Download Small galaxies are growing smaller

Galactic surveys
1: The giant spiral galaxy NGC 1365 – what we
often think of as a “typical” galaxy. (ESO VLT)
2: The Small Magellanic Cloud, an irregular galaxy
once thought to represent the faint end of the
distribution of galaxy luminosities. (AAO)
3: The dwarf elliptical galaxy M32 superimposed on
the disc of its giant neighbour M31, showing the
contrast in sizes (INT image by David Malin). M32
has a truncated light distribution suggesting outer
stars may have been stripped away by tidal forces.
Small galaxies are
New telescope technology and major sky surveys are finding more and more
dwarf galaxies. Steve Phillipps discusses how this growing population may
hold the clues to understanding a range of galaxy collisions and interactions.
G
alaxies are supposed to be huge aggregates of hundreds of billions of stars and
have the spectacular appearance we see
in “coffee table” books. Indeed, in the early days
of extragalactic astronomy it was thought that
most galaxies were of rather similar (large) luminosities. Specifically, Edwin Hubble (1936) and
his contemporaries believed that the luminosity
function (LF) of galaxies, that is the number of
galaxies per unit volume of different luminosities, was a peaked, roughly Gaussian shaped
curve. Known local galaxies spanned the range
from the giant spiral M31 in Andromeda and
our own galaxy, with luminosities of a few
1010 L (absolute magnitudes MV ≈ –21),
through smaller irregular systems like the Small
Magellanic Cloud (MV ≈ –17), down to the
dwarf elliptical galaxy companion to M31,
NGC 147, about 250 times less luminous than
M31 itself (i.e. MV ≈ –15). This was a very much
smaller range than seen for stars: even main
sequence stars have a range of at least a factor
108, with the Sun (logarithmically speaking)
somewhere near the middle. We might note
though that the range of stellar masses is quite
narrow, only around a factor 103.
This galactic uniformity was disturbed in
1938 when Hubble’s great rival Harlow Shapley
reported the discovery of two very faint Local
Group galaxies, known as the Fornax and
Sculptor dwarfs, after the constellations in
whose direction they appear to lie. Strictly
6.6
speaking they were apparently first noted by
Harvard College Observatory assistant
astronomer Sylvia Mussells on plates taken by
Shapley (see A&G 2003 45 1.18). Sculptor, in
particular, was very much less luminous (by
another factor ~100) than previously known
galaxies – with modern distance estimates it has
MV ≈ –10 – and these two were the first examples of dwarf spheroidal galaxies.
They were discovered only because they were
close enough for individual bright stars in them
to be resolved – indeed, they are satellites of our
own galaxy. Their averaged-out surface brightness (i.e. luminosity per unit area) would have
been far too low to be detectable on the photographic emulsions available in the 1930s. Even
at a dark site, the night sky brightness – due to
a combination of airglow in our atmosphere,
sunlight reflected off dust grains and the general
stellar light in our galaxy, among other things
– is an order of magnitude brighter than even
the central regions of a dwarf spheroidal. Such
dwarfs therefore suffer a problem generic to all
low surface brightness galaxies, that is a poor
contrast against the overall sky brightness. This
difficulty is epitomized by the Fornax dwarf
itself, which has a total magnitude around
mV = 7.3. A star of this magnitude would be visible with a pair of binoculars, let alone a long
photographic exposure on a large telescope.
The arrival of dwarf spheroidals on the scene
implied that galaxies could be much more
Abstract
Galaxies are not always giant collections
of billions of stars. Since the 1930s, when
Harlow Shapley discovered the first dwarf
spheroidal galaxies, technology has
allowed the detection of ever fainter
galaxies in our immediate neighbourhood.
Our galaxy is now known to have a whole
retinue of very small satellite galaxies, the
lowest luminosity examples of which can
hardly outshine one massive star. Some
galaxies appear to be getting physically
smaller. Evidence for this is found in the
streams of stars detected around our
galaxy and elsewhere and in galaxies that
appear to have had their outer regions
truncated. Recent surveys of galaxy
clusters have revealed another new class of
object, the ultra-compact dwarfs. Though
no less luminous than other dwarf
galaxies, their physical sizes, of order
20 pc, are far below anything previously
seen. They are reminiscent of the nuclei of
dE,N type galaxies and may well be
descended from them via some destructive
processes within galaxy clusters.
varied than hitherto believed. As Shapley (1943)
himself put it, “two misty patches have put us
in a fog”. In the 1950s Fritz Zwicky proposed
an alternative view. Zwicky was famous for
usually having an alternative view and in this
case it was that – as with most things astronomical, from stars to meteorites – there should
be a steady increase in numbers towards smaller
December 2004 Vol 45
Galactic surveys
4: The Fornax dwarf spheroidal galaxy, first
reported by Harlow Shapley in 1938. (UKST)
5: The very faint and diffuse Carina dwarf
spheroidal was first seen in 1977 on the UKST
plate reproduced here.
6: And IX, the most recently discovered of M31’s
companions and the lowest luminosity galaxy
currently known. (INT image from Zucker et al.
2004 ApJ)
growing smaller
objects. Support for this view, with the LF rising at the faint end, soon came from Kiang
(1961) and others, from consideration of local
galaxies, and from George Abell (1962) for
galaxies in clusters. Subsequent work has confirmed this in general terms, though the slope
of the faint end of the LF remains a contentious
issue. This is primarily because of the strong
selection effects biasing us against including low
luminosity, low surface brightness objects in
galaxy catalogues or other galaxy samples. A
similar effect had been seen for stars; the apparently brightest stars seen in the sky include a
preponderance of stars much brighter than the
Sun, but a volume-limited sample of nearby
stars reveals that most stars are really considerably less luminous than the Sun. As we have
seen, for dwarf galaxies this effect is compounded by their low surface brightnesses.
Unsurprisingly then, further detections of very
small and/or low luminosity galaxies remained
confined to the local universe, indeed to the
Local Group of galaxies. Further examples of
dwarf spheroidals were discovered in the 1950s
by Harrington and Wilson in the Palomar
Schmidt Sky Surveys. Again these were immediate companions (satellites) of our own galaxy,
less than 200 kpc away and even smaller and
fainter than Fornax. Indeed, Wilson’s (1955)
Ursa Minor and Draco dwarf spheroidals were
even fainter than Sculptor and remained the
faintest known galaxies for many years.
By 1977, when Russell Cannon, Tim
Hawarden and Sue Tritton discovered the
remarkably inconspicuous Carina dwarf spheroidal on a plate taken for the UK Schmidt
Telescope (UKST) southern sky survey, our
December 2004 Vol 45
galaxy had acquired the correct fairy-tale complement of seven dwarfs as companions. Carina
itself has a luminosity 5 × 105 L and an effective radius (i.e. the radius containing half of its
total light) about 250 pc. The galactic satellite
with the lowest luminosity is still the Ursa
Minor system (MV ≈ –8.5, L ≈ 2 × 105 L), while
the dwarf satellite with the lowest surface
brightness (as judged from star counts) is probably Draco (equivalent to 26.3B magnitudes per
square arc second, or 2 L/pc2 at the centre).
However, the physically smallest of our galaxy’s
dwarf companions is Leo II with an effective
radius re ≈ 170 pc. It is more luminous than the
likes of Draco or Carina because of its rather
higher surface brightness. (Excellent surveys of
the properties of the galaxy’s dwarf satellites are
given by Irwin and Hatzidimitriou [1995],
Mateo [1998] and van den Bergh [2000].)
We should also note that M31 has its own retinue of satellites. For many years the only ones
known were the quite moderate brightness
dwarf ellipticals M32, NGC 205, NGC 185 and
the fainter NGC 147. And I, II and III were discovered by Sydney van den Bergh in 1972, but
M31 still lagged the galaxy in terms of the number of companions until quite recently. (The
galaxy now has 12 known satellites, including
the Magellanic Clouds.) There was always the
suspicion that this shortfall might be due to the
greater difficulty of detecting faint dwarfs at the
distance of M31 (≈ 800 kpc) and the last few
years has seen the addition of And V up to
And IX (the proposed And IV is now thought
not to be a genuine galaxy). And IX, which was
detected only in the last year (Zucker et al.
2004) is probably of even lower luminosity than
the Ursa Minor dwarf, with an absolute magnitude MV ≈ –8.3, though this is still somewhat
uncertain. And IX was discovered as an excess
in projected stellar density in the Sloan Digital
Sky Survey (SDSS) data just 2.6° (a projected
distance of 34 kpc) from the centre of M31.
Allowing for the contribution of fainter, unseen
stars, the central surface brightness also seems
to be the lowest yet recorded, at 26.8 V magnitudes per square arc second (less than 1 L /pc2).
M32 remains the oddest of the companions,
though, because of its high central surface
brightness but comparatively small size. In some
ways it looks like a “chopped down” version of
a giant galaxy and it has been hypothesized that
it may represent just the central remnant of a
once much bigger galaxy that lost most of its
outer stars via gravitational tidal effects as it
passed close by M31.
If we could reverse our viewpoint and look at
our galaxy from Andromeda, we might see a
rather similar situation. In 1997 Rodrigo Ibata,
Gerry Gilmore and Mike Irwin discovered a
coherent stream of stars in nearly the same direction as, but moving at different radial velocities
to, those in the galactic centre region. This was
identified as a companion galaxy to our own,
actually plunging through the galaxy’s disc on
the opposite side to the Sun and only 12 kpc
from the centre. A huge plume of stars trails
(and precedes) the Sagittarius dwarf spheroidal
(as it became known), clear “smoking gun” evidence for the tidal destruction of satellite galaxies. As the loss of stars is preferentially from the
less tightly bound outer regions, we can anticipate that Sagittarius will end up looking like a
rather small remnant of its former self.
6.7
Galactic surveys
7: The Sagittarius is
currently plunging
through the galaxy’s disc,
on the opposite side to
where the Sun lies. The
huge trail of tidally
removed stars can be
seen projecting nearly at
right angles to the
galactic plane in this star
count plot. (From
Majewski et al. 2003 ApJ)
8: The nucleated dwarf elliptical FCC 303 in the
Fornax Cluster, at ground-based resolution. (UKST)
Interestingly, the supposed galactic globular
cluster M54 (discovered in 1770!) turns out to
be associated with the Sagittarius dwarf and it
has been suggested that it may be its nucleus.
Another stream, but with little evidence for any
surviving core, has recently been discovered and
named the Canis Major dwarf.
Returning to the detection of dwarf galaxies as
such, improved photographic technology in the
1980s made it possible to detect very low luminosity galaxies at greater distances, for example
in the Virgo and Fornax Clusters around
16–20 Mpc away. Studies of Virgo by Allan
Sandage, Bruno Binggeli and Gustav Tammann,
and of Fornax by Sandage and Harry Ferguson,
revealed many low surface brightness objects
looking just like Local Group dwarf ellipticals
(though some contain central nuclei), dwarf
irregulars and even dwarf spheroidals, the numbers implying a quite steeply rising LF. However,
at the really low luminosity end we are relying
on statistics (more objects per unit area in the
cluster direction than away from the cluster) and
on circumstantial evidence (whether they look
like previously known dwarfs) rather than any
direct, individual identification as a cluster
dwarf. The same is true, perhaps even to a
greater extent, with the more modern CCDbased studies of both these nearby clusters and
more distant ones, since it becomes increasingly
difficult to distinguish a nearby very faint dwarf
from very distant normal luminosity galaxies
that merely look faint (e.g. Driver et al. 1999,
Trentham and Hodgkin 2002). Even if we can
identify the usual types of dwarf, there is also the
possibilty of low luminosity galaxies that do not
look like those found in the Local Group.
A way around both of these problems is to
obtain individual distances to candidate dwarfs
in the direction of a cluster. This is most easily
achieved by measuring redshifts. A problem
with this is clearly the potentially large number
of galaxies that must be checked. However, technology has again come to our aid in the last
6.8
decade with the development of multi-object
spectroscopy. With the “2 degree Field” (2dF)
spectrograph on the Anglo-Australian Telescope,
for instance, we can now obtain spectra for 400
objects simultaneously. This has enabled the present generation of huge galaxy redshift surveys,
such as 2dFGRS with 220 000 redshifts (Colless
et al. 2001), or the SDSS with plans for one million redshifts (York et al. 2000).
However, taking redshifts for all the galaxies
we see down to some magnitude limit solves
only half the problem – that is, we can tell which
of the candidate faint galaxies really are in the
cluster not the background. But at the distance
of Virgo and Fornax, the nearest clusters,
200 pc subtends only 2″. Given the blurring
effect of the atmosphere, (“seeing”) a small
dwarf at these distances will be virtually unresolved and will certainly appear pretty much like
any other image 2″ in radius, regardless of any
true differences in structure. Any even smaller
objects will not only be indistinguishable from
background galaxies, but also will look like
unresolved stellar images in ordinary groundbased imaging. They will, therefore, not be
included in galaxy redshift surveys at all. The
only way out of this impasse is the extreme one
of taking spectra for all the objects in the direction of the cluster, regardless of whether they
look like galaxies or not. This has a very large
overhead since at the magnitudes generally
accessible to the large surveys (say down to
mB ~ 20), stars outnumber galaxies by a sizeable
factor. Nevertheless, again taking advantage of
multi-object spectroscopy it is possible to undertake an “all-object” survey of a cluster area.
The Fornax Cluster Spectroscopic Survey
(Phillipps 1997, Drinkwater et al. 2000a) set out
to do just this. It covers some 9 square degrees
of sky which contains about 10 000 objects
between mB = 16.5 and mB = 20. The large majority are stars or background galaxies, with a selection of cluster galaxies thrown in. Given the
magnitude limits and the distance to Fornax, the
latter are all dwarf galaxies. In fact, we would
expect many more dwarf galaxies down to our
magnitude limit than appear in the 2dF sample,
but many of the candidates are of too low a surface brightness for a spectrum to be measurable.
However, the exciting discovery came at the other
extreme. Six of the objects classified as “stars” in
the input catalogues (created by scanning UKST
photographic plates with the APM machine in
Cambridge) turned out to have redshifts that
placed them clearly at the distance of the cluster.
Despite luminosities of several million L
(–13.5 < MB < –11), these had to be dwarf cluster
members with sizes no more than 100 pc or so
(Drinkwater et al. 2000b, Phillipps et al. 2001).
Threshing, shredding and harassment
In fact, Hubble Space Telescope imaging showed
them to be even smaller, with half light radii
around 20 pc, almost – but, crucially, not quite
– as small as globular clusters (which have
re ≈ 5 pc). In addition, higher resolution spectroscopy from the ESO Very Large Telescope
and from the Keck Telescope indicated velocity
dispersions around 25 km s–1, also intermediate
between the values for known dwarf galaxies
and globular clusters. This was clearly a previously untenanted area of parameter space, making these a whole new species of small stellar
system, which we christened ultra-compact
dwarfs (UCDs; Phillipps et al. 2001, Drinkwater
et al. 2003). Their origin may be debated, of
course. They could be merely the extreme, most
massive end of the distribution of globular clusters, but the brighter UCDs are 10 times more
luminous than the brightest globular clusters
seen around our galaxy. They also seem to have
somewhat larger mass-to-light ratios. They
could be a totally new type of dwarf galaxy
formed early in the development of structure in
the universe along with the other dwarf types.
Or they could be a new variety descended from
a different galaxy type, perhaps via some of the
destructive mechanisms we discussed above for
December 2004 Vol 45
Galactic surveys
9: The Fornax Cluster showing the positions of the ultra-compact dwarfs (UCDs) discovered during our Fornax
Cluster Spectroscopic Survey. High-resolution HST images of five of the UCDs are shown, along with an
image of the nucleus of FCC 303 (c.f. figure 8). The UCDs are comparable in size to this nucleus and about
1000 times smaller than NGC 1365, also a member of the Fornax Cluster, seen in figure 1. (Image by M Hilker
and A Karlick [Fornax Cluster Spectroscopic Survey team]; original images from the Curtis Schmidt and HST)
M32 and the Sagittarius dwarf spheroidal.
One of these goes by the name of “threshing”
(there are also “shredding” and “harassment”
among others) and invokes the tidal stripping
of the outer stars from a former nucleated
dwarf elliptical (dE,N) galaxy as it passes
through the central regions of the cluster, past
the first ranked (most massive) central cluster
member (NGC 1399 in the case of Fornax). This
may leave only the nucleus itself orbiting the
centre of the cluster. Support for this sort of
mechanism has come from the more recent
detection (Jones et al. 2004) of very similar
UCDs in the middle of the Virgo Cluster, in the
region around M87, but not in the outskirts of
clusters. If UCDs are indeed such remnants then
we might expect their distribution of masses or
luminosities to follow that of the nuclei of
dE,Ns. This appears to peak at luminosities
much lower than those of the original UCDs
(around MB = –9), so a deeper search could be
expected to reveal many more UCDs.
Another search in Fornax, again using 2dF but
targeting stellar-looking objects a magnitude
fainter than before, did turn up a population of
fainter cluster members (Drinkwater et al.
2004), but there were even more of them than
expected: 46, in fact. On further consideration,
though, it is clear that we are now in the magnitude range where we will also pick up the
bright globular clusters around the central cluster galaxy NGC 1399 (which was already
known to have a large population of globulars).
Thus telling apart large globulars from small
UCDs becomes problematic, if indeed they are
December 2004 Vol 45
two independent types of object at all. (Coming
the other way, there have been suggestions [e.g.
Meylan et al. 2001] that the galaxy’s largest
“globular cluster”, ω Cen, may really be the
remnant of a former galaxy, too.) The distribution of the spectroscopically confirmed faint
UCD candidates does, in fact, suggest two populations of objects as some are closely clustered
around NGC 1399 itself (as are its globulars, of
course), while the others appear to be spread
throughout the central regions of the cluster, like
the brighter UCDs. If at least some of the new
objects are faint UCDs then, judging by the previous HST imaging, we might expect them to be
the physically smallest galaxies yet detected.
Clearly our view of the contents of the universe has changed dramatically since Shapley’s
“two misty patches” came on the scene. It is
now evident that low luminosity/low mass
galaxies in fact outnumber their more spectacular giant cousins. However, the factor by
which they do so – often called the dwarf-togiant ratio – remains to be settled. Not only is
there a large population of diffuse, hard to see,
low surface brightness dwarfs out there, but
now we have to contend with compact, high
surface brightness dwarfs masquerading as
stars, as well. Regardless of the fine detail, the
existence – and large numbers – of the varied
types of dwarf galaxy are already providing us
with vital clues and constraints on both the
original formation mechanisms for galaxies and
the dynamical, evolutionary processes that take
place during their lifetimes. The humble dwarf
is no longer to be ignored! ●
10: An illustration of a computer simulation of the
destruction (“threshing”) of a nucleated dwarf (such
as that in the upper inset) to produce an ultracompact dwarf (as in the lower inset). The trail of
tidally stripped debris is superimposed on an image
of the Fornax Cluster. (Created by U Queensland
Communications for the Fornax Cluster
Spectroscopic Survey team. Simulation by K Becker)
Steve Phillipps is a reader in astrophysics at the
University of Bristol.
Acknowledgments. The author would like to thank
his colleagues on the Fornax Cluster Spectroscopic
Survey and especially those involved in the UCD
work, Michael Drinkwater, Bryn Jones, Michael
Gregg, Kenji Bekki, Warrick Couch, Katya
Evstigneeva, Harry Ferguson, Michael Hilker, Russell
Jurek, Arna Karick, Quentin Parker, Rodney Smith
and Terry Bridges.
References
Abell G 1962 in Problems in Extragalactic Research (McMillan, NY).
Cannon R D et al. 1977 MNRAS 180 81.
Colless M et al. 2001 MNRAS 328 1039.
Drinkwater M J et al. 2000a A&A 355 915.
Drinkwater M J et al. 2000b PASA 17 227.
Drinkwater M J et al. 2003 Nature 423 519.
Drinkwater M J et al. 2004 PASA in press.
Driver S P et al. 1998 MNRAS 301 369.
Ferguson H and Sandage A 1988 AJ 96 1520.
Harrington R G and Wilson A G 1950 PASP 62 118.
Hubble E P 1936 The Realm of the Nebulae (Yale University Press,
New Haven).
Ibata R et al. 1994 Nature 370 1941.
Irwin M J and Hatzidimitriou D 1995 MNRAS 277 1354.
Jones J B 2004 AJ submitted.
Kiang T 1961 MNRAS 122 263.
Majewski S R et al. 2003 Ap J 599 1082.
Meylan G et al. 2001 AJ 122 830.
Phillipps S 1997 Wide-Field Spectroscopy (Kluwer, Dordrecht) 281.
Phillipps S et al. 1985 AJ 90 1759.
Shapley H 1938 Harvard Observatory Bulletin no. 908 1.
Shapley H 1943 Galaxies (Blakiston, Philadelphia).
Trentham N and Hodgkin S 2002 MNRAS 333 423.
van den Bergh S 1972 ApJ 171 L31.
van den Bergh S 2000 The Galaxies of the Local Group (CUP).
Wilson A G 1955 PASP 67 27.
York D G et al. 2000 AJ 120 1579.
Zucker D B et al. 2004 ApJ 612 121.
Zwicky F 1957 Morphological Astronomy (Springer, Berlin).
6.9