Ultra-luminous X-ray sources and neutron-star-black-hole mergers from very massive close binaries at low metallicity

Gravitational waves from the binary black hole (BH) merger GW150914 may enlighten our understanding of ultra-luminous X-ray sources (ULXs), as BHs>30Msun can reach luminosities>4x10^39 erg s^-1 without exceeding their Eddington limit. It is then important to study variations of evolutionary channels for merging BHs, which might instead form accreting BHs and become ULXs. It was recently shown that massive binaries with mass ratios close to unity and tight orbits can undergo efficient rotational mixing and evolve chemically homogeneously, resulting in a compact BH binary. We study similar systems by computing ~120000 detailed binary models with the MESA code covering a wide range of initial parameters. For initial mass ratios M2/M1~0.1-0.4, primaries >40Msun can evolve chemically homogeneously, remaining compact and forming a BH without undergoing Roche-lobe overflow. The secondary then expands and transfers mass to the BH, initiating a ULX phase. We predict that ~1 out of 10^4 massive stars evolves this way, and that in the local universe 0.13 ULXs per Msun yr^-1 of star-formation rate are observable, with a strong preference for low-metallicities. At metallicities log Z>-3, BH masses in ULXs are limited to 60Msun due to the occurrence of pair-instability supernovae which leave no remnant, resulting in an X-ray luminosity cut-off. At lower metallicities, very massive stars can avoid exploding as pair-instability supernovae and instead form BHs with masses above 130Msun, producing a gap in the ULX luminosity distribution. After the ULX phase, neutron-star-BH binaries that merge in less than a Hubble time are produced with a formation rate <0.2 Gpc^-3 yr^-1. We expect that upcoming X-ray observatories will test these predictions, which together with additional gravitational wave detections will provide strict constraints on the origin of the most massive BHs that can be produced by stars.