Multiscale Modeling, Optimization, and Control of Microstructural Evolution

PI: Talid R. Sinno和Warren seider机构：宾夕法尼亚大学提案号：0730971题目：微观结构进化的自适应多尺度建模、优化和控制日益强大的计算资源和新算法扩展了多尺度建模的应用范围，其中原子尺度现象与宏观加工条件相关联。这个项目的目的是开发一个分层的、多尺度的建模框架，适合在重复执行的环境中使用，比如过程优化和反馈控制。随着在先进材料加工过程中对原子种类空间分布波动的容忍度越来越严格，这一目标正变得越来越重要。本项目将考虑两类微观结构演变：(1)大块晶体半导体材料中的原子聚集，包括硅、锗和硅-锗；(2)多层金属合金（如铜-镍）中的物质再分配/偏析和相分离，这些合金是磁存储介质的模型系统。智力优势：该项目既有综合元素，也有综合元素。几个当代多尺度建模挑战将解决，以构建一个高度自适应，粗粒度晶格动力学蒙特卡罗模拟框架。第一个目标将是开发新的晶格动力学蒙特卡罗（LKMC）模拟，该模拟隐含地解释了真实系统中复杂的非晶格重排，特别是在半导体和金属加工中常见的高温下。这将通过与大规模平衡和非平衡分子动力学模拟产生的数据进行系统比较来完成。由此产生的“MD匹配”LKMC模拟将在最近引入的粗粒度策略中实现，该策略允许在控制误差的情况下进行系统的降阶。这项工作的一个关键成果将是将这种粗粒度框架应用于具有多物种之间复杂相互作用的LKMC模拟的策略。粗粒度将在一个完全自适应的框架内实现，其中粗粒度的程度在空间和时间上动态调整，这是由系统的微观结构演变以及整体控制/优化环境的分辨率需求所决定的。考虑中的许多现象已经在实验中进行了研究，在硅的情况下，与工业的持续合作提供了与聚集体形态作为热化学处理环境的函数相关的详细微观数据。更广泛的影响：本项目所考虑的材料系统和微观结构演变现象本身就具有根本性的意义，同时也是广泛问题的典型例子。原子团簇的成核和生长以及在空间异构环境中的扩散是与先进器件和材料制造相关的大量过程中的基础现象。该项目将汇集多尺度建模的几个方面，并将它们集成到一个控制环境中，目的是为多尺度优化和控制开发一个原型框架。模型开发将适用于各种各样的工艺和材料。例如，自适应粗粒度LKMC方法是一种非常强大的通用方法，它是多尺度的，而不是高度特定于系统的。该项目将汇集两个传统上不同的研究领域的元素，并将它们紧密结合在一起。大规模分子动力学和基本动力学蒙特卡罗代码已经就位，允许研究生在这个项目中工作，专注于新的方面，如自适应粗粒度和集成的优化和控制组件与多尺度模型。该项目的基本思想将用于开发与宾夕法尼亚大学化学与生物分子工程（CBE）高级设计课程相关的教材。迄今为止，PI在这门课程中开发了超越传统化学加工的新设计项目，取得了很大的成功。例如，最近的项目涉及使用有限元建模设计化学气相沉积过程，但尚未尝试包括原子尺度的建模。这项工作将为创建基于微观目标函数优化的设计模块提供基础，并将代表CBE顶点课程发展的又一重要步骤。

PI: Talid R. Sinno and Warren SeiderInstitution: University of Pennsylvania Proposal Number: 0730971Title: Adaptive Multiscale Modeling, Optimization, and Control of Microstructural EvolutionIncreasingly powerful computational resources and new algorithms have extended the application horizon of multiscale modeling, in which atomic scale phenomena are linked to macroscopic processing conditions. The aim of this project is to develop a hierarchical, multiscale-modeling framework suitable for use in repeated execution environments, such as process optimization and feedback control. This objective is becoming increasingly relevant as tolerances with respect to fluctuations in the spatial distribution of atomic species become tighter during the processing of advanced materials. Two classes of microstructural evolution will be considered in this project: (1) atomic aggregation in bulk crystalline semiconductor materials, including silicon, germanium, and silicon-germanium, and (2) species redistribution/segregation and phase separation in multilayered metallic alloys, such as copper-nickel, which are model systems for magnetic storage media.Intellectual Merit: The project has both synthetic and integrative elements. Several contemporary multiscale modeling challenges will be addressed in order to construct a highly adaptive, coarse-grained lattice kinetic Monte Carlo simulation framework. The first aim will be to develop novel lattice kinetic Monte Carlo (LKMC) simulations that implicitly account for complex off-lattice rearrangements in real systems, particularly at the elevated temperatures common in semiconductor and metals processing. This will be accomplished by systematic comparison to data generated by large-scale equilibrium and non-equilibrium molecular dynamics simulations. The resulting "MD matched" LKMC simulations will then be implemented within recently introduced coarse-graining strategies that allow for systematic order-reduction with controlled error. A key outcome of this work will be strategies for applying this coarse-graining framework to LKMC simulations with complex interactions between multiple species. The coarse-graining will be implemented within a fully adaptive framework in which the degree of coarse-graining is adjusted dynamically in space and time as dictated by the evolving microstructure of the system as well as the resolution needs of the overall control/optimization environment. Many of the phenomena under consideration have been studied experimentally, and in the case of silicon, an ongoing collaboration with industry provides access to detailed microscopic data related to aggregate morphology as a function of thermochemical processing environment.Broader Impacts: The material systems and microstructural evolution phenomena considered in this project are of fundamental interest in their own right but also are prototypical examples of a broad class of problems. Nucleation and growth of atomic clusters, and diffusion in spatially heterogeneous environments are cornerstone phenomena in a large number of processes related to the fabrication of advanced devices and materials. The project will bring together several aspects of multiscale modeling and integrate them into a control environment with the aim of developing a prototypical framework for multiscale optimization and control. The model developments will be applicable to a wide variety of processes and materials. For example, the adaptive coarse-grained LKMC method is an extremely powerful general approach that is multiscale without being highly system specific. The project will bring together elements from two traditionally different research areas and couple them closely. Large-scale molecular dynamics and basic kinetic Monte Carlo codes are already in place, allowing the graduate student working on this project to focus on novel aspects such as adaptive coarse-graining and integration of the optimization and control components with the multiscale models. The basic ideas of the project will be used to develop educational materials related to the Chemical and Biomolecular Engineering (CBE) senior design course at Penn. The development of new design projects that go beyond traditional chemicals processing into this course by the PI has been highly successful thus far. For example, recent projects have involved the design of chemical vapor deposition processes using finite element modeling, but no attempt has yet been made to include atomic-scale modeling. This work should provide a basis for creating design modules based on the optimization of microscopic objective functions and would represent yet another significant step in the evolution of the CBE capstone course.