1: \documentclass[12pt,preprint]{aastex}
2: %\documentclass{emulateapj}
3:
4: %%\usepackage{amsmath}
5: %\usepackage[]{natbib}
6:
7: \newcommand{\kms}{\mbox{$\>{\rm km\, s^{-1}}$}}
8: \newcommand{\kmskpc}{\>{\rm km}\,{\rm s}^{-1}\,{\rm kpc}^{-1}}
9: \newcommand{\pc}{\>{\rm pc}}
10: \newcommand{\kpc}{\mbox{$\>{\rm kpc}$}}
11: \newcommand{\gyr}{\mbox{$\>{\rm Gyr}$}}
12: \newcommand{\msun}{\>{\rm M_{\odot}}}
13: \newcommand{\lsun}{\>{\rm L_{\odot}}}
14: \newcommand\degrees{^\circ}
15: \newcommand{\epot}{\mbox{$\epsilon_\Phi$}}
16: \newcommand{\mdisk}{\mbox{$M_{disk}$}}
17:
18: \def\etal{{et al.}}
19: \def\eg{{\it e.g.}}
20: \def\etc{{\it etc.}}
21: \def\ie{{\it i.e.}}
22: \def\cf{{\it cf.}}
23:
24: %\voffset -4truecm
25: \slugcomment{Draft to ApJ}
26: \lefthead{Debattista \etal}
27:
28: \righthead{The Causes of Halo Shape Changes Induced by Cooling
29: Baryons}
30:
31: \begin{document}
32:
33: \title{The Causes of Halo Shape Changes Induced by Cooling Baryons:\\
34: Disks Versus Substructures}
35:
36: \author{Victor P. Debattista\altaffilmark{1,2}, Ben
37: Moore\altaffilmark{3}, Thomas Quinn\altaffilmark{1}, Stelios
38: Kazantzidis\altaffilmark{4}, Ryan Maas\altaffilmark{1}, Lucio
39: Mayer\altaffilmark{3}, Justin Read\altaffilmark{3}, Joachim
40: Stadel\altaffilmark{3}}
41:
42: \altaffiltext{1}{Astronomy Department, University of Washington, Box
43: 351580, Seattle, WA 98195, {debattis;trq;maasr@astro.washington.edu.}}
44:
45: \altaffiltext{2}{Brooks Prize Fellow}
46:
47: \altaffiltext{3}{Department of Theoretical Physics, University of
48: Z\"urich, Winterthurerstrasse 190, CH-8057, Z\"urich, Switzerland {
49: moore;lucio;justin;stadel@physik.unizh.ch.}}
50:
51: \altaffiltext{4}{Kavli Institute for Particle Astrophysics and
52: Cosmology, Department of Physics, Stanford University, P.O. Box 20450,
53: MS 29, Stanford, CA 94309; stelios@slac.stanford.edu.}
54:
55: %%\date{{\it Draft version on \today}}
56:
57: \begin{abstract}
58: Cold dark matter cosmogony predicts triaxial dark matter halos,
59: whereas observations find quite round halos. This is most likely due
60: to the condensation of baryons leading to rounder halos. We examine
61: the halo phase space distribution basis for such shape changes.
62: Triaxial halos are supported by box orbits, which pass arbitrarily
63: close to the density center. The decrease in triaxiality caused by
64: baryons is thought to be due to the scattering of these orbits. We
65: test this hypothesis with simulations of disks grown inside triaxial
66: halos. After the disks are grown we check whether the phase space
67: structure has changed by evaporating the disks and comparing the
68: initial and final states. While the halos are substantially rounder
69: when the disk is at full mass, their final shape after the disk is
70: evaporated is not much different from the initial. Likewise, the halo
71: becomes (more) radially anisotropic when the disk is grown, but the
72: final anisotropy is consistent with the initial. Only if the baryons
73: are unreasonably compact or massive does the halo change irreversibly.
74: We show that the character of individual orbits is not generally
75: changed by the growing mass. Thus the central condensation of baryons
76: does not destroy enough box orbits to cause the shape change. Rather,
77: box orbits merely become rounder along with the global potential.
78: However, if angular momentum is transferred to the halo, either via
79: satellites or via bars, a large irreversible change in the halo
80: distribution occurs. The ability of satellites to alter the phase
81: space distribution of the halo is of particular concern to galaxy
82: formation simulations since halo triaxiality can profoundly influence
83: the evolution of disks.
84: \end{abstract}
85:
86: \keywords{galaxies: evolution --- galaxies: formation --- galaxies:
87: halos --- dark matter}
88:
89: %
90: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
91: %
92:
93: \section{Introduction}
94: \label{sec:intro}
95:
96: The dark matter halos that form via hierarchical growth in the cold
97: dark matter (CDM) cosmologies are generally triaxial with mean axial
98: ratios $b/a \sim 0.6$ and $c/a \sim 0.4$, where $c < b < a$ are the
99: short, intermediate, and long axes, respectively \citep{bbks_86,
100: bar_efs_87, frenk_etal_88, dub_car_91, jin_sut_02, bai_ste_05,
101: all_etal_06}.
102: %
103: Observational constraints on halo shapes can be obtained from the
104: Milky Way \citep{iba_etal_01, joh_etal_05, helmi_04, fel_etal_06},
105: from polar ring galaxies \citep{sch_etal_83, sac_spa_90, iod_etal_03},
106: from X-ray isophotal shapes \citep{buo_can_94, buo_etal_02} \citep[but
107: see also ][]{die_sta_07}, and from gravitational lensing
108: \citep{kochan_95, bar_etal_95, koo_etal_98, ogu_etal_03}. For disk
109: galaxies, or disks surrounding elliptical galaxies, the ellipticity of
110: the potential in the mid-plane, \epot, can be constrained through
111: photometry and/or kinematics of stars or gas
112: \citep[e.g.,][]{fra_dez_92, hui_van_92, kui_tre_94, fra_etal_94,
113: sch_etal_97, and_etal_01, debatt_03, bar_sel_03, wei_etal_08}.
114: %
115: The general consensus from these studies is that dark matter halos are
116: rounder than those predicted by collisionless CDM simulations. But this
117: need not be in disagreement with CDM since the condensation of baryons
118: to the centers of halos has been shown to lead to rounder halos
119: \citep{dubins_94, kkzanm04}. For example, \citet{kkzanm04} find that
120: the principal axis ratios increase by $\sim 0.2-0.4$ in the inner
121: regions (although triaxiality is not completely erased) extending to
122: almost the virial radius.
123:
124: Slowly rotating triaxial structures can be supported by centrophilic
125: box orbits \citep{schwar_79, ger_bin_85, statle_87, udr_mar_94,
126: fri_mer_97, val_mer_98}. Several studies have shown that when a black
127: hole is present scattering of box orbits is responsible for causing an
128: elliptical galaxy to become rounder, or at least axisymmetric
129: \citep{lak_nor_83, ger_bin_85, nor_etal_85, mer_qui_98, val_mer_98,
130: hol_etal_02, kal_etal_04}. These scattering events lead to a large
131: number of orbits becoming chaotic. Chaos by itself, however, need not
132: be a fundamental limit to forming long-lived triaxial structures:
133: using orbit superposition, \citet{poo_mer_02} were able to construct
134: long-lived triaxial models of nuclei even in the presence of a large
135: fraction $(\ga 50\%)$ of chaotic orbits.
136: %
137: If axisymmetrization does occur, \citet{ger_bin_85} predict that it is
138: largely confined to the center and occurs gradually. The $N$-body
139: simulations of a cored system by \citet{mer_qui_98} instead found that
140: the axisymmetrization extends to the entire system and occurs on a
141: crossing time for black holes of mass $\sim 2\%$ of the galaxy's mass.
142: When instead the system is cuspy, \citet{hol_etal_02} found that black
143: holes do not lead to a global axisymmetrization of the system.
144: %
145: Triaxial structures in disks (\ie\ bars) can also be destroyed by
146: central mass concentrations (CMCs). The main mechanism is again
147: scattering by the CMC. Although the main bar-supporting orbit family,
148: the $x_1$ orbits \citep{contop_80}, is a centrophobic loop family,
149: stars librating about the closed $x_1$ orbits can still get close to
150: the center and then be scattered by a CMC. Simulations have shown
151: that the required mass for a soft CMC (\ie\ one with a scale of a few
152: 100 pc) is an unrealistically large $\sim 20\%$ of the disk mass,
153: while the mass required of a hard CMC (few parsecs or less scale) is
154: $\sim 5\%$ of the disk mass \citep{she_sel_04, ath_etal_05,
155: deb_etal_06}, which is much larger than typical supermassive black
156: holes.
157:
158: Likewise, it has often been assumed that the loss of halo triaxiality
159: when baryons cool inside halos is partly or mostly due to the
160: destruction of box orbits, which pass arbitrarily close to the center
161: after a sufficiently long time. The fate of box orbits in the
162: presence of disks is of interest for various reasons beside the shape
163: of the halo. Box orbits play an important role in speeding up the
164: mergers of supermassive black holes at the centers of galaxies
165: \citep{mer_poo_04}. Moreover, box
166: orbits lead to radial anisotropy, whereas the destruction of box orbits
167: results in tangential anisotropy. This in turn affects the
168: event rate and energies of dark matter detection experiments involving
169: both direct scattering and indirect annihilation from capture by the
170: Sun or the Earth \citep[see the review by][]{jun_etal_96}.
171:
172: In order to help shed light on these issues, we test whether box orbit
173: scattering is responsible for triaxial halos becoming rounder when
174: baryons cool inside them. We do this via simulations in which we first
175: grow and then evaporate disks inside triaxial halos. Such evaporation, while
176: obviously unphysical, allows us to directly assess the impact of disks
177: on halos by comparing the initial and final states.
178: %
179: After the disks are grown, we find that the halos become substantially
180: rounder and their kinematics radially anisotropic. But comparing their
181: initial and final shapes when the disk mass is zero in both cases, we
182: find that the changes are largely reversible. The destruction of box
183: orbits being irreversible, halos should not recover their initial
184: states if this is the main cause of the shape change. We also show
185: that if angular momentum is transferred to the halo (via bars or
186: satellites), then the irreversible changes are substantially larger.
187:
188: Section \ref{sec:numerics} of this paper discusses the $N$-body
189: methods used in this study. Section \ref{sec:central} presents the
190: results of simulations with growing rigid central massive objects. In
191: \S \ref{sec:angmomtran} we present simulations in which angular momentum
192: is transferred to the halo either by a live bar or by satellites.
193: Section \ref{sec:orbitanalysis} presents a preliminary analysis of the
194: orbital evolution for a subsample of the simulations. Our conclusions
195: are presented in \S \ref{sec:conc}.
196:
197: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
198:
199: \section{Numerical Methods}
200: \label{sec:numerics}
201:
202: The basis of this work is that box orbit destruction is an
203: irreversible process. Rather than following all orbits, we
204: adiabatically grow and then evaporate a disk to show that the
205: distribution function of a halo is not substantially changed despite
206: the fact that the halo appears very different when the disk is at full
207: mass. Of course, evaporating the disk is a purely numerical
208: contrivance, but this allows us to test for halo distribution function
209: changes directly.
210: %
211: Although classical mechanics are time-reversible, the random phases of
212: any scattered orbits ensure that simply evaporating the central mass
213: is not enough to return to the initial configuration. This would only
214: be possible if we had a perfect integrator and if we had reversed all
215: velocities, which we did not do.
216:
217: We formed prolate/triaxial halos via mergers, as described in
218: \citet{moo_etal_04}. The initial spherical halos were generated from
219: a distribution function using the method described in \citet{kmm04}
220: with the added refinement that each halo is composed of two mass
221: species arranged on shells. The outer shell has more massive
222: particles than the inner one, in order to increase the effective
223: resolution in the central parts.
224: %
225: Our model halo A was generated by the head-on merger of two prolate
226: halos, themselves the product of a binary merger of spherical systems.
227: The first merger placed the concentration $c=10$ halos 800 kpc apart
228: approaching each other at 50 \kms, while the second merger starts with
229: the remnant at rest, 400 kpc from an identical copy. The resulting
230: halo is highly prolate with a mild triaxiality. Halo model B was
231: produced by the merger of two spherical halos starting at rest, 800 kpc
232: apart. Both halos A and B consist of $4\times 10^6$ particles. The
233: outer particles are $\sim 18$ times more massive in halo A and $\sim
234: 5$ times more massive in halo B. A large part of the segregation by
235: particle mass persists after the mergers and the small radius regions
236: are dominated by low mass particles \citep[cf.][]{dehnen_05}.
237: Figure \ref{fig:segregation} shows the particle segregation in the
238: case of halo A. We used a softening parameter $\epsilon = 0.1\kpc$
239: for all halo particles, although we have verified that using a larger
240: softening, $\epsilon = 1 \kpc$, for the more massive species does not
241: change our results. Our force resolution was chosen to be smaller
242: than the vertical scale of the disk, thereby resolving short-range
243: forces.
244:
245: \begin{figure}
246: \centerline{
247: \includegraphics[angle=-90.,width=\hsize]{f1.ps}
248: }
249: \caption{Spherically averaged density profile of dark matter
250: particles in halo A before any baryons are introduced. The solid line
251: is the full density profile, the dashed line is for the lower mass
252: species, while the dotted line is for the higher mass species.
253: \label{fig:segregation}}
254: \end{figure}
255:
256: Once we produced the prolate/triaxial halos, we inserted a disk of
257: particles that remains rigid throughout the experiments. In all
258: mergers we have been careful to either give the halo no angular
259: momentum, or to place the disk's symmetry axis along the angular
260: momentum of the halo since otherwise additional evolution would result
261: \citep{deb_sel_99}. The disks are composed of $300K$ equal-mass
262: particles each with a softening $\epsilon = 60-100\pc$. The disk
263: distribution was, in all cases, exponential with scale length $R_{\rm
264: d}$ and Gaussian scale-height $z_{\rm d}/R_{\rm d} = 0.05$. The disks
265: were placed at various orientations within the halos. We refer to
266: these experiments by the halo axis along which the disk's symmetry
267: axis is aligned: in ``short-axis'' (S) experiments, the symmetry axis
268: of the disk is parallel to the short axis of the halo, while in
269: ``long-axis'' (L) experiments, the symmetry axis of the disk is along
270: the halo's major axis. If the halo is triaxial, then an
271: ``intermediate-axis'' (I) experiment has the disk minor axis parallel
272: to the halo's intermediate axis. Initially, the disk has negligible
273: mass, but this grows adiabatically linearly over time to a mass $M_b$
274: during a time $t_g$. After this time, we slowly evaporated it during
275: a time $t_e$. Thus,
276: %
277: \begin{equation}
278: \mdisk (t) = \left\{
279: \begin{array}{ll}
280: {M_b~ \frac{t}{t_g}} & {0 \leq t \leq t_g} \cr \cr
281: {M_b~ (1 - \frac{t-t_g}{t_e})} & {t_g \leq t \leq t_g + t_e}.
282: \end{array}\right.
283: \end{equation}
284: %
285: From $t=0$ to $t_g+t_e$ the halo particles are free to move and
286: achieve equilibrium with the disk as its mass changes, but all disk
287: particles are frozen in place. Since a triaxial global potential
288: leads to elliptical disks forming, we include one simulation with an
289: elliptical disk.
290:
291: Another key assumption in these simulations is that the disks form
292: without much transfer of angular momentum to the halo. While
293: formation of realistic galaxies requires that baryons conserve most of
294: their angular momentum \citep[e.g.,][]{som-lar_etal_99}, gas
295: condensation onto subhalos results in angular momentum transfer to the
296: halo \citep{nav_ste_97}.
297: %
298: We therefore present experiments in which a few softened
299: particles were introduced, with a mass grown in the same way. We
300: refer to these experiments by the label ``P'' subscripted by ``f'' for
301: particles frozen in place and by ``l'' for live particles free to
302: move. Lastly, we present one simulation, BA1, in which the disk at
303: $t_g$ is replaced by live particles and evolved for 10 Gyr before
304: evaporating the disk. We set initial disk particle velocities for a
305: constant Toomre-$Q = 1.5$. In setting up the disk kinematics we
306: azimuthally averaged radial and vertical forces. Thus, our disk is
307: initially not in perfect equilibrium, but was close enough that it
308: quickly settled to a new equilibrium.
309:
310: \begin{table}[!ht]
311: \begin{centering}
312: \begin{tabular}{cccccccc}\hline
313: \multicolumn{1}{c}{Run} &
314: \multicolumn{1}{c}{Halo} &
315: \multicolumn{1}{c}{$r_{200}$} &
316: \multicolumn{1}{c}{$M_{200}$} &
317: \multicolumn{1}{c}{$M_b$} &
318: \multicolumn{1}{c}{$R_{\rm d}$} &
319: \multicolumn{1}{c}{$t_g$} &
320: \multicolumn{1}{c}{$t_e$} \\
321:
322: & & [kpc] & [$10^{12} \msun$] & [$10^{11} \msun$] & [kpc] & [Gyr] & [Gyr] \\ \hline
323:
324: % 563
325: SA1 & A & 215 & 4.5 & 1.75 & 3.0 & 5 & 2.5 \\
326: % 587
327: SA2 & A & 215 & 4.5 & 5.25 & 3.0 & 5 & 2.5 \\
328: % 585
329: SA3 & A & 215 & 4.5 & 1.75 & 1.5 & 5 & 2.5 \\
330: % 534
331: IA1 & A & 215 & 4.5 & 1.75 & 3.0 & 5 & 2.5 \\
332: % 537
333: LA1 & A & 215 & 4.5 & 1.75 & 3.0 & 5 & 2.5 \\
334: % 419
335: LB1 & B & 106 & 0.65 & 1.05 & 3.0 & 15 & 7.0 \\
336: % 592
337: TA1 & A & 215 & 4.5 & 1.75 & 3.0 & 5 & 2.5 \\
338: % 593
339: EA1 & A & 215 & 4.5 & 1.75 & 3.0 & 5 & 2.5 \\
340: % 563 Replace1.EvapA
341: BA1 & A & 215 & 4.5 & 0.52 & 3.0 & 1.5 & 2.5 \\
342:
343: % 583
344: P$_l$A1 & A & 215 & 4.5 & 1.75 & 0.5 & 5 & 2.5 \\
345: % 590
346: P$_l$A2 & A & 215 & 4.5 & 1.75 & 5.0 & 5 & 2.5 \\
347:
348: % 448GR2
349: P$_l$B1 & B & 106 & 0.65 & 0.7 & 3.0 & 10 & 4 \\
350: % 448bGR2
351: P$_f$B2 & B & 106 & 0.65 & 0.7 & 3.0 & 10 & 4 \\
352: % 448SG
353: P$_l$B3 & B & 106 & 0.65 & 0.35 & 0.1 & 5 & 5 \\
354:
355: \hline
356: \end{tabular}
357: \caption{The simulations in this paper. For the particle simulations
358: (P$_l$A1-P$_l$B3), $R_{\rm d}$ refers to the softening of the
359: particle(s). For runs P$_l$A1 and P$_l$A2, the value of $M_b$ refers
360: to the combined mass of all the satellite particles at $t_g$. The
361: disk in run TA1 is tilted by $30\degrees$ relative to the one in SA1
362: whereas the disk in EA1 is elliptical. In run BA1, the disk at $t_g$
363: was replaced by live particles and evolved for 10 Gyr (during which time a
364: bar formed and then was destroyed), before being evaporated.}
365: \label{tab:simulations}
366: \end{centering}
367: \end{table}
368:
369: \begin{figure}
370: \centerline{
371: \includegraphics[angle=-90.,width=0.5\hsize]{f2a.ps}
372: \includegraphics[angle=-90.,width=0.5\hsize]{f2b.ps}
373: }
374: \caption{Azimuthally averaged rotation curves of models SA1 ({\it left})
375: and P$_l$B1/P$_f$B2 ({\it right}) measured in the midplane. In both panels
376: the solid line is the full rotation curve, the dashed line the
377: contribution of the baryons, and the dot-dashed line the contribution
378: from the halo.
379: \label{fig:rotcurves}}
380: \end{figure}
381:
382: All the simulations in this paper, which are listed in Table
383: \ref{tab:simulations}, were evolved with {\sc PKDGRAV}
384: \citep{stadel_phd}, an efficient, multi-stepping, parallel treecode.
385:
386:
387: \subsection{Measuring halo shapes}
388:
389: To measure the axis ratios $c/a$ and $b/a$ we adopt a method based on
390: \citet{katz_91} that uses the eigenvalues of the (unweighted) moment
391: of inertia tensor $I$. For each bin of $N$ particles we computed
392: $I_{ij}$ as follows:
393: %
394: \begin{equation}
395: I_{ij} = \frac
396: {\sum_{k=1}^N m_{k} r_{i,k} r_{j,k}}
397: {\sum_{k=1}^N m_k}.
398: \end{equation}
399: %
400: We then diagonalize $I$ and calculate
401: %
402: \begin{equation}
403: b/a = \sqrt{{\cal I}_{22}/{\cal I}_{11}} \hspace{0.5cm} \mbox{and}\hspace{0.5cm} c/a
404: = \sqrt{{\cal I}_{33}/{\cal I}_{11}},
405: \end{equation}
406: %
407: where the ${\cal I}_{ii}$'s are the eigenvalues of $I$ and ${\cal
408: I}_{11} \geq {\cal I}_{22} \geq {\cal I}_{33}$. A useful parameter
409: for expressing shape is the triaxiality parameter $T = (a^2 -
410: b^2)/(a^2 - c^2)$ \citep{fra_etal_91}. The cases $T=0$ and $T=1$
411: correspond to oblate and prolate shapes, respectively, while $T=0.5$ is
412: the maximally triaxial case.
413:
414: We measured shapes in shells of fixed semi-major axis widths around
415: the center of the system. Thus, these shape measurements are
416: differential, rather than integrated \citep[\cf][]{katz_91}. We use
417: the iterative procedure of \citet{katz_91} in which the convergence
418: criterion is a variation in axis ratios by $<0.01\%$. In each
419: iteration the semi major axis of the shell is held fixed; a particle
420: is included in the calculation of $I_{ij}$ if $q_{lo} < q < q_{hi}$,
421: where $q$ is the ellipsoidal radius defined as
422: %
423: \begin{equation}
424: q ^{2} = x^{2} + \left ( \frac{y}{b/a} \right )^2 + \left (
425: \frac{z}{c/a} \right )^2.
426: \end{equation}
427: %
428: We used shell widths $q_{hi}-q_{lo} = 5 \kpc$ for all
429: models.
430:
431: The center of the system is taken to be the center of mass of a sphere
432: of radius 1 \kpc\ centered on the minimum of the potential and is
433: fixed for all shells. Tests performed in which the center of mass was
434: allowed to vary by up to 0.5 \kpc\ show less than 5\% variation in the
435: axis ratios past 10 \kpc. Tests in which the limits of each shell
436: were reduced by half around the average radius of that shell gave axis
437: ratios that varied from the full resolution results by up to $0.08$
438: in the worst cases and by less than $0.05$ for most runs. Shells were
439: not prevented from overlapping; as a result, some particles are sampled
440: in more than one shell. We have verified that this does not bias our
441: shape estimates through the experiments with the halved shell widths,
442: where the shells never intersect. When the number of particles in the
443: central shell is less than $10K$, then convergence is not reached after
444: 20 iterations, or the axis ratios varies by as much as 20\%, so we
445: take this number to be a reasonable cutoff for the reliability of
446: these innermost shells and ignore shells with less particles. This
447: occurs in only two cases, and in general most inner shells have $>25K$
448: particles, which we find to be more than enough to ensure consistent
449: measurements with our method.
450:
451: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
452:
453: \section{Central Massive Objects}
454: \label{sec:central}
455: \thispagestyle{empty}
456: \setlength{\voffset}{-22mm}
457: \begin{figure*}
458: \centerline{
459: \includegraphics[width=0.45\hsize]{f3a.ps}
460: \includegraphics[width=0.45\hsize]{f3b.ps}
461: }
462: \centerline{
463: \includegraphics[width=0.45\hsize]{f3c.ps}
464: \includegraphics[width=0.45\hsize]{f3d.ps}
465: }
466: \centerline{
467: \includegraphics[width=0.45\hsize]{f3e.ps}
468: \includegraphics[width=0.45\hsize]{f3f.ps}
469: }
470: \centerline{
471: \includegraphics[width=0.45\hsize]{f3g.ps}
472: \includegraphics[width=0.45\hsize]{f3h.ps}
473: }
474: \caption{Shape evolution in runs ({\it a}) SA1, ({\it b}) SA2, ({\it c}) SA3, ({\it d}) IA1,
475: ({\it e}) LA1, ({\it f}) P$_l$A1 and P$_l$A2, ({\it g}) P$_l$B1 and P$_f$B2 and ({\it h})
476: P$_l$B3. The solid lines show $b/a$, the dashed lines show $c/a$, and
477: the dot-dashed lines show $T$ (with scale indicated on the right-hand
478: side of each panel). The black, blue, and green lines are at $t=0$,
479: $t_g$, and $t_g + t_e$, respectively. In panel ({\it a}) black/yellow/red
480: shows the evolution if, after $t_g$, the disk is held at full mass for
481: a further 5 Gyr before it is evaporated. In panel ({\it f}), the standard
482: colors are for P$_l$A1, while P$_l$A2 is indicated in black/yellow/red.
483: Likewise, in panel ({\it g}), the standard colors are for P$_f$B2, while
484: black/yellow/red are for P$_l$B1. In all panels, the vertical cyan
485: line shows $r_{200}$. The standard errors on the plotted axis ratios
486: (see text for details) are $<0.08$; shells with larger measurement
487: errors, generally at small radius, are not plotted.
488: \label{fig:shapes}}
489: \end{figure*}
490: \setlength{\voffset}{0mm}
491:
492: \subsection{Short- and Intermediate-Axis Experiments}
493:
494: \begin{figure}
495: \centerline{
496: \includegraphics[width=\hsize]{f4.ps}
497: }
498: \caption{Ellipticity of the potential, \epot, in the disk
499: mid-plane of run SA1. The solid points are at $t=0$, and the open
500: points are at $t_g$, with squares for halo only and circles for
501: disk$+$halo.
502: \label{fig:potell}}
503: \end{figure}
504:
505: \thispagestyle{empty}
506: \setlength{\voffset}{-22mm}
507: \begin{figure*}
508: \centerline{
509: \includegraphics[width=0.45\hsize]{f5a.ps}
510: \includegraphics[width=0.45\hsize]{f5b.ps}
511: }
512: \centerline{
513: \includegraphics[width=0.45\hsize]{f5c.ps}
514: \includegraphics[width=0.45\hsize]{f5d.ps}
515: }
516: \caption{Evolution of the spherically averaged density and
517: kinematics in runs SA1 ({\it top left}), P$_l$A1 ({\it top right}), P$_l$B1 and
518: P$_f$B2 ({\it bottom left}), and P$_l$B3 ({\it bottom right}). The black,
519: blue, and
520: green lines correspond to $t=0$, $t_g$, and $t_g+t_e$. The red lines
521: show $t_g+t_e$ for P$_f$B2. The top panels shows the densities. In
522: the middle panels the solid lines indicate $\sigma_r$, the dashed
523: lines $\sigma_\phi$, and the dot-dashed lines $\sigma_\theta$. Here
524: the $z$-axis from which the angle $\theta$ is measured is the long
525: axis of the halo. The bottom panels show the anisotropy parameter,
526: $\beta$.
527: \label{fig:kinematics}}
528: \end{figure*}
529: \setlength{\voffset}{0mm}
530:
531: In run SA1 we grew a disk inside halo A with the minor axes of the
532: disk and halo aligned. This orientation is a natural one for disks to
533: form in since simulations have shown that the angular momenta of halos
534: are aligned with their {\it minor} axes \citep[\eg][]{war_etal_92,
535: por_etal_02, fal_etal_05}. Once the disk is grown to its full mass it
536: dominates the inner rotation curve (see Fig
537: \ref{fig:rotcurves}, left). The shape evolution of this highly prolate,
538: mildly triaxial halo is shown in Figure \ref{fig:shapes}a. The effect
539: of the massive disk on the halo shape is large: having started out
540: with $b/a \la 0.45$ it becomes much rounder in the plane of the disk
541: ($b/a \ga 0.6$ to 40 \kpc, \ie\ $\sim 0.2~ r_{200}$, the disk
542: constituting 17\% of the mass within this radius), as shown by the
543: blue lines. The change in shape perpendicular to the disk is more
544: modest and the inner halo becomes significantly more triaxial than when it
545: started out. In the disk plane, the combined potential starts out
546: very elliptical, $\epot \simeq 0.4$, and becomes quite round, with
547: $\epot < 0.1$ over the entire extent of the disk (Fig.
548: \ref{fig:potell}). This $\epot$ is sufficiently small to be
549: consistent with the observed scatter in the Tully-Fisher relation
550: \citep{fra_dez_92}, even without the additional axisymmetrization of
551: the potential that would be caused by the disk's orthogonal response.
552: Once the disk is evaporated, the resulting halo shape, shown by the
553: green lines in Figure \ref{fig:shapes}a, is very similar to its
554: original shape, with the net increase in both $b/a$ and $c/a$ being
555: $\la 0.1$ throughout the inner 100 \kpc. The final triaxiality is
556: barely changed from the starting one, despite the fact that the inner
557: halo was almost maximally triaxial at $t_g$.
558:
559: Likewise, the final density and anisotropy profiles, shown in Figure
560: \ref{fig:kinematics}, are not significantly changed, despite the
561: factor of $\sim 3.7$ increase in halo central density at $t_g$. The
562: halo anisotropy, $\beta = 1 - \sigma_t^2/\sigma_r^2$, where
563: $\sigma_t^2 = \frac{1}{2}(\sigma_\theta^2 + \sigma_\phi^2)$,
564: starts out $\beta \simeq 0$, grows to $\beta \ga 0.2$ at $t_g$ (\ie\
565: becomes radially anisotropic), and returns to $\beta \simeq 0$ at
566: $t_g + t_e$. If box orbits had been destroyed to any significant extent,
567: we would have seen instead an increase in tangential anisotropy
568: \citep[e.g.,][]{hol_etal_02}.
569:
570: The small difference between $t=0$ and $t_g+t_e$ in halo shape and
571: kinematics suggests that the halo phase space distribution has not been
572: grossly altered by the presence of the massive disk. There is little
573: evidence for a substantial amount of box orbit scattering, and any
574: chaos induced has to be quite mild. All this is true despite the
575: quite large change in halo shape and kinematics when the disk is at
576: full mass.
577:
578: Figure \ref{fig:shapes}a ({\it yellow and red lines}) also shows the
579: evolution when we left the disk at full mass for 5 Gyr before
580: evaporating it. The halo becomes slightly rounder (by about $b/a \la
581: 0.1$) at all radii both after the additional 5 Gyr and once the disk
582: is evaporated. The difference is largest inside $ \sim 30 \kpc$ where
583: the final $b/a$ and $c/a$ are about 0.1 larger at the end of the
584: simulation. This difference must be due to orbit scattering (either
585: physical or purely numerical); the fact that the difference between
586: these two runs is so much smaller than that between $t=0$ and $t_g$ implies
587: that scattering has only a mild effect on the halo shape. The global
588: shape change at $t_g$ can therefore be attributed to orbit deformation.
589:
590: Model SA1 had a disk with $R_{\rm d} = 3 \kpc$ and a baryon-to-dark
591: matter fraction, $f_b = 0.039$, consistent with estimates for local
592: galaxies \citep{jim_etal_03}. A more massive or more compact galaxy
593: may lead to greater scattering. We explored to what extent larger
594: $f_b$ or smaller $R_{\rm d}$ affect the halo shape in two further
595: simulations.
596: %
597: Run SA2 increased $M_b$ by a factor of 3 while keeping $R_{\rm d}$
598: fixed. The halo shape is changed significantly all the way out to
599: $r_{200}$ once the disk is evaporated, but remains quite prolate, with
600: $b/a < 0.6$ and $c/a < 0.5$, as can be seen in Figure
601: \ref{fig:shapes}b. In contrast, at $t_g$ the halo has $0.5 < b/a <
602: 1.0$ within the inner 50 kpc. Even with this high $f_b \simeq 0.12$,
603: or $\sim 70\%$ of the full cosmic baryon fraction \citep{spe_etal_06},
604: the irreversible change to the halo shape is $\la 50\%$ of the full
605: change at $t_g$ out to 100\kpc. Run SA3 instead set $R_{\rm d} =
606: 1.5\kpc$, keeping the ratio $z_{\rm d}/R_{\rm d}$ fixed (and
607: decreasing all softenings appropriately). The evolution in this case
608: is shown in Figure \ref{fig:shapes}c; as in run SA1, although the halo
609: at $t=t_g$ is substantially rounder than at the start, after the disk
610: is evaporated the halo recovers most of its original shape. Of runs
611: SA2 and SA3, making the disk more massive (SA2) produced a larger
612: irreversible change in the halo than did making it more compact.
613:
614: Run IA1 explored whether having the disk orthogonal to the
615: intermediate axis makes a significant difference to the halo
616: distribution, with all other parameters as in run SA1. The resulting
617: shape evolution is presented in Figure \ref{fig:shapes}d. In this
618: simulation the halo at $t_g$ remained more elongated than that in run SA1
619: despite having the same $M_b$. The axis ratios of the halo at $t_g$
620: cross over at $\sim30 \kpc$, where the halo's flattening orthogonal to
621: the disk causes the minor axis to switch from the disk plane to the
622: orthogonal direction. Once the disk is evaporated, the halo ends very
623: nearly axisymmetric in cross section in this inner region but
624: continues to be highly prolate. As in run SA1, the net change in halo
625: shape is relatively small at $t_g + t_e$.
626:
627:
628: \subsection{Long-Axis Experiments}
629:
630: In run LA1 we placed the disk with its symmetry axis along the long
631: axis of the halo. This orientation has been suggested to be favored
632: by the distribution of satellites around the Milky Way
633: \citep{zen_etal_05} and by the Sagittarius dwarf tidal stream
634: (\citet{helmi_04} but see \citet{fel_etal_06} for a different view). Other
635: than the disk's orientation, the parameters of this model are identical to
636: those of run SA1. As in that model, the halo in run LA1 is
637: significantly deformed to large radius by the growing disk, but it
638: recovers its shape nearly completely once the disk is evaporated, as
639: shown in Figure \ref{fig:shapes}e. Likewise, the spherically averaged
640: kinematic evolution of run LA1 is indistinguishable from that of SA1,
641: as seen in Figure \ref{fig:kinematics}.
642:
643: A unique characteristic of the evolution in long-axis experiments is
644: their tendency for the major axis of the inner halo to switch
645: orientation by $90\degrees$ into the disk plane once the disk grows
646: sufficiently massive. For run LA1 this is evident in Figure
647: \ref{fig:shapes}e, which shows that the halo is axisymmetric at $\sim
648: 20 \kpc$ (the solid blue line approaches $b/a \simeq 1$, while the
649: dot-dashed blue line approaches $T \simeq 0$) but is quite
650: prolate-triaxial at smaller radii. Major axis flips are more
651: clearly illustrated by the nearly axisymmetric halo B. In run LB1, as
652: \mdisk\ increases, shells of the prolate inner halo become spherically
653: symmetric. Further increase in \mdisk\ then leads to the shell
654: becoming not only flatter vertically (relative to the disk) but also
655: acquiring an {\it elongation} with its major axis in the plane of the
656: disk, \ie\ the symmetry axis of the inner halo flips by $90\degrees$
657: and becomes orthogonal to that of the outer halo (see Figs.
658: \ref{fig:3ddensity419} and \ref{fig:angles419}). The direction along
659: which this reorientation occurs is not random since the halo is
660: initially not perfectly axisymmetric on large scales. Continued
661: increase in \mdisk\ causes the symmetry axis to flip orientation to
662: larger radii, eventually saturating at $\sim 10$ \kpc.
663: %
664: The halo orientation flips do not occur when the disk is replaced by a
665: point particle. The dotted gray line in Figure \ref{fig:angles419}
666: shows the orientation of the major axis in run P$_l$B1, at a time when
667: its mass is the same as that in LB1; no flip in the major axis
668: direction can be seen.
669:
670: \begin{figure*}
671: {
672: \includegraphics[width=\hsize]{f6.ps}
673: }
674: \caption{Evolution of the inner halo in run LB1, seen in cross section
675: in the disk plane, with only the region $|z| < 5 \kpc$ shown, where
676: the $z$-axis is the symmetry axis of both disk and halo. The panels
677: show $t=0$ ({\it left}), $t_g$ ({\it middle}) and $t_g+t_e$ ({\it right}).
678: The halo is initially axisymmetric, becomes elongated orthogonal to
679: the symmetry axis at small radii at $t_g$, and largely recovers its
680: axisymmetry at $t_g + t_e$.
681: \label{fig:3ddensity419}}
682: \end{figure*}
683:
684: \begin{figure}
685: {
686: \includegraphics[width=\hsize]{f7.ps}
687: }
688: \caption{Major axis orientation evolution in run LB1. The solid
689: black and gray lines are at $t=0$ and $t_g/3$, the dashed black and
690: gray lines are at $2 t_g/3$ and $t_g$ and the black dotted line is at
691: $t_g+t_e$. The gray dotted line shows run P$_l$B1 at $t_g$.
692: \label{fig:angles419}}
693: \end{figure}
694:
695:
696: \subsection{Inclined Disk}
697:
698: \begin{figure}
699: \centerline{
700: \includegraphics[width=0.5\hsize]{f8a.ps}
701: }
702: \centerline{
703: \includegraphics[width=0.5\hsize]{f8b.ps}
704: }
705: \centerline{
706: \includegraphics[width=0.5\hsize]{f8c.ps}
707: }
708: \caption{Shape evolution in runs ({\it a}) TA1, ({\it b}) EA1, and ({\it c})
709: BA1. The
710: solid lines show $b/a$, the dashed lines show $c/a$ and the dot-dashed
711: lines show $T$ (with scale indicated on the right hand side of each
712: panel). The black, yellow, and red lines show $t=0$, $t_g$, and $t_g +
713: t_e$. For comparison, the equivalent evolution for run SA1 is shown
714: by the blue and green lines.
715: \label{fig:nonsymmetric}}
716: \end{figure}
717:
718:
719: Simulations have found that the angular momenta of the halo and
720: gas need not be aligned \citep{vdb_etal_02, che_etal_03},
721: although the inner halo and the disk that would form settle to a
722: common plane \citep{dub_kui_95}. Therefore we also explored the
723: effect of a disk inclined relative to the main plane of the dark
724: matter halo. Run TA1 is based on run SA1 but with the disk inclined by
725: $30\degrees$ about the $y-$axis (intermediate axis), all other
726: parameters being the same. The resulting evolution is virtually
727: indistinguishable from that in run SA1 at all times, as shown in
728: Figure \ref{fig:nonsymmetric}a.
729:
730:
731: \subsection{Elliptical Disk}
732:
733: A disk forming in an elliptical potential becomes elongated with its
734: major axis orthogonal to that of the potential
735: \citep[e.g.,][]{ger_vie_86}. In run EA1 we replaced the disk in run
736: SA1 by an elliptical disk, with its long axis along the $y-$axis. We
737: obtain this oval disk by shrinking the $x$ coordinates (parallel to
738: the halo major axis) of all disk particles by a factor of 0.75;
739: i.e., the ellipticity of the disk density was $\epsilon = 1 - b/a =
740: 0.25$. The degree to which
741: elliptical disks change halo distributions is over estimated by this
742: simulation since the disk ellipticity is quite large for a massive
743: disk.
744: We then evolved this system identically to run SA1, keeping
745: the disk fixed in place. Figure \ref{fig:nonsymmetric}b compares the
746: shape evolution with that in run SA1. At $t_g + t_e$ the halo is left
747: significantly rounder within $\sim 20 \kpc$ than it was in run SA1,
748: but beyond that the evolution is very similar.
749:
750:
751: \subsection{Central Softened Point Masses}
752:
753: All of the experiments described above had rigid disks frozen in
754: place. While we have been careful to recenter the halo in position
755: and velocity after the mergers and before growing the disks, some
756: residual relative motion of the inner and outer parts of the halos
757: remained. This motion is damped as the mass of the disk increases,
758: possibly causing some scattering of orbits. In order to test for
759: artifacts associated with such damping, we resorted to simulations
760: with only a single baryonic particle and compared the evolution when
761: the particle is free to move (P$_l$B1) and when it is frozen in place
762: (P$_f$B2). The rotation curve at $t_g$ is shown in Figure
763: \ref{fig:rotcurves}. Figure \ref{fig:shapes}g shows their shape
764: evolution; in both cases, the shape is largely recovered at $t_g +
765: t_e$. If anything, P$_l$B1 is very slightly rounder at 10 \kpc\ than
766: P$_f$B2 compared with the initial halo. Figure
767: \ref{fig:kinematics} shows that their density and kinematic evolution
768: also is largely reversible. Thus, our use of rigid disks nailed in
769: place could not have induced much artificial orbit scattering.
770:
771:
772: \subsection{Ultrahard particle}
773:
774: In runs P$_l$B1 and P$_f$B2 the growing particle had a softening of
775: $\epsilon = 3 \kpc$, a reasonable size for a galaxy. In run P$_l$B3
776: we decreased the softening length of the particle to $100 \pc$ and
777: $M_b$ by half. Despite the smaller $M_b$ the final halo after $t_g +
778: t_e$ remains substantially rounder inside $20 \kpc$ than in those runs
779: (but is largely recovered at larger radii). If the central particle
780: were a black hole, its sphere of influence assuming $\sigma_0 = 100
781: \kms$ from $t=0$ would be $\sim 15 \kpc$. This is comparable to the
782: radius out to which the particle irreversibly alters the shape of the
783: halo. Likewise, the halo mass within $20\kpc$ is comparable to that
784: of the central particle: at $t=0$, the halo mass within this radius is
785: $4 M_b$.
786: %
787: Unlike runs P$_l$B1 and P$_f$B2, Figure \ref{fig:kinematics} shows
788: that the kinematic evolution is not reversible, and the halo of run
789: P$_l$B3 becomes significantly tangentially anisotropic, as expected if
790: box orbits are destroyed.
791: %
792: Despite the different final state, Figure \ref{fig:shapes}h also shows
793: that the halo shape at $t_g$ is not much different from that in
794: P$_l$B1 and P$_f$B2, implying that halo shape change is not dominated
795: by scattering.
796: %
797: Whereas run SA3 with $R_{\rm d} = 1.5 \kpc$, which is not unreasonably
798: small for most galaxies, did not significantly cause box orbit
799: destruction, the $\sim 10$ times more centrally concentrated run
800: P$_l$B3 is able to cause a large irreversible change to the halo shape
801: out to $\sim 0.3~ r_{200}$. However, $R_{\rm d}/r_{200} \simeq 0.001$
802: is unrealistically small.
803:
804: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
805:
806: \section{Angular Momentum Transport}
807: \label{sec:angmomtran}
808:
809: We next explore the effect of angular momentum transfer to the halo.
810: Such transfer is irreversible so the change inflicted on the halo must
811: also be irreversible. How strongly the halo distribution is changed
812: depends on the mechanism by which angular momentum is transferred. If
813: via bars or spirals then we may expect that the changes
814: are mostly at small radius. If
815: angular momentum is transferred by satellite galaxies, however, then
816: the effect on the halo is likely to be much more widespread.
817:
818: \subsection{Live Barred Disk}
819:
820: In run BA1 we evolved a model with a live disk for 10 Gyr after $t_g$
821: before evaporating the disk. The initial system was similar to run
822: SA1 but with only $30 \%$ of its $M_b$. We chose this lower mass
823: because the same mass as SA1 leads to a long-lived bar, whereas we are
824: interested in forming a bar that gets destroyed in order to be able
825: to evaporate the disk. A bar quickly formed and was subsequently
826: destroyed \citep{ber_etal_06}. After 10 Gyr, we fixed the disk
827: particles in place and evaporated the disk.
828: %
829: Very little of the inner halo shape is recovered after the disk is
830: evaporated. In the inner $\sim 20$ kpc the halo remains rounder than
831: at $t=0$, with both $b/a$ and $c/a$ larger by $\ga 0.1$. The
832: irreversible change in the halo is associated with the transfer of
833: angular momentum from disk to halo \citep{weinbe_85, deb_sel_98}.
834: %
835: Run BA1 produces a comparable change in the inner 20 kpc of the halo
836: as did the $10$ times more massive run SA2, but the shape change is
837: much smaller farther out. Since the bar transfers angular momentum to the
838: halo at resonances \citep{weinbe_85}, and the strongest of these are at
839: smaller radii (most of the angular momentum gained by the halo is
840: within the inner $\sim 10 \kpc$), this accounts for the relatively small
841: radial extent of the halo shape change.
842:
843: \subsection{Satellites}
844:
845: Baryons need not cool directly onto the central disk but
846: onto satellites instead. The presence of large numbers of dark
847: satellites is one of the main predictions of CDM \citep{moo_etal_99,
848: kly_etal_99, ghi_etal_00}. As they sink, satellites lose angular
849: momentum to the halo; in the process box orbits may be scattered.
850: %
851: We explored this evolution with models P$_l$A1 and P$_l$A2. Starting
852: with halo A, we selected 10 particles that stay within 200 \kpc\ but
853: otherwise at random, and adiabatically increased each of their masses
854: to give the same total baryonic mass as in run SA1. We grew these
855: satellites to full mass and then evaporated them. Each satellite had
856: $\epsilon = 0.5 \kpc$ in P$_l$A1 and $5 \kpc$ in P$_l$A2. Since we
857: only used softened point particles as satellites, which cannot be
858: tidally stripped, their effect on the halo is larger than it would be
859: in nature. Of the 10 satellites, only one remained at $r > 50 \kpc$,
860: the rest having fallen to $R < 25 \kpc$ by the end of the simulation.
861: %
862: The evolution of the halo shape in these two models is presented in
863: Figure \ref{fig:shapes}f. After the particles reach their full mass,
864: the halo of run P$_l$A1 is about as round as that in run SA1. However, the
865: halo does not recover much of its original shape after the particles
866: are evaporated. Clearly, the distribution function of the halo has
867: been altered to a large extent.
868: %
869: Figure \ref{fig:kinematics} shows that angular momentum transferred by
870: baryons to the halo can erase the cusp, in agreement with previous
871: results \citep{ton_etal_06, mas_etal_06, rea_etal_06, wei_kat_07}
872: although the contraction caused by the growing central mass masks the
873: core.
874: %
875: The halo shape change at $t_g + t_e$ is very similar in the two runs
876: because of quite similar angular momentum absorbed by the halo.
877: Because of the difference in softenings, the different baryonic potential
878: at the halo center at $t_g$ accounts for the $b/a \sim 0.1$
879: difference in shape, with the softer potential supporting the more
880: elongated shape.
881:
882: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
883:
884: \section{Orbital Evolution}
885: \label{sec:orbitanalysis}
886:
887: We explored directly the evolution of the orbital character of the
888: models by considering a subsample of 1000 particles in run SA1 and
889: following their orbits at various points in the simulation. The 1000
890: particles were randomly chosen from the $t=0$ distribution such that
891: they were inside $r=200\kpc$. We then integrated their motion as test
892: particles while holding all the other particles fixed in place. We
893: used a fixed timestep of 0.1 Myr and integrated for 15 Gyr, storing
894: the phase space coordinates of each test particle every 1 Myr. For
895: the same 1000 particles, we carried out this operation at $t=0$,
896: $t_g$, and $t_g+t_e$. Because we froze the background potential, in
897: effect we have computed the orbital character of the particles at
898: these three times. The fact that we integrated for 15 Gyr ensures
899: that we have sufficient points on each orbit to properly characterize
900: it.
901: %
902: In this paper we demonstrate with a few examples that a large fraction
903: of box-like orbits at $t=0$ return to very similar box-like orbits at
904: $t_g+t_e$, showing that deformation, not transformation, is
905: responsible for shape change in most cases. We do this by presenting
906: their configuration space projection at each of the three different times.
907: Such an analysis cannot distinguish between box orbits and mildly
908: chaotic, elongated orbits but this is unimportant anyway for our
909: present purposes since we have integrated for over a Hubble time. If
910: they are mildly chaotic, they can still support a triaxial halo. A
911: full analysis of the orbital structure using more sophisticated
912: techniques will be presented elsewhere.
913:
914: Of the 1000 orbits, we start by presenting nine particles, initially on
915: boxlike orbits, defined such that, at $t=0$, they
916: (1) remain inside $25 \kpc$, (2) do not have a fixed sense of rotation
917: relative to any of the three major axes, (3) reach a radius of at
918: least $10 \kpc$, and (4) get within $0.2 \kpc$ of the center.
919: %
920: The evolution of many of the other 991 orbits is qualitatively similar
921: to that of the nine presented here. Figure \ref{fig:orbits563} projects
922: these orbits onto the halo symmetry planes, where the $x$-axis is the
923: halo's major axis and the $z$-axis is the disk's symmetry axis. Most
924: orbits at $t_g+t_e$ are quite similar to what they looked like at
925: $t=0$. None of the orbits seem strongly chaotic, neither at $t_g$ nor
926: at $t_g + t_e$, although they may be weakly chaotic. Moreover, most
927: orbits retain a box-like shape at $t_g$, but have a significantly
928: rounder shape than those at $t=0$. At $t_g$, three of the initially
929: box-like orbits become round (orbits "a", "f" and "h"); of
930: these only orbit h changes character completely, becoming a loop
931: orbit. Some of the orbits have a slight banana shape; in the full
932: sample of orbits we found many cases of strongly banana-shaped orbits.
933: These had a tendency to become more planar but are still distinctly
934: elongated at $t_g+t_e$. In a few cases we also found the opposite
935: occurring --- slightly banana orbits becoming more strongly curved ---
936: but this was less common. Of the box-like orbits in Figure
937: \ref{fig:orbits563} some are rounder in the $(x,y)$ plane at $t_g+t_e$
938: (\eg, orbits a, f and i), but some are rounder at $t=0$
939: (\eg, orbits c and e), suggesting that differences in shape are due to
940: scattering. What little difference in orbit shape occurs between
941: $t=0$ and $t_g+t_e$ can probably be attributed to numerical noise.
942: Most importantly, while there is a clear orbit shape deformation at
943: $t_g$, little orbital transformation has occurred.
944:
945: We quantify the orbital deformation of the sample of 1000 particles by
946: plotting in Figure \ref{fig:orbittrans563} $\sigma_y/\sigma_x$,
947: where $\sigma_x^2 = \sum_t x_t^2$ and similarly for $\sigma_y^2$ and
948: the sum is over timesteps. The significantly rounder shape of orbits
949: at $t_g$ than $t=0$ is apparent, with the vast majority of
950: orbits initially aligned with the halo having $\sigma_y/\sigma_x$
951: closer to unity at the later time. Orbits initially elongated along
952: the halo's minor axis, as well as orbits initially rounder than
953: $\sigma_y/\sigma_x \ga 0.6$ end up round, with $\sigma_y/\sigma_x
954: \simeq 1$. Instead at $t_g + t_e$ the orbits tend to return to their
955: initial elongation, especially for the most elongated orbits. The
956: right panel shows the distribution of $\sigma_y/\sigma_x$; orbits
957: become substantially rounder at $t_g$ but the population as a whole
958: recovers the original distribution to a large extent once the disk is
959: evaporated.
960:
961: \begin{figure*}
962: {
963: \includegraphics[width=\hsize]{f9.ps}
964: }
965: \caption{Sample of initially boxlike orbits in run SA1. In each
966: set of panels, the top row is at $t=0$, the middle row is at $t_g$, and
967: the bottom row is at $t_g+t_e$; from left to right the panels show
968: projections onto the $(x,y)$, $(x,z)$, and $(y,z)$ planes. Each orbit
969: has been integrated for 15 Gyr from each of $t=0$, $t_g$ and
970: $t_g+t_e$.
971: \label{fig:orbits563}}
972: \end{figure*}
973:
974: \begin{figure}
975: {
976: \includegraphics[angle=-90.,width=\hsize]{f10.ps}
977: }
978: \caption{Deformation of orbits in run SA1. In the left panels,
979: the dashed line shows the diagonal. The solid points show the nine
980: particles of Fig. \ref{fig:orbits563}. In the right panel, the
981: solid, dashed, and dot-dashed lines show the distributions at $t=0$,
982: $t_g$, and $t_g+t_e$, respectively. The thick lines are for the full
983: distribution of 1000 particles, while the thin lines are for those
984: particles that remain inside $r=50\kpc$ during the 15 Gyr integration
985: starting at $t=0$.
986: \label{fig:orbittrans563}}
987: \end{figure}
988:
989: A comparison of orbital evolution in runs P$_f$B2 and P$_l$B3 provides
990: an example of orbit transformation. We again selected 1000 orbits
991: from particles within the inner $200 \kpc$ of halo B at $t=0$. We
992: integrated their orbits as above but we used the smaller timestep
993: $\delta t = 10^4 $ years in the case of P$_l$B3; for P$_f$B2 we use
994: $t_g/2$ when the central particle has the same mass as at $t_g$ in
995: model P$_l$B3. Although we are comparing the two models at the same
996: central particle mass, the orbits at time $t_g + t_e$ in model P$_f$B2
997: were computed after the central particle was evaporated from a mass
998: twice that reached in P$_l$B3. As before, we present in Figure
999: \ref{fig:orbits448} nine orbits that at $t=0$ are box-like.
1000: These boxlike orbits in P$_l$B3 are more often transformed
1001: than those in run P$_f$B2.
1002: %
1003: In model P$_f$B2, only one orbit (orbit f) appears to have
1004: changed substantially at the end of the simulation, while orbits a
1005: and d become fish orbits (although they may have been
1006: librating about fish orbits at $t=0$). Orbit b is largely unchanged and the
1007: remaining orbits are all boxlike. In model P$_l$B3, orbit e is
1008: changed about as much as orbit f in P$_f$B2. However, four of the nine
1009: orbits, f-i, are very strongly transformed by $t_g + t_e$ and are
1010: no longer able to support a triaxial shape.
1011: %
1012: Figure \ref{fig:orbittrans448} compares the distribution of all 1000
1013: particles and a depletion of elongated orbits is evident in P$_l$B3
1014: compared with run P$_f$B2.
1015:
1016: \begin{figure*}
1017: {
1018: \includegraphics[width=0.5\hsize]{f11a.ps}
1019: \includegraphics[width=0.5\hsize]{f11b.ps}
1020: }
1021: \caption{Sample of initially boxlike orbits in run P$_f$B2 ({\it left three
1022: columns}) and P$_l$B3 ({\it right three columns}). We show the evolution of the
1023: same nine starting orbits in the two different models. In each set of
1024: panels, the top row is at $t=0$, the middle row is at $t_g/2$
1025: (P$_f$B2) or $t_g$ (P$_l$B3), and the bottom row is at $t_g+t_e$; from
1026: left to right the panels show projections onto the $(x,y)$, $(x,z)$,
1027: and $(y,z)$ planes. Each orbit has been integrated for 15 Gyr from
1028: each starting point.
1029: \label{fig:orbits448}}
1030: \end{figure*}
1031:
1032: \begin{figure*}
1033: {
1034: \includegraphics[angle=-90.,width=0.5\hsize]{f12a.ps}
1035: \includegraphics[angle=-90.,width=0.5\hsize]{f12b.ps}
1036: }
1037: \caption{Identical to Fig. \ref{fig:orbittrans563} but for runs
1038: P$_f$B2 ({\it left}) and P$_l$B3 ({\it right}). The final distribution
1039: of orbits
1040: is significantly depleted of elongated orbits in run P$_l$B3 compared
1041: with run P$_f$B2. All nine orbits of Fig. \ref{fig:orbits448} are
1042: above the diagonal in run P$_l$B3 at $t_g+t_e$ but scatter about the
1043: diagonal in run P$_f$B2.
1044: \label{fig:orbittrans448}}
1045: \end{figure*}
1046:
1047:
1048: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1049:
1050: \section{Discussion}
1051: \label{sec:conc}
1052:
1053: \subsection{Timesteps}
1054:
1055: We also performed a number of
1056: tests of the numerics to verify that our results are robust. The main
1057: concern is the timestep used. \citet{she_sel_04} found that too large
1058: timesteps result in bars being destroyed too easily, because orbits
1059: are not followed accurately near the central mass. Our simulations
1060: used multi-stepping. With a base timestep $\Delta t$, particles move
1061: on timesteps $\Delta t/2^n$, where $n$ is the rung level satisfying
1062: the condition $\delta t = \Delta t/2^n < \eta (\epsilon/a)^{1/2}$, and
1063: $\epsilon$ is the particle's softening, $a$ is its acceleration and
1064: $\eta$ a tolerance parameter. We used $\eta = 0.2$, a conservative
1065: value; with a base timestep $\Delta t = 5$ Myr, simulation SA1 at
1066: $t_g$ had a range of timesteps down to $5/2^5 = 0.16$ Myr. If instead
1067: we set $\eta = 2$ the timestep distribution only reaches to $5/2 =
1068: 2.5$ Myr. The effect of these larger timesteps, shown in Figure
1069: \ref{fig:tstest}, is manifest at $r \la 20 \kpc$, which remains
1070: significantly rounder at $t_g+t_e$ than when $\eta = 0.2$. The quite
1071: modest net shape change in our simulations implies that the timesteps
1072: we used were sufficiently small to correctly follow the evolution near
1073: the center.
1074:
1075: \begin{figure}
1076: {
1077: \includegraphics[width=\hsize]{f13.ps}
1078: }
1079: \caption{Effect of timestep size on the evolution of run SA1. The
1080: black lines shows the shape at $t=0$ while the gray lines are for $t_g
1081: + t_e$, with the thick line the standard result with $\eta = 0.2$ and
1082: the thin lines for $\eta = 2$. As in Fig. \ref{fig:shapes}, the
1083: solid lines show $b/a$ while the dashed lines show $c/a$. The larger
1084: timesteps lead to an evolution that is less reversible.
1085: \label{fig:tstest}}
1086: \end{figure}
1087:
1088:
1089: \subsection{Evidence against box orbit destruction}
1090:
1091: We have demonstrated that the substantial axisymmetrization caused by
1092: a disk growing inside a dark matter halo is largely, although not
1093: wholly, reversible. If triaxial halos become rounder because of box
1094: orbit destruction \citep[\eg][]{mac_etal_07}, then these orbits would
1095: have to be repopulated in order for the halo to recover its original
1096: shape once the disk is evaporated. Apart from being unlikely for such
1097: a highly ordered system as a triaxial halo, this interpretation is not
1098: supported by our orbital analysis, which shows that box-like orbits
1099: are deformed by realistic disks but not transformed by scattering into
1100: new orbits. The strongest evidence that deformation is a more
1101: important process than transformation comes from the ability of
1102: particles to return, after the disk is evaporated, to nearly the same
1103: orbits as they started from. This makes it implausible that the
1104: shape change is due to a large increase in strong chaos. Instead we
1105: find that realistic axisymmetric disks are not concentrated or massive
1106: enough to cause substantial chaos. In a similar vein,
1107: \citet{hol_etal_02} found that chaos in triaxial ellipticals is
1108: induced by the black hole only, not by the stellar cusp surrounding
1109: it. This need not mean, however, that chaos is not enhanced by the
1110: disk. It could well be that orbits are becoming
1111: weakly chaotic but do not diffuse sufficiently on a Hubble timescale
1112: to significantly weaken triaxiality. The shape
1113: evolution due to baryons cooling onto a central galaxy can therefore
1114: be computed from adiabatic invariants.
1115:
1116: The primary role of orbit deformation over transformation is also
1117: indicated by the much smaller effects of scattering when it is clear
1118: that scattering has occurred. In run SA1 scattering occurred when the
1119: disk was maintained at full mass for an additional 5 Gyr before being
1120: evaporated. This only caused a small additional change in the final
1121: halo shape. Furthermore, comparing runs P$_l$B1 and P$_l$B3, we find
1122: that the halo shape at $t_g$ is rather similar, despite the fact that
1123: in run P$_l$B3 box orbits are significantly destroyed. The same is
1124: true for runs SA1 and P$_l$A1. These examples directly illustrate
1125: that box orbit destruction is a much smaller factor in halo shape
1126: changes than is orbit deformation.
1127:
1128: Weak chaos is consistent with the orbital characterization of
1129: \citet{mac_etal_07} but does not support their claim that the shape
1130: change is largely due to enhanced chaos from the central baryonic
1131: mass. Another possibility is that in their simulations chaos could
1132: have been caused by gas cooling inside subhalos, which we showed leads
1133: to a substantial change in the phase space distribution of the halo.
1134:
1135:
1136: \subsection{Implications for galaxy formation simulations}
1137:
1138: If baryons cool onto substructures within the halo then box orbits are
1139: very efficiently destroyed. Since the evolution of disks can be
1140: strongly influenced by halo triaxiality \citep[\eg][]{ide_hoz_00}, any
1141: process that artificially reduces triaxiality can lead to biases in
1142: the properties of galaxies forming in cosmological simulations.
1143: \citet{age_etal_07} demonstrate that the evolution of gas blobs in
1144: smoothed particle hydrodynamics (SPH) simulations is different from
1145: that found in Eulerian gas codes. They interpreted this difference as
1146: being due to the unphysically poor mixing of traditional SPH, which
1147: allows blobs to survive longer in SPH. Moreover, satellites in
1148: cosmological simulations tend to have denser, more concentrated gas
1149: components than their real counterparts, which makes them harder to
1150: strip by ram pressure and tides \citep{may_etal_07} and thus more
1151: likely to artificially enhance halo shape changes. Baryonic cooling
1152: inside substructures is strongly suppressed by feedback from
1153: supernovae \citep{dek_sil_86, gov_etal_04, gov_etal_07}. If
1154: simulations do not treat feedback properly or have low resolution,
1155: then baryons may condense into concentrated substructures biasing the
1156: global evolution of simulated galaxies.
1157:
1158:
1159: \subsection{Summary}
1160:
1161: Our results can be summarized as follows:
1162:
1163: \begin{enumerate}
1164:
1165: \item The adiabatic growth of disks with realistic sizes and masses
1166: inside prolate/triaxial halos leads to a large change in the shape of
1167: the halo. Axis ratios change by $>0.2$ out to roughly $0.5~ r_{200}$.
1168: The growth of the disk drives the halo kinematics to larger radial
1169: anisotropy. The midplane global potential ellipticity is less than
1170: 0.1, consistent with the small scatter in the Tully-Fisher relation
1171: \citep{fra_dez_92}.
1172:
1173: \item Despite these large changes, the underlying phase space
1174: distribution is not grossly altered, as we verified by artificially
1175: evaporating the disk and recovering, to a large extent, the original
1176: halo. The irreversible change in final halo structure is larger for
1177: more massive or more centrally concentrated disks, but is still a
1178: relatively small fraction of the total shape change when the disk is
1179: at full mass.
1180: %
1181: As in the case of black holes at the centers of cuspy elliptical
1182: galaxies \citep{ger_bin_85, hol_etal_02}, the bulk of the irreversible
1183: halo shape change occurs in the inner region of the galaxy. This
1184: small irreversible shape change is driven by orbit scattering.
1185:
1186: \item Box orbit destruction cannot be the right interpretation for the
1187: shape change caused by disk growth. Such a process is not reversible
1188: but we found that a large fraction of particles on box-like orbits
1189: individually return to very similar orbits after the disk is
1190: evaporated. At most only mild chaos is induced. Instead we find that
1191: box orbits become deformed by the growing disk, but retain their
1192: character, and this seems sufficient to explain the change in shape.
1193: %
1194: In the absence of angular momentum transport or extreme
1195: mass/concentration galaxies, very little of the quite large shape
1196: change that dark matter halos undergo as baryons condense inside them
1197: is due to box orbit destruction. As a result, shape change can be well
1198: approximated by adiabatic invariants.
1199:
1200: \item Very concentrated structures do lead to scattering and
1201: to large irreversible changes in halo shape and kinematics; we
1202: found that scales of $\sim 100 \pc$ are needed to
1203: accomplish this. The irreversible shape change was then
1204: restricted to the sphere of influence of this pointlike mass:
1205: $r \sim GM/\sigma^2$. Massive disks are
1206: also able to change the halo structure irreversibly, but the mass
1207: required, $\sim 70\%$ of the cosmic baryon fraction, is quite high. Even
1208: in such cases of scattering, the halo shape with the baryons at full
1209: mass is not much different from similar simulations with little
1210: scattering, suggesting that it is orbital deformation in the first
1211: place that changes the shape of the halo.
1212:
1213: \item If baryons transport angular momentum to the halo, a large
1214: irreversible change in halo shape and kinematics occurs. Such
1215: transfers can occur either because of gas condensing in satellites or
1216: nonaxisymmetric structures forming in the disk. Even quite low mass
1217: disks are able to alter the inner halo distribution if they can
1218: transport angular momentum to the halo.
1219: %
1220: The effect of satellites can be artificially large in simulations of
1221: galaxy formation because of the poor mixing in SPH \citep{age_etal_07}
1222: and the too highly concentrated satellites that form
1223: \citep{may_etal_07}.
1224:
1225: \item When the disk minor axis and halo major axis are aligned, growth
1226: of the disk leads to an elongation within the plane of
1227: the disk, even when the initial halo is very nearly axisymmetric.
1228:
1229: \end{enumerate}
1230:
1231:
1232: \acknowledgements This paper is based in part on work supported by
1233: the National Science Foundation under the following NSF programs:
1234: Partnerships for Advanced Computational Infrastructure, Distributed
1235: Terascale Facility (DTF), and Terascale Extensions: Enhancements to the
1236: Extensible Terascale Facility. Many additional simulations were
1237: carried out at the University of Z\"urich on zBox and at the Arctic
1238: Region Supercomputing Center.
1239: %
1240: V. P. D. thanks the University of Z\"urich for hospitality during part
1241: of this project. V. P. D. is supported by a Brooks Prize Fellowship in
1242: Astrophysics at the University of Washington and receives partial
1243: support from NSF ITR grant PHY-0205413. S. K. acknowledges support by
1244: the US Department of Energy through a KIPAC Fellowship at Stanford
1245: University and the Stanford Linear Accelerator Center.
1246: %
1247: We thank Andrea Macci\`{o} and Ioannis Sideris for fruitful
1248: discussions and the anonymous referee for comments that helped to
1249: improve this paper. V. P. D. thanks Monica Valluri for discussions that
1250: helped improve this paper.
1251:
1252: %%%%%%%%%%%%%%%
1253: % Reference List
1254: %%%%%%%%%%%%%%%
1255:
1256: \bibliographystyle{aj.bst}
1257: \bibliography{ms.bbl}
1258: %\bibliography{allrefs}
1259:
1260: %%%%%%%%%%%%%%%
1261: % End of Document
1262: %%%%%%%%%%%%%%%
1263:
1264: \end{document}
1265: