FERMI LARGE AREA TELESCOPE DETECTION OF EXTENDED GAMMA-RAY EMISSION FROM THE RADIO GALAXY FORNAX A

The radio galaxy Fornax A, well known for its radio lobes spanning $\sim 50^{\prime}$, with a lobe-to-lobe separation of $\sim 33^{\prime}$ (see Ekers et al. 1983), is one of the closest and brightest radio galaxies, located at a distance of only 18.6 Mpc (Madore et al. 1999). Hosted by the elliptical galaxy NGC 1316, the radio source contains a low-ionization nuclear emission-line region nucleus, which has been imaged to arcsecond-scale resolution and features a flat spectrum ($\alpha =0.4;$ ${S}_{\nu }\propto {\nu }^{-\alpha }$) core with dual-opposing "s"-shaped jets that are detected out to ∼5 kpc from the core (Geldzahler & Fomalont 1984). The radio lobes are characterized by a complex polarized filamentary structure with no observable hotspots (Fomalont et al. 1989).

Fornax A was the first radio galaxy reported to emit diffuse, non-thermal X-ray emission from within its radio lobes from observations with ROSAT (Feigelson et al. 1995) and ASCA (Kaneda et al. 1995), which were later confirmed through dedicated observations of the east lobe with XMM-Newton (Isobe et al. 2006) and the west lobe with Suzaku (Tashiro et al. 2009). The non-thermal X-rays have been widely attributed to inverse Compton (IC) emission of relativistic electrons scattering on cosmic microwave background (CMB) photons, with the same population of relativistic electrons producing both synchrotron and IC emission (e.g., Harris & Grindlay 1979). To date, similar leptonic IC/CMB emission has been detected in X-rays from tens of extended lobes in radio galaxies and quasars. In general, such detections imply that the ratio of relativistic electron pressure to magnetic field pressure within the lobes is ∼1–100 (Croston et al. 2005; Kataoka & Stawarz 2005; Isobe et al. 2011). More recently, Seta et al. (2013) reported a detection of thermal emission from the western lobe of Fornax A using combined Suzaku and XMM-Newton data. Thermal emission in the lobes of a radio galaxy is typically not seen, although evidence for this has also been reported in the giant lobes of the nearby radio galaxy Centaurus A (O'Sullivan et al. 2013; Stawarz et al. 2013).

Motivated by the observed (and presumed IC/CMB) X-ray emission from the lobes of Fornax A, Cheung (2007) predicted that the high-energy tail of the IC/CMB would be detected by the Fermi Large Area Telescope (LAT; Atwood et al. 2009) at $\gt 100\;{\rm{MeV}}$. Following this, Georganopoulos et al. (2008) predicted that the lobes would also be detected in γ-rays at higher energies by the LAT due to IC upscattering of the infrared and optical extragalactic background light (EBL) photons, analogous to the CMB photons upscattered to X-ray energies. The association of Fornax A with the Fermi-LAT second year catalog (hereafter 2FGL; Nolan et al. 2012) source 2FGL J0322.4−3717 thus raised an important question regarding the origin of the γ-ray source, which at the time had no evidence presented for significant extension. In particular, a distinction between emission arising from the lobes and possible contamination from the central core region could not be established from the γ-ray data alone, although X-ray and radio observations (Kim & Fabbiano 2003) suggested that the contribution from the core was likely to be minimal.

In a recent study by McKinley et al. (2015), the spectrum of the Fornax A lobes was modeled in multiple wavelengths using both leptonic and hadronic production scenarios without knowledge of γ-ray spatial extent or γ-ray contamination from the galaxy core. They concluded the most likely source of γ-ray production is hadronic processes within filamentary structures of the lobes. Our study follows the successful γ-ray detection of the extended lobes from Centaurus A (Abdo et al. 2010), and LAT studies of the lobes of NGC 6251 (Takeuchi et al. 2012) and Centaurus B (Katsuta et al. 2013). Gamma-ray upper limits using H.E.S.S. and Fermi-LAT observations have been used to constrain the hadronic cosmic-ray population within the radio lobes of Hydra A (Abramowski et al. 2012).

Fornax A is not associated with a γ-ray source in the most recent, third Fermi-LAT catalog based on four years of LAT data (3FGL, Acero et al. 2015; 3LAC, Ackermann et al. 2015). However, the centroid of the source 3FGL J0322.5−3721 is offset by $0\buildrel{\circ}\over{.} 15$ from the core of Fornax A. This offset is greater than the 95% position uncertainty of the 3FGL source. In the following we discuss possible reasons for this offset, detail a significant γ-ray detection of extended emission from Fornax A using 6.1 years of Fermi-LAT data, and present modeling under leptonic and hadronic scenarios. Detecting extended emission from Fornax A with the LAT is challenging because the 68% containment point-spread function (PSF) radius is $\sim 0\buildrel{\circ}\over{.} 8$ at 1 GeV, which is larger than the Fornax A lobe-to-lobe separation. The LAT PSF is energy dependent going from 5° at 100 MeV to $0\buildrel{\circ}\over{.} 1$ at 100 GeV with 68% confidence.⁵⁵

2.1. Fermi-LAT Observations

Unlike all studies mentioned in Section 1, we used 6.1 years (from 2008 August 4 to 2014 September 4) of Pass 8 LAT data. Compared to previous iterations of the LAT event-level analysis, Pass 8 provides greater acceptance and a PSF (Sgrò 2014), as well as event type partitions according to PSF⁵⁶ , which we used in this analysis. All of these factors allowed for a firm detection of extension of Fornax A. We selected from all-sky survey data at energies from 0.1 to 300 GeV extracted from a region of interest (ROI) with 10°radius centered at the J2000.0 radio position of Fornax A (R.A.  = 50673, decl.  = −37208, Geldzahler & Fomalont 1984). We used the "source" event class, recommended for individual source analysis, a zenith angle limit of 100° to greatly reduce contamination from the Earth limb, and a rocking angle limit of 52°. Fermi Science Tools v10r01p00 and instrument response functions (IRFs) P8R2_SOURCE_V6 were used for this analysis.⁵⁷

To model the LAT data, we included all sources from the 3FGL within 10° of the radio core position of Fornax A. The diffuse background was modeled using preliminary versions of Galactic diffuse and isotropic spectral templates recommended by the Fermi-LAT collaboration, of which the finalized versions have been released to the public.⁵⁸ Several tests were performed, and we determined that the results presented here with the preliminary diffuse models are compatible with those obtained with the finalized models. We used the same spectral models as in the 3FGL catalog for all background sources, and the normalization and spectral shape parameters of all point sources were left free during optimization. For the diffuse models, only the normalization parameters were left free.

We initially modeled Fornax A as a point source located at the position of the radio core, removing 3FGL J0322.5−3721 from the model since it is offset from the radio core by $0\buildrel{\circ}\over{.} 15$. We optimized the localization using the gtfindsrc tool provided in the Science Tools in unbinned mode. The best-fit localization is R.A.  = 5073, decl.  = −3728 with a 95% confidence error circle radius of $0\buildrel{\circ}\over{.} 14$, slightly southeast of the position of the Fornax A core and consistent with the reported 3FGL localization. Figure 1 shows the best-fit localization (point B) and the core (point A) as well as the 2FGL and 3FGL error contours plotted on top of the relative residual counts map. Optimizing the model with the single point source at point B, we detect γ-ray emission at a Test Statistic (TS)⁵⁹  = 121. The spectrum was modeled as a single power law with a resulting maximum-likelihood photon index ${\rm{\Gamma }}=2.08\pm 0.08$ and a full band energy flux of $(5.34\pm {0.78}_{\mathrm{stat}}\begin{array}{l}+0.03\\ -0.05\end{array}{}_{\mathrm{sys}})\times {10}^{-12}$ erg cm⁻² s⁻¹ (see Table 2). Systematic errors are due to the systematic uncertainty in the LAT effective area.⁶⁰ These fluxes and indices are consistent with those reported in the catalogs for sources 2FGL J0322.4−3717 and 3FGL J0322.5−3721.

Zoom In Zoom Out Reset image size

2.2. Extension and Morphology

In the following, we describe several tests performed to determine the morphology of the observed γ-ray emission from the direction of Fornax A, as summarized in Tables 1 and 2. All tests on extension and morphology made full use of the additional spatial information brought about by the new Pass 8 PSF event type partitions. The broadband flux and spectrum optimizations in addition to these tests were performed using all PSF types in composite likelihood.

Table 1.  Fornax A Spatial Model Comparisons

Null Hypothesis	Alternative Hypothesis	dofa	${\rm{\Delta }}\mathrm{log}{ \mathcal L }$	σb	Sections
Point source (core location A)	Point source (best-fit location B)	2	0.6	0.6	2.1
Disk (best-fit location B, $0\buildrel{\circ}\over{.} 03$ radius)	Disk (best-fit location B, $0\buildrel{\circ}\over{.} 33$ radius)	1	17.3	5.9	2.2.1
Point source (best-fit location B)	Two point sources (locations D and E)	4	16.2	4.8	2.2.2
Point source (core location A)	Radio lobes template and point source (core location A)	2	19.8	6.0	2.2.3
Radio lobes template	Radio lobes template and point source (core location A)	2	0.7	0.7	2.2.3
Radio lobes template, rotated 90° CW	Radio lobes template	1	12.9	5.1	2.2.3
Radio lobes template, rotated 90° CCW	Radio lobes template	1	18.9	6.1	2.2.3
Radio lobes template	Radio lobes template and point source (location C)	4	7.0	2.7	2.2.3

Notes. Calculated using the likelihood ratio computation as described in Section 2.

^aThe difference in the number of degrees of freedom between the two hypotheses. ^bThe alternative hypotheses are preferred over the corresponding null hypotheses by the significances σ.

Download table as:  ASCIITypeset image

Table 2.  Fornax A LAT Spectral Fit Results

Model	TS	Energy Flux (×10−12 erg cm−2 s−1)	Photon Index
Point (best-fit location B)	121	$5.34\pm {0.78}_{-0.05\;\mathrm{sys}}^{+0.03}$	$2.08\pm 0.08\pm {0.03}_{\mathrm{sys}}$
Disk (best-fit location B, $0\buildrel{\circ}\over{.} 33$ radius)	158	7 ± 1	1.99 ± 0.07
Radio lobes template	158	$7.57\pm {1.05}_{-0.08\;\mathrm{sys}}^{+0.06}$	$1.99\pm {0.07}_{-0.04\;\mathrm{sys}}^{+0.03}$
Two point sources, summed		6.6 ± 0.8
West (location D)	37	2.9 ± 0.7	2.02 ± 0.13
East (location E)	51	3.7 ± 0.8	2.05 ± 0.11
Radio lobes template and point source (location C), summed		8 ± 1
Radio lobes template	66	6 ± 1	1.97 ± 0.08
Point source (location C)	14	1.6 ± 0.6	2.01 ± 0.20

Note. Each fit was performed leaving normalization and spectral shape parameters of all sources free, except for the models for the diffuse background γ-rays which were fit with only normalization.

Download table as:  ASCIITypeset image

2.2.1. Spatial Extension

To determine if the γ-ray emission is extended beyond that of a point source we modeled Fornax A as a flat circular disk of various sizes (003–075 in steps of $0\buildrel{\circ}\over{.} 03$) by producing several disk templates centered at the best-fit location of the LAT source described in Section 2.1 (point B in Figure 1). The uniform disk is the simplest spatial model, and the use of a Gaussian profile has typically been shown to produce comparatively little difference in the overall likelihood and best-fit spectral parameters (Lande et al. 2012). Using gtlike in binned mode (with bin size $0\buildrel{\circ}\over{.} 05$), we determined the overall likelihood ${ \mathcal L }$ as a function of the disk radius r. As shown in Figure 2(a), ${ \mathcal L }$ is peaked at $r=0\buildrel{\circ}\over{.} 33\pm 0\buildrel{\circ}\over{.} 05$, which is roughly compatible with the extent of the lobes as observed in radio (1.5 GHz at $14^{\prime\prime}$ resolution, Fomalont et al. 1989). By comparing the likelihood of the peak radius with the near-zero radius of $0\buildrel{\circ}\over{.} 03$ (effectively a point source), we found that the γ-ray emission is spatial extended with $5.9\sigma$ confidence (${\rm{\Delta }}\mathrm{log}{ \mathcal L }$ = 17.3, 1 degree of freedom, Wilks 1938). See Tables 1 and 2 for more information.

Zoom In Zoom Out Reset image size

2.2.2. Blind Tests for Morphology

With the aim of making no prior assumptions about the morphology of the γ-ray emission in this region, we used the ROI fit with our position-optimized point source as described in Section 2.1 and removed that point source from the model, thereby leaving only the background sources. Using this background model, we created a map of residual counts above 1 GeV in the ROI by subtracting the model's predicted counts from the observed counts in each bin. Apart from the emission near Fornax A, the map of residuals is flat in significance and the distribution of significance values for the bins is Gaussian, indicating there are no significant systematic deviations from the ROI model. The map of residual counts shown in Figure 1 is cropped to a $1\buildrel{\circ}\over{.} 5\times 1\buildrel{\circ}\over{.} 5$ region centered around the Fornax A core. The shape of the residuals suggests a non-circular morphology that resembles the known extended radio morphology. The contours of radio emission from a VLA 1.5 GHz image at $14^{\prime\prime}$ resolution (Fomalont et al. 1989), with the radio core subtracted, are overlaid in Figure 1.

We also tested a model which included two separate point sources whose initial locations were chosen by eye based upon the residual counts. The locations of these two point sources were then optimized using gtfindsrc. The best-fit location of the western point source (labeled D in Figure 1) matches well the western lobe's centroid while the eastern point (labeled E in Figure 1) is offset from the centroid of the eastern lobe. The double point-source model is preferred over the single point-source model at a confidence level of $4.8\sigma$. Table 1 details these results and Table 2 lists the fit parameters. Note that the spectral index is statistically compatible with the single point-source model and for both point sources in the double point-source model.

2.2.3. Radio-motivated Tests for Morphology

Assuming that the same electron energy distribution (EED) determined from the radio emission scatters optical EBL photons to produce γ-rays, then a reasonable guess for the γ-ray morphology should be the observed radio structure. The lobes of Fornax A were found to emit non-thermal IC scattered X-rays, with excellent spatial coincidence to the radio structure produced by synchrotron-emitting relativistic electrons (Feigelson et al. 1995). Diffuse hard X-ray emission associated with the east lobe has been confirmed with XMM-Newton (Isobe et al. 2006), implying the magnetic field is reasonably uniform, further supporting the expectation that the γ-rays should match the radio morphology. Under these circumstances, as was observed in Centaurus A (Abdo et al. 2010), the relativistic electrons will trace the γ-ray emission (Georganopoulos et al. 2008). These assumptions could be incorrect; however, for the purpose of constraining the EBL with the γ-ray flux, using the radio structure is the best choice for the spatial distribution. Physically, Fornax A images from the Wilkinson Microwave Anisotropy Probe (WMAP) might be a better choice of template for the γ-ray emission because the synchrotron radiation within ∼23–94 GHz should be originating from the same band of relativistic electrons that IC scatter into ∼0.1–10 GeV γ-rays. We revisit this topic later in the section.

We created a spatial template of the lobe emission using VLA 1.5 GHz radio data (Fomalont et al. 1989), which offers a more-than-adequate spatial resolution of $14^{\prime\prime}$. Both the central core of Fornax A and various radio point sources inside the lobe structure were manually removed from the spatial template, the former by reduction to zero and the latter by interpolation of the adjacent lobe-dominated flux. The central core was reduced to zero because the lobes do not overlap with the central core (Ekers et al. 1983). We found this template in addition to a point-source model of the core is preferred over just the point source at the core with a confidence of $6.0\sigma$ (${\rm{\Delta }}\mathrm{log}{ \mathcal L }$ = 19.8, 2 degrees of freedom). However, this combined model is preferred over the lobes template alone by only $0.7\sigma$. Moreover, in the combined point-source-and-lobes template fit only 14% of the energy flux is assigned to the core point source. See Tables 1 and 2 for more information. We consider this as evidence that the γ-ray emission from the core of Fornax A is insignificant. Minimal γ-ray flux from the central core is expected, as it is assumed to be synchrotron self-Compton, and the radio flux from the core has been reported to be relatively weak compared to the luminous galaxy lobes (Geldzahler & Fomalont 1984).

In order to test the uniqueness of the radio template morphology and its rotational symmetry, we rotated the template around the central core in increments of $9\buildrel{\circ}\over{.} 2$ and computed the likelihood at each one. The results from this test are shown in Figure 2(b) and indicate that the template in its original orientation is preferred. In particular, we see the original orientation is a better model than the one rotated 90^◦ clockwise with $5.1\sigma$ confidence (${\rm{\Delta }}\mathrm{log}{ \mathcal L }$ = 12.9, 1 degree of freedom) and 90^◦ counterclockwise with $6.1\sigma$ confidence (${\rm{\Delta }}\mathrm{log}{ \mathcal L }$ = 18.9, 1 degree of freedom). The maximum likelihood rotation was −20° ± 10 Additionally, our plot of $\mathrm{log}{ \mathcal L }$ over template rotation angle shows a sinusoidal profile with a similar peak in likelihood around 180^◦ as around 0^◦. This degeneracy indicates that the data are not constraining enough to statistically differentiate the lobes. Modeling the lobes as separate point sources (see Section 2.2.2) results in the measurement of similar fluxes and spectral shapes for the two regions, but this is at odds with radio observations of the lobes. According to the 1.5 GHz VLA observation (Fomalont et al. 1989), the west lobe has about twice the total flux than the east. In fact, the two point-source γ-ray analysis indicated the east lobe may be slightly brighter (but is within the statistical errors). This disagreement may be a symptom of our use of 1.5 GHz VLA data instead of the very similar but more physically motivated ∼23–94 GHz WMAP data as a spatial template for the γ-rays. Indeed, the flux ratio of west to east is ∼1.3 in the WMAP 41 and 61 GHz maps (Georganopoulos et al. 2008), closer to the γ-ray result. However, Fornax A is at the resolving power of the LAT in this analysis and the PSF is broad enough ($\sim 0\buildrel{\circ}\over{.} 8$ radius at 1 GeV) that fine scale changes would be smoothed out and indistinguishable from the original, and small changes in the flux ratio between the lobes should leave the average flux nearly the same. This is highlighted by the similarity in overall likelihood we observe between 0^◦ and 180^◦ rotations of the 1.5 GHz VLA template, where the rotation of 180^◦ is identical to a flux ratio of ∼0.5. In addition, even when using the spatial morphology of a single point source, the flux is nearly consistent with the radio template flux (see Section 2.1).

With the aim of testing the region for the possibility of a contaminating background γ-ray source, we added a point source to the lobes template model and optimized its position using gtfindsrc. The optimized position of this point source (labeled C in Figure 1) was R.A. = $50\buildrel{\circ}\over{.} 67$, decl. = $-37\buildrel{\circ}\over{.} 46$ with a 95% confidence error circle radius of $0\buildrel{\circ}\over{.} 42$ (large enough to encompass the whole Fornax A emission region). The resulting fit was marginally preferred over the lobes template alone with $2.7\sigma$ significance. Therefore, we do not consider any contribution from a background point source to be significant.

2.3. Spectral and Temporal Analysis

In the following we assume the radio morphology template (without any core contribution) is the best description of the Fornax A γ-ray emission. The likelihood ratio technique cannot quantify whether the radio template is statistically preferred with respect to the best-fit disk model because these models are not nested. However, the radio template is the physically motivated model based on the leptonic scenario of γ-ray production. Further, the rotation study presented above indicates a preferred axis of the γ-ray emission which mimics the elongated emission observed at radio frequencies, in turn supporting the similarity between the γ-ray and radio emission morphology.

We tested several broadband spectral models (log-parabola, broken power law, and broken power law with an exponential cutoff), and found none were significantly preferred over the single power law. Then, we measured spectral points by fitting each of 6 equal logarithmically spaced energy bins from 0.1 to 300 GeV to a power law and optimized the flux normalizations by maximizing the likelihood function. In each bin, the normalization parameters for all sources were free, and all other parameters were fixed to the values obtained from the broadband fit. These spectral data points are shown in Figure 3 and Table 3. The source is detected in 4 of the 6 spectral bins with TS $\gt \;5$, and 95% confidence upper limits were calculated for the two lower-significance bins (at the highest energies).

Zoom In Zoom Out Reset image size

Table 3.  Fornax A Total Lobe LAT Spectral Flux

Bin Energy Range (GeV)	Energy Flux (×10−12 erg cm−2 s−1)	TS
0.10–0.38	0.7 ± 0.4	6.1
0.38–1.4	1.0 ± 0.2	49
1.4–5.5	1.1 ± 0.2	87
5.5–21	0.8 ± 0.2	22
21–79a	$\lt 1.7$	3.5
79–300a	$\lt 2.6$	4.3

Notes. Assuming radio lobes template spatial model.

^a95% confidence upper limits.

Download table as:  ASCIITypeset image

To test the γ-ray variability over the 6.1 year period, we made a 0.1–300 GeV light curve in time bins of 185 days, which was found to be the smallest possible timescale while maintaining a reasonable significance of detection in the majority of bins. For each time interval, the emission associated with Fornax A was fit to a single point source (positioned at the best-fit location B in Figure 1), as we expected any potential variable emission to be associated with a point source at the core and not the lobes. All sources included in the 6.1 year analysis were fit with all spectral shape parameters fixed to their optimized values from the full fit, while all normalizations were left free. Upper limits were calculated for time bins within which the TS fell below 4 ($\lt 2\sigma$). The significance of variability was determined following the method described in Nolan et al. (2012). Our analysis yielded a $1.3\sigma$ confidence that the emission is variable, and so we conclude that we do not observe significant variability.

Our Fermi-LAT study of the region around Fornax A consistently shows that, under all tests performed, the γ-ray emission region is significantly extended and the most likely spatial distribution is delineated by the radio lobes. Using a size-optimized flat disk model, extension beyond a point source was found to be significant at $5.9\sigma$ confidence, with a preferred radius of $0\buildrel{\circ}\over{.} 33\pm 0\buildrel{\circ}\over{.} 05$. Modeling the emission as two point sources results in a western point source well matched to the radio lobe centroid and an eastern point source offset from the eastern lobe. This model is preferred over a single point source at the $4.8\sigma$ level. Furthermore, using the 1.5 GHz VLA radio morphology (Fomalont et al. 1989) as a template in combination with a central core point source results in a significantly greater likelihood than the point source alone with $6.0\sigma$ confidence. Contamination from the core is determined to be at most 14% based on a likelihood fit with the radio lobes template and a point source at the core location. While it is difficult to determine the exact morphology of the γ-ray emission, our study shows that it cannot be fully described as a point-like source.

A few scenarios could explain the offset γ-ray point-source localization seen in the 3FGL and in the single point-source analysis presented here. First, since we now know the emission is extended (or at least not point-like), to use a point-source model to localize the emission is to start with a false assumption. The distribution of the γ-ray emitting regions may not be uniform across the lobe structure and thus would not result in a symmetric distribution of γ-ray emission. Second, based upon the offset eastern lobe point-source localization (point E in Figure 1), the existence of a background γ-ray source is not ruled out. However, adding a point source to the lobes template model and localizing with gtfindsrc yields only a slightly better fit at the $2.7\sigma$ level (see point C in Figure 1). We also find no evidence for variability in this source over ∼6 years of observations. Variability might support the presence of a common background source such as a blazar. We note that another potential source that has been investigated in this region is the Fornax cluster (Ando & Nagai 2012; Ackermann et al. 2014) whose center lies 36 northeast of the Fornax A core, and may be contributing contaminating γ-ray flux from various cluster constituents. However, no galaxy cluster has been detected in γ-rays so far.

Fermi-LAT data have been previously used by McKinley et al. (2015) to study Fornax A. They reported a photon flux above 100 MeV of $6.7\times {10}^{-9}$ ph cm⁻² s⁻¹ using a point source spatial model, and our finding using the lobes template was close at $(5.7\pm 0.9)\times {10}^{-9}$ ph cm⁻² s⁻¹. Our study establishes for the first time spatial extension of Fornax A in γ-rays, and distinguishes between γ-ray contributions from the core and lobes. This result was enabled thanks to the improvements brought about by the new Pass 8 event reconstruction, rather than the marginal increase in exposure time (i.e., 6 years of data in our study instead of 5 years in theirs).

3.1. Leptonic Modeling

3.2. Hadronic Modeling

We report the first Fermi-LAT detection of extended γ-ray emission from the radio galaxy Fornax A using 6.1 years of Fermi-LAT data. We conclude that a point-source spatial model is insufficient to describe the γ-ray emission, and our analysis indicates it is likely the emission originates in the lobes. We investigated the origin of the extended emission by assuming leptonic emission that arises due to IC scattering of EBL photons off of relativistic electrons in the radio lobes. This leptonic modeling underestimates the observed γ-ray emission for any current EBL estimate, consistent with the recent findings by McKinley et al. (2015), even after accounting for the additional contribution of IC emission off of the host galaxy light. A hadronic-only model (proton–proton interactions) requires implausibly large total cosmic-ray energy when compared to an estimate of the Fornax A outburst assumed to have created the lobes (Lanz et al. 2010), and this problem can be alleviated by invoking denser sub-structures in the lobes (McKinley et al. 2015). When we assume the lowest EBL model (Finke et al. 2010) and fit the residual γ-ray flux as hadronic production, our fit yields a total cosmic-ray energy of $\sim 5\times {10}^{60}$ erg, matching well with the Hydra A energy (Abramowski et al. 2012), but still at least 100 times greater than the estimated total energy in the Fornax A lobes (Lanz et al. 2010). Thus, even the combined leptonic and hadronic scenario may not be able to explain the γ-ray lobe emission. Given our current understanding of the content of the radio lobes and the EBL, some contribution from leptonic processes must exist. If it is true that there is a γ-ray component other than leptonic in Fornax A, we should expect to observe such a component in other nearby radio galaxies as well.

Our spatial analysis hinted at the existence of a background γ-ray source (C in Figure 1). We re-evaluated the SED of Fornax A including this source in the model, but we found this background source addition is only marginally preferred and cannot fully make up the difference between the data and the IC/EBL model. In any case, γ-ray contamination from an unresolved background source or the Fornax cluster could be present.

Our modeling was done using previously published multiwavelength data. Further analysis in other wavelengths in the future will yield a more definitive picture of the SED. More detailed observations, such as with the hard X-ray telescope NuSTAR, would help determine if the X-ray emission is contaminated by thermal processes (Seta et al. 2013), which would test our initial assumptions that the radio data traces the X-rays and γ-rays, and that the X-rays could be used to constrain the lobe magnetic field. Fermi-LAT analysis at even lower energies (below 100 MeV) may provide more information about the IC/CMB component (Cheung 2007). Potentially with greater statistics, the Fermi-LAT could extend the lobes' detection to higher energies and additionally observe the effect of the host-galaxy photons on the spatial distribution of γ-rays in the Fornax A lobes.

The Fermi-LAT Collaboration acknowledges generous ongoing support from a number of agencies and institutes that have supported both the development and the operation of the LAT as well as scientific data analysis. These include the National Aeronautics and Space Administration and the Department of Energy in the United States, the Commissariat à l'Energie Atomique and the Centre National de la Recherche Scientifique/Institut National de Physique Nucléaire et de Physique des Particules in France, the Agenzia Spaziale Italiana and the Istituto Nazionale di Fisica Nucleare in Italy, the Ministry of Education, Culture, Sports, Science and Technology (MEXT), High Energy Accelerator Research Organization (KEK) and Japan Aerospace Exploration Agency (JAXA) in Japan, and the K. A. Wallenberg Foundation, the Swedish Research Council and the Swedish National Space Board in Sweden.

Additional support for science analysis during the operations phase is gratefully acknowledged from the Istituto Nazionale di Astrofisica in Italy and the Centre National d'Études Spatiales in France.

Facility: Fermi. -