1: \documentstyle[12pt,aaspp4]{article}
2: \begin{document}
3:
4: \title{Evidence for an accelerating wind as the broad-line region in NGC 3516}
5:
6: \author{
7: J.B. Hutchings\altaffilmark{1},
8: G. A. Kriss\altaffilmark{2,3},
9: R. F. Green\altaffilmark{4},
10: M. Brotherton\altaffilmark{4},
11: M. E. Kaiser\altaffilmark{3},
12: A. P. Koratkar\altaffilmark{2},
13: W. Zheng\altaffilmark{3}
14: }
15:
16: \altaffiltext{1}{Herzberg Insitute of Astrophysics,
17: National Research Council of Canada,\\ Victoria, B.C. V8X 4M6, Canada}
18: \altaffiltext{2}{Space Telescope Science Institute,
19: 3700 San Martin Drive, Baltimore, MD 21218; gak@stsci.edu}
20: \altaffiltext{3}{Center for Astrophysical Sciences, Department of Physics and
21: Astronomy, The Johns Hopkins University, Baltimore, MD 21218--2686}
22: \altaffiltext{4}{Kitt Peak National Observatory,
23: National Optical Astronomy Observatories, P.O. Box 26732,
24: 950 North Cherry Ave., Tucson, AZ, 85726-6732}
25:
26:
27:
28:
29:
30: \begin{abstract}
31:
32: Spectroscopic data in wavelengths 900--3000 \AA\ have been obtained in a
33: low flux state of the nucleus of the Seyfert 1 galaxy NGC~3516. The line
34: profiles show P Cygni characteristics, particularly in O VI $\lambda$1032, and
35: are compared with data from an earlier higher state. The profiles are suggestive
36: of, and consistent with, an accelerating wind driven by a disk continuum
37: source, in which the ionisation radii have changed. This scenario may
38: apply to the formation of other broad emission lines in AGN.
39:
40: \end{abstract}
41:
42:
43:
44: \section{Introduction}
45:
46: While it is now known that roughly 50\% of all Seyfert galaxies
47: exhibit UV absorption lines (Crenshaw et al. 1999), the Seyfert 1
48: galaxy NGC~3516 is unique in that its absorption lines are the
49: strongest and most variable of any object in this class (Koratkar et al.
50: 1996; Goad et al. 1999). The deep, broad C IV absorption first visible
51: in {\it International Ultraviolet Explorer (IUE)} spectra (Ulrich 1988)
52: is reminiscent of the broad absorption
53: lines seen in $\sim 10$\% of radio-quiet QSOs (Weymann et al. 1991).
54: In many Seyfert 1 galaxies UV-absorbing gas appears in conjunction
55: with X-ray absorption by highly ionized gas. These X-ray ``warm absorbers"
56: are equally common in Seyferts (Reynolds 1997; George et al. 1998).
57: Crenshaw et al. (1999) note that all instances of X-ray absorption
58: also exhibit UV absorption.
59: While Mathur et al. (1994; 1995; 1997) have suggested that the same
60: gas gives rise to both the X-ray and UV absorption, the spectral complexity
61: of the UV and X-ray absorbers indicates that a wide range of physical
62: conditions are present.
63: In NGC~3516, the UV absorption in the Lyman lines, O VI, N V, Si IV,
64: and C IV observed using the {\it Hopkins Ultraviolet Telescope (HUT)} requires
65: multiple zones with differing column densities
66: and ionization parameters (Kriss et al. 1996a).
67: Likewise, multiple components with differing physical conditions
68: are present in the X-ray spectrum (Kriss et al. 1996b).
69: This inferred complexity is born out by the detection of four distinct
70: kinematic components in high-resolution spectra of the C IV region using
71: the {\it Goddard High Resolution Spectrograph (GHRS)} on the
72: {\it Hubble Space Telescope (HST)} (Crenshaw, Maran, \& Mushotzky 1998).
73:
74: With observations of only one ion in the GHRS spectrum it is impossible to
75: determine physical conditions in the absorbing gas. However, given the
76: low velocities and widths of two of the C IV components, Crenshaw et al. (1998)
77: suggest that they arise in gas in the interstellar medium or galactic halo
78: of NGC~3516. They associate the more blue-shifted components with outflowing
79: gas from the nuclear region. Given these characteristics, we planned
80: {\it Far Ultraviolet Spectroscopic Explorer (FUSE)} observations of the
81: 900--1200 \AA\ spectrum of NGC~3516 to make high-resolution measurements
82: of the O VI and Lyman-line absorption components.
83: As shown by Kriss et al. (2000) for the Seyfert 1 Mrk 509, such
84: observations can determine physical conditions in individual kinematic
85: components of the UV-absorbing gas and detect which, if any,
86: of them might be directly associated with the X-ray absorbing gas.
87: Ascertaining the physical conditions of the individual components
88: could give insight into the mechanisms giving rise to the outflow
89: and hence to the origin of associated UV absorbers in Seyfert nuclei.
90:
91: For NGC~3516, determining this structure could prove even more enlightening
92: since it is one of the rare Seyfert 1s whose narrow-line region shows a
93: bipolar morphology in the distribution of ionized gas around the nucleus
94: (e.g., Pogge 1989; Aoki et al 1994; Ferruit, Wilson, \& Mulchaey 1998).
95: In addition, NGC~3516 has an opaque intrinsic Lyman limit (Kriss et al. 1996a),
96: a feature seen only in the Seyfert 2 NGC~1068 and in only two other Seyfert 1s
97: with a clear bipolar narrow emission line morphology: NGC~4151 and
98: NGC~3227 (Kriss et al. 1997).
99: In the scheme of unified models for Seyfert galaxies, this is unexpected.
100: The standard model places the central engine and broad-line region in
101: the interior of an opaque torus (see the review by Antonucci 1993).
102: Light from the interior is collimated by the shadowing torus, and it
103: illuminates the surrounding narrow-line region with a biconical pattern.
104: Views from the pole of the torus give
105: a full view of the interior, corresponding to an object we see as a
106: Seyfert 1. Views in the plane of the torus show an obscured nucleus
107: dominated by narrow-line emission exterior to the torus, corresponding to
108: a Seyfert 2.
109: Thus, both NGC~3516 and NGC~4151 appear spectroscopically as Seyfert 1s,
110: but their narrow-line morphology resembles that of Seyfert 2s.
111: This suggests that we are viewing the interior close to the edge of the
112: collimated pattern, as noted by Hutchings et al. (1998).
113: Thus, the outflowing gas we see may be
114: associated with the more opaque material responsible for collimating the
115: ionizing radiation, or with an off-axis view of the central engine.
116:
117: In this paper we describe FUSE observations of NGC~3516 in the
118: 900--1200 \AA\ spectral range and contemporaneous HST observations
119: at longer UV wavelengths (1150 -- 3100 \AA) using the
120: {\it Space Telescope Imaging Spectrograph (STIS)} that shed new light
121: on the outflow phenomenon in Seyfert galaxies.
122:
123:
124: \section{Observations}
125:
126: FUSE was designed to obtain high-resolution (R$\sim$20,000) far-UV spectra
127: covering the 905--1187 \AA\ spectral range
128: (Moos et al. 2000; Sahnow et al. 2000).
129: Data are obtained in four independent optical channels, of which two
130: use LiF coatings on the optics, and two use SiC.
131: Two micro-channel plate detectors record individual photon events.
132: A summary of the early FUSE observations of AGN is given by Kriss (2000).
133: We observed NGC~3516 with FUSE on 2000 April 17, and with STIS
134: (gratings G230L and G140L) on 2000 April 20.
135: A full description of STIS and its performance are given by
136: Kimble et al. (1998).
137: As the AGN was in an unusually low state, the near-simultaneity of these
138: observations was of importance (and by design).
139: We thus have a good set of UV line profiles
140: in the low state to compare with the normal high state observed earlier
141: (Kriss et al. 1996a; Crenshaw et al. 1998; Goad et al. 1999).
142:
143: The FUSE data were obtained in 11 consecutive orbits
144: through the $30\arcsec \times 30\arcsec$ apertures
145: for a total exposure time of 16,382 s.
146: As described by Sahnow et al. (2000), the data were processed using the
147: standard FUSE pipeline to produce calibrated, one-dimensional spectra
148: in each of the four channels.
149: There were two exceptions to this process: first, we restricted the
150: pulse heights of acceptable data to channels 4--16 to reduce the
151: background; second, we measured the background in blank regions of each
152: detector segment and scaled the background constant to this value.
153: These steps improved the data quality on these faint exposures where
154: the continuum flux of NGC~3516 is only about
155: $1 \times 10^{-14}~\rm erg~s^{-1}~cm^{-2}~\AA^{-1}$.
156: Since the LiF1 optical channel is also the one used by the fine guidance
157: system on FUSE to maintain a precise pointing, the photometric stability
158: of the LiF1 data is assured. In the other optical channels, thermal
159: drifts can cause mis-centering of the target, leading to loss of flux
160: and wavelength-scale errors.
161: We therefore cross-correlated data from the other 3 channels with the
162: overlapping wavelength range of the LiF1 channel to adjust their
163: wavelength scales. We also applied gray-scale corrections to their
164: flux levels to bring them up to the same value as seen in the LiF1 spectrum.
165: To achieve a workable signal-to-noise ratio (S/N) while maintaining enough
166: resolution to recognize and remove H$_2$ absorption lines, the spectra were
167: binned to a resolution of $\sim$0.5 \AA.
168: As the different FUSE channels have different S/N levels and residual scattered
169: light problems, we assembled a complete 905--1187 \AA\ spectrum by selecting
170: the best data for each wavelength from among the four channels.
171:
172: The STIS data were obtained in slitless mode.
173: The G140L spectrum, covering the 1150--1700 \AA\ wavelength range
174: with a resolution of $\sim1000$, was obtained from a 600 s exposure.
175: The G230L spectrum, covering the 1650--3100 \AA\ range, was obtained in
176: a 300 s exposure in the same orbit.
177: No extended UV emission from the nuclear region of NGC~3516 was visible
178: in our images.
179: As the zero point of the wavelength calibration in slitless mode depends on
180: the location of the target in the field of view, we extracted the
181: one-dimensional STIS spectra in an iterative process.
182: Using the STIS calibration pipeline in the Space Telescope Data Analysis
183: System (STSDAS) {\tt stis} package, we first obtained a preliminary spectrum.
184: We measured the wavelengths of foreground Galactic absorption features, and
185: used these to determine appropriate offsets for the centering of the target.
186: We edited these offsets into the headers of the two-dimensional flat-fielded
187: data frames (as is usually done by the wavelength calibration process in the
188: pipeline) and re-ran the data through the pipeline to obtain our final
189: calibrated spectra with corrected wavelength scales.
190: Compared to the FUSE spectra, the STIS spectra are extracted from a much
191: smaller angular region surrounding the nucleus.
192: They have somewhat worse resolution but higher S/N.
193: In the overlap region from 1150 to 1180 \AA, the FUSE and STIS continuum levels
194: match quite well, so it appears that the continuum arises almost entirely in
195: the nuclear region with little to no contribution from starlight at larger
196: radii in the FUSE aperture.
197:
198: The diagrams illustrate the main features in the new spectra.
199: We also compare our new spectra with data of lower resolution from HUT,
200: which observed NGC~3516 in a high state in 1996 (Kriss et al. 1996a).
201: For comparison of the whole FUSE
202: wavelength range, the FUSE data were further smoothed to match the HUT data.
203: As the overall FUSE signal level was low (due to the target faintness and
204: observing time available), it was not
205: possible to obtain detailed spectra of the Lyman lines and Lyman limit in
206: NGC~3516. Therefore, the analysis and discussion below are not
207: as detailed as would be warranted by better S/N data.
208:
209:
210:
211:
212: \section{Line profile changes with nuclear flux changes}
213:
214:
215: Figure 1 shows plots of lines from our data compared with earlier,
216: high state, data. For easier comparison, the plots shown were placed on a
217: velocity scale, centred at the galaxy redshift for each line. This is
218: intended only to show the extent of the line asymmetries, which is not
219: very sensitive to the range of uncertainty in the true redshift value.
220: The 1996 data
221: were also scaled to match the 2000 low-state continuum, for better
222: comparison of the line profiles. This scaling is simply a factor applied
223: to the entire flux so that the continuum levels beyond the emission
224: lines match to within about 10\%.
225:
226: The O VI velocities are given with respect to redshift 0.0091. This is
227: higher than the published value of 0.008836 and is derived from the
228: reflection in velocity space that achieves the most symmetrical
229: outer broad emission profile for O VI. The other lines shown used the
230: published redshift value but were sufficently asymmetrical for various
231: reasons that they were not useful for defining the redshift. There is
232: uncertainty in this approach for O VI and the velocities in its plot are thus
233: uncertain by some 50 $\rm km~s^{-1}$. The ubiquity of the shortward
234: absorption in O VI, Ly$\alpha$, and C IV in this current low state of
235: the galaxy, and the strong narrow absorptions in all states, makes it
236: hard to define the line centre precisely.
237:
238: The top left FUSE O VI profile in Figure 1 is the combination of both
239: O VI lines, after approximate removal of the H$_2$ absorptions, which were
240: recognised from FUSE
241: data of different targets with high S/N spectra. The plot at lower right
242: shows both profiles without H$_2$ removal. The lower right plot also
243: shows the HUT data on the same flux scale. The doublet ratio of the O VI
244: line peaks in 2000 is nearer to unity than in the HUT 1996 data, indicating
245: that the optical depth was greater in 2000.
246:
247: The dashed lines on the left of the profiles are the reflected longward
248: profile assuming the above velocity zero. They thus show the extent of the
249: implied absorption in the profiles. The HUT Ly$\alpha$ profile is
250: heavily contaminated by local absorption (and geocoronal emission) to its
251: shortward side beyond -1000 $\rm km~s^{-1}$ or so.
252:
253:
254: The FUSE and STIS data are essentially simultaneous (obtained on
255: April 17 and 20, respectively). For comparison,
256: the dotted profiles are the HUT data from 1996, when the nucleus was in a
257: higher state. For easier profile comparison, the HUT data have been roughly
258: scaled to match the 2000 continuum levels. Thus, the flux scale shown applies
259: only to the year 2000 FUSE and STIS data. The changes in the relative
260: absorption between 1996 and 2000 are clear from these comparisons.
261: Note that the C IV absorptions seen earlier are still present, but that
262: there is extra absorption at shorter wavelengths in the 2000 low state.
263:
264: The upper right profiles show the broader Mg II and C III] profiles
265: (the latter blended on its shortward side by Si III] 1892\AA).
266: As noted by Goad et al. (1999),
267: these lines show no changes or absorption from high to low state. The
268: comparison between C IV and Mg II in centre right shows that the C IV
269: broad emission profile matches the Mg II, with the
270: absorptions lying below this broader profile.
271:
272: Figures 2 and 3 show the full wavelength ranges of FUSE and the FUV
273: channel of STIS, where other lines can be compared between the high and
274: low states of NGC~3516. Similar changes are seen in Si IV, while the weaker
275: line changes are more noisy. The large change in He II $\lambda$1640 is
276: probably due
277: to the larger aperture in the HUT data including hot stars as well as the
278: active nucleus. The Lyman absorptions seen in the HUT spectrum are not seen
279: in the low-state FUSE spectrum. However, there is severe airglow
280: contamination in Ly$\beta$, and the signal is noisy and weak at the higher
281: Lyman lines.
282:
283: \section{Low state line profiles}
284:
285:
286: The broad line profiles (both emission and absorption)
287: in the new low-state data strongly suggest that they
288: are dominated by outflow at velocities not seen in the high state.
289: The absorption and line width in O VI indicate
290: outflow velocities of some 600 $\rm km~s^{-1}$ in the hottest (highest
291: ionisation) regions, with the bulk at about 400 $\rm km~s^{-1}$.
292: There is absorption (outflow gas) at all velocities down to zero.
293: The C IV and L$\alpha$ profiles show the same velocity of maximum absorption
294: and absorption to zero velocity, but the outflow occurs to higher velocity,
295: with a second minimum at 1600 $\rm km~s^{-1}$ and a terminal velocity of
296: 2000 $\rm km~s^{-1}$. Much, but not all, of the low-velocity
297: outflow is also present in the high-state profiles, and we know that these
298: can be resolved into several sharp components, from higher resolution data
299: from HST.
300:
301: The lower resolution of the STIS spectra means that we cannot easily see
302: whether there are sharp components to any of the current outflows. The HUT
303: data on O VI have very poor resolution and noise but suggest a different O VI
304: optical depth overall. We also see that the emission from the low ionisation
305: and low density gas (C III] and Mg II) are the same as in the high state.
306: Thus, we seem to be seeing a change in the state of the high-ionisation gas.
307: As it coincides with continuum flux decrease, we speculate that there is
308: increased absorption in the inner part of the BLR responsible for the overall
309: phenomenon.
310:
311:
312: The scenario that we suggest for the broad emission and broad absorption
313: features is of an accelerating wind like a stellar
314: wind. The velocity increases and the ionisation decreases outwards. As the
315: central source flux drops, the ionisation boundaries move inwards. Wide
316: profiles with weak or non-detectable P-Cygni absorptions arise if lines are
317: formed in a large radius region at terminal velocity. This is the situation
318: in the high state (assuming the narrow low velocity absorptions arise in places
319: that lie far outside the wind region). In the low state, the high ionisation
320: lines are formed at much smaller radii from the nucleus, where P-Cygni
321: absorptions are more prominent, and can be seen through the whole absorbing
322: column to the continuum source. The highest ionisation lines are narrower
323: and have lower velocity absorptions. We describe and quantify this
324: scenario in a little more detail in the next section.
325:
326: We note that the C IV profile also contains strong low
327: velocity absorption, in both high and low states. This is resolved in
328: higher resolution HST data into several narrow features that do not
329: change, and are similar to narrow absorptions seen in other Seyferts
330: (Crenshaw et al. 1999). These arise in the general ISM and stable outflows
331: that lie well outside the BLR (as seen also in NGC~4151 by Kaiser et al. 2000),
332: and thus need to be
333: ignored or our discussion of the BLR profiles. The FUSE resolution is high
334: enough to show that the P Cygni absorption seen in O VI is not made of narrow
335: features, but is broad, as expected from the wind model.
336:
337: We note that the two higher velocity C IV absorptions resolved in high
338: resolution HST spectra have velocities that are close to those seen in the
339: FUSE O VI profiles. While the O VI profiles are not resolved into the
340: distinct components seen in C IV, it is possible that the absorbing
341: material is the same if the O VI absorbers are much more saturated.
342: If this is so, then we require a different explanation in terms of variable
343: high ionisation of dense material outside the BLR. It does not explain the
344: narrowness of the O VI emission in the FUSE data. Thus, while acknowledging
345: the possibility that the lower velocity absorbers have a different origin,
346: we will discuss in more detail the disk wind idea that appears to
347: explain more of the observed behaviour.
348:
349:
350: \section{The wind scenario for the BLR}
351:
352:
353: The profiles are broadly consistent with an accelerating wind, as present
354: in OB stars. It is worth checking this idea for consistency, since it
355: implies a very specific model for the BLR. Figure 4 outlines the scenario
356: in cartoon
357: fashion. The ionised gas that constitutes the BLR is accelerated away
358: from the central continuum source until it approaches a terminal velocity. The
359: terminal velocity is indicated by the overall width of the emission lines,
360: and where there are P-Cygni absorptions, by the maximum blue-shifted absorption.
361: There is an ionisation (temperature) gradient outwards from the central
362: source. Line profiles are formed in the range of radius (hence velocity)
363: where its ion is found. Thus, lines are formed in shells of increasing
364: radius (and velocity) as the ionisation state decreases.
365:
366: The cartoon profiles show qualitatively how the profiles differ as
367: the shell of formation
368: moves outward. These hold if the shells lie within a few times the radius of
369: the continuum source. Thus, the absorption column is a decreasing fraction of
370: the projected emission line region, as the shell radius increases. Close to
371: the continuum source, there is strong absorption, and all velocities are
372: low, so that the overall profile is narrow. It is also relatively `peaky'.
373: For shells of increasing radius, the overall profile width increases, and the
374: absorption becomes weaker and occurs only at the shortward edge of the
375: profile. Once terminal velocity is reached, the profile width remains the
376: same, and the absorption becomes negligible at larger radii. The details
377: of the profiles depend on the exact relationship between ionisation, radius,
378: and velocity.
379:
380: We note that this simple model is for spherically symmetric expansion, as
381: expected for a star. If the wind is driven by a disk, the geometry becomes
382: a further parameter, and the absorption and emission components may be
383: separated if the rotation velocity is comparable to the outflow velocity.
384: Knigge and Drew (1996) discuss
385: disk winds seen in cataclysmic binary systems, that show some of these
386: effects. Disk winds may be driven over a broad angle on each side of the
387: disk, but not near the disk plane. Since AGN jets are also driven perpendicular
388: to the disk, there may be a cone of wind avoidance where the jet lies, if
389: there is one. The Seyfert 1 paradigm is that we view the central disk and BLR
390: directly without obscuration by the opaque torus, and thus within the
391: expected sightline of a central wind. The wind profiles for
392: such a disk sightline are similar to those for a spherical wind, with slightly
393: less peaky emission profiles. Thus, the simple wind cartoon is generally
394: applicable to the expected geometry, but dependent in detail on the
395: solid angle filled by the wind.
396:
397: We note that the disk wind model of
398: Murray et al. (1995) for BALQSOs blows close to the disk
399: plane for objects of higher luminosity and outflow velocity (see also
400: Elvis 2000).
401: The hydromagnetic wind of K\"onigl and Kartje (1994), on the other hand,
402: is more collimated normal to the disk plane. Proga, Stone, and Kallman (2000)
403: discuss the dynamics for disk winds from more massive central objects,
404: and note the role of X-ray shielding and ionisation that launch and
405: accelerate the wind in such models.
406: The strong X-ray flux from the nuclear source is likely to highly ionize
407: unshielded portions of the disk wind, as emphasized by Murray et al. (1995).
408: This highly ionized portion of the wind could be associated with the X-ray
409: warm absorber. Mathur et al. (1997) discuss X-ray warm absorbers in
410: NGC~3516 in the context of an outflowing wind, but, unfortunately, we do not
411: have contemporaneous X-ray spectra that would permit us to test their
412: predictions.
413:
414:
415:
416: We now postulate that the change between high and low states of NGC~3516
417: are principally expansion or shrinking of the ionisation radii within the
418: wind, as shown in the sketch. This would arise from changes in the amount of
419: ionising radiation that are effectively the observed high and low luminosity
420: states of the nucleus. The changes seen in the profiles of high, intermediate,
421: and low ionisation, as illustrated by O VI, C IV, and Mg II, are much as
422: observed.
423: The postulated stratification is both a function of ionisation state as well
424: as the line formation mechanisms in the wind. Both thermal emission and
425: scattering contribute to the formation of wind emission lines. As in OB
426: stars, the low-ionization lines are likely dominated by thermal emission,
427: while scattering will dominate the observed high-ionization line emission.
428: Since scattered line profiles are sensitive to the velocity gradient,
429: broader lines are formed are larger radii.
430: Thus, the overall idea seems feasible, \it provided that the BLR
431: lies within a few disk radii of the continuum source. \rm
432:
433:
434:
435:
436: We can also imagine that the high to low state changes involve changes
437: in the continuum source radius or the velocity profile within the BLR.
438: The flux changes are no more than a factor of 2--3, so that we do not expect
439: disk-size or wind-acceleration changes to be large or rapid. The change of
440: the ionisation radii will be the immediate and dominant effect in any case.
441:
442: We can make a rough estimate of the radii, based on the assumption of
443: similarity with OB stellar winds. The stellar winds are driven by the radiation
444: pressure in the rest-frame UV where the strong absorptions lie that drive
445: the wind
446: by photon scattering. The nucleus of NGC~3516 has similar colour to an OB star,
447: so we may compare the magnitudes to estimate the radiation that drives the
448: winds. NGC~3516 has a redshift of $\sim$2650 $\rm km~s^{-1}$ which gives it a
449: distance of some 40 Mpc or distance modulus of about 33.
450: The galaxy V magnitude is 12.4, (with the nucleus at 14 - 16)
451: which yields an absolute magnitude of -20.6, of which the nucleus accounts
452: for about -18. An OB star with a strong wind has absolute V magnitude of -7
453: (with similar colours), so that the nuclear continuum source is equivalent to
454: some 60,000 OB stars. Assuming we require the same radiation level for the
455: disk wind of NGC~3516, this implies a continuum source radius of 250 times
456: that of an OB star, which is about 60 a.u. The BLR then should have a radius
457: of several times this, which is about 2 light days in radius. Echo mapping
458: observations indicate BLR diameters of a few light days for Seyferts.
459: In the case of NGC 3516 it is measured at 4.5 days for the total C IV emission
460: line flux (Koratkar et al 1996). Thus, within the uncertainties of the
461: luminosity estimate above, the
462: scenario seems very consistent with a radiation-driven wind for the BLR. In
463: fact, the numbers above would \it require \rm a wind to be present by the same
464: processes that we understand for stellar winds.
465:
466: It is interesting to see how the masses and mass flows scale in this
467: rough comparison. Stellar winds have terminal velocities closely related to
468: the surface escape velocity of the star. The terminal velocity in NGC~3516
469: is $\sim$2000 $\rm km~s^{-1}$,
470: very typical of stellar winds, so that if the mechanisms
471: are the same, the average effective gravity at the base of the AGN wind
472: should be comparable with an OB star. The average particle at the surface of
473: the AGN accretion disk (of radius 250 OB star radii) lies at distance
474: $\sim$200 OB star radii from the central object. To make the gravity
475: equal in both cases requires a central mass of about 10$^6$ M$_{\odot}$,
476: which again is the right order of magnitude. If we scale the mass loss
477: against typical stellar winds of about 10$^{-6}$ M$_{\odot}$ per year,
478: allowing that
479: the AGN wind may fill say 1/4 of a sphere, we find a mass-loss rate of
480: a few M$_{\odot}$ per year. This again is not unreasonable, but would be
481: significant in the accretion budget for the AGN.
482:
483:
484:
485: In in more comprehensive picture, we note that ionisation by high energy
486: photons from the AGN will alter the wind as its opacity changes with height,
487: probably giving rise to the warm absorbers seen in the X-ray spectra. The
488: wind structure that may apply in NGC~3516 will need detailed modelling, and
489: comparison with data from different luminosity states will help define the
490: important ionisation structure. Such modelling---particularly the dynamics
491: of rotation and acceleration of the wind--- must await more and better
492: contemporaneous data at FUV and X-ray wavelengths.
493:
494:
495: Our principal conclusion from the present data is that we feel there is
496: significant evidence that the BLR in NGC~3516 is an accelerating wind driven
497: by the nuclear continuum source, that is seen in sightlines within the
498: cone of UV ionisation. If this applies in a more general way, we might
499: expect to see similar profile changes in other Seyferts with very low
500: states. In the case of NGC~4151, this does seem to be true: P Cygni profiles
501: are seen in lines in low states - most notably in data taken in the current
502: (year 2000) low state by STIS. We therefore suggest that careful monitoring
503: of line profiles with nuclear flux would be a valuable way of verifying the
504: model proposed and learning more about how it works.
505:
506: Our discussion relates to the connection between broad-line profile
507: widths and the mass of the central black hole. Clearly, any motions in the
508: central region are dominated by the gravitational field, whether they be
509: in Keplerian orbit, free-falling, moving with escape velocity, or driven
510: by radiative forces. More detailed wind models may enable us to use
511: profile widths to test them, and thereby formulate the relation between
512: observed profiles and central mass. We note that the low state of NGC 3516
513: we have sampled, may reveal the inner part of the outflow not often seen in
514: more luminous and extended BLRs, where line widths are no longer radiatively
515: driven.
516:
517: Finally, we note that the highest ionisation lines, such as the Fe K line
518: seen at 7Kev in X-ray spectra, have profiles that show they arise in a
519: disk of relativistic velocities, so do not arise in the BLR wind, but in
520: the continuum source disk itself (Nandra et al. 1999).
521:
522:
523:
524: \section{Other spectral features}
525:
526: Figure 2 shows the entire FUSE range spectrum compared with the same
527: range with HUT in 1996. For this plot, the FUSE data were selected to use only
528: the best data for each wavelength region, and then smoothed to about 0.4 \AA\
529: resolution, for easier comparison of the broad weak features. The airglow lines
530: have been removed, but not the H$_2$ absorptions, which must be present
531: in both datasets. The HUT data also show zero redshift L$\beta$ absorption
532: which has not been removed. At the shortest wavelengths, there are probably
533: detector and noise effects that are not real differences.
534:
535: The lines present other than O VI are N III $\lambda$990, He II
536: $\lambda$1085, and possibly N IV $\lambda$948
537: and S IV+Si IV $\lambda$1063-66. The C III $\lambda$977 line is not detectable,
538: while the other lines are all narrower as well as weaker than in the 1996
539: spectrum. There are very broad features that appear to be associated with the
540: lines of C III, N III, O VI, and He II that are seen at both epochs, for which
541: we offer no detailed explanation. The overall continuum level in 2000 is
542: about 1/3 that of 1996, and the broad features all scale with this same factor.
543: There are differences at the positions of N III and the S IV/Si IV blend.
544: Sharper emissions are seen on top
545: of the broad wings at He II (only in 2000) and N III and O VI.
546:
547: Moving to longer wavelengths (1200 to 1800 \AA), Figure 3 shows the
548: comparison between the 1996 HUT data and the 2000 STIS data. The flux level
549: difference becomes smaller with increasing wavelength and is close to
550: 2 at 1800 \AA. Aside from emission line profile changes, the relative
551: strength of He II and N III] are lower in the 2000 low state. Since it is
552: unlikely that the forbidden line flux changes significantly, the difference is
553: probably due to the line emission arising over an extended region that is
554: differently sampled by the STIS and HUT slits. The FUSE and STIS continuum
555: levels match quite well where they join, so that it appears that the continuum
556: arises almost entirely in the nuclear region. The drop in He II $\lambda$1640
557: flux between HUT and STIS may also indicate an extended origin of this emission,
558: since the He II $\lambda$1085 line does not show the same drop in flux.
559: He II $\lambda$1640, along with He II $\lambda$4686, arises in hot stars and
560: thus may well be extended in this galaxy.
561: Thus, among the weaker emission lines from the nuclear region, both N III
562: and He II have developed sharp components in the 2000 low state.
563:
564:
565: \section{Discussion}
566:
567: The paradigm of the unified model for AGN is that in Seyfert 1
568: galaxies, we view the
569: broad line region from within the cone of light illuminated by the nucleus.
570: The cone is caused by shadowing of an equatorial region by an opaque torus
571: around the nucleus: Seyfert 2s are those nuclei hidden by the torus.
572: Inside the cone there is ionising radiation, radio jet material, and clouds
573: of narrow emission line material. Recent data from STIS (e.g. Hutchings
574: et al. 1998, Kaiser et al. 2000, Crenshaw et al. 2000) suggest that the
575: narrow-line material is moving outwards at a few hundred $\rm km~s^{-1}$ along
576: the inside surfaces of the cones, and not filling them.
577: The STIS data indicate there are also faster moving clouds that fill a wider
578: angle and arise close to the nucleus (Hutchings et al. 1999). These
579: are seen both in absorption and emission, and the former are usually stable
580: over years. Occasionally some narrow absorptions change or new ones appear.
581:
582: In addition to these narrow and relatively stable absorbers,
583: the change of broad line profiles from 1996 to 2000 suggest that they
584: may be explained by an outflowing BLR. This paper has explored the
585: way in which the change of state allows us to probe the
586: velocity structure of a broad-line wind, by varying the radii of the shells
587: in which different lines are formed. We think that the density structure may
588: not change significantly. The geometry of the wind, what happens to its
589: material as it moves outward and cools, and its relationship with the other
590: mass flows, need to be considered and tested with observations in other objects
591: as well as NGC~3516 itself as it continues to vary. Our scenario predicts
592: a close link between the BLR profiles and the continuum flux, which can be done
593: with continued monitoring, particularly in the O VI lines.
594:
595: Kriss et al. (1996a) note that the presence of Lyman limit absorption
596: and the S-shaped NLR in NGC~3516 may imply a line of sight near to the
597: edge of the cone, as in NGC~4151. This does not preclude the disk wind
598: scenario, but it would be useful to study the NLR dynamics in detail
599: for more geometrical clues, and obtain better FUSE data to look for the
600: Lyman absorption lines and limit.
601:
602: There have been other discussions of the BLR of NGC 3516, based
603: on monitoring campaigns for echo-mapping, as well as more sporadic
604: observations of high and low states and the occasional presence of
605: BAL-type absorption (e.g. Wanders et al 1993, Koratkar et al 1996, Goad et al
606: 1999). The suggested model in this paper is consistent with these other
607: discussions, none of which have proposed a detailed model.
608:
609: NGC 3516 has a NLR component, which we assume does not change with
610: the nuclear and BLR changes, since it is formed in the much more extended
611: S-shaped region noted above. This component is not seen in the UV lines
612: in Figure 1, although Goad et al (1999) model profiles with a narrow component
613: of FWHM some 1200 km s$^{-1}$. The narrow emission peak seen in O VI is much
614: narrower than this, and is unlikely to represent the NLR emission for this
615: reason as well as for its high ionisation. Aoki et al (1994) also discuss
616: the ionisation of the NLR and its relationship to the nuclear source.
617:
618: Our FUSE profiles of O VI are key to our suggested model, and we have
619: noted that a higher S/N would be very useful in checking whether these lines
620: do contain NLR components such as the narrow absorbsions and emission peak.
621: With the data in hand we favour the explanation that they arise in the
622: inner accelerating part of a central mass flow.
623:
624: There are some clues in other objects. VLBI radio maps of NGC~4151
625: shows a bend close to the nucleus (Roy et al. 1999), where the radio
626: structure becomes aligned with the outer ionisation cones and radio structure.
627: The same galaxy also shows P Cygni profiles occasionally, and notably in its
628: current low state.
629: In NGC~4151, our line of sight is deduced to be close to the edge of the
630: cone.
631:
632: The relation between line FWHM and reverberation time lag (Peterson and
633: Wandel 2000) for other AGN, indicates that in those cases the line profiles
634: become narrower with distances from the nucleus that are larger than those
635: we consider here. This would indicate that the wind velocity falls again at
636: radii where the radiation pressure has dropped and the wind material is slowed
637: by interactions with the ambient medium. It is only in cases such as the low
638: state of NGC~3516, that broad line formation is dominantly in the inner
639: accelerating region.
640:
641: We may further speculate whether a wind model for BLR applies to more
642: energetic QSOs. These have lines (and BAL velocities) that are higher
643: by a factor of several, luminosities higher by up to 100 or more, BLR sizes
644: (from variability timescales) larger by about 100, and central masses
645: probably 2 to 3 orders of magnitude higher. These numbers all scale quite
646: reasonably to winds, although the above parameters are not tightly linked
647: to each other. Good resolution and S/N studies of variations in highly
648: ionised emission lines in a range of AGN will be necessary to pursue these
649: questions.
650:
651: This work is based on data obtained for the Guaranteed Time
652: Team by the NASA-CNES-CSA FUSE mission operated by the Johns Hopkins University.
653: Financial support to U. S. participants has been provided by NASA contract
654: NAS5-32985.
655:
656: \clearpage
657:
658: \begin{thebibliography}{}
659:
660: \bibitem[Antonucci 1993]{Antonucci93}
661: Antonucci, R.R.J. 1993, \araa, 31, 473
662:
663: \bibitem[Aoki, Ohtani, Yoshida, \& Kosugi(1994)]{1994PASJ...46..539A} Aoki, K., Ohtani, H., Yoshida, M., \& Kosugi, G.\ 1994, \pasj, 46, 539
664:
665: \bibitem[Crenshaw, Maran, \& Mushotzky 1998]{Crenshaw98}
666: Crenshaw, D. M., Maran, S. P., \& Mushotzky, R. F. 1998, \apj, 496, 797
667:
668: \bibitem[Crenshaw et al. 1999]{Crenshaw99}
669: Crenshaw D.M., Kraemer S.B., Boggess A., Maran S.P., Mushotzky R.F.,
670: Wu C-C, 1999, \apj, 516, 750
671:
672: \bibitem[Crenshaw et al. 2000]{Crenshaw00}
673: Crenshaw D.M. et al. 2000, in press (astro-ph 0007017)
674:
675: \bibitem[Elvis 2000]{Elvis00}
676: Elvis M.\ 2000, \apj, 545, 63
677:
678: \bibitem[Ferruit, Wilson, \& Mulchaey 1998]{Ferruit98}
679: Ferruit, P., Wilson, A. S., \& Mulchaey, J. S. 1998, \apj, 509, 646
680:
681: \bibitem[George et al. 1998]{George98}
682: George, I. M., et al. 1998, \apjs, 114, 73
683:
684: \bibitem[Goad et al. 1999]{Goad99}
685: Goad M.R., Koratkar A.P., Axon D.J., Korista K.T., O'Brien P.T., 1999,
686: \apj, 524, 707
687:
688: \bibitem[Hutchings et al. 1998]{Hutchings98}
689: Hutchings, J.B., et al. 1998, \apj, 492, L115
690:
691: \bibitem[Hutchings et al. 1999]{Hutchings99}
692: Hutchings, J.B., et al. 1999, \aj, 118, 2101
693:
694: \bibitem[Kaiser et al. 2000]{Kaiser00}
695: Kaiser, M.E., et al. 2000, \apj, 528, 260
696:
697: \bibitem[Kimble et al. 1998]{Kimble98}
698: Kimble, R., et al. 1998, \apj, 492, L83
699:
700: \bibitem[K\"onigl \& Kartje 1994]{KK94}
701: K\"onigl, A., \& Kartje, J. F. 1994, \apj, 434, 446
702:
703: \bibitem[Koratkar et al. 1996]{Koratkar96}
704: Koratkar, A. P., et al. 1996, \apj, 470, 378
705:
706: \bibitem[Knigge \& Drew 1996]{Knigge96}
707: Knigge C., and Drew J.E., 1996, \mnras, 281, 1352
708:
709: \bibitem[Kriss et al. 1996a]{Kriss96a}
710: Kriss G.~A., et al. 1996a, \apj, 467, 622
711:
712: \bibitem[Kriss et al. 1996b]{Kriss96b}
713: Kriss G.A. Espey B.R., Krolik J.H., Tsvetanov Z., Zheng W., Davidsen A.F.,
714: 1996b, \apj, 467, 629
715:
716: \bibitem[Kriss et al. 1997]{Kriss97}
717: Kriss, G. A., et al. 1997, in ASP Conf. Ser. Vol. 113,
718: Emission Lines in Active Galaxies: New Methods and Techniques, eds.
719: B.M. Peterson, F.-Z. Cheng, \& A.S. Wilson
720: (San Francisco: ASP), 453
721:
722: \bibitem[Kriss et al. 2000]{Kriss00a}
723: Kriss, G. A., et al. 2000b, \apj, 538, L17
724:
725: \bibitem[Kriss 2000]{Kriss00b}
726: Kriss, G.A. 2000,
727: to appear in the proceedings of the workshop on ``Probing the Physics of
728: Active Galactic Nuclei by Multiwavelength Monitoring, eds. B. Peterson,
729: R. Polidan, \& R. Pogge, (ASP:San Franciso), in press (astro-ph 0008492)
730:
731: \bibitem[Mathur, Wilkes, \& Elvis 1995]{Mathur95}
732: Mathur, S., Wilkes, B., \& Elvis, M. 1995, \apj, 452, 230
733:
734: \bibitem[Mathur et al. 1994]{Mathur94}
735: Mathur, S., Wilkes, B., Elvis, M., \& Fiore, F. 1994, \apj, 434, 493
736:
737: \bibitem[Mathur, Wilkes, \& Aldcroft 1997]{Mathur97}
738: Mathur, S., Wilkes, B., \& Aldcroft, T. 1997, \apj, 478, 182
739:
740: \bibitem[Moos et al. 2000]{Moos2000}
741: Moos et al. 2000, \apj, 538, L1
742:
743: \bibitem[Murray et al. 1995]{Murray95}
744: Murray, N., Chiang, J., Grossman, S. A., Voigt, G. M. 1995, \apj, 451, 498
745:
746: \bibitem[Nandra et al. 1999]{Nandra99}
747: Nandra K., George I.M., Mushotzky R.F., Turner T.J., Yaqoob T., 1999,
748: astro-ph 9907193
749:
750: \bibitem[Peterson \& Wandel 2000]{PW00}
751: Peterson B. M., \& Wandel A., 2000, in press (astro-ph 0007147)
752:
753: \bibitem[Pogge 1989]{Pogge89}
754: Pogge, R. 1989, \aj, 98, 124
755:
756: \bibitem[Proga et al 2000]{Proga2000}
757: Proga D., Stone J.M., Kallman T.R., 2000, \apj, 543, 686
758:
759: \bibitem[Reynolds 1997]{Reynolds97}
760: Reynolds, C. S. 1997, \mnras, 286, 513
761:
762: \bibitem[Roy et al. 1997]{Roy97}
763: Roy, A.L., Ulvestad, J.S., Colbert, E.J.M., Wilson, A.S, \& Norris, R.P. 1997,
764: IAU Colloq 164
765:
766: \bibitem[Sahnow et al. 2000]{Sahnow2000}
767: Sahnow D., et al. 2000, \apj, 538, L7
768:
769: \bibitem[Ulrich 1988]{Ulrich88}
770: Ulrich, M.-H. 1988, \mnras, 230, 121
771:
772: \bibitem[Wanders et al. 1993]{Wanders93}
773: Wanders I, et al. 1993, A\&A, 269, 39
774:
775: \bibitem[Weymann et al. 1991]{Weymann91}
776: Weymann, R. J., Morris, S. L., Foltz, C. B., \& Hewett, P. C. 1991,
777: \apj, 373, 23
778:
779: \end{thebibliography}
780:
781: \clearpage
782:
783: \centerline{Figure captions}
784:
785: 1. Comparison in velocity space with respect to the galaxy velocity, of
786: profiles from 2000 low state with 1996 high state. In the left panels,
787: the dashed lines show the reflected longward profile to highlight the
788: shortward absorption. The top left O VI profile combines both lines of
789: the doublet, with the H$_2$ absorptions removed. The full O VI data are
790: shown in lower right panel, with H$_2$ lines
791: marked with vertical dashes. The 1996 fluxes
792: have been scaled to match the continuum levels from the low state, to show
793: the profile differences. FUSE data are smoothed to 0.5 \AA\ resolution and the
794: HUT data have 2 \AA\ resolution. Note the increased broad absorption in
795: the low state data.
796:
797: 2. Comparison of the low and high state spectra over the entire FUSE range.
798: The FUSE data are smoothed to match the HUT data, and the HUT data are
799: displaced in Y by 2x10$^{-14}$ (dashed line) for clarity.
800: The principal broad emission lines are identified above the HUT spectrum.
801: Other broad features with possible identifications are marked where
802: they are seen
803: in the FUSE data. In addition to the strong O VI, N III and He II
804: emission peaks, there appear to be broad components that scale with
805: the continuum. Overlapping STIS coverage is shown as dotted line.
806:
807: 3. Comparison of the STIS FUV spectral range with the HUT high state spectra.
808: Note the profile changes in Si IV, and He II as well as those shown in
809: Figure 1.
810:
811: 4. Sketch showing the way that a contraction of the ionisation structure in the
812: low state can explain the profile changes seen. While no units are given,
813: radii range from the surface of the continuum source to several times its
814: radius. The text discusses this diagram in detail.
815:
816:
817: \end{document}
818: