1:
2: %\documentclass[12pt,preprint]{aastex}
3: \documentclass[]{aastex}
4: \usepackage{emulateapj5}
5:
6:
7: %\usepackage{graphicx,ulem}
8:
9: %\input epsf
10:
11: \shorttitle{Formation of exponential disks from clump clusters}
12:
13: \shortauthors{Bournaud, Elmegreen \& Elmegreen}
14:
15: \begin{document}
16:
17: \title{Rapid formation of exponential disks and bulges at high redshift\\
18: from the dynamical evolution of clump cluster and chain galaxies}
19:
20: \author{Fr\'ed\'eric Bournaud \affil{Laboratoire AIM, CEA/DSM - CNRS - Universit\'e Paris Diderot\\Dapnia/Service d'Astrophysique,
21: CEA/Saclay, F-91191 Gif-sur-Yvette Cedex, France,
22: frederic.bournaud@cea.fr} }
23: \author{Bruce G. Elmegreen \affil{IBM Research Division, T.J. Watson
24: Research Center, P.O. Box 218, Yorktown Heights, NY 10598, USA,
25: bge@watson.ibm.com} }
26: \author{Debra Meloy Elmegreen \affil{Vassar College,
27: Dept. of Physics \& Astronomy, Box 745, Poughkeepsie, NY 12604;
28: elmegreen@vassar.edu} }
29:
30: \begin{abstract}
31: Many galaxies at high redshift have peculiar morphologies dominated
32: by $10^8-10^9$ M$_\odot$ kpc-sized clumps. Using numerical
33: simulations, we show that these ``clump clusters'' can result from
34: fragmentation in gravitationally unstable primordial disks. They
35: appear as ``chain galaxies'' when observed edge-on. In less than
36: 1~Gyr, clump formation, migration, disruption, and interaction with the disk cause
37: these systems to evolve from initially uniform disks into regular
38: spiral galaxies with an exponential or double-exponential disk profile and a
39: central bulge. The inner exponential is the initial disk size and
40: the outer exponential is from material flung out by spiral arms and
41: clump torques. A nuclear black hole may form at the same time as the
42: bulge from smaller black holes that grow inside the dense cores of
43: each clump. The properties and lifetimes of the clumps in our models
44: are consistent with observations of the clumps in high redshift
45: galaxies, and the stellar motions in our models are consistent with
46: the observed velocity dispersions and lack of organized rotation in
47: chain galaxies. We suggest that violently unstable disks are the
48: first step in spiral galaxy formation. The associated starburst
49: activity gives a short timescale for the initial stellar disk to form.
50: \end{abstract}
51:
52: \keywords{galaxies: evolution --- galaxies: formation --- galaxies: high-redshift}
53:
54: \section{Introduction}
55:
56: Galaxies at redshifts larger than $z\sim1$ become increasingly
57: clumpy with star formation in kpc-size complexes containing
58: $10^7-10^9$M$_\odot$ that are several hundred Myr old (see review in
59: Elmegreen 2007). A remarkable morphology is that of ``chain
60: galaxies'' with large visible clumps aligned along one axis (Cowie,
61: Hu \& Songaila 1995; van den Bergh et al. 1996; Moustakas et al.
62: 2004). These are rare in the Local Universe (e.g., Abraham et al.
63: 1996) but much more frequent at redshift $\sim 1$ (Elmegreen et al.
64: 2005a, hereafter EERS05). Dalcanton \& Schectman (1996) suggested that chain galaxies
65: could be edge-on LSBs. However, the clumps are much more massive
66: than those observed in the UV in nearby edge-on disks (e.g., Smith
67: et al. 2001), and the actual face-on counterparts of chain galaxies
68: at high-redshift are observed to be starburst disks with
69: kpc-sized clumps -- the so-called ``clump cluster galaxies''
70: (Elmegreen, Elmegreen \& Hirst 2004, see the example of
71: UDF~1666 in Fig.~\ref{fig:udf1666} from EERS05). Both clumpy types are found up
72: to the bandshifting limit of $z\sim5$ (Elmegreen et al. 2007a,
73: hereafter EERC07). Ring (Elmegreen \& Elmegreen 2006a) and
74: interacting (Elmegreen et al. 2007b) galaxies at high redshift are
75: also very clumpy. Distant ellipticals can be clumpy too, as found in
76: the Tadpole and A1689 cluster fields (Menanteau et al. 2004, 2005)
77: and the Ultra Deep Field (UDF; Elmegreen, Elmegreen, \& Ferguson
78: 2005). Clumpiness in high redshift galaxies has been quantified with
79: the $S$ parameter by Conselice (2003) and Conselice et al. (2004).
80:
81: O'Neil et al. (2000) suggested that chains could be knotty disks
82: seen edge-on; otherwise the edge-on counterparts of face-on disks
83: would be missing at high redshift. Taniguchi \& Shioya (2001)
84: proposed that chains are filaments of clumps about to merge into
85: elliptical remnants. Detailed observations actually confirm that
86: {\it chain} and {\it clump-cluster} galaxies are mostly a single
87: class of objects that are clumpy disks viewed in different
88: orientations: the two types have equivalent sizes, magnitudes, and
89: redshift ranges, and their clumps have similar properties (Elmegreen
90: \& Elmegreen 2005, hereafter EE05, and EERC07). Their combined
91: distribution of axial ratios is flat, as it is for a single disk
92: population (EE05). In a sample of 10 extremely clumpy galaxies in
93: the UDF, the fraction of the light in the clumps was found to be
94: $\sim40$\% and the fraction of the total stellar mass was $\sim10$\%
95: (EE05). On average in the UDF, the luminosity fraction of the clumps
96: in 178 clump-cluster galaxies is $\sim 20$\%, and the luminosity
97: fraction of the clumps in 269 spiral galaxies is $\sim5$\%, not
98: counting the bulges in the latter (EERS05).
99: Comparisons between ACS and NICMOS images of clump cluster and chain
100: galaxies confirm that the clumps are intrinsic to the mass and are
101: not rest-frame blue patches on a smooth underlying disk (EERC07).
102: The clumps therefore represent a large fraction of the mass in these
103: galaxies, unlike the clumps in modern spirals.
104:
105: Clump clusters and chains have irregular and somewhat flat
106: luminosity profiles -- different from spirals, which have bulges and
107: exponential disks. However, the azimuthally averaged number density
108: of clumps as a function of radius in UDF clump clusters, when
109: normalized to the galaxy sizes, is similar to that of the (smaller)
110: clumps in the spirals, and both distributions are close to
111: exponential (Elmegreen et al. 2005b, hereafter EEVFF05). This
112: equivalence suggests that exponential disks and bulges in spiral
113: galaxies generally form by the dissolution of clumps in clump
114: cluster and chain galaxies. Consistent with this scenario is the
115: observation that the earliest disks in the Universe are clump
116: clusters and chains (EERC07). Thus, clump cluster galaxies may
117: evolve into spirals, making the ``proto-spiral'' nomenclature for
118: this type in van den Bergh et al. (1996) remarkably prescient.
119:
120: The purpose of this article is to simulate the evolution of
121: clump-cluster galaxies and determine their final state. Noguchi
122: (1999) and Immeli et al. (2004a,b) also simulated gravitational
123: collapse in highly gaseous disks and found that they formed giant
124: clumps that interact and eventually merge into a bulge. They did not
125: consider the final disk profile and their galaxies had the masses
126: and sizes of modern galaxies. Here we consider disks that are about
127: half of this size as observed in the UDF (EERC07). We show that
128: initially uniform unstable disks evolve quickly through a clump
129: cluster phase and end up with bulges and double-exponential
130: profiles. The absence of clear rotation in chain and clumpy galaxies
131: (Bunker et al. 2000; Erb et al. 2004; Weiner et al. 2006) is also
132: explained: the massive clumps severely affect the velocity fields
133: and enhance the disk velocity dispersion.
134:
135: The numerical techniques and initial conditions are described in
136: Sect.~2. We study the properties and evolution of clump cluster and
137: chain galaxies and the formation of exponential disks in Sect.~3. In
138: Sect.~4, we discuss the results and compare with the evolution of
139: stable spiral disks. Sect.~5 considers the formation of nuclear
140: black holes. Our conclusions are summarized in Sect.~6.
141:
142: \section{Simulations}
143:
144: \subsection{Code description}
145:
146: We model the evolution of Jeans-unstable gas-rich galactic disks
147: using a particle-mesh sticky-particle code (see Bournaud \& Combes
148: 2002, 2003). The grid resolution and gravitational softening are
149: 110~pc. Stars, gas, and dark matter halo are modeled with one
150: million particles each. Unless specified otherwise, we use
151: sticky-particle parameters $\beta_r=\beta_t=0.7$ (defined as in
152: Bournaud \& Combes 2002).
153:
154: Star formation is described by a local Schmidt law. At each
155: timestep, the probability for each gas particle to be transformed
156: into a stellar particle is proportional to the local gas density to
157: the exponent 1.4 (e.g., Kennicutt 1998), the proportionality factor
158: being computed to provide a star formation rate of
159: $3.5$~M$_{\sun}$~yr$^{-1}$ in the initial uniform disk, equivalent
160: to a timescale for gas consumption of 5~Gyr (before the formation of
161: dense clumps increases the star formation rate). Feedback from star
162: formation is not implemented in this study; the consequences of this
163: choice are discussed in Section~4.2.
164:
165: \subsection{Initial conditions}
166:
167: We start the simulations with a flat and bulgeless disk of gas and
168: stars, which is Jeans-unstable and has a uniform surface density.
169: This is because clump cluster galaxies are not observed to have a
170: bulge, and they have irregular radial profiles. For instance,
171: the typical clump cluster galaxies UDF~3752
172: and UDF~82+99+103 in EEVFF05, and UDF~1666 shown here in
173: Figure~\ref{fig:udf1666}, have an irregular profile, which on large
174: scales is rather flat.
175: The azimuthally averaged radial profiles for these three galaxies
176: are irregular as well, showing a central minimum for UDF~82+99+103
177: and UDF~1666, prominent bumps at several radii, and either a
178: flattening (UDF~3752) or a slow taper from the gradient in clump
179: number density out to the edge (UDF~1666).
180: To study the formation of exponential disks from clump clusters we
181: thus have to start with a disk that does not already have a
182: concentrated nearly-exponential profile. The instabilities will
183: rapidly make it irregular.
184:
185: The initial disk radius is 6~kpc and its mass $7 \times
186: 10^{10}$~M$_{\sun}$, comparable to the observed properties of $z
187: \simeq 1$ systems (e.g., EEVFF05). The initial thickness is
188: $h=700$~pc with a $\mathrm{sech}^2\left( z/h \right)$ vertical mass
189: distribution, which ensures that, once instabilities begin to heat
190: the system, the thickness is compatible with the observed average of
191: $h=1$~kpc (fitted on $\mathrm{sech}^2\left( z/h \right)$ profiles,
192: Elmegreen \& Elmegreen 2006b). Stars initially have a Toomre
193: parameter $Q_{\mathrm{s}}=1.5$. The dark halo is a Plummer sphere of
194: radial scale-length 15~kpc. The velocity dispersion of the gas
195: $\sigma$, gas mass fraction in the disk $f_{\mathrm{G}}$, and
196: halo-to-disk mass ratio $H/D$ (measured inside the initial disk
197: radius) are varied as indicated in Table~1. The fiducial run~0 has
198: $\sigma=9$~km~s$^{-1}$, $f_{\mathrm{G}}=0.5$ and $H/D=0.5$, the
199: resulting Toomre X-parameter (Toomre 1964) is shown as a function of
200: radius on Figure~\ref{fig:xparam}; this parameter is defined by:
201: \begin{equation}
202: X(r)=\frac{r \kappa(r)^2}{2 \pi G m \mu(r)}
203: \end{equation}
204: where $\kappa$ and $\mu$ are the epicyclic frequency and surface density, respectively. We show on Figure~\ref{fig:xparam} the value of $X$ for an azimuthal wavenumber $m=2$.
205:
206: For further comparisons, we also performed a control run (run C-4)
207: with the same mass distribution as run~4, but initially stable with
208: $Q_{\mathrm{s}}=1.8$ and a gas turbulent speed computed at each
209: radius to ensure $Q_{\mathrm{g}}=1.1$. The sticky-particle
210: parameters are changed to $\beta_r=\beta_t=0.8$ to avoid global disk
211: instabilities from kinematical cooling. As a result of these
212: changes, C-4 follows the evolution of a spiral galaxy instead of a
213: clump cluster. We ran this test analogous to run~4 because this run
214: has the lowest gas fraction and is the most easily stabilized with a
215: modest gas turbulent speed.
216:
217: We also changed the sticky-particles parameters in run SP-0: the
218: initial conditions are that of run~0, but $\beta_r=0.5$ and
219: $\beta_t=1$ to conserve angular momentum with about the same rate of
220: energy dissipation.
221:
222:
223: \subsection{Clump detection and analysis}
224:
225: Because the simulated disks are unstable to fragmentation, we need
226: to detect the large clumps that form in order to measure their
227: masses and other related quantities. Because observations define
228: clumps as morphological entities, we chose to detect clumps in the
229: simulated galaxies from morphological criteria too. This is made
230: every 50~Myr, on face-on projected density maps, smoothed at a
231: resolution of 200~pc to decrease the particle noise. We compute the
232: average surface density as a function of radius $\mu_0 (r)$. First,
233: clumps must represent local overdensities compared to the average
234: density at the same radius. We then keep the pixels with $\mu > 3
235: \mu_0 (r)$ \footnote{This choice was made because in some cases
236: overdensities with $\mu > 2 \mu_0 (r)$ were found to include spiral
237: arms in spite of the maximum size criterium.}. We consider only the
238: regions less extended than 3~kpc, which was found to eliminate
239: spiral arms. The remaining regions are considered as ``clumps'' if
240: their mass is at least $2\times10^8$~M$_\sun$, for we consider only
241: the high-mass clumps in this study. The regions obtained this way
242: are split into two clumps if there are two luminosity peaks $\mu_1$
243: and $\mu_2$ that are separated by pixels with $\mu< 1/3 \min
244: (\mu_1,\mu_2)$, i.e. a significant interclump contrast, and each new
245: clump is above the mass threshold (we do not separate low-mass
246: subclumps). An example of clump detection is shown on
247: Figure~\ref{fig:detect}.
248:
249:
250: To better understand the role of clumps in fueling the bulge and the
251: disk (see below), we measured several mass fractions that are
252: indicated for each run in Table~1. All these fractions are defined
253: within the baryonic mass (gas and stars), not counting the dark
254: matter halo component.
255: \begin{itemize}
256: \item $f_\mathrm{C}$ is the fraction of the baryonic mass that is
257: concentrated in the clumps, taken at the instant of its highest
258: value, the clump borders being defined as explained above.
259: \item $f_\mathrm{g,C}$ is the gas fraction in the clumps, which is the fraction of the mass in the clumps defined above that is in the form of gas. The value in Table~1 is given at the same instant as $f_\mathrm{C}$.
260: \item $f_\mathrm{B,C}$ is the fraction of the mass from the clumps (counted in $f_\mathrm{C}$) that is in the bulge at the end of the simulation. $1-f_\mathrm{B,C}$ is the mass fraction of the clumps that ends up in the disk.
261: \item $f_\mathrm{C/B}$ is the fraction of the bulge mass that comes from the clumps, $1-f_\mathrm{C/B}$ being the fraction of the bulge that does not come from the clump material.
262: \item $f_\mathrm{C/D}$ is the fraction of the disk mass that comes from the
263: clumps.
264: \end{itemize}
265:
266: In these definitions, the {\it disk} mass is obtained by the
267: integration of the fitted double-exponential profile described
268: below. The {\it bulge} mass is then defined as the mass left over
269: above this disk profile in the central regions, without any
270: consideration of the bulge luminosity profile itself. Bulges were
271: never found to contribute to the luminosity profile beyond
272: $r=2$~kpc, so we decide that all particles at radii larger than
273: 2~kpc are disk particles. Particles at radii $r<2$~kpc have a
274: probability $\mu_{\mathrm{d}}(r)/\mu(r)$ to be disk particles, where
275: $\mu(r)$ is the measured (total) surface density profile and
276: $\mu_{\mathrm{d}}$ the (fitted) density profile of the disk
277: component. The probability that it is a bulge particle is
278: $1-\mu_{\mathrm{d}}(r)/\mu(r)$. These two probabilities are limited
279: to 1 and 0 respectively when fluctuations cause $\mu(r)$ to be
280: larger than $\mu_{\mathrm{d}}(r)$. Thus, the definition of bulge and
281: disk particles in the central regions is made on a simple
282: statistical basis that retrieves the mass and radial distribution of
283: each component. More advanced criteria might distinguish the bulge from the disk
284: based on vertical distributions and velocity dispersions. However,
285: significant differences from the present definitions would appear
286: only in a narrow radial region around $r\simeq1$ kpc, because at other
287: radii one or the other of the two components dominates.
288:
289: \section{Results}
290:
291: \subsection{Clump cluster evolution, disk and bulge fueling}
292: \label{sect:evol}
293:
294: Figure~\ref{fig:evol} shows the evolutionary sequence for the
295: fiducial run. The initial bulgeless disk is gravitationally unstable
296: and makes several large clumps of stars and gas quickly. The clumps
297: last $\simeq 400$ Myr and gather $f_\mathrm{C}=38\%$ of the total
298: baryonic mass of the system at $t=200$ Myr, which is the time at
299: which this fraction reaches its maximum. The mass fraction of the
300: stars in these clumps compared to the total stellar mass is 29\%. At
301: this stage the clumps are gas-rich with a gas mass fraction of 56\%,
302: while the gas fraction over the whole galaxy at the same instant is
303: 42\%.
304:
305: The clumps then progressively release their mass in the disk: at
306: $t=400$~Myr the clump mass fraction has decreased to 27\% from the
307: initial 38\%.
308: They have formed stars rapidly with an average star formation rate
309: between $t=200$ and $t=400$~Myr of 38~M$_{\sun}$~yr$^{-1}$, and the
310: gas mass fraction in the clumps has decreased from 56\% to 37\%. The
311: star formation rate in the model is compatible with the observed
312: average of 20~M$_{\sun}$~yr$^{-1}$ in the real clumps (EE05),
313: although the star formation efficiencies cannot be directly compared
314: because of the unknown gas mass. At the same time, their mutual
315: interaction and the friction on the halo and underlying disk bring
316: the clumps to the galaxy center (Fig.~\ref{fig:evol}) within a
317: timescale of about 4 rotations, which is 0.5--0.7 Gyr. Later on, the
318: galaxy resembles a spiral galaxy, with a central bulge and only a
319: little mass in small remaining clumps. The final system is dominated
320: by stars and shows weak density waves in the mass distribution
321: (Fig.~\ref{fig:evol}), which can be compared to smooth spiral arms
322: in the old stellar populations of a spiral galaxy seen in
323: near-infrared. The gas distribution at two late stages is shown in
324: Figure~\ref{fig:gas} and has a stronger spiral structure. Other
325: examples of simulated clump clusters are for run~6 in
326: Figure~\ref{fig:rc} and run~4 in Figure~\ref{fig:control}.
327:
328: The models in Noguchi (1999) and Immeli et al. (2004a,b) started
329: with more massive disks and developed larger clumps, most of which
330: migrated to the central spheroid, leaving only a faint disk whose
331: radial profile was not considered. With our initial conditions more
332: representative of high-redshift clump clusters, we obtain lower
333: bulge mass fractions: the bulge-to-total mass ratio (including the
334: halo) varies between 0.12 and 0.36 over our sample of simulations
335: (Table~1), which correspond to bulge-to-disk ratios between 0.14 and
336: 0.50. Hence, the galaxies resulting from clump cluster evolution are
337: still disk-dominated galaxies after the bulge has formed.
338:
339: \subsection{Mass distribution and radial profile}
340:
341: Figure~\ref{fig:radial} shows the azimuthally averaged radial
342: profiles of the disk in the fiducial run corresponding to each
343: snapshot in Figure~\ref{fig:evol}.
344: The disk starts with a flat profile that quickly becomes irregular
345: during the clump cluster phase. Clump evolution then forms a bulge
346: and an exponential disk. The exponential extends from about 1 to
347: 7~kpc in radius with a scale length $r_{\mathrm{e}}=2.1$~kpc. All
348: other runs show a similar profile with a central bulge and a massive
349: exponential disk after 1~Gyr or less (see Table~1), while the
350: luminosity profile is irregular and bulgeless during the
351: clump-cluster phase (see also Fig.~\ref{fig:rc} for another example
352: in run~6).
353:
354: The profile at $t=200$~Myr in Figure \ref{fig:radial} and the
355: corresponding image in Figure \ref{fig:evol} resemble the
356: azimuthally averaged profile and image of UDF1666 (Fig.
357: \ref{fig:udf1666}), suggesting that this clump cluster galaxy
358: is already in a somewhat advanced stage. It has no bulge yet
359: but a nearly-central clump that could be a proto-bulge, and has begun
360: to acquire a radial density gradient, but not yet a regular exponential.
361: Some other clump clusters do not have any central clump and even
362: more irregular profiles (like UDF~82+99+103 and UDF~3752 in EEVFF05)
363: and these ones correspond to earlier stages of the clump-cluster evolution
364: according to our models.
365:
366:
367: In our models with various initial conditions (see Table~1), the clumps
368: gather $f_\mathrm{C}$=23~to~38\% (average for all runs: 33\%)
369: of the system baryonic mass at the peak of the clump-cluster
370: phase. This fraction decreases with time because shear and tidal
371: fields distort the clumps, and they form stars which makes them
372: easier to disrupt. Hence, all the mass from the clumps does not
373: reach the bulge: $f_\mathrm{B,C}$=35--62 (average 50)\% of the clump
374: mass content is finally found in the bulge. The other half of the
375: clump mass has been released in the disk. This way, the bulge gets
376: most of its mass ($f_\mathrm{C/B}$=65--74, average 69\%) from the
377: clumps, the rest of its mass coming from material initially present
378: at the disk center and severely heated and thickened when the clumps
379: reach the inner regions and merge together.
380:
381: As for the disk component, 17--26\% of its mass
382: (average $f_\mathrm{C/D}$=23\%) is made of material released
383: by the clumps during their migration towards the disk center.
384: Hence the clumps have two effects on the disk:\\
385: {\it (i)} they release mass into the disk, which has been driven
386: inward by the clump migration, making the disk density
387: profile more concentrated\\
388: {\it (ii)} they act on the disk material that is not in the clumps, and
389: redistribute it through gravity torques, in the same manner as
390: spiral density waves (e.g. Pfenniger \& Friedli 1991), but more
391: rapidly (see Section~4.1).
392:
393: Hence, the material released by the clumps and the material from the
394: rest of the disk both evolve into a more concentrated distribution,
395: which our simulation shows is an exponential disk with a bulge. We
396: suggest later (Section~\ref{cc-spiral}) that it is almost entirely
397: the clumps that are responsible for this evolution.
398:
399: In addition to the main exponential disk, most runs actually end up
400: with a second, outer exponential profile. In the fiducial run, the
401: second exponential is from 7 to 13~kpc with a scale length of
402: $r'_{\mathrm{e}}=1.6$~kpc (Fig.~\ref{fig:radial}). At least six of
403: the eight runs show such a double-exponential, all with a change in
404: scalelength at around 7~kpc, comparable to the initial disk radius.
405: It is noticeable that the knee in the luminosity profile occurs at
406: the outer edge of the region initially affected by clump
407: instabilities. One can see in Figure~\ref{fig:evol} that the clumps
408: fling out some material in the form of spiral arms emerging from
409: them: this feeds the outer disk component, which has a lower density
410: and a steeper profile than the main exponential disk where the
411: clumps evolve. The scale lengths $r_{\mathrm{e}}$ and
412: $r'_{\mathrm{e}}$ indicated in Table~1 correspond to the 1--7~kpc
413: and 7--13~kpc ranges\footnote{The scale lengths have been computed
414: as the best fitting parameter over the 2--6 and 8--12~kpc range, but
415: are generally representative of the 1--7 and 7--13~kpc regions, as
416: seen on Figure~\ref{fig:radial}}. Runs 5 and 6 have a change of
417: slope that is small and might not desserve to be considered
418: double-exponential: the double-exponential fit is robust in the
419: models, but the change of slope could be difficult to detect in real
420: images with a limited sensitivity. In the six other runs the ratio
421: of the inner to outer exponential slopes is on average $\simeq 1.7$.
422: Double-exponentials have been observed in local (e.g. Pohlen et al.
423: 2002) and $z=0.6-1$ spirals (P\'erez 2004), as well as in local
424: irregulars (Hunter \& Elmegreen 2006). The ratio of slopes for the
425: observations at $z\simeq 1$ (P\'erez 2004) is $2 \pm 0.35$,
426: compatible with those found in the present models.
427:
428: \section{Discussion}
429:
430: \subsection{The role of clumps and other processes in making
431: exponential disks}\label{cc-spiral}
432:
433: We find that massive clumps in a primordial disk galaxy form an
434: exponential stellar disk with a bulge. There are other
435: theoretical ways to make
436: exponential disks too (e.g., Freeman 1970, Fall \& Efstathiou 1980,
437: Robertson et al. 2004). Pfenniger \& Friedli (1991) suggest that
438: spiral arm torques could make exponential profiles, although they
439: started with a concentrated, near-exponential profile anyway (here
440: we started with a flat profile).
441:
442: To understand the role of clumps and spirals, we compared our
443: clump-free control run~C-4 with our clump-cluster model run~4. C-4
444: develops spiral arms; after 1~Gyr its luminosity profile is somewhat
445: concentrated but still irregular and far from
446: exponential (Fig.~\ref{fig:control}). This indicates that:\\
447: {\it (i)} the formation of the exponential disk in evolving clump
448: cluster and chain galaxies is caused by the clumps, i.e., by
449: the mass they release and by their gravitational action on the rest of the disk.\\
450: {\it (i)} the timescale to redistribute mass into an exponential
451: disk through spiral arms is significantly larger than the timescale
452: to redistribute mass by clump interactions. For example, the
453: exponential profile in the model by Pfenniger \& Friedli (1991) took
454: 5~Gyr to form. Massive disk clumps are therefore expected to
455: dominate the formation of exponential disks.
456:
457: Furthermore, if the formation of exponential-like profiles were
458: driven only by density waves, then there should be non-clumpy spiral
459: disks with non-exponential profiles that have not yet had time to
460: form exponentials. This contradicts the observation that irregular
461: profiles are highly correlated with clump cluster and chain galaxies
462: (EEVFF05). Also, the spiral mechanism would not work for dwarf
463: irregulars, which have exponential disks too (Hunter \& Elmegreen
464: 2006). More likely, spirals adjust, maintain, and lengthen the first
465: exponential as random accretions and interactions distort the disk
466: over time. For example, spiral arms could increase the scale length
467: over time in the period from $z=1$ to $z=0$ (EEVFF05). Disk shear
468: viscosity with star formation can also make exponentials (Lin \&
469: Pringle 1987) but the timescale is too long, dwarfs with little
470: shear have exponentials, and exponentials in the far outer regions
471: could not be made this way if star formation is absent.
472:
473: \subsection{Modelling of the interstellar medium}
474:
475: The interstellar medium has been modeled with sticky particles. Such
476: schemes do not strictly conserve angular momentum, but the actual
477: loss is always much smaller than the angular momentum transferred by
478: spiral torques in regular spirals (Bournaud, Combes \& Semelin 2005). It should be
479: negligible compared to angular momentum redistribution by clump
480: interactions in clump clusters. To check this we ran simulation SP-0 conserving
481: angular momentum during inelastic collisions between gas particles
482: (described in Section~2.2). In this run, the bulge-to-disk mass
483: ratio was slightly smaller, and the clumps released a slightly
484: higher fraction of their mass in the disk, indicating that they are
485: less gravitationally bound (Table~1). The changes are minor,
486: however, and smaller than those corresponding to variations in
487: $\sigma$ or $H/D$.
488:
489: Feedback from star formation has not been included because its
490: effects are not well known and it should not be excessively
491: important. For example, feedback could heat the gas in the clumps
492: and cause them to release mass more rapidly. The clumps
493: already have masses of a few $10^8$ to $10^9$~M$_\sun$,
494: which is an order of magnitude above proposed masses of objects that
495: could be severely affected by supernovae (Dekel \& Woo 2003). The
496: effect of gas release would be to increase the disk mass relative to
497: the bulge, which would strengthen our conclusion that clump clusters
498: can evolve into spiral-like galaxies with massive disks. Variations
499: in the gas fraction for the initial disk and for the clumps does not
500: strongly affect the results (see Table~1, runs 4 and 5 vs. run 0).
501: Clumps with higher gas fractions are more tightly bound because the
502: mass in them is more concentrated, and then a slightly higher
503: fraction of the clump mass reaches the bulge.
504:
505: \subsection{Comparison to real clump cluster and chain galaxies,
506: and the thickness of disks}
507:
508: Figure~\ref{fig:edge} shows an edge-on view of the fiducial model of
509: Figure~\ref{fig:evol}. It resembles a chain galaxy with clumps aligned
510: along the axis of a thick disk and no central bulge. This supports
511: the statistical inference that observed chains are edge-on clump
512: clusters.
513:
514: The lifetime of the clump cluster phase in our model ($\sim0.5$~
515: Gyr) is consistent with the average age of $\sim300$~Myr observed
516: for clumps (EE05). Our simulations suggest that disk galaxies
517: undergo only one clump-cluster phase. After this, the resulting
518: bulge and exponential disk cause the galaxy to be classified as a
519: spiral. Large clumps can still form in the spiral disk, but the
520: system is more stable at this stage because of the lower gas mass
521: and higher velocity dispersion, so the clumps will be smaller and
522: less important dynamically. Multiple clump-cluster phases could be
523: possible only if a significant mass of gas accretes relatively quickly
524: onto the disk after the first phase, making it unstable to form
525: giant clumps again. Presumably these clumps would also disperse and
526: add stellar mass to the disk and bulge.
527:
528: Figure~\ref{fig:rc} shows the velocity fields and rotation curves
529: for Run~6 at each time step, obtained by viewing the disk at a 70
530: degree angle. During the clump-cluster phase, the linewidth is
531: broadened and the rotation pattern is irregular because of
532: clump-clump interactions. In the final galaxy, the velocity field is
533: typical for spirals, although the dispersion is somewhat high. We
534: predict that rotation in clump-clusters should be difficult to
535: observe because of the turbulent and irregular motions. For example,
536: if the outer parts of the disk in Figure~\ref{fig:rc} were not
537: detected beyond the clumpy region, no organized rotation would be
538: seen at all; only the high velocity dispersion would be evident.
539: This may explain why Weiner et al. (2006) found that only 30\% of
540: the chain galaxies in their survey had clear evidence for rotation.
541: F{\"o}rster Schreiber et al. (2006) and Genzel et al. (2006) also
542: found velocity dispersions higher than usual for clumpy,
543: high-redshift galaxies.
544:
545: In our models, the average values of deprojected $V/\sigma$ measured
546: between 3 and 6~kpc in radius at the end of the simulations are
547: given in Table~1. They average to 2.7, which corresponds to a high
548: velocity dispersion but still dominated by rotation. Presumably the
549: dispersion of gas and new stars during the subsequent evolution will
550: be much lower because there are no giant clumps to stir up the disk.
551: Then the final stage will be a relatively quiet and thin rotating
552: disk with a bulge and hot disk component from the clump cluster
553: phase.
554:
555: The final edge-on disk (Fig.~\ref{fig:edge}) appears to be about the
556: same thickness as the chains and edge-on spirals in the UDF, namely
557: $\sim1$ kpc (Elmegreen \& Elmegreen 2006b). Actually, the population
558: of stars already existing at the beginning of the simulation has a
559: scale-height of 1.3~kpc, having been heated by the violent
560: instabilities. The stars formed during the simulation have a
561: scale-height of $h=650$~kpc (defined as in Section~2.2). Hence the
562: final system is not made up only of a thick disk but has a massive
563: thick component in addition to a younger and thinner stellar disk.
564: The thick and thin disk components obtained here have comparable
565: masses, which is compatible with observations of same-mass spirals
566: ($V_\mathrm{circ}\simeq 120$~km~s$^{-1}$ and below, Yoachim \&
567: Dalcanton 2006). This is also compatible with the disks at high
568: redshift being thick or having a massive thick component that gets
569: redder with height (Elmegreen \& Elmegreen 2006b). Higher-mass
570: present-day spirals like the Milky Way have higher thin-to-thick
571: disk mass fractions (see Robin et al. 2003 for the Milky Way),
572: suggesting that they further evolved after the primordial unstable
573: phases, possibly having a massive thin disk grown by gas accretion
574: (Bournaud \& Combes 2002, Dekel \& Birnboim 2006). The relatively
575: thicker disks at high redshift compared to local spirals
576: (Reshetnikov, Dettmar \& Combes 2003) also suggests that present-day
577: large spirals kept on evolving through a different process after the
578: clump-cluster phase. Local spirals also have larger disks (Trujillo
579: \& Pohlen 2005) with larger exponential scale lengths (EEVFF05),
580: which further indicate that the clump-clusters and chains at high
581: redshift formed the first exponential disks, and afterwards secular
582: evolution and mass accretion enlarged them to form the present-day
583: spirals.
584:
585: \subsection{Nuclear Black Hole formation}
586:
587: According to Kormendy \& Richstone (1995) and others, bulges always
588: have black holes (BHs) with a mass proportional to the bulge mass.
589: This suggests that any bulge formation model should also account for
590: the simultaneous formation of central BHs. There should not be
591: independent BH and bulge formation mechanisms. In this case, our
592: model requires that intermediate mass BHs (IMBHs) form in the giant
593: clusters of clump-cluster galaxies and migrate inward with the
594: cluster cores. They would then have to coalesce in the galactic
595: nuclei to make supermassive BHs. Simultaneous gas accretion could
596: contribute too. To form QSOs at high redshift, this process would
597: have to occur very quickly. While this seems possible for the
598: densest and earliest disks that form -- it takes only several orbit
599: times for the giant clumps to make a bulge -- cosmological and
600: detailed disk simulations will be necessary to confirm the basic
601: model. One prediction would be that clumps in clump cluster and
602: chain galaxies should contain IMBHs, which implies the clumps should
603: be X-ray and radio sources if self-absorption is not too large.
604: Present-day observations cannot resolve the clumps well enough to
605: determine if this is the case. An alternative model suggests that BHs
606: form directly from accreted primordial gas without passing through
607: stars first (e.g., Di~Matteo et al. 2007). Observations of IMBHs in
608: disks may be able to distinguish between these two possibilities.
609:
610:
611: The clump-cluster scenario for nuclear BH formation is consistent with
612: one of the prominent models in the literature. Ebisuzaki et al. (2001) showed
613: with N-body simulations that ultra-dense clusters can form BHs from
614: stellar coalescence. They proposed that nuclear BHs grow from the
615: coalescence of these cluster BHs and obtained the observed BH/bulge
616: mass ratio of $10^{-3}$. Portegies~Zwart \& McMillan (2002) found a
617: similar result; massive clusters with relaxation times less than 25
618: Myr have mass segregation, rapid core collapse, and runaway stellar
619: coalescence that forms supermassive stars up to $10^{-3}$ of the
620: cluster mass. Portegies~Zwart et al. (2004) applied this model to
621: explain a suspected IMBH in a supermassive cluster of M82. G\"urkan,
622: Freitag \& Rasio (2004) studied a cluster with $10^7$ stars; when
623: the core-collapse time is less than the O-star lifetime, runaway
624: collisions form IMBHs with masses that are always about $10^{-3}$ of
625: the cluster mass. More detailed models of stellar coalescence in
626: dense cluster cores confirmed the formation of supermassive stars
627: (Freitag, G\"urkan, \& Rasio 2006).
628:
629: Portegies~Zwart et al. (2006) followed the evolution of dense
630: clusters in the Milky Way nuclear region and found that 10\% form
631: IMBHs during their inward migration. These IMBHs coalesce with the
632: nuclear BH at a high enough rate to account for all of the nuclear
633: BH mass. Matsubayashi, Makino, \& Ebisuzaki (2007) also studied the
634: coalescence of IMBHs to make supermassive BHs in galactic nuclei.
635: They found that dynamical friction becomes ineffective close to the
636: central BH, but that IMBH orbits become highly eccentric and spiral
637: in quickly anyway because of gravitational radiation.
638:
639: Our model for nuclear BH formation is qualitatively similar to these
640: but it occurs at a much earlier stage in the life of a galaxy when
641: the clusters are more massive and more dispersed in the main disk.
642: Also in our model, the mass of the cluster BHs could be much larger
643: than in these other models because the early disk clusters are more
644: massive. If cluster BHs have $10^{-3}$ of the cluster mass, then in
645: our model the BHs would have $\sim10^5-10^6$ M$_\odot$. This should
646: be enough to make a quick AGN in a moderate size disk galaxy, but
647: subsequent accretion would be necessary to make larger nuclear BHs,
648: and clump-cluster galaxy coalescence would be necessary to make
649: supermassive BHs in elliptical galaxies.
650:
651: \section{Conclusions}
652:
653: Our simulations suggest that gas-rich disks in the early Universe
654: collapse into giant star-forming clumps and resemble the clump
655: cluster and chain galaxies observed in the UDF. Noguchi (1999) found
656: similar fragmentation, as did Immeli et al. (2004b), who also
657: suggested such disks are a likely origin for chain galaxies. Here we
658: considered the observational consequences of clump interactions in
659: such galaxies, and we determined the resulting structures of the
660: galaxies.
661:
662: The fragmentation of unstable primordial disks and the strong
663: interactions between these fragments account for the observed
664: kinematics of chain galaxies. Chain and clump cluster galaxies
665: should be found to rotate, but the heightened turbulence and
666: irregular orbits from clump-clump interactions make this rotation
667: difficult to observe. These interactions should also add a large
668: velocity dispersion. These predictions are consistent with the
669: observations of chain galaxies by Weiner et al. (2006), and with the
670: large velocity dispersions found in clump cluster galaxies by
671: F\"orster Schreiber et al. (2006) and Genzel et al. (2006).
672:
673: The clumps in our simulations form very quickly. They interact with
674: each other, exchange angular momentum, migrate inward, dissolve, and
675: perturb the underlying disk. These processes form a bulge+disk
676: structure that is typical of spiral galaxies today, and they produce
677: an exponential profile for the main disk component. About half of
678: the clump mass goes into the disk within its initial truncation
679: radius, and in most simulations an outer, steeper exponential forms
680: from disk material that is flung out. The clump-cluster phase lasts
681: for $\sim$0.5--1~Gyr and occurs only once during the life of a
682: galaxy, unless there is a significant accretion event later.
683: There may be other ways to make bulges or pseudo-bulges (e.g., Fu,
684: Huang, \& Deng 2003; Athanassoula 2005; Kormendy \& Kennicutt 2004),
685: but only the present model has been observed at high redshift in
686: each of the important stages. Moreover, clump-cluster galaxies are
687: usually bulgeless (EEVFF05), so if they evolve into normal,
688: smooth-profile galaxies, then the bulge has to form at the same time
689: as the clumps disperse. The other secular processes can only grow
690: and evolve this first bulge later on.
691:
692: An important implication of this result is that the continuous
693: appearance of clump cluster galaxies throughout a wide range of
694: redshifts means that disk galaxies start forming over an extended
695: period of time. This long extent was also suggested by the star
696: formation timescale in comparison to the local Hubble time (EERC07).
697: Noeske et al. (2007) reach a similar conclusion based on the
698: distribution of star formation rates over mass and time.
699: When a disk becomes sufficiently massive through accretion, giant
700: clumps form at the local Jeans mass by gravitational instabilities.
701: A high velocity dispersion and a low initial disk column density put
702: this Jeans mass at a high value, around $10^8$ M$_\odot$. Clump
703: interactions redistribute disk matter into the characteristic
704: exponential form, and they thicken the disk to a $\sim$kpc scale
705: height, as observed at this stage. Clumping instabilities also cause
706: starbursts, with a star formation rate typically enhanced by a
707: factor of ten in the present models. These bursts are not related to
708: interactions, although the large-scale environment should influence
709: this kind of activity indirectly through its role in the growth of
710: disks. Such internal bursting could explain the existence of active
711: star-forming galaxies that appear to be isolated (Genzel et al.
712: 2006). Exponential disks should continue to evolve after the
713: clump-cluster phase, but the evolution should be more gradual and
714: may involve the conventional spiral and bar torques applied to
715: slowly accreting disks.
716:
717: Supermassive black holes correlate with bulges and have to form very
718: quickly in the life of a galaxy to explain QSOs and AGNs at high
719: redshift. Based on published models for runaway growth of
720: supermassive stars in dense cluster cores, we speculate that the
721: clumps in clump cluster and chain galaxies rapidly form black holes
722: with masses of $10^5-10^6$ M$_\odot$, and that these black holes
723: migrate inward as the bulges form to make nuclear black holes. This
724: whole process should operate in several rotation periods of the
725: disk, which can be fast or slow depending on the galaxy density and
726: therefore on the epoch of galaxy formation. Simulations of this BH
727: migration process will be discussed elsewhere.
728:
729: \acknowledgments
730:
731: We are grateful to an anonymous referee for constructive comments that
732: improved the presentation of the results. The simulations in this paper
733: were performed on the NEC-SX8 and SX8R
734: vectorial computers at the CEA/CCRT and IDRIS centers.
735:
736:
737: \begin{thebibliography}{}
738:
739: %\bibitem[]{675} Abraham, R.G., van den Bergh, S., Glazebrook, K., Ellis, R.S., Santiago, B.X., Surma, P., \& Griffiths, R.E. 1996, ApJS, 107, 1
740:
741: \bibitem[]{677} Abraham, R.G., Tanvir, N.R., Santiago, B.X., Ellis, R.S., Glazebrook, K., \& van den Bergh, S. 1996, MNRAS, 279, L47
742:
743: \bibitem[]{263} Athanassoula, E. 2005, in Planetary Nebulae as Astronomical
744: Tools, AIP Conf. Proceedings, Volume 804, p. 333
745:
746: \bibitem[]{266} Bournaud, F., \& Combes, F. 2002, A\&A, 392, 83
747:
748: \bibitem[]{268} Bournaud, F., \& Combes, F. 2003, A\&A, 401, 817
749:
750: \bibitem[]{686} Bournaud, F., Combes, F. \& Semelin, B. 2005, MNRAS, 364, L18
751:
752: \bibitem[]{688} Bunker, A., Spinrad, H., Stern, D., Thompson, R., Moustakas, L., Davis, M., \& Dey, A.\ 2000, ArXiv Astrophysics e-prints, arXiv:astro-ph/0004348
753:
754: \bibitem[]{690} Combes, F., \& Elmegreen, B.G. 1993, A\&A, 271, 391
755:
756: \bibitem[]{270} Conselice, C.J. 2003, ApJS, 147, 1
757:
758: \bibitem[]{272} Conselice, C.J., Grogin, N.A., \& Jogee, S. 2004, ApJ, 600, L139
759:
760: \bibitem[]{696} Cowie, L.L., Hu, E.M., \& Songaila, A. 1995, AJ, 110, 1576
761:
762: \bibitem[]{698} Dalcanton, J.~J., \& Shectman, S.~A.\ 1996, ApJL, 465, 9
763:
764: \bibitem[]{274} Dekel, A., \& Birnboim, Y. 2006, MNRAS, 368, 2
765:
766: \bibitem[]{702} Dekel, A., \& Woo, j. 2003, MNRAS, 344, 1131
767:
768: \bibitem[]{} Di Matteo, T., Colberg, J., Springel, V., Hernquist, L., \& Sijacki, D. 2007, astroph/07052269
769:
770: \bibitem[]{704} Ebisuzaki, T. et al. 2001, ApJ, 562, L19
771:
772: \bibitem[]{276} Elmegreen, D.M. 2007, in IAU Symosium 235, ed. F. Combes \& J. Palous, Cambridge: Cambridge Univ. Press, in press.
773:
774: \bibitem[]{279} Elmegreen, D.M., Elmegreen, B.G., \& Hirst, A.C., 2004, ApJ, 604, L21
775:
776: \bibitem[]{281} Elmegreen, B.G., \& Elmegreen, D.M. 2005, ApJ, 627, 623
777: (EE05)
778:
779: \bibitem[]{283} Elmegreen, D.M., Elmegreen, B.G., \& Ferguson, T.E. 2005, ApJ, 623, L71
780:
781: \bibitem[]{285} Elmegreen, D.M., Elmegreen, B.G., Rubin, D.S., \& Schaffer, M.A. 2005a, ApJ, 631, 85 (EERS05)
782:
783: \bibitem[]{287} Elmegreen, B.G., Elmegreen, D.M., Vollbach, D.R., Foster, E.R., \& Ferguson, T.E., 2005b, ApJ, 634, 101 (EEVFF05)
784:
785: \bibitem[]{289} Elmegreen, D.M., \& Elmegreen, B.G. 2006a, ApJ, 651, 676
786:
787: \bibitem[]{291} Elmegreen, B.G, \& Elmegreen, D.M. 2006b, ApJ, 650, 644
788:
789: \bibitem[]{293} Elmegreen, D.M., Elmegreen, B.G., Ravindranath, S., \& Coe, D.A., 2007a, ApJ, 658,
790: 763 (EERC07)
791:
792: \bibitem[]{295} Elmegreen, D.M., Elmegreen, B.G., Ferguson, T., \& Mullan, B. 2007b, ApJ, 663, 734
793: (astroph/07040911)
794:
795: \bibitem[]{729} Erb, D. K., Steidel, C. C., Shapley, A. E., Pettini, M., \& Adelberger, K. L. 2004, ApJ, 612, 122
796:
797: \bibitem[]{297} Fall, S. M., \& Efstathiou, G. 1980, MNRAS, 193, 189
798:
799: \bibitem[]{} F{\"o}rster Schreiber, N.~M., et al.\ 2006, ApJ, 645, 1062
800:
801: \bibitem[]{299} Freeman, K.C. 1970, ApJ, 160, 811
802:
803: \bibitem[]{735} Freitag, M., G{\"u}rkan, M.~A., \& Rasio, F.~A.\ 2006, MNRAS, 368, 141
804:
805: \bibitem[]{737} Fu, Y.~N., Huang, J.~H., \& Deng, Z.~G.\ 2003, MNRAS, 339, 442
806:
807: \bibitem[]{301} Genzel, R. et al. 2006, Nature, 442, 786
808:
809: \bibitem[]{741} G{\"u}rkan, M.~A., Freitag, M., \& Rasio, F.~A.\ 2004, ApJ, 604, 632
810:
811: \bibitem[]{303} Hunter, D., \& Elmegreen, B.G. 2006, ApJS, 162, 49
812:
813: \bibitem[]{860} Immeli, A., Samland, M., Gerhard, O., \& Westera, P. 2004a, A\&A, 413, 547
814:
815: \bibitem[]{863} Immeli, A., Samland, M., Westera, P., \& Gerhard, O. 2004b, ApJ, 611, 20
816:
817: \bibitem[]{309} Kennicutt, R.C., Jr. 1998, ApJ, 498, 541
818:
819: \bibitem[]{311} Kormendy, J., \& Kennicutt, R.C., Jr. 2004, ARA\&A, 42, 603
820:
821: \bibitem[]{} Kormendy, J., \& Richstone, D. 1995, ARAA, 33, 581
822:
823: \bibitem[]{313} Lin, D.N.C., \& Pringle, J.E. 1987, MNRAS, 325, 781
824:
825: \bibitem[]{756} Matsubayashi, T., Makino, J., \& Ebisuzaki, T.\ 2007, ApJ, 656, 879
826:
827: \bibitem[]{315} Menanteau, F., et al. 2004, ApJ, 612, 202
828:
829: \bibitem[]{317} Menanteau, F., et al. 2005, ApJ, 620, 697
830:
831: \bibitem[]{762} Moustakas, L.A. et al. 2004 ApJ, 600, 131
832:
833: \bibitem[]{} Noeske, K.G., Faber, S.M., Weiner, B.J., et al. 2007, ApJ, 660, L47
834:
835: \bibitem[]{319} Noguchi, M. 1999, ApJ, 514, 77
836:
837: \bibitem[]{766} O'Neil, K., Bothun, G.~D., \& Impey, C.~D.\ 2000, ApJS, 128, 99
838:
839: \bibitem[]{323} P\'erez, I. 2004, A\&A, 427, L17
840:
841: \bibitem[]{770} Pfenniger, D. \& Friedli, D. 1991, A\&A, 252, 75
842:
843: \bibitem[]{321} Pohlen, M., Dettmar, R.-J., L\u utticke, R., \& Aronica, G. 2002, A\&A, 392, 807
844:
845: \bibitem[]{774} Portegies~Zwart, S.~F., \& McMillan, S.~L.~W. 2002, ApJ, 576, 899
846:
847: \bibitem[]{776} Portegies~Zwart, S.~F., Baumgardt, H., Hut, P., Makino, J., \& McMillan, S.~L.~W. 2004, Nature, 428, 6984
848:
849: \bibitem[]{778} Portegies~Zwart, S.~F., Baumgardt, H., McMillan, S.~L.~W., Makino, J., Hut, P., \& Ebisuzaki, T. 2006, ApJ, 641, 319
850:
851: \bibitem[]{780} Reshetnikov, V. P., Dettmar, R.-J., \& Combes, F. 2003, A\&A, 399, 879
852:
853: \bibitem[]{325} Robertson, B., Yoshida, N., Springel, V., \& Hernquist, L. 2004, ApJ, 606, 32
854:
855: \bibitem[]{327} Robin, A.C., Reyl\'e, C., Derri\`ere, S., \& Picaud, S. 2003, A\&A, 409, 52
856:
857: \bibitem[]{786} Smith, A. M., et al. 2001, ApJ, 546, 829
858:
859: \bibitem[]{788} Taniguchi, Y., \& Shioya, Y. 2001, ApJ, 547, 146
860:
861: \bibitem[]{790} Toomre, A. 1964, ApJ, 139, 1217
862:
863: \bibitem[]{792} Trujillo, I., \& Pohlen, M.\ 2005, ApJL, 630, 17
864:
865: \bibitem[]{329} van den Bergh, S., Abraham, R.G., Ellis, R.S., Tanvir, N.R., Santiago, B.X., \& Glazebrook, K.G. 1996, AJ 112, 359
866:
867: \bibitem[]{332} Weiner, B.J., et al. 2006, ApJ, 653, 1027
868:
869: \bibitem[]{798} Yoachim, P., \& Dalcanton, J.~J. 2006, AJ, 131, 226
870:
871:
872: \end{thebibliography}{}
873:
874:
875:
876:
877:
878: \begin{table}
879: \begin{center}
880: \caption{Run parameters and results. The various fractions are defined in Section~2.3 and the bulge-to-total mass ratio $B/T$ refers to the total baryonic mass, not counting the dark matter halo.
881: \label{tbl-2}}
882: \begin{tabular}{cccccccccccccccc}
883: \tableline\tableline
884: Run && $\sigma$ & $f_{\mathrm{G}}$ & $H/D$ && $f_\mathrm{C}$ & $f_\mathrm{g,C}$ & $f_\mathrm{B,C}$ & $f_\mathrm{C/B}$ & $f_\mathrm{C/D}$ && $B/T$ & $r_{\mathrm{e}}$ & $r'_{\mathrm{e}}$ & $V/\sigma$\\
885: \tableline
886: 0 && 9 & 50\% & 0.50 && 0.38 & 0.56 & 0.53 & 0.67 & 0.26 && 0.30 & 2.1 & 1.6 & 2.9 \\
887: 1 && 5 & 50\% & 0.50 && 0.32 & 0.60 & 0.48 & 0.74 & 0.22 && 0.21 & 2.9 & 1.7 & 2.7 \\
888: 2 && 15 & 50\% & 0.50 && 0.31 & 0.52 & 0.40 & 0.66 & 0.25 && 0.19 & 2.3 & 1.4 & 2.2 \\
889: 3 && 20 & 50\% & 0.50 && 0.39 & 0.47 & 0.62 & 0.74 & 0.23 && 0.33 & 2.0 & 1.6 & 2.4 \\
890: 4 && 9 & 25\% & 0.50 && 0.26 & 0.32 & 0.36 & 0.67 & 0.20 && 0.14 & 2.4 & 1.1 & 3.0 \\
891: 5 && 9 & 75\% & 0.50 && 0.34 & 0.63 & 0.61 & 0.65 & 0.19 && 0.32 & 2.0 & (1.8) & 2.8 \\
892: 6 && 9 & 50\% & 0.25 && 0.41 & 0.51 & 0.62 & 0.71 & 0.25 && 0.36 & 2.3 & (2.0) & 2.4\\
893: 7 && 9 & 50\% & 0.80 && 0.23 & 0.57 & 0.35 & 0.67 & 0.17 && 0.12 & 2.2 & 1.2 & 3.1\smallskip \\
894: SP-0&& 9 & 50\% & 0.50 && 0.36 & 0.55 & 0.51 & 0.68 & 0.24 && 0.27 & 2.2 & 1.4 & 3.0 \smallskip \\
895: \tableline
896: \end{tabular}
897: \end{center}
898: \end{table}
899:
900:
901:
902:
903: \begin{figure}
904: \centering
905: \includegraphics[width=4.5in]{f1a.eps}\vspace{.2cm}\\
906: \includegraphics[width=4.5in]{f1b.eps}
907: \caption{upper: Clump-cluster galaxy UDF 1666 in a color optical
908: image from the UDF skywalker (left) and in NICMOS H band ( right).
909: lower: (left) The luminosity profile is shown along the major axis,
910: through the apparent center of the galaxy; this profile is irregular
911: and rather flat. (right) The azimuthally averaged radial profile has
912: a dip in the center, a peak at the brightest spot, and declines
913: gradually outward as the number density of clumps decreases. The
914: bumps correspond to radii where the main clumps occur, inside about
915: 40 pixels. The position scale in pixels corresponds to 0.03 arcsec
916: per pixel, or about 250 pc at the galaxy's photometric redshift of 1.4. This
917: profile and the corresponding image resemble step 2 in Figures 6 and
918: 4.}\label{fig:udf1666}\end{figure}
919:
920:
921: \begin{figure}
922: \centering
923: \includegraphics[width=3in]{f2.eps}
924: \caption{X-parameter (Toomre 1964) in the initial disk for the fiducial run~0, in the $m=2$ mode. Because of the initial flat profile of the disk and absence of bulge, it differs from typical spirals with values $\sim 1$ except in the central region.}\label{fig:xparam}\end{figure}
925:
926:
927:
928:
929: \begin{figure}
930: \centering
931: \includegraphics[width=3in]{f3.eps}
932: \caption{Contour of the ``clumps'' defined as entities corresponding to overdensities (above three times the average mass density at each radius) of more than $2\times10^8$~M$_\sun$ each. Two clumps are defined in the lower right region, where the density between the two peaks is lower by factor three than each density peak. We consider only regions less extended than 3 kpc in any direction, which ensures that overdense regions in the outer parts, that are spiral arms rather than clumps, are not counted.}\label{fig:detect}\end{figure}
933:
934:
935:
936: \begin{figure}
937: \centering
938: \includegraphics[width=3.9in]{f4.eps}
939: \caption{Face-on snapshots of the mass surface density in the fiducial run~0, showing the formation of a clump cluster galaxy and its evolution towards a classical spiral galaxy with a central bulge and (double-) exponential disk.}\label{fig:evol}\end{figure}
940:
941:
942:
943: \begin{figure}
944: \centering
945: \includegraphics[width=3.3in]{f5.eps}
946: \caption{Face-on snapshot of the gas mass density in the fiducial run~0 (total mass density is shown on Figure~\ref{fig:evol}) at two late instants, showing the spiral structure of the final galaxy after the clump-cluster phase.}\label{fig:gas}\end{figure}
947:
948:
949:
950:
951: \begin{figure}
952: \centering
953: \includegraphics[angle=0,width=2in]{f6a.eps}
954: \includegraphics[angle=0,width=2in]{f6b.eps}\\
955: \includegraphics[angle=0,width=2in]{f6c.eps}
956: \includegraphics[angle=0,width=2in]{f6d.eps}\\
957: \includegraphics[angle=0,width=2in]{f6e.eps}
958: \includegraphics[angle=0,width=2in]{f6f.eps}
959: \caption{Radial density profiles of the stellar disk in run~0, corresponding to each snapshot on Figure~\ref{fig:evol}. The evolution through a clump cluster phase changes the luminosity profile from irregular to exponential with a central bulge; an outer double-exponential component is also formed. The fits of the two exponential components in the 1--7 and 7--13~kpc radial ranges is shown on the last panel.}\label{fig:radial}\end{figure}
960:
961:
962: \begin{figure}
963: \centering
964: \includegraphics[angle=0,width=4in]{f7a.eps}\\
965: \includegraphics[angle=0,width=2in]{f7b.eps}
966: \includegraphics[angle=0,width=2in]{f7c.eps}\\
967: \includegraphics[angle=0,width=2in]{f7d.eps}
968: \includegraphics[angle=0,width=2in]{f7e.eps}
969: \caption{Stellar surface density, velocity field, rotation and dispersion curves, for run~6 at two different instants (during the clump cluster phase and in the final spiral galaxy). The radial density profile are also shown. The model is viewed with 70 degrees inclination, and the velocity curves are measured along the major axis indicated by the dashed line on the velocity field. The velocities here are not deprojected ; deprojected values of $V/\sigma$ are given for all models in Table~1. Clump clusters are rotating systems, but their velocity field is highly disturbed and rotation could be difficult to detect.}\label{fig:rc}\end{figure}
970:
971:
972: \begin{figure}
973: \centering
974: \includegraphics[width=3in]{f8.eps}
975: \caption{Edge-on counterparts of the snapshots shown on Figure~\ref{fig:evol} at two instants, for two perpendicular line-of-sights each. When observed edge-on, clump clusters take the appearance of chain galaxies, with several massive clumps aligned within a bulgeless underlying disk.}\label{fig:edge}\end{figure}
976:
977:
978: \begin{figure}
979: \centering
980: \includegraphics[angle=0,width=3in]{f9.eps}
981: \caption{Radial density profiles at $t=1$~Gyr for the run~4
982: (clump-cluster evolution) and run C-4 (stable spiral disk with the
983: same initial mass distribution). The clump-cluster redistributes the
984: initial flat disk into an exponential-like disk more efficiently
985: than density waves within this timescale. The snapshots of these two
986: runs are seen at $t=0.5$~Gyr, when the clumps are still present for
987: the unstable model.}\label{fig:control}\end{figure}
988:
989:
990: \end{document}
991: