REFINING THE ASSOCIATIONS OF THE FERMI LARGE AREA TELESCOPE SOURCE CATALOGS

Despite the large progress in γ-ray source localization made by the Fermi Large Area Telescope (LAT; Atwood et al. 2009), coupled with improvements in our knowledge of the diffuse Galactic γ-ray emission (e.g., Moskalenko et al. 2007; Abdo et al. 2009), the positional uncertainties of the Fermi sources are still large with respect to typical localization precisions at X-ray and lower energies, making a significant fraction of the sources in the γ-ray sky still unknown (Abdo et al. 2010a; Nolan et al. 2012). About 43% of the Fermi sources detected in the Fermi-LAT First Source Catalog (1FGL; Abdo et al. 2010a) were listed as UGSs, while there were ∼33% γ-ray objects unassociated in the Fermi-LAT 2-Year Source Catalog (2FGL, Nolan et al. 2012).

To decrease the number of unidentified γ-ray sources (UGSs), many methods based on multifrequency approaches or statistical analyses have recently been adopted. Radio follow up observations of the Fermi UGSs have already been performed (e.g., Kovalev 2009; Hovatta et al. 2012, 2014; Petrov et al. 2013; Schinzel et al. 2014) and the Swift X-ray survey for all the UGSs listed in the Fermi source catalogs is still on going¹² (e.g., Mirabal & Halpern 2009; Acero et al. 2013; Paggi et al. 2013; Cowperthwaite et al. 2013; Stroh & Falcone 2013; Takeuchi et al. 2013). Additional X-ray observations performed with Chandra and Suzaku have improved our knowledge of UGSs (e.g., Maeda et al. 2011; Cheung et al. 2012; Kataoka et al. 2012; Takahashi et al. 2012). Statistical studies based on γ-ray source properties have also allowed us to recognize the nature of the potential counterparts for UGSs (e.g., Ackermann et al. 2012; Mirabal et al. 2012; Hassan et al. 2013; Doert & Errando 2014). Moreover, a tight connection between the infrared (IR) sky seen by the Wide-Field Infrared Survey Explorer (WISE, Wright et al. 2010) and that of Fermi one has recently been discovered for blazars, the rarest class of active galaxies (Massaro et al. 2011a; DʼAbrusco et al. 2012, 2013). These works greatly decreased the fraction of UGSs with no assigned counterpart at low energies (Massaro et al. 2012a, 2012b, 2013a). More recently, low frequency radio observations (i.e., below ∼1 GHz) also revealed new spectral behavior that has allowed us to search for blazar-like sources lying within the positional uncertainty regions of UGSs (Massaro et al. 2013b; Nori et al. 2014). Additional studies have been also carried out with near-IR observations (Raiteri et al. 2014) as well as in the submillimeter range (e.g., Giommi et al. 2012; López-Caniegoet et al. 2013). Optical spectroscopic campaigns have also been crucial to disentangle and/or confirm the natures of the low-energy counterparts selected with different methods (e.g., Masetti et al. 2013; Shaw et al. 2013a, 2013b; Landoni et al. 2014; Paggi et al. 2014; Massaro et al. 2014b).

The main aim of the analysis presented here is to confirm the current status of the associations for the 1FGL and the 2FGL γ-ray sources. Thus we prepare a single list, created by merging the unique sources in the 1FGL and 2FGL catalogs. On the basis of the information reported in the First Fermi-LAT Catalog of Sources Above 10 GeV (1FHL, Ackermann et al. 2013) and in both the First and the Second Catalogs of AGNs Detected by Fermi (1LAC and 2LAC, respectively, Abdo et al. 2010b; Ackermann et al. 2011a), together with an extensive literature search in the multifrequency archives, we verify the current status of the γ-ray associations. In addition, we also present new optical spectroscopic observations of eight γ-ray blazar candidates (BCN) selected during our multifrequency study. The result is a refined and merged list of all of the associations for the 1FGL plus 2FGL catalog with updated information on the low-energy counterparts, including specific multifrequency notes.

The paper is organized as follows. In Section 2, we discuss the Fermi procedures for assigning identifications and associations with potential counterparts at low energies, and the categories of γ-ray source associations. Section 3 describes the content and main properties of both 1FGL and 2FGL. In Section 4, we introduce an updated γ-ray source classification for the counterparts of the Fermi-detected objects. Details on the correlation with multifrequency databases and catalogs are provided in Section 5. In Section 6, we present the refined version of the associations listed in the Fermi catalogs (hereafter designated 1FGLR and 2FGLR, respectively). The analysis of the IR colors for the sources associated in 1FGL is provided in Section 7 and new optical spectroscopic observations of γ-ray BCN are presented in Section 8. In Section 9, we compare our results with those achieved from statistical analyses performed on UGSs listed in both 1FGL and 2FGL. In Section 10, we speculate on both the radio–γ-ray and the IR–γ-ray connections for the Fermi blazars to check the consistency of sources classified as BCN with these multifrequency behaviors. Our summary and conclusions are provided in Section 11. We use cgs units unless stated otherwise. Spectral indices, α, are defined by flux density, ${{S}_{v}}\propto {{v}^{\alpha }}$ and WISE magnitudes at [3.4], [4.6], [12], [22] μm (i.e., the nominal bands) are in the Vega system. For numerical results, a flat cosmology was assumed with H₀ = 67.3 km s⁻¹ Mpc⁻¹, Ω_M = 0.315, and Ω_Λ = 0.685 (Planck Collaboration 2014). The astronomical survey and source class acronyms used in the paper are listed in Table 1.

In all of the Fermi catalogs, there is an important distinction between identification of low-energy counterparts for the Fermi sources and association. Identification is based on (i) spin or orbital periodicity (e.g., pulsars, binary systems), (ii) correlated variability at other wavelengths (e.g., blazars, active galaxies), or on (iii) the consistency between the measured angular sizes in γ-ray and at lower energies (e.g., supernova remnants, SNRs). On the other hand, the association designation indeed depends on the results of different procedures adopted in both Fermi catalogs (Abdo et al. 2010a; Nolan et al. 2012). These procedures are as follow.

• 1.  
  The Bayesian association method: initially applied to associate EGRET sources with flat-spectrum radio sources (e.g., Mattox et al. 1997, 2001; Abdo et al. 2010a), this method assesses the probability of association between a γ-ray source and a candidate counterpart taking into account their local densities. Local density is estimated simply by counting candidates in a nearby region of the sky.
• 2.  
  The likelihood ratio method: used to search for possible counterparts in uniform surveys in the radio and in X-ray bands, this procedure was originally proposed by Richter (1975) and subsequently applied by and modified by de Ruiter et al. (1977), Prestage & Peacock (1983), Wolstencroft et al. (1986), and Sutherland & Saunders (1992).
• 3.  
  The log N-log S association method: this is a modified version of the Bayesian method for blazars taking into account their log N-log S (see Abdo et al. 2010b; Ackermann et al. 2011a, for more details).

There are several differences between the source associations presented between 1FGL and 2FGL mostly related to the γ-ray analysis and to the improvements achieved in the models of Galactic γ-ray diffuse emission as well as in the evaluation of the LAT response functions. Improving the Fermi source localization not only makes the search for low-energy counterparts easier but also allows for more accurate estimates of the association probability since they depend on the γ-ray positional uncertainty (see, e.g., Abdo et al. 2010b; Ackermann et al. 2011a). These differences also relate to updates of comparison surveys and catalogs used to search for low-energy counterparts and on multifrequency observations performed on ugs samples.

The use of catalogs/surveys that were previously not considered for the Fermi catalog preparation, as for example the IR all-sky survey performed by WISE and the low-frequency radio catalogs, revealing new connections between the low and the high-energy skies (e.g., Massaro et al. 2011a; DʼAbrusco et al. 2013; Massaro et al. 2013b) has also decreased the unknown fraction of the γ-ray sky (e.g., Massaro et al. 2012b, 2013a). This affects the Fermi associations because the methods adopted to assign low-energy counterparts depend on the source densities of the catalogs (Ackermann et al. 2011a; Nolan et al. 2012). A better estimate of the counterpart density leads to a better estimate of the false positive associations and of the association probability and this implies that a previously unassociated source could have an assigned counterpart in a new release of the Fermi catalog.

Motivated by the changes occurring between the 1FGL and the 2FGL associations, here we propose to introduce, together with identified and associated sources, the category of "candidate associations." These are γ-ray sources that have a potential low-energy counterpart in a specific class of well-known γ-ray emitters lying within the Fermi positional uncertainty region at a 95% level of confidence and/or having angular separations between the Fermi and counterpart positions smaller than the maximum for all of the associated sources of the same class. Candidate associations have the potential to be promoted to associations in the future (see Section 4).

We considered only a few kinds of sources as counterpart classes for the candidate associations: (i) blazar-like sources, as defined in the following, (ii) known pulsars (PSRs), (iii) pulsar wind nebulae (PWNs), SNRs, and star-forming and H ii regions (hereinafter simply indicated as SFRs). The first two source classes constitute the two largest populations of known γ-ray emitters while the other choices were considered because 15 out of the 44 SNRs and 10 out of the 63 PSRs and PWNs associated in 1FGL lie within SFRs and a similar situation occurs for 2FGL.

Our choice concerning SFRs, especially for the most massive cases, is indeed motivated by the possibility that they could host type O and/or B stars, potentially associated with open star clusters, which are relatively likely to have some members that have progressed to become a PSR, PWN, and/or SNR, that are well known classes of γ-ray sources (see also Montmerle 1979, 2009).

In addition, considering SFRs as candidate associations is also motivated by the recent idea, partially supported by the Fermi discovery of γ-ray emission arising from Eta Carinae (Abdo et al. 2010c), that shells detected in the radio, IR, and X-ray could mark the termination shocks of stellar winds. These winds interacting with the stars' natal molecular clouds can be considered as plausible sites for the acceleration of particles that emit γ-rays (e.g., Araudo et al. 2007; Bosch-Ramon et al. 2010; Rowell et al. 2010). Bowshocks are associated with many source classes such as PSRs, cataclysmic variables, colliding wind binaries, cometary H ii regions, and particles accelerated therein can radiate up to the MeV–GeV energy range (see e.g., Del Valle & Romero 2012). However, we remark that evidence of γ-ray emission from colliding wind binaries associated with Wolf–Rayet stars was not found by Fermi for a select sample of seven sources (Werner et al. 2013), although, as highlighted by the same authors, searching for such γ-ray detections is extremely complicated due to the uncertainties of the diffuse emission in the Galactic plane.

Some candidate associations lack multifrequency information as well as correct spectroscopic identifications but exhibit some characteristic features, typical of known γ-ray emitters, mainly blazars and PSRs. They were not associated in the previous releases of the Fermi catalogs, mostly because they were not listed in one of the appropriate comparison catalogs of potential counterparts used to associated Fermi sources. However, once follow up observations were performed (see, e.g., Masetti et al. 2013; Paggi et al. 2013; Petrov et al. 2013; Landoni et al. 2014; Massaro et al. 2014b, for blazar-like candidates), they could be promoted to associated sources in new releases of the Fermi catalogs. In addition, within this category we also considered Fermi sources that have SFRs within their γ-ray positional uncertainty regions for which multifrequency observations could reveal the presence of unknown PSRs, PWNs and SNRs, already counterparts of ∼5% of the γ-ray sources.

We emphasize three advantages obtained from introducing the concept of candidate associations.

• 1.  
  Designating a candidate association category allows us to flag potential counterparts that might be associated or identified in future releases of the Fermi catalogs, as previously occurred for 171 1FGL recognized sources, classified and associated in 2FGL. This also facilitates planning follow up multifrequency observations, in particular, spectroscopic observations, necessary to determine and/or confirm their nature.
• 2.  
  The category of candidate associations permits us to establish the number of Fermi sources not yet having potential counterparts. Precise knowledge of this number is extremely relevant for setting better constraints on dark matter scenarios (e.g., Mirabal et al. 2012; Zechlin et al. 2012; Mirabal 2013a, 2013b; Berlin & Hooper 2014).
• 3.  
  For all of the methods adopted to associate sources in the Fermi catalogs, the resulting associations depend on the comparison catalog/survey used. The same low-energy counterpart may be associated with a Fermi source using a certain catalog but not in a different survey, even when the positions in the two catalogs are the same. This occurs because the source density of the comparison catalogs, which is used to determine the probability of false positive associations, affects the association probability. Thus, observing these candidate associations with follow up campaigns to determine their natures will permit us to add blazars and PSRs to the correct comparison catalogs and associate them in future releases of the Fermi surveys.

To reiterate the importance of having candidate associations, we highlight the recent work of Schinzel et al. (2014) on radio very long baseline interferometry observations of Fermi unassociated sources in 2FGL (see also Petrov et al. 2013). In the final list of high-confidence associations presented in Schinzel et al. (2014), more than 90% of the sources are also detected in major radio surveys such as the National Radio Astronomy Observatory (NRAO) Very Large Array (VLA) Sky Survey Catalog (NVSS; Condon et al. 1998) and the Sydney University Molonglo Sky Survey (SUMSS; Mauch et al. 2003). These counterparts were not associated using the methods adopted in the Fermi catalogs due to the high density of sources at low flux levels in the above surveys. However, thanks to the high-resolution radio follow up observations, which revealed the presence of radio compact cores, it has been possible to use a different selected sample of potential counterparts to find new associations via the likelihood procedure (Petrov et al. 2013; Schinzel et al. 2014). Thus it is extremely important to have targets for follow up observations which, once performed, will allow us to insert the potential counterparts into the correct catalogs used to search γ-ray associations.

Identifications, associations, and candidate associations' of γ-ray sources are labeled in both refined versions of the Fermi catalogs as the category of the γ-ray source association. These are defined as identified (I), associated (A), and candidate association (C) in the tables, while sources for which no potential counterpart was found are assigned to the ugs category (U).

3.1. The Fermi-LAT First Source Catalog

3.2. The Fermi-LAT Second Source Catalog

We adopt the following class definitions for Galactic counterparts of the Fermi sources summarized in Table 2: HMB, GLC, nova (NOV), millisecond pulsar (MSP), PSR, PWN, SNR, SFR, and binary star (bin). The classes that we assign for extragalactic counterparts are BZB, quasar radio loud of flat spectrum type (BZQ), blazar of uncertain type (BZU), radio galaxy (rdg), normal galaxy (gal), starburst galaxy (sbg), and Seyfert galaxy (sey). We note that we used three-letter designators to indicate the class for each counterpart; these are shown in lower case to differentiate them from the acronyms used throughout the paper (see Table 1). We emphasize that the labels used for the extragalactic blazar subclasses are used only if the counterpart corresponds to a source classified in the Roma-BZCAT(Massaro et al. 2009, 2011b). We used version 4.1 of the Roma-BZCAT in our analysis. We also emphasize that the proposed classification depends on the availability of the optical spectroscopic information for the low-energy counterparts.

We also considered BCN as a counterpart class to indicate sources having at least one of the following requirements: (1) being classified as a BL Lac candidate in Roma-BZCAT; (2) belonging to the Combined Radio All-Sky Targeted 8 GHz Survey radio catalog (CRATES; Healey et al. 2007) with a BL Lac-like or a quasar-like optical spectrum available in the literature; (3) radio sources with a blazar-like optical spectrum available in literature (i.e., a published spectrum or a description of the features found in the near-IR–optical band) and with a WISE counterpart.

Finally, we labeled as unclassified counterparts (unc) those sources satisfying at least one of the following criteria: (1) WISE IR colors similar to those of γ-ray blazars selected according to the association methods developed by DʼAbrusco et al. (2013) and/or with the kernel density estimator (KDE) analysis (Massaro et al. 2012a, 2012c, see also Section 7 for a refined analysis); (2) a flat radio spectrum at frequencies below ∼1 GHz (i.e., α < 0.5); (3) a radio source with or without an X-ray counterpart associated with the methods adopted in 1LAC and 2LAC; (4) a radio and/or X-ray source that lies at an angular separation between the Fermi and counterpart positions smaller than the maximum separation for the blazar class (i.e., 04525 corresponding to 2FGL J1746.0+2316). All of the remaining sources are still unidentified (ugs).

The proposed classification scheme has three main advantages.

• 1.  
  To have a homogeneous classification of the blazars listed in our analysis and to avoid misclassifications, we adopted the following criterion. We indicate as blazars only those sources that were previously recognized and classified in Roma-BZCAT. There are indeed several counterparts classified as BZB or BZQ listed in the 1LAC and 2LAC catalogs that do not have the same classification according to Roma-BZCAT or for which no optical spectra were found in the literature to confirm their classification.
• 2.  
  We do not classify a source with a radio and/or X-ray counterpart as an active galaxy of uncertain type (AGU) because these requirements are not strictly sufficient to claim a source as an AGN or as blazar-like, and the lack of multifrequency information does not allow us to determine precisely its nature. Consequently, we introduced the BCN and the unc classes. Once optical spectra and multifrequency observations become available, a refined classification will be provided.
• 3.  
  We introduced the SFR class for cases where the Fermi position is consistent with a known massive SFR for three reasons. First, to highlight the possibility that an unknown SNR, PWN, and/or PSR is embedded therein, for which multifrequency observations are necessary to confirm their presence. Second, because the use of this information could be adopted to refine γ-ray diffuse models in future releases of the Fermi catalogs. In particular, the largest fraction of these SFR potential associations were determined using the recent WISE Catalog of Galactic H ii Regions¹⁵ (Anderson et al. 2014), which was not available during the preparation of the currently published Fermi catalogs. This could also reveal new correspondences between the uncertainties of the diffuse γ-ray background models and regions of interstellar medium in the Galactic plane. Third, this could be useful for future investigations of the environments of Galactic γ-ray sources.

Finally, we note that for the sources indicated as candidate associations and classified as SFRs, SNRs, PWNs, and PSRs, we investigated the radio and IR images available in NVSS, the Westerbork Synthesis Radio Telescope (WSRT) Galactic Plane Compact 327 MHz Source Catalog (WSRTGP; Taylor et al. 1996), the Molonglo Galactic Plane Survey¹⁶ (MGPS; Green et al. 1999) catalogs, and in the WISE archives, searching for signatures of potential γ-ray emitters.

In Figures 1 and 2, we show two cases of γ-ray sources that have extended structures associated with an open star cluster in an SFR found within their Fermi positional uncertainty regions. These are two examples of star clusters where some members could have progressed to become a PSR, PWN, and/or an SNR, and so are plausible counterparts for the γ-ray emission.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Then, in Figures 3 and 4, we report the two best cases of SFRs for which a signature of a shell-like or ring-like extended structures was found within the Fermi positional uncertainty region which could be ascribed as a shock signature potentially responsible for the γ-ray emission.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We searched the following major radio, IR, optical, and X-ray surveys and both the NASA Extragalactic Database (NED)¹⁷ and the SIMBAD Astronomical Database¹⁸ to verify whether multifrequency information can confirm the natures of uncertain counterparts and BCN. In the following, we list all of the catalogs and the surveys searched for our investigation; the abbreviations correspond to the designators used for the multifrequency notes in Table 4.

Below ∼1 GHz radio frequency, we searched the VLA Low-Frequency Sky Survey Discrete Source Catalog (VLSS; Cohen et al. 2007,—V) and the recent revision VLLSr¹⁹ (Lane et al. 2014), both the Westerbork Northern Sky Survey (WENSS; Rengelink et al. 1997,—W) and the Westerbork in the Southern Hemisphere Source Catalog (WISH; De Breuck et al. 2002,—W), the SUMSS(Mauch et al. 2003,—S), the Parkes-MIT-NRAO Surveys (PMN; Wright et al. 1994,—Pm), the Texas Survey of Radio Sources at 365 MHz (TEXAS; Douglas et al. 1996,—T), the Parkes Southern Radio Source catalog (PKS; Wright & Otrupcek 1990,—Pk),and the Low-frequency Radio Catalog of Flat-spectrum Sources (LORCAT; Massaro et al. 2014a,—L).

For the radio counterparts (i.e., above ∼1 GHz), we searched the NRAO VLA Sky Survey (NVSS; Condon et al. 1998,—N), the VLA Faint Images of the Radio Sky at Twenty Centimeters (FIRST; Becker et al. 1995; White et al. 1997,—F), the 87 Green Bank catalog of radio sources (87GB; Gregory & Condon 1991,—87) and the Green Bank 6 cm Radio Source Catalog (GB6; Gregory et al. 1996,—GB), the Australia Telescope 20 GHz Survey (AT20G; Murphy et al. 2010,—A), the CRATES (Healey et al. 2007,—c), and the Australian Long Baseline Array (LBA) Calibrator Survey of southern compact extragalactic radio sources²⁰ (LCS1 Petrov et al. 2011,—lcs).

For the IR, we queried the WISE all-sky survey in the AllWISE Source catalog²¹ (Wright et al. 2010,—w) and 2MASS (Skrutskie et al. 2006,—M) since each WISE source is automatically matched to the closest 2MASS potential counterpart (see Cutri 2012, for details).

Then, we also searched for optical counterparts, with or without spectra available, in the Sloan Digitized Sky Survey Data Release 9 (SDSS DR9; e.g., Ahn et al. 2012,—s) and in the Six-degree-Field Galaxy Redshift Survey (6dFGS; Jones et al. 2004, 2009,—6). We also queried the United States Naval Observatory (USNO)-B1 Catalog (Monet et al. 2003,—U) for optical counterparts of the BCN and the unc sources and reported their magnitudes since they could be useful to plan future follow up observations. We also verified the presence of an ultraviolet (UV) counterpart in the GALEX GR6 archive²² (sources with a UV counterpart are marked with a "g" in the notes of Table 4).

In X-rays, we searched the ROSAT all-sky survey in both the ROSAT Bright Source Catalog (RBSC; Voges et al. 1999,—X) and the ROSAT Faint Source Catalog (RFSC; Voges et al. 2000,—X), as well as the XMM-Newton Slew Survey (XMMSL; Saxton 2008; Warwick et al. 2012,—x), the Deep Swift X-Ray Telescope Point Source Catalog (1XSPS; Evans et al. 2014,—x), the Chandra Source Catalog (CSC; Evans et al. 2010,—x), and the Swift X-ray survey for all the Fermi UGSs (Stroh & Falcone 2013). We used the same designator for the X-ray catalogs of XMM-Newton, Chandra, and Swift. These X-ray observatories perform only pointed observations, and thus there is the chance that the pointed observations present in their archives and related to the field of the Fermi source were not requested as follow up of the γ-ray source but for a different reason. Consequently, the discovery of the X-ray counterpart for an associated and/or identified source could be serendipitous.

To perform the searches over all of the surveys and the catalogs cited above, we considered the 1σ positional uncertainties reported therein, with the only exceptions being the WISE all-sky survey and SDSS DR9. In these two cases, we searched for the closest IR and optical counterparts within a maximum angular separation of 33 and 18 for the AllWISE survey and SDSS DR9, respectively. These values were derived on the basis of the statistical analysis described in DʼAbrusco et al. (2013) and Massaro et al. (2014b), developed following the approach also presented in Maselli et al. (2010) and Stephen et al. (2010).

To verify the presence of SFRs consistent with the positional uncertainty of the Fermi sources, we searched in the following surveys and catalogs: (1) the Sharpless catalog of H ii regions (Sharpless 1959), (2) the H ii regions in the InfraRed Astronomical Satellite (IRAS) point-source catalog (Hughes & MacLeod 1989; Codella et al. 1994), (3) the radio survey of H ii regions at 4.85 GHz (Kuchar & Clark 1997), (4) the new catalog of H ii regions in the Milky Way (Giveon et al. 2005), (5) the radio catalog of Galactic H ii regions from decimeter to millimeter wavelength (Paladini et al. 2003), (6) the SCUBA imaging survey of ultra-compact H ii regions (Thompson et al. 2006), (7) the recent WISE Catalog of Galactic H ii Regions (Anderson et al. 2014), (8) the QUaD Galactic Plane Survey (Culverhouse et al. 2011), (9) the 5 GHz VLA survey of the Galacitc plane (Becker et al. 1994), and (10) WSRTGP (Taylor et al. 1996). We note that for the cross-matching performed between the SFR catalogs and the Fermi sources, we took into account the size of each H ii region in combination with the Fermi uncertainty position.

To check for the presence of PSRs, we used the Australia Telescope National Facility (ATNF) Pulsar Catalog (Manchester et al. 2005), the Catalog of Galactic SNRs (Green 2009), and the census of high-energy observations of Galactic SNRs²³ (Ferrand & Safi-Harb 2012).

Table 4 summarizes all of the notes derived from the multifrequency analysis. In particular, for the PSRs, we report the information shown in the Public List of LAT-detected γ-ray PSR²⁴ and by the Pulsar Search Consortium (PSC) as well as from the literature (e.g., Hobbs et al. 2004; Ray et al. 2012, 2013; Pletsch et al. 2012a, 2012b, 2013; Espinoza et al. 2013; Hanabata et al. 2014). Symbols used for their multifrequency notes are labeled as reported in Table 3 (see Abdo et al. 2013, for more details).

The refined associations for the 1FGL and 2FGL catalogs (hereinafter 1FGLR and 2FGLR, respectively) are presented in Table 4. For each source we report the 1FGL name together with those for 2FGL and the 1FHL as derived from the associations published in those catalogs, the counterpart name, and in 37 cases also an alternative association; the category of γ-ray source association; the class of each counterpart as found in the literature; notes derived from the multifrequency investigation; and we also report whether the counterpart lies inside an SFR, and specifically for the PSRs in a PWN or SNR. In addition, in Table 4, we also report if a PWN or SNR contains a known PSR. The coordinates of each counterpart are also included in Table 4 together with the optical magnitudes in the B and R bands from the USNO-B1 Catalog (Monet et al. 2003). Within the multifrequency notes in Table 4, we also indicate if the spectral energy distribution (SED) of the source is shown in Takeuchi et al. (2013), if a redshift estimate is present in the literature (indicating the reference), and if the radio counterpart has a flat radio spectrum (i.e., designated rf in the notes of Table 4 whenever α < 0.5 in the radio band). In the online version of the table, we also report the IR analysis flag (Wright et al. 2010) for those sources classified as blazar-like with a counterpart in the AllWISE catalog within 33 as well as the column with the confirmed redshifts.

For the counterpart names we used the following priorities: if the source is a known blazar, then we adopted the Roma-BZCAT nomenclature, and for known pulsars we adopted that of the ATNF Pulsar Catalog²⁵ (Manchester et al. 2005). Radio galaxies, Seyfert galaxies, and starburst names were indicated based on the Third Cambridge Catalog of Radio Sources and revised versions (3C and 3CR Edge et al. 1959; Bennett 1962; Spinrad et al. 1985), the New General catalog and the Messier catalog, or using their radio names reported in one of the major surveys (see Section 5). Then, Galactic sources such as SNRs and PWNs have their names in Galactic coordinates, as reported in the Catalog of Galactic SNRs²⁶ (Green 2009), while the most common names were used for GLC and binaries. The counterparts that were associated with SFRs were also labeled with their name in Galactic coordinates. Finally, the remaining uncertain sources were identified by their radio names whenever possible or by their ROSAT name. A handful of counterparts were indicated using the nomenclature of optical catalogs. While adopting these choices in the names, we remark that all of the counterparts can be retrieved from the NASA/IPAC Extragalactic Database (NED) and SIMBAD archives without using their coordinates.

6.1. The Refined Associations of the 1FGL

The 1FGLR lists 111 identified sources, 880 associated, 306 candidate associations, and 154 UGSs. The sky distribution of the UGSs is shown in Figure 5 in comparison with the 1FGL associations, in which there were 66 identified objects, 755 associated, and 630 unidentified. The results of our multifrequency analysis for 1FGLR are summarized in Table 5 where we list all of the sources that are identified, associated, and candidate associations for each class separately. We note that the globular cluster GCl 94 (alias M28) associated with 1FGL J1824.5–2449 contains a known MSP PSR J1824–2452A, while 2MASS–GC01 lies in an SFR.

Zoom In Zoom Out Reset image size

Table 5.  Summary of Categories for the γ-ray Source Associations in the Refined Association List of Both Fermi Catalogs

		1FGLR				2FGLR
Class	ID.	Ass.	Cac.	Tot.	ID.	Ass.	Can.	Tot.
Galactic
bin	0	1	0	1	0	1	0	1
hmb	3	0	1	4	4	0	0	4
glc	0	9	0	9	0	11	0	11
nov	0	0	0	0	1	0	0	1
msp	26	16	8	50	26	17	8	51
psr	53	8	31	92	56	12	30	98
pwn	0	8	2	10	2	6	1	9
snr	3	38	5	46	6	55	5	66
sfr	0	0	54	54	0	0	63	63
Extragalactic
bzb	7	231	1	239	7	270	0	277
bzq	15	271	1	287	18	320	0	338
bzu	2	46	0	48	2	52	0	54
gal	1	5	0	6	2	4	0	6
rdg	0	5	0	5	1	7	0	8
sbg	0	3	0	3	0	3	0	3
sey	0	2	0	2	0	3	0	3
Uncertain
bcn	0	188	25	213	0	242	29	271
unc	1	49	178	228	1	173	146	320
ugs	0	0	0	154	0	0	0	289

Note. Col. (1) Source class. Col. (2) ID.—identified. Col. (3) Ass.—associated. Col. (4) Can.—candidate associations.

Download table as:  ASCIITypeset image

Among the PSRs and the MSPs, there are 21 sources that lie within a known PWN and 4 within SNRs as reported in the notes of Table 4. Twenty PSRs and MSPs out of 39 indicated as candidate associations in both Fermi catalogs have been identified after the release of the 2FGL catalog, as presented in the Public List of LAT-detected γ-ray Pulsars. This strongly supports our introduction of this new category of γ-ray sources. In 5 of the 10 associations of PWNs, a PSR has also been found within the nebula and the same situation occurs for 8 SNRs.

We find that 113 1FGL sources have an SFR within their positional uncertainty regions obtained by combining in quadrature the γ-ray localization at a 95% confidence level with the size of the SFR. Three of these sources are HMBs and one is the globular cluster 2MASS–GC01. Fifty six of these 113 have been tentatively associated (i.e., indicated as candidate associations) with previous UGSs, while the remaining Fermi objects all have PSRs, PWNs, and/or SNRs embedded in gas clouds and/or interacting with the interstellar medium.²⁷ Among the 54 candidate associations with SFRs, 2 contain SNRs, 1 a PWN, and 1 a known PSR. However, the potential associations with SFRs could be used to refine the γ-ray Galactic diffuse emission models and the associations in future Fermi catalogs, as well as lead to the discovery of new PWNs and/or unknown SNRs. None of the MSPs has been found to lie within a known SFR, but this occurs for 21 PSRs, 7 PWNs, and 18 SNRs.

It is worth noting that 1FGL J0503.2+4526, associated according to the 2FGL and 2LAC analyses with an unclassified source, has as an alternative potential counterpart an SNR that includes a PSR. A similar situation occurs for 1FGL J1837.5–0659c and 1FGL J1846.8–0233c, whose γ-ray emission could be ascribed to a PSR in a PWN. In addition, 1FGL J0622.2+3751 has as an alternative association with a PSR.

Within the unclassified candidate associations (i.e., unc), we noted that eight 1FGLR sources are positionally consistent with SFRs. In addition, we also listed in 1FGLR the tentative association of 1FGL J1653.6–0158 (also known as 2FGL J1653.6–0159) with the binary system that includes PSR J1653–0158 (Romani et al. 2014).

Finally, we note that in the extragalactic sky, the SMC (alias NGC 292) is identified with the Fermi source 1FGL J0101.3–7257 and is classified as a gal while the LMC is associated with all five of the remaining Fermi sources listed as normal galaxies.

6.2. The Refined Associations of the 2FGL

In the analysis of 1FGLR, we have already considered the associations for 1099 Fermi sources in common between 1FGL and 2FGL, however, we report the results for all of the refined 2FGL associations (hereinafter 2FGLR) in Table 5. As for 1FGLR, the 2FGLR catalog is dominated by blazar-like sources in the extragalactic sky and by pulsars around the Galactic plane. The 2FGLR lists 126 identified sources as in the original catalog, 1176 associated, 282 candidate associations, and 289 UGSs. The 2FGLR sources are classified as 4 HMB, 11 GLC, 51 MSPs, and 98 PSRs; 21 that lie in a PWN and 6 in an SNR, and 17 associated with an SFR, as indicated in our Table 4. In the case of 2FGLR, among the PSR candidate associations that are not in 1FGLR, only one source has been recently identified according to the Public List of LAT-detected γ-ray Pulsars (see also Abdo et al. 2013). There are also 9 PWNs and 66 sources associated with SNRs, 11 of them with a known PSR lying inside the remnant. We then list 63 Fermi sources as candidate associations with SFRs, with 1 also including an SNR. With respect to 1FGL, there is also a new association with an NOV (e.g., Cheung et al. 2014).

The extragalactic sky includes 277 BZBs, 338 BZQs, 54 BZUs, 3 sey, and 3 sbg. Four Fermi sources are associated with the γ-ray emission arising from the LMC, as occurred for those objects classified as galaxies in 1FGLR, and one with M31 (i.e., the Andromeda Galaxy; Abdo et al. 2010d). In addition, we find 271 BCN, 320 Fermi unclassified sources, and 289 UGSs. In Figure 6, we report the comparison between the sky distributions of all UGSs previously listed in 2FGL and the remaining ones determined in this refined association analysis.

Zoom In Zoom Out Reset image size

As has already been done for the AGUs (Massaro et al. 2012a) and UGSs (e.g., Paggi et al. 2013) listed in 2FGL, we performed a non-parametric analysis of the IR colors for those Fermi sources classified as BCN and unc in the merged Fermi catalog 1FGLR+2FGLR. We used KDE (see, e.g., DʼAbrusco et al. 2009; Laurino et al. 2011 and reference therein) to verify the consistency of their IR colors with the so-called WISE γ-ray Strip (e.g., Massaro et al. 2011a, 2012a, 2013c). The KDE technique allows us to estimate the probability function of a multivariate distribution and does not require any assumption concerning the shape of the "parent" distributions.

Consequently, for all BCN and unc sources with a WISE counterpart, we can provide an estimate of the probability π_KDE that such a source is consistent with the WISE γ-ray Strip. In particular, we differentiated between π_KDE,BZB and π_KDE,BZQ considering the comparison with the IR colors of the BZB and BZQ subclasses, respectively. We considered the Roma-BZCAT blazars listed in both 1FGLR and 2FGLR as training samples to build the WISE γ-ray Strip (Massaro et al. 2011a, 2013a) and to compare the IR colors. We note that the WISE counterparts of the BCN and unc sources with IR analysis flags²⁸ (Wright et al. 2010) have been considered in our analysis and their flags are reported in Table 4 together with the probabilities derived from the KDE analysis.

In Figure 7, the isodensity contours drawn from the KDE density probabilities are plotted for the Fermi-Roma-BZCAT blazars in the [3.4]-[4.6]-[12] μm color–color diagram. The IR colors of the sources listed in the refined association list of the Fermi catalogs classified as BCN and unc are also shown. Values of π_KDE,BZB and π_KDE,BZQ greater than 5% indicate that the source has IR colors consistent with the portion of the WISE γ-ray Strip constituted by BZBs and BZQs, respectively, at 95% confidence level.

Zoom In Zoom Out Reset image size

Our analysis confirms that 419 out of 499 total BCN (262) and unc (237) sources in the refined and merged associations list of the Fermi catalogs are consistent with the WISE γ-ray Strip at a 95% confidence level. For the unc sources this situation occurs for 172 sources out of 237 objects, while for the BCN it occurs in 242 out of 262 cases. We remark that 374 out of the 499 sources analyzed with KDE are associated, 1 is identified, and the remaining 124 are candidate associations. This analysis supports the introduction of the category of candidate associations, since a large fraction of BCN and unc with a WISE counterpart are likely to be blazar-like sources consistent with the WISE γ-ray Strip.

In our analysis, we took into account the correction for Galactic extinction for all the WISE magnitudes according to the Draine (2003) relation. As shown in DʼAbrusco et al. (2013), this correction only marginally affects the [3.4]-[4.6] μm color, mostly at low Galactic latitudes (i.e., 15°).

The total number of Fermi sources listed uniquely in 1FGLR and 2FGLR are 2219. We found spectroscopic information for 177 which not reported in the previous 1FGL and 2FGL catalogs, including 8 observed during our spectroscopic campaign (see below).

We have been able to confirm spectroscopically 117 BL Lacs, 21 with firm redshift estimates and 27 quasars (QSOs); an additional 8 sources appear to have a BL Lac nature, but the lack of good optical spectra in the literature or, if a spectrum is available, a low signal-to-noise ratio did not allow us to verify their natures. Thus, 24 of these 177 sources still have an uncertain classification and an uncertain redshift estimate. In particular, among these sources we found 11 listed as blazars in Ackermann et al. (2011a) plus 1 listed in Healey et al. (2007) with no optical spectra published or described in the literature. These 177 sources are all listed in Table 4 as BCNs with the exception of one radio source that appears to be an RDG; in the table notes, we report their possible classifications and their redshift estimates with a question mark (?) indicating that further investigation is required.

Optical spectroscopic observations of four candidate associations and of three associated with unclassified sources (unc) were performed with the 2.1 m telescope of the Observatorio Astronómico Nacional (OAN) in San Pedro Mártir (México) on the nights between 2014 June 28 and July 2 (UT) as reported in Table 6. The telescope carries a Boller and Chivens spectrograph and a 1024 × 1024 pixel E2V-4240 CCD. The spectrograph was tuned in the 4000–8000 Å range (grating 300 l mm⁻¹) with a resolution of 4.5 Å per pixel, which corresponds to 8 Å (FWHM), and a 25 slit. Data were wavelength calibrated using Copper–Helium–Neon–Argon lamps, while for flux calibration spectrophotometric standard stars were observed twice during every night of the observing run.

In addition, we also observed the γ-ray blazar candidate BZB J2340+8015 associated with 1FGL J2341.6+8015 at the Observatorio Astrofisico Guillermo Haro (OAGH) located at Cananea, Sonora in México. The telescope was equipped with a Boller & Chivens spectrograph with a CCD SITe 1k × 1k, tuned in the 4000–7000 Å range (grating 150 l/mm). We used a slit width of 2 5 with a resolution ∼15 Å (FWHM).

The data reduction for both telescopes and instruments was carried out using the IRAF package developed by the National Optical Astronomy Observatory, including bias subtraction, spectroscopic flat fielding, optimal extraction of the spectra, and interpolation of the wavelength solution. All spectra were reduced and calibrated employing standard techniques in IRAF and our own IDL routines (see also Matheson et al. 2008, for additional details).

These eight sources for which new spectroscopic data are provided in our analysis are listed in Table 4 as γ-ray blazar candidates (i.e., BCN). Five of them belong to 1FGLR and are all classified as BZBs according to our analysis; for three of them, the presence of several spectral features allowed us to determine their redshifts. The log of the spectroscopic observations and the results of our analysis are reported in Table 6, while in Figures 8–11 we show their optical spectra together with their finding charts. The remaining three sources are two quasars corresponding to 2FGL J1848.6+3241 and 2FGL J2021.5+0632 as shown in Figures 12 and 13 at redshifts 0.981 and 0.217, respectively, and 2FGL J2031.0+1938, a BZB with an uncertain redshift due to a unique visible emission line, potentially identified as Mg ii (see Figure 14).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

During our observing nights, we also obtained the spectra for two 2FGL sources, 2FGL J1719.3+1744 and 2FGL J1801.7+4405 (see Figure 24), each with uncertain redshift as reported in Roma-BZCAT (see the figures in the Appendix). We were able to confirm the redshift for 2FGL J1801.7+4405 while 2FGL J1719.3+1744 (alias BZB J1719+1745) is completely featureless. Results for these two additional spectra are also reported in Table 6. We also observed 1FGL J1942.7+1033 (see Figure 24), which is a BZB already classified in the literature (Tsarevsky et al. 2005; Masetti et al. 2013, and Appendix for the figure), and 1FGL J2300.4+3138 for which a series of unidentified absorption features are clearly visible in its optical spectrum. Assuming that these unidentified absorption features are due to Mg ii intervening systems along the line of sight, the source should lie at redshift ≥0.96. This source was in the sample observed by Shaw et al. (2013a), but we cannot confirm their results since we did not find any spectral feature identifiable as C iv (see Figure 15). Finally, we also report the spectrum of BZB J2340+8015 associated with 1FGL J2341.6+8015 observed at OAGH that was originally classified as a BL Lac candidate in Roma-BZCAT at redshift 0.274. We have been able to confirm the BL Lac nature of this source but not its redshift due to its featureless spectrum (see Figure 16). All of our finding charts are from the Digitized Sky Survey.²⁹

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We compared our multifrequency analysis on the candidate associations with the results obtained via statistical analyses based on the Classification Tree (CT) and on the Logistic Regression (LR) procedures performed on 1FGL sources (Ackermann et al. 2012) and using the Random Forest algorithm, named sybil, presented by Mirabal et al. (2012)³⁰ for 2FGL sources.

For the 1FGL sources the analysis executed with both the CT and LR statistical methods assesses the probability of correct classification based on the fitting a model form to the Fermi data. The result is a PSR-like or AGN-like classification, occurring when the γ-ray source properties are likely to be more consistent with those of a pulsar or an active galaxy, respectively; a similar prediction is also made for the 2FGL sources analyzed by the sybil algorithm. Thus, all of these methods provide a classification of the potential counterpart on the basis of the γ-ray properties (e.g., spectral shape, variability, etc) but do not allow us to locate it. Consequently, we can only verify if a candidate association that appears to be a blazar or a pulsar has a corresponding γ-ray source classified as AGN-like or PSR-like by the statistical analyses cited above.

There are 286 type BZB candidate associations in 1FGLR that were analyzed in Ackermann et al. (2012). Within this sample, we found that there is only 1 BZB and 1 HMB, the first statistically classified as AGN-like and the second as PSR-like. Then there are 24 sources classified as BCN, with 23 of them indicated as AGN-like while the last is classified as PSR-like.

Among the Galactic sources, there are the following classifications.

• 1.  
  Seven sources which we classified as potential MSPs: four appear to be PSR-like while three show γ-ray properties more similar to AGNs (i.e., AGN-like).
• 2.  
  Twenty-five out of 29 Galactic sources are classified in our analysis as PSRs and are confirmed as PSR-like by statistical methods; meanwhile, four are indicated as AGN-like.
• 3.  
  Two PWN candidate associations, both resembling PSR-like sources in γ-rays.
• 4.  
  Five sources reported in our Table 4 as SNRs: four are classified as PSR-like objects and one has an uncertain statistical classification (i.e., indicated as conflict in Ackermann et al. 2012).
• 5.  
  Thirty-three out of 47 candidate associations are classified as SFRs are PSR-like while only nine appear to be classifiable as AGN-like by the statistical methods; the remaining five are unclassified.

The comparison between our candidate associations in the Galactic plane and the results of the statistical analyses supports the relevance of searching for the presence of SFRs within the Fermi positional uncertainty region where PSRs, SNRs, and PWNs could be embedded, or where shocked regions could be potential sources of γ-rays that have not yet been confirmed. Finally, in the sample of 170 unc sources classified as candidate associations in our analysis, 125 appear to be AGN-like while 40 are PSR-like; the remaining 5 are conflicts between the LR and the CT methods. In particular, four out of eight unc sources that have an SFR consistent with the Fermi position are classified as PSR-like on the basis of their γ-ray behavior.

In the case of 2FGLR, we investigated 133 candidate associations with the sybil procedure and classified them accordingly. Of these, in the sample of 107 Fermi sources that are classified as AGN-like by sybil, 24 are BCN and 78 are unc type; the remaining objects are 2 SFRs, 3 PSRs, 7 MSPs (3 expected to be PSR-like γ-ray sources), and 5 SFRs all classified as AGN-like.

Finally, we highlight that the results of the statistical analyses are in agreement with the classification proposed in the refined association lists of both Fermi catalogs.

10.1. The Radio $\frac{1}{N}$ γ-ray Connection

Many attempts have been made in the past to test for correlations between the radio and γ-ray emission of AGNs and in particular for blazars (e.g., Padovani et al. 1993; Stecker et al. 1993; Salamon & Stecker 1994; Taylor et al. 2007). This connection was also used before the Fermi era to search for counterparts of the γ-ray sources (Mattox et al. 1997), thus motivating all of the past and present radio follow-up campaigns for the Fermi sources (e.g., Petrov et al. 2013; Schinzel et al. 2014).

However, biases and selection effects have to be taken into account to prove this correlation properly, since it is important to address intrinsic source variability, biases due to redshift dependence (Elvis et al. 1978), the "common distance" bias (Pavlidou et al. 2012), problems related to source misidentifications, and incorrect associations, to name a few. All of these issues can mimic a correlation (Mucke et al. 1997).

Since the launch of Fermi, several investigations have also been performed to search for a definitive answer regarding the existence of the radio–γ-ray connection (Ghirlanda et al. 2010, 2011; Mahony et al. 2010) until it was proved and described accurately in Ackermann et al. (2011b). Recently, we also suggested that a link between the radio and the γ-ray emissions in blazars can be extended well below ∼1 GHz (Massaro et al. 2013b; Nori et al. 2014; Massaro et al. 2013d).

Given the new multifrequency analysis carried out here and the candidate associations, for the BCN class in particular, we illustrate the current status of the radio–γ-ray connection. To this end, we show both the flux–flux and the luminosity–luminosity scatter plots for the blazars in our merged refined list of Fermi associations (see Figures 17 and 18).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

A study of the correlations is outside of the scope of this paper. Our main goal is to verify that sources classified as BCN are in agreement with the behavior of blazars in these parameter spaces. It is worth noting that we divided our sample of blazar-like sources into two subsamples: northern and southern samples. The former includes all of the blazars (i.e., BZB, BZQ, and BCN classes) that have a radio counterpart within the NVSS footprint, while in the latter the subsample includes those in the area covered by SUMSS. We distinguish between these two samples since the radio surveys were carried out at different frequencies.

Finally, we remark that only blazars with a firm redshift estimate as reported in our merged list of refined associations are shown in Figure 18. According to this figure, we could expect most of the BCN sources to be BZBs, although this has to be confirmed with optical spectroscopic observations.

10.2. The IR–γ-Ray Connection

DʼAbrusco et al. (2012) studied the Fermi blazar sample listed in 2FGL and lying in the footprint of the WISE Preliminary survey and reported the discovery of a correlation between their IR and γ-ray spectral indices. This is directly related to their peculiar IR colors and the WISE γ-ray Strip (Massaro et al. 2011a).

Here we report an updated scatterplot for the spectral indices (see Figure 19), as originally presented in DʼAbrusco et al. (2012). The correlation between the spectral shapes in the IR and γ-rays is still present and, as occurred in the case of the radio–γ-ray scatter plot (see Section 10.1), the locations of the BCN sources appear to be more consistent with the BL Lac population. The linear correlation coefficient for the whole data set is 0.64. Finally, we also present the connection between IR and γ-ray fluxes and luminosities in Figure 20, which is again in good agreement with our previous results (DʼAbrusco et al. 2012, 2013).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Two years after the release of the second Fermi source catalog, we present a comprehensive multifrequency investigation of all the γ-ray associations listed in both the 1FGL and 2FGL catalogs.

First, we introduced a new category of γ-ray source associations in addition to identified and associated sources. We label as candidate associations those γ-ray sources having a potential low-energy counterpart of a few specific classes of well-known γ-ray emitters lying within the Fermi positional uncertainty region and/or with angular separations between the Fermi and counterpart positions smaller than the maximum one for all of the associated sources of the same class. Then we provided a new classification scheme for the low-energy counterparts of the Fermi sources based on the multifrequency observations and on the optical spectroscopic information now available. We also presented a cross-matching between the Fermi catalogs and several surveys of SFRs (see Section 5 for details) to highlight the possibility that an unknown SNR, PWN, and/or PSR is embedded therein, and to provide information that could be used to refine models of diffuse Galactic γ-rays for future releases of the Fermi catalogs.

The total number of γ-ray sources considered in our investigation for both 1FGL and 2FGL comprises 2219 unique Fermi objects, all listed in Table 4 with each assigned counterpart and their main multifrequency properties. Overall, in the refined association list of the Fermi catalogs, we found 174 Fermi sources with an SFR consistent with their γ-ray positions. In particular, 60 Fermi objects out of these 174 do not have the "c" flag in at least one of the Fermi names and include 13 identified sources, 17 associated, and 30 candidate associations. Their counterparts are also classified as 4 HMBs; 17 PSRs, among which 3 lie in a PWN and 2 in an SNR; 3 PWNs, 1 with a PSR; 1 RDG; 13 SNRs, 2 with a PSR included; 1 binary star; 3 unclassified sources; and 19 SFRs.

We found spectroscopic information for 177 extragalactic γ-ray sources not reported in the previous version of the Fermi catalogs. We included analyses of 8 new optical spectroscopic observations performed with the 2.1 m telescope of the OAN in San Pedro Mártir and with OAGH to confirm the natures of these blazar-like sources found in our multifrequency investigation. Blazar candidates and counterparts with unknown origin (i.e., unc) were also analyzed with the KDE technique to determine the fraction that have WISE IR colors similar to known γ-ray blazars. We compared the refined associations listed in both the Fermi catalogs with the classification proposed on the basis of statistical investigations (Ackermann et al. 2012; Mirabal et al. 2012) and we found a good agreement with our results.

We note that in the combined list of refined associations for 1FGL and 2FGL, of the 2219 sources, 394 are still UGSs (i.e., ∼18% of the entire sample) with 191 having no γ-ray analysis flags. This clean sample of 191 UGSs appears to have a uniform distribution in the sky with a small excess toward the Galactic plane, as shown in Figure 21. Moreover, we conclude that the fraction of Fermi sources with plausible counterparts, combining identifications, associations, and candidate associations, within their γ-ray positional uncertainty regions, is ∼80% and up to ∼90% when considering sources with no γ-ray analysis flags.

Zoom In Zoom Out Reset image size

Finally, we remark that there are 40 sources (listed in Table 7) within the sample of 191 UGSs that do not have any γ-ray analysis flags that do not have any NVSS and/or SUMSS radio source within their positional uncertainty regions at 95% level of confidence. Their all-sky distribution, separated into those lying in the NVSS footprint (i.e., decl. greater than −40°) and the others observable from the southern hemisphere, is shown in Figure 22. The excess in the subsample of those with decl. less than −40°, clearly visible in Figure 22, could be due to shallower coverage by SUMSS relative to the NVSS catalog. It is worth mentioning that for these 40 UGSs the distributions of the γ-ray spectral index and of the γ-ray fluxes do not show any significant differences from those of BZQs and BZBs (see Figure 23). However, it is intriguing that these UGSs tend to be brighter than the BZB population in γ-rays. It is unlikely that this small ugs subsample is constituted by blazars since, given their γ-ray fluxes S_γ and assuming the typical radio–γ-ray fluxes of the blazar population, the expected radio flux densities should be greater than ∼50 mJy for a large fraction, well above the NVSS and SUMSS flux thresholds. On the other hand, we also conclude that their steep values of γ-ray spectral indices are not compatible with sources emitting via dark matter annihilation (e.g., Belikov et al. 2012; Drlica-Wagner et al. 2014).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We thank the anonymous referee for useful comments that led to improvements in the paper. F.M. and G.T. also thank L. Costamante. This investigation is supported by the NASA grants NNX12AO97G and NNX13AP20G. The work by G.T. is supported by the ASI/INAF contract I/005/12/0. H.A.S. acknowledges partial support from NASA/JPL grants RSA 1369556 and 1369565. H.O.F. was funded by a postdoctoral UNAM grant and is currently granted by a Cátedra CONACyT para Jóvenes Investigadores. V.C. acknowledges funding by CONACyT research grant 151494 (México). We thank the staff at the Observatorio Astronómico Nacional in San Pedro Mártir (México) for all their help during the observation runs. Part of this work is based on archival data, software, or online services provided by the ASI Science Data Center. This research has made use of data obtained from the high-energy Astrophysics Science Archive Research Center (HEASARC) provided by NASAʼs Goddard Space Flight Center; the SIMBAD database operated at CDS, Strasbourg, France; the NASA/IPAC Extragalactic Database (NED) operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. Part of this work is based on the NVSS (NRAO VLA Sky Survey): the National Radio Astronomy Observatory is operated by Associated Universities, Inc., under contract with the National Science Foundation and on the VLA Low-frequency Sky Survey (VLSS). The Molonglo Observatory site manager, Duncan Campbell-Wilson, and the staff, Jeff Webb, Michael White, and John Barry, are responsible for the smooth operation of Molonglo Observatory Synthesis Telescope (MOST) and the day-to-day observing programme of SUMSS. The SUMSS survey is dedicated to Michael Large, whose expertise and vision made the project possible. MOST is operated by the School of Physics with the support of the Australian Research Council and the Science Foundation for Physics within the University of Sydney. This publication makes use of data products from the Wide-field Infrared Survey Explorer, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology, funded by the National Aeronautics and Space Administration. This publication makes use of data products from 2MASS, which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation. Funding for SDSS and SDSS-II has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, the U.S. Department of Energy, the National Aeronautics and Space Administration, the Japanese Monbukagakusho, the Max Planck Society, and the Higher Education Funding Council for England. The SDSS Web Site is http://www.sdss.org/. The SDSS is managed by the Astrophysical Research Consortium for the Participating Institutions. The Participating Institutions are the American Museum of Natural History, Astrophysical Institute Potsdam, University of Basel, University of Cambridge, Case Western Reserve University, University of Chicago, Drexel University, Fermilab, the Institute for Advanced Study, the Japan Participation Group, Johns Hopkins University, the Joint Institute for Nuclear Astrophysics, the Kavli Institute for Particle Astrophysics and Cosmology, the Korean Scientist Group, the Chinese Academy of Sciences (LAMOST), Los Alamos National Laboratory, the Max-Planck-Institute for Astronomy (MPIA), the Max-Planck-Institute for Astrophysics (MPA), New Mexico State University, Ohio State University, University of Pittsburgh, University of Portsmouth, Princeton University, the United States Naval Observatory, and the University of Washington. This research has made use of the USNOFS Image and Catalog Archive operated by the United States Naval Observatory, Flagstaff Station (http://www.nofs.navy.mil/data/fchpix/). The WENSS project was a collaboration between the Netherlands Foundation for Research in Astronomy and the Leiden Observatory. We acknowledge the WENSS team consisting of Ger de Bruyn, Yuan Tang, Roeland Rengelink, George Miley, Huub Rottgering, Malcolm Bremer, Martin Bremer, Wim Brouw, Ernst Raimond, and David Fullagar for the extensive work aimed at producing the WENSS catalog. TOPCAT³¹ (Taylor 2005) for the preparation and manipulation of the tabular data and the images. The Aladin Java applet³² was used to create the finding charts reported in this paper (Bonnarel et al. 2000). It can be started from the CDS (Strasbourg—France), CFA (Harvard—USA), ADAC (Tokyo—Japan), IUCAA (Pune—India), UKADC (Cambridge—UK), or from CADC (Victoria—Canada). We acknowledge the contribution of the JHU Sloan Digital Sky Survey group to the development of this site. Many of its features were inspired by the look and feel of the SkyServer. A special thanks goes to Tamas Budavari for his help with the SDSS–GALEX matching and to Wil O'Mullane for his help with the CASJobs site setup and configuration. We also acknowledge Randy Thompson (MAST) for providing IDL IUEDAC routines and Mark Siebert for providing IDL routines to generate tile JPEG images.

Here we present the spectra of 2FGL J1719.3+1744 and 2FGL J1801.7+4405, both with uncertain redshift, as reported in Roma-BZCAT together with that of 1FGL J1942.7+1033. We confirmed the redshift for 2FGL J1801.7+4405 and the BL Lac classification of 1FGL J1942.7+1033 (Tsarevsky et al. 2005; Masetti et al. 2013) while 2FGL J1719.3+1744 (alias BZBJ1719+1745) has a completely featureless spectrum.