Extrasolar planets

Searches for extrasolar planets have started to assume center stage both in the professional arena and in the public perception of astronomy. The recent detection of a planet orbiting 51Peg (Mayor & Queloz, 1995) has generated much interest, and it is widely believed that more giant gaseous planets around solar-type stars can be found by precise radial-velocity and astrometric surveys. Both these methods are indirect, in that they measure the motion of the star around the barycenter of the star-planet system. While radial velocity searches could soon become a very efficient method to detect exoplanets, they have a serious drawback: they cannot determine separately the planetary mass and its orbital inclination; only is measurable. In contrast, astrometric observations give M directly. Of course, in planetary systems which are viewed ``pole-on'' astrometry provides the only way to detect reflex motion.

The VLTI has the potential to be an extremely powerful instrument for precise narrow-angle astrometry. For instance, the atmospheric limit for determining the separation vector between two stars which are 10'' apart is about 10 micro-arcsec for a half-hour integration. Realizing this potential is a challenging but solvable task. It requires monitoring the baseline vector inside the interferometer with ~`50m precision and measuring the differential delay between the two stars with ~0.005 micron precision. Implementing an astrometric mode in VLTI with these capabilities would enable us to detect Sun-Jupiter systems out to a distance of 1kpc and small planets (10 Earth masses) around the closest stars.

The following observing strategy could be adopted for the VLTI astrometry program: a list of ~200 target stars would be observed in the near infrared with VISA. These stars must be bright enough for fringe tracking (K<12), which will allow the astrometric reference sources to be relatively faint (K<17) and ensure that references for phasing can be found for almost any object of interest. The integration time would be half an hour per star per night. With thirty observations of each target star over ten years, this would require a total commitment of 300 nights on VISA spread over a decade. The data for each star would be used to solve for relative parallax and proper motion, with any residuals indicating the presence of planets. In practice, one would use two or three different astrometric reference stars for each target to remove ambiguities from the motions of the reference stars themselves.

The target list would include candidate planetary systems found from radial velocity searches, for which VLTI could determine the inclination and thus the planetary mass. The list could also include candidates from a large-scale astrometric survey such as the GAIA project, a mission proposed within the Horizon 2000+ program of ESA. In the GAIA data, which should also have a precision of ~10 micro-arcsec, planets would be revealed by conspicuous residuals in the astrometric fit, but the mission lifetime and temporal sampling would generally not be adequate to determine planetary orbits.

While observations of candidate exoplanets would make the VLTI program quite efficient, the target list should also include a survey of other ``interesting'' objects. Examples are the closest stars, for which the VLTI can detect planets with lower masses than can radial velocity searches, IR-excess stars like Beta Pic, and pre-main-sequence objects in low-mass star forming regions and in Orion. The astrometric mode of the VLTI would thus open new vistas in the study of the formation and evolution of planetary systems. It could also provide an input list for even more ambitious space interferometry projects aimed at spectroscopic investigations of extrasolar giant and Earth-like planets. One such project, DARWIN, is also under study as a cornerstone mission within Horizon 2000+.

While these indirect methods will certainly yield a wealth of data about extrasolar planetary systems, the direct detection of photons originating from the planet itself would enable more detailed astrophysical studies. Examples include determining the chemical composition and temperature of the planets through spectroscopy, and studying surface structure and rotation by analyzing the lightcurve. However, in the visible and near-IR regimes it is prohibitively difficult to detect planets against the glare of the parent star. The only chance lies in the mid-infrared, where the contrast is reduced by several orders of magnitude. In the 10 and 20 microns atmospheric windows, the thermal emission of a planet depends strongly on its temperature (for example, at 10microns the Earth is brighter than Jupiter). Simple sensitivity calculations show that Jupiter at a distance of 10pc would not be detectable against the thermal background in a reasonable time with an Earth-based (and therefore uncooled) 8m telescope. It should be kept in mind, however, that other planetary systems may be very different from ours. In particular, other giant planets may be warmed by internal heating, which is stronger in planets that are younger or more massive. Planets may also be warmed by strong irradiation, either because the parent star has an early spectral type or because the orbit is small (as in the case of 51Peg). There may be a realistic chance of detecting such warm giant planets in the solar neighborhood with the VLTI at 10 or 20microns, provided their temperature is at least 400K. Suitable candidate objects for such an ambitious project, which would require several hours of integration time with the full array of four 8-m telescopes, could be drawn from the astrometric survey list.