1: \documentclass[aps,prl,preprint,superscriptaddress,showpacs]{revtex4}
2: %\documentclass[aps,prl,twocolumn,showpacs]{revtex4}
3:
4: \usepackage{graphicx}
5:
6: \begin{document}
7:
8: \title{A Non-axisymmetric Magnetorotational Instability
9: of a Purely Toroidal Magnetic Field}
10:
11: \author{Rainer Hollerbach}
12: \address{Department of Applied Mathematics, University of Leeds,
13: Leeds, LS2 9JT, United Kingdom}
14: \author{G\"unther R\"udiger, Manfred Schultz, D. Elstner}
15: \address{Astrophysikalisches Institut Potsdam, An der Sternwarte 16,
16: D-14482 Potsdam, Germany}
17:
18: \date{\today}
19:
20: \begin{abstract}
21: We consider the flow of an electrically conducting fluid between
22: differentially rotating cylinders, in the presence of an externally
23: imposed toroidal field $B_0(r_i/r)\,{\bf\hat e}_\phi$. It is known
24: that the classical, axisymmetric magnetorotational instability does
25: not exist for such a purely toroidal imposed field. We show here
26: that a non-axisymmetric magnetorotational instability does exist,
27: having properties very similar to the axisymmetric magnetorotational
28: instability in the presence of an axial field.
29: \end{abstract}
30:
31: \pacs{47.20.-k, 47.65.+a, 95.30.Qd}
32:
33: \maketitle
34:
35: The magnetorotational instability (MRI) is a mechanism whereby a
36: hydrodynamically stable differential rotation flow may be destabilized by
37: the addition of a magnetic field. An obvious question then is whether
38: different orientations of the field yield different types of instability,
39: or the same, or none at all. The original view was that the axial
40: component of the field is the only important one, with any azimuthal
41: component playing no essential role, and incapable of producing any
42: instabilities on its own \cite{V59, BH91}. This view was altered
43: by the discovery that a mixed axial and azimuthal field yields
44: instabilities quite different in many ways from those found with a purely
45: axial field \cite{HR05, RH05}. In this letter we show that even a purely
46: azimuthal field yields an MRI, and compare its properties with the
47: previously known types.
48:
49: While its most important application is to astrophysical accretion disks
50: \cite{BH91}, the MRI was originally discovered in the much simpler
51: Taylor-Couette problem, consisting of the flow between differentially
52: rotating cylinders \cite{V59}. Because of its relative simplicity, this
53: geometry has proven particularly amenable both to theoretical analyses of
54: the MRI \cite{Ji,RZ,RSS}, as well as to its recent experimental realization
55: \cite{S06, R06}. We too shall examine the MRI in this context.
56:
57: Consider therefore an
58: electrically conducting fluid confined between two concentric cylinders of
59: radii $r_i$ and $r_o$, rotating at rates $\Omega_i$ and $\Omega_o$, chosen
60: to satisfy $\Omega_o r_o^2>\Omega_i r_i^2$. That is, the angular momentum
61: increases outward, so by the familiar Rayleigh criterion the flow is
62: hydrodynamically stable, with the angular velocity given by
63: $$\Omega(r)=A + B/r^2,\eqno(1)$$
64: where
65: $$A=\frac{\Omega_or_o^2-\Omega_ir_i^2}{r_o^2-r_i^2},\quad
66: B=\frac{r_i^2r_o^2(\Omega_i-\Omega_o)}{r_o^2-r_i^2}.\eqno(2)$$
67: In this work we will fix $r_o=2r_i$ and $\Omega_o=\Omega_i/2$,
68: so $A$ and $B$ simplify to $\frac{1}{3}\Omega_i$ and $\frac{2}{3}\Omega_i
69: r_i^2$, respectively. The essence of the MRI then is to ask whether the
70: addition of a magnetic field can destabilize this flow.
71:
72: If the imposed field is purely axial, ${\bf B}_0=B_0\,{\bf\hat e}_z$, the
73: profile (1) can be destabilized, provided the rotation rates are
74: sufficiently great, and the field strength $B_0$ is neither too weak nor
75: too strong \cite{Ji,RZ}. Specifically, the magnetic Reynolds number ${\rm Rm}
76: =\Omega_ir_i^2/\eta$, where $\eta$ is the magnetic diffusivity, must exceed
77: $O(10)$, and the Lundquist number ${\rm S}=B_0r_i/\eta\sqrt{\rho\mu}$, where
78: $\rho$ is the density and $\mu$ the permeability, must be around $3-10$.
79:
80: In contrast, if the imposed field is mixed axial and azimuthal, ${\bf B}_0
81: =B_0\,{\bf\hat e}_z + \beta B_0(r_i/r)\,{\bf\hat e}_\phi$, where $\beta$
82: is around $1-10$, then the profile (1) can again be destabilized, but at
83: very different rotation rates and field strengths \cite{HR05,RH05}. The
84: relevant parameter measuring the rotation rates is now not the magnetic
85: Reynolds number, but rather the hydrodynamic Reynolds number ${\rm Re}=
86: \Omega_ir_i^2/\nu$, where $\nu$ is the viscosity. Similarly, the parameter
87: measuring the field strength is not the Lundquist number, but instead the
88: Hartmann number ${\rm Ha}=B_or_i/\sqrt{\rho\mu\eta\nu}$. The MRI sets
89: in when ${\rm Re}\gtrsim O(10^3)$, and ${\rm Ha}\approx O(10)$.
90:
91: To compare these results, we note that the two sets of parameters are
92: related by ${\rm Re=Rm\,Pm}^{-1}$ and ${\rm Ha=S\,Pm}^{-1/2}$, where
93: ${\rm Pm}=\nu/\eta$ is the magnetic Prandtl number, a material property of the
94: fluid. Typical values for liquid metals are $O(10^{-6})$. Translating the
95: results for the purely axial field, we thus obtain ${\rm Re}\gtrsim O(10^7)$
96: and ${\rm Ha}\approx O(10^4)$, both several orders of magnitude greater than
97: for the mixed field. It is perhaps not surprising then that the MRI has
98: been obtained experimentally for the mixed field \cite{S06,R06}, but not
99: (yet) for the purely axial field \cite{Ji}.
100:
101: As different as they are, one feature these two types of MRI have in
102: common is that they are both axisymmetric. Non-axisymmetric modes have
103: also been explored, for both the purely axial \cite{RSS} as well as the
104: mixed fields \cite{RH05}. For a purely axial field the relevant
105: parameters are still $\rm Rm$ and $\rm S$, but the critical values $\rm Rm_c$
106: are somewhat larger than for the axisymmetric modes, indicating that the
107: axisymmetric MRI is the most unstable mode. For mixed fields, one finds
108: --- perhaps somewhat surprisingly --- that adding an azimuthal component
109: now has minimal effect, certainly far less than the reduction by four
110: orders of magnitude found for the axisymmetric modes. Evidently the
111: relevant parameters continue to be $\rm Rm$ and $\rm S$, rather than
112: $\rm Re$ and $\rm Ha$.
113:
114: What we wish to show in this letter then is that for these
115: non-axisymmetric modes, one can impose a purely azimuthal field,
116: $B_0(r_i/r)\,{\bf\hat e}_\phi$, and still obtain an MRI, having all the
117: characteristics of the previous non-axisymmetric types of MRI \cite{RSS,
118: RH05}. To this end, we solve the linear stability equations
119: $${\rm Rm}\,\frac{\partial{\bf b}}{\partial t}=\nabla^2{\bf b}
120: +\nabla\times({\bf u\times B}_0)
121: + {\rm Rm}\,\nabla\times({\bf U}_0\times{\bf b}),\eqno(3)$$
122: $${\rm Re}\,\frac{\partial{\bf u}}{\partial t}=-\nabla p + \nabla^2{\bf u}
123: + {\rm Ha}^2\,(\nabla\times{\bf b})\times{\bf B}_0
124: +{\rm Re}\,({\bf U}_0\times\nabla\times{\bf u}
125: +{\bf u\times\nabla\times U}_0),\eqno(4)$$
126: where ${\bf U}_0=r\Omega(r)\,{\bf\hat e}_\phi$ is the profile (1) whose
127: stability we are exploring, and ${\bf B}_0=B_0(r_i/r)\,{\bf\hat e}_\phi$
128: is the imposed azimuthal field. Length has been scaled by $r_i$, time by
129: $\Omega_i^{-1}$, ${\bf U}_0$ and $\bf u$ as $r_i\Omega_i$, ${\bf B}_0$ as
130: $B_0$, and $\bf b$ as ${\rm Rm}B_0$.
131:
132: Taking the $t$, $z$ and $\phi$ dependence to be of the form
133: $\exp(\sigma t+ikz+im\phi)$, and using also $\nabla\cdot{\bf b}=0$ to
134: eliminate $b_z$, the $r$ and $\phi$ components of (3) become
135: $${\rm Rm}\,\sigma b_r=\nabla^2 b_r - r^{-2}b_r
136: - 2imr^{-2}b_\phi+imr^{-2}u_r
137: -{\rm Rm}\,im\Omega\,b_r,\eqno(5)$$
138: $${\rm Rm}\,\sigma b_\phi=\nabla^2 b_\phi - r^{-2}b_\phi + 2imr^{-2}b_r
139: +imr^{-2}u_\phi
140: +2r^{-2}u_r-{\rm Rm}\,im\Omega\,b_\phi + {\rm Rm}\,\Omega'r b_r,
141: \eqno(6)$$
142: where primes denote $d/dr$. The components of (4) have a similar structure,
143: but we will not need to refer to them in the subsequent discussion, and
144: hence do not list them. The boundary conditions associated with (3) and (4)
145: are
146: $$b_r=b_\phi'+r^{-1}b_\phi=u_r=u_\phi=u_z=0\eqno(7)$$
147: at $r=r_i$ and $r_o$, corresponding to perfectly conducting, no-slip walls.
148: The resulting one-dimensional linear eigenvalue problem is solved by finite
149: differencing in $r$, as in \cite{RSS}.
150:
151: Figure 1 shows stability curves for $m=1$, the only mode
152: that appears to become unstable. We see how an MRI exists that is remarkably
153: similar to some of the results described above. In particular, as ${\rm Pm}
154: \to0$, the relevant parameters are clearly once again $\rm Rm$ and $\rm S$,
155: with the MRI arising if ${\rm Rm}\gtrsim80$, and ${\rm S}\approx40$ yielding
156: the lowest value of $\rm Rm_c$. The specific numbers are roughly an order of
157: magnitude greater than for the axisymmetric MRI in the axial field, but the
158: basic scalings, and even the detailed shape of the instability curves, are
159: identical.
160:
161: \begin{figure}[htb]
162: \includegraphics[scale=0.5]{fig1.ps}
163: \caption{The critical magnetic Reynolds number for the onset of the MRI,
164: as a function of the Lundquist number, for the different values of $\rm Pm$
165: indicated. $m=1$, $r_o/r_i=2$, $\Omega_o/\Omega_i=0.5$.}
166: \end{figure}
167:
168: Figure 2 shows the real and imaginary parts of $\sigma$ in the unstable
169: regime. Remembering that time has been scaled by $\Omega_i^{-1}$, we see
170: that we obtain growth rates as large as $0.05\Omega_i$. So again, while the
171: particular number 0.05 is about an order of magnitude smaller than for the
172: axisymmetric MRI in the axial field, this non-axisymmetric MRI is
173: clearly also growing on the basic rotational timescale.
174:
175: To understand why this non-axisymmetric MRI exists even for a purely
176: toroidal field ${\bf B}_0$, for which it is known that the axisymmetric MRI
177: fails \cite{V59,BH91}, we need to consider the details of (5) and (6). In
178: particular, note that for $m=0$, $b_r$ completely decouples from everything
179: else, and inevitably decays away. Without $b_r$ though, the MRI cannot
180: proceed, as it relies on the term ${\rm Rm}\,\Omega'r b_r$ in (6). In
181: contrast, for $m=1$, $b_r$ is coupled both to $b_\phi$, coming from
182: $\nabla^2{\bf b}$, and to $u_r$, from $\nabla\times({\bf u\times B}_0)$.
183: And once $b_r$ is coupled to the rest of the problem, the term
184: ${\rm Rm}\,\Omega'r b_r$ then allows the MRI to develop. Figure 3 presents
185: an example of these solutions, indicating how all three components of both
186: $\bf u$ and $\bf b$ are indeed present.
187:
188: \begin{figure}
189: \vbox{
190: \includegraphics[scale=0.45]{fig2a.ps}
191: \includegraphics[scale=0.45]{fig2b.ps}}
192: \caption{The real (top) and imaginary (bottom) parts of $\sigma$, as
193: functions of $\rm S$, with $\rm Rm$ fixed at 200. The real part corresponds
194: to the growth rate, the imaginary part to the azimuthal drift rate.}
195: \end{figure}
196:
197: \begin{figure}
198: \vbox{
199: \hbox{
200: \includegraphics[width=2.8cm,height=5.0cm]{fig3a.ps}
201: \includegraphics[width=2.8cm,height=5.0cm]{fig3b.ps}
202: \includegraphics[width=2.8cm,height=5.0cm]{fig3c.ps}}
203: \hbox{
204: \includegraphics[width=2.8cm,height=5.0cm]{fig3d.ps}
205: \includegraphics[width=2.8cm,height=5.0cm]{fig3e.ps}
206: \includegraphics[width=2.8cm,height=5.0cm]{fig3f.ps}}
207: }
208: \caption{The marginally stable solution at ${\rm Pm}=0.1$, ${\rm S}=47.4$
209: and ${\rm Rm}=93$ (see Fig.\ 1). From left to right the $r$, $\phi$ and
210: $z$ components of $\bf u$ (top) and $\bf b$ (bottom). The real
211: parts are solid, imaginary parts dashed. Note how both $\bf u$ and $\bf b$
212: have the $z$ components largest. The azimuthal wavenumber is $k=1.88$.}
213: \end{figure}
214:
215: There is however one aspect that is not entirely clear from this
216: analysis, namely why this non-axisymmetric MRI actually requires the term
217: ${\rm Rm}\,\Omega'r b_r$ at all. In particular, the axisymmetric MRI with
218: a mixed field does {\it not} rely on it, but instead on the term $2r^{-2}u_r$,
219: coming from $\nabla\times({\bf u\times B}_0)$ rather than ${\rm Rm}\,\nabla
220: \times({\bf U}_0\times{\bf b})$ \cite{HR05}. The non-axisymmetric MRI is
221: evidently more like the axisymmetric MRI with a purely axial field, which
222: also requires the term ${\rm Rm}\,\Omega'r b_r$, because in that case the
223: term $2r^{-2}u_r$ is absent.
224:
225: To summarize then, we have shown that even if the externally imposed magnetic
226: field is purely toroidal, one still obtains an MRI, simply non-axisymmetric
227: rather than axisymmetric. In other respects though this new instability is
228: remarkably like the classical axisymmetric MRI with a purely axial field,
229: indeed far more like it than the axisymmetric MRI with a mixed field, which
230: yielded fundamentally different scalings. In contrast, the parameters here
231: continue to be $\rm Rm$ and $\rm S$, just as in the classical MRI.
232:
233: Note though that attempting to obtain this non-axisymmetric MRI in a
234: laboratory experiment would be even more difficult than attempting to obtain
235: the classical axisymmetric MRI in an axial field. First, the required
236: rotation rates would be even greater, ${\rm Re}\gtrsim O(10^8)$, with all
237: the difficulties that entails \cite{HF}. Even more daunting, imposing an
238: azimuthal field of the required strength, ${\rm Ha}\approx O(10^5)$, would
239: require a current along the central axis in excess of $10^6$ A, surely far
240: beyond any feasible experiment.
241:
242: This new non-axisymmetric MRI could have astrophysical applications though,
243: since many astrophysical objects do have predominantly azimuthal fields, in
244: which case the results presented here suggest that this non-axisymmetric
245: MRI could be preferred over the axisymmetric MRI.
246:
247: Despite the similarities in their fundamental scalings, the fact that the
248: classical MRI is axisymmetric, whereas this new MRI is non-axisymmetric,
249: means the nonlinear equilibration, and hence the associated angular momentum
250: transport, are potentially quite different, which could again have
251: astrophysical implications. Work on this is currently in progress.
252:
253: Finally, one might ask how the results presented here change if one
254: allows for a more general toroidal field profile, ${\bf B}_0=(c_1r^{-1}
255: +c_2r)\,{\bf\hat e}_\phi$, where the term $c_2r$ corresponds to
256: an electric current flowing through the fluid itself, and not just along
257: the central axis. Eq.\ (4) then contains an additional term
258: ${\rm Ha}^2(\nabla\times{\bf B}_0)\times{\bf b}$, which opens up the
259: possibility of instabilities driven entirely by this current
260: $\nabla\times{\bf B}_0$, without any rotation necessarily present at all
261: \cite{Tay}. Understanding how these current driven instabilities (also
262: $m=1$) interact with the magnetically catalyzed but ultimately
263: rotationally driven MRI presented here is also in progress.
264:
265: \begin{acknowledgments}
266: This work was supported by the German Leibniz Gemeinschaft, under program
267: SAW.
268: \end{acknowledgments}
269:
270: \begin{thebibliography}{}
271: \bibitem{V59}
272: E. P. Velikhov, {Sov.\ Phys.\ JETP} {\bf 36}, 995 (1959).
273: \bibitem{BH91}
274: S. A. Balbus and J. F. Hawley, {Astrophys.\ J.} {\bf 376}, 214 (1991).
275: \bibitem{HR05}
276: R. Hollerbach and G. R\"udiger, {Phys.\ Rev.\ Lett.} {\bf 95}, art no
277: 124501 (2005).
278: \bibitem{RH05}
279: G. R\"udiger, R. Hollerbach, M. Schultz and D. Shalybkov, {Astron.\ Nachr.}
280: {\bf 326}, 409 (2005).
281: \bibitem{Ji}
282: H. T. Ji, J. Goodman, and A. Kageyama, {Mon.\ Not.\ Roy.\ Astron.\ Soc.}
283: {\bf 325}, L1 (2001).
284: \bibitem{RZ}
285: G. R\"udiger and Y. Zhang, {Astron.\ Astrophys.} {\bf 378}, 302 (2001).
286: \bibitem{RSS}
287: G. R\"udiger, M. Schultz, and D. Shalybkov, {Phys.\ Rev.} E {\bf 67},
288: art no 046312 (2003).
289: \bibitem{S06}
290: F. Stefani, T. Gundrum, G. Gerbeth, G. R\"udiger, M. Schultz, J. Szklarski
291: and R. Hollerbach, {Phys.\ Rev.\ Lett.} {\bf 97}, art no 184502 (2006).
292: \bibitem{R06}
293: G. R\"udiger, R. Hollerbach, F. Stefani, T. Gundrum, G. Gerbeth and R.
294: Rosner, {Astrophys.\ J.} {\bf 649}, L145 (2006).
295: \bibitem{HF}
296: R. Hollerbach and A. Fournier, in AIP Conf.\ Proc.\ 733, MHD Couette Flows:
297: Experiments and Models, ed.\ R. Rosner, G. R\"udiger and A. Bonanno,
298: (New York: AIP), 114 (2004).
299: \bibitem{Tay}
300: R. J. Tayler, {Mon.\ Not.\ Roy.\ Astron.\ Soc.} {\bf 161}, 365 (1973).
301: \end{thebibliography}
302:
303: \end{document}
304: