Collaborative Research: ITR: Global Multi-Scale Kinetic Simulations of the Earth's Magnetosphere Using Parallel Discrete Event Simulation

该项目的重点是开发基于离散事件模拟（DES）的可扩展、并行数值模型，用于模拟空间等离子体和地球磁层动力学。 太阳风与磁层的相互作用等现象是大尺度问题，其中包含小尺度动力学占主导地位的区域。 这些次区域的位置和规模随着时间的推移而变化，次区域本身的特点是主导性的物理平衡，与控制大规模动态的物理平衡不同。 在描述太阳风与地球磁层相互作用的问题中存在的制度的例子包括由电子动力学、离子物理学和磁流体力学控制的制度。 这样的问题的完全粒子模拟是不切实际的，因为需要大量的计算资源。 研究人员将开发DES方法与情况相关的物理，适合空间物理问题，然后开发所需的算法，以执行这些大规模并行计算机系统上有效地。 这将涉及研究这些方法的稳定性和准确性，现有的一维原型扩展到3D，优化并行环境，连接不同的物理平衡占主导地位的区域的方式，自适应算法，以及评估不同的方法来网格设计。 基于事件的模拟本质上是动态的，这使得编译时优化变得困难;所提出的反向执行方法使自动优化的任务进一步复杂化。 该项目将包括研究编译器技术，以改进基于事件的代码的自动优化和开发优化的可逆原语库。 这类研究可能导致空间气象建模能力的提高，并有可能对卫星通信领域产生重大影响。 要开发的建模技术可能是有用的，在其他领域的科学和工程中，多尺度，多物理机制的模拟是必需的。 该项目包括为一些研究生提供研究培训机会。

This project focuses on the development of scalable, parallel, numerical models for the simulation of space plasmas and the dynamics of the Earth's magnetosphere, based on discrete event simulation (DES). Phenomena such as the interaction of the solar wind with the magnetosphere are large-scale problems that contain regions where smaller-scale dynamics dominates. The location and scale of these sub-regions evolve in time and the sub-regions themselves are characterized by dominant physical balances different to those controlling the larger-scale dynamics. Examples of regimes present in the problem of describing the interaction of the solar wind with the Earth's magnetosphere include ones controlled by electron kinetics, by ion physics, and by magnetohydrodynamics. Full particle simulation of such a problem is impractical because of the amount of computing resources that would be required. The investigators will develop DES methods with situation-dependent physics, suitable for space physics problems, and then develop the algorithms required to execute these efficiently on massively parallel computer systems. This will involve research into the stability and accuracy of these methods, the extension of existing one-dimensional prototypes to 3D, optimization for parallel environments, ways of connecting regions in which different physical balances predominate, adaptive algorithms, and the evaluation of different approaches to mesh design. Event-based simulations are inherently dynamic in ways that make compile-time optimization difficult; the reverse-execution approach proposed further complicates the task of automatic optimization. The project will include research on compiler technology to improve the automatic optimization of event-based codes and the development of libraries of optimized reversible primitives. This type of research may lead to improvements in the ability to model space weather with the potential for significant impacts in the field of satellite communications. The modeling techniques to be developed might be useful in other areas of science and engineering where simulations of multi-scale, multi-physics regimes are required. The project includes research training opportunities for a number of graduate students.