1:
2:
3: \documentclass{elsart}
4: \bibliographystyle{elsart-harv}
5: \usepackage{natbib}
6: \usepackage[T1]{fontenc}
7: \usepackage[latin1]{inputenc}
8: \usepackage{graphicx}
9:
10: \makeatletter
11:
12: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% LyX specific LaTeX commands.
13: %% Bold symbol macro for standard LaTeX users
14: \providecommand{\boldsymbol}[1]{\mbox{\boldmath $#1$}}
15:
16: \usepackage{amssymb}
17:
18: \usepackage[english]{babel}
19: \makeatother
20: \begin{document}
21:
22: \begin{frontmatter}
23:
24:
25: \title{Axial invariance of rapidly varying
26: diffusionless motions in the Earth's core interior}
27:
28:
29: \author{Dominique Jault}
30: \ead{Dominique.Jault@obs.ujf-grenoble.fr}
31: \address{LGIT, CNRS, Universit\'e Joseph-Fourier, BP 53, 38041 Grenoble
32: Cedex 9, France}
33:
34: \begin{abstract}
35: Geostrophic jets propagating as Alfv\'en waves
36: are shown to arise in
37: a rapidly rotating spherical shell permeated by a magnetic field
38: among the transient motions set up
39: by an impulsive rotation of the inner core. These axially invariant
40: motions evolve on a time-scale which is short compared to the magnetic
41: diffusion time.
42: The numerical study is taken as illustrative of a more general point:
43: on such a fast time-scale the dimensionless number
44: appropriate to compare the rotation and magnetic forces is independent
45: of the magnetic diffusivity in contrast with the often used Elsasser number.
46: Extension of the analysis to non-axisymmetrical motions is
47: supported by published
48: studies of dynamo models and magnetic instabilities.
49: \end{abstract}
50:
51: \begin{keyword}
52: Earth's core \sep Magneto-hydrodynamic waves \sep geomagnetic field \sep
53: core flow
54: \end{keyword}
55:
56: \end{frontmatter}
57:
58:
59: \section{Introduction}
60: For the last ten years,
61: numerical simulations of the dynamo process in the Earth's core
62: have much changed the views on the interplay
63: between magnetic and rotation forces.
64: In particular, columnar flows almost invariant in the direction parallel
65: to the rotation axis and localized outside the imaginary cylinder tangent
66: to the inner core
67: have been found very often
68: even though the Elsasser number $\Lambda$, classically used
69: to estimate the ratio of magnetic to rotation forces, is of order 1
70: or larger (see e.g. \cite{olson99b,grote00}).
71: Alignment parallel to the rotation axis is caused by
72: the predominance of rotation forces.
73: Accordingly, columnar flows had been contemplated
74: previously in the context of
75: weak-field models ($\Lambda \ll 1$) alone \citep{busse75}.
76: The ``strong field regime'' ($\Lambda = O(1)$) was illustrated by
77: mean-field dynamo solutions, in which
78: the azimutal angular velocity showed instead large shears in the
79: direction parallel to the rotation axis \citep{brag78,hollerbach93,jault95}.
80: These early solutions were either steady or slowly varying
81: on the magnetic diffusion time in sharp contrast
82: with the current generation of dynamo solutions.
83:
84: With large magnetic fields, as measured by $\Lambda$, only
85: the geostrophic part of the velocity field, symmetric
86: about the rotation axis, was expected to
87: be invariant in the direction parallel to the rotation axis.
88: \cite{brag70} singled out these motions in the context of magnetostrophic
89: equilibrium, characterized
90: by the insignificance of inertial and viscous forces compared to magnetic,
91: rotation and pressure forces.
92: He found that as these azimutal velocities
93: shear the magnetic field, they
94: are subject to a restoring force, provided by the magnetic field,
95: that ensures wave propagation.
96: This is the mechanism of
97: Alfv\'en waves and indeed geostrophic velocities in a rotating spherical
98: shell permeated by a magnetic field obey an equation of Alfv\'en
99: wave type save for geometrical factors.
100: \cite{brag70} assigned these torsional Alfv\'en waves to perturbations
101: with respect to a slowly evolving basic state characterized by the
102: cancellation of the total action of magnetic forces on the geostrophic
103: cylinders. That description sets the geostrophic velocities apart.
104: My aim, in this paper, is to defend another explanation for the
105: emergence of torsional Alfv\'en waves that can be generalized to
106: nonaxisymmetric motions, such as the almost
107: axially invariant vortices found in recent geodynamo solutions characterized
108: by strong but rapidly fluctuating magnetic fields. Other examples
109: are outlined in the discussion part.
110:
111: In the next section, I introduce
112: the two dimensionless numbers
113: $\Lambda$ and $\lambda$ that
114: measure
115: the relative strength of the magnetic and rotation forces, within
116: a rapidly rotating body permeated by a magnetic field.
117: I argue that on fast diffusionless time-scales, the appropriate
118: number is $\lambda$.
119: This is illustrated in the third section,
120: which constitutes the
121: main body of the article. The competition
122: between magnetic and rotation forces is studied in a rapidly
123: rotating spherical shell immersed in a magnetic
124: field. Specifically, the
125: axisymmetrical transient
126: motions set-up by an impulsive rotation of the inner core
127: are investigated for different values of $\lambda$
128: and the Elsasser number.
129: This is followed by a general discussion, where
130: different problems are listed for which $\lambda$ rather
131: than $\Lambda$ is appropriate to compare magnetic and rotation forces.
132: The paper ends with concluding remarks.
133:
134: \section{Lehnert versus Elsasser numbers}
135:
136: \cite{elsasser46} argued that the magnetic field in the Earth's core
137: saturates when the magnetic force becomes comparable to the
138: Coriolis force and suggested the characteristic strength
139:
140: \begin{equation}
141: B= \left( \frac{2\Omega\rho}{\sigma}\right)^{1/2} \; ,
142: \label{B_Elsasser}
143: \end{equation}
144: where $\Omega$ is the angular velocity, $\rho$ is the density and
145: $\sigma$ is the electrical conductivity.
146: The Elsasser number,
147: \begin{equation}
148: \Lambda= \frac{\sigma B^2}{\Omega\rho}\; ,
149: \end{equation}
150: has subsequently been used to measure the relative strength of Coriolis and
151: magnetic forces. In order to derive the relationship (\ref{B_Elsasser}),
152: the electrical current density $j$ is estimated as $\sigma U B$.
153: This is obviously not valid when magnetic diffusion
154: is negligible compared to induction ($j\ll \sigma U B$).
155:
156: Conversely, magnetic diffusion does not enter the physics of
157: plane magnetohydrodynamic waves, of length-scale $l$ that
158: \cite{lehnert54b} studied. He used
159: another dimensionless number $\chi_0=\lambda^{-1}$, with
160: \begin{equation}
161: \lambda= \frac{B}{\Omega(\mu\rho)^{1/2}l}\; ,
162: \label{lambda}
163: \end{equation}
164: to measure the relative strength of magnetic and rotation forces.
165: The parameter
166: $\lambda$, hereinafter referred to as the Lehnert number,
167: can be defined as the ratio, in a rapidly rotating
168: and electrically conducting fluid
169: permeated by a magnetic field, of the period of the
170: inertial waves to the period of the Alfv\'en waves. Thus, the
171: typical frequency of diffusionless Alfv\'en waves is $\lambda \Omega$ and
172: the relationship
173: \begin{equation}
174: \lambda \ll 1
175: \label{petit-lehnert}
176: \end{equation}
177: states that rotation forces dominate over magnetic forces on fast
178: time-scales.
179: \cite{cardin02} argued that both $\Lambda$ and $\lambda$ are
180: important to characterize geodynamo models.
181:
182: In the spherical case, it is convenient to specify $\lambda$ using the
183: outer radius $a$ as the length-scale $l$ in the definition (\ref{lambda}).
184: Denoting by $E_M$ the magnetic Ekman number $\eta / \Omega a^2$
185: and by $E$ the ordinary Ekman number
186: $\nu/\Omega a^2$,the relationship,
187: \begin{equation}
188: E_M +E \ll \lambda ,
189: \end{equation}
190: \label{atten}
191: \noindent
192: ensures that Alfv\'en waves are not rapidly damped by either magnetic
193: or viscous diffusion ($\nu$ and $\eta$ are respectively the viscous and
194: magnetic diffusivities).
195: Indeed, the ratio of
196: $\lambda$ to $E+E_M$ is the Lundquist number $S$ which indicates
197: how far Alfv\'en waves propagate before they are quenched by diffusion
198: \citep{roberts67}. The two numbers $\lambda$ and
199: $\Lambda$ are related through the magnetic Ekman number:
200: \begin{equation}
201: \lambda ^2= \Lambda E_M .
202: \end{equation}
203:
204: The value of $\lambda$ appropriate to the Earth's core is of the order of
205: $3. \times 10^{-5} - 2. \times 10^{-4}$.
206: Indeed, $\lambda=10^{-4}$ corresponds to a magnetic field strength in the core
207: interior of the order of 3. mT. Usually quoted values range from
208: 1. mT \citep{christensen06} to 4. mT \citep{starchenko02}. Using
209: $E_M = 4. \times 10^{-9}$ yields $S$ of the order of
210: $10^{4} - 5. \times 10^{4}$.
211:
212: \section{Axisymmetric motions spawned in a spherical cavity by a
213: sudden impulse of the spin of the inner core}
214:
215: The change of the relative strengths of the magnetic and rotation
216: forces according to frequency is well illustrated by the contrast
217: between transient and steady flows in a differentially rotating spherical
218: shell in the presence of a magnetic field. Static solutions have been published
219: in the case of dipolar magnetic field and small differential rotation,
220: for which the structure of
221: the flow has been described according to the Elsasser number
222: \citep{hollerbach94,dormy98}. \cite{kleeorin97} have theoretically
223: investigated
224: steady
225: linear solutions when the imposed magnetic field is potential
226: and has dipole parity.
227: They have identified several asymptotic regimes according to values
228: of the Elsasser number in the small Ekman number limit.
229: Let us now study transient structures.
230:
231:
232: \begin{figure}
233: \centerline{
234: \includegraphics[clip=true,width=0.5\textwidth]{figure1}}
235: \caption{Field lines of the imposed poloidal axisymmetric field.}
236: \end{figure}
237:
238: \subsection{Model and governing equations}
239:
240: Consider an electrically conducting homogeneous fluid occupying
241: a spherical shell that
242: is immersed in an imposed steady magnetic field.
243: The ratio of the inner shell radius $b$ to the outer radius $a$
244: is $b/a=0.35$ as in the Earth's core.
245: The solid inner core has the
246: same electrical conductivity as the fluid and the outer boundary is insulating.
247: The fluid is rotating with the constant angular velocity $\Omega$.
248: The imposed magnetic field is chosen as:
249:
250: \begin{eqnarray}
251: \mathbf{B} &= & B_0 \mathbf{\nabla} \times (A \mathbf{e_\phi}) \\
252: A &=& (j_1(\beta_{11} r)-0.3j_1(\beta_{12} r))P_1^1(\cos\theta)
253: - 0.2\,j_3(\beta_{31} r)P_3^1(\cos\theta)
254: \label{Bbasic}
255: \end{eqnarray}
256:
257: \noindent
258: where $(r,\theta,\phi)$ are spherical coordinates,
259: $P_l^1$ are Legendre functions,
260: $(j_l)_l$ is the set of
261: spherical
262: Bessel functions of the first kind and $\beta_{ln}$ is the {\it n}th root
263: of $j_{l-1}(\beta)=0$.
264: The basic field, which is shown in figure 1,
265: has been chosen with the aim of modelling torsional
266: oscillations in the Earth's fluid core. In this context, the important
267: quantity \citep{brag70} is
268: \begin{equation}
269: \{B_s^2\}(s)=\frac{1}{2\pi (z_T-z_B)}
270: \left(\oint \int _{z_B} ^{z_T} B_s^2 d z d\phi\right),
271: \end{equation}
272: \noindent
273: evaluated on geostrophic cylinders of radius $s$ and of top and bottom
274: $z$-coordinates
275: respectively $z_T$ and $z_B$. Obviously, the choice (\ref{Bbasic})
276: is arbitrary. In contrast with the Earth's case and with the basic state
277: used by \cite{brag80} to model torsional oscillations,
278: $\{B_s^2\}$ vanishes at $s=a$ because of the imposed
279: dipole symmetry with respect to the equatorial plane.
280: Axisymmetry makes $\{B_s^2\}=0 $ at $s=0$ in contrast with the geophysical
281: case again.
282: In view of the
283: present study, the main characteristics of the field $\mathbf B$
284: defined by (\ref{Bbasic}) are that it is neither
285: parallel to the rotation axis nor rapidly decreasing with radius
286: as are current-free dipole fields.
287: The results presented below do
288: not depend on the details of the geometry of $\mathbf B$.
289:
290: Study
291: the evolution of the velocity field $\mathbf u$ and of the magnetic field
292: deviation $\mathbf b$ after an impulsive increase of the angular rotation
293: $\omega _b$ of the inner core, postulating symmetry about the axis of
294: rotation and dipole symmetry about the equatorial plane:
295: \begin{eqnarray}
296: u_r(r,\pi-\theta,\phi)&=&u_r(r,\theta,\phi), \; \;
297: u_\theta(r,\pi-\theta,\phi)=- u_\theta(r,\theta,\phi), \nonumber \\
298: u_\phi(r,\pi-\theta,\phi)&=& u_\phi(r,\theta,\phi), \; \;
299: b_r(r,\pi-\theta,\phi)=-b_r(r,\theta,\phi), \\
300: b_\theta(r,\pi-\theta,\phi)&=& b_\theta(r,\theta,\phi), \; \;
301: b_\phi(r,\pi-\theta,\phi)= -b_\phi(r,\theta,\phi). \nonumber
302: \end{eqnarray}
303: \noindent
304: The amplitude of the initial impulse
305: $\Omega _b$ is assumed to be
306: small enough so that the subsequent evolution of the dynamics
307: is independent of $\Omega _b$ within a scaling factor.
308: Using $B_0$ as unit of magnetic field, $a$ as length-scale and
309: $a (\mu\rho)^{1/2}/B_0$ as unit of time,
310: the fields $\mathbf u$ and $\mathbf b$ are
311: governed by the following linearised equations in the fluid region:
312:
313: \begin{eqnarray}
314: \frac{\partial \mathbf{u}}{\partial t}
315: + 2 \lambda ^{-1} \mathbf{e_z} \times \mathbf{u} & = & - \nabla p
316: +
317: (\mathbf{\nabla} \times \mathbf{B}) \times \mathbf{b} +
318: (\mathbf{\nabla} \times \mathbf{b}) \times \mathbf{B} \nonumber \\
319: & & +
320: P_m \lambda\,\Lambda^{-1} \nabla ^2 \mathbf{u} \, , \;
321: \label{momentum}\\
322: \frac{\partial \mathbf{b}}{\partial t} & = &
323: \mathbf{\nabla} \times (\mathbf{u} \times \mathbf{B}) +
324: \lambda\,\Lambda^{-1} \nabla ^2 \mathbf{b} ,
325: \label{induction}
326: \end{eqnarray}
327: where $P_m = \nu/\eta$ is the magnetic Prandtl number.
328: The field $\mathbf b$ is defined also in the inner solid region where:
329:
330: \begin{equation}
331: \frac{\partial \mathbf{b}}{\partial t} =
332: \lambda\,\Lambda^{-1} \nabla ^2 \mathbf{b} .
333: \end{equation}
334: Note that the steady-state solutions depend only on the two parameters
335: $\Lambda$ and $E$.
336: The velocity boundary conditions
337: \begin{eqnarray}
338: \mathbf{u} &=& 0\, , \; \; r=a \\
339: \mathbf{u} &=& s \, \omega _b (t) = s \,\Omega _b \, \delta (t-t_0), \; \; r=b
340: \end{eqnarray}
341: are written using the Dirac $\delta$ function and are
342: appropriate to rigid boundaries.
343:
344: The set of equations (\ref{momentum}) and (\ref{induction}) is discretized
345: and time-stepped from an initial state of rest:
346: \begin{equation}
347: \mathbf{u} = \mathbf{b} = \mathbf{0}
348: \end{equation}
349: \noindent
350: The Dirac $\delta$ function is approached as:
351: \begin{equation}
352: \frac{1}{\sqrt{\pi\epsilon}}e^{-(t-t_0)^2/\epsilon} .
353: \label{dirac}
354: \end{equation}
355: \noindent
356: It is a result of the simulations below that the solutions
357: are independent of the parameter $\epsilon$
358: provided that its value is set small enough.
359:
360: A poloidal/toroidal decomposition
361: \begin{eqnarray}
362: \mathbf{u} &=& u_\phi \mathbf{e_\phi} +\mathbf{\nabla} \times
363: (u_p \mathbf{e_\phi}) \\
364: \mathbf{b} &=& b_\phi \mathbf{e_\phi} +\mathbf{\nabla} \times
365: (b_p \mathbf{e_\phi})
366: \end{eqnarray}
367: is employed. The variables are expanded in associated Legendre functions, i.e.
368: \begin{equation}
369: u_\phi(s,\theta)=\sum_{l=0}^{lmax}u_\phi^l(s) P_{2l+1}^1(\cos\theta)
370: \end{equation}
371: and then discretized in radius. The minimum truncation level
372: {\it lmax} is 120 whereas at least 450 unevenly spaced points are used
373: in the radial direction.
374:
375: \begin{figure}
376: \centerline{
377: (\emph{a}) \includegraphics[clip=true,width=0.5\textwidth]{figure2a}
378: (\emph{b}) \includegraphics[clip=true,width=0.5\textwidth]{figure2b}}
379: \centerline{
380: (\emph{c}) \includegraphics[clip=true,width=0.5\textwidth]{figure2c}
381: (\emph{d}) \includegraphics[clip=true,width=0.5\textwidth]{figure2d}}
382: \caption{Contours of the induced azimutal magnetic field for
383: $\lambda=1.72\times 10^{-4}$, $\Lambda=0.52$,
384: $E=E_M=5.7\times 10^{-8}$ drawn at $t=8.6\times 10^{-3}$(a),
385: $t=1.72 \times 10^{-2}$(b),
386: $t=3.44 \times 10^{-2}$(c) and $t=6.88 \times 10^{-2}$(d). The
387: contour intervals and the frame size
388: are identical in all frames. The inner sphere
389: boundary is indicated with a thick line.}
390: \end{figure}
391:
392: \begin{figure}
393: \centerline{
394: (\emph{a}) \includegraphics[clip=true,width=0.5\textwidth]{figure3a}
395: (\emph{b}) \includegraphics[clip=true,width=0.5\textwidth]{figure3b}}
396: \centerline{
397: (\emph{c}) \includegraphics[clip=true,width=0.5\textwidth]{figure3c}
398: (\emph{d}) \includegraphics[clip=true,width=0.5\textwidth]{figure3d}}
399: \caption{Contours of constant angular velocity for
400: $\lambda=1.72\times 10^{-4}$, $\Lambda=0.52$,
401: $E=E_M=5.7\times 10^{-8}$ drawn at $t=8.6\times 10^{-2}$(a),
402: $t=0.26$(b),
403: $t=0.52$(c) and $t=1.03$(d). The
404: contour intervals are respectively $\Delta \omega$, $\Delta \omega/2$,
405: $\Delta \omega/5$ and $\Delta \omega/10$ in the frames \emph{a}, \emph{b},
406: \emph{c} and \emph{d} in order to compensate for the attenuation of
407: the velocities.}
408: \end{figure}
409:
410: \begin{figure}
411: \centerline{
412: (\emph{a}) \includegraphics[clip=true,width=0.5\textwidth]{figure4a}
413: (\emph{b}) \includegraphics[clip=true,width=0.5\textwidth]{figure4b}}
414: \centerline{
415: (\emph{c}) \includegraphics[clip=true,width=0.5\textwidth]{figure4c}
416: (\emph{d}) \includegraphics[clip=true,width=0.5\textwidth]{figure4d}}
417: \caption{Same as figure 3 for
418: $\lambda=1.72\times 10^{-4}$, $\Lambda=6.5$,
419: $E=1.42\times 10^{-8}$, $E_M=4.5\times 10^{-9}$ and same progression of
420: contour intervals.}
421: \end{figure}
422:
423:
424: \subsection{Formation and propagation of geostrophic jets}
425: \label{results}
426:
427: Let us examine
428: a typical sequence of solutions for a small value of
429: $\lambda$. Following the initial impulse, an almost geostrophic shear
430: sets up, after a few revolutions,
431: at the cylindrical surface tangent to
432: the inner core, hereafter referred to as tangent cylinder.
433: Induction of an azimutal magnetic field localized at the tangent cylinder
434: (compare the snapshot (a) to the snapshot (b) in figure 2
435: occurs about the end of this period during which the velocity field becomes
436: axially invariant. It starts from a source at the equator of the inner core.
437: The last two panels of figure 2 illustrates the following
438: period during which meridional electrical currents
439: parallel to the tangent cylinder intensify and loop further
440: and further away from the tangent cylinder.
441: This is the most noticeable feature before
442: the geostrophic shear splits up.
443: The outer shear readily transforms into a jet propagating away from the tangent
444: cylinder towards larger cylindrical radii.
445: Thereafter, the velocity
446: within the jet becomes more and more invariant along $z$ as time elapses
447: (figure 3).
448: On the other side of the tangent cylinder, a second shear propagates towards
449: the axis of rotation. Its propagation velocity slows down as $B_s$ decreases
450: to $0$ on the axis. In the event, the inner shear transforms also into a jet.
451: The comparison (figure 4) with a second sequence of
452: solutions for the same value of $\lambda$ but for $\Lambda$ multiplied by
453: a factor of 12.5 indicates that the dynamics outside the tangent cylinder is
454: almost independent of $\Lambda$. The flow remains geostrophic even though
455: $\Lambda$ is $O(1)$.
456: Note that steady flows do not reproduce this feature. Figure 5
457: shows zonal flows driven by rotating the inner sphere at a constant rate
458: for the two values of $\Lambda$ used to calculate the transient
459: solutions. For the largest value of $\Lambda$,
460: the angular velocity contours are not parallel to the rotation axis and
461: tend instead to follow the magnetic field lines, as prescribed by
462: Ferraro's law of isorotation. Steady solutions are established after
463: a period lasting a few time units $\lambda^{-1}\Omega^{-1}$ during which the
464: flow is geostrophic.
465:
466: \begin{figure}
467: \centerline{
468: (\emph{a}) \includegraphics[clip=true,width=0.5\textwidth]{figure5a}
469: (\emph{b}) \includegraphics[clip=true,width=0.5\textwidth]{figure5b}}
470: \caption{Steady azimuthal flow induced by rotating the inner and outer
471: boundaries at slightly different rates. Contours of constant angular
472: velocity for $E=4.5\times 10^{-7}$ and $\Lambda=0.52$ (a) $\Lambda=6.5$ (b).}
473: \end{figure}
474:
475: \begin{figure}
476: \centerline{\hspace*{-0.05\textwidth}
477: \includegraphics[clip=true,width=0.8\textwidth]{figure6}}
478: \caption{Scaling of the width $\delta$ of the outer geostrophic jet with
479: respect to the number $\lambda$. The magnetic Prandtl number is
480: $P_m=1$ and the Ekman number is $E=5.7\times 10^{-8}$ (crosses) or
481: $E=2.85\times 10^{-8}$ (+ signs). A line of slope $-1/4$ is shown
482: for comparison.}
483: \end{figure}
484:
485: \begin{figure}
486: \centerline{\hspace*{-0.05\textwidth}
487: \includegraphics[clip=true,width=0.8\textwidth]{figure7}}
488: \caption{Scaling of the width $\delta$ of the outer geostrophic jet with
489: respect to the Lundquist number $S^\dag=\lambda/E_M$.
490: The magnetic Prandtl number is
491: $P_m=1$ and $\lambda$ is $4.3 \times 10^{-5}$ ($\Diamond$),
492: $8.6 \times 10^{-5}$ ($\times$), $1.72 \times 10^{-4}$ ($\circ$),
493: $3.44 \times 10^{-4}$ ($+$), $6.88 \times 10^{-4}$ ($\Box$).
494: A line of slope $-1/4$ is shown
495: for comparison.}
496: \end{figure}
497:
498: The outer geostrophic jet has finite
499: width $\delta$ in the limit $\epsilon \rightarrow 0$ (see expression
500: (\ref{dirac}) for the definition of $\epsilon$).
501: Investigating the variation of $\delta$ with $\lambda$, $E$, and $E_M$
502: gives an useful insight into the mechanism of generation of the geostrophic
503: motions. Here $\delta$ is arbitrarily defined as the distance along
504: $s$ between the two cylinders, on both sides of the geostrophic jet,
505: where the angular velocity has half its
506: maximum value. For the range
507: of parameters that has been extensively explored, the outer layer is always
508: well characterized from $t=0.17$ onwards. Results are reported for this time.
509: Keeping $E$ constant and $P_m=1$, it is found that $\delta$ varies as
510: $\lambda ^{-1/4}$ (see figure 6).
511: For fixed values of $\lambda$ and $P_m=1$, $\delta$
512: varies as $E_M^{1/4}$.
513: This is illustrated by the figure 7. Indeed, assuming
514: $\delta \sim \lambda ^{-1/4}$,
515: the power law $\delta \sim E_M^{1/4}$ can simply be written:
516: \begin{equation}
517: \delta \sim \left(S^\dag\right)^{-1/4}
518: \label{asympt}
519: \end{equation}
520: using the Lundquist number $S^\dag =\lambda (E_M)^{-1}$.
521: Thus, the width $\delta$ is independent of
522: the angular velocity $\Omega$.
523: We are interested by results for $P_m < 1$ since $\nu\ll\eta$ in
524: the geophysical case and in laboratory experiments as well.
525: Decreasing $P_m$ from $P_m=1$, a slight dependence of
526: $\delta$ on $P_m$ is found (figure 8).
527: Extension of the relationship (\ref{asympt}) to solutions
528: for $P_m < 1$ is supported by these results.
529:
530:
531: \begin{figure}
532: \centerline{\hspace*{-0.05\textwidth}
533: \includegraphics[clip=true,width=0.8\textwidth]{figure8}}
534: \caption{Thickness $\delta$ of the outer geostrophic jet with
535: respect to $P_m^{-1}$ for $P_m\le 1.$ and various $\lambda$:
536: $\lambda=4.3 \times 10^{-5}$ (circles), $\lambda=8.6 \times 10^{-5}$ (crosses),
537: $\lambda=1.72 \times 10^{-4}$ (+ signs), $\lambda=3.44 \times 10^{-4}$
538: (squares).}
539: \end{figure}
540:
541:
542: The outer geostrophic jet is radiated from the tangent cylinder, which
543: touches the inner core on its equatorial circle. There, both the
544: rotation vector and the magnetic field are parallel to the inner core
545: surface and the
546: Ekman-Hartmann viscous boundary layer adjacent to the inner core is singular.
547: Thus, it is instructive to investigate the influence of the strength of
548: the magnetic field at the singularity. For this purpose, an axial uniform field
549: can be added to the magnetic field defined by (\ref{Bbasic}). It is found
550: that the jet is much thickened if the two
551: fields cancel out at the singularity. Then, the relationship (\ref{asympt})
552: does not hold.
553: On the other hand, it is also found - in the narrow parameter
554: range which has been investigated - that the
555: layer shrinks as $\left[B_z(b,0)\right]^{-1/4}$ as the
556: axial magnetic field adjacent to the
557: equatorial ring of the inner core is increased.
558: Taking this result at its face value,
559: the magnetic field strength entering the relationship (\ref{asympt}) would be
560: $B_z(b,0)$.
561: Putting these results together, it appears that the magnetic structure
562: adjacent to the equator of the inner core plays a significant role
563: in the emergence of the two propagating shear layers.
564: The transformation of the outer shear layer into an independent jet
565: detached from the inner core is promoted by the axial magnetic field.
566:
567: Once the outer jet is formed, its time evolution is given by the
568: equations of \cite{brag70}, which satisfy
569: angular momentum conservation.
570: It is possible to estimate the angular momentum
571: $ A(t)=s^2\sqrt{1-s^2}\delta u_\phi(s)$ carried
572: by the outer jet from the width $\delta$. For the solutions that have
573: been investigated, {\it A} does not change throughout the
574: propagation of the jet.
575: The strength $B_{s}$ of the magnetic field sheared by the jet decreases with
576: $s$
577: to 0 at $s=1$. Thus, the jet slows down as it approaches the outer sphere
578: equator, which it never reachs.
579:
580: The scaling (\ref{asympt}) has implications
581: for the coupling between
582: the axial rotation of the Earth's solid inner core and torsional oscillations
583: \citep{buffett05}.
584: The time unit $\lambda ^{-1} \Omega^{-1}$ corresponds to 1-10 years for
585: geophysical applications.
586: A characteristic time $\left(S^\dag\right)^{-1/4}(\lambda\Omega)^{-1}$
587: can be derived
588: from the length $\delta$ and
589: the Alfv\'en wave velocity $\lambda \Omega a$. It corresponds
590: to the inner core rotation period below which dissipative processes at
591: the tangent cylinder are important. Using a geophysical estimate for
592: $S^\dag$, the coupling mechanism presented above between
593: the rotation of the inner core and torsional Alfv\'en waves is
594: found to be efficient
595: on periods longer than a few months.
596: Investigation of magnetic fields with non-dipole symmetry will be a natural
597: follow-up of this study.
598:
599: Finally, keeping $S^\dag$ constant and increasing $\lambda$, it is found
600: that the structures radiated from the tangent cylinder lose their
601: geostrophic character for $\lambda \sim 10^{-2}$.
602:
603:
604: \section{Discussion}
605:
606: Let us now examine to what extent the parameter $\lambda$
607: is also appropriate to comment on the geometry of fast motions in other
608: problems characterized by rapid rotation and magnetic field.
609:
610:
611: \subsection{Axially invariant hydromagnetic instabilities occurring
612: at small Lehnert number}
613:
614: We can discriminate
615: between two approaches that have been followed to study the stability of a
616: magnetic field in a rotating sphere according to the parameter,
617: either $\lambda$ or $\Lambda$, used
618: to compare magnetic and rotation forces. \cite{malkus67} recently followed
619: by \cite{zhang03} studied hydromagnetic waves in a non-dissipative
620: fluid ($\Lambda \rightarrow \infty$). As a result, they wrote the condition
621: for stability as a relationship involving the Lehnert number $\lambda$, which
622: has to exceed values of the order unity. \cite{zhang94} focused
623: on the rapid
624: rotation limit ($\lambda \rightarrow 0$) instead.
625: Then, the onset of instability occurs
626: for a critical value of the Elsasser number $\Lambda _c$.
627: In accordance with the above
628: discussion on the geometry of the motions in the limit
629: ($\lambda \rightarrow 0$), they found that the (non-axisymmetric) instability
630: is characterized by nearly two-dimensional columnar fluid motions despite
631: $\Lambda _c$ being $O(10)$.
632:
633: This result stands when the instabilities are thermally driven.
634: \cite{zhang95} focused his study of rotating convection in the presence of an
635: axisymmetrical toroidal magnetic field on the limit ($E \rightarrow 0$),
636: which amounts to ($\lambda \rightarrow 0$) for finite values of the
637: Elsasser number $\Lambda$. The magnetic Prandtl number is set to $1$ and
638: the control parameters are thus $\Lambda$ together with a Rayleigh number.
639: The fluid motions, in the limit ($E \rightarrow 0$),
640: are almost two-dimensional
641: showing only slight variations along
642: the direction of the rotation axis whilst
643: $\Lambda$ is $O(1-10)$. The solutions of \cite{olson95} (outside the
644: cylindrical surface tangent to the inner core), \cite{walker97}
645: (their figures 6f, 7f, 8f, 9f) for different basic states
646: and \cite{zhang96} (their figure 4) for $\kappa/\eta \ll 1$, where
647: $\kappa$ is thermal
648: diffusivity,
649: all present similar features.
650:
651: \subsection{Columnar flow structure in geodynamo models}
652:
653: Obviously, the numbers $\lambda$ and $\Lambda$ are less directly
654: relevant to studies of dynamo simulations than to investigations of
655: models with imposed large-scale magnetic field. These two estimates
656: of the magnetic field strength come out as
657: output of the numerical runs instead of being among the initial parameters.
658: It is nevertheless true that
659: realistic values of $\lambda$ are reached in numerical
660: models of the geodynamo as $\lambda$ does not involve diffusivities.
661: Thus,
662: \cite{christensen06} conducted a statistical analysis of a set of geodynamo
663: models
664: and estimated a parameter defined as $\lambda$, using
665: the shell depth as the length-scale $l$ in (\ref{lambda}).
666: Their results correspond to $\lambda$ varying
667: from $7\times 10^{-3}$ to $3 \times 10^{-2}$,
668: keeping the core radius as length-scale.
669: \cite{christensen06} found that the narrow range of $\lambda$
670: values contrasts
671: with the wide variations of $\Lambda$.
672: They also remarked that measuring the relative strength
673: of magnetic and rotation
674: forces acting in geodynamo models
675: with the Elsasser number $\Lambda$ does not reflect the fact that
676: the Lorentz force
677: depends on the length scale of the magnetic field whereas the Coriolis
678: force is independent of the length scale of the velocity field.
679: Conversely, comparing typical periods of the Alfven waves to typical
680: periods of the inertial waves shows that the relative importance of the
681: magnetic force augments with decreasing length scales (see the expression
682: (\ref{lambda}) of $\lambda$ ).
683: In their previous systematical parameter study, \cite{christensen99} had found
684: only one case with strong deviations from columnarity ($\Lambda=14$,
685: $E_M=6 \times 10^{-5}$, $\lambda=3 \times 10^{-2}$).
686: Together, these results are consistent with the statement that the extent
687: to which rotation affects the structure of the flow depends
688: on $\lambda$, the motions remaining columnar up to
689: $\lambda=O(3 \times 10^{-2})$.
690:
691: \subsection{Torsional oscillations in convective dynamo models}
692:
693: Convection columns can excite time-dependent
694: geostrophic motions in dynamo models through magnetic and
695: Reynolds stresses. This has been illustrated by \cite{dumberry03}.
696: They separated the axisymmetric azimutal velocity field
697: obtained from the geodynamo model of \cite{kuang99}
698: - $E=E_M=4 \times 10^{-5}$, $P_r=1$ and stress-free boundary conditions -
699: into a mean flow plus a fluctuating component.
700: They showed that the quasi-static azimutal winds have large gradients
701: in the $z$-direction. On the other hand,
702: \cite{dumberry03} emphasized
703: the axial invariance of the time-varying zonal flows.
704: Their finding that, outside
705: the tangent cylinder, the whole
706: length of the geostrophic cylinders accelerates azimutally as if they
707: were rigid on time-scales $\tau \sim 0.1 \, \tau _D$ is in line
708: with the small value of $\lambda$ in this numerical experiment
709: ($\tau _D$ magnetic diffusion time).
710: Indeed, using $B \sim 2 (2 \Omega \mu \rho
711: \eta)^{1/2}$ (see fig. 10 of \cite{kuang99}),
712: we infer $\Lambda \sim 10$ and $\lambda \sim 2 \times 10^{-2}$.
713: These geostrophic motions
714: are not Alfv\'en waves as Reynolds stresses and viscous forces
715: are as important as the magnetic forces in the balance of forces
716: acting on the geostrophic cylinders.
717: More recently and
718: with Earth-like no-slip boundary conditions, \cite{taka05}
719: argued indeed that the Ekman number has to be decreased
720: down to $E=8 \times 10^{-6}$ to make the viscous torque acting
721: on the geostrophic
722: cylinders negligible and the magnetic torque predominant.
723: Finally, for the same value of
724: $E$ as \cite{taka05},
725: but with stress-free boundary conditions and small Prandtl number
726: $P_r=0.1$,
727: \cite{busse05} found a
728: dynamical state where the magnetic torques on geostrophic cylinders account
729: for most of the geostrophic acceleration. Extracting the average magnetic field
730: strength from the figure 18 of \cite{busse05} gives
731: $\lambda \sim 5 \times 10^{-3}$ - well in the domain $\lambda \ll 1$ -
732: and $\Lambda \sim 3$. The result that
733: magnetic forces dominate over Reynolds stresses in the balance of force
734: acting on the geostrophic cylinders can be related to the observation
735: that the magnetic
736: energy is much stronger than the kinetic energy in this solution.
737: Thus, sequences where geostrophic motions behave as torsional oscillations
738: begin to be detected in convective dynamo models. That requires $\lambda \ll 1$
739: - to obtain time-dependent geostrophic motions, observed for
740: $\lambda \sim 2 \times 10^{-2}$ by \cite{dumberry03} -, small $E$ and,
741: presumably,
742: kinetic energy weaker than magnetic energy.
743: In these studies, there is no evidence that the quantity $\{B_s^2\}$
744: measuring the intensity of the magnetic field sheared by the
745: geostrophic motions
746: and the geostrophic velocities
747: evolve on separate time-scales.
748: Further work is needed to decide what
749: kind of models (fully consistent dynamo models with poor separation of scales
750: versus models incorporating a steady field) better describes the Earth's core
751: dynamics.
752:
753: \subsection{Torsional oscillations and Taylor states}
754:
755: In this article, torsional oscillations are considered as part of the
756: rapid motions that
757: are dominated by rotation because magnetic diffusion is negligible.
758: \cite{brag70} had a different line of arguments.
759: He attributed torsional oscillations to departures from
760: a dynamic equilibrium where the net torque of the
761: Lorentz force on any geostrophic cylinder is zero \citep{taylor63}.
762: This condition has to be met, in spherical geometry, when only the Coriolis,
763: pressure,
764: buoyancy and magnetic forces are taken into account (MAC balance).
765: It is frequently referred
766: to as a ``Taylor state''. It describes a dynamo regime on the long
767: time-scale for which magnetic diffusion is important.
768: Reinstating the acceleration of geostrophic motions
769: $\partial u_\phi (s)/\partial t$ in the equation
770: for azimutal velocities, it has been possible
771: to exhibit inviscid solutions of the
772: model-Z of \cite{brag78} that are in a Taylor state \citep{jault95}.
773: However, this is almost
774: the unique instance where a connection between torsional oscillations
775: and idealized Taylor states has been vindicated.
776: Geodynamo numerical models showing
777: torsional oscillations that keep bringing back the magnetic field
778: towards a Taylor state have not yet been found.
779: The two viewpoints differ insofar nonzonal rapid motions are
780: considered. I envision here that they are also constrained by rotation
781: being almost $z$-invariant whereas
782: \cite{brag70} made no predictions on the geometry of these motions.
783:
784: \section{Concluding remarks}
785:
786: Focusing a numerical study on the transient motions spawned by an
787: impulse in the
788: rotation of the inner boundary of a rapidly rotating spherical
789: shell immersed
790: in a magnetic field with dipole symmetry, I have documented
791: the emergence of
792: geostrophic jets from the cylindrical surface that touches the
793: inner core at its equator, irrespectively of the value of the Elsasser number.
794: Both the poloidal motions, of which the vorticity is aligned
795: along $\mathbf{e_\phi}$, and the toroidal motions with shear
796: in the $z$ direction are rapidly eliminated.
797: The geostrophic layers travel with the velocity $\lambda a \Omega$ and
798: are governed, outside the tangent cylinder,
799: by the equations written by \cite{brag70}.
800: The jet width scales as $\left(S^\dag\right)^{-1/4}$.
801: This estimate yields the frequency
802: below which oscillations of the solid core are communicated to
803: torsional Alfv\'en waves in the fluid shell. Using Earth-like parameters,
804: it corresponds to a period of a few months.
805: This study gives an illustration of the key role
806: played - for fast flows -
807: by the parameter $\lambda$ - independent of magnetic diffusivity -
808: put forward by \cite{lehnert54b}.
809: Conversely, $\lambda$ is not appropriate to study steady
810: solutions as it cannot be derived from the two parameters $\Lambda$ and
811: $E$ that characterize the static problem.
812:
813: In the same spirit, I have been able to base a discussion of
814: earlier numerical studies of magnetic instabilities
815: and dynamos in rotating shells on that parameter
816: $\lambda$. I suggest that
817: the smallness of $\lambda$ in some
818: of these studies is the reason for the occurrence of
819: columnar motions aligned parallel to the axis of rotation
820: and also of geostrophic flows evolving as Alfv\'en waves.
821: From these earlier studies, I anticipate that the results presented here
822: can be extended to the nonaxisymmetric case.
823:
824: Thus, the parameter $\lambda$, instead of the Elsasser number $\Lambda$ ,
825: is the appropriate parameter to compare magnetic and rotation forces when
826: flows evolving on time-scales much shorter than the magnetic diffusion
827: time are considered.
828: As the value of $\lambda$ appropriate to the Earth's fluid core is
829: $O(10^{-4})$,
830: I advocate that motions
831: in the core interior
832: with fast diffusionless time-scales are
833: approximately $z$-independent and
834: columnar with vorticity
835: aligned parallel to the rotation axis. That paves the way
836: for dynamical studies of the flows responsible for the secular variation
837: of the Earth's magnetic field, generalizing to all fast motions what has
838: already been achieved for the geostrophic, axially symmetric ones
839: \citep{zatman97b}.
840:
841: \section*{Acknowledgements}
842: I thank D. Brito, A. Fournier, H.-C. Nataf and A. Pais for their careful
843: reading of the initial manuscript. A. Soward and an anonymous referee are
844: acknowledged for their detailed reports.
845: This work has been supported by a grant from the French Agence Nationale de la
846: Recherche, Research programme VS-QG (grant number BLAN06-2.155316).
847:
848: \bibliography{pepi07}
849:
850:
851: \end{document}
852:
853: