Binary-corrected velocity dispersions from single- and multi-epoch radial velocities: massive stars in R136 as a test case

Orbital motions from binary stars can broaden the observed line-of-sight velocity distribution of a stellar system, artificially inflating the measured line-of-sight velocity dispersion, which can in turn lead to erroneous conclusions about the dynamical state of the system. Cottaar et al. (2012b) proposed a maximum likelihood procedure to recover the intrinsic velocity dispersion of a resolved star cluster from a single epoch of radial velocity data of individual stars, which they achieved by simultaneously fitting the intrinsic velocity distribution of the single stars and the centres of mass of the binaries along with the velocity shifts caused by binary orbital motions. Assuming well-characterized binary properties, they showed that this procedure can accurately reproduce intrinsic velocity dispersions below 1 km s$^{-1}$ for solar-type stars. Here we investigate the systematic offsets induced in cases where the binary properties are uncertain, and we show how two epochs of radial velocity data with an appropriate baseline can help to mitigate these systematic effects. We first test the method above using Monte Carlo simulations, taking into account the large uncertainties in the binary properties of OB stars. We then apply it to radial velocity data in the young massive cluster R136, an example for which the intrinsic velocity dispersion of O-type stars is known from an intensive multi-epoch approach. For typical velocity dispersions of young massive clusters ($\gtrsim 4$ km s$^{-1}$) and with a single epoch of data, we demonstrate that the method can just about distinguish between a cluster in virial equilibrium and an unbound cluster. This is due to the higher spectroscopic binary fraction and more loosely constrained distributions of orbital parameters of OB stars compared to solar-type stars. By extending the maximum likelihood method to multi-epoch data, <abridged>