Supercomputer Helps to Clear Up Big Bang Chemical Conundrum
For roughly two decades, scientists have struggled to understand the ratio of lithium isotopes found within the bellies of the oldest stars in the universe. Now, with the aid of high-performance computing, an international group of researchers has put this cosmic problem to rest.
The team, with researchers based in Australia, Brazil, Germany, and the UK, has matched observation with theory, providing a better understanding of what happened just after the big bang, 13.8 billion years ago. To reach this solution, it took some creative thinking, starlight, and a 3D model working in parallel on supercomputing hardware at the Max Planck Institute for Astrophysics in Garching, Germany. The team published their research earlier this year in the journal Astronomy and Astrophysics.
No longer at sixes and sevens
In order to understand what the universe was like just after the big bang, scientists require a signature. This can be found in the light emitted by the oldest stars, providing a window to the early universe. Karin Lind, who is currently based at the Institute of Astronomy, University of Cambridge, UK, and who was lead author of the research paper, looked for light signals from two chemical isotopes of lithium. Isotopes are atoms which share the same name and chemical properties, but have different masses; that is, different numbers of neutrons in their core.
In astrophysics stellar data, there appeared to be more of the lithium-6 isotope (6Li), and less of the lithium-7 isotope (7Li), compared to the amounts produced in the big bang. This contradicted standard model theories of big bang nucleosynthesis (nuclear fusion), and for 20 years no one knew why. Speculation ranged from annihilation of dark matter particles called axions and neutralinos, to the acceleration of cosmic ray particles by massive stars or cosmic tidal waves.
Stepping up the calculations
After collecting 20 gigabytes of data, the research team created a multi-dimensional computer model. “To predict the synthetic spectrum from the star using more realistic physical assumptions… we performed these calculations using 3D hydro-dynamical models, instead of traditional 1D hydrostatic ones,” adds Lind.
The main challenge was accurately modeling a star’s surface convection and its interaction with the radiation field. In traditional, one-dimensional modeling, convection calculations are simplified and do not require powerful computational power.
However, Lind and her colleagues took into account interactions with the strong underlying radiation field, which are usually ignored. The models were run as parallel codes, on a supercomputer installed at the Max Planck Society computing center (RZG) in Garching, Germany. In total, this IBM system has 792 CPUs, 6 gigabytes of RAM per CPU, and a complete storage capacity of 88 terabytes.
The simulation code was written in the Fortran 90 programming language to enable the calculations to be run in parallel on the computer cluster’s multiple cores.
These simulations took several weeks. “The main bottle neck was internode communication, which is usually the most [computationally] expensive part of supercomputers. For example, running a simulation on the whole cluster is rather inefficient, while running 12 simulations in parallel is much more efficient,” says Magic.
The simulations showed that that none of the stars displayed a substantial presence of 6Li, contrary to previous scientific evidence. Gary Steigman, an expert in big bang nucleosynthesis at Ohio State University, US, who was not involved in the research, says that of all the lithium problems in early universe physics, the 6Li issue required modeling at an exquisite level of precision. “This is what the authors of this paper claim to have done,” says Steigman. “The evidence for the presence of 6Li has now disappeared; I think this is an important and interesting result. My only caveat is that others, such as Roger Cayrel and his research collaborators, have said this before and might be tempted to say: ‘I told you so’.”
Advanced code, but not advanced enough
The next step for the research team is to update their astrophysics code to extract more information from ancient stars, such as the abundance of iron. “Iron lines are currently too computationally demanding to model with our method. This is important to reduce the error bar of [our lithium isotope] analysis and make sure we fully understand the galactic production of 6Li, which becomes more significant at later times in the history of the universe,” says Lind. The team will also go back to looking at 7Li, because the amount observed in stars falls short of current big bang nucleosynthesis predictions, too. -- by Adrian Giordani, © i SGTW