1: %\documentclass[12pt,preprint]{aastex}
2: \documentclass{emulateapj}
3: \usepackage{times}
4:
5: \shorttitle{Continuum Acceleration of Black Hole Winds}
6: \shortauthors{Everett \& Ballantyne}
7:
8: %\bibliographystyle{apj}
9:
10: %\def\ltsim{\mathrel{\rlap{\lower4pt\hbox{\hskip1pt$\sim$}}
11: % \raise1pt\hbox{$<$}}} % less than or approx. symbol
12:
13: \begin{document}
14:
15: \title{Continuum Acceleration of Black Hole Winds}
16:
17: \author{John E. Everett and David R. Ballantyne}
18: \affil{Canadian Institute for Theoretical Astrophysics, University of
19: Toronto, 60 St. George Street, Toronto, ON M5S 3H8, Canada}
20: \email{everett, ballantyne@cita.utoronto.ca} \slugcomment{Accepted to ApJL}
21:
22: \begin{abstract}
23: Motivated by recent observations of high-velocity, highly ionized
24: winds in several QSOs, models of purely continuum-driven winds
25: launched from $\sim200~GM_{\rm BH}/c^2$ are presented. Launching
26: conditions are investigated, as well as the observational signatures
27: for a variety of initial conditions and illuminating continua. While
28: we verify that continuum-driven, highly-ionized outflows reach the
29: observed velocities for $L/L_{\rm Edd}\geq1$, independent of the
30: incident spectral shape, such winds are too highly ionized to exhibit
31: the observed absorption features when launched with an AGN continuum
32: (in fact, such winds are so ionized that they are driven primarily by
33: electron scattering). If the wind is instead illuminated with a
34: blackbody continuum originating from an optically thick shield, the
35: gas is too weakly ionized and does not produce high energy absorption
36: features. If high-velocity, high-ionization winds are truly launched
37: from very near the black hole, such winds must be launched under other
38: conditions or via other processes; we summarize some possibilities.
39: \end{abstract}
40:
41: \keywords{galaxies:active -- quasars:absorption lines -- accretion, accretion disks -- galaxies:individual (PG 1211+143) -- X-rays:galaxies}
42:
43: \section{Introduction}\label{Intro}
44:
45: Optical, UV, and X-ray absorption troughs in Active Galactic Nuclei
46: (AGN) spectra are central to the ongoing effort to understand not only
47: mass outflow in AGN, but to provide clues to the fueling of their
48: central supermassive black holes \citep[see][]{CKG03}. Although the
49: optical depth, metallicity, and velocity of such outflows are becoming
50: increasingly well-constrained, the mechanism accelerating the winds
51: remains elusive: thermal \citep[e.g.,][]{KK95}, continuum-driven
52: \citep[for instance,][]{CN03}, line-driven
53: \citep[e.g.,][]{MCGV95,PSK00}, and hydromagnetic \citep[e.g.,][]{KK94,
54: B97} winds may all provide insight, but none of the above models have
55: been excluded or are definitively observationally preferred for AGN
56: outflows.
57:
58: In this Letter, observational constraints on continuum-driven winds
59: launched from very near the central black hole are examined. Such
60: wind models have become significant, as observations of outflows
61: apparently moving at large fractions of the speed of light
62: ($v~\sim~$0.1--0.4c) have been published by \citet{CBG03},
63: \citet{CBGG03}, \citet{Pounds03a, Pounds03b}, and \citet{ROW03}. From
64: the measured high velocities, the winds are inferred to be launched
65: very close to the central black hole. \citet{KP03} have proposed that
66: the highly-ionized gas is subject to continuum driving near the
67: central black hole because line driving is unlikely to be important
68: for the large ionization parameters in these winds.
69:
70: A model of a continuum-driven wind is presented in
71: \S\ref{modelOverview}. This model is used to investigate terminal
72: velocities and absorption features displayed by a variety of winds
73: with different incident continua, initial shielding columns, and
74: densities at the base of the wind (\S\ref{Results}).
75:
76: %The results are
77: %summarized in \S\ref{Summary}.
78:
79: \section{Model Overview}\label{modelOverview}
80:
81: Our simple continuum-driven wind model consists of a radial streamline
82: set to rise above the disk at a constant angle of $5^\circ$ to the
83: disk, with the initial column ($N_{\rm H,0}$) and number density
84: ($n_0$) of the wind specified by the user. Gas with that initial
85: density is assumed to be in Keplerian motion and is perturbed into the
86: central, illuminating continuum with an initial velocity of roughly
87: the local sound speed. To determine the gas's motion due to continuum
88: acceleration, Cloudy \citep{Ferland02} simulations are run on a
89: %user-defined number of
90: radial rays through the wind, spaced logarithmically in co-latitude.
91: %(usually 21 such Cloudy runs are used).
92: The continuum acceleration on the gas is then computed using both the
93: ionization state and transmitted continuum (therefore taking account
94: of any self-shielding effects) at the end of each radial slice through
95: the wind, taking into account both electron scattering and bound-free
96: opacities.
97: % to calculate the continuum acceleration.
98: The radiative acceleration is calculated and then compared to gravity
99: in $\Gamma_{\rm cont}$, defined as:
100: \begin{equation}
101: \Gamma_{\rm cont} = \frac{a_{\rm cont}}{g} = \frac{\frac{1}{\rho c} \int \chi_{\nu} F_{\nu}
102: d\nu}{\frac{G M_{\rm BH}}{R^2}}
103: \end{equation}
104: where $a_{\rm cont}$ is the continuum radiative acceleration, $g$ is
105: the gravitational acceleration, $\rho$ is the density in the wind, $c$
106: is the speed of light, $\chi_{\nu}$ is the frequency-dependent
107: continuum opacity (output by Cloudy), $F_{\nu}$ is the transmitted
108: continuum (again, given by Cloudy), $G$ is the gravitational constant,
109: $M_{\rm BH}$ is the mass of the central black hole, and $R$ is the
110: spherical radius. This $\Gamma_{\rm cont}$ factor is then input into
111: the Euler equation along with the component of gravity and centrifugal
112: acceleration along the radial flowline:
113: \begin{equation}
114: v\frac{dv}{dr}=-\frac{G M_{\rm BH}(1-\Gamma_{\rm cont})}{R^2}
115: \cos(\theta-\theta_{\rm F}) +
116: \frac{v_{\phi}^2(\varpi)}{\varpi}\sin\theta_{\rm F}
117: \end{equation}
118: where $\varpi$ is the cylindrical radius, $\theta$ is the co-latitude,
119: %of the gas flowing along the radial streamline,
120: $\theta_{\rm F}$ is the (fixed) angle of the flowline from the
121: vertical, and $v_{\phi}(\varpi) = \Omega_0 r_0^2/\varpi$ is the
122: azimuthal velocity (angular momentum is conserved in outflowing gas).
123: %We
124: %non-dimensionlize this equation as in \citet{FS99} to read:
125: %\begin{equation}
126: %W' = - \frac{s + \sin\theta_{\rm F}}{(1+2s\cos\theta_{\rm
127: % F}+s^2)^{3/2}} + \frac{\sin\theta_{\rm F}}{(1+s\cos\theta_{\rm F})^3}
128: %\end{equation}
129: %where $W = (v/v_{\rm esc})^2$, $v_{\rm esc}$ is the escape velocity at
130: %the launching radius $r_0$, $s$ is the distance along the flowline in
131: %units of $r_0$, and $W'$ indicates $dW/ds$.
132:
133: We integrate this equation to compute the velocity of the wind as a
134: function of distance along the radial streamline. As that velocity
135: increases, the number density in the wind (as well as the column
136: density) decreases via mass conservation. For the first set of Cloudy
137: simulations, where no velocity data are available, the density is set
138: to fall as $r^{-2}$; this initial density distribution is replaced in
139: subsequent iterations and is unimportant for the final results. The
140: solution is iterated at least twice to ensure that the density
141: structure is consistent with the velocity profile in the wind;
142: simulations in test cases have shown that the wind structure does not
143: change after two iterations, even when ten iterations are completed.
144:
145: The wind parameters chosen in this paper reproduce the observational
146: constraints on the outflow observed by \citet{Pounds03a} in the narrow
147: emission line quasar PG~1211+143. They observed an outflow with
148: $v\sim0.08c$ and an inferred density of $n\sim3\times10^8~{\rm
149: cm}^{-3}$, situated at $3\times10^{15}~{\rm cm}$ from the black hole
150: \citep[this density is estimated observationally from the emission
151: measure and line-of-sight column density; see][]{Pounds03a}. They
152: also inferred a launching radius of $r_0=10^{15}~{\rm cm}$ by noting
153: that $v=0.08c\sim v_{\rm esc}$, the escape velocity for that $r_0$.
154: Since $\dot{M}=4 \pi b r^2 \rho v$ (where $b$ is the covering factor
155: of the flow, $r$ is the radial distance, $\rho$ is the mass density,
156: and $v$ the velocity of the wind) must be constant along the flow, an
157: original density of $\sim5\times10^{12}~{\rm cm}^{-3}$ is inferred for
158: the base of the wind if the initial velocity is of order $20~{\rm
159: km/s}$ and $b=0.8$ \citep[the value of $b$ used in][]{Pounds03a}.
160: Observations indicate that $M_{\rm BH}\approx4\times10^{7}~M_{\sun}$
161: \citep{Kaspi00}\footnote{A recent reanalysis by \citet{Peterson04}
162: yields $1.46\pm0.44\times10^8~M_{\sun}$ (with significant uncertainty
163: noted in the analysis); we retain the value of
164: $4\times10^{7}~M_{\sun}$ to compare with previous models.}, and
165: $L_{\rm Edd}\approx5\times10^{45}~{\rm erg/s}$ \citep[$L/L_{\rm Edd}
166: \sim 0.8$ as the ``Big Blue Bump'' that dominates the continuum in
167: PG~1211+143 has $L\sim 4\times10^{45}$~erg/s;][]{Pounds03a}. The
168: inferred mass outflow rate from the observations is $\sim
169: 2.8~M_{\sun}/{\rm yr}$ (for the covering fraction $b=0.8$). We assume
170: solar abundances.
171:
172: The increase of density towards the base of the wind (where the gas is
173: outflowing much more slowly) results in a large increase in the column
174: density over the observed column of $\sim5\times10^{23}~{\rm
175: cm}^{-2}$. As recognized by \citet{KP03}, this column would be
176: significantly optically thick, precluding photoionization simulations.
177: Two possibilities are therefore considered in our simulations. First,
178: ``standard'' winds are tested with $N_{\rm H,0} < 10^{24}~{\rm
179: cm}^{-2}$ at the base of the outflow and with an AGN spectral energy
180: distribution (SED). Next, winds are launched by a soft X-ray/UV
181: blackbody from the photosphere of an optically thick shield, as
182: hypothesized by \citet{KP03}.
183:
184: \section{Kinematics and Absorption Features}\label{Results}
185:
186: The most immediate check for these outflows is to see which parameters
187: lead to the observed velocities. Figure~\ref{vInfPlot} displays
188: $v_{\infty}$ as a function of the Eddington ratio ($v_{\infty}$ here
189: is the velocity at the end of the model's streamline at a distance of
190: $100r_{0}$ where $r_0$ is the launch radius of our wind; taking the
191: terminal velocity at $10^4~r_0$ only changes $v_{\infty}$ by $\sim1\%$
192: in tests). This figure shows the variation in $v_{\infty}$ for both
193: an AGN SED \citep{RE04}\footnote{The \citet{RE04} spectrum is
194: approximated by using Cloudy's ``AGN'' continuum with the Big Blue
195: Bump peaking at $T=1.5\times10^5$~K, $\alpha_{\rm OX}=-1.43$,
196: $\alpha_{\rm UV}=-0.44$, $\alpha_{\rm X}=-0.9$.} and for a blackbody
197: continuum with a range of temperatures \citep[as proposed by][]{KP03}.
198: As a simple starting case, the winds presented here are optically
199: thin, with $N_{\rm H,0}=10^{22}~{\rm cm}^{-2}$; higher columns are
200: considered below.
201:
202: As expected, when $L/L_{\rm Edd}~<~1$, the wind does not reach $v_{\rm
203: obs} \sim v_{\rm esc}$, even with the added bound-free opacity used to
204: calculating the continuum acceleration. For $L/L_{\rm Edd}~\geq~1$,
205: winds can achieve the observed velocities. Figure~\ref{vInfPlot}
206: clearly shows that the blackbody continuum can also accelerate a wind,
207: resulting in almost identical velocities as when the \citet{RE04} SED
208: is used. This seems reasonable since the majority of $\Gamma_{\rm
209: cont}$ is given by frequency-independent electron scattering.
210:
211: Next, to test the sensitivity of these results to initial parameters,
212: both the column and the initial density of the wind are varied (see
213: Fig.~\ref{vInfPlotRE04}) for winds illuminated with the AGN SED from
214: \citet{RE04}. This shows that changing the initial density by an
215: order of magnitude (which increases $\dot{M}$ by the same factor) does
216: not change the velocity structure of the wind significantly, due to
217: the fact that the continuum is strong enough to highly ionize both
218: winds. For such highly-ionized continuum driven winds, the degree of
219: acceleration is governed by $n_{\rm e}/\rho$ which is nearly identical
220: for both winds (although the higher-density wind does have a lower
221: ionization parameter near the base of the wind, yielding larger
222: accelerations there than in the lower-density wind).
223:
224: Also in Figure~\ref{vInfPlotRE04}, the shielding column at the base of
225: the wind is varied from the normal $N_{\rm H,0}=10^{22}~{\rm cm}^{-2}$
226: to $N_{\rm H,0}=5\times10^{23}~{\rm cm}^{-2}$ to investigate the
227: ``self-shielding'' of a thicker wind. This thicker column also does
228: not change the final velocities in the outflow.
229:
230: \begin{figure}[ht]
231: \begin{center}
232: %\includegraphics[width=3.4in]{f1.eps}
233: \plotone{f1.eps}
234: \caption{The final velocity of continuum-driven wind models as a
235: function of the Eddington ratio of the central continuum for a range
236: of incident SEDs. For continuum driving in such highly ionized
237: sources, the final velocity is relatively independent of the SED. The
238: observed velocity of $v_{\rm obs}=0.08c$ in PG~1211+143 is shown for
239: comparison. Note that, for $L/L_{\rm Edd} \la 1$, winds do not
240: escape the system, but we plot their velocities at $s=100r_0$ to show
241: the trend as $L/L_{\rm Edd}$ increases.
242: \label{vInfPlot}}
243: \end{center}
244: \end{figure}
245:
246: \begin{figure}[h]
247: \begin{center}
248: %\includegraphics[width=3.4in]{f2.eps}
249: \plotone{f2.eps}
250: \caption{The final velocity of continuum-driven wind models
251: vs. $L/L_{\rm Edd}$ for the \citet{RE04} incident spectrum with
252: changes to the initial density and initial wind column. Unless
253: otherwise noted in the legend, the models have $N_{\rm
254: H,0}=10^{22}~{\rm cm}^{-2}$ and $n_0=5\times10^{12}~{\rm cm}^{-3}$.
255: Neither the increased initial column nor density affect the wind
256: significantly. The observed velocity of $v_{\rm obs}=0.08c$ in
257: PG~1211+143 is shown for comparison.
258: \label{vInfPlotRE04}}
259: \end{center}
260: \end{figure}
261:
262: These figures address the range of Eddington ratios necessary to
263: launch a wind to the observed velocity. But what are the particular
264: spectral signatures of such a wind? To answer this question,
265: Figures~\ref{xiPlot} and \ref{xiPlotRE04} show the X-ray ionization
266: parameter, $\xi$, as a function of $L/L_{\rm Edd}$ for those different
267: models. We define $\xi$ in the same manner as Cloudy and XSTAR: $4
268: \pi F_{\rm X}/n_{\rm H}$, where $F_{\rm X}$ is the integrated X-ray
269: flux from 1 to 1000 Rydbergs. To pick one `fiducial' $\xi$ out of the
270: range of ionization parameters in the accelerating wind, $\xi$ is
271: extracted at a distance along the flow of $s=3r_0$ (chosen to compare
272: with the inferred distance of the wind in PG~1211+143).
273:
274: Figures~\ref{xiPlot} and \ref{xiPlotRE04} display the ionization
275: parameters in these continuum-driven winds. They show that the high
276: velocity, continuum-driven winds must by necessity have been
277: accelerated by a large incident flux, which, for AGN-like continua,
278: leads to extremely high ionization parameters in the wind ($\xi$ of
279: order $10^5-10^6$ for $L/L_{\rm Edd}\geq1$). The $T=10^5~$K blackbody
280: spectrum also has a very high ionization parameter, but all of its
281: energy in the 1-1000 Rydberg window is concentrated near 1 Rydberg.
282: Thus, even though this continuum has a large ionization parameter, the
283: lack of X-ray photons means that the ionization stages within the gas
284: are much lower than that produced by an AGN SED: for instance, the
285: predominant state of iron for the wind illuminated by the $T=10^5~$K
286: blackbody spectrum is Fe~\textsc{ix}, while the observed iron ions are
287: Fe~\textsc{xxv} and Fe~\textsc{xxvi} \citep{Pounds03a}.
288:
289: \begin{figure}[ht]
290: \begin{center}
291: \plotone{f3.eps}
292: %\includegraphics[width=3.4in]{f3.eps}
293: \figcaption{The X-ray ionization parameter, $\xi$, as a function of
294: $L/L_{\rm Edd}$ for a range of incident continua at the inferred
295: location of the absorbing gas in PG~1211. The relative values of the
296: ionization parameter are set by the amount of flux each SED has in the
297: range of 1 to 1000 Rydbergs, where $\xi$ is calculated. The AGN SED
298: yields ionization parameters too high to explain the observed line
299: absorption. While the blackbody continua can have relatively high
300: ionization parameters, they have no flux in the X-ray regime to ionize
301: iron to the states observed in PG~1211+143.
302: \label{xiPlot}}
303: \end{center}
304: \end{figure}
305: \begin{figure}[h]
306: \begin{center}
307: \plotone{f4.eps}
308: %\includegraphics[width=3.4in]{f4.eps}
309: \figcaption{As in Fig.~\ref{xiPlot}, but using the \citet{RE04}
310: continuum to investigate variations in density and shielding. Unless
311: otherwise noted in the legend, the models have $N_{\rm
312: H,0}=10^{22}~{\rm cm}^{-2}$ and $n_0=5\times10^{12}~{\rm cm}^{-3}$.
313: The larger density wind yields a somewhat smaller $\xi$, even though
314: it experiences a very similar acceleration to the lower-density model.
315: The $N_{\rm H,0}=5\times10^{23}~{\rm cm}^{-2}$ wind has a very similar
316: ionization parameter for $L/L_{\rm Edd}\geq1$.
317: \label{xiPlotRE04}}
318: \end{center}
319: \end{figure}
320: This has important implications for observations of such winds.
321: \citet{KP03} postulated that continuum driven winds could account for
322: high-velocity X-ray absorption features. As mentioned in the
323: introduction, there have been other recent observations of such
324: high-velocity X-ray absorption in a variety of QSOs \citep{CBG03,
325: CBGG03, Pounds03b, ROW03}. In observations of PG~1211+143,
326: significant equivalent widths (EWs) for absorption features in
327: Mg~\textsc{xii}, S~\textsc{xvi}, Fe~\textsc{xxv}, and Fe~\textsc{xxvi}
328: are seen. Table~\ref{ewTable} compares these observations to models
329: of a high-column wind with an AGN SED \citep{RE04}, $N_{\rm H,0} =
330: 5\times10^{23}~{\rm cm}^{-2}$, assumed line widths of 1000 km/s
331: \citep{Pounds03a} and $L/L_{\rm Edd}~=~1$. Equivalent widths are
332: calculated in the linear weak line limit:
333: \begin{equation}
334: W_{\rm eV}=\Delta E_{\rm D}
335: \sqrt{\pi}\tau_0\times[1-\tau_0/(2!\sqrt{2})+\tau_0^2/(3!\sqrt{3})+...]
336: \end{equation}
337: for $\tau_0 \la 1$, where $\tau_0$ is the optical depth in the line
338: from Cloudy and $\Delta E_{\rm D}$ is the Doppler width of the line in
339: eV (all of the lines shown here have $\tau_0 \la 1$ in the wind).
340: %, or in the limit of saturation of the Doppler core,
341: %\begin{equation}
342: %W_{\rm eV}\approx\Delta E_{\rm D}\sqrt{\ln \tau_0}
343: %\end{equation}
344: %for $\tau_0\geq1$.
345: A comparison of observed EWs to the model EWs clearly shows that,
346: because of the high ionization in the outflow, such winds cannot
347: account for the EWs observed in PG~1211+143. The EWs for winds
348: illuminated with a blackbody spectrum are even smaller, although in
349: that case the small EWs are due to the lack of high-energy ionizing
350: X-rays.
351:
352: To test the particular parameters presented by \citet{Pounds03a}, we
353: assembled a model with a constant column of $N_{\rm
354: H}=5\times10^{23}~{\rm cm}^{-2}$ present throughout the wind.
355: Although the density still declines with height above the disk as
356: before, we set the column to a constant value for every line of sight
357: through the wind. This would therefore yield the column of $N_{\rm
358: H}=5\times10^{23}~{\rm cm}^{-2}$ at $s=3r_0$ as observed by
359: \citet{Pounds03a}. This model differs strongly from the other
360: solutions where the column drops consistently as the density in the
361: wind falls. The last column of Table~\ref{ewTable} displays the EWs
362: from this model with $L/L_{\rm Edd}=1$; at higher $L/L_{\rm Edd}$, the
363: EWs only decrease. Even with the constant column supplying a greater
364: column of the relevant ions, the ionization parameter in the model is
365: so high that the line equivalent widths fall below the 90\% confidence
366: interval of the measurements.
367:
368: We have searched for some area of parameter space that might allow
369: such continuum-driven winds to show the observed equivalent widths.
370: The only possibility we have found is modifying the incident continuum
371: such that $\alpha_{\rm x}=-2$ ($\Gamma = 3$). While inconsistent with
372: the observed power law of $\Gamma \sim 1.75$, this does yield
373: equivalent widths consistent with the observations for all lines
374: except Fe~\textsc{xxv}, but only for the wind with constant column of
375: $N_{\rm H} = 5\times10^{23}~$cm$^{-2}$ along the entire outflow.
376: However, in the rather implausible case that the X-ray continuum is
377: formed outside the wind (so that the wind sees a different X-ray
378: continuum than is observed), such a model does reproduce the observed
379: equivalent widths.
380:
381: As a final consideration, we note that for a large outflow with
382: covering factor $b=0.8$, P-Cygni profiles should be observed, but
383: emission is not seen with the observed absorption lines. This may
384: perhaps indicate a smaller covering fraction for the wind.
385:
386: \begin{deluxetable}{lccc}
387: \tabletypesize{\scriptsize}
388: %\rotate
389: \tablecaption{Comparison of equivalent widths for PG~1211+143
390: \citep[from][]{Pounds03a} to a continuum driven
391: wind.\label{ewTable}}
392: \tablehead{
393: \colhead{Line} & \colhead{Observed} &
394: \colhead{Model EW [eV]\tablenotemark{b}} &
395: \colhead{Model EW [eV]\tablenotemark{b}\tablenotemark{c}} \\
396: & \colhead{EW [eV]\tablenotemark{a}} &
397: \colhead{$N_{\rm H,0}=5\times10^{23}~{\rm cm}^{-2}$} &
398: \colhead{$N_{\rm H}=5\times10^{23}~{\rm cm}^{-2}$}
399: }
400: \startdata
401: Fe~\textsc{xxvi} Ly$\alpha$ & $95 \pm 20$ & $7\times10^{-4}$ & 31 \\
402: Fe~\textsc{xxv} 1s-3p & $45 \pm 12$ & $2\times10^{-5}$ & 0.2 \\
403: Fe~\textsc{xxvi} Ly$\beta$ & $45 \pm 12$ & $1\times10^{-3}$ & 8 \\
404: S~\textsc{xvi} Ly$\alpha$ & $32 \pm 12$ & $2\times10^{-4}$ & 1.4 \\
405: Mg~\textsc{xii} Ly$\alpha$ & $15 \pm 6 $ & $6\times10^{-5}$ & 0.4 \\
406: \enddata
407:
408: \tablenotetext{a}{Errors listed correspond to 90\% confidence
409: intervals.}
410: \tablenotetext{b}{Equivalent widths are those found at $s=3r_0$, the
411: inferred location of the sight-line through the outflow in PG~1211+143.}
412: \tablenotetext{c}{The column for this calculation is held constant
413: throughout the outflow, unlike all previous models, to compare with
414: \citet{Pounds03a}.}
415:
416: \end{deluxetable}
417:
418: \section{Summary}\label{Summary}
419:
420: This Letter addresses observational constraints on winds launched from
421: near the central black hole of AGNs using pure continuum driving.
422: This has become an increasingly significant topic very recently, with
423: the observation of high-velocity, very highly ionized gas in several
424: different QSOs, where the outflow is most likely too highly ionized to
425: allow for significant line driving. As expected, continuum driving
426: can accelerate highly-ionized winds from near the central black hole
427: to the observed velocities when $L/L_{\rm Edd} \geq 1$. But the
428: models also indicate that:
429:
430: \begin{enumerate}
431: \item Changes in the initial density or shielding column do not affect
432: the final velocity of the wind. Nor do changes in the illuminating
433: continuum significantly change the final velocities in the wind,
434: although moving to softer continua does significantly lower the
435: ionization state.
436: \item Continuum-driven winds, illuminated by an AGN SED and
437: accelerating to the observed velocities quickly become very highly
438: ionized (with $\xi\sim 10^5-10^6$ for $L/L_{\rm Edd} \geq 1$),
439: completely stripping the outflowing gas, resulting in acceleration
440: primarily by electron scattering. Continuum-driven winds can be
441: launched by blackbody SEDs but such outflows do not have the
442: observed high-ionization lines.
443: \item Taking the particular example of the outflow observed in
444: PG~1211+143 \citep{Pounds03a}, wind models display much smaller EWs
445: than the observed EWs. Softening the X-ray power law to $\Gamma
446: \sim 3$ can increase the EWs, however this requires both a
447: non-standard AGN continuum and a constant-column disk wind.
448: \end{enumerate}
449:
450: If pure continuum driving has difficulty reproducing the observations,
451: what process is accelerating the wind? First, perhaps the wind is not
452: relativistic \citep[see][who finds the low-energy X-ray spectrum of
453: PG~1211+143 to be consistent with a 3000 km/s outflow]{Kaspi04}. It
454: may also be possible that line-driving is playing a role, although
455: such acceleration would have to occur only at the very base of the
456: wind, where the ionization parameter is suitably low and where
457: radiative launching would be difficult given the high column to the
458: central source near the disk \citep[but see][]{PK04}. Another option
459: for explaining these outflows is a multi-phase wind, although such a
460: wind would increase the mass outflow rate beyond its already large
461: value.
462:
463: Another possibility is that such high-velocity, highly-ionized winds
464: could be launched hydromagnetically. The addition of hydromagnetic
465: driving would decouple the ionization state from the acceleration
466: process; instead of requiring, as for continuum-driven flows, that the
467: luminosity be $L_{\rm Edd}$ or greater (which leads to
468: over-ionization), the wind could possibly be accelerated by MHD forces
469: that would allow a lower $L/L_{\rm Edd}$ and a lower ionization state.
470: Indeed, a lower $L/L_{\rm Edd}$ may already be indicated by the larger
471: $M_{\rm BH}$ (albeit with large uncertainty) found by
472: \citet{Peterson04}. MHD wind models will be addressed in a subsequent
473: paper, but we point out that the large mass outflow rates ($\ga
474: \dot{M}_{\rm Edd}$) inferred in this source may be too high for an MHD
475: outflow. Model predictions for winds driven by all of the above
476: processes will be useful in determining which forces are capable of
477: launching these outflows and which model outflows are consistent with
478: the observational signatures.
479:
480: \acknowledgments
481:
482: We thank the referee for helpful comments that improved the paper, and
483: Doron Chelouche, Arieh K\"onigl, \& Norm Murray for their comments.
484: This work is supported by the Natural Sciences and Engineering
485: Research Council of Canada.
486:
487: \begin{thebibliography}{}
488:
489: \bibitem[Bottorff et al.(1997)]{B97}Bottorff, M., Korista, K.T.,
490: Shlosman, I., \& Blandford, R.D. 1997, ApJ, 479, 200
491: \bibitem[Chartas, Brandt \& Gallagher(2003)]{CBG03}Chartas, G.,
492: Brandt, W.N., \& Gallagher, S.C. 2003, ApJ, 595, 85
493: \bibitem[Chartas et al.(2003)]{CBGG03}Chartas, G., Brandt, W.N.,
494: Gallagher, S.C., \& Garmire, G.P. 2002, ApJ, 579, 169
495: \bibitem[Chelouche \& Netzer(2003)]{CN03}Chelouche, D., \& Netzer,
496: H. 2003, MNRAS, 344, 233
497: \bibitem[Crenshaw, Kraemer, \& George(2003)]{CKG03}Crenshaw, D.M.,
498: Kraemer, S.B., \& George, I.M 2003, ARA\&A, 41, 117
499: %\bibitem[Feldmeier \& Shlosman(1999)]{FS99}Felmeier, A., \& Shlosman,
500: % I. 1999, ApJ, 526, 344
501: \bibitem[Ferland(2002)]{Ferland02}Ferland, G.J. 2002, Hazy, a Brief
502: Introduction to Cloudy, University of Kentucky Department of Physics
503: and Astronomy Internal Report
504: \bibitem[Kapsi et al.(2000)]{Kaspi00}Kaspi, S., Smith, P.S., Netzer,
505: H., Maoz, D., Jannuzi, B.T., \& Giveon, U. 2000, ApJ, 533, 631
506: \bibitem[Kaspi(2004)]{Kaspi04}Kaspi, S. 2004, in The Interplay among
507: Black Holes, Stars, and the ISM in Galactic Nuclei, eds. Th. Shorchi
508: Bergmann, L.C. Ho, \& H.R. Schmitt, astro-ph/0405563
509: \bibitem[King \& Pounds(2003)]{KP03}King, A.R., \& Pounds
510: K.A. 2003, MNRAS, 35, 657
511: \bibitem[K\"onigl \& Kartje(1994)]{KK94}K\"onigl, A., \& Kartje,
512: J.F. 1994, ApJ, 424, 446
513: \bibitem[Krolik \& Kriss(1995)]{KK95}Krolik J.H., \& Kriss G.A. 1995, ApJ, 447,
514: 512
515: \bibitem[Murray et al.(1995)]{MCGV95}Murray N., Chiang J., Grossman
516: S.A., \& Voit G.M. 1995., ApJ, 451, 498
517: \bibitem[Peterson et al.(2004)]{Peterson04}Peterson, B. et al. 2004,
518: Accepted to ApJ, astro-ph/0407299
519: \bibitem[Pounds et al.(2003a)]{Pounds03a}Pounds, K.A., Reeves, J.N.,
520: King, A.R., Page, K.L., O'Brien, P.T., \& Turner, M.J.L. 2003, MNRAS,
521: 345, 705
522: \bibitem[Pounds et al.(2003b)]{Pounds03b}Pounds, K.A., King, A.R.,
523: Page, K.L., \& O'Brien, P.T. 2003, MNRAS, 346, 1025
524: \bibitem[Proga \& Kallman(2004)]{PK04}Proga, D., \& Kallman, ApJ
525: accepted (astro-ph/0408293)
526: \bibitem[Proga, Stone, \& Kallman(2000)]{PSK00}Proga D., Stone J.M.,
527: \& Kallman T.R., 2000, ApJ, 543, 686
528: \bibitem[Reeves, O'Brien, \& Ward(2003)]{ROW03}Reeves, J.N., O'Brien,
529: P.T., \& Ward, M.J. 2003, ApJ, 593, L65
530: \bibitem[Risaliti \& Elvis(2004)]{RE04}Risaliti, G., \& Elvis, M. 2004
531: in Supermassive Black Holes in the Distant Universe, ed. A.J. Barger
532: (Dordrecht:Kluwer Academic Publishers), astro-ph/0403618
533:
534: \end{thebibliography}
535:
536:
537: \end{document}
538:
539:
540: