Abstract

The Cold Dark Matter paradigm has been well tested on large scales, yet observational constraints are weaker at sub-galactic scales. The gravitational potential of the Milky Way is generated by all of the matter – both baryonic and dark. By mapping the potential, we can thus uncover the distribution of the unseen dark component of the Milky Way. Gaia has precisely measured 6D phase-space coordinates of over 30 million stars, dramatically expanding our knowledge of stellar kinematics in the Milky Way. Prior to Gaia, highly simplified models were used to recover the gravitational potential from stellar kinematics, but the quantity and quality of the new phase-space data provided by Gaia demands new approaches that can more fully describe the richness of the data. I will discuss a new method, “Deep Potential,” which applies computational tools from Deep Learning in a physically principled way to solve the collisionless Boltzmann equation with minimal assumptions and thus to recover the underlying gravitational potential and dark matter density distribution.