NSF-EMT: A Merging Multi-scale Model for Simulations of Crystallization/Solidification of Nanostructured Materials on Large-scale Parallel Computing Systems

为了了解特定材料的微观结构和最终性能，对材料的结晶/凝固进行建模和模拟，在当今材料科学和工程中对新的纳米材料和器件的探索方面取得了重大进展。在过去的十年里，已经发展了许多计算模型来部分地模拟微观或宏观尺度上的结晶和凝固。然而，由于微观(分子、介观)计算在多个尺度上的复杂性和计算能力的有效性，还没有开发出一个通用的模型来发展微观(分子、介观)计算。本研究涉及利用网络基础设施支持的计算资源和设施，开发一个包括大规模微观和宏观纳米结构材料结晶/凝固模型的计算基础设施，以及相关的高性能计算算法和数值方法。研究人员将开发一个集成的软件包，该软件包可以系统地解决纳米结构材料形成的建模和模拟中的许多挑战。该模型从原子/分子相互作用出发，采用平行分子动力学(MD)方法，考虑了纳米级结晶、纳米粒子/晶体的形成和生长机制。研究人员使用热力学和平衡理论来解释微观液态析出和/或偏析、局部热和特殊的非平衡、微结构转变、粒子和相互作用以及相变。他们还在宏观层面上考虑了宏观输运现象和各种缺陷。计算基础设施模式完全在于国家网络基础设施孕育出的大规模并行计算技术。模型计算的大量任务分布在通过高速网络连接的多个计算平台上，而每个任务执行一次并行计算，从而增加了对具有挑战性的问题的计算能力。

Modeling and simulations of crystallization/solidification of materials for understanding a given material's microscopic structure and final properties have spawned significant advances in today's explorations of new nano-materials and devices in materials sciences and engineering. In the past decade, many computational models have been developed to partially simulate the crystallization and solidification on either microscopic or macroscopic scales. However, a generalized model to develop microscopic (molecular, meso-scale) calculations with macroscopic computations has not been developed, due to the model's complexity on multiple scales and validity of computational power.This research involves developing a computational infrastructure that includes a large-scale microscopic and macroscopic model of crystallization/solidification of nanostructure materials, and associated high performance computational algorithms and numerical methods, using the cyber-infrastructure-enabled computing resources and facilities. The investigators will develop an integrated software package which can systematically solve many challenges in the modeling and simulation of nanostructure material formation. The model begins with the atomic/molecular interactions for the origin of nucleation by using parallel molecular dynamics (MD), and considers the mechanisms of nano-scale crystallization, nano-particle/crystal formation and growth. The researchers employ thermal dynamics and the equilibrium theory to account for the microscopic liquid-state precipitation and/or segregation, local thermal and special non-equilibriums, microstructure transition, particles and interactions, and phase change. They also consider the macroscopic transport phenomena and various defects at a macroscopic level. The computational infrastructure model fully lies in the large-scale parallel computing technologies bred by national cyber-infrastructure. Numerous tasks for the model computation are distributed across multiple computing platforms connected by high-speed networks, while each task executes a parallel computation; thus increasing computational power for the challenging problem.