EDP Sciences
Free Access
Issue
A&A
Volume 598, February 2017
Article Number A134
Number of page(s) 20
Section Catalogs and data
DOI https://doi.org/10.1051/0004-6361/201628691
Published online 14 February 2017

© ESO, 2017

1. Introduction

Catalogs of γ-ray sources currently compiled by the Fermi-LAT team are based on γ-ray data only, and their standard detection method is blind with respect to information coming from other wavelengths. This approach is clean and unbiased with respect to any class of potential γ-ray emitters. However, there are populations of astrophysical objects that are now known to emit γ-rays, and the knowledge of their position in the sky can be used to facilitate the detection and identification of new γ-ray sources. Based on this principal, we select a sample of candidates to be used as seeds for a direct search of γ-ray signatures using likelihood analysis with the Fermi Science Tools.

Blazars are the most abundant γ-ray sources in the latest Fermi-LAT 3FGL catalog, being 1147 (660 BL Lacs and 487 flat spectrum radio quasars FSRQ) of the total 3034 (Acero et al. 2015). Even so, one third of the known blazars from 5BZcat1 are not confirmed as γ-ray emitters. Probably many of them are faint γ-ray sources that are hard to identify by automatic search methods only based on Fermi-LAT data. The blazar population has been extensively studied by means of a multifrequency approach considering dedicated databases on radio, microwave, infra-red (IR), optical, ultraviolet (UV), and X-ray, since they are characterized by radiation emission extending along the whole electromagnetic spectrum, up to TeV energies.

A particular family of extreme sources with the synchrotron component peaking at frequencies νpeak larger than 1015 Hz is classified as a high synchrotron peak blazar (HSP, Padovani & Giommi 1995; Abdo et al. 2010a) and is the dominant population associated with extragalactic very high-energy (VHE: E> 0.1 TeV, Rieger et al. 2013) sources in the 2nd Catalog of Hard Fermi-LAT Sources (2FHL, Ackermann et al. 2016b). Therefore, HSPs constitute a key population for the detection of point-like γ-ray sources within Fermi-LAT data.

A large sample of HSP blazars was recently assembled using a multifrequency selection procedure that exploits the unique features of their spectral energy distribution (SED). This sample is known as the 1WHSP catalog (Arsioli et al. 2015) and was built using a primary source-selection based on IR colors (following Massaro et al. 2011), later demanding all potential candidates to have a radio, IR and X-ray counterpart. The sources had to satisfy broadband spectral slope criteria (from radio to X-rays) that were fine-tuned to match the SED of typical HSP blazars. In addition, their multifrequency SEDs were inspected individually using the SED-builder tool2 fitting the synchrotron component with a third degree polynomial to determine the νpeak parameter, only keeping cases with νpeak> 1015 Hz. The catalog name “WHSP” stands for WISE high synchrotron peak blazars, since all sources have an IR counterpart from the WISE mission (Wright et al. 2010), which defines their positions. The 1WHSP catalog includes 992 objects at Galactic latitude | b | > 20°. A total of 299 1WHSPs have a confirmed γ-ray counterpart in 1FGL, 2FGL and 3FGL (Arsioli et al. 2015), but many HSPs with bright synchrotron peak are still not detected/confirmed in the γ-ray band.

Given the importance of finding new HSP blazars, an extension of the 1WHSP sample (the 2WHSP, Chang et al. 2017) has ben assembled. It considers sources located at latitudes as low as | b | = 10° with a total of 1693 sources, 439 of which have counterparts within the error circles from the 3FGL catalog. The 2WHSP sample avoids the selection based on IR colors that was used as a primary step for the 1WHSP catalog. This brings an overall improvement in completeness3, since some HSP blazars were out of the 1WHSP sample owing to the contamination of IR colors by the elliptical-galaxy thermal emission. Compared to the 1WHSP, the 2WHSP sample incorporates extra X-ray catalogs like Einstein IPC, IPC slew and Chandra (Harris et al. 1993; Elvis et al. 1992; Evans et al. 2010) as well as updated versions from 3XMM-DR5 and XMM-slew catalogs (Rosen et al. 2016; Saxton et al. 2008). In addition, Swift-XRT alone performed a series of ~160 new X-ray observation of WHSP sources (enabling us to better estimate synchrotron peak parameters) and an extensive study of X-ray extended sources helped to avoid contamination with spurious objects (more details are given by Chang et al. 2017)4. Since the 2WHSP catalog supersedes the 1WHSP (with improved selection and better estimate of synchrotron peak parameters), from now on we only refer to the 2WHSP sample.

Brightness of the synchrotron peak and detectability by Fermi-LAT The HSP blazars are characterized by hard γ-ray spectrum with average photon index ⟨ Γ ⟩ = 1.85 ± 0.01 (Arsioli et al. 2015; Ackermann et al. 2011, 2015b) favouring their detection in the high-energy band. Therefore, the 2WHSP catalog has collected an unprecedented number of remarkably rare and extreme sources that are expected to emit γ-rays.

In Fig. 1 we plot the distribution of synchrotron peak fluxes (νpeakfνpeak) for the 2WHSP detected5 and undetected γ-ray sources. As seen, most of the bright 2WHSPs with Log(νpeakfνpeak) > −11.2 erg/cm2/s have already been unveiled by Fermi-LAT. The range between −12.4 < Log(νpeakfνpeak) < −11.2 erg/cm2/s where histograms for detected and undetected sources have significant overlap, tells us that there must be a population of undetected 2WHSP blazars that is within the reach of Fermi-LAT; especially when taking into consideration integration time greater than 4 yr, as used to build the 3FGL catalog.

thumbnail Fig. 1

Distribution of Log(νpeakfνpeak) synchrotron peak flux with indigo bars that represent the γ-ray subsample of 439 2WHSP-FGL sources, and light red-bars representing the 1255 γ-ray undetected 2WHSPs. The intersection between detected and undetected distributions suggests there may be numerous 2WHSP sources close to the detection threshold from Fermi-LAT. The 2WHSP catalog lists Log(νpeakfνpeak) values in 0.1 steps, and histogram-bins are centered on those values.

Open with DEXTER

As a first step for testing the efficiency of a dedicated γ-ray sources-search, we performed a series of likelihood analysis on bright HSPs that were not included in previous Fermi-LAT catalogs 1FGL, 2FGL and 3FGL (FGL). It soon became clear that, when considering longer exposure time with improved event reconstruction of Femi-LAT Pass 8, a significant number of faint γ-ray sources could be detected. For the likelihood analysis, we define a subsample of 2WHSPs with a synchrotron component νpeakfνpeak≥ 10-12.1 erg/cm2/s simply for limiting the number of seeds to 400, indeed showing its potential for wider studies with the whole 2WHSP sample.

Our present effort for unveiling new γ-ray sources not only provides targets for future follow-up and variability studies, but also helps us to enhance the current understanding on the nature of the VHE γ-ray background, which probably has a strong contribution from unresolved point-like sources (Ajello et al. 2015). Especially at the E> 10 GeV band, unresolved HSP blazars may have increasing relevance (Giommi & Padovani 2015) as was confirmed by our results described in Sect. 4.

2. Fermi-LAT data reduction

The Fermi-LAT detector (Atwood et al. 2009) is characterized by a point spread function (PSF) which contains 68% of the 1 GeV events within 0.8°. The PSF decreases with energy, following a trend E-0.8 up 10’s GeV, and is roughly constant at 0.1° from there to the highest energies considered in this paper.

Based on the position of a potential γ-ray source, we downloaded Pass 8 processed events from the public Fermi-database6 that includes all photons recorded in a region of interest (ROI) of 25° radius from the candidate’s position, for the whole 7.2 yr of observations (MM/DD/YYYY: 08/04/2008 to 11/04/2015). In our analysis we used the Fermi Science Tools (v10r0p5), performing binned analysis to deal with the long integration time.

A series of quality cuts were applied to the raw data, starting with the selection of events having high probability of being photons (which is done by requiring evclass = 128 and evtype = 3 in the gtselect routine), working with maximum zenith angle-cut of 90°7. Given the fact that HSPs are characterized by hard γ-ray spectrum (with an average photon index ⟨ Γ ⟩ ≈ 1.85) we choose to work at E > 300 MeV, avoiding the need to calculate energy-dispersion correction during the data analysis (which is necessary for E< 300 MeV photons). With the gtmktime routine, we then generate a list of good time intervals (GTIs) to be considered in further analysis. In this step, some given flags ((DATA-QUAL > 0) and (LAT-CONFIG==1)) assure that only events acquired by LAT instrument in normal science data-taking mode are considered. Using the gtbin routine, we generate counts maps (CMAP) and counts cubes (CCUBE), having 500 × 500 and 350 × 350 pixels with 0.1°/pixel, respectively. The CCUBE is a series of CMAPs, each one having photons within a given energy bin, and here we consider 37 logarithmically spaced energy bins along 0.3500 GeV.

As an example, the CMAP (Fig. 2) have green circles corresponding to known 3FGL sources, and a magenta circle to mark the 2WHSP J031423.8+061955 position. As seen, together with our candidate, other faint γ-ray sources may be present but cannot easily be distinguished from the counts map, demanding a dedicated data reduction to test the point-like source hypothesis.

thumbnail Fig. 2

Fermi-LAT γ-ray counts map in the energy range 0.3500 GeV, over 7.2 yr, showing detected γ-ray sources, at the center of the green circles (only as indicative of their 3FGL positions). We highlight the source 2WHSP J031423.8+061955 (center of the magenta circle), which we detected in γ-rays with TS = 69.9. As seen, not all relevant sources are easy to unveil with only the CMAP inspection.

Open with DEXTER

A livetime cube is then generated using the gtltcube routine, holding information about the sky coverage as a function of inclination with respect to the LAT z-axis. An important parameter to set in this step is the cos(theta) which is related to the angle-binning, summing incoming photons from a particular solid angle; here we use 0.025° (following recommendations from the Fermi-LAT team). Later, the source map is created using the gtsrcmaps routine, and take into account the models describing all previously known γ-ray sources and background emission that are within 25° radius from the center candidate. The models that describe point-like and extended sources, as well as the diffuse Galactic and isotropic background are included in a single .xml file. This was built using the script make3FGLxml.py8 and considers the 3FGL catalog for describing the point-like and extended sources known so far, loading parameters from the source file gll-psc-v16.fit. For the diffuse Galactic background content, the high-resolution model from the source file gll-iem-v06.fit was considered, and for the isotropic component we use the model from iso-source-v06.txt. We also considered the latest data and instrument response functions (IRFs) available at the time of work, P8R2 SOURCE V6, and event selection Pass 8 processed Source: front+back.

