Recent studies of a system consisting of two colliding galaxy clusters (the so-called "Bullet Cluster") have shown clear evidence for a spatial segregation of baryonic matter and the gravitationally dominant matter component. This provides the most direct evidence so far for the existence of dark matter and contradicts models where the dark matter is explained away by modifying gravity.
I will report on a joint Chandra/HST/VLT study of a cluster merger similar to the "Bullet Cluster". The merger was serendipitously discovered in multi-wavelength follow-up observations of the first well-localized short-duration gamma-ray burst. The resulting HST image is even deeper than the original Hubble Deep Field(s), and weak lensing measurements based on these data allow for very precise mapping of the dark matter distribution in the cluster.
Finally, I will describe searches for X-ray emission from decaying or annihilating dark matter particles inside dark matter concentrations that become temporarily "baryon-poor" as a result of cluster collisions. Recently derived upper limits on such X-ray emission put strong constraints on the properties of some dark matter candidates such as sterile neutrinos and Kaluza-Klein axions.