1: \documentclass[12pt,preprint]{aastex}
2: \usepackage{graphicx}
3: \begin{document}
4: \title{On the corona of magnetars}
5:
6: \author{Yury Lyubarsky and David Eichler}
7:
8: \affil{Physics Department, Ben-Gurion University of the Negev,
9: Beer-Sheva, Israel}
10:
11:
12: \begin{abstract}
13: Slow dissipation of non-potential magnetic fields in the
14: magnetosphere of the magnetar is assumed to accelerate particles
15: to hundreds MeV along the magnetic field lines. We consider
16: interaction of fast particles with the surface of the magnetar. We
17: argue that the collisionless dissipation does not work in the
18: atmosphere of the neutron star because the two-stream instability
19: is stabilized by the inhomogeneity of the atmosphere. Rather, the
20: dominant dissipation mechanism is collisional Landau level
21: excitations followed by pair production via the deexcitation
22: gamma-rays ultimately leading to electrons with the energy below
23: the Landau energy.
24: %The energy of the beam is eventually transferred to
25: %electrons with the energy something less than the Landau energy.
26: We show that, because of the effects of the superstrong magnetic
27: field, these electrons could emerge from the surface carrying most
28: of the original energy so that a hot corona arises with the
29: temperature of $1\div 2$ MeV.
30: %Hard radiation from this atmosphere
31: %populates the magnetosphere with electron-positron pairs forming a
32: %mildly relativistic corona.
33: This extended
34: corona is better suited than a thin atmosphere to
35: convert most of the primary beam energy to non-thermal radiation
36: and, as we show, most of the coronal energy release is radiated
37: away in the hard X-ray and the soft gamma-ray bands by
38: Comptonization and bremsstrahlung. The radiation spectrum is a
39: power-law with the photon index $1<\alpha<2$. The model may
40: account for the persistent hard X-ray emission discovered recently
41: from the soft gamma-ray repeaters and anomalous X-ray pulsars and
42: predicts that the radiation spectrum is extended into the MeV
43: band.
44: \end{abstract}
45:
46: \keywords{instabilities -- plasmas -- stars:magnetic field --
47: stars:neutron }
48:
49: \section{Introduction}
50: It is now commonly accepted that Soft Gamma Repeaters (SGRs) and
51: Anomalous X-ray Pulsars (AXPs) are magnetars, neutron stars with
52: extremely high magnetic fields, $B\sim 10^{14}\div 10^{15}$ G
53: (Duncan \& Thompson 1992; Paczynski 1992; Thompson \& Duncan
54: 1995). Activity of these sources is fed by the energy of this
55: magnetic field (see Woods \& Thompson 2004 for a recent review).
56: It was found recently (Kuiper et al. 2004, 2006; Mereghetti et al.
57: 2005, Molkov et al. 2005) that the persistent pulsed X-ray
58: emission from these sources has a nonthermal spectrum extending up
59: to $\gtrsim 100$ keV. The luminosity of this tail, $L\sim 10^{36}$
60: erg/s, exceeds the thermal luminosity from the star's surface. The
61: spectra are exceptionally hard with photon indices typically in
62: the range $1\div 2$ so that the luminosity peaks above 100 keV. In
63: this paper, we discuss origin of this emission.
64:
65: According to Thompson, Lyutikov \& Kulkarni (2002), magnetic field
66: in magnetar's magnetosphere is generally non-potential. Slow
67: untwisting of the magnetic field lines generates the electric
68: field; particles are accelerated in this field until they produce
69: electron-positron pairs filling the magnetosphere and providing
70: charge carriers for the electric current (Beloborodov \& Thompson
71: 2007). Pairs are formed from gamma-photons produced by cyclotron
72: scattering of the surface radiation of the neutron star on primary
73: particles. In magnetar's magnetic field, the energy of the first
74: Landau level is high,
75: \begin{equation}
76: \varepsilon_B=m_ec^2\sqrt{1+2B/B_q}\approx 3.5\sqrt{B_{15}}\,{\rm
77: MeV};\label{Landau}
78: \end{equation}
79: so that the scattered photon is immediately converted into an
80: electron-positron pair. Here $B=10^{15}B_{15}$ G is the magnetic
81: field of the star, $B_q=m_e^2c^3/\hbar e=4.4\times 10^{13}$ G. If
82: the surface emits radiation with the characteristic photon energy
83: $\varepsilon=10\varepsilon_{10}$ keV, the cyclotron scattering
84: occurs when the Lorentz factor of the electron reaches the value
85: \begin{equation}
86: \gamma=\frac
87: B{B_q}\frac{m_ec^2}{\varepsilon}=10^3\frac{B_{15}}{\varepsilon_{10}}.
88: \label{gamma}
89: \end{equation}
90: Note that the recoil effect should be taken into account in this
91: estimate (Thompson \& Beloborodov 2005). Eventually the energy is
92: deposited in pairs with the Lorentz factor something below
93: (\ref{gamma}) such that they already could not scatter the surface
94: photons resonantly. Nonresonant scattering is strongly suppressed
95: (Beloborodov \& Thompson 2007) therefore these pairs freely flow
96: along the magnetic field lines until they hit the surface of the
97: neutron star where the energy is released. The observed hard X-ray
98: persistent emission definitely comes from an optically thin,
99: tenuous plasma; this suggests that the energy is released at a
100: small depth, $\lesssim 1$ g/cm$^2$.
101:
102: It was assumed by Thompson \& Beloborodov (2005) and by
103: Beloborodov \& Thompson (2007) that the two-stream instability
104: develops in a very narrow upper layer of the atmosphere so that
105: collisionless processes are responsible for deceleration of the
106: beam and plasma heating. The observed hard radiation was
107: attributed to bremsstrahlung from this atmosphere. However, the
108: two-stream instability is quenched when the beam is spread into a
109: plateau-like distribution so that only one half of the initial
110: beam energy is released in the optically thin domain. The beam is
111: in any case stopped finally at a larger depth by collisions with
112: the background particles. So only one half of the beam energy
113: could be radiated in the hard band. Of this, one half is directed
114: towards the star where it is mostly absorbed and reradiated back
115: in the soft band. One concludes that collisionless dissipation in
116: the atmosphere converts only 1/4 of the total energy to the hard
117: radiation and could not account for the fact that most of the
118: magnetars luminosity in the quiescent state is observed at
119: $\varepsilon\gtrsim 100$ keV. Moreover, as demonstrated below, the
120: inhomogeneity of the neutron star atmosphere stabilizes the
121: two-stream instability so that collisionless heating of the
122: atmosphere is quenched.
123:
124: An alternative explanation is that the nonthermal tail is formed
125: via resonant scattering of the thermal surface radiation at high
126: altitudes, about 10 stellar radii, where the Landau energy is
127: already nonrelativistic. This would mean that the persistent
128: emission of the magnetar is associated only with a small fraction
129: of the magnetic field lines (those rising into the resonant
130: region) whereas the currents flowing along the most of magnetic
131: field lines and carrying most of the energy do not show up.
132: Moreover, careful calculations show that the resonant scattering
133: cannot reproduce the observed rising energy spectra of the
134: persistent emission (Fern\'andes \& Thompson 2007).
135: %{\bf
136: %The two-stream instability could develop if the atmosphere is
137: %already heated to the temperatures $T> 100$ keV. If such an
138: %atmosphere was formed somehow, the collisionless heating would
139: %keep it hot. However there should be no long-term self-sustaining
140: %equilibrium because the hot electrons will slowly diffuse into the
141: %cold atmosphere eventually cooling down whereas cold electrons
142: %could not be heated up to the temperatures necessary for them to
143: %rise into the hot corona where collisionless heating operates. The
144: %reason is that collisions between one-dimensional electrons do not
145: %result in relaxation, i.e. if electrons with momenta $p_1$ and
146: %$p_2$ collide, conservation of the energy and momentum imply that
147: %after the collision, electrons with the same momenta $p_1$ and
148: %$p_2$ were obtained. Therefore hot electrons may heat the
149: %underlying matter only by ionizing ions or by heating them
150: %directly by Coulomb collisions via recoil effect. The ion
151: %temperature remains anyway below the energy of the ion Landau
152: %level $\varepsilon_{cp}=5Z/A$ keV because of efficient cyclotron
153: %cooling hence the background cool electrons could not be heated
154: %significantly by collisions with ions. As to electrons knocked out
155: %of ions by collisional ionization, their energy is generally
156: %significantly less then the energy of the ionizing electron.}
157: %there should be two interpenetrating electron fluids with the
158: %common neutralizing ion background. They form two atmospheres, one
159: %is thin and cold and another is hot and extended.
160:
161: In this paper, we reanalyze interaction of the fast particle beam
162: with the surface layers of the neutron star and show that, because
163: of the extreme physics introduced by the ultrastrong magnetic
164: field, a hot atmosphere could be formed via \emph{collisional}
165: processes and these difficulties avoided. %It was already noticed
166: %in BT that the primary beam is formed in the magnetosphere, hits
167: %the surface of the neutron star and is eventually stopped within
168: %the cold atmosphere via cascade pair production, the energy of the
169: %beam being eventually converted into the energy of electrons with
170: %the energy less than $\varepsilon_B$.
171: It will be demonstrated
172: below that even though these electrons appear at a relatively
173: large depth, they could escape upwards forming a hot layer with
174: the temperature $T\approx \varepsilon_B/2\approx 1\div 2$ MeV at
175: the top of the cold atmosphere. The reason is that the Coulomb
176: cross section sharply decreases when the electron energy becomes
177: less than $\varepsilon_B$ and the electron is restricted to move
178: only along the magnetic field like a bead on a wire. These
179: electrons could carry away a significant fraction of their initial
180: energy if the atmosphere is fully ionized so that there are no
181: ionization losses.
182: %and at the same time, the recoil effect in the course of the Coulomb
183: %collisions with ions is weak enough.
184: We show that helium atmosphere satisfies these conditions. Such an
185: atmosphere is expected in view of spallation of heavy nuclei
186: bombarded by high-energy electrons and positrons.
187:
188: Bremsstrahlung radiation from this hot layer is hard enough to
189: populate the magnetosphere by electron-positron pairs. This pair
190: corona is also heated via collisionless relaxation of the primary
191: beam because the corona, unlike the atmosphere, is extended enough
192: for the two-stream instability to develop. The observed hard
193: X-rays are radiated via Comptonization in the corona and
194: bremsstrahlung in the hot atmosphere. Our model predicts a hard
195: radiation spectrum extending to the MeV band. In the band
196: $\lesssim 100$ keV, Comptonization forms a power-law spectrum with
197: the photon slope $1<\alpha<2$ as is observed. Because the corona
198: is extended, most of radiation is not intercepted by the surface
199: of the neutron star but rather escapes to the infinity. Therefore
200: the observed persistent radiation from magnetars is dominated by
201: hard emission.
202:
203: The paper is organized as follows. In sect.2, we analyze the two
204: stream instability in the neutron star atmosphere and show that
205: the instability is suppressed by the inhomogeneity of the
206: atmosphere. In sect.3, we consider deceleration of the fast
207: particle beam via resonant Coulomb scattering
208: and development of the electron-positron cascade in the atmosphere
209: (Kotov \& Kelner 1985, Beloborodov \& Thompson 2007);
210: we show that the initial energy of the beam is eventually converted into
211: electrons with the energy something below $\varepsilon_B$ at the depth of
212: roughly 100 g/cm$^2$. In sect.4, we demonstrate that these electrons could escape
213: upwards delivering most of the original energy into a hot layer at the
214: top of the cold atmosphere. In sect.5, we argue that
215: hard radiation from this hot layer populates the magnetosphere with
216: electron-positron pairs and this pair corona emits hard radiation
217: extended into the MeV band. Conclusions are presented in
218: Sect.6. In Appendix A, we show that the energy the electron loses in the
219: atmosphere for recoil of ions is small. In Appendix B, we
220: demonstrate that helium is fully ionized in the atmospheres of magnetars.
221: In Appendix C, we show that Comptonization of soft photons on mildly relativistic electrons
222: in a super-strong magnetic field results in a power-law radiation spectrum with the
223: photon index $\alpha\ge 1$, like in the non-magnetized case.
224:
225: \section{Inefficiency of the collisionless interaction of the beam with the atmosphere}
226:
227: Thompson \& Beloborodov (2005) and Beloborodov \& Thompson (2007)
228: suggested that the energy of the magnetospheric currents is
229: dissipated when the downward electron-positron beam with the
230: Lorentz factor (\ref{gamma}) is decelerated in a thin surface
231: layer of the star where two-stream instability excites Langmuir
232: turbulence thus providing the necessary relaxation mechanism. In
233: this section, we check conditions for the development of the
234: two-stream instability. One can conveniently express the plasma
235: density in the atmosphere via the Thompson optical depth, $\tau$,
236: as
237: \begin{equation}
238: n=\frac{\tau}{\sigma_TH};\label{n}
239: \end{equation}
240: where
241: \begin{equation}
242: H=\frac{(1+Z)T}{Am_pg}=2.6\frac{(1+Z)T_7}{Ag_{14.5}}\,\rm cm
243: \label{height}
244: \end{equation}
245: is the hydrostatic height of the atmosphere, $g=10^{14.5}g_{14.5}$
246: cm$\cdot$s$^{-2}$ the surface gravity, $A$ and $Z$ the atomic
247: weight and charge, correspondingly. The beam density may be
248: expressed via the total energy release in the magnetosphere,
249: $L=10^{36}L_{36}$ erg/s, as
250: \begin{equation}
251: n_b=\frac L{4\pi R_*^2
252: m_ec^3\gamma}=4\times10^{15}\frac{L_{36}}{\gamma_3}\,\rm
253: cm^{-3};\label{n_b}
254: \end{equation}
255: where $\gamma_3=\gamma/10^3$.
256:
257: The basic regimes of the two-stream
258: instability for a relativistic beam propagating in a
259: non-relativistic plasma were studied by Fainberg, Shapiro \&
260: Shevchenko (1970). For a monochromatic beam moving with the
261: Lorentz factor $\gamma$ along the strong magnetic field, the
262: growth rate is written in the hydrodynamic regime,
263: \begin{equation}
264: \kappa_{hydr}=\frac{\sqrt{3}}{2^{4/3}}\left(\frac{n_b}n\right)^{1/3}\frac{\omega_p}{\gamma};\label{hydr}
265: \end{equation}
266: where
267: \begin{equation}
268: \omega_p=\sqrt{\frac{4\pi e^2n}{m_e}}\label{omegap}
269: \end{equation}
270: is the plasma frequency. This formula is valid provided all the
271: particles in the beam are in resonance with the excited wave,
272: $\delta v/c<\kappa/\omega_p$. For the beam with a large energy
273: spread, $\delta\gamma\sim\gamma$, this condition is written as
274: \begin{equation}
275: \left(\frac{n_b}{n}\right)^{1/3}\gamma>1.\label{hydrkin}
276: \end{equation}
277: Making use of Eqs.(\ref{n}-\ref{hydr}), one can write this
278: condition as
279: \begin{equation}
280: \gamma>400\left(\frac{A\tau
281: g_{14.5}}{(1+Z)T_7L_{36}}\right)^{1/2}.\label{hydr-kin}
282: \end{equation}
283: In the opposed limit, only some fraction of the beam particles are
284: in resonance with the excited wave (kinetic regime). Then the
285: growth rate is
286: \begin{equation}
287: \kappa_{kin}\approx\frac{n_b}{n}\frac{\omega_p}{\gamma^3\delta
288: v^2}\approx\frac{n_b\gamma}{n}\omega_p.\label{kin}
289: \end{equation}
290: Making use of these estimates, one easily finds that $\kappa H/c$
291: is very large in both regimes therefore at first glance, strong
292: Langmuir turbulence should be excited. However, careful
293: consideration shows that this is not the case because at
294: $\kappa/\omega_p\ll 1$, the instability is easily stabilized by
295: plasma inhomogeneity % either/or by non-linear processes
296: (e.g. Breizman \& Ryutov 1974; Breizman 1990). Let us consider how
297: inhomogeneity of the neutron star atmosphere affects development
298: of the two-stream instability.
299:
300: The instability excites the Langmuir waves with the dispersion
301: relation
302: \begin{equation}
303: \omega=\omega_p\left(1+\frac{3kT}{2m_ec^2}\frac{k^2c^2}{\omega_p^2}\right);\label{disp}
304: \end{equation}
305: where $T$ is the plasma temperature, $k$ the wave vector. Within
306: the atmosphere with the characteristic height (\ref{height}), the
307: plasma frequency varies with the depth, $z$, as
308: \begin{equation}
309: \frac{\delta\omega_p}{\omega_p}=\frac{\delta z}{2H}.
310: \end{equation}
311: The frequency of the propagating wave remains constant whereas the
312: wave vector varies to satisfy the dispersion equation
313: (\ref{disp}):
314: \begin{equation}
315: \delta\omega_p+\frac{3kT}{m_ec}\delta k=0.
316: \end{equation}
317: Here we take into account that for the resonance wave,
318: $\omega/k\approx\omega_p/k\approx c$. The waves are amplified only
319: if their phase velocity is close to the beam velocity,
320: \begin{equation}
321: \vert\omega-vk\vert\lesssim\kappa.\label{res}
322: \end{equation}
323: %In the case of interest, the resonance is very narrow,
324: %\begin{equation}
325: %\frac{\kappa}{\omega_p}=\frac{\sqrt{3}}{2^{4/3}\gamma}\left(\frac{n_b}n\right)^{1/3}=1.3\times
326: %10^{-6}\gamma_3^{-4/3}\left(L_{36}T_7/\tau\right)^{1/3};
327: %\end{equation}
328: %therefore development of the instability may be easily suppressed
329: %by non-linear interactions of the excited waves (Tsytovich \& Shapiro
330: %1965) or/and inhomogeneity of the plasma (Breizman \& Ryutov
331: %1970). Let us consider the fate of the excited waves propagating
332: %in the inhomogeneous atmosphere.
333: In the hydrodynamic regime, when the beam velocity spread is less
334: than the width of the resonance, the condition (\ref{res}) yields
335: $\delta k<\kappa_{hydr}/c$. Then one finds that
336: the amplification stops after the wave propagates the distance
337: \begin{equation}
338: \delta
339: z=\frac{6kT}{m_ec^2}\frac{\kappa_{hydr}}{\omega_p}H.\label{deltaz}
340: \end{equation}
341: A significant fraction of the beam energy could be dissipated
342: provided
343: \begin{equation}
344: \frac{\kappa_{hydr}\delta z}{v_g}>\Lambda;\label{instcond}
345: \end{equation}
346: where $\Lambda=10\Lambda_1$ is the logarithm of the ratio of the
347: beam energy to the initial energy of the Langmuir oscillation (it
348: is of the order of the Coulomb logarithm) and it is taken into
349: account that the wave propagates with the group velocity
350: \begin{equation}
351: v_g=\frac{d\omega}{dk}=\frac{3kT}{m_ec^2}c.\label{v_g}
352: \end{equation}
353: Now one finds making use of Eqs.(\ref{hydr}), (\ref{deltaz}) and
354: (\ref{v_g})
355: \begin{equation}
356: \frac{\kappa_{hydr}\delta z}{v_g\Lambda}=
357: %2\frac{\kappa_{hydr}^2H}{c\omega_p\Lambda}=
358: 1.3\times 10^{-6}
359: \frac{(1+Z)T_7^{7/6}L_{36}^{2/3}}{Ag_{14.5}\gamma_3^{8/3}\tau^{1/6}\Lambda_1}.
360: \label{ampl}
361: \end{equation}
362: So the condition (\ref{instcond}) is violated by a large margin.
363:
364: In the kinetic regime, the condition (\ref{res}) reduces to
365: $\delta k/k<\delta v/c\sim (c\gamma^2)^{-1}$, which yields
366: \begin{equation}
367: \delta z=\frac{6kT}{m_ec^2\gamma^2}H.
368: \end{equation}
369: Then one finds
370: \begin{equation}
371: \frac{\kappa_{kin}\delta
372: z}{v_g\Lambda}=%\frac{2\kappa_{kin}}{c\gamma^2\Lambda}=
373: 4.3\times
374: 10^{-4}\frac{1}{\Lambda_1\gamma_2^2}\sqrt{\frac{(1+Z)T_7}{A\tau
375: g_{14.5}}}.
376: \end{equation}
377: Again, the condition (\ref{instcond}) is violated by a large
378: margin. Therefore
379: %, contrary to what was claimed in TB and BT,
380: the two-stream instability does not develop in the atmosphere of
381: the neutron star and the atmosphere could not be heated via
382: collisionless dissipation.
383:
384: It will be shown below that a hot atmosphere could arise in the
385: case under consideration but as a result of \emph{collisional}
386: interaction of the beam with the surface layers of the star. Then
387: the characteristic scale of the density variation significantly
388: increases and moreover, the hot atmosphere emits hard X-rays
389: therefore the Lorentz factor of the beam decreases according to
390: Eq.(\ref{gamma}). Substituting in Eq.(\ref{ampl}) the
391: characteristic temperature of the hot atmosphere, $T=10^{10}$ K,
392: and the Lorentz factor $\gamma=100$, one can see that the
393: two-stream instability could develop in the hot atmosphere.
394: Therefore eventually some fraction of the beam energy could be
395: released via collisionless dissipation. However, at least one half
396: of the energy is in any case dissipated and delivered into the hot
397: atmosphere by collisional processes; it is these processes that
398: determine the temperature of the hot atmosphere.
399:
400: \section{Collisional deceleration of the beam}
401: Consider deceleration of a
402: relativistic electron-positron beam within the surface layers of
403: the magnetar. Let a relativistic electron move in plasma along the
404: strong magnetic field. If the electron energy exceeds the Landau
405: energy (\ref{Landau}), it efficiently scatters off ions into the
406: first Landau level and then immediately falls back emitting a
407: resonance photon; such a resonance Coulomb scattering is the main
408: mechanism of deceleration of a relativistic electron-positron beam
409: in the case under consideration (Kotov \& Kelner 1985, Beloborodov
410: \& Thompson 2007). The cross section for the transition of an
411: electron with the Lorentz factor $\gamma$ and the momentum $p$
412: from the ground to the $j$-th Landau level by scattering off an
413: ion with the charge $Ze$ is found by Bussard (1980) and Langer
414: (1981):
415: \begin{equation}
416: \sigma_{j0}=\frac
417: 3{16}\frac{B_q}B\frac{Z^2\sigma_T}{(1+\gamma)^2}\frac
418: 1{j!}\sum_{\pm}\frac{m_e^2c^2}{\mid pp_{\pm}\mid}\label{Coulomb}
419: \end{equation}
420: $$
421: \times
422: \left\{\delta_{s',-1/2}\left[(1+\gamma)^2+\frac{pp_{\pm}}{m_e^2c^2}\right]^2+
423: \delta_{s',1/2}\frac{2B}{B_q}\frac{p^2}{m_e^2c^2}\right\}
424: \epsilon_j\left(\frac{B_q}{2B}\frac{(p-p_{\pm})^2}{m_e^2c^2}\right);
425: $$$$
426: \epsilon_j(x)=\int_0^{\infty}\frac{t^jdt}{(t+x)^2}e^{-t}.
427: $$
428: Here $\sigma_T$ is the Thomson cross section, $s'$ the final spin
429: of electron and summation is over the final electron momentums
430: $p_{\pm}=\pm\sqrt{p^2-2m_ec^2jB/B_q}$. Even high energy electrons
431: jump predominantly on the first Landau level therefore one can
432: neglect excitations of higher levels. The cross section for the
433: transition $0\to 1$ is plotted in Fig.(1). In the limit
434: $mc^2\gamma\gg\varepsilon_B$, it is reduced to
435: \begin{equation}
436: \sigma_{10}=\frac{3B_q}{4B}Z^2\sigma_T\ln\left(0.413\frac{\gamma^2B_q}{B}\right).
437: \label{asymp}\end{equation}
438: %At $\gamma=2\varepsilon_B/m_ec^2$ the
439: %cross section is equal to $\sigma_{01}=(B_q/B)Z^2\sigma_T$.
440: %Note that high energy electrons efficiently split nuclei in the
441: %atmosphere (BT) therefore $Z$ may be not large.
442:
443: The electron with the Lorentz factor $\gamma>\sqrt{2B/B_q}$ jumps
444: on the first Landau level having the Lorentz factor
445: $\gamma_1=m_ec^2\gamma/\varepsilon_B=0.15\gamma B_{15}^{-1/2}$.
446: The electron immediately falls back onto the background Landau
447: level emitting a photon with the energy $(0.5\div 1)\varepsilon_B$
448: as measured in the guiding center frame of the excited electron
449: (frame moving with the Lorentz factor $\gamma_1$); due to recoil,
450: the energy of the emitted photon depends on the emission angle.
451: After the deexcitation, the Lorentz factor of the electron remains
452: on average the same because in the guiding center frame, photons
453: are emitted forward and backward with equal probability. Thus the
454: electron retains only a fraction $\xi=\gamma_1/\gamma\approx
455: 0.15B_{15}^{-1/2}$ of the total energy, most of the energy being
456: taken away by a photon.
457:
458: \begin{figure*}
459: \includegraphics[width=10 cm,scale=0.8]{f1.ps}%{sigma1num.ps}
460: \caption{Cross-section of the Coulomb excitation of the first
461: Landau level; asymptotics (\ref{asymp}) is shown by dashed lines.}
462: \end{figure*}
463:
464: The fate of the photon depends on its polarization. Two
465: polarization modes could propagate in the magnetized vacuum; the
466: so called ordinary mode is polarized in the plane set by the
467: propagation direction and the background magnetic field whereas
468: the extraordinary mode is polarized perpendicularly to this plane.
469: If an ordinary photon is emitted, it is immediately converted into
470: an electron-positron pair provided its energy exceeds the
471: threshold, $\varepsilon>2m_ec^2/\sin\theta$. In magnetar's field,
472: this condition is satisfied for most $\theta$. In the frame of the
473: scattered electron, the produced electron and positron have on
474: average equal energies and move in opposite directions. In the
475: laboratory frame, most of the energy is taken by the particle
476: moving forward; only a fraction $1/\gamma_1^2$ of the total energy
477: is taken by the second particle. This means that the energy of
478: electrons in the beam decreases only by a fraction $\xi$ in one
479: scattering. If this were the only process of the beam-plasma
480: interaction, the energy of the particles in the beam would
481: decrease $\exp{(r\xi)}$ times after $r$ scattering and the total
482: number of scattering before the particle energy becomes less than
483: the Landau energy would be large,
484: $r\approx\xi^{-1}\ln(\gamma/\varepsilon_B)\approx 20\div 30$. The
485: full length of the avalanche, $l$, could be estimated by summing
486: the free path lengths:
487: $$
488: l=\sum_r\frac
489: 1{\sigma_{01}n_i}\approx\frac{4B}{3B_qZ^2\sigma_Tn_i}
490: \int_0^r\frac{dn}{\ln(0.4\gamma^2B_q/B)-2\xi n}
491: $$$$
492: \approx \frac{2B}{3\xi
493: B_qZ^2\sigma_Tn_i}\ln\left[\ln\left(\frac{0.4\gamma^2B_q}B\right)\right].
494: $$
495: Here $n_i$ is the number density of ions. For $\gamma\sim 100\div
496: 1000$, the corresponding Thomson depth is
497: \begin{equation}
498: \tau\equiv\sigma_TZn_i l\sim \frac 2Z\left(\frac
499: B{B_q}\right)^{3/2}=200Z^{-1}B_{15}^{3/2}.\label{tau0}
500: \end{equation}
501:
502:
503: An extraordinary photon does not produce pairs directly; it first
504: splits into ordinary photons and only then pairs may be produced.
505: Then the avalanche proceeds further. As the energy per particle
506: decreases in this case roughly twice in each step, the avalanche
507: penetrates the depth significantly less than that of
508: Eq.(\ref{tau0}) provided each emitted O-photon is converted into
509: a pair. However, the energy of the photon, $(0.5\div
510: 1)\varepsilon_B\approx 2\div 3$ MeV, is now distributed between
511: two photons, both or one of them could fall below the threshold
512: for the pair production. Then these photons either are converted
513: into pairs in the Coulomb field of the nucleus or experience a
514: Compton scattering off a background electron.
515:
516: The cross section for the pair production at a nucleus is
517: \begin{equation}
518: \sigma^Z_{\pm}=\frac 7{6\pi}Z^2\alpha\sigma_T
519: \left[\ln\left(\frac{2\varepsilon}{m_ec^2}\right)-\frac{109}{42}\right];\label{Z}
520: \end{equation}
521: where $\alpha$ is the fine structure constant. The Compton
522: scattering of a photon moving along the superstrong magnetic field
523: was studied by Gonthier et al. (2000). In this case, the resonance
524: occurs only at the cyclotron fundamental
525: $\varepsilon=(B/B_q)m_ec^2=12B_{15}$ MeV. Above the resonance, the
526: scattering cross section is close to the Klein-Nishina one
527: \begin{equation}
528: \sigma_C=\frac 34\sigma_T\frac{m_ec^2}{2\varepsilon}
529: \left[\ln\left(\frac{2\varepsilon}{m_ec^2}\right)+\frac
530: 12\right].\label{Compton}
531: \end{equation}
532: The pair production at nuclei dominates at energies
533: $\varepsilon>250Z^{-1}m_ec^2$. Note that this process comes into
534: play only when the photon could not be converted directly into a
535: pair. This occurs predominantly when E-mode resonant photons are
536: emitted because these photons split into O-photons with lesser
537: energies so that these O-photons could fall below the threshold
538: for the direct pair production in the magnetic field. In this
539: case, two pairs are produced in each step therefore the energy per
540: particle decreases roughly four times in each step. This means
541: that pairs are produced at nuclei only in the first few
542: generations; after this, Compton scattering dominates. The
543: corresponding depth varies from $\tau_Z\sim 200/Z$ at small $Z$,
544: when the transition energy is high and the Compton scattering
545: comes into play already after two generations, to $\tau_Z\sim
546: 800/Z$, when $Z$ is large and three or four generations are
547: necessary.
548:
549: At the Compton stage, the avalanche proceeds further. The
550: distribution of the scattered photons in their energies,
551: $\varepsilon'$, is
552: \begin{equation}
553: d\sigma_C=\frac
554: 38\sigma_T\frac{m_ec^2d\varepsilon'}{\varepsilon^2}
555: \left[\frac{\varepsilon}{\varepsilon'}+\frac{\varepsilon'}{\varepsilon}+
556: \left(\frac{m_ec^2}{\varepsilon'}-\frac{m_ec^2}{\varepsilon}\right)^2-
557: 2\left(\frac{m_ec^2}{\varepsilon'}-\frac{m_ec^2}{\varepsilon}\right)\right];
558: \end{equation}
559: where
560: $$
561: \frac{\varepsilon}{1+2\varepsilon/m_ec^2}\le\varepsilon'\le\varepsilon.
562: $$
563: The scattered photon takes on average a fraction
564: $4/[3\ln(2\varepsilon/m_ec^2)]\approx 0.2$ of the initial energy.
565: The photon is directed at the angle
566: $\sim\sqrt{4m_ec^2/\varepsilon}$ to the magnetic field therefore
567: it is immediately converted into a pair. The rest 0.8 fraction of
568: the energy is taken by the recoil electron, which emits a resonant
569: photon
570: % so that the avalanche proceeds further. New pairs emit resonant photons
571: via Coulomb scattering off an ion. If this photon is in O-mode, it
572: is immediately converted into a pair; if it is in E-mode, it
573: splits producing two photons, the energy of each of them being
574: roughly $(1/2)\cdot 0.8(1-\xi)\approx 1/3$ of the initial energy.
575: % and one more generation of pairs is produced
576: %either directly in the magnetic field or, if they are below the
577: %threshold, again at the nucleus.
578: These two photons either produce pairs directly or, if they are
579: below the threshold for the direct pair production, experience
580: Compton scatterings and the process repeats again.
581: % Therefore one can estimate the maximal depth the
582: %avalanche penetrates assuming that pairs are produced only at nuclei.
583: The Compton scattering comes into play predominantly if an E-mode
584: photon was emitted because a resonant O-photon is converted into a
585: pair immediately whereas the E-photon splits into two O-photons,
586: which could fall below the threshold of the direct pair
587: production. Therefore in the Compton channel, the energy per
588: particle decreases by the factor of 3 in each generation. Then the
589: energy per particle becomes less than $\varepsilon_B$ in a few
590: steps and the avalanche stops. As the Compton cross section grows
591: rapidly with decreasing of the photon energy and cyclotron
592: resonances only increase the cross section, the full length of the
593: avalanche is determined by the free path of a photon with the
594: initial energy $\varepsilon=250Z^{-1}m_ec^2$. The corresponding
595: depth varies from $\tau_C\sim 100$ at $Z=1$ to $\tau_C\sim 10$ at
596: $Z=26$. We will argue in the next section that for our model to
597: be self-consistent, the avalanche should develop in the medium
598: composed from light elements. Spallation of nuclei by relativistic
599: electrons and positrons from the avalanche seems to result in
600: formation of a helium atmosphere; below we will consider only this
601: case. Then the total depth the avalanche penetrates is estimated
602: as $\tau=\tau_Z+\tau_C\sim 300/Z$.
603:
604: In real avalanches, ordinary and extraordinary photons are emitted
605: alternately with approximately the same probability. As the
606: Coulomb cross section (\ref{Coulomb}) is significantly larger than
607: (\ref{Compton}) and (\ref{Z}), the longest parts of the chain is
608: associated with emission of extraordinary photon with subsequent
609: photon splitting. On the other hand, the total number of
610: generations is also determined by emission of extraordinary
611: photons because the energy per electron decreases significantly
612: when the photon splits. Therefore the overall length of the
613: avalanche is roughly the same as if only extraordinary photons
614: were emitted. So finally the energy of the primary beam is
615: converted into electron-positron pairs with the energy $\sim
616: \varepsilon_B/2\sim 1\div 2$ MeV at the depth of
617: \begin{equation}
618: \tau_0=\sigma_TZn_i l\sim 300/Z. \label{tau}\end{equation}
619:
620: \begin{figure*}
621: \includegraphics[width=10 cm,scale=0.8]{f2.ps}%{logsigma0b15.ps}
622: \caption{Coulomb cross-section for the electron on the ground
623: Landau level, $B=10^{15}$ G.}
624: \end{figure*}
625:
626: \section{Formation of a hot atmosphere}
627: Now let us consider the fate of the newly formed pairs with the
628: energy less than $\varepsilon_B$. Their motion is purely
629: one-dimensional and conservation of the energy and momentum
630: implies that two colliding particles of equal mass may only
631: exchange their energy and momenta. Therefore electron-electron
632: collisions do not change the state of the system and may be
633: neglected. Collision of a "hot" positron with a "cold" background
634: electron results in a "cold" positron and "hot" electron.
635: Collision frequency of cold particles is high therefore the cold
636: positron does not escape but rather eventually annihilates with
637: some background electron. However, the hot electron takes most of
638: the energy in this case therefore most of the beam energy is
639: eventually stored in electrons with the characteristic energy
640: $\sim 1\div 2$ MeV. They diffuse within the medium colliding with
641: the background ions.
642:
643: Collision of a one-dimensional electron with the ion may result
644: either to the forward scattering, after which the electron energy
645: remains the same to within a small recoil, or to reflection from
646: the ion. The cross section for the Coulomb reflection is found
647: from Eq.(\ref{Coulomb}) as
648: \begin{equation}
649: \sigma_{00}=\frac
650: 34Z^2\sigma_T\frac{B_q}B\left(\frac{m_ec}p\right)^2\epsilon_0\left(\frac{2B_qp^2}{Bm_e^2c^2}\right);
651: \label{reflection}
652: \end{equation}
653: this cross section is plotted in Fig. 2. For
654: $p=0.5\varepsilon_B/c=\sqrt{B/2B_q}m_ec$ one gets
655: \begin{equation}
656: \sigma_{00}=0.6Z^2\sigma_T(B_q/B)^2=0.001Z^2\sigma_TB_{15}^{-2}.
657: \label{reflection1}
658: \end{equation}
659: It was shown in the previous section that the avalanche penetrates
660: the Thomson depth (\ref{tau}) producing electrons with the energy
661: $\sim\varepsilon_B/2$. One sees that for these electrons,
662: $\sigma_{00}n_i l=0.3B_{15}^{-2}\lesssim 1$ so they could easily
663: escape upwards taking away most of the energy of the avalanche.
664: The cross-section of the forward scattering is larger than (\ref{reflection1})
665: but because recoil is small, one can neglect this process
666: % if the atmosphere is composed of ions heavier than hydrogen
667: (see Appendix A).
668: %The electron escapes from the Thomson depth $\tau$ after
669: %experiencing $s\sim(\sigma_{00}n_i l)^2=[0.6Z(B_q/B)^2\tau]^2$
670: %collisions. Substituting the largest estimate for the depth the
671: %avalanche penetrates (\ref{tau1}) one gets $s=0.7B^{-4}_{15}$.
672: %Comparing this estimate with the number of scattering necessary
673: %for the electron to lose a significant fraction of its energy due
674: %to recoil, $Am_pc^2/\varepsilon_c$, one can see that at
675: %$B_{15}>0.16A^{-2/7}$
676: %$\tau_0\sim 300/Z$ at small $Z$. to $\tau\sim 800/Z$ at large $Z$.
677: %One sees that this layer is nearly
678: %transparent for electrons with the energy $\sim 0.5\varepsilon_B$
679: %so that they could escape upwards.
680: % after a few scatterings on ions. As the scattering is nearly elastic, the
681: %electrons take away most of the energy of the avalanche.
682:
683: The above consideration assumes that the escaping electron does
684: not lose energy on ionization so that the plasma in the atmosphere
685: is fully ionized. It is shown in the Appendix B that under the
686: condition of interest, helium is fully ionized. One can assume
687: that the layer the avalanche penetrates composed mostly of helium
688: because the avalanche electrons and gamma-quanta destroy heavy
689: nuclei(Beloborodov \& Thompson 2007). The helium is presumably
690: preferable to hydrogen because does not require transformation of
691: neutrons into protons.
692:
693: One should check how much energy the electrons lose on
694: bremsstrahlung before they escape. In the non-magnetized medium, a
695: relativistic electron loses energy on bremsstrahlung after passing
696: the depth, in Thomson units,
697: \begin{equation}
698: \tau^{\rm br}_{0}=\frac{2\pi}{3\alpha Z(\ln
699: 2\gamma-1/3)}.\label{br}
700: \end{equation}
701: Unfortunately bremsstrahlung in the superstrong magnetic field has
702: not been calculated yet however simple quasiclassical estimate
703: shows that the rate of the energy losses decreases in this case.
704:
705: In classical electrodynamics, the energy radiated by a
706: relativistic electron may be written as (e.g. Landau \& Lifshitz
707: 1995)
708: \begin{equation}
709: \delta{\cal E}
710: =\frac{2e^2}{3c^5}\int_{-\infty}^{\infty}\gamma^6[(\mathbf{v\cdot
711: w})^2+c^2w^2/\gamma^2]dt;\label{brems}
712: \end{equation}
713: where $\mathbf{w}$ is the electron acceleration. From the
714: equation of motion
715: $$
716: m_e\frac{d\gamma\mathbf{v}}{dt}=e\mathbf E
717: $$
718: one finds $w_{\|}=eE_{\|}/m_ec\gamma^3$ and
719: $w_{\bot}=eE_{\bot}/m_ec\gamma$ where the subscripts $\|$ and
720: $\bot$ refer to the projections onto the direction of velocity and
721: onto the perpendicular direction. In the non-magnetized case, both
722: components of the acceleration are presented however the first
723: term in the square brackets in Eq.(\ref{brems}) is $\gamma^2$ less
724: than the second one and may be neglected. For the electron passing
725: the nucleus at the distance $r$ one gets assuming that the motion
726: is straightforward
727: \begin{equation}
728: \delta {\cal E}=\frac{\pi
729: Z^2e^4\gamma^2}{12r^3m_e^2c^4}.\label{brems1}
730: \end{equation}
731: Integrating the obtained relation over $r$ from
732: $r_{min}=\hbar\gamma/m_ec$ defined from the condition that the
733: energy of the emitted quanta becomes equal to the electron energy,
734: one gets Eq.(\ref{br}) to within a numerical factor and the
735: logarithmic term.
736:
737: In the superstrong magnetic field, the electrons move only along
738: the field therefore only the first term in the square brackets in
739: Eq.(\ref{brems}) remains. Simple calculation shows that in this
740: case $\delta {\cal E}$ is $3\gamma^2$ less than that of
741: Eq.(\ref{brems1}). One can expect that the full bremsstrahlung
742: cross section in the superstrong magnetic field decreases by the
743: same factor. Of course quantum treatment of the process is
744: necessary to rigorously justify this conclusion, but assuming that
745: the above consideration is basically correct, the electron in the
746: superstrong magnetic field loses the energy on bremsstrahlung
747: after passing the depth $\tau^{\rm br}=3\gamma^2\tau^{\rm
748: br}_{0}$. For electrons with the energy $\sim\varepsilon_B/2$, one
749: gets $\tau^{\rm br}=6000Z^{-1}B_{15}$, which is significantly
750: larger than the depth (\ref{tau}) the avalanche penetrates. This
751: means that one can neglect bremsstrahlung energy losses.
752:
753: So hot electrons freely escape upwards taking away most of the
754: beam energy. Charge neutrality implies that the necessary amount
755: of background ions rises together with the electrons so that a hot
756: atmosphere with the temperature $T\approx \varepsilon_B/2\approx
757: 1\div 2$ MeV and the characteristic height
758: $H=26(1+Z)T_{10}/Ag_{14.5}$ m is formed.
759:
760: \section{Pair corona}
761:
762: The hot atmosphere is cooled via bremsstrahlung; radiation is
763: dominated by photons with the energy $\sim T\sim 1\div 2$ MeV. One
764: half of this radiation is directed towards the star; it is
765: absorbed and eventually is reradiated as thermal emission. The
766: other half of radiation is radiated away. In the magnetosphere,
767: hard photons could be converted into electron-positron pairs. The
768: pair production rate may be written as $\dot N=\zeta L/m_ec^2$
769: where the numerical factor $\zeta<1/2$ takes into account
770: uncertainties in the radiation spectrum and the field geometry.
771: Within the light travel time, $R_*/c$, the magnetosphere will be
772: filled by pairs with the density $n=8\times 10^{18}\zeta L_{36}$
773: cm$^{-3}$. The characteristic inhomogeneity scale of the corona is
774: of the order of the star radius, which is large enough for the
775: two-stream instability to develop in the corona. The condition for
776: the instability (\ref{instcond}) is satisfied provided the
777: left-hand side of Eq.(\ref{ampl}) exceeds unity. Substituting in
778: this relation $H$ by $R_*$ and making use of Eqs.(\ref{hydr}),
779: (\ref{omegap}) and the above estimate for the density of the
780: corona, one gets
781: $$
782: \frac{\kappa_{kin}\delta
783: z}{v_g\Lambda}=10^3\frac{L_{36}^{1/2}}{\zeta^{1/6}\gamma^{8/3}_3\Lambda_1}.
784: $$
785: Note that the condition (\ref{hydrkin}) is satisfied in the corona
786: at any reasonable parameters therefore the instability is
787: hydrodynamic. So after the corona is formed, the primary beam
788: would experience collisionless relaxation. Then about one half of
789: the beam energy is spent on heating of the corona whereas the
790: other half will be delivered to the surface of the star where the
791: the electron-positron avalanche is developed and the energy of the
792: beam is eventually transferred to electrons with the energy
793: $\sim\varepsilon_B/2\sim 1\div 2$ MeV each. It was argued in the
794: preovious sections, that these electrons escape upwards delivering
795: most of their energy into the tenious, hot atmosphere.
796: %As it was argued
797: %in the previous section, the bremsstrahlung cooling rate decreases
798: %by the same factor therefore one can write
799: %\begin{equation}
800: %\left(\frac{dE}{dt}\right)_{br}=-\alpha\sigma_Tnm_ec^3\frac{m_ec^2}{T}.
801: %\end{equation}
802: %Substituting Eq.(\ref{height}) and writing the radiation energy
803: %density as $U=L/4\pi cR^2$, one finds
804: %\begin{equation}
805: %\frac{(dE/dt)_{br}}{(dE/dt)_C}=0.5\tau L^{-1}_{36}T_{10}^{-2}
806: %\end{equation}
807: %so that generally both processes contribute comparably to the
808: %cooling rate.
809:
810: The corona is efficiently cooled by Comptonization of the thermal
811: emission from the surface. As the energy of the photons is well
812: below the Landau energy, only O-mode radiation is scattered
813: (polarized in the plane set by the direction of propagation and
814: the magnetic field). Thermal energy stored in the star interior is
815: transferred to the surface and radiated away by E-mode photons
816: because their opacities are $(Bmc^2/B_q\varepsilon)^2>>1$ times
817: less than those for O-mode photons. Soft O-mode photons could be
818: emitted only if surface layers of the star are heated. As hard
819: radiation from the hot atmosphere illuminates the underlaying cold
820: atmosphere, some fraction of this radiation will be absorbed and
821: reradiated in the soft band. It is this thermal O-mode radiation
822: that could be a soft photon source for Comptonization.
823:
824: The scattering cross section for the O-mode photons is
825: $\sigma=\sigma_T\sin^2\theta'$ where $\theta'$ is the angle
826: between the propagation direction and the magnetic field in the
827: proper electron frame. The relativistic electron "sees" radiation
828: at the angle $\theta'\sim 1/\gamma$ therefore the cooling rate
829: decreases $\gamma^2$ times as compared with the non-magnetized
830: case and may be estimated as (Beloborodov \& Thompson 2007)
831: \begin{equation}
832: \left(\frac{d{\cal E}}{dt}\right)_C=-\sigma_TUc;\label{compcool}
833: \end{equation}
834: where $\cal E$ is the energy of the electron (assumed to be larger
835: than $m_ec^2$), $U$ the radiation energy density. Writing the
836: radiation energy density as $U=L/4\pi R_*^2c$, one finds that the
837: electron cooling time
838: \begin{equation}
839: t=4\times 10^{-5} {\cal E}_{\rm MeV}L^{-1}_{36}
840: \end{equation}
841: is comparable with the light travel time.
842:
843: The observed radiation is a superposition of the bremsstrahlung
844: radiation from the hot atmosphere and the Comptonization radiation
845: from the corona; the radiation spectrum extends to the
846: characteristic particle energy of $1\div 2$ MeV. Bremsstrahlung
847: has a flat intensity spectrum (the photon index $1$) at
848: $\varepsilon\ll T$ whereas unsaturated Comptonization produces a
849: power-law spectrum with the photon spectral index $\alpha>1$
850: depending on the parameters of the system (see Appendix C). As the
851: luminosity of the corona is larger than the thermal luminosity of
852: the surface, which provides soft photons for Comptonization, the
853: slope should be hard enough $\alpha<2$. Therefore the
854: low-frequency part of the spectrum is dominated by Comptonization
855: and could exhibit a variety of spectral indices in the range
856: $1<\alpha<2$ as is observed. The observed high pulsed fraction is
857: naturally explained by the fact that the scattering cross section
858: for the ordinary mode photons is highly anisotropic.
859:
860: % The energy balance in the atmosphere is
861: %achieved when the total energy loss $VdE/dt$, where $V$ is the
862: %volume of the atmosphere, becomes equal to the luminosity $4\pi
863: %R_*^2Uc$. This is achieved when the Thomson optical depth of the
864: %atmosphere becomes of the order of a few so that the photons
865: %experience on average a few scatterings before they escape the
866: %atmosphere. As most of the energy of the primary beam is
867: %transferred to the electrons with the energy
868: %$\sim\varepsilon_B/2$, the number density of the particles
869: %escaping the surface layer may be estimated as $n=L/2\pi
870: %R_*^2\varepsilon_Bc=10^{18}B_{15}^{-1/2}$ cm$^{-3}$. The energy
871: %balance is achieved in the time $t\sim (\sigma_T nc)^{-1}\sim
872: %10^{-4}$ s. When soft photons are scattered by hot electrons, they
873: %gain, on average, energy until the photon energy becomes achieves
874: %the thermal energy of electrons, $\varepsilon\sim T$. Photons with
875: %such an energy are efficiently converted into pairs. The pair
876: %plasma is not confined by gravitation but fills the magnetosphere
877: %for the time $R_*/c$, which is of the order of $t$, forming an extended corona.
878:
879: When the pairs fill the magnetosphere, the energy release stops
880: because the pair plasma shorts out the induction electric field.
881: Energetic primary particles, which have already filled the
882: magnetosphere, disappear at the star's surface for about the light
883: travel time. At the next stage, the hot atmosphere and the corona
884: are cooled. The cooling time is of the order of the light travel
885: time. An important point that positrons could not survive for the
886: larger time because they are annihilated when hitting the surface
887: of the star. When a positron falls from the corona onto the
888: surface, it exchanges energy with an electron in the cold
889: atmosphere and annihilates\footnote{Note that even though the
890: annihilating particles are cold, the annihilation line is not
891: formed. In the superstrong magnetic field, only longitudinal
892: (along the field) component of the momentum is conserved therefore
893: the two-photon annihilation results in photons with generally
894: different energies, the annihilation spectrum being extended from
895: 0 to 1 MeV (Kaminker, Pavlov \& Mamradze 1987).}.
896:
897: The corona is expected to be extremely unsteady. Even though most
898: of the energy is taken away by the electron, which reflects from
899: an ion and goes upwards, this electron cannot rise directly into
900: the corona because of the charge neutrality and would remain in
901: the hot atmosphere. The corona is replenished by the hydrodynamic
902: expansion of this hot atmosphere, and cooled by Compton losses.
903: The Compton losses may or may not be compensated by collisionless
904: interaction with the primary beam. Therefore the corona may be
905: depleted by Compton cooling; in any case, we expect that half the
906: primary beam energy makes it down to the hot atmosphere, so that
907: new matter is constantly being added to the corona, and this very
908: likely leads to a non-steady situation. Because the coronal matter
909: can propagate the currents, the displacement currents and
910: attendant primary beam are switched off, while Compton cooling
911: continues. When the temperature of the hot atmosphere falls below
912: $m_ec^2$, the pair production stops and the plasma density in the
913: corona falls below the critical value necessary to maintain the
914: magnetospheric currents. Then the displacement current arises
915: again and the next cycle of energy release starts. So one can
916: expect strong fluctuations of the radiation at the timescale of
917: $\Delta t \ge 10^{-4}$ s. The possibility exists of revealing them
918: by analysis of the photon statistics. If a source emits radiation
919: in separate bursts of the characteristic duration $\Delta t \equiv
920: 10^{-3}\Delta t_{-3}$, the probability distribution for a pair of
921: photons to be detected within the time interval $t$ should differ
922: significantly from a Poisson distribution at $t\sim \Delta t$. For
923: a collecting area of $10^3A_3$cm$^2$, and a photon flux of
924: $10^{-3}F_{-3}$cm$^{-2}$, an observation interval of $10^5t_5$s
925: should contain $10^5A_3F_{-3}t_5$ photons and, for Poisson arrival
926: statistics, $10^2A_3F_{-3}^2t_5\Delta t_{-3}$ pairs of photons
927: arriving within $\Delta t$ of each other. The non-steady nature of
928: the hard coronal emission (this emission in fact dominates the
929: thermal surface emission already at a few keV), which produces
930: deviations from Poisson arrival statistics, is therefore
931: detectable with a sufficiently powerful detector and large
932: exposure times.
933:
934: \section{Conclusions}
935: We considered dissipation of the energy released in the
936: magnetosphere of the magnetar in the course of slow relaxation of
937: non-potential magnetic fields. A basic physical picture was
938: proposed by Thompson et al. (2002) and recently elaborated by
939: Thompson \& Beloborodov (2005) and Beloborodov \& Thompson (2007).
940: When the plasma density in the magnetosphere falls below a
941: critical value necessary to maintain the magnetospheric currents,
942: an induction electric field arises and initiates an
943: electron-positron avalanche resembling that in pulsars. The fast
944: electron-positron flow hits the surface of the star where the
945: released energy is dissipated. The observed very hard spectra of
946: the persistent X-ray emission from SGRs and AXPs (Kuiper et al.
947: 2004, 2006; Mereghetti et al. 2005, Molkov et al. 2005) imply that
948: most of the energy is released in a very hot and tenuous plasma.
949: It was assumed by Thompson \& Beloborodov (2005) and Beloborodov
950: \& Thompson (2007) that the flow loses a significant fraction of
951: its energy in a thin surface layer where the two-stream
952: instability develops so that the plasma is heated by collisionless
953: processes. Collisionless heating is balanced by bremsstrahlung
954: radiation and the equilibrium temperature about 100 keV is
955: achieved.
956:
957: We reanalyse interaction of the fast plasma flow with the surface
958: of the magnetar and conclude that this mechanism is incapable of
959: heating the atmosphere because strong density gradient in the
960: atmosphere of the neutron star suppresses the two-stream
961: instability. We propose, rather, that a hot, tenuous
962: atmosphere/corona could arise due to specific properties of
963: Coulomb scattering in the superstrong magnetic field, mainly due
964: to one-dimensional character of the electron motion. Within the
965: upper layers of the neutron star, the flow energy is transferred,
966: via an electron-positron avalanche, to electrons with the energy
967: something less than the Landau energy ($\sim 1\div 2 $ MeV in the
968: magnetar's magnetic field). This occurs at a significant depth,
969: $\sim 100$ g/cm$^2$, however these electrons are not thermalized
970: but rather escape upwards taking away most of the initial flow
971: energy. The reason is that collisions between electrons in
972: one-dimension do not result in relaxation; after the collision,
973: the two energies of the two electrons are the same as the initial
974: energies. On the other hand, collisions with ions result only in
975: nearly elastic reflection.
976: %and moreover, cross-section is very small for mildly relativistic electrons.}
977: These electrons form a hot atmosphere (the necessary amount of
978: ions accompany the electrons in order to maintain charge
979: neutrality) with the temperature $\sim 1\div 2$ MeV, which is an
980: order of magnitude larger than in the model by Thomson \&
981: Beloborodov (2005) and Beloborodov \& Thompson (2007). Hard
982: radiation from this atmosphere is generated via bremsstrahlung.
983: Pairs are easily produced in this radiation field; they fill the
984: whole magnetosphere forming a hot corona. Collisionless
985: interaction of the primary beam with the pair plasma in the corona
986: heats the pairs even more; they are cooled by Comptonization so
987: that the overall spectrum of the source is a superposition of the
988: bremsstrahlung radiation from the hot atmosphere and a
989: Comptonization radiation from the corona.
990:
991: The extended corona radiates in all directions so that only a
992: small, $<1/2$, fraction of the radiated energy is intercepted by
993: the surface of the star. Therefore the observed luminosity is
994: dominated by the hard radiation. The spectrum is extended to MeV
995: band. Unsaturated Comptonization generates a power-law spectrum,
996: which is generally steeper than the flat bremsstrahlung spectrum
997: therefore radiation from the corona dominates in the range
998: $\lesssim 100$ keV. The photon spectral slope is $1<\alpha<2$, as
999: is observed. The energy release in the magnetosphere occurs
1000: spasmodically at the time scale of at least the light travel time:
1001: the pairs short out the induction electric field in the corona and
1002: the energy release stops until the corona cools down, then the
1003: charge starvation necessitates the displacement current and the
1004: process starts again. Therefore one can expect strong fluctuations
1005: of the radiation at the time scale of $\ge 10^{-4}$s.
1006:
1007: We are grateful to Rashid Shaisultanov for help. Y.L acknowledges
1008: support from the German-Israeli Foundation for Scientific Research
1009: and Development. D.E. acknowledges support from the United States
1010: - Israel Binational Science Foundation and from an Israel Science
1011: Foundation Center of Excellence Grant.
1012:
1013: \section*{Appendix A. Electron-ion collisions: recoil effect}
1014: When the electron with the energy less than $\varepsilon_B$ passes
1015: an ion, the energy is transferred only due to recoil effect. In
1016: magnetar's field, most of the ions are in the ground Landau state
1017: and a scattering occurs if the ion makes a transition to the first
1018: level with the energy
1019: $$
1020: \varepsilon_{Bi}=\frac{ZeB}{m_ic}=5\frac ZAB_{15}\, {\rm keV};
1021: $$
1022: where $m_i=Am_p$ is the ion mass. The cross section for
1023: collisional transitions between the ion Landau levels was found by
1024: Langer (1981); for the transition between the ground and the first
1025: levels it is written as
1026: $$
1027: \sigma=\frac{3B_q}{8B}Z\sigma_T\frac{m_e^2c^2\gamma\gamma'}{pp'}(\ln\Lambda-0.577);
1028: $$
1029: where
1030: $$
1031: \Lambda^{-1}=\frac{B_q}{B}\left[\left(\frac{p-p'}{m_ec}\right)^2-(\gamma-\gamma')^2\right].
1032: $$
1033: The conservation laws imply
1034: $$
1035: m_ec\gamma+m_ic=m_ec\gamma'+\sqrt{m_i^2c^2+(p-p')^2+2\varepsilon_{Bi}m_i};
1036: $$
1037: which yields
1038: $$
1039: \gamma-\gamma'=\varepsilon_{Bi};
1040: $$
1041: and
1042: $$
1043: \Lambda^{-1}=\frac{\varepsilon_{Bi}}{m_ic^2}\left(\frac{m_ec\gamma}p-\frac
1044: 14\right).
1045: $$
1046: As the electron looses only a small fraction of the energy in a
1047: scattering, $\gamma-\gamma'\ll\gamma$, one can conveniently define
1048: the effective cross-section for the energy loss
1049: $$
1050: \tilde{\sigma}\equiv [(\gamma-\gamma')/\gamma]\sigma=2\times
1051: 10^{-3}\frac{Z^2}{A\gamma}[1+0.08\ln(A^2/ZB_{15})]\sigma_T.
1052: $$
1053: It was shown in Sect. 3 that the electron-positron avalanche stops
1054: at the depth (\ref{tau}) forming eventually a cloud of electrons
1055: with the energy $\sim\varepsilon_B/2$. The electron with this
1056: energy loses the fraction of its energy
1057: $$
1058: \frac{\Delta E}E=\tilde{\sigma}n_il=0.1\frac
1059: 1{AB^{1/2}_{15}}[1+0.08\ln(A^2/ZB_{15})]
1060: $$
1061: when rising from the depth (\ref{tau}). This estimate assumes that
1062: the electron moves straightforwardly but not diffuses upwards.
1063: This is justified because the layer of the depth (\ref{tau}) is
1064: transparent for the Coulomb reflection (see the estimate
1065: (\ref{reflection1})). So the fraction of the energy the electron
1066: loses for recoil process is small.
1067: % This cross section should be compared with
1068: %the cross section (\ref{reflection}) for the Coulomb reflection
1069: %from an ion. For the electron with the energy $\varepsilon_B/2$,
1070: %one gets
1071: %$$
1072: %\frac{\tilde{\sigma}}{\sigma_{00}}=0.6\frac{B^{3/2}_{15}}A[1+0.08\ln(A^2/ZB_{15})].
1073: %$$
1074: %One sees that in the hydrogen atmosphere, the electron loses a
1075: %significant fraction of the energy before it is reflected from an
1076: %ion however in the atmosphere of heavy nuclei, the Coulomb
1077: %reflection dominates. In this case, the electron could escape from
1078: %the atmosphere preserving the initial energy.
1079:
1080: \section*{Appendix B. Ionization equilibrium.}
1081:
1082: Ionization equilibrium in the super-strong magnetic field is still
1083: a subject of intense research, see recent reviews by Lai (2001)
1084: and Harding \& Lai (2006) and references therein. Here we present
1085: rough estimates for the helium plasma. The ionization energy of
1086: the hydrogen-like ion is
1087: $$
1088: Q=0.16Z^2\left[\ln\frac{\hbar^3B}{m_e^2e^3cZ^2}\right]^2 {\rm
1089: a.u.}
1090: $$
1091: For the helium ion, one gets $Q=2.39(1+0.086\ln B_{15})$ keV. When
1092: the atom moves, the ionization energy decreases so that the above
1093: value gives the estimate of the ionization temperature from above.
1094:
1095: The ionization equilibrium He$^{++}\rightleftarrows$He$^++e$ is
1096: written as
1097: $$
1098: \frac{n_en_{++}}{n_+}=\frac{Z_eZ_{++}}{Z_+};
1099: $$
1100: where $n_e$, $n_{++}$ and $n_+$ are the number densities of free
1101: electrons, He$^{++}$ and He$^+$, correspondingly, $Z_e$, $Z_{++}$,
1102: and $Z_+$ their partition functions. The partition function of the
1103: strongly magnetized electrons is
1104: $$
1105: Z_e=\frac{eB}{2\pi\hbar
1106: c}\left(\frac{m_eT}{2\pi\hbar^2}\right)^{1/2}. \eqno(B1)
1107: $$
1108: %where $\lambda_C=h/m_ec$ is the Compton wavelength.
1109: The ratio of partition functions of He$^{++}$ and He$^+$ is
1110: dominated by the ionization energy and may be presented as
1111: $$
1112: \frac{Z_{++}}{Z_+}\approx\exp\left(-\frac QT\right).
1113: $$
1114: %$$
1115: %\frac{\alpha_e^2}{1-\alpha_e}=\frac{eB}{\pi\hbar
1116: %c}\left(\frac{m_eT}{2\pi\hbar^2}\right)^{1/2}\exp\left(-\frac
1117: %QT\right).
1118: %$$
1119: Now the temperature of ionization (when $n_{++}=n_+$) is found as
1120: $$
1121: T_{ion}=Q\left\{\ln\left[\frac{3eB}{2\pi n\hbar
1122: c}\left(\frac{m_eT}{2\pi\hbar^2}\right)^{1/2}\right]\right\}^{-1}=
1123: 1.7\times 10^6\frac{(1+0.09\ln
1124: B_{15})^2}{1+0.06\ln\left(B_{15}/\tau g_{14.5}\right)} {\rm K}.
1125: $$
1126: This means that helium is fully ionized in magnetar's atmosphere
1127: with the temperature $T=0.5\div 1$ keV.
1128:
1129: \section*{Appendix C. Comptonization in a superstrong magnetic field}
1130:
1131: Here we demonstrate that Comptonization of soft photons in the
1132: superstrong magnetic field results in a power-law spectrum with
1133: the photon index $\alpha\ge 1$, like in the non-magnetized case.
1134: Compton scattering of soft photons on hot electrons results in a
1135: photon flux over the spectrum from the initial energy,
1136: $\varepsilon_0$, to the spectral region $\varepsilon\sim T$. If
1137: the optical depth of the source is large enough, the photons are
1138: accumulated at $\varepsilon\sim T$ and the equilibrium
1139: Bose-Einstein spectrum, $N_{\rm
1140: BE}=\{\exp[(\eta+\varepsilon)/T]-1\}^{-1}$ is formed.
1141: %if the seed photons gain energy significantly faster than they escape
1142: %from the source so that if the optical depth is large enough.
1143: This regime is called saturated Comptonization.
1144: %In the non-relativistic case, the
1145: %corresponding condition is $\tau^2\gg m_ec^2/T$. In plasmas with
1146: %relativistic temperatures, the photon gains a significant energy
1147: %in one scattering so that the condition for the saturated
1148: %Comptonization is $\tau\gg 1$.
1149: In the medium of the moderate optical depth, photons are not
1150: accumulated but rather escape and therefore a power law spectrum
1151: could be formed in the range $\varepsilon_0\ll\varepsilon\ll T$.
1152: %; Comptonization is unsaturated
1153: Here we show that the same occurs also in the magnetic
1154: field so strong that the
1155: %Landau energy exceeds the electron temperature so that
1156: electrons populate only the ground Landau level. For
1157: nonrelativistic temperatures, Comptonization in the superstrong
1158: magnetic field was studied by Lyubarskii (1987a,b) in the
1159: Focker-Plank approximation. Here we allow relativistic
1160: temperatures but restrict ourselves only to low photon energies
1161: when one can neglect recoil.
1162:
1163: Let us first consider scattering on electrons moving along the
1164: magnetic field with some momentum $p$; the number density of this
1165: electrons is $f(p)dp$; where $f(p)$ is the electron distribution
1166: function. The kinetic equation for photons is easily written in
1167: the proper frame of these electrons as
1168: $$
1169: \left(\frac{\partial}{\partial t'}+c\mathbf{l'}\frac{\partial
1170: }{\partial \mathbf{r}'}\right)
1171: n'(\mathbf{r}',\varepsilon',\mathbf{l}')=cf'dp'\int
1172: d\varepsilon'_1d\Omega'_1\delta(\varepsilon'_1-\varepsilon')
1173: \frac{\partial\sigma}{\partial \Omega'}
1174: [n(\mathbf{r}',\varepsilon'_1,\mathbf{l}'_1)-n(\mathbf{r}',\varepsilon',\mathbf{l}')];
1175: $$
1176: where $\mathbf{l}$ is the direction of propagation of photons,
1177: $n(\mathbf{r},\varepsilon,\mathbf{l})$ the phase density of
1178: photons and prime marks quantities measured in the proper electron
1179: frame. The scattering cross-section is
1180: $$
1181: \frac{\partial\sigma}{\partial \Omega'}=\frac
1182: 3{8\pi}\sigma_T\sin^2\theta'\sin^2\theta'_1;
1183: $$
1184: where $\theta$ and $\theta_1$ is the angles between the photon
1185: direction and the magnetic field before and after the scattering,
1186: correspondingly. One can transform this equation to the laboratory
1187: frame taking into account that the distribution functions are
1188: relativistic invariants, $f'(p')=f(p)$,
1189: $n'(\varepsilon',\mathbf{l}')=n(\varepsilon,\mathbf{l})$, as well
1190: as the expressions $\varepsilon d\varepsilon d\Omega$ and
1191: $\varepsilon (\frac{\partial}{\partial
1192: t}+\mathbf{l}\frac{\partial}{\partial \mathbf{r}})$. Summation
1193: over all electrons yields
1194: $$
1195: \frac{\partial n}{\partial t}+c\mathbf{l}\frac{\partial
1196: n}{\partial \mathbf{r}}=\frac 3{8\pi}c^6\sigma_T\int
1197: [n(\mathbf{r},\varepsilon_1,\mathbf{l}_1)-n(\mathbf{r},\varepsilon,\mathbf{l})]
1198: $$$$
1199: \times\delta[\varepsilon(c-v\cos\theta)-\varepsilon_1(c-v\cos\theta_1)]
1200: \frac{\sin^2\theta\sin^2\theta_1 f(p)dpd\varepsilon_1d\Omega_1}
1201: {\gamma^6(c-v\cos\theta)(c-v\cos\theta_1)^3}.
1202: $$
1203: This equation describes Comptonization of photons with the energy
1204: larger than the energy of the seed photons, $\varepsilon_0$, but
1205: small enough for the recoil effect to be neglected,
1206: $\varepsilon_0\ll\varepsilon\ll \min(T,m_ec^2/T)$.
1207:
1208: In the steady state case, $\partial/\partial t=0$, solution to
1209: this equation has a power law form
1210: $$
1211: n(\mathbf{r},\varepsilon,\mathbf{l})=J(\mathbf{r},\mathbf{l})\varepsilon^{-(2+\alpha)};
1212: $$
1213: where the spatial function $J$ satisfies the equation
1214: $$
1215: \mathbf{l}\frac{\partial J(\mathbf{r},\mathbf{l})}{\partial
1216: \mathbf{r}}=\frac 3{8\pi}c^5\sigma_T\int
1217: \left[\left(\frac{c-v\cos\theta_1}{c-v\cos\theta}\right)^{2+\alpha}J(\mathbf{r},\mathbf{l}_1)-
1218: J(\mathbf{r},\mathbf{l})\right]
1219: \frac{\sin^2\theta\sin^2\theta_1f(p)dpd\Omega_1}
1220: {\gamma^6(c-v\cos\theta)(c-v\cos\theta_1)^4}.\eqno(B1)
1221: $$
1222: The photon power index, $\alpha$, could be found as an eigenvalue
1223: of this equation. It is determined by the electron distribution
1224: function and by the geometry and the optical depth of the source.
1225: It is beyond the scope of the present paper to solve this equation
1226: (solution of a similar problem for a nonmagnetized plasma is given
1227: by Titarchuk \& Lyubarskij (1995)). Let us only demonstrate that
1228: one can expect $\alpha>1$.
1229:
1230: In the infinite homogeneous medium, the left-hand side of Eq.(B1)
1231: is zero therefore one gets the equation
1232: $$
1233: \int \left[(c-v\mu_1)^{2+\alpha}J(\mu_1)-
1234: (c-v\mu)^{2+\alpha}J(\mu)\right]
1235: \frac{(1-\mu^2)(1-\mu^2_1)f(p)dpd\mu_1}
1236: {\gamma^6(c-v\mu)^{3+\alpha}(c-v\mu_1)^4}=0;\eqno(B2)
1237: $$
1238: where $\mu=\cos\theta$. This equation has an evident solution
1239: $\alpha=-2$, $J(\mu)=\it const$, which is nothing more than the
1240: low frequency part of the equilibrium Bose-Einstein spectrum. We
1241: are interested in a solution with a non-zero photon flux over the
1242: spectrum; in the infinite medium, such a solution implies that
1243: except of the soft photon source at $\varepsilon=\varepsilon_0$,
1244: there is a sink at some large enough energy $\varepsilon_{\rm
1245: sink}$; then Eq.(B2) describes the region
1246: $\varepsilon_0<\varepsilon<\varepsilon_{\rm sink}$. In a
1247: non-magnetized plasma, the solution with a non-zero photon flux
1248: over the spectrum is $n\propto \varepsilon ^{-3}$ (Kats,
1249: Kontorovich \& Kochanov 1978) so that the intensity spectrum is
1250: flat. In order to see that the same spectrum ($\alpha=1$ in our
1251: notations) satisfies also Eq.(B2) note that at $\alpha=1$, the
1252: integrand in the left-hand side of this equation is antisymmetric
1253: with respect to exchange $\mu\leftrightarrow\mu_1$. Therefore the
1254: integral from the left-hand side of Eq.(B2) over $\mu$ vanishes
1255: identically. This means that a finite linear homogeneous set of
1256: equations, which could be obtained from Eq.(B2) by discrete
1257: approximation of the integral, is linearly dependent and therefore
1258: it has a nontrivial solution. ({\it The formal proof}. Denote the
1259: linear operator in the left-hand side of Eq. (B2) at $\alpha=1$ as
1260: $\cal L$ and introduce the standard notation for the scalar
1261: product of functions $(\psi,\phi)\equiv\int_{-1}^1\psi\phi d\mu$.
1262: Now one can write that $(e,{\cal L}J)=0$ for $e=const$ and an
1263: arbitrary $J$. Then $({\cal L}^*e,J)=0$ so that $e=const$ is a
1264: nontrivial solution to the conjugate equation ${\cal L}^*e=0$. In
1265: this case, the equation ${\cal L}J=0$ also has by the Fredholm
1266: alternative a nontrivial solution.)
1267:
1268: Thus the spectrum with the slope unity is formed in the infinite
1269: medium; then all photons produced at $\varepsilon_0$ reach
1270: $\varepsilon_{\rm sink}$. In the case of a finite optical depth,
1271: the photons escape so that the spectral photon density should
1272: decrease with the frequency faster than in the infinite medium.
1273: Therefore unsaturated Comptonization in the super-strong magnetic
1274: field generates power law spectra with the slope $\alpha>1$ like
1275: in the non-magnetized plasma.
1276:
1277: \begin{thebibliography}{}
1278:
1279: \bibitem{} Beloborodov, A. M. \& Thompson, C. 2007 ApJ 657, 967
1280: \bibitem{} Breizman B.N. 1990, in: Reviews of Plasma Physics, edited by B.B.Kadomtsev
1281: (Consultants Bureau, New York), Vol. 15, p. 61
1282: \bibitem{} Breizman, B. N. \& Ryutov, D. D. 1974, Nucl.Fusion 14, 873
1283: \bibitem{} Bussard, R. W. 1980, ApJ 237, 970
1284: \bibitem{} Duncan, R.C. \& Thompson, C. 1992, ApJ 392, L9
1285: \bibitem{} Fainberg Y.B., Shapiro V.D. \& Shevchenko V.I. 1970,
1286: Sov.Phys.JETP 30, 528
1287: \bibitem{} Fern\'andes, R. \& Thompson, C. 2007, ApJ 660, 615
1288: %\bibitem{} Godfrey, B.B., Shanahan, W.R. \& Thode, L.E. 1973, Phys.Fluids 18, 346
1289: \bibitem{} Gonthier, P.L., Harding, A.K., Baring, M.G., Costello, R.M. \& Mercer, C.L.
1290: 2000, ApJ 540, 907
1291: \bibitem{} Kaminker, A.D., Pavlov, G. G. \& Mamradze, P. G. 1987, Ap\&SS 138, 1
1292: \bibitem{} Kats, A. V., Kontorovich, V. M. \& Kochanov, A. E. 1978, Ap\&SS 57, 347
1293: \bibitem{} Kotov, Y. D. \& Kelner, S. R. 1985, Sov.Astr.Lett. 11,
1294: 392
1295: \bibitem{} Kuiper, L., Hermsen, W. \& Mendez, M. 2004, ApJ 613, 1173
1296: \bibitem{} Kuiper, L., Hermsen, W., den Hartog, P.R. \& Collmar, W., 2006,
1297: ApJ 645, 556
1298: \bibitem{} Landau, L.D. \& Lifshitz, E.M. 1995, The classical theory
1299: of fields (Butterworth-Heinemann: Oxford [England])
1300: \bibitem{} Langer, S. H. 1981, Phys.Rev.D 23, 328
1301: \bibitem{} Lyubarskii Y.E. 1988a, Astrophysics 28, 106
1302: \bibitem{} Lyubarskii Y.E. 1988b, Astrophysics 28, 253
1303: \bibitem{} Mereghetti, S., G\"otz, D., Mirabel, I. F.\& Hurley, K. 2005, A\&A 433,
1304: L9
1305: \bibitem{} Molkov, S., Hurley, K., Sunyaev, R., Shtykovsky, P., Revnivtsev, M.
1306: \& Kouveliotou, C. 2005, A\&A 433, L13
1307: \bibitem{} Paczynski, B. 1992, Acta Astr 42, 145
1308: \bibitem{} Titarchuk, L. \& Lyubarskij, Y. 1995, ApJ 450, 876
1309: \bibitem{} Thompson, C. \& Beloborodov, A. M. 2005, ApJ 634, 565
1310: \bibitem{} Thompson, C. \& Duncan, R.C. 1995, MNRAS 275, 255
1311: \bibitem{} Thompson, C., Lyutikov, M. \& Kulkarni, S. R. 2002, ApJ 574, 332
1312: %\bibitem{} Tsytovich V.N. \& Shapiro V.D. 1965, Nucl.Fusion 5, 228
1313: \bibitem{} Woods, P.M. \& Thompson, C. 2004, astro-ph/06133
1314:
1315: \end{thebibliography}
1316: \end{document}
1317: