MOSEL: Strong [OIII]5007 Å Emitting Galaxies at (3<z<4) from the ZFOURGE Survey
Kim-Vy H. Tran, Ben Forrest, Leo Y. Alcorn, Tiantian Yuan, Themiya Nanayakkara, Jonathan Cohn, Michael Cowley, Karl Glazebrook, Anshu Gupta, Glenn G. Kacprzak, Lisa Kewley, Ivo Labbe, Casey Papovich, Lee Spitler, Caroline M. S. Straatman, Adam Tomczak
DDraft version April 29, 2020
Typeset using L A TEX preprint style in AASTeX63
MOSEL : Strong [ O iii ]5007˚A Emitting Galaxies at (3 < z < from the ZFOURGE
Survey
Kim-Vy H. Tran,
1, 2, 3
Ben Forrest, Leo Y. Alcorn,
2, 5
Tiantian Yuan,
6, 3
Themiya Nanayakkara, Jonathan Cohn, Michael Cowley,
8, 9
Karl Glazebrook,
6, 3
Anshu Gupta,
1, 3
Glenn G. Kacprzak,
6, 3
Lisa Kewley,
10, 3
Ivo Labb´e, Casey Papovich, Lee Spitler,
11, 3
Caroline M. S. Straatman, Adam Tomczak, School of Physics, University of New South Wales, Kensington, Australia George P. and Cynthia W. Mitchell Institute for Fundamental Physics and Astronomy, Department of Physics &Astronomy, Texas A&M University, College Station, TX 77843, USA ARC Centre for Excellence in All-Sky Astrophysics in 3D (ASTRO 3D) Department of Physics & Astronomy, University of California, Riverside, CA 92521, USA LSSTC Data Science Fellow Swinburne University of Technology, Hawthorn, VIC 3122, Australia Leiden Observatory, Leiden University, P.O. Box 9513, NL 2300 RA Leiden, The Netherlands Centre for Astrophysics, University of Southern Queensland, West Street, Toowoomba, QLD 4350, Australia School of Chemistry, Physics and Mechanical Engineering, Queensland University of Technology, Brisbane, QLD4001, Australia Research School of Astronomy and Astrophysics, The Australian National University, Cotter Road, Weston Creek,ACT 2611, Australia Department of Physics and Astronomy, Faculty of Science and Engineering, Macquarie University, Sydney, NSW2109, Australia Sterrenkundig Observatorium, Universiteit Gent, Krijgslaan 281 S9, 9000 Gent, Belgium Department of Physics, University of California, Davis, One Shields Ave., Davis, CA 95616
ABSTRACTTo understand how strong emission line galaxies (ELGs) contribute to the overallgrowth of galaxies and star formation history of the universe, we target Strong ELGs(SELGs) from the
ZFOURGE imaging survey that have blended H β +[O iii ] rest-frameequivalent widths of > . 0. Using Keck/MOSFIRE, we measure49 redshifts for galaxies brighter than K s = 25 mag as part of our Multi-Object Spectro-scopic Emission Line ( MOSEL ) survey. Our spectroscopic success rate is ∼ 53% and z phot uncertainty is σ z =[∆ z/ (1 + z )] = 0 . < z spec < . iii ]5007˚A equivalent widthsof 100 − (cid:63) / M (cid:12) ) ∼ . − . 6] comparedto more typical star-forming galaxies. The Strong ELGs lie ∼ . z ∼ . gas (cid:38) ∼ − 100 Myr. Combined with recent results using ZFOURGE , our analysis indicatesthat 1) strong [O iii ]5007˚A emission signals an early episode of intense stellar growth Corresponding author: Kim-Vy H. [email protected] a r X i v : . [ a s t r o - ph . GA ] A p r in low mass [M (cid:63) < . (cid:63) ] galaxies and 2) many, if not most, galaxies at z > iii ]5007˚A emis-sion (EW rest > z > Keywords: Emission line galaxies (459), Galaxy evolution (594), Galaxy formation(595), Starburst galaxies (1570), Galaxy properties (615), Near infrared as-tronomy (1093) INTRODUCTIONHierarchical formation predicts that massive galaxies like our own Milky Way grow through themerger and accretion of smaller systems (Peebles 1970), thus low-mass galaxies that are chemicallypristine can provide insight into the early stages of galaxy formation. Although low-mass galaxies areabundant, identifying the ones that are the least chemically evolved via emission lines is difficult dueto their rare nature in the local universe. In the past decade, dwarf galaxies with strong [O iii ]5007˚Aemission at z (cid:46) . β +[O iii ]5007˚Aemission are ubiquitous in Lyman-break galaxies at z ∼ (cid:63) / M (cid:12) ) (cid:46) 9) at 0 < z < iii ]5007˚A emission that may bridge local “green peas” to primeval galaxies at z > 6. Slit-less near-infrared spectroscopy with the Hubble Space Telescope has revealed a population of dwarf galaxiesup to z ∼ iii ]5007˚A equivalent widths of EW rest > z ∼ iii ]5007˚A emitting galaxies at z > λ > µ m), thus only a handful of systems have been confirmedat z (cid:38) iii ]5007˚A emitting galaxies fit intoour existing picture of galaxy formation. The increasing number of Strong ELGs combined with thebrief duration of this intense starburst phase ( (cid:46) 100 Myr; Guo et al. 2016; Ceverino et al. 2018)supports a model where galaxies grow through multiple intense starbursts. For starburst systems at z (cid:38) rest (cid:38) e.g. Lyman-break galaxies at z ∼ β +[O iii ]5007˚A emis-sion (Roberts-Borsani et al. 2016), we can use SELGs to test current galaxy formation models thatcapture the intricate interplay of physics on the sub-kpc scale with the integrated galaxy propertiesthat can be measured at z > β (cid:46) − 2) and low metal-licities ( Z/Z (cid:12) (cid:46) . 2; Forrest et al. 2018; Cohn et al. 2018), i.e. the SELGs may be a tremendoussource of UV photons. By identifying the strong [O iii ]5007˚A emitting galaxies, we can then measuretheir Lyman-Continuum emission and escape fractions to infer if SELGs at z > z ∼ z > 8. However, currentnear-infrared instruments place a redshift limit of z ∼ iii ]5007˚A emitters whichare the focus of our study.An effective method to identify galaxies with strong [O iii ]5007˚A emission (EW rest > z (cid:38) iii ]5007˚A emission tend to be low-mass (log(M (cid:63) / M (cid:12) ) < . 5; e.g. Maseda et al. 2013; Maseda et al. 2014) systems, thus sensitive multi-wavelength imaging isneeded to push down in stellar mass to select candidates. Precise photometric redshifts at z > iii ]5007˚A nature of these systems be confirmed.With the advent of deep near-IR imaging surveys and sensitive near-IR spectrographs, we are nowable to identify these strong [O iii ]5007˚A emitting galaxies at z ∼ − 4. Our method is similar tostudies that couple near-IR imaging and near-IR spectroscopy to identify galaxies with strong equiv-alent widths at z ∼ − 2, e.g. (Maseda et al. 2013; Maseda et al. 2014, 2018). First we usethe ZFOURGE survey that measures precise photometric redshifts to ∼ , 000 objects by combiningdeep imaging with medium-band near-IR filters J J J H s H l K s and public multi-wavelength obser-vations (redshift uncertainties of σ z ∼ . z ∼ 3, the ZFOURGE surveyis 80% mass-complete to log(M (cid:63) / M (cid:12) ) ∼ . ∼ (cid:12) yr − (Tomczak et al. 2016).With photometry spanning observed UV to mid/far-IR, we then construct composite SEDs that aredefined by the underlying galaxy populations (Kriek et al. 2011; Forrest et al. 2016). In our analysisof ZFOURGE galaxies at 2 . 0, we discovered a population of ∼ 80 galaxies with blendedrest-frame H β +[O iii ] equivalent widths in excess of ∼ ∼ 14 galaxies with such extreme H β +[O iii ] at 1 < z < β +[O iii ] emitting galaxies from z ∼ z ∼ . β +[O iii ] emitting galaxies identified in ZFOURGE , we introduceour Multi-Object Spectroscopic Emission Line ( MOSEL ) survey. In this paper, we focus on Emis-sion Line Galaxies (ELGs) at 2 . Hubble Space Telescope . Due to the wavelength rangesof the WFC3 and ACS grisms, blind spectroscopic surveys such as (Momcheva et al. 2016), WISP (Atek et al. 2011), and PEARS (Straughn et al. 2008) are limited to SELGs at z (cid:46) . 3. Ourmedium-band NIR imaging from ZFOURGE combined with public legacy datasets enables us to reach Mosel is also one of the 13 official German wine regions (Weinbaugebiete) and known for Riesling and Pinot Noir. comparable stellar masses as the blind spectroscopic surveys (log(M (cid:63) / M (cid:12) ) ∼ . z ∼ 1; Straatmanet al. 2016). At z ∼ − (cid:38) ∼ − × − erg s − cm − ˚A − ) as the lower redshift studies.In our analysis, we use AB magnitudes and the galaxy parameters measured by Forrest et al. (2017,2018) for the ZFOURGE data-set. FAST (Kriek et al. 2009) is used to fit the SEDs assuming a ChabrierInitial Mass Function and an SED library with 1/5 solar metallicity and emission lines (see Salmonet al. 2015; Forrest et al. 2018). We assume Ω m = 0 . 3, Ω Λ =0.7, H = 70 km s − Mpc − , and a flatUniverse; the corresponding angular scale at z = 3 . DATA & METHODS2.1. Selecting Emission Line Galaxies The following summarizes the ZFOURGE observations we used to measure photometric redshifts andgalaxy properties as well as to generate the composite SEDs. For complete descriptions of the dataproducts used here, we refer the reader to the ZFOURGE survey paper by Straatman et al. (2016) andanalysis of star formation rates by Tomczak et al. (2016).2.1.1. ZFOURGE Imaging Catalogs We use the deep near-IR imaging from the FourStar Galaxy Evolution survey ( ZFOURGE ; Straatmanet al. 2016) obtained with the FourStar imager (Persson et al. 2013) on the Magellan Telescope ofthree legacy fields: CDFS (Giacconi et al. 2002), COSMOS (Scoville et al. 2007), and UDS (Lawrenceet al. 2007). ZFOURGE divides the J -band filter into J , J , and J and the H -band filter into H s and H l ; ZFOURGE also obtains deep K s imaging that is used as the detection image. In combination withexisting multi-wavelength observations, ZFOURGE provides high precision photometric redshifts with σ z = 0 . 016 (Straatman et al. 2016) for over 70,000 objects; the redshift precision is confirmed by the ZFIRE spectroscopic survey (Nanayakkara et al. 2016).We incorporate HST imaging from CANDELS (Grogin et al. 2011; Koekemoer et al. 2011) spanning0.3 µ m to 1.6 µ m to measure photometry and galaxy properties. We also use Spitzer /IRAC (allfour channels) and MIPS data (24 µ m) for the CDFS, COSMOS, and UDS fields (GOODS-S: PIDickinson, COSMOS: PI Scoville, UDS: PI Dunlop), and 100 and 160 µ m for CDFS. For CDFS only,we include public Herschel /PACS data (Elbaz et al. 2011). Total star formation rates are calculatedby combining the UV and IR contributions; see Tomczak et al. (2016, § 2) for a detailed description.2.1.2. Candidate Emission Line Galaxies at . 05, from Kriek et al. 2011; Forrest et al.2016) to collectively form separate composite groups . Observed photometry from analog galaxies ineach composite group are then de-redshifted, normalized to a common flux scale, and combined tobuild a composite SED, essentially a low resolution ( R ∼ 40) spectrum.We focus on the two composite SEDs from Forrest et al. (2017, 2018) with the steepest UV slopeand strongest blended H β +[O iii ] emission. We adopt an admittedly arbitrary definition and referto these emission line galaxies as Strong (SELG). In our analysis, we refer to the following types ofgalaxies: • Star-Forming Galaxy (SFG): composite SEDs with rest-frame H β +[O iii ] equivalent widths of < • Strong Emission Line Galaxy (SELG): the combined sample of 278 galaxies in S ELG andS ELG. • S ELG: the 60 galaxies grouped in the the composite SED with rest-frame H β +[O iii ] equivalentwidth of > • S ELG: the 218 galaxies grouped in the the composite SED with rest-frameH β +[O iii ] equiva-lent width of 230 − ZFOURGE fields, we identify a total of 278 candidate Strong ELGs from the ZFOURGE composite SEDs, the majority of which are in CDF-S (Forrest et al. 2017). Except where noted, weuse the combined sample of SELGs=(S ELG+S ELG) . We exclude Active Galactic Nuclei (AGN)identified by Cowley et al. (2016) using multi-wavelength diagnostics; we discuss this in more detailin § Keck/MOSFIRE Spectroscopy Observations We used MOSFIRE (McLean et al. 2012) on Keck I (project code Z245, PI Kewley) on 12 and13 February 2017. We observed 5 masks in COSMOS and 1 mask in CDFS. To measure H β +[O iii ]at z ∼ 3, we used the K-band with wavelength range of 1 . − . µ m and spectral dispersion of2.17 ˚A/pixel. We used an AB dither pattern with 1 . . − . 0. The remaining targets (67)primarily were star-forming galaxies at 2 4. Each mask included a flux monitor star toanchor the spectro-photometric calibrations. We follow the same reduction pipeline as in our ZFIRE survey (Tran et al. 2015; Nanayakkara et al. 2016; Tran et al. 2017) and estimate a 3 σ line-flux limitin the MOSFIRE K-band of ∼ × − erg s − cm − . Note that the angular sizes of the galaxiesare comparable to or smaller than the slit-width of 0 . (cid:48)(cid:48) (Fig. 7), i.e. there should be no significantsystematic error such as slit-loss due to the spectro-photometric calibration.2.2.2. Spectroscopic Redshifts Of the 105 galaxies targeted with MOSFIRE, we measure spectroscopic redshifts for 49 (2 . 5. In our analysis, Q z ≥ . ZFOURGE Figure 1. We compare the photometric redshifts (top) and K s magnitudes (bottom) of the 95 targetedgalaxies (small filled black circles & open black histograms) to the 49 galaxies with spectroscopic redshifts(vertical pairs of large filled+open red circles & hashed red histograms). The two samples are likely drawnfrom the same parent population: KS tests comparing the z phot and K s magnitudes of the 49 galaxies with z spec to that of the targeted sample measure probabilities of 14% and 6% respectively. The median z spec ofthese 49 galaxies is only ∼ 1% lower than their median z phot (3.19 vs. 3.22). Our analysis focuses on the 31spectroscopically confirmed galaxies at 3 < z spec < . β +[O iii ] fall in the MOSFIRE K-band (toppanel, horizontal dashed lines). photometric redshift or there are two spectral lines with the same redshift (for all definitions of Q z ,see Nanayakkara et al. 2016).On average, z phot is ∼ . 054 higher than z spec (Fig. 1) and the corresponding uncertainty is σ z =[∆ z/ (1 + z )] = 0 . z ∼ . σ z ∼ . 07. A two-sampleKolmogrov-Smirnov test shows the probability that the spectroscopically confirmed sample and thetargeted sample are drawn from the same parent z phot distribution to be 3.6%, i.e. the two distribu- Wavelength ( ˚A) − . . . . . . . F l ux ( − e r g / s / c m / ˚ A ) CDFS Wavelength ( ˚A) . . . . F l ux ( − e r g / s / c m / ˚ A ) COSMOS Wavelength ( ˚A) . . . . F l ux ( − e r g / s / c m / ˚ A ) COSMOS Wavelength ( ˚A) . . . . F l ux ( − e r g / s / c m / ˚ A ) COSMOS Wavelength ( ˚A) . . . . F l ux ( − e r g / s / c m / ˚ A ) COSMOS Figure 2. Example of fits to the MOSFIRE spectra ( § tions are different at the 2 σ level. The spectroscopically confirmed sample also is ∼ . 25 magnitudesbrighter with a K-S probability of being drawn from the same parent K s distribution as the targetedsample of 15%.Of the 13 targeted S ELG that were grouped in the composite SED with the highest H β +[O iii ]emission (EW rest > z spec =3.189 comparedto their median z phot =3.207. The corresponding uncertainty of σ z =0.42% is even lower than thatof our ZFIRE survey which targeted a broader selection of galaxies at z ∼ z ∼ . z ∼ . < z spec < . z spec = 2 . 549 (ID 4791).Of the 25 targeted S ELG that were grouped in the composite SED with the second highestH β +[O iii ] emission (230 < EW rest < z spec =3.327 compared to their median z phot =3.41. The corresponding uncertainty of σ z =1.9%for the S ELG is larger than that of the S ELG and more typical of the ZFOURGE survey as a whole(Straatman et al. 2016; Nanayakkara et al. 2016). All 13 have redshifts of 3 < z spec < . z spec =2.551 compared to their median z phot =2.612. In our analysis, we focus onthe 10 SFGs at 3 < z spec < . 8, i.e. we exclude the 17 galaxies at z spec < . z spec = 4 . λ (cid:38) . µ m, theredshift cut-off is effectively z ∼ . iii ]5007˚A emission.2.2.3. Measuring [ O iii ]5007˚A Spectral Line Emission Following Alcorn et al. (2016, 2018), we first extract a 1D spectrum from an aperture defined by the1 σ Gaussian width of the [O iii ]5007˚A emission line along the spatial direction (Fig. 2). To determinethe [OIII]5007˚A line flux, we integrate the best-fit Gaussian centered on the line emission along thewavelength direction using the 3 σ range; all line-fits are visually inspected for quality control. Wesubtract in quadrature the instrumental broadening from the measured line-width and then convertto an integrated velocity dispersion ( σ int ) using the galaxy’s measured redshift. Errors in σ int areestimated by adding sky noise to the observed spectrum and refitting 1000 times.2.2.4. Determining [ O iii ]5007˚A Equivalent Width Figure 3. Spectroscopic rest-frame [O iii ]5007˚A equivalent widths for confirmed Emission Line Galaxies(ELGs) at 3 < z spec < . 8; typical uncertainties in stellar mass are ∼ β +[O iii ] equivalent width as measured from the two compositeSEDs with the strongest emission from Forrest et al. (2017). Shown are: 1) S ELG with blended rest-frame H β +[O iii ] EW rest > 800 ˚A as measured from their Composite SED; 2) S ELG with blended rest-frame H β +[O iii ] EW rest ∼ − 800 ˚A; and 3) more typical star-forming galaxies with blended rest-frameH β +[O iii ] EW rest < 230 ˚A. For comparison, Strong ELGs at 1 < z < (Maseda et al.2013; Maseda et al. 2014) are shown in cyan; the spectral resolution and flux limit of the grism observationsselects ELGs with the highest [O iii ]5007˚A EWs. The dashed line shows the average relationship between[O iii ]5007˚A EW obs and stellar mass from MOSDEF for galaxies at z ∼ To measure [O iii ]5007˚A equivalent widths, we require both line and continuum flux. However, mostof the ELGs are too faint to directly measure their continua from the MOSFIRE spectroscopy. Weuse the method described in Nanayakkara et al. (2017) that combines our spectro-photometricallycalibrated [O iii ]5007˚A line fluxes with the deep continuum photometry from ZFOURGE (Straatmanet al. 2016). Because the ZFOURGE photometry provides a better measurement of the faint continuumrelative to the spectroscopy, the primary source of uncertainty is thus due to systematic error of thespectro-photometric calibration, and this uncertainty is of order ∼ − 20% for continuum-detectedgalaxies (Nanayakkara et al. 2016). Note that given the galaxy sizes are comparable to or smallerthan the slit-width of (Fig. 7), the systematic error due to the spectro-photometric calibration is notsignificant.To determine the continuum for each ELG, we use the FAST fits (Kriek et al. 2009) from Forrestet al. (2017, 2018) that include a template library with strong emission lines. As we show in both Cohnet al. (2018) and Forrest et al. (2018), the stellar masses for low-mass galaxies (log(M (cid:63) / M (cid:12) ) (cid:46) 10) canbe overestimated by ∼ . − . Cloudy (Ferlandet al. 2013) with BC03 simple stellar populations (Bruzual & Charlot 2003) as the ionizing source togenerate nebular emission models.Because both star formation rate and stellar mass depend on the adopted stellar metallicity, SEDfits are generated for solar ( Z = 0 . 02) and subsolar ( Z = 0 . Z = 0 . rest , onlyhow we interpret the measurements.We calculate the observed frame continuum on the blue and red side of the H β +[O iii ] lines byusing tophat filters (width of 150˚A) centered at 4675˚A and 5200˚A on the best-fit FAST SED. Wethen divide the observed [O iii ]5007˚A line flux by the average observed continuum and the galaxyredshift. EW rest (5007) = f line (5007)[( f cont (4675) + f cont (5200)] / (cid:18) 11 + z (cid:19) (1)For a line flux of 3 × − erg s − cm − and continuum flux of 5 × − erg s − cm − ˚A − (approximately K s magnitude of 24.0), the observed equivalent width is 60˚A; for a galaxy at z spec =3.0,the corresponding rest-frame equivalent width is EW rest = 15˚A. For comparison, the lowest values wemeasure for the spectral rest-frame equivalent widths using MOSFIRE are ∼ § rest down to ∼ survey quotes a 3 σ emission line flux limit for point sources of 1 . × − erg s − cm − (Momcheva et al. 2016). Assuming the same continuum flux level, their limitcorresponds to an observed equivalent width of 300˚A, i.e. approximately five times higher than MOSEL . RESULTSIn our analysis, we focus on the 31 galaxies that are spectroscopically confirmed to be at 3 < z spec < . iii ]5007˚A emission for these galaxies with our K-band spectroscopy0 Figure 4. Rest-frame U V J colors of ZFOURGE galaxies at 2 . UV+IR -based SF rates from Tomczak et al. (2016) based on solar metallicity models are robust to significantflux from line emission, the stellar masses can be overestimated by as much as a factor of ∼ Z = 0 . Strong [ O iii ]5007˚A Emission With our MOSFIRE spectroscopy and deep multi-band imaging, we estimate rest-frame[O iii ]5007˚A equivalent widths using the hybrid method described in § iii ]5007˚A EW rest ∼ − β +[O iii ] equivalent widths (EW rest (cid:38) < z spec < . iii ]5007˚A equivalent widthwith increasing stellar mass that is also observed in SELGs at z ∼ ELG and S ELG indicates that the two are not distinctly differentpopulations. Note that the SELGs at z ∼ (cid:63) / M (cid:12) ) ∼ z ∼ − (cid:63) / M (cid:12) ) (cid:38) . β +[O iii ] (cid:46) < z spec < . (cid:63) / M (cid:12) ) (cid:38) 10) and lower rest-frame [O iii ]5007˚A equivalent widths (EW rest ∼ − iii ]5007˚A line-flux, the galaxy with the brighter continuum will have a lower equivalentwidth. Our results confirm that selecting Strong ELGs from the ZFOURGE photometry is effective atidentifying galaxies with the largest [O iii ]5007˚A equivalent widths.For the S ELG, the [O iii ]5007˚A EW rest values determined using the line fluxes obtained withMOSFIRE (see § rest value estimated from the composite SED(Fig. 3). The offset is likely driven by how the continuum and emission lines are combined togenerate the template used to fit the composite SEDs. For example, underestimating the continuumwill increase the inferred EW. We refer the reader to Forrest et al. (2017) who test three fittingmethods on the composite SEDs of the strongest ELGs.3.2. Rest-frame U V J Colors With the ZFOURGE rest-frame wavelength coverage of 0 . − µ m for each galaxy, we measurecontinuum properties including rest-frame U V J colors from the individual SEDs (Tomczak et al.2014; Straatman et al. 2016). ZFOURGE galaxies at 2 . U V J diagram (see also Straatman et al. 2016). The “typical” star-forming galaxies in our spectroscopicsample have ( V − J ) (cid:46) . V < . 5; Forrestet al. 2016). None of the spectroscopically confirmed galaxies are dusty as defined using the criterionfrom Spitler et al. (2014) of ( V − J ) ≥ . V − J ) colors (Fig. 4). Their strong H β +[O iii ]emission significantly boosts their V -band fluxes to produce rest-frame values of ( V − J ) < 0; this isparticularly striking for the S ELG where virtually all have ( V − J ) < 0. Such blue U V J colors andnon-detections in the far-IR indicate that these ELGs are essentially dust-free systems. The relativedistributions of the S ELG and S ELG in the U V J diagram suggests a continuum of phases whereage and dust content increases from the Strong ELGs to the more typical star-forming galaxies, e.g.Lyman-Break Galaxies.2 Figure 5. H β +[O iii ] emitting galaxies tend to lie ∼ . z = 3 . z ∼ . ZFOURGE (Tomczak et al. 2016). Total star formationrates are based on ZFOURGE (UV+IR) fluxes, and here we plot only the galaxies that are individuallydetected in the IR. The Strong H β +[O iii ]-emitters with z spec (large stars) have the same distribution astheir respective z phot samples (small stars). The diagonal dotted lines denote mass-doubling timescales of10 and 100 Myr. Strong ELGs at 1 . < z < . Star-Formation Rate vs. Stellar Mass The Strong ELGs tend to be lower mass systems [log(M (cid:63) / M (cid:12) ) ∼ . − . 6] compared to moretypical star-forming galaxies (Fig. 3). At z ∼ 3, the log(M (cid:63) ) for the stellar luminosity function from ZFOURGE is ∼ . (cid:63) ∼ (0 . − . 08) M (cid:63) .3 Figure 6. The starburst nature of the H β +[O iii ] emitting galaxies is underlined by their high specific StarFormation Rates (SSFR) defined as their (UV+IR) SFRs divided by their stellar masses. Symbols are asin Fig. 5 and included for comparison is the SSFR-M (cid:63) at z = 3 . < 100 Myr with several systems at < 10 Myr (dottedhorizontal lines). Included for comparison are the ELGs at 1 < z < > 10 Myr and tend to be low-mass (log(M (cid:63) / M (cid:12) ) < Figure 5 shows the star formation rate to stellar mass (SFR-M (cid:63) ) for the galaxies in our sample withmeasured (UV+IR) star formation rates from ZFOURGE (Tomczak et al. 2016). Although all of ourgalaxies have measured UV fluxes, many have negative IR fluxes due to the SED fitting method (see § (cid:63) analysis. Also, note that our 3 σ line-flux limit in the MOSFIREK-band is ∼ × − erg s − cm − (see also Tran et al. 2017).4Of the 18 ELGs with positive (UV+IR) SFRs, all lie above the relation between star formationand stellar mass commonly referred to as the Star-Forming Main Sequence (SFMS; Figs. 5 & 6); weconfirm this is true even if we include UV only SFRs. The ELGs tend to lie ∼ . z = 3 . ZFOURGE . The overall distribution of the spectroscopically confirmed Strong ELGs mirrors thatof the photometrically selected sample at this epoch, i.e. most of the SELGs lie above the SFMS.With stellar mass-doubling time-scales of only ∼ − 100 Myr, virtually all of the Strong ELGsare starbursts (Figs. 5 & 6). Our results are consistent with Amor´ın et al. (2017) and Maseda et al.(2014) who find that Strong ELGs at 1 < z < (cid:63) / M (cid:12) ) ∼ − . . Galaxy Size vs. Stellar Mass Our Emission Line Galaxies lie on the galaxy size-mass (r eff -M (cid:63) ) relation measured by Allen et al.(2017) using ZFOURGE galaxies at 3 75 (Fig. 7). Here we use the effective radii (galaxy size)measured by van der Wel et al. (2012) with the WFC3/F160W imaging and consider only galaxieswith goodness of fit flag of 0. These criteria further reduce our ELG sample to 13 galaxies. We notethat relaxing the goodness of fit flag to include all ELGs with measured r eff (28) increases the scatterin the galaxy size-mass relation but does not change the overall result.Although they are virtually all starbursts (Figs. 5 & 6), our ELGs at 2 . With UV+IR luminosities from ZFOURGE and r eff from the HST/F160W imaging (van der Welet al. 2012), we use the Schmidt-Kennicutt Relation (SKR; Schmidt 1959; Kennicutt 1998, Eq. 7) toestimate the gas surface density for individual galaxies:Σ SFR = (2 . ± . × − (cid:18) Σ gas (cid:12) pc − (cid:19) . ± . M (cid:12) year − kpc − (2)Assuming that half of the gas mass is within r eff , we use Σ SFR to estimate the total gas mass:log(M gas ) = 57 log(L UV+IR ) + 27 log[ π (r eff ) ] + 1 . 52 (3)5 Figure 7. The H β +[O iii ] emitters are consistent with the ZFOURGE galaxy size-stellar mass relation at3 75 from Allen et al. (2017) (solid line), but there is considerable scatter both in the ELGsand for all ZFOURGE galaxies at 2 . 0. Here we use the effective radii measured by van der Welet al. (2012) with WFC3/F160W imaging and the symbols are as in Fig. 5. The illustrative errorbarscorrespond to a factor of two uncertainty in stellar mass and galaxy size. The strong ELGs at z ∼ (cid:63) / M (cid:12) ) < . 5) where they tend to be more compact. where r eff is measured in kpc, M gas in M (cid:12) , and L UV+IR in L (cid:12) ; see also Papovich et al. (2015). Weuse r eff defined by the stellar light; note that studies using CO (Tacconi et al. 2013) and H α (F¨orsterSchreiber et al. 2011) find r eff from gas and stars are consistent. Assuming an observational detectionlimit of L UV+IR = 10 L (cid:12) and typical galaxy size of r eff = 3 . gas /M (cid:12) )= 9 . 8. Gas fractions are defined to be M gas /(M gas +M (cid:63) ).All of our spectroscopically confirmed ELGS have inferred gas fractions of f gas > 60% (Fig. 8) whichis not surprising given the ELGs’ high specific star formation rates and our detection limits. The6 Figure 8. We infer gas masses by using the Schmidt-Kennicutt Relation (SKR; Kennicutt 1998) witheffective sizes from vdW12 and total luminosities from Tomczak et al. (2016); symbols are as in Figs. 4 & 5.The illustrative errorbars correspond to a factor of two uncertainty in stellar mass and we note that the gasmass is highly uncertain. For reference, we show the exclusion region corresponding to a gas mass limit oflog(M (cid:63) / M (cid:12) )= 9 . z (cid:38) 3. Allof the spectroscopically confirmed ELGs in our study have inferred gas mass fractions of f gas > z ∼ . 5. The SELGs will move to the right as they increase in stellar mass and, unless their gasreservoirs are replenished, downwards. high gas masses are consistent with high accretion rates that may be driving the star formation(Kacprzak et al. 2016). Our inferred gas fractions combined with measurements of Strong ELGs at z ∼ gas (cid:38) 80% for ELGs with stellar masses of log(M (cid:63) / M (cid:12) ) < 9. However, direct measurements of gas masses for the ELGs at z > (cid:63) / M (cid:12) ) < Figure 9. Left: The dynamical masses for the [O iii ]5007˚A emitting galaxies at 3 < z spec < . ∼ . iii ]5007˚A line-widths witheffective radii measured by vdW12 (see Alcorn et al. 2016, 2018). The solid diagonal line is parity and thedotted line is offset by 0.5 dex. Our results are consistent with studies showing that ELGs at z ∼ dyn > M (cid:63) (Maseda et al. 2013; Maseda et al. 2014) and with their inferred gas mass fractions off gas > Right: The same MOSEL galaxies where we now compare the total baryonic mass (sum of thestellar and estimated gas mass) to their dynamical mass. Including f gas brings the MOSEL galaxies closer toparity, but we hesitate to draw any stronger conclusions given the scatter and low number statistics. Kinematics The integrated velocity dispersions ( σ int ) based on [O iii ]5007˚A line-widths is σ int ∼ − 150 km s − for most of the ELGs with only one ELG having σ int ∼ 200 km s − (see Tables 1 & 2). Combining σ int and effective radii from vdW12, we follow Alcorn et al. (2016) and estimate virial masses:M dyn ( < R e ) = K e σ R e G (4)where for consistency with Maseda et al. (2013); Maseda et al. (2014), we adopt the virial factor K e = 3 which is typically used for disk galaxies.The dynamical masses for our ELGs at 3 < z spec < . ∼ . dyn < M (cid:63) is consistent with scatter due to errors in themeasurements. The offset between virial and stellar mass is consistent with measurements of StrongELGs at z ∼ < z spec < . 8, there is no obvious difference in the M (cid:63) -M dyn relation for Strong ELGs compared to higher mass (log(M (cid:63) / M (cid:12) ) > 9) star-forming galaxies.8When comparing the total baryonic mass (sum of the stellar and estimated gas mass) to dynamicalmass, we find that the MOSEL galaxies are closer to parity (Fig. 9, left). However, given the scatterand low number statistics, we hesitate to draw any stronger conclusions regarding the ratio of darkto baryonic mass for the MOSEL galaxies.3.7. Star Formation or Active Galactic Nuclei? Our analysis assumes that the strong [O iii ]5007˚A emission is driven by star formation and notActive Galactic Nuclei (AGN). We have used the ZFOURGE catalog by Cowley et al. (2016) to re-move AGN but recognize that at z > 3, the multi-wavelength AGN diagnostics may not be reliableespecially given the uneven coverage across these fields. However, the [O iii ]5007˚A line-widths areconsistent with star formation: most of the ELGs have σ int ∼ − 150 km s − with only one ELGhaving σ int ∼ 200 km s − (see Tables 1 & 2). Also, our recent results using Prospector to constructthe star formation histories of the strong ELGs confirms that they are dominated by starbursts span-ning the most recent ∼ 50 Myr (Cohn et al. 2018). Lastly, we note that AGN contamination is rarein low-mass galaxies (e.g., Ho et al. 1997; Trump et al. 2015).Unlike the ZFOURGE composite SEDs where H β and [O iii ]5007˚A are blended (Forrest et al. 2017),the MOSFIRE spectroscopy easily resolves these spectral features for individual ELGs. Thus we alsocan identify potential AGN by combining the ratio of [O iii ]5007˚A to H β with stellar mass (Juneauet al. 2011), although we note this method is contested at z > iii ]5007/H β value for our sample of ELGs is ∼ . z ∼ 3. Following a similar line of analysis, Masedaet al. (2014) also excluded AGN from their sample of strong ELGs at 1 < z < β is weaker than [O iii ]5007˚A and given our line-flux limit of ∼ × − erg s − cm − (3 σ ), we canonly place upper limits on the ratio of [O iii ]5007/H β for many of the ELGs. A more careful treatmentof the H β line-fluxes, e.g. by stacking the spectra, can be used to constrain ISM conditions. Furtheranalysis that includes H β , e.g. by combining [O iii ]5007/H β with stellar masses and star formationhistories, will be presented in a future MOSEL paper. DISCUSSIONWith deep multi-band photometry from ZFOURGE , we identified Emission Line Galaxies at z > . β +[O iii ] equivalent widths of (cid:38) β +[O iii ] equivalent widths (EW rest > § rest ([O iii ]5007) (cid:38) ∼ MOSEL survey, we build on recent studies to further explore how galaxiesat z ∼ . iii ]5007˚A emission fit into our current understanding of how star-forminggalaxies grow by combining Keck/MOSFIRE K-band spectroscopy with our existing multi-band pho-tometry from ZFOURGE .4.1. Strong [ O iii ]5007˚A Emission May Be Common in Early Galaxy Formation We spectroscopically confirm 31 galaxies at 3 < z spec < . (cid:63) / M (cid:12) ) ∼ . − . iii ]5007˚A equivalent widths up to ∼ V − J ) < V − J ) ∼ − U V J colors suggest that the Strong ELGs transition into more massive star-forming galaxies, e.g.Lyman-Break Galaxies.In the stellar mass range where we overlap with MOSDEF galaxies at z ∼ iii ]5007˚A EW rest and stellar mass (Fig. 3). Reddy et al. (2018)suggest that the increasing [O iii ]5007˚A EW rest with decreasing stellar mass can be explained byeither rapid enrichment of α elements or metallicities of (cid:46) . Z (cid:12) for galaxies with log(M (cid:63) / M (cid:12) ) (cid:46) z ∼ . Z (cid:63) (cid:46) . Z (cid:12) ) and higher specific star formation rates relative to SFGs (4.6 Gyr − − iii ]5007˚A emissionsignals the earliest episodes of intense star formation (see also Amor´ın et al. 2017). As the SELGsgrow in stellar mass, the growing amount of continuum light means that even during subsequentepisodes of bursty star formation, the [O iii ]5007˚A equivalent widths will not be as large as duringthe first major burst of star formation. With star formation rates of (cid:38) − 250 M (cid:12) yr − (Fig. 5)and mass-doubling times of ∼ − 100 Myr (Figs. 5 & 6), the intense [O iii ]5007˚A emission phase isbrief as these same galaxies quickly transition into more typical star-forming galaxies with H β +[O iii ]EW rest (cid:46) iii ]5007˚A emission signals the earliest stagesof stellar growth in galaxies by comparing relations between stellar mass (M (cid:63) ), galaxy size (r eff ),and virial mass (M dyn ). Our SELGs follow the same general M (cid:63) -r eff relation as that of star-forminggalaxies at z ∼ dyn -M (cid:63) as measured for SELGs at z ∼ ∼ . gas > 60% (Fig. 8).In a recent paper (Cohn et al. 2018), we derived galaxy properties from the ZFOURGE photome-try using the SED-fitting code Prospector (Leja et al. 2017) and the Flexible Stellar PopulationSynthesis package (FSPS; Conroy et al. 2009). The Prospector code finds the best fit model andestimates uncertainties by sampling the posterior probability distributions of all the free parameters.By calculating nonparametric star formation histories, Prospector can distinguish between rising,falling, and bursty star formation histories.Using Prospector , Cohn et al. (2018) show that ELGs with extreme emission (S ELG; H β +[O iii ]EW rest (cid:38) Z ∗ (cid:46) . Z (cid:12) and higher specific star formation ratescompared to star-forming galaxies: ∼ . − vs. ∼ . − . Cohn et al. (2018) inferred thatmany, if not most, star-forming galaxies at z > . β +[O iii ] emission-line phasesearly in their formation histories. As these “first burst” systems continue to form stars and chemicallyenrich to evolve into more typical SFGs, they move diagonally from the upper left to the bottomright in Figs. 6 & 8. 4.2. A Potential Source of Ionizing UV Photons 0A growing number of studies indicate that galaxies rather than AGN generated the UV photonsneeded to ionize the universe, but there are not enough massive galaxies at z (cid:38) (cid:63) / M (cid:12) ) (cid:46) 9) systems now identified at z > rest ([O iii ]5007) > f esc (cid:38) 10% (Nakajima et al. 2016; de Barros et al. 2016), the most viable source ofUV photons are these low-mass, star-bursting galaxies. However, the stellar mass function at z > rest ([O iii ]5007) > gas (cid:38) 60% (Fig. 8)and high specific star formation rates (Fig. 6) imply that the ELGs with the strongest [O iii ]5007˚Aeasily increase their stellar masses by factors > ∼ 100 Myr, i.e. these Strong ELGssignal the earliest stages of stellar growth in galaxies (see also Cohn et al. 2018).If the [O iii ]5007˚A emitters also have large [O iii ]5007/[O ii ]3727 ratios (O32 (cid:38) z ∼ 2, the ionizing photon efficiency scales with [O iii ]5007˚A emission. However,recent results by Bassett et al. (2019) of galaxies at z ∼ iii ]5007/[O ii ]3727 ratios and more Ly-C photons is weak at best, and Naidu et al. (2018) constrainthe average escape fractions for SELGs to be 8 . − . iii ]5007 emission, e.g. shocks or massive binarystars (Strom et al. 2017). By obtaining at z ∼ . iii ]5007 to well-studied emission linessuch as [O ii ]3727, H β , and Ly α (e.g. Tang et al. 2019; Bassett et al. 2019), we can better track howthe ionizing photon efficiency evolves from the first galaxies to z ∼ 0. We plan to measure [O ii ]3727˚Aemission for our ELGs to characterize their ionization conditions and constrain their production ofLyman-Continuum photons. CONCLUSIONSOur Multi-Object Spectroscopic Emission Line ( MOSEL ) survey focuses on galaxies with strong[O iii ]5007˚A emission identified using deep broad-band photometry from the ZFOURGE survey (For-rest et al. 2017, 2018). We use Keck/MOSFIRE K-band spectroscopy and measure redshifts of 49galaxies at 2 . < z spec < . ∼ 53% and z phot uncertainty is σ z =[∆ z/ (1 + z )] = 0 . § . < z spec < . iii ]5007˚A linefluxes for 31 galaxies at 3 < z spec < . ZFOURGE , we estimate rest-frame [O iii ]5007˚A equivalent widths of ∼ − rest increases with decreasing stellar mass (Fig. 3). Our analysis focuses onthe Strong Emission Line Galaxies (SELGs) grouped in the two composite SEDs with the strongestH β +[O iii ] emission (EW rest > z ∼ . iii ]5007˚A StrongEmission Line Galaxies (SELGs) mirror that of the larger photometrically selected sample. For1example, the SELGs tend to have bluer colors of ( V − J ) < V − J ) ∼ − β +[O iii ] emitting galaxies in our study have stellar masses of log(M (cid:63) / M (cid:12) ) ∼ . − . ∼ . z = 3 . ∼ − 100 Myr (Fig. 6). Theinferred gas fractions of f gas (cid:38) 60% (Fig. 8) can easily fuel a burst that increases stellar mass by > β +[O iii ]emitting galaxies bridge relations measured for Strong ELGs at 1 < z < (cid:63) / M (cid:12) ) (cid:46) 9; van derWel et al. 2011; Maseda et al. 2014) to star-forming galaxies at z ∼ . iii ]5007˚A emission (EW rest (cid:38) (cid:63) < . (cid:63) ) galaxies at z (cid:38) 3. The ELGswith the strongest [O iii ]5007˚A are a rapidly evolving population of galaxies both in number densityand stellar growth (Forrest et al. 2017; Cohn et al. 2018). The [O iii ]5007˚A ELGs are likely to evolveinto more massive and older star-forming galaxies with stable disks and bulges, e.g. Lyman-BreakGalaxies.In a recent paper (Cohn et al. 2018), we estimated that many, if not most, star-forming galaxiesat z > iii ]5007˚A emitters early in their formation history. If strong [O iii ]5007˚Aemission is a common phase in early galaxy formation, this brief episode may generate a significantnumber of ionizing UV photons. In a future paper, we will explore additional line diagnostics, e.g.the ratio of [O iii ]5007˚A to H β , to characterize ionization conditions and constrain the production ofLyman-Continuum photons in galaxies with the strongest [O iii ]5007˚A emission.ACKNOWLEDGMENTSWe are grateful to the Keck/MOSFIRE team with special thanks to M. Kassis, J. Lyke, G. Wirth,and L. Rizzi on the Keck support staff. K. Tran thanks P. Oesch, B. Holden, and M. Maseda forhelpful discussions, and B. Forrest thanks the Hagler Institute for Advanced Study at Texas A&Mfor support. We also thank the referee for a thoughtful and constructive report. This work wassupported by a NASA Keck PI Data Award administered by the NASA Exoplanet Science Institute.Data presented herein were obtained at the W. M. Keck Observatory from telescope time allocated toNASA through the agency’s scientific partnership with the California Institute of Technology and theUniversity of California. The Observatory was made possible by the generous financial support of theW. M. Keck Foundation. K. Tran acknowledges that this material is based upon work supported bythe National Science Foundation under Grant Number 1410728 and acknowledges the ARC Centrefor Excellence in All-Sky Astrophysics in 3D (ASTRO 3D) for support in preparing the manuscript.GGK acknowledges the support of the Australian Research Council through the DP170103470. Theauthors wish to recognize and acknowledge the very significant cultural role and reverence that thesummit of Mauna Kea has always had within the indigenous Hawaiian community. We are mostfortunate to have the opportunity to conduct observations from this mountain.2 T a b l e . MOSEL G a l a xy P r o p e r t i e s a F i e l d ZFOURGE b R A b D ec b z s p ec z ph o t b K o b s b ( U − V ) b ( V − J ) b l og ( M (cid:63) / M (cid:12) ) c l og ( S F R ) c r e ff d I D J J ± . ± . m ag m ag m ag M (cid:12) y r − k p c C O S M O S . . . . . . . . . . C O S M O S . . . . . . . . . . C O S M O S . . . . . . - . . < . C O S M O S . . . . . . - . . . . C O S M O S . . . . . . - . . < . C O S M O S . . . . . . - . . . . C O S M O S . . . . . . - . . . . C O S M O S . . . . . . . . . ··· C O S M O S . . . . . . - . . . . C O S M O S . . . . . . . . . ··· C O S M O S . . . . . . . . . ··· C O S M O S . . . . . . . . . . C O S M O S . . . . . . - . . < . C O S M O S . . . . . . - . . . . C O S M O S . . . . . . . . . . C O S M O S . . . . . . - . . < . C O S M O S . . . . . . . . . ··· C O S M O S . . . . . . - . . < ··· C O S M O S . . . . . . . . . ··· C O S M O S . . . . . . . . < . C O S M O S . . . . . . . . < . C O S M O S . . . . . . - . . . . C O S M O S . . . . . . - . . . . C O S M O S . . . . . . - . . < . C O S M O S . . . . . . - . . . ··· C D F S . - . . . . . - . . . . C D F S . - . . . . . . . . . C D F S . - . . . . . - . . . . C D F S . - . . . . . - . . < . C D F S . - . . . . . - . . . ··· C D F S . - . . . . . - . . < . a W e i n c l ud e o n l y MOSEL ga l a x i e s w i t h s p ec t r o s c o p i c r e d s h i f t q u a li t y fl ago f Q z ≥ . ( s ee T r a n e t a l. ; N a n a y a kk a r a e t a l. ) . b G a l a xy i d e n t i fi c a t i o nnu m b e r s , o b s e r v e d ZFOURGE K - b a nd m ag n i t ud e s , ph o t o m e t r i c r e d s h i f t s , a nd r e s t - f r a m e U V J a r e f r o m ZFOURGE ( S t r aa t m a n e t a l. ) . U n ce r t a i n t i e s o n t h e m ag n i t ud e s a nd c o l o r s a r e < . . c W e u s e t h e s t e ll a r m a ss e s f r o m F o rr e s t e t a l. ( ) a nd t h ec o m b i n e d UV + I R s t a r f o r m a t i o n r a t e s f r o m T o m cz a k e t a l. ( ) . W e r ec o mm e nd t h e r e a d e r c o n s i d e r a t y p i c a l un ce r t a i n t y o f ∼ . d e x f o r b o t hp a r a m e t e r s . d E ff ec t i v e r a d ii a r e f r o m v a nd e r W e l e t a l. ( ) a nd m e a s u r e du s i n g t h e W F C F W i m ag i n g . H e r e w e t a k e t h e s i ze s r e p o r t e d i n a r c s ec a nd c o n v e r tt o k p c u s i n g t h e a n g u l a r d i a m e t e r d i s t a n ce . Table 2. MOSEL : [O iii ]5007˚A Properties Field ZFOURGE a f (5007˚A) b σ b σ int b EW rest c ID 10 − erg s − cm − ˚A km s − ˚ACOSMOS 1877 2.2 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± a Galaxy identification numbers are from ZFOURGE (Straatman et al. 2016). b [O iii ]5007˚A line-flux, line-width ( σ ), and the corresponding integrated ve-locity dispersion ( σ int ) are measured by fitting a Gaussian to the MOSFIREspectra (see § c [O iii ]5007˚A rest-frame equivalent widths are determined using the line fluxmeasured with MOSFIRE spectroscopy and continuum flux from the best-fitFAST SED (see § Alcorn, L. Y., Tran, K.-V. H., Kacprzak, G. G.,et al. 2016, ApJL, 825, L2Alcorn, L. Y., Tran, K.-V., Glazebrook, K., et al.2018, ApJ, 858, 47Allen, R. J., Kacprzak, G. G., Glazebrook, K.,et al. 2017, ApJL, 834, L11Amor´ın, R., Fontana, A., P´erez-Montero, E., et al.2017, Nature Astronomy, 1, 0052Amor´ın, R., P´erez-Montero, E., Contini, T., et al.2015, A&A, 578, A105Amor´ın, R., P´erez-Montero, E., V´ılchez, J. M., &Papaderos, P. 2012, ApJ, 749, 185Atek, H., Siana, B., Scarlata, C., et al. 2011, TheAstrophysical Journal, 743, 121Bassett, R., Ryan-Weber, E. V., Cooke, J., et al.2019, MNRAS, 483, 5223Bian, F., Kewley, L. J., Dopita, M. A., & Juneau,S. 2016, ApJ, 822, 62Bruzual, G., & Charlot, S. 2003, MNRAS, 344,1000Cardamone, C., Schawinski, K., Sarzi, M., et al.2009, Monthly Notices of the RoyalAstronomical Society, 399, 1191Ceverino, D., Klessen, R., & Glover, S. 2018,ArXiv e-prints, arXiv:1801.10382Cohn, J. H., Leja, J., Tran, K.-V. H., et al. 2018,ApJ, 869, 141Conroy, C., Gunn, J. E., & White, M. 2009, ApJ,699, 486Cowley, M. J., Spitler, L. R., Tran, K.-V. H.,et al. 2016, MNRAS, 457, 629de Barros, S., Vanzella, E., Amor´ın, R., et al.2016, A&A, 585, A51Elbaz, D., Dickinson, M., Hwang, H. S., et al.2011, A&A, 533, A119Ferland, G. J., Porter, R. L., van Hoof,P. A. M. andWilliams, R. J. R., et al. 2013,RMxAA, 49, 137Forrest, B., Tran, K.-V. H., Broussard, A., Allen,R. J., & et al. 2017, ApJL, 838, L12Forrest, B., Tran, K.-V. H., Broussard, A., et al.2018, ApJ, 863, 131Forrest, B., Tran, K.-V. H., Tomczak, A. R., et al.2016, ApJL, 818, L26F¨orster Schreiber, N. M., Shapley, A. E., Erb,D. K., et al. 2011, ApJ, 731, 65 Giacconi, R., Zirm, A., Wang, J., et al. 2002, TheAstrophysical Journal Supplement Series, 139,369Grogin, N. A., Kocevski, D. D., Faber, S. M.,et al. 2011, ApJS, 197, 35Guo, Y., Rafelski, M., Faber, S. M., et al. 2016,ApJ, 833, 37Ho, L. C., Filippenko, A. V., & Sargent, W. L. W.1997, ApJ, 487, 568Holden, B. P., Oesch, P. A., Gonz´alez, V. G.,et al. 2016, The Astrophysical Journal, 820, 73Izotov, Y. I., Guseva, N. G., Fricke, K. J., &Henkel, C. 2016, MNRAS, 462, 4427Izotov, Y. I., Guseva, N. G., & Thuan, T. X. 2011,The Astrophysical Journal, 728, 161Izotov, Y. I., Schaerer, D., Worseck, G., et al.2018, MNRAS, 474, 4514Jaskot, A. E., & Oey, M. S. 2013, ApJ, 766, 91Juneau, S., Dickinson, M., Alexander, D. M., &Salim, S. 2011, ApJ, 736, 104Kacprzak, G. G., van de Voort, F., Glazebrook,K., et al. 2016, ApJL, 826, L11Kennicutt, R. C. 1998, ARA&A, 36, 189Koekemoer, A. M., Faber, S. M., Ferguson, H. C.,et al. 2011, ApJS, 197, 36Kriek, M., van Dokkum, P. G., Labb´e, I., et al.2009, ApJ, 700, 221Kriek, M., van Dokkum, P. G., Whitaker, K. E.,et al. 2011, ApJ, 743, 168Krumholz, M. R., Burkhart, B., Forbes, J. C., &Crocker, R. M. 2017, ArXiv e-prints,arXiv:1706.00106Lawrence, A., Warren, S. J., Almaini, O., et al.2007, MNRAS, 379, 1599Leja, J., Johnson, B. D., Conroy, C., vanDokkum, P. G., & Byler, N. 2017, ApJ, 837, 170Lofthouse, E. K., Houghton, R. C. W., & Kaviraj,S. 2017, MNRAS, 471, 2311Maseda, M. V., van der Wel, A., da Cunha, E.,et al. 2013, The Astrophysical Journal, 778, L22Maseda, M. V., van der Wel, A., Rix, H.-W.,Momcheva, I., & et al. 2018, ApJ, 854, 29Maseda, M. V., van der Wel, A., Rix, H.-W., et al.2014, ApJ, 791, 175