How galaxies are born inside computers

A look at the supercomputers unlocking the secrets of our cosmos
Written by Nick Heath, Contributor

A look at the supercomputers unlocking the secrets of our cosmos

The next time you feel like your computer is struggling to keep up with your workload, spare a thought for the physicists at the Institute for Computational Cosmology (ICC).

The researchers at the institute, based at Durham University, are tasking their machines with nothing less than recreating how galaxies are born and evolve over the course of billions of years.

Luckily, the institute is packing more power than your average desktop PC.

It uses a supercomputing cluster of about 800 AMD processor cores and 1,600GB of memory to run its cosmological models, which are built using information about the mix of matter and energy needed to create a galaxy.

This data is fed into equations representing the fundamental laws of physics - the forces which shape the development of the universe - and then crunched by the institute's supercomputer.

To take a look at the supercomputers that model the universe, see Photos: Digital galaxies and the supercomputers behind them

The models it produces allow the institute to simulate the processes that researchers believe are key to galaxy formation. This includes the influence of dark matter - invisible matter that is only detectable by the effect of its gravity - as well as the heating and cooling of the universe, the process of star formation and the impact of supernova explosions on gas inside galaxies.

supercomputer galaxy formation model

One of the institute's models of the swathe of stars, tens of thousands of light years across, enveloping a Milky Way-like galaxy.
(Image credit: A.P. Cooper and J. Helly / Virgo Consortium)

But despite having all of this computing horsepower at its disposal, the institute can still find technology doesn't always keep up with its work.

Storage presents a headache for the organisation, for example - unsurprising perhaps given a computer model of the effect of dark matter on galaxy formation can produce 20TB of data in a single run. Researchers are constantly deleting old unwanted data to free up space on the institute's 300TB storage array, although data is also archived on the institute's tape library, which in turn currently holds about 100TB of data.

The institute is also still bumping up against the limits of how accurately it is possible to model the formation of galaxies - and the billions of stars they contain - inside a computer.

Dr Lydia Heck, the ICC's computer cluster manager, said the ICC had maxed out its supercomputing cluster's processors and...

...memory by running a simulation of the effect of dark matter on how galaxies are formed.

And the maxed-out cluster is not even using large scale models.

Physicists have to simplify the cosmological models they use in order to get ones that produce data sets small enough to be accurately processed by the 64-bit chips in the supercomputing cluster, and which can fit into the cluster's available memory.

One way that the computer models are simplified is by losing some of the detail of what is being modelled - for instance, choosing to model the formation of millions of stars at a time rather than trying to model the formation of each one of the hundreds of billions of stars within a galaxy.

Another simplification method is to approximate the effect of certain forces within a galaxy, such as not trying to model the gravitational pull of every object within it.

"If you want to put the full universe into the computer, even on such a small scale as a single galaxy, and you wanted to calculate every single detail, then I don't think you could," said Heck.

"You have to make approximations, you have to decide what is of no importance and you can neglect it."

institute for computational cosmology supercomputer

The Institute for Computational Cosmology's main supercomputer, composed of 800 AMD processors and 1,600GB of memory
(Photo credit: Institute for Computational Cosmology)

The differences between models and real life galaxies
To refine the models, researchers use a process of elimination, comparing the characteristics of the galaxies produced by the computer model with observations of galaxies taken by astronomers to see if the computer model is a good fit for real life.

"We add the processes that create stars so we can make the universe shine and compare it to what we see with telescopes," Richard Bower, professor of physics at the ICC, told silicon.com.

To take a look at the supercomputers that model the universe, see Photos: Digital galaxies and the supercomputers behind them

"We can track in the computer where and when those stars form and compare it to what we see."

If the galaxies in the computer model are not a good match for real observations, the researchers simplify other parts of the simulation, and see if that model corresponds better to real life data.

Bower said that as computer processors have become more powerful, and as memory and storage capacity have increased, researchers have become able to produce more detailed and accurate models of how galaxies are formed.

"A few years ago we would be lucky to be able to make a simulation and represent a galaxy with a few thousand particles [where each particle represents anything from 10,000 to 1,000,000 stars depending on the simulation]," he said, adding: "Clearly that is not anywhere near sophisticated enough, galaxies are very complex.

"Now we're making galaxies with hundreds of thousands of particles - the state of the art right now is to model the formation of a single galaxy in a representative patch of the universe.

"Maybe over the next five years people will be able to...

...form groups of galaxies or small complete patches of the universe."

The institute's supercomputing cluster - at one time the most powerful supercomputer being used for academic research within Europe - has dropped back down the international supercomputing rankings since it was last upgraded in 2006.

To take a look at the supercomputers that model the universe, see Photos: Digital galaxies and the supercomputers behind them

However, an upgrade is not far away and by December the institute will have a new £1.6m cluster up and running. It will have a processing capacity many times more powerful than the current one, more memory, will increase the institute's storage capacity by more than 600TB and be able to transfer data between processors much faster than the ICC's existing set-up.

According to Bower, the new computing cluster will allow the researchers to create models using fewer approximations and "get a much more realistic calculation".

"As we improve this calculation presumably we are going to get a simulated universe that is much more like the universe that we see or it will show us that we're missing some exciting component.

"This leap in computing power that we are going to have is going to be really important," he said.

New discoveries
Once a model is found to produce galaxies whose characteristics match measurements taken by astronomers, cosmologists can then study the way the galaxy forms inside the computer model to discover more about the processes that lead to their formation in real life.

To date, one of the most significant discoveries thrown up by one of the institute's computer models has been to shed light on how black holes effect the formation of stars later on in a galaxy's life.

In simple terms, the model showed that the jets of energy that stream out from a black hole blast gaseous star-forming material away from galaxies, preventing stars from forming. This discovery has in turn led to more accurate models of galaxy formation and development being created.

institute for computational cosmology galaxy formation model

A computer model showing the structure of the universe one billion years after the Big Bang, the green swirls represent dark matter and the circles represent growing galaxies
(Photo credit: Sarah Noble and Vicky Greener, department of physics, Durham University)

As well as working on galaxy formation, computers at the institute are also used to model the structure of the visible universe - a process which involves modelling the creation of the "cosmic web", the name given to the strings of galaxies and vast expanses of void that make up the visible universe.

The models of the cosmic web are designed to provide an insight into the rate at which the universe is expanding and the nature of dark energy - the energy that is believed to permeate the cosmos and fuel the expansion of the universe.

The ICC does not work alone, it is a member of the Virgo Consortium - a group of scientists spread across the world who share computing resources in order to run cosmological simulations exploring areas like the formation and evolution of galaxies and the distribution of dark matter.

By helping researchers understand more about how the universe was formed, the computer models could one day even help physicists predict how alternative universes might evolve.

"We could start the universe from different initial conditions, and say things like 'What if a universe had no dark matter? Could it produce different galaxies?'," Bower said.

"Those questions are pretty far out at the moment, most of my colleagues think that I'm pretty crazy bringing them up but ultimately this is a pretty exciting direction."

Editorial standards