Further Spectroscopy of the Diffuse Ionized Gas in NGC 891 and Evidence for a Secondary Source of Ionization

Two long-slit spectra of the diffuse ionized gas (DIG) in NGC 891 are presented. The first reveals variations parallel to the major axis in emission-line ratios in the halo gas at z = 700 pc. It is found that filaments of Hα emission show lower values of [N II]/Hα, [S II]/Hα, and [O I]/Hα. Although this result is expected if the filaments represent the walls of evacuated chimneys, it merely reflects a more general correlation of these ratios with Hα surface brightness along the slit and may simply arise from radiation dilution effects. Halo regions showing low line ratios are probably relatively close to ionizing sources in the disk below. The results highlight difficulties inherent in observations of edge-on galaxies caused by lack of knowledge of structure in the in-plane directions. The [S II]/[N II] ratio shows almost no dependence on distance along the major axis or Hα surface brightness. Values of [O I]/Hα indicate that H is 80%-95% ionized (assuming T = 10 4 K), with the higher ionization fractions correlating with higher surface brightness. Much more interesting information on the nature of this gaseous halo comes from the second observation, which shows the vertical dependence of [N II]/Hα, [S II]/Hα, [O I]/Hα, and [O III]/Hβ through the brightest region of the DIG halo. The most surprising result, in complete contradiction to models in which the DIG is ionized by massive stars in the disk, is that [O III]/Hβ rises with height above the plane for z > 1 kpc (even as [N II]/Hα, [S II]/Hα, and [O I]/Hα are rising, in line with expectations from such models). The run of [S II]/[N II] is also problematic, showing essentially no contrast with z . The [O III] emission probably arises from shocks, turbulent mixing layers, or some other secondary source of ionization. Composite models in which the line emission comes from a mix of photoionized gas and shocks or turbulent mixing layers are considered in diagnostic diagrams, with the result that many aspects of the data can be explained. Problems with the run of [S II]/[N II] still remain, however. There is a reasonably large parameter space allowed for the second component. For the photoionized component, only matter-bounded models succeed, putting a fairly strong restriction on the clumpiness of the halo gas. Given the many uncertainties, the composite models can do little more than demonstrate the feasibility of these processes as secondary sources of energy input. A fairly robust result, however, is that the fraction of Hα emission arising from the second component probably increases with z . From values of [O I]/Hα, H is essentially 100% ionized at z = 0 kpc and 90% ionized at z = 1 kpc (again assuming T = 10 4 K).