基于失效物理的集成电路失效率计算方法研究

发布时间：2018-01-24 11:02

本文关键词： 失效物理失效率计算方法集成电路失效机理蒙特卡罗方法失效分布　出处：《广东工业大学》2015年硕士论文　论文类型：学位论文

【摘要】：随着半导体技术的不断发展与普及,集成电路的市场从过去要求高可靠性的军事、太空领域,逐渐拓展到更广阔的商用和民用市场。可靠性仿真与预测工具,成为加快开发进度,帮助寻求产品性能与可靠性之间平衡的有力手段。传统的“预计手册”的方法,由于更新周期、使用范围的一些限制,逐渐难以跟上行业发展的步伐。所以,以基于可靠性物理的预计方法逐步代替传统方法,成为一种趋势。可靠性物理模型在失效率与失效分布的预计方面一直有所欠缺,本文试图在这一方面做些尝试。本论文对芯片级CMOS集成电路的四种失效机理(HCI、NBTI、TDDB与EM)的物理模型进行了梳理。分别采用幸运电子模型及HU模型,得到HCI寿命模型；采用反应扩散模型,得到NBTI的寿命模型；采用渗滤模型,直接得到了TDDB的失效分布模型；采用Black方程,得到了EM的寿命模型。基于寿命模型,通过蒙特卡罗方法及连续随机变量非线性函数期望方差的近似求法两种数学方法,对HCI、NBTI及EM三种集成电路的失效机理进行Matlab仿真,得到其失效分布。通过梳理渗流模型的推导过程,整理出TDDB的失效分布模型。几种失效分布模型中的参数都来自于器件的特征参数与环境参数,这样就建立了器件四种失效机理的失效分布与器件固有参数之间的联系,使得基于失效物理的集成电路失效率预计变为可能。集成电路由大量CMOS组成,且多种失效机理造成的器件失效相互无关。本文也探讨了集成电路总的失效分布及常数失效率的计算。通过VB,将失效率的计算过程编制为一套计算程序,在输入CMOS集成电路的特征参数与环境参数后,能根据设定计算出失效率结果与失效分布图。
[Abstract]:With the continuous development and popularization of semiconductor technology, the market of integrated circuits has gradually expanded from the military and space fields which required high reliability in the past to a broader commercial and civil market, reliability simulation and prediction tools. Become a powerful means of accelerating development and helping to find a balance between product performance and reliability. The traditional "expected Manual" approach, due to the update cycle, has some limitations on the scope of use. It is gradually difficult to keep up with the development of the industry. Therefore, the traditional method is gradually replaced by the prediction method based on reliability physics. Reliability physical model has been lacking in the prediction of failure rate and failure distribution. This paper attempts to make some attempts in this field. In this paper, four failure mechanisms of chip level CMOS integrated circuits are discussed. The lucky electron model and Hu model were used to obtain the HCI lifetime model. The life model of NBTI is obtained by using the reaction diffusion model. The failure distribution model of TDDB is obtained directly by using percolation model. The life model of EM is obtained by using Black equation. Based on the life model, two mathematical methods, Monte Carlo method and approximate method to find the expected variance of nonlinear function of continuous random variables, are used to solve the problem of HCI. The failure mechanism of NBTI and EM integrated circuits is simulated by Matlab, and the failure distribution is obtained. The failure distribution model of TDDB is sorted out. The parameters of several failure distribution models come from the characteristic parameters and environmental parameters of the device. In this way, the relationship between the failure distribution of the four failure mechanisms and the intrinsic parameters of the device is established, which makes it possible to predict the failure rate of integrated circuits based on the failure physics. The integrated circuits consist of a large number of CMOS. This paper also discusses the calculation of the total failure distribution and constant failure rate of integrated circuits. Through VB, the calculation process of failure rate is compiled into a set of calculation programs. After input the characteristic parameters and environment parameters of CMOS integrated circuit, the failure rate result and failure distribution diagram can be calculated according to the setting.
【学位授予单位】：广东工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN432

【参考文献】