Formation of a Multiple-Star System

It is known that the majority of Sun-like stars are born as part of multiple-star systems. Recent studies indicate that turbulence in natal molecular clouds controls the multiplicity of newborn stars: whether stars form as multiple-star systems or as single stars like the Sun. In this simulation, we follow the formation process of multiple-star systems on various spatial scales. We start from an elongated molecular cloud, which fragments to form dense cloud cores, and multiple stars form therein. By the end of the simulation, a triple protostar system forms. The triple stars disturb the surrounding gas, creating complex structures. Long arc-like structures are particularly prominent, which are beginning to be found with ALMA.

The molecular cloud here has an elongated shape with a width of 0.1 pc (0.3 light-years), and a length of 1.5 pc (5 light-years). The color represents the density of the gas. The molecular cloud is assumed to have turbulence at the initial stage. The turbulence disturbs the molecular cloud, producing fluctuations in the gas. The dense parts of the cloud attract the surrounding gas through gravity.
The molecular cloud fragments gather the gas to form dense gas condensations. These gas condensations are called molecular cloud cores. We will magnify the central part of one of them.
The central part of the molecular cloud core has a complex density structure because of the turbulence. The gas falls toward the center due to gravity. During the infall, the turbulence disturbs the infall motion of the gas, causing rotation at the center of the cloud core.
The molecular cloud core fragments to form four protostars. The protostars are shown as the red glowing spheres. Among the four protostars, two protostars merge just after their formation, leaving three surviving protostars. At first the protostars exhibit chaotic orbital motions, but they gradually settle into stable orbits.
The protostars increase their masses by accreting the gas from the envelope. The triple system consists of a close binary pair and a distant star, which is surrounded by a large gas disk.
The accretion rates onto the stars of the close binary pair are low because the stars interfere with each other’s disks. On the other hand, the distant star accretes the gas at a high rate through its surrounding massive disk.
Here we stop the evolution of the system and look at the gas structures. The triple protostars disturb the envelope gas to form complex structures around them. Long arc-like structures with a 10,000 au length are particularly prominent. Recent ALMA observations detected similar structures, suggesting that such arc-like structures are evidence that the stars formed in a highly dynamical process, as shown in this simulation.
Acknowledgment: This simulation was partially supported by JSPS KAKENHI Grant Numbers 26400233, 26287030, 24244017, 23540270, 23403001, 23244027, 23103005, and 22244014.

Details of Numerical Simulation

PurposeStar formation in a turbulent molecular cloud, and reproducing the ALMA observations toward the molecular cloud core MC27/L1521F.
Calculation CodeSFUMATO
Calculation MethodA grid-based method with adaptive mesh refinement, multigrid method, and sink particles
ComputerCray XC30 "ATERUI"
Time Scale700,000 years (before the first star formation, scene 0:00-0:09) and 24,000 years (after the first star formation, scene 1:41)
Spatial Scale5 light-years (at the beginning), and 100 astronomical units (at highest magnification
ReserchersTomoaki Matsumoto (Hosei University)
References"An origin of arc structures deeply embedded in dense molecular cloud cores", Matsumoto, T., Onishi, T., Tokuda, K., Inutsuka, S. -I., Mon. Not. R. Astron. Soc., Vol. 449, p. L123-L127 (2015)

Details of Visualization

In this simulation, a relatively rough grid is used to represent the initial large-scale molecular cloud, while finer and finer grids are used when zooming in to the multiple-star system. Therefore, it is not possible to treat the simulation results as a simple collection of homogeneous data sets.
Thus, we have managed to use multiple data sets with different resolutions, with finer data nested in the larger and rougher data, and we can switch seamlessly between them to display the different data.


  • Simulation: Tomoaki Matsumoto
  • Visualization: Takaaki Takeda
  • Four-Dimensional Digital Universe Project, NAOJ

720p (.mp4, zip file : 196.7 MB) (.wmv, zip file : 119.2 MB)

1080p (.mp4, zip file : 362.9 MB) (.wmv, zip file : 141.7 MB) (.mov, zip file : 743.4 MB)


  • 2020.08 Version 1.0 was released.