To appear in "Chemical Enrichment of the ICM and IGM",
ASP Conference Series, Vol, xxx, 2001, eds. F. Matteucci and
R. Fusco-Femiano.
astro-ph/0107116
Abstract. I provide an observational review of the properties of starburst-driven galactic superwinds, focusing mainly on recent results pertinent to the transport of metals and energy into the IGM. Absorption-line studies are providing rich kinematic information on both neutral and ionized gas in superwinds, with observed mass flow rates similar to the star formation rate and outflow velocities comparable to or greater than the escape velocity. FUSE observations of the OVI doublet provide previously unattainable information regarding outflow velocities and radiative cooling rates in hot gas at T ~ 3 × 105 K. Emission from gas at temperatures of 104 K and ~ 5 × 106 K is now being studied with unprecedented spatial resolution using HST and Chandra, tracing the complex interaction of the still-invisible wind of SN-ejecta with the ambient ISM entrained into these outflows. I discuss the implications of these observations for our understanding of starburst-driven outflows.
INTRODUCTION
STARBURST GALAXIES IN THE LOCAL UNIVERSE
THE EFFICIENCY OF SUPERNOVA HEATING
WARM NEUTRAL GAS IN SUPERWINDS
WARM IONIZED GAS
UV ABSORPTION PROBES OF CORONAL GAS
THERMAL X-RAY EMISSION FROM SUPERWINDS
SUMMARY AND CONCLUSIONS
REFERENCES
Observations of edge-on starburst galaxies show weakly collimated
10 kpc-outflows of gas (Fig. 1),
with outflow velocities of several hundred kilometers per second
(McCarthy, Heckman,
& van Breugel 1987;
Heckman, Armus, &
Miley 1990;
Armus, Heckman, &
Miley 1990).
Tracers of warm ionized gas such as
H
emission show
filaments and arcs of emission extending outward from the nuclear regions
of the host galaxy galaxy, which outline the surfaces
of bipolar outflow cones of opening angle
~ 60deg.
The primary observational probes of these outflows have historically been
been optical emission lines
(Armus et al. 1990),
and X-ray emission
(Dahlem, Weaver, &
Heckman 1998),
although all phases of the ISM have been detected
(Dahlem 1997).
The X-ray emission correlates well spatially
with the H emission (see
Fig. 1),
although in many cases the
X-ray observations trace these outflows out to larger
galactocentric radii (
20 kpc,
Read, Ponman, &
Strickland 1997)
than the H observations.
| 
|
Figure 1. (a) Narrowband
H | |
These outflows result from the energy returned
to the ISM by the recently-formed massive stars in the starburst.
Core collapse supernovae and massive star stellar winds, from
~ 106 O & B stars in galaxies like M82 or NGC 253, return
large amounts of kinetic energy along with metal-enriched ejecta to the
ISM. Radio observations of local starbursts reveal large numbers of
young SNRs within the starburst region
(Kronberg, Biermann,
& Schwab 1981;
Muxlow et al. 1994).
Age estimates for the starburst stellar populations
(~ 10 Myr for M82, 20 - 30 Myr in NGC 253
[Satyapal et al. 1997;
Engelbracht et
al. 1998])
agree well with the dynamical ages of the outflows,
dyn ~ 10
kpc/500 km/s ~ 20 Myr.
The kinetic energy of the individual remnants and wind-blown
bubbles is thermalized via shocks as SNRs overlap
and interact, creating a hot (T ~ 108 K), high pressure
(P/k ~ 107 K cm-3)
bubble of metal-enriched gas in the starburst region
(Chevalier & Clegg
1985).
This "superbubble" expands preferentially along the
path of least resistance (i.e. lowest density),
breaking out of the disk of the galaxy along the minor axis after a
few million years. The hot gas then
expands at higher velocity (v
1000 km/s)
into the low density halo of the galaxy as a superwind,
dragging along clumps and clouds of cool dense entrained ISM
at lower velocity (see
Suchkov et al. 1994).
Many excellent reviews of both observations and theory of starburst-driven superwinds already exist (Heckman, Lehnert, & Armus 1993; Heckman 1998). In this contribution I highlight recent results related to the issue of mass, metal and energy transport by superwinds out of galaxies and into the IGM.
The average starburst galaxy in the local universe (weighted by
FIR luminosity, a convenient estimator of
the star-formation rate (SFR) in all but the most metal-poor galaxies)
has a FIR luminosity of
LFIR ~ L*
(Soifer et al. 1987)
and a SFR of a few
M
yr-1
(Heckman et al. 1993).
In general these galaxies are late type spiral galaxies with rotational
velocities vrot ~ 200 km/s,
local examples being NGC 253, NGC 3628 & NGC 4945.
It is these "typical" starbursts
that show most clearly evidence for starburst-driven outflows
(Armus et al. 1990),
and I shall mainly concentrate on discussing this class of galaxies.
In terms of overall significance, approximately ~ 25%
of all high mass star-formation
in the the local universe is in starburst galaxies
(Heckman 1998),
and it is likely that all starbursts
drive outflows. This is not a rare or exotic phenomenon.
Although outflows from dwarf galaxies have captured much of the theoretical effort on outflows (e.g. Mac Low & Ferrara 1999), under the assumption that it is easier to drive outflows in low mass systems, it is important to realize that this does not automatically mean that larger galaxies do not drive outflows. I will not discuss outflows from local ultraluminous IR galaxies (ULIRGs) which have SFRs up to ~ 30 times greater than the average starburst galaxy, or the starbursting systems seen at high redshift (see Pettini, this volume), except to note that they appear at least as powerful as winds in local starburst galaxies.
The SN rate within a typical local starburst galaxy is ~ 0.05 yr-1 (Mattila & Meikle 2001), which implies a total number of SNe exceeding ~ 2 × 106 over the lifetime of a starburst event. How much of the kinetic energy from all these SNe is available to drive the wind? This is of particular interest with respect to the possible role SN-driven winds play in imparting additional heating to the IGM and ICM (see various discussions in this volume).
It is useful to split the problem into two parts, firstly radiative losses within the starburst region, and secondly radiative losses within the larger-scale superwind. The second of these can be assessed relatively straightforwardly with observations of local superwinds, and will be addressed in Section 6. The radiative losses young SNRs suffer within the starburst region are extremely difficult to determine observationally, as these regions are heavily obscured. Consequently arguments about radiative energy losses are based purely on theory and a wide range of opinions exist. I shall present the situation as I see it, and refer the reader to the contribution by Recchi (this volume) for a different point of view.
A single isolated SNR, evolving in a uniform medium of number density
~ 1 cm-3 will lose a fraction f ~ 90% of
its initial kinetic energy
to radiation over ~ 4 × 105 years
(Thornton et
al. 1998).
Adopting the terminology of
Chevalier & Clegg
(1985),
this corresponds to a thermalization efficiency of
therm =
1 - f ~ 10%. Cooling depends sensitively on the local density as
n2 
dV.
Many authors assume that because bursts of star formation occur
in regions with large amounts of dense gas, virtually all
the energy from SNe is lost due to radiation (e.g.
Steinmetz 1999).
This ignores the multiphase nature of the ISM where the filling
factor of dense gas is low, and that the
phase structure is determined by the local SN rate
(Rosen & Bregman
1995).
In a starburst region such as at the center of M82 or
NGC 253
the SN rate per unit volume is a few
× 10-9 yr-1 pc-3,
about 5 orders of magnitude higher than the SN rate/volume in the disk
of the MW (~ 4 × 10-14 yr-1 pc-3,
Slavin & Cox
[1993]).
The average individual SNR or
wind blown bubble in a starburst
does not exist long enough to radiate away 90%
of its energy before it runs into another remnant or pre-blown
low density cavity. Once in a low density medium radiative losses
cease to be significant (see
Mac Low & McCray
1988).
As a consequence the thermalization
efficiency in starbursts must
be considerably higher than the 10%
value applicable to "normal" star-forming disks.
Numerical simulations investigating thermalization efficiency as a function
of SN rate/volume support this argument (Strickland, in preparation).
Some SNe may occur molecular cores, and suffer significant
radiative losses, but on average SNe in the starburst do not lose
a large fraction of their energy. Thermalization efficiencies
therm
50% are quite possible.
In principle, observationally measuring the temperature of the very hot
tenuous gas in the starburst region (i.e. the thermalized
SN ejecta) can directly provide the thermalization efficiency.
Using the rates of mass and energy input from the Starburst99 models
(Leitherer et
al. 1999),
Tgas = 1.2 × 108
therm
-1 K, where
1 is a measure of mass-loading
(Suchkov et al. 1996).
The faint X-ray emission from this hot gas can, in principle,
be detected in nearby starburst galaxies. Unfortunately starburst
regions are also host to large numbers of
of X-ray binaries, and possible low luminosity AGN,
making this measurement extremely difficult.
The first believable detection of this very hot gas
uses Chandra's high spatial resolution to resolve out the
X-ray binaries.
Griffiths et al. (2000)
claim to detect diffuse emission from a
T ~ 5 × 107 K gas within M82's starburst region,
which if confirmed implies
therm
~ 40 +1%.
Many local starburst galaxies show blue-shifted
Na D absorption line profiles
(Phillips 1993;
Heckman et al. 2000).
The sodium D (NaI
5890, 5896)
lines probe warm neutral gas, at a temperature
of a few thousand degrees.
In Heckman et al's sample of IR luminous starburst galaxies 19 out
of a sample of 33 showed blue shifted absorption, typically extending
out to terminal velocities between
vterm = - 200 to -700 km/s
(Fig. 2)
in galaxies with rotation velocities between 140 and 330 km/s.
Absorption is seen over a wide range in velocity, from the systemic
velocity of the galaxy out to
vterm, suggesting gas with
multiple velocity components in the outflow.
Those galaxies that show blue-shifted absorption tend to be
more face-on than those that do not show absorption.
|
Figure 2. A representative sample of Na D absorption line profiles for four starburst galaxies, adapted from the larger sample shown in Heckman et al. (2000). The dashed vertical lines show the expected centroids of the Na D doublet at the systemic velocity of the galaxy. Horizontal bars represent a blue shift of 500 km/s. Note the strongly blue-shifted Na D absorption in NGC 3526, NGC 1808 and Mrk 273 due to the superwinds in these galaxies. |
This is consistent with a model where the blue-shifted absorption features arise in cool ambient gas entrained into a weakly collimated outflow along the minor axis of the galaxy. The gas initially has low velocity, giving rise to absorption near the systemic velocity of the galaxy, but is accelerated to higher velocity by the ram pressure of the SN-ejecta wind.
This cool gas dominates the total mass in the outflow. For a typical
starburst in this sample (LFIR ~ 2 × 1011
L)
the mass of warm neutral gas is MNaD ~ 5 ×
108Msun. The mass flow rate in
this component significantly exceeds the mass injection rate
due to SNe and stellar winds and is comparable to the gas consumption rate
due to star formation
(
NaD ~ 3 -
10 × SN ~
SF). Although
not highly metal-enriched, this component is significant in terms of total
mass of metals transported out of the disk of the galaxy.
Does this gas escape the galaxy and pollute the IGM? The observed terminal velocities are typically several times the rotational velocity of the galaxy, comparable to or greater than the galactic escape velocity assuming vesc ~ 3 × vrot. It should be stressed that even if v < vesc, this does not automatically imply that the gas is retained by the galaxy. The motion of this cool gas is not simply ballistic, as the clouds are carried along by the wind. The long term fate of this gas is unknown, and depends more on on hydrodynamic forces (wind ram pressure, retardation by halo gas) than the gravitational potential of the galaxy.
A large literature exists using optical emission lines to study warm ionized gas at T ~ 104 K in superwinds (see Heckman et al. 1993 and references therein). Given this, I will only briefly mention some of the important wind diagnostics provided by these studies, before discussing what I believe to be an important observation affecting numerical estimates of mass loss from superwinds.
Balmer lines, primarily H
emission, provide kinematic information along with
morphological information regarding the structure of the outflow.
Spatially resolved kinematic studies
(McKeith et al. 1995;
Shopbell &
Bland-Hawthorn 1998;
Cecil et al. 2001)
provide information on the entrainment and acceleration of cool gas.
The [SII] doublet
( 6717, 6731) can be used as a
density diagnostic. This has been used to derive densities and pressures
in the warm clouds in superwinds
(McCarthy et al. 1987;
Armus et al. 1990).
Line ratios also provide ionization source diagnostics. The gas near the
starburst region is primarily photoionized by the intense UV radiation
from the massive stars, but at larger
distances the H emitting gas
shows line ratios indicative of shock heating
(Martin 1997).
In the future we may hope to apply the detailed
shock diagnostics used in studies of local SNRs to superwinds.
One particularly noteworthy recent development is high resolution studies with HST of the optical emission line filaments in NGC 3079 's superwind (Cecil et al. 2001) and M82's superwind (Shopbell et al, in preparation). At high resolution the filaments and arcs of warm gas break up into very small clumps or clouds. The largest clumps or clouds in NGC 3079 are ~ 30 pc in diameter, although many are unresolved even by HST. Most of the mass in a superwind is in these cool, very compact clouds. Accurately treating the entrainment and acceleration of such clouds by the wind requires that the model resolves the wind/cloud interaction. Klein, McKee, & Colella (1994) argue that this requires at least 100 cells across across a cloud diameter, implying cell sizes of << 1 pc in simulations that must cover 2-or-3-dimensional volumes 10's of kpc on a side (preferably ~ 200 kpc on a side, see Aguirre and Pettini in this volume). No current simulations achieve this level of resolution, and therefore these models may significantly underestimate mass loss in superwinds.
The launch of NASA's Far Ultraviolet Spectral Explorer ( FUSE)
mission in 1999 has finally allowed absorption from the OVI
1032, 1038 doublet to be
detected in local starburst galaxies. These lines probe collisionally
ionized gas at T ~ 3 × 105 K. These observations
provide us with kinematic information on gas ~ 30 times hotter than
probed by optical emission lines.
A theoretical prediction of both analytical and numerical models of superwinds is that hotter gas has higher outflow velocities than the cool gas. The maximum velocities are achieved by the very energetic SN-ejecta, while cooler denser ambient ISM is swept up and accelerated to a terminal velocity dependent on the column density of the clump or cloud (Chevalier & Clegg 1985). If true, hot gas might escape these galaxies even if outflow velocities in the cool phases are well below escape velocity.
If superwinds are to be stopped before reaching the IGM, it
is necessary to radiate away the majority of their energy.
Observations place strong limits on the radiative
power loss in the X-ray band at ~ 1% or so of the SN energy
injection rate 
SN,
while optical emission lines account for a few to ~ 10% of
SN. The only waveband
where appreciable energy could be emerging that has not previously
been explored is the far UV.
Blue-shifted absorption from OVI is seen in a variety of
of starbursts of different mass and star-formation intensity,
from starbursting dwarf galaxies like NGC 4214 (Martin et al, in
preparation) through to typical starbursts like NGC 3310.
Heckman et al. (2001)
present a detailed case study of one the archetypal
starbursting dwarf galaxies, NGC 1705, which shows
a complicated optical morphology suggesting that
hot gas is in the process of "blowing-out" of
a ~ 2 kpc diameter H
bubble. The FUSE observations reveal that the hot gas responsible
for the OVI absorption has a higher outflow velocity than
the warm ionized medium, which in turn has a higher outflow velocity than
warm neutral gas (vOVI = - 77 ± 10 km/s,
vWIM = - 53 ± 10 km/s,
vWNM = - 32 ± 11 km/s).
These observations are inconsistent with the standard superbubble
model
(Castor, Weaver, &
McCray 1975;
Mac Low & McCray
1988),
but agree with
the predictions gas entrainment and acceleration as hot gas flows
out through holes in a fragmented superbubble shell to form
a superwind. FUV radiative losses in NGC 1705 appear minimal,
only ~ 5% of
SN, so superwinds
appear to be inefficient radiators at any wavelength.
The motivation for studying X-ray emission from superwinds has always been the hope that the observed thermal X-ray emission provides a direct probe of the hot, metal-enriched, gas that drives these outflows. The 10-kpc-scale diffuse X-ray emission in superwinds seen by Einstein, ROSAT & ASCA had characteristic temperatures of a few million degrees. This is much cooler than the ~ 108 K expected for raw SN-ejecta within the starburst region, but as the wind expands cooling processes such as adiabatic expansion or mass-loading (Suchkov et al. 1996) might reconcile the observed temperatures with an interpretation of the X-ray emission coming from a volume-filling wind of SN-ejecta. The alternative model is that X-ray emission comes from an interaction between the hot, high velocity wind, and cooler denser ISM along the walls of the outflow or in clouds embedded within the wind (Chevalier & Clegg 1985; Suchkov et al. 1994; Strickland & Stevens 2000). A fraction of the cool dense ISM is shock-or-conductively heated to million degree temperatures.
Distinguishing between these two competing models has awaited high spatial
resolution X-ray imaging. If the X-ray emission comes from a volume-filling
wind, then the X-ray emission should smoothly fill the interior of the
cone or lobes outlined by H
emission. In the wind/ISM interaction
model the X-ray emission should be concentrated in regions with dense
cool gas, and it should appear as filamentary and
limb-brightened as the H
emission. The spatial resolution and
sensitivity of X-ray instruments prior to Chandra was not
high enough to make an exact comparison between X-ray and
H emission
in even the nearest superwinds, although a general correlation between the
two has long been noted
(Watson et al. 1984;
McCarthy et al. 1987).
Chandra's
1" spatial resolution corresponds to ~ 12
pc at the distance of nearby superwinds - the same physical scales as the
H emitting clouds. We now
have unambiguous
evidence (Fig. 3) for a 1-to-1 relationship between
the spatial distribution of the soft thermal X-ray emission and the
H emission in the inner kpc
of several superwinds
(Strickland et
al. 2000;
2001 in preparation). Over the
larger 10 kpc scales of the winds the X-ray emission still appears
filamentary and arc-like, and is associated with nearby filaments or
arcs of H
emission. Low-volume filling factor
gas dominates the X-ray emission from superwinds, and the
the gas that actually drives the outflow remains invisible.
|
Figure 3. Soft X-ray and
H |
Typical starburst galaxies (LFIR ~
L*) show high velocity (200-700 km/s)
multi-phase outflows.
The observed mass flow rates in the wind are comparable to the
gas consumption rate due to star formation
(wind
*), and
are dominated by relatively cool ambient
gas (T ~ few thousand K) that has been swept-up and accelerated
by the ram pressure of the hotter wind of SN-ejecta.
Outflow velocities are typically comparable or greater than estimates
of the galactic escape velocity, but caution should be
exercised in making claims about mass loss rates.
The gas motions are not ballistic, making it impossible to
give quantitative observational mass loss rates. Observations
of gas at much larger galactocentric radii
(100 kpc) are
needed to directly observe mass loss. Nevertheless, the observed
mass flow rates are considerable, and does seem likely that
some significant fraction of even the coolest phases may well escape
even moderately massive starburst galaxies.
Speaking as a practitioner of hydrodynamical simulations of superwinds, I am
not convinced that we know mass-loss rates theoretically. Existing
models have yet to be meaningfully tested
against observations. The lack of sub-parsec
numerical resolution in current simulations prejudices
the ability of these models to treat mass transport in winds.
FUSE observations of OVI absorption provide vital information of the kinematics and radiative losses of coronal gas at T ~ 3 × 105 K. In the dwarf starburst NGC 1705 the FUSE observations support the theoretical prediction of superwind models that the hotter phases in superwinds have higher outflow velocities than the cooler phases. This suggests that the even hotter material holding the metals is more likely to escape than the warm neutral and ionized gas. Radiative energy losses within the wind appear minimal compared to the energy injection rate from SNe, even within the FUV and X-ray wave-bands.
With the sub-arcsecond spatial resolution provided by Chandra it is now clear that the soft thermal X-ray emission seen in superwinds is due to some form of interaction between the (still invisible) high velocity hot wind and cooler denser ambient gas swept-up or overrun by the flow. This is unfortunate in the sense that X-ray observations do not provide a direct probe of the energetic metal-enriched gas driving these winds. Nevertheless the data from Chandra allow us to obtain more accurate estimates of the physical properties of the X-ray emitting gas than ever before, and provide deeper insight into the conditions within these winds.
Starburst-driven winds are difficult objects to study, due to the range of different gas phases involved and the faintness of the emission. Nevertheless, very significant progress is being made, most notably due to the new observational capabilities provided by FUSE & Chandra. The wealth of multi-wavelength data will place extremely strong constraints upon numerical models of superwinds. This is an exciting time to study superwinds, and there is no prospect of an end to new discoveries about these fascinating and important objects.
Acknowledgments. It is a pleasure to thank Tim Heckman for countless enlightening discussions over the years, Crystal Martin and Gerhardt Meurer for providing a variety of spectra and images, and Francesca Matteucci for organizing a stimulating workshop. DKS gratefully acknowledges the support from Chandra Postdoctoral Fellowship Award Number PF0-10012, issued by the Chandra X-ray Observatory Center, operated by the SAO on behalf of NASA.