Stars in the sky seem static, but when you look closely you find that the intensity and spectrum of the light they emit can vary from night to night. Some of this variability is stochastic (e.g. from periodic electromagnetic flares on the surface of the star), while some is periodic (e.g. the pulsation which arises in some classes of stars). In astronomy, these periodic stars become very important, as they give information about astrophysical processes in the star itself and let us infer the intrinsic brightness of the star, and therefore estimate the distance to the star. For example, Hubble famously utilized a particular class of these variable stars, “Cepheid Variables”, in his discovery of the expansion of the universe – now understood to be the result of the Big Bang.
Because of this importance in detecting periodicity in observations of stars, a number of automated statistical methods have been developed over the years to detect these signals in the typically noisy images that come from our telescopes. Next generation surveys such as the LSST have the potential to greatly expand the number of such objects we can detect, and use to learn about the structure of our galaxy and the universe as a whole. Unfortunately, this data, while very large and sensitive to very distant stars, is much more noisy and heterogeneous than past datasets. For this reason, past tried-and-true methods fail when applied to the types of data we expect in the next decade.
To address this, we developed a new detection of such periodicity – which we call the “multiband periogogram” – which extends the classic “Lomb-Scargle periodogram” which is well-known in the astronomical field. With it, we showed that we can expect to detect the majority of these important periodic stars with just a few months of LSST data. This type of application – developing specialized statistical methods suitable for the large datasets our field is generating – will be an essential aspect of getting the most out of these new astronomical surveys.
LSST project: http://lsst.org
Our paper preprint: http://adsabs.harvard.edu/abs/2015arXiv150201344V