Binary neutron star formation and the origin of GW170817

The first neutron star-neutron star merger was detected in by LIGO/Virgo in a galaxy in which the majority of star formation was taking place a long time ago (11 Gyr). LIGO/Virgo estimated that local cosmic NS-NS merger rate is 110-3840 Gpc^-3 yr^-1. Only some extreme evolutionary models can generate NS-NS merger rates in old host galaxies consistent with the LIGO/Virgo estimate. However, we show that these models generate rates exceeding empirical Galactic NS-NS merger rates based on the large population of Milky Way NS-NS binaries. Typically, current evolutionary models produce NS-NS merger rates that are consistent with the Milky Way empirical rates. However, these models generate local cosmic NS-NS merger rate in old host galaxies that are below the LIGO/Virgo estimate. The reason behind this tension is the predicted delay time distribution between star formation and NS-NS mergers that favors short delays. Evolutionary models produce a generic steep power-law NS-NS delay time distribution. This limits NS-NS merger rates in old host galaxies. However, we show that such distribution is consistent with observations of Galactic NS-NS binaries; 50% of which show very long merger times (much longer than Hubble time). Once model distributions are convolved with continuous prolonged (10 Gyr) star formation in the Galactic disk, then 20-70% (depending on a model) of the predicted NS-NS population has very long current Galactic merger times. Although NS-NS binaries are formed predominantly with short delay times, many of short delay time systems merge and do not make it to the present, while long delay time systems survive and contribute to the current Galactic NS-NS population. This study highlights the tension between the current evolutionary predictions and the observation of the first NS-NS merger in an old host galaxy.