We develop a method for recovering the global density distribution of the ancient Galactic stellar halo prior to disk formation , based on the present orbits of metal-poor stars observed in the solar neighborhood .
The method relies on the adiabatic invariance of the action integrals of motion for the halo population during the slow accumulation of a disk component , subsequent to earlier halo formation .
The method is then applied to a sample of local stars with [ Fe/H ] \leq - 1.5 , likely to be dominated by the halo component , taken from Beers et al. ’ s recently revised and supplemented catalog of metal-poor stars selected without kinematic bias .
We find that even if the Galactic potential is made spherical by removing the disk component in an adiabatic manner , the halo density distribution in the inner halo region ( R \leq 15 kpc ) remains moderately flattened , with axial ratio of about 0.8 for stars in the abundance range [ Fe/H ] \leq - 1.8 and about 0.7 for the more metal-rich interval -1.8 < [ Fe/H ] \leq - 1.5 .
The outer halo remains spherical for both abundance intervals .
We also find that this initial flattening of the inner halo is caused by the anisotropic velocity dispersions of the halo stars .
These results suggest that the two-component nature of the present-day stellar halo , characterized by a highly flattened inner halo and nearly spherical outer halo , is a consequence of both an initially two-component density distribution of the halo ( perhaps a signature of dissipative halo formation ) and of the adiabatic flattening of the inner part by later disk formation .
Further implications of our results for the formation of the Galaxy are also discussed , in particular in the context of the hierarchical clustering scenario of galaxy formation .