1: %\documentclass[11pt,preprint]{aastex}

2: %\documentclass{aastex}

3: %\usepackage{emulateapj5,epsfig}

4: \documentclass[useAMS, usenatbib, onecolumn]{mn2e}

5: \usepackage{epsfig}

6: \usepackage{amsmath}

7: %\usepackage{setspace}

8: %\doublespacing

9: \voffset -1.3cm

10: \def\be{\begin{equation}}

11: \def\ee{\end{equation}}

12: \def\ba{\begin{eqnarray}}

13: \def\ea{\end{eqnarray}}

14: \def\go{\mathrel{\raise.3ex\hbox{$>$}\mkern-14mu

15:              \lower0.6ex\hbox{$\sim$}}}

16: \def\lo{\mathrel{\raise.3ex\hbox{$<$}\mkern-14mu

17:              \lower0.6ex\hbox{$\sim$}}}

18: \def\bB{{\bf B}}

19: \def\bw{{\bf w}}

20: \def\bv{{\bf v}}

21: \def\bu{{\bf u}}

22: \def\br{{\bf r}}

23: \def\hate{{\hat{\bf e}}}

24: \def\bfe{{\bf e}}

25: \def\cB{{\cal B}}

26: \def\cA{{\cal A}}

27: \def\cR{{\cal R}}

28: \def\d{\partial}

29: 

30: \def\barR{{\bar R}}

31: \def\barz{{\bar z}}

32: \def\tomega{\tilde\omega}

33: \def\tomega{{\tilde{\omega}}}

34: 

35: \begin{document}

36: \title[Modes in Black Hole Accretion Discs]

37: {Corotational Instability of Inertial-Acoustic Modes in Black Hole

38: Accretion Discs and Quasi-Periodic Oscillations}

39: \author[D. Lai and D. Tsang]

40: {Dong Lai$^{1}$\thanks{Email:

41: dong@astro.cornell.edu; dtsang@astro.cornell.edu} 

42: and David Tsang$^{1,2}$\footnotemark[1] \\ 

43: $^1$Department of Astronomy, Cornell University, Ithaca, NY 14853, USA \\

44: $^2$Department of Physics,

45: Cornell University, Ithaca, NY 14853, USA \\}

46: 

47: \label{firstpage}

48: \maketitle

49: 

50: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

51: \begin{abstract}

52: We study the global stability of non-axisymmetric p-modes

53: (also called inertial-acoustic modes) trapped in the inner-most regions

54: of accretion discs around black holes.  We show that the lowest-order

55: (highest-frequency) p-modes, with frequencies $\omega=(0.5-0.7)

56: m\Omega_{\rm ISCO}$ (where $m=1,2,3,\cdots$ is the azimuthal wave

57: number, $\Omega_{\rm ISCO}$ is the disc rotation frequency at the

58: Inner-most Stable Circular Orbit, ISCO), can be overstable due to 

59: general relativistic effects, according to which the radial epicyclic

60: frequency $\kappa$ is a non-monotonic function of radius near the black

61: hole. The mode is trapped inside the corotation resonance radius $r_c$ 

62: (where the wave pattern rotation speed $\omega/m$

63: equals the disc rotation rate $\Omega$) and carries a negative energy. 

64: The mode growth arises primarily from wave absorption at the

65: corotation resonance, and the sign of the wave

66: absorption depends on the gradient of the disc vortensity,

67: $\zeta=\kappa^2/(2\Omega\Sigma)$ (where $\Sigma$ is the surface

68: density). When the mode frequency $\omega$ is

69: sufficiently high, such that $d\zeta/r>0$ at $r_c$,

70: positive wave energy is absorbed at the corotation, leading to the

71: growth of mode amplitude. The mode growth is further enhanced by

72: wave transmission beyond the corotation barrier.  We also study how

73: the rapid radial inflow at the inner edge of the disc affects the mode

74: trapping and growth. Our analysis of the behavior of the fluid perturbations

75: in the transonic flow near the ISCO indicates that, while

76: the inflow tends to damp the mode, the damping effect is sufficiently

77: small under some conditions (e.g., when the disc density decreases

78: rapidly with decreasing radius at the sonic point) so that net mode growth can 

79: still be achieved. We further clarify the role of the Rossby wave

80: instability and show that it does not operate for black hole accretion

81: discs with smooth-varying vortensity profiles.

82: Overstable non-axisymmetric p-modes driven by the

83: corotational instability provide a plausible explanation for the

84: high-frequency ($\go 100$~Hz) quasi-periodic oscillations (HFQPOs) 

85: observed from a number of black-hole X-ray binaries in the very high 

86: state. The absence of HFQPOs in the soft (thermal) state may result from 

87: mode damping due to the radial infall at the ISCO.

88: \end{abstract}

89: 

90: \begin{keywords}

91: accretion, accretion discs -- hydrodynamics -- waves -- 

92: -- black hole physics -- X-rays: binaries

93: \end{keywords}

94: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

95: \section{Introduction}

96: 

97: \subsection{Models of Quasi-Periodic X-Ray Oscillations: A Brief Review}

98: 

99: Rapid X-ray variabilities from Galactic compact binary systems have

100: been studied for decades (e.g. van del Klis 2006).  In recent years,

101: our knowledge of quasi-periodic oscillations (QPOs) in black-hole

102: X-ray binaries has greatly improved (see Remillard \& McClintock 2006

103: for a review), thanks in large part to NASA's {\it Rossi X-ray Timing

104: Explorer} (Swank 1999). The low-frequency QPOs (about 0.1-50~Hz) are

105: common, observable when the systems are in the hard state and the

106: steep power-law state (also called ``very high state''; see Done,

107: Gierlinski \& Kubota 2007), and they typically have high amplitudes

108: and high coherence ($Q>10$), and can vary in frequency on short

109: timescales (minutes).  However, it is the weaker, transient High-Frequency QPOs

110: (HFQPOs, 40--450~Hz) that have attracted more attention, since their

111: frequencies do not vary significantly in response to sizable (factors

112: of 3-4) luminosity changes and are comparable to the orbital

113: frequencies at the Innermost Stable Circular Orbit (ISCO) of black

114: holes with mass $M\sim 10 M_{\odot}$. As such, HFQPOs potentially

115: provide a probe to study the effects of strong gravity. HFQPOs are

116: usually observed in the very high state of the X-ray binaries, 

117: and have low amplitudes

118: ($0.5-2\%$ rms at 2-60~keV) and low coherence ($Q\sim 2-10$). Out of

119: the seven black-hole binaries from which HFQPOs have been reported, four

120: %(GRO J1655-40, XTE J1550-564, GRS 1915+105 and H1743-322) 

121: show pairs of QPOs (first discovered in GRO J1655-40; Strohmayer 2001)

122: with frequency ratios close to $2:3$ (300 and

123: 450~Hz in GRO J1655-40, 184 and 276~Hz in XTE J1550-564, 113 and

124: 168~Hz in GRS 1915+105, 165 and 240~Hz in H1743-322; note that GRS

125: 1915+105 also has a second pair of QPOs with $f=41$ and $67$~Hz).

126: 

127: It is worth noting that QPO (with period of $\sim 1$~hour) in X-ray emission

128: has recently been detected in the active galaxy RE J1034+396 

129: (Gierlinski et al.~2008). This could be the ``supermassive'' analog of the

130: HFQPOs detected in black-hole X-ray binaries.

131: 

132: Despite the observational progress, the origin of the HFQPOs remain 

133: unclear. A number of possibilities/models have been suggested or

134: studied to various degrees of sophistication. We comment on some of these

135: below:

136: 

137: -- Stella, Vietri \& Morsink (1999) and others (see Schnittman \&

138: Bertschinger 2004, Schnittman 2005) suggested that orbiting hot spots

139: (blobs) in the disc oscillating with epicyclic frequencies may provide

140: variability in the X-ray emission.  However the radial positions of

141: such blobs are free parameters, which must be tuned to match the

142: observed QPO frequencies, and it is also not clear that the blobs can

143: survive the differential rotation of the disc.

144: 

145: -- Abramowicz \& Kluzniak (2001) suggested that HFQPOs involve certain

146: nonlinear resonant phenomenon in the disc (e.g., coupling between the

147: radial and vertical epicyclic oscillations of the disc fluid element;

148: Kluzniak \& Abramowicz 2002).  This was motivated by the observed

149: stability of the QPO frequencies and the commensurate frequency ratio.

150: However, so far analysis has been done based only on toy models involving coupled harmonic oscillators (e.g. Rebusco 2004; Horak \& Karas 2006) and no fluid dynamical model producing these resonances

151: has been developed (see Abramowicz et al 2007 and Rebusco 2008 for

152: recent reviews). Petri (2008) considered the resonant oscillation of a

153: test mass in the presence of a spiral density wave, but the origin of

154: the wave is unclear.

155: 

156: -- Acoustic oscillation modes in pressure-supported accretion tori

157: have been suggested as a possible source of the observed QPOs

158: (Rezzolla et al. 2003; Lee, Abramowicz \& Kluziniak 2004; see also

159: Blaes, Arras \& Fragile 2006; Schnittman \& Rezzolla 2006, Blaes 

160: et al 2007, Sramkova et al 2007).  In this

161: model, the commensurate mode frequencies arise from matching the

162: radial wavelength to the size of the torus. Note that the QPO

163: frequencies are determined mainly by the radial boundaries of the

164: torus, which must be tuned to match the observed QPO frequencies. It

165: is also not clear that the accretion flow in the very high state (in

166: which HFQPOs are observed) is well represented by such a torus (e.g. Done et al 2007).

167: 

168: -- Li \& Narayan (2004) considered the dynamics of the

169: interface between the accretion

170: disc and the magnetosphere of a central compact object (see also

171: Lovelace \& Romanova 2007). 

172: The interface is generally Rayleigh-Tayor unstable and may also be

173: Kelvin-Helmholtz unstable. While such an interface is clearly relevant

174: to accreting magnetic neutron stars, Li \& Narayan suggested that it

175: may also be relevant to accreting black holes and that the strongly

176: unstable interface modes may give rise to QPOs with commensurate

177: frequencies.

178: 

179: -- Perhaps the theoretically most appealing is the relativistic

180: diskoseismic oscillation model, according to which general

181: relativistic (GR) effects produce trapped oscillation modes in the

182: inner region of the disc (Kato \& Fukue 1980; Okazaki et al.~1987;

183: Nowak \& Wagoner 1991; see Wagoner 1999; Kato 2001 for reviews; see

184: also Tassev \& Bertschinger 2007 for the kinematic description of some

185: of these wave modes).  A large majority of previous studies have

186: focused on disc g-modes (also called inertial modes or

187: inertial-gravity modes, whose wavefunctions -- such as the pressure

188: perturbation, contain nodes in the vertical direction), because the

189: trapping of the g-mode does not require a reflective inner/outer disc

190: boundary.  Kato (2003a) and Li, Goodman \& Narayan (2003) showed that

191: the g-mode that contains a corotation resonance (where the wave

192: patten frequency equals the rotation rate of the background flow) in

193: the wave zone is heavily damped. Thus the only nonaxisymmetric ($m\neq

194: 0$) g-modes of interest are those trapped around the maximum of

195: $\Omega+\kappa/m$ (where $\Omega$ is the rotational frequency,

196: $\kappa$ is the radial epicyclic frequency and $m$ is the azimuthal

197: mode number; see Fig.~1 below). Unfortunately, the frequencies of such

198: modes, $\omega\simeq m\Omega_{\rm ISCO}$, are too high (by a

199: factor of 2-3) compared to the observed values, given the measured

200: mass and the estimated spin parameter of the black hole (Silbergleit

201: \& Wagoner 2007; see also Tassev \& Bertschinger 2007).  Axisymmetric

202: g-modes ($m=0$) may still be viable in the respect, and recent studies

203: showed that they can be resonantly excited by global disc deformations

204: through nonlinear effects (Kato 2003a,2008; Ferreira \& Ogilvie 2008).

205: Numerical simulations (Arras, Blaes \& Turner 2006; Reynolds \& Miller

206: 2008), however, indicated that while axisymmetric g-mode oscillations

207: are present in the hydrodynamic disc with no magnetic field, they

208: disappear in the magnetic disc where MHD turbulence develops. Also, Fu

209: \& Lai (2008) carried out an analytic study of the effect of magnetic

210: fields on diskoseismic modes and showed that even a weak (sub-thermal)

211: magnetic field can ``destroy'' the self-trapping zone of disc g-modes,

212: and this may (at least partly) explain the disappearance of the

213: g-modes in the MHD simulations.

214: 

215: -- Tagger and collaborators (Tagger \& Pellat 1999; Varniere \& Tagger

216: 2002; Tagger \& Varniere 2006; see Tagger 2006 for a review) developed

217: the theory of accretion-ejection instability for discs threaded by

218: strong (of order or stronger than equipartition), large-scale poloidal

219: magnetic fields.  They showed that such magnetic field provides a

220: strong coupling between spiral density waves and Rossby waves at the

221: corotation, leading to the growth of the waves and energy ejection to

222: disc corona. Tagger \& Varniere (2006) suggested that normal modes

223: trapped in the inner region of the disc become strongly unstable by a

224: combination of accretion-ejection instability and an MHD form of the

225: Rossby wave instability (see Lovelace et al.~1999; Li et al.~2000; see section 6

226: below). The Tagger model has the appealing feature that the instability leads

227: to energy ejection into the disc corona, and thus explains why HFQPOs

228: manifest mainly as the variations of the nonthermal (power-law)

229: radiation from the systems.

230: 

231: 

232: 

233: %\cite{Tsang:2008p4205} explored the effect of the corotation

234: %resonance on the reflection coefficient for non-axisymmetric waves

235: %incident on the corotation barrier for perturbations with no vertical

236: %nodes. It was found that incoming waves could be super-reflected

237: %(reflection coefficient greater than unity) due to wave absorption at

238: %the corotation singularity depending on the slope of the vortensity

239: %($\zeta = \kappa^2/2\Omega\Sigma$), and explicit WKB values for the

240: %Reflection, Transmission and Absorption coefficients were calculated

241: %analytically. This absorption effect was found to dominate the

242: %outgoing flux of angular momentum that determines the growth rate of

243: %the classic corotation amplifier.

244: 

245: \subsection{This Paper}

246: 

247: In this paper we study the global corotational instability of

248: nonaxisymmetric p-modes (also called inertial-acoustic modes) trapped

249: in the inner-most region of the accretion disc around a black

250: hole. The p-modes do not have vertical structure (i.e., the

251: wavefunctions have no node in the vertical direction). We focus on

252: these modes because their basic wave properties (e.g. propagation

253: diagram) are not affected qualitatively by disc magnetic fields (Fu \&

254: Lai 2008) and they are probably robust under hydromagnetic effects

255: and disc turbulence (see Reynolds \& Miller 2008).

256: 

257: The corotational instability of p-modes studied in this paper relies

258: on the well-known GR effect of test-mass orbit around a black hole:

259: Near the black hole, the radial epicyclic frequency $\kappa$ reaches a

260: maximum (at $r=8GM/c^2$ for a Schwarzschild black hole) and goes to

261: zero at the ISCO ($r_{\rm ISCO}=6GM/c^2$).  This causes non-monotonic

262: behavior in the fluid vortensity, $\zeta=\kappa^2/(2\Sigma\Omega)$

263: (assuming the surface density $\Sigma$ is relatively smooth), such

264: that $d\zeta/dr>0$ for $r<r_{\rm peak}$ (where $r_{\rm peak}$ is the

265: radius where $\zeta$ peaks) and $d\zeta/dr<0$ for $r>r_{\rm peak}$.

266: The vortensity gradient $d\zeta/dr$ plays an important role in wave

267: absorption at the corotation resonance (Tsang \& Lai 2008a; see

268: also Goldreich \& Tremaine 1979 for corotational wave absorption due to

269: external forcing). We show that the p-modes with frequencies such that

270: the corotation radii lie inside the vortensity peak can grow in

271: amplitude due to absorption at the corotation resonance, and that the

272: overstability can be achieved for several modes with frequencies

273: closely commensurate with the azimuthal wavenumber $m$.  Tagger \&

274: Varniere (2006) have studied similar modes in discs threaded by strong

275: magnetic fields, but in our analysis magnetic fields play no role.

276: 

277: The trapping of the p-modes requires the existence of a (partially)

278: reflecting boundary at the disc inner edge, close to the ISCO. One may

279: suspect that the rapid radial inflow at the ISCO will diminish any

280: potential instabilities in the inner accretion disc (see Blaes 1987

281: for the case of thick accreting tori). Our analysis of the

282: wave perturbations in the transonic accretion flow (see section 5)

283: suggests that waves are partially reflected at the sonic point, and

284: global overstable p-modes may still be produced under certain 

285: conditions (e.g., when the surface density of the flow

286: varies on sufficiently small length scale around the sonic point).

287: Even better mode trapping (and therefore larger mode growth)

288: may be achieved when the system is an accretion

289: state such that the inner disc edge does not behave as a zero-torque

290: boundary (see section 7 for discussion and references).

291: 

292: Our paper is organized as follows. After summarizing the basic

293: fluid equations for our problem in section 2, we give a physical

294: discussion of the origin of the corotational instability of disc 

295: p-modes in section 3. We present in section 4

296: our calculations of the growing p-modes with simple reflective

297: inner disc boundary conditions. Section 5 contains our analysis of

298: the effect of the transonic radial inflow at the ISCO 

299: on the p-mode growth rate. In section 6, we discuss the role of the Rossby 

300: wave instability and show that it is not effective in typical 

301: accretion discs under consideration.

302: In section 7 we discuss the application

303: of our results to HFQPOs in black hole X-ray binaries.

304: 

305: 

306: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

307: \section{Setup and Basic equations}

308: 

309: We consider a geometrically thin disc and adopt cylindrical

310: coordinate system $(r,\phi,z)$. The flow is assumed to be barotropic,

311: so that the vertically integrated pressure, $P=\int p~dz$, depends

312: only on the surface density, $\Sigma=\int \rho~dz$.  We use

313: the pseudo-Newtonian potential of Paczynski \& Wiita (1980) 

314: \be

315: \Phi = -{GM\over r-r_S},

316: \ee

317: with $r_S = 2GM/c^2$ the Schwarzschild radius.

318: The free-particle (Keplerian) orbital and radial epicyclic (angular)

319: frequencies are

320: \be

321: \Omega_K = \left(\frac{1}{r}\frac{d\Phi}{dr}\right)^{1/2} =

322: \sqrt{\frac{GM}{r}}\frac{1}{r-r_S}~, \qquad \kappa =

323: \left[\frac{2\Omega_K}{r} \frac{d}{dr}(r^2\Omega_K)\right]^{1/2} = \Omega_K

324: \sqrt{\frac{r-3r_S}{r-r_S}}.

325: \label{eq:OmegaK}

326: \ee

327: The function $\kappa$ peaks at $r=(2+\sqrt{3})r_S$ and declines to zero at 

328: $r_{\rm ISCO}=3r_S$ (while for a Schwarzschild black hole in GR, 

329: $\kappa$ peaks at $r=4r_S$). 

330: The unperturbed flow has velocity $\bu_0=(u_r,r\Omega,0)$.

331: Since pressure is negligible for thin discs, we have

332: $\Omega\simeq \Omega_K$.

333: 

334: Neglecting the self-gravity of the disc we have the linear

335: perturbation equations:

336: \ba

337: && {\partial\over\partial t}\delta\Sigma+\nabla\cdot(\Sigma\,\delta \bu

338: +\bu_0\,\delta\Sigma)=0,\label{eq:rho}\\

339: &&{\partial \over\partial t}\delta\bu+(\bu_0\cdot\nabla)\delta\bu

340: +(\delta\bu\cdot\nabla)\bu_0

341: =-\nabla\delta h, \label{eq:u}

342: \ea

343: where $\delta\Sigma,~\delta \bu$ and $\delta h=\delta P/\Sigma$

344: are the (Eulerian) perturbations of surface density, velocity and enthalpy,

345: respectively. For barotropic flow, $\delta h$ and $\delta\Sigma$ are 

346: related by

347: \be

348: \delta h=c_s^2{\delta\Sigma\over\Sigma},

349: \ee

350: where $c_s$ is the sound speed, with $c_s^2=dP/d\Sigma$.

351: We assume all perturbed quantities to be of the form

352: $e^{im\phi - i\omega t}$, 

353: where $m$ is a positive integer, and $\omega$ is the 

354: wave (angular) frequency.

355: The perturbation equations then become

356: \ba

357: && -i\tomega {\Sigma\over c_s^2}\delta h + \frac{1}{r} \frac{\d}{\d r}

358: (\Sigma r \delta u_r) + \frac{im}{r}\Sigma \delta u_\phi 

359: +{1\over r}{\partial\over\partial r}(ru_r\delta\Sigma)

360: = 0,\label{eq:sig}\\

361: && -i\tomega \delta u_r - 2\Omega \delta u_\phi 

362: +{\partial\over\partial r}(u_r\delta u_r)= -{\partial\over\partial r}

363: \delta h,\label{eq:ur}\\

364: && -i\tomega \delta u_\phi + \frac{\kappa^2}{2\Omega}\delta u_r 

365: +{u_r\over r}{\partial\over\partial r}(r\delta u_\phi)

366: =-\frac{im}{r} \delta h,\label{eq:uphi}

367: \ea

368: where 

369: \be

370: \tomega = \omega - m \Omega,

371: \ee

372: is the wave frequency in the frame corotating with the unperturbed fluid.

373: 

374: Except very near the inner edge of the disc, $r_{\rm in}\simeq r_{\rm ISCO}$,

375: the unperturbed radial velocity is small, $|u_r|\ll r\Omega$.

376: In our calculations of the disc modes, we will neglect $u_r$ and

377: set the last terms on the left-hand sides of 

378: equations (\ref{eq:sig})-(\ref{eq:uphi}) to zero (However,

379: $u_r$ plays an important role in determining the inner boundary 

380: condition of the fluid perturbations at $r_{\rm in}$; see section 5).

381: Eliminating the velocity perturbations in favor of the enthalpy,

382: we obtain our master equation

383: \be

384: \left[\frac{d^2}{dr^2} -

385: \frac{d}{dr}\left(\ln\frac{D}{r\Sigma}\right)\frac{d}{dr} -

386: \frac{2m\Omega}{r\tomega}\left(\frac{d}{dr}\ln\frac{\Omega\Sigma}{D}\right)

387: - \frac{m^2}{r^2} - \frac{D}{c_s^2}\right]\delta h = 0,

388: \label{perturbeq1}

389: \ee

390: where 

391: \be

392: D = \kappa^2 - \tomega^2.

393: \ee

394: For concreteness we assume the surface density to have a the power-law form 

395: \be

396: \Sigma \propto r^{-p},

397: \ee

398: where $p$ is the density index.

399: 

400: The above equations adequately describe disc p-modes (also called

401: inertial-acoustic modes), which do not have vertical structure.

402: Other disc modes (g-modes and c-modes) involve the vertical

403: degree of freedom [see Kato 2001 for a review; also see Fig.~1 of

404: Fu \& Lai (2008) for a quick summary],

405: and their stability properties are studied by

406: Kato (2003a), Li et al.~(2003) and Tsang \& Lai (2008b).

407: 

408: To determine the global modes of the disc, appropriate 

409: boundary conditions must be specified. These are discussed in 

410: sections 4 and 5.

411: 

412: 

413: 

414: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

415: \section{P-modes and Their Growth Due

416: to Corotation Resonance: A Physical Discussion}

417: 

418: A WKB analysis of the wave equation (\ref{perturbeq1})

419: yields the dispersion 

420: relation for the local plane wave 

421: $\delta h\propto \exp\left[i\int^r\!k(s) ds\right]$:

422: \be

423: k^2+{D\over c_s^2}+

424: \frac{2m\Omega}{r\tomega}\left(\frac{d}{dr}\ln\frac{\Omega\Sigma}{D}\right)

425: \simeq 0.

426: \label{eq:disper}\ee

427: Far from the singularity ($\tomega=0$) at the corotation radius $r_c$, 

428: this reduces to the 

429: well-known dispersion relation of spiral density wave with no self-gravity

430: (e.g., Shu 1992), $k^2\simeq -D/c_s^2$, or 

431: \be

432: \tomega^2 = \kappa^2 + k^2 c_s^2~.

433: \ee

434: Density waves (p-modes) can propagate inside the inner Lindblad resonance

435: radius $r_{\rm IL}$ (defined by $\tomega = -\kappa$), and outside the

436: outer Lindblad resonance radius $r_{\rm OL}$ (defined by $\tomega =

437: \kappa$), i.e., in the region where

438: $\omega/m<\Omega-\kappa/m$ and 

439: $\omega/m > \Omega+\kappa/m$, respectively (see Fig.~1). 

440: Between $r_{\rm IL}$ and $r_{\rm OL}$, waves are 

441: evanescent except that a very narrow Rossby wave zone exists 

442: around the corotation radius. Indeed, 

443: in the vicinity of $\tomega=0$, equation (\ref{eq:disper}) reduces to 

444: \be

445: \tomega\simeq {2m\Omega\over r (k^2+\kappa^2/c_s^2)}

446: \left({d\ln\zeta\over dr}\right)_{r_c},

447: \ee

448: where 

449: \be

450: \zeta={\kappa^2\over 2\Omega\Sigma}

451: \ee

452: is the vortensity of the (unperturbed) flow.

453: % and the subscript ``c'' implies tha the quantity is evaluated at $r=r_c$.

454: For $(d\zeta/dr)_{r_c}>0$, the Rossby wave zone lies between

455: $r_c$ and $r_c+\Delta r_R$, where $\Delta r_R=(2c_s/\kappa)|\nu|$ and

456: $\int_{r_c}^{r_c+\Delta r_R}k\,dr=\pi\nu$, with the number of wavelengths in the

457: Rossby zone given by (Tsang \& Lai 2008a)

458: \be

459: \nu = \left(\frac{c_s}{q\kappa}\frac{d\ln \zeta}{dr} \right)_{r_c}

460: ={c_s\over q\kappa}\left[{d\over dr}\ln\left({\kappa^2\over\Omega}\right)

461: +{p\over r}\right]_{r_c},

462: \label{eq:nu}\ee

463: where $q \equiv -\left(d\ln \Omega/d \ln r\right)_{\rm r_c}$, and the second

464: equality assumes $\Sigma\propto r^{-p}$. 

465: For $(d\zeta/dr)_{r_c}<0$, the Rossby wave zone lies inside $r_c$, between

466: $r_c-|\Delta r_R|$ and $r_c$ (see Fig.~1). Note that since $\nu\sim c_s/(\kappa r)\sim

467: H/r\ll 1$, no standing Rossby wave can exist in the Rossby zone (see also section 6).

468: 

469: 

470: 

471: %===================================

472: \begin{figure}

473: \centering

474: \includegraphics[width=13.5cm]{f1.eps}

475: \caption{Wave propagation diagram for non-axisymmetric p-modes in thin

476: accretion discs around black holes.  In the upper panel, the three

477: solid curves depict the disc rotation profile $\Omega(r)$ and

478: $\Omega\pm\kappa/m$ (where $\kappa$ is the radia epicyclic frequency);

479: note that the three curves join each other at the disc inner radius

480: $r_{\rm in}=r_{\rm ISCO}$ (the inner-most circular orbit) since

481: $\kappa(r_{\rm ISCO})=0$.  The wavy lines (of height $\omega/m$)

482: indicate the propagation zones for inertial-acoustic waves. Disc

483: p-modes are trapped between $r_{\rm in}$ and the inner Lindblad

484: resonance radius (where $\omega/m = \Omega - \kappa/m$), but can

485: tunnel through the corotation barrier. The lower panel depicts the

486: disc vortensity profile, $\zeta=\kappa^2/(2\Omega\Sigma)$, which has a

487: maximum at the radius $r_{\rm peak}$ (as long as the surface density

488: $\Sigma$ does not vary too strongly with $r$).  P-modes with

489: $\omega/m>\Omega_{\rm peak}=\Omega(r_{\rm peak})$ (the upper wavy

490: line) are overstable due to wave absorption at the corotation

491: resonance radius $r_c$ (where $\omega/m=\Omega$) since

492: $(d\zeta/dr)_{r_c}>0$. P-modes with $\omega/m<\Omega_{\rm peak}$ (the

493: lower wavy line) tend to be damped by wave absorption at $r_c$ since

494: $(d\zeta/dr)_{r_c}<0$. Note that a narrow Rossby wave zone (labeled by

495: thick horizontal bars) exists just outside or inside the corotation

496: radius --- the location of this Rossby zone determines the sign of

497: the corotational wave absorption.  }

498: \label{fig1}

499: \end{figure}

500: %=======================================

501: 

502: Assuming that there exists a reflecting boundary at the inner 

503: disc radius $r_{\rm in}\simeq r_{\rm ISCO}$ (see sections 4.3 and 5),

504: normal modes can be produced, with the waves partially trapped between 

505: $r_{\rm in}$ and 

506: $r_{\rm IL}$ -- these are the p-modes that we will focus on in this paper.

507: The mode eigen-frequency $\omega=\omega_r+i\omega_i$ is generally complex,

508: with the real part $\omega_r$ determined approximately by the Sommerfeld 

509: ``quantization'' condition

510: \be

511: \int_{r_{\rm in}}^{r_{\rm IL}} {1 \over c_s} \sqrt{\tomega_r^2

512: - \kappa^2} dr = n\pi +\varphi, 

513: \ee 

514: where $\tomega_r=\omega_r-m\Omega$, 

515: $n$ is an integer and $\varphi$ (of order unity) is a phase factor

516: depending on the details of the (inner and outer) boundary conditions.

517: The overstability of the p-mode is directly related to the 

518: reflectivity of the corotation barrier between $r_{\rm IL}$ and 

519: $r_{\rm OL}$. In the WKB approximation, the imaginary part of the 

520: mode frequency is given by (Tsang \& Lai 2008a; see also Narayan et al.~1987,

521: who considered shearing-sheet model)

522: \be

523: \omega_i = \left(\frac{|{\cal R}| -1}{|{\cal R}|+1}\right)

524: \left(\int_{r_{\rm in\ }}^{r_{\rm IL}}

525: {|\tomega_r|\over c_s\sqrt{\tomega_r^2 - \kappa^2}} dr\right)^{-1},  

526: \label{eq:omegai}\ee

527: where ${\cal R}$ is the reflectivity (see below).

528: Thus the mode becomes overstable ($\omega_i > 0$) for $|{\cal R}| > 1$

529: (termed ``super-reflection'')  and stable

530: ($\omega_i < 0$) for $|{\cal R}| < 1$.

531: 

532: Super-reflection in fluid discs arises because

533: the waves inside the corotation radius and those outside carry 

534: energy or angular momentum of different signs: 

535: Since the wave inside $r_c$ has pattern speed $\omega_r/m$ less than the 

536: fluid rotation rate $\Omega(r)$, it carries negative energy; outside

537: $r_c$, we have $\omega_r/m>\Omega(r)$, the wave carries positive 

538: energy. Consider an incident wave $\delta h\propto \exp(-i\int^r k\,dr)$, 

539: carrying energy of the amount $(-1)$, propagating from small radii

540: toward the corotation barrier

541: \footnote{Note that since the group velocity 

542: of the wave has opposite sign as the phase velocity for $r<r_{\rm IL}$,

543: the wave of the form $\exp(-i\int^rk\,dr)$ (with $k>0$)

544: is outward propagating.}. The wave reflected at $r_{\rm IL}$

545: takes the form $\delta h\propto {\cal R}\exp(i\int^r k\,dr)$, and the 

546: transmitted wave in the region $r>r_{\rm OL}$ is

547: $\delta h\propto {\cal T}\exp(i\int^r k\,dr)$. 

548: Because of the corotation singularity, the wave energy can also be

549: transferred to the background flow and dissipated at the corotation 

550: radius.   Energy conservation then gives

551: $-1=(-1)|{\cal R}|^2+|{\cal T}|^2+{\cal D}_c$, or

552: \be

553: |{\cal R}|^2=1+|{\cal T}|^2+{\cal D}_c, \label{eq:Econs}

554: \ee

555: where ${\cal D}_c$ is the wave energy dissipated at the corotation. 

556: 

557: Tsang \& Lai (2008a) derived the analytical expressions (in the WKB

558: approximation) for ${\cal T}$, ${\cal R}$ and ${\cal D}_c$. 

559: Two effects determine the reflectivity.

560: (i) The transmitted wave (corresponding to the $|{\cal T}|^2$ term)

561: always carries away positive energy and thus increases 

562: $|{\cal R}|^2$.

563: (ii) Wave absorption at the corotation can have both signs, depending on 

564: $\nu$: For $\nu>0$, the Rossby wave zone lies outside $r_c$, positive wave

565: energy is dissipated and we have ${\cal D}_c>0$; for $\nu<0$, the Rossby zone

566: lies inside $r_c$ and we have ${\cal D}_c<0$ (see Fig.~1). 

567: Tsang \& Lai (2008a) showed explicitly that under most conditions,

568: $|{\cal D}_c|\gg |{\cal T}|^2$ (except when $\nu\simeq 0$, for which

569: ${\cal D}_c\simeq 0$). In the limit of $|\nu|\ll 1$, we have

570: \be

571: |{\cal T}|^2 \simeq \exp(-2\Theta_{\rm II}),\qquad

572: {\cal D}_c\simeq 2\pi\nu\exp(-2\Theta_{\rm IIa}),

573: \ee

574: where

575: \be

576: \Theta_{\rm II}\equiv \int_{r_{\rm IL}}^{r_{\rm OL}}

577: \frac{\sqrt{\kappa^2 - \tomega_r^2}}{c_s}, \qquad \Theta_{\rm IIa}\equiv

578: \int_{r_{\rm IL}}^{r_{c}} \frac{\sqrt{\kappa^2 - \tomega_r^2}}{c_s}

579: \ee

580: [these expressions are valid for $\Theta_{\rm II},\Theta_{\rm IIa}\gg 1$; see 

581: Tsang \& Lai (2008a) for more general expressions]. Thus super-reflectivity

582: ($|{\cal R}|^2>1$) and growing modes ($\omega_i>0$) are achieved when 

583: \be

584: \nu>\nu_{\rm crit}=-{1\over 2\pi}\exp(-2\Theta_{\rm IIb}),\qquad

585: {\rm with}\quad\Theta_{\rm IIb}=\Theta_{\rm II}-\Theta_{\rm IIa}

586: =\int_{r_c}^{r_{\rm OL}}

587: \frac{\sqrt{\kappa^2 - \tomega_r^2}}{c_s}.

588: \label{eq:nucrit}\ee

589: Note that typically $|\nu_{\rm crit}|\ll 1$; if $|{\cal T}|^2$ is neglected

590: compared to ${\cal D}_c$, then $\nu_{\rm crit}=0$.

591: 

592: 

593: As mentioned before, since $\kappa$ is non-monotonic near the black

594: hole, the vortensity $\zeta$ is also non-monotonic, attaining a peak

595: value at $r=r_{\rm peak}$ before dropping to zero at the

596: ISCO. Therefore, p-modes with frequencies such that the corotation

597: radius $r_c$ lies inside $r_{\rm peak}$ are expected to be overstable

598: by the corotational instability discussed above. In other words, when

599: $\omega_r/m> \Omega_{\rm peak}\equiv \Omega(r_{\rm peak})$, the

600: corotation resonance acts to grow the mode.  Note that $\Omega_{\rm

601: peak}$ depends on the surface density profile as well as the spacetime

602: curvature around the black hole (see Fig.~2).  On the other hand, when

603: $\omega_r/m< \Omega_{\rm peak}$ ($\nu<0$), the corotational wave

604: absorption acts to damp the mode.  However, when $\omega_r/m$ is only

605: slightly smaller than $\Omega_{\rm peak}$ ($\nu_{\rm crit}<\nu<0$)

606: mode growth can still be obtained due to wave leakage beyond the outer

607: Lindblad resonance, though the growth rate will be small (see section

608: 4.4 for examples).

609: 

610: 

611: %===========================================

612: \begin{figure}

613: \centering

614: \includegraphics[width=13cm]{f2.eps}

615: \caption{Critical mode frequency for corotational instability as a function 

616: of the disc surface density index $p$ (with $\Sigma\propto r^{-p}$).

617: Wave absorption at the corotation resonance acts to grow the mode only if the

618: corotation occurs in the region of positive vortensity gradient, i.e., 

619: if the mode pattern frequency 

620: $\omega/m>\Omega_{\rm peak}$ (see Fig.~1). P-mode trapping also

621: requires $\omega/m<\Omega_{\rm ISCO}=\Omega(r_{\rm ISCO})$.

622: %Thus the growth rates can only become large if the pattern

623: %frequency occurs between $\Omega_{\rm peak}$ and $\Omega_{ISCO}$. 

624: Note that weak mode growth can still occur when $\omega/m$ is slightly 

625: below $\Omega_{\rm peak}$ due to wave leakage beyond the outer Lindblad resonance

626: radius. See text for detail.

627: %however  this effect is only significant for small negative $\nu$.}

628: }

629: \label{fig2}

630: \end{figure}

631: %===========================================

632: 

633: 

634: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

635: \section{Calculations of Trapped, Overstable P-modes}

636: 

637: To determine the eigenvalues $\omega_r$ and $\omega_i$ of the trapped

638: modes, we solve equations (\ref{eq:sig})-(\ref{eq:uphi})

639: (with $u_r=0$) or equation \eqref{perturbeq1} subjected to 

640: appropriate boundary conditions at $r_{\rm in}$ and $r_{\rm out}$. 

641: 

642: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

643: \subsection{``Landau'' Integration Contour}

644: 

645: When solving eigenvalue problem using the standard method (e.g. the

646: shooting method as described in Press et al 1998), we encountered a

647: conundrum: For $\nu>0$, we could find both a growing mode and

648: a decaying mode, with almost the same $\omega_r$ but opposite

649: $\omega_i$.

650: This appears to contradict our discussion in section 3. 

651: This conundrum arises because our 

652: numerical integration is confined to the real $r$ axis. However,

653: analogous to Landau's analysis of wave damping in a plasma (e.g.,

654: Lifshitz \& Pitaevskii 1981), care must

655: be taken in defining appropriate contour of integration across the

656: corotation resonance. Indeed, at corotation, equation 

657: \eqref{perturbeq1} contains a singular term, proportional to 

658: \be

659: {1\over \tilde\omega} \propto {1\over r-R_c},

660: \ee

661: where $R_c \equiv r_c - ir_c\omega_i/(q\omega_r)$ is the complex

662: pole, $r_c$ is determined by $\omega_r=m\Omega(r_c)$ and 

663: $q=-(d\ln\Omega/d\ln r)_c>0$.

664: 

665: As discussed in Lin (1955) in the context of hydrodynamical shear flows,  

666: to obtain physically relevant solutions of the fluid system, it is 

667: necessary that the integration contour lies above the pole.

668: This is the Landau contour. In essence, only by adopting such

669: a Landau contour can one obtain the correct wave absorption (dissipation)

670: at the corotation. For growing modes ($\omega_i>0$), Im$(R_c)<0$,

671: our numerical integration along the real $r$ axis constitutes

672: the correct Landau contour. On the other hand, 

673: for decaying modes ($\omega_i<0$), the real $r$ axis is not the

674: correct Landau contour as Im$(R_c)>0$. Instead, to obtain physical

675: solutions for these decaying modes, 

676: the integration contour must be deformed so that $R_c$ lies

677: below it (see Fig.~3). As we are primarily interested in over-stable modes

678: in this paper, it is adequate to integrate along the real $r$ axis

679: in our calculation.

680:  

681: %We thus follow the procedure outlined by Lin (1955) to obtain the

682: %physically relevant branch for the given system, which is equivalent

683: %to crossing the singular points with $\omega_i > 0$ (a similar result

684: %is achieved by treating the initial value problem via Laplace

685: %transforms). Thus when integrating over the singularity to identify

686: %growing modes ($\omega_i > 0$) it is sufficient to perform the

687: %numerical integration with $\omega = \omega_r + i \omega_i$.  Such an

688: %integration is not physical for shrinking modes ($\omega_i < 0$),

689: %instead the contour must be deformed to take the same branch as the

690: %growing case (see Figure \ref{fig3}), however we are primarily

691: %interested in over-stable case.

692: 

693: \begin{figure}

694: \centering

695: \includegraphics[width=9cm]{f3.eps}

696: \caption{``Landau'' contour for integration across the corotation resonance.

697: To calculate the growing mode ($\omega_i>0$), 

698: it is adequate to integrate the fluid perturbation 

699: equations along the real $r$ axis (upper panel). To obtain the physical solution

700: for the shrinking mode ($\omega_i<0$),

701: the integration contour must deformed so that the corotational pole 

702: $R_c$ lies below the contour. 

703: %Above: The contour for integrating over the corotation singularity at

704: %$R_c$ for a growing mode. Below: The same but for a shrinking mode,

705: %taking the physically relevant branch.

706: } 

707: \label{fig3}

708: \end{figure}

709: 

710: 

711: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

712: \subsection{Outer Boundary Condition}

713: 

714: As we are interested in self-excited modes in the inner region of the disc,

715: we adopt the radiative outer boundary condition. Specifically,

716: far from the outer Lindblad resonance ($r > r_{\rm OL}$) we demand that 

717: only an outgoing wave exists:

718: \be

719: \delta h \propto A \exp\left(i\int^r k~dr \right)~,\quad

720: {\rm with}~~A=\left({D\over r\Sigma k}\right)^{1/2},

721: \ee

722: where $k = \sqrt{-D/c_s^2}$ (see

723: Tsang \& Lai 2008a). This gives the boundary condition at some 

724: $r_{\rm out} > r_{\rm OL}$: 

725: \be

726: \delta h'(r_{\rm out}) 

727: = \delta h(r_{\rm out}) 

728: %\left(ik + \frac{S'}{2S} - \frac{k'}{2k}\right)_{r_{\rm out}}.

729: \left(ik + {1\over A}{dA\over dr}\right)_{r_{\rm out}}.

730: \label{eq:outer}\ee

731: In practice, we find that $r_{\rm out}\sim 2 r_{\rm OL}$ would yield sufficiently 

732: accurate results.

733: 

734: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

735: \subsection{Inner Boundary Conditions}

736: 

737: To obtain global trapped modes, at least partial wave reflection must occur 

738: at $r_{\rm in}$. To focus on the effect of corotational instability discussed

739: in section 3, in this section we consider two simple inner boundary 

740: conditions. We defer our analysis of the effect of radial inflow

741: on the p-modes to section 5.

742: 

743: (i) At the ISCO, the flow plunges into the black hole, we expect a sudden

744: decrease in the surface density of the disc. Thus, it is reasonable to 

745: consider the free surface boundary condition, i.e.,

746: the Lagrangian pressure perturbation $\Delta P=0$.

747: Using $\delta u_r=-i\tomega\xi_r$ (where $\xi_r$ is the radial Lagrangian

748: displacement), and $\Delta P=\delta P+\xi_r dP/dr$, we have

749: \be

750: \left({\Delta P\over\Sigma}\right)_{\rm r_{\rm in}}=

751: \left(\delta h - pc_s^2\frac{i\delta u_r}{r \tomega}\right)_{r_{\rm 

752: in}}=0,

753: \label{eq:DeltaP}\ee 

754: where we have used

755: $dP/dr = (dP/d\Sigma)(d\Sigma/dr) = -pc_s^2\Sigma/r$

756: for barotropic, power-law discs ($\Sigma\propto r^{-p}$).

757: 

758: (ii) We assume that 

759: the radial velocity perturbation vanishes at the inner boundary, i.e., 

760: $\delta u_r = 0$. This was adopted by Tagger \& Varniere (2006)

761: in their calculations of overstable global modes due to accretion-ejection 

762: instability.

763: 

764: Both of these boundary conditions correspond to zero loss of wave energy

765: at the inner boundary: If a wave from large radii impinges toward 

766: $r_{\rm in}$, the reflected wave will have the same amplitude.

767: However, the phase shifts due to reflection

768: differ in the two cases, and the resulting mode frequencies $\omega_r$ 

769: are different. Since the corotational wave amplification depends on

770: $\omega_r$, the mode growth rate $\omega_i$ will also be different.

771: 

772: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

773: \subsection{Numerical Results}

774: 

775: %====================================================

776: \begin{figure}

777: \centering

778: \begin{tabular}{ccc}

779: %\epsfig{file=wf1.eps,width=0.5\linewidth,clip=} &

780: %\epsfig{file=wf2.eps,width=0.5\linewidth,clip=}

781: \epsfig{file=f4a.eps,width=0.5\linewidth,clip=} &

782: \epsfig{file=f4b.eps,width=0.5\linewidth,clip=}

783: \end{tabular}

784: \caption{Example wavefunctions for disc p-modes. The upper and 

785: middle panels show $\delta h$ and $i\delta u_r$ (the solid lines for

786: the real part and dashed lines for the imaginary part),

787: the lower panels show the angular momentum flux, all in arbitrary 

788: units [with $\delta h(r_{\rm out})=1$]. The radius $r$ is in units of

789: $GM/c^2$. The disc sound speed is $c_s =0.1r\Omega$, and the $m=2$ 

790: modes are obtained using the inner boundary condition $\Delta 

791: P(r_{\rm ISCO}) = 0$. 

792: The left panels show the p-mode for the disc with 

793: a density profile $\Sigma \propto r^{-1}$, with 

794: the eigenvalues $\omega_r = 0.467 m\Omega_{\rm ISCO},~ \omega_i/\omega_r

795: = 0.0029$ [where $\Omega_{\rm ISCO}

796: =\Omega(r_{\rm ISCO})$]; the right panels show the mode for 

797: the disc with constant surface density profile ($p=0$), with

798: eigenvalues $\omega_r =0.464 m\Omega_{\rm ISCO}, ~

799: \omega_i/\omega_r = 0.00073$. 

800: Note that the model shown on the left panels has $r_c<r_{\rm peak}$

801: (the radius of peak vortensity) and thus $F(r_c+)<F(r_c-)$, 

802: while the model shown on the right panels has $r_c>r_{\rm peak}$

803: and $F(r_c+)>F(r_c-)$.

804: In both models there is a positive flux for $r > r_c$ due to the outward

805: propagating wave. The inserts on the left panels show the blowups

806: of the real wavefunctions near the corotation radius.

807: }

808: \label{fig4}

809: \end{figure}

810: 

811: %====================================================

812: \begin{figure}

813: \centering

814: \begin{tabular}{ccc}

815: \epsfig{file=f5a.eps,width=0.5\linewidth,clip=} &

816: \epsfig{file=f5b.eps,width=0.5\linewidth,clip=}

817: \end{tabular}

818: \caption{The real and imaginary frequencies of disc p-modes (with 

819: azimuthal wave numbers $m=1,2,3$)

820: as a function of the surface density index $p$ (where $\Sigma

821: \propto r^{-p}$). The modes are calculated assuming the inner 

822: boundary condition $\Delta P(r_{\rm ISCO}) = 0$.

823: The left panels are for discs with $c_s=0.1r\Omega$ and the right

824: panels for $c_s=0.2r\Omega$. 

825: The dotted lines denote the lower bound $\omega_r/m = \Omega_{\rm

826: peak}$ for which the corotational wave absorption 

827: acts to enhance mode growth.}

828: \label{fig5}

829: \end{figure}

830: 

831: %====================================================

832: \begin{figure}

833: \centering

834: \epsfig{file=f6.eps,width=0.5\linewidth,clip=}

835: \caption{The real and imaginary frequencies of disc p-modes

836: (with $m=2,3$) as a function of the normalized sound speed

837: $c_s/(r\Omega)$. The disc is assumed to have a constant

838: density profile ($p = 0$), and the inner boundary condition

839: is $\Delta P(r_{\rm ISCO}) = 0$. The bottom panel shows

840: the $\nu$ parameters for the modes.}

841: \label{fig6}

842: \end{figure}

843: 

844: 

845: %====================================================

846: \begin{figure}

847: \centering

848: \begin{tabular}{ccc}

849: \epsfig{file=f7a.eps,width=0.5\linewidth,clip=} &

850: \epsfig{file=f7b.eps,width=0.5\linewidth,clip=}

851: \end{tabular}

852: \caption{The real and imaginary frequencies of disc p-modes (with 

853: azimuthal wave numbers $m=1,2,3$)

854: as a function of the surface density index $p$ (where $\Sigma

855: \propto r^{-p}$). The modes are calculated assuming the inner 

856: boundary condition $\delta u_r(r_{\rm ISCO}) = 0$.

857: The left panels are for discs with $c_s=0.1r\Omega$ and the right

858: panels for $c_s=0.2r\Omega$. 

859: The dotted lines denote the lower bound $\omega/m = \Omega_{\rm

860: peak}$ for which the corotational wave absorption 

861: acts to enhance mode growth.}

862: \label{fig7}

863: \end{figure}

864: %====================================================

865: 

866: We solve for the complex eigen-frequency $\omega = \omega_r + i

867: \omega_i$ using the shooting method, with a 

868: fifth-order Runge-Kutta integrator (Press et al.~1992). As

869: discussed in section 4.1 we only calculate the growing modes ($\omega_i>0$). We

870: consider disc models with different surface density profile (characterized

871: by the index $p$), sound speed $c_s$, and inner boundary

872: conditions. For a given set of disc parameters 

873: and the azimuthal mode wavenumber $m$, the 

874: lowest order (highest frequency) mode has the best chance

875: of being overstable. This is easily understood from our discussion in

876: section 3 (see Fig.~1): a low-frequency wave has to penetrate a wider 

877: evanescent barrier for the corotational amplifier to be effective,

878: and when $\omega_r<m\Omega_{\rm peak}$ the corotation resonance acts to

879: damp the mode. For most disc models we have considered, 

880: the lowest order mode (of a given $m$)

881: is the only mode that has $\omega_i>0$.

882: 

883: Our numerical results are presented in Figures \ref{fig4}--\ref{fig7}.

884: Figure 4 gives two examples of the eigenfunctions of overstable trapped 

885: p-modes, obtained with the inner boundary condition $\Delta P=0$.

886: In addition to $\delta h$ and $\delta u_r$, we also plot the

887: angular momentum flux carried by the wave across the disc

888: (e.g., Goldreich \& Tremaine 1979; Zhang \& Lai 2006)

889: \be

890: F(r) = \pi r^2\Sigma\, {\rm Re} \left(\delta u_r \delta u_\phi^\ast\right)=

891: \frac{\pi m r \Sigma}{D}\,{\rm Im}\left(\delta h\frac{d\delta

892: h^*}{dr} \right),

893: \label{eq:flux}\ee

894: where the second equality follows from equations 

895: (\ref{eq:sig})-(\ref{eq:uphi}) (with $u_r=0$).

896: We see from Fig.~4 that outside the corotation radius ($r_c$), 

897: $F$ is nearly constant since only

898: the outgoing wave exists in this region and the wave action is conserved

899: in the limit of $\omega_i\ll \omega_r$. Inside $r_c$, the interference between

900: the ingoing and outgoing waves gives rise to the variation of $F$. At 

901: $r_{\rm ISCO}$, $F$ approaches zero since no wave action is lost through 

902: the disc inner boundary when $\Delta P=0$.\footnote{

903: Note that equation (\ref{eq:flux}) is the time averaged flux and is defined

904: for waves with real $\omega$.

905: Using equation (\ref{eq:DeltaP}) and 

906: $D\delta u_r=i\tomega d\delta h/dr-(2im\Omega/r)\delta h$ (obtained from 

907: eqs.~[\ref{eq:ur}]-[\ref{eq:uphi}]) it is easy to

908: show that $F(r_{\rm in})=0$ exactly for real $\omega$.}

909: Figure 4 also shows a flux jump across the corotation.

910: In the limit of $\omega_i\ll \omega_r$, $d\delta h/dr$,

911: $\delta u_r$ and $\delta u_\phi$

912: are discontinuous across $r_c$

913: (although $\delta h$ is continuous), giving rise to the flux discontinuity

914: (see Tsang \& Lai 2008a):

915: \be

916: F(r_c+)-F(r_c-)=-{2\pi^2 m r\Sigma \nu \kappa\over c_s D}|\delta h|^2

917: \Bigl|_{r_c}.

918: \ee

919: This discontinuity signifies the corotational wave absorption, the

920: sign of which depends on $\nu\propto (d\zeta/dr)_{r_c}$, as discussed in

921: section 3. Thus, the model

922: shown on the left panels of Fig.~4 has $r_c<r_{\rm peak}$ and

923: $\nu>0$, and the mode growth is primarily driven by wave absorption 

924: at the corotation. The model shown on the right panels of Fig.~4, 

925: on the other hand, has $r_c>r_{\rm peak}$ and $\nu<0$, thus the corotational

926: wave absorption acts to damp the mode, and the overall growth of the mode

927: is due to the outgoing wave beyond $r_c$, and the growth rate 

928: is much smaller than the model shown on the left panels.

929: 

930: Figures 5--6 show the frequencies of the

931: fundamental (no node/highest frequency) growing 

932: p-modes (with $m=1,2,3$) for various disc parameters, again obtained with 

933: the inner boundary condition $\Delta P=0$. We consider 

934: $p$ in the range between $-1.5$ and 1.5, and $c_s$ 

935: up to $0.3r\Omega$. 

936: %As seen in Fig.~1, the modes are trapped between the inner boundary

937: %and the inner Lindblad resonance. 

938: For a given sound speed,

939: the real mode frequency $\omega_r$ depends very weakly on $p$, but

940: the growth rate $\omega_i$ increases with $p$ (see Fig.~5) since

941: a larger value of $p$ leads to a larger $\nu$ and enhanced wave

942: absorption at the corotation [see equation (\ref{eq:nu})].

943: In general, as the sound speed 

944: increases, the effective wavelength of the mode increases, and 

945: $\omega_r$ decreases in order ``fit in'' the trapping zone between

946: $r_{\rm in}$ and $r_{\rm IL}$ (see Fig.~6). 

947: The mode growth rates $\omega_i$ depends 

948: on $c_s$ in a non-monotonic way because of two competing effects:

949: As $c_s$ increases, less attenuation occurs in the evanescent zone, 

950: and more wave energy can be absorbed at the corotation and

951: propagate to the outer edge of the disc; these tend to increase

952: $\omega_i$. On the other hand, increasing $c_s$ also leads to

953: smaller $\omega_r$, which shifts the corotation resonance to 

954: a larger radius and leads to decreasing $\nu$ and $\omega_i$.

955: 

956: Note that the growing modes shown in Fig.~5 extend below the 

957: $\omega_r/m=\Omega_{\rm peak}$ boundary due to the propagation of 

958: waves beyond the corotation radius, as discussed in section 3

959: [see eqs.~(\ref{eq:nucrit})].

960: %up to the point that $\nu \approx -

961: %\frac{1}{2\pi} e^{-2\Theta_{\rm IIb}}$, where $\Theta_{\rm IIb} \equiv

962: %\int_{r_{\rm c}}^{r_{\rm OL}} \sqrt{\kappa^2 - \tomega_r^2}/c$. 

963: Such modes (with $\nu_{\rm crit}<\nu<0$) grow

964: significantly slower than the modes with $\nu>0$ 

965: as the flux is attenuated by the entire barrier between $r_{\rm IL}$ and $r_{\rm OL}$.

966: 

967: For comparison, Figure 7 shows the disc mode

968: frequencies and growth rates

969: when the inner boundary condition $\delta u_r=0$ is

970: adopted. The different boundary condition leads to a different phase shift

971: $\varphi$ and higher mode frequency, but 

972: the results are similar to those illustrated in Fig.~5.

973: In particular, as $p$ increases, $\omega_r$ remains approximately

974: constant while $\omega_i$ increases.

975: 

976: 

977: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

978: \section{Effect of Radial Inflow on the P-Mode Growth Rate}

979: 

980: Our mode calculations presented in section 4 neglect 

981: the radial velocity of the accretion flow and assume

982: a loss-less inner disc boundary condition (either $\Delta P=0$ or 

983: $\delta u_r=0$ at $r_{\rm ISCO}$). In real discs, the radial 

984: inflow velocity $u_r$ is not negligible as $r$ approaches 

985: $r_{\rm ISCO}$, and the flow goes through a transonic point (where

986: $u_r=-c_s$) at a radius very close to $r_{\rm ISCO}$. We expect 

987: that part of the fluid perturbations may be advected into the

988: black hole and the inner disc boundary will not be completely

989: loss-less. Here we study the effect of the transonic flow

990: on the p-mode growth rate.

991: 

992: We note that just as the accretion disc is not laminar but 

993: turbulent, the accretion flow around $r_{\rm ISCO}$

994: is complicated. General relativistic MHD simulations in 3D

995: are only beginning to shed light on the property of the black hole

996: accretion flow (e.g., Beckwith, Hawley \& Krolik 2008; Shafee et al.~2008;

997: Noble, Krolik \& Hawley 2008), and many uncertainties remain 

998: unresolved. Here, to make analytic progress, we adopt a simple viscous

999: transonic flow model, which qualitatively describes the inner accretion flow

1000: of the black hole as long as the flow remains 

1001: geometrically thin (see Afshordi \& Paczynski 2003).

1002: 

1003: 

1004: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1005: \subsection{Boundary Condition at the Sonic Point}

1006: 

1007: We rewrite equations (\ref{eq:sig})-(\ref{eq:ur}) as

1008: \ba

1009: &&{u_r\over c_s^2}\delta h'+\delta u_r'=\left[{i\tomega\over c_s^2}

1010: -u_r\left(c_s^{-1}\right)'\right]\delta h+{u_r'\over u_r}\,\delta u_r-{im\over r}

1011: \,\delta u_\phi\equiv {\cal A}_1,\\

1012: &&\delta h'+u_r\,\delta u_r'=(i\tomega-u_r')\,\delta u_r+2\Omega\,\delta u_\phi

1013: \equiv {\cal A}_2,

1014: \ea

1015: where $'$ stands for $d/dr$ and we have used $r\Sigma u_r=$ constant for 

1016: the background flow. Solving for $\delta h'$ and $\delta u_r'$ we have

1017: \ba

1018: &&\delta h'={{\cal A}_2-{\cal A}_1u_r\over 1-u_r^2/c_s^2},\\

1019: &&\delta u_r'={{\cal A}_1-(u_r/c_s^2){\cal A}_2 \over 1-u_r^2/c_s^2}.

1020: \ea

1021: Clearly, in order for the perturbation to be regular at the sonic

1022: point $r_s$, where $u_r=-c_s$, we require

1023: \be

1024: {\cal A}_2+c_s{\cal A}_1=0\quad {\rm at}~~r=r_s.

1025: \label{eq:calA}\ee

1026: For definiteness, we characterize the variations of $\Sigma$ and $c_s$ at the

1027: sonic point $r_s\simeq r_{\rm ISCO}$ by the two length scales: 

1028: \be

1029: \left({\Sigma'\over \Sigma}\right)_{r_s}={1\over L_\Sigma},\qquad

1030: \left({c_s'\over c_s}\right)_{r_s}={1\over L_c}.

1031: \label{eq:Lc}\ee

1032: From $r\Sigma u_r=$ constant, we also have $u_r'=c_s (r_s^{-1}+L_\Sigma^{-1})$ at $r=r_s$.

1033: Then equation (\ref{eq:calA}) becomes

1034: \be

1035: \left({i\tomega\over c_s}-{2\over L_c}\right)\delta h

1036: +\left[i\tomega-2c_s\left({1\over r}+{1\over L_\Sigma}\right)\right]\delta u_r

1037: +\left(2\Omega-{imc_s\over r}\right)\delta u_\phi=0\quad {\rm at}~~r=r_s.

1038: \label{eq:bc}\ee

1039: This is the boundary condition for the fluid perturbations at the sonic point.

1040:  

1041: 

1042: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1043: \subsection{Properties of the Transonic Flow}

1044: 

1045: Before exploring the effect the radial inflow on the disc modes, we first 

1046: estimate the length scale for the surface density variation, 

1047: $L_\Sigma$, using the viscous slim disc model (e.g.,

1048: Muchotrzeb \& Paczynski 1982; Matsumoto et al.~1984;

1049: Abramowicz et al.~1988)

1050: 

1051: The basic steady-state slim disc equations are

1052: \ba

1053: &&\dot M=-2\pi r\Sigma u_r,\\

1054: &&u_ru_r'=-c_s^2{\Sigma'\over\Sigma}+(\Omega^2-\Omega_K^2)r,\\

1055: &&{\dot M} l_0={\dot M} l+2\pi\nu_{\rm vis} r^3\Sigma\Omega',\label{eq:l0}

1056: \ea

1057: where $l=r^2\Omega$ is the specific angular momentum of the flow, 

1058: $\Omega$ is the actual rotation rate,

1059: $\Omega_K$ is given by equation (\ref{eq:OmegaK}), $\nu_{\rm vis}$ 

1060: is the kinetic viscosity,

1061: and $l_0$ is the eigenvalue that must be solved so that flow pass through the 

1062: sonic point smoothly. We shall use the $\alpha$-disc model,

1063: so that $\nu_{\rm vis}=\alpha Hc_s$, with $H\simeq c_s/\Omega_K$.

1064: 

1065: To estimate $L_\Sigma$, we assume $l(r)\simeq r^2\Omega_K(r)$ for $r\go r_{\rm ISCO}$

1066: and $l_0\simeq l_K(r_{\rm ISCO})$. Equation (\ref{eq:l0})

1067: gives 

1068: \be

1069: u_r(1-l_0/l)\simeq -3\nu_{\rm vis}/(2r)=-3\alpha Hc_s/(2r),

1070: \ee

1071: valid for $r\go r_{\rm ISCO}$. At the radius $r=r_{\rm ISCO}+\Delta r$, we have

1072: $u_r(\Delta r/r_{\rm ISCO})^2\simeq -4\alpha Hc_s/r_{\rm ISCO}$.

1073: The sonic point ($u_r=-c_s$) is at $\Delta r\simeq 2\sqrt{\alpha Hr_{\rm ISCO}}$,

1074: and $u_r=-c_s/2$ at $\Delta r\simeq 2\sqrt{2\alpha Hr_{\rm ISCO}}$. Thus

1075: $u_r'(r_s)\sim c_s/L_\Sigma$, with 

1076: \be

1077: L_\Sigma=\left({\Sigma\over\Sigma'}\right)_{r_s}\sim

1078: 2\sqrt{\alpha Hr_s}=2\sqrt{\alpha\beta}\,r_s,

1079: \ee

1080: where $\beta=c_s/(r\Omega)$. This should be compared to the disc thickness

1081: $H=\beta r$: depending on the value of $\alpha$, both $L_\Sigma<H$ and $L_\Sigma>H$ are

1082: possible.

1083: 

1084: The value and sign of $L_c$ depend on the thermodynamical and radiative

1085: properties of the flow, and cannot be estimated in a simple way.

1086: It is reasonable to expect $|L_c|\sim r_s$.

1087: 

1088: 

1089: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1090: \subsection{Reflectivity at the Sonic Point}

1091: 

1092: 

1093: \begin{figure}

1094: \centering

1095: \begin{tabular}{ccc}

1096: \epsfig{file=f8a.eps,width=0.5\linewidth,clip=} &

1097: \epsfig{file=f8b.eps,width=0.5\linewidth,clip=}

1098: \end{tabular}

1099: \caption{The wave reflectivity at the transonic point of the inner disc

1100: as a function of the parameter $L_\Sigma/H$, for $r_s/L_c=0$ (the left panel)

1101: or $r_s/L_c=-3$ (the right panel). In both panels, the heavier lines are for

1102: $c_s=0.1r\Omega$ and the lighter lines for $c_s=0.05r\Omega$. The 

1103: short-dashed, solid and long-dashed lines are for $m=1,2,3$, respectively.

1104: The wave frequency is set to be $\omega=0.7m\Omega(r_s)$.

1105: }

1106: \label{fig8}

1107: \end{figure}

1108: 

1109: 

1110: We can understand qualitatively the effect of the transonic boundary

1111: condition on the p-mode by calculating the reflectivity $\cR_s$

1112: of the inner boundary.

1113: 

1114: Consider a density wave $\delta h\propto \exp(i\int^r k\,dr)$

1115: in the wave zone $r_{\rm in}=r_s<r<r_{\rm IL}$, traveling toward

1116: the inner disc boundary.

1117: \footnote{Note that since the group velocity 

1118: of the wave has opposite sign as the phase velocity for $r<r_{\rm IL}$,

1119: the wave of the form $\exp(i\int^r k\,dr)$ (with $k>0$) is inward propagating.}. 

1120: Upon reflection,

1121: the wave becomes $\delta h\propto \cR_s\exp(-i\int^r k\,dr)$. 

1122: Including the correct WKB amplitude (see Tsang \& Lai 2008a),

1123: the wave outside the sonic point can be written as (up to a constant 

1124: prefactor)

1125: \be

1126: \delta h=A\left[\exp\left(i\int^r_{r_s}k\,dr\right)

1127: +\cR_s \exp\left(-i\int^r_{r_s} k\,dr\right)\right],

1128: \qquad (r_s<r<r_{\rm IL})

1129: \label{eq:deltah}\ee

1130: where 

1131: \be

1132: k={(-D)^{1/2}\over c_s},\quad

1133: A=\left({D\over r\Sigma k}\right)^{1/2}.

1134: \ee

1135: To apply the boundary condition (\ref{eq:bc})

1136: to equation (\ref{eq:deltah}), we neglect

1137: $u_r$ in equations (\ref{eq:sig})--(\ref{eq:uphi}) at $r=r_s+\varepsilon$,

1138: with $\varepsilon\ll r_s$ and $r_s\simeq r_{\rm ISCO}$. Implicit in

1139: this procedure is the assumption that the fluid perturbations

1140: do not vary significantly between $r_s$ and $r_s+\varepsilon$. 

1141: We then obtain

1142: \be

1143: \cR_s={ik+L_A^{-1}+K-2m\Omega/(r\tomega)\over

1144: ik-L_A^{-1}-K+2m\Omega/(r\tomega)}\Biggl|_{r_s},

1145: \ee

1146: where

1147: \be 

1148: L_A^{-1}=\left({A'\over A}\right)_{r_s},

1149: \ee

1150: and

1151: %\be K=\left({\tomega^2L_\Sigma\over 2c_s^2}\right)

1152: %{1-\left({mc_s\over r\tomega}\right)^2-i\left({2mc_s\Omega\over 

1153: %r\tomega^2}-{2c_s\over \tomega L_c}\right)\over

1154: %1+{L_\Sigma\over r}-i{\tomega L_\Sigma\over 2c_s}}\Biggl|_{r_s}. \ee

1155: \be

1156: K=\left({\tomega^2L_\Sigma\over 2c_s^2}\right)

1157: {1-\left({mc_s/r\tomega}\right)^2-i\left[{2mc_s\Omega/ 

1158: (r\tomega^2)}-{2c_s/(\tomega L_c)}\right]\over

1159: 1+({L_\Sigma/r})-i({\tomega L_\Sigma/2c_s})}\Biggl|_{r_s}.

1160: \label{eq:K}\ee

1161: 

1162: When considering the damping of the p-mode 

1163: %(trapped between $r_{\rm in}=r_s\simeq r_{\rm ISCO}$ and $r_{\rm IL}$) 

1164: due to the transonic flow, the quantity

1165: $|\cR_s|^2-1$ is the most relevant (see section 5.4 below). Let 

1166: \be

1167: K=|K|\exp(-i\psi),

1168: \ee

1169: we have

1170: \be

1171: |\cR_s|^2-1=-{4\,k\,|K|\sin\psi\over (L_A^{-1}+|K|\cos\psi)^2

1172: +(k+|K|\sin\psi)^2}\Biggl|_{r_s}.

1173: \ee

1174: Using $\beta =c_s/(r\Omega)$, $\tomega=-\hat\omega m\Omega_{\rm ISCO}$ (where

1175: $0<{\hat\omega}<1$), we find from equation (\ref{eq:K}) 

1176: that 

1177: \be

1178: \psi=\tan^{-1}\left({2\beta\over m{\hat\omega}^2}{1+r_s\hat\omega/L_c

1179: \over 1-\beta^2/\hat\omega^2}\right)

1180: +\tan^{-1}\left({m\hat\omega L_\Sigma\over 2\beta r_s}{1\over 1+L_\Sigma/r_s}

1181: \right).

1182: \ee

1183: 

1184: Figure 8 shows how the reflectivity depends on various parameters 

1185: of the disc inner edge. In particular, for small $L_\Sigma/H=L_\Sigma/(\beta r_s)$,

1186: i.e., when the surface density of the disc decreases rapidly at the sonic point, 

1187: $|\cR_s|^2$ is only slightly smaller than unity and the wave loss

1188: at the inner edge of the disc is small.

1189: 

1190: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1191: \subsection{Mode Growth Rate in the WKB Approximation}

1192: 

1193: Consider the p-mode trapped between $r_{\rm in}=r_s\simeq r_{\rm ISCO}$ and

1194: $r_{\rm IL}$. With the reflectivity

1195: at $r_{\rm IL}$ given by $\cR$ (see section 3), we can write the wave amplitude 

1196: for $r<r_{\rm IL}$ as\footnote{Note that this definition of $\cR$

1197: differs from that in Tsang \& Lai (2008a) by a phase factor of

1198: $\exp(i\pi/4)$.}

1199: \be

1200: \delta h \propto \left({D\over r\Sigma k}\right)^{1/2}\left[

1201: \exp\left(-i\int^r_{r_{\rm IL}}k\,dr\right)

1202: +\cR \exp\left(i\int^r_{r_{\rm IL}} k\,dr\right)\right],

1203: \qquad (r_s<r<r_{\rm IL})

1204: \label{eq:deltah2}\ee

1205: On the other hand, with the reflectivity at 

1206: $r_{\rm in}=r_s$ given by $\cR_s$, the wave can also be expressed as

1207: (\ref{eq:deltah}). For stationary waves we therefore require

1208: \be

1209: \exp(2i\Theta)=\cR\cR_s,\qquad {\rm with}~~\Theta=\int_{r_{\rm in}}^{r_{\rm IL}}k\,dr

1210: =\Theta_r+i\Theta_i,

1211: \ee

1212: where $\Theta_r$ and $\Theta_i$ are real. The real eigen-frequency 

1213: $\omega_r$ is given by

1214: \be

1215: \Theta_r=\int_{r_{\rm in}}^{r_{\rm IL}}k_r\,dr=

1216: \int_{r_{\rm in}}^{r_{\rm IL}}{\sqrt{\tomega_r^2-\kappa^2}\over c_s}\,dr=

1217: n\pi+{\varphi\over 2},

1218: \ee

1219: where $\cR\cR_s=|\cR\cR_s|\exp(i\varphi)$, and $n$ is an integer. The mode growth

1220: rate $\omega_i$ is determined by

1221: $|\cR\cR_s|=\exp(-2\Theta_i)$, or

1222: \be

1223: \tanh\Theta_i=-\left({|\cR\cR_s|-1\over |\cR\cR_s|+1}\right).

1224: \ee

1225: For $\Theta_i=\int_{r_{\rm in}}^{r_{\rm IL}}k_i\,dr\ll 1$ and

1226: $k_i\simeq \omega_i\tomega_r/(c_s\sqrt{\tomega_r^2-\kappa^2})$, we obtain

1227: \be

1228: \omega_i = \left(\frac{|\cR\cR_s| -1}{|\cR\cR_s|+1}\right)

1229: \left[\int_{r_{\rm in\ }}^{r_{\rm IL}}

1230: {|\tomega_r|\over c_s\sqrt{\tomega_r^2 - \kappa^2}} dr\right]^{-1},  

1231: \label{eq:omegai2}\ee

1232: where we have assumed $\tomega_r<0$.

1233: Equation (\ref{eq:omegai2}) is to be compared with (\ref{eq:omegai}), 

1234: where perfect reflection at $r_{\rm in}$ is assumed.

1235: Clearly, to obtain growing modes we require $|\cR\cR_s|>1$. For a given

1236: $|\cR|>1$, growing modes are possible only when the loss

1237: at the sonic point is sufficiently small (i.e., $|\cR_s|$ is sufficiently

1238: close to unity).

1239: 

1240: 

1241: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1242: \subsection{Numerical Results}

1243: 

1244: 

1245: \begin{figure}

1246: \centering

1247: \epsfig{file=f9.eps,width=0.5\linewidth,clip=}

1248: \caption{Wavefunctions for a disc p-mode. The 

1249: notations are the same as in Fig.~4.

1250: The disc has sound speed $c_s =0.1r\Omega$ and constant density profile

1251: ($p=0$), and the $m=2$ mode is calculated using 

1252: the transonic inner boundary condition (\ref{eq:bc}) with 

1253: $L_\Sigma/H=0.25$ and $L_c=\infty$.

1254: The eigenvalues are $\omega_r = 0.725m\Omega_{\rm ISCO}$ and

1255: $\omega_i/\omega_r=0.00267$.

1256: }

1257: \label{fig9}

1258: \end{figure}

1259: 

1260: 

1261: \begin{figure}

1262: \centering

1263: \begin{tabular}{ccc}

1264: \epsfig{file=f10a.eps,width=0.5\linewidth,clip=} &

1265: \epsfig{file=f10b.eps,width=0.5\linewidth,clip=}

1266: \end{tabular}

1267: \caption{The real and imaginary frequencies of disc p-modes (with 

1268: azimuthal wave numbers $m=1,2,3$)

1269: as a function of $L_\Sigma/H$ [see eq.~(\ref{eq:Lc})]

1270: The modes are calculated using the transonic inner 

1271: boundary condition (\ref{eq:bc}) with $L_c=\infty$.

1272: The disc has a constant surface density profile and the sound 

1273: speed is $c_s=0.1r\Omega$ (left panels) or $0.05r\Omega$ (right panels).

1274: }

1275: \label{fig10}

1276: \end{figure}

1277: 

1278: 

1279: \begin{figure}

1280: \centering

1281: \begin{tabular}{ccc}

1282: \epsfig{file=f11a.eps,width=0.5\linewidth,clip=} &

1283: \epsfig{file=f11b.eps,width=0.5\linewidth,clip=}

1284: \end{tabular}

1285: \caption{The real and imaginary frequencies of disc p-modes (with 

1286: azimuthal wave numbers $m=1,2,3$).

1287: The modes are calculated using the transonic inner 

1288: boundary condition (\ref{eq:bc}). 

1289: The disc has a constant surface density profile and the sound 

1290: speed is $c_s=0.1r\Omega$. The left panels show the cases with

1291: $L_c=-r_s/3$ by varying $L_\Sigma/H$, and right panels show the

1292: cases with $L_\Sigma/H=0.25$ by varying $r_s/L_c$ 

1293: [see eq.~(\ref{eq:Lc})].

1294: }

1295: \label{fig11}

1296: \end{figure}

1297: 

1298: We solve equations (\ref{eq:sig})-(\ref{eq:uphi}) (with $u_r=0$)

1299: subjected to the radiative outer boundary condition (\ref{eq:outer})

1300: and the transonic inner boundary condition (\ref{eq:bc}).

1301: 

1302: Figure 9 depicts an example of the p-mode wavefunctions.

1303: Again, the discontinuity in the angular momentum flux $F$ at 

1304: $r_c$ signifies wave absorption; since $r_c<r_{\rm peak}$, 

1305: this leads to mode growth.

1306: Comparing with Fig.~4, here the angular momentum flux at $r_{\rm in}$ is

1307: significantly nonzero, indicating wave loss through the sonic point.

1308: Nevertheless, the corotational instability is sufficiently strong to

1309: overcome the loss and makes the mode grow.

1310: 

1311: Figures 10--11 show the fundamental p-mode frequencies and growth rates

1312: as a function of the disc parameters. Consistent with the result of

1313: section 5.3 (see Fig.~8), growing modes are obtained for sufficiently

1314: small $L_\Sigma$. Negative $L_c$ also tends to reduce wave loss

1315: at $r_s$ and make the growing modes possible.

1316: Such values of $L_\Sigma$ and $L_c$ are not unreasonable for 

1317: black hole accretion discs.

1318: 

1319: It is important to note that while the mode growth rates $\omega_i$ depend

1320: sensitively on the inner disc parameters, particularly the physical

1321: property of the transonic flow near the ISCO, the real mode

1322: frequencies $\omega_r$ show only weak dependence on

1323: the inner disc parameters (e.g., $\omega_r$ decreases with increasing

1324: sound speed; see Fig.~6).

1325:  Thus we may expect that kHz

1326: QPOs appear only in certain accretion states of the black hole,

1327: and the frequencies do not vary much as the accretion rate changes.

1328: 

1329: 

1330: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1331: \section{The Role of Rossby Wave Instability}

1332: 

1333: Lovelace et al.~(1999) (see also Li et al.~2000) have shown that 

1334: when the vortensity $\zeta=\kappa^2/(2\Omega\Sigma)$ has an extremum

1335: at a certain radius ($r_{\rm peak}$) in the disc\footnote{Lovelace et al.

1336: considered non-barotropic flows, so the ``generalized vortensity'' depends

1337: on the entropy profile of the disc.}, it is possible to form 

1338: normal Rossby modes around $r_{\rm peak}$. If the trapped Rossby waves

1339: can propagate on both sides of the corotation, a standing pattern of waves of 

1340: opposite energies are formed, making the mode unstable --- This is the

1341: ``Rossby wave instability''. Tagger \& Varniere (2006) have considered

1342: the MHD version of the instability and suggested that it played a role

1343: in the diskoseismic modes around black holes (see also Tagger 2006).

1344: 

1345: We do not find any trapped Rossby modes in our calculation. To clarify the

1346: issue in light of works by Lovelace et al. and by Tagger \& Varniere,

1347: let us consider equation (\ref{perturbeq1}) and define the effective potential

1348: \be

1349: V_{\rm eff}(r)=

1350: \frac{2m\Omega}{r\tomega}\left(\frac{d}{dr}\ln\frac{\Omega\Sigma}{D}\right)

1351: +\frac{m^2}{r^2}+ \frac{D}{c_s^2}.

1352: \ee

1353: The wave equation can be approximated by $(d^2/dr^2-V_{\rm eff})\delta h\simeq 0$

1354: (see Tsang \& Lai 2008a). We will focus on modes with $r_c$ very close to $r_{\rm peak}$ 

1355: (i.e., $|r_c-r_{\rm peak}|\ll r_c$, so that 

1356: $\omega_r\simeq m\Omega_{\rm peak}$; see Fig.~1). For 

1357: $|r-r_{\rm peak}|\ll r_{\rm peak}$ in a thin disc (so that $m^2/r^2$ can be

1358: neglected compared to $\kappa^2/c_s^2$), the effective potential becomes

1359: \be

1360: V_{\rm eff}(r)\simeq -\frac{2m\Omega}{r\tomega}\left(\frac{d}{dr}\ln\zeta

1361: \right)+{\kappa^2\over c_s^2}

1362: \simeq {2\over qL_\zeta^2}\left({r-r_{\rm peak}\over r-R_c}\right)

1363: +{\kappa^2\over c_s^2},

1364: \label{eq:veff2}\ee

1365: where in the second equality we have used 

1366: $R_c=r_c-i(r_c\omega_i/q\omega_r)$, $\Omega\propto r^{-q}$, and 

1367: defined $L_\zeta$ via

1368: \be

1369: {d\ln\zeta\over dr}=-{r-r_{\rm peak}\over L_\zeta^2},\qquad ({\rm for}~

1370: |r-r_{\rm peak}|\ll r_{\rm peak}).

1371: \ee

1372: Consider the case $r_c<r_{\rm peak}$ and assume $\omega_i\ll \omega_r$. 

1373: The Rossby wave zone (where $V_{\rm eff}<0$) lies between $r_c$ and $r_c+\Delta r_R$, 

1374: with 

1375: \be

1376: \Delta r_R\simeq {r_{\rm peak}-r_c\over (q/2)(L_\zeta/H)^2},

1377: \ee

1378: where $H\simeq c_s/\kappa$, and we have used $L_\zeta/H\sim r/H\gg 1$. The

1379: number of wavelengths in the Rossby zone is 

1380: \be

1381: \int_{r_c}^{r_c+\Delta r_R}\!\!k\,dr=

1382: \int_{r_c}^{r_c+\Delta r_R}\!\!(-V_{\rm eff})^{1/2}\,dr

1383: \sim {4H (r_{\rm peak}-r_c)\over qL_\zeta^2}

1384: \ee

1385: Two points should be noted: (i) Since $\int_{r_c}^{r_c+\Delta r_R}\!k\,dr\ll 1$

1386: for $L_\zeta\sim r_{\rm peak}$, no stationary wave can form in the Rossby zone;

1387: (ii) since the Rossby zone lies only on one side of the corotation radius,

1388: even if the mode can be trapped it will not grow by the Rossby wave instability

1389: mechanism. Similar result can be obtained for the $r_c>r_{\rm peak}$ case.

1390: We conclude that for the ``smooth'' vortensity maximum (with length scale

1391: $L_\zeta\sim r$; see the lower panel of Fig.~1) considered in this paper,

1392: there is no trapped Rossby mode around $r_{\rm peak}$ 

1393: and the Rossby wave instability is ineffective.

1394: 

1395: In the hypothetical situation where the vortensity $\zeta$ has a {\it minimum}

1396: at $r=r_{\rm min}$, equation (\ref{eq:veff2}) should be replaced by

1397: \be

1398: V_{\rm eff}(r)\simeq -{2\over qL_\zeta^2}\left({r-r_{\rm min}\over r-R_c}\right)

1399: +{\kappa^2\over c_s^2},

1400: \ee

1401: where we have used 

1402: \be

1403: {d\ln\zeta\over dr}={r-r_{\rm min}\over L_\zeta^2},\qquad ({\rm for}~

1404: |r-r_{\rm min}|\ll r_{\rm min}).

1405: \ee

1406: In this case, for a mode with $\omega_r=m\Omega(r_{\rm min})$

1407: (or $r_c=r_{\rm min}$), we find 

1408: $V_{\rm eff}(r_c)\simeq -2/(qL_\zeta^2)+1/H^2$ (for $\omega_i\ll\omega_r$). 

1409: When $V_{\rm eff}(r_c)<0$, or when

1410: \be

1411: L_\zeta<\left({2\over q}\right)^{1/2}\!\!H,

1412: \ee

1413: Rossby waves can propagate on both sides of 

1414: the corotation, leading to mode growth

1415: --- this is the Rossby wave instability.

1416: Thus, the Rossby wave instability would operate if there existed 

1417: a ``sharp'' vortensity {\it minimum} in the disc (with $\zeta$ varying on the 

1418: lengthscale comparable or less than the disc thickness)

1419: --- this is not the case for typical black hole accretion discs

1420: considered in this paper.

1421: 

1422: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1423: \section{Discussion}

1424: 

1425: High-frequency QPOs (HFQPOs) in black-hole X-ray binaries have been studied

1426: observationally for more than a decade now and they provide a

1427: potentially important tool for studying the strong gravitational

1428: fields of black holes (see Remillard \& McClintock 2006).

1429: Despite much theoretical effort,

1430: the physical mechanisms that generate these QPOs remain unclear

1431: (see section 1.1 for a brief review of existing theoretical models).

1432: Ultimately, numerical simulations of realistic accretion discs

1433: around black holes may provide the answer.

1434: However, such simulations are still at their early stage of

1435: development and have their own limitations (e.g., De Villiers \& Hawley 2003; 

1436: Machida \& Matsumoto 2003, 2008; Arras et al.~2006; Fragile et al.~2007;

1437: Reynolds \& Miller 2008; Beckwith et al.~2008; Shafee et al.~2008;

1438: Noble et al.~2008), semi-analytical study remains a useful,

1439: complementary approach in order to identify the key physics involved.

1440: 

1441: In this paper, we have studied the global instability of the

1442: non-axisymmetric p-modes in black-hole accretion discs.  These modes

1443: have frequencies $\omega\sim (0.5-0.7) m\Omega_{\rm ISCO}$ (where $m$

1444: is the azimuthal wave number, $\Omega_{\rm ISCO}$ is the disc rotation

1445: frequency at the inner-most stable circular orbit), where the

1446: pre-factor (0.5-0.7) depends on the inner disc structure.  Recent

1447: works (Arras et al.~2006; Reynolds \& Miller 2008; Fu \& Lai 2008)

1448: suggested that, unlike other diskoseismic modes (g-modes and c-modes),

1449: the p-modes may be robust in the presence of disc magnetic fields and

1450: turbulence. Our linear analysis showed that due to GR effects, the

1451: p-modes may grow in amplitude due to wave absorptions at the corotation

1452: resonance. For a given $m$, only the lowest-order p-mode

1453: has sufficiently high frequency ($\omega>m\Omega_{\rm peak}$; see Fig.~1) 

1454: to be driven overstable by the corotational instability, while high-order

1455: (lower frequency) modes are damped by the corotational wave absorption.

1456: 

1457: The greatest uncertainty of our calculation of the p-mode growth rate

1458: concerns the boundary condition at the inner disc edge near the ISCO.

1459: In particular, the rapid radial inflow at the ISCO has the tendency to

1460: damp the mode (see Blaes 1986). While our analysis in section 5

1461: indicates that this damping does not completely suppress the mode

1462: growth under certain disc conditions, it suggests that mode growth may

1463: not always be achieved in real black-hole accretion 

1464: discs. Observationally, it is of interest to note that HFQPOs

1465: are observed only when the X-ray binaries are in the steep

1466: power-law state, while they do not appear in other spectral states

1467: (Remillard \& McClintock 2006). In particular, HFQPOs are absent

1468: in the thermal (soft-high) state, believed to correspond to geometrically thin 

1469: discs extending down to the ISCO. It is reasonable to expect that 

1470: in this state p-modes are damped due to the rapid radial inflow.

1471: 

1472: Our current understanding of the steep power-law state (also called

1473: very high state) of black-hole X-ray binaries is rather limited.  A

1474: thermal-radiation-emitting disc is suggested by spectral modelings,

1475: but it is not clear whether the disc is truncated at the ISCO or

1476: slightly larger radius (see Done et al.~2007). The observed power-law

1477: radiation component requires a significant corona that Compton

1478: up-scatters the disc thermal radiation. It is possible that in the

1479: steep power-law state, the inner disc behaves as a more reflective

1480: boundary (modeled in section 4) than a transonic flow (modeled in section 5), 

1481: and thus more robust p-mode growth can be achieved. 

1482: One possibility is that a significant magnetic field

1483: flux can accumulate in the inner disc when the disc accretion rate is

1484: sufficiently high (see Bisnovtyi-Kogan \& Lovelace 2007; Rothstein \& Lovelace 2008

1485: and references therein). Such a magnetic field may also

1486: enhance the corotational instability and induce variability 

1487: in the power-law radiation flux (see Tagger \& Varniere 2006).

1488: 

1489: Although the p-mode growth rates depend sensitively on a number of

1490: (uncertain) disc parameters (particularly those related to the inner

1491: disc boundary), the mode frequencies are more robust (see Figs.~5-7,

1492: 10-11). More precisely, the real mode frequency can be written as

1493: $\omega_r={\bar\omega}m\Omega_{\rm ISCO}$, where $\bar\omega<1$

1494: depends weakly on $m$ and has only modest dependence on 

1495: disc parameters (e.g. sound speed). This implies a

1496: commensurate frequency ratio as observed in HFQPOs (note that in some

1497: of our models, the $m=2,3$ modes have the largest growth rates; see

1498: Fig.~5). The fact that $\bar\omega<1$ would also make the numerical values

1499: of the p-mode frequencies more compatible with the measurements of the QPO

1500: frequencies and black hole masses. 

1501: 

1502: We note that our calculations in this work are done with a pseudo-Newtonian potential. For direct comparison with observations a fully general relativistic calculation\footnote{Previous work on relativistic diskoseismic g-modes has been done by Perez et al. (1992) and Silbergleit \& Wagoner (2008) while the c-mode was studied by Silbergleit et al. (2001). Axisymmetric p-modes were studied using a general relativistic formalism by Ortega-Rodriguez et al. (2002), but these do not include the effect of the corotation singularity.} is needed including a careful treatment of the corotation singularity. Including the effect of black hole spin would likely increase the value of $\omega$ by modifying $r_{\rm ISCO}$ and $\Omega_{\rm ISCO}$, while $\bar\omega$ will likely remain similar to the non-spinning case discussed above. We plan to study these effects in future work.

1503: 

1504: 

1505: \section*{Acknowledgments}

1506: 

1507: We thank Richard Lovelace for useful discussion. 

1508: This work has been supported in part by NASA Grant NNX07AG81G, NSF

1509: grants AST 0707628, and by {\it Chandra} grant TM6-7004X

1510: (Smithsonian Astrophysical Observatory).

1511: 

1512: 

1513: %\bibliographystyle{mn.bst}

1514: %\bibliography{pmodes}

1515: 

1516: %%%%%%%%%%%%%%%%%%%%%%

1517: \begin{thebibliography}{99}

1518: 

1519: \bibitem[]{}

1520: Afshordi, N., Paczynski, B. 2003, ApJ, 592, 354

1521: % Geometrically Thin Disk Accreting into a Black Hole

1522: 

1523: \bibitem[]{}

1524: Abramowicz, M.A., Czerny, B., Lasota, J. P., Szuszkiewicz, E. 1988,

1525: ApJ, 332, 646

1526: 

1527: \bibitem[]{}

1528: Abramowicz, M.A., Kluzniak, W. 2001, A\&A, 374, L19

1529: 

1530: \bibitem[]{}

1531: Abramowicz, M.A. et al 2007, Rev. Mexicana Astron. Astrofísica, 27, 8

1532: 

1533: \bibitem[]{}

1534: Arras, P., Blaes, O.M. \& Turner, N. J., 2006, ApJ, 645, L65

1535: 

1536: \bibitem[]{}

1537: Beckwith, K., Hawley, J.F., Krolik, J.H. 2008, MNRAS, in press

1538: (arXiv:0801.2974)

1539: %Where is the Radiation Edge in Magnetized Black Hole Accretion discs?

1540: 

1541: \bibitem[]{}

1542: Blaes, O.M. 1987, MNRAS, 227, 975

1543: 

1544: \bibitem[]{}

1545: Blaes, O.M., Arras, P., Fragile, P.C. 2006, MNRAS, 369, 1235

1546: %Oscillation modes of relativistic slender tori

1547: 

1548: \bibitem[]{}

1549: Blaes, O.M., Sramkova, E., Abramowicz, M. A., Kluzniak, W., Torkelsson, U., ApJ, 665, 642

1550: 

1551: \bibitem[]{}

1552: Balbus, S.A., Hawley, J.F. 1998, Rev. Mod. Phys., 70, 1.

1553: 

1554: \bibitem[]{}

1555: Bisnovatyi-Kogan, G.S., Lovelace, R.V.E. 2007, ApJ, 667, L167

1556: % Large-Scale B-Field in Stationary Accretion Disks

1557: 

1558: \bibitem[]{}

1559: De Villiers, J.-P., Hawley, J.F. 2003, ApJ, 592, 1060

1560: 

1561: \bibitem[]{}

1562: Done, C., Gierlinski, M., Kubota, A. 2007, Astron. Astrophys. Review, 15, 1

1563: 

1564: \bibitem[]{}

1565: Gierlinski, M., Middleton, M., Ward, M., Done, C. 2008, Nature, 455, 369

1566: % A periodicity of 1 hour in X-ray emission from the active galacy 

1567: % RE J1034_396

1568: 

1569: \bibitem[]{}

1570: Goldreich, P., Tremaine, S. 1979, ApJ, 233, 857

1571: 

1572: \bibitem[]{}

1573: Ferreira, B. T. \& Ogilvie, G. I., 2008, astro-ph/08031671

1574: 

1575: \bibitem[]{}

1576: Fragile, P.C., Blaes, O., Anninos, P., Salmonson, J.D. 2007, ApJ, 668, 417

1577: % Global General Relativistic Magnetohydrodynamic Simulation of a Tilted 

1578: % Black Hole Accretion Disk

1579: 

1580: \bibitem[]{}

1581: Fu, W., Lai, D. 2008, ApJ, in press (arXiv:0806.1938)

1582: 

1583: \bibitem[]{}

1584: Horak, J., Karas, V., 2006, A\&A, 451, 377

1585: 

1586: \bibitem[]{}

1587: Kato, S. \& Fukue, J., 1980, PASJ, 32, 377

1588: 

1589: \bibitem[]{}

1590: Kato, S., 1990, PASJ, 42, 99

1591: 

1592: \bibitem[]{}

1593: Kato, S.,  2001, PASJ, 53, 1

1594: 

1595: \bibitem[]{}

1596: Kato, S., 2003a, PASJ, 55, 257

1597: % g-mode damping

1598: 

1599: \bibitem[]{}

1600: Kato, S., 2003b, PASJ, 55, 801

1601: % Excitation by disk deformation

1602: 

1603: \bibitem[]{}

1604: Kato, S., 2008, PASJ, 60, 111

1605: 

1606: \bibitem[]{}

1607: Kluzniak, W. \& Abramowicz, M. A., 2002, astro-ph/0203314

1608: 

1609: \bibitem[]{}

1610: Lee, W. H., Abramowicz, M. A. \& Kluziniak, W., 2004, ApJ, 603, L93

1611: %Resonance in Forced Oscillations of an Accretion Disk and Kilohertz

1612: %Quasi-periodic Oscillations

1613: 

1614: \bibitem[]{}

1615: Li, H., Finn, J.M., Lovelace, R.V.E., Colgate, S.A. 2000, ApJ, 533, 1023

1616: 

1617: \bibitem[]{}

1618: Li, L., Goodman, J., Narayan, R., 2003, ApJ, 593, 980

1619: 

1620: \bibitem[]{}

1621: Li, L., Narayan, R., 2004, ApJ, 601, 414

1622: %Quasi-periodic Oscillations from Rayleigh-Taylor and Kelvin-Helmholtz

1623: %Instability at a Disk-Magnetosphere Interface

1624: 

1625: \bibitem[]{}

1626: Lifshitz, E.M., Pitaevskii, L.P. 1981, Physical Kinetics (Pergamon Press: Oxford)

1627: 

1628: \bibitem[]{}

1629: Lin, C.C. 1995, The Theory of Hydrodynamic Stability (Cambridge Univ. Press), 

1630: Chap.~8

1631: 

1632: \bibitem[]{}

1633: Lovelace, R.V.E., Li, H., Colgate, S.A., Nelson, A.F. 1999,

1634: ApJ, 513, 805

1635: 

1636: \bibitem[]{}

1637: Lovelace, R.V.E., Romanova, M.M., 2007, ApJ, 670, L13

1638: %Three Disk Oscillation Modes of Rotating Magnetized Neutron Stars

1639: 

1640: \bibitem[]{}

1641: Machida, M., Matsumoto, R., 2003, ApJ, 585, 429

1642: 

1643: \bibitem[]{}

1644: Machida, M., Matsumoto, R., 2008, PASJ, 60, 613

1645: 

1646: \bibitem[]{}

1647: Matsumoto, R., Kato, S., Fukue, J., Okazaki, A. T., 1984, PASJ, 36, 71

1648: 

1649: \bibitem[]{}

1650: Muchotrzeb, B., Paczynski, B. 1982, Acta Astron., 32, 1

1651: 

1652: \bibitem[]{}

1653: Narayan, R., Goldreich, P., Goodman, J. 1987, MNRAS, 228, 1

1654: 

1655: \bibitem[]{}

1656: Noble, S.C., Krolik, J.H., Hawley, J.F. 2008, ApJ, submitted 

1657: (arXiv:0808.3140)

1658: 

1659: \bibitem[]{}

1660: Nowak, M. A. \& Wagoner, R. V., 1991, ApJ, 378, 656

1661: 

1662: \bibitem[]{}

1663: Nowak, M. A. \& Wagoner, R. V., 1992, ApJ, 393, 697

1664: 

1665: \bibitem[]{}

1666: Okazaki, A. T., Kato, S. \& Fukue, J., 1987, PASJ, 39, 457

1667: 

1668: \bibitem[]{}

1669: Ortega-Rodriguez, M., Silbergleit, A. S. \& Wagoner, R. V., 2002, ApJ, 567, 1043

1670: 

1671: \bibitem[]{}

1672: Ortega-Rodriguez, M., Silbergleit, A. S. \& Wagoner, R. V., 2006, astro-ph/0611101

1673: 

1674: \bibitem[]{}

1675: Paczynski, B., Wiita, P.J. 1980, A\&A, 88, 23

1676: 

1677: \bibitem[]{}

1678: Petri, J. 2008, Astrophys. Space Science, in press (arXiv:0809.3115)

1679: 

1680: \bibitem[]{}

1681: Perez, C.A., Silbergleit, A.S., Wagoner, R.V., \& Lehr, D.E., 1997, ApJ, 476, 589

1682: 

1683: \bibitem[]{}

1684: Press, W.H., et al. 1998, Numerical Recipes (Cambridge Univ. Press)

1685: 

1686: \bibitem[]{}

1687: Rebusco, P., 2004, PASJ, 56, 553

1688: 

1689: \bibitem[]{}

1690: Rebusco, P., 2008, astro-ph/08013658

1691: 

1692: \bibitem[]{}

1693: Remillard, R. A. \& McClintock, J. E., 2006, ARAA, Vol. 44, pp. 49-92

1694: 

1695: \bibitem[]{}

1696: Reynolds, C. S. \& Miller, M. C., 2008, astro-ph/08052950

1697: 

1698: \bibitem[]{}

1699: Rezzolla, L., Yoshida, S'i., Maccarone \& Zanotti, O., 2003, MNRAS, 344, L37-L41

1700: %A new simple model for high-frequency quasi-periodic oscillations in 

1701: %black hole candidates

1702: 

1703: \bibitem[]{}

1704: Rothstein, D.M., Lovelace, R.V.E. 2008, ApJ, 677, 1221

1705: %Advection of Magnetic Fields in Accretion Disks: Not So Difficult After All

1706: 

1707: \bibitem[]{}

1708: Shafee, R., et al. 2008, ApJL, submitted (arXiv:0808.2860)

1709: %Three-Dimensional Simulations of Magnetized Thin Accretion 

1710: %Disks around Black Holes: Stress in the Plunging Region

1711: 

1712: \bibitem[]{}

1713: Silbergleit, A. S., Wagoner, R. V. \& Ortega-Rodriguez, M., 2001, ApJ, 548, 335

1714: 

1715: \bibitem[]{}

1716: Silbergleit, A. S. \& Wagoner, R. V., 2008, ApJ, 680, 1319

1717: 

1718: \bibitem[]{}

1719: Schnittman, J.D. 2005, ApJ, 621, 940 

1720: 

1721: \bibitem[]{}

1722: Schnittman, J.D., Bertschinger, E. 2004, ApJ, 606, 1098 

1723: 

1724: \bibitem[]{}

1725: Schnittman, J.D., Rezzolla, L. 2006, ApJ, 637, L113

1726: %Quasi-periodic Oscillations in the X-Ray Light Curves from Relativistic Tori

1727: 

1728: \bibitem[]{}

1729: Shu, F.H. 1992, The Physics of Astrophysics II: Gas Dynamics (University

1730: Science Books), Chap.~12

1731: 

1732: \bibitem[]{}

1733: Sramkova, E., Torkelsson, U., Abramowicz, M. A., 2007, A\&A, 467, 641

1734: 

1735: \bibitem[]{}

1736: Stella, L., Vietri, M., Morskink, S.M. 1999, ApJ, 524, L63 

1737: 

1738: \bibitem[]{}

1739: Strohmayer, T.E. 2001, ApJ, 552, L49

1740: % Discovery of a 450 HZ Quasi-periodic Oscillation from the Microquasar 

1741: % GRO J1655-40 with the Rossi X-Ray Timing Explorer

1742: 

1743: \bibitem[]{}

1744: Swank, J. 1999, Nucl. Phys. B, Proc. Suppl., 69, 12 (astro-ph/9802188)

1745: 

1746: \bibitem[]{}

1747: Tagger, M. 2006, arXiv:astro-ph/0612175

1748: %Global MHD instabilities: from Low Frequency to High Frequency QPOs, 

1749: %and to Sgr A*

1750: 

1751: \bibitem[]{}

1752: Tagger, M. 2006, in the proceedings of the VI Microquasar Workshop 

1753: ``Microquasars and beyond'', ed. T. Belloni (arXiv:astro-ph/0612175)

1754: 

1755: \bibitem[]{}

1756: Tagger, M., Pellat, R. 1999, A\&A, 349, 1003

1757: 

1758: \bibitem[]{}

1759: Tagger, M., Varniere, P. 2006, ApJ, 652, 1457

1760: 

1761: \bibitem[]{}

1762: Tassev, S.V., Bertschinger, E. 2007, ApJ, submitted (arXiv:0711.0065)

1763: 

1764: \bibitem[]{}

1765: Tsang, D. \& Lai, D., 2008a, MNRAS, 387, 446

1766: 

1767: \bibitem[]{}

1768: Tsang, D. \& Lai, D., 2008b, MNRAS, submitted

1769: % The c-mode paper

1770: 

1771: \bibitem[]{}

1772: Varniere, P., Tagger, M. 2002, A\&A, 394, 329

1773: % Accretion-Ejection Instability in magnetized disks: Feeding the corona 

1774: % with Alfven waves

1775: 

1776: \bibitem[]{}

1777: van der Klis, M. 2006, in Compact Stellar X-ray Sources, ed. 

1778: W.H.G. Lewin and M. van der Klis (Cambridge Univ. Press) 

1779: (astro-ph/0410551)

1780: 

1781: \bibitem[]{}

1782: Wagoner, R. V., 1999, Phys. Rep., 311, 259

1783: 

1784: \bibitem[]{}

1785: Zhang, H., Lai, D., MNRAS, 2006, 368, 917

1786: 

1787: \end{thebibliography}

1788: 

1789: 

1790: \end{document}

1791: