Annu. Rev. Astron. Astrophys. 1997. 35: 357-88 Copyright © 1997 by Annual Reviews. All rights reserved

5. STRUCTURE AND MORPHOLOGY

The spatial distribution and luminosities of the member galaxies provides further clues to the nature of compact groups. If they are primarily projections or transient configurations, the luminosity function should be the same as that of the parent systems, and the spatial distribution of the galaxies should be consistent with a random distribution. If they are bound physical systems, the luminosity function and spatial distribution might show features that reflect the origin or subsequent evolution of compact groups.

5.1 Shapes and orientations

The shapes of compact groups were first investigated by Arp (1973) who concluded that galaxy "chains" were unusually predominant. However, Rose (1977) determined that the ellipticities of his groups were consistent with a random distribution of galaxies. Using the larger HCG sample, Hickson et al. (1984) and Malykh & Orlov (1986) reached the same conclusion as Arp - the groups are typically more elongated than would be a random distribution of galaxies. An immediate consequence of this result is that compact groups cannot easily be explained as random projections or chance crossings, as this would largely erase any inherent ellipticity of a parent loose group. From static simulations, Hickson et al. (1984) concluded that the observed ellipticities are best matched by three-dimensional shapes that are intrinsically prolate. The same result was found by Oleak et al. (1995) in a recent study of the shapes of 95 Shakhbazian compact groups. These conclusions, however, are not unique. Hickson et al. (1984) also found the shapes to be consistent with those seen in dynamical simulations of compact groups seen in projection as subgroups within loose groups. In addition, one must always be concerned about possible selection biases. It may be that highly elongated groups (such as VV 172) are more easily noticed in visual searches. It will be interesting to see if these results are confirmed by studies of groups found by automated searches.

If the intrinsic shapes of compact groups are related to their formation process, one might expect to see a relationship between the orientation angle of a group and the environment. Palumbo et al. (1993) examined the environments of the HCGs, and found that the orientations of the major axes of the groups were consistent with an isotropic distribution.

If compact groups are not simply projection effects, they might be expected to show a centrally-concentrated surface density profile, as is seen in clusters of galaxies. Although the number of galaxies in individual compact groups is small, with a large sample it is possible to estimate a mean profile. By scaling and superimposing the HCGs, Hickson et al. (1984) found evidence for central concentration. Mendes de Oliveira & Girard (1994), using a similar analysis, concluded that the mean surface density profile is consistent with a King (1962) model with typical core radius of 15h^-1 kpc. Most recently, Montoya et al. (1996) have analyzed the profiles of the 42 HCG quartets which have accordant redshifts. Their technique uses the distribution of projected pair separations and thus avoids assumptions about the location of the group center. They find a smaller core radius (6h^-1 kpc for a King model). The fact that Montoya et al. (1996) find a consistent density profile for all groups, without any scaling, is particularly interesting. This would not be expected if most groups are chance alignments within loose groups. It also implies that compact groups have a unique scale, which seems counter to the concept of hierarchical clustering. Montoya et al. (1996) suggest that this arises as a result of a minimum mass density and velocity dispersion that is required for the groups to be virialized (Mamon 1994).

5.2 Compact-group galaxies

There have been several studies of the morphological types of galaxies in compact groups (Hickson 1982, Williams & Rood 1987, Sulentic 1987, Hickson et al. 1988b). Most studies agree that the fraction f_s of late type galaxies is significantly less in compact groups than in the field. Hickson et al. (1988b) find f_s = 0.49 for the HCGs; Prandoni et al. (1994) obtain f_s = 0.59 for the SCGs. Both these values are substantially lower than those found for field galaxy samples (f_s 0.82, Gisler 1980, Nilson 1973).

Also well established is morphological type concordance, observed in both the HCGs and SCGs (Sulentic 1987, Hickson et al. 1988b, Prandoni et al. 1994). A given compact group is more likely to contain galaxies of a similar type (early or late) than would be expected for a random distribution. White (1990) has pointed out that such concordance could result from a correlation of morphological type with some other property of the group. The strongest such correlation found to date is between morphological type and velocity dispersion (Hickson et al. 1988b). As Figure 3 shows, groups with higher velocity dispersions contain fewer late-type (gas-rich) galaxies. They also tend to be more luminous. The importance of velocity dispersion, in addition to local density, on the galaxy morphology had previously been emphasized by De Souza et al. (1982). A crucial clue is that the morphology-density relation seen in clusters and loose groups (Dressler 1980, Postman & Geller 1984, Whitmore & Gilmore 1991) is not the dominant correlation in compact groups (Hickson et al. 1988b), although some effect is present (Mamon 1986). This suggests that the velocity dispersion is more fundamental, ie of greater physical relevance to the formation and evolution of galaxies in compact groups, than is apparent physical density.

Figure 3. Morphology-velocity correlation for compact groups. The figure shows the cumulative distributions of velocity dispersion for spiral rich (fs > 0.5) and spiral-poor (fs < 0.5) groups. The former have typically half the velocity dispersion than the latter and a broader velocity range.

There is much evidence that interaction is occurring in a large fraction of galaxies in compact groups. The strongest direct support comes from kinematical studies. Rubin et al. (1991) found that two thirds of the 32 HCG spiral galaxies that they observed have peculiar rotation curves. These show asymmetry, irregularity and in some cases extreme distortion, characteristic of strong gravitational interaction. This result has recently been challenged by Mendes de Oliveira et al. (preprint) who obtained H velocity maps for 26 HCG spiral galaxies and found that only one third showed abnormal rotation curves. They suggest that the difference is due to the more-complete spatial sampling of their data.

In their study, Rubin et al. (1991) observed 12 HCG elliptical galaxies and detected nuclear emission in 11 of them. This high fraction suggests that interactions and mergers may be supplying gas to these galaxies. This idea received independent support from radio observations in which neutral hydrogen emission was detected in three compact groups which contain only elliptical galaxies (Huchtmeier 1994).

Zepf & Whitmore (1993) found that elliptical galaxies in compact groups tend to have lower internal velocity dispersions than do ellipticals in other environments having the same effective radii, absolute magnitudes and colors. They therefore do not lie on the fundamental plane defined by other elliptical galaxies. This discrepancy correlates with isophote shape in that those galaxies that have "disky" or irregular isophotes tend to have lower velocity dispersion. Both Zepf & Whitmore (1993) and Bettoni & Fasano (1993, 1995, 1996, Fasano & Bettoni 1994) report that HCG elliptical galaxies are less likely to have "boxy" isophotes, and more likely to have irregular isophotes. Such effects are consistent with results of simulations of tidal encounters (Balcells & Quinn 1990).

5.3 Optical luminosity function

The luminosity function of compact groups was first estimated by Heiligman & Turner (1980). They examined a sample consisting of Stefan's Quintet, Seyfert's Sextet and eight more compact groups from the Arp and Vorontsov-Velyaminov catalogs, and concluded that compact groups contain relatively fewer faint galaxies than does a comparable field galaxy sample. Analysis of the relative luminosities within individual HCGs (Hickson 1982), and studies of several Shakhbazian groups (Kodaira et al. 1991), showed a similar effect, although Tikhonov (1987) found a luminosity function similar to that of field and cluster galaxies.

The larger HCG sample allows the question of the galaxy content of compact groups to be addressed with greater certainty. The standard technique for determination of the luminosity function weights each galaxy by V_m / V, where V is the volume of the smallest geocentric sphere containing the group, and V_m is the volume of the largest such sphere within which the group could have been detected. Using this approach Sulentic and Rabaça (1994) obtained a luminosity function for HCG galaxies similar to that of field galaxies. However, Mendes de Oliveira and Hickson (1991) argued that the standard calculation does not address the selection effects of the HCG sample. For example, the luminosity range within an individual group is limited by the 3-mag range of the selection criteria. Because of this, fainter galaxies within compact groups are not included in the catalog. In order to account for such biases they used a modeling technique in which galaxies were drawn from a trial luminosity function and assigned to groups. Redshifts were given to each group according to the observed distribution, and groups that failed to meet the HCG selection criteria were rejected. The luminosity distribution of the resulting galaxy sample was then compared to the observations and the process repeated with different trial luminosity functions. Their best-fit luminosity function is deficient in faint galaxies, although a normal field-galaxy luminosity function is not excluded.

To avoid the selection problem, Ribeiro et al. (1994) obtained deeper photometry for a subsample of the HCGs in order to include the fainter galaxies explicitly. Since redshifts are not known for these galaxies, a correction for background contamination was made statistically. The luminosity function that they obtained is similar to that of field galaxies. Most recently, the luminosity function for the RSCGs has been computed by Barton et al. (1996). They find it to be mildly inconsistent with that of field-galaxies, in the same sense as that of MH for the HCGs. Figure 4 summarizes these estimates of the luminosity function of compact galaxies in terms of the Schechter (1976) parameters M* and .

Figure 4. Optical luminosity function parameters of compact groups. Triangle: Mendes de Oliveira & Hickson (1991), open square: Sulentic & Rabaça (1994), cross: Ribeiro et al. (1994), open circle: mean of the three RSCG samples (Barton et al. 1996). For comparison, the filled circle and square indicate luminosity function parameters for galaxies in the CFA-Combined (Marzke et al. 1994) and SSRS2 (da Costa et al. 1994) surveys respectively, representing galaxies in lower-density environments.

How do we interpret these apparently conflicting results? Prandoni et al. (1994) have argued that the HCG catalog is biased toward groups with a small magnitude range m, because the SCGs have a larger fraction of high m groups. However, it is not known what fraction of such groups are physically real, as few redshifts have yet been obtained. Such a bias could affect the luminosity function of Mendes de Oliveira & Hickson (1991), particularly at the faint end, but it is not evident that the bias is sufficient to account for the apparent faint-galaxy deficit. On the other hand, the small sample used by Ribeiro et al. (1994) may not be representative of compact groups in general. Hickson (1997) points out that the Ribeiro et al. (1994) sample has a spiral fraction of 0.60, substantially higher than that of the whole HCG catalogue, and contains 7 of the 16 HCGs found to be in a high density environment by Palumbo et al. (1995). This suggests that the their sample has more than the usual amount of field galaxy contamination. De Carvalho et al. (1994), note that the faint galaxies form a more extended distribution than do the brighter galaxies. Thus they may be a dynamically distinct component, or simply unrelated field galaxies. Finally, the redshift-selected RSCG sample also shows mild evidence for a faint galaxy deficiency.

While the nature of the faint end of the LF in compact groups appears to be depleted, there is evidence that the bright end may be enhanced. Limber and Matthews (1960) were the first to remark that "the members of Stefan's Quintet are to be classed among the brightest of galaxies". It is possible that this may be in part due to interaction-induced star formation, at least for the spiral galaxies. On the other hand Mendes de Oliveira & Hickson (1991) compared their luminosity function of elliptical galaxies in compact groups with those in the Virgo and Coma Clusters (as reported by Sandage et al. 1985 and Thompson & Gregory 1980) and found that that the compact-group ellipticals have a luminosity enhancement of more than 1 mag compared to cluster ellipticals. Sulentic and Rabaça (1994) find a similar enhancement in their morphological-type-specific luminosity function. This suggests that compact group elliptical galaxies may have a unique formation mechanism.

From the luminosity function, one can estimate the contribution of compact groups of galaxies _CG to the total galaxian luminosity density . MH obtained a ratio of _CG / 0.8%. Applying the same analysis to the luminosity function of Ribeiro et al. (1994) gives a ratio of 3.3%. For the RSCGs, the figure is comparable: for groups of four or more galaxies, Barton et al. (1996) obtain a compact group abundance of 1.4 x 10^-4 h^-3 Mpc^-1 which leads to a luminosity density ratio of approximately 3%. These are surprisingly high figures considering the short dynamical times of most compact groups.

5.4 Star formation and nuclear activity

Evidence is accumulating that tidal interactions play an important role in triggering starburst activity in galaxies (eg. Maccagni et al. 1990, Campos-Aguilar & Moles 1991, Kormendy & Sanders 1992, Sanders & Mirabel 1996). Compact groups, with their high galaxy density and evident signs of galaxy interaction should be ideal systems in which to study such effects. Many HCGs do in fact contain galaxies showing starbursts or harboring active galactic nuclei (AGN). For example, HCG 16 is found to contain a Seyfert 2 galaxy, two LINERs, and three starburst galaxies (Ribeiro et al. 1996). HCG 31 contains five galaxies showing signs of recent starburst activity (Rubin et al. 1990, Iglesias-Páramo & Vílchez preprint). Seyfert galaxies are also found in HCG 77, 92, 93 and 96.

The general degree of star formation activity in compact group galaxies can be determined from infrared observations. To date, studies have been based primarily on data from the IRAS satellite. Hickson et al. (1989) found sources in 40 HCG from a search of the Point Source Catalog. They concluded that the ratio of far-infrared-to-optical luminosity is greater by about a factor of two in compact group galaxies, compared to that of isolated galaxies. This result was disputed by Sulentic and De Mello Rabaça (1993) who argued that the low spatial resolution of the data made the assignment of infrared flux to individual galaxies ambiguous. They concluded that redistribution of the flux could result in little or no infrared enhancement, a conclusion echoed by . However, in cases of doubt, Hickson et al. (1989) identified the infrared galaxy on the basis of radio emission. The well-known correlation between infrared and radio continuum emission makes it unlikely that the results are much in error. Analysis of improved data (eg. Allam et al. 1996) should soon resolve questions about the identifications and infrared fluxes.

The resolution problem can be avoided by considering the infrared colors of the sources instead of the infrared/optical ratio. Zepf (1993) compared the ratio of 60 µm to 100 µm fluxes of compact group galaxies with those of isolated galaxies and also with those of galaxies believed to be currently merging. He found that the compact group sample was significantly different from both other samples, and estimated that about two thirds of the compact group galaxies had warm colors (larger 60/100 µm ratios) similar to those of merging galaxies.

Another approach to interpreting the infrared results was taken by Menon (1991) who emphasized that the strong correlation between radio and infrared radiation indicates that these likely originate from a common region. In compact group spirals the radio emission is primarily nuclear whereas in isolated spirals it originates in the disk. If this is also true for the infrared flux, there must be an enhancement of the infrared/optical ratio, in the nuclear region, of more than an order of magnitude. This idea is supported by recent millimeter-wavelength observations (Menon et al. 1996) in which CO emission is detected in 55 of 70 IRAS-selected HCG galaxies. The inferred ratio of infrared luminosity-to-H₂ mass showed an enhancement which correlates with the projected nearest-neighbor distance.

Further clues are provided by radio continuum studies. Nonthermal emission from spiral galaxies can arise from both disk and nuclear sources. Disk emission is predominantly due to supernova remnants and is thus related to the star formation rate. Nuclear emission can arise both from star formation and from an active nucleus. Menon (1995a) observed 133 spiral galaxies in 68 HCG, and found that overall they typically show less continuum emission than those in isolated environments, which is consistent with the neutral hydrogen observations. However, when considering the nuclear regions alone, the radio emission is found to be an order of magnitude higher compared to isolated spirals. The implication is that star formation and/or AGN activity is substantially enhanced in the nuclear regions of many compact group spiral galaxies. This is generally consistent with a picture in which galaxy interactions remove gas from the outer regions of galaxies, while simultaneously allowing gas to flow inwards toward the nucleus, resulting in enhanced star formation in the nuclear region, and possibly fueling an active nucleus.

Although there is a clear example of tidal interaction stimulating disk radio emission in at least one compact group (Menon 1995a), statistical evidence for a link between interactions and radio emission in compact groups is only now accumulating. If interactions are stimulating nuclear radio emission, one would expect the radio luminosity to be correlated with some index describing the degree of interaction such as the projected distance to the nearest neighbor. Evidence in support of this was found by Vettolani & Gregorini (1988) who observed that early-type galaxies have a high ratio of radio-to-optical emission show an excess of nearby neighbors. A similar effect was observed by Malumian (1996) for spiral galaxies in groups. Examining compact group galaxies, Menon (1992) found that elliptical and S0 galaxies detected at a wavelength of 20 cm had closer neighbors than the undetected galaxies. The effect was not found for spiral galaxies, but if one considers only the detected galaxies, there is a significant correlation between radio-to-optical luminosity and nearest neighbor distance for both early and late type galaxies (TK Menon, private communication).

Continuum radio emission has also been detected in a number of HCG elliptical galaxies. Unlike those found in cluster ellipticals, the radio sources are low-luminosity and compact. Where spectral indices are available, they indicate that the radio emission arises from an AGN rather than from starburst activity (TK Menon, private communication). In the HCG sample, there is a significant preference for radio-loud elliptical galaxies to be first-ranked optically (Menon & Hickson 1985, Menon 1992). The probability of radio emission does not correlate with absolute luminosity, but instead correlates with relative luminosity within a group. Spiral HCG galaxies do not show this effect. Although the tendency of radio galaxies in rich clusters to be first-ranked has been known for many years, it is surprising to find a similar effect in small groups, where the number of galaxies and luminosity range is small, the gravitational potential well is much less clearly defined, and it is unlikely that any individual galaxy holds a central location. The effect of optical rank on radio emission had been previously noted in other small groups by Tovmasian et al. (1979) although these authors made no distinction between elliptical and spiral galaxies. It is difficult to imagine any explanation for this result in which the compact group is not a true physical system. It would appear that, regardless of absolute luminosity, only the first-ranked (presumably the most massive in the group) elliptical galaxy can develop a radio source.

5.5 Diffuse light

Stars stripped from galaxies by tidal forces should accumulate in the potential well of the group and may be detectable as diffuse light. In an early photographic study, Rose (1979) found no evidence for diffuse light in his groups, and was lead to the conclusion that most of his groups must be transient configurations. However, Bergvall et al. (1981) were successful in detecting ionized gas and a common halo around a compact quartet of interacting early-type galaxies, and evidence for a common halo in VV 172 was reported by Sulentic & Lorre (1983). Diffuse light can clearly be seen in HCG 94, and has been found in HCG 55 (Sulentic 1987), but Pildis et al. (1995b) did not detect any in seven other compact groups. Analysis by Mamon (1986) indicated that while the expected diffuse light should be detectable with modern techniques, it would generally be very faint. Estimates of the total amount of diffuse light in the detected groups are rather uncertain as they depend sensitively on subtraction of the galactic light, and the sky background. Deeper photometry and improved image processing techniques may yet reveal diffuse light in other compact groups (Sulentic 1997).

5.6 Cool gas

The mass and distribution of cool galactic and intergalactic gas, can be obtained from observations of the 21-cm line of neutral hydrogen. The first such study of a large sample of compact groups is that of Williams & Rood (1987) who found a median HI mass of 2.2 x 10¹⁰ M. They concluded that compact groups are typically deficient in neutral hydrogen by about a factor of two compared to loose groups. This effect is consistent with similar deficit in continuum radio emission seen in the disks of compact group spiral galaxies (Menon 1995), and suggests that interactions in compact groups has removed much of the gas from the galaxies. Simulations suggest that in addition to an outflow of gas, inflow also occurs which may fuel nuclear star-formation, as suggested by the strongly-enhanced radio emission seen in the nuclear regions of compact group spiral galaxies (Menon 1995).

High-resolution studies of individual groups (Williams & van Gorkom 1988, Williams et al. 1991) showed clearly that the gas is not confined to the galaxies. In two of three groups studied, the radio emission originates from a common envelope surrounding the group and in the third group there are signs of tidal distortion. These results strongly indicate that at least these compact groups are physically dense systems and not chance alignments or transient configurations in loose groups. They also show that many groups have evolved to the point that gas contained within individual galaxies has been distributed throughout the group.

In contrast to the HI results, Initial CO-line observations of 15 compact-group galaxies (Boselli et al. 1996) indicated a normalized molecular gas content similar to that of isolated spiral galaxies. However, this result is based on normalizing the flux by the optical area of the galaxy, rather than by the infrared luminosity, and may be biased by the relatively small sizes of compact-group galaxies. Further CO studies, currently in progress, should soon settle this question.

5.7 Hot gas

X-ray observations of hot gas in clusters of galaxies can reveal the amount, distribution, temperature and metallicity of the gas, as well as the relative amount and distribution of the total gravitating mass. Temperature, metallicity (fraction of solar abundance), and bolometric luminosities are estimated by fitting a spectral model, such as that of Raymond & Smith (1977) to the data. X-ray emission from compact "poor clusters" was first reported by Schwartz et al. (1980), who concluded that their X-ray properties were similar to those of rich clusters. Using the Einstein observatory, Bahcall et al. (1984) first detected X-ray emission from Stepfan's Quintet. The X-ray map revealed that the emission is diffuse and not centered on individual galaxies. However, we now know that most of this emission is associated with a shock front rather than gas trapped in the group potential well (Sulentic et al. 1995). Although several other groups (Bahcall et al. 1984, Biermann & Kronberg 1984) were detected by the Einstein observatory, further progress required the improved sensitivity of the ROSAT X-ray observatory.

Pointed ROSAT observations revealed massive hydrogen envelopes surrounding the NGC 2300 group, a bright elliptical-spiral pair with two fainter members (Mulchaey et al. 1993), and HCG 62, a compact quartet of early-type galaxies (Ponman & Bertram 1993), and showed that these systems are dominated by dark matter. Subsequent investigations detected X-rays from 18 additional compact groups, either from individual galaxies, or from diffuse gas (Ebeling et al. 1994, Pildis et al. 1995a, Sarraco & Ciliegi 1995, Sulentic et al. 1995). These studies showed that the physical properties of individual systems span a wide range, but that the ratio of gas-to-stellar mass is significantly lower than in rich clusters. Moreover, the detected compact groups all contained a majority of early-type galaxies. No spiral-rich groups were detected (although Mulchaey et al. (1996b) pointed out that they might be found from QSO absorption spectra). This result is consistent with the fact that X-ray-selected groups (Henry et al. 1995) and loose groups (Mulchaey et al. 1996a) tend to be spiral poor, and led to the suggestion that spiral-rich compact groups might not be physically dense systems at all.

The most extensive X-ray study of compact groups to date is that of Ponman et al. (1996). These authors combined pointed and survey-mode observations of a complete sample of 85 HCGs and detected diffuse emission in 22 groups. They conclude that, when the detection limits are considered, diffuse emission is present in at least 75% of the systems. Significantly, they detected diffuse emission in several spiral-rich groups. In these the surface brightness is lower and the X-ray emission has a lower characteristic temperature, as would be expected given the lower velocity dispersions of spiral-rich compact groups. The diffuse X-ray luminosity was found to correlate with temperature, velocity dispersion, and spiral fraction, but not with optical luminosity. The last result suggests that the gas is mostly primordial and not derived from the galaxies. The correlations with temperature and velocity dispersion appear to be consistent with a single relation for clusters and groups (Figure 5).

Figure 5. Bolometric X-ray luminosity vs temperature. Filled circles indicate compact groups, open circles indicate X-ray selected groups (Henry et al. 1995) and squares indicate clusters. X-ray data for the compact groups and clusters are taken from Ponman et al. (1996). A single relation is consistent with clusters, groups and compact groups.

The total mass in compact groups typically exceeds the stellar and gas mass by an order of magnitude. Pildis et al. (1995) derived baryon fractions of 12-19%. Davis et al. (1996) obtained 10-16% for the NGC 2300 group. These are comparable to the fractions found for poor clusters (Dell'Antonio et al. (1995) and are about half the typical values found for rich clusters. However, the derived baryon fraction depends sensitively on the radius within which it is measured and on the assumed background level (Henriksen & Mamon 1994). Both the total mass, which is dominated by dark matter, and the gas mass continue to increase with radius. Consequently, both the baryon fraction and the gas fraction are poorly determined.

The contribution of compact groups to the X-ray luminosity function has been estimated by PBEB, who find that on the order of 4% of the total luminosity in the range 10⁴¹-10⁴³ erg s^-1 comes from HCGs. This is higher than the contribution of HCG galaxies to the local optical luminosity density estimated at 0.8% by Mendes de Oliveira & Hickson (1991), but it is comparable to the value found by Ribeiro et al. (1994).

The metallicity inferred for the X-ray-emitting gas in compact groups is relatively low. PBEB obtain a mean metallicity of 0.18 solar, compared to the value, 0.3-0.4 solar, found in rich clusters. This is comparable to the low value (< 0.11 solar) found for the NGC 2300 group (Davis et al. 1996). These figures suggest that the gas is largely primordial, a result supported by the absence of a correlation between X-ray and optical luminosity. However, given the limited spectral resolution of ROSAT, these low metalicities cannot yet be considered secure. The higher spectral resolution and sensitivity of the ASCA satellite should provide more definitive results. Recent observations of HCG 51 and the NGC 5044 group found metal abundances comparable to those of clusters (Fukazawa et al. 1996).