Rationale for the use of color information on Eddington

For the Eddington mission, the intrinsic stellar variability can be a major source of noise in the detection of extrasolar planets by the transit method. We derive that most detections of terrestrial planets (1-2 R_Earth) will occur around G or K stars with 15-16th magnitude. When these stars are 7-12 times more variable than the Sun on a 10 hour timescale, we demonstrate that the detection can be performed with a higher S/N provided composite lightcurves obtained with the combination of two colors are used instead of white ones. The level of 10 hour variability for K stars is quite uncertain. We make two "guess-estimates" of it and find that it could be several times larger than the solar value. If these estimates were relevant, the color information would not provide a significant advantage. Although we do not demonstrate a need for colors, we point out the risk of an unpleasant surprise regarding the 10 hour stellar variability. Indeed, there is presently no qualified proxy for this variability. Besides, if Eddington were designed to provide this information at the cost of added complexity but not sensitivity, white photometry by channel summation would still be as efficient. Considering the risk that 10 hour variability is higher than estimated, the Precaution Imperative points to a study of practical implementations of photometry in different colors before taking irreversible decisions about the Eddington instrument.