3 Interstellar extinction

This field FC-01863+00035 at $b=0.35\hbox{$^\circ$ }$ is close to the Galactic mid-plane where the interstellar extinction is serious. The measurement of interstellar extinction becomes then impossible in optical at large distances. Based on the optical observations of O-type to F-type stars, the ${A_{\it V}}$ value at the direction of $l=-18\hbox{$^\circ$ }$ and $b=-0.5\hbox{$^\circ$ }$which is close to the ISOGAL field, is estimated to be about 3 mag and approximately constant from a distance of $\sim$1 kpc up to $\sim$4 kpc (Neckel & Klare 1980). This may be a good reference for short distances although the patchy distribution of interstellar matter in the Galactic plane (Schultheis et al. 1999) may lead to some significant difference from the extinction in the ISOGAL field FC-01863+00035. However, the sources detected in the mid-infrared LW2 and LW3 bands extend to a much larger distance because of the much smaller extinction at these wavelengths than in optical.

Our calculation of interstellar extinction is based on two assumptions. One is that most of the objects in the ISOGAL field are luminous RGB stars or AGB stars with moderate mass loss rate. These objects are very bright with a luminosity of about one thousand to several thousand solar luminosity, and cold with an effective temperature lower than 4000 K or so. In addition, stellar winds from the photosphere have formed a circumstellar envelope for many of them. So they are strong emitters in mid-infrared. The analysis of the ISOGAL fields at (l=0.0$^\circ$, b=1.0$^\circ$) and in Baade's Window, has actually found most of the sources there are RGB and AGB stars whose colors are mildly reddened by interstellar extinction (Omont et al. 1999; Glass et al. 1999). The other assumption is that the intrinsic colors ( $J{-}K_{\rm S}$)₀ of all these sources are approximately similar as well as ($K_{\rm S}$-[7])₀, which is not true for the minority of objects with large mass-loss. According to the compilation of Wainscoat et al. (1992), the value of ( J-K)₀ of an M 0III star is about 1.0 and of an M 5 early AGB star is about 1.3 (see also van Loon et al. 2003 and references therein). In one ISOGAL field Sgr I that is inside the Baade's Window (Glass et al. 1999), the concentration of ( $J{-}K_{\rm S}$)₀and ($K_{\rm S}$-[7])₀ is very evident; for the objects with ($K_{\rm S}$-[7]) < 1.0, the average ( $J{-}K_{\rm S}$)₀ is 1.08 with standard deviation of 0.17 mag and the average ($K_{\rm S}$-[7])₀being 0.20 with standard deviation of 0.23 mag (with the revised photometry of the ISOGAL PSC, see Sect. 3.1) when the interstellar extinction is subtracted as 0.2 mag in the $K_{\rm S}$-band.

3.1 Extinction law at 7 and 15 micron

The assumption to start to extract the extinction law in mid-infrared is that the intrinsic color index ( $J{-}K_{\rm S}$)₀ of most stars in the field is the same so that their observed color $J{-}K_{\rm S}$ represents the interstellar extinction. Another presumption is the linear correlation between two color indexes if they are both originated from interstellar extinction, e.g. a linear correlation between the interstellar reddening at $J{-}K_{\rm S}$ and $K_{\rm S}$-[7], and at $J{-}K_{\rm S}$ and $K_{\rm S}$-[15]. For the first assumption, we had to remove some objects with apparently different intrinsic color. In addition to RGB stars and AGB stars with moderate mass loss, YSOs and AGB stars with high mass loss may be important members of the mid-infrared objects. Their cold and thick circumstellar envelopes absorb near-infrared radiation and radiate in mid-infrared strongly and make them easily detected by the ISOGAL survey. Their intrinsic color indexes ($K_{\rm S}$-[15])₀ and ($K_{\rm S}$-[7])₀ [and even ( $J{-}K_{\rm S}$)₀ for some of them] are redder than RGB stars or early AGB stars, and their color index [7]-[15] is also redder. Because the observed values of $J{-}K_{\rm S}$, $K_{\rm S}$-[7] and $K_{\rm S}$-[15] are significantly influenced by interstellar extinction, they can not be used as a criterion to pick up the late AGB stars or YSOs. On the other hand, the interstellar extinction in the LW2 and LW3 bands is much smaller, the color index [7]-[15] should be affected little by interstellar extinction. From the ISOGAL results in the Baade's Window (Glass et al. 1999) and from modeling (Ojha et al. 2003) we estimate that, for [7]-[15]<0.4, the effect of circumstellar dust of AGB stars is negligible on ($K_{\rm S}$-[7])₀, that it remains small ($\la$0.3) on ($K_{\rm S}$-[15])₀, and that it is similarly negligible on ( $J{-}K_{\rm S}$)₀ for [7]-[15]<1.0. Therefore we exclude the stars with ${\rm [7]{-}[15] > 0.4}$ or ${\rm [7]-[15]>1.0}$ in the discussion of the relation of the colors $K_{\rm S}$-[7] and $J{-}K_{\rm S}$, respectively, with interstellar extinction. Applying the same criteria for young stars also warrants that the effect of circumstellar dust is negligible on their intrinsic colors (see e.g. Bontemps et al. 2001).

If we set the intrinsic color ( $J{-}K_{\rm S}$)₀ of most ISOGAL stars as $C^{0}_{JK_{\rm S}}$ and of ($K_{\rm S}$-[7])₀ as $C^{0}_{K_{\rm S}7}$, and the observed color indexes as $C_{JK_{\rm S}}=J{-}K_{\rm S}$ and $C_{K_{\rm S}7}=K_{\rm S}$-[7] accordingly, the linear relationship is expected to be $C_{K_{\rm S}7}-C^{0}_{K_{\rm S}7}=k$ ( $C_{JK_{\rm S}}$- $C^{0}_{JK_{\rm S}})$. Based on model calculation (e.g. Bertelli et al. 1994) and the observation of the field Sgr I, the intrinsic color $C^{0}_{JK_{\rm S}}$ of RGB stars or early AGB stars is taken to be 1.2 on average, while the intrinsic color index $C^{0}_{K_{\rm S}7}$ of these stars is not well known (however, see van Loon et al. 2003). By leaving $C^{0}_{K_{\rm S}7}$ as a variable and assuming that the observed $C_{K_{\rm S}7}$ of sources with [7]-[15] < 0.4 is mainly caused by interstellar extinction, a robust linear fitting method, which minimizes absolute deviation and is insensitive to large departures for a small number of points, was used to fit the data and the result is displayed in the color-color diagram (Fig. 4, left). The resulted values for this linear fitting to the points decoded by pluses and triangles in Fig. 4 are: k=0.35 and $C^{0}_{K_{\rm S}7}=0.34$ (assuming $C^{0}_{JK_{\rm S}}=1.2$).

The processing of $K_{\rm S}$-[15] is similar to that of $K_{\rm S}$-[7]. The results are then k=0.39 and $C^{0}_{K_{\rm S}15}=0.39$(assuming $C^{0}_{JK_{\rm S}}=1.2$) (Fig. 4, right). In Fig. 4, the sources with [7]-[15]>0.4 are displayed by diamonds though they were abandoned during the fitting process to both the $K_{\rm S}$-[7] and $K_{\rm S}$-[15] with $J{-}K_{\rm S}$. Their large deviation to the right side from the fit line further proves that they are intrinsically redder. However, there could be a residual effect of circumstellar dust on the value of $K_{\rm S}$-[15] even for sources with [7]-[15]<0.4, especially for sources with large values of $J{-}K_{\rm S}$ which are more luminous; it could lead to an overestimation of k by maybe $\sim$10-20%.

For a blackbody radiation at temperature 4000 K typical of an M 0 RGB star, the Planck function yields $C^{0}_{K_{\rm S}7}=0.35$ and $C^{0}_{K_{\rm S}15}=0.44$ when the flux density for $K_{\rm S}=0.0$ is taken as $6.65\times10^{24}$ W/m²/Hz (Schuller et al. 2003), which is consistent with the fitting results $C^{0}_{K_{\rm S}7}=0.34$and $C^{0}_{K_{\rm S}15}=0.39$. Our values of $C^{0}_{K_{\rm S}7}$ and $C^{0}_{K_{\rm S}15}$are in reasonable agreement with those from Glass et al. (1999) in Baade's Windows when one takes into account the correction of -0.45 mag to be applied to the preliminary ISOGAL photometry used by Glass et al. (1999) (Schuller et al. 2003).

The linear coefficients between $C_{JK_{\rm S}}$ and $C_{K_{\rm S}7}$ or $C_{K_{\rm S}15}$ define the extinction values at 7 and 15 microns which depend on the extinction values in near infrared. With the extinction values in near infrared given by van de Hulst (1946), Glass (1999) and Rieke & Lebofsky (1989), the corresponding values at 7 and 15 microns are calculated and listed in Table 2 by adopting the coefficients k=0.35 and k=0.39 between $C_{JK_{\rm S}}$ and $C_{K_{\rm S}7}$, and $C_{JK_{\rm S}}$ and $C_{K_{\rm S}15}$, respectively.

While the extinction values in the near-infrared from Rieke & Lebofsky (1989) are higher, the ones preferred by Glass (1999) are just an update of those of van de Hulst (1946) and thus the two latter sets of values are quite close to each other. Similarly, the extinction values at 7 and 15 $\mu$m derived from Rieke & Lebofsky $(A_{7} \approx A_{15} \approx 0.05)$ are higher than those inferred from Glass-van de Hulst, i.e. $A_{7}\approx0.03 A_V$ and $A_{15}\approx0.025$A_V (see Table 2). As generally in other ISOGAL papers, we prefer Glass' values for the reasons exposed in Glass (1999), except possibly for the fields close to the Galactic Center and in star forming regions. To compare the values we have derived for the mid-infrared extinction with previous ones, we will distinguish the case of 7 $\mu$m where the fitting of Fig. 4 (left) is very robust, from the one of 15 $\mu$m which is more uncertain. In both cases one should note that the averages of the extinction on the wavelength range of the LW2 and LW3 broad bands yield slightly larger values than at 7 and 15 $\mu$m, respectively, which are both close to a minimum of the extinction curve.

At 7 $\mu$m the most widely used reference for the extinction value is probably Mathis (1990) which is similar to Draine & Lee (1984). They used similar extinction law in near infrared to Rieke & Lebofsky (1989), and estimated the infrared extinction from an extrapolation of the optical extinction law and near-infrared observational data, yielding $A_7\sim 0.020~A_{\it V}$. This value is slightly lower than the one we derived, 0.03 ${A_{\it V}}$. Such a difference may arise from the uncertainty in the ISOGAL photometry or in the fitting procedure in Fig. 4. It is probably included as well within the uncertainty of the extrapolation by Mathis (1990). We will use our value, 0.03 ${A_{\it V}}$, in the following of this paper, keeping in view its uncertainty. Let however note that the higher value we could derive from the near-infrared extinction law of Rieke & Lebofsky (1989) (Table 2), is also consistent with the results derived by Lutz (1999) ( $A_{7} \approx 0.045$) who used the hydrogen recombination lines and the same near-infrared extinction law towards the Galactic Center.

Figure 4: Color-color diagram of $J{-}K_{\rm S}$ vs. $K_{\rm S}$-[7] and $J{-}K_{\rm S}$vs. $K_{\rm S}$-[15]. The sources detected in the LW3 band are divided into two groups, one with ${\rm [7]{-}[15] < 0.4}$ decoded by pluses which were used for linear fitting and the other with ${\rm [7]{-}[15] > 0.4}$decoded by diamonds which were not used for linear fitting. In the left graph, the sources which are not detected in the LW3 band but are used in linear fitting are decoded by triangles. The robust linear fitting result is shown by solid line. The horizontal dot line labels where $J{-}K_{\rm S}=1.2$ which is assumed intrinsic color $J{-}K_{\rm S}$ for the sources used in fitting. In the right graph, a dash line represents the result of combining the 7 to 15 $\mu$m opacity ratio from Hennebelle et al. (2001) with the extinction values of J, $K_{\rm S}$ and 7 $\mu$m in Col. 2 of Table 2 (i.e. A15=0.044).

The extinction ratio A₇/A₁₅ is given as $\sim$1.3 by Mathis (1990). However, the value of Draine & Lee (1984) is almost twice smaller. The best value for this ratio has probably been derived by Hennebelle et al. (2001) from an analysis of infrared dark clouds from the ISOGAL survey. They give $0.7\pm0.1$ for the clouds away from the Galactic Center, which is similar to Draine & Lee (1984). The value we derived for this ratio in Table 2, Col. 2, 1.2, is more compatible with Mathis (1990) than with Hennebelle et al. (2001). In Fig. 4 (right), the 15 $\mu$m value deduced from the Hennebelle et al. ratio is shown by a dashed line derived from the extinction values at J, $K_{\rm S}$ and 7 $\mu$m taken from Col. 2 of Table 2. It is evident that their result implies more extinction at 15 $\mu$m. However, let us stress that our direct fitting is somewhat uncertain at 15 $\mu$m because of the smaller number of sources and of the possibility of residual effects of circumstellar reddening on $K_{\rm S}$-[15], especially for the distant luminous sources with large extinction. Therefore, we consider that the 15 $\mu$m value deduced from Hennebelle et al. (2001) (dashed line in Fig. 4 (right)) is still compatible with the data of our fitting in Fig. 4 (right, full line), and that our data cannot provide a really accurate value of the extinction at 15 $\mu$m. Nevertheless, we will use the values derived from our fitting in the subsequent sections, since such differences on the small extinction at 15 $\mu$m are practically negligible in the following discussions. One has to keep in mind that the extinction law may vary with the directions due to the inhomogeneous distribution of the interstellar matter in the Galactic plane.

 

 

Table 2: Extinction values in the LW2 (7  $\mu$m) and LW3 (15 $\mu$m) bands.

Band	$A_{\lambda}/A_{v}$
	vdH1	ISG2	R & L3
J	0.245	0.256	0.281
$K_{\rm S}$	0.087	0.089	0.112
7	0.032	0.031	0.0534
15	0.026	0.025	0.046

¹ Values for J and $K_{\rm S}$ derived from van de Hulst (1946) curve.
² Values for J and $K_{\rm S}$ derived from Glass (1999) (update of van de Hulst's values).
³ Values for J and $K_{\rm S}$ derived by Rieke & Lebofsky (1989) for stars towards the Galactic center.
⁴ 7 and 15 $\mu$m extinction values are then derived from the slopes k of the straight lines of Fig. 4
 (see text): $A_{K_{\rm S}}$- A₇=0.35(A_J- $A_{K_{\rm S}}$),  $A_{K_{\rm S}}$- A₁₅=0.39(A_J- $A_{K_{\rm S}}$).

3.2 Extinction structure along the line of sight

By assuming the intrinsic color index $C^{0}_{JK_{\rm S}}$ is the same for all the objects which were detected in both J and $K_{\rm S}$ bands and with [7]-[15]<1.0 if detected in both LW2 and LW3 bands as well, the interstellar extinction to individual objects can then be calculated from the observed $C_{JK_{\rm S}}$. Theoretical calculation shows that $C^{0}_{JK_{\rm S}}$ doesn't differ much for late-type RGB stars and early AGB stars which are the major components of the ISOGAL sources. Though there may be some foreground main-sequence stars and early-type AGB stars with intrinsic bluer color than late-type RGB stars that would bring about the underestimation of ${A_{\it V}}$, the number of such sources should be small and they should be mostly at small distance to be detectable by ISOGAL as seen in Fig. 4 for those with $J{-}K_{\rm S}$ < 1.0. While many of the ISOGAL sources are detected in the $K_{\rm S}$-band and not detected in the J-band, the value of ${A_{\it V}}$ may then be inferred from $C_{K_{\rm S}7}$ according to the extinction law derived above. This method is right only if the non-detection in the J-band is caused by large interstellar extinction. Because some of the non-detections in the J-band may come from serious absorption by circumstellar dust of AGB stars or YSOs, the estimation from $C_{K_{\rm S}7}$ may overestimate the real interstellar extinction, in particular when the index $C_{K_{\rm S}7}$ is large, e.g. larger than 2 that results in $A_{V}\sim30$. So when inferring the extinction value from $K_{\rm S}$-[7], we have excluded the sources with [7]-[15]>0.4 because of the risk they present the occurrence of large mass loss. For the same reason, we have excluded the sources with [7]-[15]>1.0 when inferring the extinction value from $J{-}K_{\rm S}$. However, we have thus altogether discarded a large proportion of sources with the largest extinction.

We adopted the value of intrinsic color index $C^{0}_{JK_{\rm S}}$ as 1.2 and the extinction values listed in Col. 2 of Table 2. The global distribution of ${A_{\it V}}$ is shown in Fig. 5d, where values of ${A_{\it V}}$ from $J{-}K_{\rm S}$ or $K_{\rm S}$-[7] are distinguished by the dash or dot lines respectively, while the summation of the two types is represented by the solid line. It can be seen, as expected, that the sources not detected in the J-band experience higher extinction than those detected in the J-band on average. All the sources not detected in the J-band have ${A_{\it V} > 10}$ and peak at about 28. In Fig. 5, we have added to the bin $A_{\it V}=0{-}2$all the sources with $J{-}K_{\rm S}<1.2$. Such sources should not be AGB or RGB-tip stars, but nearby earlier stars, mostly K giants with a few A-B stars. Most of them must have A_V<2-3. We have checked that the corresponding values of I-J are consistent with J-K for all these 19 sources, except for one where there is a problem with the DENIS I-J associations. As expected, all of them are associated with a visible GSC2-2 star. The combined visible/near-infrared colors are consistent with K-early or M giants for most of them, with, however, an identified A2 star.

In addition to the concentration at $A_{V}\sim 0$, the sources are distributed along ${A_{\it V}}$ unevenly. After a dip at ${A_{V}}\approx 4{-}6$, there is first a progressive increase of their number up to ${A_{\it V}} \approx 12$, and then a steeper rise until a relatively sharp maximum at ${A_{\it V}}\sim 14{-}16$. After that, their number decreases rather regularly up to ${A_{\it V}}\sim 50$.

We think that the uneven distribution of ${A_{\it V}}$ reflects the inhomogeneities in the distribution of the interstellar medium, partly along the line of sight with the crossing of the molecular ring and of several arms, but probably mainly perpendicular to the line of sight across the observed ISOGAL field. In order to see the influence of the spiral arms, we estimated the extinction values to the arms from the distribution of the emission in the radio lines of CO and HI. The kinematical distance to the interstellar clouds can be inferred from their radial velocity. The line of sight in the direction l=-18.63$^\circ$  and b=0.35$^\circ$  touches the outer edges of the Sagittarius, the Centaurus and the Norma arm at respectively 1, 4, and 6 kpc, runs through the bulge and then reaches the far side of the Norma arm at about 13 kpc. Indeed, most of the line of sight, between 3 to 13 kpc, is in the molecular ring, including the tangential point at $\sim$8 kpc ( $v\sim -130$ km s^-1).

As for the whole Galactic disk, there exist ¹²CO data for our field taken from the Milky Way survey at the coarse resolution of 0.125$^\circ$ (7.5$^\prime$) (Dame et al. 2001). Our $0.35\hbox{$^\circ$ }\times0.29\hbox{$^\circ$ }$  field thus implies nine pointings of this survey at $l=18.625\hbox{$^\circ$ }~\pm~0.125$$^\circ$ and $b=0.375\hbox{$^\circ$ }\pm0.125\hbox{$^\circ$ }$. A rapid look at these CO emission data shows that it corresponds mainly to radial velocities characteristic of the molecular rings and more precisely, of the Centaurus and Norma arms. However, there is an ambiguity for the emission between -70--80 km s^-1 which could be attributed either to the near side or to the far side of Norma. Especially in the Centaurus arm, the CO emission displays a strong gradient with respect to b across this field (Table 3). In order to take into account this gradient in the discussion of the distribution of A_V, we will consider separately the three b ranges of the pointings of the CO survey (Table 3). For each b range of the pointings we computed the CO integrated intensity, W$_{\rm CO}$, averaged over the three lpointings, for each velocity interval roughly corresponding to the different spiral arms (Table 3). Then we estimated the corresponding A_V shown in Table 3, adding a contribution from the HI regions from the HI survey of Bloemen et al. (1990).

Figure 5: Histograms of the interstellar extinction ${A_{\it V}}$, which are derived either from $J{-}K_{\rm S}$ (dot line) or from $K_{\rm S}$-[7] (dash line) when no data in the J-band are available (sources with [7]-[15] >0.4 are excluded), and from the addition of these two types (solid line). Figures  a), b) and c) display the sources with b in the quoted ranges corresponding to the overlap of the ISOGAL field with the pointings of the CO survey data by Dame et al. (2001). Figure d) shows the extinction distribution of the sources spanning the whole b range in this field. $A_{V\rm c}$ represents the average value of the accumulated extinction expected from the interstellar gas just beyond the Centaurus arm (last line of Table 3); $A_{V\rm c1}$ represents the starting value of the extinction by the Centaurus arm as deduced from the sharp rise of the ${A_{\it V}}$ histogram; and $A_{V\rm c2}$ is the symmetric value of $A_{V\rm c1}$ with respect to $A_{V\rm c}$, which could correspond to the ending value of the extinction by the arm. The same method cannot be applied to subfield "c'' because the ISOGAL range is about half, much smaller than the total range of the CO pointing; $A_{V\rm c'}$ is instead defined as the average value of the peak of the histogram.

We estimated ${A_{\it V}}$ shown in Table 3 by adopting the conversion factor ${1.8 \times 10^{20}~\rm cm^{-2}~(K~km~s^{-1})^{-1} }$ from the CO integrated line intensity W$_{\rm CO}$ to H₂ column density (Dame et al. 2001) and the factor ${\rm 10^{21} ~molecules~ cm^{-2}~mag^{-1} }$ from H₂column density to ${A_{\it V}}$. This estimation, in particular of the HI column density, suffers some uncertainty from integrating the velocity on unclear contours of the paper that could be about 30%. The W$_{\rm CO}$ to N$_{\rm H2}$ conversion factor is also known to be rather uncertain (Bachiller & Cernicharo 1986; Harjunpaa & Mattila 1996).

Similarly, we split our sample of ISOGAL sources in three unequal parts corresponding to these b ranges (Table 3). For each subsample, we have represented in Fig. 5 the histogram of the distribution of A_V. We note, as expected, important differences in the total ranges of these three distributions reflecting the gradient of the extinction with b.

In view of discussing the correspondence between the ${A_{\it V}}$distribution from the ISOGAL sources and the determination from the interstellar gas, let us stress the difficulty that any spatial inhomogeneity in the CO intensity smaller than the large CO beam is smoothed with the present CO data. Therefore, it is impossible to estimate the actual spatial dispersion of ${A_{\it V}}$ within a spiral arm. It is certainly significantly larger than the dispersion, $\sim$20%, between the three l pointings for the same b range. In each of the three histograms of Fig. 5 the average value of the accumulated extinction expected from the interstellar gas just beyond the Centaurus arm is, $A_{V\rm c}=23$, 18 and 12, respectively (Table 3). Looking at the values of the sharp rise of the ${A_{\it V}}$ distribution due the Centaurus arm, with $A_{V\rm c1}=12$, 14 and 10, respectively, one sees that these are 20-50% lower than $A_{V\rm c}$. A natural interpretation is that this difference essentially represents the dispersion of the extinction through the Centaurus arm. The case of the subfield with b>0.4375$^\circ$  is special because the b sample observed by ISOGAL, 0.4375$^\circ$ <b< 0.495$^\circ$  is about half the total range of the CO pointing, 0.4375$^\circ$ <b< 0.5625$^\circ$. Because of the strong gradient with b, it is likely that the average extinction in the region observed by ISOGAL is larger than the one in the total CO beam which yields $A_{V\rm c}=12$ (see Table 3). Indeed the main feature of Fig. 5c is approximately symmetric with respect to $A_{V\rm c'}=14$, which should be close to the actual value of the average of ${A_{\it V}}$ in the region observed by ISOGAL. However, it is unclear in this case whether the main contribution of Norma is absent because it comes from the far side with very few sources behind it, or it is included in the pedestal between $A_{\it V}=18$ and $A_{\rm V}=22$, together with the extinction of a few sources behind the tangential point region.

The situation is less clear for the other two subsamples with smaller b and larger extinction. It is true that the distribution again extends beyond the value $A_{V\rm c2}$, symmetric of $A_{V\rm c1}$, with respect to $A_{V\rm c}$, as expected from such a dispersion of the ${A_{\it V}}$distribution through the Centaurus clouds and the additional extinction beyond Centaurus. But there is a large decline of the number of the sources towards large ${A_{\it V}}$, making the distribution very asymmetric with respect to $A_{V\rm c}$. The loss of sources may be explained by the very large value of ${A_{\it V}}$ which prevents the detection at 7 $\mu$m or even in the $K_{\rm S}$-band of many sources.

Finally, we note that for the three subsamples, the total range of values of ${A_{\it V}}$ inferred from the ISOGAL sources is consistent with the maximum accumulated extinction along the line of sight expected from the interstellar gas, including the dispersion of A_V values within clouds, the Norma ambiguity and the loss of sources with very large ${A_{\it V}}$.

Despite the many uncertainties in the relation between W$_{\rm CO}$and the ISOGAL colours and in the interpretation of the ${A_{\it V}}$distribution, one feature remains intriguing. Before the sharp rise at $A_{V}\sim 10$-14 that we associate with the Centaurus arm, there is a distribution of a smaller number of sources from $A_{V}\sim6$ to $\sim$12. The local and Sagittarius arm emission of CO is quite unable to account for such extinctions. The only explanation is that such sources are located within the Centaurus arm, behind the first dust layers. The subsequent very sharp rise suggests a narrow distribution of the bulk of the Centaurus extinction. The appreciable number of sources with $A_{V}\sim6$ to $\sim$12 suggests that the first Centaurus layers of dust are distributed over a rather large distance with a substantial source density. This implies a large width for the Centaurus "arm'', which is consistent with the broad CO velocity distribution, and/or an over-density of ISOGAL sources inside it, which could be mostly young/massive K giants.