Mori. Not. R. Astron. Soc. 000, 000-000 (0000) Printed 18 April 2008 (MN WT^X. style file v2.2)
Compact groups in theory and practice - I. The spatial
properties of compact groups
oo
O
o
U
Oh!
<
00
Alan W. McConnachie 1 , Sara L. Ellison 1 , David R. Pattern 2,3
1 Department of Physics and Astronomy, University of Victoria, Victoria, B.C., V8P 1A1, Canada
2 Department of Physics and Astronomy, Trent University, 1600 West Bank Drive, Peterborough, ON, K9J 7B8, Canada
3 Visiting Researcher, Department of Physics and Astronomy, University of Victoria, Victoria, B.C., V8P 1A1, Canada
18 April 2008
6
>
00
(N
0\
(N
O
OO
O
X
ABSTRACT
We use a mock galaxy catalogue based upon the Millennium Run simulation to in-
vestigate the intrinsic spatial properties of compact groups of galaxies. We find that
approximately 30 % of galaxy associations identified in our mock catalogue are phys-
ically dense systems of four or more galaxies with no interlopers, approximately half
are close associations of 2, 3 or 4 galaxies with one or more interlopers, and the remain-
der are not physically dense (projections of looser groups and physically unassociated
galaxies). Thus the effect of interloping galaxies is significant. However, we find that
genuine compact groups are preferentially brighter and more isolated than those with
interlopers; by increasing the required minimum surface brightness of a group from the
canonical value of fi e — 26magsarcsec -2 to /.i e = 22magsarcsec -2 , we can increase
the proportion of genuinely compact systems identified with no interlopers from 29 %
to 75%. Of the genuine compact groups identified, more than half consist of a sin-
gle dark matter halo with all the member galaxies deeply embedded within it. In
some cases, there are other galaxies which share the same halo (typically with mass
~ 10 13 /i -1 M Q ) but which are not identified as being members of the compact group.
This implies that compact groups are associated with group environments, some or
all members of which are in the compact group. For those compact groups where all
galaxies are in the same halo, the three-dimensional velocity dispersion of the compact
group correlates broadly with the virial velocity of the dark matter halo. However, the
scale-size of the group - and hence the fraction of the halo mass which the group
samples - is completely uncorrelated with the properties of the dark matter halo. This
means that masses derived under the simple assumption of virial equilibrium using
the observed velocity dispersions and sizes of compact groups give incorrect estimates
of the true mass of the underlying dark matter.
Key words:
interactions
methods: statistical - catalogues - surveys - galaxies: general - galaxies:
1 INTRODUCTION
In broad terms, a compact group (CG) of galaxies is a
group of galaxies in which the typical inter-galactic sepa-
ration is of order the sca le of t he ga laxies . CGs were firs t
quantitatively defined by iRosd (119771 ) and Hickson (1982),
although their original identification significantly predates
these studies. Some of th e most famous examples of CGs in-
clude Stephan 's Quintet (|Stephanll877l ) and Seyfert's Sextet
(|Sevfertlll948l ). The advent of the first large area surveys,
especially the Palomar Observatory Sky Survey (POSS),
greatly increased the pace of discov ery of CGs. CGs make
up some of the objects identifie d by Vorontsov-Velvaminovl
i| 19591 ). IVorontsov-Vel'Yaminovl i|l977l ) and fArd (|l966l ). in
their catalogues of interacting galaxies. Studies initiated by
IShakhbazianl (1 19731 ) resulted in the identification of a few
hundr ed 'compact groups of co mpact galaxies'. Late r stud -
ies bv lHickson et alj i| 19771 ) and lHeiligman fc Turned (Il980l )
identified CGs using qualitative selection criteria based upon
their morphological appearance.
The most famous and well studie d catalogue of CGs is
the Hickson Compact Groups (HCGs; lHicksonlll982l ) which
were identified from the POSS using quantitative criteria
based on the number and flux density of galaxies in a group
(see Section 2.1 for a discussion of the selection criteria).
The original catalogue consists of 451 galaxies in 100 dif-
ferent groups. Since its creation, several other catalogues of
CGs have been compiled, inc luding (but not lim ited to) the
Digitized POSS CG catalog i|lovino et al.ll2003h . the South-
McConnachie et al.
ern CG catalogue l|Prandoni et al.lll994l; llovind[2 002), CGs
in the UZC galaxy ca talogue (IFocardi fc Kelmll2002T ). the
CfA2 redshift survey (IBarton et a l. 1996), the Las Cam-
panas redshift survey (lAllam fc Tuckerll2 000). and the SDSS
Commissioning Data (|Lee et alj|20040 .
When redshifts were measured for the members of
Stephan's Quintet and Seyfert's Sextet, it was found that
not all of the galaxies were at the same redshift. Indeed,
the physical nature of CGs has been a topic of debat e
for many dec ades (see, e.g., IfBurbidge fc Burbided Il961f ).
iMamonl (119861 ) argued that, instead of being physically dense
systems, roughly half of all CGs were line-o f-sight alignments
of gal axies within looser groups. However, iHickson fc Rood!
(1988) refuted this and found the probability of suc h an oc-
curren ce to be ~ 100 smaller than predicted by IMamonl
(|l986h . More recently, iHernquist et al.1 i|l995l ) have sug-
gested that CGs m ay be the result of vi ewing a cosmologi-
cal filament end-on. IPonman et al.l l|1996h have surveyed the
HCGs with ROSAT and implied that hot intra-group gas is
present in ~ 75 % of the systems. The presence of such gas
is strong evidence that the majority of CGs are physically
dense, even if some of their members are interlopers.
The apparent high densities of CGs c ombined with
their generally low velocity dispersions (eg. IHickson et al.l
1992), make them an intriguing environment in which to
study galaxy evolution. Mergers and interactions should be
commonplace, and CGs are potentially an ideal laboratory
to stu dy t hese processes . Calculations by ICarnevali et al.l
ljl98lTl and iBarnes! (|l989l ) suggest that CG galaxies should
merge together on a timescale equal to the crossing time of
the group, typically estimated to be of order 1 Gyr. If this is
true, then it suggests that CGs must be continually forming
in order to c ompensate for the g roups lost t hrough the merg-
ing pr ocess (|Diaferio et al.lll994r ). However. iGovernato et al.l
(|19911 ) show that, for the right choice of initial conditions,
CGs can potentially la st for as long as 9 Gyrs befo re merging
into a single remnant. I Athanassoula et al.l (|1997l ) show that
CGs can survive for much longer than a Hubble time if they
are embedded in massive dark matter haloes.
The selection criteria for CGs inevitably favor the ex-
trema of the population, or those whose orientations to our
line of sight are favourable. Thus, despite their existence
being known for many years, and despite large amounts of
observational effort in cataloging the properties of the HCGs
in particular, the physical nature of CGs and their role in
the galactic evolution process is still not clear. For example,
how dense is the CG environment? How many contain inter-
lopers? Are they dynamically stable? Are they a transient
stage in the evolution of looser groups? Do they represent a
distinct class of galaxy environment, or are they the extreme
end of a larger class? Are they the site of the formation of
early-type galaxies? What is their relation to so-called fossil
groups? What effect does the CG environment have on the
members of the group?
This is the first in a series of papers where we statisti-
cally examine the properties of CGs and their member galax-
ies based on large samples of CGs identified in cosmological
simulations and observational surveys. We begin this series
by examining the selection criteria of HCGs to determine
the three dimensional reality of these systems and the effect
of interlopers. This has been a long disputed argument in
the field. The advent of large cosmological simulations and
the ability to make realistic mock galaxy catalogues means
fresh insight can now be obtained. For the first time, we can
determine on a statistical basis how many of the HCGs are
likely to be genuinely compact systems and explore their
spatial properties.
Section 2 describes the Hickson selection criteria and
the use of the mock galaxy catalogue. In Section 3, we ap-
ply Hickson's criteria to this catalogue and examine the com-
pactness of the galaxy associations identified. In Section 4,
we optimise the selection criteria for CGs in photometric sur-
veys to minimise the effect of interlopers. In Section 5, we
take advantage of the cosmological simulations to examine
the halo properties of the CGs and their relation to observ-
able quantities. We defer the majority of the discussion of
our results until Section 6.
2 PRELIMINARIES
2.1 The Hickson criteria
IHickson! (J1982) define a CG as a set of galaxies in a photo-
metric survey with projected properties such that
(i) iV(Am = 3) >4;
(ii) 8 N >3 8 G ;
(iii) /i e < 26.0.
N (Am = 3) is the number of galaxies within 3 magnitudes
of the brightest galaxy and /x e is the effective surface bright-
ness of these galaxies in magnitudes per square arcsec where
the total flux of the galaxies is averaged over the smallest
circle which contains their geometric centres, and has an an-
gular diameter 8g ■ On is the angular diameter of the largest
concentric circle which contains no additional galaxies in
the magnitude range of the group or brighter. All magni-
tudes and surface brightnesses are measured in the r— band.
We hereafter refer to these three criteria collectively as the
'Hickson criteria'.
The identification of groups of only ~ 4 galaxies from
large photometric surveys where the only data available
are x,y projected positions and apparent magnitudes is a
difficult problem. The Hickson criteria were originally con-
structed in an attempt to quantify what had up until then
been identified by eye, and are inevitably subjective. Crite-
rion (i) ensures that the CG has enough members to con-
stitute a group. A magnitude range is imposed to ensure
all members are of broadly equivalent masses. Without it,
a galaxy and its satellites (such as the Milky Way and its
entourage of dwarf spheroidals) could be identified as a CG.
Criterion (ii) attempts to distinguish isolated groups from
parts of larger structures, such as cluster cores. Criterion (iii)
is, to first order, a distance independent means of defining
compactness.
These criteria have been subject to some controversy
since Hickson first introduced them. The basic issue is that
it is not clear what the physical, three-dimensional reality
of the identified galaxy associations are, given that the cri-
teria are applied to projected, two-dimensional datasets. In
particular, it is difficult to assess how homogeneous the iden-
tified galaxy associations are in terms of their physical com-
pactness and environment, and the contribution from inter-
lopers is unknown. These are fundamental problems for any
observational study of CGs.
Compact groups of galaxies 3
In what follows, we use the following terminology: 'Hick-
son association (HA) ' describes all of the galaxy associations
identified in the mock catalogue by application of the Hick-
son criteria. This is in recognition of the fact that some of
these associations will be genuinely compact configurations
in three dimensions, some will be projections of looser groups
and some may be entirely spurious line-of-sight chance align-
ments. The term 'compact association (CA)' is used to de-
scribe the subset of HAs which are truly compact in three
dimensions (we define 'compact' in Section 3.1).
2.2 The mock galaxy catalogue
The primary goal of this study is to assess the spatial and
dynamical properties of compact galaxy groups by using
simulations to relate three dimensional group properties to
those in redsh i ft spa ce. The Millennium Run simulation of
ISpringel et al.l (2005) is a natural starting point: this simu-
lation follows the evolution of dark matter within a cube
which is 500/i _1 Mpc on a side, from a redshift of 127
to the present day. Here, the Hubble constant is parame-
terised as H = 100 ftkms _1 Mpc _1 . Several galaxy cata-
logues have been created by applying s emi-analytical mod-
els to the output of thes e sim ulations llCroton et al.l 120061 :
|Pe Lucia fc Blaizotl 120071 ; iBower et al.1 120061 ). In principle,
any or all of these catalogues are suitable for the three-
dimensional requirements of our compact groups study.
In order to facilitate a direct comparison between simu-
lations and observations, we require a mock redshift sur-
vey from these galaxy catalogues. For this task, we use
the publicly available Blaizot Allsky catalogues available on
the Millennium websitqj. These mock catalogues were cre-
ated by applying th e Moc k Map Facility (MoMaF) c ode of
iBlaizot et all l|2005h to the lDe Lucia fc Blaizotl 1120071 ) semi-
analytic galaxy catalogues. The basic idea which creates the
mock catalogue is the replication of the data cube from the
simulations (which has periodic boundary conditions) in all
directions. Evolution is built in by using time snapshots
which correspond to the look-back time of each cube.
Five of the six available Allsky catalogs employ the
random tiling technique, which avoids spatial correlations
on large scales due to replication of the simulation cube.
We choose not to use these catalogues, since the disconti-
nuities which arise between adjacent data cubes may lead
to anomalous galaxy group detections. Instead, we use the
"Blaizot_Allsky_PT_l" catalogue, which preserves the pe-
riodicity of the density field. We note, however, that the
periodicity is not a concern on the small scales on which
compact galaxy groups are found. Moreover, even if a given
galaxy group is detected more than once, it will be viewed
from different vantage points, and is therefore likely to have
different observable properties.
The mock catalogue we use consists of ~ 5.7 million
galaxies brighter than the magnitude limit of tab — 18,
where r is the SDSS red filter. These catalogues contain the
key observable properties one would have in a flux-limited
survey: namely right ascension, declination and apparent
magnitu de. In addition, much of th e associated information
from the lDe Lucia fc Blaizotl f_2007) galaxy catalogues is also
http://www.mpa-garching.mpg.de/millennium/
provided, enabling a direct link to three dimensional posi-
tions and velocities, as well as intrinsic properties such as
absolute magnitudes and stellar mass. These properties will
be looked at in detail for th e compact groups we detect in
the next paper in this series IjBrasseur et al.ll2008l ). Here, we
are now in a position to detect compact groups in a flux
limited mock catalogue, and to relate the observed spatial
properties of these groups to their counterparts in three di-
mensions.
The results we derive are only strictly correct for the
mock galaxy catalogue from which they are derived, and
any significant change in the input physics of the simulations
on which they are based could affect our results. However,
the |Pe Lucia fe Blaizotl i|2007f ) semi-analytic model galaxy
ca talogues used in th is paper are closely related to those
of ICroton et al.l (120061 ); the latter demonstrated that their
galaxy catalogues are broadly consistent with observations
of low redshift galaxies, including the luminosity function,
the global star formation history, the Tully-Fisher relation,
and the colour-magnitu de distribution. In ad dition, the clus-
tering properties of the ICroton et alj (120061') catalog appear
to be well m atched to observations ( Springel et al.1 120051 ;
iLi et al.ll2007h .
Compact groups have, by definition, small scale radii,
and it is therefore important to check that the resolution
of the simulation is sufficient to identify these systems. The
spatial resolution of the Millennium sim ulation is 5 h~ kpc
l|Springel et al.ll2005l : ICroton et al.ll2006h . set by the soften-
ing length of the gravitational force. This resolution scale
is significantly less than the mean separation of galaxies in
compact groups (Section 3.1) and will not affect the ma-
jority of systems we identify. However, interactions between
galaxies at very small separations will not be followed accu-
rately.
Similarly, the mass resolution for the identification of
galaxy haloes in the simulation is important, particularly
during the merging of gala x ies. W hen two haloes start to
merge, |Pe Lucia fc Blaizotl ((2007) follow the merger until
the substructure drops below the mass resolution limit for
the identification of (sub-)haloes. At this point, the two
haloes are assumed to fully merge after a time t merge , which
is assumed to be twice the dynamical time. However, this
timescale is ad-hoc. Changing t„ lerge could affect the num-
ber of galaxies we identify as compact groups. To inves-
tigate whether this severely affects our results, we calcu-
late the distribution of sub-halo masses for all the galaxies
we identify as members of com pact groups (i.e., HAs; the
halo masses are provided in the lDe Lucia fc Blaizotl (|2007l )
catalogues). The mass resolution for the identification of
(sub-)haloes in the simulation is ~ 1.7 x 10 10 /i _1 Mq, and
we find that less than 1 % of galaxies in HAs have haloes
this small. Only 5% of the galaxies have haloes with mass
< 2 x lO lo /i~ 1 M0, and more than half of the haloes are
larger than the mass resolution limit by more than a fac-
tor of 10. The halo masses of CAs (genuinely dense groups
of galaxies) are higher still (Section 5). Thus possible un-
certainties associated with tm Erge are likely to have only a
small, if not negligible, effect on this study.
Future generations of cosmological simulations will have
improved spatial and mass resolution and will be able to
more accurately trace the formation and evolution of dense
groups of galaxies like compact groups. While this study -
McConnachie et al.
o 0.1 0.2 0.3 0.4
linking length (h~ Mpc)
=5 l
0.2h"'Mpc UT'Mpc
Compact assoc.
Groups?
Spurious?
-2-10 1 2
log 10 (linking length) (h~ Mpc)
Figure 1. Distribution of linking lengths of all identified Hick-
son associations (HAs) in the mock galaxy catalogue, shown on
linear (top panel) and log (bottom panel) scales. The Hickson cri-
teria preferentially identify galaxy associations with small linking
lengths, although there is a very significant tail of associations
with large linking lengths. Shaded areas show all HAs with link-
ing lengths less than 200 h _1 kpc.
like any based on mock catalogues - cannot be definitive
due to the many physical and computational uncertainties
inherent in simulating cosmological volumes, it is neverthe-
less impressive that the simulations are of sufficient quality
to start to provide a new handle on, and new insight into,
the properties of compact groups. Our results are as robust
as our current understanding allows, and should provide a
useful comparison to samples of compact groups extracted
from galaxy surveys such as the SDSS.
3 THE PREDICTED SPATIAL
CHARACTERISTICS OF COMPACT
GROUPS
3.1 Degree of compactness
We implement a search algorithm on the mock galaxy cat-
alogue using the Hickson criteria to identify all HAs. In
keeping with the original criteria, we use r— band appar-
ent magnitudes to calculate surface brightnesses. We find
15 122 HAs in the mock galaxy catalogue which satisfy the
Hickson criteria, consisting of a total of 64 525 galaxies. The
benefit of this approach is that we can now cross correlate
the two dimensional, projected, properties of each galaxy in
every HA with their de-projected properties to determine
the three dimensional reality of the HAs.
I = 0.2h"'Mpc
Compact assoc.
Iog 10 (l) (h" 1 Mpc)
Figure 2. Percentage of Hickson associations with linking lengths
< I. Approximately 29 % of all HAs (4446 out of 15122) have
linking lengths less than I = 200 h~ 1 kpc (shaded area).
The intrinsic compactness of a HA is fundamental to
studies of compact groups. The effective surface bright-
ness, /i e , is used in projection as a means of characteris-
ing compactness. Since three dimensional spatial informa-
tion is available to us, we choose to use the 'linking length',
£, as a quantitative and robust measure of the three dimen-
sional compactness of each of the 15 122 HA identified in
our mock catalogue. This parameter is used extensively in
friends-of-friends algorithms which are commonly used to
identify galaxy groups w hen redshift information is available
(jHuchra fc Gellerll 19821 ). For our purposes, we calculate the
value of £ for each HA in physical coordinates which corre-
sponds to the minimum linking length that would need to
be used to ensure all members of that HA were identified in
the simulation upon application of a friends-of-friends algo-
rithm. The use of linking length as a measure of compactness
has the added advantage that it is independent of the overall
shape of the group: for example, it does not discriminate be-
tween a roughly spherical group or an elongated filamentary
structure.
The distribution of three dimensional linking lengths for
the identified HAs is shown in Figure[T] The top panel shows
the distribution at small linking lengths (£ < 500 ft" 1 kpc)
on a linear scale whereas the lower panel shows the link-
ing lengths for all HAs on a log scale (where each bin has
been normalised by its width). The scale on which a HA
becomes a CA is, to some extent, subjective. However, the
top panel of Figure [l] reveals that Hickson's criteria prefer-
entially identify HAs with small linking lengths, of order a
few hundred kpc or less. Despite this, there is a tail to very
large values of £, as the lower panel of Figure [T] makes clear.
To quantify this further, Figure [5] shows the percentage of
HAs with linking lengths less than or equal to £.
By our definition, compact associations (CAs) have
small values of £, of order the scale of the galaxies. Looser
groups have intermediate values of £; for example, the linking
length of the brightest members of the Local Group, a rela-
tively sparse grouping of galaxies, is approximately 800 kpc.
Spurious detections, caused by line-of-sight alignments of
unassociated galaxies at different redshifts, yield very large
Compact groups of galaxies 5
values of £. Therefore, a linking length 2> 1 h^ 1 Mpc would
indicate that at least some of the galaxies in the HA are un-
likely to be physically related to one another. We have indi-
cated approximate 'boundaries' between CAs, loose groups
and spurious associations in the lower panel of Figure [l]
Given the shape of the distribution of linking lengths
in the top panel of Figure [l] and the curve of the distribu-
tion in Figure [2] £ — 200 h~ x kpc seems a natural division
below which groups are classed as compact. It is also approx-
imately the virial radius of L-k galaxy haloe s such as that of
the Milky Way and M31 JKlypin et al.ll2002T ). Since compact
groups are generally understood to be galaxy groups where
the typical inter-galactic separation is of order the scale of
the galaxies themselves, we adopt £ < 200 fo -1 kpc as the
appropriate scale for CAs. This corresponds to the shaded
areas of Figures [l] and [2] Defined in this way, ~ 29 % of all
HAs (4446 galaxy associations) are compact associations of
4 or more galaxies in three dimensions.
The boundary between compact associations and looser
arrangements of galaxies at £ = 200 ft -1 kpc is somewhat
subjective and could reasonably be set at a slightly higher
value to increase the fraction of associations we deem com-
pact. We note, however, that increasing the boundary by
50 % to £ — 300 h~ x kpc only increases the number of HAs
identified as CAs by 4% to 33%. This is probably as large
as the cut could reasonably be set given the shape of the
distribution of linking lengths in Figures 1 and 2, and it
does not significantly change the proportion of HAs we con-
sider to be compact. The cut could be set much smaller than
200 h" 1 kpc, of course, and this would reduce the number of
systems which we consider as CAs quite dramatically.
3.2 Interloping groups and interlopers within
groups
Figure[5]shows that ~ 71 % of HAs do not have small enough
values of I to be considered as CAs based on our definition of
compactness. This fraction is quite large, and would suggest
that the Hickson criteria do not efficiently identify compact
arrangements of galaxies. However, while some of the HAs
with large t will be completely spurious line-of-sight align-
ments of galaxies ('interloping groups'), another possibility
is that some of the galaxies within the HA form a compact
arrangement, and the large t value is the result of a minor-
ity of physically unassociated galaxies which happen to lie
along the same line of sight at a very different distance ('in-
terloping galaxy/galaxies'). There are famous examples of
this; for example, four out of the five galaxies which form
Stephan's Quintet are at the same redshift, whereas one is
not.
To gauge how this affects our results, we calculate how
many of the HAs with l > 200 h~ kpc can be decomposed
into a CA [I < 200/i _1 kpc) of 4 or more members and
one or more interlopers. The results are shown as a percent-
age of the total number of HAs in Figure [3] The first bin
is for all CAs with 4 or more galaxies and no interlopers,
and corresponds to the ~ 29% discussed in Section 3.1. The
remaining bins show the fraction of HAs which can be de-
composed into a CA with 4 or more genuine members plus
one or more interlopers (cross-hatched histograms). In total,
the percentage of HAs identified as CAs increases by ~ 6 %.
Of these 6 %, virtually none have more than 2 interloping
number of interlopers
Figure 3. Percentage of Hickson associations (HAs) which satisfy
our compactness criteria for compact associations (CAs) (linking
length < 200 h -1 kpc), allowing for the presence of interloping
galaxies. Approximately 35 % of HA consist of a compact configu-
ration of 4 or more galaxies, an additional 24 % consist of 3 galax-
ies in a compact configuration, and an additional 18 % of HAs
consist of a pair of galaxies separated by less than 200 ft -1 kpc.
However, many of these 'groups' would otherwise fail to satisfy
the selection criteria for HAs were it not for the interloper mem-
bers.
galaxies. Thus just over one-third (35 %) of identified HAs
are compact arrangements of 4 or more galaxies in three
dimensions, allowing for the presence of interlopers.
HAs are defined to have at least four members. How-
ever, given that interlopers are inevitably present in any ob-
servation sample of galaxy groups or clusters, we relax the
requirement that 4 member galaxies are required to define
a CA. Instead, we require a minimum of only 3 members,
still with £ < 200 /i -1 kpc (hatched histograms in Figure |3J).
By doing this, we find an additional ~ 24% of HAs can be
decomposed into a CA plus interlopers. This means that
~ 59% of identified HAs consist (at least in part) of a com-
pact arrangement of > 3 galaxies.
Finally, we determine that an additional 18 % of
HAs consist of a pair of galaxies separated by less than
200 h^ 1 kpc with at least 2 interlopers (blank histograms in
Figure [3}. The number of completely 'spurious' associations
identified by the Hickson criteria (that is, those where none
of the galaxies are closer than 200 ft -1 kpc to any other) is
therefore less than one in four; 77 % of HAs consist of 2 or
more galaxies which are physically associated on scales we
consider 'compact'.
It is important to emphasise that most of the compact
arrangements of galaxies with one or more interlopers would
otherwise fail to meet the criteria which define a HA were it
not for the interlopers; they would either not have enough
members or have too low a projected surface brightness to
qualify.
Figure 4 summarises the results of this section. The
finding that most (77%) HAs are physically close associa-
tions of galaxies is clearly significant, as is the result that
most of these physically close groupings usually consist of
one or more interlopers. Of the 64 525 galaxies which make
up all the identified HAs, 25 497 galaxies (~ 40 %) are not
6 McConnachie et al.
Galaxy pair
Galaxy triple
;
tu
Not physically
+
+
Compact
dense
interlopers
interlopers
+
Associations
( 23% )
( 18% )
( 24% )
X
o
D
en
( 29 % )
Compactness of Hickson associations
>
Figure 4. Summary of the spatial properties of the galaxy as-
sociations identified using Hickson's criteria for compact groups.
The area of the rectangles are proportional to the number of such
systems in our sample of 15 122 Hickson associations.
N
%CAs
% HAs
Am
% CAs
% HAs
4
100
29
3.0
100
29
5
19
27
2.5
79
30
6
1
22
2.0
57
29
7
1
17
1.5
31
27
e N /e G
%CAs
% HAs
ft,.
% CAs
%HAs
3
100
29
26
100
29
1
51
34
25
72
43
5
29
37
21
39
56
6
19
■11
23
16
67
Table 1 . The effect of varying the observable parameters for the
various selection criteria of HCGs. The standard Hickson criteria
have JV > 4, A m < 3 magnitudes, fi e < 26.0 mags arcsec -2 and
Sn/^g ^ 3. This identifies 4446 CAs in our catalogue (shown as
% of CAs = 100 in the table) which constitute 29 % of the HAs
identified. Each of the parameters is varied independently in this
table, with the other parameters set to their standard values.
physically close (< 200 h~ kpc) to other galaxies. The effect
of contamination from interlopers is clearly significant, and
we will discuss this in more detail in Section 6.
4 OPTIMAL SELECTION CRITERIA
Given the significant number of interloping galaxies iden-
tified, we now examine whether it is possible to alter the
Hickson criteria to reduce the 'contamination' levels, so that
fewer interloping groups and interloping galaxies are identi-
fied. The associations we are interested in are the physically
compact associations (CAs) of 4 or more galaxies with no
interlopers. This is a much more homogeneous population
than the overall sample of HAs, and are most analogous
to the systems that the Hickson criteria were designed to
identify. While pairs and triplets of galaxies are interesting
environments in their own right, they are clearly not what
the Hickson criteria were attempting to identify, insofar as
they do not have enough members to satisfy Hickson's def-
inition of a group. We therefore treat all galaxies which do
not belong to a CA as our 'contamination'.
We want to stay as true to the original definition of
(H)CGs as possib le. There f ore, w e look to fine-tune the se-
lection criteria of iHicksonl (|l982l ) to reduce contamination
rather than create a completely new set of criteria. The se-
lection criteria of Hickson detailed in Section 3.1 have 4 ob-
servable parameters whose values can be varied; these are
N (number of member galaxies), Am (r— band magnitude
range of members), /i e (effective surface brightness of group),
and 0n/8g (degree of isolation in projection). The distribu-
tion of each of these four observables for all HAs, CAs and
contamination identified in our mock catalogue are shown in
Figure [5] as blank, shaded and hatched histograms, respec-
tively.
Figure [5] shows the observed properties of the CAs com-
pared to the contaminants, and as such allows us to explore
how variations in the cuts for the observed parameters in-
crease or decrease the contamination rate. The dotted line
in each panel shows how the contamination level, expressed
as a fraction of the number of HAs, changes as a function of
each parameter (right vertical axis). Table [l] shows the net
effect of varying each of the four parameters in turn, with the
other three parameters set to their canonical values given in
Section 3.1. For each parameter, the first column gives the
value it is set to, the second column gives the number of
CAs retrieved (shown as a percentage of the number of CAs
obtained on using the standard values for all the parame-
ters, 4446) and the third column gives the number of CAs
retrieved as a percentage of the number of HAs retrieved.
Table[T]shows that the only observable parameter whose
value does not appear to significantly change the contami-
nation level is Am. Making Am smaller by a couple of
magnitudes has no significant change on the contamination
level of the sample retrieved. However, as the dotted line
in the upper-right panel of Figure [S] makes clear, the con-
tamination rate at large A m (2.5 — 3 magnitudes) and small
A m (0 — 0.5 magnitudes) is slightly higher than at interme-
diate values. Most of the galaxy associations are identified
at intermediate values, however, and so the overall contam-
ination levels for the cuts listed in Table 1 do not change
significantly.
More interestingly, Figure [5] and Table [T] reveal that the
level of contamination of more populated groups is higher
than less populated groups. This is unsurprising, given that
it is more likely that a HA with many members has an in-
terloper as a member than does a HA with fewer members.
However, in terms of reducing contamination, it is not worth
changing the requirements on how many galaxies define a
group: although contamination is higher for more populated
groups, there are very few groups which have more than 4
or 5 members in the first place.
The requirement that CAs should appear isolated can
be used to reduce the level of contamination. As Table 1 and
Figure 6 make clear, CAs preferentially appear more isolated
on the sky than the contaminants and selecting groups which
are increasingly distant from their nearest neighbour reduces
the contamination rate. The fraction of CAs present in a
sample of HAs can increase from 29% using the standard
criteria to 41% if we require 9n > 6 0G- A price is paid,
Compact groups of galaxies 7
o
o
1 1
OJ
o
Hie
Corr
_ _ - ■
CO
P
vz
1
O
^-
Cnntnminntinn X////\ .
o
CN
.
-
V////A
■i ' i ii 1 1 1
lO o
d
- r<V Q
O
o
°n/&g
K
CO 00
K
Me
Figure 5. Distribution of observable parameters in the Hickson criteria (number of galaxies, N; apparent r— band magnitude range of
members, Am; projected distance to nearest non-member galaxy, 0n/8g\ effective surface brightness, /i e ). Blank histograms show all
Hickson associations (HAs), shaded histograms show compact associations (CAs), and hatched histograms show contamination (where
we define all HAs which are not CAs as contamination). The dotted line in each panel shows the contamination level, expressed as a
fraction of the number of HAs, as a function of each parameter (right vertical axis).
however, since the sample size becomes less than one-fifth
of its original size.
By far the most striking feature of Figure 5 and Table 1
is the distribution of /j, e in the lower right panel for the CAs
compared to the contaminants. The number of contaminants
with a given surface brightness increases steeply towards
the faint end. In contrast, however, the number of CAs with
lower and lower surface brightness levels off and starts to de-
cline past a surface brightness of /i e — 24 — 25magsarcsec -2 .
This is to be expected: CAs need to have at least 4 galaxies in
a relatively compact configuration in order to be identified.
Therefore, the effective surface brightness of a CA (which
is to first order distance independent) cannot be arbitrarily
faint since the galaxies must all reside within a finite volume.
There is no such physical restriction on the contaminants.
The distribution of fi e for CAs will therefore peak in some
broad range, while that for the contamination will not.
Given the different shapes of the distribution of /x e for
CAs and the contaminants, adopting different cuts in fi e
has a dramatic effect on the contamination level of HAs, as
the dotted line in the lower right panel of Figure [5] and Ta-
ble 1 demonstrates. By requiring HAs to have a minimum
effective surface brightness of 25magsarcsec -2 instead of
26magsarcsec -2 , we increase the fraction of CAs identified
to 43% from 29%. The number of CAs identified is then 72 %
of the original number. Requiring a minimum surface bright-
ness of 24magsarcsec -2 means that over half of the identi-
fied HAs are CAs. At /i e = 23magsarcsec -2 , the proportion
is 67%. While not listed in Table 1, this trend continues to
fi e = 22magsarcsec -2 , where the proportion of CAs is 75%.
By this point, however, far fewer CAs are identified. Clearly,
searches for CAs which have higher surface brightness cuts
have significantly reduced contamination levels.
5 HALO OCCUPATION AND DYNAMICS
Despite being identified using only projected information,
the Hickson criteria do a remarkable job at retrieving dense
groupings of galaxies. With only moderate refinement of the
selection criteria we have shown that we can improve on
these criteria further to minimise interlopers. We now ask
how the observed dynamical properties of the member galax-
ies relate to the properties of the underlying dark matter.
5.1 Haloes versus galaxies
Figure 6 shows the number of separate haloes associated
with each CA previously identified in Section 3. To simplify
the analysis, we use only CAs with 4 members. Over half
of the CAs consist of only one halo, with all four galaxies
resident within that same halo. Of the rest, only a handful
contain four individual haloes (one per galaxy). There is
clearly not a one-to-one correspondence between haloes and
galaxies in CAs, and most galaxies in CAs are found within
the same halo. This is a strong affirmation of the physical
8 McConnachie et al.
-
55%
-
\
32%
I
11%
;
2%
-
.
Number of sub— haloes per CA
Figure 6. Histogram showing the number of distinct haloes in
each identified compact association with 4 members. In over half
of these compact associations all the member galaxies occupy the
same halo, confirming their physical association and co-evolution.
10
15
20
Number of galaxies per halo
Figure 7. Number of galaxies in each halo known to contain a
CA with four members. We consider only those galaxies within
three magnitudes of the brightest galaxy in the CA, to keep in
line with the original Hickson criteria. Most haloes only contain
these four galaxies, although some haloes contain a handful of
other galaxies which are not part of the CA. In more than half
of cases, CA galaxies generally appear to dominate the halo in
which they are found, although some are associated with larger
groups of typically 5-10 galaxies in total.
association of galaxies in a CA environment, and highlights
their obvi ous inter-dependence a nd co -evolutionary status.
The |Pe Lucia fc Blaizoj (|2007l ) catalogue records
whether a galaxy is at the center of its halo or if it is a
satellite. In 99 % of the cases where all member galaxies are
within the same dark matter halo, one galaxy is recorded
as being at the center of that halo. In the remaining 1% of
cases, all four galaxies are listed as satellites.
5.2 Halo occupation/environment of compact
groups
For those CAs with four members all embedded in the same
dark matter halo, we now ask if these are the only four galax-
^^^^T^
" " / '
/
/
/
/
7
' i ... .
500 1000 1500 2000
2000
4000
6000
8000
10"
Virial mass of halo (1O lo h" 1 M )
Figure 8. Top panel: histogram of the halo virial mass distri-
bution for CAs with 4 members all embedded in the same dark
matter halo. Bottom panel: the halo virial mass versus masses de-
rived using the velocity dispersion and scale radius for the CAs.
The dashed line shows a one-to-one correspondence. The inset
panel shows the same distribution on smaller scales. The derived
masses are a gross underestimate of the virial mass of the halo,
and the two are uncorrelated with one another.
ies in the halo, or if there are others embedded in the same
halo but which do not form part of the CA. This probes the
environment of CAs; if CAs are embedded in larger struc-
tures, such as groups or clusters, then we expect many other
galaxies to be resident in the same haloes.
Figure 7 shows the halo occupation distribution for
those haloes we know to contain a CA of four galaxies.
Only those galaxies within three magnitudes of the bright-
est galaxy in the CA are considered, to keep in line with the
original Hickson criteria. The large majority of these haloes
only contain the four galaxies of the CA, with the remainder
containing, at most, a handful of extra galaxies. Therefore,
most CAs dominate the haloes in which they are found, al-
though some are associated with larger groups of typically
5 — 10 galaxies in total. We have not probed the environment
of the halo, but such an analysis is beyond the scope of this
paper. We note, however, that it is unlikely that many CAs
are associated with galaxy clusters, since in this scenario we
would expect many other galaxies to share the same halo as
the CA.
5.3 Halo mass and group dynamics
The galaxies in CAs which occupy only one halo might be
viable tracers of the masses of these haloes. Indeed, the ve-
locity dispersion of compact groups is usually used to imply
a mass for the group, but small number statistics and the
necessary assumption of virial equilibrium means that it is
not clear that this method will work for these systems.
Compact groups of galaxies 9
r (h _1 Mpc)
Figure 9. Left panel: halo virial velocity, v vir , versus three-
dimensional velocity dispersion, a v , for CAs with 4 members all
embedded in the same dark matter halo. The dashed line shows
a one-to-one correspondence. The points represent the median
values of a v in lOOkms - 1 bins in v v i r . Horizontal error bars
show the width of the bin, and vertical error bars show the dis-
persion in <r„ in each bin. The two quantities correlate strongly
with one another, but the scatter in <r„ is significant, of order
100 — 300 kms -1 . Right panel: The virial radius of the halo, r v i r ,
versus scale radius of the CA (r s ; Equation 1). The dashed line
shows a one-to-one correspondence. The galaxies are deeply em-
bedded in the dark matter halo, and the two quantities are un-
corrected with one another.
demonstrated by the lack of a correlation between r V i r and
r s .
Using the observed a v and r s for the CAs, we can calcu-
late the mass that would be derived for the host halo upon
simple application of the virial theorem. The lower panel of
Figure 8 shows this mass in comparison to the true (virial)
mass of the dark matter halo. The dashed line shows a one-
to-one correspondence. The inset panel shows the same dis-
tribution on smaller scales. Clearly, the estimated mass is a
gross underestimate of the virial mass of the halo, but this
is to be expected; the galaxies are very embedded in the
dark matter halo and sample a smaller mass corresponding
to the innermost regions of the halo. Of greater interest is
the fact that the estimated mass and the virial mass are
uncorrelated. Given that r s and r v i r are uncorrelated, this
is unsurprising, but it means that simple mass estimates for
these galaxy groups do not trace, let alone estimate, the true
mass of the dark matter halo. We emphasise that the cause
of this discrepancy is due to the fact that, while the veloc-
ity dispersion of the group traces the virial velocity of the
group's dark matter halo, the scale radius of the group is
uncorrelated with the properties of the dark matter halo.
The top panel of Figure 8 shows the halo virial mass
distribution for all those CAs with 4 member galaxies which
all reside in the same halo. The typical halo mass is ~
10 13 hT 1 Mq - normal for a galaxy group - although there
are a considerable number more massive than this.
The three dimensional velocity dispersion, a v , of each
of these CAs is plotted as a function of the virial velocity
of the host halo, v v ir, in the left panel of Figure 9. Typi-
cal velocity dispersions are of order 300 — 500kms _1 . The
dashed line shows a one-to-one correspondence between v v ir
and <t v . The squares show the median velocity dispersion in
lOOkms -1 bins in virial velocity. The horizontal error bars
show the width of the bin, and the vertical error bars show
the standard deviation of the velocity dispersions in that bin.
Clearly, the velocity dispersion of the group broadly corre-
lates with the virial velocity of the host halo. The velocity
dispersion is systematically lower than the virial velocity,
and the spread about the median is very significant, of or-
der 100 — 300kms _1 . This means that adopting the velocity
dispersion as a proxy for the dark matter halo virial velocity
for any individual system is going to introduce a significant
source of uncertainty.
To calculate the mass of the halo via the virial theorem
requires a representative scale radiu s for the galaxy group.
Following iBinnev fc Tremainei (|1987l ). we use
(1)
^fcrl^K' < \Ri ~ -ftj
>
which is a measure of the typical intergalactic separation
in the group. The second panel of Figure 9 shows r s as a
function of the virial radius of the halo, r v i r . The dashed
line represents a one-to-one correspondence between these
quantities. The scale radius of the CA is much smaller than
the virial radius of the dark matter halo, showing that the
CA is deeply embedded in the central regions of the host
halo (in Section 5.1 we showed that one galaxy in the CA
is usually at the center of the halo). Further, the degree
to which the CA is embedded is independent of r v i r , as
6 DISCUSSION
In this paper we have explored the three dimensional spatial
properties of Hickson compact gr oups of galaxies iden tified
in a mock galaxy catalogue from iBlaizot et al.l i|2005h . We
now discuss our main results in the context of previous ob-
servational studies.
6.1 How compact are compact groups?
The compactness of compact groups has been an issue of
some considerable debate. There have been many previous
interpretations as to wh at compact grou ps are, summarised
in the Annual Review bv lHicksonl (|1997l ). Of particular rele-
vance, |Mamon| (1986) argued that roughly half of all HCGs
were chance al ignments of galax i es wit hin looser groups, a
view echoed bv lWalke fc Mamonl (|l989h and lMamon] l|l995l ).
We find that approximately 30 % of HAs in our mock
catalogue are physically dense systems of four or more galax-
ies with no interlopers. Nearly one-quarter of HAs are not
physically dense, even in par t. This latter fr action is signifi-
cantly less than suggested bv lMamonl (|l986T ). but it is still a
considerable fraction. The remaining HAs (48 %) are dense
configurations of 2, 3 or 4 galaxies with one or more inter-
lopers. Thus the majority (over three-quarters) of HAs are
physically dense associations of galaxies, although most of
these contain interlopers and would otherwise fail to satisfy
the selection criteria for HAs were it not for the presence of
these interloping members. In terms of individual galaxies,
25 497 of the 64 525 galaxies identified by the Hickson crite-
ria (40 %) are not found in very close proximity to at least
one other galaxy, and will no doubt a ct to bias any study o f
gala xy properties i n comp act groups (|Brasseur et al.l l2008).
IPonman et al.l (1996) surveyed most of the HCGs with
ROSAT. They calculated that the X-ray emission they de-
tected was greater than would be expected from each galaxy
in isolation, and implied that most HCGs must therefore be
physical associations of galaxies, with the excess emission
10 McConnachie et al.
associated with hot inter-galactic gas in the group. They
implied that > 75 % of the HCGs were physical associations
of galaxies, by extrapolation of the number of associations
from which they measured detectable X-ray emission. We
find that 59 % of HAs are dense groupings of 3 or more
galaxies, and this figure rises to 77% if pairs are included.
If these systems contain some intra-group ho t gas, then our
findin g would be in good agreement with the lPonman et al.l
(|l996h result.
6.2 Are Hickson's criteria optimal?
The identification of a small number of physically associ-
ated galaxies based only on projected positions and appar-
ent magnitudes is a difficult task, and we have found that
most HAs are compact groupings of galaxies. However, only
29 % of HAs consist of 4 or more galaxies in a compact con-
figuration with no interlopers. The remaining HAs can, in
some respect, be viewed as contamination of this sample.
When we compare the observable properties of CAs
with the 'contaminants' in Figure 5 and Table 1, we find
that less populated, brighter, more isolated HAs are more
likely to be CAs than more populated, fainter, and/or less
isolated HAs. Table Q] shows that the success rate of iden-
tifying HA can be significantly improved by changing the
selection criteria. The best criterion to change in this re-
spect is the surface brightness limit; increasing the cut from
fi e — 26magsarcsec -2 to /i e = 23magsarcsec -2 increases
the fraction of CAs identified from 29% to 67%. This frac-
tion can be increased to three-quarters by increasing the
surface brightness limit to fj, e — 22magsarcsec~ 2 . This is
done at a cost, of course, as fewer HAs identified with the
original criteria pass these more strict requirements. Based
upon these experiments, we conclude that, if you want to
identify a sample of compact groups with as few interlopers
as possible, then you should use selection criteria which only
acc ept the bri g htest, most isolated, compact groups.
iLee et al.l (J2004J) slightly modified the Hickson criteria
in their search for compact groups in the SDSS commission-
ing data, such that groups had to have an effective surface
brightness of fj, e < 24magsarcsec -2 . This will have reduced
the expected contamination level to ~ 44 %, significantly less
than would have been obtained using the original criteria.
In a subsequent paper in this series we use the informa-
tion presented here to derive a large catalogue of CAs from
the SDSS photometric catalogue. Importantly, for the first
time in a study of compact groups identified in photomet-
ric catalogues, we will have a strong understanding of our
sample selection and can be statistically confident in the
identification of compact groups with minimal contamina-
tion from interlopers.
6.3 What are the dark matter properties of
compact groups?
We have selected compact groups based solely on the proper-
ties of the luminous galaxies, independent of any dark mat-
ter properties. It is therefore satisfying to see that, in making
this selection, we preferentially identify galaxies which be-
long to the same dark matter halo. In over half the compact
groups (55%), all the member galaxies are found within the
same dark matter halo. A smaller number have two or three
haloes (32% and 11%, respectively), and only 2% of com-
pact groups are made of galaxies which each have their own
dark matter halo.
We interpret the above result in terms of formation
time. That is, those compact groups where each galaxy has
its own halo are dynamically young, and the haloes have not
yet had time to merge. In contrast, those compact groups
with only one halo are dynamically older, and the haloes
have fully merged. However, the mock catalogs used in this
study do not contain information about the evolutionary
history of these galaxy groups and the dark matter haloes
that they occupy. A proper treatment of the evolutionary
properties of these CAs would require analysis of the full
de Lucia catalogs. As a result, a more detailed treatment of
this topic is beyon d the s co pe of this current c ontrib ution.
IVennik et all (ll993T ). iRood fc Strublei
iRamella et all (|l99- ~
(| 19941 ) and
showed that the majority of the HCGs
seemed to be associated with loose groups and clusters, al-
though these w ere not general l y in the highest den sity en-
vironments fee. ISulenticlll987l ; IPalumbo et allll995h . More
recently, an association between loose groups and a subse t
of the HCGs was a lso f ound by Tovmassian et al.l |2006).
iDiaferio et~aJ1 (|l994T ) and lGovernato et all (|l996T ) have both
suggested that compact groups may arise as dense config-
urations within looser groups. For those CAs where all the
galaxies reside within the same halo, we find that they gener-
ally dominate the halo. However, in a significant proportion
of cases, there are several other galaxies in the halo which
are not part of the CA, implying that the CA is part of a
looser group with typically 5-10 galaxies in total. The virial
masses of these haloes are of order 10 13 — 10 14 ft -1 Mq . Thus
our results for both the halo mass and the number of galax-
ies within the halo strongly suggest that compact groups
are associated with galaxy groups, some or all of which have
formed a dense galaxy grouping. This is in good agreement
with recent observational studies, and argues against com-
pact groups being associated with galaxy clusters.
Finally, we examine whether it is possible to relate the
observed dynamics of compact groups to the mass of the
host halo. We find that there is a strong statistical corre-
lation between the host halo virial velocity and the three
dimensional velocity dispersion of the group. However, the
significant spread in this trend - likely a result of the ve-
locity dispersion being based on only four galaxies - means
that a significant uncertainty will be introduced if the ve-
locity dispersion is used as a proxy for the virial velocity for
any individual system.
Our results indicate that compact groups are deeply em-
bedded near the centre of their dark matter halo. Unfortu-
nately for observers, the degree to which they are embedded
is uncorrelated with the virial radius of the halo. This re-
sult suggests a picture in which the individual galaxies and
their haloes merge together and the galaxies sink towards
the centre of their new host halo until they are within a few
hundred kiloparsecs of each other, at which point they can
be identified as a compact group. Thus derived mass esti-
mates of the halo do not correlate with the true mass of the
halo; by the time the group is dense enough to be able to be
identified as a compact group, the linear size of the group
- and hence the fraction of the halo mass which the group
samples - is independent of the properties of the surround-
Compact groups of galaxies 11
ing dark matter halo. In practice, this result will require
more precise modeling of observed compact group dynamics
in order to determine their dark matter halo properties.
ACKNOWLEDGMENTS
The Millennium Simulation databases used in this paper
and the web application providing online access to them
were constructed as part of the activities of the German As-
trophysical Virtual Observatory. AWM acknowledges sup-
port from a Research Fellowship from the Royal Commission
for the Exhibition of 1851. He also thanks Sara Ellison and
Julio Navarro for additional financial assistance. We thank
Jon Willis, Crystal Brasseur and Jorge Penarrubia for useful
and stimulating discussions of these results. SLE and DRP
acknowledge the receipt of NSERC Discovery Grants which
funded some of this research. We thank the anonymous ref-
eree for very thoughtful and useful suggestions which led to
significant improvements in this paper.
REFERENCES
Allam, S. S. & Tucker, D. L. 2000, Astronomische
Nachrichten, 321, 101
Arp, H. 1966, ApJS, 14, 1
Athanassoula, E., Makino, J., & Bosma, A. 1997, MNRAS,
286, 825
Barnes, J. E. 1989, Nature, 338, 123
Barton, E., Geller, M., Ramella, M., Marzke, R. O., & da
Costa, L. N. 1996, AJ, 112, 871
Binney, J. & Tremaine, S. 1987, Galactic dynamics (Prince-
ton, NJ, Princeton University Press, 1987, 747 p.)
Blaizot, J., Wadadekar, Y., Guiderdoni, B., Colombi, S. T.,
Bertin, E., Bouchet, F. R., Devriendt, J. E. G, & Hatton,
S. 2005, MNRAS, 360, 159
Bower, R. G, Benson, A. J., Malbon, R., Helly, J. C,
Frcnk, C. S., Baugh, C. M., Cole, S., & Lacey, C G 2006,
MNRAS, 370, 645
Brasseur, C, McConnachie, A. W., Ellison, S., & Patton,
D. 2008, MNRAS, submitted
Burbidge, E. M. & Burbidge, G R. 1961, AJ, 66, 541
Carnevali, P., Cavaliere, A., & Santangelo, P. 1981, ApJ,
249, 449
Croton, D. J., Springel, V., White, S. D. M., De Lucia,
G, Frenk, C. S., Gao, L., Jenkins, A., Kauffmann, G,
Navarro, J. F., & Yoshida, N. 2006, MNRAS, 365, 11
De Lucia, G & Blaizot, J. 2007, MNRAS, 375, 2
Diaferio, A., Geller, M. J., & Ramella, M. 1994, AJ, 107,
868
Focardi, P. & Kelm, B. 2002, A&A, 391, 35
Governato, F., Bhatia, R., & Chincarini, G 1991, ApJ, 371,
L15
Governato, F., Tozzi, P., & Cavaliere, A. 1996, ApJ, 458,
18
Heiligman, G M. & Turner, E. L. 1980, ApJ, 236, 745
Hernquist, L., Katz, N., & Weinberg, D. H. 1995, ApJ, 442,
57
Hickson, P. 1982, ApJ, 255, 382
— . 1997, ARA&A, 35, 357
Hickson, P., Mendes de Oliveira, C, Huchra, J. P., &
Palumbo, G G 1992, ApJ, 399, 353
Hickson, P., Richstone, D. O, & Turner, E. L. 1977, ApJ,
213, 323
Hickson, P. & Rood, H. J. 1988, ApJ, 331, L69
Huchra, J. P. & Geller, M. J. 1982, ApJ, 257, 423
Iovino, A. 2002, AJ, 124, 2471
Iovino, A., de Carvalho, R. R., Gal, R. R., Odewahn, S. C,
Lopes, P. A. A., Mahabal, A., & Djorgovski, S. G 2003,
AJ, 125, 1660
Klypin, A., Zhao, H., & Somerville, R. S. 2002, ApJ, 573,
597
Lee, B. C, Allam, S. S., Tucker, D. L., Annis, J., Johnston,
D. E., Scranton, R., Acebo, Y., Bahcall, N. A., Bartel-
mann, M., Bohringer, H., Ellman, N., Grebel, E. K., In-
fante, L., Loveday, J., McKay, T. A., Prada, F., Schneider,
D. P., Stoughton, C, Szalay, A. S., Vogeley, M. S., Voges,
W., & Yanny, B. 2004, AJ, 127, 1811
Li, C, Jing, Y. P., Kauffmann, G, Borner, G, Kang, X.,
& Wang, L. 2007, MNRAS, 376, 984
Mamon, G A. 1986, ApJ, 307, 426
Mamon, G A. 1995, in Astronomical Society of the Pacific
Conference Series, Vol. 70, Groups of Galaxies, ed. O.-G
Richter & K. Borne, 83
Palumbo, G G C, Saracco, P., Hickson, P., & Mendes de
Oliveira, C. 1995, AJ, 109, 1476
Ponman, T. J., Bourner, P. D. J., Ebeling, H., & Bohringer,
H. 1996, MNRAS, 283, 690
Prandoni, I., Iovino, A., & MacGillivray, H. T. 1994, AJ,
107, 1235
Ramella, M., Diaferio, A., Geller, M. J., & Huchra, J. P.
1994, AJ, 107, 1623
Rood, H. J. & Struble, M. F. 1994, PASP, 106, 413
Rose, J. A. 1977, ApJ, 211, 311
Seyfert, C. K. 1948, AJ, 53, 203
Shakhbazian, R. K. 1973, Astrofizika, 9, 495
Springel, V., White, S. D. M., Jenkins, A., Frenk, C. S.,
Yoshida, N., Gao, L., Navarro, J., Thacker, R., Croton, D.,
Helly, J., Peacock, J. A., Cole, S., Thomas, P., Couchman,
H., Evrard, A., Colberg, J., & Pearce, F. 2005, Nature,
435, 629
Stephan, M. 1877, MNRAS, 37, 334
Sulentic, J. W. 1987, ApJ, 322, 605
Tovmassian, H., Plionis, M., & Torres-Papaqui, J. P. 2006,
A&A, 456, 839
Vennik, J., Richter, G M., & Longo, G 1993, Astronomis-
che Nachrichten, 314, 393
Vorontsov-Velyaminov, B. A. 1959, in Atlas and catalog of
interacting galaxies (1959)
Vorontsov-VePYaminov, B. A. 1977, A&AS, 28, 1
Walke, D. G & Mamon, G A. 1989, A&A, 225, 291
Live Music Archive
Librivox Free Audio
Metropolitan Museum
Cleveland Museum of Art
Internet Arcade
Console Living Room
Books to Borrow
Open Library
TV News
Understanding 9/11