2FHL: THE SECOND CATALOG OF HARD FERMI-LAT SOURCES

We explicitly modeled sources as spatially extended when a previous, dedicated, analysis found the source to be resolved by the LAT. The 25 extended sources reported in 3FGL were included in our model using the spatial templates derived in the individual source studies (see references in Acero et al. 2015). Refitting the positions and extensions of the 3FGL extended sources in this energy range is beyond the scope of this work.

Of the 25 3FGL extended sources, 19 are significantly detected here above the detection threshold (TS ≥ 25). Only 6 sources are not detected and, since all have TS < 10, are removed from the sky model (see Section 3.3.2 for details).

One extended LAT source has had a dedicated analysis published since the release of the 3FGL catalog. Abramowski et al. (2015a) reported joint H.E.S.S. and LAT observations of the very high energy (VHE) source HESS J1834-087. This source is coincident with supernova remnant (SNR) W41 and was detected as spatially extended in a wide energy range spanning 1.8 GeV to 30 TeV. In this paper, we employ the spatial model for the GeV emission determined in Abramowski et al. (2015a), leading to a significant detection of this source.

In addition to modeling the extended sources mentioned in Section 2.3.1, we performed a blind search of the Galactic plane ($| b| \lt 10^\circ$) to identify potential extended sources not included in previously published works. Our analysis pipeline is similar to that used in Hewitt et al. (2013), with some modifications tailored to searching for multiple extended sources in an ROI. The pipeline employs the pointlike binned ML package (Kerr 2010), in particular utilizing the extended source fitting tools validated by Lande et al. (2012) to simultaneously fit the position, extension, and spectra of sources in our ROI.

We created 72 ROIs of radius 10°, centered on b = 0° with neighboring ROIs overlapping and separated by 5° in Galactic longitude. Our initial model of the γ-ray emission in each ROI consisted solely of the Galactic diffuse (allowing just the normalization to be fit) and isotropic emission models (fixing the normalization), with no other sources in the ROI. Emission in the ROIs was further characterized by adding sources and fitting their spectral parameters (normalization and spectral index) in a 14° × 14° region.

A TS map, that included all significant sources found previously, made up of 01 × 01 bins across the ROI, was created at each iteration and a small radius (01) uniform disk, with a power-law spectrum was placed at the position of the peak TS pixel. The spectra of any newly added sources, as well as the position, extension, and spectral parameters of the disk were then fit. If TS_ext ≥ 16, where TS_ext = 2 log $({{ \mathcal L }}_{\mathrm{ext}}/{{ \mathcal L }}_{\mathrm{ps}})$ (i.e.,  twice the logarithm likelihood ratio of an extended to a point source, Lande et al. 2012), then the disk was kept in the model. For TS_ext < 16, the extended source was replaced by a point source with a power-law spectral model. For the point-source replacement case, spectral parameters of sources in the ROI were fit and the position of the new point source was optimized. Finally, the spatial parameters of any previously added extended sources were refit iteratively before creating a new TS map and repeating the process. We stopped adding sources when the peak TS was less than 16 for two successive sources.

To assess the impact of fitting extended sources when starting with an ROI devoid of sources, a crosscheck analysis (also using pointlike) was performed across the Galactic plane. We included 3FGL point and extended sources, the Galactic diffuse and isotropic emission, and pulsars from the second Fermi-LAT pulsar catalog (Abdo et al. 2013; as well as from 3FGL) in the preliminary source model for each region. Sources were iteratively added to account for residual emission and both these residual sources and 3FGL sources were tested for extension. Remarkably, this alternative analysis converges (i.e.,  spectral and spatial parameters for the detected extended sources are compatible in both analyses) to the initially source-devoid analysis for nearly all detected extended sources.

Extended sources detected in the analysis described in this section for which the position and extension were compatible with those found by the crosscheck were included in the ROI model at step 1 of the full ML analysis detailed in Section 2.2. Seed point sources interior to the extended sources were removed prior to the ML fit. To address the ambiguity between detecting a source as spatially extended as opposed to a combination of point sources, we utilized the algorithm detailed in Lande et al. (2012) to simultaneously fit the spectra and positions of two nearby point sources. We only consider a source to be extended if TS_ext > TS_2pts (improvement when adding a second point source). Our blind search of the Galactic plane allowed us to find 5 sources not previously detected as extended by Fermi-LAT. Further details on these sources are presented in Section 3.3.2.

The H.E.S.S array, with a field of view of about 5° and an angular resolution of approximately 012, has invested 2800 hr of exposure to survey part⁸¹ of the Galactic plane, reaching an average sensitivity of 2% of the Crab Nebula flux (i.e., 4.5 × 10⁻¹³ ph cm⁻² s⁻¹) at ≥1 TeV (Aharonian et al. 2006b; Carrigan et al. 2013). Considering that the Crab Nebula spectrum is harder in the 2FHL band than in the >1 TeV band, we estimate that the average sensitivity of 2FHL in the same region of the H.E.S.S. survey is ∼3%–4% of the 50 GeV–2 TeV Crab Nebula flux. The slightly better sensitivity allows H.E.S.S. to detect 69 sources (as reported in the TeVCat), while the LAT finds 36 objects in the same area. However, the comparable sensitivities of the two surveys allow the study of the properties of the high-energy Galactic population. In the 2FHL catalog there is almost an equal number of SNRs and PWNe in contrast to what is found in the H.E.S.S. survey where the ratio of PWNe to SNRs is 1.5 to 1. This might be because the hardest PWNe and softest SNRs are difficult to detect, respectively, in the >50 GeV and >1 TeV bands.

Of the 36 2FHL sources that fall within the footprint of the H.E.S.S. survey, 23 have already been detected at TeV energies and are associated with known counterparts, while 7 are undetected. The remaining 6 objects (2FHL J1022.0−5750, 2FHL J1505.1−5808, 2FHL J1507.4−6213, 2FHL J1703.4−4145, 2FHL J1745.1−3035 and 2FHL J1856.8+0256) are spatially coincident with TeV sources whose origin is not known. All of them have hard spectral indices (Γ < 2.2), but it is interesting to note that 4 of them (2FHL J1022.0−5750, 2FHL J1505.1−5808, 2FHL J1703.4−4145, and 2FHL J1745.1−3035) have Γ < 1.7 (see also Figure 6).

Zoom In Zoom Out Reset image size

We find that 2FHL J1022.0−5750 is spatially compatible with HESS J1023−575, an extended TeV source (Abramowski et al. 2011c), whose emission might be due to a PWN powered by PSR J1023−5746 (Acero et al. 2013). 2FHL J1505.1−5808 is spatially coincident with the unidentified object HESS J1503−582, which has a size of 026 and a flux above 1 TeV (Renaud et al. 2008) compatible with the extrapolation of the 2FHL J1505.1−5808 spectrum. Its spectrum, reminiscent of that of a PWN (e.g.,  HESS J1825−137, Grondin et al. 2011) is reported in Figure 6.

2FHL J1507.4−6213 is spatially coincident with HESS J1507−622, an extended source with a radius of 015 located 35 from the plane (Acero et al. 2011). The analysis of multiwavelength data showed that it is not possible to discriminate between a hadronic and leptonic origin of the emission, but that the latter scenario, if the emission is powered by a PWN, would require a pulsar generated in the explosion of a hyper-velocity star in order to reach the required distance from the plane (Domainko & Ohm 2012).

The sources 2FHL J1703.4−4145 and 2FHL J1745.1−3035 are the hardest sources (Γ < 1.3) among the six objects. 2FHL J1703.4−4145 is spatially coincident with the bright radio emission observed from the western side of the shell of SNR G344.7−001, a nearby mid-aged shell-type (age ∼ 3000 years and 8' diameter) SNR (Giacani et al. 2011). Both the 2FHL source and the SNR are spatially coincident with the larger, elongated and unidentified HESS J1702−420 (Aharonian et al. 2008b). It thus seems likely that SNR G344.7−001 is the counterpart of 2FHL J1703.4−4145 and perhaps also of HESS J1702−420. The combined Fermi-H.E.S.S. spectrum of this source is reported in Figure 6.

2FHL J1745.1−3035 is found to be spatially coincident with the extended source HESS J1745−303, which may be comprised of up to three different sources (Aharonian et al. 2008a). Indeed, the position of 2FHL J1745.1−3035 is compatible with the "C" emission region (the second brightest region in the complex, Aharonian et al. 2008a). However, the nature of this source is more complex, because the 2FHL source is marginally brighter at 1 TeV than the entire H.E.S.S. region and has also a harder spectrum (spectral index of 1.25 ± 0.38 in 2FHL versus 2.17 ± 0.11 as measured by H.E.S.S.).

Finally, 2FHL J1856.8+0256 is coincident with HESS J1857+026, an almost radially symmetric extended source (Aharonian et al. 2008c), whose emission likely originates from a PWN powered by PSR J1856+0245 (Rousseau et al. 2012).

In total, 31 sources are modeled as spatially extended and input into the ML analysis: 25 listed in 3FGL, 5 sources detected in the pointlike analysis (described in Section 2.3.1) that were not detected as extended at the time of 3FGL, and one, SNR W41, reported recently by both the H.E.S.S. and LAT teams (Abramowski et al. 2015a). Names and properties of the extended sources are provided in Tables 4 and 5. Six extended sources, detected in 3FGL, were not detected in 2FHL: the SMC, S 147 (the point source 2FHL J0534.1+2753 was detected inside it), the lobes of Centaurus A (although we detect its core as a point source, 2FHL J1325.6−4301), W 44, HB 21 and the Cygnus Loop.

We detect a weak source, 2FHL J1714.1−4012 (TS = 27), just outside the southwestern edge of the 3FGL spatial template used to model the emission from SNR RX J1713.7−3946 (2FHL J1713.5−3945e). 2FHL J1714.1−4012 has a hard spectral index Γ = 1.63 ± 0.38, that is within errors of the spectral index derived for the SNR, Γ = 2.03 ± 0.20. It is unclear whether 2FHL J1714.1−4012 is a distinct source separated from the SNR, or the result of un-modeled residual emission due to an imperfection in the spatial template adopted for the extended source.

2FHL J1836.5−0655e is associated with the PWN HESS J1837−069. The 3FGL catalog contains several point sources in the vicinity of the PWN. We detect three sources in the vicinity, 2FHL J1834.5−0701, 2FHL J1837.4−0717, and 2FHL J1839.5−0705, the first two of which are coincident with 3FGL sources (3FGL J1834.6−0659, 3FGL J1837.6−0717, respectively). The power-law spectral indices of the three 2FHL point sources and 2FHL J1836.5−0655e are all consistent with each other. The concentration of sources around HESS J1837−069 combined with the spectral compatibility of the sources is suggestive of a common origin to the γ-ray emission in this region. However, the surrounding γ rays could arise from other sources in the region (Gotthelf & Halpern 2008); further analysis is necessary to determine the nature of the sources in this region.

A brief description of the five new 2FHL extended sources is given below with residual TS maps for the region surrounding each source shown in Figure 7. Detailed analyses of these new extended sources will be reported in separate papers.

Zoom In Zoom Out Reset image size

2FHL J1443.2−6221e overlaps with the young, radio-detected SNR RCW 86 (G315.42.3). RCW 86 is a 42' diameter SNR that lies at a distance of 2.3–2.8 kpc and is likely associated with the first recorded supernova, SN 185 AD (Rosado et al. 1996; Sollerman et al. 2003). With more than 40 months of data and using the P7SOURCE data set, the LAT did not significantly detect the SNR, but upper limits on detection at GeV energies combined with detection of significant extension in the TeV (Aharonian et al. 2009) were sufficient to strongly favor a leptonic origin for the emission (Lemoine-Goumard et al. 2012).

An updated LAT analysis of RCW 86 using 76 months of data, as well as the Pass 8 event-level analysis, resulted in detection of the SNR by the LAT as well as significant extension measurement (Hewitt & Fermi-LAT Collaboration 2015). In this paper, we report the results derived for 2FHL J1443.2−6221e from the pointlike analysis described in Section 2.3.1.

2FHL J1419.2−6048e is a newly detected extended sources with size σ_disk = 036 ± 003, that overlaps two nearby PWN/PSR complexes in the Kookaburra region. In the southwest of Kookaburra, HESS J1418−609 (Aharonian et al. 2006a) is coincident with both the extended non-thermal X-ray "Rabbit" PWN (G313.3+0.1, Roberts et al. 1999), and the γ-ray detected pulsar PSR J1418−6058 (Abdo et al. 2009). The northeast region, called "K3," contains HESS J1420−607, coincident with PWN G313.5+0.3 and PSR J1420−6048. Acero et al. (2013) detected, with Fermi-LAT, emission from both HESS J1418−609 (with a soft spectral index, pulsar-like spectrum) and HESS J1420−607 (with a hard power-law index) above 10 GeV, but only HESS J1420−607 was significantly detected above 30 GeV. Neither showed significant extension. Our result for the fitted power-law spectral index of 2FHL J1419.2−6048e is in agreement with the previous GeV and TeV results, yet our measured radius is considerably larger than the TeV extension. To compare the extensions of the uniform disk model used for 2FHL J1419.2−6048e in this paper to the Gaussian model of Aharonian et al. (2006a), we defined the radius which contains 68% of the source's intensity as r₆₈, with r_68,Gaussian = 1.51σ, and r_{68, disk} = 0.82σ (Lande et al. 2012). We find that r₆₈ ≃ 030 for 2FHL J1419.2−6048e, and r₆₈ ≃ 009 for HESS J1420−607.

2FHL J1355.2−6430e, coincident with the VHE source HESS J1356−645, is detected as extended (σ_disk = 057 ± 002) for the first time by the LAT in this work. The source HESS J1356−645 (Abramowski et al. 2011a) is associated with the pulsar PSR J1357−6429, which was determined to be powering a surrounding extended radio and X-ray PWN (Lemoine-Goumard et al. 2011). Acero et al. (2013) detected faint emission from the nebula, and derived a 99% c.l. Bayesian upper limit on extension (σ_Gauss < 039) in the absence of significant extension. The fitted spectral index for 2FHL J1355.2−6430e is compatible with the GeV and TeV results (Acero et al. 2013; Abramowski et al. 2011a), however, the fitted disk extension is larger than that of the TeV detection, with r₆₈ ≃ 047 for 2FHL J1355.2−6430e and r₆₈ ≃ 030 for HESS J1356−645.

2FHL J1112.4−6059e is an extended source (σ_disk = 053 ± 003) newly detected by the LAT that encircles two 3FGL sources, 3FGL J1111.9−6058 and 3FGL J1111.9−6038, and has another, 3FGL J1112.0−6135, just outside its boundary (Acero et al. 2015). The extended source also partially overlaps the massive star-forming region NGC 3603.

Finally, 2FHL J0431.2+5553e is a large extended source (σ_disk = 127 ± 004), with a hard spectrum, that has not been previously detected at γ-ray energies. It overlaps the recently discovered radio SNR G150.3+4.5 (Gao & Han 2014). G150.3+4.5 is a 25 × 3° (Galactic coordinates) elliptical shell type SNR that has a steep radio synchrotron spectrum (α = −0.6), indicative of radio SNRs.

Most of the sources detected in 2FHL are extragalactic. 83% (299) are either located at $| b| \geqslant 10^\circ$ or associated with an extragalactic source. We refer to this set of sources as the extragalactic sample.

BL Lacs represent the most numerous source class (54% of the full 2FHL catalog and 65% of the extragalactic sample), while there are a few detected FSRQs (10 sources, 3% of the full 2FHL catalog). Such a low number of FSRQs is expected due to their soft spectra at >50 GeV. Most of the detected BL Lacs belong to the high-frequency synchrotron peak (HSP) class, rather than to the low-frequency synchrotron peak (LSP), or intermediate-frequency synchrotron peak (ISP) class. This is shown in Figure 8, which reports the distribution of synchrotron peak frequencies of blazars detected in the 2FHL and in the 3FGL. It is clear that the two catalogs sample different parts of the blazar population, with the 3FGL including mostly LSPs and ISPs and the 2FHL including mostly HSPs.

Zoom In Zoom Out Reset image size

There are 198 2FHL sources in the extragalactic sample that are detected both in the 3FGL and 1FHL catalogs. There are 33 other sources that are neither in the 3FGL nor in the 1FHL catalog (and not in 1FGL/2FGL either). In general, 2FHL sources not detected in 3FGL and 1FHL are harder (median Γ ∼ 2.8 versus 3.2), fainter (median F₅₀ ∼ 1.3 × 10⁻¹¹ ph cm⁻² s⁻¹ versus 2.3 × 10⁻¹¹ ph cm⁻² s⁻¹) and are detected at lower significance (median TS ∼ 30 versus 53) than those detected in the aforementioned catalogs.

One of the 33 exclusive 2FHL sources is 2FHL J1944.1+2117 (Γ = 2.73 ± 0.66) that is associated with HESS J1943+213 (Abramowski et al. 2011b). This source is potentially an extreme high-frequency peaked BL Lac, classified as bcu II, located in the Galactic plane with ${\mathrm{log}}_{10}({\nu }_{\mathrm{peak}}^{S}/{\rm{Hz}})=18.79$ (Ackermann et al. 2015). We also detect N 157B, 2FHL J0537.4−6908, with a very hard spectral index of Γ = 1.15 ± 0.37. This is an extragalactic PWN in the LMC detected by H.E.S.S. up to 18 TeV (Abramowski et al. 2012, 2015c). The other 31 sources (out of the 33 2FHL exclusive sources) are not detected by Cherenkov telescopes yet. Of the 31 sources, 16 are classified as some type of blazar whereas 14 are unclassified. One source is associated with a galaxy cluster, 2FHL J0318.0−4414 detected with a soft spectrum (Γ = 4.0 ± 1.7). The emission likely originates from PKS 0316−444, which is a bright radio source at the center of Abell 3112 (z = 0.075, Takizawa et al. 2003). Emission from the region of Abell 3112 was already marginally found by Ackermann et al. (2014a), yet it is robustly detected in the 2FHL.

We find that the number of FSRQs is strongly reduced from 1FHL (71 sources) to the 2FHL (10 sources). This is mainly due to the soft indices of this blazar population and the higher energy threshold adopted in the 2FHL with respect to the 1FHL. Another indication of this effect is that only 2 sources at z > 1 (out of a total of 7) are FSRQs, while the rest are BL Lacs. Other noteworthy examples of detected extragalactic sources are the nearby radio galaxies IC 310 (z = 0.019), NGC 1275 (z = 0.0175), PKS 0625-35 (z = 0.05494), 3C 264 (z = 0.021718), M 87 (z = 0.004283), and Centaurus A (z = 0.0018).

Figure 9 shows the SEDs of two notable sources, Mrk 421 and 3C 66A, highlighting how the LAT now resolves the descending part of the high-energy peak. Indeed, 2FHL spectra of extragalactic sources are generally softer than the corresponding 3FGL and 1FHL spectra. This is evident in Figure 10, which compares the spectral indices for a subsample of 158 BL Lacs in common between the 2FHL, 3FGL and 1FHL catalogs. There are two interesting facts illustrated by Figure 10. First, the median of the distribution shifts toward larger (softer) indices since 2FHL samples the drop of the SED. Second, the scatter of the distribution becomes larger with increasing energies. Some of the scatter arises from statistical effects since the number of detected photons at >50 GeV is generally lower than in the 3FGL and 1FHL.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Spectroscopic redshift measurements exists in the literature (e.g., Shaw et al. 2012, 2013; Masetti et al. 2013) for 43% (128) of the 299 extragalactic sources. The comparison of the 2FHL and 1FHL redshift distributions (reported in Figure 11) shows that on average 2FHL sources lie at lower redshifts than the 1FHL ones. This might be because FSRQs and LSP blazars, which tend to be located at higher redshifts, have soft spectra and are faint in the 2FHL band. High-redshift sources appear even fainter because of the EBL attenuation. The EBL contains all the photons, from the ultraviolet to the infrared, that have been emitted by star formation processes and supermassive black hole accretion throughout the history of the universe (e.g.,  Dwek & Krennrich 2013). The spectrum of extragalactic γ-ray sources, in the whole energy range of our analysis, is modified by pair-production interaction of the γ-ray photons with the EBL (e.g.,  Ackermann et al. 2012). This interaction produces an attenuation of the observed fluxes that is energy and redshift dependent (the larger the γ-ray energy and/or the redshift, the larger the attenuation). We note that despite the high energies probed by this work, seven sources have z > 1, where the EBL attenuation is not negligible. These sources are associated with PKS 0454−234 (z = 1.003), PKS 0426−380 (z = 1.105, Tanaka et al. 2013), OJ 014 (z = 1.148), TXS 0628−240 (z = 1.238), PKS B1424−418 (z = 1.55), B2 2114+33 (z = 1.6), and MG4 J000800+4712 (z = 2.1). We specially note that three sources have redshifts z > 1.5. The effect of the EBL attenuation is clearly seen in Figure 12. This figure shows the significant dependence of the spectral index on the redshift at the 2FHL energies whereas that dependence is reduced at lower energies (i.e.,  for 3FGL and 1FHL) where the EBL interaction is less relevant.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The EBL attenuation can be parameterized by an optical depth τ that can be derived from empirical EBL models (e.g.,  Franceschini et al. 2008; Finke et al. 2010; Kneiske & Dole 2010; Domínguez et al. 2011b; Gilmore et al. 2012; Helgason & Kashlinsky 2012; Stecker et al. 2012; Scully et al. 2014; Khaire & Srianand 2014). In fact, the EBL sets a distance limit from where γ-ray photons of a given energy are expected to reach us, the cosmic γ-ray horizon (CGRH, e.g.,  Abdo et al. 2010; Domínguez et al. 2013). Formally, the CGRH may be defined as the energy at which τ = 1 as a function of redshift. Figure 13 shows the HEPs in 2FHL versus the redshift of emission. As we can see from Figure 13, a few photons are from near and beyond the CGRH. These photons from significantly attenuated sources provide information about the EBL, which in turn carries fundamental information about galaxy evolution and cosmology (e.g.,  Hauser & Dwek 2001; Domínguez & Prada 2013; Biteau & Williams 2015). In fact, the most complete EBL information is typically given by the sources detected up to the highest energies, which are usually the brightest sources with the hardest spectra. The HEPs with the largest optical depths are associated with RBS 0413 (z = 0.19, τ = 1.38), PKS 0823−223 (z = 0.911, τ = 1.39), RX J0648.7+1516 (z = 0.179, τ = 1.74), 1ES 0347−121 (z = 0.188, τ = 2.43), and 1ES 0502+675 (z = 0.34, τ = 2.84), where τ is given by the observationally based EBL model by Domínguez et al. (2011b). These photons are especially interesting for testing alternative photon-propagation scenarios such as secondary cascades (Essey & Kusenko 2010) and axion-like particles (e.g., de Angelis et al. 2007; Sánchez-Conde et al. 2009; Domínguez et al. 2011a; Horns & Meyer 2012). However, we caution the reader that Figure 13 cannot be readily interpreted in terms of constraints on the EBL optical depth because brighter and harder sources stand a larger chance to have photons detected from beyond the horizon.

Zoom In Zoom Out Reset image size