A 610-MHz survey of the ELAIS-N1 field with the Giant Metrewave Radio Telescope - Observations, data analysis and source catalogue

Abstract

Observations of the ELAIS-N1 field taken at 610 MHz with the Giant Metrewave Radio Telescope are presented. Nineteen pointings were observed, covering a total area of 9 square degrees with a resolution of 6" x 5", PA +45 deg. Four of the pointings were deep observations with an rms of 40 microJy before primary beam correction, with the remaining fifteen pointings having an rms of 70 microJy. The techniques used for data reduction and production of a mosaicked image of the region are described, and the final mosaic is presented, along with a catalogue of 2500 sources detected above 6 sigma. This work complements the large amount of optical and infrared data already available on the region. We calculate 610-MHz source counts down to 270 microJy, and find further evidence for the turnover in differential number counts below 1 mJy, previously seen at both 610 MHz and 1.4 GHz. Comment: 12 pages, 18 figures, two tables. Table 1 can be found in full via http://www.mrao.cam.ac.uk/surveys/ . Accepted for publication in MNRAS

Full text

Page 1

arXiv:0710.1500v1 [astro-ph] 8 Oct 2007
Mon. Not. R. Astron. Soc. 000, 000–000 (2007) Printed 2 February 2008(MN LATEX style file v2.2)
A 610-MHz survey of the ELAIS-N1 field with the Giant Metrewave
Radio Telescope – Observations, data analysis and source catalogue
Timothy Garn⋆, David A. Green, Julia M. Riley, Paul Alexander
Astrophysics Group, Cavendish Laboratory, 19 J. J. Thomson Ave., Cambridge CB3 0HE
2 February 2008
ABSTRACT
Observations of the ELAIS-N1 field taken at 610 MHz with the Giant Metrewave Radio Tele-
scope are presented. Nineteen pointings were observed,coveringa total area of ∼ 9 deg2with
a resolution of 6 × 5 arcsec2, PA +45◦. Four of the pointings were deep observations with an
rms of ∼ 40 µJy before primary beam correction, with the remaining fifteen pointings having
an rms of ∼ 70 µJy. The techniques used for data reduction and production of a mosaicked
image of the region are described, and the final mosaic is presented, along with a catalogue
of 2500 sources detected above 6σ. This work complements the large amount of optical and
infrared data already available on the region. We calculate 610-MHz source counts down to
270 µJy, and find furtherevidencefor the turnoverin differentialnumbercounts below 1 mJy,
previously seen at both 610 MHz and 1.4 GHz.
Key words: surveys – catalogues – radio continuum: galaxies
1 INTRODUCTION
The
Lonsdale et al. 2003) survey has the largest sky coverage of
the legacy surveys being performed by the Spitzer Space Telescope
(Werner et al. 2004). A total area of ∼49 deg2of sky has been
observed withthe Infrared Array Camera (IRAC; Fazio et al. 2004)
and Multiband Imaging Photometer for Spitzer (MIPS; Reike et al.
2004) instruments at 3.6, 4.5, 5.8, 8, 24, 70 and 160 µm. The
survey is broken down into six fields, three in the northern sky –
ELAIS-N1, ELAIS-N2 and the Lockman Hole – and three in the
south – ELAIS-S1,Chandra Deep Field South and the XMM-Large
Scale Structure (XMM-LSS) field. All six regions were selected to
be away from the Galactic disk, in order to minimize background
cirrus emission.
There is a large amount of multi-wavelength information
available on all six SWIRE fields. The three ELAIS fields were ob-
served aspart of theEuropean Large-AreaISO Survey (Oliver et al.
2000), which also included another northern (-N3) and southern (-
S2) field. The Infrared Space Observatory (ISO) observed these
regions at 6.7, 15, 90 and 175 µm, and a large number of fol-
lowup observations were carried out in the optical, infrared and
radio bands. A band-merged catalogue, containing the ISO data,
along with U, g′, r′, i′, Z, J, H and K-band detections, and radio
observations at 1.4 GHz has been produced – for more details, see
Rowan-Robinson et al. (2004), and references therein.
Observations of the ELAIS-N1 region were taken with
Spitzer in 2004 January, covering ∼ 9 deg2with the IRAC and
MIPS instruments. The source catalogues have been produced,
Spitzer
Wide-area InfraredExtragalactic(SWIRE;
⋆E-mail: tsg25@cam.ac.uk
and are available online (Surace et al. 2004), containing over
280,000 sources. The UK Infrared Deep Sky Survey (UKIDSS;
Lawrence et al. 2007) intends to cover the ELAIS-N1 region in
its Deep Extragalactic Survey plan, observing the full field in
the J, H and K-bands to a depth of K = 21 mag. This will be
a great improvement over the currently available surveys, which
have a sensitivity limit of ∼18 mag in the K band. Data Release 2
(Warren et al. 2007) of UKIDSS contains early shallow data on the
ELAIS-N1 region. Further surveys have been carried out in the R-
band (Fadda et al. 2004), in Hα (Pascual et al. 2001), and with the
Chandra X-ray telescope (Manners et al. 2003; Franceshini et al.
2005). There have been several redshift surveys of the region
(Trichas et al. 2006; Berta et al. 2007), and the ELAIS-N1 region
was also partially covered by the Sloan Digital Sky Survey (SDSS;
Adelman-McCarty et al. 2007).
While there have been a great number of observations of
the ELAIS-N1 region at optical and infrared wavelengths, there is
comparatively little radio information available. The existing VLA
1.4 GHz survey of the three northern ELAIS fields (Ciliegi et al.
1999), which has been included into the band-merged catalogue
of Rowan-Robinson et al. (2004), reaches a 5σ limit of 0.135 mJy
over 0.12 deg2but only a 1.15 mJy limit over its full coverage
area of 4.22 deg2. The NVSS (Condon et al. 1998) and FIRST
(Becker et al. 1995) surveys both cover the ELAIS-N1 region, but
only to relatively shallow 5σ limits of 2.25 and 0.75 mJy respec-
tively. A recent study of polarised compact sources (Taylor et al.
2007) at 1420 MHz is underway, using the Dominion Radio As-
trophysical Observatory Synthesis Telescope (DRAO ST) cen-
tered on 16h11m00s, +55◦00′00′′and covering 7.4 deg2. The first
30 per cent of observations have been completed, with maps in

Page 2

2
T. Garn et al.
Figure 1. The 19 pointings observed in the ELAIS-N1 region. Fields B4,
C3, C5 and D4 are the deep observations.
Stokes I, Q and U being produced with a maximum sensitivity of
78 µJy beam−1, although with a resolution of ∼1 arcmin2.
Inorder toextend theinformation onthisregion, amuch larger
deep radio survey isrequired. In thispaper, we present observations
of theELAIS-N1survey field taken at 610 MHz with theGiant Me-
trewave Radio Telescope (GMRT; Ananthakrishnan 2005), cover-
ing ∼ 9 deg2of sky with a resolution of 6×5 arcsec2, PA +45◦,
centred on 16h11m00s, +55◦00′00′′(J2000 coordinates, which are
used throughout this paper). This survey, in combination with the
deep Spitzer data, will be used to study the infrared/radio correla-
tion for star-forming systems (e.g. Appleton et al. 2004), and the
link between the triggering of star formation and AGN activity, as
well as the properties of the faint radio population at 610 MHz.
In Section 2 we describe the observations and data reduction
techniques used in the creation of the survey. Section 3 presents
the mosaic and a source catalogue containing 2500 sources above
6σ, along with a sample of extended sources. In Section 4 we con-
struct the 610 MHz differential source counts, and compare them
to previous works.
2 OBSERVATIONS AND DATA REDUCTION
TheELAIS-N1region wasobserved for 25 hours, spread over three
daysin2003 August, usingtheGMRToperating at610 MHz.Nine-
teen pointings were observed, centred on 16h11m00s, +55◦00′00′′
and spaced by 36′in a hexagonal grid (as shown in Fig. 1) in order
to get nearly uniform coverage over the region. Two 16 MHz side-
bands were observed, each split into 128 spectral channels, with a
16.9 s integration time.
The flux density scale was set through observations of 3C48
or 3C286, at the beginning and end of each observing session. The
AIPS task SETJY was used to calculate 610 MHz flux densities of
29.4 and 21.1 Jy, respectively, using the AIPS implementation of
the Baars et al. (1977) scale. Each field was observed for a series of
interleaved 9 min scans in order to maximise the uv coverage, and a
nearby phase calibrator,J1634+627, was observed for four minutes
between every three scans to monitor any time-dependent phase
Figure 2. The uv coverage for pointing A3. Baselines less than 1 kλ were
not used in imaging and have been omitted from the figure.
and amplitude fluctuations of the telescope. The measured phase
typically varied by less than 10 degrees between phase calibrator
observations. Most of the fields were observed for four scans (36
mins in total), while four fields, B4, C3, C5 and D4, were observed
for significantly longer, in order to give a deeper region within the
main survey as there is deep optical information available on this
region from UKIDSS (Warren et al. 2007). B4 and D4 were ob-
served for 11 scans (99 mins) while C3 and C5 were observed for
12 scans (108 mins).
The imaging strategy was very similar to that used for our
610 MHz GMRT observations of the Spitzer extragalactic First
Look Survey field (xFLS; Garn et al. 2007). An error in the times-
tamps of the uv data was corrected using UVFXT (for further de-
tails see Garn et al. 2007). This error is only found for observations
made before the summer of 2006, and has since been fixed at the
GMRT. Initial editing of the data was performed separately on each
sideband with standard AIPS tasks, to remove bad baselines, anten-
nas, and channels that were suffering from large amounts of narrow
band interference, along with the first and last integration periods
of each scan. The flux calibrators were used to create a bandpass
correction for each antenna. In order to create a continuum chan-
nel,fivecentral frequency channels werecombined together, and an
antenna-based phase and amplitude calibration created using obser-
vations of J1634+627. This calibration was applied to the original
data, which was then compressed into 11 channels, each with band-
width of 1.25 MHz (so the first few and last few spectral channels
were omitted from the data, since they tended to be the noisiest).
The new channel width is small enough that bandwidth smearing
is not a problem in our images, and led to an effective bandwidth
of 13.75 MHz in each sideband. Further flagging was performed
on the 11 channel data set, and the two sidebands combined us-
ing UVFLP (again, see Garn et al. 2007) to improve the uv cover-
age. The coverage for one of the shallow observations is shown in
Fig. 2. Baselines shorter than 1 kλ were omitted from the imag-
ing, since the GMRT has a large number of small baselines which
would otherwise dominate the beam shape.
Each pointing was broken down into 31 smaller facets (as dis-

Page 3

A 610-MHz survey of the ELAIS-N1 field
3
cussed in Garn et al. 2007), arranged in a hexagonal grid and cov-
ering an area with diameter ∼ 1.◦8. These were imaged separately,
each with a different assumed phase centre. The large area cov-
ered (compared with the Full-Width Half-Maximum of the GMRT,
which is∼0.◦74) allowsbright sources well outside of the observed
region to be cleaned from the images, while the faceting procedure
avoids the introduction of phase errors due to the non-planar nature
of the sky. All images were made with the same elliptical synthe-
sised beam of size 6×5 arcsec2, position angle +45◦by setting the
parameters for the restoring beam within IMAGR. A pixel size of
1.5 arcsec was chosen, to ensure that the beam was well oversam-
pled.
Each pointing went through three iterations of phase self-
calibration at 10, 3 and 1 minute intervals, and then a final round
of self-calibration correcting both phase and amplitude errors, at
10 minute intervals. The overall amplitude gain was held constant
in order not to alter the flux density of sources. The self-calibration
steps improved the noise level by about 10 per cent, and signifi-
cantly reduced the residual sidelobes around the brighter sources.
The four deep pointings have a final rms of between 40 and
42 µJy before correction for the GMRT primary beam, while the
15 shallow fields have a noise level of between 66 and 73 µJy.
These figures are very close to the expected thermal noise limits of
36 and 63 µJy respectively, calculated using
√2Ts
G?n(n−1)NIF∆ντ
where the system temperature Ts≈ 92 K, and the antenna gain G≈
0.32 K Jy−1– values taken from the GMRT website1– n is the
number of working antennas, typically 27 during our observations,
NIF= 2 is the number of sidebands, ∆ν = 13.75 MHz is the effec-
tive bandwidth per sideband, and τ is the integration time for each
pointing.
In Garn et al. (2007), we detected a position-dependent error
with the GMRT primary beam. This led to a systematic difference
between the measured flux densities of sources that were present
in more than one pointing, with the fractional offset varying across
the primary beam. To check for a similar effect here, we corrected
each pointing for the nominal beam shape, using an 8th-order poly-
nomial with coefficients taken from Kantharia & Rao (2001). We
took the 557 sources present in the overlapping regions between
two pointings with peak brightness greater than 1.5 mJy beam−1,
and looked at the fractional offsets as a function of position away
from each pointing centre. Fig. 3 shows the variation in this offset
with source Right Ascension. A similar trend was seen with Decli-
nation.
We repeated the analysis of Garn et al. (2007), to model the
effective pointing centre of the telescope as having a systematic
offset from its nominal value. By shifting the phase centre of all
pointings by ∼ 2.′7 in a north-west direction before performing the
primary beam correction, we were able to remove this systematic
effect, as shown in Fig 4. The amount and direction of the correc-
tion was consistent between pointings, and did not vary between
the deep and shallow observations. This correction also removed
the systematic effect seen in the Declination direction. The size and
direction of this correction is similar to that seen in the xFLS sur-
vey.
The accuracy of the primary beam correction was then tested,
as shown in Fig. 5. The fractional offset of sources in overlapping
σ =
(1)
1http://www.gmrt.ncra.tifr.res.in/gmrt hpage/Users/Help/help.html
−0.4
−0.2
0
0.2
0.4
(I1-I2)/(I1+I2)
(I1-I2)/(I1+I2)
−40−200 2040
RA offset (arcmin)RA offset (arcmin)
Figure 3. Fractional offset in peak brightness of all sources in overlap-
ping regions, using the nominal primary beam centre. A gradient across
the pointing is seen, leading to systematic errors in measured flux density.
−0.4
−0.2
0
0.2
0.4
(I1-I2)/(I1+I2)
(I1-I2)/(I1+I2)
−40−200 20 40
RA offset (arcmin)RA offset (arcmin)
Figure 4. Fractional offset in peak brightness of sources in overlapping
regions, using the shifted primary beam centre applied to all fields.
regions is now a good fit to a Gaussian, with mean of 0.009 and
σ of 0.1, indicating a 10 per cent error in the absolute flux density
calibration of all sources.
The 19 pointings were mosaicked together, taking into ac-
count the offset primary beam and weighting the final mosaic ap-
propriately by the relative noise of each pointing. The mosaic was
cut off at the point where the primary beam correction dropped to
20 per cent of its central value. The final mosaicked image is avail-
able via http://www.mrao.cam.ac.uk/surveys/.
0
20
40
60
N N
−0.200.2
(I1-I2)/(I1+I2) (I1-I2)/(I1+I2)
Figure 5. Distribution of fractional offsets, after correction for the shifted
primary beam. The best-fit Gaussian is shown, giving a 10 per cent error in
the overall flux density calibration.

Page 4

4
T. Garn et al.
Figure 6. The rms noise of the final mosaic, calculated using Source Ex-
tractor. The grey-scale ranges between 40 and 350 µJy, and the contours
are at 60 and 120 µJy respectively.
3 RESULTS AND DISCUSSION
Source Extractor (SExtractor; Bertin & Arnouts 1996) was used to
calculatethermsnoiseσ acrossthemosaic. Agridof 16×16 pixels
was used in order to track changes in the local noise level, which
varies significantly near the brightest sources. Fig. 6 illustrates the
local noise, with the grey-scale varying between 40 µJy (the noise
level in the centre of the deep pointings) and 350 µJy (the noise
level for the shallow pointings, at the distance where the GMRT
primary beam gain was 20 per cent of its central value). The 60 and
120 µm contours are plotted, which cover the majority of the deep
and shallow survey regions respectively.
A sample region of the ELAIS-N1survey is shown in Fig. 7 to
illustratethe quality of the image. Most of the sources in our survey
are unresolved, although there are several with extended structures.
We present a sample of these in Fig. 8.
Our GMRT data suffer from dynamic range problems near
the brightest sources, and the final mosaic has increased noise and
residual sidelobes in these regions. We had fewer problems with
our survey of the xFLS region, due to the longer time spent on
each pointing and the correspondingly better uv coverage. There
were also fewer bright sources in the xFLS field, so a much smaller
region was affected by residual sidelobes. Fig. 9 shows an area
around one of the bright sources, to illustrate the problems caused
by the residual sidelobes. While the local noise calculated by SEx-
tractor increases due to these residuals, some of them still have an
apparent signal-to-noise level that is greater than 6. We therefore
opted for a two-stage selection criteria for our final catalogue.
3.1 Source fitting
An initial catalogue of 4767 sources was created using SExtrac-
tor. The mosaic was cut off at the point where the primary beam
correction dropped to 20 per cent of its central value (a radius of
0.◦53 from the outer pointings), however only sources inside the
30 per cent region (0.◦47) were included in the catalogue to avoid
Figure 9. Source GMRTEN1 J161212.4+552308, and the errors surround-
ing it. The grey-scale ranges between −0.2 and 1 mJy beam−1, and the
source has a peak brightness of 389 mJy. The region affected by an over-
density of sources is shown by the black circle, of radius 10 arcmin – see
text for more details.
the mosaic edges from affecting the estimation of local noise. The
requirements for a source to be included were that it had at least
5 connected pixels with brightness greater than 3σ, and a peak
brightness greater than 6σ. The image pixel size meant that the
beam was reasonably oversampled, so the source peak was taken
to be the value of the brightest pixel within a source. The inte-
grated flux density was calculated using the FLUX AUTO option
within SExtractor. This creates an elliptical aperture around each
object (as described in Kron 1980), and integrates the flux con-
tained within the ellipse. Comparisons between the flux density
obtained through this method and through the method developed in
Garn et al. (2007) – pixels above a given threshold were summed,
then empirically corrected for the elliptical beam shape – give good
agreement between the two techniques. Sources with Kron flux
density above 1 mJy showed no statistical difference between the
two flux density measurements, with an uncertainty of 3 per cent.
Sources below 1 mJy had a Kron flux density that was systemati-
cally larger than the Garn et al. (2007) method by 6 per cent, with
an uncertainty of 4 per cent. We chose to use the Kron method in
order to avoid the empirical correction factor.
In order to estimate the area affected by artefacts near the
bright sources, we calculated the number of (potentially spurious)
sources in a series of concentric rings centred on a bright source,
and converted this to an effective source density as a function of
distance. While the noise does vary across the map, the variation is
smooth and relatively slow away from the bright sources. Using the
more distant rings we calculated a mean source sky density µ, and
an estimate of the error in this value, σµ, which will be unaffected
by the presence of the spurious sources, and will vary slightly be-
tween each bright central source due to the changing properties of
the image.
Fig. 10 shows thesource density (inarbitrary units) away from
the source in Fig. 9. The clear over-density of sources can be seen
within 10 arcmin of the source. If there is a significant peak in the
density (greater than µ +6σµ), then we define the size of the af-
fected region by finding the first radius at which the source density
drops below µ +3σµ– for this source, the affected radius is 10 ar-
cmin. The affected region has been added to Fig. 9. Within this

Page 5

A 610-MHz survey of the ELAIS-N1 field
5
Figure 7. A 70×70 arcmin2region of the 610 MHz image, to illustrate the quality of the survey. The region is located within the deeper area of the survey,
and most sources are unresolved. The grey-scale ranges between −0.2 and 1 mJy beam−1, and the noise is relatively uniform and between 40 and 60 µJy,
apart from small regions near bright sources where the noise increases.
region, only sources with a peak brightness greater than 12σ are
included in our catalogue – this value was determined empirically.
The source density plot for a 10 mJy source is shown in Fig. 11, on
the same scale asFig. 10 – whilethere may still be a slight overden-
sity near the centre, the increased noise level in this region, along
with the 6σ cut-off reduces the risk of selecting a spurious source
near sources with weak over-densities.
This analysis was repeated for all sources with a peak greater
than 10 mJy in order to filter spurious sources. The final catalogue
contains 2500 sources – we have erred on the side of caution in or-
der to produce a catalogue with little contamination from spurious
sources. The size of the affected region is correlated with the peak
brightness of a source (with Pearson product-moment correlation
coefficient of 0.53), and the number of spurious sources is also cor-
related with the peak brightness, with correlation coefficient 0.73.
The precise size of the affected region depends on the uv coverage
for the relevant pointing, the time spent on observations and the
local noise levels.
Table 3.1 presents a sample of 60 entries in the cata-
logue, which is sorted by right ascension. The full table is avail-
able via http://www.mrao.cam.ac.uk/surveys/. Column 1
gives the IAU designation of the source, in the form GM-
RTEN1 Jhhmmss.s+ddmmss, where J represents J2000.0 coordi-
nates, hhmmss.s represents right ascension in hours, minutes and
truncated tenths of seconds, and ddmmss represents the declina-
tion in degrees, arcminutes and truncated arcseconds. Columns 2
and 3 give the right ascension and declination of the source, calcu-
lated by first moments of the relevant pixel brightnesses to give a
centroid position. Column 4 gives the brightness of the peak pixel
in each source, in mJy beam−1, and column 5 gives the local rms
noise in µJy beam−1. Columns 6 and 7 give the integrated flux
density and error, calculated from the local noise level and source