Starburst Galaxies

Some galaxies, or their nuclei, show evidence of a recent and transient increase in SFR by as much as a factor of 50. Symptoms of this may be

  • large Balmer-line luminosity and equivalent width
  • high L(IR)/LB
  • rapid gas consumption timescale M(gas)/SFR
  • unusually strong radio continuum emission

    The burst may be galaxy-wide or confined to a small region about the nucleus (few hundred parsec scales). Many (but by no means all) are associated with interacting or merging galaxies. Starbursts are strongly represented in flux-limited samples of UV-bright galaxies (the Markarian catalog) or IR-bright systems. Their optical spectra resembly those of H II regions, with a blue stellar continuum and strong line emission, as seen in the integrated spectrum of the prototype NGC 7714 from the Kennicutt spectral atlas:

    The role of reddening and obscuration in starbursts is complex, and makes detailed interpretation of their continuum and emission-line properties complicated. Calzetti et al. (1994 ApJ 429, 582) derived an effective reddening law, including effects of scattering and the mix of stars and dust, which changes systematically depending on the metallicity of the system and the far-IR fraction. There is evidence that the reddening of the gas and stars is systematically different, perhaps due to dust associated specifically with the gas emission regions, and there is different weighting of gas along the line of sight for different lines, reflecting density and reddening structure.

    Much of the star formation in starburst systems has been found to occur in very luminous, compact star clusters (up to 108 solar luminosities, dimensions of a few parsecs), which occur in bursting dwarfs, interacting galaxies, and mergers; 30 Doradus in the LMC may also be of this type. Some are apparent in this HST UV image of Markarian 357:

    Interest in these clusters is strong for several reasons. The Milky Way is not obviously forming stars in this way now, so they may represent a distinct mode of star formation different from the processes we are familiar with in our own neighborhood. If these objects have a "normal" initial mass function and remain gravitationally bound after the mass loss from massive members is complete, they will eventually look a great deal like globular clusters (which the Milky Way doesn't make anymore either). These clusters are the most dense and intense star-forming environments known, and may be analogs of typical objects in the early epochs of galaxy formation. They come as bright as MV=-15 (the nucleus of NGC 4569 may have gotten above -19 in its younger days), with characteristic sizes of a few parsecs. One is often tempted to take giant H II regions as models of starbursts, though the higher gas density in nuclei and increased role for obscuration when so much local material exists will make for differences. Even so, we see nearby luminous H II regions with (30 Doradus) and without (NGC 604) dominant dense star clusters.

    Starburst mechanisms: Much of the interest in starburst galaxies has been brought on by wondering how some galaxies, and often very small regions in their nuclei, manage to convert so much gas effectively into stars in a very short time. Often there's plenty of molecular gas as judged from CO emission, so it's not a fuelling question so much as a collection puzzle. How can so much molecular gas collect without already forning stars on the way (the analogous issue for fissile material is known as the "fizzle problem").

    The statistics of starbursts may hold a clue - starbursts are notably more common in interacting and merging systems than in more isolated galaxies. While this does not mean that more of them occur in interactions (simply because only about 10% of galaxies are in bound pairs), it does suggest that the conditions are far easier to attain during interactions and mergers. A number of indicators of star formation tell similar stories here. The majority of spirals in pairs experience an increase in SFR typically 30%, while a few experience increases of an order of magnitude. The burst is often confined to a few hundred parsecs near the nucleus, although disk-wide bursts are common. This preference for disturbed galaxies has led to a range of speculations on what causes the enhancements (and thus at least contributes to starbursts).

  • Cloud collisions in a perturbed disk. If gas clouds in a disk have orbits that pass close to one another, a relatively minor perturbation to the potential could cause collisions that would not otherwise take place. Under the widespread assumption that cloud collisions are promising sites for star formation, this could lead to a very sensitive dependence of SFR on perturbations (Lin, Pringle, and Rees 1988 ApJ 328, 103). Struck-Marcell and Scalo (1987 ApJSuppl 64, 39) find that the rate of collisions depends most sensitively on the ratio of timescales between the lifetime of a cloud and the mean collision interval, and that the SFR should undergo large excursions above and below the mean.
  • Collisions between clouds originally belonging to different galaxies. Similarly, during interpenatrating encounters or mergers, clouds might collide as a result of physical overlap of two disks. There is not necessarily any angular momentum barrier here, so a wide radial range of locations could be affected. Collision velocities could become quite high, in which case shock ionization or dissociation of molecular material would be important. Models by Olson and Kwan (1990 ApJ 349, 480) suggest that some fraction of high-velocity cloud collisions must be capable of yielding efficient star formation, to avoid complete disruption of molecular material without forming stars.
  • Bars, including tidally induced bars, and radial gas motions. If a tidally disturbed system has a large enough region with a solid-body-like rotation curve, the companion can induce a bar that last much longer than the encounter itself. Noguchi (1988 A&A 203, 459) has shown that this can lead to substantial channelling of gas into the nuclear region, perhaps leading to a nuclear starburst. In this case, interaction-induced star formation might outlive obvious morphological evidence of an encounter, except for the presence of the bar.
  • Tidally induced density waves. Tidal perturbations can produce very pronounced spiral density waves, as in the well-studied case of M51 (Toomre and Toomre 1972 ApJ 178, 623, Howard and Byrd 1990 AJ 99, 1798). These potential minima provide favorable sites for accumulation of interstellar matter and star formation, with the enhancement being in the fraction of the disk occupied by strong density peaks rather than any new process for triggering star formation.
  • Disk instabilities produced by perturbations in the potential. There is a remarkable agreement between regions of spirals in which star formation is observed, and regions in which gas is unstable by the Toomre (1964 ApJ 139, 1217) and Quirk (1972 ApJL 176, L9) dynamical criterion (Zasov and Simakov 1988 Astrof, 29, 190; Kennicutt 1989 ApJ 344, 685). A companion could distort the rotation curve (and local epicyclic frequency) by enough to render additional gas in the outer disk susceptable to collapse (perhaps via a phase change into molecular material). This seems to have been first suggested by E. Laurikainen. Unresolved issues include whether the interaction and collapse timescale are compatible, and the spatial distribution of resulting star formation.
  • Dumping of gas into E/S0 systems. Otherwise gas-poor galaxies could acquire significant material for star formation if physical transfer of gas takes place during an encounter with a gas-rich system. Sotnikova (1988 Astrof. 28, 495) found that the proper conditions for gas transfer may exist in a significant fraction of close encounters between appropriate galaxy types, but it is unclear from emission-line statistics how often this might take place. Curiously, when we clearly see such mass dumping take place, the galaxy receiving the gas may have no detectable star formation, perhaps meaning that a critical mass (or surface density) is needed to trigger a brief starburst after a prolonged accretion episode.
  • Direct impact of gas-rich dwarf satellites into disks: This is a large-scale variant of the picture of cloud collisions between two galaxies. In this case, the obvious interaction with a bright companion is not the one causing the fireworks. Statistics of faint companions are not yet well enough determined to tell how important this process might be.

    From general considerations, some of the induced star formation must be triggered by processes not requiring direct contact of disk material from different galaxies; some objects with high SFR are too far apart, and relatively undisturbed, so that internal effects of tidal stress must be responsible. Detailed modelling is thwarted by the great range of relevant physical scales in some of these cases. In testing these proposals, studies of the ISM in interacting systems, and understanding their dynamics, are crucial. For example, H2 masses in combination with SFR estimates can suggest whether the SFR goes up because of creation or accumulation of new molecular gas (and normal accompanying star formation), or via an enhancement of the "efficiency" of star formation. A survey of 13 merger candidates by Young et al. (1986 ApJL 311, L17) suggested that the SFR reflects large molecular gas content; more recent results (Young, IAU Symp. 146) extend this by suggesting that the H2/H I mass ratio is systematically larger in interacting systems than in normal spirals. CO surveys of complete and well-understood sets of both interacting and non-interacting galaxies are urgently needed (and in progress).

    For very luminous galaxies which are dusty enough that most of their power emerges in the far-IR (once known as IRAS galaxies, now sharing such acronyms as LIRG, ULIRGs, PIGs, or ELFs), it can be subtle to tell whether the dominant energy source is a starburst or AGN. Compact, flat-spectrum radio sources indicate an AGN, but more diffuse radio emission can come from star-forming nuclei as well. Condon and Broderick (1988 AJ 96, 30) have introduced a ratio of radio and far-IR flux densities as a discriminant, based on the empirical relation found for star-forming regions and the fact that powerful AGN are usually more radio-loud. Mid-IR spectra have proven to be very useful, since these photons emerge through the surrounding dust. High-ionization species indicate an AGN, while their lack and strong PAH features (destroyed by the intense hard radiation from an AGN) suggest a starburst. Laurent et al. (2000 A&A 359, 887) find that while the "unidentified" PAH band at 6.2 microns occurs only in starbursts, there is dust continuum emission from 3-10 microns which is characteristic only ofthe very hot region around an AGN.

    Of course, a relatively unobscured nucleus can be classified from its optical spectrum. Some nuclei of both flavors are so dusty that opacity effects control what we see in the visible range. As examples, I'll point out the dusty nuclei of NGC 253 and 2903, in which the dust blocks most of the star formation, looks chaotic, and shows streamers probably associated with global winds. To stress how powerful these differential opacity effects are, the nucleus of NGC 1614 has a large Balmer decrement (Hα/ Hβ = 10) but a flat UV spectrum and detectable Lyman α emission, so that we are seeing different regions at different wavelengths. Here's the optical image of the center of NGC 253, from the Hubble Heritage collection:

    The high energy densities, both in starlight and mechanical input through stellar winds and supernovae, can actually unbind the ISM from starburst galaxies. The heated ISM can set up a global (or super) wind, detetcable in optical line emission, scattered starlight, and soft X-rays (most prominently from the interface at the edge of the roughly conical outflow). Most of the escaping matter can be so hot that we don't even see it in X-rays, cooling only at the interface with less disturbed ISM. This wind may be important in forming early-type galaxies, since one has to sweep the gas out of a merger product if it's going to end up as an elliptical. Something like this seems to have happened early in the history of clusters and groups, since intracluster X-ray gas shows chemical traces of having been processed by massive stars. The best-known example of a starburst wind is blowing out of M82, as shown in this image with Hα emission coded red. Compare the Chandra image (somewhat rotated) to see how even the part of the wind that does show up can dominate the X-ray emission. Winds are also often seen via P Cygni profiles of some absorption lines - Na D has been used for optical surveys, and these lines are so strong in high-redshift Lyman-break galaxies that they make it difficult to get an accurate redshift for just the stars, much less see stellar absorption lines from many atomic species.

    M82 outflow

    Starbursts may be the best local analogs to galaxies during their formation, with large amounts of both gas and stellar energy input present. Indeed, many high-redshift galaxies shows the characteristic UV spectra of very young stellar populations. The implications of this are not really straightforward, though because of selection effects in both UV flux (the kind we see when redshifted at z=4 or so) and surface brightness (so that star-forming objects and regions within them are the easiest things to identify at large redshift). A cosmology with expanding spacetime gives a surface-brightness dimming going as 1/(1+z)4, which is very substantial for cosmologically interesting redshifts. Of special interest is Lyman α emission, which generally traces a wind (radiative-transfer effects make this the easiest geometry for the line photons to escape without being converted into something else during resonant scattering - see a spectral example).

    Starburst galaxies have calculated star-forming rates as high as hundreds of solar masses per year (exhaustion timescales of order 108 years), and correspondingly high expected supoernova rates. Searches for the expected supernovae have had mixed results. High-resolution radio observations of M82 and NGC 253 shows rich collections of small (sometimes fading and expanding) sources that are just right to be radio-bright supernova remnants, so that part checks out. Looking for the supernovae themselves has been less successful, with only a handful seen in starburst nuclei (against formidable background and confusion problems). There has been a better track record in near-IR monitoring, such as finding an obscured SN in NGC 3690 within a fairly short time. However, this becomes a very intense use of telescope time, so it has yet to be pursued on an appropriately large scale.

    What do fading starbursts look like? Stellar evolutionary models lead us to expect galaxies that are fairly blue (but rapidly reddening with time unless the burst was of large relative mass amplitude), whose spectral features are dominated by either supergiants or the upper main sequennce. This would account for the "E+A" galaxies which show a mixture of old and intermediate-age spectral features, and for the small population of cluster members with anomalously strong Hδ absorption, since this line will be the most prominent unconfused feature against an older background after ~109 years. It is still unclear whether the relative numbers of starbursts and post-starbursts are right to conclude that we understand the connection, since such different technques are use to recognize them.


    « Star formation in galaxies | Environmental effects on galaxies »
    Course Home | Bill Keel's Home Page | Image Usage and Copyright Info | UA Astronomy

    keel@bildad.astr.ua.edu
    Last changes: 9/2009	  © 2000-2009