0705.0084/ms.tex
1: \documentclass[12pt,preprint]{aastex}
2: \newcommand\simgt{\lower.5ex\hbox{$\; \buildrel > \over \sim \;$}}
3: \newcommand\simlt{\lower.5ex\hbox{$\; \buildrel < \over \sim \;$}}
4: 
5: \newcommand{\kmpers}{\mathrm{km\ s^{-1}}}
6: \newcommand{\di}{\item[--]}
7: \newcommand{\Fig}[1]{Figure \ref{#1}}
8: \newcommand{\Eq}[1]{Eq. [\ref{#1}]}
9: \newcommand{\vct}[1]{\mbox{\boldmath{$#1$}}}
10: \newcommand{\mach}{\mathcal{M}}
11: \newcommand{\step}{\mathcal{H}}
12: \newcommand{\D}{\mathcal{D}}
13: \newcommand{\I}{\mathcal{I}}
14: \newcommand{\ext}{\mathrm{ext}}
15: \newcommand{\rmin}{r_\mathrm{min}}
16: \newcommand{\rmax}{r_\mathrm{max}}
17: \newcommand{\cs}{c_\mathrm{s}}
18: \newcommand{\tR}{\tilde{R}}
19: \newcommand{\tx}{\tilde{x}}
20: \newcommand{\tz}{\tilde{z}}
21: \newcommand{\sMsml}{(\mach^2-1)^{1/2}}
22: \newcommand{\sMsRsml}{(\mach^2\tR^2-1)^{1/2}}
23: \newcommand{\IB}{\textit{IB}}
24: \newcommand{\IM}{\textit{IM}}
25: \newcommand{\OM}{\textit{OM}}
26: 
27: \shorttitle{DYNAMICAL FRICTION OF CIRCULAR-ORBIT PERTURBER}
28: \shortauthors{KIM \& KIM}
29: 
30: 
31: \begin{document}
32: \title{Dynamical Friction of a Circular-Orbit Perturber in a Gaseous Medium}
33: \author{Hyosun Kim and Woong-Tae Kim }
34: \affil{Department of Physics and Astronomy, FPRD, Seoul National University, Seoul 151-742, Korea}\email{hkim@astro.snu.ac.kr,  wkim@astro.snu.ac.kr}
35: 
36: \begin{abstract}
37: We investigate the gravitational wake due to, and dynamical friction on, a 
38: perturber moving on a circular orbit in a uniform gaseous medium
39: using a semi-analytic method.  This work is a straightforward extension
40: of \citet{ost99} who studied the case of a straight-line trajectory.  
41: The circular orbit causes the bending of the wake in the background 
42: medium along the orbit, forming a long trailing tail.  
43: The wake distribution is thus asymmetric, giving rise to the drag
44: forces in both opposite (azimuthal) and lateral (radial) directions 
45: to the motion of the perturber, although the latter does not contribute to 
46: orbital decay much.  
47: For subsonic motion, the density wake with a weak tail is simply a curved 
48: version of that in Ostriker and does not exhibit the front-back symmetry.
49: The resulting drag force in the opposite direction is
50: remarkably similar to the finite-time, linear-trajectory counterpart.
51: On the other hand, a supersonic perturber is able to overtake its
52: own wake, possibly multiple times, and develops a very pronounced tail.
53: The supersonic tail surrounds the perturber in a trailing spiral fashion,
54: enhancing the perturbed density 
55: at the back as well as far front of the perturber.
56: We provide the fitting formulae for the drag forces as functions of the
57: Mach number, whose azimuthal part is surprisingly in good agreement 
58: with the Ostriker's formula,
59: provided $V_p t=2R_p$, where $V_p$ and $R_p$ are the velocity 
60: and orbital radius of the perturber, respectively.
61: \end{abstract}
62: \keywords{hydrodynamics --- galaxies : kinematics and dynamics --- ISM: general --- shock waves}
63: 
64: 
65: 
66: \section{INTRODUCTION}
67: 
68: Dynamical friction refers to momentum loss suffered by a massive object 
69: moving through a background medium due to its gravitational interaction with 
70: its own induced wake.  The gravitational drag force removes angular 
71: momentum from an object in orbital motion, causing it to gradually 
72: spiral in toward the center of the orbit.  
73: In a pioneering study, \citet{chandra} derived the classical 
74: formula of dynamical friction in a uniform collisionless background,
75: the result of which has been applied to a number of astronomical systems.  
76: Examples include orbital decay of satellite galaxies orbiting their host 
77: galaxies
78: (e.g., \citealt{tremaine,lin83,weinberg, hashimoto, fujii}; see also
79: \citealt{bin87}),
80: dynamical fates of globular clusters near the Galactic center 
81: (e.g., \citealt{kimss03, mcm03, kimss}),
82: galaxy formation within the framework of hierarchical clustering scenario
83: (e.g., \citealt{zen03,bul05} and references therein),
84: formation of Kuiper-belt binaries \citep{gol02}, and 
85: planet migrations  \citep{planet} via interactions with planetesimals, etc.
86: 
87: Although less well recognized, dynamical friction also operates in gaseous 
88: backgrounds.  Using a time-dependent linear perturbation theory, 
89: \citet[hereafter O99]{ost99} derived the analytic expressions for 
90: the density wake and drag force 
91: for a perturber in a uniform gaseous medium.  O99 showed that 
92: resonant interactions between a perturber and pressure waves
93: make the gaseous drag more efficient than the collisionless drag
94: when the Mach number $\mach\sim1$.
95: She also found that even a subsonic perturber experiences a nonvanishing 
96: gaseous drag if interaction time between the perturber and the background
97: is finite.  This is an improvement on the previous notion that 
98: the gaseous drag is absent for subsonic perturbers 
99: because of the front-back symmetry in the steady-state density wake 
100: \citep{dokuchaev, ruderman, rep80}.
101: 
102: The results of O99 were confirmed numerically by \citet{sanchez99}
103: and have been applied to various situations including
104: massive black hole mergers in galactic nuclei
105: (e.g., \citealt{escala,liu,escala05,dotti}),  
106: orbital decay of compact objects (e.g., \citealt{narayan,kar01}) 
107: and associated viscous heating (e.g., \citealt{chang,cha03}) 
108: in accretion disks, and heating of an intracluster medium by 
109: supersonically moving galaxies in clusters 
110: (e.g., \citealt{elz04,fal05,kim}).  Without involving shocks,
111: density wakes in a collisionless medium are distributed more smoothly  
112: and achieve larger amplitudes than those in a gaseous medium \citep{mul83},
113: which led \citet{fur02} to suggest that X-ray emissions from galaxy wakes 
114: can in principle be used to discern the collisional character 
115: of dark matter in galaxy clusters.
116: 
117: While the results of \citet{chandra} and O99 are simple and 
118: provide good physical insights, they apply strictly to a mass traveling on 
119: a straight-line trajectory through an infinite homogeneous background.
120: Real astronomical systems obviously have nonuniform density distributions
121: and perturbers tend to follow curvilinear orbits. 
122: For instance, motions of galaxies in galaxy clusters, binary black holes
123: near the central parts of galaxies, and compact stars in accretion disks 
124: are better approximated by near-circular than straight-line orbits,
125: and their background media usually in hydrostatic equilibrium are 
126: stratified in the radial direction.  
127: Even for objects experiencing orbital decay, a near-circular orbit is a 
128: good approximation if the associated friction time is longer than the
129: orbital time.
130: Consideration of a circular-orbit perturber is of particular interest 
131: since it will allow the perturber, if supersonic, to overtake the backside 
132: of its wake that was created about an orbital period earlier.
133: In this case, a steady-state wake that eventually forms has
134: morphology and drag force that might be 
135: significantly different from 
136: the linear-trajectory counterparts.
137: Numerical simulations carried out by \citet{sanchez} and \citet{escala} 
138: indeed show that the density wake by a near-circular orbit perturber
139: contains a trailing spiral tail, which is absent in the linear-trajectory 
140: cases.  
141: They also showed that the resulting drag force is smaller than the 
142: estimate based on the formula given in O99,
143: which is probably due to the near-circular orbit,
144: although the effect of nonuniform backgrounds in their models 
145: cannot be ignored completely.
146: 
147: The drag formula based on perturbers moving straight in either a
148: collisionless medium or a gaseous medium depends on the Coulomb logarithm 
149: $\ln(\Lambda) \equiv \ln(\rmax/\rmin)$, where
150: $\rmin$ and $\rmax$ are the cutoff radii introduced 
151: to avoid a divergence of the force integrals.\footnote{For a perturber
152: moving with velocity $V_p$ through a gaseous medium, $\rmax=V_p t$, 
153: where $t$ denotes time elapsed since the perturber was introduced (O99).}
154: While many previous studies conventionally adopted $\rmin$ and $\rmax$ as 
155: the characteristic sizes of the perturber and the background medium,
156: respectively, the choice of $\rmax$ remains somewhat
157: ambiguous for objects moving on near-circular orbits
158: (e.g., \citealt{bin87}).
159: For {\it collisionless} backgrounds, \citet{hashimoto} and \citet{fujii} 
160: performed N-body experiments for orbital evolution of satellite galaxies
161: in a spherical halo, and found that 
162: the Coulomb logarithm with $\rmax$ varying proportionally
163: to the orbital radius rather than fixed to the system size
164: gives better fits to their numerical results (see also \citealt{tremaine}).
165: We shall show in the present work that 
166: a similar modification of the Coulomb logarithm is necessary in order 
167: to apply the results of O99 to perturbers on near-circular orbits 
168: in {\it gaseous} backgrounds, as well.
169: 
170: In this paper, we consider a perturber moving on a circular orbit in a 
171: uniform gaseous medium.  Using a linear semi-analytic approach, we 
172: explore the structure of the density wake,
173: evaluate the drag force on the perturber, 
174: and compare them with those in the straight-line trajectory cases.  
175: In \S2, we revisit the linear perturbation
176: analysis of O99 for the perturbed density response and apply it to
177: the case of a circular-orbit perturber. We solve the resulting
178: equations numerically.  In \S3, we present the numerical results 
179: for the wakes and drag forces with varying Mach number.
180: We provide simple fitting expressions to the numerical results, 
181: and show that Ostriker's formula still gives a good estimate for 
182: the drag force on a circular-orbit perturber only if 
183: the outer cutoff radius in the Coulomb logarithm is taken equal to the 
184: orbital diameter of the perturber.
185: In \S4, we summarize the present work and briefly discuss our finding. 
186: 
187: \section{FORMULATION}
188: 
189: \subsection{Formal Solution for Density Wake}
190: 
191: We consider the response of gas to 
192: a point-mass perturber moving on a circular orbit and calculate the 
193: resulting gravitational drag force on the perturber.  
194: We treat the gas as an inviscid, adiabatic fluid, and 
195: do not consider the effects of magnetic fields as well 
196: as gaseous self-gravity.
197: The governing equations for ideal hydrodynamics are
198: \begin{equation}\label{eq:con}
199:   \frac{\partial\rho}{\partial t}+\vct{\nabla}\cdot(\rho\vct{v})=0,
200: \end{equation}
201: and
202: \begin{equation}\label{eq:mom}
203:   \frac{\partial\vct{v}}{\partial t}+\vct{v}\cdot\vct{\nabla}\vct{v}
204:   = -\frac{1}{\rho}\vct{\nabla}P-\vct{\nabla}\Phi_\ext,
205: \end{equation}
206: where $\Phi_\ext$ is the gravitational potential of the perturber.
207: Other symbols have their usual meanings. 
208:   
209: Following O99,
210: we consider an initially uniform gaseous medium with density $\rho_0$.
211: Assuming that the wake induced by the perturber remains at small amplitudes, 
212: we linearize equations (\ref{eq:con}) and (\ref{eq:mom}) using
213: $\rho=\rho_0[1+\alpha(\vct{x},t)]$ and 
214: $\vct{v}=\cs\,\vct{\beta}(\vct{x},t)$, 
215: where $c_s$ is the adiabatic speed of sound in the unperturbed medium 
216: and $\alpha$ and $\vct{\beta}$ denote the 
217: dimensionless density and velocity perturbations, respectively. 
218: Eliminating $\vct{\beta}$ from the linearized equations, one 
219: obtains a three-dimensional wave equation
220: \begin{equation}\label{eq:wave}
221:   \vct{\nabla}^2\alpha-\frac{1}{\cs^2}\frac{\partial^2\alpha}{\partial t^2}
222:   = -\frac{4\pi G}{\cs^2}\rho_\ext(\vct{x},t), 
223: \end{equation}
224: where $\rho_\ext = \vct{\nabla}^2\Phi_\ext /(4 \pi G)$ represents 
225: the mass density of the perturber.
226: The formal solution to equation (\ref{eq:wave}) based on the 
227: the retarded Green function technique is given by
228: \begin{equation}\label{eq:formal}
229:   \alpha(\vct{x},t) = \frac{G}{\cs^2} \int\!\!\!\int d^3x' dt'
230:   \ \rho_\ext(\vct{x'},t')
231:   \frac{\delta\,[t'-(t-\vert\vct{x}-\vct{x'}\vert/\cs)]}
232:        {\vert\vct{x}-\vct{x'}\vert}
233: \end{equation}
234: (O99; see also \citealt{jackson}).
235: In the case of a perturber on a straight-line trajectory,
236: O99 solved equation (\ref{eq:formal}) directly to obtain an 
237: expression for the perturbed density.  The same result was found 
238: by \citet{fur02} who independently used a Fourier transform method
239: in both space and time variables. 
240: 
241: \subsection{Density Wake for Circular-Orbit Perturbers}
242: 
243: We now concentrate on the case where a point-mass perturber
244: with mass $M_p$ moves on a circular orbit with a fixed orbital radius 
245: $R_p$ and a constant velocity $V_p$ in an otherwise uniform gaseous medium;  
246: the angular speed of the perturber is $\Omega = V_p/R_p$.
247: It is convenient 
248: to work in cylindrical coordinates ($R$, $\varphi$, $z$) whose origin lies 
249: at the center of the orbit.  The $\hat{z}$-axis points perpendicular to the 
250: orbital plane.  Assuming that the perturber
251: is introduced at $(R_p, 0, 0)$ when $t=0$, one can write 
252: $\rho_\ext(\vct{x},t)=M_p\; 
253: \delta (R-R_p)\, \delta [R_p(\varphi- \Omega t)]\, 
254: \delta (z)\, \step (t)$, 
255: where $\step (t)$ is a Heaviside step function. 
256: Equation (\ref{eq:formal}) is then reduced to 
257: \begin{equation} \label{eq:delta}
258:   \alpha(\vct{x},t)=\frac{G M_p}{\cs^2 R_p} \int dw\ 
259:       \frac{\delta\left(w+s+\mach\ d\,(w;\tR,\tz)\right)}{d\,(w;\tR,\tz)}\ 
260:       \step\left(\frac{w+\varphi}{\Omega}\right),
261: \end{equation}
262: where $w \equiv \varphi^\prime-\varphi$ and 
263: $s \equiv \varphi-\Omega t$ are angular distances in 
264: the $z=0$ plane\footnote{Note that $w$ and $s$ in O99 are 
265: defined as {\it linear} distances along the line of motion, 
266: while they measure {\it angular} distances in the present work.},
267: \begin{equation}\label{eq:dist}
268: d\,(w;\tR,\tz)\equiv \frac{\vert\vct{x}-\vct{x'}\vert}{R_p} 
269: =\left(1 + \tR^2+\tz^2-2\tR\cos w\right)^{1/2},
270: \end{equation}
271: and $\mach\equiv V_p/\cs$ is the Mach number of the perturber.
272: In equation (\ref{eq:dist}), $\tR\equiv R/R_p$ and $\tz\equiv z/R_p$.
273: 
274: \Fig{sketch} schematically illustrates the situation at the orbital plane
275: and the meanings of variables used in equation (\ref{eq:delta}). 
276: At time $t$, the perturber is located at $\vct{x_p}=(R_p, \Omega t, 0)$.  
277: During its journey along the thick curve, the perturber continuously
278: launches sound waves that propagate into the background gaseous medium.  
279: The position $\vct{x}=(R,\varphi,z)$ denotes a region of interest
280: in the surrounding gas where the density response will be evaluated.
281: Since the sound waves have finite traveling time, only the signals 
282: emitted by the perturber at the location(s) $\vct{x'}=(R_p, \varphi', 0)$ 
283: at the retarded time $t'=t-\vert\vct{x}-\vct{x'}\vert/\cs$ 
284: are able to affect the point $\vct{x}$ at time $t$. 
285: Note that $s$ and $w$ represent the projected angular distances 
286: in the orbital plane between $\vct{x_p}$ and $\vct{x}$ and 
287: between $\vct{x'}$ and $\vct{x}$, respectively.  The symbol $d$ 
288: in equation (\ref{eq:dist}) refers to the three-dimensional linear distance
289: between $\vct{x'}$ and $\vct{x}$ normalized by $R_p$.
290: 
291: 
292: Using the identity $\delta\,[f(w)]=\sum_i\delta(w-w_i)/\vert f'(w_i)\vert$, 
293: where $w_i$ are the roots of an arbitrary function $f(w)$, equation 
294: (\ref{eq:delta}) is further simplified to
295: \begin{equation}\label{eq:alpha}
296:  \alpha( \vct{x},t) 
297:   = \frac{G M_p}{\cs^2 R_p}\ \D (\vct{x},t),
298: \end{equation}
299: with the dimensionless perturbed density 
300: \begin{equation} \label{eq:D}
301:   \D(\vct{x},t) = \sum_{w_i}
302:   \frac{\mach} {\vert w_i+s-\mach^2\tR\ \sin w_i \vert}
303:   \step\left(\frac{w_i+\varphi} {\Omega}\right).
304: \end{equation}
305: Here, the summation is over all possible roots $w_i$ that satisfy 
306: the condition 
307: \begin{equation} \label{eq:w} 
308: \mach\ d\,(w_i;\tR,\tz) = - ( w_i+s ),
309: \end{equation}
310: for fixed values of $\tR$, $\tz$, and $s$.  
311: The function $\step [(w_i+\varphi)/\Omega]$ in equation (\ref{eq:D}) 
312: defines the region of influence (or casual region) outside of which sonic 
313: perturbations sent off by the perturber at $t=0$ have insufficient time 
314: to reach.  For $\tR, \tz, \Omega t\gg 1$, equation (\ref{eq:w}) yields 
315: $w_i + \varphi = \Omega t - \mach (\tR^2+\tz^2)^{1/2}$, so that 
316: the region of influence corresponds roughly to a sphere with radius $c_st$
317: centered at the orbital center.
318: Appendix \ref{sec:limit} presents limiting solutions of 
319: equation (\ref{eq:D}) near the perturber. 
320: 
321: Since $d$ is a periodic function of $w$ with period 2$\pi$, 
322: equation (\ref{eq:w}) has at least one real root and may possess multiple
323: roots for $w_i$ depending on the values of $\mach$, $\tR$, $\tz$, and $s$.
324: In Appendix \ref{sec:wi}, we describe how the number of solutions of
325: equation (\ref{eq:w}) vary with the Mach number of a perturber.
326: It turns out that there is only a single root for $w_i$ 
327: everywhere for a subsonic, circular-orbit perturber, which is the same as 
328: in the straight-line trajectory case (O99). 
329: When a circular-orbit perturber moves at a supersonic speed, however, 
330: equation (\ref{eq:w}) has an odd number of roots that contribute to
331: the wake in a steady state, which is distinct from the straight-line 
332: trajectory case where
333: only one or two points along the orbit influence the wake. 
334: 
335: \subsection{Gravitational Drag Force}
336: 
337: Once the gravitational wake $\alpha( \vct{x},t)$ is found, 
338: it is straightforward to evaluate the drag force 
339: exerted on the perturber: 
340: \begin{equation}
341:   \vct{F}_\mathrm{DF} = G M_p \rho_0 \int d^3\vct{x}\ 
342:   \frac{\alpha(\vct{x},t)\ (\vct{x}-\vct{x_p})} 
343:        {\vert \vct{x}-\vct{x_p} \vert^3}\mbox{.}
344: \end{equation}
345: In the straight-line trajectory case studied by O99,
346: $\alpha(\vct{x},t)$ always remains cylindrically symmetric with respect to
347: the line of motion, resulting in the drag force 
348: in the anti-parallel direction. 
349: When the perturber is on a circular orbit, $\alpha(\vct{x},t)$ 
350: loses the cylindrical symmetry and instead becomes symmetric relative
351: to the orbital plane, making the vertical component of 
352: $\vct{F}_\mathrm{DF}$ vanish.  
353: 
354: We decompose the nonvanishing parts into
355: the radial and azimuthal components:  
356: \begin{equation}\label{eq:fdf}
357: \vct{F}_\mathrm{DF} =  -\mathcal{F} \;
358: (\I_{\rm R} \hat{\vct{R}} + \I_{\rm \varphi} \hat{\vct{\varphi}}),
359: \;\;\;\;
360: \mathcal{F} \equiv  \frac{4\pi\rho_0 (GM_p)^2}{V_p^2} 
361: \end{equation}
362: where 
363: \begin{mathletters} \label{eq:I}
364:   \begin{equation}
365:     \I_{R} = -\frac{\mach^2}{4\pi} \int d^3\tilde{\vct{x}}\ 
366:     \frac{\D({\vct{x}},t)\ (\tR\cos s -1)}
367: 	 {(1+\tz^2+\tR^2-2\tR\cos s)^{3/2}}\mbox{,}
368:   \end{equation}
369: and
370:   \begin{equation}
371:     \I_{\varphi} = -\frac{\mach^2}{4\pi} \int d^3\tilde{\vct{x}}\ 
372:     \frac{\D({\vct{x}},t)\ \tR\sin s}
373: 	 {(1+\tz^2+\tR^2-2\tR\cos s)^{3/2}}.
374:   \end{equation}
375: \end{mathletters}
376: Note that 
377: $\I_{R}$ measures the drag force along the lateral direction of the 
378: instantaneous perturber motion, 
379: while $\I_{\varphi}$ is for the backward direction.  
380: The dimensional term $\mathcal{F}$ in equation (\ref{eq:fdf}) allows to
381: directly compare $\I_{R}$ and $\I_{\varphi}$ with the
382: linear-trajectory counterparts (see eq.~[12] of O99).
383: As we shall show in \S\ref{sec:sum}, 
384: it is the azimuthal drag $\I_{\varphi}$ that 
385: is responsible for the orbital decay of a perturber.
386: 
387: 
388: 
389: \subsection{Numerical Method}
390: 
391: We solve equations (\ref{eq:D}) and  (\ref{eq:w}) 
392: numerically
393: to find the perturbed density distribution $\D(\vct{x},t)$ for given
394: $\mach$ and $t$. 
395: We first construct a three-dimensional Cartesian mesh centered at the
396: center of the orbit, and solve equation (\ref{eq:w}) for
397: $w_i$ at each grid point using a hybrid Newton-bisection method.  
398: By checking the conditions for multiple roots discussed 
399: in Appendix \ref{sec:wi}, 
400: we ensure that we do not miss any solution for $w_i$.
401: The corresponding drag forces $\I_R$ and $\I_\varphi$ are calculated
402: by direct integration of equations (\ref{eq:I}).
403: 
404: Since the density wake often exhibits sharp discontinuities especially
405: for supersonic perturbers and is distributed over a large spatial range,
406: it is important to check that the drag forces we calculate do not 
407: depend on the size of the computational box and its resolution. 
408: For fixed $\mach$, we repeated the calculations with varying box size
409: and resolution and found that depending on the Mach number, the box size 
410: of $\sim (20-100)R_p$ and resolution of $\sim (80-640)$ grids per 
411: $1R_p$ are sufficient to guarantee good convergence of the drag forces.
412: Although the density perturbations are non-zero outside the box,
413: they have very low amplitudes and are located far from the perturber, 
414: providing a negligible contribution to the drag.
415: Very high resolution calculations are required for Mach numbers 
416: near the critical values $\mach_n$, in which cases the wake tails become
417: thin and dense (see Appendix \ref{sec:wi}).
418: 
419: 
420: \section{RESULTS}\label{sec:res}
421: 
422: \subsection{Density Wake} \label{sec:density}
423: 
424: \subsubsection{Supersonic Cases}\label{sec:superwake}
425: 
426: We begin by illustrating temporal evolution of density perturbations 
427: induced by a supersonic perturber.
428: \Fig{evolution} shows snapshots of the density wake and the corresponding 
429: number of roots of equation (\ref{eq:w}) at the orbital 
430: plane for $\mach=2.0$. 
431: Time is expressed in units of $R_p$/$c_s$. The black circle represents 
432: the orbit of a perturber which, introduced at 
433: $ (R, \varphi, z) =(R_p, 0, 0)$ initially,
434: moves in the counterclockwise direction.
435: At early time ($t\simlt 1.5$), the density wake consists of a Mach cone
436: and a sonic sphere that are curved along the orbit.
437: Except the bending, the overall wake structure, not to mention the number
438: of solutions for $w_i$ which is either one or two inside the casual 
439: region, is the same as in the straight-line trajectory case of O99.
440: As the wake bends further, the Mach cone and the sonic sphere become 
441: folded at the innermost interface, creating high-density regions near 
442: the center ($t=1.8$).  The wake expands with time 
443: at a sonic speed and the Mach cone becomes elongated further. 
444: 
445: Unlike in the case of a straight-line trajectory 
446: where the Mach cone and the sonic sphere never interact with each other, 
447: the perturber (and the head of the Mach cone) on a circular orbit is 
448: able to enter its own wake, 
449: providing additional perturbations for some regions inside the sonic sphere.
450: Alternatively, this can be viewed as the sonic sphere whose center lies 
451: at the initial position of the perturber expands radially outward, 
452: swallowing a part of the elongated Mach cone.  Consequently, the overlapping 
453: of the Mach cone and the sonic sphere creates a high-density trailing tail
454: that has received perturbations three times from the perturber
455: ($t\simgt 1.9$).
456: \Fig{evolution} shows that immediate outside the sonic sphere, 
457: there still exists a region of the undisturbed Mach cone 
458: with two $w_i$'s.  As time proceeds, however, this region
459: moves away from the perturber and thus gives an increasingly small 
460: contribution to the drag force.  In a steady state which is 
461: attained at $t\rightarrow\infty$, the entire domain is affected  
462: either once or three times by the perturber.
463: 
464: \Fig{steadyM20} displays the steady-state distributions of the density wake 
465: for $\mach=2.0$ on the $x/R_p=1$, $y=0$, and $z=0$ planes, 
466: which clearly shows that the trailing tail loosely wraps around 
467: the perturber.
468: The tail becomes narrower with increasing $|z|$.
469: A close inspection of the tail at the $z=0$ plane shows that density 
470: becomes smaller away from the perturber along the tail and is largest
471: at the edges across the tail.  This is similar to the density
472: distribution inside the Mach cone in the linear trajectory
473: case where diverging flows (and reduced gravitational potential) after 
474: the shock cause density to decrease away from the shock front
475: (and the perturber). 
476: This suggests that the edges of the tail are shock fronts.
477: The outer edge of the tail shown in \Fig{steadyM20} that connects smoothly
478: all the way to the perturber corresponds indeed to the surface of the 
479: curved Mach cone (see also Fig.~\ref{sketch}).  
480: On the other hand, the inner edge of the tail traces the interface between 
481: the Mach cone and the sonic sphere that is newly swept up by the expanding
482: sonic sphere (see Fig.~\ref{evolution}).
483: Appendix \ref{sec:angle} proves that the half-opening angle of the
484: head of the curved Mach cone in the $z=0$ plane is equal to 
485: $\sin^{-1} (1/\mach)$, entirely consistent with the case of a linear 
486: trajectory.
487: 
488: The wake tail becomes thicker as $\mach$ increases from unity. 
489: \Fig{overlap} shows the steady-state density wakes 
490: as well as the number of $w_i$'s that contribute to $\D$
491: for $\mach=4.0$ and $\mach=5.0$ at the $z=0$ plane.  In both panels, 
492: the perturber moving in the counterclockwise direction is located 
493: at $x/R_p=1, y=0.$
494: When $\mach=4.0$, the tail is fat enough to cover most of the space
495: except near the center and a narrow lane between the tail edges.  
496: At $\mach_1 \approx 4.603$, the inner edge of the tail
497: eventually touches the neighboring outer edge, 
498: enabling three $w_i$ for the entire region under the conditions 
499: expressed by equation (\ref{eq:ncond}).
500: When the Mach number is slightly larger than $\mach_1$,
501: the tail overlaps itself.  This in turn creates a new tail with 
502: five $w_i$'s, as \Fig{overlap}\textit{b} displays.  
503: As $\mach$ increases further, 
504: the new tail again becomes thicker, starts to overlap
505: at $\mach_2 \approx 7.790$, and produces a narrow lane with
506: seven $w_i$'s when $\mach>\mach_2$.  The same pattern repeats with 
507: increasing $\mach$, and equation (\ref{eq:M}) determines 
508: the critical Mach numbers.
509: 
510: \subsubsection{Subsonic Cases}\label{sec:subwake}
511: 
512: Unlike in the supersonic cases where a perturber can overtake its own wake 
513: and create a tail with complicated structure, sonic perturbations generated 
514: by a subsonic perturber produce a gravitational wake that is spatially smooth
515: and does not involve a shock.  Since perturbations propagate 
516: faster than a perturber with $\mach<1$, 
517: the whole casual region is affected by a perturber 
518: just once, corresponding to a single $w_i$ at any position.
519: \Fig{steadyM05} shows the slices
520: of the perturbed density for $\mach=0.5$ in the
521: $x/R_p=1$, $y=0$, and $z=0$ planes when a steady state is reached.
522: Again, the perturber is located at $x/R_p=1$, $y=z=0$.
523: The density structure in the $z=0$ plane is simply a curved version 
524: of what a linear-trajectory perturber would produce. 
525: In particular, as Appendix \ref{sec:limit} shows, 
526: the iso-density surfaces near the perturber have the shapes of 
527: oblate spheroids with ellipticity $e=\mach$, with the short axis parallel to
528: the direction of the motion of the perturber, a characteristic feature
529: of a subsonic wake created by a linear-trajectory perturber (O99).
530: 
531: Notice, however, that the bending of wakes in circular-orbit cases 
532: makes the perturbed density distributions intrinsically asymmetric.
533: This results in nonvanishing drag forces even in a steady state,
534: and the dominant contribution to the drag comes from high-density
535: regions near the perturber.
536: This is markedly different from the purely steady-state linear-trajectory 
537: cases where a subsonic perturber experiences no drag due to
538: the front-back symmetry of a wake about the perturber 
539: (e.g., \citealt{rep80}).
540: Even if the finite interaction time between the straight-line perturber and
541: the background gas is considered, regions with symmetric perturbed
542: density close to the perturber exert zero net force (O99).
543: Nevertheless, the resulting drag force in the backward direction of motion 
544: on a circular-orbit perturber is almost the same as that in the 
545: linear-trajectory cases, as we will show in the next subsection. 
546: Compared with supersonic cases, the tail in a subsonic wake is short,
547: loosely wound, and very weak, suggesting that its contribution to
548: the drag force is negligible. 
549: 
550: 
551: \subsection{Gravitational Drag Force}
552: 
553: As sonic perturbations launched from a perturber at $t=0$ propagate 
554: radially outward, the volume of space exerting the gravitational 
555: drag on the perturber grows with time.
556: \Fig{time} plots the drag force as functions of time for a few chosen
557: Mach numbers.  The solid and dotted lines are for $\I_\varphi$ and
558: $\I_R$, respectively.  Since $\D(\vct{x},t)$ is singular at 
559: $\vct{x}=\vct{x}_p$,  only the region with $r > r_{\rm min} = R_p/10$ 
560: is taken into account in force computation,
561: where $r$ is the three-dimensional distance measured from the perturber;
562: the dependence on $\rmin$ will be checked below.
563: The drag force on a subsonic perturber increases almost linearly with
564: time before turning abruptly to a constant value, whereas a supersonic
565: wake with a high-density tail gives rise to slow fluctuations 
566: in the drag at early time.  At any event, both components of the drag 
567: force converge to respective steady-state values, 
568: typically within the sound crossing time
569: over the distance equal to the orbital diameter or 
570: within about an orbital period when $\mach$ is of order unity.
571: The primary reason for this is of course because the perturbed density 
572: decreases quite rapidly with $r$ and also because
573: gravity is an inverse-square force.
574: This is unlike the case of a straight-line trajectory where
575: the drag increases secularly as $\ln(V_p t)$ for $\mach>1$.
576: The fast convergence of the drag force guarantees that 
577: one can use the steady-state values of $\I_\varphi$ and $\I_R$ 
578: for all practical purposes.
579: 
580: 
581: Next, we check the dependence of the drag force on $\rmin (\ll R_p)$.
582: \Fig{rmin} shows the results for 
583: $\mach=0.5$, 2.0, and 4.0.  The sizes of errorbars associated with 
584: finite grid resolution are smaller than the radius of a solid circle
585: at each data point.
586: First of all, the drag force, $\I_R$, in the radial direction converges 
587: to a constant value as $\rmin$ decreases for both subsonic and 
588: supersonic cases. 
589: On the other hand, the drag force, $\I_\varphi$, in the opposite direction
590: of the orbital motion, varies as $\ln (1/\rmin) $ with decreasing $\rmin$ for
591: supersonic perturbers, while independent of $\rmin$ for subsonic
592: perturbers.
593: This dependence of $\I_\varphi$ on small $\rmin$ for 
594: circular-orbit perturbers is exactly the same as in the linear 
595: trajectory cases, which makes sense because the curvature of
596: a circular orbit is almost negligible in a tiny volume 
597: near the perturber.
598: 
599: We plot in \Fig{forces} the steady-state drag forces for a 
600: circular-orbit perturber as functions of the Mach number.  
601: For all the points, $\rmin=R_p/10$ is taken and numerical 
602: convergence is checked.  
603: Filled circles give $\I_\varphi/\mach^2$, while open circles are for 
604: $\I_R/\mach^2$, which can be compared with Figure 3 of O99.
605: For practical use, we fit the data using 
606: \begin{equation}\label{eq:IR}
607:   \I_R = \left\{\begin{array}{l l@{\ }r@{\;}c@{\,}l}
608:     \mach^2\ 10^{\ 3.51\mach-4.22},   &\textrm{for}&&\mach&<1.1,\\
609:     0.5\  \ln\big[9.33\mach^2(\mach^2-0.95)\big],
610:                                        &\textrm{for}&1.1\leq&\mach&<4.4,\\
611:     0.3\ \mach^2,                       &\textrm{for}&4.4\leq&\mach,
612:   \end{array}\right.
613: \end{equation}
614: and
615: \begin{equation}\label{eq:Iphi}
616:   \I_\varphi = \left\{\begin{array}{l l@{\ }r@{\;}c@{\,}l}
617:      0.7706\ln\left(\frac{1+\mach}{1.0004-0.9185\mach}\right)
618:        -1.4703\mach,
619:                      &\textrm{for}&&\mach&<1.0,\\
620:      \ln\left[330 (R_p/\rmin) (\mach-0.71)^{5.72}\mach^{-9.58} \right],
621:                       &\textrm{for}&1.0\leq&\mach&<4.4, \\
622:      \ln\left[(R_p/\rmin)/(0.11\mach+1.65)\right],
623:                      &\textrm{for}&4.4\leq&\mach,
624:   \end{array}\right.
625: \end{equation}
626: which are drawn as solid lines in \Fig{forces}.
627: The fits are within 4\% of our semi-analytic results for $\mach<4.4$ 
628: and within 16\% for $\mach>4.4$. 
629: Both components of the drag force peak at $\mach\sim 1.2-1.4$, 
630: analogous to the linear-trajectory cases, and $\I_\varphi$ dominates 
631: over $\I_R$ for transonic perturbers.  
632: Although $\I_R$ is larger than $\I_\varphi$ 
633: for $\mach \simgt 2.2$, its effect on orbital decay of a perturber 
634: is insignificant (see \S\ref{sec:sum}).
635: The local bumps in the drag forces at $\mach \approx 4.6$ and 7.8 are due to
636: the self-overlapping of a wake tail explained in \S\ref{sec:superwake}.
637: 
638: 
639: \Fig{forces} also plots as dotted curves the results of O99 for drag force
640: on linear-trajectory perturbers. 
641: Despite the difference in the shape of the orbits,
642: the agreement between $\I_\varphi$ and 
643: Ostriker's formula is excellent for the subsonic case.
644: This presumably reflects the fact that other than bending, the wake 
645: structure created by a subsonic circular-orbit perturber is not 
646: significantly different from the linear-trajectory counterpart 
647: (see \S\ref{sec:subwake}). 
648: Note also that Ostriker's formula for the supersonic case,
649: for which $V_pt=2R_p$ is adopted to fit our numerical results, 
650: is overall in good agreement with $\I_\varphi$ for a range of 
651: $\mach$. 
652: It is remarkable that the subsonic and supersonic drag formulae 
653: (with $V_p t$ chosen adequately in supersonic cases) 
654: obtained from perturbers moving straight still give a reasonably good 
655: estimate for the drag even on circular-orbit perturbers. 
656: 
657: 
658: 
659: \section{SUMMARY AND DISCUSSION}\label{sec:sum}
660: 
661: We have taken a semi-analytic approach to study the gravitational wake
662: and the associated drag force on a perturber moving on a circular orbit 
663: in an infinite, uniform gaseous medium.  This work extends 
664: \citet{ost99} who studied the cases with straight-line
665: trajectory perturbers.  The circular orbit generally causes the wake to 
666: bend along the orbit and creates a trailing tail.
667: For a subsonic perturber, the density wake has a weak tail and is 
668: distributed quite smoothly (see Fig.~\ref{steadyM05}). 
669: On the other hand, a supersonic perturber can catch up with its own wake,
670: possibly multiple times depending on the Mach number, 
671: forming a very pronounced trailing tail across which density 
672: changes abruptly (see Fig.~\ref{steadyM20}).  
673: Although the region influenced by the perturber keeps expanding 
674: with time into the surrounding medium, the drag
675: force promptly saturates to a steady state value within less than 
676: the crossing time of sound waves over the distance equal to 
677: the orbital diameter.
678: 
679: Because of asymmetry in the density wake, it is desirable to decompose 
680: the drag force into two components: $\I_R$ and $\I_\varphi$ in the radial and
681: azimuthal directions, respectively (eq.~[\ref{eq:fdf}]).
682: \Fig{forces} plots our main results for 
683: $\I_R$ and $\I_\varphi$ as functions of the Mach number $\mach$;
684: equations (\ref{eq:IR}) and
685: (\ref{eq:Iphi}) give the algebraic fits to the numerical results. 
686: The azimuthal drag force varying rather steeply with $\mach$ peaks 
687: at $\mach\sim1.2-1.4$, while the radial force is highly 
688: suppressed at $\mach<1$ and exceeds the azimuthal drag at 
689: $\mach\simgt 2.2$.  
690: It is remarkable that the drag on linear-trajectory perturbers given in 
691: O99 is almost identical to 
692: $\I_\varphi$ for subsonic cases, and gives a good approximation
693: for supersonic cases, too, provided $V_pt = 2R_p$.
694: 
695: A striking feature in gravitational wakes formed by circular-orbit
696: perturbers is a long tail in a trailing spiral shape.  Such a tail 
697: structure is indeed commonly seen in recent hydrodynamic simulations 
698: for black hole mergers in a gaseous medium 
699: (e.g., \citealt{escala,escala05,dotti}; see also \citealt{sanchez}).
700: Albeit much weaker, it is also apparent in N-body simulations 
701: for satellite orbital decay
702: in a collisionless background (e.g., \citealt{weinberg,her89}).
703: For supersonic perturbers, the tail is a curved Mach cone and 
704: bounded by the shock fronts that propagate radially outward.
705: As explained in Appendices \ref{sec:wi} and \ref{sec:angle}, the 
706: outer edge of the tail is described by $\Omega t - \varphi = w_+ + y_+$,
707: where $w_+$ and $y_+$ are defined through equation (\ref{eq:wy}).  
708: By taking a time derivative of this equation for fixed $\varphi$,
709: one can show that the propagation speed $\dot R_{\rm sh}$ 
710: of the spiral tail in the radial direction is given by
711: $\dot R_{\rm sh}/c_s = \mach \tR(\mach^2\tR^2-1)^{-1/2}$
712: for $\tR\equiv R/R_p\geq1$.  
713: Clearly, $\dot R_{\rm sh}=c_s$ for $R/R_p \gg 1$,
714: and $\dot R_{\rm sh} = c_s\mach/(\mach^2-1)^{1/2}$ near $R/R_p =1 $ as 
715: equation (\ref{eq:half}) implies.
716: This prediction is roughly consistent with 
717: \citet{escala} who empirically found that the tail in a model 
718: with $\mach=\sqrt{2}$ has an average propagation speed of 
719: $\dot R_{\rm sh} \approx 1.1 c_s$.
720: 
721: 
722: Many previous studies on dynamical friction 
723: adopted the drag formula based on perturbers moving straight and 
724: estimated the Coulomb logarithm by taking $\rmax=R_{\rm sys}$, 
725: where $R_{\rm sys}$ is the system size (see references in \S1). 
726: Unlike a straight-line trajectory,
727: a circular orbit naturally introduces a characteristic 
728: length scale, $R_p$. The results of our semi-analytic analyses suggest 
729: that, at least in a gaseous medium, 
730: the drag force obtained for linear-trajectory perturbers 
731: can be a reasonable approximation to that for circular-orbit 
732: perturbers with the same Mach number 
733: if $\rmax \equiv V_p t$ is taken equal to $2R_p$.
734: Since $R_{\rm sys}>2R_p$ in most relevant situations, using 
735: $\rmax =R_{\rm sys}$ usually overestimates the drag force for 
736: objects in near-circular motion. 
737: From hydrodynamic simulations for orbital decay of 
738: perturbers in a stratified gaseous sphere, 
739: \citet{sanchez} suggested $\ln(\Lambda)=
740: \ln(\rmax/\rmin) \rightarrow 2.34 \ln 
741: (0.79 R_p/\rmin)$ with an identification $\rmin = 2.25 R_{\rm soft}$, 
742: where $R_{\rm soft}$ is the softening radius of the point-mass potential
743: \citep{sanchez99}. This happens to be similar to our suggestion 
744: $\ln(\Lambda) \rightarrow \ln(2R_p/\rmin)$ for the parameter
745: range $R_p/\rmin \sim 2-6$ that they considered.  
746: 
747: Our suggestion for the Coulomb logarithm on near-circular orbit perturbers 
748: in a gaseous medium is also consistent with the common prescription 
749: for orbital decay of small satellites in collisionless backgrounds
750: (e.g., \citealt{tremaine, lin83, hashimoto, fujii}).
751: In particular, \citet{hashimoto} found that 
752: Chandrasekhar's formula with $\ln(\Lambda) \rightarrow
753: \ln[R_p/(1.4 R_{\rm soft})]$ gives excellent fits to the results of
754: their N-body simulations.
755: The discrepancy between our suggestion and their prescription may be
756: attributable in part to the effects of 
757: collisionless nature, nonuniform background density, 
758: and self-gravity of particles in their models. 
759: 
760: Density inhomogeneity in the background can also make significant 
761: changes to the classical drag force.  For collisionless backgrounds,
762: \citet{just} found that nonuniform density induces 
763: an additional drag force in the lateral direction of 
764: the perturber motion, taking up to 10\% of the drag in the 
765: backward direction.
766: They proposed the inverse of a local density gradient as 
767: an appropriate choice for $\rmax$ in the Coulomb logarithm 
768: (see also \citealt{spi03}), which was confirmed by \citet{are07} who
769: ran a number of numerical simulations for systems with a 
770: large central density concentration.
771: Hydrodynamic models in \citet{sanchez} studied the combined effects of 
772: nonuniform backgrounds and curvilinear orbits 
773: on dynamical friction in a gaseous medium,
774: although it is challenging to isolate each effect separately.
775: 
776: Finally, as a heuristic example, we consider the dynamical friction 
777: of a galaxy on a near-circular orbit subject to the drag force given by
778: equations (\ref{eq:IR}) and (\ref{eq:Iphi}) due to an intracluster
779: medium.  The cluster is dominated by dark matter 
780: whose mass distribution follows the NFW profile with the characteristic mass
781: $M_0=6.6\times 10^{14} M_\sun$ and the scale radius $r_\mathrm{s}=460$ kpc
782: \citep{NFW}.  The intracluster medium has 
783: a constant electron density $n_e=0.05\ \mathrm{cm}^{-3}$ 
784: and temperature 1 keV; the corresponding adiabatic speed of sound is 
785: $\cs = 500\ \kmpers$.  
786: Initially, a galaxy with size $\rmin=10$ kpc and 
787: mass $M_p=5\times 10^{11} M_\sun$ including 
788: a dark halo (e.g., \citealt{zen03,halkola}) is orbiting at $R_0=100$ kpc 
789: with an equilibrium velocity $V_p=10^3\ \kmpers$ about the
790: cluster center.
791: The equations of motion in the orbital plane are
792: \begin{mathletters} \label{eq:gal}
793:   \begin{equation}
794:     \ddot{R} - R\dot{\varphi}^2 = -(\mathcal{F}/M_p) \I_R - 
795:     \frac{d\Phi_{\rm NFW}}{dR}
796:      \mbox{,}\\
797:   \end{equation}
798:   \begin{equation}
799:     R\ddot{\varphi} + 2\dot{R}\dot{\varphi}  = 
800:     -(\mathcal{F}/M_p) \I_\varphi,
801:    \end{equation}
802: \end{mathletters}
803: where $\Phi_{\rm NFW}$ is the NFW potential and dots represent 
804: derivatives with respect to time.  
805: \Fig{orbitNFW} plots the 
806: resulting temporal variations of the orbital radius of the galaxy.
807: The solid line corresponds to the case with full $\I_R$ and $\I_\varphi$
808: given by equations (\ref{eq:IR}) and (\ref{eq:Iphi}),
809: while the dashed line is for a controlled case 
810: where $\I_R$ is artificially taken to be zero.  
811: Except for slight mismatches in phase,
812: both agree fairly well with each other, demonstrating that 
813: the radial drag force does not have a serious impact on the orbital decay.  
814: As equation (\ref{eq:gal}) implies, it is the azimuthal
815: drag $\I_\varphi$ that takes away most of the orbital angular momentum from 
816: the galaxy;  the radial drag changes the orbital eccentricity more 
817: than the semi-major axis (\citealt{bur76}; see also \citealt{just}).
818: 
819: \Fig{orbitNFW} also plots as dot-dashed line the decay of the orbital radius
820: blindly using Ostriker's formula with $\ln (V_pt/\rmin) = 4.6$,
821: corresponding to $V_pt=1$ Mpc.  
822: While the galaxy motion remains supersonic ($t<1.5$ Gyr),
823: this value of the Coulomb logarithm overestimates the realistic drag force,
824: on average, by a factor of 1.7 and thus brings the galaxy 
825: to $R=0.1R_0$ in 2 Gyrs, 
826: which is about 2.3 times faster than the estimate based on the realistic 
827: drag force. On the other hand, the result of using $V_p t = 2R(t)$ in 
828: the formula of O99, shown as dotted line in \Fig{orbitNFW}, is in excellent 
829: agreement with those using equations (\ref{eq:IR}) and (\ref{eq:Iphi}).
830: This demonstrates again that Ostriker's formula with $V_p t = 2R$ 
831: gives quite a good approximation to the drag force even in the case of
832: a circular-orbit perturber.
833: 
834: \acknowledgments
835: 
836: We are grateful to E.\ Ostriker for suggesting this project 
837: and making stimulating comments.  We also acknowledge a thoughtful report
838: from the referee, A.\ Escala, and helpful comments from J.\ 
839: S{\'a}nchez-Salcedo.
840: This work was supported by Korea Science and Engineering
841: Foundation (KOSEF) grant R01-2004-000-10490-0 at SNU. 
842: H.\ Kim has been partially supported by the BK21 project of the 
843: Korean Government.
844: The numerical computations presented in this work were performed on the 
845: Linux cluster at KASI (Korea Astronomy and Space Science Institute) 
846: built with funding from KASI and ARCSEC.
847: 
848: 
849: \appendix
850: \section{
851: Solutions for $\D(\vct{x},t)$ Near the Perturber}
852: \label{sec:limit}
853: 
854: In this Appendix, we explore the approximate solutions to equation 
855: (\ref{eq:D}) near the perturber ($|\tR-1|, |s|, |\tz| \ll 1$). 
856: For subsonic motion ($\mach<1$), $|w_i| \ll 1$ and 
857: $d\approx ( \delta^2 + w^2 + \tz^2)^{1/2}$ with $\delta\equiv \tR-1$. 
858: In this case, equation (\ref{eq:w}) has a single root, 
859: \begin{equation}\label{eq:www}
860: w_1 = \frac{s + \mach [ s^2 + (1-\mach^2)(\delta^2 + \tz^2)]^{1/2}}
861: {\mach^2-1}.
862: \end{equation}
863: Inserting $w_1$ into equation (\ref{eq:D}) 
864: and retaining the 
865: first-order terms in  $|\tR-1|, |s|$, and $|\tz|$, one obtains
866: $\D(\vct{x}) 
867: = [s^2 + (1-\mach^2)(\delta^2 + \tz^2)]^{-1/2}$ at a steady state,
868: whose functional form is identical to equation (9) of O99 
869: for $\mach<1$.  Obviously,
870: the equi-density surfaces of this distribution are 
871: oblate spheroids with the short axis parallel to the direction
872: of the motion of the perturber.  
873: 
874: For supersonic motion ($\mach>1$), we seek for asymptotic solutions 
875: of $\D$ just behind the perturber along its orbit (i.e., 
876: $0<-s\ll 1$, $\tR=1, \tz=0$).
877: We further restrict the discussion to not-so-highly 
878: supersonic cases\footnote{More
879: precisely, $1<\mach  < \mach_1$, where $\mach_1$ is a critical Mach
880: number discussed in Appendix \ref{sec:wi}.}, 
881: for which equation (\ref{eq:w}) has three roots:
882: \begin{mathletters}\label{eq:wthree}
883:   \begin{eqnarray}
884:     w_1 &\approx& -\frac{s}{1 + \mach}, \\
885:     w_2 &\approx& -\frac{s}{1 - \mach}, \\
886:     w_3 &\approx& w_0 -s \left(1 - \frac{\mach^2 \sin w_0}{w_0}\right)^{-1},
887:   \end{eqnarray}
888: \end{mathletters}
889: as long as $0<-s \ll (\mach-1)^{3/2}$. 
890: Here, $w_0$ is a non-zero solution of
891: \begin{equation}
892: \mach(2-2\cos w_0)^{1/2} + w_0=0
893: \end{equation}
894: for given $\mach$ ($\neq1$). 
895: Note that $w_1$ and $w_2$ are analogous to two solutions of O99 that
896: contribute the wake in the rear Mach cone for a linear-trajectory
897: perturber, while $w_3$ is a new solution introduced by a circular orbit.
898: For $\mach \rightarrow 1^{+0}$, $w_0 \rightarrow -[24(\mach-1)]^{1/2}$.
899: 
900: Substituting equations (\ref{eq:wthree}) into equation (\ref{eq:D})
901: and taking the steady-state limit ($\step \rightarrow 1$),
902: we obtain
903: \begin{equation}\label{eq:Dss}
904: \D(s) = \frac{2}{\vert s \vert} + 
905: \frac{\mach}{\mach^2\sin w_0 - w_0}, 
906: \end{equation}
907: for $-s \ll (\mach-1)^{3/2}$.
908: The first term in the right-hand side of equation (\ref{eq:Dss}) again 
909: represents the perturbed density in the linear trajectory case. 
910: The second term, albeit small compared with the first term for small $|s|$, 
911: arises as the perturber on a circular orbit can overtake its wake.  
912: The effects of circular orbits are significant in regions
913: away from the perturber, which is pursued numerically in \S\ref{sec:res}. 
914: 
915: 
916: 
917: \section{Number of Roots for $w_i$ of Equation (\ref{eq:w})} \label{sec:wi} 
918: 
919: To obtain the perturbed density $\D(\vct{x},t)$ at given time and location,
920: it is crucial to find $w_i$ that satisfies equation (\ref{eq:w}).
921: Let us define $f(w)\equiv \mach d(w; \tR, \tz)$ for a fixed set of 
922: $\mach, \tR, \tz$, and $s$.  Then, finding solutions of equation 
923: (\ref{eq:w}) is equivalent to finding the intersections of two 
924: curves $y = -w -s$ and $y=f(w)$ on the $w$ -- $y$ plane.  Since 
925: $\vert df/dw \vert \leq \mach$ for all $\tR$ and $\tz$, it is 
926: trivial to show that
927: there is only one solution for all space and time if $\mach < 1$.
928: This implies that any location inside the casual 
929: region for sound waves is influenced just once by a perturber moving 
930: at a subsonic speed.  The same is true for a 
931: subsonic perturber moving on a straight-line trajectory studied 
932: by O99.
933: 
934: A necessary condition for having more than one root is 
935: $\vert df/dw \vert > 1$ for any $w$, which after some algebra results in 
936: \begin{equation}\label{eq:ncond}
937:     \mach > 1, \quad
938:     \tR   > \mach^{-1},\quad\mbox{and} \quad
939:     |\tz|   < \mach^{-1}\ \sMsml\ \sMsRsml\mbox{.}
940: \end{equation}
941: When the conditions (\ref{eq:ncond}) are fulfilled, there are two
942: tangent points of the curve $y=f(w)$ to lines with 
943: a slope of $-1$ in the range $-\pi < w < 0$.  
944: Let ($w_+$, $y_+$) and ($w_-$, $y_-$) denote the tangent points, 
945: such that 
946: $-\pi < w_- < w_+<0$ and $0<y_+ < y_-<\mach$.  
947: The tangent condition $df/dw\vert_{w_\pm} =-1$ yields 
948: \begin{mathletters}\label{eq:wy}
949:   \begin{eqnarray}
950:     w_\pm &=& - \cos^{-1}
951:           \left(\frac{1\pm\sMsml\sMsRsml}{\mach^2\tR}\right)\mbox{,} \\
952:     y_\pm &=& \vert \sMsRsml \mp \sMsml \vert,
953:   \end{eqnarray}
954: \end{mathletters}
955: at the $z=0$ plane.
956: Note that the $2\pi$ periodicity of $y=f(w)$ ensures that 
957: ($w_\pm + 2\pi n$, $y_\pm$) for arbitrary integer $n$ are also tangent 
958: points to a line with slope of $-1$.
959: 
960: For given $\mach>1$,
961: the number of solutions of equation (\ref{eq:w}) depends on
962: the slopes of lines connecting the points $(w_- - 2\pi n, y_-)$ and 
963: $(w_+, y_+)$.  When the line passing through $(w_- - 2\pi, y_-)$ and 
964: $(w_+, y_+)$ has a slope larger than $-1$, which occurs for
965: $1 < \mach < \mach_1$ (see below for the definition of $\mach_n$ with 
966: integer $n\geq 1$), the spatial regions that satisfy equations
967: (\ref{eq:ncond}) as well as 
968: $ w_+ + y_+ < -s < w_- + y_-$ have three roots for 
969: $w_i$, while the other regions have only a single root.
970: When $\mach_1 < \mach < \mach_2$, the slope of the line connecting 
971: $(w_- - 4\pi, y_-)$ and $(w_+, y_+)$ is greater than $-1$, so that the 
972: regions bounded by the condition $ w_+ + y_+ < -s < w_- + y_- - 2\pi$
973: possess five roots; the other regions 
974: have three roots.  Note that the volume that does not satisfy conditions
975: (\ref{eq:ncond}) still has a single root.
976: As the Mach number increases further, new regions emerge to have
977: a larger (odd) number of roots for $w_i$.  
978: Generalizing the discussion given above,
979: the critical Mach number $\mach_n$ below which a steady-state wake 
980: possesses up to $2n+1$ roots can be determined  by the requirement
981: $(y_+ - y_-)/(w_+ - w_- + 2\pi n) =-1$.  Using 
982: equations (\ref{eq:wy}), this is simplified to
983: \begin{equation} \label{eq:M}
984:   2 (\mach_n^2 -1)^{1/2}-\cos^{-1}\,(2\mach_n^{-2}-1) = 2 \pi n,
985: \end{equation}
986: where we take $\tR=1$ without any loss of generality since the existence
987: of multiple roots requires $\tR > \mach^{-1}$ (eq. [\ref {eq:ncond}]).
988: The first few solutions of equation (\ref{eq:M}) are $\mach_1=4.6033$,
989: $\mach_2=7.7897$, $\mach_3=10.9499$, etc.
990: 
991: The presence of multiple roots implies that the density wake in those 
992: regions are constructed by a superposition of sonic signals that were 
993: emitted by the perturber at as many different locations (and 
994: corresponding retarded times) 
995: as the number of roots.
996: The $\mach$-dependency of the number of roots for a circular-orbit
997: supersonic
998: perturber is in sharp contrast to the case of a straight-line trajectory
999: where only one (within a sonic sphere) or two (within
1000: a rear Mach cone) roots are allowed regardless of the Mach number (O99).
1001: When a perturber is moving 
1002: at a supersonic speed along a circular orbit, it is able to catch up with
1003: its own wake (possibly multiple times), adding new sonic disturbances
1004: to the regions that were already disturbed by the perturber.
1005: Regions of multiple roots 
1006: usually take a form of a long trailing spiral, as exemplified in Figures 
1007: \ref{evolution},  \ref{steadyM20}, and  \ref{overlap}.  
1008: 
1009: 
1010: \section{HALF-OPENING ANGLE} \label{sec:angle}
1011: 
1012: Although a Mach cone is generally curved for a circular-orbit
1013: perturber, its half-opening angle in the vicinity of the perturber 
1014: is the same as in the linear trajectory counterpart.  
1015: As explained in \S\ref{sec:superwake},  the interior of a curved Mach cone 
1016: and an attached trailing tail is the region where equation (\ref{eq:w}) 
1017: has multiple roots in a steady state.  It is bounded by 
1018: $-s=w_\pm + y_\pm$ curves in the $z=0$ plane.  
1019: More specifically, the $-s=w_+ + y_+$ curve describes the inner and outer 
1020: edges of the
1021: Mach cone head near the perturber as well as the outer edge of the tail, 
1022: represented by light solid curve and dashed curve at the
1023: boundaries of the shade region displayed in \Fig{sketch}.
1024: On the other hand, the $-s=w_- + y_-$ curve corresponds to the inner 
1025: edge of the tail (dot-dashed curve in Fig.~\ref{sketch}) that 
1026: meets the Mach cone head at $\tR = \mach^{-1}$.
1027: By Taylor expanding $w_+$ and $y_+$ about $\tR=1, s=0$ 
1028: in the $z=0$ plane and adding the resulting expressions together, 
1029: one can show the head of the Mach cone is described by
1030: \begin{equation}\label{eq:half}
1031:   s = - (\mach^2 -1)^{1/2} \vert \delta \vert,
1032: \end{equation}
1033: up to the first order in $\vert \delta \vert\equiv \vert \tR-1\vert$.  
1034: The negative sign in equation (\ref{eq:half})
1035: results since $s$ is measured in the counterclockwise direction 
1036: from the perturber
1037: (Fig.~\ref{sketch}).  Let $\theta$ be the half-opening angle of the Mach 
1038: cone near the perturber. Then, $\tan \theta =  -|\delta|/s
1039: = (\mach^2-1)^{-1/2}$, or $\theta =  \sin^{-1}(1/\mach)$, the same result
1040: as in the linear trajectory case.
1041: 
1042: 
1043: 
1044: \begin{thebibliography}{}
1045: \bibitem[Arena \& Bertin(2007)]{are07} Arena, S.\ E., \& Bertin, G.\
1046: 2007, \aap, 463, 921  
1047: \bibitem[Binney \& Tremaine(1987)]{bin87} 
1048: Binney, J., \& Tremaine, S.\ 1987, Galactic Dynamics 
1049: (Princeton: Princeton Univ.\ Press)
1050: \bibitem[Bullock \& Johnston(2005)]{bul05}
1051: Bullock, J.\ S., Johnston, K.\ V.\ 2005, \apj, 635, 931
1052: \bibitem[Burns(1976)]{bur76} Burns, J.\ A.\ 1976, American Journal of Phys.,
1053:  44, 944
1054: \bibitem[Chandrasekhar(1943)]{chandra} Chandrasekhar, S.\ 1943, \apj, 97, 255
1055: \bibitem[Chang(2001)]{chang} Chang, H.-Y.\ 2001, \apjl, 551, L159
1056: \bibitem[Chang \& Choi(2003)]{cha03} Chang, H.-Y., Choi, C.-S.\ 
1057: 2003, \aap, 410, 519
1058: \bibitem[Del Popolo et al.(2003)]{planet} Del Popolo, A., Ye{\c s}ilyurt, S., \& Ercan, E.~N.\ 2003, \mnras, 339, 556
1059: \bibitem[Dokuchaev(1964)]{dokuchaev} Dokuchaev, V.~P.\ 1964, Soviet Astron., 8, 23
1060: \bibitem[Dotti et al.(2006)]{dotti} Dotti, M., Colpi, M., \& Haardt, F.\ 2006, \mnras, 367, 103
1061: \bibitem[El-Zant et al.(2004)]{elz04} El-Zant, A.~A., Kim, W.-T., \& Kamionkowski, M.\ 2004, \mnras, 354, 169
1062: \bibitem[Escala et al.(2004)]{escala} Escala, A., Larson, R.~B., Coppi, P.~S., \& Mardones, D.\ 2004, \apj, 607, 765
1063: \bibitem[Escala et al.(2005)]{escala05} Escala, A., Larson, R.~B., Coppi, P.~S., \& Mardones, D.\ 2005, \apj, 630, 152
1064: \bibitem[Faltenbacher et al.(2005)]{fal05}
1065:   Faltenbacher, A., Kravtsov, A.\ V., Nagai, D., \& Gottloeber, S.\
1066:   2005, \mnras, 358, 139
1067: \bibitem[Fujii et al.(2006)]{fujii} Fujii, M., Funato, Y., \& Makino, J.\ 2006, \pasj, 58, 743
1068: \bibitem[Furlanetto \& Loeb(2002)]{fur02} Furlanetto, S.~R., \& Loeb, A.\ 2002, \apj, 565, 854
1069: \bibitem[Goldreich et al.(2002)]{gol02} Goldreich, P., Lithwick, Y., \& Sari, R.\ 2002, \nat, 420, 643
1070: \bibitem[Halkola et al.(2007)]{halkola} 
1071:    Halkola, A., Seitz, S., \& Pannella, M.\ 2007, \apj, 656, 739
1072: \bibitem[Hashimoto et al.(2003)]{hashimoto} Hashimoto, Y., Funato, Y., \& Makino, J.\ 2003, \apj, 582, 196
1073: \bibitem[Hernquist \& Weinberg(1989)]{her89}
1074:   Hernquist, L., \& Weinberg, M.\ D.\ 1989, \mnras, 238, 407
1075: \bibitem[Jackson(1975)]{jackson} Jackson, J.~D.\ 1975, Classical Electrodynamics (New York: Wiley)
1076: \bibitem[Just \& Pe{\~n}arrubia(2005)]{just} Just, A., \& Pe{\~n}arrubia, J.\ 2005, \aap, 431, 861
1077: \bibitem[Karas \& \v{S}ubr(2001)]{kar01}
1078:   Karas, V., \v{S}ubr, L.\ 2001, \aap, 376, 686
1079: \bibitem[Kim \& Morris(2003)]{kimss03} Kim, S.~S., \& Morris, M.\ 2003, \apj, 597, 312
1080: \bibitem[Kim et al.(2004)]{kimss} Kim, S.~S., Figer, D.~F., \& Morris, M.\ 2004, \apjl, 607, L123
1081: \bibitem[Kim et al.(2005)]{kim} Kim, W.-T., El-Zant, A.~A., \& Kamionkowski, M.\ 2005, \apj, 632, 157
1082: \bibitem[Lin \& Tremaine(1983)]{lin83}
1083:  Lin, D.\ N.\ C., Tremaine, S.\ 1983, \apj, 264, 364
1084: \bibitem[Liu(2004)]{liu} Liu, F.~K.\ 2004, \mnras, 347, 1357
1085: \bibitem[McMillan \& Portegies Zwart(2003)]{mcm03}
1086:   McMillan, S.\ L.\ W., Portegies Zwart, S.\ F.\ 2003, \apj, 596, 314
1087: \bibitem[Mulder(1983)]{mul83} Mulder, W.\ A.\ 1983, \aap, 117, 9
1088: \bibitem[Narayan(2000)]{narayan} Narayan, R.\ 2000, \apj, 536, 663
1089: \bibitem[Navarro et al.(1996)]{NFW} Navarro, J.~F., Frenk, C.~S., \& White, S.~D.~M.\ 1996, \apj, 462, 563
1090: \bibitem[Ostriker(1999)]{ost99} Ostriker, E.~C.\ 1999, \apj, 513, 252 (O99)
1091: \bibitem[Rephaeli \& Salpeter(1980)]{rep80} Rephaeli, Y., \& Salpeter, E.~E.\ 1980, \apj, 240, 20
1092: \bibitem[Ruderman \& Spiegel(1971)]{ruderman} Ruderman, M.~A., \& Spiegel, E.~A.\ 1971, \apj, 165, 1
1093: \bibitem[S{\'a}nchez-Salcedo \& Brandenburg(1999)]{sanchez99} S{\'a}nchez-Salcedo, F.~J., \& Brandenburg, A.\ 1999, \apjl, 522, L35
1094: \bibitem[S{\'a}nchez-Salcedo \& Brandenburg(2001)]{sanchez} S{\'a}nchez-Salcedo, F.~J., \& Brandenburg, A.\ 2001, \mnras, 322, 67
1095: \bibitem[Spinnato et al.(2003)]{spi03}
1096:    Spinnato, P.\ F., Fellhauer, M., Portegies Z., \& Simon F.\
1097:    \mnras, 344, 22
1098: \bibitem[Tremaine(1976)]{tremaine} Tremaine, S.~D.\ 1976, \apj, 203, 72
1099: \bibitem[Weinberg(1989)]{weinberg} Weinberg, M.~D.\ 1989, \mnras, 239, 549
1100: \bibitem[Zentner \& Bullock(2003)]{zen03}
1101:    Zentner, A., \& Bullock, J.~S.\ 2003, \apj, 598, 49
1102: \end{thebibliography}
1103: 
1104: %fig 1
1105: \clearpage
1106: \begin{figure}
1107: \epsscale{0.8}
1108:   \plotone{f1.eps}
1109:   \caption{\label{sketch}
1110: Schematic diagram illustrating the situation on the $z=0$ plane at 
1111: time $t$.  A perturber initially introduced at $(R_p, 0, 0)$ moves 
1112: along a circle with radius $R_p$ at uniform angular speed
1113: $\Omega$ in the counterclockwise direction, and is currently at
1114: the position $\vct{x}_p\equiv (R_p, \Omega t, 0)$.
1115: At this time, an observer sitting at $\vct{x}=(R, \varphi, z)$ receives 
1116: a sonic signal that was emitted by the perturber when it was at 
1117: $\vct{x'}=(R_p, \Omega t', 0)$, where 
1118: $t'=t-\vert\vct{x}-\vct{x'}\vert/\cs$ is the retarded time. 
1119: The angular variables are $w\equiv \varphi'-\varphi$ and 
1120: $s\equiv \varphi-\Omega t$ along the orbital plane. 
1121: The shaded area represents a curved Mach cone and a wake tail 
1122: formed by a supersonic perturber with $\mach>1$.  
1123: The inner edge (dashed curve) of the Mach cone
1124: meets the inner edge (dot-dashed curve) of the tail 
1125: at a point on a circle with radius $R_p/\mach$.
1126: The outer edge (light solid curve) of the Mach cone defines 
1127: the outer edge of the tail.}
1128: \end{figure}
1129: 
1130: %fig 2
1131: \clearpage
1132: \begin{figure}
1133: \epsscale{1.0}
1134:   \plotone{f2.eps}
1135:   \caption{\label{evolution}
1136: Temporal evolution at the $z=0$ plane of (a) the 
1137: dimensionless density wake $\D(\vct{x},t)$
1138: in logarithmic color scale and (b) 
1139: the number of roots for $w_i$ in equation (\ref{eq:w}) 
1140: for the case of $\mach=2.0$.
1141: The time is in unit of $R_p/\cs$.  See text for details.}
1142: \end{figure}
1143: 
1144: %fig 3
1145: \clearpage
1146: \begin{figure*}
1147:   \plotone{f3.eps}
1148:   \caption{\label{steadyM20} 
1149: Density distributions of the steady-state wake for $\mach=2.0$ 
1150: on the $x/R_p=1$ (\textit{bottom right}), 
1151: $y=0$ (\textit{top left}), and $z=0$ (\textit{bottom left}) planes.  
1152: The perturber is located at $(x,\,y)=(R_p,\,0)$,
1153: and the black circle in the bottom left frame denotes the orbit of
1154: the perturber.  Color bar labels $\log \D$.}
1155: \end{figure*}
1156: 
1157: %fig 4
1158: \clearpage
1159: \begin{figure*}
1160: \epsscale{1.}
1161:   \plotone{f4.eps}
1162:   \caption{\label{overlap}
1163: Distributions of the perturbed density $\D$ in logarithmic color scale 
1164: (\textit{top}) and the corresponding number of roots for 
1165: $w_i$ in equation (\ref{eq:w}) (\textit{bottom})
1166: at the $z=0$ plane of the steady-state wake for 
1167: (a) $\mach=4.0$ and (b) $\mach=5.0$.
1168: The perturber is located at $(x,\,y)=(R_p,\,0)$.}
1169: \end{figure*}
1170: 
1171: %fig 5
1172: \clearpage
1173: \begin{figure}
1174:   \plotone{f5.eps}
1175:   \caption{\label{steadyM05} 
1176: Same as Fig.~\ref{steadyM20} except for $\mach=0.5$.}
1177: \end{figure}
1178: 
1179: %fig 6
1180: \clearpage
1181: \begin{figure}
1182:   \plotone{f6.eps}
1183:   \caption{\label{time}
1184: Time evolution of the drag force for $\mach=0.5$, 1.2, 2.0, 4.0, and 5.0. 
1185: Solid curves draw the azimuthal drag $\I_\varphi$, while
1186: dashed curves are for the radial drag $\I_R$. 
1187: For all cases, $\rmin=R_p/10$ is taken.  Note that 
1188: both $\I_\varphi$ and $\I_R$ converge typically within the
1189: sound crossing time across $2R_p$, 
1190: indicating that a steady state is reached quite rapidly.}
1191: \end{figure}
1192: 
1193: %fig 7
1194: \clearpage
1195: \begin{figure} 
1196: \epsscale{1.}
1197:   \plotone{f7.eps}  
1198:   \caption{\label{rmin}
1199: Dependence on $\rmin$ of the steady-state drag force 
1200: for $\mach=0.5$, 2.0, and 4.0.
1201: \textit{Left}: For $R_p/\rmin > 10$, the radial drag force
1202: $\I_R$ converges to a respective constant value marked by dotted line.
1203: \textit{Right}: The azimuthal drag force $\I_\varphi$ 
1204: varies as $\ln\,(R_p/\rmin)$  
1205: for small $\rmin$ when $\mach>1$, while independent of $\rmin$ 
1206: for $\mach<1$.  Dotted lines indicate a slope of unity.}
1207: \end{figure}
1208: 
1209: %fig 8
1210: \clearpage
1211: \begin{figure} 
1212: \epsscale{1.}
1213:   \plotone{f8.eps}  
1214:   \caption{\label{forces}
1215: Gravitational drag force on a circular-orbit perturber in a gaseous
1216: medium as a function of the Mach number $\mach$.  
1217: The open and filled circles give the results of our semi-analytic 
1218: calculation for the drag in the radial and azimuthal directions,
1219: respectively.  
1220: For all the points, $\rmin/R_p=0.1$ is taken.
1221: Solid lines draw the fits, equations (\ref{eq:IR}) and (\ref{eq:Iphi}),
1222: to the semi-analytic results.  Dotted line corresponding to 
1223: the force formula with $V_p t=2R_p$ in O99 for the case of 
1224: linear-trajectory perturbers is in quite a good agreement with the 
1225: azimuthal drag for circular-orbit perturbers.}
1226: \end{figure}
1227: 
1228: %fig 9
1229: \clearpage
1230: \begin{figure}
1231: \epsscale{1.}
1232:   \plotone{f9.eps}
1233:   \caption{\label{orbitNFW}
1234: Orbital decay of a model galaxy caused by dynamical friction due to an
1235: intracluster gas. Solid line corresponds to the case when 
1236: both $\I_R$ (eq.\ [\ref{eq:IR}]) and 
1237: $\I_\varphi$ (eq.\ [\ref{eq:Iphi}]) are considered, while dashed line shows 
1238: the result with only $\I_\varphi$ (i.e., with $\I_R=0$).  
1239: For comparison, the results of Ostriker's formula with fixed $V_p t=1$ Mpc 
1240: and varying $V_p t = 2R(t)$ are plotted as dot-dashed and dotted lines,
1241: respectively.}
1242: \end{figure}
1243: 
1244: \end{document}
1245: