ITR/(DMS): A Computational Environment for Multi-scale MHD Turbulence Studies

磁流体动力学（MHD）湍流模型在地球和工程应用（如聚变装置和等离子体推进）以及空间和天体物理环境（如太阳风和天体物理喷流）中变得越来越重要。 等离子体流体的磁流体力学性质对所涉及的数值程序施加了约束。 同样地，在湍流状态下的建模--因为湍流是普遍存在的--也对所采用的数值方法施加了限制和挑战。 理论方法和对计算时间和精度的要求转化为算法（或软件）的进步，这有助于硬件本身的快速发展。 在这个项目中，这些进步包括：自适应细化网格，其中仅在感兴趣区域的精细分辨率降低了整体CPU需求; MHD冲击捕获方法，提高了计算的准确性;并行计算技术，利用多处理器和存储器的可用性和高速缓存层次结构;和子网格模型，其中比湍流区域更粗糙的网格。 分析产生的大型多块多处理器数据需要先进的数据可视化工具来检查物理，计算的理由。 该项目将开发一个计算环境，包括具有自适应网格细化能力的三维MHD模拟代码，复杂的时间依赖边界条件，并优化为在各种并行计算机上运行的MPI下，以及可以处理多块多处理器MHD AMR数据的可视化系统。 从软件工程原理产生健壮的代码，到AMR所需的更智能的数据结构，到并行架构上的单节点优化和更好的通信策略，再到可视化工具的开发和实现，IT使通过模拟以更有效和有序的方式发现知识成为可能。信息技术的效率转化为科学或技术的能力。即，如果没有这些信息技术原则的应用，就不会分析物理上重要和有趣的制度。 该项目的成果将显著提高对不同流动的模拟能力，并将具体应用于日球磁场结构和太阳风湍流问题。 3DMHD模拟的日光层场和流动，包括旋转，剪切，波，湍流，电流片，压力平衡结构，和相互作用的区域，从而代表了大多数的主要物理效应的重要性，将被执行。 这项工作将使未来的研究粒子传播和磁场配置的“空间天气”对地球的影响。 该项目将直接培养一名博士后学者和一名研究生，从事相关的物理学和计算数学;项目结果将通过会议和期刊出版物和网络以及PI教授的课程向社区传播。

Modeling MHD (magneto-hydrodynamic) turbulence has become increasinglyimportant both in terrestrial and engineering applications, such asfusion devices and plasma propulsion, as well as in space andastrophysical environments such as the solar wind and astrophysicaljets. The magneto-hydrodynamic nature of the plasma fluid imposesconstraints on the numerical procedures involved. Likewise, modeling inthe turbulent regime -- for turbulence is ubiquitous -- also imposesconstraints and challenges on the numerical procedures employed. Thenumerical approach and demands of computation time and accuracytranslate into algorithmic (or software) advances which can avail thehardware -- itself advancing rapidly. In this project these advancesentail: Adaptively refined grids, where fine resolution only in regionsof interest reduces overall CPU requirements; MHD shock capturingmethods, which improve accuracy of the calculation; parallel computingtechniques, which take advantage of the availability of multipleprocessors and memory and cache hierarchy; and sub-grid models, whichallow coarser grids than otherwise for turbulent regimes. The analysisof the resulting large multi-block multi-processor data requiresadvanced data visualization tools to examine the physics, the raisond'etre for the calculations. The project will develop a computationalenvironment, consisting of a 3D MHD simulation code with adaptive gridrefinement capabilities, complex time-dependent boundary conditions andoptimized to run under MPI on various parallel computers, and avisualization system that can handle multi-block multi-processor MHD AMRdata.In this end-to-end modeling and discovery process, informationtechnology (IT) plays a critical role. From software engineeringprinciples that result in robust codes, and smarter data structures thatare required for AMR, to single-node optimization and bettercommunication strategies on parallel architectures, to visualizationtool development and implementation, IT makes possible knowledgediscovery through simulation in a more efficient and orderly fashion.The requirements for MHD turbulence modeling being so considerable, thisefficiency of IT translates into enabling the science or technology.I.e. without the application of these IT principles, physicallyimportant and interesting regimes would not be analyzed. The results ofthis project will significantly enhance the modeling capabilities ofdiverse flows and will specifically be applied to the problem ofheliospheric magnetic field configuration and solar wind turbulence. 3DMHD simulations of the heliospheric fields and flows that includerotation, shear, waves, turbulence, current sheets, pressure-balancedstructures, and interaction regions, which thus represent most of themajor physical effects of importance, will be performed. The effort willenable future studies on particle propagation and magnetic fieldconfiguration for "space weather" effects on Earth. The project willdirectly train a post-doctoral scholar and a graduate student inrelevant physics and computational mathematics; project results will becommunicated to the community via conference and journal publicationsand the web, and through classes taught by the PI.