2001ApJ...562..549Z

We analyzed the Hipparcos astrometric observations of 47 stars that were discovered to harbor giant planets and 14 stars with brown dwarf secondary candidates. The Hipparcos measurements were used together with the corresponding stellar radial velocity data to derive an astrometric orbit for each system. To find out the significance of the derived astrometric orbits, we applied a ``permutation'' technique by which we analyzed the permuted Hipparcos data to get false orbits. The size distribution of these false orbits indicated the range of possibly random, false orbits that could be derived from the true data. These tests could not find any astrometric orbit of the planet candidates with significance higher than 99%, suggesting that most if not all orbits are not real. Instead, we used the Hipparcos data to set upper limits on the masses of the planet candidates. The lowest derived upper limit is that of 47 UMa–0.014 M<sub>☉</sub>, which confirms the planetary nature of its unseen companion. For 13 other planet candidates, the upper limits exclude the stellar nature of their companions, although brown dwarf secondaries are still an option. These negate the idea that all or most of the extrasolar planets are disguised stellar secondaries. Of the 14 brown dwarf candidates, our analysis reproduced the results of Halbwachs et al., who derived significant astrometric orbits for six systems that imply secondaries with stellar masses. We show that another star, HD 164427, which was discovered only very recently, also has a secondary with stellar mass. Our findings support Halbwachs et al.'s conclusion about the possible existence of the ``brown dwarf desert'' that separates the planets and the stellar secondaries.