Annu. Rev. Astron. Astrophys. 1984. 22:
185-222 Copyright © 1984 by Annual Reviews. All rights reserved |
3.3 Optical Observations Relevant to Merging, Accretion, and Tidal Stripping
ARE cD GALAXIES BUILT FROM HOMOLOGOUS MERGERS?
Schweizer is correct in pointing out that if all giant ellipticals
have Rcore ~ 100 pc, like NGC 1316, then few galaxies
for which it has
been claimed that core radii have been measured have actually been
measured. Several overenthusiastic observers, including Dressler,
have misinterpreted data like
Oemler's (1976)
as indicating large core
radii for cD galaxies, when the true core radii may well have been
hidden by the atmospheric seeing disk (as Schweizer notes). He is
probably also correct in stating that
Hoessel's (1980)
analysis of 108
first-ranked cluster galaxies includes a large number, perhaps a
majority, of cases whose Hubble core radii are only artifacts of the
seeing. In some cases, however, the cores have radii of several
kiloparsecs and are well resolved.
All of this is probably irrelevant, however, since it is the
large-scale structure that is indicative of whether the cannibalism
model adequately describes the evolution of cD galaxies.
Hoessel (1980) and
Schneider et
al. (1983)
were really after the logarithmic
intensity gradient = d
(ln L) / d (ln r) at r = 16 kpc. The latter study
included a better (double-Gaussian) seeing model and fit the galaxies
to both a t modified Hubble law and a de Vaucouleurs profile that has
no core. These new data show that it is possible to measure quite
accurately for galaxies with z 0.2, since the radius at which is
being measured is an order of magnitude larger than the
seeing. Schneider (private communication) has repeated observations of
41 galaxies in a variety of seeing conditions, and his data show that
changes by an order of 10%
over a range of seeing of a factor of
two, again indicating that these measurements are basically unaffected
by seeing. The core radii that are derived when a Hubble law is fit
may be artifacts of the technique, but only Space Telescope
observations will tell if the core radii are actually much smaller.
In the samples of Hoessel and Schneider et al., increases with
absolute magnitude in a way that is roughly consistent with the
HO
models of homologous growth.
Morbey & Morris
(1983)
reach the same
conclusion independently for another sample of cD galaxies.
Thuan &
Romanishin (1981)
and Morbey &
Morris (1983)
both find that giant cDs
have much lower surface brightnesses within their effective radii,
again in agreement with the
HO model.
Since it is clear that cDs do seem to mimic the model of homologous
growth on the large scale, it is worthwhile to reconsider Schweizer's
comments about core radii and central surface brightnesses. Although
the N-body models consistently predict higher central surface
brightness as evolution proceeds, these simulations have a small
number of particles (~ 1000), and it is possible that two-body
relaxation effects dominate in the central regions and render these
models inappropriate. On the observational side, it is relatively easy
to find examples of giant galaxies whose central surface brightness is
low. For example, despite Schweizer's early report to the contrary,
four of the five cases studied by
Kron & Albert (1982)
have lower
central surface hrightness than other bright ellipticals in the
clusters. A2029, A2218, and A1413 are additional excellent examples of
cDs whose central surface brightness is very low. The lower surface
brightnesses of these archetype cDs are striking, and comparisons with
puny galaxies like NGC 1316 are, therefore, very misleading. Though
there may be good counterexamples (A2634, A2670), the general rule
seems clear: the most luminous cD galaxies have low central surface
brightnesses. This is again consistent with the homologous merger
picture. The exceptions might be due to the remnants of a captured galaxy.
Perhaps the best piece of evidence that the merger process is
responsible for building the insides of cD galaxies is the observation
that a large fraction (25-50%) of such objects have multiple nuclei
(Matthews et
al. 1964,
Hoessel 1980,
Schneider et
al. 1983).
This high
frequency cannot be dominated by projection effects (Schneider et
al.), so the conclusion that at least some galaxies are being consumed
by the cD seems unavoidable. Schneider et al. point out that the
frequency of multiple nuclei for the second- and third-ranked cluster
galaxies is an order of magnitude less. This datum seems to confirm
the special place of the cD in the bottom of the cluster potential
well and its unique evolution compared with other bright galaxies in
the cluster.
If none of this evidence is compelling for the case that cD galaxies
grow by ``cannibalism,'' it is sufficient to say that most optical
observations now available are consistent with the general features of
this model.
ARE THE TIME SCALES FOR DYNAMICAL FRICTION CORRECT?
This possible discrepancy is reminiscent of similar discussions
about cD galaxies and the frequency of their multiple nuclei, which
are typically 10 kpc from the cD center.
Hoessel (1980)
and
Schneider et
al. (1983)
have argued that the frequency of multiple nuclei is
consistent with the expected time scale for dynamical friction of
109
yr. (The victims are low mass.) On the other hand,
Blandford & Smarr
(1984)
claim that a much larger central mass than a cD galaxy (a black
pit) is needed to hold these victims on their spiraling
orbits. Furthermore, a smaller core radius and a larger orbital
velocity will result in a longer infall time, consistent with the high
frequency of occurrence of multiple nuclei.
Tonry's (1984b)
simulations imply, however, that black pits may not be necessary if
the galaxies are infalling on highly elliptical orbits.
Such simulations and observations suggest that there is much to be
learned about the processes that go under the label ``dynamical
friction,'' and, therefore, discrepancies between "predicted" and
``observed'' time scales of less than an order of magnitude should be
viewed with interest, but not despair.
OTHER OBSERVATIONS RELEVANT TO CANNIBALISM
Related to this topic are studies by
Dressler (1979)
of the cD in
A2029 and by
Carter et al. (1981)
of the dumbbell galaxy IC 2082,
which show that the velocity dispersions of these galaxies rise with
increasing radius, in contrast with the steady or declining velocity
dispersions in normal ellipticals
(Illingworth 1983).
Dressler (1979)
interprets this rise in velocity dispersion as the gravitational
effect of the cluster's dark matter on the stars in the cD's
envelope. Thus the rise in velocity dispersion is, in itself, neither
evidence for nor against the cannibalism model, save that it requires
the galaxies in question to lie at the true centers of their
clusters. Dressler's model for the cD in A2029 includes a normal
bright elliptical centered in the cluster potential, but it also
requires a component of intermediate M / L and velocity dispersion to
supply the additional light at R ~ 100 kpc that distinguishes this
cD. This extra component is suggestive of the cannibalism model. since
it can be attributed to the remains of accreted galaxies.
McGlynn &
Ostriker (1980)
show that the relaxation time is a
function of Bautz-Morgan type
(BM
I clusters have shorter times), as
is expected with the cannibalism model, but the dependence is weak.
van den Bergh (1983)
has argued that the number of globular clusters
per unit luminosity is very high in M87 (Virgo) and NGC 3311 (Hydra),
so that it is unlikely that these central galaxies are ``star piles''
made from more normal galaxies. Unfortunately, no data are available
on the globular cluster frequency in the truly outstanding cD
galaxies, and it is for these examples that the cannibalism model is
most appealing. Van den Bergh also cites the fact that cDs are often
quite elongated and aligned with flattened clusters (see references
above) as evidence that the formation of the cD and its cluster were
simultaneous (see also
Carter & Metcalfe
1980,
Adams et al. 1980).
On the other hand,
Binney (1977)
has pointed out that a strong anisotropy
in the velocity field of the cluster will result in an anisotropic
stellar envelope in the cD as the victims are cannibalized.
Lugger (1984)
has presented the most extensive data on colors of the
nuclei of bright ellipticals and cDs, from which she concludes that
there is a monotonic reddening with increasing luminosity (cf.
Thuan &
Romanishin 1981).
This is in contradiction with the
HO prediction that
the cDs should break from this relation and be about 0.m1
bluer in U -
B than an extrapolation from the lower luminosity ellipticals.
Reliable measurements of color gradients are only available for a
small sample of elliptical and cD galaxies
(Strom & Strom
1978a,
Gallagher et
al. 1980,
Wirth & Shaw 1983).
At present, there are
insufficient data to make a meaningful comparison, as well as
uncertainty in the models as to what extent the color gradients will
be erased in the merging process (e.g.
HO, White 1980).
EVOLUTION OF THE LUMINOSITY FUNCTION
This result is quite remarkable. It means that the processes that
determined the luminosities of galaxies were either very insensitive
to local conditions such as density, temperature, and turbulence
(angular momentum), or that these conditions varied little from
protocluster to protofield regions of space. Furthermore, it implies
that the evolution of the distribution, in dense regions for example,
is not substantial.
Nevertheless, the processes of merging, tidal stripping, and
accretion discussed above are expected to produce some evolution of
the LF, and those early-Universe models in which galaxies form after
clusters (within adiabatic perturbations) might be expected to produce
some initial variations as a function of environment. Therefore, it is
worthwhile to look beyond ``first-order universality'' for second-order
differences. A large number of LFs for rich clusters, produced mainly
by photographic surface photometry, are now available for such a
comparison.
First indications that there are systematic differences from cluster
to cluster came from
Oemler (1974),
who showed what he considered to
be weak evidence that the composite LF for ``spiral-rich'' clusters is
flatter than those for ``spiral-poor'' or ``cD'' clusters. The latter had
a more sudden rise at the bright end and then leveled off fainter than
some characteristic magnitude.
Dressler (1978b)
applied statistical tests to his sample of 12 very
rich clusters to see if such variations were significant. He fit the
cluster LFs to a functional form suggested by
Schechter (1976),
and used Monte Carlo models to test for significant variations of the
characteristic magnitude M and the faint-end (power-law) slope .
Dressler found three types of variation from the universal form.
Statistically significant deviations of M were found in at least two
clusters. Several clusters were found to have an unusually flat slope
at the faint end, and there was once again marginal evidence that the
form itself varies, much like Oemler's comparison of ``spiral-rich,''
``spiral poor,'' and ``cD'' clusters.
Unfortunately, most of the LFs produced subsequently have not been
subjected to these types of statistical tests. Only
Schneider (1982)
has checked on the universality of M. For his large sample he finds
variations in M
similar to those found by Dressler, but because the
small field size limited his study to ~ 50-100 galaxies per cluster,
these variations are not statistically significant.
On the other hand, there is some supporting evidence of variations
in faint-end slopes and steepness of the bright end. Among the 8
clusters (plus Virgo) studied by
Bucknell et
al. (1979),
there are clear examples of clusters in which the LF bright end is very steep
(A2065, A2670, A2199, and A426), and others where it is much shallower
and a distinct ``break'' at M is not obvious (A2151, A2147, and
Virgo). Similar steep bright ends have been noted
for A1146
(Carter & Godwin
1979),
A1413
(Austin & Peach
1974),
and A1930
(Austin et al. 1975),
while A1367
(Godwin & Peach
1982)
has been found to have a
flat bright end. The clusters A1656
(Godwin & Peach
1977)
and CAO 340-538
(Quintana &
Havlen 1979)
appear intermediate. These
differences in shape may reflect different mixes of morphological
types, since the clusters with the steep bright ends are rich in E and
S0 galaxies, and the shallower LFs with less obvious breaks are often
found in spiral-rich clusters.
Thompson &
Gregory's (1980)
study of
the Coma cluster shows that a composite E+S0 LF differs
from a spiral
LF in just this way. They predict that if the Coma populations are
representative, an ensemble of clusters from spiral rich to spiral
poor should have (M)
~ 0.m5, which is consistent with both Dressler's
and Schneider's results. New luminosity functions for clusters with
morphological types like
Dressler's (1980a)
sample are necessary to
test this putative dependence of the LF on cluster population.
The LF bright end is usually very steep in clusters that have a
luminous cD galaxy (see, for example,
Austin & Peach 1974,
Oemler 1974,
Dressler 1978b)
- that is, there are few other bright
galaxies. This could be interpreted as the result of dynamical
evolution, where the brightest galaxies have all merged to form the
cD. Because of the small-number statistics in each case (only a few
bright galaxies are ``missing''), it is difficult to test this
suggestion without a much larger sample.
Some large variations have been reported in the faint-end slope of the
luminosity function, as described by the power-law slope
(Schechter 1976).
The typical situation in rich clusters is for the LF to continue
to rise fainter than M, with ~
-1.25. This result has been verified to
Mv ~ -14 in Virgo by
Sandage &
Binggeli (1984)
and in
Coma by
Beckman (1983),
although the faint-end LF in Coma is rather complex. Certain
other rich clusters, however, particularly A2670
(Oemler 1973,
Dressler 1978b), A401
(Dressler 1978b),
and A1146
(Carter & Godwin
1979),
have flat LFs ( = 1.0) several
magnitudes below M. Such measurements are
plagued, unfortunately, by the uncertainties in field and incompleteness
corrections, and a thorough study of cluster membership (i.e. redshifts)
may be necessary to confirm this effect. Nevertheless, the data are
suggestive, particularly in light of simulations by
Miller (1983),
who shows that tidal stripping can have a severe effect on faint galaxies,
resulting in the flattening of the faint-end LF. (These simulations are
usually of the mass function, however, so conclusions about the
luminosity function include the additional uncertainty of the
distribution of light in a galaxy relative to its mass distribution.)
It is therefore worth noting that the rich clusters with flat faint-end
LFs are dense and have a high degree of symmetry, properties thought to
be indicative of clusters whose dynamical evolution is very advanced.
Recently, evidence of even more dramatic variations in the faint-end
LF have been reported
(Heiligman &
Turner 1980,
White & Valdes
1980),
but these variations are likely to be due to selection effects.
These indications of variation in the ``nearly universal luminosity
function'' are suggestive, but not yet compelling. The rapid growth in
the technology of digitizing photographic plates and the advent of
linear area detectors should result in many new LFs in this decade;
these should settle the issue.
EVIDENCE FOR TIDAL STRIPPING
A few observations are available, however, and they tentatively
support the standard models quoted above that predict that galaxies in
the densest regions of space may have lost up to 50% of their
luminosity (and most of their mass). A monumental data set for 400
elliptical galaxies was constructed by Strom & Strom for A1656
(1978a), A1367 and A426
(1978b), and A1228, A2151, and A2199
(1978c)
using photographic photometry from Mayall 4-m plates. The primary data
compiled are the R26 radii and the
Mv magnitudes within R26.
Strom & Strom
(1978d)
conclude that similar log (R26)
vs. Mv relations exist
for all clusters, but that elliptical galaxies in the denser,
spiral-poor clusters are, on average, 10% smaller at a given
Mv than
those in spiral-rich clusters. There is also a gradient in size
within the spiral-poor clusters, such that ellipticals within 0.5 Mpc
of the cluster center are about 5% smaller than those farther out. A
similar conclusion concerning the size of S0 galaxies in Coma as a
function of radial distance is also reported by
Strom & Strom
(1978e).
They also find that at a given central surface brightness
(within a 2" aperture), the smaller galaxies are about 0.m5
fainter. Both of these results concerning size and brightness are
consistent with the effects of tidal stripping predicted in the
studies cited above.
There is, of course, considerable observational and cosmic scatter
in these relationships, and the shortcomings of photographic surface
photometry add uncertainty to these results. It is disturbing, for
example, that two apparently dense (and evolved) clusters, A401 and
A2670
(Strom & Strom
1979),
have a similar radial dependence of galaxy
size, but that the absolute sizes of the ellipticals in these clusters
are as large at a given Mv as those in the spiral-rich clusters
studied. More data, preferably done with linear area detectors, are
necessary before any firm conclusions can be drawn on the sizes of E
and S0 galaxies as a function of environment.
Also consistent with the tidal truncation expected from galaxy
encounters is the correlation found by
Hickson et
al. (1977)
between the intergalactic spacing in dense groups and the size of the largest
galaxy. These data emphasize the importance of reliable measurements
of intergalactic light in these systems for use as an unambiguous
measure of tidal damage. For example,
Rose (1979)
used the absence of
intergalactic light to argue that the compact appearance of certain
groups of galaxies must be temporary, since there is little evidence
of tidal debris.