3.1. X-Ray Observations of Superwinds
The overall X-ray properties of starburst galaxies have been recently reviewed by Petre (1993), so we will focus on the interplay between the X-ray data and the superwind phenomenon. X-ray observations of superwinds are crucial to our understanding of the phenomenon, because (potentially at least) they offer the most direct probe of the hot, tenuous superwind fluid itself (rather than simply probing the interaction of the wind with ambient gas, as is most likely the case for the optical line emission).
The characteristic temperature corresponding to the complete thermalization
of the kinetic energy from an ensemble of supernovae and winds
from massive stars is about kT = 10 keV (e.g., 1051
ergs per
10 M
)
- see Section 2 above. Thus we expect hard
X-ray emission from a galaxy driving
a superwind to come from the hot and tenuous supernova/wind-heated gas
inside the starburst (i.e., inside the sonic radius of the wind). Since the
corresponding terminal velocity of the outflowing superwind fluid is several
thousand km s-1, hard X-rays will also arise from internal
shocks in the superwind (cf.
Tomisaka & Ikeuchi 1988;
Balsara, Suchkov, &
Heckman 1993).
In the idealized expanding bubble discussed in
Section 2 above, these
X-rays will be coming from zone 3.
Soft X-rays will be produced by dense ambient material that is shock-heated and/or evaporated by the superwind (cf. Watson, Stanger, & Griffiths 1984; White & Long 1991). In the idealized 'expanding bubble' model discussed in Section 2, this would correspond to the thin shell of shocked ambient gas at the interface with the undisturbed ambient gas (zone 4). In a more realistic situation, the soft X-rays will arise as the wind collides with, shock-heats, and evaporates density inhomogeneities ('clouds') in the galaxy halo. H I observations of starbursts show that such cool, dense material indeed exists in the environs of starbursts (see Section 3.4 below). Note that the typical outflow velocity of dense, wind-accelerated ambient material is observed in the optical to be several hundred km s-1 (see Section 3.2 below), corresponding to post-shock temperatures of kT ~ 200 eV.
This theoretical/phenomenological picture is in reasonable agreement with the limited amount of X-ray data presently available for starburst galaxies. The X-ray luminosity of starburst galaxies correlates rather well with both the IR luminosity (e.g., Griffiths & Padovani 1990; David, Jones, & Forman 1992; Green, Anderson, & Ward 1992) and the Lyman continuum luminosity (Ward 1988) of the starburst. Moreover, the observed average ratio of X-ray and bolometric starburst luminosities agrees roughly with the predictions of simple theoretical models for starburst-driven superwinds (HAM - and see the discussion in Section 2 above). X-ray spectra of starbursts imply characteristic temperatures of kT = 6 to 9 keV for the global X-ray emission (e.g., Kim, Fabbiano, & Trinchieri 1992; Ohashi et al. 1990; Petre 1993), again in agreement with simple models (see Section 2 above).
Spatially-resolved X-ray images of well-studied starburst galaxies whose large-scale stellar disks are viewed nearly edge-on (M 82, NGC 253, NGC 2146, and NGC 3628) show striking X-ray 'halos' or 'plumes' that can extend out to a radius of 10 kpc or more along the galaxy minor axis. These have been interpreted as direct evidence for galactic superwinds (e.g., Watson, Stanger, & Griffiths 1984; Fabbiano 1988; Fabbiano, Heckman, & Keel 1990; Heckman & Fabbiano 1993; Armus, Heckman, & Weaver 1993). The transgalactic-scale X-ray nebulae associated with such 'ultraluminous' IR galaxies as Arp 220, NGC 3690, and Mrk 266 are likely to be more energetic versions of this phenomenon (Eales & Arnaud 1988; Armus, Heckman, & Weaver 1993).
In general, the estimated thermal energy content and mass of the X-ray gas are consistent with the time-integrated kinetic energy and mass output of the starburst (cf. HAM and references therein). However, these estimates are based on several simplifying assumptions. The most important of these are that all the observed X-rays arise from hot gas at a uniform - and high - temperature equal to that deduced on the basis of the integrated X-ray spectrum, and that this gas has a volume filling factor of unity. As we now discuss, the actual situation is likely to be considerably more complex.
The X-ray emission from the halo of M 82 is significantly softer than
that from the central starburst
(Petre 1993),
as is also the case in NGC 3628
(Fabbiano, Heckman, & Keel
1990;
Heckman & Fabbiano 1993)
and NGC 3690
(Armus, Heckman, & Weaver
1993).
Within the context of the
superwind model, this could imply that the wind has broken free of the
galaxy and has suffered severe adiabatic cooling on its way out (cf. CC
and Equation 6 above). However, this is unlikely to be the case, since the
X-ray nebulae are far too bright to simply be an adiabatically-cooled
free-wind (e.g.,
Mathews & Doane 1993).
Our new ROSAT images of several
starbursts with superwinds also show a considerable amount of fine-scale
structure: X-ray-bright 'lumps' and 'filaments' with sizes of a few hundred
pc to a few kpc
(Heckman & Fabbiano 1993;
Armus, Heckman, & Weaver
1993).
We believe it is most likely that the soft (kT
One of the most interesting X-ray observations of a likely superwind is
the Ginga data on the center of our own Milky Way
(Yamauchi et al. 1990).
Emission-line images using the 6.7 keV Fe K line imply a region of gas
with a
temperature of about 108 K, dimensions of about 270 ×
150 pc, gas pressures
of P/k = 8 × 106 K cm-3, and total
thermal energy content of 6times1053
ergs. Since the sound speed in this gas (~ 1500 km s-1) greatly
exceeds the
escape velocity from the Galactic Center, the gas should flow out as a high
speed wind (unless it is confined by some other mechanism). If this outflow
is a steady-state phenomenon, collisional heating with a supernova rate of
about 0.6 per century within this volume is required to balance adiabatic
cooling.
To summarize, while the current situation regarding X-ray emission from
superwinds is unclear, we can expect dramatic progress over the next few
years as the avalanche of new ROSAT, BBXRT, and ASTRO-D data on
starbursts is analyzed and digested. Certainly the data already in hand are
very tantalizing.
3.2. Optical Line Emission from Superwinds
Most of the existing data on superwinds concern the optical line emission
associated with such outflows. Such data provide a whole array of
diagnostics of the dynamical and physical state of the outflow.
In the simple schematic pictures described in
Section 2 above, the optical line
emission would not arise from the very hot and tenuous superwind fluid
itself (the cooling times are excessively long, and most of the relatively
feeble
radiation that is produced there is in the form of hard X-rays). Instead, we
would expect the optical line emission to come from relatively dense ambient
material into which the wind's ram pressure drives slow (< few hundred km
s-1) radiative shocks (e.g.,
McKee & Hollenbach 1980
and see Section 2.2 above).
In the wind-blown bubble 'onion-skin' model, this material would be the
thin, dense shell of compressed and shock-heated ambient gas at the leading
edge of the bubble. In a more realistic situation in which blow-out has
occurred, and/or in which the wind is propagating through a inhomogeneous,
multi-phase medium, the optical line emission will come from clouds (e.g.,
fragments of the ruptured shell, pre-existing density inhomogeneities in the
halo, or material carried out of the galactic disk by the superwind).
We therefore expect that the optical data will provide a detailed - but
indirect - view of the superwind phenomenon. The largest compilations of
such data are found in the recent spectroscopic and narrow-band imaging
surveys by
HAM, Armus, Heckman, & Miley
(1989,
1990
- hereafter AHM89, AHM9O),
Lehnert (1992
- hereafter L92), and Lehnert & Heckman
(1993
- hereafter LH).
3.2.1. Structure and Luminosity
Galaxies selected to be both luminous and warm in the far-IR have optical
emission-line nebulae whose size and luminosity correlate reasonably well
with the far-IR luminosity, and hence with the estimated formation rate
of massive stars (AHM9O; L92; LH). The
H
We (L92, LH) have recently completed an
H
Figure 2.
H
Figure 3.
H
HAM, L92, and LH have measured radial electron density profiles in the
emission-line nebulae associated with about 20 IR-selected starburst
galaxies. Densities range from 103 to < 102
cm-3. The
observed H
In the few starburst galaxies for which both
H
Figure 4. Overlaid
H
3.2.2. Kinematics and Dynamics
The kinematics of the optical emission-line gas associated with IR-luminous
galaxies also suggest that outflows are common. Optical spectra of the
nuclei of IR-luminous galaxies show that they often have blue-asymmetric
emission-line profiles (e.g., AHM89; LH;
Phillips 1992).
Mirabel & Sanders (1988)
find that the optical emission-line velocities in the nuclei of
high-luminosity IR galaxies are blueshifted with respect to the galaxy
systemic
velocity by an average of about 100 km s-1. Both the
blue-asymmetric
profiles and blueshifts suggest an outflow of ionized gas whose redshifted
back side is obscured by dust.
For the edge-on starburst galaxies investigated by L92 and LH, in many
cases the line widths actually increase with increasing radius along the
minor
axis. The emission-lines are also systematically broader along the minor
than
along the major axis (with typical FWHM's of 150 to 350 km
s-1 vs. × 100
to 200 km s-1 respectively). The line widths along the minor
axis correlate
strongly with the starburst IR luminosity, but not with the galaxy rotation
speed, thus favoring a starburst-driven outflow over simple orbital motions
in the galaxy gravitational potential (see
Figure 5). This interpretation is
further supported by the strong trend found by L92 and LH for the largest
measured velocity shears along the minor axis to occur in the galaxies
viewed
more nearly face-on (as expected for a starburst-driven radial outflow along
the minor axis) and in the galaxies with the highest IR luminosities (see
Figure 6).
Figure 5. Summary of the kinematics of the
emission-line gas located outside the
central starburst and along the optical minor axis for the IR-selected
edge-on disk
galaxies investigated by L92 and LH. The x-axis in both plots is the
FWHM of the [N II]6584 line. This line width correlates strongly with
the IR luminosity
of the galaxy (left plot), but not with the rotation speed of the
galaxy (right plot), implying that the gas dynamics is influenced much
more strongly by the starburst-driven wind than by gravity.
Figure 6. Summary of the kinematics of the
emission-line gas located along the optical
minor axis for the IR-selected edge-on disk galaxies investigated by
L92 and LH.
These two plots show that the velocity shear along the minor axis is
largest in the
galaxies that are viewed more nearly face-on (left plot) and in the
galaxies with the largest IR luminosities (right plot).
The detailed kinematic properties of the gas located along the minor axes
of such nearby and well-studied edge-on starbursts as M 82 (e.g.,
Bland & Tully 1988;
HAM), NGC 253
(Ulrich 1978;
HAM), NGC 3079 (HAM;
Filippenko & Sargent
1992),
and NGC 4945 (HAM) provide more direct evidence
for outflows. All four galaxies have bubble-like or filamentary
emission-line
nebulae that protrude out about a kpc along the optical minor axis. The
kinematics of these structures (broad, double-peaked emission-line profiles
in the center of the bubble, narrowing to single-peaked profiles along the
bubble periphery) imply that they are either expanding bubbles or the walls
of cone-like or cylindrical outflows, with inferred outflow/expansion speeds
ranging from about 200 km s-1 to nearly 1000 km
s-1 (see Figure 7). The
sizes and expansion speeds of these structures are in good agreement
with the
quantitative predictions of the simple wind-blown-bubble model discussed in
Section 2 (see equations 9 and 10 above, and
HAM). Additional examples of similar
outflow structures are found in the composite Seyfert/starburst galaxies
NGC 1365
(Phillips et al. 1983a),
NGC 5506
(Wilson, Baldwin, &
Ulvestad 1985),
NGC 7582
(Morris et al. 1985),
and Mrk 509
(Phillips et al. 1983b).
Figure 7. Long-slit spectrum of (from left
to right) the
[N II]
Qualitatively similar, but even larger and more energetic examples of
such expanding structures, are found in several 'ultraluminous' IR galaxies
(HAM). These are the 'double-bubble' emission-line nebula in Arp 220, the
central 'hour-glass' structure in NGC 6240, and the extraordinary nebula
associated with IRAS 00182-7112 in which line-widths are nearly 1000 km
s-1
over a region 30 kpc in extent.
Colina, Lipari, &
Macchetto (1991)
have also published kinematic evidence for a superwind in the ultraluminous
galaxy IRAS 19254-7245.
On the 'micro-starburst' level,
Meurer et al. (1992)
have found a
kpc-scale bubble of ionized gas expanding at about 50 km s-1
in the core of the
post-starburst dwarf galaxy NGC 1705. They show that this expansion was
probably powered by the kinetic energy supplied by a burst of star-formation
that occurred some 10 Myr ago.
Roy et al (1991)
have found a kinematically
similar 200 pc-scale expanding bubble of ionized gas in the starbursting
dwarf irregular galaxy NGC 2366, while
de Vaucouleurs, de
Vaucouleurs, & Pence (1974)
found evidence for large-scale outflow in the similar system
NGC 1569. JHU graduate student A. Marlowe with
T. Heckman, R. Wyse,
and R. Schommer have now found an additional half-dozen similar examples
in their echelle and Fabry-Perot
H
As a closing comment on the kinematics of the optical emission-line
gas, we note that the outflow velocities as deduced from all the evidence
summarized above are quite modest (typically < several hundred km
s-1).
These velocities are much lower than the predicted outflow speed for the
hot, tenuous superwind fluid (a few thousand km s-1 - see
Section 2), but are
quite consistent with our expectations for ambient gas that is being shocked
and accelerated outward by the wind (see the discussion in
Sections 2.2 and
2.4
above). This may be seen by considering Equations 10, 16, and 18 above,
or by noting that strong optical line emission will only be produced by
material into which radiative shocks are being driven (e.g., shocks in which
the radiative cooling time behind the shock is short compared to the shock
dynamical timescale). As reviewed by
McKee & Hollenbach
(1980),
radiative shocks in the ISM generally have shock speeds less than a few
hundred km
s-1. Since we expect the velocity to which a shocked cloud is
accelerated
by the superwind to be similar to the speed of the shock driven into it, the
observed outflow velocities therefore represent additional indirect evidence
for the schematic picture we have developed above.
3.2.3. Emission-Line Ratios and the Ionization
Source
The relative emission-line intensities in the nebulae associated with
superwinds
provide indirect evidence for a high-speed outflow. Key emission-line
ratios like
[S II]
The simplest picture is that the heating and ionization of the gas in
the central starburst and in the large-scale disk of the galaxy is dominated
by photoionization by hot stars. In contrast, shocking of ambient gas by
the superwind probably makes a significant contribution to heating the gas
located out in the galaxy halo and seen most clearly along the minor axis.
The transfer of energy from the hot superwind gas to imbedded cool, dense
material via the 'mixing layer' phenomenon
(Begelman & Fabian 1990;
Slavin, Shull, & Begelman
1993)
may also contribute to the heating and
ionization of the optical emission-line gas. It is also possible that
the halo
gas is photoionized by an extremely dilute radiation field 'leaking' out
of the starburst (e.g.,
Sokolowski &
Bland-Hawthorn 1993;
see also
Dettmar 1993
and references therein).
Perhaps the strongest indirect evidence for mechanical heating of the
gas by the superwind is the strong correlation found by HAM, L92, and
LH between the dynamical state of the gas (as measured by the emission-line
widths) and the thermal/ionization state of the gas (as probed by
the strengths of the [O I], [S II], and [N II] lines relative to
H
3.2.4. Densities and Pressures as
Diagnostics
One of the most useful diagnostics of superwinds is to use the ratio of the
[S II]
The superwind model also predicts that once the wind 'blows out' of the
galactic disk and is in free expansion, the pressure within the wind is
dominated by ram pressure. The ram pressure will drop like the inverse
square
of the radius and will be proportional to the rate at which the starburst
injects momentum into the wind (see Equation 5 above). HAM have used
the measured pressure profiles in 12 IR-selected galaxies to estimate the
momentum flux of the wind, and find excellent agreement between these
estimates and the predicted rate of momentum input from the starburst.
This analysis has been extended to an additional set of galaxies by L92 and
LH, and the good agreement between theory and observation verified over
a range of nearly three orders-of-magnitude in starburst luminosity.
It is worth emphasizing that an ultraluminous starburst/superwind can
produce pressures over galactic-scale volumes that are three to four
orders-of-magnitude higher than the general pressure in the ISM in the
Milky Way:
e.g., P/k > few × 106 K cm-3 (see HAM,
L92, and LH for details). The radius
of the over-pressured region (as measured out to some fixed isobar) should
scale as the square-root of the star-formation rate, and this dependence
has been confirmed by HAM, L92, and LH. HAM and LRD have further
shown that the P
The high measured pressures and the systematic dependence of the
radial pressure profile upon the starburst luminosity clearly establish that
starbursts are able to regulate the gas pressure over very large volumes (at
least one hundred times the volume of the starburst itself). But must this
regulation be due to a wind? Might it not be possible for the measured
pressure profiles to reflect the static thermal pressure in a hot
volume-filling
phase that is gravitationally bound to the starburst galaxy?
In fact, it is rather easy to rule out the above idea for those
starbursts in
HAM that have X-ray data available. Since the X-ray luminosity depends
on density squared, and since the density (at fixed pressure) is inversely
proportional to the temperature, it follows that the hotter the
volume-filling
phase (at fixed pressure), the lower its X-ray luminosity. Specifically,
we calculate that a volume-filling hot phase in pressure-balance with the
optical emission-line clouds would have an X-ray luminosity that exceeds
the measured X-ray luminosity unless it has a temperature in excess of 4
× 107 K. The sound-speed in gas at T > 4 ×
107 K is cS > 1000 km s-1,
which is in turn much greater than the escape velocity from the starburst
potential well. Thus, any such gas could not be bound to the starburst and
will flow out as a wind (QED!).
The shapes of the measured pressure profiles can also be used to argue
against an AGN as the source of the wind. If, as would be the case with an
AGN, the region of mass/energy injection is small compared to the
seeing-disk
(which has a typical radius of 100 pc for the starbursts in HAM and
LH), then the radial pressure profile should follow an inverse square law
all the way in to the center. Such a r-2 radial
density/pressure profile is
directly measured in the classical AGN (type 1 Seyfert nucleus) NGC 4151
by Mediavilla et
al. (1992)
over a range in radii between about 70 and 400 pc
(e.g., into the approximate radius of the seeing disk).
Pronik (1989)
has used
the observed correlations between critical density and line width for many
different forbidden lines in the nucleus to conclude that the density in
NGC 4151 continues to follow a r-2 law all the way in to
sub-pc scales. In contrast,
the observed radial pressure profiles in most IR-bright galaxies show flat
central zones (within the starburst, where mass and energy are injected) and
then steepen to approximately r-2 at large radii (in
reasonable agreement
with theoretical models - cf. CC; HAM; LH). This behavior can be clearly
seen in Figure 8 for the two galaxies in HAM for
which the measured radial
pressure profiles are of the highest quality (M 82 and NGC 3256). The shapes
of the radial pressure profiles are remarkably similar in the two
galaxies, and
the larger radial scale associated with the 8-times-more-luminous starburst
in NGC 3256 is exactly as predicted by the superwind
model. Note that
both profiles are relatively flat in the central region (within the
starburst
radius denoted r* in the plot), and then steepen to an
r-2 dependence at large radii.
Figure 8. A log-log plot of the electron
density (left axis) and gas pressure (right axis)
vs. radius for two well-studied starburst galaxies (M 82 - solid dots;
NGC 3256 - hollow
dots). Note that the radial scale for NGC 3256 (upper axis) has been
shifted
by 0.45 dex (factor 2.8) with respect to that for M 82 (lower
axis). The radial
pressure profile predicted for both galaxies by the simple
spherically-symmetric
wind model of CC is indicated by the solid curve. We have also
indicated estimates
of the pressures in the X-ray and radio plasma in M 82. See HAM for
further details.
Finally, in closing our discussion of optical line emission, it must be
emphasized that imaging polarimetry of the famous M 82 nebula shows that
H
3.3. The Relativistic Plasma Associated with
Superwinds
Several starburst galaxies have been discovered to have large-scale
nonthermal
radio halos. These presumably arise as relativistic magnetized plasma
is convected by the superwind out of the central starburst, where it was
created by supernovae (e.g.,
Lerche & Schlickeiser
1980).
The galaxy M 82, studied most recently by
Seaquist & Odegard
(1991),
has a synchrotron-emitting halo extending out to a radius of at least 8
kpc. The spectrum of the radio emission steepens markedly with increasing
radius, which Seaquist & Odegard attribute to the more rapid cooling of
the most energetic electrons by inverse Compton scattering of the starburst
IR photons. They then derive an outflow speed for the relativistic plasma
of about 2000 km s-1 (consistent with the expected outflow
speed of the tenuous superwind fluid - cf. CC and the discussion in
Section 2 above). They also
use the radio data to argue that inverse Compton radiation contributes
a negligible fraction of the X-ray emission from M 82 (see
Section 3.1 above),
and they estimate that relativistic particles comprise only a few percent
of the energy content of the wind. They note that this ratio is similar to
that observed in supernova remnants, as expected in the superwind model.
Reuter et al. (1992)
have obtained higher resolution maps of the M 82 halo,
and find a considerable amount of structure (radially-oriented filaments and
gaps). They interpret this as clear evidence that magnetized plasma is being
convected out of the M 82 starburst, creating a bipolar (poloidal) magnetic
field geometry.
Hummel & Dettmar (1990)
have studied the nonthermal radio halo of
the edge-on actively-star-forming galaxy NGC 4631, and have argued that
the radial spectral index gradient they measure for the radio emission is
consistent with models in which relativistic plasma is convected out of the
galaxy by a wind, and is then cooled by synchrotron, inverse Compton,
and/or adiabatic-expansion losses. As in M 82, the magnetic field geometry
in the halo of NGC 4631 is inferred to be poloidal (field lines
directed radially
outward). Hummel & Dettmar suggest this is due to the operation of a strong
galactic wind in this galaxy (though a dynamo model for the field can not
be ruled out).
Most recently,
Carilli et al. (1992)
have discovered a radio halo around
the edge-on starburst galaxy NGC 253 which extends out to at least 9 kpc.
As in the case of M 82 and NGC 4631, the radio spectrum steepens with
increasing distance from the disk/starburst, and Carilli et
al. interpret this
as evidence for an outflow from the galactic disk into the halo. The
magnetic
field geometry in this case appears more nearly toroidal (field lines
parallel to
the galactic disk), except along a bright radio spur that attaches to
the inner disk of the galaxy.
Dettmar (1993)
has also discussed the radio halo of the
IR-bright edge-on disk galaxy NGC 5775
(Hummel 1991),
and emphasized the close connection between the radio halo and the
system of faint H
There are also less complete data indicating the presence of thick
radio disks and/or radio halos in the edge-on starburst systems NGC 1808
(Dahlem et al. 1990),
NGC 3628
(Reuter et al. 1991),
and NGC 4945
(Harnett et al. 1989).
The smaller (few-kpc-scale) radio sources extending
out along the minor axes of some edge-on radio-bright disk galaxies (e.g.,
Duric et al. 1983;
Hummel, van Gorkom, &
Kotanyl 1983)
may also be related phenomena.
3.4. The Cool Gas Associated with Superwinds
Studies of the interaction of superwinds with cool atomic and molecular
gas are potentially very important, since such gas is expected to comprise
the major fraction of the mass of the ISM inside starburst galaxies or in
their surrounding halos. One of the most important insights provided by
such studies is that starburst galaxies are often immersed in large
(transgalactic-scale) HI envelopes or halos, presumably of tidal origin
(e.g.,
Yun & Ho 1993;
Hibbard et al. 1993;
Fisher & Tully 1977;
Haynes et al. 1979;
Stanford 1990;
Stanford & Wood 1989;
Gottesman & Mahon 1990).
This is
of crucial significance, because the interaction between the wind and this
comparatively dense halo gas probably accounts for the relative brightness
of superwinds in the optical and X-ray domain. Were such gas not present
- i.e., if superwinds instead flowed into a very low density
intergalactic medium
- they would quickly adiabatically cool, their density would rapidly drop,
and they would then become undetectably faint sources of either X-rays or
optical line emission (cf.
Mathews & Doane 1993).
Thus, it is likely that the
interaction of the superwind with the H I envelope/halo is fundamental in
determining what we observe at optical and X-ray wavelengths.
Studies of the cool gas in the environs of starbursts have also provided
direct evidence for outflows. Indeed, possibly the single most convincing
piece of evidence to date for an outflow from a starburst galaxy is the
recent
analysis of the kinematics of the Na I D doublet seen in the well-known
starburst galaxy NGC 1808
(Phillips 1992).
He has studied the kinematics
of the system of radial dust-filaments that extend several kpc out along the
minor axis of this galaxy. The Na D doublet is seen in absorption on the
near
side of the outflow, and has a strongly blueshifted component (by up to 400
km s-1 with respect to the systemic velocity). Since the
kinematics are being
probed with absorption-lines in this case, there is no possible ambiguity in
the sign of the observed (outward) radial motions. Assuming a standard
gas-to-dust ratio in the outbound dust filaments, the implied mass of
the ejected material is about 6 × 107
M
There are also several pieces of evidence for outflowing molecular gas in
starbursts. High-resolution 115GHz CO observations of M 82 by
Nakai et al. (1987)
suggest that molecular gas is flowing out of the central starburst,
and Irwin & Sofue (1992)
have reached similar conclusions for the CO in the center of NGC 3079.
Turner (1985)
has found a molecular outflow traced
by the OH emission lines extending out from the nucleus to about 1.5 kpc
along the minor axis of NGC 253, and
van der Werf et
al. (1994)
have found plumes of near-IR-emitting molecular hydrogen at the base of this
apparent outflow. In the most detailed near-IR study to-date of the famous
IR-luminous galaxy NGC 6240,
van der Werf et
al. (1993)
have uncovered
convincing kinematic and morphological evidence that the relatively warm
molecular gas (as probed with the near-IR vibration-rotation lines of
H2)
in this system is being ejected from the starburst at velocities of several
hundred to 1000 km s-1 (in agreement with the kinematics of
the ionized gas discussed by HAM).
Radio observations of H I in several starburst galaxies also reveal the
impact a superwind can have on the cool ISM.
Yun, Ho, & Lo (1992)
have suggested that the wind from the starburst in M 82 has carved a channel
through the inner part of that galaxy's H I envelope.
Koribalski et al. (1992)
have found clear kinematic evidence for H I flowing out along the minor axis
of the starburst galaxy NGC 1808 at a velocity of at least 150 km
s-1. This
is the same outflow studied optically at much higher spatial resolution via
the Na I D line by Phillips
(1992 - see above).
Recent VLA H I maps of several star-forming dwarf galaxies have revealed
the presence of kpc-scale H I shells expanding at about 10 km
s-1 (e.g.,
Puche et al. 1992;
Westpfahl & Puche 1992).
These authors argue that
such expanding, hollow structures are the result of the collective effect
of multiple stellar winds and supernovae. Moreover, since the sizes and
expansion speeds of the bubbles are comparable to the total sizes and
internal velocity dispersions in the smallest dwarf galaxies, these
structures
probably play a major role in the evolution of the ISM and of star-formation
in such galaxies (e.g.,
Westpfahl & Puche 1992).
Perhaps the most spectacular phenomenon ascribed to a superwind is the
ram pressure stripping of the H I out of the outer parts of the galaxy
NGC 3073 by its starbursting, superwind-driving
neighbor NGC 3079
(Irwin et al. 1987).
The ram pressure that Irwin et al. estimate as being required to strip
the H I from NGC 3073 implies a superwind momentum flux that agrees
well with that estimated for NGC 3079 by HAM via three independent
techniques (from the kinematics of the expanding bubble, from the radial
pressure profile, and from the starburst luminosity using Equation 3 above).
luminosities of the
extra-nuclear portion of the nebulae (e.g., well outside the starburst
region) are
just consistent with the predictions of the superwind model (see Equation
11 and the associated discussion in
Section 2 and in HAM).
imaging survey of a sample
of edge-on disk galaxies selected on the basis of far-IR flux and color, and
find many examples which show emission-line filaments, loops, or bubbles
extending out one to ten kpc along the optical minor axis of the galaxy
(four examples are shown in Figure 2). In
general, we find a pronounced
excess of optical line emission along the optical minor axis compared to
what
would be expected for ordinary emission-line gas confined to the galactic
disk. At the highest levels of IR luminosity (L > few ×
1011
L), the
H
nebulae approach transgalactic dimensions (30-100 kpc), with the
large-scale morphology often dominated by filaments, loops, or bubbles
(e.g.,
Heckman, Armus, & Miley
1987;
AHM90).
Some examples of such nebulae are shown in Figure 3.
+[NII] images of IR-bright,
edge-on disk galaxies exhibiting large-scale
optical emission-line filaments extending several kpc out of the galaxy
disk and into
the halo. (a) NGC 660; (b) NGC 3079; (c) NGC 3628; and (d) NGC 4666 - see
AHM90;
Fabbiano, Heckman, & Keel
1990;
L92; LH).
+[NII] images of two galaxies
with IR luminosities greater than 1045 erg
s-1: (a) Arp 220 (from
Heckman, Armus, & Miley
1987)
and (b) Mrk 266 (from
AHM90). These represent the superwind phenomenon at the high-luminosity
end, where the emission-line nebulae are tens of kpc in size
luminosities
and the measured densities allow us to estimate the mass and the
volume-filling
factor of optical emission-line gas. The derived masses range from
105
M to
107
M, and the
volume-filling factors are typically 10-3 to
10-4.
Thus, the emission-line gas represents a modest amount of relatively dense
material distributed in the form of clumps, sheets, or filaments that occupy
only a very small fraction of the volume of the halo of the starburst
galaxy.
and X-ray images are
available, there is a clear correspondence between the two images (e.g.,
Watson, Stanger, &
Griffiths 1984;
Armus, Heckman, & Weaver
1993).
In some cases, the optical line emission is preferentially located along
the outer boundary of the X-ray nebula
(Fabbiano, Heckman, & Keel
1990;
Heckman & Fabbiano 1993;
McCarthy, Heckman, & van
Breugel 1987
- and see
Figure 4). This suggests that the optical line
emission arises at the interface
between the hot superwind and the ambient interstellar gas in the halo of
the starburst galaxy (in good agreement with both the theoretical picture
sketched above and the kinematics of the optical emission-line gas in
several
well-studied cases, as summarized below). New ROSAT images should allow
us to conduct many more such comparisons, in order to better test this
idea.
+[N II] image (greyscale) and
Einstein HRI X-ray image
(contour plot) of the central few kpc of the prototypical IR starburst
galaxy NGC 253. The starburst nucleus is located in the upper right
corner. Note how the
filamentary optical line emission to the SE of the nucleus seems to
enclose the X-ray plume (cf.
McCarthy, Heckman, & van
Breugel 1987;
Fabbiano & Trinchieri
1984;
LH; L92).
6548,
H,
[N II]6584
emission-lines in the central few kpc of NGC 253. The slit was oriented
in a position
angle of 135 degrees, along the minor axis of this edge-on galaxy. Note
that each
of the three lines has a double-peaked profile shape throughout a
region about 600
pc in extent to the SE of the nucleus. This line-splitting has a
magnitude of about
300 to 400 km s-1, and suggests either an expanding bubble
or outflow along the
surface of a hollow cone-like structure (cf. HAM). This region of
expanding gas also
corresponds to the bright X-ray plume seen in Fig. 4 (cf.
Fabbiano & Trinchieri
1984;
McCarthy, Heckman, & van
Breugel 1987).
survey of UV-bright dwarf
galaxies
(Marlowe et al. 1993).
Because the potential wells of dwarfs are so shallow,
there is a broad theoretical consensus that starburst-driven mass loss plays
a crucial role in the evolution of such systems (e.g.,
Silk, Wyse, & Shields
1987).
We will discuss related H I observations of several dwarf galaxies in
Section 3.4 below.
6716,6731 /
H,
[N II]6584 /
H, and
[O I]6300 /
H have
values that are more similar to those seen in gas that has been
shock-heated (e.g., old supernova remnants and Herbig-Haro objects - cf.
Hartigan, Raymond, &
Hartmann 1987;
Fesen, Blair, & Kirshner
1985)
than to what is observed in H II regions in spiral and irregular
galaxies (e.g.,
McCall, Rybski, & Shields
1985).
We (L92; LH) find that all these ratios tend to increase
with increasing radius along the minor axis in edge-on starburst galaxies
(see also AHM89 and HAM) and are also systematically larger along the
minor axis than along the major axis.
). This
strongly suggests that the same mechanism is responsible for both heating
and accelerating the gas. The most decisive way of testing whether the gas
is collisionally ionized or photoionized would be to determine whether the
electron temperature in the O III zone is higher than can be explained by
photoionization (e.g., T > 20,000 K - cf.
Osterbrock 1989).
HAM attempted such measurements using the ratio of the
[O III]4363 and
5007 lines, but
the results were inconclusive (in some cases shock-heating was favored and
in other cases photoionization).
6716 and
6731 lines to determine the
radial run of density within the
superwind nebulae. In many cases, the average temperature of the nebula
can also be deduced from the ratio of the
[N II]5755 and
6584 lines, so
that the radial run of pressure can be determined. According to superwind
models, the central pressure in the starburst is set by the starburst radius
and the rate at which momentum is injected by supernovae and stellar winds
(cf. CC; HAM; see Equation 4 above). HAM have used the measured central
pressures and the predicted momentum injection rates to estimate starburst
radii for 12 IR-selected galaxies. These radii agree surprisingly well
(typically
within 0.1 dex) with radii measured directly from radio, IR, and optical
imaging of the starbursts (see also L92; LH).
V work represented by
the measured pressure profiles
in the halos of starbursts with superwinds is in satisfactory agreement with
the time-integrated kinetic energy input predicted from the starburst (cf.
Equation 1 above).
emission from a portion of the starburst nucleus is being scattered into our
line-of-sight by dust grains in the halo of M 82, and that this
scattered light makes a significant contribution to the observed
H emission in the halo
(Scarrott, Eaton, & Axon
1991
and references therein). On the other hand,
the systematic spatial variations in the optical emission-line ratios
measured in the inner portion of the minor-axis nebula (that is, the gas
located within
about 30 degrees of the minor axis and within about a kpc from the nucleus)
show that the majority of the
H emission from this region
must be intrinsic
to the filament system rather than scattered light (e.g.,
McCarthy, Heckman, & van
Breugel 1987;
Bland & Tully 1988;
HAM). Indeed, the polarization
of the H within this region is
typically only 4 to 10%, compared to values
of 10 to 30% at greater angles off the minor axis
(Scarrott, Eaton, & Axon
1991).
It is also this inner region with lower polarization in which the clear
kinematic signature of an outflow is seen (HAM;
Bland & Tully 1988).
We suggest that scattered light contributes no more than about 30% of
the H
flux in this inner region. A spectropolarimetric, multi-emission-line
mapping
of the M 82 nebula with a spectral resolution of an
Ångstrom or
better would clearly be a magnificent Ph.D. thesis project!
filaments extending many kpc out of the galactic disk.
. The kinetic
energy of this material is then
about 1056 ergs, or about 20% of the time-integrated kinetic
energy of the
starburst in NGC 1808 over a period of 107 yr (the
dynamical lifetime of the
dust filament system). This is in good agreement with the prediction that
the kinetic energy in the shell of an energy-conserving wind-blown bubble
should be about 20% of the total time-integrated amount of injected kinetic
energy (cf.
Dyson 1989).