1: \documentclass[apj]{emulateapj}
2: \usepackage{float}
3: \usepackage{apjfonts}
4: \usepackage{amsmath}
5:
6: \usepackage{amsmath,amssymb}
7:
8:
9: \slugcomment{Received 2007 May 14; accepted 2007 May 29; published 2007 June
10: 27}
11: \shorttitle{MAGNETIC BUOYANCY AND COHERENT STRUCTURES}
12: \shortauthors{KERSAL\'E, HUGHES, \& TOBIAS}
13:
14: \begin{document}
15: \bibliographystyle{apj}
16:
17: % =========================================================================
18: \title{The nonlinear evolution of instabilities driven by magnetic buoyancy: A
19: new mechanism for the formation of coherent magnetic structures}
20:
21:
22: \author{Evy Kersal\'e, David W. Hughes, and Steven M. Tobias}
23: \affil{Department of Applied Mathematics, University of Leeds,
24: Leeds LS2 9JT, U.K.}
25: \email{kersale@maths.leeds.ac.uk}
26:
27:
28:
29:
30: % =========================================================================
31: \begin{abstract}
32: Motivated by the problem of the formation of active regions from a
33: deep-seated solar magnetic field, we consider the nonlinear
34: three-dimensional evolution of magnetic buoyancy instabilities resulting
35: from a smoothly stratified horizontal magnetic field. By exploring the case
36: for which the instability is continuously driven we have identified a new
37: mechanism for the formation of concentrations of magnetic flux.
38: \end{abstract}
39:
40: \keywords{Sun: interior --- Sun: magnetic fields --- instabilities --- MHD}
41:
42:
43: % =========================================================================
44: \section{Introduction}
45:
46: \label{sec:intr}
47:
48: Solar active regions are the surface manifestations of a deep-seated,
49: predominantly toroidal magnetic field. Although it is generally agreed that
50: the solar magnetic field is maintained by some sort of hydromagnetic dynamo,
51: the mechanism by which this is effected is far from understood. Consequently,
52: the strength, structure, and location of the interior magnetic field are
53: similarly unknown. However, most recent solar dynamo models, despite their
54: significant differences in other respects, postulate that the tachocline, the
55: thin layer of strong velocity shear located at the base of the convection
56: zone, plays an important role in the generation of toroidal field by the
57: shearing of a weaker poloidal component \citep[see, e.g., the review
58: by][]{smt2007}. Given this premise, it is important to address the nature of
59: the initial escape of the magnetic field from the tachocline, its subsequent
60: ascent through the convection zone, and its eventual emergence at the
61: photosphere.
62:
63: Owing to the vast range of scales across the convection zone, it is impossible
64: to model realistically all of these stages in one calculation. Here we
65: concentrate solely on the instability of a layer of magnetic field, with the
66: aim of clarifying the physics of the formation of coherent magnetic structures
67: from a much larger scale field.
68:
69: A vertically stratified, horizontal magnetic field can be unstable to magnetic
70: buoyancy instability provided that the field decreases sufficiently rapidly
71: with height. It is important to note the non-trivial distinction between
72: instability to two-dimensional modes in which the field lines remain straight
73: (interchange modes) and instability to fully three-dimensional modes. The
74: former are essentially destabilized by a decrease with height of $B / \rho$
75: (where $B$ is the magnetic field strength and $\rho$ the density), the latter
76: simply by a decrease with height of $B$. The physics underlying this
77: difference is elucidated in \citet{1987GApFD..39...65H}.
78:
79: The linear theory of magnetic buoyancy instabilities has been intensively
80: studied over a number of years \citep[e.g.][]{Newc61, 1966ApJ...145..811P,
81: 1970ApJ...162.1019G, 1978RSPTA.289..459A}. The nonlinear development of the
82: instability, particularly with respect to a deep-seated solar field, has,
83: inevitably, received less attention. \citet{1988JFM...196..323C} investigated
84: the nonlinear evolution of interchange modes resulting from a slab of uniform
85: horizontal magnetic field embedded in a convectively stable and otherwise
86: field-free atmosphere. Here the instability is driven by a density jump at the
87: upper interface of the magnetic slab --- an extreme form of magnetic buoyancy
88: instability. This interfacial instability generates a strong shear flow that
89: leads, by a secondary Kelvin-Helmholtz instability, to the formation of strong
90: vortices. The subsequent nonlinear evolution is then governed by pairwise
91: vortex interactions, which can even act so as to drag down pockets of strong
92: magnetic field. The three-dimensional instability and nonlinear evolution of
93: the same basic state was examined by \citet{1995ApJ...448..938M} and
94: \citet{2000MNRAS.318..501W}. The initial evolution again takes the form of
95: interchange modes, leading to the formation of vortex tubes. However,
96: neighbouring vortex tubes of opposite sign are unstable to a longitudinal
97: instability \citep{crow70}; this in turn causes the magnetic field to adopt an
98: arched structure. By contrast, \citet{2001ApJ...546..509F} investigated the
99: nonlinear evolution of a smoothly varying magnetic field, with a profile
100: chosen such that the initial state is unstable to three-dimensional modes, but
101: stable to interchange modes. She shows the formation of arched magnetic
102: structures, which maintain a reasonable degree of coherence as they rise. A
103: recent review of the implications of magnetic buoyancy instabilities for the
104: solar tachocline is given by \citet{dwh2007}. Related simulations of magnetic
105: buoyancy in sub-surface regions have also been performed
106: \citep[e.g.][]{2004A&A...426.1047A, 2005Natur.434..478I}.
107:
108:
109: Inspection of active regions on the solar surface suggests the emergence
110: through the photosphere of a buckled, toroidal magnetic field. However, very
111: little is known about the structure, strength and orientation of the
112: sub-surface field. For example, which properties of the observed surface field
113: result from the initial instability and which result from later interactions
114: in the convection zone? Our aim in this Letter is to seek further
115: understanding of the instability by examining the nonlinear evolution from a
116: very simple equilibrium state. We consider an atmosphere that is in both
117: magnetohydrostatic and thermal equilibrium and that has a linear magnetic
118: field profile. We concentrate on modes that are intrinsically
119: three-dimensional; i.e.\ the field gradients considered are such that the
120: basic state is unstable to three-dimensional modes but stable to interchanges.
121:
122: Furthermore, we impose boundary conditions such that the source of the
123: instability is maintained for all times, thus allowing us to consider the
124: long-term evolution. This is an important difference from the earlier works
125: cited above, all of which consider ``run-down'' experiments, in which the
126: potential energy stored in the initial field configuration is rapidly
127: converted into kinetic energy and which is then slowly dissipated.
128:
129: We consider four cases, characterised by two different field strengths and two
130: different choices of initial conditions. We have identified a completely new
131: mechanism for the formation of flux concentrations; this is of potential
132: significance in the solar context in which the formation of localised regions
133: of strong field from a weaker larger scale field is a crucial component of the
134: flux emergence process.
135:
136: \begin{figure}
137: \epsscale{0.9}
138: \plotone{f01a}
139: \plotone{f01b}
140: \figcaption{\label{fig:f01} Model with $\alpha=\frac{1}{4}$ and eigenfunction
141: perturbation. ({\sl a}) Evolution of the kinetic energy density. ({\sl b})
142: Horizontal average of the magnetic field $B_x$ vs. depth, initially
143: ({\sl dotted line}), at $t \approx 50$ ({\sl dashed line}), and in the
144: final steady state ({\sl solid line}).}
145: \end{figure}
146:
147: % =========================================================================
148: \section{Mathematical formulation and parameters}
149:
150: We consider a Cartesian layer of perfect gas governed by the equations of
151: three-dimensional compressible MHD. The dimensionless representation of these
152: equations is obtained by scaling lengths with the layer depth, time with the
153: isothermal sound crossing time at the top surface ($z=0$), and temperature,
154: density, and magnetic field with their respective values at the top surface.
155: The physical properties of the fluid are entirely determined by four
156: dimensionless numbers: the Prandtl number $P_r$, the ratio of viscous to
157: thermal diffusivity; $\zeta$, the ratio of magnetic to thermal diffusivity;
158: $\gamma$, the ratio of specific heats; and $C_k$, the nondimensional thermal
159: conductivity.
160:
161: The basic state that we consider is an exact equilibrium, dependent only on
162: depth $z$, with a unidirectional horizontal magnetic field $B_x(z) = 1 +
163: z/H_b$. The density and temperature are determined from solution of the
164: magnetohydrostatic equations. In the absence of a magnetic field, the
165: atmosphere takes the form of a plane polytrope, with polytropic index $m$ and
166: constant temperature gradient $\theta$. The strength of the magnetic field is
167: changed by varying $\alpha = 2 / \beta \propto B^2$, where the plasma-$\beta$
168: is the ratio of the thermal to the magnetic pressure. A more detailed
169: description of the governing equations and parameters is contained in
170: \citet{1998ApJ...502L.177T}.
171:
172: We consider perturbations to this basic state, subject to the following
173: boundary conditions: all variables are taken to be periodic in both horizontal
174: directions; the top and bottom boundaries are stress-free, impermeable, and
175: isothermal. For the magnetic field we choose to maintain a fixed gradient of
176: $B_x$ ($\partial B_x / \partial z = 1/H_b$) at $z=0, 1$, together with
177: $\partial B_y / \partial z = 0$ and $B_z = 0$. Our aim in this Letter is to
178: examine the nonlinear development of modes that are intrinsically
179: three-dimensional. By an extensive linear analysis we have identified the
180: regions in parameter space where three-dimensional modes ($k_x \ne 0$, $k_y
181: \ne 0$) are unstable but interchange modes ($k_x = 0$) are stable. In the next
182: section we discuss the nonlinear evolution resulting from two choices of
183: parameters and two different types of perturbation to the basic state.
184:
185: % =========================================================================
186: \section{Nonlinear evolution of three-dimensional modes}
187: % -------------------------------------------------------------------------
188:
189: The nonlinear evolution of three-dimensional magnetic buoyancy instabilities
190: is studied by solving numerically the equations of compressible MHD with a
191: modified version of the hybrid finite-difference-pseudospectral parallel code
192: used by \citet{1996ApJ...473..494B} (hydrodynamic) and
193: \citet{1998ApJ...502L.177T} (MHD). We consider the evolution from two basic
194: states, distinguished by the value of the field strength; the other parameters
195: are held fixed at $P_r = \zeta = 2\times 10^{-2}$, $C_k = 2.5 \times 10^{-2}$,
196: $\theta = 2$, $m=1.6$, $H_b=1$, and $\gamma = 5/3$. Furthermore, we consider
197: two different perturbations to each basic state: one consisting of the most
198: unstable eigenfunction, the other of random noise. In each case, the
199: horizontal dimensions of the Cartesian computational box are determined by the
200: wavenumbers $k_x$ and $k_y$ of the most unstable mode such that one wavelength
201: fits in the $x$-direction, along the initial magnetic field, and four
202: wavelengths fit transverse to the initial field, in the $y$-direction.
203:
204: %----------------------------------------------------------------------------
205: \subsection{Evolution to a Steady State}
206: \label{etass}
207:
208: Here we consider a basic state with $\alpha = 1/4$ perturbed by the most
209: unstable eigenfunction, namely, that with growth rate $\sigma = 0.16$ and with
210: wavenumbers $k_x = 0.87$ and $k_y = 8.38$ (hence, the dimensions of the
211: computational domain are $L_x = 7.24$ and $L_y=3$). The evolution is
212: essentially linear up to time $t \approx 21$. For $21 \lesssim t \lesssim 37$
213: the kinetic energy continues to grow (see Fig.~\ref{fig:f01}{\sl a}). During
214: this stage, broad upflows carry strong magnetic fields from the bottom of the
215: layer, while narrow plumes pull down weaker field, leading to the formation of
216: arched structures, as shown in Figure~\ref{fig:f02}. Figure~\ref{fig:f01}{\sl
217: b} shows the mean result of this nonlinear evolution, with a significant
218: reduction of the gradient of the horizontally averaged magnetic field $B_x$.
219: Initially constant throughout the layer ({\sl dotted line}), the vertical
220: gradient of $B_x$ becomes nearly zero in a large fraction of the domain at $t
221: \approx 50$ ({\sl dashed line}). The nonlinear reorganization has therefore
222: acted to remove the driving mechanism of the instability. There is an obvious
223: analogy with thermal convection, in which saturation is achieved by a
224: nonlinear reorganization leading to an adiabatic core with thin thermal
225: boundary layers.
226:
227: The arched configuration shown in Figure~\ref{fig:f02} is however transient;
228: following this initial nonlinear reorganization the system relaxes to a state
229: with weak flows and a magnetic field profile that is close to critical. As can
230: be seen from Figure~\ref{fig:f01}{\sl b}, the field gradient in the final
231: state is reduced everywhere in the layer compared with its initial value. This
232: subsequent evolution is permitted by the choice of boundary conditions, in
233: which the gradient of $B_x$ (but not the field itself) is fixed at the
234: boundaries.
235:
236: \begin{figure}
237: \epsscale{0.9}
238: \plotone{f02}
239: \figcaption{\label{fig:f02} Isosurface of magnetic energy $(B^2/2=1)$ at
240: $t \approx 45$ for the same model as in Fig.~\ref{fig:f01}.}
241: \end{figure}
242:
243:
244: % -------------------------------------------------------------------------
245: \subsection{Evolution to Time-dependent Concentrated Flux States}
246:
247:
248: First we consider the evolution from the same basic state as in \S\ref{etass}
249: (i.e.\ with $\alpha = 1/4$), but with the perturbation taking the form of
250: small-amplitude random noise. As shown in Figure~\ref{fig:f03}{\sl a}, the
251: linear evolution (up to $t \approx 36$) and the initial nonlinear phase, as
252: measured by the temporal evolution of gross properties such as kinetic energy,
253: are similar to those portrayed in Figure~\ref{fig:f01}{\sl a }. The first
254: nonlinear reorganization again leads to the formation of arched structures, as
255: shown in Figure~\ref{fig:f04}, although, as a result of the initial
256: conditions, these are less regular than those of Figure~\ref{fig:f02}.
257: Crucially, these irregularities play a significant role in modifying the
258: subsequent behaviour. From $100 \lesssim t \lesssim 450$ the kinetic energy
259: increases monotonically until the onset of a secondary oscillatory
260: instability. From $t\approx 1200$ the system evolves on two disparate
261: timescales: a short cycle with period $\approx 7$ modulated on larger
262: timescales $250 \lesssim t_\text{mod} \lesssim 300$.
263:
264: The underlying mechanism of the short period oscillations can be understood by
265: inspection of Figure~\ref{fig:f03}{\sl b}, which shows the temporal evolution
266: for fixed values of $x$ and $z$ of the magnetic energy, the transverse
267: horizontal velocity $v$, and the vertical velocity $w$. The left panel shows
268: the periodic formation of a concentration of magnetic energy, which drifts
269: slowly with velocity $0.03$. The central and right panels indicate that the
270: concentrations of magnetic energy result from convergent downflows associated
271: with rolls in the $yz$-plane. The magnetic flux becomes concentrated between
272: the two counter-rotating rolls; it thus becomes buoyant and rises rapidly.
273: This drives a countercell, which diverges at the flux concentration and
274: thereby destroys it. After some reorganization, the initial cellular flow is
275: reestablished but displaced relative to its initial position. Reformed
276: downflows again lead to a concentration of flux and the entire process is
277: repeated.
278:
279: Next we examine the evolution from an equilibrium with a stronger field
280: ($\alpha =1$). For this case the evolution is qualitatively similar regardless
281: of the nature of the initial perturbation; here we shall describe the case in
282: which the perturbation takes the form of random noise. The most unstable mode
283: has wavenumbers $k_x = 0.86$ and $k_y = 9.75$, and growth rate $\sigma =
284: 0.45$. Figure~\ref{fig:f05}{\sl a} shows that the temporal evolution is
285: broadly similar to that of the weaker field case portrayed in
286: Figure~\ref{fig:f03}{\sl a}. The secondary instability is evident, leading to
287: a modulated periodic state. The increase in initial field strength leads, as
288: expected, to a more vigorous instability (higher growth rate and increased
289: saturation level) and to a shorter period oscillation of the nonlinear state.
290: As in the example above, flux concentration occurs via converging downflows.
291: The peak field in the flux concentrations is significantly stronger than the
292: average initial field. In this case, as can be seen in Figure~\ref{fig:f06},
293: the flux concentrations are never completely dispersed, although their
294: strength varies as they propagate. This (modulated) travelling wave behaviour
295: for the field and flows can be clearly seen in Figure~\ref{fig:f05}{\sl b}.
296:
297: \begin{figure}
298: \epsscale{0.99}
299: \plotone{f03a}
300: \plotone{f03b}
301: \figcaption{\label{fig:f03} Model with $\alpha = \frac{1}{4}$ and random
302: perturbation. ({\sl a}) Evolution of the kinetic energy density. ({\sl b})
303: Space-time plots, with $y$ horizontal and $t$ vertical, of magnetic energy
304: $(1.17 \lesssim B^2/2 \lesssim 1.78)$, $|v| \lesssim 0.05$, and $|w|
305: \lesssim 0.06$; $ 1863 \lesssim t \lesssim 1913$; $z\approx 0.7$. }
306: \end{figure}
307:
308: % =========================================================================
309: \section{Discussion}
310: In this Letter we have identified a new, inherently nonlinear, mechanism for
311: the formation of coherent magnetic structures from a layer of weaker magnetic
312: field. The initial development of the magnetic buoyancy instability drives
313: flows that act so as to form isolated concentrations of magnetic flux. This
314: nonlinear state persists because the instability is continually driven from
315: the boundaries and saturation of the instability occurs via the net
316: redistribution of flux so as to remove average magnetic field gradients from
317: the interior of the computational domain. An intriguing feature of the flux
318: concentrations is that, once established, they travel --- taking the form of a
319: modulated wave.
320: \begin{figure}
321: \epsscale{0.9}
322: \plotone{f04}
323: \figcaption{\label{fig:f04} Isosurface of magnetic energy $(B^2/2=1)$ at
324: $t \approx 74$ for the same model as in Fig.~\ref{fig:f03}.}
325: \end{figure}
326:
327: The processes in the deep interior of the Sun that lead to the formation of
328: active regions are still poorly understood. Even the most optimistic estimates
329: from dynamo theory suggest that an upper bound for the large-scale
330: dynamo-generated field is of the order of $10^4$ G \citep[see,
331: e.g.][]{2003A&ARv..11..287O}. The energy of such a field is small, however, in
332: comparison with that of the strong downflows in the convection zone and, such
333: a field might be expected to be pinned down by the convection \citep[see,
334: e.g.][]{1998ApJ...502L.177T}. Clearly, therefore, some process must be acting
335: so as to create strong localised field structures from the dynamo field that
336: are capable of traversing the convection zone in order to form active regions.
337: Our calculations, although idealized, suggest a new mechanism that may play a
338: key role in this process.
339: \begin{figure}
340: \epsscale{1.}
341: \plotone{f05a}
342: \plotone{f05b}
343: \figcaption{\label{fig:f05} Same as Fig.~\ref{fig:f03}, but with $\alpha
344: =1$. In panel {\sl b} $0.65 \lesssim B^2/2 \lesssim 2.27$, $|v| \lesssim
345: 0.51$, $|w| \lesssim 0.56$; $153 \lesssim t \lesssim 200$; and $z \approx
346: 0.5$.}
347: \end{figure}
348: \begin{figure}
349: \epsscale{0.9}
350: \plotone{f06}
351: \figcaption{\label{fig:f06} Slices in the $yz$-plane of magnetic energy at
352: times $t \approx 152.9$, $154.7$, $156.5$. The peak magnetic energy
353: $\approx 3$.}
354: \end{figure}
355:
356: We conclude by noting that the discovery of the mechanism arose from
357: considering a system where the instability is continually driven, in contrast
358: to nearly all previous studies in which the initial magnetic buoyancy
359: instability was investigated by means of run-down calculations. Both types of
360: experiment may be of importance for understanding the dynamics of the solar
361: interior. The magnetic buoyancy instability is a fast process that might be
362: captured by run-down calculations (which usually lead to the formation of
363: arched structures). On the other hand, the replenishment of toroidal field by
364: the dynamo should lead to a continual source of instability of field at the
365: base of the solar convection zone.
366:
367: % =========================================================================
368: \acknowledgments The authors acknowledge financial support from PPARC, under
369: grant PPA/G/O/2002/00014, and the Leverhulme Trust. We are grateful to
370: N.H.~Brummell for helpful discussions and numerical advice.
371:
372:
373: % =========================================================================
374: % References
375: \bibliography{apj-jour,ms}
376:
377: \end{document}
378:
379:
380: % Figures Two ways of doing it
381: % \begin{figure}
382: % \centerline{\parbox{0.90\hsize}{(a)}}
383: % \centerline{\includegraphics[width=0.90\hsize]{f11a}}
384: % \centerline{\parbox{0.90\hsize}{(b)}}
385: % \centerline{\includegraphics[width=0.90\hsize]{f11b}}
386: % \centerline{\parbox{0.90\hsize}{(c)}}
387: % \centerline{\includegraphics[width=0.90\hsize]{f11c}}
388: % \caption{Same model as Figure~\ref{fig:f10}. Snapshot of the
389: % non-trivial steady state; (a) radial velocity; (b) azimuthal
390: % velocity; (c) vertical magnetic field.}
391: % \label{fig:f11}
392: % \end{figure}
393: