this kind of novelly new automous OPC engine could make the error-prone locations or process margin small locations good, based on the accurate simulations. We enlarged the process window and solved the layout related manufacturing problems.