1: %\documentclass[aasms4]{aastex}
2: \documentclass[12pt,preprint]{aastex}
3: %\documentclass{emulateapj}
4:
5: \shorttitle{The Counter-Streaming Instability in Dwarf Ellipticals with
6: Off-Center Nuclei}
7:
8: \shortauthors{De Rijcke \& Debattista}
9:
10: %\received{2003 December 5}
11: \begin{document}
12:
13: \title{The Counter-Streaming Instability in Dwarf Ellipticals with Off-Center
14: Nuclei\altaffilmark{1}}
15:
16: \author{Sven De Rijcke\altaffilmark{2}} \affil{Sterrenkundig
17: Observatorium, Universiteit Gent, Krijgslaan 281, S9, B-9000 Gent,
18: Belgium} \email{sven.derijcke@UGent.be} \author{Victor~P. Debattista}
19: \affil{Institut f\"ur Astronomie, ETH H\"onggerberg, HPF G4.2,
20: CH-8093 Z\"urich, Switzerland} \email{debattis@phys.ethz.ch}
21: \altaffiltext{1}{Based partly on observations collected at the European
22: Southern Observatory, Paranal, Chile (ESO Large Program
23: 165.N~0115)} \altaffiltext{2}{Research Postdoctoral Fellow of the
24: Fund for Scientific Research - Flanders (Belgium)(F.W.O.)}
25:
26: \begin{abstract}
27: In many nucleated dwarf elliptical galaxies (dE,Ns), the nucleus is
28: offset by a significant fraction of the scale radius with respect to
29: the center of the outer isophotes. Using a high-resolution $N$-body
30: simulation, we demonstrate that the nucleus can be driven off-center
31: by the $m=1$ counter-streaming instability, which is strong in
32: flattened stellar systems with zero rotation. The model develops a
33: nuclear offset of the order of one third of the exponential
34: scale-length. We compare our numerical results with photometry and
35: kinematics of FCC046, a Fornax cluster dE,N with a nucleus offset by
36: 1\farcs2; we find good agreement between the model and FCC046. We
37: also discuss mechanisms that may cause counter-rotation in dE,Ns and
38: conclude that the destruction of box orbits in an initially
39: triaxial galaxy is the most promising.
40: \end{abstract}
41: \keywords{galaxies: dwarf--- galaxies: evolution--- galaxies:
42: individual (FCC046)--- galaxies: kinematics and dynamics---
43: galaxies: structure--- galaxies: nuclei}
44:
45: \section{Introduction}
46: \label{sec:intro}
47:
48: Dwarf elliptical galaxies (dEs), faint, low-surface brightness
49: galaxies with smooth elliptical isophotes, are the most common type of
50: galaxy in clusters and groups of galaxies (see \citet{bf94}).
51: % for a review).
52: In hierarchical models of cosmological structure formation, dwarf
53: galaxies form first and subsequently merge to form larger
54: galaxies. Understanding their formation and evolution is therefore
55: vital to a complete picture of structure formation. There is evidence
56: that dEs constitute a heterogenous class of objects, kinematically
57: (rotating versus unrotating, \citet{ggv03}), chemically
58: \citep{mddzh03} and morphologically \citep{bbj02}. About half of the
59: dEs harbor a central bright nucleus and are called nucleated dEs
60: (dE,Ns). dE,Ns appear to be significantly older than normal
61: ellipticals and non-nucleated dEs \citep{rs03}; upon tidal stripping,
62: their nuclei have been suggested as sources of both the recently
63: discovered ultra compact dwarfs \citep{p01} and massive globular
64: clusters like $\omega$Cen \citep{g02}. In $\sim 20\%$ of Virgo dE,Ns,
65: the nucleus is significantly displaced with respect to the center of
66: the outer isophotes, typically by $\sim 1''$ ($\sim 100$~pc)
67: \citep{bbj00}. There is a tendency for the displacement to increase
68: with decreasing surface brightness but no relation between nuclear
69: displacement and any other structural or environmental parameter has
70: been found.
71:
72: Various models have been proposed to explain these offset nuclei. In
73: this Letter, we investigate whether the lopsided ($m=1$)
74: counter-streaming instability can reproduce a dE,N with an off-center
75: nucleus. This instability has been known since \citet{zh78}, and been
76: studied analytically \citep{s88,pp90} and with $N$-body simulations
77: \citep{ms90,lds90,sm94,sv97}. \citet{sv97} do not detect it in systems
78: rounder than E6, which is mainly due to their rounder models being
79: stabilized by a higher radial pressure. On the other hand,
80: \citet{ms90} find lopsidedness developing in systems as round as E1
81: but with negligible radial pressure. Partial rotation only introduces
82: a pattern speed in an otherwise purely growing instability;
83: simulations find that the lopsidedness produced by the instability is
84: long-lived.
85:
86: We use a realistic multi-component $N$-body model to generate a lopsided
87: system which we then compare with observations. In Section \ref{sec:theo}, we
88: present the $N$-body model and in Section \ref{sec:obs} we compare it with
89: photometry and kinematics of FCC046, an example of such a dE,N with an offset
90: nucleus. In Section \ref{sec:disc}, we discuss ways in which counter-rotation
91: may arise in dE,Ns.
92:
93: \section{$N$-body simulation}
94: \label{sec:theo}
95:
96: We have been using $N$-body simulations to explore the
97: counter-streaming instability as a mechanism for producing lopsided
98: dE,Ns.
99: %Our simulations include
100: We have performed both fully self-consistent simulations with live
101: nucleus, disk and halo run on a tree code, and restricted simulations
102: with rigid halos run on a grid code; both types of simulations will be
103: presented elsewhere. Here we present only one simulation which happens
104: to match FCC046 best, consisting of live disk and nucleus components
105: inside a rigid halo. The rigid halo was represented by a spherical
106: logarithmic potential with core-radius $r_{\rm c}$ and asymptotic circular
107: velocity $v_0$. The initial disk was modeled by an exponential disk
108: of mass $M_{\rm d}$, scale-length $R_{\rm d}$, Gaussian thickness
109: $z_{\rm d}$ and truncated at $R_{\rm t}$.
110: %
111: The nucleus of mass $M_{\rm n}$ was generated by iteratively
112: integrating a distribution function in the global potential
113: \citep{pt70}. We used a distribution function of the form
114: $f(\vec{x},\vec{v}) \propto \left\{ \left[-E(\vec{x},\vec{v})
115: \right]^{1/2} - \left[-E_{\rm max}\right]^{1/2} \right\}$. Here
116: $E_{\rm max} = \Phi_{\rm tot}(r_{\rm n})$, the total potential at the
117: nucleus truncation radius, $r_{\rm n}$, in the disk mid-plane.
118: Initial disk kinematics were set up using the epicyclic approximation
119: to give Toomre $Q = 1.3$. Vertical equilibrium was obtained by
120: integrating the vertical Jeans equation. The disk and nucleus were
121: represented by $36.8 \times 10^5$ and $3.2 \times 10^5$ equal-mass
122: particles, respectively. After setup, we switched the direction of the
123: velocity for half the particles, producing an unrotating disk and a
124: rotating nucleus. We used a quiet start \citep{s83} to cancel spurious
125: momenta and angular momenta.
126:
127: In our units, where $R_{\rm d} = M(=M_{\rm n}+M_{\rm d}) = G = 1$,
128: which give a unit of time $(R_{\rm d}^3/GM)^{1/2}$, we set $z_{\rm d}
129: = 0.2$, $R_{\rm t} = 5$, $r_{\rm n} = 0.78$, $r_{\rm c} = 5$ and $v_0
130: = 0.648$. The simulation was run on a 3-D cylindrical polar grid code
131: (described in \citet{sv97}) with $N_R\times N_\phi \times N_z = 60
132: \times 64 \times 225$. The vertical spacing of the grid planes was
133: $\delta z = 0.0125$. We used Fourier terms up to $m=8$ in the
134: potential, which was softened with the standard Plummer kernel, of
135: softening length $\epsilon = 0.0125$. Time integration was performed
136: with a leapfrog integrator using a fixed time-step $\delta t = 0.006$.
137: %
138: To scale units to FCC046, we note that its surface-brightness profile
139: declines exponentially with scale-length $R_{\rm d}=4\farcs1\approx
140: 0.4$~kpc. We estimated the stellar mass of FCC046 at $M \approx 1.0
141: \times 10^9 M_\odot$, given a $B$-band total luminosity $L_B = 2
142: \times 10^8 L_{B,\odot}$ and using a $B$-band mass-to-light ratio
143: $M/L_B \approx 5 M_\odot/ L_{B,\odot}$, typical for an old (15~Gyr),
144: rather metal-poor ([Fe/H]$=-0.5$) stellar population. With this
145: scale-length and mass, the unit of time corresponds to
146: $(R_{\rm d}^3/GM)^{1/2} \approx 4$~Myr and the velocity unit becomes
147: $(GM/R_{\rm d})^{1/2} \approx 100$~km/s.
148:
149: \clearpage
150:
151: \begin{figure}
152: \vspace{8cm}
153: \special{hscale=53 vscale=53 hsize=570 vsize=240
154: hoffset=-35 voffset=-165 angle=0 psfile="f1.eps"}
155: \caption{Panel {\bf a}:~a $B$-band image of FCC046; panel {\bf b}:~the
156: surface brightness of the $N$-body model at $t=150$. The isophotes are
157: spaced by 1~mag/arcsec$^2$ with the $\mu_B=24$~mag/arcsec$^2$ isophote
158: drawn in white. Panels {\bf c} and {\bf d}:~the model's velocity field
159: and the velocity dispersion field, respectively. The velocity field
160: perturbations have an $m=2$ symmetry and range between $-0.06$ (black)
161: and $+0.06$ (white), corresponding to $\pm 0.1 \times v_{\rm circ}$.
162: \label{modelps}}
163: \end{figure}
164:
165: \clearpage
166:
167: The global lopsidedness of the model, as measured by the amplitude of
168: the $m=1$ term in a Fourier expansion of the surface density, rose
169: exponentially with time, saturating at $t\simeq 70$, for an
170: $e$-folding time $\tau = 5.67$. %While the total angular momentum of
171: %the model is conserved, t
172: The angular momentum of the disk and the orbital angular momentum of
173: the nucleus increase at the expense of the nucleus's internal
174: (rotational) angular momentum. By the end of the simulation, the
175: nucleus has lost approximately 20~\% of its initial internal angular
176: momentum. The resulting orbital velocity of the nucleus ($v_{\rm nuc}
177: \approx 0.01$) is, however, only a few percent of the local circular
178: velocity ($v_{\rm circ} \approx 0.35$), so that dynamical friction
179: against a live dark matter halo is not likely to damp the instability,
180: as is also borne out by the live halo simulations.
181:
182: \section{FCC046:~a case study}
183: \label{sec:obs}
184:
185: FCC046 is an $m_B = 15.99$ dE4,N in the Fornax cluster. Photometry
186: and long-slit spectra, with an instrumental broadening $\sigma_{\rm
187: instr}=30$~km/s, in the wavelength region containing the strong
188: absorption lines of the near-infrared Ca{\sc ii} triplet, were
189: observed at the VLT with FORS2 in November 2001 (photometry) and 2002
190: (spectroscopy). Seeing conditions were typically 0\farcs8 FWHM. A
191: detailed analysis of the photometry has been presented in
192: \citet{ik03b}, while details of the kinematics of FCC046 will be
193: presented elsewhere. Additionally, the HST archives contain WFPC2
194: images of FCC046 in the filters F814W and F555W. Availability of all
195: these data was the sole reason for selecting this object.
196: % for our study.
197:
198: FCC046 contains a bright nucleus, which is displaced by 1\farcs2 with
199: respect to the outer isophotes, or $30\%$ of $R_{\rm d}$. The
200: lopsidedness involves not just the nucleus, but extends to $\sim 2
201: R_{\rm d}$. Color maps show no traces of large amounts of dust but
202: suggest the presence of a blue stellar population.
203: %, extending predominantly to the north of the nucleus.
204: The H$\alpha$ image reveals
205: ionized emission regions, evidence for ongoing star formation, within
206: $\sim 1 R_{\rm d}$. The position of the blue stellar population does
207: not coincide with the lopsidedness of FCC046 and is not likely to be
208: the cause of it (although it may be a consequence of it). FCC046 is
209: in the outskirts of the Fornax cluster; its nearest neighbors are
210: dwarfs over 100 kpc (in projection) away, and the nearest large galaxy
211: is $> 200$ kpc distant. Interactions are therefore not likely to
212: account for its morphology, unless a weakly damped mode has been
213: excited \citep{w94}.
214: %
215: The kinematics of FCC046 (mean velocity and velocity dispersion) were
216: extracted from the spectra following the method discussed in \citet{ik03a}.
217: Despite its E4 flattening, the main body of FCC046 has zero mean rotation.
218: The lack of rotation excludes the possibility that the eccentric instability
219: \citep{ars89,ms92,ti98} is responsible for the lopsidedness.
220:
221: The nucleus is resolved by HST; it has a FWHM$= 0\farcs27$,
222: significantly larger than the FWHM$= 0\farcs15$ of the stars in the
223: same image, and therefore cannot be a foreground star. The nucleus
224: has the same systemic velocity as the galaxy which rules out a chance
225: projection of a background object. It comprises $\sim 10~\%$ of the
226: $B$-band luminosity of FCC046 and is therefore probably quite massive.
227: We can use the fact that the outer isophotes of the $N$-body model do
228: not move appreciably and their center coincides with the model's
229: center-of-mass (COM) to estimate the relative mass-to-light ($M/L$)
230: ratios of the disk and nucleus. In order for the mean position of the
231: $B$-band light distribution to coincide with the COM, the $M/L$ ratio
232: of the nucleus has to be less than half that of the disk. If the disk
233: consists of an old, metal-poor population with $M/L_B \approx 5
234: M_\odot/L_{B,\odot}$, this would mean that the nucleus has $M/L_B < 2
235: M_\odot/L_{B,\odot}$, typical for a very young stellar population (age
236: $< 2-3$~Gyr). This is consistent with the blue color of the nucleus
237: and its strong Pa absorption \citep{mddzh03}. The rotation axis of
238: the nucleus is not aligned with that of the galaxy's main body,
239: showing evidence for minor axis rotation with an amplitude of $\approx
240: 5$~km/s.
241:
242: The surface density, the velocity field and the velocity dispersion
243: field of the $N$-body model at $t=150$ are presented in
244: Fig. \ref{modelps}. The viewing angle and the spatial scale were
245: chosen such that the model's surface density gives a fair
246: approximation to the $B$-band image of FCC046. The nucleus is offset
247: by $\sim 1''$ to the south-west with respect to the outer isophotes
248: and, in response, there is an excess in the disk density to the
249: north-east, as in FCC046. Clearly, the instability engenders a global
250: response of the whole stellar system. In Fig. \ref{compsurf}, we
251: compare the surface photometry of FCC046 with the $N$-body model
252: (convolved with a Gaussian to mimic the seeing conditions of the
253: observations) in greater detail. Simply shifting the magnitude
254: zero-point of the $N$-body model and applying the same spatial scale
255: as in Fig. \ref{modelps} sufficed to match the surface brightness
256: profile of FCC046. As in FCC046, the isophote centers of the model
257: oscillate along the major axis; in the model this keeps the total
258: center-of-mass fixed. Just outside the nucleus, the density contours
259: have a distorted shape which, in projection, results in the isophotes
260: being flatter (around $\sim 5''$ in FCC046 and around $\sim 3''$ in
261: the model) than at larger radii. On the whole, the model successfully
262: reproduces gross features in the surface photometry of FCC046.
263:
264: \clearpage
265:
266: \begin{figure}
267: \vspace{6cm}
268: \special{hscale=44 vscale=44 hsize=570 vsize=170
269: hoffset=-17 voffset=250 angle=-90 psfile="f2.eps"}
270: \caption{Surface photometry of FCC046 (left panels), compared with the
271: $N$-body model (right panels). $a$ and $b$ are respectively the
272: semi-major and minor axes of the isophotes. In the top panels, the
273: $R$-band surface brightness (in mag/arcsec$^2$) of FCC046 is
274: compared with that of the model. In the next panels, the
275: ellipticity $\epsilon = 10(1-b/a)$ of the isophotes is shown. In the
276: bottom panels, the position of the center of the isophotes is
277: presented. $\Delta {\rm maj}$ and $\Delta {\rm min}$ (in arcseconds)
278: quantify the position of the isophotes along the major and minor
279: axes, centered on the nucleus.
280: \label{compsurf}}
281: \end{figure}
282: \begin{figure}
283: \vspace{6cm}
284: \special{hscale=40 vscale=40 hsize=570 vsize=170
285: hoffset=-20 voffset=220 angle=-90 psfile="f3.eps"}
286: \caption{Part of the spectrum of FCC046, around the strongest Ca{\sc
287: ii} absorption line. The radial bins over which the spectrum has been
288: summed to boost the $S/N$ are indicated in the figure (left column,
289: top three panels:~along the major axis, towards east; right column,
290: top three panels:~along the major axis, towards west; bottom
291: panels:~along minor axis, towards north (left column) and towards
292: south (right column)). Radius is measured from the nucleus. Observed
293: spectrum:~black line; model spectrum:~grey line (see text for
294: details). This absorption line appears split into two distinct lines,
295: separated by $70 \pm 25$~km/s, in the outermost major axis
296: spectra. The $S/N$ in these panels varies from 5-10 to 20-25; adjacent
297: spectra from the left and right columns have comparable $S/N$. The
298: minor-axis spectra have the same continuum level as the outermost
299: major-axis spectra. Clearly, the Ca{\sc ii} line does not appear split
300: in the minor-axis spectra. The model reproduces rather well the
301: major-axis line width and the observed line-splitting.
302: \label{spec}}
303: \end{figure}
304:
305: \clearpage
306:
307: Fig. \ref{modelps} shows that, while the model's mean velocity is
308: zero, bi-symmetric perturbations of order $\sim 0.06$ are present. The
309: velocity dispersion profile is asymmetric, rising less rapidly towards
310: the left side of the model, in the direction of the light excess. The
311: apparent flattening of FCC046 and its zero mean rotation already hint
312: at two equal-mass counter-streaming stellar populations. Although at
313: the limit of what can be done, given the spectral resolution and the
314: low $S/N$-ratio of the data, the strongest Ca{\sc ii} absorption line
315: in the major-axis spectrum appears split into two distinct lines (see
316: Fig. \ref{spec}), separated by $70\pm 25$~km/s; the other, weaker,
317: lines have much lower $S/N$-ratios. This absorption line is unaffected
318: by the bright sky OH-emission lines. (If a badly removed sky line were
319: the cause of the line-splitting, it would also manifest itself in the
320: center of the galaxy and in the minor axis spectrum, which is not the
321: case.) What is particularly compelling about these data is that the
322: apparent bimodality is present (and absent) where the model predicts
323: it should be (before the signal disappears in the noise), and that we
324: see similar splitting in three different spatial bins. To illustrate
325: this, in Fig. \ref{spec} we plot the observed spectrum and the model
326: spectrum, consisting of a flat continuum and a very narrow
327: ($\delta$-function) absorption line, broadened with the FORS2
328: instrumental profile and the model's line-of-sight velocity
329: distribution (LOSVD), summed over the same spatial bins as the
330: observed spectrum. We did not use an observed stellar template since
331: the width of the individual LOSVD peaks is smaller than one FORS2
332: pixel, making an accurate calculation of a synthetic galaxy spectrum
333: impossible. Except for underestimating the minor-axis velocity
334: dispersion, the model reproduces rather well the major-axis line
335: widths and the observed line-splitting.
336:
337: However, these observations, while very suggestive, need to be
338: corroborated at higher spectral resolution to resolve the two peaks in
339: the LOSVDs. In Fig. \ref{losvd}, we present a mosaic of LOSVDs at
340: different points in the model. Moving out along the major axis, the
341: LOSVDs become more double-peaked due to the increasing velocity
342: separation between the two counter-streaming stellar populations and
343: the declining velocity dispersion. The small velocity-field
344: perturbations of Fig. \ref{modelps} are reflected here in the LOSVDs
345: being slightly asymmetric. Although the details of the LOSVD field in
346: Fig. \ref{losvd} are model-dependent, an observational confirmation of
347: its salient features would lend considerable support to the hypothesis
348: that the counter-streaming instability is responsible for off-center
349: nuclei in dE,Ns.
350:
351: \clearpage
352:
353: \begin{figure}
354: \vspace{6.25cm}
355: \special{hscale=40 vscale=40 hsize=570 vsize=180
356: hoffset=0 voffset=225 angle=-90 psfile="f4.eps"}
357: \caption{
358: A mosaic of the LOSVDs at different points in the projected model (the
359: middle panel is centered on the center of the outer isophotes). Big and
360: small tickmarks indicate a velocity $v = \pm 1$ and $v = \pm 0.5$,
361: respectively, as indicated below the leftmost panel. The isophotes of the
362: model, separated by 1~mag/arcsec$^2$, are plotted in grey-scale.
363: \label{losvd}}
364: \end{figure}
365:
366: \clearpage
367:
368: \section{Discussion}
369: \label{sec:disc}
370:
371: If the counter-streaming instability is indeed working in a sizable
372: fraction of the dE,Ns to drive nuclei off-center, it is natural to ask
373: why these dEs have counter-rotation in the first place. The first
374: detections of counter-rotation in galaxy disks \citep{4550,7217} lead
375: to the hypothesis that the capture of gas with retrograde rotation and
376: subsequent star-formation produced two counter-rotating stellar disks.
377: This hypothesis predicts stellar populations of different ages and
378: metallicities. Generally, the two disks will not have identical
379: masses, so some net rotation is expected; this is, therefore, an
380: unlikely explanation for FCC046.
381:
382: \citet{tr00} suggested that counter-rotation is produced when a
383: triaxial halo with an initially retrograde pattern speed slowly
384: changes to a prograde pattern speed. However, this situation probably
385: does not occur very often. Moreover, this mechanism requires evolution
386: on long timescales $\sim 10^{10}$~yr and hence requires dE,Ns to be
387: old and unperturbed.
388:
389: A more likely mechanism is the destruction of box-orbits, which are
390: the backbone of any slowly rotating triaxial mass-distribution, by a
391: growing nucleus. Stars on box orbits come very close to the nucleus
392: and thus can be scattered into loop-orbits; conservation of angular
393: momentum then requires that direct and retrograde loop orbits be
394: equally populated \citep{ec94,mq98}. If the nucleus has a mass larger
395: than $2\%$ of the total mass of the galaxy, evolution towards an
396: axisymmetric shape requires a few crossing times ($\approx 10^7 -
397: 10^8$~yr), while our simulations show that the counter-streaming
398: instability develops on similar timescales. Late infall of gas
399: \citep{co03} (perhaps as a result of harassment \citep{mo96,ma01}) or
400: globular clusters \citep{lo01} may be responsible for growing the
401: nucleus. If this explanation holds, then the two counter-rotating
402: stellar populations should have equal scale-lengths and chemical
403: properties and nearly equal mass. Additionally, the nucleus may
404: retain some of its initial angular momentum; we note that the nucleus
405: of FCC046 is indeed rotating.
406:
407: To summarize:~using $N$-body simulations, we have shown that the
408: counter-streaming instability is a viable explanation for at least the
409: dE,N galaxy FCC046. The models can produce offsets as large as
410: observed and successfully reproduce distinctive features in the
411: surface photometry of FCC046, such as the oscillation of isophote
412: centers. Although at the limit of what can reliably be extracted from
413: the spectra, we find a tantalizing hint of counter-streaming in the
414: major-axis spectra of this galaxy, which, however, needs to be
415: corroborated at higher spectral resolution.
416:
417: \acknowledgments V.P.D. thanks the Sterrenkundig Observatorium for
418: hospitality during an early stage of this project. We thank Stelios
419: Kazantzidis and Anna Pasquali for useful discussions and the referee,
420: Dr. O. Gnedin, for the suggested improvements.
421:
422: \begin{thebibliography}{}
423:
424: \bibitem[Adams {\em et al.}(1989)]{ars89} Adams, F.~C., Ruden, S.~P., Shu,
425: F.~H. 1989, \apj, 347, 959
426:
427: \bibitem[Barazza {\em et al.}(2002)]{bbj02} Barazza, F.~D., Binggeli, B., \&
428: Jerjen, H. 2002, \aap, 391, 823
429:
430: \bibitem[Binggeli \& Ferguson(1994)]{bf94} Binggeli, B., \& Ferguson, H.~C.
431: 1994, \aapr, 6, 67
432:
433: \bibitem[Binggeli {\em et al.}(2000)]{bbj00} Binggeli, B., Barazza, F.,
434: Jerjen, H. 2000, \aap, 359, 447
435:
436: \bibitem[Conselice {\em et al.}(2003)]{co03} Conselice, C.~J., O'Neil, K.,
437: Gallagher, J.~S., Wyse, R.~G. 2003, \apj, 591, 167
438:
439: \bibitem[De Rijcke {\em et al.}(2003a)]{ik03a} De Rijcke, S., Dejonghe, H.,
440: Zeilinger, W.~W., \& Hau, G.~K.~T. 2003, \aap, 400, 119
441:
442: \bibitem[De Rijcke {\em et al.}(2003b)]{ik03b} De Rijcke, S., Zeilinger,
443: W.~W., Dejonghe, H., \& Hau, G.~K.~T. 2003, \mnras, 339, 225
444:
445: \bibitem[Evans \& Collett(1994)]{ec94} Evans, N.~W., \& Collett, J.~L.
446: 1994, \apj, 420, L67
447:
448: \bibitem[Geha {\em et al.}(2003)]{ggv03} Geha, M., Guhathakurta, P., \& van
449: der Marel, R.~P. 2003, \aj, 126, 1794
450:
451: \bibitem[Gnedin {\em et al.}(2002)]{g02} Gnedin, O.~Y. , Zhao, H.-S.,
452: Pringle, J. E., Fall, S.~M., Livio, M., \& Meylan, G. 2002, \apjl,
453: 568, L23
454:
455: \bibitem[Levison {\em et al.}(1990)]{lds90} Levison, H.~F., Duncan, M.~J., \&
456: Smith, B.~F. 1990, \apj, 363, 66
457:
458: \bibitem[Lotz {\em et al.}(2001)]{lo01} Lotz, J.~M., Telford, R., Ferguson,
459: H.~C., Miller, B.~W., Stiavelli, M., \& Mack J. 2001, \apj, 552, 572
460:
461: \bibitem[Mayer {\em et al.}(2001)]{ma01} Mayer, L., Governato, F., Colpi, M.,
462: Moore, B., Quinn, T., Wadsley, J., Stadel, J., \& Lake, G. 2001, \apj, 559,
463: 754
464:
465: \bibitem[Merrifield \& Kuijken(1994)]{7217} Merrifield, M.~R., \& Kuijken, K.
466: 1994, \apj, 432, 575
467:
468: \bibitem[Merritt \& Quinlan(1998)]{mq98} Merritt, D., \& Quinlan, G.~D. 1998,
469: \apj, 498, 625
470:
471: \bibitem[Merritt \& Stiavelli(1990)]{ms90} Merritt, D., \& Stiavelli, M. 1990,
472: \apj, 358, 399
473:
474:
475: \bibitem[Michielsen {\em et al.}(2003)]{mddzh03} Michielsen, D., De Rijcke,
476: S., Dejonghe, H., Zeilinger, W. ~W., \& Hau, G.~K.~T. 2003, \apj, 597, L21
477:
478: \bibitem[Miller \& Smith(1992)]{ms92} Miller, R.~H., \& Smith, B.~F. 1992,
479: \apj, 393, 508
480:
481: \bibitem[Moore {\em et al.}(1996)]{mo96} Moore, B., Katz, N., Lake, G.,
482: Dressler, A., \& Oemler Jr., A. 1996, Nature, 379, 613
483:
484: \bibitem[Rubin {\em et al.}(1992)]{4550} Rubin, V.~C., Graham, J.~A., \&
485: Kenney, J.~D.~P 1992, \apj, 394, L9
486:
487: \bibitem[Palmer \& Papaloizou(1990)]{pp90} Palmer, P.~L., \& Papaloizou, J.
488: 1990, \mnras, 243, 263
489:
490: \bibitem[Phillipps {\em et al.}(2001)]{p01} Phillipps, S., Drinkwater,
491: M., Gregg, M., \& Jones, J. 2001, \apj, 560, 201
492:
493: \bibitem[Prendergast \& Tomer(1970)]{pt70} Prendergast, K.~H., \&
494: Tomer, E., 1970, \aj, 75, 67
495:
496: \bibitem[Rakos \& Schombert(2003)]{rs03} Rakos, K. \& Schombert, J. 2003, accepted for publication by \aj, astro-ph/0312075
497:
498: \bibitem[Sawamura(1988)]{s88} Sawamura, M. 1988, \pasj, 40, 279
499:
500: \bibitem[Sellwood \& Valluri(1997)]{sv97} Sellwood, J.~A., \& Valluri, M.
501: 1997, \mnras, 287, 124
502:
503: \bibitem[Sellwood \& Merritt(1994)]{sm94} Sellwood, J.~A., \& Merritt, D. 1994,
504: \apj, 425, 530
505:
506: \bibitem[Sellwood(1983)]{s83} Sellwood, J.~A. 1983, J. Comput. Phys., 50, 337
507:
508: \bibitem[Taga \& Iye(1998)]{ti98} Taga, M., \& Iye, M. 1998, \mnras, 299, 111
509:
510: \bibitem[Tremaine \& Yu(2000)]{tr00} Tremaine, S., \& Yu, Q., 2000, \mnras,
511: 319, 1
512:
513: \bibitem[Weinberg(1994)]{w94} Weinberg, M.~D. 1994, \apj, 421, 481
514:
515: \bibitem[Zang \& Hohl(1978)]{zh78} Zang, T.~A., \& Hohl, F. 1978, \apj, 226,
516: 521
517:
518: \end{thebibliography}
519: \end{document}
520: