0707.0576/xxx.tex
1: %style options:
2: \documentclass[twocolumn,prl,aps,showpacs]{revtex4}
3: %\documentclass[preprint,aps,showpacs]{revtex4}                             %
4: \topmargin -0.5 in
5: 
6: 
7: %\renewcommand{\baselinestretch}{2.0}
8: 
9: %%\hoffset=-1.cm
10: %%\voffset=1cm
11: %%\textwidth=16.5cm
12: %%\textheight=23cm
13: 
14: 
15: \usepackage{epsfig}
16: 
17: \begin{document}
18: %\input psfig.sty
19: 
20: 
21: %\documentclass[preprint,12pt]{aastex}               % 1 column AAS preprint
22: %\documentclass[12 pt]{article}
23: %\documentclass[longnamesfirst,apj]{emulateapj}
24: %\documentclass[apjl]{emulateapj}
25: %\usepackage{apjfonts,natbib}
26: %\usepackage{apjfonts}
27: %\bibliographystyle{apj}
28: %\topmargin 0.5 in 
29: 
30: %\parindent 0 pt
31: %\parskip 15 pt
32: %\usepackage{epsfig}
33: 
34: 
35: %\begin{document}
36: 
37: \def\beq{\begin{equation}}
38: \def\eeq{\end{equation}}
39: 
40: 
41: \title{Magnetically-dominated jets inside collapsing stars
42: as a model for gamma-ray bursts and supernova explosions}
43: 
44: \author{Dmitri A. Uzdensky}
45: \email{uzdensky@astro.princeton.edu}
46: \affiliation{Dept. of Astrophysical Sciences, Princeton University,  
47: and Center for Magnetic Self-Organization (CMSO), Princeton, NJ 08544}
48: 
49: \author{Andrew I.\ MacFadyen}
50: \email{aim@ias.edu}
51: \affiliation{Dept. of Physics, New York University, and Institute 
52: for Advanced Study, Princeton, NJ 08540}
53: 
54: 
55: \date{March 7, 2007}
56: 
57: \begin{abstract}
58: It has been suggested that magnetic fields play a dynamically-important
59: role in core-collapse explosions of massive stars. In particular, they
60: may be important in the collapsar scenario for gamma-ray bursts (GRB),
61: where the central engine is a hyper-accreting black hole or a millisecond 
62: magnetar. The present paper is focussed on the magnetar scenario, with a 
63: specific emphasis on the interaction of the magnetar magnetosphere with 
64: the infalling stellar envelope. First, the ``Pulsar-in-a-Cavity'' problem 
65: is introduced as a paradigm for a magnetar inside a collapsing star. 
66: The basic set-up of this fundamental plasma-physics problem is described, 
67: outlining its main features, and simple estimates are derived for the 
68: evolution of the magnetic field. In the context of a collapsing star, 
69: it is proposed that, at first, the ram pressure of the infalling plasma 
70: acts to confine the magnetosphere,
71: enabling a gradual build-up of the magnetic pressure. At some point, the
72: growing magnetic pressure overtakes the (decreasing) ram pressure of the
73: gas, resulting in a magnetically-driven explosion. The explosion should 
74: be highly anisotropic, as the hoop-stress of the toroidal field, confined 
75: by the surrounding stellar matter, collimates the magnetically-dominated 
76: outflow into two beamed magnetic-tower jets. This creates a clean narrow 
77: channel for the escape of energy from the central engine through the star,
78: as required for GRBs. In addition, the delayed onset of the 
79: collimated-explosion phase can explain the production of large quantities
80: of Nickel-56, as suggested by the GRB-Supernova connection. Finally, the 
81: prospects for numerical simulations of this scenario are discussed.
82: \end{abstract}    
83: \vspace{0.1 in}
84: 
85: 
86: \pacs{52.25.Xz, 52.30.Cv, 95.30.Qd, 97.60.Bw, 97.60.Jd, 98.70.Rz}
87: 
88: \maketitle
89: 
90: \newpage
91: 
92: 
93: %*******************************************************************
94: 
95: 
96: \section {Introduction}
97: \label{sec-intro}
98: 
99: %**********************************************************************
100: 
101: Long-duration gamma-ray bursts (GRBs) are believed to be strongly asymmetric 
102: explosions of massive stars. Some explosion asymmetry has also been 
103: inferred in core-collapse supernova (SN) explosions. The asymmetry 
104: is expected if the progenitor star is rapidly rotating, with the 
105: rotation axis providing a preferred direction for a jet-like outflow.
106: 
107: 
108: The leading theoretical model for long-duration GRBs based on rapid stellar 
109: rotation the {\it collapsar} model\cite{woosley-1993,paczynski-1998,
110: macfadyen-woosley-1999}. 
111: In this model the core of a massive star collapses to form a black hole. 
112: If the star is rapidly rotating, the overlying stellar material can form 
113: an accretion disk around the black hole. Because of differential
114: rotation in the disk, the magneto-rotational instability (MRI) is expected 
115: to develop ({\it e.g.}, Ref.~\cite{proga-2003}), providing angular momentum 
116: transport and dissipation of gravitational energy.  At the temperatures 
117: and densities ($T \sim 4 \times 10^{10}$ K, $\rho \sim 10^{10}$ g cm$^{-3}$) 
118: present in collapsar disks, neutrino emission cools the gas and the collapsing 
119: outer stellar core accretes at rates of $\sim 0.1\,M_{\odot}\,{\rm s}^{-1}$ 
120: for times $\gtrsim 10$~sec, long enough for the jet to break out of the star
121: and to power long GRBs. This accretion disk---black hole system thus acts 
122: as the central engine for the GRB. The energy source in this case is mainly 
123: accretion. Since most of this energy is released via neutrinos, most models 
124: have relied on neutrino luminosity to power the outflow. However, it has 
125: also been suggested that a significant amount of energy may be extracted 
126: magnetically and that strongly magnetized jets may play an active role in 
127: GRB explosions\cite{thompson-1994,meszaros-rees-1997,lee-2000,
128: vlahakis-konigl-2001,vanPutten-ostriker-2001,drenkhahn-spruit-2002,
129: lyutikov-blandford-2003,vanPutten-levinson-2003,proga-2003,lyutikov-2006,
130: mckinney-2005,um-2006a,giannios-spruit-2005,giannios-spruit-2006,
131: giannios-spruit-2007}).
132: 
133: In order for magnetic mechanisms to be viable, a fairly strong fields 
134: (of order~$10^{15}$~G) are required. Such fields, however, are expected 
135: to be generated by the MRI-driven turbulent dynamo in the disk during 
136: core-collapse\cite{proga-2003}.
137: Similar processes, resulting in similar field strengths, are also 
138: believed to be taking place in core-collapse supernovae explosions 
139: \cite{akiyama-2003,ardeljan-2005}.  
140: For example, Akiyama \& Wheeler have argued\cite{akiyama-2003} that 
141: the field may reach the level set by the equipartition with the MRI-driven 
142: turbulence, as strong as $10^{16-17}$~G. However, as they point out, this 
143: field is mostly toroidal. The large-scale {\it poloidal} magnetic field 
144: that is needed  may require a large-scale helical dynamo\cite{blackman-2006} 
145: and will probably be somewhat smaller than the toroidal field. Thus, we 
146: believe it is not unreasonable for the poloidal field at the disk surface 
147: to be a more modest $B_d\sim 10^{15}$~G (see also Ref.~\cite{proga-2003}).
148: 
149: In this paper we focus mostly on an alternative scenario, in which the 
150: central engine operating inside a collapsing star is not a black hole but 
151: a young millisecond magnetar --- a neutron star (NS) with a large-scale 
152: poloidal magnetic field of the order of $10^{15}$~G. Such a strong magnetic 
153: field can be produced by a convectively driven turbulent dynamo 
154: \cite{duncan-thompson-1992,thompson-duncan-1993} inside the young~NS,
155: or by the MRI-driven dynamo in the differentially rotating collapsing 
156: core\cite{akiyama-2003}.
157: An alternative possibility is that the progenitor core of about $10^4$~km 
158: has a magnetic field of order $10^9$~G, similar to the fields observed in 
159: some white dwarfs. When such a highly-magnetized core collapses into a 
160: neutron star of 10~km radius, flux freezing leads to the amplification 
161: of the magnetic field to $10^{15}$~G ({\it e.g.}, Ref.~\cite{um-2006a}).
162: 
163: 
164: We see that the typical values of the magnetic field strength, the rotation 
165: rate, and the size of the central engine in the classical collapsar scenario 
166: with an accreting black hole and in the millisecond-magnetar case are similar. 
167: The overall electro-magnetic luminosities should therefore be also comparable. 
168: And indeed, for a typical surface magnetic field $B_*=10^{15}$~G, 
169: a rotation rate $\Omega_*=10^4\ {\rm sec}^{-1}$, and a radius $R_*=10$~km,
170: the basic energetics and timescales make the millisecond magnetar 
171: a plausible candidate for a GRB central engine\cite{thompson-1994}. 
172: Ultimately, the energy source for the explosion is the rotational energy 
173: of the neutron star and the magnetic field acts as the agent that extracts 
174: this energy.  
175: In particular, the rotational energy of a millisecond-period neutron star 
176: is of order $E_{\rm rot}\simeq 5 \cdot 10^{52}\ {\rm erg}$, more than 
177: enough to drive a long-duration GRB.
178: The energy-extraction time scale, estimated using the usual pulsar luminosity 
179: formula $L_{\rm magn}\sim B_*^2\, R_*^6\, \Omega_*^4\, c^{-3}$, is then  
180: of order $100$~sec. 
181: Thus, the millisecond-magnetar central engine is essentially similar 
182: (with the magnetic field scaled up by three orders of magnitude and 
183: the timescale scaled down by six) to the Ostriker \& Gunn model\cite{og-1971} 
184: for powering supernovae by the spin-down magnetic power of a rapidly 
185: rotating pulsar. 
186: 
187: The idea of a millisecond magnetar as a central engine for gamma-ray bursts 
188: has been first proposed independently by Usov\cite{usov-1992} and by Duncan 
189: \& Thompson\cite{duncan-thompson-1992}. It has been further developed by 
190: a number of authors\cite{thompson-1994,yi-blackman-1998,nakamura-1998, 
191: spruit-1999,wheeler-2000,wheeler-2002,ruderman-2000,thompson-2004,um-2006b}. 
192: In addition, several magnetic explosion mechanisms have been explored in 
193: the supernova context\cite{leBlanc-wilson-1970,meier-1976,wheeler-2000,
194: wheeler-2002,akiyama-2003,thompson-2004,ardeljan-2005,blackman-2006,
195: shibata-2006,bucciantini-2006,metzger-2007,burrows-2007}). 
196: Our present paper is also devoted to investigating the millisecond-magnetar 
197: scenario for GRBs, but, importantly, viewed within the overall context of 
198: a collapsing star.
199: 
200: First we would like to stress that the plausible overall energetics
201: and timescales are, by themselves, not sufficient for making a good
202: GRB central engine model. The central engine also needs to be able 
203: to produce an energetic outflow that is (1) ultra-relativistic; 
204: (2) highly-collimated; and (3) baryon-free. Most of the previous studies 
205: have focused  mostly on the energetics and timescales, but not on 
206: the mechanisms for producing an outflow that satisfies these requirements 
207: (see, however, Refs.~\cite{wheeler-2000} and~\cite{bucciantini-2006} for 
208: a discussion of collimation). Also, most of these models, with the notable 
209: exception of Refs.~\cite{wheeler-2000} and~\cite{arons-2003}, have considered 
210: a magnetar in isolation, ignoring the effect of the surrounding stellar gas 
211: on shaping the outflow. 
212: 
213: One of the major points of our paper is that the infalling stellar gas 
214: is still present during the explosion and needs to be taken into account.
215: Thus, an important new element distinguishing our model from previous works 
216: is the consideration of the interaction between a newly-born magnetar and 
217: the stellar plasma in which it is initially embedded. Specifically, we argue 
218: that the pressure and inertia of the surrounding gas play a key role in 
219: regulating the magnetic extraction of rotational energy from the magnetar.
220: They also force the Poynting-flux-dominated outflow into two collimated 
221: jets, similar to Lynden-Bell's magnetic towers\cite{lb-1996,lb-2003,um-2006a}.
222: 
223: 
224: In Section~\ref{sec-stalled-shock} we discuss a specific situation relevant to 
225: the core collapse of a massive star: a cavity inside the stalled bounce-shock. 
226: The radius of such shock stays roughly constant allowing the magnetic fields 
227: in the cavity to grow. At the same time, both the ram pressure of the infalling
228: gas and the neutrino energy deposition inside the cavity decrease with time. 
229: We therefore argue that at some point, a fraction of a second after bounce, 
230: the magnetic field will start to dominate the force balance, leading to a 
231: magnetically-driven explosion. 
232: 
233: In order to illustrate these ideas, we introduce the ``Pulsar-in-a-Cavity'' 
234: problem as a basic-physics paradigm for this scenario. We describe this 
235: idealized problem in detail in~Section~\ref{sec-pulsar-cavity}. We first 
236: give a general description of the problem and its several versions. In 
237: Section~\ref{subsec-fixed-cavity}, we consider the simplest case of a 
238: force-free rotating magnetosphere inside a fixed rigid cavity. We demonstrate 
239: that differential rotation of the magnetic field lines is inevitably 
240: established inside the cavity, even though the pulsar itself is rotating 
241: uniformly; as a result, a strong toroidal magnetic field gradually builds up. 
242: We then study the long-term evolution of the field inside the cavity and show t
243: hat the magnetic luminosity increases with time. We also show that a massive, 
244: non-force-free plasma strip unavoidably arises in the equatorial plane 
245: beyond the light cylinder. We expect this phase to last until either the 
246: development of the kink instability modifies the situation or until the 
247: cavity walls yield to the internal magnetic stresses.
248: In Section~\ref{subsec-collimation} we discuss hoop-stress collimation and 
249: argue that external confinement and differential rotation are two important 
250: ingredients for collimating relativistic Poynting-flux dominated outflows. 
251: In either case, a magnetic tower forms. 
252: 
253: We investigate the propagation of a magnetic tower through the star in 
254: Section~\ref{sec-tower}. Section~\ref{subsec-lb-tower} is devoted to a 
255: general description of Lynden-Bell's magnetic tower model\cite{lb-1996}, 
256: whereas Sections~\ref{subsec-tower-star} --- \ref{subsec-numbers}
257: describe our modification of this model for the stellar environment. 
258: Specifically, we suggest that the external confining pressure invoked 
259: by Lynden-Bell's model is provided by the gas inside a hot cocoon behind 
260: a strong shock that the rapidly-growing magnetic tower drives into the 
261: unperturbed stellar envelope.
262: 
263: 
264: In~Section~\ref{sec-discussion}, we further explore some 
265: of the astrophysically-interesting aspects of our model. 
266: In Section~\ref{subsec-rel}, we consider the transition of the magnetic tower 
267: expansion to the relativistic regime and the final opening angle of the tower. 
268: In Section~\ref{subsec-small-scale}, we discuss a possible small-scale
269: substructure of the magnetic tower, represented as a ``train of plasmoids'',
270: that may be the outcome of the small-scale magnetic structure at the base 
271: of the outflow ({\it e.g.}, in the magnetized corona of the magnetar or of the 
272: accretion disk); it may also develop later as a result of MHD instabilities 
273: and flux conversion in the growing magnetic tower itself.
274: In Section~\ref{subsec-stability}, we discuss these stability issues, 
275: especially in regard to the Rayleigh--Taylor and kink instabilities. 
276: In Section~\ref{subsec-reconnection} we investigate the prospects for 
277: reconnection in the magnetar magnetosphere or in the magnetic tower, 
278: and argue that reconnection is ineffective in the dense environment 
279: deep inside the collapsing star. In Section~\ref{subsec-nickel}, we 
280: address an important issue of $^{56}$Ni production and argue that 
281: the two-phase nature of the explosion in our model is well-suited 
282: to explain a large amount of $^{56}$Ni inferred from observations. 
283: In Section~\ref{subsec-pulsar-kicks} we discuss the implications of 
284: our model for pulsar kicks. Finally, in~Section~\ref{subsec-numerical}, 
285: we suggest some directions for future numerical simulations of this 
286: problem. We draw our conclusions in~Section~\ref{sec-conclusions}. 
287: 
288: 
289: 
290: 
291: %*********************************************************************
292: 
293: \section{Magnetar inside a collapsing star: an outline of the general scenario}
294: \label{sec-stalled-shock}
295: 
296: 
297: Numerous studies of core-collapse supernovae have shown that, as the core 
298: of a massive star collapses into a proto-neutron star (PNS), a bounce shock 
299: is launched back into the star but quickly stalls at about 200~km (see, 
300: {\it e.g.}, Refs.~\cite{woosley-weaver-1986,bethe-1990}). 
301: The explosion then enters a long ($\sim$ 1~sec) phase 
302: (see Fig.~\ref{fig-stalled-shock}) during which the shock 
303: is quasi-stationary as determined by the balance between the ram 
304: pressure of the infalling material, which tends to quench the shock, 
305: and the thermal pressure of the post-shocked gas, supported by the continuous 
306: neutrino heating. Eventually, if neutrinos win, the shock engulfs the entire 
307: star and one gets a successful SN explosion. If they lose, the shock dies 
308: and the PNS collapses into a black hole that subsequently swallows the rest 
309: of the star, without a~SN.
310: 
311: 
312: In our model, we add a third dynamical component to this picture --- 
313: the magnetic field. The magnetic pressure is pushing out, helping the 
314: explosion, as is the thermal pressure of the neutrino-heated gas. Our 
315: main idea is that these two outward forces evolve differently with time, 
316: and thus the explosion may be a two-stage process. In particular, we 
317: suggest that the magnetic pressure is not important during the first 
318: few hundreds of msec of the stalled-shock phase. However, during this 
319: time the magnetar makes several hundred revolutions. This results in 
320: a great amplification of the toroidal magnetic flux by the differential 
321: rotation (see~Sec.~\ref{subsec-fixed-cavity}), whereas both the neutrino 
322: energy deposition and the accretion rate gradually decline (see Fig.~\ref
323: {fig-scenario}).
324: For example, assuming $R_0=3 R_{\rm LC}= 10 R_*=100$~km, and $B_*=10^{15}$~G, 
325: the entire cavity is filled (see~Sec.~\ref{subsec-fixed-cavity}) with 
326: $3\cdot 10^{14}$~G fields after about 100~turns (0.1~sec), corresponding 
327: to the magnetic pressure of $4\cdot 10^{27}$ erg/cm$^3$. 
328: The ram pressure of the infalling stellar material at $r=R_0$, compressing 
329: the magnetosphere, can be estimated roughly as $P_{\rm ram} \sim \dot{M} 
330: v_{\rm ff}/4\pi R_0^2 \simeq 8 \cdot 10^{27}\, \dot{M}_0\, M_0^{1/2}\, 
331: R_{0,7}^{-5/2} {\rm erg\ cm^{-3}}$
332: where $v_{\rm ff}=(2GM/R_0)^{1/2}\simeq 5\cdot 10^9\, M_0^{1/2}\, 
333: R_{0,7}^{-1/2}\, {\rm cm/sec}$ is the free-fall velocity at radius~$R_0$,
334: and~$M_0$ and~$\dot{M}_0$ are the mass within~$R_0$ and the accretion rate 
335: at this radius, expressed in units of~$M_\odot$ and~$M_\odot/{\rm sec}$, 
336: respectively.
337: This shows that, after a delay of a few hundreds of milliseconds, 
338: the magnetic pressure inevitably becomes an important driving force 
339: and may lead to a successful explosion. 
340: 
341: 
342: To summarize our picture, the ram pressure of the accreting material 
343: provides a nurturing womb in which the baby magnetic field grows, 
344: until it is finally strong enough to break out. Neutrino energy 
345: deposition plays an important role during this gestation period, 
346: as it prevents the magnetosphere from being completely squashed 
347: by the accreting gas.
348: Finally, if the above picture is correct and the explosion does 
349: become magnetically-driven, then the hoop-stress mechanism makes 
350: it highly collimated, thus satisfying one of the key necessary 
351: conditions for GRB (see~Sec.~\ref{subsec-collimation}). Note that 
352: this jet is driven by the magnetar-level ({\it i.e.}, $\sim10^{15}$~G) 
353: field and is thus much stronger and faster than the LeBlanc-Wilson 
354: jet\cite{leBlanc-wilson-1970} that may have been launched a few seconds 
355: earlier, during the core-collapse process\cite{wheeler-2000}.
356: 
357: 
358: 
359: 
360: %********************************************************************
361: 
362: 
363: 
364: \section{The Pulsar-in-a-Cavity Problem}
365: \label{sec-pulsar-cavity}
366: 
367: 
368: As a first step in trying to understand how a millisecond-magnetar 
369: central engine works in the collapsar context, let us consider the 
370: following basic plasma-physics problem: an axisymmetric pulsar inside 
371: a conducting cavity with a low-density plasma (see~Fig.~\ref{fig-magnetar-2}). 
372: Specifically, we consider the cavity's radius~$R_0$ to be is much larger 
373: than the pulsar light-cylinder radius~$R_{\rm LC}$. We call this idealized 
374: problem the {\it Pulsar-in-a-Cavity problem}\cite{um-2006a,um-2006b}.
375: It is a modification of the Goldreich \& Julian model for an isolated 
376: pulsar's magnetosphere\cite{gj-1969}; it has direct connections to the 
377: model proposed by Ostriker \& Gunn for powering longer-term supernova 
378: lightcurves\cite{og-1971} and it is also related to the models considered 
379: by Kardashev\cite{kardashev-1970} and by Illarionov \& 
380: Sunyaev\cite{illarionov-sunyaev-1975}.
381: 
382: 
383: Note that the behavior of Pulsar-in-a-Cavity depends on the cavity properties. 
384: For definiteness, we assume that both the walls and the plasma inside the 
385: cavity are perfect conductors. We also assume that all the field lines close 
386: back to the pulsar inside the cavity, keeping in mind that the magnetic field
387: had been generated inside the NS and then emerged through its surface.
388: 
389: 
390: At the same time, we are dealing with a whole family of problems 
391: distinguished by the assumed mechanical properties of the cavity.
392: For simplicity, we shall concentrate on the case of a spherical 
393: cavity with rigid walls (see Sec.~\ref{subsec-fixed-cavity}), 
394: which may correspond to early stages of the system's evolution. 
395: In the future we plan to consider more realistic but also more 
396: complicated cases where the shape and the size of the cavity are 
397: not fixed but instead are governed by a balance with an external 
398: pressure (similar to Lynden-Bell's magnetic tower model\cite{lb-1996}),
399: and, finally, will consider a young pulsar in a fully dynamic 
400: environment of a collapsing star (c.f., Ref.~\cite{og-1971}).
401: 
402: Each of these versions of our pulsar-in-a-cavity problem will provide 
403: important insights into the workings of a millisecond magnetar inside 
404: a collapsing star. They will probably require numerical simulations
405: using relativistic force-free  or  relativistic MHD codes (see~Sec.~\ref
406: {subsec-numerical}). 
407: To set the stage for these numerical studies, we will, in this section, 
408: qualitatively discuss a plausible physical picture of the system's evolution.
409: 
410: 
411: To investigate the interaction between the central magnetar and 
412: the surrounding stellar material, a full magnetohydrodynamic (MHD) 
413: description that includes plasma pressure and inertial effects will 
414: eventually be required. This is especially relevant if there is a 
415: strong wind driven off the~PNS by neutrinos and/or by the 
416: magneto-centrifugal mechanism ({\it e.g.}, Refs.~\cite{thompson-2004, 
417: bucciantini-2006,metzger-2007}). 
418: Of particular interest is the confinement of 
419: the expanding magnetosphere by the surrounding plasma and the dynamical 
420: response of the star to the expanding magnetosphere at its center. 
421: For simplicity, however, here we shall limit ourselves to 
422: the relativistic force-free case in which electromagnetic forces 
423: dominate the dynamics almost everywhere inside the cavity (except 
424: at the the equatorial plane outside the light cylinder --- see below).
425: While it is not realistic in the central region of a collapsing star, 
426: the force-free description may nonetheless reflect some essential physics, 
427: especially for late phases of the evolution.
428: 
429: 
430: We would like to remark that the ideal-MHD assumption is well justified 
431: due to the very large plasma densities and temperatures. Specifically, 
432: the high plasma density ensures that the plasma is highly collisional 
433: and hence is well described by resistive MHD; other non-ideal terms in 
434: generalized Ohm's law are unimportant. On the other hand, the resistivity 
435: is actually small in absolute terms ($Re_m \gg 1$). All this makes ideal MHD 
436: a good approximation in the environment of a collapsing star\cite{um-2006a}. 
437: Therefore, in this paper we shall ignore any non-ideal effects, leaving 
438: them for a future study.
439: 
440: 
441: As a final point, in contrast with the isolated pulsar, which is usually 
442: regarded as stationary, our pulsar-in-a-cavity problem is intrinsically 
443: time-dependent.
444: 
445: 
446: 
447: %********************************************************************
448: 
449: 
450: \subsection{Pulsar in a Fixed Spherical Cavity}
451: \label{subsec-fixed-cavity}
452: 
453: In this section we consider the case of a spherical cavity 
454: whose radius, $R_0$, is kept constant (or is changing slowly).
455: We are interested in the evolution of an initially dipole-like 
456: magnetosphere after the pulsar is spun up suddenly.
457: In the isolated pulsar case, the field lines extending beyond 
458: the light cylinder are swept backwards and open\cite{gj-1969}; 
459: there is no feedback of the outer region on the inner magnetosphere. 
460: In our case, however, the entire magnetosphere is contained inside 
461: the cavity and hence remain closed at all times. This is a very 
462: important difference between the two cases.
463: 
464: 
465: 
466: %-----------------------------------------------------------------------
467: 
468: 
469: The first thing to note is that a {\it differential rotation} 
470: is established in the magnetosphere. Indeed, consider a field 
471: line~$\Psi$ that extends beyond the light cylinder (Fig.~\ref
472: {fig-magnetar-2}) and compare the angular velocities at two 
473: points: point~$A$ where this line attaches to the pulsar and 
474: point~$B$ where it intersects the equator.
475: The angular velocity at point~$A$ is clearly equal to that of 
476: the neutron star: $\Omega_A = \Omega_*$. Next, what is the angular 
477: velocity at point~$B$? Since the field line in question extends 
478: beyond the light cylinder, it cannot remain purely poloidal and 
479: a toroidal field has to develop. Due to the assumed symmetry with 
480: respect to the equator, however, the toroidal field has to vanish 
481: at $z=0$, $B_\phi(z=0)\equiv 0$, and the plasma cannot slide 
482: toroidally backward along the field. Therefore, $\Omega_B$ has to 
483: equal the angular velocity of the plasma at this point.
484: Now, the toroidal field that develops in the magnetosphere off the 
485: equatorial plane continuously brakes the star down, so that there 
486: is an outward flux of angular momentum and a Poynting flux of energy 
487: along the field line. The rotational energy of the pulsar extracted 
488: by the magnetic field is partly accumulated in the magnetic form and 
489: partly transferred to the equatorial plasma. Thus, the material at 
490: point~$B$ is continuously torqued up by the magnetic field. Then, 
491: since the confining wall prevents the material from moving out freely 
492: in the radial direction, the toroidal velocity of the plasma becomes 
493: closer and closer to the speed of light. However, it can never exceed 
494: the speed of light; therefore, the angular velocity at point~$B$ is 
495: bounded:
496: $\Omega_B \simeq c/R_B = \Omega_* R_{\rm LC}/R_B$. 
497: Thus, we see that the field line experiences differential rotation 
498: at a rate $\Delta\Omega = \Omega_A-\Omega_B \geq \Omega_* (1-R_{\rm LC}/R_B)$;
499: For field lines with $R_B\gg R_{\rm LC}$ we get $\Delta\Omega\approx\Omega_*$. 
500: This differential rotation is established on the cavity's light-crossing 
501: time-scale, $t_0\equiv R_0/c \gg\Omega_*^{-1}$.
502: 
503: 
504: This differential rotation is important because it leads to 
505: a continuous injection of toroidal magnetic flux (of opposite 
506: signs) into the upper and lower hemispheres. If the cavity's 
507: size is fixed, then the toroidal magnetic field at any point 
508: grows roughly linearly with time. This is in sharp contrast 
509: with the unbounded pulsar case in whcih a steady state is 
510: established on the light travel time-scale.
511: 
512: 
513: %--------------------------------------------------------------------
514: 
515: Let us now analyze the field evolution on long time scales 
516: ($t \gg t_0 \equiv R_0/c$) and at distances $R\gg R_{\rm LC}$.
517: Note that the toroidal magnetic field continuously increases, 
518: whereas the poloidal magnetic field does not. 
519: The poloidal electric field, $E_{\rm pol}$,  may become much larger 
520: than~$B_{\rm pol}$ but in any case cannot exceed the value~$B_{\rm pol} 
521: \Omega_* R_0 /c=B_{\rm pol} R_0/R_{\rm LC}$.
522: Thus, after several light-crossing times ($t\gg t_0$) the magnetosphere 
523: outside the light cylinder is dominated by the toroidal field, $B_\phi\gg 
524: E_{\rm pol}, B_{\rm pol}$. 
525: 
526: Next, since $B_\phi\sim t$, the relative change in~$B_\phi$ 
527: over~$\Delta t\sim t_0$ becomes small at late times, $t \gg t_0$. 
528: An approximate force-free equilibrium is then established in each 
529: hemisphere, described by the relativistic force-free Grad--Shafranov 
530: equation. In the $B_\phi$-dominated limit this equation reduces to 
531: $II'(\Psi)=0$, where $I\equiv R B_\phi$ is the enclosed poloidal 
532: current. The obvious solution of this equation is $I(\Psi)=I_0={\rm const}$,
533: {\it i.e.}, a singular line current $I_0(t)$ along the rotation axis. 
534: The toroidal magnetic field is the vacuum field produced by this
535: line current, $B_\phi(t,R,Z)=I_0(t)/R$; {\it i.e.}, constant on cylinders. 
536: The main force balance in the magnetosphere is between the toroidal 
537: field tension and pressure. In other words, the poloidal current 
538: becomes spatially separated from the toroidal magnetic field: it flows out of 
539: the pulsar along the axis (in both hemispheres), then as a surface current 
540: along the cavity walls, and finally returns to the pulsar along the 
541: non-force-free equatorial current sheet present due to the sharp reversal
542: of the toroidal field across the equator. The current density in the bulk 
543: of the magnetosphere is relatively small. This is similar to the electric
544: current structure of the magnetic bubble considered by Lyutikov \& 
545: Blandford\cite{lyutikov-blandford-2003}.
546: 
547: We shall express magnetic quantities characterizing the field in the cavity 
548: in terms of the total poloidal magnetic flux~$\Psi_0$ extending beyond the 
549: light cylinder. This flux can be crudely estimated from the pure dipole 
550: magnetic field, {\it i.e.},
551: \beq
552: \Psi_0 \sim \Psi_{\rm dipole}(R_{\rm LC}) = 
553: B_*\ {{R_*^3}\over{R_{\rm LC}}} \, .
554: \label{eq-Psi0}
555: \eeq 
556: Then, the characteristic poloidal magnetic field strength 
557: in the cavity at distances $r\sim R_0$ from the center and 
558: off the equatorial plane can be estimated as
559: \beq
560: B_{\rm pol} \sim B_0 \equiv {\Psi_0\over{{R_0}^2}} \sim 
561: B_*\, {{R_*^3}\over{R_0^2 R_{\rm LC}}} \, .
562: \label{eq-B0-def}
563: \eeq
564: 
565: Let us now estimate the poloidal line current~$I_0(t)$
566: and hence the characteristic toroidal field in the cavity. 
567: The poloidal current is found by following the shape of a 
568: field line~$\Psi$:
569: \beq
570: I(\Psi,t) = \Delta\Omega t \ \biggl[ \int\limits_\Psi 
571: {{dl_{\rm pol}}\over{B_{\rm pol}R^2(l_{\rm pol})}} \biggr]^{-1} \, ,
572: \label{eq-I-twist}
573: \eeq
574: where $l_{\rm pol}$ is the path-length along the poloidal field.
575: The main contribution comes from large distances, $R\sim R_0$, 
576: and thus, using $\Delta\Omega\simeq\Omega_*=c/R_{\rm LC}$,
577: we get
578: \beq
579: I_0(t) \sim \Omega_* t \ {\Psi_0\over{R_0}} \simeq
580: {\Psi_0\over{R_{\rm LC}}} \, {t\over t_0} \, .
581: \eeq
582: We see that for $t\gg t_0$ the poloidal current becomes much stronger 
583: than that in the isolated pulsar magnetosphere ($I\sim\Psi_0/R_{\rm LC}$). 
584: Using the estimate~(\ref{eq-Psi0}) for~$\Psi_0$, we obtain
585: \beq
586: I_0(t) \sim B_* \, {{R_*^3}\over{R_{\rm LC}^2}} \, {t\over t_0} \, .
587: \label{eq-I_0}
588: \eeq
589: 
590: Correspondingly, the characteristic toroidal magnetic field at distances 
591: of order~$R_0$ is 
592: \beq
593: B_\phi(R_0) = {I_0\over{R_0}} \simeq B_0 \Omega_* t \, ,
594: \label{eq-B_phi}
595: \eeq 
596: which is similar to the estimate presented by Kardashev for the toroidal 
597: field of a pulsar inside an expanding supernova cavity\cite{kardashev-1970}.
598: Thus, after many light-crossing times across the cavity, 
599: $B_\phi(R_0)$ becomes much larger than the toroidal field 
600: of an isolated pulsar at these distances 
601: [$B_\phi^{\rm isolated} \sim \Psi_0/(R_0 R_{\rm LC}) =
602: B_0 (R_0/R_{\rm LC}) = B_0 \Omega_* t_0 \ll B_0 \Omega_* t $].
603: 
604: 
605: 
606: 
607: 
608: %-----------------------------------------------------------------------
609: 
610: 
611: As we noted above, the magnetosphere outside the pulsar light cylinder 
612: cannot be entirely force-free. Because the toroidal magnetic field 
613: reverses across the equator, the magnetic tension continuously accelerates 
614: the equatorial plasma in the toroidal direction. The tension force performs 
615: mechanical work on the equatorial plasma and so a part of the rotational 
616: energy extracted from the pulsar is deposited in the equatorial plane 
617: (the rest is stored magnetically in the bulk of the cavity).
618: Since the plasma in the equatorial plane rotates ultra-relativistically, 
619: the added energy leads to an increase in the relativistic ``mass'' of 
620: the plasma, $\Delta m\sim t^2$. Correspondingly, this relativistically
621: rotating massive equatorial sheet experiences an outward centrifugal 
622: force, $F_{\rm cent}$. 
623: This force cannot be balanced by the toroidal magnetic field because 
624: the latter is zero at the equator and so the equatorial plasma moves 
625: out towards the wall. It then pushes the poloidal magnetic flux out 
626: and concentrates it a narrow equatorial strip of ever-decreasing 
627: width $d(t)\ll R_0$ near the wall (see Fig.~\ref{fig-strip}).
628: Because of this, nearly all the poloidal flux~$\Psi_0$ that extends 
629: beyond the light cylinder crosses the equator at cylindrical 
630: radii~$R\simeq R_0$. At the same time, in the magnetosphere above and 
631: below the equatorial plane, the poloidal field lines that emanate from 
632: this strip fan out to fill the cavity volume. Thus, the characteristic 
633: poloidal magnetic field in the cavity is of the order~$B_0=\Psi_0/R_0^2$ 
634: (see eqn.~\ref{eq-B0-def}) and is much weaker (by a factor of $d/R_0$) 
635: than in the equatorial strip.
636: 
637: Let us assess the centrifugal force quantitatively.
638: The total torque on the massive equatorial strip is 
639: $\tau(t)=\int I(\Psi,t) d\Psi \simeq I_0(t)\Psi_0$
640: and the total work per unit time (the total Poynting
641: flux coming to the strip) is $P_{\rm strip}\simeq 
642: \tau c/R_0=B_{\phi} c\Psi_0$.
643: This power goes into accelerating the plasma rotation,
644: that is, into increasing the rotational $\gamma$-factor
645: and hence the relativistic mass~$m$ of the plasma in 
646: the strip:
647: \beq
648: m(t)c^2 = Pt \sim
649: \biggl({t\over{t_0}}\biggr)^2\, {{R_0}\over{R_{\rm LC}}}\,{\Psi_0^2\over{R_0}}
650: \sim {{R_{\rm LC}}\over{R_0}}\, B_\phi^2(t) R_0^3 \, .
651: \label{eq-strip-mass}
652: \eeq
653: That is, the kinetic energy in the equatorial strip is always small 
654: compared with the magnetic energy in the cavity, $B_\phi^2(t) R_0^3$.
655: The centrifugal force acting on the equatorial strip is
656: \beq
657: F_{\rm cent}(t) = {m(t)c^2\over{R_0}} \sim
658: B_0^2 R_0^2 \Omega_*^2 t^2\, {{R_{\rm LC}}\over{R_0}} \sim
659: B_\phi^2(t) R_0^2\, \biggl({{R_{\rm LC}}\over{R_0}}\biggr) \, .
660: \eeq
661: This force grows quadratically with time, just as the toroidal field 
662: pressure, but always remains small (by a factor of $R_{\rm LC}/R_0\ll 1$) 
663: compared with the total horizontal force exerted on the side wall by the 
664: toroidal field. However, since $F_{\rm cent}$ is concentrated in the thin 
665: equatorial region, it may be important in a subsequent expansion of the 
666: cavity (c.f. non-relativistic MHD simulations by Matt~{\it et~al.},
667: Ref.~\cite{matt-2006}).
668: 
669: 
670: A detailed analysis of the internal structure of the massive equatorial
671: plasma strip is an interesting problem that should be studied but it lies 
672: beyond the scope of this paper.
673: 
674: 
675: %-------------------------------------------------------------------
676: 
677: 
678: Another very important point is that the rate at which the magnetic field 
679: in a confined magnetosphere extracts rotational energy from the central 
680: rotating conductor actually grows with time. 
681: This is because the magnetic torque per unit area is proportional 
682: to the toroidal field at the conductor's surface and the latter 
683: grows linearly with time. Thus, as long as the cavity does not 
684: expand (or expands slowly) and the rotation rate of the spinning 
685: pulsar stays constant, the magnetic power generated by the pulsar 
686: inside a cavity increases linearly with time:
687: \beq
688: P(t) = I(t) \Psi_0 \Omega_*  = \Omega_*^2 t\ {\Psi_0^2\over{R_0}} \sim
689: P_{\rm isolated}\, {ct\over{R_0}} \, ,
690: \eeq
691: where $P_{\rm isolated}\sim B_*^2 R_*^6\Omega_*^4/c^3$ is the spin-down 
692: power of an isolated, unbounded pulsar. Hence, after many light-crossing 
693: times, the power of a pulsar-in-a-cavity greatly exceeds that of a 
694: classical isolated pulsar. This is resolves the apparent paradox 
695: raised by Lyutikov~\cite{lyutikov-2006}.
696: 
697: 
698: This runaway behavior can be attributed to a positive feedback 
699: between the energy that has been already extracted from the pulsar, 
700: and the strength of the agent that extracts the energy (the toroidal
701: magnetic field). Namely, most of the extracted energy is stored
702: in the toroidal magnetic field, and, since the volume is finite, 
703: the toroidal field strength increases with time. Since the magnetosphere 
704: remains in a quasi-equilibrium, the toroidal field constantly readjusts 
705: everywhere and the inner magnetosphere feels the presence of the outer 
706: confining wall. In particular, the toroidal field at the pulsar surface 
707: increases linearly with time, and so does the magnetic spin-down torque
708: on the pulsar. This picture is similar to the combustion chamber of a rocket.
709: In that case, the gas temperature and pressure increase as the chemical
710: energy of the fuel is released in the combustion process. At the same
711: time, the fuel burning rate grows with the ambient temperature. Therefore,
712: a rapid and efficient burning demands high pressure and is hence facilitated
713: by a strong confining chamber. Similarly, in our case of a pulsar placed
714: inside a cavity, the presence of strong cavity walls leads to an 
715: increased energy extraction rate from the pulsar.
716: 
717: 
718: 
719: In reality, we don't expect this power growth to last indefinitely.
720: It may saturate, for example, due to the development of the kink 
721: instability, resulting in the conversion of the toroidal flux to 
722: poloidal flux and to the dissipation of some of the magnetic energy 
723: (see Sec.~\ref{subsec-stability} for more discussion).
724: 
725: 
726: 
727: 
728: %***************************************************************
729: 
730: 
731: \subsection{Hoop-stress collimation: contrast with the isolated pulsar}
732: \label{subsec-collimation}
733: 
734: 
735: The toroidal field generated by the differential rotation exerts 
736: a constantly-growing pressure on the cavity walls. If we now relax 
737: the fixed-wall assumption, this pressure will inflate the cavity. 
738: Will this inflation be isotropic or, say, collimated along the axis?
739: 
740: 
741: Generally speaking, since the toroidal field pressure in the horizontal
742: direction is partly negated by the field's tension, one expects the 
743: resulting expansion to be mostly vertical. However, the differential 
744: rotation producing~$B_\phi$ is relativistic: $\Delta\Omega R_0\sim 
745: \Omega_* R_0\gg c$, and it is well-known that hoop-stress collimation 
746: is not a trivial issue in the relativistic case. Thus, it is not 
747: immediately obvious this mechanism can be applied to our pulsar-in-a-cavity 
748: scenario. 
749: The quintessential example of this lack of collimation for ultra-relativistic 
750: magnetically-dominated outflows is the isolated aligned pulsar wind inside 
751: the termination shock. The basic reason for this is the decollimating force 
752: due to the poloidal electric field, $E_{\rm pol}$. Indeed, in the case of 
753: an {\it unbounded} relativistic uniformly-rotating force-free magnetosphere 
754: in a steady state, $E_{\rm pol}$ and~$B_\phi$ are nearly equal at large 
755: distances from the axis\cite{gj-1969}. 
756: Importantly, this balance happens in an uncollimated, quasi-spherical 
757: poloidal magnetic field configuration; an excellent example of this is 
758: Michel's split-monopole solution\cite{michel-1973}. Here is a crude 
759: argument explaining 
760: this lack of hoop-stress collimation in the relativistic-rotation case.
761: Consider an uncollimated field configuration; the poloidal magnetic field 
762: is open outside the light cylinder and has a split-monopole geometry, 
763: {\it i.e.}, drops off as~$r^{-2}$. 
764: In a steady state, the poloidal electric field is $E_{\rm pol}=
765: B_{\rm pol}\, R/R_{\rm LC}$, where~$R$ is the cylindrical radius;
766: it hence drops off along radial rays as~$r^{-1}$. But $B_\phi$ 
767: also drops off as~$r^{-1}$. Moreover, at the light cylinder, 
768: $E_{\rm pol}=B_{\rm pol}\sim B_\phi$. Since outside the light 
769: cylinder they both decrease as~$r^{-1}$, they remain comparable 
770: to each other at large distances. In fact, as Goldreich \& Julian 
771: showed\cite{gj-1969}, $E_{\rm pol}$ and~$B_\phi$ become equal asymptotically 
772: as $r\rightarrow\infty$. The bottom line is that a quasi-spherical 
773: relativistic force-free equilibrium can be established as a balance 
774: between the collimating pinch force (the sum of the toroidal magnetic 
775: field pressure and its tension) and the opposing electric force. 
776: Hoop-stress collimation is suppressed as a result of this balance.
777: 
778: On the other hand, the case of a rotating magnetosphere enclosed 
779: inside a cavity is different and hoop-stress collimation can in 
780: fact work.  Indeed, as we showed above, at late times the toroidal
781: magnetic field filling the cavity becomes stronger than both~$B_{\rm
782: pol}$ and~$E_{\rm pol}$, in contrast to the isolated pulsar case. 
783: Furthermore, this toroidal field is distributed nonuniformly; namely,
784: $B_\phi\sim R^{-1}$. 
785: Correspondingly, the magnetic pressure pushing vertically against 
786: the top and bottom walls is much higher than that  on the side walls. 
787: Therefore, if we now allow the cavity to expand under this pressure, 
788: the expansion will be mostly vertical. The situation is then similar 
789: to the non-relativistic magnetic tower proposed by Lynden-Bell\cite{lb-1996}. 
790: We therefore envision that long-term result will be the creation of a 
791: pair of oppositely-directed magnetic towers\cite{um-2006a}. 
792: The interaction of the expanding towers with the surrounding stellar 
793: envelope aids in their confinement, similarly to jet collimation in 
794: hydrodynamical simulations of the collapsar model\cite{macfadyen-woosley-1999,
795: macfadyen-2001,zhang-2003}. In the scenario considered here, the towers are 
796: driven not by a differentially-rotating disk, but by a rapidly-rotating 
797: magnetar. This suggests that considering 
798: the pulsar magnetosphere inside a cylindrical, as opposed to spherical, 
799: cavity may represent yet another interesting topic for future research.
800: 
801: An important element in the above discussion is the fact that~$E_{\rm pol}$
802: is small compared with~$B_\phi$. This is because $B_\phi$ is generated as a 
803: result of differential rotation. This highlights the important role of {\it 
804: differential} rotation (as opposed to uniform relativistic rotation) in 
805: collimating relativistic force-free outflows.
806: 
807: 
808: 
809: %**********************************************************************
810: 
811: \section{Magnetic Tower Inside a Star}
812: \label{sec-tower}
813: 
814: For simplicity, in this section we shall mostly consider the case 
815: when the central engine is an accretion disk around a black hole.
816: However, we believe that the millisecond-magnetar case is essentially
817: similar.
818: 
819: %-------------------------------------------------------------
820: 
821: 
822: %********************************************************************
823: 
824: \subsection{Lynden-Bell's Original Magnetic Tower Model}
825: \label{subsec-lb-tower}
826: 
827: We suggest that the model most naturally suited to describe 
828: the propagation of a Poynting-flux dominated jet through a star is 
829: the {\it magnetic tower} model, originally introduced by Lynden-Bell 
830: in the AGN context\cite{lb-1996,lb-2003}.  A magnetic tower is an axisymmetric 
831: magnetic configuration that arises when a system of nested closed flux 
832: surfaces, anchored in a differentially-rotating disk, is twisted and, 
833: as a result, inflates, but when this inflation is controlled by a 
834: surrounding external pressure. The basic physical mechanism of this 
835: process can be described as follows (see Fig.~\ref{fig-LB-tower}).
836: 
837: Consider a thin conducting disk with some vertical magnetic flux frozen into
838: it. Let us assume that initially the magnetic field has a dipole-like topology 
839: (see Fig.~\ref{fig-LB-tower}a), with the
840: two footpoints of each field line located at different radii on the disk. 
841: Now let the disk rotate non-uniformly ({\it e.g.}, a Keplerian disk). 
842: Then, each field line~$\Psi$ is twisted at a rate $\Delta\Omega(\Psi)$ 
843: equal to the difference in angular velocities of its two footpoints. 
844: Correspondingly, toroidal magnetic flux is generated from the poloidal flux. 
845: The pressure of the toroidal field pushes the flux surfaces out, against the 
846: poloidal field tension. It is assumed that during the initial stages of this 
847: process the gas pressure, as well as the gravitational and inertial forces, 
848: are negligible in the magnetosphere, so the magnetic field is force-free. 
849: Then the expansion is uncollimated, typically at a $60^\circ$ angle with 
850: respect to the rotation axis, as shown in Figure~\ref{fig-LB-tower}b 
851: ({\it e.g.}, Refs.~\cite{lbb-1994,ukl-2002}). However, as was shown by 
852: Lynden-Bell\cite{lb-1996}, if there is some, no matter how small, external
853: gas pressure~$P_{\rm ext}$ surrounding the expanding magnetosphere, then 
854: the sideways expansion ceases once $B^2/8\pi$ drops down to~$P_{\rm ext}$.
855: Subsequently, as again was shown by Lynden-Bell\cite{lb-1996}, the twisted 
856: magnetosphere continues to expand in the vertical direction
857: (Fig.~\ref{fig-LB-tower}c) and eventually forms a cylindrical column
858: that Lynden-Bell termed a {\it magnetic tower} (see Fig.~\ref{fig-LB-tower}d). 
859: If the external pressure outside of the tower is kept constant and uniform, 
860: then the top of the tower rises at a constant speed. Plasma inertia never 
861: plays any role; the entire evolution is a sequence of force-free magnetostatic 
862: equilibria with a pressure balance between the external gas outside of 
863: the tower and the magnetic field inside.
864: 
865: Note that the assumption that both ends of the field lines connect to 
866: the disk is not essential. A similar behavior is encountered in the case
867: of a rotating conducting disk magnetically connected to a central star 
868: ({\it e.g.}, Refs.~\cite{lrbk-1995,ukl-2002}) or a rotating black hole\cite
869: {uzdensky-2005}. We believe that the model is also applicable to
870: the millisecond-magnetar central engine scenario for~GRBs.  
871:  
872: To get a physical feeling of how the magnetic tower grows, 
873: it is instructive to derive some simple order-of-magnitude 
874: estimates and scaling relationships. The main input parameters 
875: that set the characteristic physical scales are the total poloidal 
876: magnetic flux~$\Psi_0$ (per unit toroidal angle) in the tower, 
877: the characteristic differential rotation rate~$\Delta\Omega$, 
878: and the external pressure~$P_{\rm ext}$. Let us now ask how 
879: the main parameters of the tower, namely, its radius~$R_0$, 
880: the typical magnetic field~$B_0$, and the growth velocity~$V_{\rm top}$, 
881: scale with the three input parameters.
882: 
883: First, the radius of the tower, $R_0$, and the characteristic 
884: poloidal magnetic field, $B_{\rm pol}$ are related via 
885: \beq
886: B_{\rm pol} \sim B_0 \equiv {\Psi_0\over R_0^2} \, .
887: \label{eq-def-B0}
888: \eeq
889: The radius adjusts so that the magnetic pressure inside the tower 
890: equals~$P_{\rm ext}$. From the force-free balance inside the tower 
891: we expect $B_\phi\sim B_{\rm pol}\sim B_0$; hence, the total magnetic 
892: field strength at the outer edge of the tower is also of order~$B_0$. 
893: Then, from the the condition of pressure balance across the tower's 
894: wall we get
895: \beq
896: B_0 \sim \sqrt{8\pi P_{\rm ext}} \, , 
897: \label{eq-LB-side-equil}
898: \eeq
899: and so 
900: \beq
901: R_0 \sim \biggl( {\Psi_0^2\over{8\pi P_{\rm ext}}}\biggr)^{1\over4} \, .
902: \label{eq-LB-R0}
903: \eeq
904: 
905: To estimate the growth rate of the tower, note that the toroidal 
906: magnetic flux~$\chi$ is continuously generated from the poloidal 
907: flux~$\Psi_0$ by the differential rotation
908: \beq
909: \chi = 2\pi \Psi_0 N = \Psi_0 \Delta\Omega t   \, .
910: \eeq
911: 
912: Taking the tower to be a cylinder with radius $R_0$ and height~$Z_{\rm top}$, 
913: we get $B_\phi\sim \chi/R_0 Z_{\rm top}=(\Psi_0/R_0 Z_{\rm top})\,
914: \Delta\Omega t = B_0\, \Delta\Omega t (R_0/Z_{\rm top})$. However, 
915: as stated earlier, the typical toroidal field in the tower should 
916: be of the order of~$B_0$; therefore, the height of the tower increases 
917: steadily as
918: \beq
919: Z_{\rm top}(t) \sim R_0\Delta\Omega t \, .
920: \eeq
921: In other words, the tower grows at the speed of order of the typical
922: differential rotation velocity~$R_0\,\Delta\Omega$. If the external 
923: pressure does not change, the radius of the tower, determined by
924: Eq.~(\ref{eq-LB-R0}), stays constant during its growth; therefore, 
925: after many turns ($\Delta\Omega t\gg 1$), $Z_{\rm top}\gg R_0$, 
926: {\it i.e.}, the tower becomes slender.
927: 
928: Since the first analytical solution proposed by Lynden-Bell\cite{lb-1996}, 
929: the magnetic tower concept is becoming more and more accepted by the
930: astrophysical community. The formation and evolution of magnetic towers 
931: have been studied in numerical simulations\cite{li-2001,kato-2004a,
932: kato-2004b,kato-2006,nakamura-2006,nakamura-2007,ciardi-2007} and 
933: even in real laboratory experiments\cite{hsu-bellan-2002,lebedev-2005}. 
934: 
935: 
936: An interesting question is the flow of energy through a magnetic tower. 
937: As one can easily see, Poynting flux flows up from the disk along the 
938: inner segment of each field line and down to the disk along the outer 
939: segment. Indeed, for each field line~$\Psi$, the inner, faster-rotating 
940: footpoint~($1$) performs work on the magnetic field, $W_1\sim I_{\rm pol}
941: [\Psi(1)] \Omega(1)$ (per unit time and unit poloidal flux). 
942: The corresponding decelerating torque per unit flux is 
943: $\tau_1\sim I_{\rm pol}[\Psi(1)]$. In turn, the magnetic field exerts 
944: an accelerating torque per unit flux $\tau_2\sim I_{\rm pol}[\Psi(2)]$ 
945: on the outer disk footpoint of the same field line. Correspondingly, 
946: it performs work at a rate $W_2\sim I_{\rm pol}[\Psi(2)]\Omega(2)$. 
947: Because of force-free equilibrium in the tower, $I_{\rm pol}[\Psi(1)]=
948: I_{\rm pol}[\Psi(2)]=I_{\rm pol}(\Psi)$, and so $\tau_1=\tau_2$, {\it i.e.}, 
949: all the angular momentum extracted magnetically from point~$1$ is 
950: transferred to point~$2$. The two energy flows, on the other hand, 
951: are not equal: since $\Omega(1)>\Omega(2)$, the energy extracted 
952: from point~$1$ along the inner segment of the field line is greater 
953: than the energy that flows down along the outer segment and is 
954: deposited in the disk at point~$2$. The difference, proportional 
955: to $I_{\rm pol}(\Psi)\Delta\Omega(\Psi)$, is the power driving 
956: the expansion of the tower. A part of it goes into filling the 
957: growing volume of the tower with magnetic energy, and the rest 
958: goes into performing work against external gas pressure and driving 
959: the shock through the star. 
960: 
961: It is interesting to note that the total vertical Poynting flux in 
962: the two segments only involves the differential rotation~$\Delta\Omega
963: =\Omega(1)-\Omega(2)$, but is independent of the absolute rotation itself.
964: This is because we are dealing here with a force-free equilibrium, so that 
965: $I_{\rm pol}$ is constant along the entire length of a field line; in 
966: particular, it has the same sign on the two segments of the field line,
967: and hence so does~$B_\phi$. The situation is drastically different
968: in the relativistic-rotation case where both field-line segments extend
969: beyond their respective light cylinders. In that case, one no longer has 
970: a force-free equilibrium along the entire field line; in particular, 
971: equilibrium breaks down at the farthermost tip of the line where the 
972: two segments join. As a result, the signs of~$I_{\rm pol}$ (and hence 
973: of~$B_\phi$) on the two segments are opposite, which corresponds to 
974: both segments being swept back.
975: Consequently, the Poynting flux is outward along both segments.
976: A similar situation arises in the non-force-free MHD case;
977: the two field-line segments are then swept back by plasma
978: inertia if they extend beyond the Alfv\'en point. This 
979: again results in an outward Poynting flux along both segments.
980: In both of these cases, the total vertical Poynting flux depends
981: on the absolute rotation rates~$\Omega(1)$ and~$\Omega(2)$ themselves, 
982: as opposed to just their difference.
983: 
984: 
985: 
986: %*********************************************************************
987: 
988: \subsection{Magnetic Tower Driving a Shock through a Star}
989: \label{subsec-tower-star}
990: 
991: There are several reasons that make the magnetic tower 
992: an attractive model for the formation and propagation 
993: of a magnetically-dominated jet through a star within 
994: the collapsar model for GRBs (and core-collapse SNe). 
995: First, a configuration where all the field lines close back onto 
996: the central engine (an accretion disk or a magnetar) is natural 
997: for a field created by a dynamo with zero net flux. In addition, 
998: mixing of baryons from the stellar envelope via Kelvin-Helmholtz
999: and/or Rayleigh-Taylor instabilities may be inhibited by the magnetic 
1000: field.
1001: 
1002: In order to apply the magnetic tower model to the collapsar scenario, we 
1003: first want to make some modifications to Lynden-Bell's picture\cite{um-2006a}.
1004: Specifically, we take into account the high-pressure cocoon that 
1005: surrounds and confines the tower (see Fig.~\ref{fig-star}). 
1006: In our model, the magnetic tower grows very rapidly and acts as 
1007: a piston driving a shock ahead of itself. The shocked gas above 
1008: the tower has very high pressure; it squirts sideways and forms
1009: backflows that fill the cocoon around the tower. Therefore, the 
1010: external pressure confining the tower is no longer an arbitrary 
1011: parameter, as in Lynden-Bell's model, but is determined by the 
1012: jump conditions across the shock surrounding the cocoon and across 
1013: the contact discontinuity between the cocoon and the tower. 
1014: The external unperturbed pressure of the star is actually irrelevant; 
1015: it should thus be excluded from our three input parameters. Instead, 
1016: the expansion is controlled by the ram pressure related to the gas 
1017: {\it inertia}; therefore, we replace~$P_{\rm ext}$ by the unperturbed 
1018: stellar density~$\rho_0$ in the list of basic dimensional parameters 
1019: (along with $\Psi_0$ and~$\Delta\Omega$) that determine the physical 
1020: scales in our problem. This change is an important difference between 
1021: our model and Lynden-Bell's.
1022: 
1023: 
1024: The actual situation is complicated further by the two-dimensional 
1025: character of the problem. Since the sound speed in the cocoon is 
1026: very high, gas pressure tends to be equalized throughout the cocoon. 
1027: This expectation is supported by the hydrodynamic simulations of the 
1028: collapsar model\cite{macfadyen-2001,zhang-2003}, which show a relatively 
1029: weak (just a factor of~5 or~10) variation of the cocoon pressure along 
1030: its length. This is very moderate compared to the corresponding variations 
1031: of the unperturbed stellar density and pressure, which both vary by many 
1032: orders of magnitude along the vertical extent of the cocoon. Thus, the gas 
1033: pressure is very high everywhere in the cocoon and so the cocoon also drives 
1034: a sideways shock into the star. The boundary between the tower and the cocoon 
1035: is a contact discontinuity, whereas the boundary between the cocoon and the 
1036: rest of the star is a two-dimensional strong shock of some complicated shape.
1037: 
1038: 
1039: %************************************************************************
1040: 
1041: \subsection{Simple Estimates}
1042: \label{subsec-estimates}
1043: 
1044: Let us now show how the basic parameters of the growing magnetic tower 
1045: scale with~$\Psi_0$, $\Delta\Omega$, and~$\rho_0$. We shall ignore any 
1046: non-uniformity of the gas pressure in the cocoon. 
1047: Also, for simplicity we shall use one-dimensional shock jump conditions. 
1048: Since the pressure of the unperturbed stellar gas upstream of the shock 
1049: is neglected, the shock is strong. Assuming an adiabatic index of~5/3, 
1050: the shock velocity with respect to the unperturbed gas is $V_s=4/3 \,
1051: V_{\rm top}$, whereas the pressure in the post-shock region ({\it i.e.}, 
1052: in the cocoon) is  expressed in terms of the velocity of the piston 
1053: $V_p\equiv V_{\rm top}$ and the upstream gas density~$\rho_0$ as 
1054: \beq
1055: P_{\rm top} = {3\over 4}\, \rho_0 V_s^2 = {4\over 3}\,\rho_0 V_{\rm top}^2 \, .
1056: \label{eq-shock-pressure}
1057: \eeq
1058: 
1059: By comparing this with the pressure balance $P_{\rm top}\simeq B_0^2/8\pi$ 
1060: at the contact discontinuity at the top of the tower, we see that the tower 
1061: grows with a velocity of order the Alfv\'en speed corresponding to~$\rho_0$:
1062: \beq
1063: V_{\rm top} \sim V_A \equiv {B_0\over\sqrt{4\pi\rho_0}} =
1064: {\Psi_0\over{R_0^2\sqrt{4\pi\rho}}}   \, .
1065: \label{eq-V_top=V_A}
1066: \eeq
1067: But, as we have shown earlier, $V_{\rm top}$ should be of the order 
1068: of~$R_0\Delta\Omega$. Thus, we obtain the scaling of~$R_0$ with~$\Psi_0$, 
1069: $\Delta\Omega$, and~$\rho_0$:
1070: \beq
1071: R_0 \sim 
1072: \biggl({\Psi_0\over{\Delta\Omega}}\biggr)^{1/3}\, (4\pi\rho_0)^{-1/6} \, .
1073: \label{eq-R0}
1074: \eeq
1075: We can also relate the radius of the tower to the radius of the central 
1076: rotating conductor. Estimating the poloidal flux as~$\Psi_0\sim B_d R_d^2$,
1077: where $B_d$ and~$R_d$ are the typical magnetic field and the radius of the
1078: the base of the tower ({\it e.g.}, the inner part of the accretion disk), 
1079: we we get $B_0 \sim B_d\, (R_d/R_0)^2$ and hence
1080: \beq
1081: {R_0\over{R_d}} \sim \biggl({{\tilde{V}_{\rm A,d}}\over{V_d}}\biggr)^{1/3}\,.
1082: \label{eq-R0-Rd}
1083: \eeq
1084: Here $V_d \equiv R_d \Delta\Omega$ is the characteristic 
1085: differential rotation velocity of the central conductor 
1086: and 
1087: \beq
1088: \tilde{V}_{\rm A,d} \equiv {B_d\over{\sqrt{4\pi\rho_0}}} 
1089: \eeq
1090: is a composite Alfv\'en speed involving the disk magnetic field 
1091: and the unperturbed star's plasma density; it doesn't have a direct 
1092: physical meaning and thus can be arbitrarily high.
1093: 
1094: 
1095: Notice that, as the tower makes its way through the star, $\Psi_0$ 
1096: and~$\Delta\Omega$ remain unchanged, whereas the third parameter, 
1097: the unperturbed density~$\rho_0$ at the top of the tower, changes. 
1098: It drops rather rapidly for a typical collapsar progenitor and so
1099: the radius of the tower increases as the it grows. Thus we expect
1100: that, in a realistic situation, the tower will not be a straight
1101: cylinder, as we have assumed here; instead, its radius will be some
1102: function of the vertical coordinate~$z$. 
1103: However, as is seen from Eq.~(\ref{eq-R0}), $R_0$ scales only 
1104: weakly with~$\rho_0$ (as $\rho_0^{-1/6}$), which somewhat justifies 
1105: the constant-radius approximation.
1106: 
1107: 
1108: 
1109: %********************************************************************
1110: 
1111: \subsection{The Numbers}
1112: \label{subsec-numbers}
1113: 
1114: 
1115: Now let us make some quantitative estimates based on the above scaling 
1116: relationships; unavoidably, these estimates will be very crude.
1117: For definiteness, consider the accreting black hole scenario for 
1118: the central engine.
1119: We assume that the core of the star has collapsed into a black hole 
1120: of fiducial mass $M=3\, M_{\odot}$ with a gravitational radius $R_g 
1121: \equiv GM/c^2 \simeq 5\, {\rm km}$, and that some of the continuously 
1122: infalling material has formed an accretion disk around the black hole. 
1123: We take a fiducial disk radius of $R_d \simeq 6 R_g \simeq 3\cdot 10^6\, 
1124: {\rm cm}$ and an initial poloidal flux (per unit toroidal angle) of~$\Psi_0= 
1125: R_d^2 B_0 \simeq 10^{28} B_{15} R_{d,6.5}^2$ in cgs units.
1126: This poloidal flux is being continuously twisted by the differential
1127: rotation of the disk, with characteristic angular velocity 
1128: $\Delta\Omega = \Omega_K(R_d) \simeq 4\cdot 10^{3} {\rm sec^{-1}} 
1129: (M/3M_\odot)^{1/2}\, R_{d,6.5}^{-3/2}$.
1130: 
1131: 
1132: Taking the fiducial stellar background density to be $\rho_0=10^6\,
1133: {\rm g/cm^3}$, the tower outer radius can be estimated as
1134: \beq
1135: R_0 \sim R_d\, \biggl({{\tilde{V}_{A,d}}\over{V_d}}\biggr)^{1\over 3} =
1136: 3\, R_d\, \biggl({{B_{d,15}}
1137: \over{R_{d,6.5}\Delta\Omega_{3.5}\sqrt{\rho_{0,6}}}}\biggr)^{1/3}\, ,
1138: \label{eq-scaling-R_0}
1139: \eeq
1140: resulting in $R_0\simeq 10^7\, {\rm cm}$ for our fiducial parameter values.
1141: Then we get the following expressions for all the other parameters:
1142: \begin{eqnarray}
1143: B_0 &\equiv& {\Psi_0\over{R_0^2}} = 
1144: B_{\rm d} \biggl({R_{\rm d}\over{R_0}}\biggr)^2     \nonumber   \\
1145: &\simeq&  0.1\, B_{\rm d}\, 
1146: \biggl({{R_{d,6.5}\Delta\Omega_{3.5}}\over{B_{\rm d,15}}}\biggl)^{2/3}\, 
1147: \rho_{0,6}^{1/3} \simeq 10^{14}\, {\rm G}          
1148: \label{eq-scaling-B_0}                \, ; \\ 
1149: V_{A,0} &\equiv& {B_0\over{\sqrt{4\pi\rho_0}}} =
1150: 3\cdot 10^{10}\, {\rm cm/sec}\ B_{d,15}^{1/3}\, R_{d,6.5}^{2/3}\,
1151: \Delta\Omega_{3.5}^{2/3}\, \rho_{0,6}^{-1/6}    
1152: \label{eq-scaling-V_A}            \, .
1153: \end{eqnarray}
1154: 
1155: Notice that our crude estimate results in $V_{\rm top}\sim V_{A,0}$ 
1156: being comparable to the speed of light~$c$. Therefore, a fully-relativistic 
1157: treatment of the problem would be more appropriate (see Sec.~\ref{subsec-rel} 
1158: for discussion). Such a treatment, however, lies beyond the scope of 
1159: the current paper.
1160: 
1161: Also, we can estimate the post-shock pressure in the hot cocoon 
1162: above the tower as
1163: \beq
1164: P_{\rm top} \simeq {{B_0^2}\over{8\pi}} \simeq 
1165: 4\cdot 10^{26}\, {\rm erg\ cm^{-3}}\ B_{0,14}^2 \, .
1166: \eeq
1167: where $B_{0,14}\equiv B_0/(10^{14}\,{\rm G})$.
1168: At such high energy density the radiation pressure probably dominates 
1169: over the gas pressure; we can therefore estimate the plasma temperature 
1170: in the post-shock region as
1171: \beq
1172: T_{\rm top} \simeq \biggl({3P_{\rm top}\over a}\biggr)^{1/4} \simeq
1173: 2\cdot 10^{10}\, {\rm K} \simeq 2\, {\rm MeV}\, ,
1174: \eeq
1175: where $a\simeq 7.6\cdot 10^{-15}\, {\rm erg\, cm^{-3}\, K^{-4}}$.
1176: On the other hand, since we are dealing with a strong hydrodynamic
1177: shock between the cocoon and the unperturbed stellar material, the 
1178: baryon density in the cocoon is simply $4\rho_0$, and so the 
1179: baryon rest-mass energy density is 
1180: $4\rho_0 c^2 \simeq 4\cdot 10^{27}{\rm erg\ cm^{-3}}\,\rho_{0,6}$,
1181: and hence still exceeds the radiation/pair energy density by an 
1182: order of magnitude.
1183: The total magnetic energy contained in the tower of height~$Z_{\rm top}$
1184: can be estimated as 
1185: \beq
1186: E_{\rm mag}(t) \simeq 2\pi R_0^2\, Z_{\rm top}(t)\, {{B_0^2}\over{8\pi}}
1187: \simeq 2\cdot 10^{50}\, {\rm erg}\ R_{0,7}^2\, Z_{\rm top,9}\, B_{0,14}^2 \, ,
1188: \label{eq-Emag}
1189: \eeq
1190: which is  a noticeable fraction of a typical GRB energy.
1191: 
1192: 
1193: 
1194: 
1195: 
1196: %*********************************************************************
1197: 
1198: 
1199: \section{Discussion}
1200: \label{sec-discussion}
1201: 
1202: 
1203: 
1204: %-----------------------------------------------------------------
1205: 
1206: \subsection{Transition to Relativistic Expansion Regime 
1207: and the Jet Opening Angle}
1208: \label{subsec-rel}
1209: 
1210: As we have shown  above, the tower radius~$R_0$, and hence its growth 
1211: velocity $V_{\rm top}\sim \Delta\Omega R_0$ scale with background density
1212: as~$\rho_0^{-1/6}$. Therefore, as the tower expands into the outer regions 
1213: of the star, $V_{\rm top}$ inevitably reaches the speed of light at some 
1214: critical density~$\rho_{0,\rm rel}$.
1215: For our fiducial values $B_d= 10^{15}\ {\rm G}$, $R_d=3\cdot10^6\ {\rm cm}$, 
1216: and $\Delta\Omega= 3\cdot 10^3\ {\rm sec^{-1}}$, this critical density for 
1217: the transition to the relativistic regime is $\rho_{0,\rm rel}\sim 10^6\ 
1218: {\rm g/cm^3}$ [see eq.~(\ref{eq-scaling-V_A})]. For a typical massive 
1219: stellar GRB progenitor this corresponds to a distance $Z_{\rm rel}$ from 
1220: the center on the order of $10^8\ {\rm cm}$; according to Eq.~(\ref
1221: {eq-scaling-R_0}), the corresponding radius of the tower is 
1222: $R_{\rm rel}\sim 10^7\ {\rm cm}$. 
1223: 
1224: Our non-relativistic model becomes becomes invalid at this point and 
1225: the subsequent expansion of the tower calls for a relativistic generalization 
1226: of the magnetic tower model, which is beyond the scope of this paper. 
1227: However, we probably can derive some physical insight into the relativistic
1228: regime by looking at the results of fully-relativistic {\it hydrodynamic} 
1229: simulations\cite{zhang-2003}. 
1230: In those simulations the relativistic jet remained collimated as it 
1231: propagated through the star, across several orders of magnitude in~$\rho_0$. 
1232: This has been attributed to recollimation shocks in the cocoon and 
1233: to relativistic beaming in the jet. But the physical processes in 
1234: the cocoon should not change if we replace the inner relativistic 
1235: hydrodynamic jet with a relativistic magnetic tower. Furthermore,
1236: we expect some additional magnetic collimation due to the hoop stress.
1237: To sum up, we expect the magnetic tower to remain collimated even after 
1238: it transitions into the relativistic regime. In particular, we suggest 
1239: that the final opening angle of the outflow will be about the inverse 
1240: aspect ratio of the tower at relativistic transition:
1241: \beq
1242: \Delta\theta \lesssim {{R_{0,\rm rel}}\over{Z_{\rm rel}}} \simeq 0.1 \, .
1243: \eeq
1244: Whether this prediction is true will have to be determined by
1245: a fully-relativistic analysis and by relativistic MHD simulations,
1246: which we hope will be completed in the near future. 
1247: 
1248: 
1249: %*******************************************************************
1250: 
1251: \subsection{Small-Scale Structure of the Outflow}
1252: \label{subsec-small-scale}
1253: 
1254: The physical picture presented in this paper, with its smooth coherent
1255: magnetic structure, is just an idealization used to get the main ideas 
1256: across in the clearest way possible.
1257: The actual magnetic field, especially if it is produced by a turbulent
1258: dynamo in the PNS or in an accretion disk, will of course be different 
1259: from such a simple axisymmetric system of nested flux surfaces. Instead, 
1260: it may consist of an ensemble of loops of different sizes and orientations.
1261: It may thus have a highly-intermittent substructure on smaller scales, 
1262: both temporal and spatial. 
1263: However, each of these smaller magnetic structures is subject to the same
1264: physical processes as the simple large-scale configuration: twisting by
1265: differential rotation and a subsequent inflation controlled by the external 
1266: pressure of the cocoon and of the other loops growing at the same time. 
1267: As a result, a more realistic picture may look like a train of spheromak-like 
1268: plasmoids, pushing each other out along the axis (see Fig.~\ref{fig-train}). 
1269: Hoop stress still works inside each of them, and so the overall dynamical 
1270: effect may be qualitatively similar to that of a single tower. A similar 
1271: picture may also develop if a large-scale twisted magnetosphere becomes 
1272: unstable to the kink instability and undergoes flux conversion as a result, 
1273: breaking up into smaller plasmoids (see below). In either case, the resulting 
1274: multi-component structure of the outflow may be responsible for the observed 
1275: intermittency in GRBs.
1276: 
1277: 
1278: 
1279: %****************************************************************
1280: 
1281: 
1282: \subsection{Effect of MHD Instabilities on the Tower Evolution}
1283: \label{subsec-stability}
1284: 
1285: 
1286: One of greatest uncertainties in our model is the effect of 
1287: MHD instabilities in the highly twisted magnetic structure. 
1288: In this section we discuss two such instabilities: Rayleigh--Taylor 
1289: and kink.
1290: 
1291: 
1292: ({\it i}) Rayleigh--Taylor (or its magnetic counter-part,  
1293: Kruskal--Schwarzschild) instability may occur at the interface 
1294: between the growing magnetic bubble and the overlying colder, 
1295: denser stellar material. This instability is expected to affect
1296: purely hydrodynamic fireball models as well; if anything, strong 
1297: magnetic fields are expected to suppress it somewhat, although 
1298: probably not completely.
1299: The Rayleigh--Taylor instability may cause splitting of a coherent 
1300: magnetic structure into several separate strands interlaced with 
1301: stellar matter\cite{wheeler-2000,arons-2003}. For example, in the 
1302: Arons model\cite{arons-2003}, the stellar envelope is ``shredded'' 
1303: by the nonlinear Rayleigh-Taylor ``fingers''. This leads to creation 
1304: of several evacuated channels that allow the electromagnetic energy 
1305: produced near the central engine to escape through the star. Arons 
1306: further argued that these channels suffer only a small amount of 
1307: mixing with the non-relativistic stellar material due to the 
1308: Kelvin-Helmholtz instability. In light of his work, we think that 
1309: our magnetic cavity and/or the subsequently-formed magnetic towers 
1310: may also suffer from fragmentation into Rayleigh-Taylor ``fingers''. 
1311: However, as is well known, initially small-scale fingers quickly 
1312: merge to form a few large ones in the nonlinear stage. Therefore,  
1313: one does not expect strong mixing of the baryons from the stellar 
1314: envelope into the magnetosphere. The exact geometry of the outflow 
1315: may be affected somewhat and a strong time-variability may develop,
1316: but, overall, we expect the outflow to survive. More research is needed 
1317: in order to assess the implications of this instability for our scenario.
1318: 
1319: 
1320: ({\it ii}) As the magnetic configuration is twisted up, it may become 
1321: prone to a non-axisymmetric kink-like instability. This may happen
1322: both during the pulsar-in-a-cavity phase and during a later magnetic 
1323: tower phase.
1324: Whereas the stability of the pulsar-in-a-cavity has not yet been studied,
1325: several non-relativistic 3D MHD simulations have recently addressed\cite
1326: {kato-2004b,nakamura-2007,ciardi-2007} the kink instability of magnetic 
1327: towers (although not in the GRB context). 
1328: They seem to indicate that during the first few rotation periods, the tower
1329: is stabilized by the surrounding high-pressure gas, but at later times a 
1330: large-scale external kink does develop. As a result, the tower's general 
1331: shape becomes helical. 
1332: This, however, does not immediately lead to the overall disruption of 
1333: the tower; even though the configuration is nonaxisymmetric, its main 
1334: morphological features remain similar to those in the axisymmeitrc 
1335: case\cite{nakamura-2007}. 
1336: Similar conclusions have been reached by Nakamura \& Meier\cite
1337: {nakamura-meier-2004} in their 3D-MHD study of Poynting-flux-dominated 
1338: jets propagating through a stratified external medium. In particular, 
1339: these authors found that a steep external pressure gradient forestalls
1340: the instability onset. When the instability does eventually develop, 
1341: the resulting helical structures saturate and do not develop into 
1342: full MHD turbulence.
1343: Important theoretical evidence supporting the idea of the external 
1344: pressure stabilization follows from K\"onigl \& Choudhuri's analysis 
1345: of a force-free magnetized jet confined by an external pressure\cite
1346: {konigl-choudhuri-1985}. They showed that a non-axisymmetric helical 
1347: equilibrium state becomes energetically favorable (conserving the total 
1348: magnetic helicity in the jet) only when the pressure drops below a certain 
1349: critical value. If this happens and the external kink does go unstable, 
1350: then this non-axisymmetric equilibrium can be interpreted as the end point 
1351: of the non-linear development of the instability.
1352: 
1353: 
1354: In addition to the above non-relativistic studies, several first steps have
1355: recently been undertaken to understand the stability of relativistic jets, 
1356: in particular in the framework of relativistic force-free electrodynamics\cite
1357: {gruzinov-1999,tomimatsu-2001}. However, so far as we know, there have been 
1358: no formal stability studies of relativistic magnetic towers to date. 
1359: Such studies, both analytical and numerical, are clearly needed. 
1360: They would have to take into account several stabilizing effects. 
1361: First, as Tomimatsu {\it et al.} have found in their linear stability 
1362: analysis of a narrow rotating relativistic force-free jet, rapid 
1363: field-line rotation inhibits the kink instability\cite{tomimatsu-2001}. 
1364: Second, we expect that the tower expansion should quickly transition 
1365: to the relativistic regime (see Sec.~\ref{subsec-rel}), eventually 
1366: reaching a very large $\gamma$-factor. Once this happens, the relativistic 
1367: time delay may effectively stabilize the tower\cite{giannios-spruit-2006}. 
1368: This is because MHD instabilities grow on the local Alfv\'en-crossing time 
1369: in the fluid frame and hence much slower in the laboratory frame. 
1370: As a result, even if instabilities are excited, they do not have 
1371: enough time to develop before the break-out of the flow from the star. 
1372: 
1373: 
1374: If it does develop, the kink is probably the most dangerous instability
1375: and  may lead to a significant, although perhaps temporary, disruption. 
1376: Such a disruption, however, is not necessarily a bad thing: the tower 
1377: may be able to reform after being disrupted (as is seen in laboratory 
1378: experiments\cite{lebedev-2005}) and the resulting non-steady evolution 
1379: may provide a plausible mechanism for rapid variability seen in gamma-ray 
1380: bursts. Also, as a result of such disruption, a significant fraction of 
1381: the toroidal magnetic field energy may be dissipated into thermal energy\cite
1382: {eichler-1993,begelman-1998}, which may in fact contribute to the acceleration 
1383: of the Poynting-flux dominated outflow and to powering the prompt gamma-ray 
1384: emission at later times\cite{drenkhahn-spruit-2002,giannios-spruit-2005,
1385: giannios-spruit-2006,giannios-spruit-2007}. In any case, the kink, and 
1386: especially its nonlinear outcome, is a serious issue that needs to be 
1387: addresses in the future. Axisymmetric sausage instability also needs 
1388: to be investigated.
1389: 
1390: 
1391: An important aspect of our problem is that we are actually interested 
1392: not so much in the instability onset or its early linear development, 
1393: but rather in the long-term (many rotation periods) nonlinear evolution 
1394: and its overall effect on the outflow. An important consideration that 
1395: then needs to be taken into account is the conservation of magnetic helicity. 
1396: Differential rotation leads to a continuous injection of helicity into the 
1397: system (of opposite signs in the two hemispheres). In its nonlinear stage, 
1398: the kink instability may lead to conversion of some of the toroidal magnetic 
1399: flux to poloidal flux (in our geometry); however, it will not destroy the 
1400: magnetic helicity accumulated in the cavity. Thus, whatever the resulting 
1401: configuration will be, it will have to be consistent with a growing amount 
1402: of helicity. If some of the new poloidal flux becomes detached from the base
1403: due to reconnection (which may actually be strongly inhibited, see discussion 
1404: in~Section~\ref{subsec-reconnection}), then the resulting configuration may
1405: resemble a train of plasmoids (Fig.~\ref{fig-train}), similar to 
1406: Section~\ref{subsec-small-scale}. One may in fact imagine a cyclic process 
1407: involving the growth of the tower for several rotation periods, followed by 
1408: flux conversion due to the kink, followed by reconnection and subsequent 
1409: ejection of a plasmoid carrying the magnetic helicity (and some of the 
1410: magnetic energy) injected during the given cycle.
1411: 
1412: 
1413: It is also interesting to make the following observation.
1414: An unbounded relativistic force-free outlow from a rotating conductor 
1415: is expected to be stable. On the other hand, in the case of a confined
1416: closed magnetosphere with field lines subject to differential rotation, 
1417: such as our pulsar-in-a-cavity problem or a magnetic tower,  kink is 
1418: expected to develop. At the same time, as we discussed in Sec.~\ref
1419: {subsec-collimation}, the outflow is uncollimated in the first case
1420: but is collimated in the second case. This suggests that there may 
1421: be a deep connection between stability and lack of collimation of 
1422: axisymmetric relativistic force-free flows.
1423: 
1424: 
1425: Finally,  we would like to reiterate that a proper treatment of 
1426: these problems requires a time-dependent relativistic force-free 
1427: or full (preferably relativistic) MHD analysis and simulations 
1428: (see~Sec.~\ref{subsec-numerical}).
1429: 
1430: 
1431: %****************************************************************
1432: 
1433: \subsection{Reconnection}
1434: \label{subsec-reconnection}
1435: 
1436: Another important issue is magnetic reconnection across the equatorial 
1437: current sheet in the  pulsar-in-a-cavity magnetosphere or across the 
1438: separatrix current sheet in the magnetic tower. This process may, in 
1439: principle, lead to the break-up of a single structure into several 
1440: smaller spheromak-like plasmoids (see Fig.~\ref{fig-train}), similar 
1441: to the cyclic bevavior (involving reconnection) suggested for 
1442: magnetospheres of accreting young stars\cite{goodson-1999,uzdensky-2004}. 
1443: The expected size of the plasmoids and their production rate are 
1444: presently not known.
1445: 
1446: We would like to remark, however, that fast a reconnection is difficult 
1447: to achieve deep inside a collapsing star\cite{um-2006a}. 
1448: The reason for this is that fast Petschek-like reconnection 
1449: is now believed to be possible only in {\it collisionless} environments, 
1450: such as the Solar corona, Earth magnetosphere, and tokamak plasmas. 
1451: The plasma inside a collapsing star, on the other hand, is highly
1452: collisional\cite{um-2006a}, so that classical collisional resistivity 
1453: dominates over all other non-ideal effects in generalized Ohm's law. 
1454: It is now believed, on the basis of numerical simulations\cite{biskamp-1986,
1455: uzdensky-kulsrud-2000}, theoretical analysis\cite{kulsrud-2001,malyshkin-2005},
1456: and laboratory experiments\cite{ji-1999}, that reconnection in such a 
1457: situation proceeds in the very slow Sweet--Parker\cite{sweet-1958,parker-1957} 
1458: regime, whereas Petschek's\cite{petschek-1964} fast mechanism fails. It is of 
1459: course not obvious that this conclusion can be extended to the highly 
1460: relativistic and optically-thick electron-positron plasma in the deep 
1461: interior of a collapsing star. However, at the very least, this observation 
1462: casts a serious doubt on the possibility of fast and efficient large-scale 
1463: reconnection in this environment.
1464: As the magnetic tower grows and eventually breaks out of the star, 
1465: however, the plasma cools and the particle density in it drops rapidly. 
1466: At some point, the plasma becomes collisionless (from the reconnection
1467: point of view) and this opens up the possibility of reconnection and
1468: the corresponding delayed magnetic energy release. Indeed, post-breakout 
1469: reconnection in relativistic Poynting-flux-dominated outflows has 
1470: been invoked as a plausible mechanism for powering GRB emission\cite
1471: {lyutikov-blackman-2001,drenkhahn-spruit-2002,lyutikov-blandford-2003,
1472: giannios-spruit-2005,giannios-spruit-2006,giannios-spruit-2006,lyutikov-2006}.
1473: 
1474: 
1475: 
1476: %*********************************************************************
1477: 
1478: 
1479: \subsection{Nickel Production}
1480: \label{subsec-nickel}
1481: 
1482: 
1483: A key issue for long-duration GRBs is the required production of~$^{56}$Ni.  
1484: The supernovae observed to accompany these GRBs (SN-GRBs) belong to Type~Ibc 
1485: \cite{soderberg-2006,kaneko-2007}. They are believed to require radioactive 
1486: $^{56}$Ni to heat the ejecta after the initial expansion of the star. 
1487: The brightest SN-GRBs ({\it e.g.}, SN1998bw and~SN2003dh) require up 
1488: to several 0.1~M$_{\odot}$ of~$^{56}$Ni, as inferred from peak optical 
1489: brightness, although on average SN-GRBs do not need more $^{56}$Ni than 
1490: the local population of SNe\cite{soderberg-2006}. In fact, some SN-GRBs 
1491: ({\it e.g.}, GRB060505 and~GRB060614; see Refs.~\cite{fynbo-2006,
1492: dellaValle-2006,gal-yam-2006}) may produce little or no~$^{56}$Ni
1493: \cite{macfadyen-2003}.
1494: 
1495: In models of core-collapse SNe, $^{56}$Ni is produced in material heated to 
1496: $T\gtrsim T_{\rm Ni} \sim 5\times 10^9$~K by the explosion shock launched in 
1497: the core of the star.  
1498: The amount of $^{56}$Ni produced depends on the mass inside of the expanding
1499: shock when its temperature declines below~$T_{\rm Ni}$. This happens when its
1500: radius has reached
1501: \beq
1502: R_{\rm Ni}\sim\left(\frac{3E}{4\pi a T_{\rm Ni}^4}\right)^{1/3} \sim
1503: 4 \times 10^8\, E_{51}\, {\rm cm} \, , 
1504: \eeq
1505: where $E=E_{51}\times 10^{51}$~erg is the explosion energy and~$a$ is 
1506: the radiation constant. The mass inside $R_{\rm Ni}$ depends on the 
1507: progenitor structure and on the expansion or contraction that took place
1508: before the shock reached a given mass element. In particular, little 
1509: $^{56}$Ni is produced by a shock, even if very powerful, if it is 
1510: launched into a low density material, as it may occur if a weak initial 
1511: explosion pre-inflates the stellar core so that little mass remains 
1512: within a few $10^8$~cm when the subsequent strong shock arrives. 
1513: Production of $\sim 0.1 M_\odot$ of $^{56}$Ni usually requires 
1514: $\sim 10^{51}$~ergs to be deposited isotropically by a quasi-spherical 
1515: shock within $\sim 1$~sec, so that little pre-expansion of the star occurs.
1516: Brightest supernovae, {\it e.g.}, SN1998bw, require energies of up to
1517: $\sim 10^{52}$~ergs to make the $\sim 0.5 M_\odot$ inferred from 
1518: lightcurve modeling.
1519: 
1520: The requirement of fast ($\lesssim 1$~sec), isotropic deposition of energy 
1521: for production of $^{56}$Ni is a serious challenge for models of the SN-GRB 
1522: central engine, because the GRB engine must last for~$\gtrsim 10$~sec for 
1523: relativistic ejecta to escape the star and because GRBs are believed to 
1524: be collimated explosions.  
1525: The high degree of beaming and long timescale for energy deposition makes
1526: collapsar jets incapable of producing anywhere near the required amounts 
1527: of~$^{56}$Ni\cite{macfadyen-woosley-1999}.  
1528: Therefore, in the original collapsar model, with a black hole accretion disk 
1529: as the central engine, the $^{56}$Ni is produced by a slower bi-conical disk 
1530: wind that constitutes a distinct explosion component
1531: \cite{macfadyen-woosley-1999,macfadyen-2003}.
1532: In our magnetar model, on the other hand, $^{56}$Ni can be produced behind 
1533: a roughly spherical shock driven by the initial quasi-isotropic expansion 
1534: of the magnetosphere. 
1535: The expansion becomes collimated and a magnetic tower forms only at 
1536: a later stage. Thus, the collimation process involves both a quick 
1537: isotropic expansion followed by a beamed component.  We feel that 
1538: this modification to the magnetar scenario, {\it i.e.}, the inclusion 
1539: of the magnetosphere interaction with the surrounding star, strengthens 
1540: its viability as a model for the long GRB central engine.
1541: 
1542: 
1543: On the other hand, one cannot exclude the possibility that, if most of 
1544: the energy produced by the central engine escapes through the narrow jet 
1545: channel, giving rise to a GRB, then there may not be much energy left to 
1546: explode the rest of the star. 
1547: If this happens, then, eventually, almost all of the stellar material 
1548: falls into a black hole. This would include all the $^{56}$Ni that might 
1549: have been produced, leaving no observable supernova signature 
1550: (e.g., GRB060614).
1551: 
1552: 
1553: 
1554: %*********************************************************************
1555: 
1556: \subsection{Pulsar Kicks}
1557: \label{subsec-pulsar-kicks}  
1558: 
1559: Since most of the extracted rotational energy of the neutron star 
1560: travels vertically through the two oppositely directed channels, 
1561: a significant amount of linear momentum is also transported up and 
1562: down and hence a reactive force is exerted on the NS from both sides
1563: Even a small imbalance in the reactive magnetic force may impact 
1564: a sizable overall momentum to the~NS. For example, taking the total 
1565: initial rotational energy of the NS to be $E_{\rm rot}=5\cdot 10^{52}$~erg, 
1566: the momentum transported out in each direction is $P=E_{\rm rot}/2c \sim 
1567: 10^{42}$~cgs (it will be even larger if the propagation speed is less 
1568: than~$c$). Thus, just a 10\% imbalance may result in the NS terminal 
1569: velocity of order of $v_{\rm term}\simeq 0.1 P/M_{\rm NS}\sim 300$~km/sec.
1570: 
1571: 
1572: 
1573: %**************************************************************************
1574: 
1575: 
1576: \subsection{Suggestions for Numerical Simulations}
1577: \label{subsec-numerical}
1578: 
1579: 
1580: In this section we discuss a sequence of numerical studies of the interaction 
1581: between a magnetar's magnetosphere and its birth environment, employing a 
1582: range of plasma descriptions. Each of them will be able to address a subset 
1583: of key issues with increasing degree of realism. 
1584: For example, non-relativistic axisymmetric MHD simulations can address 
1585: the following questions:
1586: What basic magnetic configuration results when the conditions
1587: we describe are set up? What is the overall magnetic field structure? 
1588: When do magnetic towers form? How does the tower shape change as it 
1589: expands into lower-density regions?  
1590: How strongly is the magnetic field concentrated towards the axis?  
1591: How does the Poynting flux depend on radius and height? What is 
1592: the effect of the neutrino- or magneto-centrifugally-driven winds? 
1593: To what degree does the cocoon help collimate and stabilize the tower? 
1594: How rapidly do the cocoon walls spread laterally?  Can the cocoon expansion 
1595: result in the disruption of the star?  
1596: In addition, some of the physical processes described in this paper may 
1597: be relevant to other astrophysical systems, including non-relativistic 
1598: central objects ({\it e.g.}, planetary nebulae\cite{blackman-2001,matt-2006}). 
1599: For this reason, non-relativistic MHD simulations are of interest in 
1600: themselves, as well as a first step toward fully-relativistic MHD. 
1601: For example, recent non-relativistic MHD simulations indicate that 
1602: the magnetic-tower mechanism can operate successfully in a variety 
1603: of astrophysical environments\cite{romanova-2004,kato-2004a,kato-2004b,
1604: nakamura-2006,matt-2006}, including collapsing massive stars\cite
1605: {burrows-2007}.
1606: 
1607: 
1608: Eventually, however, one will have to consider relativistic effects 
1609: outside of the magnetar light cylinder or in the inner part of the black 
1610: hole's accretion disk. This can be investigated using the relativistic 
1611: force-free degenerate electrodynamics (FFDE) approach, valid in the case 
1612: of a highly magnetized plasma with negligible pressure and inertia. 
1613: Time-dependent force-free codes have recently been successfully used 
1614: to study pulsar magnetospheres\cite{komissarov-2006,mckinney-2006, 
1615: spitkovsky-2006}. In the pulsar-in-a-cavity context, the cavity wall 
1616: may be represented by a rigid conducting outer boundary. While this 
1617: case may not be directly relevant to the realistic physical environment, 
1618: some basic aspects of a bounded rotating magnetosphere may be understood 
1619: using this description. Furthermore, the full magnetar-in-a-star problem 
1620: can be investigated by a hybrid simulation employing a relativistic 
1621: force-free code inside the cavity and a relativistic hydrodynamic 
1622: simulation outside.
1623: 
1624: Finally, relativistic-MHD simulations will be able to address questions 
1625: fundamental to the application of magnetar-driven magnetic towers to GRBs,  
1626: including the beaming angle and angular distribution of energy flux and 
1627: the growth of instabilities in the relativistic outflow.
1628: There now exist several relativistic MHD codes\cite{koide-1999, 
1629: gammie-2003,delzanna-2003,devilliers-2003,anninos-2005,komissarov-2005,
1630: nishikawa-2005,shibata-2006} that can be used for this problem. 
1631: Of interest would be a set of simulations with a range of plasma~$\beta$. 
1632: The low-$\beta$ simulations should match on to the FFDE case, at least 
1633: qualitatively. Once these simulations are analyzed and the basic physical 
1634: processes elucidated, $\beta$ can be gradually increased enabling an 
1635: understanding of how plasma inertia and pressure affect the dynamics 
1636: of the magnetosphere expansion and collimation.
1637: 
1638: The basic scenario described in this paper can initially be explored 
1639: with axisymmetric simulations. However, in order to assess the role 
1640: of non-axisymmetric instabilities, such as the Rayleigh--Taylor, 
1641: Kelvin--Helmholtz, and kink instabilities (see Sec.~\ref{subsec-stability}),
1642: three-dimensional simulations will eventually be necessary. One of the goals
1643: of such an investigation will be to estimate and the degree of mixing of 
1644: the envelope baryonic material into the magnetosphere. In addition, one
1645: also would like to study the development and interaction of the MRI and 
1646: the Parker instability. 
1647: 
1648: The operation of the Parker instability, leading to the development of 
1649: a highly-magnetized low-density corona, may be strongly influenced by 
1650: neutrino cooling. Thus, a realistic treatment of the neutrino heating 
1651: and cooling processes (as it has been done, {\it e.g.}, by Burrows 
1652: {\it et~al.}\cite{burrows-2007}) is an essential physical ingredient
1653: of the overall problem.
1654: 
1655: 
1656: In summary, numerical simulations of the full problem, including a detailed 
1657: description of the central engine with relevant microphysical processes and 
1658: neutrino transport, are desirable for a comprehensive understanding of the 
1659: formation and evolution of a millisecond-magnetar-driven magnetic tower 
1660: inside a collapsing star. We believe that such simulations will very soon
1661: become technically feasible.
1662: 
1663: 
1664: 
1665: 
1666: %*********************************************************************
1667: 
1668: \section{Conclusions}
1669: \label{sec-conclusions}
1670: 
1671: The core collapse of a massive rotating star may result in two distinct
1672: outcomes, both plausible candidates for the GRB central engine.
1673: The first one is a stellar-mass black hole with an accretion disk. 
1674: The second is a neutron star. In this second scenario, the young 
1675: neutron star formed as a result of core collapse has to be a millisecond 
1676: magnetar in order to be relevant for~GRBs.
1677: 
1678: In this paper we mostly focus on the millisecond-magnetar scenario, 
1679: although many features of our model are also applicable to the black-hole 
1680: case, which we have considered in our previous paper (Ref.~\cite{um-2006a}). 
1681: Of particular interest to us is the interaction between the 
1682: rapidly-rotating magnetar's magnetosphere and the surrounding infalling 
1683: stellar envelope. We argue that the stellar material provides a confining 
1684: (ram) pressure that has a strong effect on both the size and the shape of 
1685: the magnetosphere. Namely, it can channel the highly-magnetized outflow 
1686: originating from the proto-neutron star into two collimated magnetic 
1687: towers.
1688: 
1689: More specifically, we suggest that the stalled bounce shock --- 
1690: a common feature in models of core-collapse supernovae --- plays 
1691: a role of a cavity that confines the magnetosphere. 
1692: The cavity's radius, determined by the balance between the pressure 
1693: of the hot neutrino-heated gas and the ram pressure of the infalling 
1694: material, stays quasi-stationary at $R_0\simeq 200$~km during the first 
1695: few hundreds of milliseconds after the bounce.
1696: To get a qualitative physical feeling for what happens to the magnetar 
1697: magnetosphere during this stage, we introduce an idealized fundamental-physics 
1698: problem that we call the {\it Pulsar-in-a-Cavity} problem\cite{um-2006b}. 
1699: A large part of our paper (Sec.~\ref{sec-pulsar-cavity}) is devoted to 
1700: investigating this problem.
1701: For simplicity, we consider it under the force-free assumption.
1702: We show that if the radius of the cavity is larger than the pulsar 
1703: light-cylinder radius, the magnetic field inside the cavity continuously 
1704: winds up. Then, the toroidal field strength and hence the magnetic spin-down 
1705: luminosity of the pulsar increase, roughly linearly with time. The magnetic 
1706: energy in the cavity grows quadratically with time. We then estimate that 
1707: in the context of a millisecond magnetar inside a collapsing star the magnetic 
1708: field becomes dynamically important after a few hundred turns. 
1709: This leads to a subsequent revival of the stalled shock and may result in 
1710: a successful magnetically-driven explosion. As long as the expansion of 
1711: the cavity is non-relativistic, the toroidal magnetic field inside it 
1712: remains larger than the poloidal magnetic and electric fields. As a result, 
1713: the hoop-stress collimates the Poynting-flux-dominated outflow into two 
1714: oppositely-directed vertical channels. 
1715: 
1716: We suggest that these magnetic outflows (either in the millisecond-magnetar 
1717: or the accreting black hole scenario) should be described in terms of the 
1718: magnetic tower concept, introduced by Lynden-Bell in the AGN 
1719: context\cite{lb-1996}. 
1720: Correspondingly, we investigate the propagation of a magnetic tower inside 
1721: a star (see also Ref.~\cite{um-2006a}).
1722: In our model, we modify Lynden-Bell's picture by considering that the tower 
1723: expansion is supersonic with respect to the unperturbed stellar gas. 
1724: We envision the growing magnetic tower acting as a piston that drives 
1725: a strong shock through the star. The hot shocked stellar material between 
1726: the shock and the tower forms a high-pressure cocoon that envelopes the 
1727: tower and provides the collimating pressure. In other words, the tower 
1728: in our model is confined not by the pressure of the background stellar 
1729: material, but by its inertia; the strong shock and the cocoon act as 
1730: mediators that convert the inertial support into the pressure support 
1731: ultimately acting on the tower.
1732: The entire configuration grows vertically with time and eventually reaches
1733: the star's surface, thereby providing a narrow baryon-clean channel for
1734: the Poynting-flux dominated jet, surrounded by a less-collimated hot
1735: cocoon outflow. 
1736: 
1737: 
1738: Finally, we discuss the astrophysical implications of our model for 
1739: GRBs and core-collapse supernovae, such as $^{56}$Ni production and 
1740: pulsar kicks. In addition, we discuss the role of MHD instabilities,
1741: most notably, the kink, in our scenario. We also assess the prospects 
1742: for magnetic reconnection and find that it should be strongly inhibited 
1743: in the central parts of the collapsing star owing to the high plasma 
1744: collisionality there. Finally, we outline a set of numerical studies 
1745: that we believe need to be done.
1746: 
1747: 
1748: %-------------------------------------------------------------------
1749: 
1750: \acknowledgments
1751: 
1752: We would like to thank Profs.\ Jerry Ostriker and Russell Kulsrud for 
1753: stimulating discussions and encouragement. We are also grateful to 
1754: J.~Arons, A.~Beloborodov, E.~Blackman, P.~Goldreich, J.~Goodman, 
1755: S.~Komissarov, A.~K\"onigl, H.~Li, M. Lyutikov, J.~McKinney, C. Thompson, 
1756: T.~Thompson, A. Spitkovsky, H.~Spruit, J.~C. Wheeler, and E.~Zweibel 
1757: for useful comments and suggestions. 
1758: 
1759: DAU's research has been supported by the National Science Foundation 
1760: under Grant~PHY-0215581 (PFC: Center for Magnetic Self-Organization 
1761: in Laboratory and Astrophysical Plasmas).
1762: AIM acknowledges support from the Keck Fellowship at the Institute 
1763: for Advanced Study. 
1764: 
1765: 
1766: %********************************************************************
1767: 
1768: \newpage
1769: 
1770: %\section*{REFERENCES}
1771: %\parindent 0 pt
1772: 
1773: \begin{thebibliography}{99}
1774: 
1775: \bibitem{woosley-1993}
1776:  S.~E.~Woosley, Astrophys. J., {\bf 405}, 273 (1993)
1777: 
1778: \bibitem{paczynski-1998}
1779:  B.~Paczynski, Astrophys. J., {\bf 494}, L45 (1998)
1780: 
1781: \bibitem{macfadyen-woosley-1999}
1782: A.~I.~MacFadyen \&  S.~E.~Woosley, Astrophys. J., {\bf 524}, 262 (1999)
1783: 
1784: \bibitem{proga-2003}
1785:  D.~Proga,,  A.~I.~MacFadyen, P.~J.~Armitage, \&  M.~C.~Begelman, 
1786: Astrophys. J., {\bf 599}, L5 (2003)
1787: 
1788: \bibitem{thompson-1994}
1789: C.~Thompson, Mon. Not. R. Astron. Soc., {\bf 270}, 480 (1994)
1790: 
1791: \bibitem{meszaros-rees-1997}
1792: P.~Meszaros \& M.~J.~Rees, Astrophys. J., {\bf 482}, L29 (1997)
1793: 
1794: \bibitem{lee-2000}
1795: H.~K.~Lee, R.~A.~M.~J.~Wijers, \& G.~E.~Brown, Phys. Reports, 
1796: {\bf 325}, 83 (2000)
1797:  
1798: \bibitem{vlahakis-konigl-2001}
1799: N.~Vlahakis, \& A.~K{\"o}nigl, Astrophys. J., {\bf 563}, L129 (2001)
1800: 
1801: \bibitem{vanPutten-ostriker-2001}
1802: M.~H.~P.~M.~van~Putten \& E.~C.~Ostriker, Astrophys. J., {\bf 552}, L31 (2001)
1803: 
1804: \bibitem{drenkhahn-spruit-2002}
1805: G.~Drenkhahn, \& H.~Spruit, Astron. \& Astrophys., {\bf 391}, 1141 (2002)
1806: 
1807: %\bibitem{lyutikov-blandford-2002}
1808: %Lyutikov, M. \& Blandford, R., Proceedings of the Workshop on 
1809: %{\it Beaming and Jets in Gamma Ray Bursts (NBSI)}, ed.~R.~Ouyed, 
1810: %Copenhagen, (2002) 
1811: %preprint (astro-ph/0210671)
1812: 
1813: \bibitem{lyutikov-blandford-2003}
1814: M.~Lyutikov \& R.~Blandford, ``Electromagnetic explosions in 
1815: Gamma-Ray Bursts'', Bull. Amer. Astron. Soc., {\bf 35}, 622 (2003);
1816: e-print (astro-ph/0312347) (2003)
1817: 
1818: \bibitem{vanPutten-levinson-2003}
1819: M.~H.~P.~M.~van Putten \& A.~Levinson, Astrophys. J., {\bf 584}, 937 (2003)
1820: 
1821: \bibitem{lyutikov-2006}
1822: M.~Lyutikov, New J.\ Phys., {\bf 8}, 119 (2006)
1823: %preprint (astro-ph/0512342)
1824: 
1825: \bibitem{mckinney-2005}
1826: J.~C.~McKinney, Astrophys. J., {\bf 630}, L5 (2005)
1827: 
1828: \bibitem{um-2006a}
1829: D.~A.~Uzdensky \& A.~I.~MacFadyen, Astrophys. J., {\bf 647}, 1192 (2006)
1830: 
1831: \bibitem{giannios-spruit-2005}
1832: D.~Giannios \& H.~Spruit, Astron. \& Astrophys., {\bf 430}, 1 (2005)
1833: 
1834: \bibitem{giannios-spruit-2006}
1835: D.~Giannios \& H.~Spruit, Astron. \& Astrophys., {\bf 450}, 887 (2006)
1836: 
1837: \bibitem{giannios-spruit-2007}
1838: D.~Giannios, \& H.~Spruit, ``Spectral and timing properties of 
1839: a dissipative GRB photosphere'', submitted to Astron. \& Astrophys. (2007);
1840: e-print (astro-ph/0611385)
1841: 
1842: \bibitem{akiyama-2003}
1843: S.~Akiyama, J.~C.~Wheeler, D.~L.~Meier, \& I.~Lichtenstadt, 
1844: Astrophys. J., {\bf 584}, 954 (2003)
1845: 
1846: \bibitem{ardeljan-2005}
1847: N.~V.~Ardeljan, G.~S.~Bisnovatyi-Kogan, \& S.~G.~Moiseenko, 
1848: Mon. Not. R. Astron. Soc., {\bf 359}, 333 (2005)
1849: 
1850: \bibitem{blackman-2006}
1851: E.~G.~Blackman, J.~T.~Nordhaus, \& J.~H.~Thomas, New Astron., 
1852: {\bf 11}, 452 (2006)
1853: 
1854: \bibitem{duncan-thompson-1992}
1855: R.~C.~Duncan, \& C.~Thompson, Astrophys. J., {\bf 392}, L9 (1992)
1856: 
1857: \bibitem{thompson-duncan-1993}
1858: C.~Thompson \& R.~C.~Duncan, Astrophys. J., {\bf 408}, 194 (1993)
1859: 
1860: \bibitem{og-1971}
1861: J.~P.~Ostriker \& J.~E.~Gunn,  Astrophys. J., {\bf 164}, L95 (1971)
1862: 
1863: \bibitem{usov-1992}
1864: V.~V.~Usov, Nature, {\bf 357}, 472 (1992)
1865:  
1866: \bibitem{yi-blackman-1998}
1867: I.~Yi \& E.~G.~Blackman, Astrophys. J., {\bf 494}, L163 (1998)
1868: 	
1869: \bibitem{nakamura-1998}
1870: T.~Nakamura, Progress of Theoretical Physics, {\bf 100}, 921 (1998)
1871: 
1872: \bibitem{spruit-1999}
1873: H.~Spruit, Astron. \& Astrophys., {\bf 341}, L1 (1999)
1874: 
1875: \bibitem{ruderman-2000}
1876: M.~A.~Ruderman, L.~Tao, \& W.~Kluzniak, Astrophys. J., {\bf 542}, 243 (2000)
1877: 
1878: \bibitem{wheeler-2000}
1879: J.~C.~Wheeler, I.~Yi, P.~H{\"o}flich, \& L.~Wang, Astrophys.~J., 
1880: {\bf 537}, 810 (2000)
1881: 
1882: \bibitem{wheeler-2002}
1883: J.~C.~Wheeler, D.~L.~Meier, \& J.~R.~Wilson, Astrophys. J., 
1884: {\bf 568}, 807 (2002)
1885: 
1886: \bibitem{um-2006b}
1887: D.~A.~Uzdensky \& A.~I.~MacFadyen, ``Magnetar-Driven Magnetic Tower 
1888: as a Model for Gamma-Ray Bursts and Asymmetric Supernovae'', submitted 
1889: to Astrophys.~J. (2006);
1890: e-print (astro-ph/0609047)
1891: 
1892: \bibitem{thompson-2004}
1893: T.~A.~Thompson, P.~Chang, \& E.~Quataert, Astrophys. J., {\bf 611}, 380 (2004)
1894: 
1895: \bibitem{leBlanc-wilson-1970}
1896: LeBlanc, J.~M. \& Wilson, J.~R., Astrophys. J., {\bf 161}, 541 (1970)
1897: 
1898: \bibitem{meier-1976}
1899: D.~L.~Meier, R.~I.~Epstein, W.~D.~Arnett, \& D.~N.~Schramm, 
1900: Astrophys. J., {\bf 204}, 869 (1976)
1901: 
1902: \bibitem{shibata-2006}
1903: M.~Shibata, Y.~T.~Liu, S.~L.~Shapiro, \& B.~C.~Stephens,
1904: Phys.~Rev.~D, {\bf 74}, 104026 (2006)
1905: 
1906: \bibitem{bucciantini-2006}
1907: N.~Bucciantini, T.~A.~Thompson, J.~Arons, E.~Quataert, \& L.~Del~Zanna, 
1908: Mon. Not. R. Astron. Soc., {\bf 368}, 1717 (2006)
1909: 
1910: \bibitem{metzger-2007}
1911: B.~D.~Metzger, T.~A.~Thompson, \& E.~Quataert, 
1912: ``Proto-Neutron Star Winds with Magnetic Fields and Rotation'',
1913: accepted to Astrophys.~J. (2007); e-print (astro-ph/0608682)
1914: 
1915: \bibitem{burrows-2007}
1916: A.~Burrows, L.~Dessart, E.~Livne, C.~D.~Ott, \& J.~Murphy, 
1917: ``Simulations of Magnetically-Driven Supernova and Hypernova Explosions 
1918: in the Context of Rapid Rotation'', submitted to the Astrophys.~J. (2007);
1919: e-print (astro-ph/0702539) (2007)
1920: 
1921: \bibitem{arons-2003}
1922: J.~Arons, Astrophys.~J., {\bf 589}, 871 (2003)
1923: 
1924: \bibitem{lb-1996}
1925: D.~Lynden-Bell, Mon. Not. R. Astron. Soc., {\bf 279}, 389 (1996)
1926: 
1927: \bibitem{lb-2003}
1928: D.~Lynden-Bell, Mon. Not. R. Astron. Soc., {\bf 341}, 1360 (2003)
1929: 
1930: \bibitem{woosley-weaver-1986}
1931: S.~E.~Woosley \& T.~A.~Weaver, Annual Rev. Astron. \& Astrophys., 
1932: {\bf 24}, 205 (1986)
1933: 
1934: \bibitem{bethe-1990}
1935: H.~A.~Bethe, Rev.~Mod.~Phys., {\bf 62}, 801 (1990)
1936: 
1937: \bibitem{gj-1969}
1938: P.~Goldreich \& W.~H.~Julian, Astrophys. J., {\bf 157}, 869 (1969)
1939: 
1940: \bibitem{kardashev-1970}
1941: N.~S.~Kardashev, Sov. Astron., {\bf 14}, 375 (1970)
1942: 
1943: \bibitem{illarionov-sunyaev-1975}
1944: A.~F.~Illarionov \& R.~A.~Sunyaev, Astron. \& Astrophys., {\bf 39}, 185. (1975)
1945: 
1946: \bibitem{matt-2006}
1947: S.~Matt, A.~Frank, \& E.~G.~Blackman, Astrophys.~J., {\bf 647}, L45 (2006)
1948: 
1949: \bibitem{michel-1973}
1950: F.~C.~Michel, Astrophys. J., {\bf 180}, L133 (1973)
1951: 
1952: \bibitem{macfadyen-2001}
1953: A.~I.~MacFadyen, S.~E.~Woosley, \& A.~Heger, 
1954: Astrophys. J., {\bf 550}, 410 (2001)
1955: 
1956: \bibitem{zhang-2003}
1957: W.~Zhang, S.~E.~Woosley, \& A.~I.~MacFadyen, 
1958: Astrophys. J., {\bf 586}, 356 (2003)
1959: 
1960: \bibitem{lbb-1994}
1961: D.~Lynden-Bell \& C.~Boily, Mon. Not. R. Astron. Soc., {\bf 267}, 146 (1994)
1962: 
1963: \bibitem{ukl-2002}
1964: D.~A.~Uzdensky, A.~K\"{o}nigl, \& C.~Litwin, 
1965: Astrophys. J., {\bf 565}, 1191 (2002)
1966: 
1967: \bibitem{lrbk-1995}
1968: R.~V.~E.~Lovelace, M.~M.~Romanova, \& G.~S.~Bisnovatyi-Kogan, 
1969: Mon. Not. R. Astron. Soc., {\bf 275}, 244 (1995)
1970: 
1971: \bibitem{uzdensky-2005}
1972: D.~A.~Uzdensky, Astrophys. J., {\bf 620}, 889 (2005)
1973: 
1974: \bibitem{li-2001}
1975: H.~Li, R.~V.~E.~Lovelace, J.~M.~Finn, \& S.~A.~Colgate, 
1976: Astrophys. J., {\bf 561}, 915 (2001)
1977: 
1978: \bibitem{kato-2004a}
1979: Y.~Kato, M.~R.~Hayashi, \& R.~Matsumoto, Astrophys. J., {\bf 600}, 338 (2004)
1980: 
1981: \bibitem{kato-2004b}
1982: Y.~Kato, S.~Mineshige, \& K.~Shibata, Astrophys.~J., {\bf 605}, 307 (2004)
1983: 
1984: \bibitem{kato-2006}
1985: Y.~Kato, ``Magnetic-Tower Jet Solution for Launching Astrophysical Jets'', 
1986: accepted to Astrophys.\ \& Space Sci.\ (2006)
1987: 
1988: \bibitem{nakamura-2006}
1989: M.~Nakamura, H.~Li, \& S.~Li, Astrophys. J., {\bf 652}, 1059 (2006)
1990: 
1991: \bibitem{nakamura-2007}
1992: M.~Nakamura, H.~Li, \& S.~Li, Astrophys. J., {\bf 656}, 721 (2007)
1993: 
1994: \bibitem{ciardi-2007}
1995:  A.~Ciardi, S.~V.~Lebedev, A.~Frank, {\it et~al.}, 
1996: ``The evolution of magnetic tower jets in the laboratory'', 
1997: accepted to Phys. Plasmas (2007); e-print (astro-ph/0611441)
1998: 
1999: %Ciardi, A., Lebedev, S.~V., Frank, A.,  Blackman, E.~G., Chittenden, J.~P., 
2000: %Jennings, C.~J., Ampleford, D.~J., Bland, S.~N., Bott, S.~C., Rapley, J., 
2001: %Hall, G.~N., Suzuki-Vidal, F.~A., Marocchino, A., Lery, T., \& Stehle, C.\ 
2002: %2007, accepted to Phys. Plasmas; preprint (astro-ph/0611441)
2003: 
2004: \bibitem{hsu-bellan-2002}
2005: S.~Hsu \& P.~M.~Bellan, Mon. Not. R. Astron. Soc., {\bf 334}, 257 (2002)
2006: 
2007: \bibitem{lebedev-2005}
2008: S.~V.~Lebedev, A.~Ciardi, D.~J.~Ampleford, {\it et~al.}, 
2009: Mon. Not. R. Astron. Soc., {\bf 361}, 97 (2005)
2010: 
2011: \bibitem{nakamura-meier-2004}
2012: M.~Nakamura, \& D.~L.~Meier, Astrophys. J., {\bf 617}, 123 (2004)
2013: 
2014: \bibitem{konigl-choudhuri-1985}
2015: A.~K{\"o}nigl, \& A.~R.~Choudhuri, Astrophys. J., {\bf 289}, 173 (1985)
2016: 
2017: \bibitem{gruzinov-1999}
2018: A.~Gruzinov, ``Flares on the Black Holes'', submitted to Astrophys. J. (1999) 
2019: %preprint (astro-ph/9908101) (1999)
2020: 
2021: \bibitem{tomimatsu-2001}
2022: A.~Tomimatsu, T.~Matsuoka, \& M.~Takahashi, Phys.~Rev.~D.,
2023: {\bf 64}, 123003 (2001)
2024: 
2025: \bibitem{eichler-1993}
2026: D.~Eichler, Astrophys. J., {\bf 419}, 111 (1993)
2027: 
2028: \bibitem{begelman-1998}
2029: M.~C.~Begelman, Astrophys. J., {\bf 493}, 291 (1998)
2030: 
2031: \bibitem{goodson-1999}
2032: A.~P.~Goodson, K.-H.~B\"ohm, \& R.~M.~Winglee, 
2033: Astrophys.~J., {\bf 524}, 142 (1999)
2034: 
2035: \bibitem{uzdensky-2004}
2036: D.~A.~Uzdensky, Astrophys. \& Space Sci., {\bf 292}, 573 (2004)
2037: 
2038: \bibitem{biskamp-1986}
2039: D.~Biskamp, Phys. Fluids, {\bf 29}, 1520 (1986)
2040: 
2041: \bibitem{uzdensky-kulsrud-2000}
2042: D.~A.~Uzdensky \& R.~M.~Kulsrud, Phys. Plasmas, {\bf 7}, 4018 (2000)
2043: 
2044: \bibitem{kulsrud-2001}
2045: R.~M.~Kulsrud, Earth, Planets and Space, {\bf 53}, 417 (2001)
2046: 
2047: \bibitem{malyshkin-2005}
2048: L.~M.~Malyshkin, T.~Linde, \& R.~M.~Kulsrud, 
2049: Phys. Plasmas, {\bf 12}, 102902 (2005)
2050: 
2051: \bibitem{ji-1999}
2052: H.~Ji, M.~Yamada, S.~Hsu, R.~Kulsrud, T.~Carter, \& S.~Zaharia,
2053: Phys. Plasmas, {\bf 6}, 1743 (1999)
2054: 
2055: \bibitem{sweet-1958}
2056:  P.~A.~Sweet, in IAU Symp.~6, Electromagnetic Phenomena in Cosmical 
2057: Physics, ed.\ B.~Lehnert, (Cambridge: Cambridge Univ. Press), 123 (1958)
2058: 
2059: \bibitem{parker-1957}
2060: E.~N.~Parker, J.~Geophys.~Res., {\bf 62}, 509 (1957)
2061: 
2062: \bibitem{petschek-1964}
2063: H.~E.~Petschek, AAS-NASA Symposium on Solar Flares, 
2064: (National Aeronautics and Space Administration, Washington, DC, 1964), 
2065: NASA SP50, 425 (1964)
2066: 
2067: \bibitem{lyutikov-blackman-2001}
2068: M.~Lyutikov, \& E.~G.~Blackman, Mon. Not. R. Astron. Soc., 
2069: {\bf 321}, 177 (2001)
2070: 
2071: %\bibitem{soderberg-2005}
2072: %Soderberg, A.~M., et  al., Astrophys. J., {\bf 627}, 877 (2005)
2073: 
2074: \bibitem{soderberg-2006}
2075: A.~Soderberg, ``Gamma-Ray Bursts in the Swift Era, Sixteenth
2076: Maryland Astrophysics Conference'' Eds.\ Holt, S.~S., Gehrels, N.\ 
2077: \& Nousek, J. A., AIP Conference Proceedings, {\bf 838}, 380 (2006)
2078: 
2079: \bibitem{kaneko-2007}
2080: Y.~Kaneko, E.~Ramirez-Ruiz, J.~Granot, {\it et~al.}, 
2081: Astrophys.~J., {\bf 654}, 385 (2007)
2082: 
2083: \bibitem{fynbo-2006}
2084: J.~P.~U.~Fynbo, D.~Watson, C.~C.~Th\"one, {\it et~al.}, 
2085: Nature, {\bf 444}, 1047 (2006)
2086: 
2087: \bibitem{dellaValle-2006}
2088: M.~Della Valle, G.~Chincarini, N.~Panagia, {\it et~al.}, 
2089: Nature, {\bf 444}, 1050 (2006)
2090: 
2091: \bibitem{gal-yam-2006}
2092: A.~Gal-Yam, D.~B.~Fox, P.~A.~Price, {\it et~al.}, 
2093: Nature, {\bf 444}, 1053 (2006)
2094: 
2095: \bibitem{macfadyen-2003}
2096: A.~I.~MacFadyen, ``Gamma-ray Burst and Afterglow Astronomy: 2001'' 
2097: AIP Conference Proceedings, {\bf 662}, 202 (2003)
2098: 
2099: \bibitem{blackman-2001}
2100: E.~G.~Blackman, A.~Frank, \& C.~Welch, Astrophys. J., {\bf 546}, 288 (2001)
2101: 
2102: \bibitem{romanova-2004}
2103: M.~M.~Romanova, G.~V.~Ustyugova, A.~V.~Koldoba, \& R.~V.~E.~Lovelace, 
2104: 2004, Astrophys. J., {\bf 616}, L151 (2004)
2105: 
2106: \bibitem{komissarov-2006}
2107: S.~Komissarov, Mon. Not. R. Astron. Soc., {\bf 367}, 19 (2006)
2108: 
2109: \bibitem{mckinney-2006}
2110: J.~C.~McKinney, Mon. Not. R. Astron. Soc., {\bf 368}, L30 (2006)
2111: 
2112: \bibitem{spitkovsky-2006}
2113: A.~Spitkovsky, Astrophys.~J., {\bf 648}, L51 (2006)
2114: 
2115: \bibitem{koide-1999}
2116: S.~Koide, K.~Shibata, \& T.~Kudoh, Astrophys.~J., {\bf 522}, 727 (1999)
2117: 
2118: \bibitem{gammie-2003}
2119: C.~F.~Gammie, J.~C.~McKinney, \& G.~T\'oth, 
2120: Astrophys. J., {\bf 589}, 444 (2003)
2121: 
2122: \bibitem{delzanna-2003}
2123: L.~Del~Zanna, N.~Bucciantini, \& P.~Londrillo, 
2124: Astron. \& Astrophys., {\bf 400}, 397 (2003)
2125: 
2126: \bibitem{devilliers-2003}
2127: J.-P.~De~Villiers, J.~F.~Hawley, \& J.~H.~Krolik, Astrophys. J., 
2128: {\bf 599}, 1238 (2003)
2129: 
2130: \bibitem{anninos-2005}
2131: P.~Anninos, P.~C.~Fragile, \& J.~D.~Salmonson,
2132: Astrophys. J., {\bf 635}, 723 (2005)
2133: 
2134: \bibitem{komissarov-2005}
2135: S.~S.~Komissarov, Mon. Not. R. Astron. Soc., {\bf 359}, 801 (2005)
2136: 
2137: \bibitem{nishikawa-2005}
2138: K.-I.~Nishikawa, G.~Richardson, S.~Koide, K.~Shibata, T.~Kudoh,
2139: P.~Hardee, \& G.~J.~Fishman, Astrophys. J., {\bf 625}, 60 (2005)
2140: 
2141: 
2142: \end{thebibliography}
2143: 
2144: 
2145: %*****************************************************************
2146: 
2147: 
2148: 
2149: %****************************************************************
2150: 
2151: \cleardoublepage
2152: 
2153: \hoffset=2in
2154: \voffset=1.5in
2155: 
2156: 
2157: \begin{figure} [h]
2158: \centerline
2159: {\psfig{file=stalled-shock.eps,width=5in}}
2160: \caption{Stalled shock phase of core-collapse explosion.}
2161: \label{fig-stalled-shock}
2162: \end{figure}
2163: 
2164: %-------------------------------------------------------
2165: 
2166: \cleardoublepage
2167: 
2168: \begin{figure} [h]
2169: \centerline
2170: {\psfig{file=scenario.eps,width=5in}}
2171: \caption{Schematic time evolution of the main three forces responsible 
2172: for the stalled-shock force balance.}
2173: \label{fig-scenario}
2174: \end{figure}
2175: 
2176: 
2177: %-------------------------------------------------------
2178: 
2179: \cleardoublepage
2180: 
2181: \begin{figure} [h]
2182: \centerline
2183: {\psfig{file=magnetar-2.eps,width=5in}}
2184: \caption{Aligned pulsar inside an infinitely-conducting 
2185: spherical cavity of radius~$R_0$. The vertical dashed lines represent 
2186: the pulsar's light cylinder of radius $R_{\rm LC}<R_0$. 
2187: After a time of order the
2188: light-crossing time~$R_0/c$, the poloidal field lines outside 
2189: the light cylinder expand somewhat but still remain confined within 
2190: the cavity. Because the toroidal magnetic field has to vanish 
2191: at the equatorial midplane due to reflection symmetry, the field 
2192: lines there cannot corotate with the star, $\Omega_B<c/R_B<\Omega_*$.
2193: As a result, differential rotation is established in both hemispheres,
2194: $\Delta\Omega=\Omega_*-\Omega_B\simeq\Omega_*$ (for $R_B\gg R_{\rm LC}$),
2195: which leads to continuous generation of toroidal magnetic flux.}
2196: \label{fig-magnetar-2}
2197: \end{figure}
2198: 
2199: 
2200: %-------------------------------------------------------
2201: 
2202: \cleardoublepage
2203: 
2204: \begin{figure} [h]
2205: \centerline
2206: {\psfig{file=strip.eps,width=5in}}
2207: \caption{At late times, the poloidal magnetic field is pressed against 
2208: the wall by the centrifugal force of the rotating massive equatorial sheet.}
2209: \label{fig-strip}
2210: \end{figure}
2211: 
2212: %-------------------------------------------------------
2213: 
2214: \cleardoublepage
2215: 
2216: \begin{figure} [h]
2217: \centerline
2218: {\psfig{file=lb-tower.eps,width=5in}}
2219: \caption{Development of a magnetic tower in Lynden-Bell's (1996) model.
2220: {\it Reproduced from Ref.~\cite{um-2006a} by permission of the AAS.}}
2221: \label{fig-LB-tower}
2222: \end{figure}
2223: 
2224: 
2225: %-------------------------------------------------------
2226: 
2227: \cleardoublepage
2228: 
2229: \begin{figure} [h]
2230: \centerline
2231: {\psfig{file=star.eps,width=5in}}
2232: \caption{Main components of a magnetic tower inside. 
2233: The tower grows rapidly and drives a strong shock through the star. 
2234: The shocked stellar gas behind the shock forms a hot cocoon whose 
2235: high pressure confines the tower.
2236: {\it Reproduced from Ref.~\cite{um-2006a} by permission of the AAS.}}
2237: \label{fig-star}
2238: \end{figure}
2239: 
2240: %-------------------------------------------------------
2241: 
2242: \cleardoublepage
2243: 
2244: \begin{figure} [h]
2245: \centerline
2246: {\psfig{file=train.eps,width=5in}}
2247: \caption{Magnetic tower may have a substructure represented
2248: by a train of many spheromak-like plasmoids. This situation may 
2249: arise as a result of spatial and temporal intermittency at the 
2250: base of the outflow and/or due to instabilities and reconnection
2251: in the tower.}
2252: \label{fig-train}
2253: \end{figure}
2254: 
2255: \cleardoublepage
2256: 
2257: 
2258: 
2259: %****************************************************************
2260: 
2261: 
2262: \end{document}
2263: 
2264: