A study of power spectral densities of real and simulated Kepler light curves

Abstract: During the last decade, the transit method has evolved to one of the most promising techniques in the search for extrasolar planets and the quest to find other earth-like worlds. In theory, the transit method is straight forward being based on the detection of an apparent dimming of the host star’s light due to an orbiting planet traversing in front of the observer. However, in practice, the detection of such light curve dips and their confident ascription to a planetary transit is heavily burdened by the presence of different sources of noise, the most prominent of which is probably the so called intrinsic stellar variability. Filtering out potential transit signals from background noise requires a well adjusted high-pass filter. In order to optimize such a filter, i.e. to achieve best separation between signal and noise, one typically requires access to benchmark datasets that exhibit the same light curve with and without obstructing noise. Several methods for simulating stellar variability have been proposed for the construction of such benchmark datasets. However, while such methods have been widely used in testing transit method detection algorithms in the past, it is not very well known how such simulations compare to real recorded light curves - a fact that might be contributed to the lack of large databases of stellar light curves for comparisons at that time. With the increasing amount of light curve data now available due to missions such as Kepler, I have here undertaken such a comparison of synthetic and real light curves for one particular method that simulates stellar variability based on scaled power spectra of the Sun’s flux variations. Conducting the respective comparison also in terms of estimated power spectra of real and simulated light curves, I have revealed that the two datasets exhibit substantial differences in average power, with the synthetic power spectra having generally a lower power and also lacking certain distinct power peaks present in the real light curves. The results of this study suggest that scaled power spectra of solar variability alone might be insufficient for light curve simulations and that more work will be required to understand the origin and relevance of the observed power peaks in order to improve on such light curve models.