Mapping the large scale structure around a z=1.46 galaxy cluster in 3-D using two adjacent narrow-band filters

1.4 MB·

by  Masao Hayashi

Mon.Not.R.Astron.Soc.000,1–14(2014) Printed17January2014 (MNLATEXstylefilev2.2)
Mapping the large scale structure around a z=1.46 galaxy
cluster in 3-D using two adjacent narrow-band filters
Masao Hayashi1⋆, Tadayuki Kodama2,4, Yusei Koyama2, Ken-ichi Tadaki2,
Ichi Tanaka3, Rhythm Shimakawa3,4, Yuichi Matsuda4,5, David Sobral6,7,
4 Philip N. Best8, Ian Smail9
1 1Institute for Cosmic Ray Research, The Universityof Tokyo, Kashiwa, Chiba 277-8582, Japan
0 2Optical and Infrared Astronomy Division, National Astronomical Observatory, Mitaka, Tokyo 181-8588, Japan
2 3Subaru Telescope, National Astronomical Observatory of Japan, 650 North A’ohoku Place, Hilo, HI 96720, USA
n 4Department of Astronomical Science, The Graduate Universityfor Advanced Studies(SOKENDAI), Mitaka, Tokyo 181-8588, Japan
a 5Radio Astronomy Division, National Astronomical Observatory of Japan, 2-21-1 Osawa, Mitaka, Tokyo 181-8588, Japan
J 6Leiden Observatory,Leiden University,P.O. Box 9513, NL-2300 RA Leiden, THe Netherlands
7Centro de Astronomia e Astrof´ısica da Universidade de Lisboa, Observat´orio Astrono´mico de Lisboa,
6
Tapada da Ajuda, 1349-018 Lisboa, Portugal
1
8SUPA, Institute for Astronomy, Royal Observatory of Edinburgh, Blackford Hill, Edinburgh, EH9 3HJ, UK

9Institute for Computational Cosmology, Durham University,South Road, Durham, DH1 3LE, UK
]
O
C
.
h Accepted 2014January15.Received2014January7;inoriginalform2013November7
p
-
o
ABSTRACT
r
t We present a novel method to estimate accurate redshifts of star-forming galaxies
s
by measuring the flux ratio of the same emission line observed through two adjacent
a
[ narrow-bandfilters. We apply this method to our NB912 and new NB921 data taken

with Suprime-Cam on the Subaru Telescope of a galaxy cluster, XMMXCS J2215.9-
1
1738, at z = 1.46 and its surrounding structures. We obtain redshifts for 170 [OII]
v emission line galaxies at z ∼ 1.46, among which 41 galaxies are spectroscopically
9
confirmed with MOIRCS and FMOS on the Subaru mainly, showing an accuracy of
1
σ((z−z )/(1+z ))=0.002.This allows us to revealfilamentarystructures that
9 spec spec
3 penetratetowardsthecentreofthegalaxyclusterandintersectwithotherstructures,
. consistentwiththepictureofhierarchicalclusterformation.Wealsofindthatthepro-
1
jected celestial distribution does not precisely trace the real distribution of galaxies,
0
indicatingtheimportanceofthethreedimensionalviewofstructurestoproperlyiden-
4
tify and quantify galaxy environments. We investigate the environmentaldependence
1
: of galaxy properties with local density, confirming that the median colour of galaxies
v
becomes redder in higher density region while the star-formationrate of star-forming
i
X galaxiesdoesnotdependstronglyonlocalenvironmentinthis structure.Thisimplies
that the star-forming activity in galaxies is truncated on a relatively short time scale
r
a in the cluster centre.
Key words: galaxies: clusters: general – galaxies: evolution.
1 INTRODUCTION (2dFGRS,Colless et al.2001).Galaxyclustersembeddedin
thelarge-scalestructuresaresomeofthedensestsystemsin
Galaxiesaredistributedinhomogeneouslyonsmallscalesin
the Universe and occur at intersections of the galaxy fil-
the Universe and thus define large-scale structures which
aments. While the spatial distribution of galaxies has been
consist of dense regions of galaxies such as galaxy clus-
revealedindetailinthelocalUniverse,theprocessbywhich
ters, sparse regions called voids, and filamentary structures
thegalaxy clusters form is yet to be understood in detail.
which connect the dense regions, as revealed by large spec-
troscopic surveys such as Sloan Digital Sky Survey (SDSS, Galaxy clusters are considered to originate from the
York et al.2000)and2degreeFieldGalaxyRedshiftSurvey clusteringof some younggalaxies (i.e., proto-cluster) in the
early Universe and to gradually grow into more massive
gravitationally bound systems with time while accreting
⋆ E-mail:[email protected] galaxies from the surrounding regions. The standard cold
(cid:13)c 2014RAS
2 M. Hayashi et al.
dark matter (CDM) scenario can reproduce well the pic- this galaxy cluster is a good target to investigate the pro-
ture seen in the Universe within the framework of the hier- cesses operating on galaxies assembling into the centre of
archical formation of galaxy structures (e.g., Springel et al. the cluster from the outskirts. However, our current under-
2005).Observationalstudiesthroughwide-fieldimagingsur- standingisbasedonaviewofthelargescalestructurespro-
veysofdistantclustersatz<1havealsosupportedthehier- jected on the celestial plane. We require to know the true
archical formation by finding large-scale structures around threedimensionalstructurestorevealindetailtheprocesses
galaxy clusters (e.g., Kodama et al. 2005; Gal et al. 2008; playing a critical role in formation of characteristic proper-
Tanaka et al.2009;Koyama et al.2010).Now,thediscovery ties of cluster galaxies, because the projected distribution
of large-scale structures around galaxy clusters and proto- of galaxies may not reflect the real structures. The results
clusters is being extended to z ∼ 3 (e.g., Tanaka et al. foundwithoutthethreedimensionalstructurescanbeledto
2011;Tadaki et al.2012;Hayashiet al.2012;Koyama et al. wrongconclusions.Thus,inthispaper,weaimtodefinethe
2013; Erb et al. 2011; Yamada et al. 2012; Galametz et al. threedimensionalstructuresaroundtheXCS2215clusterat
2013; Henry et al. 2014). One interesting finding suggested z=1.46.
by such observational studies is that cluster galaxies seem Theoutlineofthispaperisasfollows.Theobservations
to follow an inside-out evolution in the sense that the en- and data are described in § 2, and samples of emission line
vironment where galaxies are active shifts from the dens- galaxiesidentifiedwiththreedifferentnarrowbandfiltersare
est regions at high redshifts (z ≈ 2–3) towards less dense presentedin§3.Anewmethodtoestimatetheaccuratered-
regions at lower redshifts (z ≪ 1). Star forming activity shifts with twoadjacent narrowband filtersis introduced in
increases rapidly in the denser regions and becomes com- § 4, which allows us to reveal the three dimensional view
parable to the surrounding lower density regions at higher of the filamentary large-scale structures around the galaxy
redshifts, which implies that galaxy formation is biased to- clusterin§5.Then,wecompare[OII]emissionlinegalaxies
wardshigherdensityregionsathighredshift(Kodama et al. with Hα emission line galaxies associated with these struc-
2013,andreferencestherein).However,itisalsoknownthat tures. We also discuss the dependence of galaxy properties
theenvironmentaldependencesofgalaxy propertiesseenat ontheenvironmentdefinedbythethreedimensionaldistri-
z∼0arealreadyinplaceatz∼1(e.g.,Holden et al.2007; bution of the galaxies. Finally, we present the conclusions
van derWel et al.2007;Patel et al.2009,2011;Sobral et al. ofthispaperin§6.Throughoutthispaper,magnitudesare
2011). presented in the AB system. However, Vega magnitudes in
The next step to understand the evolution of cluster J and K, if preferred, can be obtained from our AB mag-
galaxies is todetermine precisely how galaxies changetheir nitudes using the relations: JVega = JAB−0.92 and KVega
propertiesasaresultofthehierarchicalgrowthoflarge-scale = KAB−1.90, respectively. We adopt cosmological param-
structures. However, the difficulty is the need for accurate eters of h = 0.7, Ωm = 0.3 and ΩΛ = 0.7. In z = 1.46, 1
determination of the redshift to determine the precise en- arcmin corresponds to 1.25 Mpc(co-moving) and 0.51 Mpc
vironment of the galaxies. Although photometric redshifts (physical), respectively.
estimated by fitting template spectral energy distributions
(SEDs) to the multi-wavelength photometry of galaxies is
quiteeffective,theuncertaintybecomeslargeathigherred- 2 OBSERVATIONS AND DATA
shifts. Moreover, the number density of blue star-forming
2.1 Narrowband filters targeting nebular emission
galaxies with less prominent features in the SED exceeds
lines at z=1.46
that of red quiescent galaxies which have strong breaks
in their SED at z &1.0–1.5 (e.g., Brammer et al. 2011; We used two narrowband filters which are adjacent filters
Moustakas et al. 2013; Muzzin et al. 2013; Tomczak et al. and both can detect [OII] emission at z ∼ 1.46. These are
2013).Aneffectiveway tosolvetheissueis bycarrying out the NB912 and NB921 narrowband filters installed on the
theimagingwithnarrowbandfilterstargetingnebularemis- SubaruPrimeFocusCamera(Suprime-Cam;Miyazaki et al.
sion from HIIregions of star-forming galaxies. The narrow- 2002)ontheSubaruTelescope.Thecentralwavelengthand
band imaging enables a largely unbiased sampling of star- full width half maximum (FWHM) of the NB912 (NB921)
forming galaxies down to a limiting star-formation rate of filtersare9139 (9196)˚A and134(132)˚A, whichimpliesave-
galaxy and the redshift of galaxies selected are determined locity difference of ∼ 2000 km s−1 for galaxies at z = 1.46
with the accuracy of ∆z ∼ 0.03. However, this is too un- (Figure 1). The specifications of the two narrow-band fil-
certain to accurately define the environment of galaxies as ters are shown in Table 1. Our analysis also makes use of
this redshift range corresponds to a co-moving depth of ∼ a third filter: the NB narrowband filter installed on the
H
56Mpcatz=1.46.Wethusdemonstrateanoveltechnique Wide Field Camera (WFCAM; Casali et al. 2007) on the
to circumvent thisproblem. the United Kingdom Infrared Telescope (UKIRT). This fil-
The galaxy cluster XMMXCS J2215.9-1738 (hereafter ter was made for the High Redshift Emission Line Survey
XCS2215) at z = 1.46 is one of the most distant massive (HiZELS) which is a narrowband imaging survey aiming
clustersknown(Stanford et al.2006).ExtendedX-rayemis- to detect Hα emission from galaxies at z =0.4, 0.8, 1.5,
sion from hot gas bound in the cluster suggests that it is and 2.2 with WFCAM on the UKIRT (Geach et al. 2008;
already a mature system. However, previous studies have Sobral et al.2009,2012,2013;Best et al.2010),andcande-
revealed that this galaxy cluster is still growing actively tect Hα emission at z ∼ 1.46 with the central wavelength
(Hayashiet al. 2010, 2011). We discovered a high fraction andFWHMofresponsecurvebeing1.617µmand0.0211µm.
of galaxies undergoing active star formation in the core of AsshowninFigure1,theresponsecurvesoftheNB921and
theclusterandtheexistenceoflargescalestructuresofstar- NB filters are well matched in redshift space correspond-
H
forminggalaxiesaroundthecluster.Thefindingmeansthat ing to [OII] or Hα emission lines at z ∼1.46 (Sobral et al.
(cid:13)c 2014RAS,MNRAS000,1–14
3-D structures at z = 1.46 3
Table 1.Specifications ofthethreenarrowbandfiltersandthesamplesofemissionlinegalaxies.
Filter λc ∆λ Telescope Emitter NB912+NB921[OII] LimitingSFR‡
[µm] [µm] Line Number Number Spec. confirmed† [M⊙/yr]
NB912 0.9139 0.0134 Subaru [OII] 380 2.6
170 41
NB921 0.9196 0.0132 Subaru [OII] 429 2.2
NBH 1.617 0.0211 UKIRT Hα 9 1 – 14.6
†Wespectroscopicallyconfirmed73[OII]emittersintotal;NB912+NB921=41, NB912=21,andNB921=11.
‡ThelimitingSFRsarenotcorrectedfordustextinction.
standard star, Feige110. The 5σ limiting magnitude mea-
sured within a 2′′-diameter aperture is 25.4 comparable to
thedepth of 25.2 in theexisting NB912 data (see § 2.4).
2.3 The NB and H data from HiZELS
H
As a part of HiZELS, we conducted the NB narrowband
H
imaging in the galaxy cluster. WFCAM consists of four
2048×2048HgCdTedetectorswithapixelscaleof0.4′′,and
each detector covers a 13.65′×13.65′ region. However, the
detectorsarespacedwithagapof12.83′.Thus,weobserved
the galaxy cluster in NB and H-bands by pointing one of
H
thedetectoratthecentreofcluster,resultinginthecentral
13.65′×13.65′ images being available in both bands. Note
Figure1. ResponsecurvesofNB912(blue),NB921(green)and
that the broadband image is required to estimate the con-
NBH (red) filters. Thehorizontal axis shows the redshifts where
[OII] or Hα lines enter each filter; [OII] for NB912 and NB921, tinuumflux density underneaththe Hα emission line.
and Hα for NBH. The open and hatched histograms show the Data reduction was conducted in the standard proce-
distribution of spectroscopic redshift of all cluster members and durewiththepipelineforHiZELS(PfHiZELS:Geach et al.
[OII]emittersselectedwithNB912andNB921,respectively.The 2008; Sobral et al. 2009). The integration times are 256.7
[OII] emitters are spectroscopically confirmed using MOIRCS and 49 minutes in NB and H, respectively. Since seeing
H
(Hayashietal.2011)andFMOS(§2.6)ontheSubaruTelescope. sizes were 1.02 and 1.32 arcsec in NB and H, they are
H
matched to 1.32 arcsec. The 5σ limiting magnitudes are
21.61 in NB and 22.28 in H, which are measured with
2012;Hayashi et al.2013),whiletheNB912filterissuitable H
a 2.4′′-diameter aperture.
to select [OII] emitters slightly in theforeground.
2.2 The NB921 data
The NB921 data were obtained with Suprime-Cam on the
2.4 The data set used for previous studies
SubaruTelescope, whichhasafield-of-view(FoV)of34×27
arcmin2, corresponding to 42.5×33.8 Mpc2 in co-moving Ourprevious studies in thegalaxy cluster are already pub-
scaleatz=1.46,undertheopen-useprogrammeinservice- lished in Hayashiet al. (2010, 2011) using six broadband
mode (S12A-131S, PI: M. Hayashi) on 2012 July 24. This images in B,Rc,i′,z′,J and K and theNB912 narrowband
enables us to cover from the central region to the outskirts image. Thus, we briefly summarize these data here.
oftheclusterwithasinglepointing.Wetook11frameswith Deep,wide-fieldimagingdatainopticalweretakenwith
individualexposuretimeof20minutes,whichresultsinthe Suprime-Cam on the Subaru Telescope. The PSFs in the
totalintegrationtimeof220minutes.Theobservationswere optical images are all matched to 1.09 arcsec. The 5σ lim-
conducted underphotometric conditions. The seeing varied iting magnitudes are between 25.2 and 27.0 (Hayashi et al.
between 0.6 and 0.8 arcsec. 2011). The data set in the near-infrared consists of J and
We used the data reduction package for Suprime-Cam K-band images. K-band data were taken with WFCAM
(SDFRED2: Yagi et al. 2002; Ouchiet al. 2004) to reduce on the UKIRT and overlap well with the optical images.
the data in the same way as for the other optical data However, J-band images taken with Multi-Object Infrared
(Hayashiet al.2010,2011).Notethatwealsousedthesoft- Camera and Spectrograph (MOIRCS; Ichikawaet al. 2006;
warereleasedbyYagi(2012)tocorrectforthecrosstalkseen Suzukiet al. 2008) on theSubaru telescope are available in
around bright stars in the Suprime-Cam image. The point thecentral6′×6′regiononly.PSFsinbothJ andK arealso
spread function (PSF) is matched to 1.09 arcsec, which is matchedto1.09arcsec.The5σlimitingmagnitudesare23.2
thePSFoftheotheropticaldata(see§2.4).Thephotomet- and22.7inJ andK,respectively,whicharemeasuredwith
ric zero-point is determined using the spectrophotometric a 2′′-diameter aperture(Hayashi et al. 2011).
(cid:13)c 2014RAS,MNRAS000,1–14
4 M. Hayashi et al.
2.5 Photometric catalogue no counterpart is found are cross-talk and false detection
seen around bright stars. Thus, we remove them from the
In this section, we describe the two catalogues for NB921-
catalogue. Asa result, the catalogue contains 1,324 sources
detected and NB -detected sources. The NB912-detected
H brighterthan 21.61 mag. in NB (5σ limiting magnitude).
catalogue is fully described in Hayashiet al. (2010, 2011). H
2.6 Spectroscopic data
2.5.1 NB921-detected catalogue
Hayashiet al. (2011) reported the confirmation of sixteen
A photometric catalogue for NB921-detected sources is
NB912 [OII] emitters in the central region of the galaxy
made following Hayashiet al. (2010, 2011). We use SEx-
clusterusingMOIRCSontheSubaruTelescope.Thesedata
tractor (ver. 2.8.6: Bertin & Arnouts 1996) to detect
wereusefulforshowingthevalidityofourselectionofNB912
sources on the NB921 image. Note that we use the NB921
[OII]emitters.However,thespectroscopicconfirmationwas
image before the PSF matching (i.e. FWHM of 0.81′′)
limited to [OII]emitters in the core region.
for the source detection to take advantage of the better
We have used the Fiber Multi Object Spectrograph
signal-to-noise ratio. Photometry is conducted on all the
(FMOS) on the Subaru Telescope to confirm the large
PSF-matched images by using SExtractor in dual-image
scale structure of [OII] emitters we have discovered in
mode. Colours are derived from the 2′′-diameter aperture
Hayashiet al.(2011).Itallowsformulti-objectspectroscopy
magnitudes, and mag auto magnitudes are used as total
intheJ andH-bandswith 400fibersand awideFoVof30
magnitudes. The aperture size is determined so as to be
arcmin. The FMOS spectroscopy was performed under an
about twice as large as the PSF. Magnitude errors are es-
open-use program (S12A-084, PI: T. Kodama) during one
timated from the 1σ sky noise taking account of the dif-
and half nights of 2012 June 22–23. Unfortunately, due to
ference in depth at each object position due to slightly dif-
theinstrumentaltrouble,onlytheIRS2,i.e., oneofthetwo
ferent exposuretimesand sensitivities. Magnitudesarecor-
spectrographs in FMOS, was available in our observation
rected for the Galactic absorption by the following mag-
run. Therefore, only 50% of sources were targeted, result-
nitudes; A(B)=0.10, A(R )=0.06, A(i′)=0.05, A(z′)=0.04,
c ing in sparse sampling of the targets. Targets are mainly
A(NB921)=0.04,A(J)=0.02,andA(K)=0.01,whicharede-
selected from the sample of NB912 [OII] emitters shown in
rived from the extinction law of Cardelli et al. (1989) on
Hayashiet al. (2011). We targetted 78 galaxies in total, of
the assumption of R = 3.1 and E(B−V) = 0.025 esti-
V which56areNB912[OII]emitters.Inthisrun,weusedthe
matedfromSchlegel et al.(1998).Thevaluesforthecorrec-
high resolution H-long setting which covers the wavelength
tionarethesameasusedfortheNB912-detected catalogue
of 1.6–1.8µm with a spectral resolution of R=2000, enough
of Hayashiet al. (2011). To check the NB921 zero-point,
to resolve Hα and [NII] emission lines, where present. We
wecompareNB921magnitudeswithNB912magnitudesfor
adoptedthecrossbeamswitchingmethod,thatis,halfofthe
stellar sources. We find that there is a good agreement be-
fibersavailable areallocated tothetargets andat thesame
tweenbothmagnitudes,indicatingthevalidityofzero-point
timetheotherhalfof themobservenearbyregions of blank
of magnitude in NB921 image.
sky.The total integration time was 5.5 hours. Observations
Asaresult,afterrejectingobjectsintheregionsmasked
were conducted in photometric conditions. The seeing was
due to the bad quality of the images, the catalogue con-
0.7–0.9 arcsec inR-band,whichwas measured forthecoor-
tains 40,237 sources brighter than 25.44 mag in NB921 (5σ
dinate calibration stars (so-called CCS stars in the FMOS
limiting magnitude). Among them, 36,415 galaxies are dis-
spine-to-object allocation software) near the centre of the
tinguished from 3,822 stars based on B −z′ and z′ −K
FoV,implyingthatloss of fluxescaping from the1.2-arcsec
colours. This technique was devised by Daddiet al. (2004),
diameter fibreshould besmall for unresolved objects.
and stars are actually well separated from galaxies on this
Wehadanotheropportunitytoobservethe[OII]emit-
colour-colourdiagram(B−z′vs.z′−K)(Daddiet al.2004;
ters with FMOS for about two hours on 2013 October 13
Kong et al. 2006).
(S13B-106, PI: K.-I. Tadaki). The targets are selected from
thesamplesofNB912[OII]emittersandNB921[OII]emit-
ters described below. The same setup as for the previous
2.5.2 NB -detected catalogue
H observation was used in this run. However, only one spec-
A catalogue for NB -detected sources is made by follow- trograph (IRS1) was available again. The total integration
H
ing the steps described above. Since the PSF size of the time was 1.5 hours on source. During the observations, the
NB and H images are larger than that of the others, this skywasalmostclearandtheseeingwas∼1.0arcsec.Spectra
H
catalogue is used separately to select the NBH Hα emit- were obtained for 65 [OII] emitters.
ters at z ∼ 1.46 in § 3. The NB image with PSF of 1.02′′ We used the FIBRE-pac data reduction package
H
is used for source detection, and photometry is conducted (Iwamuro et al. 2012) to reduce the FMOS data. FIBRE-
on the PSF-matched images using the dual-image mode of pac corrects the flux for the loss from the fiber which is
SExtractor. The H–NB colour is obtained from 2.4′′- estimated using a calibration star. Although further cor-
H
diameteraperturemagnitudes.Magnitudesarecorrectedfor rection may be required for extended sources like galaxies
the Galactic absorption by A(NB )=0.01 and A(H)=0.01. (Yabeet al. 2012; Stott et al. 2013), the redshift and ratio
H
TocompilethephotometryforallofthedataexceptH and of emission line fluxes can be measured reliably even if the
NB , we cross-match the NB -detected sources with those correctionisnotperformed.WefittedthreeGaussianstothe
H H
detected byNB912 orNB921. For almost all NBH-detected spectratomeasureHαandthe[NII]doubletandestimated
sources,counterpartsarefoundintheNB912orNB921cat- thesignal-to-noiseratioofemissionlinesseeninthespectra,
alogues. We note that the NB -detected sources for which and the redshift. In the Gaussian fitting, the width of the
H
(cid:13)c 2014RAS,MNRAS000,1–14
3-D structures at z = 1.46 5
Figure 2. Top: Broadband - Narrowband colours as a function of Narrowband magnitude (aperture magnitude). Bottom: two colour
diagrams used to select emissionline galaxies at z =1.46. In the left two panels for NB921 [OII] emitters, emission line galaxies (un-
)detected in K-band are shown by (open) filled red circles and then [OII] emitters identified are shown by blue filled squares. Gray
dots are all galaxies in the NB921-detected catalogue. In the left lower panel, solid black lines are our colour selection criteria defined
in Hayashietal. (2011), while dotted lines shows the original criteria for the selection of BzK galaxies (Daddietal. 2004). The right
twopanels arethe sameas theleftones, butforNBH Hαemitters.Bluesymbolsmarkedwithopensquares showHαemittershaving
adetection of[OII]emissioninNB912orNB921data.
twoGaussiansfittedtothe[NII]doubletisfixedtothesame the galaxies confirmed by our Subaru observations to the
value as Hα. In addition, the flux ratio of [NII] doublet is 44membergalaxiesinHilton et al.(2010),andsevengalax-
fixedto one-third.Errors on theemission line fluxesarees- ies turn out to be common objects. We checked that we
timated using the standard deviation of 500 measurements do not find any disagreement in the redshifts of the com-
oftheflux,whereineachmeasurement asetofthreeGaus- mon sources. Therefore, 101 galaxies are spectroscopically
siansisfittedtothespectrumtowhichtheerrorwithnormal confirmedtobeassociatedwiththestructureinandaround
distribution with 1σ spectral noiseis added.Weregard Hα thegalaxycluster.Theredshiftdistributionoftheconfirmed
emission with fluxgreater than3σ as adetection, and then galaxies is shown in Figure 1.
check thespectra judged to have a line detection, by visual
inspection. We succeed in spectroscopically confirming 48
sources by detecting Hα emission at more than 3σ in the
whole region we surveyed, among which the number of the 3 SELECTION OF EMISSION LINE GALAXIES
[OII] emitters is 41. Even if only a single emission line is
In this section, we select emission line galaxies using the
detected in the spectrum, it can be identified as being Hα,
photometriccatalogues.Wehavealreadypublishedthesam-
because we have already detected [OII] emission with our
ple of [OII] emitters at z ∼ 1.46 based on the NB912 data
narrowband imaging. The FMOS spectra demonstrate that
(Hayashiet al. 2011). Since there is only a slight difference
our selection of NB912 and NB921 [OII] emitters is valid
betweenNB912 andNB921 in thecentralwavelength ofre-
and that the structure we have found around the cluster is
sponse curve, we can apply the same method for the se-
real.
lection ofNB921[OII]emittersasforNB912 [OII]emitters
Wealsohaveadditionalspectroscopicredshiftsfromthe summarisedbelow.WereferreaderstoHayashiet al.(2011)
literature. Hilton et al. (2010) have spectroscopically con- for more details, if needed.
firmed44membergalaxies inthiscluster.Wecross-identify First,weselect galaxieswithcolourexcessinz′-NB921
(cid:13)c 2014RAS,MNRAS000,1–14
6 M. Hayashi et al.
Figure 3. CelestialdistributionofNB912/NB921[OII]emittersandNBH Hαemitters.Bluedotsshow[OII]emittersidentifiedfrom
NB912data(i.e.,NB912[OII]),whilereddotsshowthosefromNB921data(i.e.,NB921[OII]).Greendotsare[OII]emittersidentified
fromboth NB912and NB921data (i.e.,NB912+NB921 [OII]).Solidopen squares show NBH Hαemitters,followingthesamecolour-
codingasfor[OII]emitters(e.g.,theHαemittersidentifiedasNB921[OII]emittersaswellareshownbyredopensquares).Blackopen
squares indicate Hα emitters without any [OII] emission detected. The gray regions are masked due to the bad quality of the images,
and the dotted square shows the FoV of a single detector of WFCAM on UKIRT (see also the text in § 2.3). The axes shows relative
coordinatesfromthecentreofthegalaxycluster.AsalreadyfoundinHayashietal.(2011),wecanconfirmagaintheclusteringof[OII]
emittersinthecentreandthelargescalestructureexistinginthesurroundingregionevenwhentheNB921dataareused.
at more than 3σ (Figure 2). This corresponds to galaxies that the number of NB921 [OII] emitters is slightly larger
with a line flux larger than 1.2×10−17 erg s−1 cm−2 being thanthatofNB912[OII]emittersevenifthesamelimiting
selected. If the excess is due to the [OII] emission line at [OII]fluxisadoptedbetweenthetwosamples.However,the
z = 1.46, the limiting flux corresponds to a dust-free SFR numbers of [OII] emitters selected are in agreement within
of 2.2 M⊙ yr−1 according to the [OII]–SFR calibration in a 1 σ error based on Poisson statistics. The small differ-
Kennicutt(1998)whereaSalpeter(1955)initialmassfunc- ence can appear because of thestructures and clustering of
tion (IMF) is assumed. To exclude the possible contamina- galaxies at slightly different redshifts.
tionofgalaxiesduemainlytophotometricerrors,additional
We can use the NB and H images to select Hα emis-
criterionfortheobservedequivalentwidth(EW)largerthan H
35˚A (i.e., 14˚A in therest-framefor [OII]at z=1.46) is ap- sion line galaxies at z ∼1.46. Again, galaxies with a colour
excessinH–NB areselected,andthenthesamecolourse-
plied. As a result, we select 1,135 NB921 emitters. Among H
lection(Bz′K)isappliedtoidentifyHαemittersamongthe
them,789emittersaredetectedinK atmorethan2σ level.
19emissionlinecandidates.Thecutof3σcolourexcesscor-
In what follows, we deal with the K-detected emitters, be-
causeweidentifythe[OII]emittersbasedontheirB−z′and responds to galaxies with a line flux larger than 1.4×10−16
z′−K colours (Hayashiet al. 2011). With the criteria, we erg s−1 cm−2 being selected. If the excess is due to Hα at
z=1.46,thelimitingfluxcorrespondstoadust-freeSFRof
select429NB921[OII]emittersintotal(Figure2).Wenote
14.6M⊙ yr−1 accordingtoKennicutt(1998).Wealsoapply
(cid:13)c 2014RAS,MNRAS000,1–14
3-D structures at z = 1.46 7
an observed EW cut larger than 50˚A (i.e., 20˚A in the rest-
frame for Hα at z = 1.46). Although Sobral et al. (2013)
made a small correction to H–NB colours using J −H
H
colours to estimate the continuum level in the broadband
moreaccurately,wefindthatH–NB coloursaredistributed
H
around zero without any correction. Thus, no further cor-
rection is applied for the colour term and the method used
for the selection is similar to that applied for the selection
of [OII] emitters at z ∼ 1.46. With their criteria, we select
9 Hα emitters in total (Figure 2).
Now, we have three samples of emission line galaxies
at z ∼ 1.46; NB912 [OII] emitters, NB921 [OII] emitters,
and NB Hα emitters. Since they are all at similar red-
H
shifts, some emitters in the catalogues should overlap each
other.First,wecomparethecataloguesbetweenNB912and
NB921[OII]emittersbysearchingforgalaxiesforwhichthe
difference in the coordinates are within 0.5 arcsec. Among
380 NB912 and 429 NB921 [OII] emitters, it is found that
170galaxiesarecommoninbothsamples.Wealsofindthat
four NBH Hα emitters are in the [OII] emitter samples by
using a circle with 1.0 arcsec radius to search for the com-
mon galaxies. Moreover, cross-matching with the spectro-
scopic data described in § 2.6, we spectroscopically confirm
41NB912+NB921, 21NB912,11NB921[OII]emittersand
two Hα emitters (see also Figure 1). Figure 4. Top: Ratio of [OII] emission lines measured with
NB912 filter to that measured with NB921 filter as a function
Figure 3 shows a celestial distribution of NB912 or of redshift of galaxy. The broken line shows the line ratio ex-
NB921 [OII] emitters. It is worth mentioning that we can pected from the difference of the response function between the
confirm again the clustering of [OII] emitters in the cen- narrowband filters. Filled green circles show the measured line
ratios for the NB912+NB921 [OII] emitters with spectroscopic
treandthelarge-scalestructureexistinginthesurrounding
redshift. The solid line shows the linear relation between the
region which were already found in Hayashi et al. (2011),
line ratio and redshift obtained from the fitting to the data
even when the NB921 data are used. More importantly, it
of the NB912+NB921 [OII] emitters. Blue(red) symbols show
seemsthatthereisaslightdifferenceinthedistributionbe-
NB912(NB921) [OII] emitters without the detection of [OII]
tweenNB912andNB921[OII]emitters,althoughabouthalf emissioninNB921(NB921), respectively.Amongthem,opencir-
of them overlap. The prominent filamentary structure that clesshowsthegalaxieshaving[OII]fluxinadequatetobedetected
Hayashiet al. (2011) havediscovered is likely to beslightly inthe other narrowband, sincethe expected fluxissmallerthan
intheforeground,andthedistributionofNB921[OII]emit- thedetectionlimit.Inaddition,theresponsefunctionsofthetwo
ters seems to be spread along the southwestern direction filtersareshownbygraycurves.Bottom:Thecomparisonbetween
from the cluster centre. As shown in Figure 1, the NB912 theredshift,zNB,estimatedfromthelinefluxratiomeasuredby
and NB921 can detect galaxies with a difference in velocity NB912andNB921filtersandthespectroscopicredshift,zspec,for
by ∼2000 km s−1. Thus, all of the differences suggest that the NB912+NB921 [OII] emitters as a function of spectroscopic
there is a three dimensional structure around the galaxy redshift.Seethetext(§4)fordetailsofhowtheredshift,zNB,is
estimatedusingthetwonarrowbands.Thestandarddeviationof
cluster.Asimilarresulthasbeenreportedathigherredshift
thedifferenceisσ((zNB−zspec)/(1+zspec))=0.002.
aswell.Shimasaku et al.(2004)usedthesamplesofLyman
α Emitters (LAEs) selected by two narrowband filters and
foundthatthedistributionofthegalaxiesarequitedifferent
4 REDSHIFT MEASUREMENTS WITH TWO
at z = 4.79 and 4.86. These observations demonstrate that
ADJACENT NARROW-BAND FILTERS
revealing the three dimensional structure is crucial to im-
proveourunderstanding,inparticularintheearlyUniverse The two narrowband filters of NB912 and NB921 on
whengalaxiesareactiveandthestructuresaregrowing.This Suprime-Cam are very close to each other (Figure 1). The
impliesthetwodimensionalcelestialdistributioncouldlead overlapwithaslightdifferenceintheresponsecurvesallows
to errors in estimating the environment in which galaxies us to estimate the redshift where the [OII] emitters are lo-
reside due to projection effects. This is investigated in the catedbetweenz ≈1.42–1.49,basedonthedifferenceofemis-
following sections using thetwo [OII] emitter samples. sion line fluxesmeasured in the NB912 and NB921 images.
This is because the narrowband filters are not perfect top-
Figure 3 also shows a celestial distribution of NB Hα hats.Thatis,thefluxmeasuredbynarrowbandimagingfor
H
emitters.Unlikethe[OII]emitters,Hαemittersarenotclus- a galaxy with a given luminosity is dependent on its wave-
tered, but this is a likely consequence of the significantly length.Forexample,ifastar-forminggalaxyisatz=1.44,it
shallow detection limit of Hα emission when compared to shouldbedetectedasaNB912emitterbutnotdetectedasa
[OII].However,itseemsthatthedistributionofHαemitters NB921emitter.Ontheotherhand,ifastar-forming galaxy
is consistent with tracing that of [OII] emitters. is at z =1.48, it should be detected as a NB921 emitter
(cid:13)c 2014RAS,MNRAS000,1–14
8 M. Hayashi et al.
spectroscopicredshiftslessthanz∼1.46whichcorresponds
totheredshiftwherethegalaxyhastheNB912/NB921 line
ratio of unity, while those identified in only NB921 tend
to have the redshift greater than z ∼ 1.46. This is also
what we expect. Indeed, the blue(red) open circles show
the NB912(NB921) [OII] emitters for which the expected
NB921(NB912)[OII]fluxissmallerthanthedetectionlimit.
This is the case for about a third of NB912 or NB921 [OII]
emitters.Anotherthirdof[OII]emittersseemtohave[OII]
fluxeswhichintheoryshouldbeabletobedetectedinboth
narrowbands.However,theexpected[OII]fluxesinthenon-
detectionnarrowbandareclosetothedetectionlimit.Thus,
theactual[OII]fluxmaybesmallerthanthelimitduetothe
uncertaintyof[OII]fluxmeasuredbytheothernarrowband.
Finally, theremaining third of [OII]emitters should havea
[OII]fluxwhichiseasilydetectedinbothnarrowbands,and
yet are absent in one narrowband. Although the cause is
Figure 5. The distribution of the redshift estimated from unclear,thecriteriaofEWandsignificanceofcolourexcess
the line ratio measured by NB912 and NB921 filters for the
applied toselect theemitters as well as thecolour selection
NB912+NB921 [OII] emitters, is shown by the hatched his-
to identify the emission line can influence the detection of
togram.Amongthem,[OII]emitterswiththespectroscopicred-
shift are shown by the filled histogram. The two histograms are [OII] emission. Nevertheless, the redshift can be estimated
not consideredto beclearlydifferentdistributions, basedonthe from the line flux for about 70% of [OII] emitters detected
K-Stest,suggestingthatthemethodweusetoestimatethered- by NB912 or NB921 narrowband filters, which adequately
shiftisvalid. allows us to trace thethree dimensional structure.
Theredshiftestimated from thelineratio measuredby
NB912andNB921filters,z ,iscomparedwiththespectro-
NB
but not detected as a NB912 emitter. More interesting are scopicredshift,zspec,fortheNB912+NB921 [OII]emitters.
star-forminggalaxiesdetectedwithbothNB912andNB921 Wethencalculatethedifferenceofthetworedshiftsandfind
images.IftheratiooflinefluxmeasuredwithNB912tothat thatσ((zNB−zspec)/(1+zspec))=0.002(Figure4).Thissug-
with NB921 isgreater thanunity,thegalaxy isexpectedto gests that the accuracy of the redshift, zNB, is satisfactory,
be at z < 1.46, while if the ratio is smaller than unity the and despite only two narrowband datapoints being used,
galaxy is likely to be at z > 1.46. Therefore, we can reveal itisbetterthanthatofthephotometricredshiftsestimated
the three-dimensional large scale structure in and around fromphotometrywithasmanyas30-bandsintheCOSMOS
theXCS2215 galaxy cluster with these two filters. field; σ∆z/(1+zs) = 0.007 at i+AB < 22.5, σ∆z/(1+zs) = 0.012
Figure 4 shows that the expected ratio of NB912 to ati+AB <24andz<1.25,andσ∆z/(1+zs) =0.06ati+AB∼24
NB921-measured line flux is a monotonically decreasing andz∼2(Ilbert et al.2009).Thus,thelineratiomeasured
functionofredshiftbetweenz =1.430andz =1.485(shown with the adjacent narrowband filters can be a powerful es-
bythebrokenline),demonstratingthatwecanestimatethe timator of redshift enabling us to map out well the three
redshift from the flux ratio measured with the two narrow- dimensional structure of star forming galaxies in the field
bandfilters.Wethenmakesureofthevalidityofthemethod surveyedwithout spectroscopy of allgalaxies.
using the [OII] emitters which have been spectroscopically Figure 5 shows thedistributions of thetwo redshift es-
confirmed. For NB912+NB921 [OII] emitters, it is clearly timations;zNB andzspec.Thehistogramsseemtobesimilar
seen that the flux ratio is correlated with the spectroscopic to each other. Indeed, the Kolmogorov-Smirnov (K-S) test
redshiftandthefluxratiosmeasuredaredistributedaround indicates that theyare likely to show the same distribution
the values expected from the response function of the nar- and the probability of their being based on the same pop-
rowband filters. However, a systematic difference between ulations is more than 70%. Thus, we conclude that we can
the actual line ratio and the expectation is seen when the use the line ratio to estimate the redshift of [OII] emitters
redshiftisfarfromz∼1.46.Thediscrepancycanbecaused and then investigate their three dimensional distribution in
byinhomogeneityofresponsecurveovertheFoVand/orthe and around the galaxy cluster quiteaccurately.
differencebetween theactualresponsecurvewheninstalled
on the Suprime-Cam and themeasurement in a laboratory,
although it is unlikely that the response curve of the two
filtershavelargeinhomogeneity.Tocorrectforthoseeffects, 5 RESULTS
we fit a linear function to all of the data of NB912+NB921
5.1 3-D structures around the XCS2215 cluster
[OII] emitters, and the fitted relation is shown by the solid
line in Figure 4. Hereafter, we use this linear relation ob- Figure 6 shows the three dimensional distribution of
tainedfromthefittingtoestimatetheredshiftof[OII]emit- NB912+NB921 [OII] emitters. Although we have seen
ters.NotethatweremovedtheNB912+NB921[OII]emitter the filamentary structures in this field (Figure 3 and
at z = 1.438 from the fitting since the deviation from the Hayashiet al. (2011)), the 3D view reveals that it is not
othersis large, butevenif theemitteris included,we make an effect of projection but a real distribution and there are
sure that theresults shown below are not largely changed. severalnewfilamentarystructures.Itseemsthatsomestruc-
The[OII]emittersidentifiedinonlyNB912tendtohave tures extend towards the centre of the galaxy cluster, and
(cid:13)c 2014RAS,MNRAS000,1–14
3-D structures at z = 1.46 9
30
20
10
c]c] 0
pp
MM
[[ -10

hh
tt
pp
ee -20
DD

∆∆
-30
-40
-50
10
5
-20
0 -15
-10
-5 -5
∆ Dec. [Mpc] -10 0
5
-15 10 ∆ R.A. [Mpc]
15
-20 20
Figure 6. Threedimensional distribution ofNB912+NB921 [OII]emitters. The depth inz-axisis converted fromthe redshift, which
isestimatedfromtheratioof[OII]emissionlinesmeasuredwithNB912filtertothatmeasuredwithNB921filter(Figure4).Thefilled
circles are shown ingray scale based on the declination; darker colours mean lower declination. The filamentary structures are seen in
three dimensional space. The dots in the R.A.-Dec.plane are the projection on the celestial plane where the coordinates are shown in
co-movingscalerelativelytothe centreof thegalaxy cluster.Theradiusof R200 inthe XCS2215cluster is2.0Mpcinco-movingscale
(Hiltonetal.2010).
some intersect otherstructures. Thesestructurescannot be Figure 6 resembles the structures seen in cosmologi-
fullyunderstoodwithoutseeingthe3Ddistributionofgalax- cal N-body simulations based on the CDM model (e.g.,
ies as well as the projected celestial distribution. Galaxy Springel et al. 2005; Ishiyama et al. 2013). The simulations
clusters are considered to be formed at the intersection of predictthedistributionofdarkmatter.However,theconsis-
filamentary structures and grow while assembling galaxies tencybetweenthecosmological simulationandtheobserva-
bythegravityofthesystemalongthelarge-scalestructures. tionsatz ∼1.5confirmsthatthedistributionofgalaxiesin
The3Dstructureswehavefoundaroundthegalaxy cluster the Universe is governed by dark matter and grows hierar-
is fully consistent with this picture. We should be witness- chically. In addition, the three dimensional structureof the
ingthesitewheregalaxiesinthesurroundingregionarein- galaxyclusteratz =1.46providesuswithanuniqueoppor-
falling towards thecentre of galaxy cluster at z ∼1.5 when tunitytoinvestigatethedependenceofgalaxypropertieson
thevigorous evolution of galaxies occurred. the location of galaxies infalling into the cluster. At lower
Since we use the redshift to draw the distribution of redshifts, similar large-scale structures have been found
star-forminggalaxiesinthreedimension,weshouldtakecare (e.g.,Kodama et al. 2005;Nakata et al. 2005;Tanaka et al.
of theeffect of peculiar motion which distorts that view. In 2009;Sobral et al.2011;Gal et al.2008;Faloon et al.2013).
particular, thiseffect could work significantly verynear the However,thepropertiesofgalaxiesassociatedwiththeclus-
cluster core and thestructurein theredshift direction (i.e., terisdifferent,showingthatstarformationactivityweakens
depth)couldseem toshrinkrelativetowhat itis.However, in the core. It is interesting to investigate what causes the
thereisnodoubtoftheexistenceofthelargescalestructure, change of the properties in the course of galaxies infalling
sincethefilamentarystructurestendtospreadwidelyinthe along the filamentary structure. We discuss this point in
direction parallel to the celestial plane. §5.2.
(cid:13)c 2014RAS,MNRAS000,1–14
10 M. Hayashi et al.
z-K Log(F([OII]))
30 3 30 -15.4
20 2.5 20 -15.6
-15.8
10 2 10
-16
c]c] c]c]
pp 0 1.5 pp 0 -16.2
MM MM
epth [epth [ --2100 01.5 epth [epth [ --2100 ---111666...864
DD DD
∆∆ -30 0 ∆∆ -30 -17
-40 -40
-50 -50
10 10
5 0 -15 -20 5 0 -15 -20
∆ Dec. [Mpc-]5-10-15-20 20 15 10 5∆ R 0.A. -[5Mp-c1]0 ∆ Dec. [Mpc-]5-10-15-20 20 15 10 5∆ R 0.A. -[5Mp-c1]0
Figure 7. SameasFigure6,butthesymbolsarecolour-codedbyz−K colourintheleft-handpaneland[OII]fluxintheright-hand
panel.Thez−K colourofgalaxiesatz∼1.46correspondstotheU−z colourinrest-frame,implyingthatitisusefultoestimatethe
strength of theBalmer/4000˚A break. The[OII]fluxes arecorrected forthe responsecurve ofthe filter usingtheredshiftestimated for
theratioofNB912toNB921flux,andtheluminosityofthe[OII]emissionlineisanindicatorofthestarformationrateofthegalaxy.
5.1.1 Comparison between [OII] and Hα emitters agivenHαluminosity.Ontheotherhand,abouthalfofHα
emitters do not have [OII] emissions detected in NB912 or
Thankstotheuniquesetofnarrowbandfilters,wecancom- NB921data.Thesegalaxiesmaybemetal-rich.Thegalaxies
pare Hα emission with [OII] emission in an unbiased man- detectedasaNBHemittercouldalternativelybe[OII]emit-
ner for galaxies associated with the structures at z ∼ 1.46. ters at z =3.3 and/or [OIII] emitters at z =2.2. However,
Similar comparisons have been conducted for emission line since the fluxes of the emission lines detected are greater
galaxies detected with dual narrowband filters in the field than 1.4×10−16 erg s−1 cm−2, the corresponding luminosi-
(Sobral et al. 2012; Hayashi et al. 2013). Before comparing ties are quite large if the galaxies are at z ≫ 2. Thus, it is
them, we must keep in mind that thelimiting fluxesin Hα unlikely that almost all of them are contaminants.
and[OII]emissionsarequitedifferentasshownin§3orTa- In comparison with the large numbers of [OII] emit-
ble1;1.4×10−16 ergs−1 cm−2 for Hαemitters(adust-free ters being detected, thenumberof Hα emitters detected is
SFR of 14.6 M⊙ yr−1) and 1.2 or 1.4×10−17 erg s−1 cm−2 small. Typically, [OII] emitters at z ∼ 1.46 are a popula-
for [OII] emitters (a dust-free SFR of 2.2 or 2.6 M⊙ yr−1). tionwithaloweramountofdust(Hayashiet al.2013).Less
Also, note that this Hα limit is much shallower than those dustygalaxieshaveanobservedratioofHαto[OII]of∼1,
in Sobral et al. (2012) and/or Hayashi et al. (2013). suggestingthatHαluminositymaynotbestrongforgalax-
AsshowninFigure3,forfouroutofnineHαemitters, ieswithagiven[OII]luminosity.Indeed,thereareonlyfour
[OII] emissions are detected by NB912 or NB921 imaging. [OII] emitters with [OII] fluxes larger than 1.4×10−16 erg
All of them have [OII] emission detected in NB921 data, s−1 cm−2 in the region where NBH and H data by WF-
butonly oneemitterhas[OII]emission detectedin NB912. CAM are available. Even if the limit is weakened down to
This is likely due to the difference between the filter re- 1.0×10−16 erg s−1 cm−2, the number of the [OII] emitters
sponsecurvesinredshiftspace(Figure1).Notethatamong is only 12. This is of the same order of magnitude as the
the four, the Hα emitter with [OII] emission detected in numberof theHα emittersdetected among the[OII]emit-
bothNB912andNB921isspectroscopically confirmed,and tersample. This means that theratio of Hαto [OII]is not
the others are not confirmed yet. Figure 2 shows that the highbutclosetounity,whichmaysupportthatmostofthe
Hαemitterswith[OII]detectionstendtohaveredderz′−K [OII]emittersaretypicallylessdusty.Thus,adeepersurvey
colour. This may go against our intuition. We expect that of Hα emission is required so that we can obtain a larger
[OII]emissioniseasiertodetectinlessdustygalaxieswhich sample of Hα emitters at z∼1.5.
should havebluer colour. The observations seem to suggest
thatthedustextinctiondoesnotdeterminemainlywhether
5.2 Environmental dependence of galaxy
the [OII] emission is detected or not for the Hα emitters.
properties
Perhaps,it isrelated totheage,if theredderz′−K colour
meansstrongerBalmer/4000˚A breakduetoanolderstellar One of the hot topics in the evolution of clus-
populationratherthanheavierdustextinction.Wecheckthe ter galaxies is how and where the galaxies obtained
EWoftheemitters,butwedonotfindanysignificantdiffer- the properties characteristic of cluster galaxies seen
encebetweenthe[OII]emitterswithandwithoutHαdetec- in the local Universe, with redder colour, less star
tion(seealsoFigure2).TheratioofHα/[OII]isalsodepen- formation activity, and bulge-dominated morphology
dent on metallicity (e.g., Sobral et al. 2012; Hayashi et al. (e.g., Blakeslee et al. 2006; Tanaka et al. 2005, 2013;
2013).Ifthegalaxieshavelowermetallicity,[OII]luminosity Gerkeet al. 2007; van den Bosch et al. 2008; Thomas et al.
shouldbehigherandeasier tobedetectedforgalaxies with 2010; Lemaux et al. 2012; Rodriguez Del Pino et al. 2013;
(cid:13)c 2014RAS,MNRAS000,1–14