Gopal Bhatta and Niraj Dhital: Event Horizon Telescope Collaboration Et Al. 2019

Draft version February 11, 2020
Typeset using LATEX default style in AASTeX62
The Nature of γ-ray Variability in Blazars

Gopal Bhatta1 and Niraj Dhital2
1 Astronomical Observatory of the Jagiellonian University
ul. Orla 171
30-244 Kraków, Poland
2 Institute of Nuclear Physics PAN
arXiv:1911.08198v2 [astro-ph.HE] 8 Feb 2020
Radzikowskiego 152
31-342, Kraków, Poland
ABSTRACT
We present an in-depth and systematic variability study of a sample of 20 powerful blazars, including
12 BL Lacs and 8 flat spectrum radio quasars, applying various analysis tools such as flux distribution,
symmetry analysis, and time series analysis on the decade-long Fermi/LAT observations. The results
show that blazars with steeper γ-ray spectral indexes are found to be more variable; and the γ-ray flux
distribution closely resembles lognormal probability distribution function. The statistical variability
properties of the sources as studied by power spectral density analysis are consistent with flicker noise
(P (ν) ∝ 1/ν) – an indication of long-memory processes at work. Statistical analysis of the distribution
of flux rise and decay rates in the light curves of the sources, aimed at distinguishing between particle
acceleration and energy dissipation timescales, counter-intuitively suggests that both kinds of rates
follow a similar distribution and the derived mean variability timescales are in the order of a few
weeks. The corresponding emission region size is used to constrain location of γ-ray production sites
in the sources to be a few parsecs. Additionally, using Lomb-Scargle periodogram and weighted wavelet
z-transform methods and extensive Monte Carlo simulations, we detected year timescale quasi-periodic
oscillations in the sources S5 0716+714, Mrk 421, ON +325, PKS 1424-418 and PKS 2155-304. The
detection significance was computed taking proper account of the red-noise and other artifacts inherent
in the observations. We explain the results in the light of current blazar models with relativistic shocks
propagating down the jet viewed close to the line of sight.
Keywords: accretion, accretion disks — radiation mechanisms: non-thermal, γ-ray — galaxies: active
— BL Lac objects, flat spectrum radio quasars — galaxies: jets, method: time series
1. INTRODUCTION
Active galactic nuclei (AGN) are the most luminous sources (L∼ 1047 erg/s) with supermassive black holes lurking
at their centers (see Event Horizon Telescope Collaboration et al. 2019, for M87 galaxy black hole). The sources are
powered by accretion on to the supermassive black holes. A small fraction of AGN (∼10%) profusely emit in radio
frequency and thereby are known as radio-loud sources. A sub-population of radio-loud AGN that eject relativistic
jets towards us are known as blazars. The sources are known to possess extreme properties such as high luminosity,
rapid flux and polarization variability. Also, blazars are the sources of the abundant nonthermal emission that is
Doppler boosted due to the relativistic effects, which makes them appear highly variable over a wide range of spatial
and temporal frequencies. Besides, the objects could be sources of extra-solar neutrinos (see IceCube Collaboration
et al. 2018a,b). The spectral energy distribution (SED) of the broadband continuum emission from the sources is
usually characterized by two distinct spectral peaks. The low energy peak, which usually lies between the radio
and the X-ray energies, is attributed to the synchrotron emission from the relativistic particles, whereas the high
energy peak, usually observed between the UV and the γ-ray energies, is believed to originate from inverse Compton
Corresponding author: Gopal Bhatta

[email protected]
2 Bhatta & Dhital
scattering of low energy photons (for recent blazar overview, see Böttcher 2019, and references therein). However,
there is no common agreement on the source of these low energy seed photons. Of the two widely discussed models,
the synchrotron self-Compton (SSC) model (e.g., Maraschi et al. 1992; Mastichiadis & Kirk 2002), assumes that the
same population of the electrons emitting synchrotron photons up-scatters these photons to higher energies; whereas,
the external Compton (EC) model assumes that the softer seed photons are provided by various regions of AGN,
such as accretion disk (AD; Dermer & Schlickeiser 1993), broad-line region (BLR; Sikora 1994), and dusty torus (DT;
Blażejowski et al. 2000).
Based on the presence of the emission lines over the continuum in their SEDs, blazars are grouped into two sub-
classes: more luminous flat-spectrum radio quasars (FSRQ) which show emission lines over the continuum, and less
powerful BL Lacertae (BL Lac) sources which show weak or no such lines. In case of FSRQs, the synchrotron peak
is in the lower frequency, and the most plausible process responsible for the high energy emission is believed to be
EC as opposed to SSC (Ghisellini et al. 1998). This is because the sources are known to have abundant seed photons
from AD, BLR and DT (Ghisellini et al. 2011). In case of BL Lac objects, the synchrotron peak lies in the optical or
X-ray regions. These constitute an extreme class of sources featuring high energy emission from a few tens of keV to
TeV energies that results from the combination of the synchrotron and IC processes. Absence of strong circumnuclear
photon fields and relatively low accretion rates could be the possible reasons behind the apparent low luminosity for
such sources. Another scheme for blazar classification is based on the frequency of the synchrotron peak (νs ), following
which blazars are either high synchrotron peaked blazars (HSP; νs > 1015 Hz), intermediate synchrotron peaked
blazars (ISP; 1014 < νs < 1015 Hz), and low synchrotron peaked blazars (LSP; νs < 1014 Hz) (Abdo et al. 2010; see
Fan et al. 2016 for similar recent classification scheme). In the blazar sequence, a scheme to unify diverse appearance
of the sources, bolometric luminosity is found to decrease along with the γ-ray emission in the direction from FSRQ
to HSP but the peak frequencies move towards higher energies (Fossati et al. 1998; Ghisellini et al. 2017). Also, while
synchrotron and γ-ray emission are comparable in HSP sources, FSRQs are mostly γ-ray (or Compton) dominant.
Blazars exhibit variability across the electromagnetic spectrum on diverse timescales that span a few minutes to a
few decades. For the reason, multi-frequency variability studies could be one of the most relevant tools that can offer
important insights into the physical conditions prevailing the innermost regions of blazar jets, including the nature of
the dominant particle acceleration and energy dissipation mechanism, magnetic field geometry, jet content, etc. There
have been numerous attempts to model the phenomenon by relating the sources of the variability to a wide range of
possible physical processes occurring either in the accretion disk and/or in the jet; the various scenarios include emission
sites at the accretion disk revolving around the supermassive black hole, various magnetohydrodynamic instabilities
in the disk and the jets, shocks traveling down the turbulent jets, and relativistic effects due to jet orientation (e.g.
Camenzind & Krockenberger 1992; Wagner, & Witzel 1995; Bhatta et al. 2013; Marscher 2014, and the references
therein). However, the exact details of the underlying processes are still under debate. In such context, study of
γ-ray variability of blazars provides us with an important tool to probe into jet dynamics, and associated particle
acceleration and energy dissipation mechanisms resulting high energy emission.
In blazars, the flux modulations due to disk processes could easily be swamped by the Doppler boosted emission
from jets. Nonetheless, the signatures of the disk modulations should, in principle, propagate along the jet through
disk-jet coupling mechanisms such that the traces of characteristic timescales related to disk processes could be revealed
through robust time series analysis. Such timescales then can be linked to the various processes in the jet as well as the
accretion disk such as dynamical, thermal, and viscous processes (Czerny 2006). For example, for blazars with typical
masses between ∼ 108 − 109 M the dynamical, thermal and viscous time scales are in the order of a few hours to a
few years. Besides, several AGN models predict quasi-periodic oscillations (QPO) in the flux with the characteristic
timescales ranging from a few hours to a few years. For instance, in the scenario of magnetic flux paradigm (see Sikora
& Begelman 2013), magnetic field at accretion disk threads black hole in launching the jets in AGN, and consequently
it gives rise to various magnetohydrodynamical instabilities at the disk-magnetosphere interface. These instabilities in
turn can produce QPOs which subsequently could propagate along modulating the jet emission, and could be observed
in the multi-frequency observations, including γ-ray light curves. In the similar context, highly polarized optical flare
discovered by Bhatta et al. (2015) might be a signature of dominance of magnetic field near blazar cores, so called
magnetically arrested disk (MAD) scenario(see Narayan et al. 2003). In observation, detection of QPOs in various
kinds of AGNs, including both radio-loud and radio-quiet, on various timescales has been reported in several works
(see Bhatta 2019, 2018, 2017; Bhatta et al. 2016c, and the reference therein). In addition, QPOs have been observed
γ-ray variability of blazars 3
4C +21.35
27
18
0
3C 273
24
12
0
3C 279
Flux (x10-7 photon/s/cm2)
34
17
0
PKS 1502+106
26
13
0
4C +38.41
20
10
0
4500 5000 5500 6000 6500 7000 7500 8000 8500
Time (MJD - 50000; day)
Figure 1. Weekly binned Fermi/LAT light curves of 5 well studied blazars. The light curves for the rest of the sources in the
sample as listed in Table 1 are presented in Appendix A.
to naturally develop in numerical studies involving simulations of the parsec scale relativistic jets (e.g. McKinney et
al. 2012).
The state-of-the-art telescopes and detectors have enabled us to obtain a fair comprehension of these fascinating
sources. In spite of the efforts to understand them, the details of processes including the nature of accretion processes,
disk-jet connection and the role the magnetic field in launching the jets are still elusive. In such context, the current
work is primarily motivated to the characterization of the statistical properties of γ-ray variability in blazars. The
sources form a dominant group of sources that prominently shine in the γ-ray band: the recent fourth Fermi Large Area
4 Bhatta & Dhital
Table 1. The source sample of the Fermi/LAT blazars
Source name 3FGL name Source class R.A. (J2000) Dec. (J2000) Red-shift FV (%) β ± ∆β
(1) (2) (3) (4) (5) (6) (7) (8)

3C 66A 3FGL J0222.6+4301 BL Lac 02h 22m 41.6s +43d 02m 35.5s 0.444 58.43 ±1.78 0.90 ±0.17
AO 0235+164 3FGL J0238.6+1636 BL Lac 02h 38m 38.9s +16d 36m 59s 0.94 95.53 ±1.12 1.40 ±0.19
PKS 0454-234 3FGLJ0457.0-2324 BL Lac 04h 57m 03.2s −23d 24m 52s 1.003 68.25 ±1.06 1.10 ±0.09
S5 0716+714 3FGL J0721.9+7120 BL Lac 07h 21m 53.4s +71d 20m 36s 0.3 62.20 ±1.05 1.00 ±0.15
Mrk 421 3FGLJ1104.4+3812 BL Lac 11h 04m 273s +38d 12m 32s 0.03 43.65±1.45 1.00 ±0.08
TON 0599 3FGL J1159.5+2914 BL Lac 11h 59m 31.8s +29d 14m 44s 0.7247 111.69±0.88 1.30 ±0.15
ON +325 3FGL J1217.8+3007 BL Lac 12h 17m 52.1s +30d 07m 01s 0.131 43.78 ±4.60 0.80 ±0.14
W Comae 3FGL J1221.4+2814 BL Lac 12h 21m 31.7s +28d 13m 59s 0.102 24.70 ±8.87 1.10 ±0.09
4C +21.35 3FGLJ1224.9+2122 FSRQ 12h 24m 54.4s +21d 22m 46s 0.432 114.91±0.59 1.10±0.12
3C 273 3FGL J1229.1+0202 FSRQ 12h 29m 06.6997s +02d 03m 08.598s 0.158 94.66±0.98 1.20±0.17
3C 279 3FGL J1256.1-0547 FSRQ 12h 56m 11.1665s −05d 47m 21.523s 0.536 104.29± 0.46 1.10±0.16
PKS 1424-418 3FGLJ1427.9-4206 FSRQ 14h 27m 56.3s −42d 06m 19s 1.522 70.44±0.69 1.5±0.13
PKS 1502+106 3FGLJ1504.4+1029 FSRQ 15h 04m 25s .0 +10d 29m 39s 1.84 90.11 ±0.70 1.3±0.10
4C+38.41 3FGL J1635.2+3809 FSRQ 16h 35m 15.5s +38d 08m 04s 1.813 92.99 ±0.72 1.2±0.15
Mrk 501 3FGL J1653.9+3945 BL Lac 16h 53m 52.2167s +39d 45m 36.609s 0.0334 33.47± 3.76 1.10±10
1ES 1959+65 3FGL J2000.0+6509 BL Lac 19h 59m 59.8521s +65d 08m 54.652s 0.048 49.55±2.84 1.10 ±0.14
PKS 2155-304 3FGL J2158.8-3013 BL Lac 21h 58m 52.0651s −30d 13m 32.118s 0.116 45.93 ±2.02 0.90±0.20
BL Lac 3FGL J2202.7+4217 BL Lac 22h 02m 43.3s +42d 16m 40s 0.068 64.10±1.05 1.0±0.10
CTA 102 3FGL J2232.5+1143 FSRQ 22h 32m 36.4s +11d 43m 51s 1.037 117.42 ±0.37 1.20±0.19
3C 454.3 3FGL J2254.0+1608 FSRQ 22h 53m 57.7s +16d 08m 54s 0.859 81.30 ± 0.30 1.30±0.17
Telescope source catalog (4FGL) contains about 60% of the γ-detected sources as the blazar class (The Fermi-LAT
collaboration 2019). Therefore, study of γ-ray emission from blazar can compliment similar studies on the origin and
propagation of high energy emission in the Universe (see Rieger 2019; Madejski, & Sikora 2016).
In this work, we carry out systematic in-depth analysis of 20 blazars utilizing decade long Fermi/LAT observations.
In Section 2, the sample of the blazars and its physical properties are listed in Table 1. In addition, data processing
method for Fermi/LAT instrument is outlined. In Section 3, several approaches to the analysis adopting various
methods including fractional variability, flux distribution, PSD and QPO are introduced, and also the results of the
analyses on the γ-ray light curves are presented. Then discussion on the results along with their possible implications
on the nature of γ-ray emission from the sources are presented in Section 4, and we summarize our conclusions at the
end in Section 5. Additionally, one table and several figures resulting from the analyses are placed in Appendices.
2. SAMPLE SOURCES AND FERMI/LAT DATA PROCESSING

Source Sample: We included most of the Fermi/Third Source Catalog (3FGL; Acero et al. 2015) sources for which
there could be weekly flux with significant test statistic (TS) value. The sources included in the study are generally
γ-ray bright (mostly TeV blazars) and consists of 12 BL Lacs and 8 FSRQs. The source names, their 3FGL catalog
name, source classification, RA, Dec and red-shift are presented in column 1, 2, 3, 4, 5 and 6, respectively, of Table 1.
Of the 20 sources, the source with the highest red-shift is PKS 1502+106 with z = 1.84, and the closest one (z = 0.03)
is Mrk 421.
The Large Area Telescope (LAT) onboard Fermi Gamma-ray Space Telescope (Fermi) is one of the most useful
instruments in the study of the Universe in high energy. It is equipped with a large effective area (> 8000 cm2 ), wide
field of view (> 2 sr) and high angular resolution (< 3.5o around 100 MeV and < 0.15o above 10 GeV). The instrument
continuously scans the sky every 90 minutes across a wide spectral energy range that spans 20 MeV to TeV energies
(Atwood et al. 2009). However, for most of the practical purposes, the data analysis is limited between the range 100
MeV–300 GeV. This is because a large point spread function (PSF) below the range and low event statistics above
the range might render unreliable results. For this work, the Fermi/LAT observations from the period 2008-08-04 –
2018-06-22 (∼ 10 years) and in the 100 MeV–300 GeV energy range were considered for the analysis. The Fermi/LAT
1ES 1959+65
80
Lognormal Fit
Normal Fit
120
BL LAC
60
FSRQ
Linear fit
100
Counts
40
80
20
Fvar
60
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Flux ( × 10−7 ph/s/cm2 )
20 40
40
Residual
● ● ●
● ●
● ● ● ●
0
● ● ● ● ● ●
●
●
● ●
● ●
20
−40
1.6 1.8 2.0 2.2 2.4 2.6
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Index Flux ( × 10−7 ph/s/cm2 )
Figure 2. Left: Fractional variability plotted against corresponding Fermi/LAT band spectral indexes of the sources. BL Lacs
and FSRQs are shown in blue and red symbols, respectively, and the black line shows the best linear fit to the data. Right:
γ-ray flux distribution (hatched black), and normal (green) and lognormal (red) PDF fitting to the histogram of the source 1ES
1959+65. Similar plots for other sources are presented in Appendix C.
observations of the sources were processed following the standard procedures of the unbinned likelihood analysis1 . In
particular, the Fermi Science Tools were used that made the use of the Fermi/LAT catalog, Galactic diffuse emission
model and isotropic model for point sources.2 As a first step, selections of the events were made using the Fermi
tool gtselect which selected only the events in a circular region of interest (ROI) of 10◦ radius centered around the
source. To minimize the contamination of γ-rays from the Earth limb, zenith angle was limited to < 90◦ . Similarly, the
Fermi tool gtmktime was used to select the good time intervals (GTI) to ensure that the satellite was operating in the
standard science mode so that only good quality of the observations enter the final analysis. After creating an exposure
map using gtexpmap and gtltcube, a source model file was created using the Python application make3FGLxml.py3 .
Subsequently, the diffuse source response was calculated using the Galactic and extra-galactic models of the diffuse
γ-ray emission, namely, gll iem v06.fit and iso P8R2 SOURCE V6 v06.txt. To generate the light curves the data
were binned in weekly bins and the task gtlike was run to carry out maximum-likelihood analysis (Mattox et al.
1996). As one of the input parameters to the likelihood analysis, spectral index of the source model was frozen to
the average index value from the 3FGL catalog. With the set of the parameters given in the input source models,
the task attempts to maximize the probability, that the models represent the observations, by fitting all the sources
within the ROI, and consequently computes significance of the γ-ray events from the source. The maximum-likelihood
test statistic, measuring significance of a detection, is given as TS = 2 × (logL1 - logL0 ), where L1 and L0 represent
the likelihood of the data given the model with and without √ a point source at the position, respectively. Then the
significance of a source detection can be expressed by ∼ T Sσ (Abdo et al. 2010). In the current work, to ensure a
robust analysis, only the observations with TS value > 10 (equivalently & 3σ) were included (see also Bhatta 2019,
2017).
3. ANALYSIS
In order to constrain the statistical variability properties of the blazars, the γ-ray light curves of the sample sources
were intensively studied applying various analysis methods including fractional variability, flux distribution, RMS-flux
relation, and symmetry analysis. Moreover, time series analysis in the form of power spectral density analysis, Lomb-
1 https://fermi.gsfc.nasa.gov/ssc/data/analysis/scitools/likelihood tutorial.html
2 https://fermi.gsfc.nasa.gov/ssc/data/analysis/scitools/
3 https://fermi.gsfc.nasa.gov/ssc/data/analysis/user/make3FGLxml.py
6 Bhatta & Dhital
Scargle periodogram, and weighted wavelet z-transform were carried out along with an extensive Monte Carlo (MC)
simulations. The description of methods and the corresponding results of the analyses are presented below.
3.1. Fractional variability

The decade-long light curves (e.g. see Figure 1), distinctly revealing the variable nature of the sources, imply that
the sources might have undergone dramatically violent episodes to result in the observed large flux variation over the
period. The average variability during the entire period can be quantified by estimating their fractional variability
(FV; a measure of normalized excess variance) given as
s
S 2 − hσerr
2 i
Fvar = 2 , (1)
hF i
for which uncertainty can be expressed as

v s
u
u 2 i2
2 hσerr 2 i
4 hσerr
2 +
σFvar = tFvar + F 2 − Fvar (2)
N hF i4 N hF i2 var
(Vaughan et al. 2003, see also Bhatta & Webb 2018). The resulting FVs for the sample sources are listed in the 7th
column of Table 1. The analysis shows that the blazar light curves display remarkable variability in the γ-ray band
with the mean (of the sources in the sample) FV of 73.37% – the mean FV of BL Lacs is 58.44 % with standard
deviation of 24.83 % and that of the FSRQs is 95.76% with a standard deviation of 16.03 %. Of the sample sources,
the most variable source is FSRQ CTA 102 (z= 1.037) with FV ∼117%. Similarly, the next most variable sources
are FSRQ 4C+21.35 and BL Lac TON 0599 with FV ∼ 115 and ∼ 111%, respectively. Whereas the least variable
source turns out to be BL Lac W Comae with just FV ∼ 25% followed by another BL Lac Mrk 501 with FV ∼ 33%.
Although it appears that in general FSRQs sources are more variable than BL Lac sources in the sample, for a stronger
conclusion, the analysis should be carried out on larger sample.
As an attempt to characterize variability properties of the sources, we also studied correlation between the FV and
the spectral indexes of the sources taken from the 3FGL catalog. In particular, the Pearson linear correlation coefficient
between the parameters turned out to be 0.61 with a p-value of 0.004. The plot between FV and the spectral index is
shown in the left panel of Figure 2. In the figure, BL Lacs and FSRQs are shown in blue and red color, respectively,
and the best linear fit, with coefficient of determination (R2 )=0.3701 and p-value=0.004, is represented by the black
line. The fit results in a steep slope of ∼ 70 signifying high sensitiveness of variability on the spectral index. Although
the sample is small, it suggests a hint of positive correlation between the quantities and encourages to carry out similar
future studies involving large number of sources. Among the sources, two outliers are easily distinguished visually:
FSRQ 3C 454.3 with flatter spectral index and high variability and BL Lac W Comae with steeper index and lowest
FV.
3.2. Flux distribution: Lognormality and RMS-flux relation

A study of the long term flux distribution of blazars can hold some of the important clues to the origin and nature
of their variability. In particular, a statistical probe of probability density function (PDF) of gamma-ray flux can
provide important insights on the nature of high energy emission processes, and thereby help constrain the underlying
processes that drive observed variability in blazars. With such a goal, we studied γ-ray flux distribution of the blazar
samples by constructing the histograms to ascertain PDF of the distribution, which can be approximated by the model
fit to the distribution of the fluxes from the long-term light curves. We mainly attempted fitting normal and lognormal
PDFs.
A normal distribution is defined by,
(x − µ)2
fnormal (x) = exp(− ) , (3)
2σ 2
where µ and σ are the mean and the standard deviation of the normal distribution, respectively, expressed in the unit
of flux, i. e., counts/sec/cm2 . Similarly lognormal distribution is defined by,
2
1 (ln x − m)
flognormal (x) = √ exp(− ) , (4)
xs 2π 2s2
Table 2. Lognormal and normal distribution fit statistics for the γ-ray flux distribution of the Fermi/LAT
sources using maximum likelihood estimation method. Similar table using weighted least-square method is
presented in Appendix B
Lognormal fit Normal fit
Source name m s LL AIC BIC µ σ LL AIC BIC
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
3C 66A -0.03 ± 0.02 0.52 ± 0.02 -369 743 751 1.12 ± 0.03 0.70 ± 0.02 -526 1057 1065
AO 0235+164 0.43 ± 0.05 0.78 ± 0.03 -439 882 889 2.16 ± 0.13 2.09 ± 0.09 -588 1181 1188
PKS 0454-234 0.72 ± 0.03 0.71 ± 0.02 -848 1700 1708 2.61 ± 0.08 1.82 ± 0.06 -952 1909 1917
S5 0716+714 0.60 ± 0.03 0.68 ± 0.02 -799 1602 1610 2.25 ± 0.06 1.43 ± 0.05 -870 1745 1753
Mrk 421 0.65 ± 0.02 0.41 ± 0.01 -603 1210 1218 2.09 ± 0.04 0.96 ± 0.03 -699 1403 1411
TON 0599 0.35 ± 0.04 0.77 ± 0.03 -536 1077 1084 2.02 ± 0.12 2.28 ± 0.09 -796 1597 1605
ON +325 -0.32 ± 0.03 0.64 ± 0.02 -289 581 590 0.86 ± 0.02 0.51 ± 0.02 -331 666 674
W Comae -0.49 ± 0.03 0.40 ± 0.02 -3 11 18 0.66 ± 0.02 0.29 ± 0.01 -35 74 80
4C +21.35 0.91 ± 0.05 0.97 ± 0.04 -854 1712 1720 4.02 ± 0.24 4.63 ± 0.17 -1101 2206 2214
3C 273 0.96 ± 0.04 0.69 ± 0.03 -727 1459 1466 3.45 ± 0.17 3.31 ± 0.12 -950 1904 1911
3C 279 1.39 ± 0.04 0.87 ± 0.03 -1342 2688 2696 6.02 ± 0.28 6.30 ± 0.20 -1637 3277 3286
PKS 1424-418 1.52 ± 0.03 0.68 ± 0.02 -1203 2410 2418 5.69 ± 0.19 4.04 ± 0.13 -1332 2667 2676
PKS 1502+106 1.01 ± 0.04 0.87 ± 0.03 -878 1760 1768 3.93 ± 0.18 3.57 ± 0.13 -1033 2071 2078
4C +38.41 0.81 ± 0.04 0.81 ± 0.03 -937 1877 1886 3.20 ± 0.14 3.01 ± 0.10 -1164 2332 2340
Mrk 501 -0.63 ± 0.02 0.45 ± 0.01 7 -10 -1 0.59 ± 0.01 0.25 ± 0.01 -20 44 52
1ES 1959+65 -0.37 ± 0.03 0.54 ± 0.02 -182 369 377 0.80 ± 0.02 0.46 ± 0.02 -266 536 544
PKS 2155-304 0.08 ± 0.02 0.47 ± 0.01 -383 769 778 1.21 ± 0.03 0.61 ± 0.02 -467 939 947
BL Lac 1.08 ± 0.03 0.66 ± 0.02 -991 1987 1995 3.63 ± 0.11 2.38 ± 0.08 -1085 2175 2183
CTA 102 1.58 ± 0.05 1.08 ± 0.04 -1309 2623 2631 8.66 ± 0.49 10.18 ± 0.35 -1589 3182 3190
3C 454.3 2.25 ± 0.05 1.00 ± 0.03 -1694 3392 3401 14.11 ± 0.53 11.49 ± 0.38 -1784 3571 3580
Note—For the normal fit µ and σ are presented in the unit of flux in 10−7 × counts/sec/cm2 , whereas for the lognormal fit
m is in the unit of natural log of flux.
where m and s are the mean location and the scale parameters of the distribution, respectively; and m is expressed in
the unit of natural log of flux.
First, we performed curve fitting to the flux histograms of the sample sources using the above two PDFs by employing
weighted least-square (WLS) method on the binned data, and the resulted fit statistics for both lognormal and normal
PDFs are listed in Appendix B. The mean of the flux, scale and reduced χ2 for the lognormal fitting are listed in column
2, 3, and 4, respectively. Similarly, the mean, standard deviation and reduced χ2 for the normal fitting are presented in
column 5, 6, and 7, respectively. Based on the reduced χ2 i.e., χ2 /dof, we find that for most of the sources lognormal
PDF fits better than normal PDF. However, in the sources AO 0235+164, ON+325, BL Lac and 3C 454.3 the reduced
χ2 for both of the PDFs are comparable and normal distribution fits provide slightly better representation. We also
studied the flux distribution using more robust maximum likelihood estimation (MLE) method, and implemented
the PDF fitting using the software package fitdistrplus 4 (Delignette-Muller & Dutang 2015) available in the MASS
library of the public domain R statistical software system. The package attempts to fit PDFs to the unbinned flux
distribution using MLE method, as opposed to the PDF fitting on the binned histogram in WLS method. The resulting
fit statistics along with the log-likelihood (LL), Bayesian information criterion (BIC) and Akaike information criterion
(AIC) quantities for the sample sources are presented in Table 2. The smaller AIC and BIC values for the log-normal
PDF suggest that, compared to the normal PDF, it is more preferable representation of blazar flux distribution.
The fitting of histogram of the source 1ES 1959+65 using normal and lognormal distributions and the corresponding
residuals are shown in the right panel of Figure 2, and similar plots for the rest of the sources in the sample are
presented in Appendix C. From the figures, it is evident that the observed flux distribution is asymmetric with a heavy
tail.
4 https://cran.r-project.org/web/packages/fitdistrplus/index.html
8 Bhatta & Dhital
3C 279 S5 0716+714
1.0
12
σXS (×10 photon/s/m )
2
10
Flux rate (×10 ph/s /cm )

2
0.5
8
2
7
6 0.0
−7
4
−0.5
2
0
0 5 10 15 20 −1.0
0 2 4 6 8
−7 2 7 2
Mean Flux (×10 photon/s/m ) mean flux (×10 ph/s/cm )
Figure 3. Left: RMS-flux relation in the FSRQ 3C 279. The magenta line represents the linear fit to the observations.
Similar figures for other sources are presented in Appendix D. Right: Distribution of rates of flux changes over the mean of the
consecutive fluxes of the BL Lac S5 0716+714.
We further tried to characterize the flux distribution and variability properties by investigating the correlation
between the flux states and source activity as represented by root mean square (RMS), commonly known as RMS-flux
relation. For the purpose, the light curves were divided into N segments of equal lengths such that each segment
contains at least 20 observations. This is to ensure that we can conduct a meaningful statistical analysis. For each
2
segment of the light curve, the Poisson noise corrected excess variance is given as σXS = S 2 −P 2
σ̄err ; where S 2 represents
2 2 n 2
the sample variance and σ̄err is the mean square of measurement error given by σ̄err = 1/n i σerr,i . From the light
curve of the source 3C 279, the RMS values for each segment are plotted against the corresponding mean flux values
in left panel Figure 3. The magenta line on the figure represents the linear fit to the observations and also serves as
a visual guide to the trend that appears on the RMS-flux plane. Similar figures for rest of the sample sources are
presented in Appendix C. We see that a linear trend distinctly appears in most of the sources. The slope parameters
from the linear fit are listed in the 2nd column of Table 3. The mean slope of the sources is 0.47; BL Lac ON +325 has
the flattest slope of 0.03 whereas another BL Lac TON 0599 shows the steepest slope 0.82. It is seen that, in general,
BL Lacs show flatter average slope 0.43 in comparison to the steeper average slope 0.56 for FSRQs. To further quantify
the correlation between the RMS and flux in the sources, Spearmann rank correlation coefficients are estimated along
with the corresponding p-values that represent the two-sided significance of its deviation from zero, which are presented
in the 3rd and 4th column of the table, respectively. As the p-values indicate, the linear correlation between the flux
and RMS seems to be a dominant trend in most of the sample sources. It is important to note that, except for the
source ON +325, all the sources in the sample display a strong linear RMS-flux relation. Similar results were reported
in the work of Kushwaha et al. (2017) studying Fermi/LAT observations of 4 AGN.
3.3. Symmetry Analysis: Flux rise and decay profiles

To investigate into the nature of particle acceleration that results in the flux rise and energy dissipation mechanism
causing flux decay in a source light curve, the distribution of positive and negative flux rates are studied by considering
the two consecutive fluxes in the light curve. Such a statistical analysis, in principle, should reveal the inherent
difference between the nature of acceleration and cooling mechanisms. To carry out the analysis, flux rates between
two fluxes consecutive in times were estimated as rate = ∆F /∆t. This provides a simple measure of how swiftly the
fluxes rise or decay in the weekly time bin. A comparison between rise and decay rates (positive and negative rates,
respectively) in all of the sources in the sample are presented in Table 4. Here for each of the sources the ∆F s are
normalized by the mean flux of the entire light curve so that the flux change rates can be expressed in percent per
day and thereby conveniently compared with the values for the other sources in the sample. We find that the positive
and negative flux change rates for each of the sources in the sample, as listed in 2nd and 3rd column of the table,
are very similar. As an illustration, the flux rise and decay rates against the mean flux for the source S5 0716+714
are plotted in the right panel of Figure 3. It is interesting to note that a similar conclusion was inferred in Abdo et
Table 3. Relation between RMS and the mean flux of γ-ray light curves of the Fermi/LAT blazars
Source slope Spearman’s r. c. (ρ) p-value

(1) (2) (3) (4)
3C 66A 0.70 0.73 0.006

AO 0235+164 0.59 0.94 < 0.001
PKS 0454-234 0.47 0.73 0.007
S5 0716+714 0.39 0.93 0.006
Mrk 421 0.65 0.78 0.001
TON 0599 0.82 0.94 < 0.001
ON +325 0.03 0.04 0.09
W Comae 0.22 0.80 0.02
4C +21.35 0.61 0.86 < 0.001
3C 273 0.77 0.90 < 0.001
3C 279 0.66 0.96 < 0.001
PKS 1424-418 0.25 0.71 0.004
PKS 1502+106 0.46 0.96 < 0.001
4C+38.41 0.48 0.96 < 0.001
Mrk 501 0.39 0.68 0.007
1ES 1959+65 0.32 0.85 0.002
PKS 2155-304 0.39 0.68 0.007
BL Lac 0.35 0.78 < 0.001
CTA 102 0.52 0.95 < 0.001
3C 454.3 0.42 0.90 < 0.001
al. (2010) in the similar studies of a large number of sources in the Fermi/LAT observations. Furthermore, in order
to see how similar/dissimilar are the distribution of decay rates from the rise rates, Kolmogorov-Smirnov (K-S) test
was performed. The K-S statistic (D), listed in the 4th column of Table 4, specifies the maximum deviation between
the cumulative distributions of the two samples. The p-value corresponding to the K-S statistic can be used to infer
whether both samples can be considered as being drawn from the same parent population. As indicated by the p-values
in the 5th column of Table 4, none of the sources have the flux rise and decay rates that are significantly different from
each other.
Moreover, any two consecutive fluxes and their mean can be used to compute rise/decay timescales as given by
1 ∆F 1
=± , (5)
τ± ∆t hF i
where ∆t = ti − ti+1 , ∆F = Fi − Fi+1 and hF i = (Fi + Fi+1 )/2. Note that this timescale can be taken as a measure
for the flux-doubling timescale. The average of such timescales along with the corresponding 1σ are listed in the 6th
column of Table 4. We find that in almost all sources the flux rise and decay timescales are very similar, and are these
are in the order of a few weeks.
3.3.1. Time Series Analysis
POWER SPECTRAL DENSITY
Discrete Fourier periodogram (DFP) of a light curve of a variable source provides a measure for the variability power
at a given temporal frequency (or, equivalently, timescale). Mathematically, it can be given by the square of absolute
value of discrete Fourier transform. For a time series x(tj ) sampled at times tj with j = 1, 2, .., n and spanning a total
duration observations, T , the DFP at a temporal frequency ν is expressed as
2
n
1 X −i2πνtj

P (ν) = x(tj ) e . (6)
n j=1
10 Bhatta & Dhital
Table 4. Symmetry analysis of γ-ray fluxes of the source light curves
source rise rate (% d−1 ) decay rate (% d−1 ) D p-value ave. timescale (d)
(1) (2) (3) (4) (5) (6)
3C 66A 7.28 ± 6.36 7.50 ± 7.23 0.12 0.33 17.51 ± 7.72

AO 0235+164 8.36 ± 6.51 7.85 ± 6.64 0.11 0.70 21.98 ± 21.67
PKS 0454-234 6.04 ± 5.61 5.76 ± 4.55 0.08 0.63 22.19 ± 11.90
S5 716+714 7.06 ± 5.92 7.11 ± 5.96 0.06 0.93 20.10 ± 13.02
Mrk 421 5.00 ± 3.16 5.00 ± 3.01 0.10 0.51 24.94 ± 12.22
TON 0599 7.39 ± 7.31 7.04 ± 7.15 0.10 0.55 20.61 ± 20.27
ON +325 10.62 ± 6.71 11.27 ± 7.78 0.09 0.71 11.84 ± 5.93
W Comae 5.82 ± 3.78 6.29 ± 5.05 0.12 0.92 31.72 ± 29.80
4C +21.35 8.45 ± 6.48 7.49 ± 6.50 0.15 0.13 21.74 ± 21.52
3C 273 8.82 ± 9.67 7.52 ± 6.21 0.11 0.52 21.98 ± 17.22
3C 279 7.77 ± 6.77 7.34 ± 8.83 0.09 0.51 21.55 ± 14.92
PKS 1424-418 3.77 ± 2.88 4.07 ± 3.25 0.11 0.25 32.22 ± 19.14
PKS 1502+106 4.98 ± 4.05 5.66 ± 5.82 0.08 0.84 28.14 ± 27.89
4C +38.41 6.52 ± 6.49 6.44 ± 6.17 0.06 0.93 22.34 ± 15.19
Mrk 501 7.52 ± 3.65 7.48 ± 3.77 0.06 0.97 16.21 ± 7.78
1ES 1959+65 7.23 ± 4.26 7.73 ± 4.74 0.11 0.70 17.29 ± 9.74
PKS 2155-304 6.13 ± 3.96 6.77 ± 4.11 0.11 0.28 19.18 ± 9.02
BL Lac 6.40 ± 4.96 6.28 ± 5.01 0.08 0.69 20.44 ± 11.96
CTA 102 6.30 ± 6.19 6.55 ± 6.13 0.10 0.42 23.59 ± 20.87
3C 454.3 4.99 ± 4.63 4.87 ± 4.97 0.08 0.60 30.86 ± 22.03
The periodograms are computed for n/2 frequencies that are evenly sampled in log-space between the minimum
νmin = 1/T and νmax = 1/2∆t, where ∆t is mean sampling step in the light curve. Moreover, the periodogram can
2
be normalized to express it in a convenient unit. In particular, if we normalize it with a factor 2T / (nx̄) , the unit
becomes (rms/mean)2 /d. The main advantage of this normalization is that the total integrated power of periodogram
is nearly equal to the variance of the light curve - a result following from the Parseval’s theorem. The distribution
of DFP over the temporal frequencies reveal variability power at the corresponding timescales, and thereby provide
information about the underlying variability structures and dominant timescales. Power spectral density (PSD) is a
mathematical function that best approximates the shape of a source periodogram. In general, blazar periodograms
have been found to be best approximated by power-law function of the form P (ν) ∝ ν −β with spectral power index β.
However, in reality true underlying PSD of a source light curve sampled at discrete times for a finite duration often
gets distorted by the effects of sampling pattern as represented by window function. Therefore, it is important that
any robust evaluation of PSD of real astronomical observations should be able to carefully untangle the effects of the
window function on the observed PSD.
Power Spectrum Response method (PSRESP; Uttley et al. 2002) is one of such methods frequently applied in the
characterization of PSD of AGN periodogram (see Bhatta 2019; Bhatta et al. 2018b, 2016c, and the references therein).
The main merits of the method are that it properly accounts a number of important issues relating to the blazar light
curves such as dominant red-noise, discrete sampling, finite observation length and uneven sampling of the light curve.
Moreover, since the nature of distribution of periodograms of unevenly spaced light curves of power-law type PSD
are not well understood, the distribution of a large number of simulated light curves, that posses similar statistical
properties such as mean, standard deviation, sampling pattern and observation duration, are utilized to compute a
measure for the goodness of fit of a model PSD.
As mentioned earlier, the sampling properties of the observed light curve can impose distorting effects on the true
underlying PSD in many ways. In particular, variability power leakage from lower to higher frequencies owing to
the limited observation period, commonly known as red noise leakage, can alter the true PSD shape by flattening
4C +38.41
4
1.0 10
ON +325
W Comae 3
10
0.8
2
10
−1
Power [rms]
Probability
0.6 10
1
0
10
0.4
−1
10
0.2
−2
10
−3
0.0 10 −4 −3 −2 −1
0.0 0.5 1.0 1.5 2.0 10 10 10 10
Power−law index Frequency [day]
−1
Figure 4. Left: Following PSRESP, distribution of probability, that given power-law PSD model best represents the source
periodograms, is plotted over spectral indexes ranging from 0 to 2 for W Comae (red) and ON +325 (blue). Right: Discrete
Fourier periodogram of the source 4C +38.41 (black), binned periodogram (red), and best fit PSD (blue). Similar plots for
other sources are presented in Appendix E.
the high frequency tail of the power-law function (Papadakis, & Lawrence 1993). During the spectral analysis of the
blazar PSD, the effects of the red noise leak were corrected through extensive MC simulations. Particularly, to address
this issue within the scheme of the PSRESP, light curves were simulated first 10 times longer than the total source
observation duration, and then divided into segments of 10 data sets (Isobe et al. 2015). Similarly, in case of the light
curve with the finite time resolution (or inadequate sampling rate), aliasing can alter the shape of PSD also leading
to the flattening of the high frequency tail of the true PSD. The effect in general can be avoided by sampling the
periodogram up to the Nyquist frequency (see Uttley et al. 2002).
To implement PSRESP5 , 10000 light curves with a 7-day bin were simulated using single power law PSD model
with spectral power index β as given by P (ν) = ν −β + C (see Timmer & Koenig 1995). The simulated light curves
were assigned the same observational properties, e.g. mean, standard deviation, observational length and uneven
sampling, and consequently periodograms for each simulated light curves were computed. The distribution of simulated
periodograms in turn were utilized to estimate the best fit model PSDs for the source γ-ray light curves. In the left
panel of Figure 4, the probability distribution over the spectral index are presented for the sources W Comae (red)
and ON +325 (blue). The figure shows that the spectral indexes corresponding to the best-fit PSD are 1.10±0.09
(W Comae) and 0.84±0.14 (ON +325 ), where the half width at half maximum (HWHM) from the Gaussian fit to
the observations are used to represent the uncertainties in the indexes. Following similar procedure, spectral indexes
corresponding to the best-fit PSD model for the sample sources are listed in the 8th column of Table 1. In addition,
figures showing the DFP (black), binned periodogram (red) and the best-fit PSD model (blue) for the sample sources
are presented in Appendix E, whereas the plot showing the best-fit PSD for the source 4C+38.41 is presented in the
right panel of Figure 4.
It is found that the periodograms of γ-ray light curves of the 20 sources are consistent with a single power-law
of the form P (ν) ∝ ν −β where the slope index ranges between 0.8–1.5. The mean PSD index of all the sources in
the sample turns out to be 1.13 with a standard deviation of 0.18. To compare between FSRQs and BL Lacs, mean
index for BL Lacs is 1.05 with standard deviation 0.17; where as the mean of FSRQs is 1.24 with a lesser spread of
standard deviation 0.13. The results also show that the source with steepest index of 1.5 is PKS 1424-418 wheres
the one with the flattest index tuns out to be ON +325. Our results are in close agreement with the recent results
1.15 ± 0.10 by Meyer et al. (2019), and also largely in agreement with the work of Nakagawa & Mori (2013) using 4
year-long Fermi/LAT observations of 15 sources. However, in a study of first 11 months of the Fermi survey involving
several blazars, Abdo et al. (2010) reported steeper average slope indexes 1.5 for FSRQs and 1.7 for BL Lacs. The
5 We have described the PSRESP method and its implementation in detail in several of our previous works, Bhatta et al. (2016b,c);
Bhatta (2017, 2019)

12 Bhatta & Dhital
S5 0716+714
8 a
0
Mrk 421
b
0
6
ON +325
Flux (x10 photon/s/cm )
2
4
-7
0
PKS 1424-418
d
24
12
0
PKS 2155-304
6.0 e
3.0
0.0
4500 5000 5500 6000 6500 7000 7500 8000 8500
Figure 5. Decade-long Fermi/LAT light curves of blazar displaying quasi-periodic oscillations. The vertical lines mark the
tentative position of the peak or centroid of the periodic flux modulation.
discrepancies can be ascribed to the difference in the method, sampling interval and total observation duration between
the two works.
QUASI-PERIODIC OSCILLATIONS
As we saw above, the periodograms of source light curves can be largely characterized by a single power-law PSD.
But if we closely look at the periodogram structures, occasionally we find peaks at some frequencies suggesting possible
presence of (quasi-) periodic signals in the observations. In fact, several sources are known to show QPOs in their
light curves in different energy bands (see Gupta 2018; Bhatta et al. 2016c, and the references therein). To cite a few
cases, blazar OJ 287 is famous for showing characteristic double-peaks in its optical light curve that re-occur after
every ∼12 years (e. g. Sillanpää et al. 1988). In the γ-ray energy band, the first case of year-scale QPO was observed
in blazar PG 1153+113 which seemed to display a ∼ 2-year periodic modulation in the Fermi/LAT observations (see
Ackermann et al. 2015). Subsequently, a number of works have reported QPOs in the γ-ray light curves of several
blazars e. g. ∼ 230 d QPO in Mrk 501 (Bhatta 2019), 34.5 d in PKS 2247-131 (Zhou et al. 2018)), ∼ 2 yr in PKS
0301-243 (Zhang et al. 2017b) and 3.35 yr in PKS 0426-380 in (Zhang et al. 2017a). In addition to these QPO studies
focused on individual sources, search for QPOs in the γ-ray light curves in a sample of γ-ray bright sources has also
been carried out in several works (e.g., see Ait Benkhali et al. 2019; Sandrinelli et al. 2017, 2016, 2014).
The periodic γ-ray flux modulations most likely originating in the blazar jets can, through long-memory processes,
carry information about the violent processes occurring at the innermost regions of AGN. At a time when the central
engines of AGN still can not be resolved by most of our current instruments, time series analysis carrying out studies of
QPOs can serve as a probe to the nature of disk-jet connection and jet ejection. With such a motivation, we analyzed
the decade-long γ-ray observations of the sample source applying the Lomb-Scargle periodogram (LSP; Lomb 1976;
Scargle 1982), one of the most efficient methods of finding QPO signals in the data with irregularities and gaps. The
method basically tries to least-square fit sine waves of the form Xf (t) = A cos ωt + B sin ωt to the observations such
that the periodogram is given according to
(P )
2 P 2
1 [ i xi cos ω (ti − τ )] [ i xi sin ω (ti − τ )]
P = P 2
+ P 2 , (7)
2 i cos ω (ti − τ ) i sin ω (ti − τ )
P P
where τ is given by tan (2ωτ ) = i sin ωti / i cos ωti . The periodogram is evaluated for Nν number of frequencies
between the minimum, νmin = 1/T , and the maximum frequencies, νmax = 1/(2∆t). The total number of frequencies
can be empirically given as Nν = n0 T νmax , where n0 can be chosen in the range of 5 − 10 (see VanderPlas 2018).
A peak centered at a temporal frequency may potentially suggest presence of periodic signal characteristic to the
corresponding timescale. Unlike strictly periodic signals which appear as sharp peaks in periodogram, QPO signals give
rise to periodogram structures that are extended over to the frequencies nearby to a central characteristic frequency. In
case of real astronomical observations that often show irregular sampling and gaps in the data, spurious peaks can arise
due to a number of factors discussed in Section 3.3.1. More importantly, in blazar light curves, which are dominated by
variability due to red-noise processes, high amplitude QPO features can arise especially in the lower frequency region
of the periodograms. Therefore, any significance estimation method should take into account this behavior along with
the other artifacts that are prevalent in finite duration time-series sampled at discrete and irregular time steps - in
other words, artifacts introduced by window function.
To further explore the transient (in frequency and amplitude) nature of the possible QPOs, we also performed
weighted wavelet z-transform (WWZ), one of the robust wavelet methods (see Foster 1996, for details)6 . The decade-
long γ-ray observations of the blazars were analyzed to search for possible periodic flux modulations using both of the
methods, i.e., the LSP and the WWZ methods. The analyses suggested presence of year-scale QPOs in some of the
objects in the source sample as listed in Table 5. The LSP diagram for these sources are presented in the left column
panels of Figure 6 and 7; whereas the the corresponding WWZ diagrams are placed on the right column panels of the
figures. Moreover, the significant periods resulted from the LSP and WWZ methods are listed in the 2nd and the 5th
column of Table 5, respectively.
The significance of the detected periodogram features were computed by employing PSRESP method during which
extensive MC simulation were performed. In particular, the spectral distribution of 10000 simulated light curves
(simulated using their corresponding best-fit PSD models) were employed to evaluate the local 90% and 99% significance
contours (see also Bhatta et al. 2016c; Bhatta 2017; Bhatta et al. 2018b; Bhatta 2019). The estimation of local
significance only makes the use of the simulated LSP distribution at the period of the detected QPO, whereas global
significance, given that we do not have a priori knowledge of the period of the detection, considers the simulated
spectral distribution at all the temporal frequencies considered (see Bell et al. 2011; Bhatta 2017, for details). The
resulting 90% (magenta) and 99% (red) significance contours in the LSP diagram are presented in the left column
6 We are skipping here some of the details including implementation of the method and the significance estimation, as we have extensive
covered the topics in our multiple previous works (see Bhatta 2019, 2017; Bhatta et al. 2016c)
14 Bhatta & Dhital
S5 0716+714
50
a
1002 d
40
Power (Arbitrary unit)
30
346 d
20
10
0
100 1000
Period (day)
Mrk 421
50
b
40
30
285 d
20
10
0
100 1000
Period (day)
PKS 2155−304
50
c 610 d
40
30
257 d
20
10
0
100 1000
Period (day)
Figure 6. Detection of quasi-periodic oscillations in the γ-ray light curves of blazars. The LS periodograms, 90% and 99%
contours are shown in blue, magenta and red curves, respectively, in the left column panels. The right column panels show the
WWZ power in colors, mean WWZ power at a given period as the black curve, and 99% significance contour as red curve.
panels of Figure 6 and 7, and the local and the global significance of dominant periods are listed in the 3rd and 4th
columns, respectively, of Table 5. Similarly, the 99% significance contours are shown in WWZ plots as red curves and
the local and the global significance of dominant periods are listed in the 6th and 7th columns, respectively, of the
table. Moreover, a brief description of each sources showing possible QPO features are presented below.
• S5 0716+714: We detected highly significant QPO at the period centered around 340 d. The tentative peaks of
the periodic oscillation are marked in the source light curve with vertical lines at a separation of ∼ 340 d as shown
in panel a) of Figure 5. The LSP (left panel) and WWZ (right panel) diagrams along with the corresponding
significance contours are presented in panel a) of Figure 6. It is interesting to note that Prokhorov & Moraghan
(2017) in their analysis including the Fermi/LAT observations from the year 2008 to 2016 also detected exactly
the same periodicity with a high significance (99%) over power law. However, in their work, Sandrinelli et al.
(2017) did not detect the QPO. It should be pointed out that, in addition to 346 d QPO, the 1002 d QPO –
possibly the 3rd harmonics – also appears to be significant (>99%) in both of the analyses. But in such case the
light curve would contain only 3 cycles, which would leave one indecisive as to the signal being real. Interestingly,
this period is close to that of the optical QPO detected by Raiteri et al. (2003) with possible interpretation that
the γ-ray QPO could be the counterpart of the optical one.
• Mrk 421: With significance greater than 99%, we detected ∼280 d periodic flux modulations in the famous blazar
Mrk 421. Both the LSP and WWZ diagrams of the source are presented in panel b) of Figure 6, and also the
tentative peaks of the oscillation are marked with vertical lines drawn at an interval nearly equal to the period
as shown in the panel b) of Figure 5. The detection supports the previous claim by Li et al. (2016) who reported
the exact same period in the γ-ray band along with the similar one in 15 GHz radio observations. Similarly,
Benitez et al. (2015) reported a similar period of 310 d in the multi-frequency (optical, hard X-ray and γ-ray)
light curves of the source. However, in the analysis presented by Sandrinelli et al. (2017), it was not found to be
significant enough.
• PKS 2155-304: We detected ∼610 d periodic flux oscillations in the blazar PKS 2155-304. The WWZ analysis
reveals that over the year the period is gradually shifting towards slightly higher frequency. We show the tentative
peaks of the oscillation of the period which are marked with vertical lines in panel d) of Figure 5. The source LSP
and WWZ spectral powers along with the significance contours are shown in panel c) of Figure 6. A number of
previous works have also reported similar periods, e. g., 700d in the optical and γ-ray (Chevalier et al. 2019), 620
d in the γ-ray (Ait Benkhali et al. 2019), 635 d in the γ-ray (Zhang et al. 2017), 640 d in the γ-ray (Sandrinelli
et al. 2016), and 625 d in the lower energy band (300 MeV–1 GeV) of the Fermi/LAT light curve (Sandrinelli et
al. 2014). Also, it is intriguing to note that this timescale is nearly double the 317 d timescale reported by Zhang
et al. (2014) in the light curves spanning 35 yrs using multiple methods such as epoch folding, the Jurkevich
periodogram and discrete correlation function. The optical QPO was also reported in (Sandrinelli et al. 2014). In
addition, we also detected ∼ 260 d QPO with a high significance (> 99.99% using both the methods), although
the flux oscillations of the period are not visually clear in the light curve plot.
• PKS 1424-418: We found 353 d periodic flux oscillations in the γ-ray flux of the source blazar PKS 1424-418
significant above 99% over the power-law noise. The possible QPO appearing in several cycles in the data is
Table 5. List of the blazars in the sample that show significant QPO in the γ-ray light curves
LSP WWZ
source period (d) local sig. ( %) global sig. (%) period (d) local sig. ( %) global sig. (%)
(1) (2) (3) (4) (5) (6) (7)
S5 716+714 346 ± 23 99.97 99.96 349 ± 27 99.982 99.980

Mrk 421 285 ± 27 99.99 99.97 287 ± 32 99.997 99.993
PKS 2155-304 610 ± 51 99.9994 99.99841 617 ± 53 99.995 99.9981
PKS 1424-418 353 ± 21 99.98 99.95 349 ± 24 99.985 99.981
ON +325 1086 ± 63 99.9986 99.9968 1081 ± 67 99.987 99.983
16 Bhatta & Dhital
PKS 1424−418
80
d
60
40
353 d
20
0
100 1000
Period (day)
ON +325
30
e
25
1064 d
20
15
10
0
100 1000
Period (day)
BL Lac
50
f
40
30
20
10
0
100 1000
Period (day)
Figure 7. Same as Figure 6.

reported here for the first time. The LSP and the WWZ diagram along with the respective significance contours
are shown in panel e) of Figure 7; also the tentative peaks of the periodic oscillations are shown with vertical
lines in panel d) of Figure 5.
• ON +325: Both the LSP and WWZ analyses of the ON +325 light curve resulted in detection of a significant
periodic oscillation with characteristic timescales of ∼1070 d. However, since only three cycles can be seen in
the entire light curve, it is not clear if it is truly a segment of QPO oscillations. We drew the tentative peaks of
the oscillations of the period which are marked with vertical lines in panel c) of Figure 5, and the distribution
of LSP and WWZ powers of the source are shown in panel e) of Figure 7. It is noted that a ∼ 4.5 year optical
QPO was claimed in the source previously (Fan et al. 2002).
• BL Lac: As shown in panel f) of Figure 7, the structure of LSP and WWZ power distributions of the source
appears to be rather complex. There does not seem to be one dominant sinusoidal component but there are
a number of possible timescales, in particular ∼ 270, 520 and 750 d. Note that the first two periods are close
to the harmonic range. However, these peaks are well below 99% contour and therefore can not be considered
significant. We note that the 680 d γ-ray QPO claimed in the work of (Sandrinelli et al. 2017) is not visible in
the analysis.
• In addition, a ∼ 330 day period QPO in the TeV blazar Mrk 501 has been reported in Bhatta (2019).
4. DISCUSSION
In this section, we present our interpretation and discussion on the results derived from the above analyses in the
light of standard model of blazars, i.e., black hole powered central engine and the extended radio jets providing grounds
for particle acceleration and energy dissipation events.
• γ-ray variability in blazars:
In the variability analysis, quantified measure of flux modulations as observed in the γ-ray light curves of the
blazar sources was provided by computing their fractional variabilities. The numerical values listed in the 7th
column of Table 1 suggest that blazar sources are distinctly characterized by their remarkable activity in the
γ-ray band. The γ-ray variable emission can be largely ascribed to the events occurring at the kilo-parsec scale
radio jets aligned within ∼ 5o to the line of sight. These jets are primarily fed with the energy that could be
extracted from the fast spinning Kerr black hole in the presence of the magnetic field at the rotating accretion
disk (Blandford et al. 2019; Blandford, & Payne 1982; Blandford, & Znajek 1977). Shocks traveling down the
jet can produce a power-law distribution of energetic electrons N (γ) ∝ γ −p such that the spectral index of the
synchrotron emission can be related as α = (p − 1)/2. These synchrotron electrons responsible for non-thermal
emission might be accelerated to the Lorentz factors as high as ∼ 106 . Owing to the violent and energetic
events prevalent in the jets, individual radio knots appear to be moving with superluminal motion with apparent
velocity up to ∼ 78c (Jorstad et al. 2017). Observed γ-ray variability could be intrinsically linked to the combined
modulations in a number of components such as distribution of high energy particles, seed photons and ambient
magnetic field at the emission region. On the other hand, it could also be linked to extrinsic (e.g. projection)
effects associated with a ‘plasma blob’ that is moving down the jet with bulk Lorentz factors (Γ) as large as ∼ 50.
In BL Lacs, large Lorentz factors could be conceivable as viable explanation for the observed high-amplitude
variability; in FSRQs, however, it might pose problem because the high energy emission is most likely produced
through inverse Compton scattering of seed photons external to the jet such that large values of Γ enhances the
energy density of these external photons in the comoving frame by ∼ Γ2 . As a result pair production process
becomes dominant which, in turn, should lead to the reduced γ-ray emission. But in BL Lacs, due to lack of
circumnuclear seed photons and SSC being dominant process to result high energy emission, the above argument
can not be applied (see Sbarrato et al. 2011). Similarly, non-thermal emission from mini-jets which are further
embedded in larger jets can also result in rapid γ-ray variability (Giannios et al. 2009). In hadronic models of
blazar emission, γ-ray variability could arise owing to variability in synchrotron emission from extremely high
(E ∼ 1019 eV) energy protons in highly magnetized (few tens of Gauss) compact regions of the jet with a moderate
Doppler factor ∼ 15 (Aharonian 2000).
The observed linear correlation between fractional variability and γ-ray spectral index suggests that the sources
with steeper spectrum exhibit greater variability. Theoretically, one might expect such a relation in several
18 Bhatta & Dhital
cases. For instance, if the emission comes from a smaller volumes than lower-energy emission volumes, as in
the radiative shock or turbulent extreme multi-zone models, the fluctuations have higher relative amplitude and
shorter time-scales. Similarly, the observed flux could be more strongly dependent on the Doppler factor when
the spectrum is steeper. In addition, a steeper spectrum indicates that the energies of the emitting particles are
close to their upper limits, so that radiative losses are more severe and perhaps their acceleration is more sporadic,
causing greater variability. The observed steep slope of the linear fit on FV-index plane strongly supports the
last scenario.
• Flux distribution:
The analysis of flux distribution of the Fermi/LAT light curves of the sample sources suggests that for most of the
sources studied in this work, the best fit PDF closely follows lognormal distribution. Similar result is obtained
by Shah et al. (2018) who used average monthly Fermi/LAT flux for 50 bright blazars. The observed log-normal
distribution of the blazar flux has been interpreted in terms of disk processes. Accordingly, log-normal flux
distribution could act as indicative of disk-jet connection in blazars. The fluctuations in the disk, contributing to
flux variability, can take place at different radii and thereby be dictated by viscosity fluctuations in accordance
with the local viscous timescales. In turn, these modulate the mass accretion rates at larger distances from black
hole. Variable emission from accretion disks owing to variable accretion rate could be driven by uncorrelated
fluctuations in the α-parameter taking place at different radii of the disk (see Lyubarskii 1997). The observed
log-normal distribution of the blazar flux suggests multiplicative coupling of these perturbations at the disk,
as opposed to additive coupling as in shot-noise-like perturbation (Arévalo, & Uttley 2006). The radiation
being relativistically beamed, γ-variability in blazars could arise due to a combination of both source intrinsic
events such as instabilities at the disk and the jet, and source extrinsic geometrical and projection effects.
Furthermore, the radiation by the up-scattered photons depend both on the population of the seed photons as
well as high energy particles that contribute to the up-scattering. In such scenario, no single variable parameter
can be considered as dominant to the variable emission, rather all possible contributing factors such as variable
magnetic field and high energy particle density, seed photon density acted upon by the particle acceleration and
diffusion processes could be coupled in a complex manner resulting in the log-normal distribution of the variable
emission from the sources.
On the other hand, normal flux distribution can be interpreted as integrated emission from individual shock or
magnetic-reconnection events occurring stochastically in large-scale turbulent jets (e.g Xu et al. 2019; Bhatta et
al. 2013). It is possible to interpret both kinds of distributions as being special cases of a more general class of
skewed distribution, such as Pareto distributions, with variable degree of skewness. In the context of relativistic
jets, such distribution could be a natural consequence of emission from Poynting flux dominated jets that hosts
mini-jets distributed isotropically within the emission region and that get ejected close to the line of sight with
a high bulk Lorentz factor ∼ 50. In such scenario, the resulting flux distribution has been found to hold the
RMS-flux relation (see Biteau, & Giebels 2012). Similarly, in the acceleration due to shock scenario, a small
perturbation in the acceleration timescale can result in variability in particle number density that is a linear
combination of Gaussian and lognormal process. Based on the relative weight associated with these processes,
it can in turn determine the dominant shape of the flux distribution (see Sinha et al. 2018, and the references
therein). If the variability in gamma-ray emission is dictated by such variability in the number density of the
accelerated particles, then it is natural for the flux distribution to appear both as Gaussian and lognormal. Such
a scenario, where both additive and multiplicative processes operate at various degrees along the extended jet,
also looks plausible.
In blazars, although we infer the variability properties from the jet emission, the primary source of variability
could still be associated with fluctuations in the disk processes. These fluctuations could then propagate through
the relativistic jets affecting the jet processes and get altered owing to the relativistic effects e.g., flux amplification
and time dilation. In blazars, although the disk emission is often completely swamped by the non-thermal
emission from jets, a careful and detailed study of flux distribution of blazars should be able to trace the origins
of variability back to the disk, and thereby constrain the disk-jet connection.
• Symmetry analysis:
We adopted a simple yet novel approach to investigate into the blazar emission regions. For the purpose, a
statistical analysis studying flux rise and decay in the γ-ray light curves of a sample of sources was performed.
The study aimed to unravel an intrinsic difference in the distribution of flux rise and decay rates which, if intrinsic
to the source, should be associated with two inherently different mechanisms, e.g. particle acceleration and energy
dissipative processes, respectively. However, as revealed by the K-S test, we do not observe significant difference
between rising and decaying profile of the flux distribution; which we find surprising and counter-intuitive as
physical mechanisms driving the flux rise due to particle acceleration mechanisms such as shocks and magnetic
reconnection should operate in different timescale from the cooling time scales due to emission processes, mainly
considered to be inverse-Compton process. In such context, it is natural to expect characteristic difference
between the flux rise and decay rates of long term γ-ray light curves. Nevertheless these two different processes
operating in various timescales could be blended over the extent of the jet such that the overall distribution takes
the form which is not easily distinguishable.
It should be pointed out that the method of symmetry analysis presented in this work differs from the one in
which the rise and decay timescales are estimated by fitting an exponential curves to well resolved individual
flares as in the works by Meyer et al. (2019), Chatterjee et al. (2012) and Abdo et al. (2010). In such case,
the asymmetry in the flare could depend upon individual flares. But in the approach adopted in this work the
results rather provide a statistical measure of the average flux doubling timescales during a wide range of flux
changes which includes both flaring and non-flaring (quiescent) states. This is reflected in the observed wide
range of timescales corresponding to the diverse rates with large standard deviations presented as errors in the
average timescales (see 6th column in Table 4). It should be noted that, in spite of the different approaches to
the analysis, the results of this work are in close agreement with that of these works which indicate no significant
asymmetry between the rise and decay flux profiles. Interestingly, similar results were reported in the studies
using long-term optical observation of the sources S5 0716+714 (see Li et al. 2017) and BL Lac (see Guo et al.
2016).
The timescale estimated using average flux rise/decay rates (τ ) can also shed light into one of the most important
issues yet unresolved in blazar physics, namely the location of γ-ray production site in reference to the central
black hole. In literature, we mainly find two compelling arguments on the location of origin of the γ-ray emission
relative to the central engine. Based on the observed rapid (a few minutes) γ-ray variability (e.g., see Ackermann
et al. 2016; Aharonian et al. 2007), it is argued that the emission should originate at compact regions close to the
central black hole (∼ 20rg ), where bulk of the gravitational potential energy of the in-falling matter is released
and processed into radiation energy. However, to avoid an eventual depletion of the γ ray photons due pair
production in a compact region, it requires a large Doppler factor, typically δ > 60 to explain the observed
γ-rays. On the other hand, most of the γ-ray flare events have been found to coincide with the ejection of radio
knots and rotation of polarization angle at the mm-VLBI cores which lie at a distance of few kilo-parsecs (kpc)
from the central engine (see e. g., Blinov et al. 2018; Jorstad, & Marscher 2016; Marscher 2016). Also as γ-ray
flaring events are commonly observed to last a few weeks, it can be argued that γ-ray emission is produced at
the parsec scale distance away from the black hole. In such a context, the results obtained from the symmetry
analysis can be used to estimate the size of the emission region where γ-ray variability arise, and thereby obtain
a lower limit for the distance between the black hole and the γ-ray emission sites. If we let r ∼ Γ2 τ c for a typical
Γ = 15 with mean τ = 22 d, we obtain ∼ 4 pc. This supports the idea that γ-ray emission could be predominantly
produced along the jets on parsec-scale distances as opposed to regions within a few tens of gravitational radii.
To reconcile both of the ideas, it can be suggested that the blazar variability as observed in the γ-ray light curves
could be a combination of the variable emission originating at both the locations, i. e., the low-amplitude fast
variability might chiefly originate at the innermost regions – where conversion of gravitational potential energy
of in-falling matter into high energy emission is most efficient – and the γ-ray flaring events, flux brightening at
least by a factor of a few tens, that last about a few weeks could be located at a distance of a few pc.
• Power spectral density:

We find that the PSDs that best represent periodogram of γ- ray light curves of the 20 well-known sources are
consistent with a single power-law of the functional form P (ν) ∝ ν −β where the slope index ranges between
0.8–1.5. In the given sample source light curves, majority of the slope indexes tend to center around 1.0. Similar
results were obtained by Sobolewska et al. (2014) in their PSD analysis of γ-ray light curves of 13 blazars,
although the work followed different method of PSD estimation. This is interesting because β=1, often known as
flicker noise, are exactly halfway between random walk (β=2) and white noise (β=0), and are prevalent in nature
20 Bhatta & Dhital
(see Press et al. 1978). Flicker noise diverges when integrated from a finite high frequency to lower frequencies -
towards zero frequency. But the divergence, being logarithmic, is so slow that the noise maintains its appearance
over several orders of frequencies up to arbitrarily low values. Therefore, flicker noises are long-memory processes
and therefore can appear coherent over several decades in timescale. In case of blazars, although our instrument
primarily detects Doppler boosted emission from the jets, it can possess memory of the events occurring at the
accretion disk, especially the disk modulations such as changes in the accretion rates, viscosity, magnetic field
etc., that could be coupled with the jet processes such that the disk instabilities could drive the jet emission
variability. In other words, jet emission might “remember” disk processes, and this indicates strong disk-jet
connection.
In general, the power-law type PSD seen in most power spectra of blazar variability can also be explained in the
context of a turbulent flow behind propagating shock (Marscher et al. 1992) or a standing/reconfinement shock
in blazar jets (Marscher et al. 2008). If the emission from a single dominant turbulent cell get enhanced due to
Doppler boosting, it contributes to the temporal frequency corresponding to the size or velocity of the cell. The
stochastic nature of turbulence implies that cells of various sizes will be Doppler enhanced over time depending
on their velocity and angle to the line of sight. Eventually, this will result in a variability spectrum over wide
range of temporal frequencies that is consistent with the power-law noise seen in blazars (see Wiita 2011). In a
similar context, the magnetic field at the accretion disk could be fairly magnetized owing to the material accreted
over a considerable period of time. In such event, the magnetic field can extract the vast rotational energy by
threading the black hole and channel into the jet as the bulk power of the relativistic jets. Moreover, as the
radiation power is only 10% of the total jet power, a significant contribution to the jet contents could be provided
by poynting flux (Ghisellini et al. 2014), which then can facilitate the rampant magnetic reconnection events
triggering stochastic particle acceleration and energy dissipation at various temporal and spatial scales. If the
observed variable γ-ray emission is produced in such a scenario, the variability power spectrum should closely
resemble power-law shape.
• Quasi-periodic oscillations:
We found the presence of year timescale QPOs in some of the sources with a high significance over the power-law
PSDs. The detected γ-ray QPOs can potentially offer profound insights on the nature of high energy emission
processes taking place in the sources. In particular, the studies can shed light on a number of current blazar
issues such as disk-jet connection, origin of relativistic jets from the central engine, and other extreme conditions
near the fast rotating supermassive black holes. In principle, origin of QPOs can be conceived of mainly in three
scenarios: supermassive binary black holes (SMBBH) system, accretion disk and jet instabilities. Some of the
possible explanation of the origins of QPOs are discussed below.
– SMBBH system: In the context of SMBBH system, the observed timescales can be interpreted as the Kep-
lerian periods of the secondary black hole around the central black hole as given by T = 2πa3/2 (G M )−1/2
with M = Mp + Ms ∼ 109 M , where Ms and Mp are the corresponding masses, respectively, and a is the
length of the semi-major axis of the elliptic orbit. Over the long course of merging galaxies, the dynamical
friction present in the system can gradually smooth the elliptical orbits into circular orbits. Then assuming
a typical AGN total mass of M = Mp + Ms ∼ 109 M and a mass ratio Ms /Mp in the rage 0.1–0.01, the
separation between the black holes can be estimated in the order of a few parsecs. Such a binary systems
can undergo orbital decay due to emission of low frequency (a few tens of nano-Hertz) gravitational waves
(GW), which could be detected by future GW missions. For such a system, the GW-driven orbital decay
timescale can be estimated applying
−3 4
M a
τinsp = 3.05 × 10−6 years, (8)
109 M rg
(see Peters 1964), a few thousands years, rather short span of time relative to cosmic timescales.
– Accretion disk: Year time scale periodicity in blazars can be explained in the context of instabilities intrinsic
to the accretion disk. To modulate flux periodically, a bright hotspot could be revolving around the central
black hole with a Keplerian period, τk , given by
−1/2 3/2
M a
τk = 0.36 days, (9)
109 M rg
Period
10000 1000 100 10 1
8
10
P4 P3 P2 P1
6
10
Power
4
10
2
10
0
10
10000 1000 100 10 1
Period
Figure 8. Discrete Fourier periodogram of evenly space light curve simulated applying model PSD with spectral index of 1.5.
Sinusoidal waves of periods 10, 100, 500 and 2000 in arbitrary time unit are superimposed on the pure power-law PSD. The
blue line represents the log-linear fit to the periodogram and the red line shows the 99% significance contour.
where a is the length of the semi-major axis of the elliptic orbits. Assuming circular orbits, for a typical
black hole of mass of 109 M the radius of the Keplerian orbit for a year timescale can be estimated to be a
few tens of gravitational radius (rg ). Similarly, in the case of globally perturbed thick accretion disks, the
disk can undergoes p-mode oscillations with a fundamental frequency that can be approximated as,
−3/2 −1
r M
f0 ≈ 100 day−1 (10)
rg 108 M
(see An et al. 2013, and the reference therein).
To include the effects of strong gravitational field near a fast spinning supermassive black hole, the frame
dragging effect can warp the inner part of the accretion disk. This might lead to the nodal precession of
the tilted plane of the disk better known as the Lense-Thirring precession. The period of such precession
can be expressed as
3
1 M r
τLT = 0.18 9
days, (11)
as 10 M rg
where as , M and r represent dimensionless spin parameter, mass of the black hole and the radial distance
of the emission region from the black hole, respectively. For a 109 M black hole with a maximal spin
(as = 0.9), a year timescale would correspond to the inner part of accretion disk extending in the order of
a few tens of rg . In blazars, the precession of the disk also can lead to jet as precession thereby resulting in
the periodic emission (e.g Liska et al. 2018).
– Jets: The observed quasi-periodic flux modulations can also be linked to relativistic motion of the emission
regions along helical path of the magnetized jets (e.g., Camenzind & Krockenberger 1992). In particular,
when emission regions move along the helical path of a jet with a large bulk Lorentz factor, Γ, relativistic
effects become dominant such that periodic flux modulations can appear due to the periodic changes in the
viewing angle. In such scenario, The rest frame flux (Fν0 0 ) and observed flux (Fν ) are related through the
relations
Fν (ν) = δ(t)3+α Fν0 0 (ν) and δ(t) = 1/Γ (1 − βcosθ(t)) . (12)
If we let intrinsic flux of the emission region unchanged but change viewing angle, the corresponding ratio
of the observed flux to the intrinsic flux for a given change in the angle ∆θ can be expressed as
∆logF = − (3 + α) δΓβsinθ∆θ, (13)
For illustrative purpose, for blazars emission with typical γ-ray spectral index (α = 1.5) and viewing angles
in the 1 − 5o range, a slight change in the viewing angle e.g., ∼ 1o , is sufficient to produce an apparent flux
22 Bhatta & Dhital
twice as bright (see Bhatta 2018 Figure 4). Similarly, QPOs can originate in blazars jets owing to recurring
boosts of turbulent cells behind a propagating shock. If a biggest dominant structure stands out, it will
exhibit enhanced Doppler boosting contributing to QPO component. However, it is possible that, due to
the stochastic nature of turbulence, the cell would gradually decay causing the amplitude of the QPO to
diminish accordingly over a period of time (see Wiita 2011).
It should be stressed that the dominance of red-noise in blazar light curves often gives rise to a general skepticism
towards the actual presence of the QPOs in blazar, particularly QPOs at the low-frequency (LF) ends frequently
reported in the literature. Consequently, many authors tend to adopt a conservative measure for the significance,
such as ' 99.99% over the PSD, required to establish their existence. However, we argue that if we take such a
strict approach towards the significance, there could be a risk of overlooking many interesting features in AGN
and thereby we may miss exciting physics. To illustrate our point we present periodogram of a pure power-law
of spectral index 1.5 on which purely sinusoidal waves of the periods 10, 100, 500, and 2000 days but of the same
intensity are superimposed as shown in Figure 8. The simulated light curve is evenly spaced so that the 99%
significance is computed using Equation 16 in Vaughan (2005). The figure shows that for the same amplitude of
the periodic modulation, the significance of the peaks gradually decreases as we move from HF to LF such that
even in a relatively ideal situation of purely sinusoidal modulations present in the evenly spaced observations,
the corresponding spectral peaks can get drowned into the strong power-law trend which is ever-rising, and
consequently fail to pass the 99 % significance test. Similar situation might arise when LF QPOs are unable
to maintain phase coherence over more a few oscillations. In such cases, performing statistical analysis using
multiple methods e.g. carry out both frequency and time-based analysis (see Bhatta 2019) would be more useful.
Furthermore, a year-scale QPOs could arise in various scenarios discussed above (see also Bhatta 2019, 2018,
2017; Ackermann et al. 2015); now it is a challenging task to break the apparent degeneracy in the models to
single out the actual process behind the detection. The task would require an in-depth analysis of multi-frequency
light curves applying multiple approaches to the time series analysis.
5. CONCLUSION
We performed an in-depth time domain analysis of decade long (2008-2018) Fermi/LAT light curves of a sample of
20 bright blazars. We found that γ-ray emission from blazars is highly pronounced and variable over diverse timescale.
As one of characteristic features, a steep linear trend was observed in the correlation between fractional variability
and the γ-ray spectral index suggesting that the variability is highly sensitive to its spectral slope. The γ-ray flux of
the blazars is found to be distributed in a way that is closely approximated as lognormal PDF. Statistical analysis of
flux rising and decay rates in the γ-ray light curves show that both the distribution are very similar and therefore no
significant asymmetry between the flux rising and decay profiles was detected. Moreover, most of the sources appear
to exhibit a linear RMS-flux relation indicating higher flux states are often more variable. Furthermore, to constrain
the statistical nature of such variability over a wide range of temporal frequencies, extensive MC simulations were
performed to estimate the PSDs which best represent the blazar γ-ray periodogram. The study shows that the PSDs
are consistent with a single power-law, P (ν) ∝ 1/ν, with spectral indexes centered around 1.0 indicating the nature of
variability as flicker-noise and, therefore, might be driven by long-memory processes. Additionally, a closer inspection
of the Lomb-Scargle and WWZ periodograms of some of the sources in the sample, including S5 0716+714, Mrk 421,
ON +325, PKS 1424-418 and PKS 2155-304, reveal spectral features that signify presence of year timescales QPOs
that are highly significant over the possible artifacts usually found in blazar light curves.
GB acknowledges the financial support by the Narodowe Centrum Nauki (NCN) grant UMO-2017/26/D/ST9/01178.
We would like to express our gratitude to Prof. Staszek Zola for kindly allowing us to use their computational facility
for this research. We would also like to thank Prof. Michal Ostrowski and Prof. Alan Marscher for fruitful discussion
on the γ-ray variability of blazars. The authors are grateful to the anonymous referee for his/her comments that
significantly improved the quality of the work.
Facilities: Fermi/LAT
Software: HEAsoft(HEASARC2014),FTOOLS(Blackburn1995),andfitdistrplus(Delignette-Muller&Dutang2015)
REFERENCES
Abdo, A. A., Ackermann, M., Ajello, M., et al. 2010, ApJ, Blinov, D., Pavlidou, V., Papadakis, I., et al. 2018,
722, 520 MNRAS, 474, 1296
Abdo, A. A., Ackermann, M., Ajello, M., et al. 2010, ApJS, Böttcher, M. 2019, Galaxies, 7, 20
188, 405 Camenzind M., Krockenberger M., 1992, A&A, 255, 59
Ackermann, M., Anantua, R., Asano, K., et al. 2016, ApJL, Chatterjee, R., Jorstad, S. G., Marscher, A. P., et al. 2008,
824, L20 ApJ, 689, 79
Ackermann, M., Ajello, M., Albert, A., et al. 2015, ApJL, Chatterjee, R., Bailyn, C. D., Bonning, E. W., et al. 2012,
813, L41 ApJ, 749, 191
Acero, F., Ackermann, M., Ajello, M., et al. 2015, ApJS, Chevalier, J., Sanchez, D. A., Serpico, P. D., et al. 2019,
218, 23 MNRAS, 484, 749
Aharonian, F. A. 2000, New Astronomy, 5, 377. Czerny, B. 2006, Astronomical Society of the Pacific
Aharonian, F., Akhperjanian, A. G., Bazer-Bachi, A. R., et Conference Series, 360, 265
al. 2007, ApJL, 664, L71 Delignette-Muller, M. L. & Dutang, C. 2015, Journal of
Alston, W. N., Parker, M. L., Markeviciute, J., et al. 2015, Statistical Software, 64, 4
MNRAS, 449, 467 Dermer, C. D., & Schlickeiser, R. 1993, ApJ, 416, 458
An, T., Baan, W. A., Wang, J.-Y., Wang, Y., & Hong,
Event Horizon Telescope Collaboration, Akiyama, K.,
X.-Y. 2013, MNRAS, 434, 3487
Alberdi, A., et al. 2019, ApJL, 875, L6
Arévalo, P., & Uttley, P. 2006, MNRAS, 367, 801
Fan, J. H., Yang, J. H., Liu, Y., et al. 2016, ApJS, 226, 20
Atwood, W.B., Abdo, A.A. et al., 2009, ApJ, 697, 1071
Fan, J. H., Lin, R. G., Xie, G. Z., et al. 2002, A&A, 381, 1
Bell, M. E., Tzioumis, T., Uttley, P., et al. 2011, MNRAS,
The Fermi-LAT collaboration 2019, arXiv e-prints,
411, 402
arXiv:1902.10045
Ait Benkhali, F.; Hofmann, W.; Rieger, F. M., et al. 2019,
Fossati, G., Maraschi, L., Celotti, A., Comastri, A., &
2019arXiv190110246A
Ghisellini, G. 1998, MNRAS, 299, 433
Benitez, E., Cabrera, J. I., Fraija, N., et al. 2015, arXiv
Foster G., 1996, AJ, 112, 1709
e-prints, arXiv:1512.01219
Fu, W., & Lai, D. 2012, MNRAS, 423, 831
Bhatta, G. 2019, MNRAS
Rieger, F. 2019, Galaxies, 7, 28
Bhatta, G. 2018, Galaxies, 6, 136
Ghisellini, G., Righi, C., Costamante, L., & Tavecchio, F.
Bhatta, G., Mohorian M., and Bilinsky I. 2018 A&A, 619,
2017, MNRAS, 469, 255
A93
Bhatta, G., Stawarz, L., Markowitz, A., et al. 2018, ApJ, Ghisellini, G., Tavecchio, F., Maraschi, L., et al. 2014,
866, 132 Nature, 515, 376
Bhatta, G., & Webb, J. 2018, Galaxies, 6, 2 Ghisellini, G., Tavecchio, F., Foschini, L., & Ghirlanda, G.
Bhatta, G. 2017, ApJ, 847, 7 2011, MNRAS, 414, 2674
Bhatta, G., Zola S., Stawarz, L., et al. 2016c, ApJ, 832, 47 Ghisellini, G., Della Ceca, R., Volonteri, M., et al. 2010,
Bhatta, G., Stawarz, L., Ostrowski, M., et al. 2016b, ApJ, MNRAS, 405, 387
831, 92 Ghisellini, G., Celotti, A., Fossati, G., et al. 1998, MNRAS,
Bhatta, G., Goyal, A., Ostrowski, M., et al. 2015, ApJL, 301, 451
809, L27 Giannios, D., Uzdensky, D. A., & Begelman, M. C. 2009,
Bhatta, G., et. al. 2013, A&A, 558A, 92B MNRAS, 395, L29
Biteau, J., & Giebels, B. 2012, A&A, 548, A123 Guo, Y. C., Hu, S. M., Li, Y. T., et al. 2016, MNRAS, 460,
Blandford, R., Meier, D., & Readhead, A. 2019, ARA&A, 1790
57, 467 Gupta, A. 2018, Galaxies, 6, 1
Blandford, R. D., & Payne, D. G. 1982, MNRAS, 199, 883 Zhang, B.-K., Zhao, X.-Y., Wang, C.-X., et al. 2014,
Blandford, R. D., & Znajek, R. L. 1977, MNRAS, 179, 433 Research in Astronomy and Astrophysics, 14, 933941
Blażejowski, M., Sikora, M., Moderski, R., & Madejski, IceCube Collaboration, Aartsen, M. G., Ackermann, M., et
G. M. 2000, ApJ, 545, 107 al. 2018, Science, 361, eaat1378
24 Bhatta & Dhital
IceCube Collaboration, Aartsen, M. G., Ackermann, M., et Raiteri, C. M., Villata, M., Tosti, G., et al. 2003, A&A,
al. 2018, Science, 361, 147 402, 151
Isobe, N., Sato, R., Ueda, Y., et al. 2015, ApJ, 798, 27 Sandrinelli, A., Covino, S., & Treves, A. 2014, ApJL, 793,
Jorstad, S. G., Marscher, A. P., Morozova, D. A., et al. L1
2017, ApJ, 846, 98 Sandrinelli, A., Covino, S., & Treves, A. 2016, ApJ, 820, 20
Jorstad, S., & Marscher, A. 2016, Galaxies, 4, 47 Sandrinelli, A., Covino, S., Dotti, M., & Treves, A. 2016,
Kastendieck, M. A., Ashley, M. C. B., & Horns, D. 2011, AJ, 151, 54
A&A, 531, A123 Sandrinelli, A., Covino, S., Treves, A., et al. 2017, A&A,
Kushwaha, P., Sinha, A., Misra, R., et al. 2017, ApJ, 849, 600, A132
138 Sbarrato, T., Foschini, L., Ghisellini, G., et al. 2011,
Li, Y. T., Hu, S. M., Jiang, Y. G., et al. 2017, PASP, 129, Advances in Space Research, 48, 998
14101 Scargle, J. D. 1982, ApJ, 263, 835
Li, H. Z., Jiang, Y. G., Guo, D. F., et al. 2016, PASP, 128, Shah, Z.; Mankuzhiyil, N.; Sinha, A. et al. 2003, PASJ,
074101 55L3
Li, L.-X., & Narayan, R. 2004, ApJ, 601, 414 Sikora, M. 1994, ApJS, 90, 923
Liska, M., Hesp, C., Tchekhovskoy, A., et al. 2018, Sikora, M., & Begelman, M. C. 2013, ApJL, 764, L24
MNRAS, 474, L81 Sillanpää A., Haarala S., Valtonen M. J., Sundelius B.,
Liu, F. K., Zhao, G., & Wu, X.-B. 2006, ApJ, 650, 749 Byrd G. G., 1988, ApJ, 325, 628
Liu, T., Gezari, S., Heinis, S., et al. 2015, ApJL, 803, L16 Singal, J., Ko, A., & Petrosian, V. 2014, ApJ, 786, 109
Lomb, N. R. 1976, Ap& SS, 39, 447 Sinha, A., Khatoon, R., Misra, R., et al. 2018, MNRAS,
Lyubarskii, Y. E. 1997, MNRAS, 292, 679 480, L116
Madejski, G. (Greg) ., & Sikora, M. 2016, ARA&A, 54, 725 Sobolewska, M. A., Siemiginowska, A., Kelly, B. C., &
Maraschi, L., Ghisellini, G., & Celotti, A. 1992, ApJL, 397, Nalewajko, K. 2014, ApJ, 786, 143
L5 Tchekhovskoy, A., Narayan, R., & McKinney, J. C. 2011,
Mastichiadis, A., & Kirk, J. G. 2002, PASA, 19, 138 mnras, 418, L79
Marscher, A. 2016, Galaxies, 4, 37 Timmer, J., & Koenig, M. 1995, A& A, 300, 707
Marscher, A. P. 2014, ApJ, 780, 87 Uttley, P., McHardy, I. M., & Papadakis, I. E. 2002,
Marscher, A. P., Jorstad, S. G., D’Arcangelo, F. D., et al. MNRAS, 332, 231
2008, Nature, 452, 966 VanderPlas, J. T. 2018, ApJS, 236, 16
Marscher, A. P., Gear, W. K., & Travis, J. P. 1992, Vaughan, S. 2005, A&A, 431, 391
Variability of Blazars, 85 Vaughan, S., Edelson, R., Warwick, R. S., & Uttley, P.,
Mattox, J. R., Bertsch, D. L., Chiang, J., et al. 1996, ApJ, 2003, MNRAS, 345, 1271
461, 396 Wang, J.-Y., An, T., Baan, W. A., & Lu, X.-L. 2014,
McKinney, J. C., Tchekhovskoy, A., & Blandford, R. D. MNRAS, 443, 58
2012, MNRAS, 423, 3083 Wagner, S. J., & Witzel, A. 1995, ARA&A, 33, 163
Meyer, M., Scargle, J. D., & Blandford, R. D. 2019, ApJ, Wiita, P. J. 2011, Journal of Astrophysics and Astronomy,
877, 39 32, 147
Nakagawa, K., & Mori, M. 2013, ApJ, 773, 177 Wu, J., Zhou, X., Wu, X.-B., et al. 2006, AJ, 132, 1256
Narayan, R., Igumenshchev, I. V., & Abramowicz, M. A. Xie, G. Z., Yi, T. F., Li, H. Z., Zhou, S. B., & Chen, L. E.
2003, PASJ, 55, L69 2008, AJ, 135, 2212
Papadakis, I. E., & Lawrence, A. 1993, MNRAS, 261, 612 Xu, J., Hu, S., Webb, J. R., et al. 2019, ApJ, 884, 92
Papadakis, I. E., Reig, P., & Nandra, K. 2003, MNRAS, Zhang, B.-K., Zhao, X.-Y., Wang, C.-X., & Dai, B.-Z. 2014,
344, 993 Research in Astronomy and Astrophysics, 14, 933-941
Peters, P. C. 1964, Physical Review, 136, 1224 Zhang, P.-f., Yan, D.-h., Liao, N.-h., & Wang, J.-c. 2017,
Press, W. H. 1978, Comments Astrophys., 7, 103 ApJ, 835, 260
Prokhorov, D. A., & Moraghan, A. 2017, MNRAS, 471, Zhang, P.-f., Yan, D.-h., Liao, N.-h., et al. 2017, ApJ, 842,
3036 10
Raiteri, C. M., Villata, M., Aller, H. D., et al. 2001, A&A, Zhang, P.-F., Yan, D.-H., Zhou, J.-N., et al. 2017, ApJ,
377, 396 845, 82
Zheng Z.-Y., Butler N. R., Shen Y. et al., 2016, ApJ, 827,

56
Zhou, J., Wang, Z., Chen, L., et al. 2018, Nature
Communications, 9, 4599
26 Bhatta & Dhital
APPENDIX
A. DECADE-LONG FERMI/LAT LIGHT CURVES OF BLAZARS
Mrk 501
1.7
0.9
0.0
3 1ES 1959+65
0
BL Lac
Flux (x10 photon/s/cm )
2
22
11
-7
0
CTA 102
44
22
0
3C 454.3
66
33
0
4500 5000 5500 6000 6500 7000 7500 8000 8500
Figure 9. Weekly binned Fermi/LAT light curves of the sample blazar listed in Table 1
3C 66A
0
AO 0235+164
12
0
PKS 0454-234
Flux (x10-7 photon/s/cm2)
10
0
TON 0599
16
0
W Comae
1.9
1.0
0.0
4500 5000 5500 6000 6500 7000 7500 8000 8500
Figure 9 (Cont.).
28 Bhatta & Dhital
B. FITTING OF BLAZAR FLUX DISTRIBUTION USING WEIGHTED LEAST SQUARE METHOD
Table 6. Lognormal and normal distribution fit statistics for the γ-ray flux distribution of
the Fermi/LAT sources using weighted least square method.
Lognormal fit Normal fit
Source name m s χ2 /dof µ σ χ2 /dof
(8) (9) (10) (11) (12) (13) (14)

3C 66A -0.03 ± 0.02 0.52 ± 0.02 4.19/6 0.81 ± 0.07 0.75 ± 0.07 24.58/6
AO 0235+164 0.34 ± 0.08 0.90 ± 0.09 7.63/10 -24.37 ± 2.07 6.73 ± 0.53 11.66/10
PKS 0454-234 0.72 ± 0.04 0.75 ± 0.03 17.04/17 0.82 ± 0.51 2.54 ± 0.27 15.21/17
S5 0716+714 0.63 ± 0.04 0.72 ± 0.03 20.02/16 1.63 ± 0.14 1.63 ± 0.12 34.72/16
Mrk 421 0.62 ± 0.02 0.35 ± 0.01 36.22/16 1.94 ± 0.04 0.71 ± 0.03 56.54/16
TON 0599 0.25 ± 0.07 0.79 ± 0.09 19.27/12 -19.07 ± 1.42 5.33 ± 0.36 27.05/12
ON +325 -0.32 ± 0.03 0.62 ± 0.03 25.45/9 0.78 ± 0.03 0.48 ± 0.03 23.88/9
W Comae -0.51 ± 0.03 0.40 ± 0.02 11.25/10 0.61 ± 0.02 0.21 ± 0.02 41.99/10
4C +21.35 0.83 ± 0.09 1.11 ± 0.10 13.12/13 -47.73 ± 3.67 12.97 ± 0.93 23.43/13
3C 273 0.84 ± 0.04 0.70 ± 0.06 25.98/12 -34.07 ± 2.64 9.52 ± 0.65 39.21/12
3C 279 1.30 ± 0.04 0.87 ± 0.06 39.84/16 -63.96 ± 4.19 17.56 ± 1.06 57.24/16
PKS 1424-418 1.52 ± 0.03 0.66 ± 0.02 8.54/15 4.16 ± 0.26 3.51 ± 0.52 73.60/15
PKS 1502+106 1.08 ± 0.05 0.85 ± 0.05 8.26/10 -51.32 ± 5.73 14.17 ± 0.99 13.33/10
4C +38.41 0.64 ± 0.07 1.01 ± 0.07 24.30/13 -37.57 ± 2.98 10.25 ± 0.75 38.12/13
Mrk 501 -0.63 ± 0.02 0.45 ± 0.02 20.36/15 0.54 ± 0.02 0.27 ± 0.01 15.55/15
1ES 1959+65 -0.39 ± 0.03 0.53 ± 0.02 29.06/15 0.71 ± 0.03 0.35 ± 0.03 98.54/15
PKS 2155-304 0.09 ± 0.02 0.48 ± 0.02 16.45/9 1.10 ± 0.03 0.48 ± 0.03 42.73/9
BL Lac 1.03 ± 0.04 0.71 ± 0.03 35.63/17 1.05 ± 0.69 3.41 ± 0.32 23.65/17
CTA 102 1.49 ± 0.09 1.22 ± 0.11 18.90/15 -97.12 ± 11.90 26.18 ± 2.98 40.00/15
3C 454.3 2.34 ± 0.06 1.00 ± 0.05 52.79/12 -33.57 ± 28.02 27.80 ± 7.17 46.30/12
Note—For the normal fit µ and σ are presented in the unit of flux in 10−7 × counts/sec/cm2 , whereas for the lognormal fit m is in the unit of
natural log of flux.
C. PROBABILITY DENSITY FUNCTION FOR BLAZAR FLUX DISTRIBUTION
3C 66A AO 0235+164 PKS 0454−234

Lognormal Fit Lognormal Fit Lognormal Fit
80
200
Normal Fit Normal Fit Normal Fit
100
80
60
150
Counts
Counts
Counts
60
100
40
40
20
50
20
0
0
0 2 4 6 8 10 0 5 10 15 0 2 4 6 8 10
Flux ( × 10−7 ph/s/cm2 ) Flux ( × 10−7 ph/s/cm2 ) Flux ( × 10−7 ph/s/cm2 )
20 40
50
50
Residual
Residual
Residual
● ● ●
● ● ● ● ● ● ● ● ● ● ●
0
0
● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
● ● ● ●
● ● ●
●
●
−50
−50
−40
0 2 4 6 8 10 0 5 10 15 0 2 4 6 8 10
S5 0716+714 Mrk 421 TON 0599

80

100
150
60
80
100
Counts
Counts
Counts
60
40
40
50
20
20
0
0 2 4 6 8 0 2 4 6 8 0 5 10 15 20
30
20 40
●
50
Residual
Residual
Residual
10
● ● ● ●
● ●
● ●
● ● ● ● ● ● ● ● ●
● ●
0
● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ●
● ● ● ●
● ● ●
−10
● ● ●
●
●
−100
−40
−30
0 2 4 6 8 0 2 4 6 8 0 5 10 15 20
ON +325 W Comae 4C +21.35

120
150

50
100
40
100
80
Counts
Counts
Counts
30
60
20
40
50
10
20
0
0 1 2 3 4 5 0.0 0.5 1.0 1.5 2.0 0 5 10 15 20 25 30

40
10 20
20
50
Residual
Residual
Residual
● ● ●
● ● ●
● ●
● ● ● ● ● ●
● ● ●
0
● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
● ● ● ● ●
● ● ●
●
●
−100
●
−20
−40
0 1 2 3 4 5 0.0 0.5 1.0 1.5 2.0 0 5 10 15 20 25 30

Figure 10. Lognormal and normal distribution fit to the γ-ray flux distribution of the Fermi/LAT sources listed in Table 1
using MLE method.
30 Bhatta & Dhital
150
3C 273 3C 279 PKS 1424−418
150
100
80
100
100
60
Counts
Counts
Counts
40
50
50
20
0
0
0 5 10 15 20 25 30 0 10 20 30 40 0 5 10 15 20 25 30
100
50 100
60
50
Residual
Residual
Residual
20
●
● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ● ● ● ● ●
0
0
● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
● ● ● ● ●
● ● ●
● ●
−20
●
−100
−100
−60
0 5 10 15 20 25 30 0 10 20 30 40 0 5 10 15 20 25 30
PKS 1502+106 4C +38.41 Mrk 501

150
70
Normal Fit 150 Normal Fit Normal Fit
60
50
100
100
40
Counts
Counts
Counts
30
50
50
20
10
0
0
0 5 10 15 20 25 30 0 5 10 15 20 0.0 0.5 1.0 1.5 2.0
15
●
50
●
50
●
Residual
Residual
Residual
●
5
● ●
●
● ● ●
● ● ● ● ●
0
● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ●
● ●
● ●
● ●
−5
●
●
−50
●
−100
−15
●
0 5 10 15 20 25 30 0 5 10 15 20 0.0 0.5 1.0 1.5 2.0

1ES 1959+65 PKS 2155−304 BL Lac

140
80

120
60
60
100
80
Counts
Counts
Counts
40
40
60
40
20
20
20
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 2 4 6 0 2 4 6 8 10 12 14

20 40
20 40
0 20
Residual
Residual
Residual
● ● ●
● ● ● ● ●
● ● ● ●
● ● ● ● ● ● ●
● ● ● ● ● ● ●
0
● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ●
● ●
● ●
● ●
● ●
●
−40
−40
−40
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 2 4 6 0 2 4 6 8 10 12 14

CTA 102 3C 454.3

150
Lognormal Fit Lognormal Fit

Normal Fit Normal Fit
100
80
100
Counts
Counts
60
40
50
20
0
0 10 20 30 40 50 0 20 40 60 80
Flux ( × 10−7 ph/s/cm2 ) Flux ( × 10−7 ph/s/cm2 )
●
50
0 20
Residual
Residual
● ●
● ● ● ● ● ●
● ● ●
0
● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ●
● ●
●
−100
−40
0 10 20 30 40 50 0 20 40 60 80
Flux ( × 10−7 ph/s/cm2 ) Flux ( × 10−7 ph/s/cm2 )
Figure 10 (Cont.).
D. RMS-FLUX RELATION IN BLAZAR
3C 66A AO 0235+164 PKS 0454−234 S5 716+714

3.5 3.0
1.5 1.6

3.0
2
2
2.5 1.4
2.5
1.0 2.0 1.2
2.0
1.0
1.5
−7
−7
−7
−7
1.5
0.8
0.5 1.0 1.0
0.6
0.5
0.5 0.4
1.0 1.5 2.0 1 2 3 4 5 6 1 2 3 4 5 1 2 3 4
−7 2 −7 2 −7 2 −7 2
Mean Flux (×10 photon/s/m ) Mean Flux (×10 photon/s/m ) Mean Flux (×10 photon/s/m ) Mean Flux (×10 photon/s/m )
Mrk 421 TON 0599 ON +325 W Comae

2.0
0.8

2
2
4
1.5 0.20
3 0.6
1.0 0.15
−7
−7
−7
−7
2 0.4
0.5 1 0.10
0.2
1.5 2.0 2.5 3.0 1 2 3 4 5 6 0.4 0.6 0.8 1.0 1.2 0.4 0.6 0.8 1.0
−7 2 −7 2 −7 2 −7 2
4C +21.35 3C 273 3C 279 PKS 1424−418


2
2
5 8 4
6
4
6 3
4 3
−7
−7
−7
−7
4 2
2
2
1 2 1
2 4 6 8 10 12 2 4 6 8 5 10 15 5 10 15
−7 2 −7 2 −7 2 −7 2
PKS 1502+106 4C +38.41 Mrk 501 1ES 1959+65

0.6
0.30
4
2
4 0.25 0.5
3 0.20 0.4
3
0.15 0.3
−7
−7
−7
−7
2 2
0.10 0.2
1 1 0.05 0.1
2 4 6 8 10 2 4 6 8 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.4 0.6 0.8 1.0 1.2 1.4
−7 2 −7 2 −7 2 −7 2
PKS 2155−304 BL Lac CTA 102 3C 454.3

14
1.0 12 15
2
2.0
10
0.8
1.5 8 10
0.6
−7
−7
−7
−7
6
1.0
0.4 4 5
0.2 0.5 2
0.6 0.8 1.0 1.2 1.4 1.6 1.8 1 2 3 4 5 5 10 15 20 25 10 20 30

−7 2 −7 2 −7 2 −7 2
Figure 11. RMS-Flux relation in the gamma-ray light curves of the sample blazars. The magenta line represents the linear fit
to the observations.
32 Bhatta & Dhital
E. POWER SPECTRAL DENSITY OF BLAZAR
3C 66A AO 0235+164 PKS 0454−234 S5 0716+714

4 4 4 4
10 10 10 10
3 3 3 3
10 10 10 10
2 2 2 2
10 10 10 10
−1
−1
−1
−1
Power [rms]
Power [rms]
Power [rms]
Power [rms]
1 1 1 1
10 10 10 10
0 0 0 0
10 10 10 10
−1 −1 −1 −1
10 10 10 10
−2 −2 −2 −2
10 10 10 10
−3 −3 −3 −3
10 −4 −3 −2 −1
10 −4 −3 −2 −1
10 −4 −3 −2 −1
10 −4 −3 −2 −1
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
−1 −1 −1 −1
Frequency [day] Frequency [day] Frequency [day] Frequency [day]
Mrk 421 TON 0599 ON +325 W Comae

4 4 4 4
10 10 10 10
3 3 3 3
10 10 10 10
2 2 2 2
10 10 10 10
−1
−1
−1
−1
Power [rms]
Power [rms]
Power [rms]
Power [rms]
1 1 1 1
10 10 10 10
0 0 0 0
10 10 10 10
−1 −1 −1 −1
10 10 10 10
−2 −2 −2 −2
10 10 10 10
−3 −3 −3 −3
10 −4 −3 −2 −1
10 −4 −3 −2 −1
10 −4 −3 −2 −1
10 −4 −3 −2 −1
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
−1 −1 −1 −1
4C +21.35 3C 273 3C 279 PKS 1424−418

4 4 4 4
10 10 10 10
3 3 3 3
10 10 10 10
2 2 2 2
10 10 10 10
−1
−1
−1
−1
Power [rms]
Power [rms]
Power [rms]
Power [rms]
1 1 1 1
10 10 10 10
0 0 0 0
10 10 10 10
−1 −1 −1 −1
10 10 10 10
−2 −2 −2 −2
10 10 10 10
−3 −3 −3 −3
10 −4 −3 −2 −1 10 −4 −3 −2 −1
10 −4 −3 −2 −1
10 −4 −3 −2 −1
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
−1 −1 −1 −1
CTA 102
1ES 1959+65 Mrk 501 10
4 3C 454.3
4 4 4
10 10 10
3
3 3 10 3
10 10 10
2
2 2 10 2
10 10 10
−1
−1
−1
−1
Power [rms]
Power [rms]
Power [rms]
Power [rms]
10
1
10
1 10 10
1
0
10
0
10
0
10 10
0
−1
10
−1
10
−1
10 10
−1
−2 −2 −2 −2
10 10 10 10
−3 −3 −3 −3
10 −4 −3 −2 −1
10 −4 −3 −2 −1 10 −4 −3 −2 −1
10 −4 −3 −2 −1
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
−1 −1 −1 −1
4C +38.41 PKS 1502+106 PKS 2155−304 BL Lac

4 4 4 4
10 10 10 10
3 3 3 3
10 10 10 10
2 2 2 2
10 10 10 10
−1
−1
−1
−1
Power [rms]
Power [rms]
Power [rms]
Power [rms]
1 1 1 1
10 10 10 10
0 0 0 0
10 10 10 10
−1 −1 −1 −1
10 10 10 10
−2 −2 −2 −2
10 10 10 10
−3 −3 −3 −3
10 −4 −3 −2 −1
10 −4 −3 −2 −1
10 −4 −3 −2 −1
10 −4 −3 −2 −1
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
−1
Frequency [day] Frequency [day]
−1
Frequency [day]
−1
Frequency [day]
−1
Figure 12. Power spectral density of the gamma-ray light curves of the blazars. Discrete Fourier periodogram (black), binned
periodogram (red), and the best fit PSD (blue)

Gopal Bhatta and Niraj Dhital: Event Horizon Telescope Collaboration Et Al. 2019

Uploaded by

Copyright:

Available Formats

Gopal Bhatta and Niraj Dhital: Event Horizon Telescope Collaboration Et Al. 2019

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Gopal Bhatta and Niraj Dhital: Event Horizon Telescope Collaboration Et Al. 2019

Uploaded by

Copyright:

Available Formats

Draft version February 11, 2020

Typeset using LATEX default style in AASTeX62

The Nature of γ-ray Variability in Blazars

Corresponding author: Gopal Bhatta

Table 1. The source sample of the Fermi/LAT blazars

(1) (2) (3) (4) (5) (6) (7) (8)

2. SAMPLE SOURCES AND FERMI/LAT DATA PROCESSING

3.1. Fractional variability

for which uncertainty can be expressed as

3.2. Flux distribution: Lognormality and RMS-flux relation

Lognormal fit Normal fit

Source name m s LL AIC BIC µ σ LL AIC BIC

Flux rate (×10 ph/s /cm )

3.3. Symmetry Analysis: Flux rise and decay profiles

Source slope Spearman’s r. c. (ρ) p-value

3C 66A 0.70 0.73 0.006

Table 4. Symmetry analysis of γ-ray fluxes of the source light curves

3C 66A 7.28 ± 6.36 7.50 ± 7.23 0.12 0.33 17.51 ± 7.72

Bhatta (2017, 2019)

S5 716+714 346 ± 23 99.97 99.96 349 ± 27 99.982 99.980

Figure 7. Same as Figure 6.

• Power spectral density:

Zheng Z.-Y., Butler N. R., Shen Y. et al., 2016, ApJ, 827,

B. FITTING OF BLAZAR FLUX DISTRIBUTION USING WEIGHTED LEAST SQUARE METHOD

Lognormal fit Normal fit

Source name m s χ2 /dof µ σ χ2 /dof

(8) (9) (10) (11) (12) (13) (14)

C. PROBABILITY DENSITY FUNCTION FOR BLAZAR FLUX DISTRIBUTION

3C 66A AO 0235+164 PKS 0454−234

Normal Fit Normal Fit Normal Fit

S5 0716+714 Mrk 421 TON 0599

Lognormal Fit Lognormal Fit Lognormal Fit

Normal Fit Normal Fit Normal Fit

ON +325 W Comae 4C +21.35

Normal Fit Normal Fit Normal Fit

0 1 2 3 4 5 0.0 0.5 1.0 1.5 2.0 0 5 10 15 20 25 30

0 1 2 3 4 5 0.0 0.5 1.0 1.5 2.0 0 5 10 15 20 25 30

PKS 1502+106 4C +38.41 Mrk 501

0 5 10 15 20 25 30 0 5 10 15 20 0.0 0.5 1.0 1.5 2.0

1ES 1959+65 PKS 2155−304 BL Lac

Lognormal Fit Lognormal Fit Lognormal Fit

0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 2 4 6 0 2 4 6 8 10 12 14

0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 2 4 6 0 2 4 6 8 10 12 14

CTA 102 3C 454.3

Lognormal Fit Lognormal Fit

D. RMS-FLUX RELATION IN BLAZAR

3C 66A AO 0235+164 PKS 0454−234 S5 716+714

σXS (×10 photon/s/m )

σXS (×10 photon/s/m )

σXS (×10 photon/s/m )

Mrk 421 TON 0599 ON +325 W Comae

σXS (×10 photon/s/m )

σXS (×10 photon/s/m )

σXS (×10 photon/s/m )

4C +21.35 3C 273 3C 279 PKS 1424−418

σXS (×10 photon/s/m )

σXS (×10 photon/s/m )

σXS (×10 photon/s/m )

PKS 1502+106 4C +38.41 Mrk 501 1ES 1959+65