arXiv:2106.02366v2 [astro-ph.GA] 18 Oct 2021

arX

iv2

106

0236

6v2

[as

tro-

phG

A]

18

Oct

202

1Draft version October 20 2021Preprint typeset using LATEX style emulateapj v 012315

PHYSICAL PROPERTIES OF MASSIVE COMPACT STARBURST GALAXIES WITH EXTREME OUTFLOWS

Serena Perrotta1⋆ Erin R George1 Alison L Coil1 Christy A Tremonti2 David S N Rupke3 Julie DDavis2 Aleksandar M Diamond-Stanic4 James E Geach5 Ryan C Hickox6 John Moustakas7 Grayson C

Petter6 Gregory H Rudnick8 Paul H Sell9 Cameren Swiggum2 Kelly E Whalen6

1Department of Astronomy University of California San Diego CA 92092 USA2Department of Astronomy University of Wisconsin-Madison Madison WI 53706 USA

3Department of Physics Rhodes College Memphis TN 38112 USA4Department of Physics and Astronomy Bates College Lewiston ME 04240 USA

5Centre for Astrophysics Research University of Hertfordshire Hatfield Hertfordshire AL10 9AB UK6Department of Physics and Astronomy Dartmouth College Hanover NH 03755 USA7Department of Physics and Astronomy Siena College Loudonville NY 12211 USA

8Department of Physics and Astronomy University of Kansas Lawrence KS 66045 USA9Department of Astronomy University of Florida Gainesville FL 32611 USA

(Accepted for publication in ApJ)Draft version October 20 2021

ABSTRACT

We present results on the nature of extreme ejective feedback episodes and the physical condi-tions of a population of massive (Mlowast sim 1011M⊙) compact starburst galaxies at z = 04 minus 07We use data from KeckNIRSPEC SDSS GeminiGMOS MMT and MagellanMagE to measurerest-frame optical and near-IR spectra of 14 starburst galaxies with extremely high star formationrate surface densities (mean ΣSFR sim 2000M⊙yr

minus1kpcminus2) and powerful galactic outflows (maximum

speeds v98 sim 1000 minus 3000 km sminus1) Our unique data set includes an ensemble of both emission([O II]λλ37263729 Hβ [O III]λλ49595007 Hα [N II]λλ6549 6585 and [S II]λλ67166731) and ab-sorption (Mg IIλλ27962803 and Fe IIλ2586) lines that allow us to investigate the kinematics of thecool gas phase (Tsim104 K) in the outflows Employing a suite of line ratio diagnostic diagrams wefind that the central starbursts are characterized by high electron densities (median ne sim 530 cmminus3)and high metallicity (solar or super-solar) We show that the outflows are most likely driven by stellarfeedback emerging from the extreme central starburst rather than by an AGN We also present mul-tiple intriguing observational signatures suggesting that these galaxies may have substantial Lymancontinuum (LyC) photon leakage including weak [S II] nebular emission lines Our results imply thatthese galaxies may be captured in a short-lived phase of extreme star formation and feedback wheremuch of their gas is violently blown out by powerful outflows that open up channels for LyC photonsto escapeKeywords galaxies active mdash galaxies evolution mdash galaxies interactions mdash galaxies starburst

1 INTRODUCTION

Starburst galaxies represent a fundamental phase ingalaxy evolution as they are widely considered to be thetransition stage between star-forming galaxies and mas-sive passively-evolving ellipticals (eg Cimatti et al2008) According to some theoretical scenarios (egDi Matteo et al 2007 Hopkins et al 2010) this transi-tion is initiated by highly dissipative major merger events(Sanders amp Mirabel 1996) producing strong bursts ofstar formation in very dense cores and possibly trig-gering obscured black hole accretion The starburstactivity and subsequent black hole feedback can causegas depletion and removal through powerful outflows(Sanders et al 1988 Silk amp Rees 1998) leading to apassively evolving system (Kormendy amp Sanders 1992Springel et al 2005 Hopkins et al 2008)Observations have revealed that outflows are a com-

mon characteristic of star-forming galaxies over a broadrange of masses and redshifts (eg Martin 1998Pettini et al 2001 Rubin et al 2010 Martin et al 2012Arribas et al 2014 Rubin et al 2014 Chisholm et al2015 Heckman et al 2015 Heckman amp Borthakur 2016

⋆s2perrottaucsdedu

McQuinn et al 2019) The incidence and propertiesof galactic outflows have been explored from z sim 0(Chen et al 2010) through z sim 05 minus 15 (Rubin et al2010 2011 2014) to the peak epoch of cosmic star for-mation at z sim 2 minus 3 with both absorption (Steidel et al2010) and emission lines (Strom et al 2017 2018) Whenmulti-band observations of the same system are avail-able they show that outflows are multi-phase hav-ing several co-spatial possibly kinematically coherentcomponents with a wide range in density and tem-perature (Heckman et al 2017) Galactic-scale out-flows can be identified through different phase tracerscold gas (eg lt 104 K) including molecules such asCO (Fluetsch et al 2019 Spilker et al 2020) as well asneutral H I and metals such as NaD (Heckman et al2000 Chen et al 2010 Martin 2005 Concas et al 2017Bae amp Woo 2018 Rupke 2018) Cool gas (ie sim 104

K) including metal ion tracers such as Fe II Mg II[O III] (Martin amp Bouche 2009 Rubin et al 2014)C II [N II] (Heckman et al 2015 Heckman amp Borthakur2016) Si II Si III Si IV and [O I] (Chisholm et al 2015Heckman et al 2015 Heckman amp Borthakur 2016) aswell as Hα (Shapiro et al 2009 Cicone et al 2016)Warm gas (ie sim 105 K - 106 K) can also be traced

2 Perrotta et al

by ionized metals such as N V O VI (Kacprzak et al2015 Nielsen et al 2017) Finally hot gas (iegt 106 K) probed with both hard and soft X-rayemission (Lehnert et al 1999 Strickland et al 2004Strickland amp Heckman 2007 2009)While powerful outflows appear to be essential to

rapidly shut off star formation the physical drivers ofthis ejective feedback remain unclear In particular therelative role of feedback from stars versus supermassiveblack holes (SMBHs) in quenching star formation in mas-sive galaxies is still widely debated (eg Hopkins et al2012 Gabor amp Bournaud 2014 Weinberger et al 2018Kroupa et al 2020) In this context the observed corre-lations between outflow and host galaxy properties canprovide some insight (Rubin et al 2014 Tanner et al2017) For instance considering galaxy samples with awide dynamic range of intrinsic properties the outflowvelocity is found to scale with stellar mass (Mlowast) starformation rate (SFR) and SFR surface density (ΣSFR)This suggests that the faster outflows tend to be hostedin massive galaxies with high and concentrated star for-mation (eg Tanner et al 2017) implying that the star-burst phase could potentially drive impactful outflowsStudying galaxies with extreme physical conditions canprovide constraints on astrophysical feedback processesOur team has been investigating a sample of galax-

ies at z = 04 - 08 initially selected from the SloanDigital Sky Survey (SDSS York et al 2000) Data Re-lease 8 (DR8 Aihara et al 2011) to have distinct sig-natures of young post-starburst galaxies Their spec-tra are characterized by strong stellar Balmer absorp-tion from B- and A-stars and weak or absent nebularemission lines indicating minimal on-going star forma-tion They lie on the massive end of the stellar massfunction (Mlowast sim 1011 M⊙ Diamond-Stanic et al 2012)Remarkably the optical spectra of most of these ob-jects exhibit evidence of ejective feedback traced by ex-tremely blueshifted (gt 1000 km sminus1) Mg II λλ27962803interstellar absorption lines (Tremonti et al 2007 Daviset al in prep) The Mg II kinematics imply galacticoutflows much faster than the sim500 km sminus1ones typicalof massive star-forming galaxies (Chisholm et al 2017)This finding painted an interesting picture where thesegalaxies were thought to be post-starburst systems withpowerful outflows that may have played a crucial rolein quenching their star formation Surprisingly manyof these galaxies were detected in the Wide-field In-frared Survey Explorer (WISE Wright et al 2010) andthe modeling of their ultraviolet (UV) to near-IR spec-tral energy distribution (SED) suggested a high levelof heavily obscured star formation (gt 50 M⊙ yrminus1Diamond-Stanic et al 2012) Hubble Space Telescope(HST) imaging of 29 of these galaxies revealed theyare extremely compact (Re sim few 100 pc) Moreoverthese data showed complex morphologies with diffusetidal features indicative of various major merger stages(Sell et al 2014 Diamond-Stanic et al 2021) Combin-ing SFR estimates from WISE restframe mid-IR lumi-nosities with physical size measures from HST imagingwe derived extraordinarily high ΣSFR sim 103 M⊙ yrminus1

kpcminus2 (Diamond-Stanic et al 2012) approaching thetheoretical Eddington limit (Lehnert amp Heckman 1996Meurer et al 1997 Murray et al 2005 Thompson et al2005)

These results led us to draw a new scenario where thesestarburst galaxies have a dense dusty star-forming coreat the center of the galaxy and a substantial part of theirgas and dust is blown away by powerful outflows In thiscontext the high ΣSFR may reasonably be the driver ofthe exceptionally fast gas outflows seen which in turnmay be responsible for the onset of rapid star formationquenching Millimeter data for two galaxies in our sam-ple indicates that the molecular gas is being consumedby the starburst with exceptional efficiency (Geach et al2013) and expelled in an extended molecular outflow(Geach et al 2014) leading to rapid gas depletion timesInterestingly Sell et al (2014) used a suite of multiwave-length observations to assess the AGN activity in a sub-sample of these starbursts and found little evidence forcurrent AGN activity in half of the sample (lt 10 percent of the total bolometric luminosity) though pastAGN episodes could not be ruled out This finding isin line with stellar feedback being the main driver of theobserved outflows These compact starburst galaxies ex-hibit the fastest outflows (gt 1000 km sminus1) and highestΣSFR among star-forming galaxies at any redshift there-fore they are an exquisite laboratory to test the limits ofstellar feedback They could represent a brief but com-mon phase of massive galaxy evolutionOur team followed up one of these starburst galax-

ies (J2118 or Makani) with Keck Cosmic Web Imager(KCWI Morrissey et al 2018) The data reveal a spec-tacular galactic outflow traced by [O II] emission linereaching far into the circumgalactic medium (CGM) ofthe galaxy (Rupke et al 2019) The [O II] emission hasa classic bipolar hourglass limb-brightened shape andexhibits a complex structure a larger-scale slower out-flow (sim300 km sminus1) and a smaller-scale faster outflow(sim1500 km sminus1) The velocities and sizes of these twooutflows map exactly to two previous starburst episodesthat this galaxy experienced detected through the rest-frame optical emission and inferred ages of stars in thisgalaxy These outflows are therefore consistent withbeing formed during recent starburst episodes in thisgalaxyrsquos past The KCWI data on Makani directly showsthat galactic outflows feed the CGM expelling gas farbeyond the stars in galaxiesIn this paper we present new optical and near-IR ob-

servations for 14 of the most well-studied starburst galax-ies in our sample We use this in combination with someancillary data to characterize their extreme ejective feed-back events and explore their potential role in quenchingthe star formation in the host systems Our unique dataset includes both emission and absorption lines that al-low us to probe outflowing gas at different densities Weinvestigate both the nature of the outflows as well asthe physical conditions in the central dusty starburstWe use an ensemble of line ratio diagrams as crucial di-agnostics of gas ionization electron density and metal-licityThe paper is organized as follows Section 2 illustrates

the sample selection observations and data reductionSection 3 describes our measurements of the emissionline kinematics Section 4 presents our main results incomparison to other relevant galaxy samples and Section5 discusses the more comprehensive implications of ouranalysis Our conclusions are reviewed in Section 6Throughout the paper we assume a standard ΛCDM

Massive Compact Starburst Galaxies 3

cosmology with H0 = 70 km sminus1Mpcminus1 Ωm = 03 andΩΛ = 07 All spectra are converted to vacuum wave-lengths and corrected for heliocentricity

2 SAMPLE AND DATA REDUCTION

The parent sample for this analysis has beendrawn from the SDSS as described by Tremonti et al(2007) Diamond-Stanic et al (2012) Sell et al (2014)Diamond-Stanic et al (2021) and Tremonti et al (inprep) In brief this sample contains 1198 galaxies at 035lt z lt 10 with i lt 20 mag from the SDSS DR8 withpost-starburst spectral features B- or A-star dominatedstellar continua and moderately weak nebular emissionA sub-sample of 1211198 galaxies with z gt 04 (suchthat the Mg IIλλ2796 2803 doublet is readily observ-able with optical spectrographs) has been the centerof comprehensive follow-up observations with the aimof constraining the physical mechanisms responsible forlaunching their energetic feedback More details aboutthe sample selection can be found in Davis et al (inprep) and Tremonti et al (in prep) We collectedground-based spectroscopy for 50 of these galaxies withthe MMTBlue Channel MagellanMagE KeckLRISKeckHIRES andor KeckKCWI (Tremonti et al 2007Diamond-Stanic et al 2012 Sell et al 2014 Geach et al2014 Diamond-Stanic et al 2016 Geach et al 2018Rupke et al 2019) X-ray imaging with Chandra for1250 targets (Sell et al 2014) radio continuum datawith the NSFrsquos Karl G Jansky Very Large Array(JVLAVLA) for 2050 objects (Petter et al 2020) andoptical imaging with HST for 2950 galaxies (ldquoHST sam-plerdquo Diamond-Stanic et al 2012 Sell et al 2014) Forthe HST observations we first focused on the 12 mostAGN-like galaxies and then on the 17 galaxies with theyoungest derived post-burst ages (tburst lt 300 Myr)yielding a sample of galaxies with bluer U-V colors andstronger emission lines than typically found in post-starburst samples We also used multi-band HST imag-ing to investigate the physical conditions at the cen-ters of the 1229 galaxies with the largest SFR surfacedensities measured by Diamond-Stanic et al (2012) (30M⊙ yrminus1 kpcminus2 lt ΣSFR lt 2000 M⊙ yrminus1 kpcminus2) andexplored the young compact starburst component thatmakes them so extreme (Diamond-Stanic et al 2021)In this paper we focus on 13 galaxies from the HST

sample (6 from the 1229 most AGN-like galaxies and 7from the 1729 with the youngest post-burst ages) plusone additional target J1622+3145 that shows clear signsof an outflow in its spectrum The targets in our sam-ple are listed in Table 1 along with some of their mainproperties (see Section 24)

21 NIRSPEC

Near-IR spectra were obtained for the 13 targets se-lected from the HST sample using the NIRSPEC cross-dispersed echelle spectrograph (McLean et al 1998) onthe Keck II telescope Observing dates were Septem-ber 15-17 2013 and January 16-17 2014 We used theNIRSPEC-1 filter covering 0947-1121 microm correspond-ing to the photometric Y band for the 11 sources at 045lt z lt 068 and the NIRSPEC-2 filter covering 1089-1293 microm for the 2 sources at z gt 068 All targets wereobserved with a 076 arcsec times 42 arcsecond slit with a

spectral resolution of R = λ∆λ asymp 2000 Individual ex-posures were 300 seconds with total integration times of40-60 minutes per object We used the standard ABBAslit-nodding approach We reduced the data using theREDSPEC IDL package (Kim et al 2015) The expo-sures were dark subtracted and flat-fielded using an in-ternal flat-field calibration lamp We subtracted pairsof AndashB exposures to perform sky subtraction We per-formed relative flux calibrations and telluric absorptioncorrections using spectra of standard stars observed thesame night We determined the absolute flux calibra-tion for the NIRSPEC spectra using the flux-calibratedMMT spectra available for each galaxy in our sample asdescribed in Section 23

22 GMOS

Five galaxies in our NIRSPEC sample (J0826+4305J0905+5759 J1506+5402 J1613+2834 andJ1713+2817) were also observed with Gemini Multi-Object Spectrographs (GMOS Allington-Smith et al2002 Hook et al 2004) on Gemini-North Here we usethe GMOS data covering the Hβ and [O III] spectralregion for these targets and the Hα one for J1613We include in our final sample one additional targetJ1622+3145 for which the GMOS spectrum covers theHα region and which shows unambiguous signs of anoutflowThe observations were carried out in service mode us-

ing Nod-and-Shuffle spanning 16 nights from March 042019 through April 23 2019 A series of 360-secondsexposures were taken for each target giving a total ex-posure time of sim 36 minutes The spectra were obtainedwith the NS075 arcsec long-slit the Hamamatsu detec-tor binned 2 times 2 and the R400 G5305 grating with aresulting spectral resolution of R asymp 1920 and wavelengthrange from sim 036 to 103 microm We adopted 0745 0770or 0810 microm as central grating wavelength according tothe redshift of the source and spectrally dithered eachpointing by plusmn 001 microm This allows contiguous wave-length coverage in the presence of chip gaps and badcolumns on the detectorThe data were reduced using the GMOS sub-

package in the Gemini PyRAF software package (v114Labrie et al 2019) Briefly the data were bias sub-tracted and flat-fielded The sky subtraction was per-formed subtracting the two shuffled sections of the detec-tor The GMOS data were then wavelength calibratedextracted and stacked Relative flux calibration and tel-luric absorption correction were applied to the spectrabased on standard stars observed at a similar airmassas the targets We determined the absolute flux calibra-tion of the GMOS data using the flux calibrated MMTspectra described in Section 23

23 Other Optical Spectra

We obtained the rest-frame UVndashoptical spectra ofJ1341 and J1107 with the Magellan Echellette (MagE)spectrograph (Marshall et al 2008) on the MagellanClay telescope with a 1 arcsec slit and 2 hours of integra-tion time The data were reduced and calibrated usingthe MASE pipeline (Bochanski et al 2009) The spectrahave a resolution R sim 4100 over a bandpass of 3300minus9400A in 15 orders (λrest sim 2300minus 6000 A) and a signal-to-noise ratio (SNR) of sim 45 per resolution element near

4 Perrotta et al

0

10

20

30

40 J0106minus1023

0

10

20

30J1341minus0321

0

10

20 J0826minus4305

0510152025J1506+5402

0

10

20

30J0901+0314

0510152025J1613+2834

0

10

20

Flu

x(1

0minus

17

erg

sminus1

cmminus

2Aminus

1)

J0905+5759

0510152025J1622+3145

0

5

10

15 J0944+0930

0

10

20J1713+2817

05

10152025 J1107+0417

0510152025J2116minus0634

3000 3500 4000 4500 5000

05

10152025 J1125minus0145

3000 3500 4000 4500 5000

0510152025J2118+0017

Rest Wavelength (A)

Figure 1 Rest-frame near-UV and optical spectra of the 14 galaxies in our sample The black line shows the combined MMT MagEand SDSS or GMOS spectra (joined between 4500 and 4700 A or 3500 and 3800 A) The red line represents the continuum model fitoffset in the vertical direction for clarity errors from the best fit model are shown in blue The continuum model is subtracted from eachspectrum before measuring the nebular emission lines of [O II]λ3726 Hβ and [O III]λ5007 The spectra are dominated by the light of ayoung stellar population but have relatively weak nebular emission lines and strong Mg II λλ27962803 absorption originating from theinterstellar medium

the galaxyrsquos Mg IIλλ27962803 absorption lines For allthe other galaxies in our sample we collected high SNRoptical spectra with the Blue Channel spectrograph onthe 65m MMT between 2004 December and 2009 July(Tremonti et al 2007) The data were obtained using a1 arcsec long slit which produced a FWHM resolutionof 36 A (R sim 2000 near Hβ) The total exposure timefor each target was sim 45-90 min For our z = 04 minus 08galaxies this yielded rest-frame coverage from sim 2700to 3900 A The data were reduced extracted and spec-trophotometrically calibrated using the ISPEC2D datareduction package (Moustakas amp Kennicutt 2006)There is extremely good agreement between the MMT

MagE SDSS and GMOS spectra where they overlapWe join the MMT MagE SDSS and GMOS spectrawhen available in order to extend our spectral coverageThe combined spectra including the stellar continuum

fits are shown in Fig 1 The systemic redshifts usedthroughout the paper are defined by the starlightThe continuum model is built as described in

Geach et al (2018) In brief we fit the spectrum witha combination of simple stellar population (SSP) mod-els and a Calzetti et al (2000) reddening law We em-ployed the Flexible Stellar Population Synthesis code(Conroy et al 2009 Conroy amp Gunn 2010) to generateSSPs with Padova 2008 isochrones a Salpeter (1955)initial mass function (IMF) and a theoretical stellarlibrary ldquoC3Krdquo (Conroy et al 2018) with a resolutionof R sim 10000 We utilize solar metallicity SSP tem-plates with 43 ages spanning 1 Myrminus89 Gyr We per-form the fit with the Penalized Pixel-Fitting (pPXF)software (Cappellari amp Emsellem 2004 Cappellari 2017)We mask forbidden emission lines and implement twoseparate templates for broad and narrow Balmer emis-


Table 1Sample properties

ID z RA Dec log(MlowastM⊙) re SFR ΣSFR Mg II Velocity(J2000) (J2000) (kpc) (M⊙ yrminus1) (M⊙ yrminus1 kpcminus2) (km sminus1)

(1) (2) (3) (4) (5) (6) (7) (8) (9)

J0106-1023 045 16601056 -10391647 1072 0590 166+33minus31 76 -1650

J0826+4305 060 12666006 43091498 1063 0173 184+53minus41 981 -1425

J0901+0314 046 13538926 32367997 1066 0237 99+39minus26 281 -1602

J0905+5759 071 13634832 57986791 1069 0097 90+23minus20 1519 -2910

J0944+0930 051 14607437 95053855 1059 0114 88+26minus21 1074 -1679

J1107+0417 047 16676197 42840984 1060 0273 73+13minus14 155 -2093

J1125-0145 052 17132874 -17590066 1103 0600 227+104minus68 100 -2309

J1341-0321 066 20540333 -33570199 1053 0117 151+34minus23 1756 -1936

J1506+5402 061 22665124 54039095 1060 0168 116+32minus25 652 -2018

J1613+2834 045 24338552 28570772 1112 0949 172+36minus36 30 -2699

J1622+3145 044 24569628 31759132 1062 middot middot middot 151+52minus33 middot middot middot -1713

J1713+2817 058 25825161 28285631 1089 0173 229+99minus72 1218 -1298

J2116-0624 073 31910479 -65791139 1041 0284 110+55minus27 216 -2069

J2118+0017 046 31960026 02915070 1095 2240 230+93minus76 5 -1448

Note mdash ndash Column 5 Stellar mass from Prospector Column 6 Effective radii from HST Column 7SFRs from Prospector Column 8 SFR surface densities estimated using columns (6) and (7) Column 9Mg IIλ2796 A maximum velocity v98

sion lines assuming Case B recombination line ratiosBoth line and continuum are attenuated by the sameamount of dust in the pPXF fit By fitting Balmer emis-sion and absorption lines simultaneously we can take intoaccount the potential infill of the absorption line coresOne of the outputs of our pPXF fit is the stellar con-tinuum model without any nebular component (shownin Fig 1) We subtract from each spectrum our best fitpPXF model to properly remove the stellar componentMost sources in addition to having strong Balmer ab-

sorption show very blue continua indicating a recentstarburst event (sim 1minus10 Myr) that is not highly dustobscured These galaxies have morphologies of late-stagemajor mergers (Sell et al 2014) which are consistentwith having recent or on-going bursts of star formationThe MMTMagE spectra allow high SNR measurementsof the Mg IIλλ27962803 interstellar medium (ISM) linesused to search for signs of outflowing gas Mg II absorp-tion lines are detected in all sources in our sample withblueshifts with respect to the systemic redshift rangingfrom 1400 to 2900 km sminus1 Tremonti et al (2007) high-light the fact that these outflows are a factor of 2minus5 timesfaster than the outflow velocities of typical IR-luminousstar-forming galaxies (LIRGs and ULIRGs eg (LIRGsand ULIRGs eg Martin 2005 Rupke et al 2005) Wereturn to this point below in Section 4

24 Galaxy properties

Table 1 lists various relevant galaxy properties derivedfor sources in our sample Stellar mass (Mlowast) and starformation rate (SFR) estimates are derived by fitting thebroad-band UV ndash mid-IR photometry and spectra withthe Bayesian SED modelling code Prospector (Leja et al2019 Johnson et al 2021) as described in Davis et al(in prep) In brief we include the 3500 - 4200 A spec-tral region in the fit since it contains many age-sensitivefeatures (eg D4000 Hδ) and has a robust spectropho-tometric calibration SSP models are generated utilizingthe Flexible Stellar Populations Synthesis code (FSPS

Conroy et al 2009) assuming a Kroupa IMF (Kroupa2001) and adopting the MIST isochrones (Choi et al2016) and the C3K stellar theoretical libraries (Conroyet al in prep) The stellar models are very similar tothe ones described in Section 23 over the wavelengthrange of interest for this work The best fit parametersand their errors are calculated from the 16th 50th and84th percentiles of the marginalized probability distribu-tion function See Davis et al (in prep) for examplesof the SED fitting The models fit the combined pho-tometry and spectra well however the lower SNR WISEW3 and W4 photometry and the limited infrared cov-erage of the SED provide poor constraints on the dustemission properties This yields fairly tight constraintson the Mlowast (plusmn015 dex) and slightly larger errors on theSFR (plusmn02 dex) Mlowast represents the present day stel-lar mass of the galaxy and not the integral of the starformation history In this work we utilize SFRs com-puted from each galaxyrsquos star formation history averag-ing over 100 Myr timescales This is the characteristictimescale UV or IR star formation indicators are sensi-tive to (Kennicutt amp Evans 2012)Measurements of the effective radii (re) for galaxies

in our sample are discussed in Diamond-Stanic et al(2012) Sell et al (2014) Diamond-Stanic et al (2021)Briefly for 3 galaxies (J0106 J1125 and J1713) we quan-tify the morphology using optical HST UVISF814Wimages We employ GALFIT (Peng et al 2002 2010)to model the two-dimensional surface brightness profilewith a single Sersic component (defined by Sersic indexn=4 and re) adopting an empirical model point-spreadfunction (PSF) built using moderately bright stars in ourscience images For the remaining 10 galaxies with multi-band imaging (Diamond-Stanic et al 2021) we performSerscic fits to the UVISF814W and UVISF475W im-ages jointly using the GALFITM software (Hauszligler et al2013 Vika et al 2013) To avoid uncertainties producedby tidal features we fit the central region of the galaxyand extrapolate the fit to larger radii to compute re

6 Perrotta et al

Table 2Best Fit Parameters

Hα Hα Hα [O II] [O II] [O II]ID Narrow FWHM Broad FWHM voff Narrow FWHM Broad FWHM voff

(km sminus1) (km sminus1) (km sminus1) (km sminus1) (km sminus1) (km sminus1)(1) (2) (3) (4) (5) (6) (7)

J0106-1023 525 plusmn 43 middot middot middot middot middot middot 829 plusmn 39 middot middot middot middot middot middot

J0826+4305 313 plusmn 33 918 plusmn 81 -290 plusmn 56 414 plusmn 53 1761 plusmn 263 -680 plusmn 171J0901+0314 410 plusmn 42 middot middot middot middot middot middot 811 plusmn 30 middot middot middot middot middot middot

J0905+5759 294dagger plusmn 34 798dagger plusmn 56 -80dagger plusmn 16 462 plusmn 77 1139 plusmn 175 -380 plusmn 167J0944+0930 434 plusmn 61 1011 plusmn 345 -67 plusmn 13 326 plusmn 128 925 plusmn 258 -393 plusmn 249J1107+0417 481 plusmn 70 1985 plusmn 169 -43 plusmn 9 451 plusmn 61 1534 plusmn 242 20 plusmn 8J1125-0145 386 plusmn 43 middot middot middot middot middot middot 417 plusmn 108 2396 plusmn 398 -468 plusmn 174J1341-0321 483 plusmn 35 1318 plusmn 132 -205 plusmn 35 141 plusmn 29 1450 plusmn 25 -262 plusmn 11J1506+5402 358 plusmn 36 1218 plusmn 58 -143 plusmn 25 523 plusmn 31 2058 plusmn 288 -474 plusmn 158J1613+2834 397 plusmn 56 1237 plusmn 65 -257 plusmn 79 617 plusmn 25 1710 plusmn 68 -308 plusmn 37J1622+3145 482 plusmn 48 1071 plusmn 185 -102 plusmn 37 415 plusmn 102 middot middot middot middot middot middot

J1713+2817 521 plusmn 45 middot middot middot middot middot middot 357 plusmn 78 1221 plusmn 551 -577 plusmn 325J2116-0624 112 plusmn 48 631 plusmn 85 15 plusmn 9 223 plusmn 89 1607 plusmn 420 -245 plusmn 173J2118+0017 281 plusmn 31 825 plusmn 45 -231 plusmn 77 421 plusmn 42 1501 plusmn 84 -341 plusmn 51

Note mdash ndash Column 2-3 FWHMs of narrow and broad Hα emission line components from NIR-SPEC or GMOS spectra corrected for instrumental resolution Column 4 velocity offset comparedto systemic redshift of the broad Hα component Column 5-6 FWHMs of narrow and broad [O II]emission line components from MMT MagE or SDSS spectra corrected for instrumental resolutionColumn 7 velocity offset compared to systemic redshift of the broad [O II] component dagger We reportvalues from the Hβ emission line fit for J0905

The HST filters probe relatively blue (λrest(F475W) asymp

3000A λrest(F814W)asymp 5200A) emission at z sim 06 trac-ing the young unobscured stars rather than the stellarmass Typical errors on the effective radius are of theorder of 20 We do not have information on re for onegalaxy J1622We also report maximum outflow velocities derived

from the Mg IIλλ27962803 absorption lines observed inMMT spectra which show intricate velocity structuresWe use VPFIT (v104 Carswell amp Webb 2014) to fit thedoublet absorption profiles using a number of Voigt func-tions from one to six depending on the complexity of thelines We parameterize the kinematics of Mg II consid-ering only one of the doublet components and measurethe line velocity shift relative to the systemic redshiftat which 98 (v98) of the equivalent width (EW) ac-cumulates moving from red (positive velocities) to blue(negative velocities) across the line profile The derivedvalues in our sample range from -1400 to -2900 km sminus1To assess errors on v98 due to uncertainties in the fits weassume the best-fitting parameters are uncorrelated andvary them in a range of plusmn1σ and measure the resultingchange in v98 We use the largest variation of v98 as up-per limit error with typical values of 200minus400 km sminus1forour sample

3 EMISSION LINE FITTING

We quantify the kinematics of several diagnostic emis-sion lines [O II]λλ37263729 Hβ [O III]λλ49595007Hα [N II]λλ6549 6585 and [S II]λλ67166731 for eachgalaxy in our sample as follows After subtracting thebest-fitting stellar population model of the galaxy (seeSection 23) the residual emission lines are fit using acustom Python algorithm We model each emission linewith one or two Gaussian functions according to thecomplexity of the emission profiles and the SNR A sec-ond Gaussian component is added only if the improve-

ment in χ2 is statistically significant accounting for theadditional free parameters Broadened or shifted emis-sion line components trace gas with different kinematicsfrom the rest of the ionized gas in the galaxy Such com-ponents potentially trace outflowing gasThe multicomponent fits to the nebular emission lines

for the galaxies in our sample are shown in Fig 2 Thevarious emission lines are not fit simultaneously since thedata sets have different resolutions and SNR Moreoverthe lines span a broad range in wavelength and extinc-tion might impact them differently The MMTMagEdata cover the [O II] doublet spectral region We assumethe [O II] doublet lines have identical kinematics (iesame velocity widths and shifts in the Gaussian fit com-ponents) We set the flux ratio [O II]λ3729[O II]λ3726to 1005 as the spectra do not have sufficient resolutionto fit them separately We fix the [O II] ratio to reflectthe typical electron density of the ISM in our sources asestimated using the [S II] emission lines (see Section 42Sanders et al 2016) The [O II] lines generally requiretwo Gaussian components to fit their asymmetric pro-files The only exceptions are J0106 J0901 and J1622The Hβ and [O III] spectral region is covered by the

SDSS data for 814 galaxies in our sample and by theGMOS data for the remaining 614 galaxies (see Section22) As in the case of the [O II] we adopt the samekinematics for the [O III] doublet lines and we fix theiramplitude ratio [O III]λ4959[O III]λ5007 to 0337 tomatch the transition strengths (Storey amp Zeippen 2000)While we allow the Hβ profile to have a different kine-matic structure than that of [O III] we find consistentresults between the line in terms of velocity widths andcentroids of the narrow and broad components The lowSNR prevents us from performing a reliable fit of theselines for J1125 and J2116 Both Hβ and [O III] are welldescribed by one Gaussian in 3 galaxies (J0106 J0901and J1713) and by two Gaussians in the remaining 9


0

4

8J0106 M

[OII]

02

4

6 S

Hβ

01234 S

[OIII]

0

10

20 N

Hα

0

1

2

N

[SII]

0246 J0826 M

0

2

4G

0123 G

0

10

20 N

0123 N

0246 J0901 M

0

4

8S

01234 S

0102030

N

0

2

4J0905 M

01234 G

0

1

2 G

0

10

20N

0123 J0944 M

0

2

4S

0

1

2S

05

1015 N

0

2 N

0

2

4 J1107 Ma

0246 Ma

01234

Ma

0

10

20 N

0

2N

0123

Flu

x(1

0minus

17

erg

sminus1

cmminus

2Aminus

1)

J1125 M

0

5

10 N

0

2

4

6J1341 Ma

0246 Ma

0

1

2

3Ma

0

10

20N

0

2N

02468 J1506 M

0

5

10G

0

4

8 G

0102030

N

0

2

4N

0

5

10

15 J1613 M

0

2

4 G

0

1

2 G

05

1015

G

0

2G

02468 J1622 S

0246 G

0

1

2 G

05

1015 G

0

1

2 G

0123

J1713 M

0

1S

0

5

10 S

0123 N

0123 J2116 M

0

4

8 N

0

2N

3700 3720 3740

0

5

10

15J2118 M

4840 4860 4880

0246 S

4950 4990 5030

Rest Wavelength (A)

0246 S

6540 6570 6600

0

10

20 N

6690 6720 6750

0

2N

Figure 2 Fits to the nebular emission lines in the fourteen galaxies in our sample Each row represents one object and each columnfrom left to right is [O II]λλ37263729 Hβ [O III]λλ49595007 the Hα+[N II]λλ65496585 blend and [S II]λλ67176731 The grey lettersrepresent the instrument used to obtain each spectrum MMT (M) MagellanMagE (Ma) GeminiGMOS (G) KeckNIRSPEC (N) orSDSS (S) The purple solid line shows the best fit to each emission line the light blue and pink ones refer to the narrow and broad Gaussiancomponents of the fit respectively We include a broad component when it improves the reduced χ2 of the fit significantly The errorspectrum is shown as a dotted green line Spectra are omitted where the SNR is too low to identify any significant emission line Theidentification of broad emission is indicative of outflowing material and since the broad emission is seen in the forbidden lines this suggeststhat the outflow originates from the ISM (rather than any hidden AGN broad-line region)

8 Perrotta et al

galaxiesFinally we use the NIRSPEC data to fit the Hα [N II]

and [S II] emission lines for 1214 galaxies in our sampleand the GMOS data for J1613 and J1622 All the emis-sion lines in this spectral region are forced to have thesame kinematics (velocity offsets and widths) while theamplitude of each component is allowed to vary indepen-dently This choice is justified by the complex emissionline profiles of Hα and [N II] that blend together and bythe low SNR of the [S II] lines of the spectra in our sam-ple We do not fix the [N II] doublet flux ratio to be 13as the [N II] λ6549 line for some of our galaxies falls at theedge of the NIRSPEC bandpass where the spectra havehigher fluxing errors However we find the [N II] doubletflux ratio to be very close to the theoretical value in mostcases with a mean value of 038 We also perform fitsfixing the [N II] doublet ratio to 13 and find that thekinematics and fluxes of the Hα and [N II] emission lineschange by lt10 The broad [N II] doublet ratio is setto be the same as the narrow [N II] doublet ratio Theratio of the density-sensitive [S II] doublet is allowed tovary but it is restricted to be within 20 of the rangeof permitted values (043minus15 Tayal amp Zatsarinny 2010Mendoza amp Bautista 2014) The Hα and [N II] kinemat-ics are well parameterized by a single Gaussian in 514galaxies (J0106 J0901 J0905 J1125 and J1713) and bytwo Gaussian components in the remaining 914 galax-ies Although we force [S II] to have the same kinematicsas Hα and [N II] we are not able to fit a broad [S II]component in any of the galaxies where it would be ex-pected (from Hα) due to the low SNR except for J1613and J2118 Moreover the low SNR prevent us from per-forming a reliable fit of the [S II] doublet in four galaxiesin our sample (J0901 J0905 J1125 and J1713) We alsoperform a fit of the [S II] doublet lines not constrained bythe Hα and [N II] kinematics We obtain similar resultsbut with larger uncertainties due to a larger number offree parametersThree of the galaxies have slight modifications to the

fitting procedure 1) J0905 is an unusual source thatshows narrow redshifted Hα + [N II] components theseoffset features are fit separately using narrow Gaussianprofiles with the same kinematics and are excluded fromfurther analysis (marked in red in Fig 2) and 2) the[O III] kinematics for J0944 and J2118 are tied to theHβ kinematics due to the low SNR around the doubletemission linesWe correct all the emission line fluxes for dust extinc-

tion by comparing the Balmer decrement (HαHβ) withthe expected Case B value of 286 (Osterbrock 1989)Galaxies with Balmer decrements lt 286 (but consistentwith 286 within the uncertainties) are assumed to havezero extinction We adopt the Galactic extinction curvefrom Cardelli et al (1989) for galaxies with HαHβ ge

286 the interquartile range for extinction in our sampleis E(B-V) = 018minus070 with a median value of 036Table 2 lists the full widths at half-maximum (FWHM)

corrected for instrumental resolution of both the narrowand broad Gaussian components of our spectral fits forthe Hα and [O II] emission lines We also report thevelocity offset (voff ) of the broad component centroidswith respect to the systemic redshift The 1σ errors onall measurements account for uncertainties in the fit pa-rameters as well as covariance between parameters

4 RESULTS

The following sections collect the results of this workThe main goal is to characterize the physical conditionsof the starburst at the center of the galaxies in our sam-ple that is driving powerful outflows We first investi-gate the kinematics of a suite of emission and absorptionlines probing different scales of the same ionized outflow-ing gas Then we exploit an ensemble of emission lineratio diagnostics to derive quantities that regulate theemission of the H II regions like electron density metal-licity and ionization parameter Lastly we compare ourfindings with those of relevant comparison samples

41 Kinematics

The high SNR of the spectra employed in this studyprovides the unique opportunity of being able to measurethe kinematics of [O II] [O III] Hβ and Hα emission linesindependently In Fig 2 we present the various observedemission lines and best fit line results for the fourteengalaxies in our sample Although the nebular emissionlines are fit separately their line profile decompositionsin narrow and broad components agree in 1014 galax-ies Two of the remaining cases (J1125 and J1713) havethe lowest SNR spectra covering [O III] Hβ and Hα inour sample Both galaxies have [O II] that clearly ex-hibits a broad and asymmetrical line profile Howeverwe do not include a broad component to other emissionlines observed in these sources because the reduced χ2

of their fits do not improve significantly In the case ofJ1622 the [O II] kinematics are well described by nar-row lines only while the [O III] Hβ and Hα fits requirea broad component Lastly in J0905 we fit Hα usinga single narrow Gaussian while [O II] [O III] and Hβneed an additional broad line (we note however thatHα appears to have a secondary component which maypotentially be part of a broad line) We note that in allcases where a broad component is required for the bestfit the centroid of the broad component is blueshiftedrelative to that of the narrow component We quantifythe nebular emission line kinematics measured from ourspectral fits using the FWHM and voff of each compo-nent In Table 2 we report these values for Hα and [O II]only as [O III] and Hβ exhibit kinematics that are verysimilar to Hα andor [O II]Fig 3 shows a comparison of the best spectral fits for

a suite of emission and absorption lines for each galaxyin our sample Each velocity profile is first normalizedto its own emission or absorption line peak to facilitatecomparison The narrow Hα component is shown as adot-dashed magenta line in each panel and traces the sys-temic redshift of the galaxy the rest of the emission linecomponents shown are broad We note that the broad[O II] components (light blue solid line) are systemati-cally wider than the Hα broad components (pink solidline) with the exception of J0944 and J1107 The meanvalues of the broad FWHM for [O II] and Hα in our sam-ple are 1573 and 1101 km sminus1 respectively Moreover[O II] shows larger blueshifts than Hα except in sourceJ1107 The mean values of voff for [O II] and Hα are352 and 143 km sminus1 respectivelySuch line broadenings and blue velocity shifts clearly

identify outflowing gas We note that often the broadcomponents contain some redshifted gas as well com-


-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0106

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0826

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0901

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0905

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0944

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1107

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1125

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1341

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1506

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1613

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1622

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1713

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2116

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2118

Narrow Hα

Broad Hα or Hβ

Broad [OIII] 5007

Broad [OII] 3729

MgII 2796

FeII 2586

Velocity [kmsminus1]

Norm

aliz

edF

lux

Figure 3 Comparison of velocity profile fits among selected emission and absorption lines for the galaxies in our sample All profilefits are normalized to their emission or absorption flux peak to facilitate comparison The narrow Hα emission line fit is displayed as adot-dashed magenta line in each panel and represents the systemic redshift in agreement with the redshift derived by the starlight (seeSection 23) Different outflowing gas tracers are shown as different color solid lines Broad Hβ is shown for J0905 and J1622 In J2118Mg II emission is observed which obscures any underlying Mg II λ2796 absorption feature therefore we present Fe II λ2586 instead forthis galaxy using KCWI data The emission line velocity profiles show remarkable overall consistency except for [O II] λ3729 which tendsto be more blueshifted compared to systemic in several sources Emission and absorption lines probe different spatial scales of the samegas phase and exhibit comparable maximum outflowing velocities in most of the galaxies in our sample

pared to the narrow line profiles The presence of ablueshift in the velocity centroid of the broad compo-nents is attributed to dust present in the host galaxythat obscures part of the redshifted outflows We notethat their SED fitting suggests a mean attenuation ofAV sim 043 (Tremonti et al in prep) We come back tothis point in Section 51The left panel of Fig 4 compares the [O II] and Hα

broad emission line kinematics as represented by v98which is an estimate of the maximum observed outflowspeed (and is a lower limit to the actual maximum speedif the gas producing the blueshifted line wings is not mov-ing directly towards the observer) The [O II] maximumvelocity is roughly 450 km sminus1greater than that of Hαalthough their kinematics are consistent for a few galax-ies

Fig 3 compares the [O II] emission line kinematics tofits of the Mg IIλ2796 absorption lines for each galaxyMg II exhibits complex velocity profiles in our sourceswith a mean value of v98 of minus1890 km sminus1 Such largeblueshifts clearly identify outflowing gas observed in ab-sorption In the case of J2118 we do not detect Mg IIabsorption and show the fit results to Fe IIλ2586 insteadThe lack of Mg II absorption in this galaxy is most likelydue to the detected Mg II emission which fills the un-derlying absorption trough We note that 914 galaxiesin our sample have less than 5 of the Mg II EW within50 km sminus1of the systemic redshift While Mg II emissionline filling may be present for our sources it should notsubstantially affect our maximum velocity as v98 is typi-cally far greater than the velocity of Mg II when observedin emission We will present results on Mg II emission

10 Perrotta et al

using high resolution spectra in an upcoming paper (Per-rotta et al in prep) We explore the possible reasons forthe lack of Mg II absorption near the systematic velocitybelow in Section 54The various ions studied here probe the same cool gas

phase (T sim 104 K) However they could originate on dif-ferent spatial scales and their physical properties couldspan a wide range of values Most importantly emissionand absorption lines provide us different approaches tostudy outflowing gas We return to this point in Sec-tion 51

42 Electron Density

The electron density (ne) of the ISM is one of the mainphysical quantities that govern the emission of H II re-gions The nebular emission-line ratios and derived quan-tities such as the gas-phase metallicity and ionizationparameter probe the physical conditions in the centralstarburst and depend critically on measuring neThe electron density can be estimated from the ratio of

the [S II]λλ67166731 doublet The collisionally-excitedforbidden lines are produced in low density gas wherethe low number of collisions prevents the de-excitation ofthe excited state Between the low density ( 10 cmminus3)and high density (amp 104 cmminus3) regimes this ratio providesa good measurement of the nebular gas density (egOsterbrock amp Ferland 2006)We employ the diagnostic relation from Sanders et al

(2016) which assumes an electron temperature of Te

= 104 K For the two galaxies (J1613 and J2118) inour sample where the SNR is high enough to decom-pose the emission line profiles into separate narrow andbroad components we use the [S II]λ6716[S II]λ6731narrow line ratio For the rest of the sample we use the[S II]λ6716[S II]λ6731 total flux ratio The results areshown in Fig 5 The errors on each density measure-ment are determined by converting the upper and lower68th percentile uncertainties from the [S II] constrained(solid line) and unconstrained (dotted line) fits on theline ratio into electron densities The derived [S II] dou-blet ratios range from 062 to 135 which correspond toan ne range from 68 cmminus3 to 2750 cmminus3 The median nevalue across the full sample is 530 cmminus3 This densityrange is substantially elevated with respect to typicalH II regions in the local universe which generally havene sim 50minus100cmminus3 (eg Zaritsky et al 1994a)The higher average electron densities we find in our

galaxy sample are consistent with the characteristicelectron densities observed in high redshift galaxieswhich have values that are 5minus10 times higher thanzsim0 galaxies with typical ne values of asymp 200minus400cmminus3

at zsim2minus3 (eg Masters et al 2014 Steidel et al 2014Sanders et al 2016 Strom et al 2017) However obser-vations of some individual galaxies at zsim2 suggest neof sim 103 cmminus3 (Hainline et al 2009 Lehnert et al 2009Quider et al 2009 Bian et al 2010 Shirazi et al 2014)The high electron density implies the compact size of theH II regions If these high-z H II regions follow the similarne-size relation found in the local galaxies (Kim amp Koo2001) their sizes should be less than 1 pc We discusshow elevated ne values can affect the emission line pro-duction below in Section 53

43 BPT Diagnostic Diagrams

Line ratios diagrams can be employed to distin-guish between sources of ionizing radiation in emissionline galaxies Following the work by Baldwin et al(1981) Veilleux amp Osterbrock (1987) introduced thewidely-used diagnostic diagrams commonly referred toas BPT diagrams We consider the [O III]λ5007Hβvs [N II]λ6585Hα (N2-BPT) and [O III]λ5007Hβ vs[S II]λλ67176731Hα (S2-BPT) diagrams to character-ize the galaxies in our sampleFig 6 shows the N2- and S2-BPT diagrams along

with empirical and theoretical lines dividing galaxiesexcited by different mechanisms Star forming galax-ies occupy well defined regions in these diagrams Inparticular as metallicity increases the sequence of starforming galaxies in the N2-BPT space elongates fromhigh values of [O III]λ5007Hβ and low [N II]λ6585Hαand curves down to low [O III]λ5007Hβ and high[N II]λ6585Hα Moreover galaxy stellar mass increasesalong this sequence due to the correlation between stel-lar mass and gas-phase metallicity in star forming galax-ies (Tremonti et al 2004) The empirical lines divid-ing star-forming galaxies and AGN-hosted galaxies de-rived from SDSS are shown in Fig 6 as green dashedlines (Kauffmann et al 2003) and the theoretical ex-treme starburst lines determined from photoionizationand radiation transfer models are shown as blue dashedlines (Kewley et al 2001) The red and orange dashedlines represent the empirical lines separating LINER andSeyfert galaxies in the N2-BPT and S2-BPT planes asderived by Cid Fernandes et al (2010) and Kewley et al(2006) We assemble a comparison sample from theSDSS DR8 selecting galaxies within the redshift range0005lt z lt 01 to reduce aperture effects and requir-ing 3σ detection in the rest-frame optical emission linesfeatured in each diagnostic diagram Emission line mea-surements and ancillary physical parameters are drawnfrom the MPA-JHU catalog for SDSS DR81 The greycontours enclose the 30 50 70 90 and 99 ofSDSS galaxiesFig 6 shows the locations of our galaxies in the N2-

(left) and S2-BPT (right) diagrams where the top rowuses line ratios determined from the total line flux andthe bottom row shows line ratios determined from thenarrow line components onlyThe galaxies in our sample fall in or near the ldquocompos-

iterdquo region in the N2-BPT diagram with the exceptionof J1713 which is a candidate type II AGN (Sell et al2014) Comparing the line ratios determined from thetotal line flux versus the narrow line flux we find thatthere is not a bulk shift in the [N II]λ6585Hα valueswhile the [O III]λ5007 to Hβ total flux ratio in all casesexcept one (J0826) is systematically higher than the cor-responding narrow line flux ratioWe discuss in Section 52 possible AGN contribution

to the line ratiosInterestingly most galaxies in our sample exhibit

[S II]λλ67176731Hα values that are lower than normalstar forming galaxies with 59 targets having lower total[S II] to Hα ratios than 99 of SDSS galaxies We dis-cuss in Sections 53 and 54 the possible causes of suchlow [S II] to Hα ratios The S2-BPT diagram for thenarrow flux component (bottom right panel) includes the

1 Available at httpswwwsdssorgdr12spectrogalaxy_mpajhu


minus3000minus2500minus2000minus1500minus1000minus5000

Broad Hα v98 [kmsminus1]

minus3000

minus2500

minus2000

minus1500

minus1000

minus500

0

Bro

ad[O

II]v

98[k

msminus

1]


MgII v98 [kmsminus1]

J0106

J0826

J0901

J0905

J0944

J1107

J1125

J1341

J1506

J1613

J1622

J1713

J2116

J2118

Figure 4 Broad [O II] emission line kinematics compared to broad Hα emission line (left) and Mg II absorption line (right) ones asrepresented by the maximum measured velocity v98 Errors on v98 due to uncertainties in the fits are estimated varying the best-fitparameters in a range of plusmn1σ and measuring the resulting change in v98 The dotted lines represent the 1 to 1 relation The galaxies thathave no broad [O II] or Hα emission lines detected are shown as empty squares For J0905 v98 is derived from the Hβ broad emission lineinstead of the Hα For J2118 v98 is derived from the Fe II λ2586 absorption line profile instead of the Mg II λ2796 since Mg II absorptionis not detected for this galaxy Most of the objects in our sample exhibit broad [O II] maximum velocities comparable to those derivedfrom the broad Hα and Mg II absorption lines

J0106J0826

J0944J1107

J1341J1506

J1613J1622

J2116J2118

100

101

102

103

104

Ele

ctro

nD

ensi

ty[c

mminus

3]

Figure 5 Electron densities calculated following the method de-scribed by Sanders et al (2016) using narrow [S II]λ6716λ6731doublet ratio Errors on individual density measurements are es-timated by converting the upper and lower 68th percentile uncer-tainties on the line ratio into electron densities Solid error barsrepresent the errors derived using the uncertainties from the [S II]constrained fit and the dotted lines those from the [S II] uncon-strained fit

two galaxies (J1613 and J2118) with SNR high enoughto decompose the [S II] emission line profile in separatenarrow and broad components Both the total and nar-row [S II] to Hα ratios of these two galaxies agree withthose of normal star forming galaxies in the SDSS com-parison sample We also include J0106 as the emissionlines are fit with a narrow component only The [S II] toHα ratio for this galaxy is the lowest in our sample andis 037 dex lower than 99 of the DR8 SDSS galaxies ofcomparable [O III]HβIn Fig 7 we compare the locations of the line ratios

for the narrow and broad components (filled dots andopen squares respectively) in the N2- (left) and S2-BPT(right) diagrams for the galaxies where we identify broad[O III] Hβ Hα [N II] and [S II] lines In the figure

the flux ratios for the narrow and broad componentsin each galaxy are connected by a line to ease com-parison The broad [O III]λ5007Hβ ratio is routinelyhigher than the corresponding narrow line ratio with thesole exception of J0826 We find that 58 galaxies have[O III]λ5007Hβ values for the broad component in thecomposite region of the N2-BPT diagram the ratios forJ1613 and J2118 lie above the theoretical extreme star-burst line (Kewley et al 2001) and the ratios for J1622match those of normal star forming galaxies The me-dian [O III] to Hβ ratio for the narrow and broad com-ponents are 04 and 07 respectively The systematicshift between the [N II]λ6585 to Hα ratios for the broadand narrow components in our sources is less clear Themedian [N II] to Hα ratio for the narrow and broad com-ponents shift slightly higher from 067 to 069The [O III]λ5007 to Hβ ratio is sensitive to the hard-

ness of the ionizing radiation field and is useful totrace the ionization parameter of a galaxy (Baldwin et al1981) As shown in Section 41 the kinematics ofthe broad emission lines reflect that they probe out-flowing gas The higher ionization observed in thebroad components could be caused by shocks associ-ated with galactic outflows (Sharp amp Bland-Hawthorn2010) While the S2-BPT diagram can be used to iden-tify shocks unfortunately the low SNR of our spectraprevent us from exploring [S II] broad lines in most ofour sources The two galaxies where we can detect bothbroad and narrow [S II] J1613 and J2118 show similar[S II]λλ67176731Hα values for both componentsIn this section we have shown that the galaxies in our

sample fall in or very near the ldquocompositerdquo region inthe N2-BPT diagram while exhibiting low [S II] to Hαratios in the S2-BPT diagram The position of a starforming galaxy on the BPT diagrams traces the ISM con-ditions and radiation field in the galaxy Several mecha-nisms can shift its location and mimic a composite starforming-AGN system the raise of the hardness of the

12 Perrotta et al

minus1

0

1

Total Total

minus15 minus10 minus05 00 05

log([NII]λ6585Hα)

minus1

0

1

log([

OII

I]λ

5007H

β)

Narrow

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118


log([SII]Hα)

Narrow

Figure 6 N2-BPT (left) and S2-BPT (right) diagrams for the total emission line flux (top panels) and the narrow component line flux(bottom panels) for the galaxies studied here The green dashed lines delineate the empirical separation of star forming galaxies and AGNby Kauffmann et al (2003) in the N2-BPT plane The blue dashed lines are theoretical curves derived by Kewley et al (2001) to show thelocation of maximal starburst galaxies in both diagrams Red and orange dashed lines from Cid Fernandes et al (2010) and Kewley et al(2006) separate LINER and Seyfert galaxies in the N2-BPT and S2-BPT planes respectively Contours show the location of SDSS DR8galaxies for comparison (enclosing 30 50 70 90 and 99 of the galaxies) In the N2-BPT diagram our sample resides mainly inthe composite region (with the exception of J1713 a type II AGN candidate) while in the S2-BPT diagram the total line fluxes in oursample are shifted to lower [S II] to Hα ratios than in SDSS galaxies

minus15 minus10 minus05 00 05log([NII]λ6585Hα )

minus1

0

1

log([

OII

I]λ

5007H

β)

Broad

Narrow

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2118

minus15 minus10 minus05 00 05log([SII]Hα)

Figure 7 N2-BPT (left) and S2-BPT (right) diagrams comparing line ratios for the broad (open squares) and narrow (filled dots) emissionline components for the galaxies in our sample The two sources with SNR high enough to decompose the [S II] emission line profile intoseparate narrow and broad components are shown in the S2-BPT plane All dashed lines and contours are the same as in Fig 6 Thereis no obvious systematic variation of the [N II] and [S II] to Hα ratios between the narrow and broad components while the [O III] to Hβratio is routinely higher in the broad component than the narrow component in all but one galaxy in our sample

ionizing radiation field in a galaxy along the local abundance sequence or its electron density the presence of


shocks caused by galactic winds or mergers the contam-ination of the line ratios by the diffuse ionized gas (DIG)complex geometrical gas distributions As we will discussin Section 5 the composite nature of the galaxies in oursample is more likely due to their extreme physical con-ditions than the presence of a buried AGN

44 Ionization and metallicity

Knowledge of the ionization parameter is crucial in un-derstanding the properties of the ionizing sources as wellas their impact on the surrounding ISM and outflowinggas This parameter is typically measured using the ra-tio of two emission lines from the same atomic speciesthat are in different ionization states Fig 8 showsthe commonly-employed ionization parameter diagnos-tic O32 ([O III]λ5007[O II]λλ37263729) plotted againstabundance-sensitive ratios for the galaxies in our sampleand in SDSS DR8 for comparisonThe left panel shows O32 versus a widely-used optical

metallicity diagnostic the R23 ratio (([O III]λλ49595007+ [O II]λλ37263729)Hβ Pagel et al 1979) Our sam-ple exhibits similar O32 and somewhat lower R23 ratiosthan SDSS galaxies with median values of 03 and 25respectively compared to the full SDSS sample whichhas median values of 03 and 28 The blue and ma-genta contours enclose the 80 and 99 of the high (Mlowast

gt 101052 M⊙) and low (Mlowast lt 10925 M⊙) mass star-forming SDSS galaxies They have median O32 valuesof 03 (high mass) and 04 (low mass) and average R23values of 13 (high mass) and 46 (low mass) The com-posite SDSS galaxies occupy the region between thesetwo in the O32-R23 space The AGN-host galaxies (yel-low contours identified by the Kewley et al (2001) line)have average O32 and R23 values of 05 and 58 respec-tivelyThe galaxies in our sample exhibit ionization proper-

ties and R23 values consistent with those of the high masstail of SDSS star-forming galaxies We note that J1713is the only clear AGN candidate in our sample and itlies in the AGN locus with high O32 and low R23R23 is sensitive to abundance but is double-valued as

a function of metallicity It increases with metallicity atlow gas-phase OH as the number of oxygen atoms in-creases and it reaches a maximum at slightly less thansolar abundance Then R23 decreases again at highgas-phase OH because the oxygen acts as an efficientcooler reducing the gas temperature and consequentlythe number of collisionally-excited oxygen ions There-fore it is crucial to establish which solution branch ap-plies when R23 values are low The degeneracy can beresolved by the use of an additional parameter such asN2O2 ([N II]λ6585[O II]λλ37263729 Evans amp Dopita1985 1986 Dopita et al 2000) N2O2 exhibits a remark-ably tight correlation with metallicity above Z = 04Z⊙with an rms error of 004 (Kewley amp Dopita 2002) Thereasons why N2O2 is highly sensitive to metallicity aretwofold First nitrogen has a large secondary compo-nent of nucleosynthesis at high abundance which causesan increase of N2O2 and second the nebular electrontemperature declines as the abundance increases Thisleads to a strong decrease in the number of collisionalexcitations of the [O II] lines relative to the lower energy[N II] lines at high abundance Moreover N2O2 is almostindependent of the ionization parameter because of the

similar [N II]λ6594 and [O II]λ3726 ionization potentialsmaking this ratio the most reliable metallicity diagnosticin the opticalThe central panel of Fig 8 shows O32 versus N2O2

for our galaxies and the SDSS comparison sample Ourgalaxies exhibit high N2O2 ratios with an average valueof 13 in line with the most massive SDSS star-forminggalaxies suggesting high metallicities (Kewley amp Dopita2002 Kewley et al 2019) This result implies that theR23 values in our sample are low because they are partof the high abundance solution branch We apply a red-dening correction to the [N II] and [O II] lines (see Sec-tion 3) although our sample has uncertain dust contentand geometry While an accurate determination of thegas metallicity in our sample is beyond the purpose ofthis work it is clear that our galaxies have high metal-licitiesIn the right panel of Fig 8 we show O32 versus N2S2

([N II]λ6585[S II]λλ67176731 Dopita et al 2013) forour galaxies and the SDSS comparison sample At highmetallicity nitrogen is a secondary nucleosynthesis ele-ment and sulphur is a primary α-process element Theyhave similar excitation potentials and in the high metal-licity range their line ratio is a function of metallicitydue mainly to the different nucleogenic status of the twoelements The N2S2 diagnostic is not as useful as N2O2for the determination of abundance because it is consid-erably more sensitive to the ionization parameter but ithas the strong advantage that reddening corrections arenegligible Our sample exhibits high N2S2 ratios withan average value of 53 again implying high metallicity(Kewley amp Dopita 2002 Kewley et al 2019) Some ofthe targets in our sample have N2S2 values similar tothose of the most extreme high mass SDSS star-formingand AGN host galaxies However both these galaxy pop-ulations have average N2S2 of 15 more than three timeslower than the average value for our sampleLastly we note that two commonly-used metallic-

ity calibrations by McGaugh (1991) and Zaritsky et al(1994b) infer derived log(OH) + 12 = 90 and log(OH)+ 12 = 89 respectively for galaxies in our sampleThese values are in line with those inferred using theN2O2 and N2S2 diagnostics

45 Comparison with galaxy properties

In this section we investigate how the N2S2 and O32line ratios depend on the physical properties of the galax-ies studied in this paper as compared to other galaxypopulationsIn Fig 9 in the top row we show N2S2 versus the galaxy

stellar mass (Mlowast) star formation rate (SFR) and starformation rate surface density (ΣSFR) for galaxies in oursample as well as in SDSS We see in the upper left panelthe well known relation between galaxy mass and metal-licity (as seen in N2S2) in SDSS The galaxies in oursample are uniform in Mlowast with values comparable to thehigh mass tail of SDSS galaxies Our galaxies also havehigh N2S2 higher even than the typical N2S2 ratio atthe high masses of our galaxies This likely reflects thelack of S2 in our sources as seen in the S2-BPT diagramabove In the middle and right panels it is clear thatour galaxies have extremely high SFR and ΣSFR valuesbeyond SDSS galaxiesIn the lower panels we investigate the relationship be-

14 Perrotta et al

Figure 8 The ionization-sensitive ratio O32 ([O III]λ5007[O II]λλ37263729) plotted against abundance-sensitive diagnostics for oursample and the SDSS DR8 comparison sample Light grey contours enclose the 80 and 99 of the SDSS galaxies while blue and magentacontours enclose the 80 and 99 of the high (Mlowast gt 101052 M⊙) and low (Mlowast lt 10925 M⊙) mass star-forming SDSS galaxies respectivelyThe yellow contours illustrate the location of 80 and 99 of the SDSS AGN-host galaxies Left panel R23 ratio (([O III]λλ49595007+ [O II]λλ37263729)Hβ Pagel et al 1979) Central panel N2O2 ratio ([N II]λ6585[O II]λλ37263729 Evans amp Dopita 1985 1986Dopita et al 2000) Right panel N2S2 ratio ([N II]λ6585[S II]λλ67176731 Dopita et al 2013)

tween the O32 diagnostic and galaxy properties againfor galaxies in our sample and in SDSS We alsoshow known Lyman continuum (LyC) ldquoleakersrdquo at lowand high redshift (Alexandroff et al 2015 Izotov et al2016ba 2018ab Bassett et al 2019 Wang et al 2019Fletcher et al 2019) As pointed out in Section 44 oursample shows O32 ratios comparable to the most mas-sive SDSS galaxies and N2S2 ratios similar to some ofthe most extreme SDSS galaxies However the impliedaverage metallicity from N2S2 is much higher than thatof the bulk of any SDSS galaxy population As discussedin Section 54 LyC leakage may affect [N II] and [S II]differently producing a deficiency of [S II] and conse-quently anomalously high N2S2 observed valuesAn interesting comparison with our sample in the lower

panels of Fig 9 is with confirmed LyC leakers namelygalaxies with an estimated fraction of ionizing Lymancontinuum photons (λ lt 912 A) that escape into the IGMthat is greater than zero (fesc(LyC)gt 0) Our sampleexhibits some distinctive characteristics of known LyCleakers but differs in other crucial properties Most ofthe LyC leakers are substantially less massive than ourgalaxies They span a wide range (37 dex) of Mlowast withan average value of 1091 M⊙ sim15 orders of magnitudelower than the average Mlowast for our sample LyC leakersdisplay a broad range of O32 values (215 dex) Theiraverage O32 is 12 dex higher than in our sample how-ever the most massive LyC leakers overlap well with theO32 values of the compact starburst galaxies consideredin this work The SFR and ΣSFR values of the LyC leak-ers are more similar to those of our galaxies Specificallyin these samples LyC leakers have an average SFR of 37M⊙ yrminus1 and an average ΣSFR of 147 M⊙ yrminus1 kpcminus2these values are four times lower than the average values

in our sample It is worth noting that both the LyC leak-ers and our sample are entirely distinct from the SDSSgalaxy population in terms of having very high ΣSFR

valuesWhile there are not N2S2 ratios reported for the LyC

leakers presented in Fig 9 some have metallicity esti-mates ranging from log(OH) + 12 = 762 to log(OH)+ 12 = 816 (Izotov et al 2016ba 2018ab) These LyCleakers are considerably less metal-rich than our galaxiesas expected by their lower stellar masses (Such low val-ues correspond to a regime where N2S2 is not sensitiveto metallicity with values around 03 (Kewley amp Dopita2002 Kewley et al 2019) The most massive LyC leak-ers shown in Fig 9 have derived metallicity in the range818lt log(OH) + 12 lt 886 (Alexandroff et al 2015Bassett et al 2019 Wang et al 2019) where 87 corre-sponds to solar metallicity (These values imply an N2S2lt 32 Kewley amp Dopita 2002 Kewley et al 2019) Wediscuss in Section 54 below whether the galaxies in oursample are LyC leaker candidates

5 DISCUSSION

We next discuss our results including possible originsof the kinematically broad flux emission (Section 51) InSection 52 we examine the possible contribution of AGNto the observed emission lines and then consider severaladditional mechanisms that can affect the location of oursample in the line ratio diagnostic plots (Section 53) Wethen review the properties of the galaxies in this studyas potential LyC leaker candidates (Section 54)

51 Interpreting Broad Emission Lines as Tracers ofGalactic Outflows


minus10

minus05

00

05

10

15lo

g([

NII

] 65

85[

SII

] 67

176

73

1)

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2116

J2118

7 8 9 10 11 12

log(MlowastM⊙)

minus10

minus05

00

05

10

15

log

([O

III]

50

07[

OII

] 37

263

72

9)

Wang et al 2019

Bassett et al 2019

Fletcher et al 2019

Izotov et al 201618

Alexandroff et al 2015

0 1 2

log(SFRM⊙yrminus1)

minus2 0 2 4

log(ΣSFRM⊙yrminus1kpcminus2 )

Figure 9 Top panels total [N II]λ6585 to [S II]λλ67176731 flux ratio compared to stellar mass (left) star formation rate (central) andstar formation rate surface density (right) Bottom panels total [O III]λλ5007 to [O II]λλ37263729 flux ratio compared to stellar mass(left) star formation rate (central) and star formation rate surface density (right) The grey contours represent SDSS DR8 data withcontours at 25 50 75 90 and 99 Black empty symbols are Lyman continuum leaking galaxies zsim03 [S II]-weak galaxies (squaresWang et al 2019) low-redshift Green Pea galaxies (stars Izotov et al 2016ab 2018ab) low-redshift Lyman Break Analogs (trianglesAlexandroff et al 2015 zsim3 star-forming galaxies (diamonds Bassett et al 2019) and zgt3 LACES galaxies (pentagons Fletcher et al2019) Five targets from Fletcher et al (2019) are not detected in [O II] the O32 values are 3σ lower limits

Galactic winds are typically identified through theirkinematic signatures Winds seen in emission are de-tected as broad lines identified alongside a narrowercomponent resulting from star forming regions in thegalaxy (eg Newman et al 2012 Freeman et al 2019)As shown in Section 41 the emission lines in 1214galaxies in our sample require a broad+narrow Gaus-sian decomposition for at least one of the emissionlines studied in this work (ie [O II]λλ37263729Hβ [O III]λλ49595007 Hα [N II]λλ6549 6585 and[S II]λλ67166731) The mean values of the velocity dis-persion (σ) in the [O II] and Hα broad components inour sample are 670 and 470 km sminus1 respectively Thebroad components are also offset in their centroid veloc-ities from the narrow components blueshifted by meanvalues of 352 and 143 km sminus1in [O II] and Hα respec-tively Such line broadening and blueshifts are inter-preted in galactic spectra as outflowing gas In manycases for the galaxies in our sample the broad compo-nents exhibit some redshifted emission as well comparedto the narrow line profiles though the velocity centroidsare always blueshifted We attribute this to dust present

in the host galaxy that obscures a portion of the red-shifted outflowsStar formation-driven outflows are observed in galax-

ies of all stellar masses with an occurrence thatcorrelates with star formation properties specifi-cally SFR the offset from the main sequence ofstar formation and ΣSFR (eg Kornei et al 2012Rubin et al 2014 Heckman et al 2015 Chisholm et al2015 Forster Schreiber amp Wuyts 2020) Our sampleprobes high ΣSFR and as expected it presents a high in-cidence of broad emission lines However many aspectsare important in interpreting trends of outflow character-istics with galaxy properties For example the capabilityto detect an outflow depends on the strength of the windsignatures along with the SNR and spectral resolutionof the data Slower or weaker winds are more difficultto identify especially using nebular emission lines as thebroad components can be difficult to separate from thenarrow emission from star formation Therefore a noteof caution is in order when using the incidence of broadlines as a function of galaxy properties Also differencesin sample selection and assumptions made in the analysis

16 Perrotta et al

may result in different conclusions For example therehave been claims of a strong dependence of the outflowincidence on ΣSFR in high-redshift star-forming galaxiesthough the existence and location of a threshold in ΣSFR

is somewhat unclear (Newman et al 2012 Davies et al2019) In a forthcoming paper (Davis et al in prep) weinvestigate scaling relations between outflow and galaxyproperties for 46 galaxies in our parent sample that wecollected spectra for and review the biases related to theuse of different outflow tracersEmission and absorption lines provide us distinct ap-

proaches to investigate outflows While emission linesderive from the projected signal of emitting gas fillingthe whole volume in front of and behind the galaxy ab-sorption lines probe only the gas along the line of sightilluminated by the central starburst Furthermore theabsorption lines are sensitive to the density of the gasprobed while emission lines are sensitive to the densitysquared This results in absorption lines providing accessto lower density weaker gas components Comparing v98derived from the [O II] emission lines and Mg II absorp-tion lines in the right panel of Fig 4 we see that generallythe Mg II maximum velocities are higher (though theyare consistent with [O II] emission for several galaxies)This might be explained if the outflowing gas has a lowerdensity on average which makes it easier to accelerateIt is also reasonable that absorption line velocities maybe higher than emission line velocities on average asemission lines can probe gas that is both in front of andbehind the galaxy This can produce a redshifted wingin emission profiles that shifts both the central velocityand the velocity at which 98 of the total EW is detectedtowards smaller valuesBroad emission lines have also been used to constrain

outflow properties beyond kinematics The broad to nar-row flux ratio (BFR) of Hα has been shown in the litera-ture as a function of galaxy parameters and used to inferthe mass loading factor (η = outflow mass rateSFR)Adopting a model that describes the outflow geometryand physical conditions it is possible to convert the ob-served Hα BFR into an estimate of η (Steidel et al 2010Genzel et al 2011) This approach has been used toidentify a possible threshold in star formation propertiesabove which a galaxy has the ability to power outflows(eg Newman et al 2012 Freeman et al 2019) In par-ticular the inferred η has been found to strongly corre-late with ΣSFR within some galaxy samples Thereforea ΣSFR threshold has been proposed that dictates whenstar formation feedback may break through the densegas layers in the galactic disk and launch a large-scaleoutflowFor comparison to other studies we parameterize the

broad emission we measure using the BFR Fig 10shows the Hα BFR as a function of Mlowast and ΣSFR

for our sample and other relevant star-forming galaxies(Newman et al 2012 Genzel et al 2014 Swinbank et al2019 Forster Schreiber et al 2019 Freeman et al 2019)Symbols with thick contours reflect stacked spectrawhile grey symbols show results for individual galaxiesFig 10 shows that when we consider samples spanninga wide dynamic range there is no correlation betweenBFR and Mlowast or ΣSFR Additionally there is not clearevidence for a threshold in ΣSFR above which outflowsare launched Similarly such a threshold is also not ob-

served in low-redshift LIRG and ULIRG galaxies evenafter correcting for the differential fraction of the gascontent (Arribas et al 2014)Trends of BFR with Mlowast or ΣSFR observed in previ-

ous studies are often in tension with theoretical expec-tations and numerical simulations (Newman et al 2012Lilly et al 2013 Muratov et al 2015 Freeman et al2019 Forster Schreiber amp Wuyts 2020) A reasonableexplanation is that when observations are used to in-fer global properties of outflows the adopted assump-tions regarding velocity geometry temperature ioniza-tion source and gas density are too simplistic and failto capture the complexity of the outflows (Rupke et al2019) Additionally Hα traces the warm ionizedgas phase and much if not most of the outflowingmass is likely in an neutral atomic or molecular phase(Walter et al 2002 Rupke et al 2005 Rupke amp Veilleux2013 Fluetsch et al 2020 Veilleux et al 2020) Giventhe potential systematic issues in detecting outflows us-ing broad emission lines a note of caution is warrantedin interpreting any correlation between BFR and Mlowast orΣSFR especially when different sample selections or anal-yses are involved

52 AGN Contamination

All but one of the galaxies in our sample fall in thecomposite region in the N2-BPT diagram Galaxies inthis region are often interpreted as having contributionsto their line ratios from both star formation and AGNand it is therefore important to understand the possibleAGN contribution in our sourcesIn general we do not find evidence for widespread

AGN activity in our sources None of the galaxies in thisstudy show evidence of an AGN in their restframe near-ultraviolet and optical spectra (eg lack of very broadMg II Hβ or Hα) Additionally none of the sourceswould be classified as AGN based on their WISE mid-IR colors (the median W1minusW2 of our sample is 035Petter et al 2020) They also do not satisfy the W1minusW2gt 08 (Vega) criterion of Stern et al (2012) or the color-magnitude cuts of Assef et al (2013) that include faintersourcesTen galaxies in our sample (J0106 J0826 J0905

J0944 J1107 J1125 J1341 J1613 J2116 and J2118)have VLA 15 GHz continuum observations that allowus to place constraints on the ongoing radio AGN ac-tivity in these systems The derived radio luminosities(L15GHz) span a 52minus 505times1022 WHzminus1 with a medianvalue of 50times1022 WHzminus1 (Petter et al 2020) TheseL15GHz are 3σ below the radio excess threshold used bySmolcic et al (2017) to identify AGN-dominated radiosources and are compatible with being powered by thecentral starburstSix galaxies in our sample were part of a Chandra

observing program targeting the 12 galaxies in the par-ent sample with the strongest indication for possible on-going AGN activity based on emission-line properties(Sell et al 2014) Three of the galaxies in this study(J1506 J1613 and J2118) have weak detections (4 X-raycounts each) implying an X-ray luminosity of Lx asymp 1042

erg sminus1 The remaining three (J0826 J0944 and J1713)have upper limits corresponding to Lx lt 1043 erg sminus1The derived X-ray luminosities are consistent with theknown IR-based SFRs of these sources (Asmus et al


95 100 105 110 115log(MlowastM⊙)

10minus1

100

101

Fb

roadF

nar

row

minus2 0 2 4

log(ΣSFRM⊙yrminus1kpcminus2)

Genzel et al 2014

Freeman et al 2019

Swinbank et al 2019

Forster Schreiber et al 2019

Newman et al 2012

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2116

J2118

Figure 10 Broad-to-narrow Hα flux ratio as a function of stellar mass (left panel) and star formation surface density (right panel)for our galaxies and some relevant star-forming galaxy samples Stars are 10 star-forming galaxies at z sim 2 from the MOSDEF survey(Freeman et al 2019) Squares are the galaxy average values of 529 star-forming galaxies at z sim 1 from KMOS observations (Swinbank et al2019) Diamonds are 20 z = 1minus 2 galaxies from Genzel et al (2014) Pentagons are stacks of 27 z sim 2 star-forming galaxies from the SINSand zC-SINF surveys (Newman et al 2012) Triangles are stacks of 78 (left panel) and 33 (right panel) star-forming galaxies at 06 lt z lt27 from the KMOS3D survey (Forster Schreiber et al 2019)

2011 Mineo et al 2014 Sell et al 2014)Sell et al (2014) classified J1713 as the most likely

galaxy in their sample to host a type II AGN based onpseudo-BPT diagrams (eg [O III]Hβ vs [O II]Hβ)and estimated a bolometric Eddington fraction ofLbolLEdd asymp 002 minus 013 The new spectroscopic dataand resulting line ratios for this galaxy lead to the sameconclusion (see Fig 6) as this galaxy does not lie in thecomposite region but is clearly in the AGN region of theBPT diagrams Moreover J1713 is distinct from the restof our sample in the ionization and abundance diagnos-tics plots (Fig 8) and overlaps the SDSS AGN locus inthese spaces We therefore conclude that this source doescontain an AGNJ1506 exhibits a clear (sim10σ) [NeV]3426A detection

this ion has a high ionization potential and is com-monly used to trace AGN activity (eg Gilli et al 2010)Sell et al (2014) estimate a ratio of the X-ray to [NeV]luminosity LxL[NeV] = 49 implying a Compton-thickAGN (NH gt 1024 cmminus2) Under the assumption ofthe emission line being produced by an obscured AGNSell et al (2014) find that the AGN would contributesim10 of the mid-IR luminosity However [NeV] canalso be powered by a very young (less than a few Myr)stellar population containing Wolf-Rayet and O stars(Abel amp Satyapal 2008) J1506 has a very young (sim 3Myr) stellar population and the highest ΣSFR in oursample Therefore the observed [NeV] could be pro-duced by the extreme conditions of the central star-burst (Sell et al 2014) [NeV]3426A emission is also de-tected in the outflowing component of another of oursources J2118 (Rupke et al 2019) The derived lu-minosity L[NeV]=36(plusmn)1times1040 erg sminus1 is three timeslower than the averge for typical [NeV] emitters at sim-ilar redshift (Vergani et al 2018) and could be pro-duced by fast shocks with velocities of at least 300minus400km sminus1(Best et al 2000 Allen et al 2008)In summary most of the galaxies in this study show

no evidence for AGN activity based on X-ray and radio

observations optical emission lines and infrared colorsFor the galaxies that may contain a dust-obscured ac-creting SMBH the AGN contributes a small fraction ofthe bolometric luminosity While we cannot rule out pastheightened AGN activity multi-wavelength data for allof but one of these galaxies can be explained by theirknown star formation properties and the possible pres-ence of shocks

53 Interpreting the BPT diagrams

In order to interpret the position of a galaxy in the N2-and S2-BPT diagrams and understand the gas ionizationsource(s) it is key to consider the mechanisms that canaffect the integrated galaxy line ratios In addition tothe potential contribution from AGN as discussed abovehere we consider the possible contributions from diffuseionized gas (DIG) and shocksStudies based on narrowband Hα imaging have

revealed that DIG can contribute substantially to theoptical line emission in local galaxies (Zurita et al2000 Oey et al 2007) Typically DIG exhibitsenhanced forbidden-to-Balmer line ratios (eg[S II]λλ67176731Hα [N II]λ6585Hα [O II]λ3726HβHoopes amp Walterbos 2003 Madsen et al 2006Voges amp Walterbos 2006) relative to H II regionsTherefore DIG contamination can move the locationof a galaxy in the BPT diagrams towards compositeor LINER-like regions (Sarzi et al 2006 Yuan et al2010 Kehrig et al 2012 Singh et al 2013 Gomes et al2016 Belfiore et al 2016ba) Zhang et al (2017) andSanders et al (2017) have shown that DIG deviates fromH II regions more in emission-line diagrams featuring[S II] or [O II] rather than [N II] and that DIG ischaracterized by a lower ionization parameter than H IIregions Additionally the fractional contribution of DIGemission to the Balmer lines (fDIG) is found to declinewith increasing ΣSFR (Oey et al 2007 Masters et al2016 Shapley et al 2019) Indeed DIG emission isnegligible in typical high-redshift galaxies that are more

18 Perrotta et al

highly star-forming (Whitaker et al 2014) and morecompact (van der Wel et al 2014) A substantial DIGcontribution to the emission line ratios in our sampleis in contrast to the low [S II]Hα (Fig 6)observedvalues Most importantly similarly to high redshiftgalaxies our sample is characterized by extremely highΣSFR with a average value of 620 M⊙ yrminus1 kpcminus2roughly 4 order of magnitudes higher than the medianSDSS ΣSFR We therefore can safely assume negligiblecontamination from DIG (fDIG sim 0) when interpretingthe BPT diagram locations of our galaxiesAs discussed above the presence of an AGN can also

affect the location of galaxies in the BPT diagrams Asthe contribution from an AGN increases its host galaxymay migrate from the empirical sequence of H II re-gion emission toward the AGN portion of the diagnos-tic diagrams as a consequence of the increasing con-tribution from a harder ionizing radiation (Yuan et al2010) However in starburst+LINER systems the na-ture of the observed composite activity may be the resultof non-AGN sources In ultraluminous infrared galaxies(ULIRGs) extended LINER emission has been observeddue to starburst wind-driven and merger-driven shocks(Sharp amp Bland-Hawthorn 2010 Rich et al 2010 2011Soto et al 2012 Rich et al 2015) Moreover shockscommon in ongoing mergers can significantly enhance[O II] relative to [O III] thus reducing the observedO32 which is used to probe the ionization state of agalxy (Rich et al 2015 Epinat et al 2018 Bassett et al2019) Gas outflows and mergers can produce widespreadshocks throughout a galaxy which can substantially im-pact its emission line spectrum at both kpc and sub-kpcscales (Medling et al 2015) Rich et al (2014) comparedspatially resolved spectroscopy of 27 local ULIRGs to thespectra extracted from their brightest optical nuclear re-gions Interestingly they found that 75 of the galax-ies in their sample that would be classified as compositebased on optical nuclear line ratios result from a sizablecontribution from shocks to their emission line spectraTherefore shock emission combined with star formationcan mimic ldquocompositerdquo optical spectra in the absence ofAGN contributionShock excitation can affect both low and high ioniza-

tion line ratios In slow shocks (v lt 200 km sminus1) theshock front moves faster than the photoionization frontcaused by the shocked gas This type of shock producesrelatively weak high ionization lines but strong low ion-ization species such as [S II] and [N II] (Rich et al 20112015) In fast shocks (v gt 200 km sminus1) the extreme ul-traviolet and soft X-ray photons generated by the cool-ing of the hot gas behind the shock front produce a su-personic photoionization front that moves ahead of theshock front and preionizes the gas This photoionizationfront is referred to as the photoionizing precursor andit produces strong high ionization lines while the hardradiation field from the shock front itself produces an ex-tended partially ionized zone where low ionization linessuch as [OI] [NI] and [S II] are observed (Allen et al2008 Kewley et al 2019) Kewley et al (2013) showedhow local galaxies containing emission from either slowor fast shocks can result in composite locations in theBPT diagramsWhile the total luminosity of a shock depends only

on its velocity and the gas density the emission line

spectrum depends strongly on the physical and ioniza-tion structure of the shock This is determined primarilyby the shock velocity the magnetic parameter and themetallicity Moreover the density may play a crucial rolewhen it is sufficiently high for collisional de-excitation offorbidden lines to become important The magnitudeand direction of the emission line ratios shifts due toshocks are complex and difficult to predict Howevershocked emission tends to have higher [N II]Hα and[S II]Hα ratios compared to photoionized H II regions(Allen et al 2008 Rich et al 2011)Slow shock models (Rich et al 2011) can not simul-

taneously reproduce the (total and narrow) line ratiosin the N2- and S2-BPT diagrams for our galaxy sam-ple (Fig 6) Fast shock + precursor model grids fromAllen et al (2008) produce too high [N II] and [S II]to Hα ratios at given [O III]Hβ compared to the val-ues for our sample The [N II]Hα and some of the[S II]Hα observed ratios can be reproduced by somemodels that include only emission from the post-shockregion In particular model grids that simultaneouslymatch 70 of our sample in the N2- and S2-BPT di-agrams have high pre-shock density of sim 1 000 cmminus3solar or super solar metallicity a magnetic field strengthof B lt 32microG and a wide range of shock velocity val-ues spanning 200 minus 700 km sminus1 However such a highpre-shock density would imply a post-shock density of10 000 cmminus3 (Dopita amp Sutherland 1995) which is anextreme and unlikely conditionThe broad emission line ratios (Fig 7) in both BPT

diagrams reside within a wide range of model grids thatinclude emission from either the post-shock region orboth post- and pre-shock regions with shock veloci-ties of 200 minus 1000 km sminus1 and pre-shock densities of001 minus 1 000 cmminus3 Our spectra do not have sufficientSNR to study the broad [S II] component for most ofthe galaxies in our sample However in the two objectswhere we can identify a broad [S II] line the line ratiosare consistent with having a shock contribution com-bined with star formationIt is extremely challenging to investigate the ionization

source(s) in a galaxy when only a spatially-integratedspectrum is available and it is possible that the line ra-tios have contributions from multiple sources Howevershocks rarely dominate the global emission of a galaxyand if a galaxy does contains shocks there may alsobe contributions from star formation andor an AGNRich et al (2011) found that the enhanced optical lineratios from shocks are washed out by star formation andare thus easier to observe on the outskirts of galaxieswhere the level of star formation is lower In our samplewhere the optical light is dominated by young stars thatformed within the central sim few hundred parsecs duringrecent starburst events (Diamond-Stanic et al 2021) itis plausible that any signatures of potential shocks arewashed out by the intense star formation present

54 LyC leakers candidates

Next we discuss the possibility of the galaxies in oursample being Lyman continuum (LyC) leakers This isof interest as it is currently unclear what sources are re-sponsible for creating the epoch of reionization whichmarks a crucial transition phase in the early Universe inwhich hydrogen in the IGM is transformed from a neutral


to an ionized state (Fan et al 2006 Komatsu et al 2011Zahn et al 2012 Becker et al 2015 Boera et al 2019)Deep HST near-IR imaging indicates that primordialstar-forming galaxies are capable of producing the bulkof the LyC photons needed to drive reionization (egBouwens et al 2012 Oesch et al 2013 Robertson et al2015) It has been estimated that the escape fraction ofLyC (fesc(LyC)) ie the fraction of ionizing radiationreleased into the IGM should be at least 10 minus 20 per-cent on average (eg Ouchi et al 2009 Robertson et al2013 Khaire et al 2016)The increasing IGM neutral fraction at z gt 5 prevents

a direct measurement of the LyC escaped from galaxiesTherefore searches for LyC leakers are carried out atlower redshift to identify the indirect signs of LyC escapeMany groups have observed such galaxies from low red-shift up to z sim 4 Most of the confirmed LyC leakers havefesc(LyC) below 015 (Leitet et al 2013 Borthakur et al2014 Izotov et al 2016b Leitherer et al 2016) Thereare examples with estimated fesc(LyC) as high as045 minus 073 (Vanzella et al 2016 de Barros et al 2016Shapley et al 2016 Bian et al 2017 Vanzella et al2017 Fletcher et al 2019 Izotov et al 2018ab) how-ever these galaxies are remarkably rareSome of the distinct observational signatures shared

by the LyC leakers are strong Lyα emission with adouble-peaked Lyα line profile (Verhamme et al 20152017 Vanzella et al 2020) high ΣSFR high specificstar formation rate (sSFR) and high ionization pa-rameter traced by the O32 ratio (Izotov et al 2018bde Barros et al 2016 Vanzella et al 2020 Cen 2020)Some of them also show high ne (eg Guseva et al 2020)As seen in Section 45 the galaxies in our sample

show some of the features common to known LyC leak-ers In particular both populations are substantiallydistinct from the SDSS galaxy locus in terms of hav-ing high ΣSFR a physically-motivated model relatingfesc(LyC) to ΣSFR was recently proposed (Sharma et al2016 Naidu et al 2020 Cen 2020) The average O32of known LyC leakers is around 12 dex higher than inour sample however our sample overlaps well with therange of O32 values shown by the most massive LyCleakers High O32 was initially used as a primary se-lection criterion to identify LyC leaker candidates how-ever it was revealed that fesc(LyC) does not correlatestrongly with O32 (see Izotov et al 2018b Naidu et al2018 Bassett et al 2019 Nakajima et al 2020 and dis-cussions therein)Most recently [S II] deficiency has been used as an

empirical signpost to identify LyC emitter candidates(Wang et al 2019 Ramambason et al 2020) The [S II]deficiency is a tracer of gas that is optically thin toionizing radiation allowing the escape of LyC photonsIn a classical ionization-bounded H II region the [S II]lines are produced in the warm partially ionized regionnear and just beyond the outer edge of the Stromgrensphere In a density-bounded nebula the flux of ion-izing photons from the central source is so large thatthe gas between the source and the observer is fullyionized As a result the ionizing radiation can escapebecause there is little or no neutral gas between thesource and observer to absorb these photons In thismodel the outer partially-ionized [S II] zone is weakor even absent and the relative intensity of the [S II]

emission lines drop substantially (Pellegrini et al 2012Wang et al 2019 Ramambason et al 2020) As dis-cussed above the galaxies in our sample show weak[S II]λλ67176731 nebular emission-lines relative to typi-cal star-forming galaxies They have high [N II]Hα ra-tios consistent with amp solar metallicity but they exhibitanomalously weak [S II] lines (see Fig 6) which couldresult from LyC photons escaping without encounteringa low ionization outer edge of the nebulaSimilarly to O32 [S II] deficiency does not appear

to correlate strongly with the fesc(LyC) of the knownleakers However empirical correlations between lineratios and estimated fesc(LyC) may be affected by ge-ometric effects (Steidel et al 2018 Bassett et al 2019Fletcher et al 2019) and shocks (see Section 53) Ob-servations of individual local H II regions show themto be geometrically complex with significant spatialvariation in oxygen line ratios suggestive of regionsfrom which LyC could escape (eg Zastrow et al 2011Weilbacher et al 2015 Kehrig et al 2016 Keenan et al2017 Micheva et al 2018) H II regions may presentchannels carved into the ISM through which LyC fluxcould escape while other areas remain completely opaqueto high-energy radiation Single-component (one-zone)photoionization models do not capture such complex-ity failing to simultaneously reproduce the high andlow ionization lines and escape fractions of LyC leak-ers It has been shown that two-zone models combin-ing regions with a high- and low-ionization parameterwhere one of which is density-bounded do a better jobat reproducing the observed line ratios and fesc(LyC)(eg Ramambason et al 2020) Predictions from thetwo-zone models in classical BPT diagrams vary withfesc(LyC) LyC leakage does not influence [N II] stronglyas it originates from the highly excited region in the innerpart of the H II region such that it remains unaffectedwhen the edges of the H II regions are trimmed In con-trast the [S II] lines are very sensitive to LyC leakageThe complexity of H II regions could explain the variancein the O32 and [S II] values displayed by the confirmedLyC leakersIt has been reported that the LyC emitters with the

largest fesc(LyC) also exhibit high sSFR (gt 1Gyrminus1Bassett et al 2019 Kim et al 2020) Additionally hy-drodynamical simulations find a correlation between in-creasing sSFR and increasing fesc(LyC) (Yajima et al2011 Wise et al 2014) In Fig 11 we investigate the re-lationship between the O32 ratio and sSFR for our sam-ple and the known LyC leakers shown in Fig 9 The LyCleakers have a median sSFR of 10minus88 yrminus1 nearly 08dex higher than the median value in our sample How-ever they span a wide range (34 dex) in sSFR and thesSFRs values in our sample are similar to those of themost massive known LyC leakers The relatively lowersSFR values for our galaxies derive from their substan-tially higher Mlowast sim15 orders of magnitude higher thanmost known LyC leakers Despite most of the confirmedLyC leakers show high sSFR this may not be the relevantparameter for driving LyC leakage as the way in whichthey appear to be most distinct from other galaxies isnot Mlowast but ΣSFR (see Fig 9) Similarly to our samplealmost all the individual observed LyC leakers to dateshow ΣSFR higher than the average ΣSFR expected at

20 Perrotta et al

minus9 minus8 minus7

log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118

Figure 11 Total [O III]λλ5007 to [O II]λλ37263729 flux ra-tio compared to specific star formation rate The grey con-tours show SDSS DR8 galaxies with contours at 25 5075 90 and 99 Black empty symbols are known Ly-man continuum leaking galaxies zsim03 [S II]-weak galaxies(squares Wang et al 2019) low redshift Green Pea galaxies (starsIzotov et al 2016ab 2018ab) low redshift Lyman Break Analogs(triangles Alexandroff et al 2015) zsim3 star-forming galaxies (dia-monds Bassett et al 2019) and zgt3 LACES galaxies (pentagonsFletcher et al 2019) Five targets from Fletcher et al (2019) arenot detected in [O II] the O32 values are 3σ lower limits

their redshifts (Sharma et al 2016 Naidu et al 2020)An additional piece of evidence that a portion of the

LyC may be escaping the host galaxy comes from theemission and absorption line profiles Contrary to typ-ical starbursts most of the galaxies in our sample lackMg II absorption near the systemic velocity as seen inSection 41 This suggests that much of the photoelectricopacity to the LyC is in the tenuous wind itself ratherthan in the dense H II regions This is borne out bythe nebular emission lines which have usually broad lineprofiles sometimes extending over the same range of ve-locities seen in absorption (see Fig 3)Another intriguing line of evidence suggests that the

galaxies in our sample may be leaking LyC photonsThey have weak nebular emission lines while detailedstellar population synthesis modeling of their UV-opticalspectra shows that many of the galaxies have youngionizing stellar populations (lt 10 Myr) that should beproducing copious nebular emission To illustrate thiswe compiled a set of color-matched galaxies from theeBOSS sample (Dawson et al 2016) for each of the galax-ies in our sample We selected galaxies with g-r and r-iwithin plusmn06 mag and redshift within plusmn005 resultingin 10 minus 200 comparison galaxies per source We foundthat our galaxies have much lower Hβ EWs than thecolor-matched eBOSS galaxies with a median Hβ EWof 67A in our sample and 35A in the comparison sam-ple We note that dust alone can not account for the lackof strong emission in our sources The apparent Balmeremission line deficit is not an artifact of differential dustattenuation the 3 minus 5 Myr old stars producing the ion-izing photons share the same attenuation as the nebularemission lines excited by these starsHowever there may be other ways to explain these ob-

servations A possible scenario is that substantial num-bers of LyC photons are absorbed by dust before ioniz-ing hydrogen Some amount of dust absorption seems

likely in our sources as WISE 22 microm imaging shows thatthe galaxies are luminous in the restframe mid-IR How-ever these galaxies are luminous in the GALEX far-UVbands and their SEDs suggest relatively modest atten-uation (AV sim 043) Thus a complex ldquopicket fencerdquoISM geometry may be likely with some high attenua-tion sightlines and some holes enabling LyC escape Thehigh incidence of strong outflows detected in our samplemay be responsible for such holes in the ISMIn summary the galaxies in our sample show multi-

ple indirect indications that they might be leaking LyCphotons They are characterized by high ΣSFR and ion-ization parameters traced by the O32 ratio in line withthose of the most massive known LyC leakers Moreoverthey lack of gas near zero velocity and exhibit Balmeremission lines weaker than expected from stellar pop-ulation synthesis modeling of their UV-optical spectraLastly they show anomalously weak [S II] lines As ourgalaxies differ in many respects from known LyC leak-ers our sample may offer an ideal opportunity to testwhat physical property is most closely linked to LyCescape Directly measuring the LyC and determiningfesc(LyC) for our sample are necessary steps to confirmour hypothesis of potential LyC leakage If our sample isfound to have a significant fesc(LyC) this would revealthat the LyC leakage process is not exclusively drivenby low mass (lt 108 M⊙) galaxies Interestingly the re-cent model by Naidu et al (2020) suggests that lt 5 ofbright (MUV lt 18) galaxies with log(MlowastM⊙) gt 8 couldaccount for gt 80 of the reionization budget makingour sample potential analogs to the high redshift sourcesdriving the reionization

6 SUMMARY AND CONCLUSIONS

We use new optical and near-IR spectroscopy of 14compact starburst galaxies at z sim 05 in combina-tion with ancillary data to study both the nature oftheir extreme ejective feedback episodes and the physi-cal conditions in their dusty cores These galaxies aremassive (Mlowast sim 1011M⊙) compact (half-light radius sim

few hundred pc) they have high star formation rates(mean SFR sim 200M⊙yr

minus1) and star formation surfacedensities (mean ΣSFR sim 2000M⊙yr

minus1kpcminus2) and areknown to exhibit extremely fast (mean maximum veloc-ity sim minus1890 km sminus1) outflows traced by Mg II absorp-tion lines (Tremonti et al 2007 Davis et al in prep)Our unique data set consists of a suite of both emis-sion ([O II]λλ37263729 Hβ [O III]λλ49595007 Hα[N II]λλ6549 6585 and [S II]λλ67166731) and absorp-tion lines (Mg IIλλ27962803 and Fe IIλ2586) that allowus to study the kinematics of the cool gas phase (T sim 104

K) The high Mlowast SFR and ΣSFR values of these galax-ies allow us to extend the dynamic range over which toinvestigate trends of outflow characteristics with galaxyproperties Our main conclusions are as follows1) The emission lines in 1214 galaxies in our sample

show a broad and blueshifted component The [O II]and Hα broad emission lines exhibit average widths (σ)of 668 and 467 km sminus1 and offsets of their central ve-locities from the systemic redshift (voff ) of 352 and 143km sminus1 respectively (Section 41 and Fig 2) Such linebroadening and blueshifts clearly trace high velocity out-flows2) The ions studied in this work allow us to probe out-


flowing gas at different densities and distances from thecentral starburst Absorption lines are sensitive to lowerdensity gas and in our sample typically display somewhathigher maximum velocities than the emission lines (Sec-tion 41 Fig 3 and 4) This could reflect that the fastestoutflowing gas has lower density on average which maybe easier to accelerate3) We characterize the physical conditions of the com-

pact starburst using an ensemble of line ratio diagramsas key diagnostics of electron density metallicity andgas ionization Our sample exhibits high electron den-sity with a median value of 530 cmminus3 (Section 42 andFig 5) solar or super-solar metallicity and a wide rangeof ionization parameter probed by the O32 ratio rangingfrom 011 to 224 (Section 44 and Fig 8) Our resultsshow that the detected fast winds are most likely drivenby stellar feedback resulting from the extreme centralstarburst rather than by a buried AGN (Sections 52and 53)4) We present multiple intriguing observational signa-

tures that suggest that these galaxies may have substan-tial LyC photon leakage (Section 54) They have highΣSFR and ionization parameters comparable to those ofthe most massive known LyC leakers as traced by theO32 ratio They also lack gas in absorption near the sys-temic redshift and exhibit relatively weak Balmer emis-sion lines Finally they show remarkably weak [S II]lines compared to normal star-forming galaxies As ourgalaxies are distinct from known LyC leakers in manyregards (eg Mlowast and sSFR) this sample presents anexcellent chance to isolate which physical properties aremost closely connected to LyC escapeThe compact starburst galaxies in our sample provide a

unique opportunity to study star formation and feedbackat its most extreme In a related paper we find that thesegalaxies are likely observed during a short-lived but po-tentially key phase of massive galaxy evolution (Whalenet al in preparation) They have ΣSFR values approach-ing the Eddington limit associated with stellar radiationpressure feedback (Thompson et al 2005) and much oftheir gas may be violently blown out by powerful out-flows that open up channels for LyC photons to escapeIn a series of forthcoming papers based on high-

resolution KeckHIRES and integral field unitKeckKCWI spectra we will focus on deriving robustmeasurements of the physical properties morphologyand extent of the galactic outflows in our sample Suchdata on these unique galaxies provide strong observa-tional constraints to theoretical simulations that aimto produce realistic galactic outflows The comparisonof outflow characteristics between simulations andobservations will advance our understanding of galacticfeedback particularly from stellar processes during acrucial phase of massive galaxy evolution

ACKNOWLEDGEMENTS

We thank the referee for herhis time to provide a con-structive report The refereersquos thoughtful comments havehelped to improve the clarity of the manuscript We ac-knowledge support from the National Science Founda-tion (NSF) under a collaborative grant (AST-18132991813365 1814233 1813702 and 1814159) and from theHeising-Simons Foundation grant 2019-1659 S P andA L C acknowledge support from the Ingrid and Joseph

W Hibben endowed chair at UC San Diego The datapresented herein were obtained at the W M Keck Ob-servatory which is operated as a scientific partnershipamong the California Institute of Technology the Uni-versity of California and the National Aeronautics andSpace Administration The Observatory was made pos-sible by the generous financial support of the W M KeckFoundation The authors wish to recognize and acknowl-edge the very significant cultural role and reverence thatthe summit of Maunakea has always had within the in-digenous Hawaiian community We are most fortunate tohave the opportunity to conduct observations from thismountain

REFERENCES

Abel N P amp Satyapal S 2008 ApJ 678 686Aihara H Allende Prieto C An D et al 2011 ApJS 193 29Alexandroff R M Heckman T M Borthakur S Overzier R

amp Leitherer C 2015 ApJ 810 104Allen M G Groves B A Dopita M A Sutherland R S amp

Kewley L J 2008 ApJS 178 20Allington-Smith J Murray G Content R et al 2002 PASP

114 892Arribas S Colina L Bellocchi E Maiolino R amp

Villar-Martın M 2014 AampA 568 A14Asmus D Gandhi P Smette A Honig S F amp Duschl W J

2011 AampA 536 A36Assef R J Stern D Kochanek C S et al 2013 ApJ 772 26Bae H-J amp Woo J-H 2018 ApJ 853 185Baldwin J A Phillips M M amp Terlevich R 1981 PASP 93 5Bassett R Ryan-Weber E V Cooke J et al 2019 MNRAS

483 5223Becker G D Bolton J S amp Lidz A 2015 PASA 32 e045Belfiore F Maiolino R amp Bothwell M 2016a MNRAS 455

1218Belfiore F Maiolino R Maraston C et al 2016b MNRAS

461 3111Best P N Rottgering H J A amp Longair M S 2000

MNRAS 311 23Bian F Fan X McGreer I Cai Z amp Jiang L 2017 ApJL

837 L12Bian F Fan X Bechtold J et al 2010 ApJ 725 1877Bochanski J J Hennawi J F Simcoe R A et al 2009

PASP 121 1409Boera E Becker G D Bolton J S amp Nasir F 2019 ApJ

872 101Borthakur S Heckman T M Leitherer C amp Overzier R A

2014 Science 346 216Bouwens R J Illingworth G D Oesch P A et al 2012

ApJL 752 L5Calzetti D Armus L Bohlin R C et al 2000 ApJ 533 682Cappellari M 2017 MNRAS 466 798Cappellari M amp Emsellem E 2004 PASP 116 138Cardelli J A Clayton G C amp Mathis J S 1989 ApJ 345

245Carswell R F amp Webb J K 2014 VPFIT Voigt profile fitting

program ascl1408015Cen R 2020 ApJL 889 L22Chen H-W Wild V Tinker J L et al 2010 ApJL 724 L176Chisholm J Tremonti C A Leitherer C amp Chen Y 2017

MNRAS 469 4831Chisholm J Tremonti C A Leitherer C et al 2015 ApJ

811 149Choi J Dotter A Conroy C et al 2016 ApJ 823 102Cicone C Maiolino R amp Marconi A 2016 AampA 588 A41Cid Fernandes R Stasinska G Schlickmann M S et al 2010

MNRAS 403 1036Cimatti A Cassata P Pozzetti L et al 2008 AampA 482 21Concas A Popesso P Brusa M et al 2017 AampA 606 A36Conroy C amp Gunn J E 2010 ApJ 712 833Conroy C Gunn J E amp White M 2009 ApJ 699 486Conroy C Villaume A van Dokkum P G amp Lind K 2018

ApJ 854 139Davies R L Forster Schreiber N M Ubler H et al 2019

ApJ 873 122Dawson K S Kneib J-P Percival W J et al 2016 AJ 151

44de Barros S Vanzella E Amorın R et al 2016 AampA 585

A51Di Matteo P Combes F Melchior A L amp Semelin B 2007

AampA 468 61

22 Perrotta et al

Diamond-Stanic A M Coil A L Moustakas J et al 2016ApJ 824 24

Diamond-Stanic A M Moustakas J Tremonti C A et al2012 ApJ 755 L26

Diamond-Stanic A M Moustakas J Sell P H et al 2021arXiv e-prints arXiv210211287

Dopita M A Kewley L J Heisler C A amp Sutherland R S2000 ApJ 542 224

Dopita M A amp Sutherland R S 1995 ApJ 455 468Dopita M A Sutherland R S Nicholls D C Kewley L J

amp Vogt F P A 2013 ApJS 208 10Epinat B Contini T Finley H et al 2018 AampA 609 A40Evans I N amp Dopita M A 1985 ApJS 58 125mdash 1986 ApJL 310 L15Fan X Carilli C L amp Keating B 2006 ARAampA 44 415Fletcher T J Tang M Robertson B E et al 2019 ApJ 878

87Fluetsch A Maiolino R Carniani S et al 2019 MNRAS

483 4586mdash 2020 arXiv e-prints arXiv200613232Forster Schreiber N M amp Wuyts S 2020 ARAampA 58 661Forster Schreiber N M Ubler H Davies R L et al 2019

ApJ 875 21Freeman W R Siana B Kriek M et al 2019 ApJ 873 102Gabor J M amp Bournaud F 2014 MNRAS 441 1615Geach J E Hickox R C Diamond-Stanic A M et al 2013

ApJ 767 L17mdash 2014 Nature 516 68Geach J E Tremonti C Diamond-Stanic A M et al 2018

ArXiv e-prints arXiv180709789Genzel R Newman S Jones T et al 2011 ApJ 733 101Genzel R Forster Schreiber N M Lang P et al 2014 ApJ

785 75Gilli R Vignali C Mignoli M et al 2010 AampA 519 A92Gomes J M Papaderos P Kehrig C et al 2016 AampA 588

A68Guseva N G Izotov Y I Schaerer D et al 2020 MNRAS

497 4293Hainline K N Shapley A E Kornei K A et al 2009 ApJ

701 52Hauszligler B Bamford S P Vika M et al 2013 MNRAS 430

330Heckman T Borthakur S Wild V Schiminovich D amp

Bordoloi R 2017 ApJ 846 151Heckman T M Alexandroff R M Borthakur S Overzier R

amp Leitherer C 2015 ApJ 809 147Heckman T M amp Borthakur S 2016 ApJ 822 9Heckman T M Lehnert M D Strickland D K amp Armus L

2000 ApJS 129 493Hook I M Joslashrgensen I Allington-Smith J R et al 2004

PASP 116 425Hoopes C G amp Walterbos R A M 2003 ApJ 586 902Hopkins P F Hernquist L Cox T J amp Keres D 2008

ApJS 175 356Hopkins P F Quataert E amp Murray N 2012 MNRAS 421

3522Hopkins P F Younger J D Hayward C C Narayanan D

amp Hernquist L 2010 MNRAS 402 1693Izotov Y I Orlitova I Schaerer D et al 2016a Nature 529

178Izotov Y I Schaerer D Thuan T X et al 2016b MNRAS

461 3683Izotov Y I Schaerer D Worseck G et al 2018a MNRAS

474 4514Izotov Y I Worseck G Schaerer D et al 2018b MNRAS

478 4851Johnson B D Leja J Conroy C amp Speagle J S 2021 ApJS

254 22Kacprzak G G Muzahid S Churchill C W Nielsen N M

amp Charlton J C 2015 ApJ 815 22Kauffmann G Heckman T M Tremonti C et al 2003

MNRAS 346 1055Keenan R P Oey M S Jaskot A E amp James B L 2017

ApJ 848 12Kehrig C Monreal-Ibero A Papaderos P et al 2012 AampA

540 A11Kehrig C Vılchez J M Perez-Montero E et al 2016

MNRAS 459 2992Kennicutt R C amp Evans N J 2012 ARAampA 50 531Kewley L J amp Dopita M A 2002 ApJS 142 35Kewley L J Dopita M A Sutherland R S Heisler C A amp

Trevena J 2001 ApJ 556 121Kewley L J Groves B Kauffmann G amp Heckman T 2006

MNRAS 372 961Kewley L J Maier C Yabe K et al 2013 ApJL 774 L10Kewley L J Nicholls D C amp Sutherland R S 2019

ARAampA 57 511

Khaire V Srianand R Choudhury T R amp Gaikwad P 2016MNRAS 457 4051

Kim K Malhotra S Rhoads J E Butler N R amp Yang H2020 ApJ 893 134

Kim K-T amp Koo B-C 2001 ApJ 549 979Kim S Prato L amp McLean I 2015 REDSPEC NIRSPEC

data reduction ascl1507017Komatsu E Smith K M Dunkley J et al 2011 ApJS 192

18Kormendy J amp Sanders D B 1992 ApJL 390 L53Kornei K A Shapley A E Martin C L et al 2012 ApJ

758 135Kroupa P 2001 MNRAS 322 231Kroupa P Subr L Jerabkova T amp Wang L 2020 MNRAS

498 5652Labrie K Anderson K Cardenes R Simpson C amp Turner

J E H 2019 in Astronomical Society of the PacificConference Series Vol 523 Astronomical Data AnalysisSoftware and Systems XXVII ed P J Teuben M W PoundB A Thomas amp E M Warner 321

Lehnert M D amp Heckman T M 1996 ApJ 472 546Lehnert M D Heckman T M amp Weaver K A 1999 ApJ

523 575Lehnert M D Nesvadba N P H Le Tiran L et al 2009

ApJ 699 1660Leitet E Bergvall N Hayes M Linne S amp Zackrisson E

2013 AampA 553 A106Leitherer C Hernandez S Lee J C amp Oey M S 2016 ApJ

823 64Leja J Carnall A C Johnson B D Conroy C amp Speagle

J S 2019 ApJ 876 3Lilly S J Carollo C M Pipino A Renzini A amp Peng Y

2013 ApJ 772 119Madsen G J Reynolds R J amp Haffner L M 2006 ApJ 652

401Marshall J L Burles S Thompson I B et al 2008 in Society

of Photo-Optical Instrumentation Engineers (SPIE) ConferenceSeries Vol 7014 Ground-based and Airborne Instrumentationfor Astronomy II ed I S McLean amp M M Casali 701454

Martin C L 1998 ApJ 506 222mdash 2005 ApJ 621 227Martin C L amp Bouche N 2009 ApJ 703 1394Martin C L Shapley A E Coil A L et al 2012 ApJ 760

127Masters D Faisst A amp Capak P 2016 ApJ 828 18Masters D McCarthy P Siana B et al 2014 ApJ 785 153McGaugh S S 1991 ApJ 380 140McLean I S Becklin E E Bendiksen O et al 1998 in

Society of Photo-Optical Instrumentation Engineers (SPIE)Conference Series Vol 3354 Infrared AstronomicalInstrumentation ed A M Fowler 566ndash578

McQuinn K B W van Zee L amp Skillman E D 2019 ApJ886 74

Medling A M U V Rich J A et al 2015 MNRAS 448 2301Mendoza C amp Bautista M A 2014 ApJ 785 91Meurer G R Heckman T M Lehnert M D Leitherer C amp

Lowenthal J 1997 AJ 114 54Micheva G Oey M S Keenan R P Jaskot A E amp James

B L 2018 ApJ 867 2Mineo S Gilfanov M Lehmer B D Morrison G E amp

Sunyaev R 2014 MNRAS 437 1698Morrissey P Matuszewski M Martin D C et al 2018 ApJ

864 93Moustakas J amp Kennicutt Robert C J 2006 ApJ 651 155Muratov A L Keres D Faucher-Giguere C-A et al 2015

MNRAS 454 2691Murray N Quataert E amp Thompson T A 2005 ApJ 618

569Naidu R P Forrest B Oesch P A Tran K-V H amp Holden

B P 2018 MNRAS 478 791Naidu R P Tacchella S Mason C A et al 2020 ApJ 892

109Nakajima K Ellis R S Robertson B E Tang M amp Stark

D P 2020 ApJ 889 161Newman S F Genzel R Forster-Schreiber N M et al 2012

ApJ 761 43Nielsen N M Kacprzak G G Muzahid S et al 2017 ApJ

834 148Oesch P A Bouwens R J Illingworth G D et al 2013 ApJ

773 75Oey M S Meurer G R Yelda S et al 2007 ApJ 661 801Osterbrock D E 1989 Astrophysics of gaseous nebulae and

active galactic nucleiOsterbrock D E amp Ferland G J 2006 Astrophysics of gaseous

nebulae and active galactic nucleiOuchi M Mobasher B Shimasaku K et al 2009 ApJ 706

1136


Pagel B E J Edmunds M G Blackwell D E Chun M Samp Smith G 1979 MNRAS 189 95

Pellegrini E W Oey M S Winkler P F et al 2012 ApJ755 40

Peng C Y Ho L C Impey C D amp Rix H-W 2002 AJ124 266

mdash 2010 AJ 139 2097Petter G C Kepley A A Hickox R C et al 2020 ApJ 901

138Pettini M Shapley A E Steidel C C et al 2001 ApJ 554

981Quider A M Pettini M Shapley A E amp Steidel C C 2009

MNRAS 398 1263Ramambason L Schaerer D Stasinska G et al 2020 AampA

644 A21Rich J A Dopita M A Kewley L J amp Rupke D S N

2010 ApJ 721 505Rich J A Kewley L J amp Dopita M A 2011 ApJ 734 87mdash 2014 ApJL 781 L12mdash 2015 ApJS 221 28Robertson B E Ellis R S Furlanetto S R amp Dunlop J S

2015 ApJL 802 L19Robertson B E Furlanetto S R Schneider E et al 2013

ApJ 768 71Rubin K H R Prochaska J X Koo D C et al 2014 ApJ

794 156Rubin K H R Prochaska J X Menard B et al 2011 ApJ

728 55Rubin K H R Weiner B J Koo D C et al 2010 ApJ 719

1503Rupke D 2018 Galaxies 6 138Rupke D S Veilleux S amp Sanders D B 2005 ApJS 160 115Rupke D S N amp Veilleux S 2013 ApJ 768 75Rupke D S N Coil A Geach J E et al 2019 Nature 574

643Salpeter E E 1955 ApJ 121 161Sanders D B amp Mirabel I F 1996 ARAampA 34 749Sanders D B Soifer B T Elias J H et al 1988 ApJ 325 74Sanders R L Shapley A E Zhang K amp Yan R 2017 ApJ

850 136Sanders R L Shapley A E Kriek M et al 2016 ApJ 816 23Sarzi M Falcon-Barroso J Davies R L et al 2006 MNRAS

366 1151Sell P H Tremonti C A Hickox R C et al 2014 MNRAS

441 3417Shapiro K L Genzel R Quataert E et al 2009 ApJ 701

955Shapley A E Steidel C C Strom A L et al 2016 ApJL

826 L24Shapley A E Sanders R L Shao P et al 2019 ApJL 881

L35Sharma M Theuns T Frenk C et al 2016 MNRAS 458 L94Sharp R G amp Bland-Hawthorn J 2010 ApJ 711 818Shirazi M Brinchmann J amp Rahmati A 2014 ApJ 787 120Silk J amp Rees M J 1998 AampA 331 L1Singh R van de Ven G Jahnke K et al 2013 AampA 558 A43Smolcic V Novak M Delvecchio I et al 2017 AampA 602 A6Soto K T Martin C L Prescott M K M amp Armus L

2012 ApJ 757 86Spilker J S Aravena M Phadke K A et al 2020 ApJ 905

86Springel V Di Matteo T amp Hernquist L 2005 MNRAS 361

776Steidel C C Bogosavljevic M Shapley A E et al 2018 ApJ

869 123

Steidel C C Erb D K Shapley A E et al 2010 ApJ 717289

Steidel C C Rudie G C Strom A L et al 2014 ApJ 795165

Stern D Assef R J Benford D J et al 2012 ApJ 753 30Storey P J amp Zeippen C J 2000 MNRAS 312 813Strickland D K amp Heckman T M 2007 ApJ 658 258mdash 2009 ApJ 697 2030Strickland D K Heckman T M Colbert E J M Hoopes

C G amp Weaver K A 2004 ApJ 606 829Strom A L Steidel C C Rudie G C Trainor R F amp

Pettini M 2018 ApJ 868 117Strom A L Steidel C C Rudie G C et al 2017 ApJ 836

164Swinbank A M Harrison C M Tiley A L et al 2019

MNRAS 487 381Tanner R Cecil G amp Heitsch F 2017 ApJ 843 137Tayal S S amp Zatsarinny O 2010 ApJS 188 32Thompson T A Quataert E amp Murray N 2005 ApJ 630

167Tremonti C A Moustakas J amp Diamond-Stanic A M 2007

ApJ 663 L77

Tremonti C A Heckman T M Kauffmann G et al 2004ApJ 613 898

van der Wel A Franx M van Dokkum P G et al 2014 ApJ788 28

Vanzella E de Barros S Vasei K et al 2016 ApJ 825 41Vanzella E Balestra I Gronke M et al 2017 MNRAS 465

3803Vanzella E Caminha G B Calura F et al 2020 MNRAS

491 1093Veilleux S Maiolino R Bolatto A D amp Aalto S 2020

AampA Rev 28 2Veilleux S amp Osterbrock D E 1987 ApJS 63 295Vergani D Garilli B Polletta M et al 2018 AampA 620 A193Verhamme A Orlitova I Schaerer D amp Hayes M 2015

AampA 578 A7Verhamme A Orlitova I Schaerer D et al 2017 AampA 597

A13Vika M Bamford S P Hauszligler B et al 2013 MNRAS 435

623Voges E S amp Walterbos R A M 2006 ApJL 644 L29Walter F Weiss A amp Scoville N 2002 ApJL 580 L21Wang B Heckman T M Leitherer C et al 2019 ApJ 885

57Weilbacher P M Monreal-Ibero A Kollatschny W et al

2015 AampA 582 A114Weinberger R Springel V Pakmor R et al 2018 MNRAS

479 4056Whitaker K E Franx M Leja J et al 2014 ApJ 795 104Wise J H Demchenko V G Halicek M T et al 2014

MNRAS 442 2560Wright E L Eisenhardt P R M Mainzer A K et al 2010

AJ 140 1868Yajima H Choi J-H amp Nagamine K 2011 MNRAS 412 411York D G Adelman J Anderson John E J et al 2000 AJ

120 1579Yuan T T Kewley L J amp Sanders D B 2010 ApJ 709 884Zahn O Reichardt C L Shaw L et al 2012 ApJ 756 65Zaritsky D Kennicutt Robert C J amp Huchra J P 1994a

ApJ 420 87mdash 1994b ApJ 420 87Zastrow J Oey M S Veilleux S McDonald M amp Martin

C L 2011 ApJL 741 L17Zhang K Yan R Bundy K et al 2017 MNRAS 466 3217Zurita A Rozas M amp Beckman J E 2000 AampA 363 9

2 Perrotta et al

by ionized metals such as N V O VI (Kacprzak et al2015 Nielsen et al 2017) Finally hot gas (iegt 106 K) probed with both hard and soft X-rayemission (Lehnert et al 1999 Strickland et al 2004Strickland amp Heckman 2007 2009)While powerful outflows appear to be essential to

rapidly shut off star formation the physical drivers ofthis ejective feedback remain unclear In particular therelative role of feedback from stars versus supermassiveblack holes (SMBHs) in quenching star formation in mas-sive galaxies is still widely debated (eg Hopkins et al2012 Gabor amp Bournaud 2014 Weinberger et al 2018Kroupa et al 2020) In this context the observed corre-lations between outflow and host galaxy properties canprovide some insight (Rubin et al 2014 Tanner et al2017) For instance considering galaxy samples with awide dynamic range of intrinsic properties the outflowvelocity is found to scale with stellar mass (Mlowast) starformation rate (SFR) and SFR surface density (ΣSFR)This suggests that the faster outflows tend to be hostedin massive galaxies with high and concentrated star for-mation (eg Tanner et al 2017) implying that the star-burst phase could potentially drive impactful outflowsStudying galaxies with extreme physical conditions canprovide constraints on astrophysical feedback processesOur team has been investigating a sample of galax-

ies at z = 04 - 08 initially selected from the SloanDigital Sky Survey (SDSS York et al 2000) Data Re-lease 8 (DR8 Aihara et al 2011) to have distinct sig-natures of young post-starburst galaxies Their spec-tra are characterized by strong stellar Balmer absorp-tion from B- and A-stars and weak or absent nebularemission lines indicating minimal on-going star forma-tion They lie on the massive end of the stellar massfunction (Mlowast sim 1011 M⊙ Diamond-Stanic et al 2012)Remarkably the optical spectra of most of these ob-jects exhibit evidence of ejective feedback traced by ex-tremely blueshifted (gt 1000 km sminus1) Mg II λλ27962803interstellar absorption lines (Tremonti et al 2007 Daviset al in prep) The Mg II kinematics imply galacticoutflows much faster than the sim500 km sminus1ones typicalof massive star-forming galaxies (Chisholm et al 2017)This finding painted an interesting picture where thesegalaxies were thought to be post-starburst systems withpowerful outflows that may have played a crucial rolein quenching their star formation Surprisingly manyof these galaxies were detected in the Wide-field In-frared Survey Explorer (WISE Wright et al 2010) andthe modeling of their ultraviolet (UV) to near-IR spec-tral energy distribution (SED) suggested a high levelof heavily obscured star formation (gt 50 M⊙ yrminus1Diamond-Stanic et al 2012) Hubble Space Telescope(HST) imaging of 29 of these galaxies revealed theyare extremely compact (Re sim few 100 pc) Moreoverthese data showed complex morphologies with diffusetidal features indicative of various major merger stages(Sell et al 2014 Diamond-Stanic et al 2021) Combin-ing SFR estimates from WISE restframe mid-IR lumi-nosities with physical size measures from HST imagingwe derived extraordinarily high ΣSFR sim 103 M⊙ yrminus1

kpcminus2 (Diamond-Stanic et al 2012) approaching thetheoretical Eddington limit (Lehnert amp Heckman 1996Meurer et al 1997 Murray et al 2005 Thompson et al2005)

These results led us to draw a new scenario where thesestarburst galaxies have a dense dusty star-forming coreat the center of the galaxy and a substantial part of theirgas and dust is blown away by powerful outflows In thiscontext the high ΣSFR may reasonably be the driver ofthe exceptionally fast gas outflows seen which in turnmay be responsible for the onset of rapid star formationquenching Millimeter data for two galaxies in our sam-ple indicates that the molecular gas is being consumedby the starburst with exceptional efficiency (Geach et al2013) and expelled in an extended molecular outflow(Geach et al 2014) leading to rapid gas depletion timesInterestingly Sell et al (2014) used a suite of multiwave-length observations to assess the AGN activity in a sub-sample of these starbursts and found little evidence forcurrent AGN activity in half of the sample (lt 10 percent of the total bolometric luminosity) though pastAGN episodes could not be ruled out This finding isin line with stellar feedback being the main driver of theobserved outflows These compact starburst galaxies ex-hibit the fastest outflows (gt 1000 km sminus1) and highestΣSFR among star-forming galaxies at any redshift there-fore they are an exquisite laboratory to test the limits ofstellar feedback They could represent a brief but com-mon phase of massive galaxy evolutionOur team followed up one of these starburst galax-

ies (J2118 or Makani) with Keck Cosmic Web Imager(KCWI Morrissey et al 2018) The data reveal a spec-tacular galactic outflow traced by [O II] emission linereaching far into the circumgalactic medium (CGM) ofthe galaxy (Rupke et al 2019) The [O II] emission hasa classic bipolar hourglass limb-brightened shape andexhibits a complex structure a larger-scale slower out-flow (sim300 km sminus1) and a smaller-scale faster outflow(sim1500 km sminus1) The velocities and sizes of these twooutflows map exactly to two previous starburst episodesthat this galaxy experienced detected through the rest-frame optical emission and inferred ages of stars in thisgalaxy These outflows are therefore consistent withbeing formed during recent starburst episodes in thisgalaxyrsquos past The KCWI data on Makani directly showsthat galactic outflows feed the CGM expelling gas farbeyond the stars in galaxiesIn this paper we present new optical and near-IR ob-

servations for 14 of the most well-studied starburst galax-ies in our sample We use this in combination with someancillary data to characterize their extreme ejective feed-back events and explore their potential role in quenchingthe star formation in the host systems Our unique dataset includes both emission and absorption lines that al-low us to probe outflowing gas at different densities Weinvestigate both the nature of the outflows as well asthe physical conditions in the central dusty starburstWe use an ensemble of line ratio diagrams as crucial di-agnostics of gas ionization electron density and metal-licityThe paper is organized as follows Section 2 illustrates

the sample selection observations and data reductionSection 3 describes our measurements of the emissionline kinematics Section 4 presents our main results incomparison to other relevant galaxy samples and Section5 discusses the more comprehensive implications of ouranalysis Our conclusions are reviewed in Section 6Throughout the paper we assume a standard ΛCDM






21 NIRSPEC



22 GMOS






4 Perrotta et al

0

10

20

30

40 J0106minus1023

0

10

20

30J1341minus0321

0

10

20 J0826minus4305

0510152025J1506+5402

0

10

20

30J0901+0314

0510152025J1613+2834

0

10

20

Flu

x(1

0minus

17

erg

sminus1

cmminus

2Aminus

1)

J0905+5759

0510152025J1622+3145

0

5

10

15 J0944+0930

0

10

20J1713+2817

05

10152025 J1107+0417

0510152025J2116minus0634

3000 3500 4000 4500 5000

05

10152025 J1125minus0145

3000 3500 4000 4500 5000

0510152025J2118+0017

Rest Wavelength (A)









(1) (2) (3) (4) (5) (6) (7) (8) (9)

J0106-1023 045 16601056 -10391647 1072 0590 166+33minus31 76 -1650

J0826+4305 060 12666006 43091498 1063 0173 184+53minus41 981 -1425

J0901+0314 046 13538926 32367997 1066 0237 99+39minus26 281 -1602

J0905+5759 071 13634832 57986791 1069 0097 90+23minus20 1519 -2910

J0944+0930 051 14607437 95053855 1059 0114 88+26minus21 1074 -1679

J1107+0417 047 16676197 42840984 1060 0273 73+13minus14 155 -2093

J1125-0145 052 17132874 -17590066 1103 0600 227+104minus68 100 -2309

J1341-0321 066 20540333 -33570199 1053 0117 151+34minus23 1756 -1936

J1506+5402 061 22665124 54039095 1060 0168 116+32minus25 652 -2018

J1613+2834 045 24338552 28570772 1112 0949 172+36minus36 30 -2699


J1713+2817 058 25825161 28285631 1089 0173 229+99minus72 1218 -1298

J2116-0624 073 31910479 -65791139 1041 0284 110+55minus27 216 -2069

J2118+0017 046 31960026 02915070 1095 2240 230+93minus76 5 -1448








6 Perrotta et al


















0

4

8J0106 M

[OII]

02

4

6 S

Hβ

01234 S

[OIII]

0

10

20 N

Hα

0

1

2

N

[SII]

0246 J0826 M

0

2

4G

0123 G

0

10

20 N

0123 N

0246 J0901 M

0

4

8S

01234 S

0102030

N

0

2

4J0905 M

01234 G

0

1

2 G

0

10

20N

0123 J0944 M

0

2

4S

0

1

2S

05

1015 N

0

2 N

0

2

4 J1107 Ma

0246 Ma

01234

Ma

0

10

20 N

0

2N

0123

Flu

x(1

0minus

17

erg

sminus1

cmminus

2Aminus

1)

J1125 M

0

5

10 N

0

2

4

6J1341 Ma

0246 Ma

0

1

2

3Ma

0

10

20N

0

2N

02468 J1506 M

0

5

10G

0

4

8 G

0102030

N

0

2

4N

0

5

10

15 J1613 M

0

2

4 G

0

1

2 G

05

1015

G

0

2G

02468 J1622 S

0246 G

0

1

2 G

05

1015 G

0

1

2 G

0123

J1713 M

0

1S

0

5

10 S

0123 N

0123 J2116 M

0

4

8 N

0

2N

3700 3720 3740

0

5

10

15J2118 M

4840 4860 4880

0246 S

4950 4990 5030

Rest Wavelength (A)

0246 S

6540 6570 6600

0

10

20 N

6690 6720 6750

0

2N


8 Perrotta et al







4 RESULTS


41 Kinematics






-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0106

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0826

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0901

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0905

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0944

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1107

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1125

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1341

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1506

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1613

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1622

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1713

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2116

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2118

Narrow Hα

Broad Hα or Hβ

Broad [OIII] 5007

Broad [OII] 3729

MgII 2796

FeII 2586


Norm

aliz

edF

lux





10 Perrotta et al



42 Electron Density


















minus3000

minus2500

minus2000

minus1500

minus1000

minus500

0

Bro

ad[O

II]v

98[k

msminus

1]



J0106

J0826

J0901

J0905

J0944

J1107

J1125

J1341

J1506

J1613

J1622

J1713

J2116

J2118


J0106J0826

J0944J1107

J1341J1506

J1613J1622

J2116J2118

100

101

102

103

104

Ele

ctro

nD

ensi

ty[c

mminus

3]







12 Perrotta et al

minus1

0

1

Total Total


log([NII]λ6585Hα)

minus1

0

1

log([

OII

I]λ

5007H

β)

Narrow

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118


log([SII]Hα)

Narrow



minus1

0

1

log([

OII

I]λ

5007H

β)

Broad

Narrow

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2118



















14 Perrotta et al







5 DISCUSSION




minus10

minus05

00

05

10

15lo

g([

NII

] 65

85[

SII

] 67

176

73

1)

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2116

J2118

7 8 9 10 11 12

log(MlowastM⊙)

minus10

minus05

00

05

10

15

log

([O

III]

50

07[

OII

] 37

263

72

9)

Wang et al 2019

Bassett et al 2019

Fletcher et al 2019

Izotov et al 201618


0 1 2


minus2 0 2 4






16 Perrotta et al
















95 100 105 110 115log(MlowastM⊙)

10minus1

100

101

Fb

roadF

nar

row

minus2 0 2 4


Genzel et al 2014

Freeman et al 2019

Swinbank et al 2019


Newman et al 2012

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2116

J2118










18 Perrotta et al




















20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77























21 NIRSPEC



22 GMOS






4 Perrotta et al

0

10

20

30

40 J0106minus1023

0

10

20

30J1341minus0321

0

10

20 J0826minus4305

0510152025J1506+5402

0

10

20

30J0901+0314

0510152025J1613+2834

0

10

20

Flu

x(1

0minus

17

erg

sminus1

cmminus

2Aminus

1)

J0905+5759

0510152025J1622+3145

0

5

10

15 J0944+0930

0

10

20J1713+2817

05

10152025 J1107+0417

0510152025J2116minus0634

3000 3500 4000 4500 5000

05

10152025 J1125minus0145

3000 3500 4000 4500 5000

0510152025J2118+0017

Rest Wavelength (A)









(1) (2) (3) (4) (5) (6) (7) (8) (9)

J0106-1023 045 16601056 -10391647 1072 0590 166+33minus31 76 -1650

J0826+4305 060 12666006 43091498 1063 0173 184+53minus41 981 -1425

J0901+0314 046 13538926 32367997 1066 0237 99+39minus26 281 -1602

J0905+5759 071 13634832 57986791 1069 0097 90+23minus20 1519 -2910

J0944+0930 051 14607437 95053855 1059 0114 88+26minus21 1074 -1679

J1107+0417 047 16676197 42840984 1060 0273 73+13minus14 155 -2093

J1125-0145 052 17132874 -17590066 1103 0600 227+104minus68 100 -2309

J1341-0321 066 20540333 -33570199 1053 0117 151+34minus23 1756 -1936

J1506+5402 061 22665124 54039095 1060 0168 116+32minus25 652 -2018

J1613+2834 045 24338552 28570772 1112 0949 172+36minus36 30 -2699


J1713+2817 058 25825161 28285631 1089 0173 229+99minus72 1218 -1298

J2116-0624 073 31910479 -65791139 1041 0284 110+55minus27 216 -2069

J2118+0017 046 31960026 02915070 1095 2240 230+93minus76 5 -1448








6 Perrotta et al


















0

4

8J0106 M

[OII]

02

4

6 S

Hβ

01234 S

[OIII]

0

10

20 N

Hα

0

1

2

N

[SII]

0246 J0826 M

0

2

4G

0123 G

0

10

20 N

0123 N

0246 J0901 M

0

4

8S

01234 S

0102030

N

0

2

4J0905 M

01234 G

0

1

2 G

0

10

20N

0123 J0944 M

0

2

4S

0

1

2S

05

1015 N

0

2 N

0

2

4 J1107 Ma

0246 Ma

01234

Ma

0

10

20 N

0

2N

0123

Flu

x(1

0minus

17

erg

sminus1

cmminus

2Aminus

1)

J1125 M

0

5

10 N

0

2

4

6J1341 Ma

0246 Ma

0

1

2

3Ma

0

10

20N

0

2N

02468 J1506 M

0

5

10G

0

4

8 G

0102030

N

0

2

4N

0

5

10

15 J1613 M

0

2

4 G

0

1

2 G

05

1015

G

0

2G

02468 J1622 S

0246 G

0

1

2 G

05

1015 G

0

1

2 G

0123

J1713 M

0

1S

0

5

10 S

0123 N

0123 J2116 M

0

4

8 N

0

2N

3700 3720 3740

0

5

10

15J2118 M

4840 4860 4880

0246 S

4950 4990 5030

Rest Wavelength (A)

0246 S

6540 6570 6600

0

10

20 N

6690 6720 6750

0

2N


8 Perrotta et al







4 RESULTS


41 Kinematics






-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0106

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0826

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0901

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0905

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0944

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1107

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1125

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1341

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1506

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1613

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1622

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1713

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2116

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2118

Narrow Hα

Broad Hα or Hβ

Broad [OIII] 5007

Broad [OII] 3729

MgII 2796

FeII 2586


Norm

aliz

edF

lux





10 Perrotta et al



42 Electron Density


















minus3000

minus2500

minus2000

minus1500

minus1000

minus500

0

Bro

ad[O

II]v

98[k

msminus

1]



J0106

J0826

J0901

J0905

J0944

J1107

J1125

J1341

J1506

J1613

J1622

J1713

J2116

J2118


J0106J0826

J0944J1107

J1341J1506

J1613J1622

J2116J2118

100

101

102

103

104

Ele

ctro

nD

ensi

ty[c

mminus

3]







12 Perrotta et al

minus1

0

1

Total Total


log([NII]λ6585Hα)

minus1

0

1

log([

OII

I]λ

5007H

β)

Narrow

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118


log([SII]Hα)

Narrow



minus1

0

1

log([

OII

I]λ

5007H

β)

Broad

Narrow

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2118



















14 Perrotta et al







5 DISCUSSION




minus10

minus05

00

05

10

15lo

g([

NII

] 65

85[

SII

] 67

176

73

1)

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2116

J2118

7 8 9 10 11 12

log(MlowastM⊙)

minus10

minus05

00

05

10

15

log

([O

III]

50

07[

OII

] 37

263

72

9)

Wang et al 2019

Bassett et al 2019

Fletcher et al 2019

Izotov et al 201618


0 1 2


minus2 0 2 4






16 Perrotta et al
















95 100 105 110 115log(MlowastM⊙)

10minus1

100

101

Fb

roadF

nar

row

minus2 0 2 4


Genzel et al 2014

Freeman et al 2019

Swinbank et al 2019


Newman et al 2012

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2116

J2118










18 Perrotta et al




















20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77


















4 Perrotta et al

0

10

20

30

40 J0106minus1023

0

10

20

30J1341minus0321

0

10

20 J0826minus4305

0510152025J1506+5402

0

10

20

30J0901+0314

0510152025J1613+2834

0

10

20

Flu

x(1

0minus

17

erg

sminus1

cmminus

2Aminus

1)

J0905+5759

0510152025J1622+3145

0

5

10

15 J0944+0930

0

10

20J1713+2817

05

10152025 J1107+0417

0510152025J2116minus0634

3000 3500 4000 4500 5000

05

10152025 J1125minus0145

3000 3500 4000 4500 5000

0510152025J2118+0017

Rest Wavelength (A)









(1) (2) (3) (4) (5) (6) (7) (8) (9)

J0106-1023 045 16601056 -10391647 1072 0590 166+33minus31 76 -1650

J0826+4305 060 12666006 43091498 1063 0173 184+53minus41 981 -1425

J0901+0314 046 13538926 32367997 1066 0237 99+39minus26 281 -1602

J0905+5759 071 13634832 57986791 1069 0097 90+23minus20 1519 -2910

J0944+0930 051 14607437 95053855 1059 0114 88+26minus21 1074 -1679

J1107+0417 047 16676197 42840984 1060 0273 73+13minus14 155 -2093

J1125-0145 052 17132874 -17590066 1103 0600 227+104minus68 100 -2309

J1341-0321 066 20540333 -33570199 1053 0117 151+34minus23 1756 -1936

J1506+5402 061 22665124 54039095 1060 0168 116+32minus25 652 -2018

J1613+2834 045 24338552 28570772 1112 0949 172+36minus36 30 -2699


J1713+2817 058 25825161 28285631 1089 0173 229+99minus72 1218 -1298

J2116-0624 073 31910479 -65791139 1041 0284 110+55minus27 216 -2069

J2118+0017 046 31960026 02915070 1095 2240 230+93minus76 5 -1448








6 Perrotta et al


















0

4

8J0106 M

[OII]

02

4

6 S

Hβ

01234 S

[OIII]

0

10

20 N

Hα

0

1

2

N

[SII]

0246 J0826 M

0

2

4G

0123 G

0

10

20 N

0123 N

0246 J0901 M

0

4

8S

01234 S

0102030

N

0

2

4J0905 M

01234 G

0

1

2 G

0

10

20N

0123 J0944 M

0

2

4S

0

1

2S

05

1015 N

0

2 N

0

2

4 J1107 Ma

0246 Ma

01234

Ma

0

10

20 N

0

2N

0123

Flu

x(1

0minus

17

erg

sminus1

cmminus

2Aminus

1)

J1125 M

0

5

10 N

0

2

4

6J1341 Ma

0246 Ma

0

1

2

3Ma

0

10

20N

0

2N

02468 J1506 M

0

5

10G

0

4

8 G

0102030

N

0

2

4N

0

5

10

15 J1613 M

0

2

4 G

0

1

2 G

05

1015

G

0

2G

02468 J1622 S

0246 G

0

1

2 G

05

1015 G

0

1

2 G

0123

J1713 M

0

1S

0

5

10 S

0123 N

0123 J2116 M

0

4

8 N

0

2N

3700 3720 3740

0

5

10

15J2118 M

4840 4860 4880

0246 S

4950 4990 5030

Rest Wavelength (A)

0246 S

6540 6570 6600

0

10

20 N

6690 6720 6750

0

2N


8 Perrotta et al







4 RESULTS


41 Kinematics






-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0106

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0826

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0901

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0905

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0944

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1107

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1125

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1341

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1506

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1613

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1622

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1713

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2116

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2118

Narrow Hα

Broad Hα or Hβ

Broad [OIII] 5007

Broad [OII] 3729

MgII 2796

FeII 2586


Norm

aliz

edF

lux





10 Perrotta et al



42 Electron Density


















minus3000

minus2500

minus2000

minus1500

minus1000

minus500

0

Bro

ad[O

II]v

98[k

msminus

1]



J0106

J0826

J0901

J0905

J0944

J1107

J1125

J1341

J1506

J1613

J1622

J1713

J2116

J2118


J0106J0826

J0944J1107

J1341J1506

J1613J1622

J2116J2118

100

101

102

103

104

Ele

ctro

nD

ensi

ty[c

mminus

3]







12 Perrotta et al

minus1

0

1

Total Total


log([NII]λ6585Hα)

minus1

0

1

log([

OII

I]λ

5007H

β)

Narrow

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118


log([SII]Hα)

Narrow



minus1

0

1

log([

OII

I]λ

5007H

β)

Broad

Narrow

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2118



















14 Perrotta et al







5 DISCUSSION




minus10

minus05

00

05

10

15lo

g([

NII

] 65

85[

SII

] 67

176

73

1)

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2116

J2118

7 8 9 10 11 12

log(MlowastM⊙)

minus10

minus05

00

05

10

15

log

([O

III]

50

07[

OII

] 37

263

72

9)

Wang et al 2019

Bassett et al 2019

Fletcher et al 2019

Izotov et al 201618


0 1 2


minus2 0 2 4






16 Perrotta et al
















95 100 105 110 115log(MlowastM⊙)

10minus1

100

101

Fb

roadF

nar

row

minus2 0 2 4


Genzel et al 2014

Freeman et al 2019

Swinbank et al 2019


Newman et al 2012

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2116

J2118










18 Perrotta et al




















20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77





















(1) (2) (3) (4) (5) (6) (7) (8) (9)

J0106-1023 045 16601056 -10391647 1072 0590 166+33minus31 76 -1650

J0826+4305 060 12666006 43091498 1063 0173 184+53minus41 981 -1425

J0901+0314 046 13538926 32367997 1066 0237 99+39minus26 281 -1602

J0905+5759 071 13634832 57986791 1069 0097 90+23minus20 1519 -2910

J0944+0930 051 14607437 95053855 1059 0114 88+26minus21 1074 -1679

J1107+0417 047 16676197 42840984 1060 0273 73+13minus14 155 -2093

J1125-0145 052 17132874 -17590066 1103 0600 227+104minus68 100 -2309

J1341-0321 066 20540333 -33570199 1053 0117 151+34minus23 1756 -1936

J1506+5402 061 22665124 54039095 1060 0168 116+32minus25 652 -2018

J1613+2834 045 24338552 28570772 1112 0949 172+36minus36 30 -2699


J1713+2817 058 25825161 28285631 1089 0173 229+99minus72 1218 -1298

J2116-0624 073 31910479 -65791139 1041 0284 110+55minus27 216 -2069

J2118+0017 046 31960026 02915070 1095 2240 230+93minus76 5 -1448








6 Perrotta et al


















0

4

8J0106 M

[OII]

02

4

6 S

Hβ

01234 S

[OIII]

0

10

20 N

Hα

0

1

2

N

[SII]

0246 J0826 M

0

2

4G

0123 G

0

10

20 N

0123 N

0246 J0901 M

0

4

8S

01234 S

0102030

N

0

2

4J0905 M

01234 G

0

1

2 G

0

10

20N

0123 J0944 M

0

2

4S

0

1

2S

05

1015 N

0

2 N

0

2

4 J1107 Ma

0246 Ma

01234

Ma

0

10

20 N

0

2N

0123

Flu

x(1

0minus

17

erg

sminus1

cmminus

2Aminus

1)

J1125 M

0

5

10 N

0

2

4

6J1341 Ma

0246 Ma

0

1

2

3Ma

0

10

20N

0

2N

02468 J1506 M

0

5

10G

0

4

8 G

0102030

N

0

2

4N

0

5

10

15 J1613 M

0

2

4 G

0

1

2 G

05

1015

G

0

2G

02468 J1622 S

0246 G

0

1

2 G

05

1015 G

0

1

2 G

0123

J1713 M

0

1S

0

5

10 S

0123 N

0123 J2116 M

0

4

8 N

0

2N

3700 3720 3740

0

5

10

15J2118 M

4840 4860 4880

0246 S

4950 4990 5030

Rest Wavelength (A)

0246 S

6540 6570 6600

0

10

20 N

6690 6720 6750

0

2N


8 Perrotta et al







4 RESULTS


41 Kinematics






-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0106

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0826

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0901

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0905

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0944

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1107

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1125

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1341

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1506

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1613

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1622

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1713

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2116

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2118

Narrow Hα

Broad Hα or Hβ

Broad [OIII] 5007

Broad [OII] 3729

MgII 2796

FeII 2586


Norm

aliz

edF

lux





10 Perrotta et al



42 Electron Density


















minus3000

minus2500

minus2000

minus1500

minus1000

minus500

0

Bro

ad[O

II]v

98[k

msminus

1]



J0106

J0826

J0901

J0905

J0944

J1107

J1125

J1341

J1506

J1613

J1622

J1713

J2116

J2118


J0106J0826

J0944J1107

J1341J1506

J1613J1622

J2116J2118

100

101

102

103

104

Ele

ctro

nD

ensi

ty[c

mminus

3]







12 Perrotta et al

minus1

0

1

Total Total


log([NII]λ6585Hα)

minus1

0

1

log([

OII

I]λ

5007H

β)

Narrow

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118


log([SII]Hα)

Narrow



minus1

0

1

log([

OII

I]λ

5007H

β)

Broad

Narrow

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2118



















14 Perrotta et al







5 DISCUSSION




minus10

minus05

00

05

10

15lo

g([

NII

] 65

85[

SII

] 67

176

73

1)

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2116

J2118

7 8 9 10 11 12

log(MlowastM⊙)

minus10

minus05

00

05

10

15

log

([O

III]

50

07[

OII

] 37

263

72

9)

Wang et al 2019

Bassett et al 2019

Fletcher et al 2019

Izotov et al 201618


0 1 2


minus2 0 2 4






16 Perrotta et al
















95 100 105 110 115log(MlowastM⊙)

10minus1

100

101

Fb

roadF

nar

row

minus2 0 2 4


Genzel et al 2014

Freeman et al 2019

Swinbank et al 2019


Newman et al 2012

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2116

J2118










18 Perrotta et al




















20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77


















6 Perrotta et al


















0

4

8J0106 M

[OII]

02

4

6 S

Hβ

01234 S

[OIII]

0

10

20 N

Hα

0

1

2

N

[SII]

0246 J0826 M

0

2

4G

0123 G

0

10

20 N

0123 N

0246 J0901 M

0

4

8S

01234 S

0102030

N

0

2

4J0905 M

01234 G

0

1

2 G

0

10

20N

0123 J0944 M

0

2

4S

0

1

2S

05

1015 N

0

2 N

0

2

4 J1107 Ma

0246 Ma

01234

Ma

0

10

20 N

0

2N

0123

Flu

x(1

0minus

17

erg

sminus1

cmminus

2Aminus

1)

J1125 M

0

5

10 N

0

2

4

6J1341 Ma

0246 Ma

0

1

2

3Ma

0

10

20N

0

2N

02468 J1506 M

0

5

10G

0

4

8 G

0102030

N

0

2

4N

0

5

10

15 J1613 M

0

2

4 G

0

1

2 G

05

1015

G

0

2G

02468 J1622 S

0246 G

0

1

2 G

05

1015 G

0

1

2 G

0123

J1713 M

0

1S

0

5

10 S

0123 N

0123 J2116 M

0

4

8 N

0

2N

3700 3720 3740

0

5

10

15J2118 M

4840 4860 4880

0246 S

4950 4990 5030

Rest Wavelength (A)

0246 S

6540 6570 6600

0

10

20 N

6690 6720 6750

0

2N


8 Perrotta et al







4 RESULTS


41 Kinematics






-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0106

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0826

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0901

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0905

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0944

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1107

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1125

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1341

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1506

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1613

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1622

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1713

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2116

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2118

Narrow Hα

Broad Hα or Hβ

Broad [OIII] 5007

Broad [OII] 3729

MgII 2796

FeII 2586


Norm

aliz

edF

lux





10 Perrotta et al



42 Electron Density


















minus3000

minus2500

minus2000

minus1500

minus1000

minus500

0

Bro

ad[O

II]v

98[k

msminus

1]



J0106

J0826

J0901

J0905

J0944

J1107

J1125

J1341

J1506

J1613

J1622

J1713

J2116

J2118


J0106J0826

J0944J1107

J1341J1506

J1613J1622

J2116J2118

100

101

102

103

104

Ele

ctro

nD

ensi

ty[c

mminus

3]







12 Perrotta et al

minus1

0

1

Total Total


log([NII]λ6585Hα)

minus1

0

1

log([

OII

I]λ

5007H

β)

Narrow

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118


log([SII]Hα)

Narrow



minus1

0

1

log([

OII

I]λ

5007H

β)

Broad

Narrow

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2118



















14 Perrotta et al







5 DISCUSSION




minus10

minus05

00

05

10

15lo

g([

NII

] 65

85[

SII

] 67

176

73

1)

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2116

J2118

7 8 9 10 11 12

log(MlowastM⊙)

minus10

minus05

00

05

10

15

log

([O

III]

50

07[

OII

] 37

263

72

9)

Wang et al 2019

Bassett et al 2019

Fletcher et al 2019

Izotov et al 201618


0 1 2


minus2 0 2 4






16 Perrotta et al
















95 100 105 110 115log(MlowastM⊙)

10minus1

100

101

Fb

roadF

nar

row

minus2 0 2 4


Genzel et al 2014

Freeman et al 2019

Swinbank et al 2019


Newman et al 2012

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2116

J2118










18 Perrotta et al




















20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77



















0

4

8J0106 M

[OII]

02

4

6 S

Hβ

01234 S

[OIII]

0

10

20 N

Hα

0

1

2

N

[SII]

0246 J0826 M

0

2

4G

0123 G

0

10

20 N

0123 N

0246 J0901 M

0

4

8S

01234 S

0102030

N

0

2

4J0905 M

01234 G

0

1

2 G

0

10

20N

0123 J0944 M

0

2

4S

0

1

2S

05

1015 N

0

2 N

0

2

4 J1107 Ma

0246 Ma

01234

Ma

0

10

20 N

0

2N

0123

Flu

x(1

0minus

17

erg

sminus1

cmminus

2Aminus

1)

J1125 M

0

5

10 N

0

2

4

6J1341 Ma

0246 Ma

0

1

2

3Ma

0

10

20N

0

2N

02468 J1506 M

0

5

10G

0

4

8 G

0102030

N

0

2

4N

0

5

10

15 J1613 M

0

2

4 G

0

1

2 G

05

1015

G

0

2G

02468 J1622 S

0246 G

0

1

2 G

05

1015 G

0

1

2 G

0123

J1713 M

0

1S

0

5

10 S

0123 N

0123 J2116 M

0

4

8 N

0

2N

3700 3720 3740

0

5

10

15J2118 M

4840 4860 4880

0246 S

4950 4990 5030

Rest Wavelength (A)

0246 S

6540 6570 6600

0

10

20 N

6690 6720 6750

0

2N


8 Perrotta et al







4 RESULTS


41 Kinematics






-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0106

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0826

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0901

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0905

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0944

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1107

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1125

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1341

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1506

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1613

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1622

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1713

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2116

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2118

Narrow Hα

Broad Hα or Hβ

Broad [OIII] 5007

Broad [OII] 3729

MgII 2796

FeII 2586


Norm

aliz

edF

lux





10 Perrotta et al



42 Electron Density


















minus3000

minus2500

minus2000

minus1500

minus1000

minus500

0

Bro

ad[O

II]v

98[k

msminus

1]



J0106

J0826

J0901

J0905

J0944

J1107

J1125

J1341

J1506

J1613

J1622

J1713

J2116

J2118


J0106J0826

J0944J1107

J1341J1506

J1613J1622

J2116J2118

100

101

102

103

104

Ele

ctro

nD

ensi

ty[c

mminus

3]







12 Perrotta et al

minus1

0

1

Total Total


log([NII]λ6585Hα)

minus1

0

1

log([

OII

I]λ

5007H

β)

Narrow

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118


log([SII]Hα)

Narrow



minus1

0

1

log([

OII

I]λ

5007H

β)

Broad

Narrow

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2118



















14 Perrotta et al







5 DISCUSSION




minus10

minus05

00

05

10

15lo

g([

NII

] 65

85[

SII

] 67

176

73

1)

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2116

J2118

7 8 9 10 11 12

log(MlowastM⊙)

minus10

minus05

00

05

10

15

log

([O

III]

50

07[

OII

] 37

263

72

9)

Wang et al 2019

Bassett et al 2019

Fletcher et al 2019

Izotov et al 201618


0 1 2


minus2 0 2 4






16 Perrotta et al
















95 100 105 110 115log(MlowastM⊙)

10minus1

100

101

Fb

roadF

nar

row

minus2 0 2 4


Genzel et al 2014

Freeman et al 2019

Swinbank et al 2019


Newman et al 2012

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2116

J2118










18 Perrotta et al




















20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77


















8 Perrotta et al







4 RESULTS


41 Kinematics






-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0106

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0826

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0901

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0905

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0944

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1107

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1125

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1341

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1506

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1613

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1622

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1713

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2116

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2118

Narrow Hα

Broad Hα or Hβ

Broad [OIII] 5007

Broad [OII] 3729

MgII 2796

FeII 2586


Norm

aliz

edF

lux





10 Perrotta et al



42 Electron Density


















minus3000

minus2500

minus2000

minus1500

minus1000

minus500

0

Bro

ad[O

II]v

98[k

msminus

1]



J0106

J0826

J0901

J0905

J0944

J1107

J1125

J1341

J1506

J1613

J1622

J1713

J2116

J2118


J0106J0826

J0944J1107

J1341J1506

J1613J1622

J2116J2118

100

101

102

103

104

Ele

ctro

nD

ensi

ty[c

mminus

3]







12 Perrotta et al

minus1

0

1

Total Total


log([NII]λ6585Hα)

minus1

0

1

log([

OII

I]λ

5007H

β)

Narrow

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118


log([SII]Hα)

Narrow



minus1

0

1

log([

OII

I]λ

5007H

β)

Broad

Narrow

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2118



















14 Perrotta et al







5 DISCUSSION




minus10

minus05

00

05

10

15lo

g([

NII

] 65

85[

SII

] 67

176

73

1)

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2116

J2118

7 8 9 10 11 12

log(MlowastM⊙)

minus10

minus05

00

05

10

15

log

([O

III]

50

07[

OII

] 37

263

72

9)

Wang et al 2019

Bassett et al 2019

Fletcher et al 2019

Izotov et al 201618


0 1 2


minus2 0 2 4






16 Perrotta et al
















95 100 105 110 115log(MlowastM⊙)

10minus1

100

101

Fb

roadF

nar

row

minus2 0 2 4


Genzel et al 2014

Freeman et al 2019

Swinbank et al 2019


Newman et al 2012

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2116

J2118










18 Perrotta et al




















20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77



















-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0106

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0826

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0901

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0905

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J0944

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1107

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1125

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1341

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1506

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1613

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1622

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J1713

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2116

-3000 -2000 -1000 0 1000

minus10

minus05

00

05

10 J2118

Narrow Hα

Broad Hα or Hβ

Broad [OIII] 5007

Broad [OII] 3729

MgII 2796

FeII 2586


Norm

aliz

edF

lux





10 Perrotta et al



42 Electron Density


















minus3000

minus2500

minus2000

minus1500

minus1000

minus500

0

Bro

ad[O

II]v

98[k

msminus

1]



J0106

J0826

J0901

J0905

J0944

J1107

J1125

J1341

J1506

J1613

J1622

J1713

J2116

J2118


J0106J0826

J0944J1107

J1341J1506

J1613J1622

J2116J2118

100

101

102

103

104

Ele

ctro

nD

ensi

ty[c

mminus

3]







12 Perrotta et al

minus1

0

1

Total Total


log([NII]λ6585Hα)

minus1

0

1

log([

OII

I]λ

5007H

β)

Narrow

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118


log([SII]Hα)

Narrow



minus1

0

1

log([

OII

I]λ

5007H

β)

Broad

Narrow

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2118



















14 Perrotta et al







5 DISCUSSION




minus10

minus05

00

05

10

15lo

g([

NII

] 65

85[

SII

] 67

176

73

1)

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2116

J2118

7 8 9 10 11 12

log(MlowastM⊙)

minus10

minus05

00

05

10

15

log

([O

III]

50

07[

OII

] 37

263

72

9)

Wang et al 2019

Bassett et al 2019

Fletcher et al 2019

Izotov et al 201618


0 1 2


minus2 0 2 4






16 Perrotta et al
















95 100 105 110 115log(MlowastM⊙)

10minus1

100

101

Fb

roadF

nar

row

minus2 0 2 4


Genzel et al 2014

Freeman et al 2019

Swinbank et al 2019


Newman et al 2012

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2116

J2118










18 Perrotta et al




















20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77


















10 Perrotta et al



42 Electron Density


















minus3000

minus2500

minus2000

minus1500

minus1000

minus500

0

Bro

ad[O

II]v

98[k

msminus

1]



J0106

J0826

J0901

J0905

J0944

J1107

J1125

J1341

J1506

J1613

J1622

J1713

J2116

J2118


J0106J0826

J0944J1107

J1341J1506

J1613J1622

J2116J2118

100

101

102

103

104

Ele

ctro

nD

ensi

ty[c

mminus

3]







12 Perrotta et al

minus1

0

1

Total Total


log([NII]λ6585Hα)

minus1

0

1

log([

OII

I]λ

5007H

β)

Narrow

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118


log([SII]Hα)

Narrow



minus1

0

1

log([

OII

I]λ

5007H

β)

Broad

Narrow

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2118



















14 Perrotta et al







5 DISCUSSION




minus10

minus05

00

05

10

15lo

g([

NII

] 65

85[

SII

] 67

176

73

1)

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2116

J2118

7 8 9 10 11 12

log(MlowastM⊙)

minus10

minus05

00

05

10

15

log

([O

III]

50

07[

OII

] 37

263

72

9)

Wang et al 2019

Bassett et al 2019

Fletcher et al 2019

Izotov et al 201618


0 1 2


minus2 0 2 4






16 Perrotta et al
















95 100 105 110 115log(MlowastM⊙)

10minus1

100

101

Fb

roadF

nar

row

minus2 0 2 4


Genzel et al 2014

Freeman et al 2019

Swinbank et al 2019


Newman et al 2012

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2116

J2118










18 Perrotta et al




















20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77





















minus3000

minus2500

minus2000

minus1500

minus1000

minus500

0

Bro

ad[O

II]v

98[k

msminus

1]



J0106

J0826

J0901

J0905

J0944

J1107

J1125

J1341

J1506

J1613

J1622

J1713

J2116

J2118


J0106J0826

J0944J1107

J1341J1506

J1613J1622

J2116J2118

100

101

102

103

104

Ele

ctro

nD

ensi

ty[c

mminus

3]







12 Perrotta et al

minus1

0

1

Total Total


log([NII]λ6585Hα)

minus1

0

1

log([

OII

I]λ

5007H

β)

Narrow

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118


log([SII]Hα)

Narrow



minus1

0

1

log([

OII

I]λ

5007H

β)

Broad

Narrow

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2118



















14 Perrotta et al







5 DISCUSSION




minus10

minus05

00

05

10

15lo

g([

NII

] 65

85[

SII

] 67

176

73

1)

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2116

J2118

7 8 9 10 11 12

log(MlowastM⊙)

minus10

minus05

00

05

10

15

log

([O

III]

50

07[

OII

] 37

263

72

9)

Wang et al 2019

Bassett et al 2019

Fletcher et al 2019

Izotov et al 201618


0 1 2


minus2 0 2 4






16 Perrotta et al
















95 100 105 110 115log(MlowastM⊙)

10minus1

100

101

Fb

roadF

nar

row

minus2 0 2 4


Genzel et al 2014

Freeman et al 2019

Swinbank et al 2019


Newman et al 2012

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2116

J2118










18 Perrotta et al




















20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77


















12 Perrotta et al

minus1

0

1

Total Total


log([NII]λ6585Hα)

minus1

0

1

log([

OII

I]λ

5007H

β)

Narrow

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118


log([SII]Hα)

Narrow



minus1

0

1

log([

OII

I]λ

5007H

β)

Broad

Narrow

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2118



















14 Perrotta et al







5 DISCUSSION




minus10

minus05

00

05

10

15lo

g([

NII

] 65

85[

SII

] 67

176

73

1)

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2116

J2118

7 8 9 10 11 12

log(MlowastM⊙)

minus10

minus05

00

05

10

15

log

([O

III]

50

07[

OII

] 37

263

72

9)

Wang et al 2019

Bassett et al 2019

Fletcher et al 2019

Izotov et al 201618


0 1 2


minus2 0 2 4






16 Perrotta et al
















95 100 105 110 115log(MlowastM⊙)

10minus1

100

101

Fb

roadF

nar

row

minus2 0 2 4


Genzel et al 2014

Freeman et al 2019

Swinbank et al 2019


Newman et al 2012

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2116

J2118










18 Perrotta et al




















20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77

































14 Perrotta et al







5 DISCUSSION




minus10

minus05

00

05

10

15lo

g([

NII

] 65

85[

SII

] 67

176

73

1)

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2116

J2118

7 8 9 10 11 12

log(MlowastM⊙)

minus10

minus05

00

05

10

15

log

([O

III]

50

07[

OII

] 37

263

72

9)

Wang et al 2019

Bassett et al 2019

Fletcher et al 2019

Izotov et al 201618


0 1 2


minus2 0 2 4






16 Perrotta et al
















95 100 105 110 115log(MlowastM⊙)

10minus1

100

101

Fb

roadF

nar

row

minus2 0 2 4


Genzel et al 2014

Freeman et al 2019

Swinbank et al 2019


Newman et al 2012

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2116

J2118










18 Perrotta et al




















20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77



















minus10

minus05

00

05

10

15lo

g([

NII

] 65

85[

SII

] 67

176

73

1)

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2116

J2118

7 8 9 10 11 12

log(MlowastM⊙)

minus10

minus05

00

05

10

15

log

([O

III]

50

07[

OII

] 37

263

72

9)

Wang et al 2019

Bassett et al 2019

Fletcher et al 2019

Izotov et al 201618


0 1 2


minus2 0 2 4






16 Perrotta et al
















95 100 105 110 115log(MlowastM⊙)

10minus1

100

101

Fb

roadF

nar

row

minus2 0 2 4


Genzel et al 2014

Freeman et al 2019

Swinbank et al 2019


Newman et al 2012

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2116

J2118










18 Perrotta et al




















20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77


















16 Perrotta et al
















95 100 105 110 115log(MlowastM⊙)

10minus1

100

101

Fb

roadF

nar

row

minus2 0 2 4


Genzel et al 2014

Freeman et al 2019

Swinbank et al 2019


Newman et al 2012

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2116

J2118










18 Perrotta et al




















20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77



















95 100 105 110 115log(MlowastM⊙)

10minus1

100

101

Fb

roadF

nar

row

minus2 0 2 4


Genzel et al 2014

Freeman et al 2019

Swinbank et al 2019


Newman et al 2012

J0826

J0944

J1107

J1341

J1506

J1613

J1622

J2116

J2118










18 Perrotta et al




















20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77


















18 Perrotta et al




















20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77



























20 Perrotta et al


log(sSFRyrminus1)

minus10

minus05

00

05

10

15lo

g([

OII

I]5007[

OII

] 37263

729)

Wang+2019

Bassett et al 2019

Fletcher et al 2019

Izotov+201618

Alexandroff+2015

J0106

J0826

J0901

J0905

J0944

J1107

J1341

J1506

J1613

J1622

J1713

J2118





















ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77
























ACKNOWLEDGEMENTS



REFERENCES

























AampA 468 61

22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77


















22 Perrotta et al




































ARAampA 57 511





































1136


























869 123









ApJ 663 L77











































869 123









ApJ 663 L77


















arXiv:2106.02366v2 [astro-ph.GA] 18 Oct 2021

Documents

Transcript of arXiv:2106.02366v2 [astro-ph.GA] 18 Oct 2021