Data Analysis

Assuming that the emission-line profiles observed in KUG 1031+398 are the result of the contributions from several clouds, we tried to model the observed spectra with the smallest possible number of line sets, each set including three Gaussians (modeling H$\alpha$ and the [NII] lines, or H$\beta$ and the [OIII] lines) having the same velocity shift and width, with the additional constraint that the intensity ratio of the two [NII] (respectively [OIII]) lines was taken to be equal to the theoretical value of 3 (respectively 2.96) (Osterbrock 1974). In a physically meaningful and self-consistent model, the components found when fitting the blue and red spectra should have velocity shifts and widths compatible within the measurement errors.

The spectra were de-redshifted assuming z = 0.0434 and analyzed in terms of Gaussian components as described above. We discovered first that the core of the lines could not be fitted by a single set of narrow Gaussian profiles. To get a satisfactory fit, two sets of Gaussian components are needed: the first, unresolved (and subsequently taken as the origin of the velocity scales) has $\lambda$6583/H$\alpha$ = 0.55, $\lambda$5007/H$\beta$ = 1.27, and corresponds to a HII region; the second is resolved (FWHM $\sim$ 350 kms^-1, corrected for the instrumental broadening), blueshifted by $\sim$ 95 kms^-1 with respect to the narrow components and has line intensity ratios typical of a Seyfert 2 ($\lambda$6583/H$\alpha$ = 0.84, $\lambda$5007/H$\beta$ > 10).

  

Figure: Blue (a) and red (b) spectra of KUG 1031+398 in the rest frame; in (b) we also give the spectrum before correcting for the atmospheric absorption (dotted line). The narrow core components (c and d) were fitted with Gaussians and subtracted from the original data, the result being shown in (e) to (h). In (e) and (f), we show our best fit (solid line) together with the data points (crosses); the lower solid lines represent the residuals. In (g) and (h) we give the fit and residuals obtained when an ``intermediate'' component is imposed, as described in the text.

At this stage, we removed from the blue and red spectra the best fitting core (the HII region and the Seyfert 2 nebulosity, Fig. 1c and d), obtaining two spectra we shall call ``original data - core''. The blue one was then fitted with a broad H$\beta$ Gaussian component and two sets of three components modeling the narrow H$\beta$ and [OIII] lines. The result is very suggestive: one set has a strong H$\beta$ line and very weak negative [OIII] components, while the other set displays a strong [OIII] contribution and a weak negative H$\beta$ component, showing that we have in fact a H$\beta$ component with no associated [OIII] emission and [OIII] lines with a very weak (undetected) associated H$\beta$; in other words, the region producing the H$\beta$ line does not emit forbidden lines, while the [OIII] emitting region has a high $\lambda$5007/H$\beta$ ratio, which are the characteristics of the ``broad'' and ``narrow'' line regions in Seyfert 1 galaxies, respectively.

With these results in mind, we optimized this last fit by using a Lorentzian profile for the H$\beta$ line, with no associated [OIII] emission, and a set of three Gaussians for the remaining contribution coming from the ``narrow'' components; to avoid an unphysical negative intensity for the H$\beta$ line, we forced $\lambda$5007/H$\beta$ to be equal to 10, which is the ratio usually found for the narrow component in Seyfert galaxies. The best fit is presented in Fig. 1e. The H$\beta$ Lorentzian component is blueshifted by 160 kms^-1 with a width of 920 kms^-1; the [OIII] lines are blueshifted by $\sim$ 395 kms^-1 and their width is $\sim$ 1120 kms^-1.

We have also analyzed the ``original data - core'' red spectrum (Fig. 1f) with one Lorentzian H$\alpha$ component and a set of three Gaussians (for the H$\alpha$ and [NII] lines) with the constraint that $\lambda$6583/H$\alpha$ = 0.9, for which we have found a FWHM of $\sim$ 770 kms^-1 and a blueshift of 375 kms^-1. The H$\alpha$ Lorentzian component, blueshifted by 55 kms^-1, has a width of 1030 kms^-1, in reasonable agreement with the width of the corresponding H$\beta$ component. The Lorentzian Balmer components, without any measurable associated forbidden line, would qualify KUG 1031+398 as a NLS1 with, in fact, very narrow lines. The other system of lines, with a very high $\lambda$5007/H$\beta$ ratio, $\lambda$6583/H$\alpha$ $\sim$ 0.9 and FWHM $\sim$ 945 kms^-1, is analogous to what is usually found in Seyfert 2s and corresponds to a NLR cloud.

At last, we fitted the ``original data - core'' blue spectrum with a broad H$\beta$ Gaussian component and one set of three Gaussians (modeling H$\beta$ and the [OIII] lines) for which we set the $\lambda$5007/H$\beta$ ratio to the value found by Mason et al. for the ``intermediate'' component, i.e., 1.42. The red spectrum was fitted with two H$\alpha$ components, for which we fixed the redshifts to the values obtained in the blue spectrum profile fitting analysis. The resulting fits and residuals, shown in Figs. 1g and h, seem to be significantly worse than the ones given in Figs. 1e and f, showing that the presence of an ``intermediate'' component is not required by the data.