A search for Centaurus A-like features in the spectra of Fermi-LAT detected radio galaxies
MMNRAS , 1–15 (2020) Preprint 14 January 2020 Compiled using MNRAS L A TEX style file v3.0
A search for Centaurus A-like features in the spectra of
Fermi -LAT detected radio galaxies
Cameron B. Rulten, ? Anthony M. Brown, and Paula M. Chadwick, Centre for Advanced Instrumentation, Department of Physics, University of Durham, South Road, Durham, DH1 3LE, United Kingdom
Accepted 2020 January 3. Received 2019 December 16; in original form 2019 September 6
ABSTRACT
Motivated by the detection of a hardening in the γ -ray spectrum of the radiogalaxy Centaurus A, we have analysed ∼
10 years of
Fermi -LAT observations of 26radio galaxies to search for similar spectral features. We find that the majority ofthe radio galaxies’ γ -ray spectral energy distributions are best fitted with a simplepower-law model, and no spectral hardening similar to that found in Centaurus Awas detected. We show that, had there been any such spectral features present in oursample of radio galaxies, they would have been seen, but note that 7 of the radiogalaxies (3C 111, 3C 120, 3C 264, IC 4516, NGC 1218, NGC 2892 and PKS 0625-35)show evidence for flux variability on 6-month timescales, which makes the detectionof any steady spectral features difficult. We find a strong positive correlation (r = 0.9)between the core radio power at 5 GHz and the γ -ray luminosity and, using a simpleextrapolation to TeV energies, we expect around half of the radio galaxies studied willbe detectable with the forthcoming Cherenkov Telescope Array. Key words: gamma-rays: galaxies – galaxies: active – radiation mechanisms: non-thermal
In the recently released
Fermi -LAT 4FGL 8–year pointsource catalog (The Fermi-LAT collaboration 2019), nearly70% of the objects are associated with a known astrophysi-cal object. Of these associated sources, the majority are clas-sified as blazars: either of type BL Lac ( ∼ ∼ ∼
13% of the associatedpoint sources in the 4FGL. Less than 1% of all 4FGL sourcesare associated with radio galaxies.According to the unified model of active galactic nuclei(AGN) (Urry & Padovani 1995), radio galaxies are radio–loud AGN that have jets beamed at large inclination angleswith respect to the observer’s line of sight, and are thereforesometimes termed misaligned blazars. Unlike blazars, there-fore, the non-thermal radiation emitted by radio galaxies isonly modestly beamed (Rieger 2017). Thus radio galaxiesare very interesting targets because any detected γ -ray sig-nal from these AGN is not dominated by the highly beamedjet emission, meaning it might be possible to disentangle jetand core emission or even to detect other potential sourcesof γ -ray emission.The nearby, well-studied and (on Fermi -LAT ? E-mail: [email protected] timescales) non-variable object Centaurus A (Cen A) isan obvious target to search for emission which does notoriginate in the jet. A study of Brown et al. (2017) revealedevidence for a new population of energetic particles nearCen A’s core. This evidence was manifested as a statis-tically significant ( > σ ) hardening in the Fermi-LAT γ -ray spectrum, with the spectral index changing fromΓ = . ± .
02 to Γ = . ± .
07 at a break energy of2 . ± . > eV,making it difficult to explain the energy spectrum if CRsrepresent an accelerated thermal population. To solve thisproblem it might be possible to use dark matter (DM) an-nihilations as a source of non-thermal particles that can befurther accelerated in astrophysical shocks. These shock–accelerated particles should produce a power-law spectrumfor all energies, and the DM annihilation should produce aspectrum with a cut-off at the DM particle mass. The com-bination of these two effects should result in a characteristicspectrum (Lacroix et al. 2014).The lack of variability in Cen A’s emission ruled outthe possibility of jet-induced leptonic processes being re-sponsible for the spectral feature and it was found that the γ -ray spectrum of Cen A was compatible with a very large © 2020 The Authors a r X i v : . [ a s t r o - ph . H E ] J a n C. B. Rulten et al. localized enhancement (i.e. a spike) in the DM halo profile(Brown et al. 2017). However, it was noted that a popu-lation of unresolved millisecond pulsars or another popula-tion of energetic particles could also be responsible for theemission above 2.6 GeV. Recent results from the H.E.S.S.telescopes have resolved the emission above 100 GeV, andsuggest that the highest-energy emission from Cen A comesfrom a small, inner jet close to Cen A’s core (Sanchez et al.(2018); H. E. S. S. Collaboration et al. (2018b)), which couldwell be the source of the population of energetic particlespostulated in Brown et al. (2017).The discovery of a spectral hardening of Cen A’s γ -rayspectrum provides the motivation to look at other radiogalaxies detected with Fermi -LAT in order to search for sim-ilar spectral features. This work describes our analysis of aselection of such radio galaxies. In Section 2 we highlightthe
Fermi -LAT observations used, our radio galaxy selec-tion criteria, and the data analysis methods employed. InSection 3 we focus on the results of our
Fermi -LAT analysisbefore discussing possible interpretations of our findings inSection 4. Fermi -LAT OBSERVATIONS AND DATAANALYSIS
The Large Area Telescope [LAT; Atwood et al. (2009)]aboard the NASA
Fermi γ -ray Space Telescope is a wide–field pair conversion telescope sensitive to γ -rays overthe approximate energy range 30 MeV (cid:54) E (cid:54)
300 GeV. The
Fermi -LAT was launched from the Kennedy Space Centeron June 11, 2008 and started conducting science operationson 11th August 2008; it has thus recently celebrated 11 yearsof near uninterrupted service. The great majority of datataken by
Fermi -LAT during this time has been in all-sky-survey mode. This observing mode scans the entire sky every ∼
180 minutes and has produced the deepest extragalacticscan ever at γ -ray energies. The 26 radio galaxies selected and listed in Table 1 are thoseidentified and categorized as radio galaxies in the
Fermi -LAT 4FGL catalog (The Fermi-LAT collaboration 2019),excluding four well–studied, nearby radio galaxies: M 87, thePerseus cluster galaxies NGC 1275 and IC 310, all of whichhave been found to exhibit significant flux variability at γ -ray energies ((Ait Benkhali et al. 2019; Brown & Adams2011; Aleksić et al. 2014) respectively) and Cen A (Brownet al. 2017). We have included PKS 0625-35 in the list ofselected radio galaxies; however, as discussed in a recent pa-per (H. E. S. S. Collaboration et al. 2018a), there is evidenceto suggest that this galaxy could be a BL Lac object andthus its classification as a “misaligned blazar” may need tobe reconsidered. The majority of the radio galaxies in ourstudy are classified as having a Fanaroff-Riley type I (FR I)morphology (Fanaroff & Riley 1974); there are 6 Fanaroff-Riley type 2 (FR II) galaxies, and one compact radio galaxy(FR 0).Our selection does not, of course, represent a completelist of radio galaxies detected above 20 MeV, as there is al- ways the possibility that some of the 1500 + unassociatedsources in the 4FGL catalog may be radio galaxies. Roughly 10 years worth of
Fermi -LAT data were used foreach target. The exact exposure spans MJD 54682.65527778through to and including MJD 58362.0, which is equiv-alent to 317895388
Fermi -LAT seconds, 3679.34
Fermi -LAT days or 10.08
Fermi -LAT years. We consider pho-tons with energies between 0 . −
300 GeV within a 15 ◦ circular region of interest (ROI) centred on each ra-dio galaxy target. These photons were obtained from Fermi -LAT sky-survey observations in accordance withthe pass8 data analysis criteria. The
Fermi -LAT recom-mended quality cuts were used, including a zenith anglecut of 90 ◦ (to reduce γ -ray contamination originating fromthe Earth’s limb), (DATA_QUAL>0)&&(LAT_CONFIG==1) and abs ( rock_angle ) < Python package
Fermipy (Wood et al. 2017) to facilitate analysis of
Fermi -LATdata with the
Fermi -LAT
Fermitools (v1.0.1). The anal-ysis used the
P8R3_SOURCE_V2 instrument response functionand adopted the binned maximum-likelihood method (Mat-tox et al. 1996). To estimate the background, we includedsources within the region of interest (ROI) listed in the4FGL catalog (The Fermi-LAT collaboration 2019) alongwith the recommended Galactic ( gll_iem_v07.fits ) andisotropic diffuse ( iso_P8R3_SOURCE_V2_v1.txt ) templatesprovided with the
Fermitools .Since our study considered 10 years of
Fermi -LAT ob-servations, we must search for additional point sources of γ -rays not accounted for by the 8-year integrated catalogueof the 4FGL. To do this, we used the find_sources algo-rithm in Fermipy to construct a significance map centredon each radio galaxy candidate . This TS map was used toidentify additional point sources of γ -rays, with TS (cid:62) α J , β J ) of the peak excess, and a final like-lihood fit was performed with the normalisation and spectralindex of the new point sources free to vary. Once all sources of γ -rays were accounted for in our data,we conducted temporal and spectral studies for all 26 radiogalaxies considered in our research. For each, we produceda spectral energy distribution (SED); these are shown inFigures 1 and 2. The SED flux points are generated usinga separate likelihood analysis for each equally–spaced loga-rithmic energy bin. For each target we initially used 8 binsper decade, but then rebinned the flux data into a binningscheme of 2 bins per decade. Each spectral bin requires astatistical significance above background as defined by the The significance map was constructed assuming a point sourcewith an E − spectrum. MNRAS , 1–15 (2020) search for CenA-like features in the spectra of Fermi -LAT detected radio galaxies Fermi -LAT name Assoc. name l (deg.) b (deg.) z Morphology Variability index σ ( √ T S )4FGL J0322.6-3712e Fornax A 240.16 -56.68 0.0059 FR I 36.5 16.994FGL J0057.7+3023 NGC 315 124.56 -32.49 0.0164 FR II 21.0 9.034FGL J0708.9+4839 NGC 2329 168.57 22.79 0.0197 - 4.0 7.234FGL J0334.3+3920 4C +39.12 154.16 -13.43 0.0203 FR 0 20.1 8.814FGL J1144.9+1937 3C 264 † † Table 1.
Details of the radio galaxies analyzed in this study including their
Fermi -LAT variability index and detection significance(obtained in this work). The radio galaxies were selected using the
Fermi -LAT 4FGL catalog and are ordered by increasing redshift (z).A variability index > . <
1% chance of being a steady source. The two TeV-detected radio galaxies are highlighted witha † . test statistic, TS , such that √ TS (cid:62) σ , and a minimumnumber of γ -ray photons above background of γ (cid:62)
2; oth-erwise a 95% confidence–level upper limit is calculated. Foreach SED we also calculate and show the 1 σ uncertaintyband. For each radio galaxy we initially only considered thespectral model given in the 4FGL catalogue as a descriptionof the high-energy γ -ray emission. In the NGC 1218 SED(Figure 2) we see some tension between a power-law modeldescription and the highest energy bin upper limit. As a re-sult we also considered a log-parabola model for NGC 1218,and find that with a test statistic value of 235 between thelog-parabola and power-law models, the fit significantly im-proves and hence we discard the initial power-law model.As can be seen in Figures 1, 2, 3 and 4, the 10-yearSEDs produced do not share the spectral features whichwere seen in Cen A (Brown et al. 2017). Instead, the ma-jority of radio galaxy SEDs are best fitted with a simplepower-law model. The exceptions are 3C 120, NGC 1218,NGC 6251 and PKS 2152-69, which are best fitted with a log-parabola model. We also note that NGC 2484, PKS 1839-48,TXS 1303+114 and TXS 1516+064 have only 2 statisticallysignificant spectral flux bins above background; unsurpris-ingly, these targets are amongst those with the lowest detec-tion significances within this analysis. Both NGC 2484 andTXS 1303+114 fall just below the accepted 5 σ significance Defined as twice the difference between the log-likelihoods oftwo different models, 2 logL − logL , where L and L are definedas the likelihoods of individual model fits (Mattox et al. 1996). threshold in this 10-year dataset, and in both cases, theirSEDs lack sufficient statistics across the full energy band toproduce a reliable power-law fit.In most cases we detect no significant γ -ray ex-cess above 30 GeV. The exceptions to this are 3C 264,4C +39.12, B2 1447+27, Fornax A, NGC 1218, NGC 2329and PKS 0625-35. Two of these are detected at TeV ener-gies: PKS 0625-35 (H. E. S. S. Collaboration et al. 2018a)and 3C 264 (Mukherjee 2018). Apart from Fornax A, theseradio galaxies all exhibit fairly hard spectra over the en-ergy band considered in this analysis. In addition, 3C 264and 4C +39.21 share very similar SED characteristics acrossthe Fermi -LAT energy band. Intriguingly, no significant γ -ray excess above background is detected for either of thesetwo radio galaxies at energies below 1 GeV. Above 1 GeVtheir flux brightness is similar and their spectral indices arehard (0.1 and 0.14 respectively in E dNdE units), result-ing in a significant γ -ray excess up to energies of 100 GeV.Given 3C 264 was recently detected by VERITAS at TeVenergies (Mukherjee 2018), perhaps there is potential for de-tecting 4C +39.12 at TeV energies too. If detected, it wouldbe the lowest luminosity radio galaxy yet seen at TeV en-ergies apart from IC 310, a peculiar galaxy with somewhatuncertain classification (Graham et al. 2019).In addition to this spectral investigation, we alsoinvestigated the temporal characteristics by producinglightcurves for each radio galaxy studied. Relative to blazars,radio galaxies are weak γ -ray emitting sources and at theenergies we are considering, Fermi -LAT does not have thesensitivity performance to detect enough γ -ray photons for MNRAS000
2; oth-erwise a 95% confidence–level upper limit is calculated. Foreach SED we also calculate and show the 1 σ uncertaintyband. For each radio galaxy we initially only considered thespectral model given in the 4FGL catalogue as a descriptionof the high-energy γ -ray emission. In the NGC 1218 SED(Figure 2) we see some tension between a power-law modeldescription and the highest energy bin upper limit. As a re-sult we also considered a log-parabola model for NGC 1218,and find that with a test statistic value of 235 between thelog-parabola and power-law models, the fit significantly im-proves and hence we discard the initial power-law model.As can be seen in Figures 1, 2, 3 and 4, the 10-yearSEDs produced do not share the spectral features whichwere seen in Cen A (Brown et al. 2017). Instead, the ma-jority of radio galaxy SEDs are best fitted with a simplepower-law model. The exceptions are 3C 120, NGC 1218,NGC 6251 and PKS 2152-69, which are best fitted with a log-parabola model. We also note that NGC 2484, PKS 1839-48,TXS 1303+114 and TXS 1516+064 have only 2 statisticallysignificant spectral flux bins above background; unsurpris-ingly, these targets are amongst those with the lowest detec-tion significances within this analysis. Both NGC 2484 andTXS 1303+114 fall just below the accepted 5 σ significance Defined as twice the difference between the log-likelihoods oftwo different models, 2 logL − logL , where L and L are definedas the likelihoods of individual model fits (Mattox et al. 1996). threshold in this 10-year dataset, and in both cases, theirSEDs lack sufficient statistics across the full energy band toproduce a reliable power-law fit.In most cases we detect no significant γ -ray ex-cess above 30 GeV. The exceptions to this are 3C 264,4C +39.12, B2 1447+27, Fornax A, NGC 1218, NGC 2329and PKS 0625-35. Two of these are detected at TeV ener-gies: PKS 0625-35 (H. E. S. S. Collaboration et al. 2018a)and 3C 264 (Mukherjee 2018). Apart from Fornax A, theseradio galaxies all exhibit fairly hard spectra over the en-ergy band considered in this analysis. In addition, 3C 264and 4C +39.21 share very similar SED characteristics acrossthe Fermi -LAT energy band. Intriguingly, no significant γ -ray excess above background is detected for either of thesetwo radio galaxies at energies below 1 GeV. Above 1 GeVtheir flux brightness is similar and their spectral indices arehard (0.1 and 0.14 respectively in E dNdE units), result-ing in a significant γ -ray excess up to energies of 100 GeV.Given 3C 264 was recently detected by VERITAS at TeVenergies (Mukherjee 2018), perhaps there is potential for de-tecting 4C +39.12 at TeV energies too. If detected, it wouldbe the lowest luminosity radio galaxy yet seen at TeV en-ergies apart from IC 310, a peculiar galaxy with somewhatuncertain classification (Graham et al. 2019).In addition to this spectral investigation, we alsoinvestigated the temporal characteristics by producinglightcurves for each radio galaxy studied. Relative to blazars,radio galaxies are weak γ -ray emitting sources and at theenergies we are considering, Fermi -LAT does not have thesensitivity performance to detect enough γ -ray photons for MNRAS000 , 1–15 (2020)
C. B. Rulten et al. E d N / d E [ e r g c m − s − ] σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] B2 1447+27 σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] FornaxA σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] IC1531 σ γ σ γ σ γ σ γ σ γ σ γ σ γ Figure 1.
Spectral energy distributions obtained for the radio galaxies: 3C 311, 3C 120, 3C 264, 3C 303, 4C +39.12,B2 1447+27, Fornax A and IC 1531. Apart from 3C 120, all the radio galaxy SEDs in this subset are best-fitted with asimple power-law model. The binning scheme is 2 bins per decade and a 95% confidence–level upper limit is shown for binswhere √ TS < σ and the number of γ -ray photons above background in each bin is γ <
2. MNRAS , 1–15 (2020) search for CenA-like features in the spectra of
Fermi -LAT detected radio galaxies E d N / d E [ e r g c m − s − ] IC4516 σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] NGC1218 σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] NGC2329 σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] NGC2484 σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] NGC2892 σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] NGC315 σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] NGC6251 σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] PKS0625-35 σ γ σ γ σ γ σ γ σ γ σ γ σ γ Figure 2.
Spectral energy distributions obtained for the radio galaxies IC 4516, NGC 1218, NGC 2329, NGC 2484, NGC 2892,NGC 315, NGC 6251 and PKS 0625-35. Apart from NGC 1218 and NGC 6251, all the radio galaxy SEDs in this subset arebest-fitted with a simple power-law model. The binning scheme is 2 bins per decade and a 95% confidence–level upper limitis shown for bins where √ TS < σ and the number of γ -ray photons above background in each bin is γ <000
Spectral energy distributions obtained for the radio galaxies IC 4516, NGC 1218, NGC 2329, NGC 2484, NGC 2892,NGC 315, NGC 6251 and PKS 0625-35. Apart from NGC 1218 and NGC 6251, all the radio galaxy SEDs in this subset arebest-fitted with a simple power-law model. The binning scheme is 2 bins per decade and a 95% confidence–level upper limitis shown for bins where √ TS < σ and the number of γ -ray photons above background in each bin is γ <000 , 1–15 (2020) C. B. Rulten et al. E d N / d E [ e r g c m − s − ] PKS1304-215 σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] PKS1514+00 σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] PKS1839-48 σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] PKS2153-69 σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] PKS2300-18 σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] PKS2324-02 σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] PKS2338-295 σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] PictorA σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] TXS1303+114 σ γ σ γ σ γ σ γ σ γ σ γ σ γ E d N / d E [ e r g c m − s − ] TXS1516+064 σ γ σ γ σ γ σ γ σ γ σ γ σ γ Figure 3.
Spectral energy distributions obtained for the radio galaxies PKS 1304-215, PKS 1514+00, PKS 1839-48,PKS 2153-69, PKS 2300-18, PKS 2324-2, PKS 2338-295, Pictor A, TXS 1303+114 and TXS 1516+064. Apart from PKS 2153-69, all the radio galaxy SEDs in this subset are best-fitted with a simple power-law model. The binning scheme is 2 bins perdecade and a 95% confidence–level upper limit is shown for bins where √ TS < σ and the number of γ -ray photons abovebackground in each bin is γ <
2. MNRAS , 1–15 (2020) search for CenA-like features in the spectra of
Fermi -LAT detected radio galaxies Fermi -LAT name Assoc. name N (cm − s − MeV − ) Index ( Γ ) E (MeV) E pivot (MeV)4FGL J0418.2+3807 3C 111 (7 . ± . × − − . ± .
05 532.8 300.04FGL J1144.9+1937 3C 264 (2 . ± . × − − . ± .
10 3216.4 3601.24FGL J1443.1+5201 3C 303 (1 . ± . × − − . ± .
14 3426.4 3250.74FGL J0334.3+3920 4C +39.12 (7 . ± . × − − . ± .
13 5679.6 6777.74FGL J1449.5+2746 B2 1447+27 (7 . ± . × − − . ± .
20 11380.8 13137.04FGL J0322.6-3712e Fornax A (2 . ± . × − − . ± .
06 1762.5 1396.54FGL J0009.7-3217 IC 1531 (6 . ± . × − − . ± .
13 1692.2 1440.74FGL J1454.1+1622 IC 4516 (6 . ± . × − − . ± .
07 922.3 1073.64FGL J0708.9+4839 NGC 2329 (6 . ± . × − − . ± .
15 4693.8 7416.64FGL J0758.7+3746 NGC 2484 (9 . ± . × − − . ± .
17 3421.0 3346.84FGL J0931.9+6737 NGC 2892 (2 . ± . × − − . ± .
06 1459.5 1205.14FGL J0057.7+3023 NGC 315 (2 . ± . × − − . ± .
11 1124.0 1158.94FGL J0627.0-3529 PKS 0625-35 (2 . ± . × − − . ± .
04 2337.5 2215.64FGL J1306.7-2148 PKS 1304-215 (9 . ± . × − − . ± .
08 2151.2 1640.54FGL J1516.5+0015 PKS 1514+00 (8 . ± . × − − . ± .
10 744.0 852.74FGL J1843.4-4835 PKS 1839-48 (1 . ± . × − − . ± .
16 3893.3 6115.04FGL J2302.8-1841 PKS 2300-18 (8 . ± . × − − . ± .
09 1788.0 1530.94FGL J2341.8-2917 PKS 2338-295 (7 . ± . × − − . ± .
15 1548.7 1157.34FGL J0519.6-4544 Pictor A (1 . ± . × − − . ± .
10 1462.0 1397.74FGL J2326.9-0201 PKS 2324-02 (1 . ± . × − − . ± .
13 1215.0 2048.44FGL J1306.3+1113 TXS 1303+114 (7 . ± . × − − . ± .
19 4187.4 1604.04FGL J1518.6+0614 TXS 1516+064 (3 . ± . × − − . ± .
19 5831.5 10929.7
Table 2.
Details of the power-law models best fitted to the radio galaxy SEDs..
Fermi -LAT name Assoc. name N (cm − s − MeV − ) Index ( α ) Curvature ( β ) E (MeV) E pivot (MeV)4FGL J0433.0+0522 3C 120 (14 . ± . × − . ± .
07 0 . ± .
05 445.8 366.84FGL J0308.4+0407 NGC 1218 (7 . ± . × − . ± .
10 0 . ± .
03 1000.0 2518.84FGL J1630.6+8234 NGC 6251 (4 . ± . × − . ± .
04 0 . ± .
02 667.8 479.94FGL J2156.0-6942 PKS2153-69 (5 . ± . × − . ± .
23 0 . ± .
18 368.8 295.5
Table 3.
Details of the log-parabola model best fitted to the radio galaxy SEDs.. a well-sampled lightcurve at timescales under 6 months.Therefore we constructed lightcurves using 20 time bins witheach bin comprising 180 days of
Fermi -LAT data coveringthe full 10 year observation period.The lightcurves were produced using a binned likeli-hood approach where the normalisation value for the radiogalaxy of interest, all sources within 1 degree of the radiogalaxy and the diffuse background were kept free to vary.For each time bin a flux is calculated over the full energyrange 100 MeV (cid:54) E (cid:54)
300 GeV. A variability index (see Ta-ble 1) was calculated for each radio galaxy using the methoddescribed in Nolan et al. (2012) which is a simple LikelihoodRatio test between the null hypothesis (a constant source)and the alternative hypothesis (variable source). If the nullhypothesis is correct, then in accordance with this method,the variability index is distributed as χ with 19 degrees offreedom. Thus any variability index above 39.7 indicates ev-idence for variability at the (cid:62) σ level. We find statisticalevidence of flux variability on 6 month timescales for sevenof the 26 radio galaxies analysed: 3C 111, 3C 120, 3C 264,IC 4516, NGC 1218, NGC 2892 and PKS 0625-35.While fixed-time-width binning schemes, like that usedin our temporal analysis are widely used within the field,there is merit in exploring a Bayesian Block binning schemefor the radio galaxies we identify here as variable. Such ananalysis will aid the comparison of spectra obtained for flar-ing and non-flaring states and may enable us to uncover any Cen A-like spectral components potentially camouflaged byvariable components. Such an analysis will be the subject ofa future publication. Our analysis shows that the SEDs of the radio galaxies stud-ied do not exhibit any spectral features to warrant fittingany extra spectral components. Although no firm conclu-sions can be drawn without a full multi-wavelength analysison a source by source basis, it does appear as though the γ -ray emission found in the Fermi -LAT categorised radiogalaxies at energies 10 MeV (cid:54) E (cid:54)
300 GeV is dominated byjet particle acceleration and/or jet interaction. This is par-ticularly the case for the 7 objects which display evidencefor variability; such variability not only suggests a jet ori-gin for the emission, but renders it difficult to detect anysteady spectral features which may exist (Graham et al.2019). For example, in blazars we know that the variableemission we detect is dominated by emission processes intheir ultra-relativistic jets. TeV emission from blazars dueto VLBI knots is seen in jets with variability timescales typ-ically lasting minutes to hours during flaring periods. In thecase of two radio galaxies, M 87 and IC 310, the variabilitytimescales detected are much shorter than the light cross-ing time of the black hole horizon, which implies the γ -ray MNRAS000
300 GeV is dominated byjet particle acceleration and/or jet interaction. This is par-ticularly the case for the 7 objects which display evidencefor variability; such variability not only suggests a jet ori-gin for the emission, but renders it difficult to detect anysteady spectral features which may exist (Graham et al.2019). For example, in blazars we know that the variableemission we detect is dominated by emission processes intheir ultra-relativistic jets. TeV emission from blazars dueto VLBI knots is seen in jets with variability timescales typ-ically lasting minutes to hours during flaring periods. In thecase of two radio galaxies, M 87 and IC 310, the variabilitytimescales detected are much shorter than the light cross-ing time of the black hole horizon, which implies the γ -ray MNRAS000 , 1–15 (2020)
C. B. Rulten et al.
Energy [ MeV ] E d N / d E [ e r g c m − s − ] Cen A4FGL radio galaxiesCen A: power-law 3 TeV b¯b
Cen A: dark matter 3 TeV b¯b
Cen A: power-law 3 TeV τ + τ − Cen A: dark matter 3 TeV τ + τ − Cen A: power-law 400 GeV τ + τ − Cen A: dark matter 400 GeV τ + τ − Figure 4.
Shown here is a comparison of the SED shapes ob-tained for Cen A (blue dashed line) versus the best fitted power-law spectral shapes for the non-variable 4FGL radio galaxies anal-ysed in this work (orange dashed lines). Also shown is the 1 sigmaconfidence bands obtained for Cen A (blue band) and the 4FGLradio galaxies (orange bands). It is clear that if there were anyCen A-like spectral features present in the sample of radio galax-ies analysed, we would have seen them. We have also overlaidthe power-law and DM models (see legend) used to describe thetotal Cen A emission (Brown et al. 2017) for the given DM par-ticle mass and annihilation channel. Assuming the supermassiveblack holes of the 4FGL radio galaxies are of a similar mass scaleto that of Cen A’s, which implies that the total mass of DM inthe 4FGL radio galaxies is similar to that of Cen A, we see noevidence of spectral hardening in the
Fermi -LAT energy band towarrant consideration of DM scenarios. emission must be coming from a compact region (Giannioset al. 2009).It is immediately obvious from Figure 4 that if therewere any spectral features similar to that seen in Cen A,they would have been detected. However, as the supermas-sive black hole masses for the radio galaxies are not knownand the mass of the DM spike expected around the centralSMBH is related to the mass of the SMBH, detailed mod-elling of any DM signal one might expect from these radiogalaxies and how that compares to the Cen A DM modelfitted by Brown et al. (2017) is not possible; we can simplysay that there is no such component at a similar level to thatobserved in Cen A. The observational evidence from
Fermi -LAT and H.E.S.S. suggests that whatever is happening inCen A is unusual, or perhaps spectral features are simplyeasier to detect due to the object’s proximity. The lack ofevidence for variability in the γ -ray emission from Cen Ameans theorists are able to postulate scenarios where the γ -ray emission arises from larger scales i.e. not a compactemitting region. Such scenarios could include contributionsfrom undetected millisecond pulsars or dark matter (Brownet al. 2017) or hadronic processes such as the interaction ofenergetic protons with ambient matter (proton-proton in-teractions) (Sahakyan et al. 2013) or the inverse Comptonupscattering of photons on kiloparsec scales (Hardcastle &Croston 2011) or host galaxy starlight (Stawarz et al. 2003).These scenarios are largely degenerate and the only way to distinguish these models from one another is to accumulatemore and better quality radio and ground-based TeV obser-vations of radio galaxies.Six of the radio galaxies studied, 3C 264, 4C +39.12,B2 1447+27, NGC 1218, NGC 2329 and PKS 0625-35, haveparticularly hard spectra and emission above 30 GeV. Fur-thermore, two of these objects, 3C 264 and 4C +39.21, showno significant excess below 1 GeV. It is possible that thehard spectra displayed by these objects are an indicationthat the peak of the inverse Compton emission is locatedin the Fermi -LAT energy regime. This would be surprising,as radio galaxies do not have the strong Doppler-boostingnormally required to produce such a high-frequency peak intheir SED. Multi-wavelength observations would be requiredto confirm if this is the case.In the case of Cen A, the H.E.S.S. observations werereally important in the identification of a statistically signif-icant spectral hardening. Thus, with the construction of newground-based instruments like CTA about to begin, we needto look for any hints of correlation across multiwavelengthdata to try and pinpoint the best radio galaxy candidatesfor observation with IACTs, which ultimately will help usto better understand any such spectral upturns and if theyare a common occurrence in these galaxies. We thereforeused our analysis results to search for correlations betweena number of different characteristic properties and the γ -rayemission from these radio galaxies. Figures 5 and 6 showthe results of our correlation studies using data from pub-licly accessible radio and optical catalogues.In Figure 5 the left panel shows the radio luminositycalculated using the total γ -ray luminosity calculated using the integratedfluxes estimated in this work, assuming a concordance cos-mology with H =
71 km s − Mpc − , Ω M = .
27, Ω Λ = . CMB = .
725 K. Where no 5 GHz flux densities wereavailable we used the 4.8 GHz flux density measurement. Wefind a strong positive correlation between the radio and γ -ray luminosity, with the correlation coefficient r = .
8. Theright panel shows the absolute magnitude of the visual opti-cal band calculated using extinction corrected V filter dataversus the γ -ray luminosity, calculated using the integratedfluxes estimated in this work. Where no V filter data wereavailable we used B filter data. We find a weak positive cor-relation (r = .
4) between the absolute magnitude of theseradio galaxies versus their γ -ray luminosity. We also high-light and annotate the power-law modelled TeV-detected ra-dio galaxies using blue star markers, and we see no clusteringof these particular sources.The left panel of Figure 6 shows the radio power ofthe core at 5 GHz frequencies versus the γ -ray luminos-ity calculated using the integrated fluxes estimated in thiswork. Where no 5 GHz data were available we used the 4.8GHz radio flux densities to estimate the core radio power.We find a strong positive correlation between the core ra-dio power at 5 GHz frequencies versus the γ -ray luminosity,with the correlation coefficient r = .
9. The right panel showsthe γ -ray luminosity calculated using the integrated fluxesestimated in this work versus the radio core dominance pa-rameter calculated using the method highlighted in Fan &Zhang (2003). The core dominance parameter suffers froma number of different systematic uncertainties (Abdo et al.2010), and as reported elsewhere (Angioni et al. 2019) we MNRAS , 1–15 (2020) search for CenA-like features in the spectra of
Fermi -LAT detected radio galaxies L [erg s ] L G H z r a d i o [ e r g s ] L [erg s ] A b s o l u t e M a g n i t u d e M v Figure 5.
The left panel shows the radio luminosity calculated using the total 5 GHz (where not available we used 4.8 GHz) radio fluxdensity available in publicly accessible radio catalogues versus the γ -ray luminosity calculated using the integrated fluxes estimated inthis work. We find a strong positive correlation between the radio and γ -ray luminosity. The right panel shows the absolute magnitudefor the visual optical band primarily using extinction corrected V filter data (where not available we used B filter data) versus the γ -rayluminosity calculated using the integrated fluxes estimated in this work. We calculated the absolute magnitudes using photometric dataavailable in NED and Simbad (Wenger et al. 2000), except for PKS 0625-35 (Massardi et al. 2008). We find a weak correlation betweenthe optical brightness of these radio galaxies versus their γ -ray luminosity. The TeV-detected radio galaxies are annotated and highlightedusing blue star markers, and we see no clustering of these particular sources. The Cen A γ -ray luminosity was calculated assuming apower-law model and not the broken power-law model from Brown et al. (2017). find no correlation between the γ -ray luminosity and the ra-dio core dominance parameter, with correlation coefficientr = − .
3. Again we highlight and annotate the TeV-detectedradio galaxies using blue star markers, and we see no clus-tering of these particular sources.Finally, in the context of potentially correlated γ -rayand radio flux, we also investigated the possibility of ex-tended γ -ray emission associated with the kiloparsec scalejet of the radio galaxies. For each radio galaxy we pro-duced a skymap as seen in Figures 7 and 9. Each of theseshow the significance ( √ TS) for an approximate 2 degreeregion centred on the radio galaxy target (indicated with agreen x ). Significance values greater than 5 σ are enclosedwithin the solid dark-orange contour line, and values greaterthan 15 σ are enclosed within the solid light-orange contourline. We note that there is apparent evidence for extendedemission coming from the direction of 3C 111. However, oncloser inspection this extension is likely an artefact of nearby( < . ◦ ) point sources just below the detection threshold.As discussed above, radio galaxies are very interestingtargets for a host of reasons. With the forthcoming next-generation ground-based γ -ray observatory the CherenkovTelescope Array (CTA) (CTA Consortium & Ong (2019);Angioni et al. (2017)), it is hoped that a larger sample ofradio galaxies emitting radiation at very-high-energies willbe gathered. Figure 10 shows the Fermi -LAT 4FGL radiogalaxy fluxes extrapolated to 100 TeV for both CTA-South(left panel) and CTA-North (right panel) respectively. The
Fermi -LAT detected fluxes were extrapolated assuming nobreaks or features in the spectra from the GeV to TeV energyregime. CTA’s ten times better sensitivity over the core ener-gies compared to existing ground-based instruments shouldenable the detection of approximately 13 of the radio galax-ies analysed in this work, assuming a 50 hour observationusing the CTA-North and CTA-South arrays respectively. Inpractice, CTA has the potential to detect a larger number,as there are bound to be variable radio galaxies that will beseen with CTA during flaring periods.
The discovery of a distinctive break in the power-law spec-trum of Cen A posed many questions concerning the originand mechanisms behind the spectral hardening. We there-fore analysed 26 other
Fermi -LAT-detected radio galaxiesto see whether any other “similar” objects to Cen A exhibitbreaks and spectral hardening. This work has found no ev-idence for spectral hardening over a 10-year-averaged spec-trum calculated for each radio galaxy. Had there been sucha spectral feature in these objects, it would have been ap-parent in the data we analysed. This suggests that eitherCen A is unique among radio galaxies, or that any break oc-curs outside
Fermi -LAT’s energy range. We also noted thata number of the galaxies analysed show variability on the 6-month timescale, which strongly suggests a jet origin for the
MNRAS000
MNRAS000 , 1–15 (2020) C. B. Rulten et al. L [erg s ] P G H z c o r e [ W H z ] R core:5GHzext:1.4GHz L M e V E G e V [ e r g s ] Figure 6.
The left panel shows the radio core power calculated using the core 5 GHz (except for PKS 2324-02 we used 4.8 GHz) radioflux density versus the γ -ray luminosity calculated using the integrated fluxes estimated in this work. The 5 GHz radio flux densitieswere taken from the NASA/IPAC Extragalactic Database (NED) except for those objects with references listed below † . We find a strongpositive correlation between radio core power at 5GHz and γ -ray luminosity. The right panel shows the radio core dominance parameterversus the γ -ray luminosity calculated using the integrated fluxes estimated in this work. We find no correlation between the radio coredominance parameter of these radio galaxies versus their γ -ray luminosity. The TeV-detected radio galaxies are annotated and highlightedusing blue star markers, and we see no clustering of these particular sources. † Fornax A and IC 1531 (Ekers et al. 1989), NGC 1218 (Saikiaet al. 1986), NGC 2892 (Kharb & Shastri 2004) and NGC 6251 (Evans et al. 2005). γ -ray emission from these objects and would render the de-tection of any spectral break with a non-jet origin difficult,if not impossible.With the advent of new large observatories such asSKA and CTA, a new era of astronomy is upon us thatcan help to better understand the non-thermal astroparti-cle physics at play in AGN-like radio galaxies. Many openquestions still remain, such as where and how γ -rays are pro-duced in these extragalactic objects and why a fraction ofthese largest and most energetically connected objects seenin our universe produce TeV γ -rays despite having muchlower Doppler boosting factors compared to blazars? Al-though we may not get a complete understanding of howthese objects work, new discoveries and findings may helpto further our knowledge of the characteristics that distin-guish between classes of objects under unification schemes,for example multi-wavelength studies across the broad elec-tromagnetic spectrum from low frequency radio observationsthrough to very-high-energy γ -ray observations may providefurther support that these unifying characteristics are reallydown to differences in the masses of their supermassive blackholes and their spins, their accretion rates, and the anglesand distances at which we view these fascinating objects. ACKNOWLEDGEMENTS
The authors would like to acknowledge the excellent dataand analysis tools provided by the NASA
Fermi
REFERENCES
Abdo A. A., et al., 2010, ApJ, 720, 912Ait Benkhali F., Chakraborty N., Rieger F. M., 2019, A&A, 623,A2Aleksić J., et al., 2014, Science, 346, 1080Angioni R., Grandi P., Torresi E., Vignali C., Knödlseder J., 2017,Astroparticle Physics, 92, 42Angioni R., et al., 2019, A&A, 627, A148Atwood W. B., et al., 2009, ApJ, 697, 1071Brown A. M., Adams J., 2011, MNRAS, 413, 2785Brown A. M., Bœhm C., Graham J., Lacroix T., Chadwick P.,Silk J., 2017, Phys. Rev. D, 95, 063018CTA Consortium Ong R. A., 2019, in European PhysicalJournal Web of Conferences. p. 01038 ( arXiv:1904.12196 ),doi:10.1051/epjconf/201920901038MNRAS , 1–15 (2020) search for CenA-like features in the spectra of
Fermi -LAT detected radio galaxies D e c ( J ) s i g n i f i c a n c e D e c ( J ) s i g n i f i c a n c e D e c ( J ) s i g n i f i c a n c e D e c ( J ) s i g n i f i c a n c e D e c ( J ) s i g n i f i c a n c e D e c ( J ) B2 1447+27 0123456 s i g n i f i c a n c e D e c ( J ) FornaxA s i g n i f i c a n c e D e c ( J ) IC1531 012345678 s i g n i f i c a n c e Figure 7.
Skymaps showing the significance √ TS for a subset of the radio galaxies analysed. Each skymap considers allenergies between 100 MeV ≤ E ≤
300 GeV and the intensity scale in the z-axis highlights the significance. The dark-orangesolid contour line indicates the 5 σ significance boundary and the light-orange solid contour the 15 σ significance boundary.The radio galaxy position is indicated with a green × and the two orange dashed-line concentric circles in the upper leftcorner of the top left panel show the approximate Fermi -LAT PSF at 100 MeV (large) and 1 GeV (small) respectively.MNRAS000
300 GeV and the intensity scale in the z-axis highlights the significance. The dark-orangesolid contour line indicates the 5 σ significance boundary and the light-orange solid contour the 15 σ significance boundary.The radio galaxy position is indicated with a green × and the two orange dashed-line concentric circles in the upper leftcorner of the top left panel show the approximate Fermi -LAT PSF at 100 MeV (large) and 1 GeV (small) respectively.MNRAS000 , 1–15 (2020) C. B. Rulten et al. D e c ( J ) IC4516 s i g n i f i c a n c e D e c ( J ) NGC1218 s i g n i f i c a n c e D e c ( J ) NGC2329 012345678 s i g n i f i c a n c e D e c ( J ) NGC2484 012345 s i g n i f i c a n c e D e c ( J ) NGC2892 s i g n i f i c a n c e D e c ( J ) NGC315 s i g n i f i c a n c e D e c ( J ) NGC6251 s i g n i f i c a n c e D e c ( J ) PKS0625-35 s i g n i f i c a n c e Figure 8.
Skymaps showing the significance √ TS for a subset of the radio galaxies analysed. Each skymap considers allenergies between 100 MeV ≤ E ≤
300 GeV and the intensity scale in the z-axis highlights the significance. The dark-orangesolid contour line indicates the 5 σ significance boundary and the light-orange solid contour the 15 σ significance boundary.The radio galaxy position is indicated with a green × and the two orange dashed-line concentric circles in the upper leftcorner of the top left panel show the approximate Fermi -LAT PSF at 100 MeV (large) and 1 GeV (small) respectively.MNRAS , 1–15 (2020) search for CenA-like features in the spectra of
Fermi -LAT detected radio galaxies D e c ( J ) PKS1304-215 024681012 s i g n i f i c a n c e D e c ( J ) PKS1514+00 0123456789 s i g n i f i c a n c e D e c ( J ) PKS1839-48 0123456 s i g n i f i c a n c e D e c ( J ) PKS2153-69 s i g n i f i c a n c e D e c ( J ) PKS2300-18 s i g n i f i c a n c e D e c ( J ) PKS2324-02 01234567 s i g n i f i c a n c e D e c ( J ) PKS2338-295 s i g n i f i c a n c e D e c ( J ) PictorA s i g n i f i c a n c e D e c ( J ) TXS1303+114 0123456 s i g n i f i c a n c e D e c ( J ) TXS1516+064 01234567 s i g n i f i c a n c e Figure 9.
Skymaps showing the significance √ TS for a subset of the radio galaxies analysed. Each skymap considers allenergies between 100 MeV ≤ E ≤
300 GeV and the intensity scale in the z-axis highlights the significance. The dark-orangesolid contour line indicates the 5 σ significance boundary and the light-orange solid contour the 15 σ significance boundary.The radio galaxy position is indicated with a green × and the two orange dashed-line concentric circles in the upper leftcorner of the top left panel show the approximate Fermi -LAT PSF at 100 MeV (large) and 1 GeV (small) respectively.MNRAS000
300 GeV and the intensity scale in the z-axis highlights the significance. The dark-orangesolid contour line indicates the 5 σ significance boundary and the light-orange solid contour the 15 σ significance boundary.The radio galaxy position is indicated with a green × and the two orange dashed-line concentric circles in the upper leftcorner of the top left panel show the approximate Fermi -LAT PSF at 100 MeV (large) and 1 GeV (small) respectively.MNRAS000 , 1–15 (2020) C. B. Rulten et al. energy [TeV] E × d N / d E [ e r g c m s ] CTA-South 50hr sensitivity (version prod3b-v2)Fermi-LAT 10 year sensitivity (P8R2_SOURCE_V6) TS=25 > 10 photons per bin3C1203C264 (large zenith angles: 45 55 )FornaxAIC1531NGC1218PictorAPKS0625-35PKS1304-215PKS1514+00PKS2153-69 (large zenith angles: 35 55 )PKS2300-18PKS2324-02PKS2338-295TXS1303+114 (large zenith angles: 35 55 ) energy [TeV] E × d N / d E [ e r g c m s ] CTA-North 50hr sensitivity (version prod3b-v2)Fermi-LAT 10 year sensitivity (P8R2_SOURCE_V6) TS=25 > 10 photons per bin3C1113C1203C2643C3034C+39.12B2 1447+27IC4516NGC1218NGC2329NGC2892 (large zenith angles: 35 55 )NGC315NGC6251 (large zenith angles: 45 55 )PKS1304-215 (large zenith angles: 45 55 )PKS1514+00PKS2300-18 (large zenith angles: 45 55 )PKS2324-02TXS1303+114
Figure 10.
Shown here are the
Fermi -LAT 4FGL radio galaxy fluxes extrapolated up to 100 TeV for both CTA-South(top panel) and CTA-North (bottom panel). The respective sensitivity performance curves for each of the CTA sites is alsoshown (solid black line) as well as the
Fermi -LAT 10 year sensitivity (solid grey line). The
Fermi -LAT detected fluxes areextrapolated assuming no spectral breaks or features between the GeV and TeV energy range.Ekers R. D., et al., 1989, Monthly Notices of the Royal Astro-nomical Society, 236, 737Evans D. A., Hardcastle M. J., Croston J. H., Worrall D. M.,Birkinshaw M., 2005, Monthly Notices of the Royal Astro-nomical Society, 359, 363Fan J. H., Zhang J. S., 2003, A&A, 407, 899Fanaroff B. L., Riley J. M., 1974, MNRAS, 167, 31PGiannios D., Uzdensky D. A., Begelman M. C., 2009, MNRAS,395, L29Graham J. A., Brown A. M., Chadwick P. M., 2019, MNRAS, 485, 3277H. E. S. S. Collaboration et al., 2018a, MNRAS, 476, 4187H. E. S. S. Collaboration et al., 2018b, Astronomy and Astro-physics, 619, A71Hardcastle M. J., Croston J. H., 2011, MNRAS, 415, 133Kharb P., Shastri P., 2004, Astronomy and Astrophysics, 425, 825Lacroix T., Bœhm C., Silk J., 2014, Phys. Rev. D, 90, 043508Massardi M., et al., 2008, Monthly Notices of the Royal Astro-nomical Society, 384, 775Mattox J. R., et al., 1996, ApJ, 461, 396MNRAS , 1–15 (2020) search for CenA-like features in the spectra of
Fermi -LAT detected radio galaxies Mukherjee R., 2018, The Astronomer’s Telegram, 11436, 1Nolan P. L., et al., 2012, ApJS, 199, 31Rieger F. M., 2017, in 6th International Symposium on High En-ergy Gamma-Ray Astronomy. p. 020008 ( arXiv:1611.02986 ),doi:10.1063/1.4968893Sahakyan N., Yang R., Aharonian F. A., Rieger F. M., 2013, ApJ,770, L6Saikia D. J., Subrahmanya C. R., Patnaik A. R., Unger S. W.,Cornwell T. J., Graham D. A., Prabhu T. P., 1986, MonthlyNotices of the Royal Astronomical Society, 219, 545Sanchez D., Holler M., Taylor A., Rieger F., DeNaurois M., forthe H.E.S.S. collaboration 2018, in TeVPA. BerlinStawarz Ł., Sikora M., Ostrowski M., 2003, ApJ, 597, 186The Fermi-LAT collaboration 2019, arXiv e-prints, p.arXiv:1902.10045Urry C. M., Padovani P., 1995, PASP, 107, 803Wenger M., et al., 2000, Astronomy and Astrophysics SupplementSeries, 143, 9Wood M., Caputo R., Charles E., Di Mauro M., Magill J., PerkinsJ. S., Fermi-LAT Collaboration 2017, International CosmicRay Conference, 301, 824This paper has been typeset from a TEX/L A TEX file prepared bythe author.MNRAS000