Astron. Astrophys. 324, 877-887 (1997)

2. Global H I properties

Since the pioneering work by Roberts (1969), who investigated integral properties of a sample of about 100 galaxies, many studies have been published describing correlations between H I and optical properties (e.g. Shostak 1978; Huchtmeier & Richter 1988; RH). Several relations between H I content, morphological type and other global parameters were established. These studies involved large samples of galaxies (typically more than 100), observed with single-dish radio telescopes. On the other hand, synthesis observations of galaxies were made to compare the distributions of gas and light in detail (e.g. Bosma 1978, 1981; Wevers 1984; Broeils 1992). The sample sizes used in these synthesis studies are in general small, on the order of 10-20 galaxies.

The type of H I data analysed in this paper contains more spatial information than provided by single-dish observations, but less than provided by usual synthesis observations. Also the sample size (108 galaxies) lies between those of single-dish and synthesis studies. In this paper we therefore concentrate on properties that are special to these two surveys: radial H I surface density profiles, average surface densities and sizes of hydrogen discs.

The reader should keep in mind that certain correlations might be biased by the way the sample was defined. The details of the selection criteria can be found in Papers I and II, and their significance for the correlations are discussed in Sect  2.4. The properties of the sample are listed in Table 1 ¹.

Table 1. Properties of the sample galaxies.

Table 1. (continued)

Table 1. (continued)

2.1. H I and total mass

The traditional method to normalize and compare the H I content of galaxies is to use the distance-independent parameter . It has been shown that this ratio depends on morphological type (Roberts 1969; Shostak 1978; RH) in the sense that later types have larger ratios. For the galaxies in our sample this is also the case (see Fig. 1a), but the correlation is rather weak, maybe as a result of the inherent bias to this sample, that only gas-rich galaxies were selected (see Papers I and II). The average value and standard deviation of the ratio are . By analogy to the situation concerning the Tully-Fisher relation, one might expect a tighter correlation between gas content and morphological type by using infrared magnitudes, since the latter suffer less from absorption and reddening effects. Bothun (1984) has shown that the type dependence is somewhat stronger if one scales the gas content to magnitudes (i.e. the H magnitude measured within an aperture of one-third of the blue diameter, from Aaronson et al. 1982), but that within a given morphological bin the range in is as large as in the blue band. This is confirmed by our Figs. 1a and b. The average value and standard deviation of the ratio are .

Fig. 1a and b. Dependences on morphological type of the hydrogen-mass-to-B-luminosity ratio (a), the hydrogen-mass-to-H-luminosity ratio (b), the H I -to-total-mass ratio (c), and the total mass (d). In each plot the linear correlation coefficient r is given. The correlation with total mass is the strongest relation with morphological type for our sample. Comparison of panel (c) with (a) and (b) shows that the increase of gas fraction with type is more prominent when normalized by total mass than by luminosity of a galaxy

An alternative way to compare the H I contents of galaxies is to normalize the H I masses by the total masses. Assuming a spherical mass distribution, the total mass can be computed by . For V we take the maximum rotation velocity, (column 12 of Table 1), determined from the XV-maps (see Broeils 1992 and Paper II). Usually, the optical diameter (e.g. de Vaucouleurs or Holmberg diameter, see Faber & Gallagher 1979) is used to define , since the optical radius is approximately the outermost point where visible, galaxian material is detected. Because we can detect neutral hydrogen far beyond the optical boundaries, we use the H I diameter, defined at a surface density level of , to estimate the total enclosed mass. The mass thus derived can be considered as an upper limit to the mass enclosed inside the H I radius, because in most cases is somewhat larger than the rotation velocity of the last measured point of the rotation curve, and because a spherical mass distribution is used. Of course, the true total mass of a galaxy will even be larger, since there is no evidence that the outer edge of the mass distribution has been reached with the H I observations. In other words, the masses thus derived are the best determined lower total-mass limits presently available for an intermediate-size galaxy sample. In Fig. 1c the ratio of H I to total mass is plotted as a function of morphological type. It shows that late-type galaxies have much larger ratios than the classical spirals. This correlation is much stronger than the versus type relation shown in Fig. 1a, but its usefulness as "H I measuring tool" is limited by the fact that the ratio is distance-dependent. Not only the hydrogen-to-total-mass ratio correlates well with morphological type, but also the total mass itself is strongly type-dependent as is shown in Fig. 1d. With a correlation coefficient of it correlates more strongly with type than for example () or blue luminosity () do. The pronounced decrease for the dwarf systems might be somewhat exaggerated, due to the intrinsic difficulties of determining total masses for systems with narrow, Gaussian-shaped global H I profiles (Broeils 1992).

In Fig. 2 the plots of the total mass versus absolute blue magnitude and absolute near infrared magnitude are shown. The least-squares fits of these data, indicated by the solid lines, are

Fig. 2. Correlations between total enclosed mass inside the H I radius and the blue (a), and near-infrared (b) absolute magnitudes. The solid lines show the least-squares fit to all the data points; the dashed line indicates the relation for constant , which is consistent with the data for galaxies with

with and a dispersion of 0.81 mag for the blue band, and and 0.83 mag for the near infrared H band. For the blue band, the slope is considerably smaller than -2.5, the value found in earlier studies by e.g. Shostak (1978) and Huchtmeier & Richter (1988). A linear term with a slope of -2.5 in Eq.(1) would indicate that the total mass-to-light ratio of spirals would be constant. If we consider only galaxies with , then a constant of (), indicated by the dashed line in Fig. 2, is indeed consistent with the data points. The total-mass-to-light ratios of low-mass galaxies in our sample seem to be systematically smaller, possibly due to underestimations of the total masses for these galaxies as mentioned above. For the H band this effect is not visible, mainly because of the lack of H band data for dwarf galaxies. If a constant mass-to-light ratio is in any way related to the physics behind the existence of a Tully-Fisher relation (as is usually assumed), then Fig. 2 tells us that one should be careful with the inclusion of low-luminosity dwarf galaxies in the application of the Tully-Fisher relation.

2.2. H I diameters

There is a strong correlation between the (angular) H I diameter , defined at a surface density of and the optical absorption-corrected diameter , measured at the isophote:

with a linear correlation coefficient of , implying that the distance-independent diameter ratio is constant. The average value is . In Fig. 3, we plot this ratio versus morphological type (a) and luminosity (b). Fig. 3a shows no obvious correlation of the diameter ratio with type. The type dependence of the diameter ratio has been discussed before by Bosma (1981), Giovanelli & Haynes (1983), Wevers (1984) and Warmels (1986), amongst others, on the basis of smaller samples. More recently, Cayatte et al. (1994) constructed a sample of 84 galaxies with H I diameter information with types ranging from S0 to Im. (This sample has a considerable overlap with the sample analysed here, since it contains 39 out of the 48 galaxies in the sample of paper I.) These studies find a weak correlation of diameter ratio with type, in the sense that the average ratio is lower for early types. Given the width of the diameter ratio distribution per type and the relatively small number of Sa, Sab galaxies in our sample, we can not confirm these results. Fig. 3b indicates that a weak trend between the diameter ratio and absolute luminosity might exist: low luminosity galaxies seem to have slightly larger than more luminous galaxies.

Fig. 3. Ratio of H I and optical diameters, , as a function of morphological type (a) and luminosity (b). The crosses with error bars in panel (a) show the average values and dispersions obtained by averaging the data in bins indicated by the horizontal bars. The correlation coefficients are shown in the upper left corner of each panel. Two galaxies (NGC 2460 and NGC 1003) with greater than 3 have been excluded from calculation of average values

Fig. 3 further shows that for all galaxies the size of the neutral hydrogen disc is larger than that of the optical disc. Selection effects might play a role here, since there is a strong correlation between the linear H I diameter and H I mass, as shown in Fig. 4 by the filled circles (with this is actually the strongest correlation between two physically meaningful parameters for our sample); by selecting galaxies with sufficient H I flux, we have selected against galaxies with small H I diameters. As shown already by Shostak (1978), the H I mass also correlates strongly with the optical diameter with almost the same slope as that of versus . For our sample the relation between and is shown by the open circles in Fig. 4. Since both relations run parallel for our sample, a flux limited sample will always contain galaxies that have diameter ratios larger than 1, if the relations shown in Fig. 4 hold for all galaxies. The least-squares fits to these data, indicated by the two lines, are

Fig. 4. Correlation between H I mass and linear H I diameter (filled circles) and linear optical diameter (open circles). The open circles were shifted by 0.3 in the logarithm to the left-hand side of the diagram to separate them from the other points. The straight lines represent the results of least-squares fits, which both have slopes of about 2. This implies a nearly constant H I surface density when averaged over the whole disc. Note that the scatter for the H I diameters is significantly smaller than for the optical diameters

with and a dispersion of 0.13 dex for the , and and 0.19 dex for the relation. Use of (the radius enclosing 50% of the H I mass, see paper II) instead of gives basically the same result as and are tightly correlated (). It is interesting to note that the effective radius coincides with the optical radius (), which means that on average there are equal amounts of gas inside and outside the optical edge of a galaxy.

2.3. Mean H I surface density

The tight correlation between and , with a slope close to 2, implies a nearly constant mean H I surface density, averaged over the whole H I disc. This parameter, , is not correlated with either morphological type or luminosity, as shown in Fig. 5. Only galaxies of types Sa/Sab seem to have somewhat lower average surface densities. The small variation of the mean H I density among the late-type spirals and dwarf irregular systems has been demonstrated before (Bottinelli 1971, Huchtmeier & Richter 1988, and most clearly by RH) using the hybrid H I surface density, defined as the total H I mass divided by the optical surface area inside . The scatter of the distribution is however much smaller for than for the hybrid density . The mean values and standard deviations are and , respectively.

Fig. 5. True H I surface density, averaged over the whole H I disc, , as a function of morphological type (a) and luminosity (b). It shows a very weak correlation in the sense that late-type spirals have a slightly larger than early-types. The crosses indicate the average values calculated for the bins indicated by the horizontal bars

2.4. Selection biases

The above correlations might be biased by the way the galaxies were selected in the first place. The most important selection criteria were: (a) large H I contents, (b) optical diameters greater than , and (c) inclinations greater than . In particular criterion (a) has to be taken into account in judging the H I properties of this sample. In this section we discuss this by comparing our results with those from Roberts & Haynes (1994, RH), Kamphuis, Sijbring & van Albada (1996), and de Blok, McGaugh & van der Hulst (1996).

RH discussed the quantifiable properties of galaxies and their dependence on morphological type based on objects that are included both in the RC3 and in the Uppsala General Catalogue (Nilson 1973; UGC) with additional H I line data from the so-called Arecibo General Catalog (AGC) maintained by R. Giovanelli and M. Haynes. Comparison of Figs. 1 and 5a with those of RH show good agreement both in terms of absolute values and the dependence on morphological type. Note however that in at least three of the four panels in Fig. 1 the Sa/Sab galaxies do not seem to agree with the correlations found for the other types. They generally seem to have an H I content that is too high compared to the other types. RH showed that the distribution of H I masses and densities of Sa/Sab galaxies is substantially wider than those of late-type spirals. Our early-type spirals are clearly not representative for their type, but selected for their known high H I content.

Kamphuis et al. (1996) obtained short WSRT observations (similar to ours) for a sample of 57 galaxies without prior H I information in the RC3. They show that the ratios for the 42 detected galaxies agree well with the values obtained by RH (and therefore also with ours) with a slightly lower average value for the later types. de Blok et al. (1996) present VLA and WSRT H I data on 19 late-type low surface brightness (LSB) galaxies. They show that at a fixed luminosity these LSB have typically three times more H I than "normal" (HSB) galaxies, and that the H I surface density is about three times lower than those of HSB galaxies.

These comparisons indicate that by selecting galaxies with known H I properties we did not preferentially select the upper envelope of a wider distribution of H I masses for normal galaxies, but that the true distribution of H I content and densities might have a larger dispersion due to the properties of LSB galaxies. This does however apply to almost all published galaxy catalogs, since LSB galaxies are rarely included.

© European Southern Observatory (ESO) 1997

Online publication: May 5, 1998

helpdesk.link@springer.de