Towards optimal dynamical models of the Milky Way halo

Dynamical inference of the dark matter halo mass requires  modelling the phase space structure of dynamical tracers, with different tracers following different models and having different levels of sensitivity to the halo mass. Based on mock data from high resolution simulations, I will first show that some unjustified but popular assumptions in the construction of a dynamical model can introduce significant biases to the mass inference of the Milky Way halo. Steady-state modelling with minimal assumptions removes such overall biases, but still suffers from an irreducible amount of stochastic bias. This irreducible bias is determined by the level of deviation from steady-state in the tracer population, which is small for dark matter particles but as large as a factor of 2 for halo stars. By contrast, satellite galaxies show a negligible amount of such irreducible bias and is expected to be a better tracer of the halo mass than stars. Combining the steady-state information with semi-empirical knowledges of the particle orbits further improves the model accuracy.