astro-ph0503330/ms.tex
1: %\documentclass[12pt,preprint]{aastex}  % for e-submission to ApJ
2: 
3: \documentclass[manuscript]{emulateapj}  % emulateapj format
4: 
5: \usepackage{graphicx}
6: 
7: \def\etal{{\sl et al.}}
8: \def\kms{{\rm km/s}}
9: \def\Ms{{\rm M_\odot}}
10: 
11: \begin{document}
12: 
13: \title{Forming a Primordial Star in a Relic HII Region}
14: \author{Brian W. O'Shea\altaffilmark{1,2,3}, Tom Abel\altaffilmark{4},
15: Dan Whalen\altaffilmark{1,2,3} \& Michael L. Norman\altaffilmark{1}}
16: 
17: \altaffiltext{1}{Center for Astrophysics and Space Sciences,
18: University of California at San Diego, La Jolla, CA 92093, U.S.A.
19: Email: bwoshea, mnorman, dwhalen@cosmos.ucsd.edu}
20: \altaffiltext{2}{Department of Physics, University of Illinois in
21: Urbana-Champaign}
22: \altaffiltext{3}{Theoretical Astrophysics (T-6), Los Alamos National
23: Laboratories}
24: \altaffiltext{4}{Kavli Institute for Particle Astrophysics and Cosmology,
25:  Stanford University.
26: Email: tabel@stanford.edu}
27: 
28: \begin{abstract}
29: There has been considerable theoretical debate over whether photoionization
30: and supernova feedback from the first Population III stars facilitate or 
31: suppress the formation of the next generation of stars.  We present results 
32: from an Eulerian adaptive mesh refinement simulation demonstrating the 
33: formation of a primordial star within a region ionized by an earlier nearby star.  
34: Despite the higher temperatures of the ionized gas and its flow out of the dark 
35: matter potential wells, this second star formed within 23 million years of its 
36: neighbor's death.  The enhanced electron fraction within the HII region catalyzes 
37: rapid molecular hydrogen formation that leads to faster cooling in the subsequent 
38: star forming halos than in the first halos.  This ``second generation'' 
39: primordial protostar has a much lower accretion rate because, unlike the first 
40: protostar, it forms in a rotationally supported disk of 
41: $\sim 10-100~\Ms$.  This is primarily due to the much higher angular momentum of 
42: the halo in which the second star forms.  In contrast to previously published 
43: scenarios, such configurations may allow binaries or multiple systems of 
44: lower mass stars to form.  These first 
45: high resolution calculations offer insight into the impact of feedback upon  
46: subsequent populations of stars and clearly demonstrate how primordial chemistry 
47: promotes the formation of subsequent generations of stars even in the presence of
48: the entropy injected by the first stars into the IGM.
49: \end{abstract}
50: 
51: \keywords{cosmology: theory --- stars: formation --- hydrodynamics}
52: 
53: \maketitle
54: 
55: \section{Motivation}\label{motivate}
56: 
57: Calculations performed by Abel, Bryan and Norman (2002; hereafter ABN02)
58: show that rapid accretion rates driven by molecular hydrogen cooling cause 
59: the formation of solitary massive protostars in the range of $30$ to $300~\Ms$ 
60: in minihalos of $10^5-10^6~\Ms$ at redshifts $\ga 20$.  Simulations indicate 
61: that the hard UV spectra of these 10$^5$ K zero-metallicity stars 
62: will envelop them in large HII regions several kiloparsecs in diameter (Whalen, 
63: Abel \& Norman 2004; Kitayama et al. 2004). Over the main sequence lifetime of 
64: the central star (on the order of 2-6 Myr for the range of
65: $30-300~\Ms$) half of the baryons within the minihalo are 
66: driven beyond its virial radius by ionized flows that quickly steepen into shocks.  
67: These shocks exhibit expansion rates of up to ten times the escape velocity of the 
68: halo. After the death of the central star, cooling and recombination are out of 
69: equilibrium in the ionized gas, which results in significant electron fractions
70: even after its temperature has dropped to 1000 - 2000 K after 20 - 50 Myr. One 
71: dimensional, nonrotating calculations (Heger et al. 2003) predict two possible 
72: fates for the primordial stars themselves:  complete destruction by the pair 
73: instability ($140~\Ms < M_* < 260~\Ms$) which is very energetic and leaves no 
74: remnant, or direct collapse to black holes above and below this mass range, with 
75: the added possibility of SN-like precollapse mass ejections by 
76: pulsational pair instabilities from 100-140 $\Ms$ stars (Heger \& Woosley 2002).
77: 
78: An important question is whether later generations of stars can efficiently form
79: in the relatively high temperatures and ionization fractions of the relic HII 
80: regions left by the first stars.  One analytical study (Oh \& Haiman 2003) found 
81: that the first stars injected sufficient entropy into the early IGM by photoheating 
82: and supernova explosions to prevent further local star formation in their vicinity.  
83: Lyman-Werner SUV background radiation is also thought to have contributed negative 
84: feedback by photodissociating primordial H$_2$ and quenching the molecular hydrogen 
85: cooling processes allowing the first stars to form (Haiman, Abel \& Rees 2000; Machacek,
86: Bryan \& Abel 2001).  In this \textit{Letter} we present 
87: fully resolved simulations that show a second primordial star can form in the relic HII region 
88: of an earlier Pop III star.  We determine its properties, considering the effect of 
89: Lyman-Werner radiation from the resultant black hole assuming accretion rates 
90: consistent with the density fields left by ionized outflows from the parent minihalo.
91: 
92: 
93: \section{Simulation Setup}\label{setup}
94: 
95: We carried out simulations using Enzo, a publicly-available Eulerian adaptive 
96: mesh refinement (AMR) hydrodynamics + N-body code (Bryan \& Norman 1997; O'Shea 
97: et al. 2004; also see http://cosmos.ucsd.edu/enzo). We initialized a box of 
98: size $300~h^{-1}$~kpc at $z = 99$ for a cosmology with $(\Omega_M,\ \Omega_
99: \Lambda,\ \Omega_B,\ h,\ \sigma_8,\ n)=(0.3, 0.7, 0.04, 0.7, 0.9, 1)$.  We 
100: first ran a simulation with $128^3$ dark matter particles in a $128^3$ root 
101: grid with 6 total levels of adaptive mesh, refining on a dark matter overdensity 
102: of 4.0.  This model was run with dark matter alone in order to identify the most 
103: massive halo that evolves in the simulation volume, which at $z \sim 18$ had a 
104: mass $\sim 5 \times 10^5 M_{\odot}$.        
105: 
106: We then re-initialized the calculation in the original simulation volume at $z = 
107: 99$ with both baryons and dark matter using a $128^3$ root grid and three 
108: static nested subgrids, each of which was twice as refined as its parent grid 
109: and was centered on the Lagrangian volume of the peak that later evolved into 
110: the identified halo.  The effective root grid resolution was $1024^3$ in this 
111: volume, which corresponds to a comoving spatial resolution of $\sim 300~h^{-1}$ 
112: ~pc and a dark matter particle mass of $1.8~h^{-1}~M_{\odot}$ in the most highly 
113: refined region. Every dark matter particle that later enters into dark matter
114: halos of interest was within this most highly refined grid at the start of the 
115: simulation.
116: 
117: We started the simulation with this set of initial conditions at $z =
118: 99$ and followed the collapse of the first star, which occurred at a redshift
119: of 17.76. As a refinement criteria we used a baryon overdensity of 4.0
120: and a dark matter overdensity of 8.0.  In addition, to ensure appropriate
121: simulation resolution we mandated that the Jeans length must be resolved by
122: at least 16 cells at all times, which exceeds the Truelove criterion by a
123: factor of 4 along each axis (Truelove et al. 1998). At the collapse redshift 
124: the three dimensional structure was resolved with 8727 grids on nine levels 
125: containing a total of 49,641,744 unique resolution elements. 
126: 
127: To compute the extent of the HII region of the $120~\Ms$ Pop III star assumed to 
128: form in the collapse, we interpolated the density, energy, and velocity fields 
129: from the entire Enzo simulation volume at the formation redshift of this star 
130: onto a three dimensional grid of fixed resolution with $256^3$ cells for import 
131: into a static radiative transfer code.  The code utilizes the ionization front 
132: tracking technique of Abel~(2000) to calculate the boundary of the HII region 
133: along rays cast outward from the central star by the adaptive ray tracing technique 
134: of Abel \& Wandelt (2002).  
135: Within the HII region we set the ionization fraction to unity and the 
136: H$_2$ and H$^-$ fractions to zero.  We assume that the mean energy of ionization 
137: for the gas is 2.4 eV, which results in a post-ionization temperature of 
138: $\sim 18,000$~K when calculated in our multispecies ZEUS simulations.  This is
139: somewhat cooler than one might expect due to the relatively hard spectrum of
140: massive primordial stars, and is a result of our use of monochromatic radiative
141: transfer in the ZEUS code, which underestimates the UV photoheating of the halo
142: by not taking into account contributions from very high energy photons.  Whalen et
143: al. (2004) show that an increase in post-front temperatures
144: results in somewhat higher sound speeds.  These yield higher shock speeds that
145: promote the photoevaporative flow of gas from the halo in which the first star is 
146: formed and could in principle affect the dynamics of nearby halos.  We show 
147: below that in this case the outflow of gas has
148: a negligible effect on the formation of a second primordial star, which suggests
149: that our result is at worst only weakly affected by post-front temperature.
150: Higher post front temperatures will not significantly retard the cooling and 
151: recombination crucial to the formation of molecular hydrogen.
152: 
153: We approximated the dynamics of the HII region by imposing the one dimensional 
154: velocity, ionization, density and temperature profiles for a $120~\Ms$ star at
155: the end of its main sequence lifetime from Whalen et al. (2004) along every line 
156: of sight from the central star. We modified baryon densities and velocities out to 
157: $\sim 120$ pc (corresponding to the location of the shock wave in the 1D 
158: calculation) but changed only ionization fractions and temperatures beyond this
159: radius out to the boundary of the HII region determined by the ray tracing code.
160: We then mapped this HII region onto the full hierarchy of grids in the Enzo 
161: calculation, centering it on the location of the first protostar. This state 
162: corresponds to only 2.5 million years after the initial star formed (z $\simeq 
163: 17.4$), so we assume that instantaneous ionization is a reasonable approximation
164: for all gas outside the first halo (which has had the hydro profiles from the 1D
165: simulations imposed in it).  An important question is whether the satellite halos
166: are also ionized by the I-front propagating outward from the first star, 
167: an issue investigated in detail at later redshifts by Shapiro et al. (2004).  
168: Simulations we performed in 1D in ZEUS-MP indicate that the neighboring halos 
169: are photoionized by the parent star by the end of its main sequence lifetime.
170: 
171: We then continued the simulation until the collapse of the next protostar, 
172: which occurs at $z = 16.44$, 22.8 million years later. The final time that we 
173: analyzed contains 10,710 grid patches on 24 levels with 54,996,560 unique resolution 
174: elements.  In this calculation we neglect the pulsational pair instability that may eject 
175: the hydrogen envelope for this star (Heger \& Woosley 2002).  
176: 
177: As a check on our simulation setup we also ran a simulation where we simply
178: instantaneously ionized the entire simulation volume by raising the baryon temperature
179: to $\sim 10,000$ K and setting ionization fractions to one and $H_2$ fractions to zero.  This
180: simulation tests whether the addition of the one dimensional radial profiles from the Whalen 
181: et al. (2004) calculations changed the properties of the second protostar appreciably.  
182: We find that the collapse time and accretion rate of the protostar formed in this 
183: simulation are essentially identical to the results of our full setup, and only discuss
184: results from the full setup in the rest of this \textit{Letter}.
185: 
186: \section{Results} 
187: The second primordial protostar forms in a neighboring minihalo approximately 265 
188: proper parsecs from the location of the halo in which the first star formed (and where the
189: HII region originated). The halo in which this second protostar forms was completely 
190: ionized by the first star to a temperature of $\sim 
191: 10^4$ K.  Due to its relatively high density, the center of this halo cools very 
192: rapidly and molecular hydrogen formation is catalyzed by the extremely high electron 
193: fraction.  After only a few million years the core of the halo has a molecular 
194: hydrogen fraction of $\sim 5 \times 10^{-3}$, well above what one would expect for a 
195: halo which has not been ionized.  This halo is significantly smaller than the first: 
196: $\sim 2 \times 10^5~\Ms$ rather than $\sim 5 \times 10^5~\Ms$.
197: 
198: \subsection{Comparison of the First and Second Stars}\label{secondprop}
199: Figure~\ref{maccrete} compares the mass accretion times of the initial and
200: second Population III stars formed in this simulation.  In addition, this
201: figure shows the mass accretion time of the halo in ABN02 and an estimate
202: of the Kelvin-Helmholz timescale as a function of mass, using values of 
203: luminosity and effective temperature taken from Schaerer (2002). The upper 
204: and lower dotted lines correspond to an object with constant accretion rates 
205: of $10^{-3}$ and $10^{-2}~\Ms/$year, respectively. Our calculation of accretion 
206: timescales for the initial protostar agrees well with that of ABN02.  The fact 
207: that the two results are in good agreement even though the ABN02 calculations 
208: assumed a lower baryon fraction supports the analysis of Ripamonti and Abel (2004) 
209: showing that all mass scales in these calculations are set by molecular physics. 
210: Comparison of the accretion rates to the Kelvin-Helmholz timescale provides an 
211: estimate of $\sim200~\Ms$ for the upper bound of the mass of the star.  The 
212: accretion timescales suggest a reasonable lower bound of $\sim 80~\Ms$, since this 
213: much gas will accrete in $10^4$ years, an insufficient time for fusion to begin. In 
214: contrast, the accretion rate of the second protostar is over an order of magnitude 
215: lower. This is because the second protostar has a much more pronounced thick disk 
216: structure than the first protostar. The disk is rotationally supported past a 
217: radius $\sim 0.01$ pc (corresponding to an enclosed mass of $\sim 10~\Ms$), whereas 
218: the disk around the first star in the volume is not. Similar accretion timescale 
219: arguments as before suggest a mass of $\sim 5-20~\Ms$ for the second star, although 
220: accretion physics will ultimately determine the true mass, particularly given the 
221: presence of this more pronounced disk.  
222: 
223: \begin{figure}
224: \resizebox{3in}{!}{\includegraphics{f1.eps}}
225: \caption{Mass accretion time $t_a = M(r)/\dot{M} \equiv M(r)/(4\pi r^2\rho
226: v_r)$~as a function of enclosed gas mass. This is at the final output
227: corresponding to $z=16.437$. The dashed line is the corresponding data dump
228: of the initial star which had formed at $z=17.67$. The red dashed line
229: corresponds to the first star to form in this simulation.  The blue
230: dot-dashed line corresponds to the first star calculated in ABN02.  The
231: solid black line corresponds to the second star forming in this simulation,
232: and the green long-dashed line corresponds to the Kelvin-Helmholz time of a
233: representative star.  The upper and lower black dotted lines correspond to
234: objects with constant mass accretion rates of $10^{-2}$ and
235: $10^{-3}~\Ms/$yr, respectively.}
236: \label{maccrete}
237: \end{figure}
238: 
239: Examination of the net angular momentum of the two halos
240: is illuminating.  The angular momentum of a cosmological halo can be described
241: by the dimensionless spin parameter: $\lambda \equiv J |E|^{1/2}/G M^{5/2}$
242: where J is angular momentum, E is the total energy, G is the gravitational constant 
243: and M is the halo angular momentum.  This is roughly equivalent to the ratio of 
244: the angular momentum in the halo to the angular momentum needed for the halo
245: to be completely rotationally supported.  (Padmanabhan 1993)  Typical values of
246: the spin parameter for cosmological halos are $\sim 0.02 - 0.1$, 
247: with a mean of $\lambda \simeq 0.05$  
248: (Barnes \& Efstathiou 1987; Gardner 2001).  We find that the halo
249: in which the first primordial protostar forms has a spin parameter for the gas and
250: dark matter of 
251: $(\lambda_{gas} , \lambda_{dm}) = (0.0275, 0. 0363)$, which is slightly lower than
252: the mean.  The spin parameter of the second halo is 
253: $(\lambda_{gas} , \lambda_{dm}) = (0.1079, 0.1607)$, which is atypically high.
254: Examination of the evolution of angular momentum in the gas of the halos as the 
255: two protostars form shows that the angular momentum distributions are different
256: in the two clouds, and if angular momentum is conserved one would expect to see
257: a centrifugally supported disk that is approximately four times larger in the 
258: second halo.
259: 
260: \subsection{Black Hole Accretion}\label{bhaccrete}
261: 
262: Here we consider whether accretion onto a relic black hole could generate enough
263: photodissociative radiation to inhibit H$_2$ formation in the second star's halo.
264: We assume Bondi-Hoyle accretion (Bondi \& Hoyle 1944) for the 120 $\Ms$ black 
265: hole that forms after the collapse of the first star to estimate the Lyman-Werner 
266: flux from its accretion.   This rate depends on the mass of the accretor as 
267: well as the local gas temperatures, densities, and relative velocities it 
268: encounters.  To sample the local environment the black hole would traverse over 
269: the duration of the simulation, we followed the 40 dark matter particles closest 
270: to the first protostar (within $\sim 0.1$ proper pc) from the end of its main 
271: sequence lifetime until the collapse of the second protostar.  We tallied the cell
272: quantities they crossed to compile the accretion rate history each particle would 
273: have if it were the black hole.  The histories for the 40 black hole proxies appear 
274: in Figure~\ref{bhacc}. The mass accretion rates grow from $10^{-11}~\Ms$/yr to 
275: $10^{-8.5}~\Ms$/yr for most of the tracer particles.  
276: 
277: \begin{figure}
278: \resizebox{3in}{!}{\includegraphics{f2.eps}}
279: \caption{Bondi-Hoyle mass accretion rate around the black hole calculated
280: from the local gas temperature, density and relative velocity.  Integration
281: of these curves lead to estimates of growth of the black hole (initially
282: $120~\Ms$) of that range from 0.009 to 0.032~$\Ms$ over 23 Myrs}
283: \label{bhacc}
284: \end{figure}
285: 
286: To estimate the effect of Lyman-Werner radiation from the black hole on molecular 
287: hydrogen formation in nearby halos we assume a canonical 10\% radiative efficiency
288: for the accretion.  The uppermost accretion curve yields $2.2 \times 10^{37} 
289: (M/100~\Ms)$ erg/s ($\sim 4500~L_\odot$) for an upper limit to the total luminosity
290: (which is much lower than the Eddington luminosity of this object, $1.5 \times 
291: 10^{40}$ erg/s, or $\sim 4 \times 10^6~L_\odot$).  Taking this to be a blackbody
292: spectrum, the flux in the Lyman-Werner band (11.1-13.6 eV) reaching the second 
293: protostar is $\sim 1.6 \times 10^{-25} (M/100~\Ms)$ erg s$^{-1}$ cm$^{-2}$ Hz$^{-1}$, 
294: resulting in photodissociation rates that are significantly lower than the formation 
295: rates of molecular hydrogen there.  The expulsion of gas by ionized flows from the 
296: first halo prevents higher accretion rates and greater Lyman-Werner fluxes. 
297: A star in this mass range may shed its envelope just prior to collapse, 
298: resulting in a smaller black hole and making the results discussed here an upper limit.
299: 
300: 
301: \section{Discussion}\label{discuss}
302: 
303: This first high resolution three dimensional simulation of the evolution of gas 
304: within a primordial HII region demonstrates the crucial role of H$_2$ chemistry 
305: driven by photoionization in the 
306: formation of the next generation of stars.  While this has been addressed in 
307: previous work (Ricotti, Gnedin \& Shull 2002) our simulations are the first with
308: sufficient resolution to directly examine the formation of individual stars.
309: Further investigation will be necessary 
310: to determine if the lower accretion rates leading to the smaller mass of the second 
311: star are a coincidental feature of this calculation or a general trend of early star 
312: formation in halos preprocessed by HII regions.  The low accretion 
313: rate that we observe in this calculation is primarily due to the high initial angular momentum of
314: the second halo.   
315: 
316: One possible source of error lies in the method and assumptions determining 
317: whether the neighboring halos are photoionized.  While our 1D results indicate 
318: that these halos will be ionized, this issue merits further investigation 
319: with fully 3D simulations.
320: We further assume that this ionization occurred instantaneously and simply 
321: ionize the gas outside of the initial halo without changing the total density 
322: or velocity profiles of nearby halos. Instantaneous ionization appears to be a
323:  reasonable approximation since the sound crossing time of all of the 
324: ionized halos is longer than the main-sequence lifetime of the parent star.  
325: Again, full 3D radiation photo-evaporation simulations will be necessary to determine whether
326: the hydrodynamic evolution of these halos during the main sequence lifetime of
327: the parent star is unimportant.
328: 
329: We note that our HII region 
330: enveloped roughly a dozen minihalos similar to the one that formed the second star.  More
331: calculations will be required to see if these too form stars.  The evolution of 
332: the massive disk also merits examination to ascertain whether it breaks up into 
333: a multiple system or fully accretes to form a single star.  
334: The situation realized
335: in our cosmological simulation may lead to objects with initial conditions similar to 
336: the cases studied by Saigo et al. (2004).
337: Lower mass second 
338: generation stars or the possibility of binaries or multiple systems of primordial 
339: stars would have strong implications for the observability of such objects and
340: their impact on subsequent structure formation.  Less massive stars might have 
341: different nucleosynthetic signatures than those of the pair-instability supernovae 
342: that may occur in the first generation of primordial stars.  The immense size of 
343: early HII regions could also make the scenario of primordial stars forming in a 
344: relic HII region much 
345: more common than extremely massive stars forming in pristine halos.  These two facts
346: taken together may account for the lack of detection of the characteristic odd-even 
347: abundance pattern from pair-instability supernovae expected in observations of ultra 
348: metal poor halo stars (Umeda \& Nomoto 2005 and references therein). How HII regions 
349: from the first stars may regulate local star formation by suppressing the collapse of 
350: gas in local halos which have not reached relatively high densities also remains to be 
351: explored. \\
352: \\
353: 
354: \acknowledgments{BWO would like to thank Chris Fryer, Alex Heger and Falk
355: Herwig for useful discussion. This work supported in part by NASA
356: grant NAG5-12140 and NSF grant AST-0307690 for BWO, DW and MLN. 
357: BWO has been funded in part
358: under the auspices of the U.S.\ Dept.\ of Energy, and supported by its
359: contract W-7405-ENG-36 to Los Alamos National Laboratory.  TA was supported
360: by NSF CAREER award AST-0239709 from the National Science Foundation.  The
361: simulations were performed at SDSC and NCSA with computing time provided by 
362: NRAC allocation MCA98N020.  TA and MLN gratefully acknowledge the Aspen 
363: Center for Physics for its hospitality during the final phases of this project.
364: We would like to thank the referee, Nick Gnedin, for suggestions which have
365: significantly improved the quality of this paper.
366: }
367: 
368: \bibliographystyle{plainnat}
369: \begin{thebibliography}{wd}
370: \bibitem[Abel(2000)]{2000RMxAC...9..300A} Abel, T.\ 2000, Revista Mexicana
371: de Astronomia y Astrofisica Conference Series, 9, 300
372: \bibitem[Abel, Bryan, \& Norman(2002)]{2002Sci...295...93A} Abel, T., Bryan,
373: G.~L., \& Norman, M.~L.\ 2002, Science, 295, 93
374: \bibitem[Abel \& Wandelt(2002)]{2002MNRAS.330L..53A} Abel, T.~\& Wandelt,
375: B.~D.\ 2002, \mnras, 330, L53
376: \bibitem[Barnes \& Efstathiou (1987)]{} Barnes \& Efstathiou (1987), ApJ, 319, 575 
377: \bibitem[Bondi \& Hoyle (1944)]{1944MNRAS.104..273B} Bondi, H. \& Hoyle, F.
378: \ 1944, MNRAS, 104, 273
379: \bibitem[Bryan \& Norman (1997)]{313} Bryan, G. \& Norman, M.L. 1997 \ In
380: ``Workshop on Structured Adaptive Mesh  Refinement Grid Methods'', Ed. N.
381: Chrisochoides
382: \bibitem[Gardner (2001)]{2001ApJ...557..616G} Gardner, J.P. \ 2001, \apj, 557, 616
383: \bibitem[Haiman, Abel \& Rees (2002)]{2000ApJ...534...11H} Haiman, Z., Abel, T. \& Rees,
384: M. \ 2000, \apj, 534, 11
385: \bibitem[Heger \& Woosley (2002)]{2002ApJ...567..532H} Heger, A. \& Woosley, S.E. \ 2002, \apj, 567, 232
386: \bibitem[Heger et al. (2003)]{2003ApJ...591..288H} Heger, A., Fryer, C.L.,
387: Woosley, S.E., Langer, N. \& Hartmann, D.H. \ 2003, \apj, 591, 288
388: \bibitem[Kitayama, Yoshida, Susa, \& Umemura(2004)]{2004ApJ...613..631K}
389: Kitayama, T., Yoshida, N., Susa, H., \& Umemura, M.\ 2004, \apj, 613, 631
390: \bibitem[Machacek, Abel \& Bryan (2001)]{2001ApJ...548..509M} Machacek, M.E., Bryan, G.L. 
391: \& Abel, T. \ 2001, \apj, 548, 509
392: \bibitem[Oh \& Haiman (2003)]{2003MNRAS.346..456O} Oh, S.P. \& Haiman, Z. \ 2003 MNRAS, 346, 456
393: \bibitem[O'Shea et al. (2004)]{326} O'Shea, B.W., Bryan, G., Bordner, J.,
394: Norman, Michael L., Abel, T., Harknes, R.
395: and Kritsuk, A. \ 2004.  In ``Adaptive Mesh Refinement - Theory  and
396: Applications'', Eds. T. Plewa, T. Linde \& V. G. Weirs,
397: Springer Lecture  Notes in Computational Science and Engineering
398: \bibitem[Padmanabhan (1993)]{Pad1993} Padmanabhan, T., 1993, \emph{Structure formation in the universe},
399: Cambridge: Cambridge Univ. Press
400: \bibitem[Ricotti et al. (2002b)]{2002ApJ...575...49R} Ricotti, M., Gnedin, N.Y. \& Shull, J.M. \ 2002b \apj, 575, 49
401: \bibitem[Ripamonti \& Abel(2004)]{2004MNRAS.348.1019R} Ripamonti, E.~\&
402: Abel, T.\ 2004, \mnras, 348, 1019
403: \bibitem[Saigo et al.(2004)]{2004ApJ...615L..65S} Saigo, K., Matsumoto, T., 
404: \& Umemura, M.\ 2004, \apjl, 615, L65 
405: \bibitem[Schaerer(2002)]{2002A&A...382...28S} Schaerer, D.\ 2002, \aap, 382, 28
406: \bibitem[Shapiro et al.(2004)]{2004MNRAS.348..753S} Shapiro, P.~R., Iliev, 
407: I.~T., \& Raga, A.~C.\ 2004, \mnras, 348, 753 
408: \bibitem[Truelove et al. (1998)]{1998ApJ...495..821T} Truelove, J.K. et al.
409: \ 1998, \apj, 495, 821
410: \bibitem[Umeda \& Nomoto (2005)]{2005ApJ...619..427U} Umeda, H. \& Nomoto,
411: K. \ 2005, \apj, 619, 427
412: \bibitem[Whalen, Abel, \& Norman(2004)]{2004ApJ...610...14W} Whalen, D.,
413: Abel, T., \& Norman, M.~L.\ 2004, \apj, 610, 14
414: \end{thebibliography}
415: 
416: \end{document}  
417: 
418: 
419: