We present optical and ultraviolet spectra of the first electromagnetic counterpart to a gravitational wave (GW) source, the binary neutron star merger GW170817. Spectra were obtained nightly between 1.5 and 9.5 days post-merger, using the SOAR and Magellan telescopes; the UV spectrum was obtained with the Hubble Space Telescope at 5.5 days. Our data reveal a rapidly-fading blue component (T ≈ 5500 K at 1.5 days) that quickly reddens; spectra later than 4.5 days peak beyond the optical regime. The spectra are mostly featureless, although we identify a possible weak emission line at ∼ 7900 Å at t 4.5 days. The colours, rapid evolution and featureless spectrum are consistent with a "blue" kilonova from polar ejecta comprised mainly of light r-process nuclei with atomic mass number A 140. This indicates a sight-line within θ obs 45 • of the orbital axis. Comparison to models suggests ∼ 0.03 M of blue ejecta, with a velocity of ∼ 0.3c. The required lanthanide fraction is ∼ 10 −4 , but this drops to < 10 −5 in the outermost ejecta. The large velocities point to a dynamical origin, rather than a disk wind, for this blue component, suggesting that both binary constituents are neutron stars (as opposed to a binary consisting of a neutron star and a black hole). For dynamical ejecta, the high mass favors a small neutron star radius of 12 km. This mass also supports the idea that neutron star mergers are a major contributor to r-process nucleosynthesis. arXiv:1710.05456v1 [astro-ph.HE] 16 Oct 2017 2 NICHOLL ET AL.

Using the Very Long Baseline Array and the European Very Long Baseline Interferometry Network we have made a precise measurement of the radio parallax of the black hole Xray binary MAXI J1820+070, providing a model-independent distance to the source. Our parallax measurement of (0.348 ± 0.033) mas for MAXI J1820+070 translates to a distance of (2.96 ± 0.33) kpc. This distance implies that the source reached (15 ± 3) per cent of the Eddington luminosity at the peak of its outburst. Further, we use this distance to refine previous estimates of the jet inclination angle, jet velocity and the mass of the black hole in MAXI J1820+070 to be (63 ± 3) • , (0.89 ± 0.09) c and (9.2 ± 1.3) M , respectively.

We search the literature for reports on the spectral properties of neutron-star lowmass X-ray binaries when they have accretion luminosities between 10 34 and 10 36 ergs s −1 , corresponding to roughly 0.01% -1% of the Eddington accretion rate for a neutron star. We found that in this luminosity range the photon index (obtained from fitting a simple absorbed power-law in the 0.5-10 keV range) increases with decreasing 0.5-10 keV X-ray luminosity (i.e., the spectrum softens). Such behaviour has been reported before for individual sources, but here we demonstrate that very likely most (if not all) neutron-star systems behave in a similar manner and possibly even follow a universal relation. When comparing the neutron-star systems with blackhole systems, it is clear that most black-hole binaries have significantly harder spectra at luminosities of 10 34 − 10 35 erg s −1 . Despite a limited number of data points, there are indications that these spectral differences also extend to the 10 35 − 10 36 erg s −1 range, but above a luminosity of 10 35 erg s −1 the separation between neutron-star and black-hole systems is not as clear as below. In addition, the black-hole spectra only become softer below luminosities of 10 34 erg s −1 compared to 10 36 erg s −1 for the neutron-star systems. This observed difference between the neutron-star binaries and black-hole ones suggests that the spectral properties (between 0.5-10 keV) at 10 34 − 10 35 erg s −1 can be used to tentatively determine the nature of the accretor in unclassified X-ray binaries. More observations in this luminosity range are needed to determine how robust this diagnostic tool is and whether or not there are (many) systems that do not follow the general trend. We discuss our results in the context of properties of the accretion flow at low luminosities and we suggest that the observed spectral differences likely arise from the neutron-star surface becoming dominantly visible in the X-ray spectra. We also suggest that both the thermal component and the non-thermal component might be caused by low-level accretion onto the neutronstar surface for luminosities below a few times 10 34 erg s −1 .