A Polarisation Survey of Bright Extragalactic AT20G Sources
M. Massardi, S. G. Burke-Spolaor, T. Murphy, R. Ricci, M. Lopez-Caniego, M. Negrello, R. Chhetri, G. De Zotti, R. D. Ekers, R. B. Partridge, E. M. Sadler
aa r X i v : . [ a s t r o - ph . C O ] S e p Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed 13 July 2018 (MN L A TEX style file v2.2)
A Polarisation Survey of Bright ExtragalacticAT20G Sources
M. Massardi ⋆ , S. G. Burke-Spolaor , T. Murphy , , R. Ricci , M. L´opez-Caniego ,M. Negrello , R. Chhetri , , G. De Zotti , , R. D. Ekers , R. B. Partridge ,E. M. Sadler INAF - Istituto di Radioastronomia, Via Gobetti 101, I-40129, Bologna, Italy Australia Telescope National Facility, CSIRO, P.O. Box 76, Epping, NSW 1710, Australia School of Physics, University of Sydney, NSW 2006, Australia School of Information Technologies, University of Sydney, NSW 2006, Australia Instituto de F´ısica de Cantabria (CSIC-UC), Avda. los Castros s/n, E-39005 Santander, Spain INAF - Osservatorio Astronomico di Padova, vicolo dell’Osservatorio 5, I-35122, Padova, Italy Department of Astrophysics and Optics, School of Physics, University of New South Wales, NSW, 2052, Australia SISSA, Via Bonomea 265, I-34136 Trieste, Italy Haverford College Astronomy Department, 370 Lancaster Avenue, Haverford, PA, 19041 USA
13 July 2018
ABSTRACT
We present polarisation data for 180 extragalactic sources extracted from the AustraliaTelescope 20 GHz (AT20G) survey catalog, and observed with the Australia TelescopeCompact Array during a dedicated, high sensitivity run ( σ P ∼ ≃
99% complete sample of extragalactic sources brighter than S = 500 mJy atthe selection epoch with declination δ < − ◦ . The sample has a 91 .
4% detection ratein polarisation at ∼
20 GHz (94% if considering the sub-sample of point like sources).We have measurements also at 4.8 and 8.6 GHz within ∼ ∼
20 GHz.
Key words: surveys – galaxies: active – polarization – radio continuum: galaxies –techniques: polarimetric.
The study of the properties of radio source populationsabove 10 GHz has progressed greatly in recent years, fosteredby Cosmic Microwave Background (CMB) observation cam-paigns that require an accurate understanding of the con-tamination of the CMB signal by foreground sources. Ex- ⋆ Email: [email protected] tragalactic radio sources are the dominant contaminant onangular scales smaller than 30 arcmin, both in total intensityand in polarisation at frequencies of up to ≃ c (cid:13) Massardi et al.
Table 1.
AT20G and WMAP related catalogues that include data in polarisation.
References Frequency(GHz)
Massardi et al. (2008) AT20G-BSS 4.8, 8.6, 20 320 AT20G bright sampleL´opez-Caniego et al. (2009) 23,33,41 22 polarisation detection in WMAP mapJackson et al. (2010), Battye et al. (2011) 8.4, 22, 43 230 WMAP sourcesSajina et al. (2011) 4.8, 8.4, 22, 43 159 equatorial AT20G sourcesMurphy et al. (2010), Massardi et al. (2011a) AT20G 4.8, 8.6, 20 5890 AT20G 91% complete sample above 100 mJyBurke-Spolaor et al. (2009) 18 9 extended sources in the Southern hemispherecurrent paper 4.8, 8.6, 18 193 complete sample above 500 mJy erties of radio sources to high frequencies is interesting pers´e , as it provides information about the physics of the emis-sion process. In compact, Doppler boosted sources that dom-inate the high-frequency population at bright flux densitylevels, the emission at higher and higher frequencies mostlyarises from synchrotron, self-absorbed, knot-like structuresin the relativistic jet closer and closer to the active nucleus(e.g. Blandford & K¨onigl 1979). It has been argued thatthe ordering of magnetic fields should increase in the in-ner regions, and as a consequence, the polarisation degreeincreases (Tucci et al. 2004).However, the polarisation properties of high-frequencyextragalactic contaminants are still poorly constrained byobservations. Most current estimates rely on extrapolationsfrom low-frequency samples; the NVSS at 1.4 GHz (Condonet al. 1998) still constitutes the largest sample of sourcessurveyed both in total intensity and polarisation. Extrapo-lations are affected by large uncertainties since a complexcombination of effects must be considered. This includesintra-beam effects and bandwidth depolarisation, in addi-tion to intrinsic frequency-dependent changes. The propa-gation of the radiation through diffuse plasma screens be-tween the source and the observer can cause depolarisationand rotation of the polarisation angle. These effects are diffi-cult to isolate observationally, although we can benefit fromthe inverse square frequency dependence of the latter effect.Because the polarised signal in extragalactic objects istypically a few percent of the total intensity signal, deep sur-veys are necessary to collect statistically significant samples.But high-frequency, deep surveys are time-consuming fordiffraction-limited, ground-based telescopes. This has mo-tivated sensitive high frequency polarisation measurementsof source samples usually selected from surveys at < ∼ >
20 GHz has become possible thanks to theWMAP all-sky survey with a completeness limit of ≃ − ◦ (north of − ◦ the catalogue completeness islower between 14 and 20 h in right ascension). Table 1 listssome details on the AT20G and WMAP related samplesthat include data in polarisation. Larger samples of radio sources selected at higher frequencies are being provided bythe Planck mission (Planck 2011a,b, 2013).Multi-steradian samples of high-frequency selected po-larised sources are also important for identifying suitablecalibrators for CMB polarimetric experiments and upcomingmillimeter-wave telescopes. Pictor A has been identified as asuitable extragalactic polarisation calibrator for the
Planck
Low Frequency Instrument (LFI) because of its position (inthe region of the Ecliptic Pole where the satellite scans ∼ once per minute) and its lack of variability from the hotspots that dominate the polarised signal (Burke-Spolaor etal. 2009).The polarisation of WMAP sources has been investi-gated by L´opez-Caniego et al. (2009) using WMAP data;14 extragalactic sources were significantly detected in polar-isation. Follow-up observations of a complete sample of 203WMAP sources were carried out with the VLA by Jacksonet al. (2010); polarised emission was detected for 123, 169and 167 at 8.4, 22 and 43 GHz respectively.Sadler et al. (2006) presented polarisation measure-ments for a sample of 173 AT20G sources brighter than S = 100 mJy; 129 ( ≃ S >
500 mJy), finding 213 > σ polar-isation detections at 20 GHz out of a total of 320 sources( ≃ .
5% at20 GHz. The spectral indices in total intensity and in po-larisation were found to be similar on average, but therewere several sources for which the spectral shape of the po-larised emission is substantially different from the spectralshape in total intensity. The full AT20G catalog (Murphyet al. 2010; Massardi et al. 2011a) includes the 20 GHz po-larised intensity for 768 sources, 467 of which also have si-multaneous polarisation detections at 5 and/or 8 GHz, outof a total of 5890 sources. The detection limit is defined asmax(3 σ, . S , ≃
200 AT20G radio galaxies with S >
40 mJyin an equatorial field of the Atacama Cosmology Telescopesurvey; polarised flux was detected at >
95% confidencelevel for 141, 146, 89, and 59 sources, from low to high fre-quencies. The measured polarisation fractions are typically < ≃ c (cid:13) , 000–000 olarisation of bright AT20G sources The conclusions of all the polarisation studies of com-plete samples selected at high frequencies are limited by themoderate detection rates. To overcome this limitation wehave performed dedicated high-sensitivity polarisation ob-servations of a complete AT20G bright source sub-sampleachieving a ≃
93% detection rate. In addition to allowinga more thorough investigation of the polarisation propertiesof the high-frequency radio source populations, this sam-ple constitutes a legacy data set for polarisation studies inthe Southern hemisphere. In particular, this sample couldhelp in the definition of calibrator source lists for facilitiesworking in the millimetric bands, like the Atacama LargeMillimeter Array (ALMA) that, in its final configuration,will observe from 30 to 950 GHz in both total intensity andpolarisation.This paper is structured as follows. In § § §
4. The final sample thus includes 187 sources. In § § § § The selection of the sample was based on the list of con-firmed AT20G sources available at the epoch of our obser-vations (October 2006). We selected all objects with fluxdensity S >
500 mJy and declination δ < − ◦ , ex-cluding the Galactic plane region ( | b | . ◦ ) and the LargeMagellanic Cloud Region (inside a circle of 5 . ◦ radius cen-tered at α = 05 : 23 : 34 . δ = −
69 : 45 : 22). Thisresulted in a complete sample of 189 sources.Nine of them were found, with the aid of low frequencyradio imaging surveys (PMN, Griffith & Wright 1994, 1995,and SUMSS, Mauch et al. 2003), to be very extended. Thesewere observed with the mosaic mode at 20 GHz by Burke-Spolaor et al. (2009). Flux densities integrated over thewhole source are available for 5 of them. For the remaining4 objects the measured integrated flux densities, if available,refer to subregions. For these 4 objects we decided to extractthe integrated flux densities from the low resolution WMAPmaps as described in Sect. 4.It should be noted that the flux densities reported in thefinal published AT20G catalogue may be slightly differentfrom those given in the preliminary 2006 version used forour source selection. The reason is that in the case of sourcesobserved more than once (which is most likely the case forbright sources such as those considered here) the highestquality observation was listed in the final AT20G catalogue,as discussed by Murphy et al. (2010). Because of variabilitythis has the effect of moving some sources above or below theadopted flux density threshold ( S = 500 mJy). Sincethere are more sources below than above the threshold, thereare more sources moving up than moving down, and we endup with slightly more sources above threshold in the final catalogue than in the version we have used. Hence thereare 214 sources listed in the final AT20G catalogue withdeclination δ < − ◦ and flux density above our chosenthreshold.On the other hand, of the 180 sources with good qualityflux density in our ATCA observations (as discussed below)only 165 still have S >
500 mJy in the final catalogueand only 145 were found to be above this threshold in ourOctober 2006 observations. However, despite variability, thefinal sample, which is reasonably complete at the selectionepoch, is representative of the bright 20 GHz population asa whole and can be used to assess statistical properties ofthis population.
Observations were taken on October 1, 2006 using the mostcompact hybrid configuration of ATCA, “H75”, excludingthe data from the farthest antenna. The longest baseline ofthis configuration is 75 m, and its T-shape ensures adequateFourier coverage for snapshots taken on a relatively smallrange of hour angles and at high elevation.Although the linear feeds of the ATCA somewhat com-plicate the polarisation calibration procedure, the array hasseveral inherent advantages for polarisation experiments.The on-axis receivers of the telescope introduce relativelylow amounts of instrumental polarisation, while all anten-nae are fitted with a noise diode that injects a signal tocontinually track the phase difference (xy-phase) betweenthe two orthogonal feeds. In addition, since the feeds arelinearly polarised, there is very little contamination of thecircular polarisation signal by the total intensity signal. Forfurther details on the ATCA instrumental polarisation werefer readers to Sault, Reyner & Kesteven (2002).A digital correlator (later replaced by the CABB broad-band digital correlator) allowed simultaneous observationsin two bands, each with a bandwidth of 128 MHz dividedinto 32 channels. The observation frequencies within the20 GHz band, covering the 16-24 GHz range, were chosen soas to maximize sensitivity, make use of optimal system tem-peratures, and avoid correlator harmonics. The frequencybands were centered at 16.704 and 19.392 GHz. After cal-ibration, data were averaged over the 256 MHz band sothat our mean effective frequency was 18.048 GHz (hereaftertagged as “18”GHz observations). The FOV was ∼ . − ◦ . Because the effects ofshadowing and cross-talk on polarisation measurements areunknown for the instrument, the sample for H75 observa-tions was restricted to − ◦ < δ < − ◦ . Eleven δ < − ◦ sources were observed for this project during a run that tookplace on 17th October 2006 with a more extended hybrid an-tenna configuration (H214) and observation bands centeredat 18.752 and 21.056 GHz. Calibration and observation setupfor these objects matched that of AT20G follow-up obser-vations (see Murphy et al. 2010 for details). After the datareduction, the data for 3 of these objects (AT20GJ115253-834410, AT20GJ122454-831310, AT20GJ155059-825807) at20 GHz were found to be of poor quality in this run. How-ever, they had good quality observations and a polarisation c (cid:13) , 000–000 Massardi et al.
Table 2.
Catalogue of the first 10 sources of the extragalactic sample observed in the AT20G run dedicated to high sensitivity polarisation.The full source list, available as Supplementary Material online, includes 187 objects. Columns are: (1) AT20G name; (2) 18 GHz scalaraverage flux density (error given by eq. (3)); (3–5) polarised flux density, fractional polarisation, and polarisation angle at 18 GHz; (6–9)total intensity, polarised flux density, fractional polarisation and polarisation angle at 8.6 GHz; (10–13) total intensity and polarisedflux densities, polarisation degree and polarisation angle at 4.8 GHz; (14) ratio between the visibilities amplitude averaged over the longbaselines ( ∼ . ∼ . Name ‘AT20G’ S P Π φ S . P . Π . φ . S . P . Π . φ . ± ± ± -78.70 929.0 ± ± ± -51.50 865.0 ± ± ± -37.04 0.98 ...J001035-302748 595.8 ± ± ± -46.66 676.0 ± ± ± -38.97 582.0 ± ± ± -52.22 0.96 c..J001259-395426 1030.0 ± ± ± -86.42 1622.0 ± ± ± -85.56 1651.0 ± ± ± -88.13 0.96 ...J002616-351249 1136.0 ± ± ± -54.21 382.0 ± < ± < ± ± ± ± ± ± ± ± ± ± ± ± -2.21 950.0 ± ± ± ± ± ± ± ± ± -23.19 709.0 ± ± ± -72.10 573.0 ± < ± ± ± -8.26 3680.0 ± ± ± ± ± ± -62.00 ... ...J010915-604948 375.0 ± ± ± -14.55 516.0 ± ± ± -49.53 533.0 ± ± ± ± ± ± -11.21 298.0 ± ± ± -15.13 220.0 ± ± ± -31.11 0.99 ... detection in March 2006 during a previous AT20G follow-up run. Hence, we decided to include these measurements inthe present analysis, flagging them in the catalogue.For each run a bright point-like source was observed tocalculate bandpass solutions and PKS 1934–638 was used asthe primary flux density calibrator. The sample was brokeninto groups of 4–6 sources, located in the same sky region.Before each group was observed, an antenna pointing correc-tion was performed using a nearby bright source to maintaindirectional accuracy despite the vast range of sky positionsthat the telescope had to observe in a short period of time.Each target was observed in two 70-second snapshots sepa-rated by four hours.Five targets were identified as extended within the pri-mary beam, according to the extendedness criteria describedin Murphy et al. (2010). As the flux density estimation meth-ods that we use (see Sect.3) are well suited for compactobjects, the flux densities of extended sources are likely un-derestimated by an unknown amount and the polarised fluxdensity is potentially wrong. For this reason they were in-cluded in the list of extended sources for which we extractedthe polarisation information from the WMAP 9-yr coaddedmaps (see Sect. 4). In the end, after dropping 4 of the 9 veryextended sources observed by Burke-Spolaor et al. (2009)and the 5 sources found to be extended, we are left with 180sources for which we have good quality ATCA data.Five antennae were available in our 16.704 GHz band,while there was an error in the 19.392 GHz band that leftonly four usable antennae at that frequency. To save ob-serving time, no secondary calibrators were interleaved withthe target observations, with the intention of self-calibratingeach bright source during the data reduction. As described inSect. 3, this turned out to be an unwise strategy in terms ofpolarisation calibration. Nevertheless, 169 of the 180 extra-galactic sources in our ATCA observed sample (94%) weredetected in polarisation at 18 GHz. However, only 145 ofthem also have S >
500 mJy in the October 2006 ob-servations.Table 2 provides an excerpt of the catalogue containing the data for the sample of 180 objects for which we havegood quality ATCA data.
Lower-frequency observations were performed during a regu-lar AT20G follow-up run (described in Murphy et al. 2010),carried out in November 2006. In that epoch, we used a 5-antenna east-west array configuration, with a 1.5 km longestbaseline. Two 30-second observations per source providedsimultaneous 4800 and 8640 MHz snapshots. These observa-tions had 128 MHz of bandwidth per frequency. The FOVwere of 9.9 and 5.5 arcmin, respectively.Massardi et al. (2011b) demonstrated that the medianvariability of total intensity flux densities for sources in fluxdensity ranges similar to those of our sample over 3 monthtimescale is 3.5 and 6.3 per cent respectively at ∼ ∼ S >
500 mJy in the October 2006 run. 143 sourceshave a polarisation detection at all the frequencies (79% ofthe main sample), 119 of which have S >
500 mJy inthe October 2006 run.
The 16.7 and 19.4 GHz data were reduced using the
Miriad software package (Sault, Teuben, & Wright 1995). The twofrequencies were calibrated independently and then com-bined for 18 GHz imaging and flux density assessment. Opac- c (cid:13) , 000–000 olarisation of bright AT20G sources ity correction and a correction for the time-dependent in-strumental xy-phase difference was applied upon loading alldata into Miriad . After this correction, a small residual gainoffset still remained to be corrected in the following calibra-tion stages. Bandpass solutions and primary flux densitycalibration were calculated and applied using PKS 1921–293and PKS 1934–638, respectively.For polarimetric calibration with calibrators of un-known polarisation and sparse data (such as in our shortsnapshot observations), the standard
Miriad procedure sug-gests calculating the largely stable instrumental leakageterms by using an unpolarised primary calibrator. The re-maining polarisation and gain terms are then calculated foreach secondary phase calibrator.Roughly 75% of the sources in our sample are regis-tered in the ATCA calibrator database; all are sufficientlybright to determine adequate calibration solutions. However,though it is suggested that accurate Q and U values couldbe calculated from a relatively small amount of data, it wasapparent that this was not the case for our ∼ ◦ from any target source and were restricted to highelevation observations.The 4.8 and 8.6 GHz data for the main sample and the20 GHz data for the 11 δ < − ◦ sources had observationalmodes exactly coincident with the AT20G survey follow-up, and thus were flagged and reduced using the automaticpipeline developed for the AT20G (Murphy et al. 2010). Stokes-I intensities were determined from the visibili-ties to avoid the inclusion of phase instabilities inherent inimage-based measurements. This technique takes the scalaraverage of the visibility amplitudes and is robust for bright( >
200 mJy), point-like sources only.To acquire Stokes Q , U , and V flux densities, im-ages were created and deconvolved using the Miriad taskCLEAN. To correct the Stokes Q , U , and V images for decor-relation, we took advantage of the fact that Stokes parame-ters, simultaneously measured, are affected by decorrelationoriginating in atmospheric phase instabilities (as might beleft after imperfect calibration). We can thus use the frac-tional level of residual decorrelation ( χ ) in Stokes I , calcu-lated and applied to Q , U , and V flux densities as: χ = I sca I map (1) Z = χ · Z map , (2)where Z represents Stokes Q , U or V , I sca is the scalar-averaged Stokes I flux density, and I map and Z map representthe values at the position of the peak Stokes I emission in therelevant image. The image peak for all sources was sufficientto determine the decorrelated flux density measurements;the pixel size was typically 10 arcsec, and no sources in thissubsample were extended significantly beyond this.The polarised intensity and the position angles werethen calculated using standard first-order debiasing, where P = p Q + U − σ (Wardle & Kronberg 1974; Simmons& Stewart 1985). The last term, σ V , is the RMS noise in theStokes V image. Most extragalactic sources do not have sig-nificant levels of circular polarisation; therefore the Stokes V signal is usually undetected or very low for all such sources.This point makes the noise level in Stokes V a reasonable es-timate of the background noise level in the Q and U images,and thus gives a good estimate for the debiasing correction.The polarisation angles are given by φ =(1 /
2) arctan (
U/Q ), and the linear polarisation degree,Π, is given in the percentage: Π = 100 · P/I . The RMS scatter σ V provided a measurement of the noiseon the scalar average flux density of the order of 1-2%.Telescope pointing inaccuracies, considering the ∼
15 arc-sec pointing errors and the 18 GHz primary beam responsefunction are expected to cause a possible attenuation of upto 2% in all Stokes parameters.Errors in the primary flux density scaling fromPKS 1934–638 are estimated by comparing the onlineATCA calibrator catalogue data at 18.496 GHz to the high-frequency polynomial model used by
Miriad to calculate thescaling factor. The model predicts I model = 1 . I data = 1 . I sca , we ignore this calibration error. Thenet error in total intensity is thus given by: σ I = (0 . I sca ) + σ . (3)Errors in the Stokes parameters Q , U , and V can arisefrom leakage of the much brighter Stokes I signal due to c (cid:13) , 000–000 Massardi et al.
Table 3.
Polarised and total flux densities measurements (in mJy) for the 3 extended sources with a polarisation detection in the WMAP9-year co-added full-sky maps.
Name P P P P S S S S Fornax A (RA:03:22:41.7; DEC:-37:12:30) 1074 <
354 9321
PicA - AT20GJ051949-454643 457 CenA - AT20GJ132527-430104 3322 imperfections in the alignment of orthogonal receiver com-ponents. The correction of this effect using PKS 1934–634and an iterative leakage calculation using secondary cali-brators results in a negligible error compared to the systemnoise. The noise term is calculated by propagation of er-rors through eqs. (1) and (2). The main contributions tothe global error come from the antenna pointing inaccuracyand from the noise estimated by the rms levels in off-sourceregions in the restored image: σ Z = ( I sca σ Z map I map ) + ( ZI sca σ I map I map ) + (0 . Z ) , (4)where Z is either Q or U . The error on the polarised intensitycan then be derived as: σ P = Q σ Q + U σ U Q + U , (5)where Q and U are calculated from eq. (2). Note that if σ Q ≃ σ U (as expected for low polarisation sources in noisymaps), then σ P = σ Q = σ U . We defined as a non-detectionof a source P < σ P , and used 3 σ P as the upper limit onthe polarised flux density for such sources. The errors in thefractional polarisation were obtained with the usual errorpropagation from σ P and σ I . As mentioned above, our ATCA data on extended sourcesare largely incomplete. Of the 9 very extended sources with S > . , whichhas been used in the past to extract flux densities fromWMAP (Massardi et al. 2009) and Planck data (Planck http://max.ifca.unican.es/IFCAMEX Collaboration 2013 results XXVIII). The extraction of po-larised flux densities from the WMAP 9 year data has beenperformed using the IFCAPol software package used to char-acterize polarised sources in WMAP 5 year maps (L´opez-Caniego et al. 2009). This software implements the FilteredFusion approach (Arg¨ueso et al. 2011), where a maximumlikelihood estimator is obtained for the Q and U maps ofeach source. As a result, de-noised Q f and U f maps are pro-duced and the polarised flux density at the position of thesource is obtained from the map of P = q Q f + U f .Note that the WMAP polarisation maps are very noisyand it is important to assess whether or not our estimate ofthe polarised flux density at the position of a source detectedin the total intensity comes from the source or from a maxi-mum of the CMB at that position. This is done by assessingthe significance of each detection/estimation in the P map.For each source, we calculate the 99 .
90% significance level,as explained in L´opez-Caniego et al. (2009), and check thatour estimate of the flux density at the position of the sourceis at least above this level. This allows us to discriminatebetween truly significant detections in the maps of P fromrandom peaks of the background.Suitable detections were obtained at 23 GHz for 2 ofthe 9 sources in the above mentioned sub-sample (Fornax Aand Centaurus A). Extractions allowed us to define an upperlimit of the integrated polarised flux density for the other7 cases, but for 2 of these objects the extraction algorithmcould not determine the total flux density.The results of this extraction are listed in the main cat-alogue (see Table 2), flagged with ‘w’. The 23 GHz fluxdensities for the extended sources will be included in thefollowing analysis without any correction for the spectralbehaviour between the WMAP observing frequency and the18 GHz band. Hence, the full sample that will be used in thenext section includes 187 sources; it is 99% complete with S GHz > . Planck experiment), despite its steep spectrum in the re- c (cid:13) , 000–000 olarisation of bright AT20G sources Figure 1.
Radio colour-colour diagram for (from top to bottom)flux density and polarised flux density. Error bars and upper lim-its have been omitted for clarity of display. gion ∼ −
20 GHz. The detected value of 457 ±
35 mJy at 23GHz (listed in Table 3) is comparable with the 500 ±
60 mJymeasured for polarised flux density at 18 GHz by Burke-Spolaor et al. (2009) over the whole source (listed in Table 2and used in the following analysis). The WMAP detectionsseem to indicate a steep spectrum in polarised emission inthe WMAP frequency range.Table 3 lists the flux densities in total intensity andpolarisation for ForA, PicA, and CenA in all the WMAPbands. Notice that the detections at frequencies above 23GHz might refer only to fractions of the sources if they aremore extended than the WMAP beams.
Table 4.
Matrix of the number of sources classified accordingto both the total intensity and polarisation spectral behaviour.The columns are the spectral shape in polarisation, the rows thespectral shape in total intensity. The spectral types are definedin the text. S → (I) (P) (F) (S) (U)Pol. ↓ Inverted (I) 2 4 21 2 0Peaked (P) 5 10 33 10 0Flat (F) 1 3 17 7 1Steep (S) 0 4 15 9 0Upturning (U) 4 3 17 1 0
Table 5.
Median values of spectral indices in different frequencyranges in total intensity and polarisation. Spectral classes areselected according to their behaviour between 4.8 and 8.6 GHz. α ν ν All 0.09 -0.28 -0.11Flat 0.12 -0.27 -0.08Steep -0.60 -1.27 -0.95 α ν P,ν All 0.20 -0.16 -0.006Flat 0.23 -0.15 -0.006Steep -0.23 -0.43 0.04
We have defined the spectral index α as S ∝ ν α . The anal-ysis of AT20G data by Chhetri et al. (2012) has confirmedthat α = − . − . < α . . , α . < . α . . > α . < α . . < α . < α . . > α . > α . > α . . , α . < .
5) plane indicates that thespectral properties over the 5 to 18 GHz frequency rangemay be different in total intensity and in polarisation (seealso Fig. 2). The median values of the 4.8–8.6 GHz and 8.6–18 GHz spectral indices in polarisation are 0.20 and -0.16respectively. This effect could be a combination of Faradaydepolarisation operating at the lower frequencies, superposi-tion of multiple components with different polarised spectra,and different magnetic field properties for the componentsthat dominate the emission at the different frequencies.We constructed the matrix in Table 4 by comparing thetotal intensity and the polarisation spectral behaviours. Thedistribution across the cells confirms the differences of spec-tral behaviour in polarisation and in total intensity. Similarbehaviours translate in a diagonal matrix. However, even if atiny effect due to Faraday depolarisation could be the cause c (cid:13) , 000–000 Massardi et al.
Figure 2.
Comparison between the spectral index in total inten-sity and in polarisation in the ranges 4.8-8.6 GHz (upper panel)and 8.6-18 GHz. of the high number of peaked spectra in polarised emission(because of the lower level of emission at lower frequencies),more than ∼
30% of sources show a polarised emission spec-tral index on the lower frequency range higher than at thehigher frequencies, indicating that Faraday emission is notsignificantly affecting their spectral behaviour.Chhetri et al. (2012) demonstrate that low frequencyspectral index selections in flat and steep populations aremore effective in identifying compact and extended objectsthan high frequency spectral indices. For this reason we relyon the α . . to classify flat and steep spectra sources in someof the following analysis. We found 163 flat spectra objectsand only 9 steep spectra objects. Figure 1 also shows thatfor this selection criterion the total intensity and polarisa-tion spectral behaviours are different. The median of spec-tral indices in total intensity and polarised flux densities foreach class and for the full sample are in Table 5. An over-all steepening is appreciable both in total intensity and inpolarised emission. Table 6.
Distributions of the fractional polarisation at 18 GHzfor the full sample.
Π[%] Probability − +0.35 0.170 0.038 0.0471.05 0.284 0.048 0.0571.75 0.249 0.045 0.0542.45 0.159 0.036 0.0453.15 0.127 0.032 0.0413.85 0.126 0.032 0.0414.55 0.051 0.020 0.0305.25 0.046 0.020 0.0305.95 0.053 0.020 0.0306.65 0.049 0.020 0.0307.35 0.018 0.011 0.0228.05 0.030 0.016 0.0268.75 0.019 0.011 0.0219.45 0.014 0.011 0.02110.15 0.012 0.011 0.02110.85 0.008 0.007 0.01911.55 0.007 0.007 0.01912.25 < < < < We derived the distribution of the polarisation fractions inFig. 3 and Table 6 using a bootstrap re-sampling method inorder to account for the uncertainties in both the polarisedand the total intensity flux density measurements.We generated 1000 simulated catalogues by re-sampling, with repetitions, the input catalogue of polarisedflux densities. In each simulation, values for the polarisedand the total intensity flux densities were randomly assignedto each source by assuming a Gaussian distribution with amean equal to the measured values and a σ equal to thequoted errors. When only an upper limit was available onthe polarised flux density, we generated random values be-tween 0 and the quoted upper limit assuming a uniformdistribution. For each realization, the polarisation fractionis estimated for each object as the ratio between the simu-lated polarised flux density and the simulated total intensityflux density. The resulting values are then distributed intobins of polarisation fraction.The final distribution of the polarisation fractions isgiven by the mean value of the simulated polarisation frac-tions in each bin with uncertainties derived assuming aPoisson statistic, according to the prescriptions of Gehrels(1986). In Fig. 3 error bars correspond to the 68% confidenceinterval.The solid line is a fit assuming the lognormal distribu-tion: f (Π) = cost · √ πσ Π e ( − . / Π m )) /σ ) , with cost = 0 .
97, and σ = 0 . m = 2 .
14% is the medianvalue of the distribution.Table 7 summarizes the mean and quartiles of the dis-tributions of fractional polarization at 4.8, 8.6, and 18 GHzas calculated with survival analysis techniques to hold for c (cid:13) , 000–000 olarisation of bright AT20G sources Figure 3.
Distributions of the polarisation degree at 18 GHz.Errors and upper limits corresponds to a 68% c.l.. The solid lineshows the log-normal distribution with median fractional polari-sation 2.14% and σ = 0 .
90% (see the text for details).
Table 7.
Parameters describing the distributions of fractional po-larisation at 4.8, 8.6, and 18 GHz for each spectral class definedaccording to Fig. 1. The spectral types are defined in the text(Upturning class has been neglected because it includes only 1source).For each frequency and spectral class we quote the num-ber of detections, the mean fractional polarisation and its error,the first, second and third quartiles of the distribution.
F I P S18 GHz N TOT
105 12 25 29Detections 101 11 25 25 h Π i± σ h Π i ± ± ± ± quartiles N TOT
105 12 25 29Detections 100 10 24 24 h Π i± σ h Π i ± ± ± ± quartiles N TOT
105 12 25 29Detections 93 10 24 21 h Π i± σ h Π i ± ± ± ± quartiles upper limits in polarised flux density for each of the spec-tral classes identified considering the spectral behaviour inthe ranges 4.8-8.6 and 8.6-18 GHz (see Fig. 1) as discussedin the previous section. While no significant trend is visiblewith frequency for flat, upturning or peaked spectra objects,there is a tiny indication that the mean polarisation frac-tion for steep spectrum sources decreases as the frequencydecreases, probably under the effect of Faraday depolarisa-tion, but the significance is poor because we deal with fairlybroad distributions .Table 8 shows the parameters describing the distribu-tions of fractional polarisation for the full sample and forthe flat- and steep-spectrum objects as calculated with sur- Table 8.
Parameters describing the distributions of fractional po-larisation for the full sample and for the flat- ( α ν ν > − .
5) andsteep- ( α ν ν < − .
5) spectrum objects selected in the frequencyranges ν . , ν . Full sample Flat Steepselected between 4.8 and 8.6 GHz18 GHz N TOT
187 163 9Detections 171 157 5 h Π i± σ h Π i ± ± ± quartiles Prob(flat-steep) 7.8%8.6 GHz N TOT
172 163 9Detections 158 151 7 h Π i± σ h Π i ± ± ± quartiles Prob(flat-steep) 5.3%4.8 GHz N TOT
173 164 9Detections 149 143 6 h Π i± σ h Π i ± ± ± quartiles Prob(flat-steep) 0.1% vival analysis techniques to hold for upper limits in polarisedflux density, selected according to the spectral behaviour inthe 4.8-8.6 GHz frequency range only. There is a generaltrend towards an increase of the mean fractional polarisa-tion as the frequency increases for the whole sample and foreach spectral class (even if we have poor statistics for thesteep-spectrum sources). This trend is not so noticeable ifwe consider the median values, similarly to the findings inTable 7, as we deal with fairly broad distributions (see Table8). In fact, the differences among the polarisation degreedistributions at different frequencies are not statistically sig-nificant. All the generalized Wilcoxon two-sample tests avail-able in the ASURV package reject the null hypothesis, whichstates that the distributions of fractional polarisation aredrawn from the same parent distribution, for the 4.8 and8.6 GHz distributions for the full sample. In the range 8.6-18 GHz a correlation is found by the different methods at ∼ − σ significance level. Between 4.8 and 18 GHz theGehan’s and Peto & Peto version of the Wilcoxon test andthe logrank test confirm a correlation at the ∼ σ signif-icance level, while the Peto & Prentice version indicate a ∼ σ level (see Fig. 4).We then conclude that, for our sources, there is no sta-tistically significant evidence of an increase of the medianpolarisation degree with a frequency above 4.8 GHz. Thisimplies on the one hand that already at 4.8 GHz the Faradaydepolarisation is not very important, and on the other handthat the magnetic field is not substantially more ordered inthe regions dominating the emission at higher frequencies.This conclusion is in line with the results by Battyeet al. (2011) who found that the fractional polarisation of c (cid:13) , 000–000 Massardi et al.
Figure 4.
Fractional polarisation at 18 GHz as a function of 4.8and 8.6 GHz for the full sample (red diamonds indicates upperlimits at one or both the frequencies). their sources, selected from the WMAP point source cat-alogue and mostly flat-spectrum, is almost independent offrequency in the range 8.4–43 GHz, with median values inthe range 2–2.5%. Klein et al. (2003) also found that flat-spectrum sources in their sample selected at 408 MHz arecharacterized by almost constant polarisation degrees, withmedian values ≃ . Table 9.
Source counts at 18 GHz (including also WMAP data)as estimated for the current sample (upper panel) and for thewhole AT20G sample (lower panel). See the text for details. Errorsencloses the range of 68% c.l..log[ P ( Jy )] dN / dlogP (deg − ) − +-1.428 0.0119 0.0026 0.0028-1.238 0.0086 0.0020 0.0027-1.047 0.0042 0.0015 0.0019-0.857 0.0013 0.0006 0.0017-0.666 0.0014 0.0007 0.0016-0.476 0.0005 0.0004 0.0012-0.285 0.0005 0.0004 0.0012-0.095 < < < < < < Subrahmanyan et al. 2010). A similar trend was also foundby Sadler et al. (2006) for AT20G sources with S >
100 mJy. It was, however, not confirmed by the analysisof the full AT20G sample by Massardi et al. (2011a), al-though the relatively low detection rate in polarisation pre-vented a clear conclusion. Our data, with a high detectionrate, do not show any significant trend of Π with flux den-sity at any of the frequencies (4.8, 8.6 and 18 GHz) for S >
500 mJy, with only 45% probability that the null-hypothesis that sources with flux densities above and below1 Jy at 18 GHz come from the same parent distribution istrue. It must, however, be noted that our flux density rangeis limited and substantially narrower than that of Sadler etal. (2006).The median of the ratio between the fractional polar-isation at two different frequencies provides an estimate ofthe depolarisation as the frequency decreases. We found that h Π . / Π i = 79 .
7% and h Π . / Π i = 86 . We derived the differential number counts in the polarisedflux density in two ways. Results are listed in Table 9 andshown in Fig. 5. In both cases we used the bootstrap resam-pling method and performed 1000 simulations. c (cid:13) , 000–000 olarisation of bright AT20G sources Figure 5.
18 GHz differential source counts in polarisation calcu-lated with the two methods described in the text compared withthe findings by Tucci et al. (2012). The dashed line represent alinear fit of the whole AT20G sample data.
As a first method, we directly measured the numbercounts from the catalogue of polarised flux density measure-ments. For each simulation, the catalogue was resampledwith repetitions. As before, a value for each of the polarisedflux density was randomly assigned to each source by assum-ing a Gaussian distribution with a mean equal to the mea-sured values and σ equal to the quoted errors. When onlyan upper limit was available on the polarised flux density,we generated random values between 0 and the quoted up-per limit assuming a uniform distribution. The distributionof the polarised flux densities derived from the simulationswere binned into a histogram, and the mean value in each binwas taken as the measurement of the number count in thatbin, after dividing by the bin size and by the survey area.The errors on the number counts were derived assuming aPoisson statistic, according to the prescriptions of Gehrels(1986). The results are shown in Fig. 5 by the red dots witherror bars corresponding to the 68% confidence interval.For the second method we started off with the Mas-sardi et al. (2011a) catalogue of total intensity flux densi-ties. In each simulation the total intensity flux densities wasrandomly assigned to each source by assuming a Gaussiandistribution with a mean equal to the measured values andsigma equal to the quoted errors. At the same time a real-ization of the polarisation fraction was produced from thecatalogue of polarised flux densities following the proceduredescribed before. The simulated total intensity flux densityand the simulated value of the polarisation fraction werethen used to derive the corresponding polarised flux densityof each object. Finally the distribution of the polarised fluxdensities obtained from the simulations were binned into ahistogram and the mean value in each bin was taken as themeasurement of the number count in that bin, divided by thebin size and by the survey area. The errors on the numbercounts were derived assuming a Poisson statistic, accordingto the prescriptions of Gehrels (1986). The results are shownin Fig. 5 by the blue dots with error bars corresponding tothe 68% confidence interval. These points are linearly fittedwith the relation:log( dN/d log P [ deg − ]) = ( − . ± .
05) log P [ Jy ]+( − . ± . . In the same Figure the green squares represent the num-ber counts in the polarised flux densities derived by Tucci& Toffolatti (2012).
Finding suitable calibrators for polarisation studies at milli-metric wavelengths is a difficult exercise. On the one hand,extragalactic sources typically have a very low level of cir-cular polarisation, which simplifies the calibration solutions.On the other hand, as demonstrated in the previous sections,the fractional polarisation is typically a small fraction ofthe total intensity and polarised behaviour cannot be easilypredicted on the basis of total intensity properties. Further-more, source variability makes the fractional polarisationunknown. Source compactness (with respect to the beamof the used telescope) allows us to consider only on-axis ef-fects. Models of extended emissions of calibrators could beused to cope with off-axis effects, if there is no better suitedcompact calibrator.Polarised sources allow to recover gain and instrumen-tal polarisation parameters calibration, even if the polari-sation fraction of the calibrator is unknown (but non-zero)and its polarisation angle is unknown, by observing over awide range of parallactic angles (for alt-az telescope mountsand assuming the instrumental parameters do not vary inthe time of the observation, see Sault, Hamaker & Bregman1996).Statistical analysis clearly showed that spectral be-haviours in total intensity and polarisation are typically dif-ferent, with a tiny indication that the fractional polarisationincreases with frequency, at least for high-frequency-selectedsteep-spectra sources. Therefore, it is difficult to make pre-dictions with our data at frequencies > ∼
20 GHz. Any can-didate selection requires proper monitoring at the observingfrequency to confirm the calibrator properties. Hence, westress the fact that in this section we do not claim any def-inition of effective criteria to identify good calibrators, butwe only suggest a short list of targets for future calibratormonitoring programs at frequencies > ∼
20 GHz .The list is selected in the catalogue presented in theprevious sections of sources • south of − ◦ deg, excluding the Galactic plane region( | b | . ◦ ) and the LMC region • with total intensity flux density at 20 GHz S GHz >
500 mJy in the 2006 AT20G selection.Among them we selected the sources with fractionalpolarisation at 20 GHz Π GHz >
3% (corresponding to
P >
15 mJy and, according to the distribution in Sect. 5,enclosing about 40% of our sources) and at least two of thefollowing properties • flat spectral index between 8 and 20 GHz α > − . • S GHz >
500 mJy in all the AT20G epochs, and vary-ing less than 10% with respect to the October 2006 epochover 1 yr of observations or smaller than 20% over longer c (cid:13) , 000–000 Massardi et al.
Table 10.
Candidate millimetric polarisation calibrators (see the text for details on selection criteria). Columns are as follows: (1) AT20Gname; (2) 18 GHz total intensity flux density and its error in mJy; (3) 18 GHz fractional polarisation; (4-5) spectral indices α . . and α . where low frequency AT20G data are available; (6) variations of fractional polarisation at different frequencies ∆ Π = Π . GHz − Π GHz where low frequency AT20G data are available; (7-8) epoch of AT20G catalogue run (see Murphy et al. 2010; ‘1’: 2004, ‘2’: 2005,‘3-4’:2006,‘5-6’: 2007,‘7’: 2008) and relative variation of flux densities ( S AT G − S Oct ) /S AT G ; (9) ratio between the visibilities amplitudeaveraged over the long baselines ( ∼ . ∼ . σ ) in the Planck
Legacy Catalogue (PlanckCollaboration, 2013) at 30, 100, 217, and 353 GHz, in mJy; (15-16) redshift and optical identification (‘QSO-G’: quasar and galaxy asclassified in NED,‘AeB’: AGN with broad line emissions, ‘Ae’: AGN with emission features, Mahony et al. 2011).
Name ‘AT20GJ’ S20 m20 α . . α . ∆ Π Run Variab 6kmVis S ATCAcal S
Planck S Planck S Planck S Planck z Optid[mJy] [%] [%] 18 GHz ratio 3mm[mJy] 30GHz 100GHz 217GHz 353GHzJ001035-302748 595.8 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± epochs to select the most stable objects according to theavailable data; • increasing fractional polarisation between 8 and 20GHz, to favour high levels of fractional polarisation obser-vations at the higher frequencies.The selected sample contains 29 sources. They havebeen flagged in Table 2 with a ‘c’ and some of their proper-ties are summarized in Table 10. All the sources are classi-fied as point-like in the AT20G catalogue (i.e. with most oftheir emission within the synthesized beam of 10 arcsec at20 GHz).Furthermore, where available, we considered the source‘6 km visibility’ which is the ratio of the scalar averagedamplitudes at the long baselines ( ∼ ∼ > ∼ ∼
30 GHz, which may affect the higher frequencyspectra. Optical identifications (Mahony et al. 2011) indicatethat the sample is mostly composed by QSO at mean red- shift above 1.2. Almost all the objects are calibrators in theATCA database and the flux densities at 3 mm availablefor 10 of them are larger than 400 mJy. Detections withtotal intensity > σ are available in the Planck
LegacyCatalogue (Planck Collaboration 2013) at 100, 217, and 353GHz channels (roughly corresponding to ALMA Band 3, 6,and 7 frequencies) in the position of 22 sources with medianvalues above 450 mJy for all the bands. The median frac-tional polarisation of the sample is above 4.8%. The mostpolarised object is AT20GJ210933-411020 with 1.9 Jy of to-tal intensity and 10% fractional polarisation, flat spectrum,increasing fractional polarisation with frequency, and only14% relative variability over 3 yr time. In the
Planck chan-nels the spectra become steep down to 428 mJy at 100 GHzand only 128 mJy at 217 GHz.AT20GJ063546-751616 has 5.33 Jy of total intensityflux density at 20 GHz, 6.2% polarised, and remains above 1Jy up to 1 mm frequencies. It is classified as a flat-spectrumradio quasar at redshift z=0.653. Several notes indicate thepresence of a jet structure, but the 6 km visibilities identifyit as point-like and modestly variable in ATCA observationsover few years. Thanks to its position it is always visible toSouthern hemisphere telescopes like ATCA and ALMA andstands as the most suitable polarisation calibrator at highfrequencies and low declinations. c (cid:13) , 000–000 olarisation of bright AT20G sources We have conducted sensitive polarisation and total inten-sity observations on a 20 GHz flux-limited sample of 189objects selected in the Australia Telescope 20 GHz Survey,choosing sources that have S >
500 mJy in the decli-nation range δ < − ◦ in the survey scans before October2006. They have been followed up during an observing runin October 2006 designed to reach 1 mJy sensitivity in po-larisation. This strongly improved the sensitivity and thedetection rate for polarisation observations over any previ-ous sample investigated in this sky area at frequencies above10 GHz.94% of the 180 extragalactic point sources have a de-tection of polarised flux density at least at 18 GHz. 172 ofthem have been observed also at 4.8 and 8.6 GHz, and 143sources have a detected polarised flux density at all threefrequencies.The 9 sources identified as extended have poor qualityflux density measurement. So, for the sake of completeness,we extracted the values of polarised flux density from the9-yr co-added WMAP maps. We recover an upper limit for5 of them and a detection at 23 GHz for 2 of them (ForAand CenA). The final sample of 187 sources that we ana-lyzed constitutes a 99% complete sample at the 2006 surveyselection epoch with a 91.4% polarisation detection rate. Inaddition, detections have been obtained at all the WMAPfrequencies for PicA.This sample constitutes an ancillary data set for presentand future studies of polarisation in the Southern Hemi-sphere and complements other samples recently observedeither in equatorial regions (Sajina et al. 2011) or in theNorthern hemisphere (Jackson et al. 2010). Analysis of theWMAP and Planck data (L´opez-Caniego et al. 2009 andreferences therein) has demonstrated that similar source listsare crucial to improve the investigation of the CMB E andB modes in millimetric wavelength bands.Thanks to our high detection rate, to a low polarisedflux density level, to the multi-frequency observations, andto the inclusion of integrated flux densities for extendedobjects observed in mosaic mode with the ATCA (Burke-Spolaor et al. 2009) or extracted from the WMAP 9-yearmaps (updating the findings of L´opez-Caniego et al. 2009)the analysis of our sample in total intensity and polarisationallowed us to draw the following conclusions. • The spectral behaviours in total intensity and in po-larisation are different for any population of sources. Thisimplies that it is extremely difficult to make an estimation ofpolarised flux densities from total intensity measurements. • There is no statistically significant evidence of increas-ing fractional polarisation with frequency. This implies thatFaraday depolarisation is not strong enough to modify thespectral behaviour at and above ∼ . • Thanks to our high detection rate we can state thatthere is no evidence of an anticorrelation of fractional po-larisation with total intensity flux density as was previouslynoted by several surveys, which were probably biased by aselection effect: only highly polarised sources can be detectedfor faint sources, while low fractional polarisation percent- ages can be detected in bright objects; furthermore faintobjects in complete samples are typically more numerousthan bright ones. • Thanks to the high sensitivity of our observations wewere able to extend the polarisation source counts at 18 GHzof Tucci & Toffolatti (2012) and to confirm their findings. • We identified a list of 29 candidate calibrators for po-larisation at declination below − ◦ and frequencies > ∼ ∼ MM, MN, and GDZ acknowledges financial support forthis research by ASI/INAF Agreement I/072/09/0 for the
Planck
LFI activity of Phase E2. ML-C acknowledgespartial financial support from the Spanish Ministerio deEconoma y Competitividad project AYA-2012-39475-C02-01 and the Consolider Ingenio-2010 Programme projectCSD2010-00064.We thank the staff at the Australia Telescope Com-pact Array site, Narrabri (NSW), for the valuable supportthey have provided in running the telescope. The AustraliaTelescope Compact Array is part of the Australia Telescopewhich is funded by the Commonwealth of Australia for op-eration as a National Facility managed by CSIRO.We thank Elisabetta Liuzzo (INAF-IRA) and BenjaminWalter (Haevrford) for the useful discussions and proofread-ing. We thank the anonymous referee for the useful com-ments.
REFERENCES
Agudo I., Thum C., Wiesemeyer H., Krichbaum T. P.,2010, ApJS, 189, 1Arg¨ueso F., Gonz´alez-Nuevo J., Toffolatti L., 2003, ApJ,598, 86Argueso F., Sanz J. L., Herranz D., 2011, arXiv,arXiv:1101.0701Battye R. A., Browne I. W. A., Peel M. W., Jackson N. J.,Dickinson C., 2011, MNRAS, 413, 132Blandford R. D., K¨onigl A., 1979, ApJ, 232, 34Burke-Spolaor S., Ekers R. D., Massardi M., Murphy T.,Partridge B., Ricci R., Sadler E. M., 2009, MNRAS, 395,504Chhetri R., Ekers R. D., Mahony E. K., Jones P. A., Mas-sardi M., Ricci R., Sadler E. M., 2012, MNRAS, 2747Chhetri R., Ekers R. D., Jones P. A., Ricci R., 2013, arXiv,arXiv:1306.0990Condon J. J., Cotton W. D., Greisen E. W., Yin Q. F.,Perley R. A., Taylor G. B., Broderick J. J., 1998, AJ, 115,1693de Zotti G., Toffolatti L., Arg¨ueso F., Davies R. D., Maz-zotta P., Partridge R. B., Smoot G. F., Vittorio N., 1999,AIPC, 476, 204de Zotti G., Ricci R., Mesa D., Silva L., Mazzotta P., Tof-folatti L., Gonz´alez-Nuevo J., 2005, A&A, 431, 893Gehrels N., 1986, ApJ, 303, 336 c (cid:13) , 000–000 Massardi et al.
Gold B., et al., 2011, ApJS, 192, 15Grant J. K., Taylor A. R., Stil J. M., Landecker T. L.,Kothes R., Ransom R. R., Scott D., 2010, ApJ, 714, 1689Griffith M.R., Wright A.E., Burke B.F., Ekers R.D., 1994,ApJS, 90, 179Griffith M.R., Wright A.E., Burke B.F., Ekers R.D., 1995,ApJS, 97, 347Hancock P. J., et al., 2011, ExA, 32, 147Jackson N., Battye R. A., Browne I. W. A., Joshi S.,Muxlow T. W. B., Wilkinson P. N., 2007, MNRAS, 376,371Jackson N., Browne I. W. A., Battye R. A., Gabuzda D.,Taylor A. C., 2010, MNRAS, 401, 1388Klein U., Mack K.-H., Gregorini L., Vigotti M., 2003, A&A,406, 579Kurinsky N., Sajina A., Partridge B., Myers S., Chen X.,L´opez-Caniego M., 2013, A&A, 549, A133Mahony E. K., et al., 2011, MNRAS, 417, 2651L´opez-Caniego M., Massardi M., Gonz´alez-Nuevo J., LanzL., Herranz D., De Zotti G., Sanz J. L., Arg¨ueso F., 2009,ApJ, 705, 868Massardi M., et al., 2008, MNRAS, 384, 775Massardi M., et al., 2011a, MNRAS, 412, 318Massardi M., Bonaldi A., Bonavera L., L´opez-Caniego M.,de Zotti G., Ekers R. D., 2011b, MNRAS, 415, 1597Mauch T., Murphy T., Buttery H. J., Curran J., HunsteadR. W., Piestrzynski B., Robertson J. G., Sadler E. M.,2003, MNRAS, 342, 1117Mesa D., Baccigalupi C., De Zotti G., Gregorini L., MackK.-H., Vigotti M., Klein U., 2002, A&A, 396, 463Murphy T., et al., 2010, MNRAS, 402, 2403Planck, 2011a, A&A, 536, A7Planck, 2011b, A&A, 536, A13Planck, et al., 2013, arXiv, arXiv:1303.5088Ricci R., Prandoni I., Gruppioni C., Sault R. J., De ZottiG., 2004, A&A, 415, 549Sadler E. M., et al., 2006, MNRAS, 371, 898Sajina A., Partridge B., Evans T., Stefl S., Vechik N., MyersS., Dicker S., Korngut P., 2011, ApJ, 732, 45Sault R. J., Teuben P. J., Wright M. C. H., 1995, ASPC,77, 433Sault R. J., Hamaker J. P., Bregman J. D., 1996, A&AS,117, 149Sault R. J., Rayner D. P., Kesteven M. J., 2002, AIPC,609, 150Subrahmanyan R., Ekers R. D., Saripalli L., Sadler E. M.,2010, MNRAS, 402, 2792Taylor A. R., et al., 2007, ApJ, 666, 201Toffolatti L., Arg¨ueso Gomez F., de Zotti G., Mazzei P.,Franceschini A., Danese L., Burigana C., 1998, MNRAS,297, 117Toffolatti L., De Zotti G., Arg¨ueso F., Burigana C., 1999,ASPC, 181, 153Tucci M., Martinez-Gonzalez E., Toffolatti L., Gonzalez-Nuevo J., De Zotti G.; 2004; MNRAS, 349, 1267Tucci M., Toffolatti L., 2012, arXiv:1204.0427 c (cid:13)000