UK supercomputer probes dark matter and dark energy

Portsmouth University's Sciama supercomputer will boost research into mysteries of the universe...

Portsmouth University's Sciama supercomputer will boost research into mysteries of the universe...

What happened just after the Big Bang? How do stars evolve? What's powering the expansion of the universe?

These are some of the biggest questions posed by the universe, and scientists are being helped in their quest to answer them by the University of Portsmouth's Sciama supercomputer.

The University's Institute of Cosmology and Gravitation (ICG) recently went live with its first supercomputer, which has a 1,008-core Intel cluster capable of one billion calculations per second. It uses 2.66GHz Intel Xeon processors and has 85 terabytes of fast parallel storage and 10 terabytes of NFS storage.

The facility is named after Dennis Sciama, a leading figure in the international development of astrophysics and cosmology, and is also an acronym for the South East Physics Network Computing Infrastructure for Astrophysical Modelling and Analysis.

The Cerro Tololo Inter-American Observatory in Chile

The Sciama supercomputer is using data from the Cerro Tololo observatory in Chile to research dark energy
Photo: NOAO/Aura/NSA

The University of Portsmouth's ICG was founded in 2002 and now has about 50 members, including academic staff, post-doctoral researchers, PhD students and international visitors.

About half of the ICG's work concerns the theoretical physics of the universe, while the other half is focused on observational work, although some research projects combine the two.

One of the projects that will use Sciama is the Dark Energy Survey - beginning in late 2011 - which will use optical data from the Cerro Tololo Inter-American Observatory in Chile to create an image of the night sky in the southern hemisphere.

As the name suggests, the research will study dark energy, which describes the force driving the accelerating expansion of the universe - as shown by the movement of galaxies away from each other.

"All the galaxies in the universe seem to be going faster and faster away from each other, which is a very strange thing indeed. We don't understand why they're going away faster and faster. Whatever's causing that we call dark energy, and from those distortions you see, you can learn about dark energy," senior research fellow at the University of Portsmouth Dr David Bacon explained.

The project will also look at dark matter, which...

...is the theoretical matter whose gravitational pull is attributed to distorting light as it travels through the universe.

The raw images taken by the Chilean telescope will be processed and analysed to profile the shape of galaxies. When the shape of these galaxies is similar, or aligned, it means the light that has travelled from them has been distorted in the same way, indicating the presence and location of dark matter.

"If that's true, you're learning directly about gravity and the universe - because it's gravity that would be distorting those objects," Bacon said.

He added that if a human spent one second looking at each galaxy in this image it would take five years to look at them all. On the other hand, the Sciama supercomputer can process this data rapidly through the use of parallel computing.

"We realised there's an awful lot of our calculations that can be done embarrassingly easily in a parallel way, and that's what our machine is good at," Bacon told silicon.com.

The Sciama supercomputer at the University of Portsmouth

The Sciama supercomputer at the University of Portsmouth has been designed to be particularly strong at parallel processing
Image: University of Portsmouth

Parallel computing means each processor core can process small parcels of data at the same time, rather than having to store a huge image in its memory.

"You can literally chop up this vast image of the night sky into little postage stamps of the galaxies you care about. And for each of those galaxies you just want to do some fairly simple tasks - you want to fit a profile to it, you want to measure its width, you want to count up all the pixel values.

"You want to do all those little things and you can do that independently of each little galaxy in your image - you don't need to have that vast image all in memory at one time. You can just send a little postage stamp to each core independently," Bacon said.

"By examining these observations in great detail, we can measure all the properties of galaxies - the facts and the data we need - and on the other hand, we can use the supercomputer to predict what different theories of dark energy would predict for those things we're seeing. So you can compare the predictions with the observations," he added.

Other supercomputers used for cosmology and astronomy by UK researchers include the Cosmos facility in Cambridge, which the University of Portsmouth already uses extensively.

Unlike Sciama, Cosmos has a shared-memory architecture that makes it...

...suitable for different kinds of computational calculations.

Another example is the Blue Gene supercomputer in the Dutch city of Groningen, which processes data from the Low Frequency Array, or Lofar, project using images from a series of radio telescopes across Europe, including one in Chilbolton near Winchester.

Bacon said this computer creates an image of the night sky across Europe "as it glows with radio waves". Each radio telescope can create up to 24 terabytes of data per day, so the computer requires a huge amount of memory as well as fast processing.

Sciama will be used to analyse the images created by the Groningen supercomputer to further understanding of dark energy as part of the Lofar project.

The initial plan is for the University of Portsmouth to use Sciama 70 per cent of the time, with other universities in the South East Physics Network using the remainder. The South East Physics Network is a consortium comprising the universities of Kent, Oxford, Queen Mary, Southampton, Surrey and Sussex.

The Chilbolton radio telescope near Winchester

Sciama will analyse radio images of the night sky from the Lofar network, which includes the Chilbolton facility near Winchester
Photo: Science & Technology Facilities Council

There are 28 registered users of the supercomputer. They will act as testers, identifying tweaks or the need for additional technology before heavy use of the facility begins.

Bacon's PhD student is using Sciama in a project investigating unified dark matter theory in which material has properties of both dark matter and dark energy. The research is about predicting how light will be distorted if unified dark matter theory is applied to objects in the universe.

Another researcher at Portsmouth is using the supercomputer to show how galaxies are spread out and to predict which types of galaxies occur in which parts of the universe.

International projects will use Sciama in the future, including the Sloan Digital Sky Survey III, which has been gathering data from the Sloan radio telescope in New Mexico for the past year. The five-year project aims to create a 3D image of the universe showing the distribution of galaxies in the largest volume of data from the universe ever surveyed.