1: %\documentclass[12pt,preprint]{aastex}

2: \documentclass[12pt]{emulateapj}

3: %\usepackage{graphicx}

4: %\usepackage{caption2}

5: %\usepackage{amssymb}

6: %\usepackage{times}

7: %\usepackage{emulateapj}

8: %\usepackage{aasms4}

9: \usepackage{epsfig}

10: 

11: %\topmargin=1cm

12: 

13: \def\AaA{{\em Astr.~Astrophys.}}

14: \def\AJ{{\em Astr.~J.}}

15: \def\AN{{\em Astron.~Nachr.}}

16: \def\ApJ{{\em Astrophys.~J.}}

17: \def\ApJL{{\em Astrophys.~J.~Lett.}}

18: \def\ApJS{{\em Astrophys.~J.~Suppl.}}

19: \def\ARAA{{\em Ann.~Rev.~Astr.~Astrophys.}}

20: \def\MN{{\em Mon.~Not.~R.~astr.~Soc.}}

21: \def\Nat{{\em Nature}}

22: \def\PASJ{{\em Publ.~astr.~Soc.~Japan}}

23: \def\PASP{{\em Publ.~astr.~Soc.~Pacif.}}

24: \def\PhD{{\em PhD thesis}}

25: \def\Pre{Preprint}

26: \def\PS{{\em Phys.~Scr.}}

27: \def\RMP{{\em Rev.~Mod.~Phys.}}

28: \def\SSR{{\em Space Sci.~Rev.}}

29: \def\etal{{et al.\thinspace}}

30: \def\cf{{\em cf.\ }}

31: \def\eg{{\em e.g.\ }}

32: \def\etc{{\em etc.\ }}

33: \def\ie{{\em i.e.\ }}

34: \def\spose#1{\hbox to 0pt{#1\hss}}

35: 

36: \let\approxlt=\lesssim

37: \let\approxgt=\gtrsim

38: \def\multcent#1{\halign {\centerline{##}\cr #1}}

39: \def\multleft#1{\hbox to size{\vbox {\halign {\lft{##}\cr #1}}\hfill}\par}

40: \def\multright#1{\hbox to size{\vbox {\halign {\rt{##}\cr #1}}\hfill}\par}

41: \def\Mdot{\hbox{$\dot M$}}

42: \def\degmark{^\circ}

43: \def\boxit#1{\vbox{\hrule\hbox{\vrule\kern3pt\vbox{\kern3pt

44:           #1 \kern3pt}\kern3pt\vrule}\hrule}}

45: \font\big=cmr10 scaled\magstep2

46: \font\bigbf=cmbx10 scaled\magstep2

47: \font\bigit=cmti10 scaled\magstep2

48: 

49: %       Simple units

50: 

51: \def\cm{{\rm\thinspace cm}}

52: \def\dyn{{\rm\thinspace dyn}}

53: \def\erg{{\rm\thinspace erg}}

54: \def\eV{{\rm\thinspace eV}}

55: \def\g{{\rm\thinspace g}}

56: \def\ga{{\rm\thinspace gauss}}

57: \def\K{{\rm\thinspace K}}

58: \def\keV{{\rm\thinspace keV}}

59: \def\km{{\rm\thinspace km}}

60: \def\kpc{{\rm\thinspace kpc}}

61: \def\Lsun{\hbox{$\rm\thinspace L_{\odot}$}}

62: \def\m{{\rm\thinspace m}}

63: \def\Mpc{{\rm\thinspace Mpc}}

64: \def\Msun{\hbox{$\rm\thinspace M_{\odot}$}}

65: \def\pc{{\rm\thinspace pc}}

66: \def\ph{{\rm\thinspace ph}}

67: \def\s{{\rm\thinspace s}}

68: \def\yr{{\rm\thinspace yr}}

69: \def\sr{{\rm\thinspace sr}}

70: \def\Hz{{\rm\thinspace Hz}}

71: \def\chisq{\hbox{$\chi^2$}}

72: \def\delchi{\hbox{$\Delta\chi$}}

73: 

74: %       Compound units

75: 

76: \def\cmps{\hbox{$\cm\s^{-1}\,$}}

77: \def\cmsq{\hbox{$\cm^2\,$}}

78: \def\cmcu{\hbox{$\cm^3\,$}}

79: \def\pcmcu{\hbox{$\cm^{-3}\,$}}

80: \def\pcmcuK{\hbox{$\cm^{-3}\K\,$}}

81: \def\dynpcmsq{\hbox{$\dyn\cm^{-2}\,$}}

82: \def\ergcmcups{\hbox{$\erg\cm^3\ps\,$}}

83: \def\ergpcmps{\hbox{$\erg\cm^{-3}\s^{-1}\,$}}

84: \def\ergpcmsqps{\hbox{$\erg\cm^{-2}\s^{-1}\,$}}

85: \def\ergpcmsqpspA{\hbox{$\erg\cm^{-2}\s^{-1}$\AA$^{-1}\,$}}

86: \def\ergps{\hbox{$\erg\s^{-1}\,$}}

87: \def\gpcm{\hbox{$\g\cm^{-3}\,$}}

88: \def\gpcmps{\hbox{$\g\cm^{-3}\s^{-1}\,$}}

89: \def\gps{\hbox{$\g\s^{-1}\,$}}

90: \def\kmps{\hbox{$\km\s^{-1}\,$}}

91: \def\Lsunppc{\hbox{$\Lsun\pc^{-3}\,$}}

92: \def\Msunpc{\hbox{$\Msun\pc^{-3}\,$}}

93: \def\Msunpkpc{\hbox{$\Msun\kpc^{-1}\,$}}

94: \def\Msunppc{\hbox{$\Msun\pc^{-3}\,$}}

95: \def\Msunppcpyr{\hbox{$\Msun\pc^{-3}\yr^{-1}\,$}}

96: \def\Msunpyr{\hbox{$\Msun\yr^{-1}\,$}}

97: \def\pcm{\hbox{$\cm^{-3}\,$}}

98: \def\pcmsq{\hbox{$\cm^{-2}\,$}}

99: \def\pcmK{\hbox{$\cm^{-3}\K$}}

100: \def\phpcmsqps{\hbox{$\ph\cm^{-2}\s^{-1}\,$}}

101: \def\pHz{\hbox{$\Hz^{-1}\,$}}

102: \def\pmpc{\hbox{$\Mpc^{-1}\,$}}

103: \def\pmpccu{\hbox{$\Mpc^{-3}\,$}}

104: \def\ps{\hbox{$\s^{-1}\,$}}

105: \def\psqcm{\hbox{$\cm^{-2}\,$}}

106: \def\psr{\hbox{$\sr^{-1}\,$}}

107: \def\pyr{\hbox{$\yr^{-1}\,$}}

108: \def\kmpspMpc{\hbox{$\kmps\Mpc^{-1}$}}

109: \def\Msunpyrpkpc{\hbox{$\Msunpyr\kpc^{-1}$}}

110: 

111: \makeatletter

112: 

113: \makeatother

114: 

115: \begin{document}

116: 

117: \title{On the time variability of geometrically-thin black hole

118: accretion disks I : the search for modes in simulated disks.}

119: 

120: \author{Christopher~S.~Reynolds\altaffilmark{1} and 

121: M.~Coleman~Miller\altaffilmark{1}}

122: 

123: \altaffiltext{1}{Department of Astronomy and the Maryland Astronomy

124: Center for Theory and Computation, University of Maryland, College

125: Park, MD 20742-2421}

126: 

127: \begin{abstract}

128: We present a detailed temporal analysis of a set of hydrodynamic and

129: magnetohydrodynamic (MHD) simulations of geometrically-thin ($h/r\sim

130: 0.05$) black hole accretion disks.  The black hole potential is

131: approximated by the Paczynski-Wiita pseudo-Newtonian potential.  In

132: particular, we use our simulations to critically assess two widely

133: discussed models for high-frequency quasi-periodic oscillations,

134: global oscillation modes (diskoseismology) and parametric resonance

135: instabilities.  We find that initially disturbed hydrodynamic disks

136: clearly display the trapped global g-mode oscillation predicted by

137: linear perturbation theory.  In contrast, the sustained turbulence

138: produced in the simulated MHD disks by the magneto-rotational

139: instability does not excite these trapped g-modes.  We cannot say at

140: present whether the MHD turbulence actively damps the hydrodynamic

141: g-mode.  Our simulated MHD disks also fail to display any indications

142: of a parametric resonance instability between the vertical and radial

143: epicyclic frequencies.  On the other hand, we do see characteristic

144: frequencies at any given radius in the disk corresponding to local

145: acoustic waves.  We also conduct a blind search for {\it any}

146: quasi-periodic oscillation in a proxy lightcurve based on the

147: instantaneous mass accretion rate of the black hole, and place an

148: upper limit of 2\% on the total power in any such feature. We

149: highlight the importance of correcting for secular changes in the

150: simulated accretion disk when performing temporal analyses.

151: \end{abstract}

152: 

153: \keywords{{accretion: accretion disks --- black hole physics ---

154: magnetic fields --- X-ray: binaries}}

155: 

156: \section{Introduction}

157: 

158: Rapid X-ray variability is a ubiquitous characteristic of accretion

159: onto black holes.  Aperiodic X-ray fluctuations are seen from both

160: Galactic Black Hole Binaries (GBHBs) and Active Galactic Nuclei (AGN)

161: and, accounting for the inverse scaling of all relevant frequencies

162: with black hole mass, appear to have similar characteristics

163: (Uttley, McHardy \& Vaughan 2005; McHardy et al. 2006).  While it is

164: highly tempting to relate this variability to the magnetohydrodynamic

165: (MHD) turbulence that is believed to drive the accretion process

166: (Balbus \& Hawley 1991, 1998), the exact physical

167: processes underlying the observed fluctuations remain mysterious.

168: 

169: GBHBs also display quasi-periodic oscillations (QPOs) in their X-ray

170: lightcurves (see review by McClintock \& Remillard 2003)\footnote{Very

171: recently, the first convincing case of a QPO in an AGN was reported by

172: Gierlinski et al. (2008).}.  Of particular interest are the

173: high-frequency quasi-periodic oscillations (HFQPOs) that are seen in

174: the very-high (or steep power law; McClintock \& Remillard 2003) state

175: of GBHBs.  The HFQPOs have quality factors of few-to-$10$, centroid

176: frequencies of order $100\Hz$ and appear to be imprinted on the hard

177: X-ray tail of the spectrum rather than the thermal disk emission.  The

178: fact that their frequencies are stable and at least loosely comparable

179: to the orbital frequency at the innermost stable circular orbit (ISCO)

180: around the black hole suggests that their properties are set by the

181: relativistic portions of the gravitational potential.  This gives them

182: enormous promise as a diagnostic of black hole mass and spin.

183: 

184: However, the utility of HFQPOs to relativistic astrophysics is

185: severely limited by the lack of a compelling theoretical framework in

186: which to interpret measurements of the frequencies, quality factors

187: and root-mean-square (rms) powers.  There exist well defined geodesic

188: frequencies (i.e., the orbital, radial epicyclic and vertical

189: epicyclic frequencies) at any given radius in the accretion disk.

190: However, these frequencies (as well as all non-trivial linear

191: combinations) change with radius, and it is not clear why the

192: frequencies of any one particular radius would be preferentially

193: displayed in the overall power-spectrum.  

194: 

195: HFQPOs are commonly found in pairs with an approximate 3:2 frequency

196: ratio, and this has been used to suggest that a particular radius is

197: picked out due to a resonance.  In the parametric resonance model

198: (Abramowicz \& Klu\'zniak 2001, 2003; Abramowicz et al.  2002, 2003),

199: there is a resonance in the disk at the radius where the radial

200: epicyclic frequency and vertical epicyclic frequency are in small

201: integer ratios.  As discussed below, the strongest resonance occurs

202: when these frequencies are in 3:2 ratio, at least in the simplest

203: manifestation of this model.  Other resonance models have been

204: examined by Rezzolla et al. (2003a,b) and Kato (2004a,b,c).

205: 

206: Another interesting possibility is that the HFQPOs are global

207: oscillation modes of the accretion disk (i.e., ``diskoseismic''

208: modes).  Global modes have been examined analytically (using linear

209: theory) on a hydrodynamic background in spacetimes that are

210: pseudo-Newtonian (Okazaki, Kato, \& Fukue 1987; Nowak \& Wagoner 1991,

211: 1992, 1993; Markovi\'c \& Lamb 1998), Schwarzschild (Kato \& Fukue

212: 1980), and Kerr (Kato 1990, 1991, 1993; Kato \& Honma 1991; Perez et

213: al. 1997; Silbergleit et al. 2001; Wagoner et al. 2001;

214: Ortega-Rodriguez et al. 2001).  Three classes of mode are recovered

215: corresponding to pressure modes (p-modes), inertial modes

216: (conventionally referred to as g-modes even though the restoring force

217: results from rotation or inertia, depending on your frame of reference;

218: J.Pringle priv. communication) and warping/corrugation modes

219: (c-modes).  For plausible masses and spins, the fundamental ($m=0$)

220: g-mode was quickly identified as a good candidate for the first HFQPO

221: discovered, the 67\,Hz oscillation found in the system GRS~1915$+$105

222: (Nowak et al. 1997).  Current diskoseismology theory does not provide

223: a natural explanation of HFQPO pairs with small-integer ratios;

224: however, all present analyses are conducted using linear theory

225: whereas these HFQPOs pairs would likely arise from mode-coupling that

226: would only be revealed by a non-linear analysis.

227: 

228: Clearly, many open questions concerning the physics of X-ray

229: variability remain, including the correct interpretation of HFQPOs.

230: The current dominant paradigm for understanding black hole accretion

231: is that the magnetorotational instability (MRI; Balbus \& Hawley 1991)

232: drives powerful MHD turbulence, and correlated Maxwell stresses within

233: this turbulence mediate the outward transport of angular momentum

234: that allows accretion to proceed.  However, the connection between the

235: MHD turbulence paradigm and models for the aperiodic and

236: quasi-periodic variability remains highly uncertain.  For example, can

237: the MHD turbulence naturally produce the rms-flux relation noted in

238: most black hole X-ray lightcurves (Uttley \& McHardy 2001) and/or the

239: log-normal flux distribution found in Cygnus X-1 (Uttley, McHardy \&

240: Vaughan 2005)?  Are diskoseismic modes excited by turbulent

241: fluctuation (Nowak \& Wagoner 1993), or does the turbulence act to

242: damp such modes (Arras, Blaes \& Turner 2006)?  Do the Maxwell

243: stresses couple radial and vertical motions in such a way as to excite

244: parametric resonance instabilities of the type identified by

245: Abramowicz \& Klu\'zniak (2001) or any other resonant phenomena?  Does

246: the fact that the HFQPOs are imprinted on the high-energy tail provide

247: a fundamental clue to their origin, or is it a generic consequence of

248: any oscillating thermal accretion disk surrounded by a Comptonizing

249: corona (Lehr, Wagoner \& Wilms 2000)?

250: 

251: In this paper, we use a set of global hydrodynamic and MHD simulations

252: of geometrically-thin accretion disks in a pseudo-Newtonian potential

253: to begin an exploration of these issues.  Our canonical MHD simulation

254: represents a thinner disk, and is run for more orbits, than any

255: previously published well-resolved 3-d MHD disk simulation.  This

256: allows us to conduct a more extensive study of the temporal

257: variability of such disks than has previously been attempted.  In

258: \S~2, we give a brief review of the theory of both local and global

259: hydrodynamic modes of black hole accretion disks, as well as the

260: parametric resonance instability model for HFQPOs.  \S~3 presents our

261: study of ideal (inviscid) hydrodynamic disks, both with imposed

262: axisymmetry and in full 3-dimensions.  We find prominent trapped

263: g-modes in the axisymmetric simulations which remain (albeit with

264: diminished amplitude) in the full 3-d case.  We then study the MHD

265: case in \S~4, where we find that the turbulence excites neither the

266: diskoseismic modes nor the parametric resonances discussed above.

267: Instead, we find that the turbulence excites local hydrodynamic waves

268: of the type elucidated by Lubow \& Pringle (1993).  We discuss our

269: results, including a comparison to previous work, in \S~5 and conclude

270: in \S~6.

271: 

272: \section{Theoretical Preliminaries}

273: 

274: Here we provide a brief review of some previously established

275: theoretical results that are pertinent to this paper.

276: 

277: \subsection{Local oscillations and waves in accretion disks}

278: \label{sec:local_osc}

279: 

280: There is a very extensive literature on oscillations and waves in

281: accretion disks.  Here we focus on just those aspects of the field

282: that turn out to be relevant for the interpretation of our simulation

283: which have been elucidated most clearly by Lubow \& Pringle (1993).

284: 

285: Lubow \& Pringle (1993; hereafter LP93) studied three dimensional wave

286: propagation in accretion disks ignoring self-gravity.  In the case

287: where one ignores vertical motions, they show that radial waves obey

288: the well-known dispersion relation

289: \begin{equation}

290: \omega^2=\kappa^2+c_s^2k^2,

291: \label{eq:disk_dispersion}

292: \end{equation}

293: where $c_s$ is the sound speed (assumed, in this case, to be purely a

294: function of $r$) and $\kappa$ is the radial epicyclic frequency given

295: by $\kappa^2=4\Omega^2+r\,\partial \Omega^2/\partial r$ (also see

296: Binney \& Tremaine 1987).  Here, $\Omega(r)$ is the angular frequency

297: of the background Keplerian flow.  LP93 proceed to study the

298: propagation of axisymmetric waves in the case where the atmosphere has

299: a locally isothermal vertical structure.  They find two types of

300: waves.  There are low-frequency gravity waves for which

301: \begin{equation}

302: 0<\omega<\Omega.

303: \end{equation}

304: There are also high-frequency acoustic waves which have

305: \begin{equation}

306: \omega^2>(n\gamma+1)\Omega^2,

307: \end{equation}

308: where $n=0,1,2,\ldots$ and $\gamma$ is the adiabatic index.  

309: In the special case of purely vertical

310: perturbations, the inequality becomes an equality,

311: \begin{equation}

312: \omega^2=(n\gamma+1)\Omega^2.

313: \label{eq:vert_modes}

314: \end{equation}

315: The $n=0$ mode corresponds to a bulk vertical displacement of the disk

316: and subsequent vertical oscillation at the vertical epicyclic

317: frequency which, in the analysis of LP93 and in all analyses performed

318: in this paper, coincides with the orbital frequency.  In general, the

319: $n-th$ mode has $n$ vertical nodes (i.e., locations where the vertical

320: velocity perturbation vanishes), and is either even or odd depending

321: on whether $n$ is even or odd, respectively.

322: 

323: As discussed in \S~\ref{sec:mhd}, our simulations demonstrate the

324: effectiveness with which MHD turbulence excites these local acoustic

325: waves.

326: 

327: \subsection{Global oscillation modes of an accretion disk}

328: 

329: As discussed in the Introduction, several groups have studied the

330: global oscillation of black hole accretion disks using linear

331: perturbation theory, identifying three classes of normal mode

332: ($g$-modes, $p$-modes and $c$-modes).  Trapped $g$-modes have received

333: particular attention as a possible source of HFQPO, although the other

334: families of modes may well be relevant.  Here we review some of the

335: basic results of these analyses, following the approach of Nowak \&

336: Wagoner (1991, 1992; hereafter NW91 and NW92).  The NW91 and NW92

337: analyses are not fully relativistic, instead employing a

338: pseudo-Newtonian potential.  Thus, these analyses can be readily

339: compared with our pseudo-Newtonian simulations.  We also note that

340: full general relativistic MHD simulations have typically yielded

341: results for slowly rotating black holes that are very similar to those

342: obtained with pseudo-Newtonian potentials (e.g., Gammie, McKinney, \&

343: T\'oth 2003; De~Villiers \& Hawley 2003).

344: 

345: NW91 and NW92 use a Lagrangian formalism (Friedman \& Schutz 1978) and

346: a WKBJ approximation to derive the linearized equations describing

347: perturbations of an inviscid hydrodynamic thin accretion accretion

348: disk about a pure Keplerian background state.  They also initially

349: examined the special case of purely radial oscillations and found the

350: standard dispersion relation of disk theory

351: (eqn.~\ref{eq:disk_dispersion}).  Given that the gravity-modified

352: $p$-modes described by this dispersion relation become evanescent when

353: $\omega^2<\kappa^2$, global $p$-modes can be trapped between the ISCO

354: (where $\kappa=0$) and the radius at which $\omega=\kappa$.  In

355: practice, however, the ``leaky'' nature of the ISCO would seem to make

356: the trapping of these modes ineffective.

357: 

358: The more general case, including perturbations that have vertical as

359: well as radial motions, yields more promising results.  NW92 examine

360: the linearized equations describing the behavior of the scalar potential

361: \begin{equation}

362: \delta u\equiv \delta P/\rho,

363: \end{equation}

364: where $\delta P$ is the Eulerian variation in the pressure.  They show

365: that the linearized equations are approximately separable into radial

366: and vertical equations, with the separation constant being a slowly

367: varying function of $r$, $\Upsilon(r)$.  The general dispersion

368: relation for these modes becomes

369: \begin{equation}

370: [\omega^2-\gamma \Upsilon(r)\Omega^2](\omega^2-\kappa^2)=\omega^2c_s^2k^2\; .

371: \end{equation}

372: Assuming that $\gamma \Upsilon\Omega^2>\kappa^2$, non-evanescent

373: solutions exist for $\omega^2>\gamma\Upsilon\Omega^2$ (predominantly

374: radial $p$-modes) or $\omega^2<\kappa^2$ (predominantly vertical

375: $g$-modes).  Through this analysis, NW92 identify a class of global

376: $g$-modes that are trapped between two evanescent regions, $r<r_-$ and

377: $r>r_+$, where $\kappa(r_\pm)=\omega$.  In other words, these modes

378: are trapped under the peak of the epicyclic frequency.  They

379: principally focus on the $m=0$ (axisymmetric) modes, and show that the

380: mode frequency is only {\it slightly} smaller than the maximum radial

381: epicyclic frequency $\kappa_{\rm max}$.  Radial harmonics of these

382: modes are very closely spaced.  Thus, the inner radius at which the

383: mode becomes evanescent is still a finite distance (and, in plausible

384: settings, several vertical scale heights) from the ISCO.  This raises

385: the interesting possibility of having appreciable power in such modes

386: without significant leakage across the ISCO.

387: 

388: A major issue, however, is the effect of the turbulent MHD background

389: state on these modes.  The diskoseismic mode frequencies are

390: comparable to the frequencies characterizing the expected MHD

391: turbulent fluctuations (which is very different to the situation in

392: the Sun, for example, where the observed helioseismic modes have

393: frequencies that are four orders of magnitude higher than the

394: turbulent turnover frequency).  Thus, an MHD turbulent disk is likely

395: to be a hostile environment for any diskoseismic modes.  Furthermore,

396: magnetic forces can lead to a rather gradual transition in flow

397: properties around the ISCO, potentially worsening the leakage of the

398: trapped g-modes.  Considering these mode destruction/suppression

399: mechanisms, it is reasonable to suppose that mode survival becomes

400: easier in thinner disks since (1) the ratio of the typical turbulent

401: cell size to the radial extent of the resonant cavity will decrease

402: with disk thickness and (2) there are suggestions that the transition

403: in flow properties around the ISCO is sharper in thinner disks

404: (Reynolds \& Fabian 2008; Shafee et al. 2008) thereby producing less

405: mode leakage.  This raises the possibility that a $g$-mode can be

406: sustained against (or even fed by; Nowak \& Wagoner 1993) the

407: turbulence in a sufficiently thin disk.

408: 

409: Previously published global MHD disk simulations (e.g., Hawley \&

410: Krolik 2001) have modeled flows as thin as $h/r\sim 0.1$ and have not

411: reported diskoseismic modes.  However, these authors did not conduct a

412: directed search for such modes and hence it is not possible to say

413: that the modes were really not present.  Careful examination of local

414: ``shearing-box'' simulations have indeed failed to find trapped

415: $g$-modes associated with MHD turbulence (Arras, Blaes \& Turner

416: 2006), but this issue has yet to be examined in a global thin-disk

417: setting.  Searching for and characterizing these trapped $g$-modes in

418: global thin-disk simulations will be a major theme of our paper.

419: 

420: 

421: \subsection{Parametric resonance}

422: 

423: From the point of view of explaining HFQPOs, the need for global

424: oscillation modes is diminished if some process does indeed select

425: special radii in the accretion disk.  As discussed in the

426: Introduction, the discovery of pairs of HFQPOs with small integer

427: ratios has prompted several groups to examine resonance

428: models.  In particular, we shall briefly review the parametric

429: resonance model of Abramowicz \& Klu\'zniak (Abramowicz \& Klu\'zniak

430: 2001, 2003; Abramowicz et al. 2002, 2003).

431: 

432: We begin by considering an accretion disk in which the flow deviates

433: only slightly from Keplerian so that the position of a fluid element

434: (in spherical polar coordinates) is given by

435: \begin{equation}

436: r(t)=r_0+\delta r(t), \ \theta(t)=\frac{1}{2}\pi+\delta\theta(t),\ 

437: \phi(t)=\Omega t.

438: \end{equation}

439: We have specialized to the case of axisymmetric perturbations.  By

440: expanding the resulting equations of motion to third order (and

441: wrapping up the non-gravitational forces into two unspecified force

442: functions), one finds a Mathieu-type equation of motion

443: \begin{equation}

444: \delta\theta_{,\,tt}+\Omega^2[1+a\cos(\kappa t)]\delta\theta +

445: \lambda\,\delta\theta_{,\,t}=0,

446: \label{eq:mathieu}

447: \end{equation}

448: where $a$ is a small constant that describes the coupling between the

449: vertical and radial perturbations, and $\lambda$ is a (small) damping

450: constant.  Here, we have specialized the equations of Abramowicz et

451: al. (2003; hereafter A2003) to the case where the vertical epicyclic

452: frequency is the same as the orbital frequency.  This is valid for a

453: non-spinning black hole and, in particular, the pseudo-Newtonian

454: potential that we use for the simulations in this paper.

455: 

456: One expects a system described by eqn.~\ref{eq:mathieu} to exhibit a

457: parametric resonance instability when $\kappa=2\Omega/n$, where $n$ is

458: a non-zero positive integer.  Given that black hole potentials always

459: have $\kappa<\Omega$, the smallest value of $n$ for which the

460: resonance condition is obeyed is $n=3$, i.e., $\kappa=2\Omega/3$.

461: This is expected to be the strongest of the set of resonances.  It is

462: interesting that the fundamental ``test-particle'' frequencies at this

463: resonant radius have 3:2 frequency ratio in agreement with

464: observations of HFQPO pairs.  For the Paczynski-Wiita pseudo-Newtonian

465: potential we use in our simulations (Paczynski \& Wiita 1980, hereafter PW),

466: \begin{equation}

467: \Phi=-\frac{GM}{r-2r_g},\hspace{1cm}r_g\equiv \frac{GM}{c^2},

468: \label{eq:pw_pot}

469: \end{equation}

470: we have

471: \begin{equation}

472: \Omega=\frac{1}{r-2r_g}\sqrt{\frac{GM}{r}}

473: \end{equation}

474: and,

475: \begin{equation}

476: \kappa=\sqrt{\frac{GM(r-6r_g)}{r(r-2r_g)^3}}.

477: \end{equation}

478: Hence, the 3:2 resonance occurs at $r=9.2r_g$. It must be noted,

479: however, that a full integration of a toy-problem by A2003 reveals

480: that higher-order effects shift the location of the resonance, making

481: the ratio of the epicyclic frequencies extremely sensitive to the

482: strength of the coupling between the $r$ and $\theta$ perturbations.

483: While A2003 suggest that this sensitivity is a strength of the model,

484: allowing application to HFQPO pairs in neutron star systems which do

485: not have simple integer ratios, it inevitably diminishes the power of

486: this model to explain black hole systems.  In addition, there is

487: currently no physical model of the coupling between the radial

488: and vertical motions.  The simulations described in this paper allow

489: us to assess whether magnetic forces couple these motions in such a

490: way as to drive a parametric resonance instability.

491: 

492: \section{Hydrodynamic disks}

493: 

494: 

495: \subsection{Initial comments}

496: \label{sec:initial_hydro}

497: 

498: In the remainder of this paper, we construct and analyze numerical

499: simulations of thin disks in order to examine their variability

500: properties, focusing on the presence of local and global modes, as

501: well as parametric resonances.  Of course, real accretion disks are

502: believed to require at least an MHD-level description in order to

503: capture the MRI-driven turbulence that transports angular momentum and

504: drives accretion.  MHD simulations are addressed in the next section.

505: However, it is useful to begin with a discussion of ideal hydrodynamic

506: models in order to help isolate the various physical effects present

507: in these complex systems.  That will be the focus of this section.

508: 

509: In order to allow us to perform a set of simulations with

510: modest-to-high resolution, we begin with 2-d (axisymmetric)

511: hydrodynamic simulations.  We expect (and indeed show) that these

512: axisymmetric models are well suited for studying the fundamental $m=0$

513: g-mode.  However, once we move to MHD, Cowling's anti-dynamo theorem

514: (Cowling 1957) leads to qualitatively different behavior in

515: axisymmetric compared with full 3-d simulations.  Hence, all of the

516: MHD models that we shall describe are performed in 3-dimensions.

517: Since we will be directly comparing results (e.g., $g$-mode

518: amplitudes) between the hydrodynamic and MHD simulations, we also need

519: to perform a ``bridging'' 3-d hydrodynamic simulation.

520: 

521: \subsection{Basic simulation set-up}

522: \label{sec:hydro_setup}

523: 

524: To make the problem tractable despite the severe resolution

525: requirements imposed by the geometrical thinness of the accretion

526: disk, we choose to focus on only the most essential aspects of the

527: physics.  From the discussion in \S~2, it is clear that the essential

528: aspect of the relativistic potential that must be captured is the

529: nature of the radial epicyclic frequency (e.g., that it goes to zero

530: at some finite radius and hence produces an ISCO at that radius).  In

531: this sense, the PW potential (eqn.~\ref{eq:pw_pot}) is a good

532: approximation for the gravitational field of a non-rotating black

533: hole; its ISCO (at $6r_g$) and marginally bound orbit (at $4r_g$) are

534: both at the same radius as in the Schwarzschild geometry. We also

535: simplify the simulations by neglecting all radiation processes

536: (radiative heating, radiative cooling, radiative transfer, and the

537: dynamical effects of radiation pressure).  In place of a full energy

538: equation, the gas is given an adiabatic equation of state with

539: $\gamma=5/3$.

540: 

541: \begin{figure*}

542: \hbox{

543: \hspace{-1cm}

544: \psfig{figure=f1a.ps,width=0.55\textwidth}

545: \hspace{-1cm}

546: \psfig{figure=f1b.ps,width=0.55\textwidth}

547: }

548: \hbox{

549: \hspace{-1cm}

550: \psfig{figure=f1c.ps,width=0.55\textwidth}

551: \hspace{-1cm}

552: \psfig{figure=f1d.ps,width=0.55\textwidth}

553: }

554: \caption{Snapshots of evolution of the canonical axisymmetric

555: hydrodynamic simulation (HD2d\_1) at $t=0.5T_{\rm isco}$ (top-left),

556: $t=1.0T_{\rm isco}$ (top-right), $t=5T_{\rm isco}$ (bottom-left) and

557: $t=50T_{\rm isco}$ (bottom-right), where $T_{\rm isco}$ is the orbital

558: period at the ISCO.  Both the color table and contours show the

559: logarithmic density structure of the disk cross-section, with 10

560: contours per decade of density.  A range of densities spanning three

561: orders of magnitude are shown.  The curved line to the left of each

562: frame represents the event horizon.}

563: \label{fig:hydro_evol}

564: \end{figure*}

565: 

566: \begin{table*}[t]

567: \begin{center}

568: {\small

569: \begin{tabular}{cccccccc}\hline

570: Run     &   dim   & $h_2/r_{\rm isco}$ & r-domain ($r_g$)& z-domain ($r_g$)     & $\phi$-domain & $n_r\times n_z (\times n_\phi)$ & $T_{\rm stop}$ ($r_g/c$) \\

571: (1)     &    (2)  &   (3)              &   (4)           &    (5)               &     (6)       &      (7)                        &  (8)           \\\hline

572: HD2d\_1     &  2  & 0.05$^*$  & (4,16)   & (-1.5,1.5)    &    --         & $256\times 128$                 & 12320 \\

573: HD2d\_1hr   &  2  & 0.05$^*$  & (4,28)   & (-1.5,1.5)    &    --         & $1024\times 256$                & 12320 \\

574: HD2d\_2     &  2  & 0.025$^*$ & (4,16)   & (-0.75,0.75)  &    --         & $512\times 128$                 & 12320 \\

575: HD2d\_3     &  2  & 0.1$^*$   & (4,16)   & (-3,3)        &    --         & $256\times 128$                 & 12320 \\

576: HD3d\_1     &  3  & 0.05$^*$  & (4,16)   & (-1.5,1.5)    & $(0,\pi/6)$   & $240\times 128\times 32$        & 12320 \\\hline

577: MHD\_1 &       3  & 0.05  & (4,16)   & (-3,3) &    $(0,\phi/6)$  & $240\times 256\times 32$     & 38800  \\

578: MHD\_2 &       3  & 0.05  & (4,16)   & (-1.5,1.5) &    $(0,\phi/6)$  & $240\times 128\times 32$     & 38800  \\

579: MHD\_2hr &     3  & 0.05  & (4,16)   & (-1.5,1.5) &    $(0,\phi/3)$  & $480\times 256\times 64$     & 5236 \\

580: MHD\_3 &       3  & 0.05  & (4,28)   & (-1.5,1.5) &    $(0,\phi/6)$  & $960\times 128\times 32$     & 11400 \\\hline

581: \end{tabular}

582: \caption{Summary of the hydrodynamic (HD) and MHD simulations

583: discussed in this paper.  Column (1) gives the designation of the

584: simulation.  Column (2) lists the dimensionality of the simulation.

585: Column (3) gives the fractional disk thickness at the ISCO; the

586: asterisk ($*$) denotes that the simulation was started with an initial

587: vertical structure that is slightly out of equilibrium in order to

588: seed subsequent oscillation modes (as described in the text).  Columns

589: (4), (5) and (6) list the $r$-, $z$- and $\phi$- domains of the

590: simulation box.  Column (7) gives the number of computational zones

591: within the domain.  Column (8) lists the run time of the simulated

592: disk.}}

593: \end{center}

594: \end{table*}

595:  

596: 

597: All simulations are performed in cylindrical polar coordinates

598: $(r,z,\phi)$.  The initial condition consists of a disk with a

599: constant mid-plane density ($\rho_0=1$) for $r>r_{\rm isco}$.  The

600: initial density scale-height of the disk is assumed to be constant

601: with radius, implying a sound speed which falls off with radius as

602: approximately $r^{-3/2}$.  There are two motivations for choosing to

603: model a ``constant-h'' disk; (1) such a choice is well suited to the

604: cylindrical symmetry of our coordinate grid and (2) according to the

605: standard model of Shakura \& Sunyaev (1973), the radiation-pressure

606: dominated disks of real accreting black holes are likely to maintain

607: an approximately constant scale-height in their innermost regions.

608: In more detail, the vertical structure of the disk is given by

609: \begin{eqnarray}

610: \rho(r,z)&=&\rho_0\,\exp\left(-\frac{z^2}{2h_1^2}\right),\\

611: p(r,z)&=&\frac{GMh_2^2}{(R-2r_g)^2R}\,\rho(r,z),

612: \end{eqnarray}

613: where $r$ is the cylindrical radius, $z$ is the vertical height above

614: the disk midplane and $R=\sqrt{r^2+z^2}$. This corresponds to an

615: isothermal atmosphere which, when $h_1=h_2$, is in vertical

616: hydrostatic equilibrium in the PW potential.  In order to give the

617: disk an initial vertical kick, we set $h_1=1.2h_2$ (the values of

618: $h_2$ for all of our runs are detailed in Table~1).  Thus, the initial

619: disk temperature is $\sim 20\%$ too cold for the density and pressure

620: run, leading to a vertical collapse and bounce of the disk.  The

621: initial density is set to zero for $r<r_{\rm isco}$.  The initial

622: velocity field is everywhere set to

623: \begin{equation}

624: v_\phi=r\Omega=\frac{\sqrt{GMr}}{r-2r_g},\hspace{1cm}v_r=v_z=0,

625: \end{equation}

626: corresponding to rotation on cylinders and pure Keplerian motion for

627: material on the mid-plane.  We impose zero-gradient outflow boundary

628: conditions on both the radial and vertical boundaries of the

629: simulation, i.e., the fluid quantities in the ghost zones are set to

630: the values of the neighboring active zone, and a ``diode'' condition

631: is imposed on the component of the velocity perpendicular to the

632: boundary which allows outflow but disallows inflow.   

633: 

634: 

635: 

636: Table~1 details our hydrodynamic simulations.  We perform four

637: simulations in which strict axisymmetry is imposed at all times

638: ($\partial/\partial \phi=0$).  These axisymmetric simulations were

639: performed using the serial ZEUS-3D MHD code (Stone \& Norman 1992a,b)

640: in its pure hydrodynamic axisymmetric mode.  In our canonical

641: axisymmetric run (HD2d\_1), we set $h_2=0.3r_g$ (corresponding to

642: $h_2/r=0.05$ at the ISCO).  To model the dynamics of the disk in a

643: robust manner, our vertical domain must cover many scale-heights; in

644: our canonical run, the vertical domain is $z\in (-5h_2,+5h_2)$.  We

645: place 128 uniformly spaced grid cells in this vertical direction,

646: giving 13 cells per scale-height.  This allows us to resolve

647: hydrodynamic waves with wavelengths of $\sim 0.5h_2$ or greater.  The

648: vertical resolution requirements force us to consider a limited radial

649: domain.  In the canonical simulation, the radial domain extends from

650: $4r_g$ (i.e., well within the plunge region) to $16r_g$ and should

651: correspond to the range of radii where trapped diskoseismic modes or

652: A2003-type resonances occur.  Tolerating a grid-cell aspect ratio of

653: $\sim 2$, we place 256 uniformly spaced radial cells in this radial

654: domain.  The simulation was evolved for a time $200T_{\rm isco}$ where

655: \begin{equation}

656: T_{\rm isco}\approx 61.6\,GM/c^3,

657: \end{equation} 

658: is the orbital period at the ISCO.

659: 

660: In order to test the robustness of the results discussed below to

661: resolution and the limited radial domain, we performed a second

662: simulation (HD2d\_1hr) in which the spatial resolution was doubled

663: (i.e., the size of each voxel was halved in both the radial and

664: vertical dimensions) and the radial domain extended out to $28r_g$.

665: We also performed two additional axisymmetric simulations employing

666: the same set-up as the canonical simulation but with disks that are

667: half (HD2d\_2) and twice (HD2d\_3) the thickness (including

668: appropriate modifications to the vertical domain and resolution; see

669: Table~1).

670: 

671: As discussed above, we perform a 3-dimensional hydrodynamic simulation

672: in order to aid the later interpretation of the 3-d MHD simulations.

673: The set-up of our 3-d run (HD3d\_1) is identical to that for the

674: canonical 2-d run except that the computational domain has a

675: $\phi$-dimension.  To reduce computational expense while capturing the

676: essential physics, we simulate only a $\Delta\phi=30^\circ$ wedge of

677: the disk using 32 uniformly spaced grid cells, imposing periodic

678: boundary conditions on the $\phi$-boundaries.  This 3-dimensional

679: simulation was performed using an MPI-parallelized version of ZEUS

680: kindly supplied to us by Eve Ostriker (and similar to the ZEUS-MP code

681: of Vernaleo \& Reynolds 2006).

682: 

683: \subsection{Axisymmetric hydrodynamic models and the recovery of trapped 

684: $g$-modes}

685: \label{sec:axisym}

686: 

687: 

688: We now discuss the evolution of the axisymmetric hydrodynamic

689: simulations, beginning with the canonical simulation.  Starting from

690: the initial condition, the disk undergoes dynamical timescale

691: variability because of pressure gradients which push matter inside of

692: the ISCO.  The strong transients close to the ISCO launch

693: outward-radially directed waves into the disk which break rapidly to

694: become rolls.  This behavior can be seen in Fig.~\ref{fig:hydro_evol}.

695: These high amplitude transients are short lived, however, lasting only

696: $\sim 10T_{\rm isco}$.  At subsequent times, the disk settles into a

697: stationary state apart from small amplitude (and decaying) internal

698: oscillations and a very weak accretion stream driven by the numerical

699: viscosity.

700: 

701: \begin{figure}[t]

702: \centerline{

703: \psfig{figure=f2.ps,width=0.55\textwidth}

704: }

705: \caption{Change of the quantity $K=\int_{\cal D} \rho v_z^2\,dV$ with

706: time.  The integration domain ${\cal D}$ is the annulus $r\in (7,14)$.  The

707: bottom (black) line is for the canonical-2d run (HD2d\_1),

708: lower-middle (red) line is for the high-resolution 2-d run

709: (HD2d\_1hr), the upper-middle (blue) line is for the canonical 3-d run

710: (HD3d\_1), and the top (green) line is for the canonical MHD run

711: (MHD\_1).  The distinct ``ringing'' seen in runs HD2d\_1 and HD2d\_1hr

712: corresponds to the axisymmetric g-mode (see Section~\ref{sec:axisym}).

713: In run HD3d\_1, non-axisymmetric aperiodic structures mask the

714: underlying g-mode (see Section~\ref{sec:3dhydro}).  The fact that

715: $K(t)$ for the run MHD\_1 lies an order of magnitude above that for

716: HD3d\_1 demonstrates that the MRI-driven turbulence completely

717: overwhelms the hydrodynamic disturbances seen in the analogous 3-d

718: hydrodynamic simulation (see Section~\ref{sec:basicmhdevol}).}

719: \label{fig:hydro_decay}

720: \end{figure}

721: 

722: 

723: In order to study the decay of the hydrodynamic fluctuations in a more

724: quantitative manner, we start by computing the quantity,

725: \begin{equation}

726: K=\int_{\cal D} \rho v_z^2\,dV.

727: \label{eq:decay}

728: \end{equation}

729: This quantity is particularly well suited to diagnose vertical

730: oscillations of the disk, and will vanish once the hydrodynamic

731: configuration has established a stationary state.  In order to

732: diagnose the state of the body of the disk (i.e., side-stepping issues

733: of the outer radial boundary or the plunge region), we choose to

734: compute this integral over a restricted domain ${\cal D}$ consisting

735: of the annulus $r\in (7r_g,14r_g)$.  The time-dependence of

736: $K/\max(K)$ for the canonical axisymmetric simulation (HD2d\_1) and its

737: high-resolution counterpart (HD2d\_1hr) is shown in

738: Fig.~\ref{fig:hydro_decay}.  The rapid initial decline of $K(t)$ is

739: very similar for these two simulations and corresponds to the strong

740: transients described above.  At long times (after about $t\sim 4\times

741: 10^3\,GM/c^3\approx 70\,T_{\rm isco}$) the behavior of these

742: simulations begins to deviate.  HD2d\_1 continues to decay in an

743: approximately exponential manner $K(t)\propto e^{-t/\tau_0}$, where

744: $\tau_0\approx 4\times 10^3\,GM/c^3$.  Superposed on this decay is a

745: distinct oscillation.  This corresponds to (twice) the frequency of

746: the trapped $g$-mode that we shall discuss below.  The high-resolution

747: version of this simulation HD2d\_1hr shows very similar behavior

748: except that the exponential decay time is longer, $\tau_0\approx

749: 6\times 10^3\,GM/c^3$.  This suggests that the decay of these small

750: perturbations is due to numerical dissipation which scales

751: approximately as the square root of the size of the simulation grid

752: cells.

753: 

754: 

755: We now study the spatio-temporal variability of the disk and, in

756: particular, seek the diskoseismic modes predicted by linear theory.

757: Figure~\ref{fig:vr_rt} shows the mid-plane value of $v_r$ on the

758: $(r,t)$-plane, i.e., the value of the function $v_r(r,z=0;t)$, for run

759: HD2d\_1.  Note that, outside of the ISCO, the average value is

760: $\langle v_r\rangle\ll 0.001c$ so that this Figure essentially plots the

761: fluctuation of $v_r$ from its mean value.  At early times, we see

762: strong wave-like disturbances which are generated in the inner parts

763: of the disk (at $r\approx 8r_g$) and propagate to both large and small

764: radii.  Although the outer radial boundary condition is zero-gradient

765: outflow, impedance mismatching results in some reflection of these

766: initial strong waves.  After these initial transients have died out,

767: it can be seen that the highest amplitude fluctuations are limited to

768: a rather narrow range of radii in the approximate range $r=7-9r_g$.

769: The fact that these perturbations are essentially vertical on the

770: $(t,r)$-plane indicates that they are coherent across this radial

771: range, as would be expected if we are seeing a global mode.

772: 

773: \begin{figure}[b]

774: \centerline{

775: \psfig{figure=f3.ps,width=0.55\textwidth}

776: }

777: \caption{Radial component of velocity $v_r$ on the midplane of the

778: disk as a function of radius and time for the the canonical

779: axisymmetric hydrodynamic simulation (HD2d\_1).  The linear color

780: table extends from radial velocities of $v_r=-0.001c$ (black) to

781: $v_r=+0.001c$ (white).  Note that, outside of the ISCO, the average

782: value is $|v_r|\ll 0.001c$ so that this diagram essentially plots the

783: fluctuation of $v_r$ from its mean value.}

784: \label{fig:vr_rt}

785: \end{figure}

786: 

787: % FIG 4 HERE

788: \begin{figure*}

789: \hbox{

790: \psfig{figure=f4a.ps,width=0.55\textwidth}

791: \psfig{figure=f4b.ps,width=0.55\textwidth}

792: }

793: \caption{Midplane PSD for pressure fluctuations in run HD2d\_1 (left

794: panel) and its high-resolution counterpart HD2d\_1hr (right panel).

795: Note that we show the radial range $r\in (4,16)$ which is coincident

796: with the full computational domain of HD2d\_1 but only the inner half

797: of the domain for HD2d\_1hr (whose full radial domain extends from

798: $4r_g$ to $28r_g$). Also shown are the radial epicyclic frequency

799: (solid white line), orbital frequency (dashed) and the $n=1,2,3$ pure

800: vertical p-modes (from left to right dot-dashed lines).  The absolute

801: scaling of the PSD, as indicated by the color-bar, is arbitrary.}

802: \label{fig:hd_psd}

803: \vspace{0.5cm}

804: \end{figure*}

805: 

806: This temporal variability can be explored in more detail using the

807: power spectral density (PSD), defined as $P(\nu)=\alpha |\tilde

808: f(\nu)|^2$, where $\alpha$ is some normalization constant and $\tilde

809: f(\nu)$ is the Fourier transform of the time-sequence $f(t)$ under

810: consideration,

811: \begin{equation}

812: \tilde f(\nu)=\int f(t)e^{-2\pi i\nu t}\,dt.

813: \end{equation}

814: Note that, for ease of interpretation, most of the PSDs presented in

815: this paper will be in terms of the usual frequency, $\nu$, rather than

816: angular frequency, $\omega$.  Figure~\ref{fig:hd_psd} shows the PSD of

817: the mid-plane pressure as a function of $r$ and frequency for HD2d\_1

818: and HD2d\_1hr.  This is computed using approximately the final half

819: ($\Delta t=102.4T_{\rm isco}\approx 6308\,GM/c^3$) of the simulation

820: in order to avoid the initial strong transients.  These PSDs differ in

821: the range $r=12-16r_g$; run HD2d\_1hr has significantly less low

822: frequency noise than run HD2d\_1 in this radial range.  We attribute

823: this to effects related to the outer radial boundary which is at

824: $r=16r_g$ in run HD2d\_1 but substantially further out ($r=28r_g$) in

825: run HD2d\_1hr.

826: 

827: 

828: \begin{figure*}

829: \hbox{

830: \hspace{-1cm}

831: \psfig{figure=f5a.ps,width=0.55\textwidth}

832: \hspace{-1cm}

833: \psfig{figure=f5b.ps,width=0.55\textwidth}

834: }

835: \caption{{\it Left panel} : Midplane pressure PSD for run HD2d\_1

836: summed up over a range of radii $\Delta r=0.5r_g$ centered on

837: $r=8r_g$.  The dashed line shows the maximum radial epicyclic frequency.

838: {\it Right panel }: Integral of the midplane pressure PSD of those

839: frequency bins that exceed $5\times 10^{-14}$ in the left panel, as a

840: function of radius. Vertical dashed lines show the range of radii for

841: which the mean frequency of this peak is less than the radial

842: epicyclic frequency.}

843: \vspace{0.5cm}

844: \end{figure*}

845: 

846: 

847: \begin{figure*}

848: \hbox{

849: \hspace{-1cm}

850: \psfig{figure=f6a.ps,width=0.55\textwidth}

851: \psfig{figure=f6b.ps,width=0.55\textwidth}

852: }

853: \hbox{

854: \hspace{-1cm}

855: \psfig{figure=f6c.ps,width=0.55\textwidth}

856: \psfig{figure=f6d.ps,width=0.55\textwidth}

857: }

858: \caption{Maps of period-folded pressure deviation (i.e., the

859: difference between the instantaneous pressure and the time-averaged

860: pressure) for run HD2d\_1hr.  Only the last 102.4 orbits of data have

861: been included in order to avoid the transient behavior associated with

862: the initial conditions.  The folding period corresponds to the peak of

863: the PSD see in Fig.~\ref{fig:gmode}.  Phases of 0.0 (top-left), 0.25

864: (top-right), 0.5 (bottom-left) and 0.75 (bottom-right) are shown.

865: This, in essence, gives us a direct view of the eigenmode.}

866: \label{fig:folding}

867: \end{figure*}

868: 

869: 

870: Inside of $r=12r_g$, however, the PSDs of these two simulations are

871: very similar and we can trust that neither the resolution nor the

872: outer radial boundary affect the results significantly.  We note that

873: simulations in which the initial radial density profile of the disk is

874: truncated before reaching the outer boundary also produce essentially

875: identical results, again giving us confidence that noise infiltration

876: from the outer boundary is not an important issue.  In addition to

877: very low frequency noise, the most prominent feature of the inner-disk

878: PSD is a vertical ridge of enhanced power at $\nu\approx (4-5)\times

879: 10^{-3}\,c^3/GM$ extending from $r\approx 6.5r_g$ out to $r\approx

880: 9.5r_g$.  The one dimensional cuts through this ridge in

881: Fig.~\ref{fig:gmode} demonstrate it to have all of the expected

882: properties of a trapped $g$-mode.  Firstly, it exists in a rather

883: narrow range of frequencies, $\nu\approx (4-5)\times 10^{-3}\,c^3/GM$,

884: just below the maximum radial epicyclic frequency ($\kappa_{\rm

885: max}\approx 5.52\times 10^{-3}\,c^3/GM$).  Secondly, it is spatially

886: bounded by the radii at which the mode frequency becomes equal to the

887: radial epicyclic frequency.  There is, however, some leaking of the

888: mode down to the ISCO.

889: 

890: We can visualize the eigenmode by producing maps of pressure

891: deviations that have been ``period-folded'' on the period

892: corresponding to the peak power in this mode.  More precisely, we use

893: the last half of the simulation to produce maps of the difference

894: between the instantaneous pressure and the time-averaged pressure with

895: a sampling rate of $0.2T_{\rm isco}$.  We then sort these maps into 16

896: phase bins (based on the period corresponding to the peak power in

897: this mode) and average together all maps within a given phase bin.

898: The result is shown in Fig.~\ref{fig:folding}.  In addition to the

899: g-mode itself, these maps reveal a striking ``chevron'' pattern at

900: large radii.  Time-sequences of maps reveal these features to be

901: outward-radially traveling acoustic waves driven by the global

902: g-mode, refracted into the upper layers of the disk atmosphere as they

903: propagate.

904: 

905: % FIG 7 HERE

906: \begin{figure}[t]

907: \centerline{

908: \psfig{figure=f7.ps,width=0.55\textwidth}

909: }

910: \caption{Frequency of the g-modes as a function of mid-plane sound

911: speed at $r=8r_g$ for runs HD2d\_2, HD2d\_1, and HD2d\_3 (from left to

912: right).  The vertical bars indicate the range of frequencies over

913: which enhanced power is seen.}

914: \label{fig:f_vs_cs}

915: \vspace{0.5cm}

916: \end{figure}

917: 

918: 

919: As a final demonstration that we have properly identified trapped

920: $g$-modes in our axisymmetric hydrodynamic simulations, we use our set

921: of runs to study the dependence of the mode frequency on the sound

922: speed in the disk.  Figure~\ref{fig:f_vs_cs} shows the dependence of

923: the mode frequency on midplane sound speed of the disk (and hence disk

924: thickness) at $r=8r_g$.  As expected from analytic theory (see

925: Appendix~A), the difference between the mode frequency and the maximum

926: epicyclic frequency depends linearly on the sound speed.  There is,

927: however, a discrepancy in the slope of this linear relationship obtained

928: by the simulations and expected from the analytic theory.  Note that

929: the analytic treatment assumes that the gas is strictly isothermal

930: across the whole region of interest, a condition that is clearly

931: violated in the simulated disk.  Given the sensitivity of the mode to

932: the radial structure of the disk (as demonstrated by the factor of 2.5

933: difference in the analytic results between the two pseudo-Newtonian

934: potentials examined in Appendix~A), the non-isothermality of the gas

935: in the simulated disk can readily shift the mode frequency away from

936: the analytic value.

937: 

938: \subsection{Extension to 3-dimensional hydrodynamic models}

939: \label{sec:3dhydro}

940: 

941: We begin our analysis of the 3-dimensional hydrodynamic simulations

942: (HD3d\_1) by examining the decay of hydrodynamic perturbations in our

943: canonical 3-d run (HD3d\_1) using the quantity $K(t)$ as defined in

944: eqn.~\ref{eq:decay}.  The $K(t)$ behavior for HD3d\_1 is somewhat

945: different from the axisymmetric simulations at late times, reaching a

946: quasi-steady state at $K/\max(K)\sim 3\times 10^{-3}$ with aperiodic

947: fluctuations rather than continuing a ringing exponential decay

948: (Fig.~\ref{fig:hydro_decay}).  Numerical dissipation must be just as

949: effective in HD3d\_1 as compared with HD2d\_1, hence these

950: perturbations must be driven by some instability.  While it is thought

951: that free Keplerian accretion disks are stable to linear hydrodynamic

952: perturbations (see Balbus \& Hawley 1998; Hantao et al. 2006), the

953: presence of the simulation boundaries can introduce true instabilities

954: (associated with reflection from the boundaries) as well as

955: uncontrolled numerical noise, and these might explain these low level

956: sustained fluctuations.  Visual inspection of density slices in the

957: $(r,z)$ and $(r,\phi)$ planes also suggests that non-axisymmetric

958: wave-like perturbations interacting with the boundaries are

959: responsible for perturbing the 3-d hydrodynamic disk.  However, a

960: detailed study of the sustained fluctuations in the 3-d hydrodynamic

961: disks is beyond the scope of this paper.  Since these perturbations

962: exist at a low level (particularly compared with the MHD turbulence

963: discussed in \S~4; see Fig.~\ref{fig:hydro_decay} for a comparison of

964: the sustained fluctuations in the 3-d hydrodynamic run compared with

965: the canonical MHD run) and are likely to be driven by our boundaries

966: (hence, are not of astrophysical relevance), they are not of

967: importance for the principal focus of our study.

968: 

969: % FIG 8 HERE

970: \begin{figure}[b]

971: \centerline{

972: \psfig{figure=f8.ps,width=0.55\textwidth}

973: }

974: \caption{

975: Midplane PSD of azimuthally-averaged pressure for run HD3d\_1 summed

976: up over a range of radii $\Delta r=0.5r_g$ centered on $r=8r_g$.  The

977: dashed line shows the maximum epicyclic frequency.  

978: }

979: \label{fig:hd3d_psd}

980: \end{figure}

981: 

982: 

983: The principal issue to be addressed here is the impact of the

984: transition to 3-dimensions on the presence of the trapped $m=0$

985: g-modes in the simulated disks.  We might expect the power in these

986: axisymmetric modes to be reduced in the 3-d case as the free energy is

987: shared with non-axisymmetric modes.  This is indeed the

988: case. Figure~\ref{fig:hd3d_psd} shows the PSD at $r=8r_g$ of the last

989: $\Delta t=102.4T_{\rm isco}$ of the canonical 3-d run HD3d\_1.  A

990: region of enhanced power is clearly seen in the correct range of

991: frequencies $\nu\approx (4-5)\times 10^{-3}\,c^3/GM$ to be identified

992: with the axisymmetric g-modes studied in \S~\ref{sec:axisym}.

993: Furthermore, examination of the radially-resolved PSD shows that this

994: region of enhanced power is bounded by the radial epicyclic frequency

995: in precisely the manner expected for trapped g-modes.  A comparison of

996: the absolute values of the PSD across the mode does reveal, however,

997: that the mode contains almost an order of magnitude less power than in

998: the axisymmetric case.   

999: 

1000: We note the existence of a narrow but large amplitude spike just above

1001: a frequency of $7\times 10^{-3}\,c^3/GM$ in the $r=8r_g$ PSD of this

1002: run.  The identification of this feature is not clear; it is not at

1003: the frequency of any expected global g- or p-mode.  It is, however,

1004: confined to a single frequency bin, contains only a small amount of

1005: power and only shows up over a narrow range of radii.  It seems likely

1006: that this is a noise spike.

1007: 

1008: \section{Magnetohydrodynamic disks}

1009: \label{sec:mhd}

1010: 

1011: Having gained an understanding of the thin hydrodynamic disks, we move

1012: onto the more astrophysically relevant case of MHD disks.  In this

1013: section, we construct MHD simulations of thin accretion disks in a PW

1014: potential.  We then examine the properties of the broad-band noise and

1015: search for modes in the resulting turbulent MHD disks.

1016: 

1017: \subsection{Simulation set-up}

1018: 

1019: We simulate geometrically-thin MHD accretion disks by building upon

1020: our hydrodynamic computational set-up described in

1021: \S~\ref{sec:hydro_setup}.  As discussed in \S~\ref{sec:initial_hydro},

1022: we only consider 3-dimensional MHD simulations.

1023: 

1024: Table~1 details our set of MHD simulations.  Most of our discussion

1025: will center around run MHD\_1 (which we shall refer to as our

1026: canonical MHD run).  In this run, the hydrodynamic variables are

1027: set-up as for the canonical hydrodynamic run (see

1028: \S~\ref{sec:hydro_setup}) except that we begin the disk in vertical

1029: hydrostatic equilibrium ($h_1=h_2=0.3r_g$ corresponding to

1030: $h_1/r=h_2/r=0.05$ at the ISCO).  An initially weak magnetic field is

1031: introduced in the form of poloidal field loops specified in terms of

1032: their vector potential ${\bf A}=(A_r,A_z,A_\phi)$ in order to ensure

1033: that the initial field is divergence free.  We choose the explicit

1034: form for the vector potential,

1035: \begin{eqnarray}

1036: A_\phi=A_0\,f(r,z)\,p^{1/2}\,\sin \left(\frac{2\pi r}{5h_1}\right),\hspace{0.5cm}A_r

1037: =A_z=0,

1038: \end{eqnarray}

1039: where $A_0$ is a normalization constant and $f(r,z)$ is an envelope

1040: function that is unity in the body of the disk and then smoothly goes

1041: to zero at $r=r_{\rm isco}$, $r=r_{\rm out}$ and at a location three

1042: pressure scale heights away from the midplane of the disk.  The use of

1043: $f(r,z)$ keeps the initial field configuration well away from either

1044: the radial boundaries of the initial disk configuration or the

1045: vertical boundaries of the calculation domain.  The final

1046: multiplicative term produces field reversals with a radial wavelength

1047: of $5h$.  This results in a number of distinct poloidal loops

1048: throughout the disk.  The normalization constant $A_0$ is set to give

1049: an initial ratio of gas-to-magnetic pressure of $\beta=10^3$ in the

1050: inner disk.  In our canonical MHD simulation, this initial condition

1051: is evolved for a duration of $630T_{\rm isco}$ ($38800\,GM/c^3$) using

1052: the MPI-parallelized version of ZEUS described above.

1053: 

1054: We supplement the ideal MHD algorithms of ZEUS in two ways.  Firstly,

1055: it is necessary to impose a floor to the density field of $10^{-5}$

1056: times the initial maximum density in order to prevent the numerical

1057: integration from producing negative densities.  This essentially

1058: amounts to a subtle distributed mass source.  The density only reaches

1059: this floor close to the $z$-boundary (i.e., many scale heights above

1060: and below the disk plane).  Secondly, we implement the prescription of

1061: Miller \& Stone (2000) to include some effects of the displacement

1062: current, principally forcing the Alfv\'en speed to correctly limit to

1063: the speed of light as the magnetic fields becomes strong.  We note

1064: that this ``Alfv\'en speed limiter'' only plays a role within small

1065: patches of the tenuous magnetized atmosphere that forms at large

1066: vertical distances above and below the disk; it never plays a

1067: significant role in the body of the accretion disk.

1068: 

1069: Periodic boundary conditions were imposed on the $\phi$-boundaries,

1070: and zero-gradient outflow boundary conditions were imposed at both

1071: the inner and outer radial boundaries (Stone \& Norman 1992a,b).

1072: However, the choice of the $z$-boundary condition for this kind of

1073: simulation is notoriously problematic.  The most physically motivated

1074: choice would be a free outflow boundary.  However, as described in

1075: Stone et al. (1996), field-line ``snapping'' at these free boundaries

1076: can halt such a simulation.  Indeed, our own test simulations

1077: employing zero-gradient outflow boundary conditions on the

1078: $z$-boundaries were found to be subject to these difficulties, as well

1079: as occasional numerical instabilities appearing to result from an

1080: interplay of the imposition of the density floor and the free

1081: boundary.  Furthermore, these tests showed that the tenuous matter

1082: high above the disk mid-plane generally flows slowly across these

1083: boundaries at very sub-sonic and sub-Alfv\'enic speeds; strictly, this

1084: invalidates the use of such boundary conditions anyways (since the

1085: flow on the other side of the boundary should be able to act back on

1086: the simulation domain).

1087: 

1088: We adopt the solution of Stone et al. (1996) and choose to employ

1089: periodic boundary conditions in the $z$-directions.  While this is

1090: obviously unphysical in the sense that matter cannot leave the

1091: simulation domain in the vertical direction, it does guarantee

1092: mathematically reasonable behavior at the boundary (eliminating

1093: numerical instabilities) and, more importantly, appears to have no

1094: effect on the dynamics of the accretion disk itself (as diagnosed

1095: through comparisons with our vertical-outflow test runs).  In order to

1096: further isolate the simulated disk from the vertical boundaries, we

1097: expand the vertical domain (compared to the canonical hydrodynamic

1098: simulation) to $z\in(-3,3)$ (i.e., $\pm 10h_1)$.

1099: 

1100: We perform four additional simulations aimed at demonstrating the

1101: robustness of the canonical simulation.  In run MHD\_2, we restrict

1102: the vertical domain back to $z\in(-1.5,1.5)$, allowing us to gauge the

1103: importance of the location of the $z$-boundaries.  Run MHD\_2hr has an

1104: identical set-up to MHD\_2 except that the radial and vertical

1105: resolution is doubled compared with the canonical run (i.e., the voxel

1106: size is reduced by a factor of two in each of the radial and vertical

1107: directions), and the azimuthal domain is doubled to

1108: $\Delta\phi=60^\circ$ at fixed resolution.  Due to the factor of 8

1109: increase in number of computational cells (and the decrease in the

1110: timestep), this run was only integrated for a duration of $85T_{\rm

1111: isco}$ ($5236\,GM/c^3$).  While the run time is insufficient to

1112: conduct a detailed temporal study, a comparison with run MHD\_2 does

1113: allow us to investigate the effect of both the radial/vertical

1114: resolution and the extent of the $\phi$-domain on the turbulent state

1115: (see below).  As discussed below, we find that the properties of the

1116: simulated disks are very similar these two runs.

1117: 

1118: Run MHD\_3 is similar to MHD\_2 except that the outer radius is pushed

1119: to $r=28r_g$ (at fixed resolution), doubling the size of the radial

1120: domain.  Comparing MHD\_2 and MHD\_3, we find no evidence that the

1121: outer radial boundary is affecting the inner disk ($r<12r_g$) in any

1122: way.

1123: 

1124: 

1125: \subsection{Basic evolution of the MHD disks}

1126: \label{sec:basicmhdevol}

1127: 

1128: % FIG 9 HERE

1129: \begin{figure*}

1130: \hbox{

1131: \hspace{-1cm}

1132: \psfig{figure=f9a.ps,width=0.55\textwidth}

1133: \hspace{-1cm}

1134: \psfig{figure=f9b.ps,width=0.55\textwidth}

1135: }

1136: \hbox{

1137: \hspace{-1cm}

1138: \psfig{figure=f9c.ps,width=0.55\textwidth}

1139: \hspace{-1cm}

1140: \psfig{figure=f9d.ps,width=0.55\textwidth}

1141: }

1142: \caption{Snapshots of evolution of the canonical MHD simulation

1143: (MHD\_1) at $t=0$ (top-left), $t=1T_{\rm isco}$ (top-right),

1144: $t=10T_{\rm isco}$ (bottom-left) and $t=100T_{\rm isco}$

1145: (bottom-right), where $T_{\rm isco}$ is the orbital period at the

1146: ISCO.  Both the color table and contours show the logarithmic density

1147: structure of the disk cross-section, with 10 contours per decade of

1148: density.  A range of densities spanning three orders of magnitude are

1149: shown.  The curved line to the left of each frame represents the event

1150: horizon.}

1151: \label{fig:mhd_evol}

1152: \vspace{0.5cm}

1153: \end{figure*}

1154: 

1155: 

1156: We now discuss the evolution and general properties of these MHD

1157: disks, centering our discussion around run MHD\_1.  At very early

1158: times ($t<5T_{\rm isco}$) strong hydrodynamic transients dominate the

1159: evolution as radial pressure forces drive mass into the region within

1160: the ISCO.  Similarly to the hydrodynamic cases, these transients

1161: launch outwardly directed axisymmetric waves that break to form rolls.

1162: These strong hydrodynamic transients largely damp away in all but the

1163: outermost parts of the disk within $10T_{\rm isco}$.  Concurrently,

1164: the MRI amplifies the initial magnetic field until the (domain wide)

1165: volume-averaged ratio of gas-to-magnetic pressure peaks at

1166: $\langle\beta\rangle \sim 5$ (at $t\approx 10T_{\rm isco}$), at which

1167: point most of the flow has become turbulent.  The entire flow (outside

1168: of the ISCO) becomes turbulent by $t\approx 20T_{\rm isco}$.

1169: Figure~\ref{fig:mhd_evol} displays the density field across a vertical

1170: slice through the accretion disk at various times.

1171: 

1172: Between $t=10T_{\rm isco}$ and $t=20T_{\rm isco}$, the total magnetic

1173: energy in the computational domain declines until $\langle\beta\rangle

1174: \sim 20$.  After this, a quasi-steady state seems to be achieved where

1175: the magnetic field generation by the MRI-driven MHD turbulence is

1176: balanced by the removal of magnetic field energy due to numerical

1177: reconnection and magnetic buoyancy.  The fact that buoyancy is playing

1178: an important role is revealed by examining time-sequences of the

1179: strengths of B-field components in $(r,z)$ cross-sections of the disk.

1180: One clearly sees highly magnetized structures being generated close to

1181: the midplane of the disk which then propagate vertically away from the

1182: midplane\footnote{In principle, one could address the importance of

1183: magnetic buoyancy in removing magnetic energy from the mid-plane

1184: regions of the disk by comparing the vertical Poynting flux with

1185: (numerical) reconnection losses.  However, energy losses due to

1186: numerical reconnection are cannot be tracked in our simulation (due to

1187: the non-conservative nature of ZEUS algorithm) and, hence, a rigorous

1188: study of this issue is not possible.}.  From this time until the end

1189: of the simulation at $t=630T_{\rm isco}$, MHD turbulence and hence

1190: accretion is sustained.  Over the course of the simulation, the disk

1191: loses a little more than 60\% of its mass (Fig.~\ref{fig:mass_loss}).

1192: 

1193: % FIG 10 HERE

1194: \begin{figure}

1195: \centerline{

1196: \hspace{-1cm}

1197: \psfig{figure=f10.ps,width=0.55\textwidth}

1198: }

1199: \caption{Normalized total mass as a function of time for run MHD\_1.}

1200: \label{fig:mass_loss}

1201: \vspace{0.5cm}

1202: \end{figure}

1203: 

1204: \begin{figure*}

1205: \hbox{

1206: \hspace{-1cm}

1207: \psfig{figure=f11a.ps,width=0.55\textwidth}

1208: \hspace{-1cm}

1209: \psfig{figure=f11b.ps,width=0.55\textwidth}

1210: }

1211: \caption{{\it Left panel : }Total (domain integrated) magnetic and

1212: thermal energies for run MHD\_2 (black solid line) and its

1213: high-resolution counterpart (MHD\_2hr; red dashed line).  {\it Right

1214: panel : }Azimuthally-averaged plasma-$\beta$ parameter (i.e., the

1215: ratio of the thermal to magnetic pressure) as a function of vertical

1216: height in the disk at $r=8\,r_g$. To obtain this plot, data from

1217: $t=30-35T_{\rm isco}$ have been averaged together.  The solid (black)

1218: line shows run MHD\_2 whereas the dashed (red) line shows its

1219: high-resolution counterpart (MHD\_2hr).}

1220: \label{fig:mhd_compare}

1221: \end{figure*}

1222: 

1223: The total magnetic energy and thermal energy undergo a slow decline as

1224: mass is drained out of the simulated disk.  During this decline, the

1225: volume-averaged plasma-$\beta$ parameter within the domain remains in

1226: the range $\langle\beta\rangle\sim 20-30$.  However, as expected, the

1227: $\beta$ parameter within the high-density body of the disk is

1228: appreciably higher, reaching values of $\beta\approx 70-100$.  While

1229: this is significantly larger than the $\beta$ found in global

1230: simulations of thicker disks (e.g., Hawley \& Krolik 2001, 2002), it

1231: is in line with what might be expected for thin accretion disks with

1232: zero net field as diagnosed through local shearing-box simulations

1233: both with and without vertical stratification (Stone et al. 1996;

1234: Hawley, Gammie, \& Balbus 1996; Miller \& Stone 2000).

1235: 

1236: A generic concern in this class of simulation is the affect of the

1237: $z$-boundaries.  Once it achieves its quasi-steady state, our

1238: canonical MHD simulation displays a 2--3 order of magnitude drop in

1239: magnetic pressure, and a 5 order of magnitude drop in gas pressure,

1240: between the disk and the $z$-boundary.  Thus, the disk boundary seems

1241: to be well isolated from the boundary.  Further confidence is gained

1242: from an examination of run MHD\_2 in which the $z$-boundaries have

1243: been brought in from $z=\pm3$ to $z=\pm1.5$.  Despite the fact that

1244: the magnetic pressure now only drops by 1 order of magnitude from the

1245: disk to the boundary, all of the results from the canonical MHD run

1246: discussed in this paper are reproduced by MHD\_2.  More precisely, (1)

1247: the PSD of the fluid variables (\S\ref{sec:mhdfluidpsd}) are

1248: essentially indistinguishable, failing to show any evidence for global

1249: modes but displaying prominent local $p$-modes, (2) the PSD of the

1250: mass accretion rate has a broken power-law form with indices that

1251: differ from those found in MHD\_1 by $\Delta\approx 0.1$

1252: (comparable to the 1$\sigma$ error bars) and break frequencies that

1253: differ by $\Delta (\log \nu_{\rm break})\approx 0.05$ (again,

1254: comparable to the 1$\sigma$ error bars).

1255: 

1256: Another generic concern with thin disk simulations is whether the

1257: resolution in the vertical direction is adequate.  The canonical MHD

1258: simulation (MHD\_1) has approximately 13 grid cells spanning a

1259: vertical range $\Delta z=h_2$, implying that we cannot follow any

1260: modes with a wavelength smaller than $\lambda\sim h_2/2$.  This is

1261: just sufficient to follow the fastest growing MRI mode (with

1262: wavelength $\lambda\sim 2\pi h/\beta^{1/2}$) if $\beta\approxlt 100$.

1263: To ensure that we have, in fact, achieved adequate resolution, we

1264: compare runs MHD\_2 and its high resolution counterpart, run MHD\_2hr.

1265: For example, Fig.~\ref{fig:mhd_compare} compares the time-dependence

1266: of the thermal and magnetic energies, as well as the height dependence

1267: of the plasma-$\beta$ parameter.  While there is some indication of

1268: increased buoyancy driven escape of magnetic fields from the

1269: high-resolution simulation (as is apparent from the higher value of

1270: $\beta$ at intermediate heights), the two simulations generally

1271: compare very well.  Thus, we conclude that we have achieved adequate

1272: numerical resolution.

1273: 

1274: \subsection{Temporal power spectra of the basic fluid variables}

1275: 

1276: \subsubsection{The importance of correcting for secular changes 

1277: during the simulation}

1278: \label{sec:decay_corr}

1279: 

1280: The length of our canonical MHD run makes it well suited for a detailed

1281: study of temporal variability.  In particular, the long stream of

1282: simulation data facilitates the construction of PSDs.  However,  as we

1283: now discuss, significant complications arise in the analysis of these

1284: MHD simulation as compared with the pure hydrodynamic simulations.

1285: 

1286: In the case of the hydrodynamic simulations, the background (i.e.,

1287: unperturbed) flow achieves almost a stationary state once the large

1288: transients caused by the initial conditions have died out.  In

1289: particular, the lack of angular momentum transport within the

1290: hydrodynamic disk (other than that due to the very small numerical

1291: viscosity) allows the disk to achieve a non-accreting state.  Density

1292: and pressure fluctuations about this background state can then be

1293: studied via the straightforward construction of the PSD.  The constant

1294: background level does not contribute to the power spectrum and hence

1295: the PSD faithfully characterizes the fluctuations of interest.

1296: 

1297: However, even once the initial transients have been dissipated, MHD

1298: disks never achieve this kind of stationary background flow.  MHD

1299: turbulence leads to continued accretion that depletes mass from the

1300: simulated disks.  This decline in total mass leads to secular changes

1301: (with an approximately exponential form) in the density and pressure

1302: of the background flow.  Unless corrected for, even a rather slow

1303: exponential decay can have a significant influence on the PSD of the

1304: pressure or density fluctuations, severely affecting attempts to

1305: characterize the properties of the astrophysically relevant

1306: fluctuations (i.e., the fluctuations that would be present in the

1307: ideal case of a steady-state disk in which the mass was replaced from

1308: a large reservoir).

1309: 

1310: To see this, consider some quantity whose time-series $f(t)$ we extract

1311: from the simulation (for example, this could be the mid-plane

1312: density or pressure at some given radius).  Let us assume that this

1313: can be decomposed as

1314: \begin{equation}

1315: f(t)=\epsilon(t)[1+g(t)],

1316: \label{eq:fluc_decomp}

1317: \end{equation}

1318: where $g(t)$ is the time-series of the astrophysically-interesting

1319: fluctuations about some mean state (i.e., $\langle g\rangle=0$), and

1320: $\epsilon(t)$ is a decay function that describes the secular change in

1321: the background state due to the draining of mass from the simulation.

1322: It must be noted that the decomposition given in

1323: eq.~\ref{eq:fluc_decomp} is not completely general; this assumes that

1324: the amplitude of the ``real'' fluctuations is proportional to the mean

1325: background value, and that the properties of the fluctuations

1326: otherwise remain invariant as the background state decays.  This

1327: decomposition would be valid if the decay of the simulated disk simply

1328: amounted to a gradual decline in the density scale of the simulation

1329: while the temperature and velocity structures remained unchanged.  We

1330: shall refer to such (simulated) disks as {\it density-invariant

1331: disks}.  This does indeed appear to describe our simulated MHD disks

1332: (i.e., there is little or no corresponding secular change in disk

1333: thickness or characteristic turbulent velocities).  In this case, the

1334: mass accretion rate will be proportional to the density and the decay

1335: will hence have an exponential form,

1336: $\epsilon=\epsilon_0\,e^{-t/t_0}$.

1337: 

1338: In the uninteresting case where there are no fluctuations ($g(t)=0,

1339: \forall t$), the observed signal is just $f(t)=\epsilon(t)$, leading

1340: to a Fourier Transform (FT) and PSD given by

1341: \begin{equation}

1342: \tilde{\epsilon}(\omega)=\frac{A_1}{(1/t_0)+i\omega}, \hspace{0.5cm}{\rm and}\hspace{0.5cm}P_\epsilon(\omega)=\frac{A_2}{(1/t_0)^2+\omega^2},

1343: \end{equation}

1344: where $A_1$ and $A_2$ are uninteresting normalization constants.

1345: Thus, for $\omega\gg 1/t_0$, the PSD of the exponential decay goes as

1346: $P_\epsilon(\omega)\propto\omega^{-2}$.

1347: 

1348: In the more interesting case of non-zero fluctuations, the FT of the

1349: observed signal is,

1350: \begin{equation}

1351: \tilde{f}(\omega)=\tilde{\epsilon}(\omega)+\int

1352: \tilde{\epsilon}(\omega^\prime)\tilde{g}(\omega-\omega^\prime)\,d\omega^\prime,

1353: \label{eq:combined_ft}

1354: \end{equation} 

1355: i.e., the sum of the FT of the exponential decay with the convolution

1356: of the FTs of the decay and the interesting fluctuations.  When the

1357: fluctuations are small compared with the (decaying) background state,

1358: as is the case for the density and pressure, one can see that the PSD

1359: of the observed signal will be dominated by the $1/\omega^2$

1360: associated with the decay.  Even in the case where the fluctuations

1361: are large compared with the decay, the PSD will still be influenced by

1362: the exponential decay due to the convolution term in

1363: eqn.\ref{eq:combined_ft}.  In particular, regions of the ``real'' PSD

1364: which are steeper than $\omega^{-2}$ (including the high-frequency

1365: wing of any QPO or regions above a high-frequency break) will tend to

1366: get filled.

1367: 

1368: Clearly, we must correct for this decay of the background state, and

1369: be cognizant of manifestations of any remaining uncorrected effects of

1370: this decay.  This procedure plays the same role as the

1371: ``pre-whitening'' employed by Schnittman, Krolik, \& Hawley (2006;

1372: hereafter SKH).  We proceed by dividing the observed time series by a

1373: ``best-fitting'' exponential decay function.  Here, the time-constant

1374: of the exponential decay $t_0$ is estimated via two methods.  Firstly,

1375: we can set $t_0$ by the requirement that the starting and final values

1376: of the observed series are equal [$f(t=0)=f(t=T)$, where $t=T$

1377: corresponds to the end of the time-series].  This is the procedure

1378: adopted by Schnittman et al. (2006) except that they employ a linear

1379: form for the decay function.  Secondly, we can set $t_0$ via a

1380: least-square fit of an exponential form to the observed time-series.

1381: These methods give very similar values of $t_0$ and similar corrected

1382: PSDs when applied to density or pressure fluctuations, as we shall now

1383: see.

1384: 

1385: \subsubsection{Power spectra of fluid variables in disk mid-plane}

1386: \label{sec:mhdfluidpsd}

1387: 

1388: % FIG 12 HERE

1389: \begin{figure*}

1390: \begin{center}

1391: \hbox{

1392: \psfig{figure=f12a.ps,width=0.55\textwidth}

1393: \psfig{figure=f12b.ps,width=0.55\textwidth}

1394: }

1395: \hbox{

1396: \psfig{figure=f12c.ps,width=0.55\textwidth}

1397: \psfig{figure=f12d.ps,width=0.55\textwidth}

1398: }

1399: \end{center}

1400: \caption{PSD for azimuthally-averaged midplane fluid quantities from

1401: in run MHD\_1.  Panels show PSDs for radial velocity (top left),

1402: vertical velocity (top right), density (bottom left) and pressure

1403: (bottom right).  The density and pressure variables have been divided

1404: by the least-square best-fitting exponential function to correct for

1405: the secular decay of the simulation, as discussed in

1406: \S~\ref{sec:decay_corr}.  Also shown are the radial epicyclic

1407: frequency (solid), orbital frequency (dashed) and the $n=1,2,3$ pure

1408: vertical p-modes (from left to right dot-dashed lines).  The absolute

1409: scaling of the PSDs, as indicated by the color-bar, is arbitrary.  }

1410: \label{fig:mhd_psd}

1411: \end{figure*}

1412: 

1413: We now examine the PSD of the azimuthally averaged fluid variables

1414: (velocities, pressure and density) for the simulated MHD disks, as we

1415: did with the hydrodynamic disks in \S~\ref{sec:axisym}.  The top

1416: panels of Fig.~\ref{fig:mhd_psd} show the PSD for the radial and

1417: vertical components of velocity in the midplane of the disk, $v^{\rm

1418: mid}_r(r,t)$ and $v^{\rm mid}_z(r,t)$ for a duration lasting $\Delta

1419: t=409.6T_{\rm isco}$ starting at $t=100T_{\rm isco}$ (i.e., well after

1420: all of the initial transients have dissipated and a quasi-steady

1421: turbulent state has been established).  Note that, outside of the

1422: ISCO, these velocity components are themselves first-order fluctuating

1423: quantities (i.e., the radial and vertical velocity of the background

1424: state is approximately zero) and, within the density-invariant disk

1425: assumption, will have characteristic fluctuation amplitudes that

1426: remain constant throughout the simulation.  Hence, we do not need to

1427: correct these PSDs for the effect of the density decay in the

1428: simulation.

1429: 

1430: It is readily seen that the PSD of the midplane radial velocity

1431: $v_r^{\rm mid}$ is dominated by the radial epicyclic frequency from

1432: $r\sim 7-8r_g$ out to the outer radial boundary.  Inside of

1433: $r\approx 7-8r_g$, the PSD shows broad band power associated with

1434: the transition from turbulent to plunging flow.  

1435: 

1436: The PSD of the midplane vertical velocity $v^{\rm mid}_z$ is quite

1437: different and can be interpreted in the light of the {\it local} fluid

1438: oscillations discussed in \S~\ref{sec:local_osc}.  Below the radial

1439: epicyclic frequency, there is a broad spectrum of g-modes.  Above the

1440: orbital frequency, the distinct modes described by

1441: eqn.~\ref{eq:vert_modes} become apparent; the $n=0$ mode

1442: ($\omega>\Omega$) and $n=2$ mode ($\omega>2.08\Omega$) can be picked

1443: out as distinct tracks, and there are hints of higher frequency modes

1444: as well.  The fact that only even-$n$ modes are seen is readily

1445: understood given that the odd-$n$ modes have $v_z$-nodes at the

1446: midplane, as is evident from the power deficit centered on the $n=1$

1447: frequency in the $v_z$ PSD.

1448: 

1449: The bottom panels of Fig.~\ref{fig:mhd_psd} show the corresponding PSD

1450: for the density and pressure from the canonical simulation.  The

1451: displayed PSDs have been corrected for the decay of the background

1452: state by dividing through by an exponential curve that best-fits (in a

1453: least square sense) the domain-averaged density or pressure.  Very

1454: similar results are obtained using an exponential determined from just

1455: the end points.  The radial banding of the low-frequency noise is

1456: largely the effect of the pre-whitening procedure; the full

1457: (non-corrected) low-frequency power is somewhat greater and the

1458: bands correspond to the residuals between the real decay and the

1459: simple exponential model.  In addition to this low frequency noise

1460: (which is of secondary interest to the investigation presented in this

1461: paper), both PSDs show an enhancement corresponding to the $n=1$

1462: vertical p-mode (with $\omega>1.63\Omega$).  As expected, there is no

1463: enhancement in the density or pressure PSDs corresponding to the $n=0$

1464: or $n=2$ p-modes; these have pressure and density nodes at the

1465: midplane.

1466: 

1467: % FIG 13 HERE

1468: \begin{figure}

1469: \centerline{

1470: \psfig{figure=f13.ps,width=0.55\textwidth}

1471: }

1472: \caption{PSD of the midplane (decay-corrected) pressure for $\Delta

1473: r=0.5r_g$ wide zones centered on $r=7r_g$ (top, red curve),

1474: $r=8r_g$ (middle, black curve), and $r=9.2r_g$ (bottom, green

1475: curve).  The thin vertical dashed line marks the maximum radial

1476: epicyclic frequency.  A line-segment with a slope of $-2$ is shown for

1477: reference.  Also shown (heavy vertical blue lines) are the radial and

1478: vertical epicyclic frequencies at $r=9.2r_g$ where the simplest

1479: form of the parametric instability model would predict resonances.}

1480: \label{fig:mhd_1dpsd}

1481: \vspace{0.5cm}

1482: \end{figure}

1483: 

1484: 

1485: In stark contrast to the hydrodynamic simulation, the

1486: (decay-corrected) midplane pressure PSD does not show the

1487: $\kappa$-bounded vertical ``ridge'' on the frequency-radius plane that

1488: is characteristic of trapped g-modes.  The absence of an excited

1489: trapped g-mode is also confirmed by examining the PSD at $r=8r_g$

1490: (Fig.~\ref{fig:mhd_1dpsd}, normalized such that a direct comparison

1491: can be made with the 3-d hydrodynamic results of

1492: Fig.~\ref{fig:hd3d_psd}).  It is important to note, however, that a

1493: trapped g-mode of the strength seen in our 3-d hydrodynamic simulation

1494: would not stand out from the background turbulence.  Thus, while it is

1495: apparent that the fundamental trapped g-mode is not strongly excited

1496: by the turbulence (in the way that the local p-modes are, for

1497: example), we cannot say whether the g-mode is actively damped by the

1498: turbulence.

1499: 

1500: Finally, we note that there is no indication that the parametric

1501: resonance instability of Abramowicz \& Kluzniak is at work, at least

1502: in its simplest form.  As discussed in \S~2, this model predicts its

1503: strongest resonance at the location where the radial and vertical

1504: epicyclic frequencies are in a 2:3 ratio; this occurs at $r=9.2r_g$

1505: in the PW potential.  As shown in Fig.~\ref{fig:mhd_1dpsd}, the PSD of

1506: the pressure fluctuations at $r=9.2r_g$ shows no structure

1507: associated with the local epicyclic frequencies.  We have verified

1508: that a similar conclusion holds true for the PSD of the other

1509: variables.  While it is possible that non-linear effects have shifted

1510: the location of the resonance inwards from $r=9.2r_g$ (A2003), we

1511: note that the PSDs of Fig.~\ref{fig:mhd_psd} show no obvious radius at

1512: which the power at the radial epicyclic and the orbital frequencies

1513: appears to be locally enhanced.

1514: 

1515: \subsection{Temporal properties of the instantaneous black hole accretion rate}

1516: \label{sec:lightcurves}

1517: 

1518: The previous section addressed the temporal properties of the

1519: fundamental fluid variables through the body of the disk.  Of course,

1520: real observations of accretion disks measure the electromagnetic

1521: radiation generated by the accretion flow.  While our simulations

1522: miss all of the physics relevant for making a first-principles

1523: prediction of the observed lightcurve, it is still instructive to

1524: analyze a simple scalar quantity that can be generated from the

1525: simulation and may be related to the observed hard X-ray radiation

1526: (since it is the hard X-rays that carry the HFQPO signal).

1527: 

1528: % FIG 14 HERE

1529: \begin{figure}[b]

1530: \centerline{

1531: \psfig{figure=f14.ps,width=0.55\textwidth}

1532: }

1533: \caption{The instantaneous mass accretion rate onto the black hole

1534: from our canonical MHD simulation (run MHD\_1).}

1535: \label{fig:lightcurve}

1536: \end{figure}

1537: 

1538: 

1539: Here we consider one such proxy for the observed lightcurve, the

1540: {\it instantaneous mass accretion rate} into the black hole,

1541: \begin{equation}

1542: \dot{M}=\int_{\partial {\cal R}\,_i}r(-v_r)\rho\,dS,

1543: \end{equation}

1544: where the integral is performed over the surface $\partial {\cal

1545: R}\,_i$ defining the inner radial boundary of the computational

1546: domain.  Figure~\ref{fig:lightcurve} shows the raw lightcurve (i.e.,

1547: not corrected for the exponential decay of the disk).  In the

1548: remainder of this section we address the temporal properties of this

1549: lightcurve.  Of particular interest is the presence of breaks or QPOs

1550: in the PSD of this light curve.  Hence, we need a quantitative

1551: approach by which the significance of such features in the PSD can be

1552: assessed.  We begin by discussing our general statistical approach,

1553: which differs from the one advocated by SKH.

1554: 

1555: \subsubsection{Analysis method}

1556: 

1557: Our approach to PSD fitting is predicated on the assumption (also made

1558: by SKH) that at a given frequency the power density has an exponential

1559: probability distribution; if the mean is $p_0$, then the probability

1560: of measuring a power between $p$ and $p+dp$ is

1561: \begin{equation}

1562: P(p)dp={1\over p_0}e^{-p/p_0}dp\; .

1563: \end{equation}

1564: Suppose that we have a model that predicts a power density

1565: $p_{\rm mod,i}$ for frequency bin $i$, and that in our MHD

1566: simulation we actually observe a power density $p_{\rm obs,i}$

1567: in that bin.  The likelihood of the data given the model in

1568: that bin is then

1569: \begin{equation}

1570: {\cal L}_i=(1/p_{\rm mod,i})\exp(-p_{\rm obs,i}/p_{\rm mod,i}).

1571: \end{equation}

1572: The likelihood of the whole power density spectrum given the

1573: model is the product of the individual likelihoods, but the

1574: log likelihood is typically more useful:

1575: \begin{equation}

1576: \ln{\cal L}=\sum_i \left[-\ln p_{\rm mod,i}-p_{\rm obs,i}/p_{\rm mod,i}

1577: \right]\; .

1578: \end{equation}

1579: This is the figure of merit for a given model.  It can therefore be

1580: used for standard tasks such as parameter estimation and model

1581: comparison (e.g., determining if a QPO or break is required by the

1582: data).  

1583: 

1584: Note that to maximize the information content, the bin size should be

1585: the smallest possible, in this case the frequency resolution of the

1586: raw PSD.  As a result, this method does not require rebinning to

1587: coarser resolution.  Our approach therefore yields an accurate

1588: evaluation of one or more precisely specified models, as opposed to

1589: the broader but less sensitive method of trying to detect a signal in

1590: a model-independent way.

1591: 

1592: We apply a Markov Chain Monte Carlo method to search for best fits and

1593: establish confidence regions.  As discussed in

1594: \S~\ref{sec:decay_corr}, we correct for secular changes.  Given the

1595: large amplitude fluctuations in the $\dot{M}$ and $S_{\rm tot}$

1596: curves, the end-point method discussed in \S~\ref{sec:decay_corr} is

1597: not appropriate.  Hence, we employ the least-square method described

1598: in \S4.3.1 in order to determine and then divide out the best-fitting

1599: exponential decay.  

1600: 

1601: \subsubsection{Results}

1602: \label{sec:mdot_psd}

1603: 

1604: % FIG 15 HERE

1605: \begin{figure*}

1606: \hbox{

1607: \psfig{file=f15a.ps,width=0.45\textwidth}

1608: \psfig{file=f15b.ps,width=0.45\textwidth}

1609: }

1610: \caption{ Comparison of the PSD and Fourier phase of the mass

1611: accretion rate without (left panel) and with (right panel)

1612: multiplication of the time series by an exponential in time designed

1613: to compensate for the slow loss of mass from the disk.  The similarity

1614: of the two implies that at these frequencies the variability is

1615: dominated by intrinsic fluctuations instead of by secular changes in

1616: the disk.}

1617: \label{fig:mdotcomp}

1618: \end{figure*}

1619: 

1620: 

1621: The mass accretion rate $\dot{M}$ has sufficient intrinsic variability

1622: that secular changes in the disk properties have only minor effects on

1623: the PSD.  This is evident from Fig.~\ref{fig:mdotcomp}, which compares

1624: the PSD and Fourier phase of the raw $\dot{M}$ curve as a function of

1625: frequency (left panel) with these after multiplication by the

1626: exponential in time that minimizes the overall rms amplitude (right

1627: panel).  The lack of significant differences suggests that inferences

1628: drawn from the PSD are robust.  The comparison shown in

1629: Fig.~\ref{fig:mdotcomp} is for the second quarter of run MHD\_1

1630: (approximately $t=150T_{\rm isco}$ to $t=310T_{\rm isco}$); analyses

1631: of the third and fourth quarters also reveal a lack of sensitivity to

1632: the secular decay of the disk.

1633: 

1634: Having established that the secular decay of the disk is unimportant

1635: for the PSD of ${\dot M}$, we will work directly with the raw time

1636: series from run MHD\_1, without multiplication by an exponential

1637: function.  We analyze separately the second, third, and fourth quarter

1638: of the data (each encompassing a period of $\Delta t\approx 160T_{\rm

1639: isco}$) to look for trends or stability in the PSD, while discarding

1640: the first quarter as potentially biased by initial conditions.  The

1641: results are summarized in Table~\ref{tab:tabmdot}, where we compare

1642: single power law models for the PSD with models involving a broken

1643: power law.  We use the method described in the previous section; note

1644: that only differences in the log likelihood $\ln{\cal L}$ are

1645: important rather than the absolute magnitude of $\ln{\cal L}$.

1646: 

1647: \begin{table*}

1648: \centering

1649: \begin{tabular}{lccr}\hline

1650: Data segment&Model&Parameters and uncertainties&Maximum $\ln{\cal

1651: L}$\\\hline

1652: 1&Single PL&$\Gamma=2.50\pm 0.05$&9131\\

1653: &Broken PL&$\Gamma_1=0.82\pm 0.09$, $\Gamma_2=2.96\pm 0.07$,&9238\\

1654: &&$\log_{10}(\nu_{\rm break})=-1.94\pm 0.04$\\\hline

1655: 2&Single PL&$\Gamma=2.70\pm 0.08$&6508\\

1656: &Broken PL&$\Gamma_1=1.45\pm 0.11$, $\Gamma_2=2.91\pm 0.10$,&6609\\

1657: &&$\log_{10}(\nu_{\rm break})=-1.65\pm 0.07$\\\hline

1658: 3&Single PL&$\Gamma=2.16\pm 0.04$&9561\\

1659: &Broken PL&$\Gamma_1=1.47\pm 0.07$, $\Gamma_2=2.89\pm 0.09$,&9630\\

1660: &&$\log_{10}(\nu_{\rm break})=-1.60\pm 0.04$\\\hline

1661: \end{tabular}

1662: \caption{Model comparisons for PSD of mass accretion rate $\dot{M}$.

1663: $\Gamma$ is the power law index of a single power-law model for the

1664: PSD; $P(\nu)\propto \nu^{-\Gamma}$.  $\Gamma_1$ and $\Gamma_2$ are

1665: the

1666: two indices obtained from a broken power law; $P\propto

1667: \nu^{-\Gamma_1}$ for $\nu<\nu_{\rm break}$, $P\propto

1668: \nu^{-\Gamma_2}$

1669: for $\nu>\nu_{\rm break}$.  All error bars are one standard

1670: deviation.}

1671: \label{tab:tabmdot}

1672: \end{table*}

1673: 

1674: From this analysis, we find compelling evidence for a break in the

1675: power law characterizing the PSD of ${\dot M}$.  Compared to a single

1676: power law, the broken power fit is better by $\Delta\ln{\cal

1677: L}=76-90$, hence the maximum likelihood ratio is at least

1678: $\exp(76)=10^{33}$ in all three data segments independently.  The

1679: break frequency and power law slopes are consistent between the second

1680: and third data segments, but these do not match the first data

1681: segment.  The break frequencies are within a factor of two of, but not

1682: consistent with, the orbital frequency at the ISCO, $\log_{10}(\nu_{\rm

1683: ISCO})=-1.79$.  Therefore, although there is a clear steepening in the

1684: power density spectrum, it is not possible at this point to assign a

1685: specific physical meaning to the break.  We note that this general

1686: form of the PSD, i.e., an approximate power-law with curvature or a

1687: break at frequencies close to the ISCO orbital frequency, has been

1688: previously seen in the mass accretion rate of global disk simulations

1689: (Hawley \& Krolik 2001, 2002).

1690: 

1691: We see no indications of QPOs in the $\dot{M}$ PSD.  Quantitatively,

1692: we add a Lorentzian QPO to the broken power-law PSD model in which the

1693: QPO centroid frequency, full-width half-maximum, and amplitude are

1694: allowed to be free parameters, with the one restriction that the

1695: quality factor of the QPO must exceed 2.  We find that the peak power

1696: of any QPO cannot exceed 2\% of the continuum power measured at the

1697: centroid of the QPO at the 99\% confidence level.  

1698: 

1699: \section{Discussion}

1700: 

1701: \subsection{Comparison with previous numerical results}

1702: 

1703: In recent years, several groups have reported temporal analyses of MHD

1704: disk simulations.  Here, we briefly compare our work with some of

1705: these published results.

1706: 

1707: Probably the most relevant previous work is that of Arras, Blaes \&

1708: Turner (2006; hereafter ABT).  These authors perform a local, shearing

1709: box MHD simulation of a patch of an accretion disk; this provides a

1710: controlled environment in which fluid modes can be characterized.  ABT

1711: find that the MHD turbulence excites a spectrum of distinct acoustic

1712: modes as well as radial epicyclic motions.  However, they note a lack

1713: of distinct inertial modes (g-modes) and use this fact to argue

1714: against the excitation of trapped g-modes in global accretion disks.

1715: Our findings are completely in line with those of ABT, and represent

1716: an extension of ABT's conclusions to global simulations of thin

1717: accretion disks.

1718: 

1719: There have been QPOs reported from global simulations.  Kato (2004d)

1720: performed a global MHD disk simulation in a PW potential and presented

1721: an analysis of quantities derived from the mass accretion rate.

1722: Through visual inspection of the resulting PSDs from four periods of

1723: the (long) simulation, he reports a pair of transient QPOs and a pair of

1724: QPOs that are labeled as persistent (although it is not clear that they

1725: are present in the PSD of all data segments, and the statistical

1726: significance of the features is unclear).  The QPOs are attributed

1727: to resonances between the vertical and radial epicyclic frequencies,

1728: and it is found that these QPO pairs have frequency ratios that are

1729: {\it approximately} 3:2.  We find no evidence of these resonances in

1730: our simulations.  In another interesting difference, an inspection of

1731: the radially-resolved PSDs in Kato (2004d) reveal no signs of the

1732: local p-modes that seem to feature so prominently in our PSDs.  While

1733: the reason for these discrepancies is unclear, we do note that the

1734: Kato (2004d) simulations have an order of magnitude less resolution in

1735: both the azimuthal and (more importantly) the vertical direction,

1736: although his simulations do have a significantly larger computational

1737: domain.  It is possible that the Kato (2004d) simulations have failed

1738: to adequately resolve the vertical dynamics of the thin disk.

1739: 

1740: Chan et al. (2006) have also performed global MHD disk simulations in

1741: a PW potential using the pseudo-spectral algorithm of Chan, Psaltis,

1742: \& \"Ozel (2005, 2006).  Their study included a post-processing of the

1743: simulation to include detailed radiative transfer, and was explicitly

1744: targeted at understanding the variability (including the large

1745: amplitude flaring) of the hot accretion flow around the black hole at

1746: the center of the Galaxy.  They found that the turbulence of the

1747: quiescent flow could only produce a factor of two modulations in the

1748: observed luminosity.  To model the large amplitude flares, they

1749: introduce large density perturbations into the flow.  After being

1750: perturbed, the disk displays a QPO with a frequency equal to the

1751: orbital frequency at the magnetosonic point.  Given that the Chan et

1752: al. study is exploring a rather different regime of accretion than our

1753: present study (i.e., hot, thick accretion flows versus thin, cold

1754: accretion flows), it is hard to make a direct comparison.

1755: 

1756: Finally, SKH have performed detailed analyses of General Relativistic

1757: MHD (GRMHD) simulations of disks performed using the code of

1758: De~Villiers \& Hawley (2003).  In particular, they have studied a long

1759: ($6000GM/c^3$) simulation of a disk around a Schwarzschild black hole,

1760: focusing on the temporal behavior of proxy-lightcurves (rather than

1761: the underlying fluid properties discussed in this paper).  It is

1762: unclear from their analysis whether their simulated accretion disk has

1763: excited local p-modes of the type that we find in this current work.

1764: SKH do find, though, that a proxy-lightcurve based on radiative

1765: transfer through the disk assuming black body emission and free-free

1766: absorption displays QPOs with an approximate 3:2 ratio.  However,

1767: these QPOs are transient, only appearing at certain times and certain

1768: viewing inclinations.  

1769: 

1770: \subsection{Comments on 3:2 frequency ratios}

1771: 

1772: As described above, several authors have reported transient QPO pairs

1773: from MHD simulations with frequency ratio 3:2.  It is tempting to

1774: interpret these as resonance phenomena.  Here, we note that there are

1775: several ways that approximate 3:2 ratios can be generated that do not

1776: necessarily involve resonances and, hence, one should guard against

1777: over-interpreting QPO pairs.  Indeed, we must remember that some

1778: sources have QPO frequency ratios that are inconsistent with 3:2

1779: (e.g., the 67~Hz and 41~Hz QPOs from GRS~1915+105; see Strohmayer

1780: 2001).

1781: 

1782: For example, consider the local vertical p-modes of the disk

1783: (discussed in \S~\ref{sec:local_osc}).  At a given radius, the $n=1$

1784: vertical pressure mode has a frequency of $\omega_{\rm

1785: vert,1}=\sqrt{\gamma+1}\,\Omega$, where $\Omega$ is the orbital

1786: frequency.  For a gas pressure dominated disk in which $\gamma=5/3$,

1787: $\omega_{\rm vert,1}/\Omega=1.63$, and for a radiation pressure

1788: dominated disk in which $\gamma=4/3$, $\omega_{\rm

1789: vert,1}/\Omega=1.53$.  If some unspecified physical process enhances

1790: emission from a given ring of the disk, one can imagine a situation

1791: where QPOs are generated at the orbital frequency and the $n=1$

1792: vertical pressure mode, thereby giving frequency ratios compatible

1793: with the measured ratios in several sources.  Alternatively, the next

1794: lowest vertical mode that has a vertical velocity node in the midplane

1795: (and thus maximum variation of pressure and density there) has a

1796: frequency of $\omega_{\rm vert,3}=\sqrt{3\gamma+1}\,\Omega$.  As a

1797: result, when $\gamma=5/3$ we have $\omega_{\rm vert,3}/\omega_{\rm

1798: vert,1}=1.5$, and when $\gamma=4/3$ the ratio is 1.46.  Once again, if

1799: nature picks out a specific radius and the emission is modulated by

1800: the $n=1$ and $n=3$ modes, we would see QPOs with frequency ratios

1801: entirely consistent with the observations.

1802: 

1803: As another example, suppose that the disk emission is modulated at the

1804: vertical and radial epicyclic frequencies, and that the emission is

1805: distributed in radius according to a standard Page \& Thorne (1974)

1806: disk.  The radial distribution of the emission then peaks close to the

1807: radius where the radial epicyclic frequency is a maximum (and hence is

1808: slowly changing with radius).  One might then expect to see a pair of

1809: QPOs corresponding to these two epicyclic frequencies.  The frequency

1810: ratio depends only weakly on the spin parameter, ranging from 1.46 at

1811: $a/M=0$ to 1.7 at $a/M=0.9$.

1812: 

1813: \section{Conclusions}

1814: 

1815: The origin of HFQPOs remains elusive.  Our simulations of

1816: geometrically-thin accretion disks have shown that MRI-driven MHD

1817: turbulence does not excite the trapped g-modes of Nowak \& Wagoner

1818: (1992), even when those modes definitively exist in the equivalent

1819: hydrodynamic disk.  We have also shown that MHD turbulence does not

1820: excite the parametric resonance instability of Abramowicz \&

1821: Klu\'zniak (2001).  Instead, the only distinct modes found in our

1822: simulated MHD disks are local vertical p-modes and radial epicyclic

1823: oscillations.

1824: 

1825: Clearly, the failure of all simulations to date to produce stable QPO

1826: pairs of the type seen in GBHBs suggests that either the QPOs are

1827: too weak to be detected in the simulations or the models are missing

1828: some important physical ingredients.  It is an open question whether

1829: the global disk modes or the parametric instabilities discussed in

1830: this paper can be excited once one includes the full effects of GR

1831: close to rapidly spinning black holes and/or radiation physics.

1832: Indeed, the fact that HFQPOs are only seen in rather special spectral

1833: states (the soft intermediate state; e.g., see Belloni 2006) when the

1834: accretion rate is thought to be comparable to the Eddington limit

1835: suggests that radiation physics, in particular, may well be important

1836: to the HFQPO mechanism.  It is also noteworthy that the transition

1837: from the hard intermediate state into the soft intermediate state is

1838: associated with powerful relativistic ejection events.  Thus, another

1839: interesting possibility is that HFQPO production occurs in the black

1840: hole magnetosphere (i.e., the base of the jet) and not the accretion

1841: disk at all.

1842: 

1843: \section*{Acknowledgments}

1844: 

1845: We thank Eve Ostriker for insightful discussions as well as allowing

1846: us the use of her MPI-parallelized version of ZEUS.  We are also

1847: grateful to Andy Fabian, Gordon Ogilvie, Sean O'Neill, Jim Pringle,

1848: Jon Miller, Phil Uttley, Simon Vaughan, and Bob Wagoner for comments

1849: and discussion that significantly improved this paper.  All

1850: simulations described in this paper were performed on the Beowulf

1851: cluster (``The Borg'') supported by the Center for Theory and

1852: Computation (CTC) in the Department of Astronomy at the University of

1853: Maryland College Park.  CSR and MCM thank the National Science

1854: Foundation for support under grant AST~06-07428.  CSR also gratefully

1855: acknowledges the University of Maryland's Graduate Research Board

1856: Semester Award Program which supported the early phases of this work.

1857: 

1858: \section*{References}

1859: 

1860: \noindent Abramowicz, M.~A., Almergren, G.~J.~E., Klu\'zniak, W.,

1861: Thampan, A.~V., \& Wallinder, F. 2002, Class. Quant. Grav., 19, L57

1862: 

1863: \noindent Abramowicz, M.~A., Karas, V., Klu\'zniak, W., Lee, W.~H., \&

1864:  Rebusco, P. 2003, PASJ, 55, 467 (A2003)

1865: 

1866: \noindent Abramowicz, M.~A., \& Klu\'zniak, W. 2001, A\&A, 374, L19

1867: 

1868: \noindent Abramowicz, M.~A., \& Klu\'zniak, W. 2003, Gen. Rel. Grav.,

1869: 35, 69

1870: 

1871: \noindent Arras P., Blaes O.~M., \& Turner N.~J. 2006, ApJ, 645, L65 (ABT)

1872: 

1873: \noindent Balbus, S.~A., \& Hawley, J.~F. 1991, ApJ, 376, 214

1874: 

1875: \noindent Balbus, S.~A., \& Hawley, J.~F. 1998, Rev. Mod. Phys., 70, 1

1876: 

1877: \noindent Belloni, T. 2006, Ad. Space Res., 38, 2801

1878: 

1879: \noindent Binney, J., \& Tremaine, S. 1987, Galactic Dynamics (Princeton:

1880: Princeton Univ. Press), 359

1881: 

1882: \noindent Chan, C.~K., Psaltis, D., \& \"Ozel, F. 2005, ApJ, 628, 353

1883: 

1884: \noindent Chan, C.~K., Psaltis, D., \& \"Ozel, F. 2006, ApJ, 645, 506

1885: 

1886: \noindent Chan, C.~K., et al. 2006, ApJ, submitted (astro-ph/0611269)

1887: 

1888: \noindent Churazov, E., Gilfanov, M., \& Revnivtsev, M. 2001, MNRAS, 

1889: 321, 759

1890: 

1891: \noindent Cowling, T.~G. 1957, Quart. J. Mech. Appl. Math., 10, 129

1892: 

1893: \noindent De~Villiers, J.-P., \& Hawley J.~F. 2003, ApJ, 592, 1060

1894: 

1895: \noindent Friedman, J.~L., \& Schutz, B.~F. 1978, ApJ, 221, 937

1896: 

1897: \noindent Gammie, C.~F., McKinney, J.~C., \& T\'oth, G. 2003,

1898: ApJ, 589, 444

1899: 

1900: \noindent Gierlinski M., Middleton M., Ward M., Done C., 2008, Nature,

1901: in press

1902: 

1903: \noindent Gleissner T., Wilms J., Pottschmidt K., Uttley P., Nowak

1904: M.A., Staubert R., 2004, A\&A, 391, 875.

1905: 

1906: \noindent Hawley J.F., Krolik J.H., 2001, ApJ, 548, 348

1907: 

1908: \noindent Hawley J.F., Krolik J.H., 2002, ApJ, 566, 164

1909: 

1910: \noindent Hawley, J.~F., Gammie, C.~F., \& Balbus, S.~A. 1996, ApJ, 464, 690

1911: 

1912: \noindent Hantao, J., Burin, M., Schartman, E., \& Goodman, J. 2006, Nature,

1913: 444, 343

1914: 

1915: \noindent Kato, S. 1990, PASJ, 42, 99

1916: 

1917: \noindent Kato, S. 1991, PASJ, 43, 557

1918: 

1919: \noindent Kato, S. 1993, PASJ, 45, 219

1920: 

1921: \noindent Kato, S. 2004a, PASJ, 56, 559

1922: 

1923: \noindent Kato, S. 2004b, PASJ, 56, 905

1924: 

1925: \noindent Kato, S. 2004c, PASJ, 56, L25

1926: 

1927: \noindent Kato, S., \& Fukue, J. 1980, PASJ, 322, 377

1928: 

1929: \noindent Kato, S., \& Honma, F. 1991, PASJ, 43, 95

1930: 

1931: \noindent Kato, Y. 2004d, PASJ, 56, 931

1932: 

1933: \noindent Krolik, J. 1999, ApJ, 515, L73

1934: 

1935: \noindent Lehr, D., Wagoner, R.~V., \& Wilms, J., 2000, astro-ph/0004211

1936: 

1937: \noindent Lubow, S.~H., \& Pringle, J.~E. 1993, ApJ, 409, 360 (LP93)

1938: 

1939: \noindent Lyubarskii, Yu.~E. 1997, MNRAS, 292, 679

1940: 

1941: \noindent Markoff, S., Nowak, M.~A., Corbel, S., Fender, R.~P., \& Falcke, H.

1942: 2003, New AR, 47, 491

1943: 

1944: \noindent Markovi\'c, D., \& Lamb, F.~K. 1998, ApJ, 507, 316

1945: 

1946: \noindent McClintock, J.~E., \& Remillard, R.~A. 2003, astro-ph/0306213

1947: 

1948: \noindent McHardy, I.~M., K\"ording, E., Knigge, C., Uttley, P., \& Fender,

1949: R.~P. 2006, Nature, 444, 730

1950: 

1951: \noindent Miller K.A., Stone J.M., 2000, ApJ, 534, 398

1952: 

1953: \noindent Nowak, M.~A., \& Wagoner, R.~V. 1991, ApJ, 378, 656 (NW91)

1954: 

1955: \noindent Nowak, M.~A., \& Wagoner, R.~V. 1992, ApJ, 393, 697 (NW92)

1956: 

1957: \noindent Nowak, M.~A., \& Wagoner, R.~V. 1993, ApJ, 418, 187

1958: 

1959: \noindent Nowak, M.~A., Wagoner, R.~V., Begelman, M.~C., \& Lehr, D.~E.

1960: 1997, ApJ, 477, L91

1961: 

1962: \noindent Okazaki, A.~T., Kato, S., \& Fukue, J. 1987, PASJ, 39,

1963: 457

1964: 

1965: \noindent Ortega-Rodriguez, M., Silbergleit, A.~S., \& Wagoner, R.~V.

1966: 2001, ApJ, 567, 1043

1967: 

1968: \noindent Paczynski, B., \& Wiita, P.~J. 1980, A\&A, 88, 23 (PW)

1969: 

1970: \noindent Page, D.~N., \& Thorne, K.~S. 1974, ApJ, 191, 499

1971: 

1972: \noindent Perez, C.~A., Silbergleit, A.~S., Wagoner, R.~V., \&

1973: Lehr, D.~E. 1997, ApJ, 476, 589

1974: 

1975: \noindent Reynolds, C.~S., \& Fabian, A.~C., 2008, ApJ, in press

1976: 

1977: \noindent Rezzolla, L., Yoshida, S.'i., Maccarone, T.~J., \& Zanotti, O.

1978: 2003a, MNRAS, 344, L37

1979: 

1980: \noindent Rezzolla, L., Yoshida, S.'i., \& Zanotti, O. 2003b, MNRAS, 344, 978

1981: 

1982: \noindent Schnittman, J.~D., Krolik, J.~H., \& Hawley, J.~F. 

1983: 2006, ApJ, 651, 1031 (SKH)

1984: 

1985: \noindent Shakura N.I., Sunyaev R.A., 1973, A\&A, 24, 337

1986: 

1987: \noindent Shafee R., McKinney J.C., Narayan R., Tchekhovkoy A., Gammie

1988: C.F., McClintock J.E., 2008, ApJL, in press (arXive:0808.0860)

1989: 

1990: \noindent Silbergleit, A.~S., Wagoner, R.~V., \& Ortega-Rodriguez, M. 

1991: 2001, ApJ, 548, 335

1992: 

1993: \noindent Stone, J.~M., \& Norman, M.~L. 1992a, ApJS, 80, 753

1994: 

1995: \noindent Stone, J.~M., \& Norman, M.~L. 1992b, ApJS, 80, 791

1996: 

1997: \noindent Strohmayer, T.~E. 2001, ApJ, 554, L169

1998: 

1999: \noindent Uttley, P., \& McHardy, I.~M. 2001, MNRAS, 323, L26

2000: 

2001: \noindent Uttley, P., McHardy, I.~M., \& Vaughan, S. 2005, MNRAS, 359, 345

2002: 

2003: \noindent Vaughan, S., \& Uttley, P. 2005, MNRAS, 362, 235

2004: 

2005: \noindent Vernaleo, J.~C., \& Reynolds, C.~S. 2006, ApJ, 645, 83

2006: 

2007: \noindent Wagoner, R.~V., Silbergleit, A.~S., \& Ortega-Rodriguez, M.

2008: 2001, ApJ, 559, L25

2009: 

2010: 

2011: \appendix

2012: 

2013: \section{Fundamental g-mode frequency for very small sound speeds}

2014: 

2015: For very small sound speeds, there are simplifications that allow the

2016: frequency of the fundamental trapped g-mode in a hydrodynamic

2017: accretion disk to be obtained analytically.  Here we base our

2018: analysis on the equations and formalism of NW92.

2019: 

2020: NW92 examine the linearized equations describing the behavior of the

2021: scalar potential

2022: \begin{equation}

2023: \delta u\equiv \delta P/\rho,

2024: \end{equation}

2025: where $\delta P$ is the Eulerian variation in the pressure.  The

2026: radial equation for the perturbation is

2027: \begin{equation}

2028: \omega^2 c_s^2 {\partial^2\delta u\over{\partial r^2}}=

2029: -\left(\omega^2-\gamma\Upsilon\Omega^2\right)

2030: \left(\omega^2-\kappa^2\right)\delta u\; ,

2031: \label{eq:diff_eqn}

2032: \end{equation}

2033: where $\omega$ is the mode angular frequency, $\Omega$ is the orbital

2034: angular frequency, $\kappa$ is the radial epicyclic angular frequency,

2035: and $\gamma$ is the usual adiabatic index $\gamma=5/3$.   $\Upsilon$ is 

2036: determined by the quantization condition

2037: \begin{equation}

2038: {A+1/2\over{(1-4B)^{1/2}}}=j+1/2\; ,

2039: \label{eq:quantization}

2040: \end{equation}

2041: where $A=\Upsilon-\zeta$ and

2042: \begin{equation}

2043: B={\zeta\over\gamma}{(\omega^2-

2044: \gamma\Upsilon\Omega^2)\over{\omega^2}}

2045: \end{equation}

2046: with $\zeta=(\gamma-1)/\gamma= 2/5$.  For the fundamental mode, we

2047: have $j=0$ and eq.~\ref{eq:quantization} can be solved to obtain

2048: \begin{equation}

2049: \Upsilon=0.4{\Omega^2\over \omega^2}-0.2.

2050: \end{equation}

2051: 

2052: Now we concentrate on low sound speeds, $c_s\ll 1$ (within this

2053: Appendix, speeds will be given in units of $c$; frequencies in units

2054: of $c^3/(GM)$; lengths in units of $r_g$).  This implies that the

2055: mode frequency $\omega\approx\kappa_{\rm max}$, where $\kappa_{\rm

2056: max}$ is the maximum radial epicyclic frequency.  In this limit, we

2057: note that the factor ($\omega^2-\gamma\Upsilon\Omega^2$) within

2058: eqn.~\ref{eq:diff_eqn} is close to constant with radius.  Let us

2059: define $D\equiv -(\omega^2-\gamma\Upsilon\Omega^2)>0$, where we

2060: evaluate $D$ by assuming $\omega=\kappa_{\rm max}$.

2061: 

2062: We also note that near the maximum, the radial epicyclic

2063: frequency has a parabolic form, $\kappa^2=\kappa_{\rm max}^2-

2064: E(r-r_{\rm max})^2$, where $r_{\rm max}$ is the radius at which

2065: $\kappa=\kappa_{\rm max}$.  The differential equation then becomes

2066: \begin{equation}

2067: {\partial^2\delta u\over{\partial r^2}}=\left[

2068: -{D\over c_s^2}\left(\kappa_{\rm max}^2/\omega^2-1\right)

2069: +{D\over c_s^2}{E\over\omega^2}\left(r-r_{\rm max}\right)^2\right]\delta u\; .

2070: \end{equation}

2071: This has the form of a harmonic oscillator, so we try a solution

2072: of the type

2073: \begin{equation}

2074: \delta u\propto e^{-{1\over 2}C(r-r_{\rm max})^2}

2075: \end{equation}

2076: meaning that

2077: \begin{equation}

2078: {\partial^2\delta u\over{\partial r^2}}=

2079: \left[C^2(r-r_{\rm max})^2-C\right]e^{-{1\over 2}C(r-r_{\rm max})^2}\; .

2080: \end{equation}

2081: This yields the conditions

2082: \begin{eqnarray}

2083: C&=&{D\over c_s^2}\left(\kappa^2_{\rm max}/\omega^2-1\right)\\

2084: C^2&=&{D\over c_s^2}{E\over\omega^2}\; .\\

2085: \end{eqnarray}

2086: Defining $x\equiv 1/\omega^2$, these two conditions can be combined

2087: (eliminating $C$) to yield

2088: \begin{equation}

2089: {D^2\over c_s^4}\left(\kappa_{\rm max}^4x^2-2\kappa_{\rm max}^2x+1

2090: \right)={D\over c_s^2}Ex\; .

2091: \end{equation}

2092: Solving for $x$ gives

2093: \begin{equation}

2094: x={1\over{2\kappa_{\rm max}^4}}\left[2\kappa_{\rm max}^2+c_s^2E/D

2095: \pm\sqrt{4\kappa_{\rm max}^2c_s^2E/D+c_s^4E^2/D^2}\right]\; .

2096: \end{equation}

2097: Since $\omega<\kappa_{\rm max}$, we choose the positive sign.

2098: For small $c_s$, the first term in the square root dominates,

2099: and the lowest order in $c_s$ gives

2100: \begin{equation}

2101: x\approx {1\over{\kappa_{\rm max}^2}}\left(1+{c_s\over{\kappa_{\rm max}}}

2102: \sqrt{E\over D}\right)\; .

2103: \end{equation}

2104: This implies finally that

2105: \begin{equation}

2106: {\kappa_{\rm max}-\omega\over{\kappa_{\rm max}}}=

2107: {c_s\over{2\kappa_{\rm max}}}\sqrt{E/D}\; .

2108: \end{equation}

2109: This clearly demonstrates that the fractional difference of the mode

2110: frequency from the radial epicyclic maximum is linear in $c_s$.  

2111: 

2112: Evaluating this expression for the PW potential, we obtain

2113: $\kappa^2_{\rm max}=1.202\times 10^{-3}$ 

2114: (at $r=r_{\rm max}=4+2\sqrt{3}=7.464$), and 

2115: \begin{equation}

2116: {\kappa_{\rm max}-\omega\over{\kappa_{\rm max}}}=2.269c_s\; .

2117: \end{equation}

2118: In contrast, the Nowak \& Wagoner pseudo-Newtonian potential, 

2119: \begin{equation}

2120: \Phi_{\rm NW}=-{1\over r}+{3\over r^2}-{12\over r^3}\; .

2121: \end{equation}

2122: yields $\kappa^2_{\rm max}=8.607\times 10^{-4}$ (at $r_{\rm max}=7.746$) and 

2123: \begin{equation}

2124: {\kappa_{\rm max}-\omega\over{\kappa_{\rm max}}}=5.621c_s\; .

2125: \end{equation}

2126: For both potentials, the analytic frequencies agree extremely well

2127: with direct numerical solutions to eqn.~\ref{eq:diff_eqn}.

2128: 

2129: \end{document}

2130: 

2131: