What is the EAGLE project?
EAGLE (Evolution and Assembly of GaLaxies and their Environments) is a simulation aimed at understanding how galaxies form and evolve. This computer calculation models the formation of structures in a cosmological volume, 100 Megaparsecs on a side (over 300 million light-years). This is large enough to contain 10,000 galaxies of the size of the Milky Way or bigger, enabling a comparison with the whole zoo of galaxies visible in the Hubble Deep field for example. This website contains downloadable images and movies, many of which are located in Highlights or Downloads.
The simulation starts when the Universe is still very uniform - no stars nor galaxies had formed yet - with cosmological parameters motivated by observations by the Planck satellite of the cosmic microwave background. Crucial parameters are the density of dark matter - which allows structures to grow, baryonic matter - the gas from which stars form, and the cosmological constant - responsible for cosmic acceleration.
Dark matter enables structures like galaxies to form, even while the Universe is expanding rapidly. Gas falling into these dark matter structures cools and forms stars: this is how galaxies form. However core collapse supernovae, exploding massive stars, and AGN (Active Galactic Nuclei), bursting supermassive black holes, severely limit what fraction of the gas forms stars. The devastating effects of these explosions can be directly seen in starburst galaxies such as M82 and massive galaxies such as those in the Perseus cluster. Modelling these aspects accurately is key to produce a virtual universe that looks like the real one.
The image below is a slice through the simulation volume, with the intergalactic gas colour coded from blue to green to red with increasing temperature. Hot gas has temperatures of more than 100,000K, and is contained with dark matter structures that host galaxies. Such hot gas can be detected in X-rays. The insets zoom into a galaxy like the Milky Way, showing first its gas, and then its beautiful stellar disc: it looks remarkably similar to observed spirals.
The EAGLE simulation is one of the largest cosmological hydrodynamical simulations ever, using nearly 7 billion particles to model the physics. It took more than one and a half months of computer time on 4000 compute cores of the DiRAC-2 supercomputer in Durham. It was performed with a heavily modified version of the public GADGET-2 simulation code.
In the fields of cosmology and galaxy formation theory, numerical simulations play a crucial role to help scientist discriminate between various models and guide observations. Since all astrophysical events occur on very long time scales and on very large distances, astronomers cannot really perform experiments to test their theories as chemists, biologists and engineers would do in their laboratories. The observations can only tell us what a given galaxy looks like now and not what it was like in the past. Numerical simulations must hence be used to see how stars or galaxies evolve over the history of the Universe. By creating multiple simulations with different physical theories, astrophysicists can, in principle, eliminate models that lead to a virtual universe different from ours. This process should then lead astronomers towards a better understanding of the physical processes at stake in galaxy formation.
One of the main issues that simulators face is the complexity of the processes that take place and the vast range of scales that are involved. For instance, black holes at the centre of galaxies will swallow gas within 0.01 pc but the energy produced by this process will affect the galaxy or even its host halo up to almost 1 Mpc, effectively spanning 8 orders of magnitude in spacial resolution. Despite the huge computing power available nowadays in the big national facilities, we are still very far from being able to simulate a representative region of the Universe (~100 Mpc) at the resolution required to precisely simulate all the processes that are thought to be relevant for galaxy formation. Simulators then resort to so-called “subgrid models” to implement the physical processes that cannot be resolved ( simulated) due to the limited resolution. This technique of assuming a simple model to reproduce processes that are too small to be properly simulated is common to other areas of science, for instance turbulence or vortices are often simulated as subgrid models in simulations of air flows around cars or engines.
In the case of the EAGLE simulation, a resolution of 0.7 kpc is achieved (for the gravity), which lets us simulate the warm part of the gas within galaxies (ISM). For the processes occurring on smaller scales, we apply a series of physically motivated models, sometimes inspired by semi-analytic models. Our simulation includes:
- Cooling and heating of the gas due to the presence of stars and other emission,
- Formation of stars in cold and dense regions,
- Evolution and ageing of these stars,
- Distribution of the energy and metals (elements heavier than hydrogen) generated by the stars into the surrounding gas,
- Explosion of supernovae with injection of their energy in the surrounding gas,
- Formation of supermassive black holes,
- Accretion of gas onto the black holes,
- Ejection of energy due to this accretion process.
These subgrid models are all described by parameters whose values are either constrained directly by observations (the
evolution of stars is a good example) or have to be calibrated. For instance, the accretion of gas onto black holes and
the ejection of energy that is induced are still poorly understood theoretically and are anyway occurring on scales too
small for our simulation. We hence have to calibrate the parameters used in this particular model such that the effect
of the energy injection on the galaxies is reasonable. We know observationally how many galaxies of a given mass there
are in the Universe. We can hence use that information to make sure our model parameters lead to a virtual universe that
contains the right amount of galaxies of a given mass. In the case of energy injection from black holes, if the effect
is not strong enough, there will be too many big galaxies, conversely, if the effect is too strong, there won’t be
enough of them. The same applies to the injection of energy by supernovae which are one of the dominant mechanisms for
the formation of smaller galaxies.
On this figure, the black and white dots represent two different censuses of the number of galaxies of a given mass (x axis) in the Universe. The three coloured lines show the same galaxy count for our EAGLE virtual universe, with our main simulation being the blue line. The goal of the EAGLE project was to obtain a simulation set up (a series of parameters of the subgrid model) that leads to a good agreement between the observational data and the simulation. Other diagnostics plots where used to assess whether the galaxies are growing at the right rate and whether they end up having the right size.
Now, that we have obtained that reasonable-looking Universe, we can go back to the simulation and look how the galaxies evolve, form and interact. Essentially doing what other scientists can do in their lab: observe their experiment and see the physical processes unwind in front of their eyes. Making the simulation match this small set of observations was hence only the beginning. We can now start to address exciting questions such as:
-
How do galaxies stop growing? Is it because of the activity of the central black hole? Is it because they collide and merge? Is it because they are in a crowded environment?
-
How typical is our own Milky-Way? Are we in a normal galaxy in a normal part of the Universe or is there something special about where we live?
-
How do the different gas flows affect the formation of galaxies?
-
How does the presence of gas affect the observations of halo masses, lensing or dark matter? Will the presence of gas have to be taken into account when processing future observations coming from space telescopes such as Euclid ?
The completion of our simulation is only the first step in our process and the exciting science will happen now!