arXiv:astro-ph/0607141v2 28 Aug 2006
Astronomy & Astrophysics manuscript no. paper6cm.I.published
February 5, 2008
c ? ESO 2008
The ATESP 5 GHz radio survey
I. Source counts and spectral index properties of the faint radio population
I. Prandoni1, P. Parma1, M.H. Wieringa2, H.R. de Ruiter3,1, L. Gregorini4,1, A. Mignano5,1, G. Vettolani6,1, and R.D.
1INAF - Istituto di Radioastronomia, ViaGobetti 101, I–40129, Bologna, Italy
2CSIRO Australia Telescope National Facility, P.O. Box 76, Epping, NSW2121, Australia
3INAF - Osservatorio Astronomico di Bologna, Via Ranzani 1, I–40126, Bologna, Italy
4Dipartimento di Fisica, Universit` a di Bologna, Via Irnerio 46, I–40126, Bologna, Italy
5Dipartimento di Astronomia, Universit` a di Bologna, Via Ranzani 1, I–40126, Bologna, Italy
6INAF, Viale del Parco Mellini 84, I–00136, Roma, Italy
Received 29 September 2005 / Accepted 4 July 2006
Context. The nature and evolutionary properties of the faint radio population, responsible for the steepening observed in the 1.4 GHz source
counts below 1 milliJy, are not yet entirely clear. Radio spectral indices may help to constrain the origin of the radio emission in such faint
radio sources and may be fundamental in understanding eventual links to the optical light.
Aims. We study the spectral index behaviour of sources that were found in the 1.4 GHz ATESP survey (Prandoni et al. 2000a,b), considering
that the ATESP is one of the most extensive sub-mJy surveys existing at present.
Methods. Using the Australia Telescope Compact Array we observed at 5 GHz part of the region covered by the sub-mJy ATESP survey. In
particular we imaged a one square degree area for which deep optical imaging in UBVRIJK is available. In this paper we present the 5 GHz
survey and source catalogue, we derive the 5 GHz source counts and we discuss the 1.4−5 GHz spectral index properties of the ATESP
sources. The analysis of the optical properties of the sample will be the subject of a following paper.
Results. The 5 GHz survey has produced a catalogue of 111 radio sources, complete down to a (6σ) limit Slim(5 GHz) ∼ 0.4 mJy. We take
advantage of the better spatial resolution at 5 GHz (∼ 2′′compared to ∼ 8′′at 1.4 GHz) to infer radio source structures and sizes. The 5 GHz
source counts derived by the present sample are consistent with those reported in the literature, but improve significantly the statistics in the
flux range 0.4<∼S5 GHz<∼1 mJy. The ATESP sources show a flattening of the 1.4−5 GHz spectral index with decreasing flux density, which
is particularly significant for the 5 GHz selected sample. Such a flattening confirms previous results coming from smaller samples and is
consistent with a flattening of the 5 GHz source counts occurring at fluxes<∼0.5 mJy.
Key words. Surveys – Radio continuum: general – Methods: data analysis – Catalogs – Galaxies: general – Galaxies: evolution
One of the most debated issues about the sub-milliJy radio
sources, responsible for the steepening of the 1.4 GHz source
counts (Condon 1984, Windhorst et al. 1990), is the origin of
their radio emission. Understandingwhether the dominanttrig-
gering process is star formation or nuclear activity has impor-
tant implications on the study of the star formation/black hole
accretion history with radio-selected samples.
However, despite the extensive work done in the last
decade, the nature and the evolutionary properties of the faint
Send offprint requests to: I. Prandoni
Correspondence to: email@example.com
radio population are not yet entirely clear. Today we know that
the sub-mJy population is a mixture of different classes of ob-
jects (low-luminosity/high-zAGNs, star-forminggalaxies,nor-
mal elliptical and spiral galaxies), with star-forming galaxies
dominating the microJy (µJy) population (see e.g. Richards
et al. 1999), and early–type galaxies and AGNs being more
important at sub–mJy and mJy fluxes (Gruppioni et al. 1999;
Georgakakis et al. 1999; Magliocchetti et al. 2000; Prandoni et
al. 2001b). On the other hand, the relative fractions of the dif-
ferent types of objects are still quite uncertain, and very little is
known about the role played by the cosmological evolution of
the different classes of objects. Conclusions about the faint ra-
dio populationare, in fact, limited bythe incompletenessofop-
2I. Prandoni et al.: The ATESP 5 GHz radio survey. I.
tical identification and spectroscopy, since faint radio sources
have usually very faint optical counterparts. Clearly very deep
(R>25) optical follow-upfor reasonablylarge deep radiosam-
ples are critical if we want to probe such radio source popula-
Also importantmay be multi-frequencyradio observations:
radiospectral indices may help to constrainthe originof the ra-
dio emission in the faint radiosources and may actually be fun-
damental for understanding eventual links to the optical light.
This is especiallytrue if highresolutionradiodata are available
and source structures can be inferred.
Multi–frequency radio data are available only for a few
very small (<∼60 sources) sub–mJy samples. Such studies in-
dicate that most mJy radio sources are of the steep–spectrum
type (α < −0.5, assuming S ∼ να), with evidence for flatten-
ing of the spectra at lower flux densities (Donnelly et al. 1987;
Gruppioni et al. 1997; Ciliegi et al. 2003). This flattening is
consistent with the presence of many flat (α > −0.5) and/or
inverted (α > 0) spectral index sources at µJy flux densities
(Fomalont et al. 1991; Windhorst et al 1993). On the other
hand,thereis still disagreementaboutthe interpretationofsuch
In the µJy population studied by Windhorst et al. (1993)
50% of the sources have intrinsic angular size Θ ≥ 2.6±1.4
arcsec, corresponding to ≃ 5−40 kpc at the expected median
redshift of the sources. Extended (kpc–scale) steep–spectrum
radio sources suggest synchrotron emission in galactic disks,
while extended flat–spectrum sources may indicate thermal
bremsstrahlung from large scale star–formation both occasion-
ally with opaqueradiocores. On the other hand,Donnellyet al.
(1987)claim that most of the sub–mJyblue radio galaxies have
steep radio spectra and are physically quite compact (≤ 4 kpc).
emission: 1) a nuclear starburst occurringon a few kpc scale in
the galaxy center; 2) a non-thermal nucleus on parsec scales.
Only high resolution radio observations could decide between
faintradiopopulation,we imagedat 5GHz a onesquaredegree
area of the ATESP 1.4 GHz survey(Slim∼0.5 mJy,Prandoniet
al. 2000a,b),for which deep (R< 25.5) optical multi-colordata
is available, as part of an ESO public survey (e.g. Mignano
et al. 2006). Such deep optical imaging will provide optical
identification and photometric redshifts for most of the radio
We noticethat the ATESP is best suited to studythe sources
populating the flux interval 0.5−1 mJy, where starburst galax-
ies start to enter the counts, but are not yet the dominant pop-
ulation. This means that our sample can be especially useful
to study the issue of low-luminosity nuclear activity, possi-
bly related to low efficiency accretion processes and/or radio-
The present 5 GHz observations are valuable because i) the
higher resolution images probe the radio source structure at
small scales (≤2 arcsec) and thus we can hopefullydistinguish
between disk–scale and nuclear–scale radio emission, and ii)
the present 5 GHz survey is the largest at sub–mJy fluxes, by a
factor 10−100: previous samples typically cover from < 0.01
1984, 1991; Donnelly et al. 1987; Partridge et al. 1986; Ciliegi
et al. 2003).
In this paper we describe the ATESP 5 GHz survey,present
the5 GHzsourcecatalogueandcounts,anddiscuss thespectral
index properties of the ATESP radio sources. The analysis of
the optical properties of the sample will be the subject of a
following paper (Mignano et al. in preparation).
The paper is organized as follows. In Sect. 2 we briefly
present the 1.4 GHz ATESP survey and the multi-color op-
tical data coming from the ESO Deep Public Survey (DPS).
