Recent hydrodynamic simulations and studies of the x-ray halo spatial distribution have cast doubt on this interpretation, however. Current hydrodynamic simulations ([Tomisaka & Bregman 1993]; [Suchkov et al. 1994]) derive gas temperatures an order of magnitude larger than earlier estimates (e.g., [Watson, Stanger, & Griffiths 1984]). This 108 K gas does not emit as strongly in the X-ray bands, producing insufficient soft and hard X-rays to account for the observations. Even [Chevalier & Clegg 1985] admit that the density falls too rapidly in their simple model to account for the x-ray photons as thermal. X-ray spectral studies derive a range of thermal gas temperatures and suggest alternative emission mechanisms ([Fabbiano 1988]; [Schaaf et al. 1989]; [Strickland, Ponman, & Stevens 1996]). Finally, correlations between our deep H imaging and Fabry-Perot observations and high-resolution ROSAT imagery support a non-thermal origin for at least a portion of the minor-axis x-ray emission.
Figure 13 compares the spatial distribution of soft X-rays observed by ROSAT with our Fabry-Perot H flux map. As was evident at larger radii in the comparison with the deep H imagery (Fig. 8), the X-rays and optical line emission are clearly correlated. On large scales, the minor-axis x-ray flux drops at a radius of approximately 500 pc from the nucleus, as does the H emission. But also on scales as small as 10 pixels (150 pc), the x-ray and optical emission appears well correlated. This implies that soft X-rays are being produced in regions very close to those which are producing H emission, a situation which is very difficult to understand in terms of a thermal emission mechanism.
These observations lend support to the x-ray emission mechanism suggested by several authors (e.g., [Chevalier & Clegg 1985]; [Suchkov et al. 1994]; [Strickland, Ponman, & Stevens 1996]): the soft X-rays arise from shocked disk and halo ``clouds.'' This shocked gas can produce both the observed x-ray and optical emission, accounting for the spatial correlation in Figure 13. A hybrid model seems necessary in which the higher gas temperatures and densities near the nucleus create a region dominated by shocks at interfaces with disk and halo gas clouds, while the cooler temperatures and lower densities at larger radii produce a decrease in optical emission and an increase in thermally-emitted x-rays. A comparison of the scale lengths of the H and x-ray emission along the minor axis confirms the more extended nature of the x-ray component: the H surface brightness along the minor axis is fit well by an exponential function, with a scale length of 250 pc. For the most distant H emission ( kpc), this exponential can be approximated by a power law of slope -2, essentially the same power law exponent measured for the X-rays at a comparable radius ([Bregman, Schulman, & Tomisaka 1995]). Beyond this radius, the optical surface brightness falls more rapidly than does the x-ray surface brightness.
Recent detailed modeling of the x-ray emission ([Bregman, Schulman, & Tomisaka 1995]) suggests a temperature at large radii of only K, implying an increasing role for thermal emission with radius. Similarly, hydrodynamic simulations find that the majority of the wind mass must be accumulated near the starburst region, not from evaporating halo clouds ([Suchkov et al. 1996]). It should be noted, however, that observations with the Ginga x-ray satellite have made the startling discovery of faint x-ray emission extending several tens of kiloparsecs from M82 ([Tsuru et al. 1990]). Hydrodynamic simulations have modeled this emission as shock-excited in nature, assuming that the outflow is much older, years ([Tomisaka & Bregman 1993]). Although this observation has yet to be confirmed, the rapid radial decrease in the wind pressure and density could propagate the wind shock to large distances from its starburst origins.
The true importance of shocks versus thermal emission can be estimated from optical line diagnostics, at least in the inner regions of the M82 outflow. As was pointed out in Section 3, the flux-weighted [NII]/H ratio from the Fabry-Perot data is highly uniform and low in the inner kiloparsec of the outflow; values of 0.3-0.6 are typical. However, we must be careful to use line ratios for individual kinematic components when drawing conclusions regarding the physics of the gas, especially when the components have been modeled as distinct physical regions. The [NII]/H line ratio of the individual components reveals a similar low value across the spatial extent of split lines, except in the inner collimated zone, where a higher [NII]/H ratio is seen (1.0), particularly in the low-velocity component. Although we were unable to resolve separate components in the [OIII] observations, we note that the [OIII]/H ratio exhibits a strong radial gradient, unlike the [NII]/H ratio. The [OIII]/H ratio increases from a value of approximately 0.03 at the center to 0.08 at a distance of 750 pc.
In order to investigate the importance of shock excitation for the optical filaments, we have compared the observed emission line ratios from the Fabry-Perot data with the standard emission-line galaxy diagnostic diagrams of [Veilleux & Osterbrock 1987]. Although the small number of emission line diagnostics at our disposal limits our analysis, we can nevertheless make a rough assessment of the influence of shocks using the [OIII]/H versus [NII]/H diagnostic diagram (Fig. 14; from Fig. 1 of [Veilleux & Osterbrock 1987]). Using an H/H ratio of 0.25 for the outflow gas ([Heckman, Armus, & Miley 1990]), we see that the emission line ratios from the southern wind lobe of M82 rest in the region of the diagram for starburst galaxies, as expected. The ratios are comparable to those for most cooler HII regions and HII region models. This immediately suggests an emission mechanism such as photoionization for the filaments, particularly near the nucleus, where the [OIII]/H ratio is lower. As we move out in radius, however, shocks appear to become more important as an excitation mechanism, as the increasing [OIII]/H ratio drives the locus in Figure 14 toward the region for non-thermally powered AGN.
In order to more directly interpret our emission line fluxes in light of a shock mechanism, we have also compared the observed line ratios with a recent set of high-velocity shock models ([Dopita & Sutherland 1995]). These models have been computed for shocks in the velocity range of 150-500 km s-1; the deprojected gas velocity of the wind in M82 is estimated to be at the upper end of this range. Again using an H/H ratio of 0.25 for the outflow gas ([Heckman, Armus, & Miley 1990]), a comparison with the [OIII]/H versus [NII]/H diagnostic diagram (Fig. 15; from Fig. 2b of [Dopita & Sutherland 1995]) shows that it is unlikely that the observed emission line flux from the inner outflow filaments arises entirely from shocks. There is simply not enough [OIII] emission observed in the inner kiloparsec of the M82 outflow. However, the increasing [OIII]/H ratio with distance from the nucleus suggests that shocks probably become important at larger radii, just as suggested by the observational diagnostic diagrams ([Veilleux & Osterbrock 1987]). Longslit optical observations have also reached the conclusion that the line ratios become more shock-like with increasing distance from the starburst (e.g., [Heckman, Armus, & Miley 1990]), although this has often not included analysis of individual velocity components.
Additional support for a photoionization mechanism for the inner optical filaments is provided by studies of the diffuse ionized medium (DIM) in NGC 891. In that galaxy, which has no outflow or other obvious sources of shock ionization, photoionization models have been used to understand the variation of line ratios with height above the disk plane. These models show that the [NII]/H ratio gradually decreases as the ionization parameter (the ratio of ionizing photons to gas density) increases. In contrast, the [OIII]/H ratio should increase rapidly with ionization parameter ([Sokolowski 1992]). Regardless of the presence of shocks then, the low value of [NII]/H and the gradually increasing value of [OIII]/H in the innermost filaments of M82 can be understood as the result of a gradual drop in filament density, relative to the number of ionizing photons from the starburst. In the outer filaments, however, the [NII]/H ratio begins to increase, presumably as a result of dilution of the radiation field in the expanding uncollimated bubble, as well as perhaps an increasing influence of shocks.
Recent studies of the influence of halo dust on these line ratio trends in NGC 891 ([Ferrara et al. 1996]) point out that the [NII]/H ratio may appear artificially low near the disk due to dilution by scattered radiation from disk HII regions. This suggests that the low [NII]/H ratios for the inner filaments in M82 may be due in part to higher dust densities in the inner halo, scattering disk radiation from the nuclear starburst. This proposition corresponds well with the high levels of polarization detected from the filaments and the exponential nature of the observed halo, although the observed polarization levels in M82 (10-15%) are much higher than those modeled in NGC 891 (1-2%).
Based upon these comparisons and our geometric models, we propose that the optical emission from the inner kiloparsec of the M82 filament network is, at least partially, due to photoionization of the sides of the cavity created by the outflow. The hot gas in the wind itself would be quite transparent to the UV ionizing photons from the starburst region, allowing the entrained disk and halo gas to be illuminated directly. The tipped geometry of the outflow cones probably places the systemic side of each cone more directly in the path of the photoionizing radiation from the central starburst, explaining the higher fluxes seen in the low-velocity components of the wind. However, small regions of higher [NII]/H ratios in the individual velocity components suggest that a complex combination of shock and photoionization is probably required in the violent collimated zone, where the disk gas is being entrained and drawn upward by the hot wind just as it leaves the luminous starburst region. Although our small field of view and sensitivity limits restrict our analysis of the more extended optical filaments, we confirm a trend toward more shock-like line emission in the outer regions of the outflow.