集成电路工艺技术的进步、互连线电流密度的增加和芯片功率的提高，可靠性成为芯片设计关键的评价指标之一。电迁移 (Electromigration, EM) 通过原子迁移导致芯片上互连线和通孔的故障，是集成电路互连线可靠性中主要的问题之一。因长期承载单向电流且电流密度大，供电网络是集成电路互连线中最容易受电迁移影响的部分。通孔对供电网络电源完整性和电迁移可靠性有着重要影响，在供电网络设计中对通孔的分析和优化工作还不够充分。随着工艺尺寸缩小和芯片规模增大，芯片设计对供电网络电迁移模型要求更加准确和高效，传统基于最坏情况和简单经验的电迁移模型会造成保守设计或过度优化，从而显著增加芯片面积、功耗和成本。 论文对集成电路供电网络的电迁移问题开展研究，探索供电网络电迁移分析和优化方法。通过研究通孔结构、温度和应力分布对电迁移的影响，完善供电网络电迁移分析模型，提高电迁移分析精度。通过研究供电网络模型与电迁移之间的相关性，结合电流分布对供电网络进行电迁移分析，降低了分析的复杂度。通过研究供电网络结构，提出基于通孔灵敏度的供电网络快速分析和优化算法，提高供电网络优化的效率，有效地避免过度优化。论文主要研究工作包括：第一，针对现有电迁移模型缺少对通孔的全面分析，研究了通孔阵列的电迁移特性，提出通孔阵列的电迁移分析与优化模型，实现多层供电网络的电迁移分析。第二，针对现有电迁移分析模型中对温度的影响考虑不周全问题，提出了温度感知的电迁移分析方法。综合考虑焦耳热和热耦合对通孔及供电网络电迁移的影响，提高电迁移分析的准确性。第三，针对全芯片供电网络电迁移分析复杂度高的问题，提出了基于电流分布的供电网络电迁移分析方法。基于电流分布相关性，确定多段互连线最大应力通孔位置，提高供电网络电迁移分析效率。第四，针对供电网络电迁移过度优化的问题，提出了基于通孔灵敏度分析的供电网络优化方法。充分考虑供电网络结构的影响，选择关键通孔进行优化，避免电迁移过度优化，实现供电网络电迁移和电压降的协同优化。

Reliability has become a key evaluation criterion of physical designs due to the continuous scaling in the process technology of integrated circuits, the increase in the densityof interconnection currents, and the increase in chip power consumption. One major problem in the reliability of VLSI interconnections is the Electromigration (EM) phenomenon, as the atomic migration during EM can cause failures in wires and vias. Power grid networks are the most vulnerable part among interconnections to long-term reliability issues due to long-term exposure to strong and unidirectional current flows on a chip. In addition, vias have significant effects on the power integrity and the EM reliability of power grid networks, while analysis and optimization of the vias for the design of power grid networks have been insufficient. In addition, as both the number of the on-chip transistors and the scaling of process technology continue to increase aggressively, chip design requires more accurate and more efficient EM models for power grid networks. However, the conventional methods based on the worst-case or the simple empirical EM model have the nature of being over-conservative and can lead to over-design or unnecessary design iterations, which significantly increases die area, power, and cost.In this paper, we conducted an in-depth study of the EM problem of the power grid network of integrated circuits and proposed a method for analyzing and optimizing power grid networks. First, by investigating the influence of via structure, temperature, and stress distribution of EM, we improve both the EM analysis model of the power grid network and the accuracy of EM analysis. Furthermore, by exploiting the correlation between the power grid network model and EM while combining the current distribution, we achieve both analysis of the power grid network for EM and reduction in analysis complexity simultaneously. Finally, we propose a fast analysis and optimization algorithm based on via sensitivity for the power grid network. Our algorithm can expedite the optimization of power grid networks and mitigate the over-design problem. The main research works of this paper are as follows:First, in view of the lack of analysis of the via in the existing electromigration models, we studied the EM property of the via array. Our study proposes a model for analyzing and optimizing the EM of via arrays, which enables the EM analysis of multi-layer interconnections in power grid networks. Second, to involve the thermal effects into the EM analysis model in a better way, we propose a temperature-aware EM analysis algorithm in combination with an extended analysis model. By considering both the Joule heating and the effects that thermal coupling has on the EM of vias and power grid networks, we improve the accuracy of the analysis.Third, to reduce the complexity of the EM analysis of full-chip power grid networks, we propose an electromigration analysis method of the power grid network based on the current distribution. Based on the correlation between the current distribution and EM in the power grid network, the position of the maximum stress via of the multi-segment interconnection line is determined, and we can improve the efficiency of the EM analysis of power grid networks.Finally, to mitigate the over-optimization problem in the power grid network EM, we propose a power grid network optimization method based on sensitivity analysis. By considering the power grid network structure, we selectively apply optimization to key vias, whereby we avoid over-optimization of EM while achieving joint optimization of the power grid network EM and the excessive IR drop.