2.1. Adding γ-ray source candidates

Since our γ-ray candidates are not part of the latest Fermi-LAT catalog 3FGL, they have to be added to the source-input file (.xml) that contains the model parameters and the positions of all known γ-ray sources. In this work, the spectrum of each γ-ray candidate is always modeled as a power law: (1)where E0 is a scale parameter (also known as pivot energy), N0 is the pre-factor (corresponding to the flux density in units ph/cm2/s/MeV at the pivot energy E0), and Γ is the photon spectral index for the energy range under consideration. Usually, the starting guess-values were chosen to be consistent with SEDs from HSP blazars, therefore: E0 = 1000 MeV, N0 = 1.0 × 10-12 ph/cm2/s/MeV, Γ = 1.8. Both Γ and N0 are set as free parameters and further adjusted by the gtlike fitting routine. The source position and the scaling E0 are set as fixed parameters. In the source-input xml file, all sources within 10° from the candidate are set free to vary their spectral fitting parameters, therefore 3FGL models that are based on four years of observation are adjusted, since we now integrate over 7.2 yr of data. This particular choice increases the computational burden of the analysis, but is crucial for adapting the model maps to the extra 3.2 yr of exposure that is considered.

A likelihood analysis is then performed by the gtlike routine, considering all the information from the livetime cube, source maps and source input files, together with the PSF and IRFs, in order to best fit the free parameter from the source input list. Finally it is calculated the test statistic (TS) parameter, which is defined as (Mattox et al. 1996) (2)where L(no source) is the likelihood of observing a certain photon-count for a model without the candidate source (the null hypothesis), and L(source) is the likelihood value for a model with the additional candidate source at the given location. The reported TS values correspond to a full band fitting, which constrains the whole spectral distribution along 0.3500 GeV to vary smoothly with energy and assuming no spectral break. Considering we have a good description of Galactic and extragalactic diffuse components, this is a measure of how strong a source emerges from the background, also assessing the goodness of free parameters fit. A TS ≈ 25 is equivalent to a 45σ detection Abdo et al. (2010b, depending on the strength of the background in the region), and only cases with TS > 25 are considered by the Fermi-LAT team as a positive detection of point-like source. We first run the gtlike with the fitting optimiser mode DRMNFB, which generates an enhanced source-input list with all the free parameters recalculated (the first interaction of the fitting procedure). We again feed the gtlike routine with the enhanced source-input list, and use the fitting optimiser mode NEWMINUIT to generate the final model for all sources in the ROI.

3. New γ-ray detections. Validation and population properties

Here we present the γ-ray detection of 85 2WHSP sources at TS > 25 level; and we extend the γ-ray analysis by considering another 65 2WHSPs with lower significance γ-ray signal, with TS ranging between 10 and 25. In Table A.1 we list relevant information for these 150 sources, including names, positions, redshift, γ-ray model parameters and their associated uncertainties. The new γ-ray sources are named with the acronym 1BIGB for the first version of the Brazil ICRANet Gamma-ray Blazar catalog, with source designation 1BIGB JHHMMSS.s±DDMMSS, and coordinates corresponding to the 2WHSP seed-positions. We also present a few examples of TS maps (Sects. 3.2 and 3.3) both for high and low-significance γ-ray signatures, showing that they all emerge as point-like sources, and should not be taken as spurious signals.

In Sect. 3.4 we test a direct source-search using the 3FGL catalog analysis setup, showing that we could successfully probe faint γ-ray emitters and add complementary γ-ray detections. In Sect. 3.5 we calculate the γ-ray detection efficiency based on the brightness of the synchrotron peak (figure of merit). In Sect. 3.6 we plot the photon spectral index (Γ) distribution for the newly-detected γ-ray sources and compare their spectral properties with the FGL counterparts from 2WHSP sources. We also show the Γ vs. flux (1100 GeV band) so that the improvement, with respect to the flux threshold when considering detections down to TS = 10, can be evaluated. In Sect. 3.7 we comment on flux-fluctuations associated with sources close to the flux-threshold (Eddington bias effect) showing that this effect is not severe in our context.

3.1. High-significance γ-ray sources with TS > 25

In Sect. 1 we show that the natural candidates for our analysis are the bright 2WHSP sources9 that have not yet been detected by Fermi-LAT (in previous 1FGL, 2FGL and 3FGL catalogs), as suggested from Fig. 1. Therefore, by sorting the 2WHSP source based on their synchrotron peak brightness, we considered cases down to Log(νpeakfνpeak) = 12.1 erg/cm2/s, selecting 400 γ-candidates, from which 85 (~21%) have shown high-significance γ-ray signature with TS > 25.

For each source of interest, we inspected a region of 50 radius around it, checking for any previous γ-ray detections or for the presence of bright blazars that could also be potential high-energy counterparts. For this task we made use of the Sky Explorer Tool, available from the ASDC web site (tools.asdc.asi.it), which displays all radio, optical, X-ray, and γ-ray detections for a given ROI. During the preparation of this work, the 2FHL catalog (Ackermann et al. 2016b), which contains only E > 50 GeV detections was released, including six of the sources we were working with. Also Campana et al. (2015), as well as Campana et al. (2016), reported on possible counterparts of photon-clustering detected by Fermi-LAT at E > 10 GeV, which included eight of our detections (two in common with the six 2FHL). We keep them in our Table A.1, indicated with “a” and “b” superscripts, respectively, since they constitute positive detections based on our primary approach, showing the intersection between valuable methods for unveiling new γ-ray sources.

The fact that few 2FHL-sources and Campana-sources are in common with our detections is certainly due to their analysis being based only on γ-rays, applied to E > 50 GeV and E > 10 GeV, respectively. Our energy threshold at E > 300 MeV is much lower and well suited to the way we select our seeds (based on multifrequency information from radio to X-rays, not only on γ-ray data). Therefore, we are able to probe hard γ-ray sources, even if they have low flux at E > 10 GeV (we do not depend on γ-ray photon clustering to identify our seeds). In fact, both approaches are powerful and should be seen as complementary, since they all apply to the goal of enriching our description of the γ-ray sky.

3.2. The TS map and γ-ray spectrum

A TS map consists of a pixel-grid where the existence of a point-like source is tested in each pixel. This is a demanding computational task when exposure time that is longer than few months are considered. Here we study the case of 2WHSP J021631.9+231449, defining a 25 × 25 grid with 0.05°/pixel, and evaluated each grid-bin using likelihood analysis from gttsmap routine. Given the fact that WHSP blazars are expected to be hard spectrum γ-ray sources, we built a TS map that only considers photons with energy larger than 3 GeV10. The cut in photon energy helps not only to save computation time, it also has another important purpose: since the PSF improves with increasing energy, working with high-energy photons help us to determine the TS-peak position with better precision. When building the TS map from Fig. 3, the input model (.xml file) does not contain our γ-candidate, so that the map alone can test the existence of point-like sources (with no previous bias), which may manifest as a TS peak that emerges from the background.

thumbnail Fig. 3

TS map (3500 GeV), having 20 × 20 pixels and built with resolution 0.05°/pixel, integrating over 7 yr of Fermi-LAT observations. The green circle centered on + highlights the position of 2WHSP J021631.9+231449; contour dashed lines correspond, respectively, to the 68%, 95%, and 99% confinement regions (from inner to outer lines) for the γ-ray signature position.

Open with DEXTER

All sources within 7° from the grid center have their corresponding model parameters set as free to adjust for the current analysis. Also, we set as free parameters the Normalization (from the diffuse extragalactic background model) and ConstantValue (from the Galactic background model) to avoid having large TS values that are only due to an overestimated background flux. An overestimated background usually manifest as a smooth distribution of high TS values along the whole grid, therefore it is important to properly scale it in the studied region, and evaluate if the source emerges with high TS values from a low TS (0.0) background. When the background is not well described, it could affect (or mimic) a point-like source detection. Since we are mainly working at high Galactic latitude |b| > 10°, we avoid most of the Galactic diffuse emission (which is strong and highly structured at lower latitudes), preventing spurious detections. As seen in Fig. 3, the 2WHSP source is well within the 68% confinement region for γ-ray signature (Mattox et al. 1996), and TS values at the grid center contrasts with outer regions, ensuring that the observed TS peak is due to a point-like source rather than an overestimated background component.

In Fig. 4 we show the multifrequency SED11 for 2WHSP J021631.9+231449. The γ-ray SED was calculated by dividing the full energy band 0.3500 GeV into 6 bins, equally spaced in a logarithmic scale, to compensate for the lower number-counts when increasing the photon energy. The upper limits (u.l.) are computed only for energy bins with TS < 9, and considering a 95% confidence level on the integrated flux along the whole energy bin. The broadband sensitivity at a certain energy E (thin blue line in Fig. 4) is calculated as the maximum flux of a power law source at the LAT detection threshold, for any spectral index.

thumbnail Fig. 4

SED for 2WHSP J021631.9+231449 adding new γ-ray description in the full energy band 0.3500 GeV. The red line is a fitting for the nonthermal component in the synchrotron peak, the green line is the giant elliptical host galaxy template for z = 0.288, the blue line corresponds to the Fermi-LAT (7 yr) broadband sensitivity, and blue dashed line to CTA-North (50 h exposure, Bernlöhr et al. 2013).

Open with DEXTER

The source 2WHSP J021631.9+231449 (Γ = 1.97 ± 0.12) is clearly a promising candidate for observation in the VHE domain with the future Cherenkov Telescope Array (CTA, Actis et al. 2011), or even in reach for present detectors during flaring events. This new γ-ray detection is only one example within the many cases listed in Table A.1, raising our expectations for future VHE studies. Currently, we do not have TS maps and γ-ray SEDs calculated for all sources12. However, we plan to make them available in the near future as a natural extensions of the present work, given the importance of HSP blazars for upcoming observations with CTA.

3.3. Lower significance γ-ray sources

Within the 400 2WHSPs studied, 65 had faint signatures of γ-ray emission with TS values ranging between 10 and 25, and are also listed in Table A.1. We call these lower-significance detections because these sources have TS < 25 (the threshold-limit assumed by the Fermi-LAT team) but they still represent relevant findings considering the number of seed-positions used in our present approach.

As known, the significance σ can be approximated as (Mattox et al. 1996) and, working with TS = 10 threshold, implies our detections have significance of the order of ~3σ. Since we performed a series of 400 binned likelihood analyses for positions only associated with WHSP sources, the number of spurious detections (Nspur) expected is Nspur ≈ 400 × 10-3 = 0.40; therefore we do not consider spurious detections as a concern in our work. In fact, we have individuated a total of 150 γ-ray excess within TS > 10 level, which corresponds to ~37% positive signatures for all the 400 candidates tested.

It should be clear that we are not performing a blind-analysis of the whole γ-ray sky, therefore the number of seed-positions we inspect is very small. The strong threshold cut (TS > 25) adopted by the Fermi-LAT team manifest their rigour before validating any new populations of γ-ray emitters. In contrast, our candidates are a particularly small population of well-characterized (from radio to X-rays) blazars which are firmly established as a family of γ-ray emitters, therefore a 3σ detections threshold is suitable for our approach.

thumbnail Fig. 5

TS maps in the 0.6500 GeV band for six sources representing the lower-significance detections with TS between 10 to 25. At the bottom of each map, we write the corresponding source name and the reported TS value for a binned likelihood analysis when integration is over 7.2 yr of observations (along the full energy band 0.3500 GeV). The 2WHSP positions are highlighted by thick green circles with their centered on +. The contour black dashed lines are TS surfaces representing 68%, 95%, and 99% containment region for the γ-ray signature (from inner to outer lines).

Open with DEXTER

As a complementary test to evaluate if our lower-significance detections are consistent with point-like sources, we randomly choose six cases with TS in the range from 10 to 25 and calculate their corresponding TS maps (Fig. 5). Since lower-significance sources are hard to detect based only on >3 GeV photons, to improve the photon counts we go to lower energies (0.6500 GeV), which helps to individuate the γ-ray signatures13. All six candidates studied clearly emerge from the background as point sources, and are consistent with the 2WHSP positions within the 68% confident radius for the γ-ray signature. In fact, this reinforces our view that assuming a TS > 10 threshold does not contaminate our results with spurious detections.

Given the variability of one order of magnitude is often observed for HSP blazars in the GeVTeV bands14, most of these lower-significance detections may be in reach of CTA during flaring episodes (see Fig. 4). In fact, the validation of lower-significance detections associated with HSP blazars provide relevant hints about the population responsible for considerable portion of the high-energy isotropic γ-ray background (IGRB); and it is also important to account for their existence and imprints, since they add anisotropic contribution to the IGRB (Malyshev & Hogg 2011; Cuoco et al. 2012; Ackermann et al. 2012; Inoue 2014).

knowledge about the position and model-parameter for describing individual faint γ-ray sources may also help to improve tentative correlations of the IGRB with the large-scale structures/clusters (Ando et al. 2014; Prokhorov & Churazov 2014), since their contribution could be subtracted from the currently unresolved background (i.e., to clean the IGRB from faint point-like sources before trying any sort of correlation). By relying on multifrequency data to search for faint γ-ray source, we may improve our capability of resolving them. Moreover, since the present evaluation is primary driven by the position of HSP blazars, it is important to keep track of any case that shows a faint γ-ray signature even if not enough to fit in the current TS-limit for detection demanded by Fermi-LAT team.

3.4. A direct source-search as a complementary approach to probe faint γ-ray emitters

To evaluate the potential of using direct source-search as a complementary method when building γ-ray catalogs, we select 30 objects with the highest significance γ-ray signatures from our list of new-detections (all sources having TS > 45 in our analysis with Pass 8 data integrating over 7.2 yr). We then test if these sources could be detectable with high/lower-significance based on the γ-ray analysis setup used to build the 3FGL catalog. We download Pass 7 data15 corresponding to the first four years of observations (MM/DD/YYYY: 08/04/2008 to 08/04/2012) and proceed with the likelihood analysis that considers a background of extended and point-like sources built based on the gll-psc-v14.fit list, with information available at that time. For the diffuse Galactic background content we consider the source file gll-iem-v05-rev1.fit, and the iso-source-v05.txt model for the isotropic component. We also choose the IRF corresponding to the preparation of 3FGL catalog P7REP SOURCE V15, and event selection Pass 7 reprocessed source data (front+back).

The results are listed in Table A.2, showing four high-significance detections at TS > 25 level, and 17 lower-significance cases with TS in between 10 to 25. Indeed, our test shows that a direct-source search can be used as a complementary method to refine the description of the γ-ray sky; not only revealing high-significance sources, but also allowing lower-significance sources to be successfully probed. The fact that we only present four extra sources that could fit the 3FGL detection threshold should not mislead us into thinking that these types of contributions are not worth to incorporate. As discussed, this test considers only a few sources, and an extended study over the whole blazar population could add a significant complementary contribution. Also, with increasing integration time from Fermi-LAT data, we reach a lower flux threshold (S), and the source number-count may improve ~ S-1.5; therefore the impact of a direct source-search probably has increasing relevance to the building of the next γ-ray catalogs.

Clearly, most of our new high-significance detections based on 7.2 yr with Pass 8 data are mainly driven by a longer integration time (from four years, enhanced to 7.2 yr) and improved event reconstruction (from Pass 7 to Pass 8). However, all 150 detections presented in our work (Table A.1) were only possible because we knew where to look, using selected seed positions selected when considering multifrequency data. In this regard, multifrequency selected seeds are indeed very promising for driving new γ-ray detections just as γ-ray seeds are, and here we emphasise their complementarity. Also, a likelihood analysis based on seeds selected from populations of γ-ray emitter enabled us to successfully probe a population of lower-significance emitters (using Pass 74 yr) that are later confirmed with TS > 45 when working with 7.2 yr of Pass 8 data.

3.5. Detection efficiency according to FOM parameter

The figure of merit (FOM) parameter (Arsioli et al. 2015) is defined as the ratio between the synchrotron peak flux νpeakfνpeak of a given source and that of the faintest 1WHSP blazar already detected in the TeV band ( νpeakfνpeak= 10-11.3 erg/cm2/s); FOM = νpeakfνpeak / 10-11.3. The FOM then provides an objective way of assessing the likelihood for GeVTeV detection of HSP blazars, based on the synchrotron peak brightness, and is not affected by absorption of VHE photons owing to the interaction with extragalactic background light (EBL, Franceschini et al. 2008).

thumbnail Fig. 6

γ-ray detection efficiency for each bin in FOM. Red represents the 2WHSP-FGL subsample (2WHSP sources with FGL counterparts: 439 objects), and blue the 2WHSP* subsample (2WHSP sources considering all 150 new + 439 FGL γ-ray counterparts). The first bin at FOM = 1.2 condensate all cases with FOM > 1.2 (sources with the brightest synchrotron peak flux 11.2 < Log(νpeakfνpeak) <9.7) since almost all of them have been already γ-ray detected, having a 3FGL counterpart.

Open with DEXTER

Figure 6 illustrates this concept, showing the fraction of 2WHSPs already detected in γ-ray according to different FOM bins, considering the subsample of 439 2WHSP-FGL sources (in red), and the subsample of 589 2WHSP*16, which represents all γ-ray detections (in blue). Clearly, the detection efficiency increases with increasing FOM, and there is a considerable increment in the fraction of sources detected for each FOM-bin when accounting for the 150 sources listed in Table A.1. Therefore, the 2WHSP sample shows its potential for unveiling high/lower-significance γ-ray sources, emphasising the power of considering multifrequency information to select VHE γ-ray targets for CTAs, as discussed in Arsioli et al. (2015) and Chang et al. (2017). In addition, given that γ-ray detected HSP blazars have been suggested as counterparts of IceCube astrophysical neutrinos (Padovani et al. 2016), our present work may contribute to discussions in the realm of multi-messenger astrophysics, especially when studying cross-correlations between extreme γ-ray blazars and astro-particles.

3.6. The γ-ray spectral properties of 2WHSP blazars

In Fig. 7 we present the photon spectral index (Γ) distribution for the 150 new γ-ray excess signals (indigo), and compare it with the Γ distribution for the 439 2WHSP-FGL sources (red continuous line). The histogram is normalized with respect to the size of each subsample, so we can visualise their distribution-shape more accurately. A Kolmogorov-Smirnov (KS) test comparing both histograms gives a pvalue = 0.991, meaning the distributions are fully consistent with the same parent population and have similar γ-ray distribution properties. Also, the mean photon spectral index only associated with the 150 new γ-ray sources is ⟨ Γ ⟩ new = 1.94 ± 0.03 in good agreement with that calculated for the 2WHSP-FGL sample ⟨ Γ ⟩ 2WHSP−FGL = 1.89 ± 0.01. Considering all γ-ray detections together (the 2WHSP* subsample) we have ⟨ Γ ⟩ 2WHSP = 1.90 ± 0.01.

thumbnail Fig. 7

Distribution of photon spectral index Γ for the 439 2WHSP-FGL sources (red-continuous line). For the new γ-ray signatures we have: solid indigo bars considers all the 150 detections at TS > 10, red dashed line only for the 85 detections at TS > 25 level, and blue dashed line only for the 65 lower-significance detections with TS in between 10 to 25.

Open with DEXTER

When comparing the photon spectral index distribution of 2WHSP-FGL with the subsample only having our 85 new γ-ray detections at TS > 25; also with the one only having our 65 lower-significance γ-ray detections, the p-values are respectively: and . Therefore, since all the cases we compared showed pvalue> 0.05, we should not reject the hypothesis that all distributions are similar, consistent with a single-parent population.

The Γ vs. S1−100 GeV plot (Fig. 8) shows how we went into lower flux-limit (blue dashed line) compared to previous γ-ray catalogs (red dashed line). This improvement is a combination of many elements: our dedicated search for γ-ray counterparts based on WHSP positions, the larger exposure time used (since we integrate over 7.2 yr of observations), better events reconstruction (from Pass7 to Pass 8), and also the fact that we consider sources down to TS > 10.

thumbnail Fig. 8

Photon spectral index Γ plotted against total flux S1−100 GeV for the 439 2WHSP-FGL sources (in red), the 85 new detections with TS > 25 (filled-in blue), and for the 65 lower-significance detections with TS between 10 and 25 (blue outlines). The dashed lines represent the flux limit achieved by 3FGL-4 yr (red) and by our direct search based on 7.2 yr of data (blue) down to TS = 10.

Open with DEXTER

The overlapping between red and blue dots (Fig. 8) in the range 1−4 × 10-10 ph/cm2/s illustrates how we improved completeness for our HSP γ-ray sample, when considering the new detections presented in Table A.1. We note that the γ-ray threshold sensitivity for HSP blazars has little dependence on the photon spectral index down to ph/cm2/s, so that sub-samples with flux-limit > have low bias arising from Γ. On the other hand, the threshold dependence on Γ is much stronger when considering the integrated flux along the whole band 0.1100 GeV, as reported in Nolan et al. (2012) and Acero et al. (2015). Therefore, the discussion in Sect. 4 considers the 1100 GeV energy range17 for the flux distribution histogram (Fig. 9) and also for the γ-ray LogN-LogS studies (Figs. 11 and 12).

If we plot the histogram of γ-ray flux for the 2WHSP-FGL subsample, comparing it to our 150 γ-ray detections with TS > 10 (Fig. 9), we see that our sources dominate the faint-end. A KS test comparing both histograms gives a p-value of 0.062, which is relatively low, almost excluding the hypothesis that the histograms are similar with respect to the flux distribution. In fact, it shows that our new γ-ray detections (Table A.1) are part of a population of faint sources that was not probed before, and represents a contribution to the IGRB that was previously unresolved.

thumbnail Fig. 9

Histogram comparing the flux distribution S1−100 GeV for the γ-ray subsamples of 439 2WHSP-FGL (red line), and the 150 newly detected WHSPs at TS > 10 level (indigo box).

Open with DEXTER

As is clear from Fig. 9, there is a region in the S1−100 GeV histogram where our new detections overlap the 2WHSP-FGL sources. For fluxes lower than ~2.7 × 10-10 ph/cm2/s the detection efficiency from the 3FGL catalog drops considerably (also shown by a sharp cut in differential number counts dN/dS, Fig. 12 in Sect. 4). As discussed in Sect. 3.4, our new detections at the faint-end of S1−100 GeV histogram are mainly driven by longer integration time, improved event reconstruction, and a lower TS threshold; not only, a direct search can bring complementary sources that improve detection efficiency close to the flux-limit. In addition, the Fermi-LAT exposure is not uniform (Acero et al. 2015, see their Fig. 1) and therefore sky-regions inspected with lower exposure (4-yr with Pass 7) now benefit from better sensitivity owing to longer exposure, revealing new sources at the same faintest flux levels probed by the 3FGL setup. Therefore, taking all of this into consideration, we naturally expect an overlap in the S1−100 GeV faint-end.

3.7. Comments on the Eddington bias effect

Ackermann et al. (2016a) has called attention to the statistical fluctuations of photon flux, especially for faint γ-ray sources close to the Fermi-LAT detection limit, which could lead to overestimated flux-values. The statistical fluctuation of sources close to the flux threshold of any sample is known as an Eddington bias (Eddington 1913) and has a direct impact on the number counts (LogN-LogS) or any other study relying upon the measured flux. For the 2FHL catalog it has been shown through simulations (Ackermann et al. 2016a) that the measured fluxes along 50 GeV2 TeV band could be overestimated up to 10× for the faintest sources. However, we should note that this factor also has a strong dependence on the γ-ray spectral properties from individual sources.

Especially, the Γ(50 GeV−2 TeV) distribution for the 2FHL sample ranges from 1.0 to 5.5 (see Fig. 10) with mean value ⟨ Γ ⟩ 2FHL = 3.20 ± 0.08. Naturally, statistical fluctuations on photon flux measurements are more extreme for the steepest γ-ray spectra (Fig. 10, right side).

thumbnail Fig. 10

Histogram comparing the photon spectrum index (Γ) distribution for the γ-ray samples 2FHL (all sources), 2WHSP (all γ-ray detected 2WHSPs down to TS = 10), and 2WHSP(new) (only the 150 new γ-ray detections). Note that this is a qualitative comparison given that Γ parameter for the 2FHL sample is measured in the 50 GeV2 TeV energy band.

Open with DEXTER

In this case, we may be subject to the same effect, however the Γ(0.3−500 GeV) distribution for the 2WHSP sample is well confined to the 1.2 to 2.8 range, with mean value ⟨ Γ ⟩ 2WHSP = 1.90 ± 0.01. Since the mean γ-ray spectrum for the 2WHSP sources is close to flat, the effect of statistical flux fluctuations is much less severe in our sample, and does not compromise the measured-flux.

To estimate the effect of the Eddington bias in the faint-end of our sample we compare the parameter S1−100 GeV from Tables A.1 and A.2 that were calculated for the same sources, but with different flux limits (see Sect. 3.4). In the first case, the likelihood setup is based on 7.2 yr Pass 8 data, and the second one is based on 4 yr Pass 7 data; therefore different flux-thresholds.

From Table A.2 let us assume that the 17 sources with TS in between 1025 are a good representative of our lower-significance detections, for which the measured fluxes S could be overestimated. When analyzing these same 17 sources with an advanced setup of 7.2 yr with Pass 8 (Table A.1) all of them become γ-ray detected with relatively high-significance TS, and their measured fluxes can be considered as the true ones, S, since the flux threshold is now relatively improved.

We then calculate S/S⟩ = 0.79 as an estimate for the order of magnitude of flux fluctuations for 2WHSP sources close to the Fermi-LAT threshold. This is far from the 10× factor that could affect FHL sources, especially the ones with a steep γ-ray spectrum. Clearly, the effect is not representative for our sample and does not compromise further results. Moreover, in systematically overestimating the flux from faint sources, the Eddington bias would manifest as re-steepening in the number counts, which is not observed (see Figs. 11 and 12).

In conclusion, the S/S value tells us that the γ-ray variability associated with HSP blazars probably dominates eventual oscillations of the S1−100 GeV parameter when the two likelihood analysis-setups are compared. Also, it shows that the overlapping between γ-ray subsamples in Fig. 9 is mainly driven by better detection efficiency (from longer exposure time and improved event reconstruction from Pass 8) rather than statistical flux fluctuations.

4. The isotropic γ-ray background: contribution from HSP blazars to the diffuse component

Since we unveil and model a relatively large number of γ-ray emitters down to TS = 10, we try to evaluate quantitatively what is the impact of our approach for resolving the extragalactic γ-ray background (EGB), and isotropic γ-ray background (IGRB) components.

Table 1

Integral contribution from our new γ-ray sources for different energy bins, compared to EGB and IGRB fluxes reported in (Ackermann et al. 2015a).

thumbnail Fig. 11

Measured γ-ray LogN-LogS of 2WHSP sources, plotting the cumulative number counts with integrated fluxes larger than S1−100 GeV, at | b | > 10°. The dashed lines represent a broken power law fit to the 2WHSP* sample, with an early break at Sbreak−1 = 3.5 × 10-9 ph/cm2/s and fitting parameters given in Eq. (3). This plot is not corrected for nonuniform exposure from Fermi-LAT.

Open with DEXTER

thumbnail Fig. 12

Differential number counts (dN/dS) with respect to S1−100 GeV for HSP blazars at | b | > 10°. The dashed and dash-dotted lines represent the derivative for the power law fit from Eq. (3).

Open with DEXTER

Following the discussion from Ackermann et al. (2015a), we refer to EGB as the sum of all resolved and unresolved contributions from individual extragalactic sources (Blazars, misaligned AGNs and Starburst Galaxies) plus the diffuse emission coming from outer Milky Way regions (which could be related to dark-matter annihilation, intergalactic shocks, and γ-ray cascades induced by ultra high-energy cosmic rays). The exact EGB composition is a matter of intense debate18, and it is well known that contributions owing to unresolved sources may build-up a dominant fraction of the diffuse EGB. Therefore, resolving point-like sources translates directly into narrowing the window available for putative truly-diffusive components, especially when constraining the upper-limits of dark-matter annihilation cross-section as discussed by e.g. Ajello et al. (2015) and Fornasa & Sánchez-Conde (2015).

Another term commonly used is isotropic γ-ray background (IGRB), and it is obtained by subtracting the known extragalactic sources from the EGB. Therefore IGRB represents the sum of a true extragalactic diffuse component, plus the contribution of unresolved sources (which mimic and contaminate the diffuse component). According to Di Mauro et al. (2014), Di Mauro (2015) and Giommi & Padovani (2015), unresolved HSPs/BL lacs may be the dominant component of the IGRB at E> 10 GeV and indeed, here we bring evidence of a population composed of faint γ-ray HSP blazars near the detectability threshold from Fermi-LAT, that was previously undetected.

Based on the model for individual sources, we calculate their corresponding fluxes for each energy bin (Ebin are listed in Table 1), and sum over our γ-ray detections at | b | > 10° listed in Table A.1. We then normalize these values multiplying by Asky/ (4π × A| b | > 10°), where Asky = 41252.96 deg2 is the total sky area, A| b | > 10° = 34110.3 deg2 is the sky area out of the Galactic disk, and the factor 4π normalize the integral flux per unit of steradian, [ph/cm2/s/sr], written as Inew−detc. We compared Inew−detc with the IGRB and EGB intensities (IIGRB and IEGB) as reported by Ackermann et al. (2015a), following the same Ebin steps as theirs. Since our source-modeling does not account for broken power law features that may arise from EBL absorption, especially for large redshift sources, we extend our calculations up to 200 GeV only.

Table 1 lists the corresponding IGRB and EGR fractions we resolved, showing that the subsample of previously undetected γ-ray HSP blazars has increasing relevance for the background composition at higher energies. We evaluate separately the impact owing to all new detections at TS > 10 level, and also owing only to the cases reported with TS > 25. This helped us understand what to expect (in terms of ability to solve the IGRB) from dedicated source-searches based on catalogs of potential γ-ray candidates, and also to evaluate the importance of taking into account lower-significance detections from faint γ-ray blazars. As can be seen, their contribution is not negligible, showing an increment of the order of 40% larger %IGRB solved for each energy bin if we compare the subsamples of our new γ-ray sources detected at TS > 25 and TS > 10.

We also report on the LogN-LogS for HSP blazars using all γ-ray information currently available for the 2WHSP sample. Especially, we incorporate a complementary description for the HSP population at lower γ-ray fluxes by considering our 150 new γ-ray detections down to TS = 10 level. We define the 2WHSPγ-ray sample (which encloses all 2WHSP-FGL sources together with our new detections) and compare it to the 2WHSP-FGL.

In Fig. 11 we plot the cumulative number counts N [deg-2] with flux S1−100 GeV larger than the corresponding value on the x-axis. By fitting the γ-ray LogN-LogS with a broken power law (Fig. 11), we consider an Euclidian behavior for the bright-end, and an early break19 at Sbreak−1 = 3.5 × 10-9 ph/cm2/s: (3)As a test of consistency, we calculate the number of sources n predicted by the fitting (Eq. (3)), which have flux in the interval Smin<S<Smax (with Smax = 1.0 × 10-8 and Smin = 3.0 × 10-10 ph/cm2/s) where the LogN-LogS is well described by the broken power law. The number of sources predicted is: , where the parameter A| b | > 10° = 34110.3 deg2 is the sky area at high Galactic latitudes | b | > 10°, resulting in n(fit) ≈ 309.5, which is in very good agreement with the number of γ-ray sources expected from the 2WHSP sample in this same interval, n(2WHSP ∗) = 311.

An important point to mention is that our new detections only add improvements to the LogN-LogS after the second break. The region in between the first and the second break is not affected by the incompleteness of the 2WHSP γ-ray sample. However, even if the LogN-LogS for S<Sbreak−1 deviates from the Euclidian prediction, it is early to argue we are probing evolution of HSP blazars; further studies need to introduce corrections owing to nonuniform exposure from Fermi-LAT.

When plotting the differential number counts dN/dS vs. S1−100 GeV (Fig. 12), the low flux threshold (Sbreak−2) is evidenced. This manifests as a sharp cut in dN/dS at S ~ 2.7 × 10-10 ph/cm2/s, showing that incompleteness becomes severe for both samples (2WHSP-FGL and 2WHSP) at that particular flux level. Figure 12 also clearly presents how we increment the γ-ray sample completeness, specially at the faint-end where the 2WHSP* (blue) detaches from the 2WHSP-FGL (red).

If we assume there is no cut owing to the flux threshold at Sbreak−2, so that the power law (Eq. (3)) is a good description for the number counts when extrapolating to the faint-end (from Smax = 1.0 × 10-8 down to Smin = 6.0 × 10-11 ph/cm2/s)20, it is possible to estimate the integral contribution of HSP blazars to the EGB in the 1100 GeV band, I1−100 GeV: (4)where Asky = 41252.96 deg2, and the factor 4π is for normalizing the total flux per unit of sky-steradian. The total 1100 GeV flux generate by HSP blazars is of the order ph/cm2/s/sr, which represents 8.5% when compared to the total EGB content for the same energy band ph/cm2/s/sr (Ackermann et al. 2015a, their Table 3).

We should note (from Fig. 11) that the fitting presented in Eq. (3) is suitable for the flux range S>Sbreak−2 (2.7 × 10-10 ph/cm2/s), which is well described by the 2WHSP-FGL subsample even without incorporating the 150 new γ-ray detections.

In fact, our new detections mainly address the problem of incompleteness at S1−100 GeV < 2.7 × 10-10 ph/cm2/s, as evidenced from Fig. 12. However, we are now confident of extrapolating Eq. (3) down to Smin = 6.0 × 10-11 ph/cm2/s only because our new detections push to a lower flux threshold. We emphasise that the measured γ-ray LogN-LogS was calculated without corrections for nonuniform Fermi-LAT exposure. Therefore, the total flux estimated when extrapolating the LogN-LogS to lower fluxes should be regarded as a lower bound to the true contribution of HSP blazars in the 1100 GeV band, bearing in mind that HSPs have increasing relevance for the high-energy channels (Table 1).

Other intervening factors to mention, that may add corrections to the LogN-LogS fitting are:

  • The PSF and effective area from Fermi-LAT depends on energy, therefore the true sensitivity-limits rely on the intrinsic source spectrum properties.

  • The data taken mode is turned off during Fermi-LAT passages along the South Atlantic Anomaly inducing 15% sensitivity differences between north and south hemisphere.

  • Another bias is related to the incompleteness introduced by the poor all-sky coverage in X-rays, and probably an extra component owing to limitations imposed by current radio surveys (SUMMS and NVSS) that were used when building the 2WHSP sample. Evolution of the HSP Population could play an important role as well and demands further investigation.

Therefore, a refined representation of the γ-ray LogN-LogS for HSP blazars demands further corrections (see Ackermann et al. 2016a, for a practical example) that needs to incorporate parameters like the Fermi-LAT detection efficiency, sensitivity nonuniformities along the sky, and selection-efficiency of current blazar catalogs.

5. Addressing unassociated sources and confusion

In Sect. 1 we described the selection of ~400 2WHSP sources to search for their γ-ray signatures using the Fermi Science Tools. Before any likelihood analysis takes place, we inspect the region within 60 radius from all candidates, considering multifrequency databases from radio to γ-rays (using the Sky-Explorer Tool at tools.asdc.asi.it).

In this process, we found four cases where the 2WHSP γ-ray candidates were close to one of the 3FGL sources, but outside, or at the border of, 3FGL error-circles. In the following we study these fields in more details by working with energy dependent TS maps, trying to improve the γ-ray signature description and confirm the association.

The cases studied are 3FGL J0536.4-3347, 3FGL J0935.1-1736, 3FGL J0421.6+1950, and 3FGL J1838.5-6006; well representative examples of how a multifrequency approach can lead to refined scientific products, especially for γ-ray confused sources.

In particular we draw attention to 3FGL J0935.1-1736 and 3FGL J0421.6+1950 which are currently unassociated. As known, a large number of 3FGL objects (1058) have no official association to date, despite the fact that many of them have blazars and AGNs as main association-candidates (especially the 541 unassociated γ-ray sources out of the Galactic plane | b | > 10°, Fujinaga et al. 2015; Doert & Errando 2013).

Although a clear picture that accounts for the large fraction of unassociated 3FGL source is yet to be build, there is evidence of sources that are not clearly related to pulsars nor to AGNs (Acero et al. 2013). In this context, any new association may help to clarify the true nature of current unassociated γ-ray source, and therefore we report on those two cases for which we propose new associations.

5.1. 3FGL J0536.4-3347: a case of source confusion

Source 3FGL J0536.4-3347 is one of the unassociated γ-ray detections in the 3FGL catalog. At this sky position the 2FHL catalog (Ackermann et al. 2016b) reports the source 2FHLJ0536.4-3342, which has been associated with 5BZBJ0536-3343 (=2WHSPJ053628.9-334301) with SED shown in Fig. 13. The γ-ray description of this source is very rich, being detected in the 1FGL, 2FGL and 3FGL catalogs (blue/red/green points); pink dots and u.l. correspond to the 2FHL counterpart at E> 50 GeV. For this case, there is a steep+hard component (Fig. 13) in the γ-ray SED, which may be hard to explain as intrinsic emission from a single source.

thumbnail Fig. 13

SED for 2WHSP J053628.9-334301 (5BZBJ0536-3343) with red thin line showing a fitting for the synchrotron component (radio to X-rays), and its corresponding γ-ray spectrum from 3FGL J0536.4-3347 (2FHL J0536.4-3342).

Open with DEXTER

Exploring the sky area around 3FGL J0536.4-3347 with the ASDC error circle tool (Fig. 14) we note that blazar 2WHSP J053628.9-334301 is just outside the γ-ray error ellipse. Also, there is a bright FSRQ (5BZQ J0536-3401) within 15 from the 3FGL source, and it could contribute to the overall γ-ray flux that is observed.

thumbnail Fig. 14

Sky-Explorer view around 3FGL J0536.4-3347 with position indicated by ×, and γ-ray error-circle shown as dotted line. The 2WHSP J053628.9-334301 (top), and 5BZQ J0536-3401 (bottom) are indicated. X-ray and radio detections in the field are represented by blue and red circles, respectively.

Open with DEXTER

A likelihood analysis, assuming two γ-ray sources instead of one (with position corresponding to 5BZQ J0536-3401 and 2WHSP J053628.9-334301), results in a model adjustment with a very large statistical significance for both, as reported in Table 2. Each of the resolved sources is associated with a distinct γ-ray spectral component (one steep and one hard) in agreement with expectations: hard for the 2WHSP source and steep for the BZQ object.

We then consider 2WHSP J053628.9-334301 as part of the 2WHSP-FGL sample (with updated γ-ray parameters), but 5BZQ J0536-3401 does not count as part of the 150 new detections associated with 2WHSP blazars (since this source is not an HSP).

Table 2

Source model parameters from Fermi Science Tools, assuming a power law to describe the γ-ray spectrum within 0.3500 GeV, with N0 given in [ph/cm2/s/MeV].

To validate our modeling, we also calculated TS maps for the region, taking into consideration different energy bands: 3500 GeV (high-energy map), and 700800 MeV (lower-energy map21).

We build the high-energy map so that the hard γ-ray spectrum source dominates, driving the TS-map peak over the 2WHSP J053628.9-334301, left side of Fig. 16. The lower-energy map was build so that the steeper source dominates, driving the TS peak over 5BZQ J05363401, right side of Fig. 16. This approach can be applied for disentangling confused γ-ray components within 1015, just as shown in Fig. 15 where we plot the resolved SED for both sources.

One of the main reasons for source confusion is related to the PSF strong dependence on photon energy. The final position associated to the confused γ-ray sources is misplaced from their real counterparts, since the arrival direction of photons originating from distinct source are competing. In the case of close-by sources with steep/hard components, the improved PSF at high-energies may favour the association with the hard γ-ray spectrum sources (as seen in Fig. 14).

thumbnail Fig. 15

Top panel: multifrequency SED for 2WHSP J053628.9334301; source with hard γ-ray spectrum ΓWHSP = 1.76. Bottom panel: multifrequency SED for 5BZQ J05363401; source with steep γ-ray spectrum ΓBZQ = 2.75 (also showing the FSRQ template (Vanden Berk et al. 2001) as a thin blue-line along the optical to X-ray band). In the GeVTeV band, both plots show the sensitivity curve for Fermi-LAT 7 yr broadband detection, and for CTA-South considering 50h of exposure.

Open with DEXTER

thumbnail Fig. 16

Energy-dependent TS maps indicating three objects in the studied field; 3FGL J0536.4-3347’s position is highlighted with a cyan circle, centered on ×, 2WHSP J053628.9-334301 and 5BZQ J0536-3401 are highlighted with green and magenta circles, centered on + symbol. Contour dashed lines in black (from inner to outer lines) represent the 68%, 95%, and 99% containment region for the γ-ray signature. Left: high-energy TS map (20 × 20, 0.05°/pixel) taking into consideration only 3500 GeV photons; zooming into central TS-peak region, the source 2WHSP J053628.9-334301 (hard γ-ray spectrum, Γ2WHSP = 1.76) is within the 68% containment region for the high-energy γ-ray signature. Right: lower-energy TS map (20 × 20, 0.08°/pixel) considering only 700800 MeV photons. In this case, the TS-peak position is dominated by the 5BZQ J0536-3401 (steep γ-ray spectrum source, ΓBZQ = 2.75) well within the 68% containment region for the lower-energy γ-ray signature.

Open with DEXTER

thumbnail Fig. 17

High-energy TS-maps 3500 GeV with 20 × 20 grid 0.05°/pixel, integrating over 7 yr of data. In both maps the 2WHSP J093430.1-172120 position is highlighted by a thick green circle centered on +; the 3FGL J0935.1-1736 is in the center of the thick cyan circle, and a blazar candidate (possible counterpart for the 3FGL source) is indicated with a thick magenta circle. Left panel: for this map, the 3FGL source is removed from the model input, showing the pure shape of TS distribution in the region. Green dashed line represents the 99% containment region for the γ-ray signature, which is compatible both with 3FGL and blazar candidate positions. The thin contour lines refer to TS surfaces of 40-30-20, only to show how the γ-ray signature is extended, matching the 2WHSP J093430.1-172120. Right panel: residual TS map (built using the 3FGL in the background) showing excess signal consistent with a point-like source. The green dashed lines correspond to the 68%, 95%, and 99% containment region for the γ-ray signature (from inner to outer lines).

Open with DEXTER

The source 2WHSP J053628.9334301 is a promising candidate for observation with Imaging Atmosphere Cherenkov Telescope (IACTs) and probably a future targets for the CTA-South array (Rieger et al. 2013). Therefore, any dedicated γ-ray analysis has strong motivations, especially when modeling the high-energy component of TeV candidates properly. In this case we have combined multifrequency knowledge of potential GeVTeV emitters with information from the TS-maps, showing that higher quality scientific products can be extracted from the currently available data bases.

Other cases that have their γ-ray SED characterized by steep + hard component may help to identify potential cases of confusion. Source confusion between objects with similar photon spectral index seems harder to identify, but a multifrequency study in the vicinity of each γ-ray detection is useful to evaluate the presence of potential γ-ray emitters. In a hypothetical case, where a steep+hard γ-ray spectrum could emerge from a single source (multiple blobs scenario), the present treatment would help to evaluate/rule-out the possibility of source confusion for any candidate under study.

5.2. Solving a case of source confusion for the unassociated source 3FGL J0935.1-1736

The source 3FGL J0935.1-1736 is one of the unassociated 3FGL objects. Here we provide strong evidence for source confusion in γ-rays involving two objects: a blazar candidate (brighter γ-ray source in the field) and 2WHSP J093430.1-172120 (fainter γ-ray source in the field).

In Fig. 17 we study the γ-ray signature in the 3500 GeV band, to revel the TS distribution that is based on improved PSF photons. For the left-grid marked with TS map, we removed the 3FGL source from the input-model, to show the TS distribution without any bias from the 3FGL catalog. As can be seen, the TS peak matches the 3FGL position (within 99% confinement radius, shown as dashed green line), but the γ-ray signature clearly extends towards the 2WHSP source, embracing it with high-significance TSsurfaces> 30.

thumbnail Fig. 18

Sky-Explorer view around NVSS J093514-173658 (blazar candidate). Left side: XRT field with the X-ray detection marked as ×. Right side: UVOT detection indicated as +. The X-ray and UV error-circle are shown with a dotted blue line. The radio-source is marked in red, and its optical counterpart USNOB1.0 (J2000 Ra, Dec: 143.8116°,-17.6163°) is shown in green.

Open with DEXTER

Table 3

Source model parameters from Fermi Science Tools, assuming a power law to describe 3FGL J0935.1-1736 γ-ray spectrum within 0.3500 GeV, with N0 given in ph/cm2/s/MeV.

thumbnail Fig. 19

Sky-Explorer view around 3FGL J0421.6+1950 positions indicated as a red cross. The blue dash-dotted line represents the error circle for the γ-ray detection reported in the 3FGL catalog. As shown, within 15 from the unassociated γ-ray source there is a 2WHSP blazar. X-ray and radio detections in the field are represented by blue and red circles, respectively.

Open with DEXTER

To test if the extended signature is due to an extra γ-ray source in the field, we built a residual map, as shown in Fig. 17, right. It corresponds to a TS map that considers the 3FGL J0935.1-1736 source is in the background (therefore, it is part of the input-model and positioned at the center of the magenta circle). In this case, the TS distribution is a result of excess photons with respect to the modeled background, which includes point sources and the diffuse component. Clearly, the residual map shows our 2WHSP blazar matching the TS peak, well within the 68% containment region for the γ-ray signature.

thumbnail Fig. 20

High-energy 3500 GeV TS-map, integrating 7 yr of data. The contour black dashed lines are TS surfaces representing 68%, 95% and 99% containment region for the γ-ray signature (from inner to outer lines). The 2WHSP J042218.3+195054 position is highlighted by a thick green circle centered on +. The 3FGL J0421.6+1950 position is highlighted by the cyan circle centered on ×, 9.7 away from the high-energy TS peak.

Open with DEXTER

This is clearly a case of source confusion where the 2WHSP is a counterpart for the residual γ-ray signature, and the brighter source is a counterpart of an, as yet, unidentified object. However, when inspecting this region searching information from other wavelengths, we find the radio-source NVSS J093514-173658 (Fig. 18, in red).

Recent measurements with Swift satellite XRT/UVOT show both UV and an X-ray signatures matching the radio source within their error ellipses (Fig. 18). Together with IR, optical, UV and X-ray counterparts we study the multifrequency SED for NVSS J093514-173658 which turns to be a blazar candidate22 with synchrotron-peak parameters νpeak ≈ 1015.0 Hz and νfν = 10-12.0 erg/cm2/s, marked as BZ-cand in Fig. 17, and probable counterpart for the high-significance TS peak.

A likelihood analysis that considers two sources (with positions corresponding to the 2WHSP and NVSS, removing the 3FGL from the field) results in a relatively good adjustment, as shown in Table 3. Although the blazar candidate could be of HSP type, it is not part of the 2WHSP catalog (because the X-ray data was not available at the time of the sample selection) and therefore we do not count it within the 2WHSP-FGL γ-ray subsample; the source 2WHSP J093430.1-172120 is considered a lower-significance γ-ray detection, so we count it within the 150 sources listed in Table A.1.

thumbnail Fig. 21

High-energy 3300 GeV TS map, integrating 4 yr of data (in this case, the highest energy used is 300 GeV following recommendations for the use of Pass 7 data). The contour black dashed lines are TS surfaces representing 68%, 95%, and 99% containment region for the γ-ray signature (from inner to outer lines). The 2WHSP J042218.3+195054 position is highlighted by a thick green circle centered on +, well within the 68% containment, while the 3FGL J0421.6+1950 source (position highlighted by the cyan circle centered on ×) is localized within the 95% containment region.

Open with DEXTER

5.3. Improved position for the unassociated source 3FGL J0421.6+1950

In Fig. 19 we show a chart that includes the 3FGL J0421.6+1950 source (indicated with a red ×) and its corresponding error-ellipse, which is determined over the entire Fermi-LAT energy range. However, it is well known that the detection of lower-energy photons (E< 1 GeV) from point-like sources have large PSF, so that the γ-ray signature can spread along a region of the order of 1°, as shown in the right side of Fig. 16.

Since the unassociated 3FGL J0421.6+1950 is 10 away from a 2WHSP source (see Fig. 19), we tried to better evaluate the γ-ray signature localization by studying the high-energy TS map (E> 3 GeV), which benefits from smaller PSF with respect to lower-energy photons.

For the TS map in Fig. 20, we removed 3FGL J0421.6+1950 from the model-input so that the TS distribution has no bias from previous γ-ray catalogs. The TS map peaks at the position of 2WHSP J042218.3+195054 (thick green circle centered on +) well within the 68% confinement region for the γ-ray signature, which is 9.7 away from the position reported in the 3FGL catalog (thick circle centered on ×). This source is taken as part of the 2WHSP-FGL subsample, using 3FGL parameters to describe it (since there is no γ-ray confusion in this particular case).

Although the 3FGL positions are based on information associated with the full energy band 0.1300 GeV, we attempted to improve the γ-ray signature localization by selecting only high-energy photons that are know to have better PSF. In fact, we also improve the localization owing to longer exposure time (since we now integrate along seven years of Fermi-LAT observations instead of four years in the 3FGL), but it is important to mention that the high-energy maps, integrated over four years of Pass 7 data (with the same analysis setup used to build the 3FGL catalog, as described in Sect. 3.4) already enabled this kind of study, as seen in Fig. 21.

thumbnail Fig. 22

Left panel: error-circle (dash-dotted) associated with the 3FGL source. Right panel: high-energy 3500 GeV TS map, integrating over 7 yr of Fermi-LAT observations. The contour black dashed lines are TS surfaces representing 68%, 95% and 99% containment region for the γ-ray signature (from inner to outer lines). The 2WHSP J183806.7-600032 (thick green circle centered on +). The 3FGL J1838.5-6006 position is shows in cyan (centered on ×), 6.9 away from the high-energy TS peak.

Open with DEXTER

Probably, when using the full energy range from Fermi-LAT (0.1500 GeV), lower-energy photons with the largest position uncertainties could be degrading the final source localization. Indeed, based on currently available data, there is room for improvements which may bring complementary and relevant information for describing the γ-ray sky.

Working with high-energy TS maps proved to be very useful when searching for candidate-counterparts of current unassociated γ-ray sources, and could be applied systematically as a complementary refinement for the building of upcoming catalogs with the potential to improve source localization for the whole γ-ray sample. In the next subsection, we discuss a case where the 3FGL association has already been done, but we still improve the γ-ray localization using high-energy TS maps.

5.4. Improved position for the 3FGL J1838.5-6006 source

Here we study 3FGL J1838.5-6006, which is associated with the radio-source SUMSS J183806-600033 (2WHSP J183806.7-600032). In this case, the 2WHSP blazar is just at the border of the 3FGL error-circle (Fig. 22, left) and therefore we try improving the γ-ray source localization by working only with high-energy photons only. As shown in Fig. 22, the high-energy TS peak is few arcminutes away from the 3FGL position, and matches our 2WHSP source (which is within the 68% confinement radius for the γ-ray signature).

Although the association is correct, this is another example where we could improve the γ-ray signature localization (6.9 drift) just by working with E > 3 GeV photons. We also study this region at a lower-energy band (850950 MeV) and there is no evidence of another close-by source that could be the cause of the offset position. Therefore, it could be that, for some cases the determination of the source position based on the broadband counts 0.1500 GeV is non optimal, probably because of the large PSF associated with lower-energy photons. Building high-energy TS maps may help to improve source positioning, especially for cases with hard γ-ray spectrum, as shown for 3FGL J0421.6+1950 and 3FGL J1838.5-6006.

6. Conclusions and perspectives

The 2WHSP catalog was built to select promising VHE candidates for the present and future generation of Cherenkov Telescope Arrays, therefore we have tested the efficiency of a direct search for γ-ray signatures associated with 2WHSP blazars, achieving significant results.

We have detected 150 γ-ray excess signals out of 400 seed positions based on 2WHSP sources that had no counterpart in previous 1FGL, 2FGL, and 3FGL catalogs. A total of 85 sources were found with high-significance with TS > 25, and we also report on 65 lower-significance detections with TS between 10 to 25. The 150 new γ-ray sources presented in Table A.1 are named with acronym 1BIGB (first version of the Brazil ICRANet Gamma-ray Blazar catalog) which corresponds with the 2WHSP seed-positions used for our likelihood analysis. Clearly, the subsample of 2WHSP blazars that have not yet been detected by Fermi-LAT is a key representative population of faint γ-emitters, and we show how the new detections down to TS > 10 level can probe the faint-end of the flux-distribution (see Figs. 8 and 9). As discussed in Sect. 3.3, a γ-ray source-search based on the seed positions from HSP blazars can be used to unveil faint HE sources down to TS = 10 without compromising the γ-ray sample with spurious detections.

Our current work enabled us to associate a relevant fraction of the IGRB to a population of faint γ-ray emitters that had been previously unresolved. Moreover, we show the increasing relevance of faint-HSPs for the IGRB composition with respect to energy (see Table 1), specially for E > 10 GeV, reaching 68% in the 100200 GeV band. Motivated by this first assessment, we plan to perform a complete γ-ray analysis of the 2WHSP sample, down to the lowest fluxes, and probably extend the search to other blazar families with potential to improve the γ-ray description of lower-significance γ-ray blazars, also helping to constrain the origins of the extragalactic diffuse γ-ray background.

We have worked out the possibility of solving source confusion when considering multifrequency data for identifying potential γ-ray emitters in a certain ROI, and building energy dependent TS maps to help disentangle hard-steep components from confused sources.

We also addressed cases of unassociated 3FGL sources by studying high-energy TS maps to evaluate possible counterparts. This could be a key for solving cases of unassociated γ-ray sources (just as discussed in Sects. 5.2−5.4) showing that we can improve the γ-ray signature localization based on currently available databases. Certainly, it is interesting to evaluate if this kind of approach could be applied systematically as a complementary refinement for the building of upcoming γ-ray catalogs.


1

The 5BZcat (Massaro et al. 2015) is a large sample of 3561 identified blazars. Multifrequency data for the 5BZcat is available at http://www.asdc.asi.it/bzcat with a direct link to the SED-builder tool.

3

Also to improve the completeness of the final sample, known HSP sources at | b | < 10° were incorporate in the 2WHSP catalog.

4

The catalogs are available at: www.asdc.asi.it/1whsp or /2whsp; where multifrequency SEDs can be quickly built using open access online tools.

5

We may use 2WHSP-FGL when referring to the subsample of 439 2WHSPs that have counterparts from the 1FGL, 2FGL, and 3FGL catalogs.

7

The zenith angle-cut is used to avoid contamination with Earth’s limb γ-ray photons, which are induced by cosmic-ray interactions with the atmosphere, and are known as strong source of background for the low-energy band covered by Fermi-LAT.

8

The make3FGLxml.py is a python routine written by Johnson (2015), and provided by the Fermi-LAT team as an user contribution tool.

9

By bright sources we mean: 2WHSPs with the largest flux density associated to the synchrotron peak νfν−peak component.

10

For HSP sources with high-significance γ-ray signature, the cut at 3 GeV in many cases provided (and is therefore the reason why we choose it; see examples on Sect. 5) a good balance between computation time and ability to solve the γ-ray signature as a point-like source. Also, despite the fact we are dealing with HSP blazars (with mean γ-ray photon index ~ 1.9) Fermi-LAT has relatively good sensitivity along the 1100 GeV band, and improved PSF at >3 GeV, which helped to achieve better localization for the γ-ray sources when necessary.

12

The computational demand for accomplishing such task is relatively large, and requires further planning together with our Computer-Cluster partners; to cite: Joshua-Cluster from ICRANet Italy, and Gauss-Cluster from CESUP Brazil.

13

Also, we should add that the overall computation time for lower-energy TS maps integrating over 7 yr can easily became prohibitive (>weeks). Especially for bright sources; large photon-counts translate into large computational demand. Therefore, there is no absolute way to choose a working energy-range. We are always limited by the computation time, and many cases demand us to adapt (for example, see the lower-energy TS map from Sect. 5.1, where we had to work in a narrow energy range of 700800 MeV to reach results in reasonable time).

14

For dedicated studies on variability involving HSP blazars, see Krawczynski et al. (2004) reporting on 1ES 1959+650 2WHSP J195959.8+650853, or Błażejowski et al. (2005), Sahu et al. (2016) reporting on Mrk 421 2WHSP J110427.3+381230.

16

We may use 2WHSP* when referring to the total 589 sources that include: 439 2WHSP-FGL + our 150 γ-ray detections at TS > 10 level.

17

The integral flux for the energy range 1100 GeV is commonly reported in all Fermi-LAT catalogs (1FGL, 2FGL and 3FGL). For practical reasons we work in the same energy range, making it easy to combine flux information from our current list (Table A.1) with the 1100 GeV flux reported in Fermi-LAT catalogs. Moreover, since the 1100 GeV band is covered with relatively good sensitivity by Fermi-LAT, the power law modeling of faint hard-spectrum γ-ray sources is more reliable in this range.

18

Ackermann et al. (2015a) also discuss the challenges for measuring the EGB component, which demands a proper modeling of the diffuse Galactic emission (DGE) especially as a result of cosmic rays interacting with the Milky Way gas and photon fields. The DGE has intensity comparable to the EGB, and to obtain the EGB, both the DGE and the known Galactic sources have to be subtracted from the total-sky γ-ray counts. The reported EGB flux (1.1200 GeV) is IEGB ≈ 4.74 × 10-7 ph/cm2/s/sr.

19

To confirm the presence of this early break in the number counts, we extracted the LogN-LogS data from Acero et al. (2015) paper (their Fig. 29) plotting the cumulative energy flux Senergy distribution for the clean sample of HSP blazars (also uncorrected for nonuniform sensitivity and detection efficiency). Although they do not mention the fitting parameters, we found good agreement with a broken power law that has similar slopes as ours: N(S>Sbreak−1) = 3.2 × 10-19, N(S<Sbreak−1) = 1.8 × 10-15. In this case Sbreak−1 ≈ 3.5 × 10-11 erg/cm2/s, but there is also a strong cut at Sbreak−2 ≈ 7.0 × 10-12 erg/cm2/s. Therefore also manifesting two breaks that probably have the same origin as in our case.

20

We choose the faint-end to be Smin = 6.0 × 10-11 ph/cm2/s since this is consistent with our flux threshold; as can be seen in Fig. 9, there is a sharp cut in the number of γ-ray sources for fluxes lower than that.

21

In the lower-energy map we use lower resolution, to account for the larger PSF with respect to high-energy photons. In this case, the particular lower-energy range was chosen to try to balance between “going to the lowest energies probed by Fermi-LAT” and “still acceptable computation time” of the order of two weeks. A lower energy range could be used, but since it is a bright γ-ray source, photon-counts (and therefore computation time) escalate very rapidly.

22

The term blazar candidate refers to a source with multifrequency SED characteristic of blazars, but missing optical identification (optical spectrum is not available).

Acknowledgments

B.A. is supported by the Brazilian Scientific Program Ciências sem Fronteiras from Cnpq, Y.L.C. is supported by the Government of the Republic of China (Taiwan). We thank ICRANet and Prof. Carlo Bianco for the cooperation that enabled us to perform part of the data reduction at Joshua Cluster, Rome-Italy. We thank the Centro Nacional de Supercomputação (CESUP) Porto Alegre-Brazil, and Carlos Brandt, for the cooperation that enabled us to perform part of the data reduction using CESUP machines. This work was supported by the ASDC, Agenzia Spaziale Italiana Science Data Center; and University La Sapienza of Rome, Department of Physics. We thank the guidance and comments from Prof. Paolo Giommi, and the special attention from Dr. Dario Gasparrini, Dr. Sara Cutini, Dr. Stefano Ciprini and Prof. Toby Brunett. This publication makes use of public data products and software from Fermi-LAT collaboration. We also make use of archival data and bibliographic information obtained from the NASA/IPAC Extragalactic Database (NED), data and software facilities from the ASDC managed by the Italian Space Agency (ASI).

References

Appendix A: Additional tables

Table A.1

The 150 new γ-ray signatures detected down to TS > 10 level are named with the acronym 1BIGB (for the first version of “Brazil ICRANet Gamma-ray Blazar” catalog) with coordinates corresponding to those of 2WHSP seeds used for the likelihood analysis.

Table A.2

List of 30 2WHSPs detected with the largest significance within our γ-ray likelihood analysis, all having TS > 45 based Pass 8 Fermi-LAT data when integrating over 7.2 yr of observations.

All Tables

Table 1

Integral contribution from our new γ-ray sources for different energy bins, compared to EGB and IGRB fluxes reported in (Ackermann et al. 2015a).

Table 2

Source model parameters from Fermi Science Tools, assuming a power law to describe the γ-ray spectrum within 0.3500 GeV, with N0 given in [ph/cm2/s/MeV].

Table 3

Source model parameters from Fermi Science Tools, assuming a power law to describe 3FGL J0935.1-1736 γ-ray spectrum within 0.3500 GeV, with N0 given in ph/cm2/s/MeV.

Table A.1

The 150 new γ-ray signatures detected down to TS > 10 level are named with the acronym 1BIGB (for the first version of “Brazil ICRANet Gamma-ray Blazar” catalog) with coordinates corresponding to those of 2WHSP seeds used for the likelihood analysis.

Table A.2

List of 30 2WHSPs detected with the largest significance within our γ-ray likelihood analysis, all having TS > 45 based Pass 8 Fermi-LAT data when integrating over 7.2 yr of observations.

All Figures

thumbnail Fig. 1

Distribution of Log(νpeakfνpeak) synchrotron peak flux with indigo bars that represent the γ-ray subsample of 439 2WHSP-FGL sources, and light red-bars representing the 1255 γ-ray undetected 2WHSPs. The intersection between detected and undetected distributions suggests there may be numerous 2WHSP sources close to the detection threshold from Fermi-LAT. The 2WHSP catalog lists Log(νpeakfνpeak) values in 0.1 steps, and histogram-bins are centered on those values.

Open with DEXTER
In the text
thumbnail Fig. 2

Fermi-LAT γ-ray counts map in the energy range 0.3500 GeV, over 7.2 yr, showing detected γ-ray sources, at the center of the green circles (only as indicative of their 3FGL positions). We highlight the source 2WHSP J031423.8+061955 (center of the magenta circle), which we detected in γ-rays with TS = 69.9. As seen, not all relevant sources are easy to unveil with only the CMAP inspection.

Open with DEXTER
In the text
thumbnail Fig. 3

TS map (3500 GeV), having 20 × 20 pixels and built with resolution 0.05°/pixel, integrating over 7 yr of Fermi-LAT observations. The green circle centered on + highlights the position of 2WHSP J021631.9+231449; contour dashed lines correspond, respectively, to the 68%, 95%, and 99% confinement regions (from inner to outer lines) for the γ-ray signature position.

Open with DEXTER
In the text
thumbnail Fig. 4

SED for 2WHSP J021631.9+231449 adding new γ-ray description in the full energy band 0.3500 GeV. The red line is a fitting for the nonthermal component in the synchrotron peak, the green line is the giant elliptical host galaxy template for z = 0.288, the blue line corresponds to the Fermi-LAT (7 yr) broadband sensitivity, and blue dashed line to CTA-North (50 h exposure, Bernlöhr et al. 2013).

Open with DEXTER
In the text
thumbnail Fig. 5

TS maps in the 0.6500 GeV band for six sources representing the lower-significance detections with TS between 10 to 25. At the bottom of each map, we write the corresponding source name and the reported TS value for a binned likelihood analysis when integration is over 7.2 yr of observations (along the full energy band 0.3500 GeV). The 2WHSP positions are highlighted by thick green circles with their centered on +. The contour black dashed lines are TS surfaces representing 68%, 95%, and 99% containment region for the γ-ray signature (from inner to outer lines).

Open with DEXTER
In the text
thumbnail Fig. 6

γ-ray detection efficiency for each bin in FOM. Red represents the 2WHSP-FGL subsample (2WHSP sources with FGL counterparts: 439 objects), and blue the 2WHSP* subsample (2WHSP sources considering all 150 new + 439 FGL γ-ray counterparts). The first bin at FOM = 1.2 condensate all cases with FOM > 1.2 (sources with the brightest synchrotron peak flux 11.2 < Log(νpeakfνpeak) <9.7) since almost all of them have been already γ-ray detected, having a 3FGL counterpart.

Open with DEXTER
In the text
thumbnail Fig. 7

Distribution of photon spectral index Γ for the 439 2WHSP-FGL sources (red-continuous line). For the new γ-ray signatures we have: solid indigo bars considers all the 150 detections at TS > 10, red dashed line only for the 85 detections at TS > 25 level, and blue dashed line only for the 65 lower-significance detections with TS in between 10 to 25.

Open with DEXTER
In the text
thumbnail Fig. 8

Photon spectral index Γ plotted against total flux S1−100 GeV for the 439 2WHSP-FGL sources (in red), the 85 new detections with TS > 25 (filled-in blue), and for the 65 lower-significance detections with TS between 10 and 25 (blue outlines). The dashed lines represent the flux limit achieved by 3FGL-4 yr (red) and by our direct search based on 7.2 yr of data (blue) down to TS = 10.

Open with DEXTER
In the text
thumbnail Fig. 9

Histogram comparing the flux distribution S1−100 GeV for the γ-ray subsamples of 439 2WHSP-FGL (red line), and the 150 newly detected WHSPs at TS > 10 level (indigo box).

Open with DEXTER
In the text
thumbnail Fig. 10

Histogram comparing the photon spectrum index (Γ) distribution for the γ-ray samples 2FHL (all sources), 2WHSP (all γ-ray detected 2WHSPs down to TS = 10), and 2WHSP(new) (only the 150 new γ-ray detections). Note that this is a qualitative comparison given that Γ parameter for the 2FHL sample is measured in the 50 GeV2 TeV energy band.

Open with DEXTER
In the text
thumbnail Fig. 11

Measured γ-ray LogN-LogS of 2WHSP sources, plotting the cumulative number counts with integrated fluxes larger than S1−100 GeV, at | b | > 10°. The dashed lines represent a broken power law fit to the 2WHSP* sample, with an early break at Sbreak−1 = 3.5 × 10-9 ph/cm2/s and fitting parameters given in Eq. (3). This plot is not corrected for nonuniform exposure from Fermi-LAT.

Open with DEXTER
In the text
thumbnail Fig. 12

Differential number counts (dN/dS) with respect to S1−100 GeV for HSP blazars at | b | > 10°. The dashed and dash-dotted lines represent the derivative for the power law fit from Eq. (3).

Open with DEXTER
In the text
thumbnail Fig. 13

SED for 2WHSP J053628.9-334301 (5BZBJ0536-3343) with red thin line showing a fitting for the synchrotron component (radio to X-rays), and its corresponding γ-ray spectrum from 3FGL J0536.4-3347 (2FHL J0536.4-3342).

Open with DEXTER
In the text
thumbnail Fig. 14

Sky-Explorer view around 3FGL J0536.4-3347 with position indicated by ×, and γ-ray error-circle shown as dotted line. The 2WHSP J053628.9-334301 (top), and 5BZQ J0536-3401 (bottom) are indicated. X-ray and radio detections in the field are represented by blue and red circles, respectively.

Open with DEXTER
In the text
thumbnail Fig. 15

Top panel: multifrequency SED for 2WHSP J053628.9334301; source with hard γ-ray spectrum ΓWHSP = 1.76. Bottom panel: multifrequency SED for 5BZQ J05363401; source with steep γ-ray spectrum ΓBZQ = 2.75 (also showing the FSRQ template (Vanden Berk et al. 2001) as a thin blue-line along the optical to X-ray band). In the GeVTeV band, both plots show the sensitivity curve for Fermi-LAT 7 yr broadband detection, and for CTA-South considering 50h of exposure.

Open with DEXTER
In the text
thumbnail Fig. 16

Energy-dependent TS maps indicating three objects in the studied field; 3FGL J0536.4-3347’s position is highlighted with a cyan circle, centered on ×, 2WHSP J053628.9-334301 and 5BZQ J0536-3401 are highlighted with green and magenta circles, centered on + symbol. Contour dashed lines in black (from inner to outer lines) represent the 68%, 95%, and 99% containment region for the γ-ray signature. Left: high-energy TS map (20 × 20, 0.05°/pixel) taking into consideration only 3500 GeV photons; zooming into central TS-peak region, the source 2WHSP J053628.9-334301 (hard γ-ray spectrum, Γ2WHSP = 1.76) is within the 68% containment region for the high-energy γ-ray signature. Right: lower-energy TS map (20 × 20, 0.08°/pixel) considering only 700800 MeV photons. In this case, the TS-peak position is dominated by the 5BZQ J0536-3401 (steep γ-ray spectrum source, ΓBZQ = 2.75) well within the 68% containment region for the lower-energy γ-ray signature.

Open with DEXTER
In the text
thumbnail Fig. 17

High-energy TS-maps 3500 GeV with 20 × 20 grid 0.05°/pixel, integrating over 7 yr of data. In both maps the 2WHSP J093430.1-172120 position is highlighted by a thick green circle centered on +; the 3FGL J0935.1-1736 is in the center of the thick cyan circle, and a blazar candidate (possible counterpart for the 3FGL source) is indicated with a thick magenta circle. Left panel: for this map, the 3FGL source is removed from the model input, showing the pure shape of TS distribution in the region. Green dashed line represents the 99% containment region for the γ-ray signature, which is compatible both with 3FGL and blazar candidate positions. The thin contour lines refer to TS surfaces of 40-30-20, only to show how the γ-ray signature is extended, matching the 2WHSP J093430.1-172120. Right panel: residual TS map (built using the 3FGL in the background) showing excess signal consistent with a point-like source. The green dashed lines correspond to the 68%, 95%, and 99% containment region for the γ-ray signature (from inner to outer lines).

Open with DEXTER
In the text
thumbnail Fig. 18

Sky-Explorer view around NVSS J093514-173658 (blazar candidate). Left side: XRT field with the X-ray detection marked as ×. Right side: UVOT detection indicated as +. The X-ray and UV error-circle are shown with a dotted blue line. The radio-source is marked in red, and its optical counterpart USNOB1.0 (J2000 Ra, Dec: 143.8116°,-17.6163°) is shown in green.

Open with DEXTER
In the text
thumbnail Fig. 19

Sky-Explorer view around 3FGL J0421.6+1950 positions indicated as a red cross. The blue dash-dotted line represents the error circle for the γ-ray detection reported in the 3FGL catalog. As shown, within 15 from the unassociated γ-ray source there is a 2WHSP blazar. X-ray and radio detections in the field are represented by blue and red circles, respectively.

Open with DEXTER
In the text
thumbnail Fig. 20

High-energy 3500 GeV TS-map, integrating 7 yr of data. The contour black dashed lines are TS surfaces representing 68%, 95% and 99% containment region for the γ-ray signature (from inner to outer lines). The 2WHSP J042218.3+195054 position is highlighted by a thick green circle centered on +. The 3FGL J0421.6+1950 position is highlighted by the cyan circle centered on ×, 9.7 away from the high-energy TS peak.

Open with DEXTER
In the text
thumbnail Fig. 21

High-energy 3300 GeV TS map, integrating 4 yr of data (in this case, the highest energy used is 300 GeV following recommendations for the use of Pass 7 data). The contour black dashed lines are TS surfaces representing 68%, 95%, and 99% containment region for the γ-ray signature (from inner to outer lines). The 2WHSP J042218.3+195054 position is highlighted by a thick green circle centered on +, well within the 68% containment, while the 3FGL J0421.6+1950 source (position highlighted by the cyan circle centered on ×) is localized within the 95% containment region.

Open with DEXTER
In the text
thumbnail Fig. 22

Left panel: error-circle (dash-dotted) associated with the 3FGL source. Right panel: high-energy 3500 GeV TS map, integrating over 7 yr of Fermi-LAT observations. The contour black dashed lines are TS surfaces representing 68%, 95% and 99% containment region for the γ-ray signature (from inner to outer lines). The 2WHSP J183806.7-600032 (thick green circle centered on +). The 3FGL J1838.5-6006 position is shows in cyan (centered on ×), 6.9 away from the high-energy TS peak.

Open with DEXTER
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.