The Radio/gamma-ray Connection from 120 Mhz to 230 Ghz

Radio loud active galactic nuclei are composed of different spatial features, each one characterized by different spectral properties in the radio band. Among them, blazars are the most common class of sources detected at gamma-rays by Fermi, and their radio emission is dominated by the flat spectrum compact core. In this contribution, we explore the connection between emission at high energy revealed by Fermi and at radio frequencies. Taking as a reference the strong and very highly significant correlation found between gamma rays and cm-λ radio emission, we explore the different behaviours found as we change the energy range in gamma rays and in radio, therefore changing the physical parameters of the zones involved in the emitted radiation. We find that the correlation weakens when we consider (1) gamma rays of energy above 10 GeV (except for high synchrotron peaked blazars) or (2) low frequency radio data taken by the Murchison Widefield Array; on the other hand, the correlation strengthens when we consider mm-λ data taken by Atacama Large Millimeter Array (ALMA).

The basic arguments favouring the existence of such a correlation go from purely theoretical reasoning, i.e., that emission in both bands naturally arises through processes involving a population of relativistic particles, to observational matters, such as the fact that radio loud (RL) active galactic nuclei (AGN) dominate the census of gamma-ray catalogues; they made up about one half of the historical EGRET third catalogue [9], and nearly 40% of the third Fermi Large Area Telescope (LAT) catalogue (3FGL, [10]), the largest and most recent compilation of gamma-ray sources.In more detail, 98% of the RL AGNs in the third catalogue of AGN detected by Fermi-LAT (3LAC, [11]) belong to the subclass of blazars, which are RL AGNs whose jet axis is closely aligned with our line of sight; this indicates that relativistic beaming is a key element in both the radio and high energy emission, suggesting a strong connection between the two domains.Finally, simple one-zone synchrotron self-Compton (SSC) models successfully reproduce the basic features of the broadband spectral energy distribution (SED) of several blazars (particularly of BL Lac flares, [12]), and radio luminosity itself was shown to govern the physical properties and radiation mechanisms in relativistic jets up to gamma rays in the early formulation of the blazar sequence framework [13,14].
On the other hand, there are several issues that suggest a much more complex picture.First, the gamma-ray variability time scales, in particular during flares, often reach short values, down to ∼1 h [15][16][17][18].This indicates that the high energy emission region must be extremely compact and thus optically thick at radio frequency, because of the synchrotron self-absorption (SSA) mechanism.Indeed, dedicated systematic radio monitoring of gamma-ray blazars shows that there are very few sources in which there is a significant correlation between the light curves in the two regimes [19]; so-called "orphan" gamma-ray flares with no radio counterpart are also quite common [20][21][22].Moreover, while it is true that most of the gamma-ray sources are RL AGNs, and in particular blazars, the opposite is definitely not true: Radio galaxies are very seldom detected in gamma rays, and even among blazars the detection rate at high energy is far from unity, both for luminous and not-so-luminous objects [23,24].Therefore, while radio emission and beaming are necessary conditions for the presence of high energy radiation, they are not sufficient.
In this paper, we review some recent results and prospects about the existence and implications of such correlations, including the study of unidentified gamma-ray sources (UGS).In Section 2, we review the results based on Fermi and GHz-frequency radio data [1], which constitute the most comprehensive dataset ever used to address this topic.In the following sections, we consider variations of the data analysis as gamma-ray or radio data of different wavelengths are considered: in Section 3, we focus on sources detected at energies E > 10 GeV; in Section 4, on sources detected at radio frequencies as low as 120 MHz [25] by the Murchison Widefield Array (MWA); in Section 5, on sources detected as high as 230 GHz with the Atacama Large Millimeter Array (ALMA).

Radio-Gamma Ray Connection between ∼GHz and E > 100 MeV Data
The most comprehensive approach ever used for the assessment of the connection between radio and gamma-ray data in extragalactic radio sources was presented in [1].In that work, we considered all 599 "clean" AGNs (i.e., gamma-ray sources with a single high confidence association and no analysis flag) contained in the first catalogue of AGN detected by Fermi-LAT (1LAC, [26]); this was not only the largest sample ever considered but also the one in which both blazar types (BL Lacs and flat spectrum radio quasars, FSRQ) were both adequately represented; most other works were limited to very bright sources, thus excluding most of the BL Lacs.Archival interferometric data with sub-arcsecond resolution in the GHz domain (typically, at 8 GHz) were available, as well as simultaneous 15 GHz data from the Owens Valley Radio Observatory (OVRO) monitoring project, at least for a subsample ([27] see also Pearson, this conference).Finally, a dedicated statistical analysis was developed in order to assess the significance of the correlation independent of the various biases that can occur in such kinds of studies [28].The main results can be summarised as follows:

•
The entire 599-source sample shows a correlation characterised by a Pearson's coefficient r = 0.47 (see Figure 1);

•
The chance probability of obtaining this value from two intrinsically uncorrelated quantities with the same dynamic range in flux density is smaller than 10 Considering the subset of OVRO-monitored sources, both the correlation strength r and significance P improve when considering simultaneous vs archival data (chance probability decreasing from 1.9 × 10 −6 to 9 × 10 −8 ); we note however that this sample contains "only" 161 sources.
Version July 14, 2016 submitted to Galaxies 3 of 8 Adapted from [5].Magenta symbols indicate sources without a measured redshift.

Radio-gamma ray connection between VLBI and E > 10 GeV data
While the work of [5] demonstrated the existence of a strong and highly significant correlation between radio data and high energy gamma rays, the situation is far more open when we consider Very High Energy (VHE, E > 100 GeV) gamma rays.In this domain, observations are carried out in targeted mode with imaging atmospheric Cherenkov telescopes (IACT), so that no systematic and unbiased survey exists.Sources are often detected when they are in flaring states, and the optimization of the observing strategy naturally carries on a selection bias in favour of sources physically similar to those already detected.Moreover, it is still true that all the classified extragalactic VHE sources are RL AGNs, yet their radio properties are far less extreme, with low-luminosity sources, and in particular HSP BL Lacs, being more frequently detected than luminous ones, such as LSP BL Lacs and FSRQs.
In this sense, the first Fermi-LAT catalogue of sources above 10 GeV [1FHL, 6] represents an ideal resource to try and connect the HE and VHE domains.The 2FHL and the surveys that will be carried out with the Cherenkov Telescope Array (CTA) in the coming years will allow us to extend our understanding.The 1FHL is based on three years of Fermi-LAT survey data and it is as uniform and unbiased as possible.It contains 514 sources, 76% of which are AGN and 13% un-associated; the AGN fraction is larger than in 3FGL and the census leans towards extreme spectral type blazars (HSP, 41%).The fraction of UGS is lower, thanks to a generally smaller positional uncertainty ellipses when higher energy photons are considered.
We are carrying out a project aimed at studying all the sources in the 1FHL with very long baseline interferometry (VLBI), in order to characterize the parsec scale properties of this population, which several works have highlighted to be rather peculiar [14,15,21,22,36].Lico et al. [23] have presented new VLBI images for the less studied objects, proposing also associations for some of the UGS.Thanks to these new images, it is now possible to discuss the radio-gamma correlation for the sample (in particular, the northern subset) without any observational bias.If we consider the VLBI flux density and the 3FGL energy flux (i.e. the gamma-ray energy flux in the entire E > 100 MeV band), the sample shows a strong and highly significant correlation, with r = 0.73 and P chance < 10 −6 , even stronger than what presented in Sect. 2. This is likely due to the fact that VLBI observations filter out extended emission and provide a measurement of a jet region much closer to the gamma-ray zone.Adapted from [1].Magenta symbols indicate sources without a measured redshift.

Radio-Gamma ray Connection between VLBI and E > 10 GeV Data
While the work of [1] demonstrated the existence of a strong and highly significant correlation between radio data and high energy gamma rays, the situation is far more open when we consider Very High Energy (VHE, E > 100 GeV) gamma rays.In this domain, observations are carried out in targeted mode with imaging atmospheric Cherenkov telescopes (IACT), so that no systematic and unbiased survey exists.Sources are often detected when they are in flaring states, and the optimization of the observing strategy naturally carries a selection bias in favour of sources physically similar to those already detected.Moreover, it is still true that all the classified extragalactic VHE sources are RL AGNs, yet their radio properties are far less extreme, with low-luminosity sources, and in particular HSP BL Lacs, being more frequently detected than higher bolometric luminosity ones, such as LSP BL Lacs and FSRQs.
In this sense, the first Fermi-LAT catalogue of sources above 10 GeV (1FHL, [29]) represents an ideal resource to try and connect the HE and VHE domains.The 2FHL (above 50 GeV, [30]) and the surveys that will be carried out with the Cherenkov Telescope Array (CTA) in the coming years will allow us to extend our understanding.The 1FHL is based on three years of Fermi-LAT survey data and it is as uniform and unbiased as possible.It contains 514 sources, 76% of which are AGN and 13% are UGS; the AGN fraction is larger than in 3FGL and the census leans towards extreme spectral type blazars (HSP, 41%).The fraction of UGS is lower, thanks to generally smaller positional uncertainty ellipses when higher energy photons are considered.
We are carrying out a project aimed at studying all the sources in the 1FHL with very long baseline interferometry (VLBI), in order to characterize the parsec scale properties of this population, which several works have highlighted to be rather peculiar [31][32][33][34][35]. Lico et al. [36] have presented new 5 GHz VLBI images for the less studied objects, proposing also associations for some of the UGS.Thanks to these new images, it is now possible to discuss the radio-gamma correlation for the sample (in particular, the northern subset) without any observational bias.If we consider the VLBI flux density and the 3FGL energy flux (i.e., the gamma-ray energy flux in the entire E > 100 MeV band), the sample shows a strong and highly significant correlation, with r = 0.73 and P chance < 10 −6 , even stronger than what was presented in Section 2. This is likely due to the fact that VLBI observations filter out extended emission and provide a measurement of a jet region much closer to the gamma-ray zone.
However, when we consider gamma rays of energy E > 10 GeV, the correlation essentially vanishes, with r = −0.02.Both weak (typically, BL Lacs) and bright (typically, FSRQs) radio sources have lower gamma-ray energy flux in this band, which is obvious as the band itself is narrower.However, the brightest sources have softer photon indexes, which results in a more pronounced decrease of the gamma-ray energy flux for these objects (Figure 2).As a consequence, no significant trend is observed.Only if we focus on the sub-class of HSP blazars, do we find evidence of a correlation between the two bands (r = 0.60, P chance = 0.0032).
Version July 14, 2016 submitted to Galaxies 4 of 8 However, when we consider gamma rays of energy E > 10 GeV, the correlation essentially 111 vanishes, with r = −0.02.Both weak (typically, BL Lacs) and bright (typically, FSRQs) radio sources 112 have lower gamma-ray energy flux in this band, which is obvious as the band itself is narrower.
113 However, the brightest sources have softer photon indexes, which results in a more pronounced 114 decrease of the gamma-ray energy flux for these objects (Figure 2).As a consequence, no significant 115 trend is observed.Only if we focus on the sub-class of HSP blazars, we do find evidence of a 116 correlation between the two bands (r = 0.60, P chance = 0.0032).the Roma BZCat [28], finding low frequency matches for 36% of all known blazars in the MWACS sky 131 area: the detection rate is higher for FSRQs than for BL Lacs, and it is higher for gamma-ray blazars 132 than for non-gamma-ray ones (see Table 1). 133 In terms of low frequency spectral indexes, blazars are flatter than the rest of the sources in However, when we consider gamma rays of energy E > 10 GeV, the correlation essentially vanishes, with r = −0.02.Both weak (typically, BL Lacs) and bright (typically, FSRQs) radio sources have lower gamma-ray energy flux in this band, which is obvious as the band itself is narrower.
However, the brightest sources have softer photon indexes, which results in a more pronounced decrease of the gamma-ray energy flux for these objects (Figure 2).As a consequence, no significant trend is observed.Only if we focus on the sub-class of HSP blazars, we do find evidence of a correlation between the two bands (r = 0.60, P chance = 0.0032).

Radio-gamma ray connection between ∼ 100 MHz and E > 100 MeV data
Although somehow counter-intuitively, low frequency observations have been shown to be useful in studying blazars and in classifying UGS [29,30,33].The main reason is that UGSs are in general faint, and on the basis of the correlation demonstrated in [5], we can expect them to be associated to weak radio sources.Since weak radio sources are much more numerous than the bright ones, it is difficult to pinpoint the correct counterpart of a given UGS.It is therefore mandatory to add some physical information about the nature of the known gamma-ray sources and low frequency surveys are a relatively cheap way to obtain this information, e.g. by studying the spectral index.
The MWA [41] is the first operational Square Kilometer Array (SKA) precursor.It paves the way towards low frequency (< 200 MHz) radio astronomy.A commissioning survey catalogue (MWACS, presented in [19]) includes ∼ 14, 000 sources over 6, 100 deg 2 in the southern sky, with a 3σ sensitivity of ∼ 120 mJy, a positional accuracy of ∼ 3 , flux density values at 120, 150, 180 MHz, and the corresponding spectral index α low .We cross-correlated the MWACS with the fifth edition of the Roma BZCat [28], finding low frequency matches for 36% of all known blazars in the MWACS sky area: the detection rate is higher for FSRQs than for BL Lacs, and it is higher for gamma-ray blazars than for non-gamma-ray ones (see Table 1).
In terms of low frequency spectral indexes, blazars are flatter than the rest of the sources in (b)

Radio-Gamma ray Connection between ∼100 MHz and E > 100 MeV Data
Although somehow counter-intuitively, low frequency observations have been shown to be useful in studying blazars and in classifying UGS [37][38][39].The main reason is that UGSs are in general faint, and on the basis of the correlation demonstrated in [1], we can expect them to be associated with weak radio sources.Since weak radio sources are much more numerous than the bright ones, it is difficult to pinpoint the correct counterpart of a given UGS.It is therefore mandatory to add some physical information about the nature of the known gamma-ray sources and low frequency surveys are a relatively cheap way to obtain this information, e.g., by studying the radio spectral index.
The MWA [40] is the first operational Square Kilometer Array (SKA) precursor.It paves the way towards low frequency (<200 MHz) radio astronomy.A commissioning survey catalogue (MWACS , presented in [41]) includes ∼14,000 sources over 6100 deg 2 in the southern sky, with a 3σ sensitivity of ∼120 mJy, a positional accuracy of ∼3 , flux density values at 120, 150, 180 MHz, and the corresponding spectral index (We define the spectral index α such that S ν ∝ ν −α ) α low .We cross-correlated the MWACS with the fifth edition of the Roma BZCat [42], finding low frequency matches for 36% of all known blazars in the MWACS sky area: the detection rate is higher for FSRQs than for BL Lacs, and it is higher for gamma-ray blazars than for non-gamma-ray ones (see Table 1).
In terms of low frequency spectral indexes, blazars are flatter than the rest of the sources in MWACS, with α low, blazars = 0.51 ± 0.05 and α low, all = 0.81 ± 0.01 (Figure 3, left panel).Moreover, many sources whose ∼GHz flux densities are above 120 mJy are not detected in MWA, indicating that they must have inverted spectra between 120 MHz and ∼1 GHz.Taken together, these results indicate that the flat-spectrum core remains prominent also at low frequency.
On the other hand, a linear fit to the gamma-ray energy flux vs MWA flux density at 180 MHz shows no significant (r = 0.26, P chance = 0.27) correlation (Figure 3, right panel).This indicates that a substantial contribution from the non-beamed lobes enters in the total radio flux density.Assuming the lobes to have a spectral index α l = 0.81 (the MWACS mean spectral index for the non-blazar population) and the core to have α c = 0.096 (as indicated by the ∼ 1 GHz to 20 GHz spectral index of the MWACS blazars), we determined that the flux density ratio of the two components is S c /S l ∼ 0.5, 3.5, 27 at 120 MHz, 1 GHz, and 20 GHz, respectively.
A detailed presentation of this work can be found in [25].Future deeper all-sky surveys with MWA and LOFAR [43] will further improve our detection rates and provide a clearer picture of the relative contribution of these components.

Radio-Gamma Ray Connection between 230 GHz and E > 100 MeV Data
Deep observations at high frequency are certainly more challenging to obtain for large samples, as the atmospheric conditions are more problematic and the instrument field of view is smaller.Since they provide crucial information about the most compact regions, it is however important to try to obtain millimetre data for as many sources as possible.For this reason, we selected a representative subsample of the 3LAC to be observed at 230 GHz with ALMA.This sample contains 77 sources, including 26 BL Lacs, 31 FSRQs, 16 blazars of uncertain type, 1 narrow-line Seyfert 1 galaxy, and 2 radio galaxies.In terms of SED-defined sub-classes, it contains 43, 14, and 18 blazars of the LSP, ISP, and HSP types.The gamma-ray flux and photon index distribution overlap strongly with those of the entire 3LAC, so that this sub-sample is suitable to represent the entire known gamma-ray blazar population (Figure 4).
The sources were observed in ALMA Cycle 3; all targets were detected, with rms noise generally in agreement with the predicted values of 0.3-0.5 mJy•beam −1 ; some exceptions were found for the brightest sources, due to dynamic range limitations.Flux densities range between just above 1 mJy and 6.5 Jy, for the core of Cen A.
A preliminary comparison of the ALMA and Fermi-LAT data already shows some interesting findings.The 230 GHz flux density and the E > 100 MeV energy flux are correlated with a coefficient r ALMA−LAT = 0.55; the correlation coefficient is significantly larger than what we find if we consider 1.4 GHz arcsecond-scale data, i.e., r NVSS−LAT = 0.39.A more detailed analysis and discussion is in progress, including an assessment of the importance of comparing simultaneous data.
Version July 14, 2016 submitted to Galaxies 6 of 8 entire 3LAC, so that this sub-sample is suitable to represent the entire gamma-ray blazar population 158 (Figure 4).

Figure 1 .
Figure 1.Gamma-ray energy flux at E > 100 MeV vs 8 GHz radio flux density for the 1LAC sample.Adapted from [5].Magenta symbols indicate sources without a measured redshift.

Figure 1 .
Figure 1.Gamma-ray energy flux at E > 100 MeV vs 8 GHz radio flux density for the 1LAC sample.Adapted from [1].Magenta symbols indicate sources without a measured redshift.

117 4 .
Radio-gamma ray connection between ∼ 100 MHz and E > 100 MeV data 118Although somehow counter-intuitively, low frequency observations have been shown to be 119 useful in studying blazars and in classifying UGS[29,30,33].The main reason is that UGSs are 120 in general faint, and on the basis of the correlation demonstrated in[5], we can expect them to be 121 associated to weak radio sources.Since weak radio sources are much more numerous than the bright 122 ones, it is difficult to pinpoint the correct counterpart of a given UGS.It is therefore mandatory to add 123 some physical information about the nature of the known gamma-ray sources and low frequency 124 surveys are a relatively cheap way to obtain this information, e.g. by studying the spectral index.125TheMWA[41] is the first operational Square Kilometer Array (SKA) precursor.It paves the 126 way towards low frequency (< 200 MHz) radio astronomy.A commissioning survey catalogue 127 (MWACS, presented in [19]) includes ∼ 14, 000 sources over 6, 100 deg 2 in the southern sky, with 128 a 3σ sensitivity of ∼ 120 mJy, a positional accuracy of ∼ 3 , flux density values at 120, 150, 180 MHz, 129 and the corresponding spectral index α low .We cross-correlated the MWACS with the fifth edition of 130

Figure 3 .
Figure 3. (a): Cumulative distribution function for the low frequency spectral index of the entire MWACS catalogue (red line) and the blazar subset (blue line); (b): Gamma-ray energy flux at E > 100 MeV vs. 120 MHz flux density (main panel) or upper limits (side histogram).Adapted from [25].Reproduced with permission from Astronomy & Astrophysics, c ESO.

Figure 4 .
Figure 4. Gamma-ray photon flux and index distribution for the entire 3LAC (solid histogram) and the subset observed with ALMA (dashed histogram).

159 1 . 4
The sources were observed in ALMA Cycle 3; all targets were detected, with rms noise generally 160 in agreement with the predicted values of 0.3-0.5 mJy beam −1 ; some exceptions were found for the 161 brightest sources, due to dynamic range limitations.Flux densities range between just above 1 mJy 162 and 6.5 Jy, for the core of Cen A. 163 A preliminary comparison of the ALMA and Fermi-LAT data already shows some interesting 164 findings.The 230 GHz flux density and the E > 100 MeV energy flux are correlated with a coefficient 165 r ALMA−LAT = 0.55; the correlation coefficient is significantly larger than what we find if we consider 166 GHz data, i.e. r NVSS−LAT = 0.39.A more detailed analysis and discussion is in progress, including 167 an assessment of the importance of comparing simultaneous data. 168

Table 1 .
Detection rates of blazars in the MWACS.