1:
2: \documentclass[12pt,preprint]{aastex}
3: %\documentclass[12pt,article]{emulateapj}
4: \shorttitle{Nonlinear Instability of kink oscillations due to shear motions}
5: \shortauthors{Terradas et al.}
6:
7:
8: \begin{document}
9:
10: \title{Nonlinear Instability of kink oscillations due to shear motions}
11:
12:
13: \author{J. Terradas, J. Andries\altaffilmark{1}, M. Goossens}
14:
15: \email{jaume@wis.kuleuven.be}
16: %\email{jesse.andries@wis.kuleuven.be}
17: %\email{marcel.goossens@wis.kuleuven.be}
18:
19: \affil{Centre for Plasma Astrophysics and Leuven Mathematical Modeling and
20: Computational Science Centre \\ Katholieke Universiteit Leuven, Celestijnenlaan
21: 200B, B-3001 Leuven, Belgium
22: }
23:
24: \and
25:
26: \author{I. Arregui, R. Oliver, J. L. Ballester}
27:
28: \affil{Departament de F\'\i sica, Universitat de les Illes Balears,
29: E-07122 Palma de Mallorca, Spain}
30:
31:
32:
33: %\email{inigo.arregui@uib.es}
34: %\email{ramon.oliver@uib.es}
35: %\email{joseluis.ballester@uib.es}
36:
37: \altaffiltext{1}{Postdoctoral Fellow of the National Fund for Scientific
38: Research--Flanders (Belgium) (F.W.O.-Vlaanderen).}
39:
40: \begin{abstract}
41:
42: First results from a high-resolution three-dimensional nonlinear numerical study
43: of the kink oscillation are presented. We show in detail the development of a
44: shear instability in an untwisted line-tied magnetic flux tube. The instability
45: produces significant deformations of the tube boundary. An extended transition
46: layer may naturally evolve as a result of the shear instability at a sharp
47: transition between the flux tube and the external medium. We also discuss the
48: possible effects of the instability on the process of resonant absorption when
49: an inhomogeneous layer is included in the model. One of the implications of
50: these results is that the azimuthal component of the magnetic field of a stable
51: flux tube in the solar corona, needed to prevent the shear instability, is
52: probably constrained to be in a very specific range.
53:
54:
55:
56:
57: \end{abstract}
58:
59: \keywords{MHD --- Sun: corona --- Sun: magnetic fields --- waves}
60:
61:
62: \section{Introduction}
63:
64:
65:
66: Coronal loops and filament threads are magnetic flux tubes whose field lines are
67: anchored to the dense photosphere. Usually these structures are modeled as
68: circular tubes whose properties change mainly in the radial direction with a
69: sharp or a smooth transition. Using the ideal magnetohydrodynamic (MHD)
70: equations the eigenmodes of cylindrical magnetic tubes can be calculated. A
71: fundamental eigenmode of oscillation is the kink mode, with an azimuthal number
72: $m=1$, which produces a transverse displacement of the tube. Examples of the
73: calculations of the kink mode for sharp interfaces can be found in
74: \cite{spruit81,edrob83,cally86} and for smooth transition layers in
75: \cite{gooss92, rudrob02, vand04, terr06}. In the last case, the coupling of
76: compressional fast magnetohydrodynamic (MHD) waves and shear Alfv\'en waves
77: leads to the formation of resonances in the inhomogeneous layers and this
78: mechanism is a possible candidate to explain the damping of transverse coronal
79: loop oscillations \citep{hollyang88, rudrob02, gooss02} and filament threads
80: \citep{arr08}.
81:
82: So far, most of the theoretical studies about the kink mode are in the linear
83: regime and second order perturbations are neglected in the MHD equations,
84: however, there are some observational indications suggesting that this
85: assumption might be not fully justified in all the oscillating loops
86: \citep[][]{terrof04}. Nonlinearity adds new and interesting effects. For
87: example, the ponderomotive force creates a flow along the field lines
88: \citep{rank94,tik95}, and for the standing kink oscillation it tends to
89: accumulate mass at the loop apex \citep[][]{terrof04}. When gas pressure is
90: taken into account, the otherwise unlimited accumulation of mass is prevented
91: by the pressure forces which limit the secular growth and produce saturation in
92: the density.
93:
94: Another possible consequence of the nonlinearity is the generation of
95: instabilities. It is well known that even in the linear regime, a sharp
96: interface between two media in relative motion is liable to the Kelvin-Helmholtz
97: instability (KHI). In the presence of a magnetic field, KHI of a flow parallel
98: to the field lines in a homogeneous medium sets in if the velocity jump exceeds
99: the maximum Alfv\'en speed. The criteria for the KHI for an axial flow is
100: modified in the presence of a boundary layer of finite width, which also induces
101: the possibility of resonant flow instabilities
102: \citep[see][]{holl90,yangh91,and01a,and01b}. The effect of an azimuthal flow is
103: less explored in the literature. \citet{heyvpri83,browpri} showed that azimuthal
104: shear motions in the presence of a smooth transition layer can be KH unstable.
105: The most likely place for the KHI to occur is where the velocity is largest and
106: the magnetic field is perpendicular to the velocity \citep[see for
107: example][]{rank93}. For the fundamental standing kink mode this is precisely the
108: antinode of the velocity, located at half the loop length from the photosphere.
109:
110: In this Letter we report on the nonlinear evolution of the kink oscillation of a
111: tube without twist. The full nonlinear three-dimensional ideal MHD equations are
112: solved numerically. From the simulations we investigate the motions in the
113: magnetic tube and show the first results of the development of the nonlinear
114: shear instability at the tube boundary.
115:
116: \section{Tube Model and Initial Conditions}\label{model}
117:
118: The system under consideration is a compressible plasma that obeys the general
119: equations of MHD. The equilibrium magnetic field is straight, uniform, and
120: pointing in the $z-$direction. In a cylindrical coordinate system with the axis
121: along the magnetic field, the density of the circular tube changes with the
122: radial coordinate from $\rho_{\rm in}$ inside the tube to $\rho_{\rm ex}$ in the
123: coronal medium through an inhomogeneous layer of width $l$. The length of the
124: loop is $L$ and the mean radius $R$. To have magneto-hydrostatic equilibrium the
125: gas pressure is uniform, and the plasma$-\beta$ is chosen to be small, $5\times
126: 10^{-2}$ (we assume a sound speed of $c_{\rm s}=0.2\,v_{\rm Ai}$, where $v_{\rm
127: Ai}$ is the Alfv\'en velocity inside the tube). In this model the role played by
128: the non-zero gas pressure is important since it avoids the continuous
129: accumulation of mass at the tube half length due to the nonlinear ponderomotive
130: force. We apply line-tying conditions at the planes
131: located at $z=0$ and $z=L$ to mimic the anchoring in the photosphere and to
132: obtain a standing mode.
133:
134: The system is perturbed with the spatial structure of the linear fundamental
135: standing kink eigenmode ($m=1$, and longitudinal wavenumber $k_z=\pi/L$) of a
136: magnetic tube with a discontinuous density profile. This allows exciting a
137: single eigenmode (in the linear regime) and avoiding a significant excitation of
138: leaky modes. For this disturbance the radial component of the velocity is
139: continuous across the loop boundary while the azimuthal component has a jump at
140: the boundary. The initial perturbation is the seed of the instability in the
141: nonlinear regime. The amplitude of the initial velocity perturbation at the
142: center of the tube ($x=0$, $y=0$) is $v_0$. Another relevant magnitude in our
143: analysis is the axial component of the vorticity, $\Omega_z=(\nabla\times {\bf
144: {v}})\cdot {\bf {\hat e}_z}$. The initial kink perturbation is a vortex sheet at
145: the tube boundary (see Fig.~\ref{density}a).
146:
147: The radial, azimuthal, and longitudinal dependence of the initial perturbation
148: are transformed to the 3D Cartesian system of our computational box. Given the
149: initial condition the time-dependent nonlinear MHD equations are
150: numerically solved. We have used an explicit high-order numerical scheme
151: ($4th-$order in time and $3th-$order in space) based on the method of lines to
152: solve the equations in conservative form \citep[see][for further
153: details]{terr08}. Due to the resolution requirements the parallelized version of
154: the code has been run in a cluster of machines. We have used a grid resolution
155: of $512 \times 512 \times 100$ points since the small scales are in the $x-y$
156: plane while the solution is smooth in the $z-$direction.
157:
158: \section{Tube Evolution}\label{tube}
159:
160:
161:
162:
163: We have started the analysis of the evolution of the tube with very small
164: amplitudes of the initial perturbation ($v_0\ll v_{\rm Ai}$). From the fully 3D
165: problem we recover the linear results, i.e. the loop oscillates around the
166: equilibrium position as a whole with the kink mode frequency. If instead of a
167: sharp density transition between the tube and the external medium we use a
168: smooth density profile (we have used thick layers to have enough grids points
169: inside the resonant layer), then the tube attenuates due to the energy
170: conversion between the fast MHD waves and the Alfv\'en waves in the layer. We
171: have found that the period and damping time are in good agreement with the
172: theoretical linear predictions.
173:
174: The behavior of the system changes completely when we are in the nonlinear
175: regime. The first effect is the generation of flows along the tube axis produced
176: by the nonlinear terms. However, these flows are much smaller (in the regime
177: considered here) than the local Alfv\'en speed and therefore unable to generate
178: an instability. The dynamics of the system is dominated by the azimuthal flows
179: at the tube boundary.
180:
181: \subsection{Sharp transition layer}
182:
183: We first consider a sharp transition between the tube and the external medium.
184: The evolution of the density and the longitudinal component of the vorticity of
185: a representative case in a weak nonlinear regime is shown in
186: Figure~\ref{density} at different time intervals. To visualize the results of
187: the 3D simulations we concentrate on the plane at half the tube length where we
188: expect the strongest nonlinear effects. The initial perturbation produces a
189: lateral displacement of the tube in the $x-$direction. As in the linear regime
190: the tube starts to oscillate around the initial position, but small length
191: scales quickly develop at the boundary. We can appreciate this, for example,
192: around the points $x=0$ and $y=\pm R$ (see Fig.~\ref{density}b), which is the
193: position where $v_x$ has a maximum jump. These small scale structures grow with
194: time (see Fig.~\ref{density}c) and several rolls form at the loop boundary. Note
195: that around $x=\pm R$ and $y=0$ the density is almost undisturbed because there are
196: no shear motions at this position. At later stages of the evolution
197: (Fig.~\ref{density}d) the small spatial scales are still localized at the
198: boundary and eventually the system reaches a saturation state. As a result of
199: the instability the overall shape of the tube at the boundary has been
200: considerably altered and a rather inhomogeneous layer has been generated. The
201: changes at the tube boundary are also clear in the vorticity. In the early phase
202: the initial vortex sheet, located at the boundary, shows small and localized
203: deformations (Fig.~\ref{density}b). Since the tube is oscillating the flow
204: changes sign in each oscillation, and the vortex sheet evolves in a complex
205: way, showing a very undulated shape (Figs.~\ref{density}c and \ref{density}d).
206:
207: \subsection{Smooth transition layer}
208:
209: Now we assume that the equilibrium already has a transition region. The main
210: difference with respect to the previous case is that the resonant absorption
211: process induces shear motions at the inhomogeneous layer due to phase mixing.
212: The results are represented in Figure~\ref{density1} for two different widths
213: of the layer at a given time instant (same as in Fig.~\ref{density}c). We see
214: that the instability is also present (Fig.~\ref{density1}a), but now it is less
215: developed in comparison with the sharp transition case (Fig.~\ref{density}c).
216: We also see (Fig.~\ref{density1}b) that the thicker the layer the slower the
217: growth-rate of the instability (in Fig.~\ref{density1}b the instability is
218: still not present). Since the boundary of the tube eventually changes its
219: shape due to the instability, a question that arises is how this affects the
220: process of resonant absorption. We have calculated the damping rate (damping
221: time over period) of the tube from the simulations for the two cases
222: considered in Figure~\ref{density1} (calculating the displacement of the
223: central point of the tube). The numerical estimates give values around 1.9 for
224: $l=0.5R$ and 1.2 for $l=R$, while the linear values of the damping rates based
225: on eigenmode calculations are 2.4 and 1.1, respectively. The differences are
226: small, indicating that the instability, for the particular parameters
227: considered here, does not change much the efficiency of the energy conversion.
228:
229: \section{Discussion and Conclusions}\label{concl}
230:
231: The numerical results shown in this Letter indicate that the shear motions
232: involved in the kink oscillations of a magnetic tube might be unstable. The
233: instabilities are found to create small length scales in the azimuthal direction
234: and grow rapidly in time. For a cylindrical discontinuous interface between two
235: homogeneous stationary rotating fluids the following growth rates (imaginary
236: part of the frequency) of the Kelvin
237: Helmholtz instability can be readily derived: \begin{equation}\label{kh}
238: \omega_i^2=\frac{\rho_{\rm in}\rho_{\rm ex}}{(\rho_{\rm in}+\rho_{\rm
239: ex})^2}\frac{m^2}{R^2} 4 v_0^2-2 k_z^2 \frac{B_0^2}{\mu (\rho_{\rm
240: in}+\rho_{\rm ex})}. \end{equation} Here $2v_0$ is the amplitude of the velocity
241: shear at the boundary. The derivation involves the assumption that the azimuthal
242: length scales are much smaller than the longitudinal ones (which is the case for
243: the instabilities that formed in our study). Clearly the background equilibrium
244: for which equation~(\ref{kh}) is obtained is very different from the shear
245: motions associated to the kink mode oscillations, which depend on time, $\phi$
246: and $z$. Nevertheless, the small scales and the localization of the
247: instabilities in the azimuthal direction, and the fast growth rates, suggest
248: that equation~(\ref{kh}) may be considered as a local analysis and we need not
249: to worry about the azimuthal and time dependence of the shear motions. For the
250: parameters considered in \S\ref{tube}, the first unstable mode corresponds to
251: $m=5$ (which satisfies the assumption of azimuthal localization), and its growth
252: rate is $1/\omega_i\approx 5.0\tau_{\rm A}$. The important point here is that it
253: is smaller than the kink period, around $16.3\tau_{\rm A}$ (satisfying the
254: assumption of localization in time). Moreover, this calculation also clarifies
255: that the most unstable modes will have small azimuthal length scales but large
256: longitudinal length scales, since this minimizes the stabilizing force induced
257: by the bending of the field lines. However, the large longitudinal length scales
258: prevent to interpret equation~(\ref{kh}) as a local approximation in $z$ since
259: the velocity shear is less than $2v_0$ throughout most of the loop. Growth rates
260: are thus expected to be somewhat smaller, and instability will set in only for
261: azimuthal length scales smaller than those predicted by equation~(\ref{kh}). An
262: appropriate analysis thus needs to take into account the longitudinal variation
263: of the shear profile and hence necessarily involves a 2D model.
264:
265: We have found that the evolution of the tube is very sensitive to the amplitude
266: of the initial perturbation. As expected, the larger the amplitude of the
267: initial perturbation the faster the development of the instability. By changing
268: the initial amplitude of the perturbation in a broad range, an extended
269: transition layer (even much larger than the developed in Fig.~\ref{density}) may
270: naturally evolve as the result of the shear instability of a sharp transition
271: between the flux tube and the external medium. In addition, for very large
272: amplitudes the tube might show severe deformations, a wake can even form behind
273: the tube and it can interact with the main body in each oscillation, the final
274: shape of the tube being rather irregular.
275:
276: When an inhomogeneous layer between the tube and the environment is included
277: then the motions are characterized by phase-mixed scales and the instability is
278: also present. Nevertheless, the instability develops more slowly than in the
279: sharp transition case. This is probably due to the fact that the thicker the
280: layer the later the generation of small length scales due to the phase mixing
281: process, and thus the onset of the instability. This situation seems to be
282: related to the development of the KHI for torsional Alfv\'en waves ($m=0$)
283: described by \citet{browpri} \citep[see also][]{walker}. Interestingly, for the
284: regime studied here (basically thick layers) the attenuation of the central part
285: of the tube due to resonant absorption is not significantly altered by the
286: changes at the boundary due to the shear instability.
287:
288: In the context of coronal loops an immediate question that arises from the
289: results presented here is why, up to now, there is no clear evidence of such
290: instability from the observation of oscillating loops. There are several partial
291: answers. Magnetic twist, not included in our model, might decrease or even
292: suppress the instability since the presence of a magnetic field component along
293: the flow stabilizes the KHI. Several examples of the stabilizing effect of a
294: helical magnetic field component can be found in rising tubes in the convection
295: zone \citep[see for example][]{fer96} or in stellar jets. On the other hand, a
296: tube with very large azimuthal magnetic field is subject to the instabilities
297: of the linear pinch (typically for a twist larger than $2.5\pi$, although it
298: depends on the details of the equilibrium and boundary conditions). Therefore,
299: the azimuthal component of the magnetic field of a stable flux tube in the solar
300: corona is probably constrained to be in a specific range (being $2.5\pi$ an
301: upper bound for the twist). Other factors that could explain the absence of
302: observational evidences of the instability are that it is not spatially resolved
303: with the current spatial resolution of the telescopes or simply that the
304: amplitude of the oscillating loops is not strong enough to develop the
305: instability. However, the tube displacement produced by the initial perturbation
306: in our simulations is of the order of and even smaller than the observed
307: amplitudes of oscillation in loops, so this last possibility seems to be ruled
308: out.
309:
310: We have given a qualitative description of the shear instability, but a
311: quantitative analysis (with improved numerical resolution) about the growth
312: rates of the modes, and a detailed study of the effect of the instability on the
313: damping rates under different regimes is still needed. In addition, the
314: equilibrium configuration should be improved in several aspects. For example, a
315: more accurate model should incorporate a realistic variation of the density and
316: temperature from the photosphere to the corona. However, the inclusion of a
317: twisted magnetic field for the reasons mentioned before seems to be the most
318: relevant aspect that needs to be addressed.
319:
320: \acknowledgements J. Terradas is grateful to the Research Council fellowship
321: F/06/65 of the KUL. Spanish funding provided under grants AYA2006-07637 and
322: PCTIB-2005GC3-03 is acknowledged. We are grateful to Fernando Moreno-Insertis
323: for his comments and suggestions.
324:
325:
326:
327: \begin{thebibliography}{}
328:
329:
330:
331: \bibitem[Andries \& Goossens(2001a)]{and01a} Andries, J., \& Goossens, M. 2001a,
332: \aap, 368, 1083
333:
334: \bibitem[Andries \& Goossens(2001b)]{and01b} Andries, J., \& Goossens, M. 2001b,
335: \aap, 375, 1100
336:
337: \bibitem[Arregui et al.(2008)]{arr08} Arregui, I., Terradas, J., Oliver, R., \&
338: Ballester, J. L. 2008, \apj, 682, L141
339:
340: \bibitem[Browning \& Priest(1984)]{browpri} Browning, P. K., \& Priest, E. R.
341: 1984, \aap, 131, 283
342:
343: \bibitem[Cally(1986)]{cally86} Cally, P. S. 1986, \solphys, 103, 277
344:
345: \bibitem[Edwin \& Roberts(1983)]{edrob83} Edwin, P. M., \& Roberts, B. 1983,
346: \solphys, 88, 179,
347:
348:
349:
350: \bibitem[Goossens et al.(1992)]{gooss92}Goossens, M., Hollweg, J. V., \&
351: Sakurai, T. 1992, \aap, 138, 233
352:
353:
354:
355: \bibitem[Goossens et al.(2002)]{gooss02}Goossens, M., Andries, J., \&
356: Aschwanden, M. J. 2002, \aap, 394, L39
357:
358:
359: \bibitem[Heyvaerts \& Priest(1983)]{heyvpri83}Heyvaerts, J., \& Priest, E. R.
360: 1983, \aap, 117, 220
361:
362:
363:
364: \bibitem[Hollweg \& Yang(1988)]{hollyang88} Hollweg, J. V., \& Yang, G. 1988,
365: \jgr, 93, A6, 5423
366:
367: \bibitem[Hollweg et al.(1990)]{holl90}Hollweg, J.V., Yang, G., Cadez, V.M., \&
368: Gakovic, B. 1990, \apj, 349, 335
369:
370:
371:
372: \bibitem[Moreno-Insertis \& Emonet(1996)]{fer96} Moreno-Insertis, F., \& Emonet,
373: T. 1996, \apjl, 472, L53
374:
375:
376: \bibitem[Rankin et al.(1993)]{rank93} Rankin, R., Harrold, B. G., Samson, J. C.,
377: \& Frycz, P. 1993, \jgr, 98, A4, 5839
378:
379: \bibitem[Rankin et al.(1994)]{rank94} Rankin, R., Frycz, P., Tikhonchuk, V.
380: T., \& Samson, J. C. 1994, \jgr, 99, 21291
381:
382:
383: \bibitem[Ruderman \& Roberts(2002)]{rudrob02}Ruderman, M. S., \& Roberts, B.
384: 2002, \apj , 577, 129
385:
386:
387: \bibitem[Spruit(1981)]{spruit81}Spruit, H. C. 1982, \solphys, 75, 3
388:
389:
390: \bibitem[Terradas \& Ofman(2004)]{terrof04} Terradas, J., \& Ofman, L. 2004,
391: \apj, 610, 523
392:
393:
394: \bibitem[Terradas et al.(2006)]{terr06} Terradas, J., Oliver, R., \& Ballester,
395: J. L. 2006, \apj, 642, 533
396:
397:
398: \bibitem[Terradas et al.(2008)]{terr08} Terradas, J., Oliver, R., \& Ballester,
399: J. L., Keppens, R. 2008, \apj, 675, 875
400:
401:
402: \bibitem[Tikhonchuk et al.(1995)]{tik95} Tikhonchuk, V. T., Rankin, R., Frycz,
403: P., \& Samson, J. C. 1995, Phys. Plasmas, 2, 501
404:
405:
406: \bibitem[Van Doorsselaere et al.(2004)]{vand04}Van Doorsselaere, T., Andries,
407: J., Poedts, S., \& Goossens, M. 2004, \apj, 606, 1223
408:
409: \bibitem[Walker(1981)]{walker} Walker, A. D. M. 1981, Planet. Space. Sci, 29,
410: 10, 1119
411:
412: \bibitem[Yang \& Hollweg(1991)]{yangh91}Yang, G., \& Hollweg, J.V. 1991, \jgr,
413: 96, 13807
414:
415:
416: \end{thebibliography}
417:
418: \clearpage
419:
420: \begin{figure*}[!ht]
421: \center{
422: \includegraphics[width=6.4cm]{f1a.eps}
423: \hspace{1cm}
424: \includegraphics[width=6.4cm]{f1b.eps}
425: \\
426: \includegraphics[width=6.4cm]{f1c.eps}
427: \hspace{1cm}
428: \includegraphics[width=6.4cm]{f1d.eps}
429: }
430: \caption{
431: \small The top half of each panel displays the density (symmetric respect to
432: $y=0$) and the bottom half the vorticity (antisymmetric respect to $y=0$) at
433: $z=L/2$ for
434: different times. In this simulation the following parameters have been used:
435: $L=10R$, $\rho_{\rm in}/\rho_{\rm ex}$=3, $v_0=0.1v_{\rm Ai}$, and a full domain of
436: $[-6R,-6R]\times[-6R,6R]\times[0,10R]$. Lengths are normalized to the loop
437: radius, $R$ (typically of the order of $4\,000$\,km), velocities to the internal Alfv\'en velocity, $v_{\rm
438: Ai}=B_0/\sqrt{\mu \rho_{\rm in}}$ (of the order of $1\,000\,{\rm km\,s^{-1}}$), and time to the Alfv\'en crossing time,
439: $\tau_{\rm A}=R/v_{\rm Ai}$. {\em This figure is available as an mpeg animation
440: in the electronic edition of the Journal.}} \label{density} \end{figure*}
441:
442: \begin{figure*}[!ht]
443: \center{
444: \includegraphics[width=6.4cm]{f2a.eps}
445: \hspace{1cm}
446: \includegraphics[width=6.4cm]{f2b.eps}
447: }
448: \caption{ \small Density and
449: vorticity (for the same time instant as in Fig.~\ref{density}c) for two
450: different widths of the inhomogeneous layer connecting the tube and the external
451: medium, {\bf
452: a)}
453: $l=0.5R$ and {\bf b)} $l=R$. All other parameters are the same as in
454: Figure~\ref{density}.} \label{density1}
455: \end{figure*}
456:
457: \end{document}
458: