Talk abstract details

The use of a Fourier transform and eventually non-linear least squares fitting is in the base of the asteroseismic studies. In pure theory the residuals obtained after prewhitening all the sinusoidal components of the signal should have a uniform distribution in frequencies which is typical of a white noise process. Nevertheless, this is not always the case for real data coming for ultra-precise photometric measurements of asteroseismic missions. Many CoRoT light curves analysed show a non-uniform frequency distribution of the residuals when no further significant signal can be detected with Fourier techniques. This phenomenon is usually interpreted as the presence of hidden weak components or as an effect, in the form of "correlated noise", of other underlaying physical phenomena.

Here we propose a new approach raising the following question: Do these time series satisfy the necessary conditions to be represented as a sum of sinusoidal components? In other words, is the Fourier series an adequate aproximation? To answer this question we extend the Stone-Weierstrass theorem to cases of discrete sampling for 10 stellar light curves coming from different space missions: CoRoT, Kepler and SOHO. Surprisingly, our results lead to a negative answer for the question posed. We present a tool to detect deviations from the expected behaviour discovering an amazingly ubiquitous periodic feature in all these data.

Download slides