Astron. Astrophys. 331, 894-900 (1998)

5. Conclusions

Having derived the parameters that best describe the galaxy in all bands, we have created model images and compare them with the real observations. We do so in Fig. 3 where the 2D image of NGC 891 in the K, J, I, V and B bands (top to bottom) is presented, with the model in the left half of the frame and the folded real galaxy in the right half of each panel. In Fig. 4 we show the absolute value of the residuals between the observed galaxy and the model galaxy that we produced. These maps show how these residuals are distributed throughout the galaxy's image in terms of the absolute value of the percentage error. Red corresponds to areas with error less than , orange corresponds to areas with error between and , yellow corresponds to areas with error between and and black corresponds to areas with error greater than . White circles indicate the positions of the brightest foreground stars. In Table 3 we give the statistics derived from the residual maps shown in Fig. 4. In this table we give the percentage of the galaxy's image in each filter with residuals less than , , , and . We see that on average, almost of the total galaxy's image has residuals less than , while of the image is with residuals less than .

Fig. 3. Images of NGC 891 in K, J, I, V, B bands (top to bottom). The left half in each panel is the model image and the right half is the real galaxy image (folded).

Fig. 4. Colour map, showing the absolute relative error between the observed images and the model images in the K, J, I, V, B bands (top to bottom). See text for a detailed description.

Table 3. Percentage coverage of the residuals throughout the galaxy's image.

The goodness of the fit of our model to the real data is also seen in Fig. 5, where the histograms of the relative errors between the model galaxy image and the observed folded galaxy image are given for the five bands modelled. These histograms show a quite symmetrical distribution of the relative errors between positive and negative values. This means that the global fit that was done to the observed image of the galaxy accomplished the highest possible match between the smooth distribution given by the model and the real data with all the clumpiness and non-uniform structure that they have.

Fig. 5. Histograms of the relative errors between the real folded image of the galaxy and the model image.

An important result is that the dust is found to be extended in the radial direction, giving a scalelength which on average is 1.5 times larger than that of the stars. In the other (vertical) direction, the dust is concentrated in the central plane, with a scaleheight of about 1.5 times less than that of the stars. The value of 0.15 kpc obtained for the scaleheight of the young stars, is in good agreement with the scaleheight of the OB stars determined for our Galaxy (Wainscoat et al. 1992, Corradi et al. 1996) which is in the range of 0.09 - 0.11 kpc. In the K-band, we see a drop of the scaleheight of the stars to the value of 0.32 kpc, indicating a thinner stellar disk in these wavelengths, or maybe a departure from the exponential form. This is also evident in Fig. 4, where areas with high residuals are distributed symmetrically in the disk.

The filter that was used for the data modelled in KB87 was an IIIa-F emulsion filter with approximate range in wavelengths 0.58 - 0.69 µm. This puts it somewhere between the V and I passbands that we use in this study. Some of the values derived for the disk parameters in KB87 can be compared and are in good agreement with the mean values between V and I bands that have been derived in this study. For the scalelengths of the stars and the dust for example, where in both studies exponential radial distributions were used, we find a mean value of 5.1 kpc for the scalelength of the stars, while 4.9 kpc was found in KB87. For the scalelength of the dust, we found 7.5 kpc while in KB87 this parameter was not determined but it was estimated to be in the range 3.9 - 7.3 kpc. For the scaleheights of the stars and the dust though, we can not compare between the two studies since in this study we use an exponential distribution in z, while a sech² z law was used in KB87. For the central face-on optical depth, we found 0.60, while a value of 0.46 was found in KB87. This difference is mainly due to two reasons. First, the exponential distribution in the z-direction provides an upper limit to the central optical depth, because near the center of the galaxy this function is steeper than the sech² z law allowing for more dust. On the other hand, the bulge component that was added in this study, but was not taken into account in KB87, produced extra light and as a result more dust had to be added in order to absorb this extra light.

Dust and gas mass calculations for NGC 891 have been reported in Devereux & Young (1990). Assuming a distance of 14.1 Mpc for this galaxy (as opposed to 9.5 Mpc assumed by us) and using the IRAS and fluxes, the authors derive a dust mass of . For the same galaxy, the gas mass was found to be , resulting in a gas to dust mass ratio of 940. Using the results derived from our model (given in Table 1) and scaling them to the distance of 14.1 Mpc, we can calculate the total dust mass in the galaxy. The formulae that we have used are given in Sect. 5.3 of Paper I. With these calculations, the dust mass derived from the model is . Using this value, the gas to dust mass ratio becomes

which is very close to the value of 167 adopted for our Galaxy (Spitzer 1987, p.162) and also the value of 121 derived for UGC 2048 in Paper I. The dust mass derived from our model, is higher than the dust mass calculated using the IRAS fluxes. This is to be expected, because cold dust also exists, and has gone undetected by IRAS which only measures the warm dust. Our calculations are also supported by the 1.3 mm emission measured by IRAM (Guélin et al. 1993) which is nine times stronger than that predicted by the IRAS "warm dust" contribution.

From the face-on central optical depth of the dust and given a mean value for the scaleheight of the dust kpc for all the filters, we calculate the absorption coefficient in each band. We have found that , , and . These values are directly compared to the ratio of the extinction values and are plotted as a function of the effective wavelength in Fig. 6. Solid triangles correspond to the values derived with our model, while open circles are the values given by Rieke & Lebofsky (1985) for our Galaxy.

Fig. 6. The observed (open circles) values of for our Galaxy and the values calculated from the model (solid triangles) for NGC 891.

© European Southern Observatory (ESO) 1998

Online publication: March 3, 1998
helpdesk.link@springer.de