Dark matter is one of the strangest puzzles in physics: the universe seems to be composed of far more stuff than we can see directly. The evidence is overwhelming, yet indirect.Over the past few decades, astronomers have been able to measure the tug of gravitation in a huge range of environments – from galaxies, through clusters of hundreds of galaxies, to the early universe as seen in the cosmic microwave background – and we keep coming to the same conclusion, that the tug is surprisingly strong. To explain that, one needs around six times more matter than is seen directly. The missing five sixths is known as dark matter.

But attempts to detect particles of this matter have given no clear result. That doesn’t mean dark matter doesn’t exist, but it does mean we can’t be certain what it’s made from. If we can’t manufacture or find particles of it here on Earth, the only way to make progress is to keep studying its behaviour in the sky.

As our ability to infer the presence of dark matter in the night sky has improved, we can now measure how the invisible mass is distributed through individual galaxies. In the meantime, increasingly powerful computers have allowed the expected distribution to be predicted (based on assumptions about the particle’s nature). It has proved exceptionally difficult to get these two methods to agree – perhaps pointing to an exotic dark matter particle which behaves in unexpected ways; or, alternatively, that the entire dark matter idea is on the wrong track.

Using the Dirac-2 Complexity cluster, the Horizon collaboration has been studying this problem from a different perspective. Perhaps the difficulty in making theoretical ideas agree with the real Universe stems from an incomplete physical picture of what goes on inside galaxies. We know that vast amounts of energy are available – young stars pump out light and occasionally explode in supernovae. If some of this energy is transferred into the dark matter, the distribution can be radically reshaped.

We established almost two years ago that there were viable physical mechanisms for transporting stellar energy through to the dark matter distribution (dp016.6). Since then, we have been studying with higher resolution simulations how sensitive this mechanism is to fine details of how the gas behaves (e.g. dp016.7) – always the hardest thing to get right in galaxy simulations, because there are so many factors (including resolution, the cooling rates and the coupling of stellar populations to the surrounding gas). Our latest work, (Pontzen et al 2013, MNRAS submitted), shows that even if one gets many of these details intentionally wrong, the dark matter reshaping remains possible.

This is an encouraging sign that despite remaining uncertainty, we can soon determine whether data from the real universe support – or perhaps refute – the simplest dark matter models.