3D NLTE Radiation Transport with PHOENIX/3D Gauss Centre for Supercomputing e.V.

ASTROPHYSICS

3D NLTE Radiation Transport with PHOENIX/3D

Principal Investigator:
Peter Hauschildt

Affiliation:
Hamburger Sternwarte, Universität Hamburg (Germany)

Local Project ID:
PP14112588

HPC Platform used:
Hazel Hen of HLRS

Date published:

Understanding the light emitted by (magnetically) active cool stars (‘M dwarfs’) is a major challenge for astrophysics. In this project, scientists use their PHOENIX/3D code to simulate the light emitted by a ‘box’ inside the outer layers of an active M dwarf in detail. The temperatures and pressures inside the box are taken from an existing gas dynamics simulation (including magnetic field effects) by S. Wedemeyer (Oslo). The computational requirements of detailed non-equilibrium 3D radiative transfer simulations are staggering and require the largest supercomputers on Earth.

Over the last decade, a number of detailed 3D simulations of the gas dynamics in the outer layers of stars have become feasible. The most recent of these simulations include also the effect of magnetic fields on the gas (and vice versa), such calculations allow us to investigate the time dependent behavior of the outer layers of active cool stars (“M dwarfs”), called the photosphere and the chromosphere. In order to ‘connect’ such gas dynamic simulations with the observable light from the star, a very complex non local thermodynamic equilibrium (NLTE, i.e., radiation and gas are not in equilibrium) radiative transfer simulation is necessary, requiring a multiple of the computing power for gas dynamics simulations.

The PHOENIX code used in this project is a general-purpose model atmosphere simulation package. It is designed to model the structures and spectra of a wide range of astrophysical objects, from extrasolar planets (both terrestrial and gas giants) to brown dwarfs and all classes of stars, extending to novae and supernovae. The main results from the calculations are synthetic images and spectra (and derived quantities, such as colours), these can be directly compared to observed spectra and, in 3D simulations, images.

In this project, the researchers have used (and further developed) PHOENIX/3D to calculate the light emitted by a snapshot of a high-resolution gas dynamics simulation of a ‘box’ inside an active M dwarf.

The computing power and time requirements for a realistic PHOENIX/3D model are staggering. This is due to the fundamentally 7D nature of the radiative transfer modelling (for a 4D space-time). For the simulations at HLRS (on HPC system Hazel Hen) a typical full iteration with 50,000 MPI processes and 2-4 OpenMP threads per process requires approx. 3 hours, and about 100-200 iterations are needed for convergence. Thus, about 30 million core hours are needed to compute a single(!) PHOENIX/3D model for a high-resolution gas dynamics snapshot. A minimum of 50 GB of output are generated per iteration, and the final output required for the generation of images and observables is well in excess of 1 TB, taxing the capabilities of current parallel file systems.

The results allow the researchers to better interpret and to understand the observations of active M dwarfs, to ‘calibrate’ much ‘cheaper’ 1D simulations and thus to extract the maximum knowledge from observations obtained with current and next generation ground- and space-based telescopes.

Scientific Contact:

Peter H. Hauschildt 
Hamburger Sternwarte 
Gojenbergsweg 112, D-21029 Hamburg (Germany)
e-mail: yeti [@] hs.uni-hamburg.de
http://www.hs.uni-hamburg.de/~stcd101/

Tags: Universität Hamburg Astrophysics HLRS