Simulating the Universe

How do you accurately simulate the Universe on a computer?

A new programme at the Oxford Martin School, led by Pedro Ferreira of Oxford University's Department of Physics, aims to see how this and other seemingly impossible tasks can be tackled by developing new ways of handling data.

Pedro will be explaining this new approach as part of a showcase at the Royal Society this evening, but before that I caught up with him to ask about data, supercomputers and the biggest problems in science:

OxSciBlog: Why do we need new approaches to handling data?
Pedro Ferreira: We are at a threshold of a new era in Cosmology. Over the past few years we have developed very powerful instruments - groundbased, balloon born and satellite telescopes - that have a phenomenal capacity for collecting data. New surveys of galaxies, maps of cosmic radiation at various different frequencies all probe the Universe on a wide range of scales.

We want to learn what are the fundamental properties of the Universe, such as what its made of and how it is evolving. The new data sets are so massive that this can’t be done using conventional methods.

We need to be clever, innovative, pushing the boundaries of data management and statistical analysis. In particular we need to come up with radically different methods. Otherwise we won’t be able to extract the knowledge we want from the data we have in hand. And it is going to get much, much worse. 

OSB: What lessons have been learnt from astrophysics & cosmology about processing/searching large amounts of data?
PF: We have learnt a lot. For example, we have had to come up with very powerful ways of simulating the data sets.

Think about it: we want to simulate the Universe on a computer, something immense, but with enough detail that we can recognize the fine details we see around us. That has proven a challenge but as a result, members of our team have taken part in developing codes which are some of the most powerful in the world.

I mean by this that they can run on the largest supercomputers in the world, making the most use of the gigantic computing capacity of these machines. In fact, as a result of this capacity, these codes can be used to benchmark the next generation of supercomputers.  

OSB: What sort of new techniques will you look to develop?
PF: We have three strands of research that we want to pursue: First of all we want to develop methods to deal with the up and coming surveys of galaxies such as the ones that will come out of the Square Kilometre Array (SKA) or the Large Synoptic Survey Telescope (LSST). We need to be able to deal with data sets with billions, not millions of galaxies.

Second, we want to harness the capacity of the public to take part in the analysis of these data sets. Our Citizen Science project has been very successful in harnessing the creativity of hundreds of thousands of individuals online. We want to explore the potential of this new way of doing science.

Finally, we really want to push our ability to simulate the Universe using the most advanced computer codes in the field, developed by our team. These machines have to be able to correctly simulate the largest scales, the overall properties of the visible Universe while at the same time, pick out the fine details. How galaxies interact, merge and evolve to build up the complex cosmic ecology which we observe. 

OSB: How might this help researchers in fields such as oceanography, climate science and medicine?
PF: The problems we are facing in Cosmology are present in many other fields. In Climate we need to be able to simulate incredibly complex systems on a wide range of scales. In Oceanography, there are experiments which will try to map out the oceans in real time and tens of thousands at different points.

Imaging can and will play a crucial role in medicine and is amenable to the use of novel statistical methods. Citizen Science, as developed by our group, is already being deployed in a range of fields, from the classification of weather logs to the reconstruction of classical papyri.

Professor Pedro Ferreira is Director of the Computational Cosmology Programme at the Oxford Martin School and a Professor of Astrophysics.