ITR Collaborative Research: Enabling Microscopic Simulators To Perform System-Level Analysis

研究：四所大学的一个跨学科研究小组正在就这一中型信息技术研究(ITR)项目进行合作，旨在系统地弥合复杂材料系统的微观描述与直接工程重要性的系统级分析之间的差距。将开发一种数学辅助计算方法，使微观级别的模拟器能够直接执行系统级别的分析，而不需要通过宏观(偏微分或积分-微分)演化方程来对材料系统进行中间级别的描述。具体地说，基于集合平均的“粗”时间步长的计算超结构将被“包裹”在最先进的微观动态模拟器周围，例如分子动力学、动力学蒙特卡罗、格子-玻尔兹曼或混合代码。这种方法将使微观模拟器能够执行高级系统级分析：复杂、非线性分布式过程的稳定性、分叉、“粗略”集成、灵敏度和控制任务。所计划的算法将在大规模并行机器上运行。计算框架将由以下基本元素组成：(I)用于描述粗略行为的感兴趣的统计量(例如，分布矩)的选择；(Ii)宏观初始条件到一致微观构型的集合的“提升”；(Iii)根据体现对物理系统的最佳当前描述的微观模拟器，在集合中的每个初始微观构形的相同(短)时间段上的演化；(Iv)对演化的微观构型的集合进行平均(“限制”)，以提供宏观演化的系统状态；以及(V)在宏观初始条件的有限集合上执行前三个步骤。这一新方法在实现上是健壮的，在科学和工程应用范围内是可移植的。它具有普遍的适用性，适用于所有在概念上可以进行宏观描述但不能以封闭形式提供的系统。它避免了获得和关闭这种宏观模型的困难，同时通过计算准确地提取了宏观模型将获得的信息，如果该模型是封闭形式的话。这提供了ITR和一系列应用领域之间的联系。影响：这项研究的影响将是在最先进的微观级别的模拟和快速的系统级别分析能力之间建立强大和通用的联系。虽然研究集中在非均质硬材料和复杂流体中的具体问题，但计算框架适用于广泛的复杂系统，包括生物系统、它们的处理和功能。由于它有可能彻底改变工程系统水平的分析，它可能会对教育产生影响，并进一步推动微电子、生物信息学和纳米技术的进步。

Research:An interdisciplinary research team, at four universities, is collaborating on this medium-size Information Technology Research (ITR) project aimed at systematically bridging the gap between microscopic descriptions of complex material systems and systems-level analysis of direct engineering importance. A mathematics-assisted computational methodology will be developed that will enable microscopic-level simulators to perform systems-level analysis directly, without the need to pass through an intermediate level description of the material system through macroscopic (partial differential or integro-differential) evolution equations. Specifically, an ensemble-averaged "coarse" time-stepper-based computational superstructure will be "wrapped around" state-of-the-art microscopic dynamic simulators, such as molecular dynamics, kinetic Monte Carlo, Lattice-Boltzmann or hybrid codes. This methodology will enable microscopic simulators to perform advanced systems-level analysis: stability, bifurcation, "coarse" integration, sensitivity, and control tasks, of complex, nonlinear distributed processes. The planned algorithms will run on massively parallel machines.The computational framework will consist of the following basic elements: (i) choice of statistics of interest (e.g. distribution moments) for describing the coarse behavior; (ii) "lifting" of a macroscopic initial condition to an ensemble of consistent microscopic configurations; (iii) evolution over the same (short) time period of each initial microscopic configuration in the ensemble according to a microscopic simulator that embodies the best current description of the physical system; (iv) averaging ("restriction") over the ensemble of the evolved microscopic configurations to provide a macroscopic evolved system state; and (v) execution of the previous three steps over a finite set of macroscopic initial conditions. This new approach is robust in its implementation and portable in its range of scientific and engineering applications. It has general applicability to all systems for which a macroscopic description is conceptually possible, yet unavailable in closed form. It circumvents the difficulty in obtaining and closing such macroscopic models, while computationally extracting precisely the information that would be obtained by a macroscopic model, had the model been available in closed form. This provides the link between ITR and a spectrum of application areas.Impact:The impact of the research will be on establishing a powerful and general link between state-of-the-art microscopic-level simulations and fast systems level analysis capabilities. Although the research focuses on specific problems in heterogeneous hard materials and complex fluids, the computational framework is applicable to a broad range of complex systems, including biological systems, their processing and function. Since it has the potential to revolutionize engineering systems-level analysis, it could have educational impact as well as furthering advances in microelectronics, bioinformatics and nanotechnology.