91̽»¨

Abstract:

A number of space missions dedicated to the search for exoplanets via the transit method, such as COROT, Eddington and Kepler, are planned for launch over the next few years. They will need to address problems associated with the automated and efficient detection of planetary transits in light curves affected by a variety of noise sources, including stellar variability. To maximize the scientific return of these missions, it is important to develop and test appropriate algorithms in advance of their launch dates. Starting from a general-purpose maximum-likelihood approach we discuss the links between a variety of period- and transit-finding methods. The natural endpoint of this hierarchy of methods is shown to be a fast, robust and statistically efficient least-squares algorithm based on box-shaped transits. This approach is predicated on the assumption of periodic transits hidden in random noise, usually assumed to be superposed on a flat continuum with regular continuous sampling. We next show how to generalize the transit-finding method to the more realistic scenario where complex stellar (micro) variability, irregular sampling and long gaps in the data are all present. Tests of this methodology on simulated Eddington light curves, including realistic stellar microvariability, irregular sampling and gaps in the data record, are used to quantify the performance. Visually, these systematic effects can completely overwhelm the underlying signal of interest. However, in the case where transit durations are short compared to the dominant time-scales for stellar variability and data record segments, it is possible to decouple the transit signal from the remainder. We conclude that even with realistic contamination from stellar variability, irregular sampling, and gaps in the data record, it is still possible to detect transiting planets with an efficiency close to the idealized theoretical bound. In particular, space missions have the potential to approach the regime of detecting Earth-like planets around G2V-type stars.

Abstract:

A method for simulating light curves containing stellar micro-variability for a range of spectral types and ages is presented. It is based on parameter-by-parameter scaling of a multi-component fit to the solar irradiance power spectrum (based on VIRGO/PMO6 data), and scaling laws derived from ground based observations of various stellar samples. A correlation is observed in the Sun between the amplitude of the power spectrum on long (weeks) timescales and the BBSO Ca II K-line index of chromospheric activity. On the basis of this evidence, the chromospheric activity level, predicted from rotation period and B - V colour estimates according to the relationship first introduced by Noyes (1983) and Noyes et al. (1984), is used to predict the variability power on weeks time scales. The rotation period is estimated on the basis of a fit to the distribution of rotation period versus B - V observed in the Hyades and the Skumanich (1972) spin-down law. The characteristic timescale of the variability is also scaled according to the rotation period. This model is used to estimate the impact of the target star spectral type and age on the detection capability of space based transit searches such as Eddington and Kepler. K stars are found to be the most promising targets, while the performance drops significantly for stars earlier than G and younger than 2.0 Gyr. Simulations also show that Eddington should detect terrestrial planets orbiting solar-age stars in most of the habitable zone for G2 types and all of it for K0 and K5 types.