One of the clusters depicted in the simulationIllustris

Imagine exploring 13 billion years of astronomical evolution using a single desktop simulation. The Illustris project simulates the development of a cubic region of space, with a side-length of 350 million light-years, from just after the Big Bang up until the present day. The results of this incredible project – essentially the universe in-a-box – were published this week in the journal Nature. There, the research team involved describe how more than 100,000 lines of code and 8000 CPU hours were used to amalgamate all the complex physical models we have to shape cosmology into this masterpiece. To give you an idea of the enormity of the calculations and data involved, it would take an average desktop computer about 2000 years to reach this solution.

Previous simulations of the universe have made significant developments over the past 20 years; however, they were limited by incomplete physical models, and numerical inaccuracies. Inadequate computing power also meant that only very small areas of the universe could be simulated at a high resolution.

In comparison, Illustris models the composition of the universe on both a large and small scale with incredible detail; from the number (over 41,000) and morphology of galaxies, to the relative proportion of dark matter, right down to the distribution and abundance of individual elements within cosmological bodies. In fact, it is visually difficult to distinguish between images taken from this simulation and those acquired using the Hubble space telescope when observing the same area of space.

The scientific team behind Illustris was led by Mark Vogelsberger from MIT, who was working in collaboration with researchers from Germany, the US, and also Dr Deborah Sijacki here at the University of Cambridge. The team achieved this impressive feat by formulating a new theoretical model incorporating all of the other physical models of universe formation into a global solution. They then structured this theoretical model so that it was numerically appropriate for computer simulation.

The result was an incredibly beautiful and accurate representation of the universe unfolding over 13 billion years. Apart from being visually pleasing and a testament to modern computing power, the Illustris project has provided valuable insight into the validity of our current physical models. The similarity between simulated and observed data provides indirect evidence in support of the physics behind the project. This is particularly important considering ‘experiments’ to test theoretical models in astrophysics cannot be conducted as easily as in other fields of science.

Incredibly, the Illustris simulation is still not powerful enough to model rare cosmological events like the formation of black holes and suffers issues related to overestimating the rate of star formation in small galaxies. However, as Vogelsberger said when speaking to journalist Lizzy Gibney for a Nature podcast: “it’s not the successes of the model that push us forward; it is definitely the failures, because then we can learn what is wrong in our current understanding.”

Images and videos of the simulation are available through the project.