Catalog of Long-term Transient Sources in the First 10 yr of Fermi-LAT Data

We have partitioned each of the 240 monthly TBINs into 192 circular regions of interest (ROIs) centered on points defined by means of a HEALPix ⁷⁸ sky pixelization (Górski et al. 2005) with N_side = 4 and adopting celestial coordinates. The ROIs include events in cones of 15° radius about the center of the pixel. The ROIs constructed in this way are not independent, as they partially overlap.

To identify the candidate sources (seeds), we used a wavelet transform analysis (Damiani et al. 1997) using the PGWave tool (Ciprini et al. 2007). PGWave works on the square images inscribed in ROIs of 15° radius; therefore, the side of the square images is $15^\circ \sqrt{2}\sim 21.2$ degrees. We adopted a pixel size of 025 and the stereoscopic projection for the images. We performed the analysis independently on each ROI and TBIN deleting the seeds located at the border of the ROI (distance from the ROI center >10°) in order to avoid edge effects. In this way we obtained 129,802 seeds in the nominal months and 133,673 seeds in the shifted ones.

After the seed extraction, we performed an ML analysis on all seeds that had Galactic latitude ∣b∣ > 10° and that had an angular distance greater than 50' from any 4FGL-DR2 source (with this selection we excluded ∼10% of the extragalactic sky). The conservative 50' cross-correlation radius corresponds to the average plus 1σ value of the distribution of the 1FLT catalog's semimajor axis of the error ellipses at 95% confidence level (Conf_95_SemiMajor in Table A1). The ML analysis provides the evaluation of the test statistic (TS), defined as TS = 2Δlog ${ \mathcal L }$ (Mattox et al. 1996), which quantifies how significantly a source emerges from the background comparing the likelihood function ${ \mathcal L }$ with and without a source. We performed the final localization only for seeds with TS > 10 (4872 for the nominal months and 4737 for the shifted ones).

We evaluated the significance of the detection and the spectral parameters for the newly found sources by performing a binned ML with eight bins per decade in energy and 01 binning in Galactic coordinates. For each point source we assumed a power-law spectrum defined as dN/dE = K∗(E/E₀)^−Γ, where K is the flux scale factor, Γ is the spectral index, and E₀ is the reference energy (E₀ = 1 GeV). The ROI model used in the likelihood analysis contained all PGWave seeds found (including also the ones associated with 4FGL-DR2 sources) in the selected month in an ROI of 10° × 10° square, centered at the position of the target, along with the Galactic and isotropic diffuse backgrounds. ⁷⁹ To perform the analysis, we used the Fermitools 1.2.1 package available from the Fermi Science Support Center (FSSC) ⁸⁰ and the Fermipy 0.18.0 software package (Wood et al. 2017). We use TS = 25 as the detection threshold, corresponding to a significance of ∼4σ assuming a χ² distribution with 4 degrees of freedom (position and spectral parameters of a power-law source; see Mattox et al. 1996).

The evaluations of the false-positive rate and efficiency are fundamental to assess the reliability of the adopted method in the detection of transient sources. In order to calculate the false-positive, rate we simulated 120-month-long data sets, including in the model only the Galactic and the extragalactic diffuse emissions. Over month-long timescales the false positives are dominated by statistical fluctuations of the background rather than systematic effects. For the simulations we adopted the analysis of the 1FLT. We identified the seeds as described in Section 2.1 and performed the binned ML analysis assuming a power-law spectrum for each seed. The distributions of the statistical significance of the candidate sources found in the fake and true skies are shown in Figure 1. In the 120 simulated skies we found 12 spurious detections with TS > 25 over ∼24,000 seeds, and one and two detections with TS > 30 and TS > 28, respectively. To be conservative, in the estimation of the spurious detections in the real data these numbers should be multiplied by a factor of two to account also for the shifted months. Therefore, we may expect 24 fake detections with TS > 25.

Zoom In Zoom Out Reset image size

Since our adopted threshold of TS > 25 still includes a nonnegligible number of spurious sources, in the 1FLT catalog TS > 25 candidate sources are flagged as low-confidence when they are detected in only one TBIN and have a TS value between 25 and 30. A total of 72 1FLT sources are flagged as low-confidence. Each individual low-confidence source has a probability of about 34% of being spurious.

To assess the detection efficiency we used astrophysical γ-ray sources known to be steady in flux such as pulsars (PSRs) located at high latitude reported in the 4FGL catalog. The majority of sources taken in consideration for the estimation of efficiency are millisecond pulsars.

For each TBIN of data we applied the analysis procedures targeting the selected PSRs, and we compared the number of detections obtained as a function of energy flux to the total number of pulsars (efficiency of 100% when we had a significant detection of the PSRs in all 240 months). Figure 2 represents the rate of detection over 240 months for the selected PSRs as a function of energy flux. As suggested in Principe et al. (2018), we modeled the detection efficiency with a hyperbolic tangent function tanhλ(f − f₀), where the two parameters λ and f₀ are determined by fitting to the detection efficiency points. We obtained the following values: λ = 4.0 × 10⁴ MeV⁻¹ cm² s and f₀ = 8.41 × 10⁻⁷ MeV cm⁻² s⁻¹ with a χ² of 0.93 with 9 degrees of freedom.

Zoom In Zoom Out Reset image size

The 1FLT catalog lists all the sources that satisfy the following criteria: a TS > 25, a ratio between the flux error and the flux ΔFlux/Flux < 0.5, and a power-law index Γ < 4.5. We associated multiple detections with the same source when they are spatially coincident with 2D-sky cross-correlation using the error ellipses at 95% confidence level. We manually checked each individual multiple association to verify the automatic procedure, and in three cases we did not consider the sources as multiple detections since the centroids of the γ-ray emission were visually not cospatial (1FLT J0240−4657 and 1FLT J0240−4739; 1FLT J0519−3709 and 1FLT J0527−3747; 1FLT J1322−4521 and 1FLT J1323−4439). When the same source was detected on different but overlapping ROIs, we reported the results of the most significant detection. A reference of this repeated occurrence is reported in the catalog column "repROI." For instance, see Appendices A and B, as well as the catalog file.

The final 1FLT catalog is composed of 142 unique transient sources not associated with any 4FGL-DR2 emitters: 64 sources detected in the nominal 120 months and 78 in the shifted 120 months. Seventy-two 1FLT sources with 25 < TS < 30 are flagged as low-confidence (as reported in Section 2.2). The complete list of sources is available in electronic format (FITS format) as supplementary material. The columns are described in Appendix A in Tables A1–A3. In Appendix B, Table B1 has the 1FLT list. The source designation is 1FLT JHHMM+DDMM.

Out of these 142 1FLT sources, 108 1FLT sources were detected only once, i.e., in one single TBIN, while 34 other distinct 1FLT sources displayed significant gamma-ray emission in more than one TBIN. These 34 sources listed in 1FLT correspond to their most significant detection in each relative cluster of multiple flaring episodes. Forty-four remaining flaring episodes are listed in the extension FLARES of the electronic release that includes other useful information specific to each flaring episode, such as time, localization, flux, significance, and spectral parameters. The column "Flares" in Table B1 contains the number of flaring episodes of each 1FLT source. In case of multiple detections of positionally consistent candidates, the source name is based on the position of the candidate detection with the largest value of TS.

The 1FLT method allowed us to collect also 60 detections associated with the Sun and 27 associated with 14 of the brightest Fermi-LAT γ-ray bursts (GRBs). In Appendix B, Table B2 lists the Sun detections, and Table B3 contains GRB detections.

The sky locations of the 1FLT sources, together with Sun and GRB detections, are shown in Figure 3.

Zoom In Zoom Out Reset image size

To evaluate the possible contamination by the Moon of the 1FLT source catalog, we checked the times and trajectories of the passage of the Moon through the 1FLT fields. For the overlaps between the Moon's path and 1FLT source localization we extracted the 1FLT fluxes with a 1-day timescale in the referring month, to verify simultaneously the coincidence in time and space. We found no evidence for the Moon being confused with any 1FGL detection because when we detected the higher 1-day time bin flux the Moon localization was more than 5° away from the 1FLT-derived position.

For each source in the 1FLT catalog we extracted the γ-ray spectrum and monthly binned light curve. The spectral energy distributions (SEDs) are produced in three energy bands (i.e., 0.1–1 GeV, 1–10 GeV, 10–100 GeV) ⁸¹ by modeling the 1FLT source with a power-law spectrum with free normalization and a spectral index fixed to 2. When the TS of the spectral bin was less than 4 or the value of spectral data point error was greater than 50% of the measured value, we calculated the upper limit at 2σ confidence, using the Bayesian computation if TS < 1 (Helene 1984). The γ-ray spectral data points (νF_ν ) are reported in the catalog available in electronic format as supplementary material.

The light curves were produced using the binned ML technique, over the energy range of 100 MeV–300 GeV. The likelihood fit followed the same procedure described above (see Section 2), fixing the photon index of the 1FLT source to the 1FLT catalog value. The ROI model included all PGWave seed detections of the TBIN located in an ROI of 10° centered at the 1FLT catalog position of the source of interest. When the target was detected with TS < 4, or the number of its predicted photons was N_pred < 3, or the uncertainty on its flux estimate was large (ΔFlux/Flux > 0.5), upper limits were calculated based on the method of Helene (1984). For consistency with the 1FLT catalog procedure, the light curves start coherently with the nominal or shifted time selection. The statistical uncertainty on the fluxes is typically larger than the systematic uncertainty (Ackermann et al.2012), and only the former is considered in this paper. All light curves are shown in Appendix C.

In order to classify candidate γ-ray counterparts, we used either the optical classifications published in the BZCAT list (a compilation of sources classified as blazars; Massaro et al. 2015) or the spectra available in the literature or from online databases. The main ones are the Sloan Digital Sky Survey (SDSS; Alam et al. 2015; Albareti et al. 2017; Abolfathi et al. 2018) and the 6dF Galaxy Survey (Jones et al. 2009). The classes have been assigned with the following criteria:

• 1.  
  FSRQ, BL Lac-type object, compact steep-spectrum radio source (CSS), steep-spectrum radio quasar (SSRQ), and radio galaxy (RG): sources with a well-established classification in the literature and/or through a good-quality optical spectrum;
• 2.  
  blazar candidate of uncertain type (BCU): an object in the listed BZCAT as blazar of uncertain/transitional type or a source with blazar multiwavelength characteristic—flat radio spectrum and double-humped, broadband SED;
• 3.  
  nonblazar active galaxy: for these candidate counterparts the existing data do not allow an unambiguous determination of the AGN type;
• 4.  
  unassociated: sources without an unequivocal counterpart or without any plausible candidate at other wavelengths

The class designations for the 1FLT sources are listed in Table 1.

As reported in Table 1, the 1FLT includes 142 sources, with 24 FSRQs, 1 BL Lac, 1 CSS, 1 SSRQ, 3 RGs, 70 BCUs, and 2 other AGNs. This sample is composed exclusively of jetted AGNs, which is also the largest population present in the xFGL catalogs. The sample composition, however, is different when compared to that of the Fermi-LAT general catalogs. There are 24 FSRQs, which represents a higher fraction as compared to the xFGL catalogs. Twenty-seven sources are associated with the Bayesian method, 75 with the positional method. Forty sources remain unassociated (28.2%). This fraction is similar to that reported in the xFGL catalogs. The low-confidence sample includes 72 sources, among which associations are found with 29 BCUs, 1 BL Lac, 10 FSRQs, 2 RGs, 1 SSRQ, and 2 other types of AGNs. These low-confidence sources were associated with the Bayesian method in four cases (one BCU and three FSRQs). The high-confidence sample includes 70 sources, among which associations are found with 41 BCUs, 1 CSS, 14 FSRQs, and 1 RG. In this sample, the Bayesian method finds 23 associations (1 CSS, 5 FSRQs, and 17 BCUs). 1FLT includes six sources that are associated with a source from previous FGL catalogs (1FGL, 2FGL, 3FGL) but that do not have a counterpart in the 4FGL-DR2 list. These associations are reported in the catalog table in the column named ASSOC_FERMI. In that column we also include the associations (14 1FLT sources) with the preliminary 8 yr Fermi-LAT source list (FL8Y). ⁸⁴ In the catalog table column ASSOC_GAM we report the associations with other γ-ray source lists: one association in the Second AGILE catalog of gamma-ray sources (Bulgarelli et al. 2019), two associations in the Swift-BAT 105-Month Hard X-ray Survey (Oh et al. 2018), and one association in the INTEGRAL/IBIS AGN catalog (Malizia et al. 2016).

Figure 4 shows the distribution of the power-law photon index Γ for the 1FLT sources, along with that for the 4LAC sources that are associated with FSRQs and BL Lacs. The 1FLT distribution extends to softer Γ values (median value of Γ ∼ 2.7) than those of the 4LAC sources (median value of Γ ∼ 2.2 or 2.5 if we consider only 4LAC FSRQs). As shown in Figure 4, 26 1FLT sources exhibit softer spectral indices than the softest 4LAC FSRQ. We performed the Kolmogorov–Smirnov (K-S) test (Kolmogorov 1933) to evaluate whether the 1FLT spectral index distribution comes from the same parent population of 4LAC subclass distributions (null hypothesis). We found that our null hypothesis is rejected in all cases at a confidence level of 99.9% (α = 0.001). The identification of these soft-spectrum sources shows that integrating over monthly timescales allowed us to investigate the soft part of the gamma-ray source distribution, which is usually suppressed (or confused with the low-energy component of the γ-ray background) when analyzed over longer, multiyear timescales.

Zoom In Zoom Out Reset image size

The population of soft-spectrum gamma-ray sources could include members of the so-called "MeV-blazars" (blazars that are exceptionally bright at MeV energies; Bloemen et al. 1995; Blom et al. 1995). These objects have shown variability in their γ-ray flux when integrated over timescales of months. However, their variability timescales are longer than those shown by harder-spectrum blazars (Sikora et al. 2002). The MeV-peaked emission can be modeled with the external Comptonization of accretion disk radiation or of the near-IR emission of hot dust (Bednarek et al. 1996). The transient nature of the MeV-blazar phenomenon is evidence for the presence of this population of soft FSRQs with a steep Γ(E > 100 MeV) on an approximately 1-month timescale. The month-scale variability and the pronounced increase of the emission in the keV–MeV regime are also reported by Kreter et al. (2020) in their search for high-redshift blazars. The large distances of these sources and the subsequent redshifting of their emission mean that the γ-ray emission is shifted toward longer wavelengths.

For all of the 1FLT sources associated with blazars we collected the available multifrequency data in order to infer the synchrotron peak frequency ${\nu }_{\mathrm{peak}}^{{\rm{S}}}$ of the observed broadband SED, which was in turn used to perform an SED-based classification. We classified blazars as low synchrotron peaked (LSP, ${\nu }_{\mathrm{peak}}^{{\rm{S}}}\lt {10}^{14}$ Hz), intermediate synchrotron peaked (ISP, 10¹⁴ Hz < ${\nu }_{\mathrm{peak}}^{{\rm{S}}}\lt {10}^{15}$ Hz), or high synchrotron peaked (HSP, ${\nu }_{\mathrm{peak}}^{{\rm{S}}}\gt {10}^{15}$ Hz). In this procedure we used a compilation of broadband archival data as described for the 2LAC (Ackermann et al. 2011b). The estimation of ${\nu }_{\mathrm{peak}}^{{\rm{S}}}$ relies on a third-degree polynomial fit of the low-energy hump of the SED that is performed on a source-by-source basis (the procedure was adopted in Ackermann et al. 2015). With this procedure we reconstructed the ${\nu }_{\mathrm{peak}}^{{\rm{S}}}$ for only 65 1FLT sources owing to the limited multifrequency sampling. Figure 5 shows the power-law index versus the ${\nu }_{\mathrm{peak}}^{{\rm{S}}}$ for both the 1FLT and 4LAC sources. The strong correlation evident in the 4LAC sample appears to be diluted for the sources in the 1FLT catalog since we detected only LSP sources with soft spectra. We noted, however, that the number of 1FLT catalog sources with ${\nu }_{\mathrm{peak}}^{{\rm{S}}}$ estimation is just a few percent of those in the 4LAC catalog.

Zoom In Zoom Out Reset image size

For each 1FLT source that has a counterpart with a known redshift, we computed the γ-ray and radio luminosity using the ΛCDM cosmological parameters obtained by Planck (Planck Collaboration et al. 2014); in particular, we used h = 0.67, Ω_m = 0.32, and Ω_Λ = 0.68, where the Hubble constant H₀ = 100h km s⁻¹ Mpc⁻¹. Redshifts are available for a total of 30 1FLT sources: all FSRQs, 1 BCU, 1 BLL, 3 RG, and 1 nonblazar object.

Analyzing the distribution of spectral index versus γ-ray luminosity for all the sources of our sample, along with the corresponding values for all sources of the 4LAC (Ajello et al. 2020), plotted in Figure 6, we can see that the 1FLT sources and the 4LAC sources occupy different regions of the parameter space. The 1FLT sources show greater γ-ray luminosities, possibly because they are detected during flaring states, and softer γ-ray spectra than the sources detected by integrating over years-long periods. When integrating over short time intervals, sporadic flaring activity is detected from a subluminous population whose integrated emission would not have been detected over longer time intervals, due to their weak quiescent flux being merged into the integrated background gamma-ray emission. It is noteworthy that more than half of the 1FLT sources lie in a region of the parameter space that is only sparsely populated by 4LAC sources.

Zoom In Zoom Out Reset image size

Plotting the ratio of γ-ray luminosity to radio luminosity as a function of ${\nu }_{\mathrm{peak}}^{{\rm{S}}}$ (Figure 7), we can see that the ratio is clearly greater for the 1FLT than for the 4LAC; this could be explained simply by the fact that the γ-ray flux detection and the radio flux measurement are not simultaneous and the gamma-ray measurements are by definition made during γ-ray flaring episodes. These sources are in fact very bright during a flare but show a faint long-term integrated γ-ray flux, which limits their possibility of being detected significantly on the long-term periods typically considered for xFGL catalogs.

Zoom In Zoom Out Reset image size

In terms of the distribution of the radio fluxes for the 1FLT sources, our sources show fainter fluxes than those of the 4LAC FSRQs (using a K-S test, we found that our null hypothesis is rejected at a confidence level of 99.9%) but are comparable to those of the 4LAC BCUs as shown in Figure 8 (using a K-S test, we found that the two data sets are not very similar but do not differ by enough that we could reject the null hypothesis: p = 2.14% and distance = 0.16). This is because most of the 1FLT sources are BCUs.

Zoom In Zoom Out Reset image size

Figure 9 shows the redshift distributions for 4LAC FSRQs and 1FLT sources. The redshift is available for all FSRQs, 1 BCU, 1 BLL, 3 RG and 1 nonblazar object. The 1FLT redshift distribution shows a peak at low redshift (<1). This peak is due to the contribution of 1FLT misaligned AGNs. These sources show a larger jet inclination angle, where de-boosted radiation makes the relativistic jet radiation weaker and more difficult to detect in γ-rays (Abdo et al. 2010d). However, ∼50% of FSRQs are detected at z > 1, suggesting a different distribution. Another effect, in fact, that can be seen in Figure 9 is that there is an additional peak at z > 2, unlike the 4LAC distribution, which includes only a small fraction of objects at z > 2. This shows that the 1FLT method is well suited for detecting high-redshift blazars. These results are in agreement with the behavior of the population of high-redshift blazars reported in Kreter et al. (2020).

Zoom In Zoom Out Reset image size

The Fermi All-sky Variability Analysis (FAVA) searches for transient sources over the entire sky observed by the Fermi-LAT on 1-week time intervals. The second catalog of flaring γ-ray sources (2FAV) spans the first 7.4 yr of Fermi-LAT observations and reports 518 variable γ-ray sources with significances of at least 6σ. Based on positional coincidence, candidate counterparts have been found for 441 sources, most of them blazars. The catalog also provides for each source the time, location, and spectrum of each flaring episode (Abdollahi et al. 2017).

Given the variable nature of 2FAV sources, we decided to compare the 1FLT catalog to the 2FAV catalog. We have investigated positional (within the error ellipses) and time correspondence between 1FLT sources and 2FAV sources.

We found correspondence in position and time for a few sources: 1FLT J1732+1510 (TBIN 71.5) with 2FAV J1732+15.2 (reported in an Astronomer's Telegram, ⁸⁵ ATel#6395), 1FLT J1919−4543 (TBIN 21) with 2FAV J1919−45.7 (ATel#2666), 1FLT J1936+5341 (TBIN 20.5) with 2FAV J1936+53.7, and 1FLT J2010−2523 (TBIN 72.5) with 2FAV J2009−25.4 (ATel#6553). The associations with 2FAV catalog sources are reported in the column named ASSOC_FAVA in the 1FLT catalog table. These 2FAV sources have the same counterparts as the corresponding 1FLT sources except in the case of 2FAV J1936+53.7, which is not associated, and 1FLT J1936+5341, which we associated with TXS 1935+536.

Looking at γ-ray candidate seeds in the 1FLT search with TS < 25 and not already in 4FGL-DR2, we found correspondence in position and time with two other 2FAV sources. 2FAV J0539−28.8 corresponds to a detection in TBIN 55 with TS = 9 and 2FAV J2056−11.5 to a detection in TBIN 60.5 with TS = 14. We attribute this to the dilution over the month timescale of the γ-ray emission that occurred during a brief (about 1 week) flare. This result confirms the potential of Fermi to find new sources when the data are integrated over different time intervals. It also shows that choosing a monthly time interval allows us to detect a different and new population of sources than those found when the data are binned on weekly timescales.

The duty cycle for an astronomical object of a given class was defined by Romero et al. (1999) as

$$\begin{eqnarray}&&\mathrm{DC}=100\displaystyle \frac{{\sum }_{i=0}^{n}{N}_{i}1/({\rm{\Delta }}{t}_{i})}{{\sum }_{i=0}^{n}1/({\rm{\Delta }}{t}_{i})} \% ,\end{eqnarray} \tag{ 1 }$$

where Δt_i = Δt_i,obs(1 + z)⁻¹ is the duration (corrected for redshift z) of a monitoring interval of a source of the selected class and N_i is equal to 0 when the source is not variable at Δ t_i , while N_i is equal to 1 when the source is variable. We adapted the calculation to our specific analysis and used the light curves for the evaluation of the DC. For each time bin in which the source is detected (i.e., TS > 4 or N_pred > 3 or ΔFlux/Flux < 0.5), we compared the flux to the "averaged subthreshold flux" (the flux averaged over 10 yr using the evaluation of spectral index reported in 1FLT). If the flux in the monthly time bin is above the "averaged subthreshold flux," we considered it in the calculation of the DC.

In our case the DC can be evaluated as DC = 100$\displaystyle \frac{{\sum }_{i=0}^{n}{N}_{i}}{120}$ since Δt_i is 1 month for each source. We can calculate the DC for each source class independently of its redshift. We also calculated the duty cycle for each transient source class in the 4FGL using the same procedure adopted for the 1FLT. Since the light curves reported in the 4FGL have a time binning of 2 months, which is different from our sampling interval, we can only perform a qualitative comparison between the duty cycles of the 1FLT and 4FGL sources. In Figure 10 we compare between the average values of the DC for each transient source class in the 1FLT and the 4FGL catalogs. The 1FLT sources are characterized by a very low duty cycle. Clearly there is a selection effect due to the way that the 1FLT catalog was built because a bias was introduced by the definition of a significant point in the light-curve extraction. We can, however, obtain some qualitative insights about the difference between the DC of the radio galaxies and the blazar-like objects. In fact, the radio galaxy class has a greater DC than the FSRQ and BCU classes, which are characterized by shorter flaring episodes.

Zoom In Zoom Out Reset image size

We discuss here some peculiar cases that deserve a detailed investigation and for which we built SEDs collecting all archival multiwavelength data ⁸⁶ using the SSDC-SED Tool ⁸⁷ and adding our data points for each of the TBINs.

5.3.1. Soft γ-Ray Blazars Detected with the Swift Burst Alert Telescope

Two 1FLT sources were also detected with the Swift Burst Alert Telescope (Krimm et al. 2013). The source 1FLT J0010+1056 is associated with a Bayesian probability of 0.96 with Mrk 1501, which is reported as SWIFT J0010.5+1057 in the Swift-BAT 105-Month Hard X-ray Survey. ⁸⁸ Mrk 1501 is an FSRQ located in the local universe at a redshift of z = 0.089338 (Sargent & Searle 1970). This source is the nearest FSRQ in the whole 1FLT sample. It was also reported by Arsioli & Polenta (2018). They searched for sources with short-lived γ-ray emission and found a TS ∼ 26 for this source in just one monthly bin, that from 2010 June. We detected this source in the TBIN 21.5 (2010 May 20 to June 20), thus confirming this short γ-ray flare from Mrk 1501.

The SED of Mrk 1501 is shown in Figure 11. The typical two-humped, blazar-like SED is clearly reproduced. Although the source was in a γ-ray flaring state, the Compton dominance (CD_γ ), defined as the ratio of the inverse Compton to the synchrotron peak luminosity, is, in logarithmic scale, less than 1 as is typical of HSP BL Lacs (Abdo et al. 2010a; see Figure 22 in Giommi et al. 2012).

Zoom In Zoom Out Reset image size

The source 1FLT J2010−2523 is associated with a Bayesian probability of 0.99 with PMN J2010−2524, a well-known FSRQ (z = 0.825, Massaro et al. 2015). It is also reported in the Swift-BAT 105-Month Hard X-ray Survey as SWIFT J2010.6−2521 and as a γ-ray emitter by Paliya et al. (2019). We detected the source in several temporal bins. The first detection was in TBIN 6 (2009 February 3 to March 5), followed by a detection in TBINs 68 and 68.5 (2014 April 5 to May 20), and then from TBIN 72 to TBIN 73.5 (2014 August 5 to October 20). The flaring activity of the source is quite evident in the light curve reported in Appendix C.

In contrast to Mrk 1501, the SED of PMN J2010−2524 has a CD_γ greater than 1, which is fairly common among FSRQs. The CD_γ can be seen in Figure 12.

Zoom In Zoom Out Reset image size

5.3.2. Compact Steep-spectrum Source

1FLT J1416+3447 was detected in two consecutive and overlapping TBINs, TBIN 111.5 and TBIN 112 (2017 November 19 to 2018 January 4), which strengthens the reliability of the detection. It is associated, through the Bayesian method, with a probability of 0.85, with a radio CSS source, S4 1413+34. The SED of S4 1413+34 is shown in Figure 13.

Zoom In Zoom Out Reset image size

CSS sources are thought to represent the first stages of evolution of the radio sources that eventually physically expand as Fanaroff-Riley I (FR I) or Fanaroff-Riley II (FR II) radio galaxies. The γ-ray emission in CSS sources was predicted by Stawarz et al. (2008), but these sources have proved to be elusive in the xFGL catalogs. Recently, Schulz et al. (2016) reported on the results of a dedicated analysis of PKS 2004−447, which is associated with a γ-ray-loud young radio source that has a radio spectrum resembling that of a CSS and an optical spectrum that is like that of a narrow-line Seyfert 1. S4 1413+34, which does not have an optical identification, shows a core-jet structure at 5 GHz. The radio morphology (see Figure 9 of Dallacasa et al. 2013) shows a well-collimated jet, but the radio emission is dominated by the brightest compact component.

Abdo et al. (2010d) report on 11 misaligned AGNs detected at γ-ray energies, which, with the exception of NGC 1275, show no evidence for variability. Abdo et al. (2010d) suggested that sources with a larger CD, which are observed with a line of sight closer to the jet axis, are those preferentially detected with the Fermi-LAT (in contrast with what was found by Angioni et al. 2020). We evaluated the CD of S4 1413+34 at 5 GHz and obtained a value of CD = 0.28, which is consistent with the values previously found for the other misaligned AGNs. We conclude, therefore, that the observed γ-ray emission should be produced in the compact core region.

5.3.3. NGC 4278

1FLT J1219+2907 was detected in TBIN 7 (2009 March 5 to April 5) of our analysis; using the positional association procedure, we found that NGC 4278 lies within its error ellipse. The SED of NGC 4278 is shown in Figure 14. NGC 4278 (z = 0.0021) is a nearby galaxy identified as a likely radio compact symmetric object (CSO) with an FR I morphology. It shows a dominant, bright, flat-spectrum component identified as the core (see Figures 2M and 3M in Tremblay et al. 2016). From the north and south ends of the core, diffuse jets extend out to the west and east, respectively, creating an "S-shaped" symmetry as observed in other CSO radio sources. We calculated a CD of 0.71 at 366 5 GHz using the data reported in Schilizzi et al. (1983). This is a greater-than-average value of CD for an FR I, although it is reasonable for a γ-ray emitter.

Zoom In Zoom Out Reset image size

Giroletti et al. (2005) reported that the source is oriented at a very small viewing angle (2° < θ < 4°) with respect to the line of sight, although they did not find compelling evidence for strong beaming. They also proposed an alternative scenario in which the source could be oriented at a larger angle, with asymmetries related to the jet's interaction with the surrounding medium. NGC 4278 is optically classified as a low-luminosity and low-accreting AGN with a low-ionization nuclear emission-line region (LINER). All of these characteristics make NGC 4278 a very intriguing new type of γ-ray emitter. Until now, the detection of two CSOs, PKS 1718−649 (Migliori et al. 2016) and PMN J1603−4904 (Krauß et al. 2018), had been reported at γ-ray energies. Stawarz & Petrosian (2008) postulate, in their dynamical radiative model of young radio sources, that the γ-ray emission is produced via inverse Comptonization of circumnuclear (IR to UV) photon fields off relativistic electrons in compact, expanding lobes.

5.3.4. 3C 226

1FLT J0943+0940 is positionally consistent with 3C 226 and was detected in the TBIN 44 (2012 April 5 to May 5) of our analysis. 3C 226, located at z = 0.818 (Hewitt & Burbidge 1991), is an FR II radio galaxy (Laing et al. 1983) with an optical spectrum that shows the presence of narrow emission lines. The FR II radio morphology is clearly seen over a span of 260 kpc (Best et al. 1997). The source has an asymmetric radio structure, having the southeastern lobe closer to the radio core than the northwestern lobe. However, when the radio and optical observations are combined, the source shows a large opening angle. It has a low CD of −2, which makes 3C 266 an unfavored candidate for γ-ray emission. Furthermore, it can be seen from the SED, which is shown in Figure 15, that there is a discrepancy between the archival data and the γ-ray data. FR II radio galaxies are, however, found to be variable from radio to γ-ray energies, e.g., 3C 111 in Grandi et al. (2012), whose variability has been postulated to be due to a new knot component ejected from the core. A similar scenario could be at play in 3C 226, which would strengthen the case for its association with 1FLT J0943+0940.

Zoom In Zoom Out Reset image size

5.3.5. The Highest-redshift FSRQ: PMN J2219−2719

1FLT J2219−2732 is positionally consistent with PMN J2219−2719 and was detected in TBIN 96 (2016 August 04 to 2016 September 04) of our analysis. PMN J2219−2719 is a well-known FSRQ belonging to the Fifth Roma-BZCAT catalog (Massaro et al. 2014) at a redshift of z = 3.634 (Hook et al. 2002). This source has a very low duty cycle (see Section 5.2) of about 2.5%, and it lies at the highest redshift of the entire 1FLT catalog. Examination of the SED (see Figure 16) reveals a large CD_γ of ∼1.8 during a period of flaring activity. This behavior is typical of FSRQs and is usually interpreted as an external Compton process (see Figure 11 of Abdo et al. 2010b). PMN J2219−2719 is also identified as the most distant TS > 25 blazar with monthly variability detected in Kreter et al. (2020). The other source with high redshift reported in Kreter et al. (2020), i.e., 5BZQ J0422−3844, is not detected with our method. The blind search undertaken with PGWave found several monthly seeds that were positionally compatible with this source but were discarded because they were located within 50' of a 4FGL source.

Zoom In Zoom Out Reset image size