X99018


Publ. Astron. Soc. Aust., 1999, 16, 66–9.

Deep Observations with the

Parkes 21-cm Multibeam System

M. J. Disney, P. J. Boyce, G. D. Banks,

R. F. Minchin and A. E. Wright

University of Wales, Cardiff, PO Box 913, Cardiff CF2 3YB, UK
mjd@astro.cf.ac.uk, pjb@astro.cf.ac.uk, gdb@astro.cf.ac.uk, awright@atnf.csiro.au

Received 1998 November 2, accepted 1999 February 8

Abstract: We report on a preliminary analysis of a 5600 sec per point survey of
32 square degrees in Centaurus, carried out with the Parkes 13-beam system. The
signal-to-noise ratio is found to improve as

√
tobs for the whole integration. We

have detected 102 HI sources between +250 and +12,700 km s−1 either by eye or by
using the new galaxy-finding algorithm PICASSO. Over half of these are new HI
detections. Around a dozen of these are not associated with catalogued galaxies and,
in two of these cases, we have not identified an optical counterpart on the Digitized
Sky Survey. Arguments are put forward to explain why deep integrations are needed
to find low surface brightness objects.

Keywords: galaxies: distances and redshifts — galaxies: luminosity functions —
galaxies: mass function — galaxies: statistics

1 Why Go Deep?

It is worth remembering that HIPASS is a shallow
survey, with only 450 sec integrations per point, and
this may severely limit what we can find with the
survey (Staveley-Smith et al. 1996). Therefore, it
was always intended that much deeper observations
would be made of limited areas of sky, though
it was never certain how long such integrations
could be profitably carried on, i.e. for how long the
signal-to-noise ratio would continue to improve as√
t (Disney & Banks 1997). We report here, on

behalf of the DEEP team, some results from our
early experiments.

Reasons for going deep are:
(a) To find intrinsically fainter dwarf galaxies.

HIPASS (Banks et al. 1999) is capable of finding
≥107 M¯ of HI in a dwarf at a distance of 3·5
Mpc. In view of the present controversy over the
faint end of the luminosity function (Phillipps et
al. 1998; Trentham 1997) it is desirable to go deeper,
particularly in nearby groups.

(b) We should expect that low surface brightness
objects will generally have low HI column densities—
and this is born out by observations (Bothun, Impey
& McGaugh 1997). However, system noise puts a
limit to the lowest column density attainable by
any radio telescope, irrespective of size, i.e.

NHI ≥ 1018Tsys
√

∆V (km s−1)/tobs(s) (cm
−2) (1)

(e.g. Disney & Banks 1997) and NHI and surface
brightness µB (in blue mag per square arcsec) will
be related through

NHI ∼ 1020(MHI/LB)100·4(27−µB) . (2)

Thus the HIPASS survey would appear to have a
column density limit (Ts ' 23 K, ∆V ' 150 km
s−1) of around 2×1019 cm−2, which corresponds to
a mean SB (if MHI/LB ' 0·5) of 28 Bµ over the
whole hydrogen radius. If the HI radius is twice
the optical (van Zee, Haynes & Giovanelli 1995)
and if most of the light from the exponential disk
comes from within four optical scalelengths then
the central SB, µ0(B), of a galaxy with 28 Bµ as
defined above will be about 24 Bµ. Thus we cannot
expect to find many LSBGs with HIPASS. We need
to go 1–2 magnitudes fainter, i.e to integrate for 6
to 36 times as long.

(c) A rather different way to approach this is
to ask what we would have to do to find optically
invisible objects in HI:

to detect in HI:

distance d < dmax (MHI) ∼ (MHI/fmin)
1
2 ;

not to detect in optical:

distance d > dmax(LB) ∼ (LB/lmin)
1
2 ,

where fmin and lmin are the minimum detectable HI
and optical fluxes respectively. Combining these:
for an invisible HI detection

qAstronomical Society of Australia 1999 1323-3580/99/010066$05.00

https://www.cambridge.org/core/terms. https://doi.org/10.1071/AS99066
Downloaded from https://www.cambridge.org/core. Carnegie Mellon University, on 06 Apr 2021 at 01:46:04, subject to the Cambridge Core terms of use, available at

https://www.cambridge.org/core/terms
https://doi.org/10.1071/AS99066
https://www.cambridge.org/core


Deep Observations 67

MHI

LB
>
fmin(HI)
lmin(opt)

(independent of distance) . (3)

Now we can calibrate (3) for the Multibeam system,
using our recent HIPASS survey of the Cen-A group
(d = 3·5 Mpc) (Banks et al. 1998) where we found
Mmin = 107 M¯, tobs = 450 s and the faintest
optical identification at mB = 19m. We can thus
construct Table 1 for the minimum MHI/LB at
which we could find invisible objects in different
surveys.

Table 1. Minimum M HI/LB ratios for optically invisible
galaxies to be detected in different Parkes Multibeam surveys

Survey (MHI)¯/(LB)¯
Normal galaxy LSBG VLSBG

HIPASS (450 s) 25 2·5 0·6
5×HIPASS 11 1 0·25
12·5×HIPASS 7 0·7 0·2
25×HIPASS 5 0·5 0·1

The reason for the three different columns in
Table 1 is that low SB galaxies show only part of

their light above the sky. Thus a LSBG in the
table is assumed to show only 10%, the VLSBG
only 2·5%, whereas the normal galaxy is assumed
to show all (not of course strictly true).

Table 1 makes it clear that to find objects with
the Multibeam which are not on the Digital Sky
Survey and yet have MHI/LB values that are not
too extreme (≤1), it is necessary to observe for at
least five times longer than HIPASS.

2 Noise in the Deep

Given the novel nature of the multibeam system and
the active scanning mode in which data are taken,
there was no certainty that the noise would decrease
with integration time as predicted. We have carried
out an experimental survey to a depth of 12·5 ×
HIPASS, i.e. 5600 sec per beam, in a 4×8 degree
region of Centaurus, taking all the data at night. The
signal-to-noise in most parts of the frame (Figure 1)
increases with

√
tobs throughout. The main sources

of noise appear to be background continuum objects
and characteristic long-wavelength ripples (caused
by standing waves between the focus box and the

Figure 1—A portion of the Deep Cen survey cube with velocity in the horizontal direction, declination upwards. Note
the Galactic HI emission at zero velocity on the left, continuum sources at declinations of −26 and −33, a real source at
declination −30 and velocity around 4000 km s−1, edge noise at the top and bottom and the long baseline ripple with period
∼ 1000 km s−1.

https://www.cambridge.org/core/terms. https://doi.org/10.1071/AS99066
Downloaded from https://www.cambridge.org/core. Carnegie Mellon University, on 06 Apr 2021 at 01:46:04, subject to the Cambridge Core terms of use, available at

https://www.cambridge.org/core/terms
https://doi.org/10.1071/AS99066
https://www.cambridge.org/core


68 M. J. Disney et al.

Figure 2—HI spectrum and DSS image of one of the unassociated galaxies. The encircled object
(marked on the DSS image) is the only galaxy within 5′ and has a velocity of >12,000 km s−1.

surface of the dish) which look nothing like real
21-cm sources. Thus there is hope of ‘cleaning up’
such cubes and going deeper still.

3 Early Results

We have searched the 8 × 4 degree cube by
eye, and then again using PICASSO (father of

cubism), an automatic galaxy-finding algorithm we
have developed at Cardiff (Minchin 1999, present
issue p. 12). By and large the two searches agree,
though not perfectly. As a result we found a total
of 102 21-cm sources in the cube (to a flux limit of
1 Jy km s−1). This is a conservative lower limit to
the number we expect to find eventually and note

https://www.cambridge.org/core/terms. https://doi.org/10.1071/AS99066
Downloaded from https://www.cambridge.org/core. Carnegie Mellon University, on 06 Apr 2021 at 01:46:04, subject to the Cambridge Core terms of use, available at

https://www.cambridge.org/core/terms
https://doi.org/10.1071/AS99066
https://www.cambridge.org/core


Deep Observations 69

that the edges of the cube will be noisier due to
the survey strategy. Since the number of detected
sources expected per unit area should go as t3/4obs
this implies the number of detections for the whole
sky in HIPASS to be 102×(40,000)/32 × (12·5)−3/4
∼ 20,000. This, however, is a very rough estimate
because of clustering statistics.

Of the 102 sources found, 41 were identified
with catalogued galaxies with previously determined
redshifts. For most of the remaining 61 sources,
there is a catalogued galaxy without a measured
redshift lying within 4′ of the fitted HI position.

However, there are around a dozen sources for
which no such association could be made. Studies of
the Digitized Sky Survey fields around these sources
revealed possible optical counterparts in most cases.
However, there are still two for which no counterpart
has been identified. Figure 2 shows the spectrum
of one of these objects along with an image of
the surrounding DSS field. We are undertaking
CCD imaging of these candidate very low surface
brightness galaxies.

4 Future Work

To summarise:
(1) We have been awarded time to extend our

survey of this area to 23 × HIPASS integration
time.

(2) We have surveyed an area in Sculptor (8 × 8
sq. deg.) to 5 × HIPASS integration time. These
data are currently being analysed.

(3) With the aid of Quentin Parker we are
studying Tech-pan films of the Cen-A area to look
for dimmer counterparts.

(4) We are presently carrying out follow-up CCD
observations of the ‘invisibles’.

(5) We are carrying out simulations to improve
and validate the automatic galaxy-finding algorithm.

(6) We are experimenting with ways of cleaning
cubes before sending them to the ‘galaxy finder’.

Acknowledgments

Other members of the DEEP consortium include
Ian Stewart, Mark Price, Ron Ekers, Ray Haynes,
Erwin de Blok, Ken Freeman, Helmut Jerjen, Pat
Knezek, David Malin, Mary Putman, Brad Gibson,
Dan Zambonini and Quentin Parker. GB and RFM
acknowledge the support of PPARC post-graduate
research awards.

References

Banks, G. D., et al. 1998, ‘New galaxies discovered in the first
blind survey of the Centaurus A group’, ApJ, submitted

Bothun, G. D., Impey, C., & McGaugh, S. S. 1997, PASP,
109, 745

Disney, M. J., & Banks, G. D. 1997, PASA, 14, 69
Minchin, R. F. 1999, PASA, 16, 12
Phillipps, S., Parker, Q. A., Schwarzenberg, J. M., & Jones,

J. B. 1998, ApJ, 493, L59
Quintana, H., Ramirex, A., Melnick, J., Raychaudnury, S.,

& Slezak, E. 1995, AJ, 110, 463
Staveley-Smith, L., et al. 1996, PASA, 13, 243
Trentham, N. 1997, MNRAS, 286, 133
van Zee, L., Haynes, M. P., & Giovanelli, R. 1995, AJ, 109,

990

https://www.cambridge.org/core/terms. https://doi.org/10.1071/AS99066
Downloaded from https://www.cambridge.org/core. Carnegie Mellon University, on 06 Apr 2021 at 01:46:04, subject to the Cambridge Core terms of use, available at

https://www.cambridge.org/core/terms
https://doi.org/10.1071/AS99066
https://www.cambridge.org/core