In these lectures, I have attempted to summarise a physically intuitive approach toward understanding the characteristics of star formation in high-redshift galaxies making use of our intuition gained from the low-z Universe. I have devoted considerable effort towards developing an understanding of the molecular clouds (GMCs) in which most low-redshift star formation occurs in normal galaxies.
I have also provided a summary of the physics of IR emission from optically thick dust clouds. Since there is very little systematic treatment from a theoretical point of view of the observed IR SEDs, I spent considerable effort to develop a systematic approach. Most of these IR sources are centrally heated by a starburst or AGN, so the theoretical framework for the dust heating and radiative transfer can be greatly simplified and distilled. Interpreting the observed SEDs is particularly prone to mistakenly evaluating a dust temperature from the wavelength of the SED peak when in reality a significant range of dust temperatures is contributing to the emergent flux and the peak is equally determined by the overall dust opacity (since obviously the radiation at a given wavelength cannot escape if the opacity is too high at that wavelength). I underscore also the point that the long-wavelength R-J tail fluxes should be used to derive dust (and ISM) masses - not total IR luminosities (as is commonly done for submm galaxies).
Given the known internal structure of the star-forming GMCs, it is reasonable to expect that star formation activity will have two modes: 1) a quiescent mode in which stars form in GMCs at a rate roughly proportional to the mass of H2 gas (accounting for the linear correlations observed between the CO emission and star formation tracers such as FIR over the disks of local galaxies including our own) and 2) a dynamically-triggered starburst mode in which the rate per unit gas mass is elevated by factors of 10-50 accounting for the preferential formation of Hii regions in spiral arms, despite the presence of GMCs throughout the disk, and the high luminosity-to-gas ratios seen in local ULIRGs. For the starburst mode, collisions of GMCs may significantly increase the internal structure/density of the H2 gas and stimulated star formation due to supernovae and expanding Hii regions.
At high redshift, the relative importance of these modes will be shifted, albeit in different directions, by two changes: 1) higher gas-mass fractions (increasing the quiescent mode) and 2) higher rates of galactic merging (increasing the starburst mode). Some observational determinations of the galaxy merger rates at high redshifts have shown very discrepant results for the evolution. However, my recent analysis of evolution in the frequency of galaxy close pairs from COSMOS data suggests merger evolution as (1 + z)2.3, in good agreement with the dark matter halo merger rates seen in simulations.
At high redshift the conditions in the gas may be quite different from those in the star-forming clouds at low redshift. It has been speculated that the stellar IMF might be top-heavy in starbursts or low-metallicity environments. This is not an insurmountable problem - observations of the 4000 Å break may be used to constrain the mass of low-mass stars. Until such observations are done, my conservative opinion would be to avoid the `last refuge of scoundrels' since very little hard evidence is found in the local Universe to support such variations.
To test whether the ingredients discussed above provide a reasonable basis for understanding high-redshift galaxy evolution, I show the results of a Monte Carlo simulation starting at z = 6 with 10 million haloes distributed in mass and with merging rates as found in CDM simulations. Starting with galaxies for which the gas content is 90% of the baryonic mass, the galaxies are evolved including both star formation modes and gas accretion. The simulations clearly show the need for gas replenishment through accretion from the external environment; otherwise, the z = 2 and present-epoch gas contents are far too low.
This need for accretion is simply a reflection at high-z of the well-known requirement of gas replenishment for the Milky Way. Specifically, for the Milky Way the SFR ~ 3 M per year and the present gas content is ~ 3 × 109 M, implying an exhaustion timescale of only 1 Gyr. In the case of local galaxies, it does not appear that the accreting gas is Hi since the high velocity Hi clouds do not constitute a sufficient influx. More likely the inflow is in the form of diffuse ionised hydrogen (Hii) for which imaging with high sensitivity to low surface brightness emission is required. Clearly, this is an important direction for future observations.
Another conclusion from the simulations is that the starburst mode triggered by galactic merging is necessary to account for the high-luminosity power-law tail of the IR luminosity functions at high redshift. The quiescent mode cannot do this since the luminosity generated would simply reflect a scaled version of the galaxy mass function which is exponentially falling at the high mass end.
I would like to thank Zara Scoville for help in editing this manuscript and Andreas Schruba and Kevin Xu for a careful proofreading. I would also like to thank some of my close colleagues over the years who contributed much to this enjoyable research: Herve Aussel, Peter Capak, Peter Goldreich, Jeyhan Kartaltepe, Jin Koda, Colin Norman, Brant Robertson, Dave Sanders, Kartik Sheth, Phil Solomon and Min Yun.