Virtually all Star Formation Rate (SFR) tracers are fundamentally tracers of massive star formation, as these dominate the radiated energy of a stellar population. Extensive reviews ofSFRdiagnostics are given byKennicutt(1998) andKennicutt & Evans(2012), and their use to determine the evolution of star formation with redshift (Madau & Dickinson, 2014). Different observational tracers are sensitive to different stellar mass ranges, hence they respond differently as a function of age. Observational tracers of photoionised regions are sensitive to the shortest-lived OB-stars andWRstars with ages < 10 Myr, whereas the near to far-UVcontinuum is produced by stars over a broader mass range with longer lifetimes (20–100 Myr). An understanding of the underlying massive stellar population is therefore crucial for the correct interpretation of the majority of
SFRindicators, with crucial parameters including ionising luminosities and the lifetimes of the most influential evolutionary phases.
The UV luminosity of a galaxy is the most direct measure of the instantaneous
SFR density (e.g., Schiminovich et al. 2005). The far-UV luminosity, LF U V, around
(rest-frame) 1500 ˚A is commonly used as a diagnostic. Absorption by the neutral Inter- Galactic Medium (IGM) is too severe shortward of this. An advantage of this method is that this portion of a galaxy’s SED is shifted to optical wavelengths at intermediate redshift (z > 1.4), allowing ground-based observation. The conversion between LF U V
andSFR is not straightforward. It is in principle possible to predict the UV output of a stellar population and its evolution with age (e.g.,Schaerer & Vacca 1998), however these predictions are sensitive to the assumedIMF, and the specific models of stellar evolution used, particularly for the most massive stars. In Figure 1.3 I show an example of the predicted far-UV luminosity for populations of single-stars at Z and 1/10Z , showing
the dominance of young massive stars. Populations of single-stars are often assumed, yet binary interactions can significantly alter the integrated radiative output of a population
Figure 1.3: The evolution of luminosity at three different UV wavelengths in a single stellar population, at metallicities Z (solid) and 0.1 Z (dotted). Adapted from Madau
& Dickinson (2014).
of massive stars (Eldridge,2009). The most significant drawback of UVmethods is dust extinction, as UV wavelengths are heavily attenuated by even modest column densities.
Calzetti (2001) review the effects of dust extinction on UV-derived SFRs.
To supplement directUV-basedSFRindicators, various indirect measures exist, tak- ing advantage of the effects massive stars have on their surroundings. Dust that ab- sorbs UV photons re-radiates the energy at mid and far-IR wavelengths, and the IR
(∼10–300 µm) luminosity of a galaxy is generally considered to be directly proportional to this absorbed energy from star formation (Kennicutt, 1998). The far-IR luminosity of nearby late-type (star-forming) galaxies has shown consistency with other SFR in- dicators (Lonsdale & Helou, 1985; Sauvage & Thuan, 1992; Buat & Xu, 1996), but in quiescent galaxies it is thought lower-mass stars are the primary dust heaters, resulting in overestimates of SFR. Observations of both rest-frame far-UV and far-IR luminosi- ties should therefore measure all energy from star formation in a galaxy. This is possible for nearby galaxies, whose resolvable stellar contents may provide essential calibrations betweenUVandIRmethods (e.g.,Doran et al. 2013). However, while optical (and near-
IR) observations can select ‘normal’ star-forming galaxies to increasingly high redshift (z>4,Bouwens et al. 2015), the most sensitive far-IRobservatories such as Herschel are
limited to extreme objects above z ' 2 (Madau & Dickinson, 2014).
Recombination lines such as H-α, Ly-α, and Pa-α are commonly used (seeMoustakas et al. 2006for a comparison), as these trace photoionisation of theISMcaused by theUV
radiation from young massive stars. In some galaxies, an Active Galactic Nucleus (AGN) may contribute to these lines, corrupting SFR measurements. However, the harder ionising source in such galaxies can be readily diagnosed using emission lines of other elements (O & N) (Baldwin et al., 1981).
Finally, the comoving volumetric core-collapse supernova (ccSN) rate - signalling the death rate of massive stars - also contains information about the SFR. Observationally, this has been estimated to z ' 1 (Dahlen et al., 2012), showing good agreement with the other tracers discussed. This method does, however, require assumptions about the maximum stellar mass producing ccSN, which remains uncertain (Heger et al., 2003). Long Gamma Ray Bursts (GRBs) have been associated with some type Ic SNe at low redshift (e.g.,Galama et al. 1998), and have been detected to very high redshifts (z ∼ 9,
Cucchiara et al. 2011). Porciani & Madau (2001) argue that the long-GRB rate may be used as an independent measure of SFR in the early universe, but their association with massive stars is metallicity-dependent and difficult to model (seeWoosley & Bloom 2006for a review). They have also been shown to be biasedSFRtracers, occurring more frequently per unit stellar mass at earlier times (Kistler et al., 2009;Robertson & Ellis,
2012).
As well as being central to SFR diagnostics, massive stars in large numbers can have a profound effect on integrated galaxy spectra. Galaxies undergoing extraordinary star-formation (∼100 M yr−1), known as ‘starburst’ galaxies, have spectra featuring
nebular emission lines from gas photoionised by massive stars. Such galaxies were first associated with higher than average star formation rates by Searle & Sargent (1972). Nearby starburst galaxies are generally observed as either blue and irregular, or veryIR
luminous resulting from a high dust content. The latter are often referred to as (Ultra) Luminous Infra-Red Galaxys (LIRGs).
Prior to the previous two decades, few galaxies of any kind were known at redshifts, z > 1. This changed with a technique pioneered by Steidel & Hamilton (1993), utilis- ing the sharp flux discontinuity (or ‘break’) at the Lyman limit (912 ˚A), to detect large samples of star-forming galaxies at z ∼ 2–3, for which this ‘Lyman-break’ is shifted to op- tical wavelengths where adjacent filters may be used to measure a colour excess. Far-UV
emission from hot stars emphasises this flux discontinuity, as it is absorbed by Hydrogen both in the atmospheres of some hot stars and theISM, at wavelengths < 912 ˚A.
Lyman Break Galaxies (LBGs) have become a target of abundance measurements (Pettini, 2004), predominantly by two approaches: direct observation of stellar UV fea- tures, or analysis of nebular lines. Rix et al.(2004) showed that the integrated rest-frame
UVspectra of Lyman Break Galaxy (LBG)s could be used to measure metallicity to z∼3, by comparing the observed photospheric and wind lines of OB-stars to those in synthetic spectra. Heckman et al. (1998) also show that for nearby starburst galaxies, the equiv- alent widths of some UV wind lines are proportional to O/H - as a result of stronger stellar winds at higher metallicity. Such methods are of course subject to our under- standing of massive star evolution, and the metallicity-dependence of stellar winds (see Section 1.4.1). Indeed, some spectral features in the bright rest-frame UV spectra of gravitationally lensed systems cannot as yet be reproduced by population spectral syn- thesis (e.g., Quider et al. 2009). Despite the potential of rest-frame UV diagnostics, well-established nebular diagnostics using H Balmer lines, such as the R23 (McGaugh,
1991) and N2 (Pettini & Pagel, 2004) methods, have tended to prevail. For example,
Pettini et al.(2001) detect such nebular diagnostic in near-IR spectra ofLBGs at z ∼ 3, finding consistency between star formation rates derived using Hβ and LF U V, and sub-
solar oxygen abundances that seem to exceed other structures at similar redshifts (e.g., damped Lyα systems, Wolfe et al. 2005). Such measurements provide constraints on nucleosynthesis in massive stars when the Universe was only ∼ 15% of its current age. However, limited atmospheric transmission in the near-IR makes it challenging to ob- tain nebular diagnostics for LBGs at such redshifts, making rest-frame UV diagnostics more appealing. Any improvements in our understanding of massive star evolution, particularly metallicity-dependent effects, will therefore aid this exciting pursuit.