Please Note: the e-mail address(es) and any external links in this paper were correct when it was written in 1995, but may no longer be valid.
Mullard Radio Astronomy Observatory, Cavendish Laboratory, Madingley Road, Cambridge CB3 0HE, UNITED KINGDOM
The CLFST is an East-West Earth-rotation synthesis telescope operating at 151 MHz with a bandwidth of 800 kHz. A single observation takes 12 hours, sampling every 30 seconds, and has a useful field of view of about . The telescope consists of 60 steerable aerials, each composed of 4 Yagi antennae, spaced at irregular intervals along a baseline. This maximum baseline translates to a maximum resolution of . The CLFST has an extremely well-filled aperture, with 784 spacings at intervals, the lowest spacing being .
The object of any Galactic survey is to discover new Galactic objects and study their statistics. Obviously, most Galactic objects will lie close to the Galactic plane, but even directly in the plane, the vast majority of the observed radio sources are extragalactic, and so a fairly large region of the plane has to be surveyed to have a chance of obtaining a reasonable sample of Galactic objects.
Other surveys of the Galactic plane have been made, for example the Effelsberg 2.7-GHz survey of Reich et al. (1990), and a number of surveys cover a substantial fraction of the plane, e.g. the 87GB surveys of Becker et al. (1991), and Gregory & Condon (1991), however these surveys tend to be at relatively high radio frequencies with comparatively low resolution.
The usefulness of a 151 MHz survey is that at low frequencies synchrotron emitters are much more prominent than thermal emitters, so a low frequency survey is ideal to look for Galactic synchrotron emitters, e.g. pulsars and SNRs. No other survey has been made at such a low frequency with such a high resolution.
The fairly large field of view of the CLFST means that 25 survey fields were observed over the four year period to cover the whole survey area, each field being observed for at least 2 days. The fields were chosen so that the survey has complete coverage of the plane over the same latitude range as the Effelsberg survey (). Maps made from individual days' data were combined before a source fitting algorithm was applied to obtain the highest signal to noise possible.
Data processing was performed in a standard way, using interference clipping, removal of bright sources and correction for ionospheric effects. The processing software was custom-written at MRAO for the CLFST system. One thing that is not done is beam deconvolution - this is difficult because the synthesised beam shape is dependent on position, however the excellent u-v coverage of the telescope means that the synthesised beam is very ``clean''.
Interference can come from a variety of sources, and falls into two catagories, namely:
Figure 1: A quadrant in the field, before the removal of Cyg A grating responses. The greyscale is to but is uncalibrated, and should be scaled by a factor of .
Figure 2: The same quadrant as before, with the same greyscale, but with Cyg A removed. The ``shell'' SNR is clearly visible towards the centre-left of the map, along with some extended emission from the Cyg X region towards the south-west of the map.
There are a number of bright radio sources within or close to the Galactic plane which produce grating rings on maps (Figure 1), even though the sources in question may not be within the field of view of the telescope. Cyg A is nearby the field shown, near RA , DEC , and its grating responses are so strong that they dominate the map, obscuring the real sources. These rings can be removed using a standard algorithm (see Figure 2), but problems may arise due to ionospheric effects (see below) which prevent the rings being completely eliminated. This is especially a problem with the three brightest radio sources in the sky, the Crab Nebula, Cas A and Cyg A, all of which lie very close to the survey region and have 151 MHz fluxes of thousands of janskies.
Metre-wavelength radio waves are effected by ionospheric scintillation, and the most obvious problem with maps made from raw data is the ionosphere. The Sun can create waves in the ionosphere which phase-shift incoming radio waves and effectively cause point sources to ``twinkle'', thus smearing them out on a map. So long as the waves are of fairly long wavelength with relatively long coherence lengths and times, it is possible to correct for this by applying a phase correction to each individual 30 second sample. The required set of phase corrections can be calculated by choosing a ``phase calibration source'' which is known or assumed to be unresolved, and then determining the phase corrections which must be applied to in order to actually make that source look like a point source on the map. This assumes a simple linear phase gradient (i.e. position shift) across the telescope for any given sample. The gradients are then simply removed from the data. If the ionospheric waves are short wavelength or have short coherence lengths or times, a correction that is appropriate for one part of the map will not work for another part and in very bad cases the whole observation must be discarded.
The CLFST is about from East-West and is not quite straight. This means that the visibility data should be laid down in a space rather than a plane. Inverting the visibilities using 3-D Fourier transforms is impractical and so the data is extrapolated into a plane and standard 2-D transforms are used. This means that the shape of the synthesised beam depends on the distance from the observation phase centre. The beam becomes ``C''-shaped rather than being a simple Gaussian, the distortion becoming much more noticeable towards the edge of a map. Apart from making beam deconvolution very difficult, this position-dependent distortion also complicates the source fitting procedure.
Since the cause of the beam distortion is known, and the shape of the synthesised beam can be accurately calculated at any point in the sky, an appropriate synthesised beam should be fitted to each source in order to derive its flux and position. Unfortunately, this would take a prohibitive amount of computer time, and a compromise solution is used, whereby a grid of synthesised beams is calculated accurately (a ``beam set'') and the synthesised beam at any point of the map is then interpolated in a linear way from the shapes of the nearest beams to that point (Waldram & Riley (1993)).
Eventually, after a great deal of processing, map making, source fitting to theoretical beams and calibration, the result is a list of approximately 6900 small-diameter sources, some of which are Galactic objects. The sensitivity of the survey is not uniform, but rather depends on two not entirely independent quantities, namely the noise level in each survey field and the mean large-scale sky background temperature seen by the primary beam of the CLFST.
The raw noise level, which depends on the quality of the individual observations, is significantly reduced by the data processing described earlier and by combining more than one day's data together.
Figure 3: noise levels for the 151 MHz Galactic survey.
The CLFST has an automatic gain correction (AGC) system which reduces the sensitivity of the telescope (and thus the ability of the telescope to detect faint sources) when the mean background temperature is large. The mean background changes considerably over the Galactic plane, and so the sensitivity varies across the survey. The flux density scale must be scaled accordingly, in some cases by a factor more than 2.5. The flux scale used is that of Roger et al. (1973).
The noise level ranges from 24 mJy to an extreme of 314 mJy near Cyg A, with a median noise level of 39 mJy. Figure 3 shows the absolute noise level as a function of Galactic longitude for all the fields in the survey.
In the radio continuum, pulsars generally have very steep spectra, with a spectral index (defined by ) of perhaps 2 or more. This makes them very distinctive, but requires a high frequency survey that is sensitive enough to detect very faint compact sources if is to be calculated accurately.
Current pulsar catalogues are heavily influenced by selection effects. The observed concentration of known pulsars within the first and fourth Galactic quadrants does not reflect the true pulsar distribution. There should be undiscovered pulsars within the region covered by this 151 MHz survey that are detectable in the radio continuum.
Around 180 Galactic SNRs are known (Green (1991)), but some authors estimate that almost this number remain undiscovered. SNR catalogues are also heavily affected by selection effects (Green (1991)). The number of SNRs in the Galaxy at any one time and the rate of supernova explosions are quite important numbers to know, since supernova explosions are a primary mechanism for energizing the interstellar medium and re-cycling material into it.
Figure 4: resolution map of IC443.
The CLFST is ideal for looking at extended SNRs because it is sensitive to structures on scales of up to a few degrees. SNRs have synchrotron spectra with spectral indices in the range , and so even though many SNRs lie close to complex thermally emitting regions, they are clearly visible using the CLFST because at low frequencies non-thermal emission is enhanced and thermal emission depressed relative to that seen at high frequencies (e.g. in the Cyg X region, where a number of non-thermal shells are clearly visible at 151 MHz). Previous searches for SNRs at higher frequencies have discovered new extended SNRs in the very recent past, but even if no new extended SNRs are found I will be able to place an upper limit on the surface brightness of any that do exist over a large portion of the Galactic plane.
A number of well-known extended SNRs are within the survey region, for example IC443 (Figure 4), HB3 and HB9, and by comparing the 151 MHz data with images made at other frequencies fairly detailed spectral index maps of the SNRs can be made. The spectral index depends on the shape of the relativistic electron energy distribution, so these comparisons can lead to knowledge of the actual physical conditions within the SNR.
Another important question is whether or not SNRs have pulsars associated with them. A type II supernova explosion is supposed to produce a neutron star, however very few SNRs have ever had pulsars associated with them. Any new suspected pulsar that lies within the boundary of a SNR would be extremely interesting.
In addition to these large SNRs, I am also looking for more compact objects, on the scale of a few arcmin. Such objects are indistinguishable from extragalactic sources using spectral data alone. Assuming an expansion velocity of around any SNR within our own Galaxy that is around a thousand years or more old should have an angular size of a few arcmin, and will therefore be resolved at the maximum resolution of the CLFST. Thus searching for slightly resolved sources with synchrotron spectral indices may be more productive.
Another way of proceeding is to use radio-infra-red comparisons. Galactic SNRs should have a detectable infra-red (IR) flux due to dust heated by the supernova shock-wave, whereas the IR from radio galaxies is negligible in comparison. The IRAS point source catalogue is an ideal place to look for infra-red associations, because it has a very similar resolution to the 151 MHz Galactic survey. Chance matches due to associations with stars can be eliminated using the IRAS colours, and since SNRs have a distinctive ratio of to radio flux compared to that of thermal emitters (e.g. Haslam & Osbourne (1987),Fürst et al. (1987)) they can be distinguished from HII regions and PNe by this ratio or by using the IRAS colours once more. Work on this topic is currently in progress.