Milky Way-sized FDM Halo Construction by GPU Accelerated Code

Guan-Ming Su^2*, Hsi-Yu Schive^1,2,3,4

¹Institute of Astrophysics, National Taiwan University, Taipei 10617, Taiwan
²Department of Physics, National Taiwan University, Taipei 10617, Taiwan
³Center for Theoretical Physics, National Taiwan University, Taipei 10617, Taiwan
⁴Physics Division, National Center for Theoretical Sciences, Taipei 10617, Taiwan

* Presenter:Guan-Ming Su, email:d07222009@ntu.edu.tw

We report a new fuzzy dark matter (FDM) halo construction code based on the algorithm proposed by Lin et al. 2018. Our code features multi-GPU acceleration, mesh refinement, and tunable halo mass M_h and FDM particle mass m₂₂. It makes it possible, for the first time, to construct a Milky Way-sized FDM halo with M_h=10¹² M_☉ and m₂₂=0.8, which is extremely computationally expensive since the computing time scales as M_h⁴ m₂₂⁶ . We achieve it using 32 V100 GPUs for four days on the Taiwania II supercomputer at NCHC, which would otherwise take approximately one year with 16 CPU cores but without GPU. We have confirmed the following features of this halo: (i) stability (using the simulation code GAMER); (ii) virialization; (iii) well-resolved granular structure; (iv) soliton density oscillations and random walk; (v) monotonically increasing correlation time with radius. Furthermore, we discover a long-period, Keplerian-like soliton motion, which overwhelms the soliton Brownian motion in low-mass FDM halos with M_h~10¹⁰ M_☉. The capability of constructing Milky Way-sized FDM halos paves the way to exciting research topics, including, for example, the interplay between soliton and central molecular zone, bar slow-down in FDM scenario, and tidal stripping and dynamical friction of FDM subhalos.

Keywords: fuzzy dark matter, Milky Way-sized halo, soliton, GPU