Relics of structure formation: extra‐planar gas and high‐velocity clouds around the Andromeda Galaxy

Abstract

Using the 100-m radio telescope at Effelsberg, we mapped a large area around the Andromeda Galaxy in the 21 cm line emission of neutral hydrogen to search for high-velocity clouds (HVCs) out to large projected distances in excess of 100 kpc. Our 3σ H i mass sensitivity for the warm neutral medium is 8 × 104 M⊙. We can confirm the existence of a population of HVCs near M31 with typical H i masses of a few times 105 M⊙. However, we did not detect any HVCs beyond a projected distance of about 50 kpc from M31, suggesting that HVCs are generally found in proximity of large spiral galaxies at typical distances of a few 10 kpc.Comparison with cold dark matter (CDM)-based models and simulations suggests that only a few of the detected HVCs could be associated with primordial dark matter satellites, whereas others are most likely the result of tidal stripping. The lack of clouds beyond a projected distance of 50 kpc from M31 is also in conflict with the predictions of recent CDM structure formation simulations. A possible solution to this problem could be ionization of the HVCs as a result of decreasing pressure of the ambient coronal gas at larger distances from M31. A consequence of this scenario would be the presence of hundreds of mainly ionized or pure dark matter satellites around large spiral galaxies like the Milky Way and M31.

Full text

Page 1

arXiv:0808.3611v1 [astro-ph] 27 Aug 2008
Mon. Not. R. Astron. Soc. 000, 1–21 (2008) Printed 27 August 2008(MN LATEX style file v2.2)
Relics of structure formation: extra-planar gas and
high-velocity clouds around the Andromeda Galaxy
T. Westmeier1,2, C. Br¨ uns2, and J. Kerp2
1CSIRO Australia Telescope National Facility, PO Box 76, Epping NSW 1710, Australia
2Argelander-Institut f¨ ur Astronomie, Universit¨ at Bonn, Auf dem H¨ ugel 71, 53121 Bonn, Germany
Accepted 1988 December 15. Received 1988 December 14; in original form 1988 October 11
ABSTRACT
Using the 100-m radio telescope at Effelsberg, we mapped a large area around the
Andromeda Galaxy in the 21-cm line emission of neutral hydrogen to search for high-
velocity clouds (HVCs) out to large projected distances in excess of 100 kpc. Our 3σ
Hi mass sensitivity for the warm neutral medium is 8 × 104M⊙. We can confirm the
existence of a population of HVCs with typical Hi masses of a few times 105M⊙near
the disc of M31. However, we did not detect any HVCs beyond a projected distance
of about 50 kpc from M31, suggesting that HVCs are generally found in proximity of
large spiral galaxies at typical distances of a few 10 kpc.
Comparison with CDM-based models and simulations suggests that only a few of
the detected HVCs could be associated with primordial dark-matter satellites, whereas
others are most likely the result of tidal stripping. The lack of clouds beyond a pro-
jected distance of 50 kpc from M31 is also in conflict with the predictions of recent
CDM structure formation simulations. A possible solution to this problem could be
ionisation of the HVCs as a result of decreasing pressure of the ambient coronal gas
at larger distances from M31. A consequence of this scenario would be the presence of
hundreds of mainly ionised or pure dark-matter satellites around large spiral galaxies
like the Milky Way and M31.
Key words: intergalactic medium – galaxies: evolution – galaxies: individual: M31.
1INTRODUCTION
One of the currently most favoured cosmological models is
the so-called Lambda Cold Dark Matter (ΛCDM) model.
It assumes that the evolution of the universe is dominated
by dark energy and dark matter which in sum account
for 96 per cent of the total energy density of the universe
(Spergel et al. 2003). An important prediction of CDM mod-
els is the hierarchical formation of gravitationally bound
structures. The smallest dark-matter haloes are expected
to form first, whereas larger structures, ranging from spi-
ral galaxies to galaxy clusters, are formed at a later stage
through merging and accretion of smaller dark-matter haloes
(bottom-up scenario). Numerical simulations of structure
formation in CDM cosmologies have successfully reproduced
the mass function and radial distribution of galaxies on the
scales of galaxy clusters. On smaller scales, however, sim-
ulations predict significantly more dark-matter haloes than
being observed (Klypin et al. 1999; Moore et al. 1999). This
discrepancy has been named the ‘missing satellites’ problem.
To overcome this problem, Blitz et al. (1999) suggested
that high-velocity clouds (HVCs) might be the gaseous coun-
terparts of the ‘missing’ dark-matter haloes around the
Milky Way. HVCs are gas clouds observed all over the
sky in the 21-cm line emission of neutral atomic hydrogen.
They were discovered with the Dwingeloo radio telescope by
Muller et al. (1963), and they are characterised by high ra-
dial velocities of typically |vLSR| ? 100 kms−1(see Wakker
1991 for details). If HVCs were the ‘missing’ satellites, they
could not have experienced significant star formation during
their evolution. Therefore, they would appear in the form of
pure gas clouds without any noticeable stellar population.
In addition, HVCs would be spread all over the Local Group
with typical distances of hundreds of kpc and high Hi masses
of about 107M⊙.
At the same time, the expected large distances of HVCs
from the Milky Way would result in fairly decent angular
diameters of the clouds. Therefore, Braun & Burton (1999)
defined a sub-sample of compact and isolated HVCs (the so-
called CHVCs) which are characterised by angular sizes of
less than 2◦FWHM as well as isolation and separation from
neighbouring Hi emission. The overall kinematics of the
CHVC population is consistent with a distribution through-
out the Local Group (de Heij et al. 2002a), making them
promising candidates for the ‘missing’ dark-matter satellites
around the Milky Way and the Andromeda Galaxy. Putman

Page 2

2 T. Westmeier, C. Br¨ uns, and J. Kerp
Table 1. Observational parameters of the northern and southern
part of our Effelsberg Hi blind survey of M31. The final baseline
RMS is given for a system temperature of 25 K at the given
velocity resolution. The specified sensitivity is the baseline RMS
times the spectral bin width at the original velocity resolution,
converted to Hi column density. Sensitivities for the WNM were
calculated for a brightness temperature of TB > 3σ, a velocity
resolution of 20 kms−1, and a line width of 25 kms−1FWHM.
parametersouthnorthunit
autocorrelator
bandwidth
polarisations
channels per polarisation
velocity resolution
frequency switching
total integration time
final baseline RMS
sensitivity
WNM sensitivity
WNM mass sensitivity
old
6.3
new
10 MHz
22
512
2.6
4096
0.5kms−1
normal in-band
180
45
21
2.2
180
60
5.5
1.3
s
mK
1016cm−2
1018cm−2
104M⊙
85
et al. (2002) extended the CHVC catalogue into the south-
ern hemisphere, using the Hi Parkes All-Sky Survey (Barnes
et al. 2001). The data for the northern and southern hemi-
spheres were later combined by de Heij et al. (2002a) into
an all-sky catalogue of 216 CHVCs.
Several arguments have been raised against the idea of
HVCs and CHVCs being the ‘missing’ dark-matter haloes
predicted by CDM cosmologies. First of all, attempts to
identify a population of HVCs in other nearby galaxy groups
(Zwaan 2001; Braun & Burton 2001; Pisano et al. 2004)
have failed, resulting in upper distance limits for HVCs and
CHVCs from the Milky Way of the order of 150 kpc. Addi-
tional evidence for HVCs being nearby at distances of the or-
der of only 10 kpc has been provided by the detection of Hα
emission from both HVCs and CHVCs (e.g., Kutyrev 1986;
Weiner et al. 2001; Tufte et al. 2002; Putman et al. 2003)
and from the determination of distance brackets for several
HVC complexes (e.g., Danly et al. 1993; Wakker et al. 1996,
2007,?; Thom et al. 2006, 2008). Small distances from the
Milky Way are also consistent with the head-tail structures
found in numerous HVCs and CHVCs (e.g., Br¨ uns et al.
2000, 2001; Westmeier et al. 2005) which are thought to re-
sult from ram-pressure interaction of the clouds with the
ambient gas of the Galactic corona (Quilis & Moore 2001;
Konz et al. 2002).
The problem of determining the spatial distribution of
HVCs can ultimately be solved by searching for the expected
HVC population around the nearest large spiral galaxy, the
Andromeda Galaxy (M31). First, the distance of M31 is well
known so that important physical parameters of HVCs, such
as their Hi mass or their diameter, can directly be deter-
mined from the observations. Second, we will look at the
HVC population of M31 from the outside, allowing us to
determine the (projected) radial distribution of HVCs. Fi-
nally, M31 is relatively close to the Milky Way. Therefore,
the sensitivity and angular resolution of large single-dish
radio telescopes will easily allow us to detect HVCs of the
mass and size of the large HVC complexes observed near the
Milky Way. The first comprehensive search for HVCs around
M31 was carried out by Thilker et al. (2004) with the 100-m
Green Bank Telescope. They mapped an area of 7◦× 7◦in
the 21-cm line of Hi with high sensitivity and discovered a
population of about 20 HVCs out to the edge of their map at
about 50 kpc projected distance from M31. Their discovery
marked the first detection of an extensive HVC population
around a galaxy other than our own. Several of these HVCs
were studied in detail with the Westerbork Synthesis Radio
Telescope (WSRT) by Westmeier et al. (2005).
Thus, the radial extent of the population of HVCs and
CHVCs around galaxies like the Milky Way and M31 is con-
fined by two limits. An upper limit of about 150 kpc can
be derived from the non-detection of HVCs in other galaxy
groups, and a lower limit of about 50 kpc is marked by the
edge of the Hi survey of Thilker et al. (2004) out to which
HVCs were found to be present. Therefore, we decided to
carry out a complementary Hi survey with the 100-m ra-
dio telescope at Effelsberg to search for HVCs and CHVCs
around M31 out to much larger projected distances in excess
of 100 kpc. This would allow us to trace the distribution of
the HVC population of M31 over its entire radial extent and
compare our results with the predictions of CDM structure
formation scenarios.
Our paper is organized as follows. In Sect. 2 we explain
the technical aspects of our observations. In Sect. 3 our data
reduction, calibration, and analysis strategy is described. In
Sect. 4 we discuss the various completeness issues of our
survey and their implications for our results. In Sect. 5 we
describe the derived observational and physical properties of
the high-velocity clouds detected in our survey. In Sect. 6 we
discuss the results of our comparison of the observational pa-
rameters of the HVCs near M31 with different CDM-based
models and with the distribution of satellite galaxies around
M31. Sect. 7 discusses the evidence for different hypotheses
on the origin of HVCs. Finally, Sect. 8 summarises our re-
sults and conclusions.
2OBSERVATIONS
The observations for our Hi blind survey of M31 were carried
out between July 2003 and August 2004 with the 100-m
radio telescope at Effelsberg. The observational parameters
of the survey are summarised in Table 1. As a compromise
between spatial coverage and observing time we chose the
trapezoidal mapping area outlined in Fig. 1. It extends out
to an angular distance of about 10◦from M31 in the south-
eastern direction, corresponding to a projected distance of
about 140 kpc which is about two thirds of the projected
distance towards M33. In this direction we expect the least
confusion with foreground Hi emission from the Milky Way.
In the north-western direction our map extends out to about
5◦, corresponding to 70 kpc in projection.
The map was split into individual rows of constant
declination with 35 pointings each. The individual point-
ings were separated in right ascension by 760 arcsec ≈
HPBW ×√2. Adjacent rows were separated in declination
by 380 arcsec ≈ HPBW/√2 and shifted in right ascension
by the same amount, resulting in an orientation angle of the
entire map of 45◦with respect to the equatorial coordinate
system. As a consequence, the final map is beam-by-beam
sampled and oriented almost perpendicular to the major

Page 3

Relics of structure formation: HVCs around M313
12345
M31
45°
5°
10°
7.1°
Figure 1. The plot shows the geometry of the field (thick solid line) mapped around M31. The dashed circles separate the five different
regions (labelled with numbers) for which individual spatial completeness calculations have to be performed.
axis of the disc of M31. Aligning the map along the minor
axis of M31 has the advantage of avoiding confusion with
disc emission along the major axis as far as possible.
The southern part of the map was observed with the old
1024-channel autocorrelator, whereas for the northern part
we used the new 8192-channel autocorrelator (AK90). The
reason for changing from the old to the new autocorrelator
in early 2004 was that in August 2003 the quality and sta-
bility of the spectral baselines suddenly degraded. Several
maxima and minima occurred over the bandpass, requir-
ing a higher-order polynomial to approximate the shape of
the baseline. The more severe problem was that the posi-
tions and amplitudes of the maxima and minima changed
significantly over short time scales of the order of one hour,
making baseline correction difficult. As the quality of the
spectra was not sufficient for the objectives of our M31 sur-
vey we decided to carry out test observations with the new
AK90. The test observations demonstrated that the AK90
provided significantly better baseline quality and stability
than the old 1024-channel autocorrelator. Therefore, we de-
cided to continue the survey in the northern direction using
the AK90 instead of the old autocorrelator. Another advan-
tage of the AK90 is its larger bandwidth of 10 MHz. This
allowed us to use the in-band frequency switching method
which spends 100 per cent of the integration time on source.
As a result the baseline RMS in the northern part of the
map is by a factor of about
ern part. Throughout this paper we follow the conservative
approach of using the 1024-channel autocorrelator specifica-
tions as the basis for all sensitivity-related parameters.
√2 lower than in the south-
3 DATA REDUCTION AND ANALYSIS
3.1Flux calibration
The standard flux calibration source S7 was observed every
6 hours and at the beginning and end of each observing run.
The corresponding flux calibration factors for the two polar-
isations were determined by the software nautocal (based
on the results obtained by Kalberla et al. 1982), yielding
statistical calibration errors of the order of only 1 per cent
over an entire observing run of typically 10 to 15 hours.
3.2 Stray radiation
A potential problem could be that Hi emission from the disc
of M31 could have been detected through the near side lobes
of the telescope and create the impression of extra-planar gas
in proximity of the disc. The beam pattern of the Effelsberg
telescope at λ = 21 cm was studied in detail by Kalberla
et al. (1980) through combined observations of Cygnus A
and Cassiopeia A. The overall crosswise structure in their
fig. 1 results from diffraction by the four support legs. Signif-
icant side lobes can be found particularly in the north-south
direction. At 1◦angular distance from the pointing direction
the sensitivity drops to about −30 dB. Typical brightness
temperatures of 20 K, as observed along the Hi ring of M31,
would therefore be attenuated to about 20 mK which is far
below the noise level of our data at the original spectral res-
olution. The near side lobes in the east-west direction are
not as extended as those in the north-south direction, but
instead the sensitivity levels are somewhat higher near the
pointing centre. At an angular distance of 15 arcmin the
sensitivity drops to about −20 dB, resulting in an attenua-
tion of a 20 K signal to about 200 mK. This is of the same
order of magnitude as the expected signals of high-velocity
clouds around M31. The near side lobes, however, will only
affect regions of the map which are in projection close to
the Hi disc of M31. All areas beyond a projected distance
of the order of 10 kpc from the disc will not be influenced
anymore by stray radiation from M31. In addition, stray ra-
diation through the near side lobes should result in more
extended emission, whereas most HVCs and in particular
CHVCs are expected to be unresolved. Consequently, stray
radiation will not affect our observations since most of the
expected HVCs and CHVCs should be located at sufficiently
large angular distances from the disc of M31 and will likely
be unresolved by the HPBW of 9 arcmin.
3.3 Spectral baseline correction
To improve the quality of the spectral baseline we imple-
mented a two-step baseline correction procedure. The first
step was applied to each of the two polarisations separately.
All 35 spectra in a single row of the map were averaged, and

Page 4

4T. Westmeier, C. Br¨ uns, and J. Kerp
Figure 2. Spatial completeness of the M31 survey as a function
of projected distance, r, from the centre of M31. The dashed line
shows the global completeness function, CG(r), and the solid line
the local completeness function, CL(r). The vertical dotted lines
indicate the radii a, b, c, and d at which the completeness function
changes abruptly due to the special geometry of the observed field.
a polynomial of usually 8thorder over ∆v = 900 kms−1was
fitted to the averaged spectrum after manually setting line
windows. This constant reference polynomial – representing
the average baseline over a time period of about two hours
– was then subtracted from each individual spectrum of the
row. The resulting spectra were already quite smooth as the
baseline shape at Effelsberg was fairly stable over a period
of two hours. The second step was to average all spectra at
a single position on the sky and fit a polynomial to the re-
sulting spectrum. Because of the previous subtraction of the
reference polynomial, a low-order polynomial was sufficient
for the final averaged spectrum. In most cases, the baseline
shape could be sufficiently described by a 4thor 5thorder
polynomial across a velocity range of 900 kms−1. Only in
very few cases a lower (down to 1st) or higher (up to 8th)
order was necessary. After cutting out a few strong radio
frequency interference signals by hand, the resulting spectra
were Hanning-smoothed to 4.1 kms−1velocity resolution
and combined into a FITS data cube for further analysis.
3.4 Identification of sources
The final survey consists of 3586 individual pointings for
each of which an Hi spectrum in the velocity range of vLSR ≈
−750...+150 kms−1was available for further analysis. As
we were interested in the detection of HVCs and CHVCs
around M31, the survey was particularly optimised for the
detection of compact sources with narrow spectral lines. In
the in-band frequency switching mode nearly 100 per cent
of the integration time is spent on source, maximising the
sensitivity for the detection of faint CHVCs. The spectral
baselines in this special observing mode, however, are not
particularly smooth and stable, restricting the detectability
of faint, diffuse sources.
To identify HVC/CHVC candidates of M31 all spectra
were searched by eye for potential emission lines. The cri-
terion for HVC emission was a separation of the signal in
phase space from the disc emission of M31 (and the Milky
Way), where phase space is considered the observable three-
01234567
0
20
40
60
80
100
signal-to-noise ratio
fraction (per cent)
Figure 3. Fraction of detected spectral lines as a function of
signal-to-noise ratio, derived from the investigation of 500 arti-
ficial spectra as described in the text. About 50 per cent of all
signals with a peak intensity of 3.5σ were detected. Above 5σ we
are virtually complete with only a single of the 14 lines above 5σ
being missed.
dimensional sub-space consisting of the two spatial coordi-
nates in the plane of the sky and the radial velocity as the
only available kinematic component. Thus, signals could ei-
ther be located on the same line of sight as the disc emission
of M31 but at different radial velocities, or they could be lo-
cated outside the disc at any radial velocity. At the same
time, we did not consider objects with vLSR > −140 kms−1
to avoid confusion with Galactic foreground emission.
It is important to note that we have to change the defi-
nition of ‘high-velocity cloud’ for our M31 observations. The
HVCs around the Milky Way were characterised by their pe-
culiar radial velocities as seen from a point of view within
the Galactic Hi disc. In the case of M31, however, we are
looking at the entire galaxy and its HVC population from
the outside. Therefore, the term ‘HVC’ refers to all circum-
galactic gas clouds which are thought to be the equivalents of
the circumgalactic gas clouds around the Milky Way known
as ‘high-velocity clouds’. This includes clouds whose radial
velocities are compatible with the rotation curve of M31 but
which are located well beyond the edge of the Hi disc.
4 COMPLETENESS OF THE DATA
4.1 Spatial coverage of the map
An important question arising from the approach of a vi-
sual inspection of the survey is the question of detectability
and completeness of HVC/CHVC candidates. A first limi-
tation of the completeness of our M31 survey results from
the spatial incompleteness of our map at larger radii. The
geometric situation is illustrated in Fig. 1. There are five dif-
ferent regions for which analytic expressions for the spatial
completeness have to be derived separately. The five regions
are separated from each other at radii of 2.◦5, 3.◦5, 7.◦9, and
12.◦7 from the centre of M31, corresponding to projected dis-
tances of 34 kpc, 48 kpc, 108 kpc, and 174 kpc if the distance
of M31 is assumed to be 780 kpc (Stanek & Garnavich 1998).
There are two ways of calculating the spatial completeness
within these regions as a function of radial distance from the

Page 5

Relics of structure formation: HVCs around M315
log NHI (cm−2)
NHI (1021 cm−2)
20 kpc
20 kpc
Cloud
Davies’
18
19
20
21
22
0
2
3
1
HVCs
HVCs
HVCs
Figure 4. Hi column density maps of our Effelsberg blind survey of M31. Top: Integrated Hi column density in the velocity range of
vLSR= −620...−24 kms−1. The contours were drawn at 4, 8, 16, and 25 × 1020cm−2. The faint, filamentary emission all over the
map originates from Galactic Hi. Bottom: Integrated Hi column density in the velocity range of vLSR= −620...−139 kms−1, using a
logarithmic intensity scale. The black contours correspond to 2 ×1018and 1 ×1019cm−2. The white contours are those from the upper
map. Several regions of extra-planar gas and high-velocity clouds can be seen all around M31.
centre of M31. One way is to calculate the local filling factor
at distance r from the centre of M31 which is given by the
ratio of the length of the arc intersecting the map over the
total arc length, 2π r, of the full circle. The other way is to
calculate the global filling factor of the map within a certain
radius, r, which is the ratio of the area being covered by the
map within r over the total area, π r2.
The two spatial completeness functions, CL(r) and
CG(r), are plotted in Fig. 2 as a function of projected dis-
tance, r, from the centre of M31. Within r = 34 kpc the
survey is complete. Beyond r = 174 kpc the local complete-
ness drops to 0 whereas the global completeness decreases
with 1/r2. At intermediate radii both functions have a more
complex behaviour. They drop abruptly at r = 34 kpc with
the local completeness function, CL(r), decreasing much
faster than the global completeness function, CG(r). Lo-
cally, 50 per cent completeness is reached at a distance of
r50 = 48 kpc already, whereas on the global scale the cor-
responding radius is r50 = 77 kpc. At a projected distance
from M31 of 100 kpc the completeness is CL = 12 per cent
and CG = 36 per cent. This means that out to 100 kpc our
survey still covers more than a third of the total area, but
the local completeness at this distance is already fairly low
because the survey area is concentrated around the central
region.
4.2Spatial sampling of the map
Another reason for spatial incompleteness is related to the
spatial sampling of the map. The individual pointings of
the map are separated by the HPBW of the telescope of
9 arcmin, so that the sensitivity is significantly decreased at
positions which are located exactly in between four individ-
ual pointings. Assuming a two-dimensional Gaussian sensi-
tivity function for the main beam, the sensitivity of each
beam at the position in between four neighbouring beams
(i.e. the corner of each grid element) is 0.25. Averaging the
four neighbouring beams results in a decrease of the RMS
by a factor 2 and, thus, in a slightly higher effective sensitiv-
ity of 0.5. As a consequence, the sensitivity for unresolved
sources is in the range of 0.5 to 1, depending on the source