Galaxy dynamics and the mass density of the universe

Proc. Natl. Acad. Sci. USA
Vol. 90, pp. 4814-4821, June 1993
Colloquium Paper
This paper was presented at a colloquium entitled "Physical Cosmology," organized by a committee chaired by David
N. Schramm, held March 27 and 28, 1992, at the National Academy of Sciences, Irvine, CA.
Galaxy dynamics and the mass density of the universe
VERA C. RUBIN
Department of Terrestrial Magnetism, Carnegie Institution of Washington, Washington, DC 20015
ABSTRACT Dynamical evidence accumulated over the
past 20 years has convinced astronomers that luminous matter
in a spiral galaxy constitutes no more than 10% of the mass of
a galaxy. An additional 90% is inferred by its gravitational
effect on luminous material. Here I review recent observations
concerning the distribution of luminous and nonluminous
matter in the Milky Way, in galaxies, and in galaxy clusters.
Observations of neutral hydrogen disks, some extending in
radius several times the optical disk, confirm that a massive
dark halo is a major component of virtually every spiral. A
recent surprise has been the discovery that stellar and gas
motions in ellipticals are enormously complex. To date, only for
a few spheroidal galaxies do the velocities extend far enough to
probe the outer mass distribution. But the diverse kinematics
of inner cores, peripheral to deducing the overall mass distri-
bution, offer additional evidence that ellipticals have acquired
gas-rich systems after initial formation. Dynamical results are
consistent with a low-density universe, in which the required
dark matter could be baryonic. On smallest scales of galaxies
[10 kiloparsec (kpc); H. = 50 kmsec'lmegaparsec'11 the
luminous matter constitutes only 1% ofthe closure density. On
scales greater than binary galaxies (i.e., .100 kpc) all systems
indicate a density -10% of the closure density, a density
consistent with the low baryon density in the universe. If
large-scale motions in the universe require a higher mass
density, these motions would constitute the first dynamical
evidence for nonbaryonic matter in a universe of higher
density.
Twenty years ago, astronomers started accumulating obser-
vations that indicated that orbital velocities of stars and gas
in spiral galaxies remained high out to the limits of the
observable galaxy. By 1982, after a decade ofinitial disquiet,
most astronomers reluctantly accepted the conclusion that a
galaxy consists of much more than the luminous stars, gas,
and dust that can be observed at various wavelengths. In fact,
at least 90% of a galaxy consists of dark matter, matter that
is undetectable at any wavelength and that is deduced only by
its gravitational influence on the orbits ofthe stars and the gas
that we can see. But not all astronomers were reluctant to
accept the evidence. "There are reasons, increasing in num-
ber and quality, to believe that the masses of ordinary
galaxies may have been underestimated by a factor of 10 or
more," stated Ostriker, Peebles, and Yahill (1) in a truly
prescient sentence.
This revolution in our understanding came about because
theoretical arguments revealed the necessity ofstabilizing the
disks of galaxies with invisible halos (2), because observa-
tions with optical (3, 4) and with radio (5) telescopes showed
uniformly high rotation velocities, and because an enor-
mously influential review of the inferred distribution of
The publication costs ofthis article were defrayed in part by page charge
payment. This article must therefore be hereby marked "advertisement"
in accordance with 18 U.S.C. §1734 solely to indicate this fact.
matter in galaxies (6) showed the validity of a universe in
which most ofthe matter is dark. In the decade ofthe 1980s,
observational and theoretical investigations of the dark mat-
ter became one of the most active fields of extragalactic
astronomy, and excellent reviews exist (7, 8).
For this paper, I describe a handful of recent results that I
consider to be among the most significant of the last few
years. They relate to the dynamics of disk and elliptical
galaxies, to the distribution ofnonluminous matterin galaxies
and in clusters, and to estimates of the overall mass density
of the universe.
Section I. The Milky Way and the Local Group Are Massive
We live in a large, massive spiral galaxy, the Milky Way, with
a nearby equally large and massive companion spiral, the
Andromeda galaxy. Together, these contain most ofthe mass
within the Local Group. For the past several decades,
determinations of the mass within our galaxy and the Local
Group have been imprecise. This was due principally to the
lack ofaccurate velocities and lack ofknowledge oforbits for
the outlying globular clusters and small satellite galaxies
whose motions are used to map the gravitational potential of
the Galaxy, and a lack of knowledge of the evolutionary
history ofthe dwarfsatellites. Recently, more accurate radial
velocities and distances of 47 objects on the outskirts of our
galaxy have become available. Based on the assumptions that
(i) the distribution of dark and luminous matter can be
modeled by a power-law function, and (ii) that the velocity
distribution of the outer Galaxy is isotropic, Kulessa and
Lynden-Bell (9) estimate a Galaxy mass of 1.3 x 1012 solar
mass units (MO) out to the distance sampled by the test
objects, 230 kiloparsec (kpc; 1 pc = 3.09 x 1016 m). Based on
the earlier Schmidt (10) model ofthe Galaxy, a mass of =1.9
x 1011 Mo is located within 25 kpc ofthe center (11). Hence,
the new analysis increases the radius, the mass, and the
mass-to-light ratio (M/LB in solar units, with the luminosity
measured in the blue spectral band) each by nearly an order
of magnitude. At the 25-kpc limit of the optical disk, M/LB
10, and the dark matter constitutes 70% of the total mass
(11); at 230 kpc, M/Lg 50, and dark matter constitutes
>90%o of the mass (9).
A statistical analysis using the velocities of satellites or-
biting a homogeneous set of spiral galaxies similar to the
Milky Way (12) leads to a mass of order 2 x 1012 MO within
200 kpc, in agreement with that discussed above. This large
mass and mass-to-luminosity ratio for our galaxy (and hence
also for Galaxy M31) is also in comfortable agreement with
a mass of -6.4 x 1012 Me determined by Peebles (13) for the
Local Group. Note that a radius of 230 kpc reaches suspi-
ciously close to the 350-kpc half-way point between us and
the Andromeda Galaxy. If the dark halos of our Galaxy and
Abbreviations: MIL, mass-to-light ratio; HI disk, neutral hydrogen
disk; pc, parsec; M<>, solar mass unit.
4814

Proc. Natl. Acad. Sci. USA 90 (1993) 4815
M31 reach each other, the values of MILB will increase to
nearly 100, a value of MIL similar to that for other systems
of similar radial scale. I discuss MIL values as a function of
radius in Section IV below.
Section II. AB Spiral Galaxies Need Dark Matter (Halos)
For spirals more distant than M31, velocity information is
usually obtained from spectral observations of ionized hy-
drogen (Ha) sampled by a long-slit spectrograph aligned
along the major axis and from 21-cm mapping ofvelocities of
the neutral hydrogen gas. To derive a mass distribution from
a velocity distribution, it is generally assumed that Newto-
nian gravitational theory is valid, that a self-gravitating
luminous disk of constant MIL dominates at small radial
distances, that a dark isothermal halo dominates at large
radial distances, and that the internal extinction is not a
function of radius. Surely the internal extinction merits a
more sophisticated treatment, as does the distribution of
mass in the halo, but the above simplifications suffice at
present.
Outermost Regions. Virtually all observed spiral rotation
curves are flat (or rising) to the limits of the observations,
implying that the mass density decreases approximately as
1/r2. The luminosity however falls off faster, exponentially.
Thus, the ratio of mass density to luminosity density in-
creases with radial distance, requiring the presence of a halo
of dark matter, which becomes increasingly dominant with
increased nuclear distance. A major requirement for deriving
meaningful parameters for the halo and for the disk is that
rotation velocities extend from the nucleus to at least several
disk scale lengths (where one scale length equals the radial
distance at which the disk surface brightness has fallen by
lle). Velocity measures for spirals now extend this far,
especially for galaxies whose neutral hydrogen disks (HI
disks) are several times more extended than their optical
disks. I show in Fig. 1 an optical image of the M101 Galaxy,
with the superposed velocity contours of the neutral hydro-
15°50'-
a
0)
v-
(301
a
1520'I-
gen disk; an HI disk with a radius three or four times that of
the optical disk is quite impressive.
To determine the contribution to the rotational velocities
from the mass of the halo, the contribution to the velocity of
a constant M/L disk must first be subtracted in quadrature
from the observed velocities. However, even an extended
rotation curve is insufficient to constrain the mass distribu-
tion because MIL of the disk is diffi'cult to pin down. For
example, velocities in the NGC 3198 Galaxy can be well-
fitted by models with a disk ofa range ofmasses, from a disk
of maximum mass (MIL = 4.4) down to a disk of zero mass
(ref. 15; Fig. 2). During the mid-1980s, it was customary to
analyze rotation curves by adopting the maximum mass disk
consistent with the inner rise ofthe rotation curve (17). These
models generally resulted in dark and luminous mass ofabout
equal amounts within the optical galaxy, but it must be
remembered that the halo mass so determined is only a lower
limit.
A major conceptual advance came when Athanassoula and
colleagues (18) pointed out that information concerning the
spiral structure of the disk could be used as an additional
constraint on the relative halo-to-disk mass. Forgalaxies with
low mass halos, the disk is unstable against the formation of
a nonaxisymmetric m = 1 mode. Adding mass to the halo
inhibits the m = 1 mode, until the m = 2 (i.e., two-armed)
mode dominates. By requiring m = 2 (but no m = 3) modes,
disk masses of order one-half the maximum disk mass
generally result. Even more importantly, these authors dem-
onstrated that every spiral with a rotation curve extending
several disk scale lengths requires dark matter, finally putting
to rest the belief (19) that some galaxies (always only those
with not very extended rotation curves) could be modeled
with no dark halos.
In an observational survey of spirals with extended HI
disks, Casertano and van Gorkom (20) discovered two spiral
galaxies, NGC 2683 and 3521, whose extended HI disks
exhibit falling velocities beyond the optical disk. Surpris-
ingly, dark matter is required even forthese spirals. ForNGC
(194m
R.A. (1950)
1h33m
FIG. 1. An optical image of the spiral galaxy M101 with velocities (V) from 21-cm neutral hydrogen (HI) superposed (14). The hydrogen
diameter is about four times the optical diameter. Angular coordinates on the sky (right ascension and declination) are the indicated coordinates.
401-
30'[-
0
r
Colloquium Paper: Rubin
i I

Proc. Natl. Acad. Sci. USA 90 (1993)
P I
A
a - 8-5 kpc
r 2
p0 0-0075 Mo pc-3
M 4-4
Lv,disk
a - 1-3kpc
y - 205
po - 0-27 M, pc-3
M
1.3
Lv, disk B
60
-S1-5 kpc
N t s~~~~~~~~~~2-25
50 - 0 33 Mo p-3
_L__ d i s k
0 10 20 30 40 50
r, kpc
FIG. 2. Three alternative decompositions ofthe rotation curve of
NGC 3198 (15). The observed velocities (16) are shown as points,
with the model disk and halo-rotation curves plus their sum in
quadrature (curves). The model parameters are given at right, for the
assumed halo-density distribution p. Note the large ranges permitted
by the models for values of central densities po and disk MIL values.
2683, even the maximum acceptable disk (MIL = 8, set
mostly by the maximum of the velocities) predicts velocities
falling more rapidly than observed (Fig. 3). The authors argue
that there is also a lower limit to the disk mass, for too much
halo mass produces too flat a rotation curve. It is unclear how
much this conclusion would be altered if the innermost
optical velocities, discussed in Innermost regions below, had
been available. These near-nuclear velocities are not observ-
able at the low spatial resolution of the 21-cm observations.
From these and many other recent observations of veloc-
ities in the outermost regions of spirals, I draw the following
conclusions. Virtually all spirals need dark matter in amounts
significantly greater than the luminous matter in order to
reproduce the observed velocities. Beyond the optical gal-
axy, rotation velocities in the gas disk are generally constant,
implying masses that continue to rise linearly with radius,
even in those regions where the optical luminosity is essen-
tially nil. Hence, local values of MIL are enormous. Conse-
quently, only lower limits to the mass and to MIL values are
available for spiral galaxies. The only exceptions to this
generalization are tidally distorted galaxies, and a few mod-
erately peculiar spirals in dense regions near the cores of
galaxy clusters (21), whose velocities decrease with in-
creased nuclear distances.
Innermost Regions. Advances in radio astronomy tech-
niques have permitted the detection and velocity mapping of
faint outer disks and, consequently, the determination of the
mass interior to large radial distances. Corresponding ad-
vances in optical instrumentation have permitted the spectral
40
ua0)
2a
u
0
._
to
Ca
0
100 50
C
-~~ 4
0
d
v) r SNGC 2683 MAJOR
0 Ha
NE 0 [NIl]
135 270 405
Distance from nucleus on sky, arcsec
540
FIG. 3. (A) NGC 2683, an Sb spiral, from a Kitt Peak 36-inch
charge-coupled device image. Below it, a portion of a Kitt Peak 4-m
spectrum, showing Ha (Upper) and [NII] emission. Note the mul-
tivalued emission in the inner regions. (B) The measured optical
velocities in the inner region from the spectrum in A, quantifying the
rapid rise across the nucleus and the multivalued velocities. The
surface brightness profile of the galaxy is shown by the fine line. (C)
HI velocities in NGC 2683 (20). The solid line represents the
predicted velocities based on a model with no dark matter. Thus,
even the falling rotation velocities are higher than the velocities
predicted by the light. Note the change in radial scale from B to C;
all the velocities in B lie between r =
0 and the first two 21-cm points.
observations of inner-core regions at high spatial and high
velocity resolution. Moreover, the use of digital detectors
(charge-coupled devices) allows the smooth galaxy stellar
continuum, strong in the inner regions, to be easily removed
from the spectrum, thus revealing the emission ofthe ionized
4816 Colloquium Paper: Rubin
-50 -100

gas at enhanced signal-to-noise ratio. From these observa-
tions we now know that gas in central regions often exhibits
complex velocity patterns arising from a variety of causes:
warps, twists, or nonalignments ofgas disks; time-dependent
settling of recently acquired gas; nonaxisymmetric (triaxial)
orbits that may explain why some ellipticals and early-type
spirals have rotational velocities that increase too slowly for
a constant MIL disk; and even holes in the innermost regions
and warps in the outermost regions, so that gas along the
line-of-sight at various nuclear distances and various pro-
jected velocities superposes at a single radial position.
NGC 2683, one of the two spirals with a falling 21-cm
rotation curve discussed above, is a prime example exhibiting
complex inner velocities. I show in Fig. 3 an image of the
Galaxy, the Ha region of the spectrum in the inner 2' with the
underlying stellar continuum removed, the near-nuclear op-
tical velocities, and the 21-cm velocity curve from Casertano
and van Gorkom (20).
Within the inner 60", velocities in NGC 2638 are multival-
ued, extending from the upper maximum velocity down to
velocities negative with respect to the center. The steep
velocity rise implies a massive nucleus, of order5 x 108 Mo.
The upper envelope represents the rotation velocities within
the disk, while at least some of the lower-velocity gas is likely
foreground gas warped out of the plane and intruding into the
line-of-sight. Other low-velocity gas must arise from disk gas
of low angular momentum, probably derived from bulge
stars. None of these details are visible at the lower-resolution
21-cm observations. In fact, these optical observations all lie
closer to the nucleus than the second plotted point on the
21-cm velocity curve; in velocity they rise slightly with
decreased nuclear distance to the velocity V = 220 km/sec at
2" from the nucleus. Thus, a realistic mass deconvolution into
halo, disk, and bulge or central point mass (necessary to
produce the steep nuclear velocity gradient) must be based on
velocities for the inner regions obtained from high-resolution
optical observations and for outer regions on velocities from
high-sensitivity radio observations. Let us now resurrect
Bosma's (5) forgotten suggestion for such composite curves.
Complex inner regions are not rare among galaxies. Rubin
and Kenney (22) have discovered rapidly rotating cores with
kinematic properties distinct from the normal outer rotation
in --25% of the 70 Virgo cluster spirals and ellipticals they
have observed at high-velocity resolution and high spatial
scale. The galaxies are located in all regions of the cluster,
span Hubble types from E through Sc, and have a wide range
of central velocities and viewing angles. For a spiral viewed
close to edge-on, a complex velocity field can be ascribed to
an outer warp projected along the line of sight. For a galaxy
viewed more nearly face-on, the inner region itself must have
a warped or skew or twisted morphology. We are continuing
our analysis of these interesting galaxies. Complex nuclear
regions are especially prevalent in elliptical galaxies, and I
discuss these in Section VI below.
Section III. Dwarf Spirals Contain Mostly Dark Matter
It was recognized a decade ago that low-luminosity galaxies
would be important for learning about dark matter. The
systematics of galaxy rotation curves as a function of lumi-
nosity suggested the query, " . . . for the smaller galaxies,
... Is the luminous matter only a minor component of the
total galaxy mass?" (4). Moreover, the low mass, and con-
sequent low acceleration, suggests that extremely low-mass
galaxies are a fertile testing ground for theories that reject the
existence of dark matter but instead modify Newtonian
gravitational theory.
A major discovery of the last few years is the existence of
low-mass galaxies that are enormously gas rich and in which
the dark mass exceeds the luminous mass by a large amount.
DDO 154, one of the dwarfs richest in neutral hydrogen (23,
24), contains a neutral hydrogen gas disk with radial extent
five times the optical disk. The gas disk is dynamically more
important than the stellar disk; stars constitute only "1% of
the mass. It is the very low velocity gradient beyond the
nucleus, combined with the steep fall in disk luminosity,
which forces the very low-mass disk. Interior to the last
measured velocity, at least 95% of the mass is dark, and MIL
is "50, almost an order of magnitude greater than a typical
type Sc galaxy (within its optical radius).
The existence of dwarf spirals with such a small fraction of
luminous mass is remarkable. One question that naturally
follows, "Is there a galaxy that is totally dark?", has not yet
been answered.
Equally intriguing is the dwarf spiral NGC 1560 (25). Gas
rich, the best-fit model has the stellar and gas masses about
equal, together contributing about one-half the mass of the
dark matter at the last measured point. Especially remarkable
is the close similarity in form with radius ofthe total mass and
the gas mass, particularly as indicated by a dip in both the
velocity and the gas density at -5 kpc. This similarity implies
a close relation between the distribution of the HI gas and the
dark matter. It also makes spiral NGC 1560 especially
interesting for astronomers interested in efforts to modify
Newtonian dynamics to account for flat rotation curves.
Milgrom (26) has proposed modifications to Newtonian dy-
namics (MOND theory), such that at low accelerations the
attractive force varies as llr. In conventional Newtonian
dynamics, "what you see is not what you get"; that is, the
luminous galaxy is embedded in a dark halo whose mass is the
dominant component in producing the observed velocities.
Thus, the velocities predicted by the observed gas density
would be expected to have only an insignificant relation to the
total observed velocities produced by the combined gravita-
tional field of the stars, gas, and (dominant) dark matter.
In contrast, in the MOND theory, "what you see is what
you get," so the agreement of the radial variation of the
observed velocities and the (Newtonian predicted) gas-plus-
stars-predicted velocities is fundamental to the theory. While
few astrophysicists are comfortable with a modified Newto-
nian theory, such theory cannot yet be ruled out on the basis
of the dwarf spirals.
Section IV. Elliptical Galaxies Are Much More Complex
Than We Thought
Velocity Studies from Stellar Absorption Lines. Ten years
ago, discussions ofgalaxy dynamics were based on studies of
spirals only, for little was known of the kinematics of
elliptical galaxies. Ellipticals were unstudied, not because
they were considered simple but because the observations
were hard. Motions in spiral galaxies are relatively easy to
study. Orbital motions are circular, the viewing geometry is
well defined, for a given integration time the prominent gas
optical emission lines are detectable with a signal-to-noise
well above that of the absorption lines, and the neutral
hydrogen gas disks are a prime feature for 21-cm studies.
Thus, it is not surprising that during the 1970s, optical
observations with image tube plus photographic plate and
21-cm observations with new interferometers advanced our
knowledge ofthe kinematics ofdisk galaxies to a degree that
was then not matched for elliptical galaxies. But during the
last decade, knowledge ofdynamics of elliptical galaxies has
started to catch up, due, in part, to efficient digital recording
devices and, in part, to the surprise discovery of significant
emission in some ellipticals, emission that is studied in both
the optical and radio spectral regions. Thus, although an
uncertain viewing angle and complicated triaxial orbits keep
the analysis complex, unexpectedly interesting dynamical

results are now available for a significant number ofelliptical
galaxies.
It is now understood that many bright ellipticals are triaxial
in shape, supported not by their (slow or absent) rotation but
by their anisotropic velocity dispersions (27, 28). But some
ellipticals are rotating, albeit slowly. In a recent study of 22
ellipticals, many of them round, Franx, Illingworth, and
Heckman (29) detect rotation >20 km/sec in 21 of them. Of
these, 60% ofthe ellipticals had their kinematical axis within
100 ofthe photometric minor axis. Yet for six ofthe galaxies,
there is significant minor-axis rotation; in two of these the
minor-axis rotation is more rapid than the major-axis rota-
tion; in two ofthe galaxies the minor-axis rotation is approx-
imately equal to the major-axis rotation. Minor-axis rotation
in triaxial galaxies may be a clue to the formation mechanism
of these galaxies but is currently incompletely understood.
In the inner regions, still more complexity exists. Two of
the ellipticals exhibit skew-rotating small nuclear cores,
another has a counter-rotating core, and a fourth has a core
that rotates perpendicular to the rotation ofthe major axis. In
the Franx et al. (29) study, almost 20% of the galaxies
exhibited kinematically distinct nuclear cores, close to the
fraction Rubin and Kenney observe for spirals and ellipticals
in Virgo.
It is not yet possible to translate absorption line kinemat-
ical studies into statistics ofMIL values for more than a few
ellipticals (30, 31). Within a few scale lengths, MIL values
appear generally constant, and often small, an even tighter
constraint on the amount of dark matter (yet this statement
has the ring of statements made early in the history of
observations of spiral galaxies, when it was expected that
more extended observations would show rotation velocities
to fall). Important observational evidence not discussed here,
such as x-ray observations of halos of ellipticals (32) and
velocities of individual planetary nebula far out in ellipticals
(30, 33), offers impressive evidence that elliptical galaxies do
contain extended dark matter components. While absorption
line studies may not be a major contributor to learning about
the distribution ofextended matter in ellipticals, such studies
will surely teach us important details about the kinematics
and evolutionary history of ellipticals.
Velocity Studies from Gas Emission Lines. Among the many
recent surprises coming from the observations of elliptical
galaxies has been the discovery that a significant number
contain gas disks that can be used to map the gravitational
potential. A collection of rotation curves for elliptical galax-
ies, determined from emission lines, is shown in Fig. 4. Also
shown is the predicted velocity curve for a spherical galaxy
with a normal surface brightness profile; the parallel form for
the predicted and observed relations suggests that MIL is
essentially constant over these radial ranges. A more defin-
itive statement concerning values of MIL as a function of
galaxy radius can be made for only a few special galaxies,
including IC 2006.
IC 2006 (35) is an isolated spherical galaxy surrounded by
a counter-rotating outer ring ofHI. Motions within the galaxy
imply a value ofM/LB = 5 for the core; velocities in the outer
ring suggest that M/LB = 16 overall. Just as for spirals, this
increase in mass per unit of luminosity implies the existence
of nonluminous matter that becomes increasingly dominant
with increased radial distance.
But outer gas disks are rare. More common in elliptical
galaxies are inner disks of ionized gas extending only to a
small fraction of the galaxy radius. Often they are dynam-
ically decoupled from the principal rotation of the galaxy.
Such disks-e.g., NGC 5128 (36) and NGC 1316 (37)-are
now understood as the product of the acquisition of a small
gas-rich system. This gas and dust will require numerous
orbital periods to reach dynamical equilibrium and mean-
R/Re
FIG. 4. Rotation curves for six elliptical galaxies plotted as a
function of Re, the half light radius (R = radius) (34). Velocities
(VROT) are measured from their gas disks. The dotted curve is the
rotation curve for a spherical galaxy with a normal r1/4 surface
brightness profile.
while will be observed as a warped and twisted disk with
complex internal kinematics.
NGC 5077 is an elliptical galaxy with a disk of ionized gas
along its minor axis, a not uncommon form for ellipticals. A
detailed study of its gas velocities (34) shows that either
triaxial or oblate mass models with constant MIL can match
the velocity fields of the stars, the gas, and the central
velocity dispersion. Even though the rotation velocities in-
crease slowly from the nucleus, such velocities can arise from
gas in elliptical orbits; it is not necessary to invoke an MIL
increasing with radial distance. Moreover, models with the
gas located in a time-evolving polar ring predict that the ring
becomes enormously twisted, so that lines of sight intersect
the surface two or three times, thus giving rise to the complex
multivalued velocities they observe in the gas.
N-body simulations, too, produce small counter-rotating
nuclear disks. Barnes and Hernquist (38, 39) show that a
prograde merger of two gas-rich stellar disks produces a
galaxy photometrically indistinguishable from an elliptical,
which contains a kinematically distinct but small (=800 pc)
central gas core. Such cores may be skew or counter rotating,
for the tidal trauma has destroyed all knowledge of the
previous orientation ofthe gas. However, the more extended
gas, settled in a disk, rotates prograde.
Many comprehensive reviews of the current state of ob-
servations and understanding of the dynamics of elliptical
galaxies are available (40, 41). It is clear that the freedom
afforded by triaxial orbits combined with viewing angles
offers wide possibilities for models. Adding warped, twisted,
precessing, and time-dependent gas disks increases the range
ofparameters. Arriving at the correct model may be difficult,
for increasingly detailed observations may only add to the
complexity, rather than constrain the range of parameters.
However, the galaxy-by-galaxy studies that are continuing at
an accelerating pace will ultimately make clear the evolu-
tionary importance for ellipticals ofrelatively newly acquired
material; with luck we will also learn about the overall
distribution of matter in elliptical galaxies.
NGC 4550. Recently, I have been studying, in collaboration
with Graham and Kenney (42), the kinematics of spiral and
elliptical galaxies in the Virgo cluster. NGC 4550 is a fea-
tureless E7/SO galaxy located in the core ofthe Virgo cluster
only 1.2° from M87 (Fig. 5). We have discovered a unique
pattern ofstellarand gas kinematics over the inner 30% ofthe
optical galaxy. The stellar disk contains two interpenetrating
components; one system is rotating prograde and one system,
cospatial with the first, is rotating retrograde. A rotating gas
disk is coincident with one ofthe stellar disks. Thus, instead

A
B
o 800U
.4 700
._
.o 600
10 500
, 400
3.0
300,
- NGC 4550
_ *~~~200
_ * 4-r
- 0 200
_ ~ ~~~~>4-rr
S
I s
W
l I
-40 -20
Distance fron
C
oblate spheroids. Thus, it is axisymmetric, flattened by
rotation, with no apparent morphological or other kinemat-
ical peculiarities. Our limited material does not permit a.tl ...i ..:..
detailed model, although absorption line intensities and mea-
P fA v
~~~~sured velocities suggest that the two stellar systems have
jbr^X. i approximately similar dynamical properties, and are approx-
imately coplanar, and absorption line strengths within the
inner few kpc are only slightly stronger in the prograde disk.
Virtually all of the possibilities for the origin of the kine-
.........
.matics in NGC 4550 require the acquisition of secondary
material afterthe initial stellar disk ofNGC 4550 was in place.
We propose the folowing likely scenario. A retrograde gas
mass, probably small compared with the mass ofNGC 4550,
is captured at some angle intermediate between polar and
disk, settles to the disk via dissipation caused partly by
differential precession (44,45), and forms most stars after the.t '. *: ..~~~~~~~ ~~~~~~~~~~~........ ...::
gas has settled to the plane. Such an uncommon system can
be understood as an "equatorial ring" (46)-i.e., a disk, an
alternative end point to the polar ring sequence.
The lack of morphological distortions and the presence of
strong metal absorption lines suggest that the time elapsed
tPA-358° - since the acquisition is at least 109 yr, allowing time for the
_PA=35 gas to come to equilibrium in the combined gravitational
)-inch Ha,[N],[S] emissio potential and for the merger population to form and age. An
neineh Ha.[N] emrissio°n ~ age of 109 yr corresponds to :10 orbital periods at the
-inheHa absorption - half-light radius (16") and to 6 orbital periods at the limits ofneter Ha absorption
our velocity measurements, sufficient to destroy the mor-
phological evidence of the event.
Cospatial stellar systems rotatingboth clockwise and coun-
ft *z, _ terclockwise are more stable than unidirectional disks (S.
Tremaine, private communication) and have been discussed
4 by Araki (47). Earlier, Lynden-Bell (48) imagined a Maxwell
demon who reversed the direction (and violated conservation
t +N
of angular momentum) of one-half the stars in a spherical
_~N cluster to form two interpenetrating stable counter-rotating
| | I- spherical clusters. In Toomre's (49) discussion of axially
0 20 40 symmetric flattened isothermal galaxy models, he investi-
n nucleus, arcsec gated sets of models in which one-half of the stars rotate
prograde, one-half rotate retrograde, noting that nothing in
the analysis restricts the disk stars to orbit only prograde.
The discovery of galaxies like NGC 4550 confirms yet
again the enormous variety within elliptical galaxies;
Toomre's (49) models are no longer just the "elegant curi-
osities" he imagined. I envy observers living within NGC
4550 for the interesting stellar dynamics they are encounter-
ing.
FIG. 5. (Top) Virgo galaxies NGC 4550 and NGC 4551 from an
r-band image taken by C. Bailyn and Y.-C. Kim at the Cerro Tololo
InterAmerican Observatory 0.9-m telescope. There is no evidence of
morphological peculiarity in either galaxy. (Middle) Major-axis ve-
locities in NGC 4550, from the stellar Ha absorption (open symbols)
and from the counter-rotating emission disk (filled symbols). The
counter-rotating stellar disk absorption at Ha is masked by the
emission, except for that barely showing at -15' N. (Bottom)
Absorption lines from prograde and retrograde stellar disks (here
shown dark) in the region blueward of[OIII]5007 (bright line at right).
Note the X-shaped appearance of the absorption lines as they cross
the nucleus.
of a normal spectrum composed of many single absorption
lines, we observe each absorption line to be double, with the
two members opposite in velocity and crossing at the nu-
cleus. The two cospatial counter-rotating disks extend over
a sizable portion of the galaxy.
A value of V/lo = 1.4, coupled with an ellipticity, 1 - b/a
= 0.7, places NGC 4550 in the Binney (43) region of rotating
Section V. M/L 100 for the Perseus Cluster,
a "Low" Value
Nearly 60 years ago, Zwicky (50) and Smith (51) discovered
that individual galaxies in the Coma and Virgo clusters of
galaxies have velocities so large with respect to the mean
cluster velocity that their mutual gravitational attraction
would not be sufficient to bind the clusters. Hence, either
clusters are expanding and galaxies are dispersing (contrary
to observation) or the total mass in the clusters is larger than
that implied by the luminous material. Zwicky called this the
"missing mass." For many years, this conclusion was rele-
gated to the "things we don't understand" file and assumed
to be a mysterious property of galaxy clusters. Not until we
discovered the need for dark matter in individual galaxies did
these early observations join the mainstream ofastronomical
research.
Now that we recognize that dark matter is the major
constituent ofclusters, we must quantify the amount ofdark
matter if we are to set observational limits on the mass and
the ultimate fate of the universe. Is MIL averaged over a
cluster as large as 500 or 1000, implying that MIL rises with
the size ofthe region included, or is it closer to the value 100

derived for binaries and groups? During the last decade or
two, attempts to answer this question were made generally
via a virial analysis, an analysis that assumes galaxy mem-
bership in the cluster is known, that the distribution of mass
follows the distribution ofgalaxies, that the cluster is relaxed,
that the galaxy velocity distribution is isotropic, and that the
second time derivative ofthe moment ofinertia is zero (which
would not be the case, for example, if galaxy halos were
centered elsewhere than at the galaxy centers). Some years
ago, from optical data for the Perseus cluster, Kent and
Sargent (52) derived the large value of MIL 300.
In an important recent work, Eyles et al. (53) analyze x-ray
images of the Perseus cluster obtained with spatial resolu-
tions of 12' and 3' in the energy range 2.5-32 keV (1 eV =
1.602 x 10-19 J) during the Spacelab 2 mission. Because the
x-ray-emitting gas in a cluster is expected to be in hydrostatic
equilibrium, kT plays the role ofmV2 in the virial analysis and
is a good tracer of the cluster gravitational potential. From
the x-ray data, Eyles et al. (53) find that the total gravitational
cluster mass is 30-65% less than that obtained from the
application of the virial theorem, in part because the distri-
bution of mass does not follow the distribution of galaxies.
Moreover, while the dark matter is more centrally condensed
than the galaxies and is the dominant mass component out to
=1 Mpc, the x-ray gas is less so. The fraction ofthe total mass
that is in the form of the x-ray-emitting gas increases with
radius, dominating beyond 1 Mpc and reaching >55% of the
total mass and still rising at the limits of the observations.
As a result of this analysis, M/Lv is of order 100 at large
radii (rather than 300), with values between 55 and 145
permitted by the range of models and of statistical and
systematic errors. Although based on only a single cluster,
this result strengthens earlier suggestions (54) that there is not
strong evidence for very large values of MIL for systems of
very large size. New x-ray observations from ROSAT sat-
ellite are expected to enlarge the sample ofgalaxy clusters for
which determinations of the mass distribution and, conse-
quently, MIL are available.
Conclusions: MIL 100 Wi Saisfy,Al
Dynamical Observations
To discuss the observed dynamical values of MIL and the
mass density ofthe universe, I start with a few definitions. I
adopt Ho = 50 km sec-l-Mpc-1. Then:
Critical mass density Pc = 3H2/8rG 5 x 10-30 g.cm3
7 x 1010 MG-Mpc-3,
Present mass density p = 3HW2/8irG, where Q = P/Pc,
Luminosity density LB = 108 LO-Mpc3.
Hence, MILB = p/LB 750 Q, where the blue luminosity
density, surely an uncertain number, is takenfrom Efstathiou
et al. (55). Thus, a value of M/LB near 750 will close the
universe-i.e., Q will be unity, and the gravitational attrac-
tion of the mass within the universe will just halt the expan-
sion. In the values that follow, I have transformed published
values to the adopted Ho and to the blue spectral band.
It had earlier been suspected that observed values ofMIL
increase with increased enclosed volume. However, current
available data suggest that this is not so, and M/L 100 is
a limiting value required by dynamical systems on scales
from 0.1 to 200 Mpc (except the highly uncertain values
derived in studies of large-scale motions-e.g., ref. 56). This
situation is illustrated in Fig. 6, a plot of the dynamically
determined M/L values versus the appropriate radial scale
for our galaxy, galaxies, binary galaxies, the Local Group,
clusters of galaxies, and for gravitationally induced infall
motions. For scales larger than galaxies-i.e., >100 kpc, all
systems have a constant ratio ofMIL. On the smaller scales
(galaxies), the increasing dominance of the dark matter
produces values of MIL that increase with radial scale. All
Radial scale, kpc
FIG. 6. Mass-to-blue light ratio as a function of the size of the system for galaxies, binaries, clusters, and larger systems, based loosely on
and extended from Oemler (57). References to individual values come from Rubin (3, 4); Forman et al. (32); Kulessa and Lynden-Bell (9);
Schweizer (58); Huchra and Geller (59); Cowie et al. (60); Eyles et al. (53); Peebles (13); van der Kruit (11); Davis and Peebles (61); Aaronson
et al. (62); and Schramm (63); values are all transformed to H. = 50 km sec-l-Mpc-1 and blue magnitudes. A value of MIL = 750 is equivalent
to Q = 1. [For H. = 100 km sec- Mpc-1 values of MIL increase by a factor of 2, and MIL (closure) = 1500. The form of the plot relative to
[l remains unchanged]. Error bars are generic estimates. Note that beyond =100 Mpc, values of M/L do not continue to rise but are constant
near 100, independent ofthe size ofthe dynamical system. Dwarfspirals and irregulars, not plotted here, have MIL values ranging from =20-50
and radii near 10 kpc. However, their poorly known distances introduce large uncertainties in both quantities.

values fit comfortably within a universe in which the lumi-
nous mass in galaxies comprises only 1% of the closure
density; the mass required by the dynamics is ten times this
or 10% of the closure density, but not higher. Moreover,
these values do not exceed the nucleosynthetic limit on the
mass density of baryons in the universe (e.g., ref. 63).
Thus, observations tell us that we may inhabit a low-
density universe in which the gravitational attraction of the
enclosed baryonic matter is not sufficient to halt the expan-
sion. However, theoretical cosmologists like Guth and
Schramm and Turner tell us that models require that Q is
unity, that the additional 90% of the universe is still unde-
tected, and that it is required to consist ofdark, nonbaryonic
matter. To reconcile the views of the observers with the
dreams of the theorists will require heroic observational,
laboratory, and theoretical efforts but will constitute a fun-
damental step in teaching us more about the complicated
universe in which we find ourselves.
I thank my colleagues John Graham, Jeff Kenney, and Francois
Schweizer for valuable help and support; Kitt Peak and Palomar
Observatories for telescope time; and Drs. F. Bertola, F. Briggs, S.
Casertano, and J. Kormendy for permission to reproduce their
figures.
1. Ostriker, J. P., Peebles, P. J. E. & Yahill, A. (1974) Astrophys.
J. Lett. 193, Li-L4.
2. Ostriker, J. P. & Peebles, P. J. E. (1973) Astrophys. J. 186,
467-480.
3. Rubin, V. C., Ford, W. K. & Thonnard, N. (1978) Astrophys.
J. Lett. 225, L107-L111.
4. Rubin, V. C., Ford, W. K. & Thonnard, N. (1980) Astrophys.
J. 238, 471-487.
5. Bosma, A. (1978) Dissertation (Univ. ofGroningen, The Neth-
erlands).
6. Faber, S. M. & Gallagher, J. S. (1979) Annu. Rev. Astron.
Astrophys. 17, 135-187.
7. Kormendy, J. & Knapp, G. R., eds. (1987) in Dark Matter in
the Universe (Reidel, Dordrecht, The Netherlands).
8. Trimble, V. (1987) Annu. Rev. Astron. Astrophys. 25, 425-472.
9. Kulessa, A. S. & Lynden-Bell, D. (1992) Mon. Not. R. Astron.
Soc. 255, 105-118.
10. Schmidt, M. (1985) The Milky Way Galaxy, Int. Astron. Union
Symposium 106 (Reidel, Dordrecht, The Netherlands), pp.
75-84.
11. van der Kruit, P. C. (1990) The Milky Way as a Galaxy,
Saas-Fee Advanced Course 19 (Univ. Science Books, CA), pp.
185-226.
12. Zaritsky, D., Smith, R., Frenk, C. & White, S. D. M. (1993)
Astrophys. J., in press.
13. Peebles, P. J. E. (1990) Astrophys. J. 362, 1-13.
14. Kamphuis, J. & Briggs, F. (1992) Astron. Astrophys. 253, 335.
15. Kormendy, J. (1988) in Origin, Structure, and Evolution of
Galaxies, ed. Zhi, F. L. (World Sci., Singapore), pp. 252-280.
16. van Albada, T. S., Bahcall, J. B., Begeman, K. & Sancisi, R.
(1985) Astrophys. J. 295, 305-313.
17. Kent, S. M. (1988) Astron. J. 96, 514-527.
18. Athanassoula, E., Bosma, A. & Papaioannou, S. (1987) Astron.
Astrophys. 179, 23-40.
19. Kalnajs, A. J. (1985) in The Internal Kinematics andDynamics
of Galaxies, ed. Athanassoula, E. (Reidel, Dordrecht, The
Netherlands), pp. 87-88.
20. Casertano, S. & van Gorkom, J. H. (1991) Astron. J. 101,
1231-1241.
21. Whitmore, B. C., Forbes, D. A. & Rubin, V. C. (1988) Astro-
phys. J. 333, 542-560.
22. Rubin, V. C. & Kenney, J. D. P. (1991) Bull. Am. Astron. Soc.
23, 1339 (abstr.).
23. Carignan, C. & Freeman, K. C. (1988) Astrophys. J. Lett. 332,
L33-L36.
24. Carignan, C. & Beaulieu, S. (1989) Astrophys. J. 347, 760-770.
25. Broeils, A. H. (1992) Astron. Astrophys. 256, 19-32.
26. Milgrom, M. (1983) Astrophys. J. 270, 365-389.
27. Illingworth, G. (1977) Astrophys. J. Lett. 218, L43-L47.
28. Davies, R. L., Efstathiou, G., Fall, S. M., Illingworth, G. &
Schechter, P. L. (1983) Astrophys. J. 266, 41-57.
29. Franx, M., Illingworth, G. & Heckman, T. (1990) Astrophys. J.
344, 613-636.
30. Kent, S. M. (1990) in Evolution ofthe Universe ofGalaxies, ed.
Kron, R. G. (Brigham Young Univ., Provo, UT), pp. 109-118.
31. de Zeeuw, T. (1992) in Morphological and Physical Classifi-
cation ofGalaxies, eds. Longo, G., Capaccioli, M. & Busarello,
G. (Kluwer, Dordrecht, The Netherlands), pp. 139-151.
32. Forman, W., Jones, C. & Tucker, W. (1985) Astrophys. J. 293,
102-119.
33. Ciardullo, R., Jacoby, G. H. & Ford, H. C. (1989) Astrophys.
J. 344, 715-725.
34. Bertola, F., Bettoni, D., Danzinger, J., Sadler, E., Sparke, L.
& de Zeeuw, T. (1991) Astrophys. J. 373, 369-390.
35. Schweizer, F., van Gorkom, G. H. & Seitzer, P. (1989) Astro-
phys. J. 338, 770-788.
36. Graham, J. A. (1979) Astrophys. J. 232, 60-73.
37. Schweizer, F. (1980) Astrophys. J. 237, 308-318.
38. Barnes, J. E. & Hernquist, L. E. (1991) Astrophys. J. 370,
L65-L68.
39. Hernquist, L. E. & Barnes, J. E. (1991) Nature (London) 354,
210-212.
40. Wielen, E. R., ed. (1990) in Dynamics and Interactions of
Galaxies (Springer, Heidelberg).
41. de Zeeuw, T. & Franx, M. (1991) Annu. Rev. Astron. Astro-
phys. 29, 239-274.
42. Rubin, V. C., Graham, J. A. & Kenney, J. P. D. (1992) Astro-
phys. J. 394, L9-L12.
43. Binney, J. (1978) Mon. Not. R. Astron. Soc. 183, 501-514.
44. Sparke, L. S. (1986) Mon. Not. R. Astron. Soc. 219, 657-670.
45. de Zeeuw, T. (1990) in Dynamics andInteractions ofGalaxies,
ed. Wielen, E. R. (Springer, Heidelberg), pp. 263-266.
46. Schechter, P. L. (1990) in Dynamics and Interactions ofGal-
axies, ed. Wielen, E. R. (Springer, Heidelberg), pp. 508-513.
47. Araki, S. (1987) Astron. J. 94, 99-105.
48. Lynden-Bell, D. (1960) Mon. Not. R. Astron. Soc. 120, 204-
213.
49. Toomre, A. (1982) Astrophys. J. 259, 535-543.
50. Zwicky, F. (1933) Helv. Phys. Acta 6, 110.
51. Smith, S. (1936) Astrophys. J. 83, 23.
52. Kent, S. M. & Sargent, W. L. W. (1983)Astron. J. 88,697-708.
53. Eyles, C. J., Watt, M. P., Bertram, D., Church, M. J., Pon-
man, T. J., Skinner, G. K. & Willmore, A. P. (1991)Astrophys.
J. 376, 23-32.
54. Davis, M. (1987) in Dark Matter in the Universe, eds. Kor-
mendy, J. & Knapp, G. R. (Reidel, Dordrecht, The Nether-
lands), pp. 97-109.
55. Efstathiou, G., Ellis, R. S. & Peterson, B. A. (1988) Mon. Not.
R. Astron. Soc. 232, 431-461.
56. Guth, A. H. (1993) Proc. Natl. Acad. Sci. USA 90, 4871-4877.
57. Oemler, A. (1988) in The Minnesota Lectures on Clusters of
Galaxies and Large-Scale Structure, ed. Dickey, J. M.
(Brigham Young Univ., Provo, UT), pp. 19-38.
58. Schweizer, L. (1988) Astrophys. J. Suppl. Ser. 64, 427-457.
59. Huchra, J. & Geller, M. (1982) Astrophys. J. 257, 423-437.
60. Cowie, L. L., Henriksen, M. & Mushotzky, R. (1987) Astro-
phys. J. 317, 593-600.
61. Davis, M. & Peebles, P. J. E. (1983) Astrophys. J. 267, 465-
482.
62. Aaronson, M., Huchra, J., Mould, J., Schechter, P. L. & Tully,
R. B. (1982) Astrophys. J. 258, 64-76.
63. Schramm, D. N. (1993) Proc. Natl. Acad. Sci. USA 90, 4782-
4788.

Galaxy dynamics and the mass density of the universe

More Related Content

What's hot (20)

Viewers also liked (13)

Similar to Galaxy dynamics and the mass density of the universe (20)

More from Sérgio Sacani (20)

Recently uploaded (20)

Galaxy dynamics and the mass density of the universe