3. Spatial distribution
3.1. CO velocity integrated map: WCO
While in the resolution maps 3 of Dame et al. (1993) the CO emission in M 31 has the fairly smooth appearance of a ring, at the 1´ effective resolution of our survey, this ring is resolved into the well defined spiral arms (Fig. 5) studied by Baade (1963). His arm S4 is especially prominent at a galactocentric radius R 9 kpc, while closer to the center, S3 at R 6 kpc is still clearly visible, though less continuously defined. In the upper part of the map 4, near the minor axis, S3 and S4 are not easily separated owing to the steep inclination of the galaxy.
Near the major axis, S3 and S4 are linked by an ill-defined bridge which connects to S4 at the location of the large OB association NGC 206. Although multiple supernovae explosions seem rarely capable of creating perturbations of the surrounding interstellar medium on scales larger than 1 kpc, it is possible that the perturbed structure observed here results from the stellar winds and supernovae explosions associated with NGC 206, a very large young cluster containing more than 300 stars more massive than 20 (Van der Bergh 1964; Odewahn 1987). This cluster is among the most impressive sites of star formation in the Local Group.
Smaller-scale emission features are visible throughout the map. Most of the small features located farther from the center than S4 are aligned with S5 (as we will show also from the position-velocity diagrams; Sect. 4). The complexes closer to the center than S3 - some of which may be aligned with Baade's arm S2 - obey a rather perturbed kinematics, as discussed later.
As in the Milky Way, much of the CO emission in M 31 comes from large molecular complexes. In particular, the structure of the molecular arm S4 is remarkably similar to that portion of the Carina arm in the Milky Way lying at about the same galactocentric radius, as becomes evident when the Carina arm is smoothed to the linear resolution we achieved in M 31 (Fig. 6). In both instances, the emission is resolved into large complexes typically a few hundred parsecs in size with comparable CO luminosity and masses of several times 106 (assuming the Milky Way CO mass calibration for M 31). The CO profiles averaged over the complexes have similar line widths and peak temperatures in M 31 and in the Milky Way. On average, one such large complex is detected about every kiloparsec along each arm. The large sizes, masses, and line widths of these complexes indicate that they are not individual GMCs; the higher resolution maps of the Carina arm (Fig. 6g) show that each of the large complexes consists of several GMC's. Given the similar surface density distributions, it is likely that arm S4 is also comprised of groups of GMC complexes. Such complexes have indeed been resolved by the interferometric observation of Neininger et al. (1998). Hence, despite the difference in total CO luminosity between M 31 and the Milky Way, at R larger than about 8 kpc (i.e. the solar circle), both galaxies are apparently very similar in molecular clouds. The main difference is in the inner 7 kpc, where the Milky Way is bright in CO and M 31 relatively dim.
3.2. Comparison with maps of other tracers
Since the early CO observations of M 31 by Combes et al. (1977a, 1977b), it has been known that CO was preferentially detected in the direction of the dust lanes. As the present data show, there is generally a good correlation between CO and dust (Fig. 7), and the molecular arms S3 and S4 are particularly prominent as dust lanes on the optical image. The correlation is less good near the upper part of the minor axis, probably as a result of the inclination of the galaxy: the dust there is located behind most of the stars which provide the background illumination, yielding little apparent obscuration even if the dust optical depth is large. Near the lower part of the minor axis, on the other hand, S3 appears quite dark on the optical image while the associated CO emission is modest. The dust there is located in front of the very bright stellar bulge of M 31 and even small optical depth yields conspicuous obscuration. The ill-defined bridge connecting S3 and S4 near the major axis is also clearly visible on the optical image, and appears to have a structure similar to that of the CO emission. We note finally that most of the small clouds seen in the map are also visible on the optical image, especially those located to the north and the east of S4.
The overall structure of the CO emission is qualitatively fairly similar to that of the H I (Figs. 8a and 8b), the two main differences being (i) that, as is the case in the Milky Way, the CO emission in M 31 is intrinsically more clumpy than the H I ; and (ii) that substantial H I emission is associated with Baade's outer spiral arm S5 (tangent at X -70´ along the major axis), whereas relatively much less CO emission is detected there 5. There is no indication of H I enhancement at the position of Baade's inner arm S2.
The IR(100-µm) distribution (Fig. 8c) is also fairly similar to that of CO. The main difference is that the IR image shows a central peak, absent in CO. According to Soifer et al. (1986), the central IR emission comes from dust shed into the interstellar medium by late-type stars in the bulge. The 100-µm emission associated with this dust is strong because it is significantly warmer than normal interstellar dust (e.g. that in the Solar neighborhood); the associated dust mass is only about 1500 (Soifer et al. 1986). The outer arm S5 contains very little IR emission, while a faint but significant enhancement of the 100-µm emission is seen at the position of the inner arm S2.
The S3 and S4 arms also appear to have well-defined UV counterparts (Fig. 8d) along which the OB associations listed by Magnier et al. (1993) are distributed. The most obvious difference between the UV image and the CO map is that the UV image shows a central source. It is noteworthy that the larger OB associations have apparently dug holes in both the atomic (Brinks & Bajaja 1986) and molecular gas. The hole around NGC 206 is particularly impressive, but others also stand out (see especially the regions around (-8´;+12´) and (-48´;+8´)).
Near the intersection of S4 with the major axis, where our linear resolution in the galactic plane across the arm is best, the UV peak is offset outward in radius from the CO peak by several minutes of arc (Fig. 9). Out of the 15 OB associations listed by Magnier et al. along S4 between X = -30´ and X = -60´, 14 are outwards of the CO feature. This offset is unlikely to be a mere result of extinction of the UV by the dust in the arm because if that were the case, UV emission ought to be seen on both sides of the dust feature. That the offset is seen in the same direction for both the upper and the lower part of S4 further rules out absorption as the explanation for the displacement of the UV and gas/dust arms. The absence of similar offsets at other azimuthal angles might simply result from the steep inclination of the galaxy which degrades the resolution parallel to the minor axis.
It is noteworthy that at the tangent point of S4, in that region where an offset between the CO/H I and the UV peaks is detected, the arm defined by the H II regions is aligned neither with that defined by the gas, nor with that defined by the UV, but lies between the two: the H II regions are mostly found along the edge of the gaseous arm that faces the OB associations (Fig. 9). Such a "stratification" of the various tracers of the ISM and star formation across the spiral arms, also reported in the Galaxy (Roberts 1972), M 33 (Courtès & Dubout-Crillon 1971) and M51 (Vogel et al. 1987), is consistent with star formation triggered by a spiral density wave. However, the large inclination of M 31 along the line of sight and the warping of the disk at R = 6-9 kpc makes it difficult to measure accurately the linear offset between the OB associations and the gas. It was therefore not feasible to compare this distance to the ages of the associations.
If a spiral density wave is the origin of the stratification observed, associated streaming motions ought to be present across the arm. Although large local velocity shifts are detected (particularly around large OB associations such as NGC 206), no organized pattern could be discerned.
3.3. Radial distributions
To compare more quantitatively the distribution of CO with that of other tracers of the ISM and star formation, it is useful to examine radial distributions (Fig. 10). The most complete summary is obtained by averaging the maps into deprojected annuli; however, the linear resolution of the result is much worse than that along the major axis direction owing to the steep inclination of the galaxy. This degradation of linear resolution in the plane of the galaxy implies that the identification of individual spiral arms, or the detection of possible offsets between the various tracers will be difficult to obtain from such radial distributions. It is helpful therefore to examine also the radial distributions along the major axis alone.
Because the plane of M 31 seems to be slightly warped (the tangent point of S4 along the major axis is located at Y +3´), the distributions along the major axis were obtained by averaging the data between Y = -2´30" and Y = +5´00". While both radial distributions were binned every kiloparsec, the distributions averaged into annuli were computed out to a radius of 12 kpc (our coverage beyond is uncomplete), and those along the major axis were computed out to 13 kpc.
The velocity integrated CO emission (Fig. 10a) rises almost continuously from the Galactic center to a peak in the Population I ring near R 8.5 kpc. There is a possible secondary peak at R 2.5 kpc corresponding to Baade's arm S2. Beyond about 12 kpc, the CO emission drops rapidly, but does not fall to zero, even at the outer edge of our survey. Note that little significance can be given to the values found for the first CO bin (0 kpc R 1 kpc) because only 5 pixels of our map fall in that range, compared with several tens of pixels for each subsequent range.
The radial distribution of IR(100-µm) flux (Fig. 10a) shows a strong peak at the center where the emission is dominated by interstellar dust in the bulge (Soifer et al. 1986). Beyond a radius of about 4 kpc however, the warm interstellar dust of the disk is the main source of 100-µm emission, which mimics the CO, and reaches a maximum at about the same radius. The arm S2 is marginally evident at R 2 kpc as a slight bump (see also Fig. 8c).
The UV (Fig. 10b) has a very strong peak at the galactic center and decreases steadily out to a radius R of about 5 kpc. Emission coming from S3 and S4 contributes to the peak at R = 8-10 kpc. In the radial distribution along the major axis (Fig. 10b2), an offset between the peak of UV emission and that of CO emission is evident. This offset is not detected in the distributions averaged into annuli because, as explained above, such averaging degrades the resolution to that available along the minor axis. The H II regions (Fig. 10c) are also more abundant along S4, and slightly offset towards larger radii with respect to the CO peak.
On the assumption of low optical depth, the 21-cm line integrated intensity distribution (Fig. 10d) readily yields the H I column density distribution. Very low in the inner disk ( 0.5 ), the atomic gas mass surface density increases sharply between 3 and 8 kpc, where it reaches a plateau at (H I ) 3 which extends from 8 to at least 12 kpc. At the radius of S2, the column density of H I is very low, and there is no evidence for this arm. Interestingly, the CO to H I velocity integrated intensity ratio (Fig. 10e) rises steadily toward the center at least down to R = 2kpc (the two bins at smaller radii have little statistical significance). It is noteworthy that according to this result - already suspected by Dame et al. (1993) from the CfA survey - the CO to H I ratio may depend on R in much the same way in M 31 as in the Milky Way, in spite of the much higher intensity of the CO emission in the central parts of the Milky Way relative to M 31.
The CO to 21-cm line intensity ratio can be converted into a molecular-to-atomic mass surface density ratio. The H I mass surface density (corrected for inclination) is readily derived from the observed intensity of the 21-cm line: (H I ) = 3.2 10-3 . The conversion factor between the CO integrated intensity and the column density of H2 on the other hand is so far poorly determined in M 31. While for the molecular ring, a conversion factor close to the Galactic one is suggested by the similarity between Carina and S4 presented above, by observations of individual molecular complexes (e.g. Vogel et al. 1987), and by millimetric continuum observations (Neininger et al. 1998), Loinard & Allen (1998) have proposed that the value of the conversion factor in the inner few kiloparsecs might be an order of magnitude higher. In the absence of a better estimate, we will use here the Galactic value N(H2)/ = 1.9 1020 cm-2 (K )-1 (Strong & Mattox 1996), which yields (H2) = 0.65 . Under these assumptions, (H2)/(H I ) = 0.20 103 /, and the maximum of the CO to H I ratio occurs at the galactocentric radius where the total gas mass surface density is of the order of 0.6 . In the Milky Way on the other hand, this maximum occurs at a radius where the total gas mass surface density is very high ( 6 ). In the Population I ring of M 31, the molecular gas contributes about 1.3 to the total gas mass surface density.
© European Southern Observatory (ESO) 1999
Online publication: November 16, 1999