A new technique for finding galaxies leaking Lyman-continuum radiation: [SII]-deficiency
Bingjie Wang, Timothy M. Heckman, Claus Leitherer, Rachel Alexandroff, Sanchayeeta Borthakur, Roderik A. Overzier
DDraft version September 5, 2019
Typeset using L A TEX twocolumn style in AASTeX62
A New Technique for Finding Galaxies Leaking Lyman-Continuum Radiation: [SII]-Deficiency
Bingjie Wang ( 王 冰 洁 ), Timothy M. Heckman, Claus Leitherer, Rachel Alexandroff, Sanchayeeta Borthakur, and Roderik A. Overzier
5, 6 Department of Physics & Astronomy, Johns Hopkins University, Baltimore, MD 21218, USA Space Telescope Science Institute, Baltimore, MD 21218, USA Dunlap Institute for Astronomy & Astrophysics, University of Toronto, Toronto, ON M5S 3H4, Canada School of Earth & Space Exploration, Arizona State University, Tempe, AZ 85287, USA Observatorio Nacional, Rio de Janeiro, Brazil Institute of Astronomy, Geophysics and Atmospheric Sciences, Department of Astronomy, University of S˜ao Paulo, S˜ao Paulo, SP05508-090, Brazil
ABSTRACTThe source responsible for the reionization of the Universe is believed to be the population of star-forming galaxies at z ∼ z ∼ .
3. We show thatthese galaxies differ markedly in their properties from the class of leaky “Green-Pea” galaxies at similarredshifts: our sample galaxies are more massive, more metal-rich, and less extreme in terms of theirstellar population and the ionization state of the interstellar medium. Like the Green Peas, they haveexceptionally high star-formation rates per unit area. They also share some properties with the knownleaky galaxies at z ∼
3, but are significantly dustier. Our results validate a new way to identify locallaboratories for exploring the processes that made it possible for galaxies to reionize the Universe.
Keywords: extragalactic astronomy – galaxy formation – star formation – interstellar medium – inter-galactic medium INTRODUCTIONThe Epoch of Reionization (EoR) is the period dur-ing which the first stars are formed and emit light thationizes the intergalactic medium (IGM). The historyof reionization is primarily inferred from two measure-ments: large-scale anisotropies in polarization of the cos-mic microwave background (CMB) and spectroscopy ofdistant quasars. The CMB is affected by the total col-umn density of free electrons along line of sight. The pa-rameterization of its Thomson scattering optical depth τ remains to be the least constrained parameter in the Corresponding author: Bingjie [email protected]
ΛCDM model (e.g. Bennett et al. (2013); Planck Collab-oration et al. (2018a)). Observations of quasar absorp-tion lines via the Gunn-Peterson effect (Gunn & Peter-son 1965) sets the limit that reionization completes by z ∼ z ∼
12 to 6, and followed by quasars reioniz-ing helium. While deep imaging with the Hubble SpaceTelescope (HST) indicates that the ultraviolet (UV) lu-minosity density of early star-forming galaxies is highenough that they are the best candidates to provide theionizing photons necessary for reionizing the Universe(e.g. Bouwens et al. (2016)), the fraction of Lyman- a r X i v : . [ a s t r o - ph . GA ] S e p Wang et al. continuum (LyC) photons that actually escape from thegalaxies into the IGM, which is required to be significant( > α emission. In this paper, we present a new and indepen-dent signpost of leakiness that could also be measuredby future observations of galaxies during the EoR bythe James Webb Space Telescope (JWST).The new signpost is the relative weakness of the[SII]6717,6731 emission lines, defined with respect totypical star-forming galaxies. This [SII]-deficiency isa tracer of gas that is optically thin to ionizing radi-ation, allowing the escape of LyC photons. Given thatthe ionization potential for producing SII is only 10.4eV, which is significantly less than a Rydberg, much ofthe [SII] emission therefore arises in the warm partially-ionized region near and just beyond the outer edge ofthe Stromgren sphere in a classical HII region. In anHII region that is optically thin to ionizing radiation,this partially-ionized SII zone is weak or even absent,and the relative intensity of the [SII] emission lines dropsignificantly as a result (Pellegrini et al. 2012).In this paper, we validate this idea using HST far-UV observations with the Cosmic Origins Spectrograph(COS; Green et al. (2012)) of a sample of three star-forming galaxies. The structure of this paper is as fol-lows. In Section 2, we begin by detailing our definitionof the [SII]-deficiency. In Section 3, we summarize theobservational data sets, including sample selection, dataprocessing and analysis, and measured ancillary param-eters. In Section 4, we present our results, namely theescape fractions for the LyC. In Section 5, we make com-parisons of our galaxies to other known leaky galaxiesat both low and high-redshift selected in other ways, log ([SII] 6717, 6731/H ) l o g ([ O III ] / H ) D e n s i t y o f s t a r - f o r m i n g g a l a x i e s Leaky (This paper)Leaky (GP)Leaky (LBA) Non-leaky (This paper)Non-leaky (LBA)[SII] = 0
Figure 1.
This figure is used in defining [SII]-deficiency,where the flux ratio of [SII]6717,6731/H α is plotted againstthat of [OIII]5007/H β . The contours show the density distri-bution of the SDSS DR 12 star-forming galaxy sample. Theblack dotted line is fitted to the locus of the peak density ofthis distribution. The [SII]-deficiency is defined as a galaxy’sdisplacement in log([SII]/H α ) from this ridge-line. Uncer-tainties in the ridge-line are negligible except in the upperleft, where they are indicated in grey. The red triangles rep-resent the two leaky star-forming galaxies of this paper, whilethe black dot represent the non-leaky one. Also plotted areleaky Green Pea galaxies in Izotov et al. (2016a,b, 2018a,b)(pink triangles) and Lyman Break Analogs in Alexandroffet al. (2015) (orange triangles and blue dots), both of whichare discussed in Section 5. and assess the various indirect indicators of leakiness.Finally, we summarize our conclusions in Section 6.Throughout we adopt the best-fit cosmologicalparameters from the Planck 2018 analysis (theirTT,TE,EE+lowE+lensing+BAO case): H = 67 . − Mpc − , Ω M = 0 . Λ = 0 .
690 (PlanckCollaboration et al. 2018b). DEFINITION OF [SII]-DEFICIENCYThe [SII]-deficiency is established with respect to thesample of SDSS DR 12 star-forming galaxies in the planeof [SII]6717,6731/H α v s. [OIII]5007/H β , as shown inFigure 1. Here we describe the procedure as follows.First we select all the galaxies classified as “starforming” in the value added catalog provided by thePortsmouth group (Thomas et al. 2013), with a signal-to-noise cut of five in the flux measurements. We thenbin the data in log [OIII] / H β and make a histogram inlog [SII] / H α for each bin, which is subsequently fittedwith a Gaussian (or a skewed Gaussian in a few cases)to determine the peak location. Lastly we perform a SII]-Deficiency Table 1.
Observation logs.Name Galaxy z COS FUV grating Exposure time COS NUV ACQ image Date of HST(s) exposure time (s) observationJ2226 SDSSJ222634.07-090106.2 0.299 G140L 7681.728 241 2018-05-25J1119 SDSSJ111905.27+592514.1 0.290 G140L 5502.720 120 2018-09-26J0910 SDSSJ091021.35+610550.2 0.272 G140L 8336.640 161 2018-09-21J1432 SDSSJ143256.4+274249.6 0.266 G140L 5100.704 97 2018-06-25J1242 SDSSJ124206.24+011537.5 0.271 G140L 7832.864 161 2018-08-10 polynomial fit to the peaks. This is shown as the blackdotted curve in Figure 1. The resulting fitting formulais: y = − .
487 + 0 . ξ + 0 . ξ − . ξ − . ξ + 0 . ξ + 8 . ξ + 0 . ξ − . ξ (1)where ξ is the line ratio of log [OIII] / H β , and y is theline ratio of log [SII] / H α .We define the [SII]-deficiency as a galaxy’s displace-ment in log [SII] / H α from the ridge-line, denoted as∆[SII]. Uncertainties in the emission-line ratios for in-dividual galaxies are less than 0.1 dex. Uncertainties inthe location of the ridge-line are negligible except wherethe data are sparse. In these cases, we estimate un-certainties via bootstrap. These are shown in grey inFigure 1. DATA3.1.
Sample Selection
In HST program GO-15341 (PI T. Heckman) we ob-served a sample of five galaxies selected in the SDSSDR7 plus GALEX GR6 catalogs based on the followingcriteria:1. A [SII]-deficiency relative to normal star-forminggalaxies of at least 0.2 dex as shown in Figure 1.In this paper the value of ∆[SII] for J1242 is justbelow 0.2 dex. This is because, since the originalsample definition, we updated the sample of nor-mal galaxies to SDSS DR 12, which results in aslight change in the ridge-line.2. A seeing-de-convolved half-light radius of less than0 . (cid:48)(cid:48) (typically smaller than 1 kpc) based on SDSSu-band images. This mimics the small sizes ofgalaxies in the EoR.3. An estimated far-UV flux inside the COS apertureof larger than 2 × − erg cm − s − ˚A − This was derived by using SDSS u-band images to make anaperture correction to the GALEX far-UV flux.4. Redshifts higher than 0.26. This ensures that theLyman edge falls at wavelengths over which COShas high sensitivity ( > α emission-line), even thoughthe SDSS optical spectrum is dominated by a starburst.We do not discuss these targets further in this paper.For the three remaining targets, we will demonstratethat they are indeed dominated by starlight in the far-UV by using the fit of Starbust99 (hereafter SB99, Lei-therer et al. (1999)) model spectra in Section 3.3.3.2. Data Processing
All the COS far-UV spectra were obtained using theG140L grating in the 1105 setting. This covers the ob-served wavelength range from 1110 to 2150 ˚A, corre-sponding to roughly 880 to 1690 ˚A in the rest frame.The spectral resolution is about 0.5 ˚A.We first retrieve our COS data from the MAST archivewhich had been processed through the standard COSpipeline
CalCOS . The most technically challenging partof the data analysis is trying to accurately subtract thedark counts, which contribute significantly to the netcounts in the region of the LyC. Therefore, followingthe procedure in the appendix of Leitherer et al. (2016),we create a super-dark image to replace the standardCOS pipeline version. A super-dark image for a givengalaxy is obtained by selecting all the COS dark framestaken within ± ± Wang et al. dark count rate. We therefore turn off the native back-ground correction in
CalCOS , and modify the procedureto subtract the super-dark from the science exposurejust before extraction of the spectrum.By examining the individual dark frames that wereused to create a given super-dark, we estimate that thetemporal variations in the dark count rate leads to anuncertainty in the dark count rate at the time of theobservations of ± α airglow lines, or other weak emis-sion. To do so, we compare an average of five G140L ex-posures of blank fields provided by the COS team withour spectra. This comparison is shown in Figure 2, andestablishes that there is no significant sky contaminationbelow the Lyman limit.3.3. Data Analysis
Given the relatively low signal-to-noise ratio in theextracted spectra, we smooth all the spectra used witha Gaussian kernel before further analysis. The full widthat half maximum of the kernel is chosen to be about 0.5˚A to reach the native resolution.For each spectrum, we first correct for Milky Way(MW) extinction in the observed frame using the red-dening law proposed in Mathis (1990), and E(B − V) MW taken from the NASA Extragalactic Database for agiven position on the sky. We then transform the ob-served spectra to the rest frame of the galaxy using SDSSspectroscopic redshifts, conserving the quantity λF λ .Synthetic spectra are generated based on stellar evo-lutionary synthesis models using SB99. We produce ourmodels based on a star formation history of a continuousand constant rate of star formation. The stellar popula-tion is parameterized by a Kroupa initial mass function(IMF) (Kroupa 2001). The stellar population evolvesfrom the zero-age main sequence using the evolutionarymodels of the Geneva Group. The model spectra aredescribed in detail in Leitherer et al. (2010). In all, wegenerate eight sets of SB99 models based on two choiceseach for burst age (10 and 10 years), metallicity (solaror 1/7 solar), and whether or not models using stellarrotation are employed.A model spectrum is interpolated into the same wave-length array as its corresponding COS spectrum, andalso convolved with the same Gaussian kernel, ensuringthat they have the same resolution. A best-fit is chosenby eye; more specifically, we closely examine the matchbetween the synthetic and observed spectra of the two strong stellar wind features due to OVI 1032,1038 andNV 1238,1242. These P-Cygni features trace the mostmassive stars, which are the ones responsible for pro-ducing most of the ionizing continuum. For OVI wecould only examine the redshifted emission component,as the blueshifted absorption is contaminated by the [OI]airglow line. From these comparisons, we find that thebest-fits for J0910 and J1432 come from the solar metal-licity models that are of 10 year ages, and that incor-porate stellar rotation, while J1242 is better fitted witha 10 -year model. The overall best fits are shown inFigure 3, and a zoom-in on these wind lines is shownin Figure 4. As seen from the figures, each stellar spec-trum alone is a good fit to the data, and hence we inferthat the far-UV light in all three targets is in fact domi-nated by hot massive stars. The only stellar feature themodel does not fit well is the blend of the CIII 1176 andthe CIV/NIV 1169 lines. We are exploring this and willdescribe the results in a future paper dealing with thestellar populations in these galaxies.Having chosen a model, we then vary the internal (ex-tragalactic) extinction, E(B − V) int , as a free parameteruntil the slope of a given observed spectrum matches itsSB99 model. To do so, we use the extragalactic red-dening law derived in Calzetti et al. (1999). There is analternative proposed by Reddy et al. (2015, 2016), whichdeviates from the former at short wavelengths ( λ < Measured Ancillary Parameters
In this section we list important ancillary parameters,and describe how they are determined. The values areall listed in Table 2.We measure the star formation rates (SFRs) in threeways. In all cases we use the same IMF as that used inour SB99 fit (see above). SFR UV is inferred from COSUV data by taking a ratio between a dereddened galaxyflux spectrum and a SB99 spectrum generated assuminga SFR of 1 M (cid:12) yr − . SFR IR is calculated by using theWISE IR data at 12 and 22 µ m (Wright et al. 2010) toestimate the rest-frame 24 µ m luminosity, and then us-ing the relation given in Kennicutt & Evans (2012). Thishas the advantage of being independent of any uncertaincorrection to the UV fluxes. SFR H α is calculated fromextinction-corrected fluxes. The MPA-JHU catalog pro-vides the fluxes of H α and H β . We calculate E ( β − α ),defined as E ( β − α ) = A (H α ) − A (H β ) with A being theextinction in magnitude, as: E ( β − α ) = 2 . [ F (H α ) obs /F (H β ) obs ] − . [ F (H α ) /F (H β )] (2a) SII]-Deficiency Å ]101234 F l u x [ e r g s c m Å ]
1e 16 J0910 Mean blank field889 897 905 913Rest-frame wavelength [ Å ] (a) Å ]101234 F l u x [ e r g s c m Å ]
1e 16 J1432889 897 905 913Rest-frame wavelength [ Å ] (b) Å ]64202468 F l u x [ e r g s c m Å ]
1e 17 J1242889 897 905 913Rest-frame wavelength [ Å ] (c) Figure 2.
Observed spectra plotted in the region below the Lyman limit after super-dark subtraction. The orange lines arethe COS spectra of our three galaxies, while the blue line is the average of five G140L exposures of blank fields. Note that theblank sky spectrum shows no contaminating signal.
900 1000 1100 1200 1300 1400 1500Rest-frame wavelength [ Å ]10 F l u x [ e r g s c m Å ] J0910 SB99 (a)
900 1000 1100 1200 1300 1400 1500Rest-frame wavelength [ Å ]10 F l u x [ e r g s c m Å ] J1432 (b)
900 1000 1100 1200 1300 1400 1500Rest-frame wavelength [ Å ]10 F l u x [ e r g s c m Å ] J1242 (c)
Figure 3.
Spectra of the three star-forming galaxies with Milky Way extinction and internal extinction removed (in blue), andover-plotted with SB99 best fits (in coral). The extinction values are: (a). E(B − V) MW = 0 . − V) int = 0 . − V) MW = 0 . − V) int = 0 . − V) MW = 0 . − V) int = 0 . Å ]10 F l u x [ e r g s c m Å ] N V O V I J0910 SB99 (a) Å ]10 F l u x [ e r g s c m Å ] N V O V I J1432 (b) Å ]10 F l u x [ e r g s c m Å ] N V O V I J1242 (c)
Figure 4.
Same as Figure 3, but zooming in on the OVI and NV stellar wind lines, which are used for deciding the best-fitSB99 model spectra. The strongest residuals (data minus model) are due to the OI telluric airglow emission, Ly α emission, andinterstellar absorption-lines. Wang et al. (a) (b) (c)
Figure 5.
COS near-UV acquisition images of the three [SII]-weak star-forming galaxies: (a) J0910; (b) J1432; and (c) J1242.Also over-plotted in turquoise are the ellipses which enclose half of the total near-UV emitted light. The images are 1.032 (cid:48)(cid:48) by1.032 (cid:48)(cid:48) , and the color bars indicate counts per second.
Table 2.
Measured ancillary parameters.SFR UV SFR H α SFR IR A(H α ) r SFR IR /A M (cid:63) (M (cid:12) yr − ) (M (cid:12) yr − ) (M (cid:12) yr − ) (kpc) (M (cid:12) yr − kpc − ) (log M (cid:12) )J0910 128 35 125 ±
11 1.24 0.22 394 10.44J1432 209 19 134 ±
10 0.75 0.17 705 10.54J1242 100 21 55 ±
10 0.96 0.50 34 10.38EW H α EW Ly α R Ly α ∆[SII] [OIII]/[OII] 12 + log (O / H)(˚A) (˚A) (dex)J0910 138 21.84 0.75 -0.30 1.29 8.66J1432 113 24.55 0.44 -0.28 1.57 8.60J1242 125 N/A N/A -0.17 1.42 8.52
SII]-Deficiency K, which translates to anintrinsic ratio of F (H α ) /F (H β ) = 2 .
86, the extinctionin magnitude for H α is then A (H α ) = 2 . E ( β − α ) (2b)And so finally we have the extinction-corrected H α fluxas: F (H α ) corr = 10 . A (H α ) F (H α ) obs (2c)Following Table 1.1 in Calzetti (2011), we estimateSFR(H α ) in units of M (cid:12) yr − viaSFR(H α ) = 5 . × − L (H α ) (3)where L is the luminosity in erg s − .After examining the COS near-UV acquisition imagesas shown in Figure 5, we find that all targets are welllocated inside the SDSS and COS apertures, which aretaken to be 1 . (cid:48)(cid:48) and 1 . (cid:48)(cid:48) respectively. We therefore donot apply any aperture corrections to SFRs. We alsonote that the fluxes in the images are consistent withthe GALEX near-UV flux.Additionally we use the COS near-UV images to com-pute the half-light radius for a given galaxy by findingan ellipse that encloses half of the total near-UV emit-ted light; the listed value for r is therefore ( a b ) / in kpc. During the process, the background is esti-mated from the mean of an annulus of r in = 0 . (cid:48)(cid:48) andr out = 1 . (cid:48)(cid:48) . A small correction for the effect of the PSFis also applied.The values for the rest-frame equivalent width of theH α emission line are taken from the MPA-JHU catalog,and the stellar masses are taken from the median of thecorresponding PDF in the same catalog.The oxygen abundance of the interstellar medium(ISM) in each galaxy is estimated following Pettini &Pagel (2004):12 + log (O / H) = 8 . − . × O3N2 (4a)where O3N2 = log [O III ] λ / H β [N II ] λ / H α (4b)This relation is valid for − < O3N2 < .
9. Since thewavelength of H α is close to [NII] and H β is close to[OIII], this method is insensitive to dust extinction.Then we use the conversion: 12 + log (O / H) = 8 . α line, we use the followingprocedure. Each observed galaxy spectrum is first nor-malized by fitting a second-order polynomial functionto the continuum and the spectrum is divided by this function. We do the same for the corresponding best-fitSB99 spectrum, which is then subtracted from the nor-malized galaxy spectrum to remove the stellar spectralcomponent. Lastly we add a value of 1 to this differencespectrum to produce a normalized spectrum with stellarfeatures removed.To measure the Ly α equivalent widths, we fit a (multi-component-) Gaussian. We estimate that the resultingequivalent widths have errors on the order of 10%–15%dominated by systematics in the polynomial fit to thecontinuum emission and the subtraction of SB99 models.Next, we use the starlight-subtracted spectra to quan-tify the different Ly α profile shapes using the parameterR Ly α , which is defined to be the ratio of the equivalentwidth of the blueshifted portion of the profiles to that ofthe redshifted portion. We define the equivalent widthfor emission to be positive, and for absorption to be neg-ative. Therefore, a negative R Ly α indicates blueshiftedabsorption and redshifted emission (i.e. a traditional P-Cygni profile) while a positive value for R Ly α indicatessignificant blueshifted emission. RESULTSAs can be seen in Figure 2, we detect a significant fluxbelow the Lyman edge in J0910 and J1432, and measureonly an upper limit in J1242. To characterize this emis-sion, we take the mean of flux densities, uncorrected forextinction, in a spectral window from ∼
885 to 910 ˚A.Resulting values are listed in Table 3 as F . The ex-act spectral windows for each of the galaxies are alsolisted in Table 3 under the column LyC range. Theseparticular choices are motivated by avoiding the detec-tor edge where dark count rates increase significantly.The errors quoted account for both the statistical (Pois-son) errors, which are extracted from the corresponding x1d files, and the systematic errors associated with darksubtraction.In the following paragraphs, we consider three ways tomeasure the escape fraction, each with advantages anddisadvantages. Relevant measurements of flux densitiesare all listed in Table 3. The first and also the simplestway is to measure the ratio of the observed fluxes inthe LyC to those at a rest-wavelength of 1500 ˚A. Thismeasurement is made after correcting the fluxes for MWextinction only. The advantage of this parameter is thatit is most directly connected to actual observational esti-mation of the rate of escaping ionizing radiation duringthe EoR. That is, the observed luminosity density due tostar-forming galaxies at a rest-frame 1500 ˚A can be mea-sured from the far-UV luminosity functions during EoR.Knowing the mean ratio of LyC to 1500 ˚A fluxes for arepresentative ensemble of star-forming galaxies (from Wang et al.
Table 3.
Measurements of observed flux densities used in quantifying the escape fractions. The LyC ranges arewavelength ranges over which an average is taken in calculating F and F − . The first uncertainties in F areestimated from Poisson statistics, and the second ones are from dark fluctuations.LyC range F F F F b (cid:16) F − F (cid:17) obsc (cid:16) F − F (cid:17) obsd (˚A) ( × − erg cm − s − ˚A − ) ( × − erg cm − s − ˚A − )J0910 885 – 910 1 . ± . ± .
05 2.16 0.38 0.538 0.482J1432 888 – 910 2 . ± . ± .
08 2.99 0.32 0.460 0.406J1242 885 – 910 0 . ± . ± .
08 1.04 0.02 0.046 0.039 a Uncorrected for extinction. b Corrected for MW extinction only. c Corrected for MW and internal extinctions, assuming extragalactic reddenning law in Calzetti et al. (1999). d Same as c, but assuming extragalactic reddenning law in Reddy et al. (2015, 2016).
Table 4.
Relative and absolute escape fractions. The measurements quoted for J1242 are upper limits inferred from a3 σ limit on dark fluxes. The first uncertainties are estimated from Poisson statistics, and the second ones are from darkfluctuations.E(B − V) inta f esc , rel f esc , abs E(B − V) intb f esc , rel f esc , abs f esc , absc ( × − ) ( × − ) ( × − ) ( × − ) ( × − )J0910 0.239 93 . +10 . . − . − . . ± . ± . . +9 . . − . − . . ± . ± . . ± . ± . . +5 . . − . − . . ± . ± . . +6 . . − . − . . ± . ± . . ± . ± . < < < < < a Assuming reddenning law in Calzetti et al. (1999). b Assuming reddenning law in Reddy et al. (2015, 2016). c Obtained by taking the ratio between MW extinction-corrected ( F ) obs and ( F ) int inferred from SB99 given SFR IR . observations of lower-z analogs) yields an estimate of theLyC luminosity density produced by the EoR galaxies.This quantity, F /F , for the three [SII]-weak star-forming galaxies are listed in Table 3. For F we fit asimple low-order polynomial to the continuum between1100 and 1500 ˚A rest-frame and use the resulting valueat 1500 ˚A since the data are noisy at this wavelength.Next, we calculate what is sometimes referred to asthe relative escape faction, f esc , rel . This is essentiallythe ratio of the observed flux decrement across the Ly-man break (after correction for MW and internal ex-tinctions) compared to the intrinsic decrement in thebest-fit SB99 model spectrum. As such, the value of therelative escape fraction is independent of the effects ofdust extinction, and is probing only radiative transfereffects associated with the photo-electric absorption ofthe LyC due to hydrogen. In our specific case we define f esc , rel as: f esc , rel = (cid:18) F − F + (cid:19) obs (cid:18) F + F − (cid:19) int (5)where F − is the average extinction-corrected flux den-sities taken between rest-frame ∼
885 and 910 ˚A (again,the exact spectral windows are listed in Table 3), and F + is the average taken between 1050 and 1150 ˚A.The latter choice is made to avoid the effects of the Ly α airglow line and the confluence of the high-order Lymanseries lines near the Lyman edge.Finally, we note that dust can be a significant sourceof opacity for both ionizing and non-ionizing far-UV ra-diation in galaxies. We therefore measure what is com-monly referred to as the absolute escape fraction, f esc , abs (the ratio of emergent LyC flux to the intrinsic flux, in-cluding the effect of dust extinction). Conventionally SII]-Deficiency f esc , abs = f esc , rel − . A (6a)where A = κ (910˚A)E(B − V) int (6b)is the absorption at 910 ˚A. We obtain κ (910˚A) by ex-trapolating the fitting formulae provided in Calzettiet al. (1999); Reddy et al. (2015, 2016) slightly towardsshort wavelength, since the original formulae end at 1200and 915 ˚A, respectively.A major source of systematic uncertainty in Equa-tion 6 is in the UV extinction. To assess this we comparethe values for the escape fractions based on the extinc-tion laws adopted by Calzetti et al. (1999) and Reddyet al. (2015, 2016) (see Table 4). There we see thatthe effects are modest but noticeable; hence we adopta second approach to circumvent this uncertainty. Weuse SFR IR to predict the LyC flux in the best-fit SB99model, and then compare this to the observed LyC fluxcorrected only for the MW extinction. This quantity islisted in the last column in Table 4.In addition, there are systematic uncertainties in es-cape fraction associated with the intrinsic Lyman breakin the SB99 models. Therefore we compare the val-ues for both solar and 1/7 solar metallicity models, forburst ages of 10 and 10 years, and for models with andwithout stellar rotation employed (i.e. Geneva v40 and
Geneva v00 , respectively). For completeness, we list theLyman-break amplitudes defined as the ratio betweenthe average flux density over 1050-1150 ˚A and that over900-910 ˚A for different SB99 models in Table 5. Thelargest variation is with burst duration. The values wequote for the relative and absolute escape fractions forJ0910 and J1432 are obtained from SB99 models witha constant SFR for 10 years, while for J1242, they arefrom SB99 models with a constant SFR for 10 years.Those spectra better fit the OVI and NV wind lines.Taking an older burst age for the former two would in-crease the escape fractions by ∼ DISCUSSIONIn this section we will place the leaky [SII]-weak galax-ies in context. First, we will compare their properties tothose of the leaky Green Pea galaxies, which comprise alarge majority of the confirmed low-z leaky galaxies. Wewill then compare the properties of all the known low-z leaky galaxies to non-leaky low-z starburst galaxies.This will allow us to assess the robustness of the variousproposed indirect signposts of leaky galaxies. Finally,we will compare the properties of the [SII]-weak leakygalaxies to leaky galaxies at z ∼ Comparisons of [SII]-Weak and Green Pea LeakyGalaxies
For the Green Pea galaxies, M (cid:63) , [OIII]/[OII], EW H α are taken from the respective references, while the re-maining quantities are calculated the same way as pre-sented in Section 3.4 for consistency. Specifically, forSFR H α we estimate the luminosity of H α to be used inEquation 3 as 2.86 L H β , where L H β is taken from the ref-erences; for the SFR UV , we retrieve their COS spectrafrom MAST, and deredden them using the reddeninglaw of Calzetti et al. (1999). Since the Green Peas arenearly dust-free, this calculation of SFR UV is subjectto less systematic uncertainty due to internal extinctioncorrection. These properties are listed in Table 7 in theappendix, and the corresponding histograms are shownin Figure 6.As seen in Figure 6, one major difference between the[SII]-weak and Green Pea samples is the stellar mass:the median masses are 10 . and 10 . M (cid:12) for the GreenPeas and [SII]-weak galaxies respectively. This large dif-ference in mass leads to a difference in gas-phase metal-licity: median values of 12+log (O / H) of 8.6 and 7.9 forthe [SII]-weak and Green Peas samples, where a valueof 8.7 corresponds to solar metallicity.The [SII]-weak galaxies have extraordinarily high val-ues of SFR/area (mean of 550 M (cid:12) yr − kpc − ), comparedto a median value of about 75 M (cid:12) yr − kpc − for theGreen Peas. In terms of SFR/M (cid:63) , the Green Peas aremore extreme (median value 10 − yr − , about an order-of-magnitude larger than the values for the [SII]-weakgalaxies. This is consistent with the significantly lowervalues of the H α equivalent widths in the [SII]-weakgalaxies compared to the Green Peas, and together thesetwo results suggest that the current bursts in the [SII]-weak galaxies are occurring in the presence of more sig-nificant prior star-formation on timescales longer than afew Myr compared to the Green Peas.Other emission-line properties of the [SII]-weak galax-ies are also much less extreme that those of the GreenPeas. As with H α , the Ly α equivalent widths of the[SII]-weak galaxies are smaller than those of the GreenPeas by a factor of ∼ vs.
75 ˚A). Moreover, as seenin Figure 7, the [SII]-weak galaxies do not exhibit theextraordinarily high ionization level that is characteris-tic of the Green Peas (with median [OIII]/[OII] fluxesratios of 1.4 vs.
Wang et al.
Table 5.
Lyman-break amplitudes, F (1050 − A ) /F (900 −
910 ˚ A ) , for different SB99 models.The values used in calculating f esc are indicated with asterisks.Name Z (cid:12) , no rotation Z (cid:12) , rotation Z / (cid:12) , no rotation Z / (cid:12) , rotationJ0910 (10 yr) 2.084 1 . ∗ yr) 3.268 2.925 2.682 2.830J1432 (10 yr) 2.083 1 . ∗ yr) 3.268 2.923 2.684 2.830J1242 (10 yr) 2.084 1.736 1.792 1.756J1242 (10 yr) 3.268 2 . ∗ Signposts of Leakiness
There are a number of galaxy characteristics that havebeen previously identified as potential signposts of LyC-leakage from galaxies. In this section we evaluate thesesignposts in light of our discovery of this new class ofleaky galaxy. To do so, we assemble a sample of knownleaky galaxies at low-z and compare their properties toa control sample of strong starbursts at similar redshiftsthat are unlikely to be leaky. For the sake of consistency,we include only galaxies with COS data and with the setof galaxy parameters that can be measured using thespectra in the SDSS.These samples are drawn from the union of the [SII]-weak galaxies presented in this paper, the leaky GreenPeas in Izotov et al. (2016a,b, 2018a,b), and the Ly-man Break Analogs in Alexandroff et al. (2015). In thelatter sample, J0921 has been directly detected belowthe Lyman edge (Borthakur et al. 2014). For the othersample members, we use the residual intensity in theLy β absorption-line as an indicator of leakiness, follow-ing the results in Chisholm et al. (2018), and see alsoSteidel et al. (2018). This adds J0213 and J0926 asleakers, with the 13 other galaxies in Table 8 being clas-sified as non-leaky. Alexandroff et al. (2015) list all therelevant quantities, except for [OIII]/[OII], which we cal-culate using fluxes obtained from the MPA-JHU catalog.We note that our definition for [SII]-deficit differs fromthat in Alexandroff et al. (2015) by taking the horizontaldisplacement from the parametric ridge-line as shownin Figure 1 instead of the perpendicular distance be-tween each galaxy and the ridge-line, so measurementsof ∆[SII] are also re-made according to our definition.We have already compared some of the proposed sign-posts in the [SII]-weak and Green Pea galaxies in theprevious section. In Figure 6, we see that the class ofleaky galaxies as-a-whole has somewhat larger values forSFR/area than the non-leaky starbursts (median valuesof 51 vs. (cid:12) yr − kpc − ). The leaky galaxies are moreextraordinary in this regard when compared to typical low-z star-forming galaxies, which have an SFR/Area ofonly ∼ − M (cid:12) yr − kpc − (Kennicutt & Evans 2012).We also see that the leaky galaxy sample has a signifi-cantly higher median value for the Ly α equivalent widththan the non-leaky galaxies (65 and 4 ˚A respectively).Another common property of the leaky galaxies is thatthey have a significant amount of blue-shifted Ly α emis-sion (with median value for the R Ly α parameter of 0.4for leaky sample vs. α profiles and their implication for the es-cape of LyC in Green Peas is discussed in Jaskot et al.(2019). There is also a trend for the leaky galaxies tohave significantly higher SFRs based on the IR luminos-ity or the extinction-corrected far-UV luminosity thanthose based on the extinction-corrected H α emission-lineluminosity, and larger than the values in the non-leakygalaxies (median ratios of 2.3 vs. <
0. In fact, the five galax-ies with the largest [SII]-deficiency observed so far in theLyman continuum (three Green Peas and our two tar-gets) are all leaky (see Figure 7).Thus far, we opt not to discuss in depth any statis-tical significance which may be manifested in Figure 6due to the still small sample of confirmed leaky galax-ies. Rather, we think it is more suitable at present timeto describe qualitative trends among the signposts forleakiness to guide future studies. In light of this, weconclude that the following signposts appear to be ro-bust (i.e. properties that are in common among the dif-ferent classes of low-z leaky galaxies): a high SFR/area,lower values for the SFR measured from H α luminos-ity than from the far-UV plus IR continuum luminosity,strong Ly α emission with a significant fraction that isblue-shifted, and abnormally weak [SII] emission.All these signposts have physically plausible connec-tions to the escape of LyC radiation. We have already SII]-Deficiency R Ly G a l a x y c o un t s Leaky ([SII]-weak)Leaky (GP)Leaky (LBA)Non-leaky ([SII]-weak)Non-leaky (LBA) (a) log ( EW Ly ) G a l a x y c o un t s (b) log ( EW H ) G a l a x y c o un t s (c) log (SFR UV,IR /SFR H ) G a l a x y c o un t s (d) log (SFR UV,IR /A) G a l a x y c o un t s (e) log (SFR UV,IR /M ) G a l a x y c o un t s (f) log (M /M ) G a l a x y c o un t s (g)
12 + log (O/H) G a l a x y c o un t s (h) Figure 6.
Histograms of various characteristics of the low-z galaxy samples considered in this paper. Measurements of thethree [SII]-weak galaxies are tabulated in Table 2. We also provide those of the Green Peas and of the Lyman Break Analogsin Tables 7 and 8 respectively in the appendix. discussed why [SII]-weakness could be connected to LyCleakage. A high SFR/area leads directly to a high in-tensity (flux/area) of ionizing radiation, which can leadto an ISM that is optically thin to the LyC. It alsoleads to large values for radiation pressure and the ram-pressure of a starburst-driven wind (e.g. Heckman et al.(2015)). The outward forces these generate can act toexpel the ISM and create channels for the escape of ion-izing radiation. As ionizing radiation escapes the ISM,the rate of H α emission produced by recombination willdecrease. A large Ly α equivalent width implies clearchannels through which photons resonantly scattered offHI atoms can escape, and the blue-shifted emission sug- gests that we are seeing Ly α photons scattered off thenear side of an outflowing wind (e.g. Borthakur et al.(2014)).Finally, it is important to emphasize that these sign-posts are based on global/isotropic galaxy properties(i.e. properties that should depend only weakly on theobserver’s particular line-of-sight to the galaxy). Thiswould imply that leakage occurs in a rather isotropicway, instead of just along certain lines-of-sight.5.3. The Role of Dominant Central Objects
We have discussed the evidence above for a generalconnection between a high SFR/area and the escape2
Wang et al. l o g ([ O III ] / [ O II ]) Leaky (This paper)Leaky (GP)Leaky (LBA) Non-leaky (This paper)Non-leaky (LBA)
Figure 7.
Flux ratio of [OIII]5007/[OII]3727 vs. [SII] defi-ciency for the union of galaxy sample considered in this pa-per. The two leaky [SII]-weak galaxies are shown as red tri-angles, while the other non-leaky [SII]-weak galaxy is shownas a black dot. We see that the [SII]-weak leakers do notexhibit the extraordinarily high ionization level that is char-acteristic of the Green Peas (pink triangles). The remaininggalaxies are drawn from Lyman Break Analogs. of LyC radiation. Here we return to the suggestionin Heckman et al. (2011) and Borthakur et al. (2014)that this escape is made possible by the extreme feed-back effects produced by a “dominant central object”(DCO). These DCOs were discovered to be present in20% of a sample of Lyman Break Analog low-z galaxiesimaged with HST (Overzier et al. 2009). They are de-fined to be compact (marginally resolved by HST), verymassive, young objects located at or near the galacticnucleus, and much brighter in the UV than any otherstar-forming cluster or clump in the galaxy. Heckmanet al. (2011) noted that three of the four candidate leakygalaxies in the sample which they analyzed contained aDCO.As seen in Figure 5, DCOs are present in both ofthe two leaky [SII]-weak galaxies, and produce nearlyall the UV emission. In the third (non-leaky) galaxythere is a significant fraction of diffuse UV emission.While we do not have robust estimates of the masses ofjust the DCOs themselves, we can obtain rough valuesbased on the SB99 models for the far-UV spectra sinceDCOs dominate the far-UV light. The estimated SFRsof 125 and 134 M (cid:12) yr − , and ages of 10 years implythat M (cid:63), DCO > M (cid:12) ). These masses are similar tothe values derived from multi-band SED fits to the sixDCOs in Overzier et al. (2009). The measured radii are ∼
300 pc vs. a mean value of 150 pc for the DCOsin the aforementioned reference. Overzier et al. (2009) log ([SII]/H ) l o g ([ O III ] / H ) [SII] = 0 Figure 8.
Adapted from Figure 6 of Strom et al. (2018).The light purple line is our reference line from which [SII]-deficiency is quantified. The locus of z ∼ (cid:104) z (cid:105) = 2 . σ upper limits on [SII] areshown as dark green triangles. showed that the properties of the DCOs are consistentwith them being the progenitors of central “extra light”component found in the centers of cuspy elliptical galax-ies, which would have formed during a strong starburstin a dissipative galaxy merger.5.4. Comparisons at Higher-Redshift
Before proceeding to further comparisons, we wouldlike to address the validity of our selection criterionwhen it is extended to higher redshifts. Strom et al.(2018) reported spectral measurements from the KeckBaryonic Structure Survey (KBSS) of about 150 star-forming galaxies at z ∼ α but high [OIII]/H β ). We findthat the ridge-line in our Figure 1 passes right throughthe center of the data points in Figure 6 of Strom et al.(2018): see Figure 8. This shows that the method pre-sented in this paper can be straightforwardly appliedto higher redshifts, even though we drew our referencefor defining the [SII]-deficiency based on SDSS. It alsoshows that a minority population of [SII]-weak galaxiesare present at these higher-redshifts.We now compare the properties of the [SII]-weak leakygalaxies to other leaky galaxies at higher redshifts. Stei-del et al. (2018) (hereafter S18) reported the detectionof LyC flux in 15 individual galaxies at z ∼ SII]-Deficiency Table 6.
Comparisons between the mean values calculated from measurements of our two leaky [SII]-weak galaxies,and the median of the S18 sample. For the [SII]-weak sample, we use the values based on the extinction law in Reddyet al. (2015, 2016). Unless otherwise noted, the values for the S18 sample are taken directly from S18. The SFR forS18 is based on the bolometric luminosity in S18 and the prescription in Kennicutt & Evans (2012). The value forM (cid:63) assumes that these galaxies lie along the star-forming main sequence (Reddy et al. 2012). The value for R Ly α isour estimate based on the published stacked spectrum in S18.M FUV log M (cid:63) E(B − V) int EW Ly α R Ly α SFR SFR/M (cid:63) F /F f esc , rel f esc , abs (AB mag) (M (cid:12) ) (mag) (˚A) (M (cid:12) yr − ) (Gyr − )[SII] -21.4 10.5 0.25 23 0.60 130 4.2 0.35 0.77 0.04S18 -20.9 9.8 0.045 28 0.35 25 4.0 0.36 1.21 0.70 sample of 124 galaxies), and in stacked spectra binnedaccording to various galaxy properties. Marchi et al.(2017) have observed 401 galaxies at z ∼
4, and de-tected LyC flux in stacks of spectra binned in variousways. Vanzella et al. (2018) reported the highest red-shift individually-confirmed LyC-leaky galaxy at z = 4,and Vanzella et al. (2019) found evidence of a compactregion emitting LyC radiation at z ∼ <
300 pc)and with strong Ly α emission. Since S18 tabulate themedian properties of their individual detections, we di-rectly compare these values to those of our two leaky[SII]-weak galaxies. This is presented in Table 6.In many respects, the galaxies in the two samples aresimilar, including the properties of the Ly α emission-line, the specific SFR, F /F , and f esc , rel . The [SII]-weak galaxies are somewhat more massive, and havehigher SFRs. The biggest difference is in the largeramount of dust extinction in the [SII]-weak galaxies,which leads to smaller absolute escape fractions. Thismay reflect higher ( ∼ solar) ISM metal abundances(higher dust-to-gas ratio) in the [SII]-weak galaxies. CONCLUSIONSWe have reported on observations with COS on HSTof three low-z ( z ∼ .
3) starburst galaxies, selected onthe basis of the relative weakness of the [SII]6717,6731nebular emission-lines defined with respect to normalstar-forming galaxies. This is a proposed signpost forgalaxies that are optically-thin to ionizing radiation. Wedetect a significant flux below the Lyman limit in two ofthe three galaxies, with relative escape fractions of 93%and 80% respectively and absolute escape fractions of3% and 4%.We have compared these [SII]-weak galaxies to otherknown classes of “leaky” galaxies. Compared to the low-z Green Peas, the [SII]-weak leaky galaxies havesignificantly larger stellar masses, higher metallicities,larger amounts of dust extinction, a much lower ion-ization state (as traced by the nebular emission-lines),smaller Ly α emission-line equivalent widths, and haveoptical spectra that are less dominated by a very young(few Myr-year old) starburst.We have compared the properties of the entire knownset of low-z leaky galaxies to non-leaky starbursts atsimilar redshifts. We find that the leaky galaxies havehigher SFR per unit area, stronger Ly α emission-lines,and a greater fraction of the Ly α emission produced byblue-shifted material. Interestingly, we find that whilethe Green Peas were not selected based on [SII] proper-ties, they too have relatively weak [SII] emission-lines.We also find that leaky galaxies have significantly lowerSFRs based on Balmer emission-line luminosity thanthose based on the intrinsic far-UV plus IR continuumluminosity (as required if a large fraction of ionizing pho-tons escape).We have also compared the [SII]-weak galaxies to sam-ples of leaky galaxies at z ∼ α emission. Compared to the sample of galax-ies at z ∼ ∼ solar) ISM metallicity anda correspondingly higher dust/gas ratio in the [SII]-weakgalaxies. We have also shown that our technique for se-lecting [SII]-weak galaxies can be applied out to redshifts ∼ Wang et al. from Green Peas, this technique potentially expands therange of galaxy properties over which such searches forleaky galaxies can be done. This will improve our oppor-tunities to use low-z leaky galaxies as local laboratoriesin which the physical processes and characteristics thatallow LyC photons to escape can be investigated. It alsosuggests that there may be a variety of different phys-ical conditions and processes that make galaxies leaky.Finally, it gives us an additional technique to identifyleaky galaxies during the EoR using spectroscopic ob-servations with JWST.B.W. thanks Sihao Cheng and Hsiang-Chih Hwang forvaluable conversations, Kate Rowlands for assistanceson SDSS data sets, and Weichen Wang for clarifica-tions on dust extinction. This work is supported byHST-GO-15341, provided by NASA through a grantfrom the Space Telescope Science Institute, which isoperated by the Association of Universities for Re-search in Astronomy, Inc., under NASA contract NAS5- 26555. R.A.O. is grateful for financial support fromFAPESP grant 2018/02444-7. This publication madeuse of data products from the Wide-field Infrared Sur-vey Explorer, which is a joint project of the Universityof California, Los Angeles, and the Jet Propulsion Lab-oratory/California Institute of Technology, funded bythe National Aeronautics and Space Administration; theNASA/IPAC Extragalactic Database, which is operatedby the Jet Propulsion Laboratory, California Institute ofTechnology, under contract with the National Aeronau-tics and Space Administration; and the NASA Astro-physical Data System for bibliographic information.
Facilities:
HST(COS), GALEX, Sloan, WISE
Software:
Astropy (Astropy Collaboration et al. 2013,2018), CalCOS, Matplotlib (Hunter 2007), NumPy (vander Walt et al. 2011), SciPy (Jones et al. 2001), statmorph(Rodriguez-Gomez 2019)REFERENCES
Alexandroff, R. M., Heckman, T. M., Borthakur, S.,Overzier, R., & Leitherer, C. 2015, ApJ, 810, 104Astropy Collaboration, Robitaille, T. P., Tollerud, E. J.,et al. 2013, A&A, 558, A33Astropy Collaboration, Price-Whelan, A. M., Sip˝ocz, B. M.,et al. 2018, AJ, 156, 123Bennett, C. L., Larson, D., Weiland, J. L., et al. 2013,ApJS, 208, 20Borthakur, S., Heckman, T. M., Leitherer, C., & Overzier,R. A. 2014, Science, 346, 216Bouwens, R. J., Smit, R., Labb´e, I., et al. 2016, ApJ, 831,176Calzetti, D. 2011, in Secular Evolution of Galaxies, 419–458Calzetti, D., Armus, L., Bohlin, R. C., et al. 1999, ApJ,533, 682Chisholm, J., Gazagnes, S., Schaerer, D., et al. 2018, A&A,616, 30Fan, X., Carilli, C., & Keating, B. 2006, ARA&A, 44, 415Fan, X., Strauss, M. A., Becker, R. H., et al. 2006, AJ, 132,117Green, J. C., Froning, C. S., Osterman, S., et al. 2012, ApJ,744, 60Gunn, J. E., & Peterson, B. A. 1965, ApJ, 142, 1633Heckman, T. M., Alexandroff, R. M., Borthakur, S.,Overzier, R., & Leitherer, C. 2015, ApJ, 809, 147Heckman, T. M., Borthakur, S., Overzier, R., et al. 2011,ApJ, 730, doi:10.1088/0004-637X/730/1/5 Hunter, J. D. 2007, Computing in Science Engineering, 9,90Izotov, Y. I., Orlitov´a, 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, 4851Jaskot, A. E., Dowd, T., Oey, M. S., Scarlata, C., &McKinney, J. 2019, arXiv e-prints, arXiv:1908.09763Jones, E., Oliphant, T., Peterson, P., et al. 2001, SciPy:Open source scientific tools for Python, , , [Online;accessed ¡today¿]Kennicutt, R. C., & Evans, N. J. 2012, ARA&A, 14, 3Kroupa, P. 2001, MNRAS, 322, 231Leitherer, C., Hernandez, S., Lee, J. C., & Oey, M. S. 2016,ApJ, 823, 64Leitherer, C., Ortiz Ot´alvaro, P. A., Bresolin, F., et al.2010, ApJS, 189, 309Leitherer, C., Schaerer, D., Goldader, J. D., et al. 1999,ApJS, 123, 3Marchi, F., Pentericci, L., Guaita, L., et al. 2017,Astrophysics A&A, 614, 11Mathis, J. 1990, ARA&A, 28, 37Mcquinn, M. 2016, ARA&A,doi:10.1146/annurev-astro-082214-122355
SII]-Deficiency Overzier, R. A., Heckman, T. M., Tremonti, C., et al. 2009,ApJ, 706, 203Pellegrini, E. W., Oey, M. S., Winkler, P. F., et al. 2012,ApJ, 755, 40Pettini, M., & Pagel, B. E. J. 2004, MNRAS, 348, L59Planck Collaboration, Akrami, Y., Arroja, F., et al. 2018a,arXiv e-prints, arXiv:1807.06205Planck Collaboration, Aghanim, N., Akrami, Y., et al.2018b, arXiv e-prints, arXiv:1807.06209Reddy, N. A., Pettini, M., Steidel, C. C., et al. 2012, ApJ,754, 25Reddy, N. A., Steidel, C. C., Pettini, M., & Bogosavljevic,M. 2016, ApJ, arXiv:1606.00434Reddy, N. A., Kriek, M., Shapley, A. E., et al. 2015, ApJ,806, 259Robertson, B. E., Ellis, R. S., Furlanetto, S. R., & Dunlop,J. S. 2015, ApJ, 802, L19 Rodriguez-Gomez, V. 2019, vrodgom/statmorph: v0.3.3, , ,doi:10.5281/zenodo.2535876Steidel, C. C., Bogosavljevi´c, M., Shapley, A. E., et al.2018, ApJ, 869, 123Strom, A. L., Steidel, C. C., Rudie, G. C., Trainor, R. F., &Pettini, M. 2018, ApJ, 868, 117Thomas, D., Steele, O., Maraston, C., et al. 2013, MNRAS,431, 1383van der Walt, S., Colbert, S. C., & Varoquaux, G. 2011,Computing in Science Engineering, 13, 22Vanzella, E., Nonino, M., Cupani, G., et al. 2018, MNRAS,476, L15Vanzella, E., Caminha, G. B., Calura, F., et al. 2019, arXive-prints, arXiv:1904.07941Wright, E. L., Eisenhardt, P. R. M., Mainzer, A. K., et al.2010, AJ, 140, 1868 Wang et al.
APPENDIX
Table 7.
Measurements of Green Pea galaxies in Izotov et al. (2016a,b, 2018a,b).Name M (cid:63) r
50 SFR UV SFR H α SFR UV /A SFR UV / M (cid:63) EW Ly α R Ly α EW H α [OIII][OII] ∆[SII] 12 + log ( OH )(log M (cid:12) ) (kpc) (M (cid:12) yr − kpc − ) (log yr − ) (˚A) (˚A) (dex)J1152 9.59 0.49 2.33 43 -7.78 54.66 0.52 1320 5.4 -0.11 8.0J1333 8.5 0.56 6.38 32 -6.71 60.62 0.22 817 4.8 - 7.76J1442 8.96 0.25 5.01 325 -6.86 80.55 0.24 1122 6.7 -0.26 7.93J1503 8.22 0.29 2.0 102 -6.49 69.17 0.24 1438 4.9 -0.06 7.95J0925 8.91 0.35 2.32 112 -6.99 68.91 0.39 732 5.0 - 7.91J0901 9.8 0.37 1.57 24 -8.48 106.83 0.3 831 8.0 -0.32 8.16J1011 9.0 0.13 2.63 365 -7.38 74.96 0.52 1052 27.1 - 7.99J1243 7.8 0.24 1.99 86 -6.31 83.87 0.52 740 13.5 - 7.89J1248 8.2 0.25 1.19 75 -6.72 107.54 0.47 2561 11.8 -0.68 7.64J1256 8.8 0.24 1.39 29 -7.77 60.2 0.24 955 16.3 -0.26 7.87J1154 8.2 0.18 0.68 25 -7.51 86.48 0.44 1150 11.5 -0.46 7.62 SII]-Deficiency Table 8.
Measurements of Lyman Break Analogs in Alexandroff et al. (2015). “ (cid:78) ” and “x” stands for “leaky” and “non-leaky” respectively.Name leakiness M (cid:63) r
50 SFR UV SFR H α SFR UV /A SFR UV / M (cid:63) EW Ly α R Ly α EW H α [OIII][OII] ∆[SII] 12 + log ( OH )(log M (cid:12) ) (kpc) (M (cid:12) yr − kpc − ) (log yr − ) (˚A) (˚A) (dex)J0055 x 9.7 0.32 0.82 36.65 -8.33 2.32 -1.25 375 3.38 -0.1 8.28J0150 x 10.3 1.37 1.88 3.17 -8.73 3.04 -1.72 199 2.2 -0.17 8.4J0213 (cid:78) (cid:78) (cid:78)(cid:78)