astro-ph0210642/ms.tex
1: \documentstyle[12pt,aaspp]{article}
2: %\documentstyle[aaspp]{article}
3: 
4: 
5: \begin{document}
6: %\baselineskip=28pt
7: \def\MSUN{\rm M_{\odot}}
8: \def\RSUN{\rm R_{\odot}}
9: \def\MSUNYR{\rm M_{\odot}\,yr^{-1}}
10: \def\MDOT{\dot{M}}
11: 
12: \newbox\grsign \setbox\grsign=\hbox{$>$} \newdimen\grdimen \grdimen=\ht\grsign
13: \newbox\simlessbox \newbox\simgreatbox
14: \setbox\simgreatbox=\hbox{\raise.5ex\hbox{$>$}\llap
15:      {\lower.5ex\hbox{$\sim$}}}\ht1=\grdimen\dp1=0pt
16: \setbox\simlessbox=\hbox{\raise.5ex\hbox{$<$}\llap
17:      {\lower.5ex\hbox{$\sim$}}}\ht2=\grdimen\dp2=0pt
18: \def\simgreat{\mathrel{\copy\simgreatbox}}
19: \def\simless{\mathrel{\copy\simlessbox}}
20: %
21: 
22: \title{Numerical simulations  
23: of mass outflows driven from accretion disks by  radiation  and 
24: magnetic forces.}
25: 
26: \vspace{1.cm}
27: \author{ Daniel Proga }
28: \vspace{1.cm}
29: \affil{JILA, University of Colorado, Boulder, CO 80309-0440, USA;
30: proga@colorado.edu}
31: 
32: 
33: \begin{abstract}
34: 
35: We study the two-dimensional, time-dependent magnetohydrodynamics (MHD)
36: of  radiation-driven winds  from luminous accretion disks initially threaded
37: by a purely axial magnetic field. The radiation force is mediated primarily 
38: by spectral lines and is calculated  using a generalized multidimensional 
39: formulation of the Sobolev approximation. We use ideal MHD to compute 
40: numerically the evolution of Keplerian disks, varying the magnetic field 
41: strengths and the luminosity of the disk, the central accreting object  
42: or both. 
43: We find  that the magnetic fields very quickly start deviating
44: from purely axial due to the magnetorotational 
45: instability. This leads to fast growth of the
46: toroidal magnetic field as field lines wind up due to the disk rotation.
47: As a result the toroidal field dominates over the poloidal field
48: above the disk and the gradient of the former drives a slow and dense disk 
49: outflow, which conserves specific angular momentum. 
50: Depending on the strength of the magnetic field relative to the 
51: system luminosity the disk wind can be radiation- or MHD driven.
52: The pure radiation-driven wind consists of a dense, slow outflow that is 
53: bounded on the polar side by a high-velocity  stream. The mass-loss rate is 
54: mostly due to the fast stream. As the magnetic field strength increases first 
55: the slow part of the flow is affected, namely it becomes  denser and slightly 
56: faster and begins to dominate the mass-loss rate. In very strong magnetic 
57: field or pure MHD cases, the wind consists of only a dense, slow outflow 
58: without the presence of the distinctive fast stream so typical to pure 
59: radiation-driven winds. Our simulations indicate that winds launched by 
60: the magnetic fields are likely to remain dominated by the fields downstream 
61: because of their relatively high densities. The radiation force due 
62: to lines may not be able  to change a dense MHD wind because the line force
63: strongly decreases with increasing density.
64: 
65: 
66: 
67: \end{abstract}
68: 
69: \keywords{ accretion disks -- galaxies: nuclei -- binaries: close
70: -- MHD -- methods: numerical} 
71: 
72: \section{Introduction}
73: 
74: Powerful mass outflows from accretion disks are observed in many 
75: astrophysical environments such as active galactic nuclei (AGN); 
76: many types of interacting binary stars, e.g., 
77: non-magnetic cataclysmic variables (nMCVs); and young stellar objects (YSOs).  
78: Magnetic fields, the radiation force and thermal expansion  
79: have been suggested as mechanisms that can drive disk winds.
80: These three mechanisms have been studied extensively 
81: using analytic as well as  numerical methods. 
82: As a result of these studies, theoretical models have been
83: developed that allow us to estimate under what physical conditions
84: each of these mechanisms is efficient in launching, accelerating
85: and collimating disk outflows. The notion that one universal physical
86: mechanism can explain all outflows
87: from accretion disks is very appealing.
88: However, theoretical studies and observational results indicate
89: that none of the three mechanisms alone is sufficient and
90: probably such a single mechanism does not exist.
91: Therefore it make sense to
92: consider a hybrid model in which more than one mechanism
93: is involved. 
94: 
95: 
96: 
97: Magnetically driven winds from disks are the favored explanation for 
98: the outflows in many astrophysical environments.  Blandford \& Payne (1982) 
99: (see also Pelletier \& Pudritz 1992) showed that  the centrifugal force can  
100: drive a wind from  the disk if the poloidal component of the magnetic
101: field, ${\bf B_p}$ makes an angle of $> 30^o$ with respect to the normal to 
102: the disk surface. Generally, centrifugally-driven MHD  disk winds 
103: (magnetocentrifugal winds for short) require  the presence of a sufficiently
104: strong, large-scale, ordered magnetic field threading the disk 
105: with a poloidal component at least comparable to the toroidal magnetic
106: field, $|B_\phi/B_p| \simless 1$ (e.g., Cannizzo \& Pudritz 1988, 
107: Pelletier \& Pudritz 1992). Several groups have studied
108: numerically axisymmetric outflows using the Blandford \& Payne mechanism
109: (e.g., Ustyugova et al. 1995, 1999; Romanova et al. 1997; 
110: Ouyed \& Pudritz 1997a, 1997b, 1999; Krasnopolsky, Li \& Blandford 1999; 
111: Kato, Kudoh \& Shibata 2002).
112: An important feature of  magnetocentrifugal winds is that they require 
113: some assistance  to flow freely and steadily from the surface of the 
114: disk, to pass through a slow magnetosonic surface 
115: (e.g., Blandford \& Payne 1982). 
116: The numerical studies mentioned above do not resolve the vertical structure 
117: of the disk but treat it as a boundary surface through which mass 
118: is loaded on to the magnetic field lines at a specified rate.
119: 
120: Winds from disks can  driven by the magnetic pressure.
121: In particular, the toroidal magnetic field can quickly builds up
122: due to the differential rotation of the disk so that $|B_\phi/B_p| \gg 1$.
123: In such a case, the magnetic pressure of the toroidal field can give rise 
124: to a self-starting wind (e.g., Uchida \& Shibata 1985; 
125: Pudritz \& Norman 1986; Shibata \& Uchida 1986; Stone \& Norman 1994;
126: Contopoulos 1995; Kudoh \& Shibata 1997; Ouyed \& Pudritz 1997b).
127: To produce a steady outflow  driven by the magnetic pressure 
128: a steady supply of advected toroidal magnetic flux at the wind base 
129: is needed, otherwise the outflow is likely to be transient 
130: (e.g., K$\ddot{\rm o}$nigl 1993, Contopoulos 1995, Ouyed \& Pudritz 1997b). 
131: It is still not clear whether the differential rotation of the disk can 
132: produce such a supply of the toroidal magnetic flux to match 
133: the escape of magnetic flux in the wind and even 
134: if it does whether such a system will be stable  
135: (e.g., Contopoulos 1995, Ouyed \& Pudritz 1997b and references therein).
136: 
137: One of the reasons for favoring magnetic fields as an explanation
138: for mass outflows from accretion disks is the fact that magnetic fields 
139: are very likely crucial for the existence of all accretion disks.  
140: The magnetorotational instability (MRI)
141: (Balbus \& Hawley 1991; and earlier by Velikov 1959 and Chandrasekhar 1960)
142: has been shown to be  a very robust and universal mechanism to produce
143: turbulence and the transport of angular momentum in disks at all radii
144: (Balbus \& Hawley 1998). 
145: It is therefore likely that magnetic fields control mass accretion inside 
146: the disk and play a key role in producing a mass outflow from the disk. 
147: However, it has been demonstrated observationally and theoretically that
148: accretion disks are capable of losing mass also via  a radiation-driven wind, 
149: provided the disk luminousity in ultraviolet (UV) is high
150: enough.
151: 
152: Radiation-driven disk winds have been extensively modeled recently 
153: (e.g., Pereyra, Kallman \& Blondin 1997; Proga, Stone \& Drew 1998, 
154: hereafter PSD~98; Proga 1999; Proga, Stone \& Drew 1999, hereafter PSD~99; 
155: Feldmeier \& Shlosman 1999; Feldmeier, Shlosman \& Vitello 1999; 
156: Proga, Stone \& Kallman 2000; Pereyra, Kallman \& Blondin
157: 2000; Proga \& Kallman 2002).
158: These studies showed that radiation pressure due to spectral lines 
159: can drive winds from luminous disks. This result
160: has been expected (e.g., Vitello \& Shlosman 1988). 
161: These studies, in particular those by PSD~98,  
162: also showed some unexpected 
163: results. For example, the flow is unsteady in cases where the disk
164: luminosity dominates the driving radiation field.
165: Despite the complex structure of the unsteady disk wind,
166: the time-averaged mass loss rate and terminal velocity scale
167: with luminosity, as do steady flows obtained where the
168: radiation is dominated by the central object. In 
169: the most favorable conditions (i.e., high
170: UV flux and low X-ray flux) the radiation force due to spectral lines
171: (the line force) can exceed the radiation
172: force due to electron scattering by a factor as high as $\sim 2000$
173: (e.g., Castor Abbott, \& Klein 1995, hereafter CAK; Abbott 1982; Gayley
174: 1995). Thus systems with UV luminosity, $L_{UV}$ as low as  a few $10^{-4}$ 
175: of their Eddington limit, $L_{Edd}$, can produce a powerful high velocity wind.
176: 
177: Generally, one can argue 
178: that in all accretion disks, 
179: with $L_{UV} \simgreat$ a few $10^{-4} L_{Edd}$
180: mass outflows have been observed (Proga 2002). For example, accretion disks
181: around: massive black holes, white dwarfs (as in AGN and nMCVs with
182: $\Gamma_{UV} \ga 0.001$) and low mass young stellar objects 
183: (as in FU Ori stars with $\Gamma_{UV} \ga$ a few $\times$ 0.01) show 
184: powerful fast winds.
185: Systems that have too low  UV luminosities  to drive a wind 
186: include accretion disks around neutron stars and low mass black holes 
187: as in low mass X-ray binaries (LMXBs) and galactic black holes. 
188: These systems indeed 
189: do not show outflows similar to those observed in nMCVs, AGN and FU Ori. 
190: However, outflows, even in systems which appear to be luminous enough 
191: to produces radiation-driven disk winds,
192: cannot be fully explained by just line driving.
193: 
194: For example, Drew \& Proga (1999) applied results from pure line-driven 
195: (LD for short) disk wind models to nMCVs. In particular, 
196: they compared mass loss rates predicted by 
197: the  models with observational constraints. Drew \& Proga (1999) concluded 
198: that either mass accretion rates in high-state nMCVs are higher than presently 
199: thought  by a factor of 2-3 or that radiation pressure alone is not quite 
200: sufficient to drive the observed hypersonic flows. The difficulty in accounting
201: for the mass loss rate in a pure LD disk wind model for nMCVs is simply 
202: a reflection of the fact that the nMCV luminosities just barely satisfy 
203: the basic requirement, i.e., $L_{UV}\simless 7 \times 10^{-4} L_{Edd}$. 
204: Synthetic line profiles computed based on pure LD wind models confirmed 
205: Drew \& Proga's conclusion (Proga et al. 2002). 
206: Our study has been partially motivated by this conclusion
207: because if indeed radiation pressure alone does not suffice to drive 
208: the observed hypersonic flow then an obvious candidate to assist radiation 
209: pressure in these cases is MHD (e.g., Drew \& Proga 1999).
210: 
211: In this paper, we study how  magnetic fields can change disk winds driven 
212: by the line force for a given disk luminosity. 
213: We assume in our models that the transport 
214: of angular momentum in the disk is dominated by local disk viscosity, 
215: for instance due to the MRI in weakly magnetized disks 
216: (Balbus \& Hawley 1998). Here we  add  magnetic fields to the PSD~99 model 
217: and solve self-consistently the full set of ideal MHD equations.
218: 
219: The outline of this paper is as follows. We describe our calculations in 
220: Section 2. We present our results and discuss their perceived 
221: limitations in Section 3. The paper ends in Section~4,  with our conclusions.
222: 
223: \section{Method}
224: 
225: \subsection{Equations and Numerical Techniques}
226: To calculate the structure and evolution of a wind from a disk, we solve 
227: the equations of ideal MHD
228: \begin{equation}
229:    \frac{D\rho}{D t} + \rho {\bf \nabla \cdot  v} = 0,
230: \end{equation}
231: \begin{equation}
232:    \rho \frac{D{\bf v}}{Dt} = 
233: - {\bf \nabla P} - \rho {\bf \nabla} \Phi
234: + \frac{1}{4\pi} {\bf (\nabla \times B) \times B}
235:  + \rho {\bf F}^{rad} 
236: \end{equation}
237: \begin{equation}\label{eqn:induction}
238: {\partial{\bf B}\over\partial t} = {\bf\nabla\times}({\bf v\times B}).
239: \end{equation}
240: Here the convective derivative $D/Dt$ is
241: equivalent to $\partial/\partial t + {\bf v\cdot\nabla}$.
242: The dependent quantities $\rho$, ${\bf v}$, and $P$ are gas mass
243: density, velocity, and scalar isotropic gas pressure,
244: respectively, and ${\bf B}$ is the magnetic field.
245: The gas in the wind is  isothermal with a sound speed $c_s$.
246: We calculate gravitational acceleration using the Newtonian potential $\Phi$.
247: The term ${\bf F}^{rad}$ in the equation of motion (2) is
248: the total radiation force per unit mass.
249: We solve these equations in spherical polar coordinates
250: $(r,\theta,\phi)$, assuming axial symmetry about the rotational axis
251: of the accretion disk ($\theta=0^\circ$)
252: 
253: The geometry and assumptions needed to compute the radiation field from 
254: the disk and central object are as in PSD~99 (see also PSD~98).
255: The disk is flat, Keplerian, geometrically-thin and optically-thick.  
256: We specify the radiation field of the disk by assuming that the local disk
257: intensity follows the radial profile of the so-called $\alpha$-disk 
258: (Shakura \& Sunyaev 1973), and therefore depends only on the mass accretion 
259: rate in the disk, $\dot{M}_a$, and the mass  and radius of the central object, 
260: $M_\ast$  and $r_\ast$.  In particular, the disk luminosity, 
261: $L_D \equiv GM_\ast \MDOT_a/2r_\ast$. In models where the central object  
262: radiates, we take into account the irradiation of the disk, assuming that 
263: the disk re-emits all absorbed energy locally and isotropically. We express 
264: the central object luminosity $L_\ast$ in units of the disk luminosity 
265: $L_\ast=x L_D$. 
266: 
267: Our numerical algorithm for evaluating the line force is described in PSD~99.  
268: Here we briefly describe the key elements of our calculations of the line
269: force. We use the  CAK force multiplier 
270: to calculate the line-driving force.
271: In this approximation, a general form for this force  
272: at a point defined by the position vector $\bf r$ is
273: \begin{equation}
274: {\bf F}^{rad,l}~({\bf{r}})=~\oint_{\Omega} M(t) 
275: \left(\hat{n} \frac{\sigma_e I({\bf r},\hat{n}) d\Omega}{c} \right),
276: \end{equation}
277: where $I$ is the frequency-integrated continuum intensity in the direction
278: defined by the unit vector $\hat{n}$, and $\Omega$ is the solid angle
279: subtended by the disk and central object at the point W. 
280: The term in brackets is the electron-scattering radiation force, where 
281: $\sigma_e$ is  the mass-scattering coefficient for free electrons,
282: and $M(t)$ is the force multiplier -- the numerical factor which
283: parameterizes by how much spectral lines increase the scattering
284: coefficient. In the Sobolev approximation, $M(t)$ is a function
285: of the optical depth parameter
286: \begin{equation}
287: t~=~\frac{\sigma_e \rho v_{th}}
288: { \left| dv_l/dl \right|},
289: \end{equation}
290: where $v_{th}$ is the thermal velocity and 
291: $d v_l/dl$ is the velocity gradient along $\hat{n}$. 
292: The velocity gradient can be written as 
293: \begin{equation}
294: \frac{dv_l}{dl}=Q~\equiv~ \sum_{i,j}\frac{1}{2}\left(\frac{\delta v_i}{\delta r_j}
295: +\frac{\delta v_j}{\delta r_i}\right)n_in_j=\sum_{i,j}e_{ij}n_in_j,
296: \end{equation}
297: where $e_{ij}$ is the symmetric rate-of-strain tensor.
298: Expressions for the components of $e_{ij}$ in spherical polar coordinates
299: are given in Batchelor (1967).
300: 
301: We adopt the CAK  analytical expression
302: for the force multiplier as modified by Owocki, Castor \& Rybicki 
303: (1988, see also PSD98)
304: \begin{equation}
305: M(t)~=~k t^{-\alpha}~ 
306: \left[ \frac{(1+\tau_{max})^{(1-\alpha)}-1} {\tau_{max}^{(1-\alpha)}} \right]
307: \end{equation}
308: where $k$ is proportional to the total number of lines,
309: $\alpha$ is the ratio of optically thick to optically-thin lines,
310: $\tau_{max}=t\eta_{max}$, and $\eta_{max}$ is a parameter 
311: related to the opacity of the most optically thick lines.
312: The term in square brackets is the Owocki, Castor \& Rybicki correction
313: for the saturation of $M(t)$ as the wind becomes optically thin
314: even in the strongest lines, i.e., 
315: \begin{displaymath}
316: \lim_{\tau_{max} \rightarrow 0} M(t)~=~M_{max}~=~
317: k(1-\alpha)\eta_{max}^\alpha.
318: \end{displaymath}
319: 
320: 
321: We discretize the $r-\theta$ domain into zones in our calculation of
322: the wind structure. Our resolution in the $r$ and 
323: $\theta$ directions is sufficiently high  to ensure that the subsonic 
324: portion of the model outflow is sampled by at least a few grid points in 
325: both $r$ and $\theta$. This requirement and the nature of the problem combine 
326: to demand an increasingly fine mesh toward the disk plane: here the density 
327: declines dramatically with height, and, moreover, the velocity in the wind
328: increases rapidly.  Our numerical resolution consists of 100 zones in each of 
329: the $r$ and $\theta$ directions, with fixed zone size ratios, 
330: $dr_{k+1}/dr_{k}=d\theta_{l}/d\theta_{l+1} =1.05$.  
331: 
332: The boundary conditions for the hydrodynamic dependent quantities 
333: are specified as follows. At $\theta=0$, we apply an axis-of-symmetry 
334: boundary condition. For the outer radial boundary, we apply an outflow 
335: boundary condition.  For the inner radial boundary $r=r_{\ast}$ and for 
336: $\theta=90^o$, we apply reflecting boundary conditions.
337: 
338: 
339: Our simulations begin with a vertical, independent of $r$ magnetic field 
340: configuration and a Keplerian disk embedded in the rotating ambient medium 
341: of low density. The ambient medium is initially outflowing 
342: with the escape velocity  at $r=r_\ast$ in the direction
343: perpendicular to the disk midplane. Below we give the details of our initial
344: conditions and conditions in the first grid zone above the equatorial 
345: plane follows. 
346: 
347: We proceed with setting the initial conditions in the following way.
348: We start with adopting the rotational velocity
349: $v_\phi =\sqrt\frac{GM_\ast}{r \sin \theta}$ for $r \sin{\theta}> r_\ast$
350: and $v_\phi =0$ elsewhere. Thus the gas above the disk
351: rotates on cylinders with the disk Keplerian velocity whereas
352: the gas above the non-rotating central object has zero rotational 
353: velocity. 
354: 
355: 
356: Our initial density profile is given by the condition of hydrostatic 
357: equilibrium in the latitudinal direction for a gas with
358: a given initial rotational velocity. To ensure an exact numerical
359: equilibrium initially, we first initialize the density
360: in the first grid zone above the equatorial 
361: plane, $\rho_0$. We assume $\rho_0$ is radius independent.
362: Then we integrate the latitudinal equation of motion from $\theta=90^\circ$
363: to $\theta=0^\circ$ to compute the pressure using the numerical difference
364: formula in our  code. Finally, we compute the density
365: from the isothermal equation of state.
366: To reduce the problems caused by very high 
367: Alfv${\acute{\rm e}}$nic velocities 
368: in regions of very low density
369: (i.e., to prevent the time step from being prohibitively small),
370: we set a lower limit to the density on 
371: the grid  as $\rho_{min}(r)=10^{-15} (r_o/r)^2~\rm 
372: g~cm^{-3}$ and enforce it at all times in all models.
373: 
374: For the initial poloidal velocity, we adopt 
375: $v_r = \sqrt{GM/r} \sin{\theta}$ and
376: $v_\theta = \sqrt{GM/r} \cos{\theta}$ for the region where 
377: the lower limit to the density
378: is  used (see above). This choice for the poloidal
379: velocity is motivated by practical concerns 
380: (i.e., again to prevent the time step from being prohibitively small) 
381: and is particularly useful
382: for models with no radiation force due to the central object.
383: For $x=0$, the gas above the disk collapses onto the disk and 
384: the time step becomes so small that the simulations practically stop.
385: 
386: To represent steady conditions in the photosphere at the base 
387: of the wind, during the evolution of each model we continued to apply 
388: the constraint that in the first zone above the equatorial plane 
389: the density is fixed at $\rho = \rho_0$ at all times.
390: During the evolution of our standard models,
391: $\rho_0$ was fixed at  $10^{-4}\rm g~cm^{-3}$.
392: 
393: There are several differences between our initial and boundary conditions 
394: for the hydrodynamic dependent quantities 
395: and those we used in previous work (PSD~98 and PSD~99).
396: The most important difference is in our treatment of the wind base.
397: In PSD~99,  to represent steady conditions in the photosphere at the base 
398: of the wind, during the evolution of each model we continued to apply 
399: the constraints that in the first zone above the equatorial plane 
400: the radial velocity $v_r=0$, the rotational velocity $v_\phi$ remains 
401: Keplerian, and the density is fixed at $\rho = \rho_0$ at all times.  
402: Physically, $\rho_0$ is analogous to the density in the photosphere 
403: of the disk at the base of the wind provided $\rho_0$ is relatively low. 
404: In PSD~99 and PSD~98, the interior of the disk itself was treated as 
405: negligibly thin and was 
406: excluded from the models (for a disk temperature of $10^{4}$~K at 
407: $r=2 r_\ast$, the disk scale height $H$ is $H/r_\ast \sim 10^{-3}$). 
408: In PSD~98 and PSD~99, the arbitrary value  for $\rho_0$ was fixed
409: typically at $10^{-9} \rm g~cm^{-3} $.  As discussed in 
410: PSD~98,  the gross properties of  LD winds are unaffected by the 
411: value of $\rho_0$ provided it is large enough that the acceleration of the 
412: wind up to the sonic point is resolved with at least a few grid points.
413: This technique, when applied to calculations of spherically symmetric
414: LD winds from stars, produces a solution that relaxes to the
415: appropriate CAK solution within a few dynamical crossing times.
416: However, applying these constraints in MHD simulations can cause
417: very strong  evolution of  the gas close to the equator and 
418: consequently the evolution of the mass outflow.
419: We have performed many tests and found that,
420: when the above constraints are used, the disk gets
421: disrupted very quickly (within a couple orbits at $r=r_\ast$) by  MRI,
422: making it impossible to study disk winds.
423: Inclusion of magnetic fields in the model puts
424: new constraints on  $\rho_0$.
425: For strong magnetic fields and low $\rho_0$ (i.e., a small value of 
426: the plasma parameter $\beta\equiv 8\pi P/ B^2$ on the equatorial plane), 
427: the MRI or magnetic braking 
428: can very quickly reduce the density near the disk midplane.
429: In particular, we observe a dramatic drop in the density, of 
430: several orders of magnitude,
431: between the first and second zone above the equatorial plane.
432: This change in the density profile
433: makes impossible to study outflows from a `steady state' disk.
434: By choosing a large $\rho_0$ for a given magnetic field, we can reduce 
435: the dynamical importance of the magnetic field in
436: the region close to the midplane so that the base of the wind can remain 
437: in a steady state for a long time.
438: For these reasons, our value of $\rho_0$ is 
439: $\rho_0=10^{-4}\rm g~cm^{-3}$ and during the evolution of our standard models
440: we allow all the dependent quantities (including $\rho$, $v_r$, and 
441: $v_\phi$) to float everywhere on the grid.
442: For the adopted value of $\rho_0$, the plasma parameter on the 
443: equator, $\beta_0\equiv 8\pi P/ B^2=8\pi c^2_s \rho_0/ B^2$ is 
444: very high for all standard models 
445: (i.e., $ 2\times 10^5 \leq \beta_0 \leq 2 \times 10^{11}$).
446: In Section~3.3, we discuss 
447: how our results depend on the conditions
448: along the equatorial plane, in particular 
449: what difference  it makes if 
450: $\rho_0\leq10^{-9}\rm g~cm^{-3}$, 
451: and $\rho$, $v_r$,  and $v_\phi$ are set as in PSD~99.
452: 
453: We are left with describing our initial magnetic configuration and the 
454: boundary conditions for the magnetic field.
455: We assume a force-free configuration (Lorentz force
456: $\bf{ (\nabla \times B) \times B}=0$) by simply setting
457: $\bf (\nabla \times B)=0$. We consider one straightforward
458: initial magnetic configuration satisfying these constraints:
459: a uniform vertical field configuration defined by the
460: magnetic potential ${\bf A} = (A_r=0, A_\theta=0, A_\phi= A r \sin\theta)$.
461: We scale the magnitude of the magnetic field
462: using a parameter, $\beta'_w \equiv 32 \pi^2 c_s^2 \rho_w/B^2$ 
463: defined for a fiducial wind density of
464: $\rho_w= 10^{-15}\rm g~cm^{-3}$:
465: \begin{equation}
466: A =2~\pi~\sqrt{(2 c^2_s \rho_w/\beta'_w)}.
467: \end{equation}
468: 
469: The boundary conditions for the magnetic field are: at
470: $\theta=0$, we apply an axis-of-symmetry boundary condition;  for
471: the outer radial boundary, we apply an outflow boundary 
472: condition.  For the inner radial boundary $r=r_{\ast}$
473: we apply reflecting boundary conditions
474: while for $\theta=90^o$ we apply an equatorial-symmetry boundary condition
475: (Stone \& Norman 1992b).
476: In our standard models we allow all three components of
477: the magnetic field to float. 
478: 
479: In reality, gas near the disk midplane (inside the disk) is turbulent because 
480: of the MRI but in near hydrostatic equilibrium. We would like to stress that, 
481: although we include the region very close to the midplane in
482: our standard models, we do not claim that we model the disk 
483: interior. To do the latter we would need to add physical processes such 
484: as magnetic field dissipation and radiative transfer appropriate to
485: optically thick disk gas, and to solve the equation of energy. 
486: We would also need to resolve better the disk so we could capture, 
487: for example, the fastest growing modes of the MRI. 
488: The most unstable wavelength of the MRI, $\lambda \sim 2 \pi v_A/\Omega$,
489: increases with height in the disk for a given angular velocity, 
490: $\Omega$, because the Alfv${\acute{\rm e}}$n speed increases with height. 
491: In general,  near the equator, we deal with a magnetized stratified disk
492: where magnetic field is generated either by the MRI or magnetic braking.
493: It is computationally prohibitive to resolve adequately a thin disk in global
494: calculations even in two dimensions. Nevertheless, as we will discuss 
495: in section~3.3, our simulations are consistent with  high resolution local 
496: simulations of magnetized stratified disks (e.g., Stone et al. 1996; 
497: Miller and Stone 2000). 
498: 
499: To solve eqs. (1)-(3), we use the ZEUS-2D code described by 
500: Stone \& Norman (1992a, 1992b).
501: 
502: \subsection{Model Parameters}
503: 
504: As in PSD~99 and Proga (2000), we calculate disk winds with model parameters 
505: suitable for a typical nMCV (see Table~1 in PSD~99, Table~1 in Proga 2000 
506: and our Table~1). We vary the disk and central object luminosity and 
507: the strength of magnetic field (i.e., $\beta'_w$). We hold all other 
508: parameters fixed: $M_\ast=0.6~\MSUN$, $r_\ast=8.7\times10^8~\rm{cm}$,
509: $c_s=14~\rm{km~s^{-1}}$, $k=0.2$, $\alpha=0.6$ and $M_{max}=4400$
510: (see Table~1 in PSD~98). 
511: Nevertheless we can use our results to predict the wind 
512: properties for other parameters and  systems -- such as AGN and YSOs -- 
513: by applying the dimensionless form of the hydrodynamic equations and 
514: the scaling formulae as discussed in PSD~98 and Proga (1999).
515: We note that by adopting the uniform vertical field configuration,
516: our model has  two free parameters more than the pure LD
517: disk wind model of PSD~99: the plasma parameter $\beta'_w$ and
518: the initial toroidal magnetic field, $B_\phi$. In this paper, we
519: hold $B_\phi=0$ but vary $\beta'_w$. 
520: 
521: Summarizing, our model has three crucial independent parameters: 
522: the Eddington factor corresponding to the disk luminosity
523: $\Gamma_D\equiv L_D/L_{Edd}=\frac{\sigma_e \MDOT_a}{8\pi c r_\ast }$;
524: the central
525: object luminosity expressed in terms of the disk luminosity,
526: $x=L_D/L_\ast$, and the strength of the magnetic
527: field expressed in terms of the plasma parameter
528: for a fiducial density of $\rho_w= 10^{-15}\rm g~cm^{-3}$, $\beta'_w$.
529: As in PSD~99,
530: we assume that all the radiation is emitted in the ultraviolet and does
531: not evolve.
532: The model also has other parameters that can be 
533: calculated self-consistently, in principle, 
534: for a given set of $\Gamma_D$, $x$, $\beta'_w$ 
535: and the spectral distribution of the radiation field.
536: These parameters are: the sound speed $c_s$, and 
537: the parameters of  
538: the CAK force multiplier, $k$, $\alpha$ and $M_{max}$.
539: For the pure LD case, the wind solution
540: does not depend on the value of the sound speed (see CAK for 
541: LD stellar winds and  Proga 1999 for LD disk winds). 
542: The basic requirement for the strength of the disk radiation
543: to drive a wind is: 
544: \begin{equation}
545: \Gamma_D [1+ M_{max}] > 1
546: \end{equation}
547: (e.g., PSD~98; Proga 1999; 
548: see also Proga \& Kallman 2002 and Proga 2002 for a discussion
549: of the generalized version of this requirement).
550: Therefore,  $k$ and $\alpha$ do not matter 
551: as much as the value of $M_{max}$.
552: 
553: \section{Results}
554: 
555: Pure LD winds from a disk fall into two categories: 
556: 1) intrinsically unsteady with large fluctuations in density and velocity, 
557: and 2) steady with smooth density and velocity  (PSD~98 and PSD~99). 
558: The type of flow is set by the geometry of the radiation field, 
559: parametrized by $x$: if the radiation field is dominated by the disk ($x<1$) 
560: then the flow is unsteady, and if the radiation is dominated by the central 
561: object ($x\simgreat 1$) then the flow is steady. The geometry of the radiation 
562: field also determines the geometry of the flow: the wind becomes more polar 
563: as $x$ decreases. However, the mass-loss rate and terminal velocity 
564: are insensitive to geometry and depend more on the total system luminosity, 
565: $L_D+L_\ast$. Regardless of the type of  flow,  pure LD
566: winds consist of a dense, slow outflow that is bounded on the polar side by a
567: high-velocity  stream. The mass-loss rate is mostly due to the fast stream.
568: 
569: In Proga (2000), we recalculated some of the PSD~99 models to check 
570: how inclusion of the magnetocentrifugal force, corresponding
571: to purely poloidal $\bf{B}$, will change LD disk winds. 
572: We found that flows which conserve  specific angular velocity have a larger 
573: mass loss rate  than their counterparts with purely LD flow, 
574: which conserve specific angular momentum. The difference in the mass loss rate 
575: between  winds conserving specific angular momentum and those conserving
576: angular velocity can be several orders of magnitude for low disk luminosities 
577: but vanishes for high  disk luminosities. Winds which conserve angular 
578: velocity have much higher velocities than angular momentum conserving winds.
579: In Proga (2000), we  also found that fixing the wind geometry stabilizes winds 
580: which are unsteady when the geometry is derived self-consistently.
581: Additionally, as expected, the inclination angle $i$ 
582: between the poloidal velocity 
583: and the normal to the disk midplane is important. 
584: Non-zero inclination angles allow the magnetocentrifugal force to increase
585: the mass loss rate for low luminosities, and increase the wind velocity
586: for all luminosities. 
587: 
588: In this paper, we also recalculate some of the PSD~99 flows,  but this time 
589: we check, by solving the  full set of ideal MHD equations and allowing 
590: $B_\phi \neq 0$,  how {\it self-consistent} inclusion of magnetic fields to 
591: the PSD~99 model will change LD disk winds. We summarize the properties of 
592: PSD~99 models and our new simulations in Table~1. Columns (2) to (4) give 
593: the input parameters that we varied: the mass accretion rate $\MDOT_a$, 
594: the relative luminosity of the central object, $x$, and the plasma parameter
595: $\beta'_w$, respectively. Column (5) lists the final time at which we stopped 
596: each simulation (all times here are in units of 
597: $\tau=\sqrt{r^3_\ast/G M_\ast}=~2.88~ \rm sec$). Columns (6) to (8) give
598: some the gross properties of the disk wind: the wind mass loss rate, 
599: $\MDOT_w$, the wind velocity at the outer radial boundary, $v_r(10 r_\ast)$, 
600: and the wind half-opening angle, $\omega$, respectively.
601: We measure $\omega$ from the equator to the upper envelope of the wind.
602: Table~1 also  contains comments regarding some runs [column (9)]
603: and explains  our convention of  labeling our runs.
604: 
605: \subsection{Outflow from a luminous magnetized accretion disk}
606: 
607: In this section we describe the properties
608: and behavior of our model MHD-LD C0D in which 
609: $\MDOT_a~=~10^{-8}~\MSUNYR$, $x=0$,  $\beta'_w=2\times10^{-2}$. 
610: This model is a rerun of the fiducial `$x=0$' unsteady model
611: discussed in detail in PSD~98 and PSD~99.
612: 
613: Figure~1 presents a sequence of maps showing density, velocity field and
614: toroidal magnetic field (left, middle and  right panels) from model C0D, 
615: plotted in the $r,z$ plane.  The length of the arrows in the middle panels is
616: proportional to $(v_r^2 + v_\theta^2)^{1/2}$. To show better the evolution
617: of the wind with lower velocities we use the maximum lenght of the arrows 
618: in regions of high velocity, i.e., $(v_r^2 + v_\theta^2)^{1/2}\ge 
619: 200~\rm km~s^{-1}$.  We also suppress velocity vectors in regions
620: of low density (i.e., $\rho$ less than $10^{-15}~\rm g~cm^{-3}$).
621: The pattern of the direction of the arrows is an 
622: indication of the shape of the instantaneous streamlines.  The solid lines in 
623: the middle panels mark the location where the poloidal Alfv${\acute{\rm e}}$n 
624: speed, $v_{Ap}\equiv B_p/\sqrt{4\pi \rho}$, equals the poloidal fluid speed, 
625: $v_p$ (i.e. the Alfv${\acute{\rm e}}$n surface). As in the pure LD case, after 
626: $\sim 10$ time units disk material fills the grid for $\theta \simgreat 30^o$ 
627: and remains in that region for the rest of the run. In the early phases of 
628: the evolution the MHD-LD disk wind resembles its pure LD counterpart except 
629: that it is steadier. However, in the late phases the flow undergoes a dramatic 
630: change not seen in the LD case: slow and dense material rises from 
631: the disk at large radii. 
632: In the slow wind, the density increases whereas the velocity
633: decreases with time until the wind settles to a time-averaged 
634: steady state. These changes in the slow wind occur on relatively 
635: a long time scale and are caused by the evolution of the magnetic field. 
636: In particular, $B_\phi$ is generated near the base of the wind
637: as indicated by the rise of the $B_\phi$ contours in Figure~1 
638: (the strength of $B_\phi$ decreases with $z$). 
639: The gradient of $B^2_\phi$ drives the slow wind 
640: from the disk at large radii. 
641: At the end of the simulation, the LD wind 
642: is replaced by a wind driven primarily by the magnetic pressure. 
643: 
644: 
645: Figure~2  presents the density and poloidal velocity field in the wind at 
646: the end of our simulations at 680~$\tau$ (top left and middle panels). 
647: The dashed and solid line in the top right panel of Figure~2 shows 
648: the contours of the angular velocity $\Omega$ and specific angular momentum 
649: of fluid, $L\equiv r \sin{\theta} v_\phi$, respectively.
650: The bottom left, middle and right panels of Figure~2 show
651: the contours of the $\beta$ plasma parameter, 
652: the poloidal magnetic field and the contours of the toroidal magnetic
653: field, $B_\phi$, respectively. The solid lines in the bottom middle
654: panel show the location where the strength of the poloidal magnetic 
655: field equals  the toroidal magnetic field, $\mid B_p \mid =\mid B_\phi \mid$.
656: 
657: Comparing our model C0D with its pure LD counterpart, we find that model C0D 
658: has a relatively smooth density distribution and  is relatively steady. 
659: For example, the pure LD outflow is intrinsically unsteady and characterized 
660: by large amplitude velocity and density  fluctuations.  Infall as well as 
661: outflow from a disk can occur in different regions of the wind at 
662: the same time. However, in model ~C0D  the poloidal velocity
663: is mostly organized and there is no inflow onto the disk in the wind
664: domain. Model~C0D has not
665: reached a steady state even after $400~\tau$. However, we observe that 
666: this model is  close to reaching such a state; for example, the mass flux 
667: density settles to some time-averaged maximum.
668: 
669: As in the pure LD case, we find that the wind consists of 
670: a dense, slow outflow that is bounded on the polar side by a
671: high-velocity  stream. However, there is an important difference, namely that
672: the dense, slow outflow is significantly denser and somewhat faster
673: in our model C0D than the pure LD case. 
674: Finally, as one could have expected for an MHD wind, 
675: the gas pressure dominates over the magnetic
676: pressure in the region very close to the mid-plane ($\beta \gg 1$) 
677: whereas the opposite is true in the wind domain ($\beta \ll 1$,
678: see bottom left panel).
679: 
680: 
681: The above differences in the two wind solutions appear to result from
682: the Lorentz force due to the gradient of the  toroidal magnetic field. 
683: Our simulations start with zero $B_\phi$ but this situation changes very 
684: quickly as toroidal field is generated by  rotation. We note that 
685: ${\bf \nabla} B^2_\phi$  is higher than the line force in the wind domain 
686: (expect for the fast stream) by a couple of orders of magnitude. Comparing 
687: the contours of $B_\phi$ (bottom right panel of Fig. 2) and the poloidal 
688: velocity field (top middle panel), we find that the velocity field is normal 
689: to the  $B_\phi$ contours (i.e., $v_p$ is parallel to the $B^2_\phi$ gradient)
690: in the slow wind (lower right-hand corner of each panel). In contrast 
691: to the slow wind, in the fast stream where the gas is driven by 
692: the line force and the magnetic pressure is unimportant, $v_p$ is tangent 
693: to the $B_\phi$ contours.
694: 
695: The quantities presented in Figures~1 and 2 show that the outflow in model~C0D
696: is not a magnetocentrifugal wind. There are several diagnostics of 
697: magnetocentrifugal driving. For example, in the acceleration zone
698: of a steady state magnetocentrifugal wind $B_p \simgreat B_\phi$. 
699: However in model C0D, the location where $\mid B_p \mid =\mid B_\phi \mid$ 
700: is near the disk and 
701: $B_p < B_\phi$ in most of the acceleration zone 
702: (see lower middle panel in Figure~2).
703: Additionally, in a magnetocentrifugal wind the total angular
704: momentum per unit mass, 
705: $ l \equiv v_\phi r \sin{\theta}- r \sin{\theta} B_\phi B_p/(4\pi \rho v_p)$, 
706: the second term due to the twisted magnetic fields is comparable
707: to the first term.
708: Subsequently, the wind is corotating with the underlying disk
709: up to approximately the Alfv${\acute{\rm e}}$n point. In model ~C0D, 
710: the Alfv${\acute{\rm e}}$nic 
711: surface is very close to the disk (see top middle panel) 
712: and the total specific angular momentum is mostly due to the fluid
713: even near the wind base. Thus the wind is corotating with the disk only
714: over a relatively small length scale. The top right panel
715: of Figure~2 is helpful to distinguish between a magnetocentrifugal
716: wind and a fluid angular momentum conserving wind. In the former,
717: the angular velocity is conserved along the streamlines below 
718: the Alfv${\acute{\rm e}}$n point while in the latter, $L$ is conserved along 
719: the streamlines. Comparing the top right panel 
720: and the top middle panel of Figure~2, we  clearly
721: see that the contours of $L$,  not of $\Omega$, are aligned with
722: the streamlines represented by 
723: the arrows of the poloidal velocity field.
724: 
725: Next we consider the angular dependence of the flow at large radii.
726: Figure~3 shows the angular dependence of 
727: density, radial velocity, mass flux density, and accumulated mass loss rate 
728: at $r = r_o = 10 r_{\ast}$ at the end of the simulation of model~C0D.
729: The accumulated mass loss rate is given by:
730: \begin{equation}
731: d\dot{m}(\theta) =
732: 4 \pi r_o^2 \int_{0^o}^\theta \rho v_r \sin \theta d\theta.
733: \end{equation}
734: The gas density is a very strong function of angle for $\theta$ between
735: $90^o$ and $25^o$.  Between the disk mid-plane at $\theta = 90^o$ and 
736: $\theta \sim 85^o$, $\rho$ drops by $\sim 7$ orders of magnitude, as  expected
737: for a density profile determined by hydrostatic equilibrium. 
738: For $25^o \simless \theta \simless 85^o$, the wind domain, $\rho$ varies 
739: between $10^{-16}$ and $10^{-11}~\rm g~cm^{-3}$.  For $\theta \simless 25^o$, 
740: density  again decreases rapidly, but this time to so low a value that 
741: it becomes necessary to replace it by the numerical lower limit $\rho_{min}$.
742: The region with $\rho \le \rho_{min}$ is not relevant to our analysis as it 
743: has no effect on the disk flow.  The radial velocity at $10r_\ast$
744: increases gradually from  zero at the equator to 
745:  $\sim 100~\rm km~s^{-1}$ at $\theta\approx60^o$, then it drops to nearly zero
746: at $\theta\approx60^o$.
747: Over the angular range  $65^o > \theta >
748: 25^o$, $v_r$ increases from $\leq 0$ up to $1200~\rm km~s^{-1}$.
749: 
750: The cumulative mass loss rate is negligible for $\theta \simless 35^o$
751: because of the very low prevailing gas density.  Beginning at $\theta 
752: \simgreat 35^o$, $d\dot{m}$ increases to $\sim 3\times 10^{13}~\rm g~s^{-1}$ 
753: at $\theta \approx 55^o$. This increase of  $d\dot{m}$
754: is due to the fast stream. Then, in the slow dense outflow, 
755: the cumulative mass loss rate increases to $\sim 7\times 10^{14}~\rm
756: g~s^{-1}$ at $\theta \approx 86^\circ$. For even higher $\theta$, 
757: in the region close to 
758: the disk plane, where the gas density starts to rise very sharply and where 
759: the motion is subsonic and  typically more complex, 
760: the cumulative mass loss rate is 
761: subject to enormous fluctuations (some of which may even be negative).
762: In the example shown in Figure~3, the total mass loss rate through the outer 
763: boundary, $\MDOT_{tot}=$ $d\dot{m}(90^o)$ reaches $\sim  10^{19}~\rm 
764: g~s^{-1}$!  This value of $\MDOT_{tot}=$ 
765: is most certainly dominated by the contribution from 
766: the slow-moving region very close to the disk mid-plane -- a contribution 
767: that is very markedly time-dependent. We ignore the value of $\MDOT_{tot}$
768: as it is related more to subsonic oscillations of the 'disk' rather
769: than to the supersonic wind we model. In the remaining part of the
770: paper, for the wind mass loss rate we use the value of 
771: the cumulative mass loss rate at $\theta=82^\circ$.
772: 
773: We note  that the increase of the mass loss rate
774: in the slow wind is totally due to magnetic fields and  there
775: is no enhancement of the line force in this region. In fact,
776: the opposite is true -- the line force in the dense slow region in model~C0D
777: is significantly reduced by the action of the magnetic fields.
778: The latter increases the density of the outflow which in turn
779: reduces the line force. The slight increase in the velocity 
780: and its gradient in the slow wind is far from compensating
781: the reduction of the line force due to 
782: the increase of the density.
783: 
784: We conclude that magnetic fields can change qualitatively
785: and quantitatively a radiation-driven disk wind. In particular,
786: the magnetic pressure can dominate the driving of the  wind and reduce
787: the role of the line force. In model ~C0D, we find
788: that a disk loses mass via a LD wind 
789: in the inner part of the wind (the fast stream) and via a MHD-driven  wind
790: in the outer part of the wind (the dense slow wind).
791: We expect  that increasing the relative strength of the magnetic pressure 
792: to the radiation pressure 
793: (e.g., by reducing $\beta'_w$ and therefore $\beta_0$)
794: should lead to
795: the entire wind domain being driven by the magnetic force.
796: Conversely, decreasing the magnetic pressure should lead to 
797: the wind being driven by the line force.  Therefore we consider next a limited 
798: parameter survey to check  whether our expectation is correct
799: and to see how   the wind solution
800: changes  quantitatively with the strength of the radiation and magnetic fields.
801: 
802: \subsection{Parameter survey}
803: 
804: We consider only the parameter space of our models that will 
805: define the major trends in disk wind behavior.  
806: We focus on a survey of how the mass loss rate, outflow velocity and geometry
807: change with disk luminosity, relative central object luminosity and
808: strength of magnetic field. In Section~3.3 we will consider
809: how our results depend on our treatment of the conditions in the first
810: grid zone above the equatorial plane.
811: 
812: In Figure~4 we show (a) the wind mass loss rate, $\MDOT_w$, as a 
813: function of the total Eddington factor, $(1+x)\Gamma_D$ 
814: (note that $(1+x)\Gamma_D$ is proportional $\MDOT_a$ 
815: for a given $x$) for various $\beta'_w$ and (b) $\MDOT_w$, as a 
816: function of $\beta'_w$,  for  various $(1+x)\Gamma_D$.  
817: The top panel of Figure~4 also shows 
818: $\MDOT_w$ as a function of $(1+x)\Gamma_D$ 
819: as predicted by the pure LD wind model with and without taking into account
820: the fact that the force multiplier has a maximum value.
821: The thick  solid line corresponds to the prediction where $M(t)$ can
822: be arbitrarily high whereas the thin solid line corresponds
823: to the prediction where $M(t)$ reaches maximum at $M_{max}=4400$
824: (see Proga 1999).
825: In the top panel of  Figure~4, it can be seen that  
826: $\MDOT_w$ is a very strong function of $(1+x)\Gamma_D$ for $\beta'_w=\infty$.  
827: As shown in PSD~98 and Proga~(1999), two dimensional models
828: of LD disk winds predict mass loss rates (as well as velocities)
829: very similar to those predicted by the original CAK formulae
830: when the stellar Eddington factor is replaced with 
831: the the total Eddington factor and the formulae
832: are corrected for the fact that for small Eddington factors
833: the force multiplier reaches its maximum and consequently
834: the radiation force cannot exceed gravity.
835: Note that the drop in $\MDOT_w$ occurs 
836: for $(1+x)\Gamma_D\approx 0.0002=1/M_{max}$.
837: 
838: 
839: Motivated by the conclusion from section~3.1, we have
840: performed a few simulations using different values of $\beta'_w$.
841: We would like to estimate a  $\beta'_w$ range for which
842: the line force dominates and magnetic fields play a small
843: role,  and a range for which the line force is unimportant
844: and the wind is totally controlled by magnetic fields.
845: 
846: We here focus on simulations for a fiducial `$x=0$' model with 
847: $\MDOT_a=10^{-8}~\MSUNYR$
848: and a fiducial `$x=1$' steady state model with $\MDOT_a=\pi \times 
849: 10^{-8}~\MSUNYR$ and various $\beta'_w$.
850: For both series of simulations 
851: (i.e., series `C0' corresponding to runs labeled 
852: C0A, C0B, C0D, C0E, and C0F
853: and series `D1' corresponding to runs labeled 
854: D1A, D1B, D1D, D1E, D1F, and D1G, see Table~1),
855: we find that indeed for sufficiently high $\beta'_w$ the wind mass loss 
856: rate and characteristic velocity 
857: are very similar for those in the pure LD counterparts. For example,
858: we note that for models of series~D1,
859: the symbols corresponding to $\MDOT_w$ with $\beta'_w=2, 2\times10^{-2}$,
860: and $2\times10^{-3}$ overlap with the the symbol corresponding
861: to the pure LD model D1A (see also the bottom panel of Figure~4). However,
862: there is a qualitative difference between the MHD-LD winds
863: and the LD wind even for high $\beta'_w$ cases: the MHD-LD winds
864: are less unsteady than the LD winds. For high $\beta'_w$, the Lorentz
865: force can be too weak to drive a wind but if $\beta\ll 1$ 
866: in the wind (as in the cases we explored), the gradient of the
867: magnetic pressure can reduce the density fluctuation
868: near the wind base and in the wind.
869: 
870: The bottom panel of Figure~4 shows more clearly the point we made above
871: that for high $\beta'_w$, $\MDOT_w$ is as for 
872: pure LD cases and does not change with $\beta'_w$. 
873: However, starting 
874: from a certain value of $\beta'_w$, $\MDOT_w$ becomes a strong
875: function of $\beta'_w$. The value of $\beta'_w$ at which
876: $\MDOT_w$ starts to increase with decreasing $\beta'_w$ 
877: is higher for the lower luminosity system (the solid line) 
878: than for the higher luminosity system (the dashed line).
879: Generally, 
880: the strong $\MDOT_w$ dependence on $(1+x)\Gamma_D$ almost disappears
881: when strong magnetic fields are added to the model and this
882: is clear in the top panel, where $\MDOT_w$ becomes insensitive to
883: $\Gamma_D$ as $\beta'_w$ decreases.
884: 
885: As Table~1 shows, the wind geometry is also sensitive to $\beta'_w$:
886: the half-opening angle of the wind $\omega$, increases with decreasing
887: $\beta'_w$ so
888: the wind becomes more polar as $\beta'_w$ decreases. We note that
889: even the `$x=1$' models become polar for $\beta'_w,2\times10^{-4}$
890: (see models D1F and D1G).
891: 
892: Simulations for fixed $\beta'_w$ and $x$, but with varying the system 
893: luminosity,  help us to check whether the magnetic fields
894: can drive a wind for disk luminosities too low to the drive a LD wind.
895: These simulations also help us to check whether the line force
896: can `regain' control over the wind  if the system luminosity increases.
897: 
898: For the system luminosity  of model B1A, the line force 
899: is too weak to produce a supersonic outflow (e.g., PSD~99). However, 
900: in model B1F with the same radiation field as in model B1A, there is a strong 
901: robust disk outflow. 
902: In the models B1F and D1F with the same low $\beta'_w$
903: but different system luminosities, the wind is mostly driven 
904: by  MHD. However in model E1F, with the system luminosity higher
905: than in model D1F by a factor of 10 
906: the wind is LD and very similar to that with zero magnetic field 
907: (compare model E1F with model E1A).
908: 
909: We conclude that MHD driving is robust and does not require the line
910: force, e.g., 
911: the MHD driving does not require the line force to launch a wind from 
912: the disk. We find a negative feedback between 
913: the magnetic field and the line force, i.e., the higher the magnetic field 
914: the lower the line force and vice versa. 
915: We also conclude that although  MHD driving can produce
916: a strong wind it does so by driving a relatively dense wind with relatively 
917: low velocities. The latter do not seem to depend on the escape velocity
918: from the radius from which the wind is launched. 
919: We note that $v_r$ of the slow wind depends very weakly on $\theta$
920: at $r_o$. On the other hand, the line
921: force drives a relatively fast wind with a velocity sensitive
922: to the launching radius and consequently to $\theta$ (PSD~98).
923: 
924: \subsection{Dependence of wind evolution and properties on the treatment
925: of the wind base}
926: 
927: Our model for mass outflows from accretion disks is a hybrid
928: of an LD model  and an MHD-driven flow.  We have extensively tested our
929: LD disk wind model and applied it to several systems (see references in
930: Section~1). In this section we present a brief review of  our
931: test runs designed to check the MHD part of our model.
932: In particular, we have performed a few tests aimed at reproducing
933: qualitatively results already published on MHD disk winds.
934: Additionally, we have explored a parameter region of our model
935: ($0.1\simless \beta_0 \simless 10$) for which we expect to resolve
936: the fastest growing modes of the MRI.
937: 
938: There have been many numerical studies of MHD disk winds 
939: (our list of references  in the introduction is far from complete).  
940: For simplicity of our presentation we will reference
941: in more detail to the work of only two groups: Stone \& Norman (1994)
942: and Ouyed \& Pudritz (1997a; 1997b; 1999). The former
943: included the disk as well as the wind in their simulations whereas
944: the latter included only the wind and treated the disk as the lower
945: boundary of the wind. Both approaches have been commonly 
946: adopted in the literature.
947: 
948: Clearly, inclusion of the disk structure in calculations of disk outflow
949: is highly desirable. However, the physics of the disk is very complex
950: and its proper modeling is very demanding. On the other hand, one
951: would hope that it is possible to capture the key elements of a disk
952: wind without modeling the disk interior. 
953: Work by Ouyed \& Pudritz (1997a; 1997b; 1999) is an example
954: of studies of  magnetocentrifugal disk winds while PSD~98, PSD~99 
955: is an example
956: of studies of LD disk winds where the disk interior was not included.
957: However there is an important difference between modeling
958: a magnetocentrifugal disk wind and an LD disk wind, namely the treatment
959: of the lower boundary for the disk. In numerical simulations of 
960: magnetocentrifugal disk winds the mass flux density from the disk
961: is given whereas in numerical simulations of LD disk winds 
962: the mass flux density is a result. The key reason for this difference
963: is the location of the critical surface at which the mass flux density
964: is determined. In MHD simulations, this surface corresponds to the 
965: slow magneto-sonic surface which is located inside the disk, below the
966: photosphere. Thus unless the disk is included in simulations, 
967: one  must assume the mass flux density from the lower boundary. 
968: On the other hand, the critical surface in LD disk winds
969: is in the supersonic part of the wind above the photosphere.
970: Therefore, including the photosphere of the disk but not the whole disk
971: suffices to capture to the transition between the sub-critical and 
972: super-critical parts
973: of the LD wind and subsequently  the mass flux density is determined
974: by the physics of the flow 
975: (i.e., Feldmeier \& Shlosman 1999).
976: 
977: We have performed test simulations of pure MHD and  MHD-LD disk winds 
978: following the approach used in studies of magnetocentrifugal winds. 
979: Our test runs included those where we 
980: set $v_r=0$, 
981: $v_\phi=v_{Keplerian}$, $\rho=\rho_0$ and $B_\phi=0$
982: in the first grid zone above the equatorial 
983: plane at all times. This setting is very similar to the one
984: used in numerical simulations of magnetocentrifugal disk winds
985: (Ouyed \& Pudritz 1997a; 1997b; 1999) with the exception that
986: they also set the velocity in the direction perpendicular
987: to the midplane, $v_z$, to some small subsonic velocity 
988: (i.e., the mass flux density $\rho v_z$ is fixed).
989: In the test runs, we allow $v_\theta$
990: to float and let the mass flux density be fixed by the solution
991: to the problem. 
992: 
993: In short, we found such an approach to modeling MHD and MHD-LD disk winds 
994: unsatisfactory. For example, in pure MHD cases insisting on a Keplerian
995: flow in the first grid zone above the equatorial 
996: plane at all times does not allow proper  modeling
997: of  magnetic braking of the disk. By
998: setting $v_r=0$, we prevent or at least significantly reduce
999: the collapse of the disk  and dragging of magnetic field line 
1000: (e.g., Ouyed \& Pudritz 1997a).
1001: However, when we allowed all variables to float we could successfully
1002: reproduce the evolution of a collapsing disk and an MHD outflow.
1003: In particular, in strong field cases, the disk undergoes dramatic evolution
1004: on a short time scale (i.e., of order of $40~\tau$; in other words, a
1005: couple of orbital periods at $r_\ast$), consistent with the time scale
1006: for magnetic braking of an aligned rotator (Mouschovias \& Paleologou 1980). 
1007: 
1008: Similarly, we found that the evolution of the disk in a pure MHD
1009: case depends on the conditions in the first grid zone above the equatorial 
1010: plane for weak magnetic fields. We performed several test simulations 
1011: with weak magnetic field, so that the disk was unstable to MRI. In particular,
1012: we adopted $\rho_0=10^{-9} ~\rm g~cm^{-3}$ and $\beta'_w=8\times10^7$ 
1013: yielding $\beta_0=6.4$,
1014: the parameters for which the fastest growing MRI mode is resolved
1015: for our stratified disk. 
1016: For quantities in the first grid zone above the equatorial 
1017: plane set as in magnetocentrifugal models, we found that the disk
1018: evolved very quickly. Initially, we observed a characteristic
1019: exponential growth of the magnetic field, which causes the disk
1020: to separate vertically into horizontal planes. This so-called
1021: channel solution is typical of the development of the axisymmetric
1022: phase of the MRI (Hawley \& Balbus 1991; Goodman \& Xu 1994). 
1023: As expected, the gas streaming is directed  outward in the channel closest 
1024: to the equator and inwards in the channel  higher up from the equator.
1025: However, after about three orbital periods the disk is destroyed.
1026: We find many similarities in the  behavior of the disk 
1027: in our simulations and in previous simulations. 
1028: For example, Stone et al. (1996) and Miller \& Stone (2000)
1029: reported that in their three-dimensional simulations of a 
1030: stratified disk (a local shearing box approximation),
1031: an initially uniform vertical magnetic field leads to a strong radial
1032: streaming in the early phases of the evolution. However, in the late phases, 
1033: high magnetic pressures disrupt the disk and 
1034: the entire computational domain is magnetically dominated, i.e.,
1035: $\beta < 1$ even near the equatorial plane. The high
1036: magnetic pressures are created by strong large-scale radial 
1037: streaming that occurs at large heights because in a stratified disk,
1038: most unstable wavelength increases with height.
1039: As in those local three dimensional simulations, we observed that
1040: strong radial streaming occurs in the horizontal planes into which our disk
1041: separates. Miller \& Stone (2000) concluded that enforcing
1042: Keplerian rotation in the boundary conditions is
1043: appropriate for following the long-term evolution.
1044: We have arrived at the same conclusion, that it is simply impossible to
1045: model on a sufficiently long-time scale an outflow from a disk 
1046: that is quickly disrupted. We found that when
1047: we do not enforce  Keplerian rotation, more precisely when we
1048: allow all the fluid and magnetic field quantities to float, then 
1049: the disk evolves more slowly with radial streaming. 
1050: However when we add the line force and assume too low a density 
1051: the disk can be totally disrupted because of the LD wind.
1052: 
1053: 
1054: Motivated by the results from the above test simulations, 
1055: for our standard models, we decided to allow all quantities to float 
1056: in the first grid zone above the equatorial 
1057: plane so that they can evolve self-consistently in all models in Tabel~1.
1058: To ensure we retain a disk for long enough to acquire
1059: a reasonably settled outflow, we
1060: increased  $\rho_0$ from $10^{-9}~\rm g~cm^{-3}$ as in PSD~99
1061: to $10^{-4}~\rm g~cm^{-3}$.
1062: The latter change means that we do not resolve the fastest growing MRI
1063: mode in the disk. To do so we would have to increase
1064: the strength of the magnetic field for a given 
1065: resolution  or increase the resolution. Unfortunately, we cannot afford 
1066: either of these modifications
1067: because they would make our already computationally
1068: demanding simulation even more demanding.
1069: Therefore, in this first attempt to model MHD-LD disk winds
1070: we resolve  the wind from the disk but do not resolve 
1071: the disk itself. Thus we cannot resolve the fastest growing 
1072: MRI mode within the disk scale height, but only 
1073: most unstable modes at larger  heights.
1074: 
1075: We finish with a remark that in the literature 
1076: most of the numerical simulations
1077: of MHD disk winds were stopped after one or two or at most a few
1078: orbital periods. The reason for this is obvious but it is 
1079: disappointing that one cannot perform yet global simulations
1080: of magnetized disks, treating properly accretion as well as outflow
1081: and choosing physical parameters comparable to those in real systems.
1082: 
1083: \subsection{Limitations of models and future work}
1084: 
1085: The most important limitation of our model is an inadequate spatial
1086: resolution for modeling the MRI inside the disk 
1087: (see discussion above).  As is fitting for a first exploration of MHD-LD
1088: wind  models from  disks, we aim to examine only the parameter space of 
1089: our models that will define the major trends in disk wind behavior. 
1090: Therefore our priority is to set up the simulation in such a way that 
1091: the base of the wind is relatively stable and corresponds to a steady state 
1092: accretion disk. Obviously the problem with modeling MRI disappears when 
1093: the magnetic field is strong. In the context of MHD winds, a strong
1094: magnetic field case for which the disk is MRI-stable 
1095: is also  physically interesting.
1096: 
1097: We have explored  MHD-LD models where  initially $\beta < 1$
1098: everywhere on the grid. However, we found then Alfv${\acute{\rm e}}$n speeds
1099: so high that the resulting time step was extremely small.
1100: We emphasize that our choice of $\beta'_w$ is limited by the constraints
1101: on the density in the computational domain, i.e.,
1102: the need to have a relatively large contrast between the density
1103: near the equatorial plane and that high above  the plane. 
1104: Increasing arbitrarily the lower limit, $\rho_{min}$, to reduce 
1105: the density contrast and subsequently reduce the Alfv${\acute{\rm e}}$n speed
1106: is not suitable  to modeling LD winds. The line force is very sensitive to 
1107: the density and our wind solution can then include gas with a spurious 
1108: density set by the lower limit.
1109: Thus we were very cautious in choosing $\rho_{min}$, so that
1110: the region with $\rho \le \rho_{min}$ 
1111: has no effect on the disk flow.
1112: 
1113: We note that other numerical simulations explored
1114: a relatively low density contrast between the disk and the ambient
1115: gas (e.g., in Stone \& Norman 1994; Ouyed \& Pudritz 1997a, 1997b, 1999;
1116: and Krasnopolsky, Li \& Blandford 1999 the density contrast is $\simless
1117: 10^3$). 
1118: Here we deal with density contrasts several orders of magnitude higher. 
1119: 
1120: The fact that our results strongly depend on the magnetic field
1121: points to a need to explore different configurations for the initial
1122: magnetic field and 
1123: to move from two-dimensional axisymmetric simulations
1124: to fully three-dimensional simulations.
1125: We are interested in the long-time evolution of the flow. Therefore
1126: three-dimensional simulations are required as there exist no self-sustained
1127: axisymmetric dynamos. Thus, contrary to the outflows from stars,
1128: simulations of outflows from magnetized disks -- with or without radiation
1129: pressure -- should include the disks themselves, not just
1130: the disk photosphere, and should be performed
1131: in three dimensions.
1132: 
1133: \section{Conclusions}
1134: 
1135: We have studied winds from accretion disks with magnetic fields and 
1136: the radiation force due to lines. We use numerical methods to solve 
1137: the two-dimensional, time-dependent equations of  ideal MHD. We have 
1138: accounted for the radiation force using a generalized multidimensional 
1139: formulation of the Sobolev approximation. For the initial conditions,
1140: we have considered uniform vertical magnetic fields and geometrically
1141: thin, optically thick Keplerian disks. We allow the magnetic field
1142: and the disk to evolve. In particular, we do not enforce  Keplerian 
1143: rotation in the first grid zone above the equatorial plane. Although the gas 
1144: near the  equatorial plane departs only slightly from  Keplerian rotation
1145: in our self-consistent calculations, 
1146: its evolution is notably different when we enforce  Keplerian rotation.
1147: We find  that the magnetic fields very quickly start deviating from purely 
1148: vertical due to the MRI. This leads to fast growth of the toroidal magnetic 
1149: field as field lines wind up due to the disk rotation. As a result, 
1150: the toroidal field dominates over the poloidal field above  the disk and 
1151: the gradient of the former drives a slow and dense disk outflow,
1152: which conserves specific angular momentum.
1153: Depending on the strength of the magnetic field relative to the system 
1154: luminosity, the disk wind can be radiation- or MHD driven. 
1155: For example, for our model parameters $\MDOT_a~=~10^{-8}~\MSUNYR$ and $x=0$, 
1156: the wind is radiation-driven for $\beta'_w\simgreat 1$ and, as   
1157: its pure LD counterpart, consists of a dense, slow outflow that 
1158: is bounded on the polar side by a high-velocity  stream. 
1159: The mass-loss rate is mostly due to the fast stream. 
1160: As the magnetic field strength increases (i.e., $\beta'_w <10^{-1}$),  
1161: first the slow part of the flow is affected. In particular,
1162: the slow wind  becomes  denser  and faster than its pure LD 
1163: counterpart by a factor of $\simgreat 100$ and $\simgreat 3$, respectively.
1164: Consequently, the dense wind begins to dominate  the mass-loss rate. 
1165: In very strong magnetic field (i.e., $\beta'_w \simless 10^{-3}$) 
1166: or pure MHD cases, the wind consists of only 
1167: a dense, slow outflow without the presence of the distinctive fast 
1168: stream so typical of pure LD winds. Our simulations indicate 
1169: that winds launched by magnetic fields are likely to remain dominated 
1170: by the fields downstream because of their relatively high densities. 
1171: The radiation force due to lines may not be able to change a dense MHD wind 
1172: because the line force strongly decreases with the density.
1173: 
1174: Our results show that, as expected, a hybrid model predicts mass-loss rates 
1175: higher than those predicted by a pure LD model. 
1176: We plan to compute synthetic  line profile based on our simulations
1177: and check whether our MHD-LD models can resolve the problem of nMCV winds.
1178: However even now we can say that our MHD-LD models may not solve the problem
1179: because 
1180: to explain nMCV winds we need a model that predicts not only 
1181: a higher $\MDOT_w$ than a LD wind model but also $\MDOT_w$ must be mostly 
1182: due to a fast wind not a dense slow wind as we find our models
1183: (Proga et al. 2002).
1184: 
1185: 
1186: ACKNOWLEDGMENTS: We thank J.M. Stone, M.C. Begelman, J.E. Drew, P.J. Armitage, 
1187: A. Feldmeier, and M. Ruszkowski for useful discussions.
1188: In particular, we thank A. Feldmeier for sharing with us his results
1189: and conclusions on MHD-LD disk winds. 
1190: We also thank  an anonymous referee for comments
1191: that helped us clarify our presentation.
1192: We acknowledge support from NASA under LTSA grant NAG5-11736.
1193: We also acknowledge support provided by NASA through grant  AR-09532
1194: from the Space Telescope Science Institute, which is operated 
1195: by the Association of Universities for Research in Astronomy, Inc., 
1196: under NASA contract NAS5-26555.
1197: Computations were partially supported by NSF grant AST-9876887.
1198: 
1199: 
1200: \newpage
1201: \section*{ REFERENCES}
1202:  \everypar=
1203:    {\hangafter=1 \hangindent=.5in}
1204: 
1205: {
1206: 
1207:   Abbott, D.C. 1982, ApJ, 259, 282
1208: 
1209:   Balbus, S.A., \& Hawley, J.F. 1991, ApJ, 376, 214
1210: 
1211:   Balbus, S.A., \& Hawley, J.F. 1998, Rev. Mod. Phys., 70, 1 
1212: 
1213:   Batchelor G.K. 1967, An Introduction to Fluid Mechanics (Cambridge:
1214:   Cambridge University Press)
1215: 
1216:   Blandford R.D., Payne D.G. 1982, MNRAS, 199, 883
1217: 
1218: 
1219:   Cannizzo J. K., Pudritz R.E. 1988, ApJ, 327, 840
1220: 
1221:   Castor J.I., Abbott D.C.,  Klein R.I. 1975, ApJ, 195, 157 (CAK)
1222:   
1223:   Contopoulos J. 1995, ApJ, 450, 616
1224: 
1225:   Chandrasekhar, S. 1960, Proc. Nat. Acad. Sci., 46, 253
1226: 
1227:   Drew J.E., Proga D., in  {\it  ``Cataclysmic Variables'',
1228:   Symposium in Honour of Brian Warner}, Oxford 1999, 
1229:   ed. by P. Charles,  A. King, D. O'Donoghue, in press   
1230: 
1231:   Feldmeier, A., \& Shlosman, I. 1999, ApJ, 526, 344
1232: 
1233:   Feldmeier, A.,  Shlosman, I., \&, Vitello, P. 1999, ApJ, 526, 357
1234: 
1235:   Gayley, K.G. 1995, ApJ, 454, 410
1236: 
1237:   Goodman J.,  Xu G. 1994 ApJ, 432, 213
1238: 
1239:   Hawley J.F,  Balbus S.A. 1991, ApJ, 376, 214
1240: 
1241: 
1242:   Kato, S.X., Kudoh T., \& Shibata K. 2002, ApJ, 565, 1035
1243: 
1244:   K$\ddot{\rm o}$nigl A. 1993, in  {\it ``Astrophysical Jets''},
1245:   ed. by D.P. O'Dea (Cambridge: Cambridge Univ. Press), 239
1246:   
1247:   Krasnopolsky R., Li Z.-Y., Blandford R., 1999, ApJ, 526, 631 
1248: 
1249:   Kudoh, T., Matsumoto, R., \& Shibata, K. 1998, ApJ, 508, 186
1250: 
1251:   Kudoh T., Shibata K. 1997, ApJ, 474, 362
1252: 
1253:   Li, Z.-Y. 1995, ApJ, 444, 848
1254: 
1255:   Li, Z.-Y. 1996, ApJ, 465, 855
1256: 
1257:   Miller, K.A., Stone, J.M. 2000, ApJ, 534, 398
1258: 
1259:   Mouschovias, T. Ch., Paleologou, E.V. 1980, ApJ, 237, 877
1260: 
1261:   Ogilvie G.I., Livio M. 1998, ApJ, 499, 329
1262: 
1263:   Ouyed R., Pudritz R.E. 1997a, ApJ, 482, 717
1264: 
1265:   Ouyed R., Pudritz R.E. 1997b, ApJ, 484, 794
1266: 
1267:   Ouyed R., Pudritz R.E. 1999, MNRAS, 309, 233
1268: 
1269:   Owocki S.P., Cranmer S.R.,  Gayley K.G.  1996, ApJ, 472, L115 
1270: 
1271:   Pelletier G., Pudritz R.E. 1992, ApJ, 394, 117
1272: 
1273:   Pereyra N.A., Kallman T.R. Blondin J.M. 1997, ApJ, 477, 368
1274: 
1275:   Pereyra N.A., Kallman T.R. Blondin J.M. 2000, ApJ, 532, 563
1276: 
1277:   Proga D. 1999, MNRAS, 304, 938
1278: 
1279:   Proga D. 2000, ApJ, 538, 684
1280: 
1281:   Proga D. 2002, Mass Outflow in Active Galactic Nuclei: New Perspectives, 
1282:   ASP Conference Proceedings, Vol. 255. 
1283:   Edited by D. M. Crenshaw, S. B.
1284:   Kraemer, and I. M. George. ISBN: 1-58381-095-1. 
1285:   San Francisco: Astronomical Society of the Pacific, 2002., p.309
1286: 
1287:   Proga, D., \& Kallman, T.R. 2002, ApJ, 565, 455
1288: 
1289:   Proga, D., Kallman, T.R., Drew, J.E., \& Hartley, L.E., 2002, ApJ, 572, 382
1290: 
1291:   Proga D., Stone J.M., Drew J.E. 1998, MNRAS, 295, 595 (PSD~98)
1292: 
1293:   Proga D., Stone J.M., Drew J.E. 1999, MNRAS, 310, 476  (PSD~99)
1294: 
1295:   Proga D., Stone J.M., Kallman T.R. 2000, ApJ, 543, 686  
1296: 
1297:   Pudritz R.E., Norman C.A. 1986, ApJ, 301, 571 
1298: 
1299:   Romanova, M.M., Ustyugova, G.V., Koldoba, A.V., Chechetkin, V.M., \&
1300:   Lovelace, R.V.E. 1997, ApJ, 482, 708
1301: 
1302:   Shakura N.I., Sunyaev R.A. 1973 A\&A, 24, 337
1303: 
1304:   Shibata K.,  Uchida Y. 1986, PASJ, 38, 631
1305: 
1306:   Stone, J.M., Hawley, J.F., Gammie, C.F., \& Balbus, Steven A. 1996 ApJ,
1307:   463, 656
1308: 
1309:   Stone J.M., Norman M.L. 1992a, ApJS, 80, 753
1310: 
1311:   Stone J.M., Norman M.L. 1992b, ApJS, 80, 791
1312:   
1313:   Stone J.M., Norman M.L. 1994, ApJ, 433, 746
1314: 
1315:   Uchida Y., Shibata K. 1985, PASJ, 37, 515
1316: 
1317:   Ustyugova G.V., Koldoba A.V., Romanova M.M. Chechetkin V.M. Lovelace R.V.E.
1318:   1995, ApJ, 439, 39L
1319: 
1320:   Ustyugova G.V., Koldoba A.V., Romanova M.M. Chechetkin V.M. Lovelace R.V.E.
1321:   1999, ApJ, 516, 221
1322:   
1323:   Velikov, E.P. 1959, JETP, 36, 1398
1324: 
1325:   Vitello P.A.J.,  Shlosman I., 1988 ApJ, 327, 680
1326: }
1327: 
1328: \newpage
1329: Fig.~1. A sequence of density maps (left), velocity fields (middle)
1330: and contours of the toroidal magnetic field (right)
1331: from run C0D after 37.8, 166.2, 294.7, 423.1, and 551.6~$\tau$ 
1332: (time increases from top to bottom).  
1333: Run~C0D is the example of a MHD-LD flow discussed in detail in section 3.1.
1334: For clarity, we suppress velocity vectors for regions with very low
1335: density (i.e., $\rho < 10^{-15} \rm g~cm^{-3}$).  
1336: The solid line overplotted on the velocity maps marks the 
1337: the Alfv${\acute{\rm e}}$nic surface 
1338: (i.e., location where $\mid v_{Ap} \mid= \mid v_p \mid$, middle panels).
1339: The $B_\phi$ contours are for -30,-20,-10,-5.,-1, 0, and 5.
1340: Dotted lines denote negative values of $B_\phi$.
1341: 
1342: Fig.~2. Two-dimensional structure of several quantities from
1343: model C0D after 680~$\tau$.
1344: The contours of $L$ are for 1.0, 1.5, 2 ,2.5 , and 3.0 
1345: while for $\Omega$ are for 0.025, 0.05, 0.1, 0.2, 0.4, and 0.8.
1346: Both $L$ and $\Omega$ are in units of the specific angular momentum
1347: and angular velocity on the Keplerian disk at $r_\ast$.
1348: The contours of $log \beta$ are for -3, -2.5 , -2.0, and -1.0.
1349: The solid lines overplotted on the poloidal magnetic field (bottom
1350: middle panel)
1351: show the location where the strength of the poloidal magnetic 
1352: field equals  the toroidal magnetic field, $\mid B_p \mid =\mid B_\phi
1353: \mid$. The contours for $B_\phi$ are as in Figure~1.
1354: 
1355: 
1356: Fig.~3. Quantities at the outer boundary in model C0D after 680~$\tau$.
1357: The ordinate on the left hand side of each panel refers to the solid line, 
1358: while the ordinate on the right hand side refers to the dotted line.
1359: Of particular note is the continuous strong increase of  
1360: the accumulated  mass loss rate with increasing polar angle.
1361: This is associated with the fact that the mass loss rate is dominated
1362: by a slow dense wind originating at large radii.
1363: 
1364: 
1365: Fig.~4. Model mass loss rates as functions of the total Eddington factor
1366: (top panel) and the initial plasma parameter, $\beta'_w$
1367: (bottom panel). In the top panel,  open circles are for pure LD models, 
1368: and the other shapes correspond to different values of $\beta'_w$ 
1369: ($\beta'_w=2\times10^{0}$ crosses, 
1370:  $\beta'_w=2\times10^{-2}$ asterisks, 
1371:  $\beta'_w=2\times10^{-3}$ diamonds, 
1372:  $\beta'_w=2\times10^{-4}$ triangles,  and
1373:  $\beta'_w=2\times10^{-6}$ squares).
1374: Table~1 specifies other model parameters.  
1375: The thick solid line represents results in the stellar CAK
1376: case while the thin solid line represents  the stellar CAK
1377: case corrected for a finite value of the maximum force multiplier
1378: $M_{max}$ (see the text for more detail).
1379: The alternative ordinate on the
1380: right hand side of the lower panel is the dimensionless wind mass loss
1381: rate parameter $\dot{M}_{w}'$ defined in PSD~98 (see equation 22 in PSD~98).
1382: The solid line in bottom panel represents  the mass loss rate
1383: for models of series C0  while the dashed line represents  the mass loss rate
1384: for models of series D1 (see section 3.2).
1385: 
1386: 
1387: \newpage
1388: 
1389: \begin{figure}
1390: \begin{picture}(280,590)
1391: \put(100,480){\special{psfile=f1al.ps angle =90
1392: hoffset=130 voffset=-10 hscale=22 vscale=22}}
1393: \put(100,360){\special{psfile=f1bl.ps angle =90
1394: hoffset=130 voffset=-15 hscale=22 vscale=22}}
1395: \put(100,240){\special{psfile=f1cl.ps angle =90
1396: hoffset=130 voffset=-15 hscale=22 vscale=22}}
1397: \put(100,120){\special{psfile=f1dl.ps angle =90
1398: hoffset=130 voffset=-15 hscale=22 vscale=22}}
1399: \put(100,0){\special{psfile=f1el.ps angle =90
1400: hoffset=130 voffset=-15 hscale=22 vscale=22}}
1401: 
1402: \put(250,480){\special{psfile=f1am.ps angle =90
1403: hoffset=130 voffset=-10 hscale=22 vscale=22}}
1404: \put(250,360){\special{psfile=f1bm.ps angle =90
1405: hoffset=130 voffset=-15 hscale=22 vscale=22}}
1406: \put(250,240){\special{psfile=f1cm.ps angle =90
1407: hoffset=130 voffset=-15 hscale=22 vscale=22}}
1408: \put(250,120){\special{psfile=f1dm.ps angle =90
1409: hoffset=130 voffset=-15 hscale=22 vscale=22}}
1410: \put(250,0){\special{psfile=f1em.ps angle =90
1411: hoffset=130 voffset=-15 hscale=22 vscale=22}}
1412: 
1413: \put(400,480){\special{psfile=f1ar.ps angle =90
1414: hoffset=130 voffset=-10 hscale=22 vscale=22}}
1415: \put(400,360){\special{psfile=f1br.ps angle =90
1416: hoffset=130 voffset=-15 hscale=22 vscale=22}}
1417: \put(400,240){\special{psfile=f1cr.ps angle =90
1418: hoffset=130 voffset=-15 hscale=22 vscale=22}}
1419: \put(400,120){\special{psfile=f1dr.ps angle =90
1420: hoffset=130 voffset=-15 hscale=22 vscale=22}}
1421: \put(400,0){\special{psfile=f1er.ps angle =90
1422: hoffset=130 voffset=-15 hscale=22 vscale=22}}
1423: 
1424: \end{picture}
1425: \caption{ 
1426: }
1427: \end{figure}
1428: 
1429: \begin{figure}
1430: \begin{picture}(180,400)
1431: \put(50,0){\special{psfile=f2.ps angle =90
1432: hoffset=500 voffset=0 hscale=70 vscale=70}}
1433: \end{picture}
1434: \caption{ 
1435: }
1436: \end{figure}
1437: 
1438: \begin{figure}
1439: \begin{picture}(180,400)
1440: \put(50,0){\special{psfile=f3.ps angle =90
1441: hoffset=500 voffset=0 hscale=70 vscale=70}}
1442: \end{picture}
1443: \caption{ 
1444: }
1445: \end{figure}
1446: 
1447: 
1448: \begin{figure}
1449: \begin{picture}(280,590)
1450: \put(50,200){\special{psfile=f4a.ps angle =90
1451: hoffset=500 voffset=0 hscale=70 vscale=70}}
1452: 
1453: \put(50,-100){\special{psfile=f4b.ps angle =90
1454: hoffset=500 voffset=0 hscale=70 vscale=70}}
1455: \end{picture}
1456: \caption{ 
1457: }
1458: \end{figure}
1459: 
1460: 
1461: 
1462: 
1463: 
1464: 
1465: \eject
1466: 
1467: \normalsize
1468: \centerline{ Table 1. Summary of results for disc winds }
1469: %\vspace{-6mm}
1470: \begin{center}
1471: \begin{tabular}{c c c c c c c c l } \\ \hline \hline
1472:      &                         &  & &                         &                   &             \\
1473: Run$^\ast$ &  $\MDOT_a$ & $x$ & $\beta'_{w}$& $t_f$ & $\MDOT_w$ & $v_r(10 r_\ast)$ & $\omega$ & comments \\
1474:      & (M$_{\odot}$ yr$^{-1}$ ) &   & & $\tau$ & (M$_{\odot}$ yr$^{-1}$) &   
1475: $(\rm km~s^{-1})$ & degrees  &  \\  
1476: 
1477: \hline
1478: 
1479: pure LD     &                         &   & &      &                   &                   &            & \\
1480: 
1481: C0A      &$  10^{-8}$                & 0 & $\infty$ & 800 &$ 5.5\times10^{-14}$     & 900                 &  50 & run~A in PSD~99\\
1482:  & & & & &  & \\
1483: B1A    &$ \pi \times 10^{-9}$     & 1 & $\infty$ & 1300 &                         &                 &   & no supersonic outflow \\
1484: D1A    &$ \pi \times 10^{-8}$    & 1 & $\infty$ & 1000&$ 2.1\times10^{-11}$    & 3500                &  32 &  run~C in PSD~99\\
1485: E1A     &$ \pi \times 10^{-7}$    & 1 & $\infty$ & 1000&$ 1.6\times10^{-9}$     & 7000                &  32 &  \\
1486: 
1487:  & & & & &  & & \\
1488: 
1489: pure MHD     &                         &   &      &                   &                   &           &  &\\
1490: A0C     &$  0 $                    & 0 & $2\times10^{-1}$ &1500 &$6.4\times10^{-13}$     & 100                 &  65 & \\
1491:  & & & & &  &  \\
1492: LD-MHD     &                         &   &      &       &            &                   &           &  \\
1493: C0B    &$  10^{-8}$              & 0 & $2\times10^0$ & 400 &$ 6.4\times 10^{-14}$     & 900                 &  55 & \\
1494: C0D    &$  10^{-8}$              & 0 & $2\times10^{-2}$ & 680&$ \simgreat 7.1\times10^{-12}$     & 1000                 &  60 & \\
1495: C0E    &$  10^{-8}$              & 0 & $2\times10^{-3}$ & 450&$>3.2\times10^{-11}$     & 1000                 &  65 & \\
1496: C0F    &$  10^{-8}$              & 0 & $2\times10^{-4}$ & 170&$ \simgreat 1.0\times10^{-10}$     & 1300                 &  65 & \\
1497:  & & & & &  & & \\
1498: B1F    &$ \pi \times 10^{-9}$    & 1 & $2\times10^{-4}$ & 300 &$1.3\times10^{-10}$     & 600                 & 70  &    \\
1499:  & & & & &  & \\
1500: D1B    &$ \pi \times 10^{-8}$    & 1 & $2\times10^0$ &250 &$ 2.4\times10^{-11}$     & 3500                &  32 &  \\
1501: D1D    &$ \pi \times 10^{-8}$    & 1 & $2\times10^{-2}$ &460 &$ 2.7\times10^{-11}$     & 3500                &  35 &  \\
1502: D1E    &$ \pi \times 10^{-8}$    & 1 & $2\times10^{-3}$ &670 &$ \simgreat 1.0\times10^{-10}$     & 3500                &  40 &  \\
1503: D1F    &$ \pi \times 10^{-8}$    & 1 & $2\times10^{-4}$ &420 &$ \simgreat 4.8\times10^{-10}$     & 3500                &  60 &  \\
1504: D1G    &$ \pi \times 10^{-8}$    & 1 & $2\times10^{-6}$ &120 &$ \simgreat 1.9\times10^{-09}$     & 3500                &  75 &  \\
1505:  & & & & &  & \\
1506: E1F    &$ \pi \times 10^{-7}$    & 1 & $2\times10^{-4}$ &260 &$1.6\times10^{-9}$     & 7000                & 32   &  \\
1507: \hline
1508: \end{tabular}
1509: 
1510: $\ast$ We use the following convention to label our runs:
1511: the first character in the name refers to $\MDOT_a$, (i.e., 
1512: A, B, C, D, and E are for 0, $\pi\times 10^{-9}$, $10^{-8}$, 
1513: $\pi\times 10^{-8}$, $\pi\times 10^{-7}$, respectively).
1514: The second character refers to $x$ 
1515: (i.e., 0 and 1, are for 0 and 1, respectively)
1516: and finally the third character refers to $\beta'_w$
1517: (i.e., A, B, C, D, E, F, and G are for $\infty$, $2\times10^0$,
1518: $2\times10^{-1}$, $2\times10^{-2}$, $2\times10^{-3}$, $2\times10^{-4}$,
1519: and $2\times10^{-6}$, respectively).
1520: \end{center}
1521: 
1522: \end{document}
1523: