Towards a new generation of predictive galaxy formation simulations

Upcoming experiments will map galaxies and gas across unprecedented volumes and probe further back into cosmic time than ever before. These experiments have the potential to probe fundamental physics questions such as the nature of dark matter and dark energy, and the initial conditions of the Universe. But in order to extract the full scientific potential from these data, we need to understand how luminous tracers (stars and gas) are related to the underlying matter density field. Simulating galaxy formation from first principles is a huge computational challenge because of the vast range of scales and rich array of physics involved. All current large-volume simulations adopt ad-hoc phenomenological "sub-grid" recipes to treat critical physical processes such as star formation, stellar feedback, and black hole growth and feedback. I will review the current status of these simulations and highlight some of their successes, shortcomings, and challenges. I will then describe the philosophy and status of the SMAUG (Simulating Multiscale Astrophysics to Understand Galaxies) project, which aims to develop new, more physically grounded and predictive treatments of sub-grid processes in cosmological galaxy formation simulations, and present some of our recent results