0806.1739/ms.tex
1: % !iTeXMac(typeset): altpdflatex --keep-psfile ${iTMInput}
2: % !iTeXMac(compile): "./local Command"
3: %\documentclass[12pt,preprint]{aastex}
4: \documentclass{emulateapj}
5: \usepackage{apjfonts}
6: \bibliographystyle{apj}
7: 
8: %\newcommand{\scaleup}{}
9: %\newcommand{\scaledown}{}
10: %\newcommand{\plotter}{\includegraphics[scale=0.70]}
11: %\newcommand{\plotterr}{\includegraphics[scale=0.50]}
12: %\newcommand{\breaker}{\clearpage}
13: %\newcommand{\longtabler}{}
14: %\newcommand{\tableast}{\ast}
15: %\newcommand{\appendixcolumns}{}
16: \newcommand{\scaleup}{\epsscale{1.1}}
17: \newcommand{\scaledown}{\epsscale{0.9}}
18: \newcommand{\plotter}{\plotone}
19: \newcommand{\plotterr}{\plotone}
20: \newcommand{\breaker}{}
21: \newcommand{\tableast}{$\ast$}
22: \newcommand{\longtabler}{\LongTables}
23: \newcommand{\appendixcolumns}{\twocolumngrid}
24: 
25: \newcommand{\tableset}{deluxetable}
26: \newcommand{\tableclear}{\clearpage}
27: \newcommand{\Mdot}{\dot{M}}
28: \newcommand{\mdot}{\dot{m}}
29: \newcommand{\etal}{et al.}
30: \newcommand{\mbh}{M_{\rm BH}}
31: \newcommand{\mstar}{M_{\ast}}
32: \newcommand{\msun}{M_{\sun}}
33: \newcommand{\lstar}{L_{\ast}}
34: \newcommand{\phistar}{\phi_{\ast}}
35: \newcommand{\pstar}{\phistar}
36: \newcommand{\reducedchi}{\chi^{2}/\nu}
37: \newcommand{\qeos}{q_{\rm eos}}
38: \newcommand{\fgas}{f_{\rm gas}}
39: \newcommand{\mdyn}{M_{\rm dyn}}
40: \newcommand{\re}{R_{\rm e}}
41: \newcommand{\fsb}{f_{\rm sb}}
42: \newcommand{\mdisk}{M_{\rm disk}}
43: \newcommand{\scalelen}{R_{\rm d}}
44: \newcommand{\barangle}{\phi_{b}}
45: \newcommand{\orbitfreq}{\Omega_{\rm o}}
46: \newcommand{\diskfreq}{\Omega_{\rm d}}
47: \newcommand{\degree}{^{\circ}}
48: 
49: 
50: \shorttitle{How Do Disks Survive Mergers?}
51: \shortauthors{Hopkins \etal}
52: \slugcomment{Submitted to ApJ, June 5, 2008}
53: \begin{document}
54: 
55: \title{How Do Disks Survive Mergers?} 
56: \author{Philip F. Hopkins\altaffilmark{1}, 
57: Thomas J. Cox\altaffilmark{1,2}, 
58: Joshua D. Younger\altaffilmark{1}, 
59: \&\ Lars Hernquist\altaffilmark{1}
60: }
61: \altaffiltext{1}{Harvard-Smithsonian Center for Astrophysics, 
62: 60 Garden Street, Cambridge, MA 02138}
63: \altaffiltext{2}{W.~M.\ Keck Postdoctoral Fellow at the 
64: Harvard-Smithsonian Center for Astrophysics}
65: 
66: \begin{abstract}
67: 
68: We develop a general physical model for how 
69: galactic disks survive and/or are destroyed in mergers and interactions. 
70: Based on simple dynamical arguments, we show that gas primarily loses 
71: angular momentum to internal torques in a merger, induced by 
72: the gravity of the secondary. Gas within 
73: some characteristic radius, determined by the efficiency of 
74: this angular momentum loss (itself a function of the orbital parameters, 
75: mass ratio, and gas fraction of the merging galaxies), will 
76: quickly lose angular momentum to the stars sharing the perturbed host disk, fall to the 
77: center and be consumed in a starburst. We 
78: use a similar analysis to determine where violent relaxation of the pre-merger 
79: stellar disks is efficient on final coalescence. 
80: Our model describes both the dissipational and dissipationless 
81: components of the merger, and allows us to predict, for a 
82: given arbitrary encounter, the stellar and gas content of the material
83: that will survive (without significant angular momentum loss or violent relaxation) 
84: to re-form a disk in the merger remnant, versus being dissipationlessly 
85: violently relaxed or dissipationally losing angular momentum and 
86: forming a compact central starburst. 
87: We test these predictions with a large library of hydrodynamic merger 
88: simulations, and show that they agree well (with small scatter) 
89: with the properties of simulated merger remnants as a function of 
90: merger mass ratio, orbital parameters, and gas distributions, 
91: in simulations which span a wide range of parameter space 
92: in these properties as well as prescriptions for gas physics, 
93: stellar and AGN feedback, halo and initial disk structural 
94: properties, redshift, and galaxy masses. 
95: We show that, in an immediate (short-term) sense, the amount of stellar 
96: or gaseous disk that survives or re-forms 
97: following a given interaction can be understood purely 
98: in terms of simple, well-understood gravitational physics, 
99: independent of the details of the ISM gas physics or 
100: stellar and AGN feedback. 
101: This allows us to demonstrate and quantify how these physics are 
102: in fact important, in an indirect sense, to enable disks to survive 
103: mergers, by lowering star formation efficiencies in low mass systems 
104: (allowing them to retain large gas fractions) and distributing the 
105: gas to large radii. The efficiency of disk destruction in mergers is a 
106: strong function of gas content -- our model allows us to explicitly 
107: predict and demonstrate how, in sufficiently gas rich mergers (with quite 
108: general orbital parameters), even 1:1 mass-ratio mergers can yield 
109: disk-dominated remnants, and more realistic 1:3-1:4 mass-ratio 
110: major mergers can yield systems with $<20\%$ of their mass in bulges. 
111: We discuss a number of implications of this modeling for 
112: the abundance and morphology of bulges as a function of mass and 
113: redshift, and provide simple prescriptions for the implementation of our 
114: results in analytic or semi-analytic models of galaxy formation. 
115: 
116: \end{abstract}
117: 
118: \keywords{galaxies: formation --- galaxies: evolution --- galaxies: active --- 
119: galaxies: spiral --- cosmology: theory}
120: 
121: \section{Introduction}
122: \label{sec:intro}
123: 
124: In the now established ``concordance'' $\Lambda$CDM cosmology, 
125: structure grows hierarchically \citep[e.g.][]{whiterees78}, making mergers and interactions 
126: between galaxies an essential and inescapable process in 
127: galaxy formation. Indeed, mergers are widely believed to be 
128: responsible for the morphologies of spheroids \citep[bulges in disks 
129: and elliptical galaxies;][]{toomre77}, and 
130: observations find recent merger remnants in 
131: considerable abundance in the local universe 
132: \citep{schweizer82,LakeDressler86,Doyon94,ShierFischer98,James99,
133: Genzel01,tacconi:ulirgs.sb.profiles,dasyra:mass.ratio.conditions,dasyra:pg.qso.dynamics,
134: rj:profiles,rothberg.joseph:kinematics} as well as e.g.\ faint 
135: shells and tidal features common around apparently ``normal'' 
136: galaxies \citep{malin80,malin83,schweizer80,
137: schweizerseitzer92,schweizer96}, which are thought to be
138: signatures of galaxy collisions \citep[e.g.][]{hernquistquinn88,hernquist.spergel.92}. 
139: 
140: From both theoretical grounds 
141: \citep[][and references therein]{ostrikertremaine75,maller:sph.merger.rates,
142: fakhouri:halo.merger.rates,stewart:mw.minor.accretion} and 
143: observations \citep[e.g.][]{lin:merger.fraction,barton:triggered.sf,
144: woods:tidal.triggering,
145: woods:minor.mergers} it appears that ``minor'' mergers 
146: of mass ratios $\lesssim 1:10$ are ubiquitous (there are 
147: almost no galaxies without mergers of at least this mass ratio in the 
148: last few Gyr), and moreover a large fraction ($\sim1/2$) of the $\sim L_{\ast}$ 
149: galaxy population is observed and expected 
150: to have experienced a ``major'' merger 
151: (mass ratio $\lesssim1:3$) since $z\sim2-3$ 
152: \citep{lotz:merger.fraction,bell:merger.fraction,
153: bridge:merger.fractions,lin:mergers.by.type,kartaltepe:pair.fractions}. 
154: With increasing redshift, kinematic and morphological 
155: indications of recent, violent disturbance in disk-dominated galaxies 
156: appear more frequent \citep{hammer:obs.disks.w.mergers,
157: flores:tf.evolution,puech:highz.vsigma.disks,puech:tf.evol}.
158: 
159: 
160: Far from there not being enough mergers to explain the abundance 
161: of bulges and ellipticals, this has led to the 
162: concern that there may be far too {\em many} mergers to 
163: explain the survival and abundance of galactic disks in the context 
164: of our present understanding of galaxy formation. 
165: \citet{toomre72} were among the first to point out that mergers 
166: are capable of dramatically altering the morphologies of disks, 
167: transforming them into elliptical galaxies. 
168: Although their neglect of the importance of dissipational star formation 
169: and gas dynamics in the mergers led to some 
170: controversy \citep[e.g.][]{ostriker80,carlberg:phase.space,gunn87,kormendy:dissipation}, 
171: it is now increasingly well-established that 
172: major mergers between spiral galaxies (similar to those observed 
173: locally and at $z\lesssim2-3$) with gas fractions comparable to those 
174: observed yields remnants in good agreement with 
175: essentially all observed properties of low and intermediate-mass 
176: local elliptical galaxies \citep[e.g.\ morphologies, shapes, sizes, kinematics, densities, colors, 
177: black hole properties, fundamental scaling relations, 
178: stellar populations, and halo gas;][]{hernquist.89,
179: barnes.hernquist.91,barneshernquist96,
180: hernquist:phasespace,mihos:gradients,mihos:starbursts.96,
181: dimatteo:msigma,naab:gas,jesseit:kinematics,
182: cox:xray.gas,cox:kinematics,robertson:fp,springel:red.galaxies,burkert:anisotropy,
183: hopkins:clustering,
184: hopkins:cusps.ell,hopkins:cores,hopkins:cusps.fp,hopkins:groups.ell,hopkins:cusps.mergers}. 
185: 
186: Many
187: intermediate and low-luminosity ``cusp'' 
188: ellipticals (encompassing $\sim80-90\%$ of the 
189: mass density in ellipticals) contain significant embedded disks
190: \citep[perhaps all such ellipticals, given projection effects; see][]{ferrarese:type12,lauer:centers}, 
191: and they form a continuous sequence with most S0 galaxies, known to 
192: have prominent stellar (and even gaseous) disks \citep{kormendy:spheroidal1,
193: bender:ell.kinematics,ferrarese:type12,kormendy94:review,lauer:95,
194: faber:ell.centers,kormendy99,ferrarese:profiles,emsellem:sauron.rotation}. 
195: Indeed, the existence of embedded disks in simulated merger remnants is critical 
196: to matching the properties described above. 
197: 
198: A wide variety of observations including stellar populations and star formation histories 
199: \citep[e.g.][]{bender89,trager:ages,
200: mcdermid:sauron.profiles} and kinematic and structural analysis of recent 
201: merger remnants \citep{schweizer83,schweizer83:review,schweizerseitzer92,
202: schweizer:ngc34.disk,hibbard.yun:excess.light,rj:profiles} demonstrate that 
203: most of these disks are not accreted in the standard cosmological 
204: fashion after the spheroid 
205: forms -- they must somehow survive the merger or form very quickly thereafter from 
206: gas already in and around the galaxies. Therefore, despite the destruction of a 
207: large portion of a stellar disk in major mergers, {\em some} disk must survive mergers, 
208: and the amount that does so is a critical component determining 
209: many of the photometric and kinematic properties of even bulge-dominated 
210: and elliptical galaxies. 
211: 
212: Moreover, ``minor'' mergers -- at least those with mass ratios $\lesssim10:1$ (below 
213: which the difference between ``merger'' and accretion becomes increasingly blurred) -- 
214: are not generally believed to entirely destroy disks, but they 
215: are almost an order of magnitude more frequent than major 
216: mergers and as such may pose a 
217: more severe a problem for disk survival. 
218: In the $\Lambda$CDM cosmology, 
219: and from observed satellite fractions, 
220: it is unlikely than any disk (let alone a large fraction of disk galaxies) 
221: with a significant stellar age 
222: has survived $\sim5-10$\,Gyr without 
223: experiencing a merger of mass ratio $10:1$ or larger. 
224: Simulations \citep{quinn.84,quinn86:dynfric.on.sats,quinn93.minor.mergers,
225: hernquist.mihos:minor.mergers,
226: walker:disk.fragility.minor.merger,velazquezwhite:disk.heating,
227: naab:minor.mergers,
228: bournaud:minor.mergers,younger:antitruncated.disks,younger:minor.mergers} 
229: and analytic 
230: arguments \citep{ostrikertremaine75,tothostriker:disk.heating,
231: sellwood:resonant.disk.thickening} 
232: suggest that gas-poor minor mergers can convert a considerable 
233: fraction of a stellar disk into bulge and cause significant 
234: perturbation (``puffing up'' via dynamical heating) to the disk. 
235: The observed coldness of galactic disks suggests that 
236: this may be a severe problem: 
237: \citet{tothostriker:disk.heating} argued that large disks such as 
238: that in the Milky Way could not have undergone a merger of 
239: mass ratio $\lesssim10:1$ in the last $\sim10\,$Gyr. More 
240: recently e.g.\ \citet{stewart:mw.minor.accretion} and 
241: \citet{hammer:mw.no.mergers} emphasized that the tension
242: between these constraints and the expectation in CDM models that a
243: number of such mergers should occur implies either a deficit in our
244: understanding of hierarchical disk formation or a challenge to the
245: concordance cosmological model.
246: 
247: Given the successes of the $\Lambda$CDM model on large scales, 
248: and the increasing observational confirmation that disks do undergo 
249: (and therefore must somehow survive) 
250: a large number of mergers, it is likely that the problem lies in 
251: our (still relatively poor) understanding of disk galaxy formation. 
252: This has led to a great deal of focus on the problem of forming realistic 
253: disks in a cosmological context, with many different attempts and 
254: debate on the missing elements necessary to produce disks in 
255: simulations. Various groups have argued 
256: that self-consistent treatment of gas physics and star formation 
257: along with implementation of 
258: feedback of different kinds is necessary, along with greatly improved 
259: numerical resolution \citep{weil98:cooling.suppression.key.to.disks,
260: sommerlarsen99:disk.sne.fb,sommerlarsen03:disk.sne.fb,
261: thackercouchman00,thackercouchman01,abadi03:disk.structure,
262: governato04:resolution.fx,governato:disk.formation,
263: robertson:cosmological.disk.formation,
264: okamoto:feedback.vs.disk.morphology,scannapieco:fb.disk.sims}, 
265: in order to enable disks to survive 
266: their expected violent merger histories without completely 
267: losing angular momentum and transforming into systems that 
268: are too compact and have too much bulge mass (relative 
269: to real observed disks) by $z=0$. 
270: 
271: It has been known for some time \citep[see e.g.][]{hernquist:kinematic.subsystems,
272: barneshernquist96} that (even 
273: without any feedback) some
274: fraction of the gas in even a major merger of two disks 
275: can survive and form new, embedded disks in the remnant -- 
276: i.e.\ despite the problems outlined above, disks are not necessarily 
277: completely destroyed in mergers. However, 
278: early studies of this were restricted to cases with low gas 
279: content ($f_{\rm gas}\lesssim10\%$ in the progenitor disks), most of which was 
280: rapidly consumed in star formation, yielding small remnant disks in 
281: strongly bulge-dominated remnants. 
282: In seminal work, \citet{springel:spiral.in.merger} and 
283: \citet{robertson:disk.formation} showed that, in idealized merger simulations 
284: with significant stellar feedback to allow 
285: the stable evolution of extremely gas rich disks ($f_{\rm gas}\sim1$), 
286: even a major merger can produce a 
287: disk-dominated remnant. 
288: This has since been confirmed in fully cosmological simulations 
289: \citep{governato:disk.formation}. 
290: Together with other recent investigations (see references above), these works have 
291: led to the growing consensus that a combination of strong stellar feedback and 
292: large gas content is essential to the survival of disk galaxies. 
293: 
294: A large number of open questions remain, 
295: however. How, exactly, does feedback 
296: allow disks to survive mergers? What are the most important physics? 
297: Does it require fine-tuning of feedback prescriptions? How might 
298: things vary as a function of galaxy mass, redshift, gas content, 
299: merger orbits, and environment? Fundamentally, should this be expected 
300: for typical cosmological circumstances, or are these cases 
301: pathological? 
302: 
303: The ambiguity largely owes to the fact that there is no deep physical 
304: understanding of how disks survive or re-form after mergers and 
305: interactions. It has only just become possible to conduct simulations 
306: with the requisite large gas fractions, and thus far theoretical explanations 
307: have largely been restricted to phenomenological analysis, with continued 
308: efforts to improve resolution and sub-resolution prescriptions. 
309: Moreover, without a 
310: full model for how disks behave in interactions, these simulations 
311: cannot be placed into the broader context of the emergence 
312: of the entire Hubble sequence (for example asking the question, are 
313: the disks in lenticulars and embedded disks in ellipticals 
314: survivors of their pre-merger disks? Are they re-accreted? What determines 
315: how large they are? What is the key physics that gives rise to 
316: realistic embedded disks, leading to bulge-dominated galaxies 
317: with kinematic and photometric properties similar to those in the 
318: real universe?) 
319: or within a fully cosmological context.
320: 
321: The resolution requirements 
322: for full models of disk formation are severe -- limiting any 
323: attempt to properly simulate a cosmological box and still achieve 
324: the resolution necessary to reliably model a disk population -- 
325: and so models of the population of disks, largely semi-analytic, 
326: are forced to adopt simplified and un-tested prescriptions 
327: for the behavior of disks in mergers. This, in turn, has led to 
328: other well-known problems in modeling disk populations 
329: (even where prescriptions can ensure no artificial angular 
330: momentum losses); even when the 
331: cumulative (morphology-independent) galaxy mass function is 
332: correctly predicted at the low-mass end, semi-analytic models widely overproduce 
333: the relative abundance of low-mass spheroids and underproduce 
334: disks \citep[even when satellites, which have other associated 
335: model uncertainties, are removed from consideration; see e.g.][]{somerville:sam,
336: somerville:new.sam,croton:sam,
337: bower:sam,delucia:sam}. Lacking a proper, physically motivated understanding of 
338: how low-mass or gas-rich disks may or may not survive mergers, 
339: attempts to address this problem in the models have been 
340: purely phenomenological and involve arbitrary prescriptions 
341: \citep[see e.g.][]{koda:disk.survival.prescriptions}.
342: 
343: Motivated by these concerns, in this paper
344: we develop a physical, dynamical model for how 
345: disks survive and are destroyed in mergers and interactions. 
346: We show that, in an immediate (short-term) sense, the amount of stellar 
347: or gaseous disk that survives or re-forms 
348: following a given interaction can be understood purely 
349: in terms of simple, well-understood gravitational physics. 
350: Knowing these physics, we develop an analytic model that allows 
351: us to accurately predict how much of a given pre-merger 
352: stellar and cold gas disk will survive a merger, as a function of 
353: the merger mass ratio, orbital parameters, pre-merger cold gas 
354: fraction, and mass distribution of the gas and stars. 
355: We compare these predictions to the 
356: results of a large library of hundreds of hydrodynamic simulations of 
357: galaxy mergers and interactions, spanning a wide parameter space 
358: in these properties as well as prescriptions for gas physics, 
359: stellar and AGN feedback, halo and initial disk structural 
360: properties, redshift, and absolute galaxy masses. Our numerical experiments 
361: confirm that the analytic scalings accurately describe 
362: the behavior and bulge formation/disk destruction in mergers 
363: over the entire dynamic range surveyed, and 
364: confirm that the parameters not explicitly included in our 
365: model do not systematically affect either the mean predictions or 
366: the scatter of simulations about those predictions. This allows us to 
367: understand the mean behavior of systems with different orbits and 
368: mass ratios, as well as why systems with large gas fractions can 
369: form little bulge in even major mergers. 
370: 
371: This is 
372: possible because gas, in mergers, primarily loses angular momentum to 
373: internal gravitational torques (from the stars in the same disk) 
374: owing to asymmetries in 
375: the galaxy induced by the merger. Hydrodynamic torques and 
376: the direct torquing of the secondary are second-order effects, 
377: and very inefficient. Once gas is drained of angular momentum, 
378: there is little alternative but for it to fall to the center of the galaxy and 
379: form stars, regardless of the details of the prescriptions for star 
380: formation and feedback (these may change things at the 
381: $\sim10-20\%$ level by blowing out some of the gas, 
382: but they cannot fundamentally alter the 
383: fact that cold gas with no angular momentum will be largely 
384: unable to form any sort of disk, or the fact that a galaxy's worth of 
385: gas compressed to high densities and small radii 
386: will inevitably form a large mass in stars). 
387: But if the systems are sufficiently gas-rich, then there is little 
388: stellar material sharing the disk to torque on the gas in the interaction, and 
389: little or no angular momentum is lost. 
390: 
391: Feedback can dramatically alter the ability of a disk to survive in a 
392: cosmological sense: by allowing galaxies to retain large 
393: gas fractions (as opposed to no-feedback scenarios, in which cold gas 
394: in a disk is usually quickly converted into stars), they are 
395: more gas-rich when they undergo interactions, allowing them 
396: to avoid angular momentum loss for the reason above. Moreover, 
397: we show that in detail (owing to the resonant structure of 
398: interactions), it is really gas within a certain radius of the stellar disk 
399: that is drained of angular momentum. The commonly-invoked stellar wind 
400: feedback then enables cosmological disk survival in a second fashion: 
401: by redistributing gas out to large radii, it prevents angular momentum loss 
402: and allows rapid re-formation of disks after a merger. 
403: Independent of any tuning, our model allows us to quantify the 
404: disks expected as a function of interactions of arbitrary properties, 
405: and to physically, explicitly quantify what the requirements are for feedback, 
406: in a cosmological scenario, to enable disk survival. 
407: 
408: In \S~\ref{sec:sims} we describe our library of gas-rich merger simulations, 
409: which we use to test our physical model for disk destruction and 
410: survival. In \S~\ref{sec:id} we demonstrate the existence of genuine disks 
411: in remnants of even major mergers and 
412: briefly consider their properties, and compare methods to 
413: separate the disks and bulges in merger remnants. 
414: In \S~\ref{sec:form.major} we consider the question of how these disks 
415: form in and survive mergers: we identify the key components of any 
416: merger remnant in \S~\ref{sec:form.major:components}, highlighting 
417: that these disks originate from a combination of undestroyed pre-merger 
418: stellar disks and gas which avoids angular momentum loss 
419: in the merger. In \S~\ref{sec:form.major:angloss} we discuss how, in detail, 
420: that angular momentum loss proceeds. We use this, 
421: in \S~\ref{sec:model.overview}, to build a physical model for how 
422: angular momentum loss proceeds in mergers and predict the 
423: surviving disk content of merger remnants: we model and test how this 
424: depends on the gas content of the pre-merger disks (\S~\ref{sec:model.gas}), 
425: and the orbital parameters (\S~\ref{sec:model.orbit}) and 
426: mass ratio (\S~\ref{sec:model.massratio}) of the encounter. We generalize to 
427: first passage and fly-by encounters (\S~\ref{sec:model.flyby}) 
428: and demonstrate that (for otherwise fixed conditions at the time of an encounter) 
429: our conclusions are purely dynamical, independent of 
430: feedback physics or details in our 
431: treatment of e.g.\ star formation and the ISM gas physics (\S~\ref{sec:model.feedback}), 
432: although we use our model to determine exactly how these choices 
433: can have dramatic {\em indirect} consequences for disk survival (by 
434: altering the state of systems leading into a merger). 
435: We discuss some exceptions and pathological cases in \S~\ref{sec:model.exceptions}, 
436: and relate our results to the long-term secular evolution of 
437: barred systems in \S~\ref{sec:model.secular}. 
438: In \S~\ref{sec:prescriptions}, we outline how these results can and should 
439: be applied in analytic and semi-analytic models of galaxy formation, and 
440: give appropriate prescriptions derived from our numerical experiments. 
441: Finally, we summarize our results and discuss some of their cosmological implications 
442: and applications to other models and observations in \S~\ref{sec:discussion}. 
443: 
444: Throughout, we assume a $\Omega_{\rm M}=0.3$, $\Omega_{\Lambda}=0.7$,
445: $H_{0}=70\,{\rm km\,s^{-1}\,Mpc^{-1}}$ cosmology, but this has
446: little effect on our conclusions.
447: 
448: \breaker
449: \section{The Simulations}
450: \label{sec:sims}
451: 
452: 
453: Our simulations were performed with the parallel TreeSPH code {\small
454: GADGET-2} \citep{springel:gadget}, employing the fully conservative
455: formulation \citep{springel:entropy} of smoothed particle
456: hydrodynamics (SPH), which conserves energy and entropy simultaneously
457: even when smoothing lengths evolve adaptively \citep[see
458: e.g.,][]{hernquist:sph.cautions,oshea:sph.tests}.  Our simulations
459: account for radiative cooling and incorporate a sub-resolution
460: model of a multiphase interstellar medium (ISM) to describe star
461: formation and supernova feedback \citep{springel:multiphase}.
462: Feedback from supernovae is captured in this sub-resolution model
463: through an effective equation of state for star-forming gas, enabling
464: us to stably evolve disks with arbitrary gas fractions \citep[see,
465: e.g.][]{springel:models,
466: springel:spiral.in.merger,robertson:disk.formation,robertson:msigma.evolution}.
467: This is described by the parameter $\qeos$, which ranges from
468: $\qeos=0$ for an isothermal gas with effective temperature of $10^4$
469: K, to $\qeos=1$ for our full multiphase model with an effective
470: temperature $\sim10^5$ K. We have also compared with a subset of simulations
471: which adopt the star formation feedback prescription from
472: \citet{mihos:cusps,mihos:gradients,
473: mihos:starbursts.94,mihos:starbursts.96}, in which
474: the ISM is treated as a single-phase isothermal medium and feedback
475: energy is deposited as a kinetic impulse. We examine the 
476: effects of these choices in \S~\ref{sec:model.feedback}, and find they are minimal. 
477: 
478: Likewise, although they make little difference to the analysis here, 
479: supermassive black holes are usually included at the centers of both
480: progenitor galaxies.  These black holes are represented by ``sink''
481: particles that accrete gas at a rate $\Mdot$ estimated from the local
482: gas density and sound speed using an Eddington-limited prescription
483: based on Bondi-Hoyle-Lyttleton accretion theory.  The bolometric
484: luminosity of the black hole is taken to be $L_{\rm
485: bol}=\epsilon_{r}\dot{M}\,c^{2}$, where $\epsilon_r=0.1$ is the
486: radiative efficiency. We assume that a small fraction (typically
487: $\approx 5\%$) of $L_{\rm bol}$ couples dynamically to the surrounding
488: gas, and that this feedback is injected into the gas as thermal
489: energy, weighted by the SPH smoothing kernel.  This fraction is a free
490: parameter, which we determine as in \citet{dimatteo:msigma} by
491: matching the observed $M_{\rm BH}-\sigma$ relation.  For now, we do
492: not resolve the small-scale dynamics of the gas in the immediate
493: vicinity of the black hole, but assume that the time-averaged
494: accretion rate can be estimated from the gas properties on the scale
495: of our spatial resolution (roughly $\approx 20$\,pc, in the best
496: cases). While the black holes can be indirectly important, owing to 
497: their feedback ejecting gas into the halo and thus preserving it from 
498: star formation until the final merger, we find that, for a given gas content 
499: at the time of the actual merger, our results are unchanged 
500: in a parallel suite of simulations without black holes. 
501: 
502: The progenitor galaxy models are described in
503: \citet{springel:models}, and we review their properties here.  For each
504: simulation, we generate two stable, isolated disk galaxies, each with
505: an extended dark matter halo with a \citet{hernquist:profile} profile,
506: motivated by cosmological simulations \citep{nfw:profile,busha:halomass}, 
507: an exponential disk of gas and stars, and (optionally) a
508: bulge.  The galaxies have total masses $M_{\rm vir}=V_{\rm
509: vir}^{3}/(10GH[z])$, with the baryonic disk having a mass
510: fraction $m_{\rm d}=0.041$, the bulge (when present) having $m_{\rm
511: b}=0.0136$, and the rest of the mass in dark matter.  The dark matter
512: halos are assigned a
513: concentration parameter scaled as in \citet{robertson:msigma.evolution} appropriately for the 
514: galaxy mass and redshift following \citet{bullock:concentrations}. We have also 
515: varied the concentration in a subset of simulations, and find it has little 
516: effect on our conclusions (because the central regions of the 
517: galaxy are, in any case, baryon-dominated), insofar as they pertain to 
518: disk survival in mergers (it has been demonstrated that halo concentrations 
519: are important for e.g.\ the exact sizes and velocity scalings of disks, and our 
520: predicted disk sizes scale accordingly). 
521: The initial disk scale-length is computed
522: based on an assumed spin parameter $\lambda=0.033$, chosen to be near
523: the mode in the $\lambda$ distribution measured in simulations \citep{vitvitska:spin},
524: and the scale-length of an initial bulge (when present) is set to $0.2$ times this.
525: 
526: Typically, each galaxy initially consists of 168000 dark matter halo
527: particles, 8000 bulge particles (when present), 40000 gas and 40000
528: stellar disk particles, and one BH particle.  We vary the numerical
529: resolution, with many simulations using twice, and a subset up to 128
530: times, as many particles. We choose the initial seed
531: mass of the black hole either in accord with the observed $M_{\rm
532: BH}$-$\sigma$ relation or to be sufficiently small that its presence
533: will not have an immediate dynamical effect, but we have varied the seed
534: mass to identify any systematic dependences.  Given the particle
535: numbers employed, the dark matter, gas, and star particles are all of
536: roughly equal mass, and central cusps in the dark matter and bulge
537: are reasonably well resolved.
538: 
539: We consider a series of several hundred simulations of colliding
540: galaxies, described in detail in 
541: \citet{robertson:fp,robertson:msigma.evolution} and
542: \citet{cox:xray.gas,cox:kinematics}.  We vary the numerical resolution, the orbit of the
543: encounter (disk inclinations, pericenter separation), the masses and
544: structural properties of the merging galaxies, presence or absence 
545: of bulges in the progenitor galaxies, initial gas fractions,
546: halo concentrations, the parameters describing star formation and
547: feedback from supernovae and black hole growth, and initial black hole
548: masses. 
549: 
550: The progenitor galaxies have virial velocities $V_{\rm vir}=55, 80, 113, 160,
551: 226, 320,$ and $500\,{\rm km\,s^{-1}}$, and redshifts $z=0, 2, 3, {\rm
552: and}\ 6$, and the simulations span a range in final spheroid mass
553: $\mbh\sim10^{8}-10^{13}\,M_{\sun}$, covering essentially the
554: entire range of the observations we consider at all redshifts, and
555: allowing us to identify any systematic dependences in our models.  We
556: consider initial disk gas fractions (by mass) of $\fgas = 0.05,\ 0.1,\ 0.2,\ 0.4,\ 0.6,\ 
557: 0.8,\ {\rm and}\ 1.0$ for several choices of virial velocities,
558: redshifts, and ISM equations of state. 
559: The results described in this
560: paper are based primarily on simulations of equal-mass mergers;
561: however, we examine in \S~\ref{sec:model.massratio} how our results scale with 
562: mass ratio in mixed encounters, down to mass ratios $\sim1:10$ or so, 
563: below which (as we show) the encounters have little noticeable 
564: effect. In detail, the simulations studied there are 
565: described in \citet{younger:minor.mergers} and constitute a complete 
566: subset of permutations of our standard galaxy models with mass ratios 
567: uniformly sampling the range $1:1$ to $1:8$. As in 
568: our larger set of $1:1$ mergers, at each mass ratio we 
569: systematically survey the effects of different absolute galaxy mass, 
570: orbital parameters, and disk gas fraction (resulting 
571: in a typical $\sim30-40$ simulations spanning the full range of 
572: orbital parameters and gas fractions of interest, around 
573: each mass ratio $\sim 1:1,\ 1:2,\ 1:4,$ and $1:8$). 
574: We have considered 
575: more limited studies of minor mergers where we vary 
576: e.g.\ the ISM equation of state, redshift, initial disk 
577: structural properties; as we find in our studies of these 
578: parameters in the larger suite of equal-mass mergers, 
579: they make no significant difference to our conclusions here. 
580: 
581: 
582: \begin{\tableset}{lccccl}
583: \tabletypesize{\scriptsize}
584: \tablecaption{Disk Orientations\label{tbl:orbits}}
585: \tablewidth{0pt}
586: \tablehead{
587: \colhead{Name} &
588: \colhead{$\theta_{1}$} &
589: \colhead{$\phi_{1}$} &
590: \colhead{$\theta_{2}$} &
591: \colhead{$\phi_{2}$} &
592: \colhead{Comments} 
593: }
594: \startdata
595: {\bf b} & 180 & 0 & 0 & 0 & prograde-retrograde \\
596: {\bf c} & 180 & 0 & 180 & 0 & both retrograde \\
597: {\bf d} & 90 & 0 & 0 & 0 & polar-prograde \\
598: {\bf e} & 30 & 60 & -30 & 45 & ``random'' (prograde) \\
599: {\bf f} & 60 & 60 & 150 & 0 & tilted polar-retrograde \\
600: {\bf g} & 150 & 0 & -30 & 45 & retrograde-``random'' \\
601: {\bf h} & 0 & 0 & 0 & 0 & both prograde \\
602: \\
603: \hline \\
604: {\bf i} & 0 & 0 & 71 & 30 & Barnes orientations \\
605: {\bf j} & -109 & 90 & 71 & 90 &   \\
606: {\bf k} & -109 & -30 & 71 & -30 &   \\
607: {\bf l} & -109 & 30 & 180 & 0 &   \\
608: {\bf m} & 0 & 0 & 71 & 90 &   \\
609: {\bf n} & -109 & -30 & 71 & 30 &   \\
610: {\bf o} & -109 & 30 & 71 & -30 &   \\
611: {\bf p} & -109 & 90 & 180 & 0 &   \\
612: \\
613: \hline \\
614: {\bf m000} & 0 & 0 & -30 & 45 & Minor merger orientations \\
615: {\bf m030} & 30 & 0 & -30 & 45 &  \\
616: {\bf m090} & 90 & 0 & -30 & 45 &  \\
617: {\bf m150} & 150 & 0 & -30 & 45 &  \\
618: {\bf m180} & 180 & 0 & -30 & 45 &  \\
619: \enddata
620: \tablenotetext{\,}{List of disk galaxy orientations for major merger 
621: simulations. Columns show: (1) the orbit identification (used to 
622: refer to each orbit throughout); (2-3) the initial orientation 
623: of disk 1 (in standard spherical coordinates); (4-5) the initial 
624: orientation of disk 2; and (6) a brief description of some of the 
625: orientations.}
626: \end{\tableset}
627: 
628: 
629: 
630: Once built, pairs of galaxies are placed on parabolic orbits 
631: \citep[motivated by cosmological simulations; see e.g.][]{benson:cosmo.orbits,
632: khochfar:cosmo.orbits} with 
633: the spin axis of each disk specified by the angles $\theta$ and $\phi$ 
634: in standard spherical coordinates. Table~\ref{tbl:orbits} lists the 
635: orientations in different representative orbits we have sampled. 
636: The particular choice of orbits follows \citet{cox:kinematics}; 
637: there are seven idealized mergers (cases {\bf b}-{\bf h}) 
638: that represent orientations often seen in the literature 
639: (for example, case {\bf h}, where all the angular momentum 
640: vectors of the disks and orbit are initially aligned), the 
641: rest ({\bf i}-{\bf p}) follow \citet{barnes:disk.halo.mergers} 
642: by selecting unbiased initial disk orientations according 
643: to the coordinates of two oppositely directed tetrahedrons. 
644: These orbits are identical to those considered in various other studies, 
645: such as \citet{naab:minor.mergers}. 
646: For our series of orbits of various mass ratios from \citet{younger:minor.mergers}
647: (where, in minor 
648: mergers, the inclination of the secondary is less important than 
649: that of the primary) we survey the inclination of the primary 
650: in a systematic sense, considering all our mergers 
651: with $\theta_{1}=0,\,30,\,90,\,150,\,180\,\degree$ (cases {\bf m000-m180}). 
652: We examine the effect of 
653: orbits in detail in \S~\ref{sec:model.orbit}, and find that for random orbits, 
654: the differences are quantifiable but not strong -- pathological orbits 
655: (such as the aligned case {\bf h} above) are discussed in 
656: \S~\ref{sec:model.exceptions} (these pathological 
657: cases often, in fact, are the most efficient at destroying disks). 
658: We have also tested our predicted scalings with limited subsets of 
659: simulations that vary the pericentric passage distance and the 
660: energy of the orbit, described in \citet{robertson:disk.formation}
661: and \citet{cox:kinematics}, and find that our estimates are 
662: robust to these variations. 
663: 
664: Each simulation is evolved until the merger is complete and the remnants are 
665: fully relaxed, typically $\sim1-2$\,Gyr after the final merger 
666: and coalescence of the BHs. We then analyze the 
667: remnants following \citet{cox:kinematics}, in a manner designed to mirror 
668: the methods typically used by observers. For each remnant we project the 
669: stars onto a plane as if observed from a particular direction (we consider 
670: 100 viewing angles to each remnant, which uniformly sample the unit sphere). 
671: When we plot projected quantities such as $\re$, $\sigma$, and $V_{c}$, we 
672: typically show just 
673: the median value for each simulation across all $\sim100$ viewing directions.
674: The sightline-to-sightline 
675: variation in these quantities is typically smaller than the 
676: simulation-to-simulation scatter, but we explicitly note where it is large. 
677: 
678: 
679: 
680: \breaker
681: \section{The Existence of Disks in Major Merger Remnants}
682: \label{sec:id}
683: 
684: \citet{robertson:disk.formation} and \citet{springel:spiral.in.merger}, and 
685: subsequently \citet{governato:disk.formation} and \citet{naab:gas} 
686: have demonstrated that even major mergers can leave remnants with 
687: non-negligible disk components. Nevertheless, we wish to highlight 
688: several properties of these disks first, to establish their existence and 
689: nature. Moreover, 
690: we wish to ensure that we can 
691: robustly identify disks in our merger remnants, 
692: before going on to analyze the conditions for their survival. 
693: In order to do this, we have considered several methods, 
694: include e.g.\ fitting the surface brightness profiles to 
695: traditional bulge-disk decompositions, 
696: \citep[see e.g.][]{robertson:disk.formation,hopkins:cusps.mergers}, 
697: kinematic decompositions based on 
698: one and two-dimensional velocity maps 
699: \citep{cox:kinematics,hoffman:prep}, and 
700: three dimensional component fitting. 
701: These ultimately give similar results, although 
702: e.g.\ surface brightness profile fits and velocity decompositions 
703: can be considerably dependent on the viewing angle 
704: and are not especially robust at separating a small 
705: disk (in a bulge-dominated system) 
706: from e.g.\ other kinematic subcomponents 
707: or rotating bulges \citep[a well-known observational difficulty, 
708: see e.g.][and references therein]{balcells:bulge.xl,jk:profiles,
709: marinova:bar.frac.vs.freq,barazza:bar.colors}. 
710: 
711: We therefore choose to take advantage of our full three-dimensional 
712: information in the simulations to easily decompose bulges and 
713: disks in a simple, automated fashion. 
714: For convenience, let us consider the remnant in 
715: cylindrical coordinates ${\bf x}=(R,\,\phi)$ where the 
716: axis of symmetry ($\hat{z}$) is defined by the net angular momentum 
717: vector of the baryonic mass in the relaxed remnant. 
718: The effective rotational support of 
719: any given stellar or gas particle in the simulation is then 
720: \begin{equation}
721: \tilde{v}_{\rm rot} = \frac{v_{\phi}}{v_{c}(r)}
722: \end{equation}
723: where $v_{c}$ is the circular velocity 
724: \begin{equation}
725: v_{c} = \sqrt{\frac{G\,M_{\rm enc}(r)}{r}}
726: \end{equation}
727: (here $r$ is the {\em three dimensional} radius from the galaxy center). 
728: If we consider the distribution of 
729: baryonic mass in $\tilde{v}_{\rm rot}$, we find a clear 
730: segregation between bulge and disk components. 
731: 
732: \begin{figure*}
733:     \centering
734:     \scaleup
735:     %\plotone{plot_all_disk.ps}
736:     \plotone{f1.ps}
737:     \caption{Examples of merger remnants with large disks. 
738:     {\em Top:} Edge-on projected stellar surface brightness 
739:     of the galaxy. {\em Middle:} the distribution of all stars in 
740:     their rotational support, $v_{\rm rot}/v_{c}(r)$, where 
741:     $v_{c}(r)$ is the circular velocity at $r$ and $v_{\rm rot}=v_{\phi}$ 
742:     in cylindrical coordinates is the rotational velocity about the 
743:     net stellar angular momentum axis. We decompose the clearly 
744:     bimodal distributions
745:     into bulge (orange, peak near $v_{\rm rot}/v_{c}(r)\sim0$) 
746:     and disk (blue, peak near $v_{\rm rot}/v_{c}(r)\sim1$) components, 
747:     with stellar mass fraction in the disk component ($f_{\rm disk}$) 
748:     labeled. {\em Bottom:} Azimuthally averaged face-on surface brightness 
749:     profile. We show the total profile (black) and profile of each of the 
750:     components separated by their rotational support (orange and blue respectively). 
751:     We fit the total profile to a standard bulge+disk decomposition, 
752:     and show the resulting fitted bulge (red dotted) and disk (blue dotted) 
753:     components and disk mass fraction. 
754:     The two methods recover similar decompositions in almost all cases: the ``disks'' 
755:     are rotationally supported with extended exponential profiles, 
756:     the ``bulges'' are dispersion supported with compact Sersic-law profiles. 
757:     We show three example remnants typical of our simulations: 
758:     {\em Left:} Equal mass (mass ratio $1:1$) merger remnant 
759:     with $\sim40\%$ gas at the time of merger. The remnant is 
760:     a bulge dominated elliptical/lenticular, but has a prominent 
761:     smooth stellar disk with $\sim30-40\%$ of the mass. 
762:     {\em Center:} Major (mass ratio $1:2$) merger remnant with 
763:     $\sim20\%$ gas at the time of merger. The remnant is a 
764:     marginally disk-dominated S0a-type galaxy, with some spiral structure in the disk. 
765:     {\em Right:} Minor (mass ratio $1:8$) merger remnant with 
766:     $\sim15\%$ gas at the time of merger. The remnant is a 
767:     Sb/Sc disk with a flattened, compact bulge. 
768:     \label{fig:jz.distrib}}
769: \end{figure*}
770: 
771: Figure~\ref{fig:jz.distrib} shows this for three simulations with 
772: large disks in the remnant. There is clearly a bimodal 
773: distribution in $\tilde{v}_{\rm rot}$, with one component having 
774: relatively little rotation (the bulge, with a peak near 
775: $\tilde{v}_{\rm rot}\approx0$), and 
776: one component being largely rotationally 
777: supported (the disk, with a peak near $\tilde{v}_{\rm rot}\sim1$). 
778: There are two stellar populations in these remnants, with a 
779: clean division in their rotational support. 
780: 
781: We can, from this plot alone, estimate a robust disk-bulge 
782: mass ratio, from fitting e.g.\ the sum of two Gaussian components 
783: (disk and bulge) 
784: to this distribution (or in a non-parametric sense, by 
785: assuming the bulge component has a symmetric 
786: rotation distribution about its peak, and mirroring 
787: the distribution about that, taking what remains to be disk). 
788: Our results are not sensitive to the exact details of our decomposition, 
789: but we experiment with a few different methods 
790: in order to estimate uncertainties on the bulge-disk decomposition 
791: which we refer to below. Again, we have repeated our entire 
792: analysis using alternative estimators of the disk-to-bulge mass 
793: (direct profile fits and velocity profile decompositions), and 
794: find that the same scalings apply in all cases (the uncertainties in 
795: the decomposition of a given simulation do, however, increase). 
796: 
797: \begin{figure*}
798:     \centering
799:     %\scaleup
800:     %\plotone{check_SbSc.ps}
801:     \plotone{f2.ps}
802:     \caption{Bulge ({\em left}), stellar disk ({\em middle}), and 
803:     gas ({\em right}), in the remnant of a $1:2$ mass-ratio major merger 
804:     with $\sim20\%$ gas at the time of the merger, on a 
805:     typical random orbit (center panel in Figure~\ref{fig:jz.distrib}). 
806:     From top to bottom, panels show: 
807:     (a) Projected surface density (edge-on view). 
808:     (b) Face-on view. (c) Mean edge-on scale height 
809:     $H/R$ of the component as a function of circular radius $R=|x|$ (for 
810:     this projection). 
811:     (d) Edge-on velocity profile: mean velocity $v(r)$ (blue) and velocity 
812:     dispersion $\sigma(r)$ (red). (e) Rotational 
813:     support measure $v/\sigma$. 
814:     \label{fig:view.SbSc}}
815: \end{figure*}
816: 
817: Figure~\ref{fig:view.SbSc} shows one simulation from 
818: Figure~\ref{fig:jz.distrib} (S0 major-merger remnant), using this method to 
819: decompose the remnant into a stellar bulge and stellar disk. 
820: We also show the gas separately, which can cool and therefore forms an 
821: extremely thin disk. 
822: The properties are exactly what would be expected for a typical 
823: bulge-disk system: the ``bulge'' is a somewhat flattened 
824: ellipse with ellipticity $\epsilon\approx0.3-0.4$ ($H/R\approx0.6-0.7$), 
825: and is a pressure-supported system, with one-dimensional 
826: velocity dispersion $\sigma\sim120-150\,{\rm km\,s^{-1}}$ (depending 
827: on the sightline and slit width) and a 
828: rotation velocity $\sim30-50\,{\rm km\,s^{-1}}$. The resulting rotation parameter 
829: of the bulge itself 
830: ($(V/\sigma)^{\ast}\sim0.4$) is typical of reasonably rapidly rotating 
831: bulges. It is also compact (as expected), with projected $R_{e}\sim 1-2$\,kpc. 
832: The stellar disk is like that of a combined thin-thick disk 
833: system, with $H/R\sim0.15-0.2$, and exhibits a flat rotation 
834: curve with $V_{\rm max}\sim200\,{\rm km\,s^{-1}}$. 
835: The disk is rotationally supported with typical values for a 
836: disk of similar mass and overall morphology, $V/\sigma\sim2-3$, 
837: and it is far more extended than the bulge ($R_{e}\sim10\,$kpc, 
838: putting it on the observed disk size-mass relation shown below). 
839: The properties of the gas disk are similar, with the obvious exception 
840: that, since the gas can cool, it forms a very thin disk 
841: ($H/R\lesssim0.05$). 
842: 
843: 
844: \begin{figure}
845:     \centering
846:     %\scaleup
847:     %\plotone{check_SbIm.ps}
848:     \plotone{f3.ps}
849:     \caption{Bulge, disk, and gas, shown as in Figure~\ref{fig:view.SbSc}, 
850:     for the remnant of a $1:8$ mass-ratio minor merger with 
851:     $\sim15\%$ gas, on a polar orbit. 
852:     \label{fig:view.SbIm}}
853: \end{figure}
854: 
855: Figure~\ref{fig:view.SbIm} shows the components of 
856: another galaxy (Sbc minor-merger 
857: remnant) from Figure~\ref{fig:jz.distrib}, 
858: in the manner of Figure~\ref{fig:view.SbSc}. As expected given 
859: the smaller bulge-to-disk ratio in this case, the 
860: system is more flattened, with even larger rotational support 
861: ($V/\sigma\sim5-10$ and $H/R\lesssim0.1$ even 
862: in the stellar disk). The bulge is clearly a distinct dispersion-supported 
863: component, despite being relatively flattened. 
864: 
865: \begin{figure}
866:     \centering
867:     %\scaleup
868:     %\plotone{check_b3f.ps}
869:     \plotone{f4.ps}
870:     \caption{Bulge, disk, and gas, shown as in Figure~\ref{fig:view.SbSc}, 
871:     for the remnant of a $1:1$ mass-ratio major merger with 
872:     $\sim40\%$ gas, on an inclined polar orbit. 
873:     \label{fig:view.b3f}}
874: \end{figure}
875: 
876: Figure~\ref{fig:view.b3f} shows the components of 
877: the third galaxy (elliptical major-merger 
878: remnant) from Figure~\ref{fig:jz.distrib}, 
879: in the manner of Figure~\ref{fig:view.SbSc}. 
880: We show this case to demonstrate that embedded disks can be recovered 
881: reliably, and are indeed real, rotation supported ($V/\sigma\sim3-5$), 
882: and relatively thin ($H/R\lesssim0.2$ in the stellar disk, $\lesssim0.1$ in the gaseous disk)
883: kinematic objects even in simulations where they are not 
884: a majority of the mass (here, we find $B/T\sim0.7$). 
885: 
886: 
887: \begin{figure}
888:     \centering
889:     \scaleup
890:     %\plotone{disk_tf.ps}
891:     \plotone{f5.ps}
892:     \caption{Simulated disk-dominated merger remnants on the observed baryonic 
893:     Tully-Fisher relation ({\em top}) and size-stellar mass relation ({\em bottom}). 
894:     We take $V_{\rm rot}$ from the rotation curves as in Figure~\ref{fig:view.SbSc} 
895:     and $R_{e}$ is the median projected half-mass radius. 
896:     We compare with the observed relations as a function of morphology 
897:     from \citet{courteau:disk.scalings} (solid lines in each panel; dotted lines 
898:     show the observed $\pm1\,\sigma$ scatter), for 
899:     S0-Sa ({\em left}; red), Sab-Sbc ({\em center}; green), and Sc-Sd ({\em right}; 
900:     blue and black, respectively) galaxies. For convenience, we assign our 
901:     simulations a ``morphology'' based on the bulge-to-disk ratio as labeled. 
902:     Most of our simulations are major (mass ratio $1:1$) mergers that yield significant, 
903:     but not dominant disks (and are therefore not shown here). Very late types 
904:     are only produced in our limited subset of small mass ratio ($\gtrsim1:8$) 
905:     mergers. In any case, the remnant disks all lie on the observed 
906:     Tully-Fisher and size-mass relations appropriate for their morphology -- 
907:     coupled with their rotation and scale heights, we can say they are real disks 
908:     in the observable sense. 
909:     \label{fig:TF}}
910: \end{figure}
911: 
912: As a further check that these are indeed real disks, 
913: Figure~\ref{fig:TF} plots the disks in disk-dominated simulation remnants on the 
914: baryonic Tully-Fisher and stellar size-mass relations observed 
915: for disks of similar morphology
916: \citep[see e.g.][]{belldejong:tf,mcgaugh:tf,shen:size.mass,courteau:disk.scalings}. 
917: For convenience, we will use the 
918: estimated bulge-to-disk ratio as a proxy for morphology throughout, 
919: with the values as labeled in Figure~\ref{fig:TF}. We take their 
920: velocities here from the projected disk rotation curves where they are flat, and 
921: take $R_{e}$ as the projected half-mass radius (note that this is different from the 
922: exponential disk scale length $h$; for a pure exponential disk 
923: $R_{e}=1.678\,h$, and we convert the observations where necessary accordingly). 
924: 
925: For each morphological class, our simulations agree well with the observed 
926: Tully-Fisher and size-mass relations. Given our limited sampling of very minor 
927: mergers with mass ratios $\gtrsim1:8$, we have only a few simulations with 
928: final $B/T<0.2$, but those nevertheless agree (as expected, since they have only been 
929: slightly modified from the original disk). We stress that we are {\em not} 
930: claiming to reproduce the Tully-Fisher or stellar size-mass relation of disks 
931: in an {\em a priori} manner: our (pre-merger) disks are constructed, by design, to 
932: more or less lie on the observed correlations. What we are saying is that, 
933: given progenitor disks that are similar to those observed, disks 
934: that form after or survive mergers (even 
935: major mergers) will remain on the appropriate correlations for their 
936: stellar mass and morphology. In short, when disks do survive mergers, 
937: they are ``real'' disks in the observable sense, not highly flattened bulges 
938: or unusual kinematic subcomponents. 
939: 
940: 
941: \breaker
942: \section{Disk Formation in Major Mergers}
943: \label{sec:form.major}
944: 
945: Clearly, even major mergers can and do produce remnants with significant disks. 
946: We therefore ask how these disks form, and whether we can derive some 
947: analytic expectation for their masses as a function of progenitor and 
948: merger properties. 
949: 
950: 
951: \subsection{Components of the Remnant: Surviving Gas Disks}
952: \label{sec:form.major:components}
953: 
954: 
955: For simplicity, let us begin with the case of an identical $1:1$ mass ratio 
956: merger (we will generalize to arbitrary mass ratios in \S~\ref{sec:model.massratio} below). 
957: Early in the merger, the galaxies experience a first passage and begin to 
958: lose angular momentum to the halos, rapidly coalescing on a timescale of 
959: order a couple orbital periods. In the final merger and coalescence of the 
960: galaxies, the stars which are initially ``cold'' (i.e.\ pre-merger disks) will 
961: scatter and violently relax \citep{lynden-bell67}, 
962: forming a \citet{devaucouleurs}-like quasi-spherical, 
963: dispersion supported profile. Of the gas available at the time of the final 
964: merger, some will lose its angular momentum, fall into the galaxy center, 
965: and (given the sudden rapid increase in density) rapidly transform into 
966: stars in a central starburst, forming a compact, central dissipational component of the 
967: remnant bulge \citep[for a detailed study of this 
968: component, see][]{hopkins:cusps.ell,hopkins:cusps.mergers}.
969: Gas that is at sufficiently large radii that it cannot efficiently 
970: fall in, or gas that for whatever reason cannot efficiently dissipate or lose 
971: angular momentum, will rapidly see the central potential relax (the equilibration 
972: timescale of the central bulge is only $\sim10^{8}\,$yr) and, 
973: having conserved its angular momentum, will rapidly cool and re-form 
974: a thin, rotationally supported disk. \citet{barneshernquist96} outline this process and show, in detail, 
975: how the cooling gas that survives the merger rapidly settles into a 
976: typical, rotationally supported exponential disk. This will then form stars, which 
977: constitute a new stellar disk. 
978: 
979: We emphasize these three components: 
980: 
981: {\bf Pre-Merger Stars:} These (along with the dark matter) constitute the 
982: collisionless (dissipationless) component of the merger. Because they are 
983: collisionless, the stars and dark matter distributions mix in the merger. A 
984: given star, as it moves through the merging galaxies on a random orbit, 
985: feels a rapidly fluctuating potential, which deflects its orbit and allows 
986: for the phase space distribution of the particles to uniformly mix. This violent 
987: relaxation process gives rise to a pressure supported system 
988: dominated by random velocities \citep{lynden-bell67} and transforms initially 
989: exponential disk into quasi-spherical \citet{devaucouleurs}-like
990: Sersic-law profiles. In the limit of a $1:1$ mass ratio merger, it is a 
991: good approximation to assume that all of the stars are violently relaxed -- the 
992: merger is sufficiently ``violent'' that no significant component of the 
993: pre-merger stellar disks will survive the merger. This is {\em not} 
994: necessarily true at lower mass ratios (see \S~\ref{sec:model.massratio}), but it simplifies our 
995: analysis to begin, while we consider such mergers. 
996: 
997: The remaining two components of the remnant can be identified 
998: with the gas supply available at the time of the final merger. 
999: 
1000: {\bf Starburst Stars:} This is the remnant of a dissipational starburst, triggered in the 
1001: merger. Some fraction of the gas will efficiently lose its angular momentum in the merger. 
1002: Because gas can dissipate energy, it will then necessarily rapidly fall into the center 
1003: of the merging system \citep[essentially free-falling to the center until the collapsing 
1004: gas becomes self-gravitating; see][]{hopkins:cusps.mergers}. 
1005: Collecting a large gas supply in the center, the result is a rapid, highly concentrated 
1006: starburst -- in gas rich cases, this is analogous to that observed in 
1007: e.g.\ nearby merging ULIRGs \citep{soifer84a,soifer84b,scoville86,sargent87,sargent89} 
1008: and recent merger remnants \citep{kormendysanders92,hibbard.yun:excess.light,rj:profiles}. 
1009: This builds up a dense, compact central stellar distribution, that raises the 
1010: central phase space density and yields an effectively smaller, more baryon-dominated 
1011: remnant. The starburst stars, being so concentrated (and typically 
1012: having a more mixed orbital distribution owing to the random velocities of infalling gas 
1013: in the starburst), are clearly part of the bulge (although they may have slightly 
1014: different Sersic profiles and kinematics from the more extended bulge formed from 
1015: violent relaxation of the pre-merger stars). This component is important for 
1016: the structure and scalings of the bulge/spheroid component, and we study it in detail in 
1017: \citet{hopkins:cusps.ell,hopkins:cores,hopkins:cusps.mergers}. It is essentially 
1018: the dissipational component of the merger. For our purposes here, however, 
1019: this is the gas ``lost,'' which becomes part of the bulge and no longer
1020: contributes to the remnant disk. 
1021: 
1022: {\bf Surviving Gas/``Post-Merger'' Stars:} The gas that does {\em not} lose its angular 
1023: momentum will, as described above, form a new disk as the remnant relaxes. 
1024: For a $1:1$ merger, since (as noted above) the entire stellar distribution is violently 
1025: relaxed, the post-merger disks can be entirely identified with gas that survives the merger. 
1026: It is not, in this case, so much that the initial disks survive the merger intact, as it is that 
1027: some of the gas remains at large radii/with significant angular momentum, which can 
1028: rapidly re-form the disk after the merger. 
1029: Essentially then (for major mergers), 
1030: the question of how much of a disk will remain post-merger 
1031: is a question of how much of the gas (at the time of the merger) 
1032: will or will not lose its angular momentum. 
1033: 
1034: 
1035: \subsection{How Does the Gas Lose Its Angular Momentum?}
1036: \label{sec:form.major:angloss}
1037: 
1038: How, then, does the gas lose angular momentum in a merger? 
1039: The basic process has been understood since early simulations 
1040: involving highly simplified models for gas dissipation in 
1041: \citet{noguchi:merger.induced.bars.dissipationless,noguchi:merger.induced.bars.gas.forcing}, 
1042: \citet{hernquist.89},
1043: and \citet{barnes.hernquist.91}. With improved numerical models,
1044: \citet{barneshernquist96} followed this process in detail, 
1045: and showed that what happens 
1046: in a typical major merger is as follows: 
1047: the non-axisymmetric perturbation (owing to the companion) in the system 
1048: induces (largely after first passage and on the final coalescence, since this is 
1049: where the interaction is significant) a
1050: non-axisymmetric response in
1051: the disk.\footnote{In what follows, we will refer to this
1052: non-axisymmetric response as a ``bar,'' for simplicity and 
1053: because morphologically the induced feature, at least for 
1054: some time during the merger resembles bars in isolated
1055: barred spirals.  However, we caution that the formation 
1056: mechanism which excites this response may be different
1057: from that causing bars in isolated galaxies.  Furthermore,
1058: while the non-axisymmetry is present throughout the merger,
1059: it at times would not be classified as a bar morphologically,
1060: particularly during the final coalescence of the galaxy
1061: nuclei, when the resulting gas inflows are strong.}
1062: A stellar bar and gas bar form, but because the gas is 
1063: collisional and the stars are collisionless, 
1064: the stellar bar will trail or lag behind the gas bar by a small offset 
1065: (typically $\sim$a few degrees). The stellar bar therefore torques the gas bar, 
1066: draining its angular momentum, and causing the gas to collapse to 
1067: the center. 
1068: 
1069: 
1070: 
1071: 
1072: \begin{figure*}
1073:     \centering
1074:     \scaleup
1075:     %\plotone{disk_survival_illustration.ps}
1076:     %\plotone{disk_survival_illustration_mod.ps}
1077:     \plotterr{f6.ps}
1078:     \caption{Illustration of the key processes that drive starbursts in a merger. 
1079:     {\em Top:} Projected gas density (as in Figure~\ref{fig:view.SbSc}) in 
1080:     the plane of the disk at representative times in a retrograde 1:3 merger 
1081:     (left to right: before interaction, just after first passage, just after second passage/coalescence, 
1082:     after relaxation). For clarity, just the gas from the primary is shown. 
1083:     {\em Middle:} Same, for a prograde encounter. 
1084:     The passage of the secondary induces a bar-like non-axisymmetric disturbance in the primary, 
1085:     which survives after the short-lived passage and removes angular momentum from the 
1086:     gas, leading to a starburst. The same process occurs on both passages, with a larger 
1087:     (albeit less ``bar-like'') asymmetry on coalescence. The prograde encounters, being in 
1088:     resonance, induce a stronger response that extends to larger radii. 
1089:     {\em Bottom:} Quantities of interest in the merger. 
1090:     {\em Left:} Star formation rate as a function of time in the prograde encounter (solid line; retrograde 
1091:     is similar, but with a weaker enhancement in bursts). Red dotted line shows the specific 
1092:     angular momentum of the gas that will participate in either 
1093:     burst, in arbitrary units); as the gas rapidly loses angular momentum after the passages, 
1094:     it drives a central starburst. 
1095:     {\em Center Left:} Net specific torque on the primary gas that will participate in the final, 
1096:     central starburst, as a function of time in the merger (in units of the initial 
1097:     total angular momentum per Gyr). We compare the roughly numerically estimated net 
1098:     torque (diamonds; from differentiating the specific angular momentum of the gas) 
1099:     and the torque from two sources: stars in the same disk as the gas 
1100:     (internal torques; black thick line), and the secondary galaxy and extended halos 
1101:     (external torques; red dot-dashed line). The loss of angular momentum that 
1102:     drives the secondary burst at $t\sim2.2-2.5$ is driven by internal torques from 
1103:     the disturbed stellar disk; {\em not} the torque from the secondary galaxy itself. 
1104:     {\em Center Right:} Final specific angular momentum content of 
1105:     material (gas plus stars) that was originally gas at $R<R_{\rm crit}$ (our 
1106:     predicted radius where merger-induced internal torques should be efficient 
1107:     at removing angular momentum) or at $R>R_{\rm crit}$. 
1108:     There is a strong division: gas inside a characteristic radius (corresponding 
1109:     to where the internal asymmetry is strong; akin to the co-rotation resonance) 
1110:     is mostly stripped of angular momentum. Gas at larger radii conserves 
1111:     sufficient angular momentum to maintain a disk at similar $R_{e}$ and $V_{c}$. 
1112:     {\em Right:} Original (cumulative) radial distribution of gas that participates 
1113:     in the final starburst, relative to the initial disk effective radius (same in all cases) 
1114:     for the prograde and retrograde cases shown and a 
1115:     more minor retrograde merger. More resonant (prograde) and 
1116:     more major encounters induce a 
1117:     stronger response, with a larger co-rotation radius, and so torque gas out to larger 
1118:     radii and efficiently strip more gas of angular momentum as predicted. 
1119:     \label{fig:merger.demo}}
1120: \end{figure*}
1121: 
1122: 
1123: Figure~\ref{fig:merger.demo} illustrates this in a couple of
1124: representative 1:3 mass ratio major merger simulations.  For a more
1125: detailed description and illustration of the relevant physics, we
1126: refer to \citet{barneshernquist96} (particularly their Figures~3-8);
1127: but we briefly outline the scenario here.  We show the morphology of
1128: the gas before the merger, when the disk is undisturbed, and shortly
1129: after both first and second passages (the second passage leading, in
1130: these cases, to a rapid coalescence), as well as in the relaxed
1131: remnant.
1132: 
1133: The bar-like non-axisymmetric perturbation induced by the close
1134: passages is clearly evident; the stars show a similar morphology at
1135: each time, with a small phase offset in the bar pattern and (in the
1136: remnant) a stellar bulge. As we discuss in \S~\ref{sec:model.orbit}, a
1137: prograde encounter, being in resonance, produces a noticeably more
1138: pronounced bar distortion (both in amplitude -- effectively ``bar
1139: mass'' -- and spatial extent).  Shortly after each passage, this
1140: double bar system efficiently removes angular momentum from the gas,
1141: allowing it to fall into the center of the galaxy and participate in a
1142: centrally concentrated starburst. 
1143: 
1144: Following \citet{barneshernquist96}, we track the gas in the primary
1145: disk that will turn into stars in the final starburst, calculating the
1146: net instantaneous gravitational torque decelerating the disk
1147: rotation. We can coarsely infer what the total effective torque must
1148: be by simply differentiating the specific angular momentum of this gas
1149: at a given time, and compare this to the net torque from different
1150: sources. Specifically, we separate the instantaneous gravitational
1151: torques into the internal torques -- those from the stellar disk {\em
1152: in the same galaxy} as the gas, chiefly from the bar (since the
1153: axisymmetric disk, by definition, exerts no net torque) -- and the
1154: external torques -- those from the gravity of the secondary galaxy
1155: itself and the extended halos and their substructure. 
1156: 
1157: It is clear that, especially for the phases of interest shortly after
1158: second passage and leading into the final starburst, when this gas
1159: loses its angular momentum, the total torques are dominated by
1160: internal torques from the stellar disk/bar system. The agreement
1161: between these torques and the rate of change in the specific angular
1162: momentum further argues that there are no other major sources of
1163: angular momentum loss (specifically, both this comparison and direct
1164: calculation demonstrate that the ``hydrodynamic torques'' defined by
1165: pressure forces are not dominant).
1166: 
1167: As a result of these torques, gas within some critical radius where
1168: the internal torques are strong (roughly inside the ``bar radius'' in
1169: Figure~\ref{fig:merger.demo}) rapidly loses angular momentum. We
1170: define this radius more precisely in \S~\ref{sec:model.overview}
1171: below, but it is clear in the figure that at sufficiently large radius
1172: the bar perturbation is weaker (and moreover, at larger radius the
1173: potential of the disk, whether barred or unbarred, appears
1174: increasingly axisymmetric); gas outside of these radii is relatively
1175: unaffected. In general, then, the means for a more efficient encounter
1176: to consume a larger fraction of the gas in the disk is to induce a
1177: stronger bar disturbance, which is able to effectively exert internal
1178: torques out to larger radii, stripping more gas of angular momentum
1179: and bringing it into the central starburst (as evident in the stronger
1180: prograde encounter in Figure~\ref{fig:merger.demo}).  Finally, the
1181: system relaxes -- the gas that has not been subjected to strong
1182: internal torques, having retained its angular momentum (at least in
1183: large part), can rapidly re-form a disk. This may entail some
1184: redistribution of that angular momentum (``filling in'' where the bar
1185: depleted the gas of the disk), but does not lead to further
1186: significant angular momentum loss.
1187: 
1188: \citet{barneshernquist96} and 
1189: \citet{barnes:review} illustrate that this internal torquing 
1190: is by far the dominant source of 
1191: angular momentum loss, for typical orbits. This is 
1192: because the stellar bar is: (a) more or less 
1193: aligned in the plane with the gas bar, (b) trailing it 
1194: by a small amount, and (c) relatively long-lived (it lives the rest of the duration of the 
1195: merger, as opposed to the short time that is e.g.\ pericentric passage). 
1196: The companion itself (either its baryonic mass or its halo), 
1197: in most orbits, is not perfectly aligned with the 
1198: gas disk, and the torque directly from it is much weaker (the tidal 
1199: torquing drops by a factor $\sim(R_{\rm disk}/R_{\rm peri})^{3})$), 
1200: and it can act only for a short duration on pericentric passage. 
1201: There are some pathological orbits (e.g.\ perfectly coplanar prograde orbits) 
1202: where this is not true, but these are exceptional cases, and we 
1203: discuss them in \S~\ref{sec:model.exceptions}. 
1204: 
1205: At the final merger, one might image that mixing of random gas orbits or 
1206: collisions and shocks 
1207: would rapidly drain angular momentum, similar to what happens to 
1208: the stars in violent relaxation. However, this is not possible, precisely because 
1209: the gas is collisional: a Lagrangian gas element cannot go back and 
1210: forth through the galaxy, but sticks to the other gas which has 
1211: some net angular momentum. 
1212: There could in principle be some net angular momentum cancellation, but this 
1213: is inefficient -- the net angular momentum will almost always be 
1214: comparable to the initial total. Even assuming random cancellation between 
1215: two disks with comparable absolute angular momentum, the average 
1216: change in net specific angular momentum is a factor $\sim2/3$; when 
1217: one accounts for the angular momentum of the orbit -- typically comparable 
1218: or even larger than that in the disks -- there is often no change or 
1219: even a net {\em gain} in the gas specific angular momentum in a merger. 
1220: 
1221: A proper calculation 
1222: shows that over the range in mass ratios $\mu\sim 0.1-1$, for a range of typical 
1223: impact parameters $b\sim0.5-5\,\scalelen$, the expected final specific angular momentum 
1224: after cancellation is approximately equal to the initial specific angular 
1225: momentum of the primary (with $\sim20\%$ scatter). Cancellation is therefore 
1226: inefficient. Even these cancellations, we find in detail, do not 
1227: generally lead to a starburst in the same manner as a merger-induced bar, 
1228: but simply lead to moderate disk contraction (and an equal number of mergers 
1229: will scatter towards the opposite sense leading to disk expansion, keeping a mean 
1230: specific angular momentum that is constant). They do not cause a starburst because, 
1231: if two random parcels or streams of gas shock and lose angular momentum, 
1232: the alignment and relative momenta would have to be near-perfect for them 
1233: to lose, say $95\%$ of their angular momentum and fall all the way to the 
1234: central $\sim 100$pc where a nuclear starburst would occur. Rather, they will lose 
1235: some fraction of order unity of their angular momentum, fall in to a slightly smaller 
1236: radius, and continue to orbit. 
1237: 
1238: Without the 
1239: bar that can continuously drain angular momentum, the true burst is indeed 
1240: inefficient. 
1241: This initial bar-induced angular momentum loss scenario has been 
1242: well-established in subsequent numerical studies 
1243: \citep[see e.g.][]{noguchi:merger.induced.bars.dissipationless,
1244: noguchi:merger.induced.bars.gas.forcing,
1245: hernquist.89,hernquist:kinematic.subsystems,borderies:planetary.rings,
1246: barnes.hernquist.91,barneshernquist96,barnes:review,
1247: mihos:starbursts.96,springel:spiral.in.merger,robertson:disk.formation,
1248: cox:kinematics,cox:massratio.starbursts,berentzen:gas.bar.interaction,
1249: naab:gas}. 
1250: We therefore can simplify our question to ask: how efficient will a given 
1251: lagging stellar bar be at removing angular momentum from a 
1252: leading gas bar? 
1253: 
1254: 
1255: 
1256: \subsection{A Simple Model: Overview}
1257: \label{sec:model.overview}
1258: 
1259: 
1260: Consider a disk that contains a total gravitational mass $M$ (which 
1261: can include a bulge and dark matter as well; the disk mass 
1262: fraction we will denote $f_{\rm disk}$) within a 
1263: characteristic scale length $\scalelen$. Some convenient dimensional variables are: 
1264: \begin{eqnarray}
1265: \nonumber & & \mdisk = f_{\rm disk}\,M\ ({\rm disk\ mass}) \\ 
1266: \nonumber & & M_{\rm bar} = f_{\rm bar}\,M\ ({\rm stellar\ bar\ mass}) \\ 
1267: \nonumber & & v_{c} = \sqrt{\frac{G\,M}{\scalelen}}\ ({\rm characteristic\ circular\ velocity}) \\ 
1268: \nonumber & & \diskfreq = \frac{v_{c}}{\scalelen}\ ({\rm characteristic\ frequency}) \\ 
1269: & & P = \frac{2\pi}{\diskfreq}\ ({\rm rotation\ period}) \, .
1270: \end{eqnarray}
1271: We also define 
1272: the disk thickness according to a characteristic (assumed constant) 
1273: relative scale height (height $H$ versus radius $R$; $H/R=$constant)
1274: \begin{equation}
1275: \tilde{H}\equiv H/R\ ({\rm disk\ scale\ height}). 
1276: \label{eqn:scale.height}
1277: \end{equation}
1278: In these units, the circular velocity at a given 
1279: cylindrical radius $R$ is given by 
1280: \begin{equation}
1281: v_{\rm circ}(r) = v_{c}\,{\tilde{v}(r)} \equiv v_{c}\,\sqrt{\frac{M_{\rm enc}(r)}{M}\,\frac{\scalelen}{R}}
1282: \end{equation}
1283: and dimensionless lengths are defined by 
1284: \begin{eqnarray}
1285: \nonumber & & \tilde{x} = x/\scalelen \\ 
1286: \nonumber & & \tilde{y} = y/\scalelen \\ 
1287: & & \tilde{R} \equiv \sqrt{x^{2}+y^{2}}/\scalelen. 
1288: \end{eqnarray}
1289: Throughout, we will use this notation: e.g.\ the dimensional variable $u$ 
1290: is equal to the dimensionless variable $\tilde{u}$ times the appropriate combination 
1291: of dimensional constants above. 
1292: 
1293: We will show that, in such a disk, a gas bar with a lagging stellar bar will 
1294: efficiently cause gas to
1295: lose its angular momentum and dissipate into a 
1296: central starburst.
1297: This will be the case for gas interior to a radius 
1298: \begin{equation}
1299: \frac{R_{\rm gas}}{\scalelen} \le \alpha\,(1-f_{\rm gas})\, f_{\rm disk} \, F(\theta,b)\, G(\mu) \, ,
1300: \label{eqn:full.equation}
1301: \end{equation}
1302: where $\alpha\sim1$ is an appropriate 
1303: integral constant (depending weakly on details of the stellar profile shape and 
1304: bar dynamics), $f_{\rm gas}$ is the gas fraction in the disk and 
1305: $f_{\rm disk}$ is the disk mass fraction. 
1306: The factor 
1307: \begin{equation}
1308: G(\mu) \equiv \frac{2\mu}{(1+\mu)}
1309: \end{equation}
1310: contains the dependence on the merger mass ratio 
1311: (where $\mu\le1\equiv M_{2}/M_{1}$). 
1312: The term 
1313: \begin{equation}
1314: F(\theta,b) \equiv {\Bigl(}\frac{1}{1+[b/\scalelen]^{2}}{\Bigr)}^{3/2}\,\frac{1}{1-\orbitfreq/\diskfreq}
1315: \label{eqn:full.eqn.orbit.1}
1316: \end{equation}
1317: accounts for the orbital parameters: $b$ is distance of 
1318: pericentric passage on the relevant final passage before coalescence
1319: ($\sim1-$a couple $\scalelen$, for typical cosmological mergers) and 
1320: $\orbitfreq$ is the orbital frequency at pericentric passage, 
1321: \begin{eqnarray}
1322: \nonumber \frac{\orbitfreq}{\diskfreq} &=& \frac{v_{\rm peri}}{v_{c}}\,
1323: \frac{\scalelen}{b}\,\cos{(\theta)} \\
1324: \nonumber &=& \sqrt{2\,(1+\mu)}\,[1+(b/\scalelen)^{2}]^{-3/4}\,\cos{(\theta)}\\
1325: &\approx& 0.6\,\cos{(\theta)} \, ,
1326: \label{eqn:full.eqn.orbit.2}
1327: \end{eqnarray}
1328: where $\theta$ is the inclination of the orbit relative to the disk, 
1329: and the last equality comes from 
1330: adopting typical cosmological orbits and mass ratios (but in any case, 
1331: this is quite weakly dependent on the mass ratio). 
1332: 
1333: In the following sections, we derive this scaling piece by piece, and 
1334: compare each aspect to the results from our library of hydrodynamic 
1335: simulations. We show that it is robust and accurate as an approximation 
1336: to the behavior in full numerical hydrodynamic experiments over a 
1337: wide dynamic range of several orders of magnitude in surviving disk 
1338: fraction (from systems with $\sim80-100\%$ of their disks surviving a 
1339: merger to systems with $<1\%$ disk after a merger), as well 
1340: as the entire dynamic range in mass, gas content, orbital properties, 
1341: and different feedback prescriptions with which we experiment. 
1342: 
1343: 
1344: \subsubsection{Dependence on Disk Gas Content}
1345: \label{sec:model.gas}
1346: 
1347: 
1348: Let us consider an infinitely thin gas bar (a good approximation, owing to the 
1349: efficiency of gas cooling) in a potential that is otherwise cylindrically symmetric 
1350: except for the presence of a stellar bar. 
1351: For simplicity, assume that the gas bar follows a fixed pattern speed 
1352: $\Omega_{\rm b}$ ($\sim \orbitfreq$; we will derive the pattern speed later) 
1353: in the disk (while there is not exactly a constant pattern speed in the outer regions of the disk, 
1354: the torques are weak there in either case, and this approximation is globally quite good). 
1355: We take $z=0$ to be the plane of the 
1356: disk, and, without loss of generality, consider a 
1357: frame rotating with the pattern speed of the gas bar, so that the 
1358: bar lies along the $x$ axis. The material in the bar is rotationally supported, 
1359: so it has instantaneous velocity ${\bf v}_{\rm \phi} =-v_{c}\,\tilde{v}(R)\,\hat{y}$, where 
1360: $v_{c}\,\tilde{v}(R)$ (defined above) is the circular velocity at each point $x$. 
1361: 
1362: Now, consider a stellar bar of total mass $M_{\rm bar}$ also at fixed pattern speed, 
1363: but offset by some instantaneous angle $\barangle$ 
1364: from the gas bar (i.e.\ along the axis $y=\tan{(\barangle)}\,x$). 
1365: The mass per unit length in the bar at a distance $R$ along the bar 
1366: is ${\rm d}M_{\rm bar}/{\rm d}R = (M_{\rm bar}/\scalelen)\,\tilde{\Sigma}(R/\scalelen)$, 
1367: where $\tilde{\Sigma}$ is the appropriate dimensionless mass profile 
1368: and $\scalelen$ is some characteristic scale length (usually corresponding closely 
1369: to the scale length of the unperturbed disk). If the initial disk is in equilibrium, 
1370: (i.e.\ if the bar is some reasonable perturbation to the initial system), then 
1371: the unperturbed net acceleration in the $x$ direction at some point $x$ in the gas bar 
1372: will just be cancelled by the rotation of the system. Of interest here is the 
1373: torque; if the stellar bar is also thin, then 
1374: at a point $x=\tilde{x}\,\scalelen$ in the gas bar, the net torque per unit mass from the 
1375: stellar bar will be 
1376: \begin{equation}
1377: \frac{{\rm d}j}{{\rm d}t}=\tilde{x}\,\scalelen\,\frac{{\rm d}v_{y}}{{\rm d}t}=
1378: - \frac{G\,M_{\rm bar}}{\scalelen}\,
1379: I_{0}(\barangle,\,\tilde{x}),
1380: \label{eqn:bar1}
1381: \end{equation}
1382: where $I_{0}\sim1$ is a dimensionless integral which depends 
1383: weakly on $\barangle$ and $\tilde{x}$ 
1384: (at large $\barangle$, $I_{0}\rightarrow0$, 
1385: reflecting the fact that the torque is dominated by times when the 
1386: bars are close; since $\barangle\ll 1$ is expected, it is a good 
1387: approximation to ignore the $\barangle$ dependence of $I_{0}$). 
1388: 
1389: If we assume the stellar bar is infinitely thin, there is 
1390: a weak divergence in $I_{0}$ as $\barangle \rightarrow0$ 
1391: (the accelerations become large when the bars nearly overlap). 
1392: More accurately, we can allow for some finite height in the 
1393: stellar disk/bar (it will always be thicker than the gas disk/bar). 
1394: Let the stellar bar have a constant relative scale height $H/R$ given 
1395: by Equation~(\ref{eqn:scale.height}), 
1396: and for simplicity take its vertical profile to be constant 
1397: density out to a height $\pm H$ (although assuming a 
1398: more realistic vertical profile $\propto \exp{(-|z|/H)}$ or 
1399: $\propto{\rm sech}^{2}{(z/H)}$ makes almost no difference to our calculation). 
1400: The specific torque at $x$ in the gas bar now becomes 
1401: \begin{equation}
1402: \frac{{\rm d}j}{{\rm d}t}=
1403: - \frac{G\,M_{\rm bar}}{\scalelen}\, 
1404: \frac{1}{\sqrt{\sin^{2}{\barangle}+\tilde{H}^{2}}}\ 
1405: I_{1}(\barangle,\,\tilde{x},\tilde{H}),
1406: \label{eqn:bar2}
1407: \end{equation}
1408: where $I_{1}$ 
1409: is an even weaker function of $\barangle$ and $\tilde{x}$ than $I_{0}$. 
1410: The important behavior is entirely captured ignoring $I_{1}$, 
1411: namely that the finite width of the stellar bar suppresses the 
1412: numerical divergence seen earlier. 
1413: 
1414: The bar mass $M_{\rm bar}$ represents the stellar mass in the disk 
1415: that is effectively part of the bar at the appropriate instant. We can therefore 
1416: parameterize $f_{\rm bar} = M_{\rm bar}/M$ as 
1417: \begin{equation}
1418: f_{\rm bar} = (1-f_{\rm gas})\,f_{\rm disk}\,\Psi^{\prime}_{\rm bar}. 
1419: \end{equation}
1420: Here, $f_{\rm disk}$ is the disk mass fraction, and 
1421: $f_{\rm gas}$ is the gas fraction in the disk (since we are interested 
1422: in the cold gas, we explicitly ignore gas in e.g.\ a bulge or 
1423: hot halo component). Therefore, the stellar mass of the 
1424: disk is $(1-f_{\rm gas})\,f_{\rm disk}$ -- this defines the maximum 
1425: mass that could be in the stellar bar. The parameter 
1426: $\Psi^{\prime}_{\rm bar}$ thus defines the bar ``efficiency'' -- in 
1427: an instantaneous sense as we have defined it here, 
1428: $\Psi^{\prime}_{\rm bar}=0$ means there is no stellar bar, 
1429: $\Psi^{\prime}_{\rm bar}=1$ implies the maximal stellar bar.
1430: 
1431: Already, we have one significant scaling -- the bar strength, 
1432: and correspondingly the strength of the torques on the 
1433: gas, scale with $(1-f_{\rm gas})$. In very gas rich 
1434: systems where $f_{\rm gas}\rightarrow1$, there is no 
1435: stellar mass to form a lagging bar and remove angular momentum 
1436: from the gas. The gas itself may form a bar, but without a 
1437: stellar bar to drag it, the angular momentum loss 
1438: (over the timescales of relevance for a merger\footnote{At least, 
1439: in this case, a major merger. The situation becomes more complicated 
1440: in the limit of minor mergers with mass ratios $\sim$1:10; see 
1441: \S~\ref{sec:model.secular}}) is  
1442: inefficient. This is well known in e.g.\ dynamical studies of 
1443: pure gas and stellar bars \citep[e.g.][]{schwarz:disk-bar,
1444: athanassoula:bar.orbits,pfenniger:bar.dynamics,
1445: combes:pseudobulges,friedli:gas.stellar.bar.evol,oniell:bar.obs}. There might be 
1446: some angular momentum loss in such a case, between e.g.\ bar 
1447: and halo \citep[e.g.][]{hernquistweinberg92},
1448: but it will be small -- certainly nowhere near the 
1449: efficient stripping of angular momentum needed to 
1450: induce a significant starburst. 
1451: 
1452: If we consider a Lagrangian gas element at some initial 
1453: radius $x_{0}$, then its orbit will decay as it loses angular 
1454: momentum. The instantaneous rate of change in 
1455: the radius $R=|x|$ will be given by 
1456: ${\rm d}R/{\rm d}t = (R/v_{\phi})\,{\rm d}v_{\phi}/{\rm d}t$. 
1457: The characteristic timescale for the system to evolve is 
1458: given by $2\pi/\diskfreq$, where 
1459: $\diskfreq\sim v_{c}/h$ is the characteristic 
1460: frequency of the disk. Define 
1461: the timescale 
1462: \begin{equation}
1463: \tau \equiv \frac{\diskfreq}{2\pi}\,t = \frac{v_{c}}{\scalelen}\,t = \sqrt{\frac{G\,M}{\scalelen^{3}}}\,t, 
1464: \end{equation}
1465: where $M$ is the total effective gravitational mass of the 
1466: disk and $\scalelen$ is again a characteristic scale length. 
1467: We now have: 
1468: \begin{equation}
1469: \frac{{\rm d}\tilde{x}}{{\rm d}\tau} = 
1470: \frac{2\pi}{\diskfreq}\,\frac{1}{v_{\phi}}\,\frac{G\,M_{\rm bar}}{\scalelen^{2}}\,I_{1}
1471: \equiv 2\pi\,f_{\rm bar}\, I_{2}(\barangle,\ \tilde{x}) \, .
1472: \end{equation}
1473: 
1474: Because the merger occurs on a couple of dynamical timescales, 
1475: i.e.\ a time $\Delta\tau\sim1$, 
1476: to lowest order (ignoring e.g.\ the complications of different orbital 
1477: parameters) we expect that gas 
1478: within a radius 
1479: \begin{equation}
1480: \tilde{x}\ll \Delta\tau\,\frac{{\rm d}\tilde{x}}{{\rm d}\tau}
1481: \end{equation}
1482: will efficiently lose angular momentum and fall to the center, becoming 
1483: part of the central starburst, while gas at 
1484: $\tilde{x}\gg \Delta\tau\,\frac{{\rm d}\tilde{x}}{{\rm d}\tau}$ will avoid 
1485: the starburst. 
1486: This defines a 
1487: scale
1488: \begin{equation}
1489: \frac{R_{\rm gas}}{\scalelen} \lesssim (1-f_{\rm gas})\, f_{\rm disk} \,
1490: \Psi_{\rm bar}(\barangle,\tilde{H},...)
1491: \label{eqn:rtemp1}
1492: \end{equation}
1493: within which the gas will lose angular momentum. 
1494: For convenience, we have collected all of the dimensionless 
1495: integral factors, including $\Psi^{\prime}_{\rm bar}$ (the efficiency of forming 
1496: the stellar bar) and the dynamical integral factors (e.g.\ $I_{1}$, $I_{2}$) from above, 
1497: into the term $\Psi_{\rm bar}$ that represents the full solution. 
1498: We write $\Psi_{\rm bar}(\barangle,\tilde{H}...)$ because, as we will show, 
1499: this quantity (at present) encapsulates our ignorance of e.g.\ the orbital 
1500: parameters and merger mass ratio; for a $1:1$ merger on a typical orbit, 
1501: however, $\Psi_{\rm bar}\sim1$. 
1502: 
1503: The total gas mass 
1504: which will lose angular momentum and 
1505: fall into the center of the galaxy will be 
1506: $f_{\rm gas}\times f(<R_{\rm gas})$, where $f(<R_{\rm gas})$ is the 
1507: mass fraction within the characteristic radius above, according to the 
1508: details of the mass profiles and dimensionless integral above. 
1509: We consider solutions for a variety of 
1510: profiles, including e.g.\ an exponential, isothermal sphere, and a 
1511: \citet{mestel:disk.profile} $1/R$ disk profile. In general, we find that there is little difference
1512: between the predictions for these various profiles -- the differences in the 
1513: mass profile shapes tend to cancel out and leave only weak corrections to the 
1514: simple dimensional scaling. 
1515: An exponential disk 
1516: with $\Sigma\propto \exp(-R/\scalelen)$ contains a 
1517: mass fraction 
1518: \begin{equation}
1519: 1-(1+R/\scalelen)\,\exp{(-R/\scalelen)}, 
1520: \label{eqn:fburst.1}
1521: \end{equation}
1522: within a radius R; here $\Psi_{\rm bar}$ must be solved numerically. 
1523: We obtain nearly identical predictions, however, 
1524: assuming a $1/R$ disk or an isothermal sphere profile for the gas, 
1525: which allows us to analytically solve the relevant equations and 
1526: write the predicted gas fraction consumed in the form: 
1527: \begin{equation}
1528: f_{\rm burst} = f_{\rm gas}\,(1-f_{\rm gas})\, f_{\rm disk}\, \Psi_{\rm bar}(\barangle,\tilde{H},...) \, ,
1529: \label{eqn:fburst.2}
1530: \end{equation}
1531: where $\Psi_{\rm bar}\sim 1$ can be analytically calculated for 
1532: these profiles (with the equations above) under certain conditions: 
1533: if the dependence on the orbital parameters is separable 
1534: and we define $\Psi_{\rm bar}$ 
1535: by the requirement that the radius in Equation~(\ref{eqn:rtemp1}) satisfy 
1536: $\tilde{x}=\Delta\tau\,\frac{{\rm d}\tilde{x}}{{\rm d}\tau}$, then for 
1537: instantaneous bar lag of $\barangle\sim$a few degrees in a thick disk of 
1538: height $\tilde{H}\sim0.2$, $\Psi_{\rm bar}$ is given by
1539: \begin{equation}
1540: \Psi_{\rm bar} \approx F(...)\,{\Bigl\{}1 - 
1541: \exp{{\Bigl[}-\frac{\sin(2\,\barangle)}
1542: {\sin^{2}(\barangle)+\tilde{H}^{2}}{\Bigr]}}{\Bigr\}} \sim 1 \, ,
1543: \label{eqn:fburst.3}
1544: \end{equation}
1545: where we explicitly show $F(...)$ 
1546: as we have suppressed our ignorance of the orbital parameters. 
1547: Nevertheless, this simple scaling alone provides a remarkably successful 
1548: description of many of our simulations. 
1549: 
1550: 
1551: 
1552: \begin{figure}
1553:     \centering
1554:     \scaleup
1555:     %\plotone{fgas_fsb.ps}
1556:     \plotter{f7.ps}
1557:     \caption{Mass fraction formed in the central, dissipational starburst 
1558:     as a function of gas mass fraction at the time just before the starburst, 
1559:     in a suite of major $1:1$ mass-ratio mergers. 
1560:     Solid lines are our theoretical predictions ($f_{\rm burst}=f_{\rm gas}(1-f_{\rm gas})\,\Psi$, 
1561:     see Equation~\ref{eqn:fburst.2}), 
1562:     dotted line corresponds to bursting all the available gas 
1563:     ($f_{\rm burst}=f_{\rm gas}$). 
1564:     We show results here for several orbits from Table~\ref{tbl:orbits}: a typical random orbit 
1565:     with both disks inclined ({\bf e:} {\em top}), an inclined polar-prograde orbit ({\bf k:} {\em middle}), 
1566:     and a polar-polar orbit ({\bf f:} {\em bottom}). 
1567:     The simulations agree well with our analytic predictions: more gas-rich mergers 
1568:     are less efficient at torquing angular momentum away from the gas 
1569:     and funneling it into the starburst (efficiency $\sim(1-f_{\rm gas})$). 
1570:     \label{fig:fgas.fsb}}
1571: \end{figure}
1572: 
1573: Figure~\ref{fig:fgas.fsb} tests this simple prediction. For a suite of merger simulations, 
1574: we compare the mass fraction in the central starburst, $f_{\rm burst}$, 
1575: to the gas content of the 
1576: (immediately pre-merger) disks, $f_{\rm gas}$.
1577: We can either determine the starburst mass fraction 
1578: by directly measuring the gas mass that loses its angular momentum and participates 
1579: in the brief nuclear starburst, or by measuring the gas content that survives and 
1580: forms a disk (described in \S~\ref{sec:form.major}) and assuming the gas that did not survive 
1581: (relative to that available just before the final merger) was part of the burst. 
1582: In either case, we obtain a nearly identical answer for each simulation. For now, 
1583: we consider only simulations with a $1:1$ mass ratio -- we will generalize to arbitrary 
1584: mass ratios below. In all these simulations, the 
1585: pre-merger $f_{\rm disk}\approx1$, and measuring 
1586: $\barangle$ and $\tilde{H}$ just before the merger we 
1587: expect $\Psi_{\rm bar}\approx1$. We consider one set of orbits at a time -- i.e.\ compare 
1588: only systems with the same orbital parameters, so that we can temporarily suppress 
1589: the dependence on them (this yields a systematic offset between each set of 
1590: orbital parameters -- the solutions plotted account for that following our solution in the 
1591: next section). At a fixed orbit, for these mergers, then, the only parameter 
1592: that matters should be $f_{\rm gas}$. The simulations at each orbit span a wide 
1593: range in $f_{\rm gas}$, from $\sim0.01-1$. 
1594: 
1595: We compare the relation between $f_{\rm burst}$ and $f_{\rm gas}$ resulting 
1596: from the full numerical experiments to the simple scalings predicted by 
1597: Equations~(\ref{eqn:rtemp1})-(\ref{eqn:fburst.2}). In detail, we show two solutions -- 
1598: first, the scaling given by Equation~(\ref{eqn:fburst.2}), 
1599: $f_{\rm burst}\propto f_{\rm gas}(1-f_{\rm gas})$, appropriate for an isothermal 
1600: sphere or \citet{mestel:disk.profile} disk profile with 
1601: $\Phi_{\rm bar}\approx1$; and second, the appropriate numerical solution 
1602: (following Equations~\ref{eqn:rtemp1}-\ref{eqn:fburst.1}) for an 
1603: exponential disk. In either case the analytic solutions are similar, and agree 
1604: well with the trend seen in the simulations. 
1605: It is clear that the efficiency of the burst in simulations is -- as we predict -- 
1606: {\em not} constant. It is not the case that the entire gas supply is always stripped 
1607: of angular momentum and consumed in the final merger (which would yield 
1608: $f_{\rm bust}=f_{\rm gas}$). Rather, when $f_{\rm gas}$ is sufficiently high, 
1609: only a fraction $\sim(1-f_{\rm gas})$ of the available gas is able to efficiently lose 
1610: its angular momentum and participate in the starburst. 
1611: 
1612: 
1613: \begin{figure}
1614:     \centering
1615:     \scaleup
1616:     %\plotone{fgas_fdisk.ps}
1617:     \plotone{f8.ps}
1618:     \caption{Relaxed post-merger remnant disk mass fraction versus gas fraction 
1619:     just before the merger, for $1:1$ major mass-ratio mergers. In this case essentially 
1620:     all the pre-merger stellar mass is transformed (violently relaxed) into bulge -- the 
1621:     disk is formed from the gas that survives the merger. Panels consider different orbits, 
1622:     with points as Figure~\ref{fig:fgas.fsb}. Solid lines are our theoretical predictions 
1623:     ($f_{\rm disk}=f_{\rm gas}[1-(1-f_{\rm gas})\,\Psi]$, see Equation~\ref{eqn:fburst.2}), 
1624:     dotted lines correspond to all the gas surviving and forming a disk 
1625:     ($f_{\rm disk}=f_{\rm gas}$). Again, the simulations agree well with our 
1626:     analytic predictions; gas-rich mergers are inefficient at stripping angular momentum 
1627:     from the gas, leaving significant gas content that rapidly re-forms a post-merger disk. 
1628:     \label{fig:fgas.fdisk}}
1629: \end{figure}
1630: 
1631: Figure~\ref{fig:fgas.fdisk} repeats this comparison in terms of the surviving disk mass. 
1632: We argued that the gas that does not lose angular momentum in the merger will 
1633: survive to re-form a disk. Because these are $1:1$ mergers where we can safely assume 
1634: the entire stellar disks are destroyed, we expect then that 
1635: the disk mass fraction will be $f_{\rm disk} = f_{\rm gas} - f_{\rm burst}$. 
1636: Using the method described in \S~\ref{sec:form.major} to estimate the 
1637: remnant disk mass fractions, we plot $f_{\rm disk}$ versus $f_{\rm gas}$ for 
1638: each of several orbital parameter sets. Again, the exact details of the predictions 
1639: depend on orbital parameters in a manner we derive below, but for 
1640: now we are interested in whether or not they obey 
1641: the predicted scaling with $f_{\rm gas}$. 
1642: Indeed, they do. Over $2-3$ orders of magnitude in fractional disk mass 
1643: (and $\sim5-6$ in absolute disk mass), the simple scaling here agrees well with 
1644: full numerical experiments. It is clear that some of $f_{\rm gas}$ is consumed, as 
1645: expected (if all the gas survived, we would obtain $f_{\rm disk} = f_{\rm gas}$; but 
1646: in fact, especially at low $f_{\rm gas}$, the efficiency of angular momentum loss 
1647: is high as predicted and the gas participates in the starburst). 
1648: 
1649: 
1650: \begin{figure}
1651:     \centering
1652:     %\scaleup
1653:     %\plotone{fgas_fdisk_exp.ps}
1654:     \plotone{f9.ps}
1655:     \caption{Relaxed post-merger remnant disk mass fraction versus our 
1656:     analytic predictions as a function of gas fraction and orbital parameters, 
1657:     for $1:1$ mass-ratio mergers. Error bars correspond to variation using 
1658:     different methods to estimate the disk-bulge decomposition. 
1659:     \label{fig:fgas.fdisk.exp}}
1660: \end{figure}
1661: 
1662: Figure~\ref{fig:fgas.fdisk.exp} simplifies this -- we again 
1663: compare $f_{\rm disk}$ and $f_{\rm gas}$, but effectively 
1664: put all the orbits on the same footing
1665: by plotting $f_{\rm disk}$ versus the predicted $f_{\rm disk}(f_{\rm gas},...)$ (i.e.\ including 
1666: the orbital parameters according to the predictions in \S~\ref{sec:model.orbit} below). 
1667: Essentially this 
1668: amounts to implicitly including $F(...)$ in Equation~(\ref{eqn:fburst.2}) above. The 
1669: remaining scaling should just represent the predicted 
1670: $f_{\rm burst}\propto f_{\rm gas}(1-f_{\rm gas})$. For each simulation, we show 
1671: an error bar corresponding to the range of $f_{\rm disk}$ estimated using 
1672: different methods (e.g.\ a full three-dimensional kinematic decomposition, 
1673: one and two-dimensional kinematic modeling, and surface brightness profile 
1674: fits, as described in \S~\ref{sec:form.major}). The agreement is surprisingly good, 
1675: given the simplicity of our derivation. Moreover, the scatter is quite small -- a 
1676: factor $\sim2-3$ at very low $f_{\rm gas}$ and considerably smaller ($\lesssim50\%$) 
1677: at high $f_{\rm gas}$. It seems that our simple scaling indeed 
1678: captures the most important physics of angular momentum loss -- namely that 
1679: with less fractional stellar material in the disk, there is less mass 
1680: available to torque on the 
1681: gas bar in a merger, therefore less angular momentum loss in the gas. 
1682: 
1683: 
1684: 
1685: \subsubsection{Dependence on Orbital Parameters}
1686: \label{sec:model.orbit}
1687: 
1688: We now turn to how the details of the orbit affect the loss of 
1689: angular momentum. Before, we made the simplifying assumptions 
1690: that the stellar bar lagged by some constant angle $\barangle$ 
1691: and that the characteristic time for the perturbation to act 
1692: was of order the disk rotational period/dynamical time. While 
1693: these turn out to give reasonable scalings, we can improve upon 
1694: them. 
1695: 
1696: The secondary (``perturber'') galaxy will have some characteristic 
1697: orbital frequency $\orbitfreq$, 
1698: approximately given by 
1699: \begin{equation}
1700: \orbitfreq \sim \frac{v_{\rm peri}}{b} \, ,
1701: \end{equation}
1702: where $v_{\rm peri}$ is the velocity at pericentric passage and 
1703: $b$ is the impact parameter or pericentric passage distance. 
1704: Because the behavior we are interested in is relatively short-lived, this 
1705: is reasonable even for first passages or ``flyby'' encounters -- 
1706: we are interested in the orbital frequency at pericentric passage because this 
1707: is when the forcing is strongest and the bar is driven. 
1708: 
1709: Now, consider the frame rotating with the 
1710: disk/bar at a frequency $\sim\diskfreq$.  In this frame, 
1711: the secondary galaxy will have an 
1712: apparent frequency for an orbit projected into the disk plane of 
1713: \begin{equation}
1714: \Omega_{\rm eff} = \orbitfreq\,\cos{\theta}-\diskfreq, 
1715: \end{equation}
1716: where $\theta$ is the inclination of the orbit relative to the plane of the 
1717: disk (in standard parlance for orbital parameters as described in \S~\ref{sec:sims}). 
1718: Note that we are interested in the 
1719: component of the orbital motion in the plane of the disk: in terms of the 
1720: standard orbital parameters $\theta$ and $\phi$ of the primary 
1721: galaxy (angle of the angular momentum vector of the disk relative to the 
1722: plane of the orbit), this is $\orbitfreq\,\cos{\theta}$. 
1723: For a case with e.g.\ $\theta=0$ (prograde) and 
1724: a parabolic orbit with small impact parameter ($\orbitfreq\sim\diskfreq$), 
1725: the system is maximally prograde -- in the frame of the rotating disk 
1726: there is almost no net circular motion of the secondary. 
1727: For the same orbit but $\theta=180\degree$ (retrograde), 
1728: the secondary completes a circular orbit around the disk 
1729: in just half a disk dynamical time ($\Omega_{\rm eff}\sim -2\,\diskfreq$). 
1730: The time required for the secondary to complete a revolution in 
1731: this frame is therefore (in our dimensionless units $\tau = t / (2\pi/\diskfreq)$)
1732: \begin{equation}
1733: \tau_{\rm circ} = \frac{1} {1-\frac{\orbitfreq}{\diskfreq}\,\cos{\theta}}. 
1734: \end{equation}
1735: 
1736: The timescale for a gas element to lose its angular momentum and 
1737: fall to the center of the galaxy is given by our earlier 
1738: estimate of the torque, as $\tau_{\rm loss}\sim \tilde{x} / ({\rm d}\tilde{x}/{\rm d}\tau)$. 
1739: If $\tau_{\rm loss}\ll \tau_{\rm circ}$ at a given radius, then 
1740: the derivation we have obtained is essentially valid: the system 
1741: sees a quasi-static perturbation to the potential, loses its angular 
1742: momentum, and collapses before the perturbation can damp out 
1743: or circularize. However, if $\tau_{\rm loss}\gg \tau_{\rm circ}$, 
1744: then the system has not lost much angular momentum by the time 
1745: the secondary completes a revolution, and will gain 
1746: some of those losses back as the system comes around the other 
1747: side. In the limit where 
1748: $\tau_{\rm circ}$ is short (much shorter than the local 
1749: dynamical time), for example, then the potential is effectively circularized -- 
1750: the gas at these radii may undergo oscillatory motion and even have e.g.\ 
1751: spiral waves driven by this external forcing, but there is no means by 
1752: which the system can introduce a strong net asymmetry to drive inflows. 
1753: 
1754: We can therefore improve our previous estimate: instead of taking 
1755: $\Delta\tau\sim1$ (i.e.\ a disk rotation period) as 
1756: the only characteristic timescale, we 
1757: argue that gas with 
1758: \begin{equation}
1759: \frac{\tilde{x}}{{\rm d}\tilde{x}/{\rm d}\tau} \lesssim 
1760: \tau_{\rm circ} = \frac{1}{1-\frac{\orbitfreq}{\diskfreq}\,\cos{\theta}}
1761: \label{eqn:circ.x}
1762: \end{equation} 
1763: will lose its angular momentum, while gas at larger radii will not. 
1764: 
1765: In detail, we can integrate the equations from 
1766: \ref{sec:model.gas} for a parcel of gas at some initial radius 
1767: $x_{0}$, in a time-dependent potential of this nature. For simplicity, we assume that 
1768: the secondary drives a circular perturbation in the potential with frequency 
1769: $\omega=\orbitfreq$ and calculate the bar response using the 
1770: gaseous disk (assuming, again, that it 
1771: is infinitely cold) and stellar disk (assuming that the scale height $\tilde{H}$ 
1772: translates into a corresponding velocity dispersion $\sigma/v_{c}$) 
1773: wave dispersion relations from 
1774: \citet{binneytremaine}. In practice, we find this is not much different 
1775: from assuming that the lag in the stellar bar grows with 
1776: time $\propto \tau_{\rm circ}$ (i.e.\ that the stellar bar can keep up or reverse 
1777: sense tracking the perturbation without significant energy loss; 
1778: or, more or less equivalently, that the two bars 
1779: are only in phase when the perturbation is strong, and then rapidly fall out of 
1780: phase -- at the $\gtrsim5-10\degree$ level, once the perturbation is weak or 
1781: reverses its sense as the phase of the secondary reaches $\gtrsim \pi/2$). 
1782: In principle, we now have a physically motived and fully time-dependent 
1783: model for $\barangle(t)$ and the response of the gas bar. This allows us 
1784: to properly integrate out the dependence of $\Psi_{\rm bar}$ on $\barangle$ and 
1785: instantaneous conditions and replace it with the appropriate integral dependence 
1786: on orbital parameters and disk gas content and structure. 
1787: 
1788: We find that there is a strong division in expected behavior, at more or less 
1789: exactly the characteristic radius implied by Equation~(\ref{eqn:circ.x}). Within 
1790: this radius, gas (in our simple numerical calculations) is effectively torqued 
1791: efficiently as it enters the gas bar near resonance (but slightly leading 
1792: the stellar bar), and plunges to the center. Gas outside this radius begins to 
1793: feel a perturbation, but then the phase of the secondary cycles around and the sense of the 
1794: torques begin to weaken or reverse (depending on the details of the orbit), 
1795: and generate wave motion in the gas but no significant angular momentum loss 
1796: or infall. Not only is the transition between these two regimes predicted by the simple 
1797: scalings above, but we find in more detailed numerical calculations that the 
1798: width of the transition region (where behavior is more sensitive to the details of 
1799: e.g.\ the profile shapes and assumptions about the bars) is quite narrow, 
1800: $\sim20\%$ of that radius. 
1801: 
1802: This should not be surprising. Essentially, what we have derived is a rough 
1803: equivalent of the co-rotation condition, but for forced bars as opposed to 
1804: isolated self-generating (swing-amplified) bar instabilities. It is well-known 
1805: from studies of idealized bars \citep[see e.g.][]{schwarz:disk-bar,
1806: pfenniger:bar.dynamics,noguchi:merger.induced.bars.gas.forcing,binneytremaine,
1807: berentzen:gas.bar.interaction}
1808: that gas can be efficiently torqued inwards 
1809: inside of the co-rotation resonance (in the language above, given the forcing 
1810: with pattern speed $\orbitfreq$, this is interior to the radius where the relative motion of 
1811: the secondary is slow relative to the dynamical time, and so the perturbation 
1812: does not circularize). Moreover, the resonant structure around these radii 
1813: is known to be sharp; if we 
1814: follow a derivation similar to \citet{borderies:planetary.rings} 
1815: (their derivation is intended to apply to planetary disks with satellites, but the 
1816: relevant physics is similar) it is straightforward to show that the 
1817: detailed numerical prefactors will be swamped for all but a narrow range of 
1818: radii around this resonance by the strong dependence of the resonant forcing 
1819: on radius (roughly going as some large power 
1820: of $(r/r_{\rm crit})$ -- such that the forcing is strong inside the resonance 
1821: and rapidly weakens outside). 
1822: 
1823: This gives us confidence that we can adopt the scalings above and 
1824: robustly assume that there is indeed a characteristic radius (depending 
1825: in detail upon orbital parameters) interior to which the gas will lose its angular 
1826: momentum. 
1827: This resonant structure of the angular momentum loss is actually 
1828: quite convenient from an analytical perspective, as it means that more subtle 
1829: issues of e.g.\ the thermal pressure and state of the ISM, stellar and AGN 
1830: feedback, and the exact mix of e.g.\ gas and stars or density structure of the 
1831: gas will not contribute significantly to determining which gas can or cannot 
1832: lose angular momentum. Unlike e.g.\ a self-generating bar in an unstable disk, 
1833: there is no issue of stability analysis -- the torques inside this critical radius (and 
1834: the inducing perturbation) are sufficiently strong such that all the material 
1835: therein loses angular momentum in a very short time (much less than a single 
1836: orbital time, in practice). 
1837: 
1838: For example, it is well known that in isolated cases, a 
1839: pure gas disk is more unstable to gravitational perturbation than a stellar disk 
1840: \citep[see e.g.][]{christodoulou:bar.crit.1,christodoulou:bar.crit.2,
1841: mayer:lsb.disk.bars}, however in the driven case this is not applicable: the 
1842: distortion in the local stellar/gas distribution is caused by the secondary, not by 
1843: e.g.\ orbital ``pileup'' or instability in the primary. The location of the resonant 
1844: radius is not determined by the internal structure of the primary (unlike in an 
1845: isolated case, where it is determined by how e.g.\ those orbits can overlap and 
1846: where various stability criteria are satisfied), but rather by the orbital motion 
1847: of the secondary (relative to the internal motion of the primary), and therefore knows 
1848: nothing about e.g.\ the gas to stars ratio, phase structure, and feedback 
1849: situation in the primary. Inside this radius, the distortion is sufficiently strong that it 
1850: does not matter whether one configuration or another is more or less prone to 
1851: gravitational instability -- the driving force (and therefore angular momentum loss) 
1852: is large in any case. 
1853: 
1854: Exactly what the pressure support of the gas inside 
1855: this radius is may effect e.g.\ how far it free-falls after losing angular momentum 
1856: before shocking and forming a central starburst, but it will not change the fact 
1857: that the angular momentum loss is efficient. Quantitatively, the torque is 
1858: $\gg j_{\rm disk}\,\Omega_{d}$ (as it must be in order for the gas to lose its 
1859: angular momentum in much less than an orbital period); but e.g.\ the pressure 
1860: gradients resisting gas collapse cannot be larger than (in energetic terms)
1861: $\tilde{H}\,M_{d}\,V_{d}^{2}\sim\tilde{H}\,j_{\rm disk}\,\Omega_{d}\ll j_{\rm disk}\,\Omega_{d}$ 
1862: (or else the disk could not be thin) -- therefore whether or not there is even considerable 
1863: pressure support or e.g.\ thermal feedback or a modified ISM equation of state makes a 
1864: negligible correction to the behavior seen in the simulations. 
1865: 
1866: Before moving on, we would like to translate the general 
1867: scaling above in terms of $\orbitfreq$ and $\diskfreq$ into 
1868: more convenient parameters. As noted above, $\orbitfreq\sim v_{\rm peri}/b$. 
1869: We expect $b\sim1-3\,\scalelen$ for common parabolic cosmological orbits -- 
1870: as we discuss below, for orbits with larger $b$ that will eventually merge, all 
1871: that matters in terms of the end product is the impact parameter of 
1872: the final passage or two when the most dramatic forcing occurs, so even for 
1873: initially larger passages, angular momentum transfer to the halo will ensure 
1874: a value in this range towards the final stages of the merger. 
1875: 
1876: Assuming a parabolic orbit, $v_{\rm peri}$ will be given by 
1877: the infall velocity from infinity, $\sqrt{(G\,M\,(1+\mu)/b)}$ (where 
1878: $\mu$ is the merger mass ratio, discussed below). 
1879: Because the merging systems are extended, as $b\rightarrow0$ 
1880: these expressions should be replaced by a more complicated 
1881: function of $b/\scalelen$ (for the case $b=0$, the infall velocity asymptotes 
1882: to the escape velocity from the center of the primary $\sim\sqrt{G\,M/\scalelen}$), 
1883: which requires a numerical solution for an arbitrary density profile.  
1884: In practice we find that we can interpolate between the limits $b=0$ and 
1885: $b\gg \scalelen$ quite accurately by replacing $b$ 
1886: with $\sqrt{b^{2}+\scalelen^{2}}$ (which also happens to be an exact solution 
1887: for e.g.\ a Plummer sphere density profile). 
1888: Combining these 
1889: factors, we find that (for the regime of typical interest) 
1890: \begin{eqnarray}
1891: \nonumber \frac{\orbitfreq}{\diskfreq} &\sim& \frac{v_{\rm peri}}{v_{c}}\,
1892: \frac{\scalelen}{b}\,\cos{(\theta)} \\
1893: \nonumber &=& \sqrt{2\,(1+\mu)}\,[1+(b/\scalelen)^{2}]^{-3/4}\,\cos{(\theta)}\\
1894: &\approx& 0.6\,\cos{(\theta)} \, ,
1895: \end{eqnarray}
1896: where the last term comes from inserting a typical major merger 
1897: mass ratio and $b\sim2\,\scalelen$. 
1898: The orbital dependence is then -- as we would expect -- largely a function of 
1899: the inclination angle $\theta$. Prograde orbits induce a strong bar 
1900: response -- despite the fact that in these mergers the 
1901: orbital angular momenta are all aligned, we actually expect the most 
1902: angular momentum loss and least efficient disk formation. Retrograde 
1903: and polar mergers, on the other hand, despite having completely 
1904: un-aligned or cancelling total angular momentum, should most efficiently 
1905: form disks. 
1906: 
1907: Inserting this dependence on orbital parameters into our previous 
1908: derivation in Equation~(\ref{eqn:rtemp1}) allows us to effectively 
1909: replace the part of  $\Psi_{\rm bar}$ which parameterized our ignorance of 
1910: orbital parameters ($F(...)$ in Equation~\ref{eqn:fburst.2}), giving
1911: \begin{equation}
1912: \Psi_{\rm bar}(\theta,\tilde{H},...) \propto \frac{1}{1-\frac{\orbitfreq}{\diskfreq}\,\cos{\theta}}. 
1913: \label{eqn:orbit.dept.1}
1914: \end{equation}
1915: In short, our previous derivation applies, but the orbital dependence is now 
1916: explicit in $\Psi_{\rm bar}$. 
1917: 
1918: Revisiting Figures~\ref{fig:fgas.fsb} \&\ \ref{fig:fgas.fdisk}, recall that we included 
1919: this orbital dependence in the predicted curves therein. For each orbit, the 
1920: predicted curve is given by the solution for the gas mass within the critical 
1921: $R_{\rm gas}/\scalelen$ (Equation~\ref{eqn:rtemp1}) with 
1922: the dependence on orbital parameters as in Equation~\ref{eqn:orbit.dept.1} -- 
1923: we insert the appropriate orbital inclination $\theta$ and impact parameter 
1924: $b$ for the two disks in the orbit and sum their expected $f_{\rm burst}$ or $f_{\rm disk}$ 
1925: to derive the model prediction. 
1926: The difference between the different orbits does not appear dramatic 
1927: in Figure~\ref{fig:fgas.fsb} -- but this is because the burst fraction $f_{\rm burst}$ is plotted 
1928: on a linear scale, suppressing the dependence on $\theta$ at small $f_{\rm gas}$ 
1929: (most of the visible dynamic range in the plot is at large $f_{\rm gas}$ -- in this regime, however, 
1930: the stellar bar is weak in any case because there is not much stellar mass in the disk -- so 
1931: the result is that much of the gas survives and becomes part of the disk, regardless of orbital 
1932: parameters). 
1933: 
1934: However, the difference between different orbits is 
1935: much more clear in Figure~\ref{fig:fgas.fdisk}, displayed on a logarithmic scale. At 
1936: low $f_{\rm gas}$, there is much more stellar mass in the disk than gas mass, so in principle 
1937: the stellar bar could (if maximal) easily torque away all the angular momentum of the 
1938: gas. Here, however, the orbital parameters become important in determining 
1939: just how efficient this process should actually be. For the orbits close to retrograde 
1940: ($\cos{\theta}\approx-1$), 
1941: the scaling we have just derived suggests that $\Psi_{\rm bar}$ should be 
1942: suppressed by a factor $\sim2$. 
1943: But for orbits close to coplanar prograde ($\cos{\theta}\approx1$), 
1944: $\Psi_{\rm bar}$ is enhanced by a factor $\sim2-3$ -- in other words, because the orbit 
1945: is nearly resonant, the effective co-rotation resonance (the orbit interior to which the gas 
1946: can efficiently lose angular momentum to the induced stellar bar) is moved out 
1947: by a substantial factor, including a larger fraction of the disk gas (in those extremes, 
1948: only the gas at very large radii survives the merger). 
1949: 
1950: Again, Figures~\ref{fig:fgas.fsb}-\ref{fig:fgas.fdisk} demonstrate that 
1951: the simple scalings based on our model provide an accurate description of the 
1952: behavior in the full numerical experiments. Figure~\ref{fig:fgas.fdisk.exp} 
1953: combines these into a single plot -- we compare the disk fractions in our 
1954: simulations to the full expectation based on our derivation thus far as a function of 
1955: both gas fraction and orbital parameters. As noted above, the agreement is 
1956: good, with a reasonably small scatter. 
1957: 
1958: \begin{figure}
1959:     \centering
1960:     \scaleup
1961:     %\plotone{psi_theta.ps}
1962:     \plotterr{f10.ps}
1963:     \caption{Effective efficiency of bars (the parameter $\Psi$, 
1964:     efficiency at torquing gas into a starburst, removing 
1965:     its angular momentum and destroying the disk), as a function of 
1966:     the effective orbital parameter. Each point represents the effective constraint on 
1967:     $\Psi$ from fitting a correlation of the form in Equation~(\ref{eqn:fburst.2})
1968:     to a suite of simulations over a range of gas fractions, masses, and 
1969:     mass ratios (since we are interested in a comparison of orbital parameters 
1970:     here and not e.g.\ gas fractions, we choose to normalize so that $0<\Psi<1$). 
1971:     Black points are the most well-sampled orbits (e, h, k, f), 
1972:     shown in Figure~\ref{fig:fgas.fdisk}, purple points are 
1973:     more limited studies of orbits (c, i, m, d), red points are 
1974:     a study of major and minor mergers with $\phi=0$, 
1975:     $\theta=0,\ 30,\ 90,\ 150,\ 180\degree$. 
1976:     {\em Top:} $\Psi_{\rm eff}$ versus mean orbital inclination 
1977:     $\langle\cos{(\theta)}\rangle$ (for 1:1 mergers, we average the two 
1978:     inclinations). Black solid line is the simple linear scaling 
1979:     $\Psi_{\rm eff}\propto1/(1 - 0.6\,\langle\cos{\theta}\rangle)$ 
1980:     from Equation~(\ref{eqn:orbit.dept.1}); 
1981:     red dot-dashed lines are the numerical solutions for the appropriate 
1982:     $\langle\cos{\theta}\rangle$, as in Figure~\ref{fig:fgas.fdisk}. 
1983:     {\em Middle:} $\Psi_{\rm eff}$ versus the full numerical 
1984:     expectation (properly solving for the effective bar strength in each disk 
1985:     and then adding the burst fractions, rather than just taking $\Psi(\langle\cos{\theta}\rangle)$. 
1986:     The efficiency of disk destruction 
1987:     and angular momentum loss scales with orbital parameters in the 
1988:     simple manner predicted in Equation~(\ref{eqn:orbit.dept.1}). 
1989:     Over a typical random 
1990:     cosmological ensemble of orbits, we expect values similar to those 
1991:     between our typical {\bf e} and {\bf f} orbits. 
1992:     {\em Bottom:} $\Psi_{\rm eff}$ versus net specific angular momentum 
1993:     of the system (adding/cancelling the initial disk plus 
1994:     orbital angular momenta and dividing by the final baryonic mass). 
1995:     The two are actually {\em anti}-correlated, demonstrating that disks do 
1996:     not arise after or survive mergers owing to co-addition of angular 
1997:     momentum (and cancellation is inefficient at destroying disks) -- rather, 
1998:     systems with aligned angular momentum vectors are in greater resonance, 
1999:     triggering stronger internal asymmetries in the primary that drain more 
2000:     angular momentum from the gas.     
2001:     \label{fig:psi.theta}}
2002: \end{figure}
2003: 
2004: These results are for four representative orbits spanning a reasonable range in 
2005: orbital parameters -- those for which we 
2006: have a large number of simulations covering a wide range in the
2007: space of other parameters. We consider them first because this allows us 
2008: to robustly determine that the predicted orbital scalings do not depend on e.g.\ stellar 
2009: mass, halo properties, feedback prescriptions, or other varied physics in the 
2010: simulations. Having done so, we consider a more limited sampling of a much 
2011: broader range in orbits given by Table~\ref{tbl:orbits} in order to survey the 
2012: full dynamic range of orbital parameters. 
2013: 
2014: Figure~\ref{fig:psi.theta} shows the results of this. For a given suite of simulations 
2015: with some particular orbital parameters, we first construct the correlation 
2016: $f_{\rm disk}(f_{\rm gas})$ as in Figure~\ref{fig:fgas.fdisk}. Rather than adopt some 
2017: {\em a priori} model for the orbital dependence, we then fit the points in that 
2018: correlation to a function of the form in Equation~(\ref{eqn:fburst.2}) -- i.e.\ 
2019: effectively fit for the normalization or ``efficiency'' of angular momentum 
2020: removal, which we define as $\langle\Psi_{\rm eff}({\rm orbit})\rangle$. 
2021: We compare this, for our ensemble of orbits, to our analytic expectation 
2022: from the simple scaling in Equation~(\ref{eqn:orbit.dept.1}) and to 
2023: a full numerical solution (technically, for 1:1 mergers, we want to solve this separately 
2024: for each disk and add the two, although just considering the primary is a 
2025: good approximation for less major mergers). 
2026: The agreement with our analytic model is quite good across the entire range 
2027: of orbital parameters, implying that we have captured the most important 
2028: physics of resonant interactions in this simple scaling. 
2029: 
2030: We also compare this effective efficiency of disk destruction with the 
2031: net specific angular momentum of the merger remnant (assuming 
2032: pure addition/cancellation of the initial baryonic angular momenta of the disks 
2033: and the orbital angular momentum). The result is actually an anti-correlation: 
2034: systems with aligned angular momentum vectors, e.g.\ coplanar prograde mergers 
2035: being the extreme case, induce the most efficient bars and remove angular 
2036: momentum most efficiently from the gas. Systems where the angular momentum 
2037: vectors are misaligned (e.g.\ polar orbits) or anti-aligned (retrograde) 
2038: actually leave the largest disks in place. This clearly emphasizes that it 
2039: is not, in fact, any direct addition/cancellation of angular momentum that 
2040: determines or enables disks to form in and survive mergers. Rather, the cases with the 
2041: largest net angular momentum are most resonant, inducing the strongest resonant 
2042: asymmetries in the merging pair, which most effectively drains angular momentum 
2043: from the gas and leaves a compact, bulge-dominated remnant. 
2044: 
2045: 
2046: \subsubsection{Dependence on Mass Ratio}
2047: \label{sec:model.massratio}
2048: 
2049: The major remaining parameter to study is the merger mass ratio. 
2050: Thus far, we have restricted our attention to equal mass 
2051: $1:1$ mergers, which allowed us to make several convenient simplifying assumptions. 
2052: Nevertheless, most of our previous derivation applies. None of the 
2053: scalings that we have explicitly derived up to now are dependent upon 
2054: mass ratio. However, we have quantified the strength of the induced stellar and 
2055: gas bars with the parameter $\Psi_{\rm bar}$, which we expect should 
2056: scale with mass ratio. Moreover, we have made the assumption that the 
2057: pre-merger disk stars are entirely violently relaxed by the merger. While this 
2058: is a good assumption for $1:1$ mergers, it is not true for minor mergers 
2059: (a $1:10$ mass ratio merger, even with no gas, will clearly not transform the entire 
2060: primary stellar disk into bulge). 
2061: 
2062: First, consider this stellar component: there are a number of ways to derive 
2063: the disturbance of the stellar component in the merger. The simple expectation 
2064: is that the mass in galaxy $M_{1}$ which can be violently relaxed by 
2065: collision with galaxy $M_{2}$ is proportional to $M_{2}/M_{1}$ -- the net energy deposit, 
2066: tidal forces, and the mass fraction brought in from a potentially disrupted satellite 
2067: all scale in this manner. For simplicity, consider the case where the secondary 
2068: $M_{2}$ is much smaller and more dense than the primary, and falls in on a nearly radial orbit 
2069: in the final encounter (which is a good approximation, given the efficiency of 
2070: angular momentum transfer from the orbit to the halo). Since we are assuming 
2071: $M_{2}\ll M_{1}$, treat $M_{2}$ as a point mass, and consider its final orbital 
2072: decay, where it oscillates with rapidly decaying amplitude through the center of the primary 
2073: with initial impact velocity $v_{\rm i}\approx v_{c}$ and damping spatial amplitude 
2074: $\ell_{\rm max}\lesssim R_{d}$. 
2075: At some instant, then, the secondary is at 
2076: location $(R^{\prime},\phi^{\prime},\,z^{\prime})=\ell\,(\cos{\theta},\,0,\,\sin{\theta})$ 
2077: (we rotate such that the secondary orbit defines $\phi=0$ without loss of generality). 
2078: A star in the primary disk at $(R,\phi,z=0)$ then feels some potential from the 
2079: secondary ($\equiv \Phi_{2}$) and experiences
2080: a vertical deflection out of the 
2081: disk $\partial\Phi_{2}/\partial z = (G\,M_{2}/R^{3})\,\ell\,\sin{\theta}\,f(\ell/R)$, 
2082: where $f(u)=[1+u^{2}-2\,u\cos{\theta}\,\cos{\phi}]^{3/2}\sim1$. We are only interested 
2083: in the time the secondary spends at $\ell \sim R$ (when its much closer to the 
2084: disk or further away, the vertical perturbation is weak), so it effectively 
2085: acts for a time $\delta t \sim R/v_{\rm i}$ as it passes through $\ell \sim R$ 
2086: in its ringing about the center. If we know the full potential, we can 
2087: solve for the deflection as a function of time and calculate the full acceleration of 
2088: the disk stars at $(R,\phi)$, which yields an effective 
2089: net velocity deflection while the secondary is on one side of the galaxy 
2090: of $\delta{v} = (G\,M_{2}/R\,v_{\rm i})\,(\ell_{\rm max}\sin{\theta}/R)\,f(\ell_{\rm max}/R)$. 
2091: Deflection occurs when $\delta{v} \sim v$ or larger, so if 
2092: $v=\tilde{v}(r)\,v_{c}$ (where $\tilde{v}$ depends weakly on $r$)
2093: and we substitute for $v_{c}\equiv\sqrt{G\,M_{1}/R_{d}}$ here and in $v_{\rm i}$ we obtain 
2094: the criterion $G\,M_{2}/R \gtrsim v_{c}\,v_{i} \sim v_{c}^{2} \sim G\,M_{1}/R_{d}$, 
2095: i.e.\ (rearranging) $R/R_{d} \lesssim M_{2}/M_{1}\,\sin{\theta} = \mu\,\sin{\theta}$. 
2096: The $\sin{\theta}$ dependence comes because we considered only vertical 
2097: deflection of stars (i.e.\ some heating to $v_{z}^{2}$)-- 
2098: a coplanar orbit (in this limit) will obviously induce no such 
2099: heating, but will introduce deflections in the radial direction (heating 
2100: $v_{R}^{2}$). We can repeat 
2101: our derivation considering where these deflections are significant, and find 
2102: (as one would expect) $R/R_{d} \lesssim \mu\,\cos{\theta}$. So, the absolute 
2103: mass fraction scattered should be more or less angle-dependent, although 
2104: the orbital anisotropy $\beta_{z}\equiv 1 - \bar{v_{z}^{2}}/\bar{v_{R}^{2}}$ 
2105: will depend significantly on  the orbital inclination $\theta$. 
2106: 
2107: For the case of a thin 
2108: \citet{mestel:disk.profile} disk with no bulge, we can solve these equations exactly and obtain 
2109: the simple solution that 
2110: a merger with a secondary $M_{2}$ scatters exactly $M_{2}$ worth of stars in the primary, 
2111: completely independent of the inclination $\theta$, (but with an 
2112: anisotropy 
2113: $\beta_{z}(\theta) \sim 1 - \frac{2\,\sin^{2}{\theta}}{(1+2\,\cos^{2}{\theta})}$ -- 
2114: although this ignores a proper treatment of further mixing as the perturbed stars 
2115: interact with each other, and thus does not reproduce orbits quite as radial as seen 
2116: in simulations). 
2117: The full numerical solutions for arbitrary cases yield the general result that, 
2118: when inside a radius that encloses a mass $\sim M_{2}$ in the primary, 
2119: then the presence of the mass $M_{2}$ is a significant perturbation, which 
2120: scatters those stars in the primary -- i.e.\ deflections occur rapidly, so the stars 
2121: violently relax. At larger radii, where $M_{\rm enc} \rightarrow M_{1}>M_{2}$, 
2122: the motion of the $M_{2}$ secondary at the center is a small perturbation. The 
2123: disk at these radii is perturbed adiabatically by the motion of the secondary, which 
2124: can induce some warps and/or disk heating, but will not violently relax the stars. 
2125: Reversing this derivation for the secondary, it is trivial that essentially all 
2126: the mass in the secondary (we ignore stripping of the tightly bound stellar mass) 
2127: will be violently relaxed. So the total mass violently relaxed will 
2128: be $\sim M_{2}$ (in the primary) plus $M_{2}$ (the secondary), out of a total 
2129: mass $M_{1}+M_{2}$ -- i.e.\ in terms of the mass ratio $\mu\equiv M_{2}/M_{1}$, the 
2130: fraction of the pre-merger {\em stellar disk} mass which is destroyed and turned into 
2131: bulge is
2132: \begin{equation}
2133: f_{\ast,\rm disk}(\rm destroyed) = \frac{2\,\mu}{1+\mu}. 
2134: \label{eqn:minor.scaling}
2135: \end{equation}
2136: Technically this assumes the systems are initially pure disk, but the corrections 
2137: if they have pre-existing bulges are not large (generally smaller than the 
2138: simulation-to-simulation variation; although we discuss them in 
2139: more detail in \S~\ref{sec:prescriptions}), so this is a reasonable approximation 
2140: for general cases. 
2141: 
2142: Now, consider the gas. It turns out that a similar 
2143: linear scaling in Equation~(\ref{eqn:minor.scaling})
2144: is found for how the gas mass in the starburst (i.e.\ the fraction 
2145: that loses its angular momentum) scales with mass ratio, as one might expect. 
2146: In detail, though, the derivation must be revisited (and will include additional terms 
2147: depending on orbital parameters): because the gas is collisional, 
2148: even a large vertical deflection of 
2149: gas at some $R$ does not translate to a loss of that gas disk, since the gas can 
2150: dissipate the vertical energy and no loss of rotational angular momentum 
2151: has occurred. Deflections in the $R$ direction will be resisted by hydrodynamic forces. 
2152: So, for a proper derivation, we return to our model of the stellar bar torquing 
2153: the gas bar. The essential question is how the amplitude of the induced stellar bar 
2154: (our term $\Psi_{\rm bar}$) should scale with mass ratio. 
2155: 
2156: Take the thin disk limit (this is just for convenience, the final scaling is unchanged if we 
2157: allow for a finite stellar disk thickness); the 
2158: disk surface density is linear in the potential according to Poisson's equation, 
2159: \begin{equation}
2160: \nabla^{2}\Phi = 4\pi\,\Sigma(R,\phi)\,\delta{(z)}. 
2161: \end{equation}
2162: So, since the non-axisymmetric 
2163: potential of the secondary, at some distance $b$ (the impact parameter), 
2164: must scale as roughly $\Phi \sim G\,M_{2} / b^{3}$, 
2165: we expect the amplitude of the induced bar (perturbation in $\Sigma$) 
2166: should also scale as $M_{2}/b^{3}$. Fractionally, this yields 
2167: $\Psi_{\rm bar}\propto M_{\rm bar}/M_{1} \sim \mu\,(h/b)^{3}$. 
2168: 
2169: We can show this more formally: 
2170: if $\Phi_{0}$ is the (azimuthally symmetric) potential of the primary and 
2171: $\Phi_{1}$ is the perturbative potential of the secondary, which induces the 
2172: surface density perturbation $\Sigma_{1}\propto f_{\rm bar}\propto \Psi_{\rm bar}$ 
2173: that defines the bar, we 
2174: have $\nabla^{2}\Phi_{1} = 4\pi\,\Sigma_{1}(R,\phi)\,\delta{(z)}$. 
2175: We can expand any potential $\Phi_{1}$ 
2176: as $\Phi_{a}(k\,R)\,\exp{[\imath\,(m\,\phi-\omega\,t)-k\,|z|]}$, which gives the 
2177: trivial solution 
2178: $\Sigma_{1} = \Sigma_{a}(k\,R)\,\exp{[\imath\,(m\,\phi-\omega\,t)-k\,|z|]}$ 
2179: where $\Sigma_{a} = -|k|/(2\pi\,G)\,\partial^{2}\Phi_{a}/\partial r^{2}$. 
2180: We expect $\Phi_{a}\sim -G\,M_{2}/r$, so 
2181: we obtain $\partial^{2}\Phi_{a}/\partial r^{2} \sim -2\,G\,M_{2}/r^{3}$ (note 
2182: that we can generalize this to extended distributions for the secondary, 
2183: relevant for e.g.\ more major mergers, and the change is, for reasonable 
2184: profiles, equivalent to replacing $r$ with $\sqrt{r^{2}+a^{2}}$). 
2185: The details of the mode structure turn out not to be important, since 
2186: the behavior we are interested in is dominated by modes with $|k|\sim 1/\scalelen$; 
2187: but we can, for example, treat $\Phi_{a}(k\,R)$ as the potential 
2188: generated by a point source (appropriate for e.g.\ the small mass-ratio limit) 
2189: at the impact parameter $b$
2190: and expand the wave modes appropriately, then integrate over the modes 
2191: in the disk to determine the bar strength (i.e.\ the total 
2192: mass effectively contributing to the bar). In any case, up to a numerical 
2193: constant that is weakly sensitive to the mode structure, 
2194: we obtain 
2195: $M_{\rm bar} \propto \frac{\scalelen^{3}}{(b^{2}+\scalelen^{2})^{3/2}}\,M_{2}$. 
2196: In terms of $\Psi_{\rm bar} = M_{\rm bar}/M_{1}$, this gives 
2197: \begin{equation}
2198: \Psi_{\rm bar}\propto \mu\,  (1+[b/\scalelen]^{2})^{-3/2} \, ,
2199: \label{eqn:massratio.b.forcing}
2200: \end{equation}
2201: where again $b$ is the impact parameter and the 
2202: $1+[b/\scalelen]^{2}$ term effectively allows for the interpolation between 
2203: the case of an orbiting point mass and a penetrating encounter 
2204: \citep[see e.g.][]{binneytremaine}. 
2205: 
2206: As noted in \S~\ref{sec:model.orbit}, 
2207: if we are just interested in the end product of a merger -- i.e.\ we do not 
2208: care what happens on each passage separately as the companions lose 
2209: angular momentum, but only in the surviving disk fraction and total 
2210: burst fraction -- then we are not interested in some initial impact parameter 
2211: $b$ but only in the impact parameter on the final passages close to 
2212: coalescence, when angular momentum loss has made the orbits 
2213: nearly radial and the forcing is strong. We can see directly from 
2214: Equation~(\ref{eqn:massratio.b.forcing}) 
2215: that the forcing is dramatically suppressed by a factor $\sim(b/\scalelen)^{3}$ 
2216: on earlier, large-impact parameter passages, so these can be effectively 
2217: ignored in calculating the remnant properties (we confirm this is 
2218: true in a sample of simulations with much larger $R_{\rm peri}$). 
2219: Eventually, for any systems which are destined to merge, 
2220: angular momentum transfer yields a nearly radial orbit with $b\sim\scalelen$, 
2221: and this is where most of the forcing occurs, so the remnant solution is 
2222: effectively given by ignoring the $b$ dependence above (technically 
2223: summing over each passage with the appropriate $b$ is possible, 
2224: but in practice we obtain the same result to within the
2225: simulation-to-simulation scatter by assuming $b\rightarrow0$ in 
2226: Equation~\ref{eqn:massratio.b.forcing}). The final dependence on any 
2227: initial impact parameter $b$ is therefore weak, so long as the 
2228: systems are bound to merge. The dependence on mass ratio $\mu$, however,
2229: is fixed. 
2230: 
2231: If the mass enclosed is linear in $\Psi_{\rm bar}$ (the case 
2232: for e.g.\ the \citet{mestel:disk.profile} disk and an isothermal sphere, and not a 
2233: bad approximation for the regime of interest for an exponential 
2234: disk), we then have a similar result for the gas as the stellar distribution: 
2235: the secondary 
2236: induces a burst of mass $\propto M_{2}$ in the primary $M_{1}$. 
2237: Reversing the derivation, the primary (since it is larger, so 
2238: $M_{1}/M_{2}>1$ induces a burst (assuming the two have similar initial 
2239: gas fractions) $\propto M_{2}$ in the secondary (i.e.\ bursting all its gas). 
2240: The net burst mass $\propto 2\,M_{2}$ relative to the remnant 
2241: mass $M_{1}+M_{2}$ is then 
2242: \begin{equation}
2243: f_{\rm bust}\propto \frac{2\,\mu}{1+\mu}. 
2244: \end{equation}
2245: This is generally applicable for mergers; however we will note below 
2246: that, because they do not coalesce (and therefore do not 
2247: eventually come in with $b\rightarrow0$ or brake their orbital energy 
2248: interior to the stellar distribution of the primary), this is not exactly 
2249: applicable to e.g.\ fly-by or first passage scenarios. 
2250: 
2251: \begin{figure*}
2252:     \centering
2253:     \scaleup
2254:     %\plotone{massratio_fdisk.ps}
2255:     \plotone{f11.ps}
2256:     \caption{{\em Top:} Surviving disk mass fraction $f_{\rm disk}$ as a function of  
2257:     pre-merger gas fraction $f_{\rm gas}$, for a series of 
2258:     mergers of varying mass ratios (symbols). Each panel shows a series of mergers with 
2259:     different orbital parameters (orbits {\bf m000-m180}). Color encodes 
2260:     mass ratio of the merger: 1:1 (red), 1:2 (green/cyan), 1:4 (blue), 1:8 (purple). 
2261:     Dotted lines (of corresponding color) show our prediction (Equation~\ref{eqn:full.equation}) 
2262:     for the given orbital parameters and mass ratio. Note that for minor 
2263:     mergers, $f_{\rm disk}>f_{\rm gas}$ is allowed, because some of the original 
2264:     stellar disks are predicted to survive the merger as well as some of the gas 
2265:     which does not lose angular momentum.  
2266:     {\em Bottom:} Starburst mass fraction in mergers of a given mass ratio $\mu$, 
2267:     relative to our model prediction for 1:1 mass ratio mergers with 
2268:     the same orbit and pre-merger gas content (symbols, as 
2269:     top panels). Lines show our simple linear model (solid black; this does 
2270:     well for typical orbits but the bursts in nearly prograde orbits -- m000 -- 
2271:     are somewhat more efficient than predicted owing to the effects described in 
2272:     \S~\ref{sec:model.exceptions}) and full numerical calculation (dot-dashed red; 
2273:     two lines correspond to different mass profiles), as in Figure~\ref{fig:fgas.fdisk}. 
2274:     Minor mergers induce less efficient bursts, and do not completely destroy the 
2275:     primary disk: the scaling of these efficiencies with mass ratio agrees well 
2276:     with our dynamical model predictions. 
2277:     \label{fig:massratio.fdisk}}
2278: \end{figure*}
2279: 
2280: Figure~\ref{fig:massratio.pred} tests this prediction in an ensemble of 
2281: simulations spanning a range in mass ratio from 
2282: $\mu=0.1-1$. For a given set of orbital parameters (fixed), we plot 
2283: the disk and burst fractions ($f_{\rm disk}$ and $f_{\rm burst}$) of the 
2284: remnant, as a function of the immediate pre-merger 
2285: gas fraction $f_{\rm gas}$, as in Figures~\ref{fig:fgas.fsb}-\ref{fig:fgas.fdisk}. 
2286: For the $1:1$ mergers, we plot our expectation based on the 
2287: simple scaling in Equations~(\ref{eqn:rtemp1})-(\ref{eqn:orbit.dept.1}), 
2288: including the dependence on $f_{\rm gas}$ and orbital parameters ($\theta$) 
2289: following \S~\ref{sec:model.gas}-\ref{sec:model.orbit}. 
2290: We then show the prediction for mass ratios $1:2$, $1:4$, and $1:8$, according 
2291: to our derivations here. 
2292: 
2293: This includes two important corrections: 
2294: instead of assuming the entire stellar disk is turned into bulge (which 
2295: was a good approximation for the $1:1$ mergers), we allow the fraction 
2296: of the stellar disk that is destroyed (turned into bulge) to 
2297: depend on mass ratio following Equation~(\ref{eqn:minor.scaling}) -- so 
2298: some (considerable) fraction of the disk is assumed to survive in 
2299: higher mass-ratio mergers. We also include the scaling with mass ratio 
2300: in $\Psi_{\rm bar}$, used as before to calculate how much of the 
2301: gas participates in the starburst. So, in the high mass-ratio cases, 
2302: both the fraction of the gas that loses its angular momentum 
2303: (fraction of $f_{\rm gas}$) and fraction of the pre-merger primary stellar 
2304: disk turned into bulge (fraction of $(1-f_{\rm gas})$) are suppressed 
2305: by a factor $\sim\mu$. 
2306: 
2307: For each of the orbits surveyed (and the range in e.g.\ absolute masses, 
2308: gas fractions, and feedback prescriptions in our minor merger 
2309: simulations), this simple rescaling according to the merger mass 
2310: ratio provides a good approximation to the behavior in the 
2311: full hydrodynamic experiments. Both the total surviving disk fraction 
2312: (which reflects both the ability of the pre-merger stellar disks 
2313: and the pre-merger gas to survive the merger) and the burst 
2314: fractions (which reflect only how much of the gas survives/loses angular 
2315: momentum) are accurately predicted, suggesting that our derivations 
2316: are reasonable for both the dissipational and dissipationless components 
2317: of the galaxy. 
2318: 
2319: 
2320: \begin{figure}
2321:     \centering
2322:     %\scaleup
2323:     %\plotone{massratio_corr.ps}
2324:     \plotter{f12.ps}
2325:     \caption{Predicted (as a function of 
2326:     orbital parameters, pre-merger gas content, and 
2327:     mass ratio) and actual post-merger surviving disk fraction for the 
2328:     simulations in Figure~\ref{fig:massratio.fdisk} (symbol type 
2329:     and color denote orbit and mass ratio in the same style). 
2330:     {\em Top:} Comparison assuming there is no dependence on 
2331:     mass ratio (i.e.\ treating all cases as $\mu=1$). Clearly, this 
2332:     is inappropriate for minor mergers, but it is also inappropriate 
2333:     for even intermediate major mergers (mass ratios $\mu=0.3-0.5$). 
2334:     {\em Bottom:} Comparison including the predicted 
2335:     dependence on mass ratio 
2336:     of both destruction of the stellar disk and 
2337:     angular momentum loss in the gas. 
2338:     Our predictions as a function of mass ratio, orbital parameters, 
2339:     and gas fraction are accurate in the simulations to $\sim0.1$ in $f_{\rm disk}$. 
2340:     \label{fig:massratio.pred}}
2341: \end{figure}
2342: 
2343: Figure~\ref{fig:massratio.pred} summarizes these results. We first compare 
2344: the final disk fraction in the simulations to our prediction including e.g.\ 
2345: the dependence on gas content and orbital parameters but {\em without} 
2346: any accounting for mass ratio (assuming all mergers are just as efficient 
2347: as a $1:1$ merger). Unsurprisingly, this works for the $1:1$ mergers, 
2348: but is a terrible approximation to mergers of very different mass ratios. 
2349: We then compare allowing for the same scalings but including the 
2350: predicted mass ratio dependence. The agreement between full simulation 
2351: and our simple analytic expectations is good -- with a scatter for 
2352: the high disk fractions typical of intermediate and minor merger remnants 
2353: as low as $\sim20\%$. 
2354: 
2355: One important caveat here is that, for mergers of increasingly
2356: small mass ratio $\mu$,
2357: the merger timescales become long. At the smallest mass ratios we 
2358: consider, $\sim$1:10, this timescale may become sufficiently long that 
2359: the secular (i.e.\ self-amplifying) instability/response of the disk 
2360: may become important over the duration of the merger. It is not entirely 
2361: clear what the response of such an (initially driven) system will be; whether or 
2362: not, for example, the driven non-axisymmetric modes will remain locked to 
2363: their driver (the secondary orbit) or de-couple and move at the 
2364: pattern speed dictated by the internal stability properties of the disk. 
2365: This competition between secular processes (more sensitive to e.g.\ the 
2366: detailed structure, rotation, and pressure support of the disk) and 
2367: merger-driving in this regime probably contributes to 
2368: some of the increased scatter in burst fractions seen in Figure~\ref{fig:massratio.fdisk} 
2369: at the lowest $\mu$. For this reason, it is reasonable to restrict a 
2370: definition of ``mergers'' to this mass ratio and more major interactions: at smaller mass ratios, 
2371: secular/internal processes (even if initially driven by interactions) may 
2372: be more important than the direct driving from the interactions themselves 
2373: (or at least operate on comparable timescales). 
2374: 
2375: 
2376: \subsubsection{First Passage and Fly-By Encounters}
2377: \label{sec:model.flyby}
2378: 
2379: We have derived a general equation for the disk mass that should be lost in 
2380: mergers, and demonstrated that it is robust to variations in a wide range of 
2381: galaxy properties. Most of our derivation is completely generalizable as 
2382: well to encounters where the systems will not merge (or at least are not 
2383: immediately merging). Two cases of interest (which are, in the short term, essentially 
2384: equivalent) arise: first passages and ``fly-by'' encounters (in which 
2385: there is a close encounter but the velocities are sufficiently large to 
2386: delay or prevent a merger). 
2387: 
2388: In such a passage, there is of course no violent relaxation and mixing of stars, 
2389: so we assume the stellar disk is left intact (excepting the bar response). 
2390: The same physics will govern bar formation and 
2391: loss of gas angular momentum. The primary difference is the suppression 
2392: by the appropriate impact parameter $b$ in 
2393: Equation~(\ref{eqn:massratio.b.forcing}). 
2394: We argued before that the term 
2395: $\propto [1+(b/\scalelen)^{2}]^{3/2}$ should ultimately be neglected for 
2396: mergers because in the final passage(s) that dominate, the condition of merging 
2397: more or less guarantees $b\rightarrow0$. However, clearly this is not the case 
2398: on a non-merging passage. 
2399: 
2400: This introduces a non-trivial uncertainty -- 
2401: we quote $[1+(b/\scalelen)^{2}]^{3/2}$ where $b$ is the impact parameter 
2402: and $\scalelen$ is some characteristic scale length of the system. But in detail, 
2403: the appropriate ``impact parameter'' is really the {\em actual} distance of closest 
2404: approach, which is usually somewhat smaller than the distance of approach 
2405: estimated from infinity (the formal impact parameter definition), as some angular 
2406: momentum is already lost. Moreover, in detail, is 
2407: the appropriate $\scalelen$ the exponential scale length? The half-mass radius? 
2408: Any such radii are of course closely related, and all of these uncertainties in the 
2409: exact definition change the term $b/\scalelen$ only at the factor $\sim2$ level, 
2410: but since the dependence $\sim(b/\scalelen)^{3}$ is fairly strong, this is important 
2411: on a quantitative level for these fly-by situations. 
2412: 
2413: \begin{figure*}
2414:     \centering
2415:     %\scaleup
2416:     \scaledown
2417:     %\plotone{firstpassage_model.ps}
2418:     \plotterr{f13.ps}
2419:     \caption{Comparison of the burst fractions in single fly-by (or equivalently, first-passage) 
2420:     scenarios and the general application of our model scalings. In these cases, there is 
2421:     no significant violent relaxation (no stars merge), so the stellar disk is left completely 
2422:     intact. Some pseudobulge may result from the induced bar and disk heating, but 
2423:     we are not modeling that here. Some burst still results from the same induced 
2424:     non-axisymmetry in the primary, which should be described by 
2425:     our same scaling (Equation~\ref{eqn:full.equation}). The important differences from 
2426:     a case that will merge are: {\bf (1)} the suppression of the induced burst 
2427:     by a factor $\sim [1+(b/\scalelen)^{2}]^{3/2}$ (where $b$ is the impact parameter and 
2428:     we find decent agreement with our simulations when $\scalelen$ is the half-mass 
2429:     disk radius), whereas in cases that will merge $b\rightarrow0$ is appropriate, 
2430:     {\bf (2)} the lack of violent relaxation of the stellar disk, and {\bf (3)} an expected 
2431:     increased scatter, as the details of the approach are more important (and there is no 
2432:     merger/in-spiral, which tends to average out the exact details of the approach). 
2433:     {\em Top:} Burst fraction (relative to gas supply at the time of the passage) versus 
2434:     gas fraction. Our simple linear model prediction (black solid) and 
2435:     numerical predictions (red dot-dashed) are shown, with the results from 
2436:     the first passages and fly-by encounters of the simulations in Figure~\ref{fig:fgas.fdisk}, 
2437:     appropriate for each set of orbital parameters shown. These cases had 
2438:     $b/\scalelen\approx1$ (and that was used in the predictions -- the 
2439:     curves assuming $b=0$, as we used for the post-merger systems, would be a 
2440:     factor $\sim3$ higher, in conflict with the simulations). 
2441:     {\em Middle:} Same, but as a function of mass ratio for systems in 
2442:     Figure~\ref{fig:massratio.fdisk} with otherwise equal orbital parameters and 
2443:     gas fractions at the time of passage. 
2444:     {\em Bottom:} Same, but as a function of impact parameter for 1:1 mergers 
2445:     with $\fgas\approx0.2$ (black) and $0.4$ (blue). Note that ``burst'' fractions 
2446:     $\lesssim1\%$ of $\fgas$ are essentially equivalent to zero (equivalent to 
2447:     random fluctuations in isolated disks). 
2448:     Our predictions describe first passages and 
2449:     fly-by encounters reasonably well, although there is larger scatter about them 
2450:     owing to differences in the details of how the passage proceeds. 
2451:     \label{fig:firstpass}}
2452: \end{figure*}
2453: 
2454: In practice, we find that using the impact parameter $b$ defined as the halos approach 
2455: (i.e.\ neglecting detailed resonant loss of angular momentum) and 
2456: taking $\scalelen$ to be the half-mass radius of the system works well in a mean 
2457: sense. The results of this exercise are shown in Figure~\ref{fig:firstpass}. 
2458: We plot the fraction of the gas available at the time of a first passage or fly-by 
2459: encounter which is consumed in the induced burst \citep[we define the strength of the 
2460: induced burst by integrating the star formation excess over the interpolation between 
2461: the pre- and post-flyby star formation rates; see e.g.][for details]{cox:massratio.starbursts}, 
2462: as a function of the gas content, for different orbits as in Figure~\ref{fig:fgas.fdisk}. 
2463: We predict that the efficiency of channeling gas into the burst should 
2464: scale as $\sim (1-\fgas)$, as before, and that the scaling with orbital parameters 
2465: should be similar. 
2466: We also show, for cases with otherwise identical gas fraction at the time of 
2467: first passage and the same orbits as Figure~\ref{fig:massratio.fdisk}, 
2468: how this scales with merger mass ratio (again, expected to be the same as 
2469: that we derived above). Altogether, adopting our previous estimates, but 
2470: re-normalizing appropriately for the impact parameter of the passage 
2471: (here $b/\scalelen\approx1$, as we defined it above) yields a good approximation 
2472: to the typical behavior in our simulations. 
2473: We also test the behavior as a function of impact parameter, and find 
2474: that our simple scaling is a reasonable approximation, yielding rapidly 
2475: diminishing bursts as the impact parameter is increased to 
2476: $b\gg\scalelen$ (at some point here, our estimates from the simulations become 
2477: ambiguous, as a $\sim1\%$ enhancement in star formation is 
2478: below the level of random fluctuations in isolated disks). 
2479: We have also checked whether or not the pre-flyby stellar disks 
2480: are destroyed -- as expected, they are left more or less intact by
2481: fly-by encounters. The disks may be heated, and in fact some ``pseudobulge'' 
2482: can form from the buckling of the bar induced in the stars, but we are 
2483: not attempting to predict or study pseudobulge formation here (rather 
2484: considering it, as is often the case in observations, to be fundamentally 
2485: still part of the stellar disk rather than part of a violently relaxed ``classical'' 
2486: bulge). In an average sense, then, our derived scaling is generally 
2487: applicable. 
2488: 
2489: However, the details of exactly how the approach proceeds will 
2490: introduce considerable scatter in the amount of burst triggered on 
2491: first passages and in fly-by encounters. This is plain in the large (factor $\sim$ a few) 
2492: scatter in Figure~\ref{fig:firstpass}. Further, details such as the 
2493: structure of the bulge are increasingly important in the limit of weak interactions, 
2494: where distortions in the potential of the primary that would trigger gas inflows 
2495: can be suppressed by the presence of a larger bulge (and note the 
2496: caveat from \S~\ref{sec:model.massratio}, that the secular/internal response 
2497: of the disk will become relatively more important in weaker interactions 
2498: with smaller mass ratios and larger impact parameters). 
2499: We therefore expect in general that our predictions can be quite broadly 
2500: applied, but are less robust for any specific case if it is a single fly-by 
2501: as opposed to an integration over a full merger. 
2502: Fortunately, in the case of systems that will actually merge, these details 
2503: tend to average out or be unimportant, yielding the relatively 
2504: small scatter we have seen in our previous predictions. In those cases, we do not need to 
2505: be too concerned with the exact details of the impact approach, 
2506: nor the structural details of the galaxy (in particular because our predictions 
2507: are for integral quantities at the end of a merger, various effects will 
2508: tend to cancel out -- for example retaining more gas on first passage will 
2509: yield a larger supply for the second burst, etc.). 
2510: 
2511: 
2512: 
2513: \subsubsection{Independence from ``Feedback'' Physics}
2514: \label{sec:model.feedback}
2515: 
2516: Our derivation of the torques causing gas to lose angular 
2517: momentum in mergers is purely dynamical. All else being 
2518: equal (i.e.\ for systems with the same gas content and dynamical structure 
2519: at the time of the final merger), we therefore expect that the detailed 
2520: physics of e.g.\ ``feedback'' from supernovae, stellar winds, and AGN activity 
2521: should make little difference. 
2522: 
2523: \begin{figure}
2524:     \centering
2525:     \scaleup
2526:     %\plotone{qeos_fdisk.ps}
2527:     \plotone{f14.ps}
2528:     \caption{The effects of feedback on disk survival in mergers. 
2529:     We show the distribution in $f_{\rm burst}/f_{\rm burst,\ pred}$, 
2530:     i.e.\ the burst mass fraction, relative to that predicted (or equivalently, the 
2531:     mean in our simulations) for the given pre-merger gas fraction and 
2532:     orbital parameters, for simulations with two different ISM gas 
2533:     feedback prescriptions (different effective equations of state $q_{\rm eos}$). 
2534:     We also show the corresponding (but measured differently) 
2535:     disk mass fractions $f_{\rm disk}/f_{\rm disk,\ pred}$. 
2536:     There is perhaps a small offset in the sense expected (a stiffer, higher-feedback 
2537:     equation of state for the ISM suppresses bursts by an average factor $\sim1.1-1.2$), 
2538:     but this is much smaller than the simulation-to-simulation scatter. For a given 
2539:     gas content at the time of the merger, then, feedback makes almost no 
2540:     difference (true for AGN feedback and starburst winds as well). 
2541:     \label{fig:qeos}}
2542: \end{figure}
2543: 
2544: Figure~\ref{fig:qeos} demonstrates that this is indeed the case. We compare 
2545: the starburst and surviving disk gas fractions 
2546: of merger remnants, relative to those predicted by our simple dynamical model 
2547: as a function of the merger mass ratio, orbital parameters, and 
2548: gas content at the time of the merger, for suites of simulations with two different 
2549: prescriptions for supernovae feedback and the effective equation of state of 
2550: the ISM. In terms of our $\qeos$ parameter (see \S~\ref{sec:sims}), 
2551: we compare $\qeos=0.25$ simulations (a nearly isothermal equation of state 
2552: with effective temperature $\sim10^{4}\,$K) to $\qeos=1$ simulations 
2553: (the ``full'' stiff \citet{springel:multiphase} equation of state, with effective temperature 
2554: $\gtrsim10^{5}\,$K at the densities of interest here). 
2555: There is no significant systematic offset between either the median 
2556: result or the scatter about our simple analytic expectation. 
2557: At most, there may be a $\sim20\%$ systematic offset, in the sense that 
2558: more highly pressurized systems ($\qeos=1$) have slightly more 
2559: gas survive -- a small offset like this is expected
2560: because the bars in these 
2561: cases are slightly more ``puffy'' (essentially the same as a slightly 
2562: thicker disk -- for which we derive an analytic expectation in Equation~\ref{eqn:fburst.3} 
2563: that yields an expected $\sim10-20\%$ difference at most based on 
2564: the full possible range of $\qeos$). In any case, such an offset is small 
2565: relative to other systematic uncertainties in disk structure and the scatter 
2566: about the median predictions. 
2567: 
2568: \begin{figure}
2569:     \centering
2570:     \scaleup
2571:     %\plotone{sbw_fdisk.ps}
2572:     \plotone{f15.ps}
2573:     \caption{As Figure~\ref{fig:qeos}, but comparing the distribution of 
2574:     disk fractions in merger remnants relative to our simple predictions as a 
2575:     function of stellar wind and quasar feedback prescriptions. We plot 
2576:     the distribution of disk fraction $f_{\rm disk}$ in simulations 
2577:     relative to our predicted $f_{\rm disk}(f_{\rm gas},\mu,\theta)$ 
2578:     (i.e.\ our calculation as a function of immediate pre-merger 
2579:     gas fraction, orbital parameters, and merger mass ratio). 
2580:     {\em Left:} Varying starburst-driven wind prescriptions. We compare 
2581:     our usual weak stellar wind scenario (winds self-consistently can 
2582:     generate from the hot gas, but additional mass loading is only 
2583:     $\sim1\%$ of the star formation rate) to a fast winds scenario 
2584:     (with additional mass loading $\sim0.5\,\dot{M}_{\ast}$ and 
2585:     wind launch velocity $\sim800\,{\rm km\,s^{-1}}$) and a slow 
2586:     winds scenario (additional mass loading $\sim2\,\dot{M}_{\ast}$ 
2587:     and launch velocity $\sim200\,{\rm km\,s^{-1}}$). 
2588:     {\em Right:} Simulations with and without feedback from 
2589:     accreting black holes. 
2590:     For otherwise fixed merger parameters (orbit, mass ratio) 
2591:     and disk properties (cold gas content, mass profiles), 
2592:     these feedback prescriptions make no different to the starburst or 
2593:     surviving disk gas fractions. They will, however, change the 
2594:     the gas content, consumption, and distribution leading into the merger. 
2595:     \label{fig:sbw}}
2596: \end{figure}
2597: 
2598: In Figure~\ref{fig:sbw}, we perform 
2599: a similar exercise for cases with and without 
2600: central supermassive black holes (we have also 
2601: examined initial BHs with varying 
2602: initial masses from $\lesssim 10^{5}\,\msun$ to $\sim10^{7}\,\msun$), 
2603: and cases with or without a simple 
2604: implementation of starburst-driven winds where winds are launched 
2605: (in addition 
2606: to the stellar feedback implicit in our multi-phase ISM model)
2607: with a mass-loading efficiency $\dot{M}_{\rm wind} = \eta_{w}\,\dot{M}_{\ast}$ 
2608: relative to the star formation rate $\dot{M}_{\ast}$ and energy loading 
2609: efficiency $\epsilon_{w}$ relative to the total energy (for a \citet{salpeter:imf} 
2610: IMF) available for supernovae \citep[sampling the range $\eta_{w}\sim 0.01-10$ 
2611: and $\epsilon_{w}\sim 0.1-1$; see][]{cox:winds}. Shown in 
2612: Figure~\ref{fig:sbw} are our fiducial weak winds ($\eta_{w}\sim0.01$, 
2613: $\epsilon_{w}\sim0.0025$), cases with moderate mass loading 
2614: into very fast winds ($\eta_{w}=0.5$, $\epsilon_{w}=0.25$, yielding a wind 
2615: launch speed $\sim800\,{\rm km\,s^{-1}}$), and cases 
2616: with high mass loading but correspondingly slower wind velocities 
2617: ($\eta_{w}=2.0$, $\epsilon_{w}=0.0625$, yielding a wind 
2618: launch speed $\sim200\,{\rm km\,s^{-1}}$). In all these cases we find 
2619: a similar result: at otherwise fixed properties at the time of merger, 
2620: feedback makes no difference to our conclusions.
2621: 
2622: The reasons for this are described in \S~\ref{sec:model.orbit} and 
2623: in more detail below (\S~\ref{sec:model.secular}). Recall, the 
2624: distortion in the primary is driven by the secondary and as such depends only 
2625: on the gravitational physics of the merger. Given this distortion, the 
2626: gravitational torques within some characteristic radius are sufficiently strong to remove 
2627: the angular momentum from the gas in much less than an orbital time. 
2628: Feedback, then, insofar as it changes the effective pressurization or equation of 
2629: state of the gas or drives a wind, is largely irrelevant: because the angular momentum is removed 
2630: in a timescale much shorter than the orbital time, the gas (regardless of the 
2631: strength of feedback) cannot dynamically respond with these hydrodynamic forces, but 
2632: must essentially free-fall into the center of the galaxy where the starburst is triggered. 
2633: The radius interior to which the torques are strong is not a function of e.g.\ the stability of the 
2634: galaxy to perturbation, because it is not an instability in the first place, but a driven 
2635: distortion in the system. Moreover, the entire system is strongly in the non-linear regime 
2636: for the time of interest (when we consider interactions 
2637: of the magnitude simulated: mass ratios $\sim$1:8 and more major 
2638: mergers) -- no amount of 
2639: making the system more robust against linear instability would be sufficient to avoid 
2640: a strong gravitational distortion in the violent coalescence surrounding 
2641: the actual merger (this must be so, because the distortion occurs where the disturbances in 
2642: the potential are greater than order unity -- for hydrodynamic forces to resist distortion, 
2643: they would have to be stronger than large-scale gravitational forces in the equilibrium system, 
2644: negating the concept of a rotationally supported thin disk). 
2645: So what matters is instead where that coalescence occurs and how long 
2646: it introduces such a strong distortion, relative to e.g.\ the local dynamical or orbital time 
2647: of some disk element, giving rise to the simple dynamical criteria for angular 
2648: momentum loss developed here. 
2649: 
2650: 
2651: That is {\em not} to say that for fixed {\em initial} conditions (significantly pre-merger 
2652: or e.g.\ at first passage), feedback will not change the result. 
2653: There are two primary means by which feedback can indirectly have a strong 
2654: influence on disk survival: 
2655: 
2656: {\bf (1) Retaining Gas (Lowering the Stellar Mass Fraction):} 
2657: As has been demonstrated 
2658: in a number of works \citep{weil98:cooling.suppression.key.to.disks,
2659: sommerlarsen99:disk.sne.fb,sommerlarsen03:disk.sne.fb,
2660: thackercouchman00,thackercouchman01,governato:disk.formation,
2661: robertson:disk.formation,springel:models,springel:spiral.in.merger,
2662: okamoto:feedback.vs.disk.morphology,scannapieco:fb.disk.sims}, these forms of feedback can have 
2663: dramatic implications, in even a short time period, for the rates at which cooling of 
2664: new cold gas from the halo and consumption of existing gas by star formation proceed. 
2665: In cases with no feedback, star formation may exhaust gas efficiently, leading to 
2666: predicted systems that are much more gas-poor at the interesting time 
2667: of the final merger -- according to our model, then, these will 
2668: not be able to form disks as efficiently as more gas-rich systems. 
2669: In cases with strong feedback from e.g.\ star formation to lower 
2670: the effective star formation efficiency and recycle gas, the predicted gas fractions 
2671: at the time of merger (from some gas-rich initial conditions) could be much higher. 
2672: Inclusion of stellar and supernovae feedback responsible for injecting 
2673: energy and turbulent pressure into the ISM may also be necessary to 
2674: prevent the onset of clumping and disk fragmentation in 
2675: isolated gas-rich cases, enabling the stable existence and evolution of 
2676: quiescent gas-rich disks \citep[see e.g.][]{springel:multiphase,
2677: robertson:cosmological.disk.formation}. 
2678: In short, feedback may be critical to give rise to high gas fractions in the first place, 
2679: which we have shown have dramatic implications for the survival of disks -- but for a 
2680: given gas fraction (however that comes about in the first place), the results of the 
2681: merger will be (in the short term) independent of feedback. 
2682: 
2683: {\bf (2) Changing the Spatial Distribution of Gas (``Kicking Gas Out'' of $R_{\rm max}$):} 
2684: Recall, our derivations demonstrate that it is not necessarily a fixed fraction of 
2685: gas that loses its angular momentum: rather (see Equation~\ref{eqn:rtemp1} 
2686: and \S~\ref{sec:model.gas}-\ref{sec:model.orbit}) 
2687: it is the mass inside some radius $R_{\rm max}/\scalelen$ relative to that of the 
2688: stellar disk (characteristic radius $\scalelen$) which will lose its angular momentum. 
2689: If some form of feedback can change the spatial gas distribution, then, it could have 
2690: dramatic implications for disk survival. We have used the radius $R_{\rm max}$ to 
2691: estimate the mass fraction that will burst by assuming the gas density profile 
2692: is broadly similar to that of the stars (which is 
2693: true in our simulations, given their feedback 
2694: prescriptions). But one could easily imagine the extreme limit, where some 
2695: strong feedback keeps all the gas at large radii $r\gg R_{\rm max} \sim \scalelen$ (i.e.\ 
2696: a case in which there is a large hole in the gas distribution, 
2697: or in which the gas is at least much more extended 
2698: than the stellar distribution) -- 
2699: the stellar disk torques only act effectively within $R_{\rm max}$, so only a tiny 
2700: fraction of 
2701: the gas in such a case would 
2702: lose its angular momentum. Especially at high redshift, 
2703: this may be important in avoiding overcooling and the 
2704: formation of too much bulge mass in many systems 
2705: \citep[see e.g.][]{robertson:cosmological.disk.formation,
2706: governato:disk.formation, donghia:disk.ang.mom.loss,
2707: ceverino:cosmo.sne.fb,zavala:cosmo.disk.vs.fb}. Again, we stress that for a 
2708: given gas density profile at the time of merger, our calculations are independent of 
2709: feedback; but if feedback alters the gas profile -- keeping the 
2710: gas at radii $\gg R_{\rm max}$, then it will largely survive the merger. 
2711: 
2712: 
2713: 
2714: \subsubsection{Exceptions and Pathological Cases}
2715: \label{sec:model.exceptions}
2716: 
2717: We have derived a general model for how disks are destroyed in mergers 
2718: and shown that it applies to a wide range of gas fractions, orbital parameters, 
2719: galaxy mass ratios, and prescriptions for feedback and gas physics. 
2720: However, there are some pathological cases of more than academic interest, 
2721: as these can explain some small differences with previous results 
2722: as well as illustrate the important physics in our model. 
2723: 
2724: For example, consider the starburst mass fraction and surviving disks in 
2725: our ``{\bf h}'' orbits: i.e.\ a prograde-prograde, coplanar merger of two 
2726: disks. In this case, the angular momentum vectors of both disks 
2727: and the orbital angular momentum are all perfectly aligned. Naively, 
2728: one might then expect that these unique cases would create the largest disks. 
2729: In fact, the {\em opposite} is true. This is largely for the reasons we 
2730: outline in \S~\ref{sec:model.orbit} -- the alignment of angular momentum vectors 
2731: means that the system is in near-perfect resonance, so it excites the largest 
2732: tidal and bar asymmetries that rapidly drain the gas of all angular momentum. 
2733: As we have shown, the much larger space of less-aligned orbits is in 
2734: fact more favorable to disk survival. 
2735: 
2736: However, while the perfect resonance means that the bar 
2737: efficiency $\Psi_{\rm bar}$ is large, the amount of mass in the stellar bar still 
2738: scales as $1-\fgas$, so the burst fraction should vary as 
2739: $\fgas\,(1-\fgas)$ in our simple model (i.e.\ we would still expect 
2740: that a $100\%$ gas disk would have 
2741: no stellar bar, hence no burst, as we have seen for more representative orbits
2742: in Figure~\ref{fig:fgas.fsb}). In fact, though, we typically find 
2743: in these cases that the 
2744: burst fraction seems to scale as $f_{\rm burst}=\fgas$ all the way to high values of $\fgas$ -- 
2745: in short, almost all the gas always bursts -- 
2746: there is no suppression by a $1-\fgas$ factor as would be expected 
2747: if the stellar bar were doing the torquing. 
2748: 
2749: The reason for this is simple -- again, the orbits here are perfectly coplanar 
2750: and in resonance; so this is the one case where the secondary 
2751: galaxy as a whole can directly act as an efficient torque on the gas. In short, 
2752: because the systems are perfectly coplanar and in a resonant orbit, the 
2753: entire secondary galaxy (all baryons and dark matter within the 
2754: stellar $R_{e}$) acts directly to introduce a non-axisymmetric potential 
2755: perturbation (the secondary itself plays the role of the bar). 
2756: So because of this, to an even greater extreme than our scalings for 
2757: more general orbits would predict, this narrow range of orbits is pathological and biased 
2758: {\em against} disk formation. However, understanding why this is the 
2759: case, we can check and explicitly show that it is not so for  
2760: more general orbits, even nearly prograde-prograde orbits (such as case {\bf e}) -- 
2761: in all those cases, even those just slightly out of coplanar resonance, the 
2762: stellar bar is indeed the primary source of torque, and our assumptions 
2763: are justified. This example therefore nicely illustrates what the consequences would 
2764: be if our fundamental assumptions were not true, as well as showing why 
2765: they are in fact true for non-pathological cases. 
2766: 
2767: Another pathological case of interest is one in which the disks are $100\%$ 
2768: gas at the time of merger. Here, as we have said, our simple model 
2769: predicts no starburst or angular momentum loss. In practice, there will still 
2770: be some loss of angular momentum owing to direct cancellation in e.g.\ shocks 
2771: between the disks; but as discussed in \S~\ref{sec:form.major:angloss}, there will also be 
2772: the possibility of some gain owing to the angular momentum of the merger. 
2773: In fact, over the range in mass ratios $\mu\sim 0.1-1$, for a range of typical 
2774: impact parameters $b\sim0.5-5$, the expected final specific angular momentum 
2775: from after cancellation is approximately equal to the initial specific angular 
2776: momentum of the primary (with $\sim20\%$ scatter). Cancellation is therefore 
2777: inefficient. A random distribution of orbits might negate $\sim20\%$ of 
2778: the angular momentum in $\sim$ half the systems merging, but will leave 
2779: $\sim 80-100\%$ of the disk intact. Even these cancellations, we find in detail, do not 
2780: generally yield a starburst in the same manner as a merger-induced bar, 
2781: but lead to moderate disk contraction (and an equal number of mergers 
2782: will scatter towards the opposite sense leading to disk expansion, keeping a mean 
2783: specific angular momentum that is constant). They do not cause a starburst because, 
2784: if two random parcels or streams of gas shock and lose angular momentum, 
2785: the alignment and relative momenta would have to be near-perfect for them 
2786: to lose, say $95\%$ of the angular momentum and fall all the way to the 
2787: central $\sim 100$pc where a nuclear starburst would occur. Rather, they will lose 
2788: some fraction of order unity of their angular momentum, fall in to a radius smaller 
2789: by a factor $\sim2-3$ (but not to very small radii), and continue to orbit. Without the 
2790: bar that can continuously drain angular momentum, the true burst is indeed 
2791: inefficient. 
2792: 
2793: Although we show in \S~\ref{sec:model.feedback} that the physics of interest 
2794: are generally independent of feedback prescriptions, 
2795: there are some pathological feedback regimes. These are discussed in 
2796: detail in \citet{cox:winds}; here, we outline the pathological behavior. 
2797: If e.g.\ starburst-driven winds are implemented with extreme efficiencies 
2798: $\dot{M}_{\rm wind}\gg \dot{M}_{\ast}$ and with 
2799: moderate to large velocities $\gtrsim200\,{\rm km\,s^{-1}}$, then there is no definable 
2800: ``starburst'' in the simulations any more, even when the 
2801: gas loses angular momentum -- indeed, it becomes almost impossible 
2802: to trigger starbursts by {\em any} mechanism. This is because the feedback 
2803: is so extreme that any parcel of gas that begins forming stars above some 
2804: threshold rate is immediately blown apart and drives away all the surrounding 
2805: gas. However, observations suggest that these cases are almost certainly not relevant -- 
2806: observationally inferred mass-loading factors of winds are well below 
2807: the predicted threshold where we see this behavior \citep[see e.g.][]{veilleux:winds,
2808: martin99:outflow.vs.m,martin06:outflow.extend.origin,
2809: erb:lbg.metallicity-winds,sato:outflow.hosts}, 
2810: and moreover the ubiquity of starbursts and recent starburst remnants in 
2811: observed gas-rich major mergers 
2812: \citep[e.g.][]{soifer84a,soifer84b,scoville86,sargent87,sargent89} 
2813: implies that feedback, 
2814: while still potentially efficient, is not able to ``self-terminate'' a starburst 
2815: before it even begins \citep[this is in fact directly confirmed in 
2816: observations of outflows in ongoing massive, merger-induced 
2817: starbursts; see e.g.][]{martin05:outflows.in.ulirgs}. 
2818: A similar pathology can appear if we include extreme 
2819: coupling of black hole feedback to the galaxy gas (e.g.\ allowing 
2820: $100\%$ of the BH accretion energy to couple efficiently), but this is also 
2821: ruled out observationally, both for the arguments above \citep[starbursts 
2822: exist, and the winds seen are not so enormous; 
2823: see the discussion in][]{cox:winds,
2824: hopkins:lifetimes.methods,
2825: hopkins:lifetimes.interp,hopkins:lifetimes.obscuration,
2826: hopkins:lifetimes.letter,hopkins:qso.all}, 
2827: and because such a prescription 
2828: yields black hole masses orders-of-magnitude discrepant from the 
2829: observed \citep{FM00,Gebhardt00} $M_{\rm BH}-\sigma$ relation 
2830: \citep[see e.g.][]{hopkins:bhfp.theory,hopkins:bhfp.obs}
2831: 
2832: 
2833: 
2834: 
2835: 
2836: \subsubsection{Longer-Lived Perturbations: Relation to Secular Evolution}
2837: \label{sec:model.secular}
2838: 
2839: Thus far, we have focused on activity during the merger, roughly defined 
2840: as the short timescale $\sim10^{8}$\,yr following first passage and 
2841: coalescence. In this regime, we have shown that (for typical conditions), 
2842: the dominant source of angular momentum loss is the torque on 
2843: gas from stars in the same disk. However, it is well known from 
2844: studies of isolated barred galaxies \citep[e.g.][]{weinberg:bar.dynfric,
2845: hernquist:bar.spheroid.interaction,
2846: friedli:gas.stellar.bar.evol,
2847: friedli:gas.bar.ssp.gradients,
2848: athanassoula:bar.halo.growth,
2849: athanassoula:bar.vs.concentration,weinberg:bar.res.requirements,
2850: kaufmann:gas.bar.evol,foyle:two.component.disk.evol.from.bars} that 
2851: a long-lived bar (regardless of whether the bar is purely 
2852: stellar or purely gaseous) will exchange angular momentum with 
2853: itself (or e.g.\ gas/stars further out in the disk) and the 
2854: dark matter halo, allowing for further angular momentum loss 
2855: and building a central bulge or ``pseudo''-bulge \citep{patsis:gas.flow.in.bars,
2856: athanassoula:bar.evol.in.int,
2857: mayer:lsb.disk.bars,berentzen:gas.bar.interaction}. 
2858: Here, we discuss 
2859: the relation of this process to what we have described in 
2860: our merger simulations: in general, we find that it (while potentially 
2861: very important for the long-term evolution of the disk and bulge 
2862: masses and structure) is a second-order effect within the merger 
2863: itself, and on longer timescales is more appropriately considered 
2864: an independent, secular evolution process (despite being initially 
2865: triggered by a merger), whose study is better described in 
2866: simulations of idealized and long-lived bars. 
2867: 
2868: As discussed in \S~\ref{sec:model.orbit}, there is a limit to how far the analogy to 
2869: barred galaxies can be drawn. Recall, we use the term ``bar'' more generally to 
2870: represent a quadrupole moment or non-axisymmetric distortion 
2871: in the stellar disk: it does not necessarily (and, especially after second passage, 
2872: usually does not) morphologically resemble isolated barred spirals 
2873: and may not even have an $m=2$ mode structure. Critically, the distortion is 
2874: driven externally by the gravitational perturbation of the secondary orbit -- it is 
2875: not the result of an instability within the primary. As we note in \S~\ref{sec:model.orbit}, 
2876: this already gives rise to a couple of important distinctions: because the distortion is 
2877: driven by the orbital motion of the secondary, it has a characteristic frequency 
2878: (and corresponding radius) internal to which angular momentum loss is very 
2879: efficient (determined entirely by the gravitational properties and relative 
2880: motions of the systems, not 
2881: the subtleties of their internal orbital structure), 
2882: {\em regardless} of properties of the primary (e.g.\ gas phase structure, 
2883: feedback, etc.) that might otherwise make the system more or less stable to 
2884: the development of internal instabilities. 
2885: 
2886: It is straightforward to estimate the relative importance (over a 
2887: short timescale after the merger) of angular momentum loss from 
2888: the gas to the shared stellar bar/distortion induced by the merger, versus 
2889: that to itself and the dark matter halo (the standard secular scenario, which 
2890: other than the initial driving in the merger, will {\em not} be driven by the 
2891: relative gravitational motions but by the more standard bar stability and 
2892: spin-down criteria). 
2893: Approximating the gas as a rigid, thin bar of mass 
2894: $M_{\rm bar,\, gas}\approx f_{\rm gas}\,M_{\rm bar}$ and 
2895: radius $R_{\rm bar}\sim R_{d}$, we can estimate 
2896: the specific torque from the remaining gas disk 
2897: and halo in the dynamical friction limit, following 
2898: \citet{weinberg:bar.dynfric} \citep[for more detailed solutions, which ultimately 
2899: give similar results, see][]{hernquist:bar.spheroid.interaction,
2900: athanassoula:bar.slowdown,weinberg:bar.res.requirements}: 
2901: ${\rm d}j/{\rm d}t = -4\,\pi\,\alpha\,G^{2}\,M_{\rm bar,\, gas}\,\rho(R_{d})\,
2902: v_{\rm bar}^{-2}$, where $\rho(R_{d})$ is the background density and 
2903: $\alpha\sim1$ is a numerical constant (depending on the exact shape of the bar, potential, 
2904: and phase-space distribution of the background). If the ``background'' is 
2905: a \citet{mestel:disk.profile} disk or isothermal sphere, this becomes 
2906: $-2\,\alpha\,G\,M_{\rm bar,\, gas}/R_{d}$. Compare this to our 
2907: Equation~(\ref{eqn:bar2}) for the instantaneous 
2908: torque on the gas bar from the stellar bar: 
2909: $-G\,M_{\rm bar,\, \ast}/(R_{d}\,\sqrt{\sin^{2}{\barangle}+\tilde{H}^{2}})$. 
2910: Removing the common factors, the gas/halo torque 
2911: goes as $\sim f_{\rm gas}$, whereas that from the 
2912: stellar bar goes as $\sim (1-\fgas)/\sqrt{\sin^{2}{\barangle}+\tilde{H}^{2}} 
2913: \sim (1-\fgas)/\barangle$ (because $\barangle\sim\tilde{H}\ll 1$). In short, 
2914: the torque from the gas disk and halo goes as $f_{\rm gas}$ because it is a 
2915: second-order resonance effect (amplified 
2916: and trading off with the gas bar), whereas the stellar bar strength goes 
2917: as the stellar mass fraction $(1-\fgas)$, but boosted by a factor 
2918: $\sim1/\barangle$ representing the small angle of offset between the two 
2919: bars -- i.e.\ the stellar bar is in much closer spatial proximity (in particular 
2920: in spatial alignment in the disk plane) to the gas bar. 
2921: 
2922: This simple comparison gives a reasonable quantitative prediction of 
2923: the relative torques exerted by the halo and stellar disk in our 
2924: simulations. 
2925: Essentially, we have just re-derived the 
2926: well-known fact that the timescale for a bar to damp its own angular momentum 
2927: via resonant interactions with itself and/or the halo is some number ($\sim$ a few) bar 
2928: rotational periods \citep{athanassoula:bar.vs.concentration,
2929: athanassoula:bar.slowdown, 
2930: weinberg:bar.res.requirements,
2931: kaufmann:gas.bar.evol} 
2932: (each bar rotational period being $\sim1-2$ times 
2933: the disk rotational period), whereas in the typical mergers the gas is drained of 
2934: angular momentum by the much stronger local torques on a timescale much 
2935: shorter than an orbital time, allowing it to more or less free-fall into the 
2936: galactic center. Comparing these timescales gives a similar ratio of 
2937: torque strengths. 
2938: Obviously, 
2939: as $\fgas\rightarrow1$, the torque from the halo must eventually 
2940: dominate, but this will not happen until 
2941: $f_{\ast}=(1-\fgas)\lesssim \barangle\sim 0.1$ (given typical bar lags of 
2942: $\sim$ a few degrees or the ratio of the timescales above). 
2943: 
2944: In practice, such a situation is somewhat contrived (it is very difficult to 
2945: maintain a disk with a true $\gtrsim90\%$ gas fraction), and unlikely to 
2946: be of broad cosmological relevance (we have no simulations in this regime 
2947: with which to compare, in fact, because even initially $100\%$ gas disks with 
2948: low star formation efficiencies will be $\lesssim 80\%$ gas by the time of 
2949: the actual merger). However, our Equation~(\ref{eqn:full.equation}) can be trivially 
2950: modified to include these effects: the $(1-\fgas)$ term should be replaced 
2951: with a more appropriate $(1-\fgas+\epsilon_{h})$, where 
2952: $\epsilon_{h}\sim \barangle \sim0.1$ represents the contribution of 
2953: angular momentum 
2954: loss to the halo and outer gas disk during the merger. This exchange of angular momentum, 
2955: therefore, sets some minimum 
2956: bulge mass (with mass fraction $\sim10\%$) that would form even in a 
2957: pure-gaseous disk merger. 
2958: 
2959: By comparing the relative {\em instantaneous} amplitude of the torques from 
2960: the halo/gas and stellar disk, we are comparing how important they each are 
2961: in the loss of angular momentum from the gas over the same (relatively 
2962: short) merger timescale. More important is the fact that, in a gas-rich case 
2963: where the distortion to the stellar distribution may be an inefficient torque, 
2964: the gas bar could be long-lived and continue to lose angular momentum 
2965: over longer timescales. It is not necessarily clear that this would 
2966: happen, however -- a number of studies suggest that 
2967: gas and stellar bars become self-damping once a central mass concentration 
2968: (i.e.\ a nuclear starburst triggered by gas inflows, in this case) 
2969: is in place with a mass fraction larger than a few percent 
2970: \citep{bournaud:gas.bar.renewal.destruction,
2971: berentzen:bar.destruction.in.int,
2972: berentzen:self.damping.tidal.bar.generation,
2973: berentzen:gas.bar.interaction,
2974: athanassoula:bar.vs.cmc} 
2975: \citep[but see also][]{kaufmann:gas.bar.evol}. In such a case, we again arrive at the conclusion 
2976: that these processes set a minimum bulge mass from a large 
2977: bar-inducing perturbation, but do not dominate the 
2978: creation of much larger bulges in mergers. 
2979: 
2980: Regardless of this effect, it is 
2981: not clear that a bar can survive a substantial merger: 
2982: recall, the distortion following second passage 
2983: and coalescence resembles a bar only in that it introduces a rotating 
2984: quadrupole distortion in the disk potential (allowing us to describe it as a 
2985: ``bar'' for analytic convenience), not necessarily in its structure or longevity (it 
2986: does not necessarily share the orbital ``pileup'' that allows a bar to survive), 
2987: and moreover the actual coalescence of the galactic nuclei will disturb any 
2988: bar structure that may be present. Quantitatively, we find our remnants 
2989: rarely have significant long-lived 
2990: $m=2$ modes in the stars or gas -- the mode amplitude tends to damp 
2991: after merger on a timescale $\lesssim10^{8}$\,yr (i.e.\ the free-fall or dynamical 
2992: time, much slower than the typical significant number of orbital times 
2993: for standard bar self-braking). This is similar to the conclusions in the bar 
2994: studies of e.g.\ \citet{bournaud:gas.bar.renewal.destruction} and 
2995: \citet{berentzen:gas.bar.interaction}, who find that the combination of the 
2996: formation of a bulge/small central mass concentration from gas inflows 
2997: and the disturbance/heating to the bar itself in interactions prevents even 
2998: gas-rich systems from maintaining or very rapidly re-forming a bar 
2999: after a significant merger \citep[as opposed to a fly-by passage, which 
3000: may more efficiently induce long-lived bars; see e.g.][]{berentzen:self.damping.tidal.bar.generation}. 
3001: That is not to say a bar 
3002: may not form in the re-formed remnant disk, but such a bar would arise in a 
3003: standard secular fashion, and should be considered in the context of 
3004: the long-term secular evolution of the merger remnant. 
3005: 
3006: If, however, the potential distortion survives the merger to form a 
3007: stable bar, it can certainly be important to the long-term evolution of the system 
3008: and buildup of the bulge. However, this case is outside the scope of this paper, 
3009: and should be more appropriately considered as subsequent evolution of 
3010: the remnant (albeit with an initially merger-induced bar). This is because the 
3011: timescale for the bar to lose angular momentum and contract is some number 
3012: of rotational periods -- so the gas losing angular momentum will slowly 
3013: spiral inwards in some number of orbital periods (turning into stars and 
3014: possibly being ejected by feedback as it does so), rather than free-falling 
3015: into a central burst in a time much less than an orbital period. 
3016: The end result of such angular momentum loss can resemble a bulge 
3017: \citep{mayer:lsb.disk.bars,debattista:pseudobulges.a}, 
3018: although the expectation of rotational support and ``disky-ness'' 
3019: in the material lead to it more likely being a ``pseudo-bulge'' typical 
3020: of secular processes \citep{combes:pseudobulges,kuijken:pseudobulges.obs,oniell:bar.obs,
3021: kormendy.kennicutt:pseudobulge.review,athanassoula:peanuts}. 
3022: Depending on e.g.\ details of the equation of 
3023: state, feedback, and rotational support of the gas disk, it may also amount to 
3024: steady disk contraction \citep{debattista:pseudobulges.b} or emergence of a two-component disk 
3025: \citep{kaufmann:gas.bar.evol,
3026: foyle:two.component.disk.evol.from.bars}. A number of effects will be important 
3027: in this regime, including the effects of feedback in pressurizing the disk and 
3028: smoothing out substructure, and the role of accretion and mergers 
3029: in rebuilding the disk as such evolution continues (since it is occurring on 
3030: timescales $\sim$ several Gyr, comparable to the characteristic timescales 
3031: for new accretion and mergers). 
3032: 
3033: These effects make it difficult to predict the net effect of 
3034: such evolution. For example, if in a pure gas merger 
3035: of mass ratio $\mu$ the specific angular momentum (on average) is 
3036: increased (by addition of specific angular momenta plus orbital angular momentum) 
3037: by an amount $\sim \epsilon_{m}\,\mu\,j_{\rm disk}$, but then the induced 
3038: bar (of amplitude $\sim \mu$) loses its angular momentum 
3039: ($\sim \mu j_{\rm disk}$) on a timescale $\sim N\,t_{\rm rot}$, 
3040: the sequence of mergers and induced bars compete (given the 
3041: cosmologically expected timescale $\approx \mu\,t_{H}$ between 
3042: mergers of mass ratio $\mu$ in \citet{fakhouri:halo.merger.rates}, and 
3043: that for a disk of mass fraction $m_{d}$ relative to the halo, 
3044: $t_{\rm rot}\sim m_{d}\,t_{H}$, one obtains 
3045: ${\rm d}j/{\rm d}t\sim j_{\rm disk}/t_{H}\,[\epsilon_{m} - \mu/N\,m_{d}]$ -- i.e.\ 
3046: more major events will tend to lead to angular momentum loss 
3047: in gas, whereas the net effect of very minor mergers and smooth 
3048: gas accretion, even where it induces instabilities, may be to 
3049: ``spin up'' the disks). 
3050: 
3051: As discussed in \S~\ref{sec:model.massratio}, there is also an interesting 
3052: regime of parameter space, namely minor mergers with 
3053: mass ratios $\sim$1:20-1:10 or so, in which the characteristic merger 
3054: timescales and secular/internal evolution timescales are
3055: comparable. The secondary may be large enough to induce a significant 
3056: bar/non-axisymmetric response, but the merger/dynamical friction time 
3057: may be sufficiently long that the primary could respond almost as if in isolation for 
3058: several orbital periods. 
3059: In such a case it becomes less clear whether the 
3060: merger or the secular response of the disk is ultimately the dominant 
3061: driver of evolution (and the answer probably depends on e.g.\ the exact orbital 
3062: parameters and stability properties of the disk, and may be
3063: sensitive to feedback, 
3064: the gas phase structure and pressure support, and detailed halo structure).  
3065: In any event, it is clear that these processes 
3066: require study in a more complete cosmological context, and 
3067: can contribute significantly to the bulge population (especially 
3068: in less bulge-dominated galaxies, below the typical thresholds 
3069: we simulate) over a Hubble time of evolution. However, 
3070: although the bar itself may be triggered in the merger, the nature of the 
3071: relative strength of the interaction and characteristic timescale for angular momentum 
3072: loss make it not a violent process associated with the merger itself, but rather a 
3073: secular process that should be considered more analogous to bars in non-merging 
3074: systems. 
3075: 
3076: 
3077: 
3078: 
3079: \breaker
3080: \section{Application to Semi-Analytic Models}
3081: \label{sec:prescriptions}
3082: 
3083: Our results clearly have potential uses as prescriptions for 
3084: analytic and semi-analytic models of galaxy formation. 
3085: Here, we summarize and give some simple 
3086: recommendations for these applications. 
3087: 
3088: When a merger is identified in a semi-analytic model, the two key quantities 
3089: we can predict here are the mass fraction of the disk that is destroyed 
3090: (violently relaxed into a bulge) and the 
3091: fraction of the cold gas in the disk that will lose angular momentum 
3092: and contribute to the bulge by forming a compact starburst. 
3093: 
3094: First, the stellar disks: in a merger of secondary mass $M_{2}$ with 
3095: primary mass $M_{1}$, the secondary is destroyed (adding $M_{2}$ 
3096: to the bulge) and the mass within a radius enclosing $\approx M_{2}$ 
3097: in the primary is violently relaxed. If the primary were pure 
3098: disk, this would add $2\,M_{2}$ to the bulge. However, one can imagine 
3099: the limit where the primary is entirely bulge-dominated inside that 
3100: radius (with the stellar disk dominant only at much larger radii) 
3101: -- then the violent relaxation of the merger will act primarily to 
3102: heat existing bulge stars, and only a mass $1\,M_{2}$ will be added to the bulge. 
3103: Obviously, its also true that if the total disk mass of $M_{1}$ is less than 
3104: $M_{2}$, then that is a maximum to how much can be added to the bulge 
3105: (i.e.\ really ${\rm MIN}(M_{2},\,f_{\rm disk}\,M_{1})$ is added). 
3106: For most purposes, this factor $2$ possible range is not critical 
3107: in the semi-analytic models, and picking a constant (effective mean) 
3108: fraction $(0-1) \times M_{2}$ to violently relax in the primary in all mergers 
3109: is acceptable. However, if more detail is desired, an estimate of the 
3110: mass profiles of bulge plus disk components in the primary can be used to 
3111: determine the total primary 
3112: disk mass within a radius enclosing a mass $\approx M_{2}$, and 
3113: then that will be the fraction violently relaxed. For a \citet{hernquist:profile} 
3114: bulge and exponential disk obeying roughly the observed size-mass 
3115: relations from \citet{shen:size.mass}, the primary disk mass that should be 
3116: violently relaxed in a merger with mass $M_{2}$ can be 
3117: approximated as $f_{\rm disk,\,\ast}\,M_{2}/(1+[M_{1}/M_{2}]^{\alpha})$, 
3118: where $f_{\rm disk,\,\ast} \equiv (1-f_{\rm gas})\,(1 - f_{\rm bulge})$ is the  
3119: mass fraction of the stellar disk (relative to the baryonic galaxy) and 
3120: the term $(1+[M_{1}/M_{2}]^{\alpha})$ is a correction for e.g.\ the 
3121: relative sizes of the two components as a function of mass ratio and 
3122: other properties (for the assumptions above, 
3123: $\alpha\approx 0.3-0.6$, depending on the details of the disk 
3124: mass profile). 
3125: 
3126: Two clarifications should be emphasized. First, these derivations only 
3127: apply to cases where the secondary is sufficiently massive that it survives 
3128: to merge with the center of the primary. If the secondary is destroyed or 
3129: shredded by tidal forces before merger, then it will not add either its own 
3130: mass or any violently relaxed mass to the bulge. This generally occurs in 
3131: the limits of smaller mass ratios ($\lesssim1:10$, which we have 
3132: considered), but is included in some models. Second, for most applications, 
3133: the masses $M_{1}$ and $M_{2}$ should be taken to be the 
3134: {\em baryonic} masses {\em within} the galaxies (stars in the galaxy and cold gas -- not 
3135: diffuse stellar halo or pressure-supported hot gas in the extended halo). 
3136: This is how we have defined our models and fits to our simulations (although those 
3137: simulations do include dark matter and extended halo gas and stars) throughout. 
3138: The halos are much more extended, and much lower density, so they merge and mix 
3139: more efficiently, and do not strongly participate in the central 
3140: violent relaxation process that defines the bulge. 
3141: Moreover, there can be a wide range in halo masses for galaxies of similar mass -- 
3142: but most of these halos are large and often independent substructures that should 
3143: not be used to define e.g.\ the mass ratios of merging encounters. 
3144: What dark matter is 
3145: carried in with the galaxies is that enclosed in their stellar 
3146: effective radii $R_{e}$, which tends to track the baryonic mass much more closely 
3147: than, say, the total halo mass, so it is not a bad proxy to still define mass ratios, 
3148: etc.\ in terms of the baryonic masses. 
3149: 
3150: Next, in such an encounter, our analysis provides a means to estimate 
3151: the fraction of the cold gas mass in the pre-merger stellar disks that should 
3152: lose angular momentum and be funneled into a nuclear starburst. 
3153: The cold gas inside some radius $R_{\rm gas}/\scalelen$ will participate in this starburst, 
3154: where $R_{\rm gas}$ is given by Equation~(\ref{eqn:full.equation}). 
3155: There are five variables that go into this equation: (1)
3156: $f_{\rm gas}$, which we define as the mass fraction {\em of the disk} 
3157: that is in cold (rotationally supported) gas (i.e.\ if the disk is $50\%$ cold gas, then regardless of the 
3158: bulge fraction of the galaxy, $f_{\rm gas}=0.5$. Note that we only care about 
3159: cold, rotationally supported gas. Hot gas in the galactic halos can cool, of course, 
3160: and form new stars, but that process is relatively independent of the merger, 
3161: and is not related to angular momentum loss (also because the hot gas is 
3162: pressure-supported, it is fairly resistance to significant redistribution in the merger, 
3163: and if anything will tend to be shocked to even higher temperatures rather than 
3164: forming stars in the short-lived merger). 
3165: (2) $f_{\rm disk}=(1-f_{\rm bulge})$, the total (gas plus stellar) baryonic mass 
3166: fraction of the disk. (3) $\mu\equiv M_{2}/M_{1}$, the mass ratio of the 
3167: merger (defined as above). (4) $\theta$ and $b$, equivalently the 
3168: orbital parameters of the merger. As discussed in \S~\ref{sec:model.orbit}, 
3169: for cases that will merge the appropriate limit is $b\rightarrow0$, since 
3170: most of the action will occur on the final merging passages after the 
3171: angular momentum is removed. We discuss what should be adopted 
3172: for the orbital inclination $\theta$ below. 
3173: (5) $\scalelen$, the scale length of the disk stars. 
3174: 
3175: In any semi-analytic or analytic model, variables (1)-(3) should be well-known 
3176: beforehand. Given some choice of orbital parameters and an assumed 
3177: mass distribution of the disk, it is trivial then to translate 
3178: Equation~(\ref{eqn:full.equation}) into a fraction of the gas that will 
3179: burst. Because orbital parameters are generally undetermined in these 
3180: models, there are two choices for the assumed orbital inclination $\theta$. 
3181: First, one could draw a random value of $\theta$ for each merger 
3182: (uniformly sampling in $\cos(\theta)$ as appropriate for an isotropic 
3183: orbit distribution), and use 
3184: Equations~(\ref{eqn:full.eqn.orbit.1})-(\ref{eqn:full.eqn.orbit.2})
3185: for each merger. Alternatively, we can average over 
3186: a random distribution of orbits and quote an ``effective'' 
3187: orbital dependence $F(\theta,b)$ for Equation~(\ref{eqn:full.eqn.orbit.1}). 
3188: Note that this is only strictly appropriate if all disks have the same mass 
3189: profiles and those are such that the enclosed mass is linear in $R/\scalelen$ 
3190: (otherwise the appropriate average would have to be weighted by 
3191: other terms such as $(1-\fgas)$ in Equation~\ref{eqn:full.equation}). In 
3192: any case doing so yields an effective mean orbital dependence 
3193: $F(\theta,b)\approx1.2$. 
3194: 
3195: The only remaining issue is the assumed mass profile of the disk. Here, models 
3196: have some freedom. As we have emphasized, the exact profile (e.g.\ choice 
3197: of exponential disk or some other profile) does not have a dramatic effect. 
3198: What is important, however, is the assumption of how the gas is distributed 
3199: relative to the stars. Recall, Equation~(\ref{eqn:full.equation}), with the variables 
3200: above inserted, gives that the gas inside some radius 
3201: $R_{\rm gas} = x\,\scalelen$ (where $x$ is a constant depending on those 
3202: variables, and $\scalelen$ is a characteristic scale length of the {\em stellar} disk) 
3203: should lose angular momentum and participate in the burst. 
3204: Given a gas mass profile $M_{\rm gas}(R/R_{e,\, {\rm gas}})$, in terms of a 
3205: characteristic gas disk scale length $R_{e,\, {\rm gas}}$, this gives 
3206: the gas mass that bursts, $M_{\rm gas}(x\,\scalelen/R_{e,\, {\rm gas}})$. 
3207: For our simulations, we have generally assumed 
3208: (and can see that it is a good approximation) that the gas and stellar disks 
3209: initially trace one another ($\scalelen\approx R_{e,\, {\rm gas}}$). 
3210: However, since our derivation and Equation~(\ref{eqn:full.equation})
3211: show that it is the gas inside some fraction of the {\em stellar} disk half-mass 
3212: radius $\scalelen$ that loses angular momentum, then if the gas is e.g.\ much 
3213: more extended than the stars, a lower gas fraction will end up in the burst. 
3214: We discuss this in \S~\ref{sec:model.feedback}, and consider how such situations may 
3215: in fact arise owing to e.g.\ supernova feedback blowing gas out to large radii. 
3216: Semi-analytic models therefore have some freedom in adopting these prescriptions 
3217: based on their implicit assumptions about feedback and disk formation, encapsulated 
3218: effectively in our prescriptions as the ratio of the 
3219: stellar to gas disk scale lengths $\scalelen/R_{e,\, {\rm gas}}$. Lacking 
3220: some detailed model for both values in the semi-analytic models, 
3221: a constant value $\sim1$ is probably a good choice (with the exact choice 
3222: reflecting implicit assumptions about feedback and outer disk formation). 
3223: 
3224: Those prescriptions define both the violently relaxed and starburst components 
3225: induced in mergers of arbitrary mass ratios, gas content, and orbital parameters. 
3226: If desired, appropriate scatter (a factor $\sim2$) can be added to both 
3227: quantities, reflecting the scatter we see between various numerical 
3228: realizations (although it should still be ensured that, with scatter, the implied 
3229: violently relaxed and burst fractions are within the sensible physical limits). 
3230: 
3231: Although not discussed here, in \citet{hopkins:cusps.ell,
3232: hopkins:cores,hopkins:cusps.fp,hopkins:cusps.evol,
3233: hopkins:cusps.mergers} we consider 
3234: how the sizes and velocity dispersions of these components should scale, 
3235: and we refer to those papers for detailed analysis of those results. Briefly, 
3236: we note that in the absence of dissipation, it is straightforward to calculate the 
3237: size of the dissipationless component (the violently relaxed stars from the 
3238: pre-merger stellar disk), given phase space and energy conservation. 
3239: Roughly, this implies that the component will have the same (modulo 
3240: projection effects since it transforms from a disk to a sphere) 
3241: scale radius as the disk (or radius within the disk) from which it forms. 
3242: Again, conservation of energy in subsequent dissipationless re-mergers, 
3243: along with the assumption of preserved profile shape 
3244: \citep[which we demonstrate is reasonable in][]{hopkins:cores} yields 
3245: the evolution in subsequent events of these radii (in a 
3246: re-merger of masses $M_{1}$ and $M_{2}$, the dissipationless 
3247: bulge component will have final size 
3248: $R_{f}/R_{1} \approx (1+\mu)^{2}/(1+\mu^{2}\,R_{1}/R_{2})$). 
3249: Dissipation complicates this -- it is possible to solve separately for the 
3250: size of the dissipational component by allowing for energy loss 
3251: in the collision followed by (after angular momentum loss) collapse 
3252: to a self-gravitating limit, and then subsequently evolve the 
3253: component as a dissipationless body, added with the violently relaxed 
3254: components to give a total bulge effective radius. Fortunately, 
3255: \citet{covington:diss.size.expectation} 
3256: perform such an exercise and we show in \citet{hopkins:cusps.ell} 
3257: that their results can be conveniently 
3258: approximated (in both an analytic manner and as a fit to the results of 
3259: numerical simulations) by the scaling: 
3260: $R_{e}({\rm bulge}) = R_{e}(f_{\rm sb}=0)/(1+f_{\rm sb}/f_{0})$, 
3261: where $f_{\rm sb}$ is the total mass fraction of the bulge/spheroid which 
3262: originally formed dissipationally (as opposed to being violently relaxed), 
3263: $R_{e}(f_{\rm sb}=0)$ is the radius the system would have if purely 
3264: dissipationless (calculated as described above), and $f_{0}\approx 0.25-0.35$ 
3265: is a constant. 
3266: 
3267: Our modeling could also be applied in the manner described in \S~\ref{sec:model.flyby} 
3268: to fly-by (non-merging) encounters, but we caution that these 
3269: are usually ill-defined in semi-analytic models (and if adopted, the 
3270: cautions in \S~\ref{sec:model.flyby} about the appropriate meaning of the impact 
3271: parameter adopted should be borne in mind). In any case, the rapid suppression 
3272: of bursts with increasing impact parameter means that such cases should 
3273: be relatively unimportant in a representative cosmological ensemble. 
3274: 
3275: 
3276: \breaker
3277: \section{Discussion and Conclusions}
3278: \label{sec:discussion}
3279: 
3280: We have derived a general physical model for how 
3281: disks survive and/or are destroyed in mergers and interactions. 
3282: Our model describes both the dissipational and dissipationless 
3283: components of the merger, and allows us to predict, for a 
3284: given arbitrary encounter, the stellar and gas content of the 
3285: system 
3286: that will be dissipationlessly violently relaxed, dissipationally lose 
3287: angular momentum and form a compact central starburst, or 
3288: survive (without significant angular momentum loss or violent relaxation) 
3289: to re-form a disk. 
3290: We show that, in an immediate (short-term) sense, the amount of stellar 
3291: or gaseous disk that survives or re-forms 
3292: following a given interaction can be understood purely 
3293: in terms of simple, well-understood gravitational physics. 
3294: Knowing these physics, our model allows us to accurately predict the 
3295: behavior in full hydrodynamic numerical simulations across as a function of 
3296: the merger mass ratio, orbital parameters, pre-merger cold gas 
3297: fraction, and mass distribution of the gas and stars, 
3298: in simulations which span a wide range of parameter space 
3299: in these properties as well as prescriptions for gas physics, 
3300: stellar and AGN feedback, halo and initial disk structural 
3301: properties, redshift, and absolute galaxy masses. 
3302: 
3303: The fact that we can understand the complex, nonlinear behavior 
3304: in mergers with this analytic model, and moreover that (for given conditions 
3305: at the time of merger) our results are independent of the details of 
3306: prescriptions for gas physics, star formation, and feedback, owes 
3307: to the fact that the processes that strip angular momentum 
3308: from gas disks and violently relax stellar disks are fundamentally 
3309: {\em dynamical}. 
3310: 
3311: Gas, in mergers, primarily loses angular momentum to 
3312: internal gravitational torques (from the stars in the same disk) 
3313: owing to asymmetries in 
3314: the galaxy induced by the merger (on the close passages 
3315: and final coalescence of the secondary, during which phase 
3316: the potential also rapidly changes, scattering and violently relaxing the 
3317: central stellar populations of the stellar disk).\footnote{We note again 
3318: that although we have described these asymmetries as ``bars'' or 
3319: ``bar-like'' at certain points in this paper, there are a number of properties 
3320: of the non-axisymmetric distortions induced in mergers  
3321: (discussed in \S~\ref{sec:model.orbit} and \S~\ref{sec:model.secular}) that make them -- 
3322: at least over the short relaxation timescale of the merger -- dynamically 
3323: distinct from traditional bar instabilities in isolated systems.} 
3324: Hydrodynamic torques and 
3325: the direct torquing of the secondary are second-order effects, 
3326: and inefficient for all but pathological orbits. 
3327: 
3328: Once gas is efficiently drained of angular momentum, 
3329: there is little alternative but for it to fall to the center of the galaxy and 
3330: form stars, regardless of the details of the prescriptions for star 
3331: formation and feedback -- we show that even strong supernova-driven 
3332: winds (with mass loading efficiencies several times the star formation rate 
3333: and wind mass-loading velocities well above the halo escape velocity) 
3334: do not significantly effect our conclusions. Such processes, after all, 
3335: can blow out some of the gas, but they cannot fundamentally alter the 
3336: fact that cold gas with no angular momentum will be largely 
3337: unable to form any sort of disk, or the fact that a galaxy's worth of 
3338: gas compressed to high densities and small radii 
3339: will inevitably form a large mass in stars. 
3340: 
3341: For these reasons, many processes and details that are important
3342: cosmologically (systematically changing e.g.\ the pre-merger 
3343: disk gas fractions) -- in some sense setting the initial conditions for 
3344: our idealized study of what happens in mergers -- 
3345: do not alter the basic dynamical behavior within the mergers themselves, 
3346: and therefore do not change our conclusions. 
3347: 
3348: \begin{figure*}
3349:     \centering
3350:     \scaleup
3351:     %\plotone{summary_plot.ps}
3352:     \plotone{f16.ps}
3353:     \caption{Summary of our comparison between simulations and 
3354:     analytic model for the mass of disks in merger remnants as a function of 
3355:     appropriate orbital parameters, merger mass ratio, and pre-merger 
3356:     cold gas content. We plot our model prediction versus the simulation remnant 
3357:     disk fraction for all $\sim400$ full hydrodynamic merger simulations considered 
3358:     in this paper (shown in both a linear and logarithmic scale). 
3359:     Symbols encode some of the parameter studies we consider: 
3360:     orbital parameters, galaxy masses, initial merger redshift, 
3361:     choice of feedback prescription, merger mass ratio, and presence or 
3362:     absence of black holes, as labeled. For each subset of simulations, 
3363:     we sample a wide range in initial and pre-merger gas fractions $\fgas=0-1$. 
3364:     Solid line is a one-to-one relation. 
3365:     In all cases, our predictions agree well with the simulations, with no systematic 
3366:     offsets owing to any of the parameters we have varied. 
3367:     At high $f_{\rm disk}$, our predictions are accurate to an absolute uncertainty 
3368:     $\sim0.05-0.10$ in $f_{\rm disk}$. At low $f_{\rm disk}\lesssim0.1$, our predictions 
3369:     are accurate to a factor $\sim2-3$ (down to $f_{\rm disk}\lesssim1\%$, where 
3370:     it is difficult to reliably identify disks in the remnant). 
3371:     \label{fig:summary}}
3372: \end{figure*}
3373: 
3374: Figure~\ref{fig:summary} summarizes our results for the ensemble of our 
3375: simulations. We compare the fraction of the baryonic galaxy mass in the 
3376: merger remnant 
3377: that is in a surviving post-merger disk to that predicted by our simple model scalings, 
3378: and find good agreement over the entire range in disk and bulge mass 
3379: fractions sampled, with surprisingly small scatter given the complexity of 
3380: behavior in mergers. 
3381: We highlight several of the parameter studies, 
3382: showing that -- for fixed mass ratio, orbital parameters, and gas content 
3383: {\em at the time of the final merger}, none of these choices systematically 
3384: affect our predictions (note that these are not the only parameters varied -- the 
3385: complete list is discussed in \S~\ref{sec:sims}, but it is representative). 
3386: That is not to say they cannot affect them indirectly, 
3387: by e.g.\ altering how much gas is available at the time of merger -- but it 
3388: emphasizes that the processes we model and use to form our 
3389: predictions, the processes that dominate violent relaxation and the loss 
3390: of angular momentum in gas in mergers, are fundamentally dynamical. 
3391: 
3392: This allows us to make robust, accurate physical predictions independent of 
3393: the (considerable) uncertainty in feedback physics and sub-resolution 
3394: physics of the ISM. Regardless of how those physics alter the ``initial'' 
3395: conditions, they do not change basic dynamical processes, 
3396: and so do not introduce significant uncertainties in our model. 
3397: 
3398: In turn, this means that we can use our model to understand just why and how 
3399: feedback is important for the cosmological survival of disks. Why, in short, 
3400: have various works \citep[see e.g.][]{springel:spiral.in.merger,robertson:disk.formation,
3401: governato:disk.formation} concluded that strong feedback is essential for 
3402: enabling disk survival in mergers? 
3403: Our results show that it is not that feedback somehow makes the disk 
3404: more robust to the dynamical torques within the merger, in any 
3405: instantaneous sense. 
3406: These torques, at least within the critical radii where 
3407: the gravitational perturbation from the merger is large and 
3408: in resonance, are sufficiently strong that any reasonable feedback 
3409: prescription is a dynamically negligible restoring force. 
3410: Rather, feedback has two important effects that 
3411: fundamentally alter the conditions in the merger: first, 
3412: it allows the galaxy to retain much higher gas content going into the 
3413: merger. Without feedback from e.g.\ star formation and supernovae 
3414: contributing to heating and pressurizing the ISM and 
3415: redistributing gas spatially, isolated gas-rich disks may be unstable to 
3416: fragmentation. Even if fragmentation is avoided, 
3417: it is well-known that star formation 
3418: in simulations proceeds efficiently under these conditions. This would leave the 
3419: disks essentially pure stars \citep[even for idealized simulations 
3420: beginning with $\sim100\%$ gas disks; see e.g.][]{springel:models} 
3421: by the time of the merger, which guarantees that a major merger 
3422: will inevitably violently relax the stars (this is a simple collisionless 
3423: mixing process, and under such circumstances is 
3424: inescapable). With large gas fractions, however, 
3425: the system relies on stripping angular momentum from the gas to 
3426: form new bulge stars, which in turn relies on internal torques from 
3427: induced asymmetries in the stellar disk. If the gas fractions are sufficiently 
3428: large, there is little stellar disk to do any such torquing, and the 
3429: gas survives largely intact. 
3430: 
3431: Second, feedback from supernovae and stellar winds moves the 
3432: gas to large radii, where it does not feel significant torques from the 
3433: merger. Again, recall that the most efficient torquing is driven 
3434: by the internal stellar disk of the galaxy, and as such is most efficient 
3435: at torquing gas within small radii (this can be thought of as 
3436: analogous to the well-known co-rotation condition for isolated 
3437: disk bars). If star formation-driven feedback has blown much of the 
3438: gas to large radii, then there is little gas inside the radius 
3439: where torques can efficiently strip angular momentum, yielding little 
3440: induced starburst and largely preserving the gas disk at large radii. 
3441: 
3442: Not only can we qualitatively identify these requirements for feedback 
3443: processes, but we can more precisely use our model to set quantitative 
3444: limits on how much gas must be retained and/or the radii it must be 
3445: redistributed to in order to enable disk survival under various 
3446: conditions. This also clearly implies that disks must be able to avoid 
3447: fragmentation and strong local gravitational instabilities when they 
3448: achieve these gas fractions. 
3449: This provides a valuable constraint for feedback models -- 
3450: how those models affect star formation efficiencies, the ``blowout'' of 
3451: gas, and the local hydrodynamic state (effective equations of 
3452: state and phase structure) of ISM gas -- 
3453: and should be useful for calibrating their (still largely 
3454: phenomenological) implementations in both numerical 
3455: and semi-analytic models of galaxy formation. 
3456: 
3457: Our predictions are also of interest in any cosmological model for the 
3458: emergence of the Hubble sequence, since they apply not just to 
3459: disk-dominated galaxies but to small disks in bulge-dominated 
3460: systems. 
3461: We give a number of simple prescriptions for 
3462: application of our conclusions to analytic and semi-analytic models 
3463: of galaxy formation, which can be used to predict 
3464: the distribution of bulge to disk ratios in cosmological ensembles. 
3465: But even without reference to a full such model, a number of 
3466: interesting consequences are immediately apparent. 
3467: 
3468: First, it is a well-known problem that theoretical models systematically 
3469: overpredict the abundance and mass fractions of bulges in 
3470: (especially) low-mass galaxies. This is true even in e.g.\ semi-analytic 
3471: models, which are not bound by resolution requirements and can adopt a 
3472: variety of prescriptions for behavior in mergers. 
3473: However, it is also well-established observationally that disk gas fractions tend 
3474: to be very high in this regime, with large populations of gas-dominated 
3475: disks at $M_{\ast}\ll 10^{10}\,M_{\sun}$ \citep{belldejong:tf,kannappan:gfs,mcgaugh:tf}. 
3476: Our models predict that bulge formation 
3477: should, therefore, be strongly suppressed in precisely the regime 
3478: required by observations. For e.g.\ disks with $M_{\ast}<10^{9}\,M_{\sun}$ where 
3479: observations suggest typical gas fractions $\sim60-80\%$, our results 
3480: show that even a 1:1 major merger would typically yield a remnant with 
3481: only $\sim30\%$ bulge by mass -- let alone a more typical 
3482: 1:3-1:4 mass-ratio merger, which should yield a remnant with $<20\%$ bulge. 
3483: That is not to say that it is impossible to form a bulge-dominated system 
3484: at these masses, but it should be much more difficult than at high masses, 
3485: requiring either unusually gas-poor systems, violent merger histories, or 
3486: rarer merging orbits that are more efficient at destroying disks. 
3487: Our conclusions therefore have dramatic implications for the abundance of 
3488: bulges and typical morphologies and bulge-to-disk ratios 
3489: at low galaxy masses and in gas-rich systems. Low-mass systems, 
3490: when a proper dynamical model of bulge formation in mergers is considered, 
3491: should have lower bulge-to-disk ratios -- by factors of several, at least -- 
3492: than have been assumed and modeled in previous 
3493: theoretical models. 
3494: Whether this alone is sufficient to resolve the discrepancies with the observations 
3495: remains to be seen, but it is clearly of fundamental importance that future 
3496: generations of models incorporate this scaling. 
3497: 
3498: Second, the importance of this suppression owing to gas content in disks 
3499: will be even more significant at high redshifts.
3500: Observations suggest \citep[see e.g.][]{erb:lbg.gasmasses} 
3501: that by $z\sim2$, even systems with masses near $\sim L_{\ast}$ 
3502: ($M_{\ast}\sim 10^{10}-10^{11}\,M_{\sun}$) may have gas fractions as 
3503: high as $\fgas\sim0.6$. In this regime, the same argument as above should apply, 
3504: dramatically suppressing the ability of mergers to destroy disks. 
3505: Moreover, 
3506: since most of the mass density is near $L_{\ast}$, this can change not just 
3507: the behavior in a specific mass regime but significantly suppress the global 
3508: mass density of spheroids, modifying the predicted redshift history of bulge formation. 
3509: (Note that this will not change when {\em stars} form by very much, so it has little or 
3510: no effect on e.g.\ the ages of $z=0$ spheroids). 
3511: 
3512: This redshift evolution may also explain the 
3513: solution to a fundamental problem in reconciling observed disk populations 
3514: with CDM cosmologies. Integrated far enough back in time, every galaxy 
3515: is expected to have experienced a significant amount of major merging. 
3516: In extreme cases, the mass of the system when it had its last such merger 
3517: may be so small that it would not be noticed today, but in general, 
3518: it does not require going far back in redshift (to perhaps 
3519: $z\sim2-4$ before almost every $z=0$ galaxy should have had such a 
3520: merger). How, then, can the abundance of systems with relatively 
3521: small (or even no) visible bulges be explained? Our conclusions here 
3522: highlight at least part of the answer: as you go back in time, 
3523: the gas fractions of systems are also higher, nearing unity. So even though, integrating 
3524: sufficiently far in time, every system has experienced major mergers, 
3525: it is also true that the systems were increasingly gas-rich, and therefore that 
3526: the impact of those mergers was more and more suppressed. Only mergers 
3527: at later times, below certain gas fraction thresholds, will typically destroy disks. 
3528: 
3529: Third, to the extent that bulge formation is suppressed at 
3530: increasing redshifts, the existence of 
3531: an $M_{\rm BH}-M_{\rm bulge}$ relation \citep[e.g.][]{magorrian} implies 
3532: that black hole growth should also be suppressed. Indeed, 
3533: bulge formation is suppressed specifically because gas cannot efficiently 
3534: lose angular momentum in mergers if the systems are gas-dominated -- 
3535: if the gas cannot lose angular momentum efficiently, then it certainly 
3536: cannot efficiently be accreted by the nuclear black hole. 
3537: Since this pertains to gas on the scales of galactic disks, 
3538: it is probably not relevant for the formation of ``seed'' black holes at 
3539: very high redshift, but it will in general inhibit the growth of black holes 
3540: owing to early merging activity. At the same time, of course, 
3541: higher gas fractions in general imply increasing fuel supplies for black 
3542: hole growth, so the effects are not entirely clear, and more detailed 
3543: models are needed to see how this impacts the history of black hole 
3544: growth and quasar luminosity functions. Nevertheless, this may in part 
3545: explain why, above $z\sim2$ (where, for the argument above, these 
3546: effects become important for the global mass density of spheroids), 
3547: the global rate of black hole growth (i.e.\ total quasar luminosity 
3548: density) appears to decline much more rapidly with increasing 
3549: redshift than the star formation rate density \citep[compare e.g.][]{hopkinsbeacom:sfh,
3550: hopkins:groups.qso,hopkins:bol.qlf}. 
3551: 
3552: Fourth, our models imply that a large fraction of bulges and disks 
3553: survive mergers together, rather than being formed entirely separately. 
3554: It is often assumed that classical bulges -- being similar to 
3555: small ellipticals in most of their properties -- were formed initially in 
3556: major mergers, as entirely bulge-dominated systems, and then accreted 
3557: new gaseous and stellar disks at later times. Although nothing in our modeling 
3558: would prevent this from happening, our analytic and simulation results 
3559: generically lead to the expectation that a large (perhaps even dominant) fraction 
3560: of the bulge population did {\em not} form in this manner, but rather 
3561: formed {\em in situ} from minor mergers or less efficient major mergers (in e.g.\ 
3562: very gas-rich systems). 
3563: Observations tracing the evolution of disk components, 
3564: kinematics, and morphology in the last $\sim10\,$Gyr 
3565: increasingly suggest that such co-formation or disk regeneration scenario 
3566: is common \citep[see e.g.][and references therein]{hammer:obs.disks.w.mergers,
3567: conselice:tf.evolution,flores:tf.evolution,puech:tf.evol}.
3568: In short, a system with a mass fraction $\sim0.1-0.2$ in a 
3569: bulge could be the remnant of an early, violent major merger (when the system 
3570: was $\sim0.1$ times its present mass) with a re-accreted disk, or could be 
3571: the remnant of a typical (low to intermediate gas fraction) 1:10-1:5 mass ratio 
3572: minor merger, or could even be the remnant of a gas-rich major merger 
3573: (mass ratio $\lesssim1:3$, if $f_{\rm gas}$ is sufficiently large). 
3574: 
3575: Based on 
3576: a simple comparison of typical merger histories, we would actually expect that 
3577: the minor merger mechanism should be most common, but all may be 
3578: non-negligible. Fundamentally, the physics forming the bulge (torquing the gas within 
3579: some radius owing to internal asymmetries and violently relaxing stars within 
3580: a corresponding radius) are the same in all three cases, and moreover other indicators 
3581: such as their stellar populations will be quite similar \citep[in all cases, the bulge will 
3582: appear old: this is both because the central stars in even present-day disks 
3583: are much older than those at more typical radii, and because in any case star formation 
3584: will cease within the bulge itself, as opposed to the ongoing star formation in the disk, 
3585: and stellar population age estimates are primarily sensitive to the amount of 
3586: recent or ongoing star formation; see e.g.][]{trager:ages}. 
3587: This is also not to say that mergers are the only means of producing 
3588: bulges. Secular evolution of e.g.\ barred disks probably represents 
3589: an increasingly important channel for bulge evolution in later-type 
3590: and more gas-rich systems \citep[see e.g.][]{christodoulou:bar.crit.1,sheth:bar.frac.evol,
3591: mayer:lsb.disk.bars,debattista:pseudobulges.a,jogee:bar.frac.evol,
3592: kormendy.kennicutt:pseudobulge.review,marinova:bar.frac.vs.freq}, and 
3593: may even be related (albeit through longer timescales of ``isolated,'' 
3594: post-merger evolution and different physics) to initial bar formation 
3595: or ``triggering'' in mergers. 
3596: More detailed theoretical 
3597: work and analysis of cosmological simulations is needed to develop observational 
3598: probes that can distinguish between these histories. 
3599: 
3600: Further work is specifically needed to investigate the processes at work in 
3601: minor mergers with mass ratios $\sim$1:10 ($\mu\sim0.05-0.1$), 
3602: which cosmological simulations suggest 
3603: are an important contributor to the growth of disks, especially 
3604: in later-type systems 
3605: \citep{maller:sph.merger.rates,
3606: fakhouri:halo.merger.rates,stewart:mw.minor.accretion}. 
3607: In more minor mergers $\mu \ll 0.1$, the secondaries are sufficiently small and dynamical 
3608: friction times sufficiently 
3609: long that the disk is unlikely to feel significant external perturbations. 
3610: More major mergers $\mu \gtrsim 0.1$, the cases of interest here, 
3611: induce sufficiently large responses in the disk and evolve sufficiently rapidly 
3612: that they can be considered ``merger-dominated'' 
3613: for the reasons in \S~\ref{sec:model.orbit} \&\ \ref{sec:model.secular}. 
3614: But in the intermediate regime, internal amplification of instabilities in a 
3615: traditional secular fashion may occur on a timescale comparable to 
3616: or shorter than the evolution of the secondary orbit, potentially 
3617: leading to a more complex interplay between the two. It is not entirely 
3618: clear whether such a system would remain ``locked'' to the driven 
3619: perturbation, or function as a purely secular system (merely initially 
3620: driven by the presence of the secondary), or some nonlinear combination 
3621: of both. A more detailed comparison of the relevant timescales 
3622: for these processes and their relation to e.g.\ cosmological triggering of 
3623: bars and large-scale non-axisymmetric modes in disks will be the 
3624: subject of future study (in preparation). 
3625: 
3626: 
3627: Our results are also of direct interest to models of spheroid formation in 
3628: ellipticals and S0 galaxies. As discussed in \S~\ref{sec:intro}, it is increasingly 
3629: clear that embedded sub-components -- constituting surviving gaseous 
3630: and stellar disks -- are both ubiquitously observed and critical 
3631: for theoretical models to match the detailed kinematics and isophotal 
3632: shapes of observed systems \citep{naab:gas,
3633: cox:xray.gas,cox:kinematics,robertson:fp,jesseit:kinematics,
3634: hopkins:cusps.ell,hopkins:cusps.fp}. We have developed a model 
3635: that allows us to make specific predictions for how disks survive mergers, 
3636: including both the survival of some amount of the pre-merger stellar disks 
3637: and the post-merger re-formation of disks and rotationally supported 
3638: components from gas that survives the merger without losing most of its 
3639: angular momentum.
3640: 
3641: Figure~\ref{fig:summary} shows that we can extend 
3642: these predictions with reasonable accuracy to surviving rotational systems 
3643: containing as little as $\sim 1\%$ of the remnant stellar mass, comparable to 
3644: small central subcomponents and subtle features giving rise to e.g.\ 
3645: slightly disky isophotal shapes \citep[see e.g.][]{ferrarese:type12,lauer:centers,
3646: mcdermid:sauron.profiles}. Owing to 
3647: the combination of resolution requirements and desire to understand the 
3648: fundamental physics involved, most theoretical studies of these detailed 
3649: properties of ellipticals have been limited to idealized studies of individual 
3650: mergers. Our results allow these to be placed in a more global context 
3651: of cosmological models and merger histories. Moreover, our 
3652: models allow the existence of such features (or lack thereof) to be translated 
3653: into robust constraints on the possible merger histories and gas-richness 
3654: of spheroid-forming mergers. Further, \citet{hopkins:cusps.ell,hopkins:cusps.fp}, 
3655: studied how the dissipational starburst components arising in gas-rich 
3656: mergers are critical to explaining the observed properties and scaling relations 
3657: of ellipticals, and how these components can both be extracted from 
3658: and related to observed elliptical surface brightness profiles. Because both 
3659: the starburst and surviving disks arise from gas in mergers, the combination of 
3660: constraints from the central stellar populations, studied therein, with 
3661: constraints on the survival and/or loss of gas angular momentum in mergers 
3662: studied here, should be able to break some of the degeneracies in e.g.\ 
3663: pre-merger gas fractions and merger histories in order to enable new 
3664: constraints and understanding of spheroid merger histories, and 
3665: new tests of models for spheroid formation in gas-rich mergers. 
3666: 
3667: These points relate to a number of potentially testable predictions of 
3668: our models. These include the in situ formation of bulges from various types of 
3669: mergers, and possible associated stellar population signatures, the 
3670: presence of embedded disks in ellipticals, and how their sizes and mass 
3671: fractions scale with e.g.\ the masses and formation times of ellipticals 
3672: (and how this relates to gas fractions and stellar 
3673: populations in observed disks). In general, 
3674: for similar merger histories, the increasing prevalence of later type 
3675: galaxies (S0's and S0a's) at lower masses where disks are characteristically 
3676: more gas rich is a natural consequence of our predictions here, and 
3677: it is straightforward to convert our predicted scalings into detailed predictions 
3678: for the abundance and mass fractions of disks given some simplified merger histories. 
3679: To the extent that these processes also give rise to disk heating 
3680: and/or increasing velocity dispersions in disks, or changing kinematics in 
3681: both disks and bulges, then there should be corresponding relationships 
3682: between galaxy shapes, kinematics, and bulge-to-disk ratios along the Hubble 
3683: sequence. We investigate these possible correlations and tests in 
3684: subsequent papers (in preparation). 
3685: 
3686: Altogether, our results here elucidate 
3687: the relevant physics important for both dissipational and dissipationless 
3688: bulge formation in mergers. They 
3689: support a new paradigm in 
3690: which to view bulge and disk formation: gas-richness is not simply 
3691: a ``tweak'' to existing models of bulge formation and disk 
3692: destruction in mergers. Rather, if disks are sufficiently gas rich, 
3693: the qualitative character of mergers is different, with inefficient 
3694: angular momentum loss giving rise to disk-dominated 
3695: remnants. This process is not inherently governed by poorly-understood 
3696: feedback physics (although such feedback may be critical for 
3697: establishing the conditions necessary in the first place), 
3698: but rather by well-understood gravitational physics, and as such is 
3699: robust and fundamentally inescapable. Aspects of 
3700: galaxy populations such as the continuum of relative bulge 
3701: and disk mass ratios are not simply consequences of e.g.\ different 
3702: amounts of accretion, but can arise owing to the continuum in 
3703: efficiencies of disk destruction as a function of merger 
3704: mass ratios, orbital parameters, and gas content. 
3705: The relative (lack of) abundance of bulges at low galaxy masses and high 
3706: redshift is a basic consequence of the dynamics of 
3707: how gas loses angular momentum in mergers, even for similar 
3708: merger histories. In short, the baryonic physics of mergers 
3709: ensures that, despite the near self-similarity of the physics and merger histories 
3710: of their host halos, disk and bulge formation are not a self-similar 
3711: process, influenced dramatically (well out of proportion to the absolute 
3712: cold gas mass fractions) by the gas-richness of the baryonic systems. 
3713: 
3714: 
3715: 
3716: \acknowledgments 
3717: We thank Shardha Jogee and Rachel Somerville 
3718: for helpful discussions, and thank the anonymous referee for helpful 
3719: suggestions and clarification. This work
3720: was supported in part by NSF grants ACI 96-19019, AST 00-71019, AST
3721: 02-06299, and AST 03-07690, and NASA ATP grants NAG5-12140,
3722: NAG5-13292, and NAG5-13381. Support for 
3723: TJC was provided by the W.~M.\ Keck 
3724: Foundation. 
3725: 
3726: \bibliography{/Users/phopkins/Documents/lars_galaxies/papers/ms}
3727: 
3728: \end{document}
3729: