1: % mn2esample.tex
2: %
3: % v2.1 released 22nd May 2002 (G. Hutton)
4: %
5: % The mnsample.tex file has been amended to highlight
6: % the proper use of LaTeX2e code with the class file
7: % and using natbib cross-referencing. These changes
8: % do not reflect the original paper by A. V. Raveendran.
9: %
10: % Previous versions of this sample document were
11: % compatible with the LaTeX 2.09 style file mn.sty
12: % v1.2 released 5th September 1994 (M. Reed)
13: % v1.1 released 18th July 1994
14: % v1.0 released 28th January 1994
15:
16:
17: \documentclass[epsf,useAMS,usenatbib]{mn2e}
18:
19: % If your system does not have the AMS fonts version 2.0 installed, then
20: % remove the useAMS option.
21: %
22: % useAMS allows you to obtain upright Greek characters.
23: % e.g. \umu, \upi etc. See the section on "Upright Greek characters" in
24: % this guide for further information.
25: %
26: % If you are using AMS 2.0 fonts, bold math letters/symbols are available
27: % at a larger range of sizes for NFSS release 1 and 2 (using \boldmath or
28: % preferably \bmath).
29: %
30: % The usenatbib command allows the use of Patrick Daly's natbib.sty for
31: % cross-referencing.
32: %
33: % If you wish to typeset the paper in Times font (if you do not have the
34: % PostScript Type 1 Computer Modern fonts you will need to do this to get
35: % smoother fonts in a PDF file) then uncomment the next line
36: % \usepackage{Times}
37:
38: %%%%% AUTHORS - PLACE YOUR OWN MACROS HERE %%%%%
39:
40: \usepackage{lscape}
41: \usepackage{graphicx}
42: \usepackage{natbib}
43: \usepackage{graphics}
44:
45: \def \OIII {[O~{\sc iii}] 5007~\AA}
46: \def \NII {[N~{\sc ii}] 6584~\AA}
47: \def \vhel{\ifmmode{~V_{{\rm HEL}}}\else{~$V_{{\rm HEL}}$}\fi}
48: \def \vsys{\ifmmode{~V_{{\rm SYS}}}\else{~$V_{{\rm SYS}}$}\fi}
49: \def \HA {\ifmmode{{\rm\H}\alpha}\else{${\rm\ H}\alpha$}\fi}
50: \def \farcs{\hbox{$~\!\!^{\prime\prime}$}}
51:
52: % a) solar masses
53:
54: \def \msun{\ifmmode{{\rm\ M}_\odot}\else{${\rm\ M}_\odot$}\fi}
55:
56: % b) solar masses per year
57: \def \myr{\ifmmode{{\rm\ M}_\odot{\rm\ yr}^{-1}}
58: \else{${\rm\ M}_\odot$ yr$^{-1}$}\fi}
59:
60: % e) Mass loss
61: \def \mdot{\ifmmode{\dot{M}}\else{$\dot{M}$}\fi}
62: \def \tena#1 #2 {\ifmmode{#1 \times 10^{#2}}\else{$#1 \times 10^{#2}$}\fi}
63: \def \kms{\ifmmode{~{\rm km\,s}^{-1}}\else{~km s$^{-1}$}\fi}
64:
65: % Journal definitions
66: \def \philtrans{RSPT}
67: \def \apj{ApJ}
68: \def \mnras{MNRAS}
69: \def \pasp{PASP}
70: \def \araa{ARA\&A}
71: \def \aap{A\&A}
72: \def \aj{AJ}
73: \def \baas{BAAS}
74: \def \physrep{PhR}
75: \def \apjs{ApJS}
76: \def \apjl{ApJL}
77: \def \nat{Nat}
78: \def \iaucirc{IAUC.}
79: \def \aaps{A\&AS}
80: \def \apss{Ap\&SS}
81:
82: \title[Impact of tangled magnetic fields on AGN-blown bubbles]{
83: Impact of tangled magnetic fields on AGN-blown bubbles}
84: \author[M. Ruszkowski et al.]{M. Ruszkowski,$^1$\thanks{E-mail:
85: mr@mpa-garching.mpg.de (MR)} T.A. En{\ss}lin$^1$, M. Br{\"u}ggen$^2$, S. Heinz$^3$,
86: \& C. Pfrommer$^4$ \\
87: \\
88: $^1$Max Planck Institute for Astrophysics, Karl-Schwarzschild-Str. 1,
89: 85741 Garching, Germany\\
90: $^2$International University Bremen, Campus Ring 1, Bremen, Germany\\
91: $^3$Department of Astronomy, University of Wisconsin, 475 N Charter Street
92: Madison, WI 53706, USA\\
93: $^4$Canadian Institute for Theoretical Astrophysics, 60 St. George Street
94: Toronto, Ontario, M5S 3H8, Canada
95: }
96: \begin{document}
97:
98: %\date{Accepted. Received; in original form}
99: \date{Submitted 2006 December}
100:
101: \pagerange{\pageref{firstpage}--\pageref{lastpage}} \pubyear{2006}
102: \maketitle
103: \label{firstpage}
104:
105: \begin{abstract}
106: There is growing consensus that feedback from
107: active galactic nuclei (AGN) is the main
108: mechanism responsible for stopping cooling flows in clusters of galaxies.
109: AGN are known to inflate buoyant bubbles that supply mechanical power
110: to the intracluster gas (ICM). High Reynolds number hydrodynamical simulations
111: show that such bubbles get entirely disrupted within 100 Myr,
112: as they rise in cluster
113: atmospheres, which is contrary to observations. This artificial mixing
114: has consequences
115: for models trying to quantify the amount of heating and star formation
116: in cool core clusters of galaxies. It has been suggested
117: that magnetic fields can stabilize bubbles against disruption.
118: We perform magnetohydrodynamical (MHD) simulations of fossil bubbles
119: in the presence of tangled
120: magnetic fields using the high order {\it PENCIL} code.
121: We focus on the physically-motivated case where thermal pressure dominates
122: over magnetic pressure and
123: consider randomly oriented fields with and without maximum helicity
124: and a case where large scale external fields drape the bubble.
125: We find that helicity has some stabilizing effect.
126: However, unless the coherence length of magnetic fields exceeds the
127: bubble size, the bubbles are quickly shredded.
128: As observations of Hydra A suggest that lengthscale of magnetic
129: fields may be smaller then typical bubble size,
130: this may suggest that other mechanisms, such as viscosity,
131: may be responsible for stabilizing the bubbles.
132: However, since Faraday rotation observations of radio lobes
133: do not constrain large scale ICM fields well if they are aligned with
134: the bubble surface, the draping case may be a viable
135: alternative solution to the problem.
136: A generic feature found in our simulations is the formation of
137: magnetic wakes where fields are ordered and amplified. We suggest that
138: this effect could prevent evaporation by thermal conduction
139: of cold H$\alpha$ filaments observed in the Perseus cluster.
140: \end{abstract}
141:
142: \begin{keywords}
143: ICM: outflows - MHD - magnetic fields - AGN: clusters of galaxies
144: \end{keywords}
145:
146: \section{Introduction}
147:
148: Active galactic nuclei play a central role in explaining the
149: riddle of cool core clusters of galaxies. One of the unsolved problems
150: of AGN feedback in clusters is the issue of morphology and stability of buoyant
151: bubbles inflated by AGN and the efficiency of their mixing with the surrounding
152: ICM.
153: It is important to understand
154: the process of bubble fragmentation and its eventual mixing with the
155: rest of the ICM in order to quantify mass deposition
156: and star formation rates in cool cores clusters.
157: In the best-studied case of the Perseus cluster (Fabian et al. 2006)
158: observations indicate that such bubbles
159: can remain stable even far from cluster centers where they were created
160: (see cup-shaped feature northwest of the center of the
161: cluster or a similar yet tentative feature to the south in their
162: Figure 3).
163: One possible explanation for this phenomenon is that the intracluster
164: medium is viscous and that viscosity suppresses Rayleigh-Taylor
165: and Kelvin-Helmholz instabilities on their surfaces.
166: Reynolds et al. (2005) performed a series of numerical
167: experiments, compared inviscid and viscous cases and quantified how
168: much viscosity is needed to prevent bubble disruption. For viscosity
169: at the level of 25\% of the Braginskii value they obtained results
170: consistent with observations. The same problem has been considered
171: analytically by Kaiser et al. (2005) who computed the instability growth
172: rate as a function of scale and viscosity coefficient. An alternative
173: possibility is that bubbles are made more stable by
174: significant deceleration during their initial evolution
175: (Pizzolato \& Soker 2006).
176: The ``viscous solution'' is very appealing as it also
177: addresses the issue of dissipation of sound waves, weak shocks
178: as well as $g-$modes (Fabian et al. 2003a). However,
179: it is not entirely clear what the level and nature of
180: viscosity in the ICM really is. This is because the ICM
181: and the bubbles themselves are known to be
182: magnetized and magnetic fields may
183: reduce transport processes (but see
184: Lazarian 2006 and Schekochihin \& Cowley 2006).
185: Although magnetic fields in clusters are known to have
186: plasma $\beta > 1$, ($\beta \equiv P_{\rm gas}/(B^2/2\mu_{o}$); e.g.,
187: Blanton et al. 2003)
188: they may in principle have a
189: strong effect on suppressing Kelvin-Helmholz and Rayleigh-Taylor
190: instabilities. Jones \& De Young (2005) considered
191: the evolution of bubbles in a magnetized ICM
192: by performing two-dimensional numerical MHD simulations
193: and found that
194: bubbles could be prevented from shredding even when $\beta$ is as high
195: as $\sim 120$. A somewhat different conclusion was reached by
196: Robinson et al. (2004) who simulated magnetized bubbles with the {\it
197: FLASH} code (Fryxell et al. 2000) in two dimensions and found out
198: that a dynamically important magnetic field ($\beta < 1$) is required
199: to maintain bubble integrity.
200: Both of the above-mentioned simulations were performed in 2D and for very
201: idealized magnetic field configurations such as,
202: for example, doughnut-shaped fields
203: with symmetry axis parallel to the direction of motion.
204: Recently, Nakamura, Li, \& Li (2006) considered the stability of Poyinting-flux
205: dominated jets in cluster environment.
206: Here we focus on a later stage in the outflow evolution and consider
207: physically-motivated magnetic field case of
208: $\beta > 1$ and
209: more realistic (stochastically tangled) field configurations in fossil bubbles
210: (i.e., after the transition from the momentum-driven to
211: buoyancy-driven stage) than in previous buoyant stage MHD simulations.
212: We show that only when the power spectrum cutoff
213: of magnetic field fluctuations is larger than the bubble size can the
214: bubble shredding be suppressed.
215: It is possible that such ``draping''
216: case is not representative of typical cool core fields
217: (Vogt \& En{\ss}lin 2005) and, if so, other mechanism, such as viscosity,
218: would be required to keep the bubbles stable.
219: However, Vogt \& En{\ss}lin (2005) estimated magnetic power spectra
220: in Hydra A only which shows extraordinarily strong AGN outbursts.
221: We also note that such
222: Faraday rotation observations of radio lobes are only weakly sensitive
223: to large scale magnetic fields aligned with the bubble surface
224: and rely on certain untested assumptions. Therefore, we argue that the draping
225: case may be a viable solution to the problem of bubble stability.\\
226: \indent
227: In Section 2 we present the simulation setup and
228: the justification for the parameter choices. Section 3 presents
229: results and Section 4 our conclusions. The
230: Appendix discusses in detail our method
231: of setting up initial magnetic field configurations.
232:
233: \section{Simulation details}
234: \subsection{The code}
235: The simulations were performed with the {\it PENCIL} code. (Dobler et
236: al. 2003, Haugen et al. 2003, Brandenburg et al. 2004, Haugen et al. 2004)
237: Although {\it PENCIL} is non-conservative, it is a highly accurate
238: grid code that is sixth order in space and third order in time.
239: It is particularly suited for weakly compressible
240: turbulent MHD flows.
241: Magnetic fields are implemented in terms of a vector
242: potential so the field remains solenoidal throughout simulation
243: (i.e., no divergence cleaning of the magnetic field is necessary).
244: The code is memory efficient, uses Message-Passing
245: Interface and is highly parallel. The {\it PENCIL} code is better suited
246: for the simulations of stability of magnetized bubbles than
247: smoothed particle magneto-hydrodynamical codes as it
248: can better capture certain aspects of the bubble (magnetohydro) dynamics,
249: such as Kelvin-Helmholtz instability, even for high density
250: contrasts.
251: % (Agertz et al. 2006)
252: and suffers less from numerical diffusion of the magnetic field.
253: Although no formal code comparison has been made,
254: other well known
255: MHD mesh codes, such as {\it FLASH} or {\it ZEUS}, likely require
256: comparatively higher resolution to achieve the same level of magnetic flux
257: conservation as the {\it PENCIL} code.
258:
259: \subsection{Initial conditions}
260:
261: \subsubsection{Density and temperature profiles}
262: The cluster gas was assumed to be isothermal with temperature equal to
263: 10 keV (this temperature was used to minimize the ratio of code viscosity to
264: the Braginskii value; see Section 2.2.4 for more explanation).
265: The gas was initially in hydrostatic equilibrium and subject to
266: gravitational acceleration given by a sum of two isothermal potentials
267:
268: \begin{equation}
269: g(r) =-\frac{2\sigma_{a}^{2}}{(r+r_{a})}
270: -\frac{2\sigma_{b}^{2}}{(r+r_{b})},
271: \end{equation}
272:
273: \noindent
274: where $\sigma_{a}=1.41$, $\sigma_{b}=2.69$, $r_{a}=30.0$ and
275: $r_{b}=5.0$ in code units (see below) were the parameters were
276: chosen to give a pressure profile consistent with that observed in
277: clusters. Central gas density was $10^{-25}$ g cm$^{-3}$
278: (i.e., about $5.2\times 10^{-2}$ electrons per cm$^{-3}$).
279: Self-gravity of the gas was neglected.
280: The profiles of pressure and temperature are shown in Figure 1.
281: We consider ideal gas with adiabatic equation of state with $\gamma =5/3$.
282: The bubbles were underdense by a factor
283: of ten with respect to the local ICM and its temperature was
284: increased by the same factor to keep it in pressure equilibrium with the
285: surrounding gas. The remaining bubble parameters are mentioned in
286: section 2.2.4 where we discuss code units. The
287: pressure in the ICM changes significantly over the height of the bubble. It
288: is for this reason that we modified the density and temperature on a
289: "point-by-point" basis. That is, at every location within the bubble,
290: we increased the temperature by a constant factor and decreased the density by
291: the same factor while keeping the pressure at the same level as the
292: pressure at the same distance from the cluster center away from the
293: bubble. This way, the initial pressure distribution is smooth and
294: there is no strong departure from
295: perfect equilibrium.
296: Although the density contrast of the bubble is relatively low, this
297: has minimal effect on our results. This is
298: because the buoyancy velocity is proportional to $( 1 -
299: \rho_{\rm bubble}/\rho_{\rm icm} )^{1/2}$ which is insensitive to
300: $\rho_{\rm bubble}$ as long as it is much
301: smaller than $\rho_{\rm icm}$.
302: The reason we opted for such relatively low density contrast
303: is numerical: higher bubble temperatures would have required smaller
304: timesteps to achieve numerical stability.
305: We also note that some real
306: bubbles may have smaller density contrasts. A well-known extreme example
307: is the Virgo cluster where buoyant bubbles in
308: the radio show very little corresponding structure in the X-ray
309: maps.
310:
311: \subsubsection{Magnetic fields}
312: We consider magnetic fields inside the bubbles and in the ICM
313: that are dynamically unimportant in the
314: sense that their plasma $\beta$ parameter is greater than unity.
315: That is, magnetic pressure may become important when compared to
316: the bubble ram pressure associated with the gas motions in the ICM but
317: it is generally small compared to the pressure of the ICM in our
318: simulations. Analysis of cluster bubbles by Dunn et al. suggest that
319: $\beta^{-1}$ is roughly in the range (10$^{-3}$, 0.3) with the mean
320: $\langle\beta^{-1}\rangle =0.06$
321: and median of 0.03 (see sample of Dunn et al. 2004, Dunn et al. 2005 and
322: Dunn 2006, Dunn priv. communication).
323: Other evidence for high $\beta$ comes from observations of Faraday
324: rotation measure and the lack of Faraday depolarization in bright
325: X-ray shells in Abell 2052 as observed by Blanton et al. (2003). They infer
326: $\beta\sim 30$. These arguments motivated us to consider high $\beta$ cases.
327: However, we note that low-$\beta$ is not ruled out beyond reasonable
328: doubt by the above observations and theoretical considerations and, thus,
329: its consequences for bubble dynamics should be investigated further.\\
330: \indent
331: Regarding the geometry of magnetic fields, we consider
332: random isotropic fields with coherence length smaller than the bubble
333: sizes (hereafter termed ``random''),
334: isotropic helical fields with
335: coherence length smaller then the bubble size (termed ``helical''
336: below; helicity is defined as $\int {\bf A} \cdot {\bf B}dV$, where
337: ${\bf A}$ and ${\bf B}$ are vector potential and magnetic field
338: respectively), and
339: the ``draping'' case of isotropic fields characterized by coherence
340: length exceeding bubble size as well as a non-magnetic case.\\
341: \indent
342: Magnetic draping has been considered previously
343: in the context of merging cluster cores and radio bubbles
344: using analytical approach by Lyutikov (2006). He found that even
345: when magnetic fields are dynamically unimportant throughout the ICM,
346: thin layer of dynamically important fields can form around merging
347: dense substructure clumps (``bullets'') and prevent their disruption.
348: We suggest that, depending on the unknown value of magnetic diffusivity,
349: there may be some relic magnetic power spectrum
350: with a smaller amplitude than the freshly injected one (either by
351: AGN bubbles or dynamo-driven) that extends to scales larger than the
352: bubble size and that provides draping fields to stabilize the bubbles.\\
353: \indent
354: The helical case is motivated by the fact that magnetic helicity is
355: conserved for ideal MHD case.
356: Moreover, helicity is proportional to the product of
357: magnetic energy and typical lengthscale and, thus, fragmentation is
358: not energetically preferred in the sense that it would lead to local
359: increase of magnetic energy. Another justification for helical fields
360: is that they may be responsible for explaining circular polarization
361: of certain radio sources, and especially its sign persistency
362: (En{\ss}lin 2003 but see Ruszkowski \& Begelman 2002).
363: It is conceivable that magnetic helicity that could produce such signal
364: in jets could survive till the buoyant stage.\\
365: \indent
366: The ``random'' case is motivated by
367: the work of Vogt \& En{\ss}lin (2005) who estimate power spectrum of
368: magnetic field fluctuations in Hydra A
369: from Faraday rotation maps. We are conservative in our
370: choice of parameters in the sense that we use fields of somewhat higher
371: coherence length that are more likely to stabilize bubbles. \\
372:
373:
374: \subsubsection{Vector potential setup}
375: In setting up initial magnetic field configuration we
376: ensured that the following conditions are met:\\
377:
378: \begin{itemize}
379:
380: \item magnetic fields must be divergence-free
381: \item bubble and environmental fields must be isolated, i.e., no magnetic field
382: lines should penetrate the surface of the bubble
383: \item magnetic fields must be characterized by the required power spectra
384: \item when necessary, magnetic helicity may be imposed
385: \item magnetic plasma $\beta$ parameter of the bubble and the ICM
386: must be independently controllable
387: \item all considered
388: field configurations must
389: result (after suitable modifications)
390: from the same white noise Gaussian random numbers. This ensures that
391: the differences in evolution are entirely due to model parameters
392: and not due to different random realizations.
393:
394: \end{itemize}
395:
396: \indent
397: The details of the algorithm used to set up the initial magnetic field
398: configuration is presented in the Appendix..
399:
400: \subsubsection{Code units and resolution}
401: We consider the the following code units: one length unit corresponds
402: to 1 kpc, one velocity unit to 3$\times 10^7$ cm/s and one density
403: unit to 10$^{-24}$ g/cm$^3$. This corresponds to one time unit of
404: approximately 3.3$\times 10^6$ years. The box size is 100 code length
405: units on a
406: side and shock-absorbing boundary conditions were used.
407: This choice of boundary conditions was dictated by the fact that
408: we wanted to exclude the Richmyer-Meshkov instability due to reflected
409: waves passing through the bubble as a potential destabilizing mechanism.
410: The bubble size was $d=25$ in diameter and
411: its center was offset by 20 code length units from the cluster origin.
412: The bubble was underdense by a factor
413: of ten with respect to the local ICM and its temperature was
414: increased by the same factor to keep it in pressure equilibrium with the
415: surrounding gas. Our grid size is 200$^3$ zones.
416: The {\it PENCIL} code requires some viscosity and resistivity to run
417: properly. Minimum required amount of physical viscosity that the
418: code needs to prevent numerical instabilities from developing
419: decreases as the numerical resolution is increased.
420: We use viscosity $\nu = 0.07$ and magnetic resistivity
421: $\eta = 0.07$. For typical values of electron density
422: ($\sim 0.01$ cm$^{-3}$) and temperature (10 keV) in our simulations
423: this value of viscosity corresponds to 0.014 of the
424: Braginskii value. We note that for lower gas temperatures this ratio
425: would be higher which motivated our choice of temperature.
426: For the parameters considered here, in the initial
427: stages in our simulations,
428: maximum velocities in code units are
429: roughly $u_{\rm max}\sim 6$.
430: Note that for the adopted parameters the sound speed in the
431: ICM is $c_{s}=5.44$ but the gas inside the bubbles is much hotter and
432: so the actual gas velocities can be higher then $u_{\rm max}$ without
433: ``violating'' the sound speed limit. For the bubble size $d=25$
434: and gas velocity $u_{\rm max}=6.0$,
435: hydrodynamical and magnetic Reynolds numbers are
436: roughly $Re=u_{\rm max}d/\nu\sim 2000$.
437: Such values clearly lead to quick development of Rayleigh-Taylor
438: and Kelvin-Helmholtz
439: instabilities for unmagnetized bubbles.
440:
441: \section{Results}
442:
443: \begin{figure}
444: \centering
445: \rotatebox{0}{\mbox{\resizebox{8.6cm}{!}{\includegraphics{f1.ps}}}}
446: \caption{\label{}{Initial pressure (solid line) and
447: temperature profiles.}}
448: \end{figure}
449:
450: \begin{figure}
451: \centering
452: \rotatebox{0}{\mbox{\resizebox{8.6cm}{!}{\includegraphics{f2.ps}}}}
453: \caption{\label{}{One dimensional energy
454: power spectra of initial magnetic field
455: fluctuations. Solid line shows energy spectrum corresponding to our
456: ``draping''. Dashed line is for the ``random'' case.}}
457: \end{figure}
458:
459: \begin{figure}
460: \centering
461: \rotatebox{0}{\mbox{\resizebox{9.0cm}{!}{\includegraphics{f3.ps}}}}
462: \caption{\label{}{Magnetic pressure in the plane containing the
463: cluster and bubble centers. Upper and lower panels show ``draping''
464: and ``random'' cases, respectively.
465: The coherence length in the lower panel is even smaller than that used
466: in the simulations to demonstrate the robustness of the method. See Appendix for more explanation.}}
467: \end{figure}
468:
469: \begin{figure*}
470: \centering
471: \begin{minipage}[b]{\textwidth}
472: \centering
473: \includegraphics[width=1.0\textwidth, angle=0.0]{f4.ps}
474: \caption{Natural logarithm of density in the cluster. Columns show
475: time sequences of density for draping, random (i), random (ii), helical and
476: non-magnetic
477: cases from left column to the right, respectively. Rows correspond to
478: time of 15.0, 25.0, 35.0 code time units from bottom to top. Sides of all
479: boxes show densities in the planes intersecting the box and containing
480: its center.\label{}}
481: \end{minipage}
482: \end{figure*}
483:
484: \begin{figure*}
485: \centering
486: \begin{minipage}[b]{\textwidth}
487: \centering
488: \includegraphics[width=1.0\textwidth, angle=0.0]{f5.ps}
489: \caption{\label{}{Magnetic pressure structure in the cluster.
490: Columns show time sequences of magnetic
491: pressure for draping, random (i), random (ii) and helical
492: cases from left column to the right, respectively. Rows correspond to
493: time of 0.0, 15.0, 25.0, 35.0 code time units from bottom to top.}}
494: \end{minipage}
495: \end{figure*}
496:
497: \begin{figure*}
498: \centering
499: \begin{minipage}[b]{\textwidth}
500: \centering
501: \includegraphics[width=1.0\textwidth, angle=0.0]{f7.ps}
502: \includegraphics[width=1.0\textwidth, angle=0.0]{f6.ps}
503: \caption{\label{}{X-ray images of the cluster center.
504: Columns show
505: time sequences of X-ray emission for draping, random (i), random (ii)
506: helical, and non-magnetic
507: cases from left column to the right, respectively. Rows correspond to
508: time of 15.0, 25.0, 35.0 code time units from bottom to top.
509: Rows 1 to 3 correspond to a different projection axis than rows 4 to 6.}}
510: \end{minipage}
511: \end{figure*}
512:
513: \begin{figure*}
514: \centering
515: \begin{minipage}[b]{\textwidth}
516: \centering
517: \includegraphics[width=0.7\textwidth, angle=0.0]{f8.ps}
518: \caption{\label{}{Natural logarithm of density distribution.
519: Left column shows density for the random (i) case (lower panel t=15,
520: upper t = 25 code time units) but
521: twice as high mean magnetic pressure.
522: Right column corresponds to the draping case for twice the
523: mean magnetic pressure compared to the original draping case.}}
524: \end{minipage}
525: \end{figure*}
526:
527: \begin{figure*}
528: \centering
529: \begin{minipage}[b]{\textwidth}
530: \centering
531: \includegraphics[width=0.7\textwidth, angle=0.0]{f9.ps}
532: \caption{\label{}{Same as Figure 7 but for the logarithm of magnetic pressure
533: distribution. Note that in the draping case (right panel), a layer of
534: ordered and amplified magnetic field forms on the bubble-ICM interface
535: and protects the bubble against disruption.}}
536: \end{minipage}
537: \end{figure*}
538:
539: \begin{figure*}
540: \centering
541: \begin{minipage}[b]{\textwidth}
542: \centering
543: \includegraphics[width=0.7\textwidth, angle=0.0]{f10.ps}
544: \caption{\label{}{X-ray maps corresponding to Figures 8 and
545: 9. Projection orientation is the same as in the upper panels in Figure 6.}}
546: \end{minipage}
547: \end{figure*}
548:
549: \begin{figure}
550: \centering
551: \rotatebox{0}{\mbox{\resizebox{8.6cm}{!}{\includegraphics{f11.ps}}}}
552: \rotatebox{0}{\mbox{\resizebox{8.6cm}{!}{\includegraphics{f12.ps}}}}
553: \caption{\label{}{ Evolution of the mean magnetic pressure (averaged
554: in the plane containing the cluster center and the initial bubble
555: location) compared to
556: gas pressure for the draping (solid lines) and random (i) cases. Lower
557: panel is for twice the original mean initial magnetic pressure.}}
558: \end{figure}
559:
560:
561: In Figure 4 we show the evolution of the gas density. Each column corresponds
562: to one model. Time increases from bottom to top. Snapshots were taken
563: at 15, 25, 35 code time units. The following five models were considered
564: (from left to right):
565:
566: \begin{itemize}
567: \item draping model: mean $\beta\sim 40$ throughout the box
568: (i.e., relative plasma $\beta$ of the bubble with respect
569: to the ICM is set to unity)
570: \item random (i) model: constant mean $\beta$ throughout the box.
571: The initial mean $\beta$ was chosen so as to match mean $\beta$ values
572: near $t=15$ code time units in the draping case
573: \item random (ii) model : initial bubble-averaged beta inside the
574: bubble the same as the mean initial $\beta$ in case random (i)
575: Relative magnetic pressure inside the bubble was chosen to be 10
576: times higher than the typical ambient magnetic pressure
577: \item helical case: same as random (ii) but for helical fields
578: \item non-magnetic case.
579: \end{itemize}
580:
581: \indent
582: It is evident that the bubble morphology depends strongly on the
583: topology of the initial magnetic field.
584: There is a striking difference between draping case (first column)
585: and all remaining cases. Even though magnetic fields
586: are ``dynamically'' unimportant in the sense that the typical plasma
587: $\beta$ parameter is much greater than unity, the bubble clearly is more
588: coherent in the draping case than in any other case. The cup-shaped
589: morphology in the draping case
590: in the first snapshot at 15 time units is very reminiscent of the
591: fossil bubble seen in the Perseus cluster. As the bubble moves up, its
592: shape changes and it becomes more round. The Rayleigh-Taylor
593: instability is prevented and so there is no evidence for strong
594: contamination or mixing of the bubble interior with the colder ICM material.
595: However, denser and colder material lifted by the rising bubble
596: reverses its motion at a certain time and begins to fall back.
597: This leads to some stretching of the bubble in the vertical direction
598: in the later stages in its evolution.\\
599: \indent
600: The second column shows random (i) case. Here the gas behind the bubble
601: quickly penetrates its center and tends to pierce through it.
602: This is evident especially when the left hand sides of the cubes
603: in the first and second column are compared. The
604: working surface of the bubble becomes irregular and starts to
605: fragment. Similar behavior is observed in the random (ii) case (third
606: column). The two
607: cases differ in the strength of magnetic field with case (ii) having
608: stronger internal bubble fields but weaker ICM ones. Nevertheless,
609: there does not seem to be much qualitative difference between these
610: two cases in terms of density distribution.\\
611: \indent
612: The fourth column shows the helical case. The degree of
613: bubble fragmentation is lower than in the random cases even though
614: typical magnetic field strength is similar to that of random (i)
615: case. Although this appears to be a weak effect,
616: as explained in Section 2.2.2, helicity conservation
617: should tend to stabilize the flow. This column should be compared mainly
618: with the third column that shows its non-helical analog.\\
619: \indent
620: The last column shows the non-magnetic case. Both Rayleigh-Taylor and
621: Kelvin-Helmholz instabilities develop here very quickly. The bubble is
622: pierced by a column of cold gas forming a ``smoke ring'' rather then
623: cup-shaped or oval structure as in the draping case. The edges of
624: this ring are initially wrinkled by Kelvin-Helmholz instability.\\
625: \indent
626: In Figure 5 we show the distribution of magnetic pressure. As in Figure 4,
627: columns correspond to different models and are ordered
628: the same way while rows correspond to 15, 25, 35 time units from
629: bottom to top. This figure reveals that the reason for bubble
630: stability in the draping case (first column)
631: is the formation of an ``umbrella'' or a
632: thin protective magnetic layer on the bubble working surface that
633: suppresses instabilities.
634: Visual inspection of the figure comparing magnetic
635: pressure evolution in the draping, two random, and helical cases shows
636: that the field geometry does not change significantly (in the average
637: sense) far away from the bubbles.
638: The upward motion of the bubble also leads to
639: substantial amplification and ordering
640: of magnetic field in the bubble wake. This
641: has consequences for estimates of conduction in the ICM based on the
642: appearance of H$\alpha$ filaments (Fabian et al. 2003b, Hatch et al. 2006).
643: It has been argued (e.g., Nipoti \& Binney 2004)
644: that thermal conduction has to be strongly suppressed in the ICM or
645: otherwise such cold filaments would be rapidly evaporated. Our result
646: hints at a possibility that thermal conduction may be locally
647: weaker in the bubble
648: wake, thus preventing or slowing down filament evaporation.\\
649: \indent
650: The second column shows random (i) case. No ``umbrella'' effect seen in the
651: draping case is observed here. The fact that the colder gas
652: enters the bubble from beneath does not lead to compression of magnetic
653: field in the direction parallel to the bubble surface, an effect that could
654: prevent disruption. Even though the bubble does get disrupted,
655: some compression and ordering of magnetic field in the wake
656: occurs here just as in the draping case. In fact, this effect is
657: a generic feature of all the runs that we performed. The behavior of
658: magnetic pressure in random (ii) case, shown in the third column, is
659: qualitatively very similar with the difference that the magnetic wakes
660: are more pronounced. Even though magnetic fields get
661: marginally compressed near the bubble working
662: surface in this case, they remain disjoint and
663: do not prevent the shredding of the bubble.
664: \\
665: \indent
666: The last column shows magnetic pressure in the helical case.
667: There is a significant difference in the topology of the field inside
668: the bubble between this and all previous cases. Here, the upward drift of
669: the bubble appears to amplify magnetic field inside the bubble. Note
670: that while in the draping case amplification took place in a thin layer
671: coinciding with the working surface of the bubble,
672: the amplification in the helical case is distributed
673: throughout the bubble. This is
674: particularly obvious in the snapshot taken at 15 code time units.
675: We note that this means that moderate
676: stabilizing effect of helicity comes from fields internal to the
677: bubble rather than external ones as in the draping case.
678: No obvious amplification inside the bubble is seen in the random cases
679: other than the ``wake'' amplification. \\
680: \indent
681: In Figure 6 we show X-ray maps. The columns are arranged the same way as
682: in Figure 4. Rows from 1 to 3 correspond to different projection
683: direction than rows 4 to 6 (the latter ones correspond to the viewing direction
684: rotated by 90 degrees around the axis intersecting the cluster center
685: and the original location of the bubble).
686: As expected, the morphology in the draping case is very
687: different than in all other cases. In this case the
688: bubbles show up as depressions
689: in X-ray emissivity. The bubbles in the random case appear
690: irregular. They are also
691: brighter in their centers which is contrary to observations.
692: In the helical case, the bubbles seem somewhat less disturbed than in
693: the random case. It is possible that this case, when ``observed''
694: in synthetic {\it Chandra} data that includes instrument responses
695: would resemble actual bubbles more than the random case bubbles.
696: The last column shows the non-magnetic case that is clearly inconsistent with
697: observations. Here the bubble more resembles two isolated bubbles than
698: a coherent feature seen in the draping case.
699: We suggest that a hybrid model that combines internal helical magnetic
700: fields inside the bubble with
701: external draping fields may produce X-ray bubbles that closely resemble
702: bubbles seen in clusters. However, different method for setting up
703: initial conditions than the one considered here
704: would have to be employed to model such a case while
705: ensuring that the initial magnetic
706: field configuration does not suffer from any artifacts.\\
707: \indent
708: We note that the magnetic fields in our simulations are observed to decay
709: with time. The decay is expected as, apart from the bubble-induced
710: motion, the turbulence is not continuously driven in our simulations.
711: As expected, the decay is faster for more tangled fields.
712: Even though the field decays, we note that disruption (when present) is
713: initiated early on in the bubble evolution. Moreover, our field
714: strengths are greater than those in Jones \& De Young (2005) uniform
715: field case and yet we do observe disruption.
716: We quantified the decay in
717: magnetic pressure for the cases where the mean bubble and ICM $\beta$
718: parameters are the same (``draping'' and random (i) cases).
719: In Figure 10 we show the evolution of the
720: mean magnetic pressure compared to gas pressure in the plane
721: intersecting the cluster center and the original bubble location.
722: We found that, while the "random" case shows the
723: decay, the "draping" case is consistent with no decay.
724: By construction, the mean fields in the draping case and random (i) one
725: approximately match at $t=15$ code time units.
726: The decay of the field after this time is rather slow and magnetic field
727: strengths in the random and draping cases are comparable around and
728: after this time. Prior to $t=15$ the field in the random case exceeds
729: that in the draping case. It is interesting that, even though this is the
730: case, the random case fields lead to bubble disruption while the
731: draping case shows much more coherent structures.
732: This demonstrates that the difference between this two cases
733: is primarily due to the field geometry and not its strength, at least for the
734: typical plasma $\beta$ considered here. Moreover, if we
735: would have
736: considered driven turbulence then we could have afforded to start from weaker
737: fields in the random case and additional random motions due to
738: turbulence driving would be present. Both of these effects (i.e.,
739: weaker initial field in conjunction with additional random motions)
740: could only strengthen our conclusion, i.e., make the random case
741: bubbles fragment even more easily. We are thus conservative in
742: neglecting turbulence driving. We also note that our aim
743: was not to address the stability of the bubbles exposed to random motions.
744: Our objective was to discuss the effect of tangled magnetic
745: fields on the development of Rayleigh-Taylor and Kelvin-Helmholz
746: instabilities. Random
747: motions, be it due to turbulence driving or due to post-merger
748: relaxation, are inevitably going to be present in cool cluster cores
749: (even though cool cores tend to be more relaxed than the centers of
750: non-cooling flow clusters).
751: Their impact has to be evaluated in evaluated in sepatate studies.
752: Another unknown factor is how
753: magnetic fields are driven inside the cavities (if at all).
754: However, including such effects would be beyond the scope of our
755: investigation.
756:
757: \subsection{Varying magnetic pressure}
758: We simulated draping and random case (i) again for exactly the same
759: parameters as before except for twice as high mean magnetic pressures in both
760: cases. We observe the same trends with the difference that the draping
761: case results in slightly more coherent bubbles while the random case
762: in slightly more fragmented ones. This strengthens our arguments
763: presented above that it is the geometry of the field rather than its
764: strength that is responsible for stabilizing the bubbles (at least for
765: the parameters considered here). In Figures 7, 8, and 9 we show
766: density, magnetic pressure and X-ray emissivity, respectively.
767: Note that the ``umbrella'' effect mentioned above is clearly seen in the
768: draping case in the left column in Figure 8. This is to be contrasted
769: with the left hand panel corresponding to the high magnetic version of
770: the random (i) case where no such effect is observed. Lower
771: panel in Figure 10 shows the evolution of the mean magnetic pressure
772: compared to the gas pressure in the plane containing the center of the
773: cluster and the initial position of the bubble.
774:
775: \section{Conclusions}
776:
777: We considered three-dimensional MHD simulations of buoyant bubbles
778: in cluster atmospheres for varying magnetic field strengths
779: characterized by plasma $\beta > 1$ and for varying field
780: topologies. We find that field topology plays a key role in
781: controlling the mixing of bubbles with the surrounding ICM.
782: We show that large scale
783: external fields are more likely to stabilize bubbles than
784: internal ones but a moderate stabilizing effect due to magnetic
785: helicity can make internal fields play a role too.
786: We demonstrate that bubble morphology
787: closely resembling fossil bubbles in the Perseus cluster could be realized
788: if the coherence of magnetic field is greater than the typical bubble size.
789: While it is not clear if such a ``draping''
790: case is representative of typical cluster fields,
791: Vogt \& En{\ss}lin (2005) find that length scale of magnetic fields
792: in Hydra A is smaller than typical bubble size.
793: If this also holds true in other clusters
794: then other mechanisms, such as viscosity,
795: would be required to keep the bubbles
796: stable. Unfortunately, Faraday rotation method used by
797: Vogt \& En{\ss}lin (2005) is not very sensitive to large scale
798: magnetic fields if aligned with the bubble surface.
799: Moreover, their maximum Likelihood method assumes a power-law relation between
800: magnetic field and density, statistical isotropy for the purpose of
801: deprojection and a particular jet angle with respect to the line-of-sight.
802: Smaller angles and different magnetic field configurations might yield
803: a weaker decline of the power spectrum at larger scales.
804: Taking into account the above limitations, it is entirely possible that
805: the draping case offers a viable alternative solution to
806: the problem of bubble stability.
807: We also suggest that a hybrid model that combines helical fields inside the
808: bubble with external draping fields could be successful in explaining
809: morphologies of X-ray bubbles in clusters.
810: Another possibility is that
811: dynamically significant fields are be present inside
812: the bubbles and the consequences of high-$\beta$ case
813: for bubble dynamics and stability should be investigated further.
814: We note that the bubbles will most
815: likely eventually get disrupted (partially helped by ``cosmological''
816: sloshing gas motions in clusters). \\
817: \indent
818: A generic feature found in our simulations is the formation of a
819: magnetic wake where fields are ordered and amplified. We suggest that
820: this effect could prevent evaporation by thermal conduction
821: of cold H$\alpha$ filaments observed in the Perseus cluster.\\
822: \indent
823: The physical process of bubble mixing
824: in the presence of magnetic fields
825: has important consequences also for modeling of mass deposition and
826: star formation rates in cool core clusters as well as the particle
827: content of bubbles and cosmic ray diffusion from them.
828: These issues will be further complicated by the effects of anisotropy
829: of transport processes due to magnetic fields. This may give rise to
830: the onset of magneto-thermal instability on the bubble-ICM interface
831: (Balbus 2004, Parrish \& Stone 2005). Studying such effects
832: is beyond the scope of the
833: present paper but certainly deserves further investigation.
834:
835: \section{Acknowledgements}
836: We thank the referee for a detailed and insightful report.
837: MR thanks Axel Brandenburg,
838: Kandu Subramanian, Maxim Lyutikov, Eugene Churazov,
839: Debora Sijacki, Jonathan Dursi, Maxim Markevitch, Alexei Vikhlinin
840: and Alexander Schekochihin for helpful discussions.
841: Test runs were performed on IBM p690 Regatta cluster at
842: Rechenzentrum Garching at the Max-Planck-Institut f{\"u}r Plasmaphysik.
843: Final production runs were performed on the Columbia supercomputer
844: at NASA NAS Ames center. It is MR's pleasure to thank
845: the staff of NAS, and especially Johnny Chang and Art
846: Lazanoff, for their their highly professional help.
847: The Pencil Code community is thanked for making the code publicly
848: available. The main code website is located at
849: \texttt{http://www.nordita.dk/software/pencil-code/}. MB acknowledges
850: support by the Deutsche Forschungsgemeinschaft.
851:
852: \bibliographystyle{mn2e}
853: \bibliography{mn}
854:
855:
856: \section{Appendix}
857:
858: Initial magnetic fields were computed outside the main code
859: and the setup was performed in two stages. In the first phase,
860: we generated stochastic fields by three-dimensional
861: inverse Fourier transform (FFT) of magnetic
862: field that in ${\bf k}$-space had the amplitude given by
863:
864: \begin{equation}
865: B\propto k^{-11/6}\exp(-(k/k_{1})^4)\exp(-k_{2}/k),
866: \end{equation}
867:
868: \noindent
869: where $k=(k_{x}^2+k_{y}^{2}+k_{z}^2)^{1/2}$, and
870: $k_{1} = 2\pi/dx$, $k_{2} = 0.3$, $dx = x_{\rm box}/32$,
871: $x_{\rm box} = 10^{2}$ for helical or random cases
872: and $k_{1} = 2\pi/(2r_{\rm bub})$ and
873: $k_{2} = 0.05$ in the draping case, where $x_{\rm box}$ is the size of
874: the computational box in code units. (see below)\\
875: \indent
876: All three components of magnetic field were treated
877: independently which ensured that the final distribution of ${\bf
878: B}({\bf r})$ had random phase. That is, for example for the x
879: component of the magnetic field, we set up a complex field
880: such that
881:
882: \begin{equation}
883: ({\rm Re}({B_{x}}({\bf k})),{\rm Im}({B_{x}}({\bf k})))=(G(u_{1})B,G(u_{2})B),
884: \end{equation}
885:
886: \noindent
887: where $G$ is a function of a uniform random
888: deviate $u_{1}$ or $u_{2}$ that returns Gaussian-distributed values.
889: For vanishing exponential cutoff terms, the above prescription would
890: give classical Kolmogorov turbulence spectrum.
891: Whereas there is no generally accepted
892: justification for magnetic spectrum to have
893: a Kolmogorov distribution, our parameter choice for the ``random'' case
894: resembles that seen in the Hydra cluster (Vogt \& En{\ss}lin 2005).
895: One-dimensional energy power spectra $kE(k)$ [erg s$^{-1}$ cm$^{-3}]$
896: of magnetic field fluctuations are shown in Figure 2. \\
897: \indent
898: After the initial field has been set up in k-space,
899: we implement a magnetically
900: isolated bubble. This phase is performed according to the following
901: iterative scheme:\\
902:
903: \indent
904: (1) divergence cleaning in ${\bf k}$-space:
905:
906: \begin{equation}
907: {\bf B}({\bf k})\longmapsto ({\bf 1}-\hat{\bf k}\hat{\bf k}){\bf B}({\bf k}),
908: \end{equation}
909:
910: \noindent
911: where $\hat{\bf k}$ is a unit vector in ${\bf k}$-space. In the case
912: of helical fields we also act on the field with helicity operator
913: defined as:
914:
915: \begin{equation}
916: {\bf B}({\bf k})\longmapsto
917: \frac{{\bf 1}+\alpha i\hat{\bf k}\times}{(1+\alpha^2)^{1/2}}{\bf B}({\bf k}),
918: \end{equation}
919:
920: \noindent
921: where $\alpha = 1$ (maximum helicity case).
922: The helicity step is only performed in the first loop of the iteration.
923: Note that imposing helicity on the field does not change the power
924: spectrum of magnetic energy fluctuations.
925:
926: \indent
927: (2) inverse FFT of the new ${\bf B(k)}$ to real space. Each component
928: of ${\bf B(k)}$ is independently acted upon with a three-dimensional
929: inverse FFT.
930:
931: \indent
932: (3) applying a projection operator to isolate the bubble magnetically.
933: The projection operation modifies the field in the following way
934:
935: \begin{equation}
936: {\bf B}({\bf r})\longmapsto[{\bf 1}-g(r)\hat{\bf r}
937: \hat{\bf r}]{\bf B}({\bf r}),
938: \end{equation}
939:
940: \noindent
941: where $\hat{\bf r}$ is a unit
942: vector in real space, $r$ is the distance from the bubble center,
943: and $g(r)=1-|\cos[0.5\pi(x+\Delta x-1)/\Delta x]|$
944: for $1-\Delta x < x < 1+\Delta x$ and
945: $g(r)=0$ otherwise and $\Delta x=0.25$, where $x=r/r_{\rm bub}$.
946: Note that both the field just inside and outside the bubble
947: are acted upon by this operator. Applying this operator
948: results in a field that possesses some divergence.
949:
950: (4) changing relative magnetic
951: pressure inside the bubble (done only during the
952: first loop of the iteration process)
953: Plasma $\beta$ parameter is
954: given by $\beta(r) = \beta_{\rm rel}g_{b}(r)+1-g_{r}(r)$,
955: where $\beta_{\rm rel}$
956: is the relative plasma $\beta$ parameter between the ICM and the
957: bubble. The $g_{b}(r)$ term is given by
958: $g_{b}(r)=1+\cos[0.5\pi (x+\Delta_{b} x-1)/\Delta_{b} x]$,
959: where $x=r/r_{\rm bub}$ for $1 < x < 1+\Delta_{b} x$
960: and $g(r)=0$ otherwise.
961: In the above expression, $r_{\rm bub}$ is the bubble
962: radius and $\Delta_{b} x=0.15$.
963:
964: (5) computing the FFT of ${\bf B}({\bf r})$
965: and going back to (1). Iterations are
966: performed until the following convergence criterion is met:
967: $\int (\nabla\cdot {\bf B}({\bf r}))^2 dV dx^2/\int B^{2}dV<10^{-4}$, where
968: integrations are performed for $|x-1|<\Delta x$.\\
969:
970: \indent
971: At this point a convergent and divergence-free ${\bf B}({\bf r})$
972: field has been set up.
973: Inverting such field to obtain the vector potential ${\bf A}({\bf r})$
974: does not result in any loss of information.
975: Note that inverting {\bf B} to get {\bf A} before completing the
976: iteration process would result in a vector potential that
977: (by definition) would give divergence-free magnetic field but the
978: bubble would not be magnetically isolated. \\
979: \indent
980: The equation
981: ${\bf B}=\nabla\times{\bf A}$ is then solved for ${\bf A}({\bf r})$
982: using a variant of a spectral method as follows.\\
983:
984: \indent
985: (1) we Fourier transform ${\bf B}=\nabla\times{\bf A}$\\
986: \indent
987: (2) keeping ${\bf B(k)}$ and ${\bf k}$ constant we rotate
988: the k-space coordinate
989: system first around the $k_{z}$-axis until the projection of
990: the ${\bf k}$ vector on the $(k_{x},k_{y})$ plane coincides with
991: $k_{x}$ axis and then around $k_{y'}$-axis until $k_{z'}$ coincides
992: with the ${\bf k}$ vector. This way the
993: $k_{z''}$-axis is aligned with the ${\bf k}$ vector. That is,
994:
995: \begin{equation}
996: {\bf B^{''}}({\bf k^{''}})\longmapsto R(y',\alpha_{y'})R(z,\alpha_{z}){\bf B}({\bf k}),
997: \end{equation}
998:
999: \noindent
1000: where $R(i,\alpha_{i})$ are rotation matrixes around axis $i$ by angle
1001: $\alpha_{i}$.
1002:
1003: \indent
1004: (3) we invert ${\bf B}''=-i{\bf k}''\times{\bf A}''$
1005: (which, in this frame of reference, is trivial as
1006: ${\bf k^{''}}=(0,0,1)$).
1007:
1008: (4) we rotate ${\bf A}''$ back and in reverse order, i.e.,
1009:
1010: \begin{equation}
1011: {\bf A^{''}}({\bf k^{''}})\longmapsto R(z,-\alpha_{z})R(y',-\alpha_{y'}){\bf A^{''}}({\bf k^{''}}),
1012: \end{equation}
1013:
1014: \noindent
1015: where ${\bf A}({\bf k})$ is the Fourier transform of the required
1016: vector potential. Finally,
1017: we inverse FFT ${\bf A}({\bf k})$ to obtain the vector potential
1018: in real space. \\
1019:
1020: \indent
1021: The code uses the vector potential as its ``magnetic'' variable which
1022: ensures that divergence of magnetic field is strictly zero throughout
1023: the simulation. The reason our initial conditions were not set up
1024: directly in terms of the vector potential is that adding bubble as a
1025: distortion in the fluctuating background vector potential leads to
1026: significant gradients at the boundary between the bubble and the
1027: ICM. These gradients translate into very strong artificial enhancements
1028: in the magnetic field surrounding the bubble as we have seen in our
1029: experiments with that kind of setup.\\
1030: \indent
1031: We also note that the field
1032: set up in this way is not force-free. It is not possible to set up
1033: force-free field for isotropic turbulence case due to mode
1034: coupling. However, initial imbalance in magnetic forces are small
1035: compared to the buoyancy force acting on bubbles. Moreover,
1036: realistic turbulence is not expected to be force-free in any case
1037: as it has to be continuously driven to prevent its decay.\\
1038: \indent
1039: Note that the $\beta$ parameter and magnetic field strength
1040: obtained using the above method would
1041: have arbitrary overall normalization.
1042: The actual normalization of the magnetic flux is obtained by demanding
1043: that the $\beta$ parameter has a certain value inside the bubble, i.e.,
1044: ${\bf B}({\bf r})\longmapsto {\bf B}({\bf r})(P_{\rm gas}\beta_{\rm bub}^{-1}/
1045: \langle P_{B}\rangle_{\rm bub})^{1/2}$,
1046: where $P_{B}=B^{2}/2\mu_{o}$ is magnetic
1047: pressure, $P_{\rm gas}$ is the gas pressure, $\beta_{\rm bub}$ is
1048: the required plasma $\beta$ inside the bubble, and averaging is done
1049: over the bubble volume.
1050: We note that this method works best when applied to high-$\beta$ cases
1051: as then the imbalance between the total pressure in the bubble and the ICM
1052: is smallest.\\
1053: \indent
1054: The final form of our initial conditions for the distribution of
1055: magnetic pressure for the draping and random cases is shown in Figure
1056: 3. This figure shows that our method for generating initial conditions
1057: does not produce any spurious features on the bubble/ICM interface.
1058: The coherence length in the lower panel is even smaller than that used
1059: in the simulations. This has been done to demonstrate robustness of
1060: the method. The map shows natural logarithm of magnetic pressure
1061: in arbitrary units.\\
1062: \indent
1063: Note that magnetic field configurations were generated from the same
1064: random seed, which means that, despite the differences due to
1065: different power spectra, $\beta$ values, helicities, etc., the fields
1066: were as similar as possible. This permits a better comparison of the
1067: consequences of the mentioned differences.
1068:
1069: \label{lastpage}
1070: \end{document}
1071: