Fitted not long ago with 2140 processors and a near 5 terabyte memory (the equivalent to over 5000 personal computer memories), Magerit is the second most powerful supercomputer in the recently set up Spanish Supercomputing Network (RES), surpassed only by the famous MareNostrum installed at Barcelona’s National Supercomputing Centre. This incredible machine is capable of performing over 12 billion operations per second. It is then the perfect place to try to create a virtual replica of our universe.
Thanks to the installation of powerful telescopes at different sites on earth and in free space, astrophysics and cosmology have made tremendous progress over recent years, giving us a detailed picture, pieced together from information on cosmic microwave background radiation (CMB), of what the universe was like in its earliest infancy, how the primitive galaxies were formed during the universe’s adolescence, and how, in middle age, it has reached an unprecedented orderliness. We are now witnesses to the fact that the universe is entering upon an age of exponential accelerating expansion and know from this that it will have a very long and boring old age.
Even so, we still are ignorant of many stages in our universe’s long life (almost 14 thousand million years). To be able to fill in these gaps and get a fuller picture of the events that fashioned the universe as we now know it, it needs to be recreated virtually. To do this researchers have to make complicated numerical calculations in an attempt at reproducing the physical processes responsible for the formation of stars, galaxies, galaxy clusters and other structures observable through a telescope. This is a formidable endeavour that calls for tremendous computational capabilities.
This then is a mission that has to be undertaken by major international partnerships. An interdisciplinary group composed of astrophysicists from the Universidad Autónoma de Madrid, the Astrophysical Institute Potsdam in Germany and partners from Israel, the United States, Russia, Greece, etc., have joined forces to try to realistically reproduce the starting conditions that originated the observable galaxies, including the one in which we live, the Milky Way. This partnership took its name from the grand supercomputer that was the starting block for this research: The MareNostrum Numerical Cosmology Project or MNCP.
The numerical codes responsible for this task were designed to exploit the combined power of thousands of processors in Magerit and MareNostrum in Spain and other big supercomputers within the European Consortium of Supercomputing Centres. Through the use of complicated numerical algorithms, they can reproduce the gravitational and hydrodynamic processes that took place within all the constituents of the universe: the ordinary matter of which we are all made (atoms, molecules, etc.) and the famous dark matter and energy, about which know little or nothing except its quantity.
This huge computational effort generates a substantial amount of data, which, once analysed, will provide a virtual vision of how the first galaxies might have been formed, how they merged to produce larger galaxies and how these galaxies then came together to form groups and clusters of galaxies.
Like the Hollywood film studios making computer-generated movies, our researchers also have to produce a great many frames to render the universe. It is, of course, impossible to reproduce everything that happened across the known universe. The MNCP focuses on a small volume of the nearby universe, typically some 200 to 500 million light years across. In this space, the supercomputers generate similar starting conditions as would have dominated the universe around 380,000 years after the Big Bang. These conditions have been deduced from the data gathered by the WMAP and COBE satellites that thoroughly analysed cosmic background radiation.
The originality of the MNCP research lies in the fact that it aims to reproduce the type of objects that we observe around us. To do this, researchers are using a technique that can add observational links to the spatial distribution of mass and speeds derived from the galaxy catalogues. This way, they can “prepare” simulations to generate, at the end of the time period, structures that are very similar to what we see around us, including our very own Milky Way and its neighbour, Andromeda, plus a great many satellite galaxies, as shown in the image.
After analysis, these data will be an excellent basis upon which to run a great deal of experiments, that is, we will be able to observe our bit of simulated universe very like we observe the real universe through today’s telescopes. Additionally, as the objects that are formed are very similar to the real ones, the simulations and observations can be compared directly.
Additionally, unlike the real universe, an observer can move backward and forward through both the three spatial dimensions and the time dimension in a simulated world. Supercomputers like Magerit are a natural astrophysics and cosmology laboratory. This is a big advantage and will help these fields to become less speculative and gain the status of experimental sciences. Access to vast supercomputing infrastructures is now essential for progress in many sciences, known as the e-sciences, where experimentation is either impossible, as in the case of astrophysics, or extremely costly or hazardous to do.
Spain has invested a lot in observational astronomy. On the La Palma in the Canary Islands, the construction of one of the biggest telescopes in the world is close to completion. Spanish research groups are partners of most major international earth or space telescope or radio telescope development projects. Spain recently joined the European Southern Observatory, one of the most important European and world astronomy organizations.
In computational terms, Spain has taken a giant step forward and now has a supercomputing infrastructure (RES) that would have been unimaginable only a few years ago. CesViMa is one of its key nodes. Just as the huge investments in astronomical instruments have led to a spectacular development of this country’s astrophysics, sustained investments in supercomputing will provide a comparable driving force for what are known as the e-sciences, including computational astrophysics and cosmology.
Eduardo Martínez | alfa
Deep Learning predicts hematopoietic stem cell development
21.02.2017 | Helmholtz Zentrum München - Deutsches Forschungszentrum für Gesundheit und Umwelt
Sensors embedded in sports equipment could provide real-time analytics to your smartphone
16.02.2017 | University of Illinois College of Engineering
13.02.2017 | Event News
10.02.2017 | Event News
09.02.2017 | Event News
24.02.2017 | Life Sciences
24.02.2017 | Life Sciences
24.02.2017 | Trade Fair News