What can distant galaxies teach us about massive stars?
aa r X i v : . [ a s t r o - ph . GA ] F e b The Lives and Death Throws of Massive StarsProceedings IAU Symposium No. 329, 2017J.J. Eldridge, editor. c (cid:13) What can distant galaxies teach us aboutmassive stars?
Elizabeth R. Stanway Physics Department, University of Warwick, Gibbet Hill Road, Coventry, CV4 7AL, UKemail: [email protected]
Abstract.
Observations of star-forming galaxies in the distant Universe ( z >
2) are starting toconfirm the importance of massive stars in shaping galaxy emission and evolution. Inevitably,these distant stellar populations are unresolved, and the limited data available must be in-terpreted in the context of stellar population synthesis models. With the imminent launch ofJWST and the prospect of spectral observations of galaxies within a gigayear of the Big Bang,the uncertainties in modelling of massive stars are becoming increasingly important to our inter-pretation of the high redshift Universe. In turn, these observations of distant stellar populationswill provide ever stronger tests against which to gauge the success of, and flaws in, currentmassive star models.
Keywords. galaxies: evolution, galaxies: high-redshift, stars: luminosity function, mass func-tion.
1. Introduction
As studies elsewhere in these proceedings have shown, as many as ∼
70% of massivestars are expected to interact with a binary companion during their lifetimes. Thesestars dominate the integrated light of young stellar populations ( <
100 Myr) and theeffects of binary interactions are typically more pronounced at significantly sub-Solarmetallicities. This presents an interesting opportunity: the low metallicity, highly star-forming galaxies we observe at high redshifts both require an understanding of massivestellar evolution for their interpretation, and represent a laboratory in which to test thatunderstanding. This synergy has recently been recognised by a growing subset of both theextragalactic and massive stars communities, and will only become stronger as the adventof highly-multiplexed near-infrared spectroscopy from the ground is complemented by theeagerly-anticipated
James Webb Space Telescope (JWST).Here I review recent observational indications of the presence and influence of massivestars in the distant Universe, as well as discussing the the role of stellar modelling intheir interpretation and the possible insights this synergy enables.
2. Observing the Distant Universe
Over the last twenty years, the number of spectroscopically-confirmed galaxies in thedistant Universe (in this context, z >
2) has grown exponentially, from a mere handful totens of thousands of sources. The primary driver of this process is the ‘Lyman break tech-nique’ first applied on a large scale by Steidel et al. (1996), and later extended to higherredshifts (e.g. Stanway et al. 2003; Bouwens et al. 2011). This method allows the selec-tion of high redshift galaxy candidates by their distinctive photometric colours. Thesearise from a strong discontinuity imposed on their rest-frame ultraviolet spectrum dueto absorption by neutral hydrogen in the intergalactic medium (IGM). It preferentially1 E. R. Stanway F l ux / - e r g / s / c m / A Si IV Si II CIV He II C III] F l ux / - e r g / s / c m / A Composite of 811 z~3 LBGs (Shapley et al 2003)VUDS 530034129 at z=3.47 (Tasca et al 2016)
Figure 1.
Example spectra of high redshift sources, highlighting the features which confirm thepresence of massive stars. Indicated lines are Si IV, Si II, C IV, He II and C III] - all lines whicheither include a direct component from massive stars or which are powered by the ionizingradiation of massive stellar populations. Early composites of Lyman break galaxies at z ∼ z = 3 .
47 source from the VUDS survey (Tasca et al 2016). selects galaxies with high ultraviolet luminosity, and thus those with some componentof on-going star formation. The redshift of these can be confirmed (for many cases) byfollow-up spectroscopy of the rest-frame ultraviolet, redshifted into the observed optical.At very high redshift ( z ∼ −
7) spectroscopic characterisation is often restricted todetection of an isolated Lyman- α emission line ( λ rest = 1216˚A), with perhaps a sec-ond strong UV feature (e.g. He II or CIII]) to provide confirmation in rare cases (e.g.Stark et al. 2015). Beyond z ∼
7, the rising neutral hydrogen fraction in the IGM, com-bined with the shift of the Lyman break into the near-infrared, means that spectroscopicconfirmation is very seldom possible, but the properties of galaxies may still be inferredfrom their photometry (e.g. Caruana et al. 2014; Smit et al. 2014).2.1.
Rest-frame Ultraviolet Spectroscopy
Rest frame ultraviolet spectroscopy can be interpreted through the construction of com-posites which yield the ‘typical’ properties of galaxies in a population or subset thereof(e.g. Shapley et al. 2003). These have been complemented by observations of individ-ual sources, either particularly bright, or lensed targets, or simply using extremely deepspectroscopy (e.g. Tasca et al. 2016). While individual sources show variation, figure 1illustrates certain features that are common in the high redshift galaxy population. Ei-ther directly (through features arising from the stellar spectra) or indirectly (throughemission from the nebular gas of H II regions), these are diagnostic of the massive stellarpopulation in these galaxies.Clearly the strongest and most obvious of these features is the Lyman- α emission line.A full analysis of the emission and radiative transfer of this resonantly-scattered line is istant galaxies and massive stars α (mostly at z .
5, althoughwith some exceptions), the rest-UV contains other diagnostics of massive stellar pop-ulations. As figure 1 shows, the prominant absorption features of C IV 1548,1550˚A,Si IV 1393,1402˚A and Si II 1526˚A are all seen in both composite and individual spec-tra. Similarly the He II 1640˚A and CIII] 1907,1909˚A emission features appear to befar more common in the distant galaxy population than in local star forming system.Each of these may have broad (stellar wind-driven) and narrow (nebular) components,but show strengths that are difficult to reproduce with conventional stellar populations(Shapley et al. 2003). The emission lines in particular, are also indicative of a far harderionizing spectrum than that seen in local sources. A few rare He II emitting star form-ing regions in the local Universe are interpreted as hosting massive, often Wolf-Rayet,stars (Kehrig et al. 2015). Models incorporating more detailed analysis of massive stellarpopulations, either in terms of rotation or binary interaction, and exploring these effectsat sub-Solar metallicities are proving both necessary for and successful in simultaneousfitting of these line strengths in the high redshift population (e.g. Eldridge & Stanway2012; Steidel et al. 2016).2.2.
Rest-Frame Optical Spectroscopy
The recent advent of multi-object near-infrared spectrographs on 8-10m class telescopeshas had a strong impact in this field. The rest-frame optical spectra of z ∼ − α ratio isprimarily sensitive to the shape of the ionizing spectrum just above 1 Rydberg, while the[O III]/H β ratio probes the 1-3 Rydberg range. The offset in the population medians for z ∼ z < . z ∼ − . -1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0Log [NII]/H-alpha-0.50.00.51.0 L og [ O III] / H - b e t a z=2.3 galaxies Extreme z=0.1 galaxies SDSS AGN S t a r F o r m i ng
200 400 600 800 1000 1200 1400 1600Wavelength \ Angstrom0.00.20.40.60.81.01.2 L u m i no s it y p e r A ng s t r o m O • , binaries 0.1 Z O • , single stars1 Z O • , binaries 1 Z O • , single stars Figure 2.
The importance of the hardness of the ionizing spectrum. (Left) The often-used BPTdiagram which measures the hardness of the ionizing spectrum both close to 1 Rydberg (the[NII]/H α ratio) and at 1-3 Ryd ([OIII]/H β ). Greyscale indicates the distribution of local starforming galaxies at near Solar metallicity from the SDSS. Points and lines indicate the propertiesof low metallicity, intensely star forming galaxies in the local Universe (blue crosses, Greis etal 2016) and at z ∼ . galaxy locus in the BPT diagram, and so may provide more local (and hence accessible)environments in which to test our understanding of these stellar populations.2.3. Indirect Constraints
At the highest redshifts, detailed spectroscopy is seldom possible. Nevertheless, there arestrong indications that the dominant stellar population differs from that seen either at z ∼ − z > z > istant galaxies and massive stars
3. The Key Role of Population Synthesis Models
When observing distant galaxies, it is very rare that individual stars, or even indi-vidual star forming regions can be resolved. At z = 1, the angular scale is ∼ † . This uses a library of theoretical stellar evolution mod-els, and combines these with synthetic stellar atmospheres. Importantly, it also includesprescriptions for binary system evolutionary effects omitted in many other SPS codes,including mass loss and (a simple prescription for) rotation. At Solar metallicity, theoutput model spectra for continuously star forming populations are comparable to thoseof other codes for a Salpeter-like initial mass function, while at early ages, different IMFsand low metallicities the different model sets diverge. The effects of binary evolutiontend to prolong the epoch over which very blue stars dominate the spectrum. The resultspresented here are from the v2.0 BPASS data release, supplemented by lower metallicitymodels which will be made available in BPASS v2.1, and for an IMF with a broken powerlaw slope, with indices -1.35 at M < . ⊙ and -2.35 for 0 . ⊙ < M < M max .Such models can confront the very high ionizing fluxes inferred for galaxies in the dis-tant Universe. In figure 3 we illustrate the dramatic difference in ionizing flux output be-tween different stellar populations and compare these to observational constraints derivedfrom distant sources. These are usually presented using the parameter ξ ion = ˙ N ion /L UV ,i.e. the ratio between the ionizing photon production rate and the rest-frame ultravioletcontinuum luminosity. While the latter can be directly observed in the distant Universe,the former must be inferred from indirect observations - primarily of the strong nebularemission lines generated from the galaxy in question, assuming that a fraction (1 − f esc )of the ionizing continuum is absorbed by the intertellar medium (see Stark et al. 2015;Bouwens et al. 2016, for discussion). While the metallicity of stellar populations in thedistant Universe are still poorly constrained, the inferred ionizing fluxes are significantlyhigher than those predicted for a canonical, continuously forming stellar population atnear-Solar metallicity (the canonical value of Kennicutt 1998, is indicated on the figure).Instead, populations with a higher upper mass cut-off (300 M ⊙ rather than the usually-assumed 100 M ⊙ ) and those which account for binary effects (e.g. through evolutionaryeffects or rotation) are favoured. The steady increase in ξ ion measurements with redshiftis more rapid than that expected from simple cosmic metallicity evolution for single starmodels, and may indicate that binary effects at low metallicity are becoming significant.Not only the ionizing photon flux but also the hardness of its spectrum can provide † see bpass.auckland.ac.nz E. R. Stanway z=7 LBG z=3 LAEs z=2 LAEsz=4 LBGsz=5 LBGs
Kennicutt (1998)
Figure 3.
The dependence of ionizing photon efficiency, ξ ion , , on metallicity, initial mass func-tion and single vs binary star evolution, assuming f esc = 0. At each metallicity we show themodel values for binary (large, crossed symbols) and single star (smaller symbols) stellar popu-lations, and for two IMFs (M max =300 M ⊙ , upper pair, and M max =100 M ⊙ , lower pair). Shadedregions indicate observational constraints ( ± σ ) on ξ ion , from high redshift galaxy populationsfrom Bouwens et al (2016, z ∼ − z = 2 LAEs, using the β dust correction for consistency with Bouwens et al), Nakajima et al (2016, z = 3 LAEs) andStark et al (2015, z ∼ insight into the massive stellar populations in distant galaxies. In figure 2 (right) weillustrate the shape as well as the strength of the Lyman continuum emission region forstellar models with the same star formation history (constant star formation over a 30 Myrinterval) but with different metallicities and including or omitting the effect of stellarbinaries. The shape of the ionizing spectrum is quite different, and will result in shifts inthe ratios of line emission, since these probe different energy ranges. In particular, thepresence of a hard (blue) ionizing spectrum in the 1-3 Rydberg range will shift galaxiesvertically in the ionization-sensitive BPT diagram (see also Xiao et al 2017, in theseproceedings). If the observed shift in the populations of both distant galaxies and theirlocal analogues are interpreted as a stellar population effect then models suggest that lowmetallicity models which incorporate binary evolution effects are required to reproducethem (although note that models with enhanced N/H ratios may also be appropriate,Shapley et al. 2015).
4. Implications
There is clear evidence that the rest-frame ultraviolet spectra of galaxies, and thenebular emission powered by reprocessing of the same photons, have evolved over cosmictime. This is, of course, to be expected - the volume-averaged mean metallicity of theIGM and the mean stellar population age both drop towards higher redshifts simply istant galaxies and massive stars z ∼ z ∼ z = 6 . α line emission, but also shows strong He II 1640˚Aemission, with constraints on the non-detection of O III 1665˚A and C III] 1909˚A lines(Sobral et al. 2015). Together with its photometric colours, these characteristics havebeen proposed as indicative of a metal-free (Population III) stellar population, while al-ternative interpretations include a primordial Direct Collapse Black Hole (Pallottini et al.2015). Either would be somewhat surprising, given the source’s luminosity and redshift.As Bowler et al. (2016) discuss, stellar evolution models can go some way towards ad-dressing this question, but are unable to resolve the issue completely and the interpre-tation of CR7 will remain ambiguous until further observations and improved modellingof all possible scenarios are undertaken.While this is an isolated case, and the constraints placed on massive stellar modelsby galaxies (rather than the reverse) are still weak, JWST will revolutionise this field.The NIRSPEC instrument will enable deep spectroscopy of hundreds (if not thousands)of star-forming galaxies at 3 < z <
8, from the rest-frame ultraviolet through to therest-optical (see e.g. Giardino et al. 2016). As a result, it is likely to produce both moreanomalous examples and much tighter constraints on the metal abundance and interstel-lar medium properties in galaxies in the distant Universe. Direct measurements of strongline ratios and probes of the stellar and interstellar absorption lines (as opposed to juststrong line emission) should be possible on both individual and stacked sources. These E. R. Stanwaywill allow direct comparison with stellar population synthesis models, and test whetherthese recover the observed parameter spaces. Whether our understanding of the massivestar population and its behaviour is sufficient to confront the wealth of observational dataexpected in the next few years remains to be seen. It will certainly be tested. Evidenceto date is that both fields can learn from and be enriched by this synergy and it is myhope that they will continue to do so.