First Identification of 10-kpc Scale [CII] 158um Halos around Star-Forming Galaxies at z=5-7
Seiji Fujimoto, Masami Ouchi, Andrea Ferrara, Andrea Pallottini, R. J. Ivison, Christoph Behrens, Simona Gallerani, Shohei Arata, Hidenobu Yajima, Ken Nagamine
aa r X i v : . [ a s t r o - ph . GA ] S e p ApJ in press
Preprint typeset using L A TEX style emulateapj v. 12/16/11
FIRST IDENTIFICATION OF 10-kpc [C ii ] 158 µ m HALOSAROUND STAR-FORMING GALAXIES AT z = 5 − Seiji Fujimoto , Masami Ouchi , Andrea Ferrara , Andrea Pallottini , R. J. Ivison ,Christoph Behrens , Simona Gallerani , Shohei Arata , Hidenobu Yajima , and Kentaro Nagamine ApJ in press
ABSTRACTWe report the discovery of 10-kpc [Cii] µ m halos surrounding star-forming galaxies in the earlyUniverse. We choose deep ALMA data of 18 galaxies each with a star-formation rate of ≃ − M ⊙ with no signature of AGN whose [Cii] lines are individually detected at z = 5 . − . [Cii] lines and dust-continuum in the uv -visibility plane. The radial profiles of thesurface brightnesses show a 10-kpc scale [Cii] halo at the 9.2 σ level, significantly more extended thanthe HST stellar continuum data by a factor of ∼ [Cii] and Ly α halos universally found in star-forminggalaxies at this epoch, and find that the scale lengths agree within 1 σ level. While two independenthydrodynamical zoom-in simulations match the dust and stellar continuum properties, the simulationscannot reproduce the extended [C ii ] line emission. The existence of the extended [Cii] halo is theevidence of outflow remnants in the early galaxies and suggest that the outflows may be dominatedby cold-mode outflows expelling the neutral gas. Keywords: galaxies: formation — galaxies: evolution — galaxies: high-redshift INTRODUCTION
Galaxy size and morphological studies in the early Uni-verse provide important insights into the initial stage ofgalaxy formation and evolution. The size and morphol-ogy in the rest-frame ultra-violet (UV) and far-infrared(FIR) wavelengths trace the areas of young star forma-tion and the active starbursts that are less and heavilyobscured by dust, respectively. The [Cii] P / → P / fine-structure transition at 1900.5469 GHz (157.74 µ m) isa dominant coolant of the inter-stellar medium (ISM) ingalaxies (e.g., Stacey et al. 1991; De Looze et al. 2014),whose size and morphology are strong probes of ISMproperties. Comparing the size and morphology in therest-frame UV+FIR continuum and the [Cii] µ mline is thus important to comprehensively understand the [email protected] Institute for Cosmic Ray Research, The University of Tokyo,Kashiwa, Chiba 277-8582, Japan Research Institute for Science and Engineering, Waseda Uni-versity, 3-4-1 Okubo, Shinjuku, Tokyo 169-8555, Japan National Astronomical Observatory of Japan, 2-21-1, Osawa,Mitaka, Tokyo, Japan Kavli Institute for the Physics and Mathematics of the Uni-verse (WPI), University of Tokyo, Kashiwa 277-8583, Japan Scuola Normale Superiore, Piazza dei Cavalieri 7, I-56126Pisa, Italy Centro Fermi, Museo Storico della Fisica e Centro Studi eRicerche ”Enrico Fermi” Piazza del Viminale 1, Roma, 00184,Italy European Southern Observatory, Karl Schwarzschild Str. 2,D-85748 Garching, Germany Institute for Astronomy, University of Edinburgh, Royal Ob-servatory Blackford Hill, Edinburgh EH9 3HJ, UK Theoretical Astrophysics, Department of Earth and SpaceScience, Graduate School of Science, Osaka University, Toyon-aka, Osaka 560-0043, Japan Center of Computational Sciences University of Tsukuba,Ibaraki 305-8577, Japan Department of Physics & Astronomy, University of Nevada,Las Vegas, 4505 S. Maryland Pkwy, Las Vegas, NV 89154-4002,USA evolutionary process via the star-formation surroundedby the ISM.The
Hubble Space Telescope (HST) has revealed the sizeand morphological properties in the rest-frame UV wave-lengths for the high-redshift galaxies up to z ∼
10 (e.g.,Oesch et al. 2010; Ono et al. 2013; Shibuya et al. 2015;Bouwens et al. 2017; Kawamata et al. 2018). These HSTstudies show that star-forming galaxies generally havean exponential-disk profile and become compact towardhigh redshifts.The Atacama Large Millimeter / submillimeterArray (ALMA) has opened our views to the ob-scured star-formation and the [Cii] line propertiesin the rest-frame FIR wavelength up to z ∼ [Cii] line for suchhigh redshift galaxies at z ∼ −
7, where Carniani et al.(2017) report that the effective radius of the [Cii] line-emitting region is larger than that of the rest-frameUV region. However, large uncertainties still remaindue to the small number statistics and observationalchallenges.One critical challenge is sensitivity. The recent ALMAstudies show that signal-to-noise ratio (S/N) >
10 isneeded to obtain reliable size measurement results bothon the image-based and visibility-based analyses (e.g.,Simpson et al. 2015; Ikarashi et al. 2015), while the ma-jority of the previous ALMA detections of the dust con-tinuum and the [Cii] line from z ∼ − Fujimoto et al. fluctuations can make the morphology more clumpy. Toobtain the reliable size and morphological results, exten-sively deep observations are thus required.In this paper, we determine the size and morphologyfor the dust continuum and the [Cii] line in the star-forming galaxies at z = 5 − uv -visibility plane, utilizing new and archival deepALMA Band 6/7 data. In conjunction with deep HSTimages, we study the general morphology of the totalstar-formation and the ISM in the epoch of re-ionization.The structure of this paper is as follows. In Section 2,the observations and the data reduction are described.Section 3 outlines the method of [Cii] line detections, linevelocity width, source position measurements, and thestacking processes of ALMA and HST data. We reportthe results of the radial profiles of the [Cii] line, rest-frame FIR, and rest-frame UV wavelengths in Section4. In Section 5, we discuss the physical origin of theextended [Cii] line emission, comparing with the zoom-in cosmological simulation results. A summary of thisstudy is presented in Section 6.Throughout this paper, we assume a flat universe withΩ m = 0 .
3, Ω Λ = 0 . σ = 0 .
8, and H = 70 kms − Mpc − . We use magnitudes in the AB system(Oke & Gunn 1983). SAMPLE AND DATA REDUCTION
Our ALMA Sample
The sample is drawn mainly from the literature(Capak et al. 2015; Willott et al. 2015; Pentericci et al.2016; Smit et al. 2018; Carniani et al. 2018; Jones et al.2017), selecting only star-forming galaxies at z > ii ] lines have been detected (at signal-to-noise, S/N &
5) with ALMA. To obtain reliable results for rep-resentative galaxies in the early Universe, we limit oursample to galaxies with (i) star-formation rates (SFRs), <
100 M ⊙ yr − , (ii) no indication of AGN activity, (iii)no giant Ly α systems, such as Himiko (Ouchi et al. 2009)and CR7 (Matthee et al. 2015), (iv) no signs of gravi-tationally lensing, e.g. galaxies behind massive galaxyclusters, (v) [C ii ] line emission with a full width at halfmaximum (FWHM) broader than 80 km s − , and (vi)[C ii ] line detections that are reproduced in our own datareduction. We adopt (v) because the thermal noise fluc-tuation can produce peaky false source signals even withS/N >
5, when we examine the large volume data suchas the ALMA 3D data cubes. Note that our sample doesnot include the tentative [C ii ] line detections reportedin the ALMA blind line survey (Aravena et al. 2016;Hayatsu et al. 2017), because these tentative [C ii ] de-tections have not been spectroscopically confirmed. Weidentify 16 [C ii ] line sources that meet the above criteriain the literature. Table 1 summarises our sample andthe references that describe the relevant ALMA observa-tions.In addition to the literature sample, we include new[C ii ] line detections of two star-forming galaxies, NB816-S-61269 (Ouchi et al. 2008; Fujimoto et al. 2016) andWMH13 (Willott et al. 2013a) at z = 5 .
688 and 5 . ii ] detec-tions. In the velocity-integrated maps of NB816-S-61269and WMH13, the [C ii ] line is detected with peak S/N levels of 5.6 and 5.2, respectively; rest-frame FIR dustcontinuum emission is not detected from either galaxy.The details of the ALMA observations for these addi-tional sources are listed in Table 1. !" *+%,-./ !" %&%’+45666%)%757’%8 !" %&%’*’57(+%)%7576%8 ! " & ’ & %$(%) * + , $- Figure 1.
New [Cii] line detections of NB816-S-61269 (top) andWMH13 (botttom).
Left:
Natural-weighted 4 ′′ × ′′ field imageof the velocity-integrated [C ii ] line intensity (moment zero) withcontours at the − σ (white), 2 σ , 3 σ , 4 σ , and 5 σ (red) levels. TheALMA synthesized beams are presented at the bottom left. Right: [C ii ] line spectra with an aperture diameter of 1 . ′′
2. The solidcurves denote the best-fit profile of the single Gaussian with thebest-fit values of the FWHM and the frequency peak. The yellowshades present the integrated velocity ranges for the [C ii ] line in-tensity maps in the left panel. The velocities are relative to the Ly α line obtained in previous studies (Ouchi et al. 2010; Willott et al.2013b). From the literature and the additional samples, we ob-tain a total of 18 [Cii] line sources. The 18 [Cii] linesources have the spectroscopic redshifts determined bythe [Cii] lines ( z [CII] ) and the absolute rest-frame UVmagnitudes ( M UV ) in the ranges of z [CII] = 5 . − . M UV ≃ − − ≃ − M ⊙ /yr). Wesummarize the physical properties of z [CII] , M UV , and theLy α equivalent-width (EW Ly α ) in Table 1. ALMA Data
We reduce the ALMA data for our sample withthe Common Astronomy Software Applications package(CASA; McMullin et al. 2007) in the standard mannerwith the scripts provided by the ALMA observatory. Inthis process, we carry out re-calibrations for the flux den-sity and additional flagging for bad antennae if we findproblems in the final images that shows striped patternsand/or significantly higher noise levels than expected.The continuum images and line cubes are produced bythe CLEAN algorithm with the tclean task with a pixelscale of 0 . ′′
01. For the line cubes, the velocity channelwidth is re-binned to 20 km s − , where the velocity cen-ter is adjusted to the Ly α redshift. We do not CLEANthe line cubes because the [Cii] line is faint in each 20-km s − channel. The CLEAN boxes were set at the peakpixel positions with S/N ≥ σ level.We list the standard deviation of the pixel values in a fi- CII] Halo at z ∼ Table 1
Our ALMA SampleTarget R.A Dec. z [CII] ( z Ly α ) M UV EW Ly α σ cont . † Beam ALMA ID HST Ref.(J2000) (J2000) (mag) (˚A) ( µ Jy/beam) ( ′′ × ′′ )(1) (2) (3) (4) (5) (6) (7) (8)LiteratureWMH5 36.612542 − − × − − × − − × − × − − × − − × †† C18COS13679 150.099014 2.343517 7.142 (7.145) − × ‡ P16COS24108 150.197356 2.478931 6.623 (6.629) − × ‡ P16Hz1 149.971828 2.118142 5.689 (5.690) − × − × ‡ C15 (B17)Hz3 150.039247 2.3371611 5.542 (5.546) − − × − × − × − × − × − × − S − − − × − × [Cii] (Ly α ) line emission. (2) Absolute magnitudes. (3) Rest-frame Ly α EW. (4)One sigma noise measured by the standard deviation of the pixel values in the continuum map before primary beam correction. (5)Synthesized beam size of our ALMA maps (weighting = ”natural”). (6) ALMA project ID. (7) ”Y” (”N”) indicates the sources (not)included in the ALMA-HST sample. (8) ALMA (HST) data reference (W15: Willott et al. 2015, J17: Jones et al. 2017, S18: Smit et al.2018, P16: Pentericci et al. 2016, C18: Carniani et al. 2018, S15: Smit et al. 2015, C15: Capak et al. 2015, B17: Barisic et al. 2017, F16:Fujimoto et al. 2016). † Our additional flagging and difference in the imaging parameter setting may produce different values from the data references. †† Although there is the F105W data, we do not include this source in the ALMA-HST sample due to the differences in the PSF and therest-frame wavelength from the F160W data. ‡ Although there is the F160W data, we do not include these sources in the ALMA-HST sample due to the large offsets even after theastrometry correction (see text). nal natural-weighted image and a synthesized beam sizefor the continuum image in Table 1.Note that the continuum is subtracted from the uv -data of the line cubes for 4 sources (Hz4, Hz6, Hz9, andWMH5) whose continuum emission is individually de-tected (Capak et al. 2015; Willott et al. 2015). The con-tinuum level is estimated from the channels in the veloc-ity range of v > | × FWHM | in the same baseband asthe [C ii ] line emission. HST Data
To study the rest-frame UV properties of our sam-ple, we also use the HST Wide Field Camera 3 (WFC3)in F160W, 1.54 µ m ( H -band), images from the HubbleLegacy Archive, where we obtain final flat-field and flux-calibrated science products.To correct the potential offsets of the HST astrome-try (e.g., Rujopakarn et al. 2016; Dunlop et al. 2017), wecalibrate the astrometry of the H -band maps with theGaia Data Release 2 catalog (Gaia Collaboration et al.2018). First, we identify bright objects in the H -band im-ages with sextractor version 2.5.0 (Bertin & Arnouts1996). Second, we cross-match the bright H -band ob-jects and the GAIA catalog. Finally, we evaluate offsetsbetween the bright H -band object centers and the GAIAcatalog positions. We find that the bright H -band objectcenters indeed have the offsets from the GAIA catalog inthe range of ∼ . ′′ − . ′′
3. We correct the astrometry of each H -band map to match the GAIA catalog based onthese offsets. With the above procedure, the majorityof our sample shows that the [C ii ] line and the H -bandcontinuum have a consistent peak position within the off-set smaller than ∼ . ′′
1. However, the large offset over0 . ′′ ii ] line and therest-frame UV continuum (e.g., Maiolino et al. 2015). Inany cases, these objects with the large offsets cause thesmearing effect in the stacking results. To securely studythe morphological property from the stacking results, wedo not include these objects in the following HST dataanalyses. We identify that 9 out of 18 sources in oursample have been observed with the HST/ H -band whoseastrometry is successfully corrected. We refer to the 9and the 18 sources as the ”ALMA-HST” and ”ALMA-ALL” samples, respectively. In Table 1, we summarizethe HST data references and the sources included in theALMA-HST sample.Note that we confirm that the ALMA astrometry iswell consistent with the GAIA catalog within a milli-arcsec scale via the bright quasars used as the phasecalibrators in the ALMA observations. Thus, we donot carry out any astrometry corrections for our ALMAmaps. DATA ANALYSIS
Fujimoto et al. !"
Figure 2. uv -visibility coverage for individual and stacked data.For the individual data, we present Hz3 data as an example. Forthe stacked data, the uv -visibility coverage less than 500 k λ is wellsampled in circular symmetrically, which enables us to investigatethe diffuse, extended structures.
3D Position in ALMA Cube
To carry out stacking for the [C ii ] line and the rest-frame FIR continuum, we estimate source centroids forthe 18 [C ii ] line sources in the ALMA 3-dimensional datacubes via the following six steps: (1) We create fidu-cial [C ii ] velocity-integrated maps in the velocity range, ∼ − , that maximizes the S/N level of the[C ii ] line detection. (2) We measure fiducial positionalcentroids based on the peak pixel positions (pixel scale= 0 . ′′
01) in the fiducial [C ii ] velocity-integrated maps,having smoothed spatially with a uv -taper of 0 . ′′
6. (3)We produce [C ii ] spectra with an aperture diameter of1 . ′′ ii ] line emissionby fitting a single Gaussian to the [C ii ] spectra. (5) Were-create velocity-integrated maps with velocity rangesof 2 × the FWHM. (6) We measure final positional cen-troids in the new velocity-integrated map in the samemanner as step (2). Note that we use the smoothed map(via the uv -taper) instead of the naturally-weighted mapin steps (2) and (6) because Monte-Carlo simulations inthe uv -visibility plane show that smoothed maps havelower uncertainties in the positional measurements thanthe intrinsic maps (Fujimoto et al. 2018). We list thefinal positional centroids and redshifts in Table 1. ALMA Visibility-based Stacking
We carry out visibility-based stacking for our ALMAdata via the following procedure. First, we split the visi-bility data into the [C ii ] line and the rest-frame FIR con-tinuum datasets. For the [C ii ] line dataset, we extractthe visibility data with the [C ii ] line channels across avelocity range of 100 km s − (= ±
50 km s − ), where thevelocity center is the [C ii ] frequency peak (the 3D po- sition in our ALMA cubes). We do not adopt a widervelocity range because of the potential contamination ofthe close companions (Jones et al. 2017; Carniani et al.2018). For the rest-frame FIR continuum dataset, weproduce the visibility data whose [C ii ] line channels ina velocity range of 2 × FWHM are fully removed. Sec-ond, we shift the coordinate of the visibility datasets byre-writing the source center determined in Section 3.1 as”00:00:00.00 00:00:00.0” with stacker (Lindroos et al.2015). Third, we combine the visibility datasets with the concat task. Fourth, we re-calculate the data weightsfor the combined visibility datasets with the statwt task, based on the scatter of visibilities, which includesthe effects of integration time, channel width and systemtemperature. Finally, we obtain the stacked datasets ofthe [C ii ] line and the rest-frame FIR continuum. Thecentral frequency in the [C ii ] line dataset is 271.167 GHzwhich corresponds to the [C ii ] redshift at z = 6 .
01. As-suming the redshift of z = 6 .
01 as the weighted averagesource redshift of our sample, we adopt the angular scaleof 1 ′′ = 5 . H -band peak positions (Section 3.3) as thecommon stacking center for the ALMA-HST sample.Figure 2 indicates the uv -visibility coverage after thevisibility-based stacking for the ALMA datasets of theALMA-ALL sample. For comparison, the uv -visibilitycoverage for an individual dataset, before stacking, isalso plotted. In the stacked data, the uv -visibility cov-erage is well sampled, especially for the short baselines, <
500 k λ , which is important to recover the flux densityfrom diffuse, extended structures.In Figure 3, we show the natural-weighted images ofthe [Cii] line and dust continuum after the visibility-based stacking for the ALMA-ALL (ALMA-HST) sam-ple, where the standard deviation of the pixel values inthe dust continuum image achieves 4.1 (8.3) µ Jy/beamwith the synthesized beam size of 0 . ′′ × . ′′
36 (0 . ′′ × . σ (20 σ ) and 10 σ (8 σ ) significance levels for the [Cii] line and dust continuum, respectively, for the ALMA-ALL (ALMA-HST) sample. The spatially resolved [C ii ]line emission in the ALMA-ALL sample is detected atthe 9.3 σ level in the aperture radius of 10 kpc even aftermasking the emission in a central area up to 2 × FWHMof the ALMA synthesized beam, based on the random-aperture method. Because the extended structure is dif-ficult to be modeled by the clean algorithm perfectly,we use the dirty images for both the [C ii ] line and therest-frame FIR continuum in the following analyses.In Figure 4, we present the radial surface brightnessprofile of the stacked [C ii ] line, and summarize varioustests for the extended [C ii ] line structure. First, we com-pare our stacking and individual results. In the top leftpanel of Figure 4, we show the individual results for sev-eral [C ii ] line sources whose lines are detected at highS/N, with an ALMA beam size of & . ′′ ii ] line sources.Second, we evaluate the uncertainty of the sample vari-ance. We make 18 newly stacked data with 17 [C ii ] linesources, i.e., in each newly stacked data we remove onesource from the full sample, and derive the 18 [C ii ] ra- CII] Halo at z ∼ !" % &’())* ’())* Figure 3.
Natural-weighted 4 ′′ × [Cii] line and the dust continuum for the ALMA-ALL (left) and ALMA-HST (right) samples. The red and green contours denote the 2, 2 √
2, 4, ... × σ levels of the [Cii] line and the dustcontinuum emission, while the white contours indicate the − σ and − √ σ levels. The synthesized beams are presented at the bottomleft in each panel. dial profiles. In the top right panel of Figure 4, the redshaded area indicates the 16–84 percentiles of these 18 ra-dial profiles. The [C ii ] radial profile is extended up to theradius of ∼
10 kpc even including the sample variance,suggesting that the sample variance does not change ourresults of the existence of the extended [C ii ] line emis-sion. Third, we investigate whether the extended [C ii ]line structure is caused by any specific data properties.We remove the sources that are I) taken with the lowestresolutions (BDF2203, NTTDF6345, and WMH13), andII) reported to have companions (WMH5, Hz2, Hz6, andHz8; Jones et al. 2017; Carniani et al. 2018), and obtainnewly stacked data. In the top right panel of Figure4, we present the radial profiles of the [C ii ] line emis-sion in the newly stacked data. We find that the newlystacked [C ii ] line profiles reproduce the extended struc-tures that are well consistent with the original stackingresult in the ALMA-ALL sample. This indicates thatthe extended [C ii ] line structure is not caused either bythe bias to the low-resolution data or the contamina-tion of the companions. Fourth, we examine the surfacebrightness dimming effect among our sample. We di-vide the 18 [C ii ] line sources into two subsamples: low( z <
6) and high-redshift ( z >
6) samples, and obtainother newly stacked data. In the bottom left panel ofFigure 4, we show the radial profiles of the [C ii ] lineemission in both subsamples. We find that the [C ii ]line profiles in both subsamples reproduce the extendedstructures that have good agreements with the originalstacking result in the ALMA-ALL sample. This suggeststhat the surface brightness dimming effect does not sig-nificantly affect our stacking results. Fifth, we comparethe structures of the [C ii ] line and the dust continuumin the same significance level. We produce a randomnoise map smoothed by the ALMA beam, and combinethe noise and the stacked [C ii ] line maps. Changing thenoise levels, we obtain the noise-enhanced [C ii ] line mapwhose peak S/N ratio becomes comparable to the dustcontinuum one. We create the 50 noise-enhanced [C ii ]line maps. In the bottom right panel of Figure 4, weshow the 16–84th percentile of the [C ii ] radial profile in the noise-enhanced maps. We find that the [C ii ] line pro-file still exceeds more than the dust continuum in thesenoise-enhanced maps, showing that the different struc-tures between the [C ii ] line and the dust continuum arenot mimicked by the difference in the dynamic range. HST/H-band Stacking
We have performed image-based stacking for theALMA-HST sample, exploiting their deep archival HST H -band imaging. Before stacking, we carry out the fol-lowing procedure: 1) We cut out 8 ′′ × ′′ stamps fromthe H -band images, around the [C ii ] line sources, andset the pixel scale to 0 . ′′
01, which corresponds to ourALMA images. 2) We identify low-redshift contaminantswithin 2 . ′′ ii ] sources, by cross-matching the[C ii ] line source positions with photometric redshift cat-alogs (Ilbert et al. 2013; Skelton et al. 2014). 3) We re-move the low-redshift contaminants from the H -band im-ages by fitting S´ersic profiles (S´ersic 1963) with galfit (Peng et al. 2010). We then proceed to generate an av-erage stack, weighted by the noise levels of the ALMAimages of the [C ii ] line source. This is because thevisibility-based stacking for our ALMA data is weightedby the visibility scatter, which generally corresponded tothe noise levels on the ALMA images.In panel (f) of Figure 5, we show the stacked H -bandimage for the ALMA-HST sample. In the HST stack-ing, we adopted stacking centroids defined by the peakpositions of the H -band images, smoothed with the uv -tapered ALMA beams in a consistent manner with the[C ii ] line stacking.To directly compare the size and morphology of theHST and ALMA images, we need to convolve the HSTimage to obtain a PSF that resembles the one of theALMA image. We use galfit to obtain a kernel withwhich the H -band PSF can be converted to the ALMAbeam. For the kernel, we assume a sum of three inde-pendent S´ersic profiles whose positions are fixed at thecenter.In Figure 5, we present a schematic overview of con-verting the H -band PSF to the ALMA beam with the Fujimoto et al. !"
Figure 4.
Radial surface brightness profile of the [C ii ] line (red filled squares) and dust-continuum (green filled squares) emission forthe ALMA-ALL sample. Top Left:
The black solid curves denote the individual results from five [C ii ] line sources whose [C ii ] linesare detected with high S/N levels and with the ALMA beam sizes of & . ′′ Top Right:
The red shade shows the 16–84 percentile of the sample variance (see text). The open symbols indicate the re-stacked resultswithout the [C ii ] line sources that are I) taken with the lowest resolutions (BDF2203, NTTDF6345, and WMH13; upward triangle), and II)reported to have companions (WMH5, Hz2, Hz6, and Hz8; downward triangle). Bottom Left:
The re-stacked results for the low- ( z < z >
6; rightward triangle) redshift subsamples among the 18 [C ii ] line sources. Bottom Right: whose peakS/N ratio is reduced down to the level comparable to the dust continuum map. All radial profiles are normalized to the peak value of the[C ii ] line. best-fit kernel. We first convolve the HST PSF (panela) with the best-fit kernel (panel b) and derive the mockALMA beam (panel c). We then subtract the actualALMA beam (panel d) from the mock ALMA beam andproduce the residual map (panel e). Within a radiusof 1 . ′′ ∼ . H -band PSF.We finally apply the convolution to the stacked H -band image (panel f) with the best-fit kernel, and obtain themock H -band image whose PSF is almost the same asthe stacked ALMA image. RESULTS
Discovery of [Cii]
Halo
Figure 6 presents the radial surface brightness profilesof the [Cii] line, rest-frame FIR, and UV continuum,derived from the stacking results for the ALMA-HST(circles) and ALMA-ALL (squares) samples. For a fair
CII] Halo at z ∼ Figure 5.
Schematic overview to obtain the mock HST/H-band image whose spatial resolution is matched to the stacked ALMA imagefor the ALMA-HST sample: a) HST/H − band PSF, b) the best-fit kernel composed by three S´ersic profiles obtained with galfit , c) thebest-fit ALMA beam model obtained with galfit , d) the synthesized beam in the stacked ALMA image for the ALMA-HST sample, e)the residual between c) and d), f) the stacked HST/H − band image for the ALMA-HST sample, and g) the stacked HST/H − band imageobtained by convolving f) with b). The red contours present 3%, 5%, 10%, 20%, 30%, 40%, and 50% of the PSF or beam response. Theblue contours denote the 2, 2 √
2, 4, ... × σ levels of the rest-frame UV continuum emission. The cutout sizes are 2 ′′ × ′′ and 4 ′′ × ′′ forthe panels of a) − e) and f) − g), respectively. comparison, the ALMA-HST results are obtained by re-performing the ALMA visibility-based stacking with theHST/ H -band peak positions, while the ALMA-ALL re-sults are not due to the lack of the HST/ H -band images.In Figure 6, the ALMA-ALL and ALMA-HST resultsshow a good agreement in both profiles of the [C ii ] lineand the rest-FIR continuum. We find that the radialprofile of the [C ii ] line emission is extended up to a radiusof ∼
10 kpc which contrasts the rest-frame UV and FIRcontinuum. Because the typical effective radius of thenormal star-forming galaxies at z ∼ ∼ [Cii] line emission is producedin the wide CGM areas even without stellar continuum.We discuss the physical origin of the [Cii] halo in Section5. We also find that the profiles of the rest-frame FIRand UV continuum are consistent within the 1 σ errors.Note that the rest-frame FIR continuum is likely to followthe ALMA beam, while the rest-frame UV continuum is Fujimoto et al. !" !"
Figure 6.
Radial surface brightness profiles for the ALMA-HST (circles) and ALMA-ALL (squares) samples. The radial values areestimated by the median of each annulus. The red, green, and blue symbols denote the [C ii ] line, rest-frame FIR, and rest-frame UVcontinuum emission. The rest-frame UV continuum profile is directly derived from the mock HST/ H -band image whose resolution ismatched to that of the ALMA image. The black dashed and solid curves denote the ALMA synthesized beams in the stacked images ofthe ALMA-HST and ALMA-ALL samples, respectively. All radial profiles are normalized to the peak value of the [Cii] line. The greenand red symbols are slightly shifted along the x -axis for clarity. slightly resolved with the ALMA beam. This suggeststhat the intrinsic size of the rest-frame FIR continuumis smaller than that of the rest-frame UV continuum,which is consistent with the recent ALMA results of thecompact rest-frame FIR size more than the rest-frameUV and optical sizes among the star-forming galaxies at z ∼ − Effect of [C ii ]-UV offset Recent studies report a possibility that [Cii] -line emit-ting regions are physically offset from the rest-frame UVones (e.g., Maiolino et al. 2015). To evaluate the poten-tial effect from the [Cii] -UV offsets in our results, weperform the ALMA and HST stacking for the ALMA-HST sample by adopting two different stacking centers:HST/ H -band and ALMA [C ii ] line peak positions, andcompare the radial profiles from these stacking results.In Figure 7, the circle and cross symbols represent the stacking results derived with the common stacking cen-ters of the HST/ H -band continuum and ALMA [C ii ] linepeak positions, respectively. We find that the [C ii ] lineprofile is extended more than both the rest-frame FIRand UV continuum profiles in any cases. This suggeststhat the [C ii ] line originates from much wider regionsthan the continuum emission at rest-frame FIR and UVwavelengths, and clearly shows that the extended struc-ture of the [C ii ] line is not caused by the [Cii] –UV off-sets. Radial ratio of L [CII] to total SFR To test whether the extended [C ii ] line structure iscaused by satellite galaxies, we investigate radial valuesof the [C ii ] line luminosity L [CII] at a given SFR derivedfrom the rest-frame FIR and UV continuum. Because theALMA-ALL and ALMA-HST results are consistent witheach other (Figure 6), we adopt the rest-frame UV resultsfrom the ALMA-HST sample, while the [C ii ] line and CII] Halo at z ∼ !" !" Figure 7.
Radial surface brightness profiles for the ALMA-HSTsample derived with different stacking centers. The red, green,and blue symbols denote the [C ii ] line, rest-frame FIR, and rest-frame UV continuum emission. The color crosses and circles are thestacking results based on the stacking centers of the [C ii ] line andthe HST/ H -band peak positions, respectively. The rest-frame UVcontinuum profile is directly derived from the mock HST/ H -bandimage whose resolution is matched to the ALMA image. The blackdashed curve denotes the ALMA synthesized beam. All radialprofiles are normalized to the peak value of the [C ii ] line. The greenand red symbols are slightly shifted along the x -axis for clarity. rest-frame FIR continuum results from the ALMA-ALLsample to reduce the errors in the following estimates.We first estimate the radial L [CII] value. Forour sources, the weighted-average source redshift andFWHM of the [C ii ] line width are estimated to be z = 6 .
01 and 270 km s − , respectively. Since the velocity-integrated width is 100 km s − in the stacked [C ii ] linemap, we correct the velocity-integrated value in the rangefrom 100 km s − to 270 km s − , assuming a single Gaus-sian line profile, to recover the total value of L [CII] . Wesecond evaluate the radial SFR value. We derive the ob-scured (SFR IR ), un-obscured (SFR UV ), and total SFR(SFR total ) with the equations in Murphy et al. (2011) ofSFR IR [ M ⊙ yr − ] = 3 . × − L IR [erg s − ] , (1)SFR UV [ M ⊙ yr − ] = 4 . × − L UV [erg s − ] , (2)SFR total = SFR IR + SFR UV , (3)where L IR is the integrated IR flux density esti-mated by a typical modified blackbody whose spec-tral index β d and dust temperature T d are β d =1 . T d = 35 K(Coppin et al. 2008), and L UV is the rest-frame UV lu-minosity at 0.16 µ m with the HST H -band. Finally, wedivide the radial L [CII] values by the radial SFR valuesand obtain the radial ratio of L [CII] /SFR total .In Figure 8, we show the surface densities of L [CII] (Σ L [CII] ) and SFR total (Σ SFR ) as a function of radius(middle panel), and the radial ratio of L [CII] /SFR total (right panel) as a function of SFR total for our stackingresults. For comparison, the right panel of Figure 8 alsopresents global scale L [CII] /SFR total ratios of the localdwarf galaxies (De Looze et al. 2014).In the right panel of Figure 8, the red filled and opencircles denote our stacking results in the outer (radius of ≥ < L [CII] /SFR total decreases with SFR total . Thehighest ratios ( > L ⊙ M − ⊙ yr) are found at the outerregions and not compatible with typical values found inthe local dwarf galaxies ( < × L ⊙ M − ⊙ yr; black dotsin the figure). These results indicate that the [C ii ] halois not likely driven by satellite galaxies. Note that othertypes of high- z galaxies with SFR total > M ⊙ suchas star-forming, submillimeter, and quasar-host galaxiesshow ratios of 10 ∼ L ⊙ M − ⊙ yr (e.g., Capak et al.2015; Rybak et al. 2019; Venemans et al. 2019) which areyet difficult to explain the highest ratios in our stackingresults ( > L ⊙ M − ⊙ yr).D´ıaz-Santos et al. (2014) report the [C ii ] line emis-sion extended over ∼ −
10 kpc scale around localluminous infrared galaxies (LIRGs), and we also com-pare our results with this spatially resolved data. Inthe right panel of Figure 8, we show the L [CII] /SFR total ratios of the extended emission around local LIRGs.We find that the highest ratios in our stacking results( > L ⊙ M − ⊙ yr) are still higher than those around thelocal LIRGs ( < × L ⊙ M − ⊙ yr; green crosses in thefigure). The [C ii ] halo at z ∼ ii ] line emission observed in the localUniverse, which may suggest that [C ii ] halos evolve withredshift. We discuss possible origins of the [C ii ] halo inSection 5. Scale Length of [Cii]
Halo
We characterize the detail radial surface brightnessprofile of the [Cii] line emission by two-component fit-ting with galfit . Here we assume the two componentsas the central and the halo components.For the central component, we adopt the S´ersic profilewhose parameters are estimated from the rest-frame UVprofile in the stacked HST/ H -band image (Figure 5 f).We obtain the best-fit effective radius r e and the S´ersicindex n of r e = 1 . ± . n = 1 . ± .
01 that areconsistent with the average values estimated from thenormal star-forming galaxies at z ∼ α halo whichis universally identified around the high-z star-forminggalaxies (e.g., Steidel et al. 2011; Matsuda et al. 2012;Momose et al. 2014, 2016; Leclercq et al. 2017). The ex-ponential profile is described as C n exp( − r/r n ) where C n is a constant and r n is the scale length. We fix the centralpositions of both central and halo components to obtaina stable result.Top panel of Figure 9 presents the best-fit results withthe S´ersic+exponential profiles for the [Cii] line emis-sion. We obtain the best-fit scale-length values of r n =3.3 ± r e = 5 . ± ii ] halois extended ∼ Fujimoto et al.
10 kpc[CII] 158umrest-FIR
SFR total
SFR IR SFR UV L [CII] !" !" z ~ 6 star-forming Figure 8. Left:
Rest-frame UV emission of the ALMA-HST sample in the HST/ H -band 4 ′′ × ′′ image whose resolution is matched to theALMA image. The red and green contours denote the 2, 2 √
2, 4, ... × σ levels of the [Cii] line and the dust continuum emission, respectively.The ALMA synthesized beam is presented at the bottom left. Middle:
Radial profiles of Σ
SFR (left axis) and Σ L [CII] (right axis). Theblue, green, and black circles indicate Σ SFR UV , Σ SFR IR , and Σ SFR total , respectively, based on Equations (1)–(3). The red circles denoteΣ L [CII] normalized to Σ SFR with the ratio of L [CII] /SFR total = 10 [ L ⊙ M − ⊙ yr] that is the average value in the local star-forming galaxies(De Looze et al. 2014). Right:
Ratio of L [CII] /SFR total as a function of SFR total . The filled (open) red circles indicate our stackingresults at a radius of ≥ < ii ] line emission calculated from the local LIRG results in (D´ıaz-Santos et al. 2014). Weassume the area of 1 kpc for the L [CII] and SFR total estimates in our stacking and the local LIRGs results. At the radius of > total value in our stacking results becomes negative due to the noise fluctuations on the low surface brightness of the rest-frame UVand FIR continuum emission, where we evaluate the lower limit of the ratio by using the upper limit of SFR total . the central galactic component. Note that the visibility-based profile fitting with uvmultifit (Mart´ı-Vidal et al.2014) also provides us with the best-fit value of r e = 5 . ± galfit result within the error.In Figure 9, we compare the radial surface brightnessprofiles of the [Cii] with the Ly α halos universally identi-fied in the normal star-forming galaxies at z ∼ [Cii] line emission, we adopt the result from the ALMA-ALLsample due to the high significance detection. For theLy α line emission, we use the recent results with thedeep MUSE data for the high- z LAEs of Leclercq et al.(2017), where the authors estimate the best-fit radialsurface brightness profiles by fitting the two-componentS´ersic+exponential profile. We select the best-fit resultsof 6 LAEs at z > M UV . −
21 mag and EW Ly α <
100 ˚A that are consistent with the parameter spaceof our sample (Table 1). We find that the radial surfacebrightness profile of the [Cii] line emission is comparableto that of the Ly α line emission. The median r n value forthe 6 LAEs is estimated to be 3 . ± . . ± . [Cii] line emission is related to the Ly α halo.Note that we confirm that it is hard to reproduce theextended morphology of the [C ii ] line emission with thecentral component alone. In the bottom panel of Figure9, we present the residuals of the [C ii ] line emission ob-tained from the best-fit results of the one- (central) andtwo- (central+halo) component fittings with uvmulti-fit . We find that the residuals in the one-componentfitting result show a bump at a radius of ∼ ′′ over theerrors, while the residuals in the two-component fittingresult is broadly consistent with zero. This suggests thatthe extended morphology of the [C ii ] line emission con- sists of a combination of the central plus halo compo-nents. [C ii ] Stacked Spectrum We also perform the stacking for the [C ii ]-line spectraof the ALMA-ALL sample to test whether our ALMAsample has a broad wing feature which is a good probefor the on-going outflow activities. For the stacking pro-cedure, we adopt the same manner as previous ALMAstudies (Decarli et al. 2018; Bischetti et al. 2018). Herewe adopt a relatively small aperture diameter of 0 . ′′ ii ]-line spectrumwith the best-fit two Gaussian component model: thecombination of the core and broad components whosevelocity centers are fixed at 0 km/s for the stable results.The best-fit FWHMs are estimated to be 296 ±
40 km/sand 799 ±
654 km/s for the core and broad components,respectively. In the velocity range of ±
400 – 800 km/s,the velocity-integrated intensity is tentatively detected atthe 3.2 σ level. Moreover, Sugahara et al. (2019) have re-cently reported that the rest-frame UV metal absorptionlines are blue-shifted with the central outflow velocity of440 +110 − km s − from the [C ii ]-systemic redshift in thestacked Keck spectra whose stacking sample includes 6out of our 18 [C ii ] line sources. These results may sug-gest the existence of the tentative broad wing feature isproduced by the outflow.Note that there are other possibilities that produce thebroad wing feature. One possibility is that the contam-ination of the satellite galaxies. The [C ii ] line emissionfrom the individual satellite galaxies can be smoothed inthe stacking procedure for the 18 galaxies, which maybe identified as the broad wing feature. Another possi- CII] Halo at z ∼ ! " $ %& ’ ( ) " !" !" ! & %&’()" Figure 9. Top:
Two-component S´ersic+exponential profile fit-ting for the [C ii ] line, averaged over 18 galaxies. The red-dashedcurves represent the best-fit results of the central stellar continuumand outer halo components, while the solid red curve denotes thesum of the best-fit two-component results. The solid blue curveand the shaded region indicate the median and the 16–84th per-centile of the radial surface brightness profile of the Ly α lines in arecent control sample from MUSE (Leclercq et al. 2017). For theLy α line, we convolve the best-fit results of the S´ersic+exponentialprofiles with the ALMA beam. The red and blue arrows with er-ror bars show the best-fit scale lengths of the [C ii ] and the Ly α halo components, respectively. Bottom:
Residuals in the best-fitresults of one- (left) and two-component (right) profile fittings. bility is that the faint continuum emission is mistakenlyidentified as the broad wing feature. Although we haveperformed the continuum subtraction for the [C ii ] linedata cubes of the 4 galaxies whose continuum emission isindividually detected, it is possible that the faint contin-uum emission from the rest of the 14 (= 18 −
4) galaxiesappears in the deeply stacked spectrum. Since the signif-icance of the broad wing is low, we cannot draw definiteconclusions from our data.
Comparison with Model
We compare our observational results with two inde-pendent numerical simulations for star-forming galaxieswith the halo mass of M halo ∼ − M ⊙ at z ∼ M UV value of . −
21 mag (Table 1) which correspondsto M halo ≈ − M ⊙ from the M UV – M halo relation !" !" +,%-./0!" )2+%-./0345678549:7% Figure 10.
ALMA [C ii ]-line spectrum averaged over the ALMA-ALL sample. The spectrum is derived with an aperture diameterof 0 . ′′
4. The red curves denote the best-fit two (= core + broad)Gaussian component model. The shade regions indicate the veloc-ity ranges in which the velocity-integrated intensity is tentativelydetected at the 3.2 σ level. (Harikane et al. 2018).First set is a zoom-in simulation for a star-forminggalaxy, Althæa (Pallottini et al. 2017a,b; Behrens et al.2018). The hydrodynamical and dust radiative transfer(RT) simulations are combined, which provides realis-tic predictions for the spatial distribution of the [C ii ]line as well as the rest-frame FIR and UV continuumemission, with a spatial resolution of 30 pc. The hy-drodynamical and the dust RT simulations are fullydescribed in previous studies (Pallottini et al. 2017a,b;Behrens et al. 2018). Note that the dust RT is calculatedas a post-processing step on snapshots of the hydrody-namical simulation. The [C ii ] line emission is computedin post-processing (Vallini et al. 2015) by adopting thephotoionization code, cloudy (Ferland et al. 2017). Inthese processes, CMB suppression (da Cunha et al. 2013;Zhang et al. 2016; Pallottini et al. 2015; Lagache et al.2018) is included in the calculation.Second set is another cosmological hydrodynamiczoom-in simulations performed by the smoothed par-ticle hydrodynamics (SPH) code Gadget-3 (Springel2005) with the sub-grid models developed in
Overwhelm-ingly Large Simulations (OWLS) project (Schaye et al.2010) and the
First Billion Year (FiBY) project (e.g.,Johnson et al. 2013) which reproduce the general prop-erties of the high-redshift galaxy population well (e.g.,Cullen et al. 2017). For this comparison, we use fourdifferent halos: Halo-12, Halo-A, Halo-B, and Halo-Cthat have M halo = (0 . − × M ⊙ at z = 6 . − . ǫ g = 200 pc (comoving),therefore we achieve ∼
25 pc resolution at z = 7 for grav-ity. We also allow the SPH smoothing length to adaptivedown to 10% of ǫ g , therefore the hydrodynamic resolu-tion reaches a several parsecs at z = 6 −
7. The RTcalculation including the dust absorption/re-emission isperformed as a post-process with “All-wavelength Ra-2
Fujimoto et al. diative Transfer with Adaptive Refinement Tree” (ART code: Li et al. 2008; Yajima et al. 2012). This calcu-lation provides the SED over a wide wavelength range,and solves for the ionization structure of ISM/CGM. The[C ii ] line emissivity is estimated from the ionized carbonabundance. The details of hydrodynamic simulations,RT, and the [C ii ] line computations are fully describedin Yajima et al. (2017), Arata et al. (2019b), and Arataet al. (2019, in prep.).The left panel of Figure 11 presents a color compos-ite of the [Cii] line, rest-frame UV, and FIR continuumemission of Althæa at z = 6 . z = 6 . [Cii] line emission clearly shows the ex-tended structure over the 10-kpc scale with surroundingsatellite clumps and filamentary structures. The pictureof the extended [C ii ] halo around the central galaxies isroughly consistent with the observational results.To quantitatively compare the zoom-in simulation toour observational results, we carry out the stacking forthe zoom-in simulation results of the [C ii ] line, the rest-frame FIR and the rest-frame UV continuum in the samemanner as the observations. For the Althæa simula-tion, we take 12 snapshots at different redshifts within6 . ≤ z ≤ . ii ] line emissivity is calculated within 100 km s − ofthe velocity center of the galaxy to match the visibility-based stacking procedure for the ALMA data. In thisway, we obtain 48 (=12 ×
4) images of Althæa. We re-fer to Kohandel et al. (2019) for a full analysis of thedifferent morphological results from different evolution-ary stages and viewing angles. We then select 9 out of48 snapshots randomly – the same sample size used forthe stacking of the ALMA-HST sample. Finally, we per-form the stacking of the intrinsic images, and smooth thestacked images with the ALMA beam. For the second setof simulation, we calculate the surface brightness fromthree orthogonal angles for four different halos (Halo-12,Halo-A, Halo-B, Halo-C) and obtain 12 (=4 ×
3) images.We then carry out the same stacking and smoothing pro-cedures as the first simulation.In the right panel of Figure 11, we show the radialsurface brightness profiles estimated from the two inde-pendent zoom-in simulations. For comparison, we alsoplot the observational results obtained in Section 4. Wefind that both simulations reproduce the overall trend ofobservational results of rest-frame UV and FIR contin-uum within the errors. However, we also find that the [Cii] line emission in both simulations is not as extendedas the observed data. In Althæa, although it reproducesthe trend that the [C ii ] line is more extended than thatof the rest-frame FIR and UV continuum, the intensityof the [C ii ] emission at r > ii ] line is the least ex-tended. These results indicate that the existence of the[C ii ] halo challenges current hydrodynamic simulationsof galaxy formation. DISCUSSION
In Section 4, we find that the [Cii] line emis-sion is extended up to ∼ z = 5 − α halo. In con- trast to the previous reports of the 10-kpc-scale car-bon reservoirs around rare, massive galaxies, suchas dusty starbursts and quasars at z ∼ ii ] line emission and how is the carbon abundancein the circum-galactic (CG) area enriched at such earlycosmic epochs. Theoretical studies suggest the followingfive scenarios that can give rise to the extended [C ii ] lineemission with the potential association of the Ly α halo:A) satellite galaxies,B) CG photodissociation region (PDR),C) CG HII region,D) cold streams,E) outflow.These five scenarios are illustrated in Figure 12.The first scenario invokes satellite galaxies (Figure 12-A). If satellite galaxies exist around the central star-forming galaxies, the [Cii] and Ly α line emission fromthe satellite galaxies will be observed as extended struc-tures around the central galaxies. In this scenario, theextended halo size is determined by the spatial distri-bution of the satellite galaxies, which explains both ex-tended components of the [Cii] and Ly α line emission.The second scenario is a PDR extended over CG scale,referred to as CG-PDR (Figure 12-B). The ionizing pho-tons ( hν > . < hν < . [Cii] line emission is thus detected on the CGscale. Besides, the Ly α line emission is also spatiallyextended due to the resonant scattering by the neutralhydrogen in the surrounding ISM (e.g., Xue et al. 2017).The third scenario is that ionizing photons penetratethe surrounding ISM deeper and form large HII re-gions even spreading over the CGM, which we referto as CG-HII (Figure 12-C). This scenario is similarto scenario (B), but the existence of strong ionizingsources and/or ISM properties differ from scenario (B),and the HII region is larger than scenario (B) wherethe carbon is singly ionized. In this case, the Ly α line emission is extended due to the fluorescence (e.g.,Mas-Ribas & Dijkstra 2016), instead of the resonancescattering in scenario (B).The fourth scenario is cold streams (Figure 12-D). Cos-mological hydrodynamical simulations suggest that in-tense star-formation in high- z galaxies is fed by a denseand cold gas ( ∼ K) which is dubbed cold streams(e.g., Dekel et al. 2009). The cold streams radiate [Cii] as well as Ly α line emission powered by gravitational en-ergy, and produce the extended [Cii] and Ly α line emis-sion around a galaxy. Moreover, the cold stream maycause shock heating which can also produce the [Cii] and Ly α line emission.The fifth scenario is outflow (Figure 12-E). In the out- CII] Halo at z ∼ Althæa (z = 6.0)Halo-12 (z = 6.5)
10 kpc !" !" + , $ , - . / ’’ !" $% & ’ ( - - !" $% )* " - . / ’’ Figure 11. Left: ′′ × ′′ fake-color image for Althæa at z = 6 . [Cii] line,green: rest-frame FIR continuum, blue: rest-frame UV continuum). Right:
Radial surface brightness profiles of the [Cii] line (red curve),rest-frame FIR (green curve), and UV (blue curve) continuum emission estimated in the zoom-in simulations via stacking procedure. Thesolid and dashed color lines present the Althæa and Halo-12 results, respectively. The black dashed curve denotes the ALMA synthesizedbeam. The circles indicate the ALMA-HST stacking results whose colors are assigned in the same manner as the left panel. flow, the ionized carbon and hydrogen powered by theAGN and/or star-formation feedback produce the ex-tended [Cii] and Ly α line emission (see also Faisst et al.2017). The associated process of the shock heating mayalso contribute to radiating these line emission. Notethat although we choose ALMA sources not reported asAGNs, we cannot rule out the possibility that our ALMAsources contain faint AGNs and/or have the past AGNactivity.In the following subsections, we discuss these possibil-ities based on the observational and theoretical results. Hints From Observational Results
In the observational results, the [C ii ] line is more ex-tended than both the rest-frame FIR dust and UV con-tinuum beyond the errors up to a radius of at least ∼ ii ] line emissivityat a given stellar continuum (De Looze et al. 2014), thelarge gap between the radial profiles of the [C ii ] line andthe stellar continuum indicates that the stellar contin-uum is not enough to explain the large part of the [C ii ]line emissivity of the [C ii ] halo.Although the [C ii ] line emissivity may be changed fromthe central to halo areas at a given stellar continuum, themetallicity at such outer areas is expected to be ∼ − metallicity rela-tion (Mannucci et al. 2010) also suggests lower metallic-ities for the satellite galaxies. Because lower metallicityreduces the [C ii ] line emissivity for a given stellar contin-uum (Vallini et al. 2015), it would be difficult to explainthe [C ii ] halo by the same source as the stellar contin-uum. In fact, Figure 8 shows that the L [CII] /SFR total ra-tio becomes higher towards outskirts of the halo, whichcannot be explained by the dwarf galaxies. Our obser-vational results thus rule out scenario (A), and supportthe other four scenarios.In the recent [Cii] line studies at z >
5, Gallerani et al.(2018) report signatures of starburst-driven outflowsfrom 9 normal star-forming galaxies at z ∼ . [Cii] spectra. With the similar sample, therest-frame UV metal absorptions are also identified tobe blue-shifted from the [C ii ]-systemic redshift in thestacked Keck spectra (Sugahara et al. 2019). From moreluminous objects, the broad wing features are detectedin the stacked [C ii ] line spectra of z ∼ − ii ]line spectrum of a quasar at z = 6 . [Cii] spec-tra even from 23 quasars at z & Fujimoto et al. !" /0’1,*%--!*%’2,-,3!%. 40’425(6670’4"-&’1*+%,8 90’425:7;<0’=)*>-"? @ABCDE
E% % + % Figure 12.
Illustrations of five possible scenarios for the physical origin of the [C ii ] halo with the potential association of the Ly α halo:A) satellite galaxies; B) circum-galactic (CG) photodissociation region (PDR); C) CG HII region (CG-HII); D) cold stream; and E) outflow.The blue and red shades show the neutral and ionized hydrogen in ISM and CGM. The yellow stars represent the star-forming regions.The inner and outer dashed circle denote the effective radii ( r e ) of the central and halo components of the [Cii] line emission, respectively. Also, our stacked [C ii ] spectra with the 18 star-forminggalaxies does not show a clear broad wing feature, nei-ther (Section 4.5). Among scenarios of (B), (C), (D), and(E), we thus cannot conclude the most likely one fromour and recent observational results. Hints From Theoretical Results
In the simulation results, the extended profile of [C ii ]halo is not fully reproduced (Figure 11). This maysuggest that some physical processes are not sufficientlysolved in the simulations, e.g., metal enrichment, feed- back, ISM/CGM clumpiness, and the propagation of ion-izing radiation. On the other hand, if the current as-sumptions related to the [C ii ] line emissivity are correct,additional mechanism(s) are required to produce the ex-tended [C ii ] line emissivity in the simulation.There are two possibilities for such additional mech-anisms that are not included in the calculation of the[C ii ] line emissivity in the simulations. The first pos-sible mechanism is the shock heating; Appleton et al.(2013) have shown that [C ii ] can be excited on largescales from the dissipation of mechanical energy of galaxy CII] Halo at z ∼ cloudy does not adequately consider the effect ofshock heating. Since shock heating is caused by galaxymerger or gas inflow/outflow processes, the [C ii ] emis-sivity could become more enhanced if the shock heatingand associated turbulent cascade is properly treated inthe scenarios (A), (D), and (E).The second possible mechanism is the past/on-goingAGN activities, which could form a large HII region andsurrounding PDR. Moreover, the AGN feedback maycause shock heating, which also could contribute to the[C ii ] line emissivity. In this case, the scenarios of (C)and (E) are further supported.Note that if the effect of shocks and AGNs is too strong,the carbon may be doubly ionized, and then the [C ii ]line is rarely emitted. Therefore, it is hard to concludewhether the missing treatment of shocks and/or AGNsin current simulations are the major causes of the inad-equate [C ii ] halo in the simulations.It should also be noted that 7 out of 9 sources in theALMA-HST sample are placed at 5 < z < z = 6 . − .
2, from whichthe zoom-in simulation results were taken. Because theCMB effect reduces the line luminosity from the dif-fuse component (e.g., da Cunha et al. 2013; Zhang et al.2016; Pallottini et al. 2015; Lagache et al. 2018), theslight difference in the redshift range may cause the in-sufficient [Cii] line luminosity in the zoom-in simulationresults.
Physical Origin of [Cii]
Halo
We summarize the possible scenarios of what powersthe [Cii] halo based on the results of Sections 5.1 − & K) can be longer than the cos-mic time at z ∼ − ∼ ii ]-emitting coldhalos from the hot-mode outflow. On the other hand,the cold-mode outflow consists of the cold neutral hy-drogen gas that is pushed by the radiative and kinetic pressures exerted by SNe, massive stars, and AGNs. Inthis case, the majority of [Cii] line emission would beradiated from the PDR in the cold, neutral hydrogen gasclouds. Therefore our finding of the [Cii] halo suggeststhat outflows in the early star-forming galaxies may bedominated by the cold-mode outflows.Since we also find the similarity in the radial surfacebrightness profiles between the [Cii] and Ly α halos (Fig-ure 9), the physical origin of the [Cii] halo may be re-lated to the Ly α halo. Future deep observations of both [Cii] and Ly α line emission for individual high- z galaxiesare required to comprehensively understand the mecha-nism of the CGM metal enrichment with the theoreticalsimulations including the radiative transfers of these lineemission. SUMMARY
In this paper, we study the detailed morphology of [Cii] line emission via the ALMA visibility-based stack-ing method for normal star-forming galaxies whose [Cii] line have been individually detected at z = 5 . − . uv -plane, which enables usto securely investigate the diffuse emission extended overthe circum-galactic environment. In conjunction withthe deep HST/ H -band data, we examine the radial sur-face brightness profiles of the [Cii] line, rest-frame FIR,and UV continuum emission. We then discuss the physi-cal origin of the extended [Cii] line emission. The majorfindings of this paper are summarized below.1. The visibility-based stacking of our and archivaldeep ALMA data for 18 galaxies with SFR ≃ − M ⊙ yr − at z = 5 . − .
142 produces21 σ and 10 σ level detections at the peak for the [Cii] line and dust continuum emission, respec-tively. The stacked [Cii] line morphology is spa-tially extended more than that of the dust contin-uum. The radial surface brightness profiles of the [Cii] line are extended up to a radius of ∼ σ level.2. The HST/ H -band stacking for 9 out of the 18 [Cii] line sources that are taken by the deep HST ob-servations shows that the radial surface brightnessprofiles of the [Cii] line is significantly extendedmore than that of the rest-frame UV as well as therest-frame FIR continuum emission. We derive theradial ratio of L [CII] /SFR total , showing that the ra-tio becomes higher towards the outskirts of halowhere the high ratios cannot be explained by thesatellite galaxies.3. The two-component S´ersic+exponential profile fit-ting results indicate that the extended [Cii] halocomponent has the scale length of 3.3 ± α halo, universallyfound around the high- z star-forming galaxies. Interms of effective radius, the extended [Cii] halocomponent is larger than the central galactic com-ponent by a factor of ∼ Fujimoto et al.[Cii] line emission and the rest-frame FIR, com-parable to the rest-frame UV continuum emission.However, the simulations do not reproduce the fullextent of the [Cii] halo in the outskirts, where thesimulations might be missing some physical mech-anisms associated with the feedback, or still lack-ing the resolution to resolve the turbulent cascadefrom large-scale shocks down to the small scales ofmolecular clouds, if such a process is indeed im-portant for the [C ii ] emission in high- z galaxies asAppleton et al. (2013) argued.5. Although there remain several possible scenariosthat can give rise to [C ii ] line emission in theCGM, the outflow is required in any cases to en-rich the primordial CGM with carbon around theearly star-forming galaxies. Our results are thusthe evidence of outflow remnants in the early star-forming galaxies and suggest that the outflow maybe dominated by the cold-mode outflow.We thank the anonymous referee for constructivecomments and suggestions. We are grateful to IvanMarti-Vidal and the Nordic ALMA Regional Centerfor providing us with helpful CASA software toolsand advice on analyzing the data. We appreciateTohru Nagao, Jeremy Blaizot, Peter Mitchell, TakashiKojima, Shiro Mukae, Yuichi Harikane, Akio Inoue,and Rieko Momose for useful comments and sugges-tions. We are indebted for the support of the staff at the ALMA Regional Center. This paper makesuse of the following ALMA data: ADS/JAO. ALMA APPENDIX
OUR ALMA SAMPLE
The sources drawn from the literature in our ALMA sample is summarized in Table 1. For this literature sample,Figure 13 shows the [C ii ] line velocity-integrated maps and the spectra obtained from our re-analysis of the archivalALMA data. We confirm that the spatial morphology and the spectrum shape of the [C ii ] lines are consistent withthe previous studies (Capak et al. 2015; Willott et al. 2015; Pentericci et al. 2016; Smit et al. 2018; Carniani et al.2018). REFERENCES Appleton, P. N., Guillard, P., Boulanger, F., et al. 2013, ApJ,777, 66Arata, S., Yajima, H., Nagamine, K., Li, Y., & Khochfar, S.2019a, arXiv e-prints, arXiv:1908.01438—. 2019b, MNRAS, 488, 2629Aravena, M., Decarli, R., Walter, F., et al. 2016, ApJ, 833, 71Barisic, I., Faisst, A. L., Capak, P. L., et al. 2017, ApJ, 845, 41Behrens, C., Pallottini, A., Ferrara, A., Gallerani, S., & Vallini, L.2018, MNRAS, 477, 552Bertin, E., & Arnouts, S. 1996, A&A, 117, 393Bischetti, M., Maiolino, R., Fiore, S. C. F., Piconcelli, E., &Fluetsch, A. 2018, ArXiv e-prints, arXiv:1806.00786Bouwens, R. J., van Dokkum, P. G., Illingworth, G. D., et al.2017, ArXiv e-prints, arXiv:1711.02090Capak, P. L., Carilli, C., Jones, G., et al. 2015, Nature, 522, 455Carniani, S., Maiolino, R., Amorin, R., et al. 2017, ArXive-prints, arXiv:1712.03985—. 2018, MNRAS, arXiv:1712.03985Cicone, C., Maiolino, R., Gallerani, S., et al. 2015, A&A, 574, A14Coppin, K., Halpern, M., Scott, D., et al. 2008, MNRAS, 384,1597Cullen, F., McLure, R. J., Khochfar, S., Dunlop, J. S., & DallaVecchia, C. 2017, MNRAS, 470, 3006da Cunha, E., Groves, B., Walter, F., et al. 2013, ApJ, 766, 13De Looze, I., Cormier, D., Lebouteiller, V., et al. 2014, A&A, 568,A62Decarli, R., Walter, F., Venemans, B. P., et al. 2018, ApJ, 854, 97Dekel, A., Birnboim, Y., Engel, G., et al. 2009, Nature, 457, 451 D´ıaz-Santos, T., Armus, L., Charmandaris, V., et al. 2014, ApJ,788, L17Dunlop, J. S., McLure, R. J., Biggs, A. D., et al. 2017, MNRAS,466, 861Faisst, A. L., Capak, P. L., Yan, L., et al. 2017, ApJ, 847, 21Falgarone, E., Zwaan, M. A., Godard, B., et al. 2017, Nature,548, 430Ferland, G. J., Chatzikos, M., Guzm´an, F., et al. 2017, Rev.Mexicana Astron. Astrofis., 53, 385Fujimoto, S., Ouchi, M., Ono, Y., et al. 2016, ApJS, 222, 1Fujimoto, S., Ouchi, M., Shibuya, T., & Nagai, H. 2017, ApJ,850, 1Fujimoto, S., Ouchi, M., Kohno, K., et al. 2018, ApJ, 861, 7Gaia Collaboration, Brown, A. G. A., Vallenari, A., et al. 2018,A&A, 616, A1Gallerani, S., Pallottini, A., Feruglio, C., et al. 2018, MNRAS,473, 1909George, R. D., Ivison, R. J., Smail, I., et al. 2014, MNRAS, 442,1877Harikane, Y., Ouchi, M., Ono, Y., et al. 2018, PASJ, 70, S11Hashimoto, T., Inoue, A. K., Mawatari, K., et al. 2018, arXive-prints, arXiv:1806.00486Hayatsu, N. H., Matsuda, Y., Umehata, H., et al. 2017, PASJ, 69,45Heckman, T. M., & Thompson, T. A. 2017, Galactic Winds andthe Role Played by Massive Stars, ed. A. W. Alsabti &P. Murdin, 2431Hodge, J. A., Swinbank, A. M., Simpson, J. M., et al. 2016, ApJ,833, 103
CII] Halo at z ∼ !" Figure 13.
The [C ii ] line velocity-integrated map (left) and spectrum (right) obtained in our re-analysis of the archival ALMA data forthe literature sample in Table 1. Left:
Natural-weighted 4 ′′ × ′′ field image of the velocity-integrated [C ii ] line intensity (moment zero)with contours at the − σ (white), 2 σ , 3 σ , ..., 10 σ (red) levels. The synthesized beam is presented at the bottom left. North is up, andeast is to the left. Right: [C ii ] line spectrum with an aperture diameter of 1 . ′′
2. The blue curve denotes the best-fit profile of the singleGaussian. Here we perform the fitting in the velocity range of ±
300 km s − from the [C ii ] line frequency center estimated in the previousstudies. The yellow shades present the integrated velocity range for the [C ii ] line intensity map presented in the left panel. The velocitiesare relative to the center of the best-fit Gaussian. Fujimoto et al.
Figure 13. (continued)
CII] Halo at z ∼6 19