The Starburst Phenomenon

Radio observations of supernova remnants (SNR), and the relativistic electrons produced by them, greatly increase our diagnostic capabilities for investigations of the starburst phenomenon in galaxies. These observations can give a direct measure of the supernova rate, the massive star formation rate and the high end initial mass function in these galaxies. In addition, through absorption measurements, the SNR can be used to probe the ionised and neutral component of the starburst ISM on parsec scales. A further aspect of this research is that each starburst acts as a laboratory for the statistical study of well defined samples of SNR.

The work is currently strongly limited by brightness sensitivity and hence only a few nearby starbursts have been studied in depth. The sensitivity problem is illustrated with reference to recent MERLIN and VLBI observations. It is concluded that SKA would not only greatly advance research into the physics of starbursts and supernova remnants, but would also enable star formation processes to be investigated in the early universe.

The importance of Starbursts

A major aspect of understanding the evolution of galaxies is to be able to parametrise their star formation history. Several decades ago it was inferred that the Milky Way passed though an era of high star formation in its early history (e.g. Eggen, Lynden-Bell & Sandage 1962), and recent studies of the integrated ultra-violet(UV), optical and infra-red(IR) emission from field galaxies (Madau et al. 1998) have suggested that the universal star formation rate may have peaked at redshift of $z \sim 1.5$. Galaxies with high star formation rates are still relatively common in the nearby universe, and when the star formation rate cannot be sustained for the lifetime of the galaxy the phenomenon is loosely classified as a `starburst'. Studies of nearby starburst galaxies can give a unique insight into the processes and causes of high star formation rates which is not readily accessible in galaxies with high star formation rates at high redshift. However direct studies of optical and UV emission from the stars in these nearby starbursts are often hampered by extinction from the molecular clouds associated with the star formation, and many of their properties are inferred from studies of IR emission from thermal re-radiation by dust which has been heated by starlight. Radio observations of starbursts are not affected by extinction and give independent and, in many cases, more precise measurements of both the star formation rate and the structure of the starburst.

Current Radio Studies

Extended continuum emission

Radio continuum emission gives unique information on the star forming regions in nearby galaxies. Two main processes are responsible for the radio emission - thermal via the free-free mechanism and non-thermal via synchrotron mechanism, the former originating from the HIIregions associated with young stars and the latter is mainly due to relativistic electrons generated by supernova events. For most decimeter studies the non thermal emission dominates, although free-free absorption becomes increasingly important at longer wave lengths (Wills et al. 1997). The existence of a tight corelation between the non-thermal radio and far infra-red luminosity in numerous galaxy samples(e.g. Helou et al. 1985) demonstrated that non-thermal radio emission was also a strong indicator of the star formation rate in galaxies. Condon and others (e.g. Condon 1992) have quantified this process and are able to relate the supernova rate (and hence an estimate of the star formation rate) to the total radio luminosity via relatively simple equations. However it is important to consider that most of the extended emission is produced by relativistic electrons from remnants which have faded away long ago and, compared with studies of the actual remnants (see below), gives a relatively crude measurement of supernova rates.

Compact radio components

The discovery (Unger et al. 1984, Kronberg et al. 1985, Antonucci & Ulvestad 1988) of large numbers of compact radio components in nearby starburst galaxies (Fig. 1.17) has enabled a much higher degree of precision to be brought to studies of supernovae in starburst galaxies. Many of these compact components have non-thermal spectra, high brightness temperatures and show parsec scale shell structures and hence most of these objects are young supernova remnants (Muxlow et al. 1994). In M82, for example, most of these components are more luminous and compact than Cas A, and hence they almost certainly have ages of only a few hundred years. This assertion has been dramatically confirmed by recent EVN measurements of expansion velocities in the M82 remnants (Fig. 1.18) consistent with the youngest remnant having an age of $\sim$30 years (Pedlar et al. 1998). Once the age of a sample of these remnants is calibrated, then a direct measure of the supernova rate can be made by simply making high resolution radio images of these objects. In regions with high supernova rates, the rate can, in principle, be checked by observing the number of new remnants which appear over a few decades or even years. In moving from supernova rate to the rate of star-formation both the initial mass function and the cutoff below which stars do not form type II supernovae needs to be understood. Nevertheless the measurement of the luminosity function and spatial distribution of these SNR in a starburst galaxy can give an accurate measurement of the star formation rate and trace out the structure of star forming regions.

  

Figure: A $\lambda$6cm MERLIN/VLA image of nearby starburst galaxy M82. The discrete sources are mostly supernova remnants with ages less than 1000 years and compact HIIregions. The non-thermal extended background is mainly due to relativistic electrons generated by older remnants.

  

Figure 1.18: An example of a shell supernova remnant seen in M82 by the EVN. The remnant was observed at two epochs and has expanded consistent with an age of $\sim 30$ years (Pedlar et al. 1998).

An additional benefit of these observations is in the area of statistical studies of supernova remnants. Although a large number of remnants in our own galaxy have been detected and parameterised, most of the samples have uncertain distances and were observed with differing linear resolutions and sensitivities, often with different instruments. Green (1984) has outlined the selection effects and other problems with such a sample. However a single radio observation of a starburst galaxy can give large sample of SNR all at essentially at the same distance (e.g. 3200 kpc with a scatter of only 0.5 kpc within the starburst region) and observed with the same angular (hence linear) resolution and sensitivity. Studies are already underway to investigate the relation between size and luminosity, spectral index evolution as well as real time measurements of flux density decay and expansion rate.

There is much work to be done in this area, and many of the simple models of radio supernovae and SNR need to be reconsidered. However perhaps the most important area to urgently investigate is the relation between 'normal' SNR and the radio luminous `hypernovae' of which 41.95+575 in M82 is the best studied example (e.g. Wilkinson & deBruyn 1990). The recent discovery ( Smith et al. 1998) of a number of compact objects in Arp220 at a distance of 75Mpc, with a similar luminosity to 41.95+575, is particularly exciting, as, if we can relate these objects directly to the supernova/star formation rate, it will be possible to measure supernova rates at cosmological distances with the SKA.

Line studies

An increasing number of starburst galaxies have been the subject of neutral hydrogen emission studies. However the brightness temperature limitations of current instruments prevents these measurements having higher resolution than $\sim 5$ arcsec which corresponds to size scales > kpc in all but the nearest starburst galaxies. However, high angular resolution neutral hydrogen measurements can be achieved via absorption measurements, and the absorption spectrum against individual supernova remnants can be used to probe the interstellar medium of the starburst host on parsec size scales. In a recent study of M82, Wills et al. (1998) measured HI absorption against 33 remnants and investigated non-circular gas motions associated with the starburst.

Hydroxyl masers as well as molecular absorption have been observed in nearby starbursts, although much of this work is currently limited by sensitivity. A small number of starbursts, however, contain `megamasers, and can be observed to cosmological distances even with current instruments.

The Potential of SKA for Starburst Studies

Nearby starbursts

Although important advances have recently been made in radio continuum studies of nearby starbursts, almost all the work is severely limited by sensitivity. In principle we have sufficient angular resolution ( via MERLIN, EVN, VLBA etc) to resolve most remnants in starbursts out to 100Mpc (1mas $\sim 0.5$pc at 100Mpc). The major limitation is, of course, the dramatic decrease in brightness sensitivity which is a natural consequence of obtaining high angular resolution with comparable collecting areas. This is evident in the EVN observation of M82, which at 15mas resolution, only detected 5 out of the $\sim 50$ objects detected by MERLIN/VLA despite having comparable flux sensitivity.

Although there is considerable dispersion, it appears that the total flux density of the remnants in M82 decreases approximately inversely as the diameter (Muxlow et al. 1994)

$$S_{5GHz} = 30 D^{-1}_{pc}~~{\rm mJy}$$

and hence the average brightness of an SNR decreases as D^-3 which even in M82, results in the more extended (> 4pc) remnants escaping detection by MERLIN. It is indeed fortunate for present studies that we have two relatively strong starbursts in M82 and NGC253 at $\sim$3Mpc. These two are the most luminous objects within 15Mpc and hence studies of other starbursts are severely limited either by being at greater distances or having lower supernova rates, and hence MERLIN observations of a number of nearby starbursts have only detected relatively small numbers of remnants. SKA with sub $\mu$Jy sensitivity would not only enable much deeper studies to be made of older, more extended remnants in nearby strong starbursts such as M82 and NGC253, but would also result in the detection of statistically usable numbers of SNR in many nearby lower luminosity starbursts.

The ability of SKA to image starburst galaxies at low radio frequencies can also be used to constrain the distribution of ionised gas in the object via free-free absorption studies. A recent subarcsecond study of M82 at 73cm (Wills et al. 1997) has revealed extensive regions of ionised gas which are completely inaccessible in the optical because of extinction.

Finally SKA will enable subarcsecond neutral hydrogen emission studies of nearby starbursts to be made which will complement the current absorption studies and determine the gas dynamics and structures of neutral material associated with the starburst. In addition, by comparing emission and absorption spectra against individual SNR it will be possible to measure the spin temperature of neutral gas in the starburst.

Distant Starbursts

Even the weakest remnants currently seen in M82 would be detectable by SKA at $\sim 100$Mpc and hence supernova remnants in hundreds of starburst (and `normal') galaxies could be studied statistically. This would enable the effect of different environments on supernova remnant evolution to be investigated, as well as constraining the physics of the star formation process. The stronger remnants, particularly `hypernovae' such as 41.9+58 in M82 and possibly the compact objects in ARP220, would be detectable by SKA at cosmological distances ($\sim$1000Mpc), and assuming that these brighter objects can be related to the normal supernova remnants via a luminosity function, it will be possible to measure supernova and star formation rates in the early universe.

One important parameter for these studies is, of course, the angular resolution necessary to separate individual SNR in a starburst. Assuming a typical separation of 10pc then an angular resolution of $\sim$10mas is required at 100Mpc. This would, of course, require baselines of several thousand kilometers, and hence SKA would need to be linked either to the existing VLBI networks or a number of purpose built SKA elements need to be situated at these spacings. Even if SKA is just used as an additional element in existing VLBI networks, then the resulting sensitivity of $\sim 1 \mu$Jy/beam would detect statistically useable numbers of SNR in all starburst galaxies to $\sim$100Mpc.

In addition, as we have discussed in section 2.7.1, the diffuse radio emission from a starburst can be related to the star formation rate (Condon 1992). Such emission would be easily detectable by SKA. For example an object such as M82 would have a total 20cm flux density of $\sim100~\mu$Jy at 1000Mpc. Indeed several of the objects already seen in 20cm MERLIN/VLA observations of the Hubble deep field are extended and are very likely starburst galaxies (Muxlow Private Communication).

Hence it is clear that SKA would enable radio studies of starburst galaxies to be easily carried out at cosmological distances. Unlike radio loud objects, which only constitute a small percentage of all galaxies and also have a wide dispersion in intrinsic properties, starbursts are present in a large fraction of galaxies and appear to form a relatively uniform group. Such studies would be complementary to studies at other wavebands, but would not require complex modeling to account for uncertainties in extinction etc. Hence it seems likely that, with the sensitivity of SKA, starbursts can be used to constrain the parameters of the universe at cosmological distances with high precision.