The basic conceptual picture of how starburst-driven winds work was established nearly two decades ago. In the Chevalier & Clegg (1985, henceforth CC) model, the very high rate per unit volume of core collapse SNe in starburst regions leads to young SNRs colliding before they have time to lose energy radiatively. Shocks in these collisions thermalize the kinetic energy of the ejecta, creating a very hot (T ~ 108 K), high pressure (P / k ~ 107 K cm-3) and low density gas that fills the volume of the starburst region. In the analytical CC model the multiple individual SN events are treated as a time-averaged, spatially-uniform injection of mass and energy. The merged SN-ejecta expands adiabatically, becoming supersonic at the edge of the starburst region and rapidly reaching terminal velocity of (2 / )1/2 ~ 3000 km s-1. Numerical simulations show that this wind expands preferentially along the minor axis of any disk-like galaxy forming a bipolar flow, sweeping up cooler, denser, ambient disk or halo gas, and stripping ambient gas from the walls of the cavity (Tomisaka & Ikeuchi, 1988; Strickland & Stevens, 2000; Suchkov et al., 1994). The ram pressure of the merged SN-ejecta accelerates this entrained gas to velocities of a few × 100 to 1000 km s-1. The multiple cool, warm and hot gas phases observed in superwinds (Dahlem, 1997) are all entrained ambient gas.
Is this simple scenario realistic, given what we have learned about starbursts in the last 20 years? Before the Chandra observations of Griffiths et al. (2000) there had been no believable detection of the very hot gas predicted by the CC model in any starburst region (the very hot gas that in theory drives superwinds), so why was this model accepted in the first place? In fact, convincing observational evidence for the SN-driven nature of superwinds, and the quantitative accuracy of the CC model, appeared soon after its publication. Modern observations continue to validate its most basic concepts.
The classic starburst M82 is a particularly convenient laboratory for investigating starbursts and superwinds, given that it is one of the closest powerful starburst galaxies (D=3.6 Mpc), and that all current SF occurs within a 400 pc radius of the nucleus. Its FIR luminosity, dominated by the starburst activity, indicates a SN rate of ~ 0.1 per year. In terms of SF rate per unit area M82 lies close to the upper limit found in starbursts at both low and high redshift (Meurer et al., 1997). Radio observations show ~ 40 compact, but spatially resolved, SNR remnants (Muxlow et al., 1994), that outline the starburst region as seen at other wavelengths. These are not the same SNe that drive the wind, but are a subset of the SN population that occur in dense gaseous environments, although the total fraction of the starburst region filled with such dense gas must be < 0.1 (Chevalier & Fransson, 2001). Optical, near and mid-IR observations show large numbers of massive star clusters (4 < log Mcluster < 6) within the starburst region (Lipscy & Plavchan, 2004; O'Connell et al., 1995; McCrady et al., 2003). Note that the fraction of massive stars (and hence eventual SNe) within such massive clusters is only 20% of the total, consistent with the value observed in many starburst galaxies (Meurer et al., 1995). Thus, to first order, energy and mass injection from SNe is distributed over the volume of the entire starburst region, in accord with the CC model.
McCarthy et al. (1987), and later Heckman et al. (1990) and Lehnert & Heckman (1996), used optical spectroscopy to map the pressure in warm ionized gas in central regions of M82 and other starburst galaxies with superwinds. Essentially this uses the low filling factor H-emitting gas clouds as tracers of the much hotter gas they are embedded in. At large radii the pressure drops rapidly with distance, as expected for a free radial wind. Within a certain radius rP the pressure is approximately constant, indicating that energy and mass injection is distributed within this region. The radial shape and normalization of these pressure profiles are consistent with the CC model. Heckman et al. (1990) demonstrated for many galaxies with superwinds that rP is the same as the size of starburst r*.
Note that if AGN drove these winds rP would not equal r*, even in models where energy injection is distributed over large regions due to diverted or halted jets (Colbert, 1997). A thermal AGN wind driven from the vicinity of the accretion disk (Schiano, 1985) would further differ from a starburst-driven wind, even if having initially similar total mass and energy injection rates. The CC model assumes radiative losses from the thermalized gas are negligible, which is true if energy and mass injection occurs over a volume similar to the size of an entire starburst region. Silich et al. (2003) have pointed out that if the injection region is much smaller than this, the increased density of the thermalized gas leads to significant radiative losses. Their model is essentially of flows from very massive (log M > 6) super star clusters, but I wish to point out that it would also apply to centrally-driven AGN winds. Strong radiative losses in AGN winds would thus alter the physical properties observed on the larger galactic scales from those in starburst-driven superwinds.
Recent X-ray observations of superwinds have validated the use of warm gas as a tracer of the hot merged SN-ejecta, as well as detecting the very hot plasma predicted by the CC model. It is only with the Chandra X-ray Observatory's arcsecond spatial resolution that emission from the multiple X-ray binaries formed in the starburst can be cleanly separated from the diffuse X-ray emission associated with the superwind (see Fig. 1). This is especially true in the hard X-ray band (E = 2 to 8 keV) where faint diffuse bremsstrahlung and line emission from T ~ 108 K gas would be expected to be seen. Griffiths et al. (2000) were the first to robustly detect diffuse hard X-ray emission in a starburst region, once again in the archetype M82. The spatial extent of this emission in the plane of the galaxy exactly matches that of the 40"-diameter starburst region, and matches the base of the superwind as seen at other wavelengths. More recent, higher spectral-resolution, Chandra observations (Strickland et. al, in preparation) show that the diffuse hard X-ray emission not purely thermal, but do confirm the presence of a E 6.7 keV Fe K-shell emission that must come from a log T > 7.5 K plasma. In general such diffuse hard emission has not been detected in other nearby starbursts, even with Chandra, as other starburst nuclei are more distant, and have smaller angular scales, than that in M82. Weaver et al. (2002) detected diffuse hard X-ray emission within the central ~ 10" of the nuclear starburst NGC 253, but argue that the properties of this emission are more consistent with gas photoionized by that galaxy's LLAGN.
Figure 1. The center of M82 in (a) the optical R-band, (b) H emission from log T ~ 4 gas, (c) Soft X-ray emission (E = 0.3-2.0 keV), (d) Hard X-ray emission (E = 2.0-8.0 keV), (e) Diffuse soft X-ray emission from log T ~ 6.5 gas only, X-ray point sources have been removed, (f) Diffuse hard X-ray emission, including some contribution from log T ~ 7.5 - 8 gas. Each image is 2 kpc on a side, centered on the dynamical center of the galaxy. X-ray images are from Chandra ACIS observations. The location of the ~ 40 compact radio SNRs that mark the starburst region are plotted on panels e and f. The base of the superwind matches the size of the starburst region. Adapted from Strickland et al. (2004a).