Looking at Blazar Light-curve Periodicities with Gaussian Processes

Abstract

Abstract Temporal analysis of blazar flux is a powerful tool to draw inferences about the emission processes and physics of these sources. In the most general case, the available light curves are irregularly sampled and influenced by gaps, and in addition are also affected by correlated noise, making their analysis complicated. Gaussian processes may offer a viable tool to assess the statistical significance of proposed periods in light curves characterized by any sampling and noise pattern. We infer the significance of the periods proposed in the literature for two well known blazars with multiple claims of possible year-long periodicity: PG 1553 + 113 and PKS 2155–304, in the high-energy and optical bands. Adding a periodic component to the modeling gives a better statistical description of the analyzed light curves. The improvement is rather solid for PG 1553 + 113, both at high energies and in the optical, while for PKS 2155–304 at high energies the improvement is not yet strong enough to allow cogent claims, and no evidence for periodicity emerged from the analysis in the optical. Modeling a light curve by means of Gaussian processes, in spite of being relatively computationally demanding, allows us to derive a wealth of information about the data under study and suggests an original analysis framework for light curves of astrophysical interest.