© ESO, 2012

1. Introduction

Blazars are jet-dominated extragalactic objects characterized by the emission of strongly variable and polarized non-thermal radiation across the entire electromagnetic spectrum, from radio waves to γ-rays (e.g., Urry & Padovani 1995). As the extreme properties of these sources are due to the relativistic amplification of radiation emitted along a jet pointing very close to the line of sight (e.g., Blandford & Rees 1978; Urry & Padovani 1995), they are rare compared to both objects pointing their jets at random angles and radio quiet quasi stellar objects (QSOs) where the emitted radiation is due to thermal or reflection mechanisms ultimately powered by accretion onto a supermassive black hole (e.g., Abdo et al. 2010a). Despite that, the strong emission of blazars at all wavelengths makes them the dominant type of extragalactic sources in the radio, microwave, γ-ray, and TeV bands where accretion and other thermal emission processes do not produce significant amounts of radiation (Toffolatti et al. 1998; Giommi & Colafrancesco 2004; Hartman et al. 1999; Abdo et al. 2010a; Costamante & Ghisellini 2002; Colafrancesco & Giommi 2006; Weekes 2008). For these reasons, blazars are hard to distinguish from other sources at optical and X-ray frequencies, while they dominate the microwave and γ-ray sky at high Galactic latitudes. The advent of the Fermi (Atwood et al. 2009) and Planck¹ (Tauber et al. 2010; Planck Collaboration 2011a) satellites, which are surveying these two observing windows, combined with the versatility of the Swift observatory (Gehrels et al. 2004), is giving us the unprecedented opportunity to collect multi-frequency data for very large samples of blazars and open the way to a potentially much deeper understanding of the physics and demographics of these still puzzling objects.

Blazars can be categorized by their optical properties and the shape of their broad-band spectral energy distributions (SEDs). Blazar SEDs always show two broad bumps in the log ν – log νF_ν space; the lower energy one is usually attributed to synchrotron radiation, while the more energetic one is attributed to inverse Compton scattering. Blazars displaying strong and broad optical emission lines are usually called flat-spectrum radio quasars (FSRQs), while objects with no broad emission lines (i.e., rest-frame equivalent width, EW, <5Å) are called BL Lac objects. Padovani & Giommi (1995) introduced the terms LBL and HBL to distinguish between BL Lacs with low and high values of the peak frequency of the synchrotron bump (). Abdo et al. (2010a) extended this definition to all types of blazars and defined the terms LSP, ISP, and HSP (corresponding to low, intermediate, and high synchrotron peaked blazars) for the cases where  < 10¹⁴Hz, 10¹⁴Hz <   < 10¹⁵Hz, and  > 10¹⁵Hz, respectively. In the rest of this paper, we use the LSP/ISP/HSP nomenclature.

It is widely recognized that one of the most effective ways of studying the physics of blazars is through the use of multi-frequency data that is ideally simultaneous. There are several examples of studies following this approach (e.g., Giommi et al. 1995; von Montigny et al. 1995; Sambruna et al. 1996; Fossati et al. 1998; Giommi et al. 2002; Nieppola et al. 2006; Padovani et al. 2006), but in most cases the samples are heterogeneous and the data are sparse and non-simultaneous.

The compilation of simultaneous and well-sampled SEDs requires the organization of complex multi-frequency observation campaigns, involving the coordination of observations from several observatories. Such large efforts have been carried out only rarely, and almost exclusively on the occasion of large flaring events in a few bright and well-known blazars, e.g., 3C454.3 (Giommi et al. 2006; Abdo et al. 2009a; Vercellone et al. 2009), Mkn421, (Donnarumma et al. 2009; Abdo et al. 2011), and PKS2155−304 (Aharonian et al. 2009).

Significant progress has been made with the publication of a compilation of quasi-simultaneous SEDs of a large sample of γ-ray bright blazars (Abdo et al. 2010a). This is an important step forward from previous compilations, as the sample presented is statistically representative of the population of bright γ-ray selected blazars, and the data were quasi-simultaneous, that is collected within three months of the Fermi-LAT observations.

With Planck, Swift, and Fermi-LAT simultaneously in orbit, complemented by other space and ground-based observatories, it is now possible to assemble high-quality multi-frequency datasets that allow us to build simultaneous and well-sampled broad-band spectra of large and statistically well-defined samples of active galactic nuclei (AGNs).

In this and a companion paper (Planck Collaboration 2011e), we present the first results of a large cooperative program between the Planck, Fermi-LAT, and Swift satellites and a number of ground-based observatories, carried out to collect multi-frequency data on large samples of blazars selected using different criteria and observed when the sources lie in the field of view of Planck.

In this paper, we concentrate on blazars selected in the soft X-ray, hard X-ray, and γ-ray bands. We present the simultaneous data, test for flux correlations, and estimate some key parameters characterizing the SEDs. We then compare the results obtained for the different samples. Detailed fits to models, variability studies, and more complete theoretical interpretations will be presented elsewhere.

Throughout this paper, we define the radio-to-submillimetre spectral index α by S(ν) ∝ ν^α, and we adopt a ΛCDM cosmology with H₀ = 70 kms^-1Mpc^-1, Ω_m = 0.27, and Ω_Λ = 0.73 (Komatsu et al. 2011).

2. Sample selection

To explore the blazar parameter space from different viewpoints, we used several different criteria to select the blazars to be observed simultaneously by Planck, Swift, and Fermi. In this paper, we considered three samples of blazars that are flux-limited in the high-energy part of the electromagnetic spectrum: soft X-ray (0.1–2.4keV) sources from the ROSAT All-Sky Survey Bright Source Catalog (1RXS, Voges et al. 1999, hereafter RASS sample), hard X-ray (15–150keV) sources from the Swift-BAT 54-month source catalog (Cusumano et al. 2010, hereafter BAT sample), and γ-ray sources from the Fermi-LAT 3-month bright AGN source list (Abdo et al. 2009b, hereafter Fermi-LAT sample).

These high-energy-selected samples were complemented by a radio flux-limited sample of northern sources (hereafter radio sample), which is presented in the companion paper (Planck Collaboration 2011e). We used these four samples, defined in widely different parts of the electromagnetic spectrum, to try to disentangle the intrinsic properties of blazars from the heavy selection effects that often afflict blazar samples. In total we considered 175 sources.

We based our classification of different blazar types on the Roma-BZCAT catalog (Massaro et al. 2010), which is a compilation of known blazars that were carefully checked to determine their blazar type in a uniform and reliable way. Massaro et al. (2010) divided blazars into three categories: BZQ/FSRQ, in which the optical spectrum has broad emission lines; BZB/BL Lac objects, in which the optical spectrum is featureless or contains only absorption lines from the host galaxy; BZU/uncertain type, comprising objects for which the authors could not find sufficient data to safely determine the source classification, and objects that have peculiar characteristics (see Massaro et al. 2010, for details). According to this classification, 96 of our objects are of the FSRQ type, 40 are BL Lacs, and the rest are of uncertain type. About 160 were observed by Swift simultaneously with Planck, mostly by means of dedicated target of opportunity (ToO) pointings. In the following we describe the selection criteria for each high-energy selected sample. Details of the radio-selected sample are given in Planck Collaboration (2011e).

2.1. The issue of blazar classification

The classification of blazars as either featureless (BL Lacs) or broad-lined objects (FSRQs), although very simple in principle, is neither unambiguous, nor robust. The borderline between the two blazar subclasses, namely 5Å in the source rest-frame for the EW of any emission line, was originally defined as a result of the optical identification campaigns of the sources discovered in the first well-defined and complete samples of (bright) radio and X-ray selected objects (Stickel et al. 1991; Stocke et al. 1991). However, we now know that well-known BL Lac objects such as OJ287 – and BLLac itself – exhibit emission lines with EWs well above the 5Å limit on some occasions (Vermeulen et al. 1995; Corbett et al. 1996). Several other BL Lac objects have strong emission lines with EWs just below, and sometimes above the 5Å threshold, depending on the variable continuum level (see e.g. Lawrence et al. 1996; Ghisellini et al. 2011, and references therein). Well-known FSRQs such as 3C279 also appear nearly featureless during bright states (Pian et al. 1999). The detection of broad Lyman-α emission in the UV spectrum of classical BL Lacs such as Mkn421 and Mkn501 (Stocke et al. 2011) contributes to the blurring of the distinction between the two types of blazars.

It is difficult to differentiate between BL Lac objects and radio galaxies as BL Lacs have been defined as sources for which the 4000Å Ca H&K break (a stellar absorption feature in the host galaxy) is diluted by non-thermal radiation more than a certain amount that was first quantified by Stocke et al. (1991) and then revised by Marcha et al. (1996), Landt et al. (2002), and Landt et al. (2004). The level of non-thermal blazar light around 4000Å reflects the intrinsic radio power of the jet; it can be highly variable and depends strongly on the position of the peak of the synchrotron emission, thus ensuring that the border between BL Lacs and radio galaxies is quite uncertain. Giommi et al. (2012) tackled the problem of blazar classification using extensive Monte Carlo simulations and showed that the observed differences can be interpreted within a simple scenario where FSRQs and BL Lacs share the basic non-thermal emission properties.

In the present study, we relied on the blazar classification given in the Roma-BZCAT catalog (Massaro et al. 2010), which re-assessed the blazar subclass of each object after a critical review of the optical data available in the literature and in large public databases such as the SDSS (York et al. 2000). Despite that, some uncertainties remain, which may in turn influence our conclusions about the differences between BL Lacs and FSRQs. However, the large size of our samples ensures that a few misclassifications should not significantly affect our results. To assess the impact of both blazar misclassification and transitional objects in a quantitative way, it is necessary to perform detailed simulations.

2.2. The Fermi-LAT (γ-ray flux-limited) sample

Our γ-ray flux-limited sample was created from the Fermi-LAT Bright Source List² (Abdo et al. 2009b). We selected all the high Galactic latitude (|b| > 10°) blazars detected with high significance (TS > 100)³. To reduce the size of the sample and ensure that all the sources are well above the Planck sensitivity limit for one survey, we considered only the sources with radio flux density (taken from BZCAT) S_5GHz > 1Jy. We realized that this is a double cut, with a TS limit at γ-ray energies and a flux-density limit in the radio band. A TS limit translates into different γ-ray flux limits depending on the γ-ray spectral slope, with higher sensitivity to flat-spectrum sources (see Fig. 7 of Abdo et al. 2009b). Hence, for our statistical considerations we also considered the subsample with a flux cut of F(E > 100 MeV) > 8 × 10^-8 phcm^-2s^-1, which removed this dependence on the spectral slope.

The sample so defined includes 50 sources, 40 of which are brighter than the γ-ray flux limit of 8 × 10^-8 phcm^-2s^-1. Relaxing the radio flux density cut would have provided a purely γ-ray flux-limited sample and increased the number of sources to ≈ 70, but with about 40–50% of the objects with S_5GHz < 1Jy being undetected by Planck.

Details are presented in Table 1, where Col. 1 gives the source common name, Col. 2 gives the Fermi-LAT name as it appears in Abdo et al. (2010b), Cols. 3 and 4 give the source position in equatorial coordinates, Cols. 5−7 give the redshift, V magnitude, and X-ray flux (0.1−2.4 keV) from BZCAT (Massaro et al. 2010), Col. 8 gives the 1.4GHz or 843MHz flux density from NVSS (Condon et al. 1998) or from SUMSS (Mauch et al. 2003) with Dec < −40°, Col. 9 gives the γ-ray flux from Abdo et al. (2010b)⁴, and Col. 10 gives the date of the Swift ToO observation made when the source was within the Planck field of view.

2.3. The Swift/BAT (hard X-ray flux-limited) sample

We defined our hard X-ray flux-limited sample starting from the Swift-BAT 54 month source catalog⁵ (Cusumano et al. 2010), and selecting all the sources identified with blazars with X-ray flux >10^-11ergcm^-2s^-1 in the 15–150 keV energy band. The BAT catalog includes 70 known blazars that satisfy the X-ray flux condition, but many of them are too faint to be detected at millimetre wavelengths by Planck. Therefore, although a pure X-ray selection would be preferable, we have decided to add a mild radio flux-density constraint (S_5GHz > 100mJy, with S_5GHz taken from BZCAT) to select only sources that can be detected by Planck or for which Planck will be able to provide meaningful upper limits, leaving enough sources to build a statistically sizable sample. The list of the 34 sources included in this sample is given in Table 2. The column description is the same as for Table 1.

2.4. The ROSAT/RASS (soft X-ray flux-limited) sample

The soft X-ray flux-limited sample was defined starting from the RASS catalog (1RXS) (Voges et al. 1999), and selecting all the sources identified with blazars with count rates higher than 0.3 counts/s in the 0.1–2.4keV energy band, and radio flux densities (taken from BZCAT) of S_5GHz > 200mJy. The reasons for using an additional radio flux constraint are the same as for the hard X-ray flux-limited sample, where, however, we chose 200mJy to reduce the size of the sample to be comparable to those of the γ-ray and hard X-ray samples. We realize that this is a stringent cut that removes about two thirds of the sources from the purely soft X-ray selected sample. However, all the sources below the radio threshold are HSP BL Lacs, thus implying that the subsample of LBL sources remains purely X-ray flux-limited and, consequently, that high objects are strongly underrepresented. The list of the 43 sources included in this sample is given in Table 3. The column description is the same as for Tables 1 and 2.

2.5. The radio flux-density limited sample

The radio flux-density limited sample is presented in the companion Planck paper (Planck Collaboration 2011e), where all the observational details are given. The sample consists of 104 bright northern and equatorial radio-loud AGN characterized by S_37GHz > 1Jy as measured with the Metsähovi radio telescope.

Although the samples are defined by different criteria, four sources are common to all samples. These are the well-known objects 3C273, 3C279, Mkn501, and 3C454.3, which are among the brightest objects across the entire electromagnetic spectrum. A summary of the number of sources common to more than one sample is given in Table 4.

3. Data analysis

3.1. Ground-based follow-up data

Following the launch of Planck, several follow-up programs with ground-based facilities started collecting simultaneous radio and optical data. In this paper, we used data from the observatories listed in Table 5.

3.1.1. APEX

Some sources from our sample were observed in the submillimetre domain with the 12-m Atacama Pathfinder Experiment (APEX) in Chile. The observations were made using the LABOCA bolometer array centered at 345GHz. Data were taken at two epochs: September 3–4, 2009, and November 12, 2009. The data were reduced using the script package minicrush⁶, version 30-Oct.-2009. Uranus was used as a calibrator of the flux densities.

3.1.2. ATCA-PACO

The Planck-ATCA Co-eval Observations (PACO) project (Massardi et al. 2011a; Bonavera et al. 2011) observed 480 sources selected from the Australia Telescope 20GHz catalogue (AT20G, Massardi et al. 2011b), with the Australia Telescope Compact Array (ATCA) in the frequency range 4.5−40 GHz, at several epochs close in time to the Planck observations in the period July 2009 to August 2010. The PACO sample is a complete, flux-density limited, and spectrally selected sample of southern sources, with the exception of the region with Galactic latitude |b| < 5°. A total of 147 PACO point-like sources have at least one observation within ten days of the Planck observations.

3.1.3. Effelsberg and IRAM

Quasi-simultaneous cm/mm radio spectra for a larger number of Planck blazars were obtained within the framework of a Fermi monitoring program of γ-ray blazars (F-GAMMA: Fuhrmann et al. 2007; Angelakis et al. 2008) on the Effelsberg 100-m and IRAM 30-m telescopes. The frequency range was 2.64−142GHz.

The Effelsberg measurements were conducted with the secondary focus heterodyne receivers at 2.64, 4.85, 8.35, 10.45, 14.60, 23.05, 32.00, and 43.00GHz. The observations were performed quasi-simultaneously with cross-scans, that is by slewing over the source position in azimuth and elevation with the number of sub-scans chosen to reach the desired sensitivity (for details, see Fuhrmann et al. 2008; Angelakis et al. 2008). Pointing offset, gain, atmospheric opacity, and sensitivity corrections were applied to the data.

The IRAM 30-m observations were carried out with calibrated cross-scans using the EMIR horizontal and vertical polarization receivers operating at 86.2GHz and 142.3GHz. The opacity-corrected intensities were converted into the standard temperature scale and finally corrected for small remaining pointing offsets and systematic gain-elevation effects. Conversion to a standard flux density scale was based on frequent observations of primary calibrators (Mars, Uranus) and secondary calibrators (W3(OH), K3-50A, NGC7027).

From this program, radio spectra measured quasi-simultaneously with the Planck observations were collected for a total of 37 Planck blazars during the period August 2009 to June 2010. Results are reported in Tables 6 and 7.

3.1.4. Medicina

The Simultaneous Medicina Planck Experiment (SiMPlE, Procopio et al. 2011) used the 32-m Medicina single dish to make almost simultaneous observations at 5GHz and 8.3GHz of the 263 sources of the NEWPS sample (Massardi et al. 2009) with Dec > 0°, and partially overlapping with the PACO observations for −10° <  Dec < 0°. The project began in June 2010 and finished in June 2011, observing our sample several times throughout two complete Planck surveys. It does not overlap with the Planck first survey.

3.1.5. Metsähovi

The 37GHz observations were made with the 13.7-m Metsähovi radio telescope using a 1GHz bandwidth, dual-beam receiver centered at 36.8GHz. We performed ON-ON observations, by alternating between the source and the sky in each feed horn. A typical integration time for obtaininig one flux density data point was 1200–1400s. The telescope detection limit at 37GHz was ~0.2Jy under optimal conditions. Data points with a signal-to-noise ratio (S/N) smaller than four are handled as non-detections. The flux-density scale was set by observations of DR21. Sources NGC7027, 3C274, and 3C84 were used as secondary calibrators. A detailed description of the data reduction and analysis is given in Teräsranta et al. (1998). The error estimate in the flux density includes the contribution from the measurement rms and the uncertainty in the absolute calibration.

3.1.6. OVRO

Some of the sources in our samples were monitored at 15GHz using the 40-m telescope of the Owens Valley Radio Observatory as part of a larger monitoring program (Richards et al. 2011). The flux density of each source was measured approximately twice weekly, with occasional gaps due to poor weather or instrumental problems. The telescope was equipped with a cooled receiver installed at prime focus, with two symmetric off-axis corrugated horn feeds that are sensitive to left circular polarization. The telescope and receiver combination produces a pair of approximately Gaussian beams (157arcsec FWHM), separated in azimuth by 12.95arcmin. The receiver has a central frequency of 15.0GHz, a 3.0GHz bandwidth, and a noise-equivalent reception bandwidth of 2.5GHz. Measurements were made using a Dicke-switched dual-beam system, with a second level of switching in azimuth where we alternated between source and sky in each of the two horns. Our calibration is referred to 3C286, for which a flux density of 3.44Jy at 15GHz is assumed (Baars et al. 1977). Details of the observations, calibration, and analysis are given by Richards et al. (2011).

3.1.7. RATAN

A six-frequency broadband radio spectrum was obtained with the RATAN-600 radio telescope in transit mode by observing simultaneously at 1.1, 2.3, 4.8, 7.7, 11.2, and 21.7GHz (Parijskij 1993; Berlin & Friedman 1996). Data were reduced using the RATAN standard software FADPS (Flexible Astronomical Data Processing System) reduction package (Verkhodanov 1997). The flux density measurement procedure at RATAN-600 is described by Aliakberov et al. (1985).

3.1.8. UMRAO

Centimetre-band observations were obtained with the University of Michigan 26-m prime focus paraboloid equipped with radiometers operating at central frequencies of 4.8, 8.0, and 14.5GHz. Observations at all three frequencies utilized rotating polarimeter systems permitting both total flux density and linear polarization to be measured. A typical measurement consisted of 8 to 16 individual measurements over a 20–40 min time period. Frequent drift scans were made across stronger sources to verify the telescope pointing correction curves; and observations of program sources were intermixed with observations of a grid of calibrator sources to correct for temporal changes in the antenna aperture efficiency. The flux scale was based on observations of Cassiopeia A (Baars et al. 1977). Details of the calibration and analysis techniques are described by Aller et al. (1985).

3.1.9. VLA

The Very Large Array (VLA) and, since Spring 2010, the Expanded VLA (EVLA), observed a subset of the sources as simultaneously as possible. Most of the VLA and EVLA runs were performed in one to two hour chunks of time. We observed during a one-hour chunk of time, in addition to flux calibrators and phase calibrators, typically 5–8 Planck sources. In many cases, VLA flux-density and phase calibrators were themselves of interest, since they were bright enough to be detected by Planck. For these bright sources, the integration times could be extremely short.

Integration times were about 30s at 4.86GHz and 8.46GHz, 100s at 22.46GHz, and 120s at 43.34GHz. All VLA/EVLA flux density measurements were calibrated using standard values for one or both of the primary calibrator sources used by NRAO, 3C48 or 3C286, and the u-v data were flagged, calibrated, and imaged using standard NRAO software (AIPS or CASA). It is important to bear in mind that the VLA and EVLA were in different configurations at different times in the several months duration of the observations. As a consequence, the angular resolution changed, becoming, for a given configuration, much higher at higher frequencies. For that reason, sources that appeared to be resolved in any VLA configuration or at any VLA frequency were carefully flagged.

3.1.10. KVA

Optical observations were made with the 35 cm KVA (Kunliga Vetenskapsakademiet) telescope at La Palma, Canary islands. All observations were made through the R-band filter (λ_eff = 640nm) using a Santa Barbara ST-8 CCD camera with a gain factor of 2.3e⁻/ADU and readout noise of 14 electrons. Pixels were binned 2 × 2 pixels giving a plate scale of 0.98arcsec/pixel. We obtained 3–6 exposures of 180s per target. The images were reduced in the standard way of subtracting the bias and dark frames and dividing by twilight flat-fields. The fluxes of the target and 3–5 stars in the target field were measured with aperture photometry and the magnitude difference between the target and a primary reference star in the same field was determined. The use of differential mode makes the observations insensitive to variations in atmospheric transparency and accurate measurements can be obtained even in partially cloudy conditions. The R-band magnitude of the primary reference star was determined from observations made on photometric nights, using comparison stars in known blazar fields as calibrators (Fiorucci & Tosti 1996; Fiorucci et al. 1998; Raiteri et al. 1998; Villata et al. 1998; Nilsson et al. 2007) and taking into account the color term of the R-band filter employed. After the R-band magnitude of the primary reference star was determined, the object magnitudes were computed from the magnitude differences. At this phase we assumed V−R = 0.5 for the targets. Several stars in the field were used to check the quality of the photometry and stability of the primary reference. The uncertainties in the magnitudes include the contribution from both measurement and calibration errors.

3.1.11. Xinglong

The monitoring at Xinglong Station, National Astronomical Observatories of China, was performed with a 60/90 cm f/3 Schmidt telescope. The telescope is equipped with a 4096 × 4096 E2V CCD, which has a pixel size of 12μm and a spatial resolution of 13 pixel^-1 . The observations were made with an intermediate-band filter, the I filter. Its central wavelength and passband width are 6685Å and 514Å, respectively. The exposure times were mostly 120s but ranged from 60s to 180s, depending on weather and lunar phase.

3.2. Planck microwave data

Planck (Tauber et al. 2010; Planck Collaboration 2011a) is the third generation space mission to measure the anisotropy of the cosmic microwave  background (CMB). It observes the sky in nine frequency bands covering 30–857GHz with high sensitivity and angular resolution from 31′ to 5′. Full sky coverage is attained in about seven months. The Low Frequency Instrument (LFI; Mandolesi et al. 2010; Bersanelli et al. 2010; Mennella et al. 2011) covers the 30, 44, and 70GHz bands with amplifiers cooled to 20K. The High Frequency Instrument (HFI; Lamarre et al. 2010; Planck HFI Core Team 2011a) covers the 100, 143, 217, 353, 545, and 857GHz bands with bolometers cooled to 0.1K. Polarization is measured in all but the highest two bands (Leahy et al. 2010; Rosset et al. 2010). A combination of radiative cooling and three mechanical coolers produces the temperatures needed for the detectors and optics (Planck Collaboration 2011b). Two data processing centers (DPCs) check and calibrate the data and make maps of the sky (Planck HFI Core Team 2011b; Zacchei et al. 2011). Planck’s sensitivity, angular resolution, and frequency coverage make it a powerful instrument for Galactic and extragalactic astrophysics as well as cosmology. Early astrophysics results are given in Planck Collaboration (2011h–z).

The Early Release Compact Source Catalog (ERCSC, Planck Collaboration 2011c) contains all sources, both Galactic and extragalactic, detected with high confidence over the full sky during the period between August 12, 2009 and June 6, 2010 (corresponding to Planck operational days 91 to 389). The ERCSC only contains average intensity information for the sources. However, many of the sources were observed more than once during the time period spanned by the ERCSC, and some of the Swift observations used for this paper were carried out between June and October 2010. Therefore, to have simultaneous data, we produced independent maps for the first (OD 91−274), the second (OD 275–456), and the beginning of the third Planck survey (OD 457–550) through the LFI and HFI pipelines described in Zacchei et al. (2011) and Planck HFI Core Team (2011b), and we extracted the flux densities from each map using IFACMEX, which is an implementation of the Mexican Hat Wavelet 2 (MHW2) algorithm available at the LFI DPC. The MHW2 tool has been extensively used to detect point-like objects in astronomical images, both with simulations from various experiments and data from the WMAP, Planck, and Herschel satellites (González-Nuevo et al. 2006; López-Caniego et al. 2006, 2007; Massardi et al. 2009). This wavelet is defined as the fourth derivative of the two-dimensional Gaussian function, where the scale of the filter R is optimized to look for the maximum in the S/N of the sources in the filtered map. In practice, the IFCAMEX code, our implementation of the MHW algorithm, deals with flexible image transport system (FITS) maps in Healpix format (Górski et al. 2005) and can be used to detect sources in the whole sky or at the position of known objects. For this analysis, we looked for objects above the 4σ level at the positions corresponding to the 105 sources of our sample. For objects with S/N smaller than four we adopted the 4σ level as an upper limit. The results of the analysis of Planck simultaneous data are reported in Table 8, where Cols. 1 and 2 give the source name, Cols. 3 and 4 give the observation start and end times, and Cols. 5–13 give the flux densities in units of Jy at 30, 44, 70, 100, 143, 217, 353, 545, and 857GHz.

Owing to source variability, we do not expect these simultaneous flux densities to be the same as the time-averaged ERCSC measurements, except in the case of the sources that were observed only once during the ERCSC time range and for which we estimated the Planck flux densities, measured simultaneously with the Swift observation, in the same period. We verified that, for the sources fulfilling these requirements, the flux densities extracted for this paper are in good agreement with those of the ERCSC.

In addition to simultaneous Planck data, we also used ERCSC flux densities in both our analysis of flux correlations (Sect. 7.2) and the SED plots described in Sect. 5.

3.3. Swift optical, UV, and X-ray data

The Swift Gamma-Ray-Burst (GRB) Explorer (Gehrels et al. 2004) is a multi-frequency space observatory devoted to the discovery and rapid follow-up of GRBs. There are three instruments on board the spacecraft: the UV and Optical Telescope (UVOT, Roming et al. 2005), the X-Ray Telescope (XRT, Burrows et al. 2005) sensitive to the 0.3–10.0keV band, and the Burst Alert Telescope (BAT, Barthelmy et al. 2005) sensitive to the 15−150keV band. Although the primary scientific goal of the satellite is the observation of GRBs, the wide frequency coverage is suitable for blazar studies, because it covers the region where the transition between synchrotron and inverse Compton emission usually occurs.

When not engaged in GRB observations, Swift is available for target of opportunity (ToO) requests, and the Swift team decided to devote an average of three ToO observations per week to this project for simultaneous observations of blazars.

3.3.1. UVOT

The Swift UVOT telescope can produce images in each of its six lenticular filters (V, B, U, UVW1, UVM2, and UVW2). However, in an effort to reduce the use of mechanical parts after several years of orbital operations, observations are carried out using only one filter, unless specifically requested by the user. Thus images are not always available for all filters.

The photometry analysis of all our sources was performed using the standard UVOT software distributed within the HEAsoft 6.8.0 package and the calibration included in the latest release of the “Calibration Database”. A specific procedure was developed at the ASI Science Data Center (ASDC) to process all ToO observations requested for the blazar sample. Counts were extracted from apertures of 5′′ radius for all filters and converted to fluxes using the standard zero points (Poole et al. 2008). The fluxes were then de-reddened using the appropriate values of E(B−V) for each source taken from Schlegel et al. (1998) with A_λ/E(B−V) ratios calculated for UVOT filters using the mean Galactic interstellar extinction curve from Fitzpatrick (1999). No variability was detected within single exposures in any filter. The processing results were carefully validated including checks for possible contamination by nearby objects within the source and background apertures. Some sources, such as 3C273 and NGC1275, needed special analysis, and results for some other sources had to be discarded.

The results of the UVOT data analysis are summarized in Table 9, where Cols. 1 and 2 give the source name, Cols. 3 and 4 give the observation date and the Swift observation ID, and the remaining columns give the magnitudes in the six UVOT filters with errors.

3.3.2. XRT

The Swift XRT is usually operated in “auto-state” mode, which automatically adjusts the CCD read-out mode to the source brightness, in an attempt to avoid pile-up (Burrows et al. 2005; Hill et al. 2004). As a consequence, some of the data were collected using the most sensitive photon counting (PC) mode, while windowed timing (WT) mode was used for bright sources.

The XRT data were processed with the XRTDAS software package (v. 2.5.1, Capalbi et al. 2005) developed at the ASI Science Data Center (ASDC) and distributed by the NASA High Energy Astrophysics Archive Research Center (HEASARC) within the HEASoft package (v. 6.9). Event files were calibrated and cleaned with standard filtering criteria using the xrtpipeline task and the latest calibration files available in the Swift CALDB. Events in the energy range 0.3–10keV with grades 0−12 (PC mode) and 0–2 (WT mode) were used for the analysis.

Events for the spectral analysis were selected within a circle of 20 pixels (~47″) radius, which encloses about 90% of the point spread function (PSF) at 1.5keV (Moretti et al. 2005), centered on the source position. When the source count rate is above ~0.5 counts/s, the PC mode data are significantly affected by pile-up in the inner part of the PSF. In these cases, and after comparing the observed PSF profile with the analytical model derived by Moretti et al. (2005), we removed pile-up effects by excluding events detected within up to 6 pixels from the source position, and used an outer radius of 30 pixels. The value of the inner radius was evaluated individually for each observation affected by pile-up, in a way that depended on the observed source count rate.

Ancillary response files were generated with the xrtmkarf task by applying corrections for the PSF losses and CCD defects. Source spectra were binned to ensure a minimum of 20 counts per bin when utilizing the χ² minimization fitting technique. We fitted the spectra adopting an absorbed power-law model with photon index Γ_x. When deviations from a single power-law model were found, we adopted a log-parabolic law of the form F(E) = KE^{(a + blog E)} (Massaro et al. 2004), which has been shown to fit the X-ray spectrum of blazars of the HSP type well (e.g. Giommi et al. 2005; Tramacere et al. 2009). This spectral model is described by only two parameters: a, the photon index at 1keV, and b, the curvature of the parabola. For both models, the amount of hydrogen-equivalent column density (N_H) was fixed to the Galactic value along the line of sight (Kalberla et al. 2005). For a fraction of the sources fitted with a power-law model (~15%), we found evidence of an absorption excess at low energies and the hydrogen column density N_H parameter was left free.

The results of the spectral fits with a power-law model and Galactic N_H are shown in Table 10, where Cols. 1 and 2 give the source name, Col. 3 gives the Swift observation date, Col. 4 gives the best-fit photon index Γ_x, Col. 5 gives the Galactic N_H, Cols. 6 and 7 give the 0.1–2.4 and 2–10keV X-ray fluxes, Col. 8 gives the value of the reduced χ², and Col. 9 gives the number of degrees of freedom.

In Table 11, we report data obtained using a log-parabola to describe the spectrum model. Columns 1 and 2 give the source name, Col. 3 gives the Swift observation date, Cols. 4 and 5 give the log parabola parameters a and b, Col. 6 gives the Galactic N_H, Cols. 7 and 8 give the 0.1–2.4 and 2–10keV X-ray fluxes, Col. 9 gives the value of the reduced χ², and Col. 10 gives the number of degrees of freedom.

Finally, in Table 12 we report the results of the spectral fits that were performed leaving the hydrogen column density N_H to vary as a free parameter. Columns 1 and 2 give the source name, Col. 3 gives the Swift  observation date, Col. 4 gives the best-fit photon index Γ_x, Col. 5 gives the estimated N_H, Cols. 6 and 7 give the 0.1–2.4 and 2–10keV X-ray fluxes, Col. 8 gives the value of the reduced χ², and Col. 9 gives the number of degrees of freedom.

3.4. Fermi-LAT γ-ray data

The Large Area Telescope (LAT) on-board Fermi is an electron-positron pair conversion telescope sensitive to γ-rays of energies from 20MeV to > 300GeV. The Fermi-LAT consists of a high-resolution silicon microstrip tracker, a CsI hodoscopic electromagnetic calorimeter, and an anticoincidence detector for charged particle background identification. A full description of the instrument and its performance can be found in Atwood et al. (2009). The large field of view (~2.4sr) allows the LAT to observe the full sky in survey mode every 3h. The LAT point spread function (PSF) depends strongly on both the energy and the conversion point in the tracker, but less on the incidence angle.

The LAT γ-ray spectra of all AGN sources are studied in Abdo et al. (2010b) based on 11 months of Fermi-LAT data. Here we derived the γ-ray spectra of the blazars for which we built the simultaneous SEDs, integrating for two weeks encompassing the whole duration of the Planck observations.

The Fermi-LAT data considered for this analysis cover the period from August 4, 2008 to November 4, 2010 and were analyzed using the standard Fermi-LAT ScienceTools software package⁷ (version v9r16) and selecting for each source only photons of energies above 100 MeV belonging to the diffuse class (Pass6 V3 IRF; Atwood et al. 2009), which have the lowest background contamination. For each source, we selected only photons within a 15° region of interest (RoI) centered around the source itself. To avoid background contamination from the bright Earth limb, time intervals where the Earth entered the LAT Field of View (FoV) were excluded from the data sample. In addition, we excluded observations in which the source under study was viewed at zenith angles larger than 105°, where Earth’s atmospheric γ-rays increase the background contamination. The data were analyzed with a binned maximum likelihood technique (Mattox et al. 1996) using the analysis software gtlike developed by the LAT team⁸. A model accounting for the diffuse emission and nearby γ-ray sources was included in the fit.

The diffuse foreground, including Galactic interstellar emission, extragalactic γ-ray emission, and residual CR background, was modeled with gll_iem_v02⁹ for the Galactic diffuse emission and isotropiciemv02 for the extragalactic isotropic emission. Each source under study was modeled with a power-law function (1)

where both the normalization factor N and the spectral index Γ were allowed to vary in the model fit. The model also includes all the sources within a 20° RoI included in Fermi-LAT one-year catalog (Abdo et al. 2010c) and modeled using power-law functions. If a source included in the model is a pulsar belonging to the Fermi-LAT pulsar catalog (Abdo et al. 2010d), we modeled the spectrum with a power-law with exponential cut-off using the spectral parameters in the pulsar catalog.

For the evaluation of the γ-ray SEDs, the whole energy range from 100MeV to 300GeV was divided into two equal logarithmically spaced bins per decade. In each energy bin, the standard gtlike binned analysis was applied assuming power-law spectra with photon index = −2.0 for all the point sources in the model. Assuming that in each energy bin the spectral shape can be approximated by a power-law, the flux of the source in all selected energy bins was evaluated, requiring in each energy bin a TS greater than ten. If the TS is lower than ten, an upper limit (UL) was evaluated in that energy bin. Only the statistical errors in the fit parameters are shown in the plots. Systematic errors due mainly to uncertainties in the LAT effective area derived from the on-orbit estimations, are <5% near 1GeV, 10% below 0.1GeV, and 20% above 10GeV.

For each source, we considered the three different integration periods for the γ-ray data:

• Simultaneous observations: data accumulated during the period of Planck observation of the source. As the Planck instruments point in slightly different directions and the field of view depends on the frequency of observation, a typical observation covering all Planck channels takes about one week, the exact integration time depending on the position of the source.

• Quasi-simultaneous observations: data integrated over a period of two months centered on the Planck observing period of the source.

• Twenty-seven month Fermi-LAT integration: data integrated over a period of 27 months from August 4, 2008 to November 4, 2010, i.e., the entire Fermi-LAT data set available for this paper.

Tables 13 and 14 give a summary of the γ-ray detections (TS > 25) in all our samples. The fraction of sources detected by Fermi-LAT during the simultaneous integrations is not very large and varies from ~40% in the Fermi-LAT sample to just ~20% in the soft X-ray selected sample. We note that even considering all the Fermi-LAT data available at the time of writing (27 month integration), a sizable fraction of the blazars in the radio and both soft and hard X-ray selected samples were not detected.

Detailed results of the Fermi-LAT analysis are given in Tables 15–20, where the observed fluxes or upper limits are given in six or three energy bands depending on the source brightness.

Two sources (PKS0548−322 and NGC7213) appear as significant γ-ray detections in our 27-month data set, although they were not included in any of the Fermi-LAT catalogues published so far (Abdo et al. 2009b, 2010c, and ¹⁰). These should therefore be considered as new γ-ray detections.

4. The importance of simultaneity

Blazars are, by definition, highly variable sources. It is therefore important to use simultaneous multi-frequency data to build SEDs for comparison with theoretical models. In this section, we compare our measurements with data taken from the literature in order to derive an estimate of the uncertainties introduced by the use of non-simultaneous data in different parts of the spectrum.

Figure 1 plots the Planck flux density at 44GHz presented in this paper versus the WMAP flux density at 41GHz from the WMAP point source catalogs (Bennett et al. 2003; Wright et al. 2009). Some scatter is present, but most of the points lie between the two solid lines indicating a factor of two variability.

Figure 2 plots the X-ray fluxes of the sources observed by Swift simultaneously with Planck (see Table 10) against the X-ray fluxes of the same sources from the BZCAT catalog (Massaro et al. 2009, 2010). In this case, a large scatter is present, with variations of over a factor of ten.

Figure 3 shows the Fermi-LAT γ-ray fluxes of our sources measured simultaneously with Planck plotted against their γ-ray fluxes in the Fermi-LAT 1FGL catalog (Abdo et al. 2010c). As in the X-ray sample, a scatter with variations larger than a factor of ten is observed.

We conclude that SEDs built with non-simultaneous data suffer from uncertainties in the microwave region that are relatively modest and generally limited to about a factor of two, while the high energy part of the spectrum (X-ray and γ-ray) is much more affected, with uncertainties caused by flux variations of up to a factor of ten or more. The same uncertainties, of course, apply when searching for correlations in non-simultaneous multi-frequency data.

5. Spectral energy distributions

We constructed the SEDs of all the blazars in our samples from the simultaneous multi-frequency data described above using the ASDC SED Builder, an on-line service developed at the ASI Science Data Center (ASDC)¹¹ (Stratta et al. 2011). This is a WEB-based tool that allows users to build multi-frequency SEDs combining data from local catalogs and external services (e.g., NED, SDSS, USNO) with the user’s own data. The tool converts observed fluxes or magnitudes into de-reddened fluxes at a given frequency using standard recipes that take into account the instrument response and assumed average spectral slopes. The SED builder can display SEDs both in flux and in luminosity (if redshift information is given); it also provides useful tasks such as the overlay of templates for blazar host galaxies and nuclear optical emission (blue-bump), and allows users to compare the SED with models including one or more SSC components.

The SEDs of all the sources in our samples are shown in Figs. 24–41. In these figures, red points represent strictly simultaneous multi-frequency data, green points represent γ-ray data integrated over a period of two months centered on the times of the Swift/Planck observations, ground-based data taken quasi-simultaneously, and Planck-ERCSC flux densities, and blue points represent γ-ray data integrated over the full period of 27 months. In the few cases where no Swift simultaneous observations could be obtained, we plot only Planck, Fermi-LAT, and ground-based data. Two-σ upper limits are indicated by arrows.

5.1. Distinguishing the non-thermal/jet-related radiation from QSO accretion and host galaxy emission

We used the simultaneous SEDs of Figs. 24–41 to determine some parameters that can constrain the physical mechanisms powering blazars. However, before doing so we had to identify and separate the radiation that is unrelated to the non-thermal, relativistically amplified emission from the jet of the blazars; that is, radiation from accretion onto the central black hole and from the host galaxy (see, e.g., Perlman et al. 2008). To do so, we estimated the contamination by the host galaxy assuming that all blazars are hosted by giant elliptical galaxies (e.g., Kotilainen et al. 1998; Nilsson et al. 2003; León-Tavares et al. 2011a) with absolute magnitude of M_R = −23.7. As Scarpa et al. (2000) and Urry et al. (2000) demonstrated, this value is within one magnitude of the observed values in a sample of over 100 BL Lacs observed with the Hubble Space Telescope. For the spectral shape, we used the elliptical galaxy template of Mannucci et al. 2001 (bottom panel of Fig. 4), which is based on low spectral resolution observations of a number of nearby galaxies in the wavelength range 0.12–2.4μm and is a good match to the predictions of spectrophotometric models for giant ellipticals Mannucci et al. (2001).

The radiation produced by accretion was estimated from the composite optical spectrum built by Vanden Berk et al. (2001) using over 2200 optical spectra of radio-quiet QSOs taken from the SDSS database (York et al. 2000) (top panel of Fig. 4), and the expected soft X-ray emission of radio quiet AGN from Grupe et al. (2010).

The ratio of optical to soft X-ray light has been known to be a function of optical luminosity since the early obervations of the Einstein observatory (Avni & Tananbaum 1986). More recently, this dependence has been confirmed using simultaneously acquired optical and soft X-ray data from Swift (Grupe et al. 2010) and XMM-Newton (Vagnetti et al. 2010). To assess the presence of a possible thermal component in the X-ray emission, we used the relationship given by Grupe et al. (2010)(2)where L_2500Å is the rest-frame luminosity of the thermal emission at 2500Å in units of WHz^-1 and α_{UV−X(radio−quietQSO)} is the usual slope between the UV (2500Å) and the soft X-ray (2keV) flux in radio quiet QSOs (e.g., Vagnetti et al. 2010).

Examples of the emission from these components unrelated to the jet are shown in Fig. 5, which shows the SEDs of Mkn501 and Mkn421 where the optical light is dominated by the host galaxy.

Figures 6 and 7 show the SEDs of BL Lacertae where part of the UV light is thought to originate in disk emission (Raiteri et al. 2009), the nearby FSRQ 3C273 (z = 0.158), the high-redshift FSRQs 4C38.41 (z = 1.814) and 1 Jy0537−286 (z = 3.104) where the optical/UV light is heavily, or completely, contaminated by radiation coming from accretion onto the central black hole. We compared the amplitude of the optical thermal emission to non-thermal radiation using the parameter α_{R−O(Thermal)}, defined as the spectral slope between the 5GHz radio flux density and the 5000Å optical flux density that can be attributed to the blue-bump/disk/thermal emission. This quantity depends on both the relativistic amplification factor and the intrinsic ratio of non-thermal/jet radiation to disk emission. In the case of FSRQs, the optical spectrum displays emission lines by definition, therefore we were able to constrain α_{R−O(Thermal)} by adjusting the optical thermal emission to the same level as the data. When the thermal blue bump was seen directly in the optical/UV part of the SED (e.g., Figs. 6 and 7), we fit the template of Vanden Berk et al. (2001) to the observed data; when the disk emission was not obviously visible we set the intensity of the blue-bump template to just below the observed optical/UV emission where the broad emission lines had been detected.

The BL Lacs do not show emission lines in their optical spectrum, so for this class of objects we could only set upper limits on α_{R−O(Thermal)}. We did that by assuming that the template for the optical thermal emission is at least one order of magnitude below the observed data, that is sufficiently low to hamper any broad line detection. The estimation of α_{R−O(Thermal)} relies strongly on the quality of the optical data available, in particular on the simultaneous UVOT data. Therefore we define an “optical data quality” flag as follows:

• 0: no simultaneous data available;

• 1: poor quality (e.g., only one UVOT filter available);

• 2: good quality (e.g., two or three UVOT filters available);

• 3: excellent quality (all UVOT filters available).

In Fig. 8 (upper panel), we show the distribution of α_{R−O(Thermal)} for all the FSRQs with excellent optical data for the whole sample and each of our samples independently. The distribution of the α_{R−O(Thermal)} upper limits for BL Lac objects is shown in the bottom panel. The results suggest a possible difference in the α_{R−O(Thermal)} distribution between different samples. As reported in Table 21, when only sources with excellent optical data are considered we obtain ⟨α_{R−O(Thermal)}⟩ = −0.64 ± 0.05 for

the ROSAT/RASS sample, to be compared with ⟨α_{R−O(Thermal)}⟩ = −0.73 ± 0.02 for the Fermi-LAT sample.

To assess the possible presence of a thermal component in the X-ray emission, for each source we compare the predicted thermal emission from accretion with the actual X-ray spectrum, and we also include the uncertainties in the parameters of Eq. (2) in determining the uncertainties given by 1, 2, and 3σ bands. We therefore define the following X-ray thermal contamination flags:

• 0: no X-ray data available;

• 1: X-ray emission mostly or entirely due to accretion/reflection (data agree with the expectations for accretion emission within 1σ);

• 2: X-ray emission probably contaminated by the accretion component (data agree with with the expectations for accretion emission within 2σ);

• 3: X-ray emission mostly of non-thermal origin (data are at 2–3σ from the expectations for radio quiet QSOs);

• 4: X-ray emission certainly of non-thermal origin (data are more than 3σ away from the expectations for radio quiet QSOs).

Results for all the FSRQs with good or excellent optical data are shown in Fig. 9, and summarized in Table 21 for the whole sample and each sample independently. We considered as contaminated all the sources with X-ray thermal contamination flag ≤2. There is a large difference between the Fermi-LAT and Swift-BAT samples, where ≲15% of the sources have a thermal component in their X-ray emission, and the ROSAT/RASS sample, where ~50% of the sources are contaminated. This demonstrates that the thermal component in the X-ray emission of blazars cannot be neglected, even in bright sources.

5.2. SED parameter estimation

We used the SEDs of all the objects in our samples to estimate the values of important physical parameters such as , F(), , and F() (see Table 22) taking into account only the non-thermal radiation and fitting third-degree polynomials as described in Abdo et al. (2010a) (see Figs. 5–7 for examples).

For the sources that were not detected by Fermi-LAT even in the 27-month integration, we estimated limits of and F() by constraining the polynomial in the high-energy part of the SED with the 27-month Fermi-LAT upper limits, as shown in Fig. 10.

In some HSP BL Lacs with particularly high  values (see, e.g., Figs. 30, 31, 36, and 38) the Fermi-LAT data alone are insufficient to ensure a good measure of , as the spectra are still rising at the highest Fermi-LAT energies and no simultaneous TeV data are available. The  values for these sources should therefore be considered as lower limits.

6. The spectral slope of blazars in the radio-microwave region

While WMAP results are consistent with a single flat spectral index in the relatively narrow frequency range 23–94GHz (Wright et al. 2009; Gold et al. 2011), the blazar spectrum must steepen at frequencies closer to the synchrotron peak. Adding Planck data and simultaneous ground-based observations at centimetre wavelengths to the WMAP data improves the spectral coverage, and allows us to probe the spectral shape of blazars over the much wider frequency range ~1 GHz to ~1 THz.

We studied the low frequency (LF) and the high frequency (HF) regions of the centimetre to sub-millimetre blazar spectra separately to search for differences in the spectral index α and determine the frequency at which the spectral index changes. We fitted the two frequency regions independently with power-laws to estimate the spectral indices at both low frequencies (α_LF, for ν ≤ ν_Break) and high frequencies (α_HF, for ν > ν_Break), assuming a range of break frequency values, ν_Break, from 30GHz to 100GHz. In Fig. 11, we show the distributions of α_LF and α_HF for ν_Break = 70GHz and 100GHz; the blazar spectra steepen from α_LF ~ 0 to α_HF ~ −0.65.

To verify the robustness of the results we repeated the analysis by imposing the two minimum numbers of independent frequencies needed to perform the fit, namely 3 and 5, and the results are consistent (Table 23).

We also analyzed the different classes of blazars – flat spectrum radio quasars (FSRQ), BL Lac objects (BL Lac), and radio-loud AGN of unknown classification (Uncertain Type) – separately, and we have found a difference at a level of about 3σ in α_HF, which is ~−0.7 for FSRQ but ~−0.5 for BL Lacs.

The spectral steepening and the difference in α_HF are visible in Fig. 12, where we show α_LF versus (vs.) α_HF for the sources that meet the requirements for the minimum amount of independent data at both low and high frequencies.

Our results are in complete agreement with the findings of Planck Collaboration (2011e) for the radio selected sample, i.e., flat spectral index at low frequency and α_HF ~ −0.6 at high frequency with a break frequency ~70 GHz, suggesting that this is a general feature of all blazars regardless of the selection criteria. There is also general agreement with the results presented in Planck Collaboration (2011d) for all the sources in the Planck ERCSC catalogs, and the slight discrepancy in the spectral index estimated at low frequency can be explained by the fact that, unlike Planck Collaboration (2011d), we have included the ground-based observations at 5–30GHz.

7. Searching for correlations between fluxes in different energy bands

We used our simultaneous and average multi-frequency data to identify possible correlations between fluxes measured in different energy bands. As some of the blazars in our samples were not detected by either Fermi-LAT or Planck, we estimated the significance of the correlations using the ASURV code Rev1.2 (Lavalley et al. 1992), which takes account of upper limits as described in Isobe et al. (1986). We search for possible correlations using fluxes or flux-densities rather than in luminosity-luminosity space because this allows us to use all objects in the samples, including those with no redshift (and consequently luminosity) information.

7.1. Microwave vs. X-ray

Figure 13 shows the Planck 143GHz flux density versus the simultaneous Swift-XRT X-ray flux for all the sources where the X-rays are expected to be due to the inverse Compton component (that is all LSP blazars) and not significantly contaminated by X-ray emission that is unrelated to the jet, such as that produced by the accretion process (see Sect. 5.1 for details). Sources that were not detected by Planck are plotted as upper limits; all the sources were detected in the X-ray band.

A correlation, although with some scatter, is clearly present. The Spearman rank coefficients (ρ) and the corresponding probabilities that the observed correlation is the result of chance are given in Table 24. Results obtained using Planck flux densities at other frequencies from 30GHz to 217GHz are similar and not shown here.

7.2. Microwave vs. γ-ray

The relationship between radio or microwave and γ-ray fluxes is a topic that has been addressed several times in the literature. A positive correlation between the radio and γ-ray fluxes is generally found, though with a large scatter (e.g., Kovalev et al. 2009; León-Tavares et al. 2010; Giroletti et al. 2010; Ghirlanda et al. 2010; Mahony et al. 2010; Peel et al. 2011; Linford et al. 2011; Ackermann et al. 2011; León-Tavares et al. 2011b). However, in most cases radio and non-simultaneous γ-ray data for sources detected in both energy bands are compared. For the first time, we present simultaneous microwave and γ-ray data and take into account upper limits.

The top panel of Fig. 14 shows the simultaneous Fermi-LAT γ-ray  flux plotted versus the Planck flux density at 143GHz. Sources with TS < 25 in the Fermi-LAT data and below 4σ in Planck maps are shown as upper limits. The plot shows a clear trend but from Table 25, which gives the Spearman’s ρ correlation parameter and the probability P that the observed level of correlation is caused by chance, we see that the level of significance of the correlation is never very high (P is of the order of a few percent and never lower than 0.05%), especially in the hard X-ray and γ-ray selected samples and for BL Lacs. This result is partly due to the large number of upper limits to the simultaneous γ-ray flux, which was estimated typically over a period of one week. To improve the statistics, in the middle panel of Fig. 14 we show the same plot using the Fermi-LAT γ-ray  flux integrated over a two-month period centered on the time of the Planck observations. From Table 25, we see that, although the significance of the correlation in the various samples and blazar classes substantially increases, it is never very high and the scatter remains.

This result may be due to the limited flux dynamic range in the γ-ray band for the Fermi-LAT sample and to the small number of objects (in the case of BL Lacs). However, a weak correlation could occur if the microwave flux density represents the superposition of multiple synchrotron components, while the simultaneous γ-ray flux represents the emission from a single dominant component that may be active for only a short time. For this reason, in the bottom panel of Fig. 14, we plot the Planck ERCSC flux densities (which in most cases represent the flux density averaged over more than one Planck observation) versus the Fermi-LAT flux averaged over the entire 27-month period. In this case the correlation is highly significant, although, the many upper limits in both energy bands clearly imply that the dispersion is very large.

7.3. X-ray vs. γ-ray

Figure 15 plots the 2-month γ-ray flux versus the simultaneous X-ray flux for all sources observed by Swift that do not show signatures of thermal contamination in the X-ray spectrum. Open red circles represent HSP sources, where the X-ray flux is due to the tail of the synchrotron emission, while black filled circles are LSP and ISP sources for which the X-ray flux is related to the inverse Compton radiation. This distinction is needed in order to properly compare sources where the emission is produced by the same mechanism. We therefore compute the γ-ray vs. X-ray correlation coefficient only for LSP and ISP sources. The Spearman rank test shows moderate evidence of a correlation that is less significant in the longer integrations: P = 2.55% for simultaneous data, P = 6.9% for a 2-month integration, or P = 17.4% for a 27-month integration.

Figure 16 plots the power-law spectral index in the Swift XRT energy band (0.3–10keV) versus the spectral slope in the Fermi-LAT γ-ray band (100MeV–100GeV) derived using the 2-month γ-ray data. A significant correlation is present, confirming the results of Abdo et al. (2010a). The top right corner of Fig. 16 shows the same plot built with simultaneous γ-ray data; although the number of objects is smaller, the correlation is still present.

8. Discussion and conclusions

We have collected simultaneous Planck, Swift, Fermi-LAT, and ground-based multi-frequency data for 105 blazars included in three statistically well-defined samples characterized by flux limits in the soft X-ray (0.1–2.4keV, ROSAT), hard X-ray (15–150keV, Swift-BAT), and γ-ray (E > 100MeV, Fermi-LAT) energy bands, with the addition of a cut to the radio 5GHz flux density to ensure a high probability of detection by Planck. This study complements a similar study of 104 radio-bright AGN (f_{37 GHz} > 1 Jy) (Planck Collaboration 2011e). Altogether, the four samples contains a total of 175 distinct objects. The acquisition of this unprecedented multi-frequency/multi-satellite data set was possible thanks to cooperation between the Planck, Swift, and Fermi-LAT teams, who agreed to share data and organize an extensive program of multi-frequency observations involving over 160 Swift ToO pointings scheduled when the blazars were within the field of view of the Planck instruments.

We have used this unique multi-frequency dataset to build well-sampled, simultaneous SEDs of all the blazars included in our high-energy selected samples. This collection of SEDs is an improvement over previous compilations (e.g., Abdo et al. 2010a) because: a) the SEDs presented here are strictly simultaneous, while in Abdo et al. (2010a) the multi-frequency data were collected over a period of up to nine months centered on the first three months of Fermi-LAT operations; b) the sources were selected according to different statistical criteria allowing us to probe the blazar parameter space from widely different viewpoints; and c) we took care to identify and separate radiation components unrelated to the emission from the jet, such as the light from the blazar’s host galaxy and the radiation produced by the accretion onto the central black hole, which often contaminate the non-thermal blazar spectrum in the optical, UV, and X-ray bands.

Our findings are broadly consistent with those of Abdo et al. (2010a). However, our analyses of larger samples selected in different parts of the electromagnetic spectrum, wider wavelength coverages, different level of simultaneity, and the ability to separate the emission components, have allowed us to make significant progress in several areas. The use of four widely different samples has allowed us to investigate the consequences of selection effects on the estimation of critical parameters such as ⟨⟩, average Compton dominance, especially for the case of BL Lacs.

Some of our sources have been observed simultaneously by Swift, Planck, and Fermi-LAT during more than one Planck survey. In these cases, we have presented only the data collected during the first observation. Multiple simultaneous observations of a subset of our blazars and detailed model fitting of the SEDs will be the subject of future papers. The main results of this work are discussed below.

8.1. Fermi-LAT detection statistics and the effects of variability

The percentage of Fermi-LAT detected sources during the simultaneous integrations, typically lasting about one week, ranges from ≲ 40% in the Fermi-LAT sample to 20–25% in the radio and X-ray selected samples (see Table 13). When 2-month integrations centered on the Planck-Swift observations are considered, these percentages grow to 80% in the Fermi-LAT sample and to ~ 35% in the other samples. However, even when using data from the 27-month Fermi-LAT integrations available at the time of writing, many of the blazars belonging to the radio, and both soft and hard X-ray selected samples remain undetected.

We note that the detection rate is quite different for FSRQs and BL Lacs: if we exclude the Fermi-LAT sample where all the objects have been detected (by definition), the percentage of detections is 95% for BL Lacs and only ~ 60% for FSRQs, with values ranging from 72% in the radio sample to just 53% in the RASS sample (see Table 14).

A comparison of our simultaneous data with published and archival measurements shows that the use of non-simultaneous data in the SED of blazars typically introduces a scatter of about a factor of two in the microwave band, and a factor of up to ten or more in the X-ray and γ-ray  bands.

8.2. The spectral slope in the radio-sub-millimetre region

We confirm that the energy spectrum of blazars in the radio–microwave spectral region is quite flat, with an average slope of ⟨α⟩ ~ 0 (f(ν) ∝ ν ^α) up to about 70GHz, above which it steepens to ⟨α⟩ ~ −0.6. This behaviour is very similar to that observed in the sample of radio bright blazars considered by Planck Collaboration (2011e; see also Tucci et al. 2011) confirming the findings of Abdo et al. (2010a) that the radio to microwave part of the spectrum is approximately the same in all blazars (FSRQs and BL Lacs) independently of the selection band. However, the spectral slope of BL Lacs above ~70 GHz is flatter than that of FSRQs with ⟨α_HF⟩ = −0.51 ± 0.07 for BL Lacs compared to ⟨α_HF⟩ = −0.73 ± 0.04 for FSRQs (see Table 23 and Fig. 11). A KS test, performed on the subsamples of FSRQs and BL Lacs with radio flux density larger than 1 Jy, gives a probability of less than 3% that the two samples come from the same parent population. This difference in the high frequency spectral index may reflect that the radio-submm part of the spectrum is closer to in LSP than in HSP sources.

8.3. Synchrotron self-absorption

We searched for signatures of synchrotron self-absorption, which in simple homogeneous SSC models is expected to cause strong spectral flattening below ~100 GHz, but we found no evidence of any common behaviour; indeed, the average spectrum in that region steepens instead of flattening. Possible cases where some evidence of synchrotron self-absorption may be present are PKS0454−234, PKS0521−36 (Fig. 27), S40917+44 (Fig. 29), PKS1127−145 (Fig. 31), and 3C454.3 (Fig. 40).

8.4. Non-thermal versus disk radiation

In several blazars, the optical/UV light is contaminated significantly by thermal/disk radiation (known as the blue bump, see Figs. 6 and 7), while the soft X-ray flux is contaminated by radiation produced in the accretion process in approximately 25% of the blazars in our samples (see Fig. 9). In some of the closest sources, the optical light is instead either contaminated or dominated by the emission from the host galaxy. Ignoring this contamination may cause an overestimate of the position of both and by 0.5 dex or more.

We investigated the relationship between the radiation produced by accretion and the jet, which co-exist in most FSRQs, using the parameter α_{R−O(Thermal)}, defined as the spectral slope between the 5GHz radio (non-thermal) flux and the 5000Å optical flux that can be attributed to the blue-bump/disk emission. In the blazar paradigm, this quantity depends on both the amount of relativistic amplification (which only affects the non-thermal radiation from the jet) and the intrinsic ratio of non-thermal/jet radiation to disk emission. We estimated the value of α_{R−O(Thermal)} in all sources for which we had optical data of good quality. Since BL Lacs do not display broad lines, only lower limits to α_{R−O(Thermal)} can be derived.

Figure 8 shows the distribution of α_{R−O(Thermal)} for all our samples, which range from ~0.4 to just above 1.0 and have peak values between 0.6 and 0.8. The distributions are all similar, with the largest difference being between the Fermi-LAT and the RASS samples (a KS test gives a probability of 3.6% that the two distributions originate from the same parent population), possibly reflecting differences in either the amplification factor or the ratio of accretion to jet emission between γ-ray and soft X-ray selected blazars. We note that the distribution of the α_{R−O(Thermal)} limits for BL Lacs is consistent with intrinsic values of α_{R−O(Thermal)} for BL Lacs that are within the range of values observed in FSRQs.

8.5. The distribution of rest-frame peak energies

The distribution of rest-frame synchrotron peak energies () of FSRQs is very similar in all our samples with a strong peak at ≈ 10^12.5Hz, an average of ⟨⟩10^{13.1 ± 0.1}Hz, and a dispersion of only ~0.5 dex. However, for BL Lacs the value of ⟨⟩ is at least one order of magnitude larger than that of FSRQs, the exact value depending considerably on the selection method (see Table 22 and Fig. 17). Since all the sources that are below the radio flux density cut in the RASS and BAT samples are ISP or HSP blazars, their inclusion would increase the difference between the distributions. The distributions of for FSRQs and BL Lacs also differ, but not as much as those of (see Fig. 18). The majority of the sources in all the samples (both FSRQs and BL Lacs) peak between 10²¹ Hz and 10²³Hz, with a few extreme HSP BL Lacs reaching ~10²⁶ Hz.

8.6. Correlations between fluxes and other blazar parameters

Despite the strict simultaneity of our data, plots of fluxes in different spectral regions (microwave  vs. X-ray, microwave  vs. γ-ray, and X-ray vs. γ-ray) still have a large scatter (see Figs. 13−15). This is somewhat surprising, as positive correlations between the radio and γ-rays have been reported (e.g. Kovalev et al. 2009; León-Tavares et al. 2010; Giroletti et al. 2010; Ghirlanda et al. 2010; Mahony et al. 2010; Peel et al. 2011; Linford et al. 2011; Ackermann et al. 2011). The difference might be due to the different synchrotron peak energies of the objects in the samples. Even in simple SSC scenarios, this introduces a scatter in the correlation between the fluxes (e.g., for the same radio flux, an object with higher is expected to produce more γ-rays than one with a smaller ). The large scatter present in Fig. 14 could also imply that γ-ray  emission is due to components not always directly related to radiation in other energy bands, e.g., multiple SSC components (see also Abdo et al. 2010a). A good correlation is however present between the X-ray and γ-ray spectral slopes (see Fig. 16).

We confirm the correlations between the Fermi-LAT spectral index and the SED peak energies and found by Abdo et al. (2010a). As an illustration of the agreement, Fig. 19 plots the Fermi-LAT spectral slope estimated using the full 27-month data set as a function of . The gray points represent the γ-ray spectral slopes estimated using the quasi-simultaneous two-month dataset. These points, plotted without the much larger error bars to avoid confusion, clearly cluster around the 27-month data, confirming the correlation. The solid line represents the best-fit obtained by Abdo et al. (2010a).

8.7. Comparison with the expectation of simple SSC models

As discussed in Abdo et al. (2010a), simple SSC models predict that in the Thomson regime the peak frequency of the synchrotron () and inverse Compton () components are related by (3)where is the Lorentz factor of the electrons radiating at the peak energy. This is related to the observed peak frequency of the observed photon spectrum by (4)where is the synchrotron peak frequency in the rest-frame of the emitting region, B is the magnetic field, and δ is the usual Doppler factor (e.g. Urry & Padovani 1995).

In objects where is higher than ≈ 10¹⁵Hz, the Thomson approximation is no longer valid and the inverse Compton scattering occurs under the Klein-Nishina (KN) regime. Using Monte Carlo simulations, Abdo et al. (2010a) estimated the area covered by SSC models in the plane log()–log(). This area is delimited by the solid contour lines shown in Fig. 20 where we plot the ) of our sources, calculated from the observed values of and using Eq. (3), versus the rest-frame log().

As in the case of the bright γ-ray blazars considered by Abdo et al. (2010a), only a few objects are inside or close to the SSC area, implying that simple SSC models cannot explain the SED of many blazars in our samples. This conclusion is supported by the lack of a strong correlation between radio and γ-ray fluxes.

However, the of blazars that were not detected by Fermi-LAT and for which we could only infer a limit to (from 30% to 40% of the FSRQs in the radio and X-ray selected samples; see Fig. 10 for one example) are plotted as upper limits in Fig. 20; many of these limits are close to or inside the SSC area and therefore the SEDs of these objects are likely consistent with simple SSC emission.

8.8. The Compton dominance of blazars

The Compton dominance (CD, defined as the ratio of the inverse Compton to synchrotron peak luminosities) is a crucial parameter for the study of blazar physics, as it is strictly related to the location of the maximum power output in the energy spectrum of a blazar.

Figure 21 plots the CD values, estimated from our SEDs, as a function of , showing that log (CD) ranges from about −0.5 to about 2. The larger values are always associated with LSP objects, while HSP sources always have values of log (CD) lower than ≈ 0.5. In this figure, the two blazar subclasses appear to be quite different, with BL Lacs having significantly smaller CD values, even when their values are equal to those of FSRQs. To better understand this difference, in Fig. 22 we plot the CD distributions of FSRQs and BL Lacs for different samples and intervals.

The FSRQs included in the Fermi-LAT sample, which are γ-ray bright by definition, show a CD distribution peaking at large values. We note that in this sample we also applied a radio flux-density limit of 1Jy, hence the sources below the radio cut must be on average more Compton-dominated than those in our sample. This implies that the distribution of purely γ-ray selected blazars must be even more strongly peaked at high CD values than that of the Fermi-LAT sample. Considering instead FSRQs selected in the radio and the X-ray bands, we get quite a different picture, with a broader distribution extending to values of less than 1. Moreover, about 30–45% of FSRQs in the radio, soft X-ray, and hard X-ray selected samples are not detected by Fermi-LAT and therefore they must populate the part of the CD distribution with low CD values. This is shown by the dotted red histogram, which also includes upper limits to the CD estimated as the ratio of the upper limit to F() and F() where limits to and F() are obtained by fitting the X-ray data together with the 27-month Fermi-LAT upper limits as shown in Fig. 10.

8.9. The blazar sequence

The top panel of Fig. 23 plots the logarithm of the bolometric power, represented by the sum of the synchrotron and inverse Compton peak luminosities [] as a function of log() for all sources in the four samples considered in this paper and Planck Collaboration (2011e) for which an estimate of and the bolometric luminosity was possible. We use this plot to test the relationship known as the Blazar Sequence, which is the strong anti-correlation between bolometric luminosity and claimed by Fossati et al. (1998) and Ghisellini et al. (1998) that remains a subject of lively debate (e.g., Giommi et al. 1999; Padovani et al. 2003; Caccianiga & Marchã 2004; Nieppola et al. 2006; Padovani 2007; Ghisellini & Tavecchio 2008).

We stress that to robustly test for the existence of such a relationship it is mandatory to use samples that are unbiased, that is selected in such a way that no particular part of the L_Bol– diagram is more likely than others to be selected. Although our samples are statistically well-defined, they are not unbiased from this viewpoint for the following reasons:

• a)

  as shown in Figs. 17 and 18, the distribution of strongly depends on the selection method; hence, for the various samples, we obtain different samplings of the parameter ;

• the area above the dashed line, which represents the luminosity above which the non-thermal emission of a blazar completely dominates the observed optical flux¹², cannot be populated by blazars with no emission lines, such as BL Lacs, as in this case they would appear completely featureless and therefore their redshift could not be measured. In this respect, we note that over 40% of the BL Lacs in the BZCAT catalog, and an even larger fraction of the Fermi-LAT detected BL Lacs, still lack any redshift measurement (Massaro et al. 2010; Shaw et al. 2009, 2010). All the BL Lacs in our samples for which redshifts are unknown are plotted in Fig. 23 as lower limits estimated assuming that their non-thermal light is ten times brighter than the optical light of the host galaxy. Some BL Lacs of known luminosity (red points in Fig. 23) are above the red line because their redshift was measured from emission lines with rest-frame equivalent widths below the 5 Å limit; they might be objects with properties in-between those of BL Lacs and FSRQs (see. e.g., Ghisellini et al. 2011);

  Fig. 35 The SED of BZQJ1507+0415 (J1507+0415, top left), 4C−05.64 (J1510−0543, top right), APLib (J1517−2422, middle left), PG1553+113 (J1555+1111, middle right), WE1601+16W3 (J1603+1553, bottom left), and OS−237.8 (J1625−2527, bottom right). Simultaneous data are shown in red; quasi-simultaneous data, i.e. Fermi data integrated over 2 months, Planck ERCSC and non-simultaneous ground based observations are shown in green; Fermi data integrated over 27 months are shown in blue; literature or archival data are shown in light gray.

  Fig. 36 The SED of 4C38.41 (J1635+3808, top left), NRAO512 (J1640+3946, top right), 3C345 (J1642+3948, middle left), Mkn501 (J1653+3945, middle right), ARP102B (J1719+4858, bottom left), and 1ES1741+196 (J1743+1935, bottom right). Simultaneous data are shown in red; quasi-simultaneous data, i.e. Fermi data integrated over 2 months, Planck ERCSC and non-simultaneous ground based observations are shown in green; Fermi data integrated over 27 months are shown in blue; literature or archival data are shown in light gray.

  Fig. 37 The SED of OT081 (J1751+0939, top left), S51803+784 (J1800+782, top right), PKSB1830−210 (J1833−2103, middle left), PKS1833−77 (J1840−7709, middle right), 2E1908.2−201 (J1911−2006, bottom left), and PMNJ1923−2104 (J1923−2104, bottom right). Simultaneous data are shown in red; quasi-simultaneous data, i.e. Fermi data integrated over 2 months, Planck ERCSC and non-simultaneous ground based observations are shown in green; Fermi data integrated over 27 months are shown in blue; literature or archival data are shown in light gray.

  Fig. 38 The SED of OV−236 (J1924−2914, top left), 1ES1959+650 (J1959+6508, top right), 1 Jy2005−489 (J2009−4849, middle left), PKS2052−47 (J2056−4714, middle right), 1 Jy2126−158 (J2129−1538, bottom left), and S32141+17 (J2143+1743, bottom right). Simultaneous data are shown in red; quasi-simultaneous data, i.e. Fermi data integrated over 2 months, Planck ERCSC and non-simultaneous ground based observations are shown in green; Fermi data integrated over 27 months are shown in blue; literature or archival data are shown in light gray.

  Fig. 39 The SED of 1Jy2144+092 (J2147+0929, top left), 4C06.69 (J2148+0657, top right), PKS2149−307 (J2151−3027, middle left), BLLac (J2202+4216, middle right), 4C31.63 (J2203+3145, botton left), and PKS2204−54 (J2207−5346, botton right). Simultaneous data are shown in red; quasi-simultaneous data, i.e. Fermi data integrated over 2 months, Planck ERCSC and non-simultaneous ground based observations are shown in green; Fermi data integrated over 27 months are shown in blue; literature or archival data are shown in light gray.

• c)

  the different radio flux-density cuts applied to our samples imply that different subsamples probe different parts of the radio luminosity function. The radio and γ-ray selected samples, which are defined using a high radio flux limit (S ≥ 1Jy), probe the high-luminosity end of the radio luminosity function, which, for = 10¹³Hz and the observed redshift and CD distributions, translates into L_Bol ≳ 10⁴⁶ergs^-1. The soft and hard X-ray selected samples have a radio flux density cut of S = 0.1–0.2Jy, or about one order of magnitude fainter than that of the radio and Fermi-LAT samples, hence might include significantly less powerful sources, as faint as bolometric luminosities of the order of 10⁴⁴ergs^-1.

If we remove the dependence of the selection method by considering only FSRQs (which have the same distributions as all samples), we see that the luminosity values span five orders of magnitude from ~10⁴⁴ to ~10⁴⁹ ergs^-1 and show no trend with , which only ranges between ~10^12.5 and ~10¹⁴Hz. No obvious correlation is present in each sample separately or in the union of the four samples. The L-shaped distribution that is apparent in Fig. 23, if lower limits are ignored, is similar to that found by Meyer et al. (2011) who estimated both the and peak luminosities of a large sample of blazars using non-simultaneous multi-frequency data. These authors, however, instead of considering lower limits to the peak luminosities of blazars with unknown redshifts, assumed a luminosity corresponding to the redshift that the host galaxy would have for the observed blazar optical magnitude. Meyer et al. (2011) argued that the strong correlation predicted by the blazar sequence turns into ablazar envelope when partly misaligned blazars are included in the samples. However, Giommi et al. (2012), by means of detailed Monte Carlo simulations, showed that this envelope, or L-shaped distribution, is expected when blazars with no redshift measurements are not properly taken into account.

Finally, we consider the subsample of sources that satisfy the same conditions as in Fossati et al. (1998), that is S_5GHz > 2Jy for FSRQs, S_5GHz > 1Jy for radio-selected BL Lacs, no restrictions for X-ray selected BL Lacs, and the exclusion of all BL Lacs with no redshift information. This case is illustrated in Fig. 23 (bottom panel), which shows a trend very similar to that presented in Fossati et al. (1998).

Taking into account all of the above, we conclude that our data do not show a correlation of the type predicted by Fossati et al. (1998), owing to the presence of low luminosity LSP objects and the difficulty in measuring the redshifts of likely high-luminosity HSP sources. That such a correlation becomes evident when the objects are selected by the criteria of Fossati et al. (1998) supports the hypothesis that the correlation might be the result of a selection effect.

8.10. Selection effects and sample composition

Our decision to select flux-limited samples for widely different parts of the electromagnetic spectrum (radio, soft X-ray, hard X-ray, and γ-ray) has allowed us to demonstrate the strong selection biases that can affect important physical parameters, such as the peak energy of both the synchrotron and inverse Compton components (see Figs. 17 and 18) and the Compton dominance (see Fig. 22). Since FSRQs and BL Lacs have significantly different distributions, these selection biases also strongly affect the composition of the samples in terms of the relative abundances of blazar subclasses (FSRQs vs. BL Lacs, LSPs vs. HSPs), as is apparent from Table 4.

Radio-selected samples include sources that are bright in the radio band. If there is no correlation between radio flux/luminosity and other parameters such as and Compton dominance, this is the best selection for measuring the distributions of these important physical parameters. If instead there is a strong correlation between radio luminosity and , then the distribution of should strongly depend on the radio flux limit.

X-ray selection favors high (and consequently high ) sources, which are much brighter at X-ray frequencies than low for the same radio flux. X-ray flux-limited samples are therefore much richer in high BL Lacs (HBLs or HSP sources) than radio-selected samples. This selection effect has been known since the first soft X-ray surveys became available.

Selection in the γ-ray band favors bright γ-ray objects and therefore highly Compton-dominated sources. Fermi-LAT TS-limited samples contain more sources with flat γ-ray spectral slopes or high sources. This explains the overabundance of HSP blazars (only BL Lacs) and high CD blazars (only FSRQs) in Fermi-LAT catalogs.

---

Acknowledgments

We thank the entire Swift team for the help and support and especially the Science Planners and Duty Scientists for their invaluable help and professional support with the planning and execution of a large number of ToOs. The Planck Collaboration acknowledges the support of: ESA; CNES and CNRS/INSU-IN2P3-INP (France); ASI, CNR, and INAF (Italy); NASA and DoE (USA); STFC and UKSA (UK); CSIC, MICINN and JA (Spain); Tekes, AoF and CSC (Finland); DLR and MPG (Germany); CSA (Canada); DTU Space (Denmark); SER/SSO (Switzerland); RCN (Norway); SFI (Ireland); FCT/MCTES (Portugal); and DEISA (EU). A full description of the Planck Collaboration and a list of its members, indicating which technical or scientific activities they have been involved in, can be found at http://www.rssd.esa.int/Planck. We thank the Planck team and in particular the members of the Data Processing Centers for their support in the reduction of LFI and HFI data carried out specifically for this work. The Fermi-LAT Collaboration acknowledges generous ongoing support from a number of agencies and institutes that have supported both the development and the operation of the LAT as well as scientific data analysis. These include the National Aeronautics and Space Administration and the Department of Energy in the United States, the Commissariat à l’Énergie Atomique and the Centre National de la Recherche Scientifique/Institut National de Physique Nucléaire et de Physique des Particules in France, the Agenzia Spaziale Italiana (ASI) and the Istituto Nazionale di Fisica Nucleare (INFN) in Italy, the Ministry of Education, Culture, Sports, Science and Technology (MEXT), High Energy Accelerator Research Organization (KEK) and Japan Aerospace Exploration Agency (JAXA) in Japan, and the K. A. Wallenberg Foundation, the Swedish Research Council and the Swedish National Space Board in Sweden. Additional support for science analysis during the operations phase from the Istituto Nazionale di Astrofisica in Italy and the Centre National d’Études Spatiales in France is gratefully acknowledged. The Metsähovi team acknowledges the support from the Academy of Finland to our projects (numbers 212656, 210338, 121148, and others). This work was also supported by grants 127740 and 122352 of the Academy of Finland. UMRAO is supported by a series of grants from the NSF and NASA, and by the University of Michigan. This publication is partly based on data acquired with the Atacama Pathfinder Experiment (APEX). APEX is a collaboration between the Max-Planck-Institut für Radioastronomie, the European Southern Observatory, and the Onsala Space Observatory. This research is partly based on observations with the 100-m telescope of the MPIfR (Max-Planck-Institut für Radioastronomie) at Effelsberg, the IRAM 30-m telescope, and the Medicina (Noto) telescope operated by INAF – Istituto di Radioastronomia. J. Wu and X. Zhou are supported by the Chinese National Natural Science Foundation grants 10633020, 10778714, and 11073032, and by the National Basic Research Program of China (973 Program) No. 2007CB815403. The OVRO 40-m monitoring program is supported in part by NASA grants NNX08AW31G and NNG06GG1G and NSF grant AST-0808050. O. G. King acknowledges the support of a Keck Institute for Space Studies Fellowship. W. Max-Moerbeck acknowledges support from a Fulbright-CONICYT scholarship. V. Pavlidou acknowledges support provided by NASA through Einstein Postdoctoral Fellowship grant number PF8-90060 awarded by the Chandra X-ray Center, which is operated by the Smithsonian Astrophysical Observatory for NASA under contract NAS8-03060. The Australia Telescope is funded by the Commonwealth of Australia for operation as a National Facility managed by CSIRO. This paper makes use of observations obtained at the Very Large Array (VLA) which is an instrument of the National Radio Astronomy Observatory (NRAO). The NRAO is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc. We acknowledge the use of data and software facilities from the ASI Science Data Center (ASDC), managed by the ItalianSpace Agency (ASI). Part of this work is based on archival data and on bibliographic information obtained from the NASA/IPAC Extragalactic Database (NED) and from the Astrophysics Data System (ADS). We thank the anonymous referee for his/her useful and constructive comments.

References

All Tables

All Figures