arXiv:0810.1090v1 [astro-ph] 7 Oct 2008
Draft version October 7, 2008
Preprint typeset using LATEX style emulateapj v. 03/07/07
ΛCDM SATELLITES AND H I COMPANIONS —
THE ARECIBO ALFA SURVEY OF NGC 2903
Judith A. Irwin
Department of Physics, Engineering Physics and Astronomy, Queen’s University, Kingston, ON, Canada, K7L 3N6
G. Lyle Hoffman
Department of Physics, Lafayette College, Easton, PA, 18042
Department of Physics, Royal Military College of Canada, PO Box 17000, Station Forces, Kingston, ON, Canada K7K 7B4
Martha P. Haynes1
Center for Radiophysics and Space Research, Cornell University, Ithaca, NY, 14853, USA
Center for Radiophysics and Space Research, Cornell University, Ithaca, NY, 14853, USA
Suzanne M. Linder
Hamburger Sternwarte, Universit¨ at Hamburg, Gojenbergsweg 112, D-21029, Hamburg, Germany
Max-Planck-Institut f¨ ur Astrophysik, D-85748 Garching, Germany
NAIC-Arecibo Observatory, HC3 Box 53995, Arecibo, PR, 00612, USA
B¨ arbel S. Koribalski
Australia Telescope National Facility, CSIRO, Epping, NSW 1710, Australia
School of Physics and Astronomy, Cardiff University of Wales, Cardiff, CF24 3YB, UK
Centre for Astrophysics Research, Science and Technology Research Institute, University of Hertfordshire, Hatfield, AL10 9AB, UK
W. J. G. de Blok
Department of Astronomy, University of Cape Town, Rondebosch 7700, South Africa
Mary E. Putman
Department of Astronomy, Columbia University, New York, NY, 10027, USA
Wim van Driel
Observatoire de Meudon, 5 Place Jules Janssen, 92195 Meudon, France
Draft version October 7, 2008
We have conducted a deep, complete H I survey, using Arecibo/ALFA, of a field centered on the
nearby, isolated galaxy, NGC 2903, which is similar to the Milky Way in its properties. The field size
was 150 kpc × 260 kpc and the final velocity range spanned from 100 to 1133 km s−1. The ALFA
beams have been mapped as a function of azimuth and cleaned from each azimuth-specific cube prior
to forming final cubes. The final H I data are sensitive down to an H I mass of 2×105M⊙ and
column density of 2 × 1017cm−2at the 3σ2δV level, where σ is the rms noise level and δV is the
velocity resolution. NGC 2903 is found to have an H I envelope that is larger than previously known,
2 Irwin et al.
extending to at least 3 times the optical diameter of the galaxy. Our search for companions yields
one new discovery with an H I mass of 2.6 × 106M⊙. The companion is 64 kpc from NGC 2903 in
projection, is likely associated with a small optical galaxy of similar total stellar mass, and is dark
matter dominated, with a total mass > 108M⊙. In the region surveyed, there are now two known
companions: our new discovery and a previously known system that is likely a dwarf spheroidal,
lacking H I content. If H I constitutes 1% of the total mass in all possible companions, then we should
have detected 230 companions, according to ΛCDM predictions. Consequently, if this number of dark
matter clumps are indeed present, then they contain less than 1% H I content, possibly existing as
very faint dwarf spheroidals or as starless, gasless dark matter clumps.
Subject headings: galaxies: individual (NGC 2903) — galaxies: spiral — galaxies: formation —
cosmology: dark matter — radio lines: ISM
Weak gravitational lensing studies have shown that
stellar light traces dark matter on supercluster and
cluster scales (Heymans et al. 2008).
much less clear on sub-galactic scales, however, as ev-
idenced by the well-known ‘missing satellites’ prob-
lem around the Milky Way (MW). At issue is the
fact that the predicted number of satellites based on
Cold Dark Matter (CDM) and Λ Cold Dark Mat-
ter (ΛCDM) simulations of galaxy formation is signifi-
cantly greater than the observed number of dwarf MW
companions (Kauffmann et al. 1993; Klypin et al. 1999;
Moore et al. 1999; Diemand et al. 2007b).
A number of explanations for this discrepancy
have been proposed. These include the suppres-
sion of star formation due to the
the gas(Barkana & Loeb 1999;
Shaviv & Dekel 2003;Gnedin et al. 2008)
ciation of molecular hydrogen (Haiman et al. 1997),
gas-stripping due to supernovae-driven winds from
an early star-formation epoch (Hirashita et al. 1998;
Klypin et al. 1999), only
substructures forming stars
ping/stirring (Klypin et al. 1999;
Mayer et al. 2001b;Kravtsov et al. 2004),
between dwarf satellites and high velocity clouds (HVCs,
see Klypin et al. 1999 for a summary), the suppression
of small-scale power in simulations via contributions
from Warm Dark Matter (Avila-Reese et al.
Zentner & Bullock 2002), and incompleteness in the
census of MW satellites (Mateo 1998).
It is now clear that the latter explanation has played
a part, since in the past few years, the known popula-
tion of dwarf MW satellites has almost doubled based
The issue is
Benson et al. 2002;
(Stoehr et al. 2002),
Mayer et al. 2001a;
Electronic address: firstname.lastname@example.org
Electronic address: email@example.com
Electronic address: firstname.lastname@example.org
Electronic address: email@example.com
Electronic address: firstname.lastname@example.org
Electronic address: email@example.com
Electronic address: bcatinel@MPA-Garching.mpg.de
Electronic address: firstname.lastname@example.org
Electronic address: Baerbel.Koribalski@csiro.au
Electronic address: email@example.com
Electronic address: firstname.lastname@example.org
Electronic address: email@example.com
Electronic address: firstname.lastname@example.org
Electronic address: email@example.com
1National Astronomy and Ionosphere Center, Cornell Univer-
sity, Ithaca, NY, 14853.
2NAIC-Arecibo Observatory, HC3 Box 53995, Arecibo, PR,
3NRAO, PO Box O, Socorro, NM, 87801, USA
on scrutiny of the Sloan Digital Sky Survey (SDSS, York
et al. 2000, see also Koposov et al. 2008); however, al-
though the detection of the new satellites alleviates the
problem, it does not eliminate it entirely since, after mak-
ing corrections for sky coverage, the discrepancy is still
approximately a factor of 4 (see Simon & Geha 2007).
The concept that star formation might, in some way,
have been suppressed in systems of low total mass, sug-
gests the possibility that dark matter substructure could
still be traced by atomic hydrogen (H I) even though ap-
preciable stellar content may be missing. However, many
searches, of variable sensitive and coverage, have been
undertaken for low mass starless companions, with little
success (see Sect. 2) and some have suggested that small
galaxies which retain H I are likely to have developed
stars as well (Briggs 2004; Taylor & Webster 2005).
The advent of the 7-beam Arecibo L-band Feed Array
(ALFA, see Giovanelli et al. 2005a4) has now provided
an opportunity to survey nearby systems for the possible
presence of such ‘dark companions’, with unprecedented
sensitivity, coverage, and speed. We report here the re-
sults of a targeted deep survey of a single, isolated galaxy,
NGC 2903 (see Sect. 3). In this paper, we outline our
observational procedure and data reduction, we present
global results for NGC 2903 and then concentrate on
companions and their implications for primordial dark
matter searches. The details of the H I in NGC 2903, it-
self, will be left for a subsequent paper. Please note that
data related to this project, including some software that
we have developed, data from related observations, our fi-
nal cubes and beam maps can be found on our NGC 2903
In Sect. 2, we outline previous surveys that have taken
place, presenting a comparison with our approach and, in
Sect. 3, we discuss NGC 2903 and its environment. Since
this paper introduces new techniques for observing and
reducing Arecibo/ALFA data, we discuss these in detail
in Sects. 4 and Sect. 5. Our detection thresholds and data
quality are given in Sect. 6, Sect. 7 presents the results for
NGC 2903 and its environment, and Sect. 8 and Sect. 9
provide the discussion and conclusions, respectively.
2. PREVIOUS H I SURVEYS AND COMPARISON WITH
Current observational data suggest that H I clouds
tend not to be ‘intergalactic’, but rather associated
with galaxies (Briggs 2004). Various searches for faint
H I around galaxies, however, have typically been ham-
4See also http://www.naic.edu/alfa/
NGC 2903 and its Environment3
Fig. 1.— Sloan Digital Sky Survey (SDSS DR6) image of the field around NGC 2903 that has been surveyed by the Arecibo telescope.
NGC 2903 and its companion, UGC 5086, are labelled, as well as the scale. The background galaxy, NGC 2916, is also labelled (see
pered by the need to choose sensitivity at the expense of
coverage or vice versa.
For wide coverage, the blind H I Parkes All Sky Sur-
vey (HIPASS Barnes et al. 2001; Koribalski et al. 2004;
Meyer et al. 2004) revealed only one definite extragalac-
tic H I cloud in the NGC 2442 group with a high H I mass
of about 109M⊙ (Ryder et al. 2001). This cloud has
been interpreted as a remnant of a tidal interaction
(Bekki et al. 2005) and cannot be considered primordial.
As for targeted searches, lower mass limits have been
achieved. Zwaan (2001) and de Blok (2002) made in-
complete samplings of several galaxy groups to limits of
a few ×106M⊙. Minchin et al. (2003) surveyed the
Cen A group to a limit of 2 × 106M⊙, and Pisano
et al. (2004, 2007) observed 6 Local Group analogs
to 2 − 5 × 106M⊙. Kova˘ c et al. (2005) completely
surveyed the Canes Venatici group to a limit of 106-
107M⊙. Barnes & de Blok (2004) searched for faint
H I companions around NGC 1313 and Sextans A to ∼
106M⊙. Pisano & Wilcots (1999, 2003) searched for gas
rich companions around 6 isolated galaxies to an approx-
imate detection limit of only 108M⊙6. In these and other
targeted surveys (see also Kilborn et al. 2006), neither
starless H I companions nor HVCs, where the sensitiv-
ity was sufficient (e.g. Pisano et al. 2007), were detected
with the exception of H I that could again be attributed
to tidal debris (e.g. see Bekki et al. 2005).
In contrast to the targeted searches described above,
our study of a relatively isolated system (see Sect. 3) sim-
plifies the interpretation of any H I detections since tidal
explanations are much less likely. Moreover, use of the
305 m diameter Arecibo radio telescope has placed these
observations among the most sensitive yet achieved (see
Sect. 6.1). While slightly lower H I mass limits have been
claimed in deep interferometric observations of NGC 891
6Note that the various groups have not all used the same criteria
for determining their mass limits.
(Oosterloo et al. 2007), M 31 and M 33 (Westmeier et
al. 2005), our combination of low mass detection limits,
the lowest H I column density limits yet achieved in such
studies, the sensitivity to broad scale structure not pos-
sible via interferometers, and the complete (and large)
sky coverage combine to make this survey unique.
3. NGC 2903 AND ITS ENVIRONMENT
NGC 2903 (Table 1 and Fig. 1) has a number of assets
that make it a good target for deep H I mapping. It falls
within the declination range of the Arecibo telescope, it is
bright and massive so there is a reasonable expectation of
the presence of ΛCDM (or other) satellites, it is of large
angular size so is easily resolved by the Arecibo beam, it
is nearby yet lies beyond the Local Group (D = 8.9 Mpc;
Drozdovsky & Karachentsev 2000; 1′= 2.6 kpc7), and
some previous H I observations of the galaxy are available
for comparison (Begeman 1987; Begeman et al. 1991;
Hewitt et al. 1983; Haynes et al. 1998). An important
characteristic is that it is non-interacting and isolated,
in the sense that no galaxies larger than one quarter of
its optical size are present within 20 optical diameters
away (No. 0347 in the Catalogue of Isolated Galaxies,
Karachentseva 19738, see also Haynes et al. 1998).
NGC 2903 is characterized by its barred, grand-design
spiral pattern. It displays a number of ‘hot spots’ in
its nuclear region as well as a ring of star formation
(e.g. P´ erez-Ram´ ırez et al. 2000). The nuclear dust dis-
tribution is chaotic (Martini et al. 2003). The CO emis-
sion is concentrated along the bar (Regan et al. 1999)
and the star formation rate (SFR) per unit area is en-
hanced by an order of magnitude in the nucleus in com-
7Literature values range from 6.01 Mpc to 11.65 Mpc, depending
on corrections for local motions (see the NASA/IPAC Extragalactic
8At the time of writing, NGC 2903 has not been included in the
Analysis of the interstellar Medium of Isolated Galaxies (AMIGA)
catalogue (e.g. Verley et al. 2007) due to its large angular size.
4 Irwin et al.
parison to the disk (Jackson et al. 1991). A soft X-ray
halo extending to the west of the nucleus has been inter-
preted as outflow from a nuclear starburst-driven wind
(Tsch¨ oke et al. 2003).
Aside from evidence of nuclear star formation, how-
ever, NGC 2903 is a typical massive spiral whose prop-
erties are similar to those of the Milky Way. Its global
SFR (2.2 M⊙yr−1, Table 1) is comparable to the MW
value (∼ 4 M⊙yr−1, Diehl et al. 2006), considering the
different methods for estimating this value. More impor-
tantly, its rotation curve shows a rise to 210 km s−1at
a galactocentric radius of R∼4 kpc, declining slightly to
180 km s−1by R∼33 kpc, its outermost measured point
(Begeman et al. 1991). These values agree with the ro-
tation curve of the Milky Way over 4≤R (kpc)≤33 to
within error bars (Xue et al. 2008). Aside from environ-
ment, therefore, NGC 2903 appears to be an analog of
As indicated above, NGC 2903 is considered to be
an isolated galaxy, given the dearth of nearby compan-
ions sufficiently massive to perturb it.
small companions, UGC 5086 and D565-06, are known to
be associated (Drozdovsky & Karachentsev 2000) and,
from an optical search for additional possible compan-
ions within similar radii, we have now identified a third
companion, D565-10. D565-10 was found from a search
over the spatial region and velocity range within which
UGC 5086 and D565-06 have previously been found.
Since its separation from NGC 2903 in both position and
velocity space is less than that of D565-06, we include it
as a newly identified companion here. These three galax-
ies and their known properties are listed in Table 2. Of
the three, only UGC 5086 is within our surveyed field of
view and is labelled in Fig. 1.
Aside from the H I observations listed above, more re-
cent H I data from the HIPASS (Wong et al. 2006) and
the Westerbork SINGS survey (Braun et al. 2007) are
now also available. NGC 2903 is also in The H I Nearby
Galaxy Survey (THINGS, Walter et al.
makes use of Very Large Array (VLA) data.
time of our observations, five archival VLA unpublished
H I data sets were available, all of which we have reduced.
Of these, two sets produced good data. These are: a) ob-
serving run AO125, taken 29 Sept. 1996 constituting 2.03
hours on source in D configuration, and b) run AW536,
taken 22 Apr. 2000, constituting 2.68 hours on source in
C configuration. We do not reproduce the VLA cubes
here, but make them available on our NGC 2903 website
(see Footnote 5), and refer to them, as needed, only for
comparison purposes. The VLA data sets are of higher
spatial resolution than the Arecibo/ALFA data, but are
much less sensitive (see Sect. 7.1.2, for example). These
reference VLA data sets predate those of THINGS9.
Observations were carried out with the 305 m telescope
of the Arecibo Observatory10using the 7 beam ALFA
9The THINGS data set achieves a sensitivity of NHI = 4×1019
cm−2at a resolution of 30 arcsec, using the same criteria as we will
set out in Sect. 6.
10The Arecibo Observatory is part of the National Astronomy
and Ionosphere Center (NAIC), a national research center oper-
ated by Cornell University under a cooperative agreement with
the National Science Foundation (NSF).
Parameters of NGC 2903a
R.A. (h m s)
Decl. (◦ ′ ′′)
V⊙ (km s−1)b
2a × 2b (′×
(kpc × kpc)
09 32 10.11
21 30 03.0
12.6 × 6.0
32.6 × 15.5
aData from NED unless otherwise indicated.
cDrozdovsky & Karachentsev (2000).
dOptical major × minor axis diameters. The semi-major
diameter, a, is equivalent to R25, the radius in the
B-band at the 25.0magarcsec−2isophote level.
eInfra-red and far infra-red luminosity (Sanders et al. 2003),
adjusted to our distance.
fStar formation rate from LFIRand the formalism of
Kennicutt (1998) .
Companion Galaxies of NGC 2903a
R.A. (h m s)
Decl. (◦ ′ ′′)
2a × 2b (′×′)g
9 32 48.9
21 27 55
510 ± 30f
0.9 × 0.9
9 19 30.0
21 36 12
498 ± 2
0.7 × 0.6
9 30 12.8
19 59 26
562 ± 1
0.7 × 0.6
aData from NED unless otherwise indicated.
bAlternate names: D565-05 in the Low Surface Brightness
Galaxy Catalogue (LSBC); J093248.81+212756.2 in the SDSS.
cIdentifier in the LSBC.
dProjected separation from the center of NGC 2903.
eHeliocentric radial velocity.
fOptical velocity, cz, as given in the SDSS Data Release 6 (DR6).
gMajor axis × minor axis diameters.
hOptical B magnitude.
iReference for association with NGC 2903.
DK00 = Drozdovsky & Karachentsev (2000)
receiver system (see Fig. 2 of Giovanelli et al. 2005a, for
the ALFA beam geometry) with the Wideband Arecibo
Pulsar Processor (WAPP) back-end spectrometer sys-
tem. The total observing time allocated for this project
was 97 hours, divided into 37 observing blocks (Novem-
ber 28-30, December 1-6, 14-23, 26 2004; February 10-
13, 28, March 1-6, 21-26 2005) carried out during the
commissioning phase of ALFA. The observing setup is
summarized in Table 3.
Because the ALFA beams can have coma lobes as
high as 20% (7 dB), high-sensitivity observations of ex-
tended objects with this instrument must account for
contributions from stray/unwanted radiation into these
lobes; that is, we obtain a “dirty map” which must
be “cleaned”. Our basic approach is therefore to map
the field in Fig. 1 as well as the 7 ALFA beams in
a fixed number of telescope configurations. The beam
maps are used to deconvolve the sidelobe contribution to
the galaxy map in each configuration after which these
clean maps are combined to form our final datasets. In
NGC 2903 and its Environment5
Sects. 4.1 and 4.2, we describe our observing strategy for
mapping the galaxy and the beams, respectively.
4.1. Observations of NGC 2903
The mapping of NGC 2903 was conducted in a
“Fixed Azimuth Drift” mode
in the Arecibo Galaxy Environment Survey (AGES,
Auld et al. 2006)11. It is, however, somewhat less ef-
ficient than the dual-pass strategy employed by the
Arecibo Legacy Fast ALFA Survey (ALFALFA, Gio-
vanelli et al.2005a)12, mainly because we observed
NGC 2903 at 12 separate azimuths in order to increase
the effective integration time per point to 6 times the
For each of 12 azimuths (AZ = 104◦, 107◦, 109◦, 116◦,
128◦, 153◦, 200◦, 230◦, 245◦, 250◦, 253◦, 255◦), ALFA
was positioned to point 4min R.A. ahead of the source,
then held motionless while the source drifted through.
Spectra for each linear polarization of each of the 7 feeds
were recorded at a rate of once per second during the drift
scan. Thus 12 drift scans centered at a single declina-
tion could optimally be obtained in one observing session.
The azimuths were determined by calculating the min-
imum time required for slewing and resetting/adjusting
system parameters in preparation for the next drift at
the next azimuth, thus minimizing overheads.
Before each drift scan at each azimuth, the ALFA tur-
ret was rotated so as to produce equal separations in
declination between successive beams (but see below).
The azimuth and zenith angle were also both adjusted
so that a drift would cross the same declination (J2000)
any time it was repeated.
On a given night, drifts at all 12 azimuths kept the
center beam of ALFA at the same declination. On fol-
lowing nights, the declination was shifted by 4′45′′so that
the center beam drifted through the interstice between
the northermost beam of the preceding night’s drift and
the southernmost of the following. This helped to pro-
duce a map with as uniform a sensitivity as possible and
Nyquist-sampled the survey area in declination.
In the allocated time we were able to obtain full sets
of 12 drifts each, across 13 nearly constant declina-
tion tracks with the center beam spanning 21◦03′56′′
to 22◦00′56′′(J2000), and with another full set of 12
drifts spanning 21◦13′26′′to 21◦27′41′′and 21◦37′11′′to
21◦51′26′′. (The number of drifts at a given declination,
referred to as Np, is given in Fig. 5d) Compared to these
repeated drift ranges, our final map is undersensitive in
the central declination strip 21◦32′26′′, the two southern-
most strips and the two northernmost strips (see Sect. 6.1
The Fixed Azimuth Drift mode introduces some vari-
ations in declination.First of all, since the telescope
was held at fixed azimuth and zenith angle through each
drift, the declination tracked by each beam changed by
a small amount from the start to the end of the drift.
We have computed the average variation in beam posi-
tion from this effect, and find it to be less than 1 arc-
sec and therefore negligible. Secondly, the declination
spacing between ALFA beams at a given azimuth was
not exactly uniform. The variation in beam spacing is
similar to that adopted
typically ≈ 2% of the beam size, a value comparable
to the ≈5′′pointing accuracy of the Arecibo telescope.
Finally, the declination spacings vary with azimuth due
to the elliptical illumination pattern of the Arecibo tele-
scope. These declination separations varied monotoni-
cally from ∼ 123′′AZ = 180◦to ∼ 108′′at AZ = 104◦
and AZ = 255◦. We account for the latter two effects in
the data reduction (see Sect. 5.1).
Center frequency, νc (GHz)
Velocity Coverage (km s−1)a
Channel width (km s−1)
Drift Scan Duration (s)
Center frequency (GHz)
Drift Scan Duration (s)
−585.5 → 2064.2
aAll velocities in this paper are heliocentric.
4.2. Beam Mapping
The outer beams of ALFA have significant coma lobes,
and the contributions of stray radiation and outer side-
lobes to all the beams is not negligible. While precise
beam maps were produced by Cort´ es (2003), they do
not account for variations with azimuth and zenith angle
or blockage by the platform and cables.
Consequently it was necessary to map the beams by
observing strong unresolved continuum sources using a
mapping strategy similar to that described in Sect. 4.1.
In the case of the beams, however, we chose twice as
many azimuth settings (24 instead of 12) and shorter
drift scans (Table 3).
The declination of Beam 0, which is the center beam
of ALFA, was shifted by 1′53′′(about half a beamwidth)
from one night to the next, and we completed 13 sepa-
rate drifts on each source. No single source of sufficient
strength could be mapped at each of the 24 azimuths in
the time we were allotted on any one night, so we mapped
two of them: J080538+210651(0.9 Jy) at AZ ? 120◦and
J102155+215931 (1.7 Jy) at AZ ? 120◦.
This procedure gave us beam maps approximately 56′
long in R.A. and spanning 23′in decl. Only Beam 0 is
mapped to equal distances north and south of the center,
however. While the span in R.A. is sufficient to map the
first several sidelobes of each of the outer beams, the span
in declination falls short of reaching the second sidelobe
for the northernmost and southernmost beams. This will
be discussed further in Sect. 5.2.
5. DATA REDUCTION AND PROCESSING
The drift scans were bandpass-subtracted and baseline-
flattened using Interactive Data Language (IDL) proce-
dures written by P. Perillat for general use at Arecibo
Observatory, and by R. Giovanelli and B. R. Kent for
the ALFALFA precursor data (Giovanelli et al. 2005b).
Preliminary calibration was accomplished using the sys-
tem’s equivalent flux density for each beam as a function
Irwin et al.
Fig. 2.— Maps of each ALFA beam (rows) used to deconvolve sidelobe and stray radiation contributions from NGC 2903 observations at each azimuth (columns). Each panel spans 24′
in both R.A. and decl., and contours are at (-18, -12, -9, -3) dB. See fig. 2 of Giovanelli et al. 2005a for an illustration of the beam locations in the ALFA footprint. FITS files containing
the 84 beams may be downloaded from our website.
NGC 2903 and its Environment7
of zenith angle, provided by Arecibo Observatory staff.
Data reduction specific to the galaxy and beams is de-
5.1. Galaxy Data Reduction
Initial maps of NGC 2903 from the drifts showed sig-
nificant striping across the galaxy, indicating that the
calibration of the individual beams was not sufficiently
precise. An attempt to improve the calibration by in-
tegrating over the Galactic H I emission, requiring the
integral to be the same for each beam, proved unsuccess-
ful since the separate beams follow different tracks across
the Galactic emission and, on the scale of the NGC 2903
map, there is significant variation of the Galactic H I
emission between those tracks.
To improve the calibration, we sought to make use of
the continuum sources in the mapped field. Sources with
sizes small compared to the Arecibo beam and with peak
flux densities exceeding 10 mJy were selected from NED.
Gaussians were fitted to corresponding detected contin-
uum signals in our data, the distance from the fitted
peak to the catalogued source position was determined,
and the ratio of the flux expected at that position to that
observed was calculated. Only detections that fell within
half a beamwidth of the source were used. Starting with
the strongest source, then working down the list in order
of source strength, the factors by which each beam must
be multiplied to place their gains on a common scale were
determined. The resulting calibration is referenced to the
flux of the strongest source in the field, J093215+211243,
which we took to have peak flux density 562.3 mJy at the
time of our observations. This approach significantly re-
duced the striping in the final maps, though we note the
presence of residuals that will be discussed in Sect. 6.2.
The drifts for each azimuth separately were then grid-
ded into a datacube with axes, right ascension (R.A.),
declination (Decl.), and velocity (V). This was done us-
ing the ALFALFA IDL gridding tool (Giovanelli et al.
2008 in prep.) modified for the NGC 2903 drift length
and calibration method. At each defined point in the
grid, the gridding tool produced a weighted average of
emission from nearby one-second spectral samples, using
the positions recorded in the data headers for each. No
assumption of constant separation in declination from
one beam to the next was required in this process. The
beam which dominated the weight for each grid point was
also recorded in the data structure, for use later by the
cleaning software (see Sect. 5.3). The grid pixel size was
30′′in R.A. and Decl., and the velocity range was cho-
sen to extend well beyond that of NGC 2903, excluding
There was no significant radio frequency interference
from outside the observatory in our spectra.
ever, there was an internally generated “wandering
birdy” (Giovanelli et al. 2005b) present in some of our
drifts. This was an interference spike that drifted non-
monotonically in frequency during particular drifts. The
source has since been identified and corrected, and has
not been seen in any observations since early 2005, to
our knowledge. Fortunately the birdy’s wanderings did
not take it close to the NGC 2903 velocity range during
our observations for the most part. In those few spec-
tra where it would have caused difficulty in the cleaning
process, it was excised by interpolating between spectral
channels just outside the spike.
5.2. Beam Data Reduction
To produce a separate 2-dimensional (2D) map of each
beam at each azimuth, we extracted the individual beam
drifts from the sets of drift scans, then constructed maps
of the continuum source as seen by each beam, producing
7 maps for each of the 24 azimuth settings. The contin-
uum maps were written out from IDL into Flexible Im-
age Transport System (FITS) format, then read into the
Astronomical Information Processing System software,
aips++13, of the National Radio Astronomy Observa-
tory (NRAO). Standard 2D gaussian fitting routines in
aips++ were used to fit and subtract out each catalogued
continuum point source outside the much stronger target
source. Noticeable sidelobe signal from these non-target
sources was zeroed as well, as long as there was no confu-
sion with the sidelobes of the target source. Our inability
to completely erase these extraneous sources limits the
dynamic range of the final galaxy maps (see Sect. 6.2).
These maps were then read into the Astronomical Im-
age Processing System (AIPS)14for further processing.
The goal was to create a single map for each of the 7
beams at each of the 12 galaxy azimuths (84 beam maps
in total). To do this, the beam maps were placed on
the same amplitude scale and beam maps at azimuths
adjacent to the galaxy azimuth were averaged together,
weighted by distance from the galaxy azimuth. For ex-
ample, beam maps at AZ = 103◦and AZ = 106◦were
averaged to obtain an estimate of the beam at the galaxy
azimuth AZ = 104◦, with a higher weight attributed to
AZ = 103◦.
The final beam maps at each galaxy azimuth covered
54′in R.A. and 21′in decl. This was more than suffi-
cient to fully map the beam shape in R.A.. However,
as indicated in Sect. 4.2, the decl. span does not reach
the second sidelobe for the northernmost and southern-
most beams. In addition, for 4 of the 7 beams, the first
sidelobe on one side only was cut off in decl.
proximately the midpoint of its peak. For these cases,
to avoid the introduction of artifacts during the decon-
volution process (Sect. 5.3), the beam sidelobe was ex-
tended/smoothed at the edge by a gaussian of width,
All 84 beams are displayed in Fig. 2 and nicely show
the changing beam structure with changing azimuth15.
These beams were then used for the IDL-based clean
described in the next section.
5.3. Image Deconvolution and Final Cubes
To achieve high sensitivity to low-level emission from
the outer edges of the galaxy it is necessary to remove the
sidelobe and stray radiation contribution to the maps of
NGC 2903. We perform this deconvolution with a ‘clean’
algorithm analogous to that used in aperture synthesis
imaging (see Cornwell et al. 1999, for a review).
We use an image-plane IDL-based implementation of
the clean algorithm written by Buie (2008), modified by
15FITS files for these beams are available on the NGC 2903
8 Irwin et al.
Fig. 3.— R.A. – V plots at Decl. = 21◦30′03′′, for the Original cube (a), the V-smoothed cube (b), the RD-smoothed cube (c), and
the RDV-smoothed cube (d) (see Table 4). V is indicated at the bottom of the lowest frame. In each case, the greyscale range, marked at
the top of each frame, goes from -1σ to 10% of the maximum in the frame, the latter being 253.0 mJy beam−1for (a), 252.1 mJy beam−1
for (b), 285.9 mJy beam−1for (c) and 284.5 mJy beam−1for (d). Contours at 50% and 90% of the peak are also shown.
us to account for the multiple beams16. Since the dom-
inant contributing beam was recorded in the grid data
structure for each point, we were able to calculate the
contribution to that point from a point source anywhere
in the R.A. - decl. map at each velocity. Each iteration
consisted of identifying the strongest remaining emission
in the map, then using the appropriate known beams
to remove the contributions of that point to the entire
map. Iterations continued until the first negative clean
component was reached, or until the clean component
reached the level of the noise which we took to be 2 mJy.
This clean procedure was carried out at each azimuth,
producing 12 cleaned NGC 2903 datacubes.
The cleaned cubes were then read into AIPS for fur-
ther reduction and analysis. Each of these cubes was
inspected individually and some minor editing (e.g. of re-
maining wandering birdie spikes farther from the galaxy
emission) was carried out. All cubes at different azimuths
were then averaged to form a single cube. A subset of
velocity-space in the cube was then extracted so as to
avoid noisy end channels as well as contaminating Galac-
tic emission on the low velocity side. The resulting full-
resolution cube will be designated as the ‘Original’ cube.
This cube was then smoothed, in velocity alone (denoted
16See mbmclean.pro on our website.
V-smoothed), spatially alone (RD-smoothed) and both
spectrally and spatially (RDV-smoothed). The spatial
smoothing, in particular, ameliorates the striping issue
discussed in Sect. 5.1, improving the the rms noise in
the maps (see next section). Finally, residual curvature
in the baseline was removed, point by point17. The pa-
rameters of the final cubes are given in Table 4, and the
cubes themselves may be obtained from our website.
6. DETECTION LIMITS AND DATA QUALITY
A selection of R.A. - V plots and R.A. - decl. plots of
the final cubes are shown in Figs. 3 and 4, respectively.
The greyscale in the plots emphasizes low-intensity emis-
sion to illustrate the data sensitivity in low dynamic
range regions of the cube as well as residual map errors
near NGC 2903. We discuss these map properties in turn
6.1. Data Sensitivity
The mean,¯S, and root-mean-square (RMS) noise, σ,
of all regions beyond the extended envelope of NGC 2903
(see Sect. 7.1) in each cube is listed in Table 4. As ex-
pected in these low dynamic range regions, the final base-
line is consistent with zero (¯S << σ) and a histogram of
σ over all line-free channels is gaussian.
17The AIPS task, xbasl, was used.
NGC 2903 and its Environment9
Parameters of Cubes
Final velocity coverage (km s−1)a
Channel width (km s−1)
Velocity resolution, δ V (km s−1)b
Spatial resolution, θ (arcsec)c
σ (mJy beam−1)d
mean (mJy beam−1)e
100.0 → 1132.8
100.0 → 1132.8
100.0 → 1132.8
100.0 → 1132.8
aBandwidth range after removing end channels and Galactic emis-
bSince Hanning smoothing was applied, the original velocity res-
olution is not equivalent to the channel width.
cFull width at half maximum of the gaussian beam.
dRms noise over all regions of the cubes in which the galaxy emis-
sion had been blanked.
eMean over all regions of the cubes in which the galaxy emission
had been blanked.
fH I mass limit (3σ 2δ V , as described in Sect. 6.1).
gH I column density limit (3σ 2δ V , as described in Sect. 6.1).
hMaximum signal-to-noise (S/N) ratio of each cube. The S/N
cubes are described in Sect. 6.1.
The variation in σ as a function of R.A., decl. and V
was examined. While we find that σ is independent of
R.A. and V, it does vary with declination, a result that
is illustrated in Fig. 5a. This is primarily caused by the
higher gain of the central beam, Beam 0 (∼ 11K/Jy)
relative to the outer ones (∼ 8.5 K/Jy). The result is
that declinations surveyed with the former have lower
σ. The correspondence between the decl. of Beam 0 for
each drift (location of points along x axis of Fig. 5d) and
the minima in Fig. 5a illustrates this effect. Different
numbers of drift scans at some declinations (Fig. 5d),
uncertainties in beam calibration and variations in beam
spacing (see Sect. 4.1) also contribute to changes in σ
From the noise values and some assumptions, we can
compute detection limits for each cube in the low dy-
namic range regime. We consider the limiting flux inte-
gral, Slim, to be from an unresolved signal that is at a
3σ level in 2 independent channels,
Slim = 3σ2δ V Jykms−1
where δ V (km s−1) is the velocity resolution (see Ta-
ble 4) and σ is in Jy beam−1. The minimum detectable
H I mass is then,
MHI lim = 2.356 × 105D2Slim
where D is the distance (Mpc), and the minimum de-
tectable column density for a signal that uniformly fills
the beam is:
NHI lim =2.228 × 1024
where θ is the spatial resolution (arcsec, Table 4) and
νcis the central frequency (GHz, Table 3). If the signal
does not uniformly fill the beam, then the right hand side
of Eqn. 3 must be divided by an areal filling factor.
Plots of MHI lim and NHI lim as a function of decl.
are shown in Figs. 5b and 5c, respectively, and the mean
values for each cube are given in Table 4. Note that
the cubes smoothed in velocity have higher MHI limand
NHI limthan their full resolution counterparts because of
the larger δ V of the former. Note also that these results
are simply detection limits for the data, without impos-
ing assumptions about the properties of any companions
that might be present. The limits shown in Table 4 are
very low; for example, the column density limits are lower
than those of THINGS by two orders of magnitude (see
6.2. Residual Map Errors near NGC 2903
Close inspection of Figs. 3 and 4 reveal residual map
errors near NGC 2903 that remain even after the data
reduction procedure discussed in Sect. 5. These artifacts,
described below, limit the dynamic range of the data
in regions occupied by emission from NGC 2903 itself,
having a greater relative effect near the ‘edges’ of this
One artifact that is evident in Figs. 3a and 3b is a
faint ridge of emission seen on the low-RA side of the
galaxy emission, running close to and parallel with the
‘edge’ of the main galaxy emission. This ridge is due to
imperfectly cleaned sidelobes (see Sect. 5.2). It is most
evident in the data at full spatial resolution, and occurs
at typically a 2 to 5% level in comparison to the brightest
galactic emission at the same velocity.
Another type of artifact is evident in Figs. 4a and 4b,
and can be attributed to residual striping due to scan
calibration uncertainties (Sect. 5.1). These residual er-
rors vary in strength but are typically at the level of a
few percent of the peak in any given channel near the
edge of the main emission. They produce the ‘scallop-
ing’ of the edges of the H I distribution in NGC 2903 at
low column densities (e.g. Fig. 10).
Finally, the second sidelobes (outer coma lobes) of the
beams were not fully mapped in declination, thus lim-
iting the effectiveness of the cleaning in this dimension
(see Sect. 5.2). From an examination of the rather com-
plex outer lobes observed in R.A., we estimate that sec-
ond sidelobe peaks at approximately 0.6% of the central
beam peak could contribute to emission as far as 10′away
10 Irwin et al.
Fig.4.— R.A. – Decl. plots at V = 555.0 km s−1, for the
Original cube (a), the V-smoothed cube (b), the RD-smoothed
cube (c), and the RDV-smoothed cube (d) (see Table 4). RA is
indicated at the bottom of the lowest frame. In each case, the
greyscale range, marked at the top of each frame, goes from -1σ
to 10% of the maximum in the frame, the latter being 216.6 mJy
beam−1for (a), 218.0 mJy beam−1for (b), 247.1 mJy beam−1for
(c) and 247.5 mJy beam−1for (d). Contours at 50% and 90% of
the peak are also shown, as are the beam sizes in each frame.
in decl. (note, however, that this varies beam to beam
and azimuth to azimuth).
Considering these residual map errors and typical peak
H I flux densities measured for NGC 2903 in our data,
we caution against simple interpretations of emission be-
low a few percent in any given channel over the emission
region occupied by NGC 2903 itself. We estimate that
a more conservative detection criterion of ∼ 15σ2δV =
5Slim(≈ 1018cm−2in column density) is likely appro-
priate in these high-dynamic range regions; that is, the
detection limits of Table 4 should be increased by a fac-
tor of 5 for regions occupied by NGC 2903 emission it-
self. We defer the detailed analysis required to determine
the outer H I morphology of NGC 2903 to a future pa-
per. We emphasize that these artifacts arise only near
bright emission, and thus the sensitivity limits given in
Table 4 using Slimin Eqn. 1 are appropriate for regions
of the data cubes in which we search for companions (see
7.1. Basic Properties of NGC 2903
While a detailed analysis of the H I morphology and
kinematics of NGC 2903 is beyond the scope of this pa-
per, we present some of its basic properties here to illus-
trate the content and quality of our datacubes.
7.1.1. Global Parameters
Fig. 6 shows the global profile of NGC 2903, and
corresponding global parameters are given in Table 5.
The profile shape agrees well with previously published
plots (Wong et al. 2006; Hewitt et al. 1983) and our in-
tegrated flux density (Table 5) agrees with the result of
Braun et al. (2007) to within errors. There is an obvi-
ous asymmetry in the galaxy, such that the low velocity
peak (north-east side of galaxy) is higher than the high
velocity peak (south-west side). The integrated flux on
the low velocity side of the galaxy is 13% higher than on
the high velocity side, denoting an intrinsic asymmetry
in the H I distribution of that order.
7.1.2. Morphology and Kinematics
We present the integrated intensity and intensity-
weighted mean velocity fields, from the RDV-smoothed
cubes,in Fig. 7.A small,
H I companion can be seen 24.8′(64.3 kpc in projection)
to the north-west. The eastern companion, UGC 5086,
is enveloped in the H
I emission from NGC 2903.
H I companions will be discussed further in Sect. 7.2.
We find a very large H I envelope around NGC 2903,
even accounting for the Arecibo beam and residual map
errors. For comparison, the grey contour in Fig. 7a
shows the outermost significant integrated intensity
level (1×1019cm−2) in the archived D-configuration
VLA observations (see Sect. 3), smoothed to the same
resolution as the Arecibo data in the figure. The major
axis diameter at 1018cm−2in our Arecibo data – which
should be immune to residual map errors (see Sect. 6.2)
– is dHI= 40.7′(105 kpc) after correcting for the beam,
nearly twice the value measured from the VLA data.
Thus, the H I extent of NGC 2903 is at least 3.2 times its
optical diameter (Table 1) and ranks among the largest
known (Matthews et al. 2001;
Spekkens & Giovanelli 2006;
Curran et al. 2008).
The velocity field of NGC 2903 (Fig. 7b) shows reg-
ular rotation, with the north-east side advancing with
respect to the center. The contours indicate that the
outer H I disk of the galaxy is warped in spite of its
apparent isolation. A position-velocity plot along a 270′′
wide strip of the major axis is shown in Fig. 8. The inner
rotation curve of NGC 2903 appears to be regular, but
it is strongly biased by beam smearing and not a good
indicator of the gravitational potential in these regions.
We find no evidence for gas at anomalous velocities at
the sensitivity and resolution of our data.
del Rio et al. 2004;
Oosterloo et al. 2007;
NGC 2903 and its Environment11
Fig. 5.— Noise and detection thresholds as a function of Dec. ∆ DEC is relative to the optical center of NGC 2903 (Table 1). In
each panel, the solid curve represents the original cube, the dotted curve represents the V-smoothed cube, the dashed curve represents the
RD-smoothed cube, and the dash-dotted curve represents the RDV-smoothed cube (Sect. 5.3, Table 4). (a) RMS noise σ. (b) Minimum
detectable HI mass MHI lim, given by Eqn. 2. (c) Minimum detectable column density NHI lim, given by Eqn. 3. (d) Number of passes
(i.e. drifts) Np as a function of the declination of the central ALFA beam (Beam 0).
H I Properties of NGC 2903
∆V50 (km s−1)a
∆V20 (km s−1)b
Vsys (km s−1)c
SVdV (Jy km s−1)d
370 ± 2
383 ± 2
555 ± 2
255 ± 15
(4.8 ± 0.3)×109
aFull width at 50% of the average of the 2 profile peaks.
bAs in a but at 20%.
cV at the midpoint of ∆V20.
dIntegral of flux density associated with NGC 2903. The
dominant uncertainty is from the baseline flattening;
excluding this effect, the uncertainties are of order 1%.
eH I mass, from Eqn. 2 substitutingRSVdV for Slim.
fratio of H I to optical major axis diameter. dHIis
measured at 1018cm−2and dopt is from Table 2.
7.2. H I Companions of NGC 2903
7.2.1. The Known Companion, UGC 5086
UGC 5086 is the only previously known companion
of NGC 2903 that lies within the surveyed region (see
Sect. 3). This galaxy overlaps the large H I envelope of
NGC 2903 in both position and velocity space. Fig. 9,
which shows a single channel at the systemic velocity
of UGC 5086, illustrates this blending. A global profile
from a 3′× 3′box centred on UGC 5086 in the RDV-
smoothed cubes yields a spectrum (not shown) with a
peak of 4.4 mJy at V= 497km s−1, resulting in a mea-
sured H I mass of MHI = 9.0 × 106M⊙(Eqn.2 substi-
tuting the integrated flux of the profile for Slim). While
the velocity of the spectral peak agrees with the sys-
temic velocity of UGC 5086 (Table 2), the detected sig-
nal could also arise from the envelope of NGC 2903 it-
self. To help distinguish between signal from UGC 5086
and NGC 2903, we have searched our reduced VLA
D-configuration (Sect. 3) data, which clearly separate
the two galaxies spatially.
UGC 5086, and place an upper limit on the H I mass
in a single 54′′beam of MHI lim = 5.6 × 105M⊙ from
the VLA data. Thus, our Arecibo detection is primarily
from NGC 2903 iself.
UGC 5086 is well resolved in the SDSS Data Release 6
(DR6). Its image appears almost perfectly circular and it
has a red colour (B0- V0= 0.79, using SDSS magnitudes
and applying transformations referenced in Sect. 7.2.3).
Given the absence of H I and known optical parameters,
it is likely that UGC 5086 is a dwarf spheroidal galaxy
without H I content. We note that near-UV emission can
be seen in UGC 5086 also18, suggesting the presence of
a young stellar population. This is similar to the dwarf
We find no emission from
12 Irwin et al.
Fig.6.— Global profile of NGC 2903.
uncertainty in baseline flattening) are typically within 1% except
where the profile approaches zero.
Errors (ignoring any
spheroidal galaxy, Fornax, which contains both old, in-
termediate, and young stars (Battaglia et al. 2006).
7.2.2. Search for New Companions
A visual search of the datacubes has revealed a new
H I -rich companion to NGC 2903, visible in Fig. 7. This
companion, which we designate N2903-HI-1, will be dis-
cussed further in Sect. 7.2.3. In order to detect compan-
ions in a more quantitative fashion, it is important to
account for the variation in map noise, σ, with decl., il-
lustrated in Fig. 5a. To this end, we formed S/N cubes19
by dividing each datapoint by the σ corresponding to
its decl., and searched for emission exceeding Slimfrom
Eqn. 1. Specifically, we integrated each S/N cube over
all V including only those points that exceed 3σ over at
least two adjacent independent velocity resolution ele-
ments20. The resulting summed maps, which emphasize
faint emission, are shown in Fig. 10.
Fig. 10 confirms that N2903-HI-1 is the only bona fide
H I -rich companion to NGC 2903 in our data. While
other isolated non-zero pixels are also seen in the var-
ious maps, they are clearly random noise peaks. This
is corroborated by the lack of correlation between the
locations of these pixels in the different maps.
There is a possibility that companions may lie within
the spatial region over which NGC 2903 is found, but at
anomalous velocities in comparison to NGC 2903. We
have searched through this parameter space and find no
evidence for such H I clouds (see also the major axis slice
of Fig. 8).
It is also plausible that our search has missed com-
panions which overlap NGC 2903 in both position and
velocity space, a possibility raised by the location of
UGC 5086 (Sect. 7.2.1). The H I disk of NGC 2903 occu-
pies an exceedingly small percentage of the total volume
surveyed, however, making such a coincidence highly un-
likely. Moreover, if a starless H I emission feature exists
within the H I position-velocity envelope of NGC 2903,
then it becomes moot as to whether such a feature should
simply be considered part of NGC 2903 itself.
7.2.3. The New Companion, N2903-HI-1
19The S/N cubes may be download from our website.
20Our AIPS-compatible routine (xsmc), written for this pur-
pose, is available on our website.
N2903-HI-1 is separated from NGC 2903 by 24.8′spa-
tially (64.3 kpc in projection) and by +26 km s−1in
velocity (cf. Tables 1 & 6). Its global profile is shown
in Fig. 11, and related parameters are given in Table 6.
We also show the total intensity map, a position-velocity
slice, and the 1st and 2nd moments of the H I distribution
in Fig. 12. The full velocity resolution cube has been used
for the latter two maps since the profile is narrow.
Fig. 12a illustrates that N2903-HI-1 is elongated north-
east to south-west and has a ‘cometary’ or ‘head-tail’
morphology. In spite of the large beam size, the com-
panion is spatially resolved in all cubes, hence its radius,
R, can be measured. This has been done via a gaussian
fit to the highest spatial resolution data and deconvolv-
ing the beam (Table 6).
We do not find convincing evidence of systematic
motion in our data, either from the major axis slice
(Fig. 12b), the 1st moment map (Fig. 12c), or channel
maps (not shown). A rotating disk with an inclination
i > 15◦would exhibit a gradient across the disk that is
greater than the typical velocity dispersion of 6 km s−1in
Fig. 12d. Thus, if the H I represents a disk in rotation, we
might have expected, given our fine velocity resolution,
to have seen some evidence for this. Given the elongated
morphology of the H I, it seems more likely that it is be-
ing either tidally perturbed or ram pressure stripped via
passage through a gaseous medium. Head-tail morpholo-
gies are typically seen in the latter case, but because of
low spatial resolution, the former cannot be ruled out.
To determine a total dynamical mass, ideally we would
want to associate the radius of N2903-HI-1 with some ro-
tational velocity. As we do not know the precise geome-
try of the system and do not see rotation, this association
cannot be made. However, the measured line width and
radius should provide us with some measure of the to-
tal mass which we now estimate under two assumptions
that should encompass the extrema of possibilities (see
Westmeier et al. 2005a for a similar approach).
First, in the event that a rotation might have remained
undetected (for example, if the companion represents
a rotating galaxy that is near face-on), the dynamical
mass, Mdyn, can be estimated via,
where ∆V50is the width of the profile in Fig. 11 at 50%
of the peak (Table 6). ∆V50 has been adopted, rather
than ∆V20, since the latter is a measurement at a level
within the noise. With the assumption of rotation, R
may be overestimated if there is a cometary tail that
does not take part in the rotation; however, the velocity
width, which has been minimized by the choice of ∆V50,
has a greater effect.Together with what would be a
substantial correction for the unknown inclination, the
result for Mdyn(Table 6) will be a minimum.
Secondly, for comparative purposes, we assume virial
equilibrium for which,
Mvir =5R (∆V50)2
Here, we have related the mean square velocity of the
particles, ?v2?, to the observed FWHM velocity width,
∆V , according to ?v2? = 3(∆V )2/(8 ln2) (Westmeier
et al. 2005b).
NGC 2903 and its Environment13
Fig. 7.— Moment maps of NGC 2903, constructed from the RDV-smoothed cubes. The location of the galaxy, UGC 5086 is marked
with a star. Detailed maps of the companion to the north-west of NGC 2903 are in Fig. 12. (a) Total intensity HI map over the DSS2 Blue
image, the latter shown in an arbitrary greyscale. Contours are at 0.02, 0.06, 0.10, 0.20, 0.50, 1.0, 2.5, 5, 10, and 25 Jy beam−1km s−1. The
peak is 68.3 Jy beam−1km s−1. The beam is shown at lower left. A conversion to column density requires a multiplication by 1.52×1019
cm−2(Jy beam−1km s−1)−1. Note that there may be residual map errors near NGC 2903 below ∼ 1018cm−2; see Sect. 6.2. The grey
curve shows the outermost significant integrated intensity level (1×1019cm−2) in the archived D-configuration VLA observations (see
Sect. 3), after smoothing the VLA cube to the same spatial resolution as the Arecibo data. (b) Intensity-weighted mean velocity contours
over a greyscale from the same image. Contours, in km s−1, are labelled and are spaced 20 km s−1apart. The optical center of the galaxy
is marked with a cross.
Fig.8.— Position-velocity plot along the major axis of
NGC 2903, from the RDV-smoothed cube, averaged over a width
equivalent to the beam size. The center is marked with a cross.
Contours are at 1.0 (2σ), 1.8, 3.0, 6.0, 15, 30, 60, 120, 200, and 300
mJy beam−1. The north-east (NE) and south-west (SW) sides of
the galaxy are labelled and tickmarks along the position axis are
separated by 490′′. Note that there may be residual map errors
below ∼ 3% near the galaxy; see Sect. 6.2.
The results (Table 6) indicate a total mass for N2903-
HI-1 which exceeds 108M⊙by either estimate (Table 6),
and we adopt the mean, 3 × 108M⊙, as a ‘characteristic’
dynamical mass. Although the error bars are substantial,
it is nevertheless clear that the H I mass of N2903-HI-1
is only a small fraction of its total mass.
RIGHT ASCENSION (J2000)
09 33 30 15 00 32 45 30 15
Fig. 9.— A single channel of the V-smoothed cube centered on
the position and velocity (Table 2) of the companion, UGC 5086.
The companion is seen within the region of a protrusion to the
east of the HI envelope associated with NGC 2903. UGC 5086 is
marked with a star and the center of NGC 2903 is marked with a
cross to the west. Contours are at 1.6 (2σ), 3.0, 5.0, 10, 20, 50,
100, and 200 mJy beam−1. Note that there may be residual map
errors below ∼ 3%; see Sect. 6.2. The beam is shown at lower left.
14 Irwin et al.
Fig. 10.— Summed S/N maps for each of the four cubes (labelled at top left in each frame). Only points ≥ 3σ that are detected
in ≥ 2δ V independent velocity resolution elements, as described in Sect. 7.2.2, have been included in the maps. Low levels have been
emphasized, with a greyscale range (3σ − 5σ) shown at the top, and the 5σ and 50σ contours are also shown. The beam is shown at lower
left. Peak levels are 161σ, 216σ, 256σ, and 343σ for (a), (b), (c), and (d), respectively. Note that there may be residual map errors below
the 5σ contour in NGC 2903; see Sect. 6.2.
Properties of N2903-HI-1
R.A. (h m s)b
Decl. (◦ ′ ′′)b
∆V50 (km s−1)c
∆V20 (km s−1)d
Vsys (km s−1)e
SVdV (Jy km s−1)f
09 30 38
21 43 08
23.6 ± 5.2
40.2 ± 5.2
582 ± 4
0.14 ± 0.02
(2.6 ± 0.3) × 106
(1.3±1.1) × 108
(4.5±4.0) × 108
aTable 5 provides definitions when not indicated here.
bPosition of peak of the total intensity map (Fig. 12a).
The uncertainty is ∼ 15′′(1/2 the cellsize) in RA and
∼ 1′(1/4 of the beam) in decl.
cFull width at 50% of the profile peak.
The error bar encompasses variations between the cubes.
dAs in b but at 20%.
eAverage of V at the midpoint of ∆V20 and ∆V50 .
fUncertainty is dominated by variations between
the different cubes and choice of velocity window.
gAssuming the distance is the same as NGC 2903.
hRadius of N2903-HI-1, from 1/2 of the FWHM found from
deconvolved gaussian fits to the total intensity map.
iDynamical (Mdyn), and virial (Mvir) masses, as given by
Eqns. 4 and 5, respectively.
Does N2903-HI-1 have an optical counterpart?
cause of the elongated shape of the H I emission, it
is possible that an optical galaxy could be displaced
with respect to the H I central peak. We have there-
fore searched the SDSS DR6 database over the original
full beam width, (3.9′), centered on the peak position of
N2903-HI-1 (Table 6). There are 207 catalogued sources
in this area, each with photometric, rather than spec-
troscopic redshifts. Given that the dispersion between
spectroscopic and photometric redshifts is of order ≈0.06
(Csabai et al. 2003) at low redshift, we have identified
all galaxies with redshifts less than 0.06 over this spa-
tial region. The result gives 34 galaxies, all of which are
plotted in Fig. 12a. Stellar masses have been calculated
for each of these galaxies according to the formalism of
Bell et al. (2005), assuming a Kroupa (2001) Initial Mass
Function (IMF), and assuming that they are at the dis-
tance of NGC 2903. If a modified Salpeter IMF is used
instead (Bell et al. 2003), the stellar mass increases by
only a factor of 1.4.
The galaxy, SDSS J093039.96+214324.7 (see Table 7),
which is marked in red in Fig. 12a, is the most likely op-
tical counterpart for several reasons. First, the position
of this galaxy agrees with the peak of N2903-HI-1 within
errors (Tables 6, 7). It is also significantly brighter (by
at least 2.9mag in g), larger and has more stellar mass21
than other candidates in the field. A 2nd Digital Sky
Survey (DSS2) image of this galaxy is shown in Fig. 13,
and reveals a galaxy that is elongated roughly north-
south, similar to the H I morphology of N2903-HI-1 (see
Fig. 12) but at a different position angle. Assuming that
J093039.96+214324.7is at the distance of NGC 2903, we
list its properties Table 7.
At the distance of NGC 2903, the properties of
J093039.96+214324.7 indicate that it is a low luminosity
dwarf galaxy with a linear diameter of approximately 1
kpc. If this galaxy is the optical counterpart of N2903-
HI-1, then its stellar mass rivals its H I mass but is at
least two orders of magnitude less than its total mass,
leading to the conclusion that the system is dark mat-
ter dominated. (If J093039.96+214324.7 is not the opti-
cal counterpart, the same conclusion is reached.) The
H I envelope of N2903-HI-1 extends to ∼ 8 optical
21The stellar mass of the second most massive galaxy is only
37% lower, but it is separated from the peak of N2903-HI-1 by 3.9
NGC 2903 and its Environment 15
Properties of J093039.96+214324.7
(pc × pc)a
cz (km s−1)b
u, g, r (mag)c
B0, V0 (mag)d
MB, MV (mag)e
aMajor × minor axis dimensions from the 2σ contour of
bPhotometric redshift from the SDSS DR6.
cMagnitudes in SDSS u, g and r bands.
dB and V magnitudes from the transformation
of Lupton (2005), corrected for Galactic extinction
(Schlegel et al. 1998) .
eAbsolute B and V magnitudes.
fStellar mass, derived from B0 − V0 and MV assuming
a Kroupa IMF (Bell et al. 2005).
gRatio of H I mass (Table 6) to blue luminosity, the
latter from MB, with a Solar B-band magnitude of 5.48.
22 × 12
950 × 518
18.71, 18.14, 18.00
1.8 × 106
radii. Comparing J093039.96+214324.7 to known Local
Group dwarfs (Mateo 1998) or faint irregular galaxies
(Begum et al. 2008), we find that its colour, (B0- V0=
0.31), radius, line width, total mass, and H I mass to
blue light ratio (MHI/LB) fall within observed ranges
for these other known systems. Although its recessional
velocity differs from that of N2903-HI-1, the difference
falls within typical errors for photometric redshifts of
nearby systems (Csabai et al. 2003). Thus, the position
and properties of J093039.96+214324.7, in comparison
to the other galaxies in the region, all suggest that this
galaxy is the likely optical counterpart of N2903-HI-1. A
spectroscopic redshift for this system would decide the
straints(Putman et al. 2002;
Wakker et al. 2007; Wakker et al. 2008) and at the
upper end of the H I mass distribution found for the
HVCs around M 31 and M 33 (Thilker et al. 2004;
Westmeier et al. 2005a; Westmeier et al. 2007).
ever,its separation from NGC 2903 (64 kpc in
projection) is larger than the population of HVCs
around the Milky Way (less than 10 - 15 kpc typically,
Thom et al. 2008) or M 31 (within 50 kpc, Westmeier
et al. 2007). In combination with the evidence for an
optical counterpart, we consider this interpretation for
N2903-HI-1 to be less likely.
Thom et al. 2008;
In our targeted sensitive H
lated Milky-Way analogue, NGC 2903, we have discov-
ered one new companion, the H I object, N2903-HI-
1, which is dark-matter dominated (Table 6) and is
likely associated with the optical dwarf galaxy, SDSS
J093039.96+214324.7. New discoveries of dwarf satellites
of the Milky Way from the SDSS place their typical to-
tal masses in the 106−7M⊙range (Simon & Geha 2007)
which is lower than the total mass of > 108M⊙ (Ta-
ble 6) found for N2903-HI-1. Combining data from pre-
viously known MW companions, the new SDSS compan-
ions, and those of M 31, most satellites which still con-
I survey of the iso-
Fig. 11.— Global profile of the companion, N2903-HI-1, obtained
from the RD-smoothed cubes.
tain detectable H I lie beyond 300 kpc radius, the impli-
cation being that H I has been stripped in closer sys-
tems (Putman et al. 2008; Grcevich & Putman 2008),
although notable exceptions also occur (e.g. the Large
and Small Magellanic Clouds). N2903-HI-1, at a pro-
jected distance of 64 kpc from NGC 2903, would have
to lie at least 293 kpc in front of or behind NGC 2903
to be at a true separation > 300 kpc. Although this is
possible, it is more likely that the companion is closer to
NGC 2903, yet has retained its H I — a result that may
be related to its high dynamical mass in comparison to
most Milky Way systems within the same radius. Thus,
if ram pressure stripping is occurring, as suggested by the
head-tail morphology of N2903-HI-1, the process may be
slower because of its high dynamical mass.
Although the true separation and space velocity of
N2903-HI-1 are not known, we can at least adopt
the projected separation and radial velocity offset from
NGC 2903 to determine whether the above speculation
is feasible. Approximating N2903-HI-1 as a sphere, its
average ISM H I density, from the values of Table 6, is
nISM ≈ 4 × 10−4cm−3. A simple condition for strip-
where Mtotis the total mass of N2903-HI-1 and R is its
radius. Using Mtot = 3 × 108M⊙, Vrel = 26 km s−1
(Sect. 7.2.3) and R from Table 6, and solving for halo
density, we find nhalo= 2 × 10−4cm−3. This value is
within the expected range of halo densities for the Milky
Way at the projected distance of N2903-HI-1 (64 kpc)
(Grcevich & Putman 2008).
larger, stripping will be more effective, and if the true
separation is larger, stripping will be less effective since
the halo density will be lower. Nevertheless, this esti-
mate indicates that N2903-HI-1 could indeed be under-
going ram pressure stripping under the assumption that
the halo of NGC 2903 resembles the Milky Way. The
stripping timescale depends on the difference between the
two sides of Eqn. 6 which is not known. However, since
the magnitudes of the ram pressure and the internal en-
ergy density of the companion are similar, the stripping
timescale may be long, consistent with the observation
If the relative velocity is
16 Irwin et al.
Fig. 12.— HI maps of the companion, N2903-HI-1. (a) Total intensity map, from the RDV-smoothed cubes integrated over 540 to 617
km s−1, superimposed on the DSS2 Blue image. Contours are at 2, 5, 10.0, 20, 30, 50, 75, and 100 Jy beam−1m s−1. The column density
conversion is 1.52×1019cm−2(Jy beam−1km s−1)−1. Crosses mark the locations of all low redshift optical galaxies in the vicinity
(Sect. 7.2.3) and the red cross indicates the most probable companion (Table 7). (b) Position-velocity slice from the RDV-smoothed cubes,
averaged over a 4′wide swath at a position angle of 45◦. Contours are at 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, and 3.9 mJy beam−1. The angular
distance scale is marked at the bottom. (c) Intensity-weighted mean velocity, from the RD-smoothed cubes, using the same velocity range
as in (a). Contours are at 578, 580, 582, and 584 km s−1and the greyscale ranges from 567 to 593 km s−1. (d) Intensity-weighted velocity
dispersion, from the RD-smoothed cubes, with contours at 2, 4, 6, and 8 km s−1and using the same velocity range as in (a). The greyscale
ranges from 0.6 to 10 km s−1. (The black pixels to the north-east included noise and are spurious.)
of detectable H I in the companion22.
It is now interesting to ask how many Local Group
dwarf galaxies would have been detected, if they were
distributed around NGC 2903 similarly to the Milky
Way. The result is dependent on both their distribution
and H I content. Seventeen Local Group dwarf galaxies
listed in Mateo (1998) had sufficient data (distances and
H I data) that we could apply our 3σ2δ V detection cri-
terion (Sect. 6.1) to them. The result is that we could
have detected 7 of them (41%) in terms of sensitivity
limits. However, these 7 galaxies lie at radii between
490 kpc and 1.6 Mpc, in comparison to the effective
projected radius of Reff = 110 kpc (for a circularized
field) of our survey. Since our survey probes a volume
that is only 0.7% of the volume extending to 1.6 Mpc, it
is unlikely that any of these seven galaxies would have
fallen within our field of view.
not list H I data for the Large and Small Magellanic
Clouds (LMC and SMC, respectively) it is clear, how-
ever, that we would have detected these systems. Of the
Although Mateo does
22If ram pressure did strongly dominate, then the stripping
timescale would be minimized, i.e. tmin ≈ 2R/Vrel = 3 × 108
20 new SDSS dwarf Local Group galaxies listed in Simon
& Geha (2007), only one (the dwarf irregular, Leo T)
has H I content (Ryan-Weber et al. 2008), the remain-
der being dwarf spheroidals or falling within parameter
space intermediate between dwarf spheroidals and globu-
lar clusters. Leo T would be marginally detectable in our
survey but, at a distance of 420 kpc (Irwin et al. 2007),
would also likely lie outside our field of view. In spite
of our large survey region, therefore, we are still only
probing the inner region of a possible dwarf galaxy pop-
ulation, were it distributed like our own.
As for HVCs, the HVC population around the Milky
Way is not easily translated to NGC 2903 since HVC
distances (and therefore H I masses) are not well known.
Also, H I column densities which are known (typically
1019cm−2, Stanimirovic et al.
2002), will be diluted by large unknown beam filling fac-
tors at the distance of NGC 2903 (234′′= 10 kpc). We
can say, however, that the Milky Way’s HVC Complex
C, with a mass of 4.9 × 106M⊙at a distance of 10 kpc
from the Sun (Thom et al. 2008), should have been de-
tected, provided it were separated in velocity from the
bulk of the H I in NGC 2903 itself. Given the inclination
2006; Putman et al.
NGC 2903 and its Environment 17
of NGC 2903 and the nature of HVCs, we expect this
criterion to have been met. Thus, NGC 2903 lacks such
an HVC complex. Indeed, given that our search velocity
range was over 1000 km s−1(Table 4) and that the me-
dian velocity FWHM of Milky Way HVCs (36 km s−1,
Putman et al. 2002) corresponds to 28 velocity channels
in our data, it is surprising that no clear HVC detections
have been made. Either NGC 2903 lacks HVCs, possibly
due the fact that NGC 2903 is isolated, or its HVCs are
of very low mass.
No dark starless companions have been detected
around NGC 2903. This result is consistent with the AL-
FALFA survey results which indicate that all extragalac-
tic H I objects can be identified with an optical counter-
part (Saintonge et al. 2008). The discovery of one new
H I rich dwarf companion now places the total number of
companions within our surveyed field (Reff = 110 kpc)
at two: the H I companion, N2903-HI-1, likely associated
with a dwarf galaxy, J093039.96+214324.7, with a dy-
namical mass ∼ 3 × 108M⊙(mean of values in Table 6)
and UGC 5806, which is likely a dwarf spheroidal galaxy
(Sect. 7.2.1). Using the SDSS data for the latter galaxy
and the same transformations as described in Sect. 7.2.3,
the stellar mass of UGC 5806 is M⋆ = 4.7 × 107M⊙.
Gilmore et al. (2007) have shown that mass-to-(V-band)
light ratios, M/LV, for dwarf spheroidal galaxies in the
Local Group range from ≈ 4 to 600, with more luminous
galaxies having systematically lower values of M/LV.
The absolute magnitude of UGC 5086 is MV = -13.7
which is closest to Fornax, amongst Local Group dwarfs.
If UGC 5086 has similar properties, then 3 ≤ M/LV
≤ 20, implying that 1.0 × 108≤ M/M⊙ ≤ 5.2 × 108.
Adopting the mid-point of this range gives an estimate
of ≈ 3 × 108M⊙for the total mass of UGC 5086.
How many dark matter clumps of total mass Mtot ?
3×108M⊙are expected from ΛCDM predictions? Since
NGC 2903 has a total mass that is similar to the Milky
Way (Sect. 3), we can use the Via Lactea simulations of
subhalo clumps from Diemand et al. (2007a) for com-
parison. The fraction of all halo mass that is present
in substructure within a projected radius of 110 kpc
(their Fig.7) is f = 0.02.
halo mass of 1.77×1012M⊙, this yields a total mass
in halo substructure of Mhs = 3.54×1010M⊙ for the
projected radius. Over the substructure mass range
of 4.6×106≤ Msub/M⊙ ≤ 1×1010, the number of
clumps per unit mass range is given by dN/dMsub =
K/(Msub)2, where K is a constant. Since the slope does
not change with radius, we can compute K from Mhs
=?K/(Msub)2MsubdMsub, finding K = 4.6 × 109M⊙.
Finally, we compute the expected number of clumps with
masses greater than Msub from N(> Msub)=K/Msub.
For Msub = 3×108M⊙, this yields 15 clumps in com-
parison to the two observed. If the dynamical mass of
N2903-HI-1 is higher than 3×108M⊙ (for example, if
it is rotating and nearly face-on), then the discrepancy
reduces. However, its mass would have to be as high
as ∼ 5 × 109M⊙ for there to be agreement with the
theoretical expectation. Such a high value would require
N2903-HI-1 to represent a rotating system in a special
geometry with an inclination less than 10◦(Sect. 7.2.3).
Again, although this possibility cannot be ruled out com-
pletely, it is quite unlikely. Thus, it would appear that
there is a discrepancy between the expected number of
With an adopted total
RIGHT ASCENSION (J2000)
09 30 40.840.640.440.240.039.8 39.639.4 39.239.0
21 43 40
Fig. 13.— The galaxy, SDSS J093039.96+214324.7, which is the
most likely optical counterpart to NGC 2903-HI-1. The image is
from the DSS2 Blue image and contours are shown in arbitrary
units, with the first contour set at the 2σ level. The scale, for a
distance equal to that of NGC 2903, is shown at right.
companions and the number observed.
Since we have a detection threshold that is lower still
than the value of the detected companion, we can also ask
how many dark matter clumps would we expect to see,
were they H I rich. For example, our lowest H I detection
threshold is 2×105M⊙(Table 4). If Mtot/MHI≈ 100 for
a dwarf galaxy in the field (equivalent to the N2903-HI-1
value), then we could have detected companions of total
mass as low as 2×107M⊙ via their H I . Therefore,
within our field of view, using the above relation, we
should have detected 230 galaxies as opposed to the one
observed. This is strongly discordant with ΛCDM. The
conclusion is that, if this many dark matter clumps exist
in the region, then they are clearly not H I rich, i.e. they
contain no H I or they contain H I at a level lower than
Using the Arecibo telescope with the ALFA receiver,
we have mapped NGC 2903 and its environment with
very high sensitivity and coverage.
source detection limit is 2 × 105M⊙ and almost 40
thousand square kpc of sky has been fully covered. The
Arecibo ALFA beams have been carefully characterized
as a function of azimuth, allowing us to clean each beam
as a function of azimuth from the H I data cube. With
a velocity coverage of 1035 km s−1and fine velocity res-
olution (2.6 km s−1), our combination of observing pa-
rameters makes this survey unique and among the most
sensitive and complete of a nearby galaxy. Although de-
tails of NGC 2903, itself, are left to future work, our
results show that the H I envelope around NGC 2903 is
much larger than previously known, extending to at least
3 times the optical galaxy diameter.
The fact that we have targeted an apparently iso-
Our lowest point
18 Irwin et al.
lated, non-interacting galaxy to a very low sensitivity
limit has clearly been an advantage in the search for
H I companions.The discovery of only one isolated
H I companion, N2903-HI-1, which appears to have a
small optical counterpart, is a significant result. The op-
tical companion is likely a dwarf galaxy with a stellar
mass approximately equal to its H I mass with the H I in
a broad envelope, approximately 8 times larger, around
it. The best estimate of its dynamical mass is 3 × 108
M⊙. We have no convincing HVC detections.
In the field surveyed, there are now two known com-
panion galaxies, our new discovery as well as what is
likely a dwarf spheroidal galaxy, UGC 5086, the lat-
ter with a total mass likely comparable to N2903-HI-
1.In this region, ΛCDM scenarios (specifically, the
Via Lactea model) predict 15 companions for a Milky
Way-type galaxy, with masses greater than 3 × 108M⊙.
Given our H I detection limits, however, if companions to
NGC 2903 contained H I at the 1% level in comparison to
their total masses, then we should have detected 230 of
them. If these clumps are present as predicted, they do
not contain appreciable H I . They may be starless dark
clumps or very low luminosity dark-matter dominated
We are grateful to students K. Marble of Queen’s Uni-
versity and I. Bah, A. Altaf, and J. Goldstein of Lafayette
College for their assistance with the data reductions.
Many thanks also to P. Perillat and the Arecibo staff
for their knowledge and assistance. JAI gratefully ac-
knowledges a grant from the Natural Sciences and Engi-
neering Research Council of Canada. GLH gratefully ac-
knowledges grants from the Lafayette College Academic
Research Committee.This research has made use of
the NASA/IPAC Extragalactic Database (NED) which
is operated by the Jet Propulsion Laboratory, Califor-
nia Institute of Technology, under contract with the Na-
tional Aeronautics and Space Administration. Funding
for the SDSS and SDSS-II has been provided by the Al-
fred P. Sloan Foundation, the Participating Institutions,
the National Science Foundation, the U.S. Department
of Energy, the National Aeronautics and Space Adminis-
tration, the Japanese Monbukagakusho, the Max Planck
Society, and the Higher Education Funding Council for
England. The SDSS Web Site is http://www.sdss.org/.
Auld, R., et al. 2006, MNRAS, 371, 1617
Avila-Reese, V., et al. 2001, ApJ, 559, 516
Barkana, R., & Loeb, A. 1999, ApJ, 523, 54
Barnes, D. G., et al. 2001, MNRAS, 322, 486
Barnes, D. G., & de Blok, W. J. G. 2004, MNRAS, 351, 333
Battaglia, G., et al. 2006, A&A, 459, 423
Begeman, K. 1987, H I Rotation Curves of Spiral Galaxies, PhD
Thesis, Rijksuniversiteit te Groningen
Begeman, K. G., Broeils, A. H., & Sanders, R. H. 1991, MNRAS,
Begum, A., et al. 2008, MNRAS, 386, 1667
Bekki, K., Koribalski, B. S., Ryder, S. D., & Couch, W. J. 2005,
MNRAS, 357, L21
Bell, E. R., McIntosh, D. H., Katz, N., & Weinberg, M. D. 2003,
ApJS, 149, 289
Benson, A. J., Frenk, C. S., Lacey, C. G., Baugh, C. M., & Cole,
S. 2002, MNRAS, 333, 177
Braun, R., Oosterloo, T. A., Morganti, R., Klein, U., & Beck, R.
2007, A&A, 461, 455
Briggs, F. H. 2004, in Recycling Intergalactic and Interstellar
Matter, IAU Symp. 217, Eds. Duc, Braine and Brinks, Astron.
Soc. Pac., p. 26
Buie, M. 2008,
Clowe, D., Randall, S. W., & Markovitch, M. 2007, Nuclear Physics
B, Vol. 173, Proceedings of the 7th UCLA Symposium on
Sources and Detection of Dark Matter and Dark Energy in the
Universe, Proceedings of the 7th UCLA Symposium on Sources
and Detection of Dark Matter and Dark Energy in the Universe,
Corbelli, E. & Salpeter, E. E. 1994, ApJ, 419, 104
Condon, J. J., Cotton, W. D., & Broderick, J. J. 2002, AJ, 124,
Cornwell, T., Braun, R., & Briggs, D. S. 1999, Synthesis Imaging
in Radio Astronomy II, 180, 151
Cort´ es, G. 2003, Private communication
Csabai, I., et al. 2003, AJ, 125, 580
Curran, S. J., Koribalski, B. S., & Bains, I. 2008, MNRAS, in press
de Blok, W. J. G., Zwaan, M., Dijkstra, M., Briggs, F. H., &
Freeman, K. C. 2002, A&A, 382, 43
del R´ ıo, M. S., Brinks, E., & Cepa, J. 2004, AJ, 128, 89
Diehl, R., et al. 2006, Nature, 439, 45
Diemand, J., Kuhlen, M., & Madau, P. 2007, ApJ, 657, 262
Diemand, J., Kuhlen, M., Madau, P. 2007a, ApJ, 667, 859
Dove, J. B., & Shull, J. M. 1994, ApJ, 430, 222
Drozdovsky, I. O., & Karachentsev, I. D. 2000, A&AS, 142, 425
Gilmore, G., et al. 2007, ApJ, 663, 948
Giovanelli, R., et al. 2005a, AJ, 130, 2598
Giovanelli, R., et al. 2005b, AJ, 130, 2613
Gnedin, N. Y., Kravtsov, A. V., & Chen, H.-W. 2008, ApJ, 672,
Putman, M.E., et al. 2002, AJ, 123, 873
Grcevich, J., & Putman, M. E. 2008, submitted
Haiman, Z., Rees, M. J., & Loeb, A. ApJ, 476, 458
Haynes, M. P., van Zee, L., Hogg, D. E., Roberts, M. S., &
Maddalena, R. J. 1998, AJ, 115, 62
Hewitt, J. N., Haynes, M. P., & Giovanelli, R. 1983, AJ, 88, 272
Heymans, C., et al. 2008, MNRAS, preprint arXiv:0801.1156v1
Hirashita, H., Takeuchi, T. T., & Tamura, N. 1998, ApJ, 504, L83
Irwin, M. J., et al. 2007, ApJ, 656, L13
Jackson, J. M., et al. 1991, ApJ, 375, 105
Karachentseva, V. E. 1973, The Catalogue of Isolated Galaxies,
Astrof. Issledovanija Byu. Spec. Ast. Obs., Vol. 8, pp. 3 - 49
Kauffmann, G., White, S. D. M., & Guiderdoni, B. 1993, MNRAS,
Kennicutt, R. C., Jr. 1998, ApJ, 498, 541
Kilborn, V. A., Koribalski, B. S., Forbes, D. A., Barnes, D. G., &
Musgrave, R. C. 2005, MNRAS, 356, 77
Kilborn, V. A., Forbes, D. A., Koribalski, B. S., Brough, S., &
Kern, K. 2006, MNRAS, 371, 739
Klypin, A., Gottl¨ ober, S., Kravtsov, A. V., & Khokhlov, A. M.
1999, ApJ, 522, 82
Koposov, S., et al. 2008, arXiv:0706.2687v2 [astro-ph]
Koribalski, B. S., et al. 2004, AJ, 128, 16
Kova˘ c, K., Oosterloo, T. A., & van der Hulst, J. M. 2005, in Near-
Field Cosmology with Dwarf Elliptical Galaxies, Proc. of the
IAU Coll. No. 198, 351
Kravtsov, A. V., Gnedin, O. Y., & Klypen, A. A. 2004, ApJ, 609,
Kroupa, P. 2001, MNRAS, 322, 231
Linder, S. M., et al. 2004, MNRAS, 353, 201
Lupton, R. 2005,
Maloney, P. 1993, ApJ, 414, 41
Martini, P., Regan, M. W., Mulchaey, J. S., & Pogge, R. W. 2003,
ApJS, 146, 353
Mateo, M. L. 1998, ARA&A, 36, 435
Matthews, L. D., van Driel, W., & Monnier-Ragaigne, D. 2008,
A&A, 365, 1
Mayer, L., et al. 2001a, ApJ, 547, L123
Mayer, L., et al. 2001b, ApJ, 559, 754
NGC 2903 and its Environment19 Download full-text
Meyer, M. J., et al. 2004, MNRAS, 350, 1195
Minchin, R. F., et al. 2003, MNRAS, 346, 787
Moore, B., et al. 1999, ApJ, 524, L19
Oosterloo, T. A., et al. 2007, A&A, 465, 787
P´ erez-Ram´ ırez, D., Knapen, J. H., Peletier, R. F., Laine, S., Doyon,
R., & Nadeau, D. 2000, MNRAS, 317, 234
Pisano, D. J., & Wilcots, E. M. 1999, AJ, 117, 2168
Pisano, D. J., & Wilcots, E. M. 2003, ApJ, 584, 228
Pisano, D. J., et al. 2004, ApJ, 610, L17
Pisano, D. J., et al. 2007, ApJ, 662, 959
Putman, M. E., et al. 2002, AJ, 123, 873
Putman, M.E., Grcevich,J.,
Regan, M. W., Sheth, K., & Vogel, S. N. 1999, ApJ, 526, 97
Ryan-Weber, E. V., et al. 2008, MNRAS, 384, 535
Ryder, S. D., et al. 2001, ApJ, 555, 232
Sanders, D. B., Mazzarella, J. M., Kim, D.-C., Surace, J. A., &
Soifer, B. T. 2003, AJ, 126, 1607
Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525
Shaviv, N. J., & Dekel, A. 2003, arXiv:0305527 [astro-ph]
Simon, J. D., & Geha, M. 2007, ApJ, 670, 313
Saintonge, A., et al. 2008, AJ, 135, 588
Spekkens, K., & Giovanelli, R. 2006, AJ, 132, 1426
&Peek, J.E. G. 2008,
Stanimirovi´ c, S., et al. 2006, ApJ, 653, 1210
Stoehr, F., White, S. D. M., Tormen, G., & Springel, V. 2002,
MNRAS, 335, L84
Taylor, E. N., & Webster, R. L. 2005, ApJ, 634, 1067
Thilker, D. A., et al. 2004, ApJ, 601, L39
Thom, C., et al. 2008, arXiv:0712.0612v2 [astro-ph]
Tsch¨ oke, D., Hensler, G., & Junkes, N. 2003, A&A, 411, 41
Verley, S., et al. 2007, A&A, 472, 121
Wakker, B. P., et al. 2007, ApJ, 670, L113
Wakker, B. P., et al. 2008, ApJ, 672, 298
Walter, F., et al. 2008, AJ, in press
Westmeier, T., Braun, R., & Thilker, D. 2005, A&A, 436, 101
Westmeier, T., Br¨ uns, C., & Kerp, J. 2005, A&A, 432, 937
Westmeier, T., Braun, R., Br¨ uns, C., Kerp, J., & Thilker, D. A.
2007, New Astronomy Reviews, 51, 108
Wong, O. I., et al. 2006, MNRAS, 371, 1855
Xue, X.-X., et al. 2008, arXiv:0801.1232v3 [astro-ph]
York, D. G. et al. 2000, AJ, 120, 1579
Zentner, A. R., & Bullock, J. S. 2002, Phys. Rev. D, 043003-1
Zwaan, M. 2001, MNRAS, 325, 1142
Zwaan, M., et al. 2005, MNRAS, 365, 1467