1: % Use only LaTeX2e, calling the article.cls class and 12-point type.
2:
3: \documentclass[12pt,letterpaper]{article}
4:
5: \usepackage{scicite}
6: \usepackage[varg]{txfonts}
7: \usepackage{graphicx}
8:
9: \topmargin 0.0cm
10: \oddsidemargin 0.2cm
11: \textwidth 16cm
12: \textheight 21cm
13: \footskip 1.0cm
14:
15: %The next command sets up an environment for the abstract to your paper.
16: \newenvironment{sciabstract}{%
17: \begin{quote} \bf}
18: {\end{quote}}
19:
20:
21: % If your reference list includes text notes as well as references,
22: % include the following line; otherwise, comment it out.
23:
24: \renewcommand\refname{References and Notes}
25:
26: % The following lines set up an environment for the last note in the
27: % reference list, which commonly includes acknowledgments of funding,
28: % help, etc. It's intended for users of BibTeX or the {thebibliography}
29: % environment. Users who are hand-coding their references at the end
30: % using a list environment such as {enumerate} can simply add another
31: % item at the end, and it will be numbered automatically.
32:
33: \newcounter{lastnote}
34: \newenvironment{scilastnote}{%
35: \setcounter{lastnote}{\value{enumiv}}%
36: \addtocounter{lastnote}{+1}%
37: \begin{list}%
38: {\arabic{lastnote}.}
39: {\setlength{\leftmargin}{.28in}}
40: {\setlength{\labelsep}{.5em}}}
41: {\end{list}}
42:
43:
44: % Include your paper's title here
45:
46: \title{Lighting the Universe with filaments}
47:
48:
49: % Place the author information here. Please hand-code the contact
50: % information and notecalls; do *not* use \footnote commands. Let the
51: % author contact information appear immediately below the author names
52: % as shown. We would also prefer that you don't change the type-size
53: % settings shown here.
54: \author
55: {Liang Gao$^{1, \ast}$, Tom Theuns$^{1,2}$
56: \\
57: \\
58: \normalsize{$1$ Institute for Computational Cosmology, Durham
59: University, South Road, Durham DH1 3LE, UK}\\
60: \normalsize{$2$ Department of Physics, University of Antwerp,
61: Groenenborgerlaan 171, B-2020 Antwerpen, Belgium} \\
62: \normalsize{$^\ast$ E-mail: liang.gao@durham.ac.uk} \\
63: }
64: \date{}
65:
66:
67: %%%%%%%%%%%%%%%%% END OF PREAMBLE %%%%%%%%%%%%%%%%
68:
69: %%%%%%%%%%%%%%%% Authers' Defenision %%%%%%%%%%%%%%%%%
70:
71: \def\gsim{\mathrel{\rlap{\lower 4pt \hbox{\hskip 1pt $\sim$}}\raise 1pt \hbox {$>$}}}
72: \def\lsim{\mathrel{\rlap{\lower 4pt \hbox{\hskip 1pt $\sim$}}\raise 1pt \hbox {$<$}}}
73:
74: %%%%%%%%%%%%%%%%%
75:
76: \begin{document}
77:
78: % Double-space the manuscript.
79:
80: %\baselineskip24pt
81:
82: % Make the title.
83:
84: \maketitle
85:
86: % Place your abstract within the special {sciabstract} environment.
87:
88: \begin{sciabstract}
89:
90: The first stars in the Universe form when chemically pristine gas heats
91: as it falls into dark matter potential wells, cools radiatively due to
92: the formation of molecular hydrogen, and becomes self-gravitating. We
93: demonstrate with super-computer simulations that their properties
94: depend critically on the currently unknown nature of the dark matter.
95: If the dark matter particles have intrinsic velocities that wipe-out
96: small-scale structure, then the first stars form in filaments with
97: lengths of order the free-streaming scale, which can be $\sim
98: 10^{20}$~m ($\sim 3$~kpc, baryonic masses $\sim 10^7~$ solar masses)
99: for realistic \lq warm dark matter\rq~ candidates. Fragmentation of the
100: filaments forms stars with a range of masses which may explain the
101: observed peculiar element abundance pattern of extremely metal-poor
102: stars, while coalescence of fragments and stars during the filament's
103: ultimate collapse may seed the super massive black holes that lurk in
104: the centres of most massive galaxies.
105:
106: \end{sciabstract}
107:
108: Most of the matter in the Universe does not interact with light except
109: gravitationally. This \lq dark matter\rq~ is usually assumed to be \lq
110: cold\rq, meaning that its velocity dispersion is sufficiently small for
111: density perturbations imprinted in the early Universe to persist up to
112: very small scales. Although this model is able to describe the
113: large-scale distribution of galaxies in impressive detail, it may face
114: problems on the scale of galaxies and below, for example it may predict
115: too many satellite galaxies \cite{M99}, as well as too cuspy profiles
116: for the dark matter halos that surround galaxies\cite{DH01}.
117:
118: However, dark matter has yet to be detected in the laboratory, and
119: there exist many viable dark matter candidates from particle physics
120: that are not \lq cold\rq~. Warm dark matter (WDM) particles have
121: intrinsic thermal velocities and these motions quench the growth of
122: structure below a \lq free-streaming\rq~ scale (the distance over which
123: a typical WDM particle travels) which depends on the nature of the
124: particle. Because small and dense halos do not form below the
125: free-streaming-scale, the dark matter halos that surround galaxies in a
126: WDM model have far less substructure and are less concentrated as
127: compared to their cold dark matter (CDM) counterparts, which may help
128: alleviate both the satellite and galactic core problem
129: \cite{Bode01}. Structures on larger scales are similar in WDM and CDM
130: and therefore the distribution of galaxies is not affected. The first
131: generation of stars in the Universe forms when primordial gas gets
132: compressed by falling into small dark matter potential wells
133: \cite{ABN02, BCL02, Yoshida06, G07}. Because WDM affects structure
134: formation on such small scales it may influence how the first stars
135: form; we have performed simulations to analyse this in more detail.
136:
137: Large-scale power in the spectrum of density perturbations causes
138: progenitors of present-day clusters of galaxies to be amongst the first
139: objects to condense out of the initially almost smooth mass
140: distribution. We study the early formation stages of such an object by
141: identifying a massive cluster of galaxies in a dark matter simulation
142: of a large cosmological volume at redshift $z=0$, and use a multi-scale
143: technique\cite{G05, SO} to re-simulate its formation and evolution with
144: the cosmological hydrodynamics code Gadget-2 \cite{S05, SO}. Baryons
145: compressed by falling into the developing dark matter potential wells
146: cool radiatively through molecular hydrogen emission lines\cite{GP98,
147: SO}; we follow the formation of molecular hydrogen and its cooling
148: using a sophisticated chemistry network\cite{Yoshida06}. The parent
149: simulation, from which we picked the re-simulated region, followed the
150: growth of structure in a CDM dominated Universe with a cosmological
151: constant $\Lambda$ ($\Lambda$CDM); we performed the re-simulations
152: assuming both CDM and WDM as dark matter. In this WDM model, small
153: scale power in the density fluctuations is exponentially suppressed
154: below $\sim 100$~co-moving kpc, mimicking free-streaming of gravitino
155: particles with mass $m_{\rm WDM}=3$~keV\cite{Bode01}. Gravitinos are a
156: popular WDM candidate\cite{bhs05} but our results are insensitive to
157: the exact choice of WDM particle as long as its free-streaming length
158: is more than a few tens of co-moving kilo parsecs. Therefore even if
159: the gravitino were slightly more massive, or if the WDM were instead a
160: sterile neutrino (another popular WDM candidate\cite{neutrinos}), our
161: results would not change appreciably. Observations of the clustering of
162: neutral gas along sight lines to distant quasars (the Lyman-$\alpha$
163: forest of absorption lines) probe scales $1-40$~Mpc when the density
164: perturbations on these scales are still small, and the presence or
165: absence of significant substructure in these observations can constrain
166: the masses of WDM particles\cite{Viel06, Seljak06}; our choice of WDM
167: particle mass (3 keV, the corresponding mass-scale is $\sim 3\times
168: 10^8M_\odot$) is well above this lower limit ($\sim 2$ keV;
169: \cite{Viel06, Seljak06}). The initial amplitude of the imposed density
170: perturbations in our simulations is normalised to the level seen in the
171: Cosmic Microwave Background radiation\cite{wmap3}, and our simulations
172: start at redshift $z\sim 200$. (See the supplementary online material
173: for more details on these simulations.)
174:
175: The growth of structure in these re-simulations leads to a pattern of
176: filaments and sheets (Fig.1) familiar from the local large-scale
177: distribution of galaxies. This is because the assumed Gaussian spectrum
178: of density perturbations, appropriate for an inflationary model, leads
179: to collapse along one (sheet) and two directions (filament) before the
180: formation of halos. Although the large-scale filamentary pattern is
181: very similar in CDM and WDM, the structure of the filaments themselves
182: is very different: whereas the CDM filaments fragment into numerous
183: nearly spherical high density regions (\lq halos\rq; panel 1.a), the
184: WDM filaments are mostly devoid of such substructure (panel
185: 1.b). Panels 1.c and 1.d depict the WDM filament at an earlier time
186: before any of the other filaments have formed: the central density is
187: very high (hydrogen number density $n_{\rm h}\sim
188: 10^4\,$cm$^{-3}\approx 10^6\langle n_{\rm h}\rangle$) yet no dark
189: matter halo has formed yet. It is well known that the Poisson noise in
190: simulation codes that use particles to represent the dark matter leads
191: to spurious fragmentation of the filaments that form in such WDM
192: simulations\cite{Goetz02, W07}. We therefore end the analysis of our
193: WDM simulations well before filaments fragment.
194:
195: The length of the filament ($\sim 3$~kpc) is of order of the imposed
196: WDM free-streaming scale as expected, $L \sim 4$~kpc at redshift
197: $z=23.34$ when the Universe is $140$ million years old. Gas and dark
198: matter accrete perpendicular onto the filament's axis (Fig.~2). Dark
199: matter particles falling into the filament perform damped oscillations
200: as the potential well deepens. At $r\sim 50$~pc (where $r$ is the
201: distance perpendicular to the filament's axis) dark matter particles
202: falling into the well encounter particles that fell in from the other
203: side, and such successive instances of \lq orbit-crossing\rq~ give rise
204: to the steps in the density seen in Fig.~1b. Baryons do not undergo
205: orbit-crossing but the gas gets compressed to a temperature $T\approx
206: 7000$ K at $r\sim 20$~pc. Rapid build-up of ${\rm H}_2$ induces cooling
207: and the gas starts to dominate the matter density further downstream,
208: so that the ratio of gas to dark matter densities $\rho_b/\rho_{\rm
209: DM}\sim 15=100\langle\rho_{\rm b}\rangle/\langle\rho_{\rm DM}\rangle$
210: at $r=2$~pc. At $r<2$ pc where the gas dominates, the ratios of
211: principal axes of the filament are $b/a=0.123$ and $c/a=0.118$ hence
212: the filament is very nearly cylindrical. The properties of both the gas
213: and the dark matter are very uniform along the whole length of the
214: filament. The cylindrical density profile below $10$~pc is
215: approximately $\rho \propto r^{-2.8}$ for $2\le r\le 8$~pc and $\rho
216: \propto r^{-2}$ for $r<2$~pc. This contrasts with $\rho \propto
217: r^{-2.3}$ for the spherically averaged profile of the gas in CDM halos
218: on a comparable scale \cite{Yoshida06, G07}. The central ${\rm H}_2$
219: abundance reaches $10^{-3}$, higher than in the CDM case because of the
220: higher temperature reached behind the accretion shock. This higher
221: temperature, and the associated higher ionisation fraction, will also
222: enhance the importance of HD cooling at later stages \cite{HD,
223: Yoshida07}.
224:
225: The non-linear collapse into a thin filament found in these WDM
226: simulations is in sharp contrast to what happens in the CDM case.
227: There the first objects to reach high densities are discrete, nearly
228: spherical, dark matter halos (gravitationally bound concentrations of
229: dark matter) that form at tiny masses and build-up hierarchically
230: through mergers and accretion. Some halos have a sufficiently deep
231: potential well to accrete and shock baryons, enabling ${\rm H}_2$
232: formation and radiative cooling. Runaway collapse of the rapidly
233: accreting self-gravitating gas is thought to lead to the formation of a
234: single massive star per cooling halo \cite{ABN02, Yoshida06, G07}. The
235: absence of small scale power in WDM prevents halos from forming before
236: the filament itself becomes highly non-linear. Although our
237: re-simulations focus on the progenitor of a massive cluster, and hence
238: the forming object collapses unusually early on, the fact that very
239: high densities are reached in a filament, as opposed to a spherical
240: halo, is generic to WDM.
241:
242: The stability of collapsing filamentary clouds has been investigated in
243: the context of the formation of cloud cores by for example\cite{L85,
244: IM97}, and applied to early Universe filaments \cite{NU02}. The
245: inability of gas to cool sufficiently fast usually limits the collapse
246: time of the filament by the cooling time, $t_d\equiv\rho/\dot\rho\sim
247: t_c\equiv nk_{\rm Boltz}T/\Lambda$ (where $T$ is the temperature and
248: $\Lambda$ is the cooling rate due to H$_2$ and HD
249: cooling). Perturbations start growing, possibly leading to
250: fragmentation, when the dynamical time $t_d\gg t_p$, where $t_p$ is the
251: inverse growth rate of perturbations. If gas cooling is efficient,
252: $t_d\sim t_c\ll t_p$, and perturbations do not grow. At densities
253: $n=n_1\sim 10^5$~cm$^{-3}$ the level population of H$_2$ reaches LTE
254: making $\Lambda\propto n$ instead of $\propto n^2$ and $t_d$
255: increases\cite{GP98}. A sufficiently massive filament may yet survive
256: fragmentation at this stage, and at higher density $n\sim
257: 10^9$~cm$^{-3}$ three-body processes promote the formation of H$_2$ and
258: the cooling time decreases again. However when $n\geq n_2=
259: 10^{12}$~cm$^{-3}$ the gas becomes optically thick in the H$_2$ cooling
260: transitions slowing the collapse and the filament is once more in
261: danger of fragmenting. Collision induced continuum emission will again
262: decrease the cooling time when $n\sim 10^{14}$~cm$^{-3}$, until the gas
263: becomes optically thick also to this cooling radiation at densities
264: $n\geq n_3= 10^{16}$~cm$^{-3}$ \cite{Omukai98, Yoshida07}. The physics
265: of the H$_2$ molecule therefore sets three densities at which the
266: filament may fragment. The typical fragment masses are of order tens of
267: solar masses, solar masses, and sub-solar masses, for fragmentation at
268: densities $n_1$, $n_2$ and $n_3$, respectively\cite{NU02}.
269:
270: The tidal field around the filament breaks the cylindrical symmetry on
271: scales comparable to the filament's length (Fig.~1), and can trigger
272: the gas dynamical instabilities that ultimately lead to
273: fragmentation. Unfortunately our current simulations are not able to
274: follow this process in detail because these tiny deviations from
275: symmetry are overpowered by numerical noise caused by the graininess of
276: the particle distribution. This causes the filament to fragment very
277: rapidly as expected, yet the scale of the fragmentation is
278: artificial. In the WDM universe, the small-scale perturbations that
279: trigger fragmentation in the simulations are not present and need to be
280: generated through transfer of power from larger scales. Since this is a
281: relatively slow process, the central density will have reached the
282: higher value $n_2$ or even higher, implying small fragment masses of
283: order of a solar mass or below. Such fragments can coalesce to form
284: more massive clumps, as demonstrated in 2D numerical
285: simulations\cite{IM97}. Even if individual cores survive, they may
286: still grow in mass through accretion.
287:
288: Detailed observations of star-forming clouds in the Milky Way reveal
289: that the low-mass stellar mass function is very similar to that of the
290: dense, pre-stellar cloud cores within the ambient cloud, although it is
291: not yet clear what determines the cut-off at high masses. This close
292: similarity might indicate that the process of cloud fragmentation plays
293: an important role in determining the initial mass
294: function\cite{Larson05}. If this also applies to star-formation in a
295: WDM filament, then it is plausible that fragmentation will lead to a
296: burst of star formation which includes low-mass stars with masses $\sim
297: 1~M_\odot$ or below, but also much more massive stars built through
298: mergers and accretion. The more massive stars will end their short
299: lives through supernova explosions or collapse to form intermediate
300: mass black holes, and some of the very low mass stars may potentially
301: survive until today. Although the details of this scenario are
302: uncertain, it is clear that the stellar mass function will be quite
303: different from the CDM case.
304:
305: The low-mass Milky Way stars HE 0107-5240 \cite{CH02} and HE 1327-2326
306: \cite{F05} have extremely low metallicities\footnote{The notation [X]
307: $\equiv \log({\rm X}/{\rm X}_0)$, characterises the mass fraction of
308: element X in terms of the value in the Sun, X$_0$.} of [Fe/H] $\sim$
309: $-5$ and peculiar element abundances, for example [C/Fe]$>$1. An
310: initial mass function of zero metallicity first stars with a range of
311: masses, as we suggested might be the case in WDM, could explain these
312: stars either as due to self-pollution or pre-enrichment by intermediate
313: mass (tens of solar masses) supernovae\cite{BC05}. Such first
314: generation low and intermediate mass stars are not thought to form in
315: CDM\cite{ABN02, Yoshida06, G07}. Therefore, although the present
316: observational evidence is not yet unambiguous, a future detection of
317: zero-metallicity low-mass first stars may indicate that the dark matter
318: is warm.
319:
320: Free-streaming of the WDM decreases significantly the number density
321: of halos that host early star formation, which could delay
322: reionization as compared to CDM\cite{Y03}. However the additional
323: mode of star formation in filaments could partly compensate, making
324: it unclear whether reionization is indeed delayed when the dark
325: matter is warm.
326:
327: What is the ultimate fate of such $\sim 10^7\,M_\odot$ filaments?
328: Eventually the filament will collapse along its long axis, and since
329: its mean density is very high a significant number of collisions
330: between cloud cores and stars would appear inevitable. Such collisions
331: could build-up a massive object which can seed the formation of the
332: super-massive black holes that power redshift $z \sim 6$ quasars, and
333: appear to lurk in the centers of most large galaxies today.
334:
335: The different outcome of WDM versus CDM is because the thermal
336: velocities of the WDM particles prevent halos from forming before the
337: filaments themselves form stars. This happens when the free-streaming
338: length introduced by the dark matter's thermal velocities is more than
339: a few tens of co-moving kilo parsecs so that the filaments are
340: sufficiently massive to heat infalling gas to more than $\sim $1000~K,
341: enabling efficient cooling by molecular hydrogen. If we simulate a WDM
342: Universe with much shorter free-streaming length, for example $\sim
343: 20$~kpc corresponding to a gravitino mass $m_{\rm WDM}=15$~keV, star
344: formation proceeds in a similar way to the CDM case, in good agreement
345: with earlier work\cite{ON06}. The likely very different initial mass
346: function of the first stars and the rapid formation of massive black
347: holes in a WDM scenario with $m_{\rm WDM}\sim 3$~keV, as opposed to
348: CDM, implies a very different early thermal and metal enrichment
349: history of the Universe, greatly affecting subsequent galaxy
350: formation. It appears therefore that the way in which quasar, star and
351: galaxy formation started, depends strongly on the nature of the dark
352: matter.
353: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
354:
355: \begin{thebibliography}{10}
356:
357: \bibitem{M99}
358: B. Moore, S. Ghigna, F. Governato, G. Lake, T. Quinn T., J. Stadel,
359: P. Tozzi, {\it The Astrophys. J.}, {\bf 524}, L19 (1999)
360:
361: \bibitem{DH01}
362: J. J. Dalcanton, C. J. Hogan, {\it The Astrophys. J.}, {\bf 561}, 35
363: (2001)
364:
365: \bibitem{Bode01}
366: P. Bode, J. P. Ostriker, N. Turok, {\it The Astrophys. J.},{\bf 556},
367: 93 (2001)
368:
369: \bibitem{ABN02}
370: T. Abel, G. L. Bryan, M. L. Norman,{\it Science}, {\bf 295}, 93 (2002)
371:
372: \bibitem{BCL02}
373: V. Bromm, P. S. Coppi \& R. B. Larson, {\it The Astrophys. J.}, {\bf
374: 564}, 23 (2002)
375:
376: \bibitem{Yoshida06}
377: N. Yoshida, K. Omukai, L. Hernguist,T. Abel, {\it The Astrophys. J.},
378: {\bf 652}, 6 (2006)
379:
380:
381: \bibitem{G07}
382: L. Gao, N. Yoshida, T. Abel, C. S. Frenk, A. Jenkins, V. Springel,
383: {\it Mon. Not. R. Astron. Soc.}, {\bf 378}, 449 (2007)
384:
385: \bibitem{G05}
386: L. Gao , S. D. M. White, A. Jenkins, C. S. Frenk, V. Springel, {\it
387: Mon. Not. R. Astron. Soc.}, {\bf 363}, 379 (2005)
388:
389: \bibitem{SO} See supporting material on Science on line.
390:
391: \bibitem{S05}
392: V. Springel, {\it Mon. Not. R. Astron. Soc.}, {\bf 364}, 1105 (2005)
393:
394: \bibitem{GP98}
395: D. Galli, F. Palla, {\it Astron. \& Astrophys.}, {\bf 335}, 403 (1998)
396:
397: \bibitem{bhs05}
398: G. Bertone, D. Hooper, J. Silk, Particle dark matter: evidence,
399: candidates and constraints, {\it Physics Report}, {\bf 405}, 279 (2005)
400:
401: \bibitem{neutrinos}
402: S. Dodelson, L. M. Widrow, {\it Phys. Rev. Lett.}, {\bf 72}, 17 (1994)
403:
404: \bibitem{Viel06}
405: M. Viel, J. Lesgourgues, M. G. Haehnelt, S. Matarrese, A. Riotto, {\it
406: Phys. Rev. Lett.}, {\bf 97}, 071301 (2006)
407:
408: \bibitem{Seljak06}
409: U. Seljak., A. Makarov, P. McDonald, H. Trac, {\it
410: Phys. Rev. Lett.},{\bf 97}, 191303 (2006)
411:
412:
413: \bibitem{wmap3}
414: D. N. Spergel {\it et al.}, {\it The Astrophys. J.} in press, preprint
415: astro-ph/0603449 (2007)
416:
417: \bibitem{Goetz02}
418: M. G\"otz, J. Sommer-Larsen, {\it Astron. \& S. Sci.}, {\bf 281}, 415
419: (2002)
420:
421: \bibitem{W07}
422: J. Wang, S. D. M. White, preprint, astro-ph/0702575 (2007)
423:
424: \bibitem{HD}
425: J. L. Johnson, V. Bromm, {\it Mon. Not. R. Astron. Soc.}, {\bf
426: 366}, 247 (2006)
427:
428: \bibitem{Yoshida07}
429: N. Yoshida, S.~P. Oh, T. Kitayama, L. Hernquist, {\it The
430: Astrophys. J.}, {\bf 663}, 687 (2007)
431:
432: \bibitem{L85}
433: R. B. Larson, {\it Mon. Not. R. Astron. Soc.}, {\bf 214}, 379 (1985)
434:
435: \bibitem{IM97}
436: S. Inutsuka, S. M. Miyama, {\it The Astrophys. J.}, {\bf 480},
437: 681 (1997)
438:
439: \bibitem{NU02}
440: F. Nakamura, M. Umemura, {\it The Astrophys. J.}, {\bf 569}, 549 (2002)
441:
442: \bibitem{Omukai98}
443: K. Omukai, R. Nishi, {\it The Astrophys. J.}, {\bf 508}, 141 (1998)
444:
445: \bibitem{Larson05}
446: R.~B. Larson, {\it Mon. Not. R. Astron. Soc.}, {\bf 359}, 211 (2005)
447:
448: \bibitem{CH02}
449: N. Christlieb, M. S. Bessell, T. C. Beers, B. Gustafsson, A. Korn,
450: P. S. Barklem, T. Karlsson, M. Mizuno-Wiedner, S. Rossi, {\it
451: Nature}, {\bf 419}, 904 (2002)
452:
453: \bibitem{F05}
454: Frebel {\it et al.}, {\it Nature}, {\bf 434}, 871 (2005)
455:
456:
457: \bibitem{BC05}
458: T. C. Beers, N. Christlieb, {\it Annu. Rev. Astron. Astrophys.}, {\bf
459: 43}, 531 (2005)
460:
461:
462: \bibitem{Y03}
463: N. Yoshida, A. Sokasian, L. Hernquist, V. Springe, {\it The
464: Astrophys. J.}, {\bf 591}, 1 (2003)
465:
466: \bibitem{ON06}
467: B. W. O'Shea, M. L. Norman, {\it The Astrophys. J.}, {\bf 648}, 31 (2006)
468: \end{thebibliography}
469:
470: % Following is a new environment, {scilastnote}, that's defined in the
471: % preamble and that allows authors to add a reference at the end of the
472: % list that's not signaled in the text; such references are used in
473: % *Science* for acknowledgments of funding, help, etc.
474:
475: \begin{scilastnote}
476: \item It is a pleasure to thank Carlos Frenk, Richard Bower, Cedric Lacey,
477: Adrian Jenkins, Shude Mao, Volker Springel and Naoki Yoshida for
478: helpful discussions. All computations were performed on the Cosmology
479: machine of the Institute for Computational Cosmology at Durham
480: University. This work was supported by a PPARC rolling grant.
481: \end{scilastnote}
482:
483: \clearpage
484:
485: \begin{figure*}
486: \resizebox{14cm}{!}{\includegraphics{fig1.eps}}
487: \label{fig:filaments}
488: \end{figure*}
489: %\clearpage
490: \noindent Figure 1\\ {\it Top panels:} Dark matter density structure of
491: the progenitor of a redshift $z=0$ massive cluster of galaxies at
492: $z=15$ when the thermal velocities of the dark matter particles are
493: negligible (CDM; panel a) and when the dark matter is \lq warm\rq~
494: (WDM; a gravitino with $m_{\rm WDM}=3$~keV; panel b). Although both
495: models produce a characteristic filamentary pattern in the density, the
496: CDM filaments fragment into numerous nearly spherical halos whereas
497: free-streaming of the WDM prevents such substructure from forming.
498: {\it Bottom panels:} Gas temperature (panel c) and density (panel d) in
499: the WDM filament indicated by the red box in panel b, at an earlier
500: redshift ($z=23.34$) when only this filament had formed. Gas accretes
501: very uniformly onto the filament as indicated by the velocity vectors,
502: heats as it gets compressed, but further downstream cools due to the
503: formation of H$_2$, making the centre of the filament cold and
504: dense. The filament shown here is almost perfectly cylindrical, more
505: generally they are elliptical in shape.
506: \clearpage
507:
508: % Figure 2:
509: \begin{figure*}
510: \hspace{0.13cm}\resizebox{8cm}{!}{\includegraphics{fig2a.ps}}
511: \hspace{0.13cm}\resizebox{8cm}{!}{\includegraphics{fig2b.ps}}
512: \hspace{0.13cm}\resizebox{8cm}{!}{\includegraphics{fig2c.ps}}
513: \hspace{0.13cm}\resizebox{8cm}{!}{\includegraphics{fig2d.ps}}
514: \label{fig:fig2}
515: \end{figure*}
516: %\clearpage
517: \noindent Figure 2\\ Gas and dark matter profiles of the WDM filament
518: shown in panels c and d of Fig.~1. Top left: accretion of dark matter
519: particles (black) and gas particles (red); note the different
520: horizontal and vertical scales (along and perpendicular to the
521: filament, respectively). The dark matter undergoes orbit-crossing
522: (multi-valued velocities) at $r\sim 50$~pc, whereas the gas gets
523: shocked much closer in ($\sim 20$~pc) and dominates further
524: downstream. Top right: gas density profile as measured perpendicular to
525: the filament's axis. Lines $\rho\propto r^{-2}$ and $\rho\propto
526: r^{-2.8}$ are drawn to guide the eye. Bottom left: temperature profile:
527: gas heats as it gets compressed, but the rapid build-up of ${\rm H}_2$
528: cools the gas further in. Bottom right: Build-up of molecular hydrogen
529: fraction, $f_{H_2}$, during the accretion.
530:
531: \input{SOM/som.tex}
532: \end{document}
533: