Gaussian Process Modeling $\it{Fermi}$-LAT $\gamma$-ray Blazar Variability: A Sample of Blazars with $\gamma$-ray Quasi-periodicities

TL;DR

This study uses Gaussian process models to analyze Fermi-LAT gamma-ray light curves of blazars, confirming their variability as Gaussian processes and providing evidence for quasi-periodic oscillations in two specific sources.

Contribution

It demonstrates the effectiveness of Gaussian process methods in characterizing blazar gamma-ray variability and detecting QPOs, offering an alternative to traditional Fourier-based techniques.

Findings

Gamma-ray light curves are well modeled by Gaussian processes.

The PSDs of blazar gamma-ray variability follow red noise patterns.

Possible gamma-ray QPOs detected in PKS 0537-441 and PG 1553+113.

Abstract

Blazar variability may be driven by stochastic processes. On the other hand, quasi-periodic oscillation (QPO) behaviors are recently reported to be detected in $\it{Fermi}$-LAT data of blazars. However, the significances of these QPO signals given by traditional Fourier-like methods are still questioned. We analyze $\gamma$-ray light curves of the QPO blazars with two Gaussian process methods, CARMA and $\it{celerite}$, to examine the appropriateness of Gaussian processes for characterizing $\gamma$-ray light curves of blazars and the existence of the reported QPOs. We collect a sample of 27 blazars with possible $\gamma$-ray periodicity and generate their $\sim11$ years $\it{Fermi}$-LAT light curves. We apply the Gaussian process models to the $\gamma$-ray light curves, and build their intrinsic power spectral densities (PSDs). The results show that in general the $\gamma$-ray light curves can be characterized by CARMA and $\textit{celerite}$ models, indicating that $\gamma$-ray variabilities of blazars are essentially Gaussian processes. The resulting PSDs are generally the red noise shapes with slopes between $-0.6$ and $-1.7$. Possible evidence for the $\gamma$-ray QPOs in PKS 0537$-$441 and PG 1553$+$113 are found in the Gaussian process modelings.