Astron. Astrophys. 327, 72-80 (1997)

2. The ultraviolet spectrum of NLS1

2.1. The data

We have cross-correlated the NLS1 sample discussed by Boller et al. (1996) against the IUE database and found that 11 out of 30 NLS1 for which ROSAT data are available, had been observed with the short wavelength spectrograph on IUE. This subsample was increased by our own observations of three more NLS1: IRAS 13224-3809, KUG 1031+398 and Mrk 1044. The log of our IUE observations is shown in Table 1. The whole sample is described in Table 2, where we list the most common object names, the redshifts (from Boller et al. 1996) and some continuum properties as explained below.

Table 1. Log of IUE observations

Table 2. Continuum properties

The sample discussed hereafter is in no way statistically complete; it includes those galaxies that at some stage have been considered interesting by the scientific community and the IUE time allocation committees.

All the IUE spectra were re-extracted using the Final Archive processing software (Nichols et al. 1993). Foreground galactic reddening has been corrected from the HI column densities map in Dickey & Lockman (1990), converted to according to the relation: . All the spectra have been redshift corrected to get the wavelength and fluxes in the objects rest frames. In order to increase the signal-to-noise ratio, all the spectra of every single object have been averaged together (see Sect. 2.2 for a discussion of their variability properties).

The UV continuum has been measured in a 40 Å band centered at 1450 Å , which is apparently free of absorption and/or emission features. Two (IC3599 and 1652+396) out of the 14 objects were not detected; the signal-to-noise ratio (S/N) is smaller than 10 in the average spectrum of 5 objects. The average fluxes are given in Table 2. In this table we also list the fluxes in the four IRAS bands from the NED database as well as the ROSAT fluxes and spectral slopes reported by Boller et al. (1996).

2.2. Variability

One of the most general properties among normal Seyfert 1 galaxies is the variability in the UV continuum and broad emission lines. The observed UV continuum variations are a few percent on time scales of a day or less and tens of percent on time scales of several days. However, there is not very much information in the literature about optical and/or UV variability in NLS1, although they are probably the AGN's which show the fastest variations in the soft X rays. Only very recently, a sample of 12 NLS1 has been systematically monitored in order to search for optical variability (Giannuzzo & Stirpe 1996). Ten of these NLS1 showed significant variations in the optical continuum and permitted lines over a time interval of one year.

We have found significant changes in the continuum flux in 2 (Mrk 1044 and IRAS13224-3809) out of the 11 detected objects for which more than one spectrum is available. Mrk 1044 was observed with IUE on December 1 and 20, 1995. Between these two dates, the continuum increased by 38%. Unfortunately, in the first SWP spectrum the peaks of the strongest emission lines (Ly , CIV) were saturated. Nevertheless, neither the weaker lines (SiIV, HeII) nor the wings of Ly and CIV show evidences of variability. For Mrk 1044, Giannuzzo & Stirpe (1996) find variations in the H and H fluxes of -14% and -24% , respectively, between October 1993 and September 1994. They do not report on the continuum variability.

There are 11 SWP spectra of IRAS13224-3809 from January 1993 to February 1996. The variability of the continuum flux during this time is 24% (r.m.s.), with a ratio of the maximum to the minimum flux close to 2. The analysis of the spectra taken during 1993 showed changes in the profile of Ly that could be attributed to a variable narrow absorption (Mas-Hesse et al. 1994). A detailed analysis of the whole data set for this object is deferred to a later paper.

The number of IUE observations for the other NLS1 in the sample is rather small, so that it is not possible to study their variability properties.

In summary, the available data suggest that, in the UV, NLS1 can vary at least as fast as normal Seyfert 1 galaxies on time scales of several days. With the existing IUE data it has not been possible to check whether this class of objects varies faster than normal Seyfert 1s as in the soft X rays.

2.3. The continuum spectral energy distribution: comparison with normal Seyfert 1 galaxies

The spectral energy distribution (SED) of normal broad line Seyfert 1 galaxies is dominated by the "big blue" bump (Sanders et al. 1989), extending from 4 000 Å to beyond the shortest observable wavelengths in the UV region ( 1000 Å). How far this bump extends into the extreme ultraviolet (EUV) is not known because of the difficulty of observing extragalactic objects in this spectral region. In this respect, the "soft X-ray excess" found in many AGN is sometimes interpreted as the high energy end of the bump. It is generally thought that the big blue bump is due to the emission of an accretion disk around a super-massive black hole, although there are other alternatives proposed (like free-free emission from optically thin gas clouds or the non-standard model of emission from a powerful starburst). In all these models, the UV emission region would be illuminated by the hard X-ray continuum, as required by the observed properties of the X-ray and the UV spectra, together with their correlated fluctuations.

Many AGN emit similar amounts of energy in the UV-optical and FIR regions. There is growing evidence that the nuclear FIR emission in AGN is due to thermal radiation by dust, although this is still controversial. Moreover, for some objects it has been argued that the FIR emission is not directly related to the active nucleus, but to circumnuclear star forming regions (e.g. Mas-Hesse et al. 1995). The spectral index between 25 and 60 m is a good indicator of the relative contribution of star formation processes to the FIR emission in AGN (Mas-Hesse et al. 1995); steep FIR spectra are typical of star forming regions, while flat FIR spectra indicate the nuclear emission (independently of its nature) dominates.

We have compared the SED of NLS1 with that of normal broad emission line Seyfert 1 galaxies in order to investigate whether the continuum emission is originated by the same mechanisms in both types of objects. The SED (from the far IR, 100 m, to the soft X-rays, 2.4 keV) of our sample of NLS1 galaxies is listed in Table 2. For comparison we have selected the sample from the Walter & Fink (1993) study. These authors analyzed the ROSAT All Sky Survey data of 58 Seyfert 1 galaxies for which IUE spectra are available. To compare the X-ray fluxes in both samples it has been necessary to translate the Boller et al. (1996) 2 keV monochromatic fluxes to integrated fluxes in the 0.1-2.4 keV band, as given by Walter & Fink (1993). The translation has been performed using the soft X-ray spectral indexes given by Boller et al. (1996).

Five out of the 58 Walter & Fink objects are indeed NLS1, also included in the Boller et al. (1996) sample. Prior to the comparison of both samples, we have used the reported properties of the common objects to check the consistency of the two studies. We have found differences in the reported soft X-ray spectral indexes of the five NLS1 of up to 25% , but within the 3 error. Similarly, the fluxes derived by Walter & Fink (1993) are always larger than those inferred from the Boller et al. (1996) data. Nonetheless, in any case the differences are smaller than 3 , according to the errors given by Walter & Fink (1993). Therefore, we can be confident that the results of both studies are consistent within their errors bars. The 5 NLS1 in Walter & Fink (1993) have been excluded of their sample, leaving a set of 53 normal broad line Seyfert 1 galaxies. We also recall here that the Boller et al. (1996) sample of NLS1 galaxies has been enlarged with three more objects observed by us with IUE (Sect. 2.1).

In Table 3 we give the mean and standard deviation (S.D.) of some parameters in both samples, as well as the number of objects for which each parameter is available and the probability that the parameter is normally distributed. The luminosities in the IRAS bands, in the UV (1450 Å) and in the ROSAT band, as well as the ratios between some of them are given. The last column in the table is the significance of the hypothesis that both samples come from the same parent population according to the Mann-Whitney rank sum test.

Table 3. Comparison of NLS1 and normal Seyfert 1 continuum properties

The smallest significance (i.e., the highest probability that the two samples come from different parent distributions) is found for the spectral index in the ROSAT band (). However, it should be noted that the S.D. of is larger in the sample from Boller et al. (1996); many NLS1 have well within the typical values found for normal broad emission lines Seyfert 1, but there are some NLS1 that show much steeper soft X-ray spectra. Therefore, a very steep soft X-ray spectrum is not a characteristic of all NLS1, although the steepest spectra among Seyfert 1 galaxies are found in NLS1. In spite of this difference in spectral slopes, there is no statistically significant (significance 0.10) difference in the total ROSAT luminosity between NLS1 and normal Seyfert 1 galaxies.

The other parameters that are different in the two samples (at the 0.05 significance level) are the UV luminosity and the ratios in which this luminosity is involved. The NLS1 are, on average, 6 times fainter in the UV than normal broad emission lines Seyfert 1, although the total range spanned by the NLS1 ( erg s^-1) is well within the total range for normal Seyfert 1 galaxies ( erg s^-1). The average / and /L 60 ratios are correspondingly smaller in NLS1 than in normal Seyfert 1; furthermore, the smallest absolute values of these ratios correspond to NLS1. It is also worth noting that the "normal" Seyfert 1 with / and /L 60 closer to the extreme values found for NLS1 are NGC 4051 and NGC 1566. Winkler (1992) gives a FWHM for the H line in NGC 1566 of 1800 km/s and Filippenko & Sargent (1985) describe the Balmer lines of NGC 4051 as "not very broad compared with those in most Seyfert 1 galaxies".

From Table 3, the high degree of similarity in the FIR properties between NLS1 and normal broad emission lines Seyfert 1 is remarkable. This result is consistent with the Halpern & Oke (1987) results. The spectral index between 25 and 60 m () of the five NLS1 discussed by Halpern & Oke (1987) spans the whole range found for normal Seyfert 1 galaxies (Miley et al. 1985, Mas-Hesse et al. 1995). For our sample, we find that only two out of ten NLS1 detected at 25 and 60 m have steeper than -1.5, typical of star forming regions. Although our sample is rather small, it is still worthy to note that the fraction of steep FIR spectrum NLS1 is roughly consistent with that found among normal Seyfert 1 by Mas-Hesse et al. (1995).

We have compared the luminosity in all the IRAS, UV and ROSAT bands versus and L 60. We find a statistically significant correlation among the luminosities in all IRAS, UV and ROSAT bands. Moreover, there is no statistical evidence for the NLS1 and Seyfert 1 galaxies to show different slopes in the linear regression fits. However, the positive correlations found in all continuum bands are not held when the fluxes, instead of luminosties, are considered (Fig. 1). When the distance effect is removed, the correlations between low energy (IRAS) and high energy (UV, ROSAT) bands disappear, leaving only the correlation among IRAS bands on one side and between the UV and ROSAT bands on the other side. These results suggest that the mechanisms producing the UV and soft X-ray photons are strongly related, but the connection between the production of FIR radiation and UV - soft X-rays is not straightforward.

Fig. 1. Flux-flux diagrams for different continuum bands. Note that the higher energy bands (UV, soft X-rays) are not correlated with lower energy bands (FIR). Filled circles represent NLS1 galaxies and crosses normal Seyfert 1.

The conclusion that emerges from the comparison of the luminosity in the FIR, UV and soft X rays continuum bands is that the SED of NLS1 and normal broad emission lines Seyfert 1 galaxies are very similar, except in that some, but not all, NLS1 have steeper soft X-ray spectra and that NLS1 tend to be somewhat underluminous in the UV region.

2.4. Line profiles

A first look to the line profiles suggests the presence of broad wings in the strongest emission lines (Fig. 2). In order to test it, we have fitted the Ly , C 1550 and He 1640 line profiles with first only one and then two gaussian components (for the Ly profile an additional gaussian has been included to account for N 1240). The best fit parameters for the case that results in a smaller reduced chi-square ¹ ( = / , where is number of degrees of freedom) are shown in Table 4. For the seven objects where the signal-to-noise ratio (S/N) in the continuum is larger than 10, we note that the multi-component fit is preferred in at least one of the emission lines. Moreover, when the line profile is better fitted by two gaussians, we find that one is slightly broader than the IUE Point Spread Function (FWHM 1500 km/s) and the other has a FWHM similar to those found in normal Seyfert 1 galaxies (FWHM 5000 km/s). For the rest of the objects, the multi-component fit does not improve with respect to the fit with a single gaussian. However, we point out that the spectra of these objects are those which have the lowest S/N making it difficult to detect broad line wings. In any case, the line widths for the single-component fits tend to be larger than the width of the narrow component in the multigaussian fits. We want to stress that it is not our aim to identify each gaussian component in the fits with physically different regions. The purpose of the fits is to confirm or reject the presence of broad wings in the emission line profiles.

Fig. 2. Ly and CIV profiles of some NLS1 observed with IUE, showing broad wings similar to those found in normal broad line Seyfert 1 galaxies. Thin lines show the gaussian components discussed in the text, while the thick line shows the total fit.

Table 4. Results of line profile fits to the UV emission lines

Our results strongly suggest the presence of high velocity line emitting gas in the nuclei of NLS1 and that the profiles are qualitatively similar to those found in normal Seyfert 1, in the sense that they can be roughly characterized by a narrow plus a broad component.

As mentioned above, NLS1 as a class are characterized by optical hydrogen emission lines broader than their neighbour forbidden lines, but narrower than the Balmer lines found in normal Seyfert 1 galaxies (Osterbrock & Pogge, 1985). Nevertheless, after finding evidence of broad wings (FW0I 10,000 km/s and FWHM 5,000 km/s) in the permitted UV lines of some NLS1, we have specifically searched for such broad wings in the hydrogen optical lines. In their variability analysis of NLS1 galaxies, Giannuzzo & Stirpe (1996) present H and H spectra of three NLS1 objects for which we find broad wings in the UV lines (Mrk 359, Mrk 1044 and Akn 564). To check the presence of broad wings, we have done a multi-gaussian fit (similar to that performed with the UV lines) to these spectra, kindly provided by Giannuzzo & Stirpe. The fits require, first, three narrow gaussian lines whose width is determined by the spectral resolution: one gaussian for the Balmer line (H or H ) in the range and two additional ones to account for either the O 4959,5007 or the N 6548,6584 doublet. In the H range, two more gaussians, slightly broader (i.e. 1000 km/s), are included in the fits to account for the FeII emission in Akn 564 and Mrk 1044. In addition to that, both H and H in the three objects require at least two more, broader, gaussian components to account for the total profile. The first of them is of intermediate width, km/s, and the second has an FWHM less than 3000 km/s, narrower than the broadest UV component in the same objects. In order to check the existence of broader wings, we have tried another fit in which the width of the broadest line is fixed to 5000 km/s. The result is clearly worse as the r.m.s of the residuals is much larger. Therefore, we can only put upper limits to the flux of a putative optical component of 5000 km/s FWHM, as it is certainly not detected in the available spectra. These estimates can then be used to infer the upper limit of the broad Ly /H ratio. From the residuals of the multigaussian fits we obtain that, for the three NLS1 studied, the broad Ly /H ratio is larger than 100. This lower limit is also confirmed when directly estimated from the S/N in the wings of the lines.

© European Southern Observatory (ESO) 1997

Online publication: April 8, 1998
helpdesk.link@springer.de