Read recent phd theses. They always have a few chapters at the front giving an overview of the state of the art of their subject area, to prove that the candidate knows it. As for subject area, just sample widely - it's impossible to know what will be useful in advance, and in any case it's all exercise for the brain.