Draftversion January23,2017 PreprinttypesetusingLATEXstyleemulateapjv.8/13/10 MODELING OF LYMAN-ALPHA EMITTING GALAXIES AND IONIZED BUBBLES AT THE ERA OF REIONIZATION Hidenobu Yajima1,2, Kazuyuki Sugimura2, Kenji Hasegawa3, 1FrontierResearchInstitute forInterdisciplinarySciences,TohokuUniversity,Sendai,Miyagi980-8578, Japan 2AstronomicalInstitute, TohokuUniversity,Sendai,Miyagi980-8578,Japanand 3GraduateSchoolofScience,NagoyaUniversity,Furo-cho,Chikusa-ku,Nagoya, Aichi464-8602, Japan 7 Draft version January 23, 2017 1 0 ABSTRACT 2 UnderstandingLyα emitting galaxies(LAEs)canbe a keyto revealcosmicreionizationandgalaxy n formation in the early Universe. Based on halo merger trees and Lyα radiation transfer calculations, a we model redshift evolution of LAEs and their observational properties at z 6. We consider J ≥ ionized bubbles associated with individual LAEs and IGM transmission of Lyα photons. We find 9 that Lyα luminosity tightly correlates with halo mass and stellar mass, while the relation with star 1 formation rate has a large dispersion. Comparing our models with the observed luminosity function by Konno et al. (2014), we suggest that LAEs at z 7 have galactic wind of V & 150 kms−1 out ] and Hi column density of N & 1020 cm−2. Numbe∼r density of bright LAEs rapidly decreases as A HI redshiftincreases,due to bothlowerstarformationrateandsmallerHii bubbles. Our modelpredicts G futurewidedeepsurveyswithnextgenerationtelescopes,suchasJWST,E-ELTandTMT,candetect . LAEs at z 10 with a number density of n a few 10−6 Mpc−3 for the flux sensitivity of h 10−18 ergcm∼−2s−1. By combining these surveLyAsEw∼ith future×21-cm observations, it could be possible p to detect both LAEs with L & 1042 erg s−1 and their associated giant Hii bubbles with the size - Lyα o &250 kpc at z 10. ∼ r Subject headings: radiative transfer – line: profiles – galaxies: evolution – galaxies: formation – t s galaxies: high-redshift a [ 1 1. INTRODUCTION 2000; Iye et al. 2006; Gronwall et al. 2007; Ouchi et al. 2008, 2010; Bond et al. 2011; Ciardullo et al. 2012; v One of the major challenges in today’s astronomy is Yamada et al. 2012). Ouchi et al. (2010) indicated 1 revealing cosmic reionization history with galaxy evolu- 7 tion. Recent CMB observations suggested cosmic reion- that LAEs were hosted in halos with the halo mass 5 ization occurred at z 8 11 (Komatsu et al. 2011; Mh ∼ 1011 M⊙ by the clustering analysis (see also, 5 Planck Collaboration et∼al. 2−014, 2016). Gunn-Peterson Gawiser et al. 2007). Verhamme et al. (2008) suggested 0 tests by the observations of high-redshift QSOs indi- that LAEs were not dust-enriched well yet so that Lyα . photons escaped from galaxies against dust attenua- 1 cated cosmic reionization completed at z 6 (e.g., ∼ tion (see also, Yajima et al. 2014). These suggest LAEs 0 Fan et al. 2006). However, the ionization history of are likely to be in the early phase of galaxy evolu- 7 integer-galacticmedium(IGM)hasnotbeenunderstood tion (e.g., Mori & Umemura 2006). In addition, LAEs 1 yet. Recent observations of high-redshift galaxies at are one of major populations of galaxies contributing : z & 7 are gradually unveiling the cosmic star formation v the cosmic star-formation density in the early Universe history (Ouchi et al. 2010; Bouwens et al. 2012, 2015; i (Ciardullo et al. 2012). X Oesch et al. 2015, 2016), and have allowed us to spec- LAEs also have been used as a tool to investigate ulate the cosmic reionization history (Robertson et al. r the early Universe (Iye et al. 2006; Vanzella et al. 2011; a 2015). Yet, even considering the above observational Ono et al. 2012; Shibuya et al. 2012; Finkelstein et al. constraints, various reionization histories remain viable 2013;Zitrin et al.2015;Song et al.2016b). Numberden- (Cen 2003; Yajima & Khochfar 2015). One of the main sityofLAEsrapidlydecreasesatz &7(Ono et al.2012; uncertaintiesislow-massgalaxyformationwiththe halo massMh lessthan∼1012 M⊙. Duetotheirhighernum- Kgaolanxnioesetwaerl.e2s0i1g4n)ifi,cinadnitclaytaintgtetnhuaatteLdyαdufleuxtoesnferuotmraslohmye- ber density and the higher escape fraction of ionizing drogen in the IGM (e.g., Kashikawa et al. 2006). Mean- photons from them (Razoumov & Sommer-Larsen 2010; while, LAEs themselves could provide sufficient ionizing Paardekooperet al.2013;Yajima et al.2011,2014),low- photons into the IGM (Yajima et al. 2009, 2014), and mass galaxies can be responsible for main ionizing someof themcouldmake giantHiibubbles thatallowed sources. However, the detection sensitivities of current Lyα photons to reach us. In this case, galaxies can be observationsarenotsufficienttoconstraintheformation observed as LAEs. In practice, the most distant LAE of low-mass galaxies. Therefore it is important to the- has been observed even at z = 8.68 (Zitrin et al. 2015). oretically investigate the formation of low-mass galaxies Thus,LAEscanbe the keyobjectsinunderstandingthe and their contribution to cosmic reionization. galaxy formation and cosmic reionization. Mostofhigh-redshiftlow-massgalaxiesareobservedas Inthiswork,weinvestigatetheevolutionofLAEswith Lyαemitters(LAEs;Hu & McMahon1996;Steidel et al. their associated Hii bubbles. By modeling both LAEs [email protected] and Hii bubbles simultaneously, we estimate Lyα lumi- nosityfunctions(LFs),numberdensityofLAEs,sizedis- The relation between stellar and halo mass can be tribution of Hii bubbles, and the relation between Lyα written as dMstar = dlogMstar Mstar. The abundance flux and the size of Hii bubble. These estimations are matching anadlMyshis by BdelohgrMoohziMethal. (2013) indicated uSspeafcuelTfoerlesfcuotpuree(JoWbsSeTrv)aitsiognoainlgmtiossbioenlas.unJchaemdeastW20e1b8b, that ddlologgMMsthar ∼ const for Mh . 1012 M⊙. In addi- and aims to observe galaxies at z 10. Its high sensi- tion, Behroozi & Silk (2015) showed the weak depen- tivityofspectroscopywillmakeitp∼ossibletodetectLyα dence of Mstar on halo mass and redshift at the mass bflluexs.frLomategraolanx,ie3s0-imf thcleayssdgisrtoruibnudteteilnescgoiapnets,HtihiebEubu-- range Mh M∼h1011 −1013 M⊙, although MMsthar increases ropean Extremely Large Telescope (E-ELT), the Thirty with Mh at Mh . 1011 M⊙ (see also, Moster et al. 2013; Kravtsov et al. 2014). Therefore we here assume Meter Telescope (TMT), and the Giant Magellan Tele- scope (GMT), will investigate galaxies at z 10 sta- dMstar const. Thus, we estimate SFR from the growth tistically. In addition, several 21-cm observat∼ional mis- radtMehof∼halo mass with a constant tuning parameter α, sionsareongoing,e.g.,theLOwFrequencyARray(LO- i.e., SFR = αdMh. In order to estimate the growth dt FAR; Harker et al. 2010), and Murchison Widefield Ar- of halos, we use halo merger trees based on an ex- ray (MWA; Lonsdale et al. 2009). In future, Square tended Press-Schechter formalism (Somerville & Kolatt Kilometer Array phase-2 (SKA-2; Dewdney et al. 2009) 1999;Khochfar & Burkert2001,2006). Thehalomerger will perform large-scale 21-cm tomography. Since 21- trees include 5000 realizations with the halo mass range cm emission comes from Hi gas, detections of holes of of 109 1013 M⊙ at z = 6. In this work, we al- 21-cm signal indicate giant Hii bubbles around galaxies. low star−formation only for halos with the mass M > h Therefore, a combination of 21-cm and galaxy observa- 108 M⊙, because star formation in less massive halos tions will provide fruitful information about the cosmic can be significantly suppressed due to UV background reionizationandgalaxyformation(e.g.,Lidz et al.2009). or internal stellar feedback (e.g., Okamoto et al. 2008; Theoretically large-scale simulations showed that Hasegawa & Semelin 2013). In deriving statistical prop- galaxies made patchy ionization structures in an inside- erties, e.g., cosmic SFR density, stellar mass density out fashion (e.g., Iliev et al. 2006, 2012; Mellema et al. and LF, we sum the contribution of each merger tree 2006;Trac & Cen2007;Ocvirk et al.2016),whereaslow- with normalizationfactorsthat reproduce the halo mass density void regions would be ionized first, in the so- function of Sheth & Tormen (2002). We determine the calledoutside-infashion,ifAGNswerethemainionizing tuning parameter by using observed cosmic SFR den- sources. However, AGNs are unlikely to be main ion- sity and stellar mass density at z 7 8. Figure 1 izing sources because of the rapid decrease of observed shows our modeled SFR and stellar∼mas−s density with number density at redshift z > 4 (e.g., Richards et al. observations. We choose the parameter α by the least 2006), although the contribution of faint AGNs is still squarefittingtothe fourpointsofthe observations. The underthedebate(Madau & Haardt2015;Yoshiura et al. observed SFR and stellar mass densities consider only 2016). Thus, in this paper, we model LAEs assuming galaxies with MUV < 17 and Mstar 108 M⊙, respec- the reionization model with the inside-out fashion. In tively. Therefore, we c−onsider only th≥e galaxies satisfy- ordertounderstandcosmicreionizationtheoretically,we ing the above criteria for the fitting. The best fit value need high spatial resolutionto follow the detailed physi- is α=3.3 10−3. Note that, so far there is no available calprocesses ofinterstellar medium, as well as largevol- data about×dMstar at z & 9 and α can change with halo ume to consider large-scale inhomogeneity of ionization dMh mass and redshift. However, for simplicity, we assume structure. Even current numerical simulations cannot α is constant. Following the above way, we derive star resolve such a wide dynamic range. For that reason, formation history from each halo merger tree. in this work, we semi-analytically investigate the cos- We also derive UV LFs by converting SFR to UV mic reionization and galaxy formation based on simple structure formationmodels and radiative transfer calcu- flux with Lν,UV = 0.7 × 1028 erg s−1 1 MS⊙FRyr−1 lations. We use the cosmological parameters, ΩΛ = 0.7, (Madau et al.1999). Figure2showsourmo(cid:16)deledLFsa(cid:17)t ΩM =0.3, Ωb =0.045 and h=0.7(Komatsu et al. 2011; z 7and 8with thosefromthe recentobservationby Planck Collaboration et al. 2016). Bo∼uwens e∼t al. (2015). The observationindicated that a The paper is organized as follows. We describe our part of UV flux are absorbed by dust with the escape models of star formation and Hii bubbles in 2. In 3, fraction f 0.6. Our modeled LFs match the ob- wepresenttheresults,whichincludetheLyαp§roperti§es, servationewsce,lUlVwi∼th the same UV escape fraction. LFsofLAEsandredshiftevolutionofnumberdensityof Galaxies ionize the IGM as star formation proceeds. LAEs. We discuss the relation between the Lyα flux of Inaone-zoneapproximation,weestimatetimeevolution galaxiesandthesizesofHiibubblesin 5,andsummarize of cosmic ionization degree (Barkana & Loeb 2001), § in 6. § dQ 1 HII = n˙γ f α C(1+z)3n0Q , (2) 2. MODEL dt n0H ion esc,ion− B H HII Inthecurrentstandardpictureofstructureformation, where Q is the volume fraction of Hii, n0 is the halos grow via minor/major mergers. In our model, the present-dHaIyIhydrogennumberdensity( 2 10−H7cm−3), star formation rate (SFR) in a halo is assumed to be n˙γ is the intrinsic ionizing photon em∼iss×ivity per unit proportional to the growth rate of the halo as follows: ion volume, α is the case-B recombination rate, C is a B clumpiness factor of IGM, and f is escape fraction dM dM dM esc,ion star star h SFR= = . (1) of ionizing photons, which is a free parameter here. We dt dM dt h 2 estimate n˙γ from star formation history of each halo ion byusingStarburst99withtheSalpeterinitialmassfunc- tion with the metallicity of 10−2 Z⊙. The reionization history is shown in the upper panel of Figure 3. As f increases, the IGM is ionized earlier. We as- esc,ion sume the recombination rate for T = 104 K (α = B 2.6 10−13 cm3 s−1)andC =3,as suggestedby numer- × ical simulations (Pawlik et al. 2009; Jeon et al. 2014). Thecosmicreionizationhistoryisregulatedbyf . esc,ion Free electrons produced by the cosmic reionization con- tribute to the Thomson scattering optical depth (τ ) of e CMB photons, defined as zrec dt τ = σ n (z)c dz, (3) e T e dz Z0 (cid:12) (cid:12) (cid:12) (cid:12) where zrec = 1100 is the redshift (cid:12)at t(cid:12)he time of recom- (cid:12) (cid:12) bination. In this work, we assume the single ionization fraction of helium is the same as the one for hydrogen at z 3, and the double ionization takes place at z < 3 ≥ (e.g.,Wyithe et al.2010;Inoue et al.2013). Recentsim- ulations show that indeed the fraction of Heii is close to the Hii fraction at high redshifts, although the ion- Figure 1. Upper panel: Star formation rate density. Red ization fraction of helium is slightly lower than the one line represents our modeled star formation history based on the for hydrogen (Ciardi et al. 2012). The top panel of Fig- halo merger trees. Triangle symbols show the observation by ure 3 represents the ionization history of hydrogen gas Bouwensetal.(2015). Theobservedstarformationratedensities areestimatedbyintegratingtheluminosityfunctionsintherange with the different fesc,ion. The bottom panel of Figure 3 ofMUV≤−17. Ourmodelalsoconsideronlythegalaxiesbrighter shows τe with different fesc,ion. The Thomson scatter- thanthesamelimitingmagnitude. Lowerpanel: Stellarmassden- ing optical depth τ increases with redshift in a way de- sity. Redlineisthecumulatedstellarmassofourmodelconsidering pending on f edue to the different ionization histo- only the galaxies with Mstar ≥ 108 M⊙. Triangle symbols show esc,ion theobservationbySongetal.(2016a),whointegratedthederived ries. We find fesc,ion = 0.2 nicely reproduces the CMB stellarmassfunctionsforgalaxieswithMstar≥108 M⊙. observation (Planck 2016). Therefore, in this work, we adopt f = 0.2 with no redshift evolution. In fact, esc,ion Yajima et al.(2014)showedf isconstantwithred- esc,ion shift and 0.2 by cosmological simulations with radia- ∼ tive transfer calculations. 3. RESULTS 3.1. Evolution of ionized bubbles around galaxies We follow the growth history of Hii bubbles around individual galaxies with the star formation efficiency α and f estimated above. Sizes of Hii bubbles evolve esc,ion with ionizing photon emissions, recombination, and cos- mic expansion as follows (Cen & Haiman 2000): dR3 3N˙γ f HII =3H(z)R3 + ion esc,ion Cn (z)α R3 , dt HII 4πn (z) − H B HII H (4) where H(z) is the Hubble constant at specific redshifts, and N˙γ is intrinsic ionizng photon emissivity of each ion galaxy. The ionizing fronts can propagate up to the Stro¨mgren radius during the recombination time scale (Spitzer 1978). The recombination time scale is t rec 1 0.5 Gyr 1+z −3, which is longer than th∼e αBnH ∼ 11 typical time scale over which SFR changes more than (cid:0) (cid:1) factor 2. Therefore, Hii bubbles do not reach the ∼ equilibrium state, and we need to consider SFR history to estimate the sizes of Hii bubbles at given redshifts. Figure 2. UV luminosity functions. Open squares and circles We estimate probability distribution function (PDF) of represent observed LBGs at z ∼ 7 and ∼ 8 by Bouwens etal. the sizes of ionized bubbles (RHII) as shown in Fig- (2015). ure 4. In our model, higher mass halos tend to pos- sess larger ionized bubbles. Due to the decrease in the number density of halos on the high-mass end of a halo 3 Figure 3. Upperpanel: Ionizationhistory. Differentlinesshow ionizationhistorieswithdifferentescapefractionsofionizingpho- tons. Lower panel: Thomson scattering optical depth. Different Figure 4. Probabilitydistributionfunctions ofsizesofHiibub- lines are estimated with different ionization histories presented in bles associated with individual galaxies. Vertical dash lines are the upper panel. Yellow shade represents the estimation by the correspondingtotheviewingangleof1arcmin. CMBobservation(PlanckCollaborationetal.2016). scattering process. However, if dust mainly distributes mass function (Sheth & Tormen2002), the PDF ofRHII in Hi gas clumps, fesc,Lyα does not become lower than rapidly decreases at larger RHII. Although the ioniz- fesc,UV because Lyα photons are scattered by hydrogen ing front does not reach the size of Stro¨mgren sphere on surface of the clumps before interacting with dust. drsitca∝tor(N˙oiγfons)i1z/e3s(1of+Hzi)i−b2,ubitblceasn. bAesurseeddshasiftadreocurgehasiens-, fCoirarodbuslelroveetdaLl.A(E2s01a2t)zindic2a.teIdntahdadtitfieosnc,,Lycαos∼mofleosgci,UcaVl the IGM density decreases while the number density of simulations of Yajima et∼al. (2014) showed that f esc,Lyα massive halos with higher ionizing photon emissivity in- was & 0.6 and similar to f at z & 6. Therefore, in esc,UV creases. Therefore, the tail of PDF in the large-RHII this work, we assume that fesc,Lyα =fesc,UV =0.6. end shifts to larger R at lower redshift. Future 21 Next, we estimate IGM transmission as a function of HII cm observations, e.g., SKA-2, is supposed to probe the wavelength. As in Cen & Haiman (2000), we divide the IGM ionization structure with the angular resolution of paths along which Lyα photons travel from galaxies to 1′. Therefore,atz .10,the tail ofPDFatlargeR us into the two parts, i.e., outside and inside ionized HII ∼ can be observationally investigated in future. The halo bubbles,andseparatelyestimate eachcontribution. The numberdensitymonotonicallyincreasesasthehalomass transmission outside ionized bubbles is estimated as fol- decreasesatafixedredshift. Sincethe sizeofHiibubble lows: is positively related with halo mass as will be shown in zi dt Section 3.3, it seems that the PDFs in the small-RHII τ(λobs,zs)= dzc nH(z)xHIσLyα[λobs/(1+z)], end monotonically increase as RHII decreases. However, Zzr dz there are peaks in the PDFs, below which they decrease (6) amsaRssHfIoIrdestcarerafsoersm. aTthioisnisimcapuosseeddbinytthhiestwhroerskh.oldofhalo where zi ∼ zs − RHIIR×H(1+z). Here, zr is redshift when the cosmic reionization completes, which we set z = 6, r z is redshift when Lyα photons pass through ionizing 3.2. Lyα luminosity functions i front,z isredshiftofgalaxy,σ isthescatteringcross s Lyα Absorptionofionizing photons by interstellar medium sectionfor Higas,andR is the size ofthe cosmological H within galaxies results in Lyα emissions via recombina- horizon at z . We assume the outside of the bubble is s tion processes, while escaped photons cause the cosmic completely neutral, i.e., x = 1. The σ is estimated HI Lyα reionization as shown in the previous section. The Lyα by (Verhamme et al. 2006) luminosity (L ) is estimated by Lyα −1 T 2 H(x,a) LLyα =0.68(1.0−fesc,ion)fesc,LyαǫLyαN˙iγon, (5) σLyα[λ]=1.041×10−13 104 K √π . (7) (cid:18) (cid:19) where f is the escape fraction of Lyα photons esc,Lyα We set T = 104 K in this work. Here, H(x,a) is the from galaxies, ǫ = 10.2 eV is the energy of a Lyα Lyα Voigt function, photon. The escape fraction of Lyα photons f esc,Lyα can be lower than fesc,UV because the path length of a +∞ e−y2 Lyα photons until escape can be longer due to multiple H(x,a)= dy (8) π (x y)2+a2 Z−∞ − 4 where x (ν ν )/∆ν , ν = 2.466 1015 Hz is ies by combining cosmologicalSPH simulations with ra- 0 D 0 ≡ − × the line-center frequency, ν is the Doppler width, a = diationtransfercalculations. Thesecolumndensities are D ∆ν /(2∆ν ), ∆ν = 9.936 107 Hz is the natural line optically thick to ionizing photons, hence not consistent L D L × width. We here use the fitting formula of H(x,a) given with f = 0.2. However, recent simulations showed esc,ion by Tasitsiomi (2006). ionizing photons mostly escape along ionized holes cre- Even inside ionized bubbles, a tiny fraction of neutral ated by radiative and SNe feedback (e.g., Yajima et al. hydrogensexist. WeestimatetheIGMtransmissionfrom 2009, 2011; Kimm & Cen 2014). Thus, f of 20 % esc,ion ionizing front to virial radius with the neutral fraction can be considered as the fraction of viewing angle along under the ionization equilibrium state, which star forming regions are not covered by Hi gas. For simplicity, we do not take account of the effect of C r 2 N˙γ −1 1+z 3 such holes on line profiles. Note that, however,Lyα line xHI =1.5×10−5(cid:18)3(cid:19)(cid:18)kpc(cid:19) 1050iosn−1! (cid:18) 8 (cid:19) . porroofitlheesrsdoemtaeiwlehdasttrcuhcatnugreeidnuHeitogatsh(eDhijoklsetsr,ac&lumKprainmeessr (9) 2012). Notethat,theopticaldepthoutsidetheionizedbubbleis The shaded regions in Figure 6 represent the LFs us- dominant in our work. The IGM transmission is mostly ing different Lyα line profiles with the velocity range almost zero at λ . λ , where λ = 1216 ˚A is the wave- from V = 300 km s−1 to 300 km s−1. The best 0 0 out length of Lyα line center. fit velocities a−re 180 km s−1 (model A), 190 km s−1 ConsideringtheIGMtransmission,wederiveLyαLFs, (model B), and 110 km s−1 (model C), as summarized and compare them with the observation of Konno et al. in Table 1. Only model A with the velocity range (2014). Depending on the shape of Lyα line profile, the V 150 300 km s−1 can reproduce the observed out IGM transmission significantly changes. Even with re- LF w∼ell. Th−erefore we suggest LAEs are likely to have centdeepspectroscopies,however,itisdifficultto deter- high Hi column density with & 1020 cm−2 and outflow- mine intrinsic Lyα line profiles, i.e., before the IGM ex- ing Hi gas with velocity &150 km s−1. The Lyα profile tinction (e.g., Ouchi et al. 2010). The intrinsic Lyα line monotonically shifts to redder one as the outflow veloc- profiledependsonthephysicalnatureofgalaxies,e.g.,Hi ity increases in model A as shown in Figure 5. On the column density and velocity field (e.g., Verhamme et al. otherhand,asthe Hicolumndensity decreases,the Lyα 2006). In this work, we calculate intrinsic Lyα line pro- profilemovesbacktothelinecenterfrequencyatspecific filesbyLyαradiationtransfersimulationsusingthecode velocity of < 300 km s−1. This is because Lyα photons developed in Yajima et al. (2012b), and study the phys- canescapefromthecloudbeforeshiftingtolongerwave- ical nature of LAEs through the comparison with the lengthdue to loweropticaldepth. In this work,the best observation (Konno et al. 2014). The intrinsic line pro- fit velocities for model B and C roughly correspond to files are calculated based on spherically outflowing gas those producing the Lyα profiles shifted farthest away. cloud models with the following velocity structure, In addition, the width of line profile becomes smaller as the Hi column density decrease. Even with the best fit r velocities, the number densities of LAEs in the model B V(r)=V , (10) out R and C are smaller than the observation because of the (cid:18) edge(cid:19) narrower line profiles resulting in lower IGM transmis- where Redge is the edge of the spherical cloud and Vout sion. Hence, in order to get higher IGM transmission to is the outflow velocity at Redge. Although the simu- reproduce the LF, LAEs are likely to have the column lated profiles do not depend on Redge, here we consider densityhigherthan 1020 cm−2. Thus,inthiswork,we Redge Rvir. Figure 5 shows the modeled line profiles consider model A as∼our fiducial model. ∼ with various outflowing velocities. The Lyα line profiles Emergent Lyα line profiles are shown in Figure 7. get the asymmetric shape with a peak at redder wave- As Hi column density decreases, intrinsic Lyα line pro- lengthfortheoutflowvelocityfield,becauseLyαphotons filesbecomenarrowerandpeakpositionsshifttoshorter at bluer wavelength are scattered by Hi gas due to the wavelength. IGM transmission increases with wave- Doppler shift. As NHI increases, the line profiles are ex- length because photons with long wavelength redshift tended, and the peak frequency shift farther from the early to avoid from being scattered by the IGM, while line center frequency. For the inflow velocity field, the Lyαfluxneartheline centeris reducedefficiently bythe Lyα line profile becomes the mirror symmetric shape to IGMscattering. FWHMsoftheemergentlineprofilesare the one for outflow with the same absolute value of the 1.5˚A(modelA),8.9 10−1˚A(modelB)and5.2 10−2˚A velocity. × × (model C). Therefore it is difficult to distinguish the With the simulated intrinsic line profiles, we can esti- different column density models in the current spectro- mate Lyα LFs for given N and V . In this work, we HI out scopicobservationwiththeresolutionofR 1000 2000 infer typical N and V of LAEs by comparing mod- ∼ − HI out (e.g., Shibuya et al. 2012). Future high-dispersion spec- eled LFs with the observed one by Konno et al. (2014). troscopies with R 2000, e.g., Prime Focus Spectro- Figure 6 show the LFs. Here we calculate the LFs for ≫ graph on Subaru or JWST, will be able to reveal the three Hi column density models, N = 2 1020 cm−2 HI detailed shape of profile. (model A), N = 2 1019 cm−2 (mo×del B), and HI Line profiles in inflowing gas models result in the un- NHI = 2 1018 cm−2 (×model C). The column density derproductionof observable LAEs, because mostof Lyα × of model A is corresponding to Damped Lyman-α Sys- photonsarescatteredbytheIGM.Thisisconsistentwith tems (DLAs: Wolfe et al. 2005). Yajima et al. (2012a) the observation by Ouchi et al. (2010), which indicated showed that DLAs distributed at lines of sight passing LAEs at z =6.6 are likely to have outflowing gas by the star-forming regions in high-redshift star-forming galax- 5 Figure 5. Lyα line profiles from expanding spherical gas clouds with the velocity field as V(r) = Vout(cid:16)Rerdge(cid:17), where Redge is the radius of clouds and Vout is the velocity at Redge. The different panels show the Lyα profiles considering different Hi column densities: NHI = 2×1020 cm−2 (panel A), 2×1019 cm−2 (panel B), and 2×1018 cm−2 (panel C). The different lines represent the different expanding velocities at Redge. The φ(λ) is normalized to be unity when it is integrated over the wavelength. The φ(λ) of panel (B) and (C)isartificiallyreducedbyafactor2and10. composite spectrum. In addition, Shibuya et al. (2014) minosity and R tightly correlate with M . The HII star measuredoutflowvelocitiesofindividualLAEs atz 2, L -M relationdoesnotchangewith redshiftsignif- Lyα star ∼ andindicatedthatLAEswerelikelytohaveoutflowwith icantly,while theR -M becomessmallerasredshift HII star V &150 kms−1. increases. ThisisbecauseHubble constant(i.e.,expand- out Next we estimate the redshift evolution of LF based ing velocity of IGM) becomes large at higher redshift. onthe modelA with V =180kms−1. Figure8 shows Therefore, although R decreases as redshift increases out HII the modeled LFs at z =7,8,10 and 12. Note that, here due to higher IGM density, the IGM transmission does weusethesamelineprofileforallhalosandredshift. At not decreases significantly. Thus, Lyα luminosity does higherredshifts,typicalSFRissmallerduetolowerhalo not depend sensitively on redshift. mass and halo growth rate. In addition, as redshift in- The detection sensitivity of recent observations of creases,typicalsizeofHiibubblesdecreases,resultingin LAEsatz 7 8wascorrespondingtoLyαluminosityof the lowerIGM transmission. As a result, the LF rapidly &3 1042∼erg−s−1 (Ono et al.2012;Shibuya et al.2012; × shifts to the fainter side at higher redshifts. Finkelstein et al.2013;Vanzella et al.2011;Konno et al. 2014;Zitrin et al.2015). Inourmodel,medianandmin- 3.3. Relation between Lyα luminosity and stellar mass imum stellar masses producing L 3 1042 erg s−1 Lyα Figure 9 shows the sizes of Hii bubbles and Lyα lumi- at z =7.3 are 6.5 109 and 1.5 108∼M⊙×, respectively. nosities considering the IGM transmission. Lyα proper- × × Therefore, by considering the relation M 3.3 star ties are calculated by using the Lyα profile of model A. 10−3 M , we suggest that the observed LAEs at∼z =7×.3 h Tplhe.eWsheadseesertehpartesReHntIItthieghrtalnygceoorrfe2l5a%te−w7it5h%stienlltahremsaamss-, tshheoumldedbieanhohsatleodminahssalioss1w.9ith1M01h2≥M4⊙.6. ×1010 M⊙,and as R M1/3, while the relation with SFR shows a × HII ∝ star 3.4. Redshift evolution of number density of observable large dispersion. LAEs SFR rapidly increase by major merger. However, R is not so sensitive to the short-time fluctuation So far the Lyα line has been used as the most strong HII of SFR because of the longer time-scale for reaching tool to confirm the redshift of distant galaxy candidates the ionization equilibrium state. As a result, the rela- (e.g.,Iye et al. 2006; Finkelstein et al. 2013; Zitrin et al. tion between R and SFR shows the large dispersion. 2015). However,itiswidely thoughtthatLAEsatz >9 HII High-redshift galaxies have been observed as so-called are difficult to be detected because of the IGM opac- Lyman-break galaxies (LBGs: Bouwens et al. 2012), via ity. Here, we estimate the number density of LAEs the Lyman-break technique so far. Our results indicate (n ) with higher Lyα flux than specific thresholds. LAE LBGs at z & 7 with similar UV brightness can have Figure 10 shows the number density of LAEs with different Lyα fluxes due to the scatter of IGM transmis- F 10−17,10−18 and 10−19 erg s−1 cm−2. The Lyα ≥ sion. At z = 15, our sample is limited by the stellar detection limits of current observations with a reason- mass of galaxies Mstar 109 M⊙. In our model, since able integration time are 10−17 erg s−1 cm−2 (e.g., ∼ ∼ the stellar mass is simply proportional to halo mass as Shibuya et al. 2012). As explained in Sec. 3.2, the M 3.3 10−3 M , this means that there is no pro- number density of bright LAEs monotonically decreases star h genito∼rs wit×h Mh & 3 1011 M⊙ at z = 15 in our halo with increasing redshift. Given the detection limits of sample,whichisconstr×uctedtohavethemassrangefrom 10−17 ergs−1 cm−2,widefieldsurveysof 1003 Mpc3 are 109 to 1013 M⊙ at z =6. able to detect LAEs up to z 8.5. This is consistent ∼ In contrast to the SFR-R relation, both Lyα lu- with recent observed LAEs at z . 9 (Finkelstein et al. HII 6 Figure 6. Lyα luminosity functions (LFs) at z =7.3. Square symbols represent the observed LF of LAEs at z =7.3 by Konnoetal. (2014). DifferentpanelsshowmodeledLFsbasedonLyαlineprofilestodifferentHicolumndensities. Magentaandcyanshadesshowthe rangeofLFsconsideringLyαlineprofilewithdifferentoutflow andinflowvelocitywiththerange0∼±300kms−1. Blacksolidlinesare bestfittedones totheobservation. Blackdashlineintheleftpanel showstheLFbeforeconsideringIGMtransmission. Figure 8. Lyα luminosity functions at z = 7.3,8,10 and 12. OpensquaresshowobservedLAEsatz=7.3(Konnoetal.2014). 200 km s−1, the number density of LAEs with F Lyα 10−18 erg s−1 cm−2 at z 10 is a few 10−6 Mpc−≥3. Figure 7. Lyα lineprofiles. Red solidlines show emergent line AsshowninFigure9,brig∼htLAE∼sarehos×tedinmassive profiles of a halo of 5.9×1011 M⊙ at z=7.3, which evolves to a halos. The median halo and stellar mass of LAEs with halo of 1.0×1012 M⊙ at z = 6.0. Intrinsic Lyα luminosity of F 10−18 erg s−1 cm−2 at z = 10 are 1.1 1012 the halo is 1.3×1043 ergs−1. Emergent luminosities after IGM Lyα ≥ × transmissionareshowninthepanels. Blackdashlinesaretheline and 3.5 109 M⊙, respectively. It was suggested that × profilesbeforeconsideringIGMtransmission. the observed LBG at z = 11.1, GNz11, had the stellar mass of 109 M⊙ (Oesch et al. 2016), which is corre- sponding∼to F 0.2 10−18 erg s−1 cm−2 in our Lyα 2013; Oesch et al. 2015; Zitrin et al. 2015). The LAEs model. Hence, it w∼ill be c×hallenging to detect Lyα flux with FLyα 10−17 at z 10 are quite rare, with from GNz11 even by future spectroscopies with the line n 1 ≥2 Gpc−3. ∼ sensitivity of F 10−18 erg s−1 cm−2. LAE Lyα ∼ − ∼ Spectroscopies of next generation telescopes, e.g., Different line profile models predict different number JWST, are supposed to achieve the sensitivity of densities of observable LAEs. The IGM transmission 10−18 erg s−1 cm−2 with a reasonable integration tim∼e. becomes more sensitive to the intrinsic Lyα line profile If galaxies at z 10 have outflowing gas with v & models, since the typical size of Hii bubbles gets smaller ∼ 7 Table 1 Modelparameters Model NHI/cm2 Vout/kms−1 A 2×1020 180 B 2×1019 190 C 2×1018 110 fesc,ion=0.2 fesc,UV =0.6 fesc,Lyα=0.6 NOTES.—ForeachHicolumndensity, theoutflow velocityis chosentoreproducetheobservedluminosityfunctionofLAEat z=7.3(Konnoetal. 2014). Escapefractionsofionizing,UVand Lyαphotons aresameforallmodels. we estimate the total kinetic energy of outflowing gas as Figure 9. Upper panel: Sizes of Hii bubbles as a function of stellarmassandSFR.Green,redandbluelinesrepresentmedian 1M V2 f e N GMhMgas. (11) values to galaxies at z = 8,10 and 15, respectively. The shades 2 gas out ∼ conv SN SN− R show quartiles at each bin. Lower panel: Lyα luminosity as a vir functionofstellarmassandSFR. Here M = (1 ǫ)M0 is gas mass after star forma- gas − gas tion, M = ǫM0 is stellar mass, M0 is initial gas star gas gas mass, ǫ M /M0 is a star formation efficiency with with increasing redshift. As a result, the difference of ≡ star gas respecttotheinitialgasmass,f istheconversioneffi- conv n amongmodelA,BandCbecomeslargerathigher LAE ciencyfromtotalsupernovaenergyto kinetic one ofgas, redshift, and more than order unity at z 10. Future e 2 1051 erg is the releasedenergy for eachsuper- ∼ SN observation would also allow us to discriminate intrinsic ∼ × nova (e.g., Hamuy 2003), and N is the number of su- SN line profiles,whichinturnprovideinformationaboutHi pernova. For Salpeter-like IMF, N 1 10−2Mstar = column density and outflow velocity, by comparing the SN ∼ × M⊙ observed number density of LAEs with the theoretical 1 10−2 ǫ Mgas. By using the star formation effi- models. × 1−ǫ M⊙ Even next generation telescopes, e.g., GMT, E-ELT, ciency ǫ, w(cid:16)e sim(cid:17)ply estimate Vout as TMT, will be difficult to have the sensitivity of 10−19 erg s−1 cm−2. However,if future telescopes some∼- V2 +v2 1.4 103 km s−1 ǫ 12 f12 , (12) how achieve such a high sensitivity, wide field surveysof out esc ∼ × 1 ǫ conv 1003 Mpc3 would be able to reach LAEs at z 12. q (cid:18) − (cid:19) ∼In this work, we consider isolated Hii bubble∼s asso- ciated with individual LAEs in the estimation of IGM where vesc ≡ 2RGvMirh is the escape velocity of a halo. transmission. However, at lower redshift, Hii bubbles Next we obqtain fconv as a function of ǫ. Recently can be overlapped each other (e.g., Iliev et al. 2012; Kim & Ostriker(2015)showedthefinalmomentumpro- Hasegawa & Semelin 2013; Ocvirk et al. 2016). The duced by a single SN with the energy of 1051 erg as soivoenr.laTppheisdeHffieictbuwbiblllebseciannveesntihgaantceedtihneHIaGsMegatwraanesmt aisl-. follows: p = 2.8 × 105 kms−1 M⊙ 1 cnmH−3 −0.17 (see also, Cioffi et al. 1988; Thornton et al. 1998). In this (in preparation) by combining large-scale N-body with (cid:0) (cid:1) work, we ignore the weak density dependence and as- small scale radiative-hydrodynamicssimulations. sume n 1 cm−3. Since the final momentum is al- H ∼ 4. DISCUSSION most linearly proportional to the injected SN energy (Cioffi et al. 1988; Kimm & Cen 2014), we approximate 4.1. Condition for galactic outflow from LAEs the total momentum produced by multiple SNe as fol- As shown in Section 3.2, high-redshift LAEs are likely lows: P = p eSNNSN , where e N is the total en- to have galactic wind with V & 150 km s−1. Here 1051 erg SN SN we roughly derive the conditioonutfor making gas outflow ergy of super(cid:16)novae. (cid:17)Therefore, using E = P2 , we with V & 150 km s−1 in a spherical gas cloud model. 2Mgas out derive f = E =16 ǫ . Note that f can not Supernova (SN) feedback can be responsible for causing conv ESN 1−ǫ conv strong outflow in high-redshift low-mass galaxies (e.g., exceed unity according t(cid:16)o the(cid:17)energy conservation. For Kimm & Cen 2014; Kimm et al. 2015). In the assump- that reason, we set f = 1 at ǫ 0.06 because the conv ≥ tion that supernovae give feedback to all gas uniformly, above expression gives f > 1. Thus, Equation 12 is conv 8 Figure 10. Number density of LAEs as a function of redshift. Red, blue and green lines show the number densities of LAEs with FLyα ≥ 10−18 ergcm−2s−1 obtained assuming Lyα line profiles of model A, B and C, respectively. Black dash and dot lines are the number density of LAEs with FLyα ≥ 10−17 ergcm−2s−1 and FLyα ≥ 10−19 ergcm−2s−1 obtained assuming the Lyα line profile of modelA. written by using ǫ as follows: tend to have higher ionizing photon emissivity and in- trinsic Lyα luminosity for a fixed size of Hii bubbles at 5.5 103 km s−1 ǫ higherredshift. Thecombinationoftheseeffectsleadsto × 1−ǫ the weak redshift dependence in the L -R relation. qVo2ut+ve2sc ∼1.4×103 km s−1 (cid:16)1−ǫǫ(cid:17)12if ǫ<0.06 foofrT≥hze1yealrlco1mw0.isnhaaTdnhedidstrhreeeggifloionunxrecopofrrer≥seesnp1to0sn−Ltd1yh8sαeetvroigHewItsIhi−ne1gLcamAngE−l2es (cid:16) (cid:17)if ǫ 0.06. with as∼sociated Hii bubbles that are detectable both as  ≥ (13) LAEs and holes in 21-cm signal by future galaxy ob- We find that the condition of ǫ 0.04 is required to servations by JWST and 21-cm tomography by SKA- cause the galactic outflow with V ∼ = 180 kms−1 from 2, respectively. Future 21-cm observations will be able out a halo with Mh = 4.6 1010 M⊙ at z = 7.3 which is to probe giant Hii bubbles around bright LAEs with minimum halo mass to×produce the observable Lyα lu- F &10−18 erg s−1 cm−2 at z 10. Lyα minosity 3 1042 erg s−1. We can also convert the The differential brightness tem∼perature δT caused b ∼ × tuning parameter α in our star formation model to ǫ, by galaxies shows inner positive and outer negative as ǫ αΩ /Ω 0.02. This is roughly similar to the ring-like structure (e.g., Chen & Miralda-Escud´e 2004; M b ∼ ∼ value estimated above. As halo mass increases, higher ǫ Yajima & Li 2014). The detailed structure depends on is required to cause galactic outflow. Here we have es- SED. If galaxies host X-ray sources like AGNs, the pos- timated ǫ assuming all gas has same outflow velocity in itive region is extended due to partial photo-ionization the simple spherical cloud model. For example, in the heating. Therefore, future 21-cm observations may also caseofdiskgalaxies,onlyapartofgascanbe evacuated give us information about nature of X-ray and UV along the normal direction to galactic disk as shown in sources in bright LAEs. numerical simulations (e.g., Agertz et al. 2011). In this case, strong outflow can be caused even with smaller ǫ 5. SUMMARY because piled gas mass can be lower. We present models of Lyα emitting galaxies (LAEs) with IGM transmission considered at the era of reion- 4.2. Relation between Lyα luminosity and size of ization. Based on halo merger trees and a simple star ionized bubble formationmodel,we estimate cosmicstarformationand LAEscanbe responsiblefor ionizingsourcesofcosmic cosmic reionization history. Our model uses 5000 real- reionization (e.g., Yajima et al. 2009, 2011, 2014). Fig- izations of halo merger trees with the halo mass range ure 11 showsLLyα as a function ofthe size ofassociated from Mh = 109 to 1013 M⊙ at z = 6. As a result, Hii bubble R . The Lyα luminosity steeply increases our model reproduces the observed cosmic star forma- HII with R due to higher IGM transmission for large Hii tion densities, stellar mass densities, luminosity func- HII bubbles. The relation between L and R does not tions of Lyman-break galaxies with a tuning parameter, Lyα HII significantly change with redshift. IGM gas density in- α( SFR/dMh)=3.3 10−3. Our modeled star forma- creases with redshift as ρ (1+z)3, resulting in lower tio≡n history,dwtith an es×cape fraction of ionizing photons ∝ IGM transmission at higher redshift. However, galaxies f =0.2, also provides a cosmic reionizationhistory esc,ion 9 numerical simulations were performed on the computer cluster, Draco, at Frontier Research Institute for Inter- disciplinary Sciences of Tohoku University. This work is supported in part by MEXT/JSPS KAKENHI Grant Number 15H06022 (HY) and 15J03873(KS). REFERENCES Agertz,O.,Teyssier,R.,&Moore,B.2011, MNRAS,410,1391 Barkana,R.,&Loeb,A.2001,Phys.Rep.,349,125 Behroozi,P.S.,&Silk,J.2015,ApJ,799,32 Behroozi,P.S.,Wechsler,R.H.,&Conroy,C.2013,ApJ,770,57 Bond,N.,Gawiser,E.,Guaita,L.,Padilla,N.,Gronwall,C., Ciardullo,R.,&Lai,K.2011,ArXive-prints Bouwens,R.J.,Illingworth,G.D.,Oesch,P.A.,Trenti,M., Labb´e,I.,Bradley,L.,Carollo,M.,vanDokkum,P.G., Gonzalez, V.,Holwerda,B.,Franx,M.,Spitler,L.,Smit,R.,& Magee,D.2015,ApJ,803,34 Bouwens,R.J.,Illingworth,G.D.,Oesch,P.A.,Trenti,M., Labb´e,I.,Franx,M.,Stiavelli,M.,Carollo,C.M.,van Dokkum,P.,&Magee,D.2012, ApJ,752,L5 Cen,R.2003, ApJ,591,12 Cen,R.,&Haiman,Z.2000,ApJ,542,L75 Chen,X.,&Miralda-Escud´e,J.2004,ApJ,602,1 Ciardi,B.,Bolton,J.S.,Maselli,A.,&Graziani,L.2012, Figure 11. Lyα luminosity as a function of size of Hii bubble. MNRAS,423,558 Triangle,circleandsquaresymbolsrepresentmedianvaluesatz= 8,10and15,respectively. Errorbarsshowquartiles. TheLyαline Ciardullo,R.,Gronwall,C.,Wolf,C.,McCathran,E.,Bond, profileofmodelAisused. HorizontaldashlineshowsLyαfluxof N.A.,Gawiser,E.,Guaita,L.,Feldmeier,J.J.,Treister,E., 10−18 ergcm−2s−1 for z = 10. Vertical dash line represents the Padilla,N.,Francke, H.,Matkovi´c, A.,Altmann,M.,& viewingangleof1arcminforz=10. Herrera,D.2012,ApJ,744,110 Cioffi,D.F.,McKee,C.F.,&Bertschinger,E.1988,ApJ,334, 252 consistentwiththeThomsonscatteringopticaldepthin- Dewdney,P.E.,Hall,P.J.,Schilizzi,R.T.,&Lazio,T.J.L.W. dicated by Plank (2016). 2009,IEEEProceedings,97,1482 Based on the above parameters, we model LAEs and Dijkstra,M.,&Kramer,R.2012, MNRAS,424,1672 Hiibubblesusingindividualhalomergertrees. Ourmod- Fan,X.,Carilli,C.L.,&Keating,B.2006,ARA&A,44,415 els show the distribution function of the size of Hii bub- Finkelstein,S.L.,Papovich, C.,Dickinson,M.,Song,M.,Tilvi, V.,Koekemoer,A.M.,Finkelstein,K.D.,Mobasher,B., ble,andindicategiantHiibubblesassociatedwithbright Ferguson,H.C.,Giavalisco,M.,Reddy,N.,Ashby,M.L.N., LAEs at z .12 can be probed by future 21-cm observa- Dekel,A.,Fazio,G.G.,Fontana, A.,Grogin,N.A.,Huang, tion using SKA. We find that Lyα flux is tightly related J.-S.,Kocevski,D.,Rafelski,M.,Weiner,B.J.,&Willner,S.P. with stellar/halo mass, while there is a large dispersion 2013,Nature,502,524 Gawiser,E.,Francke, H.,Lai,K.,Schawinski,K.,Gronwall,C., in the relation between Lyα flux and SFR. By compar- Ciardullo,R.,Quadri,R.,Orsi,A.,Barrientos,L.F.,Blanc, ing our models with the observed luminosity function G.A.,Fazio,G.,Feldmeier,J.J.,Huang,J.-s.,Infante, L., of LAEs at z = 7.3 by Konno et al. (2014), we indi- Lira,P.,Padilla,N.,Taylor,E.N.,Treister,E.,Urry,C.M., cate that LAEs are likely to have the Hi column den- vanDokkum,P.G.,&Virani,S.N.2007,ApJ,671,278 sity of N & 1020 cm−2 and the outflowing gas with Gronwall,C.,Ciardullo,R.,Hickey,T.,Gawiser,E.,Feldmeier, HI V & 150 km s−1. Using these parameters, we pre- J.J.,vanDokkum,P.G.,Urry,C.M.,Herrera,D.,Lehmer, out B.D.,Infante,L.,Orsi,A.,Marchesini,D.,Blanc,G.A., dict that future wide deep survey can detect LAEs at Francke,H.,Lira,P.,&Treister,E.2007,ApJ,667,79 z 10 with nLAE a few 10−6 Mpc−3 and 1 Hamuy,M.2003,ApJ,582,905 10∼−4Mpc−3forthe∼fluxsensit×ivityof10−18ergcm∼−2s−×1 Harker,G.,Zaroubi,S.,Bernardi,G.,Brentjens,M.A.,de and 10−19 ergcm−2s−1, respectively. Our models pre- Bruyn,A.G.,Ciardi,B.,Jeli´c,V.,Koopmans,L.V.E., Labropoulos,P.,Mellema,G.,Offringa,A.,Pandey, V.N., dict next generation telescopes, JWST, E-ELT or TMT Pawlik,A.H.,Schaye, J.,Thomas,R.M.,&Yatawatta, S. would be able to observe galaxies at z 10 via the de- 2010,MNRAS,405,2492 ∼ tections of Lyα lines. Hasegawa,K.,&Semelin,B.2013, MNRAS,428,154 By combining future galaxy observations and 21-cm Hu,E.M.,&McMahon,R.G.1996,Nature,382,231 tomography with SKA-2, we will be able to detect both Iliev,I.T.,Mellema,G.,Pen,U.-L.,Merz,H.,Shapiro,P.R.,& LAEs with L & 1042 erg s−1 and their associated Alvarez,M.A.2006, MNRAS,369,1625 Lyα Iliev,I.T.,Mellema,G.,Shapiro,P.R.,Pen,U.-L.,Mao,Y., giantHii bubbles with RHII &250kpc. We suggestthat Koda,J.,&Ahn,K.2012,MNRAS,423,2222 a clear spatial anti-correlation between galaxies and 21- Inoue,Y.,Inoue,S.,Kobayashi,M.A.R.,Makiya,R.,Niino,Y., cm emission will be detected by focusing on such bright &Totani,T.2013,ApJ,768,197 LAEs. Iye,M.,Ota,K.,Kashikawa,N.,Furusawa,H.,Hashimoto,T., Hattori,T.,Matsuda,Y.,Morokuma,T.,Ouchi,M.,& Shimasaku,K.2006,Nature,443,186 ACKNOWLEDGMENTS Jeon,M.,Pawlik,A.H.,Bromm,V.,&Milosavljevi´c,M.2014, We thank Akira Konno and Masami Ouchi for pro- MNRAS,440,3778 viding us their recent observational data. We are Kashikawa,N.,Shimasaku,K.,Malkan,M.A.,Doi,M.,Matsuda, Y.,Ouchi,M.,Taniguchi,Y.,Ly,C.,Nagao,T.,Iye,M., grateful to Masayuki Umemura, Akio Inoue, Yoshiaki Motohara,K.,Murayama,T.,Murozono,K.,Nariai,K.,Ohta, Ono, Takatoshi Shibuya, Daisuke Nakauchi and Sadegh K.,Okamura,S.,Sasaki,T.,Shioya,Y.,&Umemura,M.2006, Khochfar for valuable discussion and comments. The ApJ,648,7 10