多行星系统中的时间可逆时间步长算法
Time-reversible timestepping in many-body planetary systems
AI总结:
本研究开发了几乎完全时间可逆的直接N体积分器MTR,可实现随系统演化的即时个体时间步长自适应,在行星系统模拟中相比GPLUM代码的Hermite模式加速达两个数量级,默认配置下速度优于树模式。
摘要:
模拟混沌行星系统演化需要解析因近距离相遇、散射、碰撞等产生的小时间尺度。通常模拟会为所有粒子分配固定时间步长,但在混沌演化系统中这效率低下且浪费计算资源,还可能导致模拟解析不足。混合方法是解析小时间尺度的粗糙工具,无法实现细粒度自适应。为行星天体分配随时间变化的个体时间步长通常会破坏辛性、可逆性等理想性质。本研究开发了一种“几乎”完全时间可逆的直接N体积分器MTR,每对行星拥有各自的时间步长,该步长可随系统演化变化。与此前工作不同,本研究实现了独立于全局时间步长的即时自适应。研究提供了该代码的C语言版本,并在多个问题上进行测试,结果显示无能量误差漂移,符合可逆方法的预期。在粒子数多达N=975的情况下,将其与最先进的GPLUM代码对比,GPLUM可用于直接Hermite模式(未针对该模式优化)或树模式(已针对该模式优化)。与Hermite模式相比,MTR的加速比高达两个数量级,且轨道分布一致;在默认配置下,尽管树模式的力计算更快且近似,MTR的速度与树模式相当或更优,加速比可达一个数量级。
英文摘要:
Simulating chaotic planetary system evolution requires resolving small time-scales due to close encounters, scattering, collisions, and more. Simulations typically assign fixed timesteps to all particles, however, which is inefficient and wastes compute resources in a chaotically evolving system; moreover, it may under-resolve the simulation. Hybrid methods are a blunt tool for resolving small time-scales as they do not allow for fine-grained adaptation. Assigning individual timesteps, which vary in time, to planetary bodies usually violates desirable properties such as symplecticity and reversibility. In this work, we develop an "almost" exactly time-reversible direct $N$-body integrator, MTR, in which each pair of planets gets its own timestep, which itself can vary as the system evolves. In contrast to our previous work, we achieve this adaptation on the fly, independent of a global timestep. We provide a C version of the code and test it on several problems, demonstrating no energy error drift, as expected of a reversible method. At large particle numbers, up to $N = 975$, we compare against the state-of-the-art GPLUM code, which we can use in direct Hermite mode (for which it is not optimized) or tree mode (for which it is). MTR provides a speedup of up to two orders of magnitude compared to the Hermite mode, while agreeing with the distribution of orbits. In our default configuration, MTR's speed matches or improves on the tree mode's speed, by up to an order magnitude, despite the tree mode's faster and approximate force calculation.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。