Interpreting high [O III]/Hbeta ratios with maturing starbursts
E. R. Stanway, J. J. Eldridge, S. M. L. Greis, L. J. M. Davies, S. M. Wilkins, M. N. Bremer
aa r X i v : . [ a s t r o - ph . GA ] A ug Mon. Not. R. Astron. Soc. , 1–8 (2014) Printed 15 October 2018 (MN L A TEX style file v2.2)
Interpreting high [O
III ]/H β ratios with maturingstarbursts Elizabeth R. Stanway ⋆ , John J. Eldridge , Stephanie M. L. Greis ,Luke J. M. Davies , Stephen M. Wilkins , Malcolm N. Bremer Department of Physics, University of Warwick, Gibbet Hill Road, Coventry, CV4 7AL, UK Department of Physics, University of Auckland, Private Bag 92019, Auckland, New Zealand ICRAR, The University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia Astronomy Centre, Department of Physics and Astronomy, University of Sussex, Brighton, BN1 9QH, U.K H. H. Wills Physics Laboratory, University of Bristol, Tyndall Avenue, Bristol, BS8 1TL, UK
Accepted 2014 August 15. Received 2014 August 05; in original form 2014 May 28
ABSTRACT
Star forming galaxies at high redshift show ubiquitously high ionization parameters,as measured by the ratio of optical emission lines. We demonstrate that local ( z < . III ]/H β = 3 . +0 . − . - comparable to all but the highest ratios seen in star forminggalaxies at z ∼ −
4. We argue that the stellar population synthesis code BPASS canexplain the high ionization parameters required through the ageing of rapidly formedstar populations, without invoking any AGN contribution. Binary stellar evolutionpathways prolong the age interval over which a starburst is likely to show elevated lineratios, relative to those predicted by single stellar evolution codes. As a result, modelgalaxies at near-Solar metallicities and with ages of up to ∼
100 Myr after a starbursttypically have a line ratio [O
III ]/H β ∼
3, consistent with those seen in Lyman breakgalaxies and local sources with similar star formation densities. This emphasises theimportance of including binary evolution pathways when simulating the nebular lineemission of young or bursty stellar populations.
Key words: galaxies: evolution – galaxies: high redshift – galaxies: star formation
Understanding the sites of star formation in the distant Uni-verse is key to developing our picture of the early stages ofgalaxy evolution. The low mass, low metallicity, intenselystar forming galaxies observed in deep field surveys are thebuilding blocks which form the more massive systems wecurrently inhabit and observe evolve. They are also the mostlikely source of the energetic photons that ionized the inter-galactic medium (IGM) at early times (e.g. Bunker et al.2010), creating the conditions which persist to the currentday. The source and spectrum of those ionizing photons arekey parameters in cosmological simulations, affecting theprocess by which small regions of ionized Hydrogen sur-rounding the first galaxies grow and eventually overlap.However, spectroscopy of such distant sources pushesthe technical limits of existing spectrographs, and is oftenimpossible. Only the most highly lensed galaxies, or extremeexamples such as submillimetre galaxies or quasar hosts,are sufficiently luminous to measure optical emission lines ⋆ E-mail: [email protected] at z >
5, where the rest-UV and optical wavelength rangeshave been shifted into the near-infrared. Nonetheless, the fit-ting of spectral energy distributions (SEDs) across a broadredshift range ( z ∼ −
8) has strongly suggested that the ma-jority of high redshift star forming galaxies may show strongoptical line emission, which contributes significantly to theirobserved flux at 3-5 microns (de Barros, Schaerer & Stark2014; Stark et al. 2013; Gonz´alez et al. 2012). Local galax-ies selected to have similar ultraviolet emission densitiesare also confirmed to have very prominent emission lines(Heckman et al. 2005; Stanway & Davies 2014).At slightly lower redshifts, z ∼ −
4, direct measure-ments of the rest-optical emission spectrum become possi-ble. Work by Holden et al. (2014) compiled K -band spec-troscopy on a sample of 67 z ∼ . III β ratios are ubiquitous in thehigh redshift population. More recently still, Steidel et al.(2014) compiled a large sample of 179 2 . < z < . c (cid:13) E. R. Stanway et al. requiring ionization parameters orders of magnitudes higherthan those typically seen in local galaxies (Dopita et al.2000; Kewley et al. 2013, 2001) or invoking an otherwise un-realistic high oxygen abundance (Contini 2014).In this letter we present measurements of the line ratiosdetermined in a sample of local galaxies which are selectedto match the distant population in photometric properties(section 2), and consider possible interpretations in the lightof models from the Binary Population and Spectral Synthe-sis (BPASS) models (section 3). In section 4 we discuss im-plications for our understanding of stellar populations in thedistant Universe, before presenting our conclusions. z >
III ] 5007˚A line to the H β z >
1) thanthe redder H α line region. Samples of distant star-forminggalaxies with rest-frame optical spectroscopy have remainedsmall nethertheless, largely due to the challenging nature ofthe observations which have required observations of indi-vidual targets from the ground or low resolution grism spec-tra from the Hubble Space Telescope (e.g. Xia et al. 2012).Lensed targets have been the most studied, for example inHainline et al. (2009) who used [O
III ]/H β and other lineratios to measure high ionization parameters in three lensedgalaxies at z ∼ − . z = 2 . III ]/H β ratio of 4.12 ± z ∼ K -band spectroscopy on a sample of 67 z ∼ III ]/H β = 4 . +0 . − . for the sample. In figure 1 we replotthe high [O III ]/H β line ratios measured by Holden et al. asa function of mass, together with the distribution of ratiosobserved in low redshift galaxies (Brinchmann et al. 2004).Distant galaxies lie well above the line ratios seen in localstar forming galaxies at the same mass. The hard ionizingradiation field required to explain these observations isvery difficult to reproduce with normal stellar populations(Kewley et al. 2013; Brinchmann, Pettini, & Charlot 2008)at the metallicities seen at z ∼ − ⊙ ,e.g. Richard et al. 2011; Hainline et al. 2009; Erb et al.2006; Pettini et al. 2001). While the vast majority of galaxies in the local Universediffer in size, star formation rate and character from thoseat high redshift, it is nonetheless possible to identify lo-cal sources which appear to match the distant popula-tion in their continuum properties. In our recent paper(Stanway & Davies 2014), we identified a pilot sample ofsuch analogues, lying at z ∼ . − .
25. Galaxies were iden-tified based on their ultraviolet luminosity and colour, andto ensure that the resulting sample matched the high spe-cific star formation densities observed at z ∼
5. Given thathigh line ratios appear ubiquitous at z >
1, we would expectthese to share that characteristic.Our sample were also selected to have spectroscopy fromthe Sloan Digital Sky Survey (SDSS), precluding the pres-ence of strong AGN activity and confirming their redshift.In figure 1 we present the [O
III ]/H β line ratios measuredin our local analogue sample, as a function of stellar mass.Line ratios are measured at high signal to noise in SDSSspectroscopy for these relatively bright sources - we use linefluxes calculated by the SDSS pipeline, but have checkedthat these agree to within 0.02 dex with the independentMPA-JPU DR7 analysis . Stellar masses are also taken fromthe MPA-JPU database. These are derived from simultane-ous fitting of age, dust, star formation history and stellarmass to the galaxies’ spectral energy distribution. Thus theyare uncertain at the ∼ . III ]/H β = 3 . +0 . − . , where the quoted uncertainty givesthe inter-quartile range, consistent with that quoted byHolden et al. (2014) for their z ∼ − The Binary Population and Spectral Synthesis (BPASS)models are a set of galaxy population synthesis mod-els which were developed to address the effects of mas-sive stars on the spectral energy distributions of galaxies(Eldridge & Stanway 2012, 2009; Eldridge, Izzard, & Tout2008). Given a young stellar population, for example inthe aftermath of a major star formation episode, the op-tical spectrum of galaxies is dominated by hot and massive see http://bpass.org.uk/ c (cid:13) , 1–8 nterpreting high [O III ]/H β ratios (Mass)-1.0-0.50.00.51.01.5 l og ( O III / H b e t a ) (Mass)-1.0-0.50.00.51.01.5 l og ( O III / H b e t a ) Figure 1.
The distribution of OIII /H β ratios seen in galaxies observed by the SDSS (Brinchmann et al 2004, greyscale). Overplottedon the left are the ratios seen at z ∼ − ∼ stars which have not yet reached the ends of their lifespan.However, as the population ages, the population averagedcolour, temperature and SED are all strongly influenced bythe evolutionary state of the remaining massive stars. Theprocesses of angular momentum transfer, mass loss or massgain due to a binary companion all modify this evolutionarystate, allowing evolved secondary stars to extend the highly-luminous phase, and boosting the population of rapidly ro-tating, hydrogen-depleted, Wolf-Rayet stars.The BPASS code tracks the evolution of stellar pop-ulations, sampled from an initial mass function and rangeof binary system properties, and creates a composite stellarspectrum at a given age. Binary evolution is treated explic-itly for initial stellar masses > = 5 M ⊙ , and empirical termsare included for the evolution of rotationally-mixed, quasi-homogeneous stars. For comparison, an equivalent popula-tion of single stars is also permitted to evolve. The radia-tive transfer of the stellar emission from both populations,through a dust and gas screen within the source galaxy, isthen modelled using the radiative tranfer code CLOUDY (Ferland et al. 1998) to assess the contribution of nebularcontinuum and line emission. The ionization parameter ofthe local radiation field is defined as a ratio of the num-ber of ionizing photons to the local gas density. It is thusnot a directly tunable parameter, but rather constructedthrough the combination of appropriate stellar atmospheremodels and a choice of gas distribution. The assumed totalHydrogen gas density of our baseline model set is 10 cm − ,distributed in a sphere around the stellar population. Thisis a fairly typical gas density for extragalactic star formingH II regions, although we note that these range over severalorders of magnitude in density (see e.g. Hunt & Hirashita2009). We explore effects of varying the gas density rela-tive to our baseline models in section 3.3. The evolution ofan instanteous, rapid burst of star formation and of a conti-nous moderate (1 M ⊙ yr − ) star formation rate are modelledseparately. In this letter we use nebular emission line fluxpredictions determined as part of the current (v1.0) BPASSmodel data release. III ]/H β with age and metallicity The time evolution of the [O
III ]/H β ratio, at a fixed hydro-gen gas density, is significantly affected by the introductionof binary stellar evolution pathways, as shown in figure 2.In both single and binary population models, a continuousstar formation rate leads to a very stable [O III ]/H β ratio,which does not vary significantly after the initial few Myrs,since the line flux is always dominated by the youngest stars.This ratio can reproduce the observed high redshift (and lo-cal analogue data) at sub-Solar metallicities, for either singleor binary populations, but at Solar metallicity, the predictedline ratio falls well below the measured values. Continuousstar formation, observed at late times, may also be consis-tent with the bulk of observed H β emission line equivalentwidths (as shown in figure 3), but struggles to reproducesome of the lower observed values.By contrast, the aging of a rapid burst leads to line ra-tios that are high at early ages, as the massive stars formedin the initial burst evolve off the main sequence and even-tually undergo supernovae. In single star population mod-els, these intervals of high line ratios ([O III ]/H β>
1) arebrief, lasting no more than ∼
10 Myr and occuring primar-ily at low metallities (5-40% of Solar). Such young stars arelikely dust embedded and heavily extincted, making this anunattractive interpretation.However, the evolution of a binary stellar population isvery different. While the initial phase of high [O
III ]/H β ra-tios is still seen, it is extended over a much longer epoch bythe formation first of lower mass Wolf-Rayet stars than arepossible in a single star population, and then longer-lived,hot, helium stars. A binary stellar population will exceeda 1:1 line ratio for up to 100 Myr after its initial formationepoch (roughly the lifetime of the minimum mass consideredfor binary stars in BPASS), and at much higher metallici-ties ( ∼ ⊙ , the binary population reaches [O III ]/H β lineratios > c (cid:13) , 1–8 E. R. Stanway et al. (Age in years)-1.0-0.50.00.51.01.5 l og ( [ O III] / H β ) single - instantsingle - continuous 0.020 0.008 0.001 6.0 6.5 7.0 7.5 8.0 8.5log (Age in years)-1.0-0.50.00.51.01.5 l og ( [ O III] / H β ) binary - instant binary - continuous 0.020 0.008 0.0016.0 6.5 7.0 7.5 8.0 8.5-0.20.00.20.40.60.81.0 6.0 6.5 7.0 7.5 8.0 8.5-0.20.00.20.40.60.81.0 Figure 2.
The distribution of [O
III ]/H β ratios predicted by BPASS models as a function of stellar population age for single star (left)and binary (right) population stellar evolution pathways. Upper panels show the ageing of an instantaneous burst. Lower panels showthe near-constant ratios seen in populations with stable star formation rates. Tracks are shown at three different metallicities. During this interval, as figure 3 shows, a binary pop-ulation model also generates Balmer line luminosities andequivalent widths consistent with those seen in both thehigh redshift sample and the low redshift analogue popu-lation presented here. The single star instantaneous burstmodel, by contrast, cannot reproduce the line ratios with-out overproducing H β line flux. Similarly, as noted above,the continuous star formation models (with or without bi-naries) struggle to reproduce the weakest H β lines whilesimultaneously maintaining the strong line ratios. For bothinstantaneous and continuous star forming models, the low-est line equivalent widths (and luminosities) are generatedat late times, as the continuum contribution from older un-derlying stars builds up. III ]/H β with gas density While the emission line strengths from star forming regionsare a stong function of irradiating spectrum, they also de-pend on the geometry and density of the emitting gas. Self-shielding and collisional excitation or de-excitation of ionscan alter the transition probabilities in either very sparseor very dense gas. As mentioned in section 3.1, the base-line BPASS model set uses a total Hydrogen gas density of10 cm − , distributed in a sphere around the stellar popu-lation to model the radiative transfer and nebular emission.While this is a reasonable gas density for extragalactic starforming H II regions (see, for example, Hunt & Hirashita2009), it is possible that a difference in the typical gas den-sity, rather than in the dominant irradiating spectrum, couldbe responsible for the line properties of the high redshiftstarburst population and their local analogues.In figures 4 we consider the effect of total hydrogen gasdensity on the line ratios as a function of age, for the instan-taneous, Z=0.08 starburst models that best fit the data atour baseline gas density (see section 3.2).The effect of changing the assumed density of the illu- minated gas has a negligible effect on the lifetime of regionswith high line ratios for single star populations - at all den-sities, the high line ratio epoch ends within 10 Myr of theonset of star formation. The effect of gas density is, however,seen in the strength of the initial [O III ]/H β line ratios mea-sured. The line ratio at a given age decreases systematicallywith decreasing gas density, except at the highest densitiesconsidered, 10 cm − - equivalent to the upper end of theHII region distribution.By contrast, when binary stellar populations are con-sidered, both the strength of the ratio and the durationover which it remains elevated are functions of the as-sumed Hydrogen gas density. At low densities, < − , the[O III ]/H β ratio reached by the population increases withgas density, as does the lifetime over which it remains at thelevels seen in the distant population. At higher densities, > − , the [O III ]/H β ratio remains very nearly constantat [O III ]/H β ∼
5, but the epoch over which this level isreached becomes shorter with increasing gas density. At adensity of 10 cm − , the lifetime of strong line emission iscomparable to that seen in the single star populations.Comparison with figure 1 suggests that moderate gasdensities, 10 − − cm − are required to reproduce the highline ratios seen in the distant population and their local ana-logues. The extended lifetime of enhanced [O III ]/H β emis-sion in binary populations at gas densities ∼ −
100 cm − and at moderate metallicities suggest that these propertiesmay be consistent with the distant population. We note thatthis is not necessarily the scenario with the highest ioniza-tion parameter (which would correspond to the lowest den-sity for a given stellar input spectrum) but is comparable tothe densities seen in star forming regions. c (cid:13) , 1–8 nterpreting high [O III ]/H β ratios (EW H β )-1.0-0.50.00.51.01.5 l og ( [ O III] / H β ) (EW H β )-1.0-0.50.00.51.01.5 l og ( [ O III] / H β ) Figure 3.
The distribution of [O
III ]/H β ratios predicted by BPASS models as a function of rest-frame H β recombination line equivalentwidth for single star (left) and binary (right) population stellar evolution pathways. Labels and linestyles are as in figure 2. Age intervalsof log(age)=0.1 are marked, and age increases to the left on each track in this parameter space. Pale grey points indicate high redshiftdata from Schenker et al. (2013) and Holden et al. (2014). Red crosses indicate our local analogue sample, the error bars for which aresmaller than the points. (Age in years)-1.0-0.50.00.51.01.5 l og ( [ O III] / H β ) -4-3-2-1 01234log(gas density) Single, Z=0.08, instantaneous burst 6.0 6.5 7.0 7.5 8.0 8.5log (Age in years)-1.0-0.50.00.51.01.5 l og ( [ O III] / H β ) -4-3-2-1 01234log(gas density) Binary, Z=0.08, instantaneous burst
Figure 4.
The distribution of [O
III ]/H β ratios predicted by BPASS models for single star (left) and binary (right) population stellarevolution pathways at Z = 0 .
08 (40% Solar), as a function of gas density.
The effect of strongly ionizing spectra (as manifest in the[O
III ]/H β ratio) has traditionally been attributed to anAGN contribution and is commonly seen in local Seyfert-type galaxies (Ferland & Netzer 1983). An alternate expla-nation could conceivably be a very high oxygen abundancerelative to hydrogen. To rule out this later possibility, weplot both observed examples of [O III ]/H β ratios and pre-dictions for binary stellar populations in the BPASS modelsagainst the largely abundance-independent [O II ]/[O III ] ra-tio (see e.g. Ferland & Netzer 1983) in figures 5 and 6. Abinary star formation model at Z = 0 .
08 and a gas den-sity of 10 cm − (our baseline model) provides a remarkablygood fit to the distribution of line ratios seen in both highand low redshift starburst galaxies, over an extended periodof order 100 Myrs. In fact, the line ratios generated by bi-nary stellar populations of ages <
100 Myr shows only mild dependence on the surrounding gas density. The ratios dohowever show a strong dependence on binary as opposed tosingle star evolution pathways, with the latter typically pro-ducing lower line ratios and showing a stronger gas densitydependence.We find that high [O
III ]/H β ratios fall naturally outof a self-consistent treatment of binary evolution in an age-ing starbust stellar population, without invoking AGN emis-sion or an unusual IMF (Masters et al. 2014). The modeststar formation rates and metallicities required to create suchspectra do not require fine tuning of the conditions, but arewell matched to the typical properties of both z ∼ − z = 0 galaxy population. Given the discretesampling in metallicity and age of the BPASS models, and c (cid:13) , 1–8 E. R. Stanway et al. the simple prescription for the dust and gas screen, it isuncertain whether the effect of binaries is sufficient to re-produce the highest observed line ratios, [O
III ]/H β ∼
10 inthe z = 2 − ∼ ⊙ .Holden et al. (2014) argued against a bursty starburstmodel for z = 2 − β , dominated bythe youngest stellar population, and the rest-frame ultravio-let continuum. However, the time-scale for establishment ofultraviolet emission is of order ∼ −
30 Myr. This wouldbe problematic for a single stellar population, but is signif-icantly shorter than the time-scales for elevated [O
III ]/H β ratios in binary populations, which would show an estab-lished UV continuum. As noted above, we have studied theBPASS prediction for Balmer line luminosity and equivalentwidth and find it consistent with the observed high redshiftdata.So is young, bursty star formation a good model fordistant galaxies with high specific star formation rates? Tosome extent, any answer depends on definitions. Degenera-cies in the possible interpretation of photometric data makethe ‘age’ of a galaxy difficult to constrain. Given the ultravi-olet selection of Lyman break galaxies (and analogues), theyinevitably have a component of relatively young ( < z ∼ − z = 3 galaxiesmay have undergone a very rapid early burst on time-scales of 50 −
100 Myr, before continuing to form stars ata significantly lower rate. In such a paradigm, the evolu-tion of that early burst may dominate the nebular con-tinuum at ages >
300 Myr. While the SEDs of LBGs at1 . < z < . z = 3 sources yields a younger median age ( ∼
320 Myr,Shapley et al. 2001).A recent study at intermediate redshifts, 2 . < z < . T eff ∼ ,
000 K) and the inclusion of binary evolution path-ways and stellar rotation is necessary to generate plausibleionizing spectra. We note that Steidel et al. also identifiedthe “extreme green pea” sample of Jaskot & Oey (2013) ashaving comparable line ratios to their z ∼ . z ∼ -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4log ( [O II]/[O III] )-0.50.00.51.0 l og ( [ O III] / H β ) binary - instant 0.020 0.008 0.001 Figure 5.
The [O II ]/[O III ] ratios seen in the data, and predictedby BPASS models for binary populations at ages 10 -10 . yrs fol-lowing an instantaneous burst. A smooth third order polynomialhas been fit through the predictions at discrete model timestepsat each metallicity. Grey points are z = 2 − It is likely that the hot, low metallicity star-burst withassociated binary evolution presented here is not a uniqueexplanation. As discussed in section 3.2, models derived forcontinuous star formation at low metallicity also recoverhigh emission line ratios for a large fraction of their lifetime,although they show some tendency to overpredict the H β equivalent width. Some fraction of ongoing star formation,following an initial starburst would also likely recover similarline ratios. Any stellar population synthesis code necessarilyexplores a limited parameter set, exploring discrete stellarmetallicities, interstellar gas densities, geometries and ages.By contrast real galaxies are a composite of many star form-ing regions, each of different age and with different physicalconditions. At best, a theoretical model can only be an ap-proximate match.The high line ratios in post-starburst conditions maywell be diminished by combination with other stellar pop-ulations. It is also challenging to entirely rule out a weakAGN contribution to the emission lines, although we notethat X-ray stacking analyses have constrained the moder-ate AGN fraction in the distant population to be < z ∼ Our main conclusions can be summarised as follows:(i) We measure the ionization-sensitive [O
III ]/H β ratioin ultraviolet-luminous local galaxies selected as z ∼ c (cid:13) , 1–8 nterpreting high [O III ]/H β ratios -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4log ( [O II]/[O III] )-0.20.00.20.40.60.81.0 l og ( [ O III] / H β ) -4-3-2-1 01234log(density):Model population:singlebinary Z=0.08, instantaneous burst
Figure 6.
As figure 5, but now varying assumed gas density ata fixed metallicity ( Z = 0 . -10 . yrs after an instantaneousburst. For binary populations (solid lines), gas density has onlyslight effect on the line ratios, such that the models occupy anarrow locus in parameter space. The single star models spana broader range, and show a more pronounced evolution in lineratios with gas density. man break analogues. We find they lie well above the localaverage ratios for their SED-derived masses, with a median[O III ]/H β = 3 . +0 . − . , similar to those seen in high redshiftgalaxy populations.(ii) We consider the line ratios derived from the BinaryPopulation and Spectral Synthesis (BPASS) models. We de-termine that they can reproduce high [O III ]/H β ∼ ∼ −
300 Myrs, at modest(0.2-1.0 Z ⊙ ) metallicities. They also accurately predict thebehaviour of the [O II ]/[O III ] line ratio.(iii) The density of the illuminated nebular gas appears tohave only small effects on the predicted line ratios in binarystellar evolution models at moderate metallicities (0.4 Z ⊙ ),and models can reproduce the data at densities seen in ex-tragalactic star forming regions. Single stellar models arerather more sensitive to gas density, with the predicted lineratios decreasing strongly with density at a given age.(iv) While continuous star formation can generate simi-larly high line ratios at moderate metallicities, they struggleto reproduce the measured H β luminosities and equivalentwidths.(v) We conclude that including binary population effectsmay be important when modelling stellar populations at <
500 Myrs, where line ratios depend sensitively on the evo-lution of the most massive stars.Ideally, the comparison of multiple emission lines, or bet-ter still modelling of the full near-ultraviolet/optical spec-trum, will be required to further constrain the star forma-tion history and properties of distant galaxies. We plan toexplore those of the local analogue population further in aforthcoming paper, in the hopes of gaining further insightsinto possible explanations of their ionizing spectra.
ACKNOWLEDGMENTS
Cloudy , last described by (Ferland et al.2013). We also thank the anonymous referee for their input.
REFERENCES
Brinchmann J., Pettini M., Charlot S., 2008, MNRAS, 385,769Brinchmann J., Charlot S., White S. D. M., Tremonti C.,Kauffmann G., Heckman T., Brinkmann J., 2004, MN-RAS, 351, 1151Bunker A. J., et al., 2010, MNRAS, 409, 855Contini M., 2014, A&A, 564, A19de Barros S., Schaerer D., Stark D. P., 2014, A&A, 563,A81Dopita M. A., Kewley L. J., Heisler C. A., Sutherland R. S.,2000, ApJ, 542, 224Eldridge J. J., Stanway E. R., 2012, MNRAS, 419, 479Eldridge J. J., Izzard R. G., Tout C. A., 2008, MNRAS,384, 1109Eldridge J. J., Stanway E. R., 2009, MNRAS, 400, 1019Erb D. K., Shapley A. E., Pettini M., Steidel C. C., ReddyN. A., Adelberger K. L., 2006, ApJ, 644, 813Eyles L. P., Bunker A. J., Ellis R. S., Lacy M., StanwayE. R., Stark D. P., Chiu K., 2007, MNRAS, 374, 910Eyles L. P., Bunker A. J., Stanway E. R., Lacy M., EllisR. S., Doherty M., 2005, MNRAS, 364, 443Ferland G. J., et al., 2013, RMxAA, 49, 137Ferland G. J., Korista K. T., Verner D. A., Ferguson J. W.,Kingdon J. B., Verner E. M., 1998, PASP, 110, 761Ferland G. J., Netzer H., 1983, ApJ, 264, 105Gonz´alez V., Bouwens R. J., Labb´e I., Illingworth G.,Oesch P., Franx M., Magee D., 2012, ApJ, 755, 148Hainline K. N., Shapley A. E., Kornei K. A., Pettini M.,Buckley-Geer E., Allam S. S., Tucker D. L., 2009, ApJ,701, 52Heckman T. M., et al., 2005, ApJ, 619, L35Holden B. P., et al., 2014, arXiv, arXiv:1401.5490Hunt L. K., Hirashita H., 2009, A&A, 507, 1327Jaskot A. E., Oey M. S., 2013, ApJ, 766, 91Kewley L. J., Dopita M. A., Leitherer C., Dav´e R., YuanT., Allen M., Groves B., Sutherland R., 2013, ApJ, 774,100Kewley L. J., Dopita M. A., Sutherland R. S., Heisler C. A.,Trevena J., 2001, ApJ, 556, 121Laird E. S., Nandra K., Hobbs A., Steidel C. C., 2006,MNRAS, 373, 217Masters D., et al., 2014, ApJ, 785, 153 c (cid:13) , 1–8 E. R. Stanway et al.
Nakajima K., Ouchi M., Shimasaku K., Hashimoto T., OnoY., Lee J. C., 2013, ApJ, 769, 3Pettini M., Shapley A. E., Steidel C. C., Cuby J.-G., Dick-inson M., Moorwood A. F. M., Adelberger K. L., Gi-avalisco M., 2001, ApJ, 554, 981Reddy N. A., Pettini M., Steidel C. C., Shapley A. E., ErbD. K., Law D. R., 2012, ApJ, 754, 25Richard J., Jones T., Ellis R., Stark D. P., Livermore R.,Swinbank M., 2011, MNRAS, 413, 643Schenker M. A., Ellis R. S., Konidaris N. P., Stark D. P.,2013, ApJ, 777, 67Shapley A. E., Steidel C. C., Erb D. K., Reddy N. A.,Adelberger K. L., Pettini M., Barmby P., Huang J., 2005,ApJ, 626, 698Shapley A. E., Steidel C. C., Adelberger K. L., DickinsonM., Giavalisco M., Pettini M., 2001, ApJ, 562, 95Stanway E. R., Davies L. J. M., 2014, MNRAS, 439, 2474Stark D. P., Schenker M. A., Ellis R., Robertson B.,McLure R., Dunlop J., 2013, ApJ, 763, 129Steidel C. C., et al., 2014, arXiv, arXiv:1405.5473Xia L., et al., 2012, AJ, 144, 28This paper has been typeset from a TEX/ L A TEX file preparedby the author. c (cid:13)000