[M-GI22] 計算科学による惑星形成・進化・環境変動研究の新展開

[MGI22-04] 大規模N体計算が切り拓く惑星形成：1億粒子、そしてその先へ

*堀 安範^1,2、押野 翔一³、藤井 通子⁴、岩澤 全規⁵ (1.自然科学研究機構　アストロバイオロジーセンター、2.国立天文台　太陽系外惑星探査プロジェクト、3.国立天文台　天文シミュレーションプロジェクト、4.東京大学　理学系研究科天文学専攻、5.理化学研究所　計算科学研究機構)

キーワード：惑星形成、N体計算、GPU

In a collisional system, close encounters play an important role in dynamical evolution. Gravitational interactions between particles undergoing a close encounter are big bottlenecks in N-body simulations because of high computational costs. In fact, a direct N-body simulation in the context of planet formation faces a wall of ten thousand particles. Toward high-resolution N-body simulations with 100 million particles and beyond, we have developed three tips to overcome this sort of big wall, implementing them into our N-body code for planet formation: (i) a tree-based hybrid N-body scheme which reduces numerical integrations of gravitational interactions among particles, PPPT method (Oshino et al. 2011), (ii) GPU clusters which allow us to handle a large number of particles, (iii) parallelization and optimization for accelerating numerical integrations, specifically, a multi-purpose platform for a parallelized particle-particle simulation, the so-called “Framework for Developing Particle Simulator” (FDPS: Iwasawa et al. 2015). In this talk, we introduce what our brand-new N-body code is like and its performance and capability. We also show our preliminary results of N-body simulations of terrestrial planet formation, using ~ 0.1-1 million planetesimals.