In many cases, lux variations are accompanied by variation in the spectral shape of the sources. In blazars, brighter luxes are often accompanied by harder spectra. This ‘harder when brighter’ behav- ior of blazar has been observed in X-rays for a sample of BL Lac objects (see, e.g.,Giommi et al.,1990). In HE gamma rays, the hardening of the spectra during laring states has been signiicantly detected for FSRQs and for LSP and ISP BL Lacs (Williamson et al.,2014;Abdo et al.,2010b). As we have seen in Section 1.3, the spectral properties of the gamma rays are related to the energy distribution of the emitting particles. Under simplifying assumptions, it can be shown that if the timescale over which particles are accelerated is smaller than the one over which the particles lose their energy, the gamma- ray spectra will harden as the lux increases. Conversely, a softening of the spectra is expected in cases where the acceleration timescale is larger than the cooling time (Kirk et al.,1998). The ‘harder when brighter’ behavior is not a standard feature of blazar lares and in many cases the situation is reversed, or more complicated, or both. When monitoring the FSRQ PKS1510−089 with the LAT it was found that the spectra was softening as the lux increased. This situation changed above a certain lux value, after which the spectra started to harden for higher luxes (Abdo et al.,2010a). In the VHE regime, the situation is similar, with a positive detection of this efect (for example, seeAharonian et al.,2002), side by side with more complex cases, similar to the one of PKS 1510−089, as for exampleAbramowski et al.(2010).
Chapter 3 The high-energy gamma-ray sky
Blazars are not the only source class characterized by this ‘harder when brighter’ behavior. This be- havior has also been observed in other types of active galaxies such as the radio galaxy NGC1275 (Brown and Adams,2011). The situation for gamma-ray binaries seems to be completely reversed. In these sources, the high-energy spectra tend to soften as the lux increases (Dubus,2013).
3.2 Variability in gamma-rays
Figure 3.9: Examples of gamma-ray light curves and their periodograms (in inset plots). Top: the FSRQ 3C 454.3, middle: the Crab pulsar, bottom: the gamma-ray binary system LS I +61 303. Although at irst glance the light curve from LSI+61 303 does not look periodic, a clear spike corresponding to the orbital period of ∼ 26.5 days appears in the periodogram. The data used to produce these plot can be downloaded from the Fermi-LAT Monitored Source List page:https://fermi.gsfc.nasa.gov/ssc/data/access/lat/msl_lc/. 45
Chapter 4
The Second Fermi All-sky Variability
Analysis Catalog
This chapter presents a catalog of variable gamma-ray sources discovered in the irst 7.4 years of
Fermi-LAT data using the Fermi All-sky Variability Analysis (FAVA). This second FAVA catalog (2FAV, Abdollahi et al.,2016) contains 4547 gamma-ray lares detected with a pre-trial signiicance of 6σ over the timescale of a week and in two energy bands: 100−800 MeV and 0.8−300 GeV. Studying the positions of these transients, we identify 518 variable gamma-ray sources, 77 of which have no likely counterpart in other gamma-ray and blazar catalogs. The spectra of the 2FAV lares is studied for the diferent source classes. A signiicant hardening of the lare spectra with higher lux levels is established as a general feature of lares associated with FSRQs, and of high-energy lares associated with BL Lacs. An observed limit on the spectral hardness of the lares (Γ ≳ 1.5) translate, under a simple leptonic scenario, to a constraint on spectra of the injected electrons, which cannot be harder than p ∼ 2. In the 2FAV, both the sources and the lares are provided together with likely gamma-ray counterparts of the sources, and detailed spatial and spectral information on every lare.
We note that an initial version of FAVA was already used by the Fermi-LAT collaboration since 2012 as an online monitoring tool and to produce a irst catalog of laring sources (1FAV, Ackermann et al.,
2013a). The work presented here consisted in a major upgrade and extension of the analysis pipeline and in the data analysis and catalog preparation which lead to the 2FAV. In addition to the analysis upgrade and the construction of the catalog, I have also prepared a publication presenting the result- ing source list. The initial version of the paper has already been released (Abdollahi et al.,2016). This chapter necessarily shares the results from this publication, but expands on the description of the methods.
Chapter 4 The Second Fermi All-sky Variability Analysis Catalog
4.1 Introduction
With its large ield of view covering 20% of the sky at any moment and almost continuous operations in survey mode, the LAT is perfectly suited to monitor the gamma-ray sky and to study variable and tran- sient phenomena. To exploit this capability, the Fermi-LAT Collaboration maintains diferent analysis pipelines dedicated to the search for and monitoring of variable gamma-ray sources. Of these, FAVA and the Fermi Flare Advocate program (Ciprini and Fermi-LAT Collaboration,2012) are both blind, all-sky variability searches. The Flare Advocate variability search uses wavelet decomposition of the all-sky counts maps to initially locate the transients. With this method, shifters monitor the vari- ability of the gamma-ray sky on timescales of 6 hours and 1 day. FAVA is instead fully automated and requires no human intervention. The reports from the FAVA online monitoring are forwarded to the Flare Advocate shifters, often leading to the circulation of an Astronomer’s Telegram when new laring sources are discovered (see, e.g., Ajello et al.,2014;Kocevski et al.,2015).
The current version of FAVA consists of two sequential steps. As a irst step, a photometric analysis is used to blindly scan the entire sky and to provide a coarse localization of the transients. Statistically signiicant excesses are further analyzed, in a second step, using maximum likelihood techniques (see, e.g.,Mattox et al.,1996). Prior to this work, FAVA consisted only of the photometric analysis. This photometric analysis searches for variability by comparing, for every direction in the sky, the number of counts detected in a given time bin ∆t with the number of counts one would expect based on a long term (≫ ∆t) average emission. With respect to maximum likelihood analysis methods, the photometric analysis has several advantages:
• it is model independent: the expected counts are derived from the data itself. The photometric analysis does not require any model of the gamma-ray difuse emission. FAVA assumes only that this difuse emission does not vary on timescales comparable to the duration of the Fermi mission. When comparing the observed counts to the expected ones, any such constant term will cancel out. FAVA is also insensitive to the spectral shape of the lares.
• it is equally sensitive to both positive and negative lux variations. In the following, we will use the word ‘lare’ to refer to both absence and excess of counts.
• it is a robust technique and it is computationally inexpensive. It can be used to perform all-sky variability searches in diferent energy bands and over diferent timescales.
Despite these advantages, the usefulness of the photometric analysis is hampered by a poor localiza- tion accuracy. As the method does not rely on a it of the spatial distribution of the events, the uncer- tainties on the position of any counts excess is determined by the LAT PSF. As seen in Section 2.2.3, the 68% containment radius of the LAT PSF is ∼ 0.8◦at 1 GeV and rapidly degrades for lower energies,