Towards High-Mass Star Formation: Confronting Simulations and Observations

Principal Investigator:
Prof. Dr. Stefanie Walch-Gassner

Physikalisches Institut, Universität zu Köln, Cologne, Germany

Local Project ID:

HPC Platform used:

Date published:


Giant clouds of molecules and dust are the birthplace of stars. Once certain conditions are met, fragments of these clouds begin to contract under the force of gravity. They begin to form compact regions of dense gas which will eventually evolve into so-called protostars. Computer simulations of these formation processes help to gain a better understanding of the highly complex dynamics within the stellar nurseries. High-mass star formation (stars with a mass larger than eight times the mass of the sun (>8Msun)) is particularly challenging to model due to highly energetic feedback effects. These include, but are not limited to, radiative heating and radiation pressure, thermal pressure, and ionizing radiation. Also, the clouds are often strongly magnetized.  It is difficult to construct models that are numerically stable when coupling these processes, and it is computationally expensive to run them.


Our simulations were performed with the (magneto-)hydrodynamic code FLASH which is constantly updated and expanded by the Cologne Theoretical Astrophysics group so that it is capable of tackling the complex physical processes necessary to model high-mass star formation. To realize this, the supercomputer JUWELS was used. As an example, we show a volume rendering of a simulation in Fig. 1.

Each simulation requires a computational time of around 1-2 Mio. core hours and uses on average 500 cores (and up to 1200 cores) simultaneously. The simulations produce around 800 files each requiring a disk space of 80 TB.

Simulation Details:

The initial conditions of the simulations, like the distribution and velocity of dense cloud material as well as the amounts of heavier elements in the molecular gas, are motivated by real observations. Several simulations with different initial conditions, how scientists expect them to be in high-mass star-forming regions, were performed. The ultimate goal of these simulations is to produce a set of synthetic observations that may be compared to real observations made with instruments like ALMA. Additionally, the influence of the feedback from massive stars, which can be turned on and off in the simulation is investigated.


We start from a spherical cloud that begins to collapse under its own gravity. Substructures that look like filaments emerge in the process due to turbulence in the cloud. Stars formation first takes place within the central region of the simulated box, but later on extends to the outer regions of the filaments as well. The number of stars which eventually form in each simulation depends on the initial conditions. In the shown simulation (see Fig. 2), 19 stars are formed, while 10 of them grow into massive ones. Each box contains a predefined amount of mass. How much of it is converted to stars, the so-called star formation efficiency, also depends on the initial conditions.

The ionizing feedback from massive stars leads to the creation of a bubble full of ionized hydrogen (see red color in Fig. 1). The pressure transferred from stellar radiation helps the bubble to grow. As a consequence, atomic hydrogen is expelled outwards radially and the further accretion of mass onto the stars is prevented.

From these simulations, synthetic observations are derived (see Fig. 3). The density profile is used to produce the emission of the dust with the radiative transfer code RADMC-3D (see Fig. 3, middle panel).  Furthermore, the software called CASA makes it possible to simulate the effects of the instrumental limitations of the ALMA telescope to produce an image of a synthetic observation (see Fig.3, right panel). The imperfect resolution of the telescope leads to a loss of the filamentary substructures so that only the brightest cores are still detectable. The comparison between simulations and synthetic observations supports the interpretation of real-world telescope data and may guide future observations of star-forming regions.


Klepitko A., Walch S., Wünsch R., Seifried D., Dinnbier F., Haid S., 2023, MNRAS, 521, 160,

Zimmermann B., Walch S., Clarke S.D., Klepitko A., Wünsch R., submitted to MNRAS in 2023, revision in process.


Andre Klepitko, PhD thesis: Tree-Based Radiative Transfer of Diffuse Sources: A Novel Scheme and its Application in Massive Star Formation,