Astron. Astrophys. 320, 378-394 (1997)

1. Introduction

Only within the last two decades has it been realised that the interplay between vigorous star formation and the state of the interstellar medium (ISM) can have profound implications for the evolution of galaxies and their environments (see for example Norman & Ikeuchi 1989). Starbursts, and in particular the galactic mass outflows or winds driven by thermalised stellar winds from massive stars and supernovae, have implications for systems of all sizes. Galactic winds may be responsible for the destruction of dwarf galaxies (Dekel & Silk 1986; Heckman et al. 1995), enrichment of the ICM and IGM in clusters and groups and removal of gas from merger remnants.

Heckman et al. (1993) provide a comprehensive review of the observational data and theory of galactic winds. Briefly, thermalised kinetic energy from stellar winds and supernovae from massive stars in the starburst creates a hot () bubble in the ISM. This expands, sweeping up ambient material into a dense shell. Eventually the bubble breaks out of the disk of the galaxy along the minor axis. The hot wind then escapes freely at several thousand kilometers per second. The dense shell fragments due to Rayleigh-Taylor instabilities, and is carried along by the wind at velocities of order hundreds of kilometers per second. This or ambient clouds overrun by the wind is the source of optical emission line filaments, and possibly the soft X-ray emission.

The archetypal starburst M 82 presents possibly the best test case, given its proximity (, Freedman et al. 1994) and the wealth of observational data available. Its high infrared luminosity (, Rieke et al. 1993), disturbed morphology, population of supernova remnants (Muxlow et al. 1994) and luminous young super star clusters (O'Connell et al. 1995) are all signatures of a strong burst of star formation. The starburst was probably caused by a close interaction with M 82's nearby (projected distance ) neighbour M81 about ago (Cottrell 1977), and a tidal bridge of HI connects the two galaxies (Yun et al. 1993).

A set of emission line filaments along M 82's minor axis show velocities consistent with gas motions along the surface of a cone at (Axon & Taylor 1978): cooler material swept out of the galaxy by the much hotter wind. The emission defines an outflow that has a radius at a distance of above the galactic plane (Götz et al. 1990), and is approximately cylindrical for . At larger z the blowout flares out to a cone with an opening angle (see Fig. 5 in McKeith et al. 1995). Within of the nucleus the inferred electron density in the filaments decreases with increasing z. McKeith et al. (1995) claim this is consistent with a model, as would be expected if the filaments were in pressure equilibrium with freely expanding wind such as that proposed by Chevalier & Clegg (1985). A similar density decrease was also inferred by McCarthy et al. (1987).

Additional evidence for a galactic wind is the synchrotron emitting radio halo, extended preferentially along the minor axis (Seaquist & Odegard 1991), due to relativistic electrons from supernovae swept out from the starburst region by the wind. This has a maximum extent comparable to the X-ray emission (this paper). A steepening of the spectral index interpreted as arising from energy loss by Inverse Compton (IC) scattering of the electrons off IR photons, allows an estimate of the speed with which the electrons are being convected outwards, assuming re-acceleration in shocks to be negligible. Seaquist and Odegard (1991) claim a conservative estimate of the wind velocity, allowing for the uncertainties, lies in the range 1000-3000 , similar to that predicted from theory (Chevalier & Clegg 1985; Heckman et al. 1993).

Schaaf et al. 's (1989) suggestion that X-rays produced by this IC scattering could be the source of the observed X-ray emission is argued against by Seaquist et al. (1991) who predict , in contrast with the value we derive below of in the ROSAT band.

X-ray observations should provide a direct method of testing the galactic wind paradigm, given that thermalised stellar wind and supernovae ejecta is expected to have a temperature of . Previous X-ray observations of M 82 have suffered from poor sensitivity, poor spectral resolution and to a lesser extent poor spatial resolution. Watson et al. (1984) detected several very luminous () sources along with diffuse emission using the Einstein HRI. The diffuse emission was seen to extend out to () to the south-east and to the north-west. Spectral fits using the Einstein IPC and MPC were inconclusive in that they were unable to distinguish between a power law or a thermal origin for the emission. Given that they were unable to separate the point sources and the diffuse emission, this is not surprising.

A reanalysis of the Einstein data by Fabbiano (1988) did attempt to separate the different components. The MPC (without any imaging capability) fitted temperature of would be dominated by the nuclear source and hence is not an estimate of the wind temperature. The IPC gives for the emission within of the nucleus, compared to for the emission between . The radial surface brightness in the IPC falls off approximately as , consistent with the expectation for a free wind.

Schaaf et al. (1989) use an EXOSAT observation together with the Einstein data. The EXOSAT spectrum is consistent with either a power law or a Raymond & Smith plasma with temperature , again without any separation of point source and diffuse components.

Although the extent of the emission seen within the EXOSAT and Einstein observations compares well with the higher sensitivity observation of ROSAT (Fig. 1), Tsuru et al. (1990) from observations with Ginga, claim evidence for a very extended, halo. Two north-south scans of a region centred on M 82 show excess flux to the north of M 82, but not to the south. A spectral fit to this emission is essentially unconstrained, with a temperature in the range 1-11 keV . Tsuru et al. argue that a single point source could not produce the observed feature, as the position of the extra source is inconsistent between the two scans. They concede this could be due to two or more point sources, but estimate the chance of finding two sources of suitable flux in such a small region as 4 square degrees is %. The dynamical age of such a halo is , hence this might be the remains of a wind from a starburst ago.

Fig. 1. Contours of X-ray emission (0.1- ) from the PSPC overlaid on a digitised sky-survey optical image of M 82. The X-ray emission has been lightly smoothed with a Gaussian of standard deviation to suppress noise. The contour levels increase in factors of two from cts arcmin-2 ( above the background).

The ROSAT HRI observations of M 82 (Bregman et al. 1995) show three sources within the nuclear region of M 82, although two of them have very low S/N values above the strong and spatially varying wind emission. A very bright source present in the Einstein data appears to have faded away completely (Collura et al. 1994), although the main nuclear source is at a position consistent with the Einstein observation.

Bregman et al. (1995)analyse the diffuse emission without the benefit of any spectral information. They conclude that the extended emission along the minor axis is consistent with an outflow of gas with opening angle that decreases with increasing radius within of the nucleus and at constant opening angle at larger radii. They model the emission successfully by adiabatically expanding gas of constant mass flux, and predict a decrease in temperature of the gas with increasing radius.

We report below, an analysis of the ROSAT PSPC and HRI observations of this X-ray emission. The PSPC's mixture of good spatial and spectral capabilities compared to any other X-ray instrument, allow the best determination yet of the properties of this emission. In particular, we can separate point source and diffuse emission, and investigate the variation of spectral properties as a function of distance from the nucleus. For the first time, we show that the diffuse emission is thermal in origin, and obtain temperatures, emission measures, metallicities, and, for an assumed geometry, electron densities, gas masses and total energies. We compare our results with Chevalier & Clegg's (1985) analytical model of a galactic wind, and a simple model in which the emission comes from shock heated clouds rather than the wind itself. Our results allow us to reject the possibility that the X-ray emission comes from the wind itself, and show that it could be consistent with emission from shock heated clouds.

© European Southern Observatory (ESO) 1997

Online publication: June 30, 1998
helpdesk.link@springer.de