FirstLight IV: Diversity in sub-L_* galaxies at cosmic dawn
Daniel Ceverino, Michaela Hirschmann, Ralf Klessen, Simon Glover, Stephane Charlot, Anna Feltre
MMNRAS , 1–9 (2021) Preprint 25 February 2021 Compiled using MNRAS L A TEX style file v3.0
FirstLight IV: Diversity in sub-L ∗ galaxies at cosmic dawn Daniel Ceverino, , , ★ Michaela Hirschmann, Ralf S. Klessen, , Simon C. O. Glover, Stéphane Charlot , Anna Feltre Departamento de Fisica Teorica, Modulo 8, Facultad de Ciencias, Universidad Autonoma de Madrid, 28049 Madrid, Spain CIAFF, Facultad de Ciencias, Universidad Autonoma de Madrid, 28049 Madrid, Spain Universität Heidelberg, Zentrum für Astronomie, Institut für Theoretische Astrophysik, Albert-Ueberle-Str. 2, 69120 Heidelberg, Germany DARK, Niels Bohr Institute, University of Copenhagen, Lyngbyvej 2, 2100, Copenhagen Ø , Denmark Universität Heidelberg, Interdisziplinäres Zentrum für Wissenschaftliches Rechnen, INF 205, 69120, Heidelberg, Germany Sorbonne Université, CNRS, UMR7095, Institut d’Astrophysique de Paris, F-75014, Paris, France INAF - Osservatorio di Astrofisica e Scienza dello Spazio di Bologna, Via P. Gobetti 93/3, 40129 Bologna, Italy
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
Using a large sample of sub-L ∗ galaxies, with similar UV magnitudes, M UV (cid:39) −
19 at 𝑧 (cid:39) ∼
40 variation in the specific star-formation rate(sSFR). This drives a ∼ 𝜆 𝛼 line ratio at a fixed sSFR. [OIII]-bright emitters have higherionization parameters and/or higher metallicities than H 𝛼 -bright galaxies. According to thesurface brightness maps in both [OIII] and H 𝛼 , [OIII]-bright emitters are more compact thanH 𝛼 -bright galaxies. H 𝛼 luminosity is higher than [OIII] if star formation is distributed overextended regions. OIII dominates if it is concentrated in compact clumps. In both cases, theH 𝛼 -emitting gas is significantly more extended than [OIII]. Key words: galaxies: evolution – galaxies: formation – galaxies: high-redshift
The current census of galaxies at the end of the reionization epoch, 𝑧 (cid:39)
6, has yielded a large number of sub-L ∗ galaxies with relativelyfaint rest-frame UV magnitudes, M 𝑈𝑉 > −
21, (Stark 2016). Theyrepresent the majority of galaxies at these high redshifts and theyprovide the bulk of the photons responsible for the reionization ofthe Universe (Robertson et al. 2013; Naidu et al. 2020). However,very little is known about their basic properties: mass, star formation(SF), size, or physical conditions in the interstellar medium (ISM).In particular, it is important to constrain the diversity of this pop-ulation. Understanding the origin of this diversity will allow us toobtain population-averaged properties from representative samplesthat will be observed with the
James Webb Space Telescope ( JWST )and the next generation ground-based facilities.Available observations of high-redshift galaxies show hints ofbright emission in rest-frame visible lines, such as [OIII] 𝜆 𝛽 or H 𝛼 (Chary et al. 2005; Stark et al. 2013; Faisst et al. 2016; De Bar-ros et al. 2019). Some galaxies exhibit extreme nebular conditions,with very high equivalent widths, EW([OIII] 𝜆𝜆 , + 𝐻𝛽 ) ≥ ★ E-mail: [email protected]
600 Å (Labbé et al. 2013). However, it is not clear what drives thesehigh values or whether these galaxies are representative examplesof the overall galaxy population at these early times.
JWST will give us the opportunity to unveil the diversity ofgalaxies at reionization. In particular, NIRSpec spectroscopy willdisentangle the individual lines in sub-L ∗ galaxies at 𝑧 ≥
6, partic-ularly in the wavelength ranges around [OIII] 𝜆 𝛼 , two ofthe brightest lines in galaxy spectra at high z (Stark 2016). Its IFUcapability will also constrain the spatial extent of these emission re-gions. These observations will tell us whether galaxies with similarrest-frame UV properties have also similar rest-frame visible lines.They will unveil the diversity in line strengths. This can give usclues about the main drivers of the observed variations. Do galaxieswith different [OIII]/H 𝛼 ratios differ in any global property or isit only related to the local conditions within HII regions? Are theproperties of [OIII]-bright galaxies different from the properties of[OIII]-faint emitters? JWST may tell us whether there are differentpopulations of line emitters at high z.Cosmological simulations of galaxy formation can give us afirst insight into this galaxy diversity at cosmic dawn (O’Shea et al.2015; Ceverino et al. 2017; Katz et al. 2019; Ma et al. 2018; Pallottiniet al. 2019). However, this requires large and unbiased samples ofsimulated galaxies. The FirstLight database is a mass-limited sam-ple of simulated galaxies at cosmic dawn (Ceverino et al. 2017). It © a r X i v : . [ a s t r o - ph . GA ] F e b Ceverino et al. successfully encompasses a large diversity of complex SF histories,characterised by frequent SF bursts of different strength and dura-tion (Ceverino et al. 2018). This translates into a large diversity ofspectral energy distributions (Ceverino et al. 2019), consistent withcurrent observations. The rest-frame UV magnitudes range fromM UV = −
12 to −
22 and the stellar masses range from 𝑀 ∗ = to10 . M (cid:12) , with a relatively large number of galaxies with similarUV magnitudes but different stellar masses.This paper uses the FirstLight database and it focuses on a nar-row range of UV magnitudes: M UV (cid:39) −
19 at 𝑧 (cid:39)
6. This allows usto study the galaxy diversity almost independently of the trends fromgalaxy scaling relations. We aim to understand the physical originof variations in the strength of two of the brightest rest-frame visiblelines, [OIII] 𝜆 𝛼 . According to the linefluxes of FirstLight galaxies described in Álvarez-Márquez et al.(2019), the median flux is about a few times 10 − erg s − cm − .The whole sample with M UV (cid:39) −
19 at 𝑧 (cid:39) 𝛼 lineratios (§3.2), and the spatial extent of this emission (§3.3). Finally,section §4 ends the paper with the summary and discussion. The parent sample is a complete mass-selected subsample fromthe FirstLight database of simulated galaxies, described fully inCeverino et al. (2017). The subsample consists of 290 halos witha maximum circular velocity ( 𝑉 max ) between 50 and 200 km s − ,selected at 𝑧 =
5. The halos cover a mass range between a few times10 and 10 M (cid:12) . This range excludes more massive and rare haloswith number densities lower than ∼ × − ( ℎ − Mpc ) − , as wellas small halos in which galaxy formation is extremely inefficient.From the parent sample, a complete luminosity-selected sam-ple was selected based on two criteria: a redshift around 𝑧 (cid:39) 𝑧 = . − .
5) and an absolute UV magnitude at 1500 Å of 𝑀 𝑈𝑉 (cid:39) −
19 (between −
19 and − . ∗ galaxies withobserved H band magnitudes between 27 and 28 at the end of thereionization epoch. On average, there are four snapshots per distinctgalaxy with a total number of 35 galaxies. Due to the burstinessof the SF histories, all snapshots can be considered as differentgalaxies without introducing any significant bias in the results. The simulations are performed with the ART code (Kravtsov et al.1997; Kravtsov 2003; Ceverino & Klypin 2009; Ceverino et al.2014, 2017), which accurately follows the evolution of a gravitating 𝑁 -body system and Eulerian gas dynamics using an adaptive mesh refinement (AMR) approach. Besides gravity and hydrodynamics,the code incorporates many of the astrophysical processes rele-vant for galaxy formation. These processes, represented via subgridphysical prescriptions, include gas cooling due to atomic hydrogenand helium, metals and molecular hydrogen, photoionization heat-ing by a constant cosmological UV background with partial self-shielding, star formation and feedback (thermal+kinetic+radiative),as described in Ceverino et al. (2017). The simulations track metalsreleased from SNe-Ia and from SNe-II, using supernovae yields thatapproximate the results from Woosley & Weaver (1995). These val-ues are given for gas cells and star particles as described in Kravtsov(2003). We assume that the effect of AGN feedback on these sub-L ∗ galaxies is very minor and it is not included.The target halos are initially selected using low-resolution 𝑁 -body only simulations of two cosmological boxes with sizes 10and 20 ℎ − Mpc, assuming WMAP5 cosmology with Ω m = . Ω b = . ℎ = . 𝜎 = .
82 (Komatsu et al. 2009). We se-lect all distinct halos with 𝑉 max at 𝑧 = 𝑉 cut / km s − ) = . ℎ − Mpc box and log ( 𝑉 cut / km s − ) = . ℎ − Mpc box. Initial conditions forthe selected halos with much higher resolution are then generatedusing a standard zoom-in technique (Klypin et al. 2011). The DMparticle mass resolution is 𝑚 DM = M (cid:12) . The minimum mass ofstar particles is 100 M (cid:12) . The maximum spatial resolution is alwaysbetween 8.7 and 17 proper pc (a comoving resolution of 109 pc after 𝑧 < Global galaxy properties, such as stellar mass, UV magnitude orslope ( 𝛽 ) are extracted from the FirstLight database. In summary,the continuum emission of the simulated galaxies is computed usingpublicly available tables from the Binary Population and SpectralSynthesis (BPASS) model (Eldridge et al. 2017) including nebularcontinuum. More details can be found in Ceverino et al. (2019).The emission-line maps are computed in two steps. First, auniform 6x6 kpc grid is laid at the face-on view of each galaxy.The 2D grid is centered with respect to the stellar distribution. Theface-on view is defined using the angular momentum of the coldgas. Each 100-pc-wide pixel stores the mass in stars younger than10 Myr, the average gas density and metallicity of the warm gas( 𝑇 < × K). The use of different pixel sizes from 50 to 300pc gives similar overall results because the typical cell size of thesimulation ( ∼
10 pc) is much smaller than the pixel size.The gas and star properties of the individual pixels are thenused to compute emission lines based on the methodology andemission line models for young stellar populations described inHirschmann et al. (2017, 2019). Specifically, we adopt the gridof nebular-emission models of star-forming galaxies computed byGutkin et al. (2016). These calculations combine the latest versionof the Bruzual & Charlot (2003) stellar population synthesis modelwith Cloudy (Ferland et al. 2013), following the method outlinedby Charlot & Longhetti (2001) for each element of a simulatedgalaxy. This implies that each pixel is composed of nebular emis-sion emerging from a steady population of ionization-bounded HIIregions. This is a reasonable assumption for modeling line emis-sion, since the majority of ionizing photons are released during thefirst 10 Myr of evolution of a single stellar population (Gutkin et al.2016). This is just two time-steps of SF in the simulation, in whichthe SFR is roughly constant in most cases.The emission line grid, table 1 in Hirschmann et al. (2017) andtable 3 in Gutkin et al. (2016), includes models in wide ranges of
MNRAS , 1–9 (2021)
L IV Figure 1.
Five examples of sub-L ∗ galaxies at 𝑧 (cid:39)
6. Each row shows projections of stars (right), gas (middle) surface density and the sSFR evolution (left) ofthe main galaxy progenitor during the previous ∼
400 Myr. The size is 10x10 kpc . The color bars range from 1 to 500 M (cid:12) pc − in 10 log-scaled ticks.MNRAS000
400 Myr. The size is 10x10 kpc . The color bars range from 1 to 500 M (cid:12) pc − in 10 log-scaled ticks.MNRAS000 , 1–9 (2021) Ceverino et al. )1.52.02.53.0 l o g ( E W _ O III / Å ) M UV -19, z
6 8.08.28.48.68.89.0 l o g ( M * / M ) )1.52.02.53.0 l o g ( E W _ O III / Å ) l o g ( S F R / ( M y r )) )1.52.02.53.0 l o g ( E W _ O III / Å ) l o g ( O III / H ) )1.52.02.53.0 l o g ( E W _ O III / Å ) l o g ( Z / Z ) )1.52.02.53.0 l o g ( E W _ O III / Å ) )1.52.02.53.0 l o g ( E W _ O III / Å ) l o g U Figure 2.
Equivalent width (EW) of the [OIII] line versus specific star formation rate. The points are coloured by stellar mass, star formation rate, [OIII]/H 𝛼 luminosity ratio, UV slope, SFR-weighted nebular metallicity, and SFR-weighted ionization parameter. The vertical lines mark the position of the star-formingmain sequence and its scatter (Ceverino et al. 2018). The EW increases with stellar mass and SFR but the scatter at fixed sSFR is correlated with the lineluminosity ratio and it is driven by the ionization parameter. interstellar (gas+dust) metallicities, ionization parameters, dust-to-metal mass ratios, HII-region densities and carbon-to-oxygen abun-dance ratios. With each galaxy pixel, we associate the SF emission-line model from the Gutkin grid (described above) with closestpixel-average values. We select the grid metallicity closest to theaverage metallicity in a given pixel. The ionization parameter, logU, is computed based on equation 1 in Hirschmann et al. (2017)using the instantaneous SFR and the filling factor = 𝑛 gas / 𝑛 H , where 𝑛 gas is the density in warm gas and 𝑛 H is the hydrogen density within a HII region. For the dust-to-metal mass ratio, the C/O ratioand the hydrogen density, we assume fixed values (0.3, solar C/Oand 𝑛 H =
100 cm − ), as these quantities are either not modelledor not resolved. We assume that the emission from old stellar pop-ulations and non-stellar sources is not significant in these activelystar-forming and young galaxies.The galaxy [OIII] and H 𝛼 luminosities given by this model arevery similar to the ones discussed by Ceverino et al. (2019), whoused the BPASS model (Eldridge et al. 2017; Xiao et al. 2018). MNRAS , 1–9 (2021)
L IV BPASS yielded a factor 0.8 lower luminosity than the ones used inthis paper. The [OIII]/H 𝛼 ratios agree to within 10%. This assures usthat the results reported in this paper are independent of a particularemission-line model. The sample of sub-L ∗ galaxies with similar UV luminosities andredshifts ( 𝑀 𝑈𝑉 (cid:39) −
19 at 𝑧 (cid:39)
6) show a large diversity. The medianvalues are 𝑀 ∗ = . M (cid:12) , SFR = . (cid:12) yr − and sSFR = − , but the sample ranges over 𝑀 ∗ = . − M (cid:12) andSFR = . − . (cid:12) yr − . Figure 1 shows five examples of thisdiversity. The top row shows a galaxy at the peak of an extremeburst of star formation, as shown by its sSFR history. Its value,sSFR =
30 Gyr − , is a factor 6 higher than the median value of thewhole sample. The burst is driven by a multiple merger. This boostsits UV luminosity in spite of its lower-than-average stellar mass, 𝑀 ∗ = M (cid:12) . The gas is distributed in a few dense and compactclumps that concentrate most of the star formation. The second rowdisplays another example of a merger-driven burst, but the peak ismore representative of a typical burst, sSFR =
16 Gyr − (Ceverinoet al. 2018). This example shows a compact gas distribution, due tothe final merger coalescence.The third row shows a typical, main-sequence star-forminggalaxy at 𝑧 (cid:39) = − . According to its sSFR his-tory, the galaxy is not in a maximum or minimum value, but some-where in between. The gas is significantly more extended than in theprevious cases and it shows smaller clumps distributed throughoutthe galaxy. The stellar mass, 𝑀 ∗ = × M (cid:12) , is higher than inthe previous starburst examples and the stellar distribution shows adense center. This is an indication of a mature population, formedin multiple bursts in the past 200-300 Myr. The fourth row illus-trates a galaxy that lies a factor of 2 below the SF main-sequence.Its gas distribution looks more diffuse and less clumpy than in theother cases. The fifth row provides an example of a quiescent, post-starburst galaxy (sSFR = . − ), 100 Myr after the last burstof star formation. Its stellar mass, 𝑀 ∗ = × M (cid:12) , is signifi-cantly higher than average and it is mostly concentrated in a densestellar center, with non-clumpy gas around it. In conclusion, thereis a factor 40 variation in sSFRs even within galaxies with simi-lar UV luminosities. This drives a large variety in stellar and gasdistributions.This galaxy diversity can also be seen in the strength of emis-sion lines. Figure 2 shows the equivalent width (EW) of the [OIII]line for all galaxies of the sample. The EW increases with sSFR,reaching EW (cid:39) (cid:39)
300 Å. Themost massive galaxies with the lowest SFR have the lowest EW ≤
100 Å (FL927 at the bottom row of Figure 1). This is a factor ∼ 𝑍 ) exhibit asimilar behaviour, driven by the diversity in stellar mass (Langanet al. 2020) and SFR.There is a significant, intrinsic scatter in the above relationbetween EW and sSFR. Therefore, there are other drivers of high )0.40.20.00.20.4 l o g ( O III / H ) M UV -19, z
6 1.41.21.00.80.60.4 l o g ( Z / Z ) )0.40.20.00.20.4 l o g ( O III / H ) l o g U Figure 3. [OIII]/H 𝛼 line luminosity ratio versus sSFR. The points arecoloured by SFR-weighted nebular metallicity (top) and ionization parame-ter (bottom). The vertical lines mark the position of the star-forming mainsequence and its scatter. At a fixed sSFR, high [OIII]/H 𝛼 ratios are drivenby a high metallicity and/or a high ionization parameter. EWs. At a fixed sSFR, galaxies with higher EW also have higher[OIII]/H 𝛼 luminosity ratio and higher SFR-weighted, ionizationparameters (Figure 2). This indicates that the nebular conditions,such as gas and SFR densities, drive this scatter. Even betweengalaxies with similar UV magnitudes, sSFR, and redshifts we expectdifferent ISM conditions. The FirstLight simulations indicate a diversity in the conditionsof star-forming regions at high z, such as variations in ionizationparameter. This translates into different [OIII]/H 𝛼 luminosity ra-tios even between galaxies with similar sSFR. For example, somegalaxies are [OIII]-bright, log([OIII]/H 𝛼 ) >
0, with [OIII] lumi-nosities up to a factor of 2 higher than H 𝛼 . They are analogous torare compact emission-line galaxies at 𝑧 ∼ 𝛽 emitters at 𝑧 ∼ − 𝛼 . There-fore, they constitute a significant fraction of the galaxy populationat the end of reionization.H 𝛼 -bright emitters, log([OIII]/H 𝛼 ) <
0, dominate the First-
MNRAS000
MNRAS000 , 1–9 (2021)
Ceverino et al. −4 −2 0 2 4x/kpc−4−3−2−101234 y / k p c FL846 Hα −4 −2 0 2 4x/kpc−4−3−2−101234 y / k p c FL921 Hα−4 −2 0 2 4x/kpc−4−3−2−101234 y / k p c FL846 OIII −4 −2 0 2 4x/kpc−4−3−2−101234 y / k p c FL921 OIII3 4 5 6 7−4 −2 0 2 4x/kpc−4−3−2−101234 y / k p c FL846 log U−3.4 −3.2 −3.0 −2.8 −2.6 −4 −2 0 2 4x/kpc−4−3−2−101234 y / k p c FL921 log U−3.4 −3.2 −3.0 −2.8 −2.6
Figure 4.
Maps of H 𝛼 (top), [OIII] luminosity (middle), and ionization parameter (bottom) for two galaxies with similar sSFR but different line ratios:log([OIII]/H 𝛼 ) = .
15 (left) and log([OIII]/H 𝛼 ) = − .
17 (right). Blue circles mark 2 times the half-light radius. The colorbar indicates the line luminosity inlog( 𝐿 (cid:12) ). Galaxies with high line ratios have systematically more compact SF regions. The distribution in H 𝛼 is significantly more extended than that in [OIII].MNRAS , 1–9 (2021) L IV Light sample. They have H 𝛼 luminosities up to a factor of 3 higherthan [OIII]. However, they are currently getting little attentionmostly because H 𝛼 becomes inaccessible for NIRSpec spectroscopyat 𝑧 > .
5. Longer wavelengths (MIRI) will be needed to unveil thispopulation at higher redshifts (Álvarez-Márquez et al. 2019) butthey are an important galaxy population at the end of reionization, 𝑧 < . 𝛼 ratio depends primarily on the ion-ization parameter and secondarily on nebular metallicity (Figure 3).In our sample of low-metallicity galaxies, an increase in metallicityleads to higher line ratios, in a way similar to the [OIII]/H 𝛽 ratio(Langan et al. 2020). This is due to the low values of metallicity,log(Z/Z (cid:12) ) < − 𝛽 -Z curve (Gutkin et al. 2016). The appendixshows the [OIII]/H 𝛽 ratio for comparison. However, metallicity alsodepends on stellar mass, and therefore, in this UV-selected sample,galaxies with higher-than-average mass and metallicity have lower-than-average sSFR but also relatively high line ratios. The oppositeis true for the metal-poor, less-massive starbursts. Consequently,the trend between line ratio and sSFR is largely due to metallic-ity effects. Around the main sequence, more metal-rich galaxieshave higher line ratios than more metal-poor galaxies with the samesSFR.The ionization parameter is the main driver of the scatter ofthe [OIII]/H 𝛼 ratio at a fixed sSFR. [OIII]-bright galaxies havehigher ionization parameters than H 𝛼 -bright galaxies. Interestingly,these [OIII] emitters do not have the highest sSFR. They are notextreme starbursts and can be found among normal main-sequencegalaxies. These conditions of high ionization can occur in galaxieswith averaged sSFR at cosmic dawn. Even some quiescent galaxieswith lower-than-averaged sSFR can host these extreme ionizationconditions. This indicates that the line ratios mostly depend on thelocal conditions within star-forming regions, characterized by theionization parameter. It depends on the local star-formation activityand the filling factor of dense gas, regardless of the overall galaxySFR or stellar mass. At this point we may wonder whether thereis any other global galaxy property that correlates with the localnebular conditions. In our sample, the diversity in line ratios is related to the differentdistribution of the SF regions within galaxies. Figure 4 shows themaps of [OIII] and H 𝛼 emission of two main-sequence galaxieswith similar sSFR but different [OIII]/H 𝛼 ratios. The left panelsshow a [OIII]-bright emitter with log([OIII]/H 𝛼 ) = .
15. The [OIII]emission is mostly concentrated in a compact central clump. Thehalf-light area, 𝜋 R , corresponds to a half-light radius of only0.5 kpc. This compact morphology may be related to the fact thatthe galaxy is at the peak of a moderate SF burst. This may bedriven by a compaction event (Dekel & Burkert 2014; Zolotovet al. 2015), in which a significant flow of gas within the galaxyreaches the galaxy center. This leads to extreme conditions of starformation with a relatively high value of the averaged ionizationparameter, log U = − .
7. These nebular conditions boost the [OIII]luminosity with respect to H 𝛼 , specially at the galaxy center, wherethe ionization parameter is the highest, log U = − . 𝛼 -bright galaxy withlog([OIII]/H 𝛼 ) = − .
17. In this case, the [OIII] luminosity is muchmore extended than in the previous example. Its half-light radius is1 kpc. However, the distribution is not smooth. Most of the [OIII] )0.40.20.00.2 l o g ( R / k p c ) M UV -19, z
6 0.40.30.20.10.00.1 l o g ( O III / H ) )0.40.20.00.2 l o g ( R _ H / k p c ) l o g ( O III / H ) Figure 5.
Half-light radius versus sSFR for [OIII] (top) and H 𝛼 (bottom).The vertical lines mark the position of the star-forming main sequenceand its scatter. The scatter at a fixed sSFR is mostly correlated with the lineratio. [OIII]-bright emitters are systematically more compact than H 𝛼 -brightgalaxies. emission is concentrated in a a few bright clumps extended over theregions with the highest ionization parameter. However, the nebularconditions are less extreme, with an averaged ionization parameterof log U = − .
3. Under these conditions, H 𝛼 dominates over [OIII].In conclusion, [OIII] emission is boosted in extremely com-pact SF clumps, where the ionization conditions can be very ex-treme with a very high ionization parameter. On the other hand, H 𝛼 dominates over [OIII] emission if the SF is distributed over severalsmaller clumps, where the ionization parameter is much lower. Asa consequence, H 𝛼 emission seems more extended than [OIII] inboth examples (Figure 4). It covers the inter-clump medium wherethe ionization parameter is lower.Figure 5 shows the half-light radius of the nebular emissionsfor the whole sample. The diversity of sizes, even between galaxiesof similar mass and SFR, is more evident. It correlates with the[OIII]/H 𝛼 ratio and the ionization parameter. [OIII]-bright galaxiestend to be more compact than H 𝛼 -bright galaxies at any sSFR. Thisis true even for galaxies below the main sequence. Therefore, anyobserved sample of extreme [OIII] emitters will be systematicallybiased towards compact, smaller objects. Many of these sourceswill be barely resolved by JWST , assuming a FWHM of around0.12 arcseconds, which corresponds to a minimum half-light radiusof ∼
300 pc at 𝑧 = .
5. The combination of [OIII]-bright and H 𝛼 - MNRAS000
5. The combination of [OIII]-bright and H 𝛼 - MNRAS000 , 1–9 (2021)
Ceverino et al. bright emitters will provide a more representative sample of galaxiesat cosmic dawn.The size of the H 𝛼 emission is significantly larger by 25% (Fig-ure 5) . Therefore, it may be better resolved in future observations.FirstLight predicts many H 𝛼 emitters with effective sizes larger than1 kpc at the end of reionization. NIRSpec IFU observations may beposible for galaxies with moderate sSFR at 𝑧 < . We have used a sample of sub-L ∗ galaxies with UV magnitudes,M UV (cid:39) −
19 at 𝑧 (cid:39)
6, extracted from the FirstLight simulations(Ceverino et al. 2017) to study the diversity of galaxies at the endof the reionization epoch. The main results can be summarized asfollows: • A factor ∼
10 variation in the equivalent width of the[OIII] 𝜆 ∼
40 variation in thespecific SFR. • Variations in nebular metallicity and ionization parameterwithin HII regions generate a scatter in the equivalent width and[OIII]/H 𝛼 line ratio at a fixed sSFR. • OIII-bright (log([OIII]/H 𝛼 ) >
0) emitters have higher ion-ization parameters and/or higher metallicities than H 𝛼 -bright(log([OIII]/H 𝛼 ) <
0) galaxies. • According to the surface brightness maps in both [OIII] andH 𝛼 , [OIII]-bright emitters are more compact than H 𝛼 -bright galax-ies. • H 𝛼 dominates over [OIII] if the star formation is distributedover extended regions. [OIII] dominates if the star formation isconcentrated in large and compact clumps. • The spatial extend of the H 𝛼 emission is significantly largerthan that of the [OIII] emission.These results indicates a large diversity in galaxy properties bythe end of reionization. Even galaxies with similar UV propertiesmay have very different rest-frame visible emission lines. Any com-pilation based only on [OIII] or H 𝛼 emitters may miss a significantand important population. [OIII]-bright emitters have on averagecompact emission regions with relatively high metallicity and/orhigh ionization parameters. On the other hand, H 𝛼 -bright galaxiestend to be significantly more extended and their SF regions haveless extreme nebular conditions.Different processes may transform H 𝛼 -bright galaxies into[OIII]-bright emitters. For example, a large inflow of gas to thegalaxy center, triggered by a compaction event (Dekel & Burkert2014; Zolotov et al. 2015) may generate a central starburst andchange the conditions of the SF regions. Violent disk instabilities(Dekel et al. 2009; Ceverino et al. 2010) produce giant clumps withsimilar gas conditions. The relevance of these processes at cosmicdawn remains to be explored in more detail in future works.One of the caveats in the present analysis is the omissionof radiative transfer effects from the intervening gas. We plan toextend this analysis to more massive galaxies in bigger cosmologicalvolumes and we will need to take into account the effect of dustattenuation. The calculation of the emission lines of this paperrelies on particular models of nebular emission. They have theirown limitations and assumptions about stellar binarity or rotation.However, the two models used in this paper give very similar lineluminosities. This assures us that the diversity reported in this paperis independent of a particular model and it is a strong prediction of the FirstLight simulations. Future observations with JWST and nextgeneration telescopes will be able to test these results.
ACKNOWLEDGEMENTS
We acknowledge stimulating discussions with Xiangcheng Ma,Sune Toft, and Luis Colina. DC is a Ramon-Cajal Researcher andis supported by the Ministerio de Ciencia, Innovación y Univer-sidades (MICIU/FEDER) under research grant PGC2018-094975-C21. This work has been funded by the ERC Advanced Grant,STARLIGHT: Formation of the First Stars (project number 339177).RSK and SCOG also acknowledge support from the DFG via SFB881 ‘The Milky Way System’ (sub-projects B1, B2 and B8) andSPP 1573 ‘Physics of the Interstellar Medium’ (grant number GL668/2-1) and KL 1358/19-2. AF acknowledges the support fromgrant PRIN MIUR2017-20173ML3WW_001. The authors grate-fully acknowledge the Gauss Center for Supercomputing for fundingthis project by providing computing time on the GCS Supercom-puter SuperMUC at Leibniz Supercomputing Centre (Project ID:pr92za). The authors acknowledge support by the state of Baden-Württemberg through bwHPC. We thank the BPASS team for shar-ing their database of SSPs and emission lines. This work madeuse of the v2.1 of the Binary Population and Spectral Synthesis(BPASS) models as last described in Eldridge et al. (2017).
DATA AVAILABILITY
The data underlying this article are available in the First-Light database, at or will be shared on reasonable requestto the corresponding author.
REFERENCES
Álvarez-Márquez J., et al., 2019, A&A, 629, A9Bouwens R. J., et al., 2016, ApJ, 833, 72Bruzual G., Charlot S., 2003, MNRAS, 344, 1000Ceverino D., Klypin A., 2009, ApJ, 695, 292Ceverino D., Dekel A., Bournaud F., 2010, MNRAS, 404, 2151Ceverino D., Klypin A., Klimek E. S., Trujillo-Gomez S., Churchill C. W.,Primack J., Dekel A., 2014, MNRAS, 442, 1545Ceverino D., Glover S. C. O., Klessen R. S., 2017, MNRAS, 470, 2791Ceverino D., Klessen R. S., Glover S. C. O., 2018, MNRAS, 480, 4842Ceverino D., Klessen R. S., Glover S. C. O., 2019, MNRAS, 484, 1366Charlot S., Longhetti M., 2001, MNRAS, 323, 887Chary R.-R., Stern D., Eisenhardt P., 2005, ApJ, 635, L5De Barros S., Oesch P. A., Labbé I., Stefanon M., González V., Smit R.,Bouwens R. J., Illingworth G. D., 2019, MNRAS, 489, 2355Dekel A., Burkert A., 2014, MNRAS, 438, 1870Dekel A., Sari R., Ceverino D., 2009, ApJ, 703, 785Eldridge J. J., Stanway E. R., Xiao L., McClelland L. A. S., Taylor G., NgM., Greis S. M. L., Bray J. C., 2017, Publ. Astron. Soc. Australia, 34,e058Endsley R., Stark D. P., Chevallard J., Charlot S., 2021, MNRAS, 500, 5229Faisst A. L., et al., 2016, ApJ, 821, 122Ferland G. J., et al., 2013, Rev. Mex. Astron. Astrofis., 49, 137Forrest B., et al., 2017, ApJ, 838, L12Gutkin J., Charlot S., Bruzual G., 2016, MNRAS, 462, 1757Hirschmann M., Charlot S., Feltre A., Naab T., Choi E., Ostriker J. P.,Somerville R. S., 2017, MNRAS, 472, 2468Hirschmann M., Charlot S., Feltre A., Naab T., Somerville R. S., Choi E.,2019, MNRAS, 487, 333 MNRAS , 1–9 (2021)
L IV )0.00.20.40.60.8 l o g ( O III / H ) M UV -19, z
6 3.43.23.02.82.6 l o g U Figure A1. [OIII]/H 𝛽 versus sSFR. There is a systematic offset in compar-ison with the [OIII]/H 𝛼 ratio due to the intrinsic ratio of 2.85 between H 𝛼 and H 𝛽 .Izotov Y. I., Guseva N. G., Thuan T. X., 2011, ApJ, 728, 161Katz H., et al., 2019, MNRAS, 487, 5902Klypin A. A., Trujillo-Gomez S., Primack J., 2011, ApJ, 740, 102Komatsu E., Dunkley J., Nolta M. R., Bennett C. L., Gold B., et al. 2009,ApJS, 180, 330Kravtsov A. V., 2003, ApJ, 590, L1Kravtsov A. V., Klypin A. A., Khokhlov A. M., 1997, ApJS, 111, 73Labbé I., et al., 2013, ApJ, 777, L19Langan I., Ceverino D., Finlator K., 2020, MNRAS, 494, 1988Ma X., et al., 2018, MNRAS, 478, 1694Naidu R. P., Tacchella S., Mason C. A., Bose S., Oesch P. A., Conroy C.,2020, ApJ, 892, 109O’Shea B. W., Wise J. H., Xu H., Norman M. L., 2015, ApJ, 807, L12Pallottini A., et al., 2019, MNRAS, 487, 1689Robertson B. E., et al., 2013, ApJ, 768, 71Stark D. P., 2016, ARA&A, 54, 761Stark D. P., Schenker M. A., Ellis R., Robertson B., McLure R., Dunlop J.,2013, ApJ, 763, 129Tang M., Stark D. P., Chevallard J., Charlot S., Endsley R., Congiu E., 2021,MNRAS, 501, 3238Woosley S. E., Weaver T. A., 1995, ApJS, 101, 181Xiao L., Stanway E. R., Eldridge J. J., 2018, MNRAS, 477, 904Zolotov A., et al., 2015, MNRAS, 450, 2327 APPENDIX A: THE [OIII]/H 𝛽 LINE RATIO
This paper was mostly focused on the comparison between[OIII] 𝜆 𝛼 , the two brightest lines in the rest-frame visi-ble. However the [OIII]/H 𝛽 line ratio is also commonly used at lowredshift. Figure A1 provides an analog to Figure 3 and it shows thatthe [OIII]/H 𝛽 ratio is systematically lower by a factor of 2.85, theintrinsic ratio between H 𝛼 and H 𝛽 . This overall offset is expectedbecause dust attenuation is very low in these sub-L ∗ galaxies. This paper has been typeset from a TEX/L A TEX file prepared by the author.MNRAS000