1: %%%\documentclass[12pt,manuscript]{aastex}
2: %\documentclass[12pt,preprint]{aastex}
3: %\doublespace
4: %%% manuscript produces a one-column, double-spaced document:
5: %%% preprint2 produces a double-column, single-spaced document:
6: %%% Apj formatinda nasil cikacagini gormek icin asagidakini
7: \documentclass[final]{emulateapj}
8: \usepackage{apjfonts}
9: %%%%%Derginin istedigi format ise asagidaki
10: %%%manuscript produces a one-column, double-spaced document:
11: %%%preprint2 produces a double-column, single-spaced document:
12: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
13: \usepackage{amssymb,latexsym,amsmath,graphics}
14: \slugcomment{} \shorttitle{QPOs AS GLOBAL MODES IN THE BOUNDARY
15: LAYERS OF ACCRETION DISKS}
16: \shortauthors{ERKUT, PSALTIS, AND ALPAR}
17:
18: \begin{document}
19:
20: \title{Quasi-periodic oscillations as global hydrodynamic modes in the boundary layers of viscous accretion disks}
21:
22: \author{M. Hakan Erkut,\altaffilmark{1,2} Dimitrios Psaltis,\altaffilmark{2,3} and M. Ali Alpar\altaffilmark{3}}
23:
24: \affil{\altaffilmark{1}Department of Mathematics and Computer
25: Science, \.Istanbul K\"{u}lt\"{u}r University, Atak\"{o}y Campus,
26: Bak\i rk\"{o}y 34156, \.Istanbul, Turkey}
27:
28: \affil{\altaffilmark{2}Physics
29: Department, University of Arizona, 1118 E. 4th St., Tucson, AZ
30: 85721}
31:
32: \affil{\altaffilmark{3}Faculty of Engineering and Natural
33: Sciences, Sabanc\i\ University, 34956, Orhanl\i, Tuzla, \.Istanbul,
34: Turkey}
35:
36: \altaffiltext{1}{m.erkut@iku.edu.tr}
37: \altaffiltext{2}{dpsaltis@physics.arizona.edu}
38: \altaffiltext{3}{alpar@sabanciuniv.edu}
39:
40: \begin{abstract}
41: The observational characteristics of quasi-periodic oscillations
42: (QPOs) from accreting neutron stars strongly indicate the
43: oscillatory modes in the innermost regions of accretion disks as a
44: likely source of the QPOs. The inner regions of accretion disks
45: around neutron stars can harbor very high frequency modes related to
46: the radial epicyclic frequency $\kappa $. The degeneracy of $\kappa
47: $ with the orbital frequency $\Omega $ is removed in a non-Keplerian
48: boundary or transition zone near the magnetopause between the disk
49: and the compact object. We show, by analyzing the global
50: hydrodynamic modes of long wavelength in the boundary layers of
51: viscous accretion disks, that the fastest growing mode frequencies
52: are associated with frequency bands around $\kappa $ and $\kappa \pm
53: \Omega $. The maximum growth rates are achieved near the radius
54: where the orbital frequency $\Omega $ is maximum. The global
55: hydrodynamic parameters such as the surface density profile and the
56: radial drift velocity determine which modes of free oscillations
57: will grow at a given particular radius in the boundary layer. In
58: accordance with the peak separation between kHz QPOs observed in
59: neutron-star sources, the difference frequency between two
60: consecutive bands of the fastest growing modes is always related to
61: the spin frequency of the neutron star. This is a natural outcome of
62: the boundary condition imposed by the rotating magnetosphere on the
63: boundary region of the inner disk.
64: \end{abstract}
65:
66: \keywords{accretion, accretion disks --- stars: neutron --- stars:
67: oscillations --- X-rays: stars}
68:
69:
70: \section{Introduction\label{intr}}
71:
72: Wave modes in the boundary region in the inner accretion disk are a
73: likely source of the distinct narrow frequency bands of
74: quasi-periodic oscillations (QPOs) (van der Klis 2000) from neutron
75: star sources in low mass X-ray binaries (LMXBs). The dependence of
76: QPO frequencies on accretion rate suggests that the observed QPOs
77: are connected with accretion disk modes. Psaltis, Belloni \& van der
78: Klis (1999) showed the existence of correlations between the
79: frequencies of different QPO bands extending over a wide span of
80: frequencies. The same correlation encompasses black hole as well as
81: white dwarf and neutron star sources. This strongly implies that the
82: frequency bands are determined by the oscillation modes of the
83: accretion disk. The nature of the compact object, whether it is a
84: black hole, white dwarf or neutron star, probably plays a role in
85: exciting the same disk modes through possibly different mechanisms
86: in different types of compact sources. The QPO bands modifying the
87: X-ray luminosity are likely to belong to the boundary region between
88: the disk and the compact object. For magnetic neutron stars, this
89: boundary region is shaped by the interaction of the disk with the
90: magnetosphere. The region where the rotation deviates from Keplerian
91: flow is not necessarily very narrow; this boundary zone may have a
92: size as large as a few tenths of the inner disk radius. We shall
93: therefore use the terms \textquotedblleft boundary\textquotedblright
94: \thinspace and \textquotedblleft transition\textquotedblright
95: \thinspace region interchangeably.
96:
97: Initial explorations of disk modes underlying QPO frequency bands
98: from LMXBs were provided by Alpar et al. (1992) and Alpar \& Y\i
99: lmaz (1997). The characteristics and possible excitation mechanisms
100: of thin-disk oscillations were reviewed and discussed by Kato
101: (2001). In many current models of QPOs, especially those of neutron
102: stars, their frequencies are identified with test-particle
103: frequencies (Stella, Vietri \& Morsink 1999; Abramowicz et al.
104: 2003). Most parts of an accretion disk, except, significantly, the
105: transition regions at the inner boundary of the disk, are
106: characterized by Keplerian rotation rates to a very good
107: approximation. At any radius beyond the transition region, acoustic,
108: magneto-acoustic and viscous corrections to the test-particle
109: Keplerian orbital frequency are negligible. In the outer disk regime
110: for disks around neutron stars and white dwarfs (and far enough from
111: a central black hole), test-particle frequencies for radial and
112: vertical perturbations of the orbit are degenerate with the
113: Keplerian orbital frequency. The degeneracy is removed if the
114: effects of general relativity are important. A Newtonian field of
115: tidal or higher multi-pole structure would also lead to a split in
116: the degeneracy. However, in Newtonian gravity, distortions of
117: stellar shape even for the most rapidly rotating neutron stars will
118: not introduce a significant level of non-degeneracy between the
119: frequencies of test-particle oscillations that is comparable to the
120: difference between the observed QPO frequency bands. Models
121: employing the Keplerian frequency have been relatively successful in
122: interpreting QPO frequency correlations. The Kepler (test particle)
123: frequency interpretation has also been employed to place constraints
124: on the masses and spins of compact objects. Empirically, predictions
125: based on test-particle frequencies applied to QPO frequency
126: correlations deviate from observations at a level of about 10\%.
127:
128: The Keplerian frequency represents the basic imprint of rotation on
129: all the dynamical responses of the disk. Albeit simple, the
130: identification of QPOs with test-particle frequencies has a number
131: of shortcomings. First, precisely in a boundary layer where QPOs are
132: expected to be excited, the dynamical frequencies of oscillations
133: are not the test-particle frequencies. In the boundary layer,
134: viscous and magnetic forces lead to deviation of orbital frequencies
135: from Keplerian test-particle frequencies (see, e.g., Erkut \& Alpar
136: 2004). The resulting band of non-Keplerian rotation frequencies
137: entail viscous and acoustic response and couplings between adjacent
138: rings of the disk fluid. Hydrodynamic effects are therefore
139: essential for an understanding of the frequency bands characterizing
140: the boundary layer. In particular, the degeneracy of test-particle
141: frequencies in the non-relativistic regime is lifted in the
142: hydrodynamic boundary layer, where the radial epicyclic frequency is
143: in fact larger than the orbital frequency. This simple observation
144: (Alpar \& Psaltis 2005) has important consequences for the
145: interpretation of kHz QPO frequencies, in particular for the
146: constraints on the neutron star mass-radius relation derived from
147: kHz QPOs.
148:
149: Second, test-particle frequencies do not distinguish between
150: azimuthal sidebands. While kinematic models of QPOs like the
151: beat-frequency model involve one specific band of frequencies, say
152: $\omega $, a large number of sidebands, of frequencies $\omega
153: _{m}\cong \omega -m\Omega $, where $\Omega $ is the orbital
154: frequency, are allowed by azimuthal symmetry. Arguments as to why
155: only one or two QPO bands are excited must involve choices imposed
156: by the symmetries of the interaction between the magnetosphere and
157: the accretion disk boundary, leading to resonances with particular
158: frequencies at one or more radial regions in the disk. Without
159: resonances, test-particle frequency spectra present no distinction
160: between the modes. The realistic hydrodynamic modes may, in some
161: parameter ranges, distinguish between both fundamental modes and all
162: their azimuthal sidebands through the different growth or decay
163: rates of these modes, \emph{even in the case of free oscillations}.
164: A reduction of the number of relevant (easily excitable) modes of
165: free oscillations is certainly an important task for an
166: understanding of accretion disks around neutron stars, white dwarfs
167: and black holes. In the case of black holes, this classification of
168: free hydrodynamic modes in terms of their growth rates is even more
169: important as resonant excitation by the black hole is not available.
170:
171: In this paper, we study the global modes of free oscillations at
172: some position $r$ in the inner disk-boundary or transition region by
173: analyzing the perturbed dynamical equations of a hydrodynamic disk,
174: including pressure gradients, viscous and magnetic stresses. We
175: consider the case of a disk around a neutron star with a
176: magnetosphere. The neutron star is taken to be a \emph{slow
177: rotator}: the star's rotation rate is less than the value of the
178: Keplerian rotation rate at the inner radius of the disk; $\Omega
179: _{\ast }<\Omega _{\mathrm{K}}(r_{\mathrm{in}})$. The actual rotation
180: rate of the disk is set by the boundary condition $\Omega
181: (r_{\mathrm{in}})=\Omega _{\ast }$. Thus in the steady state
182: solutions for the disk, the rotation rate at the inner edge of the
183: disk and in the transition region beyond is sub-Keplerian. The
184: innermost disk radius $r_{\mathrm{in}}$ determines the disk
185: magnetosphere interface which may be subject to magnetic
186: Rayleigh-Taylor or interchange instabilities. This instability
187: operates mainly at the disk-magnetosphere interface because of the
188: sharp density
189: contrast across the radius $r=r_{\mathrm{in}}$. The inner disk region for $r\gtrsim r_{\mathrm{%
190: in}}$ consists of a non-Keplerian boundary region that joins the
191: outer Keplerian disk with a continuous density distribution. As we
192: will see in \S\ \ref{ehdfom}, the surface density profile throughout
193: the non-Keplerian boundary region increases with decreasing radii.
194: Thus, our boundary region is not subject to interchange-like
195: instabilities.
196:
197: The free oscillation modes we explore by perturbing a steady state solution
198: embody information about the stability of the disk through their growth or
199: decay rates. Growing modes will provide a clue as to the origin of the QPO
200: frequency bands, which could be excited by resonant forcing of the disk by
201: time dependent interactions with the star's magnetosphere. We find, as
202: expected, that growth or decay rates are determined by the dynamical effects
203: of viscosity, with an additional dependence on the sound speed and the
204: density and rotation-rate profiles in the non-Keplerian boundary region. For
205: some boundary region parameters, not all fundamental modes of free
206: oscillations and not all azimuthal sidebands grow, and among the growing
207: modes the growth rates differ. For other choices of boundary region
208: parameters the growth rates of all sidebands are similar. The frequencies of
209: the modes differ from test-particle frequencies by amounts that can be
210: several times larger than the corresponding QPO width. Most importantly, for
211: a reasonable set of accretion disk parameters, we can show that only a
212: certain few of the hydrodynamic free oscillation modes will grow, and that
213: the frequency bands of these oscillations can be associated with the
214: observed frequency bands. The nature of free oscillation modes in the
215: boundary region depends on whether the neutron star is rotating at a rate
216: slower or faster than the rotation rates prevailing in the inner boundary of
217: the disk. In this paper, we will consider the more common and
218: straightforward case of \emph{slow rotators}, the case when the neutron star
219: rotation rate $\Omega _{\ast }$ is less than $\Omega _{\mathrm{K}}(r_{%
220: \mathrm{in}})$, the Kepler rotation rate at some representative inner disk
221: radius $r_{\mathrm{in}}$. An investigation of \emph{forced} resonant
222: excitation of these prevalent modes, and the comparison and association with
223: observed QPO bands will follow in a subsequent paper.
224:
225: \S\ 2 lays out the basic assumptions and equations, \S\ 3 displays
226: the mode analysis for the free oscillations, including a discussion
227: of hydrodynamic effects, frequency bands, and growth rates. In \S\
228: 4, we discuss the results and present our conclusions.
229:
230: \section{Basic Equations and Assumptions\label{beaa}}
231:
232: We consider the excitation of oscillations in a geometrically thin
233: disk in vertical hydrostatic equilibrium. We write the continuity
234: and the radial and angular momentum equations in cylindrical
235: coordinates, integrated vertically over the disk thickness, as
236: \begin{equation}
237: \frac{\partial \Sigma }{\partial t}+\frac{1}{r}\frac{\partial }{\partial r}%
238: (r\Sigma v_{r})+\frac{1}{r}\frac{\partial }{\partial \phi }(\Sigma v_{\phi
239: })=0, \label{avcnt}
240: \end{equation}
241: \begin{equation}
242: \frac{\partial v_{r}}{\partial t}+v_{r}\frac{\partial v_{r}}{\partial r}+%
243: \frac{v_{\phi }}{r}\frac{\partial v_{r}}{\partial \phi }-\frac{v_{\phi }^{2}%
244: }{r}=-\left( \frac{\partial \Phi }{\partial r}\right) _{z=H}-\frac{1}{\Sigma
245: }\frac{\partial \Pi }{\partial r}+\left\langle F_{r}\right\rangle ,
246: \label{avrm}
247: \end{equation}
248: and
249: \begin{equation}
250: \frac{\partial v_{\phi }}{\partial t}+v_{r}\frac{\partial v_{\phi }}{%
251: \partial r}+\frac{v_{\phi }}{r}\frac{\partial v_{\phi }}{\partial \phi }%
252: +v_{r}\frac{v_{\phi }}{r}=-\frac{1}{\Sigma r}\frac{\partial \Pi }{\partial
253: \phi }+\left\langle F_{\phi }\right\rangle . \label{avazm}
254: \end{equation}
255: Here, $H$ is the half-thickness of the disk, $v_{r}$ and $v_{\phi }$
256: are the radial and azimuthal components of the velocity field in the
257: disk, $\Phi $ is the gravitational potential,
258: \begin{equation}
259: \Sigma \equiv \int_{-H}^{H}\rho dz \label{smd}
260: \end{equation}
261: is the surface mass density,
262: \begin{equation}
263: \Pi \equiv \int_{-H}^{H}Pdz \label{vip}
264: \end{equation}
265: is the vertically integrated thermal pressure, $\rho $ is the mass density, $%
266: P$ is the sum of the gas and radiation pressures, and
267: \begin{equation}
268: \left\langle F_{j}\right\rangle \equiv \frac{1}{\Sigma }\int_{-H}^{H}F_{j}%
269: \rho dz \label{vamv}
270: \end{equation}
271: are the vertically averaged sums of the magnetic and viscous forces per unit
272: mass, and the subscript $j$ stands for the $r$ or $\phi$ component of the
273: force.
274:
275: We only consider the $\phi $-component of the viscous force as the
276: dominant shear stress between adjacent layers, because of our
277: assumption that the disk is geometrically thin (Shakura \& Sunyaev
278: 1973). We write the sum of the viscous and large-scale magnetic
279: forces (per unit mass) in the radial and azimuthal directions as
280: $F_{r}=F_{r}^{\mathrm{mag}}$ and $F_{\phi }=F_{\phi
281: }^{\mathrm{vis}}+F_{\phi }^{\mathrm{mag}}$, respectively.
282:
283: The steady, axisymmetric, equilibrium state obeys
284: \begin{equation}
285: -2\pi r\Sigma _{0}v_{r0}=\dot{M}, \label{eqcnt}
286: \end{equation}
287: \begin{equation}
288: v_{r0}\frac{dv_{r0}}{dr}-\Omega ^{2}r=-\left( \frac{\partial \Phi }{\partial
289: r}\right) _{z=H}-\frac{1}{\Sigma _{0}}\frac{d\Pi _{0}}{dr}+\left\langle
290: F_{r}^{\mathrm{mag}}\right\rangle _{0}, \label{req}
291: \end{equation}
292: and
293: \begin{equation}
294: \frac{\kappa ^{2}}{2\Omega }v_{r0}=\left\langle F_{\phi }^{\mathrm{vis}%
295: }\right\rangle _{0} +\left\langle F_{\phi }^{\mathrm{mag}}\right\rangle _{0},
296: \label{azeq}
297: \end{equation}
298: where $\dot{M}$ is the constant mass-inflow rate, $v_{r0}<0$, $v_{\phi
299: 0}=\Omega r$, $\Omega $ is the angular velocity of the disk plasma, and $%
300: \kappa \equiv \lbrack 2\Omega (2\Omega +rd\Omega /dr)]^{1/2}$ is the radial
301: epicyclic frequency.
302:
303: We introduce perturbations to this axisymmetric fluid distribution
304: and use the subscripts \textquotedblleft 0\textquotedblright\ and
305: \textquotedblleft 1\textquotedblright\ to denote the equilibrium and
306: perturbed states,
307: respectively. The linearized perturbation equations follow from equations (%
308: \ref{avcnt})--(\ref{avazm}) as
309: \begin{equation}
310: \frac{d\Sigma _{1}}{d\tau }+\frac{1}{r}\frac{\partial }{\partial r}\left[
311: r(\Sigma _{0}v_{r1}+\Sigma _{1}v_{r0})\right] +\frac{1}{r}\frac{\partial }{%
312: \partial \phi }(\Sigma _{0}v_{\phi 1})=0, \label{lcnt}
313: \end{equation}%
314: \begin{eqnarray}
315: \frac{dv_{r1}}{d\tau }-2\Omega v_{\phi 1}+\frac{\partial }{\partial r}%
316: (v_{r0}v_{r1}) &=&\frac{1}{\Sigma _{0}}\left( \frac{\Sigma _{1}}{\Sigma _{0}}%
317: \frac{d\Pi _{0}}{dr}-\frac{\partial \Pi _{1}}{\partial r}\right) \notag \\
318: &&+\left\langle F_{r}^{\mathrm{mag}}\right\rangle _{1}, \label{lrd}
319: \end{eqnarray}%
320: and
321: \begin{eqnarray}
322: \frac{dv_{\phi 1}}{d\tau }+\frac{\kappa ^{2}}{2\Omega }v_{r1}+\frac{v_{r0}}{r%
323: }v_{\phi 1}+v_{r0}\frac{\partial v_{\phi 1}}{\partial r} &=&-\frac{1}{\Sigma
324: _{0}r}\frac{\partial \Pi _{1}}{\partial \phi }+\left\langle F_{\phi }^{%
325: \mathrm{vis}}\right\rangle _{1} \notag \\
326: &&+\left\langle F_{\phi }^{\mathrm{mag}}\right\rangle _{1}, \label{liaz}
327: \end{eqnarray}%
328: where $d/d\tau \equiv \partial /\partial t+\Omega \partial /\partial \phi $
329: is the Lagrangian derivative following the motion of the fluid in the
330: azimuthal direction.
331:
332: In the following, we study perturbations varying on timescales
333: shorter than the thermal timescale and thus expect that there is no
334: energy exchange between adjacent fluid elements. In this case, the
335: fluid elements respond adiabatically to density and pressure
336: perturbations and, at the same time, the disk thickness $H$, which
337: is determined by vertical hydrostatic equilibrium, responds to the
338: change of pressure. In general, we can write (Stehle \& Spruit 1999)
339: \begin{equation}
340: \Pi _{1}=\Sigma _{1}\left( \frac{\Pi _{0}}{\Sigma _{0}}\right) \left( \Gamma
341: +\frac{\partial \ln H}{\partial \ln \rho }\right) \;, \label{dnsprt}
342: \end{equation}
343: where $\Gamma $ is the index of a polytropic equation of state. The terms in
344: the last parenthesis is of order unity. Because we will not be calculating
345: explicitly changes in the vertical equilibrium during the oscillations, we
346: will set the last term to unity. Therefore, we will be studying effectively
347: isothermal modes.
348:
349: We consider long wavelength perturbations to analyze the lowest
350: order normal modes extending globally over the radial distance scale
351: $r_{\mathrm{out}}$ of the disk. For the stability of such global
352: modes the effect of cylindrical geometry and of the structure of the
353: unperturbed disk state must be taken into account. We perform a
354: local mode analysis. Our approach differs from the usual local mode
355: analysis of short-wavelength perturbations of radial dependence
356: $e^{ikr}$ in a WKB approximation. Consideration of modes with radial
357: wavenumbers would be appropriate for the analysis of local disk
358: modes in the short wavelength regime. Instead, we study the
359: stability of the global disk modes at each particular radius $r$ in
360: the boundary layer. In this region, the unperturbed disk quantities
361: such as the surface density $\Sigma _{0}$ and the
362: rotation rate $\Omega $ change on a length scale $(\Delta r)_{%
363: \mathrm{BL}}<r_{\mathrm{in}}$. Since the variation of the global
364: long wavelength perturbations has a much longer range
365: $r_{\mathrm{out}}\gg (\Delta r)_{\mathrm{BL}}$ in the radial
366: direction, we write equations~(\ref{lcnt})--(\ref{liaz}) at any
367: particular radius $r$ in dimensionless form, neglecting the
368: $r$-dependence of the perturbations:
369: \begin{equation}
370: \frac{d\sigma }{d\tau }+(1+\beta )\Omega _{s}u_{r}+\beta \Omega _{\nu
371: }\sigma +\Omega _{s}\frac{\partial u_{\phi }}{\partial \phi }=0,
372: \label{ctlc}
373: \end{equation}%
374: \begin{equation}
375: \frac{du_{r}}{d\tau }-2\Omega u_{\phi }+(1+\beta )\Omega _{\nu }u_{r}=\beta
376: \Omega _{s}\sigma +f_{r}^{\mathrm{mag}}, \label{rmlc}
377: \end{equation}%
378: \begin{equation}
379: \frac{du_{\phi }}{d\tau }+\frac{\kappa ^{2}}{2\Omega }u_{r}-\Omega _{\nu
380: }u_{\phi }=-\Omega _{s}\frac{\partial \sigma }{\partial \phi }+f_{\phi }^{%
381: \mathrm{vis}}+f_{\phi }^{\mathrm{mag}}. \label{azlc}
382: \end{equation}%
383: Here, $\sigma \equiv \Sigma _{1}/\Sigma _{0}$ is the dimensionless density
384: perturbation, $u_{r}\equiv v_{r1}/c_{s}$ and $u_{\phi }\equiv v_{\phi
385: 1}/c_{s}$ are the dimensionless velocity perturbations, $f_{r,\phi }\equiv
386: \left\langle F_{r,\phi }\right\rangle _{1}/c_{s}$ are the force
387: perturbations in units of frequency, and $\Omega _{\nu }\equiv -v_{r0}/r$
388: and $\Omega _{s}\equiv c_{s}/r$ are the typical frequencies associated with
389: the radial accretion and the effective sound speed $c_{s}=(\Pi _{0}/\Sigma
390: _{0})^{1/2}$, respectively. The parameter
391: \begin{equation}
392: \beta \equiv \left( \frac{d\ln \Sigma _{0}}{d\ln r}\right) _{r} \label{bet}
393: \end{equation}%
394: represents the surface mass density profile at a particular radius
395: $r$ in the inner disk. Note that $\beta $ crucially represents the
396: structure of the unperturbed disk. This differs from the usual local
397: mode analysis in which the radial variation of the unperturbed
398: background quantities such as $\Sigma _{0}$ is neglected while the
399: radial derivatives of the perturbations are taken into account in
400: terms of their radial wavenumbers.
401:
402: We look for solutions of the above linear set of equations of the form%
403: \begin{equation}
404: q(\phi ,t)=\sum\limits_{m}\int q_{m}(\omega )e^{i(m\phi -\omega t)}d\omega ,
405: \label{qnt}
406: \end{equation}
407: where the integer $m\geq 1$ represents non-axisymmetric modes
408: corresponding to the prograde motion of perturbations in the
409: azimuthal direction. The Fourier decomposition of equations
410: (\ref{ctlc})--(\ref{azlc}) yields
411: \begin{equation}
412: \left[ i(m\Omega -\omega )+\beta \Omega _{\nu }\right] \sigma _{m}+(1+\beta
413: )\Omega _{s}u_{r,m}+im\Omega _{s}u_{\phi ,m}=0, \label{fdct}
414: \end{equation}
415: \begin{equation}
416: -\beta \Omega _{s}\sigma _{m}+\left[ i(m\Omega -\omega ) +(1+\beta
417: )\Omega_{\nu }\right] u_{r,m}-2\Omega u_{\phi ,m}=f_{r,m}^{\mathrm{mag}},
418: \label{fdrm}
419: \end{equation}
420: \begin{equation}
421: im\Omega _{s}\sigma _{m}+\frac{\kappa ^{2}}{2\Omega }u_{r,m}+\left[
422: i(m\Omega -\omega )-\Omega _{\nu }\right] u_{\phi ,m}=f_{\phi ,m}^{\mathrm{%
423: vis}} +f_{\phi ,m}^{\mathrm{mag}}. \label{fdaz}
424: \end{equation}
425: To proceed further in the linear mode analysis, we need to specify
426: the force perturbations $f_{r,m}$ and $f_{\phi ,m}$ in equations
427: (\ref{fdrm}) and (\ref{fdaz}).
428:
429: In this paper, we discuss the free oscillation modes. In a subsequent paper,
430: we will study oscillations forced by external perturbations, as in the case
431: of a compact object with an oblique large-scale magnetic field.
432:
433: \section{Global Free Oscillations in a Viscous Accretion Disk\label{feos}}
434:
435: When there are no perturbations introduced by large-scale magnetic
436: fields, i.e., when $f_{r,m}^{\mathrm{mag}}=f_{\phi
437: ,m}^{\mathrm{mag}}=0$, the only non-negligible force perturbation is
438: due to the kinematic viscosity $\nu $. In the absence of an analytic
439: model for the effective kinematic viscosity in a turbulent shearing
440: flow, we adopt a simple damping force prescription, i.e., $F_{\phi
441: }^{\mathrm{vis}}=-v_{\phi }/t_{\nu }$, where $t_{\nu }$ is the
442: timescale for viscous accretion. In order for our prescription to
443: yield the order-of-magnitude estimate for the viscous timescale,
444: $t_{\nu }\sim r^{2}/\nu \sim r/\left\vert v_{r0}\right\vert $, we
445: write
446: \begin{equation}
447: \left\langle F_{\phi }^{\mathrm{vis}}\right\rangle =\gamma _{\nu }\frac{%
448: v_{r0}v_{\phi }}{r}, \label{vscf}
449: \end{equation}
450: where $\gamma _{\nu }$ is a dimensionless factor. Note that, for this
451: prescription of the viscous force to satisfy equation (\ref{azeq}) for the
452: steady equilibrium state, it is necessary that
453: \begin{equation}
454: \gamma _{\nu }=\frac{\kappa ^{2}}{2\Omega ^{2}}-\frac{\langle F_{\phi }^{%
455: \mathrm{mag}}\rangle _{0}}{v_{r0}\Omega }. \label{gamnu}
456: \end{equation}
457: To linear order in the perturbed quantities, equation (\ref{vscf}) yields $%
458: \left\langle F_{\phi }\right\rangle _{1}=-\gamma _{\nu }\Omega _{\nu
459: }v_{\phi 1}$. The zeroth-order magnetic force enters the perturbation
460: equations only through the parameter $\gamma _{\nu }$ as given in equation (%
461: \ref{gamnu}).
462:
463: For the case of interest, equations (\ref{fdct})--(\ref{fdaz}) become
464: \begin{equation}
465: \left[ i(m\Omega -\omega )+\beta \Omega _{\nu }\right] \sigma _{m}+
466: (1+\beta) \Omega _{s}u_{r,m}+im\Omega _{s}u_{\phi ,m}=0, \label{hct}
467: \end{equation}
468: \begin{equation}
469: -\beta \Omega _{s}\sigma _{m}+\left[ i(m\Omega -\omega )+(1+\beta) \Omega
470: _{\nu }\right] u_{r,m}-2\Omega u_{\phi ,m}=0, \label{hrm}
471: \end{equation}
472: \begin{equation}
473: im\Omega _{s}\sigma _{m}+\frac{\kappa ^{2}}{2\Omega }u_{r,m}+\left[
474: i(m\Omega -\omega )-(1-\gamma _{\nu })\Omega _{\nu }\right] u_{\phi ,m}=0,
475: \label{azh}
476: \end{equation}
477: yielding a dispersion relation of third order in $\omega $. We provide the
478: general expressions for the dispersion relation and the three complex
479: solutions in the Appendix.
480:
481: For illustrative purposes, we write the expressions for the three
482: frequencies to leading order in $\Omega _{s}/\Omega $, i.e., for
483: small hydrodynamic corrections. The solutions of astrophysical
484: interest may actually be found in situations beyond this approximate
485: regime, as we shall discuss in \S\ 3.2. However, the approximate
486: solutions shed light on the basic parameters that determine the
487: frequencies and growth rates of the modes. In the limit of small
488: hydrodynamic corrections, i.e. $\varepsilon _{1} $, $\varepsilon
489: _{2}$, and $\varepsilon _{3}\ll 1$ (see Appendix), the axisymmetric
490: modes ($m=0$) have the frequencies and growth rates
491: \begin{eqnarray}
492: \mathrm{Re}\left[ \omega _{1,2}^{(0)}\right] &=&\pm \kappa \pm \beta \left(
493: 1+\beta \right) \left( \frac{\Omega _{s}}{2\kappa }\right) \Omega _{s}
494: \notag \\
495: &&\pm \left[ 3\left( \gamma _{\nu }-\beta -1\right) -\left( \gamma _{\nu
496: }-\beta \right) ^{2}\right] \left( \frac{\Omega _{\nu }}{6\kappa }\right)
497: \Omega _{\nu }, \label{ob0r}
498: \end{eqnarray}%
499: \begin{equation}
500: \mathrm{Im}\left[ \omega _{1,2}^{(0)}\right] =-\left( \frac{\gamma _{\nu
501: }+\beta }{2}\right) \Omega _{\nu }, \label{ob0i}
502: \end{equation}%
503: \begin{equation}
504: \mathrm{Re}\left[ \omega _{3}^{(0)}\right] =0, \label{ou0r}
505: \end{equation}%
506: and
507: \begin{equation}
508: \mathrm{Im}\left[ \omega _{3}^{(0)}\right] =-\beta \Omega _{\nu }.
509: \label{ou0i}
510: \end{equation}%
511: The corresponding expressions for the non-axisymmetric modes, with positive
512: or negative integer values of the azimuthal wavenumber $m$ are
513: \begin{eqnarray}
514: \mathrm{Re}\left[ \omega _{1,2}^{(m)}\right] &=&m\Omega \pm \kappa \pm %
515: \left[ m^{2}+\beta \left( 1+\beta \right) \right] \left( \frac{\Omega _{s}}{%
516: 2\kappa }\right) \Omega _{s} \notag \\
517: &&\pm \left[ 3\left( \gamma _{\nu }-\beta -1\right) -\left( \gamma _{\nu
518: }-\beta \right) ^{2}\right] \left( \frac{\Omega _{\nu }}{6\kappa }\right)
519: \Omega _{\nu }, \label{obmr}
520: \end{eqnarray}%
521: \begin{equation}
522: \mathrm{Im}\left[ \omega _{1,2}^{(m)}\right] =-\left( \frac{\gamma _{\nu
523: }+2\beta }{3}\right) \Omega _{\nu }, \label{obmi}
524: \end{equation}%
525: \begin{equation}
526: \mathrm{Re}\left[ \omega _{3}^{(m)}\right] =m\Omega , \label{oumr}
527: \end{equation}%
528: and
529: \begin{equation}
530: \mathrm{Im}\left[ \omega _{3}^{(m)}\right] =-\left( \frac{\gamma _{\nu
531: }+2\beta }{3}\right) \Omega _{\nu }. \label{oumi}
532: \end{equation}%
533: Modes whose real frequencies differ only by a sign are of course identical.
534: We note that in these approximate expressions to leading order in $\Omega
535: _{s}$ the imaginary parts have no dependence on $m$. This means that the
536: fundamental modes and the infinite sequence of sidebands shifted in
537: frequency by $\pm |m|\Omega $ from the fundamental modes of interest, all
538: have the same growth or decay rates in the limit of small hydrodynamic
539: corrections, i.e. $\Omega _{\nu }\ll \Omega $ and $\Omega _{s}\ll \Omega $.
540: Any distinction between the free oscillation modes arises only in a regime
541: in which acoustic and viscous hydrodynamic effects are important.
542:
543: It is possible to check the consistency of our stability analysis
544: using equations (\ref{ob0i}), (\ref{ou0i}), (\ref{obmi}), and
545: (\ref{oumi}) in the limit of non-magnetic local modes for which the
546: radial gradient in the background surface density at any given
547: location is negligible, that is, $d\ln \Sigma _{0}/d\ln r\ll 1$.
548: Note that none of the modes grows if the local
549: gradient of the surface density is neglected $(\beta =0)$ since $%
550: \gamma _{\nu }>0$ for $\langle F_{\phi }^{\mathrm{mag}}\rangle
551: _{0}=0$. There is no instability for local hydrodynamic modes in the
552: long wavelength limit. The global modes grow in the inner disk where
553: the radial gradient of $\Sigma _{0}$ cannot be neglected $(\beta
554: \neq 0).$
555:
556: \subsection{Hydrodynamic Corrections to Test-particle Frequencies\label%
557: {hctpf}}
558:
559: In the ideal case of $\Omega _{\nu }=\Omega _{s}=0$, the three
560: characteristic frequencies coincide with appropriate combinations of
561: test-particle frequencies in the disk, i.e., $\omega
562: _{1}^{(m)}\simeq \kappa +|m|\Omega $, $\omega _{2}^{(m)}\simeq
563: \kappa -|m|\Omega $, and $\omega _{3}^{(m)}\simeq |m|\Omega $. This
564: justifies, to zeroth order, the identification of the observed QPO
565: frequencies with frequencies of test particles in kinematic models
566: of the QPOs. For example, in the sonic-point model (Miller et al.
567: 1998), the higher kHz QPO would be the $\omega _{3}^{(1)}$ mode at
568: the sonic radius. In the relativistic precession model (Stella et
569: al.\ 1999), the two high-frequency QPOs would be the $\omega
570: _{3}^{(1)}$ and $\omega _{2}^{(1)}$ modes.
571:
572: In the absence of external forcing, high-frequency modes in an
573: accretion disk can be excited only in the presence of viscosity.
574: This is a well-known result, discussed in detail by Kato (2001).
575: Because the disk cannot respond thermally at the high-frequencies of
576: interest here, viscosity can excite the modes only through its
577: dynamical effect (mechanism II in Kato 2001) and not through its
578: thermal effect (mechanism I in Kato 2001). Thus free oscillation
579: modes in the accretion disk modes can have positive growth rates only if $%
580: \Omega _{\nu }$ is not zero, i.e., through the effect of viscosity,
581: and, hence, only if the hydrodynamic contributions to their
582: frequencies are operative. Indeed, all models of hydrodynamic
583: accretion disk modes show that non-negligible hydrodynamic
584: corrections modify the test-particle frequencies of growing modes
585: (see, e.g., Wagoner 1999; Kato 2001; Psaltis \& Norman 2000).
586:
587: We estimate the expected hydrodynamic corrections to the oscillatory
588: mode frequencies typically identified with high-frequency QPOs using
589: the scaling of the viscous and thermal frequencies in an alpha disk,
590: i.e.,
591: \begin{equation}
592: \Omega _{\nu }\simeq \alpha \left( \frac{H}{r}\right) ^{2} \left( \frac{%
593: \Omega_{\mathrm{K}}}{\Omega}\right) ^{2} \Omega \label{onuad}
594: \end{equation}
595: and
596: \begin{equation}
597: \Omega _{s}\simeq \left( \frac{H}{r}\right) \left( \frac{\Omega_{\mathrm{K}}%
598: }{\Omega}\right) \Omega. \label{ados}
599: \end{equation}
600: The fractional hydrodynamic correction to, e.g., the $\mathrm{Re}\left[
601: \omega _{2}^{(1)}\right]\simeq \kappa -\Omega $ mode is determined by the
602: acoustic response provided that $\alpha $ is not close to unity:
603: \begin{equation}
604: \frac{\delta \omega _{2}^{(1)}}{\omega _{2}^{(1)}}\simeq \frac{\Omega
605: _{s}^{2}}{\kappa (\kappa -\Omega )}\simeq \left( \frac{H}{r}\right)
606: ^{2}\left( \frac{\kappa }{\Omega }\right) ^{-1} \left( \frac{\Omega_{\mathrm{%
607: K}}}{\Omega}\right) ^{2} \left\vert 1-\frac{\kappa }{\Omega }\right\vert
608: ^{-1}. \label{eq:corr}
609: \end{equation}
610:
611: The growth rate of the same modes is $\simeq \Omega _{\nu }$ and hence the
612: quality factor $Q$ of the corresponding QPO will be at most
613: \begin{equation}
614: Q\simeq \frac{\kappa -\Omega }{\Omega _{\nu }} \simeq \alpha ^{-1}\left(
615: \frac{H}{r}\right) ^{-2} \left( \frac{\Omega_{\mathrm{K}}}{\Omega}\right)
616: ^{-2} \left\vert 1-\frac{\kappa }{\Omega }\right\vert \label{eq:Q}
617: \end{equation}
618: and the resulting width of the QPO will be $\simeq |\Omega -\kappa |/Q
619: \simeq \Omega _{\nu }$. Combining equations~(\ref{eq:corr}) and (\ref{eq:Q})
620: we obtain for the hydrodynamic correction to the mode frequency in units of
621: the width of the corresponding QPO the relation
622: \begin{equation}
623: \frac{\delta \omega _{2}^{(1)}}{\omega _{2}^{(1)}/Q}\simeq \frac{1}{\alpha }
624: \left( \frac{\Omega }{\kappa }\right). \label{hydcrc}
625: \end{equation}
626: It is clear from this expression that, even for relatively large values of
627: the parameter $\alpha \simeq 0.1$, the QPOs will have frequencies that are
628: distinct from the corresponding test-particle frequencies by amounts that
629: are several times larger than the QPO widths.
630:
631: \begin{figure}[h]
632: \epsscale{1.0} \plotone{f1.eps} \caption{The radial profiles of the
633: orbital frequency $\Omega (r)$,
634: epicyclic frequency $\protect\kappa (r)$, Keplerian frequency $\Omega _{%
635: \mathrm{K}}(r)$, and the vertically integrated dynamical viscosity
636: $f(r)$ throughout the inner disk. This example is based on a typical
637: boundary region model from Erkut \& Alpar (2004). The frequencies
638: are given in units of $\Omega _{\mathrm{K}}(r_{\mathrm{in}})$. In
639: this particular example, the actual orbital frequency at the inner
640: disk boundary and the rotation rate of the neutron star are half the
641: Keplerian value at $r_{\mathrm{in}}$; $\Omega
642: (r_{\mathrm{in}})=\Omega _{*}= \Omega
643: _{\mathrm{K}}(r_{\mathrm{in}})/2$. \label{fig1}}
644: \end{figure}
645:
646: \subsection{Excitation of Global Hydrodynamic Free Oscillation Modes\label%
647: {ehdfom}}
648:
649: The growth rates of free oscillation modes are determined by a
650: number of parameters characterizing the boundary region in the inner
651: disk. In this region, the angular velocity of the disk matter makes
652: a transition from the Keplerian rotation to match the stellar
653: rotation rate. Our analysis depends on the local values of these
654: parameters at any particular radius $r$ within this transition or
655: boundary region. The main parameters are the ratio of the
656: radial epicyclic frequency $\kappa $ to the orbital angular frequency $%
657: \Omega $, the radial surface-density profile $\beta $ (equation \ref{bet}), $%
658: \Omega _{\nu }/\Omega _{s}$, the ratio of the radial drift velocity to the
659: sound speed, and $\Omega _{s}/\Omega $, the inverse timescale or the typical
660: frequency associated with the sound speed in units of the angular velocity.
661: In this section, we relate these parameters to the angular velocity profile $%
662: \Omega (r)$ and the dynamical viscosity $\nu \Sigma _{0} $ in the
663: accretion disk. We then study the local excitation of the modes in
664: the inner disk or boundary region where the hydrodynamic effects are
665: important.
666:
667: The ratio $\kappa /\Omega $ of the epicyclic and orbital frequencies is
668: related to the orbital frequency profile $\Omega (r)$ through the relation
669: \begin{equation}
670: \frac{\kappa ^{2}}{\Omega ^{2}}=4\left( 1+\frac{1}{2}\frac{d\ln \Omega }{%
671: d\ln r}\right) . \label{kappaOmega}
672: \end{equation}
673: In a non-Keplerian boundary-transition region of the accretion disk
674: around a neutron star that is a slow rotator, $\Omega(r)$ is less
675: than the Keplerian value $\Omega_{\mathrm{K}}(r)$. Proceeding from
676: the Keplerian outer disk through the transition region, $\Omega(r)$
677: goes through a maximum and then decreases to match the star's
678: rotation rate $\Omega _{\ast }$ at the inner
679: radius of the disk. The epicyclic frequency $\kappa$, degenerate with $%
680: \Omega_{\mathrm{K}}$ in the outer disk, increases through the
681: transition region. The ratio $\kappa /\Omega$ equals 2 at the radius
682: where $\Omega$ is maximum, has values between 1 and 2 in the outer
683: parts of the non-Keplerian transition region, and is larger than 2
684: in the inner parts. Fig.~1 shows the run of $\Omega(r)$ and $\kappa
685: (r)$ in a typical transition region. The numerical data for this
686: example are obtained from the model curve shown in the panel (a) of
687: Fig.~2 in Erkut \& Alpar (2004).
688:
689: For a viscous accretion disk, we write the turbulent viscosity,
690: \begin{equation}
691: \nu =\alpha \frac{c_{s}^{2}}{\Omega }, \label{vscty}
692: \end{equation}
693: using the $\alpha $-prescription (Shakura \& Sunyaev 1973). The
694: angular velocity profile in a disk depends on the radial profile of
695: the vertically integrated dynamical viscosity,
696: \begin{equation}
697: \nu \Sigma _{0} =\frac{\dot{M}}{3\pi }f(r), \label{vidv}
698: \end{equation}
699: and the appropriate boundary conditions. Here, $f(r)$ is a
700: dimensionless function for the dynamical viscosity. The run of $f$
701: as a function of the radial distance in the typical disk-transition
702: region is also shown in Fig.~1 (see Erkut \& Alpar 2004). The radial
703: profile of $f$ is independent of the particular prescription for
704: $\nu $. However, once we specify the kinematic viscosity as in
705: equation (\ref{vscty}), we can relate the surface
706: density profile (see equation \ref{bet}), using equations (\ref{vscty}) and (%
707: \ref{vidv}), to the orbital frequency profile and other parameters of the
708: boundary region through
709: \begin{equation}
710: \beta =\frac{d\ln f}{d\ln r}+\frac{d\ln \Omega }{d\ln r}-2\frac{d\ln c_{s}}{%
711: d\ln r} \equiv \beta _{0}-2\frac{d\ln c_{s}}{d\ln r}. \label{beta}
712: \end{equation}
713: The first two terms, which we delineate as $\beta _{0}$ characterize each of
714: the boundary region models developed in Erkut \& Alpar (2004), with values
715: in the range $-15<\beta _{0}<3$.
716:
717: Using equations (\ref{eqcnt}), (\ref{vscty}), and (\ref{vidv}), we
718: obtain the explicit dependence of the ratio of the radial drift
719: velocity and the sound speed on the disk-boundary region parameters
720: as
721: \begin{equation}
722: \frac{\Omega _{\nu }}{\Omega _{s}}=-\frac{v_{r0}}{c_{s}}=\frac{3}{2} \left(%
723: \frac{\alpha }{f} \right) \frac{\Omega _{s}}{\Omega }. \label{onuovos}
724: \end{equation}
725: Note that this ratio, for the given values of $\alpha $ and $\Omega
726: _{s}/\Omega $, takes its maximum value for the minimum value of $f$. As $%
727: \Omega _{\nu }$, which determines the growth rates of the modes, depends on
728: the ratio $\Omega _{s}/\Omega $, the growth rates differ for sufficiently
729: large hydrodynamic corrections, i.e., $\Omega _{s}\lesssim \Omega $, and the
730: fastest growing modes can be identified. This is the regime of only slightly
731: supersonic azimuthal flow. In a magnetic boundary layer or transition zone
732: where $\Omega <\Omega _{\mathrm{K}}$, the speed of sound can be written as $%
733: c_{s}\simeq \Omega _{\mathrm{K}}\sqrt{Hr}$ whereas in the weakly
734: magnetized outer disk, $c_{s}\simeq \Omega _{\mathrm{K}}H$ (see
735: Erkut \& Alpar 2004). We therefore expect to find larger values for
736: $\Omega _{s}/ \Omega $ in a
737: non-Keplerian boundary region; $\Omega _{s}/ \Omega \simeq (\Omega _{\mathrm{%
738: K}}/ \Omega)(H/r)^{1/2}$. The larger the ratio $\Omega _{s}/\Omega $, the
739: stronger is the effect of hydrodynamic corrections on the growth rates of
740: the modes.
741:
742: We explore the run of the frequencies and the growth rates of free
743: oscillations excited at a characteristic radius $r$ in the innermost
744: disk as the hydrodynamic parameter $\Omega _{s}/\Omega $ changes at
745: that radius. As a guideline, we have employed the model solutions of
746: Erkut \& Alpar (2004). These models qualitatively represent
747: conditions in a boundary-transition region. The key model parameters
748: can be easily translated into the
749: parameters in equations (\ref{kappaOmega}), (\ref{beta}), and (\ref{onuovos}%
750: ). The zeroth-order magnetic force in equation (\ref{gamnu}) can also be
751: estimated from these model solutions with a range $0\leq \langle F_{\phi }^{%
752: \mathrm{mag}}\rangle _{0}/v_{r0}\Omega \lesssim 1$ in the inner boundary
753: region. The maximum value of this parameter corresponds to the radius where $%
754: \Omega $ is maximum. Near the innermost radius of the disk, the
755: azimuthal magnetic force vanishes as the relative shear between the
756: magnetosphere and the disk becomes negligible, i.e., $\Omega -\Omega
757: _{*}\simeq 0$.
758:
759: To elucidate how the hydrodynamic parameters give different growth rates for
760: different modes, we investigate the excitation of free oscillations in the
761: typical boundary region shown in Fig.~1. To see the hydrodynamic effects on
762: the growth rates of the modes, we use the full eigenfrequency solutions for
763: axisymmetric and non-axisymmetric perturbations given in equations (\ref%
764: {afeo})--(\ref{thef}) and (\ref{fof})--(\ref{ftf}) (see Appendix). The panel
765: (a) of Fig.~2 shows the real parts of the complex mode frequencies at the
766: radius where $\Omega $ is maximum for the $m=0$ and $m=1$ cases. For the
767: boundary layer model chosen, the model parameters at this radius are $\kappa
768: /\Omega =2$, $\beta =-14$, $f=0.067$, and $\langle F_{\phi }^{\mathrm{mag}%
769: }\rangle _{0}/v_{r0}\Omega =1$. The panel (a) of Fig.~2 is obtained for $%
770: \alpha =0.1$. Although the real parts of the mode frequencies do not change
771: under a different value of the viscosity parameter, say $\alpha =0.01$ at
772: this radius, the imaginary parts do (see panel b of Fig.~2). The unstable
773: modes grow faster for $\alpha =0.1$ than they do for $\alpha =0.01$, as
774: shown in Fig.~2 (panel b). This is because the growth rates are primarily
775: determined by $\Omega _{\nu }/\Omega _{s}\propto \alpha /f$ (see equation %
776: \ref{onuovos}). In Fig.~3, we display the real (panel a) and imaginary
777: (panel b) parts of the mode frequencies at the innermost disk radius where $%
778: \Omega =\Omega _{*}$ for $m=0$ and $m=1$. The model parameters illustrated
779: in Fig.~3 are $\kappa /\Omega =2.5$, $\beta =-5$, $f=1$, and $\langle
780: F_{\phi }^{\mathrm{mag}}\rangle _{0}/v_{r0}\Omega =0$. Fig.~3 is plotted for
781: $\alpha =0.1$. Note that the growth rates, for the same value of the
782: viscosity parameter $\alpha $, are lower than those at the radius where $%
783: \Omega $ is maximum.
784:
785: The analysis for the particular transition zone discussed above can also be
786: extended to other sample boundary regions with different model parameters.
787: Our conclusion that the fastest growing modes are excited near the radius $%
788: r=r_{0}$ where $\Omega $ is maximum remains valid for all
789: sub-Keplerian inner disks around slow rotators: The common property
790: of all boundary regions is that the parameters $|\beta |$ and
791: $\alpha /f$, which determine the growth rates of the unstable modes,
792: attain highest values near $r_{0}$ where $\kappa /\Omega \simeq 2$.
793: Note that the growing mode frequencies
794: significantly exceed test-particle frequencies in the boundary region for $%
795: \Omega _{s}/\Omega \lesssim 1$. In the limit of small hydrodynamic
796: corrections ($\Omega _{s}/\Omega \ll 1$), the mode frequencies converge to
797: the corresponding test-particle frequencies (see panel a of Fig.~2 and panel
798: a of Fig.~3) with negligible growth rates as expected (see panel b of Fig.~2
799: and panel b of Fig.~3).
800:
801: The frequency branches that map to the test-particle frequencies $\omega
802: _{1}^{(0)} =\kappa $ and $\omega _{1,2}^{(1)} =\kappa \pm \Omega $ in the
803: limit of small hydrodynamic corrections, $\Omega _{s}/\Omega \ll 1$, have
804: positive and rather fast growth rates at $r\lesssim r_{0}$, where $\kappa
805: /\Omega \gtrsim 2$, when $\Omega _{s}/\Omega \lesssim 1$ (see panel a of
806: Fig.~2 and panel a of Fig.~3). We identify $\omega _{3}^{(0)} =0$ and $%
807: \omega _{3}^{(1)} \simeq \Omega $ as the quasi-stable or decaying mode
808: frequencies in the same range of $\Omega _{s}/\Omega $ as shown in Fig.~2
809: (panel b) and Fig.~3 (panel b). Note that while each of the growing branches
810: differs from test-particle frequencies as $\Omega _{s}/\Omega $ increases,
811: the difference frequency between consecutive bands is always $\Delta \omega
812: = \omega _{1}^{(1)}-\omega _{1}^{(0)}\lesssim \Omega $ or $\Delta \omega =
813: \omega _{1}^{(0)}-\omega _{2}^{(1)}\lesssim \Omega $ over a wide range of $%
814: \Omega _{s}/\Omega $ (see panel a of Fig.~2 and panel a of Fig.~3). As $%
815: \Omega \gtrsim \Omega _{*}$ in a boundary layer, we find $\Delta \omega
816: \simeq \Omega _{*}$ as observed for the relatively slow rotators among the
817: neutron-star LMXBs that exhibit kHz QPOs.
818:
819: \section{Discussion and Conclusions}
820:
821: We have studied the global hydrodynamic modes of long wavelength in
822: the boundary or transition region of viscous accretion disks as a
823: possible source of kHz QPOs in neutron-star LMXBs. The stability of
824: the eigenmodes strongly depends on the global disk structure imposed
825: by the boundary conditions as expected from the mode analysis in the
826: long wavelength regime. Our local treatment takes account of the
827: local effects of the global disk parameters on the excitation of
828: hydrodynamic free oscillations.
829:
830: We find that the frequencies and growth rates of the modes are
831: mainly determined by global disk parameters such as $\kappa /\Omega
832: $, the ratio of the radial epicyclic frequency to the orbital
833: frequency, the radial surface-density profile $\beta $, $\Omega
834: _{\nu }/\Omega _{s}$, the ratio of the radial drift velocity to the
835: sound speed, and $\Omega _{s}/\Omega $, the
836: ratio of the sound speed to the azimuthal velocity. The local values of $%
837: \kappa /\Omega $ and $\beta $ directly follow from the global
838: solution for the rotational dynamics of a boundary-transition region
839: model. The parameters $\Omega _{\nu }/\Omega _{s}$ and $\Omega
840: _{s}/\Omega $ depend additionally on the particular viscosity
841: prescription and the details of the structure of accretion flow in
842: the inner disk.
843:
844: The hydrodynamic effects on the frequencies and growth rates of the
845: modes are reflected by the parameters $\Omega _{\nu }/\Omega _{s}$
846: and $\Omega _{s}/\Omega $. We have found that the growth rates of
847: different modes for a given azimuthal wavenumber $m$ differ
848: significantly when the hydrodynamic corrections are sufficiently
849: large, i.e. $\Omega _{\nu}/\Omega _{s}\lesssim 1 $ and $\Omega
850: _{s}/\Omega \lesssim 1$. In this regime, the growing mode
851: frequencies significantly exceed test-particle frequencies for a
852: plausible range of $\Omega _{s}/\Omega $ at any particular radius
853: within the boundary region. This is because the effect of
854: hydrodynamic corrections on the growth rates of the modes is
855: expected to be strong in a magnetic boundary layer or transition
856: zone as discussed in \S\ 3.2. In the limit of small hydrodynamic
857: corrections, i.e. $\Omega _{\nu}/\Omega _{s}\ll 1$ and $\Omega
858: _{s}/\Omega \ll 1$, the eigenmodes have test-particle frequencies
859: with negligible growth rates. Our analysis shows that taking account
860: of hydrodynamic effects has an important outcome: the frequencies of
861: the growing modes gain higher values above test-particle frequencies
862: as their growth rates increase for larger hydrodynamic corrections.
863:
864: Throughout the boundary region, we have identified the growing mode
865: frequencies by the eigenvalues that approach the test-particle frequencies $%
866: \omega _{1}^{(0)} =\kappa $, $\omega _{1}^{(1)} =\kappa +\Omega $, and $%
867: \omega _{2}^{(1)} =\kappa -\Omega $ in the limit of small hydrodynamic
868: corrections (see panel a of Fig.~2). Modes such as $\omega _{3}^{(0)} =0$
869: and $\omega _{3}^{(1)}\simeq \Omega $ are not excited for the range of $%
870: \Omega _{s}/\Omega $ used in Fig.~2 (panel b) and Fig.~3 (panel b).
871: The difference frequency between successive bands of growing modes
872: is $\Delta \omega \simeq \Omega _{*}$ as observed for the relatively
873: slowly rotating neutron stars in LMXBs that show kHz QPOs (\S\ 3.2).
874: In a boundary layer or transition region, the orbital frequency of
875: the disk matter is close to the rotation frequency of the star
876: ($\Omega \sim \Omega _{*}$). As $\Delta \omega \simeq \Omega$, the
877: separation between consecutive mode frequencies is always related to
878: the spin frequency of the neutron star. This is the result of the
879: boundary condition imposed by the rotating magnetosphere on the
880: innermost disk. The modes of the test-particle frequencies $\omega
881: _{1}^{(0)}\cong \kappa $ and $\omega _{2}^{(1)}\cong \kappa -\Omega
882: $ were originally proposed by Alpar \& Psaltis (2005) to be
883: associated with the upper and lower kHz QPOs observed in
884: neutron-star LMXBs. The frequency separation between the higher and
885: lower-frequency kHz QPO peaks decreases by a few tens of Hz as both
886: QPO frequencies increase by hundreds of Hz in a range of $200-1200$
887: Hz per source in all observed sources, namely, Sco~X-1,
888: 4U~$1608-52$, 4U~$1728-34$ and 4U~$1735-44$ (van der Klis 2000).
889: This pair of frequencies satisfies the observed behavior that the
890: difference frequency $\Delta \omega $ between two kHz QPOs decreases
891: as both frequencies increase. According to our present analysis, the
892: pairs of observed kHz QPO bands could be either $\omega
893: _{1}^{(1)}\simeq \kappa +\Omega $ and $\omega _{1}^{(0)}\simeq
894: \kappa $ or $\omega _{1}^{(0)}\simeq \kappa $ and $\omega
895: _{2}^{(1)}\simeq \kappa -\Omega $, respectively, since the
896: identification of the observed QPOs with $\omega _{1}^{(1)}\simeq
897: \kappa +\Omega $ and $\omega _{1}^{(0)}\simeq \kappa $ also meets
898: the observed condition that $\Delta \omega $ decreases when both QPO
899: frequencies increase. The growth rates of free oscillations produce
900: a wide spectrum of sidebands, $\omega _{1,2}^{(m)}\cong \kappa \pm
901: \vert m\vert \Omega $, some of which fall in the range of observed
902: power spectra. The reduction to only two of kHz QPO bands as
903: observed in neutron star sources must therefore be a product of
904: forced oscillations, resonances, and boundary conditions. The role
905: of forced oscillations and resonances in the excitation of
906: particular modes will be discussed in future work.
907:
908: We find that the fastest growing modes with frequency branches
909: $\kappa $ and $\kappa \pm \Omega $ are excited near the disk radius
910: $r\lesssim r_{0}$, where the orbital frequency is maximum. In the
911: boundary layer, the loss of centrifugal support near $r_{0}$ leads
912: to some radial acceleration of the disk matter. The subsequent rise
913: in the radial drift velocity of the disk matter is accompanied by a
914: sudden drop in the surface density as the vertically integrated
915: dynamical viscosity is minimized (see \S\ 3.2). This picture is
916: common to all boundary layers or transition zones for which magnetic
917: braking is efficient (see Erkut \& Alpar 2004 and references
918: therein). Thus, independent of the boundary conditions and the
919: magnetic field configuration in the boundary region, there exists a
920: specific disk radius where the hydrodynamical background quantities
921: such as the surface density $\Sigma _{0}$ and the radial drift
922: velocity $v_{r0}$ change dramatically. The steepness in the change
923: of the surface density in the
924: radial direction is reflected through the parameter $\beta $ (see equation %
925: \ref{bet}). The sign and magnitude of $\beta $ (the surface density
926: profile in the disk) control whether the free oscillation modes will
927: grow or decay. We see that the modes of frequency bands $\kappa $
928: and $\kappa \pm \Omega $
929: can be excited with the largest growth rates only for the disk radii $%
930: r\lesssim r_{0}$, where $\beta $ is minimum and $\Omega $ is maximum
931: (see \S\ 3.2 for the range of $\beta $). The inverse timescale,
932: $\Omega _{\nu} $, associated with radial accretion also obtains its
933: highest value near the same radii yielding the maximum growth rates
934: in the accretion disk.
935:
936: The boundary layers with sub-Keplerian rotation rates are usually
937: expected to be realized in the innermost regions of accretion disks
938: around \emph{slow rotators}. For a \emph{slow rotator}, the Kepler
939: rotation rate at the inner disk radius $\Omega
940: _{\mathrm{K}}(r_{\mathrm{in}})$ prevails over the stellar rotation
941: rate $\Omega _{*}$. As we have illustrated in \S\ 3.2, the
942: difference frequency between two consecutive bands of growing modes
943: in the boundary region is close to the spin frequency of the neutron star ($%
944: \Delta \omega \simeq \Omega _{*}$). This result agrees with the kHz QPO
945: observations of the relatively slow rotators with spin frequencies below $%
946: \sim 400$~Hz in neutron-star LMXBs. In these sources, for example,
947: in 4U~$1728-34$, the lower kHz QPO frequencies increase from 600 to
948: 900 Hz and the upper kHz QPO frequencies increase from 950 to 1200
949: Hz, while the difference between the two observed kHz QPO frequency
950: bands decreases slightly, of the order of 10 Hz, while remaining
951: still commensurate with the spin frequency of the neutron star
952: (M\'{e}ndez \& van der Klis 1999). A second class of kHz QPO
953: sources, such as the sources KS~$1731-260$, Aql~X-1, 4U~$1636-53$,
954: which have spin frequencies above $\sim 400$~Hz is usually depicted
955: as \emph{fast rotators} (see Wijnands et al. 2003). In this second
956: class of sources, for example in 4U~$1636-53$, the lower kHz QPO
957: frequencies increase from 900 to 950 Hz and the upper kHz QPO
958: frequencies increase from 1150 to 1190 Hz, while the difference
959: between the two observed kHz QPO frequency bands again decreases by
960: amounts of the order of 10 Hz (see van der Klis 2000 and references
961: therein). The frequency separation between the kHz QPOs is close to
962: half the spin frequency of the neutron star if it is a \emph{fast
963: rotator}. The emergence of two seemingly different classes of
964: neutron-star LMXBs could be related to whether the rotation rate of
965: the neutron star is less or greater than that of the inner disk
966: matter. It is probable that $\Omega _{*}>\Omega
967: _{\mathrm{K}}(r_{\mathrm{in}})$ for the relatively fast rotating
968: neutron star sources for which the difference frequency between twin
969: QPO peaks is $\Delta \omega \simeq \Omega _{*}/2$. Recent work by
970: M\'{e}ndez \& Belloni (2007) actually suggests a more
971: continuous distribution between $\Delta \omega \simeq \Omega _{*}$ and $%
972: \Delta \omega \simeq \Omega _{*}/2$. The rotational dynamics of
973: accretion flow in the innermost regions of accretion disks around
974: these \emph{fast rotators} could be quite different from those of
975: the transition regions we consider here. We plan to address this
976: issue in a subsequent paper.
977:
978: \acknowledgments
979: We acknowledge support from the Marie Curie FP6 Transfer of Knowledge
980: Project ASTRONS, MKTD-CT-2006-042722. M.\,H.\,E.\ was partially supported by
981: a T\"{U}B\.{I}TAK (The Scientific and Technical Research Council of Turkey)
982: postdoctoral fellowship. D.\,P.\ was partially supported by NASA grant
983: NAG-513374. M.\,A.\,A.\ acknowledges support from the Turkish Academy of
984: Sciences and the Sabanc{\i} University Astrophysics and Space Forum.
985:
986: \appendix
987:
988: \section{General Expressions}
989:
990: The dispersion relation of the global force-free hydrodynamic modes
991: we consider in \S\ 3 is
992: \begin{equation}
993: (\omega -m\Omega )\left[(\omega -m\Omega )^{2}-\Omega _{a}^{2}\right]-\Omega
994: _{b}^{3} -i\left[ (\omega -m\Omega )^{2}\Omega _{c}+\Omega _{d}^{3}\right]
995: =0, \label{dr}
996: \end{equation}
997: where
998: \begin{equation}
999: \Omega _{a}^{2} \equiv \kappa ^{2}+\left[ m^{2}+\beta (1+\beta )\right]%
1000: \Omega_{s}^{2} +\left[ \beta (1+\beta )-(1+2\beta )(1-\gamma _{\nu })\right]%
1001: \Omega _{\nu }^{2}, \label{oma}
1002: \end{equation}
1003: \begin{equation}
1004: \Omega _{b}^{3}\equiv 2m(1+\beta )\Omega _{s}^{2}\Omega +m\beta \Omega
1005: _{s}^{2}\left( \frac{\kappa ^{2}}{2\Omega }\right) , \label{omb}
1006: \end{equation}
1007: \begin{equation}
1008: \Omega _{c}\equiv -(\gamma _{\nu }+2\beta )\Omega _{\nu }, \label{omc}
1009: \end{equation}
1010: and
1011: \begin{equation}
1012: \Omega _{d}^{3} \equiv \beta \Omega _{\nu }\kappa ^{2}-\beta (1+\beta
1013: )(1-\gamma _{\nu })\Omega _{\nu }^{3} -(1+\beta )\left[ \beta (1-\gamma
1014: _{\nu })-m^{2}\right] \Omega _{\nu }\Omega _{s}^{2}\;. \label{omd}
1015: \end{equation}
1016:
1017: Note that the solution of equation (\ref{dr}), in the ideal case of
1018: an inviscid $\Omega _{\nu }=0$, cold disk $\Omega _{s}\simeq 0$,
1019: where the gas particles barely interact with each other, is given by
1020: $\omega
1021: _{1,2}^{(m)}\simeq m\Omega \pm \kappa $ and $\omega_{3}^{(m)}\simeq m\Omega $%
1022: . In general, equation (\ref{dr}), for the realistic case of $\Omega _{\nu
1023: }\neq 0$ and $\Omega _{s}\neq 0$, yields
1024: \begin{equation}
1025: \omega _{\mathrm{I}}^{(0)}=(\chi -\psi )\Omega _{d}+i\frac{\Omega _{c}}{3},
1026: \label{afeo}
1027: \end{equation}
1028: \begin{equation}
1029: \omega _{\mathrm{II}}^{(0)}=\frac{1}{2}(\psi -\chi )\Omega _{d}+i\frac{%
1030: \Omega _{c}}{3}+i\frac{\sqrt{3}}{2}(\chi +\psi )\Omega _{d}, \label{ttef}
1031: \end{equation}
1032: and
1033: \begin{equation}
1034: \omega _{\mathrm{III}}^{(0)}=\frac{1}{2}(\psi -\chi )\Omega _{d}+i\frac{%
1035: \Omega _{c}}{3}-i\frac{\sqrt{3}}{2}(\chi +\psi )\Omega _{d} \label{thef}
1036: \end{equation}
1037: as the eigenfrequency solutions for axisymmetric perturbations ($m=0$),
1038: where
1039: \begin{equation}
1040: \psi \equiv \frac{1}{9\chi }\left( \frac{\Omega _{c}}{\Omega _{d}}\right)
1041: ^{2}-\frac{1}{3\chi }\left( \frac{\Omega _{a}}{\Omega _{d}}\right) ^{2},
1042: \label{cgmm}
1043: \end{equation}
1044: \begin{equation}
1045: \chi \equiv \left( i\frac{\eta _{1}}{4}+\frac{1}{2}\sqrt{\eta _{2}-\eta _{1}}%
1046: \right) ^{1/3}, \label{chi}
1047: \end{equation}
1048: \begin{equation}
1049: \eta _{1}\equiv 2+\frac{2}{3}\left( \frac{\Omega _{a}^{2}\Omega _{c}}{\Omega
1050: _{d}^{3}}\right) -\frac{4}{27}\left( \frac{\Omega _{c}}{\Omega _{d}}\right)
1051: ^{3}, \label{feta}
1052: \end{equation}
1053: and
1054: \begin{equation}
1055: \eta _{2}\equiv 1-\frac{4}{27}\left( \frac{\Omega _{a}}{\Omega _{d}}\right)
1056: ^{6}+\frac{1}{27}\left( \frac{\Omega _{a}^{4}\Omega _{c}^{2}}{\Omega
1057: _{d}^{6} }\right). \label{seta}
1058: \end{equation}
1059:
1060: For small hydrodynamic corrections, $\eta _{2}<\eta _{1}$, we obtain
1061: \begin{equation}
1062: \mathrm{Re}\left[ \omega _{1}^{(0)}\right] = \frac{\left( \sqrt{%
1063: 1+\varepsilon _{1}}+\varepsilon _{2}\right) ^{1/3}\Omega _{a}}{2} +\frac{%
1064: \left(1-\varepsilon_{3}\right)\Omega _{a}}{2\left( \sqrt{1+\varepsilon _{1}}
1065: +\varepsilon _{2}\right) ^{1/3}}, \label{aef}
1066: \end{equation}
1067: \begin{equation}
1068: \mathrm{Im}\left[ \omega _{1}^{(0)}\right] =\frac{\left( \sqrt{1+\varepsilon
1069: _{1}}+\varepsilon _{2}\right) ^{1/3}\Omega _{a}}{2\sqrt{3}} -\frac{\left(
1070: 1-\varepsilon _{3}\right) \Omega _{a}}{2\sqrt{3}\left( \sqrt{1+\varepsilon
1071: _{1}}+\varepsilon _{2}\right) ^{1/3}}+\frac{\Omega _{c}}{3}, \label{iomo}
1072: \end{equation}
1073: \begin{equation}
1074: \mathrm{Re}\left[ \omega _{2}^{(0)}\right] =-\,\mathrm{Re}\left[ \omega
1075: _{1}^{(0)}\right] , \label{rotmz}
1076: \end{equation}
1077: \begin{equation}
1078: \mathrm{Im}\left[ \omega _{2}^{(0)}\right] =\mathrm{Im}\left[ \omega
1079: _{1}^{(0)}\right] , \label{iotmz}
1080: \end{equation}
1081: \begin{equation}
1082: \mathrm{Re}\left[ \omega _{3}^{(0)}\right] =0, \label{otrmz}
1083: \end{equation}
1084: \begin{equation}
1085: \mathrm{Im}\left[ \omega _{3}^{(0)}\right] =\Omega _{c}-2\,\mathrm{Im}\left[
1086: \omega _{1}^{(0)}\right] \label{otimz}
1087: \end{equation}
1088: as the only non-degenerate eigenfrequency solutions to the dispersion
1089: relation for axisymmetric perturbations, where
1090: \begin{equation}
1091: \varepsilon _{1}\equiv \frac{27}{4}\left( \frac{\Omega _{d}}{\Omega _{a}}%
1092: \right) ^{6}+\frac{9\Omega _{c}\Omega _{d}^{3}}{2\Omega _{a}^{4}}-\frac{%
1093: \Omega _{d}^{3}\Omega _{c}^{3}}{\Omega _{a}^{6}}-\frac{1}{4}\left( \frac{%
1094: \Omega _{c}}{\Omega _{a}}\right) ^{2}, \label{epso}
1095: \end{equation}
1096: \begin{equation}
1097: \varepsilon _{2}\equiv \frac{3\sqrt{3}}{2}\left( \frac{\Omega _{d}}{\Omega
1098: _{a}}\right) ^{3}+\frac{\sqrt{3}\Omega _{c}}{2\Omega _{a}}-\frac{\sqrt{3}}{9}%
1099: \left( \frac{\Omega _{c}}{\Omega _{a}}\right) ^{3}, \label{epst}
1100: \end{equation}
1101: and
1102: \begin{equation}
1103: \varepsilon _{3}\equiv \frac{1}{3}\left( \frac{\Omega _{c}}{\Omega _{a}}%
1104: \right) ^{2}. \label{epsh}
1105: \end{equation}
1106:
1107: The general solutions to the dispersion equation for $m\neq 0$ are
1108: \begin{equation}
1109: \omega _{\mathrm{I}}^{(m)}=m\Omega +(\xi +\Lambda )\Omega _{b}+i\frac{\Omega
1110: _{c}}{3}, \label{fof}
1111: \end{equation}
1112: \begin{equation}
1113: \omega _{\mathrm{II}}^{(m)}=m\Omega -\frac{1}{2}(\xi +\Lambda )\Omega _{b}+i%
1114: \frac{\Omega _{c}}{3}+i\frac{\sqrt{3}}{2}(\xi -\Lambda )\Omega _{b},
1115: \label{ff}
1116: \end{equation}
1117: and
1118: \begin{equation}
1119: \omega _{\mathrm{III}}^{(m)}=m\Omega -\frac{1}{2}(\xi +\Lambda )\Omega _{b}+i%
1120: \frac{\Omega _{c}}{3}-i\frac{\sqrt{3}}{2}(\xi -\Lambda )\Omega _{b},
1121: \label{ftf}
1122: \end{equation}
1123: where
1124: \begin{equation}
1125: \Lambda \equiv \frac{1}{3\xi }\left( \frac{\Omega _{a}}{\Omega _{b}}\right)
1126: ^{2}-\frac{1}{9\xi }\left( \frac{\Omega _{c}}{\Omega _{b}}\right) ^{2},
1127: \label{lmbd}
1128: \end{equation}
1129: \begin{equation}
1130: \xi \equiv \left( \frac{1}{2}+i\frac{\zeta _{1}}{4}+\frac{1}{2}\sqrt{\zeta
1131: _{2}+i\zeta _{1}}\right) ^{1/3}\, \label{ksi}
1132: \end{equation}
1133: \begin{equation}
1134: \zeta _{1}\equiv \frac{2}{3}\left( \frac{\Omega _{a}^{2}\Omega _{c}}{\Omega
1135: _{b}^{3}}\right) +2\left( \frac{\Omega _{d}}{\Omega _{b}}\right) ^{3}-\frac{4%
1136: }{27}\left( \frac{\Omega _{c}}{\Omega _{b}}\right) ^{3}, \label{zto}
1137: \end{equation}
1138: and
1139: \begin{equation}
1140: \zeta _{2}\equiv 1-\frac{4}{27}\left( \frac{\Omega _{a}}{\Omega _{b}}\right)
1141: ^{6}+\frac{1}{27}\left( \frac{\Omega _{a}^{4}\Omega _{c}^{2}}{\Omega _{b}^{6}%
1142: }\right) +\left( \frac{\Omega _{d}}{\Omega _{b}}\right) ^{6}-\zeta
1143: _{1}\left( \frac{\Omega _{d}}{\Omega _{b}}\right) ^{3}. \label{ztt}
1144: \end{equation}
1145:
1146: In the limit of small hydrodynamic corrections, $|\zeta _{1}|\,\ll |\zeta
1147: _{2}|$, we find
1148: \begin{equation}
1149: \mathrm{Re}\left[ \omega _{1}^{(m)}\right] =m\Omega \pm \frac{\Omega
1150: _{b}\Omega _{a}}{\left\vert \Omega _{b}\right\vert }\mp \left( \frac{\Omega
1151: _{c}}{\Omega _{a}}\right) ^{2}\frac{\left\vert \Omega _{b}\right\vert \Omega
1152: _{a}}{6\Omega _{b}}, \label{naor}
1153: \end{equation}
1154: \begin{equation}
1155: \mathrm{Im}\left[ \omega _{1}^{(m)}\right] =\frac{\Omega _{c}}{3}\left[ 1\pm
1156: \left( \frac{\Omega _{c}}{\Omega _{a}}\right) \frac{\left\vert \Omega
1157: _{b}\right\vert }{2\sqrt{3}\Omega _{b}}\right] , \label{naoi}
1158: \end{equation}
1159: \begin{equation}
1160: \mathrm{Re}\left[ \omega _{2}^{(m)}\right] =m\Omega \mp \frac{\Omega
1161: _{b}\Omega _{a}}{\left\vert \Omega _{b}\right\vert }\pm \left( \frac{\Omega
1162: _{c}}{\Omega _{a}}\right) ^{2}\frac{\left\vert \Omega _{b}\right\vert \Omega
1163: _{a}}{6\Omega _{b}}, \label{natr}
1164: \end{equation}
1165: \begin{equation}
1166: \mathrm{Im}\left[ \omega _{2}^{(m)}\right] =\mathrm{Im}\left[ \omega
1167: _{1}^{(m)}\right] , \label{nati}
1168: \end{equation}
1169: \begin{equation}
1170: \mathrm{Re}\left[ \omega _{3}^{(m)}\right] =m\Omega , \label{rnat}
1171: \end{equation}
1172: \begin{equation}
1173: \mathrm{Im}\left[ \omega _{3}^{(m)}\right] =\Omega _{c}-2\,\mathrm{Im}\left[
1174: \omega _{1}^{(m)}\right] \label{inat}
1175: \end{equation}
1176: as the only non-degenerate eigenfrequency solutions to the dispersion
1177: relation for non-axisymmetric perturbations.
1178:
1179: \begin{thebibliography}{99}
1180: \bibitem[]{} Abramowicz, M. A., Karas, V., Klu{\'z}niak, W., Lee, W. H., \& Rebusco, P.
1181: 2003, \pasj, 55, 467
1182:
1183: \bibitem[]{} Alpar, M. A., Hasinger, G., Shaham, J., \& Yancopoulos, S. 1992, \aap,
1184: 257, 627
1185:
1186: \bibitem[]{} Alpar, M. A., \& Psaltis, D. 2005, preprint (astro-ph/0511412)
1187:
1188: \bibitem[]{} Alpar, M. A., \& Y\i lmaz, A. 1997, NewA, 2, 225
1189:
1190: \bibitem[]{} Erkut, M. H., \& Alpar, M. A. 2004, \apj, 617, 461
1191:
1192: \bibitem[]{} Kato, S. 2001, \pasj, 53, 1
1193:
1194: \bibitem[]{} M\'{e}ndez, M., \& Belloni, T. 2007, \mnras, 381, 790
1195:
1196: \bibitem[]{} M\'{e}ndez, M., \& van der Klis, M. 1999, \apj, 517, L51
1197:
1198: \bibitem[]{} Miller, M. C., Lamb, F. K., \& Psaltis, D. 1998, \apj, 508, 791
1199:
1200: \bibitem[]{} Psaltis, D., Belloni, T., \& van der Klis, M. 1999, \apj, 520, 262
1201:
1202: \bibitem[]{} Psaltis, D., \& Norman, C. 2000, preprint (astro-ph/0001391)
1203:
1204: \bibitem[]{} Shakura, N. I., \& Sunyaev, R. A. 1973, \aap, 24, 337
1205:
1206: \bibitem[]{} Stehle, R., \& Spruit, H. C. 1999, \mnras, 304, 674
1207:
1208: \bibitem[]{} Stella, L., Vietri, M., \& Morsink, S. M. 1999, \apj, 524, L63
1209:
1210: \bibitem[]{} van der Klis, M. 2000, \araa, 38, 717
1211:
1212: \bibitem[]{} Wagoner, R. W. 1999, \physrep, 311, 259
1213:
1214: \bibitem[]{} Wijnands, R., van der Klis, M., Homan, J., Chakrabarty, D.,
1215: Markwardt, C. B., \& Morgan, E. H. 2003, \nat, 424, 44
1216: \end{thebibliography}
1217:
1218: %\clearpage
1219:
1220:
1221:
1222: \clearpage
1223:
1224: \begin{figure}[tbp]
1225: \epsscale{1.0} \plottwo{f2a.eps}{f2b.eps} \caption{The real and
1226: imaginary parts of the solutions to the dispersion relation for
1227: axisymmetric ($m=0$) and non-axisymmetric ($m=1$) perturbations
1228: excited at the radius where the orbital frequency $\Omega $ is
1229: maximum. The
1230: model parameters are $\protect\kappa /\Omega =2$, $\protect\beta =-14$, $%
1231: f=0.067$, and $\langle F_{\protect\phi }^{\mathrm{mag}}\rangle
1232: _{0}/v_{r0}\Omega =1$. The curves are labelled with the values of
1233: the
1234: frequencies in the test-particle limit. The $\protect\kappa $ and $\protect%
1235: \kappa \pm \Omega $ modes have the same positive growth rates,
1236: growing more rapidly for larger $\protect\alpha $. The
1237: zero-frequency and $\Omega $
1238: modes, indicated by the overlapping dotted curves, do not grow for any $%
1239: \protect\alpha $. \label{fig2}}
1240: \end{figure}
1241:
1242: %\clearpage
1243:
1244: \begin{figure}[tbp]
1245: \epsscale{1.0} \plottwo{f3a.eps}{f3b.eps} \caption{The real and
1246: imaginary parts of the solutions to the dispersion relation for
1247: axisymmetric ($m=0$) and non-axisymmetric ($m=1$) perturbations
1248: excited at the innermost disk radius where $\Omega =\Omega _{*}$.
1249: The model parameters are $\protect\kappa /\Omega =2.5$,
1250: $\protect\beta =-5$, $f=1$, and $\langle F_{\protect\phi
1251: }^{\mathrm{mag}}\rangle _{0}/v_{r0}\Omega =0$. The frequencies and
1252: growth rates are obtained for $\protect\alpha =0.1$. The
1253: growing modes are associated with the $\protect\kappa +\Omega $, $\protect%
1254: \kappa $, and $\protect\kappa -\Omega $ branches, with $\protect\kappa %
1255: +\Omega $ having the highest growth rate. The zero-frequency and
1256: $\Omega $ modes grow for lower values of $\Omega _{s}/\Omega $, but
1257: decay in the
1258: regime of $\Omega _{s}/\Omega \lesssim 1$, where the $\protect\kappa $ and $%
1259: \protect\kappa \pm \Omega $ modes have the fastest growth rates.
1260: \label{fig3}}
1261: \end{figure}
1262:
1263: \end{document}
1264: