A Rapid Prototyping Approach to Semiconductor Device Manufacturing Process Simulation

[00:82381]我们建议开发计算框架，使微电子器件制造工艺系统的模拟器开发的快速原型方法成为可能。本研究旨在填补基于超级计算机的高分辨率模拟和集总参数模型之间存在的差距，集总参数模型是目前由过程模拟、分析、模型简化和参数估计的专门方法所跨越的范围-通常彼此不兼容的数值工具。所提出的过程模拟方法是基于开发计算元素，这些计算元素与实现高维投影、频谱滤波和高级加权残差方法（如降基和非线性Galerkin投影）时所采取的步骤具有一对一的对应关系。模型效度评估方法；解分析（例如，稳定性）和离散误差将在这个框架中讨论。事实上，由于它依赖于参数估计方法，在这种仿真框架中总会产生一个简化模型预测能力的绝对度量。这些相互关联的计算工具将在MATLAB环境中开发，利用MATLAB的面向对象编程特性来简化复杂系统的模拟，并创建一个将我们的模拟工具整合到谱元素和其他商业软件包中的途径。为了提高化学气相沉积和其他半导体材料制造单元操作中的跨晶圆加工均匀性，需要生成计算效率高、经过验证的模拟，这是本研究的主要动机。简化模型适用于实时和运行到运行的控制，高效的工艺配方优化，以及基于模型的不可测量过程的感知和估计，如微特征演化。该提案将支持PI和英特尔技术CAD组之间的联合研究，通过研究生实习，旨在在公司研究环境中测试模拟工具。该方法的基本要素也将在位于马里兰大学的商业化学气相沉积（CVD）簇工具的背景下开发，其目标是产生一个经过验证的简化模型，用于晶圆温度预测和保形CVD研究。＊＊＊

0082381AdomaitisWe propose to develop the computational framework that will make possible a rapid proto-typing approach to simulator development for microelectronic device fabrication process systems. This research is intended to fill the gap that exists between supercomputer-based, highly resolved simulations and lumped-parameter models, a range currently spanned by specialized methods for process simulation, analysis, model reduction, and parameter estimation - numerical tools that are generally incompatible with each other. The proposed approach to process simulation is based on developing computational elements that have a one-to-one correspondence to the steps taken when implementing high-dimensional projection, spectral filtering and advanced weighted residual methods, such as the reduced basis and nonlinear Galerkin projections. Methods for assessing model validity; solution analysis (e.g., stability), and discretization error will be addressed in this framework. Indeed, because of its reliance on parameter estimation methods, an absolute measure of reduced-model predictive power always will be produced in this simulation framework. These interconnected computational tools will be developed in the MATLAB environment, taking advantage of MATLAB's object-oriented programming features to simplify simulations of complex systems and to create a pathway to incorporating our simulation tools in spectral element and other commercial software packages.The need to generate computationally efficient, validated simulations for improving across-wafer processing uniformity in chemical vapor deposition and other semiconductor materials manufacturing unit operations provides the primary motivation for this research. Reduced models are suitable for real-time and run-to-run control, efficient process recipe optimization, and model-based sensing and estimation of unmeasurable processes, such as microfeature evolution. This proposal will support the joint research between the PI and the Tecbnology CAD group of Intel through a graduate student internship aimed at testing the simulation tools in a corporate research environment. The basic elements of this approach also will be developed in the context of a commercial chemical vapor deposition (CVD) cluster tool located at the University of Maryland, with the goal of producing a validated, reduced model to be used for wafer temperature prediction and conformal CVD studies. ***