Compact starbursts in z~3-6 submillimeter galaxies revealed by ALMA
Soh Ikarashi, R. J. Ivison, Karina I. Caputi, Itziar Aretxaga, James S. Dunlop, Bunyo Hatsukade, DavidH. Hughes, Daisuke Iono, Takuma Izumi, Ryohei Kawabe, Kotaro Kohno, ClaudiaD. P. Lagos, Kentaro Motohara, Koichiro Nakanishi, Kouji Ohta, Yoichi Tamura, Hideki Umehata, Grant W. Wilson, Kiyoto Yabe, Min S. Yun
aa r X i v : . [ a s t r o - ph . GA ] A ug D RAFT VERSION O CTOBER
12, 2018
Preprint typeset using L A TEX style emulateapj v. 08/22/09
COMPACT STARBURSTS IN Z ∼ S OH I KARASHI , R. J. I
VISON , K
ARINA
I. C
APUTI , I TZIAR A RETXAGA , J AMES
S. D
UNLOP , B UNYO H ATSUKADE ,D AVID
H. H
UGHES , D AISUKE I ONO , T
AKUMA I ZUMI , R YOHEI K AWABE , K
OTARO K OHNO , C
LAUDIA
D. P. L
AGOS ,K ENTARO M OTOHARA , K OUICHIRO N AKANISHI , K
OUJI O HTA , Y OICHI T AMURA , H IDEKI U MEHATA , G RANT
W. W
ILSON ,K IYOTO Y ABE , M IN S. Y UN E UROPEAN S OUTHERN O BSERVATORY , K
ARL S CHWARZSCHILD S TR . 2, D-85748 G ARCHING , G
ERMANY I NSTITUTE OF A STRONOMY , U
NIVERSITY OF T OKYO , 2-21-1 O
SAWA , M
ITAKA , T
OKYO
APAN K APTEYN A STRONOMICAL I NSTITUTE , U
NIVERSITY OF G RONINGEN , P.O. B OX RONINGEN , T HE N ETHERLANDS ; sikarash@astro . rug . nl I NSTITUTE FOR A STRONOMY , U
NIVERSITY OF E DINBURGH , R
OYAL O BSERVATORY , B
LACKFORD H ILL , E
DINBURGH
EH9 3HJ, UK I NSTITUTO N ACIONAL DE A STROFÍSICA , Ó
PTICA Y E LECTRÓNICA (INAOE), A
PTDO . P
OSTAL Y UEBLA , M
EXICO N ATIONAL A STRONOMICAL O BSERVATORY OF J APAN , M
ITAKA , T
OKYO
APAN SOKENDAI (T HE G RADUATE U NIVERSITY FOR A DVANCED S TUDIES ), S
HONAN V ILLAGE , H
AYAMA , K
ANAGAWA
APAN R ESEARCH C ENTER FOR THE E ARLY U NIVERSE , S
CHOOL OF S CIENCE , U
NIVERSITY OF T OKYO , 7-3-1 H
ONGO , B
UNKYO , T
OKYO
APAN J OINT
ALMA O
BSERVATORY , A
LONSO DE C ORDOVA
ITACURA , S
ANTIAGO
763 0355, C
HILE D EPARTMENT OF A STRONOMY , K
YOTO U NIVERSITY , K
ITASHIRAKAWA -O IWAKE -C HO , S AKYO - KU , K YOTO
APAN D EPARTMENT OF A STRONOMY , U
NIVERSITY OF M ASSACHUSETTS , A
MHERST , MA 01003, USA
Draft version October 12, 2018
ABSTRACTWe report the source size distribution, as measured by ALMA millimetric continuum imaging, of a sampleof 13 AzTEC-selected submillimeter galaxies (SMGs) at z phot ∼ L IR ∼ × L ⊙ and ∼ ⊙ yr - , respectively. The sizes of these SMGsrange from 0 ′′ .10 to 0 ′′ .38, with a median of 0 ′′ .20 + ′′ . - ′′ . (FWHM), corresponding to a median circularizedeffective radius ( R c , e ) of 0.67 + . - . kpc, comparable to the typical size of the stellar component measured incompact quiescent galaxies at z ∼ R e ∼ + - M ⊙ yr - kpc - , comparable to that seen in local merger-driven (U)LIRGs rather than in extended diskgalaxies at low and high redshifts. The discovery of compact starbursts in z & z ∼ z ∼ z & ∼
90% of the age of the Universe.
Subject headings: submillimeter: galaxies — galaxies: evolution — galaxies: formation — galaxies: high-redshift INTRODUCTIONThe most massive galaxies in the local Universe are thoughtto have evolved to their current state via a series of dry merg-ers of relatively gas-poor galaxies over the last 10 Gyr (e.g.Newman et al. 2012; Oser et al. 2012; Carollo et al. 2013;Krogager et al. 2014). Their ancestors – the so-called ‘com-pact quiescent galaxiesâ ˘A ´Z (cQGs) – are found at z ∼ ∼ × smaller effective radii ( R e ∼ & × denserthan their local descendants (e.g. van Dokkum et al. 2008;Onodera et al. 2010; Newman et al. 2012) and the process bywhich they form remains a mystery. Recent attempts to probetheir star-forming phase using conventional NIR observationsresulted in the discovery of a relatively unobscured starburst,seen around z ∼ . z ∼ z ∼ . z ∼ ∼ ′′ .5 (full width athalf maximum; FWHM) corresponding to a radius of ∼ z ∼ z & z & z & z ∼ z = 6 . ∼ . L IR > L ⊙ ) and we needto image more typical SMGs with L IR ∼ L ⊙ .Sensitivity limitations of existing arrays meant we needed Ikarashi et al.to wait for ALMA in order to measure the far-infrared (FIR)sizes of z & H = 70 km s - Mpc - , Ω M = 0 . Ω Λ = 0 . AZTEC-SELECTED Z & µ m map contains a total of 221 millimeter sources overa contiguous area of 950 arcmin . We selected our ALMAtargets based on their faintness in the Herschel images( S ν (250 µ m) < . - ; 3 σ ; Oliver et al. 2012)and VLA 1.4-GHz map ( . µ Jy; 5 σ ) (Arumugam et al.,in preparation). With the ALMA observations we detected35 significant ( ≥ σ ) SMGs (hereafter ASXDF sources)associated with 30 AzTEC sources.Given the strong negative K-correction at λ ∼ µ m, the faintness in the Herschel and radio bands indi-cate that these 1100- µ m-selected galaxies are expected to beat high redshifts, i.e., z &
3. Note that these galaxies consti-tute a complementary population to that studied in early sub-millimetre galaxy studies, which were biased towards radio-bright sources, and found to lie at lower redshifts z ≈ S / N ≥
10 in the ALMA 1100- µ m continuum map.Such a high S / N threshold ensures that we can study theirsizes with good accuracy with the ALMA continuum data(See details in Section 3.1). As the focus of this paper ison z & z phot ≥ .
8, or are faintin IRAC ( F . µ m ≥ AB ) and detected in at most fouroptical/near-mid-IR broad bands, indicating a likely high red-shift, as we explain below.At these shorter wavelengths (optical through mid-IR), wehave performed the spectral energy distribution (SED) analy-sis of our sources based on 12 bands, namely B , V , Rc , i , z ′ , J , H , Ks -bands and IRAC 3.6, 4.5, 5.8 and 8.0 µ m (Ikarashiet al. in prep), using the same method described in Caputi etal. (2012). For three out of our 13 sources, we obtained red-shift estimates, z phot , and derived parameters (Table 1). Theremaining ten sources are only detected in four or less broadbands, so no robust z phot can be obtained from the SED fitting.Figure 1 shows a 4.5- µ m–redshift plot for ALMA sourcesand those in the ALMA-identified SMG sample (ALESS) re-ported by Simpson et al. (2014). The dashed line in this plotindicates the median 4.5- µ m–redshift relation, and the solidline corresponds to this same relation minus the 1 σ scatter,which we have derived using the average SED of the ALESSsources (see Figure 8 in Simpson et al. 2014). We expectmore than 85% of the ASXDF sources with F . µ m ≥ .
75 m AB to be located at z &
3. About 15% of SMGs are expectedto have F . µ m fainter than the solid black curve in Figure 1at each redshift, and 15% of SMGs at z = 3 are expected tohave F . µ m ≥ .
75 m AB . So, by selecting only those galaxieswith no redshift estimate in our sample, with F . µ m ≥ . z & z & µ m (PACS and SPIRE), 1100 µ m (ALMA) and 21 cm(VLA) with the SED of the averaged SMGs ( T d = 32 K) at z =3, 4 and 5. All of the stacked fluxes and errors are basedon bootstrapping analysis. We see that the stacked submil-limeter/radio fluxes are best fitted at z ∼
4. Given a stackedVLA 1.4-GHz flux density of 15 . ± . µ Jy and the observedALMA 1100- µ m flux density, we expect a photometric red-shift, z = 4 . + . - . , for our ASXDF sources, based on their ra-dio/(sub)millimeter color (e.g. Carilli & Yun 1999). Note that,for this exercise, we have considered the radio-FIR SED tem-plate of the averaged SMGs (Swinbank et al. 2014) derivedfrom the ALMA-identified SMGs at z ∼ z & T d follows the trend that SMGs at higher redshiftshave higher T d , then the redshifts should indeed be z &
4. Atthe moment, it is difficult to be confident of such a trend asthe samples of known z & z & µ m range from 1.5 to3.4 mJy, corresponding to star formation rates (SFRs) of ∼ ⊙ yr - and L IR ∼ × L ⊙ . The median SFRand L IR are 340 + - M ⊙ yr - and 3.4 + . - . × L ⊙ , respectively.The properties of our sample are summarized in Table 1. TheSFRs and L IR are estimated from the average SMG SEDs. Weconsidered uniform redshift probability density at z = 3 - z phot determination, and a 1 σ error forsources which do have a z phot . SOURCE-SIZE MEASUREMENTS3.1.
Data, method and results
We measured the source sizes of our ASXDF sources us-ing ALMA continuum data centered at 265 GHz. Our ALMAobservations were obtained in three blocks, with only smalldifferences in antenna configurations between blocks. Sevenof the 13 ASXDF sources were observed with 25 working 12-m antennas, mainly covering baselines up to 400 k λ , corre-sponding to physical baseline lengths of 440 m. The remain-ing six sources were observed with three more 12-m anten-nas, deployed for tests on longer baselines, covering up to1200 k λ or 1320 m. The extended-baseline data from 400 to1200 k λ are used here only as supplementary data because oftheir limited uv coverage (Figure 3). On-source observationtimes were 3.6–4.5 minutes, sufficient to achieve r.m.s. noiselevels of 70–88 µ Jy beam - . The synthesized beam size inour ALMA continuum images, using baselines up to 400 k λ ,is ∼ ′′ . & Figure 1.
Observed 4.5- µ m flux of submillimeter galaxies as a functionof redshift. Blue dots mark ALMA-identified LABOCA (ALESS) sources(Simpson et al. 2014). Red points mark ASXDF sources with photometricredshifts in our sample for source size measurements. Black curves show theredshift-4.5- µ m relation expected from the absolute H -band flux distribu-tion of ALESS sources and the optical/NIR SED of average ALESS sources(Simpson et al. 2014); Dashed line is for SMGs with the median absolute H -band flux; solid shows the absolute H -band flux distribution minus 1 σ . Lightred bars mark the 4.5- µ m flux of the ASXDF sources without photometricredshifts. Solid green horizontal bar marks F . µ m at 22.75 m AB , which isthe threshold for selection of z & σ lower limits of redshift for each source (thesevalues are listed in Table 1). Figure 2.
Stacked submillimeter/radio SED of the ASXDF sources withoutphoto- z . Error are estimated by Bootstrapping analysis. Colored SED is thatof average SMGs (Swinbank et al. 2014) for z =3, 4 and 5, as best fit to theALMA flux. For PACS 100 and 160 µ m and SPIRE 250 µ m data, we plot3 σ upper limits. the beam size by a CASA task, IMFIT ; about the 5 of the 13are unresolved or point-like.In this paper we have measured source sizes using the visi-bility data directly – on uv –amplitude plots (hereafter uv -ampplots) – assuming a symmetrical Gaussian as was done inprevious studies, to exploit the long-baseline ( ≤
400 k λ ) datafor source size measurements (Figure 4). Source-size mea-surements using uv -amp plots have often been made in previ-ous studies using, e.g., SMA and CARMA, in order to betterconstrain the size of largely unresolved sources in an image(e.g. Iono et al. 2006; Younger et al. 2008; Ivison et al. 2010;Ikarashi et al. 2011). This is equivalent to measuring the cir-cularized effective radius, R c , e .In this paper, we have been able to polish this method,owing to the high data quality from ALMA. We have eval-uated the accuracy of our source-size measurements usinga Monte-Carlo simulation, for the purpose of correcting forany systematics and obtaining more reliable source sizes. Wegenerated 82000 mock sources with a symmetric Gaussianprofile in noisy visibility data, for a range of source sizesand flux densities that cover the putative parameter range ofour ASXDF sources. We measured source sizes and createdcleaned continuum images in the same manner as we had donefor our real targets, in order to derive a relation between theinput source size, the measured source size and the signal-to-noise ratio in a continuum image. Figure 5 shows the derivedrelation between measured source size found by fitting in uv -amp plots and the ‘actual’ size input for the simulation, eachversus source size for continuum detections of 10 and 15 σ .This plot demonstrates that our source size measurement isaccurate to within 1 σ and that actual source sizes are system-atically a little bit larger than the measured source sizes. Wetherefore adopt source sizes after making a correction basedon this relation between measured and actual source sizes; thecorrection is done using the probability distribution of actualsource size for the appropriate signal-to-noise ratio and mea-sured size of each source. In this paper, we measure millime-ter size of ASXDF sources with ≥ σ continuum detections.This is because size measurement in the visibility data for 10 σ sources gets less sensitive at FWHM < ′′ .2, i.e., losing inear-ity (Figure 5). Measurement of our sample is safe from thisissue; there is just one ASXDF source with S/N=10, but itsmeasured size is ∼ ′′ .3, and we checked that our measure-ment for the second lowest S/N of 11.3, is sensitive down to0 ′′ .1.The measured source sizes are listed in Table 1. The sourcesizes of our sample range from 0 ′′ .10 to 0 ′′ .38 with a medianof 0 ′′ .20 + ′′ . - ′′ . (Figure 6). We also check the dependency ofthe circularized size measured in uv -amp plots on ellipticityvia simulations, for minor/major axis ratio of 0.5, 0.6, 0.7,0.8, 0.9 and 1.0 (Figure 5). These simulation suggest thatthe measured circularized size does not depend strongly onthe ellipticity; however, a weak dependency exists: at an axisratio of 0.5, the circularized size can be over-estimated about10 percent. 3.2. Ancillary long-baseline data
As we noted above, six of 13 ASXDF sources in our sam-ple were observed with additional three long-baseline anten-nas covering up to 1200 k λ (Table 1) which enable us to If we measure the size of a disk-like source using a Gaussian fit, theactual size of the disk ( R e , Disk ) is empirically ∼ . × larger than the sizemeasured ( R e , Gauss ) for measurements at ≤
400 k λ . Ikarashi et al.
Figure 3.
The uv coverage for ASXDF sources with long-baseline antennasfor Schedule Block 1 in Table. 1. We use visibilities at uv distances of ≤
400 k λ for source-size measurements where u – v coverage is well sampled.We use visibilities at 400–1200 k λ only to check the consistency between theexpected long-baseline visibilities and the measured size. make millimeter images of the sources with an angular res-olution of 0 ′′ .2 (FWHM). Using the long-baseline data, wecheck some concerns in our source size measurements —assumptions made about source multiplicity (i.e. that thereis none), the possibility of source ellipticity, and the possi-bility of faint, extended emission. In Figure 4, we presenthigh-resolution ALMA continuum images and uv -amp plotsof ASXDF1100.013.1, 27.1, 45.1, 45.2, 49.1 and 53.1, whichare covered by long-baseline data individually. Moreoverin order to check the properties of fainter sources with bet-ter signal-to-noise — ASXDF1100.027.1, 045.1, 045.2 and049.1 — we stacked the visibility data of these four sourcesusing the CASA code, STACKER (Lindroos et al. 2015). Wealso stacked the six sources with long-baseline data to checktheir average properties. Hereafter we refer to the former andlatter stacked data as stacked faint and stacked all .First, uv -amp plots of ASXDF1100.013.1, 27.1, 45.1, 45.2,49.1 and 53.1 and stacked faint and stacked all demonstratethat estimating their size by uv -amp analysis using up to400 k λ yields results consistent with their long-baseline vis-ibility data up to 1200 k λ (Figure 4).Second, we created high-angular-resolution ALMA mil-limeter continuum images (hereafter high-res images) using200–1200 k λ baselines (Figure 4). The resultant synthesizedbeam is 0 ′′ .23 × ′′ .19 (PA = 21 ◦ ). The r.m.s. of the imagesof ASXDF1100.053.1, 13.1, 27.1, 45.1, 45.2, 49.1, stackedfaint and stacked all are 126, 124, 124, 122, 122, 126, 88 and67 µ Jy beam - , respectively. The sources are detected with S / N peak =11, 9, 7, 7, 6, 6, 10 and 15 σ , respectively. We mea-sured their millimeter sizes and fluxes in the image using aCASA task, IMFIT : 0 ′′ .16-0 ′′ .29 (major axis). As suggestedby our source size measurements via uv -amp fitting, each ofour z & uv -distances <
200 k λ in order tosharpen the synthesized beam, so we need to take into accountany missing flux. In this paper, then, we compare fluxes mea-sured using IMFIT on the high-res ALMA images with fluxesexpected at a uv -distance of ≥
200 k λ . We adopt the fluxesmeasured at a uv -distance of 200 k λ (Figure 4) as the flux ex-pected for the measured size. Figure 7 shows the compari-son between the flux measured at a uv -distance of ≥
200 k λ and flux measured via IMFIT . The comparison shows that thefluxes measured by
IMFIT are almost the same as the fluxesmeasured using the visibilities. The relation between fluxesfrom the image and from the visibility data can be fit by F image , λ = 1 . + . - . × F visibility , λ , (1)where F image , λ is the flux measured by IMFIT and F visibility , λ is that measured from the visibilities. This in-dicates that ∼
100 percent of the rest-frame FIR emission in z & IMFIT are shown on the high-res ALMA images in Fig-ure 4, and tend to show ellipticity, i.e., minor/major axis ratioof a fitted asymmetric Gaussian <
1; the minor/major axis ra-tio of ASXDF1100.053.1, 013.1, 027.1, 045.1 and 045.2 are0.5 ± ± ± ± ± S / N peak ∼
10 (Figure 4). In order to test whetherthe elliptical feature is real, we checked the empirical accu-racy of
IMFIT using Monte Carlo simulations for major/minoraxis ratios of 0.5, 0.6, 0.7, 0.8, 0.9 and 1.0 and circularizedsizes of 0 ′′ .20, 0 ′′ .30 and 0 ′′ .40 (FWHM) and S / N peak ∼ IMFIT tends to return major/minor axis ratio of < IMFIT also tends to report smallersizes for larger input circularized sizes such as 0 ′′ .30 and0 ′′ .40. The latter is partly because of missing flux in our high-res images and partly because their detection of S / N peak ∼ IMFIT on thestacked images are not inconsistent with symmetric Gaussianemission (minor/major axis ratio = 1). Given the measuredellipticity for the stacked all and stacked faint are 0.7 and 0.5,respectively, and their R c , e are 0 ′′ .24 and 0 ′′ .17, respectively,the simulations indicate that they can in fact be symmetricGaussians. When we see the input-output ellipticity plot for R c , e = 0 ′′ .20 (bottom in Figure 8), the measured ellipticity forthe stacked faint looks off 1 σ error of the ellipticity of 1 butwith in 1.4 σ .Next, we investigate individual ASXDF sources with ∼ σ detections in the high-res ALMA image, ASXDF1100.053.1and 013.1. ASXDF1100.053.1 has a size estimated via its vis-ibility data of 0 ′′ .28 and shows an ellipticity of 0.5 via IMFIT .Because of our simulations we cannot exclude the possibilityof symmetric Gaussian emission in ASXDF1100.053.1, butan ellipticity of . . R c , e = 0 ′′ .30 (bottom in Figure 8), themeasured ellipticity of ASXDF1100.053.1 has a probabilityof only 1.3 per cent that ASXDF1100.053.1 has ellipticity of ≥ .
8. Thus ASXDF1100.053.1 has ellipticity of . . Figure 4.
Size measurements for six of the sources with long-baseline data (400–1200 k λ ) in our sample and stacked visibility data. Stacked (all) includes all ofASXDF sources with long-baseline data (ASXDF1100.013.1, 27.1, 45.1, 45.2, 49.1 and 53.1). Stacked (faint) includes faint ASXDF sources with long-baselinedata (ASXDF1100.027.1, 45.1, 45.2 and 49.1). ( Left: ) Black and grey points show the uv visibilities up to 400 and 1200 k λ , respectively. The black line is a uv -amp model of the best-fitted Gaussian component. The blue line and shaded area are possible solutions for the corrected source size, with errors listed inTable 1. The blue line and shaded area are plotted for the total amplitude of the best-fitted model. ( Right: ) ALMA 1100- µ m continuum images with synthesizedbeam sizes of ∼ ′′ .2, generated by using 200–1200 k λ data. The r.m.s. in images of ASXDF1100.053.1, 13.1, 27.1, 45.1, 45.2, 49.1 and stacked faint and stackedall are 126, 124, 124, 122, 122, 126, 88 and 67 µ Jy beam - , respectively. Contours are shown at +4 σ and +8 σ . These uv -amp plots and high-angular-resolutionimages using ≤ λ data imply that these sources have a single, compact component, as shown by our source size analysis using ≤ λ data. Ikarashi et al.
Figure 4.
Continued ompact starbursts in high-redshift SMGs 7sibly. Moving on to ASXDF1100.013.1, there is not a mocksource with the size of ASXDF1100.013.1, 0 ′′ .16 × ′′
14, inthe simulation for an input circularized size of 0 ′′ .30. Thisimplies that ASXDF1100.013.1 may have a starburst regionmore concentrated in the center than a Gaussian profile. Theseshapes that are unlikely to be symmetric Gaussians may begiving us a hint of complex star-forming structure in the smallemission area. ARE Z ∼ Z . R c , e ) of starburst nuclei in z ∼ + . - . kpc(Table 1). In the conversion from intrinsic angular source sizeto physical scale, we assume uniform redshift probability at z = 3–6 for sources without photo- z and within 1 σ error ofphoto- z for the sources with photo- z . Figure 6 reveals thatour measured sizes are more than a factor 2 × smaller thanthose of SMGs at z ∼ ′′ .5 or 2.5 kpcas measured via radio continuum – Biggs & Ivison 2008 –or ∼ ′′ . J CO emission –Tacconi et al. 2006). These radio and CO sizes were mea-sured by Gaussian fitting, as with our measurements, so thecomparison is fair. Errors in calibration cause smearing ininterferometric data, but we have no reason to suspect thatthese larger measurements are due to flawed calibration. Ifradio and (sub)millimeter continuum and CO line emissiontrace star-forming regions, Figure 6 implies that z ∼ z & z ∼ z ∼ > µ m) source sizemeasurements by ALMA of SCUBA2 sources including 23SMGs with > σ detections covering z ∼ - z ∼
3; me-dian) is also reported (Simpson et al. 2015). Their ALMAdata were taken by an array configuration similar to ourSchedule Block 1 yielding a median synthesized beam of0 ′′ .35 × ′′ .25 with the benefit of shorter observing wave-length than ours. Their sample consists mainly of SMGs withoptical/NIR-detections and photo- z by optical/NIR data con-trary to our sample consisting mainly of SMGs faint at opti-cal/NIR wavelength. Their sample is typically twice brighter(5.7 × L ⊙ ; median) than our sample in infrared luminos-ity. They report a median size of 0 ′′ .30 ± Figure 5. (Top) Relationship between ‘raw’ measured sizes from fitting in uv -amp plot and ‘actual’ sizes derived by our Monte Carlo simulation in noisevisibility data for ALMA sources with 10 and 15 σ ALMA continuum detec-tions. Grey dots mark mock sources with 15 σ detections. This plot showshow the input size for mock sources compares with the size measured byfitting to the uv -amp plot. Error bars show 1 σ for the input source size distri-butions. The plot indicates that measurements for low signal-to-noise sourcesget less effective at . ′′ .10, requiring larger corrections. (Bottom) Relation-ship between intrinsic minor/major axis ratio and measured circularized sizeby uv -amp plot based on another simulation. We plot adding offset of - + θ min / θ maj for visibility. FWHM) and 1.2 ± R e ) by Gaussian fitting in imagesnot visibilities. The median in SCUBA2 sample ( R e = 1.2 kpc)seems to be ∼ × . R c , e =0.67 kpc). However, we need to take into account the fact thatthey measured R e of major axis and we measured circularized R c , e . Then we can not compare our sizes with theirs morein details here, but both of our sample and SCUBA2 sources Ikarashi et al. Figure 6.
Size distribution of z & µ m, in comparison with the radio sizes (Biggs & Ivison2008) and CO emission-line sizes (Tacconi et al. 2006) of z ∼ z & z ∼ show smaller FIR continuum sizes of star forming region ofSMGs than the previous radio and CO sizes in spite of thedifferent luminosity and redshift distributions in the two sam-ples. In order to reveal the possible relation in FIR-continuumsize and redshift (and L IR ), we need higher-angular resolutionimaging of SMGs with various properties by ALMA. Z & Z ∼ z ∼ z ∼ z ∼ z & z & z & z & z & z ∼ z ∼ z ∼ Figure 7.
Comparison between flux at uv -distance ≥
200 k λ expected fromthe size by uv -amplitude plot and flux measured by imfit on the high-res im-age. Error in flux by visibility comes from the measured size uncertaintyshown in Figure 4. Error in flux by IMFIT is output by imfit. The dashed grayline shows flux by
IMFIT = 1.07 × flux by visibility. ing the idea that compact dust-obscured starbursts in z ∼ z ∼ z & z &
3. Our results make the evolution-ary scenario suggested by Toft et al. (2014) more plausible.Given that z ∼ z ∼ SURFACE STAR FORMATION RATE DENSITY OF Z & Σ SFR ) ishelpful to understand the origin of the compact but huge star-formation activity in z & Σ SFR of oursample using estimated R c , e (Table 1). Σ SFR of our sampleare in the range of ∼ ⊙ yr - kpc - with a medianof 100 + - M ⊙ yr - kpc - . We can find that ASXDF sourceswith a millimeter size of ∼ ′′ .10 (FWHM) show large un-certainty in their Σ SFR (Table 1). These large uncertainties inompact starbursts in high-redshift SMGs 9
Figure 8. (Top) Relationship between intrinsic minor/major axis ratio andmeasured major axis size by
IMFIT on high-res ALMA image by the MonteCarlo simulation. Dashed colored curves are expected sizes of major axisfrom input visibility model for each size. (Middle) Relationship between in-trinsic minor/major axis ratio and measured minor axis size by
IMFIT on high-res ALMA image by the simulation. Dashed colored curves are expectedsizes of minor axis from input visibility model for each size. (Bottom) Rela-tionship between intrinsic minor/major axis ratio and measured major/minoraxis ratio by
IMFIT on high-res ALMA image by the simulation. We plotadding offset of - + θ min / θ maj for visibility. Σ SFR come from the large fraction of their size errors to theirmillimeter sizes which contribute to Σ SFR by ∝ R - , e .First we compare with local galaxies. Given that Σ SFR for local merger-driven (U)LIRGs is 5–4500 M ⊙ yr - kpc - with a median of 29 + - M ⊙ yr - kpc - and that Σ SFR forlocal disks is 0.01–1 M ⊙ yr - kpc - with a median of0.04 + . - . M ⊙ yr - kpc - (Rujopakarn et al. 2011) , z & Σ SFR distributions of high- z SMGs and local (U)LIRGs couldarise by chance, and thus the two distribution are consistentwith a significant level of 3.5 percent. The range of the in-frared luminosities of local (U)LIRG sample is 10 . - . L ⊙ which is a little bit fainter than that of our sample. It isworth mentioning that a brighter half of local (U)LIRGs with10 . - . L ⊙ shows more similar Σ SFR distribution to thatof high- z SMGs; a KS test gives a probability of 17.1 per-cent that the differences could arise by chance indicating thatthe Σ SFR distributions of local brighter (U)LIRGs and high- z SMGs are consistent with a significant level of 17.1 percent.On the other hand, the remaining fainter local LIRGs show aless similar Σ SFR to high- z SMGs; a KS test shows a proba-bility of 0 percent that the differences could arise by chance.A KS test also shows that Σ SFR distributions of local disksand high- z SMGs do not match with a probability of 0 indi-cating that star-formation of high- z SMGs is different fromthat in local disks. Next we compare with high- z extendeddisk galaxies. We took four BzK galaxies at z ∼ J =2–1 sizes from Daddi et al. (2010) and 14 high-redshift diskgalaxies at z ∼ J =3–2 sizes from Tacconi et al.(2013) as our sample of high-redshift disk galaxies. TheseCO sizes have been derived assuming Gaussian profiles. Thesample of 18 high- z disk galaxies is among 42 disk galaxieswith CO detections in Tacconi et al. (2013) including 6 BzKsby Daddi et al. (2010). The remaining 24 sources tend to beobserved by low angular resolutions and are unresolved in COimages. Here we adopt the distribution of CO sizes of the 18high- z disk galaxies as the representative of the 42 sourcesbecause of following facts. Tacconi et al. (2013) shows thata ratio of R e (optical)/ R e (CO) is = 1 . ± .
06. We checkedthat optical size distributions of the high- z disk galaxies withCO sizes and those without CO sizes are consistent with asignificant level of 99.7 percent by a KS test. The SFRsof these galaxies are ∼ ⊙ yr - and their sizes ( R e )range within 2–12 kpc. The distribution of Σ SFR for the high-redshift disk galaxies ranges 0.1-7.0 M ⊙ yr - kpc - with a me-dian of 0.5 + . - . M ⊙ yr - kpc - . A KS test shows that there isa probability of 0 percent that the differences between Σ SFR distributions of high- z SMGs and the high- z extended diskscould arise by chance.We should take into account a possible difference in COand (sub)millimeter sizes in high- z SMGs and star-forminggalaxies. However, we do not know a size correction factor ofCO/(sub)millimeter sizes at this moment. Then we adopt hereCO/(sub)millimeter size ratio of 2.9 and 1.7 derived from COsizes of SMGs in Tacconi et al. (2006) and millimeter sizesin our ASXDF sources, and submillimeter sizes of SMGsin Simpson et al. (2014), respectively. The estimated Σ SFR Rujopakarn et al. (2011) measured the sizes of local galaxies after con-volving high-resolution images to compare with high-redshift sources. Wederived Σ SFR from the surface infrared luminosity densities of local galaxiesin Rujopakarn et al. (2011) for a Chabrier IMF (Chabrier 2003).
Figure 9.
Relationship between redshift and sizes for z & z ∼ z ∼ µ m size – that of the starburst nuclei– for z & z & ∼ .
05 kpc) (Barro et al. 2014b). for the size correction factor of 2.9 is 0.8–59 M ⊙ yr - kpc - with a median of 4 M ⊙ yr - kpc - and one for the size correc-tion factor of 1.7 is 0.3–20 M ⊙ yr - kpc - with a median of1 M ⊙ yr - kpc - . Then the z & & × larger Σ SFR than local and high- z disks. KS testsgives probabilities of 0 percent that Σ SFR of high- z extendeddisks with both of the size correction factor of 2.9 and 1.7 areconsistent with that of high- z SMGs.Figure 10 shows the distributions of Σ SFR of our sample incomparison with local (U)LIRGs, extended disks in high andlow redshifts. We derived the expected distribution of Σ SFR of z & Σ SFR of z & Σ SFR sim-ilar to local (U)LIRGs rather than those of extended disks athigh and low redshifts. Even though the compact star-formingregion could consist of more compact clumps spreading in ≤ Σ SFR of z & z & z & z & z that would be seen as SMGs (Cowley et al.2015). Older versions of the same semi-analytic model ar-gued for merger-driven starbursts as the main formation chan-nel of SMGs (Baugh et al. 2005). The predicted sizes result-ing in starbursts from both disk instabilities and mergers inthe GALFORM are in the range ∼ . - R e ). However,the simplicity of the angular momentum evolution models ap-plied to semi-analytic model prevents us from ruling out thepossibility that the high SFR surface densities can only beachieved during galaxy mergers. Thus at this stage we cannotdistinguish the trigger of starbursts of z & z ∼ z SMGs.Regardless of the triggering mechanism of the z & z & z ∼ SUMMARYWe have exploited new ALMA 1100- µ m continuum datato measure the size of dusty, starburst regions in a sample of13 SMGs at z &
3. The radii of z ∼ L IR ∼ × L ⊙ ranges from 0 ′′ .10 to 0 ′′ .38 (FWHM) with aompact starbursts in high-redshift SMGs 11 Figure 10.
Surface star fomation rate density ( Σ SFR ) distribution of z & z extended disks (Daddi et al. 2010; Tacconi et al. 2013).Here, we have not applied any correction of a possible size difference be-tween CO line and (sub)millimeter continuum sizes in high- z disks. We de-rived the expected distribution of Σ SFR of z & Σ SFR of z & z & Σ SFR distribution similar to local (U)LIRGs. median of 0 ′′ .20 + . - . , corresponding to a median circularizedeffective radius R c , e of 0.67 + . - . kpc. Our results demonstratethat the star-forming regions of z & × smaller than measured previously using radio continuumand CO data for z ∼ z & z & z ∼
2, meaning thatwe can now trace the evolutionary path of the most massivegalaxies over a period encompassing ∼
90% of the age of theUniverse.This paper makes use of the following ALMA data:ADS/JAO.ALMA
Facilities:
ALMA, ASTE.
REFERENCESBarro, G., Faber, S. M., Pérez-González, P. G., et al. 2014a, ApJ, 791, 52Barro, G., Trump, J. R., Koo, D. C., et al. 2014b, ApJ, 795, 145Baugh, C. M., Lacey, C. G., Frenk, C. S., et al. 2005, MNRAS, 356, 1191Bezanson, R., van Dokkum, P. G., Tal, T., et al. 2009, ApJ, 697, 1290Biggs, A. D., & Ivison, R. J. 2008, MNRAS, 385, 893Blain, A. W., Chapman, S. C., Smail, I., & Ivison, R. 2004, ApJ, 611, 725Carollo, C. M., Bschorr, T. J., Renzini, A., et al. 2013, ApJ, 773, 112Caputi, K. I., Dunlop, J. S., McLure, R. J., et al. 2012, ApJ, 750, LL20Carilli, C. L., & Yun, M. S. 1999, ApJ, 513, L13Ceverino, D., Dekel, A., & Bournaud, F. 2010, MNRAS, 404, 2151Chabrier, G. 2003, PASP, 115, 763Chapman, S. C., Blain, A. W., Smail, I., & Ivison, R. J. 2005, ApJ, 622, 772Chen, C.-C., Smail, I., Swinbank, A. M., et al. 2015, ApJ, 799, 194Cowley, W. I., Lacey, C. G., Baugh, C. M., & Cole, S. 2015, MNRAS, 446,1784Lagos, C. d. P., Bayet, E., Baugh, C. M., et al. 2012, MNRAS, 426, 2142Daddi, E., Bournaud, F., Walter, F., et al. 2010, ApJ, 713, 686Daddi, E., Renzini, A., Pirzkal, N., et al. 2005, ApJ, 626, 680Dekel, A., Sari, R., & Ceverino, D. 2009, ApJ, 703, 785Elbaz, D., Dickinson, M., Hwang, H. S., et al. 2011, A&A, 533, AA119Gonzalez-Perez, V., Lacey, C. G., Baugh, C. M., et al. 2014, MNRAS, 439,264Hatsukade, B., Kohno, K., Aretxaga, I., et al. 2011, MNRAS, 411, 102Hughes, D. H., Aretxaga, I., Chapin, E. L., et al. 2002, MNRAS, 335, 871Hughes, D. H., Serjeant, S., Dunlop, J., et al. 1998, Nature, 394, 241Ikarashi, S., Kohno, K., Aretxaga, I., et al. 2013, New Trends in RadioAstronomy in the ALMA Era: The 30th Anniversary of Nobeyama RadioObservatory, 476, 265Ikarashi, S., Kohno, K., Aguirre, J. E., et al. 2011, MNRAS, 415, 3081Iono, D., Peck, A. B., Pope, A., et al. 2006, ApJ, 640, L1Ivison, R. J., Smail, I., Papadopoulos, P. P., et al. 2010, MNRAS, 404, 198Ivison, R. J., Greve, T. R., Dunlop, J. S., et al. 2007, MNRAS, 380, 199Ivison, R. J., Smail, I., Le Borgne, J.-F., et al. 1998, MNRAS, 298, 583Krogager, J.-K., Zirm, A. W., Toft, S., Man, A., & Brammer, G. 2014, ApJ,797, 17Lindroos, L., Knudsen, K. K., Vlemmings, W., Conway, J., & Martí-Vidal, I.2015, MNRAS, 446, 3502Nelson, E., van Dokkum, P., Franx, M., et al. 2014, Nature, 513, 394Newman, A. B., Ellis, R. S., Bundy, K., & Treu, T. 2012, ApJ, 746, 162Oliver, S. J., Bock, J., Altieri, B., et al. 2012, MNRAS, 424, 1614Onodera, M., Daddi, E., Gobat, R., et al. 2010, ApJ, 715, L6Oser, L., Naab, T., Ostriker, J. P., & Johansson, P. H. 2012, ApJ, 744, 63Riechers, D. A., Bradford, C. M., Clements, D. L., et al. 2013, Nature, 496,329Rujopakarn, W., Rieke, G. H., Eisenstein, D. J., & Juneau, S. 2011, ApJ,726, 93Sakamoto, K., Wang, J., Wiedner, M. C., et al. 2008, ApJ, 684, 957Sakamoto, K., Ho, P. T. P., Iono, D., et al. 2006, ApJ, 636, 685Sakamoto, K., Scoville, N. Z., Yun, M. S., et al. 1999, ApJ, 514, 68Simpson, J. M., Smail, I., Swinbank, A. M., et al. 2015, ApJ, 799, 81Simpson, J. M., Swinbank, A. M., Smail, I., et al. 2014, ApJ, 788, 125Smail, I., Ivison, R. J., & Blain, A. W., 1997, ApJ, 490, L5Swinbank, A. M., Simpson, J. M., Smail, I., et al. 2014, MNRAS, 438, 1267Tacconi, L. J., Neri, R., Genzel, R., et al. 2013, ApJ, 768, 74Tacconi, L. J., Genzel, R., Smail, I., et al. 2008, ApJ, 680, 246Tacconi, L. J., Neri, R., Chapman, S. C., et al. 2006, ApJ, 640, 228Targett, T. A., Dunlop, J. S., Cirasuolo, M., et al. 2013, MNRAS, 432, 2012Toft, S., Smolˇci´c, V., Magnelli, B., et al. 2014, ApJ, 782, 68van Dokkum, P. G., Franx, M., Kriek, M., et al. 2008, ApJ, 677, L5Wuyts, S., Cox, T. J., Hayward, C. C., et al. 2010, ApJ, 722, 1666Wilson, C. D., Rangwala, N., Glenn, J., et al. 2014, ApJ, 789, L36Wilson, C. D., Petitpas, G. R., Iono, D., et al. 2008, ApJS, 178, 189Younger, J. D., Fazio, G. G., Wilner, D. J., et al. 2008, ApJ, 688, 59
Table 1
Summary of ASXDF source size measurements.
Name R.A. Dec. SNR S µ m SFR † L † IR Photo Size (FWHM) R c , e ‡ Σ SFR z raw corrected(J2000) (J2000) (mJy) (M ⊙ yr - ) (10 L ⊙ ) (arcsec) (arcsec) (kpc) (M ⊙ yr - kpc )Schedule Block 1 (covering 1200 k λ )ASXDF1100.013.1 02:16:45.86 - ± + - + . - . > + . - . + . - . + - ASXDF1100.027.1 02:17:20.95 - ± + - + . - . + . - . + . - . + . - . + - ASXDF1100.045.1 02:18:16.04 - ± + - + . - . > < ∗ + . - . + . - . + - ASXDF1100.045.2 02:18:14.89 - ± + - + . - . > + . - . + . - . + - ASXDF1100.049.1 02:17:32.86 - ± + - + . - . > + . - . + . - . + - ASXDF1100.053.1 02:16:48.20 - ± + - + . - . > + . - . + . - . + - Stacked faint (ASXDF1100.027.1, 45.1, 45.2 and 49.1) 25.0 1.90 ± + - + . - . — 0.170 0.18 + . - . + . - . + - Stacked all (faint + ASXDF1100.013.1 and 053.1) 31.0 2.37 ± + - + . - . — 0.240 0.24 + . - . + . - . + - Schedule Block 2, 3 (covering 400 k λ )ASXDF1100.073.1 02:18:10.04 - ± + - + . - . > + . - . + . - . + - ASXDF1100.083.1 02:17:12.42 - ± + - + . - . > + . - . + . - . + - ASXDF1100.090.1 02:17:23.04 - ± + - + . - . > < ∗ + . - . + . - . + - ASXDF1100.110.1 02:17:43.59 - ± + - + . - . + . - . < ∗ + . - . + . - . + - ASXDF1100.127.1 02:17:33.36 - ± + - + . - . > + . - . + . - . + - ASXDF1100.230.1 02:17:59.39 - ± + - + . - . + . - . < ∗ + . - . + . - . + - ASXDF1100.231.1 02:17:59.65 - ± + - + . - . > + . - . + . - . + - All values in this table are measured on ALMA data not on AzTEC data. † L IR and SFR assume average SED of ALMA-identified SMGs (Swinbank et al. 2014), with a Chabrier IMF (Chabrier 2003). We also assume uniform redshiftprobability at z = 3–6 for sources without photo- z , and in 1 σ error for sources with photo- z . ‡ R c , e is derived from the half width at half maximum (HWHM) assuming a symmetric Gaussian profile. HWHM corresponds to R c , e in a symmetric Gaussianprofile. We also assume a same redshift probability as we do for L IR and SFR. * Our fitting stops at 0 ′′′′