1: \documentclass[12pt,preprint]{emulateapj}

2: 

3: %\usepackage{natbib}

4: %\usepackage{amsmath}

5: 

6: \newcommand{\eps}{\varepsilon}

7: 

8: \begin{document}

9: 

10: \title{Radiative MHD simulation of sunspot structure}

11: \shorttitle{Radiative MHD simulation of sunspot structure}

12: \shortauthors{Rempel, Sch\"ussler \& Kn\"olker}

13: 

14: \author{M. Rempel\altaffilmark{1}, M. Sch\"ussler\altaffilmark{2}  

15:   and M. Kn\"olker\altaffilmark{1}}

16: 

17: \email{rempel@hao.ucar.edu}

18: 

19: \altaffiltext{1}{High Altitude Observatory,

20:     NCAR, P.O. Box 3000, Boulder, Colorado 80307, USA}

21: \altaffiltext{2}{Max-Planck-Institut f\"ur Sonnensystemforschung,

22:     Max-Planck-Str. 2, 37191 Katlenburg-Lindau, Germany}

23: 

24: \begin{abstract}

25:   Results of a 3D MHD simulation of a sunspot with a photospheric size

26:   of about 20~Mm are presented. The simulation has been carried out with

27:   the MURaM code, which includes a realistic equation of state with

28:   partial ionization and radiative transfer along many ray

29:   directions. The largely relaxed state of the sunspot shows a division 

30:   in a central dark umbral region with bright dots and a penumbra showing

31:   bright filaments of about $2$ to $3$ Mm length with central dark lanes. 

32:   By a process similar to the formation of umbral dots, the penumbral

33:   filaments result from magneto-convection in the form of upflow plumes,

34:   which become elongated by the presence of an inclined magnetic field:

35:   the upflow is deflected in the outward direction while the magnetic

36:   field is weakened and becomes almost horizontal in the upper part of the 

37:   plume near the level of optical depth unity. A dark lane forms owing to 

38:   the piling up of matter near the cusp-shaped top of the rising plume

39:   that leads to an upward bulging of the surfaces

40:   of constant optical depth. The simulated penumbral structure

41:   corresponds well to the observationally inferred interlocking-comb

42:   structure of the magnetic field with Evershed outflows along

43:   dark-laned filaments with nearly horizontal magnetic field and

44:   overturning perpendicular (`twisting') motion, which are embedded in a

45:   background of stronger and less inclined field. Photospheric spectral

46:   lines are formed at the very top and somewhat above the upflow plumes,

47:   so that they do not fully sense the strong flow as well as the large

48:   field inclination and significant field strength reduction in the

49:   upper part of the plume structures.

50: \end{abstract}

51: 

52: \keywords{MHD -- convection -- radiative transfer -- sunspots}

53: 

54: %\received{}

55: %\accepted{}

56: 

57: %\maketitle

58: 

59: \section{Introduction}

60: 

61: The physical understanding of sunspot structure has been hampered for

62: decades by 1) insufficient resolution of the fine structure by

63: observations, 2) lack of information about the layers below the visible

64: surface, and 3) insufficient computational power to perform ab-initio 3D

65: MHD simulation of a full sunspot including the surrounding

66: granulation. Recently, we have seen considerable progress on all three

67: of these fronts: 1) adaptive optics, image selection and reconstruction

68: at ground-based telescopes and the advent of spectro-polarimetry in the

69: visible from space with the {\em Hinode} satellite have led to a wealth

70: of new information about the fine structure of sunspot umbrae and

71: penumbrae \citep[e.g.][]{Bharti:etal:2007, Langhans:etal:2007,

72: Ichimoto:etal:2007, Riethmueller:etal:2008, Rimmele:Marino:2006} 2)

73: local helioseismology has started to probe the sub-surface structure of

74: sunspots \citep[e.g.,][]{Cameron:etal:2008}, and 3) the ever-increasing

75: computational power of parallel computers have made ab-initio

76: simulations of full sunspots come into reach.

77: 

78: While \citet{Cameron:etal:2007b} simulated solar pores of up to about

79: 3~Mm diameter and did not find indications for the development of a

80: penumbral structure, the first attempt to simulate a sunspot together

81: with the surrounding granulation is due to

82: \citet{Heinemann:etal:2007}. They considered a rectangular section of a

83: (slab-like) small sunspot of about 4~Mm diameter. The main result of

84: this simulation is the formation of filamentary structures in the outer

85: part of the spot, various properties of which (such as dark cores,

86: outflows, and strongly inclined magnetic field) are consistent with

87: observational results.  However, the filaments found by

88: \citet{Heinemann:etal:2007} are much shorter than the typical lengths of

89: real penumbral filaments and the overall extension of the simulated

90: penumbra is very small.

91: 

92: Here we report about results of a sunspot simulation with the {\em

93: MURaM} code \citep{Voegler:etal:2005}. The simulated sunspot has a

94: total diameter of about 20~Mm, shared about equally by umbra and

95: penumbra. 

96: 

97: \section{Simulation setup}

98: 

99: \begin{figure*}

100:   \centering 

101:   \resizebox{\hsize}{!}{\includegraphics{f1.eps}}

102:   \caption{Continuum intensity image at 630~nm of the simulated sunspot

103:     and its environment (doubled in the $y$-direction). The bright

104:     umbral dots and penumbral filaments have peak intensities between

105:     40\% and 90\% of the average value outside the spot. The penumbral

106:     filaments reach lengths of 2--3 Mm. The white frame indicates the

107:     filament studied in detail in Sec.~3.2.}

108:   \label{fig:intensity_global}

109: \end{figure*}

110: 

111: \begin{figure*}

112:   \centering 

113:   \resizebox{\hsize}{!}{\includegraphics{f2.eps}}

114:   \caption{Vertical structure of the simulated sunspot. Shown is a color

115:   coding of the square root of the field strength on a cut through the

116:   simulation in $x$-$z$ plane at $y=3.84\,$Mm. White color corresponds

117:   to the maximum field strength of 9$\,$kG, black to zero field. The

118:   whitish line indicates the height of the level $\tau_{630}=1.$ The

119:   average Wilson depression of the spot umbra is about 450~km. Separated

120:   from the main sunspot, two pore-like structures have formed at

121:   $x\simeq31\,$Mm and $x\simeq35\,$Mm, respectively.}

122:   \label{fig:bb_vert_global}

123: \end{figure*}

124: 

125: \begin{figure}

126:   \centering 

127:   \resizebox{\hsize}{!}{\includegraphics{f3a.ps}}

128:   \resizebox{\hsize}{!}{\includegraphics{f3b.ps}}

129:   \caption{Large-scale structure of the simulated sunspot: quantities

130:   averaged over the horizontal $y$-direction as functions of $x$.  {\em

131:   Upper panel:} vertical magnetic field, $B_z$, at $\tau_{630}=0.1$

132:   (solid), modulus of the horizontal field component, $|B_x|$, at

133:   $\tau_{630}=0.1$ (dashed), and normalized continuum intensity at 630~nm

134:   (dotted). {\em Lower panel:} magnetic field inclination with respect

135:   to the vertical at $\tau_{630}=0.1$ (solid) and horizontal velocity,

136:   $v_x$, at $\tau_{630}=1.$ (dashed).}

137:   \label{fig:means_global}

138: \end{figure}

139: 

140: 

141: The primary numerical challenges of performing a large-scale sunspot

142: simulation including umbra, penumbra and granulation are the significant

143: variations of $\beta=8\pi p/B^2$ and Alfv{\'e}n velocity encountered at a

144: given geometrical height.  While a very low value of $\beta$ primarily

145: imposes stability problems, very high Alfv{\'e}n velocities of more that

146: $1000$~km$\cdot$s$^{-1}$ above the umbra of a spot lead to unacceptably

147: small time steps for an explicit code. In order to cope with these

148: problems, we have modified the {\em MURaM} code as outlined in

149: Appendix~\ref{appendix}.

150: 

151: The basic concept of our simulation is similar to that of

152: \citet{Heinemann:etal:2007}: we consider a slender rectangular section

153: of a sunspot. However, our computational domain is considerably wider

154: and deeper than theirs, permitting us to simulate a much larger sunspot.

155: The rectangular box spans $36.864\,{\rm Mm}\times 4.608\,{\rm Mm}$ in

156: the horizontal $(x,y)$ directions and $6.144$ Mm in the vertical $(z)$

157: direction. The side boundaries are periodic. The top boundary is closed

158: for the flow and the magnetic field is matched to a potential field

159: above.  At the bottom, the boundary is open for the flow and the

160: magnetic field kept vertical \citep[for details,

161: see][]{Voegler:etal:2005}.  In regions of strong magnetic field

162: ($B>1000$~G), the bottom boundary is closed to avoid outflow

163: instabilities in long runs. The radiative transfer is treated in the

164: grey approximation.

165: 

166: The simulation was started from a thermally relaxed non-magnetic run by

167: introducing a 2D (slab) field configuration in the $x$-$z$ plane with a

168: width of $5$ Mm and a strength of $10$ kG at the bottom of the box,

169: expanding to $15$ Mm at the top.  Using a moderate grid resolution of

170: $48\,{\rm km}\times48\,{\rm km}\times32\,{\rm km}$ we run the

171: simulation for about $12$ hours solar time. After a very dynamic

172: adjustment phase of about $2$ hours, elongated filaments with dark cores

173: of about $2$ to $3$ Mm length started forming in the periphery of the

174: umbra, their heads moving inward. After about $7.5$ hours of evolution we

175: restarted from a snapshot and increased the resolution to $32\,{\rm

176: km}\times32\,{\rm km}\times21.33\,{\rm km}$.

177: 

178: \section{Results}

179: \begin{figure}

180:   \centering 

181:   \resizebox{\hsize}{!}{\includegraphics{f4.eps}}

182:   \caption{Continuum intensity image showing details of penumbral

183:     filaments.  The vertical lines indicate the positions of the 

184:     vertical cuts presented in Fig.~\ref{fig:cuts_color}.}

185:   \label{fig:intensity_detail}

186: \end{figure}

187: 

188: \begin{figure}

189:   \centering 

190:   \resizebox{\hsize}{!}{\includegraphics{f5.eps}}

191:   \caption{Vertical cuts in the $y$-$z$ plane of magnetic field strength

192:     (left), inclination angle (middle) and horizontal velocity (right)

193:     for the positions indicated in Fig.~\ref{fig:intensity_detail} (the top

194:     row corresponding to the leftmost cut). The magnetic field strength is

195:     saturated at $4\,$kG, the velocity at $2\,$km$\cdot$s$^{-1}$. The

196:     filaments correspond to about $2\,$Mm deep channels of weaker and

197:     almost horizontal field and an outflow of a few km$\cdot$s$^{-1}$. 

198:   }

199:   \label{fig:cuts_color}

200: \end{figure}

201: 

202: \begin{figure}

203:   \centering 

204:   \resizebox{\hsize}{!}{\includegraphics{f6.eps}}

205:   \caption{Horizontal and vertical profiles of various physical

206:     quantities. {\em Top panel:} Horizontal cut along the middle line in

207:     Fig.~\ref{fig:intensity_detail} at height $z=5$ Mm, near the

208:     $\tau=1$ level in the filament. Shown are the vertical field

209:     component ($B_z$, black), the horizontal field component ($B_x$,

210:     blue), and the horizontal velocity component ($v_x$, green). The

211:     vertical field drops to very low values in the filaments, resulting

212:     in an almost horizontal field of $1-1.5\,$kG. Along the centers of

213:     the filaments, we have an outward directed horizontal flow. {\em Middle

214:     panel:} Vertical profiles of the same quantities through the center

215:     of the second filament from the left in the top panel. In addition,

216:     the red line shows the temperature profile. The horizontal outflow

217:     peaks close to the $\tau=1$ level, where the magnetic field has the

218:     largest inclination. {\em Bottom panel:} Vertical profiles at the

219:     center of the fourth filament from the left in the top panel.}

220:   \label{fig:profiles}

221: \end{figure}

222: 

223: \begin{figure*}

224:   \centering 

225:   \resizebox{\hsize}{!}{\includegraphics{f7a.eps}

226: 	               {\includegraphics{f7b.eps}}}

227:   \resizebox{\hsize}{!}{\includegraphics{f7c.eps}

228: 	               {\includegraphics{f7d.eps}}}

229:   \resizebox{\hsize}{!}{\includegraphics{f7e.eps}

230: 	               {\includegraphics{f7f.eps}}}

231:   \resizebox{\hsize}{!}{\includegraphics{f7g.eps}

232: 	               {\includegraphics{f7h.eps}}}

233:   \caption{Grey-scale maps of physical quantities on the surface

234:     $\tau_{630}=0.1$ for the filament indicated by the rectangular box

235:     in Fig.~\ref{fig:intensity_global}. The ranges between minimum

236:     (black) and maximum (white) of the various quantities

237:     are: $|B_x|$: 670$\,$G ... 1970$\,$G; $B_y$: $-940\,$G ... 640$\,$G;

238:     $B_z$: 980$\,$G ... 3050$\,$G; $B$ inclination with respect to

239:     vertical: 17.5$\,$deg ... 61$\,$deg; $-v_x$:

240:     $-1.4\,$km$\cdot$s$^{-1}$ ...  3.3$\,$km$\cdot$s$^{-1}$; $v_y$:

241:     $-0.95\,$km$\cdot$s$^{-1}$ ...  0.66$\,$km$\cdot$s$^{-1}$; $v_z$:

242:     $-2.1\,$km$\cdot$s$^{-1}$ ...  1.5$\,$km$\cdot$s$^{-1}$; $I_{630}$:

243:     0.13 ... 1.02 of the average value outside the spot.}

244:   \label{fig:filament_hor}

245: \end{figure*}

246: 

247: \begin{figure}

248:   \centering 

249:   \resizebox{\hsize}{!}{\includegraphics{f8a.ps}}

250:   \resizebox{\hsize}{!}{\includegraphics{f8b.ps}}

251:   \caption{Profiles of the magnetic field and velocity components at

252:            optical depth $\tau_{630}=0.1$ along a cut perpendicular to

253:            the filament shown in Fig.~\ref{fig:filament_hor} (cut in

254:            $y$-direction at $x=2\,$arcsec). {\em Upper panel:} $v_x$

255:            (solid), $v_y$ (dashed), $v_z$ (dotted). {\em Lower panel:}

256:            $B_x$ (solid), $B_y$ (dashed), $B_z$ (dotted).  }

257:   \label{fig:B_v_profiles}

258: \end{figure}

259: 

260: \begin{figure*}

261:   \centering 

262:   \resizebox{\hsize}{!}{

263:     \includegraphics{f9a.ps}

264:     \includegraphics{f9b.ps}

265:     \includegraphics{f9c.ps}}

266:   \resizebox{\hsize}{!}{

267:     \includegraphics{f9d.ps} 

268:     \includegraphics{f9e.ps}

269:     \includegraphics{f9f.ps}} 

270:   \caption{Vertical cuts at $x=2\,$arcsec perpendicular through the

271:    filament shown in Fig.~\ref{fig:filament_hor}. The vertical scale is

272:    the same as used in Figs.~\ref{fig:cuts_color} and 

273:    \ref{fig:profiles}, where zero height corresponds to the bottom of 

274:    the computational domain. The

275:    grey-scale ranges between minimum (black) and maximum (white) of the

276:    quantities shown are: $|B|$: 480$\,$G ... 4220$\,$G; $B$ inclination

277:    with respect to vertical: 20.1$\,$deg ... 90$\,$deg; $T$:

278:    $3890\,$K ... $15410\,$K; $-v_x$:

279:    $1.88\,$km$\cdot$s$^{-1}$ ...  2.09$\,$km$\cdot$s$^{-1}$; $v_y$:

280:    $-0.7\,$km$\cdot$s$^{-1}$ ...  0.8$\,$km$\cdot$s$^{-1}$; $v_z$:

281:    $-2.3\,$km$\cdot$s$^{-1}$ ...  2.2$\,$km$\cdot$s$^{-1}$ . 

282:    The white lines indicate the levels of 

283:    $\tau_{630}=1.$, 0.1, and 0.01, respectively.}

284:   \label{fig:filament_vertx}

285: \end{figure*}

286: 

287: \begin{figure*}

288:   \centering 

289:   \resizebox{\hsize}{!}{\includegraphics{f10a.ps}

290: 	               {\includegraphics{f10b.ps}}}

291:   \resizebox{\hsize}{!}{\includegraphics{f10c.ps}

292: 	               {\includegraphics{f10d.ps}}}

293:   \caption{Vertical cuts along the center of the filament 

294:    at constant $y=0.9\,$arcsec in Fig.~\ref{fig:filament_hor}.

295:    The grey-scale ranges between

296:    minimum (black) and maximum (white) of the

297:    quantities shown are: $|B|$: 480$\,$G ... 4220$\,$G; $B$ inclination

298:    with respect to vertical: 20.1$\,$deg ... 90$\,$deg; $-v_x$:

299:    $-1.3\,$km$\cdot$s$^{-1}$ ...  3.6$\,$km$\cdot$s$^{-1}$; $v_z$:

300:    $-2.7\,$km$\cdot$s$^{-1}$ ...  2.5$\,$km$\cdot$s$^{-1}$.}

301:   \label{fig:filament_vertz}

302: \end{figure*}

303: 

304: \begin{figure}

305:   \centering 

306:   \resizebox{\hsize}{!}{\includegraphics{f11.eps}}

307:   \caption{Schematic illustration of the basic structure of penumbral

308:   filaments as suggested by \citet{Zakharov:etal:2008} on the basis of

309:   spectro-polarimetric observations. Shown is a vertical cut

310:   perpendicular to the filament axes. The bright semicircular areas

311:   indicate the uppermost part of the upflow plume, the curved arrows

312:   represent the overturning convective flow. Circled crosses indicate the

313:   almost horizontal magnetic field and velocity along the filaments. The

314:   lines in the grey area are projections of the less inclined magnetic

315:   field lines outside the filaments. The intrusion of the plume and flux

316:   transport by the overturning motion have pushed aside the less inclined

317:   field between the filaments.}

318:   \label{fig:sketch}

319: \end{figure}

320: 

321: For the detailed analysis of the simulation results, we consider a

322: snapshot taken $30$ solar minutes after the start of the high-resolution

323: run. We first present results concerning the global structure of the

324: simulated spot and its average properties and then consider the

325: penumbral filaments.

326:  

327: \subsection{Large-scale structure}

328: 

329: Fig.~\ref{fig:intensity_global} presents a continuum intensity image

330: ($\lambda=630\,$nm) of the sunspot and its environment. For better

331: visibility, the snapshot has been doubled in $y$-direction.  The spot

332: umbra has a width of about $10$ Mm and is surrounded on both sides by a

333: penumbra of about $4-5$ Mm width, harboring filaments with dark

334: cores. The umbra shows umbral dots similar to those found by

335: \citet{Schuessler:Voegler:2006} in a local simulation.

336: 

337: Fig.~\ref{fig:bb_vert_global} displays the square root of the magnetic

338: field strength on a vertical cut through the simulation box at

339: $y=3.84$~Mm. The height expansion of the sunspot due to the decreasing

340: gas pressure is well visible. Comparison with

341: Fig.~\ref{fig:intensity_global} shows that the regions corresponding to

342: the penumbra (roughly in the ranges $x=9...14\,$Mm and

343: $x=23...28\,$Mm) mostly have a deep underlying magnetic structure. In

344: fact, about the same amount of vertical flux emerges in the penumbral

345: part of the simulated spot as in the umbra, so that the penumbra cannot

346: be shallow \citep[cf.][]{Solanki:Schmidt:1993}.

347: 

348: Fig.~\ref{fig:bb_vert_global} also indicates that the magnetic field is

349: rather inhomogeneous between the level $\tau_{630}=1$ (indicated by the

350: white line) and about 2~Mm below. This is caused by magneto-convection

351: in the form of upflow plumes which lead to reduced field strength owing

352: to expansion and flux transport by the overturning flow. As we shall

353: discuss further below, the underlying processes for the formation of

354: penumbral filaments are the same as those for the umbral dots

355: \citep{Schuessler:Voegler:2006}. Above $\tau_{630}=1$, the field again

356: becomes rather smooth: owing to the low plasma $\beta$ in these layers

357: overlying umbra and penumbra, gas pressure gradients cannot maintain

358: strong field inhomogeneities and the field structure has to become

359: almost force-free. 

360: 

361: The sunspot has already lost some amount of flux to its environment,

362: which shows a plage-like mean vertical flux density of about 200~G at

363: $\tau_{630}=1$. Most of this flux has been assembled in small-scale flux

364: concentrations in the intergranular lanes (bright structures in

365: Fig.~\ref{fig:intensity_global}), but also pore-like structures have

366: formed, probably supported by the periodic boundary condition.

367: 

368: A more quantitative account of the large-scale structure of the

369: simulated sunspot is provided by Fig.~\ref{fig:means_global}, which

370: shows horizontal profiles of average quantities determined in the layers

371: accessible to observation. The profiles of the magnetic field components

372: and its inclination with respect to the vertical agree well with the

373: observational curves determined by \citet{Keppens:Martinez:1996},

374: although the field strength in the simulated penumbra is somewhat

375: high. Average outward horizontal flows with peak velocities of about

376: $2\,$km$\cdot$s$^{-1}$ are clearly visible in both penumbral

377: regions. The main discrepancy between the observed and simulated

378: penumbra is the lower mean brightness in the simulation (about 60\% of

379: the quiet Sun) compared to the observed value of about 75\%

380: \citep[e.g.,][]{Schlichenmaier:Solanki:2003}. Guided by our experience

381: with umbra simulations, we conjecture that this discrepancy could be due

382: to insufficient thermal relaxation of the magnetic region, so that the

383: magneto-convective processes providing the structuring and the energy

384: transport in the penumbra are still not completely developed. 

385: 

386: \subsection{Penumbral filaments}

387: Fig.~\ref{fig:intensity_detail} shows an enlargement of the penumbral

388: region on the right-hand side of the continuum intensity map

389: (Fig.~\ref{fig:intensity_global}). The most conspicuous structures are

390: dark-cored filaments of up to a few Mm length. On their end facing the

391: umbra, the filaments typically show bright `heads', which propagate into

392: the umbra. The typical lifetime of the filaments is about one hour. The

393: three lines in Fig.~\ref{fig:intensity_detail} indicate the positions of

394: vertical cuts shown in Fig.~\ref{fig:cuts_color}, which give the

395: magnetic field strength, $B$, the inclination angle with respect to

396: the vertical, ${\arcsin}(B_x/\vert B\vert)$, and the horizontal flow

397: velocity, $v_x$.

398: 

399: It is obvious from Fig.~\ref{fig:cuts_color} that the penumbral

400: filaments correspond to slender structures of about $2$ Mm depth (near

401: the umbra) and a few hundred km width with significantly reduced overall

402: field strength. The vertical field component, $B_z$, is reduced to close

403: to zero near the top of the filament, so that an inclination angle of

404: almost $90\deg$ results, i.e., the field becomes nearly horizontal. In

405: the inner penumbra, the filaments develop within regions of strong

406: magnetic field; only in the periphery of the spot, where the vertical

407: thickness of the penumbra drops to less than $2$ Mm, the filaments

408: become more connected to the almost field-free convection zone

409: beneath. Therefore, the filaments in this simulation do not originate

410: from convection penetrating into the penumbra from beneath; they are

411: rather similar to the plume structures leading to umbral dots in the

412: simulation of \citet{Schuessler:Voegler:2006}, here modified by the

413: presence of a significant horizontal field component. 

414: 

415: The third column in Fig.~\ref{fig:cuts_color} shows that the penumbral

416: filaments are associated with strong horizontal flows away from the

417: umbra. These flows reach their largest velocities of a few

418: km$\cdot$s$^{-1}$ in the upper and outer parts of the filaments, where

419: the magnetic field is most strongly inclined.

420: 

421: Fig.~\ref{fig:profiles} provides a detailed view of the physical

422: conditions in the filaments by showing horizontal and vertical profiles

423: of magnetic field and velocity components for the cut shown in the

424: middle row of Fig.~\ref{fig:cuts_color}. The horizontal profiles at

425: $z=5$ Mm in the top panel illustrate the connection between horizontal

426: outflow and almost horizontal magnetic field. The vertical cuts at the

427: positions $y=1.94$ Mm and $y=2.6$ Mm, respectively, shown in the middle

428: and bottom panels of Fig.~\ref{fig:profiles} indicate that the outflows

429: are concentrated in the near-surface layers.  It is tempting to

430: associate these flows with the Evershed effect

431: \citep[cf.][]{Scharmer:etal:2008}.

432: 

433: In what follows, we will study in more detail one prototypical filament

434: (outlined by the white square in Fig.~\ref{fig:intensity_global}).

435: Fig.~\ref{fig:filament_hor} shows a magnified continuum intensity image

436: at 630~nm (bottom left panel) together with maps of various quantities

437: calculated at the surface $\tau_{630}=0.1$, roughly corresponding to the

438: layer dominating the spectro-polarimetric information obtainable with

439: the often used neutral iron lines at 630.15~nm and 630.25~nm. This gives

440: a first idea of the actually observable consequences of the physical

441: processes underlying the penumbral filamentation, although definite

442: results will require a detailed comparison with synthetic Stokes

443: profiles from a non-grey simulation, also taking into account image

444: degradation by a realistic point-spread function and noise. Such a study

445: is beyond the scope of this paper.

446: 

447: The intensity image in Fig.~\ref{fig:filament_hor} shows a bright

448: filament of a few hundred km width, pervaded by a dark lane of about

449: 100~km width. The head of the filament has moved some way into the umbra

450: and has nearly disconnected from the filament, forming a peripheral

451: umbral dot. The maps at constant optical depth $\tau_{630}=0.1$ show an

452: upflow ($v_z>0$) of up to 1.5 km$\cdot$s$^{-1}$ at the center of the

453: filament, which is connected to downflows at both sides via outflows in

454: $\pm y$-direction, perpendicular to the filament axis. In combination

455: with the longitudinal flow along the filament, such a flow pattern is

456: consistent with the recent observations of `twisting motions' in

457: penumbral filaments \citep{Ichimoto:etal:2007,Zakharov:etal:2008}.  The

458: strong horizontal flow away from the umbra ($-v_x>0$) along the dark

459: lane is associated with a weaker, laterally more extended, inward return

460: flow at the periphery of the filament. The profiles of the three

461: velocity components along a cut in $y$-direction through the filament at

462: $x=2\,$arcsec shown in the upper panel of Fig.~\ref{fig:B_v_profiles}

463: clearly reveal the flow pattern of a rising plume with laterally

464: overturning motion and a strong longitudinal outflow.

465: 

466: The maps of the magnetic field components in Fig.~\ref{fig:filament_hor}

467: and the corresponding profiles along the perpendicular cut at

468: $x=2\,$arcsec (lower panel of Fig.~\ref{fig:B_v_profiles}) show a

469: reduction of the vertical component, $B_z$, while the component along

470: the filament, $B_x$, stays almost constant. Accordingly, the field is

471: more strongly inclined and its strength reduced in the filament.  The

472: component perpendicular to the filament axis reverses sign at the

473: filament center (above the dark lane), corresponding to the cusp-shaped

474: geometry of the top part of the filament: the strong field sideways of

475: the filament expands and `closes' above the filament. This configuration

476: has recently been confirmed observationally by

477: \citet{Borrero:etal:2008}.

478: 

479: The structure below the visible filament and its relationship to the

480: surfaces of constant optical depth that are relevant for observations

481: are illustrated by vertical cuts perpendicular to and along the

482: filament, which are shown in Figs.~\ref{fig:filament_vertx} and

483: \ref{fig:filament_vertz}, respectively. The perpendicular cuts at

484: $x=2\,$arcsec (Fig.~\ref{fig:filament_vertx}) reveal a structure which

485: is very similar to that underlying the umbral dots in the simulations of

486: \citet{Schuessler:Voegler:2006}: a strong upflow plume in the center

487: with narrow downflows at its sides leads to an elevation of the

488: iso-$\tau$ surfaces by about 200~km, also clearly visible in the

489: temperature profile. The magnetic field is strongly reduced in the plume

490: since 1) plasma moving upward expands owing to the strong pressure

491: stratification and 2) the overturning horizontal flow transports flux to

492: the downflow regions at the edges of the filament. The latter effect

493: preferentially weakens the vertical field component while horizontal

494: field is replenished by the upflow. The presence of the strongly

495: inclined magnetic field deflects the central upflow outward, leading to

496: a horizontal flow of about $2\,$km$\cdot$s$^{-1}$ away from the umbra of

497: the spot.

498: 

499: There is some indication that occasionally part of the overturning flow

500: is recirculated into the upflow plume, so that the flow pattern becomes

501: reminiscent to roll convection as originally suggested by

502: \citet{Danielson:1961}. To see this, consider the velocity components

503: shown in the lower panels of Fig.~\ref{fig:filament_vertx}: the central

504: upflow ($v_z$) drives perpendicular outflows ($v_y$) near the top of the

505: filament, which feed the downflows adjacent to it. About 300~km deeper,

506: the sign of $v_y$ is reversed, so that we now have a perpendicular

507: inflow, which converges with the central upflow and closes the roll. It

508: is not clear at this moment 1) whether this is a robust or a transient

509: feature of the flow pattern, 2) whether it is really roll-type

510: convection or more related to Kelvin-Helmholtz instability of the

511: downflow, and 3) whether it is realistic at all since the simulation

512: still has a much larger effective magnetic diffusivity than the plasma

513: in a real sunspot. Anyway, its rather small depth extension indicates

514: that such roll convection probably does not contribute much to the convective

515: energy transport, which is mainly provided by the deep upflow plume. 

516: 

517: 

518: Fig.~\ref{fig:filament_vertz} shows physical quantities on a vertical

519: cut roughly along the dark lane. It reveals that the structure

520: underlying the visible filament is extended in depth as far as the

521: upflows and the reduction of the field strength are concerned.  On the

522: other hand, the horizontal outflow ($v_x$) and the field inclination

523: sharply peak near optical depth unity and lead to the formation of a

524: narrow, almost horizontal flow channel.

525: 

526: Altogether, the structure of the simulated penumbral filaments may well

527: be represented by the sketch shown in Fig.~\ref{fig:sketch}, which

528: illustrates the essence of the high-resolution spectro-polarimetric

529: observations of \citet{Zakharov:etal:2008}.

530: 

531: A direct connection of the Evershed flow with the basic

532: magneto-convective process in the penumbra has been suggested by

533: \citet{Scharmer:etal:2008} on the basis of the simulations of

534: \citet{Heinemann:etal:2007}.  In fact, we find a dominant outward flow

535: due to the deflection of the upward flow in the central part of the

536: filament, which is turned outward by the presence of the inclined

537: magnetic field. The corresponding return flow is mainly located in the

538: regions with less inclined and stronger magnetic field between the

539: filaments (see Figs.~\ref{fig:filament_hor} and \ref{fig:B_v_profiles});

540: it therefore has a smaller horizontal component at the same (optical)

541: depth. Consequently, the observable average horizontal velocity is

542: outward (see the lower panel of Fig.~\ref{fig:means_global}) in

543: accordance with the observed Evershed effect, even though there is no

544: significant overall net outward mass flux in the penumbra. This possibly

545: resolves the long-standing problem of the source and disposal of the

546: mass transported by the Evershed flow

547: \citep[e.g.,][]{Solanki:etal:1994}. The existence of the weaker inward

548: return flows is a prediction of the simulation that should be testable

549: with high-resolution observations. Note that \citet{Ichimoto:etal:2008}

550: have found first indications for flows between the Evershed flow

551: channels. In our current simulation, the penumbral filaments appear to

552: be more separated from each other than shown by observations. While the

553: return flow is clearly visible around separated individual filaments,

554: there are indications that it might be less easily detectable in the

555: case of more densely packed filaments. This needs to be investigated in

556: the future with simulations carried out at higher spatial resolution.

557: 

558: The hot upflow extends over most of the length of the filaments, which

559: provides a rather efficient energy transport to supply the radiative

560: losses of the penumbra. In contrast, the moving-flux-tube model with

561: only localized upflows is far too inefficient to explain the average

562: brightness of the penumbra

563: \citep{Schlichenmaier:Solanki:2003}. Similarly, the localized plumes

564: underlying umbral dots can only maintain the much lower brightness of

565: the umbra in comparison with the penumbra, although the basic

566: magneto-convective process is very similar.

567: 

568: 

569: The location of the iso-$\tau$ lines in Figs.~\ref{fig:filament_vertx}

570: and \ref{fig:filament_vertz} demonstrates that the main part of the

571: upflow plume and the overturning motion underlying the filament is below

572: the visible layers. Spectro-polarimetric observations corresponding to

573: $\tau_{630}\simeq 0.1\dots0.01$ just scratch the `tip of the iceberg'

574: and reveal only the uppermost part of the Evershed flow and magnetic

575: cusp structure, which is already laterally much more homogeneous than

576: the underlying main part of the filament.  This situation is very

577: similar to that of umbral dot observations and in the past probably has

578: led, together with the effect of insufficient spatial resolution, to

579: ambiguities in the interpretation of spectroscopic observations of

580: velocity and magnetic field.

581: 

582: 

583: \section{Discussion}

584: 

585: The properties of our simulated sunspot are consistent with the general

586: picture of sunspot structure that has emerged from observational

587: studies. This applies to the overall structure (e.g., distinction

588: between umbra and penumbra, average `radial' profiles of the magnetic

589: field components and inclination angle, average outflow in the penumbra)

590: as well as to the detailed properties of the fine structure of umbra and

591: penumbra. The penumbral filamentation results from magneto-convective

592: energy transport in the form of hot rising plumes, very similar to the

593: process giving rise to umbral dots \citep{Schuessler:Voegler:2006}. The

594: inclined magnetic field near the periphery of the spot causes a symmetry

595: breaking which leads to elongated filaments with strong outflows along

596: flow tubes of nearly horizontal field near optical depth unity. In

597: addition to the flow along the filament, the upflow also turns over into

598: a motion perpendicular to the filament axis. Dark lanes appear above the

599: strongest upflows owing to the upward bulging of the surface of optical

600: depth unity and the piling up of plasma in a cusp-shaped region at the

601: top of the filament, above which the less inclined field outside the

602: filament becomes laterally fairly homogeneous. The horizontal outflows

603: are concentrated along the dark lanes. All these properties are

604: consistent with recent observational results

605: \citep[e.g.,][]{Bellot-Robio:etal:2005, Rimmele:Marino:2006,

606: Langhans:etal:2007, Ichimoto:etal:2007, Borrero:etal:2008,Noort:Rouppe:2008,

607: Zakharov:etal:2008}.

608: 

609: Our results are also consistent with many properties of the short

610: penumbral filaments found by \citet{Heinemann:etal:2007}, who gave

611: interpretations along very similar lines. The fact that simulations with

612: two rather different numerical codes lead to basically the same picture

613: for the formation of penumbral structure indicates that, in spite of all

614: differences in detail, the simulations have indeed captured 

615: essential physical processes.  The explanation for the Evershed effect

616: as a natural consequence of rising plumes in an inclined field

617: \citep[cf.][]{Scharmer:etal:2008} connects this flow directly to the

618: basic magneto-convective structure of the penumbra, so that it should

619: occur whenever penumbral structure is present.  The geometry of the flow

620: pattern is such that the observable average outflow velocity need not to

621: be connected with a net outflux of mass. This does not exclude the

622: additional presence of siphon flows \citep[e.g.,][]{Degenhardt:1993,

623: Montesinos:Thomas:1997, Solanki:etal:1994}, but it is much less clear

624: whether the pressure gradients required for a sustained outward siphon

625: flow are maintained always and everywhere in all penumbrae.

626: 

627: What can be said on the basis of our simulation results about the

628: various models that have been proposed to explain the penumbral

629: structure? First of all, we do not see evidence for the `moving flux

630: tube' model and interchange convection

631: \citep[e.g.][]{Schlichenmaier:etal:1998b, Schlichenmaier:etal:1998a}.

632: In our simulations, the progression of the filament heads toward the

633: umbra during their formation phase is not caused by the inward motion of

634: a narrow flux tube, but rather due to the expansion of the sheet-like

635: upflow plumes along the filament.

636: 

637: Furthermore, the simulations combine various aspects of the `embedded

638: flux tube' model \citep{Solanki:Montavon:1993, Borrero:etal:2005,

639: Borrero:etal:2006} and the `gappy penumbra' model

640: \citep{Spruit:Scharmer:2006,Scharmer:Spruit:2006}. However, the

641: simulated penumbral filaments are neither intrusions of field-free

642: plasma from below nor are they confined to almost horizontal flux tubes

643: disconnected from their environment, particularly in depth. The

644: uppermost part of the plume structure forming the penumbral filament

645: with its strong horizontal flow and almost horizontal field could

646: possibly be represented by a kind of embedded flux tube. The main

647: feature missing in the embedded flux tube models is the overturning

648: convection within the filament and the deep-reaching upflow of plasma

649: that provides the primary energy supply. In this respect, the underlying

650: plume structure with its reduced (albeit non-vanishing) field strength

651: has much similarity with the `gappy' configuration of the penumbra. This

652: scenario captures the convective origin of the penumbral filamentation,

653: even though the gaps form within the strong magnetic field. As a

654: consequence, the gaps contain a horizontal field and, in most cases, are

655: not connected to the almost field free convecting plasma below the

656: penumbra.

657: 

658: There is no clear evidence in our simulations that the penumbral

659: structure is affected or even caused by the fluting instability as

660: suggested by \citet{Weiss:etal:2004}. However, the periodic boundary

661: condition in the horizontal ($x$) direction used in the simulation

662: implicitly corresponds to the existence of identical sunspots just

663: about 20~Mm from the penumbral boundaries, which certainly affects the

664: field structure, particularly the inclination, in the outer

665: penumbra. Therefore, the simulation possibly does not well represent the

666: convective pumping effect suggested by \citet{Weiss:etal:2004} as a

667: mechanism for the downward dragging of magnetic flux in the outer

668: penumbra. This might provide a possible explanation for the still rather

669: small extension of the simulated penumbra. Observations in fact indicate

670: that penumbrae are often suppressed on the side of a sunspot which faces

671: a nearby spot of the same polarity.

672: 

673: Altogether, our results indicate a new level of realism in the

674: theoretical modelling of sunspot structure. The properties of the

675: simulated penumbral filaments are consistent with a variety of

676: observational results and provide a basis for a physical understanding

677: of umbral and penumbral structure in terms of magneto-convective

678: processes. On the other hand, there are still clear discrepancies

679: between the numerical results and real sunspots, so that there is some

680: way ahead to be covered towards a completely satisfactory model. Our

681: penumbral structure does not yet appear to be fully evolved and the

682: overall extension of the penumbra is still somewhat small. The average

683: intensity profile indicates that we have simulated the development of an

684: inner penumbra, while the outer penumbra might be more strongly affected

685: by convective pumping \citep{Weiss:etal:2004}. 

686: 

687: The lower boundary condition remains arbitrary since we still have no

688: reliable observational constraints concerning the subsurface structure

689: of sunspots.  Computational limits have forced us to use a rather coarse

690: spatial resolution of 32~km and a fairly small computational

691: box. Furthermore, we could only cover a relatively short overall

692: evolution time. As a consequence, the effective diffusivities in the

693: simulation are still much larger than the real values. Test calculations

694: with different resolution show that 1) first indications for filamentary

695: structure appear already at a horizontal resolution of 96~km and 2) the

696: reduction of field strength in the plumes increases somewhat when we

697: move to a resolution of 24~km.  On that basis, the fundamental physical

698: process of sheet-like plume convection appears to be a robust

699: feature. The results will certainly change in detail (and, hopefully,

700: become even more similar to the observed penumbrae) as resolution

701: increases, but we do not expect totally new processes replacing those

702: that we have described here.

703: 

704: The rapid increase in available computational power and the foreseeable

705: progress in local helioseismology will soon alleviate some of the

706: limitations of the present approach and thus enable us to carry out even

707: more realistic simulations. 

708: 

709: \acknowledgements

710: Vasily Zakharov kindly provided Fig.~\ref{fig:sketch}.

711: M. Rempel wishes to thank the Institute for Pure and Applied Mathematics

712: (IPAM) of UCLA, Los Angeles for their support to attend the program on

713: `Grand Challenge Problems in Computational Astrophysics', from which

714: this work has benefited significantly. M. Rempel also thanks J.~M. Borrero

715: for in-depth discussions regarding observations and models of sunspot

716: structure. The National Center for Atmospheric Research (NCAR) is sponsored 

717: by the National Science Foundation.

718: 

719: 

720: \appendix

721: \section{Numerical scheme}

722: \label{appendix}

723: %

724: To alleviate the time step constraint for an explicit code brought about

725: by high Alfv{\'e}n velocity, we limit the strength of the Lorentz

726: force in the regions of low $\beta$, such that the resulting Alfv{\'e}n

727: velocity has a given upper bound, $c_{\rm max}$:

728: \begin{equation}

729:   \vec{F}_L=\frac{c_{\rm max}^2}{\sqrt{c_{\rm max}^4+v_{\rm A}^4}}\,

730:     \vec{J}\times\vec{B}\;.

731: \end{equation}

732: Here $v_{\rm A}$ denotes the value that the Alfv{\'e}n

733: velocity would have without limitation. The modification of the Lorentz

734: force is applied to regions where $v_{\rm A} > c_{\rm max}$

735: or, in terms of the $\beta$ value, where

736: \begin{equation}

737:   \beta<\frac{8\pi\,p}{\varrho c_{\rm max}^2}\approx 

738:   \left(\frac{c_{\rm sound}}{c_{\rm max}}\right)^2 \approx 0.05\;.

739: \end{equation}

740: For the simulations presented in this paper, we have used a value of

741: $c_{\rm sound}=7$ km$\cdot$s$^{-1}$ to estimate the speed of sound above

742: the sunspot umbra and set $c_{\rm max}=31$ km$\cdot$s$^{-1}$.  Since the

743: magnetic field is already in an almost force free state for

744: $\beta<0.05$, this correction has only a minor influence on the overall

745: force balance. We have tested this with a 2D run for which we varied the

746: value of $c_{\rm max}$ by a factor of $3$ and found no significant

747: difference. This approach increases the explicit time step limit by

748: about $2$ orders of magnitude at zero computational expense, in contrast

749: to the alternative of an implicit treatment of the Lorentz force.

750: 

751: The presence of high and low $\beta$ regions as well as high and low

752: Mach number flows presents a significant challenge for a numerical

753: scheme to properly resolve all regimes without being too diffusive in

754: any of them.  To this end, we have changed the artificial diffusivity

755: scheme of the {\em MURaM} code. After a piecewise linear reconstruction of the

756: solution, $u_i$, where the reconstruction slope in each cell $\Delta

757: u_i$ is limited by an appropriate slope limiter, we use the extrapolated

758: values at the interface $u_l=u_i+0.5\,\Delta u_i$ and

759: $u_r=u_{i+1}-0.5\,\Delta u_{i+1}$ to compute the diffusive flux

760: \begin{equation}

761:   F_{i+\frac{1}{2}}=\frac{1}{2}\, c_{i+\frac{1}{2}}\,

762:   \phi(u_r-u_l,u_{i+1}-u_{i})\,\left(u_r-u_l\right)\;,

763: \end{equation}

764: where $c$ denotes the characteristic velocity. Using $\phi=1$ reduces

765: the scheme to a standard (at least) second order flux (depending on

766: limiter) such as used in a second-order Lax Friedrichs scheme. For the

767: $4$th order MHD scheme of the {\em MURaM} code, we found that a choice of

768: $\phi=\left(\frac{u_r-u_l}{u_{i+1}-u_{i}}\right)^2$ for

769: $(u_r-u_l)\cdot(u_{i+1}-u_{i})>0$ and $\phi=0$ otherwise (no artificial

770: steepening) represents the best compromise between maximum stability and

771: minimum diffusion. The diffusivity of the scheme is controlled through

772: the slope limiter, for which we use a linear combination of the most

773: diffusive Minmod and least diffusive Superbee limiter 

774: \citep[see, e.g.][]{LeVeque:1990}:

775: $\eps\,{\rm Minmod}+(1-\eps)\,{\rm Superbee}$ and allow $\eps$ to vary

776: as function of $\beta$ (typically $0.5$ in high- and less than $0.2$ in

777: low-$\beta$ regions).  Furthermore it turned out that it is sufficient

778: to use $c=0.1\,c_{\rm sound}+v+v_{\rm alf}$ for the characteristic

779: velocity, which significantly reduces the diffusivity in low Mach number

780: flows. We apply this scheme to all MHD variables and account for effects

781: of (unfortunately unavoidable) mass diffusion in the momentum and energy

782: fluxes. The div $\vec{B}$ error produced by the diffusion scheme is

783: controlled by iterating

784: \begin{equation}

785:   \frac{\partial \vec{B}}{\partial t}=\mu (\Delta x)^2

786:        {\rm grad} ({\rm div}\vec{B}) \;.

787: \end{equation}

788: For values of $\mu=0.3\ldots 0.5$, typically less than $10$ iterations

789: are required to satisfy ${\rm max}\,(\Delta x\,\vert{\rm

790: div}\vec{B}\vert\,) < 10^{-3} B_{\rm rms}$, which was found to be

791: sufficient. In the results presented in this paper, we did not use any

792: explicit viscosity or magnetic diffusivity. The artificial diffusivities

793: are added once every time step, after completion of the Runge-Kutta loop

794: of the $4^{\rm th}$ order MHD scheme.  The treatment of diffusivity

795: outlined above leads to a scheme that is fully shock-capturing, at least

796: $4$th order accurate in smooth regions (higher order is possible

797: depending on slope limiter used), minimally diffusive for low Mach

798: number flows, and stable to $\beta$ values as low as $10^{-4}$.

799: 

800: A very low value of $\beta$ also leads to problems in codes that use the

801: conservative formulation of the energy equation, since the determination

802: of the internal energy requires to compute the small difference between

803: the nearly equal values of the total and the magnetic energies.  To

804: avoid this problem, we switch to an isothermal equation of state in

805: regions where $E_{\rm int}<10^{-3}E_{\rm mag}$ and prevent too small

806: $\beta$ values by imposing a lower limit for the density when $E_{\rm

807: int}<10^{-5}E_{\rm mag}$. An alternative to this procedure would have

808: been to directly solve an equation for the internal energy at the

809: expense of loosing the advantage of the conservative formulation. Since

810: most of the dynamics are driven beneath the photosphere and stability

811: problems only occur in layers with an average density at least $3$

812: orders of magnitude less than the photosphere (and therefore of little

813: dynamical importance), we have preferred to stay with the conservative

814: formulation.

815: 

816: \begin{thebibliography}{30}

817: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi

818: 

819: \bibitem[{{Bellot Rubio} {et~al.}(2005){Bellot Rubio}, {Langhans}, \&

820:   {Schlichenmaier}}]{Bellot-Robio:etal:2005}

821: {Bellot Rubio}, L.~R., {Langhans}, K., \& {Schlichenmaier}, R. 2005, \aap, 443,

822:   L7

823: 

824: \bibitem[{{Bharti} {et~al.}(2007){Bharti}, {Jain} \&

825:   {Jaaffrey}}]{Bharti:etal:2007}

826: {Bharti}, L., {Jain}, R., \& {Jaaffrey}, S.~N.~A.  2007, \apj, 665, L79

827: 

828: %\bibitem[{{Bharti} {et~al.}(2007b){Bharti}, {Joshi} \&

829: %  {Jaaffrey}}]{Bharti:etal:2007b}

830: %{Bharti}, L., {Joshi}, C., \& {Jaaffrey}, S.~N.~A.  2007b, \apj, 669, L57

831: 	

832: \bibitem[{{Borrero} {et~al.}(2005){Borrero}, {Lagg}, {Solanki}, \&

833:   {Collados}}]{Borrero:etal:2005}

834: {Borrero}, J.~M., {Lagg}, A., {Solanki}, S.~K., \& {Collados}, M. 2005, \aap,

835:   436, 333

836: 

837: \bibitem[{{Borrero} {et~al.}(2008){Borrero}, {Lites}, \&

838:   {Solanki}}]{Borrero:etal:2008}

839: {Borrero}, J.~M., {Lites}, B.~W., \& {Solanki}, S.~K. 2008, \aap, 481, L13

840: 

841: \bibitem[{{Borrero} {et~al.}(2006){Borrero}, {Solanki}, {Lagg},

842:   {Socas-Navarro}, \& {Lites}}]{Borrero:etal:2006}

843: {Borrero}, J.~M., {Solanki}, S.~K., {Lagg}, A., {Socas-Navarro}, H., \&

844:   {Lites}, B. 2006, \aap, 450, 383

845: 

846: \bibitem[{{Cameron} {et~al.}(2008){Cameron}, {Gizon}, \&

847:   {Duvall}}]{Cameron:etal:2008}

848: {Cameron}, R., {Gizon}, L., \& {Duvall}, Jr., T.~L. 2008, \solphys, 251, 291 

849: 

850: \bibitem[{{Cameron} {et~al.}(2007){Cameron}, {Sch{\"u}ssler}, {V{\"o}gler}, \&

851:   {Zakharov}}]{Cameron:etal:2007b}

852: {Cameron}, R., {Sch{\"u}ssler}, M., {V{\"o}gler}, A., \& {Zakharov}, V. 2007,

853:   \aap, 474, 261

854: 

855: \bibitem[{{Danielson}(1961)}]{Danielson:1961}

856: {Danielson}, R.~E. 1961, \apj, 134, 289

857: 

858: \bibitem[{{Degenhardt}(1993)}]{Degenhardt:1993}

859: {Degenhardt}, D. 1993, \aap, 277, 235

860: 

861: \bibitem[{{Heinemann} {et~al.}(2007){Heinemann}, {Nordlund}, {Scharmer}, \&

862:   {Spruit}}]{Heinemann:etal:2007}

863: {Heinemann}, T., {Nordlund}, {\AA}., {Scharmer}, G.~B., \& {Spruit}, H.~C.

864:   2007, \apj, 669, 1390

865: 

866: \bibitem[{{Ichimoto} {et~al.}(2007){Ichimoto}, {Suematsu}, {Tsuneta},

867:   {Katsukawa}, {Shimizu}, {Shine}, {Tarbell}, {Title}, {Lites}, {Kubo}, \&

868:   {Nagata}}]{Ichimoto:etal:2007}

869: {Ichimoto}, K., {Suematsu}, Y., {Tsuneta}, S., {Katsukawa}, Y., {Shimizu}, T.,

870:   {Shine}, R.~A., {Tarbell}, T.~D., {Title}, A.~M., {Lites}, B.~W., {Kubo}, M.,

871:   \& {Nagata}, S. 2007, Science, 318, 1597

872: 

873: \bibitem[{{Ichimoto} {et~al.}(2008){Ichimoto}, {Tsuneta}, {Suematsu},

874:   {Katsukawa}, {Shimizu}, {Lites}, {Kubo}, {Tarbell}, {Shine}, {Title}, \&

875:   {Nagata}}]{Ichimoto:etal:2008}

876: {Ichimoto}, K., {Tsuneta}, S., {Suematsu}, Y., {Katsukawa}, Y., {Shimizu}, T.,

877:   {Lites}, B.~W., {Kubo}, M., {Tarbell}, T.~D., {Shine}, R.~A., {Title}, A.~M.,

878:   \& {Nagata}, S. 2008, \aap, 481, L9

879: 

880: \bibitem[{{Keppens} \& {Martinez Pillet}(1996)}]{Keppens:Martinez:1996}

881: {Keppens}, R. \& {Martinez Pillet}, V. 1996, \aap, 316, 229

882: 

883: \bibitem[{{Langhans} {et~al.}(2007){Langhans}, {Scharmer}, {Kiselman}, \&

884:   {L{\"o}fdahl}}]{Langhans:etal:2007}

885: {Langhans}, K., {Scharmer}, G.~B., {Kiselman}, D., \& {L{\"o}fdahl}, M.~G.

886:   2007, \aap, 464, 763

887: 

888: \bibitem[{{LeVeque}(1990)}]{LeVeque:1990}

889: {LeVeque}, R.~J. 1990, Numerical Methods for Conservation Laws (Basel:

890:   Birkhauser-Verlag)

891: 

892: \bibitem[{{Montesinos} \& {Thomas}(1997)}]{Montesinos:Thomas:1997}

893: {Montesinos}, B. \& {Thomas}, J.~H. 1997, \nat, 390, 485

894: 

895: \bibitem[{{Riethm{\"u}ller} {et~al.}(2008){Riethm{\"u}ller}, {Solanki}, \&

896:   {Lagg}}]{Riethmueller:etal:2008}

897: {Riethm{\"u}ller}, T.~L., {Solanki}, S.~K., \& {Lagg}, A. 2008, \apjl, 678,

898:   L157

899: 

900: \bibitem[{{Rimmele} \& {Marino}(2006)}]{Rimmele:Marino:2006}

901: {Rimmele}, T. \& {Marino}, J. 2006, \apj, 646, 593

902: 

903: \bibitem[{{Scharmer} {et~al.}(2008){Scharmer}, {Nordlund}, \&

904:   {Heinemann}}]{Scharmer:etal:2008}

905: {Scharmer}, G.~B., {Nordlund}, {\AA}., \& {Heinemann}, T. 2008, \apjl, 677,

906:   L149

907: 

908: \bibitem[{{Scharmer} \& {Spruit}(2006)}]{Scharmer:Spruit:2006}

909: {Scharmer}, G.~B. \& {Spruit}, H.~C. 2006, \aap, 460, 605

910: 

911: \bibitem[{{Schlichenmaier} {et~al.}(1998{\natexlab{a}}){Schlichenmaier},

912:   {Jahn}, \& {Schmidt}}]{Schlichenmaier:etal:1998b}

913: {Schlichenmaier}, R., {Jahn}, K., \& {Schmidt}, H.~U. 1998{\natexlab{a}}, \apj,

914:   493, L121

915: 

916: \bibitem[{{Schlichenmaier} {et~al.}(1998{\natexlab{b}}){Schlichenmaier},

917:   {Jahn}, \& {Schmidt}}]{Schlichenmaier:etal:1998a}

918: ---. 1998{\natexlab{b}}, \aap, 337, 897

919: 

920: \bibitem[{{Schlichenmaier} \& {Solanki}(2003)}]{Schlichenmaier:Solanki:2003}

921: {Schlichenmaier}, R. \& {Solanki}, S.~K. 2003, \aap, 411, 257

922: 

923: \bibitem[{{Sch{\"u}ssler} \& {V{\"o}gler}(2006)}]{Schuessler:Voegler:2006}

924: {Sch{\"u}ssler}, M. \& {V{\"o}gler}, A. 2006, \apjl, 641, L73

925: 

926: \bibitem[{{Solanki} \& {Montavon}(1993)}]{Solanki:Montavon:1993}

927: {Solanki}, S.~K. \& {Montavon}, C.~A.~P. 1993, \aap, 275, 283

928: 

929: \bibitem[{{Solanki} {et~al.}(1994){Solanki}, {Montavon}, \&

930:   {Livingston}}]{Solanki:etal:1994}

931: {Solanki}, S.~K., {Montavon}, C.~A.~P., \& {Livingston}, W. 1994, \aap, 283,

932:   221

933: 

934: \bibitem[{{Solanki} \& {Schmidt}(1993)}]{Solanki:Schmidt:1993}

935: {Solanki}, S.~K. \& {Schmidt}, H.~U. 1993, \aap, 267, 287

936: 

937: \bibitem[{{Spruit} \& {Scharmer}(2006)}]{Spruit:Scharmer:2006}

938: {Spruit}, H.~C. \& {Scharmer}, G.~B. 2006, \aap, 447, 343

939: 

940: \bibitem[{{van Noort} \& {Rouppe van der Voort}(2008)}]{Noort:Rouppe:2008}

941: {van Noort}, M.~J. \& {Rouppe van der Voort}, L.~H.~M.  2008, \aap, 489, 429

942: 

943: \bibitem[{{V{\" o}gler} {et~al.}(2005){V{\" o}gler}, {Shelyag}, {Sch{\"

944:   u}ssler}, {Cattaneo}, {Emonet}, \& {Linde}}]{Voegler:etal:2005}

945: {V{\" o}gler}, A., {Shelyag}, S., {Sch{\" u}ssler}, M., {Cattaneo}, F.,

946:   {Emonet}, T., \& {Linde}, T. 2005, \aap, 429, 335

947: 

948: \bibitem[{{Weiss} {et~al.}(2004){Weiss}, {Thomas}, {Brummell}, \&

949:   {Tobias}}]{Weiss:etal:2004}

950: {Weiss}, N.~O., {Thomas}, J.~H., {Brummell}, N.~H., \& {Tobias}, S.~M. 2004,

951:   \apj, 600, 1073

952: 

953: \bibitem[{{Zakharov} {et~al.}(2008){Zakharov}, {Hirzberger}, {Riethm{\"u}ller},

954:   {Solanki}, \& {Kobel, P.}}]{Zakharov:etal:2008}

955: {Zakharov}, V., {Hirzberger}, J., {Riethm{\"u}ller}, T., {Solanki}, S., \&

956:   {Kobel, P.} 2008, \aap, 488, L17

957: 

958: \end{thebibliography}

959: 

960: \end{document}

961: