A&A 453, 459-475 (2006)
DOI: 10.1051/0004-6361:20035672
Ch. Nieten1 - N. Neininger1,2,3 - M. Guélin3 - H. Ungerechts4 - R. Lucas3 - E. M. Berkhuijsen1 - R. Beck1 - R. Wielebinski1
1 - Max-Planck-Institut für Radioastronomie,
Auf dem Hügel 69,
53121 Bonn, Germany
2 - Radioastronomisches Institut der Universität Bonn,
Auf dem Hügel 71,
53121 Bonn, Germany
3 - Institut de Radioastronomie Millimétrique,
300 rue de la piscine, 38406 St. Martin d' Hères,
France
4 - Instituto de Radioastronomía Milimétrica,
Avenida Divina Pastora 7, 18012 Granada, Spain
Received 13 November 2003 / Accepted 5 December 2005
Abstract
Aims. We study the distribution of the molecular gas in the Andromeda galaxy (M 31) and compare this with the distributions of the atomic gas and the emission from cold dust at
m.
Methods. We obtained a new 12CO(J = 1-0)-line survey of the Andromeda galaxy with the highest resolution to date (
,
or 85 pc along the major axis), observed On-the-Fly with the IRAM 30-m telescope. We fully sampled an area of
with a velocity resolution of
.
In several selected regions we also observed the 12CO(2-1)-line.
Results. Emission from the 12CO(1-0) line was detected from galactocentric radius R=3 kpc to R=16 kpc with a maximum in intensity at kpc. The molecular gas traced by the (velocity-integrated) (1-0)-line intensity is concentrated in narrow arm-like filaments, which often coincide with the dark dust lanes visible at optical wavelengths. Between R=4 kpc and R=12 kpc the brightest CO filaments define a two-armed spiral pattern that is described well by two logarithmic spirals with a pitch angle of 7
-8
.
The arm-interarm brightness ratio averaged over a length of 15 kpc along the western arms reaches about 20 compared to 4 for H I at an angular resolution of
.
For a constant conversion factor
,
the molecular fraction of the neutral gas is enhanced in the spiral arms and decreases radially from 0.6 on the inner arms to 0.3 on the arms at
kpc. The apparent gas-to-dust ratios
and
increase by a factor of
20 between the centre and
,
whereas the ratio
only increases by a factor of 4.
Conclusions. Either the atomic and total gas-to-dust ratios increase by a factor of 20 or the dust becomes colder towards larger radii. A strong variation of
with radius seems unlikely. The observed gradients affect the cross-correlations between gas and dust. In the radial range R=8-14 kpc total gas and cold dust are well correlated; molecular gas correlates better with cold dust than atomic gas.
The mass of the molecular gas in M 31 within a radius of 18 kpc is
at the adopted distance of 780 kpc. This is 7% of the total neutral gas mass in M 31.
Key words: ISM: molecules - galaxies: individual: M 31 - galaxies: ISM - galaxies: spiral - radio lines: galaxies
Star formation and spiral structure in galaxies require the coupling of processes operating on linear scales so different that they are hard to study in a single galaxy. The small structures are difficult to observe in external galaxies, whereas large structures are hard to see in the Milky Way due to distance ambiguities. Single-dish telescopes were used to survey CO in galaxies (e.g. Nakano et al. 1987; Braine et al. 1993; Young et al. 1995) but with limited angular resolution. Molecular spiral arms were barely resolved in these surveys even in the nearest galaxies (e.g. Koper et al. 1991; Garcia-Burillo et al. 1993; Loinard et al. 1996; Heyer et al. 2004). Only in the Magellanic Clouds did the single-dish surveys resolve giant molecular clouds (Israel et al. 1993). More recently mm-wave interferometer surveys like the BIMA SONG (Regan et al. 2001) gave vastly improved data on nearby galaxies like M 51, resolving molecular arms into cloud complexes. This instrument was also used for an all-disk survey of M 33, about ten times closer to us than M 51, in which individual molecular clouds are recognized (Engargiola et al. 2003). The IRAM Plateau de Bure interferometer has resolved molecular clouds in M 31 into components (Neininger et al. 2000a) enabling close comparisons with molecular clouds in the Milky Way.
The nearest large spiral is the Andromeda Nebula, M 31. Its distance
of
Mpc (Stanek & Garnavich 1998) ranks among the best known for any galactic or
extragalactic nebula; the accuracy of this distance allows us to derive
accurate luminosities and masses. At this distance,
along the
major axis corresponds to
pc. The large inclination of
M 31,
,
degrades the resolution along the minor axis by a
factor of 4.6, but has the advantage of yielding accurate in-plane
velocities. The proximity of M 31 gives us the chance to see many
details of the distribution and kinematics of the gas, as well as the
relation of the gas to the spiral structure and to star formation.
The contents of stars, dust, and atomic gas in M 31 are well known.
The whole galaxy has been mapped in the 21 cm line of H I with
resolution by Brinks & Shane
(1984, hereafter B&S) and its northeastern half with
resolution by Braun (1990). It has been
entirely mapped in the mid and far infrared by the IRAS, ISO, and
Spitzer satellites (see Haas et al. 1998 and Schmidtobreick
et al. 2000 for the ISOPHOT map at
m,
and Gordon et al. 2004 for the MIPS maps at
m,
m and
m). Furthermore, M 31 was
partially mapped with ISOCAM (5.1-
m) at
resolution (see e.g. Pagani et al. 1999). Comparisons of
the emission in different wavelength ranges - like UV, optical,
H I, FIR (
m and
m) and radio continuum
emissions - have also been reported (Loinard et al.
1999; Pagani et al. 1999;
Keel 2000; Lequeux 2000; Nieten et al.
2000; Berkhuijsen et al. 2000; Gordon et al.
2004).
So far, the situation was not as favourable for the molecular gas. Prior
to ours, the only complete CO survey of M 31 was made with a 1.2-m
diameter telescope and had a resolution of
(Koper et al. 1991; Dame et al. 1993). More recently, a
survey of the southwestern half, made at an angular resolution of
with the FCRAO 14-m telescope, was published by Loinard et al. (1996, 1999). The latter authors
(1999, their Table 2) give a nearly complete overview of
previous CO observations of M 31. Loinard et al. (1999)
and Heyer et al. (2000) found many similarities, but also
clear differences, between properties of the molecular gas in M 31
and those in the Milky Way.
Our survey, made with the IRAM 30-m telescope in the
12CO(J = 1-0) line, has a resolution of
corresponding to 85 pc along the major axis. It is much more sensitive
than the previous surveys and detects all clouds with
(=
rms noise).
In this article we present the CO distribution
in the bright disk of the galaxy. We derive some important basic
results using simple assumptions, e.g., a constant conversion factor
from CO intensity to molecular column density.
We discuss the spiral-arm structure of the neutral gas
and the arm-interam brightness contrast in Sect. 3. In addition
to the 12CO(J = 1-0) line, we observed several
selected areas covering bright arm segments in the
12CO(J = 2-1) line with high sensitivity; we discuss the
line ratios in Sect. 3.3. In a previous publication, based on one third
of the present data (Neininger et al. 1998), we reported a
tight correlation between the CO sources and the dark dust lanes.
In Sect. 4 we return to this point and compare the CO distribution with
those of H I, FIR (
m) and
20 cm radio
continuum. Radial profiles of the various constituents are discussed in
Sect. 4.1 and correlations between CO, H I and FIR (
m)
in Sect. 4.2. In Sect. 4.3 we derive the molecular and total gas mass.
The CO velocity field is described in Sect. 5. Our results are
summarized in Sect. 6. Preliminary reports on this survey
are given by Guélin et al. (2000), Neininger et al.
(1998, 2000b) and Nieten et al.
(2000).
Our survey was carried out with the IRAM 30-m telescope between
November 1995 and August 2001. The observations were made On-the-Fly in two steps: in a first step a field typically
in size was scanned back and forth in the
direction parallel to M 31's minor axis, Y, at a speed of
.
The successive scans were spaced by
in
the orthogonal X direction. At the beginning and at the end of each
scan, a reference position, located
or
away
from the major axis
and free of CO or H I emission,
was observed for 30 s. Every 1-2 h the telescope pointing was
checked on planets and nearby quasars. A second reference position,
located within M 31 and showing strong CO emission, was observed for
calibration purposes (see below). The telescope focus was checked
several times a day, in particular
after sunrise and sunset. In a second step the observations were
repeated by scanning the same field in the orthogonal direction,
parallel to the major axis. The data recorded by the backends were read
every second of time, so that the data cube obtained by combining the
two orthogonal maps was fully sampled on a
grid.
The reduction procedure was described in some detail by Neininger
et al. (2000b). After calibration (see below),
subtraction of the off-source reference spectrum and a baseline
for individual spectra in a map, two orthogonal maps were combined
using "basket-weaving'', the de-striping technique of Emerson &
Gräve (1988). This code was adapted to
work on two orthogonal channel maps before averaging them
(Hoernes 1997). Examples of the technique are shown in
Neininger et al. (2000b), Hoernes (1997) and
Emerson & Gräve (1988). In this process, the CO
map was smoothed from
to
FWHM.
The maps shown in Fig. 1 are the combination of 12 individual fields, as listed in Table 1. Each field is fully sampled and all fields together contain nearly 1.7 million spectra (before gridding) obtained in about 500 h of effective observing time. We present individual spectra for selected regions in Fig. 2.
![]() |
Figure 1:
a) The velocity-integrated intensity distribution of the
12CO(1-0) spectrum,
![]() ![]() ![]() ![]() |
Open with DEXTER |
Table 1: Data on observed fields of M 31.
When observing the CO(1-0) line, the 30-m telescope allows
simultaneous observations of two polarizations of the (2-1) line.
The (2-1) line, however, is weaker than the (1-0) line, and the
integration time per beam is about
smaller because the beam
area is
smaller. Moreover, the receivers are
noisier and the sky opacity is higher at 230 GHz (2-1) than at 115 GHz (1-0). This made it difficult to detect the (2-1)-line emission in
our survey, except for the brightest clouds. In order to improve the
signal-to-noise ratio and the sampling for (2-1), we re-observed
several rectangular regions,
-
wide by
-
long, with half the scanning velocity and twice
the sampling, when the zenith opacity at 230 GHz was favourably low
(
0.2). We discuss some general results of these observations in
Sect. 3.3.
In order to confirm the reliability of the OTF method, to integrate some
emission-free positions to a lower noise level and to check several
apparent discrepancies with previous CO observations, we re-observed
about 200 positions, located inside as well as outside the arms,
in the position-switching mode with integration times of
10-30 min. Some of these results are used in Sect. 3.2
.
We used two SIS receivers with orthogonal polarizations to observe the
(1-0) line and a similar system to observe the (2-1) line. The
receiver temperatures in the standard reference plane (before
the polarization splitter) were close to 90 K (SSB) at 115 GHz at the
beginning of our survey (fields 1-3, see Table 1) and
close to 50 K at the end (fields 8-11). After addition of the
atmospheric contribution, the system temperature was between 200 and
400 K at both 115 GHz and at 230 GHz. The backends consisted of two
MHz filterbanks at 115 GHz and of 2 autocorrelators with
resolutions of 0.8 MHz and total bandwidth of 320 MHz at 230 GHz.
A channel width of 1 MHz corresponds to a velocity resolution of
for the (1-0) line and of
for the (2-1) line.
![]() |
Figure 2:
Sample of spectra extracted from the data cube of the
12CO(1-0) survey of M 31 (full lines). At some positions
the 12CO(2-1) spectrum is also shown (dotted lines,
shifted down by 0.1 K for clarity). The velocity resolution is
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
The standard calibration at the 30-m telescope is equivalent to the
chopper-wheel calibration method for observations at millimeter
wavelengths; it gives the antenna temperature
,
corrected for atmospheric losses and forward efficiency,
(for details see Downes 1989). The
main-beam brightness temperature,
,
can then be
calculated from
![]() |
(1) |
A corollary of the high beam efficiency of the 30-m telescope at
115 GHz is the low error beam. Greve et al. (1998) found that
the far-beam pattern can be described by 3 error beams with half-power
widths of
,
and
,
respectively. The
first two could pick up signals from regions with similar radial
velocities as observed in the main beam. As they contributed only a few
percent to the main-beam signals, they hardly affected the observed
spectra. The contribution of the third error beam, which is larger than
the disk of M 31, was negligible.
The effective integration time per
beam for the large
12CO(1-0) map was 64 s,
yielding a rms noise of
33 mK per 1 MHz channel
(
scale) in the southern fields (1-4) and
25 mK per 1 MHz in the northern ones (see Table 1). The
corresponding values for the small maps are
,
173 s
and 15 mK for the (1-0) line, and
,
57 s
and 35 mK per 1 MHz channel for the (2-1) line.
The distribution of velocity-integrated CO-line intensities,
,
is shown in
Fig. 1a. The rms noise in the velocity-integrated
emission varies between the fields, but is typically about
in the southern part and about
in the northern half of the
survey or
for the total map (see
Table 1). We note that the sensitivity of our survey to
point-like and extended sources exceeds that of Loinard et al.
(1999) by factors >8 and >1.5, respectively.
The survey of the Andromeda galaxy presented in this paper is the largest and most detailed molecular-line survey ever made of an extragalactic object. Most of the emission in Fig. 1 appears concentrated on radii between 3 and 12 kpc and occurs as long and narrow filaments that strongly suggest a spiral arm structure. In addition, we see a number of scattered CO clouds of weak intensity between the spiral arms; sometimes these form bridges. The inner arms at radii near 5 kpc are remarkably bright, especially in the NE half of the galaxy.
Close to the centre of M 31, CO emission is very weak. Melchior et al.
(2000) have found CO emission of about
mK in a dark-cloud complex located at a distance of 350 pc from
the centre using position switching and long integrations. The
integration time per beam was about 6 h and the rms noise 2 mK
per
-wide channel, which is much better than the noise in the
present survey.
Only a few scattered clouds are visible at large distances from the centre. The most distant cloud in this survey was found at a deprojected radius of 19.4 kpc. The spectrum of this cloud is shown in Fig. 2a. The central velocity of the emission fits well to the velocity field of M 31 (see Fig. 1b). The cloud is located in the outermost part of the spiral arm A-N modelled by Braun (1991) near several H II regions.
In the following subsections we model the spiral pattern and analyse the arm-interarm contrast.
Although M 31 was classified at the beginning by Hubble
(1929) as an Sb type spiral, attempts to draw its
spiral pattern have mostly been inconclusive. For example, Baade
(1963) used the young stars and Hodge (1979) the
open star clusters as spiral arm tracers. These results were
summarized by Hodge (1981b). Due to absorption of optical
light by dust-rich lanes, the presence of a bright optical
bulge and the lack of H I radio line emission in the inner disk,
there was no consensus even on whether the arms are trailing or
leading. More recently, Braun (1991) proposed a non-planar
trailing two-armed spiral pattern with varying pitch angle on the basis
of the H I interferometric data. His model accounts fairly
well for the outer H I arms, but it does not trace the innermost
structures. Based on their analysis of the m dust emission,
Haas et al. (1998) suggested that M 31 may be closer to a
ring galaxy than to a spiral.
The CO emission, which traces the dense molecular gas, is better
suited to determining M 31's spiral pattern because the CO arms are:
(i) thinner than the H I arms; (ii) less patchy than the
H
arms; and (iii) not affected by absorption. As discussed
below, the CO arm-interarm contrast is also much higher than in H I. Furthermore, the linear resolution of the CO survey is
sufficient for distinguishing neighbouring arms even on the minor axis.
In order to determine the spiral pattern in an objective way, we
decomposed the CO map into 170 individual "clouds''
using the CLUMPFIND analysis program (Williams et al.
1994). CLUMPFIND provided the position of "clouds'',
deprojected them (assuming an inclination of the molecular disk of
)
and transformed them into polar coordinates. The
resulting data points, plotted in semi-logarithmic coordinates, were
then least-square fitted with two straight lines representing two
simple logarithmic spirals (see Figs. 3a and b). The
fits led to very similar pitch angles for the two spirals, namely
and
,
and to a phase shift of
roughly
.
The derived pitch angles agree well with
the value of
for the mean of the 2 spiral arms
observed in H I (Braun 1991) and with the
mean value of
for the optical spiral arms (Baade & Arp 1964). The oscillations visible in Fig. 3b
indicate variations in pitch angle that could be partly due to
variations in inclination angle observed by Braun (1991).
We note that these geometrical spiral fits required centre positions
shifted to the NE with respect to the nucleus of M 31, i.e.
for the full-line
spiral and
for the
dashed spiral (Figs. 3a and b). Such a displacement is
not surprising, as the bulge of M 31 is likely to host a bar (Stark &
Binney 1994), which makes it difficult to trace the
spiral arms down to the nucleus.
This displacement and the fact that on the minor axis the outer CO
arms lie slightly outside the fitted spirals
may also indicate that the molecular spiral arms have
inclinations that differ from the adopted inclination of
of
the main plane of M 31, like the H I arms (Braun 1991). For a detailed comparison with the H I arms, a
more complete analysis of the molecular arm structure should be made
that also includes the velocity structure of the arms. This could
be the subject of a future study.
For comparison, we tried to represent the CO emission at R=8-12 kpc by a circular ring, as suggested by Haas et al. (1998). The result of the fit is shown in Fig. 3c. Even at these radii the spiral pattern appears to be a better description of the CO distribution than a ring-like structure.
The derived spiral pattern is a good fit to the CO emission up to about
12 kpc from the centre, except for the region of the giant stellar
association NGC 206 (
), the velocity of
which strongly deviates from the normal rotation velocity.
In particular, it accounts for the splitting of the CO emission along
the eastern minor axis into two distinct arms. These arms are clearly
seen in the
m emission from hot dust (MSX satellite map, Moshir
et al. 1999; ISOCAM map of the central bulge of Willaime
et al. 2001, see Fig. 7 below) and are also
predicted by Braun's model although they are only marginally visible
in H I and H
emissions.
At radii 12 kpc the fitted spiral pattern starts to deviate from
the observed spiral. There the CO emission becomes fainter and the
filaments more difficult to trace. We note, however, that the ISO
m map of Haas et al. (1998) and the H
image
of Walterbos (2000) show a clear arm-like structure that
seems to lie on the extension of the fitted spiral to the NE, far
beyond our CO image.
The CO peak-to-dip ratio reaches a maximum >20 in all quadrants,
showing that the arm-interarm intensity contrast is high almost
everywhere. The shallow dip at
in the SE quadrant reflects
that the two brightest spiral arms (
and 12 kpc) are only
partly resolved in this quadrant (see Fig. 1). The
shoulders visible in Fig. 4 at
(NE quadrant)
and at 10 kpc (SW quadrant) come from two weak, nearly symmetrical
arm-like substructures that are clearly visible in Fig. 1
(e.g. from
to
in the NE quadrant).
![]() |
Figure 3:
Spiral structure in M 31.
a) Two logarithmic spirals (solid and dashed lines) fitted
through the distribution of CO "clouds'' in the radial range
R=4-12 kpc. They are overlayed onto the CO brightness distribution
(contours).
b) Least-square fits in the
![]() ![]() ![]() ![]() |
Open with DEXTER |
![]() |
Figure 4:
a) A family of logarithmic spiral segments (defined in the plane
of M 31) fitting the CO arms in the NW quadrant, superimposed onto the
velocity-integrated CO emission, I1-0, smoothed to a resolution of
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
The apparent half-power width of the arms in Fig. 4b
is 1-2 kpc. This includes the true arm width within the plane of M 31,
the arm thickness perpendicular to the plane and the effect of
beam-smoothing. Nieten (2001) attempted to disentangle these
3 effects for the NW part of the 10 kpc arm and derived a "true'' width
of pc and an arm thickness of
pc.
We repeated the above analysis for the H I map of Brinks &
Shane (1984), smoothed to the same resolution of
,
for the 20 cm radio continuum map of Beck et al.
(1998) at the original
resolution
and for the
m map of Haas et al. 1998) at
resolution
(see Fig. 4c). The arm-interarm contrast of the
integrated H I-line brightness is
4, which is 5 times
smaller than that of CO in the same disk section, and the apparent
half-power arm width of
is 3 times larger. In contrast to
CO emission, H I emission is detected everywhere between the arms
at 5 kpc and 10 kpc. At
20 cm the contrast is about 2.5 and the
arm width is between the widths in CO and H I. This was also
noted by Berkhuijsen et al. (1993), who found that the width
at
20 cm corresponds to the width of the total gas arm, at a
resolution of
.
The molecular arms traced by CO are much narrower and thinner than the
H I arms, and the arm-interarm contrast is much higher. This
indicates that the molecular phase is short-lived compared to the
lifetime of the H I gas.
![]() |
Figure 5:
Two maps showing the integrated emission of the 12CO(2-1)
transition at a resolution of
![]() ![]() ![]() ![]() |
Open with DEXTER |
From a comparison of CO emission and visual extinction in a wide strip centred on the SW bright arm, Neininger et al. (1998) concluded that CO(1-0) line emission is a good tracer of the molecular gas in M 31, including the interarm region. The >5 times higher arm-interarm ratio in CO compared to H I thus implies that the molecular gas has almost vanished in the interarm regions. Yet molecular and atomic clouds have about the same velocity dispersion (see Sect. 5), have roughly the same response to the stellar potential and follow the same orbits with the same orbital velocity. This indicates that molecular clouds become mostly atomic when leaving the arms.
An accurate determination of the molecular cloud lifetimes would
require a density-wave model of M 31 that explains the observed CO
arm pattern. Such a model is outside the scope of the present paper,
although a crude upper limit to this lifetime can be estimated from
the CO and H I arm-interarm ratios. Consider a gas cloud
orbiting around M 31's centre and crossing a spiral arm. From
Fig. 4c we estimate that the total gas arm-interarm
ratio is 4-5. This means that, due to streaming motions, the clouds
stay 4-5 times longer inside the arms than they would if they followed
purely circular orbits at a constant velocity. For an arm width of
0.5 kpc, a pitch angle of
,
and an orbital velocity of
,
the time spent in an arm is
yr, which is an upper
limit to the age of the molecular clouds.
![]() |
Figure 6:
Contours of 12CO(1-0) emission from M 31 at
![]() ![]() ![]() |
Open with DEXTER |
Our observations of several selected regions on a finer grid yielded
maps with high signal-to-noise ratio in the (2-1) line as well as in
(1-0). Two example maps of the velocity-integrated CO (2-1) line
intensity (Fig. 5) cover some bright segments of spiral
arms, about 2-4 kpc long: one located near the major axis about 6 kpc
from
the centre (Fig. 5a), the other west of the main axis and
about 9 kpc from the centre (Fig. 5b). The ratio R21of the (2-1)-to-(1-0) line intensities is often regarded as a first
indicator of excitation conditions in the clouds emitting the CO lines.
To apply this test to our data, we smoothed the (2-1) map to the same
resolution (23
)
as the (1-0) map, selected those points where
both line intensities were above
,
and computed the line
ratio. In the region of Fig. 5a we found an average ratio
of
with values reaching up to 1.0 near D615 (
)
;
in the clouds of Fig. 5b we found an average of
with ratios up to 0.7 around D84 (
).
These values are compatible with simple standard assumptions about the
CO excitation: i.e., (i) both lines are optically thick; (ii) they
have the same excitation temperature
that equals the
gas kinetic temperature
;
and (iii) they sample the
same gas of uniform conditions. The expected line ratio under these
conditions is
R21 = 0.49 for
K,
R21 = 0.67 for T=6 K, and
R21 = 0.79 for T=10 K.
The line ratios we found in the spiral arms of M 31 are also similar to
values found widely over the Milky Way (e.g., Sakamoto et al.
1997), as well as to those of 0.7 to 0.8 seen in M 51
(Garcia-Burillo et al. 1993) and 0.60 to 0.85 in
NGC 891 (Garcia-Burillo et al. 1992). For these
reasons the CO excitation in the bright regions of our survey is clearly
different from that of gas that shows very low "subthermal'' values of
to
0.4, as found, e.g., by Allen & Lequeux
(1993) and Loinard et al. (1995) towards
some positions in M 31. Of course these papers mostly concern
the positions of small clouds or weak emission, whereas our values are
selected for positions with relatively strong lines and therefore
probably high ratios R21. We note, however, that Melchior et al.
(2000) reported an R21 ratio of 0.65 towards the
weak CO complex associated with D395A, located only 350 pc from the
centre of M 31. This value is quite similar to the ratios that we find.
A more complete study of CO excitation in M 31 would have to include
positions of weaker emission and also observations of optically thin CO
isotopes, but this is beyond the scope of our present paper.
![]() |
Figure 7:
Contours of 12CO(1-0) emission from M 31 at
![]() ![]() ![]() ![]() |
Open with DEXTER |
Figure 8 collects the observed brightness distributions
of CO, H I, total neutral gas, FIR at m and 20 cm radio
continuum. The
m emission mainly traces cold dust at
temperatures near 16 K (Haas et al. 1998).
The difference between the distributions of CO and
H I is striking. The H I arms are much smoother and wider
than the CO arms and weak H I emission is seen nearly everywhere
outside the spiral arms. We obtained the distribution of the total gas
column density,
,
after
smoothing the CO map to the angular resolution of
of the H I map (Brinks & Shane 1984). We used the Galactic conversion factor
given by Strong & Mattox
(1996), which was assumed to be constant across the
CO map. This conversion factor is supported by 1.2 mm observations of
the thermal emission from dust in two extended regions in the disk
(Zylka & Guélin, in prep.) and by virial mass estimates of several
molecular cloud complexes in M 31 (Muller 2003; Guélin et al. 2004). The H I column density was calculated
from the relation
,
valid for optically thin lines.
This relation may lead to a significant underestimate of
if
K (Braun & Walterbos 1992), i.e. at the crest of the arms. The
distribution of the total gas calculated this way is shown in
Fig. 8c.
![]() |
Figure 8:
Distributions of neutral gas, cold dust and radio continuum in M 31.
From top to bottom:
a) emission observed in the 12CO(1-0) line (this
paper);
b) emission observed in the H I line (Brinks & Shane 1984);
c) emission from the total neutral gas, N(H I)+2N(H2),
with
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
Comparing the various distributions in Fig. 8, we notice that
the nuclear area is only prominent at m and
20 cm
radio continuum, but all distributions in Fig. 8 show
the pronounced ring of emission at about 10 kpc from the centre, where
most of the star-formation regions are also located. The spiral arm
closer to the nucleus, clearly visible in CO and at
m, is
hardly seen in H I and radio continuum. This indicates that the
CO-to-H I brightness ratio increases towards smaller radii.
Only the H I and the
m distributions show extended
weak emission between the spiral arms. Thus the CO emission is
concentrated in regions of the denser clouds seen at
m and in
H I, located in the spiral arms, and does not trace weak
and extended interarm emission at
m, which is especially
visible at
.
In Sect. 4.2 we present the first results of a
correlation study between CO, H I and
m. Radial
distributions and the gas-to-dust ratios are shown in Sect. 4.1.
![]() |
Figure 9:
Radial profiles of I1-0 (full lines, left-hand scale) and
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
![]() |
Figure 10:
Radial variation of the molecular gas fraction
![]() ![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
The radio continuum emission at 20 cm (Beck et al.
1998) is also concentrated in the main spiral arms that form
the emission ring. Since at this wavelength most of the emission is
nonthermal, magnetic fields and cosmic rays are concentrated in this
ring. Using the CO data of Koper et al. (1991), Berkhuijsen
(1997) compared nonthermal emission with CO and H I
data at a resolution of
.
Although significant correlations
indeed exist at this resolution, detailed correlations with the new
data are required to enable an interpretation in terms of the coupling
between magnetic fields and gas. This will be the subject of a
forthcoming study.
![]() |
Figure 11:
Distribution of the molecular gas fraction
![]() ![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
![]() |
Figure 12:
Radial variations for the northern ( left panel) and the southern ( right
panel) halves of M 31 of
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
![]() |
Figure 13:
Radial variation of apparent
gas-to-dust ratios for the northern ( left panel) and the southern
halves ( right panel) of M 31 derived from the profiles
in Fig. 12. Thin full line -
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
Our CO profiles of the southern half show only half as much emission
as those in Fig. 10d of Loinard et al. (1999),
whereas the H I profiles are nearly identical. This may indicate a
difference in the adopted CO intensity scales, e.g. in the calibration
or in the corrections for beam efficiencies. In Sect. 4.3 we find
close agreement of our CO intensity scale with that
of the 1.2-m telescope at the Center for Astrophysics (Dame et al.
1993), to which the standard Galactic
is
calibrated. We also find that even after smoothing our map to the
effective resolution of
of the map of Loinard et al.,
the half-intensity width of the CO arms in their Figs. 5 and 6a is
about 1.5 times larger than in our map. We may attribute this
discrepancy to the much higher error beam of the FCRAO telescope
compared to that of the IRAM 30-m dish.
The fraction of molecular-to-total gas,
,
generally decreases with increasing
distance from the centre (see Figs. 10 and 11). The
highest values occur on the weak inner arm at
(2-3 kpc) and on the bright inner arm at
(5-6 kpc)
where the molecular fraction is nearly 0.5 in the north and about 0.25
in the south (see Fig. 10). On the bright ring at
,
the molecular fraction is about 0.2 both in the northern and
in the southern parts, although the northern part of the ring contains
about 1.5 times more neutral gas than the southern one. The averages
in circular rings underestimate the ratios on the arms.
Figures 4c and 11 yield a decrease from
0.6 on the arms at 5 kpc to 0.3-0.4 on the arms at 10 kpc.
The observed values of
and its radial decrease are typical of
nearby galaxies (Honma et al. 1995). The decrease in the
molecular fraction in Fig. 10 confirms the decrease in the
fraction of molecular gas mass along the major axis of M 31 reported by
Dame et al. (1993). Along the arms the molecular fraction
varies considerably (see Fig. 11). The highest fraction
detected is 0.96 for a cloud near the northern major axis at
,
which seems purely molecular.
![]() |
Figure 14:
Distribution of the apparent
dust-to-total gas ratio,
![]() ![]() ![]() ![]() |
Open with DEXTER |
We smoothed the distributions of CO, H I, total gas and 20 cm
radio continuum in Fig. 8 to the angular resolution of
of the
m map and compare their radial
distributions in Fig. 12. In all constituents the main ring at
is brighter in the north than in the south. The
pronounced molecular arm in the north at
is
invisible in radio continuum, possibly because of a lack of relativistic
electrons (Moss et al. 1998). When we disregard the central
region
,
the profiles of molecular gas and
m
emission from cold dust are the most alike.
Using the profiles in Fig. 12, we calculated the apparent
gas-to-dust ratios
,
and
as a function of radius presented in
Fig. 13. The ratio
is clearly
enhanced in the spiral arms at
and
,
especially in the north, whereas
continuously increases steeply from the centre outwards by nearly a
factor 20. As H I is the dominant gas component, the apparent
total gas-to-dust ratio
increases by
about a factor 20 between
and
.
However, the physical reality of this strong increase may be
questionable. First, the H I line opacity changes with radius;
second, the
conversion factor may vary with radius;
third, I175 may not reflect the dust column density if the dust
absorption cross section and, especially, the dust temperature vary
with radius.
The first two causes are unlikely. Braun & Walterbos
(1992) showed that the atomic gas temperature in
M 31 increases towards larger radii, while the 21-cm line opacity
decreases. This variation, however, is less than 20% and cannot
account for the strong gradient in
.
The behaviour of
with radius has been the subject of
several studies of the Milky Way and nearby galaxies (Wilson 1995; Sodroski et al. 1995; Strong et al.
2004). The value of
is found to increase
with increasing radius, perhaps in relation to decreasing metallicity.
Such an increase, if present at all in M 31, would only enhance the
radial variations in
and
.
A study of 50 bright cloud complexes at a resolution of at
least
,
which is high enough to alleviate rotation velocity
gradients, at radial distances between 5 and 12 kpc (Muller 2003; Guélin et al. 2004) yield
without any radial dependence.
Melchior et al. (2000) and Israel et al.
(1998) find similar values for two complexes at
and
.
Hence, except perhaps for the observation by
Allen & Lequeux (1993) of a dark cloud (D268,
)
with low CO luminosity and large
linewidth, which may not be in equilibrium, there is no evidence
of a radial variation of
in M 31.
It is more difficult to rule out a variation of the dust temperature
with radius. The thermal emission of cold dust at
m varies like
to the power 4-5, so even a mild
decrease in
outwards could cause a strong decrease in
I175. In fact, the apparent increase in
from 1 to 14 kpc could be explained by a decrease in the dust
temperature from
25 K to 15 K. Only extensive mapping of this
emission at longer wavelengths will allow us to discriminate between a
temperature effect and a dust-to-gas variation.
An increase in the apparent gas-to-dust ratio with increasing radius
in M 31 was found by several authors from comparisons of
and optical or UV extinction (Bajaja & Gergely 1977; Walterbos & Kennicutt 1988; Xu & Helou 1996;
Nedialkov et al. 2000; Savcheva & Tassev 2002). From a study of globular clusters, Savcheva
& Tassev found a similar increase in the apparent gas-to-dust radio
as we do. The distribution across M 31 of the reversed ratio, i.e. the
apparent dust-to-total gas ratio
,
is shown in
Fig. 14. The general decrease from the centre outwards is
clearly visible. The ratio
tends to be low on
the gas arms where the CO emission is strong, and even the many
roundish minima in the distribution of
coincide with CO clouds.
It is interesting to note that on the brightest spot of emission at
m, about
west of the nucleus, the value of
is normal for that radius. This bright emission
comes from one of the coldest dust clouds in M 31 analysed by
Schmidtobreick et al. (2000). Willaime et al.
(2001) showed that this cloud is located at a junction or
superposition of several thin dust lanes seen at
m (see
Fig. 7), which may explain its normal dust-to-gas ratio.
![]() |
Figure 15:
Examples of correlations between neutral gas and cold dust (16 K) at
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
Table 2:
Correlations between CO, H I and m.
In Sect. 3.2 we compared the CO and H I distributions across
the spiral arms. We now compare the general distribution of CO with
that of H I and each of them with the m FIR emission
using the CO and H I maps smoothed to
,
the resolution
of the FIR map.
At this resolution the CO and H I arms near
merge to a
broad emission ring, similar to the one seen at
m.
The comparison was restricted to radii
(=13.6 kpc) and intensities above
the rms noise at
resolution. Therefore, the noise in the CO map determined
the selection of points in correlations involving CO intensities;
likewise, the noise in the H I map determined the selection
of points in correlations involving intensities of H I or
total gas. In spite of this difference, the selected pixels are largely
the same; most pixels are located on the broad emission ring.
To obtain sets of independent data points, i.e.
a beam area overlap of
5%, only pixels spaced by
the beamwidth were considered. The resulting
correlations between CO, H I and the FIR are listed in
Table 2, and examples of correlation plots are shown in
Fig. 15.
In view of the distinct differences between the CO and H I distributions, it is not surprising that their velocity-integrated
intensities are not well correlated (see Figs. 15a and 15b). In
the panel showing the full radial range,
,
two
maxima occur that correspond to the inner bright CO arm and the main
emission ring, respectively. In the inner arm the molecular fraction is
much larger than in the main ring, and the decrease of this fraction
outwards obviously contributes to the large spread in the points.
Therefore we also analysed the inner (
)
and outer
(
)
radial ranges separately. In the inner part
I1-0 and
are not correlated (the correlation
coefficient
,
see Table 2), but for the outer
range some correlation is visible (Fig. 15b), with
.
Note that the correlation follows a power law with
exponent
.
An exponent of 2 is expected if the formation
and destruction rates of H2 are balanced (Reach &
Boulanger 1998). This occurs in the denser
H I phase at temperatures near 80 K, which are also found in
M 31 (Braun & Walterbos 1992). For a better
check of this possibility in M 31, this exponent could be determined in
narrow radial ranges at the original resolution of
of the H I map, which may yield better correlations
than obtained here.
The middle panels of Fig. 15 show that in the interval
both
and I1-0 are well
correlated with the dust emission at
m. With correlation
coefficients of
and
,
respectively, both correlations are highly significant
(Table 2). The relationship between
and
I175 is nearly linear, whereas that between I1-0 and I175follows a power law with exponent
(see
Table 2). This difference reflects the power-law
dependence of I1-0 on
in Fig. 15b. The
two branches visible at
in Fig. 15c are
caused by the strong radial gradient in the apparent
atomic gas-to-dust ratio
plotted in Fig. 13: points in the
upper branch are at larger radius than points in the lower one. This
branching does not occur in the correlation between I1-0 and
I175 (Fig. 15d), because
increases much less with R than
.
Correlations between the total gas column density and the m
emission for both the inner and the outer part of M 31 are
better than those between each of the gas components alone and I175(see Table 2 and Figs. 15e and f). The
correlation for the outer part is very good (
),
much better than for the inner part (
). Because in
the inner part
is dominated by the molecular gas, but
in the outer part by the atomic gas, the power-law exponent in the inner
part (
)
is larger than in the outer part (
).
The latter exponent is consistent with a linear correlation between
and I175. Linear correlations between gas and
dust in M 31 were also found by other authors. Pagani et al.
(1999) found a linear dependence of the intensity in the
ISOCAM LW2 filter (
m) on
in a field
centred on the southern major axis near
.
In a spiral-arm
region centred near
in the SW quadrant,
Neininger et al. (1998) obtained a good linear
correlation between
and the apparent red opacity.
In the Milky Way, gas and dust were found to correlate in detail (Bohlin
et al. 1978; Boulanger & Pérault 1988; Boulanger et al. 1996).
![]() |
Figure 16:
Position-velocity diagram along the major axis obtained by adding the
intensities along lines parallel to the Y-axis as function of position
on the X-axis and velocity. The grey scale represents the intensities
starting at contour level 0.5 K. The dotted line indicates the system
velocity of
![]() |
Open with DEXTER |
One may ask why the m dust emission follows
more closely than
alone, since the gas is mainly
atomic (see Table 2; see also Buat et al. 1989).
There are several reasons for this. First, emission from dust in
H I clouds, as well as from dust in H2 clouds, contributes to
the
m emission. Second, according to dust
models (see e.g. Mezger et al. 1990), the dust absorption
coefficient and the dust emissivity per H-atom are two (or more) times
larger in the dense molecular clouds than in the diffuse atomic clouds.
Hence, for equal dust column densities and temperatures, the emission
from dust in CO clouds exceeds that from dust in H I clouds as
soon as
.
The data presented in Sects. 4.1 and 4.2 show that a radial gradient in
the apparent gas-to-dust ratio exists that influences the
correlations between the gas components and the dust.
Corrections for H I-line opacity, better knowledge of the
conversion factor
and, probably most important, a
determination of the dust temperature as a function of radius are
needed to understand whether this radial gradient reflects a real
increase in the dust-to-gas ratio in M 31. In the meantime
comparisons of gas and dust on small scales in narrow radial ranges
would be of interest.
The CO intensity integrated over the area of M 31 gives an estimate of
the total molecular mass if one assumes a conversion factor between
I1-0 and
.
We integrated the CO intensities in
Fig. 1a out to a radius of
in the plane of
M 31, which corresponds to R=18 kpc. With the same conversion factor
as used above we find a molecular mass of
.
Along the major axis our map
only extends to
,
so we missed the emission near the
major axis between
and
detected by Koper et al. (1991). From Figs. 3b and 10c in Dame et al.
(1993) we estimate that this contributes about 5% of the
total. After correcting for these 5%, we get
within a radius R=18 kpc. This is in excellent
agreement with the value obtained by Dame et al. (1993), which
is
after scaling it to
the distance of 780 kpc adopted here. Given that Dame et al. used the
same value for
as we do, we can also conclude that their
CO intensity scale of radiation temperature,
,
closely
agrees with our scale of main beam brightness temperature,
.
To obtain the atomic gas mass within R=18 kpc, we integrated the
Westerbork H I map of Brinks & Shane (1984)
giving
in the optically thin case.
Thus for R<18 kpc, the H2 mass is 14% of the H I mass and
12% of the neutral gas mass. The total H I mass of M 31 at the
distance of 780 kpc is
(Cram et al.
1980, corrected by Dame et al. 1993), which gives a
total neutral gas mass in M 31 of
and a molecular mass fraction
.
The bulk of the dust in M 31 is cold dust at a temperature of
K. Haas et al. (1998) estimated a total
dust mass of
at D =
780 kpc. However, Schmidtobreick et al. (2000) found
using the same data but
a different method of calculation. Hence the apparent gas-to-dust
mass ratio in M 31 is
.
If the
H I masses are corrected for opacity of the H I, they
increase by 19% (Braun & Walterbos 1992) and the
ratio becomes
-488. For the Milky
Way, Sodroski et al. (1994) obtained a gas-to-dust mass
ratio for the entire Galaxy of
,
similar to the lower estimates for M 31.
Figure 1b shows the first moment of the CO spectral
data cube, i.e. the intensity-weighted mean CO velocity as a function
of position. This figure can be compared with Fig. 15a of Brinks
& Burton (1986 - hereafter B&B), which shows the
first moment of the H I data cube. In contrast to H I, CO
emission becomes very weak at galactocentric radii R> 18 kpc,
where the H I disk starts to flare and to warp (scaled to
the distance of M 31 of 0.78 Mpc). The velocity field in
Fig. 1b is therefore little affected by foreground and
background emission from the warp and is more appropriate to deriving
the velocity field in the disk than the velocity field of H I.
Moreover, as M 31 is at a galactic latitude of
,
contamination by foreground Milky Way clouds at the velocities of
interest here (
)
is unlikely. The average deviation between
the H I and CO mean velocities is only
,
which
may imply that the bulk of the H I emission in the region
covered by our CO map arises from M 31's disk, although the different
velocity resolutions (
for H I and
for CO) may
contribute to the differences.
The distribution of the mean CO velocities is dominated by the
rotation around the dynamical centre of the galaxy that has a
velocity
.
Only one small cloud near the centre
of the galaxy and a large cloud complex near the northwestern
minor axis show strong deviations from circular rotation. The small
cloud is located at
.
Its
non-circular orbit may be caused by the central bar (Berman 2001; Berman & Loinard 2002).
The cloud complex near the minor axis (
),
which appears particularly bright at
m, shows complex line
profiles that may be related to the complex filamentary structure
visible in Fig. 7. Alternatively, they may trace streaming
motions (Muller et al., in prep.).
The width of the observed CO-line profiles varies strongly from position
to position. The average linewidth (second moment of the CO data cube)
computed over the whole disk is
;
it is higher in the arms
(
)
and lower at the arm edges and in the interarm regions
(
). The average CO linewidth agrees with the
average linewidth of the H I emission of
(at
resolution) arising from the disk (Unwin 1983).
Some of the line profiles are very broad (40-65
,
e.g.
Fig. 2c) and show multiple velocity components. This is
particularly the case for the cloud complexes associated with the dark
clouds D 47 and D 39 (
and
,
see Figs. 2g and 2i), which show
complex profiles with total widths of 50 and
,
respectively. We
note that in the case of D 47 and D 39 the multiple profiles arise in
the vicinity of bright H II regions, but this possible
connection is not a general feature.
Other line profiles appear to be very narrow (
;
see
Fig. 2b). The profiles associated with D 153 (
,
see Fig. 2f) have
a particularly high peak intensity combined with a narrow
line width of only about
.
A detailed analysis of the cloud-to-cloud velocity dispersions inside the molecular cloud complexes and of the interplay between density-wave driven motions and peculiar motions linked to H II regions or supernova remnants, will be the subject of a forthcoming paper (Muller et al., in prep.).
The position-velocity diagram of the CO (1-0) intensity integrated
along the minor axis is shown in Fig. 16. The dominant
features directly reflect the general velocity field in the disk
(Fig. 1b) and the fact that most CO emission is
concentrated in spiral arms (Figs. 1a, 3a). Due to the rotation of the M 31 disk, CO emission
from the arms marked in Fig. 3a is seen in a broad band
of about
width, stretching from
to
.
Most of the
emission in this band is from regions near R=10 kpc. This
large-scale picture agrees, of course, with that in the low-resolution
survey by Dame et al. (1993), who also find some emission
near
,
outside the area of our survey.
The CO spiral arms appear as loops within the range of this band
(compare Braun (1991) for H I); most clearly seen in
Fig. 16 is the loop at the most negative X corresponding to
the "dashed'' spiral arm in Fig. 3a. Emission from the
inner part of the "solid'' spiral arm in Fig. 3a is
clearly visible at
above the main
band of emission. Extended and relatively strong emission at velocities
around
and X about
to
is from clouds
located inside the "solid'' spiral arm in Fig. 3a, i.e.
closer to the major axis (compare Fig. 1b). While we
cannot reliably trace spiral arms from the CO emission in this inner
region, we note that these CO features agree in position and velocity
with the inner loops of the spiral arms identified by Braun (1991) from
H I surveys. In a similar fashion, most smaller and weaker
features at velocities that are different from those of the main CO
spiral arms in Figs. 1b and 3a are on
other loops of Braun's H I spiral arms.
The new 12CO(J = 1-0)-line survey of the Andromeda galaxy
presented here covers an area of about
,
which is
fully sampled with a velocity resolution of
and an angular
resolution of
,
the highest angular resolution to date
of a map of this extent.
At the adopted distance of 780 kpc the linear resolution is
in the plane of M 31. The On-the-Fly
method of observing made it possible to measure nearly 1.7 million
spectra (before gridding) in about 500 h of effective observing
time on M 31. The rms noise in the CO(1-0) line per 1 MHz channel
is 25 mK in the northern half of the map and 33 mK in the southern half.
The velocity-integrated distribution, I1-0, and the velocity field
are shown in Fig. 1. The molecular gas is concentrated in
narrow, filamentary arms between 4 and 12 kpc from the centre with a
maximum near 10 kpc. The inner arm at kpc is remarkably
bright. Only a few clouds, often forming bridge-like structures,
were detected between the arms above
rms noise of
typically
.
The region within 2 kpc from the centre is
almost free of molecular clouds that are brighter than the sensitivity
of this survey.
The thin CO arms define a two-armed spiral pattern that can be
described well by two logarithmic spirals with a constant pitch angle of
.
At a resolution of
,
the arm-interarm
contrast reaches a maximum of 20 in I1-0 compared to 4 in
H I for the western bright arms. The H I arms are much
wider than the molecular arms, and diffuse H I exists everywhere
between the arms and at large radii. Few molecular clouds are
visible outside a radius of 16 kpc.
The velocity field of the molecular gas is very similar to that of the disk component in H I. At some positions perturbed velocity profiles occur that are possibly caused by nearby H II regions.
Several selected regions were also observed in the 12CO(2-1) line.
At a resolution of
,
the line ratios are nearly constant with
mean values of
I2-1/I1-0 = 0.5-0.7 in the arms. These line
ratios are similar to those observed in other galaxies and show no
indication of subthermal excitation.
Averaged radial profiles of the velocity-integrated CO and H I
distributions show that, for a constant conversion factor of
,
the molecular fraction of the neutral gas is enhanced
on the spiral arms and decreases radially from about 0.6 on the inner
arms to about 0.3 on the arms at
kpc (see
Fig. 11). The molecular fraction also varies considerably
along the arms. Comparisons of the H I and CO profiles with the
averaged radial profile of the
m emission from
cold (16 K) dust revealed a strong, continuous increase of the
apparent atomic gas-to-dust ratio of nearly a factor 20 between the
centre and
kpc, whereas the apparent molecular
gas-to-dust ratio increases by about a factor of 4. The apparent
total gas-to-dust ratio also increases by about a factor of 20.
The strong apparent gradients in the molecular fraction and the
gas-to-dust ratios influence the cross-correlations between CO,
H I and FIR (m) intensities. In the radial range
(about 8-14 kpc), the best correlation exists
between total neutral gas and
m, followed by the correlations
between CO and
m and between H I and
m.
The relationships between H I and
m and between
the total gas and
m are close to linear, whereas that between
CO and
m is a power law with exponent 1.6 due to a possible
non-linearity between CO and H I. In the inner part of M 31,
(but outside the nuclear area), only total gas and
FIR (
m) are reasonably well correlated (see
Table 2).
The total molecular mass of M 31 within a radius of 18 kpc is
,
when using the above-mentioned value of
.
As the total H I mass (without correction for
opacity) is
,
the total mass of the
neutral gas is
.
The total mass of the
cold dust is
,
so the total
gas-to-dust mass ratio is 410-137. The lower value agrees with the
Galactic one.
The wealth of information contained in our new 12CO(1-0) survey of M 31 enables a number of new investigations into the physical relationships between molecular gas, atomic hydrogen gas, cold and warm dust, ionized gas, relativistic electrons and magnetic fields. Such studies will be the subject of forthcoming papers.
The data shown in Fig. 1 can be obtained from M. Guélin (guelin@iram.fr).
Acknowledgements
C.N. acknowledges support from the Studienstiftung des Deutschen Volkes. We thank W. Brunswig, C. Kramer, G. Paubert, J. Schraml and A. Sievers for their important contributions to the On-the-Fly observing mode at the IRAM 30-m telescope and P. Hoernes for observations and software adjustment. S. Muller kindly communicated to us some results of his Ph.D. Thesis prior to publication. We thank M.-C. Willaime for sending us an updated version of the ISOCAM LW6 map that we used in Fig. 7. The extensive comments and suggestions of an anonymous referee were a great help in improving the manuscript. This work was started when M.G. was visiting the MPIfR as a von Humboldt Fellow.