What motivates our theoretical attempts to model the SEDs of galaxies? Put another way, what do we hope to learn from such an exercise? Firstly, we should recall that the bolometric luminosity of a starburst galaxy is dominated by the young stars it contains. Thus regardless of how much or how little of this luminosity is reprocessed through the dusty interstellar medium either through thermal emission in the IR of dust grains, through fluorescent processes or through heating and re-emission in an ionized medium, the pan-spectral SED tells us what the star formation rate currently is, or has been in the recent past. The first objective of pan-spectral SED modelling is therefore to be able to reliably infer star formation rates in galaxies and to provide likely error estimates using observational data sets which may in practice be restricted to only certain emission lines or spectral features.
The importance of being able to reliably measure star formation cannot
be under-estimated, since the star formation history of the universe is
a fundamental indicator of galaxy evolution. The famous Lilly-Madau Plot
[62]
provides this quantity on a global scale. This plot shows the star
formation rate (in solar masses per year) per (cosmology corrected)
co-moving volume (Mpc3) vs. the redshift, z. As shown
notably by Ascasibar et al.
[6],
an unacceptable scatter attaches not only to the observational
estimates, and also to theoretical estimates based on the
CDM cosmology. For
the observational data, much of this scatter results from uncertain
corrections for the absorption by dust, and from the application of
different methods with different observational biases and
uncertainties. For the theoretical curves the uncertainties are mostly
due to cosmic variance, the choice of the "prescription" of star
formation which has been adopted, and to the treatment of feedback
between the star formation and the interstellar medium.
Beyond the global picture of the star formation history of the universe
provided by the Madau Plot, there is increasing interest in deriving the
star formation rates of individual galaxies at high redshift to
determine how galaxies were assembled as a function of environment and
of time. At high redshift, massive dusty protogalaxies have been
observed to be forming stars at many thousands of solar masses per year
[70].
If continued, such bursts would convert all the galactic gas into stars
in roughly a dynamical timescale; ~ 108 years. By contrast,
there is increasing evidence that the Lyman break galaxies represent a
more moderate (and probably much less massive) class of object with, in
many cases, little dust obscuration and star formation rates
~ 50 M
yr-1. In order to interpret the SEDs of such distant objects,
we first need to understand how the form of the SED is controlled by the
interstellar physics and the geometry of the stars with respect to the
gas. To do this, we would be best advised to first study and model
nearby star-bursting objects such as Arp 220 in which we can better
resolve the structure of a starburst region, and for which we frequently
have much better quality data over more wavebands than we have for most
high-redshift objects. Once understood, such objects can then be used to
gain physical insight into these more distant sources.
Such theoretical studies of nearby objects provide the second motivation for SED modelling; to gain insights into the physical parameters of starburst galaxies, such as the stellar populations, the atomic and molecular gas content, the star and gas-phase metallicities, the physical parameters such as pressure or mean density of their interstellar media, and the nature of the interstellar dust, its composition and spatial distribution. These physical parameters then drive insight into the physical processes which control them. These include as supernova processing, shock-induced star formation, dust grain processing, synchrotron losses as well as many others.