Oscar Agertz on the Comparison of Hydrodynamical Codes

Emerging Research Fronts Commentary, August 2011

Oscar Agertz

Article: Fundamental differences between SPH and grid methods

Authors: Agertz, O, et al.
Journal: MON NOTIC ROY ASTRON SOC, 380 (3): 963-978 SEP 21 2007
Addresses: Univ Zurich, Inst Theoret Phys, CH-8057 Zurich, Switzerland.
Univ Zurich, Inst Theoret Phys, CH-8057 Zurich, Switzerland.
ETH, Inst Astron, Dept Phys, CH-8093 Zurich, Switzerland.
Nicholas Copernicus Astron Ctr, PL-00716 Warsaw, Poland.
Univ Chicago, Dept Astron & Astrophys, Chicago, IL 60637 USA.
Univ Copenhagen, Niels Bohr Inst, DK-2100 Copenhagen, Denmark.
Univ Nottingham, Sch Phys & Astron, Nottingham NG7 2RD, England.
Univ Valencia, Dept Astron & Astrophys, E-46100 Burjassot, Spain.
(Addresses have been truncated)

Oscar Agertz talks with ScienceWatch.com and answers a few questions about this month's Emerging Research Front paper in the field of Space Science.

SW: Why do you think your paper is highly cited?

It was a very timely publication for the astrophysical community. The field is today, more than ever, strongly influenced by numerical simulations and many advanced simulation codes are publicly available. We managed to demonstrate that different numerical methods are unable to converge on analytically expected results, and could pinpoint why this is so. This has led to plenty of development in the field, including several new algorithmic methods, as well as a greater awareness on how to interpret simulation results.

Validating scientific methods is often of little direct interest to the general public. However, this type of work is a great service to the scientific community, as it exposes weaknesses and drawbacks in currently used methods. This spurs development, often leading to improved tools for the community.

SW: Does it describe a new discovery, methodology, or synthesis of knowledge?

Some of the things we discovered had been known in the astrophysical community, but not published and discussed in a concise way. Similar discoveries had also been made in the engineering community in various contexts. In essence, we presented an old lingering issue in a rather dramatic and clear way, improved on previous analysis, and exposed drawbacks of specific numerical methods.

SW: Would you summarize the significance of your paper in layman's terms?

This paper concerned the way in which we simulate the dynamics of astrophysical fluids. So why is this important, you may ask. Shortly after the Big Bang, all visible matter existed as a primordial gas, or plasma to be precise, consisting almost solely of ionized hydrogen and helium, with some small amounts of heavier elements. It is the subsequent cosmic evolution of this gas, via gravitational and gas dynamical process, that ultimately leads to the formation of galaxies, stars, and planets.

We know the equations that govern these processes, although they must be solved numerically via powerful supercomputers. There are many different ways to translate well defined analytical equations into a powerful numerical scheme, which is why code comparisons must be carried out to make sure that they all agree and are doing what they are supposed to do.

We considered a simple test problem for computer codes; a wind tunnel experiment. A dense cloud of gas was exposed to a strong wind of hot gas. This setup is known to be unstable and the cloud should be destroyed as the hot wind forces the cloud to break up into small pieces and ultimately mix with the ambient medium (see image below). This numerical experiment was then run with a myriad of codes commonly used in the community.

A snapshot from the numerical simulation described in the paper. As the surrounding flow is supersonic, a pronounced bow shock has developed in front of the cloud. The smooth flow quickly becomes very turbulent as fluid instabilities on the initially spherical cloud's surface grow stronger. This process destroys the cloud and mixes it with the ambient medium.

A snapshot from the numerical simulation described in the paper. As the surrounding flow is supersonic, a pronounced bow shock has developed in front of the cloud. The smooth flow quickly becomes very turbulent as fluid instabilities on the initially spherical cloud's surface grow stronger. This process destroys the cloud and mixes it with the ambient medium.

As it turns out, the results were very different among the methods. We then simplified the analysis and were able to pinpoint why these differences occurred and why some methods were highly inappropriate to use in certain astrophysical scenarios.

SW: How did you become involved in this research, and how would you describe the particular challenges, setbacks, and successes that you've encountered along the way?

I was introduced to this project by Prof. Ben Moore early on in my Ph.D. years at the Institute for Theoretical Physics at the University of Zurich. In the fall of 2004, one year before I started my Ph.D., a workshop on numerical astrophysics was organized in the beautiful Wengen in the Swiss Alps. The test problem mentioned above was part of a large code comparison suite discussed during the workshop. No one had properly looked into the results afterwards and they were simply left on the Institute's hard drives for more than a year.

When you dig into a project like this, you are initially faced with an enormous amount of unstructured data in various formats. It was very time consuming to go through this to get a clear picture of the issue at hand. After a while I realized that it was time better spent to repeat the majority of the code comparison by myself. This is when I started seeing the systematic differences presented in the paper. Moving away from data mining to actually carrying out my own experiments, and in addition constructing new and simpler ones, made it relatively easy to pinpoint the underlying problem.

In a project like this, with many strong minds contributing along the way, it can often be difficult to come to a consensus and get the scientific results into publishable shape. However, this was never an issue and we could finish the project on a relatively short timescale. 

SW: Where do you see your research leading in the future?

As I mentioned above, this research has led to many improvements on existing methods, as well as plenty of new ingenious ones. To be confident in the accuracy and validity of simulation results is always extremely important, and I believe the community is nowadays more critical when drawing conclusions from simulations than before.

At the moment, my collaborators and I use high-resolution numerical methods, some of which are discussed in the paper, to understand how galaxies form. Currently we are investigating how more complex physics, such as the radiation from massive stars, influence galactic evolution and star formation. Future work by me and my colleagues will hopefully lead to a better understanding of galaxy formation.

SW: Do you foresee any social or political implications for your research?

Our research has a very different social impact than what is expected of, e.g., medicine or biochemistry. The ultimate goal is to understand the formation and evolution of galaxies, stars, and planets. Our research has in this sense a massive social impact, as it aims to understand our origins and our place in the universe.End

Dr. Oscar Agertz
KICP Fellow
Kavli Institute for Cosmological Physics
University of Chicago
Chicago, IL, USA



   |   BACK TO TOP