"Gas accretion and galactic winds are two of the most important and poorly understood ingredients of models for the formation and evolution of galaxies. We propose to take advantage of two unique opportunities to embark on a multi-disciplinary program to advance our understanding of the circumgalactic medium (CGM).We will use MUSE, a massive optical integral field spectrograph that we helped to develop and that will be commissioned on the VLT in 2012, to study the CGM in both absorption and emission. We will use 200 hours of guaranteed time to carry out deep redshift surveys of fields centred on bright z≈3.5 and z≈5 QSOs. This will yield hundreds of faint galaxies (mainly Lyα emitters) within 250 kpc of the lines of sight to the background QSOs, an order of magnitude increase compared to the best existing sample (bright, z≈2.3 galaxies). This will allow us to map the CGM in absorption in 3-D using HI and metal lines and to identify, for the first time, the counterparts to most metal absorbers. MUSE will also enable us to detect Lyα emission from the denser CGM (also using another 300 hours of guaranteed time targeting deep HST fields) and thus to directly explore its connection with galaxies and QSO absorbers.We will use the new supercomputer of the Virgo consortium to carry out cosmological hydro simulations that contain 1-2 orders of magnitude more resolution elements than the largest existing (spatially adaptive) runs. We will use the results of our previous work to guide our choice of parameters in order to obtain a better match to the observed mass function of galaxies. In parallel, we will carry out a complementary program of zoomed simulations of individual galaxies. These will have the physics and resolution to include a cold gas phase and hence to bypass much of the ""subgrid"" physics used in the cosmological runs. Both types of simulations will be used to study the physics of gas flows around galaxies and to guide the interpretation of our observations."