HPC: ASTROPHYSICS AND COSMOLOGY
The universe in a C Shell
What simulation could be more impressive than that of the entire universe? With only four per cent of the cosmos even visible to observers, and 13.7 billion years of history behind us, there is much for these simulations to show, as Stephen Mounsey discovers
‘S
pace is big. Really big. You just won’t believe how vastly hugely mind-bogglingly big it is…’ So
wrote Douglas Adams in The Hitchhiker’s Guide to the Galaxy. As a species, we’ve managed over the last 1,000 years or so to scratch together a detailed working theory of the cosmos, but in a universe this big, there’s much left to learn. At the forefront of astrophysics are groups of researchers making full use of some of the fastest HPC installations with which this small planet can provide them. The length and time scales that such simulations are run at vary by many orders of magnitude, from groups of atoms to billions of light-years, depending on exactly what it is that is to be modelled by each group. Working at the entire-universe end of the scale is Dr Kentaro Nagamine, an assistant professor of astrophysics the University of Nevada, Las Vegas. His group looks at the formation and evolution of galaxies, from the beginning of the universe, 13.75 billion years ago, up to and beyond the present. He is currently preparing code to be run on Blue Waters, the petascale supercomputer under construction in Chicago – expected to become the world’s fastest supercomputer when it comes online next year. Nagamine explains what access to more computing power will mean for his simulations: ‘Dynamic range is the fundamental problem of cosmology; we want to be able to simulate from large scale to small scale, all at once. The universe is big, and we are small. If we want to simulate from the early universe to the late universe, and simulate that evolution as
32
a function of time, then we need to simulate from millions of light-years down to the astronomical unit scale – the planetary scale,’ he explains, noting that an astronomical unit (or AU) is the distance from the Earth to the Sun – about 93m miles, or 8.3 light- minutes. ‘That’s a significant dynamic range in both space and time, and by using a larger computer we can cover a larger dynamic range than we do currently, allowing us to generate a more consistent picture across all scales at once. Even with Blue Waters, we’ll never reach the AU scale – the one-planet scale. We hope at least to reach the scale of single galaxies, so that we can start to address the morphology of galaxies – observing how disk galaxies form, for example.’ During the expansion of the universe,
matter clumped together to form galaxies and galaxy clusters, a phenomenon Nagamine hopes to replicate in silico. He explains that there are three approaches to modelling the expanding universe: the simplest option, he says, is an Eulerian mesh, which expands uniformly with the universe. ‘The problem here is that as structures and gasses form [as the universe develops], we want to zoom in on those regions,’ he says. The second, the adaptive mesh refinement (AMR) technique allows this ‘zooming-in’ to interesting features, increasing computational efficiency by disregarding empty or irrelevant areas of the mesh. The third option is smooth particle hydrodynamics (SPH), in which the gasses of the universe are modelled as particles. ‘There are good and bad points about each of these codes; the trade-off is between the
SCIENTIFIC COMPUTING WORLD AUGUST/SEPTEMBER 2010
accuracy and the efficiency of the code. With an AMR code you can do a more accurate job of resolving developments, but it will be slower than some more-efficient SPH codes,’ explains Nagamine, whose ENZO code is based on an AMR technique. Also preparing the ENZO code for Blue
Waters is Brian O’Shea of Michigan State University. O’Shea points out that simulating a single galaxy is not that useful. With one hundred trillion galaxies in the universe, he says, there is a galaxy out there somewhere to match any simulation. ‘We’re not looking at how one galaxy evolves over time, we’re looking at tens of thousands of galaxies at a time; we see population changes, but any given galaxy does not change in a human lifetime.’ O’Shea’s simulations use a ‘periodic box’, in which objects leaving the box at one side come back into it in the other side, with dimensions of up to two billion light-years.
www.scientific-computing.com
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48