(PDF) Sub-millimetre source identifications and the microjansky source population at 8.4 GHz in the William Herschel Deep Field

Mon. Not. R. Astron. Soc. 000, 1–19 (20XX) Printed 8 January 2014 (MN LATEX style file v2.2) Sub-millimetre source identifications and the micro-Jansky source population at 8.4 GHz in the William Herschel Deep Field I. Heywood1,3⋆, R. M. Bielby2, M. D. Hill2, N. Metcalfe2, S. Rawlings1,T. Shanks2 , arXiv:1209.4660v1 [astro-ph.CO] 20 Sep 2012 and O. M. Smirnov3,4 1 Astrophysics, Department of Physics, University of Oxford, Keble Road, Oxford, OX1 3RH, UK 2 Department of Physics, Durham University, South Road, Durham, DH1 3LE, UK 3 Department of Physics and Electronics, Rhodes University, PO Box 94, Grahamstown, 6140, South Africa 4 SKA South Africa, 3rd Floor, The Park, Park Road, Pinelands, 7405, South Africa Accepted 20XX Month XX. Received 20XX Month XX; in original form 20XX Month XX ABSTRACT Sub-millimetre observations of the William Herschel Deep Field (WHDF) using the Large Apex Bolometer Camera (LABOCA) revealed possible sub-mm counterparts for two X-ray absorbed quasars. The primary aim here is to exploit Expanded Very Large Array (EVLA) radio continuum imaging at 8.4 GHz to establish the absorbed quasars as radio/sub-mm sources. The main challenge in reducing the WHDF EVLA data was the presence of a strong 4C source at the field edge. A new calibration algorithm was applied to the data to model and subtract this source. The resulting thermal noise limited radio map covers a sky area which includes the 16′ ×16′ Extended WHDF. It contains 41 radio sources above the 4σ detection threshold, 17 of which have primary beam corrected flux densities. The radio observations show that the two absorbed AGN with LABOCA detections are also coincident with radio sources, confirming the tendency for X-ray absorbed AGN to be sub-mm bright. These two sources also show strong ultraviolet excess (UVX) which suggest the nuclear sightline is gas- but not dust-absorbed. Of the three remaining LABOCA sources within the ≈ 5′ half-power diameter of the EVLA primary beam, one is identified with a faint nuclear X-ray/radio source in a nearby galaxy, one with a faint radio source and one is unidentified in any other band. More generally, differential radio source counts calculated from the beam-corrected data are in good agreement with previous observations, showing at S < 50µJy a sig- nificant excess over a pure AGN model. In the full area, of ten sources fainter than this limit, six have optical counterparts of which three are UVX (i.e. likely quasars) including the two absorbed quasar LABOCA sources. The other faint radio counter- parts are not UVX but are only slightly less blue and likely to be star-forming/merging galaxies, predominantly at lower luminosities and redshifts. The four faint, optically unidentified radio sources may be either dust obscured quasars or galaxies. These high redshift obscured AGN and lower redshift star-forming populations are thus the main candidates to explain the observed excess in the faint source counts and hence also the excess radio background found previously by the Absolute Radiometer for Cosmology, Astrophysics and Diffuse Emission (ARCADE2) experiment. Key words: galaxies: high-redshift – quasars: general – radio continuum: galaxies – techniques: interferometric 1 INTRODUCTION The William Herschel Deep Field (WHDF; 7 × 7 arcminutes ⋆

, SEPnet fellow at 00h 22m 30s +00◦ 21m 00s) has been comprehensively c 20XX RAS 2 Heywood et al. targeted by many authors using many observatories for over 2 OBSERVATIONS AND DATA REDUCTION two decades (Metcalfe et al., 1991, 1995, 2001, 2006; Mc- The WHDF was observed with the EVLA in its most com- Cracken et al., 2000). It is one of the deepest survey fields in pact D-configuration between 28 March 2010 and 15 June terms of optical and near infrared photometry, having more 2010. A total of 35 hours were awarded under the Open than 150 hours of CCD imaging on the 4.2 m William Her- Shared Risk Observing program, of which 30 hours were ob- schel Telescope (WHT) and 3.9 m telescope at Kitt Peak Na- served. The observations were recorded in 25 independent tional Observatory, resulting in magnitude limits of U = 27, scheduling blocks (SBs) with durations of either 1.5 or 3.5 B = 28.2, R = 27 and I = 25.5. The magnitude apertures hours, resulting in 25 sets of visibility data. vary slightly with band / seeing but are ≈ 1.′′ 3 radius for the The X-band receivers were used and the correlator was faintest stars. All magnitudes quoted are in the Vega system. configured to deliver two spectral windows centred at 8.396 More details are given by Metcalfe et al. (2001). There is also and 8.524 GHz, each providing 64 × 2 MHz channels giving deep H band imaging following 30 hours of exposure with the 256 MHz of contiguous frequency coverage. For this contin- Calar Alto 3.5 m telescope (Vallbe-Mumbru, 2004). Deep, uum experiment an 8:1 frequency averaging was applied to high-resolution I-band imaging is available through Hub- make the data volume more manageable. No time averag- ble Space Telescope (HST) Advanced Camera for Surveys ing was applied, so the standard 1 second integration time (ACS) imaging (Bohm & Zeigler, 2006). Observations with was preserved. Each of the 25 Measurement Sets were then the Low Dispersion Survey Spectrograph 2 (LDSS-2) spec- split further in order to separate each of the two spectral trograph on the 6.5 m Magellan telescope (Vallbe-Mumbru, windows, resulting in a total of 50 unique data sets to be 2004) and Very Large Telescope (VLT, 8.2 m) observations independently edited and calibrated. using FORS2 (Bohm & Zeigler, 2006) have yielded ∼200 Editing of the data was performed using the PlotMS spectroscopic redshifts in the field. tool within the NRAO CASA1 package. In general the data were free from corruptions save for the occasional bursts Moving away from optical and infrared wavelengths, the of radio frequency interference which were simple to locate WHDF has deep X-ray imaging via 75 ks of Chandra ACIS- and excise. Two SBs exhibited severe amplitude drifts on a I data (Vallbe-Mumbru, 2004), and Bielby et al. (2012) majority of baselines and in both spectral windows and were recently obtained sub-mm observations of the field using discarded outright. the Large Apex Bolometer Camera (LABOCA), reaching For flux calibration a single observation of J0137+3309 a depth of 1 mJy at a wavelength of 870 µm and resulting (3C48) was performed at the start of each SB, and the phase in the detection of eleven sources. calibrator J0022+0014 (4C 00+02; 0.64 Jy at X-band) was observed for 40 seconds for every 9 minutes of target obser- vation. This paper presents the latest addition to the multi- Standard calibration procedures were followed using wavelength coverage of the WHDF, namely 8.4 GHz ob- CASA. The flux scale for the amplitude calibrator was set servations using the Expanded Very Large Array (here- against a model image as 3C48 is slightly resolved in X-band after EVLA, since renamed the Karl G. Jansky Very Large observations, and this source was also used to calibrate the Array). These observations were motivated by the recent bandpass. The per-antenna complex gain solutions were gen- LABOCA data with the high resolution afforded by the ra- erated for the phase calibrator source and then interpolated dio observations being required to confirm whether any of across the target field. A Python script was constructed to the bright sub-mm sources in the field were associated with apply this standard procedure to each of the 50 flagged Mea- the significant sample of both absorbed and unabsorbed surement Sets automatically. Plots of the gain solutions were quasars (QSOs) also present in the WHDF. In particular, generated and inspected in order to ensure successful cali- Bielby et al. (2012) suggested two X-ray absorbed high red- bration. Phase stability was excellent throughout. As an ad- shift active galactic nuclei (AGN), one of which is a Type 2 ditional diagnostic, calibrated maps of the phase calibrator quasar showing only narrow lines in the optical, may be asso- and target field were also generated by the script. ciated with LABOCA sources, in line with previous sugges- Inspection of the images of the target field immediately tions of a connection between X-ray absorbed AGN and sub- shows why the WHDF is a challenge for radio interferome- mm emission (Gunn 1999; Page et al. 2004; Hill & Shanks ter observations, as the phase calibrator is situated approxi- 2011). It is therefore important to establish the reality of the mately 6 arcminutes south of the target field and completely association between the sub-mm and these absorbed AGN dominates the radio emission in each map. Furthermore, as as far as possible, using these radio data. with many well-studied multiwavelength fields, the WHDF is situated close to the celestial equator (Declination +20 arcminutes). When observing the celestial equator with an The WHDF is hitherto largely unexplored at radio interferometer the locus traced by a given baseline in the uv wavelengths due to the presence of a strong 4C source 6 plane as the Earth rotates exhibits a strong east-west bias. arcminutes south of the field centre, however this has been The Fourier transform of the uv coverage determines the cleanly subtracted from the observations described in this point-spread function (PSF, also known as the dirty beam) paper using a new calibration algorithm. A description of of the observation, thus a uv plane sampling pattern that the observations and the calibration process follows in Sec- tion 2. The radio map and source catalogue are presented in Section 3. Section 4 matches the radio sources with coun- terparts at other wavebands. These results are discussed in 1 Common Astronomy Software Applications: Section 5, and concluding remarks are made in Section 6. http://casa.nrao.edu c 20XX RAS, MNRAS 000, 1–19 The micro-Jy source population at 8.4 GHz in the WHDF 3 is dominated by east-west structure results in strong north- Data are averaged over five minute intervals and in south sidelobes in the PSF. As can be seen in the upper pairs of frequency channels (i.e. four blocks across the aver- panel of Figure 1 the sidelobes associated with the phase aged band). A solution is generated for each of these time- calibrator completely cover the target field, and these ef- frequency tiles. This has the effect of reducing the degrees of fects must be mitigated in order to obtain a scientifically freedom in the fit, whilst simultaneously time-smearing out useful radio map. contributions from the rest of the field as well as accounting for any spectral behaviour in either the direction-dependent gain corruptions or the source itself. In this regime, the use of 2.1 Subtracting the phase calibrator from the the differential gains algorithm is analogous to the more gen- target field erally employed “peeling” algorithm (e.g. Noordam, 2004), however the advantage of the differential gains technique for Straightforward multifrequency synthesis deconvolution of more general applications is one of flexibility: it can generate the phase calibrator from the target field using the CLEAN additional gain terms for many individual sources simulta- algorithm did not yield the result that one might hope for, neously, whilst simultaneously solving on shorter timescales and residual sidelobe structure still swamped the target for the traditional complex receiver gains derived from an field: the upper panel of Figure 1 actually shows the cali- all-inclusive sky model. Peeling generally requires a cum- brated target field after deconvolution has been attempted. bersome iterative approach whereby dominating sources are Similarly, subtraction of a visibility model derived from the treated in order of brightness. clean components followed by imaging of the residual data The best fit visibility model derived from this process set resulted in a map that was still corrupted by strong was subtracted from the observed visibilities, and the resid- north-south emission associated with the phase calibrator. ual data were imaged. Following this procedure the confus- Examining a model of the EVLA primary beam at 8.4 ing source is completely removed leaving no residual emis- GHz (Walter Brisken, private communication; Figure A2) sion above the noise. As a quantitative example of the suc- reveals that when the array is pointing at the target field cess of this of this technique, the root-mean-square (rms) the phase calibrator lies at the boundary of the first null and background level in a map formed from one of the spectral the first sidelobe, with its radial separation from the point- windows of one of the 3.5-hour scheduling blocks was 21 µJy ing centre changing with frequency. The first sidelobe of the / beam (as measured away from the dominating residual EVLA beam has four-way azimuthal structure caused by the north-south structure) following either deconvolution of the Cassegrain optics of the dishes, and the azimuth-elevation confusing source, or subtraction of an inverted clean com- mount of the EVLA dishes causes this pattern to rotate on ponent model from the visibilities. This value dropped to the sky as a source is tracked. Since the primary beam can 12 µJy / beam following the subtraction of the confusing to first order be thought of as the gain of the instrument as a source using the differential gains algorithm. The radio im- function of direction, these effects conspire to impart appar- ages corresponding to these “before” and “after” scenarios ent temporal variability to the confusing source (as well as can be seen in Figure 1 with further details provided in the significant spectral effects). Sources which vary significantly caption. on timescales shorter than the observation have corrupted PSFs associated with them, and this is why traditional de- Following the discarding of the two corrupted SBs, the convolution and model subtraction cannot cope. first step in the calibration produced 46 Measurement Sets which were then split in order to contain only calibrated A calibration scheme which can solve for these direction-dependent effects must be employed, and for these visibilities from the target field. MeqTrees can be operated non-interactively, allowing the source subtraction process to observations the differential gains algorithm2 (Smirnov, be scripted and applied automatically to each of these Mea- 2011b) was used, implemented using the Calico frame- work within the MeqTrees3 software package (Noordam & surement Sets. Again, calibrated maps were produced for each individual Measurement Set and the differential gain Smirnov, 2011). The algorithm works by solving for addi- solutions were examined to check for problems. The gain so- tional complex gain terms against an assumed sky model for a subset of sources. In the case of these observations the lutions themselves encode information as to the origin of the direction-dependent corruptions, and these are discussed in solutions are generated for a sky model consisting solely of the Appendix. a point source at the location of the phase calibrator. Since this source dominates the radio emission, the gain solutions will also be dominated by corruptions in the direction of that source. There is no need to provide a stringent measure of the source flux for the sky model, as any errors in this pa- rameter will be subsumed into the gain solutions, however 2.2 Concatenation and imaging the flux in the sky model was fixed at 5 mJy, as measured from the image generated following the initial calibration The calibrated data with the confusing source removed were pass. Note the high attenuation of this source by comparing concatenated into a final Measurement Set which was im- this measured value to its intrinsic flux measured by imaging aged to form the final map. At this stage, only 23 of the the phase calibration scans. 46 viable Measurement Sets were included. The ones which were not included were shorter duration runs at low eleva- tions. The observations had relaxed hour angle constraints 2 Also known as “The Flyswatter” for reasons that will become when they were scheduled, and the inclusion of these data obvious. did not improve the quality or depth of the final map, which 3 http://www.astron.nl/meqwiki is presented in Section 3.1. c 20XX RAS, MNRAS 000, 1–19 4 Heywood et al. 3 RESULTS The following two sections present the final radio map of the WHDF at 8.4 GHz and the source catalogue derived from it. 3.1 Radio map Figure 2 shows the final, deconvolved 8.4 GHz radio image covering the WHDF. Deconvolution was performed using the CASA clean task in multi-frequency synthesis mode with w-term correction. The restoring beam is a circular Gaussian with a full-width at half-maximum spanning 8 arcseconds, as shown by the filled circle in the lower left-hand corner of the map. The rms of the background noise level in this map is 2.5 µJy. This is consistent with the theoretical thermal noise level expected in this observation to within 3%. The contours √ on Figure 2 begin at 3σ and increase in multiples of 2. There is a single dashed negative contour at -3σ. The area of sky covered by this image is 22.2 × 22.2 arcminutes, although most of the multiwavelength observa- tions occupy a 7 × 7 arcminute patch near the map centre. This is well-matched to the primary beam of the EVLA an- tennas, the approximate half-power point of which is shown by the large dashed circle in Figure 2. Imaging over this ex- tended area in the radio was done to search for bright radio sources in the 16 × 16 arcminute area covered by the shal- lower Extended WHDF observations. The position of the phase calibrator source (4C 00+02) that was confusing the central region before subtraction is also marked. No resid- ual emission is present above the noise. Also noteworthy is the spatially extended radio source towards the western edge of the map (WHDF-EVLA-3 in Table 1). This source is present in the Sloan Digital Sky Survey (SDSS; Abazajian et al., 2009) which lists it as a galaxy at z = 0.2609. The crosses on Figure 2 show the positions of the spectroscopically-confirmed quasars in the WHDF (Vallbe- Mumbru, 2004). The small circles show the locations of the LABOCA-detected sub-mm sources in the field, with the size of the circles being the positional uncertainties in the detections. The nomenclature of these two classes of sources on Figure 2 and in the discussion that follows is a concise version of that employed by Bielby et al. (2012): three digit identifications are confirmed quasars and two digit identifi- Figure 1. Radio images of the target field generated from a sin- gle 3.5-hour Measurement Set. The upper panel shows the image cations correspond to LABOCA sources (prefixed by WHD- that results from the initial complex gain calibration and a multi- FCH and WHDF-LAB respectively in the aforementioned frequency synthesis deconvolution of the phase calibrator / con- article). Radio sources are referred to using the full WHDF- fusing source using the CLEAN algorithm. The time-dependent EVLA prefix throughout, with the exception of the labels gains imparted by the primary beam of the array in that direction on Figure 3. This covers the same sky area as Figure 2 but corrupt the PSF associated with this source and even following shows the location of each source according to its identifi- deconvolution the residual sidelobe structure completely covers cation as listed in Table 1. Figure 3 also shows the central the science targets. The lower panel shows the map that results and extended regions of the WHDF. Higher magnification following subtraction of the confusing source using the differential thumbnails of individual radio sources can be found in Fig- gains algorithm (Smirnov, 2011b). The position of the confusing ures 4 and 5, along with images of the counterparts identified source is marked with the cross. Radio sources in the target field are now visible. The contour levels in these maps are the same, at other wavebands (see Section 4.2). and begin at 3σ (where σ = 12 µJy, the background rms of the lower map) and increase in multiples of 2. A single, dashed neg- ative countour is included with a value of -3σ. 3.2 Source catalogue The radio map presented in Section 3.1 was searched for peaks of emission exceeding 4σ (= 10 µJy) using the SAD task within the Astronomical Image Processing System (AIPS; Greisen, 2003) package. The resulting catalogue was c 20XX RAS, MNRAS 000, 1–19 The micro-Jy source population at 8.4 GHz in the WHDF 5 Figure 2. Radio contour map of the William Herschel √ Deep Field at 8.4 GHz. The rms noise (σ) in the map is 2.5 µJy / beam. Contours begin at 3σ and increase in multiples of 2. There is a single dashed negative contour at -3σ. The numbered crosses show the spectroscopically confirmed quasars in the WHDF (Vallbe-Mumbru, 2004) and the numbered circles indicate the locations of the LABOCA-detected sub-mm sources (Bielby et al., 2012). The numbers associated with these two classes of sources are consistent with those used by Bielby et al. (2012), see Section 3.1 for details. The size of the circles used to show the locations of the LABOCA sources indicates the positional uncertainty. The large dashed circle shows the approximate half-power point of the EVLA primary beam. The position of the phase calibrator 4C 00+02 is also marked although no detectable emission remains. Note also the extended SDSS source to the west. The filled circle in the lower left-hand corner of the plot shows the 8 arcsecond extent of the circular restoring beam applied following deconvolution. manually pruned as two of the sources (WHDF-EVLA-3 and Of the sources detected in the map, 17 of them were WHDF-EVLA-6) were resolved into multiple Gaussian com- within the region where the gain of the main lobe of the ponents. In these cases the flux was determined by summing EVLA primary beam exceeded 0.2. These are indicated by up the components. The 22.2 × 22.2 arcminute map shown the circles in Table 1 (see also Figure 3). An analytic expres- in Figure 2 contains 41 discrete sources with flux densities sion is available to correct for the beam gain of the (E)VLA exceeding 4σ and the properties of these are listed in de- antennas as per the AIPS task PBCOR, and this has been scending order of brightness in Table 1. applied to the 17 sources in the main lobe. Thus bold-face c 20XX RAS, MNRAS 000, 1–19 6 Heywood et al. type in Table 1 represents intrinsic flux values and plain WHDF-EVLA-8 correspond to the X-ray absorbed quasars type entries have apparent flux values. The measured sep- 007 (z = 1.33) and 008 (z = 2.12) respectively. aration from the pointing centre that was used to derive The reliability of the radio and sub-mm sources be- the primary beam correction factor for each source is also ing counterparts (and not chance alignments) is evaluated listed. Uncertainties in the intrinsic fluxes include contribu- here using the corrected Poisson probability (P ) as used tions from errors in the fit to the main lobe of the beam. by Downes et al. (1986). Based on the number densities of The subtracted phase calibrator (0.64 Jy intrinsic flux) radio sources at the fluxes of those presented here, the prob- is not included in this table. If one were interested in this abilities that radio sources WHDF-EVLA-1,WHDF-EVLA- source it would be more useful to simply image the phase cal- 6, WHDF-EVLA-8, and WHDF-EVLA-32 are chance align- ibrator scans during which the array was pointing directly at ments with the LABOCA sub-mm sources are P = 0.003, it rather than trying to correct for its complicated behaviour P = 0.023, P = 0.006 and P = 0.032 respectively. These in the target scans. probabilities of . 3% indicate that the radio and sub-mm signals do indeed originate from the same sources. Summarising, of the 5 sub-mm sources in the central WHDF area surveyed by EVLA, 4 are radio sources and 4 SUB-MM AND OPTICAL/NEAR-INFRARED 3 were already known to be candidate X-ray sources. The RADIO SOURCE COUNTERPARTS EVLA observations confirm the positional coincidence of 4.1 Sub-mm counterparts the three sub-mm source with X-ray sources on the as- sumption that the LABOCA and EVLA sources are the High redshift extragalactic sources which are luminous at same. One further LABOCA source is close to an EVLA sub-mm wavelengths are generally thought to be driven by source (02/WHDF-EVLA-32) but neither source is close to high star formation rates, with the high sub-mm flux in an optical/near-infrared object. The final sub-mm source the system arising due to the reprocessing of the intense (04) is unidentified in either radio or optical/near-infrared. stellar radiation by large amounts of dust. The low tem- Thus three out of five sub-mm sources are AGN as indi- perature of the dust is thought to favour such a heating cated by their X-ray properties and the other two are blank source rather than AGN. However the high redshift sub-mm in the optical/near-infrared. Two of the three X-ray sources sources clearly have the bolometric luminosities of AGN. are hard X-ray sources and likely to be cold gas absorbed. Moreover, explaining the observed sub-mm source counts as Although the numbers of sub-mm sources are small, the frac- being of purely star-formation origin in a Cold Dark Mat- tion that are X-ray sources is high and so is the gas-absorbed ter cosmology context requires invocation of top heavy ini- fraction. tial mass functions (Baugh et al. 2005). Recently, evidence Hill & Shanks (2011a), following Gunn & Shanks has mounted for AGN making significant contributions to (1999), suggested that the FIR flux re-radiated by the dust sub-mm source counts. (e.g. Lutz et al. 2010; Hill & Shanks might be proportional to the amount of X-ray radiation 2011a) following earlier identifications of AGN in sub-mm absorbed and in this case the obscured AGN contribution sources (eg Ivison et al. 1998, Brandt et al. 2001) to the sub-mm background might reach 40%. Hill & Shanks These findings motivated the 870 µm LABOCA ob- (2011a) found strong evidence for this through statistical servations of Bielby et al. (2012) as the WHDF conve- analyses of the Extended Chandra Deep Field South niently contains 15 spectroscopically confirmed quasars, four (ECDFS) X-ray and sub-mm data. Aided by the EVLA of which are classified as obscured via the assumption that confirmations of the identification of the 2 absorbed X- their high X-ray hardness ratio4 is caused by the presence ray quasars with the sub-mm sources means that there of large amounts of absorbing hydrogen. Eleven sub-mm is now further evidence for this hypothesis in the new sources were detected in the LABOCA observations, two WHDF/EVLA dataset. These points will be revisited in of which appeared to be associated with heavily obscured the discussion in Section 5. quasars. None of the unobscured quasars were near to robust sub-mm counterparts, although a stacking analysis revealed that absorbed X-ray quasars showed the most significant 4.2 Optical/near-infrared counterparts and sub-mm emission. Individual radio sources Since sub-mm observations of this kind have relatively This section presents possible (near-)infrared, visible and low resolution, radio observations have been used to de- ultraviolet counterparts for the radio sources and discusses tect counterparts for sub-mm sources by several authors noteworthy individual sources. Cut-out images from exist- (e.g. Ivison et al., 2002), with the improved positional ac- ing observations are plotted for each radio source in the curacy afforded by the radio observations then being used central 7’ × 7’ area of the WHDF in Figure 4, and in the to confirm counterparts at other wavebands. Recalling the extended area in Figure 5. Photometric bands U (365 nm), naming scheme defined in Section 3.1, the EVLA obser- B (445 nm), R (658 nm), I (806 nm) and H (1630 nm) are vations detect radio counterparts at the positions of four shown, with the telescope used indicated above each col- sub-mm sources: 06, 11, 05 and 02 corresponding to radio umn. The final column shows the radio image with over- sources WHDF-EVLA-1, WHDF-EVLA-6, WHDF-EVLA-8 laid contours.√The base contour level is 2σ and increases in and WHDF-EVLA-32. Radio sources WHDF-EVLA-6 and multiples of 2. Each thumbnail spans 25 arcseconds. The nearest counterparts to the radio sources are listed in Ta- 4 If H and S represent photon counts in the hard (2–8 keV) and ble 2. Photometric colours are listed, as are redshifts, where soft (0.5–2 keV) X-ray bands then the hardness ratio is defined available. as (H - S) / (H + S). WHDF-EVLA-1: The brightest radio source (aside from c 20XX RAS, MNRAS 000, 1–19 The micro-Jy source population at 8.4 GHz in the WHDF 7 ID Right Ascension Declination (σRA , σDec ) Flux density Radius Beam-corrected (J2000) (J2000) (arcsec) (µJy) (arcmin) flux density (µJy) ◦ WHDF-EVLA-1 0h 22m 31.82s +0d 21m 28.03s (0.1,0.1) 164.42 (±6.0) 1.389 201.23 (±7.55) ◦ WHDF-EVLA-2 0h 22m 15.86s +0d 21m 28.32s (0.08,0.09) 106.23 (±4.0) 2.884 275.71 (±18.4) WHDF-EVLA-3 0h 21m 50.09s +0d 22m 3.15s (0.64,0.77) 87.58 (±10.0) 9.33 N/A ◦ WHDF-EVLA-4 0h 22m 31.07s +0d 22m 24.81s (0.13,0.12) 70.88 (±4.0) 1.939 106.16 (±6.84) ◦ WHDF-EVLA-5 0h 22m 29.15s +0d 18m 19.78s (0.24,0.26) 43.85 (±5.0) 2.482 86.98 (±9.76) ◦ WHDF-EVLA-6 0h 22m 24.89s +0d 20m 10.21s (0.68,0.61) 43.1 (±7.0) 0.794 45.99 (±7.81) WHDF-EVLA-7 0h 22m 39.97s +0d 18m 38.2s (0.31,0.28) 33.38 (±5.0) 3.867 N/A ◦ WHDF-EVLA-8 0h 22m 22.85s +0d 20m 13.98s (0.48,0.43) 32.51 (±5.0) 1.169 37.47 (±6.28) ◦ WHDF-EVLA-9 0h 22m 32.6s +0d 23m 28.48s (0.36,0.36) 29.11 (±5.0) 3.053 86.04 (±14.85) ◦ WHDF-EVLA-10 0h 22m 20.53s +0d 19m 1.52s (0.55,0.56) 28.3 (±6.0) 2.375 52.75 (±10.6) WHDF-EVLA-11 0h 21m 55.42s +0d 25m 41.58s (1.37,1.3) 26.01 (±8.0) 9.32 N/A WHDF-EVLA-12 0h 22m 10.95s +0d 19m 50.99s (0.89,0.66) 25.52 (±6.0) 4.122 N/A WHDF-EVLA-13 0h 22m 54.37s +0d 30m 39.98s (1.85,1.59) 24.75 (±9.0) 12.036 N/A WHDF-EVLA-14 0h 22m 24.83s +0d 16m 47.18s (1.37,1.25) 24.66 (±7.0) 4.008 N/A ◦ WHDF-EVLA-15 0h 22m 30.09s +0d 19m 41.63s (1.04,0.94) 23.09 (±7.0) 1.31 27.62 (±8.02) ◦ WHDF-EVLA-16 0h 22m 14.93s +0d 21m 4.12s (0.5,0.43) 22.97 (±5.0) 3.046 67.52 (±14.43) WHDF-EVLA-17 0h 22m 42.96s +0d 19m 56.28s (1.35,1.01) 21.34 (±7.0) 4.063 N/A WHDF-EVLA-18 0h 22m 37.79s +0d 23m 36.78s (0.67,0.46) 21.15 (±5.0) 3.922 N/A ◦ WHDF-EVLA-19 0h 22m 22.66s +0d 20m 42.82s (0.46,0.59) 20.72 (±5.0) 1.093 23.45 (±5.46) WHDF-EVLA-20 0h 22m 53.44s +0d 16m 57.03s (1.48,1.04) 20.17 (±7.0) 7.618 N/A WHDF-EVLA-21 0h 22m 23.89s +0d 11m 14.61s (1.04,1.21) 20.12 (±7.0) 9.547 N/A WHDF-EVLA-22 0h 21m 58.45s +0d 22m 41.49s (1.26,1.17) 18.88 (±7.0) 7.405 N/A ◦ WHDF-EVLA-23 0h 22m 32.02s +0d 18m 11.78s (1.51,1.02) 18.67 (±7.0) 2.846 47.13 (±17.25) WHDF-EVLA-24 0h 21m 47.32s +0d 15m 59.91s (1.19,1.33) 17.94 (±7.0) 11.011 N/A ◦ WHDF-EVLA-25 0h 22m 36.86s +0d 20m 55.57s (0.86,1.1) 16.43 (±6.0) 2.461 32.19 (±11.18) WHDF-EVLA-26 0h 22m 13.91s +0d 18m 5.77s (0.98,0.82) 15.95 (±5.0) 4.225 N/A ◦ WHDF-EVLA-27 0h 22m 16.16s +0d 19m 46.78s (0.97,1.04) 15.01 (±5.0) 2.891 39.16 (±13.9) WHDF-EVLA-28 0h 22m 15.75s +0d 31m 11.39s (1.31,1.23) 15.0 (±6.0) 10.808 N/A ◦ WHDF-EVLA-29 0h 22m 21.64s +0d 21m 49.48s (0.87,1.02) 13.8 (±5.0) 1.722 18.91 (±7.06) WHDF-EVLA-30 0h 22m 31.64s +0d 31m 25.44s (0.8,1.24) 12.99 (±5.0) 10.729 N/A WHDF-EVLA-31 0h 22m 59.13s +0d 11m 20.34s (0.68,0.75) 12.95 (±4.0) 12.372 N/A ◦ WHDF-EVLA-32 0h 22m 28.25s +0d 21m 47.24s (0.79,0.54) 12.94 (±4.0) 1.073 14.58 (±4.88) WHDF-EVLA-33 0h 21m 43.99s +0d 18m 5.18s (0.79,0.73) 12.11 (±4.0) 11.09 N/A WHDF-EVLA-34 0h 21m 43.24s +0d 11m 24.65s (1.07,1.26) 11.96 (±5.0) 14.397 N/A WHDF-EVLA-35 0h 22m 59.7s +0d 12m 34.81s (1.21,0.86) 11.93 (±5.0) 11.554 N/A WHDF-EVLA-36 0h 21m 59.92s +0d 20m 41.05s (0.96,1.01) 11.09 (±5.0) 6.779 N/A WHDF-EVLA-37 0h 21m 47.34s +0d 12m 23.71s (0.54,0.76) 10.71 (±4.0) 12.977 N/A WHDF-EVLA-38 0h 22m 30.07s +0d 30m 2.58s (0.91,0.64) 10.3 (±4.0) 9.316 N/A ◦ WHDF-EVLA-39 0h 22m 30.84s +0d 22m 53.9s (0.8,0.79) 10.08 (±4.0) 2.345 18.46 (±7.65) WHDF-EVLA-40 0h 22m 48.38s +0d 14m 49.05s (0.72,0.56) 9.36 (±4.0) 7.982 N/A WHDF-EVLA-41 0h 22m 11.57s +0d 25m 41.96s (0.68,0.67) 9.27 (±4.0) 6.277 N/A Table 1. Properties of the 41 radio sources detected in the EVLA map centred on the WHDF and presented in Figure 2. Sources marked with a circle (◦) in the left-hand column are within the region of the main lobe of the EVLA primary beam where the direction dependent gain is greater than 0.2. These are corrected with a primary beam model, thus the flux density values in bold face type represent intrinsic values, whereas the ones in plain type are apparent values. Please refer to the text for further details. 4C 00+02) in the central region of the WHDF corresponds cal colours show an ultraviolet excess (UVX) while the R-H to a spiral galaxy at z = 0.046 which is also associated with colours are quite red. At HST resolution the source may a sub-mm LABOCA detection and an X-ray source. just be resolved. The redder near-infrared colours may be WHDF-EVLA-3: This source is a low redshift explained by host domination in these bands. Assuming all (z = 0.2609) galaxy. the R band flux comes from the QSO and R-K ≈ 2 (e.g. Mad- WHDF-EVLA-4: Near infrared imaging of this source dox et al., 2012), the QSO would have K ≈ 20.6 implying shows an edge-on spiral which is clumpy in ultraviolet imag- K ≈ 18.1 for the host galaxy. Such galaxy magnitudes are ing. not uncommon at this redshift in surveys such as the K20 WHDF-EVLA-6: This radio source corresponds to one of survey (Cimatti et al., 2002), and as evidenced further by the LABOCA-detected obscured quasars at z = 1.33 and a the K-band luminosity function derived from the UKIDSS faint infrared counterpart can be seen at the centre of the Ultra Deep Survey (Cirasuolo et al., 2010). Note also the major radio peak. The radio emission is extended and re- general tendency for AGN at all epochs to lie in the most solved into two components. There is no clear optical or in- massive galaxies (Dunlop et al., 2003). frared counterpart for the lower radio peak suggesting that WHDF-EVLA-8: The second LABOCA-detected ob- the quasar may host a radio jet. Paradoxically, the opti- scured quasar at z = 2.12. In the HST I band image the c 20XX RAS, MNRAS 000, 1–19 8 Heywood et al. Radio ID R-band RA R-band Dec Offset R U-B B-R R-I R-Z R-H H-K Redshift (J2000) (J2000) (′′ ) mag Colours within 3′′ diameter apertures WHDF-EVLA-1 0h 22m 31.86s +0d 21m 27.5s 0.8 16.07 0.04 1.42 0.73 0.89 2.49 0.32 0.046 WHDF-EVLA-2 0h 22m 15.87s +0d 21m 28.8s 0.5 19.20 0.52 2.70 0.92 1.11 3.04 - 0.259 WHDF-EVLA-4 0h 22m 31.04s +0d 22m 24.7s 0.5 23.29 0.32 2.23 1.64 1.82 4.62 1.22 0.961 WHDF-EVLA-5 0h 22m 29.12s +0d 18m 19.2s 0.7 21.62 -0.12 2.01 0.97 1.32 3.59 0.79 0.433 WHDF-EVLA-6 0h 22m 24.83s +0d 20m 11.8s 1.8 22.63 -0.72 1.04 1.06 1.33 3.71 0.98 1.32 WHDF-EVLA-7 0h 22m 39.94s +0d 18m 38.6s 0.6 22.44 1.20 3.42 1.67 2.05 4.27 0.90 - WHDF-EVLA-8 0h 22m 22.81s +0d 20m 14.6s 0.9 24.02 -1.25 0.79 0.71 1.12 3.44 0.85 2.11 WHDF-EVLA-9 0h 22m 32.56s +0d 23m 27.6s 1.1 19.31 -0.29 1.62 0.78 1.04 2.72 0.76 0.382 WHDF-EVLA-10 0h 22m 20.48s +0d 19m 0.3s 1.4 20.15 -0.18 1.76 0.86 1.04 2.82 0.78 0.432 WHDF-EVLA-12 0h 22m 11.08s +0d 19m 50.9s 2.0 24.18 -0.33 0.48 0.47 - 3.85 - - WHDF-EVLA-14 0h 22m 24.99s +0d 16m 47.8s 2.5 21.74 -0.26 1.58 0.43 0.57 1.56 - - WHDF-EVLA-16 0h 22m 14.86s +0d 21m 4.2s 1.1 20.50 0.06 1.92 0.88 2.06 3.16 - - WHDF-EVLA-17 0h 22m 42.84s +0d 19m 54.8s 2.3 21.60 0.02 2.24 0.96 1.16 2.92 0.75 - WHDF-EVLA-18 0h 22m 37.79s +0d 23m 38.1s 1.3 17.39 -0.14 1.31 0.70 0.82 2.42 0.56 - WHDF-EVLA-19 0h 22m 22.62s +0d 20m 43.0s 0.6 22.12 -0.27 0.87 0.56 0.89 2.72 1.11 - WHDF-EVLA-20 0h 22m 53.22s +0d 16m 59.8s 4.3 18.43 0.06 1.69 0.78 0.80 2.57 - - WHDF-EVLA-23 0h 22m 32.09s +0d 18m 11.1s 1.3 24.91 0.16 1.29 1.36 - 4.79 0.73 - WHDF-EVLA-25∗ 0h 22m 36.78s +0d 20m 56.5s 1.5 19.82 - - - - - 1.01 - WHDF-EVLA-27 0h 22m 16.22s +0d 19m 45.7s 1.4 21.39 -0.21 2.01 0.78 1.06 2.51 - - WHDF-EVLA-29 0h 22m 21.80s +0d 21m 47.3s 3.2 22.29 -0.68 0.83 0.91 0.66 2.52 0.88 - WHDF-EVLA-32 0h 22m 28.24s +0d 21m 53.5s 5.0 22.62 -0.17 0.97 0.41 -0.11 1.51 - - WHDF-EVLA-39 0h 22m 30.95s +0d 22m 54.5s 1.8 23.77 -0.41 1.47 1.00 1.31 3.63 1.26 - ∗ Position and magnitude are taken from the H-band image. Table 2. Optical and near-infrared properties of the nearest counterpart within 5′′ of the positions of the radio sources in Table 1. Magnitude, colours and positions are taken from imaging described in Metcalfe et al. (2001) and Metcalfe et al. (2006). Spectra are unpublished data taken with LDSS2 on the Magellan 6.5m telescope. Radio ID LABOCA Comment R U-B 870µm 8.4GHz LAB-EVLA Redshift ID mag (mJy) (µJy) Separation WHDF-EVLA-6 LAB-11 absorbed QSO/UVX 22.63 -0.72 3.4 48 7.′′ 7 1.33 WHDF-EVLA-8 LAB-05 absorbed QSO/UVX 24.02 -1.25 4.0 37 3.′′ 1 2.12 WHDF-EVLA-15 - blank - - - 28 - - WHDF-EVLA-19 - merger-train 22.12 -0.27 - 23 - - WHDF-EVLA-23 - stellar - QSO? 24.91 0.16 - 47 - - WHDF-EVLA-25∗ - i dropout 19.82 - - 32 - - WHDF-EVLA-27 - merger-double 21.39 -0.21 - 39 - - WHDF-EVLA-29 - blank - - - 19 - - WHDF-EVLA-32 LAB-02 blank - - 4.3 15 5.′′ 4 - WHDF-EVLA-39 - merger-train/UVX 23.77 -0.41 - 18 - - WHDF-EVLA-1 LAB-06 spiral galaxy 16.07 0.04 3.9 201 5.′′ 5 0.046 ∗ Magnitude is taken from the H-band image. Table 3. Summary of Tables 1, 2 and Table 1 of Bielby et al. (2012) for complete sample of 10 faint EVLA sources with S < 50µJy at 8.4GHz. For completeness of the sub-mm fluxes and LABOCA-EVLA separations, WHDF-EVLA-1 is also listed, although too bright at 8.4GHz for inclusion in the faint sample. object has no nucleus and looks like an edge-on spiral, but a WHDF-EVLA-19: This object shows a stellar nucleus in nucleus is seen both in the blue and the near-infrared imag- the ultraviolet, is nearly UVX and appears among a string ing. Again, the colours of the quasar are anomalous, having of fainter objects. It is a possible quasar or merger. U-B = -1.25, i.e. very blue, whereas R-H = 3.44 i.e. very red. Again, the redder near-infrared colours may be explained by WHDF-EVLA-23: This source has a stellar appearance increasing host domination in these bands. On the same as- in the HST I band images. It is very red at R-H and H-K. sumptions as for WHDF-EVLA-6 the host galaxy could have The U-B colour is blue but not UVX so it could be a z & 2.2 K ≈ 19.8 and galaxies are seen out to z ≈ 2 at this limit in QSO or a modestly reddened QSO at lower redshift, perhaps the K20 survey. contaminated in the infrared by the host galaxy. WHDF-EVLA-9: This is a low-redshift spiral galaxy WHDF-EVLA-25: This is a drop-out at I-band and other (z = 0.382) which harbours an optical jet and has associated optical bands. It is only detected in H imaging, where it is X-ray emission. relatively bright. c 20XX RAS, MNRAS 000, 1–19 The micro-Jy source population at 8.4 GHz in the WHDF 9 4.3 Faint radio source counts at 8.4 GHz Determining the counts of extragalactic sources at radio wavelengths has been an active area of study for several decades; de Zotti et al. (2010) present a review of both the history and the state of the art. Early radio surveys provided a key fulcrum in Steady State versus Big Bang cosmology debates in the 1950s, and observations since then have re- vealed much about cosmology and the evolution of radio sources with cosmic time, and have (particularly at higher frequencies) proved essential for categorising extragalactic foregrounds for Cosmic Microwave Background experiments. Source counts at the faint (61 mJy) end of the distribution exhibit a turn-up. This is generally explained by the increase in the dominance of star-forming galaxies over AGN at these low luminosities (e.g. Padovani et al., 2009) and it has also been claimed that radio-weak AGN make a significant con- tribution (e.g. Jarvis & Rawlings, 2004, Smolˇci´c et al., 2009, Simpson et al., 2006, 2012), although the exact nature of this excess remains a source of debate. While most studies of source counts derived from radio surveys have been con- ducted at L-band, the turn up persists in higher frequency observations including those such as the X-band observa- tions presented in this paper. Figure 3. Positions and IDs of the radio sources listed in Table The recent Absolute Radiometer for Cosmology, As- 1 and the coverage of the optical and infrared frames from which trophysics and Diffuse Emission (ARCADE2) experiment the thumbnails in Figure 4 have been extracted. The concentric (Fixsen et al., 2011) has also piqued interest in the faint grey contours show the main lobe of the EVLA primary beam. end of radio source populations due to the measured excess The outer contour is the 20% gain level and the contours have in the sky brightness temperature at 3 GHz (Seiffert et al., increments of 10%. Also shown is the area covered by the extended 2011). Vernstrom et al. (2011) use source count data from WHDF observations. 150 MHz to 8.4 GHz to predict the contribution to the sky temperature background from measured source populations and conclude that if the ARCADE2 result is correct there must be an additional significant population of radio galax- WHDF-EVLA-27: This is a possible double object, a can- ies at fluxes fainter than those hitherto reached by radio didate merger with relatively blue colours. surveys. This clearly motivates the need for deeper radio WHDF-EVLA-39: The UVX here implies probable continuum surveys. quasar. Infrared and HST I-band imaging resolves this The source counts derived from the EVLA observations source into multiple components so it could be a merger. of the WHDF do not reach the depths needed to begin to Otherwise, a possible strong gravitational lens? In the lat- address the ARCADE2 result; the counts are presented here ter case, the uppermost pair of images may form a partial for completeness. It is noteworthy however that the depth Einstein ring. of the observations in this paper (2.5 µJy) is approaching Of the 12 sources within the ≈ 5′ half-power point, 9 that of the deepest 8.4 GHz observations (1.49 µJy in the have optical/near-infrared counterparts within 2′′ and 3 are SA13 field; Fomalont et al. 2002) which required 190 hours UVX. 4 are also LABOCA sources. Similarly, in the full area, of integration time with the old VLA system. The survey of the complete sample of 10 faint (< 50µJy) sources (see speed advantage is brought about by the huge increase in Table 3) 6 have optical/near-infrared counterparts of which available bandwidth (∆ν), since the noise level in a syn- 3 are UVX (i.e. likely quasars) including the two absorbed thesis image σ ∝ ∆ν −0.5 . Increasing the bandwidth of the quasar LABOCA sources. The other 3 faint radio counter- observation does not change the traditional limiting factor parts are not UVX but are only slightly less blue and likely for high frequency survey work, which is that the sky area to be star-forming galaxies, predominantly at lower lumi- covered by a single pointing is proportional to ν −2 where nosities and redshifts. Of the 10, there are 3 possible merg- ν is the observing frequency, but it does mean that far less ers, one of which contains a UVX source. The 4 faint opti- time is required to reach a certain depth per pointing. High cally unidentified radio sources may be either dust obscured frequency surveys covering significant sky areas thus become quasars or galaxies. One of these sources is identified only far more economical. Note that since the WHDF radio ob- at H and K and it appears resolved even at ground-based servations were taken the bandwidth available for general resolution. users of the EVLA has increased by a factor of 8. The LABOCA 870µm fluxes of the four radio sources The source counts from the WHDF observation are with LABOCA counterparts are also given for completeness shown in Figure 6 along with three other previously pub- in Table 3. Their high ratio of sub-mm relative to radio flux lished data sets. The counts are normalized to those ex- suggests that all four sources are likely to be dust rather pected in a non-expanding Euclidean universe for ease of than synchrotron dominated in the sub-mm. comparison to the published data. The origins of the previ- c 20XX RAS, MNRAS 000, 1–19 10 Heywood et al. Figure 4. Near infrared / ultraviolet thumbnail images centred on the EVLA radio sources within the central area of the WHDF. The radio sources are presented in the right hand column. Photometric wavebands and the instrument used are noted above each column, √ radio source IDs are noted at the end of each row. The base contour level on the radio cutouts is 2σ and increases in multiples of 2. Each thumbnail spans 25 arcseconds. This figure continues on the next page. c 20XX RAS, MNRAS 000, 1–19 The micro-Jy source population at 8.4 GHz in the WHDF 11 Table 4. Source counts and bin details for the data plotted in which were collated and listed by de Zotti et al. (2010). Only Figure 6. the 17 sources from the WHDF observations which could be reliably corrected for beam attenuation effects were included Bin Central flux (µJy) Bin width (µJy) N in the count. Omitting this step would clearly artificially bias the counts towards fainter levels. Sources were counted 1 15.2 10.4 3 in five bins between 10 and 350 µJy with logarithmically 2 30.9 22.0 5 increasing widths. Error bars are simply derived from Pois- 3 62.9 43.0 4 4 128.2 87.5 3 son statistics. The solid angle covered by the single pointing 5 260.9 178.1 2 within the region where the beam gain is greater than a factor of 0.2 is 0.0119 deg2 for the central frequency. For reference, also shown on Figure 6 are the source ously published points are shown on the figure, the values of counts predicted by two models down to a flux limit of c 20XX RAS, MNRAS 000, 1–19 12 Heywood et al. Figure 5. Near infrared / ultraviolet thumbnail images centred on the EVLA radio sources within the extended area of the WHDF. As per Figure 4, photometric wavebands and the instrument used are noted above each column. Radio source IDs are noted at the end of each row and the radio sources √ themselves are shown in the right hand column. The base contour level on the radio cutouts is 2σ and increases in multiples of 2. Each thumbnail spans 25 arcseconds. 1 µJy. The small circles show the source counts generated model counts at the faintest limits may explain the more by binning all ∼260 million galaxies in the full 20 × 20 deg2 ≈1:1 split of AGN (both radio-loud and -weak) and star- sky area of the semi-empirical galaxy simulation of Wilman forming/merging galaxies seen in Table 2 compared to the et al. (2008). Linear extrapolation of the 4.86 and 18 GHz ≈2:1 ratio in favour of star-forming galaxies predicted by fluxes offered by the simulation was used to generate the the model. Further discussion of the nature of the faint ra- source count data. The solid line shows the model of de dio source population in the WHDF is presented in Sections Zotti et al. (2005), solely for AGN-powered sources. The 5 and 6. faint count data lie above the AGN model, illustrating the need for another component at faint fluxes. In the Wilman et al. (2008) model, this is represented by star-forming galax- ies (67%) and also radio-weak AGN (20%), with the lat- ter therefore making a comparable contribution to radio- loud AGN (13%)5 . The apparent overestimation of the total then binned according to their star formation and AGN types: 5 These fractions were determined by extracting a sample of 20508 radio quiet AGN; 11694 FR-I; 0 FR-II; 1069 gigahertz galaxies with fluxes between 10 and 50 µJy from 25 square degrees peaked spectrum; 55894 quiescent star forming galaxies; 11279 of the Wilman et al. (2008) simulation. These 100444 sources were starburst galaxies. c 20XX RAS, MNRAS 000, 1–19 The micro-Jy source population at 8.4 GHz in the WHDF 13 increases with hydrogen column density, the re-radiated far- infrared flux increases in proportion and they therefore pre- dict that there will be more sub-mm radiation from X-ray absorbed AGN. Such explanations therefore have implica- tions for the unified model (e.g. Antonucci, 1993) which holds that the differing X-ray and optical properties of ob- scured and unobscured AGN can be explained in terms of the inclination of the system with respect to the line of sight of the observer. The amount of sub-mm emission, originating in the dusty torus surrounding the system, should then be independent of its orientation. The differing sub-mm proper- ties between two classes of AGN thus cannot be explained by invoking the unified model. Previously it was suggested that sub-mm-bright AGN are being observed during a different evolutionary phase, whereby the AGN experiences a period of growth within a dusty, starforming galaxy environment (Page et al., 2004). It is speculated here that X-ray absorbed AGN might be expected to be strong sub-mm sources if the presence of cold, neutral gas in an AGN powered source also indicated the presence of cold dust. In this case the dust torus would need to extend up to ∼1 kpc to maintain a dust temperature as low as 30–35 K (e.g. Kuraszkiewicz et al. 2003). It is also noted that the optical and near-infrared colours of the two X-ray absorbed quasars are anomalous, being quite UVX in U-B yet quite red in R-H. If this is not due to variability then it might suggest that the Figure 6. Euclidean-normalized differential source counts at 8.4 optical colours are active nucleus dominated whereas the GHz. Observational data are from Windhorst et al. (1993), Fo- near-infrared colours are host dominated. Clearly this could malont et al. (2002) and Henkel & Partridge (2005), collated by favour a picture where the nuclear sight line is cold gas ab- de Zotti et al. (2010), and also from this paper. Only sources which have been corrected for primary beam attenuation were sorbed as evidenced by both the hard X-rays and the optical used when determining the source counts for the latter. Sources narrow lines. But the dust would then have to form a more were placed into five bins of logarithmically increasing width be- clumpy or toroidal structure around the nucleus to leave the tween 10 and 350 µJy, the lower limit being equivalent to the nuclear sight-line unobscured by dust. The fact that narrow, completeness limit. The error bars on the counts from this paper high-ionisation lines like C IV are seen suggest the gas ab- are derived from Poisson statistics. Simulated source counts are sorption is on smaller scales than the narrow line region, from de Zotti et al. (2005) and from Wilman et al. (2008). however there appears to be little dust on the nuclear sight- line. This behaviour may be more characteristic of a unified model but such a model does not explain the basic sub-mm 5 DISCUSSION - neutral gas absorption correlation. A non-unified model The EVLA survey of the WHDF confirms the associa- which does explain this correlation would then also need tions between LABOCA sub-mm sources and X-ray ab- to invoke a mechanism, maybe a jet, to destroy cold dust sorbed AGN. Two of the three X-ray sources associated just along the nuclear sightline while leaving the neutral gas with LABOCA sources are hard X-ray sources and likely more or less in place. It is interesting that the sub-mm ab- to be cold gas absorbed, with the third X-ray source be- sorbed X-ray AGN, WHDF-EVLA-6, does show evidence of ing associated with the nucleus of a nearby galaxy and too a radio jet. faint to measure its hardness ratio. All three sub-mm+X- The two absorbed QSOs also have radio and ray sources are associated with radio sources. Although the sub-mm fluxes that put them close to the FIR- numbers are small, the X-ray fraction is high and so is the radio correlation (e.g. Jarvis et al. 2010 and ref- gas-absorbed fraction. Previously, Page et al. (2004), who erences therein). For example, WHDF-EVLA-08 at studied samples of absorbed (density of Hydrogen atoms z = 2.12 has S870µm = 6.9 × 1025 WHz−1 and 23 −1 NH > 1022 cm−3 ) and unabsorbed quasars, suggested a sim- L8.4GHz = 8.6 × 10 WHz . An approximate con- ilar difference in the sub-mm properties of these two classes version assuming dust temperature components of 35- of sources. Hill & Shanks (2011a) themselves found that the 60 K in the wavelength range 8-1000 µm gives LABOCA ECDFS sub-mm Survey (LESS, Wardlow et al. LF IR ≈ 1 × 1039 W = 3 × 1012 L⊙ . Similarly, an approxi- 2011) showed clear evidence for ≈ 20% of sub-mm sources mate conversion of the radio flux assuming a ν −1 spectrum being associated with X-ray absorbed AGN. gives L1.4GHz ≈ 5 × 1024 WHz−1 . This gives qIR ≈ 1.75 (see Models whereby absorbed AGN are used to fit the X- Jarvis et al 2010) compared to the average for star-forming ray background with the absorbed photons being re-radiated galaxies of qIR ≈ 2.2, indicating some degree of contamina- in the infrared have been discusssed by Gunn (1999) and tion by non-thermal radio emission. Gunn & Shanks (1999) and more recently by Hill & Shanks Hill & Shanks (2011a) have argued that in a non-unified (2011a). In these models, as the absorbed X-ray fraction model the absorbed AGN may make up a significant frac- c 20XX RAS, MNRAS 000, 1–19 14 Heywood et al. tion of the bright sub-mm sources. Also, Mart´ınez-Sansigre For QSO dust components at higher temperatures mea- et al. (2005) have inferred from radio observations that most sured at shorter wavelengths, there may be more evidence black hole accretion is obscured, implying that the contri- for a unified model from finding correlations between AGN bution of obscured AGN to the sub-mm background may be radio and IR/FIR dust emission if the radio flux is assumed very significant. to be orientation independent. Shi et al. (2005) found some This non-unified picture where absorbed AGN are pref- indication of a correlation between 70µm and radio fluxes erentially sub-mm loud is also supported by several more for a sample of AGN observed with Spitzer MIPS. Previ- recent results. For example, Page et al. (2012) find that ously, Polletta et al. (2000) found that in a sample of 22 most of the AGN detected in the Chandra Deep Field AGN, the amount of IR/FIR dust emission was reasonably North at 250 µm by Herschel SPIRE show strong X-ray ab- independent of whether the AGN were radio loud or not, sorption, although these authors emphasised more the lack again broadly consistent with a unified picture. Whether of far-infrared emission from the intrinsically bright X-ray these results present a problem for either of the non-unified AGN. Simpson et al. (2012) have also reported that two high scenarios discussed above awaits further data. redshift 2SLAQ quasars detected in Herschel ATLAS as The identification of optical/near-infrared counterparts strong far-infrared sources appear to show heavily absorbed for 10 faint (<50 µJy) radio sources show a mixture of ≈ Lyman-α lines. Rovilos et al. (2012, priv. comm.) using Her- 30% lower redshift, blue, starforming/merging radio galaxies schel PACS/SPIRE have also found that the cold dust com- (3/10) and ≈ 30% higher redshift quasars/AGN (3/10), in- ponent in ECDFS AGN is as significant a component in cluding the absorbed quasars identified as sub-mm sources. AGN spectral energy distributions (SEDs) as their hot dust The remaining 4/10 faint sources are optically blank, al- components. Hickox et al. (2012) have also found that the though one is detected in the near-infrared as a resolved galaxy group environment and clustering of sub-mm sources source. Otherwise, there are few sources identified with red, are completely consistent with those of QSOs. Finally, a crit- early-type galaxies. Thus it might be concluded that this icism of the Hill & Shanks (2011a) model is that the dust mix of sources at 8.4 GHz is not dissimilar to the mix of mass required will have to be around 5 × 106 M⊙ for L∗ sources found in the sub-mm samples at 350 GHz. Indeed, QSOs, implying a cold gas mass of perhaps ≈5 × 108 M⊙ . Hill & Shanks (2011b) suggested that this mix persists to It is interesting to note that Molinari et al. (2011) have de- the Herschel SPIRE bands at 350 µm, i.e. 850 GHz, with tected a ≈100 pc ring with ≈5 × 107 M⊙ of gas and dust the AGN component then becoming increasingly less domi- around the Sgr A black hole candidate in the centre of the nant at higher frequencies. The results presented above seem Milky Way. If this was associated with a ≈1043 ergs−1 AGN to support the idea that a strong AGN component is seen at outburst then it would at least be of the right size and mass all frequencies from the radio to the far-infrared, with the to produce a cold sub-mm component in the Milky Way. AGN component synchrotron-dominated at low frequencies A more typical non-unified model invokes an evolu- and increasingly cold dust-dominated at higher frequencies. tionary relation between absorbed and unabsorbed QSOs (e.g. Sanders et al., 1988, Fabian 1999). Here a merger trig- gers the birth of an obscured QSO, accompanied by a burst of star-formation. Once the spheroid is formed, expulsion of cool gas and dust quenches the star-formation, turns off the sub-mm emission and leaves the QSO unobscured. This model implies that more absorbed QSOs will be more 6 CONCLUSIONS FIR/sub-mm bright. Such a model would thus also be con- sistent with X-ray absorbed QSOs being preferentially sub- Radio continuum observations can now achieve extreme mm bright as observed in this paper. The main difference depths with relatively short integration times due to the with the model of Hill & Shanks (2011a) is that there the vastly increased correlator bandwidths that are now avail- dust is AGN heated whereas in this other case it is star- able. This makes surveys much more economical, thus burst heated. Hill & Shanks (2011a) argued that starburst prospects are good for future, deep high-frequency surveys heating leaves the similarity of the sub-mm sources’ lumi- covering significant sky areas with instruments such as the nosities to that of QSOs looking like a coincidence. High res- EVLA and MeerKAT (e.g. Jarvis, 2011; Heywood et al., olution ALMA observations could also measure the size of 2011). There is a caveat here however: both deeper sensi- the sub-mm emitting region and help differentiate between tivity limits and broad bandwidths conspire to increase the these possibilities. calibration complexity. Direction dependent effects (which Herschel ATLAS results from Bonfield et al. (2011) find are often also highly frequency-dependent) need to be prop- a correlation between optical QSO and FIR luminosity at erly accounted for during calibration in order to remove the fixed redshift as well as with redshift. This is consistent artefacts associated with off-axis sources. As observations with the model of Hill & Shanks (2011a) since at fixed col- are pushed deeper these effects become more apparent (see umn and redshift, higher optical/X-ray luminosity implies also Smirnov 2011), and if the science targets are extremely higher FIR luminosity. The amount of sub-mm emission faint they are in danger of being swamped by residual cali- from QSOs with relatively low absorption of log(NH ) ≈ 20.5 bration errors. It is fair to describe the location of the con- or AV ≈ 0.1-0.2 mag, implies that 10-20% of the QSO optical fusing source which blighted the WHDF observations as the luminosity will be re-radiated by dust, implying no disagree- ‘worst-case scenario’ (see Appendix for details) yet the suc- ment with the substantial FIR luminosities for these objects cess with which it was removed is encouraging if future deep as observed by Bonfield et al. (2011; see also Hatziminaoglou radio continuum surveys are to routinely produce images et al., 2010, Serjeant et al., 2010). which are limited by thermal noise as opposed to artefacts c 20XX RAS, MNRAS 000, 1–19 The micro-Jy source population at 8.4 GHz in the WHDF 15 brought about by deficiencies in either the model of the sky The South East Physics Network (SEPnet). This work is or the instrument.6 based upon research supported by the South African Re- Applying these techniques to EVLA data, a deep 8.4 search Chairs Initiative of the Department of Science and GHz radio image of an area covering the Extended WHDF Technology and National Research Foundation. We thank has been generated, and a catalogue of 41 radio sources with Gianfranco de Zotti for providing the model count data flux densities exceeding 4σ has been derived. The central, which features in Figure 6, and Walter Brisken for his Cass- deepest area of the WHDF is well matched to the main lobe beam software which was used to generate the EVLA an- of the EVLA primary beam, and within this area 17 sources tenna beam patterns in Figure A2. We also thank the other are detected that can have their apparent fluxes corrected participants of the Second Workshop on 3rd Generation Cal- for the beam attenuation. ibration in Radio Astronomy, where a discussion session in- Two of these radio sources (WHDF-EVLA-6 and spired the simulations in the Appendix. We thank Vivek WHDF-EVLA-8) confirm the association of two high red- Dhawan for useful discussions regarding the EVLA pointing shift (z = 1.33 and 2.12 respectively) X-ray absorbed quasars accuracy. Some of the figures in this paper were produced with the sub-mm sources detected by Bielby et al. (2012). with APLpy, an open-source astronomical plotting package Such sources warrant further investigation due to the uncer- for Python (http://aplpy.github.com). This research has tainty surrounding the contributions that AGN make to the made use of NASA’s Astrophysics Data System. This paper sub-mm background. Both the sub-mm absorbed quasars is dedicated to Steve Rawlings. show unabsorbed nuclear colours in the blue that would re- quire a mechanism to remove cold dust but not neutral gas from the quasar sightline. It is interesting that one of the sub-mm sources, WHDF-EVLA-6, appears to harbour a ra- APPENDIX A: INTERPRETATION OF THE dio jet. Certainly the EVLA + LABOCA results from the DIFFERENTIAL GAIN SOLUTIONS WHDF support the previous result of Hill & Shanks (2011a) Self-calibration of a radio interferometric data set involves from ECDFS/LESS that X-ray absorbed AGN are signifi- the generation of a model visibility set from assumed models cant sub-mm sources. This in turn may support their non- of the sky and the instrument. Model visibilities are usually unified absorbed AGN X-ray background model where AGN predicted by evaluating the radio interferometer measure- dust emission is proportional to their X-ray column which ment equation (RIME; e.g. Smirnov, 2011a). In the most predicts that up to ≈40% of the sub-mm background may commonly employed method of calibration the RIME has a be due to AGN. solvable complex gain term per feed, per antenna, and a nu- The beam-corrected source fluxes are used to determine merical algorithm is used to minimise the difference between differential source counts which are in good agreement with the observed and the model visibility data. The best-fitting previously published values covering the distribution down complex gain terms (which are functions of both time and to ≈50-100 µJy and the semi-empirical extragalactic simu- frequency) are applied to correct the observed data and re- lation of Wilman et al. (2008). The counts show a signifi- move the instrumental gain drifts. These corrected data are cant excess over the purely AGN-powered count model of then imaged for deconvolution or to refine the sky model for de Zotti et al. (2005). The optical identifications suggest further calibration. the WHDF faint source counts are composed of 30% high In the case where there is a single dominating source redshift AGN, 30% low redshift star-forming galaxies and in the field (e.g. a typical observation of a phase calibrator) 40/30% optical/near-infrared blank fields, likely either to both the observed and model visibility function, and the be dusty star-forming galaxies or the more heavily absorbed associated complex gain corrections, will be dominated by AGN needed to explain the X-ray background. Various in- the contribution of this source. If the dominating source is dividually interesting counterparts within this faint radio away from the pointing centre (as is the case for the observa- source population have been noted. tions presented in this paper) then the time- and frequency- Future EVLA observations over a wider area of the dependent behaviour of the complex gain solutions derived WHDF will test how representative the results for the from self-calibration will be dominated by any instrumental present area are. Deeper optical and near-infrared obser- or atmospheric effects which are specific to the direction to- vations are needed to determine further the nature of the wards that source. In the case where the target field contains population of faint radio sources responsible for both the many strong sources, each of which is subject to a direction- turn-up in faint source counts and the excess radio sky tem- dependent effect (DDE; e.g. primary beam or ionospheric perature detected by the ARCADE2 experiment. effects) the traditional approach of solving for a single com- plex gain term per antenna is insufficient if high fidelity or high dynamic range imaging is required. Continuum imaging in the presence of sources which ACKNOWLEDGMENTS are blighted by strong DDEs results in maps whose dy- We thank the anonymous referee for very useful comments. namic range is limited by calibration artefacts rather than The National Radio Astronomy Observatory is a facility of thermal noise, as is the ideal outcome. Self-calibration of the National Science Foundation operated under coopera- the field will generate solutions dominated by the strongest tive agreement by Associated Universities, Inc. IH thanks source(s). Deriving a single complex gain correction is in- sufficient, and deficiencies in either the instrumental or sky model will manifest themselves as corrupted versions of the 6 “If it’s bright enough to cause trouble it’s bright enough to be point spread function centred on the fainter sources, which solved for.”– J. E. Noordam, ipse dixit. deconvolution is unable to remove. c 20XX RAS, MNRAS 000, 1–19 16 Heywood et al. Figure A1. Mean amplitudes of the differential gain solutions for three different Measurement Sets (rows A, B and C). Each column corresponds to an antenna in the array. The usual approach to mitigating these effects is to em- ately apparent. As severe as the presence of this confusing ploy some form of peeling algorithm (e.g. Noordam, 2004). source was in terms of obtaining a scientifically useful map, This is an iterative process whereby the sources are treated the DDEs that it was subject to can be tracked by the solver, in order of decreasing brightness. Self-calibration is per- and despite the factor of ∼100 by which 4C 00+02 was at- formed on a per-source basis on a visibility set which is tenuated by the primary beam response there is still ample phase-rotated to the position of the source in question. The signal to noise to form the solutions. best fitting model is computed and subtracted from the orig- Discontinuities in the solutions (e.g. 27-B) are indicative inal data. These residuals are then used as the starting point of amplitude spikes on antennas that were missed during for the second brightest problem source, and the process re- initial flagging. As the amplitude jumps suddenly the gain peats until an acceptable map is achieved. Aside from the correction drops to compensate. fact that this process is somewhat unwieldy and prone to Can the origin of the observed temporal gain variations user error, it may have trouble converging in the scenario be deduced? The most significant causes of DDEs in inter- where there are multiple confusing sources of similar bright- ferometry data are the ionosphere and the primary beam ness. response of the elements that make up the array. The iono- A more flexible and generic alternative now exists in the sphere manifests itself as a dynamic phase screen over the form of the differential gains algorithm (Smirnov, 2011b). array (Intema et al., 2009) and is a significant problem for This avoids the generation of intermediate data products observing with long baseline arrays at low frequencies, how- by forming simultaneous solutions for the complex receiver ever the X-band observations presented in this paper are gains against an all-inclusive sky model on short timescales, unlikely to be significantly affected by it. as well as solving for additional gain terms on longer timescales for a subset of dominating sources. The solution interval for the second component should be matched to the A2 The EVLA primary beam maximum interval over which one expects the DDE to be As mentioned in Section 2.1 it is speculated that the gain roughly constant. This maximises the signal to noise in the drifts are caused by the a combination of the structure in measurement, minimises the degrees of freedom in the fit, the primary beam pattern and its apparent rotation on the and effectively time-smears out contributions from the other sky as the observation progresses. Figure A2 shows a simu- sources in the field. For source subtraction purposes (as im- lated complex beam pattern for the EVLA at 8.4 GHz gen- plemented in this paper) the best fitting visibility models erated using the Cassbeam package (Walter Brisken, private for each of the sources for which differential gain solutions communication). Cassbeam takes a parametrized model of are computed are subtracted from the data and the residual the Cassegrain antenna optics (Brisken, 2003) and uses ray data are imaged. tracing to compute a 2×2 Jones matrix (Smirnov, 2011a and references therein) describing the effect of the antenna on the incoming radiation as a function of direction. This A1 Gain solutions complex-valued matrix is visualised in Figure A2. The complex gain solutions themselves encode much infor- The real components of the diagonal terms (LL and RR) mation (Smirnov, 2011c). For the calibration process to be show clearly the main lobe of the beam and the azimuthal successful there is an implicit (and justifiable) assumption structure beyond this induced by the antenna optics. The that (direction-dependent) gains vary smoothly with fre- off-diagonal terms (LR and RL) effectively show the instru- quency and time. As mentioned in Section 2.1 examining the mental polarization of the EVLA, the so-called ‘beam squint’ solutions can provide a valuable diagnostic by highlighting introduced by the fact that the two receptors sensitive to or- deviations from smooth behaviour. The mean amplitudes of thogonal polarization modes are not co-spatial. Polarization the differential gain solutions for 4C 00+02 for three differ- will not be discussed further. ent Measurement Sets (dubbed A, B and C) can be see in Such beam models can also be used by the A-projection Figure A1. Each column corresponds to a single EVLA an- algorithm (Bhatnagar et al., 2008) both to predict model tenna. The plots show the mean amplitude of each solution visibilities during calibration, and to apply a correction for as a function of time, averaged across the band. Rows A known beam effects during imaging. and B are derived from shorter 1.5 hour SBs and row C is It has already been noted that the confusing source in derived from a 3.5 hour observation. the observations of the WHDF was situated close to the The smoothness of these solutions in time is immedi- first sidelobe. Rotation of the beam on the sky will therefore c 20XX RAS, MNRAS 000, 1–19 The micro-Jy source population at 8.4 GHz in the WHDF 17 cause the gain to drift according to the azimuthal asymme- tries. This clearly does not tell the whole story as the gains in Figure A1 do not exhibit similar behaviour from antenna to antenna. The hypothesised explanation for this is point- ing error: radio telescopes have varying degrees of pointing accuracy which has the effect of shifting the beam patterns shown in Figure A2 away from the nominal pointing centre. The chance positioning of the confusing source in a region of the beam that has azimuthal structure and steep radial gradients causes the gain to exhibit strong behaviour in time and frequency. The shifting of the beam pattern on the sky due to pointing error causes these variations to differ from antenna to antenna. The next step is to attempt a simulation which replicates the observations. A3 Simulated observation The CASA sm tool was used to generate a Measurement Set with pointing direction and frequency range consistent with the real observations but with a six-hour track length, and antenna positions matching those of the the EVLA D- configuration with all 27 dishes. As per the averaged real data 8 × 32 MHz channels were used. The Measurement Set was then filled with simulated visibilities using the Siamese framework within MeqTrees. The key part of this framework is the BeamSims module which is able to read the simulated beams presented above as gridded FITS images. The module performs interpola- tion of the beam patterns as well as sky rotation. Crucially, pointing errors can also be applied. Each antenna is assigned a random pointing offset in two orthogonal directions on the sky, with a value between 0 and 10 arcseconds, a conserva- tive estimate of the true pointing accuracy of the EVLA (Vivek Dhawan, private communication). Two scenarios are now simulated. The first is with a source that is 6 arcminutes from the pointing centre, consis- tent with the location of 4C 00+02 in the WHDF observa- tions. The second is a simulation of a source at a radius of 9 arcminutes, which places it in the approximate centre of the first sidelobe. As the visibilities are computed the complex values of the applied beam gain for each of these sources, per integration time, per channel, per antenna are exported from the RIME tree. Two channels corresponding to 8.364 and 8.556 GHz are selected and plotted as a function of time Figure A2. Complex antenna beam patterns for all four polar- in Figure A3. The upper two rows show the amplitude val- izations products as generated by the Cassbeam software. The ues and the lower two rows show the phases. The thickest polarization product and whether the beam pattern corresponds line represents the lowest frequency channel. to the real or imaginary component is indicated above each frame. This simulation at least qualitatively reproduces the ob- These are rendered as FITS images and fed through a MeqTrees served behaviour. For the source with a separation of 6 ar- simulation module to investigate the effects of primary beam pat- cminutes there is large vertical scatter in the beam gains be- terns on various source configurations. For the diagonal terms in tween antennas. There are periods where gains rise as others the Jones matrix (LL and RR) the base contour level is ±0.005, fall and the variation is strongly chromatic. By contrast the with adjacent √ positive and negative contours increasing in mul- tiples of 2. The off-diagonal terms (RL and LR) have a base source at a radial separation of 9 arcminutes, although it contour level of ±0.0003 and the same increments as the diagonal exhibits temporal variation due to the rotation of the beam terms. The images span a sky area of approximately 54 arcmin- pattern, exhibits much less variation between antennas. The utes. source at 6 arcminutes is in a part of the beam which has strong gain gradients, and the pointing error exacerbates the disparities between antennas. c 20XX RAS, MNRAS 000, 1–19 18 Heywood et al. Figure A3. Simulated values of the complex beam gain (blue = amplitude, normalized to the maximal gain close to the beam centre; green = phase, in radians) as a function of time for each antenna in a six-hour EVLA simulation. Note that the phases in the r = 9’ case have been rotated by π radians to make the plot clearer. A random pointing error with a maximum value of 10” is applied to each antenna, equivalent to shifting the beam patterns shown in Figure A2 by a random amount in two orthogonal directions on the sky. For each complex component the plots show the behaviour for a source at a radius of 6 arcminutes from the phase centre (as per the EVLA observations of the WHDF) and at a radius of 9 arcminutes from the phase centre (corresponding to the approximate centre of the first sidelobe). The two lines in each panel correspond to two different channels in the simulation in order to show the strong chromatic effect in the r = 6’ case. The thick line is the 8.364 GHz and the thin line is 8.556 GHz, corresponding to the channels that span most of the bandwidth of the actual observations. The salient point that this plot makes is that a combination of pointing error, azimuthal structure in the beam and the chance positioning of a strong source in a part of the beam exhibiting strong radial gradients can result in beam gains with significantly different temporal and spectral behaviour from antenna to antenna. REFERENCES Hickox R. C., et al., 2012, MNRAS, 421, 284 Hill M. D., Shanks T., 2011, MNRAS, 410, 762 Abazajian, K. N., Adelman-McCarthy, J. K., Ag¨ ueros, Hill M. D., Shanks T., 2011, MNRAS, 414, 1875 M. A., et al. 2009, ApJ Suppl., 182, 543 Intema, H. T., van der Tol, S., Cotton, W. D., et al. 2009, Antonucci, R. 1993, ARA&A, 31, 473 A&A, 501, 1185 Bhatnagar, S., Cornwell, T. J., Golap, K., & Uson, J. M. Ivison R. J., Smail I., Le Borgne J.-F., Blain A. W., Kneib 2008, A&A, 487, 419 J.-P., Bezecourt J., Kerr T. H., Davies J. K., 1998, MN- Baugh C. M., Lacey C. G., Frenk C. S., Granato G. L., Silva RAS, 298, 583 L., Bressan A., Benson A. J., Cole S., 2005, MNRAS, 356, Ivison, R. J., Greve, T. R., Smail, I., et al. 2002, MNRAS, 1191 337, 1 Bielby, R. M., Hill, M. D., Metcalfe, N., & Shanks, T. 2012, MNRAS, 419, 1315 Jarvis, M. J., Rawlings S., 2004, NewAR, 48, 1173 B¨ohm, A., & Ziegler, B. L. 2007, ApJ, 668, 846 Jarvis, M. J., et al., 2010, MNRAS, 409, 92 Bonfield D. G., et al., 2011, MNRAS, 416, 13 Jarvis, M. J. 2011, arXiv:1107.5165 Brandt W. N., et al., 2001, AJ, 122, 2810 Kuraszkiewicz J. K., et al., 2003, ApJ, 590, 128 Brisken, W. 2003, Using GRASP8 to study Lutz D., et al., 2010, ApJ, 712, 1287 the VLA primary beam, EVLA Memo 58, Maddox N., Hewett P. C., P´eroux C., Nestor D. B., http://www.aoc.nrao.edu/evla/memolist.shtml Wisotzki L., 2012, MNRAS, 424, 2876 Cimatti A., et al., 2002, A&A, 391, L1 Mart´ınez-Sansigre, A., Rawlings, S., Lacy, M., et al. 2005, Cirasuolo, M., McLure, R. J., Dunlop, J. S., et al. 2010, Nature, 436, 666 MNRAS, 401, 1166 McCracken, H. J., Shanks, T., Metcalfe, N., Fong, R., & de Zotti, G., Massardi, M., Negrello, M., & Wall, J. 2010, Campos, A. 2000, MNRAS, 318, 913 A&A Rev., 18, 1 Metcalfe, N., Shanks, T., Fong, R., & Jones, L. R. 1991, Downes, A. J. B., Peacock, J. A., Savage, A., & Carrie, MNRAS, 249, 498 D. R. 1986, MNRAS, 218, 31 Metcalfe, N., Shanks, T., Fong, R., & Roche, N. 1995, MN- Dunlop, J. S., McLure, R. J., Kukula, M. J., et al. 2003, RAS, 273, 257 MNRAS, 340, 1095 Metcalfe, N., Shanks, T., Campos, A., McCracken, H. J., Fabian A. C., 1999, MNRAS, 308, L39 & Fong, R. 2001, MNRAS, 323, 795 Fixsen, D. J., Kogut, A., Levin, S., et al. 2011, ApJ, 734, 5 Metcalfe, N., Shanks, T., Weilbacher, P. M., et al. 2006, Fomalont, E. B., Kellermann, K. I., Partridge, R. B., Wind- MNRAS, 370, 1257 horst, R. A., & Richards, E. A. 2002, AJ, 123, 2402 Molinari S., et al., 2011, ApJ, 735, L33 Greisen, E. W. 2003, Information Handling in Astronomy Noordam, J. E. 2004, Proceedings of the SPIE, 5489, 817 - Historical Vistas, 285, 109 Noordam, J. E., & Smirnov, O. M. 2010, A&A, 524, A61 Gunn K. F. & Shanks T., 1999, arXiv:astro-ph/9909089 Padovani, P., Mainieri, V., Tozzi, P., et al. 2009, ApJ, 694, Gunn K. F., 1999, PhD Thesis, Durham University 235 Hatziminaoglou E., et al., 2010, A&A, 518, L33 Page, M. J., Stevens, J. A., Ivison, R. J., & Carrera, F. J. Henkel, B., & Partridge, R. B. 2005, ApJ, 635, 950 2004, ApJ Letters, 611, L85 Heywood, I., Armstrong, R. P., Booth, R., et al. 2011, Page M. J., et al., 2012, Natur, 485, 213 arXiv:1103.0862 Polletta M., Courvoisier T. J.-L., Hooper E. J., Wilkes c 20XX RAS, MNRAS 000, 1–19 The micro-Jy source population at 8.4 GHz in the WHDF 19 B. J., 2000, A&A, 362, 75 Sanders D. B., Soifer B. T., Elias J. H., Madore B. F., Matthews K., Neugebauer G., Scoville N. Z., 1988, ApJ, 325, 74 Seiffert, M., Fixsen, D. J., Kogut, A., et al. 2011, ApJ, 734, 6 Serjeant S., et al., 2010, A&A, 518, L7 Shi Y., et al., 2005, ApJ, 629, 88 Simpson, C., Mart´ınez-Sansigre, A., Rawlings, S., et al. 2006, MNRAS, 372, 741 Simpson C., et al., 2012, MNRAS, 421, 3060 Simpson J. M., et al., 2012, arXiv, arXiv:1206.4692 Smirnov, O. M. 2011a, A&A, 527, A106 Smirnov, O. M. 2011b, A&A, 527, A107 Smirnov, O. M. 2011c, A&A, 527, A108 Smolˇci´c V., et al., 2009, ApJ, 696, 24 Vallbe-Mumbru M., 2004, PhD thesis, Durham University Vernstrom, T., Scott, D., & Wall, J. V. 2011, MNRAS, 415, 3641 Wardlow J. L., et al., 2011, MNRAS, 415, 1479 Wilman, R. J., Miller, L., Jarvis, M. J., et al. 2008, MN- RAS, 388, 1335 Windhorst, R. A., Fomalont, E. B., Partridge, R. B., & Lowenthal, J. D. 1993, ApJ, 405, 498 This paper has been typeset from a TEX/ LATEX file prepared by the author. c 20XX RAS, MNRAS 000, 1–19