Sect. 3 describes the ATESP 5 GHz survey observations and
data reduction. The radio mosaics we produced are discussed
in Sect. 4, while Sect. 5 describes the source extraction and pa-
rameterization procedure. The 5 GHz source catalogue is pre-
structure. The 5 GHz source counts derived from the present
survey and the 1.4−5 GHz spectral index properties of the
ATESP sources are presented in Sect. 7. A summaryis givenin
2. The 1.4 GHz ATESP Survey and Related Optical
The ATESP 1.4 GHz survey (Prandoni et al. 2000a) was car-
ried out with the Australia Telescope Compact Array (ATCA);
it consists of 16 mosaics with 8′′×14′′resolution and uni-
formsensitivity (1σ noiselevel ∼79µJy), coveringtwo narrow
strips of 21◦×1◦and 5◦×1◦near the SGP, at decl. −40◦. The
ATESP 1.4GHz surveyhas produceda catalogueof 2967radio
sources, down to a flux limit (6σ) of S ∼ 0.5 mJy (Prandoni et
In order to alleviate the identification work, the area cov-
ered by the ATESP survey was chosen to overlap with the re-
gion where Vettolani et al. (1997) made the ESP (ESO Slice
Project) redshift survey. They performed a photometric and
spectroscopic study of all galaxies down to bJ∼ 19.4. The ESP
survey yielded 3342 redshifts (Vettolani et al. 1998), to a typi-
cal depth of z = 0.1 and a completeness level of 90%.
In the same region lies the ESO Imaging Survey (EIS)
Patch A (∼ 3◦×1◦square degrees, centered at 22h40m, −40◦),
mainlyconsistingofimagesin the I-bandout ofwhicha galaxy
catalogue95%completeto I =22.5has beenextracted(Nonino
et al. 1999). This catalogue allowed us to identify ∼57% of the
386 ATESP sources present in that region. A first radio/optical
analysis of a magnitude-limited sub-sample of 70 sources was
presented by Prandoni et al. (2001b).
More recently, a different strip of 2◦× 0.5◦within the
ATESP region at 22h50m40s, −40◦13′was selected for a very
deep multi-color ESO public survey: the Deep Public Survey
(DPS), which was carried out with the Wide Field Imager
(WFI) at the 2.2 m ESO telescope. This one square degree re-
gion is coveredby 4 WFI fields (referredto as DEEP1a,b,c and
d). In this region UBVRI imaging down to very faint magni-
tudes is available (see Mignano et al. 2006): U ∼ 25, B ∼ 25.8,
V ∼ 25.2, R ∼ 25.5, I ∼ 24 (5σ, 2 arcsec aperture magnitudes).
In addition, DEEP1a and b have been observed in the infrared
with SOFI at NTT down to KAB∼ 21.3, while deeper J- and
I. Prandoni et al.: The ATESP 5 GHz radio survey. I.3
Fig.1. Sketch of the sky coverage of the radio (ATESP at 1.4 and 5 GHz) and optical (ESP, EIS, DPS) surveys relevant for this
study. See text for details.
K-band images (JAB< 23.4 and KAB< 22.7) have been taken
for selected sub-regions (see Olsen et al., 2006)
A sketch of the sky coverage of the surveys described in
this section is given in Fig. 1.
3. The Observations
3.1. Observing Strategy
We imaged the entire DEEP1 2×0.5 degree region at 5 GHz,
since it has the best optical coverage. The area was spanned
with a radio mosaic consisting of 21×6 pointings (fields) at
6′spacing, i.e. FWHM/√2, where FWHM= 10′is the full
width at half maximum of the primary beam (see Prandoni
et al. 2000a). If we aim at virtually detecting all the 1.4 GHz
ATESP sources with radio spectral index α ≥ −0.7, we need to
reach a 5 GHz point source detection limit of 3σ ≃ 0.2 mJy.
Of course, correct spectral index determination can only be
made if the 5 and 1.4 GHz beams have the same size, as severe
incompleteness would result if the extra resolution at 5 GHz
were used. Therefore we produced 5 GHz radio mosaics with
the same spatial resolution as for the ATESP 1.4 GHz mosaics.
We thus used the Compact Array in the 1.5 km configuration.
On the other hand we also requested the 6 km antenna. While
we want to extract the source cataloguefromthe low resolution
mosaic, the longerbaselines to the 6km antennawere exploited
to get additional information on the radio source structure (see
Each field (with 2×128 MHz bandwidth and 10 baselines)
was observed for 98 minutes, which results in a final uniform
noise level of ∼ 64 µJy. Therefore all 21×6 = 126 fields could
be observedin 18 blocks of 12 hours (allowingalso for calibra-
tion time). Care was taken to obtain good hour angle coverage,
by cycling continually through the individual field during the
observing process. Since we wanted to observe the entire re-
gion also with the 6 km configuration, an additional 12 hours
were used for that purpose.
The observing log is given in Table 1. Note that the use of
different1.5 km arrays is not relevantfor the present study. The
two 128 MHz bands were set at 4800 and 5056 MHz.
The flux density calibration was performed through ob-
servations of the source PKS B1934-638, which is the stan-
Table 1. Log of the observations.
dard primary calibrator for ATCA observations (S = 5.8 Jy at
ν = 4800 MHz as revised by Reynolds 1994, Baars et al. 1977
flux scale). The phase and gain calibration was based on obser-
vations of a secondary calibrator (source 2254-367) selected
from the ATCA calibrator list.
3.2. Data Reduction
For the data reduction we used the Australia Telescope
National Facility (ATNF) release of the Multichannel Image
Reconstruction, Image Analysis and Display (MIRIAD) soft-
ware package (Sault et al. 1995).
Every single 12hrun and each of the two observing bands
were flagged and calibrated following standard procedures for
ATCA observations, as described in Prandoni et al. (2000a).
Sensitivity and u−v coverage were improved for each field
by merging, before imaging and cleaning, the visibilities com-
ing from all the observing runs and from the two observing
bands. Imaging and deconvolution was done simultaneously
for several pointings. This is not only simpler and faster, but
also produces better results, as overlappingpointings can make
use of a higher number of visibilities and side lobes of sources
in contiguous fields can be easlily cleaned.
We producedmosaics at bothlow andfullresolution(∼10′′
entire 2◦×0.5◦region, while in full resolution 30 overlapping
mosaics of 9 pointings each were produced.Final images were
obtained by cleaning after (phase only) self calibration.
4I. Prandoni et al.: The ATESP 5 GHz radio survey. I.
Table 2. Main parameters for the 2 low resolution mosaics and average values from the 30 full resolution mosaics.
fld x to y
< σ >
22 47 39.57
22 52 38.16
-40 13 00.0
-40 13 00.0
full res. mosaicsf
ax and y refer to the first and last field columns composing the mosaic.
bJ2000 reference frame.
cP.A. is defined from North through East.
dLow res. mosaic overlapping the 1.4 GHz ATESP mosaic fld05to11 (see Prandoni et al. 2000a).
eLow res. mosaic overlapping the 1.4 GHz ATESP mosaic fld10to15 (see Prandoni et al. 2000a).
fAverage values from the 30 full resolution mosaics.
Snapshot surveys like the present one are typically affected
by the clean bias effect: the deconvolutionprocess can produce
a systematic underestimation of the source fluxes, as conse-
quence of the loose constraints to the cleaning algorithm due
to sparse u −v coverage (see White et al. 1997; Condon et
al. 1998). The clean bias effect has been discussed in great
detail (Prandoni et al. 2000a), and we repeat here only that
such a systematic effect can be kept under control if cleaning is
stopped well beforethe maximumresidual flux has reachedthe
theoretical noise level. Specifically, we set the cleaning limit
at 4 times the theoretical noise, since simulations made by us
show that this cut-off ensures that the clean bias does not affect
Another systematic effect that has to be taken into account
is bandwidth smearing. It is well known that at large distance
from the pointing centre bandwidth smearing tends to reduce
the peak flux and increase the apparent source size in the radial
direction, such that total flux remains conserved. Also this ef-
fect has been discussed by us extensively in an earlier paper on
the ATESP 1.4 GHz survey (Prandoni et al. 2000b), in partic-
ular in the context of radio mosaics. Considering that the pass-
band width is 4 MHz, for the multichannel 32×4 MHz con-
tinuum mode observations, and the observing frequency about
5000 MHz, it is easily seen from equation (8) in Prandoni et
al. (2000b)that the ratio between smeared and unsmearedpeak
flux is between 0.9999 and 1. Consequently bandwidth smear-
ing is of no concern for our 5 GHz survey.
4. The Radio Mosaics
4.1. Production of the Mosaics
As mentioned before, we needed to produce mosaics at ex-
actly the same resolution as the 1.4 GHz images (see Prandoni
et al. 2000a), in order to be able to compare the data at two
frequencies and determine reliable spectral indices. We there-
fore used only the 10 baselines shorter than 3 km. Radio maps
with 520×520 pixels of 2.5 arcsec were made, and combined
into two mosaics. In this way all the flux in the primary beam
(which is 20.6 arcmin at 4800 MHz) is recovered. Details on
the two low resolution mosaics (which are of the order of
2000× 1200 pixels. and have a small overlap) are given in
Table 2: we list the number of fields composing the mosaics
(columns × rows), the tangent point (sky position used for ge-
ometry calculations) and the restoring synthesized beam (size
in arcsec and position angle).
Since the aim of the low resolution imaging is basically
sensitivity, natural weighting was used in the deconvolution
process. However, this choice may introduce some spatially
correlated features and this may affect the zero level of faint
radio sources. This problem can be avoided by removing all
baselines shorter than 60 m from the data prior to imaging and
Althoughthis meansthat ∼3−7% ofthe visibilities in 1.5B
and 1.5F configurations had to be rejected, this had hardly any
adverse effect on the quality of the mosaics. The lack of the
shortest spacings would in principle lead to an increased insen-
sitivity to sources larger than 90′′, but in reality less than one
source with angular size > 90′′is expected in the area and flux
range covered by the present survey (as discussed in Sect. 6.2).
Therefore the effect on completeness and flux densities should
In order to assess the radio source structures full resolution
imageswere produced.The2×0.5sq.degr.regionwas covered
by a grid of 30 overlapping mosaics, each composed by 3×3
or 3×2 fields. A size of 2060×2060 pixels (with a pixel size
of 0.6′′) for each field in the mosaics ensured complete recov-
ery of the whole flux in the field of the primary beam. Some
average parameters of the full resolution mosaics are given in
4.2. Noise Analysis of the Mosaics
The 5 GHz survey was designed to give uniform noise in the
central 1.0◦×0.5◦regions of the two low resolution mosaics,
which together coverthe area of the DEEP1 optical survey(see
Mignano et al. 2006). In the following our noise analysis al-
ways refers to this region. In column 7 of Table 2 we list the
minimum (negative) flux Sminfound in the image, in column 8
I. Prandoni et al.: The ATESP 5 GHz radio survey. I.5
Fig.2. Top: Grey scale of the noise maps obtained from the two low resolution mosaics: fld1to11 (Top) and fld10to21 (Bottom).
Darker regions correspond to higher noise. The rectangular boxes indicate the region corresponding to the DEEP1 area, where
the survey was designed to provide a uniform noise level. The contour images represent the theoretical sensitivity computed by
taking into account the observing time actually spent on the mosaiced single fields: contours refer to 1.1, 1.0, 0.9, 0.8 × the mean
noise value estimated in the noise maps (see Fig. 3). As expected the actual noise variations trace reasonably well the theoretical
the noise σfitestimated as the FWHM of the flux distribution
in the pixels in the range ±Smin, and in column 9 the noise
< σ > estimated as the standard deviation of the average flux
in several source-free sub-regionsof the mosaics. σfitwas used
to check the presence of correlated noise, while < σ > gives an
idea of the uniformity of the noise over the area of the mosaic.
As expected, noise variations in general do not exceed ∼ 10%,
with the sole exception of one of the full resolution mosaics,
due to the presence of an S = 27 mJy source that could not ef-
fectively be self-calibrated. On average we find a noise level
around ∼ 70 µJy; therefore a 3σ detection limit for a 5 GHz
source at the position of a 1.4 GHz source is ∼ 0.21 mJy. In all
mosaics the noise is essentially Gaussian.