The Fourth Catalog of Active Galactic Nuclei Detected by the Fermi Large Area Telescope

3.1.1. The Bayesian Association Method

3.1.2. The Likelihood Ratio Association Method

The threshold adopted for the association probability is 0.80 in either method. This value represents a compromise between association efficiency and purity. The fraction of sources associated by both methods is 73% (2082/2863), with $684$ and $97$ sources being solely associated with the Bayesian and LR methods, respectively. The overall false-positive rate is 1.6%, where ${N}_{\mathrm{false}}$ is calculated as described in Section 3.1.1. The estimated number of false positives among the $1353$ sources not previously reported in 3LAC is $35.6$.

As in previous LAT AGN catalogs, we define a Clean Sample as those 4LAC sources that did not have any cautionary analysis flags, as described in Section 3.7.3 of the 4FGL paper (The Fermi-LAT collaboration 2020). The most frequent flags are flag 5 (source close to a brighter neighbor), flag 3 (large flux variation when changing diffuse emission model), and flag 2 (large position shift when changing diffuse emission model). Table 2 compares the performance of the two methods in terms of total number of associations ${N}_{\mathrm{assoc}}$, estimated number of false associations ${N}_{\mathrm{false}}$, and number of sources associated solely via a given method, ${N}_{S}$, for the full and Clean samples.

Table 2.  Comparison of Association Methods in Terms of Total Number of Associations, ${N}_{\mathrm{assoc}}$, Estimated Number of False Associations, ${N}_{\mathrm{false}}$, and Number of Sources Associated Only via a Given Method, ${N}_{{\rm{S}}}$

Sample	All Methods		Bayesian Method			LR Method
	${N}_{\mathrm{assoc}}$	${N}_{\mathrm{false}}$	${N}_{\mathrm{assoc}}$	${N}_{\mathrm{false}}$	${N}_{{\rm{S}}}$	${N}_{\mathrm{assoc}}$	${N}_{\mathrm{false}}$	${N}_{{\rm{S}}}$
Full Sample	2863	44.4	2766	34.8	684	2179	143.1	97
Clean Sample	2614	36.5	2529	27.9	596	2018	130.9	85

Download table as:  ASCIITypeset image

The classification of a source as an AGN primarily relies on its optical spectrum. Other characteristics, like the radio loudness, the presence of a flat/steep radio spectrum between 1.4 and 5 GHz, the broadband emission, the core compactness or the level of radio extended emission, the detection of variability, and the degree of polarization observed in different bands are used as ancillary information. If available, earlier classifications reported in the literature have been checked.

3.3.1. Optical Classification

3.3.2. Classification Based on the Broadband Spectral Energy Distribution

Table 4 summarizes the 4LAC statistics. The 4LAC includes $2863$ sources, with $655$ FSRQs, $1067$ BL Lacs, $1077$ BCUs, and $64$ other AGNs. A total of $1353$ sources were not reported in the previous 3LAC catalog, although some of these have been reported elsewhere, e.g., Arsioli & Chang (2017) and Arsioli & Polenta (2018). The new sources include $204$ FSRQs, $290$ BL Lacs, $822$ BCUs, and $36$ nonblazar AGNs. The Clean Sample contains $2614$ sources, with $591$ FSRQs, $1027$ BL Lacs, $941$ BCUs, and $55$ other AGNs. The figures shown in the following only include Clean Sample sources, unless specified otherwise

The SED-based scheme was able to classify 92% of the FSRQs and 85% of the BL Lacs in 4FGL. These values are lower than in 3LAC (99% and 97% for FSRQs and BL Lacs, respectively) due to the more conservative classification procedure used here. The classified fraction decreases to 60% for BCUs, for which fewer broadband archival data are available. Figure 3 shows the ${\nu }_{{\rm{s}},\mathrm{peak}}$ distributions for the different classes. FSRQs are overwhelmingly of the LSP class; therefore, no distinction based on SED-based classes is made for them in figures and tallies. In addition to leaving more sources unclassified, the new procedure has produced a decrease in the the share of ISPs and HSPs among FSRQs, from 10% to 2%. Of the five HSP FSRQs, all located at ${\rm{z}}\lt 0.63$, two are new: B3 0038 + 377 and PKS 1555−140.

Zoom In Zoom Out Reset image size

BL Lacs are fairly evenly distributed among the LSP, ISP, and HSP subclasses. The 4LAC ${\nu }_{{\rm{s}},\mathrm{peak}}$ distribution differs substantially from the 3LAC one, where HSPs were the most abundant subclass. The new classification procedure caused 15% of the AGN previously classified as HSP to be reclassified as ISP.

As a testimony to the substantial follow-up observational efforts already mentioned, 144 sources reported as BCUs in the 3LAC paper (either at high or low Galactic latitudes) are now classified as BL Lacs and 17 as FSRQs. Three BCUs have been reclassified as radio galaxies (IC 1531, TXS 0149+710, PKS 1304−215). Eight sources have changed from a FSRQ to a BL Lac (RGB J0250+172, NVSS J040324−242946, GB6 J0941+2721, 2MASS J11303636+1018245, PKS 1144−379, 4C +15.54, TXS 1951−115, PKS 2233−173) and three more from a BL Lac to a FSRQ (PMN J0709−0255, B2 2234+28A, TXS 2241+406).

The 3LAC sources that are not present in 4LAC are listed in Table 5, along with the various reasons for this situation. Fifty-five 3LAC sources have not been detected in 4FGL. Nineteen 3LAC sources were duplicate associations for which the smaller 4FGL error boxes relative to 3FGL have enabled the association ambiguity to be lifted. Fifteen sources reported in 3LAC have lost their associations (becoming unassociated), while three others are now associated with non-AGN counterparts, all due to improved localizations.

Table 6 lists the 70 nonblazar AGNs included in the 4FGL, $64$ of which belong to the 4LAC and $6$ are part of the low-latitude sample. Nonblazar sources represent about 2% of the total number of AGNs in the 4FGL, a fraction that is basically identical to that found in the 3LAC. Nonblazar sources are further separated into six different classes: 41 radio galaxies,⁷⁸ 9 NLSy1s, 5 CSSs, 2 SSRQs, 1 Seyfert galaxy, and 11 other AGNs.

A total of $36$ new nonblazar AGNs are reported in the 4LAC, 22 of which are radio galaxies. The median 1.4 GHz radio luminosity of the newly detected radio galaxies is about ${10}^{24.4}$ W Hz⁻¹, with the distribution ranging over more than 4 decades (from below 10²² W Hz⁻¹, for NGC 2892, to above 10²⁶ W Hz⁻¹, for PKS 2324−02). The detection of the FR I radio galaxy 3C 120 in gamma-rays was first reported by Abdo et al. (2010c) using 15 months of LAT data, but it is reported in a LAT catalog for the first time. Its absence from previous catalogs can be attributed to periods of flaring interspersed with long periods of low activity (Tanaka et al. 2015). TXS 1303+114 and TXS 1516+064 are members of the FRICAT (Capetti et al. 2017a), although they were earlier proposed as candidate low-power BL Lacs in Capetti & Raiteri (2015) based on their mid-infrared and optical emission. Among the sources already present in previous LAT catalogs, Fornax A stands out because it is, for the first time, detected as an extended source (see Ackermann et al. 2016). Three 3FGL sources have changed class from BCU to radio galaxy (IC 1531, TXS 0149+710, and PKS 1304−215). We associate 4FGL J1346.3−6026 with Cen B, although its location is not coincident with that of the radio-galaxy core but points to the southern radio jet. Similarly, we associate 4FGL J1516.5+0015 with the radio galaxy 4C +00.56 (PKS 1514+00, z = 0.0525) while the gamma-ray position is closer to the lobes than to the core of the radio galaxy.

The only new SSRQ is 3C 212, for which X-ray emission associated with both lobes was detected by Chandra (Aldcroft et al. 2003). Three new CSSs are reported in the 4LAC: 3C 138, 3C 216, and 3C 309.1. These new CSSs are hosted in quasars. At the pc scale, they show a core-jet structure with apparent superluminal motion, indicating Doppler effects and small viewing angles (Paragi et al. 2000; Shen et al. 2001; Lister et al. 2019). While the CSSs all have very high radio luminosity, at the opposite end of the radio luminosity distribution are two other new LAT sources, NGC 3894 and NGC 6328. Based on the small extent of their radio emission, two-sided parsec scale morphology (Taylor et al. 1998; Tingay & de Kool 2003), and low radio luminosity, these sources are excellent candidates for being young radio galaxies (see also Migliori et al. 2016; Principe et al. 2019).

The classification of a source as an NLSy1 relies on three criteria, as reported in Osterbrock & Pogge (1985), Goodrich (1989), and Pogge (2000): (i) a full width half maximum (FWHM) of the ${{\rm{H}}}_{\beta }$ line <2000 km s⁻¹; (ii) a [O iii]λ5007/H_β ratio <3; and (iii) unusually strong Fe ii lines. Nine NLSy1 are reported in the 4LAC. Four of them are new with respect to the 3LAC: 1ERS B1303+515, B3 1441+476, MG2 J164443+2618, and TXS 2116−077. B3 1441+476, MG2 J164443+2618, and TXS 2116−077 were previously reported by D'Ammando et al. (2015a) and Paliya et al. (2018), respectively. The Circinus galaxy (Hayashida et al. 2013) remains the only radio-quiet Seyfert galaxy detected by the LAT.

Among other AGNs, there are two remarkable cases. One is PKS 0521−36, previously classified as BCU in the 3LAC, which shows a knotty VLBA radio structure similar to misaligned AGN. Based on the broad emission lines in the optical and ultraviolet bands and the steep radio spectrum, a possible classification as an intermediate object between broad-line radio galaxies and SSRQ has been suggested by D'Ammando et al. (2015b). The new source PKS 2331−240 was the subject of a multiwavelength study revealing features of a giant radio galaxy restarted as a blazar (Hernández-García et al. 2017).

Finally, 10 nonblazar AGNs reported in the 3LAC are not confirmed in the 4LAC. Five of them had a double association in the 3LAC and are now firmly associated with the other counterpart; two have changed associations (formerly with 3C 221 and 3C 275.1); one has been reclassified as a BCU (GB 1310+487); while two 3FGL nonblazar AGNs are missing in 4FGL (TXS 0348+013, PKS 1617−251).

In addition to high-latitude ($| b| \gt 10^\circ$) 4LAC sources, we present a low-latitude sample. This sample is less complete than the 4LAC because the LAT detection flux limit is higher in this region (by factors of a few) and the counterpart catalogs suffer from Galactic extinction. Because a large contamination of Galactic sources is present in the radio and X-ray surveys used in the LR association method, the classes of the resulting associations are highly uncertain. Consequently, these associations were not considered here. The census of the low-latitude sample is given in the last column of Table 4. The fraction of BCUs (62%) is overwhelming, as expected from the observational hindrance mentioned above.

Although many 4FGL/4LAC sources show significant spectral curvature in the gamma-ray band (The Fermi-LAT collaboration 2020), the gamma-ray power-law photon index, Γ, represents a convenient way to compare the spectral hardness of different sources across various classes and flux values. This index is plotted versus the 8 yr average energy flux above 100 MeV, S₂₅, in Figure 4. The flux detection limit ranges from (1–4) $\times {10}^{-12}$ erg cm⁻² s⁻¹ (i.e., close to 1 eV cm⁻² s⁻¹), with a slight dependence on the spectral hardness. This dependence is stronger than in 3LAC due to the introduction of weights in the 4FGL likelihood to better account for systematic uncertainties. BCUs are rare among the brightest sources (i.e., with ${S}_{25}\gt {10}^{-11}$ erg cm⁻² s⁻¹), while they are dominant close to the flux limit.

Zoom In Zoom Out Reset image size

Figure 5 displays the photon index distributions for the different optical blazar classes, for 4LAC sources already present in 3LAC and the new ones. Newly detected gamma-ray FSRQs, i.e., not reported in previous LAT AGN catalogs, have somewhat softer spectra (difference in median Γ ≃ 0.08) than the previously reported ones, possibly indicating the emergence of sources with SED peaking at lower energy. The difference between the FSRQ and BL Lac distributions is striking. The Γ medians and rms are 2.44 ± 0.20 and 2.02 ± 0.21 for FSRQs and BL Lacs respectively. The relative separation in gamma-ray spectral hardness between FSRQs and BL Lacs already reported in previous LAT catalogs is confirmed: 89% of FSRQs and 86% of BL Lacs have Γ greater and lower than 2.25, respectively. This feature by itself carries significant discrimination power between the two classes. The photon index varies among the BL Lac subclasses, with medians and rms in Γ of 2.17 ± 0.16, 2.05 ± 0.19, and 1.88 ± 0.14 for LSPs, ISPs, and HSPs, respectively. The BCU index distribution straddles that of the two classes and extends to Γ = 3. The expectation that both the BCU photon-index and ${\nu }_{{\rm{s}},\mathrm{peak}}$ distributions correspond to linear combinations of the observed FSRQ and BL Lac distributions is tested in Figure 6. Composite distributions were built assuming the same fractions of FSRQs and BL Lacs as in the observed sample for the photon index distributions. For the ${\nu }_{{\rm{s}},\mathrm{peak}}$ distribution, a slight correction (a factor of 0.92) in the normalization was introduced in disfavor of FSRQ to account for the difference in efficiency for fitting the SEDs successfully, due to better broadband data for FSRQs (see Section 4.1). The reasonable agreement between the composite and BCU distributions for both photon index and ${\nu }_{{\rm{s}},\mathrm{peak}}$ seen in Figure 6 supports the idea that the sample of unclassified blazars (i.e., BCUs) is of a composition similar to that of the classified sample in 4LAC.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 7 displays the photon index as a function of the observed synchrotron peak frequency. Even though ${\nu }_{{\rm{s}},\mathrm{peak}}$ should be corrected by a factor 1+z to obtain the rest-frame value and make the correlation more physical, the fairly strong correlation already noted in previous catalogs is clearly visible. The correlation obtained in 3LAC was reproduced theoretically by Dermer et al. (2015) using an equipartition blazar model with a log-parabola description of the electron energy distribution. In the region of ${\nu }_{{\rm{s}},\mathrm{peak}}$ where BL Lac LSPs and FSRQs overlap, their photon index distributions are very similar. This is the expected region for objects that might be transitioning between being FSRQs and BL Lacs. Ruan et al. (2014) found six such transitioning objects. Five of them, all of the LSP subclass, are present in 4LAC: OJ 451 (4FGL J0833.9+4223, FSRQ, ${\rm{\Gamma }}=2.44\pm 0.07$), TXS 1013+054 (4FGL J1016.0+0512, FSRQ, ${\rm{\Gamma }}=2.18\pm 0.04$), PKS 1247+025 (4FGL J1250.6+0217, BLL, ${\rm{\Gamma }}=2.00\pm 0.10$), 5C 12.291 (4FGL J1308.5+3547, FSRQ, ${\rm{\Gamma }}=2.29\pm 0.06$), and PMN J2206−0031 (4FGL J2206.8−0032, BLL, ${\rm{\Gamma }}=2.25\pm 0.05$). Three of these sources have Γ very close to the Γ = 2.25 limit outlined above. The four HSP FSRQs in the Clean Sample all have ${\rm{\Gamma }}\lt 2.02$, as expected from the class of "HFSRQs," which is not fitting with the "blazar sequence" (see, e.g., Padovani 2007).

Zoom In Zoom Out Reset image size

Some blazars have undergone statistically significant spectral changes since 3LAC: PKS 1349−439 (BL Lac, ${\rm{\Delta }}{\rm{\Gamma }}\equiv {{\rm{\Gamma }}}_{4\mathrm{FGL}}-{{\rm{\Gamma }}}_{3\mathrm{FGL}}=-0.37\pm 0.12$), RX J1415.5+4830 (BL Lac, ${\rm{\Delta }}{\rm{\Gamma }}=-0.70\pm 0.16$), PKS 1532+01 (FSRQ, ${\rm{\Delta }}{\rm{\Gamma }}=-0.41\pm 0.12$), and S4 1800+44 (FSRQ, ${\rm{\Delta }}{\rm{\Gamma }}\,=-0.33\pm 0.09$). The changes in photon index are all such that ${{\rm{\Gamma }}}_{4\mathrm{FGL}}$ is closer to the median of the class than was ${{\rm{\Gamma }}}_{3\mathrm{FGL}}$. Inspection of the light curves reveals that all four sources show enhanced activity in the last 4 yr of the 4LAC period relative to that of 3LAC, which may be correlated with the observed spectral hardening. Fourteen other sources have experienced spectral variations greater than 3σ but of lower amplitudes than these four.

As noted above, many FSRQs and BL Lacs show significant spectral curvature. The comparison of the TS distributions of sources with significantly curved spectra to those of the whole sample of FSRQs and BL Lacs (Figure 8) demonstrates that essentially all bright blazars have curved spectra in the LAT energy range. A total of 212 FSRQs, 172 BL Lacs, and 70 BCUs have significantly curved spectra. To enable comparison, if we apply the more stringent threshold on the curvature significance (i.e., approximately 4σ instead of 3σ) used in 3LAC, these numbers become 146 FSRQs, 112 BL Lacs, and 26 BCUs. In 3LAC, the comparable numbers were 57, 32, and 8, respectively. It is therefore likely that fainter blazars have curved spectra as well, but the current data do not allow their curvature to be established with high confidence.

Zoom In Zoom Out Reset image size

We conducted a literature search for spectroscopic redshifts. Well-established redshifts (999) came from BZCAT or the optical campaigns mentioned in Section 3.3.1. For the other sources (656), remaining contamination from photometric values or from bad signal-to-noise ratio optical spectra cannot be excluded. We found redshifts for all the FSRQs, but were unable to find those for 36% of the BL Lacs in our sample (compared to 50% of the BL Lacs without redshifts in 3LAC). This clear improvement has been primarily achieved thanks to follow-up observations of 3LAC blazars (see Section 3.3.1 for references). The fraction without redshifts is similar for the three BL Lac subclasses (41%, 42%, and 28% for LSPs, ISPs, and HSPs, respectively).

The redshift distributions are displayed in Figure 9 for FSRQs and BL Lacs. The FSRQ distribution shows a broad peak around z = 1. This trend confirms the conclusion that the number density of FSRQs grows dramatically up to redshift ≃0.5–2.0 and declines thereafter (Ajello et al. 2012). For BL Lacs, the overall peak lies at $z\simeq 0.3$. For the sake of comparison, the distributions for previously and newly reported AGNs are plotted separately. The redshifts of the 3LAC and newly detected blazars are similar, with respective medians and widths of 1.14 ± 0.62 and 1.09 ± 0.68 for FSRQs, and of 0.34 ± 0.42 and 0.36 ± 0.34 for BL Lacs. The median redshifts decrease between BL Lac LSPs, ISPs, and HSPs from 0.47 to 0.36 to 0.25, respectively. While the maximum redshift for an FSRQ was 3.1 in earlier LAC catalogs, five counterparts to 4LAC sources have higher redshifts: GB 1508+5714 (z = 4.31), PKS 1351−018 (z = 3.72), PKS 0335−122 (z = 3.44), MG3 J163554+3629 (z = 3.65), and PMN J0833−0454 (z = 3.45). GB 1508+5714 and PKS 0335−122 are not present in the Clean Sample, however, due to analysis flags. The detections of three of these sources (PKS 1351−018, GB 1508+5714, MG3 J163554+3629 were reported earlier in Ackermann et al. (2017). Two other high-z sources (NVSS J064632+445116 and NVSS J212912-153841), whose detections were also announced in Ackermann et al. (2017), are absent in the 4FGL/4LAC, possibly due to variability effects.

Zoom In Zoom Out Reset image size

Figure 10 displays the photon index versus the gamma-ray luminosity for the different blazar classes. The trend of softer spectra with higher luminosity observed in earlier catalogs is confirmed. We reiterate the word of caution expressed in 3LAC: this trend is only significant when considering the whole sample of 4LAC blazars. It is not significant when considering the different blazar classes/subclasses individually. Figure 11 shows the corresponding plot for the nonblazar sources. Radio galaxies show a large scatter in photon index, while sources of the other classes have fairly soft spectra akin to those of FSRQs.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Variability is a key property of blazars and is known to depend on the energy band considered (e.g., Aleksić et al. 2015). This feature can be naturally explained as emitting electrons (assuming a leptonic scenario) of different energies, and thus different acceleration/cooling times contribute preferentially to the distinct bands. The assessment of variability in 4FGL does not only depend on intrinsic variability, but also on the overall significance of the source detection. Two different values of the variability index⁷⁹ are provided in 4FGL, derived from the sets of 1 yr and 2 month light curves. The 1 yr light curves allow the variability of fainter sources to be established compared to the monthly light curves used in early FGL catalogs. Because the variability indices are distributed as ${\chi }^{2}$ functions with N_dof degrees of freedom, a source is defined as variable if at least one of the variability indices is greater than the 99% confidence limit of 18.48 and 72.44 for the 1 yr (N_dof = 7) and 2 month (N_dof = 47) light curves, respectively.

Figure 12 displays the TS distributions of variable sources compared to those of the whole sets of FSRQs and BL Lacs. All bright blazars are found to be variable. The fraction of variable sources goes down from 79% (464/591) for FSRQs to 35% for BL Lacs (362/1027). A monotonic trend is observed for the BL Lac subclasses, with fractions of 49% (140/288), 36% (98/270), and 31% (98/316) for LSPs, ISPs, and HSPs, respectively. Only 17% (157/941) of the BCUs show variability, as expected from the fact that these sources tend to be fainter (Section 5.1). The median fractional variability amplitude is 0.63 for FSRQs and 0.27 for BL Lacs. More extended studies about variability of the LAT-detected blazars can be found in the 3LAC paper.

Zoom In Zoom Out Reset image size

Among the radio galaxies, IC 310, NGC 1275, 3C 120, NCG 1218, 3C 111, NGC 2892, and IC 4516 are found to be variable, hinting at a blazar-like behavior for these sources. All NLSy1s show significant variability, except for IERS B1303+515, B3 1441+476, TXS 2116−077. Three out of five CSS sources are variable (3C 138, 3C 309.1, 3C 380), as well as one SSRQ (3C 207). The three variable sources of the AGN class are: PKS 0521−36, PMN J1118−0413, and CGCG 050−083.

Some Fermi-LAT sources show blazar-like flaring activity during limited time intervals but do not meet the detection significance criterion to be included in the 4FGL, which is based on a summed 8 yr data set. We present here information about some of these transient sources spatially consistent with AGNs.

When undergoing periods of enhanced activity in 6 hr and 1 day time intervals, some sources can be caught in near-real time by the Fermi-LAT Flare Advocate Gamma-ray Sky Watcher (FA-GSW) service (Ciprini & Fermi-LAT Collaboration 2012; Thompson et al. 2015, and references therein). The brightest sources are often reported to the community in Astronomer's Telegrams (ATels). A total of 371 ATels (plus three errata) were posted on behalf of the LAT Collaboration in the 8 yr period considered in the 4FGL/4LAC (2008 August 4 to 2016 August 2), and a total of 472 ATels were published from 2008 July 24 (ATel 1628) to 2019 August 15 (ATel 13032). At the time of writing, announced transient sources that were positionally consistent with blazars or other AGNs include: NVSS J104516+275136 (ATel 12906); S5 0532+82 (ATel 12902; PKS 2247−131 (ATels 9285, 9620, 11141); PKS 1915−458 (z = 2.47, ATels 2666, 2679, and already reported as missing in the 3FGL/3LAC catalogs); TXS 0135+291 (ATel 12888); 2MASX J15441967−0649156 ($z\simeq 0.04$, ATel 10482); PKS 1251−71 (ATel 8215), PMN J0508−5628 (ATel 6658); PMN J2010−2524 (z = 0.825, ATel 6553); TXS 1731+152A (ATel 6395, 6410); PKS 2136−642 (ATel 5695); and PKS 1510−319 (z = 1.71, ATel 2528).

Sources detected on a one-week timescale were reported in the Second Fermi-LAT All-sky Variability Analysis (FAVA) Catalog (2FAV, Abdollahi et al. 2017), based on the first 7.4 yr of Fermi-LAT mission data. In the FAVA catalog,⁸⁰ the analysis was run in 100–800 MeV and 0.8–300 GeV energy bands, leading to the identification of 518 flaring gamma-ray sources. Among these sources, 13 were associated with established blazars or blazar candidates that are included in neither 4LAC nor 3LAC: PMN J0231−4746, 2MASS J06164292−4021527, TXS 0723+220, 4C +38.28 (B2 0913+39), PKS 1200−051, PMN J1322−8419, PKS 1354−17, RX J1410.5+6100, PKS 1510−319, TXS 1534+378, TXS 1731+152A, PKS 1824−582, 1RXS J235018.0−055928.

For longer timescales, the first catalog of gamma-ray transient sources (1FLT, The Fermi LAT Collaboration in preparation) reports detections on monthly time intervals during the first 96 months of Fermi-LAT operation. This catalog contains 64 new gamma-ray sources not present in 4LAC/4FGL or earlier LAT catalogs. Their mean photon spectral index Γ of 2.6 indicates softer spectra than exhibited by the 4LAC sources (mean ${\rm{\Gamma }}\sim 2.2$). These new sources include 24 BCUs (e.g., PKS 1649−031, TXS 0209+168, TXS 1601+160, PKS 2108−326 and others), 20 FSRQs (e.g., PKS 1524−13, TXS 1226+046, PKS 1200−051, PMN J2010−2524, PKS 1706+006 and others), one BL Lac object (1RXS J112100.6+014515), one NLSy1 (Mkn 1501), and two radio galaxies (S5 1733+71 and PKS 2236−364).

The highest-energy photons detected for each source were selected within the purest—i.e., with the lowest instrumental background—class (P8R3_ULTRACLEANVETO_V2). Based on the energy-dependent PSF, we required a probability⁸¹ greater than 0.95 for the photons to belong to the source being considered. Blazar gamma-ray spectra provide insight into the density of the EBL via the effect of photon-photon absorption, often defined in terms of an optical depth τ = 1, where a photon has only a 1/e probability of reaching the observer (e.g., Abdollahi et al. 2018). A comparison between the energy of the highest-energy photon measured for a given source with the energy computed for τ = 1 at that redshift by theoretical models of the EBL provides a simple and direct test of these models. Figure 13 compares the 4LAC highest-energy photons with the optical depth predictions of Finke et al. (2010), Domínguez et al. (2011), and Gilmore et al. (2012). Only a few photons exceed the τ = 1 mark, thus remaining compatible with the predictions of these models.

Zoom In Zoom Out Reset image size

The blazars detected in gamma-rays after 8 yr of LAT operation represent a sizable fraction of the whole population of known blazars as listed in BZCAT. This catalog represents an exhaustive list of all sources ever classified as blazars, but is by no means complete. Although a comparison between the gamma-ray-detected and nondetected blazars within that sample has no strong statistical meaning in terms of relative weights, it is nevertheless useful to look for general trends.

In total, the 4LAC includes $29$% ($562$/$1909$), $66$% ($764$/$1151$), and $38$% ($88$/$227$) of the BZCAT (Massaro et al. 2015) FSRQs, BL Lacs, and BCUs, respectively. Out of the $1353$ new 4FGL sources, $405$ are present in BZCAT. Figure 14 compares the distributions of radio flux at 1.4 GHz, optical R-band magnitude, and X-ray (0.1–2.4 keV) fluxes for the gamma-ray-detected and nondetected BZCAT blazars, as well as the fraction of gamma-ray-detected blazars relative to the total as a function of the different fluxes. The gamma-ray-detected blazars are somewhat brighter on average in all bands, confirming previous findings (Lister et al. 2011; Böck et al. 2016; Paliya et al. 2017). The fraction of 4LAC blazars in the total population of BZCAT blazars remains non-negligible even at the faint ends of the radio, optical, and X-ray flux distributions, in particular for BL Lacs. This observation is a clue that even the faintest known blazars could eventually shine in gamma-rays at LAT-detection levels.

Zoom In Zoom Out Reset image size

Table 7 shows the list of 78 AGNs detected by ground-based Cerenkov telescopes, as listed in TeVCat.⁸² All are present in 4LAC, and the table shows their optical and SED-based classes, redshifts, and 4LAC photon indices. The 3LAC catalog included 55 of the 56 VHE AGNs with detections published or announced at that time; only HESS J1943+213 was missing. The overall mean of the 4LAC photon indices for these VHE AGNs is 1.91 ± 0.20. For the most numerous subclass, the HSP BL Lacs, we find the mean index to be 1.81 ± 0.08, i.e., a very hard gamma-ray spectrum. Of the 78, 56 4LAC AGNs are variable at a significance greater than 99%.

In the course of the 4LAC analysis, we found a number of individual sources that have changed classification or have unclear associations. We include this information about them for completeness. In case of conflicting associations, we only retained the most probable one in the 4LAC catalog.

• 1.  
  4FGL J0140.6−0758 is associated with the BL Lac RX J0140.7−0758. Another object, SDSS J014040.63−075857.2, is located 9'' away (i.e., 0.04r₉₅, where r₉₅ is the 95% confidence radius), at a redshift of 2.674 coming from a broad-line SDSS optical spectrum. This source has no reported radio emission.
• 2.  
  4FGL J0242.3+5216 has two possible high-confidence associations: with GB6 J0242+5209 and TXS 0239+520, which have similar brightness in the radio band.
• 3.  
  4FGL J0337.8−1157 is associated with PKS 0335−122, a distant FSRQ at z = 3.442 that has a damped Lyα absorption system at z = 3.180 along the line of sight (Kanekar & Chengalur 2003).
• 4.  
  4FGL J0618.1+7819 is associated with the starburst galaxy NGC 2146 (z = 0.002975), but the FSRQ 6C 060948+781625 (z = 1.43) has a similar association probability.
• 5.  
  4FGL J0647.7−4418 is associated with the high-mass X-ray binary RX J0648.0−4418 in 4FGL, but the blazar candidate SUMSS J064744−441946 has a just barely lower association probability (0.80 versus 0.85).
• 6.  
  4FGL J0720.0−6237 is associated with the FSRQ PMN J0716−6240, but another FSRQ, PMN J0719−6218, lies just outside the 95% confidence region and might contribute to the gamma-ray emission.
• 7.  
  4FGL J0814.4+2941 is associated with the BL Lac EXO 0811.2+2949, but a broad-line quasar, SDSS J081425.89+294115.6, also lies within the 95% confidence region.
• 8.  
  4FGL J0941.9+2724 may have a double association with the BL Lac 5BZBJ0941+2722 and the FSRQ MG2 J094148+2728, which was an association in 3FGL.
• 9.  
  4FGL J1300.4+1416 (3FGL J1300.2+1416) is associated with OW 197 (PKS 1257+145, z = 1.1085), as it was in 3LAC. At about 5' distance from OW 197, a blazar candidate NVSS J130041+141728 lies in both the 3FGL and 4FGL 95% confidence ellipses, despite a reduction in size of the ellipse by a factor of 4. The 4FGL source position is about midway between OW 197 and NVSS J130041+141728.
• 10.  
  4FGL J1625.7+4134 may have a double association, with 4C +41.32 and B3 1624+414.