Collabortive Research: Petascale Simulations of Merging Black Holes and Neutron Stars

该研究的主要科学目标是利用 Blue Waters 从理论上支持和提高 NSF LIGO 实验从观测到的引力波中提取引力波携带的丰富信息的能力。 更一般的目标是促进我们对相对论天体物理系统的理解，并有助于开发能够对高度动态弯曲时空进行稳健和准确模拟的数值方法和计算机代码。 该计划还将作为年轻物理学家和天体物理学家的培训基地。它将教他们各种各样的计算技术。该研究将包括开发一种名为“幽灵”的下一代超级计算机代码，该代码将使用全新的技术。 Spectre 将作为开源发布，模拟输出将公开。这将使其他研究能够从这项工作中受益。小组成员将积极开展广泛的活动，接触更广泛的科学界和公众。该研究将由两个部分组成。首先，研究人员将使用光谱爱因斯坦代码（SpEC）在 Blue Waters 上进行数值模拟，以产生双黑洞的引力波形。这些波形需要数周到数月的时间才能生成，然后将用于开发数值替代模型，该模型可以在几毫秒内评估单个波形，同时保留完整数值模拟的准确性；该替代模型适合直接用于 LIGO 数据分析。 第二个组件将开发并在 Blue Waters 上运行，这是用于物质和辐射数值相对论模拟的新代码幽灵。 Spectre 使用不连续伽辽金有限元方法。它遵循新颖的基于任务的并行化范例，旨在在当前和未来的超级计算机上实现高精度和高可扩展性。 SpECTRE 的初始版本已经实现了广义相对论磁流体动力学。 它将升级以处理中子星-中子星、黑洞-中子星合并和恒星塌缩的动态时空，包括基于核理论的热状态方程和中微子。 Spectre 将用于计算 Blue Waters 上的高精度超长中子星螺旋模拟，以预测引力波信号并帮助 LIGO 约束核状态方程。

The primary scientific objective of the research is to use Blue Waters to theoretically underpin and improve the ability of NSF's LIGO experiment to extract from observed gravitational waves the rich information that the waves carry. More general objectives are to contribute to our understanding of relativistic astrophysical systems, and contribute to the development of numerical methods and computer codes capable of carrying out robust and accurate simulations of highly dynamical curved space-time. This program will also serve as a training ground for young physicists and astrophysicists. It will teach them a wide variety of computational techniques. The research will include developing a next-generation supercomputer code called SpECTRE that will use radically new techniques. SpECTRE will be released as open source and simulation outputs will be made publicly available. This will enable other research to benefit from the work. Group members will vigorously pursue a wide range of activities that reach out to the broader scientific community and the general public.The research will consist of two components. In the first, the researchers will run numerical simulations on Blue Waters using the Spectral Einstein Code (SpEC) to produce gravitational waveforms for binary black holes. These waveforms, which take weeks to months to produce, will then be used to develop a numerical surrogate model that can evaluate a single waveform in milliseconds while retaining the accuracy of full numerical simulations; this surrogate model will be suitable for direct use in LIGO data analysis. The second component will develop, and run on Blue Waters, the new code SpECTRE for numerical relativity simulations with matter and radiation. SpECTRE uses Discontinuous Galerkin finite element methods. It is designed for high accuracy and high scalability on current and future supercomputers by following a novel task-based parallelization paradigm. SpECTRE's initial version already implements general-relativistic magnetohydrodynamics. It will be upgraded to handle the dynamical spacetimes of neutron star-neutron star and black hole-neutron star mergers and stellar collapse, including nuclear-theory based hot equations of state and neutrinos. SpECTRE will be applied to compute high-accuracy ultra-long neutron star inspiral simulations on Blue Waters to predict gravitational wave signals and help LIGO constrain the nuclear equation of state.