0706.2574/ms.tex
1: \documentclass[12pt,preprint]{aastex}
2: 
3: \shorttitle{The NLR of type 1 AGN}
4: \shortauthors{Xu et al.}
5: 
6: \newcommand{\mbh}{M_\bullet}
7: \newcommand{\hst}{{\it HST}}
8: \newcommand{\kms}{km s$^{-1}$}
9: \newcommand{\oiii}{[\ion{O}{3}]}
10: \newcommand{\oii}{[\ion{O}{2}]}
11: \newcommand{\oi}{[\ion{O}{1}]}
12: \newcommand{\peroi}{\ion{O}{1}}
13: \newcommand{\caii}{\ion{Ca}{2}}
14: \newcommand{\siii}{\ion{S}{3}]}
15: \newcommand{\ciii}{\ion{C}{3}]}
16: \newcommand{\sii}{[\ion{S}{2}]}
17: \newcommand{\nii}{[\ion{N}{2}]}
18: \newcommand{\neiii}{[\ion{Ne}{3}]}
19: \newcommand{\nev}{[\ion{Ne}{5}]}
20: \newcommand{\fex}{[\ion{Fe}{10}]}
21: \newcommand{\fexi}{[\ion{Fe}{11}]}
22: \newcommand{\feii}{\ion{Fe}{2}}
23: \newcommand{\heii}{\ion{He}{2}}
24: \newcommand{\hii}{\ion{H}{2}}
25: \newcommand{\hb}{H$\beta$}
26: \newcommand{\hbb}{H$\beta_{\rm b}$}
27: \newcommand{\hbn}{H$\beta_{\rm n}$}
28: \newcommand{\han}{H$\alpha_{\rm n}$}
29: \newcommand{\ha}{H$\alpha$}
30: \newcommand{\LLedd}{${\rm L/L_{Edd}}$}
31: \newcommand{\LLeddbol}{${\rm L_{bol}/L_{Edd}}$}
32: \newcommand{\xLLedd}{${\rm L_X/L_{Edd}}$}
33: \newcommand{\n}{$n_{\rm e}$}
34: \newcommand{\te}{$T_{\rm e}$}
35: \newcommand{\cm}{cm$^{-3}$}
36: 
37: \slugcomment{ApJ accepted [00 00 2007]}
38: 
39: \begin{document}
40: 
41: \title{The narrow-line region of narrow-line and broad-line type 1 
42: Active Galactic Nuclei I. A zone of avoidance in density}
43: 
44: \author{Dawei Xu}
45: \affil{National Astronomical Observatories, Chinese Academy of
46:        Sciences, Beijing 100012, China; dwxu@bao.ac.cn}
47: 
48: \and
49: \author{Stefanie Komossa}
50: \affil{Max-Planck-Institut f\"ur extraterrestrische Physik,
51: Giessenbachstrasse 1, 85748 Garching, Germany; skomossa@mpe.mpg.de}
52: 
53: \and
54: \author{Hongyan Zhou} 
55: \affil{Department of Astronomy, University of Florida, Gainesville, FL32611, 
56: USA; and Max-Planck-Institut f\"ur extraterrestrische Physik,
57: Giessenbachstrasse 1, 85748 Garching, Germany;
58: and Center for Astrophysics, University of Science and 
59: Technology of China, Hefei, China; zhou@astro.ufl.edu}
60: 
61: \and
62: \author{Tinggui Wang}
63: \affil{Center for Astrophysics, University of Science and Technology of China,
64:        Hefei, China; twang@ustc.edu.cn}
65: 
66: \author{Jianyan Wei}
67: \affil{National Astronomical Observatories, Chinese Academy of
68:        Sciences, Beijing 100012, China; wjy@bao.ac.cn}
69: 
70: \begin{abstract}
71: The properties of narrow-line Seyfert\,1 (NLS1) galaxies,
72: the links and correlations between them, and the physics
73: behind them, are still not well understood. 
74: Apart from accretion rates and black hole masses,
75: density and outflows were speculated to be among the main drivers
76: of the NLS1 phenomenon.
77: Here, we utilize the diagnostic power of the \sii\ 
78: $\lambda\lambda 6716,6731$ intensity ratio to measure the density 
79: of the narrow-line region (NLR)
80: systematically and homogeneously for a large sample of NLS1 galaxies, 
81: and we perform a comparison with a sample of broad-line type\,1 
82: Active Galactic Nuclei (AGN). 
83: We report the discovery of a '{\em zone of avoidance}' in  
84: density in the sense that AGN       
85: with broad lines (FWHM(H$\beta$) $>$ 2000\,\kms) avoid low densities, 
86: while NLS1 galaxies show a wider distribution in the NLR density, 
87: including a significant number of objects with low densities.
88: A correlation analysis further shows that the Eddington ratio 
89: \LLedd\ anti-correlates with density. 
90: We investigate a number of different models  
91: for the '{\em zone of avoidance}' in density. 
92: Supersolar metallicities and temperature effects,
93: a strong starburst contribution in NLS1 galaxies, 
94: different NLR extents and selective obscuration are considered unlikely. 
95: Possible differences in the fraction of matter-bounded clouds
96: and differences in the interstellar media of the host galaxies
97: of NLS1 galaxies and broad-line Seyfert 1 (BLS1) galaxies 
98: can only be tested further with future observations.  
99: We tentatively favor the effects of winds/outflows, stronger
100: in NLS1 galaxies than in BLS1 galaxies, to explain the observations.  
101: \end{abstract}
102: 
103: \keywords{galaxies: density -- galaxies: ISM -- 
104: galaxies: emission lines -- galaxies: active -- galaxies: Seyfert}
105: 
106: \section{Introduction}
107: Optical and X-ray observations over the past few decades revealed a new
108: sub-class of AGN, termed Narrow-line Seyfert\,1 (NLS1) galaxies 
109: (e.g., Gaskell 1984; Osterbrock \& Pogge 1985). 
110: NLS1 galaxies are intriguing due to their
111: extreme emission line and continuum properties. Their optical broad
112: lines are narrower (FWHM \hb\ $\le$ 2000 \kms) than in `normal' 
113: broad-line Seyfert\,1 (BLS1) galaxies and they show strong \feii\
114: emission. Their X-ray spectra are sometimes, but not always, very soft
115: (e.g., Zhou et al. 2006, and references therein). Many emission-line
116: and continuum 
117: properties of AGN were found to correlate strongly with each other
118: (e.g., Boroson \& Green 1992, BG92 hereafter; Wang et al. 1996; 
119: Lawrence et al. 1997; Grupe et al. 1999; Vaughan et al. 2001; 
120: Xu et al. 2003; Sulentic et al. 2000, 2003; Grupe 2004).
121: The strongest variance, often referred to as `Eigenvector\,1' (EV1), is
122: defined by the correlation between the width of the \hb\ emission
123: line and the strength of the \oiii/\hb\ emission line ratio, and the
124: anti-correlation with the \feii/\hb\ ratio (e.g. BG92).
125: NLS1 galaxies are placed at one extreme end of EV1 
126: parameter space. The most common interpretation 
127: is that this regime is goverened by  
128: the highest Eddington accretion rates and/or lowest black hole
129: masses (e.g., BG92; Sulentic et al. 2000). 
130: 
131: Among other parameters%
132: \footnote{See \citet{stefanie06b} for a recent discussion. 
133: Briefly speaking,  other parameters considered to
134: be, perhaps, relevant in explaining NLS1 properties are
135: orientation (e.g., Osterbrock \& Pogge 1985; Puchnarewicz et al. 1992;
136: Collin et al. 2006; Zhang \& Wang 2006), the 
137: physics behind NLS1's radio properties (e.g. Komossa et al. 2006b),
138: metallicity (e.g., Mathur 2000; Komossa \& Mathur 2001; Nagao et al. 2002; 
139: Shemmer \& Netzer 2002; Romano et al. 2004; Fields et al. 2005), 
140: (ionized) absorption (e.g., Komossa \& Meerschweinchen 2000; 
141: Gierlinski \& Done 2004), 
142: and the galaxies' location on the $M_{\rm BH}-\sigma$ plane 
143: (e.g. Mathur et al. 2001; Wang \& Lu 2001; Grupe \& Mathur 2004;
144: Botte et al. 2005; Mathur \& Grupe 2005; Komossa \& Xu 2007)},
145: the density of an outflowing wind was 
146: firstly speculated to be a prominent driver of EV1 by 
147: Lawrence et al. (1997), given 
148: the connection of \feii\ strength with the presence of 
149: low-ionization, blueshifted broad absorption lines, 
150: and with blue-asymmetric emission lines. 
151: %    
152: Winds and outflows play a crucial role in 
153: understanding the physics and evolution of AGN
154: (e.g., Elvis 2000, Hopkins et al. 2005),
155: and there is ample observational evidence for 
156: winds and outflows in AGN from
157: sub-kpc to galactic scale (see, e.g., Sulentic, Marziani \& Dultzin-Hacyan 2000;
158: Veilleux, Cecil, \& Bland-Hawthorn 2005, for recent reviews).
159: In enriching the nuclear environment with matter from the central region, 
160: winds may have an important impact on the 
161: gas densities in the emission-line regions. 
162: 
163: Regarding NLS1 galaxies, their high ratios of \LLedd\
164:  are likely particularly efficient in driving outflows. 
165: The large \oiii\ $\lambda$5007 blueshifts observed in some 
166: NLS1 galaxies are interpreted 
167: straightforwardly as the result of an outflow (e.g., Zamanov et al. 2002; 
168: Aoki et al. 2005, Boroson 2005). 
169: \citet{dewangan01} suggested that stronger outflows would likely push 
170: the broad-line region (BLR) further radially outward thereby resulting 
171: in narrower \hb\ lines
172: in NLS1 galaxies. They further speculated that the observed flux  ratios from 
173: the NLR and BLR can be explained in terms of density enhancements 
174: (see also Wills et al. 2000).
175: Outflows in NLS1 galaxies have been observed in 
176: terms of both blueshifted UV absorption lines 
177: (e.g., Laor et al. 1997a,b; Goodrich 2000) 
178: and UV emission lines (e.g., Leighly \& Moore 2004). 
179: 
180: Regarding the {\em density} of the emission-line regions
181: of NLS1 galaxies, only few previous estimates exist
182: (e.g., Wills et al. 2000; Kuraszkiewicz et al. 2000; Sulentic, Marziani \&
183: Dultzin-Hacyan 2000; Marziani et al. 2001; Bachev et al. 2004). 
184: Those which do exist, actually 
185: lead to partially conflicting results. 
186: While some sample studies \citep{kur00,wil00,mar01,bach04} suggested 
187: a high-density BLR in NLS1 galaxies, based on  
188: the large \ion{Si}{3}]~$\lambda1892$/\ion{C}{3}]~$\lambda1909$ ratios
189: measured from UV spectra
190: \footnote{One of the alternative explanations is 
191: that the carbon has been removed from gas 
192: phase by depletion onto dust grains \citep{cre02}.}, 
193: other inquiries favored low density emission-line clouds. 
194: \citet{rod00a} studied the emission-line properties of a sample of Seyfert\,1 
195: galaxies, including 7 NLS1 galaxies using optical and near-IR spectroscopy.
196: They found that the typical density of the \sii\ emitting zone is lower 
197: in NLS1 galaxies than in BLS1 galaxies.
198: \citet{rod97} tentatively favored a low-density BLR in NLS1 galaxies, inferred
199: from their photoionization modeling of the UV emission lines.
200: \citet{ferland89} derived densities similar to the canonical BLR value 
201: in strong \feii\ emitters, based on the ratio of the forbidden lines 
202: relative to the \caii\ triplet. 
203: \citet{baskin05} reported that the density of the \oiii\ $\lambda$5007 
204: emitting gas in the NLR decreases with steeper observed soft X-ray slope.
205: 
206: 
207: In addition, a few individual objects
208: were inspected more closely \citep{lao97b,lei04,veron04,veron06}.
209: In particular, analyses of the high S/N UV spectrum of the 
210: prototype NLS1 galaxy, I\,Zw\,1, indicated a high BLR 
211: density of 10$^{11}\,$cm$^{-3}$
212: and NLR density of $5\times 10^{5}\,$cm$^{-3}$ \citep{lao97b}. 
213: \citet{veron04} found on the same object 
214: from optical spectroscopy that the density 
215: of the low-ionization part of the NLR is of 
216: the order of 10$^{6-7}\,$cm$^{-3}$.
217: 
218: Given that few systematic measurements exist at all, 
219: and that those which do produce partially
220: conflicting results, it is important to explore this topic further.
221: We present for the first time a study of the NLR density 
222: for one of the largest homogeneously analyzed NLS1 samples to date 
223: and compare it with that of BLS1 galaxies. 
224: The electron density of the NLR can be measured by 
225: making use of the density-sensitive
226: line ratios
227: \oii\ $\lambda3729/\lambda3726$ and \sii\ $\lambda6716/\lambda6731$ (e.g., Osterbrock 1989).
228: In practice, the most important density diagnostic is the \sii\ line
229: ratio. Usually the density can not be directly inferred from the \oii\ line
230: ratio because the \oii\ $\lambda\lambda3726,3729$ doublet is often unresolved.
231: More indirectly, other line ratios  will also change with density.
232: For instance, a higher-density NLR would strongly boost
233: the \oi\ $\lambda$6300 line, thus
234: the intensity of \oi\ can be used to probe higher density regions 
235: (e.g., Komossa \& Schulz 1997; Barth et al. 2001).
236: 
237: This work is part of a series of papers investigating the properties of
238: the emission-line regions of type\,1 AGN, including NLS1 galaxies 
239: and BLS1 galaxies. 
240: In this first paper of the series,
241: we focus on topics related to density. Specifically, we attempt 
242: to answer the following key questions: 
243: (1) is there any difference in the NLR density between NLS1 galaxies 
244: and BLS1 galaxies?
245: (2) If so, do trends in density correlate with other parameters? and
246: (3) what are the key physical drivers to explain differences in the 
247:     NLR of NLS1 galaxies and BLS1 galaxies? 
248: 
249: This paper is organized as follows.
250: We present the data base and the sample selection in Sect.\,2.
251: In Sect.\,3, we describe our method of the optical spectral analysis.
252: The sample classification and an investigation
253: of selection effects is provided in Sect.\,4.
254: The key result, the detection of a '{\em zone of avoidance}' in the NLR density,
255: is reported in Sect.\,5. 
256: In Sect.\,6 we discuss the reality of the zone of avoidance,
257: followed by a discussion on its origin in Sect.\,7. 
258: We summarize our conclusions in Sect.\,8.
259: 
260: We use the terms NLS1 galaxies and BLS1 galaxies collectively for
261: high-luminosity and low-luminosity objects, i.e., NLS1 galaxies
262: for narrow-line type~1 quasars and for narrow-line type~1 Seyfert galaxies,
263: and BLS1 galaxies for broad-line type~1 quasars and for
264: broad-line type~1 Seyfert galaxies, respectively.
265: Throughout this paper, a cosmology with $H_{\rm 0}=70$\,\kms\,Mpc$^{-1}$,
266: $\Omega_{\rm M}=0.3$ and $\Omega_{\rm \Lambda}=0.7$ is adopted.
267: 
268: \section{Sample selection}
269: The combination of X-ray and optical observations has
270: proven to be an efficient way in detecting NLS1 galaxies over the last decades
271: (e.g., see Pogge 2000 and V\'{e}ron-Cetty,  V\'{e}ron \& Gon\c{c}alves 2001,
272: hereafter VVG01, for reviews).
273: The uniform optical spectroscopic galaxy survey known as the Sloan Digital Sky Survey 
274: (SDSS, York et al. 2000) is an excellent data base which enables us to study
275: AGN properties and, in particular, optical emission
276: line properties in a homogeneous way.
277: 
278: In this paper, in order to measure NLR densities,
279: we homogeneously analyze and compare optical emission line 
280: properties of a large number of NLS1 galaxies with BLS1 galaxies based on spectra obtained
281: in the course of the SDSS. We use the 3rd data release, DR3 \citep{sloandr3}.
282: The DR3 spectroscopic program covers an area of about 4188 deg$^2$. 
283: The two double spectrographs 
284: produce data covering a  wavelength range 3800--9200\AA\, 
285: at a spectral resolution $\approx 2000$. 
286: Exposure times for spectroscopy are determined such that a signal-to-noise (S/N)
287: of at least 4 pixel$^{-1}$ at $g=20.2$ is reached. 
288: The SDSS spectroscopic pipeline 
289: (Stoughton et al. 2002) 
290: performs spectral extraction, sky subtraction, removal of the atmospheric 
291: absorption bands, wavelength and flux calibration, and estimates the error 
292: spectrum. The processed DR3 spectra have not been corrected for Galactic 
293: extinction, but spectrophotometric calibration has been considerably improved 
294: since the First Data Release (DR1). We refer the reader to \citet{sloandr3}
295: for details of the changes.
296:   
297: \subsection{The NLS1 sample}
298: First, we extracted all NLS1 galaxies included in the the 11th 
299: edition of the "Catalogue of Quasars and AGN" (V\'{e}ron-Cetty \& V\'{e}ron 
300: 2003; VV03 hereafter.). The defining criterion of NLS1 galaxies in VV03 is 
301: FWHM(\hb) $\le$ 2000 \kms. We further take into account only NLS1 galaxies 
302: with redshift $z$ less than 0.3. The redshift cut is imposed to ensure 
303: that \sii\ $\lambda\lambda$6716,6731 is observable, 
304: and furthermore, to ensure that the doublet is in a region 
305: free of noise from strong 
306: night-sky emission lines. The list of 309 selected NLS1 galaxies was then 
307: cross-correlated with the SDSS DR3, in order to obtain a homogeneous 
308: set of spectra of a large sample of NLS1 galaxies for 
309: uniform spectral analyses. 
310: This procedure resulted in the selection of 119 sources. 
311: In order to get an accurate density measurement, we 
312: require \sii\ $\lambda\lambda$6716,6731 to have S/N 
313: greater than 5. We also removed a handful of objects with spectral defects 
314: (e.g. problems with sky subtraction, or missing signal over a range of 
315: wavelengths; see Strateva et al. 2003) by visual inspection. 
316: This leaves a total of 58 objects.
317: 
318: \subsection{The BLS1 sample}
319: Recently, Boroson (2003) presented a sample of 107 low-redshift 
320: (redshift $z<0.5$) type~1 AGN from the SDSS Early Data Release (EDR, Stoughton 
321: et al. 2002)%
322: \footnote{We do not divide the Seyfert 1 class into subclasses such as 
323: Seyfert 1.5-1.9 (as defined by \citet{ost85}), since the definitions of 
324: Seyfert classifications depend on the resolution of the spectra used and 
325: the noise in the spectra \citep{gru99,good89}.}.  
326: 82 out of these objects match our redshift constraint, i.e., $z<0.3$.
327: We use this sample to build the control BLS1 sample imposing the same 
328: redshift cut and S/N limit of \sii\ $\lambda\lambda$6716,6731 (i.e., $z<0.3$ 
329: and S/N $>5$) as to the NLS1 sample.   
330: The processed DR3 spectra were downloaded for spectral inspection in 
331: accordance with the data set of the NLS1 sample.
332: With these restrictions, 48 objects survive, 
333: 13 of which overlap with the NLS1 sample. 
334: 
335: \section{Spectral Analysis}
336: The optical emission-line properties (line widths and line ratios) are among
337: the defining criteria of the NLS1 phenomenon. The exact and homogeneous 
338: measurements of the emission-line parameters of AGN provide us with basic
339: knowledge about emission-line regions, and enable us to 
340: investigate the relationship between NLS1 galaxies and BLS1 galaxies. 
341: However, the
342: published data were compiled in a heterogeneous way. The use of different line
343: profiles (e.g., Lorentzian vs Gaussian profiles) leads to strong differences in
344: the \hb\ widths (e.g., VVG01).
345: In addition to the Lorentzian and Gaussian representations of line profiles,
346: a direct measurement of the width at half of their maximum intensity
347: is sometimes adopted 
348: (e.g., Williams, Pogge \& Mathur 2002; Boroson et al. 2003).
349: Also, the strong \feii\ contamination makes it difficult to
350: measure \hb\ and brings large uncertainties in determining the \hb\
351: line width. 
352: 
353: Here we perform a homogeneous spectral analysis for the NLS1 and BLS1 samples. 
354: The objects are then re-grouped into the NLS1 or BLS1 sample for further 
355: investigation, according to the widths of the broad \hb\ line (Sect.\,3.2) 
356: obtained in our spectral analysis. 
357: 
358: In a first step, the SDSS spectra were corrected for 
359: the Galactic extinction using the reddening map 
360: of Schlegel et al. (1998), 
361: and then shifted to their rest wavelength, adopting the SDSS value 
362: of the redshift from the header of each spectrum.
363: Once these steps were completed, we removed the stellar continuum, 
364: subtracted the \feii\ complexes and performed the spectral analysis for the 
365: emission lines following the procedures described below.
366: 
367: \subsection{Subtraction of starlight and nuclear continuum}
368: SDSS spectra are acquired with a pair of fiber-fed spectrographs. 
369: Each fiber subtends a diameter of 3$''$, corresponding
370: to $\sim 6.5$~kpc at $z=0.1$. This aperture is large enough
371: to let through not only the emission from the nucleus, but also
372: a substantial amount of starlight from the host galaxy (e.g., Vanden
373: Berk et al. 2001, Hao et al. 2005). The accurate removal of the stellar
374: contribution is essential to reliably measure the emission-line
375: spectrum, such as the line widths and line strengths. Particularly,
376: the reliable classification of NLS1 galaxies is strongly dependent on the
377: measurement accuracy of the width of the broad \hb\ line. Furthermore, in
378: many of the spectra there is a clear contribution from blends of
379: \feii\ line emission. Well-studied NLS1 galaxies usually show strong optical
380: \feii\ emission features on both the blue and red sides of the
381: \hb\--\oiii\ complex (e.g., V\'{e}ron-Cetty \& V\'{e}ron 2003). 
382: In order to reliably measure line
383: parameters, we choose those wavelength ranges as pseudo-continuum,
384: which are not affected by prominent emission lines, and 
385: then decompose the spectra into the following 4 components (see
386: Zhou et al. 2006 for details):
387: \begin{itemize}
388:     \item A starlight component modeled by 6 synthesized galaxy
389:     templates, which were built from the synthetic spectral library of Bruzual
390:     \& Charlot (2003) using the algorithm of Ensemble Learning for
391:     Independent Component Analysis (EL-ICA, Lu et al. 2006).
392:     These templates were broadened by convolution with a Gaussian to match the
393:     stellar velocity dispersion of the host galaxy.
394:    \item A power-law continuum to describe the emission from the active nucleus.
395:     \item An \feii\ template obtained by V\'{e}ron-Cetty \& V\'{e}ron (2003).
396:     This template covers the wavelengths between 3535$-$7534 \AA,
397:     extending further to both the blue and red wavelength ranges than
398:     the \feii\ template used in BG92. This makes it more advantageous
399:     in modeling the \feii\ emission in the SDSS spectra. We assume
400:     that \feii\ has the same profile as the broad component of \hb\
401:     (see the next subsection).
402:    \item A Balmer continuum generated in the same way as Dietrich et al. (2003).
403: \end{itemize}
404: 
405: The modeling is performed by minimizing the reduced $\chi^2$ in the
406: fitting process. The final multi-component fit is then subtracted
407: from the observed spectrum.
408: An example of the residual spectrum is plotted in Fig.\,1.
409: 
410: \subsection{Decomposition techniques}
411: The multicomponent-subtracted spectra are used to measure the 
412: non-\feii\ line properties. 
413: The broad Balmer lines in AGN exhibit a wide variety of profile shapes and a 
414: large range in width \citep{ost82,de85,cre86,str91,mil92,veron01}, and they 
415: are often strongly asymmetric \citep{cor95}. In many cases the Balmer lines
416: are mixtures of broad and narrow components. Differences in the relative 
417: strengths of these components account for much of the diversity of broad 
418: line profiles \citep{fra92,wil93,bro94,cor95,cor97,veron01}.
419: A proper decomposition of the NLR and BLR line emission contribution of 
420: the Balmer lines is of great importance to address the 
421: physical properties of the emission-line regions. 
422: Particularly, the width of the broad \hb\ profile has significant 
423: impact on the reliable classification of a galaxy as a NLS1 galaxy. 
424: 
425: In order to measure the parameters of the BLR emission lines, 
426: the NLR line emission
427: contribution has to be removed first. Using homogeneous sets of spectra, 
428: previous studies (e.g., Filippenko \& Sargent 1988; Ho et al. 1997; 
429: Greene \& Ho 2004; Zhou et al. 2006) have shown that the narrow Balmer emission
430: profiles are well matched to those of \sii\ $\lambda\lambda$6716, 6731. 
431: Moreover, the widths of \sii\ and \nii\ doublets trace the stellar velocity 
432: dispersion of galaxy bulges better than that of \oiii\ within the 
433: uncertainties (Greene \& Ho 2004). Therefore, we use
434: the strong profiles of \sii\ $\lambda\lambda$6716, 6731 (those with S/N $>$5) 
435: as a NLR template profile for the narrow component of the Balmer lines% 
436: \footnote{Occasionally, we also make use
437: of spectra with weak or absent \sii\ emission. These are
438: used to measure \oi\ (see Sect. 5.3). 
439: In such a case, the \oiii\ profile is used as a substitute for \sii.}.
440: 
441: To characterize the NLR emission-line profiles, we fit these lines using
442: Gaussian profiles. Most \sii\ $\lambda\lambda$6716,6731 and 
443: \nii\ $\lambda\lambda$6548,6583 lines can be well fit employing one single
444: Gaussian profile% 
445: \footnote{ Using spatially-resolved {\em HST} spectra, Rice et al. (2006) 
446: identified both blue and red asymmetries in some [S II] line profiles, 
447: which are primarily due to nuclear line-emitting gas, rather than more 
448: symmetric emission from the NLR on larger scales.}.
449: A large fraction of the \oiii\ $\lambda\lambda$4959,5007 lines show strongly 
450: asymmetric profiles. In those cases, a second component is then added 
451: to represent the line wings. When we fit \sii\ $\lambda\lambda$6716,6731,
452: \nii\ $\lambda\lambda$6548,6583 and \oiii\ $\lambda\lambda$4959,5007 lines, 
453: the separation of the lines of each doublet is fixed to the laboratory value. 
454: Each pair is assumed to have the same profile%
455: \footnote{The line width of each component of the doublet may be slightly 
456: different due
457: to stratification in the NLR. However, Rice et al. (2006) found the widths 
458: of the two lines differ by less than 3\%, which is within the error of 
459: the Gaussian fit parameters and hard to discern with data of moderate S/N.}. 
460: While the flux ratios of the \nii\ and \oiii\ doublets 
461: are fixed to the theoretical ratio of $3:1$, the intensity ratio of the \sii\ 
462: doublet is measured, and then used to derive the density.
463: 
464: We used the IRAF package SPECFIT \citep{kriss94}
465: to measure blended lines and separate the NLR from the BLR emission. 
466: The actual profile shape of the broad Balmer lines of 
467: NLS1 galaxies is still an issue debated. While the broad Balmer line component 
468: can be well represented by 
469: Gaussian profiles, particularly by a 
470: combination of multiple Gaussian components
471: \citep{rod00b,nagao02,xu03,greene05a,greene05b,dietrich05}, successful 
472: fitting can also be accomplished with a single Lorentzian profile 
473: \citep{lei99,veron01} for some NLS1 galaxies.  
474: \citet{veron01} suggested that the broad Balmer lines of 
475: NLS1 galaxies were better fitted with a single Lorentzian profile than 
476: a single Gaussian profile (see also Sulentic et al. 2002), since many NLS1 
477: galaxies were located in the H\,II region in their diagnostic diagrams if a 
478: single Gaussian profile was adopted. However, as noted in \citet{eva88}, 
479: the choice of Gaussian or Lorentzian profiles as representatives of the 
480: observed emission lines may bear no physical meaning. Furthermore, the 
481: broad-line profiles show complex shapes and asymmetries that can not be 
482: described with a single component, indicating the presence of two 
483: or multiple components, 
484: independent from the type of profile used to fit \citep{dietrich05}. 
485: 
486: In order to isolate the narrow and broad components of the Balmer lines, 
487: we fix the width of the narrow component (determined from the width of \sii\
488: as described above) and only leave its strength as fit parameter.  
489: The broad part of the line profiles we fit by using a combination of 
490: two Gaussian profiles, as well as a single Lorentzian profile. 
491: For the approach with a multi-Gaussian components fit, we measure the FWHM of 
492: the profile and its integrated flux from the final combined model for 
493: the broad component. The individual Gaussian components have no physical 
494: significance by themselves but are only used to serve as a description of 
495: the complex line shapes of the broad components as far as the data 
496: quality allows. 
497: While comparable and equally reasonable results can be achieved with both
498: single Lorentzian and two Gaussian profiles for most broad components
499: of the Balmer lines of NLS1 galaxies, 
500: generally no acceptable fit is possible when 
501: employing the Lorentzian profiles to fit BLS1 galaxies. 
502: We will use the results of the multiple Gaussian fit for NLS1 galaxies 
503: for classification and further correlation analysis, for its simplicity, 
504: and furthermore, for a direct comparison with the BLS1 control sample and  
505: with previous studies (e.g. BG92; Grupe et al. 1999; Vaughan et al. 2001).
506: However, for comparison purposes, we still report results from
507: Lorentzian fitting in our key figures (Fig. 5, 6). 
508: We further checked the location of NLS1 galaxies and BLS1 galaxies in the 
509: diagnostic diagrams using the standard  
510: emission-line ratios. 
511: In contrast to Fig.\,8 of VVG01, 
512: most of our NLS1 galaxies locate in the AGN regime
513: for both Gaussian and Lorentzian fits. 
514: 
515: The broad component of \hb\ is referred as \hbb, and the narrow 
516: component as ${\rm H\beta_n}$. 
517: The dominant uncertainties in the line parameter measurements 
518: can result from the continuum subtraction and component decomposition. 
519: The latter depends on the line profile shape.
520: Although the line measurements of Gaussian and 
521: Lorentzian fits are precise, they
522: are only true estimates if the lines can be correctly represented by the 
523: profiles. The average error in the width of narrow lines is of approximately 
524: 5\%. Errors in the width of broad lines, introduced by the
525: different profile types, e.g., two Gaussians or a single Lorentzian, 
526: are about 10\%. For most objects, the uncertainties of flux measurements 
527: of emission lines are less than 10\%. 
528: The typical measurement error of the \sii\ doublet is about 5\% for our sample.
529: 
530: 
531: \section{Sample re-classification and selection effects}
532: 
533: \subsection{Re-classification of the sample} 
534: In order to follow the standard practice to distinguish between 
535: NLS1 galaxies and BLS1 galaxies by FWHM(\hbb), we re-classified the
536: objects of our sample accordingly, after having carefully measured
537: FWHM(\hbb).
538: A total of 55 objects with
539: FWHM(\hbb) $\le 2000$ \kms are included in the final NLS1 sample, 
540: while 39 with FWHM(\hbb) $>$ 2000 \kms\ are included in the BLS1 sample.
541: 
542: \subsection{Luminosity and redshift distributions}
543: As described in Sect.\,2, we do not set any selection criteria upon our 
544: samples, which combines data from various different sources and thus
545: is not statistically complete. It is therefore necessary to examine 
546: whether or not the NLS1 and BLS1 samples are drawn from comparative 
547: populations such that 
548: luminosity effects (in particular, the number of quasars vs Seyferts in each
549: sample) and evolutionary effects can be excluded.
550: In order to check the possible biases between the two samples, we look into
551: the absolute magnitude distribution and the redshift distribution.
552: 
553: First, we investigate whether or not there is a systematic difference in the 
554: luminosity distribution between the two samples. Here we trace the luminosity 
555: with the absolute $i$\,magnitude, which is calculated from the $i$\_psf 
556: magnitude (e.g. Schneider et al. 2005; Vanden Berk et al. 2004), 
557: by correcting the $i$\,measurements for Galactic extinction 
558: (Schlegel et al. 1998) and assuming 
559: a powerlaw spectral energy distribution (SED)
560: ($f_v \propto \nu^\alpha)$, where $\alpha=-0.5$ (e.g., 
561: Vanden Berk et al. 2001). 
562: Following the previous studies of the SDSS AGN (Schneider et al. 2003, 2005; 
563: Nagao et al. 2005), we define objects that have luminosities larger than
564: M$_i=-22.0$ as quasars. 
565: We show the histograms of the absolute $i$\,magnitudes 
566: for the NLS1 galaxies and BLS1 galaxies
567: in Fig.\,3. We apply the Kolmogorov-Smirnov (K-S) statistical test to check 
568: how closely the distributions resemble each other.  
569: The resultant K-S test probability is 0.33, which means 
570: the two distributions are statistically indistinguishable.
571: 
572: Secondly, we explore the redshift distribution. The histograms of the 
573: sources' redshifts are shown in Fig.\,3. The NLS1 and BLS1 samples 
574: possess similar average 
575: redshifts. The value and 1$\sigma$ deviations for 
576: the two samples are $0.128 \pm 0.068$ and $0.142 \pm 0.066$, respectively.
577: Moreover, since both samples are compiled from low-redshift objects with 
578: $z<0.3$, no strong redshift bias is expected, anyway.
579: We conclude that our two samples do not show significantly different
580: luminosity and redshift distributions.
581: 
582: \subsection{Optical \feii\ emission versus \oiii\ emission}
583: In this section we wish to demonstrate that our (NL)S1 sample follows
584: the same correlations found previously for (NL)S1s as a class, and 
585: therefore, that our NLS1 galaxies are representative for the NLS1 population as 
586: a whole, as far as previously known correlations are concerned.
587: Among optical emission lines of NLS1 galaxies and BLS1 galaxies, 
588: the most striking correlation is the so called EV1 
589: (e.g., BG92; Grupe et al. 1999; Sulentic et al. 2002, 2003; Xu et al. 2003), 
590: i.e., that \oiii/\hb\ emission 
591: is weak in objects with strong \feii\ emission and vice versa. NLS1 galaxies 
592: are at the extreme end of this correlation, often 
593: showing strong \feii\ emission and weak \oiii/\hb$_{\rm total}$ emission. 
594: 
595: The optical \feii\ strength, R4570, is defined as the ratio of the flux of 
596: the \feii\ complex between the rest wavelength 4434\AA\, and 4684\AA\, to 
597: the total \hb\ flux, including the narrow component (e.g. BG92, VVG01).  
598: The average values of R4570 for our NLS1 and BLS1 samples are 0.75 and 0.30, 
599: respectively. The regime of high values of R4570 is solely 
600: occupied by NLS1 galaxies. 
601: The results are consistent with previous studies (e.g., Rice 
602: et al. 2006 and references therein). 
603: In Fig.\,4, we plot R5007, the ratio of \oiii$\lambda5007$ flux to the total 
604: \hb\ flux, as a function of R4570. Both emission lines span almost the 
605: whole range from the low to high end observed in 
606: NLS1 galaxies and BLS1 galaxies (e.g., 
607: Fig.\,11 in VVG01), suggesting there is no preferential selection effect 
608: to objects with extreme properties. Moreover, an anti-correlation between 
609: the two parameters is present, with a correlation coefficient 
610: r$_{\rm s}=-0.37$. The probability of the null correlation is
611: P$_{\rm null}<10^{-3}$. 
612: These quantities are calculated from the Spearman rank-order 
613: correlation analysis. Hence, 
614: the NLS1 and BLS1 samples under study can be considered as representative
615: samples of these two subclasses, though they are not statistically complete.
616: 
617: \section{A zone of avoidance in the NLR density}
618: \subsection{Measurement of density}
619: We use the {\em density} diagnostic \sii\ $\lambda6716$/$\lambda6731$ 
620: to measure the NLR electron density.
621: The electron density is calculated following \citet{ost89} using 
622: the task {\em temden} of the IRAF package STSDAS \citep{shaw94}. 
623: This approach comes with two limitations.
624: Firstly, around densities of 10$^3$--10$^4$\,\cm\ we approach the critical densities
625: of \sii\,$\lambda6716$~($1.5 \times 10^3$\,\cm) 
626: and \sii\,$\lambda6731$~($3.9 \times 10^3$\,\cm).
627: Secondly, the \sii\ line ratio also depends on temperature. However,
628: the dependence is only weak in the range of temperatures of the regions studied.
629: We fix the electron temperature at \te\,$=10,000$K, typical for photoionized 
630: gas in the NLR, and comment on possible temperature effects later.
631: The \sii-based density measurement allows us to probe the \sii-emitting
632: part of the NLR. In order to get a first impression also on the 
633: high-density part of the NLR, later we make use of the strength of
634: the \oi\ line (Sect. 5.3){\footnote{We note that the NLR is most likely
635: composed of  
636: emission-line clouds with a range in densities 
637: (e.g., Komossa \& Schulz 1997; Peterson 1997; 
638: Rodriguez-Ardila, Pastoriza \& Donzelli 2000a; Brinkman et al. 2000),
639: and that NLR density increases  
640: toward small radii (e.g., 
641: Fraquelli, Storchi-Bergmann \& Binette 2000; Bennert et al. 2006a,b).}}. 
642: 
643: The [SII] ratio is only a good density diagnostic in a certain range of 
644: densities because depopulation of the upper levels changes from being primarily 
645: radiative at low densities to primarily collisional at high densities. 
646: The {\rm low-density} limit is $\sim 10$\,\cm,
647: while for densities higher than $10^{4}$\,\cm, collisional de-excitation
648: becomes more and more important. Most of our sources distribute in a regime 
649: where the methods are well applicable.
650: In the {\rm low-density} limit (\n\,$< 10$\,\cm), the error
651: in density is larger due to the saturation of the relation 
652: between the line ratio and electron density, so these derived 
653: densities should be treated with caution. 
654: An \sii\ ratio of 1.42 corresponds to a density of $\sim 10$\,\cm.
655: Only two objects, SDSS J011448.68-002946.1 and SDSS J161809.38+361957.8, 
656: have ratios greater than 1.42, which places the sources at the 
657: low end in the density distribution.
658: 
659: \paragraph{Notes on individual objects.}
660: Among the whole sample, only objects which are above a certain S/N ratio
661: in the \sii\ line 
662: were kept for further analysis (Sect.\,2). Among these, we then individually
663: re-inspect all objects below a density threshold of 
664: \n\,$=140$\,\cm (log\,\n\,$=2.15$\,\cm, 
665: i.e., those in regime "A" in Fig.\,5) 
666: to check for peculiarities in 
667: spectral features, and to check the robustness of spectral fitting and 
668: thus reliability of the density estimate in the 
669: {\em low-density} regime.  
670: 
671: We comment here on those objects 
672: which are extreme in the \sii\ ratio, FWHM or other spectral features: 
673: (1) There is one single broad-line object actually located in the 
674: '{\em zone of avoidance}' in density. This is SDSS J011448.68-002946.1. While
675: its \sii\ ratio is extreme (R(\sii)$=1.47 \pm 0.03$), and within the 
676: errors beyond
677: the '{\em low-density}' regime, its other emission lines are 
678: not unusual. The profile of \sii\ $\lambda6731$ deviates somewhat from
679: a Gaussian, but there is nothing else very peculiar about it. 
680: This is the only outlier in Fig.\,5. 
681: (2) The low-density NLS1 SDSS J092247.03+512038.0 is peculiar in that 
682: the peak of the \oiii\ $\lambda5007$ line is blueshifted with respect to the 
683: low-ionization forbidden lines (e.g., \oii, \nii\ and \sii) and the narrow
684: component of \hb\ by more than 400~\kms. The large velocity shift places 
685: it among the \oiii\ blue outliers \citep{zamanov02}.
686: Moreover, \oiii\ $\lambda5007$ is almost as broad as \hbb\ 
687: (FWHM 1060 \kms\ and 1250 \kms\ for \oiii\ $\lambda5007$ and \hbb,
688: respectively.), while the low-ionization forbidden lines such as \sii\
689: are narrow (FWHM(\sii) $=220$~\kms). It is a strong \feii\ emitter
690: with ${\rm R4570}=1.32$. 
691: (3) SDSS J161809.38+361957.8 is the object with the most extreme
692: combination of \hbb\ line width 
693: (FWHM(\hbb) $=700$~\kms\ for a Lorentzian profile fit) and density 
694: (\n\,$=2.0$\,\cm). 
695: The emission lines have a very wide range of excitation, the highest
696: corresponding to coronal lines of \fex\ $\lambda6374$ and \fexi\ $\lambda7892$.
697: We also detect the broad low-ionization line \peroi\,$\lambda8446$, which 
698: is generally produced in a region with very high density 
699: (e.g. Komossa \& Bade 1999).  
700: It is the only low-density source of our sample which has 
701: the \peroi\,$\lambda8446$ line detected.         
702: (4) The low-density NLS1 SDSS J120226.76-012915.3 shows
703: exceptionally strong \feii\ emission of ${\rm R4570}=2.74$.
704: It is the most extreme \feii\ emitter in our sample.
705: (5) Finally, we mention that among the high-density sources,
706: the spectrum of 
707: the BLS1 SDSS J013527.85-004448.0 
708: is special in that it shows the highest ratio of \oii/\oiii\ of 
709: the whole sample (\oii/\oiii$_{\rm obs} = 1.6$) 
710: {\footnote{More detailed account on multi-wavelength properties
711: and on (unusual) line profiles and line ratios of individual sources 
712: will be given in a follow-up paper (Xu et al. 2007, in prep.)}}.  
713: 
714: \subsection{Density versus FWHM(\hbb): a zone of avoidance}
715: One of our main goals is to examine whether or not there is a difference in
716: electron density \n\ 
717: between NLS1 galaxies and BLS1 galaxies, in order to test different NLS1 models.
718: The \sii\ $\lambda6716/\lambda6731$ line ratio in our sample ranges
719: from 0.87 to 1.47.
720: We plot the ratio versus the FWHM of \hbb\ in Fig.\,5.
721: The typical error on the ratio is $\sim$0.06.
722: The histograms of the ratios for the NLS1 and BLS1 samples are also shown.
723: NLS1 galaxies show 
724: \sii\ ratios in the range from 0.94 to 1.43. 
725: 17 out of 55 NLS1 galaxies have a ratio higher than 1.28,
726: while only one out of 39 BLS1 galaxies has a high ratio (i.e., 1.47). 
727: The other 38 BLS1 galaxies occupy the range from 0.87 to 1.27.
728: The average ratios and 1$\sigma$ deviations of the two samples are
729: $1.23 \pm 0.12$ and $1.12 \pm 0.10$, respectively.
730: 
731: In Fig.\,5 (lower panel), we display the inferred electron density 
732: against the FWHM of \hbb. We find that the sources do not homogeneously 
733: populate the \n\,--FWHM(\hbb) diagram.
734: The key detection is a '{\em zone of avoidance}' in the diagram.
735: While the 38 BLS1 galaxies avoid low average densities, and all show 
736: \n\,$>140$\,\cm\ (regime\,C), 
737: NLS1 galaxies show a larger scatter in density 
738: in the range \n\,$=2 \sim 770$\,\cm, 
739: including a significant number of objects with low densities. 
740: 17 out of the 55 NLS1 galaxies under study show 
741: \n\,$<140$\,\cm\ (regime\,A) and are 
742: clearly separated from the range occupied by BLS1 galaxies. 
743: The other 38 NLS1 galaxies overlap well with the range in density 
744: for BLS1 galaxies (regime\,B)% 
745: \footnote{Results are robust, no matter whether a Gaussian or Lorentzian 
746: profile is used for the broad component of \hb. 
747: A larger scatter, rather than a strict cut-off in density 
748: for BLS1 galaxies might  
749: appear if the sample size increases significantly, 
750: but we expect our findings do still hold on average.}. 
751: The average electron densities for zone A, B and C are 69, 294 and 380\,\cm, 
752: respectively. 
753: We apply the K-S statistical test 
754: on the distributions of density for NLS1 galaxies vs BLS1 galaxies.
755: The resultant K-S test probability is 0.0002, which means
756: that the two density distributions are significantly different.
757: 
758: \subsection{\oi\ emission versus FWHM(\hbb)}
759: In order to probe also the high-density NLR regime, well above the critical 
760: densities of the two \sii\ lines, we concentrate on the emission line \oi\ $\lambda6300$. 
761: Even though the ratio \oi/\ha\ is also influenced by other parameters, 
762: \oi\ is strongly boosted for high densities (e.g., Fig.\,4 and 8 of 
763: Komossa \& Schulz 1997) and we thus checked for any trends
764: and correlations between \oi\ intensity and FWHM(\hbb)
765: (Fig.\,6), which we regard as a supplement to  the \n\,--\,FWHM(\hbb) 
766: diagram (Fig.\,5). 
767: In Fig.\,6, we plot the ratio \oi\ $\lambda6300$/\han\
768:  as a function of the FWHM(\hbb).
769: An anti-correlation between the two parameters is found with a
770: correlation coefficient r$_{\rm s}=-0.34$, and a probability of the null
771: correlation P$_{\rm null}<10^{-2}$, calculated from the Spearman
772: rank-order correlation analysis. The trend meets our expectation 
773: from Fig.\,5, in the sense that
774: BLS1 galaxies show, on average, higher \oi/H$\alpha$ than NLS1 galaxies. 
775: 
776: \section{Reality of the zone of avoidance}
777: The sources of our sample do not homogeneously populate the \n\,--\,FWHM(\hbb) 
778: diagram, but show a '{\em zone of avoidance}', in the sense that 
779: BLS1 galaxies lack low average NLR densities. 
780: Is there any data analysis or selection effect that could cause a spurious 
781: {\em zone of avoidance} in the BLS1 galaxies regime, or could mimic a larger 
782: density scatter in the NLS1 galaxies regime? We discuss and reject several 
783: possibilities in turn.
784: 
785: \paragraph{Magnitude distribution.}  
786: Firstly, we note that the magnitude distributions of the NLS1 and the BLS1 
787: sample are similar (also in dependence of redshift; Fig.\,3).
788: We note that both extremes of the distribution in the \n\,--\,FWHM(\hbb)
789: diagram include both quasars and Seyferts (but more Seyferts, 
790: since the total number of Seyferts is higher than the quasars).
791: 
792: \paragraph{Profile shape.}
793: Secondly, our results are robust, independent of the profile the
794: broad emission lines are fit with, either a Lorentzian profile or a 
795: combination of two Gaussian profiles (see Fig.\,5). 
796: 
797: \paragraph{Atmospheric absorption effects.} 
798: We then checked whether the density estimate for the small-FWHM regime 
799: is more unreliable (thus, a larger scatter) because at a certain redshift 
800: range ($z=$\,0.130--0.140), the atmospheric O$_2$ absorption line at 7620\AA\
801: overlaps with one or the other sulfur line, making line estimates 
802: more unreliable (even though atmospheric absorption is generally already 
803: corrected for the released SDSS spectra).%
804: \footnote{The other atmospheric O$_2$ absorption line at 6870\AA\ will not 
805: overlap the sulfur lines, given the redshift range 0.034--0.289 of the
806: objects in our sample.} 
807: In that case, sources in the regime with the larger range in densities 
808: (zones A \& B in Fig.\,5) should have a specific, narrow redshift range. 
809: This is not the case, however. The low density objects vary in redshift 
810: between $z=$\,0.034--0.280, while the rest of the sample 
811: shows $z=$\,0.038--0.289.  
812: 
813: \paragraph{Signal/Noise.}
814: In order to get an accurate \sii\ line ratio and thus density measurement,
815: we required \sii\ $\lambda\lambda$6716,6731 to have S/N greater than 5
816: which is 
817: quite a common cut-off imposed on SDSS data.
818: 90\% of our spectra are actually of S/N (\sii) $\ge$ 10. 
819: In order to see whether the S/N of our spectra affects in
820: any kind of way our measurement of the [SII] ratio, we
821: have checked whether one class of objects (those with low density) or the other class
822: (with high density) shows an excess of sources 
823: with low S/N. 
824: However, this is not the case. 
825: We have run a Spearman rank-order correlation analysis on 
826: the distribution of the \sii\ ratio in dependence of S/N, 
827: and we do not find any correlation (r$_{\rm s}=-0.05$, P$_{\rm null}=0.63$).
828: 
829: \paragraph{Faint broad wings in the Balmer lines.}
830: In case of very broad Balmer lines, if the broad wings of the line are over- or
831: underestimated, then a corresponding small under- or over-estimate in the \sii\ lines
832: might result. The \sii\ $\lambda$6716 line might then be slightly more strongly 
833: affected, and that might marginally change the measured \sii\ ratio.  However, in 
834: most of our sources the [SII] lines are well separated from faint broad wings (see Fig.\,2)
835: 
836: \paragraph{High and low density limits.}
837: \sii\ $\lambda$6716/$\lambda$6731 is only a good density 
838: diagnostic in a certain 
839: range of densities. The low-density limit is ${\rm \sim 10}$\,\cm,
840: while for densities higher than $10^{4}$\,\cm, collisional de-excitation 
841: becomes more and more important. Most of our sources distribute in a regime 
842: where the methods are applicable. Since the conversion of the \sii\ ratio 
843: into density is non-linear, the estimate of density in the low-density regime 
844: (Fig.\,5.3 of Osterbrock 1985) comes with larger errors. These are shown in 
845: Fig.\,5. Within these errors, objects with small FWHM(\hbb) still have, 
846: on average, lower densities. 
847: 
848: The whole diagram (i.e., across the whole FWHM(\hbb) range), may, to a minor extent, 
849: be biased toward excluding very high densities by the original selection effect 
850: of using only spectra with significant S/N in the \sii\ lines. This 
851: pre-selection cannot be avoided. 
852: However, since \sii\ is collisionally de-excited for densities well above 
853: $10^4$\,\cm, the pre-selection criterion then is prone to excluding 
854: objects with extremely high densities, should they exist at all.
855:     
856: \paragraph{Influence of temperature.}
857: The estimate of density using the ratio \sii\ $\lambda$6716/$\lambda$6731 also 
858: depends on temperature. As described in Sect.\,5.1, we fix \te\,$=10000$\,K. 
859: An estimate of the temperature is possible with the temperature-sensitive
860: emission-line ratio \oiii\ $\lambda$4363/\oiii\ $\lambda$5007.
861: However, \oiii\ $\lambda$4363 is a faint line which generally
862: comes with large measurement errors. 
863: We have inspected the highest-quality spectra available in the 
864: \oiii\ $\lambda$4363 wavelength range (high S/N and/or easy deblending of 
865: H$\gamma$ and \oiii); both, among the lowest density object, and, for 
866: comparison, a similar number among the higher density objects. We do not 
867: find systematically higher temperatures for the low-density objects. 
868: We further note that given the scaling of density with temperature, 
869: $n_{\rm e}(T)=n_{\rm e}({\rm obs}) \times  \sqrt{T/10\,000}$ \citep{ost89}, 
870: even an increase
871: in temperature up to 40,000K (close to the average of four Sy1 galaxies
872: of Bennert et al. 2006b), would only change density by a factor of 2, much 
873: less than the scatter in the observed density values. We come back to this 
874: point in Sect.\,7.2. 
875: 
876: \paragraph{Starburst contribution.}
877: If there was a strong systematic starburst/\hii\ contribution to the 
878: emission lines, 
879: then we would expect the density estimates to be biased 
880: toward lower values, more 
881: typical for \hii\ regions.   
882: Since our sample shows unambiguous AGN-like line ratios, 
883: we expect the vast majority of the emission to come from the NLR, 
884: rather than star-forming regions. 
885: We further use the star-forming indicator, \oii\ $\lambda$3727 
886: (e.g, Hippelein et al. 2003; Kewley \& Geller 2004; Ho 2005), 
887: to track the starburst contribution to the optical   
888: emission lines of the AGN of our sample. 
889: \oii\ $\lambda$3727 is prominent in \hii\ regions,  
890: while the ratio \oii/\oiii\ is observed and predicted to be relatively 
891: weaker in Seyfert galaxies.
892: We compare the line ratios \oii\ $\lambda$3727/\oiii\ $\lambda$5007 
893: of NLS1 galaxies and BLS1 galaxies, in order to check whether or not the 
894: \oii\ emission is on average stronger in NLS1 galaxies.
895: In Fig.\,7 we show the distributions of the ratio 
896: prior to reddening correction and after reddening correction%
897: \footnote{For dust reddening correction we use the average reddening curve
898: of Osterbrock (1989, his Table\, 7.2) and an intrinsic value of 
899: \ha$_{\rm n}$/\hbn\,=\,3.1.}. 
900: No difference is present between the \oii/\oiii\ ratio for NLS1 galaxies and BLS1 galaxies. 
901: Thus, we conclude that the starburst contribution
902: is not a possible explanation for the \n\,--\,FWHM(\hbb) distribution.
903: The average observed \oii/\oiii\ ratios for the NLS1 and BLS1 sample are
904: 0.32 and 0.29, while the average reddening corrected ratios are 0.35 and 0.33,
905: respectively.
906: 
907: 
908: \paragraph{Spatially resolved NLRs and viewing angle effects.}
909: Finally, we checked whether nearby objects might be spatially resolved, 
910: such that the low-density outer part of the NLR is missed in the fiber
911: which would then result in an overestimate of the density in
912: these objects.  
913: We find that the fiber diameter of 3$''$ corresponds to 
914: $\sim 2$\,kpc at ${\rm z=0.034}$ (the lowest redshift in our sample), 
915: while the gas in the NLR is typically distributed over a distance 
916: $r \sim 10-1000$\,pc from the nucleus. 
917: Therefore, SDSS fibers would always cover the entire NLR.
918: 
919: \section{On the origin of the zone of avoidance}
920: 
921: The key result of this study is the detection of a {\em zone of avoidance} in 
922: the density - FWHM(\hbb) diagram (Fig.\,5): BLS1 galaxies (FWHM(\hbb) $>$ 2000~\kms) 
923: avoid low average densities, and all show \n\ $>140$\,\cm. 
924: On the other hand, NLS1 galaxies show a larger scatter in densities in the range 
925: \n\,$=2 \sim 770$\,\cm, including a significant number of objects with 
926: low densities. The results obtained for \sii\ are consistent with 
927: the \oi\,--\,FWHM(\hbb) diagram as shown in Fig.\,6, 
928: which shows lower average \oi/H$\alpha$ in NLS1 galaxies.
929: In the following, we first confront our results on density with predictions
930: or indications on density effects of existing NLS1 models, and then discuss
931: further possibilities.
932: 
933: \subsection{Coupling between NLR and BLR?}
934: 
935: Given occasional reports of signs of lower or higher-than-average
936: BLR density of NLS1 galaxies (e.g., Kuraszkiewicz et al. 2000; 
937: Rodriguez-Ardila et al.2000a; Marziani et al. 2001;
938: Komossa \& Mathur 2001; Xu et al. 2003; Bachev et al. 2004)
939: we then also expect the same trend to reflect in the NLR densities
940: -- {\em if} the properties of NLR and BLR are closely   
941: linked{\footnote{The exact relation between BLR and NLR in AGN is still
942: under examination. While some models do predict a close link, others
943: assume or indicate that both cloud components are of different and
944: unrelated origin. Models that argue against a common link include indications
945: that NLRs are just normal interstellar medium (ISM) in the host galaxy, 
946: while the BLR has separate
947: origin. Also, the fact that the FWHMs of the Balmer lines of BLR and NLR do not
948: generally correlate (e.g., BG92; Grupe et al. 1999; Vaughan et al. 2001; 
949: Xu et al. 2003) 
950: suggests different kinematic components. 
951: On the other hand, there are models and observations that do predict
952: a link, including common wind outflows (e.g., Schiano et al. 1986); the
953: suggestion that the apparent gap between BLR and NLR is solely caused by
954: dust effects (Netzer and Laor 1993);
955: and observational links between BLR and NLR parameters, in particular,
956: between line strength, line asymmetry and the shift of the
957: line centroids 
958: (e.g., Xu et al. 2003).}}. 
959: Such models can then be tested by measuring the NLR density.
960: We start with a short review of previous measurements,
961: then come back to our new results. 
962: 
963: Few NLS1 galaxies have been studied with respect to the density of
964: their NLR, so far.
965: \citet{rod00a} reported NLR density measurements of seven NLS1 galaxies,
966: based on the \sii~$\lambda6716$/\sii~$\lambda6731$ line ratio.
967: They found lower average density in the \sii\ emitting zone in NLS1 galaxies
968: than in BLS1 galaxies.
969: V\'{e}ron-Cetty et al. (2004) presented a high S/N optical
970: spectrum of the NLS1 galaxy I\,Zw\,1 and concluded that the bulk of the NLR is
971: unlike that of most Sy1 galaxies. It is of unusually low excitation and
972: dominated by lines of high critical density, while lines like 
973: \oiii\ $\lambda$5007
974: and \sii\ $\lambda\lambda$6716,6731 are weak. They infer a density of the
975: low-ionization part of the NLR of \n\,$=10^{6-7}$\,\cm.
976: \citet{lao97b} detected on the same object
977: very weak [\ion{C}{3}] $\lambda1907$ and [\ion{Si}{3}] $\lambda1883$ emission
978: in a high S/N UV spectrum, suggesting a NLR component with
979: \n\,$ \sim 5 \times 10^5$\,\cm. 
980: 
981: Rodriguez-Pascual et al. (1997) analyzed the UV properties of a sample of NLS1 galaxies.
982: Based on their photoionization modeling of the emission lines, they
983: tentatively favored a {\em low-density} BLR in NLS1 galaxies.
984: \citet{ferland89} presented observations of \caii\ emission lines from AGN
985: with strong \feii\ emission. They suggested 
986: that the BLR density in strong \feii\ emitters is not higher 
987: than in other sources, based on the ratio of the forbidden lines
988: relative to the \caii\ triplet.
989: On the other hand, other studies (e.g., Kuraszkiewicz et al. 2000; 
990: Wills et al. 2000; Marziani et al. 2001; Bachev et al. 2004) 
991: favored high-density BLRs, based on the 
992: \ion{Si}{3}]~$\lambda1892$/\ion{C}{3}]~$\lambda1909$ ratio
993: which is sensitive to density; such results are consistent with
994: predictions by Gaskell (1985) that UV spectra of NLS1 galaxies
995: would show a larger \ion{Si}{3}]~$\lambda1892$/\ion{C}{3}]~$\lambda1909$
996: ratio. 
997: More recently, in studies of the optical-UV emission-line spectra of AGN,
998: Marziani et al. (2001) and Bachev et al. (2004) reported indications
999: of a systematic increase in density toward AGN with smaller FWHMs of
1000: the broad component of \hb\, based on the line ratio
1001: \ion{Si}{3}]~$\lambda1892$/\ion{C}{3}]~$\lambda1909$ (they could not exclude
1002: the alternative interpretation of varying metal abundances, though).
1003: Comastri et al. (1998) reported detection of a deviation of the BLR Balmer
1004: line ratio \ha/\hb\ from the recombination value, such that the ratio
1005: was below the recombination value. They interpreted this as an indication
1006: of high density of the BLR of the NLS1 galaxies.
1007: 
1008: Our finding is that NLS1 galaxies have lower average NLR density. Therefore, 
1009: if NLR and BLR are closely coupled, then our results favor models
1010: which also predict lower average BLR density in NLS1 galaxies (i.e., a stronger
1011: scatter including objects with lower density).
1012: 
1013: \subsection{Supersolar metallicities and temperature effects}
1014: There are several indications that NLS1 galaxies have supersolar metallicities (e.g.,
1015: Mathur 2000; Komossa \& Mathur 2001; Nagao et al. 2002; 
1016: Shemmer and Netzer 2002; Romano et al. 2004; Fields et al. 2005). 
1017: We already noted in Sect.\,6 that those objects with reliable 
1018: \oiii\ $\lambda$4363 measurements do not indicate enhanced temperatures 
1019: in the low-density objects. However, the number of good spectra is still 
1020: relatively small. Nagao et al. (2001) also examined the temperature-sensitive 
1021: \oiii\ ratio for a small sample of NLS1 galaxies and concluded that they do not 
1022: significantly deviate from BLS1 galaxies.
1023: On the theoretical side, whether an increase in metals, and specific 
1024: elements in particular, leads to increased heating or cooling of the gas, 
1025: and thus an increase or decrease of its temperature, needs to be assessed
1026: by detailed photoionization calculations. 
1027: In general, increasing the oxygen abundance first leads to a {\em decrease} 
1028: in temperature since oxygen is an important coolant. Such an effect would 
1029: shift the data points in the \n\,--\,FWHM(\hbb) diagram to even lower densities 
1030: (because of the dependence of the \sii\ ratio on temperature). 
1031: In summary, we do not expect metallicity effects to play a dominant role 
1032: in explaining the \n\,--\,FWHM(\hbb) diagram.  
1033: 
1034: 
1035: \subsection{Starburst contribution}
1036: If we had a strong starburst contribution in a fraction of our sources, 
1037: then this would lead to lower measured density because H\,II
1038: regions have lower average density (e.g. Osterbrock 1989; Ho et al. 1997).
1039: If NLS1 galaxies are "young objects" (e.g., Mathur 2000; 
1040: Mathur, Kuraszkiewicz \& Czerny 2001; Grupe \& Mathur 2004; 
1041: Mathur \& Grupe 2005), still in the process of growing their black holes, 
1042: then they may possibly also show enhanced starburst activity. 
1043: However, as already being checked in Sect.6, NLS1 galaxies, including 
1044: {\em low-density} objects, do not show signs of stronger starburst 
1045: contribution to their optical emission lines than BLS1 galaxies. 
1046: We thus conclude 
1047: that starburst effects are negligible% 
1048: \footnote{
1049: In a study of a sample of 74 post-starburst type\,I AGN, which underwent a
1050: strong recent star formation epoch but stopped forming stars, \citet{zhou05}
1051: found that more than half of them are NLS1 galaxies. We checked whether their 
1052: sample, which also includes BLS1 galaxies, shows any special preference 
1053: for certain 
1054: density values, particularly whether the post-starburst NLS1 galaxies would 
1055: preferentially have low density. 
1056: We find that their sources 
1057: populate similar areas in the \n\,--\,FWHM(\hbb) diagram. 
1058: We have few post-starbursts among the low-density NLS1 galaxies in our sample, 
1059: based on the lack of evidence 
1060: of strong post-burst features in their host galaxies
1061: (equivalent width of H$\delta$ absorption line EW(H$\delta$) $>$ 5 \AA\ 
1062: for post-starbursts.}.
1063: 
1064: \subsection{NLR extent}
1065: BLS1 galaxies would lack an observable {\em low-density} NLR component, if the 
1066: {\em low-density} part of the NLR was selectively obscured. However,
1067: spatially-resolved spectroscopy of the NLRs of nearby Seyfert galaxies 
1068: shows that density declines as a function of cloud distance from the center. 
1069: This would imply that it is the outer parts being obscured,
1070: leading to a peculiar and unlikely geometry. 
1071: 
1072: Alternatively, lower average NLR density of the small--FWHM(\hbb) objects 
1073: may imply that their NLRs are, on average, more extended, since density 
1074: declines outward.  Indeed, there were early 
1075: suggestions that the emission-line regions of NLS1 galaxies are at larger nuclear 
1076: separations than those of BLS1 galaxies (e.g., Giannuzzo et al. 1999).
1077: Such a NLR model is consistent with a larger BLR distance in NLS1 galaxies (e.g.,  
1078: Wandel \& Boller 1998, Puchnarewicz et al. 2001). 
1079: 
1080: However, we note, that other parameters also play a role
1081: in determining NLR extent. 
1082: In particular, there appears to be a scaling between luminosity of \oiii\ and 
1083: NLR extent (Bennert et al. 2002, Schmitt et al. 2003), such that more luminous 
1084: objects have more extended emission-line regions. 
1085: We checked whether our sample shows a correlation between density and
1086: \oiii\ luminosity. No obvious trend is found.
1087: 
1088: \subsection{Fraction of matter-bounded clouds}
1089: A substantial part of the \sii\ emission is produced in the
1090: partially ionized zone of the NLR clouds.
1091: {\em If} the fraction of matter-bounded clouds varies 
1092: as a function of the distance of the NLR clouds from the nucleus,
1093: and {\em if} the fraction of matter-bounded clouds in BLS1 galaxies
1094: is {\em higher} at {\em larger} nuclear distances than it is in
1095: NLS1 galaxies, then (given observations of density stratification
1096: within the NLR from high to low density) some of the
1097: low-density clouds in BLS1 galaxies partly escape
1098: our measurement, with the consequence that the average
1099: measured density in BLS1 galaxies is higher. Or, to put it the other way,
1100: a larger scatter in the number of matter-bounded low-density clouds
1101: in NLS1 galaxies would lead to a larger scatter in their average measured
1102: densities.
1103: 
1104: There are no direct measurements of the fraction of matter-bounded
1105: clouds in NLS1 galaxies and BLS1 galaxies.  However, 
1106: several observations do indicate a wide of range of
1107: column densities of the emission line
1108: clouds in NLS1 galaxies: Ferland \& Persson (1989) need very high column
1109: densities to reproduce the strength of the Calcium emission in 
1110: objects with strong \feii\ emission, while Rodriguez-Pascual et al. (1997) 
1111: infer a higher fraction of
1112: matter-bounded BLR clouds in NLS1 galaxies, based on UV observations.
1113: Regarding the NLR of NLS1 galaxies, Contini et al. (2003) require
1114: matter-bounded clouds in order to
1115: reproduce the strength of high-ionization iron lines.   
1116: Photoionization models of the NLR
1117: of NLS1 galaxies presented by Rodriguez-Ardila et al. (2000b)  
1118: invoke a mixture of matter-bounded and ionization-bounded clouds,
1119: the inner, high-density clouds mostly matter-bounded
1120: and producing high-ionization lines; the outer,
1121: low-density clouds mostly ionization-bounded and producing
1122: mostly low-ionization lines. However, 
1123: Rodriguez-Ardila et al. (2005) discuss a different type
1124: of NLR models, involving shocks and photoionization,
1125: and in that model it is shocks which produce the 
1126: low-ionization lines.    
1127: 
1128: Future spatially resolved long-slit spectroscopy of the NLRs of nearby
1129: NLS1 galaxies will allow a direct comparison of the density profiles of
1130: BLS1 galaxies (e.g., Bennert et al. 2006b and references therein) 
1131: with those of NLS1 galaxies. 
1132: Making use of other line ratios in addition to \sii\ 
1133: may also allow us to determine the fraction of matter-bounded clouds  
1134: as a function of radius, and enable us to measure the NLR extent of NLS1 galaxies.
1135: 
1136: \subsection{ISM of the host galaxy}
1137: The NLR clouds are most likely directly related to the ISM of the host
1138: galaxy. The NLR properties then might also reflect different gas enrichment
1139: mechanisms of the ISM. For instance, the gas could be due to local stellar
1140: processes or transported inward from much larger scales
1141: (or else could arise from the inner parts in case of outflowing winds,
1142: driven by the central engine; see Sect.\,7.7).
1143: The absence of low-density clouds in BLS1 galaxies may thus reflect
1144: the properties of the ISM in the host galaxy.
1145: 
1146: The host galaxies of NLS1 galaxies are often spirals, but not much is 
1147: known about their systematic properties.
1148: There are indications that the host galaxies of NLS1 galaxies and 
1149: BLS1 galaxies differ, particularly in the sense 
1150: that large-scale bars are more common in NLS1 galaxies \citep{cre03,ohta07}, 
1151: and that 
1152: NLS1 galaxies have more grand-design dust spirals 
1153: and a higher fraction of nuclear star-forming rings \citep{deo06}. 
1154: Moreover, \citet{kron01} examined the host galaxies 
1155: and found NLS1 galaxies reside in galaxies with smaller 
1156: diameters than BLS1 galaxies. 
1157: Previous studies on mass--luminosity/sigma relation led to 
1158: conflicting possibilities, i.e., NLS1 galaxies are hosted by 
1159: less luminous galaxies
1160: (e.g., Wang \& Lu 2001; Botte et al. 2004, 2005) 
1161: or on the contrary, more luminous galaxies for a given black hole mass
1162: (e.g., Mathur, Kuraszkiewicz \& Czerny 2001;
1163: Grupe \& Mathur 2004; Bian \& Zhao 2004; 
1164: Ryan et al. 2007).  
1165: 
1166: Many of the morphological differences of the host galaxies of 
1167: NLS1 galaxies and BLS1 galaxies 
1168: are due to the presence or absence of a large-scale stellar bar \citep{deo06}.
1169: Dynamical models show that a bar potential can efficiently drive gas from the 
1170: outer regions (several kpc) to within $\sim$1\,kpc from the nucleus, at which 
1171: point the bar-driven gas flow slows or even stalls at the inner Lindblad 
1172: resonance and the infalling gas will form a disk \citep{shl90}. 
1173: Gas inflow along a galactic stellar bar \citep{simkin80,shl00}, has been 
1174: proposed to be one possible trigger of AGN activity
1175: (e.g. Shlosman, Begeman \& Frank 1990).
1176: However, how these processes affect the ISM density distribution
1177: on the scales of the NLR is presently unclear.
1178: Further studies of ISM properties of NLS1 and BLS1 galaxies on sub-kpc
1179: scales are needed to further address this issue. 
1180: 
1181: \subsection{Outflows}
1182:  
1183: Lawrence et al. (1997) suggested that the density of an outflowing wind
1184: might be an important ingredient in understanding emission-line parameters 
1185: of AGN and correlations among them. They loosely suggested strong \feii\ 
1186: emitters have the denser winds, in the context of a model where
1187: parts of the BLR are mechanically heated and produce the \feii\ emission. 
1188: We do find that BLS1 galaxies and NLS1 galaxies differ indeed 
1189: in the density of their NLRs, 
1190: but that it is actually the BLS1 galaxies which do harbor the 
1191: higher density clouds.
1192: A correlation analysis further shows that \n\ is 
1193: anti-correlated with R4570 ($r_{\rm s}=-0.47$, $P_{\rm null}<10^{-4}$; Fig.\,8).
1194: The {\em low-density} objects show larger-than-average R4570 compared to
1195: the {\em high-density} NLS1 galaxies. 
1196: The average R4570 for the {\em low-density} objects and the 
1197: {\em high-density} NLS1 galaxies is 0.95 and 0.61, respectively.
1198: 
1199: \citet{schiano86} predicted that the average NLR density should be 
1200: higher in very luminous objects than in lower luminosity objects.
1201: The prediction was based on their '{\em quasar wind }' model, 
1202: i.e., the NLR is the result of the interaction of 
1203: AGN ionizing photons and a thermal wind on
1204: dense, massive interstellar clouds. 
1205: However, our result conflicts with this prediction.
1206: The {\em low-density} regime in the \n\--FWHM \hbb\ 
1207: diagram (Fig.\,5) is not dominated by low-luminosity systems, 
1208: but a mixture of high- and low-luminosity NLS1 galaxies. The quasars and
1209: low-luminosity Sy1 galaxies 
1210: have almost the same average NLR density
1211: of $\approx$ 290\,\cm.
1212: 
1213: There are indications that many (but not all; e.g., Xu et al. 2003; 
1214: Williams, Mathur \& Pogge 2004) 
1215: NLS1 galaxies accrete close to or even above the Eddington limit
1216: (e.g., Boroson \& Green 1992; Wang et al. 1996;
1217: Boller et al. 1996; Laor et al. 1997; 
1218: Sulentic, Marziani \& Dultzin-Hacyan 2000; Boroson 2003; Grupe 2004; 
1219: Grupe \& Mathur 2004; Collin \& Kawaguchi 2004).
1220: This likely comes with the presence of strong outflows. 
1221: If these still propagate up into the NLR, then we may expect that the
1222: NLR gas in such objects is actually more tenuous.
1223: Radiation-pressure driven wind models predict a decrease of accretion rate with
1224: increasing width of the broad component of \hb\ (e.g., Nicastro 2000; 
1225: Witt et al. 1997).
1226: Among the NLS1 population itself, 
1227: it should then be the objects with accretion rates closer to Eddington 
1228: that drive the stronger winds and thus have more tenuous, {\em low-density} 
1229: NLR components. 
1230: In order to test this, in Fig.\,9, 
1231: we plot Eddington ratio \LLeddbol\ as a function of density \n. We
1232: estimate the bolometric luminosities using 
1233: $L_{\rm bol} \approx 9\lambda L_{\lambda}(5100 \AA)$ \citep{kaspi00}, 
1234: while the Eddington luminosities are calculated using the black hole masses
1235: determined using the BLR radius and the velocity of the BLR gas
1236: (e.g. Peterson 1997). We find that \LLeddbol\ is higher in 
1237: NLS1 galaxies than in BLS1 galaxies.
1238: An anti-correlation of decreasing electron density with
1239: increasing Eddington ratio can be seen across our entire
1240: sample of NLS1 and BLS1 galaxies 
1241: ($r_{\rm s}=-0.42$, $P_{\rm null}=10^{-4}$).
1242: However, among the NLS1 population itself,
1243: the {\em low-density} objects do not show higher-than-average 
1244: \LLeddbol\ compared to the {\em high-density} NLS1 galaxies, which may suggest
1245: that higher \LLeddbol\ is a necessary 
1246: but not a sufficient 
1247: condition to lower density. 
1248: Independent calculation of \LLedd,
1249: e.g., estimation from X-ray observations, 
1250: is of great importance to check the trend.
1251: 49 out of 93 objects have X-ray counterparts in the RASS catalogs
1252: (e.g. Voges et al. 1999). 
1253: We had a first look at the X-ray data. 40 of them have enough photon 
1254: counts for an estimation of the X-ray slope and thus X-ray luminosity in 
1255: the ROSAT band. The anti-correlation between \n\ and 
1256: \xLLedd\ is even stronger ($r_{\rm s}=-0.47$, $P_{\rm null}=3 \times 10^{-3}$),
1257: particularly in the sense that {\em low-density} objects do have 
1258: higher-than-average \xLLedd\ compared to the {\em high-density} NLS1 galaxies, 
1259: with average \xLLedd\ = 3.2 and 1.2, respectively. 
1260: Further careful study of the X-ray spectra and follow-up X-ray 
1261: observations of the {\em low-density} objects will be 
1262: crucial to understand the cause of the correlation. 
1263: 
1264: If outflow was a key mechanism to explain the lower average NLR density 
1265: in NLS1 galaxies, 
1266: then we would expect that the density (\sii\ ratio) scales with 
1267: the \oiii\ outflow velocity (blueshift). 
1268: We checked for both, NLS1 galaxies vs. BLS1 galaxies, and within 
1269: the NLS1 sample 
1270: ({\em high-density} objects vs. {\em low-density} objects) whether 
1271: the density correlates with the \oiii\ (peak) blueshift,
1272: and only found a weak correlation 
1273: ($r_{\rm s}=-0.29$, $P_{\rm null}=7 \times 10^{-3}$)%
1274: \footnote{This does not yet exclude that the whole NLR is in outflow. 
1275: The \oiii\ peak blueshift was calculated relative to the
1276: the low-ionization lines; i.e., \sii\ and \nii. Ideally,
1277: one should measure the shift between host galaxy absorption lines and NLR
1278: emission lines, but most of our spectra are AGN dominated with few absorption
1279: lines detected.}. 
1280: However, it is interesting to note that the peak blueshift of \oiii\ does
1281: strongly correlate with \LLeddbol\
1282: ($r_{\rm s}=0.51$, $P_{\rm null}<10^{-4}$).
1283: This correlation then indicates that outflows are more common
1284: in objects with high accretion rates.
1285: We also checked whether there is a correlation between density and 
1286: the blueshift of the {\em blue wing} of \oiii%
1287: \footnote{57 out of 93 objects clearly show blue wings.
1288: The blueshift of the peak of the blue wing was measured 
1289: against the peak position of the core of the [O III] line.}, since a 
1290: preferred interpretation of blue wings is the existence of outflows
1291: (or inflows) combined with viewing angle effects (e.g., Boroson 2005).
1292: A correlation is seen with 
1293: $r_{\rm s}=-0.43$ ($P_{\rm null}=1\times10^{-3}$).
1294: This correlation shows that outflows are stronger in 
1295: the {\em low-density} objects (Fig.\,10), 
1296: even if the bulk of the NLR does not participate in the outflow.
1297: 
1298: In summary, we find several indications which point toward a link between
1299: NLR density and outflows, and we tentatively favor the role of outflows 
1300: in explaining the difference in the NLR density between
1301: NLS1 galaxies and BLS1 galaxies. 
1302: 
1303: \section{Summary and conclusions}
1304: We have studied one of the largest homogeneously analyzed 
1305: sample of NLS1 galaxies 
1306: in order to examine whether or not there is a difference in electron density
1307: \n\ between NLS1 galaxies and BLS1 galaxies.
1308: We employ a powerful diagnostic,  
1309: the density-sensitive line ratio \sii\ $\lambda6716/\lambda6731$, to 
1310: measure the NLR density. We show that the galaxies do not homogeneously 
1311: populate the \n\,--FWHM(\hbb) diagram.
1312: Our key finding is the detection of a
1313: '{\em zone of avoidance}' in the \n\,--\,FWHM(\hbb) plane:
1314: BLS1 galaxies (FWHM(\hbb) $>$ 2000 \kms) avoid low average densities,
1315: and all show \n\,$>140$\,\cm. 
1316: On the other hand, NLS1 galaxies show a larger scatter in densities in the range
1317: \n\,$=2 \sim 770$\,\cm, including a significant number of objects with
1318: low densities.
1319: The results obtained for \sii\ are consistent with the \oi\,--\,FWHM(\hbb)
1320: diagram, which shows higher average \oi/H$\alpha$ intensity for BLS1 galaxies.
1321: 
1322: We investigated a number of different explanations for the 
1323: '{\em zone of avoidance}' in density. 
1324: We find that supersolar metallicities
1325: and temperature effects, a strong starburst contribution in NLS1 galaxies,
1326: and the effect of NLR extent are unlikely explanations.
1327: Consequences of the fraction of matter bounded clouds,
1328: and different properties of the ISM in the host galaxies,
1329: can only be further judged with  
1330: future observations.
1331: We find several lines of evidence that outflows
1332: play a significant role in driving the difference in the NLR
1333: between NLS1 galaxies and BLS1 galaxies, and favor these as explanation
1334: for the zone of avoidance in the density-FWHM(\hbb) diagram. 
1335: 
1336: \acknowledgments
1337: DX and SK thank the Chinese National Science
1338: Foundation (NSF) for support under grant NSFC-10503005.
1339: DX acknowledges the Max-Planck-Institut f\"ur extraterrestrische Physik and
1340: the Max-Planck-Gesellschaft for financial support.
1341: We are grateful to Hongling Lu for running her independent component 
1342: analysis software.
1343: We also thank Chen Cao, Biwei Jiang and Xiaobo Dong for helpful discussion 
1344: on various softwares used in the spectral analysis, and Martin Gaskell,
1345: Caina Hao and Jing Wang for useful conversations. 
1346: This research made use of the SDSS archives and the Catalogue of Quasars 
1347: and Active Nuclei. 
1348: Funding for the
1349: creation and the distribution of the SDSS Archive has been provided
1350: by the Alfred P. Sloan Foundation, the Participating Institutions,
1351: the National Aeronautics and Space Administration, the National
1352: Science Foundation, the U.S. Department of Energy, the Japanese
1353: Monbukagakusho, and the Max Planck Society. The SDSS is managed by
1354: the Astrophysical Research Consortium (ARC) for the Participating
1355: Institutions. The Participating Institutions are The University of
1356: Chicago, Fermilab, the Institute for Advanced Study, the Japan
1357: Participation Group, The Johns Hopkins University, Los Alamos
1358: National Laboratory, the Max-Planck-Institute for Astronomy (MPIA),
1359: the Max-Planck-Institute for Astrophysics (MPA), New Mexico State
1360: University, Princeton University, the United States Naval
1361: Observatory, and the University of Washington.
1362: 
1363: \begin{thebibliography}{}
1364: \bibitem[Abazajian et al.(2005)]{sloandr3} Abazajian, K., Adelman-McCarthy, 
1365: J.K., Ag{\"u}eros, M.A., et al. 2005, \aj, 129, 1755
1366: \bibitem[Aoki, Kawaguchi \& Ohta(2005)]{aoki05} Aoki, K., Kawaguchi, T.,
1367:   \& Ohta, K. 2005, \apj, 618, 601
1368: \bibitem[Bachev et al.(2004)]{bach04} Bachev, R., Marziani, P., Sulentic, J.W.,
1369:   Zamanov, R., Calvani, M., \& Dultzin-Hacyan, D. 2004, \apj, 617, 171
1370: \bibitem[Baskin \& Laor(2005)]{baskin05} Baskin, A., \& Laor, A. 2005,
1371:   \mnras, 358, 1043
1372: \bibitem[Bennert et al.(2002)]{bennert02} Bennert, N., Falcke H., Schulz, 
1373:   H. Wilson, A.S., \& Wills, B.J. 2002, \apj, 574, 105
1374: \bibitem[Bennert et al.(2006a)]{bennert06a} Bennert, N., Jungwiert, B.,
1375:   Komossa, S., Haas, M., \& Chini, R. 2006a, \aap, 459, 55
1376: \bibitem[Bennert et al.(2006b)]{bennert06b} Bennert, N., Jungwiert, B., 
1377:   Komossa, S., Haas, M., \& Chini, R. 2006b, \aap, 456, 953 
1378: \bibitem[Bian \& Zhao(2004)]{bian04} Bian, W., \& Zhao, Y. 2004, 
1379:   \mnras, 347, 607 
1380: \bibitem[Boller et al.(1996)]{bol96} Boller, Th., Brandt, W.N., \& Fink, H. 
1381:   1996, \aaps, 305, 53
1382: \bibitem[Boroson \& Green(1992)]{bg92} Boroson, T.A., \& Green, R.F. 1992, 
1383:   \apjs, 80, 109
1384: \bibitem[Boroson(2002)]{boroson02} Boroson, T.A. 2002, \apj, 565, 78
1385: \bibitem[Boroson(2003)]{boroson03} Boroson, T.A. 2003, \apj, 585, 647
1386: \bibitem[Botte et al.(2004)]{botte04}
1387:    Botte, V., Ciroi, S., Rafanelli, P., \& Di Mille, F. 2004, \aj, 127, 3168
1388: \bibitem[Botte et al.(2005)]{bott05}
1389:    Botte, V., Ciroi, S., P., Di Mille, F., Rafanelli, P., \& Romano, A. 2005,
1390:   \mnras, 356, 789  
1391: \bibitem[Brotherton et al.(1994)]{bro94} Brotherton, M.S., Wills, B.J., 
1392:   Francis, P.J., \& Steidel, C.C. 1994, \apj, 430, 495
1393: \bibitem[Collin \& Kawaguchi(2004)]{collin04} Collin, S., \& Kawaguchi, T. 2004,
1394:   \aap, 426, 797
1395: \bibitem[Collin et al.(2006)]{collin06} Collin, S., Kawaguchi, T., 
1396:   Peterson, B.M., \& Vestergaard, M. 2006, \aap, 456, 75 
1397: \bibitem[Comastri et al.(1998)]{com98} Comastri, A., Fiore, F., Guainazzi, M., 
1398:   et al. 1998, \aap, 333, 31
1399: \bibitem[Constantin \& Shields(2004)]{con04} Constantin, S., \& 
1400:   Shields, J. 2003, \pasp, 115, 592
1401: \bibitem[Contini et al.(2003)]{conti03} Contini, M., Rodriguez-Ardila, A., 
1402:     \& Viegas, S.M. 2003, \aap, 408, 101 
1403: \bibitem[Contini \& Viegas(2001)]{contini01} Contini, M., \& Viegas, S.M. 
1404:   2001, \apjs, 132, 211
1405: \bibitem[Corbin(1995)]{cor95} Corbin, M.R. 1995, \apj, 447, 496
1406: \bibitem[Corbin(1997)]{cor97} Corbin, M.R. 1997, \apjs, 113, 245
1407: \bibitem[Crenshaw(1986)]{cre86} Crenshaw, D.M. 1986, \apjs, 62, 821
1408: \bibitem[Crenshaw et al.(2002)]{cre02} Crenshaw, D.M., Kraemer, S.B., 
1409:   \& Turner, T.J., et al. 2002, \apj, 566, 187
1410: \bibitem[Crenshaw, Kraemer \& Gabel(2003)]{cre03} Crenshaw, D.M., 
1411:   Kraemer, S.B., \& Gabel, J.R. 2003, \aj, 126, 1690
1412: \bibitem[de Robertis(1985)]{de85} de Robertis, M.M. 1985, \apj, 289, 67
1413: \bibitem[Deo, Crenshaw \& Kraemer(2006)]{deo06} Deo, R.P., Crenshaw, D.M., 
1414:   \& Kraemer, S.B. 2006, \aj, 132, 321
1415: \bibitem[Dewangan et al.(2001)]{dewangan01} Dewangan, G.C., Singh, K.P., 
1416:   Jones, L.R., M$^c$Hardy, I.M., Mason, K.O., \& Newsam, A.M. 2001,
1417:   \mnras, 325, 1616
1418: \bibitem[Dietrich, Grenshaw \& Kraemer(2005)]{dietrich05} Dietrich, M., 
1419:   Grenshaw, D.M., \& Kraemer, S.B. 2005, \apj, 623, 700
1420: \bibitem[Elvis(2000)]{elvis00} Elvis, M. 2000, \apj, 545, 63
1421: \bibitem[Evans(1988)]{eva88} Evans, I.N. 1988, \apjs, 67, 373
1422: \bibitem[Ferland \& Persson(1989)]{ferland89} Ferland, G.J., \& Persson, S.E. 
1423:   1989, \apj, 347, 656
1424: \bibitem[Fields et al.(2005)]{fields05} Fields, D.L., Mathur, S., 
1425:   Pogge, R.W., Nicastro, F., Komossa, S., \& Krongold, Y. 2005, \apj, 634, 928
1426: \bibitem[Filippenko \& Sargent(1988)]{fil88} Filippenko, A.V., \& 
1427:   Sargent, W.L.W. 1988, \apj, 324, 134
1428: \bibitem[Francis et al.(1992)]{fra92} Francis, P.J., Hewett, P.C., Foltz, C.B.,
1429:  \& Chaffee, F.H. 1992, \apj, 398, 476
1430: \bibitem[Francis et al.(1996)]{francis96} Francis, P.J. 1996, PASA, 13, 212
1431: \bibitem[Fraquelli, Storchi-Bergmann \& Binette(2000)]{fraquelli00}
1432:   Fraquelli, H.A., Storchi-Bergmann, T., \& Binette, L. 2000, \apj, 532, 867 
1433: \bibitem[Gaskell(1984)]{gas84} Gaskell, C.M. 1984, ApL, 24, 43 
1434: \bibitem[Gaskell(1985)]{gas85} Gaskell, C.M. 1985, \apj, 291, 112
1435: \bibitem[Giannuzzo et al.(1999)]{gia99} Giannuzzo M.E., Mignoli M., 
1436:   Stirpe G.M., Comastri A. 1999, ASP Conf. Series, 175, p.291, 
1437:   C.M. Gaskell et al. (eds)
1438: \bibitem[Gierlinski \& Done(2004)]{gie04} Gierlinski, M., \& Done, C. 2004, 
1439:   \mnras, 347, 885
1440: \bibitem[Goodrich(1989)]{good89} Goodrich, R.W. 1989, \apj, 340, 190
1441: \bibitem[Goodrich(2000)]{good00} Goodrich, R.W. 2000, NewA Rev., 44, 519 
1442: \bibitem[Grupe et al.(1998)]{gru98} Grupe, D., Beuermann, K., Thomas, H.-C., 
1443:   Mannheim, K., \& Fink H.H. 1999, \aap, 350, 31
1444: \bibitem[Grupe et al.(1999)]{gru99} Grupe, D., Beuermann, K., Mannheim, K., 
1445:   \& Thomas, H.-C. 1999, \aap, 350, 31
1446: \bibitem[Grupe \& Mathur(2004)]{grumat04} Grupe D., \& Mathur S. 2004, 
1447:   \apj, 606, L31
1448: \bibitem[Grupe(2004)]{gru04} Grupe, D. 2004, \aj, 127, 1799
1449: \bibitem[Greene \& Ho(2005a)]{greene05a} Greene, J.E., \& Ho, L.C. 2005a,
1450:   \apj, 627, 721 
1451: \bibitem[Greene \& Ho(2005b)]{greene05b} Greene, J.E., \& Ho, L.C. 2005b,
1452:   \apj, 630, 122 
1453: \bibitem[Greene \& Ho(2004)]{greene04} Greene, J.E., \& Ho, L.C. 2004,
1454:    \apj, 610, 722
1455: \bibitem[Hao(2005)]{hao05} Hao, L., Strauss, M.A., Tremonti, C.A., et al. 2005,
1456:   \aj, 129, 1783
1457: \bibitem[Hippelein et al.(2003)]{hip03} Hippelein, H., Maier, C., 
1458:   Meisenheimer, K., et al. 2003, \aap, 402, 65
1459: \bibitem[Ho(1997)]{ho97} Ho, L. C., Filippenko, A.V., \& Sargent, W.L.W. 1997,
1460:   \apj, 487, 579
1461: \bibitem[Ho(2005)]{ho05} Ho, L.C. 2005, \apj, 629, 680 
1462: \bibitem[Hopkins et al.(2005)]{hopkins05} Hopkins, P.F., Hernquist, L.,
1463:  Cox, T.J., Di Matteo, T., Martini, P., Robertson, B., \& Springel, V. 2005,
1464:  \apj, 630, 705
1465: \bibitem[Kaspi et al.(2000)]{kaspi00} Kaspi, S., Smith, P.S., Netzer, H., 
1466:   et al. 2000, \apj, 533, 631
1467: \bibitem[Kauffmann et al.(2003)]{kau03} Kauffmann, G., et al. 2003, \mnras, 
1468:   346, 1055
1469: \bibitem[Kewley \& Geller(2004)]{kewley04} Kewley, L.J., \& Geller, M.J.,
1470:   2004, \aj, 127, 2002 
1471: \bibitem[Komossa \& Bade(1999)]{koba99} Komossa, S., \& Bade, N. 1999, 
1472:   \aap, 343, 775 
1473: \bibitem[Komossa \& Schulz(1997)]{stefanie97} Komossa, S., \& Schulz, H. 1997,
1474:   \aap, 323, 31
1475: \bibitem[Komossa \& Meerschweinchen(2000)]{stefanie00} Komossa, S., 
1476:   \& Meerschweinchen, J. 2000, \aap, 354, 411 
1477: \bibitem[Komossa et al.(2001)]{stefanie01a} Komossa, S., Grupe, D., 
1478:   \& Janek, M. 2001,
1479:   in AIP Conf. Proc. 599, X-ray Astronomy '999: stellar endpoints, 
1480:   AGN and the diffuse background, ed. N. White, G. Malaguti, \& G. Palumboi,
1481:   686
1482: \bibitem[Komossa \& Mathur(2001)]{stefanie01b} Komossa, S., \& Mathur, S. 2001,
1483:   \aap,  374, 914 
1484: \bibitem[Komossa et al.(2006a)]{stefanie06a} Komossa, S., Voges, W., 
1485:  Adorf, H.-M., Xu, D., Mathur, S., \& Anderson, S.F. 2006a, \apj, 639, 710 
1486: \bibitem[Komossa et al.(2006b)]{stefanie06b} Komossa, S., Voges, W., Xu, D., 
1487:  Mathur, S., Adorf, H.-M., Lemson, G., Duschl, W., \& Grupe, D. 2006b, \aj, 
1488:  132, 531
1489: \bibitem[Komosssa \& Xu(2007)]{kxu07} Komossa, S., \& Xu, D., 2007, ApJ Letters, submitted
1490: \bibitem[Krongold et al.(2001)]{kron01} Krongold et al., 2001, \aj, 121, 702 
1491: \bibitem[Kuraszkiewicz et al(2000)]{kur00} Kuraszkiewicz, J., Wilkes, B.J.,
1492:   Czerny, B., \& Mathur, S. 2000, \apj, 542, 692
1493: \bibitem[Kuraszkiewicz et al(2000)]{kur01}Kuraszkiewicz, J., Mason, K.O.,
1494:   Siemiginowska, A., et al. 2001, \apj, 550, 644
1495: \bibitem[Kriss(1994)]{kriss94} Kriss, G.A. 1994, in ASP Conf. Series 61, 
1496:   Astronomical Data Analysis Software and Systems III, ed. D.R. Crabtree, 
1497:   R.J. Hanisch, \& J. Barnes (San Francisco: ASP), 437
1498: \bibitem[Laor et al.(1997a)]{lao97} Laor, A., Fiore, F., Martin, E., et al. 
1499:   1997a, \apj, 477, 93
1500: \bibitem[Laor et al.(1997b)]{lao97b} Laor, A., Jannuzi, B.T., Green, R.F.,
1501:   \& Boroson, T. 1997b, \apj, 489, 656
1502: \bibitem[Lawrence et al.(1997)]{law97} Lawrence, A., Elvis, M., Wilkes, B.J.,
1503:  McHardy, I., \& Brandt, N. 1997, \mnras, 285, 879
1504: \bibitem[Leighly(1999)]{lei99} Leighly, K.M. 1999, \apjs, 125, 317
1505: \bibitem[Leighly \& Moore(2004)]{lei04} Leighly, K.M. \& Moore, J.R. 2004,
1506:  \apj, 611, 107
1507: \bibitem[Lu et al.(2006)]{lu06} Lu, H., Zhou, H., Wang, T., Dong, X., \& 
1508:   Li, C. 2006, \aj, 131, 790
1509: \bibitem[Mathur(2000)]{mat00} Mathur, S. 2000, \mnras, 314, L17 
1510: \bibitem[Mathur, Kuraszkiewicz \& Czerny(2001)]{mat01} Mathur, S., 
1511:  Kuraszkiewicz, J., \& Czerny, B. 2001, NewA, 6, 321
1512: \bibitem[mathur \& Grupe(2005)]{mat05} Mathur, S., \& Grupe, D. 2005, \apj,
1513:     633, 688
1514: \bibitem[Marziani et al.(2001)]{mar01} Marziani, P., Sulentic, J.W., 
1515:   Zwitter, T., Dultzin-Hacyan, D., \& Calvai, M. 2001, \apj, 558, 560
1516: \bibitem[Miller et al.(1992)]{mil92} Miller, P., Rawlings, S., Saunders, R., 
1517:   \&  Eales, S. 1992, \mnras, 254, 93
1518: \bibitem[Nagao et al.(2001)]{nagao01} Nagao, T., Murayama, T., \& 
1519:   Yoshiaki, T. 2001, \apj, 546, 744
1520: \bibitem[Nagao et al.(2002)]{nagao02} Nagao, T., Murayama, T., Shioya, Y.,
1521:  \& Taniguchi Y. 2002, \apj, 575, 721
1522: \bibitem[Netzer \& Laor(1993)]{netzer93} Netzer, H., \& Laor, A. 1993,
1523:    \apj, 404, L51
1524: \bibitem[Netzer et al.(2004)]{netzer04} Netzer, H., Shemmer, O., Maiolino, R.,
1525:  Oliva, E., Croom, S., Corbett, E., \& Fabrizio, L.D. 2004, \apj, 614, 558
1526: \bibitem[Ohta et al.(2007)]{ohta07} Ohta, K., Aoki, K., Kawaguchi, T.,
1527:   \& Kiuchi G. 2007, \apjs, 169, 1
1528: \bibitem[Osterbrock \& Shuder(1982)]{ost82} Osterbrock, D.E., \& Shuder, J.M. 
1529:   1982, \apjs, 49, 149
1530: \bibitem[Osterbrock \& Pogge(1985)]{ost85} Osterbrock, D.E., \& Pogge, R.W. 
1531:   1985, \apj, 297, 166
1532: \bibitem[Osterbrock(1989)]{ost89} Osterbrock, D.E. 1989, Astrophysics
1533:   of Gaseous Nebulae and Active Galactic Nuclei (University Science Books:
1534:   Mill Valley, CA)
1535: \bibitem[Peterson(1993)]{peterson93} Peterson, B.M. 1993, \pasp, 105, 247
1536: \bibitem[Peterson(1997)]{peterson97} Peterson, B.M. 1997, An Introduction to 
1537:   Active Galactic Nuclei (Cambridge University Press)
1538: \bibitem[Puchnarewicz et al.(1992)]{pun92} Puchnarewicz, E.M., Mason, K.O., 
1539: C'ordova, F.A., et al. 1992, \mnras, 256, 589
1540: \bibitem[Puchnarewicz et al.(2001)]{pun01} Puchnarewicz, E.M., Mason, K.O.,
1541:  Siemiginowska, A., Fruscione, A., Comastri, A., Fiore, F., \& 
1542:  Cagnoni, I. 2001, \apj, 550, 644
1543: \bibitem[Rice et al.(2006)]{rice06} Rice, M.S., Martini, P., Greene, J.,
1544:   et al. 2006, \apj, 636, 654 
1545: \bibitem[Rodriguez-Ardila et al.(2000a)]{rod00a} Rodriguez-Ardila, A.,
1546:   Pastoriza, M.G., \& Donzelli, C.J. 2000a, \apjs, 126, 63
1547: \bibitem[Rodriguez-Ardila et al.(2000b)]{rod00b} Rodriguez-Ardila, A., 
1548:   Binette, L., Pastoriza, M.G., \& Donzelli, C.J. 2000b, \apj, 538, 581
1549: \bibitem[Rodriguez-Ardila et al.(2005)]{rodr05} Rodriguez-Ardila, A.,  
1550:  Contini, M., \& Viegas, S.M. 2005, \mnras 357, 220
1551: \bibitem[Rodriguez-Pascual et al.(1997)]{rod97} Rodriguez-Pascual P.M., 
1552:   Mas-Hesse J.M., \& Santos-Lleo M. 1997, \aap, 327, 72
1553: \bibitem[Romano et al.(2004)]{romano2004} Romano, P., Mathur, S., Turner, T.J.,
1554:   et al. 2004, \apj, 602, 635
1555: \bibitem[Ryan et al.(2007)]{ryan07} Ryan, C.J., De Robertis, M.M., Viani, S.
1556:  Laor, A., \& Dawson, P.C. 2007, \apj, 654, 799
1557: \bibitem[Schiano(1986)]{schiano86} Schiano A.V.R., 1986, ApJ, 302, 81
1558: \bibitem[Schmitt et al.(2003)]{schmit03} Schmitt, H.R., Donley, J.L., 
1559:  Antonucci, R.R.J. 2003, \apjs, 148, 327
1560: \bibitem[Shaw \& Dufour(1994)]{shaw94} Shaw, R.A., \& Dufour, R.J. 1994, 
1561:  in ASP Conf. Series 61, Astronomical Data Analysis Software and Systems III, 
1562:  ed. D.R. Crabtree, R.J. Hanisch, \& J. Barnes (San Francisco: ASP), 327
1563: \bibitem[Shemmer \& Netzer(2002)]{shemmer02} Shemmer, O., \& Netzer, H. 2002,
1564:   \apj, 567, L19
1565: \bibitem[Shlosman, Begelman \& Frank(1990)]{shl90} Shlosman, I.,
1566:   Begelman, M.C., \& Frank, J. 1990, \nat, 345, 679
1567: \bibitem[Shlosman, Peletier \& Knapen(2000)]{shl00} Shlosman, I., 
1568:    Peletier, R.F., \& Knapen, J. 2000, \apj, 535, L83
1569: \bibitem[Simkin, Su \& Schwarz(1980)]{simkin80} Simkin, S.M., Su, H.J., \& 
1570:   Shwarz, M.P. 1980, \apj, 237, 404
1571: \bibitem[Strateva et al.(2003)]{Strateva03} Strateva, I.V., Strauss, M.A., 
1572:   Hao, L., et al. 2003, \apj, 126, 1720
1573: \bibitem[Stirpe(1991)]{str91} Stirpe, G.M. 1991, \aap, 247, 3
1574: \bibitem[Sulentic et al.(2000)]{sulentic00a}
1575:  Sulentic, J.W., Zwitter, T.,  Marziani, P., \& Dultzin-Hacyan, D. 2000, 
1576:   \apj, 536, 5
1577: \bibitem[Sulentic, Marziani \& Dultzin-Hacyan(2000)]{sulentic00b} 
1578:    Sulentic, J.W., Marziani, P., \& Dultzin-Hacyan, D. 2000, 
1579:   \araa, 38, 521
1580: \bibitem[Sulentic et al.(2002)]{sulentic02} Sulentic, J.W., Marziani, P.,
1581:  Zamanov, R., Bachev, R., Calvani, M., \& Dultzin-Hacyan, D. 2002, 
1582:  \apj, 566, L71 
1583: \bibitem[Sulentic et al.(2003)]{sulentic03} Sulentic, J.W., Zamfir, S.,
1584:  Marziani, P., Bachev, R., Calvani, M., \& Dultzin-Hacyan, D. 2003, 
1585:  \apj, 597, L17
1586: \bibitem[Vaughan et al.(2001)]{vau01} Vaughan, S., Edelson, R., Warwick, R.S., 
1587:   et al. 2001, \mnras, 327, 673
1588: \bibitem[Veilleux, Cecil \& Bland-Hawthorn(2005)]{vei05} Veilleux, S., 
1589:   Cecil G., \& Bland-Hawthorn, J. 2005, \araa, 43, 769
1590: \bibitem[Vanden Berk et al.(2004)]{vanden04} Vanden Berk, D.E.,
1591:  et al. 2004, \aj, 601, 692
1592: \bibitem[Vanden Berk et al.(2001)]{vanden01} Vanden Berk, D.E., 
1593:   et al. 2001, \aj, 122, 549
1594: \bibitem[V\'{e}ron-Cetty \& V\'{e}ron(2000)]{veron00} V\'{e}ron-Cetty, M.P., 
1595:   \& V\'{e}ron, P. 2000, \aapr, 10, 81
1596: \bibitem[V\'{e}ron-Cetty, V\'{e}ron \& Gon\c{c}alves(2001)]{veron01} 
1597:   V\'{e}ron-Cetty, M.P., V\'{e}ron, P., \& Gon\c{c}alves, A.C. 2001, 
1598:   \aap, 372, 730
1599: \bibitem[V\'{e}ron-Cetty, Joly \& V\'{e}ron(2004)]{veron04} V\'{e}ron-Cetty,
1600:  M.-P., Joly, M., \& V\'{e}ron, P. 2004, \aap, 417, 515
1601: \bibitem[V\'{e}ron-Cetty et al.(2006)]{veron06} V\'{e}ron-Cetty,
1602:  M.-P., Joly, M., V\'{e}ron, P., et al. 2006, \aap, 451, 851
1603: \bibitem[Voges et al.(1999)]{voges99} Voges, W., Aschenbach, B., 
1604:   Boller, Th., et al. 1999, \aap, 349, 389
1605: \bibitem[Wang, Brinkmann \& Bergeron(1996)]{wan96} Wang, T., Brinkmann, W., 
1606:  \& Bergeron, J. 1996, \aap, 309, 81
1607: \bibitem[Wandel \& Boller(1998)]{wandel98} Wandel, A. \& Boller, Th.
1608:    1998, \aap, 331, 884
1609: \bibitem[Watson, Mathur, \& Grupe(2006)]{watson06} Watson, L., Mathur, S., 
1610:   \& Grupe, D. 2006, \aj, in press (astro-ph/0611819)
1611: \bibitem[Wei et al.(1999)]{wei99} Wei, J.Y., Xu, D.W., Dong, X.Y., 
1612:   \& Hu, J.Y. 1999, \aaps, 139, 575 
1613: \bibitem[Williams, Pogge, \& Mathur(2002)]{WPM02} 
1614:    Williams, J.M., Pogge, R.W., \& Mathur, S. 2002, \aj, 124, 3042
1615: \bibitem[Williams, Mathur, \& Pogge(2004)]{wil04}
1616:    Williams, J.M., Mathur, S., \& Pogge, R.W. 2004, \apj, 610, 737
1617: \bibitem[Wills et al.(2000)]{wil00} Wills, B.J., Shang, Z., \& Yuan, J.M. 
1618:   2000, NewA Rev., 44, 511
1619: \bibitem[Wills et al.(1993)]{wil93} Wills, B.J., Brotherton, M.S., Fang, D., 
1620:   Steidel, C.C., \& Sargent, W.L.W. 1993, \apj, 415, 563
1621: \bibitem[Witt et al.(1997)]{witt97} Witt, H.J., Czerny, B., \& Solarzycki, P.T. 
1622:   1997, \mnras, 286, 848
1623: \bibitem[Xu et al.(1999)]{xu99} Xu, D.W., Wei, J.Y., \& Hu, J.Y. 1999, 
1624:   \apj, 517, 622
1625: \bibitem[Xu et al.(2003)]{xu03} Xu, D.W., Komossa, S., Wei, J.Y., Qian, Y.,
1626:  \& Zheng, X.Z. 2003, \apj, 590, 73
1627: \bibitem[Zamanov et al.(2002)]{zamanov02} Zamanov, R., Marziani, P., 
1628:   Sulentic, J.W., Calvani, M., Dultzin-Hacyan, D., \& Bachev, R. 2002,
1629:  \apj, 576, L9 
1630: \bibitem[Zhang \& Wang(2006)]{zhang06} Zhang, E.-P., \& J.-M. Wang 2006,
1631:   \apj, 653, 137
1632: \bibitem[Zhou et al.(2005)]{zhou05} Zhou, H., Wang, T., Dong, X., Wang, J.
1633:   \& Lu, H. 2005, Mem. Soc. Astron. Italiana, 76, 93 
1634: \bibitem[Zhou et al.(2006)]{zhou06} Zhou, H., Wang, T., Yuan, W., Lu, H.,
1635:   Dong, X., Wang, J. \& Lu, Y. 2006, \apjs, 166, 128 
1636: \end{thebibliography}
1637: 
1638: \begin{figure}
1639: \plotone{f1.eps}
1640: \caption{
1641: Examples of the decomposition of the spectrum into starlight and 
1642: nuclear continuum.
1643: F$_\lambda$ in units of 10$^{-17}$~erg~s$^{-1}$~cm$^{-2}$~\AA$^{-1}$
1644: is plotted against wavelength in \AA.
1645: In each panel, the original spectrum, the power-law continuum 
1646: of the nucleus, the host galaxy spectrum, the \feii\ template and the 
1647: residual spectrum are shown from top to bottom. For clarity, the residual
1648: spectrum is offset by an additive constant. 
1649: The three examples are drawn from the regimes A, B and C  
1650: of Fig.\,5, defined by the width of the broad component of \hb\ and 
1651: electron density.
1652: \label{fig1}}
1653: \end{figure}
1654: 
1655: \begin{figure}
1656: \plotone{f2.eps}
1657: \caption{Examples of the decompsoition of the H$\beta+$\oiii\ and 
1658: H$\alpha+$\nii\ emission-line profiles. 
1659: F$_\lambda$ in units of 10$^{-17}$~erg~s$^{-1}$~cm$^{-2}$~\AA$^{-1}$
1660: is plotted against wavelength in \AA. 
1661: The narrow emission-lines are fit by Gaussian profiles, while the 
1662: broad Balmer components are fit by either Gaussian (G) or 
1663: Lorentzian (L) profiles.
1664: For Balmer lines the narrow and broad components are shown, while for 
1665: the forbidden lines the resulting narrow line profiles are plotted. 
1666: To illustrate the difference between the data and the fit, the resulting
1667: residuals are shown at the bottom of each panel. 
1668: For clarity, the residual spectrum is offset by an additive constant.
1669: The three examples are drawn from regimes A, B and C of Fig.\,5, 
1670: defined by the widths of \hbb\ and electron density.
1671: \label{fig2}}
1672: \end{figure}
1673: 
1674: \begin{figure}
1675: \plotone{f3.eps}
1676: \caption{Distributions of $i$-band absolute magnitude (left) and redshift
1677: (right). The open histograms plot the NLS1 galaxies, and the shaded histograms
1678: plot the BLS1 galaxies of our samples. 
1679: \label{fig3}}
1680: \end{figure}
1681: 
1682: \begin{figure}
1683: \plotone{f4.eps}
1684: \caption{
1685: Ratio of \oiii\ $\lambda5007$ to the total H$\beta$ flux vs. R4570, 
1686: the ratio of the flux of the \feii\ complex between $\lambda$4434 and 
1687: $\lambda$4684 to that of H$\beta$ for NLS1 galaxies (filled circles) 
1688: and BLS1 galaxies (open circles).
1689: The dashed line and dot-dashed line 
1690: mark the mean R4570 for NLS1 galaxies and BLS1 galaxies, 
1691: respectively. The large filled circles (pink) represent the {\em low-density} 
1692: objects from regime\,A of Fig.\,5. 
1693: \label{fig4}}
1694: \end{figure}
1695: 
1696: \begin{figure}
1697: \plotone{f5.eps}
1698: %
1699: \vspace{-1.2cm}
1700: %
1701: \caption{$\bold{Top}$: \sii\ $\lambda6716/\lambda6731$ intensity ratio
1702: vs. FWHM of the broad component of \hb\ for our sample. 
1703: Filled and open symbols represent the broad \hb\ components modeled by 
1704: Gaussian (G) and Lorentzian (L) profiles, respectively. 
1705: Squares correspond to QSOs; triangles 
1706: to  Seyfert~1s. The median error bar is given at the upper left corner. 
1707: The vertical dot-dashed line marks the boundary between NLS1 galaxies
1708: and BLS1 galaxies in terms of FWHM(\hbb). 
1709: Histograms of the \sii\ $\lambda6716/\lambda6731$ ratio 
1710: of NLS1 galaxies and BLS1 galaxies are plotted 
1711: in the left and right panels, respectively. 
1712: $\bold{Bottom}$: Electron density obtained from the 
1713: \sii\ $\lambda6716/\lambda6731$ ratio in dependence of FWHM of the broad 
1714: component of \hb. 
1715: Symbols as in the top panel.
1716: The arrow points to the location of one outlier which is off the plot.
1717: The dot-dashed lines distinguish areas populated by: 
1718: (A) NLS1 galaxies with small width of \hbb\ and low density; 
1719: (B) NLS1 galaxies small width of \hbb\ and high density; and 
1720: (C) BLS1 galaxies with large width of \hbb\ and high density. 
1721: Median error bars of each regime are given.
1722: Distributions of the electron density of NLS1 galaxies and BLS1 galaxies 
1723: are plotted 
1724: in the left and right panels, respectively.
1725: \label{fig5}}
1726: \end{figure}
1727: 
1728: \begin{figure}
1729: \plotone{f6.eps}
1730: \caption{\oi$\lambda6300$/H$\alpha_n$ ratio plotted against 
1731: FWHM of the broad component of \hb. 
1732: Filled triangles are data 
1733: derived by fitting the broad Balmer components with Gaussian profiles, 
1734: while open triangles are data derived by modeling the broad Balmer 
1735: components with Lorentzian profiles. Large symbols represent the 
1736: {\em low-density} objects from regime\,A of Fig.\,5. 
1737: The \oi$\lambda6300$/H$\alpha_n$ ratio 
1738: is correlated with FWHM(\hbb) ($r_{\rm s}=0.41$, $P_{\rm null}<10^{-3}$). 
1739: The solid line shows the ordinary 
1740: least-square regression fit to filled triangles, 
1741: the dotted line the fit to open triangles. 
1742: The dot-dashed line marks the boundary between NLS1 galaxies and BLS1 galaxies.
1743: \label{fig6}}
1744: \end{figure}
1745: 
1746: \begin{figure}
1747: \plotone{f7.eps}
1748: \caption{Histograms showing the \oii\ $\lambda3727$/\oiii\ $\lambda5007$ 
1749: distribution prior to reddening correction (left) and after reddening
1750: correction (right). The open histograms refer to NLS1 galaxies, the shaded 
1751: histograms to BLS1 galaxies.
1752: \label{fig7}}
1753: \end{figure}
1754: 
1755: \begin{figure}
1756: \plotone{f8.eps}
1757: \caption{
1758: Electron density plotted against R4570 for NLS1 galaxies (filled circles) 
1759: and BLS1 galaxies
1760: (open circles). The large filled circles represent the {\em low-density}
1761: objects from regime\,A of Fig.\,5. 
1762: The sources which are off the plot, are indicated by arrows. 
1763: The density is anti-correlated with R4570
1764: ($r_{\rm s}=-0.47$, $P_{\rm null}<10^{-4}$).
1765: \label{fig8}}
1766: \end{figure}
1767: 
1768: \begin{figure}
1769: \plotone{f9.eps}
1770: \caption{Electron density plotted against the Eddington ratio, 
1771: ${\rm L_{bol}/L_{Edd}}$. Filled symbols correspond to NLS1 galaxies; 
1772: open symbols to BLS1 galaxies.
1773: Symbols are the same as in Fig.\,8.
1774: A trend of decreasing electron density with 
1775: increasing Eddington ratio can be 
1776: seen ($r_{\rm s}=-0.42$, $P_{\rm null}=10^{-4}$). 
1777: \label{fig9}}
1778: \end{figure}
1779: 
1780: \begin{figure}
1781: \plotone{f10.eps}
1782: \caption{Electron density plotted against 
1783: the blueshift of the blue wing of \oiii. 
1784: Symbols are the same as in Fig.\,8.
1785: \label{fig10}}
1786: \end{figure}
1787: 
1788: \end{document} 
1789: