0807.2060/ms.tex
1: % version 1; submitted to ApJ (Letters); June 12, 2008
2: % version 2; resubmitted to ApJ (Letters); Jul 09, 2008
3: 
4: %\documentclass[12pt,preprint]{aastex}
5: \documentclass{emulateapj}
6: %\documentclass[preprint2]{aastex}
7: %\documentclass[manuscript]{aastex}
8: %\documentclass[preprint2,longabstract]{aastex}
9: \usepackage{apjfonts}
10: 
11: \newcommand{\ang}{\AA}
12: \newcommand{\civ}{{\sc C iv}}
13: \newcommand{\feii}{Fe {\sc ii}}
14: \newcommand{\hb}{${\rm H}\beta$}
15: \newcommand{\hbi}{${\rm H}\beta_{\rm IC}$}
16: \newcommand{\hbb}{${\rm H}\beta_{\rm BC}$}
17: \newcommand{\hbn}{${\rm H}\beta_{\rm NC}$}
18: \newcommand{\hbvb}{${\rm H}\beta_{\rm VBC}$}
19: \newcommand{\kms}{$\rm km~s^{-1}$}
20: \newcommand{\ks}{$\chi^2$}
21: \newcommand{\mgii}{Mg {\sc ii}}
22: \newcommand{\msun}{$M_{\odot}$}
23: \newcommand{\myemail}{chenhu@bao.ac.cn}
24: \newcommand{\oii}{{\sc [O ii]}}
25: \newcommand{\oiii}{{\sc [O iii]}}
26: 
27: \slugcomment{Accepted for publication in ApJ (Letters).}
28: 
29: \shorttitle{Intermediate-line Region in Quasars}
30: \shortauthors{Hu et al.}
31: 
32: \begin{document}
33: 
34: \title{\hb\ Profiles in Quasars:
35: Evidence for an Intermediate-line Region}
36: 
37: \author{Chen Hu\altaffilmark{1,2,3}, Jian-Min Wang\altaffilmark{2,4},
38: Luis C. Ho\altaffilmark{5}, Yan-Mei Chen\altaffilmark{2,3}, 
39: Wei-Hao Bian\altaffilmark{2,6} and Sui-Jian Xue\altaffilmark{1}}
40: 
41: \altaffiltext{1}{National Astronomical Observatories of China,
42: Chinese Academy of Sciences, Beijing 100012, China; \myemail}
43:                                                                                 
44: \altaffiltext{2}{Key Laboratory for Particle Astrophysics,
45: Institute of High Energy Physics, Chinese Academy of Sciences,
46: Beijing 100049, China.}
47:                                                                                 
48: \altaffiltext{3}{Graduate University of the Chinese Academy of Sciences,
49: Beijing 100049, China.}
50: 
51: \altaffiltext{4}{Theoretical Physics Center for Science Facilities (TPCSF), 
52: Chinese Academy of Sciences, China.}
53: 
54: \altaffiltext{5}{The Observatories of the Carnegie Institution of Washington,
55: 813 Santa Barbara Street, Pasadena, CA 91101, USA.}
56: 
57: \altaffiltext{6}{Department of Physics and Institute of Theoretical Physics,
58: Nanjing Normal University, Nanjing 210097, China.}
59: 
60: \begin{abstract}
61: We report on a systematic investigation of the \hb\ and \feii\ emission lines 
62: in a sample of 568 quasars within $z < 0.8$ selected from the Sloan Digital 
63: Sky Survey.  The conventional broad \hb\ emission line can be decomposed into 
64: two components---one with intermediate velocity width and another with very 
65: broad width. The velocity shift and equivalent width of the intermediate-width 
66: component do not correlate with those of the very broad component of \hb, but 
67: its velocity shift and width do resemble \feii.  Moreover, the width of the very broad component is 
68: roughly 2.5 times that of the intermediate-width component. These 
69: characteristics 
70: strongly suggest the existence of an intermediate-line region, whose 
71: kinematics seem to be dominated by infall, located at the outer portion of the 
72: broad-line region.  
73: \end{abstract}
74: 
75: \keywords{galaxies: nuclei --- (galaxies:) quasars: emission lines ---
76: (galaxies:) quasars: general --- galaxies: Seyfert --- line: profiles}
77: 
78: \section{Introduction}
79: 
80: The geometry and kinematics of the broad-line region (BLR) in active galactic
81: nuclei (AGNs) have been studied for about three decades but the details are
82: far from well understood.  It is widely accepted that the BLR is stratified:
83: high-ionization lines originate from small radii and low-ionization lines
84: arise further out \citep{collin88}. This stratification picture is
85: supported by the results of reverberation mapping, which show that lines of 
86: different ionization have different lags (e.g., \citealt{peterson99}).  
87: The dependence of the systemic velocities of the emission lines on ionization
88: \citep[e.g.,][]{gaskell82,sulentic00a,richards02,shang07} suggests that the 
89: BLR may originate from a wind and a disk \citep[e.g.,][and references 
90: therein]{leighly04a,leighly04b}.  However, the profiles of the broad emission 
91: lines often contain multiple velocity components, suggesting that the
92: structure of the BLR may be more complex than can be described by a 
93: simple stratification or wind $+$ disk model.
94: 
95: It has been known that the broad \hb\ line profiles are generally not well
96: described by a single Gaussian. Two Gaussians \citep[e.g.,][]{netzer07} or a
97: Gauss-Hermite function \citep[e.g.,][]{salviander07} are often used.
98: Additionally, the profiles show great diversity from object to object.  Sources
99: with narrower \hb\ lines tend to have stronger line wings, while those 
100: with broader \hb\ lines are dominated by the line core 
101: \citep[e.g.,][]{sulentic02}. Some sources have asymmetric and shifted \hb\ 
102: profiles, suggesting that the BLR has a structure more complex than a single 
103: virialized component. The \hb\ profile of OQ 208, for example, has an 
104: additional redshifted \hb\ component of intermediate width that closely
105: resembles the kinematics of \feii\ \citep{marziani93}. Additional evidence 
106: that the BLR contains two or more kinematically distinct components comes 
107: from differential variability between the line core and wing
108: \citep[e.g.,][]{ferland90,peterson00}.  The existence of an
109: intermediate-line region and a very broad-line region for \hb\ emission was
110: suggested by some previous studies \citep[e.g.,][]{brotherton96,sulentic00b}.
111: 
112: In a recent spectral decomposition of a large sample of quasars selected from 
113: the Sloan Digital Sky Survey (SDSS), Hu et al. (2008, hereinafter Paper I) 
114: find that the majority of quasars show \feii\ emission that is both redshifted 
115: and narrower than \hb.  Moreover, the magnitude of the \feii\ redshift 
116: correlates inversely with the Eddington ratio.  These characteristics suggest 
117: that \feii\ originates from an exterior portion of the BLR, whose dynamics may 
118: be dominated by infall.  These findings offer fresh insights into the 
119: structure of the BLR. 
120: 
121: In light of the trends associated with \feii\ emission, we expect that a 
122: portion of the \hb-emitting gas may be related to an inflowing component too.
123: In this Letter, we systematically study the \hb\ profiles of SDSS quasars to
124: try to answer two questions: do all quasars have an intermediate-width \hb\ 
125: component similar to OQ 208, and, if so, is this component also associate with 
126: the \feii\ emission?  We show that the conventional broad \hb\ line actually 
127: consists of two kinematically linked components, one of which originates
128: from the same region that emits \feii.
129: 
130: \section{Sample and Data Analysis}
131: 
132: Our sample is selected from the SDSS Fifth Data Release (\citealt{adelman07}) 
133: quasar catalog \citep{schneider07}.  We choose objects with redshifts $z<0.8$ 
134: to ensure that \oiii\ $\lambda$5007 lies within the SDSS spectral coverage. We 
135: also require a signal-to-noise ratio (S/N) $>$ 10 in the restframe wavelength 
136: range 4430--5550 \ang\ (covering \hb, \oiii, and the most prominent features 
137: of optical \feii\ emission) and that no more than 1/3 of the pixels are masked 
138: by the
139: SDSS pipeline in this region. 7601 quasars satisfy these criteria.  The 
140: spectral analysis, whose details are described in Paper I, involves fitting a 
141: continuum model in a set of windows devoid of strong emission lines that 
142: consists of (1) a single power law, (2) Balmer continuum supplemented with 
143: high-order Balmer emission lines, and (3) a pseudo-continuum due to blended
144: \feii\ emission.  The full width at half maximum (FWHM) and shift of \feii\ are
145: measured from the continuum decomposition. After subtracting the continuum 
146: model, the \hb\ line is decomposed into narrow (\hbn) and broad (\hbb) 
147: components.  \hbn\ is forced to have the same profile as \oiii, a shift of up 
148: to 600 \kms\ relative to \oiii, and an intensity constrained to lie between 
149: 1/20 and 1/3 of that of \oiii. \hbb\ is modeled using a Gauss-Hermite function 
150: \citep{vandermarel93}, whose best fit yields FWHM(\hbb).
151: The rest frame is defined by the peak of the \oiii\ $\lambda$5007 (see Paper
152: I for more details).
153: 
154: \begin{figure}
155:   \includegraphics[angle=-90,width=0.47\textwidth]{f1.eps}
156: \caption{\footnotesize ({\it a}) Variation of \hb\ profile with line width. 
157: $P$ denotes the probability that the double-Gaussian model cannot improve 
158: the fit.  A high value of $P$ means that the \hb\ line can be adequately 
159: described by a single Gaussian, whereas a low value of $P$ indicates that 
160: two Gaussians are needed.  The dashed line is $1-P = 99.73 \%$. The sources 
161: below this line need two Gaussians to fit their broad \hb\ lines. ({\it b}) 
162: The critical EW ratio of \hbi/\hbvb\ for requiring the double-Gaussian model
163: as a function of signal-to-noise ratio.} 
164: \label{fig-hbwfp}
165: \end{figure}
166: 
167: Although the Gauss-Hermite function provides a good and convenient mathematical 
168: description of the \hbb\ profile, for the purpose of this
169: Letter---investigating whether there is an intermediate-width \hb\ component
170: associated with \feii\ emission---we use the following simpler method to
171: fit \hbb. First, we use only one Gaussian to model the broad \hb\ line
172: (hereinafter the single-Gaussian model).  Then we fit broad \hb\ with two 
173: Gaussians (hereinafter the double-Gaussian model), an intermediate-width 
174: component (\hbi) and a very broad component (\hbvb).  In both models, \hbn\ is 
175: fitted in the same way as in Paper I.  First, we compare the \ks\ of the
176: double-Gaussian model and the Gauss-Hermite. There are 630 sources with 
177: \ks\ 20\% larger in the double-Gaussian model than in the Gauss-Hermite model
178: \citep{hao05}; this means that the double-Gaussian fit is still inadequate for 
179: these sources\footnote{From a statistical view for the whole sample, however,
180: this does not mean that the double-Gaussian model is worse than the 
181: Gauss-Hermite model. There are 510 sources with \ks\ 20\% smaller for the 
182: double-Gaussian model than for the Gauss-Hermite model, and for most sources 
183: both models are equally good.}.
184: Then, for the sources that can be well fitted by the
185: double-Gaussian model, we compare the reduced \ks\ of the double-Gaussian
186: and single-Gaussian model, and use the F-test \citep[Chapter 12.1]{lupton93}
187: to calculate how significantly the double-Gaussian model improves the fit.
188: Whether a source needs two Gaussians or not depends on its \hbb\ width.
189: Figure \ref{fig-hbwfp}{\it a} shows the probability $P$ that the
190: double-Gaussian model cannot improve the fit, as a function of FWHM(\hbb).
191: The 2435 sources below the dashed line ($\sim 32\%$ of the whole sample)
192: require two Gaussians at a significance greater than 3 $\sigma$.  For
193: sources with FWHM(\hbb) $\gtrsim$ 5000 \kms, one Gaussian is enough to
194: describe the profile of \hbb.  This is consistent with \citet{collin06}, who
195: find that the ratio FWHM/$\sigma$ ($\sigma$ is the line dispersion) of \hbb\
196: depends on its line width (see their Fig. 3). For sources with broader \hbb,
197: this ratio is close to 2.35, the value for a single Gaussian. As discussed
198: later, this means that broader sources have a weaker \hbi\ component.
199: 
200: We use Monte Carlo simulations to determine the detection threshold of \hbi.
201: We generate artificial \hb\ lines using two components [FWHM(\hbvb) is set to
202: 2.5 FWHM(\hbi) as shown in Fig. \ref{fig-hbihbvb}], fit it using a single 
203: and a double Gaussian, and then calculate the probability $P$ as before. The 
204: EW ratio of \hbi/\hbvb\ is increased progressively until $1-P$ exceeds 
205: 99.73\%. (All EW measurements refer to the continuum at 5100 \ang.) 
206: The critical EW ratio of \hbi/\hbvb\ depends on S/N, as shown in 
207: Figure \ref{fig-hbwfp}{\it b}. In order to detect an \hbi\ component with an 
208: EW that is 10\% larger than the EW of \hbvb, the S/N should be larger than 
209: $22.5$.  Thus, in the analysis below, we remove all sources with (1) 
210: S/N $< 22.5$, (2) \ks\ of double-Gaussian 20\% larger than the \ks\ of
211: Gauss-Hermite and $1-P < 99.73$\%, and (3) EW(Fe) $<$ 25 \ang\ (to insure
212: that the \feii\ measurements are as reliable as those in Paper I).
213: These three cuts left 1499, 811, and 568 sources, respectively. Note that
214: the fraction of sources that require a double-Gussian model (811/1499 $\approx$
215: 54\%) is larger than the fraction when the S/N threshold is 10 (2435/7601 
216: $\approx$ 32\%), consistent with the simulations above: fainter \hbi\ can
217: be detected with higher S/N. The analysis below focuses on the most
218: stringent subset of 568 sources ($\sim 38\%$ of the sources with S/N
219: $>$ 22.5). The reduced \ks\ for the double-Gaussian model fit has a median
220: value of 1.004.
221: 
222: To constrain the relative strengths of \feii\ emission associated with the
223: very broad and intermediate-width components, we generated simulated spectra 
224: using two \feii\ components, and then fit them using only one component, whose 
225: width is fixed to that of the input intermediate-width component. The EW of 
226: the input very broad component is increased progressively until the reduced \ks\ 
227: of the fit exceeds 1 $\sigma$ from the expect value. For a typical \feii\ EW 
228: of 75 \ang\ and S/N $\approx$ 25 in our final sample, the flux ratio of \feii\ 
229: between the input very broad component and intermediate-width component is 
230: $\sim$0.3. This is the upper limit of the \feii\ emission coming from the 
231: very broad component. 
232: 
233: \begin{figure*}
234:   \includegraphics[angle=-90,width=0.98\textwidth]{f2.eps}
235:   \caption{\footnotesize Examples of \feii\ measurement and emission-line
236:   fitting for ({\it a}) SDSS J094603.94$+$013923.6, ({\it b}) SDSS
237:   J092008.22$+$032245.4, and ({\it c}) SDSS J103859.58$+$422742.2.  For each
238:   source, the top panel shows multi-Gaussian fitting for \hb\ and \oiii.
239:   \hbvb\ is in green and \hbi\ is in magenta. The two blue dashed lines mark
240:   the rest-frame wavelength of \hb\ and \oiii\ $\lambda$5007. The magenta
241:   dotted line is the position of \hb\ at the same velocity as \feii.  Note the 
242:   consistency between the \hbi\ peak and the dotted line.
243:   \hbn\ and \oiii\ are in blue. The red line is the sum of each component. The
244:   bottom panel shows the emission-line spectrum after subtracting the power-law 
245:   continuum. The green portions of the spectrum denote the windows for fitting 
246:   the \feii\ emission, whose model is given in red.  The blue dashed line marks 
247:   the peak of \feii\ $\lambda$4924 at zero velocity shift. 
248:   }
249:   \label{fig-example}
250: \end{figure*}
251: 
252: Figure \ref{fig-example}
253: shows the emission-line fitting and \feii\ emission measurement for three
254: typical sources. Example {\it a} is an extreme case that has a large \feii\
255: redshift (1533$\pm$24 \kms) and \hb\ with an isolated red peak. The
256: \feii\ in example {\it b} is moderately redshifted (590$\pm$98 \kms), and its
257: \hb\ core shows only a strong red asymmetry rather than another peak. A more 
258: common situation is seen in example {\it c}, in which \feii\ and \hbi\ have no 
259: shift and the \hb\ profile is symmetrical.  Table \ref{tab-all} lists the 
260: measurements of the three sources shown in Figure \ref{fig-example}.
261: 
262: \begin{table*}
263: \begin{center}
264:   \caption{Line Measurements
265:   \label{tab-all}}
266: {\scriptsize
267: \begin{tabular}{cccccccccccccc}
268:   \hline\hline
269:  & &\multicolumn{3}{c}{\feii} & & \multicolumn{3}{c}{\hbi} & & 
270: \multicolumn{3}{c}{\hbvb} & \\
271: \cline{3-5} \cline{7-9} \cline{11-13}
272: SDSS Name & $z$ & EW & FWHM & Shift & & EW & FWHM & Shift & &
273: EW & FWHM & Shift & \ks \\
274:  & &(\ang)&(\kms)&(\kms)& &(\ang)&(\kms)&(\kms)& &(\ang)&(\kms)&(\kms)& \\ 
275:  \hline
276: 094603.94$+$013923.6 & 0.220 & 60$\pm$1 & 1543$\pm$55 & 1533$\pm$24 & &
277: 54$\pm$1 & 1428$\pm$18 & 1754$\pm$7 & & 
278: 125$\pm$1 & 5730$\pm$ 58 & 400$\pm$25 & 3.506 \\
279: 092008.22$+$032245.4 & 0.334 & 45$\pm$2 & 2548$\pm$251 & 590$\pm$98 & &
280: 27$\pm$2 & 2281$\pm$103 & 589$\pm$42 & & 
281: 42$\pm$2 & 6924$\pm$332 & $-$195$\pm$109 & 0.715 \\
282: 103859.58$+$422742.2 & 0.221 & 72$\pm$1 & 1297$\pm$40 & $-$10$\pm$18 & &
283: 13$\pm$1 & 1259$\pm$71 & $-$16$\pm$26 & &
284: 44$\pm$1 & 3893$\pm$87 & $-$279$\pm$29 & 1.361 \\ \hline
285: \end{tabular}
286: }
287: \parbox{0.9\textwidth}{
288: \tablecomments{Table \ref{tab-all} is available in its entirety 
289:   electronically.}
290: }
291: %\vspace{0.5cm}
292: \end{center}
293: \end{table*}
294: 
295: \section{Results and Discussions}
296: 
297: \begin{figure} 
298:   \includegraphics[angle=-90,width=0.47\textwidth]{f3.eps}
299:   \caption{\footnotesize Correlations between ({\it a}) \feii\ and \hbi\
300:   shifts and ({\it b}) \feii\ and \hbi\ widths. The solid diagonal lines 
301:   denote that \feii\ and \hbi\ have the same shifts and widths. The contours
302:   in panel ({\it a}) show the density of the data points.} 
303:   \label{fig-hbife} 
304: \end{figure}
305: 
306: \begin{figure}
307:   \includegraphics[angle=-90,width=0.47\textwidth]{f4.eps}
308:   \caption{\footnotesize Plots of ({\it a}) \feii\ vs. \hbvb\ shifts and
309:   ({\it b}) \feii\ vs. \hbvb\ widths. The two lines have different shifts
310:   and widths.}
311:   \label{fig-hbvbfe}
312: \end{figure}
313: 
314: We find that \hbi\ and \feii\ emission have similar kinematics. The similarity 
315: can be seen not only in individual sources, but also statistically for the 
316: whole sample.  From examples {\it a} to {\it c} in Figure \ref{fig-example}, the
317: \hb\ profiles change progressively while the \feii\ shifts become lower and
318: lower. Figure \ref{fig-hbife}{\it a} shows a strong correlation between \hbi\
319: and \feii\ shifts. Pearson's correlation coefficient $r_{\rm P}$ is 0.22,
320: and the probability $P$ of a chance correlation is $< 1\times10^{-5}$. 
321: \hbi\ and \feii\ have approximately the
322: same shifts. The widths of \hbi\ and \feii\ are also well correlated and
323: roughly equal (Fig. \ref{fig-hbife}{\it b}); $r_{\rm P}$ = 0.48 and $P <
324: 1\times10^{-5}$. Except for some sources to the upper left of the solid lines
325: whose errors are large, the majority of sources follow the relation that \hbi\ and
326: \feii\ have the same shifts and widths. By contrast, \feii\ and \hbvb\ have
327: different shifts and widths (Fig. \ref{fig-hbvbfe}). We also find
328: that the kinematic
329: connection between \hbi\ and \hbvb\ is complicated.  No correlation between
330: \hbi\ and \hbvb\ shifts is seen (Fig. \ref{fig-hbihbvb}{\it a}), but the
331: \hbi\ and \hbvb\ widths are strongly linked, such that FWHM(\hbvb) $\approx$
332: 2.5 FWHM(\hbi) (Fig. \ref{fig-hbihbvb}{\it b}).
333: 
334: \begin{figure}
335:   \includegraphics[angle=-90,width=0.47\textwidth]{f5.eps}
336:   \caption{\footnotesize Plots of ({\it a}) \hbvb\ vs. \hbi\ shifts and
337:   ({\it b}) \hbvb\ vs. \hbi\ widths. The two components have different shifts
338:   and widths. The solid line in panel ({\it b}) denotes FWHM(\hbvb) $= 2.5$
339:   FWHM(\hbi).}
340:   \label{fig-hbihbvb}
341: \end{figure}
342: 
343: The observational results described above suggest a scenario in which the
344: conventional BLR consists of two components---an intermediate-line region 
345: (ILR) and a very broad-line region (VBLR).
346: If both regions are virialized, so that $R \propto v^{-2}$, then the ILR is 
347: about $2.5^2$ or 6.25 times farther from the center than the VBLR.  Because of 
348: its redshift, the kinematics of the ILR may be dominated by infall. \hb\ 
349: emission emerges from both the ILR and VBLR, while most of the \feii\ emission 
350: comes from the ILR.  
351: 
352: The ILR and VBLR defined in this paper are essentially similar to those
353: described in \citet{corbin95}, \citet{brotherton96}, \citet{sulentic00b},
354: and \citet{zhu08} but
355: differ from those in \citet{brotherton94} or \citet{sulentic99}, 
356: which refer to the \civ-emitting region.  Note that the ILR and VBLR of the 
357: \hb-emitting region are distinct from those of the \civ-emitting region
358: because the \civ\ ILR usually has the systemic redshift while the \civ\ VBLR 
359: shows a blueshift. 
360: 
361: \begin{figure}
362:   \includegraphics[angle=-90,width=0.47\textwidth]{f6.eps}
363:   \caption{\footnotesize Plots of ({\it a}) \hbvb\ vs. \hbi\ EWs and
364:   ({\it b}) \feii\ vs. \hbi\ EWs. No correlations are found.}
365:   \label{fig-ew}
366: \end{figure}
367: 
368: The lack of correlation between the EWs of \hbi\ and \hbvb\ (Fig. 
369: \ref{fig-ew}{\it a}) strongly suggests that the two components are emitted 
370: from different regions. If both are photoionized, they must have different 
371: covering factors. The relative strength between \hbi\ and \hbvb\ determines 
372: the final \hb\ profiles.  In Figure \ref{fig-hbwfp}, sources broader than 
373: $\sim$5000 \kms\ can be fitted well using one Gaussian. This trend can be 
374: interpreted in our two-component BLR scenario. It is well known that sources 
375: with broad \hb\ tend to have weak \feii/\hb\ 
376: \citep[e.g.,][]{bg92,sulentic00a}.  The ILR is weak in these systems because
377: \feii\ is weak.  Their profiles show little deviation from a single Gaussian 
378: under the typical S/N level of SDSS spectra. The composite spectra of sources 
379: with large \feii\ redshifts, on the other hand, do show red asymmetry in the 
380: \hb\ profiles (see Fig. 13 of Paper I).  The variation in the relative 
381: strength of the \hbvb\ and \hbi\ components in different sources reflects
382: the competition between the two components,
383: although they apparently do so in such a manner that their kinematics
384: remained coupled.
385: 
386: It is of interest to note that the strengths of \hbi\ and \feii\
387: are not correlated (Fig. \ref{fig-ew}{\it b}). The wide range of 
388: \feii/\hbi\ ratios reflects either the complexity of the excitation mechanism
389: of \feii\ emission \citep[e.g.,][and references therein]{baldwin04} or 
390: large variations in quasar metallicities \citep[e.g.,][]{netzer07}.
391: 
392: 
393: \section{Summary}
394: 
395: We have studied the profiles of the \hb\ emission line using a large sample of
396: quasars selected from SDSS. Comparing the \hb\ profiles with the properties of 
397: \feii\ emission given in Paper I, we deduced the existence of two \hb\ 
398: emission regions---an intermediate-line region and a very broad-line region. 
399: The observational evidence can be summarized as follows:
400: \begin{enumerate}
401: \item{The velocity shifts and widths of the \hb\ intermediate-width component 
402: are approximately the same as those of \feii, indicating that they originate
403: from the same region.  However, the \feii/\hbi\ ratios vary greatly from 
404: object to object, reflecting variations in either metal abundance or 
405: excitation conditions in the \feii-emitting region.}
406: \item{The velocity width of the very broad component of \hb\ is roughly 
407: 2.5 times larger than that of the intermediate-width component, but no 
408: correlation exists between their radial velocities.  The equivalent widths 
409: of the two components are also unrelated, suggesting that they have very 
410: different covering factors and geometry.  The conventional BLR seems to 
411: consist of two different, physically distinct regions.  We suggest 
412: that the intermediate-width component of \hb\ and \feii\ trace an infalling 
413: region in the outskirts of the BLR, likely located in between the molecular 
414: torus and the accretion disk.}
415: \end{enumerate}
416: 
417: %\bigskip
418: The properties of the \hb-emitting region discussed here, in conjunction with 
419: those of \feii\ emission summarized in Paper I, offer important new 
420: constraints on models of the broad emission-line regions.  This will be 
421: the subject of a forthcoming paper.
422: 
423: \acknowledgments
424: 
425: We appreciate extensive discussions among the members of the IHEP AGN group. 
426: We thank an anonymous referee for helpful comments.
427: This research is supported by NSFC via NSFC-10325313, 10733010 and
428: 10521001, and by CAS via KJCX2-YW-T03.
429: 
430: \begin{thebibliography}{}
431: \bibitem[Adelman-McCarthy et al.(2007)]{adelman07}Adelman-McCarthy, J., et al. 2007, \apjs, 172, 634
432: \bibitem[Baldwin et al.(2004)]{baldwin04}Baldwin, J. A., Ferland, G. J., Korista, K. T., Hamann, F., \& LaCluyz\'e, A. 2004, \apj, 615, 610
433: \bibitem[Boroson \& Green(1992)]{bg92}Boroson, T. A., \& Green, R. F. 1992, \apjs, 80, 109
434: \bibitem[Brotherton(1996)]{brotherton96}Brotherton, M. S. 1996, \apjs, 102, 1
435: \bibitem[Brotherton et al.(1994)]{brotherton94}Brotherton, M. S., Wills, B. J., Francis, P. J., \& Steidel, C. C. 1994, \apj, 430, 495
436: \bibitem[Collin et al.(2006)]{collin06}Collin, S., Kawaguchi, T., Peterson, B. M., \& Vestergaard, M., 2006, \aap, 456, 75
437: \bibitem[Collin-Souffrin \& Lasota(1988)]{collin88}Collin-Souffrin, S., \& Lasota, J.-P. 1988, \pasp, 100, 1041
438: \bibitem[Corbin(1995)]{corbin95}Corbin, M. R. 1995, \apj, 447, 496
439: \bibitem[Ferland et al.(1990)]{ferland90}Ferland, G. J., Korista, K. T., \& Peterson, B. M., 1990, \apj, 363, L21
440: \bibitem[Gaskell(1982)]{gaskell82}Gaskell, C. M. 1982, \apj, 263, 79
441: \bibitem[Hao et al.(2005)]{hao05}Hao, L., et al. 2005, \aj, 129, 1783
442: \bibitem[Hu et al.(2008)]{hu08}Hu, C., et al. 2008, ApJ, in press (arXiv:0807.2059)~(Paper I)
443: \bibitem[Leighly(2004)]{leighly04b}Leighly, K. M., 2004, \apj, 611, 125
444: \bibitem[Leighly \& Moore(2004)]{leighly04a}Leighly, K. M., \& Moore, J. R., 2004, \apj, 611, 107
445: \bibitem[Lupton(1993)]{lupton93}Lupton, R. H. 1993, Statistics in Theory and Practice (Princeton: Princeton Univ. Press)
446: \bibitem[Marziani et al.(1993)]{marziani93}Marziani, P., Sulentic, J. W., Calvani, M., P\'erez, E., Moles, M., \& Penston, M. V. 1993, \apj, 410, 56
447: \bibitem[Netzer \& Trakhtenbrot(2007)]{netzer07}Netzer, H., \& Trakhtenbrot, B. 2007, \apj, 654, 754
448: \bibitem[Peterson et al.(2000)]{peterson00}Peterson, B. M., et al. 2000, \apj, 542, 161
449: \bibitem[Peterson \& Wandel(1999)]{peterson99}Peterson, B. M., \& Wandel, A. 1999, \apj, 521, L95
450: \bibitem[Richards et al.(2002)]{richards02}Richards, G. T., et al. 2002, \aj, 124, 1
451: \bibitem[Salviander et al.(2007)]{salviander07}Salviander, S., Shields, G. A., Gebhardt, K., \& Bonning, E. W. 2007, \apj, 662, 131
452: \bibitem[Schneider et al.(2007)]{schneider07}Schneider, D. P., et al. 2007, \aj, 134, 102
453: \bibitem[Shang et al.(2007)]{shang07}Shang, Z., Wills, B. J., Wills, D., \& Brotherton, M. S. 2007, \aj, 134, 294
454: \bibitem[Sulentic \& Marziani(1999)]{sulentic99}Sulentic, J. W., \& Marziani, P. 1999, \apj, 518, L9
455: \bibitem[Sulentic et al.(2000a)]{sulentic00a}Sulentic, J. W., Marziani, P., \& Dultzin-Hacyan, D. 2000a, \araa, 38, 521
456: \bibitem[Sulentic et al.(2002)]{sulentic02}Sulentic, J. W., Marziani, P., Zamanov, R., Bachev, R., Calvani, M., \& Dultzin-Hacyan, D. 2002, \apj, 566, L71
457: \bibitem[Sulentic et al.(2000b)]{sulentic00b}Sulentic, J. W., Marziani, P., Zwitter, T., Dultzin-Hacyan, D., \& Calvani, M. 2000b, \apj, 545, L15
458: \bibitem[van der Marel \& Franx(1993)]{vandermarel93}van der Marel, R. P., \& Franx, M. 1993, \apj, 407, 525
459: \bibitem[Zhu et al.(2008)]{zhu08}Zhu, L., Zhang, S. N., \& Tang, S. M. 2008, arXiv:0807.3992
460: \end{thebibliography}
461: 
462: \end{document}
463: