Title: Where are the extremely metal-poor stars in the Milky Way and Andromeda? Expectations from TNG50

Abstract: We analyse the location of extremely metal-poor stars (EMPs, [Fe/H]$ < -3$) in 198 Milky Way (MW)/M31-like galaxies at $z=0$ in the TNG50 simulation. Each system is divided into four kinematically-defined morphological stellar components based on stellar circularity and galactocentric distance, namely bulge, cold disk, warm disk, and stellar halo, in addition to satellites (with stellar mass $\ge 5\times10^6\,M_\odot$). According to TNG50 and across all simulated systems, the stellar halo of the main galaxy and satellites present the highest frequency of EMPs (largest $M_{\mathrm{EMP, comp}}$-to-$M_{\mathrm{tot, comp}}$ stellar mass ratio), and thus the highest chances of finding them. Such frequency is larger in lower-mass than high-mass satellites. Moreover, TNG50 predicts that the stellar halo of the main galaxy always hosts and thus contributes the majority of the EMPs of the system. Namely, it has the highest mass ratio of EMPs in it to all the EMPs in the system (largest $M_{\mathrm{EMP, comp}}$-to-$M_\mathrm{EMP} (<300\mathrm{kpc})$). However, notably, we also find that 33 MW/M31-like galaxies in TNG50 have cold disks that contribute more than 10 per cent to the total EMP mass, each with $\gtrsim 10^{6.5-7}\, M_\odot$ of EMPs in cold circular orbits. These qualitative statements do not depend on the precise definition of EMP stars, i.e. on the adopted metallicity threshold. The results of this work provide a theoretical prediction for the location of EMP stars from both a spatial and kinematic perspective and across an unprecedented number of well-resolved MW/M31-like systems.