Molecular gas in the inner 0.7kpc-radius ring of M31

Abstract

The study of the gas kinematic in the central 1.5kpc x 1.5kpc region of M31
has revealed several surprises. The starting point of this investigation was
the detection at the IRAM-30m telescope of molecular gas with very large line
splittings up to 260km/s within the beam (40 pc). In this region, which is
known for its low gas content, we also detect an ionised gas outflow in the
circumnuclear region (within 75pc from the centre) extending to the whole area
in X-ray. Relying on atomic, ionised, and molecular gas, we account for most
observables with a scenario that assumes that a few hundreds Myr ago, M31
underwent a frontal collision with M32, which triggered some star-formation
activity in the centre, and this collision explains the special configuration
of M31 with two rings observed at 0.7kpc and 10kpc. The inner disc (whose
rotation is detected in HI and ionised gas ([NII])) has thus been tilted
(inclination: 43deg, PA: 70deg) with respect to the main disc (inclination:
77deg, PA: 35deg). One of the CO velocity components is compatible with this
inner disc, while the second one comes from a tilted ring-like material with
40deg inclination and PA=-35deg. The relic star formation estimated by previous
works to have occurred more than 100Myr ago could have been triggered by the
collision and could be linked to the outflow detected in the ionised gas. Last,
we demonstrate that the amplitude of the line splittings detected in CO centred
on the systemic velocity with a relatively high spatial resolution (40pc)
cannot be accounted for by a possible weak bar that is roughly aligned along
the minor axis. Although M31 has a triaxial bulge, there are no bar indicators
in the gas component (photometry, no strong skewness of the isovelocities,
etc.).

Full text

Page 1

arXiv:1103.3392v1 [astro-ph.CO] 17 Mar 2011
Astronomy & Astrophysics manuscript no. amelchior
March 18, 2011
c ? ESO 2011
Molecular gas in the inner 0.7kpc-radius ring of M31
A.-L. Melchior1,2and F. Combes1
1LERMA, Observatoire de Paris, LERMA, UMR8112, 61, avenue de l’Observatoire, Paris, F-75 014, France
e-mail: A.L.Melchior@obspm.fr,Francoise.Combes@obspm.fr
2Universit´ e Pierre et Marie Curie-Paris 6, 4, Place Jussieu, F-75252 Paris Cedex 05, France
ABSTRACT
The gaseous disc of M31, traced by atomic, ionised and molecular gas together with dust emission, is mainly characterised by a large
ring of 10kpc radius, and an inner ring of 0.7kpc radius, in addition to an underlying spiral structure. This inner ring is conspicuous
in dust emission and extinction, and co-exists with an inner disc detected in ionised and atomic gas. Up to now, there was very
little neutral gas detected in these inner parts, and the dynamics of this ring was unknown. We have observed with the IRAM-30m
telescope the CO lines in several regions of the inner kpc of M31. While the inner ring is offset by 0.5kpc from the nucleus, the
observed positions, selected on the extinction map, are located on the North-Western side of the inner disc, corresponding to the near
side. We detect CO in most of these regions, with a complex kinematics. In several points, multiple velocity components are detected
with line splittings up to 260 kms−1. One of the component, the broader one, corresponds to the rotation pattern of the galaxy, while
the narrower is in counter-rotation. In addition, the velocities measured in the molecular gas do not match the one measured with
ionised gas, especially along the minor axis, where one would expect the gas to have the systemic velocity, for a disc in circular
rotation. The North-East part of the ring is weak but detected in HI, while a disc speed pattern is tentatively detected after removal
of the main disc contribution. We have studied several scenarios to account for the abnormal kinematics, in terms of bars, warps or
tilted rings. Our best model which reproduces the main observed features assumes that the peculiar component comes from a tilted
ring-like material with 40deg inclination and PA = −35deg. This ring corresponds to that discovered in the dust emission by Block et
al. (2006). It could have formed in the annular wave provoked by the head-on collision, when the companion M32 crossed the plane
of M31. The reasons why the ring appears in counter-rotation are discussed.
Key words. (Galaxies:) Local Group, Galaxies: spiral, (Galaxies:) bulges, Methods: data analysis, Hydrodynamics, ISM: molecules
1. Introduction
The merging processes that probably took place during the as-
sembly of the galaxies of the Local Group is now actively stud-
ied (e.g. Klimentowski et al., 2010). While the Milky Way has
not suffered any major merger for several billion years and
presentonlyaseriesofdwarfgalaxiesstellarstreams(seeHelmi,
2008), the other main spiral of the Local Group (M31) has had
a more perturbed history. Giant stellar loops and tidal streams
are observed in the surroundings, extending up to the neigh-
bour M33, which is suspected to have interacted with M31
(e.g. McConnachie et al., 2009). Many coherent structures ob-
served around M31 are interpreted as the disruption of small
dwarf galaxies, which are numerous in the vicinity of M31 (e.g.
Ibata et al., 2004).
In parallel, it has been known for a long time that the
M31 galaxy exhibits an unusual morphology (e.g. Helfer et al.,
2003). The young stellar population tracers in the disc are
all concentrated in the so-called 10kpc ring (Chemin et al.,
2009; Nieten et al., 2006), which is highly contrasted and super-
posed to only a few spiral structures (e.g. Nieten et al., 2006).
One striking feature is the little amount of gas present in the
central region: neither Braun et al. (2009) nor Chemin et al.
(2009) detect any significant HI component, while a sim-
ilar depletion is observed in CO-emission by Nieten et al.
(2006). It is usually described as a quiescent galaxy with
little star formation, with an ultra-weak nuclear activity
Send offprint requests to: A.L.Melchior@obspm.fr
(del Burgo et al., 2000). However, ionised gas is detected in the
central field (e.g. Ciardullo et al., 1988; Boulesteix et al., 1987;
Bogd´ an & Gilfanov, 2008; Liu et al., 2010), usually interpreted
intermofshocks,andMelchior et al.(2000)hasdetectedmolec-
ular gas inside the inner ring.
In this empty central region, interest focused on the super-
massive black hole (Bacon et al., 1994) and the circumnuclear
region (inner few hundred parsecs). Barmby et al. (2006) ob-
served the whole galaxy in the mid-infrared with the Infrared
Array Camera on board the Spitzer Space Telescope revealing
spectacular dust rings and spiral arms. Block et al. (2006) stress
the presence of an inner ring with projected diameters 1.5kpc
by 1kpc and propose a completely new interpretation for the
morphology on this galaxy. Both rings at 1kpc and 10kpc are
off-centred. Their respective radii do not correspond to what
is expected from resonant rings in a barred spiral galaxy. The
most likely scenario to form such rings is a head-on collision of
the Cartwheel type. Unlike the Cartwheel, where the compan-
ion is about 1/3rd of the mass of the target (major merger), in
Andromeda,the collision can be called a minor merger,and pro-
duces much less contrasted rings in the main disc. Block et al.
(2006) propose that the collision partner was M32, with about
1/10thof the mass (dark matter included)at the beginning.After
stripping experienced in the collision, the M32 mass is now
1/23 that of the main target M31. The M32 plunging head-on
through the centre of M31 has triggered the propagation of an
annular wave, which is now identified with the 10kpc ring,
and a second wave propagates more slowly afterwards (see e.g.
Appleton & Struck-Marcell, 1996), and would correspondto the

Page 2

2 A.-L. Melchior and F. Combes: Molecular gas in the inner 0.7kpc-radius ring of M31
inner ring. In addition, the inner ring has formed in a tilted and
warped disc, which accounts for its almost face-on appearance,
in contrast with the inclined main disc of M31. This scenario
explains why the cold gas has been expelled from the central re-
gionas well as the presenceof shocks and hot gas. In this article,
we focus on the gas content of the inner ring and inside.
We presentCO(1-0)andCO(2-1)observationsobtainedwith
IRAM-30m in different positions located in the North-Western
part of the inner ring, as well as a few positions inside the ring.
While, the CO intensities are correlated in first instance with the
ABextinction, the velocity distributions are quite unexpected.In
the inner ring, the velocities are spread between -450 kms−1and
-150 kms−1for a few positions while the systemic velocity (-
310 kms−1) is expected in this area along the minor axis. In or-
der to get a better understandingof these velocities, we compare
our measurements with the velocities available for the ionised
andHIgasandtheydonotreallymatcheachother.Whileneither
the main HI warp of the disc nor a bar along the line of sight can
explainthewideamplitudevelocitysplittingsobservedalongthe
minor axis, the main configuration compatible with the data dis-
cussed here is to have a ring inclined with respect to the nuclear
disc, and corresponding in projection to the inner dust ring, de-
tected in extinction and in infra-red emission. This ring could
beattributedtoahead-oncollisionwithM32(Block et al.,2006)
or to some accretion of gas from a M31-M33 gas bridge or tidal
loop detected in HI (Braun & Thilker, 2004).
In Section 2, we describe the observations performed at
IRAM-30m. In Section 3, we describe the reduction of the C0
data. In Section 4, we present an analysis of the CO data and
compare the molecular gas with the ionised gas, together with
information from other wavelengths. In Section 5, we explore
various scenarios and propose one modelling explaining the ob-
servations. In Section 6, we discuss our results and conclude.
2. Observations
The observations were carried out on 1999 June 13-15 and 2000
July 14-17 with the IRAM 30-m telescope. Most of the observ-
ing was made in the symmetrical wobbler switching mode, in
which the secondary mirror nutates up to a maximum limit of
±240 arcsec in azimuth. The beam throw was determined as a
function of the hour angle in such a way that OFF positions lie
in extinction-free regions as described in Melchior et al. (2000).
It neverthelessoccurredin a few positionsthat some signal is de-
tected in the OFF positions. Near transit, we had to use position-
switching mode, taking an extinction-free OFF position located
at a given position from the nucleus as indicated in the last col-
umn of Table 1. During the second epoch of observations, in
order to sample in CO(2-1) the CO(1-0) beam and to compute
the CO(2-1)/CO(1-0) line ratio, little maps of 7 points with a
spacing of 5” have been done for 4 positions (A, C, D and G).
Pointing and focus calibration were regularly checked. In total,
10 ON positions in the inner disc of M31 have thus been ob-
served, as summarised in Table 1 and presented in Figure 1.
We used four receivers simultaneously, two for12CO(1-0) at
115GHz and two for12CO(2-1) at 230GHz. At 115GHz, each
receiver was connected to two autocorrelator sub-bands (shifted
by 40MHz fromeach other)and each sub-bandconsisted of 225
channelsseparated by1.25MHz. At 230GHz, each receiverwas
connected to a filter-bank consisting of 512 channels of 1MHz
width.
3. Data reduction
TheCLASS1packagehasbeenusedforthedatareduction.After
checking the quality of each single spectra, the data have been
averaged with inverse variance weights. A first-order baseline
has been fitted to the resulting spectrum and subtracted. For a
few spectra, a higher-order polynomial has been carefully sub-
tracted.Finally,thespectrahavebeensmoothedtoavelocityres-
olution of 3.2 kms−1(resp. 2.6 kms−1) in CO(1-0) (resp. CO(2-
1)).
The maps are presented in Figure 2 and illustrate the ob-
serving procedure. The spectra are located arbitrarily close to
the beam position they correspond to. Spatial variations in the
spectra are in relatively good agreement with the extinction dis-
tribution. For M31G, the CO intensity is clearly stronger in the
area where the extinction is stronger. For M31A and M31D, we
cannot exclude that some pointing errors and/or a clumpy distri-
bution of the underlying gas to explain the spatial variations of
the CO distribution.
Theresults of thefitting are presentedin Table 2. A Gaussian
function is fitted to each line to determine its area, central veloc-
ity V0, width σ and peak temperature Tpeak. The baseline RMS
is provided for each line. We provide main-beam temperatures
(unless said so) throughout this paper with Bef f = 64.2±3 and
Fef f = 91±2 (resp. Bef f = 42.±3 and Fef f = 86±2) at 115GHz
(resp. 230GHz). For the positions for which maps have been ob-
served, the spectra are convolvedin order to obtain an 24′′beam
(
line ratio.
The spectra are displayed in Figure 3. In the final spectra,
the signal present in the OFF positions is mainly seen for the
position M31G. It is at the systemic velocity and corresponds
probably to the gas in the main disc along the minor axis on the
Northofthe M31Gposition(as the OFF positionsare symmetric
in azimuth with respect to the ON position).
√112+ 212) forbothlines, enablinga directcomputationofthe
4. Analysis
4.1. Characteristics of the CO emission and gas-dust
connection
The CO gas, that has been searched for and discussed in this
paper, is located mainly around the minor axis of the main disc
of M31. The detected molecular gas complexes correspond to
the strongest ABextinction complexes that we observed. For the
positionswhereno(M31GB)extinctionis present,we donotde-
tect any signal, while no line has been detected in the centre nei-
ther. For M31Ewhere the extinctionis weak, we have a tentative
detection at 4σ. The extinction has been computed assuming a
fractionofforegroundlightof x = 0 (seeMelchior et al. (2000)),
which is probably correct for the North-Western part, where we
observed, as an inclination of 45deg is usually assumed for the
nuclear spiral (Ciardullo et al., 1988). We have also computed
the NUV and FUV extinction maps from archive GALEX data
with the same method and found the same structures. In addi-
tion, we have subtracted from the original GALEX images the
bulge emission thus modelled, and do not find any trace of UV
emission in the inner ring.
The position M31F, which exhibits a weak CO(1-0) signal
with no detection in CO(2-1), corresponds to a position with
small ABextinction. The other positions are detected with high
signal to noise ratios at least in CO(1-0). We have computed the
1Continuum
http://www.iram.fr/IRAMFR/GILDAS
andLineAnalysisSingle-disSoftware,

Page 3

A.-L. Melchior and F. Combes: Molecular gas in the inner 0.7kpc-radius ring of M313
Table 1. Summary of observations
Position PAOffsetR
(arcsec)
RA
(J2000)
DEC
(J2000)
dateTexp
(min)
mapOff
sig-
nal
Position
switch
(arsec)
M31A161.1 -26,76 80.300:42:42.141:17:21.113/06/99
14-
17/07/00
630
1160 yes
nono
no
122,321
385,57;
274,23;
379,34
385,57
122,321
404,-41
122,321
122,321
122,321
419,-7
122,321
540,-23
543,-43
508,-93
M31C
M31D
–
161.3
0,0
-51,151
0
159.4
00:42:44.4
00:42:39.9
41:16:09.2
41:18:39.6
2000
14/06/99
16/07/00
14/06/99
15/06/99
15/06/99
15/07/00
15/06/99
17/07/00
17/07/00
17/07/00
256
180
112
360
442
180
175
78
42
22
90
yes
no
yes
no
no
no
yes
no
no
no
no
yes
yes
no
no
no
yes
yes
no
no
no
no
M31E
M31F
M31G
47.1
125.6
121.8
72,67
-134,96
-171, 106
98.4
164.8
201.2
00:42:50.8
00:42:32.5
00:42:29.2
41:17:15.0
41:17:44.0
41:17:54.0
M31GB
M31I
M31J
M31K
170.7
123.8
127.8
143.1
-22,134
-172,115
-175,136
-139,185
135.8
206.9
221.6
231.4
00:42:42.4
00:42:29.1
00:42:28.9
00:42:32.0
41:18:22.0
41:18:3.6
41:18:24.3
41:19:13.7
Table 2. Characteristics of the CO lines. CO(2-1) (resp. CO(2-1)) spectra are reduced to a 2.6 kms−1(resp. 3.2 kms−1) resolution
Position
12CO
(beam size)
lineICO(K km s−1) =
?
TmbdV
V0(km s−1)
σ (km s−1)Tpeak
(mK)
baseline
rms
(mK)
NH2(cm−2)
ΣH2
(M⊙pc−2)
M31A 1-0 (24′′)
2-1 (24′′)
1-0 (21′′)
2-1 (11′′)
1-0 (24′′)
2-1 (24′′)
1-0 (24′′)
2-1 (24′′)
1-0 (21′′)
2-1 (11′′)
1-0 (21′′)
2-1 (11′′)
1-0 (21′′)
2-1 (11′′)
1-0 (24′′)
2-1 (24′′)
1-0 (21′′)
2-1 (11′′)
1-0 (21′′)
2-1 (11′′)
1-0 (21′′)
1-0 (21′′)
1-0 (21′′)
1-0 (21′′)
2-1 (11′′)
2-1 (11′′)
2-1 (11′′)
2-1 (11′′)
1-0 (21′′)
1-0 (21′′)
1-0 (21′′)
2-1 (11′′)
2-1 (11′′)
2-1 (11′′)
1-0 (21′′)
2-1 (11′′)
0.75±0.03
0.88±0.06
0.72±0.04
0.78±0.06
−
−
2.43±0.10
3.45±0.16
2.47±0.11
2.12±0.12
0.18±0.04
0.33±0.10
1.56±0.14
−
2.98±0.18
2.39±0.14
3.01±0.23
2.18±0.20
−
−
3.22±0.11
4.60±0.77
2.50±0.31
3.69±0.89
3.28±0.12
7.28±0.47
4.04±0.47
2.96±0.41
12.81±0.33
1.06±0.11
7.34±0.57
15.22±0.78
0.86±0.41
1.94±0.69
2.66±0.14
2.61±0.43
-154.9±0.6
-151.1±0.7
-153.6±0.7
-152.8±1.0
−
−
-74.3±0.4
-73.2±0.5
-74.9±0.5
-74.9±0.6
-132.2±1.9
-128.2±3.1
-360.6±3.3
−
-147.4±0.4
-146.7±0.5
-146.7±0.6
-146.6±0.9
−
−
-148.5±0.2
-411.9±1.0
-278.7±2.6
-376.4±11.1
-148.6±0.2
-409.5±1.3
-283.4±2.7
-353.3±5.7
-413.2±0.2
-145.5±0.6
-350.1±7.5
-412.7±0.9
-142.1±3.0
-268.6±7.1
-300.8±1.1
-292.5±4.9
28.7±1.3
22.9±1.9
28.7±1.7
24.9±2.6
−
−
23.7±1.2
19.7±1.1
23.6±1.4
21.1±1.4
15.8±3.1
15.3±6.9
68.0±6.7
−
13.6±1.0
15.4±1.1
15.3±1.4
18.5±1.9
−
−
9.3±0.4
34.1±3.1
46.3±5.4
88.5±16.5
7.7±0.4
40.0±3.5
43.5±5.0
46.6±4.5
33.3±0.8
11.9±1.5
178.2±13.2
35.1±2.0
9.0±5.2
35.4±10.8
41.7±2.9
58.5±11.0
24.8
35.7
23.7
29.4
−
−
96.4
164.6
98.0
93.8
10.5
19.4
20.6
−
205.9
145.2
185.2
110.6
−
−
324.3
126.8
50.7
39.2
400.9
171.0
87.7
59.6
361.4
83.8
38.8
407.0
88.5
51.4
59.8
42.0
2.1
4.7
2.4
5.5
3.7
7.8
7.2
15.5
8.7
10.6
4.1
13.1
6.4
16.3
18.0
16.1
21.2
20.4
7.7
14.7
12.5
12.5
12.5
12.5
29.0
29.0
29.0
29.0
12.2
12.2
12.2
54.5
54.5
54.5
7.5
24.1
1.73×1020
−
1.67×1020
−
<2.55×1021
−
5.58×1020
−
5.68×1020
−
4.25×1019
−
3.59×1020
−
6.86×1020
−
6.92×1020
−
<5.29×1021
−
7.41×1020
1.06×1021
5.75×1020
8.49×1020
−
−
−
−
2.95×1021
2.45×1020
1.69×1021
−
−
−
6.11×1020
−
2.94
2.83
M31C
<43.31
M31D9.49
9.66
M31Ea
0.72
M31F6.11
M31G11.66
11.77
M31GB
<89.95
M31I12.60
17.99
9.77
14.44
M31J50.08
4.16
28.70
M31K 10.38
aTentative detection.

Page 4

4 A.-L. Melchior and F. Combes: Molecular gas in the inner 0.7kpc-radius ring of M31
Fig.1. Right: Positions (J2000) observed in CO in the central part of M31. The circles indicate 1-mm (green) and 3-mm (red)
beams of the 10 positions observed superimposed on the ABextinction map obtained in Melchior et al. (2000). As discussed in
this paper, the extinction with the central 2 arcsec is not well defined on this map. The (blue) cross indicates the centre of M31
(αJ2000=00h42m44s.371, δJ2000=41◦16′08′′.34, Crane et al. (1992).) Left: Velocity field measured in CO for the whole galaxy by
Nieten et al. (2006). A small rectangle in the central part indicates the field studied in this paper. This global view reminds the large
scale configuration of this galaxy: a very inclined disc at 77deg with a position angle of 35deg.
Table 3. ABextinction derived from the map based on Ciardullo et al. (1988) data as obtained in Melchior et al. (2000). For each
configuration, E(B-V) has then been computed assuming RVin the range [2.1,3]. These values have been computed assuming a
fraction of foreground light x = 0. The ICO/E(B − V) ratio is then provided for the strongest lines. For M31J, we also provide the
ratio corresponding to the ring velocity.
PositionABat 24′′
(mag beam−1)
ABat 21′′(11′′)
(mag beam−1)
E(B-V) at 24′′
(mag beam−1)
E(B-V) at 21′′
(mag beam−1)
E(B-V) at 11′′
(mag beam−1)
ICO/E(B-V) at 24′′
(K km s−1mag−1)
ICO/E(B-V) at 21′′
(K km s−1mag−1)
M31A
M31D
M31F
M31G
M31I
M31J
M31K
0.095
0.067
0.074
0.182
0.210
0.203
0.138
0.100 (0.166)
0.076 (0.111)
0.081 (0.123)
0.193 (0.256)
0.219 (0.245)
0.208 (0.241)
0.141 (0.159)
0.027±0.003
0.019±0.002
0.021±0.003
0.052±0.007
0.060±0.008
0.058±0.007
0.040±0.005
0.029±0.004
0.022±0.003
0.023±0.003
0.055±0.007
0.063±0.008
0.060±0.008
0.040±0.005
0.048±0.006
0.032±0.004
0.035±0.004
0.073±0.009
0.070±0.009
0.069±0.009
0.046±0.006
28±5
129±22
59±11
26± 5
116±20
69±15
56±11
52± 8
219±33(18± 4)
67±12
ABextinction values for each position convolved with the dif-
ferent beams, as provided in Table 3, and there is not a one-
to-one correspondence with the intensity of the detected CO.
As also seen in Figure 1, M31D has a much stronger signal in
CO than M31A while its ABextinction is weaker. One plau-
sible explanation could be that the dust clumps corresponding
to M31D does not lie in the same plane as M31A: if the fore-
ground light (x) is larger than 0, we have underestimated the
extinction. Assuming the ICO/ABratio measured for M31A ap-
plies to M31D, one would expect a peak intensity Areal
for M31D while it is measured Ameas.
B
= 1.56
B
= 0.26.This configuration
corresponds to a fraction of light in front of the dust x = 0.72,
meaning that M31D lies on the back side of the bulge. It is also
probablethat the gas is very clumpy and that the non-linearityof
extinction biases somehow our ABestimate.
Our detections are concentrated in an area2of 415pc ×
570pc (in projection), located in the North Western part of the
M31 within 3.8arcmin from the centre, correspondingto 880pc
(resp. 1.2kpc if deprojected). While the area explored is quite
localised, the detected velocities span from -73 to -413 kms−1,
2Like in Melchior et al. (2000), we assume a distance of M31 of
780kpc i.e. 1arcsec=3.8pc

Page 5

A.-L. Melchior and F. Combes: Molecular gas in the inner 0.7kpc-radius ring of M315
Fig.2. Small maps performed at in CO(2-1) (left panels) and CO(1-0) (right panels) for the positions A, D and G. The green (resp.
red) circles correspond to the FWHM of the CO(2-1) (resp. CO(1-0)) beams at the various positions observed. The corresponding
offsets are: (0,0), (+5,0), (+2,+4) (-2,+4), (-5,0), (-2,-4), (+2, -4). The spectra are superimposed on the corresponding ABmaps.
They are positioned arbitrarily close to the beam corresponding to the observations.
while velocities along the minor axis of a rotating disc are ex-
pected to be at the systemic velocity. M31K is the sole position
exhibiting this expected velocity, as well as a component for
M31G, M31I and M31J, and possibly M31D. Figure 4 shows
a superimposition of the spectra obtained for M31G, M31I and
M31J (and reduced to a 24′′beam): the various velocity pat-
terns appear in each spectra with different relative intensities.
Beside the velocity line detected at -145 kms−1, a broadband
signal is detected with velocities between -450 and -250 kms−1.
As discussed below, such a range of velocities cannot be ex-

Page 6

6A.-L. Melchior and F. Combes: Molecular gas in the inner 0.7kpc-radius ring of M31
Fig.3. Spectra with a detected signal. M31A, M31D and M31G (resp. M31E, M31F, M31I and M31J) spectra are reduced to a 24′′
beam (with the original beam) in CO(2-1) and CO(1-0). The y-axis provides the antenna temperature in K.

Page 7

A.-L. Melchior and F. Combes: Molecular gas in the inner 0.7kpc-radius ring of M317
Fig.4. Superimposition of the G, I and J spectra at the nominal spatial resolution, namely 21′′for CO(1-0) and 11′′for CO(2-1).
Table 4. Lineratios ofthecomplexesreducedto a24′′beam.We
provide both the ratios of the line intensities r12and the ratios of
the peak temperatures R.
Complexesr12
R
M31A
M31D
M31G
1.17±0.40
1.43±0.45
0.80±0.30
0.19
1.44±0.3
1.71±0.36
0.71±0.14
0.1
0.21
0.12
0.13
0.05
plained by a single disc inclined at 45deg in regular rotation
and are the signature of a peculiar structure. The line widths of
the detected lines are quite different from one location to an-
other suggestinga spatial extensionof the emitting areas. We as-
sume a Galactic XCO= NH2/ICO= 2.3×1020cm−2(K kms−1)−1
following Strong et al. (1988). For the positions, where we did
not detect any signal, we do provide a 3σ upper limit based on
the dispersion (rms) computed on the baseline. Some CO(2-1)
measurements (M31I,M31J, M31K) are at the limit of detection
and are providedfor a crude comparisonwith CO(1-0) measure-
ments.
In the centre (< 2′′), our ABmap does not measure the ex-
tinction (due to the method based on ellipse fitting). The map of
8µm emission (after subtraction of the scaled stellar continuum
at 3.6µm) also displays a defect in the central part, so it is diffi-
culttobeconclusiveabouttheextinctionintheinnerarcseconds.
However, Garcia et al. (2000), relying on simple modelling of
Chandra data have estimated AV = 1.5 ± 0.6 in the central 3-
arcsec region. Our millimetre observations exclude the presence
of CO(1-0) (resp. CO(2-1)) at the 1σ level of 3.7mK (resp.
7.8mK). On the basis of this upper limit, the gas present within
80pc from M31’s centre does not exceed 43M⊙/pc−2(3σ), as-
suminga conservativevalue for XCOtypical of the Galactic disc.
This is an extremely small amount compared to the gas mass
computed in the central region of the Milky Way (Oka et al.,
1998).
We use the 3 positions with observations reduced to a 24′′
beam, to compute the CO(2-1) to CO(1-0) line ratio. As dis-
played in Table 4, we compute both the ratio of the integrated
intensitiesandtheratioofthepeaktemperatures,whicharecom-
patible within error bars. M31G exhibits a line ratio compara-
ble with the mean value found for the Milky Way spiral arms
(Sakamoto et al., 1997).
M31DandM31AexhibitsCOlineratioslargerthan1.Given
theNH2columndensitiescomputed(respectively5.6×1020cm−2
and 1.7 × 1020cm−2), the gas is probably not optically thin. In
addition, the velocity dispersion of CO(2-1) are systematically
smaller than the one of CO(1-0), suggesting that these com-
plexes are probably externally heated and composed of dense
gas. This hypothesis is compatible, as later discussed, with the
combined presence of an intense stellar radiation field due to
bulge stars (Stephens et al., 2003), the outflow detected in X-
ray in this area (Bogd´ an & Gilfanov, 2008) and the ionised gas
(Jacoby et al., 1985). This supports the view that no star forma-
tion is currently active inside the inner ring of M31.
4.2. Ionised gas
4.2.1. Spatial distribution
The ionised gas in the central regions of M31 has been mapped
by Ciardullo et al. (1988), who produced the most high S/N ra-
tio map of Hα+[NII] based on the compilation of a 5-years nova
survey, published so far. It exhibits, as displayed in the top left
panel of Figure 5, a turbulent spiral, which appears more face-
on than does the main disc. These authors described it as ly-
ing in a disc, warped such that the gas south of the nucleus is
viewed closer to face-on than the gas in the northern half of
the galaxy. This perturbed morphology is usually attributed to
shocks. Rubin & Ford (1971) argues that in this area [NII] is 2
times stronger than Hα, which supports the shock hypothesis as
also discussed by Jacoby et al. (1985) . According to Liu et al.
(2010), half of the kinematic temperature of hot gas in the cen-

Page 8

8 A.-L. Melchior and F. Combes: Molecular gas in the inner 0.7kpc-radius ring of M31
Fig.5. Hα+[NII] (left) and [NII] (right) intensities from Ciardullo et al. (1988) and Boulesteix et al. (1987) are displayed with an
arbitrarynormalisationon thetop panels,with an arbitrarynormalisation.Thebottomleft panelpresentsthe CO velocitiesmeasured
in this paper, superimposed on the gas kinematics based on (1) Hβ and [Oiii] lines measured in slits by Saglia et al. (2010), (2) [Oii]
and [Neiii] measured in slits by Ciardullo et al. (1988) and (3) Hα and [Nii] measured in slits by Rubin & Ford (1971). The bottom
right panel displays the velocity field obtained with [Nii] λ6584˙A observations (cf top right panel) based on a Fabry-Perotdevice by
Boulesteix et al. (1987). On each panel, the positions of our CO observations are indicated with circles correspondingto the beams.
tral bulge is accounted for by the stellar dispersion, but addi-
tional heating is expected from type-Ia supernovae, even though
the issue of the iron enrichment is not entirely clear. A head-on
collision with M32 suggested by Block et al. (2006) could ac-
count for the additional heating without advocating SNIa iron
ejecta and stellar mass loss not well mixed in the gas.
In parallel, Boulesteix et al. (1987) observed this same area
with a Fabry-Perot in order to map the velocity field, and pro-
duced an [NII] map, displayed in the top right panel of Figure
5. The centre has been masked for technical reasons. Beside the
lower resolution and smaller S/N ratio (due to a smaller inte-
gration time), it is instructive to observe that it exhibits different
patterns when compared with the Hα+[NII] map. It is striking
that the filament, which crosses the position M31A is not clearly
detected in [NII], so it should be mainly associated to Hα com-
ponent. The comparison of these two maps enhances the fact
that obviously forbidden lines and Balmer lines do not sample
exactly the same regions: (1) the gas does not have the same
Hα/[NII] ratio; (2) the extinction further complicates the obser-
vations.
4.2.2. Single-component velocity field
Inordertocomparetheionisedgaswithourmoleculardetection,
wetryto reconstructtheionisedgasvelocityfieldfromdatapub-
lished in the literature. Various lines have been used, but only
a single velocity has been measured for each spectrum. In the
bottom right panel of Figure 5, we display the velocity map of
Boulesteix et al. (1987). The best S/N ratio is of course observed
where the [NII] intensity is strongest. It exhibits a not very reg-
ular disc in rotation, with obvious perturbations along the mi-
nor axis. Saglia et al. (2010) interpreted these perturbations as
counter-rotations. In the bottom left panel of Figure 5, we dis-
play the slit measurements we gather from the literature, using
Dexter3when necessary, namely Hβ and [Oiii] from Saglia et al.
(2010); [Oii] and [Neiii] from Ciardullo et al. (1988); Hα and
[Nii] from Rubin & Ford (1971). We also superimpose our CO
velocities, and find a very marginal agreement. In figure 6, we
superimpose the two figures. We find a good overall agreement
for the various ionised gas measurements with complicated fea-
3http://dc.zah.uni-heidelberg.de/sdexter

Page 9

A.-L. Melchior and F. Combes: Molecular gas in the inner 0.7kpc-radius ring of M319
Fig.6. Superimposition of the various gas velocities (see Fig. 5
bottom left) on the Boulesteix et al. (1987) Fabry-Perot-derived
velocity map.
tures in the central part and along the minor axis, with the most
striking discrepancy at position angles of 128deg and 142deg.
It is difficult to understand the discrepancies observed in the
ionised gas velocity field, as the technics are quite different and
the Boulesteix et al. (1987) data suffers relatively low resolution
(even though excellent given the fact that it has been obtained
in 1985!). We can consider several explanations: (1) the various
lines might have different relative ratios from one region to an-
otherandmightbeaffecteddifferentlybyextinction.(2)it ispos-
sible that the ionised gas, like the molecular gas, exhibits several
velocity component, and it has been not explored so far, but by
Boulesteix et al. (1987), who mentioned line splittings greater
than 30 kms−1in the central region. As discussed in Sect. 4.2.3,
given the large intrinsic velocity dispersion of the ionised gas,
small line splittings (∼ 50 kms−1) will be difficult to detect and
might explain observed discrepancies in the ionised gas.
In addition, the CO velocities do not follow the regular pat-
tern.M31AandM31Ddonotreallymatchtheionisedgasveloc-
ities. However, some CO positions might have a velocity (or at
least one component) compatible with the ionised gas, namely
M31E, M31F, M31K, M31I, M31J and M31G. This suggests
that only part of the molecular gas is kinetically decoupled from
the ionised gas. Also, we wanted to figure out whether similar
line splittings were present in the ionised data.
4.2.3. Line splittings and associated velocities
As discussed by Saglia et al. (2010), the intrinsic velocity dis-
persions of the gas is smaller than 80 kms−1, while the in-
strumental resolution achieved by these authors is 57 kms−1.
Boulesteix et al. (1987), who detected line splittings larger than
30 kms−1in the central area, reached a spectral resolution of
14 kms−1. It is thus challenging to detect line splittings of
this order of magnitude, but such splittings might be underly-
ing and explain part of the discrepancies. As we observe large
(> 200 kms−1) line splittings in CO, we try to investigateif such
double components can be detected from the ionised gas.
Optical spectra of M31 are heavily dominated by the stel-
lar continuum of its bulge. Therefore, in order to study emis-
sion lines in that spectral region we approximated the calibrated
sky subtracted spectra of M31 kindly provided by R. Saglia by
Fig.7. Line splittings detected in [Oiii]50007˙A
stellar population models using the NBursts full spectral fitting
technique(Chilingarian et al., 2007b,a). We excludednarrowre-
gions around Hβ, [Oiii], and [Ni] lines from the fitting. Then
we subtracted the best-fitting stellar population models from the
original spectra and analysed the fitting residuals. This allowed
us to reliably measure parameters of these rather faint emission
lines which are barely visible in the original data. We then in-
spect visually the spectra to determine the positions where high
S/N line splittings is present. We then fit a two Gaussian func-
tionswiththeCLASS packageonthe[Oiii]5007˙Aline.Template
mismatch and the relative low S/N affect the measurementof the
Hβ, which is weaker and not always detected. For none of our
measurements (but one), the [NI] line is detected. Our detection
are summarised in Table 5 and Figure 7. We identify 14 posi-
tions with a double component detected in [Oiii] and also in Hβ,
within 23′′( 86pc) from the centre.
Several (8/14) of the positions with line splittings have a
component with a large [Oiii]/Hβ ratio. Such a large ionisation
is compatible with planetary nebulae. del Burgo et al. (2000)
observed several high-ionisation “clouds” in this area. These
authors discussed that the intensity of 3 of these sources is
much larger than those of planetary nebulae detected in M31
by Ciardullo et al. (2002). However, in this central region, it
might be possible that these sources are multiple. Interestingly,
one of the sources (D) detected by del Burgo et al. (2000) ex-
hibits a line splittings (-270,-527) with an amplitude compara-
ble with ours, even though none of the velocities really match.
Also, the strongest component we detect in the inner arcsec
region matches in first approximation with the source A of
del Burgo et al. (2000).
An interpretation of the other positions with line splittings
detected within 20′′of the centre is difficult to constrain. In ad-
dition, one can note that this gas does not need to lie in a plane
and could be a 3D feature (cf. the outflow detected in X-ray by
(Bogd´ an & Gilfanov, 2008)).
4.2.4. Modelling of a tilted inclined disc
We create a velocity field map with the CCDVEL task
within the NEMO software package (Teuben, 1995). Following
Ciardullo et al. (1988), we consider a tilted ring model with an
inclination of 45deg, a position angle of 39.8deg and a systemic
velocity of -310 kms−1. In order to understand the behaviour of

Page 10

10 A.-L. Melchior and F. Combes: Molecular gas in the inner 0.7kpc-radius ring of M31
Table 5. Line splittings detected on [Oiii] and Hβ lines from Saglia et al. (2010). We provide the velocities, signal-to-noise ratio of
the integrated line and [Oiii]/Hβ line ratio for each detected component (1 and 2). We also give for comparison purposes the single
values published by Saglia et al. (2010).
offsets R(′′)
-22.8
-18.2
-16.2
-12.5
-10.9
-6.2
-0.9
0.7
1.0
1.5
1.2
9.3
11.7
17.2
PA(deg)v[Oiii]5007
1
-548.6±12.0
-462.8±6.0
-449.5±9.1
-485.8±9.9
-454.3±14.4
-319.9±20.0
-281.3±8.9
-319.9±19.8
-332.6±11.0
-322.3±13.0
-321.5±13.2
-598.0±4.7
-593.7±5.5
-575.1±3.8
(kms−1)S/N1
4.7
13.1
10.4
4.9
3.3
3.3
8.7
3.3
6.8
4.9
5.2
14.1
8.4
6.4
[Oiii]/Hβ
0.51+0.20
0.29+0.09
0.73+0.20
1.17+inf
0.15+0.23
0.14+0.34
0.78+1.05
-0.20+0.14
0.60+0.22
-0.06+0.17
0.20+0.26
0.92+0.37
1.17+0.56
-0.16+0.17
v[Oiii]5007
2
-360.6±16.7
-234.0±4.0
-225.4±5.2
-375.0±6.7
-272.5±13.3
-144.8±15.9
-85.9±16.2
-144.8±16.1
-64.9±7.5
-56.0±3.5
-78.7±4.8
-319.2±14.0
-373.3±29.2
-444.8±35.2
(kms−1)S/N2
5.2
15.4
12.2
3.9
5.3
3.6
3.2
3.7
7.5
18.2
12.7
8.1
3.7
4.5
[Oiii]/Hβ
-0.22+0.10
0.79+0.37
0.86+0.32
-0.23+0.20
0.41+0.22
0.19+0.32
0.00+0.72
-0.04+0.20
1.04+inf
0.92+0.32
0.57+0.22
0.18+0.12
-0.18+0.16
0.28+0.17
vgas
Saglia
-15.2,16.9
-13.5,-12.2
-12.0,-10.8
-9.3,-8.4
-8.1,-7.3
-1.9,-5.9
-0.2,0.9
0.1,-0.7
0.7,0.6
1.1,1.0
1.2,0.3
8.8,-2.9
11.2,-3.6
5.3,16.4
138
48
48
48
48
18
168
168
48
48
78
108
108
18
−0.19
−0.08
−0.16
−0.71
−0.25
−0.31
−0.33
−0.18
−0.18
−0.17
−0.22
−0.22
−0.28
−0.16
−0.12
0.22
-469.1±8.4
-323.7±7.4
-314.8±5.4
-418.5±10.7
-314.4±9.2
-342.7
-241.8±10.0
-221.7±11.6
-313.6±7.8
-330.6±7.8
-327.8±11.9
-343.6±13.9
-376.8±18.1
-291.4±6.6
0.21
0.21
0.20
0.28
0.41
0.22
−0.40
0.20
0.17
−0.12
−0.19
−0.18
Fig.8. Superposition of the various gas velocities (see Fig. 5 bottom left) (left:) on a simple model of a galactic disc with an
inclination of 45deg, a position angle of 39.8deg and a systemic velocity of -310 kms−1; (right:) similar disc with variable position
angles
the gas in the central part of M31, we superimpose the veloc-
ities measured in this paper (ionised gas and CO gas) on this
modelling in Figure 8. Even though there is a velocity field, as
shownwithionisedgasbyBoulesteix et al.(1987)anddisplayed
in the bottom right panel of Figure 6, it is not regular and does
not exhibit a well-defined zero-velocity curve. In addition, the
velocities of the CO measurements do not fit with such a regular
speed pattern. Varying the velocity angle in the centre mimics
the velocity pattern in the inner region (< 10arcsec) but does
not explain the minor axis configuration.
4.3. HI emission
ThedetailedmapsfromBraun et al.(2009)haverevealedclearly
the HI deficiency in the central parts of M31. There is however
some HI emission in the central kiloparsec, but most of it comes
near the systemic velocity, and is likely to be projected emission
fromthe externaltilted gas orbits,whichare warpedto an almost
edge-on inclination (e.g. Corbelli et al., 2010). To subtract this
projected large-radii emission, and better see the residual com-
ingfromtheactualcenter,withalargevelocitygradient,wehave
summed all channels avoiding ±100km/s around the systemic
velocity of -310km/s. A weak signal can then be distinguished,
with a morphology of an incomplete ring, with emission cor-
responding to the east side of the inner ring delineated by the
dust emission revealed in the Spitzer map. Figure 9 shows the
dust contours superposed on the residual high-velocity central
HI emission. In this picture, we can see that the main HI resid-
ual component still follows the large-scale (NE-SW) arms seen
in projection, and coinciding with the dust features, in the direc-
tion of the major axis of M31. However, there remains a weak
component perpendicular (NW-SE) to it. The main concentra-
tion of this residual HI emission is well aligned with the North-
East part of the dust ring and it could correspond to the weak
componentseen by Brinks (1983) on the minor axis position ve-
locity diagram (his figure 1b). However, the remaining parts of
the ring is devoid of HI emission; the atomic gas must have been
transformed into the molecular phase in the dense parts of the
ring. The Figure in the Appendix reveals that most of the CO
strong emission at high velocity in the center has no HI counter-
part.

Page 11

A.-L. Melchior and F. Combes: Molecular gas in the inner 0.7kpc-radius ring of M3111
Fig.9. Image of the HI emission in the very centre, avoiding
±100km/s around the systemic velocity of -310km/s. This cut
has been made to subtract the large-scale HI emission seen
in projection superposed to the center, expected near the sys-
temic velocity. The contours are the dust emission from Spitzer
(Block et al., 2006). The HI cube is from Braun et al (2009).
Fig.10. Velocity field of the HI emission, corresponding to
Figure 10, avoiding large-scale HI emission.
The HI component associated to the dust ring is, however,
too weak to be seen in the velocity map, as shown from Figure
10. The velocity field in the HI selected with high speed with
respect to the systemic velocity, is still compatible to the “nor-
mal” inner disc, and similar to the velocity field found with the
ionised gas in Hα, although with possible perturbations in the
vicinity of the ring.
4.4. Other wavelengths information
While the centre of M31 hosts a massive black hole with a
mass of 0.7 − 1.4 × 108M⊙ (Bacon et al., 2001; Bender et al.,
2005), it is one of the most underluminous supermassive black
hole (Garcia et al., 2010). The A-star cluster, detected in the
P3 component, can be associated to a recent star formation
episode. Bender et al. (2005) estimate a single burst, which oc-
curred 200Myr ago, with a total mass in the range 104– 106M⊙,
corresponding to an accretion rate of 10−4– 10−2M⊙yr−1. It
might be triggered by the possible frontal collision with M32
(Block et al., 2006). The detection of an ionised gas outflow in
X-raysalongtheminoraxisofthegalaxybyBogd´ an & Gilfanov
(2008), perpendicular to the main disc, could be linked to this
recent star formation activity. However, this is still controver-
sial as such a burst does not have enough energy to power the
galactic wind required (Bogd´ an & Gilfanov, 2008). In the left
panel of Figure 11, Hα+[Nii] contours are superimposed (with
the same resolution) on the Chandra soft X-ray emission map
from Bogd´ an & Gilfanov (2008). The relative intensity of the
outflow on both sides is compatible with the intensity of ABex-
tinction: the NW side is more extinguished than the SE side.
This is compatible with the modeling described in the next sec-
tion. Last, the right panel of Figure 11 displays superimposition
of the contours of the ABextinction map on the PAH-dust emis-
sion at 8µm detected by Spitzer. The extinction features match
exactly the ring detected at 8µm.
5. Interpretation
The molecular, atomic and ionised gas exhibit different radial
distributions in disc galaxies as first discussed by Kennicutt
(1989) (see also Bigiel et al., 2008, for a recent review). The HI
gasis knowntoextendat muchlargerradius.M31hasHI gasex-
tendingat least upto 40kpc (Corbelli et al., 2010; Chemin et al.,
2009) with high velocity clouds up to 50kpc (Westmeier et al.,
2008). As discussed below, the projected warp dominates the HI
emission in the central part (see also A). The CO and ionised gas
are usually more concentrated.In M31,while CO is knownto be
depleted in its centre (Nieten et al., 2006), the ionised gas is de-
tected in the inner parts (Ciardullo et al., 1988; Boulesteix et al.,
1987) and is strongest there (Devereux et al., 1994).
While the atomic and ionised gas exhibit the presence a per-
turbed disc, the molecular gas detected in CO displays unex-
pected kinematic signatures, with significant line splits close to
the minor axis and a very weak when present signal close to the
systemic velocity. Several scenarios could be invoked to explain
the existence of two well separated high S/N velocity compo-
nents (∆V = 260 kms−1on a scale of 40pc) traced by the molec-
ular component (CO emission) in the center of M31. One com-
ponent is in the sense of the expected rotation, the other com-
ponent is counter-rotating. The widths of these two components
are also very different, as displayed in Figures 3 and 4.
InSect. 5.1, we first explorethe possible scenarios to explain
the observations. In Sect. 5.2, we discuss a possible modelling
accounting for the various observables.
5.1. Various possible scenarios
5.1.1. Scenario 1: large scale warp
We could think that the second component, the counter-rotating
one, is just observed in projection, and comes from the exter-
nal warp observed in the outer parts of the M31 disc, with a
different inclination and position angle than the normal M31
disk. The warp is visually conspicuous, and remarkable in HI
(Corbelli et al., 2010). Farther in the North-East disc (beyond
10kpc), Casoli & Combes (1988) already noted that CO emis-
sion could come from the main disc and the warped component,
both on the same line of sight. In that case, the CO emission
(T∗
16kpc from the centre.
Thisscenariois, however,impossible,since: (1)the expected
molecular gas has negligible emission at the large radii where
A∼ 40mK) was coming from material distant by more than

Page 12

12A.-L. Melchior and F. Combes: Molecular gas in the inner 0.7kpc-radius ring of M31
Fig.11. Left: Superposition of the Hα+[Nii] map (upper left panel of figure 5) of Ciardullo et al. (1988) on the Chandra soft X-ray
emission from Bogd´ an & Gilfanov (2008, see their Figure 7). The Hα+[Nii] map has been smoothed with the same smoothing
lengths as those used for the Chandra map (A. Bogdan, private communication). Right: Superposition of the ABextinction on the
8µm Spitzer map after subtraction of a scaled version of the 3.6 µm image Block et al. (2006). The ABcontours are fixed to 0.05
(blue) and 0.15 (white).
this warped component should be (e.g. Neininger et al., 1998);
(2)thecorrespondingprojectedvelocityonthelineofsightclose
to the centre is expected aroundthe systemic velocity.It is likely
that the gas at large distance (16-30kpc) is on nearly circular
orbits, and when projected to the centre, has no radial veloc-
ity. Accordingly, the correspondingwarped material was always
found at systemic velocity. Whatever inclination it has, the gra-
dient in velocity across a small region of 40pc should be negli-
gible. Ad hoc hypothesis, with HI infall/outflow with important
radial velocity, are then required, which are incompatible with
the HI observations(e.g. Corbelli et al 2010).We thereforethink
that, unless the gas is completely out of equilibrium (which is
notreallysupportedbytheobservationsofCorbelli et al.(2010);
Casoli & Combes (1988)), the peculiar component cannot come
fromtheouterwarp.However,wehavesometentativedetections
close to the systemic velocity (M31K, M31I, M31G, M31J),
which could be accounted for by the external warp. Some po-
sitions (e.g. M31G) suffer from signal subtraction due to the off
positions (wobblermode ofobservation),and we cannotexclude
that the real signal could be a bit larger.
5.1.2. Scenario 2: large scale tidal streams
The velocity range measured in CO is comparable with the
measurements performed by Ibata et al. (2004); Chapman et al.
(2008) for stellar streams and for extra-planar gas and high-
velocity clouds detected in HI outside the disc Braun & Thilker
(2004); Westmeier et al. (2008). Casoli & Combes (1988) has
detected CO up to 16kpc. Braun & Thilker (2004) detected a
faint bridge of HI emission which appears to join the systemic
velocities of M31 with that of M33 and continues beyond M31
to the North-West. Davies’s cloud (Westmeier et al., 2008) lies
on the North-West part of M31’s disc and could belong to this
bridge. Similarly to the stellar streams observed in the South-
East part of M31’s disc (Ibata et al., 2007), the HI gas ex-
hibits large design loop-like features (e.g. the Magellanic stream
Kalberla & Haud, 2006). The second component detected in the
molecular gas we have detected could be associated to the M31-
M33bridgeor anygaseous-looprelics. It would thencorrespond
to some accretion of gas onto the disc generating an inner po-
lar ring and leading to compression and excitation of the CO.
Such a configurationcan be comparedto the inner polar gaseous
disc discussed by Sil’chenko et al. (2011) in NGC7217 where
the disc is face-on.
While HI studies mention contamination by the Galaxy
(Westmeier et al., 2008), our CO detectionsare related to extinc-
tion patterns, clearly associated to M31. The CO detected for
M31D has a mean velocity of -74 kms−1is typically below the
limit for Galactic contamination used for HI. However, typical
velocity dispersion of HVC in the Galaxy is 8.5 kms−1, while
the dispersion measured for M31D is 24 kms−1. Furthermore,it
exhibits a larger than expected ICO/E(B-V) ratio, rather suggest-
ing that it lies behindthe bulge and not in frontof it. Last, no CO
from the Milky Way has been detected by Dame et al. (2001) in
the vicinity of M31.
Last,wecannotexcludethatsomenon-circularvelocitiesde-
tected in CO result from relics of M31’s formation: scattered de-
bris or other extra-planar gas.
5.1.3. Scenario 3: an inner bar
Another a priori plausible scenario could be that the second ve-
locity componentis associated to a possible bar.Bar are frequent
in spiral galaxies (60 to 75% are barred according to the bar
strength) (e.g. Verley et al., 2007), and it is likely that M31 is
barred (Beaton et al., 2007).
To account for non-circular motions, elliptical orbits of
the gas due to a bar potential are an obvious solution.
There have been several propositions of bar models for M31:
Stark & Binney (1994) assume the existence of an analytic
Ferrers bar potential in the center, to explain some non-circular
motions there, but there was no complete velocity field, and
they do not try to fit the observed stellar distribution with a bar.
Athanassoula & Beaton (2006) investigatethe twist of isophotes
of the NIR 2MASS image, which shows a boxy shape bulge.
They compare this morphology with simulated stellar models
with bars, but without gas. It is possible that a bar exists inside
the bulgeof M31.However,thebulgealonecouldberesponsible

Page 13

A.-L. Melchior and F. Combes: Molecular gas in the inner 0.7kpc-radius ring of M3113
for the isophote twist and the boxy shape, and it is not possible
to derive any bar orientation in the disc. Athanassoula & Beaton
(2006) also compare a 2D gaseous velocity field in an analyti-
cal bar potential (not fitted to M31), put then projected edge-on.
The model 2D gas morphology clearly shows a bar, with emp-
tied regions, that do not correspond to the de-projected view of
M31 gaseous disc (e.g. Braun, 1991). From the HI-derived gas
morphology, Braun (1991) suggests a two-armed trailing den-
sity wave, with a pattern speed of 15 kms−1kpc−1, leading to
a corotation at 16kpc, an ILR at 5kpc and an OLR at 22kpc.
He proposes as the driver of the spiral structure the collision
with M32, which could also be responsible for the many tilts
of the M31 plane. The nearly head-on collision with M32 has
been developped in detail by Block et al. (2006), which allows
them to account for the two off-centered gas rings observed in
M31 at 0.7kpc and 10kpc radius. Indeed, they note that the gas
morphology in M31 does not show the usual signatures of the
response to a bar. Although it is possible that the galaxy disc
possesses a stellar bar, aligned with the triaxial bulge, oriented
roughly along the major axis, the observed 0.7kpc dust ring,
is aligned almost along the minor axis, and not aligned to the
potential bar along the major axis. The size of the inner ring
does not correspond to a possible inner Lindblad resonance, nor
the outer ring at 10kpc radius to a correlated outer resonance.
Moreover, the inner disc appears highly inclined with respect to
the main disc, and the inner ring is off-centred by 40% of its ra-
dius, which is not expected in barred galaxies. This offset centre
and the tilted plane of the inner disc strongly point towards a
perturbed origin of the inner features of M31.
Beside these dynamical arguments, there is no signature of
the bar in the gas component, and the best models for the gas
distribution are in terms of rings or tightly wound spiral arms
(Braun, 1991; Block et al., 2006). The only possibility is the bar
is along the minor axis, a way to hide its signature almost com-
pletely. Indeed, the elongated gas distribution would become al-
most round by the factor ∼ 4.4 stretch due to the inclination of
77deg, and also there would be no non-circular motions, since
the bar is aligned along the symmetry axis of the projection. We
could then expect to observe two velocity components along a
single line of sight, due to different orbits. This is equivament
to the shocks predicted by hydrodynamical models at the spi-
ral arms, or dust lanes leading the bar. However, the amplitude
of such a velocity gradient at maximum for strong bars is much
lower than is observed here in the center of M31.
Several strongly barred galaxies have been studied in Hα
spectroscopy,determining the kinematics with high spatial reso-
lution, comparable to what we have in the CO gas towards M31.
NGC1365 is the prototypeof a very strong bar. The position an-
gle of the bar is not coincident with the symmetry axis of the
projection, so that the bar signatures are obvious. With an incli-
nation of i=42deg, and a distance of D=20Mpc, the observed
velocity gradient across the bar dust lanes is determined to be
at maximum 40kms−1, with a spatial resolution below 100pc
(Lindblad et al., 1996). If deprojected, the maximum would be
60kms−1. NGC 1300 is also a remarkable and strongly barred
spiral galaxy, where the kinematics has been studied in detail in
Hα, on scales smaller than 100pc. Lindblad et al. (1997) present
a projected velocity gradient, across the dust lanes shocks of 20-
30 kms−1projected on the sky. When deprojected, this becomes
a maximum of 35-50 kms−1. It is to be noted that to have these
strongvelocitygradients,a verycontrastedgas structuremust be
found, which is not the case of the M31 central region.
In the molecular component, interferometric observations
of nearby barred galaxies can yield now sub-arcsec reso-
lution, corresponding to scales smaller than 100pc, there-
fore comparable to the single dish resolution towards M31.
Garc´ ıa-Burillo et al. (2005) have presentednon-circularmotions
due to bars, which are of the same orders of magnitude ∼ 60-
70 kms−1. More recently, with even higher sensitivity and res-
olution, van der Laan et al. (2011) have for the first time seen
dedoubled velocity profiles in the CO data, towards the barred
galaxy NGC 6951. If interpreted in terms of crowding of orbits,
they reveal velocity differences of 80 kms−1, which in deprojec-
tion would come up to 110 kms−1. The CO velocity profiles are,
however, very broad in the center of this galaxy, showing a star-
burst ring, and it is not sure what is due to turbulence or to the
bar. The molecular gas surface density in the center is in average
100M⊙/pc2, which may explain the unstable and turbulent gas
forming stars, while in M31, the surface density is at least one
order of magnitude lower. The velocity dispersion in the distinct
velocity components observed in CO is quite low (15 kms−1),
which corresponds to a stable gas layer. We conclude that the
bar is not a likely explanation for the special kinematical phe-
nomenon observed towards the M31 center.
5.1.4. Scenario 4: head-on collision with M32
Jacoby et al. (1985) have shown that the inner disc of M31 is
tilted with respect to the main disc. While the inclination of the
large-scale disc is 77◦, the nuclear disc, of size ∼ 1kpc, is in-
clined by only about 40◦on the sky plane. Block et al. (2006)
discovered in the Spitzer-IRAC maps an inner dust disc of scale
1kpc by 1.5kpc, which also appear with such a low inclination.
The ring morphology of this tilted structure, together with the
ring (10kpc)morphologyof the large-scale disc led them to pro-
pose a head-on collision with a small companion, with a proba-
ble candidate being M32.
In this scenario, we assume that the peculiar component
comes from a tilted ring-like material, likely coming from the
perturbed gas due to the recent M32 collision. The collision
might have perturbed the nuclear disc, already tilted, and pro-
duce some local warp, before the annular density wave propa-
gates outwards. Beside its interest for the inner configuration of
M31, this scenario also proposes an explanation for the 10kpc
ring, resulting from the propagation of the initial annular wave.
The resulting configuration is similar to the scenario 2 pro-
posed in Sect. 5.1.2. In the following, we discuss some mod-
elling which enables one to explain the observations.
5.2. Modelling of an inner disc with an inner ring
We use a model for the gas, through static simulations of dy-
namicalcomponentsrepresentedbyparticles,withdifferentgeo-
metrical orientations on the sky, but embedded in a gravitational
potential representing the observed rotation curve of M31.
We represent the gaseous disc of M31 by a rather homo-
geneous Miyamoto-Nagai disc of particles (with radial scale of
1kpc,andheightof0.2kpc),tobe ableto varyeasily the inclina-
tion and position angle as a functionof radius. We used typically
half a million particles to have enough statistics. We plunge the
gas disc into a potential made of a stars and a dark matter halo.
The stellar component is composed of a bulge and a disc. The
bulge is initially distributed as a Plummer sphere, with a poten-
tial:
Φb(r) = −
GMb
?
r2+ r2
b
(1)

Page 14

14A.-L. Melchior and F. Combes: Molecular gas in the inner 0.7kpc-radius ring of M31
where Mband rbare the mass and characteristic radius of the
bulge, respectively (see Table 6).
The stellar disc is initially a Kuzmin-Toomre disc of surface
density
Σ(r) = Σ0(1 + r2/r2
d)−3/2
(2)
with a mass Md, and characteristic radius rd.
The dark matter halo is also a Plummer sphere, with mass
MDMand characteristic radius rDM. A summary of the adopted
parameters is given in Table 6. The resulting rotational velocity
curve is rising until a maximum of 300kms−1, reached at 3kpc,
and thenslightly decreases, remainingclose to 300kms−1in the
centralregionsofinteresthere.Thiscorrespondstotherotational
velocity observed (e.g. Carignan et al., 2006). All the molecular
clouds we want to reproduce are inside the radius of maximum
rotational velocity.
The gas particles are distributed in this potential, with veloc-
ity dispersion corresponding to a Toomre Q parameter of 1.3.
Table 6. Model parameters
Componentradius
[kpc]
0.2
4.0
10.
Mass
[M⊙]
1.6e10
7.e10
27.e10
Mass fraction
[ %]
4.5
20.
75.5
Bulge
Disc
Halo
Fig.12. Mean density of CO emission from the homogeneous
nuclear disc plus tilted ring of the model. The field of view is
200 arcsec in radius, or 0.76kpc in radius. The density scale,
indicated on the wedge, is in arbitrary units.
In order to reproduce the observations, we adopt an inclina-
tion of 43◦for the gaseous nuclear disc, inside 1.5kpc radius. At
the boundary of this disc, we assume a progressive warping of
the disc plane,so thatthe inclinationon thesky growsfrom43 to
77degrees, over 300pc. The details of this transition, however,
are notconstrainedby theobservations,since the latter are all in-
side1kpcradius.Asforthepositionangleonthesky,thenuclear
disc has PA= 60◦, unlike the main disc, which has PA= 35◦. In
Fig.13. First moment of the velocity field of the simulated nu-
clear region, in the same field of view as Fig 12. Signatures of
the tilted ring can be seen in the boarder of the field. The wedge
gives the velocity scale in kms−1.
Fig.14. Second moment (velocity dispersion) of the simulated
velocity field. The wedge gives the velocity scale in kms−1. The
locations of double components, with counter-rotating features
correspond to the blue and purple regions.
projection on the sky, the nuclear gas disc model gives veloci-
ties in agreement with the observations, at least with the main
velocity component. In many observed points only this main
component is observed, with a broad line-width (FWHM=50-
70 kms−1), located on the blueshifted side in the SW and on
the redshifted side on the NE of the major axis. In some of the
points, there is an additional peculiar velocity component, lo-
cated on the opposite side (redshifted on the SW), and narrower
(20 kms−1).
Trying several parameters, we found that material on a ring-
like orbit, with 40◦inclination, and PA= -35◦gave kinemati-
cal features compatible with the observations (cf Figure 12-15).
The modelled ring has a width of 0.4kpc, and a mean radius
of 0.6kpc, coinciding with the observed dust ring. The adopted

Page 15

A.-L. Melchior and F. Combes: Molecular gas in the inner 0.7kpc-radius ring of M31 15
positionangle and inclinationof the variouscomponentsare dis-
played in Table 7.
Table 7. Geometrical parameters
ComponentMain disc
[◦]
35
77
Nuclear disc
[◦]
60
43
Inner ring
[◦]
-35
40
PA
Inclination
Fig.15. Spectra extracted from the simulated cube, at the (RA,
DEC) offset in arcseconds indicated in each panels. The vertical
scale is in arbitrary units.
To draw these maps, we built data cubes corresponding to
the observations, with a pixel of 6 arcsecond, and channels of 12
kms−1, and the data smoothed to a beam of 12′′. The pixel size
ofthecubesweretherefore(60,60,40)to welldescribetheinner
parts studied here in CO lines and dust extinction. To take into
account that the gas and dust distribution is not homogeneous,
but patchy, we smoothed to the same spatial resolution the dust
map obtained by Block et al. (2006) from Spitzer IRAC images
(the8µm mapwiththe3.6µm contributionsubtracted),andused
it as a multiplicative filter to our homogeneous disc map. This
took into account that the dust inner ring is offset by 0.5kpc
(131”) from the centre of the galaxy.
Typical spectra, in the zone where double features are ex-
pected, have been extracted, at the offset indicated in Fig 15 in
arcseconds. The broad characteristics of the blue-shifted veloc-
ity component is due to beam smearing of the velocity gradient,
Fig.16. Schematic view of the interpretation proposed for the
CO velocities observed. The inner disc is presented with a PA
of 60deg and an inclination of 43deg. The inner ring is superim-
posed with a similar inclination but a position angle of -35deg.
The straight line indicates the position of the major axis of the
main disc inclined by 77deg with a PA of 35deg.
as illustrated here with a 12′′beam. Interestingly, some emis-
sion at the systemic velocity is expected due to the inclination of
a plain disc. It is thus not necessary to evoke the large scale warp
to explain the weak emission at the systemic velocity, which can
be accounted for by the inner disc.
6. Discussion and conclusions
We have detected CO emission, tracing the presence of molec-
ular gas in several positions in the central kpc of M31, corre-
sponding to dust extinction. Some of these positions correspond
also to the conspicuous inner ring, detected in dust emission by
IRAC on Spitzer(Block et al., 2006).(This ringwas also present
on ISO data studied by Willaime et al. (2001).) The velocity in-
formation brought by the CO lines reveal a complex kinemati-
cal structure, different from the ionised and atomic gas dynam-
ics. In the NW inner ring positions, two main high S/N velocity
components are detected, on each side of the systemic velocity.
Thecomponentwiththeexpectedvelocity,accordingtothe rota-
tion curve and the positive position angle, is broad, and is likely
to correspond to the nuclear disc. Exploring a wider region of
the strong central rotational gradient, the beam averaged spec-
trum may be 100km/s broad. The second component, redshifted
on the other side of the systemic velocity, has a small velocity
width, since it is not a full disc, but a relatively narrow ring. For
the model, we used radii from 0.4 to 0.8kpc.
We summarise in Figure 16 the geometry we propose in our
modelling. The peculiar component appears counter-rotating,
since the gas is in a tilted ring, almost perpendicular to that of
the inner disc. Its inclination on the plane of the sky is similar,
whichexplainswhythe amplitudeofthe projectedrotationis the
same as for the regular component. In the Figure, the near sides
of the two components are coloured, while the far sides are kept
white. The red and blue colours correspond to the redshift (and
blueshift) relative to the systemc velocity. The arrows indicate
the apparent sense of rotation. The two discs appear to rotate in
the same sense together, and the same sense as the main M31