Searching for the most distant quasars

A quasar, the glowing disk of gas around the super-massive black hole at the centre of a galaxy, can be hundreds of times more luminous than its host galaxy, and the brightest quasars are the most powerful non-transient objects in the Universe. Quasars can hence be seen to great distances, and so provide one of the best ways to find out about conditions in the first billion years after the Big Bang and, in particular, the first super-massive black holes. (The most distant known quasar, which we discovered at Imperial, is seen as it was when the Universe was just ~5% of its current age.) Finding more quasars in the early Universe would hugely aid such studies - but they are also rare, so identifying these objects efficiently is a very challenging problem in astronomical data analysis. This project will be to develop machine learning (and other) methods to identify the most distant quasars in up-coming data-sets from the Euclid satellite mission (scheduled for launch in 2020) and the Large Synoptic Survey Telescope (LSST, first light expected in 2020). A part of any such rare object search is the visual inspection of a large number of candidates, which has traditionally been done by expert astronomers, but will now, thanks to the sheer data volume, have to be done by automatically. This project is hence to develop machine learning algorithms to replace humans in this task, and make it possible to extend rare object searches to the next generation of astronomical survey. It will involve a combination of algorithm development, simulations and testing/refinement on real data.