astro-ph0406071/ms.tex
1: %\documentclass[12pt,preprint]{aastex}
2: \documentclass[onecolumn,10pt]{emulateapj}
3: %\documentclass[10pt,preprint]{aastex} \usepackage{amssymb}
4: \usepackage{times} 
5: 
6: 
7: \setlength{\parskip}{1.175ex plus 0.5ex minus 0.2ex}
8: 
9: \newcommand{\D}[2]{\frac{\partial #2}{\partial #1}}
10: \newcommand{\DD}[2]{\frac{\partial2 #2}{\partial #12}}
11: \newcommand\bb[1] {\mbox{\boldmath{$#1$}}}
12: \newcommand\del{\bb{\nabla}} \newcommand\bcdot{\bb{\cdot}}
13: \newcommand\btimes{\bb{\times}} \newcommand\beq{\begin{equation}}
14: \newcommand\eeq{\end{equation}} \newcommand{\Alfven}{Alfv\'{e}n }
15: \newcommand{\Alfvennospace}{Alfv\'{e}n}
16: 
17: \begin{document}
18: %\submitted{Draft, \today}
19: \shorttitle{\textsc{MAGNETIZED ROTATING PLASMAS WITH SUPERTHERMAL FIELDS}}
20: \shortauthors{\textsc{PESSAH \& PSALTIS}}
21: 
22:   
23: \title{\textsc{THE STABILITY OF MAGNETIZED ROTATING PLASMAS WITH
24:     SUPERTHERMAL FIELDS}} 
25: 
26: %\received{} \revised{}    \accepted{}
27: 
28: \author{\textsc{Martin E. Pessah}\altaffilmark{1,2} and
29: \textsc{Dimitrios Psaltis}\altaffilmark{2,1}}
30: \altaffiltext{1}{Astronomy Department, University of Arizona, 933
31: N. Cherry Ave., Tucson, AZ, 85721; mpessah@as.arizona.edu }
32: \altaffiltext{2}{Physics Department, University of Arizona, 1118 E.
33: 4th St., Tucson, AZ, 85721; dpsaltis@physics.arizona.edu}
34: 
35: 
36: \begin{abstract}
37: During the last decade it has become evident that the
38: magnetorotational instability is at the heart of  the enhanced angular
39: momentum transport in weakly magnetized accretion disks around neutron
40: stars and black holes.  In this paper, we investigate the local linear
41: stability of differentially rotating, magnetized flows and the
42: evolution of the magnetorotational instability beyond the weak-field
43: limit. We show that, when superthermal  toroidal fields are
44: considered,  the effects of both compressibility and magnetic tension
45: forces, which are related to  the curvature of toroidal field lines,
46: should be taken fully into account.  We demonstrate that 
47: the presence of a strong toroidal component
48: in the magnetic field plays a non-trivial role. When strong fields are
49: considered, the strength of the toroidal  magnetic field not only
50: modifies the growth rates of the unstable modes but also determines
51: which modes are subject to instabilities. We find that, for
52: rotating configurations with Keplerian laws,
53: the magnetorotational instability is stabilized at low wavenumbers for
54: toroidal \Alfven speeds exceeding the geometric mean of the sound
55: speed and the rotational speed. For a broad range of magnetic field
56: strengths, we also find that two additional distinct instabilities are
57: present; they both appear as the result of coupling between the modes
58: that become the \Alfven and the slow modes in the limit of no
59: rotation. We discuss the significance of our  findings for the
60: stability of cold, magnetically dominated, rotating fluids and argue
61: that, for these systems, the curvature of toroidal field lines cannot
62: be neglected even when short wavelength perturbations are considered.
63: We also comment on the implications of our results for the validity of
64: shearing box simulations in which superthermal toroidal fields are
65: generated.
66: \end{abstract}
67: 
68: 
69: \keywords{accretion, accretion disks --- MHD --- instabilities ---
70: plasmas}
71: 
72: \section{\textsc{INTRODUCTION}}
73: Linear mode analyses provide a useful  tool in gaining important
74: insight into the relevant physical  processes determining the
75: stability of magnetized accretion flows. Studies of local linear modes
76: of accretion disks threaded by weak magnetic fields have pointed out
77: important clues on viable mechanisms for angular momentum transport
78: and the subsequent accretion of matter onto the central objects
79: \citep{BH91, BH98, BH02, SM99, Balbus03}. They have also provided
80: simplified physical models and analogies over which more complex
81: physics can, in principle, be added \citep{BH92, BH98, Quataert02}.
82: These treatments were carried out in the magnetohydrodynamic (MHD)
83: limit  (but see Quataert, Dorland \& Hammett 2002 who studied the
84: kinetic limit) and invoked a number of approximations appropriate to
85: the study of the evolution of short-wavelength perturbations, when
86: weak fields are considered. In this context, the strength of the
87: magnetic field, $B$, is inferred by comparing the thermal pressure,
88: $P$,  to the magnetic pressure and is characterized by a plasma
89: parameter, $\beta \equiv 8\pi P/ B^2 > 1$.
90: 
91: 
92: It is not hard to find situations of astrophysical interest, however,
93: in which the condition of weak magnetic fields is not satisfied.  A
94: common example of such a situation is the innermost  region of an
95: accretion disk around a magnetic neutron star.  It is widely accepted
96: that X-ray pulsars are powered by accretion of matter onto the polar
97: caps of magnetic neutron stars.  For this to occur, matter in the
98: nearly Keplerian accretion disk has to be funneled along the field
99: lines. This suggests that, at some radius, centrifugal forces and
100: thermal pressure have to be overcome by magnetic stresses, leading
101: naturally to regions where $\beta \lesssim 1$.
102: 
103: 
104: In the context of accretion disks, the presence of superthermal fields
105: in rarefied coronae also seems hard to avoid, if coronal heating is a
106: direct consequence of the internal dynamics of the disk itself rather
107: than being produced by external irradiation from the central
108: object. Three-dimensional MHD simulations by \citet{MS00} showed  that
109: magnetic turbulence can effectively couple with buoyancy to transport
110: the  magnetic energy produced by the magnetorotational instability
111: (MRI) in weakly magnetized disks to create a strongly magnetized
112: corona within a few scale heights from the disk plane.   On long time
113: scales, the average vertical disk structure consists of a weakly
114: magnetized ($\beta \simeq 50$) turbulent core below two scale heights
115: and a strongly magnetized ($\beta \lesssim 0.1$) non-turbulent corona
116: above it. The late stages of evolution in these models show that the
117: disks themselves become magnetically dominated.  \citet{MHM00} also
118: found that the average plasma $\beta$ in  disk coronae is $\simeq
119: 0.1-1$ and the volume filling factor for regions with $\beta \lesssim
120: 0.3$ is up to $0.1$.  Even in the absence of an initial toroidal
121: component, simulations carried out by \citet{KMS02} showed that
122: low-$\beta$ regions develop near the equator of the disk  because of a
123: strong toroidal component of the magnetic field generated by shear.
124: On more theoretical grounds, strong toroidal magnetic fields produced
125: by strong shear in the boundary layer region  have been suggested as
126: responsible for the observed bipolar outflows in young stellar objects
127: \citep{Pringle89}.  More recently, \citet{PBB03} found self-consistent
128: solutions for thin magnetically-supported accretion disks and pointed
129: out the necessity of assessing the stability properties of such
130: configurations.
131: 
132: Another case in which magnetic fields seem to play an important
133: dynamical role in rotating fluid configurations is that of
134: magnetically supported molecular clouds.  Observations of both large
135: Zeeman line-splitting and of broad molecular  lines support
136: the presence of superthermal fields 
137: (see Myers \& Goodman 1988 for further references
138: and Bourke \& Goodman 2003 for a review on the current understanding
139: on the role of magnetic fields in molecular clouds).  Values of the
140: plasma $\beta$ of the order of $0.1-0.01$ have also been used in
141: numerical studies of the structural properties of giant molecular
142: clouds \citep{OSG01}.
143: 
144: 
145: As a last example of astrophysical interest,  we mention
146: magnetocentrifugally driven winds,  such as those observed  in
147: protostars. These outflows seem to play an important role in the
148: evolution of young  stellar objects and in the dynamics of the parent
149: clouds  by providing a source of turbulent energy.  Magnetocentrifugal
150: jets typically involve internal \Alfven speeds comparable to the flow
151: speeds.   These structures are supported mainly by magnetic pressure
152: due to strong toroidal fields. The ratio of magnetic pressure in the
153: jet to the gas pressure of the ambient medium can be of the order of
154: $10^6$ (for an extensive study of MHD driven instabilities in these
155: systems see Kim \& Ostriker 2000 and \S \ref{subsec:Comparison to
156: previous analytical studies}).
157: 
158: In this paper, we investigate the local linear stability of
159: differentially rotating flows without imposing any \emph{a priori}
160: restrictions on the strength of the magnetic field. We do, however,
161: restrict our attention to rotationally supported flows 
162: (we loosely use this term to refer to flows with
163: internal \Alfven speeds smaller than the rotational speed). Our intent
164: is to demonstrate that the effects of the finite curvature of the
165: toroidal field lines on the stability of small-wavelength vertical
166: perturbations (i.e., on the most unstable modes present in the
167: weak-field MRI) cannot be neglected when  superthermal toroidal fields
168: are present.  In order to achieve this task, we relax the Boussinesq
169: approximation (see also Papaloizou \& Szuszkiewicz 1992 and Blaes \&
170: Balbus 1994), which is valid only when the toroidal component of the
171: field is subthermal \citep{BH91}. We thus consider the MHD fluid to be
172: fully compressible. Moreover, even though we perform a local analysis,
173: we do consider curvature terms when evaluating magnetic forces, for
174: they become important in the strong-field regime (see also Knobloch
175: 1992 and Kim \& Ostriker 2000).
176: 
177: In most early studies addressing the MRI,  it was found that the only
178: role played by a toroidal component in the magnetic field is to quench
179: the growth rates of the modes that are already unstable when only weak
180: vertical fields are considered (Balbus \& Hawley 1991; Blaes \& Balbus
181: 1994; see Quataert, Dorland,  \& Hammett 2002 for the kinetic limit;
182: see also Kim \& Ostriker 2000  for the cold MHD limit). Here, we show
183: that, when strong fields are considered and the approximations usually
184: invoked in the study of the weak-field MRI  are relaxed, the presence
185: of a toroidal component of the magnetic field plays a crucial role not
186: only in the growth rates of the unstable modes but also in determining
187: which modes are subject to instabilities\footnote{We will comment
188: later in more detail on the paper by \citet{CP95} who outlined the
189: effects of a dynamically important toroidal field (in the case of an
190: incompressible MHD flow) and address the similitudes and differences
191: with our findings.}. As expected, the presence of a toroidal component
192: breaks the symmetry of the problem, also giving rise to traveling
193: modes. Moreover, for a broad range of magnetic field strengths, we
194: find that two different instabilities are present. They both appear as
195: the result of coupling between the modes that become the \Alfven and
196: the slow mode in the limit of no rotation.
197: 
198: The paper is organized as follows. In \S  \ref{sec:mhd equations and
199: dispersion relation},  we describe the physical setup to be studied,
200: present the dispersion relation to be solved, and discuss the
201: importance of curvature terms in the limit of superthermal fields.  In
202: \S \ref{sec:numerical solutions}, we solve numerically the dispersion
203: relation in some interesting regimes.   In \S \ref{sec:onset of
204: instabilities},  we study the onset of instabilities as a function of
205: magnetic field strengths and present some useful approximate criteria
206: that enable us to study analytically some aspects of the full
207: problem. In \S \ref{sec:discussion}, we compare our results to
208: previous investigations and discuss some of the implications of this
209: study. Finally, in  \S \ref{sec:summary and conclusions}, we present a
210: brief summary and our conclusions.
211: 
212: \section{\textsc{MHD EQUATIONS FOR PERTURBATIONS AND THE DISPERSION RELATION}}
213: \label{sec:mhd equations and dispersion relation}
214: 
215: We start with the set of equations that govern the behavior of a
216: polytropic fluid in the MHD approximation,
217: \begin{equation}
218: \label{eq:continuity}
219: \D{t}{\rho} + \del\bcdot(\rho\bb{v}) = 0 ~,
220: \end{equation}
221: \begin{eqnarray}
222: \label{eq:euler}
223: \rho\D{t}{\bb{v}} + \left(\rho\bb{v}\bcdot\del\right)\bb{v} =  -
224: \rho\del\Phi - \del\left(P + \frac{\bb{B}^2}{8\pi}\right)  +
225: \left(\frac{\bb{B}}{4\pi}\bcdot\del\right)\bb{B} ~,
226: \end{eqnarray}
227: \begin{equation}
228: \label{eq:induction}
229: \D{t}{\bb{B}} + \left( \del \bcdot \bb{v} \right) \bb{B} - \left(
230: \bb{B} \bcdot \del \right) \bb{v} + \left( \bb{v} \bcdot \del \right)
231: \bb{B} = 0 ~,
232: \end{equation}
233: and
234: \begin{equation}
235: \label{eq:energy}
236: P = P_0 \left(\frac{\rho}{\rho_0} \right)^\Gamma ~.
237: \end{equation}
238: In these equations, $\rho$ is the mass density, $\bb{v}$ the velocity,
239: $P$ the gas pressure, and $\Gamma$ the polytropic index; $\bb{B}$ is
240: the magnetic field and $\Phi$ the gravitational potential.   For
241: convenience we adopt a cylindrical set of coordinates $(r,\phi,z)$
242: with origin in the central object  (i.e., the neutron star or black
243: hole).  We assume a steady axisymmetric background flow
244: characterized by a velocity field of the form  $\bb{v} = v_{\phi}(r,
245: z) \hat \phi$ and threaded by a background magnetic field. For
246: consistency, our analysis is restricted to background fields of the
247: form $\bb{B} = B_{\phi} \hat \phi + B_z \hat z$ since the effect of
248: including a radial component in the field is to generate a linear
249: growth in time of the toroidal component \citep{BH91}.  Under these
250: circumstances, all the background quantities depend on the radial and
251: vertical coordinates only.   In the present treatment, we neglect the
252: self gravity of the fluid.  In fact, \citet{PBB03} showed that
253: magnetically dominated accretion disks have lower surface and volume
254: densities for a fixed accretion rate. This suggests that these systems
255: are  lighter than standard disks and thus are not subject to
256: self-gravity instabilities.
257: 
258: \subsection{Equations for the Perturbations}
259: \label{sec:equations for the perturbations}
260: 
261: In order to perform the local linear
262: mode analysis, we perturb the set of equations
263: (\ref{eq:continuity})-(\ref{eq:energy}) by substituting every physical
264: variable $f$ by $f + \delta f$ and retain only linear orders in
265: $\delta f$.   In the following, we focus our analysis on the study of
266: axisymmetric perturbations in an axisymmetric background.
267: 
268: We first consider, in some detail, the  radial component  of the
269: momentum equation (\ref{eq:euler}), which becomes
270: \begin{eqnarray}
271: \label{eq:euler_r_linear_step1}
272: \D{t}{\delta v_r} &-& 2\Omega \delta v_\phi + \frac{1}{\rho}  \left[
273: \D{r}{\delta P} + \frac{1}{4\pi}  \left( \D{r}{B_\phi} \delta B_\phi +
274: B_\phi  \D{r}{\delta B_\phi} + \D{r}{B_z} \delta B_z + B_z
275: \D{r}{\delta B_z} \right)\right] -\D{r}{P} \frac{\delta \rho}{\rho^2}
276: \nonumber \\ &-&\frac{1}{4\pi\rho} \left( B_\phi  \D{r}{B_\phi}
277: + B_z  \D{r}{B_z} \right) \frac{\delta \rho}{\rho} -
278: \frac{1}{4\pi\rho} \left[ B_z \D{z}{\delta B_r}   - 2 \frac{B_\phi}{r}
279: \delta B_\phi \right] - \frac{1}{4\pi\rho}  \frac{B_\phi^2}{r}
280: \frac{\delta \rho}{\rho} =0 ~.%\nonumber \\
281: \end{eqnarray}
282: The coefficients in this linear equation for the perturbed
283: variables depend in general on $r$ and $z$,  [e.g., the angular
284: velocity $\Omega$ is in general $\Omega(r,z)$].  Therefore, at this
285: point, the decomposition of perturbed quantities in Fourier modes --
286: e.g., symbolically $\delta f = \sum \delta f_k  e^{i(k_r r+k_z z -
287: \omega t)}$ -- would not result in any particular simplification of
288: the problem.  This can be seen by taking the Fourier transform of
289: equation (\ref{eq:euler_r_linear_step1}) which results in a sum of
290: convolutions of the Fourier transforms of background and perturbed
291: quantities.  Further progress can be made if we restrict the
292: wavelengths of the perturbations for which our stability analysis is
293: valid. To this end, we choose a fiducial point, $\bb{r}_0 = (r_0,
294: \phi_0, z_0)$, around which we perform the local stability analysis.
295: The choice of the particular value of $\phi_0$ is, of course,
296: irrelevant in the axisymmetric case under study.
297: 
298: We expand all the background quantities in equation
299: (\ref{eq:euler_r_linear_step1}) in Taylor series around $\bb{r}_0$ and
300: retain only the zeroth order in terms of the local coordinates
301: $\xi_r=r-r_0$ and $\xi_z=z-z_0$ to obtain
302: \begin{eqnarray}
303: \label{eq:euler_r_linear_step2}
304: \D{t}{\delta v_r}  &-& 2\Omega_0 \delta v_\phi  + \frac{1}{\rho_0}
305: \left[ \D{\xi_r}{\delta P}  + \frac{1}{4\pi} \left(
306:   \left.\D{r}{B_\phi}\right|_0 \delta B_\phi  + B^0_\phi
307:   \D{\xi_r}{\delta B_\phi} +  \left.\D{r}{B_z}\right|_0 \delta B_z  +
308:   B^0_z  \D{\xi_r}{\delta B_z} \right)\right] -
309: \left.\D{r}{P}\right|_0 \frac{\delta \rho}{\rho_0^2} \nonumber \\
310: &-&\frac{1}{4\pi\rho_0} \left( B^0_\phi  \left. \D{r}{B_\phi} \right|_0 +
311: B^0_z  \left.\D{r}{B_z} \right|_0\right) \frac{\delta \rho}{\rho_0} -
312: \frac{1}{4\pi\rho_0} \left( B^0_z \D{\xi_z}{\delta B_r}   - 2
313: \frac{B^0_\phi}{r_0} \delta B_\phi \right) - \frac{1}{4\pi\rho_0}
314: \frac{(B^0_\phi)^2}{r_0} \frac{\delta \rho}{\rho_0} =0 ~.%\nonumber \\
315: \end{eqnarray}
316: Here, $\Omega_0$, $\rho_0$, $B_\phi^0$, and $B_z^0$  stand for the
317: angular velocity,  background density, and magnetic field components
318: at the fiducial point $\bb{r}_0$ and the subscript ``$0$'' in the
319: derivatives with respect to the radial coordinate $r$ indicates that they
320: are evaluated at $\bb{r}_0$.  Equation (\ref{eq:euler_r_linear_step2})
321: is a linear partial-differential  equation in the local variables for
322: the perturbed quantities  but with constant coefficients.  This is a
323: good approximation as long as the departures $(\xi_r, \xi_z)$ are
324: small compared to the length scales over which there are significant
325: variations in the background quantities, i.e,  $\xi_r\ll L_r$ and
326: $\xi_z\ll L_z$, where $L_r$ and $L_z$ are the characteristic length
327: scales in the radial and vertical directions, respectively.
328: 
329: 
330: It is only now that it is useful to expand the perturbed quantities
331: in equation (\ref{eq:euler_r_linear_step2}) in Fourier modes.  We can thus
332: write each one of the perturbed quantities as
333: \begin{equation}
334: \delta f = \delta   f(k_r, k_z, \omega)  \, e^{i(k_r \xi_r+k_z \xi_z -
335: \omega t)}  ~,
336: \end{equation}
337: and write the radial momentum equation for each mode as
338: \begin{eqnarray}
339: \label{eq:euler_r_linear_step3}
340: - i \omega \delta   v_r &-& 2\Omega_0 \delta   v_\phi  - ik_z
341: \frac{B^0_z \delta   B_r}{4\pi\rho_0} + \left( ik_r + \frac{2}{r_0} +
342: \left.\D{r}{\ln B_\phi}\right|_0 \right) \frac{B_\phi^0 \delta
343:   B_\phi}{4\pi\rho_0} + \left( ik_r + \left.\D{r}{\ln B_z}\right|_0
344: \right)  \frac{B_z^0 \delta   B_z}{4\pi\rho_0} \nonumber \\ &+& \left
345: \{ \left(ik_r -  \left.\D{r}{\ln \rho}\right|_0 \right)  (c^0_{\rm
346:   s})^2  -\left[ \left( \frac{1}{r_0}
347:   +\left.\D{r}{\ln B_\phi}\right|_0 \right) (v^0_{{\rm A} \phi})^2 +
348:   \left.\D{r}{\ln B_z}\right|_0  (v^0_{{\rm A} z})^2 \right] \right \}
349: \frac{\delta \rho}{\rho_0}=0 ~.%\nonumber \\
350: \end{eqnarray}
351: Here, we have introduced the quantities $c^0_{\rm s}, v^0_{{\rm A}
352: \phi}$ and  $v^0_{{\rm A} z}$ that stand for the local sound speed and
353: local \Alfven speeds associated with the toroidal and vertical
354: components of the local magnetic field and are defined by
355: \begin{equation}
356: \label{eq:speeds}
357: c^0_{\rm s} \equiv \sqrt{\Gamma \frac{P_0}{\rho_0}} ~, \qquad
358: v^0_{{\rm A}\phi} \equiv  \frac{B^0_\phi}{\sqrt{4\pi \rho_0}} ~,
359: \qquad \textrm{and} \qquad v^0_{{\rm A} z} \equiv
360: \frac{B^0_z}{\sqrt{4\pi \rho_0}}  ~.
361: \end{equation}
362: Note that, for brevity, we have omitted the dependences $(k_r, k_z,
363: \omega)$ in the Fourier amplitudes. For consistency, the validity of
364: the analysis  is now restricted to modes with wavenumbers satisfying
365: $k_r L_r \gg 1$ and $k_z L_z \gg 1$.  Without loss of generality, we
366: assume that the fiducial point is inside the disk so we can write
367: the local conditions on the wavenumbers as $k_r r_0 \gg 1$ and $k_z
368: z_0 \gg 1$ . Moreover, for fiducial points such that $r_0 \ge z_0$ the
369: latter condition  also implies $k_z r_0 \gg 1$.
370: 
371: 
372: At this point, it is also convenient to define a new set of
373: independent variables  $(\delta v_{{\rm A} r}, \delta v_{{\rm A}
374: \phi},\delta v_{{\rm A} z} )$ defined in terms of $(\delta B_r, \delta
375: B_\phi, \delta B_z)$ in such a way that $\delta \bb{v}_{A} \equiv \delta
376: \bb{B}/\sqrt{4\pi \rho_0}$.  In this case, equation
377: (\ref{eq:euler_r_linear_step3}) reads
378: \begin{eqnarray}
379: \label{eq:euler_r_linear_step4}
380: - i \omega \delta v_r &-& 2\Omega_0 \delta v_\phi  - ik_z v^0_{{\rm A}
381:   z} \delta v_{{\rm A} r} + \left( ik_r + \frac{2}{r_0} +
382: \left.\D{r}{\ln B_\phi}\right|_0 \right) v^0_{{\rm A} \phi} \delta
383: v_{{\rm A} \phi} + \left( ik_r + \left.\D{r}{\ln B_z}\right|_0 \right)
384: v^0_{{\rm A} z} \delta v_{{\rm A} z} \nonumber \\ &+& \left \{
385: \left(ik_r -  \left.\D{r}{\ln \rho}\right|_0 \right) (c^0_{\rm s})^2 -
386: \left[ \left( \frac{1}{r_0} + \left.\D{r}{\ln B_\phi}\right|_0 \right)
387:   (v^0_{{\rm A} \phi})^2  +  \left.\D{r}{\ln B_z}\right|_0 (v^0_{{\rm
388:       A} z})^2\right] \right \}\frac{\delta \rho}{\rho_0}=0 ~.%\nonumber \\
389: \end{eqnarray}
390: 
391: As a last step, it is useful to work with dimensionless quantities.
392: To this end, we define dimensionless variables by scaling  all the
393: frequencies with the local rotational frequency $\Omega_0$ and all
394: speeds with the local circular velocity $\Omega_0 r_0$.  It is also
395: convenient to define  dimensionless wavenumbers by multiplying the
396: physical wavenumber by the radial coordinate $r_0$.  In summary, we
397: define
398: \begin{eqnarray}
399: \tilde \omega  = \omega/\Omega_0 ~,  \qquad  &\tilde k_r = k_r r_0 ~,&
400: \qquad  \tilde k_z = k_z r_0 ~, \\ \tilde c^0_{\rm s} = c^0_{\rm
401: s}/\Omega_0 r_0 ~,  \qquad  &\tilde v^0_{{\rm A} \phi} =  v^0_{{\rm A}
402: \phi}/\Omega_0 r_0 ~, &\qquad  \tilde v^0_{{\rm A} z} =  v^0_{{\rm A}
403: z}/\Omega_0 r_0 ~, \\ \delta \tilde  \rho = \delta \rho/\rho_0 ~,
404: \qquad  &\delta \tilde{\bb{v}} = \delta \bb{v}/\Omega_0 r_0 ~,& \qquad
405: \delta \tilde{\bb{v}}_{\rm A} = \delta \bb{v}_{\rm A}/\Omega_0 r_0 ~.
406: \end{eqnarray}
407: For completeness,  we introduce here the epicyclic frequency $\kappa$
408: and its local dimensionless counterpart
409: \begin{equation}
410: \label{eq:epicyclic}
411: \kappa = 2\Omega \left[1 + \frac{1}{2}\frac{d\ln \Omega}{d\ln
412:     r}\right]^{1/2} \qquad \textrm{and} \qquad \tilde{\kappa}_0 =
413:     \frac{\kappa_0}{\Omega_0}  = 2 \left[1 + \frac{1}{2}
414:     \left. \frac{d\ln \Omega}{d\ln r} \right|_0  \right]^{1/2} ~.
415: \end{equation}
416: This quantity appears naturally in stability analyses of
417: differentially rotating configurations and it is the frequency at
418: which all the flow variables oscillate around their background values
419: in the absence of magnetic fields. For a rotational profile given by a
420: power law, the epicyclic frequency is proportional to the angular
421: frequency at all radii.
422: 
423: Finally, the dimensionless version of equation
424: (\ref{eq:euler_r_linear_step4}) reads,
425: \begin{eqnarray}
426: \label{eq:euler_r_linear_step_5}
427: - i \tilde \omega \delta \tilde v_r &-& 2\delta \tilde  v_\phi  - i
428: \tilde k_z \tilde v^0_{{\rm A} z} \delta \tilde v_{{\rm A} r} + \left(
429: i\tilde k_r + 2 + \left.\D{\ln r}{\ln B_\phi}\right|_0 \right) \tilde
430: v^0_{{\rm A} \phi} \delta \tilde v_{{\rm A} \phi} + \left( i\tilde k_r
431: + \left.\D{\ln r}{\ln B_z}\right|_0 \right)  \tilde v^0_{{\rm A} z}
432: \delta \tilde v_{{\rm A} z} \nonumber \\ &+& \left \{ \left(i\tilde
433: k_r -  \left.\D{\ln r}{\ln \rho}\right|_0 \right) (\tilde c^0_{\rm
434:   s})^2 - \left[ \left( 1 + \left.\D{\ln r}{\ln B_\phi}\right|_0
435:   \right) (\tilde v^0_{{\rm A}\phi})^2  +  \left.\D{\ln r}{\ln
436:     B_z}\right|_0 (\tilde v^0_{{\rm A} z})^2\right] \right \}\delta
437: \tilde \rho =0 ~.%\nonumber \\
438: \end{eqnarray} 
439: 
440: Following a similar procedure with the remaining equations in the
441: system (\ref{eq:continuity})-(\ref{eq:energy}),  we arrive  to the
442: linear set of equations needed to perform the local stability
443: analysis. For brevity, we now drop the hat in all the dimensionless
444: variables and  the superscript in $\tilde{c}^0_{\rm s},
445: \tilde{v}^0_{{\rm A} \phi}$ and  $\tilde{v}^0_{{\rm A} z}$. We then
446: write
447: \begin{eqnarray}
448: \label{eq:pert_cont_grads}
449: - i  \omega \delta  \rho + \left( i k_r + \epsilon_4 + \left.\D{\ln
450:   r}{\ln \rho}\right|_0 \right) \delta   v_r +  \left( i k_z  +
451: \frac{r_0}{z_0}\left.\D{\ln z}{\ln \rho}\right|_0 \right) \delta v_z
452: =0 ~,
453: \end{eqnarray}
454: %
455: \begin{eqnarray}
456: \label{eq:pert_euler_r_grads}
457: - i  \omega \delta  v_r &-& 2\delta   v_\phi  - i  k_z  v_{{\rm A} z}
458: \delta  v_{{\rm A} r} + \left( i k_r + 2\epsilon_1 + \left.\D{\ln
459:   r}{\ln B_\phi}\right|_0 \right)  v_{{\rm A} \phi} \delta  v_{{\rm A}
460:   \phi} + \left( i k_r + \left.\D{\ln r}{\ln B_z}\right|_0 \right)  v
461: _{{\rm A} z} \delta  v_{{\rm A} z} \nonumber \\ &+& \left \{ \left(i
462: k_r -  \left.\D{\ln r}{\ln \rho}\right|_0 \right)  c_{\rm s}^2 -
463: \left[ \left( \epsilon_2 + \left.\D{\ln r}{\ln B_\phi}\right|_0
464:   \right) v_{{\rm A} \phi}^2  +  \left.\D{\ln r}{\ln B_z}\right|_0
465:   v_{{\rm A} z}^2\right] \right \}\delta  \rho =0 ~,
466: \end{eqnarray}
467: %
468: \begin{eqnarray}
469: \label{eq:pert_euler_phi_grads}
470: - i  \omega \delta  v_\phi + \frac{ \kappa^2}{2} \delta   v_r +
471: \frac{r_0}{z_0}  \left.\D{\ln z}{\ln \Omega}\right|_0  \delta   v_z
472: &-&  \left( \epsilon_3 + \left.\D{\ln r}{\ln B_\phi}\right|_0  \right)
473: v_{{\rm A} \phi}  \delta  v_{{\rm A} r}  \nonumber \\ &-& i k_z
474: v_{{\rm A} z} \delta v_{{\rm A} \phi} - \frac{r_0}{z_0} \left.\D{\ln
475:   z}{\ln B_\phi}\right|_0  v_{{\rm A} \phi}  \delta  v_{{\rm A} z} =
476: 0~, 
477: \end{eqnarray}
478: %
479: \begin{eqnarray}
480: \label{eq:pert_euler_z_grads}
481: - i  \omega \delta  v_z 
482: &-&\left.\D{\ln r}{\ln B_z}\right|_0  v_{{\rm A} z} 
483: \delta  v_{{\rm A} r} + \left( i k_z +
484: \frac{r_0}{z_0}\left.\D{\ln z}{\ln B_\phi}\right|_0  \right) v_{{\rm
485: A} \phi} \delta  v_{{\rm A} \phi} \nonumber  \\ &+& \left[ \left(i
486: k_z -  \frac{r_0}{z_0}\left.\D{\ln z}{\ln \rho}\right|_0 \right)
487: c_{\rm s}^2 - \frac{r_0}{z_0} \left(\left.\D{\ln z}{\ln B_\phi}
488: \right|_0  v_{{\rm A} \phi}^2  
489: +  \left.\D{\ln z}{\ln B_z}\right|_0 v_{{\rm A} z}^2 \right)
490: \right] \delta  \rho =0 ~, \nonumber \\
491: \end{eqnarray}
492: %
493: \begin{equation}
494: \label{eq:pert_induc_r_grads}
495: i \omega \delta v_{{\rm A} r} + ik_z v_{{\rm A} z} \delta v_r = 0 ~,
496: \end{equation}
497: %
498: \begin{eqnarray}
499: \label{eq:pert_induc_phi_grads}
500: - i \omega \delta v_{{\rm A} \phi}  
501: - \left.\frac{d \ln \Omega}{d \ln r} \right|_0  \delta v_{{\rm A} r}  
502: - \left.\frac{r_0}{z_0}\frac{d \ln \Omega}{d \ln z}\right|_0  \delta v_{{\rm A} z}  
503: &+& \left(i k_r + \left.\D{\ln r}{\ln B_\phi}\right|_0 \right)  v_{{\rm A} \phi}
504: \delta  v_r - i k_z  v_{{\rm A} z} \delta v_\phi  \nonumber  \\ &+& \left(i k_z +
505: \frac{r_0}{z_0}\left.\D{\ln z}{\ln B_\phi}\right|_0 \right)  
506: v_{{\rm A} \phi} \delta  v_z =0 ~,  
507: \end{eqnarray}
508: and
509: \begin{equation}
510: \label{eq:pert_induc_z_grads}
511: - i \omega \delta v_{{\rm A} z} + \left(ik_r + \epsilon_4   +
512: \left.\D{\ln r}{\ln B_z}\right|_0 \right) v_{{\rm A} z} \delta v_r = 0
513: ~,
514: \end{equation}
515: where we have used  equation (\ref{eq:energy}) to recast the pressure
516: perturbations in terms of density perturbations.  The local conditions
517: over the wavenumbers now read $k_r \gg 1$ and $k_z \gg 1$.
518: 
519: The factors $\epsilon_i$, with $i=1,2,3,4$, are just convenient dummy
520: variables that we introduce in order to help us keep track of the
521: terms that account for the finite curvature  of the background  and
522: are usually neglected in local studies of the weak-field MRI. Their
523: numerical values are to be regarded as unity, unless otherwise
524: mentioned.   The terms proportional to $\epsilon_1$ and $\epsilon_2$
525: in equation (\ref{eq:pert_euler_r_grads}) and the term  proportional
526: to $\epsilon_3$ in equation (\ref{eq:pert_euler_phi_grads}) are due to
527: the effects of magnetic tension and they appear naturally when a
528: cylindrical coordinate system is adopted. The terms proportional to
529: $\epsilon_4$ in equations  (\ref{eq:pert_cont_grads}) and
530: (\ref{eq:pert_induc_z_grads}) are related to flux conservation in
531: cylindrical coordinates.   Although the three terms labeled by
532: $\epsilon_1$, $\epsilon_2$, and $\epsilon_3$ share the same physical
533: origin (i.e., magnetic tension introduced by the curvature of toroidal
534: field lines), it is useful to be able to distinguish among  them
535: because the one labeled with $\epsilon_2$ vanishes in the limit of an
536: incompressible flow.   Note that  equations
537: (\ref{eq:pert_induc_r_grads}) and (\ref{eq:pert_induc_z_grads}) ensure
538: a divergence-free perturbed magnetic field, i.e., 
539: $\del\bcdot \delta \bb{B} =0$, only when the finite curvature of 
540: the background is  accounted for (i.e., $\epsilon_4=1$).
541: 
542: 
543: Up to this point, our intention has been to keep the discussion as
544: general as possible in order to clearly state all the assumptions that
545: we have made to obtain the set of equations for the
546: perturbations to perform a  local linear mode analysis. For the sake
547: of simplicity, and to avoid the parametric study from being too
548: extensive, we further invoke the following assumptions. We choose the
549: fiducial point $\bb{r}_0$ to lie in the disk mid-plane and assume
550: that, locally, the vertical gradients in all background quantities are
551: negligible and set them to zero. This will be a good approximation as
552: long as we consider equilibrium configurations such that
553: \begin{equation}
554: \label{eq:logarithmic_derivatives_z}
555: \left|\frac{d\ln \rho}{d\ln z}  \right|_0  \ll 1  ~, \qquad
556: \left|\frac{d\ln B_\phi}{d\ln z}\right|_0  \ll 1  ~, \qquad
557: \textrm{and} \qquad \left|\frac{d\ln \Omega}{d\ln z}\right|_0  \ll 1
558: ~. \qquad
559: \end{equation}
560: Note that the solenoidal character of the magnetic field ensures that,
561: for $\bb{B} = B_{\phi}(r,z) \hat \phi + B_z(r,z) \hat z$, the condition
562: $\partial B_z/\partial z=0$ holds for arbitrary $z$.
563: 
564: In general, the forces induced by the curvature terms (e.g., the one
565: proportional to $\epsilon_2$ in  eq.~[\ref{eq:pert_euler_r_grads}])
566: and those induced by background (logarithmic) gradients in the radial
567: direction, (e.g., the term proportional to   $d\ln B_\phi/d\ln r|_0$
568: in the same equation) will not cancel each other.  As a single
569: exception, for the case in which $B_\phi \propto r^{-1}$, the most
570: important effects due to the finite curvature of toroidal  field lines
571: are canceled out by the gradients in the toroidal field.  This
572: particular case, however, might not be completely relevant to
573: realistic rotating flows since, in order to ensure force balance when
574: the thermal pressure can be neglected against magnetic stresses, the
575: magnetic field strength must decline more slowly than $r^{-1}$ (see,
576: e.g.,  Kim \& Ostriker 2000). For simplicity, we further focus our
577: attention on the study of differentially rotating, axisymmetric MHD
578: flows with locally negligible radial gradients in the background 
579: density and magnetic field, i.e.,
580: \begin{equation}
581: \label{eq:logarithmic_derivatives_r}
582: \left|\frac{d\ln \rho}{d\ln r}  \right|_0  \ll 1  ~, \qquad
583: \left|\frac{d\ln B_\phi}{d\ln r}\right|_0  \ll 1  ~, \qquad
584: \textrm{and} \qquad  \left|\frac{d\ln B_z}{d\ln r}   \right|_0  \ll 1
585: ~. \qquad
586: \end{equation}
587: In the rest of the paper, we consider that the only background flow
588: variable with a non-negligible local radial gradient is the angular
589: velocity $\Omega \propto r^{-q}$ and set all other radial gradients to
590: zero.  Note that this assumption is widely invoked in many
591: investigations of the weak-field MRI (e.g., Blaes \& Balbus 1994;
592: Balbus \& Hawley 1998; Blaes \& Socrates 2001; Quataert, Dorland, \&
593: Hammett 2002; Balbus 2003).  The assumption that the only background
594: variable with a non-negligible  radial gradient is the angular
595: velocity is also generally a part of the initial set of conditions
596: used in many numerical analyses of the MRI in the shearing box
597: approximation (e.g., Hawley, Gammie, \& Balbus 1994, 1995, 1996;
598: Miller \& Stone 2000).
599: 
600: In spite of being linear in the perturbed quantities, the terms
601: proportional to $\epsilon_i$  have been neglected in previous local
602: studies  of the MRI under the assumption that $k_r \gg 1$ and $k_z
603: \gg1$ (but see also Knobloch 1992 and Kim \& Ostriker 2000).  Although
604: comparing an imaginary term against a real one in a stability analysis
605: might seem particularly risky, this might not be a bad argument in
606: order to neglect the terms proportional to $\epsilon_1$  in equation
607: (\ref{eq:pert_euler_r_grads})  or $\epsilon_4$ in equations
608: (\ref{eq:pert_cont_grads}) and (\ref{eq:pert_induc_z_grads}) against
609: $ik_r$ (but see the discussion in Appendix A). The same could be said
610: about the terms proportional to $\epsilon_2$ in equation
611: (\ref{eq:pert_euler_r_grads})  or $\epsilon_3$ in equation
612: (\ref{eq:pert_euler_phi_grads}) in the limit of a very weak toroidal
613: component in the magnetic field, given that both of them are
614: proportional to $v_{{\rm A} \phi}$.  It is not  evident, however, that
615: we can neglect the terms proportional to either $\epsilon_2$ in
616: equation  (\ref{eq:pert_euler_r_grads}) or $\epsilon_3$ in equation
617: (\ref{eq:pert_euler_phi_grads}) if we are to explore the regime of
618: strong toroidal fields. There are two different reasons for this.  In
619: order to neglect the term proportional to  $\epsilon_2$ against the
620: one proportional to $k_r$ in  equation (\ref{eq:pert_euler_r_grads})
621: we should be able to ensure that the condition
622: $(\epsilon_2/k_r)(v_{{\rm A}\phi}^2/c_{\rm s}^2) \ll 1$ is always
623: satisfied, since both terms are proportional to $\delta \rho$.  In
624: this particular case, neglecting the forces induced by the bending of
625: toroidal field lines  becomes a progressively worse approximation the
626: colder the disk is and is not well justified in the limit $c_{\rm s}
627: \rightarrow 0$.  The case presented in equation
628: (\ref{eq:pert_euler_phi_grads}) is even harder to justify a priori
629: since now we would need to guarantee that the  condition
630: $(\epsilon_3/k_z)(v_{{\rm A} \phi}/v_{{\rm A} z}) (\delta v_{{\rm A}
631: r} /\delta v_{{\rm A}\phi}) \ll 1$ is always satisfied.  However, this
632: ratio is not only proportional to  $v_{{\rm A} \phi}/v_{{\rm A} z}$,
633: which might not be negligible in many astrophysical contexts but,
634: through the ratio  $\delta v_{{\rm A} r} /\delta v_{{\rm A}\phi}$,  is
635: also a function of $k_r$, $k_z$, and $\omega(k_r, k_z)$; the magnitude
636: of this term is therefore  unknown until we solve the problem fully.
637: A similar situation to this one is encountered if we aim to compare
638: the term $\epsilon_2  v_{{\rm A} \phi}^2 \delta \rho$  with the term
639: proportional to $k_z v_{{\rm A} z} \delta v_{{\rm A} r}$ in equation
640: (\ref{eq:pert_euler_r_grads})   (see \S \ref{subsec:importance} for
641: further discussion).
642: 
643: For the sake of consistency and in order  not to impose a constraint
644: on the magnitude of the toroidal \Alfven speed with respect to the
645: sound speed we keep all the terms proportional to the parameters 
646: $\epsilon_i$. We will later show that the term proportional to 
647: $\epsilon_1$ is  negligible when superthermal toroidal fields 
648: are considered. We will also discuss under which conditions the terms
649: proportional to $\epsilon_4$ can be neglected and why the terms
650: proportional to $\epsilon_2$ and $\epsilon_3$ are particularly
651: important.
652: 
653: 
654: \subsection{Dispersion Relation}
655: \label{sec:dispersion relation}
656: 
657: In order to seek for non-trivial solutions of the homogeneous system
658: of linear equations
659: (\ref{eq:pert_cont_grads})-(\ref{eq:pert_induc_z_grads}) we set its
660: determinant to zero. The resulting characteristic polynomial is
661: \begin{eqnarray}
662: \label{eq:disp_full_nodim_kr}
663: \omega^6  &-& \{(k_z^2+k_r^2)(c_{\rm s}^2+v_{{\rm A} \phi}^2 + v_{{\rm
664: A}z}^2)  - ik_r [(2\epsilon_1 -\epsilon_2) v_{{\rm A} \phi}^2   +
665: \epsilon_4 (c_{\rm s}^2 + v_{{\rm   A}z}^2)] + k_z^2 v_{{\rm  A}z}^2 +
666: \kappa^2 + \epsilon_2 \epsilon_4 v_{{\rm A} \phi}^2 \}  \omega^4 \nonumber \\
667: &-& \left( 2\epsilon_1 + \epsilon_3 \right) 2 k_z v_{{\rm A}\phi}
668: v_{{\rm A}z}  \omega^3 + \bigg \lbrace k_z^2 v_{{\rm A}z}^2 [
669: (k_z^2+k_r^2-\epsilon_4ik_r)  (2c_{\rm s}^2 + v_{{\rm A}\phi}^2 +
670: v_{{\rm A}z}^2) + ik_r (\epsilon_2-\epsilon_3) v_{{\rm A}\phi}^2 ]
671: \nonumber \\  &+& k_z^2\left[\kappa^2(c_{\rm s}^2 + v_{{\rm
672: A}\phi}^2)   + 2 \epsilon_1 \epsilon_4  c_{\rm s}^2 v_{{\rm A}\phi}^2  
673: + \epsilon_2 \epsilon_4  v_{{\rm A}\phi}^4  + 2 \frac{d \ln
674: \Omega}{d \ln r} v_{{\rm A}z}^2 + (\epsilon_2 \epsilon_4   -  2
675: \epsilon_1 \epsilon_3) v_{{\rm A}\phi}^2 v_{{\rm A}z}^2 \right]
676: \bigg \rbrace \omega^2 \nonumber \\ &+& 2 k_z^3 v_{{\rm A}\phi}
677: v_{{\rm A}z} \left[  ( 2 \epsilon_1 + \epsilon_3 + \epsilon_4)
678: c_{\rm s}^2+ ( \epsilon_2 + \epsilon_3 ) v_{{\rm A}\phi}^2\right]
679: \omega \nonumber  \\  &-& k_z^4 v_{{\rm A}z}^2 \left[
680: (k_z^2+k_r^2-\epsilon_4 ik_r) c_{\rm s}^2 v_{{\rm A}z}^2 + 2
681: \frac{d \ln \Omega}{d \ln r} c_{\rm s}^2 - 2 \epsilon_1 \epsilon_3
682: c_{\rm s}^2 v_{{\rm A}\phi}^2 - \epsilon_2 \epsilon_3 v_{{\rm
683: A}\phi}^4 \right] =0 ~,
684: \end{eqnarray} 
685: where we have dropped the subscript ``$0$'' in the radial logarithmic
686: derivative of the angular frequency.  This is the most general
687: dispersion relation under our  current set of assumptions.  When all
688: the parameters $\epsilon_i$ are set equal to zero, we recover the results of
689: previous analyses where the curvature of the toroidal field lines  was
690: not considered (e.g., Blaes \& Balbus 1994; Balbus \& Hawley 1998),
691: while when they are set equal to unity we obtain our full dispersion relation.
692: 
693: Although the original linear system
694: (\ref{eq:pert_cont_grads})-(\ref{eq:pert_induc_z_grads}) related seven
695: variables (recall that we had eliminated $\delta P$ in terms of
696: $\delta \rho$ using eq.~[\ref{eq:energy}] which is time-independent),
697: the characteristic polynomial is only of $6^{{\rm th}}$ degree. This
698: is easily understood by noting that  equations
699: (\ref{eq:pert_induc_r_grads}) and (\ref{eq:pert_induc_z_grads}) can be
700: combined into one single equation expressing the solenoidal character
701: of the perturbations in the magnetic field, $\del\bcdot \delta \bb{B}
702: =0$. This implies a relationship between $\delta B_r$ and $\delta B_z$
703: (or equivalently between $\delta v_{{\rm A} r}$ and $\delta v_{{\rm A}
704: z}$) that must be satisfied at all times and is, therefore,
705: independent of $\omega$. The fact that the dispersion relation
706: (\ref{eq:disp_full_nodim_kr}) is of $6^{th}$ and not of $4^{{\rm th}}$
707: degree is because  we are taking
708: into account the effects of finite compressibility.  This can be seen
709: immediately by taking the limit $c_{\rm s} \rightarrow \infty$.
710: 
711: Once all the
712: dimensionless variables have been properly defined, it is not evident
713: that the magnetic-tension terms, proportional to $\epsilon_1$,
714: $\epsilon_2$, and  $\epsilon_3$, will play a negligible role in
715: determining the eigenfrequencies $\omega$.  This is because the
716: non-vanishing toroidal component of the magnetic field introduces odd
717: powers in the dispersion relation and hence break its even
718: symmetry. In fact, small modifications in the odd-power coefficients
719: can and do have an important impact on the nature (real vs. complex)
720: of the solutions.  As we will see in \S \ref{sec:numerical solutions}
721: and describe in further detail in   \S \ref{sec:onset of
722: instabilities},  these curvature terms introduce further coupling
723: between the radial and toroidal directions, which in turn result in a
724: strong coupling between the \Alfven and the slow mode.
725: 
726: Also important is the fact that some of the coefficients in the
727: dispersion relation (\ref{eq:disp_full_nodim_kr})  are no longer real
728: due to the  factors $ik_r$. The presence of these  terms  does not
729: allow us to affirm that complex roots will appear in conjugate pairs.
730: As we discuss in Appendix A, the terms proportional to $ik_r$ play an
731: important role in determining the stability of modes for which the
732: ratio $k_r/k_z$ is non-negligible, even in the local limit, i.e.,
733: when $k_r \gg 1$.  Of course the smaller the ratio $k_r/k_z$, the
734: smaller the effects  of the factors $ik_r$ will be. If we consider the
735: limit $k_z \gg k_r$ in equation
736: (\ref{eq:disp_full_nodim_kr}), 
737: the imaginary part of all the coefficients in the dispersion
738: relation will become negligible. 
739: In this limiting case, whenever a
740: given complex root is a solution of the dispersion relation
741: (\ref{eq:disp_full_nodim_kr}) so is its complex conjugate, for the
742: dispersion relation has real coefficients (see Appendix A).
743: 
744: In the next section, we will show that the dispersion relation
745: (\ref{eq:disp_full_nodim_kr}) reduces to the  dispersion relations
746: previously derived in many local studies in different regimes.  It is
747: important to emphasize, however, that this dispersion relation fully
748: considers the effects of compressibility and magnetic tension simultaneously
749: without imposing any restrictions on the field strength or geometry.
750: This feature is crucial in determining the stability properties of the MHD
751: flow when strong toroidal fields are considered.
752: 
753: 
754: \subsection{Previous Treatments}
755: \label{sec:previous treatments}
756: There has been some discussion in the past about the importance of the
757: curvature terms for the stability of magnetized Keplerian flows
758: \citep{Knobloch92, GB94}.  In studies in which these terms were
759: considered \citep{Knobloch92, DK93}, compressibility effects were
760: neglected. On the other hand, there have also been treatments in which
761: compressibility was addressed but the curvature terms were neglected
762: \citep{BB94}.  Both types of studies provided arguments for and
763: against the importance of these terms. 
764: The limit of cold MHD flows has been addressed by Kim \& Ostriker (2000).
765: These authors concluded that when the magnetic field strength  is
766: superthermal, the inclusion of toroidal fields tends to suppress the
767: growth of the MRI and that for quasi-toroidal field configurations 
768: no axisymmetric MRI takes place in the limit $c_{\rm s} \rightarrow 0$. 
769: 
770: Because of the generality of
771: our treatment, in which both curvature terms and compressibility
772: effects are fully taken into account, we are able to address all of these
773: issues in \S \ref{sec:discussion}. For the time being, and as a check, 
774: we can take the appropriate limits in the general
775: dispersion relation (\ref{eq:disp_full_nodim_kr}) to recover the
776: dispersion relations derived in the aforementioned works. 
777: 
778: \emph{Compressibility with no field curvature ---\/}
779: Setting $\epsilon_i = 0$, for $i=1,2,3,4$,  
780: and considering perturbations propagating only in the vertical
781: direction (this can be formally done by taking the limit $k_z \gg k_r$
782: in equation
783: [\ref{eq:disp_full_nodim_kr}]) we recover the dispersion relation derived 
784: in the compressible, weak-field limit by  \citet{BB94},
785: \begin{eqnarray}
786: \label{eq:disp_blaes}
787: && \omega^6 - [k_z^2(c_{\rm s}^2+v_{{\rm A}\phi}^2+2v_{{\rm A}z}^2) +
788: \kappa^2] \omega^4 \nonumber \\ &+& k_z^2\left[ k_z^2 v_{{\rm A}z}^2 (2c_{\rm
789: s}^2 + v_{{\rm A}\phi}^2 + v_{{\rm A}z}^2) + \kappa^2(c_{\rm
790: s}^2 + v_{{\rm A}\phi}^2) + 2 \frac{d\ln \Omega}{d \ln r} v_{{\rm
791: A}z}^2 \right]  \omega^2 \nonumber \\ &-& k_z^4 v_{{\rm A}z}^2
792: c_{\rm s}^2 \left( k_z^2 v_{{\rm A}z}^2 + 2 \frac{d \ln \Omega}{d \ln r}
793: \right) = 0 ~.
794: \end{eqnarray}
795: The stability criterion derived from this dispersion relation is not
796: different from the one derived, within the Boussinesq approximation, 
797: by \citet{BH91}. All the perturbations with vertical wavenumber smaller than 
798: the critical wavenumber $k_{\rm BH}$ are unstable, with 
799: \begin{eqnarray}
800: \label{eq:k_BH}
801: k_{BH}^2 v_{{\rm A}z}^2 \equiv  - 2 \frac{d \ln \Omega}{d \ln r} ~.
802: \end{eqnarray}
803: In this case, the strength of the toroidal component of the magnetic
804: field does not play any role in deciding which modes are subject to
805: instabilities.
806: 
807: \emph{Field curvature with no compressibility ---\/}
808: It is important to stress that even in the incompressible limit
809: not all the terms proportional to $\epsilon_i$ in the dispersion relation
810: (\ref{eq:disp_full_nodim_kr}) are negligible (of course, the ones
811: proportional to $\epsilon_2$ are).  
812: To see that this is the case, we can take the limit 
813: $c_{\rm s} \rightarrow \infty$ in  the dispersion relation 
814: (\ref{eq:disp_full_nodim_kr}) to obtain
815: \begin{eqnarray}
816: \label{eq:disp_incomp_nodim}
817: (k_z^2+k_r^2- ik_r \epsilon_4)  \omega^4 
818: -  k_z^2  [ 2 v_{{\rm A}z}^2 (k_z^2+k_r^2-\epsilon_4ik_r) 
819: +  \kappa^2  + 2 \epsilon_1 \epsilon_4  v_{{\rm A}\phi}^2 ] \omega^2
820: - 2 k_z^3 v_{{\rm A}\phi} v_{{\rm A}z} (2 \epsilon_1 + \epsilon_3 +
821: \epsilon_4) \omega   \nonumber  \\  
822: + k_z^4 v_{{\rm A}z}^2 \left[
823:   (k_z^2+k_r^2-\epsilon_4 ik_r) v_{{\rm A}z}^2 + 2  \frac{d \ln \Omega}{d \ln r} -
824:   2 \epsilon_1 \epsilon_3 v_{{\rm A}\phi}^2 \right] =0 ~,   %\nonumber  \\  
825: \end{eqnarray} 
826: where  we have explicitly left the factors $\epsilon_i$ that should be
827: considered as unity. This incompressible version of our dispersion
828: relation is to be compared with the one obtained by \citet{DK93} as
829: the local limit of the corresponding eigenvalue problem.  Note that, 
830: in order to compare expression (\ref{eq:disp_incomp_nodim}) with the dispersion
831: relation (eq.~[37]) presented in \citet{DK93}, it is necessary to 
832: considered the limit $\partial v_{{\rm A}\phi}/\partial r,  \partial v_{{\rm A}
833: z}/\partial r \rightarrow 0$ in their equation (9).  We also note that
834: the radial wavenumber $k_r$ appears in equation
835: (\ref{eq:disp_incomp_nodim}) only in the combination
836: $k_z^2+k_r^2-ik_r$ while in equation (37) in \citet{DK93} we only find
837: it as  $k_z^2+k_r^2$ (i.e., $n^2+k^2$ in their notation). This is
838: because when taking the local limit, $k_r \gg 1 $, in the process of
839: deriving their equation (37) from  their equation (9),  the terms
840: proportional to $ik_r$ were neglected against $k_r^2$ by \citet{DK93}.
841: 
842: When the toroidal component of the magnetic field is negligible, i.e.,
843: when $v_{{\rm A} \phi} \rightarrow 0$ in equation
844: (\ref{eq:disp_incomp_nodim}), and we consider vertical modes ($k_z \gg
845: k_r$), we recover the dispersion relation for the incompressible MRI;
846: the onset of unstable modes is still given by expression
847: (\ref{eq:k_BH}).  For weak toroidal fields, i.e., when $v_{{\rm A} \phi}
848: \ll 1$, we can read off the small corrections to the critical
849: wavenumber from the constant coefficient,
850: \begin{eqnarray}
851: \label{eq:k_c}
852: (k_z^{0i})^2 v_{{\rm A}z}^2 =  - 2 \frac{d \ln \Omega}{d \ln r} + 2
853: \epsilon_1 \epsilon_2 v^2_{A \phi}  ~.
854: \end{eqnarray}
855: For stronger fields, however, the $\omega = 0$ mode is no longer
856: unstable (see the Appendix B for a general discussion on the
857: stability of the $\omega=0$ mode when compressibility and curvature
858: terms are considered) and it is necessary to solve equation
859: (\ref{eq:disp_incomp_nodim}) in order to find the critical wavenumber
860: for the onset of the instability. Roughly speaking, we would expect
861: the solutions of equation (\ref{eq:disp_incomp_nodim}) to depart
862: significantly from the solutions to the incompressible version of
863: the dispersion relation (\ref{eq:disp_blaes}) 
864: when $v^2_{{\rm A} \phi} \gtrsim  |d\ln \Omega/d
865: \ln r|$, or $v_{{\rm A} \phi} \gtrsim 1.2$ for a Keplerian disk.
866: Since in this paper we consider rotationally supported configurations (i.e.,
867: $v_{{\rm A} \phi} \lesssim 1 $), we will not address the
868: modifications to the mode structure caused by curvature terms in
869: incompressible MHD flows.
870: 
871: It is important to stress that, for both dispersion relations
872: (\ref{eq:disp_blaes}) and (\ref{eq:disp_incomp_nodim}), in the case of
873: rotationally supported disks, the stability criterion is insensitive
874: (or, at most, very weakly sensitive, in the incompressible case) to the
875: magnitude of the toroidal component of the field. As we will see
876: throughout our study, the stability criteria that emerge from equation
877: (\ref{eq:disp_full_nodim_kr}) are significantly different from the ones
878: discussed in this section, when we consider fields for which
879: $v_{{\rm A} \phi} > c_{\rm s}$. We will also see that the term
880: proportional to $\epsilon_2$, which depends on curvature and
881: compressibility effects and is, therefore, absent from either equation
882: (\ref{eq:disp_blaes}) or (\ref{eq:disp_incomp_nodim}), plays an
883: important role in determining the mode structure in the general case.
884: 
885: \emph{Cold limit with no field curvature ---\/} Another limit of
886: interest is the one corresponding to the cold, MHD, cylindrical
887: shearing flows usually involved in the modeling of cold disk winds
888: (i.e., far away from the disk). In this context, Kim \& Ostriker
889: (2000) addressed the behavior of the compressible axisymmetric MRI in
890: the limit $c_{\rm s} \rightarrow 0$.  These authors obtained a 
891: dispersion relation considering  both vertical and radial wavenumbers
892: and derived the criterion for instability associated with it. Their
893: dispersion relation in the  fully compressible case [eq.~(57)] reads
894: \begin{eqnarray}
895: \label{eq:disp_cseq0}
896: && \omega^6 - [(k_z^2 + k_r^2)(c_{\rm s}^2  + v_{{\rm
897: A}\phi}^2+v_{{\rm A}z}^2) + k_z^2 v_{{\rm A}z}^2 + \kappa^2] \omega^4
898: \nonumber \\  &+& k_z^2 \left[(k_z^2 + k_r^2) v_{{\rm A}z}^2 (2c_{\rm
899: s}^2  + v_{{\rm A}\phi}^2 + v_{{\rm A}z}^2)  + \kappa^2(c_{\rm s}^2  +
900: v_{{\rm A}\phi}^2)  + 2 \frac{d\ln \Omega}{d \ln r} v_{{\rm A}z}^2
901: \right]  \omega^2 \nonumber \\  &-& k_z^4 v_{{\rm A}z}^2 c_{\rm s}^2
902: \left[ (k_z^2+k_r^2)  v_{{\rm A}z}^2 + 2 \frac{d \ln \Omega}{d \ln r}
903: \right] = 0 ~.
904: \end{eqnarray}
905: This dispersion relation can be obtained from equation
906: (\ref{eq:disp_full_nodim_kr}) if we set  $\epsilon_i = 0$, for
907: $i=1,2,3,4$. Note that if we take the limit  $k_z \gg k_r$  in equation
908: (\ref{eq:disp_cseq0}) we recover equation (\ref{eq:disp_blaes}).
909: 
910: For extremely cold flows we can take the limit $c_{\rm s} \rightarrow
911: 0$ in equation (\ref{eq:disp_cseq0}) to obtain
912: \begin{eqnarray}
913: \label{eq:disp_cseq0_cold}
914: \omega^4 - [(k_z^2+k_r^2)(v_{{\rm A}\phi}^2+ v_{{\rm A}z}^2) + 
915:   k_z^2 v_{{\rm A}z}^2  + \kappa^2 ] \omega^2  + k_z^2 v_{{\rm A}z}^2
916:   (k_z^2+k_r^2)(v_{{\rm A}\phi}^2+v_{{\rm A}z}^2)  + \kappa^2 k_z^2
917:   v_{{\rm A}\phi}^2 + 2 k_z^2  v_{{\rm A}z}^2 \frac{d\ln \Omega}{d \ln
918:   r} =0 ~. \nonumber \\
919: \end{eqnarray}
920: On the other hand, taking the limit $c_{\rm s} \rightarrow 0$ in
921: equation (\ref{eq:disp_full_nodim_kr}), we obtain the more general dispersion
922: relation
923: \begin{eqnarray}
924: \label{eq:disp_cold_nodim}
925: \omega^6  &-& \{(k_z^2+k_r^2)(v_{{\rm A} \phi}^2 + v_{{\rm   A}z}^2)
926: - ik_r [(2\epsilon_1 -\epsilon_2) v_{{\rm A} \phi}^2   + \epsilon_4
927: v_{{\rm   A}z}^2] + k_z^2 v_{{\rm  A}z}^2 +  \kappa^2 + \epsilon_2
928: \epsilon_4 v_{{\rm A} \phi}^2 \}  \omega^4 \nonumber \\  &-& \left(
929: 2\epsilon_1 + \epsilon_3 \right) 2 k_z v_{{\rm A}\phi} v_{{\rm A}z}
930: \omega^3 + \bigg \lbrace k_z^2 v_{{\rm A}z}^2 [
931: (k_z^2+k_r^2-\epsilon_4ik_r)   ( v_{{\rm A}\phi}^2 + v_{{\rm A}z}^2) +
932: ik_r (\epsilon_2-\epsilon_3) v_{{\rm A}\phi}^2 ]  \nonumber \\  &+&
933: k_z^2\left[\kappa^2 v_{{\rm A}\phi}^2   +  \epsilon_2 \epsilon_4
934: v_{{\rm A}\phi}^4  + 2 \frac{d \ln \Omega}{d \ln r} v_{{\rm A}z}^2 +
935: (\epsilon_2 \epsilon_4   -  2 \epsilon_1 \epsilon_3) v_{{\rm A}\phi}^2
936: v_{{\rm A}z}^2 \right]  \bigg \rbrace \omega^2 \nonumber \\ &+& 2
937: k_z^3 v_{{\rm A}\phi} v_{{\rm A}z} ( \epsilon_2 + \epsilon_3 ) v_{{\rm
938: A}\phi}^2 \omega + k_z^4 v_{{\rm A}z}^2 \epsilon_2 \epsilon_3 v_{{\rm
939: A}\phi}^4 =0 ~.
940: \end{eqnarray} 
941: Note that in this expression, as it was also the case in the
942: incompressible limit, several of the terms that are due to the finite
943: curvature of the toroidal field lines are still present.
944: 
945: Analyzing the limit  $c_{\rm s} \rightarrow 0$  in the  dispersion
946: relation (\ref{eq:disp_cseq0}),  Kim \& Ostriker (2000) concluded
947: that toroidal fields tend to suppress the growth of the MRI and that,
948: for a Keplerian rotation law,  no axisymmetric MRI occurs if
949: $i<30\degr$, where $i$ is the local pitch angle of the magnetic fields
950: defined by $i \equiv \tan^{-1} (v_{{\rm A} z}/v_{{\rm A} \phi})$.  
951: However, the eigenfrequencies satisfying the dispersion relations 
952: (\ref{eq:disp_full_nodim_kr}) and (\ref{eq:disp_cseq0}) 
953: in the limit $c_{\rm s} \rightarrow 0$  
954: are different and so are the criteria
955: for instability  which they are subject to. In \S
956: \ref{subsec:Comparison to previous analytical studies}, we comment in
957: more detail on how the solutions to the dispersion relations
958: (\ref{eq:disp_full_nodim_kr}) and (\ref{eq:disp_cseq0}) 
959: differ in the limit  $c_{\rm s} \rightarrow 0$ and on the 
960: implications  regarding the stabilization of the MRI 
961: in cold MHD shearing flows.
962: 
963: In order to investigate how previous results from local stability
964: analyses of the weak field MRI are modified as the strength of the
965: toroidal field component increases, we will focus our attention on the
966: stability of modes with $k_z \gg k_r$.\footnote{In Appendix A, we
967: briefly describe how this results are modified when finite ratios
968: $k_r/k_z$ are considered.}  This approach is physically motivated,
969: since vertical modes correspond to the most unstable modes in the well
970: studied MRI, and is also more tractable mathematically.  In the next
971: two sections, we will perform a thorough numerical and semi-analytical study
972: of the general dispersion relation (\ref{eq:disp_full_nodim_kr}) in
973: the limit  $k_z\gg k_r$, with particular emphasis on the
974: case of strong toroidal fields.    We will then be in a better
975: position to understand  the similitudes and differences of our
976: findings  with those of the aforementioned studies and  we will
977: address them in \S\ref{sec:discussion}.
978: 
979: \section{\textsc{NUMERICAL SOLUTIONS}}
980: \label{sec:numerical solutions}
981: 
982: 
983: \begin{figure}[tbh]
984: \begin{center}
985: \includegraphics[width=\textwidth]{f1.eps}
986: \caption{The real parts of the numerical solutions to the dispersion
987:   relation (\ref{eq:disp_full_nodim_kr}) corresponding to a Keplerian
988:   disk with $c_{\rm s}= 0.05$ and $v_{{\rm A}z}=0.01$.
989:   \emph{Left panel}: solutions to the full problem
990:   ($\epsilon_1=\epsilon_2=\epsilon_3=\epsilon_4=1$). \emph{Central
991:   panel}: the case in which compressibility is neglected in the
992:   curvature terms ($\epsilon_1=\epsilon_3=\epsilon_4=1$ and
993:   $\epsilon_2=0$). \emph{Right panel}: the case in which all curvature
994:   terms are neglected
995:   ($\epsilon_1=\epsilon_2=\epsilon_3=\epsilon_4=0$). 
996:   Open circles
997:   indicate unstable modes (i.e., those with positive imaginary part).
998:   Long-dashed, short-dashed, and point-dashed lines show the fast,
999:   \Alfvennospace, and slow modes, respectively, in the limit of no
1000:   rotation.}
1001: \label{fig:comp_real} 
1002: \end{center}
1003: \end{figure}
1004: 
1005: \begin{figure}[tbh]
1006: \begin{center}
1007: \includegraphics[width=\textwidth]{f2.eps}
1008: \caption{The imaginary parts for the cases discussed in Figure
1009: \ref{fig:comp_real}.}
1010: \label{fig:comp_imag} 
1011: \end{center}
1012: \end{figure}
1013: 
1014: We solved numerically the dispersion relation
1015: (\ref{eq:disp_full_nodim_kr})   for the frequency $\omega$ as a
1016: function of the wavenumber $k_z$, employing Laguerre's root finding
1017: method (Press et al.~1992).   As a typical situation of interest, we
1018: consider a Keplerian disk with  $c_{\rm s}= 0.05$ and $v_{{\rm
1019: A}z}=0.01$.  As it will be seen from the range of values of $k_z$ in
1020: which the various instabilities occur, the case of quasi-toroidal
1021: superthermal fields is perfectly suited to be studied in the local
1022: approximation, i.e., when $k_z \gg 1$, provided that the vertical component of
1023: the magnetic field  is weak enough (i.e., $v_{{\rm A}z} \ll 1$).
1024: 
1025: To better appreciate the effects that the curvature terms have on the
1026: stability of the modes, a set of solutions to the dispersion relation
1027: (\ref{eq:disp_full_nodim_kr}) is shown in Figures \ref{fig:comp_real}
1028: and \ref{fig:comp_imag}.\footnote{Some animations of the results
1029: presented in Figs. \ref{fig:comp_real}, \ref{fig:comp_imag},
1030: \ref{fig:dlog}, and \ref{fig:sound} are available at
1031: \texttt{http://www.physics.arizona.edu/$\sim$mpessah/research/}}  Each
1032: of the three panels, in both figures, shows the real and imaginary
1033: parts of the solutions for different values of the toroidal field
1034: strength, parameterized by $v_{{\rm A}\phi}$.  The left panel shows
1035: the solutions to the full dispersion relation
1036: (\ref{eq:disp_full_nodim_kr}), i.e., when
1037: $\epsilon_1=\epsilon_2=\epsilon_3=\epsilon_4=1$. The central panel
1038: shows the solutions to equation (\ref{eq:disp_full_nodim_kr}) when
1039: compressibility is neglected in the curvature terms, i.e., when
1040: $\epsilon_1=\epsilon_3=\epsilon_4=1$  and $\epsilon_2=0$.  For the
1041: sake of comparison, the right panel shows the solutions to the
1042: dispersion relation (\ref{eq:disp_blaes}), in which all curvature
1043: terms are neglected.
1044: 
1045: We first analyze Figure \ref{fig:comp_real}.  When all magnetic
1046: tension terms are neglected (right panel), the qualitative structure
1047: of the normal modes of the plasma is insensitive to the magnitude of
1048: the toroidal field component (see Blaes \& Balbus 1994). However, the
1049: situation is very different when the magnetic tension terms are
1050: included. For weak toroidal fields, i.e., when  $v_{{\rm A}\phi} \lesssim
1051: 0.1$, the solutions seem quite insensitive to the curvature terms;
1052: indeed these terms do not seem to play a significant role in altering
1053: the local stability properties of magnetized Keplerian flows compared
1054: to what is quoted elsewhere in the literature.  As we will see
1055: later, for a Keplerian disk, the presence of the curvature terms is
1056: significant once $v_{{\rm A}\phi}^2 \gtrsim c_{\rm s}$, which in this
1057: case translates into $v_{{\rm A}\phi} \gtrsim 0.22$.
1058: 
1059: For stronger toroidal fields, i.e., when $v_{{\rm A}\phi} \gtrsim
1060: 0.2$, the modes with the longest wavelengths become stable when all
1061: curvature terms are included, in sharp contrast to the case in which
1062: $\epsilon_2=0$. For even stronger toroidal fields, i.e., when $v_{{\rm
1063:     A}\phi} \gtrsim 0.3$, a second instability appears at long
1064: wavelengths, while the original instability is suppressed.  When
1065: $v_{{\rm A}\phi} \gtrsim 0.4$, both instabilities coexist as separate
1066: entities and the original instability reaches smaller and smaller
1067: spatial scales, when the magnitude of the toroidal field
1068: increases. For even higher toroidal fields, i.e., when  $v_{{\rm
1069:     A}\phi} \gtrsim 0.7$, the largest unstable wavenumber of the
1070: instability that developed for $v_{{\rm A}\phi} \gtrsim 0.3$
1071: approaches $k_{\rm BH}$ (see eq.~[\ref{eq:k_BH}]).  The major
1072: implication of neglecting compressibility in the curvature  terms is
1073: that the original instability seems to be totally suppressed  for
1074: toroidal fields larger than the ones corresponding to $v_{{\rm A}\phi}
1075: \gtrsim 0.3$.
1076: 
1077: As it is clear from the dispersion relation
1078: (\ref{eq:disp_full_nodim_kr}),  the presence of the toroidal component
1079: in the field introduces odd powers of the mode frequency $\omega$ and
1080: hence breaks the symmetry between positive and negative real parts of
1081: the solutions. The physical meaning of this is clear.  The phase
1082: velocities of the instabilities are no longer zero and they are
1083: propagating vertically throughout the disk. This, of course, is not
1084: the case for the unstable solutions to the dispersion relation 
1085: (\ref{eq:disp_blaes})
1086: regardless of the magnitude of $v_{{\rm A}\phi}$. In that case, the
1087: most noticeable effect of an increasing toroidal field is to reduce
1088: the phase velocity of the stable modes beyond $k_{\rm BH}$ (which is
1089: itself independent of $v_{{\rm A}\phi}$).
1090: 
1091: It is also interesting to analyze how the presence of the curvature
1092: terms modifies the growth rates of the unstable modes as a function of
1093: the toroidal magnetic field. This is shown in Figure
1094: \ref{fig:comp_imag}. Again, there are no significant changes for
1095: $v_{{\rm A}\phi} \lesssim 0.1$; however, quite significant
1096: modifications  to the growth rates are present for $v_{{\rm A}\phi}
1097: \gtrsim 0.2$.   The sequence of plots in the left panel shows more
1098: clearly the suppression of the original instability, the appearance of
1099: the instability at low wavenumbers, the return of the instability at
1100: high wavenumbers, and finally the fusion of these last two.  
1101: The right panel in this figure shows the effects that the presence of a strong
1102: toroidal component has on the mode structure when curvature terms are
1103: not considered. In this case, the critical wavenumber for the onset of
1104: instabilities is not modified while there is a clear 
1105: reduction in the growth rate of the non-propagating unstable modes as
1106: the magnitude of the toroidal field component increases.
1107: When the curvature terms are
1108: considered fully, the effects are more dramatic.  Note also that the
1109: growth rate of the original instability is reduced faster from the
1110: first to the second plot in the left panel in Figure
1111: \ref{fig:comp_imag} with respect to their counterparts in the right
1112: panel of the same figure.
1113: 
1114: 
1115: \section{\textsc{THE ONSET OF INSTABILITIES}}
1116: \label{sec:onset of instabilities}
1117: 
1118: \subsection{Unstable Modes}
1119: \label{subsec:unstable modes}
1120: 
1121: In \S \ref{sec:numerical solutions} we presented how the structure of
1122: the various modes evolves as a function of the toroidal field strength
1123: and noted that, for a range of field strengths,  two different
1124: instabilities are clearly  distinguishable.  Here, we obtain the
1125: conditions (i.e., the range of wavenumbers and toroidal field
1126: strengths) for which these unstable modes are present. We start by plotting
1127: in Figure \ref{fig:unstable_regions_num} the range of unstable
1128: wavenumbers as a function of the toroidal field strength. As a
1129: reference, we have plotted the case for a Keplerian disk.  The black
1130: dots in the diagram represent the unstable vertical wavenumbers, in
1131: units of  $k_z v_{{\rm A} z}/c_{\rm s}$, for a given toroidal \Alfven
1132: speed, in terms of $v_{{\rm A}\phi}/c_{\rm s}$.  Three regions of
1133: unstable modes are clearly distinguishable:
1134: \begin{itemize}
1135: 
1136: \item \emph{Region I\/} shows the evolution of the original
1137: instability present in the topmost three plots in the left panel in
1138: Figure \ref{fig:comp_real}.  This is the region where the MRI
1139: lives. Strictly speaking, the MRI is confined to the region where
1140: $v_{{\rm A} \phi}/c_{\rm s} \ll 1$. As we will comment in \S
1141: \ref{subsec:analytic approximations}, instability I is no longer
1142: incompressible beyond this point.  The maximum wavenumber for which
1143: this instability exists is independent of $v_{{\rm A}\phi}$ and
1144: corresponds to the critical wavenumber for the onset of the MRI (i.e.,
1145: $k_{\rm BH}$ in eq.~[\ref{eq:k_BH}]).   The stabilization of the
1146: long-wavelength perturbations beyond a critical value of the toroidal
1147: \Alfven speed is also evident in this region. For larger toroidal
1148: field strengths, shorter and shorter wavelengths are stabilized up to the ones
1149: corresponding to  $k_{\rm BH}$.
1150: 
1151: \item  \emph{Region II\/} represents   the evolution of the
1152: instability that is only present for wavenumbers $k_z > k_{ \rm
1153: BH}$. Note that $k_{\rm BH}$ is  now the minimum wavenumber for the
1154: onset of instability II. In this case, increasing $v_{{\rm A}
1155: \phi}/c_{\rm s}$ gives rise to unstable modes with even shorter
1156: wavelengths (two bottommost plots in the left panel of either Figure
1157: \ref{fig:comp_real} or \ref{fig:comp_imag}).
1158: 
1159: \item \emph{Region III\/} shows the instability  that appears for
1160: intermediate wavenumbers (see for example the third plot in the left
1161: panel in Figure \ref{fig:comp_real}).  Note that the shortest unstable
1162: wavelength in this region approaches $k_{\rm BH}$ for large values of
1163: $v_{{\rm A} \phi}/c_{\rm s}$ (i.e., bottommost plot in the left panel
1164: of either Figure \ref{fig:comp_real} or \ref{fig:comp_imag}).
1165: \end{itemize}
1166: 
1167: 
1168: 
1169: \begin{figure}[!tbh]
1170: \begin{center}
1171: \includegraphics[width=.75\textwidth]{f3.eps}
1172: \caption{The black dots represent unstable modes obtained from solving the
1173:         dispersion relation (\ref{eq:disp_full_nodim_kr}) numerically for
1174:         a Keplerian disk with $c_{\rm s}=0.05$ and $v_{{\rm A} z}=
1175:         0.01$. The solid ($k_z^0$) and dashed ($k_z^{0i}$)  lines correspond
1176:         to the critical wavenumber for which the
1177:         $\omega = 0$ mode exists in the case of a compressible (see
1178:         Appendix B) and an incompressible (discussed in \S
1179:         \ref{sec:mhd equations and dispersion relation}) flow,
1180:         respectively. For strong toroidal fields, compressibility
1181:         plays a crucial role in the stability of the $\omega = 0$
1182:         mode. Note that, in the limit of small $v_{{\rm A}\phi}/c_{\rm
1183:         s}$ we have $k_z^0, k_z^{0i} \rightarrow k_{\rm BH}$, and the trivial
1184:         mode becomes unstable.}
1185: \label{fig:unstable_regions_num}
1186: \end{center}
1187: \end{figure}
1188: 
1189: 
1190: 
1191: 
1192: \subsection{Analytic Approximations}
1193: \label{subsec:analytic approximations}
1194: 
1195: In this section we obtain analytical approximations to  the dispersion
1196: relation (\ref{eq:disp_full_nodim_kr}) in various limits,  which will
1197: help us identify the different critical curves in  Figure
1198: \ref{fig:unstable_regions_num}.
1199: 
1200: The fast (or magnetosonic) modes are reasonably
1201: well decoupled from the rest of the oscillations (see left panels in
1202: Fig.\ref{fig:comp_real}). By studying the modes that satisfy the
1203: condition $\omega^2 \ll k_z^2 c^2_{\rm s}$, we effectively eliminate
1204: the fast modes from our analysis. This can be done for strong toroidal
1205: fields because, even in the presence of rotation, the magnetosonic
1206: modes are well described by  $\omega^2 \simeq k_z^2 (c^2_{\rm s} +
1207: v^2_{{\rm A}})$. Note that imposing $\omega^2 \ll k_z^2 c^2_{\rm s}$  is
1208: a distinct and weaker condition than asking for the MHD fluid to be
1209: incompressible ($c_{\rm s} \rightarrow \infty$).  By eliminating these
1210: fastest modes, it is possible to find a $4^{\rm th}$ degree dispersion
1211: relation in $\omega$, with solutions that constitute a very good
1212: approximation to the interesting modes seen in Figures
1213: \ref{fig:comp_real} and \ref{fig:comp_imag}.
1214: 
1215: We first write the equations for the evolution of the perturbations in the
1216: magnetic field. For the sake of clarity, we present the intermediate
1217: steps with the appropriate physical dimensions but we drop the 
1218: index indicating local values. Substituting equations
1219: (\ref{eq:pert_cont_grads}),  (\ref{eq:pert_induc_r_grads}), 
1220: and (\ref{eq:pert_induc_z_grads}) 
1221: in equation (\ref{eq:pert_euler_z_grads}) we obtain $\delta v_z$ 
1222: in terms of $\delta B_\phi$ and $\delta B_z$,
1223: \begin{equation}
1224: \label{eq:vz_Bphi}
1225: \delta v_z = - \frac{k_z \omega c_{\rm s}^2  } {(k_zc_{\rm s})^2 -
1226: \omega^2} \left[ \frac{v_{{\rm A} \phi}^2}{c_{\rm s}^2}\frac{\delta
1227: B_\phi}{ B_\phi} +  \frac{\delta B_z}{B_z} \right] ~.
1228: \end{equation}
1229: Using this result in equation (\ref{eq:pert_induc_phi_grads}) we find,
1230: \begin{equation}
1231: \label{eq:vphi_BrBphi}
1232: i k_z B_z \delta v_\phi = - \frac{d\Omega}{d \ln r} \delta B_r  -i\omega
1233: \delta B_\phi - i \omega B_\phi \frac{(k_zc_{\rm s})^2 } {(k_zc_{\rm s})^2
1234:   - \omega^2} \left[ \frac{v_{{\rm A} \phi}^2}{c_{\rm
1235:       s}^2}\frac{\delta B_\phi}{B_\phi} +  \frac{\delta B_z}{ B_z} \right] ~.
1236: \end{equation}
1237: From  equations (\ref{eq:pert_cont_grads}), (\ref{eq:pert_induc_z_grads}), and
1238: (\ref{eq:vz_Bphi}) we can recast $ \delta \rho$ in terms of $\delta
1239: B_\phi$ and  $\delta B_z$ as
1240: \begin{equation}
1241: \label{eq:rho_Bphi}
1242: \frac{\delta \rho}{\rho} =   - \frac{k_z^2 v_{{\rm A} \phi}^2}{(k_zc_{\rm
1243:     s})^2 - \omega^2} \frac{\delta B_\phi}{ B_\phi} +  \left[ 1 -
1244:   \frac{(k_zc_{\rm s})^2} {(k_zc_{\rm s})^2 - \omega^2} \right]
1245: \frac{\delta B_z}{ B_z} ~.
1246: \end{equation}
1247: 
1248: Finally, we can write equations (\ref{eq:vz_Bphi})-(\ref{eq:rho_Bphi}) for the
1249: modes with frequencies such that $\omega^2 \ll k_z^2 c^2_{\rm s}$ as,
1250: \begin{equation}
1251: \label{eq:vz_Bphi_limit}
1252: \delta v_z = - \frac{\omega}{k_z}  \left[ \left(\frac{v_{{\rm
1253: A}\phi}}{c_{\rm s}}\right)^2 \frac{\delta B_\phi}{B_\phi} +
1254: i\frac{\epsilon_4}{k_z r}\frac{\delta B_r}{B_z} \right] ~,
1255: \end{equation}
1256: \begin{equation}
1257: \label{eq:vphi_BrBphi_limit}
1258: i k_z B_z \delta v_\phi = - \left[\frac{d\Omega}{d \ln r}  -
1259:   \frac{\epsilon_4}{r} \frac{\omega}{k_z}\frac{B_\phi}{B_z}
1260:   \right]\delta B_r - \left[ 1+  \left( \frac{v_{{\rm A}\phi}}{c_{\rm
1261:       s}}\right)^2 \right] i \omega \delta B_\phi ~,
1262: \end{equation}
1263: and
1264: \begin{equation}
1265: \label{eq:rho_Bphi_limit}
1266: \frac{\delta \rho}{\rho} = -  \frac{v^2_{A \phi}}{c^2_s} \frac{\delta
1267:  B_\phi}{B_\phi} ~,
1268: \end{equation}
1269: where we have used equations (\ref{eq:pert_induc_r_grads}) and
1270: (\ref{eq:pert_induc_z_grads}) to recast $\delta B_z$ in terms of
1271: $\delta B_r$. Note that, neglecting the factor  $\omega^2$ against
1272: $k_z^2 c^2_{\rm s}$ in equation (\ref{eq:vz_Bphi}), and
1273: therefore in equations (\ref{eq:vphi_BrBphi}) and (\ref{eq:rho_Bphi}),
1274: effectively reduces to neglecting the term proportional to $\omega$ in
1275: equation (\ref{eq:pert_euler_z_grads}). Thus, for the modes of
1276: interest, the condition  $\omega^2 \ll k_z^2 c^2_{\rm s}$ is a
1277: statement about force balance in the vertical direction, which is made
1278: explicit in equation  (\ref{eq:rho_Bphi_limit}). In this way,  we can
1279: see how important perturbations in the density are,  in the presence
1280: of strong toroidal fields (see also Balbus \& Hawley 1991).  For
1281: $v_{{\rm A} \phi} \gg c_{\rm s}$,  even small variation in the
1282: toroidal component of the field can have an important impact on the
1283: dynamics of the perturbations.  For this reason, the assumption of an
1284: incompressible MHD flow is not valid, whenever superthermal toroidal
1285: fields are considered. Note that, in order to recover the
1286: incompressible MRI when  $\epsilon_i = 0$, for $i=1,2,3,4$, we have
1287: not neglected the factor unity against $(v_{{\rm A} \phi}/ c_{\rm
1288: s})^2$, in equation (\ref{eq:vphi_BrBphi_limit}).
1289: 
1290: 
1291: \begin{figure}[tbh]
1292: \begin{center}
1293: \includegraphics[width=\textwidth]{f4a.eps}
1294: \includegraphics[width=\textwidth]{f4b.eps}
1295: \caption{\emph{Left panels}: Solutions to the full dispersion relation
1296:   (\ref{eq:disp_full_nodim_kr}). \emph{Central panels}: Solutions to
1297:   the $4^{\rm th}$ order, approximate dispersion relation
1298:   (\ref{eq:disp_large_alpha}), with
1299:   $\epsilon_2=\epsilon_3=\epsilon_4=1$.  \emph{Right panels}:
1300:   Solutions to the $2^{\rm nd}$ order, approximate dispersion relation
1301:   (\ref{eq:disp_2nd}), with $\epsilon_2=\epsilon_3=1$ and
1302:   $\epsilon_4=0$. All solutions correspond to a Keplerian disk with
1303:   $c_{\rm s}= 0.05$, $v_{{\rm A}z}=0.01$, and $v_{{\rm A}
1304:   \phi}=0.40$.  Open circles in upper panels indicate unstable
1305:   modes. Note that the phase velocities of the two instabilities (seen
1306:   in either the leftmost or central upper panels and corresponding to
1307:   Region II and III in \S \ref{subsec:unstable modes}) are similar to
1308:   the phase velocities, positive and negative respectively, of the
1309:   slow mode (point-dashed line) in the limit of no rotation. The fast
1310:   magnetosonic modes can barely be seen close to the left axis in the
1311:   upper left panel.}
1312: \label{fig:approx}
1313: \end{center}
1314: \end{figure}
1315: 
1316: 
1317: We now have all the elements to write equations
1318: (\ref{eq:pert_euler_r_grads}) and (\ref{eq:pert_euler_phi_grads}) in
1319: terms of $\delta B_r$ and $\delta B_\phi$. Using equations
1320: (\ref{eq:vz_Bphi_limit})-(\ref{eq:rho_Bphi_limit}), valid in the limit
1321: $\omega^2 \ll k_z^2 c^2_{\rm s}$, we obtain, in terms of the
1322: dimensionless variables,
1323: \begin{eqnarray}
1324: \label{eq:motion_Br}
1325: - \omega^2 \delta B_r + 2i\omega \left[ 1 + \left(\frac{v_{{\rm A}
1326:   \phi}}{c_{\rm s}} \right)^2 \right] \delta B_\phi = &-&  \left[
1327:   2\frac{d \ln \Omega}{d \ln r} + (k_z v_{{\rm A} z})^2   -
1328:   2\epsilon_4 \frac{\omega}{k_z}\frac{v_{{\rm A} \phi}}{v_{{\rm A}
1329:   z}}\right] \delta B_r  \nonumber \\ &-& i k_z v_{{\rm A} \phi}
1330:   v_{{\rm A}z}    \left[ 2\epsilon_1 + \epsilon_2 \left(\frac{v_{{\rm
1331:   A} \phi}}{c_{\rm s}} \right)^2 \right] \delta B_\phi ~,
1332: \end{eqnarray}
1333: \begin{equation}
1334: \label{eq:motion_Bphi}
1335:  - \omega^2 \delta B_\phi \left[ 1 + \left(\frac{v_{{\rm
1336:  A}\phi}}{c_{\rm s}}\right)^2 \right]  - i \omega \left[2 + \epsilon_4
1337:  \frac{\omega}{k_z}\frac{v_{{\rm A} \phi}}{v_{{\rm A} z}} \right]\delta
1338:  B_r = - (k_z v_{{\rm A} z})^2 \delta B_\phi + i k_z v_{{\rm A} \phi}
1339:  v_{{\rm A}z} \epsilon_3 \delta B_r ~.
1340: \end{equation}
1341: These equations are the generalization of the set of equations used to
1342: illustrate the physics behind the weak-field 
1343: MRI as a system of masses coupled by
1344: a spring in a differentially rotating background. Indeed, in the
1345: incompressible limit and neglecting the curvature terms proportional
1346: to $\epsilon_i$, for $i=1,2,3,4$,
1347: we recover the set of equations presented elsewhere
1348: (Balbus \& Hawley 1992, 1998).
1349: 
1350: 
1351: 
1352: \begin{figure}[tbh]
1353: \begin{center}
1354: \includegraphics[width=.75\textwidth]{f5.eps}
1355: \caption{The black dots represent unstable modes satisfying the
1356:         approximate instability criteria (\ref{eq:D_4}), described in
1357:         \S \ref{subsec:analytic approximations}.  The dashed lines,
1358:         labeled by $k_z^{{\rm c} 1}$ and $k_z^{{\rm c} 2}$, 
1359: 	are the limits of Regions I
1360:         and II obtained analytically, also in \S \ref{subsec:analytic
1361:         approximations}. The onset of instability III is labeled by
1362:         $k_z^{{\rm c} 3}$. As in Figure
1363:         \ref{fig:unstable_regions_num}, we have assumed a
1364:         Keplerian disk with $c_{\rm s}=0.05$ and $v_{{\rm A}z}= 0.01$.}
1365: \label{fig:unstable_regions_anal}
1366: \end{center}
1367: \end{figure}
1368: 
1369: 
1370: Setting the determinant of the linear system
1371: (\ref{eq:motion_Br})-(\ref{eq:motion_Bphi}) equal to zero and taking
1372: the limit $v_{{\rm A} \phi}  \gg  c_{\rm s} $ provides the following
1373: approximate dispersion relation that is valid for strong toroidal
1374: fields\footnote{Note that,  had we taken the opposite limit, i.e.,  $
1375: c_{\rm s} \gg v_{{\rm A} \phi}$,  we would have recovered the
1376: dispersion relation (\ref{eq:disp_incomp_nodim}) in the limit 
1377: $k_r/k_z \rightarrow 0$.},
1378: \begin{equation}
1379: \label{eq:disp_large_alpha}
1380: \omega^4 - (\kappa^2 +  k_z^2 v_{{\rm A} z}^2  +\epsilon_2\epsilon_4
1381:  v_{{\rm A} \phi}^2) \omega^2 - 2 k_z v_{{\rm A} \phi} v_{{\rm A} z}
1382:  (\epsilon_2 + \epsilon_3) \omega  + k_z^2 v_{{\rm A} z}^2 \left[
1383:  \frac{c^2_{\rm s}}{v^2_{{\rm A} \phi}} \left( k_z^2 v_{{\rm A} z}^2 + 2
1384:  \frac{d \ln \Omega}{d \ln r}\right) - \epsilon_2 \epsilon_3 v_{{\rm
1385:  A} \phi}^2 \right]=0 ~.
1386: \end{equation}
1387: Note that we have not neglected the factor  $ c^2_{\rm s}/v^2_{{\rm
1388: A} \phi}$ in the last term in equation (\ref{eq:disp_large_alpha})
1389: because its contribution is non-negligible at large wavenumbers.
1390: The solutions to the dispersion relation (\ref{eq:disp_large_alpha}),
1391: for a Keplerian disk with $c_{\rm s}= 0.05$, $v_{{\rm A}z}=0.01$, 
1392: and $v_{{\rm A} \phi}=0.4$, are shown in the central panels in Figure
1393: \ref{fig:approx}. For the sake of comparison, the left panels in the
1394: same figure show the solutions of the full dispersion relation
1395: (\ref{eq:disp_full_nodim_kr}). The solutions to the approximate
1396: dispersion relation (\ref{eq:disp_large_alpha}) 
1397: are in excellent agreement with the solutions to 
1398: the general dispersion relation (\ref{eq:disp_full_nodim_kr})
1399: for which $\omega^2 \ll k_z^2 c^2_{\rm s}$.
1400: 
1401: 
1402: Note that the term proportional to $\epsilon_1$ is not present in
1403: equation (\ref{eq:disp_large_alpha}).  This feature has important
1404: consequences for us to understand the physics behind the stability of
1405: strongly magnetized compressible flows.  It has been suggested
1406: \citep{CP95} that the magnetic tension term $ B_\phi \delta B_\phi /
1407: r_0$ (i.e., the one proportional to $\epsilon_1$ in
1408: eq.~[\ref{eq:pert_euler_r_grads}]) is responsible for the
1409: stabilization of long-wavelength perturbations via the restoring
1410: forces provided by strong toroidal field lines in  incompressible MHD
1411: flows.  This argument sounds compelling, but  we can see from the last
1412: term in equation (\ref{eq:motion_Br}) that the term proportional to
1413: $\epsilon_1$  is not dynamically important for compressible flows in
1414: which $v_{{\rm A} \phi} \gg c_{\rm s}$.  At least in the radial
1415: direction, it is rather the term proportional to $\epsilon_2$ the one
1416: governing the deviation compared to the stability properties of weak
1417: toroidal fields. This is in complete agreement with equation
1418: (\ref{eq:rho_Bphi_limit}).
1419: 
1420: The dispersion relation (\ref{eq:disp_large_alpha}) is of the form
1421: \begin{equation}
1422: \omega^4 + b_2\omega^2 + b_1 \omega  + b_0 = 0 ~.
1423: \end{equation}
1424: For this $4^{\rm th}$ order equation to have complex roots
1425: (corresponding to unstable modes), its discriminant has to be
1426: negative, i.e., 
1427: \begin{equation}
1428: \label{eq:D_4}
1429: D_4(v_{{\rm A}\phi}, k_z v_{{\rm A} z})=  -4 b_2^3 b_1^2 - 27 b_1^4 + 16
1430: b_0 b_2^4 -128 b_2 b_0^2 + 144 b_2 b_1^2 b_0 + 256 b_0^3 < 0 ~.
1431: \end{equation}
1432: 
1433: The modes satisfying this condition are shown as black dots in Figure
1434: \ref{fig:unstable_regions_anal}. This analytical criterion agrees well
1435: with the numerical results for most of the parameter space
1436: ($v_{{\rm A}\phi}/c_{\rm s}$, $k_z v_{{\rm A} z}/c_{\rm s}$) with the
1437: exception of some of the unstable modes close to the separatrix of the
1438: Regions I and II, defined in \S \ref{subsec:unstable modes}.
1439: 
1440: \emph{Limiting wavenumbers for Regions I and II.---\/} The modes
1441: satisfying the condition $D_4=0$ correspond to the limits of Regions
1442: I, II, and III in Figure \ref{fig:unstable_regions_anal}. Their
1443: analytical expressions, however, are complicated. Yet, some more
1444: progress can be made by realizing that the solutions to the second
1445: order equation obtained by simply dropping the $\omega^4$ term in
1446: equation (\ref{eq:disp_large_alpha}),
1447: \begin{equation}
1448: \label{eq:disp_2nd}
1449: (\kappa^2 +  k_z^2 v_{{\rm A} z}^2 + \epsilon_2\epsilon_4  v_{{\rm A}
1450:  \phi}^2) \omega^2 + 2 k_z v_{{\rm A} \phi} v_{{\rm A} z} (\epsilon_2 +
1451:  \epsilon_3) \omega  - k_z^2 v_{{\rm A} z}^2 \left[ \frac{c^2_{\rm
1452:  s}}{v^2_{{\rm A} \phi}} \left( k_z^2 v_{{\rm A} z}^2 + 2 \frac{d \ln
1453:  \Omega}{d \ln r}\right) - \epsilon_2 \epsilon_3 v_{{\rm A} \phi}^2
1454:  \right]=0 ~, \nonumber \\
1455: \end{equation}
1456: constitute a very good approximation to the solutions of the
1457: dispersion relation (\ref{eq:disp_full_nodim_kr}) whenever the frequencies
1458: of the modes satisfy $\omega^2 \ll 1$.  This can be
1459: appreciated by comparing the left and right panels in Figure
1460: \ref{fig:approx}. The physics behind this approximation is not as
1461: direct as the physics behind the condition $\omega^2 \ll k_z^2 c^2_{\rm
1462:   s}$, but it can also be understood in terms of force
1463: balance, this time in the radial direction. The dispersion relation
1464: (\ref{eq:disp_2nd}) can be obtained by neglecting the term
1465: proportional to $\omega^2$
1466: in equation (\ref{eq:motion_Br}), setting to zero the determinant
1467: of the resulting linear system given by equations 
1468: (\ref{eq:motion_Br})-(\ref{eq:motion_Bphi}) and taking the limit 
1469: $v_{{\rm A}\phi} \gg c_{\rm s}$.
1470: This approximation is equivalent to neglecting the term proportional to
1471: $\omega$ in equation (\ref{eq:pert_euler_r_grads}) and hence related
1472: to neglecting  the radial acceleration experienced by a displaced fluid
1473: element.
1474: 
1475: Setting the discriminant of equation (\ref{eq:disp_2nd}) to zero,
1476: gives an equation in $k_z$ with solutions that are the limiting wavenumbers
1477: for the onset of instabilities I and II in Figure
1478: \ref{fig:unstable_regions_anal}, i.e.,
1479: \begin{eqnarray}
1480: \label{eq:D_2}
1481: D_2(v_{{\rm A}\phi}, k_z v_{{\rm A} z}) =  (k_z v_{{\rm A} z})^4 &+&
1482: \left[ \kappa^2 + 2\frac{d \ln \Omega} {d \ln r} - v_{{\rm A} \phi}^2
1483: \left( \frac{v_{{\rm A} \phi}^2}{c_{\rm s}^2} - \epsilon_4
1484: \frac{c_{\rm s}^2}{v_{{\rm A} \phi}^2} \right) \right] (k_z v_{{\rm A}
1485: z})^2  \nonumber \\ &+& 2\frac{d \ln \Omega} {d \ln r} \left[\kappa^2
1486: - v_{{\rm A} \phi}^2 \left( \frac{v_{{\rm A} \phi}^2}{c_{\rm s}^2} -
1487: \epsilon_4 \frac{c_{\rm s}^2}{v_{{\rm A} \phi}^2} \right) \right] -
1488: \epsilon_4 v_{{\rm A} \phi}^4 = 0 ~.
1489: \end{eqnarray}
1490: Here, we have set $\epsilon_2=\epsilon_3=1$ but have explicitly left
1491: $\epsilon_4$ to show that its contribution to the onset of
1492: instabilities I and II is not important when $v_{{\rm A}\phi} \gg
1493: c_{\rm s}$, as long as we are considering a rotationally supported
1494: disk. We mention, however, that the numerical solutions  show
1495: that the contribution of the term proportional to  $\epsilon_4$ is
1496: small but not negligible for the unstable modes in region III.
1497: Neglecting the terms proportional to $\epsilon_4$, the solutions to
1498: equation (\ref{eq:D_2}) are simply
1499: \begin{equation}
1500: (k_z^{\rm c} v_{{\rm A} z})^2 = \frac{1}{2} \left[ \frac{v_{{\rm A}
1501: \phi}^4}{c_{\rm s}^2} -  \left(\kappa^2 + 2\frac{d \ln \Omega}{d \ln
1502: r}\right) \right]  \pm  \frac{1}{2} \left| \frac{v_{{\rm A}
1503: \phi}^4}{c_{\rm s}^2} - 4 \right| ~.
1504: \end{equation}
1505: One of these solutions coincides always with $k_{\rm BH}$
1506: (eq.~[\ref{eq:k_BH}]),
1507: \begin{equation}
1508: \label{eq:kc1}
1509: (k_z^{{\rm c} 1} v_{{\rm A} z})^2 = - 2\frac{d \ln \Omega}{d \ln r} ~,
1510: \end{equation}
1511: and the other one is
1512: \begin{equation}
1513: \label{eq:kc2}
1514: (k_z^{{\rm c} 2} v_{{\rm A} z})^2 = \frac{v_{{\rm A} \phi}^4}{c_{\rm s}^2} -
1515: \kappa^2 ~.
1516: \end{equation}
1517: 
1518: 
1519: The modes with wavenumbers in the range 
1520: $[min(k_z^{{\rm c} 1},k_z^{{\rm c} 2}),max(k_z^{{\rm c} 1},k_z^{{\rm c} 2})]$ 
1521: are unstable.
1522: In Figure \ref{fig:unstable_regions_anal},
1523: the critical curves $k_z^{{\rm c} 1}(v_{{\rm A} \phi})$ and 
1524: $k_z^{{\rm c} 2}(v_{{\rm A} \phi})$ are shown, with the proper
1525: normalization, as dashed lines.
1526: The critical wavenumber $k_z^{{\rm c} 2}$ in equation (\ref{eq:kc2}) will be
1527: positive only for toroidal \Alfven speeds larger than
1528: \begin{equation}
1529: \label{eq:vphic_limit_I}
1530: v_{{\rm A} \phi}^{{\rm I}} = \sqrt{\kappa  c_{\rm s}} ~.
1531: \end{equation}
1532: This is the critical value of the \Alfven speed beyond which the modes
1533: with longest wavelength in Region I (see
1534: Fig. \ref{fig:unstable_regions_anal}) begin to be stable.  For a
1535: Keplerian disk, the epicyclic frequency coincides with the orbital
1536: frequency and thus, in dimensionless units, $\kappa^2=1$.  In this
1537: case, the critical \Alfven speed for $k_z^{{\rm c} 2}$ to be positive
1538: corresponds to $v_{{\rm A} \phi} = 0.223$. This is the reason for
1539: which the long-wavelength modes are already stable in the second plot
1540: in the left panel in Figure \ref{fig:comp_real}, where $v_{{\rm A}
1541: \phi}=0.25$.
1542: 
1543: Incidentally, we find that the values of toroidal \Alfven speeds for
1544: which the standard MRI gives the appropriate range of unstable modes
1545: are not as restricted to $v_{{\rm A} \phi} \ll c_{\rm s}$ but rather
1546: to $v_{{\rm A} \phi} \ll  \sqrt{\kappa  c_{\rm s}}$.  For  $v_{{\rm A}
1547: \phi} \gtrsim  \sqrt{\kappa  c_{\rm s}}$, the standard MRI is
1548: stabilized at low wavenumbers.   We point out that, Papaloizou \&
1549: Szuszkiewicz (1992) found, by means of a global stability analysis of a
1550: compressible flow, that for a  slim disk threaded only by a
1551: vertical field, the flow is stable if the vertical \Alfven speed
1552: exceeds, within a factor of order unity, the geometrical mean of the
1553: sound speed and the rotational speed. In dimensionless units, this
1554: stability criterion translates into $v_{{\rm A} z} \gtrsim
1555: \sqrt{c_{\rm s}}$.
1556: 
1557: The limiting case in which $k_z^{{\rm c} 1}=k_z^{{\rm c} 2}$, is reached for 
1558: \begin{equation}
1559: \label{eq:vphic_limit_II}
1560: v^{{\rm II}}_{{\rm A} \phi} = \sqrt{2 c_{\rm s}} ~. 
1561: \end{equation}
1562: Note that, for $c_{\rm s} =0.05$, this corresponds to a
1563: value for the critical toroidal \Alfven speed of $v_{{\rm A} \phi} =
1564: 0.316$. This situation is to be compared with the mode structure in
1565: the third plot in the left panel in Figure \ref{fig:comp_real}, where
1566: $v_{{\rm A} \phi} = 0.32$. 
1567: 
1568: \emph{Limiting wavenumbers for Region III.---\/} In the previous section
1569: we presented some useful analytical approximations to describe the
1570: dependence of the critical values of the toroidal \Alfven speeds and
1571: wavenumbers defining Regions I and II on the different quantities
1572: characterizing the MHD flow.  We could not, however, find simple 
1573: analytical expressions to describe satisfactorily the corresponding
1574: behavior of the critical values defining Region III. 
1575: We will describe next how the different unstable regions 
1576: in Figure \ref{fig:unstable_regions_num} depend on the
1577: magnitude of the sound speed and the steepness of the rotation
1578: profile.
1579: 
1580: \section{\textsc{DISCUSSION}}
1581: \label{sec:discussion}
1582: 
1583: In this section, we address several issues related to the importance
1584: of the curvature terms in determining the stability criteria  obeyed
1585: by the solutions to the dispersion relation
1586: (\ref{eq:disp_full_nodim_kr}).  We comment on some controversies raised
1587: by previous investigations that have treated the standard MRI taking
1588: into account, in various ways, either compressibility or curvature of
1589: the background magnetic field. We also comment on the 
1590: importance in the outcome of the instabilities played  by the magnetic
1591: tension produced by toroidal filed lines in the limit of cold MHD
1592: flows. We highlight the similarities and differences of our findings
1593: with the results of \citet{CP95}, who also found the emergence of a
1594: new (but different) instability for strong toroidal fields in the case
1595: of an incompressible MHD flow.  Finally, we address the potential
1596: implications of our findings for shearing box models in which magnetic tension
1597: terms, induced by the curvature of the background field, are not considered.
1598: 
1599: \subsection{Importance of Curvature Terms}
1600: \label{subsec:importance}
1601: 
1602: \begin{figure}[tbh]
1603: \begin{center}
1604: \includegraphics[width=.75\textwidth, angle=90]{f6.eps}
1605: \caption{The importance of the curvature terms proportional to
1606:   $\epsilon_2$ and $\epsilon_3$, as defined by the ratios $R_1$
1607:   (eq.~[\ref{eq:R1_nodim}]) and $R_2$ (eq.~[\ref{eq:R2_nodim}]).  For
1608:   illustrative purposes, we have considered a Keplerian disk with
1609:   $c_{\rm s}=0.05$, $v_{{\rm A}z}= 0.01$, and $v_{{\rm A} \phi}= 0.4$.}
1610: \label{fig:ratios}
1611: \end{center}
1612: \end{figure}
1613: 
1614: In section \S\ref{sec:mhd equations and dispersion relation} we
1615: mentioned that the terms proportional to $\epsilon_i$, for
1616: $i=1,2,3,4$, are usually neglected in local stability analyses
1617: due to their $1/r_0$ dependence. Some
1618: of these terms, however, are also proportional to the magnitude of the
1619: toroidal field. In this paper, we found that, when
1620: strong toroidal fields are considered, these terms led to substantial
1621: modifications to the stability criteria of MHD modes known to be valid
1622: in the limit of weak fields.
1623: After solving the full problem, we are in a better
1624: position to understand why this is the case.
1625: 
1626: To illustrate the point, consider the ratio of the term proportional to $k_z$
1627: to the one proportional to $\epsilon_2$ in equation 
1628: (\ref{eq:pert_euler_r_grads}) and the ratio of the term 
1629: proportional to $k_z$ to the one proportional to $\epsilon_3$ in equation
1630: (\ref{eq:pert_euler_phi_grads}), i.e.,
1631: \begin{equation}
1632: R_1 \equiv \frac{\epsilon_2}{ik_z} \frac{v_{{\rm A} \phi}^2}{v_{{\rm A} z}}
1633: \frac{\delta \rho}{\delta v_{{\rm A} r}}    \qquad {\rm and} \qquad R_2 \equiv
1634: \frac{\epsilon_3}{ik_z} \frac{v_{{\rm A} \phi}}{v_{{\rm A} z}}
1635: \frac{\delta v_{{\rm A} r}}{\delta v_{{\rm A} \phi}}~.
1636: \end{equation}
1637: In order to ensure that the contributions due to curvature are
1638: negligible in a local analysis regardless of the magnitude of the
1639: toroidal field component, we should be able to ensure that the
1640: conditions $R_1 \ll 1$ and  $R_2 \ll 1$ hold in the limit of large $k_z$
1641: for any value of  $v_{{\rm A} \phi} \lesssim 1$.   While it is
1642: encouraging that both dimensionless ratios are proportional to $1/k_z$,
1643: they are also proportional to the ratio of perturbed quantities, which
1644: we do not know \emph{a priori}.   It is only after having found
1645: the eigenfrequencies $\omega(k_z)$ by taking into account all the
1646: curvature terms that we can properly address this issue.
1647: 
1648: 
1649: \begin{figure}[tbh]
1650: \begin{center}
1651: \includegraphics[width=.475\textwidth]{f7a.eps}
1652: \includegraphics[width=.475\textwidth]{f7b.eps}
1653: \includegraphics[width=.475\textwidth]{f7c.eps}
1654: \includegraphics[width=.475\textwidth]{f7d.eps}
1655: \caption{The black dots represent unstable modes obtained from solving
1656:         numerically the dispersion relation (\ref{eq:disp_full_nodim_kr})
1657:         as a function of the toroidal \Alfven speed. As an example, we have
1658:         assumed $c_{\rm s}=0.05$ and $v_{{\rm A} z}=
1659:         0.01$. In each plot, we consider different values of the rotational
1660:         profile, $q = - d\ln \Omega/ d\ln r$. Note that,
1661:         the highest value of the local toroidal \Alfven speed considered
1662:         here corresponds to the local circular velocity.}
1663: \label{fig:dlog}
1664: \end{center}
1665: \end{figure}
1666: 
1667: 
1668: 
1669: We can calculate how the ratios $R_1$ and $R_2$ depend on the
1670: wavenumber $k_z$ as follows.
1671: The ratio $R_1$ can be recast using equations
1672: (\ref{eq:rho_Bphi_limit}) and (\ref{eq:motion_Br}) as
1673: \begin{equation}
1674: \label{eq:R1_nodim}
1675: R_1 = \frac{\epsilon_2}{k_z} \frac{v_{{\rm A}\phi}}{v_{{\rm A} z}}
1676:      \left(\frac{v_{{\rm A}\phi}}{c_{\rm s}}\right)^2 
1677:      \frac{ \omega^2 - \left(2 \frac{d\ln \Omega}{d\ln r} + 
1678:      k_z^2 v_{{\rm A} z}^2 -2 \epsilon_4\frac{\omega}{k_z}\frac{v_{{\rm
1679:      A}\phi}}{v_{{\rm A} z}} \right) } 
1680:      {2\omega \left[ 1+\left(\frac{v_{{\rm A}\phi}}{c_{\rm
1681:      s}}\right)^2  \right] + k_z v_{{\rm A} z} v_{{\rm A}\phi}
1682:      \left[ 2\epsilon_1 + \epsilon_2 \left(\frac{v_{{\rm A}\phi}}
1683:      {c_{\rm s}}\right)^2  \right]} ~.
1684: \end{equation}
1685: In a similar way, we can rewrite the ratio $R_2$ using equation
1686: (\ref{eq:motion_Bphi}) as
1687: \begin{equation}
1688: \label{eq:R2_nodim}
1689: R_2 = \frac{\epsilon_3}{k_z} \frac{v_{{\rm A}\phi}}{v_{{\rm A} z}}
1690: \frac{\omega^2 \left[ 1+\left(\frac{v_{{\rm A}\phi}}
1691: {c_{\rm s}}\right)^2  \right] - k_z^2 v^2_{{\rm A} z}}
1692: {\omega \left[ 2+ \epsilon_4  \frac{\omega}{k_z}\frac{v_{{\rm
1693: A}\phi}}{v_{{\rm A} z}} \right] +\epsilon_3 k_z v_{{\rm A} z} 
1694: v_{{\rm A}\phi} } ~.
1695: \end{equation}
1696: For the sake of simplicity, let us consider a given value for the
1697: toroidal \Alfven speed, e.g., $v_{{\rm A}\phi}=0.4$.  Figure
1698: \ref{fig:ratios} shows the dependence of the ratios $R_1$ and $R_2$
1699: on wavenumber for the unstable.  The eigenfrequencies
1700: $\omega(k_z)$ were obtained by solving equation
1701: (\ref{eq:disp_full_nodim_kr}) with $\epsilon_i=1$, for $i=1,2,3,4$,
1702: considering a Keplerian disk with  $c_{\rm s}=0.05$ and $v_{{\rm
1703: A}z}=0.01$. The ratios $R_1$ and $R_2$ for the unstable modes (with
1704: $v_{{\rm A}\phi} = 8 c_{\rm s}$) in Regions II and III in Figure
1705: \ref{fig:unstable_regions_num} are clearly identified.  The complete
1706: mode structure corresponding to this case can be seen 
1707: in the left panels of Figure \ref{fig:approx}.
1708: 
1709: It is important to stress that neither the real nor the imaginary parts of
1710: either $R_1$ or $R_2$  are negligible compared to unity even for
1711: \Alfven speeds of order a few times the sound speed.  In fact, for the
1712: unstable modes, the ratio $R_1$ is of order unity   and the ratio
1713: $R_2$ is in some cases larger than one by one order of magnitude. Their
1714: functional form is significantly different than the assumed $1/k_z$.
1715: 
1716: 
1717: \subsection{Magnetorotational Instabilities with Superthermal Fields}
1718: \label{subsec:mri_superthermal}
1719: 
1720: 
1721: 
1722: \begin{figure}[tbh]
1723: \begin{center}
1724: \includegraphics[width=.475\textwidth]{f8a.eps}
1725: \includegraphics[width=.475\textwidth]{f8b.eps}
1726: \includegraphics[width=.475\textwidth]{f8c.eps}
1727: \includegraphics[width=.475\textwidth]{f8d.eps}
1728: \caption{The black dots represent unstable modes obtained from solving the
1729:         dispersion relation (\ref{eq:disp_full_nodim_kr}) numerically, for
1730:         a Keplerian disk with $v_{{\rm A}z}= 0.01$. In each plot, 
1731: 	different values of the local sound speed, $c_{\rm s}$, are
1732:         considered. Note that in this case, the axes are not normalized
1733:         by the particular value of the local sound speed, but rather
1734:         by our initial choice of dimensionless variables,
1735: 	see \S \ref{sec:mhd equations and dispersion relation}.}
1736: \label{fig:sound}
1737: \end{center}
1738: \end{figure}
1739: 
1740: 
1741: 
1742: In \S\ref{sec:numerical solutions} we demonstrated that, when the
1743: toroidal magnetic field in a differentially rotating MHD flow becomes
1744: superthermal, three distinct instabilities can be identified, which we
1745: denote by roman numerals I, II, and III in
1746: Figure~\ref{fig:unstable_regions_num}. We summarize the 
1747: qualitative characteristics of these instabilities below.
1748: 
1749: In contrast to the weak-field MRI, all three instabilities correspond
1750: to compressible MHD modes. Moreover, while the traditional MRI
1751: corresponds to perturbations with negligible displacements along the
1752: vertical direction, this is not true for any of the three
1753: instabilities with superthermal toroidal fields. Instead, vertical
1754: displacements are an important characteristic of these instabilities
1755: and they occur with negligible acceleration, under a force balance
1756: between thermal and magnetic pressure. Finally, as in the case of the
1757: MRI, there is no significant acceleration along the radial direction
1758: but rather a force balance between magnetic tension, magnetic
1759: pressure, and thermal pressure.
1760: 
1761: In Figures~\ref{fig:dlog} and \ref{fig:sound} we study numerically the
1762: dependences of the three instabilities on the properties of the
1763: background flow. As also shown in the case of the weak-field MRI
1764: (Balbus \& Hawley, 1991),  instability I occurs only in 
1765: differentially rotating flows, with radially decreasing angular velocity. 
1766: However, instability I also requires the presence of a
1767: non-negligible thermal pressure. Either a radially increasing angular
1768: velocity or a superthermal toroidal field can suppress instability I
1769: and hence the traditional MRI. 
1770: 
1771: Instability II is ubiquitous, whenever the background toroidal field
1772: of the flow is significantly superthermal. Indeed, it occurs even for
1773: flat (see Fig.~\ref{fig:dlog}) or very cold (see Fig.~\ref{fig:sound})
1774: flows. In a sense opposite to instability I, the steepness of the
1775: rotational profile determines the minimum unstable wavenumber, whereas
1776: the magnitude of the sound speed determines the minimum toroidal field
1777: strength required for the instability to occur.
1778: This instability seems to correspond to a generalization of the
1779: axisymmetric toroidal buoyancy (ATB) modes identified in \cite{KO00},
1780: where the case $c_{\rm s}=0$ was studied.  In a similar way to
1781: instability II, the ATB modes with $c_{\rm s}=0$ become unstable
1782: for all wavenumbers exceeding a critical value (for vertical modes this
1783: value is just given by $k_{\rm BH}$).  
1784: When a finite sound speed is considered, however, 
1785: thermal effects play an important
1786: role at small scales by completely stabilizing all the modes with
1787: wavenumbers larger than $k_z^{{\rm c} 2}$ (eq.~[\ref{eq:kc2}]). 
1788: 
1789: Finally, instability III depends strongly on the rotational profile
1790: but very weakly on the sound speed. For rotationally supported flows
1791: (i.e., for $v_{\rm A \phi}\ll 1$), instability III occurs only for
1792: significantly steep rotational profiles, e.g.,  
1793: $q=|d\ln \Omega/ d\ln r|\gtrsim 1.0$, for
1794: the parameters depicted in Figure~\ref{fig:dlog}.
1795: 
1796: 
1797: \newpage
1798: 
1799: \begin{figure}[tbh]
1800: \begin{center}
1801: \includegraphics[width=\textwidth]{f9.eps}
1802: \caption{The growth rate evolution of the different instabilities
1803: defined in \S \ref{subsec:unstable modes} for increasing magnetic
1804: field strength parameterized in terms of  the ratio $c_{\rm s}^2/
1805: v_{{\rm A}}^2$ for a fixed pitch angle  $i\equiv\tan^{-1} (v_{{\rm A}
1806: z}/v_{{\rm A} \phi}) =25 \degr$.  \emph{Left panels}: Growth rates of
1807: the unstable solutions to the full dispersion relation
1808: (\ref{eq:disp_full_nodim_kr}), when all curvature terms are taken into
1809: account. \emph{Right panels}: Growth rates of the unstable solutions
1810: to the dispersion relation (\ref{eq:disp_cseq0}),  i.e., when all
1811: curvature terms are neglected. In both cases, we have considered a
1812: vanishing ratio $k_r/k_z$, $v_{{\rm A} z}=0.05$, and a  Keplerian
1813: disk.  The dotted lines, in both panels,  show the 
1814: stabilization of the standard MRI 
1815: as the magnetic field becomes superthermal (instability
1816: I). The solid and dashed lines on the left panel show the growth rates
1817: corresponding to instability III and II respectively. As discussed in
1818: \S \ref{sec:numerical solutions}, these instabilities do not have a
1819: counterpart when the magnetic tension induced by bending of toroidal
1820: field lines is neglected.  For completeness we have included, in the
1821: right panel, with solid and dashed lines the unstable solutions to
1822: equation (44) in Kim \& Ostriker (2000). Instability II corresponds to
1823: a generalization of the ATB mode when thermal effects are accounted for.}
1824: \label{fig:growth_rates}
1825: \end{center}
1826: \end{figure}
1827: 
1828: \subsection{Comparison to Previous Analytical Studies}
1829: \label{subsec:Comparison to previous analytical studies}
1830: 
1831: Soon after the original paper by \citet{BH91}, \citet{Knobloch92}
1832: critiqued their approach to the study of local instabilities
1833: for lacking the contributions of curvature terms.
1834: \citet{Knobloch92} 
1835: formulated the stability analysis of a vertically unstratified,
1836: incompressible disk as an eigenvalue problem in the radial
1837: coordinate. He found that the presence of a toroidal field
1838: component changes the conditions for the presence of the instability
1839: as well as the character of the unstable modes from purely exponentials to
1840: overstable (i.e., $Re[\omega] \ne 0$). \citet{GB94} argued against
1841: Knobloch's findings regarding overstability, stating that it arose as a
1842: consequence of having kept only  small order terms (like
1843: $v_{\rm A}/c_{\rm s}$ and $v_{\rm A}/\Omega_0 r_0$).  They concluded
1844: that these contributions would have been negligible had the flow been
1845: considered compressible.
1846: 
1847: As we comment in \S \ref{subsec:analytic approximations},  Knobloch's
1848: dispersion relation is correct even in the limit $c_{\rm s} \gg 
1849: v_{{\rm A}\phi}$  (i.e., without the necessity of imposing strict
1850: incompressibility).  Formally speaking, the linear term in $\omega$ in
1851: equation (\ref{eq:disp_incomp_nodim}) does break the symmetry of the
1852: problem allowing for unstable modes with $Re(\omega) \ne 0$.  But it
1853: is also the case that, in the limit  $c_{\rm s} \gg v_{{\rm A}\phi}$, because
1854: of the relative magnitude of the coefficients in the dispersion
1855: relation (\ref{eq:disp_incomp_nodim}), we do not expect the stability
1856: properties of the flow to differ greatly from those described by the
1857: incompressible MRI.  As we mention in \S \ref{sec:previous treatments},
1858: in order to see significant differences, the \Alfven speed would 
1859: have to be of the
1860: order of the circular speed and therefore we do not expect the
1861: curvature terms to play a significant role on the stability of
1862: incompressible, rotationally supported flows.
1863: On the other hand, if we allow the MHD fluid to be compressible and
1864: consider the curvature of the background flow, the
1865: mode structure can be radically different from what is expected for
1866: the compressible MRI (c.f. Blaes \& Balbus 1994). This is the case, even if the
1867: toroidal \Alfven speed exceeds the sound speed by a factor of a few
1868: without the necessity of violating the condition of a rotationally
1869: supported disk (see Fig. \ref{fig:comp_real}). 
1870: 
1871: The stability of axisymmetric perturbations in weakly ionized and
1872: weakly magnetized shear flows was considered by \citet{BB94}. They
1873: showed that, when ionization equilibrium is considered in the
1874: two-fluid approach, strong toroidal fields can fully stabilize the
1875: flow. As part of their study, they relaxed the Boussinesq
1876: approximation in the case of a single fluid and argued that, to all
1877: orders in the field strength, the magnitude of  
1878: $B_\phi$ does not affect the stability
1879: criterion.  
1880: As noted by \citet{CP95}, this conclusion was reached
1881: because the terms proportional to
1882: $B_\phi/r_0$ were not included in the local analysis.
1883: 
1884: 
1885: 
1886: 
1887: The behavior of the MRI in cold MHD shearing flows,  has been
1888: addressed by Kim \& Ostriker (2000).  When performing their local
1889: analysis, these authors obtained the compressible version of the standard
1890: dispersion relation for the MRI and studied its solutions  for
1891: different values of the ratio $c_{\rm s}^2/v_{\rm A}^2$.  
1892: Analyzing their dispersion relation (i.e.,
1893: their equation [57] which is equivalent to equation [\ref{eq:disp_cseq0}] in
1894: this study), Kim \& Ostriker (2000) concluded that, when the magnetic
1895: field is superthermal, the inclusion of a toroidal component
1896: suppresses the growth rate of the MRI. Moreover, they found that, 
1897: for a Keplerian rotation law, 
1898: no axisymmetric MRI takes place in very cold MHD flows
1899: if $i<30\degr$, where $i$ is the pitch angle of the local 
1900: magnetic fields, $i\equiv\tan^{-1}(v_{{\rm A} z}/v_{{\rm A} \phi})$.
1901: 
1902: \begin{figure}[tbh]
1903: \begin{center}
1904: \includegraphics[width=\textwidth]{f10.eps}
1905: \caption{The real parts of the solutions to the dispersion relation
1906: (\ref{eq:disp_full_nodim_kr})  corresponding to  $c_{\rm
1907: s}^2/v_{{\rm A}}^2=0.01$  for a pitch angle   $i=\tan^{-1} (v_{{\rm A}
1908: z}/v_{{\rm A} \phi}) =25 \degr$, a vanishing ratio $k_r/k_z$,
1909: $v_{{\rm A} z}=0.05$, and a Keplerian disk.  \emph{Left
1910: panel}: Solutions to the full dispersion relation
1911: (\ref{eq:disp_full_nodim_kr}), when all curvature terms are taken into
1912: account. \emph{Right panel}: Solutions  to the dispersion relation
1913: (\ref{eq:disp_blaes}), i.e., when all curvature terms are
1914: neglected. Open circles indicate unstable modes.  In
1915: both cases, the unstable modes with the shortest  wavelength correspond to
1916: $k_{\rm BH} v_{{\rm A} z}$.}  
1917: \label{fig:mri_stable}
1918: \end{center}
1919: \end{figure}
1920: 
1921: 
1922: In  \S \ref{sec:onset of instabilities} we showed that,
1923: depending on the strength of the toroidal field component,
1924: accounting for the finite curvature of the background magnetic field
1925: and the finite compressibility of the flow could be crucial in
1926: establishing which modes are subject to instabilities.
1927: In particular,
1928: we stated that both effects should be considered simultaneously
1929: whenever the local value of the toroidal \Alfven speed exceeds the
1930: geometric mean of the local sound speed and the local rotational
1931: speed (for a Keplerian disk). 
1932: However, this analytic criterion was found to be relevant for
1933: the modes with frequencies satisfying the condition $\omega^2 \ll
1934: k_z^2 c_{\rm s}^2$.   Therefore, it is not obvious that we can trust
1935: this criterion in the  limit $c_{\rm s} \rightarrow 0$.  
1936: 
1937: 
1938: 
1939: 
1940: 
1941: 
1942: 
1943: In order to
1944: see whether finite  curvature effects do play a role in the stability
1945: of cold MHD  flows we solved the complete  dispersion relation
1946: (\ref{eq:disp_full_nodim_kr}) for a pitch angle $i=25\degr$ and
1947: considered different values for the ratio $c_{\rm s}^2/v_{\rm
1948: A}^2$. 
1949: Figure \ref{fig:growth_rates} shows the growth rates for the
1950: unstable modes of our dispersion relation (i.e.,
1951: eq.~[\ref{eq:disp_full_nodim_kr}]) and  compares them to the ones of
1952: the dispersion relation obtained when the curvature terms are
1953: neglected (i.e., eq.~[\ref{eq:disp_cseq0}]).
1954: In both cases, we have considered
1955: $k_z \gg k_r$, $v_{{\rm A} z}=0.05$, and a  Keplerian disk. The
1956: stabilization of the standard MRI (i.e., instability I) as the
1957: magnetic field becomes superthermal ($v_{\rm A}>c_{\rm s}$)  is
1958: evident (dotted lines in both panels).  When the effects of magnetic
1959: tension are considered, not only does the growth rate of the MRI
1960: decrease faster for low values of  $c_{\rm s}^2/v_{\rm A}^2$ but the
1961: modes with longest wavelengths are no longer unstable
1962: (e.g. when   $c_{\rm s}^2/v_{\rm A}^2=0.01$). Because of
1963: this, the  MRI is completely stabilized even for finite values of
1964: $c_{\rm s}^2/v_{\rm A}^2$. In contrast, when the curvature of the 
1965: field lines is neglected, the growth rates decrease but the range
1966: of unstable modes remains unchanged as  $c_{\rm s} \rightarrow 0$
1967: (right panel in Figure \ref{fig:growth_rates});  
1968: it is only when $c_{\rm s} = 0$ that the MRI is completely suppressed. 
1969: For completeness, we present in Figure \ref{fig:mri_stable}  the real parts
1970: of the  solutions to the dispersion relation
1971: (\ref{eq:disp_full_nodim_kr})  for $i=25\degr$, $k_z \gg k_r$,  $v_{{\rm
1972: A} z}=0.05$, $c_{\rm s}^2/v_{\rm A}^2=0.01$ and a Keplerian disk, in
1973: the cases where curvature terms are considered (left panel) and
1974: neglected (right panel). The stabilization of the MRI at low
1975: wavenumbers and the emergence of instability III are evident.
1976: Note that, from Figure \ref{fig:mri_stable}, 
1977: it is clear that the inclusion of magnetic tension 
1978: terms can cause modifications to the mode structure
1979: when $v_{\rm A} \gg c_{\rm s}$, 
1980: even  when the toroidal and vertical components of the
1981: magnetic field are comparable.
1982: 
1983: 
1984: In \S \ref{subsec:mri_superthermal} we mentioned that instability
1985: II seems to be a generalization of the axisymmetric toroidal buoyancy
1986: (ATB) modes, identified by Kim \& Ostriker (2000), that accounts for
1987: finite temperature effects. Further indication that this is indeed the
1988: case can be found in Figure \ref{fig:growth_rates} where we have
1989: plotted (dashed lines) the growth rates corresponding to instability
1990: II and the one corresponding to the ATB mode (solutions  of equation
1991: [44] with $\omega^2 \ll k_z^2 v_{{\rm A} z}^2$  in  Kim \& Ostriker
1992: 2000).  Although finite compressibility suppresses instability II at
1993: large wavenumbers, it is clear that, as $c_{\rm s} \rightarrow 0$, the
1994: growth rates associated with instability II tend continuously to the
1995: growth rate of the cold ATB mode.  For completeness, we have also
1996: included, in the right panel of Figure \ref{fig:growth_rates},  the
1997: growth rate corresponding to the remaining unstable solution  of
1998: equation (44) in  Kim \& Ostriker (2000) (solid line).  This growth
1999: rate should be interpreted with great care since the aforementioned
2000: equation was derived
2001: under the condition $\omega^2\ll k_z^2 v_{\rm A}^2$, which is not
2002: satisfied by the corresponding unstable mode.   Nonetheless, we have
2003: included it to show the similarities that it shares with the growth
2004: rates corresponding to instability III (solid lines in the left panel)
2005: as  $c_{\rm s} \rightarrow 0$ .  Note that the growth rates
2006: corresponding to instability III increase as  $c_{\rm s} \rightarrow
2007: 0$ and they saturate at  $c_{\rm s} =0$.  Although the higher critical
2008: wavenumber and the growth rate around this critical wavenumber seem to
2009: be the same for both instabilities, the differences between them at
2010: low wavenumbers is also evident. These differences become more
2011: dramatic as the pitch angle increases.
2012: 
2013: 
2014: Curvature terms cannot, of course, be neglected in global treatments
2015: of magnetized accretion disks.  It is, therefore, not surprising that
2016: new  instabilities, distinct from the MRI, have already been found in
2017: global studies in which strong fields were considered.  In
2018: particular, \citet{CP95} performed a global stability analysis to
2019: linear axisymmetric perturbations of an incompressible, differentially
2020: rotating fluid, threaded by vertical and toroidal fields. They
2021: considered power-law  radial profiles for the angular velocity and the
2022: toroidal and vertical components of the field. Each of these were
2023: parameterized as  $\Omega \propto r^{-a}$, $B_\phi \propto r^{-b+1}$,
2024: and $B_z \propto r^{-c+1}$, respectively. Most of their analysis dealt
2025: with a constant vertical field and they allowed variations of the
2026: exponents $(a,b)$,  with the restriction that they correspond to a
2027: physical equilibrium state with a stationary pressure distribution.
2028: Although the majority of that paper dealt with global characteristics,
2029: they also performed a WKB analysis and concluded that, for $ 3/2 \le
2030: a=b \le 2$  and $v_{{\rm A} \phi} < 1 $, the growth rate of unstable
2031: modes is suppressed on both short and long wavelengths and it
2032: approaches zero when $v_{{\rm A} \phi} \rightarrow 1$. On the other
2033: hand, for  $ a = b \ne 2$ and $v_{{\rm A} \phi} > 1 $,  they found a
2034: new instability, with a growth rate that increased with $v_{{\rm A}
2035: \phi}$. They call this the Large Field Instability (LFI) and showed
2036: that it can be stabilized for sufficiently large  $v_{{\rm A} z}$.
2037: 
2038: It is worth mentioning the major qualitative differences between the
2039: LFI and the new instability discussed in \S \ref{sec:numerical
2040: solutions} that emerges for $k_z < k_{\rm BH}$ after the stabilization
2041: of the MRI. Although it is true that, for our instability to be
2042: present, it is necessary for the toroidal \Alfven speed to exceed the
2043: local sound speed, there is no need to invoke \Alfven speeds larger
2044: than the local rotational speed. This is in sharp contrast with the
2045: LFI which only appears for $v_{{\rm A} \phi} > 1$. Regarding the range
2046: of unstable wavenumbers,  the LFI remains unstable for $k_z \rightarrow
2047: 0$, albeit with diminishing growth rate for large values of $v_{{\rm
2048: A} z}$. This is not the case for the new instability present at low
2049: wavenumbers in our study.  This can be seen, for example, in the
2050: left panels in Figure \ref{fig:approx}. Perhaps
2051: the most noticeable difference is that the two instabilities in
2052: \citet{CP95} that are present in the case $a = b \ne 2$ do not seem
2053: to coexist under any particular circumstances. The instability present
2054: for $v_{{\rm A} \phi} < 1$ reaches zero growth for $v_{{\rm A} \phi}
2055: \rightarrow 1$, while the LFI appears for $v_{{\rm A} \phi} =1$ and
2056: its growth rate is proportional to $v_{{\rm A} \phi}$. When
2057: compressibility is considered, however, the two new instabilities found in
2058: our study can coexist even for \Alfven speeds smaller than the local
2059: rotational speed.
2060: 
2061: 
2062: 
2063: \subsection{Implications for Shearing Box Simulations}
2064: \label{sec:implications}
2065: 
2066: In an attempt to capture the most relevant physics without all the
2067: complexities involved in global simulations,  the shearing box
2068: approach has been widely used in numerical studies of magnetized
2069: accretion disks (see, e.g., Hawley, Gammie, \& Balbus 1994).  The aim
2070: of the shearing box approximation is to mimic a small region of a
2071: larger disk. The size of the box is usually  $H_z \times 2\pi H_z
2072: \times H_z$, with $H_z$ the thermal scale height of the isothermal
2073: disk.  In this approach, it is common to adopt a pseudo-Cartesian
2074: local system centered at  $r_0$ and in corotation with the disk with
2075: an angular frequency $\Omega_0$, with coordinates $x=r-r_0$,
2076: $y=r_0(\phi-\Omega_0 t)$, and $z$. The effects of differential
2077: rotation are then considered by imposing a velocity gradient in the
2078: radial direction. For a Keplerian accretion disk this is achieved by
2079: setting $v_y= -(3/2) \Omega_0 x$.
2080: 
2081: 
2082: \begin{figure}[tbh]
2083: \begin{center}
2084: \includegraphics[width=0.7\textwidth, angle=90]{f11.eps}
2085: \caption{The implication of our study for shearing box simulations.
2086: \emph{Left panels}: Solutions to the full dispersion relation
2087: (\ref{eq:disp_full_nodim_kr}), when all curvature terms are taken into
2088: account. \emph{Right panels}: Solutions to the dispersion relation
2089: (\ref{eq:disp_blaes}), i.e., when all curvature terms are
2090: neglected. In both cases, we have considered 
2091: $c_{\rm s}= 0.007$, $v_{{\rm A}z}=0.01$, $v_{{\rm A} \phi}=0.1$, 
2092: and a Keplerian disk.
2093: Open circles in upper panels indicate unstable modes. The vertical
2094: line indicates the minimum wavenumber (i.e., largest wavelength) that
2095: can be accommodated in the simulations of a strongly-magnetized corona
2096: above a weakly magnetized disk by \citet{MS00}.}
2097: \label{fig:shearing_box}
2098: \end{center}
2099: \end{figure}
2100: 
2101: 
2102: In most studies of unstratified shearing boxes, \Alfven speeds rarely
2103: exceed the value of the local sound speed (see, e.g., Hawley, Gammie,
2104: \& Balbus 1995, 1996). This is mainly because they are designed to
2105: simulate the mid-plane of the disk where the flow is relatively
2106: dense. In \S \ref{subsec:analytic approximations}, we have seen that,
2107: as long as the toroidal \Alfven speed does not exceed the critical
2108: value $v_{{\rm A} \phi}^2 = c_s \kappa $, neglecting  magnetic tension
2109: due to the curvature of toroidal field lines does not seem to affect the
2110: outcome of the MRI and hence the shearing box approach is well
2111: justified.  However, when stratification is taken into account,
2112: usually by adopting a density profile of the form $\rho \propto
2113: \exp[-z^2/(2H^2)]$ in the case of isothermal disks, the steep drop in
2114: the density beyond a few scale heights can potentially lead to a
2115: magnetically dominated flow, with \Alfven speeds larger than the
2116: critical value $v_{{\rm A} \phi}^2 = \kappa c_{\rm s}$.
2117: 
2118: As discussed in the introduction, \citet{MS00} carried out
2119: three-dimensional MHD simulations to study the evolution of a
2120: vertically stratified, isothermal, compressible, magnetized
2121: shear flow.  The simulations were local in the plane of the disk but
2122: vertically extended up to $\pm 5$ thermal scale heights. This allowed
2123: them to follow the highly coupled dynamics of the weakly magnetized
2124: disk core and the rarefied magnetically-dominated (i.e., $\beta < 1$)
2125: corona that formed above the disk, for several (10 to 50) orbital
2126: periods.
2127: 
2128: Miller \& Stone (2000) considered a variety of models, all sharing the same initial
2129: physical background, but different initial field configurations.  In
2130: particular, they considered the following values: $\Omega_0=10^{-3}$,
2131: $2c^2_{\rm s}=10^{-6}$ (so that $H_z=\sqrt{2} c_{\rm s}/\Omega_0= 1$),
2132: and $r_0=100$. We mention some of the results they obtained for the
2133: models with initial toroidal fields (BY), which were qualitatively
2134: similar to the zero net $z$-field (ZNZ) models. After a few orbital
2135: periods, the presence of a highly magnetized (with plasma $\beta
2136: \simeq 0.1 - 0.01$) and rarefied (with densities two orders of
2137: magnitude lower than the disk mid-plane density) corona above $\sim 2$
2138: scale heights is evident.  Within both the disk and the corona, the
2139: ``toroidal'' component of the field ($B_y$), favored by differential
2140: rotation, dominates the poloidal component of the field by more than
2141: one order of magnitude (with $B_x^2 \simeq B_z^2$).
2142: 
2143: We can compare the predictions of our study to the mode structure that
2144: one might expect from the standard  compressible MRI for the
2145: particular values of sound and \Alfven speeds found in the strongly
2146: magnetized corona by \citet{MS00}.  To this end, we consider as
2147: typical (dimensionless) values  $c_{\rm s}=0.007$, $v_{{\rm A}
2148: \phi}=0.1$ and $v_{{\rm A} z}=0.01$, where we have assumed $\beta =
2149: 2(c_{\rm s}/ v_{\rm A})^2 \simeq 0.01$ and $B_\phi=10 B_z$.   
2150: The largest features that their simulations are able
2151: to accommodate are those with $k_z \sim 60$ (corresponding to a
2152: wavelength of 10 in the vertical direction). As it can be seen in
2153: Figure \ref{fig:shearing_box},  the role of the curvature terms is not
2154: negligible in two different respects. First, it completely stabilizes
2155: the perturbations on the longest scales well inside the numerical
2156: domain. Second, the growth rate for the unstable modes is
2157: significantly reduced.
2158: 
2159: It is difficult to extrapolate from the present work to address how
2160: the instabilities discussed here would couple to buoyancy in the
2161: presence of a stratified medium like the one considered by
2162: \citet{MS00}. Shearing boxes might also suffer from other problems
2163: when used to model strongly magnetized plasmas (e.g., the shearing
2164: sheet boundary conditions in the radial direction might not be
2165: appropriate for strong fields). The question is raised, however, about
2166: whether, because of their own Cartesian nature, they constitute a good
2167: approach at all to simulating compressible flows in which superthermal
2168: toroidal fields are present.  Despite the fact that the generation of
2169: strongly magnetized regions via the MRI in stratified disks seems hard
2170: to avoid, their stability properties will ultimately depend on both
2171: the use of proper boundary conditions and  proper accounting of the
2172: field geometry.
2173: 
2174: \section{\textsc{SUMMARY AND CONCLUSIONS}}
2175: \label{sec:summary and conclusions}
2176: 
2177: In this paper we have addressed the role of toroidal fields on the
2178: stability of local axisymmetric perturbations in compressible,
2179: differentially rotating, MHD flows, when the geometrical curvature  of
2180: the background is taken into account. In order to accomplish this
2181: task, without imposing restrictions on the strength of the background
2182: equilibrium field, we relaxed the Boussinesq approximation.  In
2183: particular, we have studied under which circumstances the curvature
2184: terms, intimately linked to magnetic tension in cylindrical coordinate
2185: systems, cannot be neglected.  We have shown that the MRI is
2186: stabilized and two distinct instabilities appear for strong toroidal
2187: fields.  At least for large wavenumbers,  the structure of the modes
2188: seems to be the result of a purely local effect that is accounted for
2189: when compressibility and curvature terms are consistently taken into
2190: account. In particular, we have demonstrated that, even for
2191: rotationally supported cylindrical flows, 
2192: both curvature terms and flow compressibility  have to be
2193: considered if, locally,  the toroidal \Alfven speed exceeds 
2194: the critical value given 
2195: by $v_{{\rm A} \phi}^2 = (\kappa/\Omega) c_{\rm s} \Omega r $ 
2196: (in physical units).
2197: 
2198: There is little doubt that a realistic treatment of normal modes in
2199: magnetized accretion disks has to contemplate gradients in the flow
2200: variables over large scales and should, therefore, be global in
2201: nature.  The results presented in this paper, however,  provide the
2202: complete dispersion relation and, more importantly, analytic
2203: expressions for some of its solutions that should be recovered, in
2204: the appropriate limit, by a study of global modes in  magnetized
2205: accretion disks, where compressibility effects are likely to be non
2206: negligible. We will address these issues in a future paper.
2207: 
2208: \begin{figure}[tb]
2209: \begin{center}
2210: \includegraphics[width=\textwidth]{f12a.eps}
2211: \includegraphics[width=\textwidth]{f12b.eps}
2212: \caption{Solutions to the dispersion relation
2213: (\ref{eq:disp_full_nodim_kr})  for different values of the ratio
2214: $k_r/k_z$. The left, central, and right panels show the results for
2215: $k_r/k_z=0.01, 0.1$, and 1 respectively.  In all three cases, we have
2216: considered $c_{\rm s}=0.05$, $v_{{\rm A} z}=0.01$, $v_{{\rm A}
2217: \phi}=0.4$, and a Keplerian disk.  Open circles
2218: in upper panels indicate unstable modes.  The same instabilities that
2219: were present in the case  with vanishing ratio $k_r/k_z$ 
2220: (left panels in Figure  \ref{fig:approx}) can be clearly identified.}
2221: 
2222: \label{fig:kr_ne_0}
2223: \end{center}
2224: \end{figure}
2225: 
2226: \newpage
2227: 
2228: \section*{\textsc{APPENDIX A}}
2229: \label{sec:APPENDIX A}
2230: 
2231: The central dispersion relation obtained in this study,  i.e.,
2232: equation (\ref{eq:disp_full_nodim_kr}), was derived considering
2233: axisymmetric perturbations in both the vertical and the radial
2234: directions. Throughout the majority of our analysis,  however, we
2235: focused our attention  on the modes with vanishingly small ratios
2236: $k_r/k_z$. Particular emphasis has been given to the study of these
2237: modes in the literature of weakly magnetized, differentially rotating
2238: disks, since these are the modes that exhibit the fastest growth rates
2239: \citep{BH92, BH98, Balbus03}.   After having analyzed the role played
2240: by magnetic tension forces due  to the finite curvature of strong
2241: toroidal field lines on the stability of these modes, we are in a
2242: better position to understand their effects on the modes for which the
2243: ratio $k_r/k_z$ is finite.
2244: 
2245: In Figure \ref{fig:kr_ne_0}, we present the solutions to the dispersion
2246: relation (\ref{eq:disp_full_nodim_kr}) for three different ratios of
2247: the radial to the vertical wavenumber, i.e., $k_r/k_z = 0.01, 0.1$,
2248: and 1. For the sake of comparison, we have used in this figure the
2249: same parameters that we used in obtaining  the left panels in Figure
2250: \ref{fig:approx}, for which the ratio $k_r/k_z$ was considered to be
2251: vanishingly small;
2252: we have assumed $c_{\rm s}= 0.05$, $v_{{\rm A}z}=0.01$,
2253: $v_{{\rm A} \phi}=0.4$ and a Keplerian disk.   In all the cases, the
2254: same instabilities (II and III) that were present in the case $k_z \gg
2255: k_r$ can be clearly identified.   Note however, that even for very
2256: small ratios $k_r/k_z$ (e.g., left panels in Figure \ref{fig:kr_ne_0})
2257: some of the modes that were stable in the case $k_z \gg k_r $ 
2258: become unstable,  albeit with negligible growth rate,
2259: when compared with the other unstable modes.  It is also evident that
2260: in the limit $k_z \gg k_r$, the mode structure in Figure
2261: \ref{fig:kr_ne_0} tends continuously toward the mode structure in the
2262: left panels in Figure \ref{fig:approx}. It is this continuous behavior
2263: that ultimately  justifies the study of modes with negligible ratio
2264: $k_r/k_z$ in a local stability analysis.
2265: 
2266: In the case $k_r =k_z$ (right panels in Figure \ref{fig:kr_ne_0}),
2267: the value of the critical vertical wavenumbers
2268: for the onset of instabilities, i.e., $k_z^{{\rm c} 1}$ and $k_z^{{\rm c}
2269: 2}$ in  equations (\ref{eq:kc1}) and (\ref{eq:kc2}) respectively, are
2270: different with respect to the case in which $k_z \gg k_r$ by a factor
2271: $\sqrt{2}$. This  indicates that $k_z$ and $k_r$ play similar roles in
2272: establishing these critical wavenumbers. The growth rates of all these
2273: modes are reduced with respect to the case with  $k_z \gg k_r$.
2274: This behavior is similar to the one observed in the case of weak
2275: magnetic fields.  It is important to stress that,  even for the modes
2276: with comparable values of vertical and radial wavenumbers, 
2277: the general characteristics of
2278: the  instabilities for strong toroidal fields  that we discussed in \S
2279: \ref{subsec:unstable modes}  are insensitive to the inclusion of a
2280: non-negligible $k_r$.
2281: 
2282: The completely new feature in Figure \ref{fig:kr_ne_0}  is the
2283: appearance of another instability with a growth rate that does not
2284: seem to depended on wavenumber; the terms  proportional to $ik_r$ in
2285: equation (\ref{eq:disp_full_nodim_kr}) are crucial for the appearance
2286: of this new instability.  The mode that is unstable seems to
2287: correspond to the mode that becomes the \Alfven mode in the limit of
2288: no rotation.  With increasing values of the ratio $k_r/k_z$, the
2289: growth rates of the instabilities studied in \S \ref{subsec:unstable
2290: modes} go to zero, but the new instability in Figure \ref{fig:kr_ne_0}
2291: persists.  Note that for $k_r \simeq k_z$ the growth rates of all the
2292: instabilities in  Figure \ref{fig:kr_ne_0} are comparable.
2293: 
2294: As an aside, we point out that all the terms that are proportional to
2295: $i k_r$, as opposed to $k_r^2$, in equation
2296: (\ref{eq:disp_full_nodim_kr}) are also proportional to some factor
2297: $\epsilon_i$ with $i=1, 2, 3, 4$. All of these terms are negligible
2298: for sufficiently small ratios $k_r/k_z$ no matter how strong the
2299: toroidal field is.  Indeed,  if we consider perturbations with small
2300: enough radial wavelengths, at some point, curvature effects will not
2301: be important. However, this is not true in the vertical direction. In
2302: that case, we can ignore the curvature terms only when the toroidal
2303: field is weak.
2304: 
2305: \section*{\textsc{APPENDIX B}}
2306: \label{sec:APPENDIX B}
2307: 
2308: In the various sections of the present study, we have seen that
2309: the toroidal component of the magnetic field does play a role
2310: in determining the stability criteria. In fact, for superthermal 
2311: fields and quasi toroidal configurations. it dictates the values of
2312: some of the critical wavenumbers for the onset of instabilities
2313: (see \S \ref{subsec:analytic approximations}).
2314: A particular, simple case, in which the importance of
2315: considering both compressibility and magnetic tension terms can be
2316: appreciated, is the study of modes with negligible frequency, i.e.,
2317: with $\omega \ll 1$. We can obtain the wavenumber of these modes  by imposing
2318: $\omega = 0$ to be a solution of the dispersion relation
2319: (\ref{eq:disp_full_nodim_kr}). We obtain, in physical dimensions,
2320: \begin{equation}
2321: \label{eq:k_critical}
2322: (k_z^0)^2 = - 2 \left.\frac{d \ln \Omega}{d \ln r}\right|_0  \left(
2323: \frac{\Omega_0 r_0}{v_{{\rm A} z}} \right)^2  + 2 \epsilon_1
2324: \epsilon_3  \left(\frac{v_{{\rm A} \phi}}{v_{{\rm A} z}} \right)^2 +
2325: \epsilon_2 \epsilon_3 \left(\frac{v_{{\rm A} \phi}}{v_{{\rm A} z}}
2326: \right)^2 \left(\frac{v_{{\rm A}\phi}}{c_{\rm s}} \right)^2 ~.
2327: \end{equation}
2328: 
2329: In what follows, let us consider a rotationally supported disk whose
2330: rotational profile is not too steep  (i.e., $|d \ln \Omega / d \ln r|$
2331: is of order unity).  The \Alfven speed $v_{{\rm A} z}$ appears in all
2332: three terms on the right hand side of equation (\ref{eq:k_critical})
2333: and therefore it does not play a role in determining the relative
2334: magnitudes between them.  Unlike the second term, the third term is
2335: not necessarily small  with respect to the first one, when
2336: superthermal fields are considered.  In this case, it seems again
2337: (see \S \ref{subsec:analytic approximations}) safe
2338: to neglect the curvature term proportional to  $\epsilon_1 \delta
2339: v_{{\rm A}\phi}$ in equation  (\ref{eq:pert_euler_r_grads}). However,
2340: had we neglected the curvature term proportional to $ \epsilon_2
2341: \delta \rho$ in equation (\ref{eq:pert_euler_r_grads}) or the one
2342: proportional to $\epsilon_3 \delta v_{{\rm A} r}$  in equation
2343: (\ref{eq:pert_euler_phi_grads}), we would have missed the important
2344: impact that the third term in equation (\ref{eq:k_critical}) has on
2345: the stability of modes with $\omega \rightarrow 0$,  in the limit of
2346: strong toroidal fields (see Fig. \ref{fig:unstable_regions_num}).
2347: This is somewhat counterintuitive because there does not seem to be
2348: any \emph{a priori} indication about which of the magnetic tension
2349: terms (related to the curvature of the toroidal field component) is
2350: less relevant in the original set of equations
2351: (\ref{eq:pert_cont_grads})-(\ref{eq:pert_induc_z_grads}) for the
2352: perturbations. This particular example illustrates the risks
2353: associated with neglecting terms that are not strictly $2^{\rm nd}$
2354: order in the perturbed quantities but rather address the 
2355: geometrical characteristics of the 
2356: background in which the (local) analysis is being carried out.
2357: 
2358: From equation (\ref{eq:k_critical}) it is also straightforward to see
2359: under which circumstances it is safe to neglect the curvature terms.
2360: For subthermal fields, $k_z^0$ will not differ significantly from
2361: $k_{\rm BH}$ (eq.~[\ref{eq:k_BH}]) regardless of the geometry of the
2362: field configuration. This is because, if the field is weak enough
2363: ($v_{{\rm A}} \ll c_{\rm s}$), no matter how strong of a (subthermal)
2364: $B_\phi$ component we consider, the second and third term are
2365: negligible with respect to the first one.  We conclude this short
2366: analysis by commenting that, while a strong vertical field plays a
2367: stabilizing role, in the sense that it drives $k_z^0$ toward small
2368: values leaving all modes with shorter wavelengths stable, the
2369: consequences of considering strong toroidal fields is a little more
2370: subtle as it can be seen in the evolution of the structure of the
2371: modes in Figure \ref{fig:comp_real}.
2372: 
2373: \acknowledgments{We thank Eliot Quataert, Jim Stone, Charles Gammie,
2374: and Wolfgang Duschl for useful comments and discussion in different
2375: stages of this study. We thank an anonymous referee for
2376: pointing out the connection between instability II and the ATB modes.
2377: We are also grateful to Ethan Vishniac for encouraging us to address 
2378: the limit discussed in Appendix A. We also thank the Institute  for Advanced
2379: Study for their hospitality during part of this  investigation.
2380: This work was partially supported by NASA grant NAG-513374}
2381: 
2382: \begin{thebibliography}{}
2383: 
2384: \bibitem[Balbus(2003)]{Balbus03}
2385: {Balbus, S.\ A. 2003, ARA \& A, 41, 555}
2386: 
2387: \bibitem[Balbus \& Hawley(1991)]{BH91}
2388: {Balbus, S.\ A., \& Hawley, J.\ F. 1991, ApJ, 376, 214}
2389: 
2390: \bibitem[Balbus \& Hawley(1992)]{BH92}
2391: {---------. 1992, ApJ, 392, 662}
2392: 
2393: \bibitem[Balbus \& Hawley(1998)]{BH98}
2394: {---------. 1998, Rev. Mod. Phys., 70, 1}
2395: 
2396: \bibitem[Balbus \& Hawley(2002)]{BH02}
2397: {---------. 2002, ApJ, 573, 749}
2398: 
2399: \bibitem[Blaes \& Balbus(1994)]{BB94}
2400: {Blaes, O.\ M., \& Balbus, S.\ A. 1994, ApJ, 421, 163}
2401: 
2402: \bibitem[Blaes \& Socrates(2001)]{BS01}
2403: {Blaes, O.\ M., \& Socrates, A. 2001, ApJ, 553, 987}
2404: 
2405: \bibitem[Bourke \& Goodman(2003)]{BG03}
2406: {Bourke, T.\ L., \&  Goodman, A.\ A. 2003, In Star Formation at High
2407:   Angular Resolution, ASP Conf. Ser., Vol. S-221, eds.
2408:  M.\ G. Burton, R. Jayawardhana, and T.\ L. Bourke (astro-ph/0401281)} 
2409: 
2410: \bibitem[Curry \& Pudritz(1995)]{CP95}
2411: {Curry, C., \& Pudritz, R.\ E. 1995, ApJ, 453, 697}
2412: 
2413: \bibitem[Dubrulle \& Knobloch(1993)]{DK93}
2414: {Dubrulle, B., \& Knobloch, E. 1993, A\&A, 274, 667}
2415: 
2416: \bibitem[Gammie \& Balbus(1994)]{GB94}
2417: {Gammie, C.\ F., \& Balbus, S.\ A. 1994, MNRAS, 270, 138}
2418: 
2419: \bibitem[Hawley, Gammie, \& Balbus(1994)]{HGB94}
2420: {Hawley, J.\ F., Gammie, C.\ F., \& Balbus, S.\ A. 1994, in ASP
2421:   Conf. Ser. 54, The First Stromlo Symposium: The Physics of Active Galaxies. ed.
2422:  G.\ V. Bicknell, M.\ A. Dopita, \& P.J. Quinn (San Francisco:ASP), 53} 
2423: 
2424: \bibitem[Hawley, Gammie, \& Balbus(1995)]{HGB95}
2425: {---------. 1995, ApJ, 440, 742}
2426: 
2427: \bibitem[Hawley, Gammie, \& Balbus(1996)]{HGB96}
2428: {---------. 1996, ApJ, 464, 690}
2429: 
2430: \bibitem[Kim \& Ostriker(2000)]{KO00}
2431: {Kim, W.\ T., \& Ostriker,  E.\ C. 2000, ApJ, 540, 372}
2432: 
2433: \bibitem[Knobloch(1992)]{Knobloch92}
2434: {Knobloch, E. 1992, MNRAS, 255, 25}
2435: 
2436: \bibitem[Kudoh, Matsumoto, \& Shibata(2002)]{KMS02}
2437: {Kudoh, T., Matsumoto, R., \& Shibata, K. 2002, PASJ, 54, 121}
2438: 
2439: \bibitem[Machida, Hayashi, \& Matsumoto(2000)]{MHM00}
2440: {Machida, M., Hayashi, M.\ R., \& Matsumoto R. 2000, ApJ, 532, L67}
2441: 
2442: \bibitem[Miller \& Stone(2000)]{MS00}
2443: {Miller, K.\ A., \& Stone,  J.\ M. 2000, ApJ, 534, 398}
2444: 
2445: \bibitem[Myers \& Goodman(1988)]{MG88}
2446: {Myers, P.\ C., \& Goodman A.\ A. 1988, ApJ, 326, L27}
2447: 
2448: \bibitem[Ostriker, Stone, \& Gammie(2001)]{OSG01}
2449: {Ostriker, E.\ C.,  Stone, J.\ M., \& Gammie, C.\ F. 2001, ApJ, 577, 524}
2450: 
2451: \bibitem[Pariev, Blackman, \& Boldyrev(2003)]{PBB03}
2452: {Pariev, V.\ I., Blackman, E.\ G., \& Boldyrev, S.\ A. 2003, A\&A, 407, 403}
2453: 
2454: \bibitem[Papaloizou \& Szuszkiewicz(1992)]{PSB92}
2455: {Papaloizou J., \& Szuszkiewicz E. 1992, Geophys. Astrophys. Fluid Dyn., 66, 223}
2456: 
2457: \bibitem[Press et al.(1992)]{Press92}
2458: {Press, W.\ H., Flannery, B.\ P., Teukolsky, S.\ A., \& Vetterling, W.\
2459:   T. 1992, Numerical Recipes (Cambridge: University Press)}
2460: 
2461: \bibitem[Pringle(1989)]{Pringle89}
2462: {Pringle, J.\ E. 1989, MNRAS, 236, 107}
2463: 
2464: \bibitem[Quataert, Dorland, \& Hammett(2002)]{Quataert02}
2465: {Quataert, E., Dorland, W., \& Hammett, G.\ W. 2002, ApJ, 577, 524}
2466: 
2467: \bibitem[Sano \& Miyama(1999)]{SM99}
2468: {Sano, T., \& Miyama, S.\ M. 1999, ApJ, 515, 776}
2469: 
2470: \end{thebibliography}
2471: 
2472: \end{document}
2473: 
2474: