1: \documentclass[preprint2]{aastex}

2: 

3: \bibliographystyle{apj}

4: \citestyle{aa}

5: 

6: \usepackage{graphicx}

7: %\graphicspath{{figures/}}

8: 

9: \newcommand{\tensor}[1]{\mathsf{#1}}

10: \newcommand{\text}[1]{\mbox{#1}}

11: \newcommand{\agt}{\ga}

12: \newcommand{\alt}{\la}

13: 

14: \newcommand{\grad}[1]{\nabla #1}

15: \renewcommand{\div}[1]{\nabla \cdot #1}

16: \newcommand{\curl}[1]{\nabla \times #1}

17: 

18: \newcommand{\B}{\vec{B}}

19: \renewcommand{\b}{\hat{b}}

20: \newcommand{\W}{\tensor{W}}

21: \newcommand{\E}{\vec{E}}

22: \newcommand{\J}{\vec{J}}

23: \renewcommand{\v}{\vec{v}}

24: \newcommand{\z}{\hat{z}}

25: \renewcommand{\L}{\tensor{L}}

26: \newcommand{\R}{\tensor{R}}

27: \newcommand{\V}{\vec{V}}

28: \renewcommand{\r}{\vec{r}}

29: 

30: 

31: 

32: 

33: \begin{document}

34: 

35: \title{Finite Larmor Radius Effects on the Magnetorotational

36:   Instability}

37: \author{Nathaniel M. Ferraro} 

38: \affil{Princeton Plasma Physics Laboratory, Princeton, NJ, 08543-0451}

39: 

40: \begin{abstract}

41:   The linear dispersion relation for the magnetorotational instability

42:   (MRI) is derived including finite Larmor radius (FLR) effects.  In

43:   particular, the Braginskii form of the ion gyroviscosity, which

44:   represents the first-order FLR corrections to the two-fluid

45:   equations, is retained.  It is shown that FLR effects are the most

46:   important effects in the limit of weak magnetic fields, and are much

47:   more important than the Hall effect when $\beta_i \gg 1$, where

48:   $\beta_i$ is the ratio of the ion thermal pressure to the magnetic

49:   pressure.  FLR effects may completely stabilize even MRI modes

50:   having wavelengths much greater than the ion Larmor radius.  Some

51:   implications for astrophysical accretion disks are discussed.

52: \end{abstract}

53: 

54: \keywords{accretion, accretion disks---instabilities---MHD---plasmas}

55: 

56: \maketitle 

57: 

58: \section{Introduction}

59: 

60: The magnetorotational instability is a local instability which may be

61: present in accretion disks having sheared azimuthal flow and a weak

62: magnetic field \citep{Balbus91}. Turbulence resulting from this

63: instability is thought to play an important role in the radial

64: transport of angular momentum in such systems \citep{Balbus98}.  Here

65: we explore the modifications to the local, two-fluid theory of the MRI

66: in the linear regime due to finite Larmor radius (FLR) effects.

67: 

68: \cite{Rosenbluth62} have shown using kinetic theory that FLR effects

69: can be stabilizing to ``weakly unstable'' modes---defined as modes

70: having a linear growth rate much smaller than the ion cyclotron

71: frequency---even when the mode wavelength is much larger than the ion

72: Larmor radius.  It was later shown by \cite{Roberts62} that this

73: result could be obtained from fluid theory by retaining the

74: gyroviscous stress component of the ion pressure tensor.  In most

75: physical scenarios the MRI is weakly unstable in the sense of

76: Rosenbluth \textit{et al.}, and indeed we show that the MRI may be

77: completely stabilized by gyroviscous effects at scales much larger

78: than the ion Larmor radius.  In some cases this stabilization

79: significantly constrains the spectrum of linearly unstable modes.

80: 

81: The effect of the Hall term, which accounts for differences between

82: the electron and ion fluid velocities, has been examined previously by

83: \cite{Wardle99}, \cite{Balbus01}, \cite{Salmeron03}, and

84: \cite{Krolik06}.  In particular, it was found that the Hall effect may

85: be either stabilizing or destabilizing, depending on whether the

86: equilibrium magnetic field is aligned or anti-aligned to the

87: equilibrium angular velocity.  It was also found that the Hall effect

88: is important only when the ion cyclotron frequency is comparable to,

89: or smaller than, the orbital frequency.  This situation may occur in

90: early galaxy formation where the magnetic fields are still weak, or in

91: weakly ionized protostellar disks.  \cite{Krolik06} have suggested

92: that, in this limit, short-wavelength modes are likely suppressed by

93: viscous or resistive damping, leaving only slowly growing,

94: long-wavelength modes as the magnetic field get sufficiently weak.

95: However, their analysis is restricted to low-$\beta_i$ plasmas as they

96: do not consider FLR effects, which we show to be much more important

97: than the Hall effect in the weak-field limit.  The strong FLR

98: stabilization of the MRI in the weak-field limit may have important

99: implications for the possible role of the MRI in the amplification of

100: weak, primordial magnetic fields.

101: 

102: The gyroviscous stress is defined as the traceless, perpendicular part

103: of the ion stress tensor which does not depend explicitly on the

104: collision frequency \citep{Ramos05}.  In typical cases, this stress

105: arises primarily from variations in particle drift velocities over the

106: scale of a Larmor orbit \citep{Kaufman60}.  However, other effects may

107: contribute to this stress, including gradients in heat fluxes.

108: Braginskii's form of the ion gyroviscous stress is appropriate for

109: collisional plasmas (in the sense that the ion mean-free-path is small

110: compared to the hydrodynamic perturbation length-scale), and in the

111: limit where the ion cyclotron period is short compared to collisional

112: and hydrodynamic time-scales \citep{Braginskii65}.  More general, but

113: more complicated, expressions for the gyroviscous force have been

114: derived, which are applicable to a broader range of collisionality

115: regimes and dynamical time scales; \cite{Ramos05} and references

116: therein provide derivations and discussions of these alternate forms.

117: We choose to work with Braginskii's form here because it is

118: (relatively) simple and applies to a broad range of astrophysical

119: objects.

120: 

121: The MRI in the collisionless regime, where the collisional

122: mean-free-path is greater than the mode wavelength, has been explored

123: by \cite{Quataert02} and \cite{Sharma03} using kinetic closures.

124: \cite{Islam05} have extended the single-fluid MHD treatment to lower

125: collisionality regimes by including the \cite{Braginskii65} form of

126: the parallel viscosity, and have obtained results similar to those

127: obtained using kinetic closure.  These various analyses have found the

128: linear growth rate to be enhanced by a factor of order unity at lower

129: collisionality when an azimuthal component of the magnetic field is

130: present, but the criterion for instability was found not to differ

131: from the MHD result.  We do include the parallel viscous stress for

132: completeness in our analysis, as formally it may be larger than the

133: gyroviscous stress. However, for the sake of simplicity we restrict

134: the MRI mode wavevector and magnetic field to be normal to the

135: accretion disk, in which case the parallel viscosity has no effect on

136: the MRI.  This case is the most unstable one in the collisional limit,

137: which is the limit in which we are mainly interested.

138: 

139: 

140: \section{Linear Theory}

141: 

142: \subsection{Model}

143: 

144: We consider the two-fluid MHD equations:

145: \begin{mathletters}

146:   \label{eq:model}

147: \begin{eqnarray}

148:   \frac{\partial n}{\partial t} & = & -\div{(n \v)} 

149:   \\ 

150:   n \frac{\partial \v}{\partial t} & = & -n \v \cdot \grad{\v} + 

151:   \frac{\J \times \B}{c}  - \mbox{} \\

152:   && \mbox{} - \grad{p} - \div{\tensor{\Pi}} - n g(r) \hat{r}

153:   \\

154:   \frac{\partial \B}{\partial t} & = & -c \curl{\E}

155: \end{eqnarray}

156: \end{mathletters}

157: where

158: \begin{eqnarray*}

159:   & \E = -\frac{1}{c} \v \times \B + \frac{1}{n e} 

160:   (\frac{1}{c}\J \times \B - \nabla p_e) 

161:   \\ 

162:   & \J = \frac{c}{4\pi} \curl{\B};\quad p = p_i + p_e. &

163: \end{eqnarray*}

164: The terms representing two-fluid effects are the Hall term ($\J \times

165: \B/n e c$) and the electron pressure gradient in the definition of $\E$.

166: We assume barytropic pressure variations of the form

167: \begin{equation}

168:   dp_s = \Gamma T_s dn

169: \end{equation}

170: for each species $s$ where, for example, $\Gamma=5/3$ for an adiabatic

171: equation of state.  For $\tensor{\Pi}$ we use the leading order terms

172: in the \cite{Braginskii65} closure:

173: \begin{displaymath}

174:   \tensor{\Pi} = \tensor{\Pi}^{v} + \tensor{\Pi}^{gv},

175: \end{displaymath}

176: where

177: \begin{eqnarray}

178:   \tensor{\Pi}^{v} & = & \eta_0 \frac{p_i}{2 \nu_i} 

179:     \left( \tensor{I} - 3 \b \b \right)

180:     \left( \b\cdot\W\cdot\b \right)

181:   \\

182:   \label{eq:gyroviscosity}

183:   \tensor{\Pi}^{gv} & = & \frac{p_i}{4 \omega_{c i}} \left\{

184:     \b \times \W \cdot (\tensor{I} + 3 \b\b) + \mbox{} \right. \\ 

185:     & & \left. \mbox{} + 

186:     \left[\b \times \W \cdot (\tensor{I} + 3 \b\b)\right]^\top \right\}

187: \end{eqnarray}

188: where $\b = \B/B$, $\omega_{c i} = e B_{z 0}/m_i c$ is the ion

189: cyclotron frequency, $\nu_i$ is the ion collision frequency, and the

190: coefficient $\eta_0 \approx 0.96$ is a factor determined by kinetic

191: analysis \citep{Braginskii65}.  The rate-of-strain tensor is

192: \begin{displaymath}

193:   \W = \nabla \v + (\nabla \v)^\top - \frac{2}{3} \tensor{I} \div{\v}.

194: \end{displaymath}

195: $\tensor{\Pi}^{v}$ is the parallel viscosity term considered by

196: \cite{Balbus04} and \cite{Islam05}, which represents the lowest-order

197: in $\partial_t/\nu_i$ correction to the fluid equations.  (It will be

198: shown, however, that the parallel viscosity plays no role in the MRI

199: in the configuration we choose to examine.)  $\tensor{\Pi}^{gv}$ is

200: the gyroviscous force, which represents the lowest-order FLR

201: correction to the fluid equations.  This form of $\tensor{\Pi}$ is

202: appropriate in the limit where $\omega_{c i} \gg \nu_i$.  Together

203: with the short mean-free-path condition ($k_\parallel \lambda_{mfp}

204: \ll 1$, where $\lambda_{mfp} \sim v_t / \nu_i$) necessary to justify

205: the fluid closure, this restricts the validity of our results to the

206: parameter regime where

207: \begin{equation}

208: \label{eq:validity}

209: k_\parallel v_t \ll \nu_i \ll \omega_{c i}.

210: \end{equation}

211: In the case of an accretion disk where $k_\parallel \agt 1/d$ is

212: limited by the disk height $d \sim v_t/\Omega$, this validity

213: condition becomes $\Omega \ll \nu_i \ll \omega_{c i}$.

214: 

215: \subsection{Equilibrium}

216: 

217: We consider a cylindrical equilibrium, using the standard coordinates

218: $(r,\theta,z)$, and assume axisymmetry in the azimuthal coordinate

219: ($\partial_\theta = 0$).  For simplicity, we assume the equilibrium

220: magnetic field is of the form $\B_0 = B_{z 0} \hat{z}$, and that the

221: equilibrium fluid velocity is of the form $\v_0 = r \Omega(r)

222: \hat{\theta}$.  Without loss of generality, we orient our coordinate

223: system so that $\Omega(r) > 0$ at the radius of interest.  For such a

224: configuration, radial force balance is satisfied when

225: \begin{equation}

226:   \label{eq:radial_force_balance}

227:   g(r) = r \Omega^2(r) +

228:   \frac{\rho_i^2 \omega_{c i}}{2 r^2}\frac{\partial}{\partial r}

229:   \left[r^3 \Omega'(r)\right],

230: \end{equation}

231: where $\rho_i = v_t/\omega_{c i}$ is the ion Larmor radius, and $v_t =

232: \sqrt{T_i/m_i}$ is the ion thermal velocity.  We neglect any

233: equilibrium structure in the $z$-direction.  In the following

234: derivation, we choose to use equation~(\ref{eq:radial_force_balance})

235: to eliminate $g(r)$ in favor of $\Omega(r)$ (hereafter we will drop

236: the explicit dependence of $\Omega$ on $r$).

237: 

238: 

239: \subsection{Local Linear Dispersion Relation}

240: 

241: \begin{figure*}[t]

242:   \includegraphics[width=2.8in]{f1a.eps}

243:   \includegraphics[width=2.8in]{f1b.eps}

244:   \includegraphics[width=0.6in]{f1c.eps}

245:   \caption{\label{fig:MRI_gamma} Contours of the growth rate of the

246:   unstable solution to equation~(\ref{eq:dispersion_relation}) are

247:   plotted versus $k v_A/\Omega$.  \emph{Left:} The growth rate is

248:   plotted for various values of $F$, which measures the importance of

249:   FLR effects, in the limit where $\Omega/\omega_{c i} \ll 1$ (no Hall

250:   effect).  $F = 0$ is the ideal MHD result.  \emph{Right:} The growth

251:   rate is plotted for various values of $\Omega/\omega_{c i}$, which

252:   measures the importance of the Hall effect, in the limit where $F

253:   \ll 1$ (no FLR effects).  $\Omega/\omega_{c i}=0$ is the ideal MHD

254:   result.}

255: \end{figure*}

256: 

257: We consider linear perturbations about this equilibrium having scale

258: lengths $1/k$ much smaller than the equilibrium flow gradient scale

259: length $L = \Omega/\Omega'$, so that $\delta = 1/ |k L| \ll 1$.  In

260: this limit, we may assume that the normal modes of the system are

261: plane waves to lowest order in $\delta$.  We restrict our analysis to

262: perturbations $\propto e^{i(k z - \omega t)}$ for simplicity, as this

263: is the most unstable case in both the MHD and collisionless limits

264: \citep{Quataert02}.  Carrying out the linearization of

265: equations~(\ref{eq:model}) yields the following dispersion relation,

266: to lowest order in $\delta$:

267: \begin{eqnarray}

268:   0 & = & (W^2 + W \frac{8 i \eta_0}{3} \frac{\omega_{c i}}{\nu_i}A - C)

269:   \times \mbox{} \nonumber \\ & & \mbox{} \times

270:   \left\{ W^4 + a_2 W^2 + a_0 \right\} \label{eq:dispersion_relation}

271: \end{eqnarray}

272: where

273: \begin{eqnarray*}

274:   a_2 & = & -2(2+R+K) + \mbox{} \\

275:   && \mbox{} + 2 A (4+R-2A) - H(R+H)

276:   \\

277:   a_0 & = & \left[K - A(R+2H) + 2(R+H) \right] \times \mbox{} \\

278:    && \mbox{} \times \left[K - A(R+2H) + H(2+R) \right].

279: \end{eqnarray*}

280: We define the dimensionless quantities

281: \begin{mathletters}

282: \begin{eqnarray}

283:   W &=& \omega/\Omega\\

284:   K &=& (v_A k/\Omega)^2\\

285:   A &=& K \beta_i (\Omega/\omega_{c i})/4 \equiv K F \label{eq:def_A}\\

286:   H &=& K (\Omega/\omega_{c i})\\

287:   C &=& (c_s k/\Omega)^2\\

288:   R &=& r \Omega'/\Omega

289: \end{eqnarray}

290: \end{mathletters}

291: and characteristic velocities $v_A^2 = B_{z 0}^2/4\pi n_0 m_i$ and

292: $c_s^2 = \Gamma (T_e + T_i)/m_i$.  Here $\beta_i = 8 \pi p_{i

293: 0}/B_0^2$ is the ratio of ion thermal pressure to magnetic pressure.

294: The dimensionless parameter $A$ measures the importance of the

295: gyroviscous force, and setting $A = 0$ is equivalent to omitting $\div

296: \tensor{\Pi}^{gv}$ in the ion force equation.  Similarly, $H$ measures

297: the importance of the Hall term in Ohm's law, and $K$ measures the

298: importance of magnetic tension.  $R$ is the ratio of the radial

299: coordinate to the equilibrium flow gradient scale length, and is taken

300: to be $\sim \mathcal{O}(1)$.  For a Keplerian disk, $R = -3/2$.  For

301: convenience we have also defined

302: \begin{displaymath} 

303:   F = \frac{\beta_i}{4} \frac{\Omega}{\omega_{c i}},

304: \end{displaymath}

305: which is the ratio of the gyroviscous force to the magnetic tension

306: force.

307: 

308: Equation~(\ref{eq:dispersion_relation}) contains two uncoupled modes.

309: The first factor contains the acoustic mode, which may be damped by

310: the parallel viscosity when $A \ne 0$, and is not of interest here.

311: (While $\omega_{c i}/\nu_i$ is generally large, $A \omega_{c i}/\nu_i

312: \sim C \Omega/\nu_i \ll C$ in the collisional regime, so the effect on

313: the acoustic mode is small.)  The second factor, enclosed in braces,

314: contains the MRI.  In the limit where $A \to 0$ (no FLR effects), $H

315: \to 0$ (no Hall effect), and $\Omega/\nu_i \to 0$ (collisional

316: regime), the dispersion relation of \cite{Balbus98} is recovered.

317: Note that the parallel viscosity ($\propto \eta_0$) affects only the

318: acoustic mode and not the MRI.  Evidently, there is no

319: $\mathcal{O}(\Omega/\nu_i)$ correction to the MHD result for the MRI

320: when $\B_0 = B_{z 0} \z$, which is in agreement with the findings of

321: \cite{Sharma03} and \cite{Islam05}.  Here we are interested in

322: corrections to the collisional mode, for which the $\B_0 = B_{z 0} \z$

323: case is the most unstable.  Extending this analysis to a more general

324: magnetic field configuration substantially complicates the analysis.

325: 

326: It should also be noted that $\omega_{c i}$ is a signed quantity since

327: it is proportional to $B_z$, which may be positive or negative.  Since

328: we have chosen the coordinate system so that $\Omega$ is positive,

329: $\text{sign }\omega_{c i} = \text{sign }\B \cdot \vec{\Omega} =

330: \text{sign } F$.  It has been shown previously that the effect of the

331: Hall term on the MRI depends strongly on the sign relative signs of

332: $\omega_{c i}$ and $\Omega$ \citep{Wardle99}.  The effect of the

333: gyroviscous force has a similar dependence.

334: 

335: The growth rate $\gamma = \text{Im }\omega$ of the unstable solution

336: to equation~(\ref{eq:dispersion_relation}) is plotted in

337: figure~\ref{fig:MRI_gamma} for a Keplerian rotation profile

338: ($R=-3/2$).  Note that the abscissa should be read as a normalized

339: wavenumber and not a normalized magnetic field strength, because $F$

340: and $\Omega/\omega_{c i}$ are dependent on $B$.  When $\omega_{c i} >

341: 0$, and hence $F>0$ also, both the FLR and the Hall effects can be

342: seen to move the most unstable mode to lower wavenumbers, and to

343: reduce the value of $K$ at which the MRI is completely stabilized.

344: Also, FLR effects increase the growth rate of the most unstable mode.

345: When $\omega_{c i} < 0$, and hence $F<0$, both effects are seen

346: instead to increase the cutoff value of $K$ all the way to the point

347: where modes of any wavelength for which this analysis is valid are

348: unstable.  Presumably, the inclusion of a finite resistivity would

349: damp this resonance, as it does in the case of the resonance due to

350: the Hall effect \citep{Balbus01}.  When $F$ or $\Omega/\omega_{c i}$

351: becomes sufficiently negative ($F < -2/3$ when $\Omega/\omega_{c i}

352: \to 0$, or $\Omega/\omega_{c i} < -2$ when $F \to 0$), all values of

353: $k$ are suddenly completely stabilized.  (It has been shown by

354: \cite{Balbus01} that this stabilization is less sudden when finite

355: resistivity is included.)

356: 

357: \subsection{Instability Criterion}

358: 

359: Applying the Routh-Hurwitz theorem to

360: equation~(\ref{eq:dispersion_relation}), we find that the condition

361: for stability of an MRI mode is that:

362: \begin{mathletters}

363: \label{eq:stability_criterion}

364: \begin{eqnarray*}

365:   a_2 < 0 \mbox{ and } a_0 > 0.

366: %  0 & < & 2(2+R+K) - 2 A (4+R-2A) + H(R+H)  \\

367: %  0 & < & \left[K - A(R+2H) + 2(R+H) \right]\left[K - A(R+2H) + H(2+R) \right]

368: \end{eqnarray*}

369: \end{mathletters}

370: This criterion is highly complicated, and for general values of $A$

371: and $H$, there may be multiple stable and unstable regions in

372: $K$-space.

373: 

374: In the ideal limit, when $A \to 0$ and $H \to 0$, the instability

375: criterion of \cite{Balbus98}, $K < -2 R$, is recovered.  This limit is

376: well understood, and in this case stabilization at high-$K$ is due to

377: the effect of magnetic tension.  In this limit, instability does not

378: exist in flows in which the angular velocity increases with radius

379: ($R>0$).

380: 

381: The limit $A \to 0$, in which the Hall effect is dominant over the FLR

382: effects, has also been considered before.  In this limit,

383: equation~(\ref{eq:dispersion_relation}) reduces to the dispersion

384: relation of \cite{Balbus01}.  Since $A/H = \beta_i/4$, this limit

385: describes accretion disks having $\beta_i \ll 1$.  Formally, the

386: instability criterion in this case remains somewhat complicated

387: because the signs and relative magnitudes of most of the terms are

388: undetermined in general.  There is some discussion of the instability

389: criterion in this case by \cite{Wardle99} and \cite{Balbus01}, as well

390: as insight into its physical meaning.  We will not repeat this

391: discussion, except to mention a few interesting points.  The first is

392: that there may exist some values of $\Omega/\omega_{c i}$ for which

393: modes of any wavelength are unstable (this is true in the Keplerian

394: case for $-2 < \Omega/\omega_{c i} < -1/2$).  Also, some unstable

395: modes may be present in disks in which angular velocity increases with

396: radius ($R > 0$), in contrast to the ideal result \citep{Balbus01}.

397: 

398: We are more interested in the opposite limit, $\beta_i \gg 1$, in

399: which FLR effects are dominant over the Hall effect.  Taking $H \to

400: 0$, the dispersion relation for the MRI reduces to

401: \begin{eqnarray}

402: \label{eq:flr_dispersion_relation}

403: 0 & = & W^4 - \mbox{} \\

404:  && \mbox{} - 2 \left[ 2+R+K - A (4+R-2A) \right] W^2 + \mbox{} \nonumber \\

405:  && \mbox{} + \left(K- A R + 2 R \right) \left(K - A R \right) \nonumber

406: \end{eqnarray}

407: and the criterion for instability is found to be

408: \begin{equation}

409:   K  < -\frac{2 R}{1 - F R}.

410: \end{equation}

411: For the usual case where $R < 0$, all modes are completely stabilized

412: if $F < -1/|R|$.  As with the Hall effect, gyroviscosity allows

413: unstable modes to exist when $R > 0$; in this case, unstable modes may

414: exist when $F>1/R$.  In the limit where the gyroviscous force

415: dominates the force of magnetic tension, $F \gg 1$, the instability

416: criterion becomes simply

417: \begin{equation}

418:   \label{eq:FLR_instability_criterion}

419:   A < 2 \mbox{ and } F > 0

420: \end{equation}

421: or, equivalently, $(k \rho_i)^2 < 4 \Omega/\omega_{c i}$, In the case

422: where $F<0$, there are no unstable modes in this limit.

423: 

424: \section{Discussion and Conclusions}

425: 

426: The gyroviscous stress arises from changes to the guiding-center

427: drifts due to the gradients of magnetohydrodynamic forces (the

428: electric field, in this case) across a gyro-orbit.  A more complete

429: and quantitative discussion of this effect is discussed by

430: \cite{Kaufman60}.  Due to their larger Larmor radius, the ions are

431: more affected by this modification that the electrons, leading to the

432: generation of currents.  Essentially, the stabilization (or

433: destabilization) of the MRI by FLR effects is due to the currents

434: generated by the gyroviscous force being out of phase (or in phase)

435: with the current of the ideal-MHD MRI eigenmode.

436: 

437: The effect of the gyroviscous stress becomes relatively more important

438: to the behavior of the MRI as the magnetic field strength is

439: decreased.  The relative importance of this stress to the magnetic

440: tension is $A/K = F \propto B^{-3}$, and the relative importance of

441: the gyroviscous stress to the Hall effect is $A/H = \beta_i/4 \propto

442: B^{-2}$.  Note that this means analyses of the MRI in the

443: ``weak-field'' limit (in the sense that $\Omega/\omega_{c i} \ga 1$)

444: which did not consider FLR effects are valid only for $\beta_i \ll 1$.

445: If the magnetic field is sufficiently weak, the validity condition for

446: the Braginskii closure, equation~(\ref{eq:validity}), may be violated

447: since $k v_t / \omega_{c i} = \sqrt{2 H F} \propto B^{-1}$, in which

448: case the above analysis no longer strictly applies.  Of course,

449: gyroviscosity does not shut off at this point; indeed, FLR effects are

450: expected to become increasingly powerful as the magnetic field is

451: decreased further, though not in a way that is correctly described by

452: equation~({\ref{eq:gyroviscosity}).  Therefore it is probable that the

453: low-$K$ modes which survive at the lowest magnetic field strengths

454: covered by this analysis will be completely stabilized as the magnetic

455: field decreases further.  This is the proper resolution to the

456: inconsistency of the ideal MHD result than the MRI remains unstable as

457: $B \to 0$ in the non-dissipative case.  Because the MRI is not present

458: at arbitrarily low magnetic fields, its role in the amplification of

459: primordial astrophysical magnetic fields is severely restricted.

460: 

461: It should also be noted that FLR effects may completely stabilize MRI

462: modes having wavelengths much greater than the ion Larmor radius,

463: where $k \rho_i \ll 1$.  As was shown by \cite{Rosenbluth62}, this is

464: possible for ``weakly unstable'' modes like the MRI.  Restricting our

465: analysis to the FLR-dominated limit ($F \gg 1$), MRI modes are

466: stabilized when $(k \rho_i)^2 > 4 (\Omega/\omega_{c i})$, where

467: $\Omega/\omega_{c i}$ is typically small in astrophysical accretion

468: disks.  This stabilization may be made more important by the fact that

469: in an accretion disk the lower bound on $k$ is set by the height of the

470: disk $d$, which may be much smaller than the equilibrium flow gradient

471: scale length $L$.  Although a proper understanding of this phenomenon

472: must take into account the $z$-stratification of the disk equilibrium,

473: we may estimate that the smallest wavenumber present in the disk is

474: $\sim \pi/d$.  The criterion for complete stabilization by FLR effects

475: of \emph{all} MRI modes within an accretion disk at some distance from

476: the central mass is then

477: \begin{equation}

478:   \pi^2 (\rho_i / d)^2 \agt 4 \Omega/\omega_{c i}.

479: \end{equation}

480: In the typical case where $d \sim v_t/\Omega$ this inequality reduces

481: to $\omega_{c i}/\Omega \alt \pi^2/4$.  While this is not typically

482: satisfied in astrophysical accretion disks, it may be satisfied in

483: nascent galaxies with weak magnetic fields, or weakly ionized

484: protostellar disks \citep{Krolik06}.  It is important to recall here

485: that the condition for validity of our analysis for wavelengths of

486: this scale requires $\Omega \ll \omega_{c i}$; whether FLR effects are

487: more or less stabilizing than our result would predict outside this

488: range of validity is a matter for further research.

489: 

490: \begin{figure}[t]

491:   \begin{center}

492:     \includegraphics[width=3in]{f2.eps}

493:   \end{center}

494:   \caption{\label{fig:galactic_MRI} The growth rate of the MRI is

495:     plotted versus wavelength for a Keplerian, $10^7 M_\sun$ active

496:     galactic nucleus in a weak magnetic field ($B = 10$ nG), at a

497:     distance $10^{16}$ cm.  The solutions to the dispersion relation

498:     including (solid line), and excluding (broken line) the FLR

499:     correction are plotted.  Gyroviscosity completely stabilizes the

500:     shorter-wavelength modes, and enhances the longer-wavelength

501:     modes.}

502: \end{figure}

503: 

504: For a concreteness, we consider the example of an active galactic

505: nucleus (AGN) in a weak magnetic field, for which the conditions for

506: validity of the Braginskii equations are met, and in which the effect

507: of the gyroviscous force is particularly strong.  We assume Keplerian

508: rotation ($R = -3/2$), a central mass of $10^7 M_\sun$, $B=10$ nG, $n

509: = 10$ cm$^{-3}$, and $T_i = 1$ eV, at a distance of $10^{16}$ cm.  For

510: this case, the relevant dimensionless parameters are $F \approx 384$,

511: $H \approx 3.81\times10^{-4}$, and $K \approx 1.96 \times 10^{-5}$ at

512: $k = \Omega/v_t$.  The growth rates of the MRI under these conditions

513: are plotted in figure~\ref{fig:galactic_MRI}.  From that figure it can

514: be seen that in both cases the FLR correction increases the wavelength

515: of the fastest-growing mode from $\approx 10^{11}$ cm to $\approx

516: 25\times 10^{11}$ cm, and similarly increases the cutoff wavelength

517: below which there are no unstable modes.  In this example, the

518: weakness of the magnetic field is crucial to the importance of the FLR

519: effect.  If the magnetic field strength is raised by two orders of

520: magnitude, the FLR correction becomes insignificant.  Therefore,

521: ``primordial'' accretion disks in weak magnetic fields will be most

522: affected by the FLR correction, whereas disks in stronger fields ($B

523: \ge 10$ $\mu$G) may be completely unaffected, unless they are

524: unusually hot and dense, or have an orbital frequency not much smaller

525: than the ion cyclotron frequency.

526: 

527: The nonlinear saturation of the MRI cannot be properly addressed by

528: linear analysis.  Because gyroviscosity is not dissipative, and there

529: is no energy associated with the gyroviscous term, one might expect

530: that the ultimate nonlinear saturated state should not be affected

531: strongly by the gyroviscosity.  Ultimately, questions of nonlinear

532: saturation should be addressed by numerical simulation.

533: 

534: 

535: \acknowledgments{The author is grateful for helpful discussions with

536:   Dr. Gregory Hammett and Ian Parrish.}

537: 

538: %\bibliography{ms}

539: 

540: \begin{thebibliography}{14}

541: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi

542: 

543: \bibitem[{Balbus(2004)}]{Balbus04}

544: Balbus, S.~A. 2004, ApJ, 616, 857

545: 

546: \bibitem[{Balbus \& Hawley(1991)}]{Balbus91}

547: Balbus, S.~A. \& Hawley, J.~F. 1991, ApJ, 376, 214

548: 

549: \bibitem[{Balbus \& Hawley(1998)}]{Balbus98}

550: ---. 1998, Rev. Mod. Phys., 70, 1

551: 

552: \bibitem[{Balbus \& Terquem(2001)}]{Balbus01}

553: Balbus, S.~A. \& Terquem, C. 2001, ApJ, 552, 235

554: 

555: \bibitem[{Braginskii(1965)}]{Braginskii65}

556: Braginskii, S.~I. 1965, in Reviews of Plasma Physics, ed. M.~A. Leontovich,

557:   Vol.~1 (New York: Consultants Bureau), 205--311

558: 

559: \bibitem[{Islam \& Balbus(2005)}]{Islam05}

560: Islam, T. \& Balbus, S. 2005, ApJ, 633, 328

561: 

562: \bibitem[{Kaufman(1960)}]{Kaufman60}

563: Kaufman, A.~N. 1960, Phys. Fluids, 3, 610

564: 

565: \bibitem[{Krolik \& Zweibel(2006)}]{Krolik06}

566: Krolik, J.~H. \& Zweibel, E.~G. 2006, ApJ, 644, 651

567: 

568: \bibitem[{Quataert {et~al.}(2002)Quataert, Dorland, \& Hammett}]{Quataert02}

569: Quataert, E., Dorland, W., \& Hammett, G. 2002, ApJ, 577, 524

570: 

571: \bibitem[{Ramos(2005)}]{Ramos05}

572: Ramos, J.~J. 2005, Phys. Plasmas, 12, 112301

573: 

574: \bibitem[{Roberts \& Taylor(1962)}]{Roberts62}

575: Roberts, K.~V. \& Taylor, J.~B. 1962, Phys. Rev. Lett., 8, 197

576: 

577: \bibitem[{Rosenbluth {et~al.}(1962)Rosenbluth, Krall, \&

578:   Rostoker}]{Rosenbluth62}

579: Rosenbluth, M.~N., Krall, N.~A., \& Rostoker, N. 1962, Nucl. Fusion Suppl., 1,

580:   143

581: 

582: \bibitem[{Salmeron \& Wardle(2003)}]{Salmeron03}

583: Salmeron, R. \& Wardle, M. 2003, MNRAS, 345, 992

584: 

585: \bibitem[{Sharma {et~al.}(2003)Sharma, Hammett, \& Quataert}]{Sharma03}

586: Sharma, P., Hammett, G.~W., \& Quataert, E. 2003, ApJ, 596, 1121

587: 

588: \bibitem[{Wardle(1999)}]{Wardle99}

589: Wardle, M. 1999, MNRAS, 307, 849

590: 

591: \end{thebibliography}

592: 

593: 

594: \end{document}