astro-ph0502509/ms.tex
1: %\documentstyle[emulateapj]{article}
2: %\documentstyle[aas2pp4]{article}
3: %\documentstyle[aaspp]{article}
4: %\documentstyle[apjpt]{article}
5: %\documentstyle[12pt]{article}
6: \documentclass[12pt,preprint]{aastex}
7: 
8: \begin{document}
9: 
10: 
11: %\documentclass[12pt,preprint]{aastex}
12: %\usepackage{natbib}
13: %\citestyle{aa}
14: 
15: %\received{}
16: %\accepted{}
17: %\journalid{}{}
18: %\articleid{}{}
19: 
20: 
21: 
22: \title{On MHD jet production in the collapsing and rotating envelope}
23: 
24: \author{Daniel Proga,\altaffilmark{1}}
25: 
26: \affil{$^1$ Princeton University Observatory, Peyton Hall, Princeton, NJ 08544.e-mail: dproga@astro.princeton.edu}
27: 
28: 
29: 
30: \def\LSUN{\rm L_{\odot}}
31: \def\MSUN{\rm M_{\odot}}
32: \def\RSUN{\rm R_{\odot}} 
33: \def\MSUNYR{\rm M_{\odot}\,yr^{-1}}
34: \def\MSUNS{\rm M_{\odot}\,s^{-1}}
35: \def\MDOT{\dot{M}}
36: 
37: \newbox\grsign \setbox\grsign=\hbox{$>$} \newdimen\grdimen \grdimen=\ht\grsign
38: \newbox\simlessbox \newbox\simgreatbox
39: \setbox\simgreatbox=\hbox{\raise.5ex\hbox{$>$}\llap
40:      {\lower.5ex\hbox{$\sim$}}}\ht1=\grdimen\dp1=0pt
41: \setbox\simlessbox=\hbox{\raise.5ex\hbox{$<$}\llap
42:      {\lower.5ex\hbox{$\sim$}}}\ht2=\grdimen\dp2=0pt
43: \def\simgreat{\mathrel{\copy\simgreatbox}}
44: \def\simless{\mathrel{\copy\simlessbox}}
45: 
46: 
47: \begin{abstract}
48: We present results from axisymmetric, time-dependent 
49: hydrodynamical (HD) and magnetohydrodynamical
50: (MHD) simulations of a gaseous envelope collapsing onto a  black hole (BH). 
51: We consider gas with so small angular momentum that after 
52: an initial transient, the flow in the HD case, accretes directly onto a BH
53: without forming a rotationally support torus.
54: However, in the MHD case even with a very weak initial magnetic field, 
55: the flow settles into a configuration with four components: 
56: (i) an equatorial inflow,
57: (ii) a bipolar outflow, (iii) polar funnel outflow, and
58: (iv) polar funnel inflow.
59: We focus our analysis on the second flow component of the MHD flow
60: which represents a simple yet robust example 
61: of a well-organized inflow/outflow solution to the problem
62: of MHD jet formation.
63: The jet is heavy, highly magnetized, and 
64: driven by magnetic and centrifugal forces. 
65: A significant fraction of the total energy in the
66: jet is carried out by a large scale magnetic field.
67: We review previous simulations, where
68: specific angular momentum was higher than that assumed here, 
69: and conclude that our bipolar outflow develops
70: for a wide range of the properties of the flow
71: near the equator
72: and near the poles. Future work
73: on such a simple inflow/outflow solution
74: will help to pinpoint the key elements
75: of real jets/outflows as well as help to 
76: interpret much more complex simulations aimed at 
77: studying jet formation and collapse of magnetized envelopes.
78: \end{abstract}
79: 
80: \keywords{accretion, accretion disks  -- methods:
81: numerical -- MHD -- stars: winds, outflows} 
82: 
83: \section{Introduction}
84: The common occurrence and importance of astrophysical jets
85: have stimulated many theoretical studies.
86: Both analytic and numerical studies 
87: provide strong support for the scenario where jets are 
88: magnetohydrodynamic (MHD) in character and are associated
89: with an accretion disk, a rotating accretor, or both
90: (e.g., Blandford \& Payne 1982; Blandford 1990;
91: De Villier et al. 2003). In fact,
92: most MHD simulations of accretion flows show outflows
93: (e.g., Uchida \& Shibata 1985; Stone \& Norman 1994;
94: Hawley \& Balbus 2002; De Villier et al. 2004; Mizuno et al. 2004; 
95: Kato et al. 2004; McKinney \& Gammie 2004).
96: 
97: Here, we describe a study of the time evolution of
98: MHD flows in the vicinity of a stationary black hole (BH). This study
99: has been motivated by the results from our previous work 
100: (Proga \& Begelman 2003b, hereafter PB03b; Proga et al 2003) 
101: which showed that large-scale magnetic fields can produce
102: a jet from a rotationally supported accretion disk or torus
103: and also from extremely low angular momentum gas 
104: that almost radially accretes onto a BH.
105: This falling gas and associated outflow can be well-organized and
106: a simple, self-consistent solution for the MHD jet problem.
107: However, in our previous simulations and likely in some other simulations, 
108: this simple inflow/outflow is a component of a complex convolution
109: of a rotationally supported,
110: MHD turbulent torus, the torus corona and outflow.
111: To  articulate the basic physics that occurs in jet
112: production, we focus on a flow with angular momentum so low
113: that, if not for the effects of MHD, 
114: the flow would accrete directly onto a BH without forming a disk.
115: 
116: \section{Method}
117: 
118: To calculate the flow structure and evolution, we solve the equations of 
119: ideal MHD:
120: 
121: \begin{equation}
122:    \frac{D\rho}{Dt} + \rho \nabla \cdot {\bf v} = 0,
123: \end{equation}
124: 
125: \begin{equation}
126:    \rho \frac{D{\bf v}}{Dt} = - \nabla P - \rho \nabla \Phi+ \frac{1}{4\pi} {\bf (\nabla \times B) \times B},
127: \end{equation}
128: 
129: \begin{equation}
130:    \rho \frac{D}{Dt}\left(\frac{e}{ \rho}\right) = -P \nabla \cdot {\bf v},
131: \end{equation}
132: 
133: \begin{equation}
134: {\partial{\bf B}\over\partial t} = {\bf\nabla\times}({\bf v\times B}),
135: \end{equation}
136: where $\rho$ is the mass density, $P$ is the total gas pressure plus radiation
137: pressure, 
138: ${\bf v}$ is the fluid velocity, $e$ is the internal energy density,
139: $\Phi$ is the gravitational potential, and 
140: $\bf B$ is the magnetic field vector. We adopt an adiabatic
141: equation of state $P=(\gamma-1) e$ and consider models with
142: $\gamma=5/3$.
143: 
144: We perform simulations using the
145: pseudo-Newtonian potential of the central mass $\Phi_{PW} =
146: GM/(r-R_S)$, where $R_S = 2GM/c^2$ is the Schwarzschild radius,
147: introduced by Paczy\'{n}ski \& Wiita (1980).  
148: This potential approximates general relativistic effects
149: in the inner regions, for a nonrotating black hole.
150: In particular, the Paczy\'{n}ski--Wiita  potential
151: reproduces the last stable circular orbit at $r=3 R_S$
152: as well as  the marginally bound orbit at $r=2 R_S$.
153: 
154: Our calculations are performed in spherical polar coordinates
155: $(r,\theta,\phi)$. We assume axial symmetry about the rotational axis
156: of the accretion flow ($\theta=0^\circ$ and $180^\circ$).  The
157: computational domain occupies the radial range
158: $r_i~=~1.5~R_S \leq r \leq \ r_o~=~ 1200~R_S$, and the angular range
159: $0^\circ \leq \theta \leq 180^\circ$. The $r-\theta$ domain is
160: discretized on a non-uniform grid as in PB03b.
161: 
162: Our calculations use the ZEUS-2D code described by Stone \& Norman
163: (1992a,b).  We adopt PB03b's boundary and initial conditions but
164: we make two modifications to the initial conditions (see below).
165: An important element of the initial conditions is that the 
166: rotating gas has constant specific angular momentum, $l$, 
167: and is  confined in a wedge near the equator between 
168: $\theta=90^\circ -\theta_0$ and $\theta=90^\circ +\theta_0$.
169: The wedge of the rotating gas is truncated at the radius, $r(v_r=c_\infty)$
170: defined as the radius where
171: the initial radial velocity (as predicted by modified
172: Bondi velocity law, see Proga \& Begelman 2003a, hereafter PB03a) equals the sound speed at infinity, $c_\infty$.
173: 
174: The two modifications of the PB03b initial conditions are:
175: (i) we set the initial conditions exactly
176: as in PB03b and then reduce the density and internal energy
177: by a factor of 100
178: in the part of the computational domain where $v_\phi=0$ (outside
179: the rotating wedge) and (ii)
180: we consider two field geometries: 
181: a purely radial magnetic field as in PB03b and
182: a vertical magnetic field defined by the potential 
183: ${\bf A} = (A_r=0, A_\theta=0, A_\phi= A r \sin\theta)$.
184: For $r \sin{\theta} > r(v_r=c_\infty)$, we 
185: scale the magnitude of the magnetic field
186: using a parameter, $\beta_o \equiv 8 \pi P_B(r_o)/B^2$
187: defined as the plasma parameter $\beta\equiv 8\pi P/ B^2$ at 
188: $r_o$, so that
189: $A =(2 \pi P_B(r_o)/\beta_o)^{1/2}$
190: (where $P_B$ is the gas pressure associated with the Bondi solution
191: at $r_o$).
192: For $r \sin{\theta} < r(v_r=c_\infty)$, we set the constant A to
193: a very small value. 
194: 
195: We introduce these two modifications to 
196: reduce the flow complexity. In particular,
197: we want the rotating flow to be dominant
198: and we want to follow the evolution of the rotating flow with 
199: as  little interference
200: as possible from other flow components.
201: 
202: We choose the following units: 
203: $r_0=R_S$,
204: $t_0=4\pi(R_S^3/2GM)^{1/2}$, 
205: $v_0=c$, 
206: $B_0=(4\pi\rho_\infty c^2)^{1/2}$, 
207: $\rho_0=\rho_\infty$
208: (the density at infinity for a classic  Bondi flow). The force
209: is in units of $f_0=c^4/4GM$ and the specific angular momentum
210: is in units of $l_0= 2 R_S c$.
211: 
212: 
213: \section{Results}
214: 
215: To simulate a simple inflow/outflow solution we must appropriately 
216: set the model parameters.
217: Our initial conditions help to promote 
218: the dominance the rotating flow and reduce effects
219: of the non-rotating flow. 
220: The importance of the non-rotating flow depends not 
221: only on the initial conditions
222: but also on the minimum value of the density allowed in 
223: the simulations, the so-called density floor, $\rho_f$.
224: We set $\rho_f=(1/r)^{1/2}$. 
225: For comparison, 
226: the maximum density at small radii achieved
227: during the simulation is about $10^4$.
228: To aid the rotating flow in reaching an organized solution
229: we set $l$  smaller
230: than the critical angular momentum, $2 R_S c$. [Otherwise the rotating gas will
231: form a rotationally supported torus which becomes turbulent
232: and generates a magnetized complex corona and outflow (e.g., PB03b).]
233: Here we describe results from the simulations with $l=0.8$.
234: We set $\theta_0=56^\circ$ (as in a fiducial model in PB03b).
235: 
236: For our choice of $l$, the flow near the equator is sub-Keplerian at all radii
237: and will likely stay  sub-Keplerian during the evolution with appreciable 
238: radial velocity (i.e., the equatorial flow may continue to accrete 
239: supersonically). This feature distinguishes our simulations from 
240: many previous ones where the focus was on the outflows from rotationally 
241: supported disks or torii. Our main focus is on formation of an outflow
242: from nearly radially falling gas away from the equator.
243: 
244: We have performed numerous simulations varying the magnitude and geometry
245: of the magnetic field, the specific angular momentum, the numerical resolution
246: and the density floor. 
247: Here we present results from two models: model A without a magnetic field
248: and model B for which $\beta_o=10^3$ and the initial magnetic field 
249: is vertical.
250: Model A is a reference model which illustrates the flow pattern when
251: accretion proceeds directly whereas
252: model B illustrates the dynamics and properties of
253: simplest accretion flow which generates outflow
254: without formation of rotationally supported disk and without development
255: of magnetorotational instability.
256: 
257: In the early phase of the evolution, when the flow
258: relaxes from the initial conditions, both models show the same behaviour. 
259: First, the gas near the equator falls in nearly radially onto  a BH. However,
260: the rotating gas closer to the poles diverges away from the equator
261: because of a lack of pressure equilibrium in the $\theta$ direction
262: due to the density and internal energy difference between
263: the equatorial wedge and the polar region. 
264: The flow at small radii becomes gradually radial,
265: with the density decreasing  between the
266: equator and the poles. After this early phase, the evolution of the 
267: two models proceeds differently. In particular, the flow in model A
268: settles down into a steady state of direct accretion whereas
269: in model B an outflow as well as a direct accretion flow form.
270: 
271: 
272: Figure~1 presents the flow pattern for model A at small radii 
273: at $t =3550$.
274: The figure shows the density map overplotted with
275: the direction of the poloidal velocity.
276: The flow has two components: 
277: (i) an equatorial inflow and 
278: (ii) a polar funnel inflow. The equatorial inflow has non-zero $l$
279: but its cicularization radius is inside the last stable orbit
280: therefore direct accretion occurs. The polar
281: funnel inflow has zero $l$. The density contrast between the two components
282: is due to the fact that at the outer boundary and for the initial conditions
283: the density of the non rotating gas is smaller than the density of the rotating
284: gas by a factor of 100. 
285: At $t=500$, the mass accretion rate, $\MDOT_a=0.08$
286: in units of the Bondi rate ($\MDOT_B$) and then gradually increases.
287: At the end of the simulation, $t=3550$, $\MDOT_a=0.123$ and still continues
288: to grow but very slowly. We estimate that $\MDOT_a$ will saturate
289: at the level of $\approx 0.125$. We stress that there is no indication of
290: an outflow in model A. This contrasts with the MHD counterpart
291: of this model.
292: 
293: In model B, the establishment of the equatorial inflow
294: is soon (at $t>715$) followed by a development of a bipolar outflow from  
295: the 'shoulders' of the inflowing gas. At first, the outflow is
296: confined to a very narrow range of $\theta$ but with time
297: this range increases. In particular, at t=5700,
298: the outflow is at $26^\circ \lesssim \theta \lesssim 48^\circ$ and
299: $132^\circ \lesssim \theta \lesssim 154^\circ$ at $r=20$. 
300: In other words, we observe that with time the boundary
301: between the equatorial inflow and the  bipolar outflow moves toward 
302: the equator. The mass loss rate associated with the outflow, 
303: $\MDOT_W$ increases with time. For example,
304: at $t=1500$, $\MDOT_W=0.0008$  while at $t=5700$, $\MDOT_W=0.008$
305: ($\MDOT_w$ was measured at $r=175$). The late time evolution of $\MDOT_a$
306: for model B differs from that for model A. Namely,
307: for model B, $\MDOT_a= 0.08$ at $t=500$ 
308: and then gradually increases to 0.11 at $t=1500$. The mass accretion
309: rate stays at this level until $t=5000$ and then decreases
310: to 0.102 at $t=5700$.
311: 
312: The magnetic field evolves from vertical to nearly radial. 
313: This change in the field geometry is a natural consequence of
314: radially falling gas because the field lines are dragged in with
315: the gas. However, even at the end of the simulation,
316: the field did not evolve into a split monopole 
317: configuration due to the initial nonradial evolution of the flow.
318: 
319: 
320: Of course, the evolution of the field configuration
321: is accompanied by the growth of the field strength.
322: In particular, shear generates the toroidal field in the rotating flow.
323: The toroidal field is fastest growing component of the field in the
324: rotating gas and eventually dominates 
325: the poloidal field, $B_p=(B_r+B_\theta)^{1/2}$.
326: Additionally, 
327: the continuously growing field becomes dynamically important
328: (we find that an outflow forms in the regions where 
329: $\beta_\phi \equiv 8 \pi P/B_\phi^2<1$). 
330: In the polar region, the poloidal field is dominant,
331: $\beta_p \equiv 8 \pi P/B_p^2<1$) whereas
332: $\beta_\phi >1$). 
333: 
334: 
335: 
336: Figure~2 shows the flow pattern at small radii at $t =5700$ for model B.
337: The left and right panels show density and $|B_\phi|$ maps, respectively.
338: The figure shows also the direction of the poloidal velocity and
339: poloidal magnetic field as well as an example of a streamline.
340: The flow has  four components: 
341: (i) a radial equatorial inflow,
342: (ii) a bipolar outflow
343: (the streamline shown in Fig. 2 is typical for this flow component), 
344: (iii) a polar funnel outflow, and
345: (iv) a polar funnel inflow.
346: The third and fourth components (we call collectively 
347: the funnel flow) are separated from the inflow/outflow solution
348: by the centrifugal barrier. We note that the funnel outflow
349: can be hardly distinguished from the bipolar outflow based
350: on the direction of the poloidal
351: velocity. Comparison of the panels of Fig. 2 
352: reveals  that across the boundary between the bipolar outflow
353: and the funnel flow, three significant changes  occur: 
354: (i) the density dramatically decreases; (ii)
355: $B_p$ changes suddenly direction, and (iii) $B_\phi$ increases.
356: We also note that the polar funnel inflow is confined to
357: a very narrow region along the rotational axis and its
358: properties/dynamics are difficult to capture by our numerical
359: approach because of the imposed density floor
360: and gas heating by artificial viscosity. These effects can
361: occasionally lead to production of a thermal outflow in the funnel 
362: (e.g., see the lower half of the left panel for $r\lesssim - 11$ 
363: near the rotational axis).
364: In the remaining part of the paper, we focus on 
365: the second component of the MHD flow -- the inflow/outflow solution.
366: 
367: 
368: The relative simplicity and slow time evolution of the inflow/
369: outflow solution helps us to  identify the forces responsible
370: for production of the outflow from the infalling gas.
371: For example, we have analyzed and compared all the terms in the equation of
372: motion along various streamlines.
373: The important force components in the radial direction are: 
374: gravity, $f_g$, gas pressure gradient, $f_p$, centrifugal force, $f_c$
375: and gradient of the  toroidal  field pressure, $f_m$ (the other
376: components of the Lorentz force are negligible).
377: In the latitudinal direction  $f_p$, $f_c$ and $f_m$ 
378: determine the total force, $f_t$.
379: 
380: Figure~3 presents various fluid and magnetic field
381: properties and the forces acting on the flow as a function
382: of the pathlength (the pathlength is measured from
383: $r=14$ and $\theta=53^\circ$)
384: along the typical streamline for the inflow/outflow solution.
385: To generate this plot, we assume that the solution
386: is steady. This assumption is justified
387: for our purposes here because the time changes in the flow
388: occur on the scales significantly longer than the time needed for the 
389: flow element to follow the segment of the streamline we analyze.
390: 
391: Fig.~3a shows that the motion in the radial direction is 
392: determined in the following way:
393: (1) the infall slows down and eventually stops
394: due to the centrifugal force. Up to the stagnation point, 
395: the fluid specific angular momentum stays nearly constant
396: (dash-dotted line in Fig. 3f).
397: The centrifugal force can balance gravity, despite
398: sub-Keplerian rotation on the equator, because $f_c$
399: increases from the equator toward the poles  
400: [$f_c \propto 1/(r^3 \sin^2{\theta})$ for $l=const$].
401: Near the stagnation point, two significant changes occur in the force
402: balance: (i) $f_m$ becomes significant and is directed
403: outward and (ii) $f_c$ increases compared to gravity because
404: the magnetic torque increases somewhat specific angular momentum of the fluid. 
405: The combined centrifugal and $B_\phi$ pressure 
406: forces overcome the combined gravity and gas pressure 
407: (the latter is directed inwards)
408: and accelerate the flow outward. Although 
409: $f_c > f_m$ in the outflow, $f_c$ alone
410: is not strong enough to accelerate the flow.
411: Therefore, the flow is not 'flung out' by the centrifugal
412: force as in Blandford \& Payne's (1982) magnetocentrifugal wind
413: but is  much more gradually pushed by the pressure of the toroidal field
414: (see Spruit 1996 and also below for a discussion of heavy loaded MHD winds).
415: 
416: 
417: Fig. 3b shows that as the infalling gas approaches 
418: the stagnation point, $f_p$ pushes the gas away from the equator against 
419: the centrifugal force. Near the stagnation point, two significant changes 
420: occur in the force balance in the $\theta$ direction: (i)  $f_m$ chances sign 
421: and (ii) $f_p$ increases so that it
422: alone can 'collimate' the gas during the early phase of the outflow
423: (this is done against not only $f_c$ and also against $f_m$).
424: 
425: Farther downstream from the stagnation point,  $f_m$ again changes sign
426: and becomes stronger than $f_p$. 
427: Thus, $f_m$ appears to play an important role in the outflow collimation 
428: at large radii whereas $f_p$ pushes the outflow
429: away from the equator during the initial acceleration. 
430: However, when the $\theta$ component of $f_m$ changes sign for 
431: the second time, $B_r$ changes sign too
432: (the dashed vertical lines in all panels of Fig. 3 mark the location
433: where $B_r$ changes sign downstream from the stagnation point). 
434: In ideal MHD, the magnetic field should be dragged in with the flow
435: and no changes in the field orientation should occur
436: within an organized outflow due to field freezing. However, 
437: in our solution of MHD equations using the grid based code,
438: the change of $B_r$ sign is due to
439: annihilation (reconnection) of oppositely
440: oriented fluxes (i.e., the polar funnel flux vs. the bipolar
441: outflow flux) on the grid-spacing scale.
442: We note this numerical limitation
443: and focus on the properties of the bipolar outflow 
444: upstream from the point where the polar funnel flux
445: annihilates the bipolar outflow flux.
446: 
447: The bipolar outflow is  unbound and it is heavy loaded.
448: To quantify the latter, we 
449: compute the so-called mass loading parameter
450: $\mu\equiv (v_p v_\phi/v^2_{Ap})^{1/2}$  
451: (where $v_{Ap}\equiv (B_p^2/4\pi\rho)^{1/2}$ 
452: is the poloidal Alfven spead) introduced by
453: Spruit (1996, see also eq. 2 in Anderson et al. 2004; note a difference
454: between the definiation of $\mu$ in the two references). 
455: At the stagnation point, $\mu=30$.
456: Fig.~3d shows that 
457: the toroidal field is higher than the poloidal field 
458: (compare solid, dotted, and dashed lines)
459: while Fig.~3f shows that
460: the outflow does not corotate with its base and the outflow
461: does not gain much of specific angular momemtum
462: (see  dashed and dash-dotted lines in Fig.~3f for
463: the angular velocity, $\Omega$ and specific angular momentum, respectively).
464: All these flow characteristics 
465: are consistent with the high value of $\mu$ at the stagnation point
466: (e.g., Spruit 1996).
467: 
468: 
469: Fig. 3e shows that the flow is super Alfvenic (dotted line).  
470: This figure also shows that
471: the ratio between the Poynting flux and kinetic energy flux, $F_P/F_K$
472: increases downstream from the stagnation point.
473: The energy flux ratio becomes of order of unity 
474: near the location of the bipolar outflow 'merging' with the funnel flow 
475: (i.e., the location mark by the vertical dashed line). 
476: 
477: \section{Summary and Concluding Remarks}
478: 
479: We present results from asymmetric HD and MHD simulations of 
480: a rotating gas collapsing onto a stationary BH. We consider
481: extreme cases where the gas rotation is so small that the circularization
482: radius is inside the last stable orbit.  Therefore one expects direct accretion
483: of all gas without formation of a rotationally supported torus
484: no mention of formation of an outflow.
485: After an initial transient, the HD flow, as expected, 
486: settles into a configuration where direct accretion
487: occurs and there is not outflow. This contrasts with the MHD flow,
488: which even for a very weak initial magnetic field, 
489: settles into a configuration with four
490: components: (i) a radial equatorial inflow,
491: (ii) a bipolar outflow, (iii) polar funnel outflow, and
492: (iv) polar funnel inflow.
493: Our focus here is on the second, unexpected component of the MHD flow.
494: 
495: The bipolar outflow is driven by 
496: the magnetic and centrifugal forces. Following the streamlines
497: of the bipolar outflow, we identify
498: four stages of the flow motion:
499: (i) a gravitational collapse, (ii) a gradual slow down and eventual
500: termination of the collapse by the centrifugal force,
501: (iii) launching of the outflow by gas pressure which
502: redirects the flow away from the equator and by the centrifugal force
503: which redirects the flow away from a BH in the radial direction 
504: (the latter force is enhanced owning to
505: a magnetic torque spinning up the gas), and (iv)
506: an acceleration by the centrifugal force and gradient
507: of the toroidal field pressure. 
508: The outflow is highly magnetized ($\beta< 1$) and 
509: the Poynting flux carries a significant fraction of the total
510: energy. 
511: 
512: Despite performing the simulation for thousands of orbits at the inner radius
513: the flow did not reach a steady state.
514: One indication of the time evolution is  that
515: an increasingly larger fraction of the equatorial inflow 
516: turns into the bipolar outflow. Additionally,  the
517: inflow very close to the equator can start turning back, 
518: the equatorial symmetry breaks, and
519: the flow near the equator starts to circulate. 
520: However, not all flow properties evolve with time.
521: For example, the angular velocity profile along the equator
522: is time steady and closely follows the  $1/r$ scaling
523: as expected for the constant $l$ flow.
524: Generally,
525: the flow evolves over a very long time scale 
526: into a state where the bipolar outflow 'sandwiches'
527: a smooth radial inflow or a complex flow near the equator 
528: (our test runs with $l=0.5$ show that
529: the equatorial flow can remain a smooth inflow for as long
530: as the simulations). We have observed a similar bipolar outflow in
531: almost all of our simulations
532: performed recently or performed in PB03b and Proga et al. (2003).
533: Our inflow/outflow solution resembles 
534: inflow/outflow self-similar solutions studied in the context of 
535: the core collapse leading to star formation 
536: (e.g., see the circulation region in Fig. 1 of Lery el. al 2002; 
537: see also Henriksen, R.N., \& Valls-Gabaud 1995,
538: and reference therein). We also note that our solution could be qualitatively
539: similar to inflow-outflow circulation found in the simulations
540: of the magnetized cloud contracting under self-gravity (Tomisaka 1998)
541: and to the funnel-wall jet found 
542: in fully relativistic numerical simulations of accretion disks in 
543: the Kerr metric (De Villiers t al. 2004). 
544: 
545: We finish with two remarks: (i) The inflow/outflow
546: solution is relevant to several astrophysical problems
547: where some of the magnetized collapsing gas may have very low
548: angular momentum (e.g., star forming regions, the Galactic center
549: and other supermassive BH environments where the BH is surrounded by
550: the diffused gas and mass losing stars, some GBRs and supernovae
551: where it is believed that the explosion is related to a collapse
552: of a massive rotating stellar envelope). (ii) Although
553: our physics, initial and boundary conditions are 
554: simple compared to those  in real systems,
555: our inflow/outflow solution is 
556: simple and probable robust, and will help to pinpoint the key elements
557: of real jets/outflows from low $l$ accretion flows.
558: 
559: ACKNOWLEDGMENTS: We thank Philp Armitage, Mitch Begelman, Roger Blandford, 
560: Scott Kenyon, Tom Gardiner, Bohdan Paczy\'{n}ski, 
561: Jim Stone, and Dimitri Uzdensky
562: for useful discussions. We also thank an anonymous referee for useful
563: comments that helped us clarify our presentation.
564: We acknowledge support from NASA under ATP grant NNG05GB68G
565: and support provided by NASA through grant  HST-AR-10305.05-A
566: from the Space Telescope Science Institute, which is operated
567: by the Association of Universities for Research in Astronomy, Inc.,
568: under NASA contract NAS5-26555.
569: 
570: 
571: \begin{thebibliography}{99}
572: 
573: \bibitem[]{} Blandford, R.D., \& Payne D.G. 1982, MNRAS, 199, 883
574: 
575: \bibitem[]{}  Blandford, R.D., {\it Active Galactic Nuclei} (Eds
576: R. Blandford, H. Netzer \& L. Woltjer), Berlin: Springer 1990
577: 
578: \bibitem[]{}  Balbus, S. A., \& Hawley, J. F. 2002, ApJ, 573, 738
579: 
580: \bibitem[]{}  De Villiers, J.-P., Hawley J.F., Krolik J.H., Hirose S. 2004, ApJ,in press  (astro-ph/0407092)
581:  
582: \bibitem[]{}  Henriksen, R.N., \& Valls-Gabaud, D. 1994, MNRAS, 266, 681
583: 	
584: \bibitem[]{}  Lery, T., Henriksen, R.N., Fiege, J.D., Ray, T.P, Frank, A., 
585: \& Bacciotti, F. 2002, A\&A, 387, 187
586: 
587: \bibitem[]{}  Kato, Y., Mineshige, S., \& Shibata, K. 2004, ApJ, 605, 307
588: 
589: \bibitem[]{}  McKinney, J.C., \& Gammie, C.F. 2004, ApJ, 611, 977
590: 
591: \bibitem[]{} Mizuno, Y., Yamada, S., Koide, S., \& Shibata, K. 
592: 2004, ApJ, 606, 395
593: 
594: \bibitem[]{} Paczy\'{n}ski, B., \& Wiita, P. J. 1980, A\&A, 88, 23 
595: 
596: \bibitem[]{} Proga, D., \& Begelman, M.C. 2003 ApJ, 582, 69 (PB03a)
597: 
598: \bibitem[]{} Proga, D., \& Begelman, M.C. 2003 ApJ, 592, 767 (PB03b)
599: 	
600: \bibitem[]{}  Proga, D., MacFadyen, A.I., Armitage, P.J., \& Begelman, M.C. 
601:  2003, ApJ, 599, L5
602: 
603: \bibitem[]{} Spruit, H. C. 1996, in Evolutionary Processes in Binary Stars, 
604: ed. R. A. M. J. Wijers, M. B. Davies, \& C. A. Tout (Dordrecht: Kluwer), 249
605: 
606: \bibitem[]{}  Stone, J.M., \& Norman, M.L. 1994, ApJ, 433, 746
607: 
608: \bibitem[]{}  Stone, J.M., \& Norman, M.L. 1992a, ApJS, 80, 753
609: 
610: \bibitem[]{}  Stone, J.M., \& Norman, M.L. 1992b, ApJS, 80, 791
611: 
612: \bibitem[]{}  Tomisaka K. 1998, ApJ, 502, L167
613: 
614: \bibitem[]{} Uchida, Y,  \& Schibata K. 1985, PASJ, 37, 515
615: 
616: 
617: \end{thebibliography}
618: 
619: \clearpage
620: 
621: \begin{figure}
622: \begin{picture}(180,200)
623: \put(380,-10){\special{psfile=f1.ps angle =90
624: hoffset=130 voffset=-15 hscale=70 vscale=70}}
625: \end{picture}
626: \caption{
627: A map of logarithmic density overplotted with
628: the direction of the poloidal velocity for model A 
629: (a zero magnetic field case).}
630: \end{figure}
631: 
632: \clearpage
633: 
634: \begin{figure}
635: \begin{picture}(180,200)
636: \put(380,-10){\special{psfile=f2a.ps angle =90
637: hoffset=130 voffset=-15 hscale=70 vscale=70}}
638: \put(630,-10){\special{psfile=f2b.ps angle =90
639: hoffset=130 voffset=-15 hscale=70 vscale=70}}
640: \end{picture}
641: \caption{
642: Maps of logarithmic density (left panel) and toroidal magnetic field 
643: (right panel) overplotted with an example of a streamline 
644: corresponding to an inflow/outflow for model~B 
645: (a non-zero magnetic field case). 
646: The maps are also overplotted with
647: the direction of the poloidal velocity and
648: the direction of the poloidal field 
649: (the left  and right panels, respectively).
650: }
651: \end{figure}
652: 
653: 
654: \begin{figure}
655: \begin{picture}(180,480)
656: \put(100,460){\special{psfile=f3a.ps angle =90
657: hoffset=200 voffset=-150 hscale=40 vscale=40}}
658: \put(100,340){\special{psfile=f3b.ps angle =90
659: hoffset=200 voffset=-150 hscale=40 vscale=40}}
660: \put(100,220){\special{psfile=f3c.ps angle =90
661: hoffset=200 voffset=-150 hscale=40 vscale=40}}
662: 
663: \put(340,460){\special{psfile=f3d.ps angle =90
664: hoffset=200 voffset=-150 hscale=40 vscale=40}}
665: \put(340,340){\special{psfile=f3e.ps angle =90
666: hoffset=200 voffset=-150 hscale=40 vscale=40}}
667: \put(340,220){\special{psfile=f3f.ps angle =90
668: hoffset=200 voffset=-150 hscale=40 vscale=40}}
669: 
670: \end{picture}
671: \caption{\small{
672: {\it From top to bottom:}  The dominant forces
673: in the radial and latitudinal
674: directions (a and b panels, respectively)
675: along the streamline presented in Fig. 2.
676: The fluid and magnetic properties as a function of
677: the pathlength along the streamline (c-f panels).
678: For details see the labels and the main text.
679: Note that the specific angular momentum, $l$ is reduced
680: by a factor of 10 while the radius along the stream
681: line is reduced by a factor of 50
682: (dot-dashed line and solid line
683: in  panel f, respectively).
684: The solid  vertical lines in all panels correspond to the location where
685: the flow reaches the stagnation point 
686: (the minimum radius along the streamline).
687: The dashed  vertical lines in all panels correspond to the location where
688: the radial component of the magnetic field changes sign
689: and indicates the location where the magnetic field associated
690: with the bipolar outflow is advected into the grid cell
691: with a much stronger, oppositely  oriented magnetic field associated
692: with the polar funnel.
693: }}
694: \end{figure}
695: 
696: 
697: 
698: \end{document}
699: 
700: