Ultraviolet tails and trails in cluster galaxies: A sample of candidate gaseous stripping events in Coma

Abstract

We have used new deep observations of the Coma cluster from GALEX to identify 13 star-forming galaxies with asymmetric morphologies in the ultraviolet. Aided by optical broad-band and H-alpha imaging, we interpret the asymmetric features as being due to star formation within gas stripped from the galaxies by interaction with the cluster environment. The selected objects display a range of structures from broad fan-shaped systems of filaments and knots (`jellyfish') to narrower and smoother tails extending up to 100 kpc in length. Some of the features have been discussed previously in the literature, while others are newly identified here. As an ensemble, the candidate stripping events are located closer to the cluster centre than other star-forming galaxies; their radial distribution is similar to that of all cluster members, dominated by passive galaxies. The fraction of blue galaxies which are undergoing stripping falls from 40% in the central 500 kpc, to less than 5% beyond 1 Mpc. We find that tails pointing away from (i.e. galaxies moving towards) the cluster centre are strongly favoured (11/13 cases). From the small number of `outgoing' galaxies with stripping signatures we conclude that the stripping events occur primarily on first passage towards the cluster centre, and are short-lived compared to the cluster crossing time. Using infall trajectories from simulations, the observed fraction of blue galaxies undergoing stripping can be reproduced if the events are triggered at a threshold radius of ~1 Mpc and detectable for ~500 Myr. HST images are available for two galaxies from our sample and reveal compact blue knots coincident with UV and H-alpha emission, apparently forming stars within the stripped material. Our results confirm that stripping of gas from infalling galaxies, and associated star formation in the stripped material, is a widespread phenomenon in rich clusters. Comment: MNRAS, in press

Full text

Page 1

arXiv:1006.4867v2 [astro-ph.CO] 28 Jun 2010
Mon. Not. R. Astron. Soc. 000, 1−16 (2010)Printed 30 June 2010 (MN LATEX style file v2.2)
Ultraviolet tails and trails in cluster galaxies:
A sample of candidate gaseous stripping events in Coma
Russell J. Smith1⋆, John R. Lucey1, Derek Hammer2,3, Ann E. Hornschemeier3,2,
David Carter4, Michael J. Hudson5,6,7, Ronald O. Marzke8, Mustapha Mouhcine4,
Sareh Eftekharzadeh9, Phil James4, Habib Khosroshahi9, Ehsan Kourkchi9, Arna Karick4
1Department of Physics, University of Durham, Durham DH1 3LE
2Department of Physics and Astronomy, Johns Hopkins University, 3400 North Charles Street, Baltimore, MD 21218, USA
3NASA Goddard Space Flight Centre, Code 662.0, Greenbelt, MD 20771, USA
4Astronomical Research Institute, Liverpool John Moores University, Twelve Quays House, Egerton Wharf , Birkenhead CH41 1LD
5Department of Physics and Astronomy, University of Waterloo, 200 University Avenue West, Waterloo, Ontario N2L 3G1, Canada
6Institut d’Astrophysique de Paris, UMR 7095 / Universit´ e Pierre et Marie Curie, 98bis boulevard Arago, 75014, Paris, France
7Perimeter Institute for Theoretical Physics, 31 Caroline St. N., Waterloo, ON, N2L 2Y5, Canada
8Department of Physics and Astronomy, San Francisco State University, San Francisco, CA 94132, USA
9School of Astronomy, Institute for Research in Fundamental Sciences, PO Box 19395-5531, Tehran, Iran
30 June 2010
ABSTRACT
We have used new deep observations of the Coma cluster from Galaxy Evolution Explorer
to visually identify 13 star-forming galaxies with asymmetric morphologiesin the ultraviolet.
Aidedbywide-fieldopticalbroad-bandandHα imaging,we interpretthe asymmetricfeatures
as being due to star formation within gas stripped from the galaxies by interaction with the
cluster environment.The selected objects display a range of structures from broad fan-shaped
systems of filaments and knots (“jellyfish”) to narrower and smoother tails extending up to
100kpc in length. Some of the features have been discussed previouslyin the literature, while
others are newly identified here. We assess the ensemble properties of the sample. The candi-
date stripping events are located closer to the cluster centre than other star-forming galaxies;
their radial distribution is more similar to that of all cluster members, dominated by passive
galaxies. The fraction of blue galaxies which are undergoing stripping falls from 40per cent
in the central 500kpc, to less than 5per cent beyond 1Mpc. We find that tails pointing away
from (i.e. galaxies moving towards) the cluster centre are strongly favoured (11/13 cases).
From the small numberof “outgoing”galaxies with stripping signatures, we conclude that the
strippingevents occur primarilyon first passage towards the cluster centre, and are short-lived
compared to the cluster crossing time. Using galaxy infall trajectories extracted from a cos-
mologicalsimulation, we find that the observedfraction of blue galaxies undergoingstripping
can be reproduced if the events are triggered at a threshold radius of ∼1Mpc and detectable
for∼500Myr. HubbleSpaceTelescope images are available for two galaxies fromoursample
and revealcompactblue knots coincidentwith UV andHα emission,apparentlyformingstars
within the stripped material. Our results confirm that stripping of gas from infalling galaxies,
and associated star formation in the stripped material, is a widespread phenomenon in rich
clusters. Deep UV imaging of additional clusters is a promising route to constructing a statis-
tically powerful sample of stripping events and constraining models for the truncation of star
formation in clusters.
Key words: galaxies: clusters: individual: Coma — galaxies: evolution
⋆Email: russell.smith@durham.ac.uk
1INTRODUCTION
In rich galaxy clusters, the evolution of the member galaxies and
the intra-cluster medium (ICM) are intricately linked, through vari-
ous forms of ejection of material from galaxies into their surround-
c ? 2010 RAS

Page 2

2
Russell J. Smith et al.
ings. Galaxies in dense environments are susceptible to loss of stel-
lar mass through tidal stripping, removal of cold gas from their
disks by ram-pressure stripping (e.g. Gunn & Gott 1972; Quilis,
Moore & Bower 2000), and stripping of hot gas from their halos
by the same mechanism (e.g. Larson, Tinsley & Caldwell 1980;
McCarthy et al. 2008). The eventual consequences are as varied
as the physical processes involved: depletion of cold gas truncates
star formation the disks, perhaps leading to the formation of clus-
ter S0 galaxies; halo-stripping causes a more gradual decline in the
star-formation history (“strangulation” or “starvation”); tidal strip-
ping transfers individual stars and globular clusters to the intra-
cluster population (e.g. Peng et al. 2010), and may disrupt dwarfs
enough to affect the luminosity function (e.g. Henriques, Bertone
& Thomas 2008); stripped cold gas may be heated and mixed with
the ICM, or survive in clumps that can heat the cluster core (Dekel
& Birnboim 2008), or fragment to form new stars or clusters in the
intra-cluster population (Yoshidaet al. 2008; Puchwein et al.2010).
There is substantial observational evidence for gaseous (as op-
posed to purely stellar) stripping events in clusters. Notable exam-
ples are seen in Virgo(e.g. Phookun & Mundy 1995; Crowl & Ken-
ney 2006; Vollmer et al. 2009), Abell 3627 (Sun, Donahue & Voit
2007; Woudt et al. 2008; Sun et al. 2010), Abell 1367 (e.g. Gavazzi
et al. 2001; Scott et al. 2010), as well as in Coma (e.g. Vollmer et
al. 2001; Yagi et al. 2007; Yoshida et al. 2008). Two particularly
spectacular cases have been highlighted by Cortese et al. (2007) in
Abell 1689 and Abell 2667 at z ∼ 0.2. Another example in a dis-
tant cluster (Abell 2125 at z = 0.25) is discussed by Owen et al.
(2006).
Simple calculations based on the arguments of Gunn & Gott
(1972) suggest that ram-pressure stripping iseffective mainly inthe
cores of clusters. Several groups have recently performed detailed
simulations of individual galaxies to investigate the properties of
the stripped tails (e.g. Roediger & Br¨ uggen 2008; Kapferer et al.
2009; Tonnesen & Bryan 2010). The simulations with radiative
cooling (Kapferer et al.;Tonnesen & Bryan) show narrow, highly
structured tails, with dense clouds embedded within them. When
star formation isincluded (Kapferer et al.),the starscreated inthese
dense clouds in some cases fall back towards the stripped galaxy,
adding to its bulge.
Until now, observational studies of ongoing stripping in clus-
ters were based on one or two galaxies per cluster, in clusters of
disparate properties and a range in redshift. As a result, meaning-
ful statistical conclusions about the stripped galaxies could not be
drawn. Inthispaper, weusewide-fieldUV andoptical imagingdata
to identify a sample of candidate gaseous stripping events in a sin-
gle nearby cluster (Coma), primarily on the basis of emission from
young stars formed in the stripped gas. Rather than concentrate on
details of the individual cases, we instead focus on the ensemble
statistics of galaxies undergoing this process, including their radial
and redshift distributions compared to other cluster galaxies, and
the orientation of their projected velocities relative to the cluster
centre.
The paper is structured as follows: The observations and se-
lection of candidate stripping events are described in Section 2,
followed by a brief description of each of the selected objects, with
reference to previous work. The spatial and redshift distribution of
the sample is analysed in Section 3 and interpreted in Section 4.
Our conclusions are summarized in Section 5.
Throughout the paper, we adopt a distance of 100Mpc for
Coma, so that the distance modulus is 35.0 and one arcminute cor-
responds to 29kpc.
2SAMPLE CONSTRUCTION
2.1Observations
Our primary observational resources are deep ultraviolet (UV)
imaging from the Galaxy Evolution Explorer (GALEX) satellite,
and ugi optical imaging from the 3.6m Canada–France–Hawaii
Telescope (CFHT).The GALEXdata are especially useful tosearch
for stripping events because the formation of young stars in the
stripped material leads to high contrast in the UV. Accordingly, we
limit our study to the area covered by two deep GALEX observa-
tions.
In the core of the cluster, we obtained a deep observation in
GALEX Cycle 5 (PI: Smith), in ‘petal-pattern’ mode to ensure de-
tector safety due to the presence of UV-bright stars in the field.
The total exposure time in the co-added data used for our analysis
is 14.7ksec in the NUV detector and 13.6ksec in the FUV. In the
south-west part of the cluster, we use the deep observation made
in GALEX Cycle 2 (PI: Hornschemeier), and analysed by Hammer
et al. (2010a). The co-added images used for the present analy-
sis, have a depth of 21.0/19.8ksec (NUV/FUV), slightly less than
the total integration used by Hammer et al. Together, the GALEX
fields extend nearly to the virial radius of the cluster, and include
thewell-known infalling group centred on NGC 4839 (Briel, Henry
& B¨ ohringer 1992). In the core field, the surface brightness limit
for detecting diffuse structures (which we define as a 3σ fluctu-
ation on 10×10arcsec areas) is 29.7magarcsec−2in NUV and
30.1magarcsec−2in FUV.
Complementary optical data have been obtained with Mega-
Cam at the CFHT (PI: Hudson), imaging a 3×3deg2area centred
on the cluster core. The exposure times were 1360s, 300s, 300s in
u, g and i bands. After pipeline processing and stacking, the 80per
cent completeness depths for point sources are at least 24.5, 24.0
and 23.5 in u, g and i, with variations from field to field due to
varying sky brightness. In the central field, the surface brightness
limit (defined as above) is 27.6, 27.4 and 26.5 magarcsec−2in u,
g and i. The image quality in the i-band ranges from 0.5arcsec to
0.8arcsec FWHM, with a median of ∼0.6arcsec.
Withinthecentral partof thecluster,wealsoemployed amuch
deeper u-band MegaCam dataset obtained from the archive (PI:
Adami), using a custom stack kindly generated by Dr S. Gwyn,
with a total integration of 23400s, and FWHM 1.3arcsec. The sur-
facebrightness limitis28.5magarcsec−2. Thisimagewas not used
for construction of catalogues, but only as an additional visual ref-
erence for faint features in the u-band.
We made similar incidental use of Hα imaging data from
the 2.5m Isaac Newton Telescope (INT, PI: Mouhcine). The Hα
data are part of a wide-field survey covering ∼2.5deg2of the
cluster and sensitive to Hα emission in the redshift range cz =
4900 − 10800kms−1. These data will be presented and analysed
elsewhere. Here, we used continuum-subtracted Hα images of the
GALEX-selected candidate stripping events, to assess the distribu-
tion of ongoing star formation in the stripped systems.
2.2 Catalogues
In order to measure matched-aperture colours, the MegaCam im-
ages were resampled, using SWARP (Bertin et al. 2002), to the same
1.5-arcsec pixels as the GALEX data, and smoothed to match the
∼5arcsec FWHM point-spread function (PSF) of the UV images.
In the latter step, we applied a circular gaussian PSF, constant over
all of the data. Although variation and ellipticity of the PSF should
c ? 2010 RAS, MNRAS 000, 1−16

Page 3

Gaseous stripping candidates in Coma
3
Mi
NUV − i
−24−23−22−21−20−19−18−17
1
2
3
4
5
6
7
Figure 1. The UV versus optical colour–magnitude relation for
spectroscopically-confirmed cluster members within the two GALEX fields.
No galactic extinction corrections or k-corrections have been applied. The
galaxies plotted in red are outliers from a robust fit to the red sequence, indi-
cated by the solid line. Objects bluer than the dashed line at NUV −i = 4
form the sample examined for signs of gaseous stripping. The 13 candidate
stripping events identified in this paper are highlighted by green crosses.
be taken into account for photometry of higher precision, these ap-
proximations are justified for our purposes.
Photometry was performed using SEXTRACTOR (Bertin &
Arnouts 1996) in dual-image mode. The MegaCam i-band im-
ages were used for object detection, deblending and aperture def-
inition. Magnitudes for the detected objects were then computed
within common apertures for the two GALEX and three MegaCam
bandpasses. In this paper, all magnitudes and colours are based on
counts within the i-band elliptical Kron-like aperture.
Finally, the catalogues were matched to a compilation of red-
shiftdata, including adeepsurvey madewithHectospec atthe6.5m
MMT (Marzke et al. in preparation). Within the surveyed fields,
there are 589 known member galaxies, defined by 3000kms−1<
cz <13000kms−1, having i ≤ 18 (Mi ≤ −17), and all but two
of these were recovered in the GALEX data with positive flux. Fig-
ure 1 shows the NUV − i colour–magnitude relation for the re-
sulting galaxy sample. A clear red sequence of passive galaxies is
seen, with a scatter of 0.38mag around a linear fit, which is typi-
cal for the NUV colours (e.g. Rawle et al. 2008). The systematic
deviation from the linear fit for the brightest galaxies is likely due
to contamination from the UV-excess sources (old hot stars) which
dominate at FUV but also contribute substantial flux at NUV.
2.3Visual selection of candidate gaseous stripping events
We aim to identify stripping events that are due to “gaseous” in-
teraction with the surrounding medium. Under this definition, we
include several processes including ram-pressure stripping, viscous
stripping, and also tidal interactions perturbing the gas. However,
we do not include tidal disturbance or disruption of purely stellar
systems, which mayalsobe common inclusters(e.g. Gregg&West
1998), but which are not sensitively probed by UV data. Although
gaseous interactionsdirectly perturbthegasdistribution(which can
be observed in HI 21cm emission), they may also lead to the for-
mation of new stars within the stripped material which strongly
affect the appearance of the galaxy in the UV regime. Asymmet-
ric features in the UV morphology thus provide an indirect probe
of ongoing (or at least very recent) gaseous stripping. The primary
criterion we apply to select a sample of such events is therefore
UV asymmetry visible in the GALEX image, without a strong cor-
responding disturbance in the redder bands of the MegaCam data.
An initial visual search in the central GALEX field revealed a
number of candidates for recent gaseous stripping, including sev-
eral previously noted in the literature. All have blue UV-to-optical
colours, with NUV −i < 4, compared to red-sequence colours of
5 < (NUV −i) < 6. To construct a more controlled sample, all 80
galaxies with NUV − i < 4 and Mi < −17 (see Figure 1) were
subsequently examined closely in the UV and optical images, to
search for comparable features. The simple constant colour thresh-
old was adopted for simplicity; examining the galaxies that would
be selected by a sloping cut that follows the red sequence, no addi-
tional similar objects were found. During the visual examination of
blue galaxies, ten were judged to be due to blends with neighbour-
ing galaxies or contaminated by image edges, ghosts, halos from
bright stars, etc. These are removed from the sample of blue ob-
jects when computing fractions and distributions in the following
sections.
We identified 13 galaxies showing UV asymmetry which
we interpret as evidence for ongoing gaseous stripping (Table 1
and Figure 2). Some galaxies show clearly distorted images, with
the optical data resolving multiple filaments offset asymmetrically
from the galaxy. These are similar to the “jellyfish” galaxies in
more distant clusters described by Cortese et al. (2007). In other
cases, the features we identify are trails of UV emission, in which
the optical counterpart takes the form of narrow streams or clumps
in the u-band image. These objects appear more like “tadpoles”,
and include for instance the galaxy with a 60kpc Hα tail discussed
by Yagi et al. (2007).
2.4 Comments on individual objects
In this section we briefly discuss the characteristics of each galaxy
in our sample, and the features which led to its identification as a
gaseous stripping candidate (see also Figure 2). Where relevant, we
summarise any previous literature discussion of each galaxy in the
context of gas stripping.
GMP 2559 = IC 4040
This is a spiral galaxy with an irregular, patchy distribution of
dust obscuration. It is one of only two galaxies in our sample that
fall within the footprint of the Coma HST/ACS Treasury Survey1
(Carter et al. 2008, Hammer et al. 2010b). The GALEX imaging
reveals a plume of emission towards the south-east, and three com-
pact sources to the south-west. At very low surface brightness lev-
els, the region between these two features appears to be filled by
faint diffuse UV emission.
The UV plume to the south-east extends ∼1.8arcmin
(∼50kpc) from the centre of GMP 2559, and can also be seen in
the deep Adami u-band image, though there is no clear counter-
part in our shallower MegaCam data. The trail passes close to a
neighbouring faint red-sequence galaxy GMP 2529, which is offset
by 775kms−1in radial velocity. Co-located with the UV plume is
an Hα trail extending at least 1.4arcmin (40kpc), with secondary
1The custom-reduced ACS imaging from this survey is publicly available
from http://archive.stsci.edu/prepds/coma
c ? 2010 RAS, MNRAS 000, 1−16

Page 4

4
Russell J. Smith et al.
Figure 2. Images of the candidate stripping events. For each galaxy, the first panel shows a greyscale image of the combined GALEX FUV and NUV bands,
while the second panel is a GALEX/MegaCam composite with NUV and FUV combined for the blue channel, g and u combined for the green and i for the red.
The yellow line points along the direction we identify as a “trail”, while the blue line points away from the cluster centre. Hα emission, from INT narrow-band
imaging, is overlaid as contours for selected objects (GMP 2559, GMP 2599, GMP 2910, GMP 3816, GMP 4060, GMP 4471 and GMP 4629). The third panel
shows the Adami deep MegaCam u-band data (except for GMP 5422 where we show our shallower image).
c ? 2010 RAS, MNRAS 000, 1−16

Page 5

Gaseous stripping candidates in Coma
5
Figure 2 – continued
c ? 2010 RAS, MNRAS 000, 1−16

Page 6

6
Russell J. Smith et al.
Figure 2 – continued
c ? 2010 RAS, MNRAS 000, 1−16

Page 7

Gaseous stripping candidates in Coma
7
Table 1. Identifications, co-ordinates and other data for the galaxies discussed in this paper. GMP numbers are from Godwin, Metcalfe & Peach (1983). P.A.
is the average position angle (in degrees east of north), of the stripped material estimated by eye from the images. The angle θclusis measured between the
tail and the cluster-centric vector, such that θclus= 0◦is a tail directed away from the cluster centre, and θclus= 180◦is a tail directed towards the cluster
centre. The column headed “type” gives the morphological class assigned by Dressler (1980). Hα indicates whether extended or asymmetric Hα emission is
observed in our INT data. The “comp” column indicates whether the galaxy has a bright projected neighbour galaxy with a small radial velocity difference
(< 500kms−1). The final column is the leading identification in the NASA Extragalactic Database.
ID R.A.
(J2000)
Dec
(J2000)
Dcl
[kpc]
cziNUV − i
[mag]
P.A.
[deg]
θclus
[deg]
TypeHα
comp NED name
[kms−1][mag]
GMP 2559
GMP 2599
GMP 2640
GMP 2910
GMP 3016
GMP 3816
GMP 4060
GMP 4232
GMP 4471
GMP 4555
GMP 4570
GMP 4629
GMP 5422
195.158
195.140
195.122
195.038
195.004
194.759
194.678
194.628
194.523
194.491
194.486
194.459
194.119
28.057
27.638
27.515
27.866
28.082
28.116
27.760
27.564
28.243
28.061
27.992
28.170
27.291
448
694
874
277
278
360
501
831
786
675
659
796
1727
7845
7497
7429
5297
7765
9431
8756
7283
7193
8163
4595
6936
7532
14.37
15.04
15.26
15.08
17.40
14.82
16.36
17.45
13.22
14.82
16.50
16.75
14.57
2.55
1.94
3.20
3.46
3.05
2.23
2.83
2.51
2.69
2.88
1.20
1.48
2.72
177
141
336
62
40
326
180
138
332
235
304
344
246
110Scd
Sb
S0p
I
—
Sbc
—
—
Scd
I
I
—
—
y
y
n
y
n
y
y
y
y
y
n
y
n
n
y
n
n
n
n
y
n
n
n
y
n
IC 4040
KUG 1258+279A
KUG 1258+277
MRK 0060 NED01
MAPS-NGP 0 323 0924669
NGC 4858
SDSS J125842.58+274537.8
MAPS-NGP 0 323 1033883
NGC 4848
KUG 1255+283
SDSS J125756.79+275930.3
SDSS J125750.23+281013.2
IC 3913
5
179
65
3
4
45
73
21
51
29
46
20—
Figure 2 – continued
Hα peaks at 12, 23 and 28kpc from the nucleus. The ACS imag-
ing appears to show compact, fairly red objects centrally located
within these peaks, but these may simply be chance alignments.
The galaxy is deficient in HI (DefHI = 0.56, Gavazzi et al. 2006),
and Bravo-Alfaro et al. (2001) note that the HI distribution is off-
set towards the south-east (i.e. the same direction as the Hα tail).
A displacement in the same sense is seen for both the X-ray and
radio continuum emission (Finoguenov et al. 2004; Miller, Horn-
schemeier & Mobasher 2009).
The three UV-bright objects to the south-west, which have not
previously been discussed in the literature, are also seen in the Hα
emission maps, extending 0.8arcmin (23kpc) from the nucleus.
The central object is catalogued as GMP 2572, and has been con-
firmed using spectroscopy from LRIS at Keck (Chiboucas et al.,
in preparation) to be a strong emission-line source at a velocity of
7600kms−1(i.e. 250kms−1lower than GMP 2559). Wolf–Rayet
features are observed in the spectrum, indicating a very young star-
burst (∼5Myr). In the Carter et al. (2008) ACS imaging, the peaks
of the Hα emission to the south-west correspond to very blue, ir-
regular systems of compact knots, reminiscent of the “fireballs” in
GMP 4060 (Yoshida et al. 2008)2. The south-west clumps also ap-
pear in archival Spitzer/MIPS data of Bai et al. (2006), as a faint
extension of the 24µm emission from GMP 2559, coincident with
the UV/optical/Hα detections.
GMP 2599 = KUG 1259+279A
This spiral galaxy appears distorted in the GALEX image, showing
broad streaks that point south-east away from the cluster centre.
Some faint knots are visible in the MegaCam images. Miller et al.
(2009) note the radio emission is offset from the galaxy and sug-
gestive of ram-pressure stripping, while Finoguenov et al. (2004)
found that the X-ray source is offset to the east of the galaxy.
Gavazzi et al. (2006) detected the galaxy in HI but indicate that
it is strongly gas-deficient (DefHI = 0.76). The Hα data show an
offset in the emission to the east of the galaxy centre.
2The “fireballs” in GMP 2559 and GMP 4060 are described in greater
detail in Section 4.3.
c ? 2010 RAS, MNRAS 000, 1−16

Page 8

8
Russell J. Smith et al.
cz [km/s]
050001000015000
0
50
100
200
stripping
events
All members
cz [km/s]
05000 1000015000
0
5
10
15
20
25
stripping
events
Blue control sample
D [kpc]
0 50010001500200025003000
0
20
60
100
140
stripping
events
All members
D [kpc]
050010001500200025003000
0
2
4
6
8
10
12
stripping
events
Blue control sample
Figure 3. The redshift and radial distributions of the galaxies with stripping features, compared to the matched Coma member galaxies (upper panels) and
to the blue galaxies without ongoing stripping (lower panels). The vertical tick above each histogram marks the median of the comparison sample, while the
open symbols show the stripping galaxies (at arbitrary vertical position).
GMP 2640 = KUG 1258+277
This appears to be an edge-on S0 or spiral galaxy, which presents
a very long trail in the GALEX imaging, extending 3.5arcmin
(100kpc) north towards the cluster centre. In the MegaCam data,
the trail is faintly visible and the inner morphology is clearly dis-
turbed. The galaxy is an HI non-detection in Bravo-Alfaro et al.
(2000), with an upper limit which implies that it is gas deficient
(DefHI > 0.7). In the INT data, we do not detect any Hα emission,
either fromthegalaxy itselfor fromthetrail.Closeinprojection are
an elliptical (GMP 2670 at 0.5arcmin or 15kpc, with 360kms−1
radial velocity difference) and a spiral (GMP 2601 at 1.2arcmin
or 35kpc, with 1800kms−1radial velocity difference); hence this
could possibly be acaseof tidal strippingby aneighbouring galaxy,
rather interaction with the cluster itself.
GMP 2910 = MRK 0060 NED01
This irregular or spiral galaxy has a post-starburst spectrum in the
disk region, with the burst age estimated at 250Myr and ongoing
star formation in the nucleus (Caldwell, Rose & Dendy 1999). In
our GALEX image, we observe a narrow tail of length ∼0.5arcmin
(15kpc), extending to the north-east. Stripping in this galaxy was
first discussed by Yagi et al. (2007), who reported a 60kpc Hα
tail, which is also seen in our INT Hα imaging. This remarkably
narrow and straight feature is also clearly seen in the deep Adami
u-band image, and is co-located with the UV trail. The presence of
continuum emission suggests that star formation is taking place in
the stripped material, not merely ionization of a purely gaseous tail
as proposed by Yagi et al. The galaxy is seen close in projection
to an elliptical (GMP 2897, at 0.3arcmin or 10kpc) further away
is another early-type galaxy (GMP 2852, at 1.0arcmin or 30kpc),
but both have large differences in radial velocity (4700kms−1and
2100kms−1respectively) and are unlikely to be physically asso-
ciated with GMP 2910.
GMP 3016
This is among the less convincing cases in the sample. GMP 3016
itself is a faint irregular galaxy. About 0.5arcmin (15kpc) to the
north-east, there is an elongated UV source which points towards
the galaxy which could be a trail similar to the case of GMP 2910.
In the Adami u-band imaging, the ‘trail’ appears to extend up to
1arcmin (30kpc), with numerous clumps. A companion is seen in
projection to the south-west, along the same axis as the ‘trail’. Its
radial velocity is 8314kms−1, which is 550kms−1larger than
that of GMP 3016, so the two galaxies are potentially associated,
however it is much fainter (i = 18.7) and unlikely to be caus-
ing a major tidal disturbance. There is no clear detection of star
formation in Hα. It is possible that this system could be a chance
alignment of unrelated galaxies.
c ? 2010 RAS, MNRAS 000, 1−16

Page 9

Gaseous stripping candidates in Coma
9
GMP 3816 = NGC 4858
This disturbed barred spiral presents a “jellyfish” morphology in
the GALEX image and in the MegaCam data (especially u-band).
Several tails and knots, are seen in a broad fan-like distribution,
extending 0.5arcmin (15kpc) to the north-west, away from cluster
centre. There are strong variations in Hα-vs-UV flux ratio among
the various filaments; in particular the brightest UV structure is not
coincident with the strong central Hα feature. The galaxy is defi-
cient in HI (DefHI > 1.1), based on the upper limit to the gas mass
published by Gavazzi et al. (2006). The galaxy is observed close in
projection (0.6arcmin or 20kpc) to a large elliptical, GMP 3792,
with1500kms−1radial velocitydifference. Itispossiblethat these
galaxies are physically interacting, although their relative velocity
is quite large compared to the characteristic internal velocities, and
hence strong mutual interactions are unlikely.
GMP 4060
This galaxy, discussed extensively by Yoshida et al. (2008), has a
disturbed optical morphology suggestive of a merger remnant, and
a post-starburst spectrum (Poggianti et al. 2004). There is some
tidal stellar debris prominent in the optical images, to the west of
the galaxy. In the GALEX images, the more spectacular feature is
a broad fan of emission extending ∼1arcmin (30kpc) south of the
galaxy, with several bright knots. Yoshida et al. used deep broad-
band and Hα imaging to reveal a complex network of filaments ex-
tending up to 2.7arcmin 80kpc from the galaxy. Embedded within
these filaments are bright star-forming knots, which Yoshida et al.
refer to as ‘fireballs’, which in several cases are co-located with the
UV emission peaks. The galaxy and part of the filamentary struc-
ture fall within the Carter et al. (2008) HST/ACS imaging survey,
which reveals an extended clumpy substructure within two of the
fireballs, very similar to the knots south-west of GMP 2599 (see
Section 4.3). GMP 4060 is observed close in projection to a faint
elliptical (GMP 4035 at 0.8arcmin or 24kpc), but there is a large
radial velocity difference (2100kms−1), so a tidal interaction is
not likely.
GMP 4232
This is a distorted possibly spiral galaxy with an apparent trail to
thesouth-east intheGALEXimaging. TheMegaCamu-band image
shows a fan of faint filaments/knots 0.5arcmin (15kpc) south-east
of the galaxy; beyond this the GALEX “trail” is probably spurious,
caused by aligned unrelated galaxies. GMP 4232 is close in projec-
tion and radial velocity to an edge-on S0 (GMP 4255, at 0.6arcmin
or 20kpc, with 290kms−1radial velocity difference), so tidal in-
teraction is a possibility. The INT Hα imaging shows one bright
knot and some faint extended emission to the SE, coincident with
the UV features.
GMP 4471 = NGC 4848
This is the brightest galaxy in the sample, a spiral with a disturbed
inner morphology and a faint extension to the north-west visible in
theMegaCam imaging. In theGALEXdata, we observe trailspoint-
ing north-west away from cluster centre, certainly to 1.4arcmin
(40kpc), and tentatively to twice this distance. (The uncertainty
arises because this galaxy is close to the edge of the GALEX field.)
The INT Hα images show an asymmetric distribution of emission
in the core, and a stream of emitting knots to the north-west of the
galaxy, some of which are coincident with compact blue sources in
the MegaCam imaging. GMP 4471 has been the subject of previous
studies: Gavazzi et al. (1998) showed the Hα morphology of the
inner region, but no extended structure. Bravo-Alfaro et al. (2001)
noted an unusual HI distribution peaking 0.3arcmin (10 kpc) north-
west of the core. It is also deficient in total HI (DefHI > 0.49,
Gavazzi et al. 2006). Vollmer et al. (2001) discussed the peculiar
morphology inCO, HI and Hα, and proposed this object as a “post-
stripping” galaxy, i.e. one which has already passed through the
cluster. Finally Finoguenov et al. (2004) noted a tail of X-ray emis-
sion to the north-west of GMP 4471, making it one of very few
galaxies known to exhibit a stripping trail in hot gas.
GMP 4555 = KUG 1255+283
The GALEX imaging reveals a UV plume to the west of this dis-
turbed galaxy, oriented away from the cluster centre. In the Mega-
Cam u-band data, the inner part of the plume is delineated by faint
filaments extending to the south-west The Hα emission is also
asymmetric and extended to the south-west. Miller et al. (2009)
note that the radio emission is offset from the galaxy and sugges-
tive of ram-pressure stripping. The X-ray morphology also appears
extended to the west in figure 3 of Finoguenov et al. (2004), al-
though they do not specifically discuss this source.
GMP 4570
This isa distorted spiral galaxy, for which the GALEX image shows
a clumpy UV extension to the west, away from the cluster centre.
Hα is not detected from this object in the INT narrow-band imag-
ing, because its large velocity with respect to the cluster shifts the
emission line outside of the filter profile. The u-band MegaCam
image shows some trails and knots, coincident with the UV peaks,
extending at least ∼0.4arcmin (12kpc) from the galaxy centre.
GMP 4629
This distorted spiral shows an asymmetric extension in the GALEX
image. In the MegaCam data, multiple blue knots are seen to the
north-west of the galaxy, away from cluster centre. The Hα imag-
ing shows secondary emission peaks to the north and north-west of
the core. GMP 4629 lies close in projection (0.7arcmin or 20kpc)
to an elliptical galaxy (GMP 4648), with 300kms−1radial veloc-
ity difference, and hence a tidal interaction is possible in this case.
GMP 5422 = IC 3913
This is perhaps the least convincing example of stripping in our
sample. The galaxy is a spiral with a distinct opening of the spiral
arms to the south-west, and an enhancement of star formation to
the north-east. The galaxy is part of the NGC 4839 group to the
south-west of the cluster core, and lies beyond the extent of our Hα
observations. Previous studies in HI (Bravo-Alfaro et al. 2001), in
Hα (Gavazzi et al. 1998), and in radio continuum (Miller et al.
2009) have not suggested ongoing stripping in this galaxy.
2.5 Independent Hα identifications
During revision of this work, Yagi et al. (2010) submitted a paper
based on Hα imaging of the Coma cluster core from Subaru (much
c ? 2010 RAS, MNRAS 000, 1−16

Page 10

10
Russell J. Smith et al.
projected clustercentric radius [kpc]
fraction of blue galaxies undergoing stripping
0 50010001500 2000
0.0
0.2
0.4
0.6
0.8
1.0
Figure 4. The fraction of blue (NUV − i < 4) galaxies undergoing
gaseous stripping, as a function of distance from the cluster centre. The
light and dark intervals represent 90per cent and 68per cent confidence in-
tervals, based on binomial statistics. The upper solid line shows the fraction
of bright late-type galaxies from Gavazzi et al. (2006) that are more HI-
deficient than 95per cent of galaxies beyond 3Mpc. The dotted lines show
the 68per cent interval on the HI-deficient fraction.
deeper than the INT data), in which they independently select can-
didate gaseous stripping galaxies based on extended Hα features.
Yagi et al. confirm all six of our GALEX-selected objects within the
region of overlapping observations (GMP 2559, GMP 2910, GMP
3016, GMP 3816, GMP 4060 and GMP 4232). It is notable that
two of these (GMP 3016 and GMP 4232) were among the less se-
cure identifications discussed in the previous section. The Yagi et
al. observations hence provide independent support for our sample
selection methods in general.
3STATISTICS OF GASEOUS STRIPPING EVENTS
The objects identified in Section 2 were selected primarily on the
basis of asymmetric UV light distribution, indicating a disturbance
in the distribution of recently-formed stars, without corresponding
asymmetry in the old stars probed by redder optical bands. In gen-
eral we assume these features to result from interaction with the en-
vironment, i.e. either with neighbouring galaxies, or with the clus-
ter itself, via hydrodynamical or gravitational tidal forces. These
forces act on gas in the parent galaxies, not only stripping gas
from the galaxy disks but also triggering the formation of new stars
which become visible in the UV. In the hydrodynamical case, only
the gas is directly affected, but the newly-formed stars provide the
observed evidence. In the case of tidal interactions, existing stars
are affected as well as the gas, but the formation of new stars, fun-
damentally a gaseous process, causes the asymmetry to be espe-
cially prominent in the UV.
We therefore describe the UV trails discussed above as being
due to “gaseous stripping events” (GSEs hereafter): the production
of UV asymmetries occurs through stripping of gas, star formation
within that gas, and subsequent UV emission from the new stars.
The term GSE, as we use it here, does not distinguish between var-
ious possible physical mechanisms: tidal interaction with neigh-
bours or with the cluster potential, ram-pressure stripping, viscous
stripping etc (see Boselli & Gavazzi 2006 for detailed discussion of
these processes). In what follows, however, we muster support for a
picture in which ram-pressure stripping is a major (though not nec-
essarily the only) process responsible for the objects we observe.
Instead of considering the details of individual galaxies, our
approach is to use the extensive data for Coma to analyse the en-
semble properties of the sample. In this section, we examine the
incidence and distribution of the GSE galaxies, in comparison to
the cluster galaxy population at large, which is dominated by the
red sequence members. We also compare the GSEs to a blue con-
trol sample comprised of the 57 galaxies with NUV −i < 4 which
are not identified as GSEs (after excluding the ten galaxies judged
to have unreliable colours during the visual inspection). Implicitly
we assume that these star-forming galaxies are the “parent” popu-
lationfrom which the GSEsweretriggered, and therefore statistical
differences between the GSEs and control sample indicate the con-
ditions responsible for the stripping events.
The left-hand panels of Figure 3 show the redshift distribution
of the GSE galaxies, compared to the other Coma cluster members.
The GSE galaxies on average have slightly higher radial velocities
than thefull sample of matched members, but given thelarge veloc-
itydispersion (∼1000kms−1) inComa, theoffset isnot significant
(316 ± 370kms−1). The sample of blue members not undergoing
gaseous stripping is offset in the same sense, relative to the galaxy
population at large (by 560±250 kms−1). Thus the GSEsare con-
sistent with having been drawn from the same velocity distribution
as either the full sample or the blue control sample.
The right-hand panels of Figure 3 present equivalent compar-
isons for the radial distribution, adopting a cluster centre coinci-
dent with the giant elliptical galaxy NGC 4874. The GSE galax-
ies are markedly more concentrated towards the cluster core than
are the blue control galaxies, with 12/13 (92per cent) lying within
1Mpc, compared to 26/57 (47per cent) of the control sample. A
Kolmogorov–Smirnov (KS) test yields a 0.4per cent probability
that the GSE galaxies were drawn from the same radial distribution
as the control sample galaxies. The GSEs are, in fact, consistent
with being drawn from the much more concentrated distribution
followed by all matched cluster members, which is dominated by
passive galaxies. Note also that the outermost GSE galaxy is GMP
5422 which is the least certain case of stripping identified here.
Removing it from the sample would strengthen the discrepancy be-
tween GSE and control-sample galaxies in terms of the radial dis-
tribution.
Next we consider the fraction of all star-forming (NUV −i <
4) galaxies which are currently undergoing gaseous stripping. For
the surveyed region as a whole, the GSE galaxies form a fraction
NGSE/Nblue = 13/70 = 0.19+0.06
the incidence of stripping is concentrated towards the cluster core.
Within a radius of 1Mpc from the cluster centre, the fraction be-
comes NGSE,1Mpc/Nblue,1Mpc = 12/38 = 0.32 ± 0.07. These
fractions refer to thestripping features detectable inour data; future
deeper imaging might reveal even an larger incidence of low-level
stripping activity. Figure 4 shows graphically the variation in strip-
ping fraction with radius. Clearly we have insufficient numbers to
draw any secure conclusions regarding the form of the decline, e.g.
whether there is really a sharp cut-off somewhere around 1Mpc,
which could indicate a threshold ICM density for stripping.
We have visually estimated the position angles of the
streams/trails/tails, and calculated the angle relative to the cluster-
−0.05. However, as shown above,
c ? 2010 RAS, MNRAS 000, 1−16

Page 11

Gaseous stripping candidates in Coma
11
Figure 5. The projected distribution and orientation of the galaxies with stripping features. The dashed circles show the extent of the GALEX imaging, while
large open circles show the locations of giant ellipticals NGC 4839, NGC 4874 and NGC 4889, for reference. Small circles show all matched Coma cluster
members in our catalogues, highlighting in blue those with NUV − i < 4. The GSE galaxies are marked with large filled symbols and vectors denoting the
direction of their projected motion, as inferred from their streams or tails. Galaxies apparently approaching the cluster centre are marked in blue, while those
receding from it are plotted in red. Dotted lines indicate the cluster-centric direction at the position of each GSE. The thick green line-segment joins the two
residual X-ray maxima in the western structure identified by Neumann et al. (2003). The inset shows the distribution of the alignment angle θclusbetween
tails and the cluster-centric vector. We define θclusto be near zero for galaxies apparently approaching (i.e. tails pointing away from) the cluster centre and
near 180◦for galaxies apparently receding from (i.e. tails pointing towards) the cluster centre.
centric vector for each galaxy, again adopting the position of NGC
4874 for the cluster centre. In some galaxies the position angle of
the debris is unclear, and a the adopted angle is open to debate.
For the case of GMP 2559, we interpret both the south-east trail
and the south-west clumps as the brightest parts of a broad fan of
stripped material, and hence assign a position angle which is close
to south. For GMP 4060, we ignore the tidal stellar debris to the
west, and assign a position angle close to south, to describe the
star-forming knots and filaments. In each case, the position angles
are shown in the images of Figure 2. Figure 5 shows the location
and orientation of all the GSEs within the cluster, and summarises
the distribution of alignment angles. Among the thirteen objects
selected here, six show streams extending away from the cluster
centre (within 30◦of the cluster-centric vector), compatible with
stripping occurring on passage towards the cluster centre. Only one
galaxy(GMP 2640) has astreampointing directlytowards theclus-
ter centre, which instead implies that the galaxy has already passed
through the cluster core. In the remaining six objects, the asymme-
tries are at large angle to the cluster-centric axis, but favour galax-
ies approaching, rather than receding from the cluster core. The
distribution of cluster-centric angle is incompatible with a uniform
distribution at the 99per cent level, according to a KS test.
c ? 2010 RAS, MNRAS 000, 1−16

Page 12

12
Russell J. Smith et al.
Projected distance from cluster centre [kpc]
Projected distance to nearest neighbour [kpc]
0500100015002000 2500
0
50
100
150
200
Figure 6. Projected distance to the nearest bright neighbour for cluster
members, as a function of distance from the cluster centre. To identify po-
tential cases of tidal interaction, we select only the neighbours that are no
more than 0.5mag fainter (in i) than the target galaxy. The overall broad
correlation arises because the density of cluster members is highest near
the cluster centre. The black points show the GSE galaxies, the blue points
are the control sample of blue cluster members, and the grey points are the
red cluster members. The points highlighted in green are GSE galaxies with
bright neighbours close in projection, and at similar radial velocity (offset
< 500kms−1).
4DISCUSSION
In this section, we discuss our results in the context of other obser-
vations of stripping in dense environments. In particular, we com-
pare withother cases of UV/Hαtrailsin other rich clusters(Cortese
et al. 2007; Sun et al. 2010), and with observations of HI tails in
Virgo (Chung et al. 2007, 2009). We also make reference to sim-
ulations, especially those of Kapferer et al. (2009), to interpret the
various observed structures.
4.1Ram-pressure stripping on first passage into cluster
The key results from Section 3 are that blue galaxies with tails
and other stripping features are preferentially located closer to
the cluster core than blue galaxies without such features, and that
the stripped tails predominantly point away from the cluster cen-
tre. Both results suggest that the stripping events are in general
triggered by interaction with the cluster itself; while ram-pressure
stripping by the intra-cluster gas is probably the key process in-
volved, other mechanisms such as tidal interactions may contribute
to the features observed.
As noted in Section 2.4, it is possible that at least some of the
GSE galaxies are interacting with neighbouring galaxies. Such in-
teractions are not simply an alternative to ram-pressure stripping,
but may act in combination with it. Simulation work (e.g. Vollmer
2003; Kapferer et al. 2008) has demonstrated that gravitational in-
teractions between galaxies can act to enhance the efficacy of ram-
pressure stripping, by pulling gas to larger radius where it can be
more easily removed by the intra-cluster wind. In Figure 6, we
show the projected distance to the nearest neighbour for galaxies
in the matched member sample. The neighbours are drawn from
the same sample, i.e. confirmed cluster members with Mi < −17,
and further restricted to be no more than 0.5mag fainter (in the i
band) than the “target” galaxy, since tidal stripping is most effi-
cient for companions of comparable mass. The figure demonstrates
that most of the GSE candidates do not have closer neighbours than
non-GSE galaxies at similar distance from the cluster centre. How-
ever, in three cases (GMP 2640, GMP 4232 and GMP 4629) the
nearest neighbour is very close and also has similar radial velocity
to the target galaxy, as required for tidal interactions. (The other
three galaxies with close projected neighbours have velocity differ-
ences >1500kms−1, and are therefore unlikely to be physically
interacting.) GMP 2640 is the single clear exception to the the ten-
dency for tails to be aligned away from the cluster centre, so it is
tempting to invoke tidal stripping of stars and/or gas, rather than
(or in combination with) ram-pressure stripping to account for this
galaxy. Finally GMP 4060 is worthy of note here; it is probably a
merger remnant, and thus a case in which ram-pressure and tidal
interaction may have acted together to form its spectacular system
of knots and filaments. We conclude fromthis test that tidal interac-
tion with neighbouring galaxies is not the mechanism responsible,
in general for the stripping events identified in this paper, but may
play a role in some specific cases.
An alternative possibility which is more difficult to distin-
guish is that the stripping may be due to tidal interactions, not with
neighbouring galaxies, but with the cluster potential itself. Such
effects would produce disturbances oriented preferentially along
the cluster-centric vector, as in the case of ram-pressure stripping.
There is qualitative evidence against purely tidal interaction, in that
the old stellar material, as traced by the i-band luminosity distri-
bution, appears fairly undisturbed in most of the GSE galaxies, but
this is hard to quantify. Moreover, as shown by Smith et al. (2010)
from a GALEX atlas of interacting field galaxies, the triggering of
star formation from gas in tidally-stripped material can lead to tails
that are accentuated in the UV even in the absence of ram-pressure
stripping. Ultimately, the head–tail morphologies of many of the
GSE galaxies provide the strongest, though not conclusive, evi-
dence that the stripping events are indeed caused by ram pressure,
rather than tidal interaction.
Interpreting the observed trails as stripped material behind
the galaxies that have suffered gaseous stripping, we infer that the
stripping events are occurring mainly on approach to the cluster
centre. Furthermore, the absence of GSEs moving away from the
cluster centre indicates strongly that the stripping event is seen on
first approach to the cluster centre, and generally completed on a
time-scale short compared to the crossing time. Combined with the
typical cluster-centric radius of 700kpc, and a velocity dispersion
of 1500kms−1, this yields a time-scale for the stripping event of
TGSE<
∼500Myr.
This crossing-time argument is supported by analysing more
realistic infall trajectories extracted from the Millenium Simula-
tion (Springel et al. 2005). We selected all galaxies with stellar
mass greater than 109M⊙ in the model of Font et al. (2008), that
are members of the four most massive clusters in the simulation
(halo masses ∼ 1015M⊙). Ignoring the star-formation histories as-
signed by the semi-analytic machinery, we instead construct a sim-
ple model which assumes each galaxy enters the cluster forming
stars (i.e. is “blue”) and is then stripped of gas when it first comes
within a threshold distance of 1Mpc from the centre of the clus-
ter. The galaxy is assumed to be observable as an ongoing stripping
event (i.e. a GSE) for a period TGSE after crossing this threshold
c ? 2010 RAS, MNRAS 000, 1−16

Page 13

Gaseous stripping candidates in Coma
13
radius, and to remain “blue” for a longer interval Tblue = 1.5Gyr
after stripping begins, while young stars are still present. Identify-
ing the stage reached in this process for all galaxies at z = 0, the
predicted ratio of GSEs to all blue galaxies can be compared to our
observed GSE fraction as a function of projected radius from the
cluster centre. To match the assumptions used in interpreting the
observations, we use the angle between the projected velocity vec-
tor and the direction to the cluster centre to determine whether the
simulated GSEs are apparently approaching or receding from the
cluster centre.
Theresultsof thisanalysisareshown inFigure7,for threeval-
ues of the stripping time-scaleTGSE. Asexpected, the total fraction
of model GSEs increases with with increasing TGSE. The fraction
of “outgoing” GSEs is small until TGSE becomes comparable to
the time taken by an typical galaxy to travel from the 1Mpc thresh-
old radius to the peri-centre of its orbit. To match the observed to-
tal GSE fraction of ∼30per cent, we find that a stripping dura-
tion longer than ∼300Myr is required. However, the duration must
be also be shorter than ∼700Myr, to avoid predicting too many
outgoing GSEs. An intermediate time-scale of TGSE ≈ 500Myr,
yields acceptable agreement, with ∼25per cent of blue galaxies
undergoing stripping, and ∼80per cent of GSEs approaching the
cluster centre, although these averages mask substantial cluster-to-
cluster variation. The threshold radius at which the GSE phase is
assumed to begin (held at 1Mpc in all panels of Figure 7) is con-
strained by the radial distribution of GSEs and by their total frac-
tion. The threshold radius must be larger than ∼750kpc to match
the GSE fraction in the second bin. The threshold radius must also
be smaller than ∼ 1.5Mpc; otherwise either the fraction of GSEs
falls too low at low radius (for short GSE time-scale) or else the
overall fraction of GSE is much too large (for longer GSE time-
scales).
In summary, this fairly crude modelling suggests that the frac-
tions of objects undergoing stripping are consistent with what may
be expected from galaxy accretion paths within clusters, and yields
approximate constraints on the threshold radius and duration of the
observed stripping events.
4.2 Typical density for gas stripping
In the UV, we observe gaseous stripping occurring predominantly
within a projected radius of 1Mpc. For Coma this corresponds, us-
ing the beta-model fit of Briel et al. (1992), to a hot gas density of
ρ ≈ 10−27gcm−3.
For comparison, the “jellyfish” galaxies identified by Cortese
et al. (2007) in clusters at z = 0.2 are located at projected dis-
tances of ∼300kpc from their cluster centres, corresponding to
ρ ≈ 10−25gcm−3using the beta-model parameters tabulated in
their paper. ESO137–001 inAbell3627 isprojected ∼200kpc from
the cluster centre, where the ambient density is ρ ≈ 10−27gcm−3
(Sun et al. 2010). The galaxy in Abell 2125 discussed by Owen
et al. (2006) is also close to the core, at a projected distance of
∼100kpc. Since these are all very rich clusters, these projected
radii are directly comparable with Coma, and we may conclude
that our GSE galaxies typically lie at larger radii (both absolute
and relative to viral radius), by a factor of ∼3 than these previous
examples. Of course, previous detection of GSEs at large radii will
have been limited by the requirement for wide fields of view (es-
pecially for nearby clusters) or HST observations (for distant clus-
ters),introducing abiastowardsthemost central objects.Moreover,
the density required for effective stripping will likely depend on
galaxy mass, sothat thegiant galaxies studiedinprevious workwill
undergo stripping only in the centres of clusters, while the much
fainter objects identified here (typically M⋆
to gas loss at larger radii.
Our results can also be compared to the incidence of neutral
gas tails identified behind spirals in the Virgo cluster by Chung
et al. (2007, 2009). From a survey sampling cluster-centric radii
0.3–3.3Mpc, they identify seven such galaxies, all of which lie at
intermediate radii, 0.6–1.0Mpc. The HI tails are all pointing away
from the centre of Virgo. The radii are similar in absolute terms
to those of our GSEs in Coma, but relative to the corresponding
cluster virial radii, the HI tailsin Virgoare located further out in the
cluster, by a factor of ∼4, and at much lower external gas density,
ρ ≈ 2×10−28gcm−3. A simple interpretation of this comparison
would be that removal of neutral gas and production of HI tails can
occur at a lower ICM density, while higher densities are required
to trigger significant star formation in the stripped material. This
is supported by the simulations of Kapferer et al. (2009), whose
figures 6 and 7 show the fraction of all newly-formed stars that are
located in the wake. For ambient densities ρ = 10−28gcm−3, and
relative velocity 1000kms−1, this fraction is ∼10per cent, while
for ρ > 10−27gcm−3, a majority of the triggered star formation
occurs in the wake.
Finally, we consider whether localised enhancements of the
ICM might be responsible for the stripping events we observe.
Analysing the XMM mosaic observation of Coma, Neumann et al.
(2003) showed the presence of an extended enhancement in emis-
sion (relative to a symmetric model) in the west of the cluster, at
projected radii 400–900kpc, with two maxima in the X-ray resid-
ual map. Four of our GSE galaxies lie just 100–200kpc beyond
the north-west maximum. Although we cannot yet quantify this as-
sociation, it is at least suggestive that some stripping events could
be triggered by encounters with local density enhancements in the
ICM. This possibility is reminiscent of the claim by Poggianti et
al. (2004) that young post-starburst galaxies in Coma are related
to local features in the ICM, indeed including the same Neumann
et al. western structure. The most prominent X-ray substructure in
the region we have studied is the group centred on NGC 4839 to the
south-west of the cluster core. Only one GSE galaxy, GMP 5422, is
identified in this region, and as noted in Section 2.4, this is among
the least certain cases. Inspecting the XMM mosaic, the local pro-
jected X-ray surface brightness close to GMP 5422 is comparable
to that at ∼1Mpc elsewhere in the cluster. If GMP 5422 were to
be confirmed as a GSE, it would provide further evidence for local
ICM structures as the driver for gaseous stripping in clusters.
i+ 1.5) are susceptible
4.3Intra-cluster star formation
We now turn to consider the fate of the stars which form in the
stripped material. If the gas is able to cool sufficiently, new stars
may be formed beyond the extent of the original source galaxy, as
seen in the simulations of Kapferer et al. (2009). Individual exam-
ples of intra-cluster star formation have been observed previously
in nearby and distant clusters (Gerhard et al. 2002; Cortese et al.
2004, 2007; Boquien et al. 2007; Sun et al. 2007; Reverte et al.
2007; Yoshida et al. 2008), and our Coma GSE sample shows that
this process may be widespread. Unlike the gas from which they
formed, the stars do not experience ram pressure from the ICM,
and are free to move under gravity, falling back towards their par-
ent galaxy. The “de-coupling” of newly-formed stars from the ICM
wind leads to a characteristic gradient in the stripped trails. While
the star-formationevent isongoing, the stripped material will be lu-
minous both in the UV and in Hα emission. However, because the
c ? 2010 RAS, MNRAS 000, 1−16

Page 14

14
Russell J. Smith et al.
TGSE= Myr
250
projected clustercentric radius [kpc]
fractions
0 500100015002000
0.0
0.2
0.4
0.6
0.8
1.0
TGSE= Myr
500
projected clustercentric radius [kpc]
fractions
05001000 15002000
0.0
0.2
0.4
0.6
0.8
1.0
GSE fraction
Outgoing GSE fraction
Post−stripping fraction
TGSE= Myr
750
projected clustercentric radius [kpc]
fractions
0500100015002000
0.0
0.2
0.4
0.6
0.8
1.0
Figure 7. A comparison between the observed stripping fraction and predictions from a simplistic model based on galaxy orbits in the four richest clusters of
the Millenium Simulation. The model assumes that galaxies enter as “blue” objects, become visible as ongoing stripping events for a duration TGSEafter
first passing within 1Mpc of the cluster centre, and turn from “blue” to “red” 1.5Gyr after crossing this threshold. The blue lines (one for each cluster) show
the fraction of all blue galaxies that are undergoing stripping at z = 0, for comparison with the observed GSE fractions (grey boxes as in Figure 4). The green
lines show the ongoing stripping events that are apparently moving away from the cluster centre (only <10per cent of all GSEs in the observed sample). The
red lines show the fraction of all “blue” galaxies that have passed through the stripping threshold radius, which may be compared to the observed HI-deficient
fraction (solid line as in Figure 4, the error bounds are not shown here, for clarity). Only for stripping time-scales of TGSE≈ 500Mpc can we reproduce the
fraction of GSEs without generating many more “outgoing” GSEs than are observed.
Hα emission fades rapidly after star formation ceases (∼10Myr),
the streams of back-falling stars will be most prominent in the UV
continuum, for which the timescaleis 100–1000Myr. Thisscenario
issupported insomeindividual cases, notably GMP4060 wherethe
“fireballs” at the ends of the filaments are luminous in Hα but the
filaments closer to the galaxy are seen only in UV. In some config-
urations, the newly-formed stars will remain bound to the source
galaxy, and may be able to rejoin it, contributing to the growth
of a central bulge, as in the simulations of Kapferer et al. (2009).
Given the low total mass of stars likely being produced in the tails,
however, this process seems unlikely to resolve the discrepancy be-
tween bulge masses in cluster S0s and their supposed spiral pro-
genitors (e.g. Kodama & Smail 2001; Wilman et al. 2009).
Alternatively, the newly-formed stars may become unbound
from the source galaxy, contributing to a genuinely intra-cluster
population of stars or star clusters. Such intra-cluster star formation
has been tentatively identified in recent hydrodynamic simulations
of galaxy clusters by Puchwein et al. (2010), where it contributes
some 30per cent of the total intra-cluster light. If the stars formed
in the fireballs were to remain internally bound in compact sys-
tems, these would constitute an unusual class of objects analogous
to the “tidal dwarf galaxies” (TDG) identified in some interacting
galaxies (e.g. Hancock et al. 2009). Two galaxies from our sample
(GMP 2559 and GMP 4060) were observed in the HST survey of
Carter et al. (2008), providing high-resolution images of the star-
forming clumps in these objects (Figure 8). From the this data, we
find that the brightest clumps have luminosities 106−107L⊙in the
I-band. Based on the Starburst99 models of Leitherer et al. (1999),
their colours of B − I ≈ 0.2 suggest I-band mass-to-light ratios
of ∼0.01M⊙/L⊙ (neglecting dust). Thus the stellar mass of these
structures is probably ∼ 104−105M⊙, similar to globular clusters.
From their Hα fluxes, we estimate the star-formation rates to be
0.001–0.004M⊙yr−1, and hence their star-formation time-scales
are <1.0Gyr. The fireball masses are similar to those estimated for
the TDGs in Arp 305 by Hancock et al. (2009). For TDGs in inter-
actingfieldgalaxies, itisamatterof disputewhether theywilleven-
tually become independent of the galaxies which originally hosted
their gas. By contrast, in the case of ram-pressure stripped clus-
ter galaxies, the ultimate detachment of the clumps from the host
galaxy seems quite likely, and if they survive as bound systems
they may evolve into stellar systems resembling intra-cluster glob-
ular clusters or compact dwarf galaxies. Similar objects have been
noted in simulations of ram-pressure stripping by Kapferer et al.
(2008), who term them “stripped baryonic dwarfs”.
4.4Post-stripping galaxies and HI deficiency
Finally, we consider the destiny of the galaxies themselves after
stripping is completed. It has been known for a long time that clus-
ter spirals are deficient in neutral gas, relative to their counterparts
in the field (e.g. Haynes & Giovanelli 1984). In Coma, Gavazzi et
al. (2006) find that a significant average HI deficiency extends from
the cluster core out to ∼ 2Mpc, the gas content becoming consis-
tent with a field reference sample at ∼3Mpc. (A similar trend is
seen for Virgo, e.g. Cayatte et al. 1994). The HI deficiency data for
Coma show an apparently sharp transition at a radius of ∼1Mpc,
within which nearly all cluster members are gas-poor compared to
field spirals.
Figure 4 compares the Gavazzi et. al. HI-deficient fraction
to the incidence of ongoing stripping events identified in this pa-
per. For this test, the galaxies are flagged as gas-deficient if they
have DefHI > 0.64, corresponding to the 95th percentile of DefHI
among galaxies beyond 3Mpc from the Coma core. It is notable
that a similar characteristic radius of ∼1Mpc seems to apply to
both phenomena. On the other hand we found no ongoing strip-
ping events, with the uncertain exception of GMP 5422, beyond
1Mpc, where 20–30per cent of spirals are HI-deficient. This result
can be understood in terms of a “backsplash” population (Sanchis
et al. 2002; Gill, Knebe & Gibson 2004): although stripping itself
is only effective within ∼1Mpc, HI-deficient post-stripping galax-
ies can be observed at larger radii after the initial stripping event is
complete and the galaxy has passed through the cluster core. This
c ? 2010 RAS, MNRAS 000, 1−16

Page 15

Gaseous stripping candidates in Coma
15
is confirmed by Figure 7, which shows that our simple stripping
model, tuned to reproduce the fraction of GSE galaxies, also pro-
duces a post-stripping population consistent with the observed HI-
deficient galaxy fraction.
5CONCLUSIONS
We have used UV and optical imaging to identify a sample of can-
didate gaseous stripping events in the Coma cluster. The stripped
galaxies are characterised by tails or trails of UV-bright debris,
which we interpret as young stars formed within gas stripped by
ram pressure from the intra-cluster medium. Some of these cases
have been noted as peculiar in previous work, in a variety of wave-
bands (Vollmer et al. 2001; Finoguenov et al. 2004; Yagi et al.
2007; Yoshida et al. 2008; Miller et al. 2009), while others are
newly identified here as possible stripping events.
The trails are predominantly oriented away from the cluster
centre, indicating that the galaxies are falling into the cluster for
the first time, along fairly radial orbits, and that the stripping events
are completed rapidly compared to the orbital time-scale. All but
one uncertain case lie at projected radii of 300–900kpc from the
cluster centre. The radial distribution of these galaxies is much
more centrally concentrated than the distribution of blue galaxies
from which they were selected, and more similar to the distribution
of passive galaxies. Within 1Mpc projected radius, some 30per
cent of blue galaxies are currently undergoing stripping, a fraction
which is compatible with a ∼500Myr time-scale for the stripping
events. Theradiuswithinwhich UV trailsareobserved corresponds
to an ICM density of ∼ 10−27gcm−3, in agreement with simula-
tions which show significant star formation in the stripped wake in
this density regime for infall velocities ∼1000kms−1(Kapferer et
al. 2009). There are hints that some stripping events are associated
with local enhancements in the ICM density, e.g. the western struc-
ture and the NGC 4839 group, but a firm link can not be concluded
from the present data.
We propose an interpretation of these objects as a stage in
ram-pressure stripping that is subsequent to the HI gas-tail phase
(Chung et al. 2007), and occurring at higher ambient densities. The
star formation triggered in the stripping events may add mass to
the galaxy bulge, if newly-formed stars fall back into the source
galaxy. Alternatively they may escape, forming intra-cluster stellar
systems that could evolve into objects resembling globular clusters
or compact dwarf galaxies. After the initial stripping, the infalling
galaxies will remain as gas-deficient spirals before fading slowly
into S0s as they exhaust their remaining gas.
As stressed by Sun et al. (2010), a fuller understanding of the
relationship between different manifestations of gas stripping (HI
deficiency, and tails in HI, UV, Hα and X-ray) will be made pos-
sible by improving the overlap between observations in the vari-
ous wavebands, for the same galaxy cluster. Our work has assem-
bled acomprehensive wide-fieldoptical, UV, Hα and spectroscopic
dataset for Coma, complemented by archival XMM imaging and
the radio continuum survey of Miller et al. (2009). A key missing
element is high-sensitivity 21cm HI mapping of a large sample of
Coma member galaxies, which should be possible in the next few
years using the Expanded Very Large Array.
ACKNOWLEDGMENTS
We are grateful to Stephen Gwyn for generating a custom stack of
theAdami deep u-band dataforour use, toMasafumi Yagifor com-
municating the Subaru Hα results in advance of submission, and to
Neal Miller for helpful comments on this paper. RJS was supported
for this work by STFC Rolling Grant PP/C501568/1 “Extragalactic
Astronomy and Cosmology at Durham 2008–2013”. This work is
based on observations made with the NASA Galaxy Evolution Ex-
plorer (GALEX). GALEX is a NASA Small Explorer, developed in
cooperation with the Centre National d’Etudes Spatiales of France
and the Korean Ministry of Science and Technology. This work
is based on observations obtained with MegaPrime/MegaCam, a
joint project of CFHT and CEA/DAPNIA, at the Canada–France–
Hawaii Telescope (CFHT) which is operated by the National Re-
search Council (NRC) of Canada, the Institute National des Sci-
ences de l’Univers of the Centre National de la Recherche Scien-
tifique of France, and the University of Hawaii. The work has made
use of data products produced at the TERAPIX data center located
at the Institut d’Astrophysique de Paris. The Isaac Newton Tele-
scope is operated on the island of La Palma by the Isaac Newton
Group in the Spanish Observatorio del Roque de los Muchachos of
the Instituto de Astrof´ ısica de Canarias. This work has made use of
the NASA/IPAC Extragalactic Database (NED) which is operated
by the Jet Propulsion Laboratory, California Institute of Technol-
ogy, under contract with the National Aeronautics and Space Ad-
ministration. The Millennium Simulation databases used in this pa-
per and the web application providing online access to them were
constructed as part of the activities of the German Astrophysical
Virtual Observatory.
REFERENCES
Bai L., Rieke G.H., Rieke M.J., Hinz J.L., Kelly D.M., Blaylock M. 2006,
ApJ, 639, 827
Bertin E., Arnouts S. 1996, A&AS, 317, 393
Bertin E., Mellier Y., Radovich M., Missonnier G., Didelon P., Morin B.
2002, ASPC, 281, 228
Boquien M., Duc P.-A., Braine J., Brinks E., Lisenfeld U., Charmandaris V.
2007, A&A, 467, 93
Boselli A., Gavazzi G. 2006, PASP, 118, 517
Bravo-Alfaro H., Cayatte V.,van Gorkom J.H.,Balkowski C. 2000, AJ,119,
580
Bravo-Alfaro H., Cayatte V., van Gorkom J.H., Balkowski C. 2001, A&A,
379, 347
Briel U.G., Henry J.P., B¨ ohringer H. 1992, A&A, 259, L31
Caldwell N., Rose J.A., Dendy K. 1999, AJ, 117, 140
Carter D. et al. 2008, ApJS, 176, 424
Cayatte V., Kotanyi C.,Balkowski C., van Gorkom J.H. 1994, AJ, 107, 1003
Chung A., van Gorkom J.H., Kenney J.D.P., Vollmer B. 2007, ApJ, 659,
L115
Chung A., van Gorkom J.H., Kenney J.D.P., Crowl H., Vollmer B. 2009,
AJ, 138,1741
Cortese L., Gavazzi G., Boselli A., Iglesias-Paramo J. 2004, A&A, 416, 119
Cortese L. et al. 2007, MNRAS, 376, 157
Crowl H.H., Kenney J.D.P. 2006, ApJ, 649, L75
Dressler 1980, ApJS, 42, 565
Finoguenov A., Briel U.G., Henry J.P., Gavazzi G., Iglesias-Paramo J.,
Boselli A. 2004, A&A, 2004, 419, 47
Font A.S. et al. 2008, MNRAS, 389, 1619
Forbes D.A., Lasky P., Graham A.W., Spitler L. 2008, 389, 1924
Gavazzi G., O’Neil K., Boselli A., van Driel W. 2006, A&A, 449, 929
Gavazzi G., Boselli A., Mayer L., Iglesias-Paramo J., V´ ılchez J.M., Car-
rasco L. 2001, ApJ, 563, L23
c ? 2010 RAS, MNRAS 000, 1−16