1: \documentclass[12pt,preprint]{aastex}
2:
3: %%\usepackage{latexsym}
4: %%\usepackage{epsfig}
5: %\usepackage{appendix}
6:
7: %\slugcomment{}
8:
9: \shorttitle{Transverse oscillations of two coronal loops}
10: \shortauthors{M. Luna et al.}
11: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
12: \begin{document}
13:
14: \title{Transverse oscillations of two coronal loops}
15:
16: \author{M. Luna\altaffilmark{1}, J.
17: Terradas\altaffilmark{1,2}, R. Oliver\altaffilmark{1}, and J.L.
18: Ballester\altaffilmark{1}}
19: %\email{manuel.luna@uib.es}
20: %\email{jaume.terradas@uib.es}
21: %\email{ramon.oliver@uib.es}
22: %\email{dfsjlb0@uib.es}
23:
24: \altaffiltext{1}{Departament de F\'{\i}sica, Universitat de les Illes Balears,
25: 07122 Palma de Mallorca, Spain. Email: manuel.luna@uib.es,
26: jaume.terradas@uib.es, ramon.oliver@uib.es and joseluis.ballester@uib.es}
27: \altaffiltext{2}{Centre for Plasma Astrophysics, Katholieke Universiteit Leuven,
28: Celestijnenlaan 200B, B-3001 Leuven, Belgium}
29:
30: \begin{abstract}
31: We study transverse fast magnetohydrodynamic waves in a system of two coronal
32: loops modeled as smoothed, dense plasma cylinders in a uniform magnetic field.
33: The collective oscillatory properties of the system due to the interaction
34: between the individual loops are investigated from two points of view. Firstly,
35: the frequency and spatial structure of the normal modes are studied. The system
36: supports four trapped normal modes in which the loops move rigidly in the
37: transverse direction. The direction of the motions is either parallel or
38: perpendicular to the plane containing the axes of the loops. Two of these modes
39: correspond to oscillations of the loops in phase, while in the other two they
40: move in antiphase. Thus, these solutions are the generalization of the kink mode
41: of a single cylinder to the double cylinder case. Secondly, we analyze the
42: time-dependent problem of the excitation of the pair of tubes. We find that
43: depending on the shape and location of the initial disturbance, different normal
44: modes can be excited. The frequencies of normal modes are accurately recovered
45: from the numerical simulations. In some cases, because of the simultaneous
46: excitation of several eigenmodes, the system shows beating.
47: \end{abstract}
48:
49: \keywords{Sun: corona--magnetohydrodynamics (MHD)--waves}
50:
51: \section{Introduction}
52:
53: Transverse coronal loop oscillations have been studied in recent years after
54: being observed for the first time by the Transition Region and Coronal Explorer
55: (TRACE) in 1998 \citep[see for example][]{aschwanden1999, aschwanden2002,
56: schrijver2002, verwichte2004}. These oscillations were initiated shortly after a
57: solar flare that disturbed the loops. Much before the TRACE observations, the
58: theory of loop oscillations was developed
59: \citep{spruit1981,edwin&roberts1983,cally1983} and the different kinds of
60: oscillations were studied. The observed transverse motions have been interpreted
61: in terms of the excitation of the fast magnetohydrodynamic (MHD) fundamental
62: kink mode.
63:
64: Most analytical studies about transverse loop oscillations have only considered
65: the properties of individual loops, but in many cases loops belong to complex
66: active regions where they are usually not isolated. For example,
67: \citet{verwichte2004} reported complex transverse motions of loops in a
68: post-flare arcade. In particular, loops D and E \citep[see Fig.~1
69: of][]{verwichte2004} show bouncing displacements with oscillations in phase and
70: antiphase that repeat in time. The same behavior of the movements in a loop
71: bundle can be observed in the event of March 23, 2000 of the compact flare
72: recorded by TRACE \citep[see][]{schrijver2002}. Additionally, antiphase
73: oscillations of adjacent loops have also been reported in
74: \citet{schrijver2000,schrijver2002}. These observations suggest that there are
75: interactions between neighboring loops and that the dynamics of loop systems is
76: not simply the sum of the dynamics of the individual loops.
77:
78: On the other hand, it is currently debated whether active region coronal loops
79: are monolithic \citep{aschwanden2005} or multi-stranded
80: \citep{klimchuk2006,deforest2007}. The strands are considered as mini-loops for
81: which the heating and plasma properties are approximately uniform in the
82: transverse direction. In the multi-stranded model it is suggested that loops are
83: formed by bundles of several tens or several hundreds of physically related
84: strands \citep{klimchuk2006}. \citet{lopezfuentes2006} suggest that these
85: strands wrap around each other in complicated ways due to the random motion of
86: the foot points in the solar surface. These models explain the constant width
87: and symmetry of the loops as observed with current X-ray and EUV telescopes.
88:
89: From the observations, it is thus necessary to study not only individual loops
90: but also how several loops or strands can oscillate as a whole, since their
91: joint dynamics can be different from those of a single loop. Little work has
92: been done on composite structures so far. \citet{berton1987} studied the
93: magnetohydrodynamic normal modes of a periodic magnetic medium, while other
94: authors, for example \citet{bogdan&fox1991, keppens1994}, analyzed the
95: scattering and absorption of acoustic waves by bundles of magnetic flux tubes
96: with sunspot properties. \citet{kris93, kris94} studied numerically the
97: propagation of fast waves in two slabs unbounded in the longitudinal direction.
98: On the other hand, in \citet{diaz2005} the oscillations of the prominence thread
99: structure were investigated. These authors found that in a system of equal
100: fibrils the only non-leaky mode is the symmetric one, with all fibrils
101: oscillating in spatial phase with the same frequency. Finally, \citet{luna}
102: found that in a system of two coronal slabs, the symmetric and antisymmetric
103: modes can be trapped and that an initial disturbance can excite these modes,
104: which are readily detectable after a brief transient phase. If the fundamental
105: symmetric mode and the antisymmetric first harmonic are excited at the same
106: time, a beating phenomenon takes place. In such a case, the loops interchange
107: energy periodically. In any case, all these authors found that a system of
108: several loops behave differently from an individual loop.
109:
110: Here we consider a more complex system than those studied in previous works. Our
111: model consists of two parallel cylinders, without gravity and curvature. This
112: model allows us to study the interaction between loops and the collective
113: behavior of the system. We study the normal modes and also solve the
114: time-dependent problem of the excitation of transverse coronal loop
115: oscillations. We concentrate on a planar pulse excitation and compare the
116: results of the simulations with the eigenmodes of the configuration.
117:
118:
119:
120: This paper is organized as follows. In \S \ref{loop_model} the loop model is
121: presented. In \S \ref{normal_modes_sec} the normal modes are calculated and the
122: frequencies and spatial distribution of the eigenfunctions are studied. The
123: time-dependent problem is considered in \S \ref{time_dependent_analysis}, where
124: the velocity and pressure field distribution are analyzed for different
125: incidence angles of the initial perturbation. In \S \ref{2d_beating} the loop
126: motions are studied and the beating is analyzed. Finally, in \S
127: \ref{discussion_conclusions} the results are summarized and the main conclusions
128: are drawn.
129:
130: \section{Equilibrium configuration and basic equations}
131: \label{loop_model}
132:
133: The simplest way to investigate the interaction of a set of loops is to consider
134: a pair of loops in slab geometry. In \citet{luna} this model was studied in
135: detail using the ideal MHD equations and the zero-$\beta$ plasma limit. Here a
136: more realistic model is considered. The equilibrium configuration consists of a
137: system of two parallel homogeneous straight cylinders of radius $a$, length
138: $L$, and separation between centers $d$ (see Fig.~~\ref{sketch}). We assume the
139: following equilibrium plasma density profile:
140: \begin{eqnarray*}
141: \rho_\mathrm{0} (x,y)=\left\{
142: \begin{array}{lll}
143: \rho_\mathrm{e}, & \;{\textrm{\normalsize if $r_\mathrm{1}> a$ and $r_\mathrm{2}
144: > a$},} \\
145: \rho_\mathrm{i}, & \;{\textrm{\normalsize if
146: $r_\mathrm{1} \leq a$ or $r_\mathrm{2} \leq a$},}
147: \end{array}
148: \right.
149: \end{eqnarray*}
150: where $x$, $y$ are the Cartesian coordinates and $r_\mathrm{1}$ and
151: $r_\mathrm{2}$, defined as $r_\mathrm{1}^2=(x+d/2)^2+y^2$ and
152: $r_\mathrm{2}^2=(x-d/2)^2+y^2$, are the distances from the point $(x,y)$ to the
153: centers of the left and right loops, respectively. In the previous expression
154: $\rho_\mathrm{e}$ and $\rho_\mathrm{i}$ are the densities in the external medium
155: or corona and the loop ($\rho_\mathrm{i}>\rho_\mathrm{e}$), respectively.
156: Hereafter, we use a density contrast $\rho_\mathrm{i}/\rho_\mathrm{e}=10$.
157: \clearpage
158: \begin{figure}[!ht]
159: \center
160: \resizebox{8cm}{!}{\includegraphics{f1.eps}}
161: \caption{
162: \label{sketch}
163: \small Sketch of the model, that consists of two straight cylinders immersed in
164: the coronal medium. Hereafter the total pressure and the velocity fields are
165: plotted in the $xy$-plane, shown as a white slice.}
166: \end{figure}
167: \clearpage
168: The loop centers lie on the $x$-axis at $x=d/2$ for the right loop and $x=-d/2$
169: for the left loop. The configuration is symmetric with respect to the $yz$-plane
170: and the $z$-axis is parallel to the axes of the cylinders. The tubes and the
171: environment are permeated by a uniform magnetic field along the $z$-direction
172: ($\mathbf{B} = B_\mathrm{0} \mathbf{e_\mathrm{z}}$). The Alfv\'en speed,
173: $v_\mathrm{A} = B_\mathrm{0}/\sqrt{\mu \rho}$, takes the value $v_\mathrm{A i}$
174: inside the loop and $v_\mathrm{Ae}$ in the surrounding corona ($v_\mathrm{A i} <
175: v_\mathrm{A e}$).
176:
177: Linear perturbations about this equilibrium for a perfectly conducting fluid in
178: the zero-$\beta$ limit can be readily described using the ideal MHD equations in
179: Cartesian coordinates. The velocity is denoted by $\mathbf{v}= \left(v_x, v_y, 0
180: \right)$ and $\mathbf{B}= \left(B_x, B_y, B_z \right)$ is the magnetic field
181: perturbation. We have assumed a $z$-dependence of the perturbations of the form
182: $e^{-i k_z z}$. In this model we consider the photosphere as two infinitely
183: dense planes located at $z=\pm L/2$. The loop feet are anchored in these planes
184: and so the fluid velocity is zero at these positions (this is the so-called
185: line-tying effect). This condition produces a quantization of the $z$-component
186: of the wave-vector to $k_z=n \pi/L$. Hereafter we concentrate on the fundamental
187: mode, with $n=1$. The total pressure perturbation is
188: \begin{equation}\label{total_pressure}
189: p_\mathrm{T}=\frac{B_\mathrm{0}}{\mu} B_z,
190: \end{equation}
191: and coincides with the magnetic pressure perturbation in the zero-$\beta$ limit.
192:
193: \section{Normal modes}
194: \label{normal_modes_sec}
195:
196: Analytical solutions to the eigenvalue problem of the previous model (assuming
197: a temporal dependence of the form $e^{i \omega t}$) are very difficult to derive
198: due to the geometry of the system. The methods used for a single cylinder
199: \citep[see][]{edwin&roberts1983} cannot be applied to the study of two tubes.
200: One way to solve the problem is to use scattering theory, see for example
201: \citet{edwin&roberts1983}, \citet{bogdan&knolker1990},
202: \citet{bogdan&zeibel1985}, \citet{bogdan&fox1991} and \citet{keppens1994}.
203: Another way is to solve the eigenvalue problem given by the ideal MHD equations
204: numerically. We have used this approach and we have done the computations with
205: the PDE2D code \citep{sewell}. We have used bicylindrical orthogonal
206: coordinates, defined by the transformation
207: \begin{eqnarray}
208: x=\frac{d/2 \sinh v}{\cosh v - \cos u} ~ ,~ y=\frac{d/2 \sin u}{\cosh v - \cos
209: u} ~ ,
210: \end{eqnarray}
211: where $0 \le u < 2\pi$ and $-\infty < v < \infty$. The loop boundaries are
212: coordinate surfaces at $v=\pm \mathrm{arcsinh} ~\frac{d}{2 a}$, where the
213: positive and negative signs correspond to the right and left tubes,
214: respectively. We impose the restriction that the solutions tend to zero at large
215: distances from the cylinders, i.e. we seek trapped mode solutions.
216:
217: We find four collective fundamental trapped modes (see Fig.~\ref{normal_modes}).
218: There are other harmonics but we concentrate on the fundamental kink-like modes
219: because they produce the largest transverse displacement of the loops axes. The
220: velocity field is more or less uniform in the interior of the loops, and so they
221: move basically as a solid body, while the external velocity field has a more
222: complex structure. The four velocity field solutions have a well defined
223: symmetry with respect to the $y$-axis. In Figure \ref{normal_modes}a, we see
224: that the velocity field inside the tubes lies in the $x$-direction and is
225: symmetric with respect to the $y$-axis. We call this mode $S_x$, where $S$
226: refers to the symmetry of the velocity field and the subscript $x$ refers to the
227: direction of the velocity inside the tube. The same nomenclature is used for the
228: other modes. In Figure \ref{normal_modes}b the velocity inside the cylinders is
229: mainly in the $x$-direction and antisymmetric with respect to the $y$-axis, so
230: we call this mode $A_x$. Similarly, in Figure \ref{normal_modes}c the velocity
231: lies in the $y$-direction and is symmetric with respect to the $y$-axis, while
232: it is antisymmetric in Figure \ref{normal_modes}d. Hence, we call these modes
233: $S_y$ and $A_y$, respectively. The pressure field of the $A_x$ and $S_y$ modes
234: is symmetric with respect to the $y$-axis, while that of the $S_x$ and $A_y$
235: modes is antisymmetric.
236: \clearpage
237: \begin{figure}[!ht]
238: \center
239: \mbox{\hspace{0.cm}\hspace{8.cm}\hspace{8.cm}}\vspace{-2.cm}
240: \includegraphics[width=7.5cm]{f2a_color.eps}\hspace{0.8cm}\includegraphics[width=7.5cm]{f2b_color.eps}\vspace{-4.cm}
241: \mbox{\hspace{-0.cm}\hspace{8.cm}\hspace{8.cm}}
242: \includegraphics[width=7.5cm]{f2c_color.eps}\hspace{0.8cm}\includegraphics[width=7.5cm]{f2d_color.eps}\vspace{-1.cm}
243: \caption{
244: Total pressure perturbation (color field) and velocity field (arrows) of the
245: fast four collective normal modes (plotted in the $xy$-plane, see
246: Fig.~\ref{sketch}). The modes are labeled as (a) $S_x$, with the loops moving
247: in phase in the $x$-direction; (b) $A_x$, the tubes move in the $x$-direction
248: but in antiphase; (c) $S_y$, the tubes move in the $y$-direction in phase; and
249: finally (d) $A_y$, the loops move in antiphase in the $y$-direction. Here, the
250: loop radii are $a=0.1L$ and the distance between centers is $d=6 a$.
251: }
252: \label{normal_modes}
253: \end{figure}
254: \clearpage
255: The frequencies of oscillation of these four modes as a function of the loop
256: separation, $d$, are displayed in Figure \ref{w_vs_d}. For large separations
257: between the tubes, the modes tend to the kink mode of an individual loop (see
258: dotted line). On the other hand, for smaller separations, they split in four
259: branches associated to the four oscillatory modes described before. The
260: splitting effect was noticed in \citet{diaz2005} and \citet{luna} in a
261: configuration of several slabs. The frequency difference between the modes
262: increases when the interaction between the loops becomes stronger, i.e. when
263: the distance between them is small. When the loops are very close ($d\sim 2a$),
264: the frequencies of the $S_x$ and $A_y$ modes tend to the value $\omega=3.33/
265: \tau_\mathrm{A i}$, which is similar to the internal cut-off frequency,
266: $\omega_\mathrm{c i}=k_z v_\mathrm{A i}=3.14/ \tau_\mathrm{A i}$ (the difference
267: is only around 6$\%$). Here $\tau_\mathrm{A i}$ is the Alfv\'en transit time,
268: defined as $\tau_\mathrm{A i}=L/v_\mathrm{A i}$. On the other hand, in this
269: limit, the $S_y$ and $A_x$ frequencies are quite large in comparison to the kink
270: mode frequency.
271:
272: It is interesting to note that when both tubes move symmetrically in the
273: $x$-direction, i.e. in the $S_x$ mode, the fluid between follows the loops
274: motion (see Fig.~\ref{normal_modes}a). On the other hand, when the loops
275: oscillate antisymmetrically, i.e. in the $A_x$ mode, the intermediate fluid is
276: compressed and rarefied (see Fig.~\ref{normal_modes}b), producing a more forced
277: motion than that of the symmetric mode. This is the reason for the $S_x$ ($A_x$)
278: mode having a smaller (larger) frequency than that of the individual loop. For
279: the modes polarized in the $y$-direction the behavior is somehow similar,
280: although in this case the antisymmetric mode (see Fig.~\ref{normal_modes}d) has
281: a lower frequency than the symmetric mode (see Fig.~\ref{normal_modes}c). When
282: one of the loops moves upwards the surrounding fluid near the other loop moves
283: downwards. This helps to push the other loop in this direction and produces the
284: antisymmetric motion. The situation is different for the $S_y$ mode, for which
285: the direction of motion of the surrounding fluid is opposite to that of the
286: other tube. This explains why the frequency of the $A_y$ solution is smaller
287: than that of the $S_y$ mode.
288:
289: \clearpage
290: \begin{figure}[!ht]
291: \center{
292: {\resizebox{15cm}{!}{\includegraphics{f3.eps}}}
293: }
294: \caption{
295: Frequency, $\omega$, as a function of the separation between cylinders, $d$, for
296: a density enhancement $\rho_\mathrm{i}/\rho_\mathrm{e}=10$ and loop radius
297: $a=0.1 L$. The lower horizontal thick dotted line is the internal cut-off
298: frequency, $\omega_\mathrm{c i} = 3.14/\tau_\mathrm{Ai}$. The thin dotted line
299: is the kink frequency ($\omega= 4.12/\tau_\mathrm{A i}$) of an individual loop.
300: The calculated frequencies from the time-dependent results in \S
301: \ref{time_dependent_normal_modes} are also plotted as triangles ($S_x$), squares
302: ($A_y$), asterisks ($S_y$), and diamonds ($A_x$).
303: \label{w_vs_d}
304: }
305: \end{figure}
306: \clearpage
307: \section{Time-dependent analysis: numerical simulations}
308: \label{time_dependent_analysis}
309:
310: The initial perturbation that we have used when solving numerically the ideal
311: MHD equations is a planar pulse in the velocity field of the form
312: \begin{equation}\label{initial_condition}
313: \mathbf{v}_\mathrm{0} = \mathbf{K} ~ e^{\left[-\mathbf{K}\cdot
314: (\mathbf{r}-\mathbf{r}_\mathrm{0}) / \Delta \right]^2} ,
315: \end{equation}
316: i.e. a Gaussian profile (of width $\Delta=0.2 L$ centered at
317: $\mathbf{r}_\mathrm{0} = \left( d/2 , 0, 0\right)$) and direction of propagation
318: along $\mathbf{K}=-\left( \cos\alpha , \sin \alpha, 0\right)$, $\alpha$ being
319: the angle between the wavevector and the $x$-axis. Here $\mathbf{K}$ also
320: defines the initial polarization of $\mathbf{v}$, which is perpendicular to the
321: planar pulse. The initial value of the magnetic field perturbation is zero, and
322: thus the same applies to the total pressure perturbation. In the
323: simulations a spatial domain of size $30a \times 30a$ is used and the boundaries
324: are far from the loops. These boundaries are open, which ensures that the
325: numerical reflections are negligible.
326:
327: In Figures \ref{t_evol_90}, \ref{t_evol_0}, and \ref{t_evol_45} three examples
328: of the time evolution are shown for $\alpha= 90^\circ, 0^\circ$ and $45^\circ$,
329: respectively, and for a fixed distance between loops $d=6 a$, identical to the
330: one used in Figure~\ref{normal_modes} (see the time evolution in Movie 1, Movie
331: 2, and Movie 3).
332: These three cases illustrate the time evolution of the system after a
333: perturbation, which consists of two regimes: the transient and the stationary
334: phases. The stationary phase is characterized by oscillations in one or several
335: fundamental trapped normal modes (see \S \ref{normal_modes_sec}). On the other
336: hand, in the transient phase there are leaky modes and internal reflections and
337: refractions.
338: \clearpage
339: \begin{figure}[!ht]
340: \center
341: \mbox{\hspace{-1.cm}\hspace{8.cm}\hspace{8.cm}}\vspace{-0.cm}
342: \includegraphics[width=7.5cm]{f4a_color.eps}\hspace{0.8cm}\includegraphics[width=7.5cm]{f4b_color.eps}
343: \mbox{\hspace{-0.4cm}\hspace{8.cm}\hspace{8.cm}}\vspace{-0.5cm}
344: \includegraphics[width=7.5cm]{f4c_color.eps}\hspace{0.8cm}\includegraphics[width=7.5cm]{f4d_color.eps}
345: \caption{
346: Time-evolution of the velocity field (arrows) and total pressure field (colored
347: contours), for a separation between loops $d=6 a$ and an initial pulse with an
348: angle $\alpha=90^\circ$. The two circles mark the positions of the loops at
349: $t=0$. The panels show different evolution times. In (a) the initial condition
350: over the velocity field is represented. In (b) the velocity and pressure field
351: shortly after the initial disturbance, that is, during the transient phase, are
352: shown. Both tubes are excited at the same time. In panel (b) the tubes are in
353: the transient phase. In panels (c) and (d) the system oscillates in the
354: stationary phase with the $S_y$ normal mode. This time evolution is also
355: available as an mpeg animation in Movie 1. }
356: \label{t_evol_90}
357: \end{figure}
358: \clearpage
359: In Figure \ref{t_evol_90} (see Movie 1) the time evolution for the
360: $\alpha=90^\circ$ initial disturbance is shown, for which, the pulse front lies
361: along the $x$-axis and excites the $v_y$ component. The loops are perturbed at
362: the same time (as can be appreciated in Fig.~\ref{t_evol_90}b) and as a
363: consequence they oscillate symmetrically. In Figure \ref{t_evol_90}b the system
364: is in the transient phase, characterized by internal reflections related with
365: the emission of leaky modes. The external medium has not relaxed yet. Finally,
366: the system reaches the stationary phase (see Figs. \ref{t_evol_90}c and
367: \ref{t_evol_90}d) and oscillates with the $S_y$ trapped mode (compare the
368: velocity field and the pressure distribution with Fig.~\ref{normal_modes}c).
369:
370: \clearpage
371: \begin{figure}[!ht]
372: \center
373: \mbox{\hspace{-1.cm}\hspace{8.cm}\hspace{8.cm}}\vspace{-0.cm}
374: \includegraphics[width=7.5cm]{f5a_color.eps}\hspace{0.8cm}\includegraphics[width=7.5cm]{f5b_color.eps}
375: \mbox{\hspace{-0.4cm}\hspace{8.cm}\hspace{8.cm}}\vspace{-0.5cm}
376: \includegraphics[width=7.5cm]{f5c_color.eps}\hspace{0.8cm}\includegraphics[width=7.5cm]{f5d_color.eps}
377: \caption{
378: Same as Figure~\ref{t_evol_90} for an initial pulse with an angle
379: $\alpha=0^\circ$. Here the stationary phase is governed by a superposition of
380: the $S_x$ and $A_x$ normal modes. The whole time evolution is presented in Movie
381: 2.
382: }
383: \label{t_evol_0}
384: \end{figure}
385: \clearpage
386:
387: In Figure \ref{t_evol_0} (and Movie 2), the time evolution
388: for the $\alpha=0^\circ$ initial disturbance is shown. Now the pulse is centered
389: on the right loop (see Fig.~\ref{t_evol_0}a) and excites the $v_x$ component. In
390: Figure \ref{t_evol_0}b, the pulse reaches the left tube and passes through it,
391: the system still being in the transient phase. On the other hand, in Figures
392: \ref{t_evol_0}c and \ref{t_evol_0}d the system oscillates in the stationary
393: phase. It is interesting to note that this particular initial disturbance does
394: not excite the left loop; neither at $t=0$ nor during the transient phase.
395: Nevertheless, the oscillatory amplitude in the left loop grows with time in the
396: stationary phase, while the amplitude in the right loop decreases in the time
397: interval shown in Figures \ref{t_evol_0}c and \ref{t_evol_0}d (see also Movie
398: 2). Then, it is clear that the left tube acquires its movement through the
399: interaction with the right loop, i.e. by a transfer of energy from the right
400: loop to the left loop. This process is reversed and repeated periodically: once
401: the left loop has gained most of the energy retained by the loops system, so
402: that the right loop is almost at rest, the left tube starts giving away its
403: energy to the right cylinder, and so on. This is simply a beating phenomenon,
404: that can be explained in terms of the normal modes excited in this numerical
405: simulation. In fact, the initial disturbance excites the $S_x$ and $A_x$ modes
406: with the same amplitude and for this reason the excitation is initially maximum
407: on the right tube and zero on the left tube. A more detailed discussion about
408: this issue is given in \S \ref{2d_beating}.
409:
410: \clearpage
411: \begin{figure}[!h]
412: \center
413: \mbox{\hspace{-1.cm}\hspace{8.cm}\hspace{8.cm}}\vspace{-0.cm}
414: \includegraphics[width=7.5cm]{f6a_color.eps}\hspace{0.8cm}\includegraphics[width=7.5cm]{f6b_color.eps}
415: \mbox{\hspace{-0.4cm}\hspace{8.cm}\hspace{8.cm}}\vspace{-0.5cm}
416: \includegraphics[width=7.5cm]{f6c_color.eps}\hspace{0.8cm}\includegraphics[width=7.5cm]{f6d_color.eps}
417: \caption{
418: Same as Figure~\ref{t_evol_90} for an initial pulse with an angle
419: $\alpha=45^\circ$. Here the stationary phase is governed by a superposition of
420: the $S_x$, $A_x$, $S_y$, and $A_y$ normal modes. The whole time evolution is
421: presented in Movie 3. Movie 4 contains the time evolution
422: for much larger times.
423: }
424: \label{t_evol_45}
425: \end{figure}
426: \clearpage
427:
428: Finally, we discuss the results for an excitation with $\alpha=45^\circ$. This
429: simulation is the most complex and general of all (see Movie 3).
430: As we can see in Figure \ref{t_evol_45}a now both components of the velocity are
431: excited. In Figure \ref{t_evol_45}b the initial pulse reaches the left tube and
432: passes through it, but only leaky modes are excited. In Figures \ref{t_evol_45}c
433: and \ref{t_evol_45}d the system oscillates in the stationary phase, which is a
434: combination of the four modes $S_x$, $A_x$, $S_y$ and $A_y$. As in the previous
435: case, there is beating but now it is present in both the $x$- and $y$-velocity
436: components. Like for the previous simulation, the left loop is almost still
437: until the stationary phase (see also dotted curves in
438: Figs.~\ref{time_evol_example_1}a and \ref{time_evol_example_1}d) despite that in
439: this simulation the pulse directly hits the left loop without the obstacle of
440: the right loop. In \S \ref{2d_beating} details about the behavior of the system
441: are given.
442:
443: Once we know the general features of the excitation of the two cylinders we can
444: perform a parametric study of the effect of the distance between the loops and
445: also the angle of excitation on the loops motion.
446:
447: \subsection{Effect of the distance between loops}
448:
449: \label{time_dependent_normal_modes}
450:
451: We generate an initial disturbance with an angle of $45^\circ$ for different
452: distances $d$ and measure the velocity in the loops as a function of time. From
453: this information we can extract the frequencies of oscillation. As we have seen,
454: since the velocity field inside the loops is more or less uniform (see
455: Fig.~\ref{normal_modes}), it is enough to measure the velocity at the center of
456: the loops to describe their global motion. The reason for choosing the initial
457: disturbance with $\alpha=45^\circ$ is that it excites the four normal modes, so
458: that with a single simulation we can measure their frequencies.
459:
460:
461: \clearpage
462: \begin{figure}[!ht]
463: \center
464: \mbox{\hspace{-0.cm}\hspace{4.cm}\hspace{4.cm}}\vspace{-0.cm}
465: \includegraphics[width=4.5cm]{f7a.eps}\includegraphics[width=4.5cm]{f7b.eps}\includegraphics[width=4.5cm]{f7c.eps}
466: \mbox{\hspace{-0.cm}\hspace{4.cm}\hspace{4.cm}}\vspace{-0.8cm}
467: \includegraphics[width=4.5cm]{f7d.eps}\includegraphics[width=4.5cm]{f7e.eps}\includegraphics[width=4.5cm]{f7f.eps}
468: \caption{
469: (a) $x$-component and (d) $y$-component of the velocity at the center of the
470: right (solid line) and left (dotted line) loops for the numerical simulation of
471: Figure~\ref{t_evol_45} (i.e. with an initial incidence angle $\alpha=45^\circ$).
472: With the method explained in \S \ref{time_dependent_normal_modes} the normal
473: mode velocities are extracted and plotted in (b) for the $S_x$ (solid line) and
474: the $A_x$ (three-dot-dashed line) modes and in (e) for the $S_y$ (dashed line)
475: and $A_y$ (dot-dashed line) modes. The corresponding power spectra are plotted
476: with the same line styles in (c) and (f). Power maxima allow us to determine the
477: frequency of the normal modes from the numerical simulation.
478: }
479: \label{time_evol_example_1}
480: \end{figure}
481:
482: \begin{figure}[!ht]
483: \center
484: \mbox{\hspace{-0.cm}\hspace{4.cm}\hspace{4.cm}}\vspace{-0.1cm}
485: \includegraphics[width=4.5cm]{f8a.eps}\includegraphics[width=4.5cm]{f8b.eps}\includegraphics[width=4.5cm]{f8c.eps}
486: \mbox{\hspace{-0.cm}\hspace{4.cm}\hspace{4.cm}}\vspace{-0.8cm}
487: \includegraphics[width=4.5cm]{f8d.eps}\includegraphics[width=4.5cm]{f8e.eps}\includegraphics[width=4.5cm]{f8f.eps}
488: \caption{
489: Same as Figure~\ref{time_evol_example_1} for an initial incidence angle
490: $\alpha=70^\circ$.
491: }
492: \label{time_evol_example_2}
493: \end{figure}
494: \clearpage
495:
496: In Figures \ref{time_evol_example_1}a and \ref{time_evol_example_1}d the $x$-
497: and $y$-components of the velocity at the center of each loop are plotted. In
498: these figures we see that, after a very brief transient characterized by
499: short-period oscillations, the system oscillates with the sum of normal modes.
500: The frequencies of the modes are quite similar, and it is difficult to resolve
501: them. Although the frequencies of these modes are present in the time-dependent
502: signal, this information cannot be easily extracted from the data because in
503: these simulations the maximum evolution time (which is determined by the
504: numerical damping) is $T=6\tau_\mathrm{Ai}$. With this maximum time we have a
505: frequency resolution $2/T \simeq 0.3/\tau_\mathrm{Ai}$, but, as evidenced by
506: Figure~\ref{w_vs_d}, the difference in frequency between the eigenmodes is
507: typically less than $0.1/\tau_\mathrm{Ai}$ so we have not enough frequency
508: resolution. For this reason we extract the frequencies with another method
509: considering that the velocity field is the addition of normal modes with
510: symmetric and antisymmetric spatial functions with respect to the $y$-axis. We
511: measure the velocity in the loop centers $(x=-d/2,y=0)$ and $(x=d/2,y=0)$, i.e.
512: two symmetric points with respect to $x=0$. Then, the sum of both measured
513: velocities in these points is twice the symmetric velocity. Dividing this
514: velocity by two we obtain the $v_x$ of the $S_x$ mode and the $v_y$ of the $S_y$
515: mode in these points. On the other hand, the subtraction of the measured
516: velocities is twice the antisymmetric velocity. Similarly, dividing this
517: velocity by two we obtain the $v_x$ of the $A_x$ mode and the $v_y$ of the $A_y$
518: mode. The obtained mode velocities are plotted in
519: Figures~\ref{time_evol_example_1}b and \ref{time_evol_example_1}e. Next, we
520: compute a periodogram of these signals (plotted in
521: Figs.~\ref{time_evol_example_1}c and \ref{time_evol_example_1}f), from which the
522: frequencies of the collective modes are determined. The periodogram is preferred
523: over the FFT as it allows to precisely identify these frequencies.
524:
525: The above procedure has been applied to numerical simulations for different
526: separations between loops and the frequencies of the four fundamental eigenmodes
527: have been obtained. The calculated frequencies have been superimposed to the
528: normal mode values in Figure~\ref{w_vs_d} using symbols. A good agreement
529: between the normal mode calculations and the time-dependent results can be
530: appreciated.
531:
532: \subsection{Effect of the incidence angle} \label{study_alpha}
533:
534:
535: We next study the evolution of the system for different incidence angles,
536: $\alpha$, of the planar pulse and a fixed distance between loops ($d=6 a$). Some
537: examples of the time evolution have already been discussed and shown in
538: Figures~\ref{t_evol_90}, \ref{t_evol_0}, and \ref{t_evol_45}. The mode
539: excitation depends on the width, $\Delta$, the incidence angle, $\alpha$, and
540: the position, $\mathbf{r}_\mathrm{0}$, of the initial disturbance, but here we
541: only consider the dependence on the incidence angle. The angles considered in
542: our simulations vary from $\alpha=0^\circ$ to $90^\circ$ with steps of
543: $5^\circ$. Using the method of \S \ref{time_dependent_normal_modes} it is also
544: possible to extract the amplitude of each normal mode, given by the amplitude of
545: the sinusoidal oscillations in the stationary phase. Two examples of the
546: extraction method are plotted in Figure \ref{time_evol_example_1} for
547: $\alpha=45^\circ$ and Figure \ref{time_evol_example_2} for $\alpha=70^\circ$.
548:
549: \clearpage
550: \begin{figure}[!th]
551: \center
552: \includegraphics[width=10.cm]{f9.eps}
553: \caption{
554: Amplitudes of the four normal modes as a function of the incidence angle
555: $\alpha$. The separation between loops is $d=6 a$.
556: } \label{amplitudes}
557: \end{figure}
558: \clearpage
559:
560: In Figure \ref{amplitudes} the amplitude of the four collective modes is plotted
561: as a function of the incidence angle. The behavior of the amplitude can be
562: divided in two parts, namely for $0^\circ\le\alpha\le50^\circ$ and for
563: $50^\circ<\alpha\le90^\circ$. In the first interval the amplitudes of the $S_x$
564: and $A_x$ modes are more or less equal (see Figs.~\ref{time_evol_example_1}b and
565: \ref{time_evol_example_1}e as an example) and can be approximated by $0.3 \cos
566: \alpha$. The same occurs for the amplitudes of the $S_y$ and $A_y$ modes, which
567: vary roughly as $0.29 \sin \alpha$. In the second interval these amplitudes can
568: be quite different (see Figs. \ref{time_evol_example_2}b and
569: \ref{time_evol_example_2}e as an example) and the $S_x$, $A_x$, and $A_y$
570: amplitudes go to zero at $\alpha=90^\circ$. On the other hand, the $S_y$
571: amplitude increases and reaches its maximum value at $\alpha=90^\circ$.
572: Furthermore, for $\alpha=0^\circ$ the amplitudes of the $S_x$ and $A_x$ modes
573: have a maximum around $0.3$ while the amplitudes of $S_y$ and $A_y$ modes are
574: zero. This is because for $\alpha=0^\circ$ the initial disturbance drives the
575: $x$-component of the velocity and so only the $S_x$ and $A_x$ modes are excited.
576: Similarly, for the perturbation with $\alpha=90^\circ$ only the $S_y$ and $A_y$
577: modes can be excited, although the shape of our initial perturbation prevents
578: the $A_y$ mode from being driven and so the $S_y$ mode reaches the largest
579: amplitude of all modes. On the other hand, the excitation of the antisymmetric
580: modes requires the initial disturbance to hit the right and left loops at
581: different times. For this reason, the amplitudes of these modes decrease with
582: $\alpha$. In fact, when $\alpha=90^\circ$ this time difference is zero since
583: both loops are excited at the same time and the amplitude of the $A_x$ and $A_y$
584: vanishes. Finally, it is interesting to note that for $\alpha=45^\circ$ the four
585: modes are excited with almost the same amplitude.
586:
587: \section{Study of the loops motions: beating}
588: \label{2d_beating}
589:
590: As we have shown in the previous section, loop motions can be very complex. This
591: is even more clear in Movie 4, in which the time-evolution for a simulation with
592: identical parameters to those used in Figure~\ref{t_evol_45} but for much larger
593: times is represented. In \S \ref{time_dependent_analysis} we mentioned that the
594: initial disturbance excites the right loop but does not perturb the left loop.
595: After a short time the left tube starts to oscillate due to the interaction with
596: the right one. At this stage, the right loop oscillates with the velocity
597: polarization of the initial pulse, whereas the left tube oscillates in a
598: direction perpendicular to that of the initial disturbance. The reason for the
599: complexity of the loops motions is that their oscillations are a superposition
600: of four normal modes with different velocity polarizations, parities, and
601: frequencies.
602:
603: \clearpage
604: \begin{figure}[!ht]
605: \center
606: \includegraphics[width=8cm]{f10a.eps}
607: \includegraphics[width=8cm]{f10b.eps}
608: \caption{
609: Temporal variation of the velocity components (a) $v_x$ and (b) $v_y$ at the
610: center of the right loop (solid line) and left loop (dashed line). These results
611: correspond to the simulation shown in Figure \ref{t_evol_45} and illustrate the
612: beating of the pair of loops. Damping caused by numerical dissipation causes a
613: slight decrease of the amplitude during the numerical simulation. The
614: time-evolution is also available as an mpeg animation in Movie 4. }
615: \label{beating_vxy}
616: \end{figure}
617: \clearpage
618: We next analyze this case in detail. The $x$- and $y$-components of the velocity
619: at the center of the loops are represented in Figures \ref{beating_vxy}a and
620: \ref{beating_vxy}b, respectively. There is a clear beating, characterized by the
621: periodic interchange of the $x$- and $y$-components of the velocity between the
622: loops. The two velocity components are modulated in such a way that $v_x$
623: reaches its maximum value in the left tube and becomes zero in the right tube at
624: the same time (around $t\simeq40 \tau_\mathrm{A i}$). This process is reversed
625: at $t\simeq80 \tau_{A i}$ and repeats periodically.
626:
627: The loops motions can be studied theoretically. In the stationary phase, during
628: which the system oscillates in the normal modes $S_x$, $A_x$, $S_y$, and $A_y$,
629: the velocity field components are
630: \begin{eqnarray}
631: \label{ss_beat_x}
632: v_x (x,y,t)&=& C_x^\mathrm{S} (x,y) \cos \left( \omega_x^\mathrm{S} t+
633: \phi_x^S\right)+C_x^\mathrm{A} (x,y)\cos \left( \omega_x^\mathrm{A} t+
634: \phi_x^A\right) ,\\
635: \label{ss_beat_y}
636: v_y (x,y,t)&=& C_y^\mathrm{S} (x,y)\cos \left( \omega_y^\mathrm{S} t+
637: \phi_y^S\right)+C_y^\mathrm{A} (x,y)\cos \left( \omega_y^\mathrm{A} t+
638: \phi_y^\mathrm{A}\right).
639: \end{eqnarray}
640: The $\mathrm{S}$ and $\mathrm{A}$ superscripts refer to the symmetric and
641: antisymmetric modes, respectively. The functions $C_x^\mathrm{S}$,
642: $C_x^\mathrm{A}$, $C_y^\mathrm{S}$, and $C_y^\mathrm{A}$ represent the spatial
643: distribution of the four normal modes (see Fig.~\ref{normal_modes}) and their
644: amplitude accounts for the energy deposited by the initial disturbance in each
645: of them. The normal mode frequencies are represented by their frequencies,
646: $\omega$, while $\phi$ are their initial phases.
647:
648: Let us turn our attention to the results in
649: Figure~\ref{time_evol_example_1}. In the loops centers the symmetric and
650: antisymmetric modes have a very similar amplitude (see also
651: Fig.~\ref{beating_vxy} for $\alpha=45^\circ$), which means that $C_x^\mathrm{S}
652: (d/2,0) = C_x^\mathrm{A} (d/2,0)$. Then, taking into account the parity of
653: $C_x^\mathrm{S}$ and $C_x^\mathrm{A}$ about $x=0$, we have $C_x^\mathrm{S}
654: (-d/2,0) = -C_x^\mathrm{A} (-d/2,0)$. Inserting these expressions into
655: equations~(\ref{ss_beat_x}) and (\ref{ss_beat_y}) evaluated at the loop centers
656: we obtain
657: \begin{eqnarray}
658: \label{right_velocity_approax}
659: \mathbf{v}_\mathrm{right} (t) &=&\Big(C_x \cos ( \frac{\omega_x^\mathrm{A}-\omega_x^\mathrm{S}}{2}~t
660: )\cos ( \frac{\omega_x^\mathrm{A}+\omega_x^\mathrm{S}}{2}~t
661: ) , C_y \cos (\frac{\omega_y^\mathrm{A}-\omega_y^\mathrm{S}}{2}~t )\cos
662: (\frac{\omega_y^\mathrm{A}+\omega_y^\mathrm{S}}{2}~t) \Big) ,\\
663: \label{left_velocity_approax}
664: \mathbf{v}_\mathrm{left} (t) &=&-\Big(C_x \sin (\frac{\omega_x^\mathrm{A}-\omega_x^\mathrm{S}}{2}~t
665: ) \sin (\frac{\omega_x^\mathrm{A}+\omega_x^\mathrm{S}}{2}~t
666: ) ,C_y \sin (\frac{\omega_y^\mathrm{A}-\omega_y^\mathrm{S}}{2}~t
667: ) \sin (\frac{\omega_y^\mathrm{A}+\omega_y^\mathrm{S}}{2}~t
668: ) \Big).
669: \end{eqnarray}
670: where $\mathbf{v}_\mathrm{right}$ and $\mathbf{v}_\mathrm{left}$ are the
671: velocity of the right and left loop, respectively. We have defined $C_x=2
672: C_x^\mathrm{S}(d/2,0)$ and $C_y=2 C_y^\mathrm{S}(d/2,0)$ and have assumed
673: $\phi_x^\mathrm{S}=\phi_x^\mathrm{A}=\phi_y^\mathrm{S}=\phi_y^\mathrm{A}=0$
674: because the initial disturbance is over the right loop. The beating curves shown
675: in Figure~\ref{beating_vxy} are accurately described by these equations.
676:
677: These formulae contain products of two harmonic functions. Then, the
678: temporal evolution during the stationary phase is governed by four periods: the
679: two oscillatory periods,
680: \begin{eqnarray}
681: \label{oscillating_period_x}
682: T_x={ 4 \pi \over \omega_x^\mathrm{A}+\omega_x^\mathrm{S}} ,\\
683: \label{oscillating_period_y}
684: T_y={ 4 \pi \over \omega_y^\mathrm{A}+\omega_y^\mathrm{S}} ,
685: \end{eqnarray}
686: {giving the mean periods of the time signal;} and two beating periods,
687: \begin{eqnarray}
688: \label{beating_period_x}
689: T_{bx}={ 4 \pi \over \omega_x^\mathrm{A}-\omega_x^\mathrm{S}} ,\\
690: \label{beating_period_y}
691: T_{by}={ 4 \pi \over \omega_y^\mathrm{A}-\omega_y^\mathrm{S}}.
692: \end{eqnarray}
693: giving the periods of the envelop of the time signal. To apply these expressions
694: to the numerical simulation of Figure~\ref{time_evol_example_1} we insert the
695: values of $\omega_x^\mathrm{S}$, $\omega_x^\mathrm{A}$, $\omega_y^\mathrm{S}$,
696: and $\omega_y^\mathrm{A}$ for $d=6 a$ into
697: equations~(\ref{oscillating_period_x})--(\ref{beating_period_y}). Then we obtain
698: $T_x=1.52\tau_\mathrm{A i}$, $T_y=1.52\tau_\mathrm{A i}$,
699: $T_{\mathrm{b}x}=159.96\tau_\mathrm{A i}$, and $T_{\mathrm{b}y}=479.88
700: \tau_\mathrm{A i}$. The two oscillating periods are equal because the frequency
701: distribution is approximately symmetric around the central value (the kink
702: frequency of an individual loop) for sufficiently large distances (see Fig.
703: \ref{w_vs_d}). The two beating periods derived from the numerical simulations
704: match very well these values because Figure~\ref{beating_vxy} gives
705: $T_{\mathrm{b}x}/4\simeq 40 \tau_\mathrm{A i}$ and $T_{\mathrm{b}y}/4\simeq 120
706: \tau_\mathrm{A i}$.
707:
708: The $\pi/2$ phase difference between $\mathbf{v}_\mathrm{right}$ and
709: $\mathbf{v}_\mathrm{left}$ (see Figs.~\ref{time_evol_example_1}a and
710: \ref{time_evol_example_1}d) is due to the fact that our system of two loops
711: basically behaves as a pair of driven-forced oscillators. Considering $v_x$, the
712: left loop has initially a $\pi/2$ delay with respect to the right loop because
713: it behaves as a driven oscillator and the left one like a forced oscillator.
714: After half beating period, $T_\mathrm{b x}/2$, the roles are exchanged and left
715: loop becomes the driver and right one the forced oscillator. The $y$-components
716: of $\mathbf{v}_\mathrm{right}$ and $\mathbf{v}_\mathrm{left}$ exhibit the same
717: behavior (see Fig.~\ref{time_evol_example_1}d). This was already shown by
718: \citet{luna} in the case of two slabs.
719:
720: As we have seen, the polarization of the oscillations changes with time (see
721: Movie 4 for an example). In the beating range, we can see this from the
722: equations by calculating the scalar product of the velocity at the loop centers,
723: \begin{eqnarray}\nonumber{}
724: \mathbf{v}_\mathrm{right} \cdot \mathbf{v}_\mathrm{left}= -C_x^2 \sin \Big(2
725: (\omega_x^\mathrm{A}-\omega_x^\mathrm{S}) t\Big)\sin \Big(2 (\omega_x^\mathrm{A}+\omega_x^\mathrm{S})t\Big)\\
726: -C_y^2 \sin \Big(2
727: (\omega_y^\mathrm{A}-\omega_y^\mathrm{S})t\Big)\sin \Big(2 (\omega_y^\mathrm{A}+\omega_y^\mathrm{S})t\Big) .
728: \end{eqnarray}
729: This product gives the relative polarization of the loop oscillations and we see
730: that is zero at $t=0$ and approximately zero for sufficiently small times. Thus,
731: the left loop does not oscillate initially and it starts to oscillate
732: perpendicularly to the right loop during the first oscillations. This feature is
733: shown in Figure \ref{t_evol_45} and Movie~3 and Movie~4.
734:
735:
736: Similar beating features are recovered for incidence angles of the initial
737: disturbance in the range $0^\circ \le \alpha \lesssim 50^\circ$ (what we call
738: the beating range). The cause of this behavior is explained by
739: Figure~\ref{amplitudes}: for these values of $\alpha$ a similar amount of energy
740: is deposited in the $S_x$ and $A_x$ modes, so the beating of the $v_x$ component
741: is possible. Obviously, an analogous argument applies to $v_y$. This is not the
742: case for $50^\circ \lesssim \alpha \le 90^\circ$ for which the symmetric and
743: antisymmetric modes receive different amounts of energy from the initial
744: excitation and then their relative amplitude is different (see Fig.
745: \ref{time_evol_example_2} for an example). Simulations for angles
746: $\alpha>50^\circ$ do not clearly exhibit beating and the trajectories of the
747: loops are much more complex than those in the beating range.
748:
749: \section{Discussion and conclusions}
750: \label{discussion_conclusions}
751:
752:
753:
754: In this work, we have investigated the transverse oscillations of a system of
755: two coronal loops. We have considered the low-$\beta$, ideal MHD equations and
756: have studied both the normal modes of this configuration and the time-dependent
757: problem. The results of this work can be summarized as follows:
758:
759: \begin{itemize}
760:
761: \item The system has four fundamental normal modes, somehow similar to the kink
762: mode of a single cylinder. These modes are collective, i.e. the system
763: oscillates with a unique frequency, different for each mode. When arranged in
764: increasing frequency the modes, are $S_x$, $A_y$, $S_y$, and $A_x$, where
765: $S$($A$) stand for symmetric (antisymmetric) velocity oscillations with respect
766: to the plane in the middle of the two loops and $x$ ($y$) stands for the
767: polarization of motions. These modes produce transverse motions of the tubes, so
768: they are kink-like modes.
769:
770: \item We have studied the eigenfrequencies as a function of the separation of
771: loops. For large distances between cylinders, they behave as a two independent
772: loops, i.e. the frequency tends to the individual kink mode frequency. When the
773: distance decreases the frequency splits in four branches, two of which
774: correspond to the $S_x$ and the $A_y$ modes and are below the frequency of the
775: individual tube, and the other two are related to the $S_y$ and $A_x$ modes and
776: lie above the kink frequency of a single tube. Roughly speaking, there is a
777: certain parallelism between our system of two loops and a mechanical system of
778: two coupled oscillators with $f$ degrees of freedom, which has $f\times n$
779: collective normal modes. This parallelism is possible because a slab or a
780: cylinder oscillating with the kink mode moves more or less like a solid body.
781: The number of translational degrees of freedom one for an individual slab
782: ($f=1$) and two for an individual loop. Then, the parallelism with the
783: mechanical system tells us that in a two slab system there are two collective
784: normal modes \citep{luna}, while in a two cylinder system there are four.
785:
786: \item For small distances between the loops, the frequency of the $S_x$ and
787: $A_y$ modes is quite similar and tends to the internal cut-off frequency. This
788: is different to the behavior in a configuration of two slabs \citep[see][]{luna}
789: where, for small distances between the slabs, the system behaves as an
790: individual loop of double width. On the other hand, for the two cylinders the
791: frequency is much lower than that of a loop with double radius.
792:
793: \item We have also studied the temporal evolution of the system after an initial
794: planar pulse. We have shown that, depending on the incidence angle, the system
795: oscillates with a combination of several normal modes. The frequencies of
796: oscillation calculated from the numerical simulations agree very well with the
797: normal mode eigenfrequencies.
798:
799: \item In the beating range ($0^\circ \le \alpha \lesssim 50^\circ$), the system
800: beats in the $x$- and $y$-components of the velocity and the left and right
801: loops are $\pi/2$ out of phase for each velocity component. They behave as a
802: pair of driven-forced oscillators, with one loop giving energy to the other and
803: forcing its transverse oscillations. The role of the two loops is interchanged
804: every half beating period. On the other hand, for perturbations with
805: $\alpha>50^\circ$ the loops motions are much more complex than those in the
806: beating range. The phase lag cannot be clearly appreciated and it strongly
807: depends on the incidence angle of the initial pulse.
808:
809: \end{itemize}
810:
811: From this work, we conclude that a loop system clearly shows a collective
812: behavior, its fundamental normal modes being quite different from the kink mode
813: of a single loop. These collective normal modes are not a combination of
814: individual loop modes. This suggests that the observed oscillations reported in
815: \citet{aschwanden1999, aschwanden2002, schrijver2002, verwichte2004} are in fact
816: caused by one or a superposition of some collective modes. Moreover, the
817: antiphase movements reported by \citet{nakariakov1999} can be easily explained
818: using our model. The same applies to the bounce movement of loops D and E
819: studied in \citet{verwichte2004}. These motions can be interpreted by assuming
820: that there is beating between the loops produced by the simultaneous excitation
821: of the fundamental $S_x$ and $A_x$ modes.
822:
823: It should be noted that the observations indicate a very rapid damping of
824: transverse oscillations, such that in a few periods the amplitude of oscillation
825: of the loops is almost zero. This fast attenuation may hide the beating
826: produced by the simultaneous excitation of several normal modes of the system.
827: However, in some situations, for example, for small loop separations and high
828: density contrast loops, the beating periods decrease. Then, under such
829: conditions the beating could be detectable in the observation interval. In
830: any case, the beating is just one particular collective behavior, and there is
831: always interaction between the individual loops in short time scales (typically
832: of the order of $2d/v_\mathrm{A e}$). The consequences of this interaction are
833: the collective normal modes of the system. The presence of the normal modes
834: could be also clear from a frequency analysis. Unfortunately, due to the
835: temporal resolution, these observations do not allow us to perform such
836: analysis, but the frequency extraction method derived in \S
837: \ref{time_dependent_normal_modes} is suitable to be applied to the
838: observations.
839:
840:
841: Finally, in order to have more realistic models additional effects need to be
842: included. In this work, we have studied two loops with exactly the same density
843: and radii, so the next step is to analyze the behavior of a system of $n$ loops
844: with different properties. This study could also be extended to understand
845: the possible effect of internal structure (multi-stranded models and small
846: filling factors) on the oscillating loops by considering a set of very thin
847: tubes with different densities and radii. We expect that the dynamical behavior
848: and frequencies of multi-stranded loops differ from those of the monolithic
849: models. Preliminary work has been done by \citet{arregui2007} who have
850: studied the effects on the dynamics of the possibly unresolved internal
851: structure of a coronal loop composed of two very close, parallel, identical
852: coronal slabs in Cartesian geometry.
853:
854:
855:
856: \acknowledgments
857: M. Luna is grateful to the Spanish Ministry of Science and Education for an FPI
858: fellowship, which is partially supported by the European Social Fund. He also
859: thanks the members of the Departament of Mathematics of K. U. Leuven for their
860: warm hospitality during his brief stay at this University and for their worthy
861: comments. J. Terradas thanks the Spanish Ministry of Science and Education for
862: the funding provided under a Juan de la Cierva fellowship. The authors
863: acknowledge the Spanish Ministry of Science and Education and the Conselleria
864: d'Economia, Hisenda i Innovaci\'o of the Goverment of the Balearic Islands for
865: the funding provided under grants AYA2006-07637, PRIB-2004-10145 and
866: PCTIB-2005-GC3-03, respectively. We are grateful to the referee for his/her
867: comments and suggestions that helped to improve the manuscript.
868:
869:
870: \begin{thebibliography}{}
871: \bibitem[Arregui et al.(2007)]{arregui2007} Arregui, I., Terradas, J., Oliver,
872: R., \& Ballester, J. L. 2007, \aap, 466, 1145
873: \bibitem[Aschwanden et al.(1999)]{aschwanden1999} Aschwanden, M. J., Fletcher,
874: L., Schrijver, C. J., \& Alexander, D. 1999, \apj, 520, 880
875: \bibitem[Aschwanden et al.(2002)]{aschwanden2002} Aschwanden, M. J., De Pontieu,
876: B., Schrijver, C. J., \& Title, A. M. 2002, \solphys, 206, 99
877: \bibitem[Aschwanden et al.(2005)]{aschwanden2005} Aschwanden, M. J., \&
878: Nightingale, W. 2005, \apj, 633, 499
879: \bibitem[Berton and Heyvaerts(1987)]{berton1987} Berton, R., \& Heyvaerts, J.
880: 1987, \solphys, 109, 201
881: \bibitem[Bogdan and Zweibel(1985)]{bogdan&zeibel1985} Bogdan, T. J., \& Zweibel,
882: E. G. 1985, \apj, 298, 867
883: \bibitem[Bogdan and Kn\"olker(1991)]{bogdan&knolker1990} Bogdan, T. J., \&
884: Kn\"olker, M. 1990, \apj, 369, 219
885: \bibitem[Bogdan and Fox(1991)]{bogdan&fox1991} Bogdan, T. J., \& Fox, D. C.
886: 1991, \apj, 379, 758
887: \bibitem[Cally(1986)]{cally1983} Cally, P. S. 1986, \solphys, 103, 277
888: \bibitem[DeForest(2007)]{deforest2007} DeForest, c. E. 2007, \apj, 661, 532
889: \bibitem[Edwin and Roberts(1983)]{edwin&roberts1983} Edwin, P. M., \& Roberts,
890: B. 1983, \solphys, 88, 179
891: \bibitem[Hudson et al.(2004)]{hudson2004} Hudson, H. S., \& Warmuth, A. 2004,
892: \apj, 614, L85
893: %\bibitem[D\'{\i}az et al.(2004)]{diaz2004} D\'{\i}az, A. J., Oliver, R. \&
894: %Ballester, J. L., Roberts, B. 2004, \aap, 424, 1055-1064
895: \bibitem[D\'{\i}az et al.(2005)]{diaz2005} D\'{\i}az, A. J., Oliver, R., \&
896: Ballester, J. L. 2005, \aap, 440, 1167-1175
897: %\bibitem[Goedbloed (1983)]{goedbloed1983} Goedbloed, J. P. 1983, Lecture notes
898: %on Ideal Magnetohydrodynamics, Rijnhuizen Rep., 76
899: \bibitem[Keppens et al.(1994)]{keppens1994} Keppens, R., Bogdan, T. M., \&
900: Goossens, M. 1994, \apj, 436, 372
901: \bibitem[Klimchuk(2006)]{klimchuk2006} Klimchuk, J. A. 2006, \solphys, 234,41
902: \bibitem[Leveque(2002)]{leveque} Leveque, R. J. 2002, Finite Volume Methods for
903: Hyperbolic Problems, Cambridge University Press
904: \bibitem[L\'opez Fuentes et al.(2006)]{lopezfuentes2006} L\'opez Fuentes, M.
905: C., \& Klimchuk, J. A. 2006, \apj, 639, 459
906: \bibitem[Luna et al.(2006)]{luna} Luna M., Terradas J., Oliver R., \& Ballester J. L. 2006, \aap, 457, 1071-1079
907: \bibitem[Murawski(1993)]{kris93} Murawski, K. 1993, Acta Astronomica, 43, 2,
908: 161
909: \bibitem[Murawski and Roberts(1994)]{kris94} Murawski, K., \& Roberts, B. 1994,
910: \solphys, 151, 305
911: \bibitem[Nakariakov et al.(1999)]{nakariakov1999} Nakariakov, V. M., Ofman, L.,
912: DeLuca, E. E., Roberts, B., \& Davila, J. M. 1999, Science, 285, 862
913: %\bibitem[Ogrodowczyk and Murawski (2007)]{ogro2007} Ogrodowczyk, R. \&
914: %Murawski, K. 2007, \aap, 461, 1133
915: \bibitem[Sewell(2005)]{sewell} Sewell, G. 2005, The Numerical Solution of Ordinary and
916: Partial Differential Equations, Wiley-Interscience
917: \bibitem[Schrijver and Brown (2000)]{schrijver2000} Schrijver, C. J., \& Brown,
918: D. S. 2000, \apj, 537, L69
919: \bibitem[Schrijver et al.(2002)]{schrijver2002} Schrijver, C. J., Aschwanden, M.
920: J., \& Title, A. M. 2002, \solphys, 206,69
921: \bibitem[Spruit(1981)]{spruit1981} Spruit, H. C. 1982, \solphys, 75, 3
922: \bibitem[Terradas et al.(2005)]{terradas2005} Terradas, J., Oliver, R., \&
923: Ballester, J. L. 2005, \aap, 441, 371
924: \bibitem[Verwichte et al.(2004)]{verwichte2004} Verwichte, E., Nakariakov, V. M.,
925: Ofman, L., \& Deluca, E. E. 2004, \solphys, 223, 77
926: %\bibitem[Wang et al. (2003)]{wang2003} Wang, T. J., Solanki, S. K., Curdt, W.,
927: %Innes, D. E., Dammasch I. E., \& Kliem, B. 2003, \aap, 406, 1105
928: \end{thebibliography}
929:
930:
931:
932:
933:
934:
935: \end{document}
936: