Comparison with UV Observations

The inner regions of the outflow in M82 have been detected in UV images taken with the Ultraviolet Imaging Telescope (UIT; [Stecher et al. 1992]). Figure 16 compares our deep H$\alpha$ image with the UV image observed in the 2490Å band by the UIT. The southern outflow is clearly evident in ultraviolet emission, out to at least a kiloparsec. The northern outflow is only barely visible in the UIT data, presumably because of the high disk extinction at ultraviolet wavelengths. The projected morphology of the southern UV outflow is similar to that at H$\alpha$ wavelengths, consisting of a relatively smooth, broad fan of emission. The ratio of the H$\alpha$ and UV surface brightnesses (each in units of ergs s^-1 cm^-2 arcsec^-2 Å^-1) in the outflow region is $\sim$8.8 ([Hennessy 1996]).

 

Figure: A deep H$\alpha$ image overlaid with a contour map of the 2490Å ultraviolet emission observed by the UIT. Tick marks are spaced at 1' intervals.  

Previous analyses of the UIT data ([Maran et al. 1991]; [Hennessy 1993]), as well as earlier balloon observations ([Courvoisier et al. 1990]; [Blecha et al. 1990]) have interpreted the minor-axis ultraviolet emission in M82 in terms of dust scattering of photons from the nuclear starburst by particles in the outflow. It has been suggested that filamentary structures can be seen in the UIT data ([Maran et al. 1991]), presumably because the higher relative densities in the filaments enhance scattering in those regions. The spatial distribution of the UV emission does not correlate well with the H$\alpha$,X-ray, radio continuum, or IR morphologies, although this may be due to variations in opacity ([Courvoisier et al. 1990]). One difficulty with the scattering picture for the production of the southern UV outflow is the surprising non-detection of emission in the UIT 1600Å far-ultraviolet (FUV) band ([Hennessy 1993]). The naive expectation for Rayleigh scattering would be to expect increased scattered flux at shorter wavelengths, even in the presence of substantial extinction along the line of sight. Ultraviolet observations of reflection nebula (e.g., [Calzetti et al. 1995]) and starburst galaxies ([Calzetti, Kinney, & Storchi-Bergmann 1994]) derive extinction curves that confirm this expectation.

An alternative explanation for the UV light is ``two photon'' continuum emission from ionized gas within the optical filaments. This emission arises from the spontaneous two-photon decay of the 2 ²S level of HI and is commonly seen in AGN and regions of low-density ionized gas. The spectral distribution of two-photon emission is symmetric, in photons per frequency interval, about a peak at 2431Å ([Osterbrock 1989]), very close to the UIT 2490Å band. Unlike the scattering function, two-photon flux decreases ($\sim\lambda^{-1}$) toward shorter wavelengths (see also [Hennessy 1996]). Preliminary analysis of the [SII] $\lambda\lambda$6719,6731 lines in optical spectra of the outflow from the Keck telescope confirm previous studies (e.g., [Rodriguez & Chaisson 1980]; [Houck et al. 1984]; [Heckman, Armus, & Miley 1990]) that the inner filament densities rarely exceed 500 cm^-3. As this is significantly below the critical density at which collisional effects influence the intensity of two-photon emission ($n_e\sim10^{4}$ cm^-3; [Osterbrock 1989]), we may anticipate the filamentary nature of the outflow in the UV imagery.

The observed H$\beta$/H$\alpha$ ratio of 0.25 for the outflow gas ([Heckman, Armus, & Miley 1990]) implies an H$\beta$/UV ratio of $\sim$2.2. This is approximately a factor of ten greater than the ratio of H$\beta$ to two-photon flux predicted by photoionization models ($\sim$0.1; [Dopita, Binette, & Schwartz 1982]; [Dopita 1997]), suggesting the presence of 2.5 mag more extinction at 2500Å. The observed extinction law for starburst galaxies ([Calzetti, Kinney, & Storchi-Bergmann 1994]; Fig. 5), combined with the observed H$\alpha$ extinction factor of 1.8 in the outflow ([Heckman, Armus, & Miley 1990]) indicates that this level of obscuration is entirely reasonable. Shock-induced two-photon emission is almost certainly ruled out, as it would require extinction by a factor of $\sim4$ to match the observed UV intensity, well above the value observed in the outflow regions. Spectral observations near the Balmer limit, $\sim$3650Å, where the two-photon continuum emission is enhanced relative to line emission, could be useful for determining the importance of this mechanism.