1: % 8/1/07
2: \documentclass[usegraphicx,usenatbib]{mn2e}
3: %%%\documentstyle[useAMS,referee,psfig,graphicx]{mn2e}
4: %%%%%%%%%%%%%%%%%%%%%%%%%%%
5: \renewcommand\[{\begin{equation}}
6: \renewcommand\]{\end{equation}}
7: \def\parn{\par\noindent}
8: \font\boldsym=cmmib10\font\boldsym=cmmib10
9: \def\parn{\par\noindent}
10: \def\xv{{\bf x}}
11: \def\yv{{\bf y}}
12: \def\av{{\bf a}}
13: \def\xci{x_i}
14: \def\yci{y_i}
15: \def\Phic{\Phi_c}
16: \def\Psic{\Psi_c}
17: \def\phieff{\Phi_{\rm eff}}
18: \def\rhoc{\rho_c}
19: \def\zetac{\zeta_c}
20: \def\rc{r_c}
21: \def\vel{v_{\varphi}}
22: \def\velq{\vel^2}
23:
24: \catcode`\@=11
25: \def\gsim{\ifmmode{\mathrel{\mathpalette\@versim>}}
26: \else{$\mathrel{\mathpalette\@versim>}$}\fi}
27: \def\lsim{\ifmmode{\mathrel{\mathpalette\@versim<}}
28: \else{$\mathrel{\mathpalette\@versim<}$}\fi}
29: \def\@versim#1#2{\lower 2.9truept \vbox{\baselineskip 0pt \lineskip
30: 0.5truept \ialign{$\m@th#1\hfil##\hfil$\crcr#2\crcr\sim\crcr}}}
31: \catcode`\@=12
32:
33: \arraycolsep0.35mm % makes multiple equations look nice
34: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
35:
36: \title[Exact density-potential pairs]{Exact density-potential pairs
37: from the holomorphic Coulomb field}
38:
39: \author[Ciotti \& Giampieri]
40: {Luca Ciotti$^1$ and Giacomo Giampieri$^2$
41: \\ $^1$Astronomy Department, University of Bologna,
42: via Ranzani 1, 40127 Bologna, Italy
43: \\ $^2$Jet Propulsion Laboratory, California Institute of
44: Technology, Pasadena CA 91109, USA
45: }
46:
47: \date{Accepted version, January 12, 2007}
48: \pubyear{2007}
49:
50: \begin{document}
51: \maketitle
52:
53: \begin{abstract}
54:
55: We show how the complex-shift method introduced by Appell in gravity
56: to the case of a point mass (and applied among others in
57: electrodynamics by Newman, Carter, Lynden-Bell, and Kaiser to
58: determine remarkable properties of the electromagnetic field of
59: rotating charged configurations), can be extended to obtain new and
60: explicit density-potential pairs for self-gravitating systems
61: departing significantly from spherical symmetry. The rotational
62: properties of two axysimmetric baroclinic gaseous configurations
63: derived with the proposed method are illustrated.
64:
65: \end{abstract}
66:
67: \begin{keywords}
68: celestial mechanics -- stellar dynamics -- galaxies: kinematics and dynamics
69: \end{keywords}
70:
71: \section{Introduction}
72:
73: For the discussion of many astrophysical problems where gravity is
74: important, a major difficulty is set by the potential theory. In
75: general, to calculate the gravitational potential associated with a
76: given density distribution one has to evaluate a three-dimensional
77: integral. Except for special circumstances, where a solution can be
78: found in terms of elementary functions, one has to resort to numerical
79: techniques and sophisticated tools, such as expansions in orthogonal
80: functions or integral transforms.
81:
82: Under spherical symmetry, the density-potential relation can be
83: reduced to a one-dimensional integral, while for axisymmetric systems
84: one in general is left with a (usually non-trivial) two-dimensional
85: integral. As a result, the majority of the available explicit
86: density-potential pairs refers to spherical symmetry and only few
87: axially symmetric pairs are known (e.g., see Binney \& Tremaine 1987,
88: hereafter BT). In special cases (in particular, when a
89: density-potential pair can be expressed in a suitable parametric form)
90: there exist systematic procedures to generate new non-trivial
91: density-potential pairs (e.g., see the case of Miyamoto \& Nagai 1975
92: and the related Satoh 1980 disks; see also Evans \& de Zeeuw 1992; de
93: Zeeuw \& Carollo 1996). For non-axisymmetric systems the situation is
94: worse. One class of triaxial density distributions for which the
95: potential can be expressed in a tractable integral form is that of the
96: stratified homeoids, such as the Ferrers (1887) distributions (e.g.,
97: see Pfenniger 1984, Lanzoni \& Ciotti 2003, and references therein)
98: and special cases of the family considered by de Zeeuw \& Pfenniger
99: (1988). Additional explicit density-potential pairs are given by the
100: Evans (1994) models and by those constructed with the Kutuzov-Osipkov
101: method (1980, see also Kutuzov 1998). Another way to construct in a
102: systematic way explicit density-potential pairs with finite deviations
103: from spherical symmetry has been presented in Ciotti \& Bertin
104: (2005). This technique is based on an elementary property of the
105: asymptotic expansion of the homeoidal potential quadrature formula for
106: small flattenings\footnote{Such expansion can be traced back to the treatise
107: on geodesy by Sir H. Jeffreys (1970, and references therein; see also
108: Hunter 1977).}, and recently, it has been applied to the modeling of gaseous
109: halos in clusters (Lee \& Suto 2003, 2004), to the study of the
110: dynamics of elliptical galaxies (Muccione \& Ciotti 2003, 2004), and
111: finally to a possible interpretation of the rotational field of the
112: extraplanar gas in disk galaxies (Barnab\`e et al. 2005, 2006).
113:
114: In the context of classical electrodynamics a truly remarkable and
115: seamingly unrelated result has been obtained by Newman (1973, see also
116: Newman \& Janis 1965, Newman et al. 1965), who considered the case of
117: the electromagnetic field of a point charge displaced on the imaginary
118: axis. Successive analysis of Carter (1968), Lynden-Bell (2000, 2002,
119: 2004ab, and references therein; see also Teukolsky, 1973;
120: Chandrasekhar, 1976; Page, 1976), and Kaiser (2004, and references
121: therein) revealed the astonishing properties of the resulting
122: holomorphic "magic" field. As pointed out to us by the Referee, the
123: complex-shift method was first introduced by Appell (1887, see also
124: Whittaker \& Watson 1950) to the case of the gravitational field of a
125: point mass, and successively used in General Relativity (e.g., see
126: Gleiser \& Pullin 1989; Letelier \& Oliveira 1987, 1998; D'Afonseca,
127: Letelier \& Oliveira, 2005 and references therein).
128:
129: In this paper we show how the complex-shift method can be also applied
130: in classical gravitation to obtain exact and explicit
131: density-potential pairs deviating significantly from spherical
132: simmetry; we are not aware that this possibility has been noticed in
133: the literature. The paper is organized as follows. In Sect. 2 we
134: present the idea behind the method, and we discuss its application to
135: the case of spherical systems, describing two new self-gravitating and
136: axisymmetric systems obtained from the Plummer and Isochrone spheres.
137: In Sect. 3 we focus on the rotational fields of the new pairs when
138: interpreted as gaseous systems. Section 4 summarizes the main results
139: obtained, while in the Appendix we present the peculiar behavior of
140: the singular isothermal sphere under the action of the complex shift.
141:
142: \section{General considerations}
143:
144: We start by extending the complexification of a point charge Coulomb
145: field discussed by Lynden-Bell (2004b), to the gravitational potential
146: $\Phi(\xv)$ generated by a density distribution $\rho(\xv)$. From now
147: on $\xv =(x,y,z)$ will indicate the position vector, while
148: $<\xv,\yv>\equiv\xci\yci$ is the standard inner product over the reals
149: (repeated index summation convention implied).
150:
151: Let assume that the (nowhere negative) density distribution $\rho
152: (\xv)$ satisfies the Poisson equation
153: \[
154: \nabla^2\Phi=4\pi G\rho,
155: \label{poisson}
156: \]
157: and that the associated complexified potential $\Phic$ with shift $i
158: \av$ is defined as
159: \[
160: \Phic(\xv)\equiv \Phi(\xv -i\av),
161: \]
162: where $i^2=-1$ is the imaginary unit and $\av$ is a real vector. The
163: idea behind the proposed method is based on the recognition that
164: 1) the Poisson equation is a linear PDE, and that 2) the complex shift
165: is a linear coordinate transformation. From these two properties,
166: and from equations (1) and (2) it follows that
167: \[
168: \nabla^2\Phic=4\pi G\rhoc,
169: \]
170: where
171: \[
172: \rhoc(\xv)\equiv \rho(\xv -i\av).
173: \]
174: Thus, by separating the real and imaginary parts of $\Phic$ and
175: $\rhoc$ obtained from the shift of a known real density-potential pair
176: one obtains {\it two} real density-potential pairs.
177:
178: A distinction is in order here between electrostatic and gravitational
179: problems: in fact, while in the former case a density (charge)
180: distribution with negative and positive regions can be (at least
181: formally) accepted, in the gravitational case the obtained density
182: components have physical meaning only if they do not change sign (see
183: however Sect. 4). Quite interestingly, some general result about the
184: sign of the real and imaginary parts of the shifted density can be
185: obtained by considering the behavior of the complexified
186: self-gravitational energy and total mass. In fact, from the linearity
187: of the shift, it follows that the volume integral over the entire
188: space
189: \begin{eqnarray}
190: W_c&\equiv&{1\over 2}\int\rhoc\Phic d^3\xv\nonumber\\
191: &=&{1\over 2}\int [\Re(\rhoc)\Re(\Phic) - \Im(\rhoc)\Im(\Phic)] d^3\xv
192: \end{eqnarray}
193: coincides with the self-gravitational energy $W=0.5\int\rho\Phi
194: d^3\xv$ of the real unshifted seed density. Thus the imaginary part
195: of $W_c$ is zero, i.e.
196: \begin{eqnarray}
197: {1\over 2}\int [\Re(\rhoc)\Im(\Phic)+ \Im(\rhoc)\Re(\Phic)]
198: d^3\xv&=&\nonumber\\
199: -G\int\int{\Re[\rhoc (\xv)]\Im[\rhoc(\xv ')]\over ||\xv-\xv'||}
200: d^3\xv d^3\xv '&=&0,
201: \label{wcim}
202: \end{eqnarray}
203: and $W_c$ is the difference of the gravitational energies of the real
204: and the imaginary parts of the shifted density. The vanishing of the
205: double integral (\ref{wcim}) shows that the integrand necessarily
206: changes sign, i.e., the complex shift cannot generate {\it two}
207: physically acceptable densities. Additional informations are provided
208: by considering that, for the same reasons just illustrated, also the
209: total mass of the complexified distribution $M_c=\int\rhoc d^3\xv$
210: coincides with the total (real) mass of the seed density distribution
211: $M=\int\rho d^3\xv$, so that
212: \[
213: \int\Im(\rhoc) d^3\xv =0.
214: \label{imM}
215: \]
216: Identities~(\ref{wcim})-(\ref{imM}) then leave open the question
217: whether at least $\Re(\rhoc)$ can be characterized by a single sign
218: over the whole space. It results that either cases can happen,
219: depending on the specific seed density and shift vector adopted. In
220: fact, in the following Section we show that simple seed densities
221: exist so that $\Re(\rhoc)$ is positive everywhere, while in the
222: Appendix we present a simple case in which also $\Re(\rhoc)$ changes
223: sign.
224:
225: \subsection{Spherically symmetric ``seed'' potential}
226:
227: We now restrict the previous general considerations to a spherically
228: symmetric real potential $\Phi(r)$, where $r\equiv
229: ||\xv||=\sqrt{<\xv,\xv>}$ is the spherical radius, and $|| ... ||$ is
230: the standard Euclidean norm. After the complex shift the norm must be
231: still interpreted over the reals\footnote{If one adopt the standard
232: inner product over the complex field, one would obtain $||\xv
233: -i\av||^2 = r^2 +||\av||^2$.} (e.g., see Lynden-Bell 2004b), so that
234: \[
235: ||\xv -i\av||^2=r^2 -2i<\xv,\av> -a^2,
236: \]
237: where $a^2\equiv ||\av ||^2$. Without loss of generality we assume
238: $\av=(0,0,a)$, and the shifted radial coordinate becomes
239: \[
240: \rc^2 = r^2-2iar\mu -a^2;\quad\quad \mu\equiv\cos\theta,
241: \]
242: where $\theta$ is the colatitude of the considered point, $\mu r=z$,
243: and
244: \[
245: \Phic=\Phi\left(\sqrt{r^2-2iaz -a^2}\right),
246: \label{eqpsic}
247: \]
248: \[
249: \rhoc=\rho\left(\sqrt{r^2-2iaz -a^2}\right).
250: \label{eqrhoc}
251: \]
252: Note that, when starting from a spherically symmetric seed system, the
253: real and immaginary parts of the shifted density $\rhoc$ can be
254: obtained 1) from evaluation of the Laplace operator applied to the
255: real and imaginary parts of the potential $\Phic$, 2) by expansion of
256: the complexified density in equation~(\ref{eqrhoc}), or finally 3) by
257: considering that for spherically symmetric systems $\rho=\rho(\Phi)$,
258: and so the real and imaginary parts of $\rhoc(\Phic)$ can be expressed
259: (at least in principle) as functions of the real and imaginary part of
260: the shifted potential.
261:
262: \subsection{The shifted Plummer sphere}
263:
264: As a first example, in this Section we apply the complex shift to the
265: Plummer (1911) sphere. We start from the relative potential
266: $\Psi=-\Phi$, where
267: \[
268: \Psi={GM\over b}{1\over\zeta},\quad\zeta\equiv\sqrt{1+r^2},
269: \label{plumpot}
270: \]
271: and $r$ is normalized to the model scale-lenght $b$, so that the
272: associated density distribution is
273: \[
274: \rho={3M\over 4\pi b^3}{1\over\zeta^5}
275: \]
276: (e.g., see BT). For ease of notation, from now
277: on we will use normalized density (to $M/b^3$) and potential (to
278: $GM/b$), and so
279: \[
280: \rhoc={3\over 4\pi}[\Re(\Psic)+i\Im(\Psic)]^5.
281: \label{rhopcplu}
282: \]
283: From substitution (8) (where also the shift $a$ is expressed in $b$
284: units), the shifted potential $\Psic=1/\zetac$ depends on the square
285: root of $\zetac^2=1-a^2+r^2 -2iaz\equiv d{\rm e}^{i\varphi}$, with
286: \[
287: d\equiv |\zetac|^2=\sqrt{(1-a^2+r^2)^2 +4a^2z^2},
288: \]
289: and
290: \[
291: \cos\varphi={1-a^2+r^2\over d},\quad\sin\varphi=-{2az\over d}.
292: \]
293: Note that $\cos\varphi >0$ everywhere for $a<1$, and in
294: the following discussion we restrict to this case. The square
295: root
296: \[
297: \zetac=\sqrt{d} {\rm e}^{\pi k i+\varphi i/2}, \quad\quad(k=0,1),
298: \label{zeta}
299: \]
300: is made a single-valued function of $(r,z)$ by cutting the complex
301: plane along the negative real axis (which is never touched by
302: $\zetac^2$) and assuming $k=0$, so that the model equatorial plane is
303: mapped into the line $\varphi=0$. With this choice the principal
304: determination of $\Psic$ reduces to $\Psi$ when $a=0$, and simple
305: algebra shows that
306: \[
307: \cos{\varphi\over 2}={\sqrt{1+\cos\varphi}\over\sqrt{2}},\quad
308: \sin{\varphi\over 2}=-{\sqrt{2}az\over d\sqrt{1+\cos\varphi}}.
309: \]
310: The real and imaginary parts of $\Psic$ are then given by
311: \[
312: \Re (\Psic)=\Re\left({\bar\zetac\over |\zetac|^2}\right)
313: ={\sqrt{d+1-a^2+r^2}\over \sqrt{2}\,d},
314: \label{plumre}
315: \]
316: \[
317: \Im (\Psic)=\Im\left({\bar\zetac\over |\zetac|^2}\right)
318: ={az\over d^2\Re(\Psic)},
319: \label{plumim}
320: \]
321: respectively, and from equation~(\ref{rhopcplu}) we obtain the
322: expressions of the (normalized) axysimmetric densities:
323: \[
324: \Re(\rhoc)={3\Re(\Psic)\over 4\pi}\left[
325: \Re(\Psic)^4-{10 a^2z^2\over d^4}+
326: {5a^4z^4\over d^8\Re(\Psic)^4}
327: \right],
328: \label{plumrerho}
329: \]
330: \[
331: \Im(\rhoc)={3\Im(\Psic)\over 4\pi}\left[
332: 5\Re(\Psic)^4-{10a^2z^2\over d^4}+
333: {a^4z^4\over d^8\Re(\Psic)^4}
334: \right].
335: \label{plumimrho}
336: \]
337: We verified that the two new density-potential pairs
338: (\ref{plumre})-(\ref{plumrerho}) and (\ref{plumim})-(\ref{plumimrho})
339: satisfy the Poisson equation~(\ref{poisson}).
340:
341: Note that, at variance with $\Re(\Psic)$, the potential $\Im(\Psic)$
342: change sign crossing the equatorial plane of the system. In accordance
343: with this change of sign, also $\Im(\rhoc)$ is negative for $z<0$, as
344: can be seen from equation~(\ref{plumimrho}), and thus cannot be used
345: to describe a gravitational system; we do not discuss this pair any
346: further, while we focus on the real components of the shifted
347: density-potential pair. Near the center $d\sim 1-a^2
348: +r^2[1+2a^2\mu^2/(1-a^2)]+O(\mu^2 r^4)$, and the leading terms of the
349: asymptotic expansion of equations~(\ref{plumre}) and (\ref{plumrerho})
350: are
351: \[
352: \Re(\Psic)\sim {1\over\sqrt{1-a^2}}
353: -{r^2(1-a^2+3a^2\mu^2)\over 2 (1-a^2)^{5/2}}
354: +O(r^4),
355: \]
356: \[
357: \Re(\rhoc)\sim {3\over 4\pi (1-a^2)^{5/2}}
358: -{15r^2 (1-a^2 +7a^2\mu^2)\over 8\pi (1-a^2)^{9/2}}
359: +O(r^4);
360: \]
361: in particular, the isodenses are oblate ellipsoids with minor-to-major
362: squared axis ratio $(1-a^2)/(1+6a^2)$. For $r\to\infty$, $d\sim r^2
363: + 1-a^2+2a^2\mu^2 + O(\mu^2 r^{-2})$, and
364: \[
365: \Re(\Psic)\sim {1\over r}-{1-a^2+3a^2\mu^2\over 2 r^3}+O(r^{-5}),
366: \]
367: \[
368: \Re(\rhoc)\sim {15\over 4\pi r^5}
369: \left({1\over 5} - {1-a^2+7a^2\mu^2\over 2 r^2}\right)+O(r^{-9}),
370: \]
371: so that $\Re(\rhoc)$ coincides with the unshifted seed density (13)
372: and it is spherically symmetric and positive\footnote{This is a
373: general property of shifted spherical systems, as can be seen just by
374: expanding equation~(\ref{eqrhoc}) for $r\to\infty$.}. Thus, near the
375: center and in the far field $\Re(\rhoc)>0$ for $0\leq a<1$. In
376: addition, on the model equatorial plane $z=0$ (where $d=1-a^2+R^2$,
377: and $R$ is the cylindrical radius), $\Re(\Psic)$ coincides with the
378: potential of a Plummer sphere of scale-lenght $\sqrt{1-a^2}$, and from
379: equation~(\ref{plumrerho}) it follows that
380: $\Re(\rhoc)=3\Re(\Psic)^5/(4\pi)>0$ for $0\leq a<1$.
381:
382: However, for $z\neq 0$ a negative term is present in
383: equation~(\ref{plumrerho}), and the positivity of $\Re(\rhoc)$ is not
384: guaranteed for a generic value of the shift parameter in the range
385: $0\leq a<1$. In fact, a numerical exploration reveals that
386: $\Re(\rhoc)$ becomes negative on the symmetry axis $R=0$ at $z\simeq
387: 0.81$ for $a=a_m\simeq 0.588$; the negative density region then
388: expands around this critical point for increasing $a>a_m$. The
389: isodensity contours of $\Re(\rhoc)$ in a meridional plane are shown in
390: the top panels of Fig. 1, for shift parameter values $a=1/2$ (left)
391: and $a =23/40=0.575$ (right). The most salient property is the
392: resulting toroidal shape of the model with the large shift, which
393: reminds similar structures known in the literature, as the Lynden-Bell
394: (1962) flattened Plummer sphere, the densities associated with the
395: Binney (1981) logarithmic potential and the Evans (1994) scale-free
396: potentials, the Toomre (1982) and Ciotti \& Bertin (2005) tori (see
397: also Ciotti, Bertin, \& Londrillo 2004), and the exact MOND
398: density-potential pairs discussed in Ciotti, Nipoti \& Londrillo
399: (2006). Unfortunately, we were not able to find the explanation (if
400: any) behind this similarity (see however Sect. 4).
401:
402: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
403: \begin{figure*}
404: \vskip -0.7truecm
405: \hskip -1.7truecm
406: %%%\psfig{file=fig1.eps,width=0.9\hsize}
407: \includegraphics[height=0.7\textheight,width=0.9\textwidth]{fig1}
408: \caption{Isodensity contours in the $(R,z)$ plane of $\Re(\rho_c)$ of
409: the shifted Plummer sphere for $a=1/2$ (top left) and $a=23/40$ (top
410: right), and of the shifted Isochrone sphere for $a=1/2$ (bottom left)
411: and $a=4/5$ (bottom right). The coordinates are normalized to the
412: scale-lenght $b$ of the corresponding seed spherical model.}
413: \label{isodensity}
414: \end{figure*}
415: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
416:
417: \subsection{The shifted Isochrone sphere}
418:
419: Following the treatment of the Plummer sphere, we now consider the
420: slightly more complicate case of the shifted Isochrone sphere. Its
421: relative potential and density are given by
422: \[
423: \Psi={GM\over b}{1\over 1+\zeta},
424: \]
425: \[
426: \rho={M\over 4\pi b^3}{3(1+\zeta)+2r^2\over
427: (1+\zeta)^3\zeta^3},
428: \]
429: where $\zeta$ is defined in equation (12), and again all lenghts are
430: normalized to the scale $b$ (H\'enon 1959, see also BT). In
431: analogy with equation (14), the normalized shifted density can be
432: written as a function of the normalized shifted potential
433: \[
434: \rhoc={\Psic^2\over 4\pi\zetac^3}\left(3+2\rc^2\Psic\right).
435: \]
436: The real and imaginary parts of
437: $\Psic=(1+\bar\zetac)/[1+|\zetac|^2+2\Re(\zetac)]$ are easily obtained
438: from equations (15)-(18) as
439: %\[
440: %\Re(\Psic)={1+\sqrt{d}\cos(\varphi/2)\over
441: % 1 +2\sqrt{d}\cos (\varphi/2)+d},
442: %\]
443: %\[
444: %\Im(\Psic)={\sqrt{2}az\over d\sqrt{1+\cos\varphi}
445: % [1+2\sqrt{d}\cos (\varphi/2) +d]}.
446: %\]
447: \[
448: \Re(\Psic)={1+\sqrt{d+1-a^2+r^2}/\sqrt{2}\over
449: 1 +d+\sqrt{2(d+1-a^2+r^2)}},
450: \]
451: \[
452: \Im(\Psic)={\sqrt{2}az\over
453: \sqrt{d(d+1-a^2+r^2)}
454: [1+d+\sqrt{2(d+1-a^2+r^2)}]},
455: \]
456: while $\Re(\rhoc)$ and $\Im(\rhoc)$ can be obtained by expansion of
457: equation (29); however their expression is quite cumbersome, and so
458: not reported here. Again, $\Im(\Psic)$ changes sign when crossing the
459: model equatorial plane, revealing that also $\Im(\rhoc)$ changes sign
460: in order to produce the resulting vertical force field near $z=0$, and
461: the vanishing of the integral in equation~(\ref{imM}). In accordance
462: with equation~(\ref{eqrhoc}) on the model equatorial plane
463: $\Re(\rhoc)$ is functionally identical to an Isochrone sphere of
464: scale-lenght $\sqrt{1-a^2}$ (and so positive for $0\leq a<1$), while
465: for $r\to\infty$
466: \[
467: \Re(\Psic)\sim {r-1\over r^2}+{1+a^2-3a^2\mu^2\over 2 r^3}+O(r^{-4}),
468: \]
469: \[
470: \Re(\rhoc)\sim {1\over 4\pi r^5}
471: \left(2r-3 +4a^2{1-6\mu^2\over r}\right)+O(r^{-7}).
472: \]
473: As for the shifted Plummer sphere, also in this case $\Re(\rhoc)$
474: coincides in the far field with the seed density, and so it is
475: positive for $0\leq a<1$, while it becomes negative on the $z$ axis
476: at $z\simeq 0.648$ for $a>a_m\simeq 0.804$. In the bottom panels of
477: Fig. 1 we show the isodensity contours in the meridional plane of
478: $\Re(\rhoc)$ for the representative values of the shift parameter
479: $a=1/2$ (left) and $a=4/5$ (right): again, the toroidal shape and the
480: critical regions on the symmetry axis are apparent in the case of the
481: larger shift.
482:
483: \section{Rotational fields}
484:
485: A natural question to ask is about the nature of the kinematical
486: fields that can be supported by the found density-potential pairs
487: $\Re(\Psic)-\Re(\rhoc)$ when considered as self-gravitating stellar
488: systems. In general, the associated two-integrals Jeans equations
489: (e.g., BT, Ciotti 2000) can be solved numerically
490: after having fixed the relative amount of ordered streaming motions
491: and velocity dispersion in the azimuthal direction (e.g., see Satoh
492: 1980, Ciotti \& Pellegrini 1996). For simplicity here we restrict to
493: the investigation of the isotropic case, when the resulting Jeans
494: equations are formally identical to the equations describing
495: axysimmetric, self-gravitating gaseous systems in permanent rotation:
496: \[
497: \left\{
498: \begin{array}{lcl}
499: \displaystyle {1\over\rho}\, {\partial P\over\partial z} & = &
500: \displaystyle -{\partial\Phi\over\partial z} \: ,\\
501: &&\\
502: \displaystyle {1\over\rho}\, {\partial P\over\partial R} & = &
503: \displaystyle -{\partial\Phi\over\partial R} + \Omega^2 R ,
504: \end{array}
505: \right.
506: \label{eq.hydro}
507: \]
508: (in order to simplify the notation, in the following we intend
509: $\rho=\Re[\rhoc]$ and $\Phi=\Re[\Phic]$). The quantities $\rho$, $P$
510: and $\Omega$ denote the fluid density, pressure and angular velocity,
511: respectively; the rotational (i.e., streaming) velocity is
512: $\vel=\Omega R$, while $v_R =v_z =0$.
513:
514: In problems where $\rho$ and $\phi$ are assigned, the standard
515: approach for the solution of equations~(\ref{eq.hydro}) is to
516: integrate for the pressure with boundary condition $P(R,\infty)=0$,
517: \[
518: P=\int_z^{\infty}\rho {\partial\Phi\over \partial z'} \, {\rm d} z',
519: \label{pressione}
520: \]
521: and then to obtain the rotational velocity field from the radial
522: equation
523: \[
524: \velq={R\over\rho}{\partial P\over\partial R} +
525: R{\partial\Phi\over\partial R}.
526: \label{velocita}
527: \]
528: However, $\velq$ can be also obtained without previous knowledge of
529: $P$, because by combining equations~(\ref{pressione})-(\ref{velocita})
530: and integrating by parts with the assumption that
531: $P=\rho\partial\Phi/\partial R=0$ for $z=\infty$ it follows that
532: \[
533: {\rho\velq\over R}=\int_z^{\infty}\left(
534: {\partial\rho\over\partial R}
535: {\partial\Phi\over\partial z'}-
536: {\partial\rho\over\partial z'}
537: {\partial\Phi\over\partial R}
538: \right)\, {\rm d}z'.
539: \label{commutatore}
540: \]
541: We remark that this ``commutator-like'' relation is not new (e.g., see
542: Rosseland 1926, Waxmann 1978, and, in the context of stellar dynamics,
543: Hunter 1977).
544:
545: Before solving equations~(\ref{eq.hydro}) for the real parts of the
546: density-potential pairs of Sect. 2, it is useful to recall some basic
547: property of rotating fluid configurations. For example, in several
548: astrophysical application (e.g., the set-up of initial conditions for
549: hydrodynamical simulations), equations~(\ref{eq.hydro}) are solved for
550: the density under the assumption of a barotropic pressure distribution
551: (and neglecting the gas self-gravity), and for assigned
552: $\rho(R,\infty)=0$ or $\rho(R,0)=\rho_0(R)$.
553: %Thus, a gravitational potential $\Phi$ (for example produced by a
554: %stellar disk or a dark matter halo) and a specific function $P(\rho)$
555: %are fixed, the first of equations~(\ref{eq.hydro}) is integrated with
556: %boundary $\rho(R,0)$ or imposing $\rho(R,\infty)=0$, and finally the
557: %angular velocity $\Omega$ is obtained from the second of
558: %equations~(\ref{eq.hydro}).
559:
560: As well known, this approach can lead only to hydrostatic ($\Omega
561: =0$) or {\it cylindrical rotation} ($\Omega =\Omega[R]$) kinematical
562: fields. In fact, according to the Poincar\'e-Wavre theorem (e.g., see
563: Lebovitz 1967, Tassoul 1980), cylindrical rotation is {\it equivalent}
564: to barotropicity, or to the fact that the acceleration field at the
565: r.h.s. of equations~(\ref{eq.hydro}) derives from the effective
566: potential
567: \[
568: \phieff\equiv\Phi - \int_{R_0}^R\Omega^2(R')R'dR'
569: \]
570: (where $R_0$ is an arbitrary but fixed radius), so that the gas
571: density and pressure are stratified on $\phieff$.
572:
573: However, barotropic equilibria are just a very special class of
574: solutions of equations~(\ref{eq.hydro}). For example, in the context
575: of galactic dynamics isotropic axisymmetric galaxy models often show
576: streaming velocities dependent on $z$ (for simple examples see, e.g.,
577: Lanzoni \& Ciotti 2003, Ciotti \& Bertin 2005). Such {\it baroclinic}
578: configurations (i.e., fluid systems in which $P$ cannot be expressed
579: as a function of $\rho$ only, and $\phieff$ does not exist at all),
580: have been studied in the past for problems ranging from geophysics, to
581: the theory of sunspots and stellar rotation (e.g., see Rosseland
582: 1926), and to the problem of modeling the decrease of rotational
583: velocity of the extraplanar gas in disk galaxies for increassing $z$
584: (e.g., see Barnab\'e et al. 2005, 2006). Note that a major problem
585: posed by the construction of baroclinic solutions is the fact that the
586: existence of {\it physically acceptable solutions}
587: (i.e. configurations for which $\velq\geq 0$ everywhere) is not
588: guaranteed for arbitrary choices of $\rho$ and $\Phi$. However, the
589: positivity of the integrand in equation~(\ref{commutatore}) for $z\geq
590: 0$ is a sufficient condition to have $\velq \geq 0$ everywhere. In
591: fact, in Barnab\'e et al. (2006) several theorems on baroclinic
592: configurations have been proved starting from
593: equation~(\ref{commutatore}): in particular {\it toroidal} gas
594: distributions are strongly favoured in order to have $\velq\geq 0$ in
595: presence of a dominating disk gravitational field. We note that also
596: in the present cases the density is of toroidal shape, altough the
597: distribution is self-gravitating.
598: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
599: \begin{figure*}
600: \vskip -0.7truecm
601: \hskip -1.7truecm
602: \includegraphics[height=0.7\textheight,width=0.9\textwidth]{fig2}
603: \caption{Isorotational contours in the meridional plane of
604: $\Re(\rho_c)$ of Plummer sphere for $a=1/2$ (top left), and $a=23/40$
605: (top right). In the bottom panels contours of constant pressure for
606: the same models.}
607: \label{plumrot}
608: \end{figure*}
609: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
610:
611: In the case of the $\Re(\rhoc)-\Re(\Psic)$ pair of the shifted Plummer
612: sphere, the (normalized) commutator in equation~(\ref{commutatore}) is
613: a very simple function
614: \[
615: {\partial\rho\over\partial R}{\partial\Phi\over\partial z'}-
616: {\partial\rho\over\partial z'}{\partial\Phi\over\partial R}=
617: {15a^2\over\pi} {R z (1-a^2+r^2)\over d^7},
618: \label{complum}
619: \]
620: which is positive for $z>0$ and $0\leq a <1$, so that $\velq\geq 0$ as
621: far as $\Re(\rhoc)$ is positive everywhere, i.e., for $0\leq a<a_m$.
622: Remarkably, the integration of equation~(\ref{commutatore}) can be
623: easily carried out in terms of elementary functions
624: \begin{eqnarray}
625: &&\rho\velq={3 a^2 R^2\over 2\pi d^5}\Bigg\{1 -
626: {d^4 (1+a^2+r^2)\over 12a^4(1+R^2)^3}\times\nonumber\\
627: &&
628: \left[{6a^4(1+R^2)^2\over d^4}-
629: {2a^2(1+R^2)\over d^2}+{1+a^2+r^2-d\over 1+a^2+r^2}
630: \right]
631: \Bigg\},
632: \end{eqnarray}
633: while the direct integration of the pressure equation is much more
634: complicate (again showing the relevance of
635: equation~[\ref{commutatore}] in applications), and we report its
636: normalized expression on the equatorial plane only
637: \[
638: P(R,0)={2(1+R^2)-a^2\over 16\pi (1+R^2)^2(1-a^2+R^2)^2}.
639: \]
640: Note that $\velq\propto a^2$, and so it vanishes (as a consequence of
641: the imposed isotropy) when reducing to the spherically symmetric seed
642: density: in fact, it can be proved that the argument in parenthesis in
643: equation (40) is regular for $a\to 0$. The baroclinic nature of the
644: equilibrium is apparent, and confirmed by Fig. 2, where we show the
645: contours of constant $\velq$ for $a=1/2$ and $a=23/40$ (top panels,
646: correspondig to the models plotted in the top panels of Fig. 1). The
647: normalized
648: streaming velocity field in the equatorial plane is
649: \[
650: \velq(R,0)={a^2 R^2\over 3}
651: {6(1+R^2)^2 - 4a^2(1+R^2) +a^4\over
652: (1+R^2)^3(1-a^2+R^2)^{3/2}},
653: \]
654: and has a maximum at $R\lsim 1$, while the normalized circular
655: velocity is the same as that of a Plummer sphere of scale-lenght
656: $\sqrt{1-a^2}$ (cf. the comment after equation [26]), i.e.,
657: \[
658: v_c^2(R)=R{\partial\Phi(R,0)\over\partial R}={R^2\over (1-a^2+R^2)^{3/2}}.
659: \]
660: It is easy to verify that equations (41) and (42) satisfy equation
661: (36).
662:
663: Unfortunately, we were unable to obtain the explicit formulae for the
664: analogous quantities for the shifted Isochrone model. However, it is
665: easy to prove that also in this case the commutator is proportional to
666: $a^2$, and so it vanishes (as expected) for $a\to 0$. A comparison of
667: the rotational fields of the two models in the far field can be
668: obtained by using the asymptotic expressions (25)-(26) and (32)-(33)
669: can be used. In particular, for the real part of the shifted Plummer
670: sphere we found
671: \[
672: \velq={2a^2R^2\over r^5}-{5a^2 R^2(3-a^2-9a^2\mu^2)\over 3 r^7}+
673: O(R^2r^{-9}),
674: \]
675: while for $\Re(\rhoc)$ of the Isochrone shifted sphere
676: \begin{eqnarray}
677: \velq&=&{a^2R^2\over r^5}
678: \left({4\over 3}-{13\over 5 r}\right)+\nonumber\\
679: &&{a^2R^2(3159+5780a^2+21780a^2\mu^2)\over r^7}+
680: O(R^2r^{-8}).
681: \end{eqnarray}
682: Note that, even though the radial asymptotic behavior of the density
683: is different in the two considered cases, the obtained velocity field
684: has the same radial dependence in the far field, and in both cases
685: $\velq$ decreases for increasing $z$ and fixed $R$. In addition,
686: $\velq\to 0$ for $r\to\infty$, as expected from the spherical symmetry
687: of the density and potential for $r\to\infty$: in fact, it is apparent
688: from equation~(\ref{commutatore}) that spherical (additive) components
689: of $\rho$ and $\Phi$ (e.g., the leading terms in the asymptotic
690: expansions [25]-[26] and [32]-[33]) mutually cancel in the commutator
691: evaluation.
692:
693: \section{Conclusions}
694:
695: In this paper we have shown that the complex-shift method, introduced
696: in electrodynamics by Newman, and throughly studied among others by
697: Carter, Lynden-Bell and Kaiser, can be extended to classical
698: gravitation to produce new and explicit density-potential pairs with
699: finite deviation from spherical symmetry. In particular, we showed
700: that after separation of the real and imaginary parts of the
701: complexified density-potential pair, the imaginary part of the density
702: corresponds to a system of null total mass, while the real component
703: can be positive everywhere (depending on the original seed density
704: distribution and the amount of the complex shift).
705:
706: As simple application of the proposed method we illustrated the
707: properties of the axysimmetric systems resulting from the shift of the
708: Plummer and Isochrone spheres. The analysis revealed that the
709: obtained densities are nowhere negative only for shift values in some
710: restricted range. This property is linked to the presence of a flat
711: "core" in the adopted (spherically symmetric) seed distributions. It
712: is thus expected that also other distributions, such as the King
713: (1972) model, and the Dehenen (1993) and Tremaine et al. (1994)
714: $\gamma=0$ model, will lead to physically acceptable shifted systems.
715: For the two new exact density-potential pairs we also found that the
716: associated rotational fields (obtained under the assumption of global
717: isotropy of the velocity dispersion tensor) correspond to baroclinic
718: gaseous configurations.
719:
720: An interesting issue raised by the present analysis is the toroidal
721: shape of the obtained densities. While we were not able to find a
722: general explanation of this phenomenon, some hints can be obtained by
723: considering the behavior of the complex-shift for ${\bf a}\to 0$. In
724: fact, in this case it can be easily proved that while the {\it odd}
725: terms of the expansion correspond to the imaginary part fo the shifted
726: density (and contain the factor $z$), the {\it even} terms are real
727: and, to the second order in $a$,
728: \[
729: \Re(\rhoc)\sim\rho-{a^2\over 2r}{d\rho\over dr} +
730: {a^2 z^2\over 2 r^3}
731: \left ({d\rho\over dr}-r{d^2\rho\over dr^2}\right) + O(a^4).
732: \]
733: This expansion is functionally similar to that obtained by Ciotti \&
734: Bertin (2005) for the case of oblate homeoidal distributions, where
735: the term dependent on $z^2$ is immediately identified with a spherical
736: harmonic (e.g., Jackson 1999) responsible of the toroidal shape. It
737: is clear that when restricting to the small-shift approximation (46),
738: all the computations presented in this paper can be carried out
739: explicitly for simple seed distributions. Moreover, from equation
740: (46) it is easy to prove that negative values of density appears on
741: the $z$ axis - no matter how much the shift parameter is small - when
742: the seed density has a central cusp $\rho\propto r^{-\gamma}$ with
743: $\gamma >0$ (e.g., see the explicit example in Appendix).
744:
745: In this paper we restricted for simplicity to spherically simmetric
746: seed systems, which produce axysimmetric systems. A natural extension
747: of the present investigation would be the study of the
748: density-potential pairs originated by the complex shift of seed disk
749: distributions, such as the Miyamoto \& Nagay (1975) and Satoh (1980)
750: disks, or even the generalization of homeoidal quadrature formulae
751: (e.g., see Kellogg 1953, Chandrasekhar 1969, BT). Note that in such
752: cases, at variance with the spherical cases here discussed, the shift
753: {\it direction} is important: for example, the complex shift of a disk
754: along the $z$ axis will lead to a different system than
755: an equatorial shift, that would produce a triaxial object.
756:
757: A second interesting line of study could be the use of the
758: complex-shift to produce more elaborate density-potential pairs by
759: adding models with a weight function $w(\av)$, i.e., by considering
760: the behavior of the linear operator
761: \[
762: \varrho_c=\int\rho(\xv -i\av) w(\av) d^3\av,
763: \]
764: where the integration is extended over some suitably chosen region.
765: We do not pursue this line of investigation, but it is obvious that
766: also shifted density components with negative regions can be accepted
767: in this context, as far as one is able to construct a weight function
768: leading to a final distribution nowhere negative (as it happens, for
769: example, in spherical harmonics expansions).
770:
771: \section*{Acknowledgments}
772:
773: The paper is dedicated to the memory of my best friend and colleague
774: Dr. Giacomo Giampieri, who tragically passed away on September 3, 2006
775: while working on this project. L.C. thanks Donald Lynden-Bell for
776: enligthening discussions, and Tim de Zeeuw, Scott Tremaine and an
777: anonymous Referee for interesting comments. This work was supported
778: by the Italian MIUR grant CoFin2004 "Collective phenomena in the
779: dynamics of galaxies".
780:
781: \appendix
782:
783: \section{The shifted singular isothermal sphere}
784:
785: In this case the normalized density and potential are $\rho=1/r^2$ and
786: $\Phi=4\pi\ln r$, so that from equations~(\ref{eqpsic})-(\ref{eqrhoc})
787: \[
788: \rhoc={r^2-a^2\over (r^2-a^2)^2 + 4 a^2 z^2}+ i{2az\over
789: (r^2-a^2)^2 + 4 a^2 z^2},
790: \]
791: \[
792: \Phic=2\pi\ln (r^2-a^2 -2iaz).
793: \]
794: Note how, at variance with the Plummer and Isochrone spheres, in the
795: present case $\Re(\rhoc)$ describes a density distribution negative
796: inside the open ball $r<a$ and positive outside. The density on the
797: surface $r=a$ is zero except on the singular ring $R=a$ in the
798: equatorial plane. In the far field the density decreases as $1/r^2$,
799: while near the center it flattens to $-1/a^2$, while the potential
800: generated by $\Re(\rhoc)$ is given by
801: \[
802: \Re(\Phic)=\pi\ln [(r^2-a^2)^2 +4a^2z^2].
803: \]
804: As for the other two cases discussed in this paper, the density
805: $\Im(\rhoc)$ vanishes on the equatorial plane, is positive above and
806: negative below.
807:
808: \begin{thebibliography}{99}
809:
810: \bibitem{} Appell, P. 1887,
811: Ann. Math. Lpz., 30, 155
812:
813: \bibitem{} Barnab\'e, M., Ciotti, L., Fraternali, F., \& Sancisi, R. 2005,
814: in ``Extra-planar Gas'' Conference, ASP Conf. Series, ed. R.~Braun,
815: vol. 331, p. 231
816:
817: \bibitem{} Barnab\'e, M., Fraternali, F., Ciotti, L., \& Sancisi, R. 2006,
818: A\&A, 446, 61
819:
820: \bibitem{} Binney, J. 1981,
821: MNRAS, 196, 455
822:
823: \bibitem{} Binney, J., \& Tremaine, S. 1987,
824: Galactic Dynamics (Princeton University Press, Princeton) (BT)
825:
826: \bibitem{} Carter, B. 1968,
827: Commun. Math. Phys., 10, 280
828:
829: \bibitem{} Chandrasekhar, S. 1969,
830: Ellipsoidal figures of equilibrium
831: (Yale University Press, New Haven)
832:
833: \bibitem{} Chandrasekhar, S. 1976,
834: Proc. R. Soc. London A, 349, 571
835:
836: \bibitem{} Ciotti, L. 2000,
837: Lectures Notes on Stellar Dynamics,
838: Scuola Normale Superiore (Pisa) editor
839:
840: \bibitem{} Ciotti, L. \& Bertin, G. 2005,
841: A\&A, 437, 419
842:
843: \bibitem{} Ciotti, L. \& Pellegrini, S. 1996,
844: MNRAS, 279, 240
845:
846: \bibitem{} Ciotti, L., Bertin, G., \& Londrillo, P. 2004,
847: in Plasmas in the laboratory and in the universe:
848: new insights and new challenges, ed. G. Bertin, D. Farina, \&
849: R. Pozzoli, AIP Conf. Proc. 703, 322
850:
851: \bibitem{} Ciotti, L., Nipoti, C., \& Londrillo, P. 2006,
852: ApJ, 640, 741
853:
854: \bibitem{} D'Afonseca, L.A., Letelier, P.S., \& Oliveira, S.R. 2005,
855: Class. Quantum Grav., 22, 3803
856:
857: \bibitem{} Dehenen, W. 1993,
858: MNRAS, 265, 250
859:
860: \bibitem{} de Zeeuw, P.T., \& Pfenniger, D. 1988,
861: MNRAS, 235, 949
862:
863: \bibitem{} de Zeeuw, P.T., \& Carollo, C.M. 1996,
864: MNRAS, 281, 1333
865:
866: \bibitem{} Evans, N.W. 1994,
867: MNRAS, 267, 333
868:
869: \bibitem{} Evans, N.W., \& de Zeeuw, P.T. 1992,
870: MNRAS, 257, 152
871:
872: \bibitem{} Ferrers, N.M. 1877,
873: Quart. J. Pure and Appl. Math., 14, 1
874:
875: \bibitem{} Gleiser, R., \& Pullin, J. 1989,
876: Class. Quantum Grav., 6, 977
877:
878: \bibitem{} H\'enon, M. 1959,
879: Ann. d'Astrophys., 22, 126
880:
881: \bibitem{} Hunter, C. 1977,
882: AJ, 82, 271
883:
884: \bibitem{} Jeffreys, Sir H. 1970,
885: The Earth (fifth edition) (Cambridge University Press, Cambridge)
886:
887: \bibitem{} Kaiser, G. 2004,
888: J. Phys. A: Math. Gen. 37, 8735
889:
890: \bibitem{} Kellogg, O.D. 1953,
891: Foundations of potential theory (Dover, New York)
892:
893: \bibitem{} King, I. 1972,
894: ApJL, 174, L123
895:
896: \bibitem{} Kutuzov, S.A. 1998,
897: Astronomy Letters, 24, n.5, 645
898:
899: \bibitem{} Kutuzov, S.A., \& Osipkov, L.P. 1980,
900: Astron. Zh., 57, 28
901:
902: \bibitem{} Lanzoni, B., \& Ciotti, L. 2003,
903: A\&A, 404, 819
904:
905: \bibitem{} Lebovitz, N. R. 1967, ARAA, 5, 465
906:
907: \bibitem{} Lee, J., \& Suto, Y. 2003,
908: ApJ, 585, 151
909:
910: \bibitem{} Lee, J., \& Suto, Y. 2004,
911: ApJ, 601, 599
912:
913: \bibitem{} Letelier, P.S., \& Oliveira S.R. 1987,
914: J. Math. Phys., 28, 165
915:
916: \bibitem{} Letelier, P.S., \& Oliveira S.R. 1998,
917: Class. Quantum. Grav., 15, 421
918:
919: \bibitem{} Lynden-Bell, D. 1962,
920: MNRAS, 123, 447
921:
922: \bibitem{} Lynden-Bell, D. 2000,
923: MNRAS, 312, 301
924:
925: \bibitem{} Lynden-Bell, D. 2002,
926: preprint (astro-ph/0207064)
927:
928: \bibitem{} Lynden-Bell, D. 2004a,
929: Phys. Rev. D, 70, 104021
930:
931: \bibitem{} Lynden-Bell, D. 2004b,
932: Phys. Rev. D, 70, 105017
933:
934: \bibitem{} Miyamoto, M., \& Nagai, R. 1975,
935: PASJ, 27, 533
936:
937: \bibitem{} Muccione, V., \& Ciotti, L. 2003,
938: in Galaxies and Chaos, Lecture Notes on Physics,
939: ed. G. Contopoulos \& N. Voglis, 626, 387
940: (Springer-Verlag, New York)
941:
942: \bibitem{} Muccione, V., \& Ciotti, L. 2004,
943: A\&A, 421, 583
944:
945: \bibitem{} Newman, E.T. 1973
946: J. Math. Phys., 14, 102
947:
948: \bibitem{} Newman, E.T., \& Janis, A.I. 1965
949: J. Math. Phys., 6, 915
950:
951: \bibitem{} Newman, E.T., Couch, E.C., Chinnapared, K., Exton, A., Prakash, A.,
952: \& Torrence, R. 1965
953: J. Math. Phys., 6, 918
954:
955: \bibitem{} Page, D.N. 1976,
956: Phys. Rev. D, 14 1509
957:
958: \bibitem{} Pfenniger, D. 1984,
959: A\&A, 134, 373
960:
961: \bibitem{} Plummer, H.C. 1911,
962: MNRAS, 71, 460
963:
964: \bibitem{} Rosseland, S. 1926, ApJ, 63, 342
965:
966: \bibitem{} Satoh, C. 1980,
967: PASJ, 32, 41
968:
969: \bibitem{} Tassoul, J.L. 1978,
970: Theory of rotating stars (Princeton University Press, Princeton)
971:
972: \bibitem{} Teukolsky, S. 1973,
973: ApJ, 185, 635
974:
975: \bibitem{} Toomre, A. 1982,
976: ApJ, 259, 535
977:
978: \bibitem{} Tremaine, S., et al. 1994,
979: AJ, 107, 634
980:
981: \bibitem{} Waxman, A. M. 1978,
982: ApJ, 222, 61
983:
984: \bibitem{} Whittaker, E.T., \& Watson, G.N. 1950,
985: A course of modern analysis (Cambridge University Press, Cambridge)
986:
987: \end{thebibliography}
988: \end{document}
989: