1: %\documentclass[manuscript]{aastex}
2: %\documentclass[preprint,flushrt]{aastex}
3: \documentclass{article}\usepackage{natbib,emulateapj}
4: \newcommand{\beq}{\begin{equation}}
5: \newcommand{\eeq}{\end{equation}}
6: \bibpunct[,]{(}{)}{;}{a}{}{,}
7: \begin{document}
8: \title{Particle Heating by Nonlinear Alfv\'enic Turbulence in ADAFs}
9: \author{Mikhail V. Medvedev\altaffilmark{1} }
10: \affil{Harvard-Smithsonian Center for Astrophysics,
11: 60 Garden St., Cambridge, MA 02138}
12: \altaffiltext{1}{Also at the Institute for Nuclear Fusion, RRC ``Kurchatov
13: Institute'', Moscow 123182, Russia; E-mail: mmedvedev@cfa.harvard.edu;
14: URL: http://cfa-www.harvard.edu/\~{ }mmedvede/ }
15:
16:
17: \begin{abstract}
18: Particle heating in advection-dominated accretion flows (ADAFs) by nonlinear
19: MHD (Alfv\'enic) turbulence is investigated. Such turbulence with
20: highly-fluctuating magnetic fields, $\tilde B\sim B_0$, is believed to be
21: naturally produced by the magnetic shearing instability near the nonlinear
22: saturation. It is shown that the energy is dissipated in the parallel cascade,
23: which occurs due to nonlinear compressibility of high-amplitude turbulence, and
24: predominantly heats protons, but not electrons.
25: The conservative limit on the electron--to--proton heating fraction
26: is $\delta\lesssim{\rm few}\times10^{-2}$.
27: \end{abstract}
28: \keywords{accretion, accretion discs --- MHD --- plasmas --- turbulence ---
29: waves}
30:
31: \section{Introduction}
32:
33: The Advection-Dominated Accretion Flows (ADAFs) are a class of hot, optically
34: thin accretion solutions
35: \citetext{\citealt{Ichimaru77,Rees-etal82,NY95,Abramowitcz-etal95};
36: see e.g., \citealt{NMQ98} for review}
37: which describe quite well the spectral characteristics of a number of low
38: luminosity accreting black hole systems, e.g., black hole binaries
39: \citep{NBMc97,Hameury-etal97} and low luminosity galactic nuclei
40: \citetext{e.g., Sgr A$^*$:
41: \citealt{NYM95,Manmoto-etal97,Mahadevan98,Narayan-etal98};
42: NGC 4258: \citealt{Lasota-etal96,GNB98};
43: M87 and other ellipticals: \citealt{Reynoldsetal96,Mahadevan97,DMetal98}}.
44: In ADAFs, the protons and electrons are thermally decoupled
45: (Coulomb collisions are rare), so that all the
46: energy generated by turbulent viscous stresses is stored as thermal energy
47: of the protons and ultimately advected beyond the black hole horizon,
48: while the electrons remain relatively cool and radiate much less energy, hence
49: the sub-Eddington luminosities. In such hot accretion flows, however, the
50: particle mean free path is often comparable to the size of the system, and thus
51: collective plasma effects are likely to be significant \citep{Rees-etal82}.
52:
53: In the absence of collisions, strong MHD turbulence is required for
54: the angular momentum transport and energy dissipation in accretion flows.
55: The way MHD turbulence dissipates in hot accretion flows turns out to be
56: crucial for the relevance of ADAF models. Let's denote $P_e$ and $P_p$ the
57: amounts of energy that heat the electrons and protons, respectively.
58: In traditional ADAF models, the branching ratio parameter $\delta=P_e/P_p$ is
59: commonly assumed to be less than or similar to $10^{-2}$. The predicted spectra
60: are only weakly sensitive to the exact value of $\delta$ so long as
61: $\delta\lesssim10^{-2}$, while for larger values of $\delta$ the changes in
62: spectrum characteristics are drastic because the electrons become hot and
63: radiate a large fraction of the internal energy of the flow via the synchrotron
64: emission so that the flow is no longer advection-dominated. The above
65: constraint on the value of $\delta$ has been recently relaxed in the
66: modified ADAF model with wind outflows \citep{BB98}. Because of the lower
67: accretion rates (due to the mass loss) at the inner regions,
68: where most of the radiation is produced, the density of the inflowing gas and
69: the magnetic field strength decrease (assuming constant $\beta$) and result in
70: a much lower radiative efficiency of the flow. It was demonstrated by
71: \cite{QN98} that the values of $\delta$ as high as $\delta\sim0.3$ are still
72: plausible in the advection-dominated flows with strong winds.
73: The question of turbulent particle heating
74: and non-thermal coupling of electrons and protons in hot accretion flows has
75: been addressed in several works \citep{Quataert98,QG98,BkL97,BC88,NY95}.
76: To proceed further, we need to understand
77: what type of MHD turbulence is most likely to be realized in ADAFs.
78:
79: It is commonly believed that MHD turbulence in accretion flows is generated
80: by the magneto-rotational instability \citep{BalbusHawley91}, which
81: generates the large-scale magnetic field as well as produces random gas
82: motions and strong fluctuating magnetic fields in the nonlinear regime. Hence,
83: a variety of plasma waves may be generated. Only low-frequency waves may be
84: efficiently excited by large scale motions and carry a significant fraction
85: of the turbulent energy. \citet{Quataert98} shows that among the low-frequency
86: MHD waves, fast and slow magnetosonic waves are strongly damped in the
87: two-temperature $T_p\sim10^{12}\textrm{ K},\ T_e\sim10^9\textrm{ K}$
88: ($T_p$ and $T_e$ are the proton and electron temperatures) ADAF plasma
89: by collisionless dissipation (known as Landau damping and
90: its magnetic analog the transit time damping) while Alfv\'en waves are
91: still very weakly damped in the incompressible approximation.
92: As argued by \citet{Quataert98}, the ``impedance mismatch'' will likely inhibit
93: excitation of heavily damped modes and thus the MHD turbulence in ADAFs will
94: be (predominantly) {\em Alfv\'enic}. Additionally, it is natural to believe
95: that the turbulent fluctuating magnetic fields are of order of the mean, large
96: scale magnetic field, $\tilde B\sim B_0$, and the Balbus-Hawley instability is
97: nonlinearly saturated.
98:
99: A detailed study particle heating processes by Alfv\'enic turbulence in ADAFs
100: neglecting the effects of compressibility has been performed by
101: \citet{Quataert98} and \citet{QG98}. They assumed that the energy which is
102: injected into the system on large scales comparable to the size of the
103: accretion flow, $L\sim R$, will cascade in $k$-space to small scales,
104: typically the proton Larmor scale, $l\sim\rho_p={v_{\rm th}}_p/\Omega_p$
105: (with ${v_{\rm th}}$ being the thermal particle
106: velocity and $\Omega=eB_0/mc$ being the cyclotron frequency) or even smaller,
107: where it dissipates. Since the Alfv\'en wave cascade is likely anisotropic
108: \citep{GS95}, the $k$ components are related as
109: $k_\|\sim k_\bot^{2/3}L^{-1/3}$. Thus, when the cascade reaches $k_\bot$ such
110: that $k_\bot\rho_p\sim1$, the wave energy dissipates only via Landau damping
111: (Cherenkov resonance) and may preferentially heat either the electrons or the
112: protons, depending on the value of plasma $\beta$ ($\beta=8\pi p/B_0^2$ is the
113: ratio of gas to magnetic pressure).\footnotemark
114: %
115: \footnotetext{The plasma physics $\beta$ used in this paper is different
116: from that used in ADAF models and is related to it as
117: $\beta_{\rm adaf}=\beta/(\beta+1)$.}
118: %
119: The cyclotron damping which may be very
120: efficient to heat protons is unimportant in this case, because $k_\|\ll
121: k_\bot$ and $\omega=k_\|v_A\ll\Omega_p$ (where $v_A^2=B_0^2/4\pi m_pn$ is the
122: Alfv\'en speed and $n$ is the particle density), so that the cyclotron resonance
123: ($\omega-k_\|v-\Omega_p=0$) is satisfied for a negligibly small population of
124: very energetic particles from the tail of the Maxwellian distribution function.
125: \citet{QG98} also argue that only a fraction of Alfv\'enic energy may be
126: dissipated on the scale $k_\bot\sim\rho_p^{-1}$ and the rest of it passes this
127: `barrier'. On scales $k_\bot\gg\rho_p^{-1}$ Alfv\'en waves may be converted
128: to whistler waves and cascade even further where dissipation on electrons
129: may be dominant. Because the details of how Alfv\'en waves are converted into
130: whistler waves on scales $k_\bot\sim\rho_p^{-1}$ are not known, they estimate
131: that the transition from the electron dominated heating regime to the proton
132: dominated one occurs somewhere inbetween $\beta\sim5$ and $\beta\sim10^2$.
133:
134: The above analysis is accurate for low-amplitude turbulence,
135: $\tilde B/B_0\ll1$, where one can neglect the effect of finite magnetic field
136: pressure, $\tilde B^2/8\pi$, (i.e., incompressible plasma)
137: and use the linear theory. In a high-amplitude turbulence, $\tilde B/B_0\sim1$,
138: which is likely to be present in accretion flows,
139: finite magnetic pressure of waves nonlinearly couples Alfv\'enic energy
140: to ion-acoustic-like (i.e., density) quasi-mode perturbations, which are, in
141: general, dissipative. In this paper, we show that:
142: \begin{enumerate}
143:
144: \item
145: The Alfv\'enic {\em parallel} cascade towards higher $k_\|$ exists
146: in a compressible MHD turbulence as a result of nonlinear wave steepening
147: (modulational instability), in addition to the perpendicular
148: (incompressible) cascade in $k_\bot$. This parallel cascade is
149: nonlinearly dissipative because it proceeds via the excitation of ion-acoustic
150: (i.e., compressible) quasi-modes which are damped via the Landau mechanism.
151: Even for relatively small amplitudes of turbulence, $\tilde B/B_0\gtrsim0.2$,
152: and for $\beta\gtrsim3$, much or almost {\em all} magnetic energy of Alfv\'en
153: waves dissipates before it reaches the perpendicular `dissipation barrier',
154: $k_\bot\rho_p\sim1$. The perpendicular cascade, thus, turns out to be
155: energetically unimportant. Dissipation in the nonlinear parallel cascade
156: preferentially heats the protons and yields
157: $\delta\lesssim{\rm few}\times10^{-2}$.
158:
159: \item
160: If $\beta\lesssim3$, the turbulence bifurcates to another `phase' where the
161: Afv\'enic dissipation in the parallel cascade is weak and other dissipation
162: mechanisms (at small scales) are required to maintain a steady-state.
163: Since the parallel cascade proceeds up to
164: $k_\|\simeq4(\tilde B/B_0)(\Omega_p/v_A)$, the Alfv\'enic turbulent energy may
165: be efficiently dissipated due the cyclotron damping on protons,
166: if $\omega=k_\|v_A\simeq\Omega_p$, i.e., if $\tilde B/B_0\gtrsim0.3$. Only
167: protons may be heated in this regime. For lower amplitudes, the perpendicular
168: Goldreich-Sridhar cascade is likely to dominate and energy dissipates, as
169: discussed by \citet{Quataert98} and \citet{QG98}. The three-wave cascade
170: \citep{NB96}, if it dominated the above four-wave cascade, turns out to
171: be insignificant for electron energetics.
172: \end{enumerate}
173: We also consider other `higher-order' nonlinear effects, such as particle
174: trapping, and their relevance to particle heating in ADAFs.
175:
176: The paper is organized as follows. In \S \ref{S:N-KNLS} we discuss the
177: noisy-KNLS model of strong MHD turbulence. In \S \ref{S:ADAF} we apply the
178: model to the conditions in ADAFs. We compare the model at hand with other
179: competing turbulent processes in \S \ref{S:COMPETE}.
180: Section \S \ref{S:CONCL} is the conclusion.
181:
182:
183: \section{The model of nonlinear Alfv\'en wave turbulence \label{S:N-KNLS}}
184:
185: \subsection{General considerations}
186:
187: It is known that in an incompressible plasma, $\nabla\cdot{\bf v}=0$, the
188: Reynolds and magnetic stresses (i.e., the fluid and magnetic nonlinearities)
189: in a finite-amplitude Alfv\'en wave cancel each other exactly.
190: Such a wave behaves as if it is linear. Hence, only the Goldreich-Sridhar
191: Alfv\'enic perpendicular, $k_\bot$, cascade occurs in a turbulent regime.
192: The mutual cancellation of Reynolds
193: and magnetic stresses in the wave breaks down when plasma is compressible.
194: Indeed, the magnetic field pressure, $\tilde B^2/8\pi$, associated with the
195: wave field exerts an additional stress onto an ionized gas and change its
196: local density, $n$. Variations in $n$ affect, in turn, the local wave phase
197: speed, $v_A$, and introduce a positive nonlinear feedback into the wave
198: dynamics \citep{CohenKulsrud71}. This nonlinear phase speed--amplitude
199: coupling in {\em nonlinear Alfv\'en} waves results in the {\em modulational
200: instability} which is ultimately responsible for the wave-front steepening,
201: parallel, $k_\|$, Alfv\'enic cascade, and formation of collisionless shocks.
202: The compressional nonlinearity may also be viewed as an effective parametric
203: coupling of a finite-amplitude Alfv\'en wave to ion-acoustic (i.e., density)
204: wave-like perturbations in a medium.\footnotemark
205: %
206: \footnotetext{
207: Such ion-acoustic modes are {\em not} plasma eigenmodes.
208: They are driven by the Alfv\'en wave and, hence, propagate with the
209: Alfv\'en speed, which is, in general, different from the sound speed in
210: plasma.}
211: %
212: These ion-acoustic quasi-modes are compressional and have a longitudinal
213: component of the electric field (with respect to the ambient magnetic field).
214: Hence, they efficiently dissipate the energy via Landau damping.
215: A quantitative theory of nonlinear Alfv\'en waves is now well established
216: \citetext{see a short review by \citealt{KNLSreview} and references therein}.
217: Note that the dissipation of the finite-amplitude Alfv\'en waves is
218: intrinsically nonlinear process and leads to a number of unusual properties
219: of such waves \citep{KNLSreview}.
220:
221: \subsection{The model \label{S:MODEL}}
222:
223: There is only one (to our knowledge) model of strong nonlinear Alfv\'en wave
224: turbulence in a compressible plasma which self-consistently includes the
225: effects of wave nonlinearity and particle kinetics, e.g., Landau damping
226: \citep{MD97}. It is formulated as a stochastic modification of the
227: dynamic equation of Alfv\'en waves, --- the kinetic nonlinear Schr\"odinger
228: equation (KNLS). This model of turbulence, referred to as the ``noisy-KNLS,''
229: is thus structure-based, meaning that it describes the turbulence as a
230: collection of strongly interacting {\em coherent} structures, such as
231: shocks, solitons, cnoidal waves, rotational discontinuities,
232: etc., generated by an external random noise source.\footnotemark
233: %
234: \footnotetext{In this respect, the model is analogous to the noisy-Burgers
235: model of hydrodynamic turbulence which describes the turbulence
236: as a collection of randomly interacting collisional shocks.}
237: %
238: The noisy-KNLS equation is
239: \beq
240: \frac{\partial b}{\partial \tau}+v_A\frac{\partial}{\partial x}
241: \left(N_1 b\widetilde{|b|^2}
242: +N_2 b\hat{\cal H}\bigl[\widetilde{|b|^2}\bigr]\right)
243: + i\frac{v_A^2}{2\Omega_p}\frac{\partial^2 b}{\partial x^2}=\tilde f ,
244: \label{noisy-knls}
245: \eeq
246: where $\hat{\cal H}$ is the {\em integral operator}, referred to as the Hilbert
247: operator (or the Hilbert integral transform), acting on the wave field:
248: \beq
249: \hat{\cal H}\bigl[\widetilde{|b|^2}\bigr]=\frac{1}{\pi}
250: \int_{-\infty}^{\infty}\frac{{\cal P}}{(x'-x)}
251: \widetilde{|b(x')|^2}{\rm d}x' .
252: \eeq
253: Here also $b=(\tilde B_y+i\tilde B_z)/B_0$ is the normalized complex wave
254: magnetic field, $\widetilde{|b|^2}=|b|^2-\langle|b|^2\rangle$ is the
255: fluctuating component of the magnetic pressure, $\langle\dots\rangle$ means
256: appropriate averaging over the spatial domain, ${\cal P}$ means principal
257: value integration, $\tilde f$ is the random (turbulent) noise, and the
258: coefficients $N_1$ and $N_2$ are complicated functions of parameters of a
259: plasma and are discussed below.
260:
261: The first nonlinear term in equation (\ref{noisy-knls}) describes the nonlinear
262: (ponderomotive) coupling of finite-amplitude Alfv\'en waves to ion-acoustic
263: modes. Indeed, the first two terms are similar to a continuity equation for
264: the wave magnetic field, $\partial_\tau b+\partial_x(ub)=0$, where the
265: ``hydrodynamic'' velocity, $u$, is generated by the wave pressure,
266: $u\propto v_A\widetilde{|b|^2}$. The process of nonlinear steepening stops
267: when it is balanced by linear Alfv\'en wave dispersion which occurs at very
268: small scales comparable to proton's Larmor radius, $k_\|\sim\rho_p^{-1}$,
269: as represented by the last, linear in $b$, term.
270: The second nonlinear term in equation (\ref{noisy-knls}) describes Landau
271: (i.e., collisionless) damping of the induced density perturbations, not the
272: magnetic fluctuations. This collective kinetic effect is represented by the
273: integral Hilbert operator. We emphasize that the
274: damping is {\em nonlocal} which can be interpreted as the effect of particle
275: ``memory'' which is associated with particle transit through the wave packet.
276: Another interesting and important feature is the absence of any particular
277: scale beyond which the damping dominates. Indeed, the Fourier representation
278: of the Hilbert operator $\hat{\cal H}[f]$ acting on a function $f$ is
279: $(ik_\|/|k_\||)f_k\equiv i\,{\rm sgn}(k_\|)\,f_k$, i.e., independent of the
280: magnitude of $k_\|$. Thus, Landau damping of nonlinear Alfv\'en waves occurs
281: at all scales; it is {\em scale invariant}.
282: We should comment that both nonlinearities result in the wave steepening and
283: formation of sharp fronts (shocks) which is, practically, equivalent to the
284: parallel cascade of wave energy to smaller scales. Unlike the conventional,
285: incoherent cascade paradigm, however, this cascade proceeds
286: through random interactions of coherent wave structures (such as shocks,
287: nonlinear waves, etc.). Hence, collective wave--particle interactions,
288: such as the nonlinear Landau damping and particle trapping also contribute.
289:
290: We now briefly discuss basic assumptions used in the noisy-KNLS model. First,
291: equation (\ref{noisy-knls}) is the {\em envelope} equation. That is, Alfv\'en
292: waves themselves (carrier waves) are assumed to be linear and, thus,
293: obey the dispersion $\omega=k_\|v_A$ while the amplitude of these waves may
294: vary in space and time and is described by equation (\ref{noisy-knls}). The time
295: variable $\tau$ is the ``slow'' time of the large-scale dynamics of the
296: wave envelope, $\tau=(B_0/\tilde B)^2t_A$, where $t_A=\omega^{-1}$ is the
297: typical ``Alfv\'enic'' time. It has been assumed in derivation that
298: $(\tilde B/B_0)^2$ is small but finite. Comparison of the predictions of
299: this model with observations of nonlinear Alfv\'en waves in the solar wind by
300: spacecrafts shows that the KNLS theory is ``robust'' and supports its validity
301: even on the margin of applicability, $(\tilde B/B_0)\simeq1$.
302:
303: Second, the noisy-KNLS also assumes that waves are propagating in one
304: direction, only. The existence of counter-propagating waves allows for
305: another type of parametric wave--wave interactions, ---
306: the decay instability, due to which an Alfv\'en wave may decay into an acoustic
307: wave and a counter-propagating Alfv\'en wave, --- which greatly complicates any
308: analytical treatment of the problem. The decay instability, however, may be
309: greatly suppressed in hot, two-temperature ADAF conditions. An
310: acoustic (compressible) wave that is to be generated is heavily damped by
311: collisionless dissipation. The same argument that the ``impedance mismatch''
312: inhibits resonant excitation of strongly damped modes suggests here (but does
313: not prove, though) that the parametric resonance discussed above will be
314: inefficient.
315:
316: Third, the KNLS-based theory considers planar waves, only. That is, no wave
317: packet modulations are allowed in the plane perpendicular to the direction of
318: wave propagation; the wave fronts are always flat. The dynamics of waves is,
319: thus, purely one-dimensional. This strong limitation of the dynamics and
320: evolution of wave structures in the Alfv\'enic turbulence has, probably,
321: very little effect (if any) on the processes of particle heating.
322: Indeed, on large scales, particles move along field lines, only. Thus,
323: heating of particles may occur (in Cherenkov resonance) via increase of
324: their parallel velocity, only. The perpendicular wave evolution, therefore,
325: hardly affects particle energy, unless $k_\bot\sim\rho_p^{-1}$.
326: Note, the effective one-dimensionality of the
327: problem does not mean that waves may propagate along the ambient magnetic
328: field, ${\bf B_0}$, only. It has been shown that waves propagating at angle
329: are still described by equation (\ref{noisy-knls}) if one formally writes the
330: wave field as $b=(\tilde B_y+B_0\sin{\Theta}+i\tilde B_z)/B_0$, where
331: $\Theta$ is the angle\footnotemark
332: %
333: \footnotetext{ Strictly speaking, the angle $\Theta$ cannot be pushed close
334: to $90^\circ$ because dispersion becomes different for left-hand-
335: and right-hand-polarized waves (dispersion on electrons vs. dispersion
336: on protons). This sets the limit on the angle
337: $(\pi/2-\Theta)\gg\sqrt{m_e/m_p}$. }
338: %
339: between ${\bf k}$ and ${\bf B_0}$.
340:
341: Fourth, the nonlinear coupling coefficients, $N_1$ and $N_2$, presented in
342: the Appendix are calculated by solving the particle kinetic (Vlasov) equation
343: \citep{Spangler89,Spangler90} and agree with those calculated using the guiding
344: center formalism \citep{MW88}. Both methods, however, used perturbative
345: techniques and do not include nonlinear effects such as particle trapping.
346: Maxwellian distribution was assumed for bulk particles. \citet{GQ99} showed
347: that in a strong Goldreich-Sridhar turbulence the time-asymptotic distribution
348: function of protons may be different. It is not clear whether this situation
349: occurs in ADAF because (i) the infall (accretion) time may be shorter than
350: the velocity diffusion time required to establish a new distribution and
351: (ii) the emergent distribution should, in general, be anisotropic (because
352: the turbulence is anisotropic) and thus may be unstable against, for instance,
353: fire-hose instability which will isotropize and thermalize the particles.
354: Since Landau damping is a resonant process, it cares about the local slope
355: of a distribution function, but its overall shape is irrelevant. Therefore,
356: using the Maxwellian distribution is reasonable.
357:
358: Finally, we do not specify the origin of random forcing for the moment.
359: In general, involvement of a particular mechanism would force re-ordering
360: in equation (\ref{noisy-knls}) and changes in $N_1$ and $N_2$.
361: This complication is automatically eliminated in the renormalization-group
362: approach, the key idea of which is to use bare nonlinear couplings to
363: compute the noise-renormalized (i.e., turbulent) quantities \citep{MD97}.
364: Going ahead, the reason why the bifurcation point (see next section) seems
365: to be determined by the bare $N$'s and not the turbulent ones is simple:
366: both nonlinearities are cubic and renormalize in the same way.
367:
368:
369: \subsection{The structure of Alfv\'enic turbulence}
370:
371: We are interested in a steady-state turbulence in which the energy input
372: due to random forcing is balanced by Landau dissipation. Since this
373: dissipation is independent of $|k|$, it occurs at {\em all scales}; thus,
374: no inertial interval exists. An analytical study of the model for arbitrary
375: noise is an extremely laborious task. That is why it has been analyzed only
376: in the simplest
377: case of $\delta$-correlated in space and time, zero-mean, white in $k$ noise
378: \citep{MD97}. A one-loop renormalization group technique has been utilized.
379: The noise-dependent renormalized turbulent transport coefficients has been
380: obtained. Quite importantly, they are mostly determined by properties of the
381: noise source on the largest scales (i.e., the system size), --- the so called
382: infrared divergences. The large-$k$ tail of the noise spectrum seems to be
383: relatively insignificant to control global properties of turbulence.
384: Thus, the assumption of white noise is likely to be unproblematic
385: in the context of turbulence in hot accretion flows, because the spectrum of
386: turbulence is not known, anyway.
387:
388: The existence of two distinct {\em phases} (or regimes) of compressible
389: Alfv\'enic turbulence constitutes the most important and interesting
390: prediction of the theory. The system will {\em bifurcate} (rather than
391: smoothly transit) from one phase to another, as plasma parameters smoothly
392: change. The bifurcation point is determined by the ratio of the
393: nonlinearity-to-damping coefficients:
394: \beq
395: \left.\frac{|N_1|}{|N_2|}\right|_{\rm bif}\simeq1.3\, .
396: \label{bif}
397: \eeq
398: This result has been obtained rigorously from the analysis of the
399: self-consistency of the infrared cutoff scalings of the solutions for turbulent
400: transport coefficients. The physical interpretation, however, is very
401: simple and clear. If $|N_1|/|N_2|<1.3$, the collisionless dissipation always
402: dominates the nonlinear parallel energy cascade (i.e., the wave steepening).
403: Most of the energy injected on a large scale dissipates {\em before} it reaches
404: the dispersion scale, and {\em all} the energy is dissipated {\em via} Landau
405: damping in the Alfv\'enic cascade. Thus, the steady-state, so-called
406: ``hydrodynamic'' ($\omega,k\to0$) turbulence with no steep fronts is predicted.
407: In the opposite case $|N_1|/|N_2|>1.3$, no mathematically rigorous prediction
408: can be made, since no solution to renormalized equations exists. Speculatively,
409: however, it is clear that the collisionless damping is weak compared to the
410: nonlinear steepening. Only a small fraction of the injected energy
411: dissipates, while most of it cascades along $k_\|$ to the smallest scales set by
412: the linear dispersion. Since the source continuously injects energy into the
413: system, it will be {\em accumulated} at the smallest scales.
414: Hence, a non-steady-state,
415: small-scale turbulence of Alfv\'enic shocklets is expected. Stationarity may
416: be enforced by including other dissipation mechanisms into the model.
417: The most efficient one is the cyclotron damping on protons because
418: it operates at nearly the same scales
419: (i.e., the proton Larmor radius) where the wave energy is accumulated.
420: Another mechanism which dominates for linear and weakly nonlinear waves
421: ($\tilde B/B_0\ll1$) is the perpendicular cascade \citep{GS95,NB96}. The energy
422: may cascade through $k_\bot\sim\rho_p$ to smaller scales where Alfv\'en waves
423: are probably converted into whistler waves. How energy is dissipated in
424: this case is unclear \citep{QG98}.
425: The discussion of these issues and quantitative estimates of the rates
426: of the perpendicular vs. parallel cascade and cyclotron damping are given in
427: \S \ref{S:COMPETE}.
428:
429:
430: \section{Application to hot accretion flows \label{S:ADAF} }
431:
432: \subsection{Preliminary notes}
433:
434: In this section, we primarily focus on the ADAF solutions which are the only
435: known thermally stable, self-consistent models of hot, two-temperature
436: accretion flows. The self-similar scalings of plasma parameters relevant for
437: our problem are \citetext{\citealt{NY95}, see also \citealt{Quataert98}}:
438: %
439: \begin{mathletters}
440: \begin{eqnarray}
441: T_p&\simeq&2\times10^{12}\frac{\beta}{\beta+1}r^{-1}\textrm{ K}, \\
442: T_e&\sim&10^9-10^{10}\textrm{ K}, \\
443: \beta&=&const.,\\
444: B_0&\simeq&10^9\alpha^{-1/2}(1+\beta)^{-1/2}m^{-1/2}\dot m^{1/2}r^{-5/4}
445: \textrm{ G}, \\
446: \frac{\rho_p}{R}&\simeq&6\times10^{-9}\alpha^{1/2}\beta^{-1/2}
447: m^{-1/2}\dot m^{-1/2}r^{-1/4}, \\
448: \frac{v_A}{c}&\simeq&0.9(1+\beta)^{-1/2}r^{-1/2}, \\
449: \frac{v_r}{c}&\simeq&0.37\alpha r^{-1/2},
450: \end{eqnarray}
451: \label{adaf}
452: \end{mathletters}
453: %
454: where $\alpha$ is the Shakura-Sunyaev turbulent viscosity parameter,
455: $m=M/M_{\sun}$ is the mass of the central object in solar mass units,
456: $\dot m=\dot M/\dot M_{\rm Edd}$ is the accretion rate in Eddington units
457: ($\dot M_{\rm Edd}=1.39\times10^{18}m\textrm{ g s}^{-1}$), and $r=R/R_S$ is the
458: local radius in Schwarzschild units ($R_S=2.95\times10^5m\textrm{ cm}$).
459: Here $B_0$ is the strength of the large-scale magnetic field, $\beta$ is
460: assumed to be independent of $R$ and takes, likely, sub-equipartition values
461: ($\beta\gtrsim1-10$), $\rho_p/R$ is the proton Larmor scale in terms of
462: the local scale of the accretion flow, and $v_r/c$ and $v_A/c$ are the
463: radial inflow velocity and the non-relativistic Alfv\'en speed,\footnotemark\
464: respectively, in units of the speed of light.
465: %
466: \footnotetext{In the relativistic case, the displacement current cannot be
467: neglected in Maxwell's equations. The relativistic Alfv\'en
468: velocity then reads $v_A^{\rm rel}=v_A/\sqrt{1+v_A^2/c^2}$.
469: We neglect this relativistic effect in our paper because it may be
470: significant very close to the inner edge of the flow, near the
471: last stable orbit, $r\sim3$, only. }
472: %
473: The electron temperature is nearly constant throughout the accretion flow
474: and starts to decrease at large radii, $r\gtrsim10^2-10^3$. The efficient
475: cooling of relativistic electrons prevents higher temperatures at smaller
476: radii. Thus the proton-to-electron temperature ratio ranges from
477: $T_p/T_e\sim10^3$ deeply inside the flow at $R\sim\textrm{few}\times R_S$
478: to $T_p/T_e\sim1$ in the outer regions, $R/R_s\gtrsim10^3$.
479:
480: The MHD turbulence in hot, collisionless accretion flows is likely to
481: originate as a result of the magneto-rotational instability
482: \citep{BalbusHawley91}. In the linear phase, the Balbus-Hawley instability
483: predominantly amplifies the toroidal component of the large-scale magnetic
484: field, so that the fluctuations are weak, $\tilde B\ll B_0$. When the
485: instability reaches nonlinear amplitudes, it becomes large scale,
486: $k_{BH}\sim R^{-1}$, so that long wavelength MHD waves are excited and
487: the MHD turbulence results, as seen from numerical simulations
488: \citetext{see, e.g., a review by \citealt{BH98} and references therein}.
489: This turbulence
490: is generated on large scales and, hence, carries most of the gravitational
491: potential energy of the accreting gas. Since the ADAF gas/plasma is
492: two-temperature, with the protons much hotter than the electrons, all
493: compressional MHD modes are heavily damped by the collisionless (Landau and
494: transit time) damping and thus are hard to excite, as suggested by
495: \citet{Quataert98}. Thus, only noncompressive (Alfv\'en) MHD modes may
496: be efficiently excited and reach nonlinear amplitudes. The nonlinear saturation
497: of the Halbus-Hawley instability means that the local fluctuating magnetic
498: fields in the accretion flow become comparable to the averaged global magnetic
499: field and prevent its further growth; thus in the absence of
500: dissipation\footnotemark\ $\tilde B/B_0\sim1$.
501: %
502: \footnotetext{This condition does not place any constraint on the value of
503: $B_0$.}
504: %
505: At such amplitudes, noncompressibility of Alfv\'en waves is no longer a
506: good approximation and the nonlinear theory of strong MHD turbulence
507: discussed in the previous section should be used instead. In this case,
508: Landau damping will decrease the amplitude of turbulence to a level when
509: dissipation balances energy input. Given the noise properties, the
510: amplitude of turbulence may be estimated from \citet{MD97}.
511:
512:
513: \subsection{Dissipation-dominated parallel cascade}
514:
515: As we discussed in the previous section (\S \ref{S:N-KNLS}), there is a regime
516: of strong nonlinear Alfv\'enic turbulence in which all energy that is injected
517: into a system on large scales dissipates in the cascade and does not reach the
518: proton Larmor radius scales. The bifurcation point, which is given by
519: Eq.\ (\ref{bif}), is a function of $T_p/T_e$ and $\beta$, only
520: (via $N_1$ and $N_2$, see Appendix). It is independent of the fluctuation level
521: $b\propto\tilde B/B_0$ because steepening and damping are both cubic in $b$
522: and the wave amplitude cancels out. (The rate of dissipation, however, does
523: depend on the amplitude of turbulence, as discussed below in
524: \S \ref{SSS:RATES-PAR}.) Using the expressions for $N_1$ and $N_2$ given in
525: Appendix and Eq.\ (\ref{bif}), we now plot the borderline between the two
526: regimes, as a function of plasma parameters.
527:
528: The $T_p/T_e$-$\beta$-diagram of state of the
529: nonlinear Alfv\'enic turbulence is shown in Fig.\ \ref{fig-state}.
530: The shaded region corresponds to $|N_1/N_2|<1.3$ and the unshaded region ---
531: to $|N_1/N_2|>1.3$. The bold-faced part of the boundary curve represents the
532: range of the temperature ratio relevant for ADAFs:
533: $T_p/T_e$ decreases from $\sim10^3$ in the inner regions of the accretion
534: flow at radii $R\lesssim10R_S$ to $\sim1$ in the outer regions at
535: $R\gtrsim10^2-10^3R_S$. As is seen from the figure, the boundary between the
536: two phases of turbulence is insensitive to the temperature ratio for the
537: ADAF conditions and set only by plasma $\beta$.
538: If $\beta>\beta_{\rm crit}\simeq2.6$, the Alfv\'enic turbulence in an ADAF
539: is in the regime in which the parallel cascade is strongly dissipative.
540: If additionally the parallel cascade is dominant (i.e., faster than any other
541: cascade/dissipation process), then it dissipates all injected energy.
542: Alternatively, for $\beta<\beta_{\rm crit}\simeq2.6$, Alfv\'en wave energy
543: does not dissipate in the parallel cascade (even if it is dynamically dominant)
544: and reaches $k_\|\sim\rho_p^{-1}$ where it may be dissipated by other
545: mechanisms, or cascade beyond this point to even smaller scales.
546:
547: \subsection{Electron vs. proton heating in the parallel cascade
548: \label{SS:EL-HEAT} }
549:
550: We can also investigate the electron-to-proton heating ratio, $\delta$, using
551: the equations for $N_1$ and $N_2$ as follows. The electron and proton
552: contributions to the wave dissipation are proportional to the imaginary part of
553: the plasma dispersion function, Eq.\ (\ref{plasma-disp}). The factor in
554: Eq.\ (\ref{plasma-disp}) with integral over the distribution function
555: has been explicitly incorporated into the wave equation,
556: Eq.\ (\ref{noisy-knls}), as the Hilbert operator. The numerical factor
557: was left in the coefficients, Eqs.\ (\ref{f's}) -- (\ref{g's}), and is
558: proportional to $e^{-X^2_e}$ and $e^{-X^2_p}$ for the electrons and protons,
559: respectively. By vanishing by hands all those terms which are multiplied by
560: $e^{-X^2_e}$, we may easily determine how much the electrons contribute to the
561: overall dissipation of waves. Calculaing the ratio $(1-N_2^{*})/N_2^{*}=\delta$,
562: where $N_2^{*}$ is the nonlinear dissipation coupling coefficient without the
563: contribution of the electrons to dissipation, we obtain the heating ratio.
564: Note, that in $N_2^{*}$ the electrons still contribute to the linear and
565: nonlinear dispersion of waves, as represented by the real part of the plasma
566: distribution function.
567:
568: A contour plot showing the levels of constant heating ratio,
569: $\delta=.005,\, .01,\, \dots,\, .035$, is superimposed onto the Alfv\'enic
570: dominated dissipation region in Fig.\ \ref{fig-state}. Fig. \ref{fig-heat}
571: also depicts the heating ratio as a function of the temperature ratio
572: $T_p/T_e$ for two values of $\beta$,\ $\beta=3$ and $40$.
573: Clearly, $\delta$ depends on $\beta$ only weakly and ranges with
574: temperature from $\sim.025$ to $\sim.005$ for the typical ADAF conditions.
575: Since, the spectral ADAF models weakly depend on $\delta$ so long as
576: $\delta\lesssim\textrm{few}\times10^{-2}$, we set a conservative upper limit
577: on the electron heating in the dissipative {\em parallel} cascade of nonlinear
578: Alfv\'enic turbulence:
579: \beq
580: \delta_\|\equiv\left.\frac{P_e}{P_p}\right|_{\|\,{\rm casc}}\lesssim.025 ,
581: \quad\textrm{ if }\ \beta>\beta_{\rm crit}\simeq2.6 .
582: \label{delta-limit}
583: \eeq
584:
585:
586: \section{Other competing processes \label{S:COMPETE} }
587:
588: In the previous section (\S \ref{SS:EL-HEAT}) we demonstrated that the
589: energy which heats the electrons in the compressible Alfv\'enic dissipative
590: parallel cascade in typical ADAF conditions constitues a few percents of the
591: total energy which goes into the protons. If this parallel cascade would be
592: the only one which operates, then equation (\ref{delta-limit}) gives a solid
593: prediction. In fact, there are other competing processes which transfer
594: Alfv\'enic energy in $k$-space and thus may affect the energetics.
595: We consider them in order.
596:
597: In addition to the compressible parallel cascade, there is an incompressible
598: perpendicular cascade due to direct interactions of linear Alfv\'en wave
599: packets. Recently two different theories of this process have been proposed.
600: \citet{GS95} (hereafter GS) showed that 4-wave interactions lead to a
601: critically balanced anisotropic cascade in which $k_\|$ and $k_\bot$ are
602: uniquely related. A nice paper by \citet{NB96} (hereafter NB) demonstrates
603: that 3-wave interactions of $k_\|=0$ perturbations are not empty and result
604: in a more conventional Iroshnikov-Kraichnan-type cascade. Although there are
605: some indications from both theoretical arguments and numerical simulations
606: \citep{GS97,Vishniac00} that the GS cascade dominates over the NB cascade,
607: the debate is still not completely resolved. For this reason, we consider both
608: possibilities below. Briefly, we obtained that the GS cascade puts some
609: limitations on our predictions of \S \ref{SS:EL-HEAT}. The NB cascade,
610: however, does not alter our conclusions about the electron vs proton heating,
611: despite that the total wave energy in the perpendicular direction may be
612: substantial.
613:
614:
615: \subsection{Parallel vs. GS cascade \label{SSS:RATES-PAR} }
616:
617: The critically balanced GS cascade \citep{GS95} is an anisotropic,
618: (almost) perpendicular cascade in which $k_\|\ll k_\bot\propto k_\|^{3/2}$.
619: This cascade involves linear Alfv\'en waves and, thus, is independent of the
620: amplitude of the turbulence. No damping occurs in this process. The MHD
621: (Alfv\'en) energy is transferred from the scales where it is injected,
622: $k_{\bot i}\sim R^{-1}$, to the scales where it is dissipated or converted into
623: other waves, $k_{\bot f}\sim\rho_p^{-1}$. We have to compare it with the
624: competing compressible parallel cascade during which much of injected energy
625: is dissipated. The rate of
626: the energy transfer depends on the turbulence level via the compressibility.
627: Since these two cascade processes are independent (at least in the KNLS
628: approximation), we may compare relative contribution of each of them to
629: the electron heating as a function of the amplitude of MHD turbulence in ADAFs.
630:
631: The characteristic rate (i.e., the inverse $e$-folding time) of the
632: GS cascade is the Alfv\'enic frequency,
633: $\gamma_{\rm GS}\sim\omega_A$. The characteristic dissipation rate in the
634: parallel cascade and the nonlinear cascade rate itself are comparable and
635: estimated from Eq.\ (\ref{noisy-knls}) to be
636: $\gamma_{\rm NL}\sim\omega_A|N_2|\widetilde{|b|^2}$. Note, although the rate
637: of the GS cascade, $\gamma_{\rm GS}\propto k_\|$, increases
638: with $k_\|$, so does the energy dissipation rate. Thus, the total energy
639: dissipated in the cascade is determined by the width of the ``inertial range,''
640: $R^{-1}\lesssim k\lesssim\rho_p^{-1}$. The energy dissipation rate is, by
641: definition,
642: \beq
643: \frac{d\epsilon}{dt}=\frac{d\epsilon}{dk_\|}\frac{dk_\|}{dt}
644: =-\gamma_{\rm NL}\,\epsilon .
645: \eeq
646: Using the definition for the total energy cascade rate,
647: $dk_\|/dt=(\gamma_{\rm GS}+\gamma_{\rm NL})k_\|$, and integrating the
648: resultant equation, we obtain
649: \beq
650: \frac{\ln\epsilon/\epsilon_0}{\ln k_{\|m}/k_{\|0}}
651: =-\frac{|N_2|\widetilde{|b|^2}}{|N_2|\widetilde{|b|^2}+1} ,
652: \eeq
653: where $\epsilon_0$ and $k_{\|0}$ are initial values. The largest scale is
654: set by the size of the accretion flow, $k_{\|0}\simeq R^{-1}$. The smallest
655: scale is estimated from the scaling, $k_\|\simeq k_\bot^{2/3}R^{-1/3}$, of the
656: GS cascade with $k_{\bot m}\sim\rho_p^{-1}$.\footnote{
657: Note, here we set a lower limit on $k_{\|m}$. Since both
658: cascades proceed independently, $k_{\|m}$ may only be larger
659: (i.e., closer to $\rho_p^{-1}$) due to the parallel cascade.}
660: Thus, $k_{\|m}/k_{\|0}\simeq (R/\rho_p)^{2/3}$.
661: \citet{QG98} argue that much of the energy dissipated at
662: $k_\bot\sim\rho_p^{-1}$ may heat electrons, depending on the value of $\beta$
663: and numerical constants estimated from numerical simulations which are not
664: very well constrained. We assume here the {\em worst} possible case in which
665: {\em all the wave energy} reaching the Larmor scale heats electrons.
666:
667: For the electron heating to dominate by the compressible cascade, the amount of
668: energy that reaches the Larmor scale must be smaller than the fraction of
669: energy that is dissipated on electrons in the compressible cascade:
670: \beq
671: \epsilon(k_{\|m})/\epsilon_0 < \delta_\|,
672: \eeq
673: where $\delta$ is given by Eq.\ (\ref{delta-limit}). Combining all
674: together, this inequality may now be re-written as follows:
675: \beq
676: \widetilde{|b|^2}>-\frac{1}{|N_2|}\,\frac{\frac{3}{2}\ln\delta_\|/\ln(R/\rho_p)}
677: {1+\frac{3}{2}\ln\delta_\|/\ln(R/\rho_p)} .
678: \eeq
679: The nonlinear coupling to dissipation, $N_2$, depends on plasma parameters.
680: The dependence of $|N_2|$ vs. $\beta$ for two values of the
681: proton-to-electron temperature ratio: $T_p/T_e=1,\ 10^3$
682: is shown in Fig.\ \ref{fig-rates}. Clearly, it weakly
683: depends on $T_p/T_e$ and is roughly $\propto\sqrt\beta$. To estimate the lower
684: bound on $|b|$, we take the {\em lowest} value of $|N_2|\sim1$ for
685: $\beta\sim3$, i.e., near the bifurcation value. The normalized amplitude may
686: be written in a general case as follows (see \S \ref{S:MODEL}):
687: \begin{eqnarray}
688: \widetilde{|b|^2}&=&(\tilde B + B_0\,k_\bot/k)^2/B_0^2-
689: \langle(\tilde B + B_0\,k_\bot/k)^2/B_0^2\rangle \nonumber\\
690: &=&(\tilde B/B_0)^2-\langle(\tilde B/B_0)^2\rangle
691: +2(\tilde B/B_0)k_\bot/k \nonumber\\
692: &\sim&2(\tilde B/B_0)
693: \end{eqnarray}
694: for $\tilde B\lesssim B_0$ and an oblique and nearly perpendicular
695: angle of propagation,
696: $k_\bot\sim k$. We take [see Eqs.\ (\ref{delta-limit}) and (\ref{adaf})]
697: $\delta\lesssim2.5\times10^{-2}$ and $\rho_p/R\gtrsim6\times10^{-9}$. Then the
698: conservative limit on the fluctuation amplitude of MHD turbulence at which the
699: electron heating is dominated by the compressible cascade is readily estimated
700: as
701: \beq
702: \frac{\tilde B}{B_0}\gtrsim0.2\, .
703: \eeq
704: Thus, for $\tilde B/B_0\sim0.2$, the electron heating due to the parallel,
705: $\delta_\|$, and perpendicular, $\delta_\bot$, cascades are of comparable
706: value and the total electron heating is $\delta=\delta_\|+\delta_\bot\sim0.05$.
707: Hence, for the levels of turbulence which are believed to exist in ADAFs,
708: $\tilde B/B_0\sim1$, the heating of electrons is dominated by the compressible
709: cascade by at least an order of magnitude and yields $\delta\lesssim0.025$,
710: provided $\beta>\beta_{\rm crit}\simeq2.6$.
711:
712:
713: \subsection{Parallel vs. NB cascade \label{SSS:NB} }
714:
715: In a similar way we consider competition of the NB and parallel cascades.
716: The NB cascade \citep{NB96} proceeds in $k_\bot$ only, hence it is totally
717: independent of $k_\|$. There is no dissipation during the NB cascade.
718: The Alfv\'en wave energy may be damped only when it reaches the
719: $k_\bot\sim\rho_p^{-1}$
720: scale and, as discussed in \S \ref{SSS:RATES-PAR}, will predominantly heat
721: the electrons. To estimate the energy lost via the NB cascade, it is sufficient
722: to compare the rate, $\gamma_{\rm NB}(\rho_p^{-1})$, of the NB cascade at
723: $k_\bot\sim\rho_p^{-1}$ to the maximum damping rate in the $k_\|$ cascade,
724: $\gamma_{\rm NL,max}$. The latter is $\gamma_{\rm NL,max}\sim
725: \omega_{A,max}|N_2|\widetilde{|b|^2}\sim v_A\rho_p^{-1}|N_2|\widetilde{|b|^2}$.
726: The former is estimated from \citet{NB96} as follows. The energy transfer rate
727: is
728: \beq
729: \dot\epsilon\sim v^2/(N\tau)\sim\textrm{ constant},
730: \eeq
731: where $v^2\sim E_kk$ is the energy per unit mass, $N\sim(\delta v/v)$ is the
732: number of interactions, $\tau\sim(\kappa_\|v_A)^{-1}$ is the interaction time,
733: $\delta v$ is the perturbation, and $\kappa_\|$is a longitudinal scale of
734: interacting Alfv\'en wave packets. Therefore,
735: \begin{eqnarray}
736: \gamma_{\rm NB}&\sim&(N\tau)^{-1}\sim\dot\epsilon/v^2\sim\dot\epsilon(E_kk)^{-1}
737: \nonumber\\
738: &\sim&(E_{k0}k_0)^{-1}\left(\frac{k}{k_0}\right)^{1/2}\sim
739: \left(\frac{\delta v_0}{v_0}\right)^2\kappa_{\|0}v_A,
740: \end{eqnarray}
741: where we used the NB scaling $E_k\propto k^{-3/2}$ and the subscript ``0''
742: denotes quantities at the outer scale of the turbulence. Assuming strong
743: turbulence, $\delta v_0\sim v_0$, we obtain an upper limit on $\gamma_{\rm NB}$.
744: Assuming also that on the outer scale $k_0\sim\kappa_0\sim\kappa_{\|0}\sim
745: \kappa_{\bot0}\sim R^{-1}$, we have
746: \beq
747: \gamma_{\rm NB}(\rho_p^{-1})\sim
748: \left.v_A(\kappa_0 k)^{1/2}\right|_{k\sim\rho_p^{-1}}\sim
749: v_A(R\rho_p)^{-1/2}.
750: \eeq
751: Thus, the ratio is
752: \beq
753: \frac{\gamma_{\rm NB}(\rho_p^{-1})}{\gamma_{\rm NL,max}}\sim
754: \left(|N_2|\widetilde{|b|^2}\right)^{-1}\left(\frac{\rho_p}{R}\right)^{1/2}
755: \sim10^{-4} .
756: \eeq
757: Thus, even if all the energy cascaded in $k_\bot$ heats the electrons,
758: it constitutes about $10^{-4}$ of the total energy dissipated on the protons
759: (unless $\tilde B/B_0\ll1$), i.e., is completely negligible compared to
760: equation (\ref{delta-limit}). Note that the above result means only that
761: the NB cascade is ``slower'' than the parallel cascade, while the total
762: wave energy in $k_\bot$ may be non-negligible.
763:
764:
765: \subsection{Proton-cyclotron damping vs. perpendicular cascade
766: \label{SSS:RATES-CYC} }
767:
768: The nonlinear interaction of finite-amplitude wave-packets results in
769: generation of high-$k$ harmonics, as discussed in \S \ref{S:MODEL}. Such a
770: cascade transfers wave energy along the local magnetic field to high $k_\|$,
771: where the cyclotron resonance, $\omega-k_\|v-\Omega_p=0$, may be satisfied and
772: the cyclotron damping on (bulk) protons becomes very efficient. This effect was
773: completely missing from previous studies because both the GS and NB cascades
774: are (nearly) perpendicular; $k_\|$ is always small, so that only a very
775: small number of particles from the tail of a particle distribution function
776: are in the resonance. The proton-cyclotron damping is known to heat protons
777: only, no energy goes to electrons, since $\Omega_p\ll\Omega_e$ and electrons
778: are off resonance. It is, however, difficult to rigorously estimate the
779: $\widetilde{|b|^2}$, because for the cyclotron damping to dominate, it must be
780: faster than the typical GS or NB cascading time, i.e.,
781: $\gamma_c\gtrsim\omega_A$. Such a situation may occur only when
782: $\omega_A\simeq\Omega_p$, so that (i) Alfv\'en waves are heavily damped and
783: (ii) they may convert to proton cyclotron waves, which greatly complicates a
784: rigorous analysis.
785:
786: One can, however, roughly estimate the fluctuation level at which the
787: proton-cyclotron damping is so strong that most of the wave energy dissipates
788: on protons and does not convert into whistlers. The parallel cascade
789: proceeds until it hits a scale where the nonlinear steepening is
790: balanced by the wave dispersion [last term in Eq.\ (\ref{noisy-knls})]. Thus,
791: the maximum parallel wave-number to which wave energy cascades in the
792: compressible cascade is readily estimated from Eq.\ (\ref{noisy-knls}) to be
793: ${k_\|}_{\rm max}\sim\widetilde{|b|^2}(2\Omega_p/v_A|N_{1,2}|)$,
794: where $|N_{1,2}|$ is the maximum of $|N_1|$ and $|N_2|$. The proton-cyclotron
795: damping is strong when $\omega_A=k_\|v_A\simeq\Omega_p$. Thus, if
796: ${k_\|}_{\rm max}\gtrsim\Omega_p/v_A$, the MHD turbulence heats protons only.
797: This occurs when
798: \beq
799: \frac{\tilde B}{B_0}\gtrsim\frac{1}{4|N_{1,2}|} ,
800: \eeq
801: where we again took $\widetilde{|b|^2}\simeq2(\tilde B/B_0)$. Here
802: $|N_{1,2}|\sim1$ for $\beta\sim3$ and increases with $\beta$ and
803: weakly with $T_p/T_e$ as shown in Fig.\ \ref{fig-rates}. For a range of
804: $\beta$'s, $1\lesssim\beta\lesssim2.6$, i.e., above the bifurcation threshold
805: (dissipation in the compressible cascade is weak), but below equipartition,
806: the conservative condition on $\tilde B$ reads:
807: \beq
808: \frac{\tilde B}{B_0}\gtrsim0.3\, .
809: \eeq
810:
811:
812: \subsection{Is nonlinear trapping important?}
813:
814: The parallel (coherent) cascade of the wave energy occurs due to the formation
815: and nonlinear evolution of coherent wave structures, such as shocks, Alfv\'enic
816: discontinuities, solitary wave-packets, etc., as discussed in
817: \S \ref{S:MODEL}. The resonant particles, i.e., those which take the energy
818: from waves and result in damping, turn out to be trapped in the wave potential
819: created by the wave magnetic field. Those resonant particles which move a little
820: slower than the wave will be reflected from the rear part of the wave potential,
821: gain energy, and start to move faster than the wave. The particles moving
822: faster than the wave will be reflected from the front part of the wave
823: potential and, correspondingly, give energy to the wave and decelerate. Upon a
824: half bounce time, $\tau_{\rm b}/2$, the particles which have been reflected
825: from the rear of the potential cross though it and reach its front part,
826: where they are reflected back and give their energy to the wave. The slow
827: particles do the opposite at the rear part. Such a process may repeat over
828: and over again, provided the wave amplitude is maintained constant by an
829: external source (e.g., the Balbus-Hawley instability in ADAFs),
830: to prevent particles from escape. Clearly, the number of slow ($v<v_A$)
831: and fast ($v>v_A$) particles will oscillate with time, so that the wave
832: damping rate will change in time too and may become negative
833: (i.e., the wave amplitude will grow) during some periods of time.
834: The bounce period of a particle depends on its energy, because a wave trapping
835: potential is, in general, anharmonic. Due to the difference in the bounce
836: periods, groups of particles with different energies gain a phase shift which
837: grows with time. Upon many cycles (bounces), the particles near the resonance,
838: $v\sim v_A$, become completely randomized in phase and form an equilibrium,
839: {\em plateau} distribution. At this moment, the wave collisionless dissipation
840: quenches.\footnote{This process does not affect the cyclotron damping on
841: protons, which operates at the cyclotron, and not Cherenkov, resonance.}
842: The process described above is referred to as the nonlinear
843: Landau damping \citetext{see \citealt{Metal98} for more details}.
844: Numerical simulations of the process show that it takes
845: $\sim10^2-10^3\tau_{\rm b}$ or even longer for particles to phase mix and
846: quench collisionless damping. We note here that the nonlinear Landau damping is
847: significant for large amplitude waves $\tilde B/B_0\sim1$, for which the
848: number of trapped particles may be large.
849:
850: We now estimate whether particle trapping is important for the physics of
851: ADAFs and, in particular, whether the Alfv\'en wave damping during the parallel
852: (compressible) cascade in ADAFs quenches. First,
853: the kinetic energy of a particle in a potential is of order its potential
854: energy, so that the typical velocity of a particle of mass $m_j$
855: ($j=e,p$) is $v_j\sim(\tilde B^2/8\pi m_j n)^{1/2}\sim
856: {v_A}_j(\tilde B/B_0)$. The typical scale of the potential, $\lambda$,
857: is set by the scale at which the energy is pumped into the system, i.e.,
858: the scale of the Balbus-Hawley instability, $\lambda\sim k_{BH}^{-1}\sim R$.
859: Thus the characteristic oscillation period of a particle (the bounce time)
860: at a given radial position in the accretion flow is
861: $\tau_{\rm b}\sim\lambda/v_j$, which yields
862: \beq
863: {\tau_{\rm b}}_p\simeq R/(v_A\tilde B/B_0), \qquad
864: {\tau_{\rm b}}_e=(m_e/m_p)^{1/2}\,{\tau_{\rm b}}_p
865: \eeq
866: for protons and electrons, respectively.
867: Second, accreting gas flows into the central object and, thus, continuously
868: replenished with a fresh material on the outer edge of the ADAF.
869: The typical infall time at a given radius $R$ is
870: \beq
871: \tau_{\rm r}\simeq R/v_r ,
872: \eeq
873: where $v_r$ is the radial velocity of a gas given by Eqs.\ (\ref{adaf}).
874: Estimating the number of bounces of a particle of species $j$ in an
875: Alfv\'en wave at a radial position $R$ as
876: ${\cal N}_j(R)\sim\tau_{\rm r}/{\tau_{\rm b}}_j$, we obtain
877: \beq
878: {\cal N}_p\simeq0.4\alpha(1+\beta)^{1/2}(\tilde B/B_0), \qquad
879: {\cal N}_e\simeq43{\cal N}_p .
880: \eeq
881: Note that ${\cal N}$ is independent of radius. The above equations show that
882: the protons in the accretion flow are always in the linear regime of damping
883: (no trapping, ${\cal N}_p\lesssim1$). The electrons may experience
884: up to a few tens of bounces
885: within the infall time, which is probably not enough to quench damping on
886: electrons at all, but still may lower its value a little. We thus conclude
887: that the damping process in the compressible parallel cascade is hardly
888: affected by nonlinear particle trapping and the value of
889: $\delta\sim{\rm few}\times10^{-2}$ represents a conservative upper
890: limit on the the electron to proton heating ratio in hot, advection-dominated
891: accretion flows with strong magnetic field turbulence, $\tilde B\sim B_0$.
892:
893:
894: \section{Conclusions \label{S:CONCL} }
895:
896: The particle heating is one of the key issues for the advection-dominated
897: accretion flow models. All the models assume that protons receive most of the
898: energy released due to viscous (turbulent) heating of the accretion gas, while
899: electrons remain relatively cold; hence the low luminosities of ADAFs.
900: In the collisionless plasma of the ADAFs, particle heating is determined
901: solely by excitation and dissipation of plasma collective motions (waves).
902: The instability of Balbus \& Hawley which is believed to operate and produce
903: magnetic fields in such differentially rotating flows may naturally result
904: in strong MHD turbulence with highly fluctuating magnetic fields,
905: $\tilde B\sim B_0$, during the nonlinear stages of its evolution. A detailed
906: analysis of particle heating by such turbulence is presented in this paper.
907:
908: We show that in the most natural case of the ADAF parameters,
909: $\beta\gtrsim3$ and $\tilde B/B_0\gtrsim0.2$, dissipation of the magnetic
910: energy occurs predominantly in the parallel, compressible cascade (the effect
911: absent for linear, low-amplitude Alfv\'en waves). Most of the dissipated energy
912: heats protons, while electrons receive only $\delta\sim{\rm few}\%$ of the
913: energy, which in agreement with the empirical values inferred from the spectral
914: fits for various, low-luminosity accreting systems.
915: If, alternatively, the magnetic field in ADAFs is close to equipartition,
916: $\beta\lesssim3$, the energy of MHD turbulence is dissipated only on protons
917: via the cyclotron damping mechanism, provided the amplitude of the turbulence
918: is somewhat higher, $\tilde B/B_0\gtrsim{\rm few}\times10^{-1}$. The electron
919: heating is negligible in this case. At lower amplitudes, however, the
920: nonlinear effects are weak and the results of the linear analysis for
921: the dissipation of the linear Alfv\'en waves, addressed by other authors, hold.
922:
923:
924: \acknowledgements
925: The author is grateful to Ramesh Narayan for his interest in this work and
926: helpful discussions, to the referee Amitava Bhattacharjee for many insightful
927: comments and suggestions which helped to greatly improve the manuscript
928: and stimulated further work, and to Eliot Quataert for discussions.
929: This work was supported by NASA grant NAG~5-2837 and NSF grant PHY~9507695.
930:
931: \begin{appendix}
932: \section{Exact formulae for the nonlinear coefficients $N_1$ and $N_2$}
933:
934: The nonlinear coefficients in Eq.\ (\ref{noisy-knls}) has been calculated from
935: the full kinetic treatment by solving the Vlasov equation for arbitrary wave
936: profile and determining the wave-induced plasma density perturbation
937: \citep{Spangler89,Spangler90}. They are
938: \beq
939: N_1=M_1-m_1, \qquad N_2=M_2-m_2 ,
940: \eeq
941: where $M_1$ and $M_2$ are the isotropic pressure contributions and
942: $m_1$ and $m_2$ are the contributions due to pressure anisotropy
943: (i.e., the temperatures and, hence, pressures may be different along the
944: ambient magnetic field and perpendicular to it, $p_\|\not=p_\bot$):
945: %
946: \begin{eqnarray}
947: M_1&=&-\frac{1}{4}
948: \left(f_7+\frac{f_5\left(f_1f_2-\pi f_3f_4\right)
949: +\pi f_6\left(f_3f_2+f_1f_4\right)}{\left(f_2^2+\pi f_4^2\right)}\right), \\
950: M_2&=&-\frac{\sqrt{\pi}}{4}
951: \left(f_6+\frac{f_6\left(f_1f_2-\pi f_3f_4\right)
952: -f_5\left(f_3f_2+f_1f_4\right)}{\left(f_2^2+\pi f_4^2\right)}\right), \\
953: m_1&=&-\frac{1}{4X_p^2}
954: \left(g_1+\frac{g_4\left(f_1f_2-\pi f_3f_4\right)
955: +\pi g_3\left(f_3f_2+f_1f_4\right)}{2\left(f_2^2+\pi f_4^2\right)}\right), \\
956: m_2&=&-\frac{\sqrt{\pi}}{4X_p^2}
957: \left(g_2+\frac{g_3\left(f_1f_2-\pi f_3f_4\right)
958: -g_4\left(f_3f_2+f_1f_4\right)}{2\left(f_2^2+\pi f_4^2\right)}\right),
959: \end{eqnarray}
960: %
961: where $\sqrt{\pi}$ appears in numerators of $M_2$ and $m_2$ because the
962: Hilbert operator carries an extra ($1/\pi$) and additional (1/2) in all four
963: coefficients absorbs it from the wave equation, as compared to
964: \citet{Spangler89,Spangler90}.
965: The coefficients $f_1,\ f_2, \dots,\ f_7, g_1, \dots,\ g_4$ are defined as
966: %
967: \begin{eqnarray}
968: f_1&=&-X_pZ_R(X_p)+X_eZ_R(X_e), \label{f's} \\
969: f_2&=&[1+X_pZ_R(X_p)]+T_r[1+X_eZ_R(X_e)], \\
970: f_3&=&X_pe^{-X_p^2}-X_ee^{-X_e^2}, \\
971: f_4&=&X_pe^{-X_p^2}+T_rX_ee^{-X_e^2}, \\
972: f_5&=&1+X_pZ_R(X_p), \\
973: f_6&=&X_pe^{-X_p^2}, \\
974: f_7&=&X_pZ_R(X_p),
975: \end{eqnarray}
976: %
977: \begin{eqnarray}
978: g_1&=&[X_p^2+X_p^3Z_R(X_p)-X_pZ_R(X_p)]
979: +(1/T_r)[X_e^2+X_e^3Z_R(X_e)-X_eZ_R(X_e)], \\
980: g_2&=&(X_p^3-X_p)e^{-X_p^2}+(1/T_r)(X_e^3-X_e)e^{-X_e^2}, \\
981: g_3&=&(2X_p^3-X_p)e^{-X_p^2}-(2X_e^3-X_e)e^{-X_e^2}, \\
982: g_4&=&[2X_p^2+2X_p^3Z_R(X_p)-X_pZ_R(X_p)]-[2X_e^2+2X_e^3Z_R(X_e)-X_eZ_R(X_e)].
983: \label{g's}
984: \end{eqnarray}
985: %
986: Here $Z_R(X)$ is the real part of the plasma dispersion function
987: for argument $X$:
988: \beq
989: Z(X)=\frac{1}{\sqrt{\pi}}
990: \int_{-\infty}^\infty\frac{\exp(-\xi^2)}{\xi-X}\;{\rm d}\xi
991: =2i\,e^{-X^2}\int_{-\infty}^{iX}e^{-\xi^2}{\rm d}\xi .
992: \label{plasma-disp}
993: \eeq
994: The imaginary part of $Z(X)$ appears explicitly in Eq.\ (\ref{noisy-knls})
995: as the Hilbert integral operator. Physically, $X=X_R+iX_I$ is the ratio of
996: wave phase velocity to thermal velocity. For Alfv\'en waves, $X_p$ and $X_e$
997: become:
998: \beq
999: X_p=\left(1+\frac{T_e}{T_p}\right)^{1/2}\frac{1}{\sqrt\beta}, \qquad
1000: X_e=\left(\frac{T_p}{T_e}+1\right)^{1/2}\sqrt{\frac{m_e}{m_p}}\;
1001: \frac{1}{\sqrt\beta}
1002: \eeq
1003: At last, $T_r=T_p/T_e$ is the proton-to-electron temperature ratio and
1004: $m_e/m_p\simeq1/1836$ is the electron-to-proton mass ratio.
1005: \end{appendix}
1006:
1007:
1008: \begin{thebibliography}{DUM}
1009: %
1010: \bibitem[Abramowitz et al.(1995)]{Abramowitcz-etal95} %------*-ADAF model
1011: Abramowitcz, M., Chen, X., Granath, M., \& Lasota, J. P. 1996, \apj, 471, 762
1012: %
1013: \bibitem[Balbus \& Hawley(1991)]{BalbusHawley91} %-------Balbus&Hawley instab.
1014: Balbus, S. A., \& Hawley, J. F. 1991, \apj, 376, 214
1015: %
1016: \bibitem[Balbus \& Hawley(1998)]{BH98} %-------Balbus&Hawley review
1017: Balbus, S. A., \& Hawley, J. F. 1998, Rev. Mod. Phys., 70, 1
1018: %
1019: \bibitem[Begelman \& Chiueh(1988)]{BC88} %------drift waves in ADAFs
1020: Begelman, M. C., \& Chiueh, T. 1988, \apj, 332, 872
1021: %
1022: \bibitem[Bisnovatyi-Kogan \& Lovelace(1997)]{BkL97} %------turb.heatng.ADAF
1023: Bisnovatyi-Kogan, G. S., \& Lovelace, R. V. E. 1997, \apj, 486, L43
1024: %
1025: \bibitem[Blandford \& Begelman(1999)]{BB98} %------ADAF => winds
1026: Blandford, R. D. \& Begelman, M. C. 1999, \mnras, 303, L1
1027: %
1028: \bibitem[Cho \& Vishniac(2000)]{Vishniac00} %----Gold-Sridhar numerical
1029: Cho, J., \& Vishniac, E. T. 2000, \apj, accepted (astro-ph/0003403)
1030: %
1031: \bibitem[Cohen \& Kulsrud(1971)]{CohenKulsrud71} %------NL Alfven
1032: Cohen, R. H., \& Kulsrud, R. M. 1971, Phys Fluids, 17, 2215
1033: %
1034: \bibitem[Di Matteo et al.(1999)]{DMetal98} %------
1035: Di Matteo, T., Fabian, A. C., Rees, M. J., Carili, C. L., \& Ivison, R. J.
1036: 1999, \mnras, 305, 493
1037: %
1038: \bibitem[Gammie, Narayan, \& Blandford(1999)]{GNB98} %----NGC4258 ADAFspect.
1039: Gammie, C. F., Narayan, R., \& Blandford, R. 1999, \apj, 516, 177
1040: %
1041: \bibitem[Gruzinov \& Quataert(1999)]{GQ99}
1042: Gruzinov, A., \& Quataert, E. 1999, \apj, 520, 849
1043: %
1044: \bibitem[Goldreich \& Sridhar(1995)]{GS95} %------GS turbulence
1045: Goldreich, S., \& Sridhar, S. 1995, \apj , 438, 763
1046: %
1047: \bibitem[Goldreich \& Sridhar(1997)]{GS97}
1048: Goldreich, S., \& Sridhar, S. 1997, \apj, 485, 680
1049: %
1050: \bibitem[Hameury et al.(1997)]{Hameury-etal97} %-----BH bin.ADAFspect.
1051: Hameury, J. M., Lasota, J. P., McClintock, J. E., \& Narayan, R. 1997,
1052: \apj, 489, 234
1053: %
1054: \bibitem[Ichimaru(1977)]{Ichimaru77} %------*-ADAF first
1055: Ichimaru, S. 1977, \apj, 214, 840
1056: %
1057: \bibitem[Lasota et al.(1996)]{Lasota-etal96} %------NGC4258 ADAF sp.
1058: Lasota, J. P., Abramowitcz, M., Chen, X., Krolik, J., Narayan, R., \& Yi, I.
1059: 1996, \apj, 462, 142
1060: %
1061: \bibitem[Mahadevan(1997)]{Mahadevan97} %------
1062: Mahadevan, R. 1997, \apj, 477, 585
1063: %
1064: \bibitem[Mahadevan(1998)]{Mahadevan98} %------
1065: Mahadevan, R. 1998, Nature, 394, 651
1066: %
1067: \bibitem[Manmoto et al.(1997)]{Manmoto-etal97} %------Sgr A* ADAF spec
1068: Manmoto, T., Mineshige, S., \& Kukunose, M. 1997, \apj, 489, 791
1069: %
1070: \bibitem[Medvedev(1999)]{KNLSreview} %------KNLS-review
1071: Medvedev, M. V. 1999, Phys. Plasmas, 6, 2191
1072: %
1073: \bibitem[Medvedev et al.(1998)]{Metal98} %------trapping
1074: Medvedev, M. V., Diamond, P. H., Rosenbluth, M. N., \& Shevchenko V. I.,
1075: 1998, \prl, 81, 5824
1076: %
1077: \bibitem[Medvedev \& Diamond(1997)]{MD97} %------noisy-KNLS
1078: Medvedev, M. V., \& Diamond, P. H. 1997, \pre, 56, R2371
1079: %
1080: \bibitem[Mj\o lhus \& Wyller(1988)]{MW88} %------KNLS coefficients
1081: Mj\o lhus, E., \& Wyller, J. 1988, J. Plasma Phys., 40, 299
1082: %
1083: \bibitem[Narayan, Barret, \& McClintock(1997)]{NBMc97} %------BH bin.ADAF spec
1084: Narayan, R., Barret, D., \& McClintock, J. E. 1997, \apj, 482, 448
1085: %
1086: \bibitem[Narayan et al.(1998)]{Narayan-etal98} %------Sgr A* ADAF spec
1087: Narayan, R., Mahadevan, R., Grindlay, J. E., Popham, R. G., \& Gammie, C. F.
1088: 1998, \apj, 492, 554
1089: %
1090: \bibitem[Narayan, Mahadevan, \& Quataert(1998)]{NMQ98} %------>>-ADAF-review
1091: Narayan, R., Mahadevan, R., \& Quataert, E. 1998,
1092: in {\it The Theory of Black Hole Accretion Discs},
1093: eds. M. A. Abramowitz, G. Bjornsson, and J. E. Pringle
1094: (Cambridge: Cambridge University Press)
1095: %
1096: \bibitem[Narayan \& Yi(1995)]{NY95} %------*-ADAF model
1097: Narayan, R., \& Yi, I. 1995 \apj, 452, 710.
1098: %
1099: \bibitem[Narayan, Yi, \& Mahadevan(1995)]{NYM95} %------
1100: Narayan, R., Yi, I., \& Mahadevan, R. 1995, Nature, 374, 623
1101: %
1102: \bibitem[Ng \& Bhattacharjee(1996)]{NB96}
1103: Ng, C.S., \& Bhattacharjee, A. 1996, \apj, 465, 845
1104: %
1105: \bibitem[Oraevskii(1983)]{OsnFizPlazm}
1106: Oraevskii, B. N. 1983, in Basic plasma physics, Vol. 1,
1107: ed. A. A. Galeev \& R. N. Sudan (New York : North-Holland Pub.), 241
1108: %
1109: \bibitem[Quataert(1998)]{Quataert98} %------heating Alfven
1110: Quataert, E. 1998, \apj, 500, 978
1111: %
1112: \bibitem[Quataert \& Gruzinov(1999)]{QG98} %------heating Alfven
1113: Quataert, E., \& Gruzinov, A. 1999, \apj, 520, 248
1114: %
1115: \bibitem[Quataert \& Narayan(1999)]{QN98} %------ADAF with winds
1116: Quataert, E., \& Narayan, R. 1999, \apj, 520, 849
1117: %
1118: \bibitem[Rees et al.(1982)]{Rees-etal82} %------*-ADAF model
1119: Rees, M. J., Begelman, M. C., Blandford, R. D., \& Phinney, E. S. 1982,
1120: Nature, 295, 17
1121: %
1122: \bibitem[Reynolds et al.(1996)]{Reynoldsetal96} %------M87
1123: Reynolds, C. S., Di Matteo, T., Fabian, A. C., Hwang, U., \& Canizares, C. R.
1124: 1996, \mnras, 283, L111
1125: %
1126: \bibitem[Spangler(1989)]{Spangler89} %------KNLS coefficients
1127: Spangler, S. R. 1989, Phys. Fluids B, 1, 1738
1128: %
1129: \bibitem[Spangler(1990)]{Spangler90} %------KNLS coefficients
1130: Spangler, S. R. 1989, Phys. Fluids B, 2, 407
1131: %
1132: %\bibitem[]{} %------
1133: %
1134:
1135: %\bibitem[Sridhar \& Goldreich (1994)]{GS1}
1136: %Sridhar, S. \& Goldreich 1994, \apj, 432, 612
1137: %
1138: %\bibitem[Narayan \& Yi 1994]{NY94} %------*-ADAF model
1139: %Narayan, R., \& Yi, I. 1994 \apj , 428, L13.
1140: %
1141: %\bibitem[Narayan, McClintock, \& Yi 1996]{NMcY96} %------BH bin.ADAF spec
1142: %Narayan, R., McClintock, J. E., \& Yi, I. 1996, \apj, 457, 821
1143: %
1144: \end{thebibliography}
1145:
1146: \figcaption[adaf-state-elheat.eps]{A $T_p/T_e$-$\beta$-diagram of state of
1147: nonlinear Alfv\'enic turbulence. The shaded region corresponds to the
1148: compressional cascade dominated regime (phase) of turbulence. The contour lines
1149: represent the electron-to-proton heating ratio, $\delta$, in MHD turbulence in
1150: ADAFs with $\tilde B/B_0\gtrsim0.1$. The unshaded region corresponds to the
1151: small-scale dissipation dominated phase, where the Goldreich-Sridhar cascade
1152: is dominant. The boldfaced part of the dividing (bifurcation) line
1153: corresponds to the range of the $T_p/T_e$ parameter typical of ADAFs.
1154: \label{fig-state} }
1155: %
1156: \figcaption[adaf-elheat.eps]{The electron-to-proton heating ratio vs. $T_p/T_e$
1157: for $\beta=3$ and $\beta=40$. \label{fig-heat} }
1158: %
1159: \figcaption[adaf-rates.eps]{The coefficients of nonlinear damping, $|N_2|$
1160: (solid curves), and nonlinear steepening, $N_1$ (dashed curves), vs. $\beta$
1161: for two values of the temperature ratio,\ $T_p/T_e=10^3$ (boldface curves)
1162: and $T_p/T_e=1$ (thin curves). \label{fig-rates} }
1163:
1164: \vfill
1165: \plottwo{f1.eps}{f2.eps}~ \\ ~\\
1166: Figs. \ref{fig-state},\ref{fig-heat}\vskip1cm
1167: \plottwo{f3.eps}{empty.eps}~ \\ ~\\
1168: Fig. \ref{fig-rates}
1169:
1170: %\newpage
1171: %\plotone{adaf-state-elheat.eps}\vskip0.3cm Fig. \ref{fig-state}\vskip1cm
1172: %\plotone{adaf-elheat.eps}\vskip0.3cm Fig. \ref{fig-heat}\vskip1cm
1173: %\plotone{adaf-rates.eps}\vskip0.3cm Fig. \ref{fig-rates}
1174:
1175: \end{document}
1176:
1177:
1178: