On the Difference in Statistical Behavior Between Astrometric and Radial-Velocity Planet Detections

Abstract

Astrometric and radial-velocity planet detections track very similar motions, and one generally expects that the statistical properties of the detections would also be similar after they are scaled to the signal-to-noise ratio of the underlying observations. I show that this expectation is realized for periods small compared to the duration of the experiment P/T << 1, but not when P/T >~ 1. At longer periods, the fact that models of astrometric observations must take account of an extra nuisance parameter causes the mass error to begin deteriorating at P/T ~ 0.8, as compared to P/T ~ 1.0 for RV. Moreover, the deterioration is much less graceful. This qualitative difference carries over to the more complicated case in which the planet is being monitored in the presence of a distant companion that generates an approximately uniform acceleration. The period errors begin deteriorating somewhat earlier in all cases, but the situation is qualitatively similar to that of the mass errors. These results imply that to preserve astrometric discovery space at the longest accessible orbits (which nominally have the lowest-mass sensitivity) requires supplementary observations to identify or rule out distant companions that could contribute quasi-uniform acceleration.